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 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94
30 * $FreeBSD$
31 */
32
33 #include "opt_ipsec.h"
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 #include "opt_tcpdebug.h"
37
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/malloc.h>
41 #include <sys/kernel.h>
42 #include <sys/sysctl.h>
43 #include <sys/mbuf.h>
44 #ifdef INET6
45 #include <sys/domain.h>
46 #endif /* INET6 */
47 #include <sys/socket.h>
48 #include <sys/socketvar.h>
49 #include <sys/protosw.h>
50 #include <sys/proc.h>
51 #include <sys/jail.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 */
87 extern char *tcpstates[]; /* XXX ??? */
88
89 static int tcp_attach(struct socket *);
90 static int tcp_connect(struct tcpcb *, struct sockaddr *,
91 struct thread *td);
92 #ifdef INET6
93 static int tcp6_connect(struct tcpcb *, struct sockaddr *,
94 struct thread *td);
95 #endif /* INET6 */
96 static struct tcpcb *
97 tcp_disconnect(struct tcpcb *);
98 static struct tcpcb *
99 tcp_usrclosed(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 */
116 static int
117 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
118 {
119 int error;
120 struct inpcb *inp;
121 struct tcpcb *tp = 0;
122 TCPDEBUG0;
123
124 INP_INFO_WLOCK(&tcbinfo);
125 TCPDEBUG1();
126 inp = sotoinpcb(so);
127 if (inp) {
128 error = EISCONN;
129 goto out;
130 }
131
132 error = tcp_attach(so);
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
139 inp = sotoinpcb(so);
140 tp = intotcpcb(inp);
141 out:
142 TCPDEBUG2(PRU_ATTACH);
143 INP_INFO_WUNLOCK(&tcbinfo);
144 return error;
145 }
146
147 /*
148 * pru_detach() detaches the TCP protocol from the socket.
149 * If the protocol state is non-embryonic, then can't
150 * do this directly: have to initiate a pru_disconnect(),
151 * which may finish later; embryonic TCB's can just
152 * be discarded here.
153 */
154 static int
155 tcp_usr_detach(struct socket *so)
156 {
157 int error = 0;
158 struct inpcb *inp;
159 struct tcpcb *tp;
160 TCPDEBUG0;
161
162 INP_INFO_WLOCK(&tcbinfo);
163 inp = sotoinpcb(so);
164 if (inp == NULL) {
165 INP_INFO_WUNLOCK(&tcbinfo);
166 return error;
167 }
168 INP_LOCK(inp);
169 tp = intotcpcb(inp);
170 TCPDEBUG1();
171 tp = tcp_disconnect(tp);
172
173 TCPDEBUG2(PRU_DETACH);
174 if (tp)
175 INP_UNLOCK(inp);
176 INP_INFO_WUNLOCK(&tcbinfo);
177 return error;
178 }
179
180 #define INI_NOLOCK 0
181 #define INI_READ 1
182 #define INI_WRITE 2
183
184 #define COMMON_START() \
185 TCPDEBUG0; \
186 do { \
187 if (inirw == INI_READ) \
188 INP_INFO_RLOCK(&tcbinfo); \
189 else if (inirw == INI_WRITE) \
190 INP_INFO_WLOCK(&tcbinfo); \
191 inp = sotoinpcb(so); \
192 if (inp == 0) { \
193 if (inirw == INI_READ) \
194 INP_INFO_RUNLOCK(&tcbinfo); \
195 else if (inirw == INI_WRITE) \
196 INP_INFO_WUNLOCK(&tcbinfo); \
197 return EINVAL; \
198 } \
199 INP_LOCK(inp); \
200 if (inirw == INI_READ) \
201 INP_INFO_RUNLOCK(&tcbinfo); \
202 tp = intotcpcb(inp); \
203 TCPDEBUG1(); \
204 } while(0)
205
206 #define COMMON_END(req) \
207 out: TCPDEBUG2(req); \
208 do { \
209 if (tp) \
210 INP_UNLOCK(inp); \
211 if (inirw == INI_WRITE) \
212 INP_INFO_WUNLOCK(&tcbinfo); \
213 return error; \
214 goto out; \
215 } while(0)
216
217 /*
218 * Give the socket an address.
219 */
220 static int
221 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
222 {
223 int error = 0;
224 struct inpcb *inp;
225 struct tcpcb *tp;
226 struct sockaddr_in *sinp;
227 const int inirw = INI_WRITE;
228
229 sinp = (struct sockaddr_in *)nam;
230 if (nam->sa_len != sizeof (*sinp))
231 return (EINVAL);
232 /*
233 * Must check for multicast addresses and disallow binding
234 * to them.
235 */
236 if (sinp->sin_family == AF_INET &&
237 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
238 return (EAFNOSUPPORT);
239
240 COMMON_START();
241 error = in_pcbbind(inp, nam, td->td_ucred);
242 if (error)
243 goto out;
244 COMMON_END(PRU_BIND);
245 }
246
247 #ifdef INET6
248 static int
249 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
250 {
251 int error = 0;
252 struct inpcb *inp;
253 struct tcpcb *tp;
254 struct sockaddr_in6 *sin6p;
255 const int inirw = INI_WRITE;
256
257 sin6p = (struct sockaddr_in6 *)nam;
258 if (nam->sa_len != sizeof (*sin6p))
259 return (EINVAL);
260 /*
261 * Must check for multicast addresses and disallow binding
262 * to them.
263 */
264 if (sin6p->sin6_family == AF_INET6 &&
265 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
266 return (EAFNOSUPPORT);
267
268 COMMON_START();
269 inp->inp_vflag &= ~INP_IPV4;
270 inp->inp_vflag |= INP_IPV6;
271 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
272 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
273 inp->inp_vflag |= INP_IPV4;
274 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
275 struct sockaddr_in sin;
276
277 in6_sin6_2_sin(&sin, sin6p);
278 inp->inp_vflag |= INP_IPV4;
279 inp->inp_vflag &= ~INP_IPV6;
280 error = in_pcbbind(inp, (struct sockaddr *)&sin,
281 td->td_ucred);
282 goto out;
283 }
284 }
285 error = in6_pcbbind(inp, nam, td->td_ucred);
286 if (error)
287 goto out;
288 COMMON_END(PRU_BIND);
289 }
290 #endif /* INET6 */
291
292 /*
293 * Prepare to accept connections.
294 */
295 static int
296 tcp_usr_listen(struct socket *so, struct thread *td)
297 {
298 int error = 0;
299 struct inpcb *inp;
300 struct tcpcb *tp;
301 const int inirw = INI_WRITE;
302
303 COMMON_START();
304 SOCK_LOCK(so);
305 error = solisten_proto_check(so);
306 if (error == 0 && inp->inp_lport == 0)
307 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
308 if (error == 0) {
309 tp->t_state = TCPS_LISTEN;
310 solisten_proto(so);
311 }
312 SOCK_UNLOCK(so);
313 COMMON_END(PRU_LISTEN);
314 }
315
316 #ifdef INET6
317 static int
318 tcp6_usr_listen(struct socket *so, struct thread *td)
319 {
320 int error = 0;
321 struct inpcb *inp;
322 struct tcpcb *tp;
323 const int inirw = INI_WRITE;
324
325 COMMON_START();
326 SOCK_LOCK(so);
327 error = solisten_proto_check(so);
328 if (error == 0 && inp->inp_lport == 0) {
329 inp->inp_vflag &= ~INP_IPV4;
330 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
331 inp->inp_vflag |= INP_IPV4;
332 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
333 }
334 if (error == 0) {
335 tp->t_state = TCPS_LISTEN;
336 solisten_proto(so);
337 }
338 SOCK_UNLOCK(so);
339 COMMON_END(PRU_LISTEN);
340 }
341 #endif /* INET6 */
342
343 /*
344 * Initiate connection to peer.
345 * Create a template for use in transmissions on this connection.
346 * Enter SYN_SENT state, and mark socket as connecting.
347 * Start keep-alive timer, and seed output sequence space.
348 * Send initial segment on connection.
349 */
350 static int
351 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
352 {
353 int error = 0;
354 struct inpcb *inp;
355 struct tcpcb *tp;
356 struct sockaddr_in *sinp;
357 const int inirw = INI_WRITE;
358
359 sinp = (struct sockaddr_in *)nam;
360 if (nam->sa_len != sizeof (*sinp))
361 return (EINVAL);
362 /*
363 * Must disallow TCP ``connections'' to multicast addresses.
364 */
365 if (sinp->sin_family == AF_INET
366 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
367 return (EAFNOSUPPORT);
368 if (td && jailed(td->td_ucred))
369 prison_remote_ip(td->td_ucred, 0, &sinp->sin_addr.s_addr);
370
371 COMMON_START();
372 if ((error = tcp_connect(tp, nam, td)) != 0)
373 goto out;
374 error = tcp_output(tp);
375 COMMON_END(PRU_CONNECT);
376 }
377
378 #ifdef INET6
379 static int
380 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
381 {
382 int error = 0;
383 struct inpcb *inp;
384 struct tcpcb *tp;
385 struct sockaddr_in6 *sin6p;
386 const int inirw = INI_WRITE;
387
388 sin6p = (struct sockaddr_in6 *)nam;
389 if (nam->sa_len != sizeof (*sin6p))
390 return (EINVAL);
391 /*
392 * Must disallow TCP ``connections'' to multicast addresses.
393 */
394 if (sin6p->sin6_family == AF_INET6
395 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
396 return (EAFNOSUPPORT);
397
398 COMMON_START();
399 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
400 struct sockaddr_in sin;
401
402 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
403 error = EINVAL;
404 goto out;
405 }
406
407 in6_sin6_2_sin(&sin, sin6p);
408 inp->inp_vflag |= INP_IPV4;
409 inp->inp_vflag &= ~INP_IPV6;
410 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
411 goto out;
412 error = tcp_output(tp);
413 goto out;
414 }
415 inp->inp_vflag &= ~INP_IPV4;
416 inp->inp_vflag |= INP_IPV6;
417 inp->inp_inc.inc_isipv6 = 1;
418 if ((error = tcp6_connect(tp, nam, td)) != 0)
419 goto out;
420 error = tcp_output(tp);
421 COMMON_END(PRU_CONNECT);
422 }
423 #endif /* INET6 */
424
425 /*
426 * Initiate disconnect from peer.
427 * If connection never passed embryonic stage, just drop;
428 * else if don't need to let data drain, then can just drop anyways,
429 * else have to begin TCP shutdown process: mark socket disconnecting,
430 * drain unread data, state switch to reflect user close, and
431 * send segment (e.g. FIN) to peer. Socket will be really disconnected
432 * when peer sends FIN and acks ours.
433 *
434 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
435 */
436 static int
437 tcp_usr_disconnect(struct socket *so)
438 {
439 int error = 0;
440 struct inpcb *inp;
441 struct tcpcb *tp;
442 const int inirw = INI_WRITE;
443
444 COMMON_START();
445 tp = tcp_disconnect(tp);
446 COMMON_END(PRU_DISCONNECT);
447 }
448
449 /*
450 * Accept a connection. Essentially all the work is
451 * done at higher levels; just return the address
452 * of the peer, storing through addr.
453 */
454 static int
455 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
456 {
457 int error = 0;
458 struct inpcb *inp = NULL;
459 struct tcpcb *tp = NULL;
460 struct in_addr addr;
461 in_port_t port = 0;
462 TCPDEBUG0;
463
464 if (so->so_state & SS_ISDISCONNECTED) {
465 error = ECONNABORTED;
466 goto out;
467 }
468
469 INP_INFO_RLOCK(&tcbinfo);
470 inp = sotoinpcb(so);
471 if (!inp) {
472 INP_INFO_RUNLOCK(&tcbinfo);
473 return (EINVAL);
474 }
475 INP_LOCK(inp);
476 INP_INFO_RUNLOCK(&tcbinfo);
477 tp = intotcpcb(inp);
478 TCPDEBUG1();
479
480 /*
481 * We inline in_setpeeraddr and COMMON_END here, so that we can
482 * copy the data of interest and defer the malloc until after we
483 * release the lock.
484 */
485 port = inp->inp_fport;
486 addr = inp->inp_faddr;
487
488 out: TCPDEBUG2(PRU_ACCEPT);
489 if (tp)
490 INP_UNLOCK(inp);
491 if (error == 0)
492 *nam = in_sockaddr(port, &addr);
493 return error;
494 }
495
496 #ifdef INET6
497 static int
498 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
499 {
500 struct inpcb *inp = NULL;
501 int error = 0;
502 struct tcpcb *tp = NULL;
503 struct in_addr addr;
504 struct in6_addr addr6;
505 in_port_t port = 0;
506 int v4 = 0;
507 TCPDEBUG0;
508
509 if (so->so_state & SS_ISDISCONNECTED) {
510 error = ECONNABORTED;
511 goto out;
512 }
513
514 INP_INFO_RLOCK(&tcbinfo);
515 inp = sotoinpcb(so);
516 if (inp == 0) {
517 INP_INFO_RUNLOCK(&tcbinfo);
518 return (EINVAL);
519 }
520 INP_LOCK(inp);
521 INP_INFO_RUNLOCK(&tcbinfo);
522 tp = intotcpcb(inp);
523 TCPDEBUG1();
524 /*
525 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
526 * copy the data of interest and defer the malloc until after we
527 * release the lock.
528 */
529 if (inp->inp_vflag & INP_IPV4) {
530 v4 = 1;
531 port = inp->inp_fport;
532 addr = inp->inp_faddr;
533 } else {
534 port = inp->inp_fport;
535 addr6 = inp->in6p_faddr;
536 }
537
538 out: TCPDEBUG2(PRU_ACCEPT);
539 if (tp)
540 INP_UNLOCK(inp);
541 if (error == 0) {
542 if (v4)
543 *nam = in6_v4mapsin6_sockaddr(port, &addr);
544 else
545 *nam = in6_sockaddr(port, &addr6);
546 }
547 return error;
548 }
549 #endif /* INET6 */
550
551 /*
552 * This is the wrapper function for in_setsockaddr. We just pass down
553 * the pcbinfo for in_setsockaddr to lock. We don't want to do the locking
554 * here because in_setsockaddr will call malloc and can block.
555 */
556 static int
557 tcp_sockaddr(struct socket *so, struct sockaddr **nam)
558 {
559 return (in_setsockaddr(so, nam, &tcbinfo));
560 }
561
562 /*
563 * This is the wrapper function for in_setpeeraddr. We just pass down
564 * the pcbinfo for in_setpeeraddr to lock.
565 */
566 static int
567 tcp_peeraddr(struct socket *so, struct sockaddr **nam)
568 {
569 return (in_setpeeraddr(so, nam, &tcbinfo));
570 }
571
572 /*
573 * Mark the connection as being incapable of further output.
574 */
575 static int
576 tcp_usr_shutdown(struct socket *so)
577 {
578 int error = 0;
579 struct inpcb *inp;
580 struct tcpcb *tp;
581 const int inirw = INI_WRITE;
582
583 COMMON_START();
584 socantsendmore(so);
585 tp = tcp_usrclosed(tp);
586 if (tp)
587 error = tcp_output(tp);
588 COMMON_END(PRU_SHUTDOWN);
589 }
590
591 /*
592 * After a receive, possibly send window update to peer.
593 */
594 static int
595 tcp_usr_rcvd(struct socket *so, int flags)
596 {
597 int error = 0;
598 struct inpcb *inp;
599 struct tcpcb *tp;
600 const int inirw = INI_READ;
601
602 COMMON_START();
603 tcp_output(tp);
604 COMMON_END(PRU_RCVD);
605 }
606
607 /*
608 * Do a send by putting data in output queue and updating urgent
609 * marker if URG set. Possibly send more data. Unlike the other
610 * pru_*() routines, the mbuf chains are our responsibility. We
611 * must either enqueue them or free them. The other pru_* routines
612 * generally are caller-frees.
613 */
614 static int
615 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
616 struct sockaddr *nam, struct mbuf *control, struct thread *td)
617 {
618 int error = 0;
619 struct inpcb *inp;
620 struct tcpcb *tp;
621 const int inirw = INI_WRITE;
622 #ifdef INET6
623 int isipv6;
624 #endif
625 TCPDEBUG0;
626
627 /*
628 * Need write lock here because this function might call
629 * tcp_connect or tcp_usrclosed.
630 * We really want to have to this function upgrade from read lock
631 * to write lock. XXX
632 */
633 INP_INFO_WLOCK(&tcbinfo);
634 inp = sotoinpcb(so);
635 if (inp == NULL) {
636 /*
637 * OOPS! we lost a race, the TCP session got reset after
638 * we checked SBS_CANTSENDMORE, eg: while doing uiomove or a
639 * network interrupt in the non-splnet() section of sosend().
640 */
641 if (m)
642 m_freem(m);
643 if (control)
644 m_freem(control);
645 error = ECONNRESET; /* XXX EPIPE? */
646 tp = NULL;
647 TCPDEBUG1();
648 goto out;
649 }
650 INP_LOCK(inp);
651 #ifdef INET6
652 isipv6 = nam && nam->sa_family == AF_INET6;
653 #endif /* INET6 */
654 tp = intotcpcb(inp);
655 TCPDEBUG1();
656 if (control) {
657 /* TCP doesn't do control messages (rights, creds, etc) */
658 if (control->m_len) {
659 m_freem(control);
660 if (m)
661 m_freem(m);
662 error = EINVAL;
663 goto out;
664 }
665 m_freem(control); /* empty control, just free it */
666 }
667 if (!(flags & PRUS_OOB)) {
668 sbappendstream(&so->so_snd, m);
669 if (nam && tp->t_state < TCPS_SYN_SENT) {
670 /*
671 * Do implied connect if not yet connected,
672 * initialize window to default value, and
673 * initialize maxseg/maxopd using peer's cached
674 * MSS.
675 */
676 #ifdef INET6
677 if (isipv6)
678 error = tcp6_connect(tp, nam, td);
679 else
680 #endif /* INET6 */
681 error = tcp_connect(tp, nam, td);
682 if (error)
683 goto out;
684 tp->snd_wnd = TTCP_CLIENT_SND_WND;
685 tcp_mss(tp, -1);
686 }
687
688 if (flags & PRUS_EOF) {
689 /*
690 * Close the send side of the connection after
691 * the data is sent.
692 */
693 socantsendmore(so);
694 tp = tcp_usrclosed(tp);
695 }
696 if (tp != NULL) {
697 if (flags & PRUS_MORETOCOME)
698 tp->t_flags |= TF_MORETOCOME;
699 error = tcp_output(tp);
700 if (flags & PRUS_MORETOCOME)
701 tp->t_flags &= ~TF_MORETOCOME;
702 }
703 } else {
704 SOCKBUF_LOCK(&so->so_snd);
705 if (sbspace(&so->so_snd) < -512) {
706 SOCKBUF_UNLOCK(&so->so_snd);
707 m_freem(m);
708 error = ENOBUFS;
709 goto out;
710 }
711 /*
712 * According to RFC961 (Assigned Protocols),
713 * the urgent pointer points to the last octet
714 * of urgent data. We continue, however,
715 * to consider it to indicate the first octet
716 * of data past the urgent section.
717 * Otherwise, snd_up should be one lower.
718 */
719 sbappendstream_locked(&so->so_snd, m);
720 SOCKBUF_UNLOCK(&so->so_snd);
721 if (nam && tp->t_state < TCPS_SYN_SENT) {
722 /*
723 * Do implied connect if not yet connected,
724 * initialize window to default value, and
725 * initialize maxseg/maxopd using peer's cached
726 * MSS.
727 */
728 #ifdef INET6
729 if (isipv6)
730 error = tcp6_connect(tp, nam, td);
731 else
732 #endif /* INET6 */
733 error = tcp_connect(tp, nam, td);
734 if (error)
735 goto out;
736 tp->snd_wnd = TTCP_CLIENT_SND_WND;
737 tcp_mss(tp, -1);
738 }
739 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
740 tp->t_force = 1;
741 error = tcp_output(tp);
742 tp->t_force = 0;
743 }
744 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
745 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
746 }
747
748 /*
749 * Abort the TCP.
750 */
751 static int
752 tcp_usr_abort(struct socket *so)
753 {
754 int error = 0;
755 struct inpcb *inp;
756 struct tcpcb *tp;
757 const int inirw = INI_WRITE;
758
759 COMMON_START();
760 tp = tcp_drop(tp, ECONNABORTED);
761 COMMON_END(PRU_ABORT);
762 }
763
764 /*
765 * Receive out-of-band data.
766 */
767 static int
768 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
769 {
770 int error = 0;
771 struct inpcb *inp;
772 struct tcpcb *tp;
773 const int inirw = INI_READ;
774
775 COMMON_START();
776 if ((so->so_oobmark == 0 &&
777 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
778 so->so_options & SO_OOBINLINE ||
779 tp->t_oobflags & TCPOOB_HADDATA) {
780 error = EINVAL;
781 goto out;
782 }
783 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
784 error = EWOULDBLOCK;
785 goto out;
786 }
787 m->m_len = 1;
788 *mtod(m, caddr_t) = tp->t_iobc;
789 if ((flags & MSG_PEEK) == 0)
790 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
791 COMMON_END(PRU_RCVOOB);
792 }
793
794 /* xxx - should be const */
795 struct pr_usrreqs tcp_usrreqs = {
796 tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind,
797 tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach,
798 tcp_usr_disconnect, tcp_usr_listen, tcp_peeraddr, tcp_usr_rcvd,
799 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
800 tcp_sockaddr, sosend, soreceive, sopoll, in_pcbsosetlabel
801 };
802
803 #ifdef INET6
804 struct pr_usrreqs tcp6_usrreqs = {
805 tcp_usr_abort, tcp6_usr_accept, tcp_usr_attach, tcp6_usr_bind,
806 tcp6_usr_connect, pru_connect2_notsupp, in6_control, tcp_usr_detach,
807 tcp_usr_disconnect, tcp6_usr_listen, in6_mapped_peeraddr, tcp_usr_rcvd,
808 tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
809 in6_mapped_sockaddr, sosend, soreceive, sopoll, in_pcbsosetlabel
810 };
811 #endif /* INET6 */
812
813 /*
814 * Common subroutine to open a TCP connection to remote host specified
815 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
816 * port number if needed. Call in_pcbconnect_setup to do the routing and
817 * to choose a local host address (interface). If there is an existing
818 * incarnation of the same connection in TIME-WAIT state and if the remote
819 * host was sending CC options and if the connection duration was < MSL, then
820 * truncate the previous TIME-WAIT state and proceed.
821 * Initialize connection parameters and enter SYN-SENT state.
822 */
823 static int
824 tcp_connect(tp, nam, td)
825 register struct tcpcb *tp;
826 struct sockaddr *nam;
827 struct thread *td;
828 {
829 struct inpcb *inp = tp->t_inpcb, *oinp;
830 struct socket *so = inp->inp_socket;
831 struct tcptw *otw;
832 struct rmxp_tao tao;
833 struct in_addr laddr;
834 u_short lport;
835 int error;
836
837 bzero(&tao, sizeof(tao));
838
839 if (inp->inp_lport == 0) {
840 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
841 if (error)
842 return error;
843 }
844
845 /*
846 * Cannot simply call in_pcbconnect, because there might be an
847 * earlier incarnation of this same connection still in
848 * TIME_WAIT state, creating an ADDRINUSE error.
849 */
850 laddr = inp->inp_laddr;
851 lport = inp->inp_lport;
852 error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
853 &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
854 if (error && oinp == NULL)
855 return error;
856 if (oinp) {
857 if (oinp != inp &&
858 (oinp->inp_vflag & INP_TIMEWAIT) &&
859 (ticks - (otw = intotw(oinp))->t_starttime) < tcp_msl &&
860 otw->cc_recv != 0) {
861 inp->inp_faddr = oinp->inp_faddr;
862 inp->inp_fport = oinp->inp_fport;
863 (void) tcp_twclose(otw, 0);
864 } else
865 return EADDRINUSE;
866 }
867 inp->inp_laddr = laddr;
868 in_pcbrehash(inp);
869
870 /* Compute window scaling to request. */
871 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
872 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
873 tp->request_r_scale++;
874
875 soisconnecting(so);
876 tcpstat.tcps_connattempt++;
877 tp->t_state = TCPS_SYN_SENT;
878 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
879 tp->iss = tcp_new_isn(tp);
880 tp->t_bw_rtseq = tp->iss;
881 tcp_sendseqinit(tp);
882
883 /*
884 * Generate a CC value for this connection and
885 * check whether CC or CCnew should be used.
886 */
887 if (tcp_do_rfc1644)
888 tcp_hc_gettao(&inp->inp_inc, &tao);
889
890 tp->cc_send = CC_INC(tcp_ccgen);
891 if (tao.tao_ccsent != 0 &&
892 CC_GEQ(tp->cc_send, tao.tao_ccsent)) {
893 tao.tao_ccsent = tp->cc_send;
894 } else {
895 tao.tao_ccsent = 0;
896 tp->t_flags |= TF_SENDCCNEW;
897 }
898
899 if (tcp_do_rfc1644)
900 tcp_hc_updatetao(&inp->inp_inc, TCP_HC_TAO_CCSENT,
901 tao.tao_ccsent, 0);
902
903 return 0;
904 }
905
906 #ifdef INET6
907 static int
908 tcp6_connect(tp, nam, td)
909 register struct tcpcb *tp;
910 struct sockaddr *nam;
911 struct thread *td;
912 {
913 struct inpcb *inp = tp->t_inpcb, *oinp;
914 struct socket *so = inp->inp_socket;
915 struct tcptw *otw;
916 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
917 struct in6_addr *addr6;
918 struct rmxp_tao tao;
919 int error;
920
921 bzero(&tao, sizeof(tao));
922
923 if (inp->inp_lport == 0) {
924 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
925 if (error)
926 return error;
927 }
928
929 /*
930 * Cannot simply call in_pcbconnect, because there might be an
931 * earlier incarnation of this same connection still in
932 * TIME_WAIT state, creating an ADDRINUSE error.
933 */
934 error = in6_pcbladdr(inp, nam, &addr6);
935 if (error)
936 return error;
937 oinp = in6_pcblookup_hash(inp->inp_pcbinfo,
938 &sin6->sin6_addr, sin6->sin6_port,
939 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
940 ? addr6
941 : &inp->in6p_laddr,
942 inp->inp_lport, 0, NULL);
943 if (oinp) {
944 if (oinp != inp &&
945 (oinp->inp_vflag & INP_TIMEWAIT) &&
946 (ticks - (otw = intotw(oinp))->t_starttime) < tcp_msl &&
947 otw->cc_recv != 0) {
948 inp->inp_faddr = oinp->inp_faddr;
949 inp->inp_fport = oinp->inp_fport;
950 (void) tcp_twclose(otw, 0);
951 } else
952 return EADDRINUSE;
953 }
954 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
955 inp->in6p_laddr = *addr6;
956 inp->in6p_faddr = sin6->sin6_addr;
957 inp->inp_fport = sin6->sin6_port;
958 /* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
959 inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
960 if (inp->in6p_flags & IN6P_AUTOFLOWLABEL)
961 inp->in6p_flowinfo |=
962 (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
963 in_pcbrehash(inp);
964
965 /* Compute window scaling to request. */
966 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
967 (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
968 tp->request_r_scale++;
969
970 soisconnecting(so);
971 tcpstat.tcps_connattempt++;
972 tp->t_state = TCPS_SYN_SENT;
973 callout_reset(tp->tt_keep, tcp_keepinit, tcp_timer_keep, tp);
974 tp->iss = tcp_new_isn(tp);
975 tp->t_bw_rtseq = tp->iss;
976 tcp_sendseqinit(tp);
977
978 /*
979 * Generate a CC value for this connection and
980 * check whether CC or CCnew should be used.
981 */
982 if (tcp_do_rfc1644)
983 tcp_hc_gettao(&inp->inp_inc, &tao);
984
985 tp->cc_send = CC_INC(tcp_ccgen);
986 if (tao.tao_ccsent != 0 &&
987 CC_GEQ(tp->cc_send, tao.tao_ccsent)) {
988 tao.tao_ccsent = tp->cc_send;
989 } else {
990 tao.tao_ccsent = 0;
991 tp->t_flags |= TF_SENDCCNEW;
992 }
993 if (tcp_do_rfc1644)
994 tcp_hc_updatetao(&inp->inp_inc, TCP_HC_TAO_CCSENT,
995 tao.tao_ccsent, 0);
996
997 return 0;
998 }
999 #endif /* INET6 */
1000
1001 /*
1002 * The new sockopt interface makes it possible for us to block in the
1003 * copyin/out step (if we take a page fault). Taking a page fault at
1004 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now
1005 * use TSM, there probably isn't any need for this function to run at
1006 * splnet() any more. This needs more examination.)
1007 */
1008 int
1009 tcp_ctloutput(so, sopt)
1010 struct socket *so;
1011 struct sockopt *sopt;
1012 {
1013 int error, opt, optval;
1014 struct inpcb *inp;
1015 struct tcpcb *tp;
1016
1017 error = 0;
1018 INP_INFO_RLOCK(&tcbinfo);
1019 inp = sotoinpcb(so);
1020 if (inp == NULL) {
1021 INP_INFO_RUNLOCK(&tcbinfo);
1022 return (ECONNRESET);
1023 }
1024 INP_LOCK(inp);
1025 INP_INFO_RUNLOCK(&tcbinfo);
1026 if (sopt->sopt_level != IPPROTO_TCP) {
1027 INP_UNLOCK(inp);
1028 #ifdef INET6
1029 if (INP_CHECK_SOCKAF(so, AF_INET6))
1030 error = ip6_ctloutput(so, sopt);
1031 else
1032 #endif /* INET6 */
1033 error = ip_ctloutput(so, sopt);
1034 return (error);
1035 }
1036 tp = intotcpcb(inp);
1037
1038 switch (sopt->sopt_dir) {
1039 case SOPT_SET:
1040 switch (sopt->sopt_name) {
1041 #ifdef TCP_SIGNATURE
1042 case TCP_MD5SIG:
1043 error = sooptcopyin(sopt, &optval, sizeof optval,
1044 sizeof optval);
1045 if (error)
1046 break;
1047
1048 if (optval > 0)
1049 tp->t_flags |= TF_SIGNATURE;
1050 else
1051 tp->t_flags &= ~TF_SIGNATURE;
1052 break;
1053 #endif /* TCP_SIGNATURE */
1054 case TCP_NODELAY:
1055 case TCP_NOOPT:
1056 error = sooptcopyin(sopt, &optval, sizeof optval,
1057 sizeof optval);
1058 if (error)
1059 break;
1060
1061 switch (sopt->sopt_name) {
1062 case TCP_NODELAY:
1063 opt = TF_NODELAY;
1064 break;
1065 case TCP_NOOPT:
1066 opt = TF_NOOPT;
1067 break;
1068 default:
1069 opt = 0; /* dead code to fool gcc */
1070 break;
1071 }
1072
1073 if (optval)
1074 tp->t_flags |= opt;
1075 else
1076 tp->t_flags &= ~opt;
1077 break;
1078
1079 case TCP_NOPUSH:
1080 error = sooptcopyin(sopt, &optval, sizeof optval,
1081 sizeof optval);
1082 if (error)
1083 break;
1084
1085 if (optval)
1086 tp->t_flags |= TF_NOPUSH;
1087 else {
1088 tp->t_flags &= ~TF_NOPUSH;
1089 error = tcp_output(tp);
1090 }
1091 break;
1092
1093 case TCP_MAXSEG:
1094 error = sooptcopyin(sopt, &optval, sizeof optval,
1095 sizeof optval);
1096 if (error)
1097 break;
1098
1099 if (optval > 0 && optval <= tp->t_maxseg &&
1100 optval + 40 >= tcp_minmss)
1101 tp->t_maxseg = optval;
1102 else
1103 error = EINVAL;
1104 break;
1105
1106 default:
1107 error = ENOPROTOOPT;
1108 break;
1109 }
1110 break;
1111
1112 case SOPT_GET:
1113 switch (sopt->sopt_name) {
1114 #ifdef TCP_SIGNATURE
1115 case TCP_MD5SIG:
1116 optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
1117 break;
1118 #endif
1119 case TCP_NODELAY:
1120 optval = tp->t_flags & TF_NODELAY;
1121 break;
1122 case TCP_MAXSEG:
1123 optval = tp->t_maxseg;
1124 break;
1125 case TCP_NOOPT:
1126 optval = tp->t_flags & TF_NOOPT;
1127 break;
1128 case TCP_NOPUSH:
1129 optval = tp->t_flags & TF_NOPUSH;
1130 break;
1131 default:
1132 error = ENOPROTOOPT;
1133 break;
1134 }
1135 if (error == 0)
1136 error = sooptcopyout(sopt, &optval, sizeof optval);
1137 break;
1138 }
1139 INP_UNLOCK(inp);
1140 return (error);
1141 }
1142
1143 /*
1144 * tcp_sendspace and tcp_recvspace are the default send and receive window
1145 * sizes, respectively. These are obsolescent (this information should
1146 * be set by the route).
1147 */
1148 u_long tcp_sendspace = 1024*32;
1149 SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW,
1150 &tcp_sendspace , 0, "Maximum outgoing TCP datagram size");
1151 u_long tcp_recvspace = 1024*64;
1152 SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
1153 &tcp_recvspace , 0, "Maximum incoming TCP datagram size");
1154
1155 /*
1156 * Attach TCP protocol to socket, allocating
1157 * internet protocol control block, tcp control block,
1158 * bufer space, and entering LISTEN state if to accept connections.
1159 */
1160 static int
1161 tcp_attach(so)
1162 struct socket *so;
1163 {
1164 register struct tcpcb *tp;
1165 struct inpcb *inp;
1166 int error;
1167 #ifdef INET6
1168 int isipv6 = INP_CHECK_SOCKAF(so, AF_INET6) != 0;
1169 #endif
1170
1171 INP_INFO_WLOCK_ASSERT(&tcbinfo);
1172
1173 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1174 error = soreserve(so, tcp_sendspace, tcp_recvspace);
1175 if (error)
1176 return (error);
1177 }
1178 error = in_pcballoc(so, &tcbinfo, "tcpinp");
1179 if (error)
1180 return (error);
1181 inp = sotoinpcb(so);
1182 #ifdef INET6
1183 if (isipv6) {
1184 inp->inp_vflag |= INP_IPV6;
1185 inp->in6p_hops = -1; /* use kernel default */
1186 }
1187 else
1188 #endif
1189 inp->inp_vflag |= INP_IPV4;
1190 tp = tcp_newtcpcb(inp);
1191 if (tp == 0) {
1192 int nofd = so->so_state & SS_NOFDREF; /* XXX */
1193
1194 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */
1195
1196 INP_LOCK(inp);
1197 #ifdef INET6
1198 if (isipv6)
1199 in6_pcbdetach(inp);
1200 else
1201 #endif
1202 in_pcbdetach(inp);
1203 so->so_state |= nofd;
1204 return (ENOBUFS);
1205 }
1206 tp->t_state = TCPS_CLOSED;
1207 return (0);
1208 }
1209
1210 /*
1211 * Initiate (or continue) disconnect.
1212 * If embryonic state, just send reset (once).
1213 * If in ``let data drain'' option and linger null, just drop.
1214 * Otherwise (hard), mark socket disconnecting and drop
1215 * current input data; switch states based on user close, and
1216 * send segment to peer (with FIN).
1217 */
1218 static struct tcpcb *
1219 tcp_disconnect(tp)
1220 register struct tcpcb *tp;
1221 {
1222 struct inpcb *inp = tp->t_inpcb;
1223 struct socket *so = inp->inp_socket;
1224
1225 INP_INFO_WLOCK_ASSERT(&tcbinfo);
1226 INP_LOCK_ASSERT(inp);
1227
1228 if (tp->t_state < TCPS_ESTABLISHED)
1229 tp = tcp_close(tp);
1230 else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1231 tp = tcp_drop(tp, 0);
1232 else {
1233 soisdisconnecting(so);
1234 sbflush(&so->so_rcv);
1235 tp = tcp_usrclosed(tp);
1236 if (tp)
1237 (void) tcp_output(tp);
1238 }
1239 return (tp);
1240 }
1241
1242 /*
1243 * User issued close, and wish to trail through shutdown states:
1244 * if never received SYN, just forget it. If got a SYN from peer,
1245 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1246 * If already got a FIN from peer, then almost done; go to LAST_ACK
1247 * state. In all other cases, have already sent FIN to peer (e.g.
1248 * after PRU_SHUTDOWN), and just have to play tedious game waiting
1249 * for peer to send FIN or not respond to keep-alives, etc.
1250 * We can let the user exit from the close as soon as the FIN is acked.
1251 */
1252 static struct tcpcb *
1253 tcp_usrclosed(tp)
1254 register struct tcpcb *tp;
1255 {
1256
1257 INP_INFO_WLOCK_ASSERT(&tcbinfo);
1258 INP_LOCK_ASSERT(tp->t_inpcb);
1259
1260 switch (tp->t_state) {
1261
1262 case TCPS_CLOSED:
1263 case TCPS_LISTEN:
1264 tp->t_state = TCPS_CLOSED;
1265 tp = tcp_close(tp);
1266 break;
1267
1268 case TCPS_SYN_SENT:
1269 case TCPS_SYN_RECEIVED:
1270 tp->t_flags |= TF_NEEDFIN;
1271 break;
1272
1273 case TCPS_ESTABLISHED:
1274 tp->t_state = TCPS_FIN_WAIT_1;
1275 break;
1276
1277 case TCPS_CLOSE_WAIT:
1278 tp->t_state = TCPS_LAST_ACK;
1279 break;
1280 }
1281 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
1282 soisdisconnected(tp->t_inpcb->inp_socket);
1283 /* To prevent the connection hanging in FIN_WAIT_2 forever. */
1284 if (tp->t_state == TCPS_FIN_WAIT_2)
1285 callout_reset(tp->tt_2msl, tcp_maxidle,
1286 tcp_timer_2msl, tp);
1287 }
1288 return (tp);
1289 }
1290
Cache object: cda52e520d8c25e42ddaff68dec94db2
|