xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 17dd36da8292193180754d5047c0926dbb56818c)
1 /*	$NetBSD: tcp_usrreq.c,v 1.61 2001/03/20 20:07:52 thorpej 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, 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 	/*
595 	 * TCP slow timer went off; going through this
596 	 * routine for tracing's sake.
597 	 */
598 	case PRU_SLOWTIMO:
599 		tp = tcp_timers(tp, (long)nam);
600 		req |= (long)nam << 8;		/* for debug's sake */
601 		break;
602 
603 	default:
604 		panic("tcp_usrreq");
605 	}
606 	if (tp && (so->so_options & SO_DEBUG))
607 		tcp_trace(TA_USER, ostate, tp, NULL, req);
608 
609 release:
610 	splx(s);
611 	return (error);
612 }
613 
614 int
615 tcp_ctloutput(op, so, level, optname, mp)
616 	int op;
617 	struct socket *so;
618 	int level, optname;
619 	struct mbuf **mp;
620 {
621 	int error = 0, s;
622 	struct inpcb *inp;
623 #ifdef INET6
624 	struct in6pcb *in6p;
625 #endif
626 	struct tcpcb *tp;
627 	struct mbuf *m;
628 	int i;
629 	int family;	/* family of the socket */
630 
631 	family = so->so_proto->pr_domain->dom_family;
632 
633 	s = splsoftnet();
634 	switch (family) {
635 #ifdef INET
636 	case PF_INET:
637 		inp = sotoinpcb(so);
638 #ifdef INET6
639 		in6p = NULL;
640 #endif
641 		break;
642 #endif
643 #ifdef INET6
644 	case PF_INET6:
645 		inp = NULL;
646 		in6p = sotoin6pcb(so);
647 		break;
648 #endif
649 	default:
650 		splx(s);
651 		return EAFNOSUPPORT;
652 	}
653 #ifndef INET6
654 	if (inp == NULL)
655 #else
656 	if (inp == NULL && in6p == NULL)
657 #endif
658 	{
659 		splx(s);
660 		if (op == PRCO_SETOPT && *mp)
661 			(void) m_free(*mp);
662 		return (ECONNRESET);
663 	}
664 	if (level != IPPROTO_TCP) {
665 		switch (family) {
666 #ifdef INET
667 		case PF_INET:
668 			error = ip_ctloutput(op, so, level, optname, mp);
669 			break;
670 #endif
671 #ifdef INET6
672 		case PF_INET6:
673 			error = ip6_ctloutput(op, so, level, optname, mp);
674 			break;
675 #endif
676 		}
677 		splx(s);
678 		return (error);
679 	}
680 	if (inp)
681 		tp = intotcpcb(inp);
682 #ifdef INET6
683 	else if (in6p)
684 		tp = in6totcpcb(in6p);
685 #endif
686 	else
687 		tp = NULL;
688 
689 	switch (op) {
690 
691 	case PRCO_SETOPT:
692 		m = *mp;
693 		switch (optname) {
694 
695 		case TCP_NODELAY:
696 			if (m == NULL || m->m_len < sizeof (int))
697 				error = EINVAL;
698 			else if (*mtod(m, int *))
699 				tp->t_flags |= TF_NODELAY;
700 			else
701 				tp->t_flags &= ~TF_NODELAY;
702 			break;
703 
704 		case TCP_MAXSEG:
705 			if (m && (i = *mtod(m, int *)) > 0 &&
706 			    i <= tp->t_peermss)
707 				tp->t_peermss = i;  /* limit on send size */
708 			else
709 				error = EINVAL;
710 			break;
711 
712 		default:
713 			error = ENOPROTOOPT;
714 			break;
715 		}
716 		if (m)
717 			(void) m_free(m);
718 		break;
719 
720 	case PRCO_GETOPT:
721 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
722 		m->m_len = sizeof(int);
723 
724 		switch (optname) {
725 		case TCP_NODELAY:
726 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
727 			break;
728 		case TCP_MAXSEG:
729 			*mtod(m, int *) = tp->t_peermss;
730 			break;
731 		default:
732 			error = ENOPROTOOPT;
733 			break;
734 		}
735 		break;
736 	}
737 	splx(s);
738 	return (error);
739 }
740 
741 #ifndef TCP_SENDSPACE
742 #define	TCP_SENDSPACE	1024*16;
743 #endif
744 int	tcp_sendspace = TCP_SENDSPACE;
745 #ifndef TCP_RECVSPACE
746 #define	TCP_RECVSPACE	1024*16;
747 #endif
748 int	tcp_recvspace = TCP_RECVSPACE;
749 
750 /*
751  * Attach TCP protocol to socket, allocating
752  * internet protocol control block, tcp control block,
753  * bufer space, and entering LISTEN state if to accept connections.
754  */
755 int
756 tcp_attach(so)
757 	struct socket *so;
758 {
759 	struct tcpcb *tp;
760 	struct inpcb *inp;
761 #ifdef INET6
762 	struct in6pcb *in6p;
763 #endif
764 	int error;
765 	int family;	/* family of the socket */
766 
767 	family = so->so_proto->pr_domain->dom_family;
768 
769 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
770 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
771 		if (error)
772 			return (error);
773 	}
774 	switch (family) {
775 #ifdef INET
776 	case PF_INET:
777 		error = in_pcballoc(so, &tcbtable);
778 		if (error)
779 			return (error);
780 		inp = sotoinpcb(so);
781 #ifdef INET6
782 		in6p = NULL;
783 #endif
784 		break;
785 #endif
786 #ifdef INET6
787 	case PF_INET6:
788 		error = in6_pcballoc(so, &tcb6);
789 		if (error)
790 			return (error);
791 		inp = NULL;
792 		in6p = sotoin6pcb(so);
793 		break;
794 #endif
795 	default:
796 		return EAFNOSUPPORT;
797 	}
798 #ifdef IPSEC
799 #ifdef INET
800 	if (inp) {
801 		error = ipsec_init_policy(so, &inp->inp_sp);
802 		if (error != 0) {
803 			in_pcbdetach(inp);
804 			return (error);
805 		}
806 	}
807 #endif
808 #ifdef INET6
809 	if (in6p) {
810 		error = ipsec_init_policy(so, &in6p->in6p_sp);
811 		if (error != 0) {
812 			in6_pcbdetach(in6p);
813 			return (error);
814 		}
815 	}
816 #endif
817 #endif /*IPSEC*/
818 	if (inp)
819 		tp = tcp_newtcpcb(family, (void *)inp);
820 #ifdef INET6
821 	else if (in6p)
822 		tp = tcp_newtcpcb(family, (void *)in6p);
823 #endif
824 	else
825 		tp = NULL;
826 
827 	if (tp == 0) {
828 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
829 
830 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
831 #ifdef INET
832 		if (inp)
833 			in_pcbdetach(inp);
834 #endif
835 #ifdef INET6
836 		if (in6p)
837 			in6_pcbdetach(in6p);
838 #endif
839 		so->so_state |= nofd;
840 		return (ENOBUFS);
841 	}
842 	tp->t_state = TCPS_CLOSED;
843 	return (0);
844 }
845 
846 /*
847  * Initiate (or continue) disconnect.
848  * If embryonic state, just send reset (once).
849  * If in ``let data drain'' option and linger null, just drop.
850  * Otherwise (hard), mark socket disconnecting and drop
851  * current input data; switch states based on user close, and
852  * send segment to peer (with FIN).
853  */
854 struct tcpcb *
855 tcp_disconnect(tp)
856 	struct tcpcb *tp;
857 {
858 	struct socket *so;
859 
860 	if (tp->t_inpcb)
861 		so = tp->t_inpcb->inp_socket;
862 #ifdef INET6
863 	else if (tp->t_in6pcb)
864 		so = tp->t_in6pcb->in6p_socket;
865 #endif
866 	else
867 		so = NULL;
868 
869 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
870 		tp = tcp_close(tp);
871 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
872 		tp = tcp_drop(tp, 0);
873 	else {
874 		soisdisconnecting(so);
875 		sbflush(&so->so_rcv);
876 		tp = tcp_usrclosed(tp);
877 		if (tp)
878 			(void) tcp_output(tp);
879 	}
880 	return (tp);
881 }
882 
883 /*
884  * User issued close, and wish to trail through shutdown states:
885  * if never received SYN, just forget it.  If got a SYN from peer,
886  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
887  * If already got a FIN from peer, then almost done; go to LAST_ACK
888  * state.  In all other cases, have already sent FIN to peer (e.g.
889  * after PRU_SHUTDOWN), and just have to play tedious game waiting
890  * for peer to send FIN or not respond to keep-alives, etc.
891  * We can let the user exit from the close as soon as the FIN is acked.
892  */
893 struct tcpcb *
894 tcp_usrclosed(tp)
895 	struct tcpcb *tp;
896 {
897 
898 	switch (tp->t_state) {
899 
900 	case TCPS_CLOSED:
901 	case TCPS_LISTEN:
902 	case TCPS_SYN_SENT:
903 		tp->t_state = TCPS_CLOSED;
904 		tp = tcp_close(tp);
905 		break;
906 
907 	case TCPS_SYN_RECEIVED:
908 	case TCPS_ESTABLISHED:
909 		tp->t_state = TCPS_FIN_WAIT_1;
910 		break;
911 
912 	case TCPS_CLOSE_WAIT:
913 		tp->t_state = TCPS_LAST_ACK;
914 		break;
915 	}
916 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
917 		struct socket *so;
918 		if (tp->t_inpcb)
919 			so = tp->t_inpcb->inp_socket;
920 #ifdef INET6
921 		else if (tp->t_in6pcb)
922 			so = tp->t_in6pcb->in6p_socket;
923 #endif
924 		else
925 			so = NULL;
926 		soisdisconnected(so);
927 		/*
928 		 * If we are in FIN_WAIT_2, we arrived here because the
929 		 * application did a shutdown of the send side.  Like the
930 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
931 		 * a full close, we start a timer to make sure sockets are
932 		 * not left in FIN_WAIT_2 forever.
933 		 */
934 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tcp_maxidle > 0))
935 			TCP_TIMER_ARM(tp, TCPT_2MSL, tcp_maxidle);
936 	}
937 	return (tp);
938 }
939 
940 static const struct {
941 	 unsigned int valid : 1;
942 	 unsigned int rdonly : 1;
943 	 int *var;
944 	 int val;
945 	 } tcp_ctlvars[] = TCPCTL_VARIABLES;
946 
947 /*
948  * Sysctl for tcp variables.
949  */
950 int
951 tcp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
952 	int *name;
953 	u_int namelen;
954 	void *oldp;
955 	size_t *oldlenp;
956 	void *newp;
957 	size_t newlen;
958 {
959 
960 	/* All sysctl names at this level are terminal. */
961 	if (namelen != 1)
962 		return (ENOTDIR);
963 
964 	if (name[0] < sizeof(tcp_ctlvars)/sizeof(tcp_ctlvars[0])
965 	    && tcp_ctlvars[name[0]].valid) {
966 		if (tcp_ctlvars[name[0]].rdonly)
967 			return (sysctl_rdint(oldp, oldlenp, newp,
968 			    tcp_ctlvars[name[0]].val));
969 		else
970 			return (sysctl_int(oldp, oldlenp, newp, newlen,
971 			    tcp_ctlvars[name[0]].var));
972 	}
973 
974 	return (ENOPROTOOPT);
975 }
976