tcp_usrreq.c revision 25201
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 1.30 1997/02/21 16:30:31 wollman Exp $
35 */
36
37#include <sys/param.h>
38#include <sys/queue.h>
39#include <sys/systm.h>
40#include <sys/kernel.h>
41#include <sys/sysctl.h>
42#include <sys/malloc.h>
43#include <sys/mbuf.h>
44#include <sys/socket.h>
45#include <sys/socketvar.h>
46#include <sys/protosw.h>
47#include <sys/errno.h>
48#include <sys/stat.h>
49
50#include <net/if.h>
51#include <net/route.h>
52
53#include <netinet/in.h>
54#include <netinet/in_systm.h>
55#include <netinet/ip.h>
56#include <netinet/in_pcb.h>
57#include <netinet/in_var.h>
58#include <netinet/ip_var.h>
59#include <netinet/tcp.h>
60#include <netinet/tcp_fsm.h>
61#include <netinet/tcp_seq.h>
62#include <netinet/tcp_timer.h>
63#include <netinet/tcp_var.h>
64#include <netinet/tcpip.h>
65#ifdef TCPDEBUG
66#include <netinet/tcp_debug.h>
67#endif
68
69/*
70 * TCP protocol interface to socket abstraction.
71 */
72extern	char *tcpstates[];	/* XXX ??? */
73
74static int	tcp_attach __P((struct socket *, struct proc *));
75static int	tcp_connect __P((struct tcpcb *, struct mbuf *,
76				 struct proc *));
77static struct tcpcb *
78		tcp_disconnect __P((struct tcpcb *));
79static struct tcpcb *
80		tcp_usrclosed __P((struct tcpcb *));
81
82#ifdef TCPDEBUG
83#define	TCPDEBUG0	int ostate
84#define	TCPDEBUG1()	ostate = tp ? tp->t_state : 0
85#define	TCPDEBUG2(req)	if (tp && (so->so_options & SO_DEBUG)) \
86				tcp_trace(TA_USER, ostate, tp, 0, req)
87#else
88#define	TCPDEBUG0
89#define	TCPDEBUG1()
90#define	TCPDEBUG2(req)
91#endif
92
93/*
94 * TCP attaches to socket via pru_attach(), reserving space,
95 * and an internet control block.
96 */
97static int
98tcp_usr_attach(struct socket *so, int proto, struct proc *p)
99{
100	int s = splnet();
101	int error;
102	struct inpcb *inp = sotoinpcb(so);
103	struct tcpcb *tp = 0;
104	TCPDEBUG0;
105
106	TCPDEBUG1();
107	if (inp) {
108		error = EISCONN;
109		goto out;
110	}
111
112	error = tcp_attach(so, p);
113	if (error)
114		goto out;
115
116	if ((so->so_options & SO_LINGER) && so->so_linger == 0)
117		so->so_linger = TCP_LINGERTIME * hz;
118	tp = sototcpcb(so);
119out:
120	TCPDEBUG2(PRU_ATTACH);
121	splx(s);
122	return error;
123}
124
125/*
126 * pru_detach() detaches the TCP protocol from the socket.
127 * If the protocol state is non-embryonic, then can't
128 * do this directly: have to initiate a pru_disconnect(),
129 * which may finish later; embryonic TCB's can just
130 * be discarded here.
131 */
132static int
133tcp_usr_detach(struct socket *so)
134{
135	int s = splnet();
136	int error = 0;
137	struct inpcb *inp = sotoinpcb(so);
138	struct tcpcb *tp;
139	TCPDEBUG0;
140
141	if (inp == 0) {
142		splx(s);
143		return EINVAL;	/* XXX */
144	}
145	tp = intotcpcb(inp);
146	TCPDEBUG1();
147	if (tp->t_state > TCPS_LISTEN)
148		tp = tcp_disconnect(tp);
149	else
150		tp = tcp_close(tp);
151
152	TCPDEBUG2(PRU_DETACH);
153	splx(s);
154	return error;
155}
156
157#define	COMMON_START()	TCPDEBUG0; \
158			do { \
159				     if (inp == 0) { \
160					     splx(s); \
161					     return EINVAL; \
162				     } \
163				     tp = intotcpcb(inp); \
164				     TCPDEBUG1(); \
165		     } while(0)
166
167#define COMMON_END(req)	out: TCPDEBUG2(req); splx(s); return error; goto out
168
169
170/*
171 * Give the socket an address.
172 */
173static int
174tcp_usr_bind(struct socket *so, struct mbuf *nam, struct proc *p)
175{
176	int s = splnet();
177	int error = 0;
178	struct inpcb *inp = sotoinpcb(so);
179	struct tcpcb *tp;
180	struct sockaddr_in *sinp;
181
182	COMMON_START();
183
184	/*
185	 * Must check for multicast addresses and disallow binding
186	 * to them.
187	 */
188	sinp = mtod(nam, struct sockaddr_in *);
189	if (sinp->sin_family == AF_INET &&
190	    IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
191		error = EAFNOSUPPORT;
192		goto out;
193	}
194	error = in_pcbbind(inp, nam, p);
195	if (error)
196		goto out;
197	COMMON_END(PRU_BIND);
198
199}
200
201/*
202 * Prepare to accept connections.
203 */
204static int
205tcp_usr_listen(struct socket *so, struct proc *p)
206{
207	int s = splnet();
208	int error = 0;
209	struct inpcb *inp = sotoinpcb(so);
210	struct tcpcb *tp;
211
212	COMMON_START();
213	if (inp->inp_lport == 0)
214		error = in_pcbbind(inp, (struct mbuf *)0, p);
215	if (error == 0)
216		tp->t_state = TCPS_LISTEN;
217	COMMON_END(PRU_LISTEN);
218}
219
220/*
221 * Initiate connection to peer.
222 * Create a template for use in transmissions on this connection.
223 * Enter SYN_SENT state, and mark socket as connecting.
224 * Start keep-alive timer, and seed output sequence space.
225 * Send initial segment on connection.
226 */
227static int
228tcp_usr_connect(struct socket *so, struct mbuf *nam, struct proc *p)
229{
230	int s = splnet();
231	int error = 0;
232	struct inpcb *inp = sotoinpcb(so);
233	struct tcpcb *tp;
234	struct sockaddr_in *sinp;
235
236	COMMON_START();
237
238	/*
239	 * Must disallow TCP ``connections'' to multicast addresses.
240	 */
241	sinp = mtod(nam, struct sockaddr_in *);
242	if (sinp->sin_family == AF_INET
243	    && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
244		error = EAFNOSUPPORT;
245		goto out;
246	}
247
248	if ((error = tcp_connect(tp, nam, p)) != 0)
249		goto out;
250	error = tcp_output(tp);
251	COMMON_END(PRU_CONNECT);
252}
253
254/*
255 * Initiate disconnect from peer.
256 * If connection never passed embryonic stage, just drop;
257 * else if don't need to let data drain, then can just drop anyways,
258 * else have to begin TCP shutdown process: mark socket disconnecting,
259 * drain unread data, state switch to reflect user close, and
260 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
261 * when peer sends FIN and acks ours.
262 *
263 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
264 */
265static int
266tcp_usr_disconnect(struct socket *so)
267{
268	int s = splnet();
269	int error = 0;
270	struct inpcb *inp = sotoinpcb(so);
271	struct tcpcb *tp;
272
273	COMMON_START();
274	tp = tcp_disconnect(tp);
275	COMMON_END(PRU_DISCONNECT);
276}
277
278/*
279 * Accept a connection.  Essentially all the work is
280 * done at higher levels; just return the address
281 * of the peer, storing through addr.
282 */
283static int
284tcp_usr_accept(struct socket *so, struct mbuf *nam)
285{
286	int s = splnet();
287	int error = 0;
288	struct inpcb *inp = sotoinpcb(so);
289	struct tcpcb *tp;
290
291	COMMON_START();
292	in_setpeeraddr(so, nam);
293	COMMON_END(PRU_ACCEPT);
294}
295
296/*
297 * Mark the connection as being incapable of further output.
298 */
299static int
300tcp_usr_shutdown(struct socket *so)
301{
302	int s = splnet();
303	int error = 0;
304	struct inpcb *inp = sotoinpcb(so);
305	struct tcpcb *tp;
306
307	COMMON_START();
308	socantsendmore(so);
309	tp = tcp_usrclosed(tp);
310	if (tp)
311		error = tcp_output(tp);
312	COMMON_END(PRU_SHUTDOWN);
313}
314
315/*
316 * After a receive, possibly send window update to peer.
317 */
318static int
319tcp_usr_rcvd(struct socket *so, int flags)
320{
321	int s = splnet();
322	int error = 0;
323	struct inpcb *inp = sotoinpcb(so);
324	struct tcpcb *tp;
325
326	COMMON_START();
327	tcp_output(tp);
328	COMMON_END(PRU_RCVD);
329}
330
331/*
332 * Do a send by putting data in output queue and updating urgent
333 * marker if URG set.  Possibly send more data.
334 */
335static int
336tcp_usr_send(struct socket *so, int flags, struct mbuf *m, struct mbuf *nam,
337	     struct mbuf *control, struct proc *p)
338{
339	int s = splnet();
340	int error = 0;
341	struct inpcb *inp = sotoinpcb(so);
342	struct tcpcb *tp;
343
344	COMMON_START();
345	if (control && control->m_len) {
346		m_freem(control); /* XXX shouldn't caller do this??? */
347		if (m)
348			m_freem(m);
349		return EINVAL;
350	}
351
352	if(!(flags & PRUS_OOB)) {
353		sbappend(&so->so_snd, m);
354		if (nam && tp->t_state < TCPS_SYN_SENT) {
355			/*
356			 * Do implied connect if not yet connected,
357			 * initialize window to default value, and
358			 * initialize maxseg/maxopd using peer's cached
359			 * MSS.
360			 */
361			error = tcp_connect(tp, nam, p);
362			if (error)
363				goto out;
364			tp->snd_wnd = TTCP_CLIENT_SND_WND;
365			tcp_mss(tp, -1);
366		}
367
368		if (flags & PRUS_EOF) {
369			/*
370			 * Close the send side of the connection after
371			 * the data is sent.
372			 */
373			socantsendmore(so);
374			tp = tcp_usrclosed(tp);
375		}
376		if (tp != NULL)
377			error = tcp_output(tp);
378	} else {
379		if (sbspace(&so->so_snd) < -512) {
380			m_freem(m);
381			error = ENOBUFS;
382			goto out;
383		}
384		/*
385		 * According to RFC961 (Assigned Protocols),
386		 * the urgent pointer points to the last octet
387		 * of urgent data.  We continue, however,
388		 * to consider it to indicate the first octet
389		 * of data past the urgent section.
390		 * Otherwise, snd_up should be one lower.
391		 */
392		sbappend(&so->so_snd, m);
393		if (nam && tp->t_state < TCPS_SYN_SENT) {
394			/*
395			 * Do implied connect if not yet connected,
396			 * initialize window to default value, and
397			 * initialize maxseg/maxopd using peer's cached
398			 * MSS.
399			 */
400			error = tcp_connect(tp, nam, p);
401			if (error)
402				goto out;
403			tp->snd_wnd = TTCP_CLIENT_SND_WND;
404			tcp_mss(tp, -1);
405		}
406		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
407		tp->t_force = 1;
408		error = tcp_output(tp);
409		tp->t_force = 0;
410	}
411	COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
412		   ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
413}
414
415/*
416 * Abort the TCP.
417 */
418static int
419tcp_usr_abort(struct socket *so)
420{
421	int s = splnet();
422	int error = 0;
423	struct inpcb *inp = sotoinpcb(so);
424	struct tcpcb *tp;
425
426	COMMON_START();
427	tp = tcp_drop(tp, ECONNABORTED);
428	COMMON_END(PRU_ABORT);
429}
430
431/*
432 * Receive out-of-band data.
433 */
434static int
435tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
436{
437	int s = splnet();
438	int error = 0;
439	struct inpcb *inp = sotoinpcb(so);
440	struct tcpcb *tp;
441
442	COMMON_START();
443	if ((so->so_oobmark == 0 &&
444	     (so->so_state & SS_RCVATMARK) == 0) ||
445	    so->so_options & SO_OOBINLINE ||
446	    tp->t_oobflags & TCPOOB_HADDATA) {
447		error = EINVAL;
448		goto out;
449	}
450	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
451		error = EWOULDBLOCK;
452		goto out;
453	}
454	m->m_len = 1;
455	*mtod(m, caddr_t) = tp->t_iobc;
456	if ((flags & MSG_PEEK) == 0)
457		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
458	COMMON_END(PRU_RCVOOB);
459}
460
461/* xxx - should be const */
462struct pr_usrreqs tcp_usrreqs = {
463	tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind,
464	tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach,
465	tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd,
466	tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
467	in_setsockaddr, sosend, soreceive, soselect
468};
469
470/*
471 * Common subroutine to open a TCP connection to remote host specified
472 * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
473 * port number if needed.  Call in_pcbladdr to do the routing and to choose
474 * a local host address (interface).  If there is an existing incarnation
475 * of the same connection in TIME-WAIT state and if the remote host was
476 * sending CC options and if the connection duration was < MSL, then
477 * truncate the previous TIME-WAIT state and proceed.
478 * Initialize connection parameters and enter SYN-SENT state.
479 */
480static int
481tcp_connect(tp, nam, p)
482	register struct tcpcb *tp;
483	struct mbuf *nam;
484	struct proc *p;
485{
486	struct inpcb *inp = tp->t_inpcb, *oinp;
487	struct socket *so = inp->inp_socket;
488	struct tcpcb *otp;
489	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
490	struct sockaddr_in *ifaddr;
491	int error;
492	struct rmxp_tao *taop;
493	struct rmxp_tao tao_noncached;
494
495	if (inp->inp_lport == 0) {
496		error = in_pcbbind(inp, (struct mbuf *)0, p);
497		if (error)
498			return error;
499	}
500
501	/*
502	 * Cannot simply call in_pcbconnect, because there might be an
503	 * earlier incarnation of this same connection still in
504	 * TIME_WAIT state, creating an ADDRINUSE error.
505	 */
506	error = in_pcbladdr(inp, nam, &ifaddr);
507	if (error)
508		return error;
509	oinp = in_pcblookuphash(inp->inp_pcbinfo,
510	    sin->sin_addr, sin->sin_port,
511	    inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr
512						: ifaddr->sin_addr,
513	    inp->inp_lport,  0);
514	if (oinp) {
515		if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
516		otp->t_state == TCPS_TIME_WAIT &&
517		    otp->t_duration < TCPTV_MSL &&
518		    (otp->t_flags & TF_RCVD_CC))
519			otp = tcp_close(otp);
520		else
521			return EADDRINUSE;
522	}
523	if (inp->inp_laddr.s_addr == INADDR_ANY)
524		inp->inp_laddr = ifaddr->sin_addr;
525	inp->inp_faddr = sin->sin_addr;
526	inp->inp_fport = sin->sin_port;
527	in_pcbrehash(inp);
528
529	tp->t_template = tcp_template(tp);
530	if (tp->t_template == 0) {
531		in_pcbdisconnect(inp);
532		return ENOBUFS;
533	}
534
535	/* Compute window scaling to request.  */
536	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
537	    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
538		tp->request_r_scale++;
539
540	soisconnecting(so);
541	tcpstat.tcps_connattempt++;
542	tp->t_state = TCPS_SYN_SENT;
543	tp->t_timer[TCPT_KEEP] = tcp_keepinit;
544	tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
545	tcp_sendseqinit(tp);
546
547	/*
548	 * Generate a CC value for this connection and
549	 * check whether CC or CCnew should be used.
550	 */
551	if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) {
552		taop = &tao_noncached;
553		bzero(taop, sizeof(*taop));
554	}
555
556	tp->cc_send = CC_INC(tcp_ccgen);
557	if (taop->tao_ccsent != 0 &&
558	    CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
559		taop->tao_ccsent = tp->cc_send;
560	} else {
561		taop->tao_ccsent = 0;
562		tp->t_flags |= TF_SENDCCNEW;
563	}
564
565	return 0;
566}
567
568int
569tcp_ctloutput(op, so, level, optname, mp, p)
570	int op;
571	struct socket *so;
572	int level, optname;
573	struct mbuf **mp;
574	struct proc *p;
575{
576	int error = 0, s;
577	struct inpcb *inp;
578	register struct tcpcb *tp;
579	register struct mbuf *m;
580	register int i;
581
582	s = splnet();
583	inp = sotoinpcb(so);
584	if (inp == NULL) {
585		splx(s);
586		if (op == PRCO_SETOPT && *mp)
587			(void) m_free(*mp);
588		return (ECONNRESET);
589	}
590	if (level != IPPROTO_TCP) {
591		error = ip_ctloutput(op, so, level, optname, mp, p);
592		splx(s);
593		return (error);
594	}
595	tp = intotcpcb(inp);
596
597	switch (op) {
598
599	case PRCO_SETOPT:
600		m = *mp;
601		switch (optname) {
602
603		case TCP_NODELAY:
604			if (m == NULL || m->m_len < sizeof (int))
605				error = EINVAL;
606			else if (*mtod(m, int *))
607				tp->t_flags |= TF_NODELAY;
608			else
609				tp->t_flags &= ~TF_NODELAY;
610			break;
611
612		case TCP_MAXSEG:
613			if (m && (i = *mtod(m, int *)) > 0 && i <= tp->t_maxseg)
614				tp->t_maxseg = i;
615			else
616				error = EINVAL;
617			break;
618
619		case TCP_NOOPT:
620			if (m == NULL || m->m_len < sizeof (int))
621				error = EINVAL;
622			else if (*mtod(m, int *))
623				tp->t_flags |= TF_NOOPT;
624			else
625				tp->t_flags &= ~TF_NOOPT;
626			break;
627
628		case TCP_NOPUSH:
629			if (m == NULL || m->m_len < sizeof (int))
630				error = EINVAL;
631			else if (*mtod(m, int *))
632				tp->t_flags |= TF_NOPUSH;
633			else
634				tp->t_flags &= ~TF_NOPUSH;
635			break;
636
637		default:
638			error = ENOPROTOOPT;
639			break;
640		}
641		if (m)
642			(void) m_free(m);
643		break;
644
645	case PRCO_GETOPT:
646		*mp = m = m_get(M_WAIT, MT_SOOPTS);
647		m->m_len = sizeof(int);
648
649		switch (optname) {
650		case TCP_NODELAY:
651			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
652			break;
653		case TCP_MAXSEG:
654			*mtod(m, int *) = tp->t_maxseg;
655			break;
656		case TCP_NOOPT:
657			*mtod(m, int *) = tp->t_flags & TF_NOOPT;
658			break;
659		case TCP_NOPUSH:
660			*mtod(m, int *) = tp->t_flags & TF_NOPUSH;
661			break;
662		default:
663			error = ENOPROTOOPT;
664			break;
665		}
666		break;
667	}
668	splx(s);
669	return (error);
670}
671
672/*
673 * tcp_sendspace and tcp_recvspace are the default send and receive window
674 * sizes, respectively.  These are obsolescent (this information should
675 * be set by the route).
676 */
677u_long	tcp_sendspace = 1024*16;
678SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace,
679	CTLFLAG_RW, &tcp_sendspace , 0, "");
680u_long	tcp_recvspace = 1024*16;
681SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace,
682	CTLFLAG_RW, &tcp_recvspace , 0, "");
683
684/*
685 * Attach TCP protocol to socket, allocating
686 * internet protocol control block, tcp control block,
687 * bufer space, and entering LISTEN state if to accept connections.
688 */
689static int
690tcp_attach(so, p)
691	struct socket *so;
692	struct proc *p;
693{
694	register struct tcpcb *tp;
695	struct inpcb *inp;
696	int error;
697
698	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
699		error = soreserve(so, tcp_sendspace, tcp_recvspace);
700		if (error)
701			return (error);
702	}
703	error = in_pcballoc(so, &tcbinfo, p);
704	if (error)
705		return (error);
706	inp = sotoinpcb(so);
707	tp = tcp_newtcpcb(inp);
708	if (tp == 0) {
709		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
710
711		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
712		in_pcbdetach(inp);
713		so->so_state |= nofd;
714		return (ENOBUFS);
715	}
716	tp->t_state = TCPS_CLOSED;
717	return (0);
718}
719
720/*
721 * Initiate (or continue) disconnect.
722 * If embryonic state, just send reset (once).
723 * If in ``let data drain'' option and linger null, just drop.
724 * Otherwise (hard), mark socket disconnecting and drop
725 * current input data; switch states based on user close, and
726 * send segment to peer (with FIN).
727 */
728static struct tcpcb *
729tcp_disconnect(tp)
730	register struct tcpcb *tp;
731{
732	struct socket *so = tp->t_inpcb->inp_socket;
733
734	if (tp->t_state < TCPS_ESTABLISHED)
735		tp = tcp_close(tp);
736	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
737		tp = tcp_drop(tp, 0);
738	else {
739		soisdisconnecting(so);
740		sbflush(&so->so_rcv);
741		tp = tcp_usrclosed(tp);
742		if (tp)
743			(void) tcp_output(tp);
744	}
745	return (tp);
746}
747
748/*
749 * User issued close, and wish to trail through shutdown states:
750 * if never received SYN, just forget it.  If got a SYN from peer,
751 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
752 * If already got a FIN from peer, then almost done; go to LAST_ACK
753 * state.  In all other cases, have already sent FIN to peer (e.g.
754 * after PRU_SHUTDOWN), and just have to play tedious game waiting
755 * for peer to send FIN or not respond to keep-alives, etc.
756 * We can let the user exit from the close as soon as the FIN is acked.
757 */
758static struct tcpcb *
759tcp_usrclosed(tp)
760	register struct tcpcb *tp;
761{
762
763	switch (tp->t_state) {
764
765	case TCPS_CLOSED:
766	case TCPS_LISTEN:
767		tp->t_state = TCPS_CLOSED;
768		tp = tcp_close(tp);
769		break;
770
771	case TCPS_SYN_SENT:
772	case TCPS_SYN_RECEIVED:
773		tp->t_flags |= TF_NEEDFIN;
774		break;
775
776	case TCPS_ESTABLISHED:
777		tp->t_state = TCPS_FIN_WAIT_1;
778		break;
779
780	case TCPS_CLOSE_WAIT:
781		tp->t_state = TCPS_LAST_ACK;
782		break;
783	}
784	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
785		soisdisconnected(tp->t_inpcb->inp_socket);
786		/* To prevent the connection hanging in FIN_WAIT_2 forever. */
787		if (tp->t_state == TCPS_FIN_WAIT_2)
788			tp->t_timer[TCPT_2MSL] = tcp_maxidle;
789	}
790	return (tp);
791}
792
793