1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California.
6 * Copyright (c) 2006-2007 Robert N. M. Watson
7 * Copyright (c) 2010-2011 Juniper Networks, Inc.
8 * All rights reserved.
9 *
10 * Portions of this software were developed by Robert N. M. Watson under
11 * contract to Juniper Networks, Inc.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94
38 */
39
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD: stable/12/sys/netinet/tcp_usrreq.c 372961 2023-03-03 21:30:21Z git2svn $");
42
43 #include "opt_ddb.h"
44 #include "opt_inet.h"
45 #include "opt_inet6.h"
46 #include "opt_ipsec.h"
47 #include "opt_tcpdebug.h"
48
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/limits.h>
52 #include <sys/malloc.h>
53 #include <sys/refcount.h>
54 #include <sys/kernel.h>
55 #include <sys/sysctl.h>
56 #include <sys/mbuf.h>
57 #ifdef INET6
58 #include <sys/domain.h>
59 #endif /* INET6 */
60 #include <sys/socket.h>
61 #include <sys/socketvar.h>
62 #include <sys/protosw.h>
63 #include <sys/proc.h>
64 #include <sys/jail.h>
65 #include <sys/syslog.h>
66
67 #ifdef DDB
68 #include <ddb/ddb.h>
69 #endif
70
71 #include <net/if.h>
72 #include <net/if_var.h>
73 #include <net/route.h>
74 #include <net/vnet.h>
75
76 #include <netinet/in.h>
77 #include <netinet/in_kdtrace.h>
78 #include <netinet/in_pcb.h>
79 #include <netinet/in_systm.h>
80 #include <netinet/in_var.h>
81 #include <netinet/ip_var.h>
82 #ifdef INET6
83 #include <netinet/ip6.h>
84 #include <netinet6/in6_pcb.h>
85 #include <netinet6/ip6_var.h>
86 #include <netinet6/scope6_var.h>
87 #endif
88 #include <netinet/tcp.h>
89 #include <netinet/tcp_fsm.h>
90 #include <netinet/tcp_seq.h>
91 #include <netinet/tcp_timer.h>
92 #include <netinet/tcp_var.h>
93 #include <netinet/tcp_log_buf.h>
94 #include <netinet/tcpip.h>
95 #include <netinet/cc/cc.h>
96 #include <netinet/tcp_fastopen.h>
97 #include <netinet/tcp_hpts.h>
98 #ifdef TCPPCAP
99 #include <netinet/tcp_pcap.h>
100 #endif
101 #ifdef TCPDEBUG
102 #include <netinet/tcp_debug.h>
103 #endif
104 #ifdef TCP_OFFLOAD
105 #include <netinet/tcp_offload.h>
106 #endif
107 #include <netipsec/ipsec_support.h>
108
109 /*
110 * TCP protocol interface to socket abstraction.
111 */
112 static int tcp_attach(struct socket *);
113 #ifdef INET
114 static int tcp_connect(struct tcpcb *, struct sockaddr *,
115 struct thread *td);
116 #endif /* INET */
117 #ifdef INET6
118 static int tcp6_connect(struct tcpcb *, struct sockaddr *,
119 struct thread *td);
120 #endif /* INET6 */
121 static void tcp_disconnect(struct tcpcb *);
122 static void tcp_usrclosed(struct tcpcb *);
123 static void tcp_fill_info(struct tcpcb *, struct tcp_info *);
124
125 #ifdef TCPDEBUG
126 #define TCPDEBUG0 int ostate = 0
127 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0
128 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
129 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
130 #else
131 #define TCPDEBUG0
132 #define TCPDEBUG1()
133 #define TCPDEBUG2(req)
134 #endif
135
136 /*
137 * tcp_require_unique port requires a globally-unique source port for each
138 * outgoing connection. The default is to require the 4-tuple to be unique.
139 */
140 VNET_DEFINE(int, tcp_require_unique_port) = 0;
141 SYSCTL_INT(_net_inet_tcp, OID_AUTO, require_unique_port,
142 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_require_unique_port), 0,
143 "Require globally-unique ephemeral port for outgoing connections");
144 #define V_tcp_require_unique_port VNET(tcp_require_unique_port)
145
146 /*
147 * TCP attaches to socket via pru_attach(), reserving space,
148 * and an internet control block.
149 */
150 static int
tcp_usr_attach(struct socket * so,int proto,struct thread * td)151 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
152 {
153 struct inpcb *inp;
154 struct tcpcb *tp = NULL;
155 int error;
156 TCPDEBUG0;
157
158 inp = sotoinpcb(so);
159 KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
160 TCPDEBUG1();
161
162 error = tcp_attach(so);
163 if (error)
164 goto out;
165
166 inp = sotoinpcb(so);
167 tp = intotcpcb(inp);
168 out:
169 TCPDEBUG2(PRU_ATTACH);
170 TCP_PROBE2(debug__user, tp, PRU_ATTACH);
171 return error;
172 }
173
174 /*
175 * tcp_detach is called when the socket layer loses its final reference
176 * to the socket, be it a file descriptor reference, a reference from TCP,
177 * etc. At this point, there is only one case in which we will keep around
178 * inpcb state: time wait.
179 *
180 * This function can probably be re-absorbed back into tcp_usr_detach() now
181 * that there is a single detach path.
182 */
183 static void
tcp_detach(struct socket * so,struct inpcb * inp)184 tcp_detach(struct socket *so, struct inpcb *inp)
185 {
186 struct tcpcb *tp;
187
188 INP_INFO_LOCK_ASSERT(&V_tcbinfo);
189 INP_WLOCK_ASSERT(inp);
190
191 KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
192 KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
193
194 tp = intotcpcb(inp);
195
196 if (inp->inp_flags & INP_TIMEWAIT) {
197 /*
198 * There are two cases to handle: one in which the time wait
199 * state is being discarded (INP_DROPPED), and one in which
200 * this connection will remain in timewait. In the former,
201 * it is time to discard all state (except tcptw, which has
202 * already been discarded by the timewait close code, which
203 * should be further up the call stack somewhere). In the
204 * latter case, we detach from the socket, but leave the pcb
205 * present until timewait ends.
206 *
207 * XXXRW: Would it be cleaner to free the tcptw here?
208 *
209 * Astute question indeed, from twtcp perspective there are
210 * four cases to consider:
211 *
212 * #1 tcp_detach is called at tcptw creation time by
213 * tcp_twstart, then do not discard the newly created tcptw
214 * and leave inpcb present until timewait ends
215 * #2 tcp_detach is called at tcptw creation time by
216 * tcp_twstart, but connection is local and tw will be
217 * discarded immediately
218 * #3 tcp_detach is called at timewait end (or reuse) by
219 * tcp_twclose, then the tcptw has already been discarded
220 * (or reused) and inpcb is freed here
221 * #4 tcp_detach is called() after timewait ends (or reuse)
222 * (e.g. by soclose), then tcptw has already been discarded
223 * (or reused) and inpcb is freed here
224 *
225 * In all three cases the tcptw should not be freed here.
226 */
227 if (inp->inp_flags & INP_DROPPED) {
228 in_pcbdetach(inp);
229 if (__predict_true(tp == NULL)) {
230 in_pcbfree(inp);
231 } else {
232 /*
233 * This case should not happen as in TIMEWAIT
234 * state the inp should not be destroyed before
235 * its tcptw. If INVARIANTS is defined, panic.
236 */
237 #ifdef INVARIANTS
238 panic("%s: Panic before an inp double-free: "
239 "INP_TIMEWAIT && INP_DROPPED && tp != NULL"
240 , __func__);
241 #else
242 log(LOG_ERR, "%s: Avoid an inp double-free: "
243 "INP_TIMEWAIT && INP_DROPPED && tp != NULL"
244 , __func__);
245 #endif
246 INP_WUNLOCK(inp);
247 }
248 } else {
249 in_pcbdetach(inp);
250 INP_WUNLOCK(inp);
251 }
252 } else {
253 /*
254 * If the connection is not in timewait, we consider two
255 * two conditions: one in which no further processing is
256 * necessary (dropped || embryonic), and one in which TCP is
257 * not yet done, but no longer requires the socket, so the
258 * pcb will persist for the time being.
259 *
260 * XXXRW: Does the second case still occur?
261 */
262 if (inp->inp_flags & INP_DROPPED ||
263 tp->t_state < TCPS_SYN_SENT) {
264 tcp_discardcb(tp);
265 in_pcbdetach(inp);
266 in_pcbfree(inp);
267 } else {
268 in_pcbdetach(inp);
269 INP_WUNLOCK(inp);
270 }
271 }
272 }
273
274 /*
275 * pru_detach() detaches the TCP protocol from the socket.
276 * If the protocol state is non-embryonic, then can't
277 * do this directly: have to initiate a pru_disconnect(),
278 * which may finish later; embryonic TCB's can just
279 * be discarded here.
280 */
281 static void
tcp_usr_detach(struct socket * so)282 tcp_usr_detach(struct socket *so)
283 {
284 struct inpcb *inp;
285 int rlock = 0;
286 struct epoch_tracker et;
287
288 inp = sotoinpcb(so);
289 KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
290 if (!INP_INFO_WLOCKED(&V_tcbinfo)) {
291 INP_INFO_RLOCK_ET(&V_tcbinfo, et);
292 rlock = 1;
293 }
294 INP_WLOCK(inp);
295 KASSERT(inp->inp_socket != NULL,
296 ("tcp_usr_detach: inp_socket == NULL"));
297 tcp_detach(so, inp);
298 if (rlock)
299 INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
300 }
301
302 #ifdef INET
303 /*
304 * Give the socket an address.
305 */
306 static int
tcp_usr_bind(struct socket * so,struct sockaddr * nam,struct thread * td)307 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
308 {
309 int error = 0;
310 struct inpcb *inp;
311 struct tcpcb *tp = NULL;
312 struct sockaddr_in *sinp;
313
314 sinp = (struct sockaddr_in *)nam;
315 if (nam->sa_len != sizeof (*sinp))
316 return (EINVAL);
317 /*
318 * Must check for multicast addresses and disallow binding
319 * to them.
320 */
321 if (sinp->sin_family == AF_INET &&
322 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
323 return (EAFNOSUPPORT);
324
325 TCPDEBUG0;
326 inp = sotoinpcb(so);
327 KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
328 INP_WLOCK(inp);
329 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
330 error = EINVAL;
331 goto out;
332 }
333 tp = intotcpcb(inp);
334 TCPDEBUG1();
335 INP_HASH_WLOCK(&V_tcbinfo);
336 error = in_pcbbind(inp, nam, td->td_ucred);
337 INP_HASH_WUNLOCK(&V_tcbinfo);
338 out:
339 TCPDEBUG2(PRU_BIND);
340 TCP_PROBE2(debug__user, tp, PRU_BIND);
341 INP_WUNLOCK(inp);
342
343 return (error);
344 }
345 #endif /* INET */
346
347 #ifdef INET6
348 static int
tcp6_usr_bind(struct socket * so,struct sockaddr * nam,struct thread * td)349 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
350 {
351 int error = 0;
352 struct inpcb *inp;
353 struct tcpcb *tp = NULL;
354 struct sockaddr_in6 *sin6;
355 u_char vflagsav;
356
357 sin6 = (struct sockaddr_in6 *)nam;
358 if (nam->sa_len != sizeof (*sin6))
359 return (EINVAL);
360 /*
361 * Must check for multicast addresses and disallow binding
362 * to them.
363 */
364 if (sin6->sin6_family == AF_INET6 &&
365 IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
366 return (EAFNOSUPPORT);
367
368 TCPDEBUG0;
369 inp = sotoinpcb(so);
370 KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
371 INP_WLOCK(inp);
372 vflagsav = inp->inp_vflag;
373 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
374 error = EINVAL;
375 goto out;
376 }
377 tp = intotcpcb(inp);
378 TCPDEBUG1();
379 INP_HASH_WLOCK(&V_tcbinfo);
380 inp->inp_vflag &= ~INP_IPV4;
381 inp->inp_vflag |= INP_IPV6;
382 #ifdef INET
383 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
384 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
385 inp->inp_vflag |= INP_IPV4;
386 else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
387 struct sockaddr_in sin;
388
389 in6_sin6_2_sin(&sin, sin6);
390 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
391 error = EAFNOSUPPORT;
392 INP_HASH_WUNLOCK(&V_tcbinfo);
393 goto out;
394 }
395 inp->inp_vflag |= INP_IPV4;
396 inp->inp_vflag &= ~INP_IPV6;
397 error = in_pcbbind(inp, (struct sockaddr *)&sin,
398 td->td_ucred);
399 INP_HASH_WUNLOCK(&V_tcbinfo);
400 goto out;
401 }
402 }
403 #endif
404 error = in6_pcbbind(inp, nam, td->td_ucred);
405 INP_HASH_WUNLOCK(&V_tcbinfo);
406 out:
407 if (error != 0)
408 inp->inp_vflag = vflagsav;
409 TCPDEBUG2(PRU_BIND);
410 TCP_PROBE2(debug__user, tp, PRU_BIND);
411 INP_WUNLOCK(inp);
412 return (error);
413 }
414 #endif /* INET6 */
415
416 #ifdef INET
417 /*
418 * Prepare to accept connections.
419 */
420 static int
tcp_usr_listen(struct socket * so,int backlog,struct thread * td)421 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
422 {
423 int error = 0;
424 struct inpcb *inp;
425 struct tcpcb *tp = NULL;
426
427 TCPDEBUG0;
428 inp = sotoinpcb(so);
429 KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
430 INP_WLOCK(inp);
431 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
432 error = EINVAL;
433 goto out;
434 }
435 tp = intotcpcb(inp);
436 TCPDEBUG1();
437 SOCK_LOCK(so);
438 error = solisten_proto_check(so);
439 INP_HASH_WLOCK(&V_tcbinfo);
440 if (error == 0 && inp->inp_lport == 0)
441 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
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
453 if (IS_FASTOPEN(tp->t_flags))
454 tp->t_tfo_pending = tcp_fastopen_alloc_counter();
455
456 out:
457 TCPDEBUG2(PRU_LISTEN);
458 TCP_PROBE2(debug__user, tp, PRU_LISTEN);
459 INP_WUNLOCK(inp);
460 return (error);
461 }
462 #endif /* INET */
463
464 #ifdef INET6
465 static int
tcp6_usr_listen(struct socket * so,int backlog,struct thread * td)466 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
467 {
468 int error = 0;
469 struct inpcb *inp;
470 struct tcpcb *tp = NULL;
471 u_char vflagsav;
472
473 TCPDEBUG0;
474 inp = sotoinpcb(so);
475 KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
476 INP_WLOCK(inp);
477 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
478 error = EINVAL;
479 goto out;
480 }
481 vflagsav = inp->inp_vflag;
482 tp = intotcpcb(inp);
483 TCPDEBUG1();
484 SOCK_LOCK(so);
485 error = solisten_proto_check(so);
486 INP_HASH_WLOCK(&V_tcbinfo);
487 if (error == 0 && inp->inp_lport == 0) {
488 inp->inp_vflag &= ~INP_IPV4;
489 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
490 inp->inp_vflag |= INP_IPV4;
491 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
492 }
493 INP_HASH_WUNLOCK(&V_tcbinfo);
494 if (error == 0) {
495 tcp_state_change(tp, TCPS_LISTEN);
496 solisten_proto(so, backlog);
497 #ifdef TCP_OFFLOAD
498 if ((so->so_options & SO_NO_OFFLOAD) == 0)
499 tcp_offload_listen_start(tp);
500 #endif
501 }
502 SOCK_UNLOCK(so);
503
504 if (IS_FASTOPEN(tp->t_flags))
505 tp->t_tfo_pending = tcp_fastopen_alloc_counter();
506
507 if (error != 0)
508 inp->inp_vflag = vflagsav;
509
510 out:
511 TCPDEBUG2(PRU_LISTEN);
512 TCP_PROBE2(debug__user, tp, PRU_LISTEN);
513 INP_WUNLOCK(inp);
514 return (error);
515 }
516 #endif /* INET6 */
517
518 #ifdef INET
519 /*
520 * Initiate connection to peer.
521 * Create a template for use in transmissions on this connection.
522 * Enter SYN_SENT state, and mark socket as connecting.
523 * Start keep-alive timer, and seed output sequence space.
524 * Send initial segment on connection.
525 */
526 static int
tcp_usr_connect(struct socket * so,struct sockaddr * nam,struct thread * td)527 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
528 {
529 int error = 0;
530 struct inpcb *inp;
531 struct tcpcb *tp = NULL;
532 struct sockaddr_in *sinp;
533
534 sinp = (struct sockaddr_in *)nam;
535 if (nam->sa_len != sizeof (*sinp))
536 return (EINVAL);
537 /*
538 * Must disallow TCP ``connections'' to multicast addresses.
539 */
540 if (sinp->sin_family == AF_INET
541 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
542 return (EAFNOSUPPORT);
543 if ((sinp->sin_family == AF_INET) &&
544 (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST))
545 return (EACCES);
546 if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
547 return (error);
548
549 TCPDEBUG0;
550 inp = sotoinpcb(so);
551 KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
552 INP_WLOCK(inp);
553 if (inp->inp_flags & INP_TIMEWAIT) {
554 error = EADDRINUSE;
555 goto out;
556 }
557 if (inp->inp_flags & INP_DROPPED) {
558 error = ECONNREFUSED;
559 goto out;
560 }
561 tp = intotcpcb(inp);
562 TCPDEBUG1();
563 if ((error = tcp_connect(tp, nam, td)) != 0)
564 goto out;
565 #ifdef TCP_OFFLOAD
566 if (registered_toedevs > 0 &&
567 (so->so_options & SO_NO_OFFLOAD) == 0 &&
568 (error = tcp_offload_connect(so, nam)) == 0)
569 goto out;
570 #endif
571 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
572 error = tp->t_fb->tfb_tcp_output(tp);
573 out:
574 TCPDEBUG2(PRU_CONNECT);
575 TCP_PROBE2(debug__user, tp, PRU_CONNECT);
576 INP_WUNLOCK(inp);
577 return (error);
578 }
579 #endif /* INET */
580
581 #ifdef INET6
582 static int
tcp6_usr_connect(struct socket * so,struct sockaddr * nam,struct thread * td)583 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
584 {
585 int error = 0;
586 struct inpcb *inp;
587 struct tcpcb *tp = NULL;
588 struct sockaddr_in6 *sin6;
589 u_int8_t incflagsav;
590 u_char vflagsav;
591
592 TCPDEBUG0;
593
594 sin6 = (struct sockaddr_in6 *)nam;
595 if (nam->sa_len != sizeof (*sin6))
596 return (EINVAL);
597 /*
598 * Must disallow TCP ``connections'' to multicast addresses.
599 */
600 if (sin6->sin6_family == AF_INET6
601 && IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
602 return (EAFNOSUPPORT);
603
604 inp = sotoinpcb(so);
605 KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
606 INP_WLOCK(inp);
607 vflagsav = inp->inp_vflag;
608 incflagsav = inp->inp_inc.inc_flags;
609 if (inp->inp_flags & INP_TIMEWAIT) {
610 error = EADDRINUSE;
611 goto out;
612 }
613 if (inp->inp_flags & INP_DROPPED) {
614 error = ECONNREFUSED;
615 goto out;
616 }
617 tp = intotcpcb(inp);
618 TCPDEBUG1();
619 #ifdef INET
620 /*
621 * XXXRW: Some confusion: V4/V6 flags relate to binding, and
622 * therefore probably require the hash lock, which isn't held here.
623 * Is this a significant problem?
624 */
625 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
626 struct sockaddr_in sin;
627
628 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
629 error = EINVAL;
630 goto out;
631 }
632 if ((inp->inp_vflag & INP_IPV4) == 0) {
633 error = EAFNOSUPPORT;
634 goto out;
635 }
636
637 in6_sin6_2_sin(&sin, sin6);
638 if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
639 error = EAFNOSUPPORT;
640 goto out;
641 }
642 if (ntohl(sin.sin_addr.s_addr) == INADDR_BROADCAST) {
643 error = EACCES;
644 goto out;
645 }
646 if ((error = prison_remote_ip4(td->td_ucred,
647 &sin.sin_addr)) != 0)
648 goto out;
649 inp->inp_vflag |= INP_IPV4;
650 inp->inp_vflag &= ~INP_IPV6;
651 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
652 goto out;
653 #ifdef TCP_OFFLOAD
654 if (registered_toedevs > 0 &&
655 (so->so_options & SO_NO_OFFLOAD) == 0 &&
656 (error = tcp_offload_connect(so, nam)) == 0)
657 goto out;
658 #endif
659 error = tp->t_fb->tfb_tcp_output(tp);
660 goto out;
661 } else {
662 if ((inp->inp_vflag & INP_IPV6) == 0) {
663 error = EAFNOSUPPORT;
664 goto out;
665 }
666 }
667 #endif
668 if ((error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr)) != 0)
669 goto out;
670 inp->inp_vflag &= ~INP_IPV4;
671 inp->inp_vflag |= INP_IPV6;
672 inp->inp_inc.inc_flags |= INC_ISIPV6;
673 if ((error = tcp6_connect(tp, nam, td)) != 0)
674 goto out;
675 #ifdef TCP_OFFLOAD
676 if (registered_toedevs > 0 &&
677 (so->so_options & SO_NO_OFFLOAD) == 0 &&
678 (error = tcp_offload_connect(so, nam)) == 0)
679 goto out;
680 #endif
681 tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
682 error = tp->t_fb->tfb_tcp_output(tp);
683
684 out:
685 /*
686 * If the implicit bind in the connect call fails, restore
687 * the flags we modified.
688 */
689 if (error != 0 && inp->inp_lport == 0) {
690 inp->inp_vflag = vflagsav;
691 inp->inp_inc.inc_flags = incflagsav;
692 }
693
694 TCPDEBUG2(PRU_CONNECT);
695 TCP_PROBE2(debug__user, tp, PRU_CONNECT);
696 INP_WUNLOCK(inp);
697 return (error);
698 }
699 #endif /* INET6 */
700
701 /*
702 * Initiate disconnect from peer.
703 * If connection never passed embryonic stage, just drop;
704 * else if don't need to let data drain, then can just drop anyways,
705 * else have to begin TCP shutdown process: mark socket disconnecting,
706 * drain unread data, state switch to reflect user close, and
707 * send segment (e.g. FIN) to peer. Socket will be really disconnected
708 * when peer sends FIN and acks ours.
709 *
710 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
711 */
712 static int
tcp_usr_disconnect(struct socket * so)713 tcp_usr_disconnect(struct socket *so)
714 {
715 struct inpcb *inp;
716 struct tcpcb *tp = NULL;
717 struct epoch_tracker et;
718 int error = 0;
719
720 TCPDEBUG0;
721 INP_INFO_RLOCK_ET(&V_tcbinfo, et);
722 inp = sotoinpcb(so);
723 KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
724 INP_WLOCK(inp);
725 if (inp->inp_flags & INP_TIMEWAIT)
726 goto out;
727 if (inp->inp_flags & INP_DROPPED) {
728 error = ECONNRESET;
729 goto out;
730 }
731 tp = intotcpcb(inp);
732 TCPDEBUG1();
733 tcp_disconnect(tp);
734 out:
735 TCPDEBUG2(PRU_DISCONNECT);
736 TCP_PROBE2(debug__user, tp, PRU_DISCONNECT);
737 INP_WUNLOCK(inp);
738 INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
739 return (error);
740 }
741
742 #ifdef INET
743 /*
744 * Accept a connection. Essentially all the work is done at higher levels;
745 * just return the address of the peer, storing through addr.
746 */
747 static int
tcp_usr_accept(struct socket * so,struct sockaddr ** nam)748 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
749 {
750 int error = 0;
751 struct inpcb *inp = NULL;
752 struct tcpcb *tp = NULL;
753 struct in_addr addr;
754 in_port_t port = 0;
755 TCPDEBUG0;
756
757 if (so->so_state & SS_ISDISCONNECTED)
758 return (ECONNABORTED);
759
760 inp = sotoinpcb(so);
761 KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
762 INP_WLOCK(inp);
763 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
764 error = ECONNABORTED;
765 goto out;
766 }
767 tp = intotcpcb(inp);
768 TCPDEBUG1();
769
770 /*
771 * We inline in_getpeeraddr and COMMON_END here, so that we can
772 * copy the data of interest and defer the malloc until after we
773 * release the lock.
774 */
775 port = inp->inp_fport;
776 addr = inp->inp_faddr;
777
778 out:
779 TCPDEBUG2(PRU_ACCEPT);
780 TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
781 INP_WUNLOCK(inp);
782 if (error == 0)
783 *nam = in_sockaddr(port, &addr);
784 return error;
785 }
786 #endif /* INET */
787
788 #ifdef INET6
789 static int
tcp6_usr_accept(struct socket * so,struct sockaddr ** nam)790 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
791 {
792 struct inpcb *inp = NULL;
793 int error = 0;
794 struct tcpcb *tp = NULL;
795 struct in_addr addr;
796 struct in6_addr addr6;
797 struct epoch_tracker et;
798 in_port_t port = 0;
799 int v4 = 0;
800 TCPDEBUG0;
801
802 if (so->so_state & SS_ISDISCONNECTED)
803 return (ECONNABORTED);
804
805 inp = sotoinpcb(so);
806 KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
807 INP_INFO_RLOCK_ET(&V_tcbinfo, et);
808 INP_WLOCK(inp);
809 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
810 error = ECONNABORTED;
811 goto out;
812 }
813 tp = intotcpcb(inp);
814 TCPDEBUG1();
815
816 /*
817 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
818 * copy the data of interest and defer the malloc until after we
819 * release the lock.
820 */
821 if (inp->inp_vflag & INP_IPV4) {
822 v4 = 1;
823 port = inp->inp_fport;
824 addr = inp->inp_faddr;
825 } else {
826 port = inp->inp_fport;
827 addr6 = inp->in6p_faddr;
828 }
829
830 out:
831 TCPDEBUG2(PRU_ACCEPT);
832 TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
833 INP_WUNLOCK(inp);
834 INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
835 if (error == 0) {
836 if (v4)
837 *nam = in6_v4mapsin6_sockaddr(port, &addr);
838 else
839 *nam = in6_sockaddr(port, &addr6);
840 }
841 return error;
842 }
843 #endif /* INET6 */
844
845 /*
846 * Mark the connection as being incapable of further output.
847 */
848 static int
tcp_usr_shutdown(struct socket * so)849 tcp_usr_shutdown(struct socket *so)
850 {
851 int error = 0;
852 struct inpcb *inp;
853 struct tcpcb *tp = NULL;
854 struct epoch_tracker et;
855
856 TCPDEBUG0;
857 INP_INFO_RLOCK_ET(&V_tcbinfo, et);
858 inp = sotoinpcb(so);
859 KASSERT(inp != NULL, ("inp == NULL"));
860 INP_WLOCK(inp);
861 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
862 error = ECONNRESET;
863 goto out;
864 }
865 tp = intotcpcb(inp);
866 TCPDEBUG1();
867 socantsendmore(so);
868 tcp_usrclosed(tp);
869 if (!(inp->inp_flags & INP_DROPPED))
870 error = tp->t_fb->tfb_tcp_output(tp);
871
872 out:
873 TCPDEBUG2(PRU_SHUTDOWN);
874 TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN);
875 INP_WUNLOCK(inp);
876 INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
877
878 return (error);
879 }
880
881 /*
882 * After a receive, possibly send window update to peer.
883 */
884 static int
tcp_usr_rcvd(struct socket * so,int flags)885 tcp_usr_rcvd(struct socket *so, int flags)
886 {
887 struct inpcb *inp;
888 struct tcpcb *tp = NULL;
889 int error = 0;
890
891 TCPDEBUG0;
892 inp = sotoinpcb(so);
893 KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
894 INP_WLOCK(inp);
895 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
896 error = ECONNRESET;
897 goto out;
898 }
899 tp = intotcpcb(inp);
900 TCPDEBUG1();
901 /*
902 * For passively-created TFO connections, don't attempt a window
903 * update while still in SYN_RECEIVED as this may trigger an early
904 * SYN|ACK. It is preferable to have the SYN|ACK be sent along with
905 * application response data, or failing that, when the DELACK timer
906 * expires.
907 */
908 if (IS_FASTOPEN(tp->t_flags) &&
909 (tp->t_state == TCPS_SYN_RECEIVED))
910 goto out;
911 #ifdef TCP_OFFLOAD
912 if (tp->t_flags & TF_TOE)
913 tcp_offload_rcvd(tp);
914 else
915 #endif
916 tp->t_fb->tfb_tcp_output(tp);
917
918 out:
919 TCPDEBUG2(PRU_RCVD);
920 TCP_PROBE2(debug__user, tp, PRU_RCVD);
921 INP_WUNLOCK(inp);
922 return (error);
923 }
924
925 /*
926 * Do a send by putting data in output queue and updating urgent
927 * marker if URG set. Possibly send more data. Unlike the other
928 * pru_*() routines, the mbuf chains are our responsibility. We
929 * must either enqueue them or free them. The other pru_* routines
930 * generally are caller-frees.
931 */
932 static int
tcp_usr_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct thread * td)933 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
934 struct sockaddr *nam, struct mbuf *control, struct thread *td)
935 {
936 int error = 0;
937 struct inpcb *inp;
938 struct tcpcb *tp = NULL;
939 struct epoch_tracker net_et;
940 #ifdef INET
941 #ifdef INET6
942 struct sockaddr_in sin;
943 #endif
944 struct sockaddr_in *sinp;
945 #endif
946 #ifdef INET6
947 int isipv6;
948 #endif
949 u_int8_t incflagsav;
950 u_char vflagsav;
951 bool restoreflags;
952 TCPDEBUG0;
953
954 /*
955 * We require the pcbinfo lock if we will close the socket as part of
956 * this call.
957 */
958 if (flags & PRUS_EOF)
959 INP_INFO_RLOCK_ET(&V_tcbinfo, net_et);
960 inp = sotoinpcb(so);
961 KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
962 INP_WLOCK(inp);
963 vflagsav = inp->inp_vflag;
964 incflagsav = inp->inp_inc.inc_flags;
965 restoreflags = false;
966 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
967 if (control)
968 m_freem(control);
969 /*
970 * In case of PRUS_NOTREADY, tcp_usr_ready() is responsible
971 * for freeing memory.
972 */
973 if (m && (flags & PRUS_NOTREADY) == 0)
974 m_freem(m);
975 error = ECONNRESET;
976 goto out;
977 }
978 tp = intotcpcb(inp);
979 TCPDEBUG1();
980 if (nam != NULL && tp->t_state < TCPS_SYN_SENT) {
981 switch (nam->sa_family) {
982 #ifdef INET
983 case AF_INET:
984 sinp = (struct sockaddr_in *)nam;
985 if (sinp->sin_len != sizeof(struct sockaddr_in)) {
986 if (m)
987 m_freem(m);
988 error = EINVAL;
989 goto out;
990 }
991 if ((inp->inp_vflag & INP_IPV6) != 0) {
992 if (m)
993 m_freem(m);
994 error = EAFNOSUPPORT;
995 goto out;
996 }
997 if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
998 if (m)
999 m_freem(m);
1000 error = EAFNOSUPPORT;
1001 goto out;
1002 }
1003 if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) {
1004 if (m)
1005 m_freem(m);
1006 error = EACCES;
1007 goto out;
1008 }
1009 if ((error = prison_remote_ip4(td->td_ucred,
1010 &sinp->sin_addr))) {
1011 if (m)
1012 m_freem(m);
1013 goto out;
1014 }
1015 #ifdef INET6
1016 isipv6 = 0;
1017 #endif
1018 break;
1019 #endif /* INET */
1020 #ifdef INET6
1021 case AF_INET6:
1022 {
1023 struct sockaddr_in6 *sin6;
1024
1025 sin6 = (struct sockaddr_in6 *)nam;
1026 if (sin6->sin6_len != sizeof(*sin6)) {
1027 if (m)
1028 m_freem(m);
1029 error = EINVAL;
1030 goto out;
1031 }
1032 if ((inp->inp_vflag & INP_IPV6PROTO) == 0) {
1033 if (m != NULL)
1034 m_freem(m);
1035 error = EAFNOSUPPORT;
1036 goto out;
1037 }
1038 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
1039 if (m)
1040 m_freem(m);
1041 error = EAFNOSUPPORT;
1042 goto out;
1043 }
1044 if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1045 #ifdef INET
1046 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1047 error = EINVAL;
1048 if (m)
1049 m_freem(m);
1050 goto out;
1051 }
1052 if ((inp->inp_vflag & INP_IPV4) == 0) {
1053 error = EAFNOSUPPORT;
1054 if (m)
1055 m_freem(m);
1056 goto out;
1057 }
1058 restoreflags = true;
1059 inp->inp_vflag &= ~INP_IPV6;
1060 sinp = &sin;
1061 in6_sin6_2_sin(sinp, sin6);
1062 if (IN_MULTICAST(
1063 ntohl(sinp->sin_addr.s_addr))) {
1064 error = EAFNOSUPPORT;
1065 if (m)
1066 m_freem(m);
1067 goto out;
1068 }
1069 if ((error = prison_remote_ip4(td->td_ucred,
1070 &sinp->sin_addr))) {
1071 if (m)
1072 m_freem(m);
1073 goto out;
1074 }
1075 isipv6 = 0;
1076 #else /* !INET */
1077 error = EAFNOSUPPORT;
1078 if (m)
1079 m_freem(m);
1080 goto out;
1081 #endif /* INET */
1082 } else {
1083 if ((inp->inp_vflag & INP_IPV6) == 0) {
1084 if (m)
1085 m_freem(m);
1086 error = EAFNOSUPPORT;
1087 goto out;
1088 }
1089 restoreflags = true;
1090 inp->inp_vflag &= ~INP_IPV4;
1091 inp->inp_inc.inc_flags |= INC_ISIPV6;
1092 if ((error = prison_remote_ip6(td->td_ucred,
1093 &sin6->sin6_addr))) {
1094 if (m)
1095 m_freem(m);
1096 goto out;
1097 }
1098 isipv6 = 1;
1099 }
1100 break;
1101 }
1102 #endif /* INET6 */
1103 default:
1104 if (m)
1105 m_freem(m);
1106 error = EAFNOSUPPORT;
1107 goto out;
1108 }
1109 }
1110 if (control) {
1111 /* TCP doesn't do control messages (rights, creds, etc) */
1112 if (control->m_len) {
1113 m_freem(control);
1114 if (m)
1115 m_freem(m);
1116 error = EINVAL;
1117 goto out;
1118 }
1119 m_freem(control); /* empty control, just free it */
1120 }
1121 if (!(flags & PRUS_OOB)) {
1122 sbappendstream(&so->so_snd, m, flags);
1123 if (nam && tp->t_state < TCPS_SYN_SENT) {
1124 /*
1125 * Do implied connect if not yet connected,
1126 * initialize window to default value, and
1127 * initialize maxseg using peer's cached MSS.
1128 */
1129 #ifdef INET6
1130 if (isipv6)
1131 error = tcp6_connect(tp, nam, td);
1132 #endif /* INET6 */
1133 #if defined(INET6) && defined(INET)
1134 else
1135 #endif
1136 #ifdef INET
1137 error = tcp_connect(tp,
1138 (struct sockaddr *)sinp, td);
1139 #endif
1140 /*
1141 * The bind operation in tcp_connect succeeded. We
1142 * no longer want to restore the flags if later
1143 * operations fail.
1144 */
1145 if (error == 0 || inp->inp_lport != 0)
1146 restoreflags = false;
1147
1148 if (error)
1149 goto out;
1150 if (IS_FASTOPEN(tp->t_flags))
1151 tcp_fastopen_connect(tp);
1152 else {
1153 tp->snd_wnd = TTCP_CLIENT_SND_WND;
1154 tcp_mss(tp, -1);
1155 }
1156 }
1157 if (flags & PRUS_EOF) {
1158 /*
1159 * Close the send side of the connection after
1160 * the data is sent.
1161 */
1162 INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
1163 socantsendmore(so);
1164 tcp_usrclosed(tp);
1165 }
1166 if (!(inp->inp_flags & INP_DROPPED) &&
1167 !(flags & PRUS_NOTREADY)) {
1168 if (flags & PRUS_MORETOCOME)
1169 tp->t_flags |= TF_MORETOCOME;
1170 error = tp->t_fb->tfb_tcp_output(tp);
1171 if (flags & PRUS_MORETOCOME)
1172 tp->t_flags &= ~TF_MORETOCOME;
1173 }
1174 } else {
1175 /*
1176 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
1177 */
1178 SOCKBUF_LOCK(&so->so_snd);
1179 if (sbspace(&so->so_snd) < -512) {
1180 SOCKBUF_UNLOCK(&so->so_snd);
1181 m_freem(m);
1182 error = ENOBUFS;
1183 goto out;
1184 }
1185 /*
1186 * According to RFC961 (Assigned Protocols),
1187 * the urgent pointer points to the last octet
1188 * of urgent data. We continue, however,
1189 * to consider it to indicate the first octet
1190 * of data past the urgent section.
1191 * Otherwise, snd_up should be one lower.
1192 */
1193 sbappendstream_locked(&so->so_snd, m, flags);
1194 SOCKBUF_UNLOCK(&so->so_snd);
1195 if (nam && tp->t_state < TCPS_SYN_SENT) {
1196 /*
1197 * Do implied connect if not yet connected,
1198 * initialize window to default value, and
1199 * initialize maxseg using peer's cached MSS.
1200 */
1201
1202 /*
1203 * Not going to contemplate SYN|URG
1204 */
1205 if (IS_FASTOPEN(tp->t_flags))
1206 tp->t_flags &= ~TF_FASTOPEN;
1207 #ifdef INET6
1208 if (isipv6)
1209 error = tcp6_connect(tp, nam, td);
1210 #endif /* INET6 */
1211 #if defined(INET6) && defined(INET)
1212 else
1213 #endif
1214 #ifdef INET
1215 error = tcp_connect(tp,
1216 (struct sockaddr *)sinp, td);
1217 #endif
1218 /*
1219 * The bind operation in tcp_connect succeeded. We
1220 * no longer want to restore the flags if later
1221 * operations fail.
1222 */
1223 if (error == 0 || inp->inp_lport != 0)
1224 restoreflags = false;
1225
1226 if (error)
1227 goto out;
1228 tp->snd_wnd = TTCP_CLIENT_SND_WND;
1229 tcp_mss(tp, -1);
1230 }
1231 tp->snd_up = tp->snd_una + sbavail(&so->so_snd);
1232 if (!(flags & PRUS_NOTREADY)) {
1233 tp->t_flags |= TF_FORCEDATA;
1234 error = tp->t_fb->tfb_tcp_output(tp);
1235 tp->t_flags &= ~TF_FORCEDATA;
1236 }
1237 }
1238 TCP_LOG_EVENT(tp, NULL,
1239 &inp->inp_socket->so_rcv,
1240 &inp->inp_socket->so_snd,
1241 TCP_LOG_USERSEND, error,
1242 0, NULL, false);
1243 out:
1244 /*
1245 * If the request was unsuccessful and we changed flags,
1246 * restore the original flags.
1247 */
1248 if (error != 0 && restoreflags) {
1249 inp->inp_vflag = vflagsav;
1250 inp->inp_inc.inc_flags = incflagsav;
1251 }
1252 TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
1253 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1254 TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB :
1255 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1256 INP_WUNLOCK(inp);
1257 if (flags & PRUS_EOF)
1258 INP_INFO_RUNLOCK_ET(&V_tcbinfo, net_et);
1259 return (error);
1260 }
1261
1262 static int
tcp_usr_ready(struct socket * so,struct mbuf * m,int count)1263 tcp_usr_ready(struct socket *so, struct mbuf *m, int count)
1264 {
1265 struct inpcb *inp;
1266 struct tcpcb *tp;
1267 int error;
1268
1269 inp = sotoinpcb(so);
1270 INP_WLOCK(inp);
1271 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1272 INP_WUNLOCK(inp);
1273 for (int i = 0; i < count; i++)
1274 m = m_free(m);
1275 return (ECONNRESET);
1276 }
1277 tp = intotcpcb(inp);
1278
1279 SOCKBUF_LOCK(&so->so_snd);
1280 error = sbready(&so->so_snd, m, count);
1281 SOCKBUF_UNLOCK(&so->so_snd);
1282 if (error == 0)
1283 error = tp->t_fb->tfb_tcp_output(tp);
1284 INP_WUNLOCK(inp);
1285
1286 return (error);
1287 }
1288
1289 /*
1290 * Abort the TCP. Drop the connection abruptly.
1291 */
1292 static void
tcp_usr_abort(struct socket * so)1293 tcp_usr_abort(struct socket *so)
1294 {
1295 struct inpcb *inp;
1296 struct tcpcb *tp = NULL;
1297 struct epoch_tracker et;
1298 TCPDEBUG0;
1299
1300 inp = sotoinpcb(so);
1301 KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
1302
1303 INP_INFO_RLOCK_ET(&V_tcbinfo, et);
1304 INP_WLOCK(inp);
1305 KASSERT(inp->inp_socket != NULL,
1306 ("tcp_usr_abort: inp_socket == NULL"));
1307
1308 /*
1309 * If we still have full TCP state, and we're not dropped, drop.
1310 */
1311 if (!(inp->inp_flags & INP_TIMEWAIT) &&
1312 !(inp->inp_flags & INP_DROPPED)) {
1313 tp = intotcpcb(inp);
1314 TCPDEBUG1();
1315 tp = tcp_drop(tp, ECONNABORTED);
1316 if (tp == NULL)
1317 goto dropped;
1318 TCPDEBUG2(PRU_ABORT);
1319 TCP_PROBE2(debug__user, tp, PRU_ABORT);
1320 }
1321 if (!(inp->inp_flags & INP_DROPPED)) {
1322 SOCK_LOCK(so);
1323 so->so_state |= SS_PROTOREF;
1324 SOCK_UNLOCK(so);
1325 inp->inp_flags |= INP_SOCKREF;
1326 }
1327 INP_WUNLOCK(inp);
1328 dropped:
1329 INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
1330 }
1331
1332 /*
1333 * TCP socket is closed. Start friendly disconnect.
1334 */
1335 static void
tcp_usr_close(struct socket * so)1336 tcp_usr_close(struct socket *so)
1337 {
1338 struct inpcb *inp;
1339 struct tcpcb *tp = NULL;
1340 struct epoch_tracker et;
1341 TCPDEBUG0;
1342
1343 inp = sotoinpcb(so);
1344 KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
1345
1346 INP_INFO_RLOCK_ET(&V_tcbinfo, et);
1347 INP_WLOCK(inp);
1348 KASSERT(inp->inp_socket != NULL,
1349 ("tcp_usr_close: inp_socket == NULL"));
1350
1351 /*
1352 * If we still have full TCP state, and we're not dropped, initiate
1353 * a disconnect.
1354 */
1355 if (!(inp->inp_flags & INP_TIMEWAIT) &&
1356 !(inp->inp_flags & INP_DROPPED)) {
1357 tp = intotcpcb(inp);
1358 TCPDEBUG1();
1359 tcp_disconnect(tp);
1360 TCPDEBUG2(PRU_CLOSE);
1361 TCP_PROBE2(debug__user, tp, PRU_CLOSE);
1362 }
1363 if (!(inp->inp_flags & INP_DROPPED)) {
1364 SOCK_LOCK(so);
1365 so->so_state |= SS_PROTOREF;
1366 SOCK_UNLOCK(so);
1367 inp->inp_flags |= INP_SOCKREF;
1368 }
1369 INP_WUNLOCK(inp);
1370 INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
1371 }
1372
1373 /*
1374 * Receive out-of-band data.
1375 */
1376 static int
tcp_usr_rcvoob(struct socket * so,struct mbuf * m,int flags)1377 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1378 {
1379 int error = 0;
1380 struct inpcb *inp;
1381 struct tcpcb *tp = NULL;
1382
1383 TCPDEBUG0;
1384 inp = sotoinpcb(so);
1385 KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1386 INP_WLOCK(inp);
1387 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1388 error = ECONNRESET;
1389 goto out;
1390 }
1391 tp = intotcpcb(inp);
1392 TCPDEBUG1();
1393 if ((so->so_oobmark == 0 &&
1394 (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1395 so->so_options & SO_OOBINLINE ||
1396 tp->t_oobflags & TCPOOB_HADDATA) {
1397 error = EINVAL;
1398 goto out;
1399 }
1400 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1401 error = EWOULDBLOCK;
1402 goto out;
1403 }
1404 m->m_len = 1;
1405 *mtod(m, caddr_t) = tp->t_iobc;
1406 if ((flags & MSG_PEEK) == 0)
1407 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1408
1409 out:
1410 TCPDEBUG2(PRU_RCVOOB);
1411 TCP_PROBE2(debug__user, tp, PRU_RCVOOB);
1412 INP_WUNLOCK(inp);
1413 return (error);
1414 }
1415
1416 #ifdef INET
1417 struct pr_usrreqs tcp_usrreqs = {
1418 .pru_abort = tcp_usr_abort,
1419 .pru_accept = tcp_usr_accept,
1420 .pru_attach = tcp_usr_attach,
1421 .pru_bind = tcp_usr_bind,
1422 .pru_connect = tcp_usr_connect,
1423 .pru_control = in_control,
1424 .pru_detach = tcp_usr_detach,
1425 .pru_disconnect = tcp_usr_disconnect,
1426 .pru_listen = tcp_usr_listen,
1427 .pru_peeraddr = in_getpeeraddr,
1428 .pru_rcvd = tcp_usr_rcvd,
1429 .pru_rcvoob = tcp_usr_rcvoob,
1430 .pru_send = tcp_usr_send,
1431 .pru_ready = tcp_usr_ready,
1432 .pru_shutdown = tcp_usr_shutdown,
1433 .pru_sockaddr = in_getsockaddr,
1434 .pru_sosetlabel = in_pcbsosetlabel,
1435 .pru_close = tcp_usr_close,
1436 };
1437 #endif /* INET */
1438
1439 #ifdef INET6
1440 struct pr_usrreqs tcp6_usrreqs = {
1441 .pru_abort = tcp_usr_abort,
1442 .pru_accept = tcp6_usr_accept,
1443 .pru_attach = tcp_usr_attach,
1444 .pru_bind = tcp6_usr_bind,
1445 .pru_connect = tcp6_usr_connect,
1446 .pru_control = in6_control,
1447 .pru_detach = tcp_usr_detach,
1448 .pru_disconnect = tcp_usr_disconnect,
1449 .pru_listen = tcp6_usr_listen,
1450 .pru_peeraddr = in6_mapped_peeraddr,
1451 .pru_rcvd = tcp_usr_rcvd,
1452 .pru_rcvoob = tcp_usr_rcvoob,
1453 .pru_send = tcp_usr_send,
1454 .pru_ready = tcp_usr_ready,
1455 .pru_shutdown = tcp_usr_shutdown,
1456 .pru_sockaddr = in6_mapped_sockaddr,
1457 .pru_sosetlabel = in_pcbsosetlabel,
1458 .pru_close = tcp_usr_close,
1459 };
1460 #endif /* INET6 */
1461
1462 #ifdef INET
1463 /*
1464 * Common subroutine to open a TCP connection to remote host specified
1465 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
1466 * port number if needed. Call in_pcbconnect_setup to do the routing and
1467 * to choose a local host address (interface). If there is an existing
1468 * incarnation of the same connection in TIME-WAIT state and if the remote
1469 * host was sending CC options and if the connection duration was < MSL, then
1470 * truncate the previous TIME-WAIT state and proceed.
1471 * Initialize connection parameters and enter SYN-SENT state.
1472 */
1473 static int
tcp_connect(struct tcpcb * tp,struct sockaddr * nam,struct thread * td)1474 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1475 {
1476 struct inpcb *inp = tp->t_inpcb, *oinp;
1477 struct socket *so = inp->inp_socket;
1478 struct in_addr laddr;
1479 u_short lport;
1480 int error;
1481
1482 INP_WLOCK_ASSERT(inp);
1483 INP_HASH_WLOCK(&V_tcbinfo);
1484
1485 if (V_tcp_require_unique_port && inp->inp_lport == 0) {
1486 error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1487 if (error)
1488 goto out;
1489 }
1490
1491 /*
1492 * Cannot simply call in_pcbconnect, because there might be an
1493 * earlier incarnation of this same connection still in
1494 * TIME_WAIT state, creating an ADDRINUSE error.
1495 */
1496 laddr = inp->inp_laddr;
1497 lport = inp->inp_lport;
1498 error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1499 &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1500 if (error && oinp == NULL)
1501 goto out;
1502 if (oinp) {
1503 error = EADDRINUSE;
1504 goto out;
1505 }
1506 /* Handle initial bind if it hadn't been done in advance. */
1507 if (inp->inp_lport == 0) {
1508 inp->inp_lport = lport;
1509 if (in_pcbinshash(inp) != 0) {
1510 inp->inp_lport = 0;
1511 error = EAGAIN;
1512 goto out;
1513 }
1514 }
1515 inp->inp_laddr = laddr;
1516 in_pcbrehash(inp);
1517 INP_HASH_WUNLOCK(&V_tcbinfo);
1518
1519 /*
1520 * Compute window scaling to request:
1521 * Scale to fit into sweet spot. See tcp_syncache.c.
1522 * XXX: This should move to tcp_output().
1523 */
1524 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1525 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1526 tp->request_r_scale++;
1527
1528 soisconnecting(so);
1529 TCPSTAT_INC(tcps_connattempt);
1530 tcp_state_change(tp, TCPS_SYN_SENT);
1531 tp->iss = tcp_new_isn(&inp->inp_inc);
1532 if (tp->t_flags & TF_REQ_TSTMP)
1533 tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
1534 tcp_sendseqinit(tp);
1535
1536 return 0;
1537
1538 out:
1539 INP_HASH_WUNLOCK(&V_tcbinfo);
1540 return (error);
1541 }
1542 #endif /* INET */
1543
1544 #ifdef INET6
1545 static int
tcp6_connect(struct tcpcb * tp,struct sockaddr * nam,struct thread * td)1546 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1547 {
1548 struct inpcb *inp = tp->t_inpcb;
1549 int error;
1550
1551 INP_WLOCK_ASSERT(inp);
1552 INP_HASH_WLOCK(&V_tcbinfo);
1553
1554 if (V_tcp_require_unique_port && inp->inp_lport == 0) {
1555 error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1556 if (error)
1557 goto out;
1558 }
1559 error = in6_pcbconnect(inp, nam, td->td_ucred);
1560 if (error != 0)
1561 goto out;
1562 INP_HASH_WUNLOCK(&V_tcbinfo);
1563
1564 /* Compute window scaling to request. */
1565 while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1566 (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1567 tp->request_r_scale++;
1568
1569 soisconnecting(inp->inp_socket);
1570 TCPSTAT_INC(tcps_connattempt);
1571 tcp_state_change(tp, TCPS_SYN_SENT);
1572 tp->iss = tcp_new_isn(&inp->inp_inc);
1573 if (tp->t_flags & TF_REQ_TSTMP)
1574 tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
1575 tcp_sendseqinit(tp);
1576
1577 return 0;
1578
1579 out:
1580 INP_HASH_WUNLOCK(&V_tcbinfo);
1581 return error;
1582 }
1583 #endif /* INET6 */
1584
1585 /*
1586 * Export TCP internal state information via a struct tcp_info, based on the
1587 * Linux 2.6 API. Not ABI compatible as our constants are mapped differently
1588 * (TCP state machine, etc). We export all information using FreeBSD-native
1589 * constants -- for example, the numeric values for tcpi_state will differ
1590 * from Linux.
1591 */
1592 static void
tcp_fill_info(struct tcpcb * tp,struct tcp_info * ti)1593 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1594 {
1595
1596 INP_WLOCK_ASSERT(tp->t_inpcb);
1597 bzero(ti, sizeof(*ti));
1598
1599 ti->tcpi_state = tp->t_state;
1600 if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1601 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1602 if (tp->t_flags & TF_SACK_PERMIT)
1603 ti->tcpi_options |= TCPI_OPT_SACK;
1604 if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1605 ti->tcpi_options |= TCPI_OPT_WSCALE;
1606 ti->tcpi_snd_wscale = tp->snd_scale;
1607 ti->tcpi_rcv_wscale = tp->rcv_scale;
1608 }
1609 if (tp->t_flags & TF_ECN_PERMIT)
1610 ti->tcpi_options |= TCPI_OPT_ECN;
1611
1612 ti->tcpi_rto = tp->t_rxtcur * tick;
1613 ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick;
1614 ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1615 ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1616
1617 ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1618 ti->tcpi_snd_cwnd = tp->snd_cwnd;
1619
1620 /*
1621 * FreeBSD-specific extension fields for tcp_info.
1622 */
1623 ti->tcpi_rcv_space = tp->rcv_wnd;
1624 ti->tcpi_rcv_nxt = tp->rcv_nxt;
1625 ti->tcpi_snd_wnd = tp->snd_wnd;
1626 ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */
1627 ti->tcpi_snd_nxt = tp->snd_nxt;
1628 ti->tcpi_snd_mss = tp->t_maxseg;
1629 ti->tcpi_rcv_mss = tp->t_maxseg;
1630 ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1631 ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1632 ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1633 #ifdef TCP_OFFLOAD
1634 if (tp->t_flags & TF_TOE) {
1635 ti->tcpi_options |= TCPI_OPT_TOE;
1636 tcp_offload_tcp_info(tp, ti);
1637 }
1638 #endif
1639 }
1640
1641 /*
1642 * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1643 * socket option arguments. When it re-acquires the lock after the copy, it
1644 * has to revalidate that the connection is still valid for the socket
1645 * option.
1646 */
1647 #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do { \
1648 INP_WLOCK(inp); \
1649 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) { \
1650 INP_WUNLOCK(inp); \
1651 cleanup; \
1652 return (ECONNRESET); \
1653 } \
1654 tp = intotcpcb(inp); \
1655 } while(0)
1656 #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */)
1657
1658 int
tcp_ctloutput(struct socket * so,struct sockopt * sopt)1659 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1660 {
1661 int error;
1662 struct inpcb *inp;
1663 struct tcpcb *tp;
1664 struct tcp_function_block *blk;
1665 struct tcp_function_set fsn;
1666
1667 error = 0;
1668 inp = sotoinpcb(so);
1669 KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1670 if (sopt->sopt_level != IPPROTO_TCP) {
1671 #ifdef INET6
1672 if (inp->inp_vflag & INP_IPV6PROTO) {
1673 error = ip6_ctloutput(so, sopt);
1674 /*
1675 * In case of the IPV6_USE_MIN_MTU socket option,
1676 * the INC_IPV6MINMTU flag to announce a corresponding
1677 * MSS during the initial handshake.
1678 * If the TCP connection is not in the front states,
1679 * just reduce the MSS being used.
1680 * This avoids the sending of TCP segments which will
1681 * be fragmented at the IPv6 layer.
1682 */
1683 if ((error == 0) &&
1684 (sopt->sopt_dir == SOPT_SET) &&
1685 (sopt->sopt_level == IPPROTO_IPV6) &&
1686 (sopt->sopt_name == IPV6_USE_MIN_MTU)) {
1687 INP_WLOCK(inp);
1688 if ((inp->inp_flags &
1689 (INP_TIMEWAIT | INP_DROPPED))) {
1690 INP_WUNLOCK(inp);
1691 return (ECONNRESET);
1692 }
1693 inp->inp_inc.inc_flags |= INC_IPV6MINMTU;
1694 tp = intotcpcb(inp);
1695 if ((tp->t_state >= TCPS_SYN_SENT) &&
1696 (inp->inp_inc.inc_flags & INC_ISIPV6)) {
1697 struct ip6_pktopts *opt;
1698
1699 opt = inp->in6p_outputopts;
1700 if ((opt != NULL) &&
1701 (opt->ip6po_minmtu ==
1702 IP6PO_MINMTU_ALL)) {
1703 if (tp->t_maxseg > TCP6_MSS) {
1704 tp->t_maxseg = TCP6_MSS;
1705 }
1706 }
1707 }
1708 INP_WUNLOCK(inp);
1709 }
1710 }
1711 #endif /* INET6 */
1712 #if defined(INET6) && defined(INET)
1713 else
1714 #endif
1715 #ifdef INET
1716 {
1717 error = ip_ctloutput(so, sopt);
1718 }
1719 #endif
1720 return (error);
1721 }
1722 INP_WLOCK(inp);
1723 if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1724 INP_WUNLOCK(inp);
1725 return (ECONNRESET);
1726 }
1727 tp = intotcpcb(inp);
1728 /*
1729 * Protect the TCP option TCP_FUNCTION_BLK so
1730 * that a sub-function can *never* overwrite this.
1731 */
1732 if ((sopt->sopt_dir == SOPT_SET) &&
1733 (sopt->sopt_name == TCP_FUNCTION_BLK)) {
1734 INP_WUNLOCK(inp);
1735 error = sooptcopyin(sopt, &fsn, sizeof fsn,
1736 sizeof fsn);
1737 if (error)
1738 return (error);
1739 INP_WLOCK_RECHECK(inp);
1740 blk = find_and_ref_tcp_functions(&fsn);
1741 if (blk == NULL) {
1742 INP_WUNLOCK(inp);
1743 return (ENOENT);
1744 }
1745 if (tp->t_fb == blk) {
1746 /* You already have this */
1747 refcount_release(&blk->tfb_refcnt);
1748 INP_WUNLOCK(inp);
1749 return (0);
1750 }
1751 if (tp->t_state != TCPS_CLOSED) {
1752 /*
1753 * The user has advanced the state
1754 * past the initial point, we may not
1755 * be able to switch.
1756 */
1757 if (blk->tfb_tcp_handoff_ok != NULL) {
1758 /*
1759 * Does the stack provide a
1760 * query mechanism, if so it may
1761 * still be possible?
1762 */
1763 error = (*blk->tfb_tcp_handoff_ok)(tp);
1764 } else
1765 error = EINVAL;
1766 if (error) {
1767 refcount_release(&blk->tfb_refcnt);
1768 INP_WUNLOCK(inp);
1769 return(error);
1770 }
1771 }
1772 if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) {
1773 refcount_release(&blk->tfb_refcnt);
1774 INP_WUNLOCK(inp);
1775 return (ENOENT);
1776 }
1777 /*
1778 * Release the old refcnt, the
1779 * lookup acquired a ref on the
1780 * new one already.
1781 */
1782 if (tp->t_fb->tfb_tcp_fb_fini) {
1783 /*
1784 * Tell the stack to cleanup with 0 i.e.
1785 * the tcb is not going away.
1786 */
1787 (*tp->t_fb->tfb_tcp_fb_fini)(tp, 0);
1788 }
1789 #ifdef TCPHPTS
1790 /* Assure that we are not on any hpts */
1791 tcp_hpts_remove(tp->t_inpcb, HPTS_REMOVE_ALL);
1792 #endif
1793 if (blk->tfb_tcp_fb_init) {
1794 error = (*blk->tfb_tcp_fb_init)(tp);
1795 if (error) {
1796 refcount_release(&blk->tfb_refcnt);
1797 if (tp->t_fb->tfb_tcp_fb_init) {
1798 if((*tp->t_fb->tfb_tcp_fb_init)(tp) != 0) {
1799 /* Fall back failed, drop the connection */
1800 INP_WUNLOCK(inp);
1801 soabort(so);
1802 return(error);
1803 }
1804 }
1805 goto err_out;
1806 }
1807 }
1808 refcount_release(&tp->t_fb->tfb_refcnt);
1809 tp->t_fb = blk;
1810 #ifdef TCP_OFFLOAD
1811 if (tp->t_flags & TF_TOE) {
1812 tcp_offload_ctloutput(tp, sopt->sopt_dir,
1813 sopt->sopt_name);
1814 }
1815 #endif
1816 err_out:
1817 INP_WUNLOCK(inp);
1818 return (error);
1819 } else if ((sopt->sopt_dir == SOPT_GET) &&
1820 (sopt->sopt_name == TCP_FUNCTION_BLK)) {
1821 strncpy(fsn.function_set_name, tp->t_fb->tfb_tcp_block_name,
1822 TCP_FUNCTION_NAME_LEN_MAX);
1823 fsn.function_set_name[TCP_FUNCTION_NAME_LEN_MAX - 1] = '\0';
1824 fsn.pcbcnt = tp->t_fb->tfb_refcnt;
1825 INP_WUNLOCK(inp);
1826 error = sooptcopyout(sopt, &fsn, sizeof fsn);
1827 return (error);
1828 }
1829 /* Pass in the INP locked, called must unlock it */
1830 return (tp->t_fb->tfb_tcp_ctloutput(so, sopt, inp, tp));
1831 }
1832
1833 /*
1834 * If this assert becomes untrue, we need to change the size of the buf
1835 * variable in tcp_default_ctloutput().
1836 */
1837 #ifdef CTASSERT
1838 CTASSERT(TCP_CA_NAME_MAX <= TCP_LOG_ID_LEN);
1839 CTASSERT(TCP_LOG_REASON_LEN <= TCP_LOG_ID_LEN);
1840 #endif
1841
1842 int
tcp_default_ctloutput(struct socket * so,struct sockopt * sopt,struct inpcb * inp,struct tcpcb * tp)1843 tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp)
1844 {
1845 int error, opt, optval;
1846 u_int ui;
1847 struct tcp_info ti;
1848 struct cc_algo *algo;
1849 char *pbuf, buf[TCP_LOG_ID_LEN];
1850 size_t len;
1851
1852 /*
1853 * For TCP_CCALGOOPT forward the control to CC module, for both
1854 * SOPT_SET and SOPT_GET.
1855 */
1856 switch (sopt->sopt_name) {
1857 case TCP_CCALGOOPT:
1858 INP_WUNLOCK(inp);
1859 if (sopt->sopt_valsize > CC_ALGOOPT_LIMIT)
1860 return (EINVAL);
1861 pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO);
1862 error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize,
1863 sopt->sopt_valsize);
1864 if (error) {
1865 free(pbuf, M_TEMP);
1866 return (error);
1867 }
1868 INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP));
1869 if (CC_ALGO(tp)->ctl_output != NULL)
1870 error = CC_ALGO(tp)->ctl_output(tp->ccv, sopt, pbuf);
1871 else
1872 error = ENOENT;
1873 INP_WUNLOCK(inp);
1874 if (error == 0 && sopt->sopt_dir == SOPT_GET)
1875 error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize);
1876 free(pbuf, M_TEMP);
1877 return (error);
1878 }
1879
1880 switch (sopt->sopt_dir) {
1881 case SOPT_SET:
1882 switch (sopt->sopt_name) {
1883 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
1884 case TCP_MD5SIG:
1885 if (!TCPMD5_ENABLED()) {
1886 INP_WUNLOCK(inp);
1887 return (ENOPROTOOPT);
1888 }
1889 error = TCPMD5_PCBCTL(inp, sopt);
1890 if (error)
1891 return (error);
1892 goto unlock_and_done;
1893 #endif /* IPSEC */
1894
1895 case TCP_NODELAY:
1896 case TCP_NOOPT:
1897 INP_WUNLOCK(inp);
1898 error = sooptcopyin(sopt, &optval, sizeof optval,
1899 sizeof optval);
1900 if (error)
1901 return (error);
1902
1903 INP_WLOCK_RECHECK(inp);
1904 switch (sopt->sopt_name) {
1905 case TCP_NODELAY:
1906 opt = TF_NODELAY;
1907 break;
1908 case TCP_NOOPT:
1909 opt = TF_NOOPT;
1910 break;
1911 default:
1912 opt = 0; /* dead code to fool gcc */
1913 break;
1914 }
1915
1916 if (optval)
1917 tp->t_flags |= opt;
1918 else
1919 tp->t_flags &= ~opt;
1920 unlock_and_done:
1921 #ifdef TCP_OFFLOAD
1922 if (tp->t_flags & TF_TOE) {
1923 tcp_offload_ctloutput(tp, sopt->sopt_dir,
1924 sopt->sopt_name);
1925 }
1926 #endif
1927 INP_WUNLOCK(inp);
1928 break;
1929
1930 case TCP_NOPUSH:
1931 INP_WUNLOCK(inp);
1932 error = sooptcopyin(sopt, &optval, sizeof optval,
1933 sizeof optval);
1934 if (error)
1935 return (error);
1936
1937 INP_WLOCK_RECHECK(inp);
1938 if (optval)
1939 tp->t_flags |= TF_NOPUSH;
1940 else if (tp->t_flags & TF_NOPUSH) {
1941 tp->t_flags &= ~TF_NOPUSH;
1942 if (TCPS_HAVEESTABLISHED(tp->t_state))
1943 error = tp->t_fb->tfb_tcp_output(tp);
1944 }
1945 goto unlock_and_done;
1946
1947 case TCP_MAXSEG:
1948 INP_WUNLOCK(inp);
1949 error = sooptcopyin(sopt, &optval, sizeof optval,
1950 sizeof optval);
1951 if (error)
1952 return (error);
1953
1954 INP_WLOCK_RECHECK(inp);
1955 if (optval > 0 && optval <= tp->t_maxseg &&
1956 optval + 40 >= V_tcp_minmss)
1957 tp->t_maxseg = optval;
1958 else
1959 error = EINVAL;
1960 goto unlock_and_done;
1961
1962 case TCP_INFO:
1963 INP_WUNLOCK(inp);
1964 error = EINVAL;
1965 break;
1966
1967 case TCP_CONGESTION:
1968 INP_WUNLOCK(inp);
1969 error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1);
1970 if (error)
1971 break;
1972 buf[sopt->sopt_valsize] = '\0';
1973 INP_WLOCK_RECHECK(inp);
1974 CC_LIST_RLOCK();
1975 STAILQ_FOREACH(algo, &cc_list, entries)
1976 if (strncmp(buf, algo->name,
1977 TCP_CA_NAME_MAX) == 0)
1978 break;
1979 CC_LIST_RUNLOCK();
1980 if (algo == NULL) {
1981 INP_WUNLOCK(inp);
1982 error = EINVAL;
1983 break;
1984 }
1985 /*
1986 * We hold a write lock over the tcb so it's safe to
1987 * do these things without ordering concerns.
1988 */
1989 if (CC_ALGO(tp)->cb_destroy != NULL)
1990 CC_ALGO(tp)->cb_destroy(tp->ccv);
1991 CC_DATA(tp) = NULL;
1992 CC_ALGO(tp) = algo;
1993 /*
1994 * If something goes pear shaped initialising the new
1995 * algo, fall back to newreno (which does not
1996 * require initialisation).
1997 */
1998 if (algo->cb_init != NULL &&
1999 algo->cb_init(tp->ccv) != 0) {
2000 CC_ALGO(tp) = &newreno_cc_algo;
2001 /*
2002 * The only reason init should fail is
2003 * because of malloc.
2004 */
2005 error = ENOMEM;
2006 }
2007 INP_WUNLOCK(inp);
2008 break;
2009
2010 case TCP_KEEPIDLE:
2011 case TCP_KEEPINTVL:
2012 case TCP_KEEPINIT:
2013 INP_WUNLOCK(inp);
2014 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2015 if (error)
2016 return (error);
2017
2018 if (ui > (UINT_MAX / hz)) {
2019 error = EINVAL;
2020 break;
2021 }
2022 ui *= hz;
2023
2024 INP_WLOCK_RECHECK(inp);
2025 switch (sopt->sopt_name) {
2026 case TCP_KEEPIDLE:
2027 tp->t_keepidle = ui;
2028 /*
2029 * XXX: better check current remaining
2030 * timeout and "merge" it with new value.
2031 */
2032 if ((tp->t_state > TCPS_LISTEN) &&
2033 (tp->t_state <= TCPS_CLOSING))
2034 tcp_timer_activate(tp, TT_KEEP,
2035 TP_KEEPIDLE(tp));
2036 break;
2037 case TCP_KEEPINTVL:
2038 tp->t_keepintvl = ui;
2039 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
2040 (TP_MAXIDLE(tp) > 0))
2041 tcp_timer_activate(tp, TT_2MSL,
2042 TP_MAXIDLE(tp));
2043 break;
2044 case TCP_KEEPINIT:
2045 tp->t_keepinit = ui;
2046 if (tp->t_state == TCPS_SYN_RECEIVED ||
2047 tp->t_state == TCPS_SYN_SENT)
2048 tcp_timer_activate(tp, TT_KEEP,
2049 TP_KEEPINIT(tp));
2050 break;
2051 }
2052 goto unlock_and_done;
2053
2054 case TCP_KEEPCNT:
2055 INP_WUNLOCK(inp);
2056 error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2057 if (error)
2058 return (error);
2059
2060 INP_WLOCK_RECHECK(inp);
2061 tp->t_keepcnt = ui;
2062 if ((tp->t_state == TCPS_FIN_WAIT_2) &&
2063 (TP_MAXIDLE(tp) > 0))
2064 tcp_timer_activate(tp, TT_2MSL,
2065 TP_MAXIDLE(tp));
2066 goto unlock_and_done;
2067
2068 #ifdef TCPPCAP
2069 case TCP_PCAP_OUT:
2070 case TCP_PCAP_IN:
2071 INP_WUNLOCK(inp);
2072 error = sooptcopyin(sopt, &optval, sizeof optval,
2073 sizeof optval);
2074 if (error)
2075 return (error);
2076
2077 INP_WLOCK_RECHECK(inp);
2078 if (optval >= 0)
2079 tcp_pcap_set_sock_max(
2080 (sopt->sopt_name == TCP_PCAP_OUT) ?
2081 &(tp->t_outpkts) : &(tp->t_inpkts),
2082 optval);
2083 else
2084 error = EINVAL;
2085 goto unlock_and_done;
2086 #endif
2087
2088 case TCP_FASTOPEN: {
2089 struct tcp_fastopen tfo_optval;
2090
2091 INP_WUNLOCK(inp);
2092 if (!V_tcp_fastopen_client_enable &&
2093 !V_tcp_fastopen_server_enable)
2094 return (EPERM);
2095
2096 error = sooptcopyin(sopt, &tfo_optval,
2097 sizeof(tfo_optval), sizeof(int));
2098 if (error)
2099 return (error);
2100
2101 INP_WLOCK_RECHECK(inp);
2102 if ((tp->t_state != TCPS_CLOSED) &&
2103 (tp->t_state != TCPS_LISTEN)) {
2104 error = EINVAL;
2105 goto unlock_and_done;
2106 }
2107 if (tfo_optval.enable) {
2108 if (tp->t_state == TCPS_LISTEN) {
2109 if (!V_tcp_fastopen_server_enable) {
2110 error = EPERM;
2111 goto unlock_and_done;
2112 }
2113
2114 if (tp->t_tfo_pending == NULL)
2115 tp->t_tfo_pending =
2116 tcp_fastopen_alloc_counter();
2117 } else {
2118 /*
2119 * If a pre-shared key was provided,
2120 * stash it in the client cookie
2121 * field of the tcpcb for use during
2122 * connect.
2123 */
2124 if (sopt->sopt_valsize ==
2125 sizeof(tfo_optval)) {
2126 memcpy(tp->t_tfo_cookie.client,
2127 tfo_optval.psk,
2128 TCP_FASTOPEN_PSK_LEN);
2129 tp->t_tfo_client_cookie_len =
2130 TCP_FASTOPEN_PSK_LEN;
2131 }
2132 }
2133 tp->t_flags |= TF_FASTOPEN;
2134 } else
2135 tp->t_flags &= ~TF_FASTOPEN;
2136 goto unlock_and_done;
2137 }
2138
2139 #ifdef TCP_BLACKBOX
2140 case TCP_LOG:
2141 INP_WUNLOCK(inp);
2142 error = sooptcopyin(sopt, &optval, sizeof optval,
2143 sizeof optval);
2144 if (error)
2145 return (error);
2146
2147 INP_WLOCK_RECHECK(inp);
2148 error = tcp_log_state_change(tp, optval);
2149 goto unlock_and_done;
2150
2151 case TCP_LOGBUF:
2152 INP_WUNLOCK(inp);
2153 error = EINVAL;
2154 break;
2155
2156 case TCP_LOGID:
2157 INP_WUNLOCK(inp);
2158 error = sooptcopyin(sopt, buf, TCP_LOG_ID_LEN - 1, 0);
2159 if (error)
2160 break;
2161 buf[sopt->sopt_valsize] = '\0';
2162 INP_WLOCK_RECHECK(inp);
2163 error = tcp_log_set_id(tp, buf);
2164 /* tcp_log_set_id() unlocks the INP. */
2165 break;
2166
2167 case TCP_LOGDUMP:
2168 case TCP_LOGDUMPID:
2169 INP_WUNLOCK(inp);
2170 error =
2171 sooptcopyin(sopt, buf, TCP_LOG_REASON_LEN - 1, 0);
2172 if (error)
2173 break;
2174 buf[sopt->sopt_valsize] = '\0';
2175 INP_WLOCK_RECHECK(inp);
2176 if (sopt->sopt_name == TCP_LOGDUMP) {
2177 error = tcp_log_dump_tp_logbuf(tp, buf,
2178 M_WAITOK, true);
2179 INP_WUNLOCK(inp);
2180 } else {
2181 tcp_log_dump_tp_bucket_logbufs(tp, buf);
2182 /*
2183 * tcp_log_dump_tp_bucket_logbufs() drops the
2184 * INP lock.
2185 */
2186 }
2187 break;
2188 #endif
2189
2190 default:
2191 INP_WUNLOCK(inp);
2192 error = ENOPROTOOPT;
2193 break;
2194 }
2195 break;
2196
2197 case SOPT_GET:
2198 tp = intotcpcb(inp);
2199 switch (sopt->sopt_name) {
2200 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
2201 case TCP_MD5SIG:
2202 if (!TCPMD5_ENABLED()) {
2203 INP_WUNLOCK(inp);
2204 return (ENOPROTOOPT);
2205 }
2206 error = TCPMD5_PCBCTL(inp, sopt);
2207 break;
2208 #endif
2209
2210 case TCP_NODELAY:
2211 optval = tp->t_flags & TF_NODELAY;
2212 INP_WUNLOCK(inp);
2213 error = sooptcopyout(sopt, &optval, sizeof optval);
2214 break;
2215 case TCP_MAXSEG:
2216 optval = tp->t_maxseg;
2217 INP_WUNLOCK(inp);
2218 error = sooptcopyout(sopt, &optval, sizeof optval);
2219 break;
2220 case TCP_NOOPT:
2221 optval = tp->t_flags & TF_NOOPT;
2222 INP_WUNLOCK(inp);
2223 error = sooptcopyout(sopt, &optval, sizeof optval);
2224 break;
2225 case TCP_NOPUSH:
2226 optval = tp->t_flags & TF_NOPUSH;
2227 INP_WUNLOCK(inp);
2228 error = sooptcopyout(sopt, &optval, sizeof optval);
2229 break;
2230 case TCP_INFO:
2231 tcp_fill_info(tp, &ti);
2232 INP_WUNLOCK(inp);
2233 error = sooptcopyout(sopt, &ti, sizeof ti);
2234 break;
2235 case TCP_CONGESTION:
2236 len = strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
2237 INP_WUNLOCK(inp);
2238 error = sooptcopyout(sopt, buf, len + 1);
2239 break;
2240 case TCP_KEEPIDLE:
2241 case TCP_KEEPINTVL:
2242 case TCP_KEEPINIT:
2243 case TCP_KEEPCNT:
2244 switch (sopt->sopt_name) {
2245 case TCP_KEEPIDLE:
2246 ui = TP_KEEPIDLE(tp) / hz;
2247 break;
2248 case TCP_KEEPINTVL:
2249 ui = TP_KEEPINTVL(tp) / hz;
2250 break;
2251 case TCP_KEEPINIT:
2252 ui = TP_KEEPINIT(tp) / hz;
2253 break;
2254 case TCP_KEEPCNT:
2255 ui = TP_KEEPCNT(tp);
2256 break;
2257 }
2258 INP_WUNLOCK(inp);
2259 error = sooptcopyout(sopt, &ui, sizeof(ui));
2260 break;
2261 #ifdef TCPPCAP
2262 case TCP_PCAP_OUT:
2263 case TCP_PCAP_IN:
2264 optval = tcp_pcap_get_sock_max(
2265 (sopt->sopt_name == TCP_PCAP_OUT) ?
2266 &(tp->t_outpkts) : &(tp->t_inpkts));
2267 INP_WUNLOCK(inp);
2268 error = sooptcopyout(sopt, &optval, sizeof optval);
2269 break;
2270 #endif
2271 case TCP_FASTOPEN:
2272 optval = tp->t_flags & TF_FASTOPEN;
2273 INP_WUNLOCK(inp);
2274 error = sooptcopyout(sopt, &optval, sizeof optval);
2275 break;
2276 #ifdef TCP_BLACKBOX
2277 case TCP_LOG:
2278 optval = tp->t_logstate;
2279 INP_WUNLOCK(inp);
2280 error = sooptcopyout(sopt, &optval, sizeof(optval));
2281 break;
2282 case TCP_LOGBUF:
2283 /* tcp_log_getlogbuf() does INP_WUNLOCK(inp) */
2284 error = tcp_log_getlogbuf(sopt, tp);
2285 break;
2286 case TCP_LOGID:
2287 len = tcp_log_get_id(tp, buf);
2288 INP_WUNLOCK(inp);
2289 error = sooptcopyout(sopt, buf, len + 1);
2290 break;
2291 case TCP_LOGDUMP:
2292 case TCP_LOGDUMPID:
2293 INP_WUNLOCK(inp);
2294 error = EINVAL;
2295 break;
2296 #endif
2297 default:
2298 INP_WUNLOCK(inp);
2299 error = ENOPROTOOPT;
2300 break;
2301 }
2302 break;
2303 }
2304 return (error);
2305 }
2306 #undef INP_WLOCK_RECHECK
2307 #undef INP_WLOCK_RECHECK_CLEANUP
2308
2309 /*
2310 * Attach TCP protocol to socket, allocating
2311 * internet protocol control block, tcp control block,
2312 * bufer space, and entering LISTEN state if to accept connections.
2313 */
2314 static int
tcp_attach(struct socket * so)2315 tcp_attach(struct socket *so)
2316 {
2317 struct tcpcb *tp;
2318 struct inpcb *inp;
2319 struct epoch_tracker et;
2320 int error;
2321
2322 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
2323 error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
2324 if (error)
2325 return (error);
2326 }
2327 so->so_rcv.sb_flags |= SB_AUTOSIZE;
2328 so->so_snd.sb_flags |= SB_AUTOSIZE;
2329 INP_INFO_RLOCK_ET(&V_tcbinfo, et);
2330 error = in_pcballoc(so, &V_tcbinfo);
2331 if (error) {
2332 INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
2333 return (error);
2334 }
2335 inp = sotoinpcb(so);
2336 #ifdef INET6
2337 if (inp->inp_vflag & INP_IPV6PROTO) {
2338 inp->inp_vflag |= INP_IPV6;
2339 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
2340 inp->inp_vflag |= INP_IPV4;
2341 inp->in6p_hops = -1; /* use kernel default */
2342 }
2343 else
2344 #endif
2345 inp->inp_vflag |= INP_IPV4;
2346 tp = tcp_newtcpcb(inp);
2347 if (tp == NULL) {
2348 in_pcbdetach(inp);
2349 in_pcbfree(inp);
2350 INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
2351 return (ENOBUFS);
2352 }
2353 tp->t_state = TCPS_CLOSED;
2354 INP_WUNLOCK(inp);
2355 INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
2356 TCPSTATES_INC(TCPS_CLOSED);
2357 return (0);
2358 }
2359
2360 /*
2361 * Initiate (or continue) disconnect.
2362 * If embryonic state, just send reset (once).
2363 * If in ``let data drain'' option and linger null, just drop.
2364 * Otherwise (hard), mark socket disconnecting and drop
2365 * current input data; switch states based on user close, and
2366 * send segment to peer (with FIN).
2367 */
2368 static void
tcp_disconnect(struct tcpcb * tp)2369 tcp_disconnect(struct tcpcb *tp)
2370 {
2371 struct inpcb *inp = tp->t_inpcb;
2372 struct socket *so = inp->inp_socket;
2373
2374 INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
2375 INP_WLOCK_ASSERT(inp);
2376
2377 /*
2378 * Neither tcp_close() nor tcp_drop() should return NULL, as the
2379 * socket is still open.
2380 */
2381 if (tp->t_state < TCPS_ESTABLISHED &&
2382 !(tp->t_state > TCPS_LISTEN && IS_FASTOPEN(tp->t_flags))) {
2383 tp = tcp_close(tp);
2384 KASSERT(tp != NULL,
2385 ("tcp_disconnect: tcp_close() returned NULL"));
2386 } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
2387 tp = tcp_drop(tp, 0);
2388 KASSERT(tp != NULL,
2389 ("tcp_disconnect: tcp_drop() returned NULL"));
2390 } else {
2391 soisdisconnecting(so);
2392 sbflush(&so->so_rcv);
2393 tcp_usrclosed(tp);
2394 if (!(inp->inp_flags & INP_DROPPED))
2395 tp->t_fb->tfb_tcp_output(tp);
2396 }
2397 }
2398
2399 /*
2400 * User issued close, and wish to trail through shutdown states:
2401 * if never received SYN, just forget it. If got a SYN from peer,
2402 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
2403 * If already got a FIN from peer, then almost done; go to LAST_ACK
2404 * state. In all other cases, have already sent FIN to peer (e.g.
2405 * after PRU_SHUTDOWN), and just have to play tedious game waiting
2406 * for peer to send FIN or not respond to keep-alives, etc.
2407 * We can let the user exit from the close as soon as the FIN is acked.
2408 */
2409 static void
tcp_usrclosed(struct tcpcb * tp)2410 tcp_usrclosed(struct tcpcb *tp)
2411 {
2412
2413 INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
2414 INP_WLOCK_ASSERT(tp->t_inpcb);
2415
2416 switch (tp->t_state) {
2417 case TCPS_LISTEN:
2418 #ifdef TCP_OFFLOAD
2419 tcp_offload_listen_stop(tp);
2420 #endif
2421 tcp_state_change(tp, TCPS_CLOSED);
2422 /* FALLTHROUGH */
2423 case TCPS_CLOSED:
2424 tp = tcp_close(tp);
2425 /*
2426 * tcp_close() should never return NULL here as the socket is
2427 * still open.
2428 */
2429 KASSERT(tp != NULL,
2430 ("tcp_usrclosed: tcp_close() returned NULL"));
2431 break;
2432
2433 case TCPS_SYN_SENT:
2434 case TCPS_SYN_RECEIVED:
2435 tp->t_flags |= TF_NEEDFIN;
2436 break;
2437
2438 case TCPS_ESTABLISHED:
2439 tcp_state_change(tp, TCPS_FIN_WAIT_1);
2440 break;
2441
2442 case TCPS_CLOSE_WAIT:
2443 tcp_state_change(tp, TCPS_LAST_ACK);
2444 break;
2445 }
2446 if (tp->t_state >= TCPS_FIN_WAIT_2) {
2447 soisdisconnected(tp->t_inpcb->inp_socket);
2448 /* Prevent the connection hanging in FIN_WAIT_2 forever. */
2449 if (tp->t_state == TCPS_FIN_WAIT_2) {
2450 int timeout;
2451
2452 timeout = (tcp_fast_finwait2_recycle) ?
2453 tcp_finwait2_timeout : TP_MAXIDLE(tp);
2454 tcp_timer_activate(tp, TT_2MSL, timeout);
2455 }
2456 }
2457 }
2458
2459 #ifdef DDB
2460 static void
db_print_indent(int indent)2461 db_print_indent(int indent)
2462 {
2463 int i;
2464
2465 for (i = 0; i < indent; i++)
2466 db_printf(" ");
2467 }
2468
2469 static void
db_print_tstate(int t_state)2470 db_print_tstate(int t_state)
2471 {
2472
2473 switch (t_state) {
2474 case TCPS_CLOSED:
2475 db_printf("TCPS_CLOSED");
2476 return;
2477
2478 case TCPS_LISTEN:
2479 db_printf("TCPS_LISTEN");
2480 return;
2481
2482 case TCPS_SYN_SENT:
2483 db_printf("TCPS_SYN_SENT");
2484 return;
2485
2486 case TCPS_SYN_RECEIVED:
2487 db_printf("TCPS_SYN_RECEIVED");
2488 return;
2489
2490 case TCPS_ESTABLISHED:
2491 db_printf("TCPS_ESTABLISHED");
2492 return;
2493
2494 case TCPS_CLOSE_WAIT:
2495 db_printf("TCPS_CLOSE_WAIT");
2496 return;
2497
2498 case TCPS_FIN_WAIT_1:
2499 db_printf("TCPS_FIN_WAIT_1");
2500 return;
2501
2502 case TCPS_CLOSING:
2503 db_printf("TCPS_CLOSING");
2504 return;
2505
2506 case TCPS_LAST_ACK:
2507 db_printf("TCPS_LAST_ACK");
2508 return;
2509
2510 case TCPS_FIN_WAIT_2:
2511 db_printf("TCPS_FIN_WAIT_2");
2512 return;
2513
2514 case TCPS_TIME_WAIT:
2515 db_printf("TCPS_TIME_WAIT");
2516 return;
2517
2518 default:
2519 db_printf("unknown");
2520 return;
2521 }
2522 }
2523
2524 static void
db_print_tflags(u_int t_flags)2525 db_print_tflags(u_int t_flags)
2526 {
2527 int comma;
2528
2529 comma = 0;
2530 if (t_flags & TF_ACKNOW) {
2531 db_printf("%sTF_ACKNOW", comma ? ", " : "");
2532 comma = 1;
2533 }
2534 if (t_flags & TF_DELACK) {
2535 db_printf("%sTF_DELACK", comma ? ", " : "");
2536 comma = 1;
2537 }
2538 if (t_flags & TF_NODELAY) {
2539 db_printf("%sTF_NODELAY", comma ? ", " : "");
2540 comma = 1;
2541 }
2542 if (t_flags & TF_NOOPT) {
2543 db_printf("%sTF_NOOPT", comma ? ", " : "");
2544 comma = 1;
2545 }
2546 if (t_flags & TF_SENTFIN) {
2547 db_printf("%sTF_SENTFIN", comma ? ", " : "");
2548 comma = 1;
2549 }
2550 if (t_flags & TF_REQ_SCALE) {
2551 db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
2552 comma = 1;
2553 }
2554 if (t_flags & TF_RCVD_SCALE) {
2555 db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
2556 comma = 1;
2557 }
2558 if (t_flags & TF_REQ_TSTMP) {
2559 db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
2560 comma = 1;
2561 }
2562 if (t_flags & TF_RCVD_TSTMP) {
2563 db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
2564 comma = 1;
2565 }
2566 if (t_flags & TF_SACK_PERMIT) {
2567 db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
2568 comma = 1;
2569 }
2570 if (t_flags & TF_NEEDSYN) {
2571 db_printf("%sTF_NEEDSYN", comma ? ", " : "");
2572 comma = 1;
2573 }
2574 if (t_flags & TF_NEEDFIN) {
2575 db_printf("%sTF_NEEDFIN", comma ? ", " : "");
2576 comma = 1;
2577 }
2578 if (t_flags & TF_NOPUSH) {
2579 db_printf("%sTF_NOPUSH", comma ? ", " : "");
2580 comma = 1;
2581 }
2582 if (t_flags & TF_MORETOCOME) {
2583 db_printf("%sTF_MORETOCOME", comma ? ", " : "");
2584 comma = 1;
2585 }
2586 if (t_flags & TF_LQ_OVERFLOW) {
2587 db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
2588 comma = 1;
2589 }
2590 if (t_flags & TF_LASTIDLE) {
2591 db_printf("%sTF_LASTIDLE", comma ? ", " : "");
2592 comma = 1;
2593 }
2594 if (t_flags & TF_RXWIN0SENT) {
2595 db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
2596 comma = 1;
2597 }
2598 if (t_flags & TF_FASTRECOVERY) {
2599 db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
2600 comma = 1;
2601 }
2602 if (t_flags & TF_CONGRECOVERY) {
2603 db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
2604 comma = 1;
2605 }
2606 if (t_flags & TF_WASFRECOVERY) {
2607 db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
2608 comma = 1;
2609 }
2610 if (t_flags & TF_SIGNATURE) {
2611 db_printf("%sTF_SIGNATURE", comma ? ", " : "");
2612 comma = 1;
2613 }
2614 if (t_flags & TF_FORCEDATA) {
2615 db_printf("%sTF_FORCEDATA", comma ? ", " : "");
2616 comma = 1;
2617 }
2618 if (t_flags & TF_TSO) {
2619 db_printf("%sTF_TSO", comma ? ", " : "");
2620 comma = 1;
2621 }
2622 if (t_flags & TF_ECN_PERMIT) {
2623 db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
2624 comma = 1;
2625 }
2626 if (t_flags & TF_FASTOPEN) {
2627 db_printf("%sTF_FASTOPEN", comma ? ", " : "");
2628 comma = 1;
2629 }
2630 }
2631
2632 static void
db_print_toobflags(char t_oobflags)2633 db_print_toobflags(char t_oobflags)
2634 {
2635 int comma;
2636
2637 comma = 0;
2638 if (t_oobflags & TCPOOB_HAVEDATA) {
2639 db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
2640 comma = 1;
2641 }
2642 if (t_oobflags & TCPOOB_HADDATA) {
2643 db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
2644 comma = 1;
2645 }
2646 }
2647
2648 static void
db_print_tcpcb(struct tcpcb * tp,const char * name,int indent)2649 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
2650 {
2651
2652 db_print_indent(indent);
2653 db_printf("%s at %p\n", name, tp);
2654
2655 indent += 2;
2656
2657 db_print_indent(indent);
2658 db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n",
2659 TAILQ_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
2660
2661 db_print_indent(indent);
2662 db_printf("tt_rexmt: %p tt_persist: %p tt_keep: %p\n",
2663 &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
2664
2665 db_print_indent(indent);
2666 db_printf("tt_2msl: %p tt_delack: %p t_inpcb: %p\n", &tp->t_timers->tt_2msl,
2667 &tp->t_timers->tt_delack, tp->t_inpcb);
2668
2669 db_print_indent(indent);
2670 db_printf("t_state: %d (", tp->t_state);
2671 db_print_tstate(tp->t_state);
2672 db_printf(")\n");
2673
2674 db_print_indent(indent);
2675 db_printf("t_flags: 0x%x (", tp->t_flags);
2676 db_print_tflags(tp->t_flags);
2677 db_printf(")\n");
2678
2679 db_print_indent(indent);
2680 db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: x0%08x\n",
2681 tp->snd_una, tp->snd_max, tp->snd_nxt);
2682
2683 db_print_indent(indent);
2684 db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n",
2685 tp->snd_up, tp->snd_wl1, tp->snd_wl2);
2686
2687 db_print_indent(indent);
2688 db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n",
2689 tp->iss, tp->irs, tp->rcv_nxt);
2690
2691 db_print_indent(indent);
2692 db_printf("rcv_adv: 0x%08x rcv_wnd: %u rcv_up: 0x%08x\n",
2693 tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
2694
2695 db_print_indent(indent);
2696 db_printf("snd_wnd: %u snd_cwnd: %u\n",
2697 tp->snd_wnd, tp->snd_cwnd);
2698
2699 db_print_indent(indent);
2700 db_printf("snd_ssthresh: %u snd_recover: "
2701 "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
2702
2703 db_print_indent(indent);
2704 db_printf("t_rcvtime: %u t_startime: %u\n",
2705 tp->t_rcvtime, tp->t_starttime);
2706
2707 db_print_indent(indent);
2708 db_printf("t_rttime: %u t_rtsq: 0x%08x\n",
2709 tp->t_rtttime, tp->t_rtseq);
2710
2711 db_print_indent(indent);
2712 db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n",
2713 tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
2714
2715 db_print_indent(indent);
2716 db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u "
2717 "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
2718 tp->t_rttbest);
2719
2720 db_print_indent(indent);
2721 db_printf("t_rttupdated: %lu max_sndwnd: %u t_softerror: %d\n",
2722 tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
2723
2724 db_print_indent(indent);
2725 db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
2726 db_print_toobflags(tp->t_oobflags);
2727 db_printf(") t_iobc: 0x%02x\n", tp->t_iobc);
2728
2729 db_print_indent(indent);
2730 db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n",
2731 tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
2732
2733 db_print_indent(indent);
2734 db_printf("ts_recent: %u ts_recent_age: %u\n",
2735 tp->ts_recent, tp->ts_recent_age);
2736
2737 db_print_indent(indent);
2738 db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: "
2739 "%u\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
2740
2741 db_print_indent(indent);
2742 db_printf("snd_ssthresh_prev: %u snd_recover_prev: 0x%08x "
2743 "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
2744 tp->snd_recover_prev, tp->t_badrxtwin);
2745
2746 db_print_indent(indent);
2747 db_printf("snd_numholes: %d snd_holes first: %p\n",
2748 tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
2749
2750 db_print_indent(indent);
2751 db_printf("snd_fack: 0x%08x rcv_numsacks: %d sack_newdata: "
2752 "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
2753
2754 /* Skip sackblks, sackhint. */
2755
2756 db_print_indent(indent);
2757 db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n",
2758 tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
2759 }
2760
DB_SHOW_COMMAND(tcpcb,db_show_tcpcb)2761 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
2762 {
2763 struct tcpcb *tp;
2764
2765 if (!have_addr) {
2766 db_printf("usage: show tcpcb <addr>\n");
2767 return;
2768 }
2769 tp = (struct tcpcb *)addr;
2770
2771 db_print_tcpcb(tp, "tcpcb", 0);
2772 }
2773 #endif
2774