xref: /openbsd-src/sys/netinet/tcp_usrreq.c (revision a28daedfc357b214be5c701aa8ba8adb29a7f1c2)
1 /*	$OpenBSD: tcp_usrreq.c,v 1.99 2008/05/24 19:48:32 thib Exp $	*/
2 /*	$NetBSD: tcp_usrreq.c,v 1.20 1996/02/13 23:44:16 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)COPYRIGHT	1.1 (NRL) 17 January 1995
33  *
34  * NRL grants permission for redistribution and use in source and binary
35  * forms, with or without modification, of the software and documentation
36  * created at NRL provided that the following conditions are met:
37  *
38  * 1. Redistributions of source code must retain the above copyright
39  *    notice, this list of conditions and the following disclaimer.
40  * 2. Redistributions in binary form must reproduce the above copyright
41  *    notice, this list of conditions and the following disclaimer in the
42  *    documentation and/or other materials provided with the distribution.
43  * 3. All advertising materials mentioning features or use of this software
44  *    must display the following acknowledgements:
45  * 	This product includes software developed by the University of
46  * 	California, Berkeley and its contributors.
47  * 	This product includes software developed at the Information
48  * 	Technology Division, US Naval Research Laboratory.
49  * 4. Neither the name of the NRL nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
54  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
55  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
56  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
57  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
58  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
59  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
60  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
61  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
62  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
63  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
64  *
65  * The views and conclusions contained in the software and documentation
66  * are those of the authors and should not be interpreted as representing
67  * official policies, either expressed or implied, of the US Naval
68  * Research Laboratory (NRL).
69  */
70 
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/mbuf.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/protosw.h>
77 #include <sys/stat.h>
78 #include <sys/sysctl.h>
79 #include <sys/domain.h>
80 #include <sys/kernel.h>
81 #include <sys/pool.h>
82 
83 #include <dev/rndvar.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_systm.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet/ip_var.h>
94 #include <netinet/tcp.h>
95 #include <netinet/tcp_fsm.h>
96 #include <netinet/tcp_seq.h>
97 #include <netinet/tcp_timer.h>
98 #include <netinet/tcp_var.h>
99 #include <netinet/tcpip.h>
100 #include <netinet/tcp_debug.h>
101 
102 /*
103  * TCP protocol interface to socket abstraction.
104  */
105 extern	char *tcpstates[];
106 extern	int tcptv_keep_init;
107 
108 extern int tcp_rst_ppslim;
109 
110 /* from in_pcb.c */
111 extern	struct baddynamicports baddynamicports;
112 
113 #ifndef TCP_SENDSPACE
114 #define	TCP_SENDSPACE	1024*16
115 #endif
116 u_int	tcp_sendspace = TCP_SENDSPACE;
117 #ifndef TCP_RECVSPACE
118 #define	TCP_RECVSPACE	1024*16
119 #endif
120 u_int	tcp_recvspace = TCP_RECVSPACE;
121 
122 int *tcpctl_vars[TCPCTL_MAXID] = TCPCTL_VARS;
123 
124 struct	inpcbtable tcbtable;
125 
126 int tcp_ident(void *, size_t *, void *, size_t, int);
127 
128 /*
129  * Process a TCP user request for TCP tb.  If this is a send request
130  * then m is the mbuf chain of send data.  If this is a timer expiration
131  * (called from the software clock routine), then timertype tells which timer.
132  */
133 /*ARGSUSED*/
134 int
135 tcp_usrreq(so, req, m, nam, control, p)
136 	struct socket *so;
137 	int req;
138 	struct mbuf *m, *nam, *control;
139 	struct proc *p;
140 {
141 	struct sockaddr_in *sin;
142 	struct inpcb *inp;
143 	struct tcpcb *tp = NULL;
144 	int s;
145 	int error = 0;
146 	short ostate;
147 
148 	if (req == PRU_CONTROL) {
149 #ifdef INET6
150 		if (sotopf(so) == PF_INET6)
151 			return in6_control(so, (u_long)m, (caddr_t)nam,
152 			    (struct ifnet *)control, 0);
153 		else
154 #endif /* INET6 */
155 			return (in_control(so, (u_long)m, (caddr_t)nam,
156 			    (struct ifnet *)control));
157 	}
158 	if (control && control->m_len) {
159 		m_freem(control);
160 		if (m)
161 			m_freem(m);
162 		return (EINVAL);
163 	}
164 
165 	s = splsoftnet();
166 	inp = sotoinpcb(so);
167 	/*
168 	 * When a TCP is attached to a socket, then there will be
169 	 * a (struct inpcb) pointed at by the socket, and this
170 	 * structure will point at a subsidiary (struct tcpcb).
171 	 */
172 	if (inp == 0 && req != PRU_ATTACH) {
173 		error = so->so_error;
174 		if (error == 0)
175 			error = EINVAL;
176 		splx(s);
177 		/*
178 		 * The following corrects an mbuf leak under rare
179 		 * circumstances
180 		 */
181 		if (m && (req == PRU_SEND || req == PRU_SENDOOB))
182 			m_freem(m);
183 		return (error);
184 	}
185 	if (inp) {
186 		tp = intotcpcb(inp);
187 		/* WHAT IF TP IS 0? */
188 #ifdef KPROF
189 		tcp_acounts[tp->t_state][req]++;
190 #endif
191 		ostate = tp->t_state;
192 	} else
193 		ostate = 0;
194 	switch (req) {
195 
196 	/*
197 	 * TCP attaches to socket via PRU_ATTACH, reserving space,
198 	 * and an internet control block.
199 	 */
200 	case PRU_ATTACH:
201 		if (inp) {
202 			error = EISCONN;
203 			break;
204 		}
205 		error = tcp_attach(so);
206 		if (error)
207 			break;
208 		if ((so->so_options & SO_LINGER) && so->so_linger == 0)
209 			so->so_linger = TCP_LINGERTIME;
210 		tp = sototcpcb(so);
211 		break;
212 
213 	/*
214 	 * PRU_DETACH detaches the TCP protocol from the socket.
215 	 * If the protocol state is non-embryonic, then can't
216 	 * do this directly: have to initiate a PRU_DISCONNECT,
217 	 * which may finish later; embryonic TCB's can just
218 	 * be discarded here.
219 	 */
220 	case PRU_DETACH:
221 		tp = tcp_disconnect(tp);
222 		break;
223 
224 	/*
225 	 * Give the socket an address.
226 	 */
227 	case PRU_BIND:
228 #ifdef INET6
229 		if (inp->inp_flags & INP_IPV6)
230 			error = in6_pcbbind(inp, nam, p);
231 		else
232 #endif
233 			error = in_pcbbind(inp, nam, p);
234 		if (error)
235 			break;
236 		break;
237 
238 	/*
239 	 * Prepare to accept connections.
240 	 */
241 	case PRU_LISTEN:
242 		if (inp->inp_lport == 0) {
243 #ifdef INET6
244 			if (inp->inp_flags & INP_IPV6)
245 				error = in6_pcbbind(inp, NULL, p);
246 			else
247 #endif
248 				error = in_pcbbind(inp, NULL, p);
249 		}
250 		/* If the in_pcbbind() above is called, the tp->pf
251 		   should still be whatever it was before. */
252 		if (error == 0)
253 			tp->t_state = TCPS_LISTEN;
254 		break;
255 
256 	/*
257 	 * Initiate connection to peer.
258 	 * Create a template for use in transmissions on this connection.
259 	 * Enter SYN_SENT state, and mark socket as connecting.
260 	 * Start keep-alive timer, and seed output sequence space.
261 	 * Send initial segment on connection.
262 	 */
263 	case PRU_CONNECT:
264 		sin = mtod(nam, struct sockaddr_in *);
265 
266 #ifdef INET6
267 		if (sin->sin_family == AF_INET6) {
268 			struct in6_addr *in6_addr = &mtod(nam,
269 			    struct sockaddr_in6 *)->sin6_addr;
270 
271 			if (IN6_IS_ADDR_UNSPECIFIED(in6_addr) ||
272 			    IN6_IS_ADDR_MULTICAST(in6_addr) ||
273 			    (IN6_IS_ADDR_V4MAPPED(in6_addr) &&
274 			    ((in6_addr->s6_addr32[3] == INADDR_ANY) ||
275 			    IN_MULTICAST(in6_addr->s6_addr32[3]) ||
276 			    in_broadcast(sin->sin_addr, NULL)))) {
277 				error = EINVAL;
278 				break;
279 			}
280 
281 			if (inp->inp_lport == 0) {
282 				error = in6_pcbbind(inp, NULL, p);
283 				if (error)
284 					break;
285 			}
286 			error = in6_pcbconnect(inp, nam);
287 		} else if (sin->sin_family == AF_INET)
288 #endif /* INET6 */
289 		{
290 			if ((sin->sin_addr.s_addr == INADDR_ANY) ||
291 			    IN_MULTICAST(sin->sin_addr.s_addr) ||
292 			    in_broadcast(sin->sin_addr, NULL)) {
293 				error = EINVAL;
294 				break;
295 			}
296 
297 			if (inp->inp_lport == 0) {
298 				error = in_pcbbind(inp, NULL, p);
299 				if (error)
300 					break;
301 			}
302 			error = in_pcbconnect(inp, nam);
303 		}
304 
305 		if (error)
306 			break;
307 
308 		tp->t_template = tcp_template(tp);
309 		if (tp->t_template == 0) {
310 			in_pcbdisconnect(inp);
311 			error = ENOBUFS;
312 			break;
313 		}
314 
315 		so->so_state |= SS_CONNECTOUT;
316 
317 		/* Compute window scaling to request.  */
318 		tcp_rscale(tp, so->so_rcv.sb_hiwat);
319 
320 		soisconnecting(so);
321 		tcpstat.tcps_connattempt++;
322 		tp->t_state = TCPS_SYN_SENT;
323 		TCP_TIMER_ARM(tp, TCPT_KEEP, tcptv_keep_init);
324 		tcp_set_iss_tsm(tp);
325 		tcp_sendseqinit(tp);
326 #if defined(TCP_SACK)
327 		tp->snd_last = tp->snd_una;
328 #endif
329 #if defined(TCP_SACK) && defined(TCP_FACK)
330 		tp->snd_fack = tp->snd_una;
331 		tp->retran_data = 0;
332 		tp->snd_awnd = 0;
333 #endif
334 		error = tcp_output(tp);
335 		break;
336 
337 	/*
338 	 * Create a TCP connection between two sockets.
339 	 */
340 	case PRU_CONNECT2:
341 		error = EOPNOTSUPP;
342 		break;
343 
344 	/*
345 	 * Initiate disconnect from peer.
346 	 * If connection never passed embryonic stage, just drop;
347 	 * else if don't need to let data drain, then can just drop anyways,
348 	 * else have to begin TCP shutdown process: mark socket disconnecting,
349 	 * drain unread data, state switch to reflect user close, and
350 	 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
351 	 * when peer sends FIN and acks ours.
352 	 *
353 	 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
354 	 */
355 	case PRU_DISCONNECT:
356 		tp = tcp_disconnect(tp);
357 		break;
358 
359 	/*
360 	 * Accept a connection.  Essentially all the work is
361 	 * done at higher levels; just return the address
362 	 * of the peer, storing through addr.
363 	 */
364 	case PRU_ACCEPT:
365 #ifdef INET6
366 		if (inp->inp_flags & INP_IPV6)
367 			in6_setpeeraddr(inp, nam);
368 		else
369 #endif
370 			in_setpeeraddr(inp, nam);
371 		break;
372 
373 	/*
374 	 * Mark the connection as being incapable of further output.
375 	 */
376 	case PRU_SHUTDOWN:
377 		if (so->so_state & SS_CANTSENDMORE)
378 			break;
379 		socantsendmore(so);
380 		tp = tcp_usrclosed(tp);
381 		if (tp)
382 			error = tcp_output(tp);
383 		break;
384 
385 	/*
386 	 * After a receive, possibly send window update to peer.
387 	 */
388 	case PRU_RCVD:
389 		/*
390 		 * soreceive() calls this function when a user receives
391 		 * ancillary data on a listening socket. We don't call
392 		 * tcp_output in such a case, since there is no header
393 		 * template for a listening socket and hence the kernel
394 		 * will panic.
395 		 */
396 		if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) != 0)
397 			(void) tcp_output(tp);
398 		break;
399 
400 	/*
401 	 * Do a send by putting data in output queue and updating urgent
402 	 * marker if URG set.  Possibly send more data.
403 	 */
404 	case PRU_SEND:
405 		sbappendstream(&so->so_snd, m);
406 		error = tcp_output(tp);
407 		break;
408 
409 	/*
410 	 * Abort the TCP.
411 	 */
412 	case PRU_ABORT:
413 		tp = tcp_drop(tp, ECONNABORTED);
414 		break;
415 
416 	case PRU_SENSE:
417 		((struct stat *) m)->st_blksize = so->so_snd.sb_hiwat;
418 		splx(s);
419 		return (0);
420 
421 	case PRU_RCVOOB:
422 		if ((so->so_oobmark == 0 &&
423 		    (so->so_state & SS_RCVATMARK) == 0) ||
424 		    so->so_options & SO_OOBINLINE ||
425 		    tp->t_oobflags & TCPOOB_HADDATA) {
426 			error = EINVAL;
427 			break;
428 		}
429 		if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
430 			error = EWOULDBLOCK;
431 			break;
432 		}
433 		m->m_len = 1;
434 		*mtod(m, caddr_t) = tp->t_iobc;
435 		if (((long)nam & MSG_PEEK) == 0)
436 			tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
437 		break;
438 
439 	case PRU_SENDOOB:
440 		if (sbspace(&so->so_snd) < -512) {
441 			m_freem(m);
442 			error = ENOBUFS;
443 			break;
444 		}
445 		/*
446 		 * According to RFC961 (Assigned Protocols),
447 		 * the urgent pointer points to the last octet
448 		 * of urgent data.  We continue, however,
449 		 * to consider it to indicate the first octet
450 		 * of data past the urgent section.
451 		 * Otherwise, snd_up should be one lower.
452 		 */
453 		sbappendstream(&so->so_snd, m);
454 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
455 		tp->t_force = 1;
456 		error = tcp_output(tp);
457 		tp->t_force = 0;
458 		break;
459 
460 	case PRU_SOCKADDR:
461 #ifdef INET6
462 		if (inp->inp_flags & INP_IPV6)
463 			in6_setsockaddr(inp, nam);
464 		else
465 #endif
466 			in_setsockaddr(inp, nam);
467 		break;
468 
469 	case PRU_PEERADDR:
470 #ifdef INET6
471 		if (inp->inp_flags & INP_IPV6)
472 			in6_setpeeraddr(inp, nam);
473 		else
474 #endif
475 			in_setpeeraddr(inp, nam);
476 		break;
477 
478 	default:
479 		panic("tcp_usrreq");
480 	}
481 	if (tp && (so->so_options & SO_DEBUG))
482 		tcp_trace(TA_USER, ostate, tp, (caddr_t)0, req, 0);
483 	splx(s);
484 	return (error);
485 }
486 
487 int
488 tcp_ctloutput(op, so, level, optname, mp)
489 	int op;
490 	struct socket *so;
491 	int level, optname;
492 	struct mbuf **mp;
493 {
494 	int error = 0, s;
495 	struct inpcb *inp;
496 	struct tcpcb *tp;
497 	struct mbuf *m;
498 	int i;
499 
500 	s = splsoftnet();
501 	inp = sotoinpcb(so);
502 	if (inp == NULL) {
503 		splx(s);
504 		if (op == PRCO_SETOPT && *mp)
505 			(void) m_free(*mp);
506 		return (ECONNRESET);
507 	}
508 #ifdef INET6
509 	tp = intotcpcb(inp);
510 #endif /* INET6 */
511 	if (level != IPPROTO_TCP) {
512 		switch (so->so_proto->pr_domain->dom_family) {
513 #ifdef INET6
514 		case PF_INET6:
515 			error = ip6_ctloutput(op, so, level, optname, mp);
516 			break;
517 #endif /* INET6 */
518 		case PF_INET:
519 			error = ip_ctloutput(op, so, level, optname, mp);
520 			break;
521 		default:
522 			error = EAFNOSUPPORT;	/*?*/
523 			break;
524 		}
525 		splx(s);
526 		return (error);
527 	}
528 #ifndef INET6
529 	tp = intotcpcb(inp);
530 #endif /* !INET6 */
531 
532 	switch (op) {
533 
534 	case PRCO_SETOPT:
535 		m = *mp;
536 		switch (optname) {
537 
538 		case TCP_NODELAY:
539 			if (m == NULL || m->m_len < sizeof (int))
540 				error = EINVAL;
541 			else if (*mtod(m, int *))
542 				tp->t_flags |= TF_NODELAY;
543 			else
544 				tp->t_flags &= ~TF_NODELAY;
545 			break;
546 
547 		case TCP_MAXSEG:
548 			if (m == NULL || m->m_len < sizeof (int)) {
549 				error = EINVAL;
550 				break;
551 			}
552 
553 			i = *mtod(m, int *);
554 			if (i > 0 && i <= tp->t_maxseg)
555 				tp->t_maxseg = i;
556 			else
557 				error = EINVAL;
558 			break;
559 
560 #ifdef TCP_SACK
561 		case TCP_SACK_ENABLE:
562 			if (m == NULL || m->m_len < sizeof (int)) {
563 				error = EINVAL;
564 				break;
565 			}
566 
567 			if (TCPS_HAVEESTABLISHED(tp->t_state)) {
568 				error = EPERM;
569 				break;
570 			}
571 
572 			if (tp->t_flags & TF_SIGNATURE) {
573 				error = EPERM;
574 				break;
575 			}
576 
577 			if (*mtod(m, int *))
578 				tp->sack_enable = 1;
579 			else
580 				tp->sack_enable = 0;
581 			break;
582 #endif
583 #ifdef TCP_SIGNATURE
584 		case TCP_MD5SIG:
585 			if (m == NULL || m->m_len < sizeof (int)) {
586 				error = EINVAL;
587 				break;
588 			}
589 
590 			if (TCPS_HAVEESTABLISHED(tp->t_state)) {
591 				error = EPERM;
592 				break;
593 			}
594 
595 			if (*mtod(m, int *)) {
596 				tp->t_flags |= TF_SIGNATURE;
597 #ifdef TCP_SACK
598 				tp->sack_enable = 0;
599 #endif /* TCP_SACK */
600 			} else
601 				tp->t_flags &= ~TF_SIGNATURE;
602 			break;
603 #endif /* TCP_SIGNATURE */
604 		default:
605 			error = ENOPROTOOPT;
606 			break;
607 		}
608 		if (m)
609 			(void) m_free(m);
610 		break;
611 
612 	case PRCO_GETOPT:
613 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
614 		m->m_len = sizeof(int);
615 
616 		switch (optname) {
617 		case TCP_NODELAY:
618 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
619 			break;
620 		case TCP_MAXSEG:
621 			*mtod(m, int *) = tp->t_maxseg;
622 			break;
623 #ifdef TCP_SACK
624 		case TCP_SACK_ENABLE:
625 			*mtod(m, int *) = tp->sack_enable;
626 			break;
627 #endif
628 #ifdef TCP_SIGNATURE
629 		case TCP_MD5SIG:
630 			*mtod(m, int *) = tp->t_flags & TF_SIGNATURE;
631 			break;
632 #endif
633 		default:
634 			error = ENOPROTOOPT;
635 			break;
636 		}
637 		break;
638 	}
639 	splx(s);
640 	return (error);
641 }
642 
643 /*
644  * Attach TCP protocol to socket, allocating
645  * internet protocol control block, tcp control block,
646  * bufer space, and entering LISTEN state if to accept connections.
647  */
648 int
649 tcp_attach(so)
650 	struct socket *so;
651 {
652 	struct tcpcb *tp;
653 	struct inpcb *inp;
654 	int error;
655 
656 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
657 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
658 		if (error)
659 			return (error);
660 	}
661 	error = in_pcballoc(so, &tcbtable);
662 	if (error)
663 		return (error);
664 	inp = sotoinpcb(so);
665 	tp = tcp_newtcpcb(inp);
666 	if (tp == NULL) {
667 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
668 
669 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
670 		in_pcbdetach(inp);
671 		so->so_state |= nofd;
672 		return (ENOBUFS);
673 	}
674 	tp->t_state = TCPS_CLOSED;
675 #ifdef INET6
676 	/* we disallow IPv4 mapped address completely. */
677 	if (inp->inp_flags & INP_IPV6)
678 		tp->pf = PF_INET6;
679 	else
680 		tp->pf = PF_INET;
681 #else
682 	tp->pf = PF_INET;
683 #endif
684 	return (0);
685 }
686 
687 /*
688  * Initiate (or continue) disconnect.
689  * If embryonic state, just send reset (once).
690  * If in ``let data drain'' option and linger null, just drop.
691  * Otherwise (hard), mark socket disconnecting and drop
692  * current input data; switch states based on user close, and
693  * send segment to peer (with FIN).
694  */
695 struct tcpcb *
696 tcp_disconnect(tp)
697 	struct tcpcb *tp;
698 {
699 	struct socket *so = tp->t_inpcb->inp_socket;
700 
701 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
702 		tp = tcp_close(tp);
703 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
704 		tp = tcp_drop(tp, 0);
705 	else {
706 		soisdisconnecting(so);
707 		sbflush(&so->so_rcv);
708 		tp = tcp_usrclosed(tp);
709 		if (tp)
710 			(void) tcp_output(tp);
711 	}
712 	return (tp);
713 }
714 
715 /*
716  * User issued close, and wish to trail through shutdown states:
717  * if never received SYN, just forget it.  If got a SYN from peer,
718  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
719  * If already got a FIN from peer, then almost done; go to LAST_ACK
720  * state.  In all other cases, have already sent FIN to peer (e.g.
721  * after PRU_SHUTDOWN), and just have to play tedious game waiting
722  * for peer to send FIN or not respond to keep-alives, etc.
723  * We can let the user exit from the close as soon as the FIN is acked.
724  */
725 struct tcpcb *
726 tcp_usrclosed(tp)
727 	struct tcpcb *tp;
728 {
729 
730 	switch (tp->t_state) {
731 
732 	case TCPS_CLOSED:
733 	case TCPS_LISTEN:
734 	case TCPS_SYN_SENT:
735 		tp->t_state = TCPS_CLOSED;
736 		tp = tcp_close(tp);
737 		break;
738 
739 	case TCPS_SYN_RECEIVED:
740 	case TCPS_ESTABLISHED:
741 		tp->t_state = TCPS_FIN_WAIT_1;
742 		break;
743 
744 	case TCPS_CLOSE_WAIT:
745 		tp->t_state = TCPS_LAST_ACK;
746 		break;
747 	}
748 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
749 		soisdisconnected(tp->t_inpcb->inp_socket);
750 		/*
751 		 * If we are in FIN_WAIT_2, we arrived here because the
752 		 * application did a shutdown of the send side.  Like the
753 		 * case of a transition from FIN_WAIT_1 to FIN_WAIT_2 after
754 		 * a full close, we start a timer to make sure sockets are
755 		 * not left in FIN_WAIT_2 forever.
756 		 */
757 		if (tp->t_state == TCPS_FIN_WAIT_2)
758 			TCP_TIMER_ARM(tp, TCPT_2MSL, tcp_maxidle);
759 	}
760 	return (tp);
761 }
762 
763 /*
764  * Look up a socket for ident or tcpdrop, ...
765  */
766 int
767 tcp_ident(oldp, oldlenp, newp, newlen, dodrop)
768 	void *oldp;
769 	size_t *oldlenp;
770 	void *newp;
771 	size_t newlen;
772 	int dodrop;
773 {
774 	int error = 0, s;
775 	struct tcp_ident_mapping tir;
776 	struct inpcb *inp;
777 	struct tcpcb *tp = NULL;
778 	struct sockaddr_in *fin, *lin;
779 #ifdef INET6
780 	struct sockaddr_in6 *fin6, *lin6;
781 	struct in6_addr f6, l6;
782 #endif
783 	if (dodrop) {
784 		if (oldp != NULL || *oldlenp != 0)
785 			return (EINVAL);
786 		if (newp == NULL)
787 			return (EPERM);
788 		if (newlen < sizeof(tir))
789 			return (ENOMEM);
790 		if ((error = copyin(newp, &tir, sizeof (tir))) != 0 )
791 			return (error);
792 	} else {
793 		if (oldp == NULL)
794 			return (EINVAL);
795 		if (*oldlenp < sizeof(tir))
796 			return (ENOMEM);
797 		if (newp != NULL || newlen != 0)
798 			return (EINVAL);
799 		if ((error = copyin(oldp, &tir, sizeof (tir))) != 0 )
800 			return (error);
801 	}
802 	switch (tir.faddr.ss_family) {
803 #ifdef INET6
804 	case AF_INET6:
805 		fin6 = (struct sockaddr_in6 *)&tir.faddr;
806 		error = in6_embedscope(&f6, fin6, NULL, NULL);
807 		if (error)
808 			return EINVAL;	/*?*/
809 		lin6 = (struct sockaddr_in6 *)&tir.laddr;
810 		error = in6_embedscope(&l6, lin6, NULL, NULL);
811 		if (error)
812 			return EINVAL;	/*?*/
813 		break;
814 #endif
815 	case AF_INET:
816 	  	fin = (struct sockaddr_in *)&tir.faddr;
817 		lin = (struct sockaddr_in *)&tir.laddr;
818 		break;
819 	default:
820 		return (EINVAL);
821 	}
822 
823 	s = splsoftnet();
824 	switch (tir.faddr.ss_family) {
825 #ifdef INET6
826 	case AF_INET6:
827 		inp = in6_pcbhashlookup(&tcbtable, &f6,
828 		    fin6->sin6_port, &l6, lin6->sin6_port);
829 		break;
830 #endif
831 	case AF_INET:
832 		inp = in_pcbhashlookup(&tcbtable,  fin->sin_addr,
833 		    fin->sin_port, lin->sin_addr, lin->sin_port);
834 		break;
835 	}
836 
837 	if (dodrop) {
838 		if (inp && (tp = intotcpcb(inp)) &&
839 		    ((inp->inp_socket->so_options & SO_ACCEPTCONN) == 0))
840 			tp = tcp_drop(tp, ECONNABORTED);
841 		else
842 			error = ESRCH;
843 		splx(s);
844 		return (error);
845 	}
846 
847 	if (inp == NULL) {
848 		++tcpstat.tcps_pcbhashmiss;
849 		switch (tir.faddr.ss_family) {
850 #ifdef INET6
851 		case AF_INET6:
852 			inp = in6_pcblookup_listen(&tcbtable,
853 			    &l6, lin6->sin6_port, 0, NULL);
854 			break;
855 #endif
856 		case AF_INET:
857 			inp = in_pcblookup_listen(&tcbtable,
858 			    lin->sin_addr, lin->sin_port, 0, NULL);
859 			break;
860 		}
861 	}
862 
863 	if (inp != NULL && (inp->inp_socket->so_state & SS_CONNECTOUT)) {
864 		tir.ruid = inp->inp_socket->so_ruid;
865 		tir.euid = inp->inp_socket->so_euid;
866 	} else {
867 		tir.ruid = -1;
868 		tir.euid = -1;
869 	}
870 	splx(s);
871 
872 	*oldlenp = sizeof (tir);
873 	error = copyout((void *)&tir, oldp, sizeof (tir));
874 	return (error);
875 }
876 
877 /*
878  * Sysctl for tcp variables.
879  */
880 int
881 tcp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
882 	int *name;
883 	u_int namelen;
884 	void *oldp;
885 	size_t *oldlenp;
886 	void *newp;
887 	size_t newlen;
888 {
889 	int error, nval;
890 
891 	/* All sysctl names at this level are terminal. */
892 	if (namelen != 1)
893 		return (ENOTDIR);
894 
895 	switch (name[0]) {
896 #ifdef TCP_SACK
897 	case TCPCTL_SACK:
898 		return (sysctl_int(oldp, oldlenp, newp, newlen,
899 		    &tcp_do_sack));
900 #endif
901 	case TCPCTL_SLOWHZ:
902 		return (sysctl_rdint(oldp, oldlenp, newp, PR_SLOWHZ));
903 
904 	case TCPCTL_BADDYNAMIC:
905 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
906 		    baddynamicports.tcp, sizeof(baddynamicports.tcp)));
907 
908 	case TCPCTL_IDENT:
909 		return (tcp_ident(oldp, oldlenp, newp, newlen, 0));
910 
911 	case TCPCTL_DROP:
912 		return (tcp_ident(oldp, oldlenp, newp, newlen, 1));
913 
914 #ifdef TCP_ECN
915 	case TCPCTL_ECN:
916 		return (sysctl_int(oldp, oldlenp, newp, newlen,
917 		   &tcp_do_ecn));
918 #endif
919 	case TCPCTL_REASS_LIMIT:
920 		nval = tcp_reass_limit;
921 		error = sysctl_int(oldp, oldlenp, newp, newlen, &nval);
922 		if (error)
923 			return (error);
924 		if (nval != tcp_reass_limit) {
925 			error = pool_sethardlimit(&tcpqe_pool, nval, NULL, 0);
926 			if (error)
927 				return (error);
928 			tcp_reass_limit = nval;
929 		}
930 		return (0);
931 #ifdef TCP_SACK
932 	case TCPCTL_SACKHOLE_LIMIT:
933 		nval = tcp_sackhole_limit;
934 		error = sysctl_int(oldp, oldlenp, newp, newlen, &nval);
935 		if (error)
936 			return (error);
937 		if (nval != tcp_sackhole_limit) {
938 			error = pool_sethardlimit(&sackhl_pool, nval, NULL, 0);
939 			if (error)
940 				return (error);
941 			tcp_sackhole_limit = nval;
942 		}
943 		return (0);
944 #endif
945 
946 	case TCPCTL_STATS:
947 		if (newp != NULL)
948 			return (EPERM);
949 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
950 		    &tcpstat, sizeof(tcpstat)));
951 
952 	default:
953 		if (name[0] < TCPCTL_MAXID)
954 			return (sysctl_int_arr(tcpctl_vars, name, namelen,
955 			    oldp, oldlenp, newp, newlen));
956 		return (ENOPROTOOPT);
957 	}
958 	/* NOTREACHED */
959 }
960