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
| 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
|
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;
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 prison_remote_ip(p, 0, &sinp->sin_addr.s_addr);
243
244 if ((error = tcp_connect(tp, nam, p)) != 0)
245 goto out;
246 error = tcp_output(tp);
247 COMMON_END(PRU_CONNECT);
248}
249
250/*
251 * Initiate disconnect from peer.
252 * If connection never passed embryonic stage, just drop;
253 * else if don't need to let data drain, then can just drop anyways,
254 * else have to begin TCP shutdown process: mark socket disconnecting,
255 * drain unread data, state switch to reflect user close, and
256 * send segment (e.g. FIN) to peer. Socket will be really disconnected
257 * when peer sends FIN and acks ours.
258 *
259 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
260 */
261static int
262tcp_usr_disconnect(struct socket *so)
263{
264 int s = splnet();
265 int error = 0;
266 struct inpcb *inp = sotoinpcb(so);
267 struct tcpcb *tp;
268
269 COMMON_START();
270 tp = tcp_disconnect(tp);
271 COMMON_END(PRU_DISCONNECT);
272}
273
274/*
275 * Accept a connection. Essentially all the work is
276 * done at higher levels; just return the address
277 * of the peer, storing through addr.
278 */
279static int
280tcp_usr_accept(struct socket *so, struct sockaddr **nam)
281{
282 int s = splnet();
283 int error = 0;
284 struct inpcb *inp = sotoinpcb(so);
285 struct tcpcb *tp;
286
287 COMMON_START();
288 in_setpeeraddr(so, nam);
289 COMMON_END(PRU_ACCEPT);
290}
291
292/*
293 * Mark the connection as being incapable of further output.
294 */
295static int
296tcp_usr_shutdown(struct socket *so)
297{
298 int s = splnet();
299 int error = 0;
300 struct inpcb *inp = sotoinpcb(so);
301 struct tcpcb *tp;
302
303 COMMON_START();
304 socantsendmore(so);
305 tp = tcp_usrclosed(tp);
306 if (tp)
307 error = tcp_output(tp);
308 COMMON_END(PRU_SHUTDOWN);
309}
310
311/*
312 * After a receive, possibly send window update to peer.
313 */
314static int
315tcp_usr_rcvd(struct socket *so, int flags)
316{
317 int s = splnet();
318 int error = 0;
319 struct inpcb *inp = sotoinpcb(so);
320 struct tcpcb *tp;
321
322 COMMON_START();
323 tcp_output(tp);
324 COMMON_END(PRU_RCVD);
325}
326
327/*
328 * Do a send by putting data in output queue and updating urgent
329 * marker if URG set. Possibly send more data.
330 */
331static int
332tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
333 struct sockaddr *nam, struct mbuf *control, struct proc *p)
334{
335 int s = splnet();
336 int error = 0;
337 struct inpcb *inp = sotoinpcb(so);
338 struct tcpcb *tp;
339
340 COMMON_START();
341 if (control && control->m_len) {
342 m_freem(control); /* XXX shouldn't caller do this??? */
343 if (m)
344 m_freem(m);
345 error = EINVAL;
346 goto out;
347 }
348
349 if(!(flags & PRUS_OOB)) {
350 sbappend(&so->so_snd, m);
351 if (nam && tp->t_state < TCPS_SYN_SENT) {
352 /*
353 * Do implied connect if not yet connected,
354 * initialize window to default value, and
355 * initialize maxseg/maxopd using peer's cached
356 * MSS.
357 */
358 error = tcp_connect(tp, nam, p);
359 if (error)
360 goto out;
361 tp->snd_wnd = TTCP_CLIENT_SND_WND;
362 tcp_mss(tp, -1);
363 }
364
365 if (flags & PRUS_EOF) {
366 /*
367 * Close the send side of the connection after
368 * the data is sent.
369 */
370 socantsendmore(so);
371 tp = tcp_usrclosed(tp);
372 }
373 if (tp != NULL) {
374 if (flags & PRUS_MORETOCOME)
375 tp->t_flags |= TF_MORETOCOME;
376 error = tcp_output(tp);
377 if (flags & PRUS_MORETOCOME)
378 tp->t_flags &= ~TF_MORETOCOME;
379 }
380 } else {
381 if (sbspace(&so->so_snd) < -512) {
382 m_freem(m);
383 error = ENOBUFS;
384 goto out;
385 }
386 /*
387 * According to RFC961 (Assigned Protocols),
388 * the urgent pointer points to the last octet
389 * of urgent data. We continue, however,
390 * to consider it to indicate the first octet
391 * of data past the urgent section.
392 * Otherwise, snd_up should be one lower.
393 */
394 sbappend(&so->so_snd, m);
395 if (nam && tp->t_state < TCPS_SYN_SENT) {
396 /*
397 * Do implied connect if not yet connected,
398 * initialize window to default value, and
399 * initialize maxseg/maxopd using peer's cached
400 * MSS.
401 */
402 error = tcp_connect(tp, nam, p);
403 if (error)
404 goto out;
405 tp->snd_wnd = TTCP_CLIENT_SND_WND;
406 tcp_mss(tp, -1);
407 }
408 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
409 tp->t_force = 1;
410 error = tcp_output(tp);
411 tp->t_force = 0;
412 }
413 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
414 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
415}
416
417/*
418 * Abort the TCP.
419 */
420static int
421tcp_usr_abort(struct socket *so)
422{
423 int s = splnet();
424 int error = 0;
425 struct inpcb *inp = sotoinpcb(so);
426 struct tcpcb *tp;
427
428 COMMON_START();
429 tp = tcp_drop(tp, ECONNABORTED);
430 COMMON_END(PRU_ABORT);
431}
432
433/*
434 * Receive out-of-band data.
435 */
436static int
437tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
438{
439 int s = splnet();
440 int error = 0;
441 struct inpcb *inp = sotoinpcb(so);
442 struct tcpcb *tp;
443
444 COMMON_START();
445 if ((so->so_oobmark == 0 &&
446 (so->so_state & SS_RCVATMARK) == 0) ||
447 so->so_options & SO_OOBINLINE ||
448 tp->t_oobflags & TCPOOB_HADDATA) {
449 error = EINVAL;
450 goto out;
451 }
452 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
453 error = EWOULDBLOCK;
454 goto out;
455 }
456 m->m_len = 1;
457 *mtod(m, caddr_t) = tp->t_iobc;
458 if ((flags & MSG_PEEK) == 0)
459 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
460 COMMON_END(PRU_RCVOOB);
461}
462
463/* xxx - should be const */
464struct pr_usrreqs tcp_usrreqs = {
465 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind,
466 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach,
467 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd,
468 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
469 in_setsockaddr, sosend, soreceive, sopoll
470};
471
472/*
473 * Common subroutine to open a TCP connection to remote host specified
474 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
475 * port number if needed. Call in_pcbladdr to do the routing and to choose
476 * a local host address (interface). If there is an existing incarnation
477 * of the same connection in TIME-WAIT state and if the remote host was
478 * sending CC options and if the connection duration was < MSL, then
479 * truncate the previous TIME-WAIT state and proceed.
480 * Initialize connection parameters and enter SYN-SENT state.
481 */
482static int
483tcp_connect(tp, nam, p)
484 register struct tcpcb *tp;
485 struct sockaddr *nam;
486 struct proc *p;
487{
488 struct inpcb *inp = tp->t_inpcb, *oinp;
489 struct socket *so = inp->inp_socket;
490 struct tcpcb *otp;
491 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
492 struct sockaddr_in *ifaddr;
493 struct rmxp_tao *taop;
494 struct rmxp_tao tao_noncached;
495 int error;
496
497 if (inp->inp_lport == 0) {
498 error = in_pcbbind(inp, (struct sockaddr *)0, p);
499 if (error)
500 return error;
501 }
502
503 /*
504 * Cannot simply call in_pcbconnect, because there might be an
505 * earlier incarnation of this same connection still in
506 * TIME_WAIT state, creating an ADDRINUSE error.
507 */
508 error = in_pcbladdr(inp, nam, &ifaddr);
509 if (error)
510 return error;
511 oinp = in_pcblookup_hash(inp->inp_pcbinfo,
512 sin->sin_addr, sin->sin_port,
513 inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr
514 : ifaddr->sin_addr,
515 inp->inp_lport, 0);
516 if (oinp) {
517 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
518 otp->t_state == TCPS_TIME_WAIT &&
519 otp->t_duration < TCPTV_MSL &&
520 (otp->t_flags & TF_RCVD_CC))
521 otp = tcp_close(otp);
522 else
523 return EADDRINUSE;
524 }
525 if (inp->inp_laddr.s_addr == INADDR_ANY)
526 inp->inp_laddr = ifaddr->sin_addr;
527 inp->inp_faddr = sin->sin_addr;
528 inp->inp_fport = sin->sin_port;
529 in_pcbrehash(inp);
530
531 tp->t_template = tcp_template(tp);
532 if (tp->t_template == 0) {
533 in_pcbdisconnect(inp);
534 return ENOBUFS;
535 }
536
537 /* Compute window scaling to request. */
538 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
539 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
540 tp->request_r_scale++;
541
542 soisconnecting(so);
543 tcpstat.tcps_connattempt++;
544 tp->t_state = TCPS_SYN_SENT;
545 tp->t_timer[TCPT_KEEP] = tcp_keepinit;
546 tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
547 tcp_sendseqinit(tp);
548
549 /*
550 * Generate a CC value for this connection and
551 * check whether CC or CCnew should be used.
552 */
553 if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) {
554 taop = &tao_noncached;
555 bzero(taop, sizeof(*taop));
556 }
557
558 tp->cc_send = CC_INC(tcp_ccgen);
559 if (taop->tao_ccsent != 0 &&
560 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
561 taop->tao_ccsent = tp->cc_send;
562 } else {
563 taop->tao_ccsent = 0;
564 tp->t_flags |= TF_SENDCCNEW;
565 }
566
567 return 0;
568}
569
570/*
571 * The new sockopt interface makes it possible for us to block in the
572 * copyin/out step (if we take a page fault). Taking a page fault at
573 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now
574 * use TSM, there probably isn't any need for this function to run at
575 * splnet() any more. This needs more examination.)
576 */
577int
578tcp_ctloutput(so, sopt)
579 struct socket *so;
580 struct sockopt *sopt;
581{
582 int error, opt, optval, s;
583 struct inpcb *inp;
584 struct tcpcb *tp;
585
586 error = 0;
587 s = splnet(); /* XXX */
588 inp = sotoinpcb(so);
589 if (inp == NULL) {
590 splx(s);
591 return (ECONNRESET);
592 }
593 if (sopt->sopt_level != IPPROTO_TCP) {
594 error = ip_ctloutput(so, sopt);
595 splx(s);
596 return (error);
597 }
598 tp = intotcpcb(inp);
599
600 switch (sopt->sopt_dir) {
601 case SOPT_SET:
602 switch (sopt->sopt_name) {
603 case TCP_NODELAY:
604 case TCP_NOOPT:
605 case TCP_NOPUSH:
606 error = sooptcopyin(sopt, &optval, sizeof optval,
607 sizeof optval);
608 if (error)
609 break;
610
611 switch (sopt->sopt_name) {
612 case TCP_NODELAY:
613 opt = TF_NODELAY;
614 break;
615 case TCP_NOOPT:
616 opt = TF_NOOPT;
617 break;
618 case TCP_NOPUSH:
619 opt = TF_NOPUSH;
620 break;
621 default:
622 opt = 0; /* dead code to fool gcc */
623 break;
624 }
625
626 if (optval)
627 tp->t_flags |= opt;
628 else
629 tp->t_flags &= ~opt;
630 break;
631
632 case TCP_MAXSEG:
633 error = sooptcopyin(sopt, &optval, sizeof optval,
634 sizeof optval);
635 if (error)
636 break;
637
638 if (optval > 0 && optval <= tp->t_maxseg)
639 tp->t_maxseg = optval;
640 else
641 error = EINVAL;
642 break;
643
644 default:
645 error = ENOPROTOOPT;
646 break;
647 }
648 break;
649
650 case SOPT_GET:
651 switch (sopt->sopt_name) {
652 case TCP_NODELAY:
653 optval = tp->t_flags & TF_NODELAY;
654 break;
655 case TCP_MAXSEG:
656 optval = tp->t_maxseg;
657 break;
658 case TCP_NOOPT:
659 optval = tp->t_flags & TF_NOOPT;
660 break;
661 case TCP_NOPUSH:
662 optval = tp->t_flags & TF_NOPUSH;
663 break;
664 default:
665 error = ENOPROTOOPT;
666 break;
667 }
668 if (error == 0)
669 error = sooptcopyout(sopt, &optval, sizeof optval);
670 break;
671 }
672 splx(s);
673 return (error);
674}
675
676/*
677 * tcp_sendspace and tcp_recvspace are the default send and receive window
678 * sizes, respectively. These are obsolescent (this information should
679 * be set by the route).
680 */
681u_long tcp_sendspace = 1024*16;
| 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;
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 prison_remote_ip(p, 0, &sinp->sin_addr.s_addr);
243
244 if ((error = tcp_connect(tp, nam, p)) != 0)
245 goto out;
246 error = tcp_output(tp);
247 COMMON_END(PRU_CONNECT);
248}
249
250/*
251 * Initiate disconnect from peer.
252 * If connection never passed embryonic stage, just drop;
253 * else if don't need to let data drain, then can just drop anyways,
254 * else have to begin TCP shutdown process: mark socket disconnecting,
255 * drain unread data, state switch to reflect user close, and
256 * send segment (e.g. FIN) to peer. Socket will be really disconnected
257 * when peer sends FIN and acks ours.
258 *
259 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
260 */
261static int
262tcp_usr_disconnect(struct socket *so)
263{
264 int s = splnet();
265 int error = 0;
266 struct inpcb *inp = sotoinpcb(so);
267 struct tcpcb *tp;
268
269 COMMON_START();
270 tp = tcp_disconnect(tp);
271 COMMON_END(PRU_DISCONNECT);
272}
273
274/*
275 * Accept a connection. Essentially all the work is
276 * done at higher levels; just return the address
277 * of the peer, storing through addr.
278 */
279static int
280tcp_usr_accept(struct socket *so, struct sockaddr **nam)
281{
282 int s = splnet();
283 int error = 0;
284 struct inpcb *inp = sotoinpcb(so);
285 struct tcpcb *tp;
286
287 COMMON_START();
288 in_setpeeraddr(so, nam);
289 COMMON_END(PRU_ACCEPT);
290}
291
292/*
293 * Mark the connection as being incapable of further output.
294 */
295static int
296tcp_usr_shutdown(struct socket *so)
297{
298 int s = splnet();
299 int error = 0;
300 struct inpcb *inp = sotoinpcb(so);
301 struct tcpcb *tp;
302
303 COMMON_START();
304 socantsendmore(so);
305 tp = tcp_usrclosed(tp);
306 if (tp)
307 error = tcp_output(tp);
308 COMMON_END(PRU_SHUTDOWN);
309}
310
311/*
312 * After a receive, possibly send window update to peer.
313 */
314static int
315tcp_usr_rcvd(struct socket *so, int flags)
316{
317 int s = splnet();
318 int error = 0;
319 struct inpcb *inp = sotoinpcb(so);
320 struct tcpcb *tp;
321
322 COMMON_START();
323 tcp_output(tp);
324 COMMON_END(PRU_RCVD);
325}
326
327/*
328 * Do a send by putting data in output queue and updating urgent
329 * marker if URG set. Possibly send more data.
330 */
331static int
332tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
333 struct sockaddr *nam, struct mbuf *control, struct proc *p)
334{
335 int s = splnet();
336 int error = 0;
337 struct inpcb *inp = sotoinpcb(so);
338 struct tcpcb *tp;
339
340 COMMON_START();
341 if (control && control->m_len) {
342 m_freem(control); /* XXX shouldn't caller do this??? */
343 if (m)
344 m_freem(m);
345 error = EINVAL;
346 goto out;
347 }
348
349 if(!(flags & PRUS_OOB)) {
350 sbappend(&so->so_snd, m);
351 if (nam && tp->t_state < TCPS_SYN_SENT) {
352 /*
353 * Do implied connect if not yet connected,
354 * initialize window to default value, and
355 * initialize maxseg/maxopd using peer's cached
356 * MSS.
357 */
358 error = tcp_connect(tp, nam, p);
359 if (error)
360 goto out;
361 tp->snd_wnd = TTCP_CLIENT_SND_WND;
362 tcp_mss(tp, -1);
363 }
364
365 if (flags & PRUS_EOF) {
366 /*
367 * Close the send side of the connection after
368 * the data is sent.
369 */
370 socantsendmore(so);
371 tp = tcp_usrclosed(tp);
372 }
373 if (tp != NULL) {
374 if (flags & PRUS_MORETOCOME)
375 tp->t_flags |= TF_MORETOCOME;
376 error = tcp_output(tp);
377 if (flags & PRUS_MORETOCOME)
378 tp->t_flags &= ~TF_MORETOCOME;
379 }
380 } else {
381 if (sbspace(&so->so_snd) < -512) {
382 m_freem(m);
383 error = ENOBUFS;
384 goto out;
385 }
386 /*
387 * According to RFC961 (Assigned Protocols),
388 * the urgent pointer points to the last octet
389 * of urgent data. We continue, however,
390 * to consider it to indicate the first octet
391 * of data past the urgent section.
392 * Otherwise, snd_up should be one lower.
393 */
394 sbappend(&so->so_snd, m);
395 if (nam && tp->t_state < TCPS_SYN_SENT) {
396 /*
397 * Do implied connect if not yet connected,
398 * initialize window to default value, and
399 * initialize maxseg/maxopd using peer's cached
400 * MSS.
401 */
402 error = tcp_connect(tp, nam, p);
403 if (error)
404 goto out;
405 tp->snd_wnd = TTCP_CLIENT_SND_WND;
406 tcp_mss(tp, -1);
407 }
408 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
409 tp->t_force = 1;
410 error = tcp_output(tp);
411 tp->t_force = 0;
412 }
413 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
414 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
415}
416
417/*
418 * Abort the TCP.
419 */
420static int
421tcp_usr_abort(struct socket *so)
422{
423 int s = splnet();
424 int error = 0;
425 struct inpcb *inp = sotoinpcb(so);
426 struct tcpcb *tp;
427
428 COMMON_START();
429 tp = tcp_drop(tp, ECONNABORTED);
430 COMMON_END(PRU_ABORT);
431}
432
433/*
434 * Receive out-of-band data.
435 */
436static int
437tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
438{
439 int s = splnet();
440 int error = 0;
441 struct inpcb *inp = sotoinpcb(so);
442 struct tcpcb *tp;
443
444 COMMON_START();
445 if ((so->so_oobmark == 0 &&
446 (so->so_state & SS_RCVATMARK) == 0) ||
447 so->so_options & SO_OOBINLINE ||
448 tp->t_oobflags & TCPOOB_HADDATA) {
449 error = EINVAL;
450 goto out;
451 }
452 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
453 error = EWOULDBLOCK;
454 goto out;
455 }
456 m->m_len = 1;
457 *mtod(m, caddr_t) = tp->t_iobc;
458 if ((flags & MSG_PEEK) == 0)
459 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
460 COMMON_END(PRU_RCVOOB);
461}
462
463/* xxx - should be const */
464struct pr_usrreqs tcp_usrreqs = {
465 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind,
466 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach,
467 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd,
468 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
469 in_setsockaddr, sosend, soreceive, sopoll
470};
471
472/*
473 * Common subroutine to open a TCP connection to remote host specified
474 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
475 * port number if needed. Call in_pcbladdr to do the routing and to choose
476 * a local host address (interface). If there is an existing incarnation
477 * of the same connection in TIME-WAIT state and if the remote host was
478 * sending CC options and if the connection duration was < MSL, then
479 * truncate the previous TIME-WAIT state and proceed.
480 * Initialize connection parameters and enter SYN-SENT state.
481 */
482static int
483tcp_connect(tp, nam, p)
484 register struct tcpcb *tp;
485 struct sockaddr *nam;
486 struct proc *p;
487{
488 struct inpcb *inp = tp->t_inpcb, *oinp;
489 struct socket *so = inp->inp_socket;
490 struct tcpcb *otp;
491 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
492 struct sockaddr_in *ifaddr;
493 struct rmxp_tao *taop;
494 struct rmxp_tao tao_noncached;
495 int error;
496
497 if (inp->inp_lport == 0) {
498 error = in_pcbbind(inp, (struct sockaddr *)0, p);
499 if (error)
500 return error;
501 }
502
503 /*
504 * Cannot simply call in_pcbconnect, because there might be an
505 * earlier incarnation of this same connection still in
506 * TIME_WAIT state, creating an ADDRINUSE error.
507 */
508 error = in_pcbladdr(inp, nam, &ifaddr);
509 if (error)
510 return error;
511 oinp = in_pcblookup_hash(inp->inp_pcbinfo,
512 sin->sin_addr, sin->sin_port,
513 inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr
514 : ifaddr->sin_addr,
515 inp->inp_lport, 0);
516 if (oinp) {
517 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
518 otp->t_state == TCPS_TIME_WAIT &&
519 otp->t_duration < TCPTV_MSL &&
520 (otp->t_flags & TF_RCVD_CC))
521 otp = tcp_close(otp);
522 else
523 return EADDRINUSE;
524 }
525 if (inp->inp_laddr.s_addr == INADDR_ANY)
526 inp->inp_laddr = ifaddr->sin_addr;
527 inp->inp_faddr = sin->sin_addr;
528 inp->inp_fport = sin->sin_port;
529 in_pcbrehash(inp);
530
531 tp->t_template = tcp_template(tp);
532 if (tp->t_template == 0) {
533 in_pcbdisconnect(inp);
534 return ENOBUFS;
535 }
536
537 /* Compute window scaling to request. */
538 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
539 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
540 tp->request_r_scale++;
541
542 soisconnecting(so);
543 tcpstat.tcps_connattempt++;
544 tp->t_state = TCPS_SYN_SENT;
545 tp->t_timer[TCPT_KEEP] = tcp_keepinit;
546 tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
547 tcp_sendseqinit(tp);
548
549 /*
550 * Generate a CC value for this connection and
551 * check whether CC or CCnew should be used.
552 */
553 if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) {
554 taop = &tao_noncached;
555 bzero(taop, sizeof(*taop));
556 }
557
558 tp->cc_send = CC_INC(tcp_ccgen);
559 if (taop->tao_ccsent != 0 &&
560 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
561 taop->tao_ccsent = tp->cc_send;
562 } else {
563 taop->tao_ccsent = 0;
564 tp->t_flags |= TF_SENDCCNEW;
565 }
566
567 return 0;
568}
569
570/*
571 * The new sockopt interface makes it possible for us to block in the
572 * copyin/out step (if we take a page fault). Taking a page fault at
573 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now
574 * use TSM, there probably isn't any need for this function to run at
575 * splnet() any more. This needs more examination.)
576 */
577int
578tcp_ctloutput(so, sopt)
579 struct socket *so;
580 struct sockopt *sopt;
581{
582 int error, opt, optval, s;
583 struct inpcb *inp;
584 struct tcpcb *tp;
585
586 error = 0;
587 s = splnet(); /* XXX */
588 inp = sotoinpcb(so);
589 if (inp == NULL) {
590 splx(s);
591 return (ECONNRESET);
592 }
593 if (sopt->sopt_level != IPPROTO_TCP) {
594 error = ip_ctloutput(so, sopt);
595 splx(s);
596 return (error);
597 }
598 tp = intotcpcb(inp);
599
600 switch (sopt->sopt_dir) {
601 case SOPT_SET:
602 switch (sopt->sopt_name) {
603 case TCP_NODELAY:
604 case TCP_NOOPT:
605 case TCP_NOPUSH:
606 error = sooptcopyin(sopt, &optval, sizeof optval,
607 sizeof optval);
608 if (error)
609 break;
610
611 switch (sopt->sopt_name) {
612 case TCP_NODELAY:
613 opt = TF_NODELAY;
614 break;
615 case TCP_NOOPT:
616 opt = TF_NOOPT;
617 break;
618 case TCP_NOPUSH:
619 opt = TF_NOPUSH;
620 break;
621 default:
622 opt = 0; /* dead code to fool gcc */
623 break;
624 }
625
626 if (optval)
627 tp->t_flags |= opt;
628 else
629 tp->t_flags &= ~opt;
630 break;
631
632 case TCP_MAXSEG:
633 error = sooptcopyin(sopt, &optval, sizeof optval,
634 sizeof optval);
635 if (error)
636 break;
637
638 if (optval > 0 && optval <= tp->t_maxseg)
639 tp->t_maxseg = optval;
640 else
641 error = EINVAL;
642 break;
643
644 default:
645 error = ENOPROTOOPT;
646 break;
647 }
648 break;
649
650 case SOPT_GET:
651 switch (sopt->sopt_name) {
652 case TCP_NODELAY:
653 optval = tp->t_flags & TF_NODELAY;
654 break;
655 case TCP_MAXSEG:
656 optval = tp->t_maxseg;
657 break;
658 case TCP_NOOPT:
659 optval = tp->t_flags & TF_NOOPT;
660 break;
661 case TCP_NOPUSH:
662 optval = tp->t_flags & TF_NOPUSH;
663 break;
664 default:
665 error = ENOPROTOOPT;
666 break;
667 }
668 if (error == 0)
669 error = sooptcopyout(sopt, &optval, sizeof optval);
670 break;
671 }
672 splx(s);
673 return (error);
674}
675
676/*
677 * tcp_sendspace and tcp_recvspace are the default send and receive window
678 * sizes, respectively. These are obsolescent (this information should
679 * be set by the route).
680 */
681u_long tcp_sendspace = 1024*16;
|
687
688/*
689 * Attach TCP protocol to socket, allocating
690 * internet protocol control block, tcp control block,
691 * bufer space, and entering LISTEN state if to accept connections.
692 */
693static int
694tcp_attach(so, p)
695 struct socket *so;
696 struct proc *p;
697{
698 register struct tcpcb *tp;
699 struct inpcb *inp;
700 int error;
701
702 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
703 error = soreserve(so, tcp_sendspace, tcp_recvspace);
704 if (error)
705 return (error);
706 }
707 error = in_pcballoc(so, &tcbinfo, p);
708 if (error)
709 return (error);
710 inp = sotoinpcb(so);
711 tp = tcp_newtcpcb(inp);
712 if (tp == 0) {
713 int nofd = so->so_state & SS_NOFDREF; /* XXX */
714
715 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */
716 in_pcbdetach(inp);
717 so->so_state |= nofd;
718 return (ENOBUFS);
719 }
720 tp->t_state = TCPS_CLOSED;
721 return (0);
722}
723
724/*
725 * Initiate (or continue) disconnect.
726 * If embryonic state, just send reset (once).
727 * If in ``let data drain'' option and linger null, just drop.
728 * Otherwise (hard), mark socket disconnecting and drop
729 * current input data; switch states based on user close, and
730 * send segment to peer (with FIN).
731 */
732static struct tcpcb *
733tcp_disconnect(tp)
734 register struct tcpcb *tp;
735{
736 struct socket *so = tp->t_inpcb->inp_socket;
737
738 if (tp->t_state < TCPS_ESTABLISHED)
739 tp = tcp_close(tp);
740 else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
741 tp = tcp_drop(tp, 0);
742 else {
743 soisdisconnecting(so);
744 sbflush(&so->so_rcv);
745 tp = tcp_usrclosed(tp);
746 if (tp)
747 (void) tcp_output(tp);
748 }
749 return (tp);
750}
751
752/*
753 * User issued close, and wish to trail through shutdown states:
754 * if never received SYN, just forget it. If got a SYN from peer,
755 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
756 * If already got a FIN from peer, then almost done; go to LAST_ACK
757 * state. In all other cases, have already sent FIN to peer (e.g.
758 * after PRU_SHUTDOWN), and just have to play tedious game waiting
759 * for peer to send FIN or not respond to keep-alives, etc.
760 * We can let the user exit from the close as soon as the FIN is acked.
761 */
762static struct tcpcb *
763tcp_usrclosed(tp)
764 register struct tcpcb *tp;
765{
766
767 switch (tp->t_state) {
768
769 case TCPS_CLOSED:
770 case TCPS_LISTEN:
771 tp->t_state = TCPS_CLOSED;
772 tp = tcp_close(tp);
773 break;
774
775 case TCPS_SYN_SENT:
776 case TCPS_SYN_RECEIVED:
777 tp->t_flags |= TF_NEEDFIN;
778 break;
779
780 case TCPS_ESTABLISHED:
781 tp->t_state = TCPS_FIN_WAIT_1;
782 break;
783
784 case TCPS_CLOSE_WAIT:
785 tp->t_state = TCPS_LAST_ACK;
786 break;
787 }
788 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
789 soisdisconnected(tp->t_inpcb->inp_socket);
790 /* To prevent the connection hanging in FIN_WAIT_2 forever. */
791 if (tp->t_state == TCPS_FIN_WAIT_2)
792 tp->t_timer[TCPT_2MSL] = tcp_maxidle;
793 }
794 return (tp);
795}
796
| 687
688/*
689 * Attach TCP protocol to socket, allocating
690 * internet protocol control block, tcp control block,
691 * bufer space, and entering LISTEN state if to accept connections.
692 */
693static int
694tcp_attach(so, p)
695 struct socket *so;
696 struct proc *p;
697{
698 register struct tcpcb *tp;
699 struct inpcb *inp;
700 int error;
701
702 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
703 error = soreserve(so, tcp_sendspace, tcp_recvspace);
704 if (error)
705 return (error);
706 }
707 error = in_pcballoc(so, &tcbinfo, p);
708 if (error)
709 return (error);
710 inp = sotoinpcb(so);
711 tp = tcp_newtcpcb(inp);
712 if (tp == 0) {
713 int nofd = so->so_state & SS_NOFDREF; /* XXX */
714
715 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */
716 in_pcbdetach(inp);
717 so->so_state |= nofd;
718 return (ENOBUFS);
719 }
720 tp->t_state = TCPS_CLOSED;
721 return (0);
722}
723
724/*
725 * Initiate (or continue) disconnect.
726 * If embryonic state, just send reset (once).
727 * If in ``let data drain'' option and linger null, just drop.
728 * Otherwise (hard), mark socket disconnecting and drop
729 * current input data; switch states based on user close, and
730 * send segment to peer (with FIN).
731 */
732static struct tcpcb *
733tcp_disconnect(tp)
734 register struct tcpcb *tp;
735{
736 struct socket *so = tp->t_inpcb->inp_socket;
737
738 if (tp->t_state < TCPS_ESTABLISHED)
739 tp = tcp_close(tp);
740 else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
741 tp = tcp_drop(tp, 0);
742 else {
743 soisdisconnecting(so);
744 sbflush(&so->so_rcv);
745 tp = tcp_usrclosed(tp);
746 if (tp)
747 (void) tcp_output(tp);
748 }
749 return (tp);
750}
751
752/*
753 * User issued close, and wish to trail through shutdown states:
754 * if never received SYN, just forget it. If got a SYN from peer,
755 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
756 * If already got a FIN from peer, then almost done; go to LAST_ACK
757 * state. In all other cases, have already sent FIN to peer (e.g.
758 * after PRU_SHUTDOWN), and just have to play tedious game waiting
759 * for peer to send FIN or not respond to keep-alives, etc.
760 * We can let the user exit from the close as soon as the FIN is acked.
761 */
762static struct tcpcb *
763tcp_usrclosed(tp)
764 register struct tcpcb *tp;
765{
766
767 switch (tp->t_state) {
768
769 case TCPS_CLOSED:
770 case TCPS_LISTEN:
771 tp->t_state = TCPS_CLOSED;
772 tp = tcp_close(tp);
773 break;
774
775 case TCPS_SYN_SENT:
776 case TCPS_SYN_RECEIVED:
777 tp->t_flags |= TF_NEEDFIN;
778 break;
779
780 case TCPS_ESTABLISHED:
781 tp->t_state = TCPS_FIN_WAIT_1;
782 break;
783
784 case TCPS_CLOSE_WAIT:
785 tp->t_state = TCPS_LAST_ACK;
786 break;
787 }
788 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
789 soisdisconnected(tp->t_inpcb->inp_socket);
790 /* To prevent the connection hanging in FIN_WAIT_2 forever. */
791 if (tp->t_state == TCPS_FIN_WAIT_2)
792 tp->t_timer[TCPT_2MSL] = tcp_maxidle;
793 }
794 return (tp);
795}
796
|