proc->thread stage 4: rework the VFS and DEVICE subsystems to take thread
[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 $
dadab5e9 35 * $DragonFly: src/sys/netinet/tcp_usrreq.c,v 1.3 2003/06/25 03:56:04 dillon Exp $
984263bc
MD
36 */
37
38#include "opt_ipsec.h"
39#include "opt_inet6.h"
40#include "opt_tcpdebug.h"
41
42#include <sys/param.h>
43#include <sys/systm.h>
44#include <sys/kernel.h>
45#include <sys/sysctl.h>
46#include <sys/mbuf.h>
47#ifdef INET6
48#include <sys/domain.h>
49#endif /* INET6 */
50#include <sys/socket.h>
51#include <sys/socketvar.h>
52#include <sys/protosw.h>
53
54#include <net/if.h>
55#include <net/route.h>
56
57#include <netinet/in.h>
58#include <netinet/in_systm.h>
59#ifdef INET6
60#include <netinet/ip6.h>
61#endif
62#include <netinet/in_pcb.h>
63#ifdef INET6
64#include <netinet6/in6_pcb.h>
65#endif
66#include <netinet/in_var.h>
67#include <netinet/ip_var.h>
68#ifdef INET6
69#include <netinet6/ip6_var.h>
70#endif
71#include <netinet/tcp.h>
72#include <netinet/tcp_fsm.h>
73#include <netinet/tcp_seq.h>
74#include <netinet/tcp_timer.h>
75#include <netinet/tcp_var.h>
76#include <netinet/tcpip.h>
77#ifdef TCPDEBUG
78#include <netinet/tcp_debug.h>
79#endif
80
81#ifdef IPSEC
82#include <netinet6/ipsec.h>
83#endif /*IPSEC*/
84
85/*
86 * TCP protocol interface to socket abstraction.
87 */
88extern char *tcpstates[]; /* XXX ??? */
89
dadab5e9 90static int tcp_attach __P((struct socket *, struct thread *));
984263bc 91static int tcp_connect __P((struct tcpcb *, struct sockaddr *,
dadab5e9 92 struct thread *));
984263bc
MD
93#ifdef INET6
94static int tcp6_connect __P((struct tcpcb *, struct sockaddr *,
dadab5e9 95 struct thread *));
984263bc
MD
96#endif /* INET6 */
97static struct tcpcb *
98 tcp_disconnect __P((struct tcpcb *));
99static struct tcpcb *
100 tcp_usrclosed __P((struct tcpcb *));
101
102#ifdef TCPDEBUG
103#define TCPDEBUG0 int ostate = 0
104#define TCPDEBUG1() ostate = tp ? tp->t_state : 0
105#define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
106 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
107#else
108#define TCPDEBUG0
109#define TCPDEBUG1()
110#define TCPDEBUG2(req)
111#endif
112
113/*
114 * TCP attaches to socket via pru_attach(), reserving space,
115 * and an internet control block.
116 */
117static int
dadab5e9 118tcp_usr_attach(struct socket *so, int proto, struct thread *td)
984263bc
MD
119{
120 int s = splnet();
121 int error;
122 struct inpcb *inp = sotoinpcb(so);
123 struct tcpcb *tp = 0;
124 TCPDEBUG0;
125
126 TCPDEBUG1();
127 if (inp) {
128 error = EISCONN;
129 goto out;
130 }
131
dadab5e9 132 error = tcp_attach(so, td);
984263bc
MD
133 if (error)
134 goto out;
135
136 if ((so->so_options & SO_LINGER) && so->so_linger == 0)
137 so->so_linger = TCP_LINGERTIME;
138 tp = sototcpcb(so);
139out:
140 TCPDEBUG2(PRU_ATTACH);
141 splx(s);
142 return error;
143}
144
145/*
146 * pru_detach() detaches the TCP protocol from the socket.
147 * If the protocol state is non-embryonic, then can't
148 * do this directly: have to initiate a pru_disconnect(),
149 * which may finish later; embryonic TCB's can just
150 * be discarded here.
151 */
152static int
153tcp_usr_detach(struct socket *so)
154{
155 int s = splnet();
156 int error = 0;
157 struct inpcb *inp = sotoinpcb(so);
158 struct tcpcb *tp;
159 TCPDEBUG0;
160
161 if (inp == 0) {
162 splx(s);
163 return EINVAL; /* XXX */
164 }
165 tp = intotcpcb(inp);
166 TCPDEBUG1();
167 tp = tcp_disconnect(tp);
168
169 TCPDEBUG2(PRU_DETACH);
170 splx(s);
171 return error;
172}
173
174#define COMMON_START() TCPDEBUG0; \
175 do { \
176 if (inp == 0) { \
177 splx(s); \
178 return EINVAL; \
179 } \
180 tp = intotcpcb(inp); \
181 TCPDEBUG1(); \
182 } while(0)
183
184#define COMMON_END(req) out: TCPDEBUG2(req); splx(s); return error; goto out
185
186
187/*
188 * Give the socket an address.
189 */
190static int
dadab5e9 191tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
984263bc
MD
192{
193 int s = splnet();
194 int error = 0;
195 struct inpcb *inp = sotoinpcb(so);
196 struct tcpcb *tp;
197 struct sockaddr_in *sinp;
198
199 COMMON_START();
200
201 /*
202 * Must check for multicast addresses and disallow binding
203 * to them.
204 */
205 sinp = (struct sockaddr_in *)nam;
206 if (sinp->sin_family == AF_INET &&
207 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
208 error = EAFNOSUPPORT;
209 goto out;
210 }
dadab5e9 211 error = in_pcbbind(inp, nam, td);
984263bc
MD
212 if (error)
213 goto out;
214 COMMON_END(PRU_BIND);
215
216}
217
218#ifdef INET6
219static int
dadab5e9 220tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
984263bc
MD
221{
222 int s = splnet();
223 int error = 0;
224 struct inpcb *inp = sotoinpcb(so);
225 struct tcpcb *tp;
226 struct sockaddr_in6 *sin6p;
227
228 COMMON_START();
229
230 /*
231 * Must check for multicast addresses and disallow binding
232 * to them.
233 */
234 sin6p = (struct sockaddr_in6 *)nam;
235 if (sin6p->sin6_family == AF_INET6 &&
236 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
237 error = EAFNOSUPPORT;
238 goto out;
239 }
240 inp->inp_vflag &= ~INP_IPV4;
241 inp->inp_vflag |= INP_IPV6;
242 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
243 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
244 inp->inp_vflag |= INP_IPV4;
245 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
246 struct sockaddr_in sin;
247
248 in6_sin6_2_sin(&sin, sin6p);
249 inp->inp_vflag |= INP_IPV4;
250 inp->inp_vflag &= ~INP_IPV6;
dadab5e9 251 error = in_pcbbind(inp, (struct sockaddr *)&sin, td);
984263bc
MD
252 goto out;
253 }
254 }
dadab5e9 255 error = in6_pcbbind(inp, nam, td);
984263bc
MD
256 if (error)
257 goto out;
258 COMMON_END(PRU_BIND);
259}
260#endif /* INET6 */
261
262/*
263 * Prepare to accept connections.
264 */
265static int
dadab5e9 266tcp_usr_listen(struct socket *so, struct thread *td)
984263bc
MD
267{
268 int s = splnet();
269 int error = 0;
270 struct inpcb *inp = sotoinpcb(so);
271 struct tcpcb *tp;
272
273 COMMON_START();
274 if (inp->inp_lport == 0)
dadab5e9 275 error = in_pcbbind(inp, (struct sockaddr *)0, td);
984263bc
MD
276 if (error == 0)
277 tp->t_state = TCPS_LISTEN;
278 COMMON_END(PRU_LISTEN);
279}
280
281#ifdef INET6
282static int
dadab5e9 283tcp6_usr_listen(struct socket *so, struct thread *td)
984263bc
MD
284{
285 int s = splnet();
286 int error = 0;
287 struct inpcb *inp = sotoinpcb(so);
288 struct tcpcb *tp;
289
290 COMMON_START();
291 if (inp->inp_lport == 0) {
292 inp->inp_vflag &= ~INP_IPV4;
293 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
294 inp->inp_vflag |= INP_IPV4;
dadab5e9 295 error = in6_pcbbind(inp, (struct sockaddr *)0, td);
984263bc
MD
296 }
297 if (error == 0)
298 tp->t_state = TCPS_LISTEN;
299 COMMON_END(PRU_LISTEN);
300}
301#endif /* INET6 */
302
303/*
304 * Initiate connection to peer.
305 * Create a template for use in transmissions on this connection.
306 * Enter SYN_SENT state, and mark socket as connecting.
307 * Start keep-alive timer, and seed output sequence space.
308 * Send initial segment on connection.
309 */
310static int
dadab5e9 311tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
984263bc
MD
312{
313 int s = splnet();
314 int error = 0;
315 struct inpcb *inp = sotoinpcb(so);
316 struct tcpcb *tp;
317 struct sockaddr_in *sinp;
318
319 COMMON_START();
320
321 /*
322 * Must disallow TCP ``connections'' to multicast addresses.
323 */
324 sinp = (struct sockaddr_in *)nam;
325 if (sinp->sin_family == AF_INET
326 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
327 error = EAFNOSUPPORT;
328 goto out;
329 }
330
dadab5e9 331 prison_remote_ip(td, 0, &sinp->sin_addr.s_addr);
984263bc 332
dadab5e9 333 if ((error = tcp_connect(tp, nam, td)) != 0)
984263bc
MD
334 goto out;
335 error = tcp_output(tp);
336 COMMON_END(PRU_CONNECT);
337}
338
339#ifdef INET6
340static int
dadab5e9 341tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
984263bc
MD
342{
343 int s = splnet();
344 int error = 0;
345 struct inpcb *inp = sotoinpcb(so);
346 struct tcpcb *tp;
347 struct sockaddr_in6 *sin6p;
348
349 COMMON_START();
350
351 /*
352 * Must disallow TCP ``connections'' to multicast addresses.
353 */
354 sin6p = (struct sockaddr_in6 *)nam;
355 if (sin6p->sin6_family == AF_INET6
356 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
357 error = EAFNOSUPPORT;
358 goto out;
359 }
360
361 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
362 struct sockaddr_in sin;
363
364 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
365 error = EINVAL;
366 goto out;
367 }
368
369 in6_sin6_2_sin(&sin, sin6p);
370 inp->inp_vflag |= INP_IPV4;
371 inp->inp_vflag &= ~INP_IPV6;
dadab5e9 372 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
984263bc
MD
373 goto out;
374 error = tcp_output(tp);
375 goto out;
376 }
377 inp->inp_vflag &= ~INP_IPV4;
378 inp->inp_vflag |= INP_IPV6;
379 inp->inp_inc.inc_isipv6 = 1;
dadab5e9 380 if ((error = tcp6_connect(tp, nam, td)) != 0)
984263bc
MD
381 goto out;
382 error = tcp_output(tp);
383 COMMON_END(PRU_CONNECT);
384}
385#endif /* INET6 */
386
387/*
388 * Initiate disconnect from peer.
389 * If connection never passed embryonic stage, just drop;
390 * else if don't need to let data drain, then can just drop anyways,
391 * else have to begin TCP shutdown process: mark socket disconnecting,
392 * drain unread data, state switch to reflect user close, and
393 * send segment (e.g. FIN) to peer. Socket will be really disconnected
394 * when peer sends FIN and acks ours.
395 *
396 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
397 */
398static int
399tcp_usr_disconnect(struct socket *so)
400{
401 int s = splnet();
402 int error = 0;
403 struct inpcb *inp = sotoinpcb(so);
404 struct tcpcb *tp;
405
406 COMMON_START();
407 tp = tcp_disconnect(tp);
408 COMMON_END(PRU_DISCONNECT);
409}
410
411/*
412 * Accept a connection. Essentially all the work is
413 * done at higher levels; just return the address
414 * of the peer, storing through addr.
415 */
416static int
417tcp_usr_accept(struct socket *so, struct sockaddr **nam)
418{
419 int s = splnet();
420 int error = 0;
421 struct inpcb *inp = sotoinpcb(so);
422 struct tcpcb *tp = NULL;
423 TCPDEBUG0;
424
425 if (so->so_state & SS_ISDISCONNECTED) {
426 error = ECONNABORTED;
427 goto out;
428 }
429 if (inp == 0) {
430 splx(s);
431 return (EINVAL);
432 }
433 tp = intotcpcb(inp);
434 TCPDEBUG1();
435 in_setpeeraddr(so, nam);
436 COMMON_END(PRU_ACCEPT);
437}
438
439#ifdef INET6
440static int
441tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
442{
443 int s = splnet();
444 int error = 0;
445 struct inpcb *inp = sotoinpcb(so);
446 struct tcpcb *tp = NULL;
447 TCPDEBUG0;
448
449 if (so->so_state & SS_ISDISCONNECTED) {
450 error = ECONNABORTED;
451 goto out;
452 }
453 if (inp == 0) {
454 splx(s);
455 return (EINVAL);
456 }
457 tp = intotcpcb(inp);
458 TCPDEBUG1();
459 in6_mapped_peeraddr(so, nam);
460 COMMON_END(PRU_ACCEPT);
461}
462#endif /* INET6 */
463/*
464 * Mark the connection as being incapable of further output.
465 */
466static int
467tcp_usr_shutdown(struct socket *so)
468{
469 int s = splnet();
470 int error = 0;
471 struct inpcb *inp = sotoinpcb(so);
472 struct tcpcb *tp;
473
474 COMMON_START();
475 socantsendmore(so);
476 tp = tcp_usrclosed(tp);
477 if (tp)
478 error = tcp_output(tp);
479 COMMON_END(PRU_SHUTDOWN);
480}
481
482/*
483 * After a receive, possibly send window update to peer.
484 */
485static int
486tcp_usr_rcvd(struct socket *so, int flags)
487{
488 int s = splnet();
489 int error = 0;
490 struct inpcb *inp = sotoinpcb(so);
491 struct tcpcb *tp;
492
493 COMMON_START();
494 tcp_output(tp);
495 COMMON_END(PRU_RCVD);
496}
497
498/*
499 * Do a send by putting data in output queue and updating urgent
500 * marker if URG set. Possibly send more data. Unlike the other
501 * pru_*() routines, the mbuf chains are our responsibility. We
502 * must either enqueue them or free them. The other pru_* routines
503 * generally are caller-frees.
504 */
505static int
506tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
dadab5e9 507 struct sockaddr *nam, struct mbuf *control, struct thread *td)
984263bc
MD
508{
509 int s = splnet();
510 int error = 0;
511 struct inpcb *inp = sotoinpcb(so);
512 struct tcpcb *tp;
513#ifdef INET6
514 int isipv6;
515#endif
516 TCPDEBUG0;
517
518 if (inp == NULL) {
519 /*
520 * OOPS! we lost a race, the TCP session got reset after
521 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a
522 * network interrupt in the non-splnet() section of sosend().
523 */
524 if (m)
525 m_freem(m);
526 if (control)
527 m_freem(control);
528 error = ECONNRESET; /* XXX EPIPE? */
529 tp = NULL;
530 TCPDEBUG1();
531 goto out;
532 }
533#ifdef INET6
534 isipv6 = nam && nam->sa_family == AF_INET6;
535#endif /* INET6 */
536 tp = intotcpcb(inp);
537 TCPDEBUG1();
538 if (control) {
539 /* TCP doesn't do control messages (rights, creds, etc) */
540 if (control->m_len) {
541 m_freem(control);
542 if (m)
543 m_freem(m);
544 error = EINVAL;
545 goto out;
546 }
547 m_freem(control); /* empty control, just free it */
548 }
549 if(!(flags & PRUS_OOB)) {
550 sbappend(&so->so_snd, m);
551 if (nam && tp->t_state < TCPS_SYN_SENT) {
552 /*
553 * Do implied connect if not yet connected,
554 * initialize window to default value, and
555 * initialize maxseg/maxopd using peer's cached
556 * MSS.
557 */
558#ifdef INET6
559 if (isipv6)
dadab5e9 560 error = tcp6_connect(tp, nam, td);
984263bc
MD
561 else
562#endif /* INET6 */
dadab5e9 563 error = tcp_connect(tp, nam, td);
984263bc
MD
564 if (error)
565 goto out;
566 tp->snd_wnd = TTCP_CLIENT_SND_WND;
567 tcp_mss(tp, -1);
568 }
569
570 if (flags & PRUS_EOF) {
571 /*
572 * Close the send side of the connection after
573 * the data is sent.
574 */
575 socantsendmore(so);
576 tp = tcp_usrclosed(tp);
577 }
578 if (tp != NULL) {
579 if (flags & PRUS_MORETOCOME)
580 tp->t_flags |= TF_MORETOCOME;
581 error = tcp_output(tp);
582 if (flags & PRUS_MORETOCOME)
583 tp->t_flags &= ~TF_MORETOCOME;
584 }
585 } else {
586 if (sbspace(&so->so_snd) < -512) {
587 m_freem(m);
588 error = ENOBUFS;
589 goto out;
590 }
591 /*
592 * According to RFC961 (Assigned Protocols),
593 * the urgent pointer points to the last octet
594 * of urgent data. We continue, however,
595 * to consider it to indicate the first octet
596 * of data past the urgent section.
597 * Otherwise, snd_up should be one lower.
598 */
599 sbappend(&so->so_snd, m);
600 if (nam && tp->t_state < TCPS_SYN_SENT) {
601 /*
602 * Do implied connect if not yet connected,
603 * initialize window to default value, and
604 * initialize maxseg/maxopd using peer's cached
605 * MSS.
606 */
607#ifdef INET6
608 if (isipv6)
dadab5e9 609 error = tcp6_connect(tp, nam, td);
984263bc
MD
610 else
611#endif /* INET6 */
dadab5e9 612 error = tcp_connect(tp, nam, td);
984263bc
MD
613 if (error)
614 goto out;
615 tp->snd_wnd = TTCP_CLIENT_SND_WND;
616 tcp_mss(tp, -1);
617 }
618 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
619 tp->t_force = 1;
620 error = tcp_output(tp);
621 tp->t_force = 0;
622 }
623 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
624 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
625}
626
627/*
628 * Abort the TCP.
629 */
630static int
631tcp_usr_abort(struct socket *so)
632{
633 int s = splnet();
634 int error = 0;
635 struct inpcb *inp = sotoinpcb(so);
636 struct tcpcb *tp;
637
638 COMMON_START();
639 tp = tcp_drop(tp, ECONNABORTED);
640 COMMON_END(PRU_ABORT);
641}
642
643/*
644 * Receive out-of-band data.
645 */
646static int
647tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
648{
649 int s = splnet();
650 int error = 0;
651 struct inpcb *inp = sotoinpcb(so);
652 struct tcpcb *tp;
653
654 COMMON_START();
655 if ((so->so_oobmark == 0 &&
656 (so->so_state & SS_RCVATMARK) == 0) ||
657 so->so_options & SO_OOBINLINE ||
658 tp->t_oobflags & TCPOOB_HADDATA) {
659 error = EINVAL;
660 goto out;
661 }
662 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
663 error = EWOULDBLOCK;
664 goto out;
665 }
666 m->m_len = 1;
667 *mtod(m, caddr_t) = tp->t_iobc;
668 if ((flags & MSG_PEEK) == 0)
669 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
670 COMMON_END(PRU_RCVOOB);
671}
672
673/* xxx - should be const */
674struct pr_usrreqs tcp_usrreqs = {
675 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind,
676 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach,
677 tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd,
678 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
679 in_setsockaddr, sosend, soreceive, sopoll
680};
681
682#ifdef INET6
683struct pr_usrreqs tcp6_usrreqs = {
684 tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind,
685 tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach,
686 tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd,
687 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
688 in6_mapped_sockaddr, sosend, soreceive, sopoll
689};
690#endif /* INET6 */
691
692/*
693 * Common subroutine to open a TCP connection to remote host specified
694 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
695 * port number if needed. Call in_pcbladdr to do the routing and to choose
696 * a local host address (interface). If there is an existing incarnation
697 * of the same connection in TIME-WAIT state and if the remote host was
698 * sending CC options and if the connection duration was < MSL, then
699 * truncate the previous TIME-WAIT state and proceed.
700 * Initialize connection parameters and enter SYN-SENT state.
701 */
702static int
dadab5e9 703tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
984263bc
MD
704{
705 struct inpcb *inp = tp->t_inpcb, *oinp;
706 struct socket *so = inp->inp_socket;
707 struct tcpcb *otp;
708 struct sockaddr_in *sin = (struct sockaddr_in *)nam;
709 struct sockaddr_in *ifaddr;
710 struct rmxp_tao *taop;
711 struct rmxp_tao tao_noncached;
712 int error;
713
714 if (inp->inp_lport == 0) {
dadab5e9 715 error = in_pcbbind(inp, (struct sockaddr *)0, td);
984263bc
MD
716 if (error)
717 return error;
718 }
719
720 /*
721 * Cannot simply call in_pcbconnect, because there might be an
722 * earlier incarnation of this same connection still in
723 * TIME_WAIT state, creating an ADDRINUSE error.
724 */
725 error = in_pcbladdr(inp, nam, &ifaddr);
726 if (error)
727 return error;
728 oinp = in_pcblookup_hash(inp->inp_pcbinfo,
729 sin->sin_addr, sin->sin_port,
730 inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr
731 : ifaddr->sin_addr,
732 inp->inp_lport, 0, NULL);
733 if (oinp) {
734 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
735 otp->t_state == TCPS_TIME_WAIT &&
736 (ticks - otp->t_starttime) < tcp_msl &&
737 (otp->t_flags & TF_RCVD_CC))
738 otp = tcp_close(otp);
739 else
740 return EADDRINUSE;
741 }
742 if (inp->inp_laddr.s_addr == INADDR_ANY)
743 inp->inp_laddr = ifaddr->sin_addr;
744 inp->inp_faddr = sin->sin_addr;
745 inp->inp_fport = sin->sin_port;
746 in_pcbrehash(inp);
747
748 /* Compute window scaling to request. */
749 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
750 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
751 tp->request_r_scale++;
752
753 soisconnecting(so);
754 tcpstat.tcps_connattempt++;
755 tp->t_state = TCPS_SYN_SENT;
756 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
757 tp->iss = tcp_new_isn(tp);
758 tp->t_bw_rtseq = tp->iss;
759 tcp_sendseqinit(tp);
760
761 /*
762 * Generate a CC value for this connection and
763 * check whether CC or CCnew should be used.
764 */
765 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
766 taop = &tao_noncached;
767 bzero(taop, sizeof(*taop));
768 }
769
770 tp->cc_send = CC_INC(tcp_ccgen);
771 if (taop->tao_ccsent != 0 &&
772 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
773 taop->tao_ccsent = tp->cc_send;
774 } else {
775 taop->tao_ccsent = 0;
776 tp->t_flags |= TF_SENDCCNEW;
777 }
778
779 return 0;
780}
781
782#ifdef INET6
783static int
dadab5e9 784tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
984263bc
MD
785{
786 struct inpcb *inp = tp->t_inpcb, *oinp;
787 struct socket *so = inp->inp_socket;
788 struct tcpcb *otp;
789 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
790 struct in6_addr *addr6;
791 struct rmxp_tao *taop;
792 struct rmxp_tao tao_noncached;
793 int error;
794
795 if (inp->inp_lport == 0) {
dadab5e9 796 error = in6_pcbbind(inp, (struct sockaddr *)0, td);
984263bc
MD
797 if (error)
798 return error;
799 }
800
801 /*
802 * Cannot simply call in_pcbconnect, because there might be an
803 * earlier incarnation of this same connection still in
804 * TIME_WAIT state, creating an ADDRINUSE error.
805 */
806 error = in6_pcbladdr(inp, nam, &addr6);
807 if (error)
808 return error;
809 oinp = in6_pcblookup_hash(inp->inp_pcbinfo,
810 &sin6->sin6_addr, sin6->sin6_port,
811 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
812 ? addr6
813 : &inp->in6p_laddr,
814 inp->inp_lport, 0, NULL);
815 if (oinp) {
816 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
817 otp->t_state == TCPS_TIME_WAIT &&
818 (ticks - otp->t_starttime) < tcp_msl &&
819 (otp->t_flags & TF_RCVD_CC))
820 otp = tcp_close(otp);
821 else
822 return EADDRINUSE;
823 }
824 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
825 inp->in6p_laddr = *addr6;
826 inp->in6p_faddr = sin6->sin6_addr;
827 inp->inp_fport = sin6->sin6_port;
828 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != NULL)
829 inp->in6p_flowinfo = sin6->sin6_flowinfo;
830 in_pcbrehash(inp);
831
832 /* Compute window scaling to request. */
833 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
834 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
835 tp->request_r_scale++;
836
837 soisconnecting(so);
838 tcpstat.tcps_connattempt++;
839 tp->t_state = TCPS_SYN_SENT;
840 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
841 tp->iss = tcp_new_isn(tp);
842 tp->t_bw_rtseq = tp->iss;
843 tcp_sendseqinit(tp);
844
845 /*
846 * Generate a CC value for this connection and
847 * check whether CC or CCnew should be used.
848 */
849 if ((taop = tcp_gettaocache(&tp->t_inpcb->inp_inc)) == NULL) {
850 taop = &tao_noncached;
851 bzero(taop, sizeof(*taop));
852 }
853
854 tp->cc_send = CC_INC(tcp_ccgen);
855 if (taop->tao_ccsent != 0 &&
856 CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
857 taop->tao_ccsent = tp->cc_send;
858 } else {
859 taop->tao_ccsent = 0;
860 tp->t_flags |= TF_SENDCCNEW;
861 }
862
863 return 0;
864}
865#endif /* INET6 */
866
867/*
868 * The new sockopt interface makes it possible for us to block in the
869 * copyin/out step (if we take a page fault). Taking a page fault at
870 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now
871 * use TSM, there probably isn't any need for this function to run at
872 * splnet() any more. This needs more examination.)
873 */
874int
875tcp_ctloutput(so, sopt)
876 struct socket *so;
877 struct sockopt *sopt;
878{
879 int error, opt, optval, s;
880 struct inpcb *inp;
881 struct tcpcb *tp;
882
883 error = 0;
884 s = splnet(); /* XXX */
885 inp = sotoinpcb(so);
886 if (inp == NULL) {
887 splx(s);
888 return (ECONNRESET);
889 }
890 if (sopt->sopt_level != IPPROTO_TCP) {
891#ifdef INET6
892 if (INP_CHECK_SOCKAF(so, AF_INET6))
893 error = ip6_ctloutput(so, sopt);
894 else
895#endif /* INET6 */
896 error = ip_ctloutput(so, sopt);
897 splx(s);
898 return (error);
899 }
900 tp = intotcpcb(inp);
901
902 switch (sopt->sopt_dir) {
903 case SOPT_SET:
904 switch (sopt->sopt_name) {
905 case TCP_NODELAY:
906 case TCP_NOOPT:
907 error = sooptcopyin(sopt, &optval, sizeof optval,
908 sizeof optval);
909 if (error)
910 break;
911
912 switch (sopt->sopt_name) {
913 case TCP_NODELAY:
914 opt = TF_NODELAY;
915 break;
916 case TCP_NOOPT:
917 opt = TF_NOOPT;
918 break;
919 default:
920 opt = 0; /* dead code to fool gcc */
921 break;
922 }
923
924 if (optval)
925 tp->t_flags |= opt;
926 else
927 tp->t_flags &= ~opt;
928 break;
929
930 case TCP_NOPUSH:
931 error = sooptcopyin(sopt, &optval, sizeof optval,
932 sizeof optval);
933 if (error)
934 break;
935
936 if (optval)
937 tp->t_flags |= TF_NOPUSH;
938 else {
939 tp->t_flags &= ~TF_NOPUSH;
940 error = tcp_output(tp);
941 }
942 break;
943
944 case TCP_MAXSEG:
945 error = sooptcopyin(sopt, &optval, sizeof optval,
946 sizeof optval);
947 if (error)
948 break;
949
950 if (optval > 0 && optval <= tp->t_maxseg)
951 tp->t_maxseg = optval;
952 else
953 error = EINVAL;
954 break;
955
956 default:
957 error = ENOPROTOOPT;
958 break;
959 }
960 break;
961
962 case SOPT_GET:
963 switch (sopt->sopt_name) {
964 case TCP_NODELAY:
965 optval = tp->t_flags & TF_NODELAY;
966 break;
967 case TCP_MAXSEG:
968 optval = tp->t_maxseg;
969 break;
970 case TCP_NOOPT:
971 optval = tp->t_flags & TF_NOOPT;
972 break;
973 case TCP_NOPUSH:
974 optval = tp->t_flags & TF_NOPUSH;
975 break;
976 default:
977 error = ENOPROTOOPT;
978 break;
979 }
980 if (error == 0)
981 error = sooptcopyout(sopt, &optval, sizeof optval);
982 break;
983 }
984 splx(s);
985 return (error);
986}
987
988/*
989 * tcp_sendspace and tcp_recvspace are the default send and receive window
990 * sizes, respectively. These are obsolescent (this information should
991 * be set by the route).
992 */
993u_long tcp_sendspace = 1024*32;
994SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW,
995 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size");
996u_long tcp_recvspace = 57344; /* largest multiple of PAGE_SIZE < 64k */
997SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
998 &tcp_recvspace , 0, "Maximum incoming TCP datagram size");
999
1000/*
1001 * Attach TCP protocol to socket, allocating
1002 * internet protocol control block, tcp control block,
1003 * bufer space, and entering LISTEN state if to accept connections.
1004 */
1005static int
dadab5e9 1006tcp_attach(struct socket *so, struct thread *td)
984263bc
MD
1007{
1008 register struct tcpcb *tp;
1009 struct inpcb *inp;
1010 int error;
1011#ifdef INET6
1012 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != NULL;
1013#endif
1014
1015 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1016 error = soreserve(so, tcp_sendspace, tcp_recvspace);
1017 if (error)
1018 return (error);
1019 }
dadab5e9 1020 error = in_pcballoc(so, &tcbinfo, td);
984263bc
MD
1021 if (error)
1022 return (error);
1023 inp = sotoinpcb(so);
1024#ifdef INET6
1025 if (isipv6) {
1026 inp->inp_vflag |= INP_IPV6;
1027 inp->in6p_hops = -1; /* use kernel default */
1028 }
1029 else
1030#endif
1031 inp->inp_vflag |= INP_IPV4;
1032 tp = tcp_newtcpcb(inp);
1033 if (tp == 0) {
1034 int nofd = so->so_state & SS_NOFDREF; /* XXX */
1035
1036 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */
1037#ifdef INET6
1038 if (isipv6)
1039 in6_pcbdetach(inp);
1040 else
1041#endif
1042 in_pcbdetach(inp);
1043 so->so_state |= nofd;
1044 return (ENOBUFS);
1045 }
1046 tp->t_state = TCPS_CLOSED;
1047 return (0);
1048}
1049
1050/*
1051 * Initiate (or continue) disconnect.
1052 * If embryonic state, just send reset (once).
1053 * If in ``let data drain'' option and linger null, just drop.
1054 * Otherwise (hard), mark socket disconnecting and drop
1055 * current input data; switch states based on user close, and
1056 * send segment to peer (with FIN).
1057 */
1058static struct tcpcb *
1059tcp_disconnect(tp)
1060 register struct tcpcb *tp;
1061{
1062 struct socket *so = tp->t_inpcb->inp_socket;
1063
1064 if (tp->t_state < TCPS_ESTABLISHED)
1065 tp = tcp_close(tp);
1066 else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1067 tp = tcp_drop(tp, 0);
1068 else {
1069 soisdisconnecting(so);
1070 sbflush(&so->so_rcv);
1071 tp = tcp_usrclosed(tp);
1072 if (tp)
1073 (void) tcp_output(tp);
1074 }
1075 return (tp);
1076}
1077
1078/*
1079 * User issued close, and wish to trail through shutdown states:
1080 * if never received SYN, just forget it. If got a SYN from peer,
1081 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1082 * If already got a FIN from peer, then almost done; go to LAST_ACK
1083 * state. In all other cases, have already sent FIN to peer (e.g.
1084 * after PRU_SHUTDOWN), and just have to play tedious game waiting
1085 * for peer to send FIN or not respond to keep-alives, etc.
1086 * We can let the user exit from the close as soon as the FIN is acked.
1087 */
1088static struct tcpcb *
1089tcp_usrclosed(tp)
1090 register struct tcpcb *tp;
1091{
1092
1093 switch (tp->t_state) {
1094
1095 case TCPS_CLOSED:
1096 case TCPS_LISTEN:
1097 tp->t_state = TCPS_CLOSED;
1098 tp = tcp_close(tp);
1099 break;
1100
1101 case TCPS_SYN_SENT:
1102 case TCPS_SYN_RECEIVED:
1103 tp->t_flags |= TF_NEEDFIN;
1104 break;
1105
1106 case TCPS_ESTABLISHED:
1107 tp->t_state = TCPS_FIN_WAIT_1;
1108 break;
1109
1110 case TCPS_CLOSE_WAIT:
1111 tp->t_state = TCPS_LAST_ACK;
1112 break;
1113 }
1114 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
1115 soisdisconnected(tp->t_inpcb->inp_socket);
1116 /* To prevent the connection hanging in FIN_WAIT_2 forever. */
1117 if (tp->t_state == TCPS_FIN_WAIT_2)
1118 callout_reset(tp->tt_2msl, tcp_maxidle,
1119 tcp_timer_2msl, tp);
1120 }
1121 return (tp);
1122}
1123