tcp_usrreq.c revision 28270
1/*
2 * Copyright (c) 1982, 1986, 1988, 1993
3 *	The Regents of the University of California.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 *    must display the following acknowledgement:
15 *	This product includes software developed by the University of
16 *	California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 *	From: @(#)tcp_usrreq.c	8.2 (Berkeley) 1/3/94
34 *	$Id: tcp_usrreq.c,v 1.32 1997/08/02 14:32:58 bde Exp $
35 */
36
37#include <sys/param.h>
38#include <sys/systm.h>
39#include <sys/kernel.h>
40#include <sys/sysctl.h>
41#include <sys/mbuf.h>
42#include <sys/socket.h>
43#include <sys/socketvar.h>
44#include <sys/protosw.h>
45
46#include <net/if.h>
47#include <net/route.h>
48
49#include <netinet/in.h>
50#include <netinet/in_systm.h>
51#include <netinet/in_pcb.h>
52#include <netinet/in_var.h>
53#include <netinet/ip_var.h>
54#include <netinet/tcp.h>
55#include <netinet/tcp_fsm.h>
56#include <netinet/tcp_seq.h>
57#include <netinet/tcp_timer.h>
58#include <netinet/tcp_var.h>
59#include <netinet/tcpip.h>
60#ifdef TCPDEBUG
61#include <netinet/tcp_debug.h>
62#endif
63
64/*
65 * TCP protocol interface to socket abstraction.
66 */
67extern	char *tcpstates[];	/* XXX ??? */
68
69static int	tcp_attach __P((struct socket *, struct proc *));
70static int	tcp_connect __P((struct tcpcb *, struct sockaddr *,
71				 struct proc *));
72static struct tcpcb *
73		tcp_disconnect __P((struct tcpcb *));
74static struct tcpcb *
75		tcp_usrclosed __P((struct tcpcb *));
76
77#ifdef TCPDEBUG
78#define	TCPDEBUG0	int ostate
79#define	TCPDEBUG1()	ostate = tp ? tp->t_state : 0
80#define	TCPDEBUG2(req)	if (tp && (so->so_options & SO_DEBUG)) \
81				tcp_trace(TA_USER, ostate, tp, 0, req)
82#else
83#define	TCPDEBUG0
84#define	TCPDEBUG1()
85#define	TCPDEBUG2(req)
86#endif
87
88/*
89 * TCP attaches to socket via pru_attach(), reserving space,
90 * and an internet control block.
91 */
92static int
93tcp_usr_attach(struct socket *so, int proto, struct proc *p)
94{
95	int s = splnet();
96	int error;
97	struct inpcb *inp = sotoinpcb(so);
98	struct tcpcb *tp = 0;
99	TCPDEBUG0;
100
101	TCPDEBUG1();
102	if (inp) {
103		error = EISCONN;
104		goto out;
105	}
106
107	error = tcp_attach(so, p);
108	if (error)
109		goto out;
110
111	if ((so->so_options & SO_LINGER) && so->so_linger == 0)
112		so->so_linger = TCP_LINGERTIME * hz;
113	tp = sototcpcb(so);
114out:
115	TCPDEBUG2(PRU_ATTACH);
116	splx(s);
117	return error;
118}
119
120/*
121 * pru_detach() detaches the TCP protocol from the socket.
122 * If the protocol state is non-embryonic, then can't
123 * do this directly: have to initiate a pru_disconnect(),
124 * which may finish later; embryonic TCB's can just
125 * be discarded here.
126 */
127static int
128tcp_usr_detach(struct socket *so)
129{
130	int s = splnet();
131	int error = 0;
132	struct inpcb *inp = sotoinpcb(so);
133	struct tcpcb *tp;
134	TCPDEBUG0;
135
136	if (inp == 0) {
137		splx(s);
138		return EINVAL;	/* XXX */
139	}
140	tp = intotcpcb(inp);
141	TCPDEBUG1();
142	tp = tcp_disconnect(tp);
143
144	TCPDEBUG2(PRU_DETACH);
145	splx(s);
146	return error;
147}
148
149#define	COMMON_START()	TCPDEBUG0; \
150			do { \
151				     if (inp == 0) { \
152					     splx(s); \
153					     return EINVAL; \
154				     } \
155				     tp = intotcpcb(inp); \
156				     TCPDEBUG1(); \
157		     } while(0)
158
159#define COMMON_END(req)	out: TCPDEBUG2(req); splx(s); return error; goto out
160
161
162/*
163 * Give the socket an address.
164 */
165static int
166tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
167{
168	int s = splnet();
169	int error = 0;
170	struct inpcb *inp = sotoinpcb(so);
171	struct tcpcb *tp;
172	struct sockaddr_in *sinp;
173
174	COMMON_START();
175
176	/*
177	 * Must check for multicast addresses and disallow binding
178	 * to them.
179	 */
180	sinp = (struct sockaddr_in *)nam;
181	if (sinp->sin_family == AF_INET &&
182	    IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
183		error = EAFNOSUPPORT;
184		goto out;
185	}
186	error = in_pcbbind(inp, nam, p);
187	if (error)
188		goto out;
189	COMMON_END(PRU_BIND);
190
191}
192
193/*
194 * Prepare to accept connections.
195 */
196static int
197tcp_usr_listen(struct socket *so, struct proc *p)
198{
199	int s = splnet();
200	int error = 0;
201	struct inpcb *inp = sotoinpcb(so);
202	struct tcpcb *tp;
203
204	COMMON_START();
205	if (inp->inp_lport == 0)
206		error = in_pcbbind(inp, (struct sockaddr *)0, p);
207	if (error == 0)
208		tp->t_state = TCPS_LISTEN;
209	COMMON_END(PRU_LISTEN);
210}
211
212/*
213 * Initiate connection to peer.
214 * Create a template for use in transmissions on this connection.
215 * Enter SYN_SENT state, and mark socket as connecting.
216 * Start keep-alive timer, and seed output sequence space.
217 * Send initial segment on connection.
218 */
219static int
220tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
221{
222	int s = splnet();
223	int error = 0;
224	struct inpcb *inp = sotoinpcb(so);
225	struct tcpcb *tp;
226	struct sockaddr_in *sinp;
227
228	COMMON_START();
229
230	/*
231	 * Must disallow TCP ``connections'' to multicast addresses.
232	 */
233	sinp = (struct sockaddr_in *)nam;
234	if (sinp->sin_family == AF_INET
235	    && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
236		error = EAFNOSUPPORT;
237		goto out;
238	}
239
240	if ((error = tcp_connect(tp, nam, p)) != 0)
241		goto out;
242	error = tcp_output(tp);
243	COMMON_END(PRU_CONNECT);
244}
245
246/*
247 * Initiate disconnect from peer.
248 * If connection never passed embryonic stage, just drop;
249 * else if don't need to let data drain, then can just drop anyways,
250 * else have to begin TCP shutdown process: mark socket disconnecting,
251 * drain unread data, state switch to reflect user close, and
252 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
253 * when peer sends FIN and acks ours.
254 *
255 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
256 */
257static int
258tcp_usr_disconnect(struct socket *so)
259{
260	int s = splnet();
261	int error = 0;
262	struct inpcb *inp = sotoinpcb(so);
263	struct tcpcb *tp;
264
265	COMMON_START();
266	tp = tcp_disconnect(tp);
267	COMMON_END(PRU_DISCONNECT);
268}
269
270/*
271 * Accept a connection.  Essentially all the work is
272 * done at higher levels; just return the address
273 * of the peer, storing through addr.
274 */
275static int
276tcp_usr_accept(struct socket *so, struct sockaddr **nam)
277{
278	int s = splnet();
279	int error = 0;
280	struct inpcb *inp = sotoinpcb(so);
281	struct tcpcb *tp;
282
283	COMMON_START();
284	in_setpeeraddr(so, nam);
285	COMMON_END(PRU_ACCEPT);
286}
287
288/*
289 * Mark the connection as being incapable of further output.
290 */
291static int
292tcp_usr_shutdown(struct socket *so)
293{
294	int s = splnet();
295	int error = 0;
296	struct inpcb *inp = sotoinpcb(so);
297	struct tcpcb *tp;
298
299	COMMON_START();
300	socantsendmore(so);
301	tp = tcp_usrclosed(tp);
302	if (tp)
303		error = tcp_output(tp);
304	COMMON_END(PRU_SHUTDOWN);
305}
306
307/*
308 * After a receive, possibly send window update to peer.
309 */
310static int
311tcp_usr_rcvd(struct socket *so, int flags)
312{
313	int s = splnet();
314	int error = 0;
315	struct inpcb *inp = sotoinpcb(so);
316	struct tcpcb *tp;
317
318	COMMON_START();
319	tcp_output(tp);
320	COMMON_END(PRU_RCVD);
321}
322
323/*
324 * Do a send by putting data in output queue and updating urgent
325 * marker if URG set.  Possibly send more data.
326 */
327static int
328tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
329	     struct sockaddr *nam, struct mbuf *control, struct proc *p)
330{
331	int s = splnet();
332	int error = 0;
333	struct inpcb *inp = sotoinpcb(so);
334	struct tcpcb *tp;
335
336	COMMON_START();
337	if (control && control->m_len) {
338		m_freem(control); /* XXX shouldn't caller do this??? */
339		if (m)
340			m_freem(m);
341		return EINVAL;
342	}
343
344	if(!(flags & PRUS_OOB)) {
345		sbappend(&so->so_snd, m);
346		if (nam && tp->t_state < TCPS_SYN_SENT) {
347			/*
348			 * Do implied connect if not yet connected,
349			 * initialize window to default value, and
350			 * initialize maxseg/maxopd using peer's cached
351			 * MSS.
352			 */
353			error = tcp_connect(tp, nam, p);
354			if (error)
355				goto out;
356			tp->snd_wnd = TTCP_CLIENT_SND_WND;
357			tcp_mss(tp, -1);
358		}
359
360		if (flags & PRUS_EOF) {
361			/*
362			 * Close the send side of the connection after
363			 * the data is sent.
364			 */
365			socantsendmore(so);
366			tp = tcp_usrclosed(tp);
367		}
368		if (tp != NULL)
369			error = tcp_output(tp);
370	} else {
371		if (sbspace(&so->so_snd) < -512) {
372			m_freem(m);
373			error = ENOBUFS;
374			goto out;
375		}
376		/*
377		 * According to RFC961 (Assigned Protocols),
378		 * the urgent pointer points to the last octet
379		 * of urgent data.  We continue, however,
380		 * to consider it to indicate the first octet
381		 * of data past the urgent section.
382		 * Otherwise, snd_up should be one lower.
383		 */
384		sbappend(&so->so_snd, m);
385		if (nam && tp->t_state < TCPS_SYN_SENT) {
386			/*
387			 * Do implied connect if not yet connected,
388			 * initialize window to default value, and
389			 * initialize maxseg/maxopd using peer's cached
390			 * MSS.
391			 */
392			error = tcp_connect(tp, nam, p);
393			if (error)
394				goto out;
395			tp->snd_wnd = TTCP_CLIENT_SND_WND;
396			tcp_mss(tp, -1);
397		}
398		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
399		tp->t_force = 1;
400		error = tcp_output(tp);
401		tp->t_force = 0;
402	}
403	COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
404		   ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
405}
406
407/*
408 * Abort the TCP.
409 */
410static int
411tcp_usr_abort(struct socket *so)
412{
413	int s = splnet();
414	int error = 0;
415	struct inpcb *inp = sotoinpcb(so);
416	struct tcpcb *tp;
417
418	COMMON_START();
419	tp = tcp_drop(tp, ECONNABORTED);
420	COMMON_END(PRU_ABORT);
421}
422
423/*
424 * Receive out-of-band data.
425 */
426static int
427tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
428{
429	int s = splnet();
430	int error = 0;
431	struct inpcb *inp = sotoinpcb(so);
432	struct tcpcb *tp;
433
434	COMMON_START();
435	if ((so->so_oobmark == 0 &&
436	     (so->so_state & SS_RCVATMARK) == 0) ||
437	    so->so_options & SO_OOBINLINE ||
438	    tp->t_oobflags & TCPOOB_HADDATA) {
439		error = EINVAL;
440		goto out;
441	}
442	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
443		error = EWOULDBLOCK;
444		goto out;
445	}
446	m->m_len = 1;
447	*mtod(m, caddr_t) = tp->t_iobc;
448	if ((flags & MSG_PEEK) == 0)
449		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
450	COMMON_END(PRU_RCVOOB);
451}
452
453/* xxx - should be const */
454struct pr_usrreqs tcp_usrreqs = {
455	tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind,
456	tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach,
457	tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd,
458	tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
459	in_setsockaddr, sosend, soreceive, soselect
460};
461
462/*
463 * Common subroutine to open a TCP connection to remote host specified
464 * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
465 * port number if needed.  Call in_pcbladdr to do the routing and to choose
466 * a local host address (interface).  If there is an existing incarnation
467 * of the same connection in TIME-WAIT state and if the remote host was
468 * sending CC options and if the connection duration was < MSL, then
469 * truncate the previous TIME-WAIT state and proceed.
470 * Initialize connection parameters and enter SYN-SENT state.
471 */
472static int
473tcp_connect(tp, nam, p)
474	register struct tcpcb *tp;
475	struct sockaddr *nam;
476	struct proc *p;
477{
478	struct inpcb *inp = tp->t_inpcb, *oinp;
479	struct socket *so = inp->inp_socket;
480	struct tcpcb *otp;
481	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
482	struct sockaddr_in *ifaddr;
483	int error;
484	struct rmxp_tao *taop;
485	struct rmxp_tao tao_noncached;
486
487	if (inp->inp_lport == 0) {
488		error = in_pcbbind(inp, (struct sockaddr *)0, p);
489		if (error)
490			return error;
491	}
492
493	/*
494	 * Cannot simply call in_pcbconnect, because there might be an
495	 * earlier incarnation of this same connection still in
496	 * TIME_WAIT state, creating an ADDRINUSE error.
497	 */
498	error = in_pcbladdr(inp, nam, &ifaddr);
499	if (error)
500		return error;
501	oinp = in_pcblookuphash(inp->inp_pcbinfo,
502	    sin->sin_addr, sin->sin_port,
503	    inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr
504						: ifaddr->sin_addr,
505	    inp->inp_lport,  0);
506	if (oinp) {
507		if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
508		otp->t_state == TCPS_TIME_WAIT &&
509		    otp->t_duration < TCPTV_MSL &&
510		    (otp->t_flags & TF_RCVD_CC))
511			otp = tcp_close(otp);
512		else
513			return EADDRINUSE;
514	}
515	if (inp->inp_laddr.s_addr == INADDR_ANY)
516		inp->inp_laddr = ifaddr->sin_addr;
517	inp->inp_faddr = sin->sin_addr;
518	inp->inp_fport = sin->sin_port;
519	in_pcbrehash(inp);
520
521	tp->t_template = tcp_template(tp);
522	if (tp->t_template == 0) {
523		in_pcbdisconnect(inp);
524		return ENOBUFS;
525	}
526
527	/* Compute window scaling to request.  */
528	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
529	    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
530		tp->request_r_scale++;
531
532	soisconnecting(so);
533	tcpstat.tcps_connattempt++;
534	tp->t_state = TCPS_SYN_SENT;
535	tp->t_timer[TCPT_KEEP] = tcp_keepinit;
536	tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
537	tcp_sendseqinit(tp);
538
539	/*
540	 * Generate a CC value for this connection and
541	 * check whether CC or CCnew should be used.
542	 */
543	if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) {
544		taop = &tao_noncached;
545		bzero(taop, sizeof(*taop));
546	}
547
548	tp->cc_send = CC_INC(tcp_ccgen);
549	if (taop->tao_ccsent != 0 &&
550	    CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
551		taop->tao_ccsent = tp->cc_send;
552	} else {
553		taop->tao_ccsent = 0;
554		tp->t_flags |= TF_SENDCCNEW;
555	}
556
557	return 0;
558}
559
560int
561tcp_ctloutput(op, so, level, optname, mp, p)
562	int op;
563	struct socket *so;
564	int level, optname;
565	struct mbuf **mp;
566	struct proc *p;
567{
568	int error = 0, s;
569	struct inpcb *inp;
570	register struct tcpcb *tp;
571	register struct mbuf *m;
572	register int i;
573
574	s = splnet();
575	inp = sotoinpcb(so);
576	if (inp == NULL) {
577		splx(s);
578		if (op == PRCO_SETOPT && *mp)
579			(void) m_free(*mp);
580		return (ECONNRESET);
581	}
582	if (level != IPPROTO_TCP) {
583		error = ip_ctloutput(op, so, level, optname, mp, p);
584		splx(s);
585		return (error);
586	}
587	tp = intotcpcb(inp);
588
589	switch (op) {
590
591	case PRCO_SETOPT:
592		m = *mp;
593		switch (optname) {
594
595		case TCP_NODELAY:
596			if (m == NULL || m->m_len < sizeof (int))
597				error = EINVAL;
598			else if (*mtod(m, int *))
599				tp->t_flags |= TF_NODELAY;
600			else
601				tp->t_flags &= ~TF_NODELAY;
602			break;
603
604		case TCP_MAXSEG:
605			if (m && (i = *mtod(m, int *)) > 0 && i <= tp->t_maxseg)
606				tp->t_maxseg = i;
607			else
608				error = EINVAL;
609			break;
610
611		case TCP_NOOPT:
612			if (m == NULL || m->m_len < sizeof (int))
613				error = EINVAL;
614			else if (*mtod(m, int *))
615				tp->t_flags |= TF_NOOPT;
616			else
617				tp->t_flags &= ~TF_NOOPT;
618			break;
619
620		case TCP_NOPUSH:
621			if (m == NULL || m->m_len < sizeof (int))
622				error = EINVAL;
623			else if (*mtod(m, int *))
624				tp->t_flags |= TF_NOPUSH;
625			else
626				tp->t_flags &= ~TF_NOPUSH;
627			break;
628
629		default:
630			error = ENOPROTOOPT;
631			break;
632		}
633		if (m)
634			(void) m_free(m);
635		break;
636
637	case PRCO_GETOPT:
638		*mp = m = m_get(M_WAIT, MT_SOOPTS);
639		m->m_len = sizeof(int);
640
641		switch (optname) {
642		case TCP_NODELAY:
643			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
644			break;
645		case TCP_MAXSEG:
646			*mtod(m, int *) = tp->t_maxseg;
647			break;
648		case TCP_NOOPT:
649			*mtod(m, int *) = tp->t_flags & TF_NOOPT;
650			break;
651		case TCP_NOPUSH:
652			*mtod(m, int *) = tp->t_flags & TF_NOPUSH;
653			break;
654		default:
655			error = ENOPROTOOPT;
656			break;
657		}
658		break;
659	}
660	splx(s);
661	return (error);
662}
663
664/*
665 * tcp_sendspace and tcp_recvspace are the default send and receive window
666 * sizes, respectively.  These are obsolescent (this information should
667 * be set by the route).
668 */
669u_long	tcp_sendspace = 1024*16;
670SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace,
671	CTLFLAG_RW, &tcp_sendspace , 0, "");
672u_long	tcp_recvspace = 1024*16;
673SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace,
674	CTLFLAG_RW, &tcp_recvspace , 0, "");
675
676/*
677 * Attach TCP protocol to socket, allocating
678 * internet protocol control block, tcp control block,
679 * bufer space, and entering LISTEN state if to accept connections.
680 */
681static int
682tcp_attach(so, p)
683	struct socket *so;
684	struct proc *p;
685{
686	register struct tcpcb *tp;
687	struct inpcb *inp;
688	int error;
689
690	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
691		error = soreserve(so, tcp_sendspace, tcp_recvspace);
692		if (error)
693			return (error);
694	}
695	error = in_pcballoc(so, &tcbinfo, p);
696	if (error)
697		return (error);
698	inp = sotoinpcb(so);
699	tp = tcp_newtcpcb(inp);
700	if (tp == 0) {
701		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
702
703		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
704		in_pcbdetach(inp);
705		so->so_state |= nofd;
706		return (ENOBUFS);
707	}
708	tp->t_state = TCPS_CLOSED;
709	return (0);
710}
711
712/*
713 * Initiate (or continue) disconnect.
714 * If embryonic state, just send reset (once).
715 * If in ``let data drain'' option and linger null, just drop.
716 * Otherwise (hard), mark socket disconnecting and drop
717 * current input data; switch states based on user close, and
718 * send segment to peer (with FIN).
719 */
720static struct tcpcb *
721tcp_disconnect(tp)
722	register struct tcpcb *tp;
723{
724	struct socket *so = tp->t_inpcb->inp_socket;
725
726	if (tp->t_state < TCPS_ESTABLISHED)
727		tp = tcp_close(tp);
728	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
729		tp = tcp_drop(tp, 0);
730	else {
731		soisdisconnecting(so);
732		sbflush(&so->so_rcv);
733		tp = tcp_usrclosed(tp);
734		if (tp)
735			(void) tcp_output(tp);
736	}
737	return (tp);
738}
739
740/*
741 * User issued close, and wish to trail through shutdown states:
742 * if never received SYN, just forget it.  If got a SYN from peer,
743 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
744 * If already got a FIN from peer, then almost done; go to LAST_ACK
745 * state.  In all other cases, have already sent FIN to peer (e.g.
746 * after PRU_SHUTDOWN), and just have to play tedious game waiting
747 * for peer to send FIN or not respond to keep-alives, etc.
748 * We can let the user exit from the close as soon as the FIN is acked.
749 */
750static struct tcpcb *
751tcp_usrclosed(tp)
752	register struct tcpcb *tp;
753{
754
755	switch (tp->t_state) {
756
757	case TCPS_CLOSED:
758	case TCPS_LISTEN:
759		tp->t_state = TCPS_CLOSED;
760		tp = tcp_close(tp);
761		break;
762
763	case TCPS_SYN_SENT:
764	case TCPS_SYN_RECEIVED:
765		tp->t_flags |= TF_NEEDFIN;
766		break;
767
768	case TCPS_ESTABLISHED:
769		tp->t_state = TCPS_FIN_WAIT_1;
770		break;
771
772	case TCPS_CLOSE_WAIT:
773		tp->t_state = TCPS_LAST_ACK;
774		break;
775	}
776	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
777		soisdisconnected(tp->t_inpcb->inp_socket);
778		/* To prevent the connection hanging in FIN_WAIT_2 forever. */
779		if (tp->t_state == TCPS_FIN_WAIT_2)
780			tp->t_timer[TCPT_2MSL] = tcp_maxidle;
781	}
782	return (tp);
783}
784
785