xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 73704c4ce4ee2a60eb617e693ce7e9f03902613e)
1 /*	$NetBSD: tcp_usrreq.c,v 1.84 2003/09/29 21:39:35 tls 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. Neither the name of the University nor the names of its contributors
82  *    may be used to endorse or promote products derived from this software
83  *    without specific prior written permission.
84  *
85  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
86  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
87  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
88  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
89  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
90  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
91  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
92  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
93  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
94  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
95  * SUCH DAMAGE.
96  *
97  *	@(#)tcp_usrreq.c	8.5 (Berkeley) 6/21/95
98  */
99 
100 #include <sys/cdefs.h>
101 __KERNEL_RCSID(0, "$NetBSD: tcp_usrreq.c,v 1.84 2003/09/29 21:39:35 tls Exp $");
102 
103 #include "opt_inet.h"
104 #include "opt_ipsec.h"
105 #include "opt_tcp_debug.h"
106 #include "opt_mbuftrace.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_space.h"
150 
151 #ifdef IPSEC
152 #include <netinet6/ipsec.h>
153 #endif /*IPSEC*/
154 
155 /*
156  * TCP protocol interface to socket abstraction.
157  */
158 extern	char *tcpstates[];
159 
160 static int tcp_sysctl_ident(int *, u_int, void *, size_t *, void *, size_t);
161 
162 /*
163  * Process a TCP user request for TCP tb.  If this is a send request
164  * then m is the mbuf chain of send data.  If this is a timer expiration
165  * (called from the software clock routine), then timertype tells which timer.
166  */
167 /*ARGSUSED*/
168 int
169 tcp_usrreq(so, req, m, nam, control, p)
170 	struct socket *so;
171 	int req;
172 	struct mbuf *m, *nam, *control;
173 	struct proc *p;
174 {
175 	struct inpcb *inp;
176 #ifdef INET6
177 	struct in6pcb *in6p;
178 #endif
179 	struct tcpcb *tp = NULL;
180 	int s;
181 	int error = 0;
182 #ifdef TCP_DEBUG
183 	int ostate = 0;
184 #endif
185 	int family;	/* family of the socket */
186 
187 	family = so->so_proto->pr_domain->dom_family;
188 
189 	if (req == PRU_CONTROL) {
190 		switch (family) {
191 #ifdef INET
192 		case PF_INET:
193 			return (in_control(so, (long)m, (caddr_t)nam,
194 			    (struct ifnet *)control, p));
195 #endif
196 #ifdef INET6
197 		case PF_INET6:
198 			return (in6_control(so, (long)m, (caddr_t)nam,
199 			    (struct ifnet *)control, p));
200 #endif
201 		default:
202 			return EAFNOSUPPORT;
203 		}
204 	}
205 
206 	if (req == PRU_PURGEIF) {
207 		switch (family) {
208 #ifdef INET
209 		case PF_INET:
210 			in_pcbpurgeif0(&tcbtable, (struct ifnet *)control);
211 			in_purgeif((struct ifnet *)control);
212 			in_pcbpurgeif(&tcbtable, (struct ifnet *)control);
213 			break;
214 #endif
215 #ifdef INET6
216 		case PF_INET6:
217 			in6_pcbpurgeif0(&tcbtable, (struct ifnet *)control);
218 			in6_purgeif((struct ifnet *)control);
219 			in6_pcbpurgeif(&tcbtable, (struct ifnet *)control);
220 			break;
221 #endif
222 		default:
223 			return (EAFNOSUPPORT);
224 		}
225 		return (0);
226 	}
227 
228 	s = splsoftnet();
229 	switch (family) {
230 #ifdef INET
231 	case PF_INET:
232 		inp = sotoinpcb(so);
233 #ifdef INET6
234 		in6p = NULL;
235 #endif
236 		break;
237 #endif
238 #ifdef INET6
239 	case PF_INET6:
240 		inp = NULL;
241 		in6p = sotoin6pcb(so);
242 		break;
243 #endif
244 	default:
245 		splx(s);
246 		return EAFNOSUPPORT;
247 	}
248 
249 #ifdef DIAGNOSTIC
250 #ifdef INET6
251 	if (inp && in6p)
252 		panic("tcp_usrreq: both inp and in6p set to non-NULL");
253 #endif
254 	if (req != PRU_SEND && req != PRU_SENDOOB && control)
255 		panic("tcp_usrreq: unexpected control mbuf");
256 #endif
257 	/*
258 	 * When a TCP is attached to a socket, then there will be
259 	 * a (struct inpcb) pointed at by the socket, and this
260 	 * structure will point at a subsidary (struct tcpcb).
261 	 */
262 #ifndef INET6
263 	if (inp == 0 && req != PRU_ATTACH)
264 #else
265 	if ((inp == 0 && in6p == 0) && req != PRU_ATTACH)
266 #endif
267 	{
268 		error = EINVAL;
269 		goto release;
270 	}
271 #ifdef INET
272 	if (inp) {
273 		tp = intotcpcb(inp);
274 		/* WHAT IF TP IS 0? */
275 #ifdef KPROF
276 		tcp_acounts[tp->t_state][req]++;
277 #endif
278 #ifdef TCP_DEBUG
279 		ostate = tp->t_state;
280 #endif
281 	}
282 #endif
283 #ifdef INET6
284 	if (in6p) {
285 		tp = in6totcpcb(in6p);
286 		/* WHAT IF TP IS 0? */
287 #ifdef KPROF
288 		tcp_acounts[tp->t_state][req]++;
289 #endif
290 #ifdef TCP_DEBUG
291 		ostate = tp->t_state;
292 #endif
293 	}
294 #endif
295 
296 	switch (req) {
297 
298 	/*
299 	 * TCP attaches to socket via PRU_ATTACH, reserving space,
300 	 * and an internet control block.
301 	 */
302 	case PRU_ATTACH:
303 #ifndef INET6
304 		if (inp != 0)
305 #else
306 		if (inp != 0 || in6p != 0)
307 #endif
308 		{
309 			error = EISCONN;
310 			break;
311 		}
312 		error = tcp_attach(so);
313 		if (error)
314 			break;
315 		if ((so->so_options & SO_LINGER) && so->so_linger == 0)
316 			so->so_linger = TCP_LINGERTIME;
317 		tp = sototcpcb(so);
318 		break;
319 
320 	/*
321 	 * PRU_DETACH detaches the TCP protocol from the socket.
322 	 */
323 	case PRU_DETACH:
324 		tp = tcp_disconnect(tp);
325 		break;
326 
327 	/*
328 	 * Give the socket an address.
329 	 */
330 	case PRU_BIND:
331 		switch (family) {
332 #ifdef INET
333 		case PF_INET:
334 			error = in_pcbbind(inp, nam, p);
335 			break;
336 #endif
337 #ifdef INET6
338 		case PF_INET6:
339 			error = in6_pcbbind(in6p, nam, p);
340 			if (!error) {
341 				/* mapped addr case */
342 				if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
343 					tp->t_family = AF_INET;
344 				else
345 					tp->t_family = AF_INET6;
346 			}
347 			break;
348 #endif
349 		}
350 		break;
351 
352 	/*
353 	 * Prepare to accept connections.
354 	 */
355 	case PRU_LISTEN:
356 #ifdef INET
357 		if (inp && inp->inp_lport == 0) {
358 			error = in_pcbbind(inp, (struct mbuf *)0,
359 			    (struct proc *)0);
360 			if (error)
361 				break;
362 		}
363 #endif
364 #ifdef INET6
365 		if (in6p && in6p->in6p_lport == 0) {
366 			error = in6_pcbbind(in6p, (struct mbuf *)0,
367 			    (struct proc *)0);
368 			if (error)
369 				break;
370 		}
371 #endif
372 		tp->t_state = TCPS_LISTEN;
373 		break;
374 
375 	/*
376 	 * Initiate connection to peer.
377 	 * Create a template for use in transmissions on this connection.
378 	 * Enter SYN_SENT state, and mark socket as connecting.
379 	 * Start keep-alive timer, and seed output sequence space.
380 	 * Send initial segment on connection.
381 	 */
382 	case PRU_CONNECT:
383 #ifdef INET
384 		if (inp) {
385 			if (inp->inp_lport == 0) {
386 				error = in_pcbbind(inp, (struct mbuf *)0,
387 				    (struct proc *)0);
388 				if (error)
389 					break;
390 			}
391 			error = in_pcbconnect(inp, nam);
392 		}
393 #endif
394 #ifdef INET6
395 		if (in6p) {
396 			if (in6p->in6p_lport == 0) {
397 				error = in6_pcbbind(in6p, (struct mbuf *)0,
398 				    (struct proc *)0);
399 				if (error)
400 					break;
401 			}
402 			error = in6_pcbconnect(in6p, nam);
403 			if (!error) {
404 				/* mapped addr case */
405 				if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr))
406 					tp->t_family = AF_INET;
407 				else
408 					tp->t_family = AF_INET6;
409 			}
410 		}
411 #endif
412 		if (error)
413 			break;
414 		tp->t_template = tcp_template(tp);
415 		if (tp->t_template == 0) {
416 #ifdef INET
417 			if (inp)
418 				in_pcbdisconnect(inp);
419 #endif
420 #ifdef INET6
421 			if (in6p)
422 				in6_pcbdisconnect(in6p);
423 #endif
424 			error = ENOBUFS;
425 			break;
426 		}
427 		/* Compute window scaling to request.  */
428 		while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
429 		    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
430 			tp->request_r_scale++;
431 		soisconnecting(so);
432 		tcpstat.tcps_connattempt++;
433 		tp->t_state = TCPS_SYN_SENT;
434 		TCP_TIMER_ARM(tp, TCPT_KEEP, TCPTV_KEEP_INIT);
435 		tp->iss = tcp_new_iss(tp, 0);
436 		tcp_sendseqinit(tp);
437 		error = tcp_output(tp);
438 		break;
439 
440 	/*
441 	 * Create a TCP connection between two sockets.
442 	 */
443 	case PRU_CONNECT2:
444 		error = EOPNOTSUPP;
445 		break;
446 
447 	/*
448 	 * Initiate disconnect from peer.
449 	 * If connection never passed embryonic stage, just drop;
450 	 * else if don't need to let data drain, then can just drop anyways,
451 	 * else have to begin TCP shutdown process: mark socket disconnecting,
452 	 * drain unread data, state switch to reflect user close, and
453 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
454 	 * when peer sends FIN and acks ours.
455 	 *
456 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
457 	 */
458 	case PRU_DISCONNECT:
459 		tp = tcp_disconnect(tp);
460 		break;
461 
462 	/*
463 	 * Accept a connection.  Essentially all the work is
464 	 * done at higher levels; just return the address
465 	 * of the peer, storing through addr.
466 	 */
467 	case PRU_ACCEPT:
468 #ifdef INET
469 		if (inp)
470 			in_setpeeraddr(inp, nam);
471 #endif
472 #ifdef INET6
473 		if (in6p)
474 			in6_setpeeraddr(in6p, nam);
475 #endif
476 		break;
477 
478 	/*
479 	 * Mark the connection as being incapable of further output.
480 	 */
481 	case PRU_SHUTDOWN:
482 		socantsendmore(so);
483 		tp = tcp_usrclosed(tp);
484 		if (tp)
485 			error = tcp_output(tp);
486 		break;
487 
488 	/*
489 	 * After a receive, possibly send window update to peer.
490 	 */
491 	case PRU_RCVD:
492 		/*
493 		 * soreceive() calls this function when a user receives
494 		 * ancillary data on a listening socket. We don't call
495 		 * tcp_output in such a case, since there is no header
496 		 * template for a listening socket and hence the kernel
497 		 * will panic.
498 		 */
499 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0)
500 			(void) tcp_output(tp);
501 		break;
502 
503 	/*
504 	 * Do a send by putting data in output queue and updating urgent
505 	 * marker if URG set.  Possibly send more data.
506 	 */
507 	case PRU_SEND:
508 		if (control && control->m_len) {
509 			m_freem(control);
510 			m_freem(m);
511 			error = EINVAL;
512 			break;
513 		}
514 		sbappendstream(&so->so_snd, m);
515 		error = tcp_output(tp);
516 		break;
517 
518 	/*
519 	 * Abort the TCP.
520 	 */
521 	case PRU_ABORT:
522 		tp = tcp_drop(tp, ECONNABORTED);
523 		break;
524 
525 	case PRU_SENSE:
526 		/*
527 		 * stat: don't bother with a blocksize.
528 		 */
529 		splx(s);
530 		return (0);
531 
532 	case PRU_RCVOOB:
533 		if (control && control->m_len) {
534 			m_freem(control);
535 			m_freem(m);
536 			error = EINVAL;
537 			break;
538 		}
539 		if ((so->so_oobmark == 0 &&
540 		    (so->so_state & SS_RCVATMARK) == 0) ||
541 		    so->so_options & SO_OOBINLINE ||
542 		    tp->t_oobflags & TCPOOB_HADDATA) {
543 			error = EINVAL;
544 			break;
545 		}
546 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
547 			error = EWOULDBLOCK;
548 			break;
549 		}
550 		m->m_len = 1;
551 		*mtod(m, caddr_t) = tp->t_iobc;
552 		if (((long)nam & MSG_PEEK) == 0)
553 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
554 		break;
555 
556 	case PRU_SENDOOB:
557 		if (sbspace(&so->so_snd) < -512) {
558 			m_freem(m);
559 			error = ENOBUFS;
560 			break;
561 		}
562 		/*
563 		 * According to RFC961 (Assigned Protocols),
564 		 * the urgent pointer points to the last octet
565 		 * of urgent data.  We continue, however,
566 		 * to consider it to indicate the first octet
567 		 * of data past the urgent section.
568 		 * Otherwise, snd_up should be one lower.
569 		 */
570 		sbappendstream(&so->so_snd, m);
571 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
572 		tp->t_force = 1;
573 		error = tcp_output(tp);
574 		tp->t_force = 0;
575 		break;
576 
577 	case PRU_SOCKADDR:
578 #ifdef INET
579 		if (inp)
580 			in_setsockaddr(inp, nam);
581 #endif
582 #ifdef INET6
583 		if (in6p)
584 			in6_setsockaddr(in6p, nam);
585 #endif
586 		break;
587 
588 	case PRU_PEERADDR:
589 #ifdef INET
590 		if (inp)
591 			in_setpeeraddr(inp, nam);
592 #endif
593 #ifdef INET6
594 		if (in6p)
595 			in6_setpeeraddr(in6p, nam);
596 #endif
597 		break;
598 
599 	default:
600 		panic("tcp_usrreq");
601 	}
602 #ifdef TCP_DEBUG
603 	if (tp && (so->so_options & SO_DEBUG))
604 		tcp_trace(TA_USER, ostate, tp, NULL, req);
605 #endif
606 
607 release:
608 	splx(s);
609 	return (error);
610 }
611 
612 int
613 tcp_ctloutput(op, so, level, optname, mp)
614 	int op;
615 	struct socket *so;
616 	int level, optname;
617 	struct mbuf **mp;
618 {
619 	int error = 0, s;
620 	struct inpcb *inp;
621 #ifdef INET6
622 	struct in6pcb *in6p;
623 #endif
624 	struct tcpcb *tp;
625 	struct mbuf *m;
626 	int i;
627 	int family;	/* family of the socket */
628 
629 	family = so->so_proto->pr_domain->dom_family;
630 
631 	s = splsoftnet();
632 	switch (family) {
633 #ifdef INET
634 	case PF_INET:
635 		inp = sotoinpcb(so);
636 #ifdef INET6
637 		in6p = NULL;
638 #endif
639 		break;
640 #endif
641 #ifdef INET6
642 	case PF_INET6:
643 		inp = NULL;
644 		in6p = sotoin6pcb(so);
645 		break;
646 #endif
647 	default:
648 		splx(s);
649 		return EAFNOSUPPORT;
650 	}
651 #ifndef INET6
652 	if (inp == NULL)
653 #else
654 	if (inp == NULL && in6p == NULL)
655 #endif
656 	{
657 		splx(s);
658 		if (op == PRCO_SETOPT && *mp)
659 			(void) m_free(*mp);
660 		return (ECONNRESET);
661 	}
662 	if (level != IPPROTO_TCP) {
663 		switch (family) {
664 #ifdef INET
665 		case PF_INET:
666 			error = ip_ctloutput(op, so, level, optname, mp);
667 			break;
668 #endif
669 #ifdef INET6
670 		case PF_INET6:
671 			error = ip6_ctloutput(op, so, level, optname, mp);
672 			break;
673 #endif
674 		}
675 		splx(s);
676 		return (error);
677 	}
678 	if (inp)
679 		tp = intotcpcb(inp);
680 #ifdef INET6
681 	else if (in6p)
682 		tp = in6totcpcb(in6p);
683 #endif
684 	else
685 		tp = NULL;
686 
687 	switch (op) {
688 
689 	case PRCO_SETOPT:
690 		m = *mp;
691 		switch (optname) {
692 
693 		case TCP_NODELAY:
694 			if (m == NULL || m->m_len < sizeof (int))
695 				error = EINVAL;
696 			else if (*mtod(m, int *))
697 				tp->t_flags |= TF_NODELAY;
698 			else
699 				tp->t_flags &= ~TF_NODELAY;
700 			break;
701 
702 		case TCP_MAXSEG:
703 			if (m && (i = *mtod(m, int *)) > 0 &&
704 			    i <= tp->t_peermss)
705 				tp->t_peermss = i;  /* limit on send size */
706 			else
707 				error = EINVAL;
708 			break;
709 
710 		default:
711 			error = ENOPROTOOPT;
712 			break;
713 		}
714 		if (m)
715 			(void) m_free(m);
716 		break;
717 
718 	case PRCO_GETOPT:
719 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
720 		m->m_len = sizeof(int);
721 		MCLAIM(m, so->so_mowner);
722 
723 		switch (optname) {
724 		case TCP_NODELAY:
725 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
726 			break;
727 		case TCP_MAXSEG:
728 			*mtod(m, int *) = tp->t_peermss;
729 			break;
730 		default:
731 			error = ENOPROTOOPT;
732 			break;
733 		}
734 		break;
735 	}
736 	splx(s);
737 	return (error);
738 }
739 
740 #ifndef TCP_SENDSPACE
741 #define	TCP_SENDSPACE	1024*32
742 #endif
743 int	tcp_sendspace = TCP_SENDSPACE;
744 #ifndef TCP_RECVSPACE
745 #define	TCP_RECVSPACE	1024*32
746 #endif
747 int	tcp_recvspace = TCP_RECVSPACE;
748 
749 /*
750  * Attach TCP protocol to socket, allocating
751  * internet protocol control block, tcp control block,
752  * bufer space, and entering LISTEN state if to accept connections.
753  */
754 int
755 tcp_attach(so)
756 	struct socket *so;
757 {
758 	struct tcpcb *tp;
759 	struct inpcb *inp;
760 #ifdef INET6
761 	struct in6pcb *in6p;
762 #endif
763 	int error;
764 	int family;	/* family of the socket */
765 
766 	family = so->so_proto->pr_domain->dom_family;
767 
768 #ifdef MBUFTRACE
769 	so->so_mowner = &tcp_mowner;
770 	so->so_rcv.sb_mowner = &tcp_rx_mowner;
771 	so->so_snd.sb_mowner = &tcp_tx_mowner;
772 #endif
773 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
774 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
775 		if (error)
776 			return (error);
777 	}
778 	switch (family) {
779 #ifdef INET
780 	case PF_INET:
781 		error = in_pcballoc(so, &tcbtable);
782 		if (error)
783 			return (error);
784 		inp = sotoinpcb(so);
785 #ifdef INET6
786 		in6p = NULL;
787 #endif
788 		break;
789 #endif
790 #ifdef INET6
791 	case PF_INET6:
792 		error = in6_pcballoc(so, &tcbtable);
793 		if (error)
794 			return (error);
795 		inp = NULL;
796 		in6p = sotoin6pcb(so);
797 		break;
798 #endif
799 	default:
800 		return EAFNOSUPPORT;
801 	}
802 	if (inp)
803 		tp = tcp_newtcpcb(family, (void *)inp);
804 #ifdef INET6
805 	else if (in6p)
806 		tp = tcp_newtcpcb(family, (void *)in6p);
807 #endif
808 	else
809 		tp = NULL;
810 
811 	if (tp == 0) {
812 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
813 
814 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
815 #ifdef INET
816 		if (inp)
817 			in_pcbdetach(inp);
818 #endif
819 #ifdef INET6
820 		if (in6p)
821 			in6_pcbdetach(in6p);
822 #endif
823 		so->so_state |= nofd;
824 		return (ENOBUFS);
825 	}
826 	tp->t_state = TCPS_CLOSED;
827 	return (0);
828 }
829 
830 /*
831  * Initiate (or continue) disconnect.
832  * If embryonic state, just send reset (once).
833  * If in ``let data drain'' option and linger null, just drop.
834  * Otherwise (hard), mark socket disconnecting and drop
835  * current input data; switch states based on user close, and
836  * send segment to peer (with FIN).
837  */
838 struct tcpcb *
839 tcp_disconnect(tp)
840 	struct tcpcb *tp;
841 {
842 	struct socket *so;
843 
844 	if (tp->t_inpcb)
845 		so = tp->t_inpcb->inp_socket;
846 #ifdef INET6
847 	else if (tp->t_in6pcb)
848 		so = tp->t_in6pcb->in6p_socket;
849 #endif
850 	else
851 		so = NULL;
852 
853 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
854 		tp = tcp_close(tp);
855 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
856 		tp = tcp_drop(tp, 0);
857 	else {
858 		soisdisconnecting(so);
859 		sbflush(&so->so_rcv);
860 		tp = tcp_usrclosed(tp);
861 		if (tp)
862 			(void) tcp_output(tp);
863 	}
864 	return (tp);
865 }
866 
867 /*
868  * User issued close, and wish to trail through shutdown states:
869  * if never received SYN, just forget it.  If got a SYN from peer,
870  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
871  * If already got a FIN from peer, then almost done; go to LAST_ACK
872  * state.  In all other cases, have already sent FIN to peer (e.g.
873  * after PRU_SHUTDOWN), and just have to play tedious game waiting
874  * for peer to send FIN or not respond to keep-alives, etc.
875  * We can let the user exit from the close as soon as the FIN is acked.
876  */
877 struct tcpcb *
878 tcp_usrclosed(tp)
879 	struct tcpcb *tp;
880 {
881 
882 	switch (tp->t_state) {
883 
884 	case TCPS_CLOSED:
885 	case TCPS_LISTEN:
886 	case TCPS_SYN_SENT:
887 		tp->t_state = TCPS_CLOSED;
888 		tp = tcp_close(tp);
889 		break;
890 
891 	case TCPS_SYN_RECEIVED:
892 	case TCPS_ESTABLISHED:
893 		tp->t_state = TCPS_FIN_WAIT_1;
894 		break;
895 
896 	case TCPS_CLOSE_WAIT:
897 		tp->t_state = TCPS_LAST_ACK;
898 		break;
899 	}
900 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
901 		struct socket *so;
902 		if (tp->t_inpcb)
903 			so = tp->t_inpcb->inp_socket;
904 #ifdef INET6
905 		else if (tp->t_in6pcb)
906 			so = tp->t_in6pcb->in6p_socket;
907 #endif
908 		else
909 			so = NULL;
910 		soisdisconnected(so);
911 		/*
912 		 * If we are in FIN_WAIT_2, we arrived here because the
913 		 * application did a shutdown of the send side.  Like the
914 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
915 		 * a full close, we start a timer to make sure sockets are
916 		 * not left in FIN_WAIT_2 forever.
917 		 */
918 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tcp_maxidle > 0))
919 			TCP_TIMER_ARM(tp, TCPT_2MSL, tcp_maxidle);
920 	}
921 	return (tp);
922 }
923 
924 static const struct {
925 	 unsigned int valid : 1;
926 	 unsigned int rdonly : 1;
927 	 int *var;
928 	 int val;
929 	 } tcp_ctlvars[] = TCPCTL_VARIABLES;
930 
931 /*
932  * Sysctl for tcp variables.
933  */
934 int
935 tcp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
936 	int *name;
937 	u_int namelen;
938 	void *oldp;
939 	size_t *oldlenp;
940 	void *newp;
941 	size_t newlen;
942 {
943 	int error, saved_value = 0;
944 
945 	if (name[0] == TCPCTL_IDENT)
946 		return tcp_sysctl_ident(&name[1], namelen - 1, oldp, oldlenp,
947 		    newp, newlen);
948 
949 	/* All remaining sysctl names at this level are terminal. */
950 	if (namelen != 1)
951 		return (ENOTDIR);
952 
953 	if (name[0] < sizeof(tcp_ctlvars)/sizeof(tcp_ctlvars[0])
954 	    && tcp_ctlvars[name[0]].valid) {
955 		if (tcp_ctlvars[name[0]].rdonly) {
956 			return (sysctl_rdint(oldp, oldlenp, newp,
957 			    tcp_ctlvars[name[0]].val));
958 		} else {
959 			switch (name[0]) {
960 			case TCPCTL_MSSDFLT:
961 				saved_value = tcp_mssdflt;
962 				break;
963 			}
964 			error = sysctl_int(oldp, oldlenp, newp, newlen,
965 			    tcp_ctlvars[name[0]].var);
966 			if (error)
967 				return (error);
968 			switch (name[0]) {
969 			case TCPCTL_MSSDFLT:
970 				if (tcp_mssdflt < 32) {
971 					tcp_mssdflt = saved_value;
972 					return (EINVAL);
973 				}
974 				break;
975 			}
976 			return (0);
977 		}
978 	}
979 
980 	return (ENOPROTOOPT);
981 }
982 
983 
984 static int
985 tcp_sysctl_ident(int *name, u_int namelen, void *oldp, size_t *oldlenp,
986     void *newp, size_t newlen)
987 {
988 	struct inpcb *inb;
989 	struct in_addr laddr, raddr;
990 	u_int lport, rport;
991 	uid_t uid;
992 	int error;
993 
994 	if (*oldlenp != sizeof(uid_t))
995 		return ENOMEM;
996 	if (!oldp || *oldlenp != sizeof(uid_t))
997 		return ENOMEM;
998 	if (namelen != 4)
999 		return EINVAL;
1000 
1001 	raddr.s_addr = (uint32_t)name[0];
1002 	rport = (u_int)name[1];
1003 	laddr.s_addr = (uint32_t)name[2];
1004 	lport = (u_int)name[3];
1005 
1006 	inb = in_pcblookup_connect(&tcbtable, raddr, rport, laddr, lport);
1007 	if (inb) {
1008 		struct socket *sockp = inb->inp_socket;
1009 		if (sockp)
1010 			uid = sockp->so_uid;
1011 		else
1012 			return ESRCH;
1013 	} else
1014 		return ESRCH;
1015 
1016 	if ((error = copyout(&uid, oldp, sizeof(uid))) != 0)
1017 		return error;
1018 
1019 	return 0;
1020 }
1021