xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 1ffa7b76c40339c17a0fb2a09fac93f287cfc046)
1 /*	$NetBSD: tcp_usrreq.c,v 1.76 2003/04/19 20:58:36 christos 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 <sys/cdefs.h>
105 __KERNEL_RCSID(0, "$NetBSD: tcp_usrreq.c,v 1.76 2003/04/19 20:58:36 christos Exp $");
106 
107 #include "opt_inet.h"
108 #include "opt_ipsec.h"
109 #include "opt_tcp_debug.h"
110 
111 #include <sys/param.h>
112 #include <sys/systm.h>
113 #include <sys/kernel.h>
114 #include <sys/malloc.h>
115 #include <sys/mbuf.h>
116 #include <sys/socket.h>
117 #include <sys/socketvar.h>
118 #include <sys/protosw.h>
119 #include <sys/errno.h>
120 #include <sys/stat.h>
121 #include <sys/proc.h>
122 #include <sys/domain.h>
123 #include <sys/sysctl.h>
124 
125 #include <net/if.h>
126 #include <net/route.h>
127 
128 #include <netinet/in.h>
129 #include <netinet/in_systm.h>
130 #include <netinet/in_var.h>
131 #include <netinet/ip.h>
132 #include <netinet/in_pcb.h>
133 #include <netinet/ip_var.h>
134 
135 #ifdef INET6
136 #ifndef INET
137 #include <netinet/in.h>
138 #endif
139 #include <netinet/ip6.h>
140 #include <netinet6/in6_pcb.h>
141 #include <netinet6/ip6_var.h>
142 #endif
143 
144 #include <netinet/tcp.h>
145 #include <netinet/tcp_fsm.h>
146 #include <netinet/tcp_seq.h>
147 #include <netinet/tcp_timer.h>
148 #include <netinet/tcp_var.h>
149 #include <netinet/tcpip.h>
150 #include <netinet/tcp_debug.h>
151 
152 #include "opt_tcp_space.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 static int tcp_sysctl_ident(void *, size_t *, void *, size_t);
164 
165 /*
166  * Process a TCP user request for TCP tb.  If this is a send request
167  * then m is the mbuf chain of send data.  If this is a timer expiration
168  * (called from the software clock routine), then timertype tells which timer.
169  */
170 /*ARGSUSED*/
171 int
172 tcp_usrreq(so, req, m, nam, control, p)
173 	struct socket *so;
174 	int req;
175 	struct mbuf *m, *nam, *control;
176 	struct proc *p;
177 {
178 	struct inpcb *inp;
179 #ifdef INET6
180 	struct in6pcb *in6p;
181 #endif
182 	struct tcpcb *tp = NULL;
183 	int s;
184 	int error = 0;
185 #ifdef TCP_DEBUG
186 	int ostate = 0;
187 #endif
188 	int family;	/* family of the socket */
189 
190 	family = so->so_proto->pr_domain->dom_family;
191 
192 	if (req == PRU_CONTROL) {
193 		switch (family) {
194 #ifdef INET
195 		case PF_INET:
196 			return (in_control(so, (long)m, (caddr_t)nam,
197 			    (struct ifnet *)control, p));
198 #endif
199 #ifdef INET6
200 		case PF_INET6:
201 			return (in6_control(so, (long)m, (caddr_t)nam,
202 			    (struct ifnet *)control, p));
203 #endif
204 		default:
205 			return EAFNOSUPPORT;
206 		}
207 	}
208 
209 	if (req == PRU_PURGEIF) {
210 		switch (family) {
211 #ifdef INET
212 		case PF_INET:
213 			in_pcbpurgeif0(&tcbtable, (struct ifnet *)control);
214 			in_purgeif((struct ifnet *)control);
215 			in_pcbpurgeif(&tcbtable, (struct ifnet *)control);
216 			break;
217 #endif
218 #ifdef INET6
219 		case PF_INET6:
220 			in6_pcbpurgeif0(&tcb6, (struct ifnet *)control);
221 			in6_purgeif((struct ifnet *)control);
222 			in6_pcbpurgeif(&tcb6, (struct ifnet *)control);
223 			break;
224 #endif
225 		default:
226 			return (EAFNOSUPPORT);
227 		}
228 		return (0);
229 	}
230 
231 	s = splsoftnet();
232 	switch (family) {
233 #ifdef INET
234 	case PF_INET:
235 		inp = sotoinpcb(so);
236 #ifdef INET6
237 		in6p = NULL;
238 #endif
239 		break;
240 #endif
241 #ifdef INET6
242 	case PF_INET6:
243 		inp = NULL;
244 		in6p = sotoin6pcb(so);
245 		break;
246 #endif
247 	default:
248 		splx(s);
249 		return EAFNOSUPPORT;
250 	}
251 
252 #ifdef DIAGNOSTIC
253 #ifdef INET6
254 	if (inp && in6p)
255 		panic("tcp_usrreq: both inp and in6p set to non-NULL");
256 #endif
257 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
258 		panic("tcp_usrreq: unexpected control mbuf");
259 #endif
260 	/*
261 	 * When a TCP is attached to a socket, then there will be
262 	 * a (struct inpcb) pointed at by the socket, and this
263 	 * structure will point at a subsidary (struct tcpcb).
264 	 */
265 #ifndef INET6
266 	if (inp == 0 && req != PRU_ATTACH)
267 #else
268 	if ((inp == 0 && in6p == 0) && req != PRU_ATTACH)
269 #endif
270 	{
271 		error = EINVAL;
272 		goto release;
273 	}
274 #ifdef INET
275 	if (inp) {
276 		tp = intotcpcb(inp);
277 		/* WHAT IF TP IS 0? */
278 #ifdef KPROF
279 		tcp_acounts[tp->t_state][req]++;
280 #endif
281 #ifdef TCP_DEBUG
282 		ostate = tp->t_state;
283 #endif
284 	}
285 #endif
286 #ifdef INET6
287 	if (in6p) {
288 		tp = in6totcpcb(in6p);
289 		/* WHAT IF TP IS 0? */
290 #ifdef KPROF
291 		tcp_acounts[tp->t_state][req]++;
292 #endif
293 #ifdef TCP_DEBUG
294 		ostate = tp->t_state;
295 #endif
296 	}
297 #endif
298 
299 	switch (req) {
300 
301 	/*
302 	 * TCP attaches to socket via PRU_ATTACH, reserving space,
303 	 * and an internet control block.
304 	 */
305 	case PRU_ATTACH:
306 #ifndef INET6
307 		if (inp != 0)
308 #else
309 		if (inp != 0 || in6p != 0)
310 #endif
311 		{
312 			error = EISCONN;
313 			break;
314 		}
315 		error = tcp_attach(so);
316 		if (error)
317 			break;
318 		if ((so->so_options & SO_LINGER) && so->so_linger == 0)
319 			so->so_linger = TCP_LINGERTIME;
320 		tp = sototcpcb(so);
321 		break;
322 
323 	/*
324 	 * PRU_DETACH detaches the TCP protocol from the socket.
325 	 */
326 	case PRU_DETACH:
327 		tp = tcp_disconnect(tp);
328 		break;
329 
330 	/*
331 	 * Give the socket an address.
332 	 */
333 	case PRU_BIND:
334 		switch (family) {
335 #ifdef INET
336 		case PF_INET:
337 			error = in_pcbbind(inp, nam, p);
338 			break;
339 #endif
340 #ifdef INET6
341 		case PF_INET6:
342 			error = in6_pcbbind(in6p, nam, p);
343 			if (!error) {
344 				/* mapped addr case */
345 				if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
346 					tp->t_family = AF_INET;
347 				else
348 					tp->t_family = AF_INET6;
349 			}
350 			break;
351 #endif
352 		}
353 		break;
354 
355 	/*
356 	 * Prepare to accept connections.
357 	 */
358 	case PRU_LISTEN:
359 #ifdef INET
360 		if (inp && inp->inp_lport == 0) {
361 			error = in_pcbbind(inp, (struct mbuf *)0,
362 			    (struct proc *)0);
363 			if (error)
364 				break;
365 		}
366 #endif
367 #ifdef INET6
368 		if (in6p && in6p->in6p_lport == 0) {
369 			error = in6_pcbbind(in6p, (struct mbuf *)0,
370 			    (struct proc *)0);
371 			if (error)
372 				break;
373 		}
374 #endif
375 		tp->t_state = TCPS_LISTEN;
376 		break;
377 
378 	/*
379 	 * Initiate connection to peer.
380 	 * Create a template for use in transmissions on this connection.
381 	 * Enter SYN_SENT state, and mark socket as connecting.
382 	 * Start keep-alive timer, and seed output sequence space.
383 	 * Send initial segment on connection.
384 	 */
385 	case PRU_CONNECT:
386 #ifdef INET
387 		if (inp) {
388 			if (inp->inp_lport == 0) {
389 				error = in_pcbbind(inp, (struct mbuf *)0,
390 				    (struct proc *)0);
391 				if (error)
392 					break;
393 			}
394 			error = in_pcbconnect(inp, nam);
395 		}
396 #endif
397 #ifdef INET6
398 		if (in6p) {
399 			if (in6p->in6p_lport == 0) {
400 				error = in6_pcbbind(in6p, (struct mbuf *)0,
401 				    (struct proc *)0);
402 				if (error)
403 					break;
404 			}
405 			error = in6_pcbconnect(in6p, nam);
406 			if (!error) {
407 				/* mapped addr case */
408 				if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr))
409 					tp->t_family = AF_INET;
410 				else
411 					tp->t_family = AF_INET6;
412 			}
413 		}
414 #endif
415 		if (error)
416 			break;
417 		tp->t_template = tcp_template(tp);
418 		if (tp->t_template == 0) {
419 #ifdef INET
420 			if (inp)
421 				in_pcbdisconnect(inp);
422 #endif
423 #ifdef INET6
424 			if (in6p)
425 				in6_pcbdisconnect(in6p);
426 #endif
427 			error = ENOBUFS;
428 			break;
429 		}
430 		/* Compute window scaling to request.  */
431 		while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
432 		    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
433 			tp->request_r_scale++;
434 		soisconnecting(so);
435 		tcpstat.tcps_connattempt++;
436 		tp->t_state = TCPS_SYN_SENT;
437 		TCP_TIMER_ARM(tp, TCPT_KEEP, TCPTV_KEEP_INIT);
438 		tp->iss = tcp_new_iss(tp, 0);
439 		tcp_sendseqinit(tp);
440 		error = tcp_output(tp);
441 		break;
442 
443 	/*
444 	 * Create a TCP connection between two sockets.
445 	 */
446 	case PRU_CONNECT2:
447 		error = EOPNOTSUPP;
448 		break;
449 
450 	/*
451 	 * Initiate disconnect from peer.
452 	 * If connection never passed embryonic stage, just drop;
453 	 * else if don't need to let data drain, then can just drop anyways,
454 	 * else have to begin TCP shutdown process: mark socket disconnecting,
455 	 * drain unread data, state switch to reflect user close, and
456 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
457 	 * when peer sends FIN and acks ours.
458 	 *
459 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
460 	 */
461 	case PRU_DISCONNECT:
462 		tp = tcp_disconnect(tp);
463 		break;
464 
465 	/*
466 	 * Accept a connection.  Essentially all the work is
467 	 * done at higher levels; just return the address
468 	 * of the peer, storing through addr.
469 	 */
470 	case PRU_ACCEPT:
471 #ifdef INET
472 		if (inp)
473 			in_setpeeraddr(inp, nam);
474 #endif
475 #ifdef INET6
476 		if (in6p)
477 			in6_setpeeraddr(in6p, nam);
478 #endif
479 		break;
480 
481 	/*
482 	 * Mark the connection as being incapable of further output.
483 	 */
484 	case PRU_SHUTDOWN:
485 		socantsendmore(so);
486 		tp = tcp_usrclosed(tp);
487 		if (tp)
488 			error = tcp_output(tp);
489 		break;
490 
491 	/*
492 	 * After a receive, possibly send window update to peer.
493 	 */
494 	case PRU_RCVD:
495 		/*
496 		 * soreceive() calls this function when a user receives
497 		 * ancillary data on a listening socket. We don't call
498 		 * tcp_output in such a case, since there is no header
499 		 * template for a listening socket and hence the kernel
500 		 * will panic.
501 		 */
502 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0)
503 			(void) tcp_output(tp);
504 		break;
505 
506 	/*
507 	 * Do a send by putting data in output queue and updating urgent
508 	 * marker if URG set.  Possibly send more data.
509 	 */
510 	case PRU_SEND:
511 		if (control && control->m_len) {
512 			m_freem(control);
513 			m_freem(m);
514 			error = EINVAL;
515 			break;
516 		}
517 		sbappendstream(&so->so_snd, m);
518 		error = tcp_output(tp);
519 		break;
520 
521 	/*
522 	 * Abort the TCP.
523 	 */
524 	case PRU_ABORT:
525 		tp = tcp_drop(tp, ECONNABORTED);
526 		break;
527 
528 	case PRU_SENSE:
529 		/*
530 		 * stat: don't bother with a blocksize.
531 		 */
532 		splx(s);
533 		return (0);
534 
535 	case PRU_RCVOOB:
536 		if (control && control->m_len) {
537 			m_freem(control);
538 			m_freem(m);
539 			error = EINVAL;
540 			break;
541 		}
542 		if ((so->so_oobmark == 0 &&
543 		    (so->so_state & SS_RCVATMARK) == 0) ||
544 		    so->so_options & SO_OOBINLINE ||
545 		    tp->t_oobflags & TCPOOB_HADDATA) {
546 			error = EINVAL;
547 			break;
548 		}
549 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
550 			error = EWOULDBLOCK;
551 			break;
552 		}
553 		m->m_len = 1;
554 		*mtod(m, caddr_t) = tp->t_iobc;
555 		if (((long)nam & MSG_PEEK) == 0)
556 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
557 		break;
558 
559 	case PRU_SENDOOB:
560 		if (sbspace(&so->so_snd) < -512) {
561 			m_freem(m);
562 			error = ENOBUFS;
563 			break;
564 		}
565 		/*
566 		 * According to RFC961 (Assigned Protocols),
567 		 * the urgent pointer points to the last octet
568 		 * of urgent data.  We continue, however,
569 		 * to consider it to indicate the first octet
570 		 * of data past the urgent section.
571 		 * Otherwise, snd_up should be one lower.
572 		 */
573 		sbappendstream(&so->so_snd, m);
574 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
575 		tp->t_force = 1;
576 		error = tcp_output(tp);
577 		tp->t_force = 0;
578 		break;
579 
580 	case PRU_SOCKADDR:
581 #ifdef INET
582 		if (inp)
583 			in_setsockaddr(inp, nam);
584 #endif
585 #ifdef INET6
586 		if (in6p)
587 			in6_setsockaddr(in6p, nam);
588 #endif
589 		break;
590 
591 	case PRU_PEERADDR:
592 #ifdef INET
593 		if (inp)
594 			in_setpeeraddr(inp, nam);
595 #endif
596 #ifdef INET6
597 		if (in6p)
598 			in6_setpeeraddr(in6p, nam);
599 #endif
600 		break;
601 
602 	default:
603 		panic("tcp_usrreq");
604 	}
605 #ifdef TCP_DEBUG
606 	if (tp && (so->so_options & SO_DEBUG))
607 		tcp_trace(TA_USER, ostate, tp, NULL, req);
608 #endif
609 
610 release:
611 	splx(s);
612 	return (error);
613 }
614 
615 int
616 tcp_ctloutput(op, so, level, optname, mp)
617 	int op;
618 	struct socket *so;
619 	int level, optname;
620 	struct mbuf **mp;
621 {
622 	int error = 0, s;
623 	struct inpcb *inp;
624 #ifdef INET6
625 	struct in6pcb *in6p;
626 #endif
627 	struct tcpcb *tp;
628 	struct mbuf *m;
629 	int i;
630 	int family;	/* family of the socket */
631 
632 	family = so->so_proto->pr_domain->dom_family;
633 
634 	s = splsoftnet();
635 	switch (family) {
636 #ifdef INET
637 	case PF_INET:
638 		inp = sotoinpcb(so);
639 #ifdef INET6
640 		in6p = NULL;
641 #endif
642 		break;
643 #endif
644 #ifdef INET6
645 	case PF_INET6:
646 		inp = NULL;
647 		in6p = sotoin6pcb(so);
648 		break;
649 #endif
650 	default:
651 		splx(s);
652 		return EAFNOSUPPORT;
653 	}
654 #ifndef INET6
655 	if (inp == NULL)
656 #else
657 	if (inp == NULL && in6p == NULL)
658 #endif
659 	{
660 		splx(s);
661 		if (op == PRCO_SETOPT && *mp)
662 			(void) m_free(*mp);
663 		return (ECONNRESET);
664 	}
665 	if (level != IPPROTO_TCP) {
666 		switch (family) {
667 #ifdef INET
668 		case PF_INET:
669 			error = ip_ctloutput(op, so, level, optname, mp);
670 			break;
671 #endif
672 #ifdef INET6
673 		case PF_INET6:
674 			error = ip6_ctloutput(op, so, level, optname, mp);
675 			break;
676 #endif
677 		}
678 		splx(s);
679 		return (error);
680 	}
681 	if (inp)
682 		tp = intotcpcb(inp);
683 #ifdef INET6
684 	else if (in6p)
685 		tp = in6totcpcb(in6p);
686 #endif
687 	else
688 		tp = NULL;
689 
690 	switch (op) {
691 
692 	case PRCO_SETOPT:
693 		m = *mp;
694 		switch (optname) {
695 
696 		case TCP_NODELAY:
697 			if (m == NULL || m->m_len < sizeof (int))
698 				error = EINVAL;
699 			else if (*mtod(m, int *))
700 				tp->t_flags |= TF_NODELAY;
701 			else
702 				tp->t_flags &= ~TF_NODELAY;
703 			break;
704 
705 		case TCP_MAXSEG:
706 			if (m && (i = *mtod(m, int *)) > 0 &&
707 			    i <= tp->t_peermss)
708 				tp->t_peermss = i;  /* limit on send size */
709 			else
710 				error = EINVAL;
711 			break;
712 
713 		default:
714 			error = ENOPROTOOPT;
715 			break;
716 		}
717 		if (m)
718 			(void) m_free(m);
719 		break;
720 
721 	case PRCO_GETOPT:
722 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
723 		m->m_len = sizeof(int);
724 		MCLAIM(m, so->so_mowner);
725 
726 		switch (optname) {
727 		case TCP_NODELAY:
728 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
729 			break;
730 		case TCP_MAXSEG:
731 			*mtod(m, int *) = tp->t_peermss;
732 			break;
733 		default:
734 			error = ENOPROTOOPT;
735 			break;
736 		}
737 		break;
738 	}
739 	splx(s);
740 	return (error);
741 }
742 
743 #ifndef TCP_SENDSPACE
744 #define	TCP_SENDSPACE	1024*16
745 #endif
746 int	tcp_sendspace = TCP_SENDSPACE;
747 #ifndef TCP_RECVSPACE
748 #define	TCP_RECVSPACE	1024*16
749 #endif
750 int	tcp_recvspace = TCP_RECVSPACE;
751 
752 /*
753  * Attach TCP protocol to socket, allocating
754  * internet protocol control block, tcp control block,
755  * bufer space, and entering LISTEN state if to accept connections.
756  */
757 int
758 tcp_attach(so)
759 	struct socket *so;
760 {
761 	struct tcpcb *tp;
762 	struct inpcb *inp;
763 #ifdef INET6
764 	struct in6pcb *in6p;
765 #endif
766 	int error;
767 	int family;	/* family of the socket */
768 
769 	family = so->so_proto->pr_domain->dom_family;
770 
771 #ifdef MBUFTRACE
772 	so->so_mowner = &tcp_mowner;
773 	so->so_rcv.sb_mowner = &tcp_rx_mowner;
774 	so->so_snd.sb_mowner = &tcp_tx_mowner;
775 #endif
776 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
777 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
778 		if (error)
779 			return (error);
780 	}
781 	switch (family) {
782 #ifdef INET
783 	case PF_INET:
784 		error = in_pcballoc(so, &tcbtable);
785 		if (error)
786 			return (error);
787 		inp = sotoinpcb(so);
788 #ifdef INET6
789 		in6p = NULL;
790 #endif
791 		break;
792 #endif
793 #ifdef INET6
794 	case PF_INET6:
795 		error = in6_pcballoc(so, &tcb6);
796 		if (error)
797 			return (error);
798 		inp = NULL;
799 		in6p = sotoin6pcb(so);
800 		break;
801 #endif
802 	default:
803 		return EAFNOSUPPORT;
804 	}
805 	if (inp)
806 		tp = tcp_newtcpcb(family, (void *)inp);
807 #ifdef INET6
808 	else if (in6p)
809 		tp = tcp_newtcpcb(family, (void *)in6p);
810 #endif
811 	else
812 		tp = NULL;
813 
814 	if (tp == 0) {
815 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
816 
817 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
818 #ifdef INET
819 		if (inp)
820 			in_pcbdetach(inp);
821 #endif
822 #ifdef INET6
823 		if (in6p)
824 			in6_pcbdetach(in6p);
825 #endif
826 		so->so_state |= nofd;
827 		return (ENOBUFS);
828 	}
829 	tp->t_state = TCPS_CLOSED;
830 	return (0);
831 }
832 
833 /*
834  * Initiate (or continue) disconnect.
835  * If embryonic state, just send reset (once).
836  * If in ``let data drain'' option and linger null, just drop.
837  * Otherwise (hard), mark socket disconnecting and drop
838  * current input data; switch states based on user close, and
839  * send segment to peer (with FIN).
840  */
841 struct tcpcb *
842 tcp_disconnect(tp)
843 	struct tcpcb *tp;
844 {
845 	struct socket *so;
846 
847 	if (tp->t_inpcb)
848 		so = tp->t_inpcb->inp_socket;
849 #ifdef INET6
850 	else if (tp->t_in6pcb)
851 		so = tp->t_in6pcb->in6p_socket;
852 #endif
853 	else
854 		so = NULL;
855 
856 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
857 		tp = tcp_close(tp);
858 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
859 		tp = tcp_drop(tp, 0);
860 	else {
861 		soisdisconnecting(so);
862 		sbflush(&so->so_rcv);
863 		tp = tcp_usrclosed(tp);
864 		if (tp)
865 			(void) tcp_output(tp);
866 	}
867 	return (tp);
868 }
869 
870 /*
871  * User issued close, and wish to trail through shutdown states:
872  * if never received SYN, just forget it.  If got a SYN from peer,
873  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
874  * If already got a FIN from peer, then almost done; go to LAST_ACK
875  * state.  In all other cases, have already sent FIN to peer (e.g.
876  * after PRU_SHUTDOWN), and just have to play tedious game waiting
877  * for peer to send FIN or not respond to keep-alives, etc.
878  * We can let the user exit from the close as soon as the FIN is acked.
879  */
880 struct tcpcb *
881 tcp_usrclosed(tp)
882 	struct tcpcb *tp;
883 {
884 
885 	switch (tp->t_state) {
886 
887 	case TCPS_CLOSED:
888 	case TCPS_LISTEN:
889 	case TCPS_SYN_SENT:
890 		tp->t_state = TCPS_CLOSED;
891 		tp = tcp_close(tp);
892 		break;
893 
894 	case TCPS_SYN_RECEIVED:
895 	case TCPS_ESTABLISHED:
896 		tp->t_state = TCPS_FIN_WAIT_1;
897 		break;
898 
899 	case TCPS_CLOSE_WAIT:
900 		tp->t_state = TCPS_LAST_ACK;
901 		break;
902 	}
903 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
904 		struct socket *so;
905 		if (tp->t_inpcb)
906 			so = tp->t_inpcb->inp_socket;
907 #ifdef INET6
908 		else if (tp->t_in6pcb)
909 			so = tp->t_in6pcb->in6p_socket;
910 #endif
911 		else
912 			so = NULL;
913 		soisdisconnected(so);
914 		/*
915 		 * If we are in FIN_WAIT_2, we arrived here because the
916 		 * application did a shutdown of the send side.  Like the
917 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
918 		 * a full close, we start a timer to make sure sockets are
919 		 * not left in FIN_WAIT_2 forever.
920 		 */
921 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tcp_maxidle > 0))
922 			TCP_TIMER_ARM(tp, TCPT_2MSL, tcp_maxidle);
923 	}
924 	return (tp);
925 }
926 
927 static const struct {
928 	 unsigned int valid : 1;
929 	 unsigned int rdonly : 1;
930 	 int *var;
931 	 int val;
932 	 } tcp_ctlvars[] = TCPCTL_VARIABLES;
933 
934 /*
935  * Sysctl for tcp variables.
936  */
937 int
938 tcp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
939 	int *name;
940 	u_int namelen;
941 	void *oldp;
942 	size_t *oldlenp;
943 	void *newp;
944 	size_t newlen;
945 {
946 	int error, saved_value = 0;
947 
948 	/* All sysctl names at this level are terminal. */
949 	if (namelen != 1)
950 		return (ENOTDIR);
951 
952 	if (name[0] == TCPCTL_IDENT)
953 		return tcp_sysctl_ident(oldp, oldlenp, newp, newlen);
954 
955 	if (name[0] < sizeof(tcp_ctlvars)/sizeof(tcp_ctlvars[0])
956 	    && tcp_ctlvars[name[0]].valid) {
957 		if (tcp_ctlvars[name[0]].rdonly) {
958 			return (sysctl_rdint(oldp, oldlenp, newp,
959 			    tcp_ctlvars[name[0]].val));
960 		} else {
961 			switch (name[0]) {
962 			case TCPCTL_MSSDFLT:
963 				saved_value = tcp_mssdflt;
964 				break;
965 			}
966 			error = sysctl_int(oldp, oldlenp, newp, newlen,
967 			    tcp_ctlvars[name[0]].var);
968 			if (error)
969 				return (error);
970 			switch (name[0]) {
971 			case TCPCTL_MSSDFLT:
972 				if (tcp_mssdflt < 32) {
973 					tcp_mssdflt = saved_value;
974 					return (EINVAL);
975 				}
976 				break;
977 			}
978 			return (0);
979 		}
980 	}
981 
982 	return (ENOPROTOOPT);
983 }
984 
985 
986 static int
987 tcp_sysctl_ident(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
988 {
989 	struct sysctl_tcp_ident_args args;
990 	struct socket *sockp;
991 	struct inpcb *inb;
992 	uid_t uid;
993 	int error;
994 
995 	if (newlen != sizeof(args))
996 		return EINVAL;
997 	if (!newp)
998 		return EFAULT;
999 	if (*oldlenp != sizeof(uid_t))
1000 		return ENOMEM;
1001 	if (!oldp || *oldlenp != sizeof(uid_t))
1002 		return ENOMEM;
1003 	if ((error = copyin(newp, &args, newlen)) != 0)
1004 		return error;
1005 
1006 	inb = in_pcblookup_connect(&tcbtable, args.raddr, args.rport,
1007 	    args.laddr, args.lport);
1008 	if (inb) {
1009 		sockp = inb->inp_socket;
1010 		if (sockp)
1011 			uid = sockp->so_uid;
1012 		else
1013 			return ESRCH;
1014 	} else
1015 		return ESRCH;
1016 
1017 	if ((error = copyout(&uid, oldp, sizeof(uid))) != 0)
1018 		return error;
1019 
1020 	return 0;
1021 }
1022