1/*-
2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California.
4 * Copyright (c) 2006-2007 Robert N. M. Watson
5 * Copyright (c) 2010-2011 Juniper Networks, Inc.
6 * All rights reserved.
7 *
8 * Portions of this software were developed by Robert N. M. Watson under
9 * contract to Juniper Networks, Inc.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 4. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 *
35 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94
36 */
37
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD: head/sys/netinet/tcp_usrreq.c 289276 2015-10-14 00:35:37Z hiren $");
40
41#include "opt_ddb.h"
42#include "opt_inet.h"
43#include "opt_inet6.h"
44#include "opt_tcpdebug.h"
45
46#include <sys/param.h>
47#include <sys/systm.h>
48#include <sys/limits.h>
49#include <sys/malloc.h>
50#include <sys/kernel.h>
51#include <sys/sysctl.h>
52#include <sys/mbuf.h>
53#ifdef INET6
54#include <sys/domain.h>
55#endif /* INET6 */
56#include <sys/socket.h>
57#include <sys/socketvar.h>
58#include <sys/protosw.h>
59#include <sys/proc.h>
60#include <sys/jail.h>
61
62#ifdef DDB
63#include <ddb/ddb.h>
64#endif
65
66#include <net/if.h>
67#include <net/if_var.h>
68#include <net/route.h>
69#include <net/vnet.h>
70
71#include <netinet/cc.h>
72#include <netinet/in.h>
73#include <netinet/in_kdtrace.h>
74#include <netinet/in_pcb.h>
75#include <netinet/in_systm.h>
76#include <netinet/in_var.h>
77#include <netinet/ip_var.h>
78#ifdef INET6
79#include <netinet/ip6.h>
80#include <netinet6/in6_pcb.h>
81#include <netinet6/ip6_var.h>
82#include <netinet6/scope6_var.h>
83#endif
84#include <netinet/tcp_fsm.h>
85#include <netinet/tcp_seq.h>
86#include <netinet/tcp_timer.h>
87#include <netinet/tcp_var.h>
88#include <netinet/tcpip.h>
89#ifdef TCPPCAP
90#include <netinet/tcp_pcap.h>
91#endif
92#ifdef TCPDEBUG
93#include <netinet/tcp_debug.h>
94#endif
95#ifdef TCP_OFFLOAD
96#include <netinet/tcp_offload.h>
97#endif
98
99/*
100 * TCP protocol interface to socket abstraction.
101 */
102static int tcp_attach(struct socket *);
103#ifdef INET
104static int tcp_connect(struct tcpcb *, struct sockaddr *,
105 struct thread *td);
106#endif /* INET */
107#ifdef INET6
108static int tcp6_connect(struct tcpcb *, struct sockaddr *,
109 struct thread *td);
110#endif /* INET6 */
111static void tcp_disconnect(struct tcpcb *);
112static void tcp_usrclosed(struct tcpcb *);
113static void tcp_fill_info(struct tcpcb *, struct tcp_info *);
114
115#ifdef TCPDEBUG
116#define TCPDEBUG0 int ostate = 0
117#define TCPDEBUG1() ostate = tp ? tp->t_state : 0
118#define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
119 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
120#else
121#define TCPDEBUG0
122#define TCPDEBUG1()
123#define TCPDEBUG2(req)
124#endif
125
126/*
127 * TCP attaches to socket via pru_attach(), reserving space,
128 * and an internet control block.
129 */
130static int
131tcp_usr_attach(struct socket *so, int proto, struct thread *td)
132{
133 struct inpcb *inp;
134 struct tcpcb *tp = NULL;
135 int error;
136 TCPDEBUG0;
137
138 inp = sotoinpcb(so);
139 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
140 TCPDEBUG1();
141
142 error = tcp_attach(so);
143 if (error)
144 goto out;
145
146 if ((so->so_options & SO_LINGER) && so->so_linger == 0)
147 so->so_linger = TCP_LINGERTIME;
148
149 inp = sotoinpcb(so);
150 tp = intotcpcb(inp);
151out:
152 TCPDEBUG2(PRU_ATTACH);
153 TCP_PROBE2(debug__user, tp, PRU_ATTACH);
154 return error;
155}
156
157/*
158 * tcp_detach is called when the socket layer loses its final reference
159 * to the socket, be it a file descriptor reference, a reference from TCP,
160 * etc. At this point, there is only one case in which we will keep around
161 * inpcb state: time wait.
162 *
163 * This function can probably be re-absorbed back into tcp_usr_detach() now
164 * that there is a single detach path.
165 */
166static void
167tcp_detach(struct socket *so, struct inpcb *inp)
168{
169 struct tcpcb *tp;
170
171 INP_INFO_LOCK_ASSERT(&V_tcbinfo);
172 INP_WLOCK_ASSERT(inp);
173
174 KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
175 KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
176
177 tp = intotcpcb(inp);
178
179 if (inp->inp_flags & INP_TIMEWAIT) {
180 /*
181 * There are two cases to handle: one in which the time wait
182 * state is being discarded (INP_DROPPED), and one in which
183 * this connection will remain in timewait. In the former,
184 * it is time to discard all state (except tcptw, which has
185 * already been discarded by the timewait close code, which
186 * should be further up the call stack somewhere). In the
187 * latter case, we detach from the socket, but leave the pcb
188 * present until timewait ends.
189 *
190 * XXXRW: Would it be cleaner to free the tcptw here?
191 *
192 * Astute question indeed, from twtcp perspective there are
193 * three cases to consider:
194 *
195 * #1 tcp_detach is called at tcptw creation time by
196 * tcp_twstart, then do not discard the newly created tcptw
197 * and leave inpcb present until timewait ends
198 * #2 tcp_detach is called at timewait end (or reuse) by
199 * tcp_twclose, then the tcptw has already been discarded
200 * (or reused) and inpcb is freed here
201 * #3 tcp_detach is called() after timewait ends (or reuse)
202 * (e.g. by soclose), then tcptw has already been discarded
203 * (or reused) and inpcb is freed here
204 *
205 * In all three cases the tcptw should not be freed here.
206 */
207 if (inp->inp_flags & INP_DROPPED) {
208 KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && "
209 "INP_DROPPED && tp != NULL"));
210 in_pcbdetach(inp);
211 in_pcbfree(inp);
212 } else {
213 in_pcbdetach(inp);
214 INP_WUNLOCK(inp);
215 }
216 } else {
217 /*
218 * If the connection is not in timewait, we consider two
219 * two conditions: one in which no further processing is
220 * necessary (dropped || embryonic), and one in which TCP is
221 * not yet done, but no longer requires the socket, so the
222 * pcb will persist for the time being.
223 *
224 * XXXRW: Does the second case still occur?
225 */
226 if (inp->inp_flags & INP_DROPPED ||
227 tp->t_state < TCPS_SYN_SENT) {
228 tcp_discardcb(tp);
229 in_pcbdetach(inp);
230 in_pcbfree(inp);
231 } else {
232 in_pcbdetach(inp);
233 INP_WUNLOCK(inp);
234 }
235 }
236}
237
238/*
239 * pru_detach() detaches the TCP protocol from the socket.
240 * If the protocol state is non-embryonic, then can't
241 * do this directly: have to initiate a pru_disconnect(),
242 * which may finish later; embryonic TCB's can just
243 * be discarded here.
244 */
245static void
246tcp_usr_detach(struct socket *so)
247{
248 struct inpcb *inp;
249 int rlock = 0;
250
251 inp = sotoinpcb(so);
252 KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
253 if (!INP_INFO_WLOCKED(&V_tcbinfo)) {
254 INP_INFO_RLOCK(&V_tcbinfo);
255 rlock = 1;
256 }
257 INP_WLOCK(inp);
258 KASSERT(inp->inp_socket != NULL,
259 ("tcp_usr_detach: inp_socket == NULL"));
260 tcp_detach(so, inp);
261 if (rlock)
262 INP_INFO_RUNLOCK(&V_tcbinfo);
263}
264
265#ifdef INET
266/*
267 * Give the socket an address.
268 */
269static int
270tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
271{
272 int error = 0;
273 struct inpcb *inp;
274 struct tcpcb *tp = NULL;
275 struct sockaddr_in *sinp;
276
277 sinp = (struct sockaddr_in *)nam;
278 if (nam->sa_len != sizeof (*sinp))
279 return (EINVAL);
280 /*
281 * Must check for multicast addresses and disallow binding
282 * to them.
283 */
284 if (sinp->sin_family == AF_INET &&
285 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
286 return (EAFNOSUPPORT);
287
288 TCPDEBUG0;
289 inp = sotoinpcb(so);
290 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
291 INP_WLOCK(inp);
292 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
293 error = EINVAL;
294 goto out;
295 }
296 tp = intotcpcb(inp);
297 TCPDEBUG1();
298 INP_HASH_WLOCK(&V_tcbinfo);
299 error = in_pcbbind(inp, nam, td->td_ucred);
300 INP_HASH_WUNLOCK(&V_tcbinfo);
301out:
302 TCPDEBUG2(PRU_BIND);
303 TCP_PROBE2(debug__user, tp, PRU_BIND);
304 INP_WUNLOCK(inp);
305
306 return (error);
307}
308#endif /* INET */
309
310#ifdef INET6
311static int
312tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
313{
314 int error = 0;
315 struct inpcb *inp;
316 struct tcpcb *tp = NULL;
317 struct sockaddr_in6 *sin6p;
318
319 sin6p = (struct sockaddr_in6 *)nam;
320 if (nam->sa_len != sizeof (*sin6p))
321 return (EINVAL);
322 /*
323 * Must check for multicast addresses and disallow binding
324 * to them.
325 */
326 if (sin6p->sin6_family == AF_INET6 &&
327 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
328 return (EAFNOSUPPORT);
329
330 TCPDEBUG0;
331 inp = sotoinpcb(so);
332 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
333 INP_WLOCK(inp);
334 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
335 error = EINVAL;
336 goto out;
337 }
338 tp = intotcpcb(inp);
339 TCPDEBUG1();
340 INP_HASH_WLOCK(&V_tcbinfo);
341 inp->inp_vflag &= ~INP_IPV4;
342 inp->inp_vflag |= INP_IPV6;
343#ifdef INET
344 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
345 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
346 inp->inp_vflag |= INP_IPV4;
347 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
348 struct sockaddr_in sin;
349
350 in6_sin6_2_sin(&sin, sin6p);
351 inp->inp_vflag |= INP_IPV4;
352 inp->inp_vflag &= ~INP_IPV6;
353 error = in_pcbbind(inp, (struct sockaddr *)&sin,
354 td->td_ucred);
355 INP_HASH_WUNLOCK(&V_tcbinfo);
356 goto out;
357 }
358 }
359#endif
360 error = in6_pcbbind(inp, nam, td->td_ucred);
361 INP_HASH_WUNLOCK(&V_tcbinfo);
362out:
363 TCPDEBUG2(PRU_BIND);
364 TCP_PROBE2(debug__user, tp, PRU_BIND);
365 INP_WUNLOCK(inp);
366 return (error);
367}
368#endif /* INET6 */
369
370#ifdef INET
371/*
372 * Prepare to accept connections.
373 */
374static int
375tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
376{
377 int error = 0;
378 struct inpcb *inp;
379 struct tcpcb *tp = NULL;
380
381 TCPDEBUG0;
382 inp = sotoinpcb(so);
383 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
384 INP_WLOCK(inp);
385 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
386 error = EINVAL;
387 goto out;
388 }
389 tp = intotcpcb(inp);
390 TCPDEBUG1();
391 SOCK_LOCK(so);
392 error = solisten_proto_check(so);
393 INP_HASH_WLOCK(&V_tcbinfo);
394 if (error == 0 && inp->inp_lport == 0)
395 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
396 INP_HASH_WUNLOCK(&V_tcbinfo);
397 if (error == 0) {
398 tcp_state_change(tp, TCPS_LISTEN);
399 solisten_proto(so, backlog);
400#ifdef TCP_OFFLOAD
401 if ((so->so_options & SO_NO_OFFLOAD) == 0)
402 tcp_offload_listen_start(tp);
403#endif
404 }
405 SOCK_UNLOCK(so);
406
407out:
408 TCPDEBUG2(PRU_LISTEN);
409 TCP_PROBE2(debug__user, tp, PRU_LISTEN);
410 INP_WUNLOCK(inp);
411 return (error);
412}
413#endif /* INET */
414
415#ifdef INET6
416static int
417tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
418{
419 int error = 0;
420 struct inpcb *inp;
421 struct tcpcb *tp = NULL;
422
423 TCPDEBUG0;
424 inp = sotoinpcb(so);
425 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
426 INP_WLOCK(inp);
427 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
428 error = EINVAL;
429 goto out;
430 }
431 tp = intotcpcb(inp);
432 TCPDEBUG1();
433 SOCK_LOCK(so);
434 error = solisten_proto_check(so);
435 INP_HASH_WLOCK(&V_tcbinfo);
436 if (error == 0 && inp->inp_lport == 0) {
437 inp->inp_vflag &= ~INP_IPV4;
438 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
439 inp->inp_vflag |= INP_IPV4;
440 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
441 }
442 INP_HASH_WUNLOCK(&V_tcbinfo);
443 if (error == 0) {
444 tcp_state_change(tp, TCPS_LISTEN);
445 solisten_proto(so, backlog);
446#ifdef TCP_OFFLOAD
447 if ((so->so_options & SO_NO_OFFLOAD) == 0)
448 tcp_offload_listen_start(tp);
449#endif
450 }
451 SOCK_UNLOCK(so);
452
453out:
454 TCPDEBUG2(PRU_LISTEN);
455 TCP_PROBE2(debug__user, tp, PRU_LISTEN);
456 INP_WUNLOCK(inp);
457 return (error);
458}
459#endif /* INET6 */
460
461#ifdef INET
462/*
463 * Initiate connection to peer.
464 * Create a template for use in transmissions on this connection.
465 * Enter SYN_SENT state, and mark socket as connecting.
466 * Start keep-alive timer, and seed output sequence space.
467 * Send initial segment on connection.
468 */
469static int
470tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
471{
472 int error = 0;
473 struct inpcb *inp;
474 struct tcpcb *tp = NULL;
475 struct sockaddr_in *sinp;
476
477 sinp = (struct sockaddr_in *)nam;
478 if (nam->sa_len != sizeof (*sinp))
479 return (EINVAL);
480 /*
481 * Must disallow TCP ``connections'' to multicast addresses.
482 */
483 if (sinp->sin_family == AF_INET
484 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
485 return (EAFNOSUPPORT);
486 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
487 return (error);
488
489 TCPDEBUG0;
490 inp = sotoinpcb(so);
491 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
492 INP_WLOCK(inp);
493 if (inp->inp_flags & INP_TIMEWAIT) {
494 error = EADDRINUSE;
495 goto out;
496 }
497 if (inp->inp_flags & INP_DROPPED) {
498 error = ECONNREFUSED;
499 goto out;
500 }
501 tp = intotcpcb(inp);
502 TCPDEBUG1();
503 if ((error = tcp_connect(tp, nam, td)) != 0)
504 goto out;
505#ifdef TCP_OFFLOAD
506 if (registered_toedevs > 0 &&
507 (so->so_options & SO_NO_OFFLOAD) == 0 &&
508 (error = tcp_offload_connect(so, nam)) == 0)
509 goto out;
510#endif
511 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
512 error = tcp_output(tp);
513out:
514 TCPDEBUG2(PRU_CONNECT);
515 INP_WUNLOCK(inp);
516 return (error);
517}
518#endif /* INET */
519
520#ifdef INET6
521static int
522tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
523{
524 int error = 0;
525 struct inpcb *inp;
526 struct tcpcb *tp = NULL;
527 struct sockaddr_in6 *sin6p;
528
529 TCPDEBUG0;
530
531 sin6p = (struct sockaddr_in6 *)nam;
532 if (nam->sa_len != sizeof (*sin6p))
533 return (EINVAL);
534 /*
535 * Must disallow TCP ``connections'' to multicast addresses.
536 */
537 if (sin6p->sin6_family == AF_INET6
538 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
539 return (EAFNOSUPPORT);
540
541 inp = sotoinpcb(so);
542 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
543 INP_WLOCK(inp);
544 if (inp->inp_flags & INP_TIMEWAIT) {
545 error = EADDRINUSE;
546 goto out;
547 }
548 if (inp->inp_flags & INP_DROPPED) {
549 error = ECONNREFUSED;
550 goto out;
551 }
552 tp = intotcpcb(inp);
553 TCPDEBUG1();
554#ifdef INET
555 /*
556 * XXXRW: Some confusion: V4/V6 flags relate to binding, and
557 * therefore probably require the hash lock, which isn't held here.
558 * Is this a significant problem?
559 */
560 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
561 struct sockaddr_in sin;
562
563 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
564 error = EINVAL;
565 goto out;
566 }
567
568 in6_sin6_2_sin(&sin, sin6p);
569 inp->inp_vflag |= INP_IPV4;
570 inp->inp_vflag &= ~INP_IPV6;
571 if ((error = prison_remote_ip4(td->td_ucred,
572 &sin.sin_addr)) != 0)
573 goto out;
574 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
575 goto out;
576#ifdef TCP_OFFLOAD
577 if (registered_toedevs > 0 &&
578 (so->so_options & SO_NO_OFFLOAD) == 0 &&
579 (error = tcp_offload_connect(so, nam)) == 0)
580 goto out;
581#endif
582 error = tcp_output(tp);
583 goto out;
584 }
585#endif
586 inp->inp_vflag &= ~INP_IPV4;
587 inp->inp_vflag |= INP_IPV6;
588 inp->inp_inc.inc_flags |= INC_ISIPV6;
589 if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
590 goto out;
591 if ((error = tcp6_connect(tp, nam, td)) != 0)
592 goto out;
593#ifdef TCP_OFFLOAD
594 if (registered_toedevs > 0 &&
595 (so->so_options & SO_NO_OFFLOAD) == 0 &&
596 (error = tcp_offload_connect(so, nam)) == 0)
597 goto out;
598#endif
599 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
600 error = tcp_output(tp);
601
602out:
603 TCPDEBUG2(PRU_CONNECT);
604 TCP_PROBE2(debug__user, tp, PRU_CONNECT);
605 INP_WUNLOCK(inp);
606 return (error);
607}
608#endif /* INET6 */
609
610/*
611 * Initiate disconnect from peer.
612 * If connection never passed embryonic stage, just drop;
613 * else if don't need to let data drain, then can just drop anyways,
614 * else have to begin TCP shutdown process: mark socket disconnecting,
615 * drain unread data, state switch to reflect user close, and
616 * send segment (e.g. FIN) to peer. Socket will be really disconnected
617 * when peer sends FIN and acks ours.
618 *
619 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
620 */
621static int
622tcp_usr_disconnect(struct socket *so)
623{
624 struct inpcb *inp;
625 struct tcpcb *tp = NULL;
626 int error = 0;
627
628 TCPDEBUG0;
629 INP_INFO_RLOCK(&V_tcbinfo);
630 inp = sotoinpcb(so);
631 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
632 INP_WLOCK(inp);
633 if (inp->inp_flags & INP_TIMEWAIT)
634 goto out;
635 if (inp->inp_flags & INP_DROPPED) {
636 error = ECONNRESET;
637 goto out;
638 }
639 tp = intotcpcb(inp);
640 TCPDEBUG1();
641 tcp_disconnect(tp);
642out:
643 TCPDEBUG2(PRU_DISCONNECT);
644 TCP_PROBE2(debug__user, tp, PRU_DISCONNECT);
645 INP_WUNLOCK(inp);
646 INP_INFO_RUNLOCK(&V_tcbinfo);
647 return (error);
648}
649
650#ifdef INET
651/*
652 * Accept a connection. Essentially all the work is done at higher levels;
653 * just return the address of the peer, storing through addr.
654 */
655static int
656tcp_usr_accept(struct socket *so, struct sockaddr **nam)
657{
658 int error = 0;
659 struct inpcb *inp = NULL;
660 struct tcpcb *tp = NULL;
661 struct in_addr addr;
662 in_port_t port = 0;
663 TCPDEBUG0;
664
665 if (so->so_state & SS_ISDISCONNECTED)
666 return (ECONNABORTED);
667
668 inp = sotoinpcb(so);
669 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
670 INP_WLOCK(inp);
671 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
672 error = ECONNABORTED;
673 goto out;
674 }
675 tp = intotcpcb(inp);
676 TCPDEBUG1();
677
678 /*
679 * We inline in_getpeeraddr and COMMON_END here, so that we can
680 * copy the data of interest and defer the malloc until after we
681 * release the lock.
682 */
683 port = inp->inp_fport;
684 addr = inp->inp_faddr;
685
686out:
687 TCPDEBUG2(PRU_ACCEPT);
688 TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
689 INP_WUNLOCK(inp);
690 if (error == 0)
691 *nam = in_sockaddr(port, &addr);
692 return error;
693}
694#endif /* INET */
695
696#ifdef INET6
697static int
698tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
699{
700 struct inpcb *inp = NULL;
701 int error = 0;
702 struct tcpcb *tp = NULL;
703 struct in_addr addr;
704 struct in6_addr addr6;
705 in_port_t port = 0;
706 int v4 = 0;
707 TCPDEBUG0;
708
709 if (so->so_state & SS_ISDISCONNECTED)
710 return (ECONNABORTED);
711
712 inp = sotoinpcb(so);
713 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
714 INP_INFO_RLOCK(&V_tcbinfo);
715 INP_WLOCK(inp);
716 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
717 error = ECONNABORTED;
718 goto out;
719 }
720 tp = intotcpcb(inp);
721 TCPDEBUG1();
722
723 /*
724 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
725 * copy the data of interest and defer the malloc until after we
726 * release the lock.
727 */
728 if (inp->inp_vflag & INP_IPV4) {
729 v4 = 1;
730 port = inp->inp_fport;
731 addr = inp->inp_faddr;
732 } else {
733 port = inp->inp_fport;
734 addr6 = inp->in6p_faddr;
735 }
736
737out:
738 TCPDEBUG2(PRU_ACCEPT);
739 TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
740 INP_WUNLOCK(inp);
741 INP_INFO_RUNLOCK(&V_tcbinfo);
742 if (error == 0) {
743 if (v4)
744 *nam = in6_v4mapsin6_sockaddr(port, &addr);
745 else
746 *nam = in6_sockaddr(port, &addr6);
747 }
748 return error;
749}
750#endif /* INET6 */
751
752/*
753 * Mark the connection as being incapable of further output.
754 */
755static int
756tcp_usr_shutdown(struct socket *so)
757{
758 int error = 0;
759 struct inpcb *inp;
760 struct tcpcb *tp = NULL;
761
762 TCPDEBUG0;
763 INP_INFO_RLOCK(&V_tcbinfo);
764 inp = sotoinpcb(so);
765 KASSERT(inp != NULL, ("inp == NULL"));
766 INP_WLOCK(inp);
767 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
768 error = ECONNRESET;
769 goto out;
770 }
771 tp = intotcpcb(inp);
772 TCPDEBUG1();
773 socantsendmore(so);
774 tcp_usrclosed(tp);
775 if (!(inp->inp_flags & INP_DROPPED))
776 error = tcp_output(tp);
777
778out:
779 TCPDEBUG2(PRU_SHUTDOWN);
780 TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN);
781 INP_WUNLOCK(inp);
782 INP_INFO_RUNLOCK(&V_tcbinfo);
783
784 return (error);
785}
786
787/*
788 * After a receive, possibly send window update to peer.
789 */
790static int
791tcp_usr_rcvd(struct socket *so, int flags)
792{
793 struct inpcb *inp;
794 struct tcpcb *tp = NULL;
795 int error = 0;
796
797 TCPDEBUG0;
798 inp = sotoinpcb(so);
799 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
800 INP_WLOCK(inp);
801 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
802 error = ECONNRESET;
803 goto out;
804 }
805 tp = intotcpcb(inp);
806 TCPDEBUG1();
807#ifdef TCP_OFFLOAD
808 if (tp->t_flags & TF_TOE)
809 tcp_offload_rcvd(tp);
810 else
811#endif
812 tcp_output(tp);
813
814out:
815 TCPDEBUG2(PRU_RCVD);
816 TCP_PROBE2(debug__user, tp, PRU_RCVD);
817 INP_WUNLOCK(inp);
818 return (error);
819}
820
821/*
822 * Do a send by putting data in output queue and updating urgent
823 * marker if URG set. Possibly send more data. Unlike the other
824 * pru_*() routines, the mbuf chains are our responsibility. We
825 * must either enqueue them or free them. The other pru_* routines
826 * generally are caller-frees.
827 */
828static int
829tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
830 struct sockaddr *nam, struct mbuf *control, struct thread *td)
831{
832 int error = 0;
833 struct inpcb *inp;
834 struct tcpcb *tp = NULL;
835#ifdef INET6
836 int isipv6;
837#endif
838 TCPDEBUG0;
839
840 /*
841 * We require the pcbinfo lock if we will close the socket as part of
842 * this call.
843 */
844 if (flags & PRUS_EOF)
845 INP_INFO_RLOCK(&V_tcbinfo);
846 inp = sotoinpcb(so);
847 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
848 INP_WLOCK(inp);
849 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
850 if (control)
851 m_freem(control);
852 /*
853 * In case of PRUS_NOTREADY, tcp_usr_ready() is responsible
854 * for freeing memory.
855 */
856 if (m && (flags & PRUS_NOTREADY) == 0)
857 m_freem(m);
858 error = ECONNRESET;
859 goto out;
860 }
861#ifdef INET6
862 isipv6 = nam && nam->sa_family == AF_INET6;
863#endif /* INET6 */
864 tp = intotcpcb(inp);
865 TCPDEBUG1();
866 if (control) {
867 /* TCP doesn't do control messages (rights, creds, etc) */
868 if (control->m_len) {
869 m_freem(control);
870 if (m)
871 m_freem(m);
872 error = EINVAL;
873 goto out;
874 }
875 m_freem(control); /* empty control, just free it */
876 }
877 if (!(flags & PRUS_OOB)) {
878 sbappendstream(&so->so_snd, m, flags);
879 if (nam && tp->t_state < TCPS_SYN_SENT) {
880 /*
881 * Do implied connect if not yet connected,
882 * initialize window to default value, and
883 * initialize maxseg/maxopd using peer's cached
884 * MSS.
885 */
886#ifdef INET6
887 if (isipv6)
888 error = tcp6_connect(tp, nam, td);
889#endif /* INET6 */
890#if defined(INET6) && defined(INET)
891 else
892#endif
893#ifdef INET
894 error = tcp_connect(tp, nam, td);
895#endif
896 if (error)
897 goto out;
898 tp->snd_wnd = TTCP_CLIENT_SND_WND;
899 tcp_mss(tp, -1);
900 }
901 if (flags & PRUS_EOF) {
902 /*
903 * Close the send side of the connection after
904 * the data is sent.
905 */
906 INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
907 socantsendmore(so);
908 tcp_usrclosed(tp);
909 }
910 if (!(inp->inp_flags & INP_DROPPED) &&
911 !(flags & PRUS_NOTREADY)) {
912 if (flags & PRUS_MORETOCOME)
913 tp->t_flags |= TF_MORETOCOME;
914 error = tcp_output(tp);
915 if (flags & PRUS_MORETOCOME)
916 tp->t_flags &= ~TF_MORETOCOME;
917 }
918 } else {
919 /*
920 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
921 */
922 SOCKBUF_LOCK(&so->so_snd);
923 if (sbspace(&so->so_snd) < -512) {
924 SOCKBUF_UNLOCK(&so->so_snd);
925 m_freem(m);
926 error = ENOBUFS;
927 goto out;
928 }
929 /*
930 * According to RFC961 (Assigned Protocols),
931 * the urgent pointer points to the last octet
932 * of urgent data. We continue, however,
933 * to consider it to indicate the first octet
934 * of data past the urgent section.
935 * Otherwise, snd_up should be one lower.
936 */
937 sbappendstream_locked(&so->so_snd, m, flags);
938 SOCKBUF_UNLOCK(&so->so_snd);
939 if (nam && tp->t_state < TCPS_SYN_SENT) {
940 /*
941 * Do implied connect if not yet connected,
942 * initialize window to default value, and
943 * initialize maxseg/maxopd using peer's cached
944 * MSS.
945 */
946#ifdef INET6
947 if (isipv6)
948 error = tcp6_connect(tp, nam, td);
949#endif /* INET6 */
950#if defined(INET6) && defined(INET)
951 else
952#endif
953#ifdef INET
954 error = tcp_connect(tp, nam, td);
955#endif
956 if (error)
957 goto out;
958 tp->snd_wnd = TTCP_CLIENT_SND_WND;
959 tcp_mss(tp, -1);
960 }
961 tp->snd_up = tp->snd_una + sbavail(&so->so_snd);
962 if (!(flags & PRUS_NOTREADY)) {
963 tp->t_flags |= TF_FORCEDATA;
964 error = tcp_output(tp);
965 tp->t_flags &= ~TF_FORCEDATA;
966 }
967 }
968out:
969 TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
970 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
971 TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB :
972 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
973 INP_WUNLOCK(inp);
974 if (flags & PRUS_EOF)
975 INP_INFO_RUNLOCK(&V_tcbinfo);
976 return (error);
977}
978
979static int
980tcp_usr_ready(struct socket *so, struct mbuf *m, int count)
981{
982 struct inpcb *inp;
983 struct tcpcb *tp;
984 int error;
985
986 inp = sotoinpcb(so);
987 INP_WLOCK(inp);
988 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
989 INP_WUNLOCK(inp);
990 for (int i = 0; i < count; i++)
991 m = m_free(m);
992 return (ECONNRESET);
993 }
994 tp = intotcpcb(inp);
995
996 SOCKBUF_LOCK(&so->so_snd);
997 error = sbready(&so->so_snd, m, count);
998 SOCKBUF_UNLOCK(&so->so_snd);
999 if (error == 0)
1000 error = tcp_output(tp);
1001 INP_WUNLOCK(inp);
1002
1003 return (error);
1004}
1005
1006/*
1007 * Abort the TCP. Drop the connection abruptly.
1008 */
1009static void
1010tcp_usr_abort(struct socket *so)
1011{
1012 struct inpcb *inp;
1013 struct tcpcb *tp = NULL;
1014 TCPDEBUG0;
1015
1016 inp = sotoinpcb(so);
1017 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
1018
1019 INP_INFO_RLOCK(&V_tcbinfo);
1020 INP_WLOCK(inp);
1021 KASSERT(inp->inp_socket != NULL,
1022 ("tcp_usr_abort: inp_socket == NULL"));
1023
1024 /*
1025 * If we still have full TCP state, and we're not dropped, drop.
1026 */
1027 if (!(inp->inp_flags & INP_TIMEWAIT) &&
1028 !(inp->inp_flags & INP_DROPPED)) {
1029 tp = intotcpcb(inp);
1030 TCPDEBUG1();
1031 tcp_drop(tp, ECONNABORTED);
1032 TCPDEBUG2(PRU_ABORT);
1033 TCP_PROBE2(debug__user, tp, PRU_ABORT);
1034 }
1035 if (!(inp->inp_flags & INP_DROPPED)) {
1036 SOCK_LOCK(so);
1037 so->so_state |= SS_PROTOREF;
1038 SOCK_UNLOCK(so);
1039 inp->inp_flags |= INP_SOCKREF;
1040 }
1041 INP_WUNLOCK(inp);
1042 INP_INFO_RUNLOCK(&V_tcbinfo);
1043}
1044
1045/*
1046 * TCP socket is closed. Start friendly disconnect.
1047 */
1048static void
1049tcp_usr_close(struct socket *so)
1050{
1051 struct inpcb *inp;
1052 struct tcpcb *tp = NULL;
1053 TCPDEBUG0;
1054
1055 inp = sotoinpcb(so);
1056 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
1057
1058 INP_INFO_RLOCK(&V_tcbinfo);
1059 INP_WLOCK(inp);
1060 KASSERT(inp->inp_socket != NULL,
1061 ("tcp_usr_close: inp_socket == NULL"));
1062
1063 /*
1064 * If we still have full TCP state, and we're not dropped, initiate
1065 * a disconnect.
1066 */
1067 if (!(inp->inp_flags & INP_TIMEWAIT) &&
1068 !(inp->inp_flags & INP_DROPPED)) {
1069 tp = intotcpcb(inp);
1070 TCPDEBUG1();
1071 tcp_disconnect(tp);
1072 TCPDEBUG2(PRU_CLOSE);
1073 TCP_PROBE2(debug__user, tp, PRU_CLOSE);
1074 }
1075 if (!(inp->inp_flags & INP_DROPPED)) {
1076 SOCK_LOCK(so);
1077 so->so_state |= SS_PROTOREF;
1078 SOCK_UNLOCK(so);
1079 inp->inp_flags |= INP_SOCKREF;
1080 }
1081 INP_WUNLOCK(inp);
1082 INP_INFO_RUNLOCK(&V_tcbinfo);
1083}
1084
1085/*
1086 * Receive out-of-band data.
1087 */
1088static int
1089tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1090{
1091 int error = 0;
1092 struct inpcb *inp;
1093 struct tcpcb *tp = NULL;
1094
1095 TCPDEBUG0;
1096 inp = sotoinpcb(so);
1097 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1098 INP_WLOCK(inp);
1099 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1100 error = ECONNRESET;
1101 goto out;
1102 }
1103 tp = intotcpcb(inp);
1104 TCPDEBUG1();
1105 if ((so->so_oobmark == 0 &&
1106 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1107 so->so_options & SO_OOBINLINE ||
1108 tp->t_oobflags & TCPOOB_HADDATA) {
1109 error = EINVAL;
1110 goto out;
1111 }
1112 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1113 error = EWOULDBLOCK;
1114 goto out;
1115 }
1116 m->m_len = 1;
1117 *mtod(m, caddr_t) = tp->t_iobc;
1118 if ((flags & MSG_PEEK) == 0)
1119 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1120
1121out:
1122 TCPDEBUG2(PRU_RCVOOB);
1123 TCP_PROBE2(debug__user, tp, PRU_RCVOOB);
1124 INP_WUNLOCK(inp);
1125 return (error);
1126}
1127
1128#ifdef INET
1129struct pr_usrreqs tcp_usrreqs = {
1130 .pru_abort = tcp_usr_abort,
1131 .pru_accept = tcp_usr_accept,
1132 .pru_attach = tcp_usr_attach,
1133 .pru_bind = tcp_usr_bind,
1134 .pru_connect = tcp_usr_connect,
1135 .pru_control = in_control,
1136 .pru_detach = tcp_usr_detach,
1137 .pru_disconnect = tcp_usr_disconnect,
1138 .pru_listen = tcp_usr_listen,
1139 .pru_peeraddr = in_getpeeraddr,
1140 .pru_rcvd = tcp_usr_rcvd,
1141 .pru_rcvoob = tcp_usr_rcvoob,
1142 .pru_send = tcp_usr_send,
1143 .pru_ready = tcp_usr_ready,
1144 .pru_shutdown = tcp_usr_shutdown,
1145 .pru_sockaddr = in_getsockaddr,
1146 .pru_sosetlabel = in_pcbsosetlabel,
1147 .pru_close = tcp_usr_close,
1148};
1149#endif /* INET */
1150
1151#ifdef INET6
1152struct pr_usrreqs tcp6_usrreqs = {
1153 .pru_abort = tcp_usr_abort,
1154 .pru_accept = tcp6_usr_accept,
1155 .pru_attach = tcp_usr_attach,
1156 .pru_bind = tcp6_usr_bind,
1157 .pru_connect = tcp6_usr_connect,
1158 .pru_control = in6_control,
1159 .pru_detach = tcp_usr_detach,
1160 .pru_disconnect = tcp_usr_disconnect,
1161 .pru_listen = tcp6_usr_listen,
1162 .pru_peeraddr = in6_mapped_peeraddr,
1163 .pru_rcvd = tcp_usr_rcvd,
1164 .pru_rcvoob = tcp_usr_rcvoob,
1165 .pru_send = tcp_usr_send,
1166 .pru_ready = tcp_usr_ready,
1167 .pru_shutdown = tcp_usr_shutdown,
1168 .pru_sockaddr = in6_mapped_sockaddr,
1169 .pru_sosetlabel = in_pcbsosetlabel,
1170 .pru_close = tcp_usr_close,
1171};
1172#endif /* INET6 */
1173
1174#ifdef INET
1175/*
1176 * Common subroutine to open a TCP connection to remote host specified
1177 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
1178 * port number if needed. Call in_pcbconnect_setup to do the routing and
1179 * to choose a local host address (interface). If there is an existing
1180 * incarnation of the same connection in TIME-WAIT state and if the remote
1181 * host was sending CC options and if the connection duration was < MSL, then
1182 * truncate the previous TIME-WAIT state and proceed.
1183 * Initialize connection parameters and enter SYN-SENT state.
1184 */
1185static int
1186tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1187{
1188 struct inpcb *inp = tp->t_inpcb, *oinp;
1189 struct socket *so = inp->inp_socket;
1190 struct in_addr laddr;
1191 u_short lport;
1192 int error;
1193
1194 INP_WLOCK_ASSERT(inp);
1195 INP_HASH_WLOCK(&V_tcbinfo);
1196
1197 if (inp->inp_lport == 0) {
1198 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1199 if (error)
1200 goto out;
1201 }
1202
1203 /*
1204 * Cannot simply call in_pcbconnect, because there might be an
1205 * earlier incarnation of this same connection still in
1206 * TIME_WAIT state, creating an ADDRINUSE error.
1207 */
1208 laddr = inp->inp_laddr;
1209 lport = inp->inp_lport;
1210 error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1211 &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1212 if (error && oinp == NULL)
1213 goto out;
1214 if (oinp) {
1215 error = EADDRINUSE;
1216 goto out;
1217 }
1218 inp->inp_laddr = laddr;
1219 in_pcbrehash(inp);
1220 INP_HASH_WUNLOCK(&V_tcbinfo);
1221
1222 /*
1223 * Compute window scaling to request:
1224 * Scale to fit into sweet spot. See tcp_syncache.c.
1225 * XXX: This should move to tcp_output().
1226 */
1227 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1228 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1229 tp->request_r_scale++;
1230
1231 soisconnecting(so);
1232 TCPSTAT_INC(tcps_connattempt);
1233 tcp_state_change(tp, TCPS_SYN_SENT);
1234 tp->iss = tcp_new_isn(tp);
1235 tcp_sendseqinit(tp);
1236
1237 return 0;
1238
1239out:
1240 INP_HASH_WUNLOCK(&V_tcbinfo);
1241 return (error);
1242}
1243#endif /* INET */
1244
1245#ifdef INET6
1246static int
1247tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1248{
1249 struct inpcb *inp = tp->t_inpcb;
1250 int error;
1251
1252 INP_WLOCK_ASSERT(inp);
1253 INP_HASH_WLOCK(&V_tcbinfo);
1254
1255 if (inp->inp_lport == 0) {
1256 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1257 if (error)
1258 goto out;
1259 }
1260 error = in6_pcbconnect(inp, nam, td->td_ucred);
1261 if (error != 0)
1262 goto out;
1263 INP_HASH_WUNLOCK(&V_tcbinfo);
1264
1265 /* Compute window scaling to request. */
1266 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1267 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1268 tp->request_r_scale++;
1269
1270 soisconnecting(inp->inp_socket);
1271 TCPSTAT_INC(tcps_connattempt);
1272 tcp_state_change(tp, TCPS_SYN_SENT);
1273 tp->iss = tcp_new_isn(tp);
1274 tcp_sendseqinit(tp);
1275
1276 return 0;
1277
1278out:
1279 INP_HASH_WUNLOCK(&V_tcbinfo);
1280 return error;
1281}
1282#endif /* INET6 */
1283
1284/*
1285 * Export TCP internal state information via a struct tcp_info, based on the
1286 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently
1287 * (TCP state machine, etc). We export all information using FreeBSD-native
1288 * constants -- for example, the numeric values for tcpi_state will differ
1289 * from Linux.
1290 */
1291static void
1292tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1293{
1294
1295 INP_WLOCK_ASSERT(tp->t_inpcb);
1296 bzero(ti, sizeof(*ti));
1297
1298 ti->tcpi_state = tp->t_state;
1299 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1300 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1301 if (tp->t_flags & TF_SACK_PERMIT)
1302 ti->tcpi_options |= TCPI_OPT_SACK;
1303 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1304 ti->tcpi_options |= TCPI_OPT_WSCALE;
1305 ti->tcpi_snd_wscale = tp->snd_scale;
1306 ti->tcpi_rcv_wscale = tp->rcv_scale;
1307 }
1308
1309 ti->tcpi_rto = tp->t_rxtcur * tick;
1310 ti->tcpi_last_data_recv = (long)(ticks - (int)tp->t_rcvtime) * tick;
1311 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1312 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1313
1314 ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1315 ti->tcpi_snd_cwnd = tp->snd_cwnd;
1316
1317 /*
1318 * FreeBSD-specific extension fields for tcp_info.
1319 */
1320 ti->tcpi_rcv_space = tp->rcv_wnd;
1321 ti->tcpi_rcv_nxt = tp->rcv_nxt;
1322 ti->tcpi_snd_wnd = tp->snd_wnd;
1323 ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */
1324 ti->tcpi_snd_nxt = tp->snd_nxt;
1325 ti->tcpi_snd_mss = tp->t_maxseg;
1326 ti->tcpi_rcv_mss = tp->t_maxseg;
1327 if (tp->t_flags & TF_TOE)
1328 ti->tcpi_options |= TCPI_OPT_TOE;
1329 ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1330 ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1331 ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1332}
1333
1334/*
1335 * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1336 * socket option arguments. When it re-acquires the lock after the copy, it
1337 * has to revalidate that the connection is still valid for the socket
1338 * option.
1339 */
1340#define INP_WLOCK_RECHECK(inp) do { \
1341 INP_WLOCK(inp); \
1342 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
1343 INP_WUNLOCK(inp); \
1344 return (ECONNRESET); \
1345 } \
1346 tp = intotcpcb(inp); \
1347} while(0)
1348
1349int
1350tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1351{
1352 int error, opt, optval;
1353 u_int ui;
1354 struct inpcb *inp;
1355 struct tcpcb *tp;
1356 struct tcp_info ti;
1357 char buf[TCP_CA_NAME_MAX];
1358 struct cc_algo *algo;
1359
1360 error = 0;
1361 inp = sotoinpcb(so);
1362 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1363 INP_WLOCK(inp);
1364 if (sopt->sopt_level != IPPROTO_TCP) {
1365#ifdef INET6
1366 if (inp->inp_vflag & INP_IPV6PROTO) {
1367 INP_WUNLOCK(inp);
1368 error = ip6_ctloutput(so, sopt);
1369 }
1370#endif /* INET6 */
1371#if defined(INET6) && defined(INET)
1372 else
1373#endif
1374#ifdef INET
1375 {
1376 INP_WUNLOCK(inp);
1377 error = ip_ctloutput(so, sopt);
1378 }
1379#endif
1380 return (error);
1381 }
1382 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1383 INP_WUNLOCK(inp);
1384 return (ECONNRESET);
1385 }
1386
1387 switch (sopt->sopt_dir) {
1388 case SOPT_SET:
1389 switch (sopt->sopt_name) {
1390#ifdef TCP_SIGNATURE
1391 case TCP_MD5SIG:
1392 INP_WUNLOCK(inp);
1393 error = sooptcopyin(sopt, &optval, sizeof optval,
1394 sizeof optval);
1395 if (error)
1396 return (error);
1397
1398 INP_WLOCK_RECHECK(inp);
1399 if (optval > 0)
1400 tp->t_flags |= TF_SIGNATURE;
1401 else
1402 tp->t_flags &= ~TF_SIGNATURE;
1403 goto unlock_and_done;
1404#endif /* TCP_SIGNATURE */
1405
1406 case TCP_NODELAY:
1407 case TCP_NOOPT:
1408 INP_WUNLOCK(inp);
1409 error = sooptcopyin(sopt, &optval, sizeof optval,
1410 sizeof optval);
1411 if (error)
1412 return (error);
1413
1414 INP_WLOCK_RECHECK(inp);
1415 switch (sopt->sopt_name) {
1416 case TCP_NODELAY:
1417 opt = TF_NODELAY;
1418 break;
1419 case TCP_NOOPT:
1420 opt = TF_NOOPT;
1421 break;
1422 default:
1423 opt = 0; /* dead code to fool gcc */
1424 break;
1425 }
1426
1427 if (optval)
1428 tp->t_flags |= opt;
1429 else
1430 tp->t_flags &= ~opt;
1431unlock_and_done:
1432#ifdef TCP_OFFLOAD
1433 if (tp->t_flags & TF_TOE) {
1434 tcp_offload_ctloutput(tp, sopt->sopt_dir,
1435 sopt->sopt_name);
1436 }
1437#endif
1438 INP_WUNLOCK(inp);
1439 break;
1440
1441 case TCP_NOPUSH:
1442 INP_WUNLOCK(inp);
1443 error = sooptcopyin(sopt, &optval, sizeof optval,
1444 sizeof optval);
1445 if (error)
1446 return (error);
1447
1448 INP_WLOCK_RECHECK(inp);
1449 if (optval)
1450 tp->t_flags |= TF_NOPUSH;
1451 else if (tp->t_flags & TF_NOPUSH) {
1452 tp->t_flags &= ~TF_NOPUSH;
1453 if (TCPS_HAVEESTABLISHED(tp->t_state))
1454 error = tcp_output(tp);
1455 }
1456 goto unlock_and_done;
1457
1458 case TCP_MAXSEG:
1459 INP_WUNLOCK(inp);
1460 error = sooptcopyin(sopt, &optval, sizeof optval,
1461 sizeof optval);
1462 if (error)
1463 return (error);
1464
1465 INP_WLOCK_RECHECK(inp);
1466 if (optval > 0 && optval <= tp->t_maxseg &&
1467 optval + 40 >= V_tcp_minmss)
1468 tp->t_maxseg = optval;
1469 else
1470 error = EINVAL;
1471 goto unlock_and_done;
1472
1473 case TCP_INFO:
1474 INP_WUNLOCK(inp);
1475 error = EINVAL;
1476 break;
1477
1478 case TCP_CONGESTION:
1479 INP_WUNLOCK(inp);
1480 bzero(buf, sizeof(buf));
1481 error = sooptcopyin(sopt, &buf, sizeof(buf), 1);
1482 if (error)
1483 break;
1484 INP_WLOCK_RECHECK(inp);
1485 /*
1486 * Return EINVAL if we can't find the requested cc algo.
1487 */
1488 error = EINVAL;
1489 CC_LIST_RLOCK();
1490 STAILQ_FOREACH(algo, &cc_list, entries) {
1491 if (strncmp(buf, algo->name, TCP_CA_NAME_MAX)
1492 == 0) {
1493 /* We've found the requested algo. */
1494 error = 0;
1495 /*
1496 * We hold a write lock over the tcb
1497 * so it's safe to do these things
1498 * without ordering concerns.
1499 */
1500 if (CC_ALGO(tp)->cb_destroy != NULL)
1501 CC_ALGO(tp)->cb_destroy(tp->ccv);
1502 CC_ALGO(tp) = algo;
1503 /*
1504 * If something goes pear shaped
1505 * initialising the new algo,
1506 * fall back to newreno (which
1507 * does not require initialisation).
1508 */
1509 if (algo->cb_init != NULL)
1510 if (algo->cb_init(tp->ccv) > 0) {
1511 CC_ALGO(tp) = &newreno_cc_algo;
1512 /*
1513 * The only reason init
1514 * should fail is
1515 * because of malloc.
1516 */
1517 error = ENOMEM;
1518 }
1519 break; /* Break the STAILQ_FOREACH. */
1520 }
1521 }
1522 CC_LIST_RUNLOCK();
1523 goto unlock_and_done;
1524
1525 case TCP_KEEPIDLE:
1526 case TCP_KEEPINTVL:
1527 case TCP_KEEPINIT:
1528 INP_WUNLOCK(inp);
1529 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1530 if (error)
1531 return (error);
1532
1533 if (ui > (UINT_MAX / hz)) {
1534 error = EINVAL;
1535 break;
1536 }
1537 ui *= hz;
1538
1539 INP_WLOCK_RECHECK(inp);
1540 switch (sopt->sopt_name) {
1541 case TCP_KEEPIDLE:
1542 tp->t_keepidle = ui;
1543 /*
1544 * XXX: better check current remaining
1545 * timeout and "merge" it with new value.
1546 */
1547 if ((tp->t_state > TCPS_LISTEN) &&
1548 (tp->t_state <= TCPS_CLOSING))
1549 tcp_timer_activate(tp, TT_KEEP,
1550 TP_KEEPIDLE(tp));
1551 break;
1552 case TCP_KEEPINTVL:
1553 tp->t_keepintvl = ui;
1554 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1555 (TP_MAXIDLE(tp) > 0))
1556 tcp_timer_activate(tp, TT_2MSL,
1557 TP_MAXIDLE(tp));
1558 break;
1559 case TCP_KEEPINIT:
1560 tp->t_keepinit = ui;
1561 if (tp->t_state == TCPS_SYN_RECEIVED ||
1562 tp->t_state == TCPS_SYN_SENT)
1563 tcp_timer_activate(tp, TT_KEEP,
1564 TP_KEEPINIT(tp));
1565 break;
1566 }
1567 goto unlock_and_done;
1568
1569 case TCP_KEEPCNT:
1570 INP_WUNLOCK(inp);
1571 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1572 if (error)
1573 return (error);
1574
1575 INP_WLOCK_RECHECK(inp);
1576 tp->t_keepcnt = ui;
1577 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1578 (TP_MAXIDLE(tp) > 0))
1579 tcp_timer_activate(tp, TT_2MSL,
1580 TP_MAXIDLE(tp));
1581 goto unlock_and_done;
1582
1583#ifdef TCPPCAP
1584 case TCP_PCAP_OUT:
1585 case TCP_PCAP_IN:
1586 INP_WUNLOCK(inp);
1587 error = sooptcopyin(sopt, &optval, sizeof optval,
1588 sizeof optval);
1589 if (error)
1590 return (error);
1591
1592 INP_WLOCK_RECHECK(inp);
1593 if (optval >= 0)
1594 tcp_pcap_set_sock_max(TCP_PCAP_OUT ?
1595 &(tp->t_outpkts) : &(tp->t_inpkts),
1596 optval);
1597 else
1598 error = EINVAL;
1599 goto unlock_and_done;
1600#endif
1601
1602 default:
1603 INP_WUNLOCK(inp);
1604 error = ENOPROTOOPT;
1605 break;
1606 }
1607 break;
1608
1609 case SOPT_GET:
1610 tp = intotcpcb(inp);
1611 switch (sopt->sopt_name) {
1612#ifdef TCP_SIGNATURE
1613 case TCP_MD5SIG:
1614 optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
1615 INP_WUNLOCK(inp);
1616 error = sooptcopyout(sopt, &optval, sizeof optval);
1617 break;
1618#endif
1619
1620 case TCP_NODELAY:
1621 optval = tp->t_flags & TF_NODELAY;
1622 INP_WUNLOCK(inp);
1623 error = sooptcopyout(sopt, &optval, sizeof optval);
1624 break;
1625 case TCP_MAXSEG:
1626 optval = tp->t_maxseg;
1627 INP_WUNLOCK(inp);
1628 error = sooptcopyout(sopt, &optval, sizeof optval);
1629 break;
1630 case TCP_NOOPT:
1631 optval = tp->t_flags & TF_NOOPT;
1632 INP_WUNLOCK(inp);
1633 error = sooptcopyout(sopt, &optval, sizeof optval);
1634 break;
1635 case TCP_NOPUSH:
1636 optval = tp->t_flags & TF_NOPUSH;
1637 INP_WUNLOCK(inp);
1638 error = sooptcopyout(sopt, &optval, sizeof optval);
1639 break;
1640 case TCP_INFO:
1641 tcp_fill_info(tp, &ti);
1642 INP_WUNLOCK(inp);
1643 error = sooptcopyout(sopt, &ti, sizeof ti);
1644 break;
1645 case TCP_CONGESTION:
1646 bzero(buf, sizeof(buf));
1647 strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
1648 INP_WUNLOCK(inp);
1649 error = sooptcopyout(sopt, buf, TCP_CA_NAME_MAX);
1650 break;
1651 case TCP_KEEPIDLE:
1652 case TCP_KEEPINTVL:
1653 case TCP_KEEPINIT:
1654 case TCP_KEEPCNT:
1655 switch (sopt->sopt_name) {
1656 case TCP_KEEPIDLE:
1657 ui = tp->t_keepidle / hz;
1658 break;
1659 case TCP_KEEPINTVL:
1660 ui = tp->t_keepintvl / hz;
1661 break;
1662 case TCP_KEEPINIT:
1663 ui = tp->t_keepinit / hz;
1664 break;
1665 case TCP_KEEPCNT:
1666 ui = tp->t_keepcnt;
1667 break;
1668 }
1669 INP_WUNLOCK(inp);
1670 error = sooptcopyout(sopt, &ui, sizeof(ui));
1671 break;
1672#ifdef TCPPCAP
1673 case TCP_PCAP_OUT:
1674 case TCP_PCAP_IN:
1675 optval = tcp_pcap_get_sock_max(TCP_PCAP_OUT ?
1676 &(tp->t_outpkts) : &(tp->t_inpkts));
1677 INP_WUNLOCK(inp);
1678 error = sooptcopyout(sopt, &optval, sizeof optval);
1679 break;
1680#endif
1681 default:
1682 INP_WUNLOCK(inp);
1683 error = ENOPROTOOPT;
1684 break;
1685 }
1686 break;
1687 }
1688 return (error);
1689}
1690#undef INP_WLOCK_RECHECK
1691
1692/*
1693 * Attach TCP protocol to socket, allocating
1694 * internet protocol control block, tcp control block,
1695 * bufer space, and entering LISTEN state if to accept connections.
1696 */
1697static int
1698tcp_attach(struct socket *so)
1699{
1700 struct tcpcb *tp;
1701 struct inpcb *inp;
1702 int error;
1703
1704 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1705 error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
1706 if (error)
1707 return (error);
1708 }
1709 so->so_rcv.sb_flags |= SB_AUTOSIZE;
1710 so->so_snd.sb_flags |= SB_AUTOSIZE;
1711 INP_INFO_RLOCK(&V_tcbinfo);
1712 error = in_pcballoc(so, &V_tcbinfo);
1713 if (error) {
1714 INP_INFO_RUNLOCK(&V_tcbinfo);
1715 return (error);
1716 }
1717 inp = sotoinpcb(so);
1718#ifdef INET6
1719 if (inp->inp_vflag & INP_IPV6PROTO) {
1720 inp->inp_vflag |= INP_IPV6;
1721 inp->in6p_hops = -1; /* use kernel default */
1722 }
1723 else
1724#endif
1725 inp->inp_vflag |= INP_IPV4;
1726 tp = tcp_newtcpcb(inp);
1727 if (tp == NULL) {
1728 in_pcbdetach(inp);
1729 in_pcbfree(inp);
1730 INP_INFO_RUNLOCK(&V_tcbinfo);
1731 return (ENOBUFS);
1732 }
1733 tp->t_state = TCPS_CLOSED;
1734 INP_WUNLOCK(inp);
1735 INP_INFO_RUNLOCK(&V_tcbinfo);
1736 return (0);
1737}
1738
1739/*
1740 * Initiate (or continue) disconnect.
1741 * If embryonic state, just send reset (once).
1742 * If in ``let data drain'' option and linger null, just drop.
1743 * Otherwise (hard), mark socket disconnecting and drop
1744 * current input data; switch states based on user close, and
1745 * send segment to peer (with FIN).
1746 */
1747static void
1748tcp_disconnect(struct tcpcb *tp)
1749{
1750 struct inpcb *inp = tp->t_inpcb;
1751 struct socket *so = inp->inp_socket;
1752
1753 INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
1754 INP_WLOCK_ASSERT(inp);
1755
1756 /*
1757 * Neither tcp_close() nor tcp_drop() should return NULL, as the
1758 * socket is still open.
1759 */
1760 if (tp->t_state < TCPS_ESTABLISHED) {
1761 tp = tcp_close(tp);
1762 KASSERT(tp != NULL,
1763 ("tcp_disconnect: tcp_close() returned NULL"));
1764 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
1765 tp = tcp_drop(tp, 0);
1766 KASSERT(tp != NULL,
1767 ("tcp_disconnect: tcp_drop() returned NULL"));
1768 } else {
1769 soisdisconnecting(so);
1770 sbflush(&so->so_rcv);
1771 tcp_usrclosed(tp);
1772 if (!(inp->inp_flags & INP_DROPPED))
1773 tcp_output(tp);
1774 }
1775}
1776
1777/*
1778 * User issued close, and wish to trail through shutdown states:
1779 * if never received SYN, just forget it. If got a SYN from peer,
1780 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1781 * If already got a FIN from peer, then almost done; go to LAST_ACK
1782 * state. In all other cases, have already sent FIN to peer (e.g.
1783 * after PRU_SHUTDOWN), and just have to play tedious game waiting
1784 * for peer to send FIN or not respond to keep-alives, etc.
1785 * We can let the user exit from the close as soon as the FIN is acked.
1786 */
1787static void
1788tcp_usrclosed(struct tcpcb *tp)
1789{
1790
1791 INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
1792 INP_WLOCK_ASSERT(tp->t_inpcb);
1793
1794 switch (tp->t_state) {
1795 case TCPS_LISTEN:
1796#ifdef TCP_OFFLOAD
1797 tcp_offload_listen_stop(tp);
1798#endif
1799 tcp_state_change(tp, TCPS_CLOSED);
1800 /* FALLTHROUGH */
1801 case TCPS_CLOSED:
1802 tp = tcp_close(tp);
1803 /*
1804 * tcp_close() should never return NULL here as the socket is
1805 * still open.
1806 */
1807 KASSERT(tp != NULL,
1808 ("tcp_usrclosed: tcp_close() returned NULL"));
1809 break;
1810
1811 case TCPS_SYN_SENT:
1812 case TCPS_SYN_RECEIVED:
1813 tp->t_flags |= TF_NEEDFIN;
1814 break;
1815
1816 case TCPS_ESTABLISHED:
1817 tcp_state_change(tp, TCPS_FIN_WAIT_1);
1818 break;
1819
1820 case TCPS_CLOSE_WAIT:
1821 tcp_state_change(tp, TCPS_LAST_ACK);
1822 break;
1823 }
1824 if (tp->t_state >= TCPS_FIN_WAIT_2) {
1825 soisdisconnected(tp->t_inpcb->inp_socket);
1826 /* Prevent the connection hanging in FIN_WAIT_2 forever. */
1827 if (tp->t_state == TCPS_FIN_WAIT_2) {
1828 int timeout;
1829
1830 timeout = (tcp_fast_finwait2_recycle) ?
1831 tcp_finwait2_timeout : TP_MAXIDLE(tp);
1832 tcp_timer_activate(tp, TT_2MSL, timeout);
1833 }
1834 }
1835}
1836
1837#ifdef DDB
1838static void
1839db_print_indent(int indent)
1840{
1841 int i;
1842
1843 for (i = 0; i < indent; i++)
1844 db_printf(" ");
1845}
1846
1847static void
1848db_print_tstate(int t_state)
1849{
1850
1851 switch (t_state) {
1852 case TCPS_CLOSED:
1853 db_printf("TCPS_CLOSED");
1854 return;
1855
1856 case TCPS_LISTEN:
1857 db_printf("TCPS_LISTEN");
1858 return;
1859
1860 case TCPS_SYN_SENT:
1861 db_printf("TCPS_SYN_SENT");
1862 return;
1863
1864 case TCPS_SYN_RECEIVED:
1865 db_printf("TCPS_SYN_RECEIVED");
1866 return;
1867
1868 case TCPS_ESTABLISHED:
1869 db_printf("TCPS_ESTABLISHED");
1870 return;
1871
1872 case TCPS_CLOSE_WAIT:
1873 db_printf("TCPS_CLOSE_WAIT");
1874 return;
1875
1876 case TCPS_FIN_WAIT_1:
1877 db_printf("TCPS_FIN_WAIT_1");
1878 return;
1879
1880 case TCPS_CLOSING:
1881 db_printf("TCPS_CLOSING");
1882 return;
1883
1884 case TCPS_LAST_ACK:
1885 db_printf("TCPS_LAST_ACK");
1886 return;
1887
1888 case TCPS_FIN_WAIT_2:
1889 db_printf("TCPS_FIN_WAIT_2");
1890 return;
1891
1892 case TCPS_TIME_WAIT:
1893 db_printf("TCPS_TIME_WAIT");
1894 return;
1895
1896 default:
1897 db_printf("unknown");
1898 return;
1899 }
1900}
1901
1902static void
1903db_print_tflags(u_int t_flags)
1904{
1905 int comma;
1906
1907 comma = 0;
1908 if (t_flags & TF_ACKNOW) {
1909 db_printf("%sTF_ACKNOW", comma ? ", " : "");
1910 comma = 1;
1911 }
1912 if (t_flags & TF_DELACK) {
1913 db_printf("%sTF_DELACK", comma ? ", " : "");
1914 comma = 1;
1915 }
1916 if (t_flags & TF_NODELAY) {
1917 db_printf("%sTF_NODELAY", comma ? ", " : "");
1918 comma = 1;
1919 }
1920 if (t_flags & TF_NOOPT) {
1921 db_printf("%sTF_NOOPT", comma ? ", " : "");
1922 comma = 1;
1923 }
1924 if (t_flags & TF_SENTFIN) {
1925 db_printf("%sTF_SENTFIN", comma ? ", " : "");
1926 comma = 1;
1927 }
1928 if (t_flags & TF_REQ_SCALE) {
1929 db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
1930 comma = 1;
1931 }
1932 if (t_flags & TF_RCVD_SCALE) {
1933 db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
1934 comma = 1;
1935 }
1936 if (t_flags & TF_REQ_TSTMP) {
1937 db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
1938 comma = 1;
1939 }
1940 if (t_flags & TF_RCVD_TSTMP) {
1941 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
1942 comma = 1;
1943 }
1944 if (t_flags & TF_SACK_PERMIT) {
1945 db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
1946 comma = 1;
1947 }
1948 if (t_flags & TF_NEEDSYN) {
1949 db_printf("%sTF_NEEDSYN", comma ? ", " : "");
1950 comma = 1;
1951 }
1952 if (t_flags & TF_NEEDFIN) {
1953 db_printf("%sTF_NEEDFIN", comma ? ", " : "");
1954 comma = 1;
1955 }
1956 if (t_flags & TF_NOPUSH) {
1957 db_printf("%sTF_NOPUSH", comma ? ", " : "");
1958 comma = 1;
1959 }
1960 if (t_flags & TF_MORETOCOME) {
1961 db_printf("%sTF_MORETOCOME", comma ? ", " : "");
1962 comma = 1;
1963 }
1964 if (t_flags & TF_LQ_OVERFLOW) {
1965 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
1966 comma = 1;
1967 }
1968 if (t_flags & TF_LASTIDLE) {
1969 db_printf("%sTF_LASTIDLE", comma ? ", " : "");
1970 comma = 1;
1971 }
1972 if (t_flags & TF_RXWIN0SENT) {
1973 db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
1974 comma = 1;
1975 }
1976 if (t_flags & TF_FASTRECOVERY) {
1977 db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
1978 comma = 1;
1979 }
1980 if (t_flags & TF_CONGRECOVERY) {
1981 db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
1982 comma = 1;
1983 }
1984 if (t_flags & TF_WASFRECOVERY) {
1985 db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
1986 comma = 1;
1987 }
1988 if (t_flags & TF_SIGNATURE) {
1989 db_printf("%sTF_SIGNATURE", comma ? ", " : "");
1990 comma = 1;
1991 }
1992 if (t_flags & TF_FORCEDATA) {
1993 db_printf("%sTF_FORCEDATA", comma ? ", " : "");
1994 comma = 1;
1995 }
1996 if (t_flags & TF_TSO) {
1997 db_printf("%sTF_TSO", comma ? ", " : "");
1998 comma = 1;
1999 }
2000 if (t_flags & TF_ECN_PERMIT) {
2001 db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
2002 comma = 1;
2003 }
2004}
2005
2006static void
2007db_print_toobflags(char t_oobflags)
2008{
2009 int comma;
2010
2011 comma = 0;
2012 if (t_oobflags & TCPOOB_HAVEDATA) {
2013 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
2014 comma = 1;
2015 }
2016 if (t_oobflags & TCPOOB_HADDATA) {
2017 db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
2018 comma = 1;
2019 }
2020}
2021
2022static void
2023db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
2024{
2025
2026 db_print_indent(indent);
2027 db_printf("%s at %p\n", name, tp);
2028
2029 indent += 2;
2030
2031 db_print_indent(indent);
2032 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n",
2033 LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
2034
2035 db_print_indent(indent);
2036 db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n",
2037 &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
2038
2039 db_print_indent(indent);
2040 db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", &tp->t_timers->tt_2msl,
2041 &tp->t_timers->tt_delack, tp->t_inpcb);
2042
2043 db_print_indent(indent);
2044 db_printf("t_state: %d (", tp->t_state);
2045 db_print_tstate(tp->t_state);
2046 db_printf(")\n");
2047
2048 db_print_indent(indent);
2049 db_printf("t_flags: 0x%x (", tp->t_flags);
2050 db_print_tflags(tp->t_flags);
2051 db_printf(")\n");
2052
2053 db_print_indent(indent);
2054 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n",
2055 tp->snd_una, tp->snd_max, tp->snd_nxt);
2056
2057 db_print_indent(indent);
2058 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n",
2059 tp->snd_up, tp->snd_wl1, tp->snd_wl2);
2060
2061 db_print_indent(indent);
2062 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n",
2063 tp->iss, tp->irs, tp->rcv_nxt);
2064
2065 db_print_indent(indent);
2066 db_printf("rcv_adv: 0x%08x rcv_wnd: %lu rcv_up: 0x%08x\n",
2067 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
2068
2069 db_print_indent(indent);
2070 db_printf("snd_wnd: %lu snd_cwnd: %lu\n",
2071 tp->snd_wnd, tp->snd_cwnd);
2072
2073 db_print_indent(indent);
2074 db_printf("snd_ssthresh: %lu snd_recover: "
2075 "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
2076
2077 db_print_indent(indent);
2078 db_printf("t_maxopd: %u t_rcvtime: %u t_startime: %u\n",
2079 tp->t_maxopd, tp->t_rcvtime, tp->t_starttime);
2080
2081 db_print_indent(indent);
2082 db_printf("t_rttime: %u t_rtsq: 0x%08x\n",
2083 tp->t_rtttime, tp->t_rtseq);
2084
2085 db_print_indent(indent);
2086 db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n",
2087 tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
2088
2089 db_print_indent(indent);
2090 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u "
2091 "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
2092 tp->t_rttbest);
2093
2094 db_print_indent(indent);
2095 db_printf("t_rttupdated: %lu max_sndwnd: %lu t_softerror: %d\n",
2096 tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
2097
2098 db_print_indent(indent);
2099 db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
2100 db_print_toobflags(tp->t_oobflags);
2101 db_printf(") t_iobc: 0x%02x\n", tp->t_iobc);
2102
2103 db_print_indent(indent);
2104 db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n",
2105 tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
2106
2107 db_print_indent(indent);
2108 db_printf("ts_recent: %u ts_recent_age: %u\n",
2109 tp->ts_recent, tp->ts_recent_age);
2110
2111 db_print_indent(indent);
2112 db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: "
2113 "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
2114
2115 db_print_indent(indent);
2116 db_printf("snd_ssthresh_prev: %lu snd_recover_prev: 0x%08x "
2117 "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
2118 tp->snd_recover_prev, tp->t_badrxtwin);
2119
2120 db_print_indent(indent);
2121 db_printf("snd_numholes: %d snd_holes first: %p\n",
2122 tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
2123
2124 db_print_indent(indent);
2125 db_printf("snd_fack: 0x%08x rcv_numsacks: %d sack_newdata: "
2126 "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
2127
2128 /* Skip sackblks, sackhint. */
2129
2130 db_print_indent(indent);
2131 db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n",
2132 tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
2133}
2134
2135DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
2136{
2137 struct tcpcb *tp;
2138
2139 if (!have_addr) {
2140 db_printf("usage: show tcpcb <addr>\n");
2141 return;
2142 }
2143 tp = (struct tcpcb *)addr;
2144
2145 db_print_tcpcb(tp, "tcpcb", 0);
2146}
2147#endif
2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California.
4 * Copyright (c) 2006-2007 Robert N. M. Watson
5 * Copyright (c) 2010-2011 Juniper Networks, Inc.
6 * All rights reserved.
7 *
8 * Portions of this software were developed by Robert N. M. Watson under
9 * contract to Juniper Networks, Inc.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 4. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 *
35 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94
36 */
37
38#include <sys/cdefs.h>
39__FBSDID("$FreeBSD: head/sys/netinet/tcp_usrreq.c 289276 2015-10-14 00:35:37Z hiren $");
40
41#include "opt_ddb.h"
42#include "opt_inet.h"
43#include "opt_inet6.h"
44#include "opt_tcpdebug.h"
45
46#include <sys/param.h>
47#include <sys/systm.h>
48#include <sys/limits.h>
49#include <sys/malloc.h>
50#include <sys/kernel.h>
51#include <sys/sysctl.h>
52#include <sys/mbuf.h>
53#ifdef INET6
54#include <sys/domain.h>
55#endif /* INET6 */
56#include <sys/socket.h>
57#include <sys/socketvar.h>
58#include <sys/protosw.h>
59#include <sys/proc.h>
60#include <sys/jail.h>
61
62#ifdef DDB
63#include <ddb/ddb.h>
64#endif
65
66#include <net/if.h>
67#include <net/if_var.h>
68#include <net/route.h>
69#include <net/vnet.h>
70
71#include <netinet/cc.h>
72#include <netinet/in.h>
73#include <netinet/in_kdtrace.h>
74#include <netinet/in_pcb.h>
75#include <netinet/in_systm.h>
76#include <netinet/in_var.h>
77#include <netinet/ip_var.h>
78#ifdef INET6
79#include <netinet/ip6.h>
80#include <netinet6/in6_pcb.h>
81#include <netinet6/ip6_var.h>
82#include <netinet6/scope6_var.h>
83#endif
84#include <netinet/tcp_fsm.h>
85#include <netinet/tcp_seq.h>
86#include <netinet/tcp_timer.h>
87#include <netinet/tcp_var.h>
88#include <netinet/tcpip.h>
89#ifdef TCPPCAP
90#include <netinet/tcp_pcap.h>
91#endif
92#ifdef TCPDEBUG
93#include <netinet/tcp_debug.h>
94#endif
95#ifdef TCP_OFFLOAD
96#include <netinet/tcp_offload.h>
97#endif
98
99/*
100 * TCP protocol interface to socket abstraction.
101 */
102static int tcp_attach(struct socket *);
103#ifdef INET
104static int tcp_connect(struct tcpcb *, struct sockaddr *,
105 struct thread *td);
106#endif /* INET */
107#ifdef INET6
108static int tcp6_connect(struct tcpcb *, struct sockaddr *,
109 struct thread *td);
110#endif /* INET6 */
111static void tcp_disconnect(struct tcpcb *);
112static void tcp_usrclosed(struct tcpcb *);
113static void tcp_fill_info(struct tcpcb *, struct tcp_info *);
114
115#ifdef TCPDEBUG
116#define TCPDEBUG0 int ostate = 0
117#define TCPDEBUG1() ostate = tp ? tp->t_state : 0
118#define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
119 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
120#else
121#define TCPDEBUG0
122#define TCPDEBUG1()
123#define TCPDEBUG2(req)
124#endif
125
126/*
127 * TCP attaches to socket via pru_attach(), reserving space,
128 * and an internet control block.
129 */
130static int
131tcp_usr_attach(struct socket *so, int proto, struct thread *td)
132{
133 struct inpcb *inp;
134 struct tcpcb *tp = NULL;
135 int error;
136 TCPDEBUG0;
137
138 inp = sotoinpcb(so);
139 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
140 TCPDEBUG1();
141
142 error = tcp_attach(so);
143 if (error)
144 goto out;
145
146 if ((so->so_options & SO_LINGER) && so->so_linger == 0)
147 so->so_linger = TCP_LINGERTIME;
148
149 inp = sotoinpcb(so);
150 tp = intotcpcb(inp);
151out:
152 TCPDEBUG2(PRU_ATTACH);
153 TCP_PROBE2(debug__user, tp, PRU_ATTACH);
154 return error;
155}
156
157/*
158 * tcp_detach is called when the socket layer loses its final reference
159 * to the socket, be it a file descriptor reference, a reference from TCP,
160 * etc. At this point, there is only one case in which we will keep around
161 * inpcb state: time wait.
162 *
163 * This function can probably be re-absorbed back into tcp_usr_detach() now
164 * that there is a single detach path.
165 */
166static void
167tcp_detach(struct socket *so, struct inpcb *inp)
168{
169 struct tcpcb *tp;
170
171 INP_INFO_LOCK_ASSERT(&V_tcbinfo);
172 INP_WLOCK_ASSERT(inp);
173
174 KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
175 KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
176
177 tp = intotcpcb(inp);
178
179 if (inp->inp_flags & INP_TIMEWAIT) {
180 /*
181 * There are two cases to handle: one in which the time wait
182 * state is being discarded (INP_DROPPED), and one in which
183 * this connection will remain in timewait. In the former,
184 * it is time to discard all state (except tcptw, which has
185 * already been discarded by the timewait close code, which
186 * should be further up the call stack somewhere). In the
187 * latter case, we detach from the socket, but leave the pcb
188 * present until timewait ends.
189 *
190 * XXXRW: Would it be cleaner to free the tcptw here?
191 *
192 * Astute question indeed, from twtcp perspective there are
193 * three cases to consider:
194 *
195 * #1 tcp_detach is called at tcptw creation time by
196 * tcp_twstart, then do not discard the newly created tcptw
197 * and leave inpcb present until timewait ends
198 * #2 tcp_detach is called at timewait end (or reuse) by
199 * tcp_twclose, then the tcptw has already been discarded
200 * (or reused) and inpcb is freed here
201 * #3 tcp_detach is called() after timewait ends (or reuse)
202 * (e.g. by soclose), then tcptw has already been discarded
203 * (or reused) and inpcb is freed here
204 *
205 * In all three cases the tcptw should not be freed here.
206 */
207 if (inp->inp_flags & INP_DROPPED) {
208 KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && "
209 "INP_DROPPED && tp != NULL"));
210 in_pcbdetach(inp);
211 in_pcbfree(inp);
212 } else {
213 in_pcbdetach(inp);
214 INP_WUNLOCK(inp);
215 }
216 } else {
217 /*
218 * If the connection is not in timewait, we consider two
219 * two conditions: one in which no further processing is
220 * necessary (dropped || embryonic), and one in which TCP is
221 * not yet done, but no longer requires the socket, so the
222 * pcb will persist for the time being.
223 *
224 * XXXRW: Does the second case still occur?
225 */
226 if (inp->inp_flags & INP_DROPPED ||
227 tp->t_state < TCPS_SYN_SENT) {
228 tcp_discardcb(tp);
229 in_pcbdetach(inp);
230 in_pcbfree(inp);
231 } else {
232 in_pcbdetach(inp);
233 INP_WUNLOCK(inp);
234 }
235 }
236}
237
238/*
239 * pru_detach() detaches the TCP protocol from the socket.
240 * If the protocol state is non-embryonic, then can't
241 * do this directly: have to initiate a pru_disconnect(),
242 * which may finish later; embryonic TCB's can just
243 * be discarded here.
244 */
245static void
246tcp_usr_detach(struct socket *so)
247{
248 struct inpcb *inp;
249 int rlock = 0;
250
251 inp = sotoinpcb(so);
252 KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
253 if (!INP_INFO_WLOCKED(&V_tcbinfo)) {
254 INP_INFO_RLOCK(&V_tcbinfo);
255 rlock = 1;
256 }
257 INP_WLOCK(inp);
258 KASSERT(inp->inp_socket != NULL,
259 ("tcp_usr_detach: inp_socket == NULL"));
260 tcp_detach(so, inp);
261 if (rlock)
262 INP_INFO_RUNLOCK(&V_tcbinfo);
263}
264
265#ifdef INET
266/*
267 * Give the socket an address.
268 */
269static int
270tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
271{
272 int error = 0;
273 struct inpcb *inp;
274 struct tcpcb *tp = NULL;
275 struct sockaddr_in *sinp;
276
277 sinp = (struct sockaddr_in *)nam;
278 if (nam->sa_len != sizeof (*sinp))
279 return (EINVAL);
280 /*
281 * Must check for multicast addresses and disallow binding
282 * to them.
283 */
284 if (sinp->sin_family == AF_INET &&
285 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
286 return (EAFNOSUPPORT);
287
288 TCPDEBUG0;
289 inp = sotoinpcb(so);
290 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
291 INP_WLOCK(inp);
292 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
293 error = EINVAL;
294 goto out;
295 }
296 tp = intotcpcb(inp);
297 TCPDEBUG1();
298 INP_HASH_WLOCK(&V_tcbinfo);
299 error = in_pcbbind(inp, nam, td->td_ucred);
300 INP_HASH_WUNLOCK(&V_tcbinfo);
301out:
302 TCPDEBUG2(PRU_BIND);
303 TCP_PROBE2(debug__user, tp, PRU_BIND);
304 INP_WUNLOCK(inp);
305
306 return (error);
307}
308#endif /* INET */
309
310#ifdef INET6
311static int
312tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
313{
314 int error = 0;
315 struct inpcb *inp;
316 struct tcpcb *tp = NULL;
317 struct sockaddr_in6 *sin6p;
318
319 sin6p = (struct sockaddr_in6 *)nam;
320 if (nam->sa_len != sizeof (*sin6p))
321 return (EINVAL);
322 /*
323 * Must check for multicast addresses and disallow binding
324 * to them.
325 */
326 if (sin6p->sin6_family == AF_INET6 &&
327 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
328 return (EAFNOSUPPORT);
329
330 TCPDEBUG0;
331 inp = sotoinpcb(so);
332 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
333 INP_WLOCK(inp);
334 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
335 error = EINVAL;
336 goto out;
337 }
338 tp = intotcpcb(inp);
339 TCPDEBUG1();
340 INP_HASH_WLOCK(&V_tcbinfo);
341 inp->inp_vflag &= ~INP_IPV4;
342 inp->inp_vflag |= INP_IPV6;
343#ifdef INET
344 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
345 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
346 inp->inp_vflag |= INP_IPV4;
347 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
348 struct sockaddr_in sin;
349
350 in6_sin6_2_sin(&sin, sin6p);
351 inp->inp_vflag |= INP_IPV4;
352 inp->inp_vflag &= ~INP_IPV6;
353 error = in_pcbbind(inp, (struct sockaddr *)&sin,
354 td->td_ucred);
355 INP_HASH_WUNLOCK(&V_tcbinfo);
356 goto out;
357 }
358 }
359#endif
360 error = in6_pcbbind(inp, nam, td->td_ucred);
361 INP_HASH_WUNLOCK(&V_tcbinfo);
362out:
363 TCPDEBUG2(PRU_BIND);
364 TCP_PROBE2(debug__user, tp, PRU_BIND);
365 INP_WUNLOCK(inp);
366 return (error);
367}
368#endif /* INET6 */
369
370#ifdef INET
371/*
372 * Prepare to accept connections.
373 */
374static int
375tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
376{
377 int error = 0;
378 struct inpcb *inp;
379 struct tcpcb *tp = NULL;
380
381 TCPDEBUG0;
382 inp = sotoinpcb(so);
383 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
384 INP_WLOCK(inp);
385 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
386 error = EINVAL;
387 goto out;
388 }
389 tp = intotcpcb(inp);
390 TCPDEBUG1();
391 SOCK_LOCK(so);
392 error = solisten_proto_check(so);
393 INP_HASH_WLOCK(&V_tcbinfo);
394 if (error == 0 && inp->inp_lport == 0)
395 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
396 INP_HASH_WUNLOCK(&V_tcbinfo);
397 if (error == 0) {
398 tcp_state_change(tp, TCPS_LISTEN);
399 solisten_proto(so, backlog);
400#ifdef TCP_OFFLOAD
401 if ((so->so_options & SO_NO_OFFLOAD) == 0)
402 tcp_offload_listen_start(tp);
403#endif
404 }
405 SOCK_UNLOCK(so);
406
407out:
408 TCPDEBUG2(PRU_LISTEN);
409 TCP_PROBE2(debug__user, tp, PRU_LISTEN);
410 INP_WUNLOCK(inp);
411 return (error);
412}
413#endif /* INET */
414
415#ifdef INET6
416static int
417tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
418{
419 int error = 0;
420 struct inpcb *inp;
421 struct tcpcb *tp = NULL;
422
423 TCPDEBUG0;
424 inp = sotoinpcb(so);
425 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
426 INP_WLOCK(inp);
427 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
428 error = EINVAL;
429 goto out;
430 }
431 tp = intotcpcb(inp);
432 TCPDEBUG1();
433 SOCK_LOCK(so);
434 error = solisten_proto_check(so);
435 INP_HASH_WLOCK(&V_tcbinfo);
436 if (error == 0 && inp->inp_lport == 0) {
437 inp->inp_vflag &= ~INP_IPV4;
438 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
439 inp->inp_vflag |= INP_IPV4;
440 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
441 }
442 INP_HASH_WUNLOCK(&V_tcbinfo);
443 if (error == 0) {
444 tcp_state_change(tp, TCPS_LISTEN);
445 solisten_proto(so, backlog);
446#ifdef TCP_OFFLOAD
447 if ((so->so_options & SO_NO_OFFLOAD) == 0)
448 tcp_offload_listen_start(tp);
449#endif
450 }
451 SOCK_UNLOCK(so);
452
453out:
454 TCPDEBUG2(PRU_LISTEN);
455 TCP_PROBE2(debug__user, tp, PRU_LISTEN);
456 INP_WUNLOCK(inp);
457 return (error);
458}
459#endif /* INET6 */
460
461#ifdef INET
462/*
463 * Initiate connection to peer.
464 * Create a template for use in transmissions on this connection.
465 * Enter SYN_SENT state, and mark socket as connecting.
466 * Start keep-alive timer, and seed output sequence space.
467 * Send initial segment on connection.
468 */
469static int
470tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
471{
472 int error = 0;
473 struct inpcb *inp;
474 struct tcpcb *tp = NULL;
475 struct sockaddr_in *sinp;
476
477 sinp = (struct sockaddr_in *)nam;
478 if (nam->sa_len != sizeof (*sinp))
479 return (EINVAL);
480 /*
481 * Must disallow TCP ``connections'' to multicast addresses.
482 */
483 if (sinp->sin_family == AF_INET
484 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
485 return (EAFNOSUPPORT);
486 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
487 return (error);
488
489 TCPDEBUG0;
490 inp = sotoinpcb(so);
491 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
492 INP_WLOCK(inp);
493 if (inp->inp_flags & INP_TIMEWAIT) {
494 error = EADDRINUSE;
495 goto out;
496 }
497 if (inp->inp_flags & INP_DROPPED) {
498 error = ECONNREFUSED;
499 goto out;
500 }
501 tp = intotcpcb(inp);
502 TCPDEBUG1();
503 if ((error = tcp_connect(tp, nam, td)) != 0)
504 goto out;
505#ifdef TCP_OFFLOAD
506 if (registered_toedevs > 0 &&
507 (so->so_options & SO_NO_OFFLOAD) == 0 &&
508 (error = tcp_offload_connect(so, nam)) == 0)
509 goto out;
510#endif
511 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
512 error = tcp_output(tp);
513out:
514 TCPDEBUG2(PRU_CONNECT);
515 INP_WUNLOCK(inp);
516 return (error);
517}
518#endif /* INET */
519
520#ifdef INET6
521static int
522tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
523{
524 int error = 0;
525 struct inpcb *inp;
526 struct tcpcb *tp = NULL;
527 struct sockaddr_in6 *sin6p;
528
529 TCPDEBUG0;
530
531 sin6p = (struct sockaddr_in6 *)nam;
532 if (nam->sa_len != sizeof (*sin6p))
533 return (EINVAL);
534 /*
535 * Must disallow TCP ``connections'' to multicast addresses.
536 */
537 if (sin6p->sin6_family == AF_INET6
538 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
539 return (EAFNOSUPPORT);
540
541 inp = sotoinpcb(so);
542 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
543 INP_WLOCK(inp);
544 if (inp->inp_flags & INP_TIMEWAIT) {
545 error = EADDRINUSE;
546 goto out;
547 }
548 if (inp->inp_flags & INP_DROPPED) {
549 error = ECONNREFUSED;
550 goto out;
551 }
552 tp = intotcpcb(inp);
553 TCPDEBUG1();
554#ifdef INET
555 /*
556 * XXXRW: Some confusion: V4/V6 flags relate to binding, and
557 * therefore probably require the hash lock, which isn't held here.
558 * Is this a significant problem?
559 */
560 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
561 struct sockaddr_in sin;
562
563 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
564 error = EINVAL;
565 goto out;
566 }
567
568 in6_sin6_2_sin(&sin, sin6p);
569 inp->inp_vflag |= INP_IPV4;
570 inp->inp_vflag &= ~INP_IPV6;
571 if ((error = prison_remote_ip4(td->td_ucred,
572 &sin.sin_addr)) != 0)
573 goto out;
574 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
575 goto out;
576#ifdef TCP_OFFLOAD
577 if (registered_toedevs > 0 &&
578 (so->so_options & SO_NO_OFFLOAD) == 0 &&
579 (error = tcp_offload_connect(so, nam)) == 0)
580 goto out;
581#endif
582 error = tcp_output(tp);
583 goto out;
584 }
585#endif
586 inp->inp_vflag &= ~INP_IPV4;
587 inp->inp_vflag |= INP_IPV6;
588 inp->inp_inc.inc_flags |= INC_ISIPV6;
589 if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
590 goto out;
591 if ((error = tcp6_connect(tp, nam, td)) != 0)
592 goto out;
593#ifdef TCP_OFFLOAD
594 if (registered_toedevs > 0 &&
595 (so->so_options & SO_NO_OFFLOAD) == 0 &&
596 (error = tcp_offload_connect(so, nam)) == 0)
597 goto out;
598#endif
599 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
600 error = tcp_output(tp);
601
602out:
603 TCPDEBUG2(PRU_CONNECT);
604 TCP_PROBE2(debug__user, tp, PRU_CONNECT);
605 INP_WUNLOCK(inp);
606 return (error);
607}
608#endif /* INET6 */
609
610/*
611 * Initiate disconnect from peer.
612 * If connection never passed embryonic stage, just drop;
613 * else if don't need to let data drain, then can just drop anyways,
614 * else have to begin TCP shutdown process: mark socket disconnecting,
615 * drain unread data, state switch to reflect user close, and
616 * send segment (e.g. FIN) to peer. Socket will be really disconnected
617 * when peer sends FIN and acks ours.
618 *
619 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
620 */
621static int
622tcp_usr_disconnect(struct socket *so)
623{
624 struct inpcb *inp;
625 struct tcpcb *tp = NULL;
626 int error = 0;
627
628 TCPDEBUG0;
629 INP_INFO_RLOCK(&V_tcbinfo);
630 inp = sotoinpcb(so);
631 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
632 INP_WLOCK(inp);
633 if (inp->inp_flags & INP_TIMEWAIT)
634 goto out;
635 if (inp->inp_flags & INP_DROPPED) {
636 error = ECONNRESET;
637 goto out;
638 }
639 tp = intotcpcb(inp);
640 TCPDEBUG1();
641 tcp_disconnect(tp);
642out:
643 TCPDEBUG2(PRU_DISCONNECT);
644 TCP_PROBE2(debug__user, tp, PRU_DISCONNECT);
645 INP_WUNLOCK(inp);
646 INP_INFO_RUNLOCK(&V_tcbinfo);
647 return (error);
648}
649
650#ifdef INET
651/*
652 * Accept a connection. Essentially all the work is done at higher levels;
653 * just return the address of the peer, storing through addr.
654 */
655static int
656tcp_usr_accept(struct socket *so, struct sockaddr **nam)
657{
658 int error = 0;
659 struct inpcb *inp = NULL;
660 struct tcpcb *tp = NULL;
661 struct in_addr addr;
662 in_port_t port = 0;
663 TCPDEBUG0;
664
665 if (so->so_state & SS_ISDISCONNECTED)
666 return (ECONNABORTED);
667
668 inp = sotoinpcb(so);
669 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
670 INP_WLOCK(inp);
671 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
672 error = ECONNABORTED;
673 goto out;
674 }
675 tp = intotcpcb(inp);
676 TCPDEBUG1();
677
678 /*
679 * We inline in_getpeeraddr and COMMON_END here, so that we can
680 * copy the data of interest and defer the malloc until after we
681 * release the lock.
682 */
683 port = inp->inp_fport;
684 addr = inp->inp_faddr;
685
686out:
687 TCPDEBUG2(PRU_ACCEPT);
688 TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
689 INP_WUNLOCK(inp);
690 if (error == 0)
691 *nam = in_sockaddr(port, &addr);
692 return error;
693}
694#endif /* INET */
695
696#ifdef INET6
697static int
698tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
699{
700 struct inpcb *inp = NULL;
701 int error = 0;
702 struct tcpcb *tp = NULL;
703 struct in_addr addr;
704 struct in6_addr addr6;
705 in_port_t port = 0;
706 int v4 = 0;
707 TCPDEBUG0;
708
709 if (so->so_state & SS_ISDISCONNECTED)
710 return (ECONNABORTED);
711
712 inp = sotoinpcb(so);
713 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
714 INP_INFO_RLOCK(&V_tcbinfo);
715 INP_WLOCK(inp);
716 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
717 error = ECONNABORTED;
718 goto out;
719 }
720 tp = intotcpcb(inp);
721 TCPDEBUG1();
722
723 /*
724 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
725 * copy the data of interest and defer the malloc until after we
726 * release the lock.
727 */
728 if (inp->inp_vflag & INP_IPV4) {
729 v4 = 1;
730 port = inp->inp_fport;
731 addr = inp->inp_faddr;
732 } else {
733 port = inp->inp_fport;
734 addr6 = inp->in6p_faddr;
735 }
736
737out:
738 TCPDEBUG2(PRU_ACCEPT);
739 TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
740 INP_WUNLOCK(inp);
741 INP_INFO_RUNLOCK(&V_tcbinfo);
742 if (error == 0) {
743 if (v4)
744 *nam = in6_v4mapsin6_sockaddr(port, &addr);
745 else
746 *nam = in6_sockaddr(port, &addr6);
747 }
748 return error;
749}
750#endif /* INET6 */
751
752/*
753 * Mark the connection as being incapable of further output.
754 */
755static int
756tcp_usr_shutdown(struct socket *so)
757{
758 int error = 0;
759 struct inpcb *inp;
760 struct tcpcb *tp = NULL;
761
762 TCPDEBUG0;
763 INP_INFO_RLOCK(&V_tcbinfo);
764 inp = sotoinpcb(so);
765 KASSERT(inp != NULL, ("inp == NULL"));
766 INP_WLOCK(inp);
767 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
768 error = ECONNRESET;
769 goto out;
770 }
771 tp = intotcpcb(inp);
772 TCPDEBUG1();
773 socantsendmore(so);
774 tcp_usrclosed(tp);
775 if (!(inp->inp_flags & INP_DROPPED))
776 error = tcp_output(tp);
777
778out:
779 TCPDEBUG2(PRU_SHUTDOWN);
780 TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN);
781 INP_WUNLOCK(inp);
782 INP_INFO_RUNLOCK(&V_tcbinfo);
783
784 return (error);
785}
786
787/*
788 * After a receive, possibly send window update to peer.
789 */
790static int
791tcp_usr_rcvd(struct socket *so, int flags)
792{
793 struct inpcb *inp;
794 struct tcpcb *tp = NULL;
795 int error = 0;
796
797 TCPDEBUG0;
798 inp = sotoinpcb(so);
799 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
800 INP_WLOCK(inp);
801 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
802 error = ECONNRESET;
803 goto out;
804 }
805 tp = intotcpcb(inp);
806 TCPDEBUG1();
807#ifdef TCP_OFFLOAD
808 if (tp->t_flags & TF_TOE)
809 tcp_offload_rcvd(tp);
810 else
811#endif
812 tcp_output(tp);
813
814out:
815 TCPDEBUG2(PRU_RCVD);
816 TCP_PROBE2(debug__user, tp, PRU_RCVD);
817 INP_WUNLOCK(inp);
818 return (error);
819}
820
821/*
822 * Do a send by putting data in output queue and updating urgent
823 * marker if URG set. Possibly send more data. Unlike the other
824 * pru_*() routines, the mbuf chains are our responsibility. We
825 * must either enqueue them or free them. The other pru_* routines
826 * generally are caller-frees.
827 */
828static int
829tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
830 struct sockaddr *nam, struct mbuf *control, struct thread *td)
831{
832 int error = 0;
833 struct inpcb *inp;
834 struct tcpcb *tp = NULL;
835#ifdef INET6
836 int isipv6;
837#endif
838 TCPDEBUG0;
839
840 /*
841 * We require the pcbinfo lock if we will close the socket as part of
842 * this call.
843 */
844 if (flags & PRUS_EOF)
845 INP_INFO_RLOCK(&V_tcbinfo);
846 inp = sotoinpcb(so);
847 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
848 INP_WLOCK(inp);
849 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
850 if (control)
851 m_freem(control);
852 /*
853 * In case of PRUS_NOTREADY, tcp_usr_ready() is responsible
854 * for freeing memory.
855 */
856 if (m && (flags & PRUS_NOTREADY) == 0)
857 m_freem(m);
858 error = ECONNRESET;
859 goto out;
860 }
861#ifdef INET6
862 isipv6 = nam && nam->sa_family == AF_INET6;
863#endif /* INET6 */
864 tp = intotcpcb(inp);
865 TCPDEBUG1();
866 if (control) {
867 /* TCP doesn't do control messages (rights, creds, etc) */
868 if (control->m_len) {
869 m_freem(control);
870 if (m)
871 m_freem(m);
872 error = EINVAL;
873 goto out;
874 }
875 m_freem(control); /* empty control, just free it */
876 }
877 if (!(flags & PRUS_OOB)) {
878 sbappendstream(&so->so_snd, m, flags);
879 if (nam && tp->t_state < TCPS_SYN_SENT) {
880 /*
881 * Do implied connect if not yet connected,
882 * initialize window to default value, and
883 * initialize maxseg/maxopd using peer's cached
884 * MSS.
885 */
886#ifdef INET6
887 if (isipv6)
888 error = tcp6_connect(tp, nam, td);
889#endif /* INET6 */
890#if defined(INET6) && defined(INET)
891 else
892#endif
893#ifdef INET
894 error = tcp_connect(tp, nam, td);
895#endif
896 if (error)
897 goto out;
898 tp->snd_wnd = TTCP_CLIENT_SND_WND;
899 tcp_mss(tp, -1);
900 }
901 if (flags & PRUS_EOF) {
902 /*
903 * Close the send side of the connection after
904 * the data is sent.
905 */
906 INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
907 socantsendmore(so);
908 tcp_usrclosed(tp);
909 }
910 if (!(inp->inp_flags & INP_DROPPED) &&
911 !(flags & PRUS_NOTREADY)) {
912 if (flags & PRUS_MORETOCOME)
913 tp->t_flags |= TF_MORETOCOME;
914 error = tcp_output(tp);
915 if (flags & PRUS_MORETOCOME)
916 tp->t_flags &= ~TF_MORETOCOME;
917 }
918 } else {
919 /*
920 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
921 */
922 SOCKBUF_LOCK(&so->so_snd);
923 if (sbspace(&so->so_snd) < -512) {
924 SOCKBUF_UNLOCK(&so->so_snd);
925 m_freem(m);
926 error = ENOBUFS;
927 goto out;
928 }
929 /*
930 * According to RFC961 (Assigned Protocols),
931 * the urgent pointer points to the last octet
932 * of urgent data. We continue, however,
933 * to consider it to indicate the first octet
934 * of data past the urgent section.
935 * Otherwise, snd_up should be one lower.
936 */
937 sbappendstream_locked(&so->so_snd, m, flags);
938 SOCKBUF_UNLOCK(&so->so_snd);
939 if (nam && tp->t_state < TCPS_SYN_SENT) {
940 /*
941 * Do implied connect if not yet connected,
942 * initialize window to default value, and
943 * initialize maxseg/maxopd using peer's cached
944 * MSS.
945 */
946#ifdef INET6
947 if (isipv6)
948 error = tcp6_connect(tp, nam, td);
949#endif /* INET6 */
950#if defined(INET6) && defined(INET)
951 else
952#endif
953#ifdef INET
954 error = tcp_connect(tp, nam, td);
955#endif
956 if (error)
957 goto out;
958 tp->snd_wnd = TTCP_CLIENT_SND_WND;
959 tcp_mss(tp, -1);
960 }
961 tp->snd_up = tp->snd_una + sbavail(&so->so_snd);
962 if (!(flags & PRUS_NOTREADY)) {
963 tp->t_flags |= TF_FORCEDATA;
964 error = tcp_output(tp);
965 tp->t_flags &= ~TF_FORCEDATA;
966 }
967 }
968out:
969 TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
970 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
971 TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB :
972 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
973 INP_WUNLOCK(inp);
974 if (flags & PRUS_EOF)
975 INP_INFO_RUNLOCK(&V_tcbinfo);
976 return (error);
977}
978
979static int
980tcp_usr_ready(struct socket *so, struct mbuf *m, int count)
981{
982 struct inpcb *inp;
983 struct tcpcb *tp;
984 int error;
985
986 inp = sotoinpcb(so);
987 INP_WLOCK(inp);
988 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
989 INP_WUNLOCK(inp);
990 for (int i = 0; i < count; i++)
991 m = m_free(m);
992 return (ECONNRESET);
993 }
994 tp = intotcpcb(inp);
995
996 SOCKBUF_LOCK(&so->so_snd);
997 error = sbready(&so->so_snd, m, count);
998 SOCKBUF_UNLOCK(&so->so_snd);
999 if (error == 0)
1000 error = tcp_output(tp);
1001 INP_WUNLOCK(inp);
1002
1003 return (error);
1004}
1005
1006/*
1007 * Abort the TCP. Drop the connection abruptly.
1008 */
1009static void
1010tcp_usr_abort(struct socket *so)
1011{
1012 struct inpcb *inp;
1013 struct tcpcb *tp = NULL;
1014 TCPDEBUG0;
1015
1016 inp = sotoinpcb(so);
1017 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
1018
1019 INP_INFO_RLOCK(&V_tcbinfo);
1020 INP_WLOCK(inp);
1021 KASSERT(inp->inp_socket != NULL,
1022 ("tcp_usr_abort: inp_socket == NULL"));
1023
1024 /*
1025 * If we still have full TCP state, and we're not dropped, drop.
1026 */
1027 if (!(inp->inp_flags & INP_TIMEWAIT) &&
1028 !(inp->inp_flags & INP_DROPPED)) {
1029 tp = intotcpcb(inp);
1030 TCPDEBUG1();
1031 tcp_drop(tp, ECONNABORTED);
1032 TCPDEBUG2(PRU_ABORT);
1033 TCP_PROBE2(debug__user, tp, PRU_ABORT);
1034 }
1035 if (!(inp->inp_flags & INP_DROPPED)) {
1036 SOCK_LOCK(so);
1037 so->so_state |= SS_PROTOREF;
1038 SOCK_UNLOCK(so);
1039 inp->inp_flags |= INP_SOCKREF;
1040 }
1041 INP_WUNLOCK(inp);
1042 INP_INFO_RUNLOCK(&V_tcbinfo);
1043}
1044
1045/*
1046 * TCP socket is closed. Start friendly disconnect.
1047 */
1048static void
1049tcp_usr_close(struct socket *so)
1050{
1051 struct inpcb *inp;
1052 struct tcpcb *tp = NULL;
1053 TCPDEBUG0;
1054
1055 inp = sotoinpcb(so);
1056 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
1057
1058 INP_INFO_RLOCK(&V_tcbinfo);
1059 INP_WLOCK(inp);
1060 KASSERT(inp->inp_socket != NULL,
1061 ("tcp_usr_close: inp_socket == NULL"));
1062
1063 /*
1064 * If we still have full TCP state, and we're not dropped, initiate
1065 * a disconnect.
1066 */
1067 if (!(inp->inp_flags & INP_TIMEWAIT) &&
1068 !(inp->inp_flags & INP_DROPPED)) {
1069 tp = intotcpcb(inp);
1070 TCPDEBUG1();
1071 tcp_disconnect(tp);
1072 TCPDEBUG2(PRU_CLOSE);
1073 TCP_PROBE2(debug__user, tp, PRU_CLOSE);
1074 }
1075 if (!(inp->inp_flags & INP_DROPPED)) {
1076 SOCK_LOCK(so);
1077 so->so_state |= SS_PROTOREF;
1078 SOCK_UNLOCK(so);
1079 inp->inp_flags |= INP_SOCKREF;
1080 }
1081 INP_WUNLOCK(inp);
1082 INP_INFO_RUNLOCK(&V_tcbinfo);
1083}
1084
1085/*
1086 * Receive out-of-band data.
1087 */
1088static int
1089tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1090{
1091 int error = 0;
1092 struct inpcb *inp;
1093 struct tcpcb *tp = NULL;
1094
1095 TCPDEBUG0;
1096 inp = sotoinpcb(so);
1097 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1098 INP_WLOCK(inp);
1099 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1100 error = ECONNRESET;
1101 goto out;
1102 }
1103 tp = intotcpcb(inp);
1104 TCPDEBUG1();
1105 if ((so->so_oobmark == 0 &&
1106 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1107 so->so_options & SO_OOBINLINE ||
1108 tp->t_oobflags & TCPOOB_HADDATA) {
1109 error = EINVAL;
1110 goto out;
1111 }
1112 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1113 error = EWOULDBLOCK;
1114 goto out;
1115 }
1116 m->m_len = 1;
1117 *mtod(m, caddr_t) = tp->t_iobc;
1118 if ((flags & MSG_PEEK) == 0)
1119 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1120
1121out:
1122 TCPDEBUG2(PRU_RCVOOB);
1123 TCP_PROBE2(debug__user, tp, PRU_RCVOOB);
1124 INP_WUNLOCK(inp);
1125 return (error);
1126}
1127
1128#ifdef INET
1129struct pr_usrreqs tcp_usrreqs = {
1130 .pru_abort = tcp_usr_abort,
1131 .pru_accept = tcp_usr_accept,
1132 .pru_attach = tcp_usr_attach,
1133 .pru_bind = tcp_usr_bind,
1134 .pru_connect = tcp_usr_connect,
1135 .pru_control = in_control,
1136 .pru_detach = tcp_usr_detach,
1137 .pru_disconnect = tcp_usr_disconnect,
1138 .pru_listen = tcp_usr_listen,
1139 .pru_peeraddr = in_getpeeraddr,
1140 .pru_rcvd = tcp_usr_rcvd,
1141 .pru_rcvoob = tcp_usr_rcvoob,
1142 .pru_send = tcp_usr_send,
1143 .pru_ready = tcp_usr_ready,
1144 .pru_shutdown = tcp_usr_shutdown,
1145 .pru_sockaddr = in_getsockaddr,
1146 .pru_sosetlabel = in_pcbsosetlabel,
1147 .pru_close = tcp_usr_close,
1148};
1149#endif /* INET */
1150
1151#ifdef INET6
1152struct pr_usrreqs tcp6_usrreqs = {
1153 .pru_abort = tcp_usr_abort,
1154 .pru_accept = tcp6_usr_accept,
1155 .pru_attach = tcp_usr_attach,
1156 .pru_bind = tcp6_usr_bind,
1157 .pru_connect = tcp6_usr_connect,
1158 .pru_control = in6_control,
1159 .pru_detach = tcp_usr_detach,
1160 .pru_disconnect = tcp_usr_disconnect,
1161 .pru_listen = tcp6_usr_listen,
1162 .pru_peeraddr = in6_mapped_peeraddr,
1163 .pru_rcvd = tcp_usr_rcvd,
1164 .pru_rcvoob = tcp_usr_rcvoob,
1165 .pru_send = tcp_usr_send,
1166 .pru_ready = tcp_usr_ready,
1167 .pru_shutdown = tcp_usr_shutdown,
1168 .pru_sockaddr = in6_mapped_sockaddr,
1169 .pru_sosetlabel = in_pcbsosetlabel,
1170 .pru_close = tcp_usr_close,
1171};
1172#endif /* INET6 */
1173
1174#ifdef INET
1175/*
1176 * Common subroutine to open a TCP connection to remote host specified
1177 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
1178 * port number if needed. Call in_pcbconnect_setup to do the routing and
1179 * to choose a local host address (interface). If there is an existing
1180 * incarnation of the same connection in TIME-WAIT state and if the remote
1181 * host was sending CC options and if the connection duration was < MSL, then
1182 * truncate the previous TIME-WAIT state and proceed.
1183 * Initialize connection parameters and enter SYN-SENT state.
1184 */
1185static int
1186tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1187{
1188 struct inpcb *inp = tp->t_inpcb, *oinp;
1189 struct socket *so = inp->inp_socket;
1190 struct in_addr laddr;
1191 u_short lport;
1192 int error;
1193
1194 INP_WLOCK_ASSERT(inp);
1195 INP_HASH_WLOCK(&V_tcbinfo);
1196
1197 if (inp->inp_lport == 0) {
1198 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1199 if (error)
1200 goto out;
1201 }
1202
1203 /*
1204 * Cannot simply call in_pcbconnect, because there might be an
1205 * earlier incarnation of this same connection still in
1206 * TIME_WAIT state, creating an ADDRINUSE error.
1207 */
1208 laddr = inp->inp_laddr;
1209 lport = inp->inp_lport;
1210 error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1211 &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1212 if (error && oinp == NULL)
1213 goto out;
1214 if (oinp) {
1215 error = EADDRINUSE;
1216 goto out;
1217 }
1218 inp->inp_laddr = laddr;
1219 in_pcbrehash(inp);
1220 INP_HASH_WUNLOCK(&V_tcbinfo);
1221
1222 /*
1223 * Compute window scaling to request:
1224 * Scale to fit into sweet spot. See tcp_syncache.c.
1225 * XXX: This should move to tcp_output().
1226 */
1227 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1228 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1229 tp->request_r_scale++;
1230
1231 soisconnecting(so);
1232 TCPSTAT_INC(tcps_connattempt);
1233 tcp_state_change(tp, TCPS_SYN_SENT);
1234 tp->iss = tcp_new_isn(tp);
1235 tcp_sendseqinit(tp);
1236
1237 return 0;
1238
1239out:
1240 INP_HASH_WUNLOCK(&V_tcbinfo);
1241 return (error);
1242}
1243#endif /* INET */
1244
1245#ifdef INET6
1246static int
1247tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1248{
1249 struct inpcb *inp = tp->t_inpcb;
1250 int error;
1251
1252 INP_WLOCK_ASSERT(inp);
1253 INP_HASH_WLOCK(&V_tcbinfo);
1254
1255 if (inp->inp_lport == 0) {
1256 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1257 if (error)
1258 goto out;
1259 }
1260 error = in6_pcbconnect(inp, nam, td->td_ucred);
1261 if (error != 0)
1262 goto out;
1263 INP_HASH_WUNLOCK(&V_tcbinfo);
1264
1265 /* Compute window scaling to request. */
1266 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1267 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1268 tp->request_r_scale++;
1269
1270 soisconnecting(inp->inp_socket);
1271 TCPSTAT_INC(tcps_connattempt);
1272 tcp_state_change(tp, TCPS_SYN_SENT);
1273 tp->iss = tcp_new_isn(tp);
1274 tcp_sendseqinit(tp);
1275
1276 return 0;
1277
1278out:
1279 INP_HASH_WUNLOCK(&V_tcbinfo);
1280 return error;
1281}
1282#endif /* INET6 */
1283
1284/*
1285 * Export TCP internal state information via a struct tcp_info, based on the
1286 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently
1287 * (TCP state machine, etc). We export all information using FreeBSD-native
1288 * constants -- for example, the numeric values for tcpi_state will differ
1289 * from Linux.
1290 */
1291static void
1292tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1293{
1294
1295 INP_WLOCK_ASSERT(tp->t_inpcb);
1296 bzero(ti, sizeof(*ti));
1297
1298 ti->tcpi_state = tp->t_state;
1299 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1300 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1301 if (tp->t_flags & TF_SACK_PERMIT)
1302 ti->tcpi_options |= TCPI_OPT_SACK;
1303 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1304 ti->tcpi_options |= TCPI_OPT_WSCALE;
1305 ti->tcpi_snd_wscale = tp->snd_scale;
1306 ti->tcpi_rcv_wscale = tp->rcv_scale;
1307 }
1308
1309 ti->tcpi_rto = tp->t_rxtcur * tick;
1310 ti->tcpi_last_data_recv = (long)(ticks - (int)tp->t_rcvtime) * tick;
1311 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1312 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1313
1314 ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1315 ti->tcpi_snd_cwnd = tp->snd_cwnd;
1316
1317 /*
1318 * FreeBSD-specific extension fields for tcp_info.
1319 */
1320 ti->tcpi_rcv_space = tp->rcv_wnd;
1321 ti->tcpi_rcv_nxt = tp->rcv_nxt;
1322 ti->tcpi_snd_wnd = tp->snd_wnd;
1323 ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */
1324 ti->tcpi_snd_nxt = tp->snd_nxt;
1325 ti->tcpi_snd_mss = tp->t_maxseg;
1326 ti->tcpi_rcv_mss = tp->t_maxseg;
1327 if (tp->t_flags & TF_TOE)
1328 ti->tcpi_options |= TCPI_OPT_TOE;
1329 ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1330 ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1331 ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1332}
1333
1334/*
1335 * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1336 * socket option arguments. When it re-acquires the lock after the copy, it
1337 * has to revalidate that the connection is still valid for the socket
1338 * option.
1339 */
1340#define INP_WLOCK_RECHECK(inp) do { \
1341 INP_WLOCK(inp); \
1342 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
1343 INP_WUNLOCK(inp); \
1344 return (ECONNRESET); \
1345 } \
1346 tp = intotcpcb(inp); \
1347} while(0)
1348
1349int
1350tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1351{
1352 int error, opt, optval;
1353 u_int ui;
1354 struct inpcb *inp;
1355 struct tcpcb *tp;
1356 struct tcp_info ti;
1357 char buf[TCP_CA_NAME_MAX];
1358 struct cc_algo *algo;
1359
1360 error = 0;
1361 inp = sotoinpcb(so);
1362 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1363 INP_WLOCK(inp);
1364 if (sopt->sopt_level != IPPROTO_TCP) {
1365#ifdef INET6
1366 if (inp->inp_vflag & INP_IPV6PROTO) {
1367 INP_WUNLOCK(inp);
1368 error = ip6_ctloutput(so, sopt);
1369 }
1370#endif /* INET6 */
1371#if defined(INET6) && defined(INET)
1372 else
1373#endif
1374#ifdef INET
1375 {
1376 INP_WUNLOCK(inp);
1377 error = ip_ctloutput(so, sopt);
1378 }
1379#endif
1380 return (error);
1381 }
1382 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1383 INP_WUNLOCK(inp);
1384 return (ECONNRESET);
1385 }
1386
1387 switch (sopt->sopt_dir) {
1388 case SOPT_SET:
1389 switch (sopt->sopt_name) {
1390#ifdef TCP_SIGNATURE
1391 case TCP_MD5SIG:
1392 INP_WUNLOCK(inp);
1393 error = sooptcopyin(sopt, &optval, sizeof optval,
1394 sizeof optval);
1395 if (error)
1396 return (error);
1397
1398 INP_WLOCK_RECHECK(inp);
1399 if (optval > 0)
1400 tp->t_flags |= TF_SIGNATURE;
1401 else
1402 tp->t_flags &= ~TF_SIGNATURE;
1403 goto unlock_and_done;
1404#endif /* TCP_SIGNATURE */
1405
1406 case TCP_NODELAY:
1407 case TCP_NOOPT:
1408 INP_WUNLOCK(inp);
1409 error = sooptcopyin(sopt, &optval, sizeof optval,
1410 sizeof optval);
1411 if (error)
1412 return (error);
1413
1414 INP_WLOCK_RECHECK(inp);
1415 switch (sopt->sopt_name) {
1416 case TCP_NODELAY:
1417 opt = TF_NODELAY;
1418 break;
1419 case TCP_NOOPT:
1420 opt = TF_NOOPT;
1421 break;
1422 default:
1423 opt = 0; /* dead code to fool gcc */
1424 break;
1425 }
1426
1427 if (optval)
1428 tp->t_flags |= opt;
1429 else
1430 tp->t_flags &= ~opt;
1431unlock_and_done:
1432#ifdef TCP_OFFLOAD
1433 if (tp->t_flags & TF_TOE) {
1434 tcp_offload_ctloutput(tp, sopt->sopt_dir,
1435 sopt->sopt_name);
1436 }
1437#endif
1438 INP_WUNLOCK(inp);
1439 break;
1440
1441 case TCP_NOPUSH:
1442 INP_WUNLOCK(inp);
1443 error = sooptcopyin(sopt, &optval, sizeof optval,
1444 sizeof optval);
1445 if (error)
1446 return (error);
1447
1448 INP_WLOCK_RECHECK(inp);
1449 if (optval)
1450 tp->t_flags |= TF_NOPUSH;
1451 else if (tp->t_flags & TF_NOPUSH) {
1452 tp->t_flags &= ~TF_NOPUSH;
1453 if (TCPS_HAVEESTABLISHED(tp->t_state))
1454 error = tcp_output(tp);
1455 }
1456 goto unlock_and_done;
1457
1458 case TCP_MAXSEG:
1459 INP_WUNLOCK(inp);
1460 error = sooptcopyin(sopt, &optval, sizeof optval,
1461 sizeof optval);
1462 if (error)
1463 return (error);
1464
1465 INP_WLOCK_RECHECK(inp);
1466 if (optval > 0 && optval <= tp->t_maxseg &&
1467 optval + 40 >= V_tcp_minmss)
1468 tp->t_maxseg = optval;
1469 else
1470 error = EINVAL;
1471 goto unlock_and_done;
1472
1473 case TCP_INFO:
1474 INP_WUNLOCK(inp);
1475 error = EINVAL;
1476 break;
1477
1478 case TCP_CONGESTION:
1479 INP_WUNLOCK(inp);
1480 bzero(buf, sizeof(buf));
1481 error = sooptcopyin(sopt, &buf, sizeof(buf), 1);
1482 if (error)
1483 break;
1484 INP_WLOCK_RECHECK(inp);
1485 /*
1486 * Return EINVAL if we can't find the requested cc algo.
1487 */
1488 error = EINVAL;
1489 CC_LIST_RLOCK();
1490 STAILQ_FOREACH(algo, &cc_list, entries) {
1491 if (strncmp(buf, algo->name, TCP_CA_NAME_MAX)
1492 == 0) {
1493 /* We've found the requested algo. */
1494 error = 0;
1495 /*
1496 * We hold a write lock over the tcb
1497 * so it's safe to do these things
1498 * without ordering concerns.
1499 */
1500 if (CC_ALGO(tp)->cb_destroy != NULL)
1501 CC_ALGO(tp)->cb_destroy(tp->ccv);
1502 CC_ALGO(tp) = algo;
1503 /*
1504 * If something goes pear shaped
1505 * initialising the new algo,
1506 * fall back to newreno (which
1507 * does not require initialisation).
1508 */
1509 if (algo->cb_init != NULL)
1510 if (algo->cb_init(tp->ccv) > 0) {
1511 CC_ALGO(tp) = &newreno_cc_algo;
1512 /*
1513 * The only reason init
1514 * should fail is
1515 * because of malloc.
1516 */
1517 error = ENOMEM;
1518 }
1519 break; /* Break the STAILQ_FOREACH. */
1520 }
1521 }
1522 CC_LIST_RUNLOCK();
1523 goto unlock_and_done;
1524
1525 case TCP_KEEPIDLE:
1526 case TCP_KEEPINTVL:
1527 case TCP_KEEPINIT:
1528 INP_WUNLOCK(inp);
1529 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1530 if (error)
1531 return (error);
1532
1533 if (ui > (UINT_MAX / hz)) {
1534 error = EINVAL;
1535 break;
1536 }
1537 ui *= hz;
1538
1539 INP_WLOCK_RECHECK(inp);
1540 switch (sopt->sopt_name) {
1541 case TCP_KEEPIDLE:
1542 tp->t_keepidle = ui;
1543 /*
1544 * XXX: better check current remaining
1545 * timeout and "merge" it with new value.
1546 */
1547 if ((tp->t_state > TCPS_LISTEN) &&
1548 (tp->t_state <= TCPS_CLOSING))
1549 tcp_timer_activate(tp, TT_KEEP,
1550 TP_KEEPIDLE(tp));
1551 break;
1552 case TCP_KEEPINTVL:
1553 tp->t_keepintvl = ui;
1554 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1555 (TP_MAXIDLE(tp) > 0))
1556 tcp_timer_activate(tp, TT_2MSL,
1557 TP_MAXIDLE(tp));
1558 break;
1559 case TCP_KEEPINIT:
1560 tp->t_keepinit = ui;
1561 if (tp->t_state == TCPS_SYN_RECEIVED ||
1562 tp->t_state == TCPS_SYN_SENT)
1563 tcp_timer_activate(tp, TT_KEEP,
1564 TP_KEEPINIT(tp));
1565 break;
1566 }
1567 goto unlock_and_done;
1568
1569 case TCP_KEEPCNT:
1570 INP_WUNLOCK(inp);
1571 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1572 if (error)
1573 return (error);
1574
1575 INP_WLOCK_RECHECK(inp);
1576 tp->t_keepcnt = ui;
1577 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1578 (TP_MAXIDLE(tp) > 0))
1579 tcp_timer_activate(tp, TT_2MSL,
1580 TP_MAXIDLE(tp));
1581 goto unlock_and_done;
1582
1583#ifdef TCPPCAP
1584 case TCP_PCAP_OUT:
1585 case TCP_PCAP_IN:
1586 INP_WUNLOCK(inp);
1587 error = sooptcopyin(sopt, &optval, sizeof optval,
1588 sizeof optval);
1589 if (error)
1590 return (error);
1591
1592 INP_WLOCK_RECHECK(inp);
1593 if (optval >= 0)
1594 tcp_pcap_set_sock_max(TCP_PCAP_OUT ?
1595 &(tp->t_outpkts) : &(tp->t_inpkts),
1596 optval);
1597 else
1598 error = EINVAL;
1599 goto unlock_and_done;
1600#endif
1601
1602 default:
1603 INP_WUNLOCK(inp);
1604 error = ENOPROTOOPT;
1605 break;
1606 }
1607 break;
1608
1609 case SOPT_GET:
1610 tp = intotcpcb(inp);
1611 switch (sopt->sopt_name) {
1612#ifdef TCP_SIGNATURE
1613 case TCP_MD5SIG:
1614 optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
1615 INP_WUNLOCK(inp);
1616 error = sooptcopyout(sopt, &optval, sizeof optval);
1617 break;
1618#endif
1619
1620 case TCP_NODELAY:
1621 optval = tp->t_flags & TF_NODELAY;
1622 INP_WUNLOCK(inp);
1623 error = sooptcopyout(sopt, &optval, sizeof optval);
1624 break;
1625 case TCP_MAXSEG:
1626 optval = tp->t_maxseg;
1627 INP_WUNLOCK(inp);
1628 error = sooptcopyout(sopt, &optval, sizeof optval);
1629 break;
1630 case TCP_NOOPT:
1631 optval = tp->t_flags & TF_NOOPT;
1632 INP_WUNLOCK(inp);
1633 error = sooptcopyout(sopt, &optval, sizeof optval);
1634 break;
1635 case TCP_NOPUSH:
1636 optval = tp->t_flags & TF_NOPUSH;
1637 INP_WUNLOCK(inp);
1638 error = sooptcopyout(sopt, &optval, sizeof optval);
1639 break;
1640 case TCP_INFO:
1641 tcp_fill_info(tp, &ti);
1642 INP_WUNLOCK(inp);
1643 error = sooptcopyout(sopt, &ti, sizeof ti);
1644 break;
1645 case TCP_CONGESTION:
1646 bzero(buf, sizeof(buf));
1647 strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
1648 INP_WUNLOCK(inp);
1649 error = sooptcopyout(sopt, buf, TCP_CA_NAME_MAX);
1650 break;
1651 case TCP_KEEPIDLE:
1652 case TCP_KEEPINTVL:
1653 case TCP_KEEPINIT:
1654 case TCP_KEEPCNT:
1655 switch (sopt->sopt_name) {
1656 case TCP_KEEPIDLE:
1657 ui = tp->t_keepidle / hz;
1658 break;
1659 case TCP_KEEPINTVL:
1660 ui = tp->t_keepintvl / hz;
1661 break;
1662 case TCP_KEEPINIT:
1663 ui = tp->t_keepinit / hz;
1664 break;
1665 case TCP_KEEPCNT:
1666 ui = tp->t_keepcnt;
1667 break;
1668 }
1669 INP_WUNLOCK(inp);
1670 error = sooptcopyout(sopt, &ui, sizeof(ui));
1671 break;
1672#ifdef TCPPCAP
1673 case TCP_PCAP_OUT:
1674 case TCP_PCAP_IN:
1675 optval = tcp_pcap_get_sock_max(TCP_PCAP_OUT ?
1676 &(tp->t_outpkts) : &(tp->t_inpkts));
1677 INP_WUNLOCK(inp);
1678 error = sooptcopyout(sopt, &optval, sizeof optval);
1679 break;
1680#endif
1681 default:
1682 INP_WUNLOCK(inp);
1683 error = ENOPROTOOPT;
1684 break;
1685 }
1686 break;
1687 }
1688 return (error);
1689}
1690#undef INP_WLOCK_RECHECK
1691
1692/*
1693 * Attach TCP protocol to socket, allocating
1694 * internet protocol control block, tcp control block,
1695 * bufer space, and entering LISTEN state if to accept connections.
1696 */
1697static int
1698tcp_attach(struct socket *so)
1699{
1700 struct tcpcb *tp;
1701 struct inpcb *inp;
1702 int error;
1703
1704 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1705 error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
1706 if (error)
1707 return (error);
1708 }
1709 so->so_rcv.sb_flags |= SB_AUTOSIZE;
1710 so->so_snd.sb_flags |= SB_AUTOSIZE;
1711 INP_INFO_RLOCK(&V_tcbinfo);
1712 error = in_pcballoc(so, &V_tcbinfo);
1713 if (error) {
1714 INP_INFO_RUNLOCK(&V_tcbinfo);
1715 return (error);
1716 }
1717 inp = sotoinpcb(so);
1718#ifdef INET6
1719 if (inp->inp_vflag & INP_IPV6PROTO) {
1720 inp->inp_vflag |= INP_IPV6;
1721 inp->in6p_hops = -1; /* use kernel default */
1722 }
1723 else
1724#endif
1725 inp->inp_vflag |= INP_IPV4;
1726 tp = tcp_newtcpcb(inp);
1727 if (tp == NULL) {
1728 in_pcbdetach(inp);
1729 in_pcbfree(inp);
1730 INP_INFO_RUNLOCK(&V_tcbinfo);
1731 return (ENOBUFS);
1732 }
1733 tp->t_state = TCPS_CLOSED;
1734 INP_WUNLOCK(inp);
1735 INP_INFO_RUNLOCK(&V_tcbinfo);
1736 return (0);
1737}
1738
1739/*
1740 * Initiate (or continue) disconnect.
1741 * If embryonic state, just send reset (once).
1742 * If in ``let data drain'' option and linger null, just drop.
1743 * Otherwise (hard), mark socket disconnecting and drop
1744 * current input data; switch states based on user close, and
1745 * send segment to peer (with FIN).
1746 */
1747static void
1748tcp_disconnect(struct tcpcb *tp)
1749{
1750 struct inpcb *inp = tp->t_inpcb;
1751 struct socket *so = inp->inp_socket;
1752
1753 INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
1754 INP_WLOCK_ASSERT(inp);
1755
1756 /*
1757 * Neither tcp_close() nor tcp_drop() should return NULL, as the
1758 * socket is still open.
1759 */
1760 if (tp->t_state < TCPS_ESTABLISHED) {
1761 tp = tcp_close(tp);
1762 KASSERT(tp != NULL,
1763 ("tcp_disconnect: tcp_close() returned NULL"));
1764 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
1765 tp = tcp_drop(tp, 0);
1766 KASSERT(tp != NULL,
1767 ("tcp_disconnect: tcp_drop() returned NULL"));
1768 } else {
1769 soisdisconnecting(so);
1770 sbflush(&so->so_rcv);
1771 tcp_usrclosed(tp);
1772 if (!(inp->inp_flags & INP_DROPPED))
1773 tcp_output(tp);
1774 }
1775}
1776
1777/*
1778 * User issued close, and wish to trail through shutdown states:
1779 * if never received SYN, just forget it. If got a SYN from peer,
1780 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1781 * If already got a FIN from peer, then almost done; go to LAST_ACK
1782 * state. In all other cases, have already sent FIN to peer (e.g.
1783 * after PRU_SHUTDOWN), and just have to play tedious game waiting
1784 * for peer to send FIN or not respond to keep-alives, etc.
1785 * We can let the user exit from the close as soon as the FIN is acked.
1786 */
1787static void
1788tcp_usrclosed(struct tcpcb *tp)
1789{
1790
1791 INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
1792 INP_WLOCK_ASSERT(tp->t_inpcb);
1793
1794 switch (tp->t_state) {
1795 case TCPS_LISTEN:
1796#ifdef TCP_OFFLOAD
1797 tcp_offload_listen_stop(tp);
1798#endif
1799 tcp_state_change(tp, TCPS_CLOSED);
1800 /* FALLTHROUGH */
1801 case TCPS_CLOSED:
1802 tp = tcp_close(tp);
1803 /*
1804 * tcp_close() should never return NULL here as the socket is
1805 * still open.
1806 */
1807 KASSERT(tp != NULL,
1808 ("tcp_usrclosed: tcp_close() returned NULL"));
1809 break;
1810
1811 case TCPS_SYN_SENT:
1812 case TCPS_SYN_RECEIVED:
1813 tp->t_flags |= TF_NEEDFIN;
1814 break;
1815
1816 case TCPS_ESTABLISHED:
1817 tcp_state_change(tp, TCPS_FIN_WAIT_1);
1818 break;
1819
1820 case TCPS_CLOSE_WAIT:
1821 tcp_state_change(tp, TCPS_LAST_ACK);
1822 break;
1823 }
1824 if (tp->t_state >= TCPS_FIN_WAIT_2) {
1825 soisdisconnected(tp->t_inpcb->inp_socket);
1826 /* Prevent the connection hanging in FIN_WAIT_2 forever. */
1827 if (tp->t_state == TCPS_FIN_WAIT_2) {
1828 int timeout;
1829
1830 timeout = (tcp_fast_finwait2_recycle) ?
1831 tcp_finwait2_timeout : TP_MAXIDLE(tp);
1832 tcp_timer_activate(tp, TT_2MSL, timeout);
1833 }
1834 }
1835}
1836
1837#ifdef DDB
1838static void
1839db_print_indent(int indent)
1840{
1841 int i;
1842
1843 for (i = 0; i < indent; i++)
1844 db_printf(" ");
1845}
1846
1847static void
1848db_print_tstate(int t_state)
1849{
1850
1851 switch (t_state) {
1852 case TCPS_CLOSED:
1853 db_printf("TCPS_CLOSED");
1854 return;
1855
1856 case TCPS_LISTEN:
1857 db_printf("TCPS_LISTEN");
1858 return;
1859
1860 case TCPS_SYN_SENT:
1861 db_printf("TCPS_SYN_SENT");
1862 return;
1863
1864 case TCPS_SYN_RECEIVED:
1865 db_printf("TCPS_SYN_RECEIVED");
1866 return;
1867
1868 case TCPS_ESTABLISHED:
1869 db_printf("TCPS_ESTABLISHED");
1870 return;
1871
1872 case TCPS_CLOSE_WAIT:
1873 db_printf("TCPS_CLOSE_WAIT");
1874 return;
1875
1876 case TCPS_FIN_WAIT_1:
1877 db_printf("TCPS_FIN_WAIT_1");
1878 return;
1879
1880 case TCPS_CLOSING:
1881 db_printf("TCPS_CLOSING");
1882 return;
1883
1884 case TCPS_LAST_ACK:
1885 db_printf("TCPS_LAST_ACK");
1886 return;
1887
1888 case TCPS_FIN_WAIT_2:
1889 db_printf("TCPS_FIN_WAIT_2");
1890 return;
1891
1892 case TCPS_TIME_WAIT:
1893 db_printf("TCPS_TIME_WAIT");
1894 return;
1895
1896 default:
1897 db_printf("unknown");
1898 return;
1899 }
1900}
1901
1902static void
1903db_print_tflags(u_int t_flags)
1904{
1905 int comma;
1906
1907 comma = 0;
1908 if (t_flags & TF_ACKNOW) {
1909 db_printf("%sTF_ACKNOW", comma ? ", " : "");
1910 comma = 1;
1911 }
1912 if (t_flags & TF_DELACK) {
1913 db_printf("%sTF_DELACK", comma ? ", " : "");
1914 comma = 1;
1915 }
1916 if (t_flags & TF_NODELAY) {
1917 db_printf("%sTF_NODELAY", comma ? ", " : "");
1918 comma = 1;
1919 }
1920 if (t_flags & TF_NOOPT) {
1921 db_printf("%sTF_NOOPT", comma ? ", " : "");
1922 comma = 1;
1923 }
1924 if (t_flags & TF_SENTFIN) {
1925 db_printf("%sTF_SENTFIN", comma ? ", " : "");
1926 comma = 1;
1927 }
1928 if (t_flags & TF_REQ_SCALE) {
1929 db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
1930 comma = 1;
1931 }
1932 if (t_flags & TF_RCVD_SCALE) {
1933 db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
1934 comma = 1;
1935 }
1936 if (t_flags & TF_REQ_TSTMP) {
1937 db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
1938 comma = 1;
1939 }
1940 if (t_flags & TF_RCVD_TSTMP) {
1941 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
1942 comma = 1;
1943 }
1944 if (t_flags & TF_SACK_PERMIT) {
1945 db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
1946 comma = 1;
1947 }
1948 if (t_flags & TF_NEEDSYN) {
1949 db_printf("%sTF_NEEDSYN", comma ? ", " : "");
1950 comma = 1;
1951 }
1952 if (t_flags & TF_NEEDFIN) {
1953 db_printf("%sTF_NEEDFIN", comma ? ", " : "");
1954 comma = 1;
1955 }
1956 if (t_flags & TF_NOPUSH) {
1957 db_printf("%sTF_NOPUSH", comma ? ", " : "");
1958 comma = 1;
1959 }
1960 if (t_flags & TF_MORETOCOME) {
1961 db_printf("%sTF_MORETOCOME", comma ? ", " : "");
1962 comma = 1;
1963 }
1964 if (t_flags & TF_LQ_OVERFLOW) {
1965 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
1966 comma = 1;
1967 }
1968 if (t_flags & TF_LASTIDLE) {
1969 db_printf("%sTF_LASTIDLE", comma ? ", " : "");
1970 comma = 1;
1971 }
1972 if (t_flags & TF_RXWIN0SENT) {
1973 db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
1974 comma = 1;
1975 }
1976 if (t_flags & TF_FASTRECOVERY) {
1977 db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
1978 comma = 1;
1979 }
1980 if (t_flags & TF_CONGRECOVERY) {
1981 db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
1982 comma = 1;
1983 }
1984 if (t_flags & TF_WASFRECOVERY) {
1985 db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
1986 comma = 1;
1987 }
1988 if (t_flags & TF_SIGNATURE) {
1989 db_printf("%sTF_SIGNATURE", comma ? ", " : "");
1990 comma = 1;
1991 }
1992 if (t_flags & TF_FORCEDATA) {
1993 db_printf("%sTF_FORCEDATA", comma ? ", " : "");
1994 comma = 1;
1995 }
1996 if (t_flags & TF_TSO) {
1997 db_printf("%sTF_TSO", comma ? ", " : "");
1998 comma = 1;
1999 }
2000 if (t_flags & TF_ECN_PERMIT) {
2001 db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
2002 comma = 1;
2003 }
2004}
2005
2006static void
2007db_print_toobflags(char t_oobflags)
2008{
2009 int comma;
2010
2011 comma = 0;
2012 if (t_oobflags & TCPOOB_HAVEDATA) {
2013 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
2014 comma = 1;
2015 }
2016 if (t_oobflags & TCPOOB_HADDATA) {
2017 db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
2018 comma = 1;
2019 }
2020}
2021
2022static void
2023db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
2024{
2025
2026 db_print_indent(indent);
2027 db_printf("%s at %p\n", name, tp);
2028
2029 indent += 2;
2030
2031 db_print_indent(indent);
2032 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n",
2033 LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
2034
2035 db_print_indent(indent);
2036 db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n",
2037 &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
2038
2039 db_print_indent(indent);
2040 db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", &tp->t_timers->tt_2msl,
2041 &tp->t_timers->tt_delack, tp->t_inpcb);
2042
2043 db_print_indent(indent);
2044 db_printf("t_state: %d (", tp->t_state);
2045 db_print_tstate(tp->t_state);
2046 db_printf(")\n");
2047
2048 db_print_indent(indent);
2049 db_printf("t_flags: 0x%x (", tp->t_flags);
2050 db_print_tflags(tp->t_flags);
2051 db_printf(")\n");
2052
2053 db_print_indent(indent);
2054 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n",
2055 tp->snd_una, tp->snd_max, tp->snd_nxt);
2056
2057 db_print_indent(indent);
2058 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n",
2059 tp->snd_up, tp->snd_wl1, tp->snd_wl2);
2060
2061 db_print_indent(indent);
2062 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n",
2063 tp->iss, tp->irs, tp->rcv_nxt);
2064
2065 db_print_indent(indent);
2066 db_printf("rcv_adv: 0x%08x rcv_wnd: %lu rcv_up: 0x%08x\n",
2067 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
2068
2069 db_print_indent(indent);
2070 db_printf("snd_wnd: %lu snd_cwnd: %lu\n",
2071 tp->snd_wnd, tp->snd_cwnd);
2072
2073 db_print_indent(indent);
2074 db_printf("snd_ssthresh: %lu snd_recover: "
2075 "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
2076
2077 db_print_indent(indent);
2078 db_printf("t_maxopd: %u t_rcvtime: %u t_startime: %u\n",
2079 tp->t_maxopd, tp->t_rcvtime, tp->t_starttime);
2080
2081 db_print_indent(indent);
2082 db_printf("t_rttime: %u t_rtsq: 0x%08x\n",
2083 tp->t_rtttime, tp->t_rtseq);
2084
2085 db_print_indent(indent);
2086 db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n",
2087 tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
2088
2089 db_print_indent(indent);
2090 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u "
2091 "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
2092 tp->t_rttbest);
2093
2094 db_print_indent(indent);
2095 db_printf("t_rttupdated: %lu max_sndwnd: %lu t_softerror: %d\n",
2096 tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
2097
2098 db_print_indent(indent);
2099 db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
2100 db_print_toobflags(tp->t_oobflags);
2101 db_printf(") t_iobc: 0x%02x\n", tp->t_iobc);
2102
2103 db_print_indent(indent);
2104 db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n",
2105 tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
2106
2107 db_print_indent(indent);
2108 db_printf("ts_recent: %u ts_recent_age: %u\n",
2109 tp->ts_recent, tp->ts_recent_age);
2110
2111 db_print_indent(indent);
2112 db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: "
2113 "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
2114
2115 db_print_indent(indent);
2116 db_printf("snd_ssthresh_prev: %lu snd_recover_prev: 0x%08x "
2117 "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
2118 tp->snd_recover_prev, tp->t_badrxtwin);
2119
2120 db_print_indent(indent);
2121 db_printf("snd_numholes: %d snd_holes first: %p\n",
2122 tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
2123
2124 db_print_indent(indent);
2125 db_printf("snd_fack: 0x%08x rcv_numsacks: %d sack_newdata: "
2126 "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
2127
2128 /* Skip sackblks, sackhint. */
2129
2130 db_print_indent(indent);
2131 db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n",
2132 tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
2133}
2134
2135DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
2136{
2137 struct tcpcb *tp;
2138
2139 if (!have_addr) {
2140 db_printf("usage: show tcpcb <addr>\n");
2141 return;
2142 }
2143 tp = (struct tcpcb *)addr;
2144
2145 db_print_tcpcb(tp, "tcpcb", 0);
2146}
2147#endif