xref: /freebsd/sys/netinet/tcp_usrreq.c (revision 7d0d268b8a67f28ccefdd0b8ce6fb38acac78d80)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.
4  * Copyright (c) 2006-2007 Robert N. M. Watson
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 4. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	From: @(#)tcp_usrreq.c	8.2 (Berkeley) 1/3/94
32  */
33 
34 #include <sys/cdefs.h>
35 __FBSDID("$FreeBSD$");
36 
37 #include "opt_ddb.h"
38 #include "opt_inet.h"
39 #include "opt_inet6.h"
40 #include "opt_tcpdebug.h"
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/malloc.h>
45 #include <sys/kernel.h>
46 #include <sys/sysctl.h>
47 #include <sys/mbuf.h>
48 #ifdef INET6
49 #include <sys/domain.h>
50 #endif /* INET6 */
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/protosw.h>
54 #include <sys/proc.h>
55 #include <sys/jail.h>
56 #include <sys/vimage.h>
57 
58 #ifdef DDB
59 #include <ddb/ddb.h>
60 #endif
61 
62 #include <net/if.h>
63 #include <net/route.h>
64 
65 #include <netinet/in.h>
66 #include <netinet/in_systm.h>
67 #ifdef INET6
68 #include <netinet/ip6.h>
69 #endif
70 #include <netinet/in_pcb.h>
71 #ifdef INET6
72 #include <netinet6/in6_pcb.h>
73 #endif
74 #include <netinet/in_var.h>
75 #include <netinet/ip_var.h>
76 #ifdef INET6
77 #include <netinet6/ip6_var.h>
78 #include <netinet6/scope6_var.h>
79 #endif
80 #include <netinet/tcp.h>
81 #include <netinet/tcp_fsm.h>
82 #include <netinet/tcp_seq.h>
83 #include <netinet/tcp_timer.h>
84 #include <netinet/tcp_var.h>
85 #include <netinet/tcpip.h>
86 #ifdef TCPDEBUG
87 #include <netinet/tcp_debug.h>
88 #endif
89 #include <netinet/tcp_offload.h>
90 #include <netinet/vinet.h>
91 
92 /*
93  * TCP protocol interface to socket abstraction.
94  */
95 static int	tcp_attach(struct socket *);
96 static int	tcp_connect(struct tcpcb *, struct sockaddr *,
97 		    struct thread *td);
98 #ifdef INET6
99 static int	tcp6_connect(struct tcpcb *, struct sockaddr *,
100 		    struct thread *td);
101 #endif /* INET6 */
102 static void	tcp_disconnect(struct tcpcb *);
103 static void	tcp_usrclosed(struct tcpcb *);
104 static void	tcp_fill_info(struct tcpcb *, struct tcp_info *);
105 
106 #ifdef TCPDEBUG
107 #define	TCPDEBUG0	int ostate = 0
108 #define	TCPDEBUG1()	ostate = tp ? tp->t_state : 0
109 #define	TCPDEBUG2(req)	if (tp && (so->so_options & SO_DEBUG)) \
110 				tcp_trace(TA_USER, ostate, tp, 0, 0, req)
111 #else
112 #define	TCPDEBUG0
113 #define	TCPDEBUG1()
114 #define	TCPDEBUG2(req)
115 #endif
116 
117 /*
118  * TCP attaches to socket via pru_attach(), reserving space,
119  * and an internet control block.
120  */
121 static int
122 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
123 {
124 	struct inpcb *inp;
125 	struct tcpcb *tp = NULL;
126 	int error;
127 	TCPDEBUG0;
128 
129 	inp = sotoinpcb(so);
130 	KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
131 	TCPDEBUG1();
132 
133 	error = tcp_attach(so);
134 	if (error)
135 		goto out;
136 
137 	if ((so->so_options & SO_LINGER) && so->so_linger == 0)
138 		so->so_linger = TCP_LINGERTIME;
139 
140 	inp = sotoinpcb(so);
141 	tp = intotcpcb(inp);
142 out:
143 	TCPDEBUG2(PRU_ATTACH);
144 	return error;
145 }
146 
147 /*
148  * tcp_detach is called when the socket layer loses its final reference
149  * to the socket, be it a file descriptor reference, a reference from TCP,
150  * etc.  At this point, there is only one case in which we will keep around
151  * inpcb state: time wait.
152  *
153  * This function can probably be re-absorbed back into tcp_usr_detach() now
154  * that there is a single detach path.
155  */
156 static void
157 tcp_detach(struct socket *so, struct inpcb *inp)
158 {
159 	struct tcpcb *tp;
160 #ifdef INVARIANTS
161 	INIT_VNET_INET(so->so_vnet);
162 #endif
163 
164 	INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
165 	INP_WLOCK_ASSERT(inp);
166 
167 	KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
168 	KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
169 
170 	tp = intotcpcb(inp);
171 
172 	if (inp->inp_vflag & INP_TIMEWAIT) {
173 		/*
174 		 * There are two cases to handle: one in which the time wait
175 		 * state is being discarded (INP_DROPPED), and one in which
176 		 * this connection will remain in timewait.  In the former,
177 		 * it is time to discard all state (except tcptw, which has
178 		 * already been discarded by the timewait close code, which
179 		 * should be further up the call stack somewhere).  In the
180 		 * latter case, we detach from the socket, but leave the pcb
181 		 * present until timewait ends.
182 		 *
183 		 * XXXRW: Would it be cleaner to free the tcptw here?
184 		 */
185 		if (inp->inp_vflag & INP_DROPPED) {
186 			KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && "
187 			    "INP_DROPPED && tp != NULL"));
188 			in_pcbdetach(inp);
189 			in_pcbfree(inp);
190 		} else {
191 			in_pcbdetach(inp);
192 			INP_WUNLOCK(inp);
193 		}
194 	} else {
195 		/*
196 		 * If the connection is not in timewait, we consider two
197 		 * two conditions: one in which no further processing is
198 		 * necessary (dropped || embryonic), and one in which TCP is
199 		 * not yet done, but no longer requires the socket, so the
200 		 * pcb will persist for the time being.
201 		 *
202 		 * XXXRW: Does the second case still occur?
203 		 */
204 		if (inp->inp_vflag & INP_DROPPED ||
205 		    tp->t_state < TCPS_SYN_SENT) {
206 			tcp_discardcb(tp);
207 			in_pcbdetach(inp);
208 			in_pcbfree(inp);
209 		} else
210 			in_pcbdetach(inp);
211 	}
212 }
213 
214 /*
215  * pru_detach() detaches the TCP protocol from the socket.
216  * If the protocol state is non-embryonic, then can't
217  * do this directly: have to initiate a pru_disconnect(),
218  * which may finish later; embryonic TCB's can just
219  * be discarded here.
220  */
221 static void
222 tcp_usr_detach(struct socket *so)
223 {
224 	INIT_VNET_INET(so->so_vnet);
225 	struct inpcb *inp;
226 
227 	inp = sotoinpcb(so);
228 	KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
229 	INP_INFO_WLOCK(&V_tcbinfo);
230 	INP_WLOCK(inp);
231 	KASSERT(inp->inp_socket != NULL,
232 	    ("tcp_usr_detach: inp_socket == NULL"));
233 	tcp_detach(so, inp);
234 	INP_INFO_WUNLOCK(&V_tcbinfo);
235 }
236 
237 /*
238  * Give the socket an address.
239  */
240 static int
241 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
242 {
243 	INIT_VNET_INET(so->so_vnet);
244 	int error = 0;
245 	struct inpcb *inp;
246 	struct tcpcb *tp = NULL;
247 	struct sockaddr_in *sinp;
248 
249 	sinp = (struct sockaddr_in *)nam;
250 	if (nam->sa_len != sizeof (*sinp))
251 		return (EINVAL);
252 	/*
253 	 * Must check for multicast addresses and disallow binding
254 	 * to them.
255 	 */
256 	if (sinp->sin_family == AF_INET &&
257 	    IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
258 		return (EAFNOSUPPORT);
259 
260 	TCPDEBUG0;
261 	INP_INFO_WLOCK(&V_tcbinfo);
262 	inp = sotoinpcb(so);
263 	KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
264 	INP_WLOCK(inp);
265 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
266 		error = EINVAL;
267 		goto out;
268 	}
269 	tp = intotcpcb(inp);
270 	TCPDEBUG1();
271 	error = in_pcbbind(inp, nam, td->td_ucred);
272 out:
273 	TCPDEBUG2(PRU_BIND);
274 	INP_WUNLOCK(inp);
275 	INP_INFO_WUNLOCK(&V_tcbinfo);
276 
277 	return (error);
278 }
279 
280 #ifdef INET6
281 static int
282 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
283 {
284 	INIT_VNET_INET(so->so_vnet);
285 	int error = 0;
286 	struct inpcb *inp;
287 	struct tcpcb *tp = NULL;
288 	struct sockaddr_in6 *sin6p;
289 
290 	sin6p = (struct sockaddr_in6 *)nam;
291 	if (nam->sa_len != sizeof (*sin6p))
292 		return (EINVAL);
293 	/*
294 	 * Must check for multicast addresses and disallow binding
295 	 * to them.
296 	 */
297 	if (sin6p->sin6_family == AF_INET6 &&
298 	    IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
299 		return (EAFNOSUPPORT);
300 
301 	TCPDEBUG0;
302 	INP_INFO_WLOCK(&V_tcbinfo);
303 	inp = sotoinpcb(so);
304 	KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
305 	INP_WLOCK(inp);
306 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
307 		error = EINVAL;
308 		goto out;
309 	}
310 	tp = intotcpcb(inp);
311 	TCPDEBUG1();
312 	inp->inp_vflag &= ~INP_IPV4;
313 	inp->inp_vflag |= INP_IPV6;
314 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
315 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
316 			inp->inp_vflag |= INP_IPV4;
317 		else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
318 			struct sockaddr_in sin;
319 
320 			in6_sin6_2_sin(&sin, sin6p);
321 			inp->inp_vflag |= INP_IPV4;
322 			inp->inp_vflag &= ~INP_IPV6;
323 			error = in_pcbbind(inp, (struct sockaddr *)&sin,
324 			    td->td_ucred);
325 			goto out;
326 		}
327 	}
328 	error = in6_pcbbind(inp, nam, td->td_ucred);
329 out:
330 	TCPDEBUG2(PRU_BIND);
331 	INP_WUNLOCK(inp);
332 	INP_INFO_WUNLOCK(&V_tcbinfo);
333 	return (error);
334 }
335 #endif /* INET6 */
336 
337 /*
338  * Prepare to accept connections.
339  */
340 static int
341 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
342 {
343 	INIT_VNET_INET(so->so_vnet);
344 	int error = 0;
345 	struct inpcb *inp;
346 	struct tcpcb *tp = NULL;
347 
348 	TCPDEBUG0;
349 	INP_INFO_WLOCK(&V_tcbinfo);
350 	inp = sotoinpcb(so);
351 	KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
352 	INP_WLOCK(inp);
353 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
354 		error = EINVAL;
355 		goto out;
356 	}
357 	tp = intotcpcb(inp);
358 	TCPDEBUG1();
359 	SOCK_LOCK(so);
360 	error = solisten_proto_check(so);
361 	if (error == 0 && inp->inp_lport == 0)
362 		error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
363 	if (error == 0) {
364 		tp->t_state = TCPS_LISTEN;
365 		solisten_proto(so, backlog);
366 		tcp_offload_listen_open(tp);
367 	}
368 	SOCK_UNLOCK(so);
369 
370 out:
371 	TCPDEBUG2(PRU_LISTEN);
372 	INP_WUNLOCK(inp);
373 	INP_INFO_WUNLOCK(&V_tcbinfo);
374 	return (error);
375 }
376 
377 #ifdef INET6
378 static int
379 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
380 {
381 	INIT_VNET_INET(so->so_vnet);
382 	int error = 0;
383 	struct inpcb *inp;
384 	struct tcpcb *tp = NULL;
385 
386 	TCPDEBUG0;
387 	INP_INFO_WLOCK(&V_tcbinfo);
388 	inp = sotoinpcb(so);
389 	KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
390 	INP_WLOCK(inp);
391 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
392 		error = EINVAL;
393 		goto out;
394 	}
395 	tp = intotcpcb(inp);
396 	TCPDEBUG1();
397 	SOCK_LOCK(so);
398 	error = solisten_proto_check(so);
399 	if (error == 0 && inp->inp_lport == 0) {
400 		inp->inp_vflag &= ~INP_IPV4;
401 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
402 			inp->inp_vflag |= INP_IPV4;
403 		error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
404 	}
405 	if (error == 0) {
406 		tp->t_state = TCPS_LISTEN;
407 		solisten_proto(so, backlog);
408 	}
409 	SOCK_UNLOCK(so);
410 
411 out:
412 	TCPDEBUG2(PRU_LISTEN);
413 	INP_WUNLOCK(inp);
414 	INP_INFO_WUNLOCK(&V_tcbinfo);
415 	return (error);
416 }
417 #endif /* INET6 */
418 
419 /*
420  * Initiate connection to peer.
421  * Create a template for use in transmissions on this connection.
422  * Enter SYN_SENT state, and mark socket as connecting.
423  * Start keep-alive timer, and seed output sequence space.
424  * Send initial segment on connection.
425  */
426 static int
427 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
428 {
429 	INIT_VNET_INET(so->so_vnet);
430 	int error = 0;
431 	struct inpcb *inp;
432 	struct tcpcb *tp = NULL;
433 	struct sockaddr_in *sinp;
434 
435 	sinp = (struct sockaddr_in *)nam;
436 	if (nam->sa_len != sizeof (*sinp))
437 		return (EINVAL);
438 	/*
439 	 * Must disallow TCP ``connections'' to multicast addresses.
440 	 */
441 	if (sinp->sin_family == AF_INET
442 	    && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
443 		return (EAFNOSUPPORT);
444 	if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
445 		return (error);
446 
447 	TCPDEBUG0;
448 	INP_INFO_WLOCK(&V_tcbinfo);
449 	inp = sotoinpcb(so);
450 	KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
451 	INP_WLOCK(inp);
452 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
453 		error = EINVAL;
454 		goto out;
455 	}
456 	tp = intotcpcb(inp);
457 	TCPDEBUG1();
458 	if ((error = tcp_connect(tp, nam, td)) != 0)
459 		goto out;
460 	error = tcp_output_connect(so, nam);
461 out:
462 	TCPDEBUG2(PRU_CONNECT);
463 	INP_WUNLOCK(inp);
464 	INP_INFO_WUNLOCK(&V_tcbinfo);
465 	return (error);
466 }
467 
468 #ifdef INET6
469 static int
470 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
471 {
472 	INIT_VNET_INET(so->so_vnet);
473 	int error = 0;
474 	struct inpcb *inp;
475 	struct tcpcb *tp = NULL;
476 	struct sockaddr_in6 *sin6p;
477 
478 	TCPDEBUG0;
479 
480 	sin6p = (struct sockaddr_in6 *)nam;
481 	if (nam->sa_len != sizeof (*sin6p))
482 		return (EINVAL);
483 	/*
484 	 * Must disallow TCP ``connections'' to multicast addresses.
485 	 */
486 	if (sin6p->sin6_family == AF_INET6
487 	    && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
488 		return (EAFNOSUPPORT);
489 
490 	INP_INFO_WLOCK(&V_tcbinfo);
491 	inp = sotoinpcb(so);
492 	KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
493 	INP_WLOCK(inp);
494 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
495 		error = EINVAL;
496 		goto out;
497 	}
498 	tp = intotcpcb(inp);
499 	TCPDEBUG1();
500 	if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
501 		struct sockaddr_in sin;
502 
503 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
504 			error = EINVAL;
505 			goto out;
506 		}
507 
508 		in6_sin6_2_sin(&sin, sin6p);
509 		inp->inp_vflag |= INP_IPV4;
510 		inp->inp_vflag &= ~INP_IPV6;
511 		if ((error = prison_remote_ip4(td->td_ucred,
512 		    &sin.sin_addr)) != 0)
513 			goto out;
514 		if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
515 			goto out;
516 		error = tcp_output_connect(so, nam);
517 		goto out;
518 	}
519 	inp->inp_vflag &= ~INP_IPV4;
520 	inp->inp_vflag |= INP_IPV6;
521 	inp->inp_inc.inc_flags |= INC_ISIPV6;
522 	if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
523 		goto out;
524 	if ((error = tcp6_connect(tp, nam, td)) != 0)
525 		goto out;
526 	error = tcp_output_connect(so, nam);
527 
528 out:
529 	TCPDEBUG2(PRU_CONNECT);
530 	INP_WUNLOCK(inp);
531 	INP_INFO_WUNLOCK(&V_tcbinfo);
532 	return (error);
533 }
534 #endif /* INET6 */
535 
536 /*
537  * Initiate disconnect from peer.
538  * If connection never passed embryonic stage, just drop;
539  * else if don't need to let data drain, then can just drop anyways,
540  * else have to begin TCP shutdown process: mark socket disconnecting,
541  * drain unread data, state switch to reflect user close, and
542  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
543  * when peer sends FIN and acks ours.
544  *
545  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
546  */
547 static int
548 tcp_usr_disconnect(struct socket *so)
549 {
550 	INIT_VNET_INET(so->so_vnet);
551 	struct inpcb *inp;
552 	struct tcpcb *tp = NULL;
553 	int error = 0;
554 
555 	TCPDEBUG0;
556 	INP_INFO_WLOCK(&V_tcbinfo);
557 	inp = sotoinpcb(so);
558 	KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
559 	INP_WLOCK(inp);
560 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
561 		error = ECONNRESET;
562 		goto out;
563 	}
564 	tp = intotcpcb(inp);
565 	TCPDEBUG1();
566 	tcp_disconnect(tp);
567 out:
568 	TCPDEBUG2(PRU_DISCONNECT);
569 	INP_WUNLOCK(inp);
570 	INP_INFO_WUNLOCK(&V_tcbinfo);
571 	return (error);
572 }
573 
574 /*
575  * Accept a connection.  Essentially all the work is
576  * done at higher levels; just return the address
577  * of the peer, storing through addr.
578  */
579 static int
580 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
581 {
582 	INIT_VNET_INET(so->so_vnet);
583 	int error = 0;
584 	struct inpcb *inp = NULL;
585 	struct tcpcb *tp = NULL;
586 	struct in_addr addr;
587 	in_port_t port = 0;
588 	TCPDEBUG0;
589 
590 	if (so->so_state & SS_ISDISCONNECTED)
591 		return (ECONNABORTED);
592 
593 	inp = sotoinpcb(so);
594 	KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
595 	INP_INFO_RLOCK(&V_tcbinfo);
596 	INP_WLOCK(inp);
597 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
598 		error = ECONNABORTED;
599 		goto out;
600 	}
601 	tp = intotcpcb(inp);
602 	TCPDEBUG1();
603 
604 	/*
605 	 * We inline in_getpeeraddr and COMMON_END here, so that we can
606 	 * copy the data of interest and defer the malloc until after we
607 	 * release the lock.
608 	 */
609 	port = inp->inp_fport;
610 	addr = inp->inp_faddr;
611 
612 out:
613 	TCPDEBUG2(PRU_ACCEPT);
614 	INP_WUNLOCK(inp);
615 	INP_INFO_RUNLOCK(&V_tcbinfo);
616 	if (error == 0)
617 		*nam = in_sockaddr(port, &addr);
618 	return error;
619 }
620 
621 #ifdef INET6
622 static int
623 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
624 {
625 	struct inpcb *inp = NULL;
626 	int error = 0;
627 	struct tcpcb *tp = NULL;
628 	struct in_addr addr;
629 	struct in6_addr addr6;
630 	in_port_t port = 0;
631 	int v4 = 0;
632 	TCPDEBUG0;
633 
634 	if (so->so_state & SS_ISDISCONNECTED)
635 		return (ECONNABORTED);
636 
637 	inp = sotoinpcb(so);
638 	KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
639 	INP_WLOCK(inp);
640 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
641 		error = ECONNABORTED;
642 		goto out;
643 	}
644 	tp = intotcpcb(inp);
645 	TCPDEBUG1();
646 
647 	/*
648 	 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
649 	 * copy the data of interest and defer the malloc until after we
650 	 * release the lock.
651 	 */
652 	if (inp->inp_vflag & INP_IPV4) {
653 		v4 = 1;
654 		port = inp->inp_fport;
655 		addr = inp->inp_faddr;
656 	} else {
657 		port = inp->inp_fport;
658 		addr6 = inp->in6p_faddr;
659 	}
660 
661 out:
662 	TCPDEBUG2(PRU_ACCEPT);
663 	INP_WUNLOCK(inp);
664 	if (error == 0) {
665 		if (v4)
666 			*nam = in6_v4mapsin6_sockaddr(port, &addr);
667 		else
668 			*nam = in6_sockaddr(port, &addr6);
669 	}
670 	return error;
671 }
672 #endif /* INET6 */
673 
674 /*
675  * Mark the connection as being incapable of further output.
676  */
677 static int
678 tcp_usr_shutdown(struct socket *so)
679 {
680 	INIT_VNET_INET(so->so_vnet);
681 	int error = 0;
682 	struct inpcb *inp;
683 	struct tcpcb *tp = NULL;
684 
685 	TCPDEBUG0;
686 	INP_INFO_WLOCK(&V_tcbinfo);
687 	inp = sotoinpcb(so);
688 	KASSERT(inp != NULL, ("inp == NULL"));
689 	INP_WLOCK(inp);
690 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
691 		error = ECONNRESET;
692 		goto out;
693 	}
694 	tp = intotcpcb(inp);
695 	TCPDEBUG1();
696 	socantsendmore(so);
697 	tcp_usrclosed(tp);
698 	error = tcp_output_disconnect(tp);
699 
700 out:
701 	TCPDEBUG2(PRU_SHUTDOWN);
702 	INP_WUNLOCK(inp);
703 	INP_INFO_WUNLOCK(&V_tcbinfo);
704 
705 	return (error);
706 }
707 
708 /*
709  * After a receive, possibly send window update to peer.
710  */
711 static int
712 tcp_usr_rcvd(struct socket *so, int flags)
713 {
714 	struct inpcb *inp;
715 	struct tcpcb *tp = NULL;
716 	int error = 0;
717 
718 	TCPDEBUG0;
719 	inp = sotoinpcb(so);
720 	KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
721 	INP_WLOCK(inp);
722 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
723 		error = ECONNRESET;
724 		goto out;
725 	}
726 	tp = intotcpcb(inp);
727 	TCPDEBUG1();
728 	tcp_output_rcvd(tp);
729 
730 out:
731 	TCPDEBUG2(PRU_RCVD);
732 	INP_WUNLOCK(inp);
733 	return (error);
734 }
735 
736 /*
737  * Do a send by putting data in output queue and updating urgent
738  * marker if URG set.  Possibly send more data.  Unlike the other
739  * pru_*() routines, the mbuf chains are our responsibility.  We
740  * must either enqueue them or free them.  The other pru_* routines
741  * generally are caller-frees.
742  */
743 static int
744 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
745     struct sockaddr *nam, struct mbuf *control, struct thread *td)
746 {
747 	INIT_VNET_INET(so->so_vnet);
748 	int error = 0;
749 	struct inpcb *inp;
750 	struct tcpcb *tp = NULL;
751 	int headlocked = 0;
752 #ifdef INET6
753 	int isipv6;
754 #endif
755 	TCPDEBUG0;
756 
757 	/*
758 	 * We require the pcbinfo lock in two cases:
759 	 *
760 	 * (1) An implied connect is taking place, which can result in
761 	 *     binding IPs and ports and hence modification of the pcb hash
762 	 *     chains.
763 	 *
764 	 * (2) PRUS_EOF is set, resulting in explicit close on the send.
765 	 */
766 	if ((nam != NULL) || (flags & PRUS_EOF)) {
767 		INP_INFO_WLOCK(&V_tcbinfo);
768 		headlocked = 1;
769 	}
770 	inp = sotoinpcb(so);
771 	KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
772 	INP_WLOCK(inp);
773 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
774 		if (control)
775 			m_freem(control);
776 		if (m)
777 			m_freem(m);
778 		error = ECONNRESET;
779 		goto out;
780 	}
781 #ifdef INET6
782 	isipv6 = nam && nam->sa_family == AF_INET6;
783 #endif /* INET6 */
784 	tp = intotcpcb(inp);
785 	TCPDEBUG1();
786 	if (control) {
787 		/* TCP doesn't do control messages (rights, creds, etc) */
788 		if (control->m_len) {
789 			m_freem(control);
790 			if (m)
791 				m_freem(m);
792 			error = EINVAL;
793 			goto out;
794 		}
795 		m_freem(control);	/* empty control, just free it */
796 	}
797 	if (!(flags & PRUS_OOB)) {
798 		sbappendstream(&so->so_snd, m);
799 		if (nam && tp->t_state < TCPS_SYN_SENT) {
800 			/*
801 			 * Do implied connect if not yet connected,
802 			 * initialize window to default value, and
803 			 * initialize maxseg/maxopd using peer's cached
804 			 * MSS.
805 			 */
806 			INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
807 #ifdef INET6
808 			if (isipv6)
809 				error = tcp6_connect(tp, nam, td);
810 			else
811 #endif /* INET6 */
812 			error = tcp_connect(tp, nam, td);
813 			if (error)
814 				goto out;
815 			tp->snd_wnd = TTCP_CLIENT_SND_WND;
816 			tcp_mss(tp, -1);
817 		}
818 		if (flags & PRUS_EOF) {
819 			/*
820 			 * Close the send side of the connection after
821 			 * the data is sent.
822 			 */
823 			INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
824 			socantsendmore(so);
825 			tcp_usrclosed(tp);
826 		}
827 		if (headlocked) {
828 			INP_INFO_WUNLOCK(&V_tcbinfo);
829 			headlocked = 0;
830 		}
831 		if (tp != NULL) {
832 			if (flags & PRUS_MORETOCOME)
833 				tp->t_flags |= TF_MORETOCOME;
834 			error = tcp_output_send(tp);
835 			if (flags & PRUS_MORETOCOME)
836 				tp->t_flags &= ~TF_MORETOCOME;
837 		}
838 	} else {
839 		/*
840 		 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
841 		 */
842 		SOCKBUF_LOCK(&so->so_snd);
843 		if (sbspace(&so->so_snd) < -512) {
844 			SOCKBUF_UNLOCK(&so->so_snd);
845 			m_freem(m);
846 			error = ENOBUFS;
847 			goto out;
848 		}
849 		/*
850 		 * According to RFC961 (Assigned Protocols),
851 		 * the urgent pointer points to the last octet
852 		 * of urgent data.  We continue, however,
853 		 * to consider it to indicate the first octet
854 		 * of data past the urgent section.
855 		 * Otherwise, snd_up should be one lower.
856 		 */
857 		sbappendstream_locked(&so->so_snd, m);
858 		SOCKBUF_UNLOCK(&so->so_snd);
859 		if (nam && tp->t_state < TCPS_SYN_SENT) {
860 			/*
861 			 * Do implied connect if not yet connected,
862 			 * initialize window to default value, and
863 			 * initialize maxseg/maxopd using peer's cached
864 			 * MSS.
865 			 */
866 			INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
867 #ifdef INET6
868 			if (isipv6)
869 				error = tcp6_connect(tp, nam, td);
870 			else
871 #endif /* INET6 */
872 			error = tcp_connect(tp, nam, td);
873 			if (error)
874 				goto out;
875 			tp->snd_wnd = TTCP_CLIENT_SND_WND;
876 			tcp_mss(tp, -1);
877 			INP_INFO_WUNLOCK(&V_tcbinfo);
878 			headlocked = 0;
879 		} else if (nam) {
880 			INP_INFO_WUNLOCK(&V_tcbinfo);
881 			headlocked = 0;
882 		}
883 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
884 		tp->t_flags |= TF_FORCEDATA;
885 		error = tcp_output_send(tp);
886 		tp->t_flags &= ~TF_FORCEDATA;
887 	}
888 out:
889 	TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
890 		  ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
891 	INP_WUNLOCK(inp);
892 	if (headlocked)
893 		INP_INFO_WUNLOCK(&V_tcbinfo);
894 	return (error);
895 }
896 
897 /*
898  * Abort the TCP.  Drop the connection abruptly.
899  */
900 static void
901 tcp_usr_abort(struct socket *so)
902 {
903 	INIT_VNET_INET(so->so_vnet);
904 	struct inpcb *inp;
905 	struct tcpcb *tp = NULL;
906 	TCPDEBUG0;
907 
908 	inp = sotoinpcb(so);
909 	KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
910 
911 	INP_INFO_WLOCK(&V_tcbinfo);
912 	INP_WLOCK(inp);
913 	KASSERT(inp->inp_socket != NULL,
914 	    ("tcp_usr_abort: inp_socket == NULL"));
915 
916 	/*
917 	 * If we still have full TCP state, and we're not dropped, drop.
918 	 */
919 	if (!(inp->inp_vflag & INP_TIMEWAIT) &&
920 	    !(inp->inp_vflag & INP_DROPPED)) {
921 		tp = intotcpcb(inp);
922 		TCPDEBUG1();
923 		tcp_drop(tp, ECONNABORTED);
924 		TCPDEBUG2(PRU_ABORT);
925 	}
926 	if (!(inp->inp_vflag & INP_DROPPED)) {
927 		SOCK_LOCK(so);
928 		so->so_state |= SS_PROTOREF;
929 		SOCK_UNLOCK(so);
930 		inp->inp_vflag |= INP_SOCKREF;
931 	}
932 	INP_WUNLOCK(inp);
933 	INP_INFO_WUNLOCK(&V_tcbinfo);
934 }
935 
936 /*
937  * TCP socket is closed.  Start friendly disconnect.
938  */
939 static void
940 tcp_usr_close(struct socket *so)
941 {
942 	INIT_VNET_INET(so->so_vnet);
943 	struct inpcb *inp;
944 	struct tcpcb *tp = NULL;
945 	TCPDEBUG0;
946 
947 	inp = sotoinpcb(so);
948 	KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
949 
950 	INP_INFO_WLOCK(&V_tcbinfo);
951 	INP_WLOCK(inp);
952 	KASSERT(inp->inp_socket != NULL,
953 	    ("tcp_usr_close: inp_socket == NULL"));
954 
955 	/*
956 	 * If we still have full TCP state, and we're not dropped, initiate
957 	 * a disconnect.
958 	 */
959 	if (!(inp->inp_vflag & INP_TIMEWAIT) &&
960 	    !(inp->inp_vflag & INP_DROPPED)) {
961 		tp = intotcpcb(inp);
962 		TCPDEBUG1();
963 		tcp_disconnect(tp);
964 		TCPDEBUG2(PRU_CLOSE);
965 	}
966 	if (!(inp->inp_vflag & INP_DROPPED)) {
967 		SOCK_LOCK(so);
968 		so->so_state |= SS_PROTOREF;
969 		SOCK_UNLOCK(so);
970 		inp->inp_vflag |= INP_SOCKREF;
971 	}
972 	INP_WUNLOCK(inp);
973 	INP_INFO_WUNLOCK(&V_tcbinfo);
974 }
975 
976 /*
977  * Receive out-of-band data.
978  */
979 static int
980 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
981 {
982 	int error = 0;
983 	struct inpcb *inp;
984 	struct tcpcb *tp = NULL;
985 
986 	TCPDEBUG0;
987 	inp = sotoinpcb(so);
988 	KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
989 	INP_WLOCK(inp);
990 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
991 		error = ECONNRESET;
992 		goto out;
993 	}
994 	tp = intotcpcb(inp);
995 	TCPDEBUG1();
996 	if ((so->so_oobmark == 0 &&
997 	     (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
998 	    so->so_options & SO_OOBINLINE ||
999 	    tp->t_oobflags & TCPOOB_HADDATA) {
1000 		error = EINVAL;
1001 		goto out;
1002 	}
1003 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1004 		error = EWOULDBLOCK;
1005 		goto out;
1006 	}
1007 	m->m_len = 1;
1008 	*mtod(m, caddr_t) = tp->t_iobc;
1009 	if ((flags & MSG_PEEK) == 0)
1010 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1011 
1012 out:
1013 	TCPDEBUG2(PRU_RCVOOB);
1014 	INP_WUNLOCK(inp);
1015 	return (error);
1016 }
1017 
1018 struct pr_usrreqs tcp_usrreqs = {
1019 	.pru_abort =		tcp_usr_abort,
1020 	.pru_accept =		tcp_usr_accept,
1021 	.pru_attach =		tcp_usr_attach,
1022 	.pru_bind =		tcp_usr_bind,
1023 	.pru_connect =		tcp_usr_connect,
1024 	.pru_control =		in_control,
1025 	.pru_detach =		tcp_usr_detach,
1026 	.pru_disconnect =	tcp_usr_disconnect,
1027 	.pru_listen =		tcp_usr_listen,
1028 	.pru_peeraddr =		in_getpeeraddr,
1029 	.pru_rcvd =		tcp_usr_rcvd,
1030 	.pru_rcvoob =		tcp_usr_rcvoob,
1031 	.pru_send =		tcp_usr_send,
1032 	.pru_shutdown =		tcp_usr_shutdown,
1033 	.pru_sockaddr =		in_getsockaddr,
1034 	.pru_sosetlabel =	in_pcbsosetlabel,
1035 	.pru_close =		tcp_usr_close,
1036 };
1037 
1038 #ifdef INET6
1039 struct pr_usrreqs tcp6_usrreqs = {
1040 	.pru_abort =		tcp_usr_abort,
1041 	.pru_accept =		tcp6_usr_accept,
1042 	.pru_attach =		tcp_usr_attach,
1043 	.pru_bind =		tcp6_usr_bind,
1044 	.pru_connect =		tcp6_usr_connect,
1045 	.pru_control =		in6_control,
1046 	.pru_detach =		tcp_usr_detach,
1047 	.pru_disconnect =	tcp_usr_disconnect,
1048 	.pru_listen =		tcp6_usr_listen,
1049 	.pru_peeraddr =		in6_mapped_peeraddr,
1050 	.pru_rcvd =		tcp_usr_rcvd,
1051 	.pru_rcvoob =		tcp_usr_rcvoob,
1052 	.pru_send =		tcp_usr_send,
1053 	.pru_shutdown =		tcp_usr_shutdown,
1054 	.pru_sockaddr =		in6_mapped_sockaddr,
1055  	.pru_sosetlabel =	in_pcbsosetlabel,
1056 	.pru_close =		tcp_usr_close,
1057 };
1058 #endif /* INET6 */
1059 
1060 /*
1061  * Common subroutine to open a TCP connection to remote host specified
1062  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
1063  * port number if needed.  Call in_pcbconnect_setup to do the routing and
1064  * to choose a local host address (interface).  If there is an existing
1065  * incarnation of the same connection in TIME-WAIT state and if the remote
1066  * host was sending CC options and if the connection duration was < MSL, then
1067  * truncate the previous TIME-WAIT state and proceed.
1068  * Initialize connection parameters and enter SYN-SENT state.
1069  */
1070 static int
1071 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1072 {
1073 	struct inpcb *inp = tp->t_inpcb, *oinp;
1074 	struct socket *so = inp->inp_socket;
1075 	INIT_VNET_INET(so->so_vnet);
1076 	struct in_addr laddr;
1077 	u_short lport;
1078 	int error;
1079 
1080 	INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1081 	INP_WLOCK_ASSERT(inp);
1082 
1083 	if (inp->inp_lport == 0) {
1084 		error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1085 		if (error)
1086 			return error;
1087 	}
1088 
1089 	/*
1090 	 * Cannot simply call in_pcbconnect, because there might be an
1091 	 * earlier incarnation of this same connection still in
1092 	 * TIME_WAIT state, creating an ADDRINUSE error.
1093 	 */
1094 	laddr = inp->inp_laddr;
1095 	lport = inp->inp_lport;
1096 	error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1097 	    &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1098 	if (error && oinp == NULL)
1099 		return error;
1100 	if (oinp)
1101 		return EADDRINUSE;
1102 	inp->inp_laddr = laddr;
1103 	in_pcbrehash(inp);
1104 
1105 	/*
1106 	 * Compute window scaling to request:
1107 	 * Scale to fit into sweet spot.  See tcp_syncache.c.
1108 	 * XXX: This should move to tcp_output().
1109 	 */
1110 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1111 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1112 		tp->request_r_scale++;
1113 
1114 	soisconnecting(so);
1115 	V_tcpstat.tcps_connattempt++;
1116 	tp->t_state = TCPS_SYN_SENT;
1117 	tcp_timer_activate(tp, TT_KEEP, tcp_keepinit);
1118 	tp->iss = tcp_new_isn(tp);
1119 	tp->t_bw_rtseq = tp->iss;
1120 	tcp_sendseqinit(tp);
1121 
1122 	return 0;
1123 }
1124 
1125 #ifdef INET6
1126 static int
1127 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1128 {
1129 	struct inpcb *inp = tp->t_inpcb, *oinp;
1130 	struct socket *so = inp->inp_socket;
1131 	INIT_VNET_INET(so->so_vnet);
1132 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
1133 	struct in6_addr *addr6;
1134 	int error;
1135 
1136 	INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1137 	INP_WLOCK_ASSERT(inp);
1138 
1139 	if (inp->inp_lport == 0) {
1140 		error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1141 		if (error)
1142 			return error;
1143 	}
1144 
1145 	/*
1146 	 * Cannot simply call in_pcbconnect, because there might be an
1147 	 * earlier incarnation of this same connection still in
1148 	 * TIME_WAIT state, creating an ADDRINUSE error.
1149 	 * in6_pcbladdr() also handles scope zone IDs.
1150 	 */
1151 	error = in6_pcbladdr(inp, nam, &addr6);
1152 	if (error)
1153 		return error;
1154 	oinp = in6_pcblookup_hash(inp->inp_pcbinfo,
1155 				  &sin6->sin6_addr, sin6->sin6_port,
1156 				  IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
1157 				  ? addr6
1158 				  : &inp->in6p_laddr,
1159 				  inp->inp_lport,  0, NULL);
1160 	if (oinp)
1161 		return EADDRINUSE;
1162 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
1163 		inp->in6p_laddr = *addr6;
1164 	inp->in6p_faddr = sin6->sin6_addr;
1165 	inp->inp_fport = sin6->sin6_port;
1166 	/* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
1167 	inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
1168 	if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
1169 		inp->inp_flow |=
1170 		    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
1171 	in_pcbrehash(inp);
1172 
1173 	/* Compute window scaling to request.  */
1174 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1175 	    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
1176 		tp->request_r_scale++;
1177 
1178 	soisconnecting(so);
1179 	V_tcpstat.tcps_connattempt++;
1180 	tp->t_state = TCPS_SYN_SENT;
1181 	tcp_timer_activate(tp, TT_KEEP, tcp_keepinit);
1182 	tp->iss = tcp_new_isn(tp);
1183 	tp->t_bw_rtseq = tp->iss;
1184 	tcp_sendseqinit(tp);
1185 
1186 	return 0;
1187 }
1188 #endif /* INET6 */
1189 
1190 /*
1191  * Export TCP internal state information via a struct tcp_info, based on the
1192  * Linux 2.6 API.  Not ABI compatible as our constants are mapped differently
1193  * (TCP state machine, etc).  We export all information using FreeBSD-native
1194  * constants -- for example, the numeric values for tcpi_state will differ
1195  * from Linux.
1196  */
1197 static void
1198 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1199 {
1200 
1201 	INP_WLOCK_ASSERT(tp->t_inpcb);
1202 	bzero(ti, sizeof(*ti));
1203 
1204 	ti->tcpi_state = tp->t_state;
1205 	if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1206 		ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1207 	if (tp->t_flags & TF_SACK_PERMIT)
1208 		ti->tcpi_options |= TCPI_OPT_SACK;
1209 	if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1210 		ti->tcpi_options |= TCPI_OPT_WSCALE;
1211 		ti->tcpi_snd_wscale = tp->snd_scale;
1212 		ti->tcpi_rcv_wscale = tp->rcv_scale;
1213 	}
1214 
1215 	ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1216 	ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1217 
1218 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1219 	ti->tcpi_snd_cwnd = tp->snd_cwnd;
1220 
1221 	/*
1222 	 * FreeBSD-specific extension fields for tcp_info.
1223 	 */
1224 	ti->tcpi_rcv_space = tp->rcv_wnd;
1225 	ti->tcpi_rcv_nxt = tp->rcv_nxt;
1226 	ti->tcpi_snd_wnd = tp->snd_wnd;
1227 	ti->tcpi_snd_bwnd = tp->snd_bwnd;
1228 	ti->tcpi_snd_nxt = tp->snd_nxt;
1229 	ti->__tcpi_snd_mss = tp->t_maxseg;
1230 	ti->__tcpi_rcv_mss = tp->t_maxseg;
1231 	if (tp->t_flags & TF_TOE)
1232 		ti->tcpi_options |= TCPI_OPT_TOE;
1233 }
1234 
1235 /*
1236  * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1237  * socket option arguments.  When it re-acquires the lock after the copy, it
1238  * has to revalidate that the connection is still valid for the socket
1239  * option.
1240  */
1241 #define INP_WLOCK_RECHECK(inp) do {					\
1242 	INP_WLOCK(inp);							\
1243 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {		\
1244 		INP_WUNLOCK(inp);					\
1245 		return (ECONNRESET);					\
1246 	}								\
1247 	tp = intotcpcb(inp);						\
1248 } while(0)
1249 
1250 int
1251 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1252 {
1253 	INIT_VNET_INET(so->so_vnet);
1254 	int	error, opt, optval;
1255 	struct	inpcb *inp;
1256 	struct	tcpcb *tp;
1257 	struct	tcp_info ti;
1258 
1259 	error = 0;
1260 	inp = sotoinpcb(so);
1261 	KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1262 	INP_WLOCK(inp);
1263 	if (sopt->sopt_level != IPPROTO_TCP) {
1264 #ifdef INET6
1265 		if (inp->inp_vflag & INP_IPV6PROTO) {
1266 			INP_WUNLOCK(inp);
1267 			error = ip6_ctloutput(so, sopt);
1268 		} else {
1269 #endif /* INET6 */
1270 			INP_WUNLOCK(inp);
1271 			error = ip_ctloutput(so, sopt);
1272 #ifdef INET6
1273 		}
1274 #endif
1275 		return (error);
1276 	}
1277 	if (inp->inp_vflag & (INP_TIMEWAIT | INP_DROPPED)) {
1278 		INP_WUNLOCK(inp);
1279 		return (ECONNRESET);
1280 	}
1281 
1282 	switch (sopt->sopt_dir) {
1283 	case SOPT_SET:
1284 		switch (sopt->sopt_name) {
1285 #ifdef TCP_SIGNATURE
1286 		case TCP_MD5SIG:
1287 			INP_WUNLOCK(inp);
1288 			error = sooptcopyin(sopt, &optval, sizeof optval,
1289 			    sizeof optval);
1290 			if (error)
1291 				return (error);
1292 
1293 			INP_WLOCK_RECHECK(inp);
1294 			if (optval > 0)
1295 				tp->t_flags |= TF_SIGNATURE;
1296 			else
1297 				tp->t_flags &= ~TF_SIGNATURE;
1298 			INP_WUNLOCK(inp);
1299 			break;
1300 #endif /* TCP_SIGNATURE */
1301 		case TCP_NODELAY:
1302 		case TCP_NOOPT:
1303 			INP_WUNLOCK(inp);
1304 			error = sooptcopyin(sopt, &optval, sizeof optval,
1305 			    sizeof optval);
1306 			if (error)
1307 				return (error);
1308 
1309 			INP_WLOCK_RECHECK(inp);
1310 			switch (sopt->sopt_name) {
1311 			case TCP_NODELAY:
1312 				opt = TF_NODELAY;
1313 				break;
1314 			case TCP_NOOPT:
1315 				opt = TF_NOOPT;
1316 				break;
1317 			default:
1318 				opt = 0; /* dead code to fool gcc */
1319 				break;
1320 			}
1321 
1322 			if (optval)
1323 				tp->t_flags |= opt;
1324 			else
1325 				tp->t_flags &= ~opt;
1326 			INP_WUNLOCK(inp);
1327 			break;
1328 
1329 		case TCP_NOPUSH:
1330 			INP_WUNLOCK(inp);
1331 			error = sooptcopyin(sopt, &optval, sizeof optval,
1332 			    sizeof optval);
1333 			if (error)
1334 				return (error);
1335 
1336 			INP_WLOCK_RECHECK(inp);
1337 			if (optval)
1338 				tp->t_flags |= TF_NOPUSH;
1339 			else {
1340 				tp->t_flags &= ~TF_NOPUSH;
1341 				error = tcp_output(tp);
1342 			}
1343 			INP_WUNLOCK(inp);
1344 			break;
1345 
1346 		case TCP_MAXSEG:
1347 			INP_WUNLOCK(inp);
1348 			error = sooptcopyin(sopt, &optval, sizeof optval,
1349 			    sizeof optval);
1350 			if (error)
1351 				return (error);
1352 
1353 			INP_WLOCK_RECHECK(inp);
1354 			if (optval > 0 && optval <= tp->t_maxseg &&
1355 			    optval + 40 >= V_tcp_minmss)
1356 				tp->t_maxseg = optval;
1357 			else
1358 				error = EINVAL;
1359 			INP_WUNLOCK(inp);
1360 			break;
1361 
1362 		case TCP_INFO:
1363 			INP_WUNLOCK(inp);
1364 			error = EINVAL;
1365 			break;
1366 
1367 		default:
1368 			INP_WUNLOCK(inp);
1369 			error = ENOPROTOOPT;
1370 			break;
1371 		}
1372 		break;
1373 
1374 	case SOPT_GET:
1375 		tp = intotcpcb(inp);
1376 		switch (sopt->sopt_name) {
1377 #ifdef TCP_SIGNATURE
1378 		case TCP_MD5SIG:
1379 			optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
1380 			INP_WUNLOCK(inp);
1381 			error = sooptcopyout(sopt, &optval, sizeof optval);
1382 			break;
1383 #endif
1384 
1385 		case TCP_NODELAY:
1386 			optval = tp->t_flags & TF_NODELAY;
1387 			INP_WUNLOCK(inp);
1388 			error = sooptcopyout(sopt, &optval, sizeof optval);
1389 			break;
1390 		case TCP_MAXSEG:
1391 			optval = tp->t_maxseg;
1392 			INP_WUNLOCK(inp);
1393 			error = sooptcopyout(sopt, &optval, sizeof optval);
1394 			break;
1395 		case TCP_NOOPT:
1396 			optval = tp->t_flags & TF_NOOPT;
1397 			INP_WUNLOCK(inp);
1398 			error = sooptcopyout(sopt, &optval, sizeof optval);
1399 			break;
1400 		case TCP_NOPUSH:
1401 			optval = tp->t_flags & TF_NOPUSH;
1402 			INP_WUNLOCK(inp);
1403 			error = sooptcopyout(sopt, &optval, sizeof optval);
1404 			break;
1405 		case TCP_INFO:
1406 			tcp_fill_info(tp, &ti);
1407 			INP_WUNLOCK(inp);
1408 			error = sooptcopyout(sopt, &ti, sizeof ti);
1409 			break;
1410 		default:
1411 			INP_WUNLOCK(inp);
1412 			error = ENOPROTOOPT;
1413 			break;
1414 		}
1415 		break;
1416 	}
1417 	return (error);
1418 }
1419 #undef INP_WLOCK_RECHECK
1420 
1421 /*
1422  * tcp_sendspace and tcp_recvspace are the default send and receive window
1423  * sizes, respectively.  These are obsolescent (this information should
1424  * be set by the route).
1425  */
1426 u_long	tcp_sendspace = 1024*32;
1427 SYSCTL_ULONG(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLFLAG_RW,
1428     &tcp_sendspace , 0, "Maximum outgoing TCP datagram size");
1429 u_long	tcp_recvspace = 1024*64;
1430 SYSCTL_ULONG(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
1431     &tcp_recvspace , 0, "Maximum incoming TCP datagram size");
1432 
1433 /*
1434  * Attach TCP protocol to socket, allocating
1435  * internet protocol control block, tcp control block,
1436  * bufer space, and entering LISTEN state if to accept connections.
1437  */
1438 static int
1439 tcp_attach(struct socket *so)
1440 {
1441 	INIT_VNET_INET(so->so_vnet);
1442 	struct tcpcb *tp;
1443 	struct inpcb *inp;
1444 	int error;
1445 
1446 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1447 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
1448 		if (error)
1449 			return (error);
1450 	}
1451 	so->so_rcv.sb_flags |= SB_AUTOSIZE;
1452 	so->so_snd.sb_flags |= SB_AUTOSIZE;
1453 	INP_INFO_WLOCK(&V_tcbinfo);
1454 	error = in_pcballoc(so, &V_tcbinfo);
1455 	if (error) {
1456 		INP_INFO_WUNLOCK(&V_tcbinfo);
1457 		return (error);
1458 	}
1459 	inp = sotoinpcb(so);
1460 #ifdef INET6
1461 	if (inp->inp_vflag & INP_IPV6PROTO) {
1462 		inp->inp_vflag |= INP_IPV6;
1463 		inp->in6p_hops = -1;	/* use kernel default */
1464 	}
1465 	else
1466 #endif
1467 	inp->inp_vflag |= INP_IPV4;
1468 	tp = tcp_newtcpcb(inp);
1469 	if (tp == NULL) {
1470 		in_pcbdetach(inp);
1471 		in_pcbfree(inp);
1472 		INP_INFO_WUNLOCK(&V_tcbinfo);
1473 		return (ENOBUFS);
1474 	}
1475 	tp->t_state = TCPS_CLOSED;
1476 	INP_WUNLOCK(inp);
1477 	INP_INFO_WUNLOCK(&V_tcbinfo);
1478 	return (0);
1479 }
1480 
1481 /*
1482  * Initiate (or continue) disconnect.
1483  * If embryonic state, just send reset (once).
1484  * If in ``let data drain'' option and linger null, just drop.
1485  * Otherwise (hard), mark socket disconnecting and drop
1486  * current input data; switch states based on user close, and
1487  * send segment to peer (with FIN).
1488  */
1489 static void
1490 tcp_disconnect(struct tcpcb *tp)
1491 {
1492 	struct inpcb *inp = tp->t_inpcb;
1493 	struct socket *so = inp->inp_socket;
1494 #ifdef INVARIANTS
1495 	INIT_VNET_INET(so->so_vnet);
1496 #endif
1497 
1498 	INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1499 	INP_WLOCK_ASSERT(inp);
1500 
1501 	/*
1502 	 * Neither tcp_close() nor tcp_drop() should return NULL, as the
1503 	 * socket is still open.
1504 	 */
1505 	if (tp->t_state < TCPS_ESTABLISHED) {
1506 		tp = tcp_close(tp);
1507 		KASSERT(tp != NULL,
1508 		    ("tcp_disconnect: tcp_close() returned NULL"));
1509 	} else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
1510 		tp = tcp_drop(tp, 0);
1511 		KASSERT(tp != NULL,
1512 		    ("tcp_disconnect: tcp_drop() returned NULL"));
1513 	} else {
1514 		soisdisconnecting(so);
1515 		sbflush(&so->so_rcv);
1516 		tcp_usrclosed(tp);
1517 		if (!(inp->inp_vflag & INP_DROPPED))
1518 			tcp_output_disconnect(tp);
1519 	}
1520 }
1521 
1522 /*
1523  * User issued close, and wish to trail through shutdown states:
1524  * if never received SYN, just forget it.  If got a SYN from peer,
1525  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1526  * If already got a FIN from peer, then almost done; go to LAST_ACK
1527  * state.  In all other cases, have already sent FIN to peer (e.g.
1528  * after PRU_SHUTDOWN), and just have to play tedious game waiting
1529  * for peer to send FIN or not respond to keep-alives, etc.
1530  * We can let the user exit from the close as soon as the FIN is acked.
1531  */
1532 static void
1533 tcp_usrclosed(struct tcpcb *tp)
1534 {
1535 #ifdef INVARIANTS
1536 	INIT_VNET_INET(tp->t_inpcb->inp_vnet);
1537 #endif
1538 
1539 	INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1540 	INP_WLOCK_ASSERT(tp->t_inpcb);
1541 
1542 	switch (tp->t_state) {
1543 	case TCPS_LISTEN:
1544 		tcp_offload_listen_close(tp);
1545 		/* FALLTHROUGH */
1546 	case TCPS_CLOSED:
1547 		tp->t_state = TCPS_CLOSED;
1548 		tp = tcp_close(tp);
1549 		/*
1550 		 * tcp_close() should never return NULL here as the socket is
1551 		 * still open.
1552 		 */
1553 		KASSERT(tp != NULL,
1554 		    ("tcp_usrclosed: tcp_close() returned NULL"));
1555 		break;
1556 
1557 	case TCPS_SYN_SENT:
1558 	case TCPS_SYN_RECEIVED:
1559 		tp->t_flags |= TF_NEEDFIN;
1560 		break;
1561 
1562 	case TCPS_ESTABLISHED:
1563 		tp->t_state = TCPS_FIN_WAIT_1;
1564 		break;
1565 
1566 	case TCPS_CLOSE_WAIT:
1567 		tp->t_state = TCPS_LAST_ACK;
1568 		break;
1569 	}
1570 	if (tp->t_state >= TCPS_FIN_WAIT_2) {
1571 		soisdisconnected(tp->t_inpcb->inp_socket);
1572 		/* Prevent the connection hanging in FIN_WAIT_2 forever. */
1573 		if (tp->t_state == TCPS_FIN_WAIT_2) {
1574 			int timeout;
1575 
1576 			timeout = (tcp_fast_finwait2_recycle) ?
1577 			    tcp_finwait2_timeout : tcp_maxidle;
1578 			tcp_timer_activate(tp, TT_2MSL, timeout);
1579 		}
1580 	}
1581 }
1582 
1583 #ifdef DDB
1584 static void
1585 db_print_indent(int indent)
1586 {
1587 	int i;
1588 
1589 	for (i = 0; i < indent; i++)
1590 		db_printf(" ");
1591 }
1592 
1593 static void
1594 db_print_tstate(int t_state)
1595 {
1596 
1597 	switch (t_state) {
1598 	case TCPS_CLOSED:
1599 		db_printf("TCPS_CLOSED");
1600 		return;
1601 
1602 	case TCPS_LISTEN:
1603 		db_printf("TCPS_LISTEN");
1604 		return;
1605 
1606 	case TCPS_SYN_SENT:
1607 		db_printf("TCPS_SYN_SENT");
1608 		return;
1609 
1610 	case TCPS_SYN_RECEIVED:
1611 		db_printf("TCPS_SYN_RECEIVED");
1612 		return;
1613 
1614 	case TCPS_ESTABLISHED:
1615 		db_printf("TCPS_ESTABLISHED");
1616 		return;
1617 
1618 	case TCPS_CLOSE_WAIT:
1619 		db_printf("TCPS_CLOSE_WAIT");
1620 		return;
1621 
1622 	case TCPS_FIN_WAIT_1:
1623 		db_printf("TCPS_FIN_WAIT_1");
1624 		return;
1625 
1626 	case TCPS_CLOSING:
1627 		db_printf("TCPS_CLOSING");
1628 		return;
1629 
1630 	case TCPS_LAST_ACK:
1631 		db_printf("TCPS_LAST_ACK");
1632 		return;
1633 
1634 	case TCPS_FIN_WAIT_2:
1635 		db_printf("TCPS_FIN_WAIT_2");
1636 		return;
1637 
1638 	case TCPS_TIME_WAIT:
1639 		db_printf("TCPS_TIME_WAIT");
1640 		return;
1641 
1642 	default:
1643 		db_printf("unknown");
1644 		return;
1645 	}
1646 }
1647 
1648 static void
1649 db_print_tflags(u_int t_flags)
1650 {
1651 	int comma;
1652 
1653 	comma = 0;
1654 	if (t_flags & TF_ACKNOW) {
1655 		db_printf("%sTF_ACKNOW", comma ? ", " : "");
1656 		comma = 1;
1657 	}
1658 	if (t_flags & TF_DELACK) {
1659 		db_printf("%sTF_DELACK", comma ? ", " : "");
1660 		comma = 1;
1661 	}
1662 	if (t_flags & TF_NODELAY) {
1663 		db_printf("%sTF_NODELAY", comma ? ", " : "");
1664 		comma = 1;
1665 	}
1666 	if (t_flags & TF_NOOPT) {
1667 		db_printf("%sTF_NOOPT", comma ? ", " : "");
1668 		comma = 1;
1669 	}
1670 	if (t_flags & TF_SENTFIN) {
1671 		db_printf("%sTF_SENTFIN", comma ? ", " : "");
1672 		comma = 1;
1673 	}
1674 	if (t_flags & TF_REQ_SCALE) {
1675 		db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
1676 		comma = 1;
1677 	}
1678 	if (t_flags & TF_RCVD_SCALE) {
1679 		db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
1680 		comma = 1;
1681 	}
1682 	if (t_flags & TF_REQ_TSTMP) {
1683 		db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
1684 		comma = 1;
1685 	}
1686 	if (t_flags & TF_RCVD_TSTMP) {
1687 		db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
1688 		comma = 1;
1689 	}
1690 	if (t_flags & TF_SACK_PERMIT) {
1691 		db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
1692 		comma = 1;
1693 	}
1694 	if (t_flags & TF_NEEDSYN) {
1695 		db_printf("%sTF_NEEDSYN", comma ? ", " : "");
1696 		comma = 1;
1697 	}
1698 	if (t_flags & TF_NEEDFIN) {
1699 		db_printf("%sTF_NEEDFIN", comma ? ", " : "");
1700 		comma = 1;
1701 	}
1702 	if (t_flags & TF_NOPUSH) {
1703 		db_printf("%sTF_NOPUSH", comma ? ", " : "");
1704 		comma = 1;
1705 	}
1706 	if (t_flags & TF_NOPUSH) {
1707 		db_printf("%sTF_NOPUSH", comma ? ", " : "");
1708 		comma = 1;
1709 	}
1710 	if (t_flags & TF_MORETOCOME) {
1711 		db_printf("%sTF_MORETOCOME", comma ? ", " : "");
1712 		comma = 1;
1713 	}
1714 	if (t_flags & TF_LQ_OVERFLOW) {
1715 		db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
1716 		comma = 1;
1717 	}
1718 	if (t_flags & TF_LASTIDLE) {
1719 		db_printf("%sTF_LASTIDLE", comma ? ", " : "");
1720 		comma = 1;
1721 	}
1722 	if (t_flags & TF_RXWIN0SENT) {
1723 		db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
1724 		comma = 1;
1725 	}
1726 	if (t_flags & TF_FASTRECOVERY) {
1727 		db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
1728 		comma = 1;
1729 	}
1730 	if (t_flags & TF_WASFRECOVERY) {
1731 		db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
1732 		comma = 1;
1733 	}
1734 	if (t_flags & TF_SIGNATURE) {
1735 		db_printf("%sTF_SIGNATURE", comma ? ", " : "");
1736 		comma = 1;
1737 	}
1738 	if (t_flags & TF_FORCEDATA) {
1739 		db_printf("%sTF_FORCEDATA", comma ? ", " : "");
1740 		comma = 1;
1741 	}
1742 	if (t_flags & TF_TSO) {
1743 		db_printf("%sTF_TSO", comma ? ", " : "");
1744 		comma = 1;
1745 	}
1746 	if (t_flags & TF_ECN_PERMIT) {
1747 		db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
1748 		comma = 1;
1749 	}
1750 }
1751 
1752 static void
1753 db_print_toobflags(char t_oobflags)
1754 {
1755 	int comma;
1756 
1757 	comma = 0;
1758 	if (t_oobflags & TCPOOB_HAVEDATA) {
1759 		db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
1760 		comma = 1;
1761 	}
1762 	if (t_oobflags & TCPOOB_HADDATA) {
1763 		db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
1764 		comma = 1;
1765 	}
1766 }
1767 
1768 static void
1769 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
1770 {
1771 
1772 	db_print_indent(indent);
1773 	db_printf("%s at %p\n", name, tp);
1774 
1775 	indent += 2;
1776 
1777 	db_print_indent(indent);
1778 	db_printf("t_segq first: %p   t_segqlen: %d   t_dupacks: %d\n",
1779 	   LIST_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
1780 
1781 	db_print_indent(indent);
1782 	db_printf("tt_rexmt: %p   tt_persist: %p   tt_keep: %p\n",
1783 	    &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
1784 
1785 	db_print_indent(indent);
1786 	db_printf("tt_2msl: %p   tt_delack: %p   t_inpcb: %p\n", &tp->t_timers->tt_2msl,
1787 	    &tp->t_timers->tt_delack, tp->t_inpcb);
1788 
1789 	db_print_indent(indent);
1790 	db_printf("t_state: %d (", tp->t_state);
1791 	db_print_tstate(tp->t_state);
1792 	db_printf(")\n");
1793 
1794 	db_print_indent(indent);
1795 	db_printf("t_flags: 0x%x (", tp->t_flags);
1796 	db_print_tflags(tp->t_flags);
1797 	db_printf(")\n");
1798 
1799 	db_print_indent(indent);
1800 	db_printf("snd_una: 0x%08x   snd_max: 0x%08x   snd_nxt: x0%08x\n",
1801 	    tp->snd_una, tp->snd_max, tp->snd_nxt);
1802 
1803 	db_print_indent(indent);
1804 	db_printf("snd_up: 0x%08x   snd_wl1: 0x%08x   snd_wl2: 0x%08x\n",
1805 	   tp->snd_up, tp->snd_wl1, tp->snd_wl2);
1806 
1807 	db_print_indent(indent);
1808 	db_printf("iss: 0x%08x   irs: 0x%08x   rcv_nxt: 0x%08x\n",
1809 	    tp->iss, tp->irs, tp->rcv_nxt);
1810 
1811 	db_print_indent(indent);
1812 	db_printf("rcv_adv: 0x%08x   rcv_wnd: %lu   rcv_up: 0x%08x\n",
1813 	    tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
1814 
1815 	db_print_indent(indent);
1816 	db_printf("snd_wnd: %lu   snd_cwnd: %lu   snd_bwnd: %lu\n",
1817 	   tp->snd_wnd, tp->snd_cwnd, tp->snd_bwnd);
1818 
1819 	db_print_indent(indent);
1820 	db_printf("snd_ssthresh: %lu   snd_bandwidth: %lu   snd_recover: "
1821 	    "0x%08x\n", tp->snd_ssthresh, tp->snd_bandwidth,
1822 	    tp->snd_recover);
1823 
1824 	db_print_indent(indent);
1825 	db_printf("t_maxopd: %u   t_rcvtime: %lu   t_startime: %lu\n",
1826 	    tp->t_maxopd, tp->t_rcvtime, tp->t_starttime);
1827 
1828 	db_print_indent(indent);
1829 	db_printf("t_rttime: %d   t_rtsq: 0x%08x   t_bw_rtttime: %d\n",
1830 	    tp->t_rtttime, tp->t_rtseq, tp->t_bw_rtttime);
1831 
1832 	db_print_indent(indent);
1833 	db_printf("t_bw_rtseq: 0x%08x   t_rxtcur: %d   t_maxseg: %u   "
1834 	    "t_srtt: %d\n", tp->t_bw_rtseq, tp->t_rxtcur, tp->t_maxseg,
1835 	    tp->t_srtt);
1836 
1837 	db_print_indent(indent);
1838 	db_printf("t_rttvar: %d   t_rxtshift: %d   t_rttmin: %u   "
1839 	    "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
1840 	    tp->t_rttbest);
1841 
1842 	db_print_indent(indent);
1843 	db_printf("t_rttupdated: %lu   max_sndwnd: %lu   t_softerror: %d\n",
1844 	    tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
1845 
1846 	db_print_indent(indent);
1847 	db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
1848 	db_print_toobflags(tp->t_oobflags);
1849 	db_printf(")   t_iobc: 0x%02x\n", tp->t_iobc);
1850 
1851 	db_print_indent(indent);
1852 	db_printf("snd_scale: %u   rcv_scale: %u   request_r_scale: %u\n",
1853 	    tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
1854 
1855 	db_print_indent(indent);
1856 	db_printf("ts_recent: %u   ts_recent_age: %lu\n",
1857 	    tp->ts_recent, tp->ts_recent_age);
1858 
1859 	db_print_indent(indent);
1860 	db_printf("ts_offset: %u   last_ack_sent: 0x%08x   snd_cwnd_prev: "
1861 	    "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
1862 
1863 	db_print_indent(indent);
1864 	db_printf("snd_ssthresh_prev: %lu   snd_recover_prev: 0x%08x   "
1865 	    "t_badrxtwin: %lu\n", tp->snd_ssthresh_prev,
1866 	    tp->snd_recover_prev, tp->t_badrxtwin);
1867 
1868 	db_print_indent(indent);
1869 	db_printf("snd_numholes: %d  snd_holes first: %p\n",
1870 	    tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
1871 
1872 	db_print_indent(indent);
1873 	db_printf("snd_fack: 0x%08x   rcv_numsacks: %d   sack_newdata: "
1874 	    "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
1875 
1876 	/* Skip sackblks, sackhint. */
1877 
1878 	db_print_indent(indent);
1879 	db_printf("t_rttlow: %d   rfbuf_ts: %u   rfbuf_cnt: %d\n",
1880 	    tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
1881 }
1882 
1883 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
1884 {
1885 	struct tcpcb *tp;
1886 
1887 	if (!have_addr) {
1888 		db_printf("usage: show tcpcb <addr>\n");
1889 		return;
1890 	}
1891 	tp = (struct tcpcb *)addr;
1892 
1893 	db_print_tcpcb(tp, "tcpcb", 0);
1894 }
1895 #endif
1896