xref: /netbsd-src/sys/netinet/tcp_usrreq.c (revision 27578b9aac214cc7796ead81dcc5427e79d5f2a0)
1 /*	$NetBSD: tcp_usrreq.c,v 1.65 2001/09/10 20:15:14 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 #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/ucred.h>
120 #include <sys/domain.h>
121 
122 #include <uvm/uvm_extern.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_recvspace.h"
153 #include "opt_tcp_sendspace.h"
154 
155 #ifdef IPSEC
156 #include <netinet6/ipsec.h>
157 #endif /*IPSEC*/
158 
159 /*
160  * TCP protocol interface to socket abstraction.
161  */
162 extern	char *tcpstates[];
163 
164 /*
165  * Process a TCP user request for TCP tb.  If this is a send request
166  * then m is the mbuf chain of send data.  If this is a timer expiration
167  * (called from the software clock routine), then timertype tells which timer.
168  */
169 /*ARGSUSED*/
170 int
171 tcp_usrreq(so, req, m, nam, control, p)
172 	struct socket *so;
173 	int req;
174 	struct mbuf *m, *nam, *control;
175 	struct proc *p;
176 {
177 	struct inpcb *inp;
178 #ifdef INET6
179 	struct in6pcb *in6p;
180 #endif
181 	struct tcpcb *tp = NULL;
182 	int s;
183 	int error = 0;
184 	int ostate;
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(&tcb6, (struct ifnet *)control);
218 			in6_purgeif((struct ifnet *)control);
219 			in6_pcbpurgeif(&tcb6, (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 		ostate = tp->t_state;
279 	}
280 #endif
281 #ifdef INET6
282 	if (in6p) {
283 		tp = in6totcpcb(in6p);
284 		/* WHAT IF TP IS 0? */
285 #ifdef KPROF
286 		tcp_acounts[tp->t_state][req]++;
287 #endif
288 		ostate = tp->t_state;
289 	}
290 #endif
291 	else
292 		ostate = 0;
293 
294 	switch (req) {
295 
296 	/*
297 	 * TCP attaches to socket via PRU_ATTACH, reserving space,
298 	 * and an internet control block.
299 	 */
300 	case PRU_ATTACH:
301 #ifndef INET6
302 		if (inp != 0)
303 #else
304 		if (inp != 0 || in6p != 0)
305 #endif
306 		{
307 			error = EISCONN;
308 			break;
309 		}
310 		error = tcp_attach(so);
311 		if (error)
312 			break;
313 		if ((so->so_options & SO_LINGER) && so->so_linger == 0)
314 			so->so_linger = TCP_LINGERTIME;
315 		tp = sototcpcb(so);
316 		break;
317 
318 	/*
319 	 * PRU_DETACH detaches the TCP protocol from the socket.
320 	 */
321 	case PRU_DETACH:
322 		tp = tcp_disconnect(tp);
323 		break;
324 
325 	/*
326 	 * Give the socket an address.
327 	 */
328 	case PRU_BIND:
329 		switch (family) {
330 #ifdef INET
331 		case PF_INET:
332 			error = in_pcbbind(inp, nam, p);
333 			break;
334 #endif
335 #ifdef INET6
336 		case PF_INET6:
337 			error = in6_pcbbind(in6p, nam, p);
338 			if (!error) {
339 				/* mapped addr case */
340 				if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
341 					tp->t_family = AF_INET;
342 				else
343 					tp->t_family = AF_INET6;
344 			}
345 			break;
346 #endif
347 		}
348 		break;
349 
350 	/*
351 	 * Prepare to accept connections.
352 	 */
353 	case PRU_LISTEN:
354 #ifdef INET
355 		if (inp && inp->inp_lport == 0) {
356 			error = in_pcbbind(inp, (struct mbuf *)0,
357 			    (struct proc *)0);
358 			if (error)
359 				break;
360 		}
361 #endif
362 #ifdef INET6
363 		if (in6p && in6p->in6p_lport == 0) {
364 			error = in6_pcbbind(in6p, (struct mbuf *)0,
365 			    (struct proc *)0);
366 			if (error)
367 				break;
368 		}
369 #endif
370 		tp->t_state = TCPS_LISTEN;
371 		break;
372 
373 	/*
374 	 * Initiate connection to peer.
375 	 * Create a template for use in transmissions on this connection.
376 	 * Enter SYN_SENT state, and mark socket as connecting.
377 	 * Start keep-alive timer, and seed output sequence space.
378 	 * Send initial segment on connection.
379 	 */
380 	case PRU_CONNECT:
381 #ifdef INET
382 		if (inp) {
383 			if (inp->inp_lport == 0) {
384 				error = in_pcbbind(inp, (struct mbuf *)0,
385 				    (struct proc *)0);
386 				if (error)
387 					break;
388 			}
389 			error = in_pcbconnect(inp, nam);
390 		}
391 #endif
392 #ifdef INET6
393 		if (in6p) {
394 			if (in6p->in6p_lport == 0) {
395 				error = in6_pcbbind(in6p, (struct mbuf *)0,
396 				    (struct proc *)0);
397 				if (error)
398 					break;
399 			}
400 			error = in6_pcbconnect(in6p, nam);
401 			if (!error) {
402 				/* mapped addr case */
403 				if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr))
404 					tp->t_family = AF_INET;
405 				else
406 					tp->t_family = AF_INET6;
407 			}
408 		}
409 #endif
410 		if (error)
411 			break;
412 		tp->t_template = tcp_template(tp);
413 		if (tp->t_template == 0) {
414 #ifdef INET
415 			if (inp)
416 				in_pcbdisconnect(inp);
417 #endif
418 #ifdef INET6
419 			if (in6p)
420 				in6_pcbdisconnect(in6p);
421 #endif
422 			error = ENOBUFS;
423 			break;
424 		}
425 		/* Compute window scaling to request.  */
426 		while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
427 		    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
428 			tp->request_r_scale++;
429 		soisconnecting(so);
430 		tcpstat.tcps_connattempt++;
431 		tp->t_state = TCPS_SYN_SENT;
432 		TCP_TIMER_ARM(tp, TCPT_KEEP, TCPTV_KEEP_INIT);
433 		tp->iss = tcp_new_iss(tp, 0);
434 		tcp_sendseqinit(tp);
435 		error = tcp_output(tp);
436 		break;
437 
438 	/*
439 	 * Create a TCP connection between two sockets.
440 	 */
441 	case PRU_CONNECT2:
442 		error = EOPNOTSUPP;
443 		break;
444 
445 	/*
446 	 * Initiate disconnect from peer.
447 	 * If connection never passed embryonic stage, just drop;
448 	 * else if don't need to let data drain, then can just drop anyways,
449 	 * else have to begin TCP shutdown process: mark socket disconnecting,
450 	 * drain unread data, state switch to reflect user close, and
451 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
452 	 * when peer sends FIN and acks ours.
453 	 *
454 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
455 	 */
456 	case PRU_DISCONNECT:
457 		tp = tcp_disconnect(tp);
458 		break;
459 
460 	/*
461 	 * Accept a connection.  Essentially all the work is
462 	 * done at higher levels; just return the address
463 	 * of the peer, storing through addr.
464 	 */
465 	case PRU_ACCEPT:
466 #ifdef INET
467 		if (inp)
468 			in_setpeeraddr(inp, nam);
469 #endif
470 #ifdef INET6
471 		if (in6p)
472 			in6_setpeeraddr(in6p, nam);
473 #endif
474 		break;
475 
476 	/*
477 	 * Mark the connection as being incapable of further output.
478 	 */
479 	case PRU_SHUTDOWN:
480 		socantsendmore(so);
481 		tp = tcp_usrclosed(tp);
482 		if (tp)
483 			error = tcp_output(tp);
484 		break;
485 
486 	/*
487 	 * After a receive, possibly send window update to peer.
488 	 */
489 	case PRU_RCVD:
490 		/*
491 		 * soreceive() calls this function when a user receives
492 		 * ancillary data on a listening socket. We don't call
493 		 * tcp_output in such a case, since there is no header
494 		 * template for a listening socket and hence the kernel
495 		 * will panic.
496 		 */
497 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0)
498 			(void) tcp_output(tp);
499 		break;
500 
501 	/*
502 	 * Do a send by putting data in output queue and updating urgent
503 	 * marker if URG set.  Possibly send more data.
504 	 */
505 	case PRU_SEND:
506 		if (control && control->m_len) {
507 			m_freem(control);
508 			m_freem(m);
509 			error = EINVAL;
510 			break;
511 		}
512 		sbappend(&so->so_snd, m);
513 		error = tcp_output(tp);
514 		break;
515 
516 	/*
517 	 * Abort the TCP.
518 	 */
519 	case PRU_ABORT:
520 		tp = tcp_drop(tp, ECONNABORTED);
521 		break;
522 
523 	case PRU_SENSE:
524 		/*
525 		 * stat: don't bother with a blocksize.
526 		 */
527 		splx(s);
528 		return (0);
529 
530 	case PRU_RCVOOB:
531 		if (control && control->m_len) {
532 			m_freem(control);
533 			m_freem(m);
534 			error = EINVAL;
535 			break;
536 		}
537 		if ((so->so_oobmark == 0 &&
538 		    (so->so_state & SS_RCVATMARK) == 0) ||
539 		    so->so_options & SO_OOBINLINE ||
540 		    tp->t_oobflags & TCPOOB_HADDATA) {
541 			error = EINVAL;
542 			break;
543 		}
544 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
545 			error = EWOULDBLOCK;
546 			break;
547 		}
548 		m->m_len = 1;
549 		*mtod(m, caddr_t) = tp->t_iobc;
550 		if (((long)nam & MSG_PEEK) == 0)
551 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
552 		break;
553 
554 	case PRU_SENDOOB:
555 		if (sbspace(&so->so_snd) < -512) {
556 			m_freem(m);
557 			error = ENOBUFS;
558 			break;
559 		}
560 		/*
561 		 * According to RFC961 (Assigned Protocols),
562 		 * the urgent pointer points to the last octet
563 		 * of urgent data.  We continue, however,
564 		 * to consider it to indicate the first octet
565 		 * of data past the urgent section.
566 		 * Otherwise, snd_up should be one lower.
567 		 */
568 		sbappend(&so->so_snd, m);
569 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
570 		tp->t_force = 1;
571 		error = tcp_output(tp);
572 		tp->t_force = 0;
573 		break;
574 
575 	case PRU_SOCKADDR:
576 #ifdef INET
577 		if (inp)
578 			in_setsockaddr(inp, nam);
579 #endif
580 #ifdef INET6
581 		if (in6p)
582 			in6_setsockaddr(in6p, nam);
583 #endif
584 		break;
585 
586 	case PRU_PEERADDR:
587 #ifdef INET
588 		if (inp)
589 			in_setpeeraddr(inp, nam);
590 #endif
591 #ifdef INET6
592 		if (in6p)
593 			in6_setpeeraddr(in6p, nam);
594 #endif
595 		break;
596 
597 	default:
598 		panic("tcp_usrreq");
599 	}
600 #ifdef TCP_DEBUG
601 	if (tp && (so->so_options & SO_DEBUG))
602 		tcp_trace(TA_USER, ostate, tp, NULL, req);
603 #endif
604 
605 release:
606 	splx(s);
607 	return (error);
608 }
609 
610 int
611 tcp_ctloutput(op, so, level, optname, mp)
612 	int op;
613 	struct socket *so;
614 	int level, optname;
615 	struct mbuf **mp;
616 {
617 	int error = 0, s;
618 	struct inpcb *inp;
619 #ifdef INET6
620 	struct in6pcb *in6p;
621 #endif
622 	struct tcpcb *tp;
623 	struct mbuf *m;
624 	int i;
625 	int family;	/* family of the socket */
626 
627 	family = so->so_proto->pr_domain->dom_family;
628 
629 	s = splsoftnet();
630 	switch (family) {
631 #ifdef INET
632 	case PF_INET:
633 		inp = sotoinpcb(so);
634 #ifdef INET6
635 		in6p = NULL;
636 #endif
637 		break;
638 #endif
639 #ifdef INET6
640 	case PF_INET6:
641 		inp = NULL;
642 		in6p = sotoin6pcb(so);
643 		break;
644 #endif
645 	default:
646 		splx(s);
647 		return EAFNOSUPPORT;
648 	}
649 #ifndef INET6
650 	if (inp == NULL)
651 #else
652 	if (inp == NULL && in6p == NULL)
653 #endif
654 	{
655 		splx(s);
656 		if (op == PRCO_SETOPT && *mp)
657 			(void) m_free(*mp);
658 		return (ECONNRESET);
659 	}
660 	if (level != IPPROTO_TCP) {
661 		switch (family) {
662 #ifdef INET
663 		case PF_INET:
664 			error = ip_ctloutput(op, so, level, optname, mp);
665 			break;
666 #endif
667 #ifdef INET6
668 		case PF_INET6:
669 			error = ip6_ctloutput(op, so, level, optname, mp);
670 			break;
671 #endif
672 		}
673 		splx(s);
674 		return (error);
675 	}
676 	if (inp)
677 		tp = intotcpcb(inp);
678 #ifdef INET6
679 	else if (in6p)
680 		tp = in6totcpcb(in6p);
681 #endif
682 	else
683 		tp = NULL;
684 
685 	switch (op) {
686 
687 	case PRCO_SETOPT:
688 		m = *mp;
689 		switch (optname) {
690 
691 		case TCP_NODELAY:
692 			if (m == NULL || m->m_len < sizeof (int))
693 				error = EINVAL;
694 			else if (*mtod(m, int *))
695 				tp->t_flags |= TF_NODELAY;
696 			else
697 				tp->t_flags &= ~TF_NODELAY;
698 			break;
699 
700 		case TCP_MAXSEG:
701 			if (m && (i = *mtod(m, int *)) > 0 &&
702 			    i <= tp->t_peermss)
703 				tp->t_peermss = i;  /* limit on send size */
704 			else
705 				error = EINVAL;
706 			break;
707 
708 		default:
709 			error = ENOPROTOOPT;
710 			break;
711 		}
712 		if (m)
713 			(void) m_free(m);
714 		break;
715 
716 	case PRCO_GETOPT:
717 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
718 		m->m_len = sizeof(int);
719 
720 		switch (optname) {
721 		case TCP_NODELAY:
722 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
723 			break;
724 		case TCP_MAXSEG:
725 			*mtod(m, int *) = tp->t_peermss;
726 			break;
727 		default:
728 			error = ENOPROTOOPT;
729 			break;
730 		}
731 		break;
732 	}
733 	splx(s);
734 	return (error);
735 }
736 
737 #ifndef TCP_SENDSPACE
738 #define	TCP_SENDSPACE	1024*16;
739 #endif
740 int	tcp_sendspace = TCP_SENDSPACE;
741 #ifndef TCP_RECVSPACE
742 #define	TCP_RECVSPACE	1024*16;
743 #endif
744 int	tcp_recvspace = TCP_RECVSPACE;
745 
746 /*
747  * Attach TCP protocol to socket, allocating
748  * internet protocol control block, tcp control block,
749  * bufer space, and entering LISTEN state if to accept connections.
750  */
751 int
752 tcp_attach(so)
753 	struct socket *so;
754 {
755 	struct tcpcb *tp;
756 	struct inpcb *inp;
757 #ifdef INET6
758 	struct in6pcb *in6p;
759 #endif
760 	int error;
761 	int family;	/* family of the socket */
762 
763 	family = so->so_proto->pr_domain->dom_family;
764 
765 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
766 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
767 		if (error)
768 			return (error);
769 	}
770 	switch (family) {
771 #ifdef INET
772 	case PF_INET:
773 		error = in_pcballoc(so, &tcbtable);
774 		if (error)
775 			return (error);
776 		inp = sotoinpcb(so);
777 #ifdef INET6
778 		in6p = NULL;
779 #endif
780 		break;
781 #endif
782 #ifdef INET6
783 	case PF_INET6:
784 		error = in6_pcballoc(so, &tcb6);
785 		if (error)
786 			return (error);
787 		inp = NULL;
788 		in6p = sotoin6pcb(so);
789 		break;
790 #endif
791 	default:
792 		return EAFNOSUPPORT;
793 	}
794 	if (inp)
795 		tp = tcp_newtcpcb(family, (void *)inp);
796 #ifdef INET6
797 	else if (in6p)
798 		tp = tcp_newtcpcb(family, (void *)in6p);
799 #endif
800 	else
801 		tp = NULL;
802 
803 	if (tp == 0) {
804 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
805 
806 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
807 #ifdef INET
808 		if (inp)
809 			in_pcbdetach(inp);
810 #endif
811 #ifdef INET6
812 		if (in6p)
813 			in6_pcbdetach(in6p);
814 #endif
815 		so->so_state |= nofd;
816 		return (ENOBUFS);
817 	}
818 	tp->t_state = TCPS_CLOSED;
819 	return (0);
820 }
821 
822 /*
823  * Initiate (or continue) disconnect.
824  * If embryonic state, just send reset (once).
825  * If in ``let data drain'' option and linger null, just drop.
826  * Otherwise (hard), mark socket disconnecting and drop
827  * current input data; switch states based on user close, and
828  * send segment to peer (with FIN).
829  */
830 struct tcpcb *
831 tcp_disconnect(tp)
832 	struct tcpcb *tp;
833 {
834 	struct socket *so;
835 
836 	if (tp->t_inpcb)
837 		so = tp->t_inpcb->inp_socket;
838 #ifdef INET6
839 	else if (tp->t_in6pcb)
840 		so = tp->t_in6pcb->in6p_socket;
841 #endif
842 	else
843 		so = NULL;
844 
845 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
846 		tp = tcp_close(tp);
847 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
848 		tp = tcp_drop(tp, 0);
849 	else {
850 		soisdisconnecting(so);
851 		sbflush(&so->so_rcv);
852 		tp = tcp_usrclosed(tp);
853 		if (tp)
854 			(void) tcp_output(tp);
855 	}
856 	return (tp);
857 }
858 
859 /*
860  * User issued close, and wish to trail through shutdown states:
861  * if never received SYN, just forget it.  If got a SYN from peer,
862  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
863  * If already got a FIN from peer, then almost done; go to LAST_ACK
864  * state.  In all other cases, have already sent FIN to peer (e.g.
865  * after PRU_SHUTDOWN), and just have to play tedious game waiting
866  * for peer to send FIN or not respond to keep-alives, etc.
867  * We can let the user exit from the close as soon as the FIN is acked.
868  */
869 struct tcpcb *
870 tcp_usrclosed(tp)
871 	struct tcpcb *tp;
872 {
873 
874 	switch (tp->t_state) {
875 
876 	case TCPS_CLOSED:
877 	case TCPS_LISTEN:
878 	case TCPS_SYN_SENT:
879 		tp->t_state = TCPS_CLOSED;
880 		tp = tcp_close(tp);
881 		break;
882 
883 	case TCPS_SYN_RECEIVED:
884 	case TCPS_ESTABLISHED:
885 		tp->t_state = TCPS_FIN_WAIT_1;
886 		break;
887 
888 	case TCPS_CLOSE_WAIT:
889 		tp->t_state = TCPS_LAST_ACK;
890 		break;
891 	}
892 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
893 		struct socket *so;
894 		if (tp->t_inpcb)
895 			so = tp->t_inpcb->inp_socket;
896 #ifdef INET6
897 		else if (tp->t_in6pcb)
898 			so = tp->t_in6pcb->in6p_socket;
899 #endif
900 		else
901 			so = NULL;
902 		soisdisconnected(so);
903 		/*
904 		 * If we are in FIN_WAIT_2, we arrived here because the
905 		 * application did a shutdown of the send side.  Like the
906 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
907 		 * a full close, we start a timer to make sure sockets are
908 		 * not left in FIN_WAIT_2 forever.
909 		 */
910 		if ((tp->t_state == TCPS_FIN_WAIT_2) && (tcp_maxidle > 0))
911 			TCP_TIMER_ARM(tp, TCPT_2MSL, tcp_maxidle);
912 	}
913 	return (tp);
914 }
915 
916 static const struct {
917 	 unsigned int valid : 1;
918 	 unsigned int rdonly : 1;
919 	 int *var;
920 	 int val;
921 	 } tcp_ctlvars[] = TCPCTL_VARIABLES;
922 
923 /*
924  * Sysctl for tcp variables.
925  */
926 int
927 tcp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
928 	int *name;
929 	u_int namelen;
930 	void *oldp;
931 	size_t *oldlenp;
932 	void *newp;
933 	size_t newlen;
934 {
935 
936 	/* All sysctl names at this level are terminal. */
937 	if (namelen != 1)
938 		return (ENOTDIR);
939 
940 	if (name[0] < sizeof(tcp_ctlvars)/sizeof(tcp_ctlvars[0])
941 	    && tcp_ctlvars[name[0]].valid) {
942 		if (tcp_ctlvars[name[0]].rdonly)
943 			return (sysctl_rdint(oldp, oldlenp, newp,
944 			    tcp_ctlvars[name[0]].val));
945 		else
946 			return (sysctl_int(oldp, oldlenp, newp, newlen,
947 			    tcp_ctlvars[name[0]].var));
948 	}
949 
950 	return (ENOPROTOOPT);
951 }
952