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 * $FreeBSD$
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 */
69 extern char *tcpstates[]; /* XXX ??? */
70
71 static int tcp_attach __P((struct socket *, struct proc *));
72 static int tcp_connect __P((struct tcpcb *, struct sockaddr *,
73 struct proc *));
74 static struct tcpcb *
75 tcp_disconnect __P((struct tcpcb *));
76 static 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 */
94 static int
95 tcp_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;
115 tp = sototcpcb(so);
116 out:
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 */
129 static int
130 tcp_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 */
167 static int
168 tcp_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 */
198 static int
199 tcp_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 */
221 static int
222 tcp_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 */
259 static int
260 tcp_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 */
277 static int
278 tcp_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 */
293 static int
294 tcp_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 */
312 static int
313 tcp_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. Unlike the other
328 * pru_*() routines, the mbuf chains are our responsibility. We
329 * must either enqueue them or free them. The other pru_* routines
330 * generally are caller-frees.
331 */
332 static int
333 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
334 struct sockaddr *nam, struct mbuf *control, struct proc *p)
335 {
336 int s = splnet();
337 int error = 0;
338 struct inpcb *inp = sotoinpcb(so);
339 struct tcpcb *tp;
340 TCPDEBUG0;
341
342 if (inp == NULL) {
343 /*
344 * OOPS! we lost a race, the TCP session got reset after
345 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a
346 * network interrupt in the non-splnet() section of sosend().
347 */
348 if (m)
349 m_freem(m);
350 if (control)
351 m_freem(control);
352 error = ECONNRESET; /* XXX EPIPE? */
353 goto out;
354 }
355 tp = intotcpcb(inp);
356 TCPDEBUG1();
357 if (control) {
358 /* TCP doesn't do control messages (rights, creds, etc) */
359 if (control->m_len) {
360 m_freem(control);
361 if (m)
362 m_freem(m);
363 error = EINVAL;
364 goto out;
365 }
366 m_freem(control); /* empty control, just free it */
367 }
368 if(!(flags & PRUS_OOB)) {
369 sbappend(&so->so_snd, m);
370 if (nam && tp->t_state < TCPS_SYN_SENT) {
371 /*
372 * Do implied connect if not yet connected,
373 * initialize window to default value, and
374 * initialize maxseg/maxopd using peer's cached
375 * MSS.
376 */
377 error = tcp_connect(tp, nam, p);
378 if (error)
379 goto out;
380 tp->snd_wnd = TTCP_CLIENT_SND_WND;
381 tcp_mss(tp, -1);
382 }
383
384 if (flags & PRUS_EOF) {
385 /*
386 * Close the send side of the connection after
387 * the data is sent.
388 */
389 socantsendmore(so);
390 tp = tcp_usrclosed(tp);
391 }
392 if (tp != NULL) {
393 if (flags & PRUS_MORETOCOME)
394 tp->t_flags |= TF_MORETOCOME;
395 error = tcp_output(tp);
396 if (flags & PRUS_MORETOCOME)
397 tp->t_flags &= ~TF_MORETOCOME;
398 }
399 } else {
400 if (sbspace(&so->so_snd) < -512) {
401 m_freem(m);
402 error = ENOBUFS;
403 goto out;
404 }
405 /*
406 * According to RFC961 (Assigned Protocols),
407 * the urgent pointer points to the last octet
408 * of urgent data. We continue, however,
409 * to consider it to indicate the first octet
410 * of data past the urgent section.
411 * Otherwise, snd_up should be one lower.
412 */
413 sbappend(&so->so_snd, m);
414 if (nam && tp->t_state < TCPS_SYN_SENT) {
415 /*
416 * Do implied connect if not yet connected,
417 * initialize window to default value, and
418 * initialize maxseg/maxopd using peer's cached
419 * MSS.
420 */
421 error = tcp_connect(tp, nam, p);
422 if (error)
423 goto out;
424 tp->snd_wnd = TTCP_CLIENT_SND_WND;
425 tcp_mss(tp, -1);
426 }
427 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
428 tp->t_force = 1;
429 error = tcp_output(tp);
430 tp->t_force = 0;
431 }
432 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
433 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
434 }
435
436 /*
437 * Abort the TCP.
438 */
439 static int
440 tcp_usr_abort(struct socket *so)
441 {
442 int s = splnet();
443 int error = 0;
444 struct inpcb *inp = sotoinpcb(so);
445 struct tcpcb *tp;
446
447 COMMON_START();
448 tp = tcp_drop(tp, ECONNABORTED);
449 COMMON_END(PRU_ABORT);
450 }
451
452 /*
453 * Receive out-of-band data.
454 */
455 static int
456 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
457 {
458 int s = splnet();
459 int error = 0;
460 struct inpcb *inp = sotoinpcb(so);
461 struct tcpcb *tp;
462
463 COMMON_START();
464 if ((so->so_oobmark == 0 &&
465 (so->so_state & SS_RCVATMARK) == 0) ||
466 so->so_options & SO_OOBINLINE ||
467 tp->t_oobflags & TCPOOB_HADDATA) {
468 error = EINVAL;
469 goto out;
470 }
471 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
472 error = EWOULDBLOCK;
473 goto out;
474 }
475 m->m_len = 1;
476 *mtod(m, caddr_t) = tp->t_iobc;
477 if ((flags & MSG_PEEK) == 0)
478 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
479 COMMON_END(PRU_RCVOOB);
480 }
481
482 /* xxx - should be const */
483 struct pr_usrreqs tcp_usrreqs = {
484 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind,
485 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach,
486 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd,
487 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
488 in_setsockaddr, sosend, soreceive, sopoll
489 };
490
491 /*
492 * Common subroutine to open a TCP connection to remote host specified
493 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
494 * port number if needed. Call in_pcbladdr to do the routing and to choose
495 * a local host address (interface). If there is an existing incarnation
496 * of the same connection in TIME-WAIT state and if the remote host was
497 * sending CC options and if the connection duration was < MSL, then
498 * truncate the previous TIME-WAIT state and proceed.
499 * Initialize connection parameters and enter SYN-SENT state.
500 */
501 static int
502 tcp_connect(tp, nam, p)
503 register struct tcpcb *tp;
504 struct sockaddr *nam;
505 struct proc *p;
506 {
507 struct inpcb *inp = tp->t_inpcb, *oinp;
508 struct socket *so = inp->inp_socket;
509 struct tcpcb *otp;
510 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
511 struct sockaddr_in *ifaddr;
512 struct rmxp_tao *taop;
513 struct rmxp_tao tao_noncached;
514 int error;
515
516 if (inp->inp_lport == 0) {
517 error = in_pcbbind(inp, (struct sockaddr *)0, p);
518 if (error)
519 return error;
520 }
521
522 /*
523 * Cannot simply call in_pcbconnect, because there might be an
524 * earlier incarnation of this same connection still in
525 * TIME_WAIT state, creating an ADDRINUSE error.
526 */
527 error = in_pcbladdr(inp, nam, &ifaddr);
528 if (error)
529 return error;
530 oinp = in_pcblookup_hash(inp->inp_pcbinfo,
531 sin->sin_addr, sin->sin_port,
532 inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr
533 : ifaddr->sin_addr,
534 inp->inp_lport, 0);
535 if (oinp) {
536 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
537 otp->t_state == TCPS_TIME_WAIT &&
538 otp->t_duration < TCPTV_MSL &&
539 (otp->t_flags & TF_RCVD_CC))
540 otp = tcp_close(otp);
541 else
542 return EADDRINUSE;
543 }
544 if (inp->inp_laddr.s_addr == INADDR_ANY)
545 inp->inp_laddr = ifaddr->sin_addr;
546 inp->inp_faddr = sin->sin_addr;
547 inp->inp_fport = sin->sin_port;
548 in_pcbrehash(inp);
549
550 tp->t_template = tcp_template(tp);
551 if (tp->t_template == 0) {
552 in_pcbdisconnect(inp);
553 return ENOBUFS;
554 }
555
556 /* Compute window scaling to request. */
557 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
558 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
559 tp->request_r_scale++;
560
561 soisconnecting(so);
562 tcpstat.tcps_connattempt++;
563 tp->t_state = TCPS_SYN_SENT;
564 tp->t_timer[TCPT_KEEP] = tcp_keepinit;
565 #ifdef TCP_COMPAT_42
566 tp->iss = tcp_iss;
567 tcp_iss += TCP_ISSINCR/2;
568 #else /* TCP_COMPAT_42 */
569 tp->iss = tcp_rndiss_next();
570 #endif /* !TCP_COMPAT_42 */
571 tcp_sendseqinit(tp);
572
573 /*
574 * Generate a CC value for this connection and
575 * check whether CC or CCnew should be used.
576 */
577 if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) {
578 taop = &tao_noncached;
579 bzero(taop, sizeof(*taop));
580 }
581
582 tp->cc_send = CC_INC(tcp_ccgen);
583 if (taop->tao_ccsent != 0 &&
584 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
585 taop->tao_ccsent = tp->cc_send;
586 } else {
587 taop->tao_ccsent = 0;
588 tp->t_flags |= TF_SENDCCNEW;
589 }
590
591 return 0;
592 }
593
594 /*
595 * The new sockopt interface makes it possible for us to block in the
596 * copyin/out step (if we take a page fault). Taking a page fault at
597 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now
598 * use TSM, there probably isn't any need for this function to run at
599 * splnet() any more. This needs more examination.)
600 */
601 int
602 tcp_ctloutput(so, sopt)
603 struct socket *so;
604 struct sockopt *sopt;
605 {
606 int error, opt, optval, s;
607 struct inpcb *inp;
608 struct tcpcb *tp;
609
610 error = 0;
611 s = splnet(); /* XXX */
612 inp = sotoinpcb(so);
613 if (inp == NULL) {
614 splx(s);
615 return (ECONNRESET);
616 }
617 if (sopt->sopt_level != IPPROTO_TCP) {
618 error = ip_ctloutput(so, sopt);
619 splx(s);
620 return (error);
621 }
622 tp = intotcpcb(inp);
623
624 switch (sopt->sopt_dir) {
625 case SOPT_SET:
626 switch (sopt->sopt_name) {
627 case TCP_NODELAY:
628 case TCP_NOOPT:
629 case TCP_NOPUSH:
630 error = sooptcopyin(sopt, &optval, sizeof optval,
631 sizeof optval);
632 if (error)
633 break;
634
635 switch (sopt->sopt_name) {
636 case TCP_NODELAY:
637 opt = TF_NODELAY;
638 break;
639 case TCP_NOOPT:
640 opt = TF_NOOPT;
641 break;
642 case TCP_NOPUSH:
643 opt = TF_NOPUSH;
644 break;
645 default:
646 opt = 0; /* dead code to fool gcc */
647 break;
648 }
649
650 if (optval)
651 tp->t_flags |= opt;
652 else
653 tp->t_flags &= ~opt;
654 break;
655
656 case TCP_MAXSEG:
657 error = sooptcopyin(sopt, &optval, sizeof optval,
658 sizeof optval);
659 if (error)
660 break;
661
662 if (optval > 0 && optval <= tp->t_maxseg)
663 tp->t_maxseg = optval;
664 else
665 error = EINVAL;
666 break;
667
668 default:
669 error = ENOPROTOOPT;
670 break;
671 }
672 break;
673
674 case SOPT_GET:
675 switch (sopt->sopt_name) {
676 case TCP_NODELAY:
677 optval = tp->t_flags & TF_NODELAY;
678 break;
679 case TCP_MAXSEG:
680 optval = tp->t_maxseg;
681 break;
682 case TCP_NOOPT:
683 optval = tp->t_flags & TF_NOOPT;
684 break;
685 case TCP_NOPUSH:
686 optval = tp->t_flags & TF_NOPUSH;
687 break;
688 default:
689 error = ENOPROTOOPT;
690 break;
691 }
692 if (error == 0)
693 error = sooptcopyout(sopt, &optval, sizeof optval);
694 break;
695 }
696 splx(s);
697 return (error);
698 }
699
700 /*
701 * tcp_sendspace and tcp_recvspace are the default send and receive window
702 * sizes, respectively. These are obsolescent (this information should
703 * be set by the route).
704 */
705 u_long tcp_sendspace = 1024*16;
706 SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace,
707 CTLFLAG_RW, &tcp_sendspace , 0, "");
708 u_long tcp_recvspace = 1024*16;
709 SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace,
710 CTLFLAG_RW, &tcp_recvspace , 0, "");
711
712 /*
713 * Attach TCP protocol to socket, allocating
714 * internet protocol control block, tcp control block,
715 * bufer space, and entering LISTEN state if to accept connections.
716 */
717 static int
718 tcp_attach(so, p)
719 struct socket *so;
720 struct proc *p;
721 {
722 register struct tcpcb *tp;
723 struct inpcb *inp;
724 int error;
725
726 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
727 error = soreserve(so, tcp_sendspace, tcp_recvspace);
728 if (error)
729 return (error);
730 }
731 error = in_pcballoc(so, &tcbinfo, p);
732 if (error)
733 return (error);
734 inp = sotoinpcb(so);
735 tp = tcp_newtcpcb(inp);
736 if (tp == 0) {
737 int nofd = so->so_state & SS_NOFDREF; /* XXX */
738
739 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */
740 in_pcbdetach(inp);
741 so->so_state |= nofd;
742 return (ENOBUFS);
743 }
744 tp->t_state = TCPS_CLOSED;
745 return (0);
746 }
747
748 /*
749 * Initiate (or continue) disconnect.
750 * If embryonic state, just send reset (once).
751 * If in ``let data drain'' option and linger null, just drop.
752 * Otherwise (hard), mark socket disconnecting and drop
753 * current input data; switch states based on user close, and
754 * send segment to peer (with FIN).
755 */
756 static struct tcpcb *
757 tcp_disconnect(tp)
758 register struct tcpcb *tp;
759 {
760 struct socket *so = tp->t_inpcb->inp_socket;
761
762 if (tp->t_state < TCPS_ESTABLISHED)
763 tp = tcp_close(tp);
764 else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
765 tp = tcp_drop(tp, 0);
766 else {
767 soisdisconnecting(so);
768 sbflush(&so->so_rcv);
769 tp = tcp_usrclosed(tp);
770 if (tp)
771 (void) tcp_output(tp);
772 }
773 return (tp);
774 }
775
776 /*
777 * User issued close, and wish to trail through shutdown states:
778 * if never received SYN, just forget it. If got a SYN from peer,
779 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
780 * If already got a FIN from peer, then almost done; go to LAST_ACK
781 * state. In all other cases, have already sent FIN to peer (e.g.
782 * after PRU_SHUTDOWN), and just have to play tedious game waiting
783 * for peer to send FIN or not respond to keep-alives, etc.
784 * We can let the user exit from the close as soon as the FIN is acked.
785 */
786 static struct tcpcb *
787 tcp_usrclosed(tp)
788 register struct tcpcb *tp;
789 {
790
791 switch (tp->t_state) {
792
793 case TCPS_CLOSED:
794 case TCPS_LISTEN:
795 tp->t_state = TCPS_CLOSED;
796 tp = tcp_close(tp);
797 break;
798
799 case TCPS_SYN_SENT:
800 case TCPS_SYN_RECEIVED:
801 tp->t_flags |= TF_NEEDFIN;
802 break;
803
804 case TCPS_ESTABLISHED:
805 tp->t_state = TCPS_FIN_WAIT_1;
806 break;
807
808 case TCPS_CLOSE_WAIT:
809 tp->t_state = TCPS_LAST_ACK;
810 break;
811 }
812 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
813 soisdisconnected(tp->t_inpcb->inp_socket);
814 /* To prevent the connection hanging in FIN_WAIT_2 forever. */
815 if (tp->t_state == TCPS_FIN_WAIT_2)
816 tp->t_timer[TCPT_2MSL] = tcp_maxidle;
817 }
818 return (tp);
819 }
820
Cache object: 1ba3fd3120bc5a62047df939c0cedd3d
|