Initial import from FreeBSD RELENG_4:
[dragonfly.git] / sys / netinet / tcp_usrreq.c
CommitLineData
984263bc
MD
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: src/sys/netinet/tcp_usrreq.c,v 1.51.2.17 2002/10/11 11:46:44 ume Exp $
35 */
36
37#include "opt_ipsec.h"
38#include "opt_inet6.h"
39#include "opt_tcpdebug.h"
40
41#include <sys/param.h>
42#include <sys/systm.h>
43#include <sys/kernel.h>
44#include <sys/sysctl.h>
45#include <sys/mbuf.h>
46#ifdef INET6
47#include <sys/domain.h>
48#endif /* INET6 */
49#include <sys/socket.h>
50#include <sys/socketvar.h>
51#include <sys/protosw.h>
52
53#include <net/if.h>
54#include <net/route.h>
55
56#include <netinet/in.h>
57#include <netinet/in_systm.h>
58#ifdef INET6
59#include <netinet/ip6.h>
60#endif
61#include <netinet/in_pcb.h>
62#ifdef INET6
63#include <netinet6/in6_pcb.h>
64#endif
65#include <netinet/in_var.h>
66#include <netinet/ip_var.h>
67#ifdef INET6
68#include <netinet6/ip6_var.h>
69#endif
70#include <netinet/tcp.h>
71#include <netinet/tcp_fsm.h>
72#include <netinet/tcp_seq.h>
73#include <netinet/tcp_timer.h>
74#include <netinet/tcp_var.h>
75#include <netinet/tcpip.h>
76#ifdef TCPDEBUG
77#include <netinet/tcp_debug.h>
78#endif
79
80#ifdef IPSEC
81#include <netinet6/ipsec.h>
82#endif /*IPSEC*/
83
84/*
85 * TCP protocol interface to socket abstraction.
86 */
87extern char *tcpstates[]; /* XXX ??? */
88
89static int tcp_attach __P((struct socket *, struct proc *));
90static int tcp_connect __P((struct tcpcb *, struct sockaddr *,
91 struct proc *));
92#ifdef INET6
93static int tcp6_connect __P((struct tcpcb *, struct sockaddr *,
94 struct proc *));
95#endif /* INET6 */
96static struct tcpcb *
97 tcp_disconnect __P((struct tcpcb *));
98static struct tcpcb *
99 tcp_usrclosed __P((struct tcpcb *));
100
101#ifdef TCPDEBUG
102#define TCPDEBUG0 int ostate = 0
103#define TCPDEBUG1() ostate = tp ? tp->t_state : 0
104#define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
105 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
106#else
107#define TCPDEBUG0
108#define TCPDEBUG1()
109#define TCPDEBUG2(req)
110#endif
111
112/*
113 * TCP attaches to socket via pru_attach(), reserving space,
114 * and an internet control block.
115 */
116static int
117tcp_usr_attach(struct socket *so, int proto, struct proc *p)
118{
119 int s = splnet();
120 int error;
121 struct inpcb *inp = sotoinpcb(so);
122 struct tcpcb *tp = 0;
123 TCPDEBUG0;
124
125 TCPDEBUG1();
126 if (inp) {
127 error = EISCONN;
128 goto out;
129 }
130
131 error = tcp_attach(so, p);
132 if (error)
133 goto out;
134
135 if ((so->so_options & SO_LINGER) && so->so_linger == 0)
136 so->so_linger = TCP_LINGERTIME;
137 tp = sototcpcb(so);
138out:
139 TCPDEBUG2(PRU_ATTACH);
140 splx(s);
141 return error;
142}
143
144/*
145 * pru_detach() detaches the TCP protocol from the socket.
146 * If the protocol state is non-embryonic, then can't
147 * do this directly: have to initiate a pru_disconnect(),
148 * which may finish later; embryonic TCB's can just
149 * be discarded here.
150 */
151static int
152tcp_usr_detach(struct socket *so)
153{
154 int s = splnet();
155 int error = 0;
156 struct inpcb *inp = sotoinpcb(so);
157 struct tcpcb *tp;
158 TCPDEBUG0;
159
160 if (inp == 0) {
161 splx(s);
162 return EINVAL; /* XXX */
163 }
164 tp = intotcpcb(inp);
165 TCPDEBUG1();
166 tp = tcp_disconnect(tp);
167
168 TCPDEBUG2(PRU_DETACH);
169 splx(s);
170 return error;
171}
172
173#define COMMON_START() TCPDEBUG0; \
174 do { \
175 if (inp == 0) { \
176 splx(s); \
177 return EINVAL; \
178 } \
179 tp = intotcpcb(inp); \
180 TCPDEBUG1(); \
181 } while(0)
182
183#define COMMON_END(req) out: TCPDEBUG2(req); splx(s); return error; goto out
184
185
186/*
187 * Give the socket an address.
188 */
189static int
190tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
191{
192 int s = splnet();
193 int error = 0;
194 struct inpcb *inp = sotoinpcb(so);
195 struct tcpcb *tp;
196 struct sockaddr_in *sinp;
197
198 COMMON_START();
199
200 /*
201 * Must check for multicast addresses and disallow binding
202 * to them.
203 */
204 sinp = (struct sockaddr_in *)nam;
205 if (sinp->sin_family == AF_INET &&
206 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
207 error = EAFNOSUPPORT;
208 goto out;
209 }
210 error = in_pcbbind(inp, nam, p);
211 if (error)
212 goto out;
213 COMMON_END(PRU_BIND);
214
215}
216
217#ifdef INET6
218static int
219tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
220{
221 int s = splnet();
222 int error = 0;
223 struct inpcb *inp = sotoinpcb(so);
224 struct tcpcb *tp;
225 struct sockaddr_in6 *sin6p;
226
227 COMMON_START();
228
229 /*
230 * Must check for multicast addresses and disallow binding
231 * to them.
232 */
233 sin6p = (struct sockaddr_in6 *)nam;
234 if (sin6p->sin6_family == AF_INET6 &&
235 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
236 error = EAFNOSUPPORT;
237 goto out;
238 }
239 inp->inp_vflag &= ~INP_IPV4;
240 inp->inp_vflag |= INP_IPV6;
241 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
242 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
243 inp->inp_vflag |= INP_IPV4;
244 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
245 struct sockaddr_in sin;
246
247 in6_sin6_2_sin(&sin, sin6p);
248 inp->inp_vflag |= INP_IPV4;
249 inp->inp_vflag &= ~INP_IPV6;
250 error = in_pcbbind(inp, (struct sockaddr *)&sin, p);
251 goto out;
252 }
253 }
254 error = in6_pcbbind(inp, nam, p);
255 if (error)
256 goto out;
257 COMMON_END(PRU_BIND);
258}
259#endif /* INET6 */
260
261/*
262 * Prepare to accept connections.
263 */
264static int
265tcp_usr_listen(struct socket *so, struct proc *p)
266{
267 int s = splnet();
268 int error = 0;
269 struct inpcb *inp = sotoinpcb(so);
270 struct tcpcb *tp;
271
272 COMMON_START();
273 if (inp->inp_lport == 0)
274 error = in_pcbbind(inp, (struct sockaddr *)0, p);
275 if (error == 0)
276 tp->t_state = TCPS_LISTEN;
277 COMMON_END(PRU_LISTEN);
278}
279
280#ifdef INET6
281static int
282tcp6_usr_listen(struct socket *so, struct proc *p)
283{
284 int s = splnet();
285 int error = 0;
286 struct inpcb *inp = sotoinpcb(so);
287 struct tcpcb *tp;
288
289 COMMON_START();
290 if (inp->inp_lport == 0) {
291 inp->inp_vflag &= ~INP_IPV4;
292 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
293 inp->inp_vflag |= INP_IPV4;
294 error = in6_pcbbind(inp, (struct sockaddr *)0, p);
295 }
296 if (error == 0)
297 tp->t_state = TCPS_LISTEN;
298 COMMON_END(PRU_LISTEN);
299}
300#endif /* INET6 */
301
302/*
303 * Initiate connection to peer.
304 * Create a template for use in transmissions on this connection.
305 * Enter SYN_SENT state, and mark socket as connecting.
306 * Start keep-alive timer, and seed output sequence space.
307 * Send initial segment on connection.
308 */
309static int
310tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
311{
312 int s = splnet();
313 int error = 0;
314 struct inpcb *inp = sotoinpcb(so);
315 struct tcpcb *tp;
316 struct sockaddr_in *sinp;
317
318 COMMON_START();
319
320 /*
321 * Must disallow TCP ``connections'' to multicast addresses.
322 */
323 sinp = (struct sockaddr_in *)nam;
324 if (sinp->sin_family == AF_INET
325 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
326 error = EAFNOSUPPORT;
327 goto out;
328 }
329
330 prison_remote_ip(p, 0, &sinp->sin_addr.s_addr);
331
332 if ((error = tcp_connect(tp, nam, p)) != 0)
333 goto out;
334 error = tcp_output(tp);
335 COMMON_END(PRU_CONNECT);
336}
337
338#ifdef INET6
339static int
340tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
341{
342 int s = splnet();
343 int error = 0;
344 struct inpcb *inp = sotoinpcb(so);
345 struct tcpcb *tp;
346 struct sockaddr_in6 *sin6p;
347
348 COMMON_START();
349
350 /*
351 * Must disallow TCP ``connections'' to multicast addresses.
352 */
353 sin6p = (struct sockaddr_in6 *)nam;
354 if (sin6p->sin6_family == AF_INET6
355 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
356 error = EAFNOSUPPORT;
357 goto out;
358 }
359
360 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
361 struct sockaddr_in sin;
362
363 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
364 error = EINVAL;
365 goto out;
366 }
367
368 in6_sin6_2_sin(&sin, sin6p);
369 inp->inp_vflag |= INP_IPV4;
370 inp->inp_vflag &= ~INP_IPV6;
371 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, p)) != 0)
372 goto out;
373 error = tcp_output(tp);
374 goto out;
375 }
376 inp->inp_vflag &= ~INP_IPV4;
377 inp->inp_vflag |= INP_IPV6;
378 inp->inp_inc.inc_isipv6 = 1;
379 if ((error = tcp6_connect(tp, nam, p)) != 0)
380 goto out;
381 error = tcp_output(tp);
382 COMMON_END(PRU_CONNECT);
383}
384#endif /* INET6 */
385
386/*
387 * Initiate disconnect from peer.
388 * If connection never passed embryonic stage, just drop;
389 * else if don't need to let data drain, then can just drop anyways,
390 * else have to begin TCP shutdown process: mark socket disconnecting,
391 * drain unread data, state switch to reflect user close, and
392 * send segment (e.g. FIN) to peer. Socket will be really disconnected
393 * when peer sends FIN and acks ours.
394 *
395 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
396 */
397static int
398tcp_usr_disconnect(struct socket *so)
399{
400 int s = splnet();
401 int error = 0;
402 struct inpcb *inp = sotoinpcb(so);
403 struct tcpcb *tp;
404
405 COMMON_START();
406 tp = tcp_disconnect(tp);
407 COMMON_END(PRU_DISCONNECT);
408}
409
410/*
411 * Accept a connection. Essentially all the work is
412 * done at higher levels; just return the address
413 * of the peer, storing through addr.
414 */
415static int
416tcp_usr_accept(struct socket *so, struct sockaddr **nam)
417{
418 int s = splnet();
419 int error = 0;
420 struct inpcb *inp = sotoinpcb(so);
421 struct tcpcb *tp = NULL;
422 TCPDEBUG0;
423
424 if (so->so_state & SS_ISDISCONNECTED) {
425 error = ECONNABORTED;
426 goto out;
427 }
428 if (inp == 0) {
429 splx(s);
430 return (EINVAL);
431 }
432 tp = intotcpcb(inp);
433 TCPDEBUG1();
434 in_setpeeraddr(so, nam);
435 COMMON_END(PRU_ACCEPT);
436}
437
438#ifdef INET6
439static int
440tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
441{
442 int s = splnet();
443 int error = 0;
444 struct inpcb *inp = sotoinpcb(so);
445 struct tcpcb *tp = NULL;
446 TCPDEBUG0;
447
448 if (so->so_state & SS_ISDISCONNECTED) {
449 error = ECONNABORTED;
450 goto out;
451 }
452 if (inp == 0) {
453 splx(s);
454 return (EINVAL);
455 }
456 tp = intotcpcb(inp);
457 TCPDEBUG1();
458 in6_mapped_peeraddr(so, nam);
459 COMMON_END(PRU_ACCEPT);
460}
461#endif /* INET6 */
462/*
463 * Mark the connection as being incapable of further output.
464 */
465static int
466tcp_usr_shutdown(struct socket *so)
467{
468 int s = splnet();
469 int error = 0;
470 struct inpcb *inp = sotoinpcb(so);
471 struct tcpcb *tp;
472
473 COMMON_START();
474 socantsendmore(so);
475 tp = tcp_usrclosed(tp);
476 if (tp)
477 error = tcp_output(tp);
478 COMMON_END(PRU_SHUTDOWN);
479}
480
481/*
482 * After a receive, possibly send window update to peer.
483 */
484static int
485tcp_usr_rcvd(struct socket *so, int flags)
486{
487 int s = splnet();
488 int error = 0;
489 struct inpcb *inp = sotoinpcb(so);
490 struct tcpcb *tp;
491
492 COMMON_START();
493 tcp_output(tp);
494 COMMON_END(PRU_RCVD);
495}
496
497/*
498 * Do a send by putting data in output queue and updating urgent
499 * marker if URG set. Possibly send more data. Unlike the other
500 * pru_*() routines, the mbuf chains are our responsibility. We
501 * must either enqueue them or free them. The other pru_* routines
502 * generally are caller-frees.
503 */
504static int
505tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
506 struct sockaddr *nam, struct mbuf *control, struct proc *p)
507{
508 int s = splnet();
509 int error = 0;
510 struct inpcb *inp = sotoinpcb(so);
511 struct tcpcb *tp;
512#ifdef INET6
513 int isipv6;
514#endif
515 TCPDEBUG0;
516
517 if (inp == NULL) {
518 /*
519 * OOPS! we lost a race, the TCP session got reset after
520 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a
521 * network interrupt in the non-splnet() section of sosend().
522 */
523 if (m)
524 m_freem(m);
525 if (control)
526 m_freem(control);
527 error = ECONNRESET; /* XXX EPIPE? */
528 tp = NULL;
529 TCPDEBUG1();
530 goto out;
531 }
532#ifdef INET6
533 isipv6 = nam && nam->sa_family == AF_INET6;
534#endif /* INET6 */
535 tp = intotcpcb(inp);
536 TCPDEBUG1();
537 if (control) {
538 /* TCP doesn't do control messages (rights, creds, etc) */
539 if (control->m_len) {
540 m_freem(control);
541 if (m)
542 m_freem(m);
543 error = EINVAL;
544 goto out;
545 }
546 m_freem(control); /* empty control, just free it */
547 }
548 if(!(flags & PRUS_OOB)) {
549 sbappend(&so->so_snd, m);
550 if (nam && tp->t_state < TCPS_SYN_SENT) {
551 /*
552 * Do implied connect if not yet connected,
553 * initialize window to default value, and
554 * initialize maxseg/maxopd using peer's cached
555 * MSS.
556 */
557#ifdef INET6
558 if (isipv6)
559 error = tcp6_connect(tp, nam, p);
560 else
561#endif /* INET6 */
562 error = tcp_connect(tp, nam, p);
563 if (error)
564 goto out;
565 tp->snd_wnd = TTCP_CLIENT_SND_WND;
566 tcp_mss(tp, -1);
567 }
568
569 if (flags & PRUS_EOF) {
570 /*
571 * Close the send side of the connection after
572 * the data is sent.
573 */
574 socantsendmore(so);
575 tp = tcp_usrclosed(tp);
576 }
577 if (tp != NULL) {
578 if (flags & PRUS_MORETOCOME)
579 tp->t_flags |= TF_MORETOCOME;
580 error = tcp_output(tp);
581 if (flags & PRUS_MORETOCOME)
582 tp->t_flags &= ~TF_MORETOCOME;
583 }
584 } else {
585 if (sbspace(&so->so_snd) < -512) {
586 m_freem(m);
587 error = ENOBUFS;
588 goto out;
589 }
590 /*
591 * According to RFC961 (Assigned Protocols),
592 * the urgent pointer points to the last octet
593 * of urgent data. We continue, however,
594 * to consider it to indicate the first octet
595 * of data past the urgent section.
596 * Otherwise, snd_up should be one lower.
597 */
598 sbappend(&so->so_snd, m);
599 if (nam && tp->t_state < TCPS_SYN_SENT) {
600 /*
601 * Do implied connect if not yet connected,
602 * initialize window to default value, and
603 * initialize maxseg/maxopd using peer's cached
604 * MSS.
605 */
606#ifdef INET6
607 if (isipv6)
608 error = tcp6_connect(tp, nam, p);
609 else
610#endif /* INET6 */
611 error = tcp_connect(tp, nam, p);
612 if (error)
613 goto out;
614 tp->snd_wnd = TTCP_CLIENT_SND_WND;
615 tcp_mss(tp, -1);
616 }
617 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
618 tp->t_force = 1;
619 error = tcp_output(tp);
620 tp->t_force = 0;
621 }
622 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
623 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
624}
625
626/*
627 * Abort the TCP.
628 */
629static int
630tcp_usr_abort(struct socket *so)
631{
632 int s = splnet();
633 int error = 0;
634 struct inpcb *inp = sotoinpcb(so);
635 struct tcpcb *tp;
636
637 COMMON_START();
638 tp = tcp_drop(tp, ECONNABORTED);
639 COMMON_END(PRU_ABORT);
640}
641
642/*
643 * Receive out-of-band data.
644 */
645static int
646tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
647{
648 int s = splnet();
649 int error = 0;
650 struct inpcb *inp = sotoinpcb(so);
651 struct tcpcb *tp;
652
653 COMMON_START();
654 if ((so->so_oobmark == 0 &&
655 (so->so_state & SS_RCVATMARK) == 0) ||
656 so->so_options & SO_OOBINLINE ||
657 tp->t_oobflags & TCPOOB_HADDATA) {
658 error = EINVAL;
659 goto out;
660 }
661 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
662 error = EWOULDBLOCK;
663 goto out;
664 }
665 m->m_len = 1;
666 *mtod(m, caddr_t) = tp->t_iobc;
667 if ((flags & MSG_PEEK) == 0)
668 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
669 COMMON_END(PRU_RCVOOB);
670}
671
672/* xxx - should be const */
673struct pr_usrreqs tcp_usrreqs = {
674 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind,
675 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach,
676 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd,
677 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
678 in_setsockaddr, sosend, soreceive, sopoll
679};
680
681#ifdef INET6
682struct pr_usrreqs tcp6_usrreqs = {
683 tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind,
684 tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach,
685 tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd,
686 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
687 in6_mapped_sockaddr, sosend, soreceive, sopoll
688};
689#endif /* INET6 */
690
691/*
692 * Common subroutine to open a TCP connection to remote host specified
693 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
694 * port number if needed. Call in_pcbladdr to do the routing and to choose
695 * a local host address (interface). If there is an existing incarnation
696 * of the same connection in TIME-WAIT state and if the remote host was
697 * sending CC options and if the connection duration was < MSL, then
698 * truncate the previous TIME-WAIT state and proceed.
699 * Initialize connection parameters and enter SYN-SENT state.
700 */
701static int
702tcp_connect(tp, nam, p)
703 register struct tcpcb *tp;
704 struct sockaddr *nam;
705 struct proc *p;
706{
707 struct inpcb *inp = tp->t_inpcb, *oinp;
708 struct socket *so = inp->inp_socket;
709 struct tcpcb *otp;
710 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
711 struct sockaddr_in *ifaddr;
712 struct rmxp_tao *taop;
713 struct rmxp_tao tao_noncached;
714 int error;
715
716 if (inp->inp_lport == 0) {
717 error = in_pcbbind(inp, (struct sockaddr *)0, p);
718 if (error)
719 return error;
720 }
721
722 /*
723 * Cannot simply call in_pcbconnect, because there might be an
724 * earlier incarnation of this same connection still in
725 * TIME_WAIT state, creating an ADDRINUSE error.
726 */
727 error = in_pcbladdr(inp, nam, &ifaddr);
728 if (error)
729 return error;
730 oinp = in_pcblookup_hash(inp->inp_pcbinfo,
731 sin->sin_addr, sin->sin_port,
732 inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr
733 : ifaddr->sin_addr,
734 inp->inp_lport, 0, NULL);
735 if (oinp) {
736 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
737 otp->t_state == TCPS_TIME_WAIT &&
738 (ticks - otp->t_starttime) < tcp_msl &&
739 (otp->t_flags & TF_RCVD_CC))
740 otp = tcp_close(otp);
741 else
742 return EADDRINUSE;
743 }
744 if (inp->inp_laddr.s_addr == INADDR_ANY)
745 inp->inp_laddr = ifaddr->sin_addr;
746 inp->inp_faddr = sin->sin_addr;
747 inp->inp_fport = sin->sin_port;
748 in_pcbrehash(inp);
749
750 /* Compute window scaling to request. */
751 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
752 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
753 tp->request_r_scale++;
754
755 soisconnecting(so);
756 tcpstat.tcps_connattempt++;
757 tp->t_state = TCPS_SYN_SENT;
758 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
759 tp->iss = tcp_new_isn(tp);
760 tp->t_bw_rtseq = tp->iss;
761 tcp_sendseqinit(tp);
762
763 /*
764 * Generate a CC value for this connection and
765 * check whether CC or CCnew should be used.
766 */
767 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
768 taop = &tao_noncached;
769 bzero(taop, sizeof(*taop));
770 }
771
772 tp->cc_send = CC_INC(tcp_ccgen);
773 if (taop->tao_ccsent != 0 &&
774 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
775 taop->tao_ccsent = tp->cc_send;
776 } else {
777 taop->tao_ccsent = 0;
778 tp->t_flags |= TF_SENDCCNEW;
779 }
780
781 return 0;
782}
783
784#ifdef INET6
785static int
786tcp6_connect(tp, nam, p)
787 register struct tcpcb *tp;
788 struct sockaddr *nam;
789 struct proc *p;
790{
791 struct inpcb *inp = tp->t_inpcb, *oinp;
792 struct socket *so = inp->inp_socket;
793 struct tcpcb *otp;
794 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
795 struct in6_addr *addr6;
796 struct rmxp_tao *taop;
797 struct rmxp_tao tao_noncached;
798 int error;
799
800 if (inp->inp_lport == 0) {
801 error = in6_pcbbind(inp, (struct sockaddr *)0, p);
802 if (error)
803 return error;
804 }
805
806 /*
807 * Cannot simply call in_pcbconnect, because there might be an
808 * earlier incarnation of this same connection still in
809 * TIME_WAIT state, creating an ADDRINUSE error.
810 */
811 error = in6_pcbladdr(inp, nam, &addr6);
812 if (error)
813 return error;
814 oinp = in6_pcblookup_hash(inp->inp_pcbinfo,
815 &sin6->sin6_addr, sin6->sin6_port,
816 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
817 ? addr6
818 : &inp->in6p_laddr,
819 inp->inp_lport, 0, NULL);
820 if (oinp) {
821 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
822 otp->t_state == TCPS_TIME_WAIT &&
823 (ticks - otp->t_starttime) < tcp_msl &&
824 (otp->t_flags & TF_RCVD_CC))
825 otp = tcp_close(otp);
826 else
827 return EADDRINUSE;
828 }
829 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
830 inp->in6p_laddr = *addr6;
831 inp->in6p_faddr = sin6->sin6_addr;
832 inp->inp_fport = sin6->sin6_port;
833 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL)
834 inp->in6p_flowinfo = sin6->sin6_flowinfo;
835 in_pcbrehash(inp);
836
837 /* Compute window scaling to request. */
838 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
839 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
840 tp->request_r_scale++;
841
842 soisconnecting(so);
843 tcpstat.tcps_connattempt++;
844 tp->t_state = TCPS_SYN_SENT;
845 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
846 tp->iss = tcp_new_isn(tp);
847 tp->t_bw_rtseq = tp->iss;
848 tcp_sendseqinit(tp);
849
850 /*
851 * Generate a CC value for this connection and
852 * check whether CC or CCnew should be used.
853 */
854 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
855 taop = &tao_noncached;
856 bzero(taop, sizeof(*taop));
857 }
858
859 tp->cc_send = CC_INC(tcp_ccgen);
860 if (taop->tao_ccsent != 0 &&
861 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
862 taop->tao_ccsent = tp->cc_send;
863 } else {
864 taop->tao_ccsent = 0;
865 tp->t_flags |= TF_SENDCCNEW;
866 }
867
868 return 0;
869}
870#endif /* INET6 */
871
872/*
873 * The new sockopt interface makes it possible for us to block in the
874 * copyin/out step (if we take a page fault). Taking a page fault at
875 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now
876 * use TSM, there probably isn't any need for this function to run at
877 * splnet() any more. This needs more examination.)
878 */
879int
880tcp_ctloutput(so, sopt)
881 struct socket *so;
882 struct sockopt *sopt;
883{
884 int error, opt, optval, s;
885 struct inpcb *inp;
886 struct tcpcb *tp;
887
888 error = 0;
889 s = splnet(); /* XXX */
890 inp = sotoinpcb(so);
891 if (inp == NULL) {
892 splx(s);
893 return (ECONNRESET);
894 }
895 if (sopt->sopt_level != IPPROTO_TCP) {
896#ifdef INET6
897 if (INP_CHECK_SOCKAF(so, AF_INET6))
898 error = ip6_ctloutput(so, sopt);
899 else
900#endif /* INET6 */
901 error = ip_ctloutput(so, sopt);
902 splx(s);
903 return (error);
904 }
905 tp = intotcpcb(inp);
906
907 switch (sopt->sopt_dir) {
908 case SOPT_SET:
909 switch (sopt->sopt_name) {
910 case TCP_NODELAY:
911 case TCP_NOOPT:
912 error = sooptcopyin(sopt, &optval, sizeof optval,
913 sizeof optval);
914 if (error)
915 break;
916
917 switch (sopt->sopt_name) {
918 case TCP_NODELAY:
919 opt = TF_NODELAY;
920 break;
921 case TCP_NOOPT:
922 opt = TF_NOOPT;
923 break;
924 default:
925 opt = 0; /* dead code to fool gcc */
926 break;
927 }
928
929 if (optval)
930 tp->t_flags |= opt;
931 else
932 tp->t_flags &= ~opt;
933 break;
934
935 case TCP_NOPUSH:
936 error = sooptcopyin(sopt, &optval, sizeof optval,
937 sizeof optval);
938 if (error)
939 break;
940
941 if (optval)
942 tp->t_flags |= TF_NOPUSH;
943 else {
944 tp->t_flags &= ~TF_NOPUSH;
945 error = tcp_output(tp);
946 }
947 break;
948
949 case TCP_MAXSEG:
950 error = sooptcopyin(sopt, &optval, sizeof optval,
951 sizeof optval);
952 if (error)
953 break;
954
955 if (optval > 0 && optval <= tp->t_maxseg)
956 tp->t_maxseg = optval;
957 else
958 error = EINVAL;
959 break;
960
961 default:
962 error = ENOPROTOOPT;
963 break;
964 }
965 break;
966
967 case SOPT_GET:
968 switch (sopt->sopt_name) {
969 case TCP_NODELAY:
970 optval = tp->t_flags & TF_NODELAY;
971 break;
972 case TCP_MAXSEG:
973 optval = tp->t_maxseg;
974 break;
975 case TCP_NOOPT:
976 optval = tp->t_flags & TF_NOOPT;
977 break;
978 case TCP_NOPUSH:
979 optval = tp->t_flags & TF_NOPUSH;
980 break;
981 default:
982 error = ENOPROTOOPT;
983 break;
984 }
985 if (error == 0)
986 error = sooptcopyout(sopt, &optval, sizeof optval);
987 break;
988 }
989 splx(s);
990 return (error);
991}
992
993/*
994 * tcp_sendspace and tcp_recvspace are the default send and receive window
995 * sizes, respectively. These are obsolescent (this information should
996 * be set by the route).
997 */
998u_long tcp_sendspace = 1024*32;
999SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW,
1000 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size");
1001u_long tcp_recvspace = 57344; /* largest multiple of PAGE_SIZE < 64k */
1002SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
1003 &tcp_recvspace , 0, "Maximum incoming TCP datagram size");
1004
1005/*
1006 * Attach TCP protocol to socket, allocating
1007 * internet protocol control block, tcp control block,
1008 * bufer space, and entering LISTEN state if to accept connections.
1009 */
1010static int
1011tcp_attach(so, p)
1012 struct socket *so;
1013 struct proc *p;
1014{
1015 register struct tcpcb *tp;
1016 struct inpcb *inp;
1017 int error;
1018#ifdef INET6
1019 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL;
1020#endif
1021
1022 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1023 error = soreserve(so, tcp_sendspace, tcp_recvspace);
1024 if (error)
1025 return (error);
1026 }
1027 error = in_pcballoc(so, &tcbinfo, p);
1028 if (error)
1029 return (error);
1030 inp = sotoinpcb(so);
1031#ifdef INET6
1032 if (isipv6) {
1033 inp->inp_vflag |= INP_IPV6;
1034 inp->in6p_hops = -1; /* use kernel default */
1035 }
1036 else
1037#endif
1038 inp->inp_vflag |= INP_IPV4;
1039 tp = tcp_newtcpcb(inp);
1040 if (tp == 0) {
1041 int nofd = so->so_state & SS_NOFDREF; /* XXX */
1042
1043 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */
1044#ifdef INET6
1045 if (isipv6)
1046 in6_pcbdetach(inp);
1047 else
1048#endif
1049 in_pcbdetach(inp);
1050 so->so_state |= nofd;
1051 return (ENOBUFS);
1052 }
1053 tp->t_state = TCPS_CLOSED;
1054 return (0);
1055}
1056
1057/*
1058 * Initiate (or continue) disconnect.
1059 * If embryonic state, just send reset (once).
1060 * If in ``let data drain'' option and linger null, just drop.
1061 * Otherwise (hard), mark socket disconnecting and drop
1062 * current input data; switch states based on user close, and
1063 * send segment to peer (with FIN).
1064 */
1065static struct tcpcb *
1066tcp_disconnect(tp)
1067 register struct tcpcb *tp;
1068{
1069 struct socket *so = tp->t_inpcb->inp_socket;
1070
1071 if (tp->t_state < TCPS_ESTABLISHED)
1072 tp = tcp_close(tp);
1073 else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1074 tp = tcp_drop(tp, 0);
1075 else {
1076 soisdisconnecting(so);
1077 sbflush(&so->so_rcv);
1078 tp = tcp_usrclosed(tp);
1079 if (tp)
1080 (void) tcp_output(tp);
1081 }
1082 return (tp);
1083}
1084
1085/*
1086 * User issued close, and wish to trail through shutdown states:
1087 * if never received SYN, just forget it. If got a SYN from peer,
1088 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1089 * If already got a FIN from peer, then almost done; go to LAST_ACK
1090 * state. In all other cases, have already sent FIN to peer (e.g.
1091 * after PRU_SHUTDOWN), and just have to play tedious game waiting
1092 * for peer to send FIN or not respond to keep-alives, etc.
1093 * We can let the user exit from the close as soon as the FIN is acked.
1094 */
1095static struct tcpcb *
1096tcp_usrclosed(tp)
1097 register struct tcpcb *tp;
1098{
1099
1100 switch (tp->t_state) {
1101
1102 case TCPS_CLOSED:
1103 case TCPS_LISTEN:
1104 tp->t_state = TCPS_CLOSED;
1105 tp = tcp_close(tp);
1106 break;
1107
1108 case TCPS_SYN_SENT:
1109 case TCPS_SYN_RECEIVED:
1110 tp->t_flags |= TF_NEEDFIN;
1111 break;
1112
1113 case TCPS_ESTABLISHED:
1114 tp->t_state = TCPS_FIN_WAIT_1;
1115 break;
1116
1117 case TCPS_CLOSE_WAIT:
1118 tp->t_state = TCPS_LAST_ACK;
1119 break;
1120 }
1121 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
1122 soisdisconnected(tp->t_inpcb->inp_socket);
1123 /* To prevent the connection hanging in FIN_WAIT_2 forever. */
1124 if (tp->t_state == TCPS_FIN_WAIT_2)
1125 callout_reset(tp->tt_2msl, tcp_maxidle,
1126 tcp_timer_2msl, tp);
1127 }
1128 return (tp);
1129}
1130