tcp_usrreq.c revision 42902
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 1.39 1998/12/07 21:58:42 archie Exp $
35 */
36
37#include "opt_tcpdebug.h"
38
39#include <sys/param.h>
40#include <sys/systm.h>
41#include <sys/kernel.h>
42#include <sys/sysctl.h>
43#include <sys/mbuf.h>
44#include <sys/socket.h>
45#include <sys/socketvar.h>
46#include <sys/protosw.h>
47
48#include <net/if.h>
49#include <net/route.h>
50
51#include <netinet/in.h>
52#include <netinet/in_systm.h>
53#include <netinet/in_pcb.h>
54#include <netinet/in_var.h>
55#include <netinet/ip_var.h>
56#include <netinet/tcp.h>
57#include <netinet/tcp_fsm.h>
58#include <netinet/tcp_seq.h>
59#include <netinet/tcp_timer.h>
60#include <netinet/tcp_var.h>
61#include <netinet/tcpip.h>
62#ifdef TCPDEBUG
63#include <netinet/tcp_debug.h>
64#endif
65
66/*
67 * TCP protocol interface to socket abstraction.
68 */
69extern	char *tcpstates[];	/* XXX ??? */
70
71static int	tcp_attach __P((struct socket *, struct proc *));
72static int	tcp_connect __P((struct tcpcb *, struct sockaddr *,
73				 struct proc *));
74static struct tcpcb *
75		tcp_disconnect __P((struct tcpcb *));
76static struct tcpcb *
77		tcp_usrclosed __P((struct tcpcb *));
78
79#ifdef TCPDEBUG
80#define	TCPDEBUG0	int ostate
81#define	TCPDEBUG1()	ostate = tp ? tp->t_state : 0
82#define	TCPDEBUG2(req)	if (tp && (so->so_options & SO_DEBUG)) \
83				tcp_trace(TA_USER, ostate, tp, 0, req)
84#else
85#define	TCPDEBUG0
86#define	TCPDEBUG1()
87#define	TCPDEBUG2(req)
88#endif
89
90/*
91 * TCP attaches to socket via pru_attach(), reserving space,
92 * and an internet control block.
93 */
94static int
95tcp_usr_attach(struct socket *so, int proto, struct proc *p)
96{
97	int s = splnet();
98	int error;
99	struct inpcb *inp = sotoinpcb(so);
100	struct tcpcb *tp = 0;
101	TCPDEBUG0;
102
103	TCPDEBUG1();
104	if (inp) {
105		error = EISCONN;
106		goto out;
107	}
108
109	error = tcp_attach(so, p);
110	if (error)
111		goto out;
112
113	if ((so->so_options & SO_LINGER) && so->so_linger == 0)
114		so->so_linger = TCP_LINGERTIME * hz;
115	tp = sototcpcb(so);
116out:
117	TCPDEBUG2(PRU_ATTACH);
118	splx(s);
119	return error;
120}
121
122/*
123 * pru_detach() detaches the TCP protocol from the socket.
124 * If the protocol state is non-embryonic, then can't
125 * do this directly: have to initiate a pru_disconnect(),
126 * which may finish later; embryonic TCB's can just
127 * be discarded here.
128 */
129static int
130tcp_usr_detach(struct socket *so)
131{
132	int s = splnet();
133	int error = 0;
134	struct inpcb *inp = sotoinpcb(so);
135	struct tcpcb *tp;
136	TCPDEBUG0;
137
138	if (inp == 0) {
139		splx(s);
140		return EINVAL;	/* XXX */
141	}
142	tp = intotcpcb(inp);
143	TCPDEBUG1();
144	tp = tcp_disconnect(tp);
145
146	TCPDEBUG2(PRU_DETACH);
147	splx(s);
148	return error;
149}
150
151#define	COMMON_START()	TCPDEBUG0; \
152			do { \
153				     if (inp == 0) { \
154					     splx(s); \
155					     return EINVAL; \
156				     } \
157				     tp = intotcpcb(inp); \
158				     TCPDEBUG1(); \
159		     } while(0)
160
161#define COMMON_END(req)	out: TCPDEBUG2(req); splx(s); return error; goto out
162
163
164/*
165 * Give the socket an address.
166 */
167static int
168tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
169{
170	int s = splnet();
171	int error = 0;
172	struct inpcb *inp = sotoinpcb(so);
173	struct tcpcb *tp;
174	struct sockaddr_in *sinp;
175
176	COMMON_START();
177
178	/*
179	 * Must check for multicast addresses and disallow binding
180	 * to them.
181	 */
182	sinp = (struct sockaddr_in *)nam;
183	if (sinp->sin_family == AF_INET &&
184	    IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
185		error = EAFNOSUPPORT;
186		goto out;
187	}
188	error = in_pcbbind(inp, nam, p);
189	if (error)
190		goto out;
191	COMMON_END(PRU_BIND);
192
193}
194
195/*
196 * Prepare to accept connections.
197 */
198static int
199tcp_usr_listen(struct socket *so, struct proc *p)
200{
201	int s = splnet();
202	int error = 0;
203	struct inpcb *inp = sotoinpcb(so);
204	struct tcpcb *tp;
205
206	COMMON_START();
207	if (inp->inp_lport == 0)
208		error = in_pcbbind(inp, (struct sockaddr *)0, p);
209	if (error == 0)
210		tp->t_state = TCPS_LISTEN;
211	COMMON_END(PRU_LISTEN);
212}
213
214/*
215 * Initiate connection to peer.
216 * Create a template for use in transmissions on this connection.
217 * Enter SYN_SENT state, and mark socket as connecting.
218 * Start keep-alive timer, and seed output sequence space.
219 * Send initial segment on connection.
220 */
221static int
222tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
223{
224	int s = splnet();
225	int error = 0;
226	struct inpcb *inp = sotoinpcb(so);
227	struct tcpcb *tp;
228	struct sockaddr_in *sinp;
229
230	COMMON_START();
231
232	/*
233	 * Must disallow TCP ``connections'' to multicast addresses.
234	 */
235	sinp = (struct sockaddr_in *)nam;
236	if (sinp->sin_family == AF_INET
237	    && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
238		error = EAFNOSUPPORT;
239		goto out;
240	}
241
242	if ((error = tcp_connect(tp, nam, p)) != 0)
243		goto out;
244	error = tcp_output(tp);
245	COMMON_END(PRU_CONNECT);
246}
247
248/*
249 * Initiate disconnect from peer.
250 * If connection never passed embryonic stage, just drop;
251 * else if don't need to let data drain, then can just drop anyways,
252 * else have to begin TCP shutdown process: mark socket disconnecting,
253 * drain unread data, state switch to reflect user close, and
254 * send segment (e.g. FIN) to peer.  Socket will be really disconnected
255 * when peer sends FIN and acks ours.
256 *
257 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
258 */
259static int
260tcp_usr_disconnect(struct socket *so)
261{
262	int s = splnet();
263	int error = 0;
264	struct inpcb *inp = sotoinpcb(so);
265	struct tcpcb *tp;
266
267	COMMON_START();
268	tp = tcp_disconnect(tp);
269	COMMON_END(PRU_DISCONNECT);
270}
271
272/*
273 * Accept a connection.  Essentially all the work is
274 * done at higher levels; just return the address
275 * of the peer, storing through addr.
276 */
277static int
278tcp_usr_accept(struct socket *so, struct sockaddr **nam)
279{
280	int s = splnet();
281	int error = 0;
282	struct inpcb *inp = sotoinpcb(so);
283	struct tcpcb *tp;
284
285	COMMON_START();
286	in_setpeeraddr(so, nam);
287	COMMON_END(PRU_ACCEPT);
288}
289
290/*
291 * Mark the connection as being incapable of further output.
292 */
293static int
294tcp_usr_shutdown(struct socket *so)
295{
296	int s = splnet();
297	int error = 0;
298	struct inpcb *inp = sotoinpcb(so);
299	struct tcpcb *tp;
300
301	COMMON_START();
302	socantsendmore(so);
303	tp = tcp_usrclosed(tp);
304	if (tp)
305		error = tcp_output(tp);
306	COMMON_END(PRU_SHUTDOWN);
307}
308
309/*
310 * After a receive, possibly send window update to peer.
311 */
312static int
313tcp_usr_rcvd(struct socket *so, int flags)
314{
315	int s = splnet();
316	int error = 0;
317	struct inpcb *inp = sotoinpcb(so);
318	struct tcpcb *tp;
319
320	COMMON_START();
321	tcp_output(tp);
322	COMMON_END(PRU_RCVD);
323}
324
325/*
326 * Do a send by putting data in output queue and updating urgent
327 * marker if URG set.  Possibly send more data.
328 */
329static int
330tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
331	     struct sockaddr *nam, struct mbuf *control, struct proc *p)
332{
333	int s = splnet();
334	int error = 0;
335	struct inpcb *inp = sotoinpcb(so);
336	struct tcpcb *tp;
337
338	COMMON_START();
339	if (control && control->m_len) {
340		m_freem(control); /* XXX shouldn't caller do this??? */
341		if (m)
342			m_freem(m);
343		error = EINVAL;
344		goto out;
345	}
346
347	if(!(flags & PRUS_OOB)) {
348		sbappend(&so->so_snd, m);
349		if (nam && tp->t_state < TCPS_SYN_SENT) {
350			/*
351			 * Do implied connect if not yet connected,
352			 * initialize window to default value, and
353			 * initialize maxseg/maxopd using peer's cached
354			 * MSS.
355			 */
356			error = tcp_connect(tp, nam, p);
357			if (error)
358				goto out;
359			tp->snd_wnd = TTCP_CLIENT_SND_WND;
360			tcp_mss(tp, -1);
361		}
362
363		if (flags & PRUS_EOF) {
364			/*
365			 * Close the send side of the connection after
366			 * the data is sent.
367			 */
368			socantsendmore(so);
369			tp = tcp_usrclosed(tp);
370		}
371		if (tp != NULL) {
372			if (flags & PRUS_MORETOCOME)
373				tp->t_flags |= TF_MORETOCOME;
374			error = tcp_output(tp);
375			if (flags & PRUS_MORETOCOME)
376				tp->t_flags &= ~TF_MORETOCOME;
377		}
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, sopoll
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 sockaddr *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 = (struct sockaddr_in *)nam;
490	struct sockaddr_in *ifaddr;
491	struct rmxp_tao *taop;
492	struct rmxp_tao tao_noncached;
493	int error;
494
495	if (inp->inp_lport == 0) {
496		error = in_pcbbind(inp, (struct sockaddr *)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_pcblookup_hash(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
568/*
569 * The new sockopt interface makes it possible for us to block in the
570 * copyin/out step (if we take a page fault).  Taking a page fault at
571 * splnet() is probably a Bad Thing.  (Since sockets and pcbs both now
572 * use TSM, there probably isn't any need for this function to run at
573 * splnet() any more.  This needs more examination.)
574 */
575int
576tcp_ctloutput(so, sopt)
577	struct socket *so;
578	struct sockopt *sopt;
579{
580	int	error, opt, optval, s;
581	struct	inpcb *inp;
582	struct	tcpcb *tp;
583
584	error = 0;
585	s = splnet();		/* XXX */
586	inp = sotoinpcb(so);
587	if (inp == NULL) {
588		splx(s);
589		return (ECONNRESET);
590	}
591	if (sopt->sopt_level != IPPROTO_TCP) {
592		error = ip_ctloutput(so, sopt);
593		splx(s);
594		return (error);
595	}
596	tp = intotcpcb(inp);
597
598	switch (sopt->sopt_dir) {
599	case SOPT_SET:
600		switch (sopt->sopt_name) {
601		case TCP_NODELAY:
602		case TCP_NOOPT:
603		case TCP_NOPUSH:
604			error = sooptcopyin(sopt, &optval, sizeof optval,
605					    sizeof optval);
606			if (error)
607				break;
608
609			switch (sopt->sopt_name) {
610			case TCP_NODELAY:
611				opt = TF_NODELAY;
612				break;
613			case TCP_NOOPT:
614				opt = TF_NOOPT;
615				break;
616			case TCP_NOPUSH:
617				opt = TF_NOPUSH;
618				break;
619			default:
620				opt = 0; /* dead code to fool gcc */
621				break;
622			}
623
624			if (optval)
625				tp->t_flags |= opt;
626			else
627				tp->t_flags &= ~opt;
628			break;
629
630		case TCP_MAXSEG:
631			error = sooptcopyin(sopt, &optval, sizeof optval,
632					    sizeof optval);
633			if (error)
634				break;
635
636			if (optval > 0 && optval <= tp->t_maxseg)
637				tp->t_maxseg = optval;
638			else
639				error = EINVAL;
640			break;
641
642		default:
643			error = ENOPROTOOPT;
644			break;
645		}
646		break;
647
648	case SOPT_GET:
649		switch (sopt->sopt_name) {
650		case TCP_NODELAY:
651			optval = tp->t_flags & TF_NODELAY;
652			break;
653		case TCP_MAXSEG:
654			optval = tp->t_maxseg;
655			break;
656		case TCP_NOOPT:
657			optval = tp->t_flags & TF_NOOPT;
658			break;
659		case TCP_NOPUSH:
660			optval = tp->t_flags & TF_NOPUSH;
661			break;
662		default:
663			error = ENOPROTOOPT;
664			break;
665		}
666		if (error == 0)
667			error = sooptcopyout(sopt, &optval, sizeof optval);
668		break;
669	}
670	splx(s);
671	return (error);
672}
673
674/*
675 * tcp_sendspace and tcp_recvspace are the default send and receive window
676 * sizes, respectively.  These are obsolescent (this information should
677 * be set by the route).
678 */
679u_long	tcp_sendspace = 1024*16;
680SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace,
681	CTLFLAG_RW, &tcp_sendspace , 0, "");
682u_long	tcp_recvspace = 1024*16;
683SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace,
684	CTLFLAG_RW, &tcp_recvspace , 0, "");
685
686/*
687 * Attach TCP protocol to socket, allocating
688 * internet protocol control block, tcp control block,
689 * bufer space, and entering LISTEN state if to accept connections.
690 */
691static int
692tcp_attach(so, p)
693	struct socket *so;
694	struct proc *p;
695{
696	register struct tcpcb *tp;
697	struct inpcb *inp;
698	int error;
699
700	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
701		error = soreserve(so, tcp_sendspace, tcp_recvspace);
702		if (error)
703			return (error);
704	}
705	error = in_pcballoc(so, &tcbinfo, p);
706	if (error)
707		return (error);
708	inp = sotoinpcb(so);
709	tp = tcp_newtcpcb(inp);
710	if (tp == 0) {
711		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
712
713		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
714		in_pcbdetach(inp);
715		so->so_state |= nofd;
716		return (ENOBUFS);
717	}
718	tp->t_state = TCPS_CLOSED;
719	return (0);
720}
721
722/*
723 * Initiate (or continue) disconnect.
724 * If embryonic state, just send reset (once).
725 * If in ``let data drain'' option and linger null, just drop.
726 * Otherwise (hard), mark socket disconnecting and drop
727 * current input data; switch states based on user close, and
728 * send segment to peer (with FIN).
729 */
730static struct tcpcb *
731tcp_disconnect(tp)
732	register struct tcpcb *tp;
733{
734	struct socket *so = tp->t_inpcb->inp_socket;
735
736	if (tp->t_state < TCPS_ESTABLISHED)
737		tp = tcp_close(tp);
738	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
739		tp = tcp_drop(tp, 0);
740	else {
741		soisdisconnecting(so);
742		sbflush(&so->so_rcv);
743		tp = tcp_usrclosed(tp);
744		if (tp)
745			(void) tcp_output(tp);
746	}
747	return (tp);
748}
749
750/*
751 * User issued close, and wish to trail through shutdown states:
752 * if never received SYN, just forget it.  If got a SYN from peer,
753 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
754 * If already got a FIN from peer, then almost done; go to LAST_ACK
755 * state.  In all other cases, have already sent FIN to peer (e.g.
756 * after PRU_SHUTDOWN), and just have to play tedious game waiting
757 * for peer to send FIN or not respond to keep-alives, etc.
758 * We can let the user exit from the close as soon as the FIN is acked.
759 */
760static struct tcpcb *
761tcp_usrclosed(tp)
762	register struct tcpcb *tp;
763{
764
765	switch (tp->t_state) {
766
767	case TCPS_CLOSED:
768	case TCPS_LISTEN:
769		tp->t_state = TCPS_CLOSED;
770		tp = tcp_close(tp);
771		break;
772
773	case TCPS_SYN_SENT:
774	case TCPS_SYN_RECEIVED:
775		tp->t_flags |= TF_NEEDFIN;
776		break;
777
778	case TCPS_ESTABLISHED:
779		tp->t_state = TCPS_FIN_WAIT_1;
780		break;
781
782	case TCPS_CLOSE_WAIT:
783		tp->t_state = TCPS_LAST_ACK;
784		break;
785	}
786	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
787		soisdisconnected(tp->t_inpcb->inp_socket);
788		/* To prevent the connection hanging in FIN_WAIT_2 forever. */
789		if (tp->t_state == TCPS_FIN_WAIT_2)
790			tp->t_timer[TCPT_2MSL] = tcp_maxidle;
791	}
792	return (tp);
793}
794
795