xref: /freebsd/sys/netinet/tcp_usrreq.c (revision cc426dd31990b8b50b210efc450e404596548ca1)
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("$FreeBSD$");
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 attaches to socket via pru_attach(), reserving space,
138  * and an internet control block.
139  */
140 static int
141 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
142 {
143 	struct inpcb *inp;
144 	struct tcpcb *tp = NULL;
145 	int error;
146 	TCPDEBUG0;
147 
148 	inp = sotoinpcb(so);
149 	KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
150 	TCPDEBUG1();
151 
152 	error = tcp_attach(so);
153 	if (error)
154 		goto out;
155 
156 	if ((so->so_options & SO_LINGER) && so->so_linger == 0)
157 		so->so_linger = TCP_LINGERTIME;
158 
159 	inp = sotoinpcb(so);
160 	tp = intotcpcb(inp);
161 out:
162 	TCPDEBUG2(PRU_ATTACH);
163 	TCP_PROBE2(debug__user, tp, PRU_ATTACH);
164 	return error;
165 }
166 
167 /*
168  * tcp_detach is called when the socket layer loses its final reference
169  * to the socket, be it a file descriptor reference, a reference from TCP,
170  * etc.  At this point, there is only one case in which we will keep around
171  * inpcb state: time wait.
172  *
173  * This function can probably be re-absorbed back into tcp_usr_detach() now
174  * that there is a single detach path.
175  */
176 static void
177 tcp_detach(struct socket *so, struct inpcb *inp)
178 {
179 	struct tcpcb *tp;
180 
181 	INP_INFO_LOCK_ASSERT(&V_tcbinfo);
182 	INP_WLOCK_ASSERT(inp);
183 
184 	KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
185 	KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
186 
187 	tp = intotcpcb(inp);
188 
189 	if (inp->inp_flags & INP_TIMEWAIT) {
190 		/*
191 		 * There are two cases to handle: one in which the time wait
192 		 * state is being discarded (INP_DROPPED), and one in which
193 		 * this connection will remain in timewait.  In the former,
194 		 * it is time to discard all state (except tcptw, which has
195 		 * already been discarded by the timewait close code, which
196 		 * should be further up the call stack somewhere).  In the
197 		 * latter case, we detach from the socket, but leave the pcb
198 		 * present until timewait ends.
199 		 *
200 		 * XXXRW: Would it be cleaner to free the tcptw here?
201 		 *
202 		 * Astute question indeed, from twtcp perspective there are
203 		 * four cases to consider:
204 		 *
205 		 * #1 tcp_detach is called at tcptw creation time by
206 		 *  tcp_twstart, then do not discard the newly created tcptw
207 		 *  and leave inpcb present until timewait ends
208 		 * #2 tcp_detach is called at tcptw creation time by
209 		 *  tcp_twstart, but connection is local and tw will be
210 		 *  discarded immediately
211 		 * #3 tcp_detach is called at timewait end (or reuse) by
212 		 *  tcp_twclose, then the tcptw has already been discarded
213 		 *  (or reused) and inpcb is freed here
214 		 * #4 tcp_detach is called() after timewait ends (or reuse)
215 		 *  (e.g. by soclose), then tcptw has already been discarded
216 		 *  (or reused) and inpcb is freed here
217 		 *
218 		 *  In all three cases the tcptw should not be freed here.
219 		 */
220 		if (inp->inp_flags & INP_DROPPED) {
221 			in_pcbdetach(inp);
222 			if (__predict_true(tp == NULL)) {
223 				in_pcbfree(inp);
224 			} else {
225 				/*
226 				 * This case should not happen as in TIMEWAIT
227 				 * state the inp should not be destroyed before
228 				 * its tcptw.  If INVARIANTS is defined, panic.
229 				 */
230 #ifdef INVARIANTS
231 				panic("%s: Panic before an inp double-free: "
232 				    "INP_TIMEWAIT && INP_DROPPED && tp != NULL"
233 				    , __func__);
234 #else
235 				log(LOG_ERR, "%s: Avoid an inp double-free: "
236 				    "INP_TIMEWAIT && INP_DROPPED && tp != NULL"
237 				    , __func__);
238 #endif
239 				INP_WUNLOCK(inp);
240 			}
241 		} else {
242 			in_pcbdetach(inp);
243 			INP_WUNLOCK(inp);
244 		}
245 	} else {
246 		/*
247 		 * If the connection is not in timewait, we consider two
248 		 * two conditions: one in which no further processing is
249 		 * necessary (dropped || embryonic), and one in which TCP is
250 		 * not yet done, but no longer requires the socket, so the
251 		 * pcb will persist for the time being.
252 		 *
253 		 * XXXRW: Does the second case still occur?
254 		 */
255 		if (inp->inp_flags & INP_DROPPED ||
256 		    tp->t_state < TCPS_SYN_SENT) {
257 			tcp_discardcb(tp);
258 			in_pcbdetach(inp);
259 			in_pcbfree(inp);
260 		} else {
261 			in_pcbdetach(inp);
262 			INP_WUNLOCK(inp);
263 		}
264 	}
265 }
266 
267 /*
268  * pru_detach() detaches the TCP protocol from the socket.
269  * If the protocol state is non-embryonic, then can't
270  * do this directly: have to initiate a pru_disconnect(),
271  * which may finish later; embryonic TCB's can just
272  * be discarded here.
273  */
274 static void
275 tcp_usr_detach(struct socket *so)
276 {
277 	struct inpcb *inp;
278 	int rlock = 0;
279 	struct epoch_tracker et;
280 
281 	inp = sotoinpcb(so);
282 	KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
283 	if (!INP_INFO_WLOCKED(&V_tcbinfo)) {
284 		INP_INFO_RLOCK_ET(&V_tcbinfo, et);
285 		rlock = 1;
286 	}
287 	INP_WLOCK(inp);
288 	KASSERT(inp->inp_socket != NULL,
289 	    ("tcp_usr_detach: inp_socket == NULL"));
290 	tcp_detach(so, inp);
291 	if (rlock)
292 		INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
293 }
294 
295 #ifdef INET
296 /*
297  * Give the socket an address.
298  */
299 static int
300 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
301 {
302 	int error = 0;
303 	struct inpcb *inp;
304 	struct tcpcb *tp = NULL;
305 	struct sockaddr_in *sinp;
306 
307 	sinp = (struct sockaddr_in *)nam;
308 	if (nam->sa_len != sizeof (*sinp))
309 		return (EINVAL);
310 	/*
311 	 * Must check for multicast addresses and disallow binding
312 	 * to them.
313 	 */
314 	if (sinp->sin_family == AF_INET &&
315 	    IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
316 		return (EAFNOSUPPORT);
317 
318 	TCPDEBUG0;
319 	inp = sotoinpcb(so);
320 	KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
321 	INP_WLOCK(inp);
322 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
323 		error = EINVAL;
324 		goto out;
325 	}
326 	tp = intotcpcb(inp);
327 	TCPDEBUG1();
328 	INP_HASH_WLOCK(&V_tcbinfo);
329 	error = in_pcbbind(inp, nam, td->td_ucred);
330 	INP_HASH_WUNLOCK(&V_tcbinfo);
331 out:
332 	TCPDEBUG2(PRU_BIND);
333 	TCP_PROBE2(debug__user, tp, PRU_BIND);
334 	INP_WUNLOCK(inp);
335 
336 	return (error);
337 }
338 #endif /* INET */
339 
340 #ifdef INET6
341 static int
342 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
343 {
344 	int error = 0;
345 	struct inpcb *inp;
346 	struct tcpcb *tp = NULL;
347 	struct sockaddr_in6 *sin6p;
348 
349 	sin6p = (struct sockaddr_in6 *)nam;
350 	if (nam->sa_len != sizeof (*sin6p))
351 		return (EINVAL);
352 	/*
353 	 * Must check for multicast addresses and disallow binding
354 	 * to them.
355 	 */
356 	if (sin6p->sin6_family == AF_INET6 &&
357 	    IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
358 		return (EAFNOSUPPORT);
359 
360 	TCPDEBUG0;
361 	inp = sotoinpcb(so);
362 	KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
363 	INP_WLOCK(inp);
364 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
365 		error = EINVAL;
366 		goto out;
367 	}
368 	tp = intotcpcb(inp);
369 	TCPDEBUG1();
370 	INP_HASH_WLOCK(&V_tcbinfo);
371 	inp->inp_vflag &= ~INP_IPV4;
372 	inp->inp_vflag |= INP_IPV6;
373 #ifdef INET
374 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
375 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
376 			inp->inp_vflag |= INP_IPV4;
377 		else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
378 			struct sockaddr_in sin;
379 
380 			in6_sin6_2_sin(&sin, sin6p);
381 			if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
382 				error = EAFNOSUPPORT;
383 				INP_HASH_WUNLOCK(&V_tcbinfo);
384 				goto out;
385 			}
386 			inp->inp_vflag |= INP_IPV4;
387 			inp->inp_vflag &= ~INP_IPV6;
388 			error = in_pcbbind(inp, (struct sockaddr *)&sin,
389 			    td->td_ucred);
390 			INP_HASH_WUNLOCK(&V_tcbinfo);
391 			goto out;
392 		}
393 	}
394 #endif
395 	error = in6_pcbbind(inp, nam, td->td_ucred);
396 	INP_HASH_WUNLOCK(&V_tcbinfo);
397 out:
398 	TCPDEBUG2(PRU_BIND);
399 	TCP_PROBE2(debug__user, tp, PRU_BIND);
400 	INP_WUNLOCK(inp);
401 	return (error);
402 }
403 #endif /* INET6 */
404 
405 #ifdef INET
406 /*
407  * Prepare to accept connections.
408  */
409 static int
410 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
411 {
412 	int error = 0;
413 	struct inpcb *inp;
414 	struct tcpcb *tp = NULL;
415 
416 	TCPDEBUG0;
417 	inp = sotoinpcb(so);
418 	KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
419 	INP_WLOCK(inp);
420 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
421 		error = EINVAL;
422 		goto out;
423 	}
424 	tp = intotcpcb(inp);
425 	TCPDEBUG1();
426 	SOCK_LOCK(so);
427 	error = solisten_proto_check(so);
428 	INP_HASH_WLOCK(&V_tcbinfo);
429 	if (error == 0 && inp->inp_lport == 0)
430 		error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
431 	INP_HASH_WUNLOCK(&V_tcbinfo);
432 	if (error == 0) {
433 		tcp_state_change(tp, TCPS_LISTEN);
434 		solisten_proto(so, backlog);
435 #ifdef TCP_OFFLOAD
436 		if ((so->so_options & SO_NO_OFFLOAD) == 0)
437 			tcp_offload_listen_start(tp);
438 #endif
439 	}
440 	SOCK_UNLOCK(so);
441 
442 	if (IS_FASTOPEN(tp->t_flags))
443 		tp->t_tfo_pending = tcp_fastopen_alloc_counter();
444 
445 out:
446 	TCPDEBUG2(PRU_LISTEN);
447 	TCP_PROBE2(debug__user, tp, PRU_LISTEN);
448 	INP_WUNLOCK(inp);
449 	return (error);
450 }
451 #endif /* INET */
452 
453 #ifdef INET6
454 static int
455 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
456 {
457 	int error = 0;
458 	struct inpcb *inp;
459 	struct tcpcb *tp = NULL;
460 
461 	TCPDEBUG0;
462 	inp = sotoinpcb(so);
463 	KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
464 	INP_WLOCK(inp);
465 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
466 		error = EINVAL;
467 		goto out;
468 	}
469 	tp = intotcpcb(inp);
470 	TCPDEBUG1();
471 	SOCK_LOCK(so);
472 	error = solisten_proto_check(so);
473 	INP_HASH_WLOCK(&V_tcbinfo);
474 	if (error == 0 && inp->inp_lport == 0) {
475 		inp->inp_vflag &= ~INP_IPV4;
476 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
477 			inp->inp_vflag |= INP_IPV4;
478 		error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
479 	}
480 	INP_HASH_WUNLOCK(&V_tcbinfo);
481 	if (error == 0) {
482 		tcp_state_change(tp, TCPS_LISTEN);
483 		solisten_proto(so, backlog);
484 #ifdef TCP_OFFLOAD
485 		if ((so->so_options & SO_NO_OFFLOAD) == 0)
486 			tcp_offload_listen_start(tp);
487 #endif
488 	}
489 	SOCK_UNLOCK(so);
490 
491 	if (IS_FASTOPEN(tp->t_flags))
492 		tp->t_tfo_pending = tcp_fastopen_alloc_counter();
493 
494 out:
495 	TCPDEBUG2(PRU_LISTEN);
496 	TCP_PROBE2(debug__user, tp, PRU_LISTEN);
497 	INP_WUNLOCK(inp);
498 	return (error);
499 }
500 #endif /* INET6 */
501 
502 #ifdef INET
503 /*
504  * Initiate connection to peer.
505  * Create a template for use in transmissions on this connection.
506  * Enter SYN_SENT state, and mark socket as connecting.
507  * Start keep-alive timer, and seed output sequence space.
508  * Send initial segment on connection.
509  */
510 static int
511 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
512 {
513 	int error = 0;
514 	struct inpcb *inp;
515 	struct tcpcb *tp = NULL;
516 	struct sockaddr_in *sinp;
517 
518 	sinp = (struct sockaddr_in *)nam;
519 	if (nam->sa_len != sizeof (*sinp))
520 		return (EINVAL);
521 	/*
522 	 * Must disallow TCP ``connections'' to multicast addresses.
523 	 */
524 	if (sinp->sin_family == AF_INET
525 	    && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
526 		return (EAFNOSUPPORT);
527 	if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
528 		return (error);
529 
530 	TCPDEBUG0;
531 	inp = sotoinpcb(so);
532 	KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
533 	INP_WLOCK(inp);
534 	if (inp->inp_flags & INP_TIMEWAIT) {
535 		error = EADDRINUSE;
536 		goto out;
537 	}
538 	if (inp->inp_flags & INP_DROPPED) {
539 		error = ECONNREFUSED;
540 		goto out;
541 	}
542 	tp = intotcpcb(inp);
543 	TCPDEBUG1();
544 	if ((error = tcp_connect(tp, nam, td)) != 0)
545 		goto out;
546 #ifdef TCP_OFFLOAD
547 	if (registered_toedevs > 0 &&
548 	    (so->so_options & SO_NO_OFFLOAD) == 0 &&
549 	    (error = tcp_offload_connect(so, nam)) == 0)
550 		goto out;
551 #endif
552 	tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
553 	error = tp->t_fb->tfb_tcp_output(tp);
554 out:
555 	TCPDEBUG2(PRU_CONNECT);
556 	TCP_PROBE2(debug__user, tp, PRU_CONNECT);
557 	INP_WUNLOCK(inp);
558 	return (error);
559 }
560 #endif /* INET */
561 
562 #ifdef INET6
563 static int
564 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
565 {
566 	int error = 0;
567 	struct inpcb *inp;
568 	struct tcpcb *tp = NULL;
569 	struct sockaddr_in6 *sin6p;
570 
571 	TCPDEBUG0;
572 
573 	sin6p = (struct sockaddr_in6 *)nam;
574 	if (nam->sa_len != sizeof (*sin6p))
575 		return (EINVAL);
576 	/*
577 	 * Must disallow TCP ``connections'' to multicast addresses.
578 	 */
579 	if (sin6p->sin6_family == AF_INET6
580 	    && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
581 		return (EAFNOSUPPORT);
582 
583 	inp = sotoinpcb(so);
584 	KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
585 	INP_WLOCK(inp);
586 	if (inp->inp_flags & INP_TIMEWAIT) {
587 		error = EADDRINUSE;
588 		goto out;
589 	}
590 	if (inp->inp_flags & INP_DROPPED) {
591 		error = ECONNREFUSED;
592 		goto out;
593 	}
594 	tp = intotcpcb(inp);
595 	TCPDEBUG1();
596 #ifdef INET
597 	/*
598 	 * XXXRW: Some confusion: V4/V6 flags relate to binding, and
599 	 * therefore probably require the hash lock, which isn't held here.
600 	 * Is this a significant problem?
601 	 */
602 	if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
603 		struct sockaddr_in sin;
604 
605 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
606 			error = EINVAL;
607 			goto out;
608 		}
609 		if ((inp->inp_vflag & INP_IPV4) == 0) {
610 			error = EAFNOSUPPORT;
611 			goto out;
612 		}
613 
614 		in6_sin6_2_sin(&sin, sin6p);
615 		if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
616 			error = EAFNOSUPPORT;
617 			goto out;
618 		}
619 		inp->inp_vflag |= INP_IPV4;
620 		inp->inp_vflag &= ~INP_IPV6;
621 		if ((error = prison_remote_ip4(td->td_ucred,
622 		    &sin.sin_addr)) != 0)
623 			goto out;
624 		if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
625 			goto out;
626 #ifdef TCP_OFFLOAD
627 		if (registered_toedevs > 0 &&
628 		    (so->so_options & SO_NO_OFFLOAD) == 0 &&
629 		    (error = tcp_offload_connect(so, nam)) == 0)
630 			goto out;
631 #endif
632 		error = tp->t_fb->tfb_tcp_output(tp);
633 		goto out;
634 	} else {
635 		if ((inp->inp_vflag & INP_IPV6) == 0) {
636 			error = EAFNOSUPPORT;
637 			goto out;
638 		}
639 	}
640 #endif
641 	inp->inp_vflag &= ~INP_IPV4;
642 	inp->inp_vflag |= INP_IPV6;
643 	inp->inp_inc.inc_flags |= INC_ISIPV6;
644 	if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
645 		goto out;
646 	if ((error = tcp6_connect(tp, nam, td)) != 0)
647 		goto out;
648 #ifdef TCP_OFFLOAD
649 	if (registered_toedevs > 0 &&
650 	    (so->so_options & SO_NO_OFFLOAD) == 0 &&
651 	    (error = tcp_offload_connect(so, nam)) == 0)
652 		goto out;
653 #endif
654 	tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
655 	error = tp->t_fb->tfb_tcp_output(tp);
656 
657 out:
658 	TCPDEBUG2(PRU_CONNECT);
659 	TCP_PROBE2(debug__user, tp, PRU_CONNECT);
660 	INP_WUNLOCK(inp);
661 	return (error);
662 }
663 #endif /* INET6 */
664 
665 /*
666  * Initiate disconnect from peer.
667  * If connection never passed embryonic stage, just drop;
668  * else if don't need to let data drain, then can just drop anyways,
669  * else have to begin TCP shutdown process: mark socket disconnecting,
670  * drain unread data, state switch to reflect user close, and
671  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
672  * when peer sends FIN and acks ours.
673  *
674  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
675  */
676 static int
677 tcp_usr_disconnect(struct socket *so)
678 {
679 	struct inpcb *inp;
680 	struct tcpcb *tp = NULL;
681 	struct epoch_tracker et;
682 	int error = 0;
683 
684 	TCPDEBUG0;
685 	INP_INFO_RLOCK_ET(&V_tcbinfo, et);
686 	inp = sotoinpcb(so);
687 	KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
688 	INP_WLOCK(inp);
689 	if (inp->inp_flags & INP_TIMEWAIT)
690 		goto out;
691 	if (inp->inp_flags & INP_DROPPED) {
692 		error = ECONNRESET;
693 		goto out;
694 	}
695 	tp = intotcpcb(inp);
696 	TCPDEBUG1();
697 	tcp_disconnect(tp);
698 out:
699 	TCPDEBUG2(PRU_DISCONNECT);
700 	TCP_PROBE2(debug__user, tp, PRU_DISCONNECT);
701 	INP_WUNLOCK(inp);
702 	INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
703 	return (error);
704 }
705 
706 #ifdef INET
707 /*
708  * Accept a connection.  Essentially all the work is done at higher levels;
709  * just return the address of the peer, storing through addr.
710  */
711 static int
712 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
713 {
714 	int error = 0;
715 	struct inpcb *inp = NULL;
716 	struct tcpcb *tp = NULL;
717 	struct in_addr addr;
718 	in_port_t port = 0;
719 	TCPDEBUG0;
720 
721 	if (so->so_state & SS_ISDISCONNECTED)
722 		return (ECONNABORTED);
723 
724 	inp = sotoinpcb(so);
725 	KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
726 	INP_WLOCK(inp);
727 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
728 		error = ECONNABORTED;
729 		goto out;
730 	}
731 	tp = intotcpcb(inp);
732 	TCPDEBUG1();
733 
734 	/*
735 	 * We inline in_getpeeraddr and COMMON_END here, so that we can
736 	 * copy the data of interest and defer the malloc until after we
737 	 * release the lock.
738 	 */
739 	port = inp->inp_fport;
740 	addr = inp->inp_faddr;
741 
742 out:
743 	TCPDEBUG2(PRU_ACCEPT);
744 	TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
745 	INP_WUNLOCK(inp);
746 	if (error == 0)
747 		*nam = in_sockaddr(port, &addr);
748 	return error;
749 }
750 #endif /* INET */
751 
752 #ifdef INET6
753 static int
754 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
755 {
756 	struct inpcb *inp = NULL;
757 	int error = 0;
758 	struct tcpcb *tp = NULL;
759 	struct in_addr addr;
760 	struct in6_addr addr6;
761 	struct epoch_tracker et;
762 	in_port_t port = 0;
763 	int v4 = 0;
764 	TCPDEBUG0;
765 
766 	if (so->so_state & SS_ISDISCONNECTED)
767 		return (ECONNABORTED);
768 
769 	inp = sotoinpcb(so);
770 	KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
771 	INP_INFO_RLOCK_ET(&V_tcbinfo, et);
772 	INP_WLOCK(inp);
773 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
774 		error = ECONNABORTED;
775 		goto out;
776 	}
777 	tp = intotcpcb(inp);
778 	TCPDEBUG1();
779 
780 	/*
781 	 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
782 	 * copy the data of interest and defer the malloc until after we
783 	 * release the lock.
784 	 */
785 	if (inp->inp_vflag & INP_IPV4) {
786 		v4 = 1;
787 		port = inp->inp_fport;
788 		addr = inp->inp_faddr;
789 	} else {
790 		port = inp->inp_fport;
791 		addr6 = inp->in6p_faddr;
792 	}
793 
794 out:
795 	TCPDEBUG2(PRU_ACCEPT);
796 	TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
797 	INP_WUNLOCK(inp);
798 	INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
799 	if (error == 0) {
800 		if (v4)
801 			*nam = in6_v4mapsin6_sockaddr(port, &addr);
802 		else
803 			*nam = in6_sockaddr(port, &addr6);
804 	}
805 	return error;
806 }
807 #endif /* INET6 */
808 
809 /*
810  * Mark the connection as being incapable of further output.
811  */
812 static int
813 tcp_usr_shutdown(struct socket *so)
814 {
815 	int error = 0;
816 	struct inpcb *inp;
817 	struct tcpcb *tp = NULL;
818 	struct epoch_tracker et;
819 
820 	TCPDEBUG0;
821 	INP_INFO_RLOCK_ET(&V_tcbinfo, et);
822 	inp = sotoinpcb(so);
823 	KASSERT(inp != NULL, ("inp == NULL"));
824 	INP_WLOCK(inp);
825 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
826 		error = ECONNRESET;
827 		goto out;
828 	}
829 	tp = intotcpcb(inp);
830 	TCPDEBUG1();
831 	socantsendmore(so);
832 	tcp_usrclosed(tp);
833 	if (!(inp->inp_flags & INP_DROPPED))
834 		error = tp->t_fb->tfb_tcp_output(tp);
835 
836 out:
837 	TCPDEBUG2(PRU_SHUTDOWN);
838 	TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN);
839 	INP_WUNLOCK(inp);
840 	INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
841 
842 	return (error);
843 }
844 
845 /*
846  * After a receive, possibly send window update to peer.
847  */
848 static int
849 tcp_usr_rcvd(struct socket *so, int flags)
850 {
851 	struct inpcb *inp;
852 	struct tcpcb *tp = NULL;
853 	int error = 0;
854 
855 	TCPDEBUG0;
856 	inp = sotoinpcb(so);
857 	KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
858 	INP_WLOCK(inp);
859 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
860 		error = ECONNRESET;
861 		goto out;
862 	}
863 	tp = intotcpcb(inp);
864 	TCPDEBUG1();
865 	/*
866 	 * For passively-created TFO connections, don't attempt a window
867 	 * update while still in SYN_RECEIVED as this may trigger an early
868 	 * SYN|ACK.  It is preferable to have the SYN|ACK be sent along with
869 	 * application response data, or failing that, when the DELACK timer
870 	 * expires.
871 	 */
872 	if (IS_FASTOPEN(tp->t_flags) &&
873 	    (tp->t_state == TCPS_SYN_RECEIVED))
874 		goto out;
875 #ifdef TCP_OFFLOAD
876 	if (tp->t_flags & TF_TOE)
877 		tcp_offload_rcvd(tp);
878 	else
879 #endif
880 	tp->t_fb->tfb_tcp_output(tp);
881 
882 out:
883 	TCPDEBUG2(PRU_RCVD);
884 	TCP_PROBE2(debug__user, tp, PRU_RCVD);
885 	INP_WUNLOCK(inp);
886 	return (error);
887 }
888 
889 /*
890  * Do a send by putting data in output queue and updating urgent
891  * marker if URG set.  Possibly send more data.  Unlike the other
892  * pru_*() routines, the mbuf chains are our responsibility.  We
893  * must either enqueue them or free them.  The other pru_* routines
894  * generally are caller-frees.
895  */
896 static int
897 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
898     struct sockaddr *nam, struct mbuf *control, struct thread *td)
899 {
900 	int error = 0;
901 	struct inpcb *inp;
902 	struct tcpcb *tp = NULL;
903 	struct epoch_tracker net_et;
904 #ifdef INET
905 #ifdef INET6
906 	struct sockaddr_in sin;
907 #endif
908 	struct sockaddr_in *sinp;
909 #endif
910 #ifdef INET6
911 	int isipv6;
912 #endif
913 	TCPDEBUG0;
914 
915 	/*
916 	 * We require the pcbinfo lock if we will close the socket as part of
917 	 * this call.
918 	 */
919 	if (flags & PRUS_EOF)
920 		INP_INFO_RLOCK_ET(&V_tcbinfo, net_et);
921 	inp = sotoinpcb(so);
922 	KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
923 	INP_WLOCK(inp);
924 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
925 		if (control)
926 			m_freem(control);
927 		/*
928 		 * In case of PRUS_NOTREADY, tcp_usr_ready() is responsible
929 		 * for freeing memory.
930 		 */
931 		if (m && (flags & PRUS_NOTREADY) == 0)
932 			m_freem(m);
933 		error = ECONNRESET;
934 		goto out;
935 	}
936 	tp = intotcpcb(inp);
937 	TCPDEBUG1();
938 	if (nam != NULL && tp->t_state < TCPS_SYN_SENT) {
939 		switch (nam->sa_family) {
940 #ifdef INET
941 		case AF_INET:
942 			sinp = (struct sockaddr_in *)nam;
943 			if (sinp->sin_len != sizeof(struct sockaddr_in)) {
944 				if (m)
945 					m_freem(m);
946 				error = EINVAL;
947 				goto out;
948 			}
949 			if ((inp->inp_vflag & INP_IPV6) != 0) {
950 				if (m)
951 					m_freem(m);
952 				error = EAFNOSUPPORT;
953 				goto out;
954 			}
955 			if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
956 				if (m)
957 					m_freem(m);
958 				error = EAFNOSUPPORT;
959 				goto out;
960 			}
961 			if ((error = prison_remote_ip4(td->td_ucred,
962 			    &sinp->sin_addr))) {
963 				if (m)
964 					m_freem(m);
965 				goto out;
966 			}
967 #ifdef INET6
968 			isipv6 = 0;
969 #endif
970 			break;
971 #endif /* INET */
972 #ifdef INET6
973 		case AF_INET6:
974 		{
975 			struct sockaddr_in6 *sin6p;
976 
977 			sin6p = (struct sockaddr_in6 *)nam;
978 			if (sin6p->sin6_len != sizeof(struct sockaddr_in6)) {
979 				if (m)
980 					m_freem(m);
981 				error = EINVAL;
982 				goto out;
983 			}
984 			if (IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
985 				if (m)
986 					m_freem(m);
987 				error = EAFNOSUPPORT;
988 				goto out;
989 			}
990 			if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
991 #ifdef INET
992 				if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
993 					error = EINVAL;
994 					if (m)
995 						m_freem(m);
996 					goto out;
997 				}
998 				if ((inp->inp_vflag & INP_IPV4) == 0) {
999 					error = EAFNOSUPPORT;
1000 					if (m)
1001 						m_freem(m);
1002 					goto out;
1003 				}
1004 				inp->inp_vflag &= ~INP_IPV6;
1005 				sinp = &sin;
1006 				in6_sin6_2_sin(sinp, sin6p);
1007 				if (IN_MULTICAST(
1008 				    ntohl(sinp->sin_addr.s_addr))) {
1009 					error = EAFNOSUPPORT;
1010 					if (m)
1011 						m_freem(m);
1012 					goto out;
1013 				}
1014 				if ((error = prison_remote_ip4(td->td_ucred,
1015 				    &sinp->sin_addr))) {
1016 					if (m)
1017 						m_freem(m);
1018 					goto out;
1019 				}
1020 				isipv6 = 0;
1021 #else /* !INET */
1022 				error = EAFNOSUPPORT;
1023 				if (m)
1024 					m_freem(m);
1025 				goto out;
1026 #endif /* INET */
1027 			} else {
1028 				if ((inp->inp_vflag & INP_IPV6) == 0) {
1029 					if (m)
1030 						m_freem(m);
1031 					error = EAFNOSUPPORT;
1032 					goto out;
1033 				}
1034 				inp->inp_vflag &= ~INP_IPV4;
1035 				inp->inp_inc.inc_flags |= INC_ISIPV6;
1036 				if ((error = prison_remote_ip6(td->td_ucred,
1037 				    &sin6p->sin6_addr))) {
1038 					if (m)
1039 						m_freem(m);
1040 					goto out;
1041 				}
1042 				isipv6 = 1;
1043 			}
1044 			break;
1045 		}
1046 #endif /* INET6 */
1047 		default:
1048 			if (m)
1049 				m_freem(m);
1050 			error = EAFNOSUPPORT;
1051 			goto out;
1052 		}
1053 	}
1054 	if (control) {
1055 		/* TCP doesn't do control messages (rights, creds, etc) */
1056 		if (control->m_len) {
1057 			m_freem(control);
1058 			if (m)
1059 				m_freem(m);
1060 			error = EINVAL;
1061 			goto out;
1062 		}
1063 		m_freem(control);	/* empty control, just free it */
1064 	}
1065 	if (!(flags & PRUS_OOB)) {
1066 		sbappendstream(&so->so_snd, m, flags);
1067 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1068 			/*
1069 			 * Do implied connect if not yet connected,
1070 			 * initialize window to default value, and
1071 			 * initialize maxseg using peer's cached MSS.
1072 			 */
1073 #ifdef INET6
1074 			if (isipv6)
1075 				error = tcp6_connect(tp, nam, td);
1076 #endif /* INET6 */
1077 #if defined(INET6) && defined(INET)
1078 			else
1079 #endif
1080 #ifdef INET
1081 				error = tcp_connect(tp,
1082 				    (struct sockaddr *)sinp, td);
1083 #endif
1084 			if (error)
1085 				goto out;
1086 			if (IS_FASTOPEN(tp->t_flags))
1087 				tcp_fastopen_connect(tp);
1088 			else {
1089 				tp->snd_wnd = TTCP_CLIENT_SND_WND;
1090 				tcp_mss(tp, -1);
1091 			}
1092 		}
1093 		if (flags & PRUS_EOF) {
1094 			/*
1095 			 * Close the send side of the connection after
1096 			 * the data is sent.
1097 			 */
1098 			INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
1099 			socantsendmore(so);
1100 			tcp_usrclosed(tp);
1101 		}
1102 		if (!(inp->inp_flags & INP_DROPPED) &&
1103 		    !(flags & PRUS_NOTREADY)) {
1104 			if (flags & PRUS_MORETOCOME)
1105 				tp->t_flags |= TF_MORETOCOME;
1106 			error = tp->t_fb->tfb_tcp_output(tp);
1107 			if (flags & PRUS_MORETOCOME)
1108 				tp->t_flags &= ~TF_MORETOCOME;
1109 		}
1110 	} else {
1111 		/*
1112 		 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
1113 		 */
1114 		SOCKBUF_LOCK(&so->so_snd);
1115 		if (sbspace(&so->so_snd) < -512) {
1116 			SOCKBUF_UNLOCK(&so->so_snd);
1117 			m_freem(m);
1118 			error = ENOBUFS;
1119 			goto out;
1120 		}
1121 		/*
1122 		 * According to RFC961 (Assigned Protocols),
1123 		 * the urgent pointer points to the last octet
1124 		 * of urgent data.  We continue, however,
1125 		 * to consider it to indicate the first octet
1126 		 * of data past the urgent section.
1127 		 * Otherwise, snd_up should be one lower.
1128 		 */
1129 		sbappendstream_locked(&so->so_snd, m, flags);
1130 		SOCKBUF_UNLOCK(&so->so_snd);
1131 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1132 			/*
1133 			 * Do implied connect if not yet connected,
1134 			 * initialize window to default value, and
1135 			 * initialize maxseg using peer's cached MSS.
1136 			 */
1137 
1138 			/*
1139 			 * Not going to contemplate SYN|URG
1140 			 */
1141 			if (IS_FASTOPEN(tp->t_flags))
1142 				tp->t_flags &= ~TF_FASTOPEN;
1143 #ifdef INET6
1144 			if (isipv6)
1145 				error = tcp6_connect(tp, nam, td);
1146 #endif /* INET6 */
1147 #if defined(INET6) && defined(INET)
1148 			else
1149 #endif
1150 #ifdef INET
1151 				error = tcp_connect(tp,
1152 				    (struct sockaddr *)sinp, td);
1153 #endif
1154 			if (error)
1155 				goto out;
1156 			tp->snd_wnd = TTCP_CLIENT_SND_WND;
1157 			tcp_mss(tp, -1);
1158 		}
1159 		tp->snd_up = tp->snd_una + sbavail(&so->so_snd);
1160 		if (!(flags & PRUS_NOTREADY)) {
1161 			tp->t_flags |= TF_FORCEDATA;
1162 			error = tp->t_fb->tfb_tcp_output(tp);
1163 			tp->t_flags &= ~TF_FORCEDATA;
1164 		}
1165 	}
1166 	TCP_LOG_EVENT(tp, NULL,
1167 	    &inp->inp_socket->so_rcv,
1168 	    &inp->inp_socket->so_snd,
1169 	    TCP_LOG_USERSEND, error,
1170 	    0, NULL, false);
1171 out:
1172 	TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
1173 		  ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1174 	TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB :
1175 		   ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1176 	INP_WUNLOCK(inp);
1177 	if (flags & PRUS_EOF)
1178 		INP_INFO_RUNLOCK_ET(&V_tcbinfo, net_et);
1179 	return (error);
1180 }
1181 
1182 static int
1183 tcp_usr_ready(struct socket *so, struct mbuf *m, int count)
1184 {
1185 	struct inpcb *inp;
1186 	struct tcpcb *tp;
1187 	int error;
1188 
1189 	inp = sotoinpcb(so);
1190 	INP_WLOCK(inp);
1191 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1192 		INP_WUNLOCK(inp);
1193 		for (int i = 0; i < count; i++)
1194 			m = m_free(m);
1195 		return (ECONNRESET);
1196 	}
1197 	tp = intotcpcb(inp);
1198 
1199 	SOCKBUF_LOCK(&so->so_snd);
1200 	error = sbready(&so->so_snd, m, count);
1201 	SOCKBUF_UNLOCK(&so->so_snd);
1202 	if (error == 0)
1203 		error = tp->t_fb->tfb_tcp_output(tp);
1204 	INP_WUNLOCK(inp);
1205 
1206 	return (error);
1207 }
1208 
1209 /*
1210  * Abort the TCP.  Drop the connection abruptly.
1211  */
1212 static void
1213 tcp_usr_abort(struct socket *so)
1214 {
1215 	struct inpcb *inp;
1216 	struct tcpcb *tp = NULL;
1217 	struct epoch_tracker et;
1218 	TCPDEBUG0;
1219 
1220 	inp = sotoinpcb(so);
1221 	KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
1222 
1223 	INP_INFO_RLOCK_ET(&V_tcbinfo, et);
1224 	INP_WLOCK(inp);
1225 	KASSERT(inp->inp_socket != NULL,
1226 	    ("tcp_usr_abort: inp_socket == NULL"));
1227 
1228 	/*
1229 	 * If we still have full TCP state, and we're not dropped, drop.
1230 	 */
1231 	if (!(inp->inp_flags & INP_TIMEWAIT) &&
1232 	    !(inp->inp_flags & INP_DROPPED)) {
1233 		tp = intotcpcb(inp);
1234 		TCPDEBUG1();
1235 		tp = tcp_drop(tp, ECONNABORTED);
1236 		if (tp == NULL)
1237 			goto dropped;
1238 		TCPDEBUG2(PRU_ABORT);
1239 		TCP_PROBE2(debug__user, tp, PRU_ABORT);
1240 	}
1241 	if (!(inp->inp_flags & INP_DROPPED)) {
1242 		SOCK_LOCK(so);
1243 		so->so_state |= SS_PROTOREF;
1244 		SOCK_UNLOCK(so);
1245 		inp->inp_flags |= INP_SOCKREF;
1246 	}
1247 	INP_WUNLOCK(inp);
1248 dropped:
1249 	INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
1250 }
1251 
1252 /*
1253  * TCP socket is closed.  Start friendly disconnect.
1254  */
1255 static void
1256 tcp_usr_close(struct socket *so)
1257 {
1258 	struct inpcb *inp;
1259 	struct tcpcb *tp = NULL;
1260 	struct epoch_tracker et;
1261 	TCPDEBUG0;
1262 
1263 	inp = sotoinpcb(so);
1264 	KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
1265 
1266 	INP_INFO_RLOCK_ET(&V_tcbinfo, et);
1267 	INP_WLOCK(inp);
1268 	KASSERT(inp->inp_socket != NULL,
1269 	    ("tcp_usr_close: inp_socket == NULL"));
1270 
1271 	/*
1272 	 * If we still have full TCP state, and we're not dropped, initiate
1273 	 * a disconnect.
1274 	 */
1275 	if (!(inp->inp_flags & INP_TIMEWAIT) &&
1276 	    !(inp->inp_flags & INP_DROPPED)) {
1277 		tp = intotcpcb(inp);
1278 		TCPDEBUG1();
1279 		tcp_disconnect(tp);
1280 		TCPDEBUG2(PRU_CLOSE);
1281 		TCP_PROBE2(debug__user, tp, PRU_CLOSE);
1282 	}
1283 	if (!(inp->inp_flags & INP_DROPPED)) {
1284 		SOCK_LOCK(so);
1285 		so->so_state |= SS_PROTOREF;
1286 		SOCK_UNLOCK(so);
1287 		inp->inp_flags |= INP_SOCKREF;
1288 	}
1289 	INP_WUNLOCK(inp);
1290 	INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
1291 }
1292 
1293 /*
1294  * Receive out-of-band data.
1295  */
1296 static int
1297 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1298 {
1299 	int error = 0;
1300 	struct inpcb *inp;
1301 	struct tcpcb *tp = NULL;
1302 
1303 	TCPDEBUG0;
1304 	inp = sotoinpcb(so);
1305 	KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1306 	INP_WLOCK(inp);
1307 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1308 		error = ECONNRESET;
1309 		goto out;
1310 	}
1311 	tp = intotcpcb(inp);
1312 	TCPDEBUG1();
1313 	if ((so->so_oobmark == 0 &&
1314 	     (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1315 	    so->so_options & SO_OOBINLINE ||
1316 	    tp->t_oobflags & TCPOOB_HADDATA) {
1317 		error = EINVAL;
1318 		goto out;
1319 	}
1320 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1321 		error = EWOULDBLOCK;
1322 		goto out;
1323 	}
1324 	m->m_len = 1;
1325 	*mtod(m, caddr_t) = tp->t_iobc;
1326 	if ((flags & MSG_PEEK) == 0)
1327 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1328 
1329 out:
1330 	TCPDEBUG2(PRU_RCVOOB);
1331 	TCP_PROBE2(debug__user, tp, PRU_RCVOOB);
1332 	INP_WUNLOCK(inp);
1333 	return (error);
1334 }
1335 
1336 #ifdef INET
1337 struct pr_usrreqs tcp_usrreqs = {
1338 	.pru_abort =		tcp_usr_abort,
1339 	.pru_accept =		tcp_usr_accept,
1340 	.pru_attach =		tcp_usr_attach,
1341 	.pru_bind =		tcp_usr_bind,
1342 	.pru_connect =		tcp_usr_connect,
1343 	.pru_control =		in_control,
1344 	.pru_detach =		tcp_usr_detach,
1345 	.pru_disconnect =	tcp_usr_disconnect,
1346 	.pru_listen =		tcp_usr_listen,
1347 	.pru_peeraddr =		in_getpeeraddr,
1348 	.pru_rcvd =		tcp_usr_rcvd,
1349 	.pru_rcvoob =		tcp_usr_rcvoob,
1350 	.pru_send =		tcp_usr_send,
1351 	.pru_ready =		tcp_usr_ready,
1352 	.pru_shutdown =		tcp_usr_shutdown,
1353 	.pru_sockaddr =		in_getsockaddr,
1354 	.pru_sosetlabel =	in_pcbsosetlabel,
1355 	.pru_close =		tcp_usr_close,
1356 };
1357 #endif /* INET */
1358 
1359 #ifdef INET6
1360 struct pr_usrreqs tcp6_usrreqs = {
1361 	.pru_abort =		tcp_usr_abort,
1362 	.pru_accept =		tcp6_usr_accept,
1363 	.pru_attach =		tcp_usr_attach,
1364 	.pru_bind =		tcp6_usr_bind,
1365 	.pru_connect =		tcp6_usr_connect,
1366 	.pru_control =		in6_control,
1367 	.pru_detach =		tcp_usr_detach,
1368 	.pru_disconnect =	tcp_usr_disconnect,
1369 	.pru_listen =		tcp6_usr_listen,
1370 	.pru_peeraddr =		in6_mapped_peeraddr,
1371 	.pru_rcvd =		tcp_usr_rcvd,
1372 	.pru_rcvoob =		tcp_usr_rcvoob,
1373 	.pru_send =		tcp_usr_send,
1374 	.pru_ready =		tcp_usr_ready,
1375 	.pru_shutdown =		tcp_usr_shutdown,
1376 	.pru_sockaddr =		in6_mapped_sockaddr,
1377 	.pru_sosetlabel =	in_pcbsosetlabel,
1378 	.pru_close =		tcp_usr_close,
1379 };
1380 #endif /* INET6 */
1381 
1382 #ifdef INET
1383 /*
1384  * Common subroutine to open a TCP connection to remote host specified
1385  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
1386  * port number if needed.  Call in_pcbconnect_setup to do the routing and
1387  * to choose a local host address (interface).  If there is an existing
1388  * incarnation of the same connection in TIME-WAIT state and if the remote
1389  * host was sending CC options and if the connection duration was < MSL, then
1390  * truncate the previous TIME-WAIT state and proceed.
1391  * Initialize connection parameters and enter SYN-SENT state.
1392  */
1393 static int
1394 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1395 {
1396 	struct inpcb *inp = tp->t_inpcb, *oinp;
1397 	struct socket *so = inp->inp_socket;
1398 	struct in_addr laddr;
1399 	u_short lport;
1400 	int error;
1401 
1402 	INP_WLOCK_ASSERT(inp);
1403 	INP_HASH_WLOCK(&V_tcbinfo);
1404 
1405 	if (inp->inp_lport == 0) {
1406 		error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1407 		if (error)
1408 			goto out;
1409 	}
1410 
1411 	/*
1412 	 * Cannot simply call in_pcbconnect, because there might be an
1413 	 * earlier incarnation of this same connection still in
1414 	 * TIME_WAIT state, creating an ADDRINUSE error.
1415 	 */
1416 	laddr = inp->inp_laddr;
1417 	lport = inp->inp_lport;
1418 	error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1419 	    &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1420 	if (error && oinp == NULL)
1421 		goto out;
1422 	if (oinp) {
1423 		error = EADDRINUSE;
1424 		goto out;
1425 	}
1426 	inp->inp_laddr = laddr;
1427 	in_pcbrehash(inp);
1428 	INP_HASH_WUNLOCK(&V_tcbinfo);
1429 
1430 	/*
1431 	 * Compute window scaling to request:
1432 	 * Scale to fit into sweet spot.  See tcp_syncache.c.
1433 	 * XXX: This should move to tcp_output().
1434 	 */
1435 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1436 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1437 		tp->request_r_scale++;
1438 
1439 	soisconnecting(so);
1440 	TCPSTAT_INC(tcps_connattempt);
1441 	tcp_state_change(tp, TCPS_SYN_SENT);
1442 	tp->iss = tcp_new_isn(&inp->inp_inc);
1443 	if (tp->t_flags & TF_REQ_TSTMP)
1444 		tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
1445 	tcp_sendseqinit(tp);
1446 
1447 	return 0;
1448 
1449 out:
1450 	INP_HASH_WUNLOCK(&V_tcbinfo);
1451 	return (error);
1452 }
1453 #endif /* INET */
1454 
1455 #ifdef INET6
1456 static int
1457 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1458 {
1459 	struct inpcb *inp = tp->t_inpcb;
1460 	int error;
1461 
1462 	INP_WLOCK_ASSERT(inp);
1463 	INP_HASH_WLOCK(&V_tcbinfo);
1464 
1465 	if (inp->inp_lport == 0) {
1466 		error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1467 		if (error)
1468 			goto out;
1469 	}
1470 	error = in6_pcbconnect(inp, nam, td->td_ucred);
1471 	if (error != 0)
1472 		goto out;
1473 	INP_HASH_WUNLOCK(&V_tcbinfo);
1474 
1475 	/* Compute window scaling to request.  */
1476 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1477 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1478 		tp->request_r_scale++;
1479 
1480 	soisconnecting(inp->inp_socket);
1481 	TCPSTAT_INC(tcps_connattempt);
1482 	tcp_state_change(tp, TCPS_SYN_SENT);
1483 	tp->iss = tcp_new_isn(&inp->inp_inc);
1484 	if (tp->t_flags & TF_REQ_TSTMP)
1485 		tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
1486 	tcp_sendseqinit(tp);
1487 
1488 	return 0;
1489 
1490 out:
1491 	INP_HASH_WUNLOCK(&V_tcbinfo);
1492 	return error;
1493 }
1494 #endif /* INET6 */
1495 
1496 /*
1497  * Export TCP internal state information via a struct tcp_info, based on the
1498  * Linux 2.6 API.  Not ABI compatible as our constants are mapped differently
1499  * (TCP state machine, etc).  We export all information using FreeBSD-native
1500  * constants -- for example, the numeric values for tcpi_state will differ
1501  * from Linux.
1502  */
1503 static void
1504 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1505 {
1506 
1507 	INP_WLOCK_ASSERT(tp->t_inpcb);
1508 	bzero(ti, sizeof(*ti));
1509 
1510 	ti->tcpi_state = tp->t_state;
1511 	if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1512 		ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1513 	if (tp->t_flags & TF_SACK_PERMIT)
1514 		ti->tcpi_options |= TCPI_OPT_SACK;
1515 	if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1516 		ti->tcpi_options |= TCPI_OPT_WSCALE;
1517 		ti->tcpi_snd_wscale = tp->snd_scale;
1518 		ti->tcpi_rcv_wscale = tp->rcv_scale;
1519 	}
1520 	if (tp->t_flags & TF_ECN_PERMIT)
1521 		ti->tcpi_options |= TCPI_OPT_ECN;
1522 
1523 	ti->tcpi_rto = tp->t_rxtcur * tick;
1524 	ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick;
1525 	ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1526 	ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1527 
1528 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1529 	ti->tcpi_snd_cwnd = tp->snd_cwnd;
1530 
1531 	/*
1532 	 * FreeBSD-specific extension fields for tcp_info.
1533 	 */
1534 	ti->tcpi_rcv_space = tp->rcv_wnd;
1535 	ti->tcpi_rcv_nxt = tp->rcv_nxt;
1536 	ti->tcpi_snd_wnd = tp->snd_wnd;
1537 	ti->tcpi_snd_bwnd = 0;		/* Unused, kept for compat. */
1538 	ti->tcpi_snd_nxt = tp->snd_nxt;
1539 	ti->tcpi_snd_mss = tp->t_maxseg;
1540 	ti->tcpi_rcv_mss = tp->t_maxseg;
1541 	ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1542 	ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1543 	ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1544 #ifdef TCP_OFFLOAD
1545 	if (tp->t_flags & TF_TOE) {
1546 		ti->tcpi_options |= TCPI_OPT_TOE;
1547 		tcp_offload_tcp_info(tp, ti);
1548 	}
1549 #endif
1550 }
1551 
1552 /*
1553  * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1554  * socket option arguments.  When it re-acquires the lock after the copy, it
1555  * has to revalidate that the connection is still valid for the socket
1556  * option.
1557  */
1558 #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do {			\
1559 	INP_WLOCK(inp);							\
1560 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {		\
1561 		INP_WUNLOCK(inp);					\
1562 		cleanup;						\
1563 		return (ECONNRESET);					\
1564 	}								\
1565 	tp = intotcpcb(inp);						\
1566 } while(0)
1567 #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */)
1568 
1569 int
1570 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1571 {
1572 	int	error;
1573 	struct	inpcb *inp;
1574 	struct	tcpcb *tp;
1575 	struct tcp_function_block *blk;
1576 	struct tcp_function_set fsn;
1577 
1578 	error = 0;
1579 	inp = sotoinpcb(so);
1580 	KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1581 	INP_WLOCK(inp);
1582 	if (sopt->sopt_level != IPPROTO_TCP) {
1583 #ifdef INET6
1584 		if (inp->inp_vflag & INP_IPV6PROTO) {
1585 			INP_WUNLOCK(inp);
1586 			error = ip6_ctloutput(so, sopt);
1587 			/*
1588 			 * In case of the IPV6_USE_MIN_MTU socket option,
1589 			 * the INC_IPV6MINMTU flag to announce a corresponding
1590 			 * MSS during the initial handshake.
1591 			 * If the TCP connection is not in the front states,
1592 			 * just reduce the MSS being used.
1593 			 * This avoids the sending of TCP segments which will
1594 			 * be fragmented at the IPv6 layer.
1595 			 */
1596 			if ((error == 0) &&
1597 			    (sopt->sopt_dir == SOPT_SET) &&
1598 			    (sopt->sopt_level == IPPROTO_IPV6) &&
1599 			    (sopt->sopt_name == IPV6_USE_MIN_MTU)) {
1600 				INP_WLOCK(inp);
1601 				if ((inp->inp_flags &
1602 				    (INP_TIMEWAIT | INP_DROPPED))) {
1603 					INP_WUNLOCK(inp);
1604 					return (ECONNRESET);
1605 				}
1606 				inp->inp_inc.inc_flags |= INC_IPV6MINMTU;
1607 				tp = intotcpcb(inp);
1608 				if ((tp->t_state >= TCPS_SYN_SENT) &&
1609 				    (inp->inp_inc.inc_flags & INC_ISIPV6)) {
1610 					struct ip6_pktopts *opt;
1611 
1612 					opt = inp->in6p_outputopts;
1613 					if ((opt != NULL) &&
1614 					    (opt->ip6po_minmtu ==
1615 					    IP6PO_MINMTU_ALL)) {
1616 						if (tp->t_maxseg > TCP6_MSS) {
1617 							tp->t_maxseg = TCP6_MSS;
1618 						}
1619 					}
1620 				}
1621 				INP_WUNLOCK(inp);
1622 			}
1623 		}
1624 #endif /* INET6 */
1625 #if defined(INET6) && defined(INET)
1626 		else
1627 #endif
1628 #ifdef INET
1629 		{
1630 			INP_WUNLOCK(inp);
1631 			error = ip_ctloutput(so, sopt);
1632 		}
1633 #endif
1634 		return (error);
1635 	}
1636 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1637 		INP_WUNLOCK(inp);
1638 		return (ECONNRESET);
1639 	}
1640 	tp = intotcpcb(inp);
1641 	/*
1642 	 * Protect the TCP option TCP_FUNCTION_BLK so
1643 	 * that a sub-function can *never* overwrite this.
1644 	 */
1645 	if ((sopt->sopt_dir == SOPT_SET) &&
1646 	    (sopt->sopt_name == TCP_FUNCTION_BLK)) {
1647 		INP_WUNLOCK(inp);
1648 		error = sooptcopyin(sopt, &fsn, sizeof fsn,
1649 		    sizeof fsn);
1650 		if (error)
1651 			return (error);
1652 		INP_WLOCK_RECHECK(inp);
1653 		blk = find_and_ref_tcp_functions(&fsn);
1654 		if (blk == NULL) {
1655 			INP_WUNLOCK(inp);
1656 			return (ENOENT);
1657 		}
1658 		if (tp->t_fb == blk) {
1659 			/* You already have this */
1660 			refcount_release(&blk->tfb_refcnt);
1661 			INP_WUNLOCK(inp);
1662 			return (0);
1663 		}
1664 		if (tp->t_state != TCPS_CLOSED) {
1665 			/*
1666 			 * The user has advanced the state
1667 			 * past the initial point, we may not
1668 			 * be able to switch.
1669 			 */
1670 			if (blk->tfb_tcp_handoff_ok != NULL) {
1671 				/*
1672 				 * Does the stack provide a
1673 				 * query mechanism, if so it may
1674 				 * still be possible?
1675 				 */
1676 				error = (*blk->tfb_tcp_handoff_ok)(tp);
1677 			} else
1678 				error = EINVAL;
1679 			if (error) {
1680 				refcount_release(&blk->tfb_refcnt);
1681 				INP_WUNLOCK(inp);
1682 				return(error);
1683 			}
1684 		}
1685 		if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) {
1686 			refcount_release(&blk->tfb_refcnt);
1687 			INP_WUNLOCK(inp);
1688 			return (ENOENT);
1689 		}
1690 		/*
1691 		 * Release the old refcnt, the
1692 		 * lookup acquired a ref on the
1693 		 * new one already.
1694 		 */
1695 		if (tp->t_fb->tfb_tcp_fb_fini) {
1696 			/*
1697 			 * Tell the stack to cleanup with 0 i.e.
1698 			 * the tcb is not going away.
1699 			 */
1700 			(*tp->t_fb->tfb_tcp_fb_fini)(tp, 0);
1701 		}
1702 #ifdef TCPHPTS
1703 		/* Assure that we are not on any hpts */
1704 		tcp_hpts_remove(tp->t_inpcb, HPTS_REMOVE_ALL);
1705 #endif
1706 		if (blk->tfb_tcp_fb_init) {
1707 			error = (*blk->tfb_tcp_fb_init)(tp);
1708 			if (error) {
1709 				refcount_release(&blk->tfb_refcnt);
1710 				if (tp->t_fb->tfb_tcp_fb_init) {
1711 					if((*tp->t_fb->tfb_tcp_fb_init)(tp) != 0)  {
1712 						/* Fall back failed, drop the connection */
1713 						INP_WUNLOCK(inp);
1714 						soabort(so);
1715 						return(error);
1716 					}
1717 				}
1718 				goto err_out;
1719 			}
1720 		}
1721 		refcount_release(&tp->t_fb->tfb_refcnt);
1722 		tp->t_fb = blk;
1723 #ifdef TCP_OFFLOAD
1724 		if (tp->t_flags & TF_TOE) {
1725 			tcp_offload_ctloutput(tp, sopt->sopt_dir,
1726 			     sopt->sopt_name);
1727 		}
1728 #endif
1729 err_out:
1730 		INP_WUNLOCK(inp);
1731 		return (error);
1732 	} else if ((sopt->sopt_dir == SOPT_GET) &&
1733 	    (sopt->sopt_name == TCP_FUNCTION_BLK)) {
1734 		strncpy(fsn.function_set_name, tp->t_fb->tfb_tcp_block_name,
1735 		    TCP_FUNCTION_NAME_LEN_MAX);
1736 		fsn.function_set_name[TCP_FUNCTION_NAME_LEN_MAX - 1] = '\0';
1737 		fsn.pcbcnt = tp->t_fb->tfb_refcnt;
1738 		INP_WUNLOCK(inp);
1739 		error = sooptcopyout(sopt, &fsn, sizeof fsn);
1740 		return (error);
1741 	}
1742 	/* Pass in the INP locked, called must unlock it */
1743 	return (tp->t_fb->tfb_tcp_ctloutput(so, sopt, inp, tp));
1744 }
1745 
1746 /*
1747  * If this assert becomes untrue, we need to change the size of the buf
1748  * variable in tcp_default_ctloutput().
1749  */
1750 #ifdef CTASSERT
1751 CTASSERT(TCP_CA_NAME_MAX <= TCP_LOG_ID_LEN);
1752 CTASSERT(TCP_LOG_REASON_LEN <= TCP_LOG_ID_LEN);
1753 #endif
1754 
1755 int
1756 tcp_default_ctloutput(struct socket *so, struct sockopt *sopt, struct inpcb *inp, struct tcpcb *tp)
1757 {
1758 	int	error, opt, optval;
1759 	u_int	ui;
1760 	struct	tcp_info ti;
1761 	struct cc_algo *algo;
1762 	char	*pbuf, buf[TCP_LOG_ID_LEN];
1763 	size_t	len;
1764 
1765 	/*
1766 	 * For TCP_CCALGOOPT forward the control to CC module, for both
1767 	 * SOPT_SET and SOPT_GET.
1768 	 */
1769 	switch (sopt->sopt_name) {
1770 	case TCP_CCALGOOPT:
1771 		INP_WUNLOCK(inp);
1772 		if (sopt->sopt_valsize > CC_ALGOOPT_LIMIT)
1773 			return (EINVAL);
1774 		pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO);
1775 		error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize,
1776 		    sopt->sopt_valsize);
1777 		if (error) {
1778 			free(pbuf, M_TEMP);
1779 			return (error);
1780 		}
1781 		INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP));
1782 		if (CC_ALGO(tp)->ctl_output != NULL)
1783 			error = CC_ALGO(tp)->ctl_output(tp->ccv, sopt, pbuf);
1784 		else
1785 			error = ENOENT;
1786 		INP_WUNLOCK(inp);
1787 		if (error == 0 && sopt->sopt_dir == SOPT_GET)
1788 			error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize);
1789 		free(pbuf, M_TEMP);
1790 		return (error);
1791 	}
1792 
1793 	switch (sopt->sopt_dir) {
1794 	case SOPT_SET:
1795 		switch (sopt->sopt_name) {
1796 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
1797 		case TCP_MD5SIG:
1798 			if (!TCPMD5_ENABLED()) {
1799 				INP_WUNLOCK(inp);
1800 				return (ENOPROTOOPT);
1801 			}
1802 			error = TCPMD5_PCBCTL(inp, sopt);
1803 			if (error)
1804 				return (error);
1805 			goto unlock_and_done;
1806 #endif /* IPSEC */
1807 
1808 		case TCP_NODELAY:
1809 		case TCP_NOOPT:
1810 			INP_WUNLOCK(inp);
1811 			error = sooptcopyin(sopt, &optval, sizeof optval,
1812 			    sizeof optval);
1813 			if (error)
1814 				return (error);
1815 
1816 			INP_WLOCK_RECHECK(inp);
1817 			switch (sopt->sopt_name) {
1818 			case TCP_NODELAY:
1819 				opt = TF_NODELAY;
1820 				break;
1821 			case TCP_NOOPT:
1822 				opt = TF_NOOPT;
1823 				break;
1824 			default:
1825 				opt = 0; /* dead code to fool gcc */
1826 				break;
1827 			}
1828 
1829 			if (optval)
1830 				tp->t_flags |= opt;
1831 			else
1832 				tp->t_flags &= ~opt;
1833 unlock_and_done:
1834 #ifdef TCP_OFFLOAD
1835 			if (tp->t_flags & TF_TOE) {
1836 				tcp_offload_ctloutput(tp, sopt->sopt_dir,
1837 				    sopt->sopt_name);
1838 			}
1839 #endif
1840 			INP_WUNLOCK(inp);
1841 			break;
1842 
1843 		case TCP_NOPUSH:
1844 			INP_WUNLOCK(inp);
1845 			error = sooptcopyin(sopt, &optval, sizeof optval,
1846 			    sizeof optval);
1847 			if (error)
1848 				return (error);
1849 
1850 			INP_WLOCK_RECHECK(inp);
1851 			if (optval)
1852 				tp->t_flags |= TF_NOPUSH;
1853 			else if (tp->t_flags & TF_NOPUSH) {
1854 				tp->t_flags &= ~TF_NOPUSH;
1855 				if (TCPS_HAVEESTABLISHED(tp->t_state))
1856 					error = tp->t_fb->tfb_tcp_output(tp);
1857 			}
1858 			goto unlock_and_done;
1859 
1860 		case TCP_MAXSEG:
1861 			INP_WUNLOCK(inp);
1862 			error = sooptcopyin(sopt, &optval, sizeof optval,
1863 			    sizeof optval);
1864 			if (error)
1865 				return (error);
1866 
1867 			INP_WLOCK_RECHECK(inp);
1868 			if (optval > 0 && optval <= tp->t_maxseg &&
1869 			    optval + 40 >= V_tcp_minmss)
1870 				tp->t_maxseg = optval;
1871 			else
1872 				error = EINVAL;
1873 			goto unlock_and_done;
1874 
1875 		case TCP_INFO:
1876 			INP_WUNLOCK(inp);
1877 			error = EINVAL;
1878 			break;
1879 
1880 		case TCP_CONGESTION:
1881 			INP_WUNLOCK(inp);
1882 			error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1);
1883 			if (error)
1884 				break;
1885 			buf[sopt->sopt_valsize] = '\0';
1886 			INP_WLOCK_RECHECK(inp);
1887 			CC_LIST_RLOCK();
1888 			STAILQ_FOREACH(algo, &cc_list, entries)
1889 				if (strncmp(buf, algo->name,
1890 				    TCP_CA_NAME_MAX) == 0)
1891 					break;
1892 			CC_LIST_RUNLOCK();
1893 			if (algo == NULL) {
1894 				INP_WUNLOCK(inp);
1895 				error = EINVAL;
1896 				break;
1897 			}
1898 			/*
1899 			 * We hold a write lock over the tcb so it's safe to
1900 			 * do these things without ordering concerns.
1901 			 */
1902 			if (CC_ALGO(tp)->cb_destroy != NULL)
1903 				CC_ALGO(tp)->cb_destroy(tp->ccv);
1904 			CC_DATA(tp) = NULL;
1905 			CC_ALGO(tp) = algo;
1906 			/*
1907 			 * If something goes pear shaped initialising the new
1908 			 * algo, fall back to newreno (which does not
1909 			 * require initialisation).
1910 			 */
1911 			if (algo->cb_init != NULL &&
1912 			    algo->cb_init(tp->ccv) != 0) {
1913 				CC_ALGO(tp) = &newreno_cc_algo;
1914 				/*
1915 				 * The only reason init should fail is
1916 				 * because of malloc.
1917 				 */
1918 				error = ENOMEM;
1919 			}
1920 			INP_WUNLOCK(inp);
1921 			break;
1922 
1923 		case TCP_KEEPIDLE:
1924 		case TCP_KEEPINTVL:
1925 		case TCP_KEEPINIT:
1926 			INP_WUNLOCK(inp);
1927 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1928 			if (error)
1929 				return (error);
1930 
1931 			if (ui > (UINT_MAX / hz)) {
1932 				error = EINVAL;
1933 				break;
1934 			}
1935 			ui *= hz;
1936 
1937 			INP_WLOCK_RECHECK(inp);
1938 			switch (sopt->sopt_name) {
1939 			case TCP_KEEPIDLE:
1940 				tp->t_keepidle = ui;
1941 				/*
1942 				 * XXX: better check current remaining
1943 				 * timeout and "merge" it with new value.
1944 				 */
1945 				if ((tp->t_state > TCPS_LISTEN) &&
1946 				    (tp->t_state <= TCPS_CLOSING))
1947 					tcp_timer_activate(tp, TT_KEEP,
1948 					    TP_KEEPIDLE(tp));
1949 				break;
1950 			case TCP_KEEPINTVL:
1951 				tp->t_keepintvl = ui;
1952 				if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1953 				    (TP_MAXIDLE(tp) > 0))
1954 					tcp_timer_activate(tp, TT_2MSL,
1955 					    TP_MAXIDLE(tp));
1956 				break;
1957 			case TCP_KEEPINIT:
1958 				tp->t_keepinit = ui;
1959 				if (tp->t_state == TCPS_SYN_RECEIVED ||
1960 				    tp->t_state == TCPS_SYN_SENT)
1961 					tcp_timer_activate(tp, TT_KEEP,
1962 					    TP_KEEPINIT(tp));
1963 				break;
1964 			}
1965 			goto unlock_and_done;
1966 
1967 		case TCP_KEEPCNT:
1968 			INP_WUNLOCK(inp);
1969 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1970 			if (error)
1971 				return (error);
1972 
1973 			INP_WLOCK_RECHECK(inp);
1974 			tp->t_keepcnt = ui;
1975 			if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1976 			    (TP_MAXIDLE(tp) > 0))
1977 				tcp_timer_activate(tp, TT_2MSL,
1978 				    TP_MAXIDLE(tp));
1979 			goto unlock_and_done;
1980 
1981 #ifdef TCPPCAP
1982 		case TCP_PCAP_OUT:
1983 		case TCP_PCAP_IN:
1984 			INP_WUNLOCK(inp);
1985 			error = sooptcopyin(sopt, &optval, sizeof optval,
1986 			    sizeof optval);
1987 			if (error)
1988 				return (error);
1989 
1990 			INP_WLOCK_RECHECK(inp);
1991 			if (optval >= 0)
1992 				tcp_pcap_set_sock_max(TCP_PCAP_OUT ?
1993 					&(tp->t_outpkts) : &(tp->t_inpkts),
1994 					optval);
1995 			else
1996 				error = EINVAL;
1997 			goto unlock_and_done;
1998 #endif
1999 
2000 		case TCP_FASTOPEN: {
2001 			struct tcp_fastopen tfo_optval;
2002 
2003 			INP_WUNLOCK(inp);
2004 			if (!V_tcp_fastopen_client_enable &&
2005 			    !V_tcp_fastopen_server_enable)
2006 				return (EPERM);
2007 
2008 			error = sooptcopyin(sopt, &tfo_optval,
2009 				    sizeof(tfo_optval), sizeof(int));
2010 			if (error)
2011 				return (error);
2012 
2013 			INP_WLOCK_RECHECK(inp);
2014 			if (tfo_optval.enable) {
2015 				if (tp->t_state == TCPS_LISTEN) {
2016 					if (!V_tcp_fastopen_server_enable) {
2017 						error = EPERM;
2018 						goto unlock_and_done;
2019 					}
2020 
2021 					tp->t_flags |= TF_FASTOPEN;
2022 					if (tp->t_tfo_pending == NULL)
2023 						tp->t_tfo_pending =
2024 						    tcp_fastopen_alloc_counter();
2025 				} else {
2026 					/*
2027 					 * If a pre-shared key was provided,
2028 					 * stash it in the client cookie
2029 					 * field of the tcpcb for use during
2030 					 * connect.
2031 					 */
2032 					if (sopt->sopt_valsize ==
2033 					    sizeof(tfo_optval)) {
2034 						memcpy(tp->t_tfo_cookie.client,
2035 						       tfo_optval.psk,
2036 						       TCP_FASTOPEN_PSK_LEN);
2037 						tp->t_tfo_client_cookie_len =
2038 						    TCP_FASTOPEN_PSK_LEN;
2039 					}
2040 					tp->t_flags |= TF_FASTOPEN;
2041 				}
2042 			} else
2043 				tp->t_flags &= ~TF_FASTOPEN;
2044 			goto unlock_and_done;
2045 		}
2046 
2047 #ifdef TCP_BLACKBOX
2048 		case TCP_LOG:
2049 			INP_WUNLOCK(inp);
2050 			error = sooptcopyin(sopt, &optval, sizeof optval,
2051 			    sizeof optval);
2052 			if (error)
2053 				return (error);
2054 
2055 			INP_WLOCK_RECHECK(inp);
2056 			error = tcp_log_state_change(tp, optval);
2057 			goto unlock_and_done;
2058 
2059 		case TCP_LOGBUF:
2060 			INP_WUNLOCK(inp);
2061 			error = EINVAL;
2062 			break;
2063 
2064 		case TCP_LOGID:
2065 			INP_WUNLOCK(inp);
2066 			error = sooptcopyin(sopt, buf, TCP_LOG_ID_LEN - 1, 0);
2067 			if (error)
2068 				break;
2069 			buf[sopt->sopt_valsize] = '\0';
2070 			INP_WLOCK_RECHECK(inp);
2071 			error = tcp_log_set_id(tp, buf);
2072 			/* tcp_log_set_id() unlocks the INP. */
2073 			break;
2074 
2075 		case TCP_LOGDUMP:
2076 		case TCP_LOGDUMPID:
2077 			INP_WUNLOCK(inp);
2078 			error =
2079 			    sooptcopyin(sopt, buf, TCP_LOG_REASON_LEN - 1, 0);
2080 			if (error)
2081 				break;
2082 			buf[sopt->sopt_valsize] = '\0';
2083 			INP_WLOCK_RECHECK(inp);
2084 			if (sopt->sopt_name == TCP_LOGDUMP) {
2085 				error = tcp_log_dump_tp_logbuf(tp, buf,
2086 				    M_WAITOK, true);
2087 				INP_WUNLOCK(inp);
2088 			} else {
2089 				tcp_log_dump_tp_bucket_logbufs(tp, buf);
2090 				/*
2091 				 * tcp_log_dump_tp_bucket_logbufs() drops the
2092 				 * INP lock.
2093 				 */
2094 			}
2095 			break;
2096 #endif
2097 
2098 		default:
2099 			INP_WUNLOCK(inp);
2100 			error = ENOPROTOOPT;
2101 			break;
2102 		}
2103 		break;
2104 
2105 	case SOPT_GET:
2106 		tp = intotcpcb(inp);
2107 		switch (sopt->sopt_name) {
2108 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
2109 		case TCP_MD5SIG:
2110 			if (!TCPMD5_ENABLED()) {
2111 				INP_WUNLOCK(inp);
2112 				return (ENOPROTOOPT);
2113 			}
2114 			error = TCPMD5_PCBCTL(inp, sopt);
2115 			break;
2116 #endif
2117 
2118 		case TCP_NODELAY:
2119 			optval = tp->t_flags & TF_NODELAY;
2120 			INP_WUNLOCK(inp);
2121 			error = sooptcopyout(sopt, &optval, sizeof optval);
2122 			break;
2123 		case TCP_MAXSEG:
2124 			optval = tp->t_maxseg;
2125 			INP_WUNLOCK(inp);
2126 			error = sooptcopyout(sopt, &optval, sizeof optval);
2127 			break;
2128 		case TCP_NOOPT:
2129 			optval = tp->t_flags & TF_NOOPT;
2130 			INP_WUNLOCK(inp);
2131 			error = sooptcopyout(sopt, &optval, sizeof optval);
2132 			break;
2133 		case TCP_NOPUSH:
2134 			optval = tp->t_flags & TF_NOPUSH;
2135 			INP_WUNLOCK(inp);
2136 			error = sooptcopyout(sopt, &optval, sizeof optval);
2137 			break;
2138 		case TCP_INFO:
2139 			tcp_fill_info(tp, &ti);
2140 			INP_WUNLOCK(inp);
2141 			error = sooptcopyout(sopt, &ti, sizeof ti);
2142 			break;
2143 		case TCP_CONGESTION:
2144 			len = strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
2145 			INP_WUNLOCK(inp);
2146 			error = sooptcopyout(sopt, buf, len + 1);
2147 			break;
2148 		case TCP_KEEPIDLE:
2149 		case TCP_KEEPINTVL:
2150 		case TCP_KEEPINIT:
2151 		case TCP_KEEPCNT:
2152 			switch (sopt->sopt_name) {
2153 			case TCP_KEEPIDLE:
2154 				ui = TP_KEEPIDLE(tp) / hz;
2155 				break;
2156 			case TCP_KEEPINTVL:
2157 				ui = TP_KEEPINTVL(tp) / hz;
2158 				break;
2159 			case TCP_KEEPINIT:
2160 				ui = TP_KEEPINIT(tp) / hz;
2161 				break;
2162 			case TCP_KEEPCNT:
2163 				ui = TP_KEEPCNT(tp);
2164 				break;
2165 			}
2166 			INP_WUNLOCK(inp);
2167 			error = sooptcopyout(sopt, &ui, sizeof(ui));
2168 			break;
2169 #ifdef TCPPCAP
2170 		case TCP_PCAP_OUT:
2171 		case TCP_PCAP_IN:
2172 			optval = tcp_pcap_get_sock_max(TCP_PCAP_OUT ?
2173 					&(tp->t_outpkts) : &(tp->t_inpkts));
2174 			INP_WUNLOCK(inp);
2175 			error = sooptcopyout(sopt, &optval, sizeof optval);
2176 			break;
2177 #endif
2178 		case TCP_FASTOPEN:
2179 			optval = tp->t_flags & TF_FASTOPEN;
2180 			INP_WUNLOCK(inp);
2181 			error = sooptcopyout(sopt, &optval, sizeof optval);
2182 			break;
2183 #ifdef TCP_BLACKBOX
2184 		case TCP_LOG:
2185 			optval = tp->t_logstate;
2186 			INP_WUNLOCK(inp);
2187 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2188 			break;
2189 		case TCP_LOGBUF:
2190 			/* tcp_log_getlogbuf() does INP_WUNLOCK(inp) */
2191 			error = tcp_log_getlogbuf(sopt, tp);
2192 			break;
2193 		case TCP_LOGID:
2194 			len = tcp_log_get_id(tp, buf);
2195 			INP_WUNLOCK(inp);
2196 			error = sooptcopyout(sopt, buf, len + 1);
2197 			break;
2198 		case TCP_LOGDUMP:
2199 		case TCP_LOGDUMPID:
2200 			INP_WUNLOCK(inp);
2201 			error = EINVAL;
2202 			break;
2203 #endif
2204 		default:
2205 			INP_WUNLOCK(inp);
2206 			error = ENOPROTOOPT;
2207 			break;
2208 		}
2209 		break;
2210 	}
2211 	return (error);
2212 }
2213 #undef INP_WLOCK_RECHECK
2214 #undef INP_WLOCK_RECHECK_CLEANUP
2215 
2216 /*
2217  * Attach TCP protocol to socket, allocating
2218  * internet protocol control block, tcp control block,
2219  * bufer space, and entering LISTEN state if to accept connections.
2220  */
2221 static int
2222 tcp_attach(struct socket *so)
2223 {
2224 	struct tcpcb *tp;
2225 	struct inpcb *inp;
2226 	struct epoch_tracker et;
2227 	int error;
2228 
2229 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
2230 		error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
2231 		if (error)
2232 			return (error);
2233 	}
2234 	so->so_rcv.sb_flags |= SB_AUTOSIZE;
2235 	so->so_snd.sb_flags |= SB_AUTOSIZE;
2236 	INP_INFO_RLOCK_ET(&V_tcbinfo, et);
2237 	error = in_pcballoc(so, &V_tcbinfo);
2238 	if (error) {
2239 		INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
2240 		return (error);
2241 	}
2242 	inp = sotoinpcb(so);
2243 #ifdef INET6
2244 	if (inp->inp_vflag & INP_IPV6PROTO) {
2245 		inp->inp_vflag |= INP_IPV6;
2246 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
2247 			inp->inp_vflag |= INP_IPV4;
2248 		inp->in6p_hops = -1;	/* use kernel default */
2249 	}
2250 	else
2251 #endif
2252 	inp->inp_vflag |= INP_IPV4;
2253 	tp = tcp_newtcpcb(inp);
2254 	if (tp == NULL) {
2255 		in_pcbdetach(inp);
2256 		in_pcbfree(inp);
2257 		INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
2258 		return (ENOBUFS);
2259 	}
2260 	tp->t_state = TCPS_CLOSED;
2261 	INP_WUNLOCK(inp);
2262 	INP_INFO_RUNLOCK_ET(&V_tcbinfo, et);
2263 	TCPSTATES_INC(TCPS_CLOSED);
2264 	return (0);
2265 }
2266 
2267 /*
2268  * Initiate (or continue) disconnect.
2269  * If embryonic state, just send reset (once).
2270  * If in ``let data drain'' option and linger null, just drop.
2271  * Otherwise (hard), mark socket disconnecting and drop
2272  * current input data; switch states based on user close, and
2273  * send segment to peer (with FIN).
2274  */
2275 static void
2276 tcp_disconnect(struct tcpcb *tp)
2277 {
2278 	struct inpcb *inp = tp->t_inpcb;
2279 	struct socket *so = inp->inp_socket;
2280 
2281 	INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
2282 	INP_WLOCK_ASSERT(inp);
2283 
2284 	/*
2285 	 * Neither tcp_close() nor tcp_drop() should return NULL, as the
2286 	 * socket is still open.
2287 	 */
2288 	if (tp->t_state < TCPS_ESTABLISHED &&
2289 	    !(tp->t_state > TCPS_LISTEN && IS_FASTOPEN(tp->t_flags))) {
2290 		tp = tcp_close(tp);
2291 		KASSERT(tp != NULL,
2292 		    ("tcp_disconnect: tcp_close() returned NULL"));
2293 	} else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
2294 		tp = tcp_drop(tp, 0);
2295 		KASSERT(tp != NULL,
2296 		    ("tcp_disconnect: tcp_drop() returned NULL"));
2297 	} else {
2298 		soisdisconnecting(so);
2299 		sbflush(&so->so_rcv);
2300 		tcp_usrclosed(tp);
2301 		if (!(inp->inp_flags & INP_DROPPED))
2302 			tp->t_fb->tfb_tcp_output(tp);
2303 	}
2304 }
2305 
2306 /*
2307  * User issued close, and wish to trail through shutdown states:
2308  * if never received SYN, just forget it.  If got a SYN from peer,
2309  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
2310  * If already got a FIN from peer, then almost done; go to LAST_ACK
2311  * state.  In all other cases, have already sent FIN to peer (e.g.
2312  * after PRU_SHUTDOWN), and just have to play tedious game waiting
2313  * for peer to send FIN or not respond to keep-alives, etc.
2314  * We can let the user exit from the close as soon as the FIN is acked.
2315  */
2316 static void
2317 tcp_usrclosed(struct tcpcb *tp)
2318 {
2319 
2320 	INP_INFO_RLOCK_ASSERT(&V_tcbinfo);
2321 	INP_WLOCK_ASSERT(tp->t_inpcb);
2322 
2323 	switch (tp->t_state) {
2324 	case TCPS_LISTEN:
2325 #ifdef TCP_OFFLOAD
2326 		tcp_offload_listen_stop(tp);
2327 #endif
2328 		tcp_state_change(tp, TCPS_CLOSED);
2329 		/* FALLTHROUGH */
2330 	case TCPS_CLOSED:
2331 		tp = tcp_close(tp);
2332 		/*
2333 		 * tcp_close() should never return NULL here as the socket is
2334 		 * still open.
2335 		 */
2336 		KASSERT(tp != NULL,
2337 		    ("tcp_usrclosed: tcp_close() returned NULL"));
2338 		break;
2339 
2340 	case TCPS_SYN_SENT:
2341 	case TCPS_SYN_RECEIVED:
2342 		tp->t_flags |= TF_NEEDFIN;
2343 		break;
2344 
2345 	case TCPS_ESTABLISHED:
2346 		tcp_state_change(tp, TCPS_FIN_WAIT_1);
2347 		break;
2348 
2349 	case TCPS_CLOSE_WAIT:
2350 		tcp_state_change(tp, TCPS_LAST_ACK);
2351 		break;
2352 	}
2353 	if (tp->t_state >= TCPS_FIN_WAIT_2) {
2354 		soisdisconnected(tp->t_inpcb->inp_socket);
2355 		/* Prevent the connection hanging in FIN_WAIT_2 forever. */
2356 		if (tp->t_state == TCPS_FIN_WAIT_2) {
2357 			int timeout;
2358 
2359 			timeout = (tcp_fast_finwait2_recycle) ?
2360 			    tcp_finwait2_timeout : TP_MAXIDLE(tp);
2361 			tcp_timer_activate(tp, TT_2MSL, timeout);
2362 		}
2363 	}
2364 }
2365 
2366 #ifdef DDB
2367 static void
2368 db_print_indent(int indent)
2369 {
2370 	int i;
2371 
2372 	for (i = 0; i < indent; i++)
2373 		db_printf(" ");
2374 }
2375 
2376 static void
2377 db_print_tstate(int t_state)
2378 {
2379 
2380 	switch (t_state) {
2381 	case TCPS_CLOSED:
2382 		db_printf("TCPS_CLOSED");
2383 		return;
2384 
2385 	case TCPS_LISTEN:
2386 		db_printf("TCPS_LISTEN");
2387 		return;
2388 
2389 	case TCPS_SYN_SENT:
2390 		db_printf("TCPS_SYN_SENT");
2391 		return;
2392 
2393 	case TCPS_SYN_RECEIVED:
2394 		db_printf("TCPS_SYN_RECEIVED");
2395 		return;
2396 
2397 	case TCPS_ESTABLISHED:
2398 		db_printf("TCPS_ESTABLISHED");
2399 		return;
2400 
2401 	case TCPS_CLOSE_WAIT:
2402 		db_printf("TCPS_CLOSE_WAIT");
2403 		return;
2404 
2405 	case TCPS_FIN_WAIT_1:
2406 		db_printf("TCPS_FIN_WAIT_1");
2407 		return;
2408 
2409 	case TCPS_CLOSING:
2410 		db_printf("TCPS_CLOSING");
2411 		return;
2412 
2413 	case TCPS_LAST_ACK:
2414 		db_printf("TCPS_LAST_ACK");
2415 		return;
2416 
2417 	case TCPS_FIN_WAIT_2:
2418 		db_printf("TCPS_FIN_WAIT_2");
2419 		return;
2420 
2421 	case TCPS_TIME_WAIT:
2422 		db_printf("TCPS_TIME_WAIT");
2423 		return;
2424 
2425 	default:
2426 		db_printf("unknown");
2427 		return;
2428 	}
2429 }
2430 
2431 static void
2432 db_print_tflags(u_int t_flags)
2433 {
2434 	int comma;
2435 
2436 	comma = 0;
2437 	if (t_flags & TF_ACKNOW) {
2438 		db_printf("%sTF_ACKNOW", comma ? ", " : "");
2439 		comma = 1;
2440 	}
2441 	if (t_flags & TF_DELACK) {
2442 		db_printf("%sTF_DELACK", comma ? ", " : "");
2443 		comma = 1;
2444 	}
2445 	if (t_flags & TF_NODELAY) {
2446 		db_printf("%sTF_NODELAY", comma ? ", " : "");
2447 		comma = 1;
2448 	}
2449 	if (t_flags & TF_NOOPT) {
2450 		db_printf("%sTF_NOOPT", comma ? ", " : "");
2451 		comma = 1;
2452 	}
2453 	if (t_flags & TF_SENTFIN) {
2454 		db_printf("%sTF_SENTFIN", comma ? ", " : "");
2455 		comma = 1;
2456 	}
2457 	if (t_flags & TF_REQ_SCALE) {
2458 		db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
2459 		comma = 1;
2460 	}
2461 	if (t_flags & TF_RCVD_SCALE) {
2462 		db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
2463 		comma = 1;
2464 	}
2465 	if (t_flags & TF_REQ_TSTMP) {
2466 		db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
2467 		comma = 1;
2468 	}
2469 	if (t_flags & TF_RCVD_TSTMP) {
2470 		db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
2471 		comma = 1;
2472 	}
2473 	if (t_flags & TF_SACK_PERMIT) {
2474 		db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
2475 		comma = 1;
2476 	}
2477 	if (t_flags & TF_NEEDSYN) {
2478 		db_printf("%sTF_NEEDSYN", comma ? ", " : "");
2479 		comma = 1;
2480 	}
2481 	if (t_flags & TF_NEEDFIN) {
2482 		db_printf("%sTF_NEEDFIN", comma ? ", " : "");
2483 		comma = 1;
2484 	}
2485 	if (t_flags & TF_NOPUSH) {
2486 		db_printf("%sTF_NOPUSH", comma ? ", " : "");
2487 		comma = 1;
2488 	}
2489 	if (t_flags & TF_MORETOCOME) {
2490 		db_printf("%sTF_MORETOCOME", comma ? ", " : "");
2491 		comma = 1;
2492 	}
2493 	if (t_flags & TF_LQ_OVERFLOW) {
2494 		db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
2495 		comma = 1;
2496 	}
2497 	if (t_flags & TF_LASTIDLE) {
2498 		db_printf("%sTF_LASTIDLE", comma ? ", " : "");
2499 		comma = 1;
2500 	}
2501 	if (t_flags & TF_RXWIN0SENT) {
2502 		db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
2503 		comma = 1;
2504 	}
2505 	if (t_flags & TF_FASTRECOVERY) {
2506 		db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
2507 		comma = 1;
2508 	}
2509 	if (t_flags & TF_CONGRECOVERY) {
2510 		db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
2511 		comma = 1;
2512 	}
2513 	if (t_flags & TF_WASFRECOVERY) {
2514 		db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
2515 		comma = 1;
2516 	}
2517 	if (t_flags & TF_SIGNATURE) {
2518 		db_printf("%sTF_SIGNATURE", comma ? ", " : "");
2519 		comma = 1;
2520 	}
2521 	if (t_flags & TF_FORCEDATA) {
2522 		db_printf("%sTF_FORCEDATA", comma ? ", " : "");
2523 		comma = 1;
2524 	}
2525 	if (t_flags & TF_TSO) {
2526 		db_printf("%sTF_TSO", comma ? ", " : "");
2527 		comma = 1;
2528 	}
2529 	if (t_flags & TF_ECN_PERMIT) {
2530 		db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
2531 		comma = 1;
2532 	}
2533 	if (t_flags & TF_FASTOPEN) {
2534 		db_printf("%sTF_FASTOPEN", comma ? ", " : "");
2535 		comma = 1;
2536 	}
2537 }
2538 
2539 static void
2540 db_print_toobflags(char t_oobflags)
2541 {
2542 	int comma;
2543 
2544 	comma = 0;
2545 	if (t_oobflags & TCPOOB_HAVEDATA) {
2546 		db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
2547 		comma = 1;
2548 	}
2549 	if (t_oobflags & TCPOOB_HADDATA) {
2550 		db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
2551 		comma = 1;
2552 	}
2553 }
2554 
2555 static void
2556 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
2557 {
2558 
2559 	db_print_indent(indent);
2560 	db_printf("%s at %p\n", name, tp);
2561 
2562 	indent += 2;
2563 
2564 	db_print_indent(indent);
2565 	db_printf("t_segq first: %p   t_segqlen: %d   t_dupacks: %d\n",
2566 	   TAILQ_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
2567 
2568 	db_print_indent(indent);
2569 	db_printf("tt_rexmt: %p   tt_persist: %p   tt_keep: %p\n",
2570 	    &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
2571 
2572 	db_print_indent(indent);
2573 	db_printf("tt_2msl: %p   tt_delack: %p   t_inpcb: %p\n", &tp->t_timers->tt_2msl,
2574 	    &tp->t_timers->tt_delack, tp->t_inpcb);
2575 
2576 	db_print_indent(indent);
2577 	db_printf("t_state: %d (", tp->t_state);
2578 	db_print_tstate(tp->t_state);
2579 	db_printf(")\n");
2580 
2581 	db_print_indent(indent);
2582 	db_printf("t_flags: 0x%x (", tp->t_flags);
2583 	db_print_tflags(tp->t_flags);
2584 	db_printf(")\n");
2585 
2586 	db_print_indent(indent);
2587 	db_printf("snd_una: 0x%08x   snd_max: 0x%08x   snd_nxt: x0%08x\n",
2588 	    tp->snd_una, tp->snd_max, tp->snd_nxt);
2589 
2590 	db_print_indent(indent);
2591 	db_printf("snd_up: 0x%08x   snd_wl1: 0x%08x   snd_wl2: 0x%08x\n",
2592 	   tp->snd_up, tp->snd_wl1, tp->snd_wl2);
2593 
2594 	db_print_indent(indent);
2595 	db_printf("iss: 0x%08x   irs: 0x%08x   rcv_nxt: 0x%08x\n",
2596 	    tp->iss, tp->irs, tp->rcv_nxt);
2597 
2598 	db_print_indent(indent);
2599 	db_printf("rcv_adv: 0x%08x   rcv_wnd: %u   rcv_up: 0x%08x\n",
2600 	    tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
2601 
2602 	db_print_indent(indent);
2603 	db_printf("snd_wnd: %u   snd_cwnd: %u\n",
2604 	   tp->snd_wnd, tp->snd_cwnd);
2605 
2606 	db_print_indent(indent);
2607 	db_printf("snd_ssthresh: %u   snd_recover: "
2608 	    "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
2609 
2610 	db_print_indent(indent);
2611 	db_printf("t_rcvtime: %u   t_startime: %u\n",
2612 	    tp->t_rcvtime, tp->t_starttime);
2613 
2614 	db_print_indent(indent);
2615 	db_printf("t_rttime: %u   t_rtsq: 0x%08x\n",
2616 	    tp->t_rtttime, tp->t_rtseq);
2617 
2618 	db_print_indent(indent);
2619 	db_printf("t_rxtcur: %d   t_maxseg: %u   t_srtt: %d\n",
2620 	    tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
2621 
2622 	db_print_indent(indent);
2623 	db_printf("t_rttvar: %d   t_rxtshift: %d   t_rttmin: %u   "
2624 	    "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
2625 	    tp->t_rttbest);
2626 
2627 	db_print_indent(indent);
2628 	db_printf("t_rttupdated: %lu   max_sndwnd: %u   t_softerror: %d\n",
2629 	    tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
2630 
2631 	db_print_indent(indent);
2632 	db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
2633 	db_print_toobflags(tp->t_oobflags);
2634 	db_printf(")   t_iobc: 0x%02x\n", tp->t_iobc);
2635 
2636 	db_print_indent(indent);
2637 	db_printf("snd_scale: %u   rcv_scale: %u   request_r_scale: %u\n",
2638 	    tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
2639 
2640 	db_print_indent(indent);
2641 	db_printf("ts_recent: %u   ts_recent_age: %u\n",
2642 	    tp->ts_recent, tp->ts_recent_age);
2643 
2644 	db_print_indent(indent);
2645 	db_printf("ts_offset: %u   last_ack_sent: 0x%08x   snd_cwnd_prev: "
2646 	    "%u\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
2647 
2648 	db_print_indent(indent);
2649 	db_printf("snd_ssthresh_prev: %u   snd_recover_prev: 0x%08x   "
2650 	    "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
2651 	    tp->snd_recover_prev, tp->t_badrxtwin);
2652 
2653 	db_print_indent(indent);
2654 	db_printf("snd_numholes: %d  snd_holes first: %p\n",
2655 	    tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
2656 
2657 	db_print_indent(indent);
2658 	db_printf("snd_fack: 0x%08x   rcv_numsacks: %d   sack_newdata: "
2659 	    "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
2660 
2661 	/* Skip sackblks, sackhint. */
2662 
2663 	db_print_indent(indent);
2664 	db_printf("t_rttlow: %d   rfbuf_ts: %u   rfbuf_cnt: %d\n",
2665 	    tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
2666 }
2667 
2668 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
2669 {
2670 	struct tcpcb *tp;
2671 
2672 	if (!have_addr) {
2673 		db_printf("usage: show tcpcb <addr>\n");
2674 		return;
2675 	}
2676 	tp = (struct tcpcb *)addr;
2677 
2678 	db_print_tcpcb(tp, "tcpcb", 0);
2679 }
2680 #endif
2681