xref: /freebsd/sys/netinet/tcp_usrreq.c (revision 2b15cb3d0922bd70ea592f0da9b4a5b167f4d53f)
1 /*-
2  * Copyright (c) 1982, 1986, 1988, 1993
3  *	The Regents of the University of California.
4  * Copyright (c) 2006-2007 Robert N. M. Watson
5  * Copyright (c) 2010-2011 Juniper Networks, Inc.
6  * All rights reserved.
7  *
8  * Portions of this software were developed by Robert N. M. Watson under
9  * contract to Juniper Networks, Inc.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	From: @(#)tcp_usrreq.c	8.2 (Berkeley) 1/3/94
36  */
37 
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include "opt_ddb.h"
42 #include "opt_inet.h"
43 #include "opt_inet6.h"
44 #include "opt_tcpdebug.h"
45 
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/limits.h>
49 #include <sys/malloc.h>
50 #include <sys/kernel.h>
51 #include <sys/sysctl.h>
52 #include <sys/mbuf.h>
53 #ifdef INET6
54 #include <sys/domain.h>
55 #endif /* INET6 */
56 #include <sys/socket.h>
57 #include <sys/socketvar.h>
58 #include <sys/protosw.h>
59 #include <sys/proc.h>
60 #include <sys/jail.h>
61 
62 #ifdef DDB
63 #include <ddb/ddb.h>
64 #endif
65 
66 #include <net/if.h>
67 #include <net/if_var.h>
68 #include <net/route.h>
69 #include <net/vnet.h>
70 
71 #include <netinet/cc.h>
72 #include <netinet/in.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/in_systm.h>
75 #include <netinet/in_var.h>
76 #include <netinet/ip_var.h>
77 #ifdef INET6
78 #include <netinet/ip6.h>
79 #include <netinet6/in6_pcb.h>
80 #include <netinet6/ip6_var.h>
81 #include <netinet6/scope6_var.h>
82 #endif
83 #include <netinet/tcp_fsm.h>
84 #include <netinet/tcp_seq.h>
85 #include <netinet/tcp_timer.h>
86 #include <netinet/tcp_var.h>
87 #include <netinet/tcpip.h>
88 #ifdef TCPDEBUG
89 #include <netinet/tcp_debug.h>
90 #endif
91 #ifdef TCP_OFFLOAD
92 #include <netinet/tcp_offload.h>
93 #endif
94 
95 /*
96  * TCP protocol interface to socket abstraction.
97  */
98 static int	tcp_attach(struct socket *);
99 #ifdef INET
100 static int	tcp_connect(struct tcpcb *, struct sockaddr *,
101 		    struct thread *td);
102 #endif /* INET */
103 #ifdef INET6
104 static int	tcp6_connect(struct tcpcb *, struct sockaddr *,
105 		    struct thread *td);
106 #endif /* INET6 */
107 static void	tcp_disconnect(struct tcpcb *);
108 static void	tcp_usrclosed(struct tcpcb *);
109 static void	tcp_fill_info(struct tcpcb *, struct tcp_info *);
110 
111 #ifdef TCPDEBUG
112 #define	TCPDEBUG0	int ostate = 0
113 #define	TCPDEBUG1()	ostate = tp ? tp->t_state : 0
114 #define	TCPDEBUG2(req)	if (tp && (so->so_options & SO_DEBUG)) \
115 				tcp_trace(TA_USER, ostate, tp, 0, 0, req)
116 #else
117 #define	TCPDEBUG0
118 #define	TCPDEBUG1()
119 #define	TCPDEBUG2(req)
120 #endif
121 
122 /*
123  * TCP attaches to socket via pru_attach(), reserving space,
124  * and an internet control block.
125  */
126 static int
127 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
128 {
129 	struct inpcb *inp;
130 	struct tcpcb *tp = NULL;
131 	int error;
132 	TCPDEBUG0;
133 
134 	inp = sotoinpcb(so);
135 	KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
136 	TCPDEBUG1();
137 
138 	error = tcp_attach(so);
139 	if (error)
140 		goto out;
141 
142 	if ((so->so_options & SO_LINGER) && so->so_linger == 0)
143 		so->so_linger = TCP_LINGERTIME;
144 
145 	inp = sotoinpcb(so);
146 	tp = intotcpcb(inp);
147 out:
148 	TCPDEBUG2(PRU_ATTACH);
149 	return error;
150 }
151 
152 /*
153  * tcp_detach is called when the socket layer loses its final reference
154  * to the socket, be it a file descriptor reference, a reference from TCP,
155  * etc.  At this point, there is only one case in which we will keep around
156  * inpcb state: time wait.
157  *
158  * This function can probably be re-absorbed back into tcp_usr_detach() now
159  * that there is a single detach path.
160  */
161 static void
162 tcp_detach(struct socket *so, struct inpcb *inp)
163 {
164 	struct tcpcb *tp;
165 
166 	INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
167 	INP_WLOCK_ASSERT(inp);
168 
169 	KASSERT(so->so_pcb == inp, ("tcp_detach: so_pcb != inp"));
170 	KASSERT(inp->inp_socket == so, ("tcp_detach: inp_socket != so"));
171 
172 	tp = intotcpcb(inp);
173 
174 	if (inp->inp_flags & INP_TIMEWAIT) {
175 		/*
176 		 * There are two cases to handle: one in which the time wait
177 		 * state is being discarded (INP_DROPPED), and one in which
178 		 * this connection will remain in timewait.  In the former,
179 		 * it is time to discard all state (except tcptw, which has
180 		 * already been discarded by the timewait close code, which
181 		 * should be further up the call stack somewhere).  In the
182 		 * latter case, we detach from the socket, but leave the pcb
183 		 * present until timewait ends.
184 		 *
185 		 * XXXRW: Would it be cleaner to free the tcptw here?
186 		 *
187 		 * Astute question indeed, from twtcp perspective there are
188 		 * three cases to consider:
189 		 *
190 		 * #1 tcp_detach is called at tcptw creation time by
191 		 *  tcp_twstart, then do not discard the newly created tcptw
192 		 *  and leave inpcb present until timewait ends
193 		 * #2 tcp_detach is called at timewait end (or reuse) by
194 		 *  tcp_twclose, then the tcptw has already been discarded
195 		 *  and inpcb is freed here
196 		 * #3 tcp_detach is called() after timewait ends (or reuse)
197 		 *  (e.g. by soclose), then tcptw has already been discarded
198 		 *  and inpcb is freed here
199 		 *
200 		 *  In all three cases the tcptw should not be freed here.
201 		 */
202 		if (inp->inp_flags & INP_DROPPED) {
203 			KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && "
204 			    "INP_DROPPED && tp != NULL"));
205 			in_pcbdetach(inp);
206 			in_pcbfree(inp);
207 		} else {
208 			in_pcbdetach(inp);
209 			INP_WUNLOCK(inp);
210 		}
211 	} else {
212 		/*
213 		 * If the connection is not in timewait, we consider two
214 		 * two conditions: one in which no further processing is
215 		 * necessary (dropped || embryonic), and one in which TCP is
216 		 * not yet done, but no longer requires the socket, so the
217 		 * pcb will persist for the time being.
218 		 *
219 		 * XXXRW: Does the second case still occur?
220 		 */
221 		if (inp->inp_flags & INP_DROPPED ||
222 		    tp->t_state < TCPS_SYN_SENT) {
223 			tcp_discardcb(tp);
224 			in_pcbdetach(inp);
225 			in_pcbfree(inp);
226 		} else {
227 			in_pcbdetach(inp);
228 			INP_WUNLOCK(inp);
229 		}
230 	}
231 }
232 
233 /*
234  * pru_detach() detaches the TCP protocol from the socket.
235  * If the protocol state is non-embryonic, then can't
236  * do this directly: have to initiate a pru_disconnect(),
237  * which may finish later; embryonic TCB's can just
238  * be discarded here.
239  */
240 static void
241 tcp_usr_detach(struct socket *so)
242 {
243 	struct inpcb *inp;
244 
245 	inp = sotoinpcb(so);
246 	KASSERT(inp != NULL, ("tcp_usr_detach: inp == NULL"));
247 	INP_INFO_WLOCK(&V_tcbinfo);
248 	INP_WLOCK(inp);
249 	KASSERT(inp->inp_socket != NULL,
250 	    ("tcp_usr_detach: inp_socket == NULL"));
251 	tcp_detach(so, inp);
252 	INP_INFO_WUNLOCK(&V_tcbinfo);
253 }
254 
255 #ifdef INET
256 /*
257  * Give the socket an address.
258  */
259 static int
260 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
261 {
262 	int error = 0;
263 	struct inpcb *inp;
264 	struct tcpcb *tp = NULL;
265 	struct sockaddr_in *sinp;
266 
267 	sinp = (struct sockaddr_in *)nam;
268 	if (nam->sa_len != sizeof (*sinp))
269 		return (EINVAL);
270 	/*
271 	 * Must check for multicast addresses and disallow binding
272 	 * to them.
273 	 */
274 	if (sinp->sin_family == AF_INET &&
275 	    IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
276 		return (EAFNOSUPPORT);
277 
278 	TCPDEBUG0;
279 	inp = sotoinpcb(so);
280 	KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
281 	INP_WLOCK(inp);
282 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
283 		error = EINVAL;
284 		goto out;
285 	}
286 	tp = intotcpcb(inp);
287 	TCPDEBUG1();
288 	INP_HASH_WLOCK(&V_tcbinfo);
289 	error = in_pcbbind(inp, nam, td->td_ucred);
290 	INP_HASH_WUNLOCK(&V_tcbinfo);
291 out:
292 	TCPDEBUG2(PRU_BIND);
293 	INP_WUNLOCK(inp);
294 
295 	return (error);
296 }
297 #endif /* INET */
298 
299 #ifdef INET6
300 static int
301 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
302 {
303 	int error = 0;
304 	struct inpcb *inp;
305 	struct tcpcb *tp = NULL;
306 	struct sockaddr_in6 *sin6p;
307 
308 	sin6p = (struct sockaddr_in6 *)nam;
309 	if (nam->sa_len != sizeof (*sin6p))
310 		return (EINVAL);
311 	/*
312 	 * Must check for multicast addresses and disallow binding
313 	 * to them.
314 	 */
315 	if (sin6p->sin6_family == AF_INET6 &&
316 	    IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
317 		return (EAFNOSUPPORT);
318 
319 	TCPDEBUG0;
320 	inp = sotoinpcb(so);
321 	KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
322 	INP_WLOCK(inp);
323 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
324 		error = EINVAL;
325 		goto out;
326 	}
327 	tp = intotcpcb(inp);
328 	TCPDEBUG1();
329 	INP_HASH_WLOCK(&V_tcbinfo);
330 	inp->inp_vflag &= ~INP_IPV4;
331 	inp->inp_vflag |= INP_IPV6;
332 #ifdef INET
333 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
334 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
335 			inp->inp_vflag |= INP_IPV4;
336 		else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
337 			struct sockaddr_in sin;
338 
339 			in6_sin6_2_sin(&sin, sin6p);
340 			inp->inp_vflag |= INP_IPV4;
341 			inp->inp_vflag &= ~INP_IPV6;
342 			error = in_pcbbind(inp, (struct sockaddr *)&sin,
343 			    td->td_ucred);
344 			INP_HASH_WUNLOCK(&V_tcbinfo);
345 			goto out;
346 		}
347 	}
348 #endif
349 	error = in6_pcbbind(inp, nam, td->td_ucred);
350 	INP_HASH_WUNLOCK(&V_tcbinfo);
351 out:
352 	TCPDEBUG2(PRU_BIND);
353 	INP_WUNLOCK(inp);
354 	return (error);
355 }
356 #endif /* INET6 */
357 
358 #ifdef INET
359 /*
360  * Prepare to accept connections.
361  */
362 static int
363 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
364 {
365 	int error = 0;
366 	struct inpcb *inp;
367 	struct tcpcb *tp = NULL;
368 
369 	TCPDEBUG0;
370 	inp = sotoinpcb(so);
371 	KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
372 	INP_WLOCK(inp);
373 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
374 		error = EINVAL;
375 		goto out;
376 	}
377 	tp = intotcpcb(inp);
378 	TCPDEBUG1();
379 	SOCK_LOCK(so);
380 	error = solisten_proto_check(so);
381 	INP_HASH_WLOCK(&V_tcbinfo);
382 	if (error == 0 && inp->inp_lport == 0)
383 		error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
384 	INP_HASH_WUNLOCK(&V_tcbinfo);
385 	if (error == 0) {
386 		tcp_state_change(tp, TCPS_LISTEN);
387 		solisten_proto(so, backlog);
388 #ifdef TCP_OFFLOAD
389 		if ((so->so_options & SO_NO_OFFLOAD) == 0)
390 			tcp_offload_listen_start(tp);
391 #endif
392 	}
393 	SOCK_UNLOCK(so);
394 
395 out:
396 	TCPDEBUG2(PRU_LISTEN);
397 	INP_WUNLOCK(inp);
398 	return (error);
399 }
400 #endif /* INET */
401 
402 #ifdef INET6
403 static int
404 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
405 {
406 	int error = 0;
407 	struct inpcb *inp;
408 	struct tcpcb *tp = NULL;
409 
410 	TCPDEBUG0;
411 	inp = sotoinpcb(so);
412 	KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
413 	INP_WLOCK(inp);
414 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
415 		error = EINVAL;
416 		goto out;
417 	}
418 	tp = intotcpcb(inp);
419 	TCPDEBUG1();
420 	SOCK_LOCK(so);
421 	error = solisten_proto_check(so);
422 	INP_HASH_WLOCK(&V_tcbinfo);
423 	if (error == 0 && inp->inp_lport == 0) {
424 		inp->inp_vflag &= ~INP_IPV4;
425 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
426 			inp->inp_vflag |= INP_IPV4;
427 		error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
428 	}
429 	INP_HASH_WUNLOCK(&V_tcbinfo);
430 	if (error == 0) {
431 		tcp_state_change(tp, TCPS_LISTEN);
432 		solisten_proto(so, backlog);
433 #ifdef TCP_OFFLOAD
434 		if ((so->so_options & SO_NO_OFFLOAD) == 0)
435 			tcp_offload_listen_start(tp);
436 #endif
437 	}
438 	SOCK_UNLOCK(so);
439 
440 out:
441 	TCPDEBUG2(PRU_LISTEN);
442 	INP_WUNLOCK(inp);
443 	return (error);
444 }
445 #endif /* INET6 */
446 
447 #ifdef INET
448 /*
449  * Initiate connection to peer.
450  * Create a template for use in transmissions on this connection.
451  * Enter SYN_SENT state, and mark socket as connecting.
452  * Start keep-alive timer, and seed output sequence space.
453  * Send initial segment on connection.
454  */
455 static int
456 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
457 {
458 	int error = 0;
459 	struct inpcb *inp;
460 	struct tcpcb *tp = NULL;
461 	struct sockaddr_in *sinp;
462 
463 	sinp = (struct sockaddr_in *)nam;
464 	if (nam->sa_len != sizeof (*sinp))
465 		return (EINVAL);
466 	/*
467 	 * Must disallow TCP ``connections'' to multicast addresses.
468 	 */
469 	if (sinp->sin_family == AF_INET
470 	    && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
471 		return (EAFNOSUPPORT);
472 	if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
473 		return (error);
474 
475 	TCPDEBUG0;
476 	inp = sotoinpcb(so);
477 	KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
478 	INP_WLOCK(inp);
479 	if (inp->inp_flags & INP_TIMEWAIT) {
480 		error = EADDRINUSE;
481 		goto out;
482 	}
483 	if (inp->inp_flags & INP_DROPPED) {
484 		error = ECONNREFUSED;
485 		goto out;
486 	}
487 	tp = intotcpcb(inp);
488 	TCPDEBUG1();
489 	if ((error = tcp_connect(tp, nam, td)) != 0)
490 		goto out;
491 #ifdef TCP_OFFLOAD
492 	if (registered_toedevs > 0 &&
493 	    (so->so_options & SO_NO_OFFLOAD) == 0 &&
494 	    (error = tcp_offload_connect(so, nam)) == 0)
495 		goto out;
496 #endif
497 	tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
498 	error = tcp_output(tp);
499 out:
500 	TCPDEBUG2(PRU_CONNECT);
501 	INP_WUNLOCK(inp);
502 	return (error);
503 }
504 #endif /* INET */
505 
506 #ifdef INET6
507 static int
508 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
509 {
510 	int error = 0;
511 	struct inpcb *inp;
512 	struct tcpcb *tp = NULL;
513 	struct sockaddr_in6 *sin6p;
514 
515 	TCPDEBUG0;
516 
517 	sin6p = (struct sockaddr_in6 *)nam;
518 	if (nam->sa_len != sizeof (*sin6p))
519 		return (EINVAL);
520 	/*
521 	 * Must disallow TCP ``connections'' to multicast addresses.
522 	 */
523 	if (sin6p->sin6_family == AF_INET6
524 	    && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr))
525 		return (EAFNOSUPPORT);
526 
527 	inp = sotoinpcb(so);
528 	KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
529 	INP_WLOCK(inp);
530 	if (inp->inp_flags & INP_TIMEWAIT) {
531 		error = EADDRINUSE;
532 		goto out;
533 	}
534 	if (inp->inp_flags & INP_DROPPED) {
535 		error = ECONNREFUSED;
536 		goto out;
537 	}
538 	tp = intotcpcb(inp);
539 	TCPDEBUG1();
540 #ifdef INET
541 	/*
542 	 * XXXRW: Some confusion: V4/V6 flags relate to binding, and
543 	 * therefore probably require the hash lock, which isn't held here.
544 	 * Is this a significant problem?
545 	 */
546 	if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
547 		struct sockaddr_in sin;
548 
549 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
550 			error = EINVAL;
551 			goto out;
552 		}
553 
554 		in6_sin6_2_sin(&sin, sin6p);
555 		inp->inp_vflag |= INP_IPV4;
556 		inp->inp_vflag &= ~INP_IPV6;
557 		if ((error = prison_remote_ip4(td->td_ucred,
558 		    &sin.sin_addr)) != 0)
559 			goto out;
560 		if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
561 			goto out;
562 #ifdef TCP_OFFLOAD
563 		if (registered_toedevs > 0 &&
564 		    (so->so_options & SO_NO_OFFLOAD) == 0 &&
565 		    (error = tcp_offload_connect(so, nam)) == 0)
566 			goto out;
567 #endif
568 		error = tcp_output(tp);
569 		goto out;
570 	}
571 #endif
572 	inp->inp_vflag &= ~INP_IPV4;
573 	inp->inp_vflag |= INP_IPV6;
574 	inp->inp_inc.inc_flags |= INC_ISIPV6;
575 	if ((error = prison_remote_ip6(td->td_ucred, &sin6p->sin6_addr)) != 0)
576 		goto out;
577 	if ((error = tcp6_connect(tp, nam, td)) != 0)
578 		goto out;
579 #ifdef TCP_OFFLOAD
580 	if (registered_toedevs > 0 &&
581 	    (so->so_options & SO_NO_OFFLOAD) == 0 &&
582 	    (error = tcp_offload_connect(so, nam)) == 0)
583 		goto out;
584 #endif
585 	tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
586 	error = tcp_output(tp);
587 
588 out:
589 	TCPDEBUG2(PRU_CONNECT);
590 	INP_WUNLOCK(inp);
591 	return (error);
592 }
593 #endif /* INET6 */
594 
595 /*
596  * Initiate disconnect from peer.
597  * If connection never passed embryonic stage, just drop;
598  * else if don't need to let data drain, then can just drop anyways,
599  * else have to begin TCP shutdown process: mark socket disconnecting,
600  * drain unread data, state switch to reflect user close, and
601  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
602  * when peer sends FIN and acks ours.
603  *
604  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
605  */
606 static int
607 tcp_usr_disconnect(struct socket *so)
608 {
609 	struct inpcb *inp;
610 	struct tcpcb *tp = NULL;
611 	int error = 0;
612 
613 	TCPDEBUG0;
614 	INP_INFO_WLOCK(&V_tcbinfo);
615 	inp = sotoinpcb(so);
616 	KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
617 	INP_WLOCK(inp);
618 	if (inp->inp_flags & INP_TIMEWAIT)
619 		goto out;
620 	if (inp->inp_flags & INP_DROPPED) {
621 		error = ECONNRESET;
622 		goto out;
623 	}
624 	tp = intotcpcb(inp);
625 	TCPDEBUG1();
626 	tcp_disconnect(tp);
627 out:
628 	TCPDEBUG2(PRU_DISCONNECT);
629 	INP_WUNLOCK(inp);
630 	INP_INFO_WUNLOCK(&V_tcbinfo);
631 	return (error);
632 }
633 
634 #ifdef INET
635 /*
636  * Accept a connection.  Essentially all the work is done at higher levels;
637  * just return the address of the peer, storing through addr.
638  */
639 static int
640 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
641 {
642 	int error = 0;
643 	struct inpcb *inp = NULL;
644 	struct tcpcb *tp = NULL;
645 	struct in_addr addr;
646 	in_port_t port = 0;
647 	TCPDEBUG0;
648 
649 	if (so->so_state & SS_ISDISCONNECTED)
650 		return (ECONNABORTED);
651 
652 	inp = sotoinpcb(so);
653 	KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
654 	INP_WLOCK(inp);
655 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
656 		error = ECONNABORTED;
657 		goto out;
658 	}
659 	tp = intotcpcb(inp);
660 	TCPDEBUG1();
661 
662 	/*
663 	 * We inline in_getpeeraddr and COMMON_END here, so that we can
664 	 * copy the data of interest and defer the malloc until after we
665 	 * release the lock.
666 	 */
667 	port = inp->inp_fport;
668 	addr = inp->inp_faddr;
669 
670 out:
671 	TCPDEBUG2(PRU_ACCEPT);
672 	INP_WUNLOCK(inp);
673 	if (error == 0)
674 		*nam = in_sockaddr(port, &addr);
675 	return error;
676 }
677 #endif /* INET */
678 
679 #ifdef INET6
680 static int
681 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
682 {
683 	struct inpcb *inp = NULL;
684 	int error = 0;
685 	struct tcpcb *tp = NULL;
686 	struct in_addr addr;
687 	struct in6_addr addr6;
688 	in_port_t port = 0;
689 	int v4 = 0;
690 	TCPDEBUG0;
691 
692 	if (so->so_state & SS_ISDISCONNECTED)
693 		return (ECONNABORTED);
694 
695 	inp = sotoinpcb(so);
696 	KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
697 	INP_INFO_RLOCK(&V_tcbinfo);
698 	INP_WLOCK(inp);
699 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
700 		error = ECONNABORTED;
701 		goto out;
702 	}
703 	tp = intotcpcb(inp);
704 	TCPDEBUG1();
705 
706 	/*
707 	 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
708 	 * copy the data of interest and defer the malloc until after we
709 	 * release the lock.
710 	 */
711 	if (inp->inp_vflag & INP_IPV4) {
712 		v4 = 1;
713 		port = inp->inp_fport;
714 		addr = inp->inp_faddr;
715 	} else {
716 		port = inp->inp_fport;
717 		addr6 = inp->in6p_faddr;
718 	}
719 
720 out:
721 	TCPDEBUG2(PRU_ACCEPT);
722 	INP_WUNLOCK(inp);
723 	INP_INFO_RUNLOCK(&V_tcbinfo);
724 	if (error == 0) {
725 		if (v4)
726 			*nam = in6_v4mapsin6_sockaddr(port, &addr);
727 		else
728 			*nam = in6_sockaddr(port, &addr6);
729 	}
730 	return error;
731 }
732 #endif /* INET6 */
733 
734 /*
735  * Mark the connection as being incapable of further output.
736  */
737 static int
738 tcp_usr_shutdown(struct socket *so)
739 {
740 	int error = 0;
741 	struct inpcb *inp;
742 	struct tcpcb *tp = NULL;
743 
744 	TCPDEBUG0;
745 	INP_INFO_WLOCK(&V_tcbinfo);
746 	inp = sotoinpcb(so);
747 	KASSERT(inp != NULL, ("inp == NULL"));
748 	INP_WLOCK(inp);
749 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
750 		error = ECONNRESET;
751 		goto out;
752 	}
753 	tp = intotcpcb(inp);
754 	TCPDEBUG1();
755 	socantsendmore(so);
756 	tcp_usrclosed(tp);
757 	if (!(inp->inp_flags & INP_DROPPED))
758 		error = tcp_output(tp);
759 
760 out:
761 	TCPDEBUG2(PRU_SHUTDOWN);
762 	INP_WUNLOCK(inp);
763 	INP_INFO_WUNLOCK(&V_tcbinfo);
764 
765 	return (error);
766 }
767 
768 /*
769  * After a receive, possibly send window update to peer.
770  */
771 static int
772 tcp_usr_rcvd(struct socket *so, int flags)
773 {
774 	struct inpcb *inp;
775 	struct tcpcb *tp = NULL;
776 	int error = 0;
777 
778 	TCPDEBUG0;
779 	inp = sotoinpcb(so);
780 	KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
781 	INP_WLOCK(inp);
782 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
783 		error = ECONNRESET;
784 		goto out;
785 	}
786 	tp = intotcpcb(inp);
787 	TCPDEBUG1();
788 #ifdef TCP_OFFLOAD
789 	if (tp->t_flags & TF_TOE)
790 		tcp_offload_rcvd(tp);
791 	else
792 #endif
793 	tcp_output(tp);
794 
795 out:
796 	TCPDEBUG2(PRU_RCVD);
797 	INP_WUNLOCK(inp);
798 	return (error);
799 }
800 
801 /*
802  * Do a send by putting data in output queue and updating urgent
803  * marker if URG set.  Possibly send more data.  Unlike the other
804  * pru_*() routines, the mbuf chains are our responsibility.  We
805  * must either enqueue them or free them.  The other pru_* routines
806  * generally are caller-frees.
807  */
808 static int
809 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
810     struct sockaddr *nam, struct mbuf *control, struct thread *td)
811 {
812 	int error = 0;
813 	struct inpcb *inp;
814 	struct tcpcb *tp = NULL;
815 #ifdef INET6
816 	int isipv6;
817 #endif
818 	TCPDEBUG0;
819 
820 	/*
821 	 * We require the pcbinfo lock if we will close the socket as part of
822 	 * this call.
823 	 */
824 	if (flags & PRUS_EOF)
825 		INP_INFO_WLOCK(&V_tcbinfo);
826 	inp = sotoinpcb(so);
827 	KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
828 	INP_WLOCK(inp);
829 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
830 		if (control)
831 			m_freem(control);
832 		/*
833 		 * In case of PRUS_NOTREADY, tcp_usr_ready() is responsible
834 		 * for freeing memory.
835 		 */
836 		if (m && (flags & PRUS_NOTREADY) == 0)
837 			m_freem(m);
838 		error = ECONNRESET;
839 		goto out;
840 	}
841 #ifdef INET6
842 	isipv6 = nam && nam->sa_family == AF_INET6;
843 #endif /* INET6 */
844 	tp = intotcpcb(inp);
845 	TCPDEBUG1();
846 	if (control) {
847 		/* TCP doesn't do control messages (rights, creds, etc) */
848 		if (control->m_len) {
849 			m_freem(control);
850 			if (m)
851 				m_freem(m);
852 			error = EINVAL;
853 			goto out;
854 		}
855 		m_freem(control);	/* empty control, just free it */
856 	}
857 	if (!(flags & PRUS_OOB)) {
858 		sbappendstream(&so->so_snd, m, flags);
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 #ifdef INET6
867 			if (isipv6)
868 				error = tcp6_connect(tp, nam, td);
869 #endif /* INET6 */
870 #if defined(INET6) && defined(INET)
871 			else
872 #endif
873 #ifdef INET
874 				error = tcp_connect(tp, nam, td);
875 #endif
876 			if (error)
877 				goto out;
878 			tp->snd_wnd = TTCP_CLIENT_SND_WND;
879 			tcp_mss(tp, -1);
880 		}
881 		if (flags & PRUS_EOF) {
882 			/*
883 			 * Close the send side of the connection after
884 			 * the data is sent.
885 			 */
886 			INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
887 			socantsendmore(so);
888 			tcp_usrclosed(tp);
889 		}
890 		if (!(inp->inp_flags & INP_DROPPED) &&
891 		    !(flags & PRUS_NOTREADY)) {
892 			if (flags & PRUS_MORETOCOME)
893 				tp->t_flags |= TF_MORETOCOME;
894 			error = tcp_output(tp);
895 			if (flags & PRUS_MORETOCOME)
896 				tp->t_flags &= ~TF_MORETOCOME;
897 		}
898 	} else {
899 		/*
900 		 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
901 		 */
902 		SOCKBUF_LOCK(&so->so_snd);
903 		if (sbspace(&so->so_snd) < -512) {
904 			SOCKBUF_UNLOCK(&so->so_snd);
905 			m_freem(m);
906 			error = ENOBUFS;
907 			goto out;
908 		}
909 		/*
910 		 * According to RFC961 (Assigned Protocols),
911 		 * the urgent pointer points to the last octet
912 		 * of urgent data.  We continue, however,
913 		 * to consider it to indicate the first octet
914 		 * of data past the urgent section.
915 		 * Otherwise, snd_up should be one lower.
916 		 */
917 		sbappendstream_locked(&so->so_snd, m, flags);
918 		SOCKBUF_UNLOCK(&so->so_snd);
919 		if (nam && tp->t_state < TCPS_SYN_SENT) {
920 			/*
921 			 * Do implied connect if not yet connected,
922 			 * initialize window to default value, and
923 			 * initialize maxseg/maxopd using peer's cached
924 			 * MSS.
925 			 */
926 #ifdef INET6
927 			if (isipv6)
928 				error = tcp6_connect(tp, nam, td);
929 #endif /* INET6 */
930 #if defined(INET6) && defined(INET)
931 			else
932 #endif
933 #ifdef INET
934 				error = tcp_connect(tp, nam, td);
935 #endif
936 			if (error)
937 				goto out;
938 			tp->snd_wnd = TTCP_CLIENT_SND_WND;
939 			tcp_mss(tp, -1);
940 		}
941 		tp->snd_up = tp->snd_una + sbavail(&so->so_snd);
942 		if (!(flags & PRUS_NOTREADY)) {
943 			tp->t_flags |= TF_FORCEDATA;
944 			error = tcp_output(tp);
945 			tp->t_flags &= ~TF_FORCEDATA;
946 		}
947 	}
948 out:
949 	TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
950 		  ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
951 	INP_WUNLOCK(inp);
952 	if (flags & PRUS_EOF)
953 		INP_INFO_WUNLOCK(&V_tcbinfo);
954 	return (error);
955 }
956 
957 static int
958 tcp_usr_ready(struct socket *so, struct mbuf *m, int count)
959 {
960 	struct inpcb *inp;
961 	struct tcpcb *tp;
962 	int error;
963 
964 	inp = sotoinpcb(so);
965 	INP_WLOCK(inp);
966 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
967 		INP_WUNLOCK(inp);
968 		for (int i = 0; i < count; i++)
969 			m = m_free(m);
970 		return (ECONNRESET);
971 	}
972 	tp = intotcpcb(inp);
973 
974 	SOCKBUF_LOCK(&so->so_snd);
975 	error = sbready(&so->so_snd, m, count);
976 	SOCKBUF_UNLOCK(&so->so_snd);
977 	if (error == 0)
978 		error = tcp_output(tp);
979 	INP_WUNLOCK(inp);
980 
981 	return (error);
982 }
983 
984 /*
985  * Abort the TCP.  Drop the connection abruptly.
986  */
987 static void
988 tcp_usr_abort(struct socket *so)
989 {
990 	struct inpcb *inp;
991 	struct tcpcb *tp = NULL;
992 	TCPDEBUG0;
993 
994 	inp = sotoinpcb(so);
995 	KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
996 
997 	INP_INFO_WLOCK(&V_tcbinfo);
998 	INP_WLOCK(inp);
999 	KASSERT(inp->inp_socket != NULL,
1000 	    ("tcp_usr_abort: inp_socket == NULL"));
1001 
1002 	/*
1003 	 * If we still have full TCP state, and we're not dropped, drop.
1004 	 */
1005 	if (!(inp->inp_flags & INP_TIMEWAIT) &&
1006 	    !(inp->inp_flags & INP_DROPPED)) {
1007 		tp = intotcpcb(inp);
1008 		TCPDEBUG1();
1009 		tcp_drop(tp, ECONNABORTED);
1010 		TCPDEBUG2(PRU_ABORT);
1011 	}
1012 	if (!(inp->inp_flags & INP_DROPPED)) {
1013 		SOCK_LOCK(so);
1014 		so->so_state |= SS_PROTOREF;
1015 		SOCK_UNLOCK(so);
1016 		inp->inp_flags |= INP_SOCKREF;
1017 	}
1018 	INP_WUNLOCK(inp);
1019 	INP_INFO_WUNLOCK(&V_tcbinfo);
1020 }
1021 
1022 /*
1023  * TCP socket is closed.  Start friendly disconnect.
1024  */
1025 static void
1026 tcp_usr_close(struct socket *so)
1027 {
1028 	struct inpcb *inp;
1029 	struct tcpcb *tp = NULL;
1030 	TCPDEBUG0;
1031 
1032 	inp = sotoinpcb(so);
1033 	KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
1034 
1035 	INP_INFO_WLOCK(&V_tcbinfo);
1036 	INP_WLOCK(inp);
1037 	KASSERT(inp->inp_socket != NULL,
1038 	    ("tcp_usr_close: inp_socket == NULL"));
1039 
1040 	/*
1041 	 * If we still have full TCP state, and we're not dropped, initiate
1042 	 * a disconnect.
1043 	 */
1044 	if (!(inp->inp_flags & INP_TIMEWAIT) &&
1045 	    !(inp->inp_flags & INP_DROPPED)) {
1046 		tp = intotcpcb(inp);
1047 		TCPDEBUG1();
1048 		tcp_disconnect(tp);
1049 		TCPDEBUG2(PRU_CLOSE);
1050 	}
1051 	if (!(inp->inp_flags & INP_DROPPED)) {
1052 		SOCK_LOCK(so);
1053 		so->so_state |= SS_PROTOREF;
1054 		SOCK_UNLOCK(so);
1055 		inp->inp_flags |= INP_SOCKREF;
1056 	}
1057 	INP_WUNLOCK(inp);
1058 	INP_INFO_WUNLOCK(&V_tcbinfo);
1059 }
1060 
1061 /*
1062  * Receive out-of-band data.
1063  */
1064 static int
1065 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1066 {
1067 	int error = 0;
1068 	struct inpcb *inp;
1069 	struct tcpcb *tp = NULL;
1070 
1071 	TCPDEBUG0;
1072 	inp = sotoinpcb(so);
1073 	KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1074 	INP_WLOCK(inp);
1075 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1076 		error = ECONNRESET;
1077 		goto out;
1078 	}
1079 	tp = intotcpcb(inp);
1080 	TCPDEBUG1();
1081 	if ((so->so_oobmark == 0 &&
1082 	     (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1083 	    so->so_options & SO_OOBINLINE ||
1084 	    tp->t_oobflags & TCPOOB_HADDATA) {
1085 		error = EINVAL;
1086 		goto out;
1087 	}
1088 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1089 		error = EWOULDBLOCK;
1090 		goto out;
1091 	}
1092 	m->m_len = 1;
1093 	*mtod(m, caddr_t) = tp->t_iobc;
1094 	if ((flags & MSG_PEEK) == 0)
1095 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1096 
1097 out:
1098 	TCPDEBUG2(PRU_RCVOOB);
1099 	INP_WUNLOCK(inp);
1100 	return (error);
1101 }
1102 
1103 #ifdef INET
1104 struct pr_usrreqs tcp_usrreqs = {
1105 	.pru_abort =		tcp_usr_abort,
1106 	.pru_accept =		tcp_usr_accept,
1107 	.pru_attach =		tcp_usr_attach,
1108 	.pru_bind =		tcp_usr_bind,
1109 	.pru_connect =		tcp_usr_connect,
1110 	.pru_control =		in_control,
1111 	.pru_detach =		tcp_usr_detach,
1112 	.pru_disconnect =	tcp_usr_disconnect,
1113 	.pru_listen =		tcp_usr_listen,
1114 	.pru_peeraddr =		in_getpeeraddr,
1115 	.pru_rcvd =		tcp_usr_rcvd,
1116 	.pru_rcvoob =		tcp_usr_rcvoob,
1117 	.pru_send =		tcp_usr_send,
1118 	.pru_ready =		tcp_usr_ready,
1119 	.pru_shutdown =		tcp_usr_shutdown,
1120 	.pru_sockaddr =		in_getsockaddr,
1121 	.pru_sosetlabel =	in_pcbsosetlabel,
1122 	.pru_close =		tcp_usr_close,
1123 };
1124 #endif /* INET */
1125 
1126 #ifdef INET6
1127 struct pr_usrreqs tcp6_usrreqs = {
1128 	.pru_abort =		tcp_usr_abort,
1129 	.pru_accept =		tcp6_usr_accept,
1130 	.pru_attach =		tcp_usr_attach,
1131 	.pru_bind =		tcp6_usr_bind,
1132 	.pru_connect =		tcp6_usr_connect,
1133 	.pru_control =		in6_control,
1134 	.pru_detach =		tcp_usr_detach,
1135 	.pru_disconnect =	tcp_usr_disconnect,
1136 	.pru_listen =		tcp6_usr_listen,
1137 	.pru_peeraddr =		in6_mapped_peeraddr,
1138 	.pru_rcvd =		tcp_usr_rcvd,
1139 	.pru_rcvoob =		tcp_usr_rcvoob,
1140 	.pru_send =		tcp_usr_send,
1141 	.pru_ready =		tcp_usr_ready,
1142 	.pru_shutdown =		tcp_usr_shutdown,
1143 	.pru_sockaddr =		in6_mapped_sockaddr,
1144 	.pru_sosetlabel =	in_pcbsosetlabel,
1145 	.pru_close =		tcp_usr_close,
1146 };
1147 #endif /* INET6 */
1148 
1149 #ifdef INET
1150 /*
1151  * Common subroutine to open a TCP connection to remote host specified
1152  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
1153  * port number if needed.  Call in_pcbconnect_setup to do the routing and
1154  * to choose a local host address (interface).  If there is an existing
1155  * incarnation of the same connection in TIME-WAIT state and if the remote
1156  * host was sending CC options and if the connection duration was < MSL, then
1157  * truncate the previous TIME-WAIT state and proceed.
1158  * Initialize connection parameters and enter SYN-SENT state.
1159  */
1160 static int
1161 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1162 {
1163 	struct inpcb *inp = tp->t_inpcb, *oinp;
1164 	struct socket *so = inp->inp_socket;
1165 	struct in_addr laddr;
1166 	u_short lport;
1167 	int error;
1168 
1169 	INP_WLOCK_ASSERT(inp);
1170 	INP_HASH_WLOCK(&V_tcbinfo);
1171 
1172 	if (inp->inp_lport == 0) {
1173 		error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1174 		if (error)
1175 			goto out;
1176 	}
1177 
1178 	/*
1179 	 * Cannot simply call in_pcbconnect, because there might be an
1180 	 * earlier incarnation of this same connection still in
1181 	 * TIME_WAIT state, creating an ADDRINUSE error.
1182 	 */
1183 	laddr = inp->inp_laddr;
1184 	lport = inp->inp_lport;
1185 	error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1186 	    &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1187 	if (error && oinp == NULL)
1188 		goto out;
1189 	if (oinp) {
1190 		error = EADDRINUSE;
1191 		goto out;
1192 	}
1193 	inp->inp_laddr = laddr;
1194 	in_pcbrehash(inp);
1195 	INP_HASH_WUNLOCK(&V_tcbinfo);
1196 
1197 	/*
1198 	 * Compute window scaling to request:
1199 	 * Scale to fit into sweet spot.  See tcp_syncache.c.
1200 	 * XXX: This should move to tcp_output().
1201 	 */
1202 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1203 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1204 		tp->request_r_scale++;
1205 
1206 	soisconnecting(so);
1207 	TCPSTAT_INC(tcps_connattempt);
1208 	tcp_state_change(tp, TCPS_SYN_SENT);
1209 	tp->iss = tcp_new_isn(tp);
1210 	tcp_sendseqinit(tp);
1211 
1212 	return 0;
1213 
1214 out:
1215 	INP_HASH_WUNLOCK(&V_tcbinfo);
1216 	return (error);
1217 }
1218 #endif /* INET */
1219 
1220 #ifdef INET6
1221 static int
1222 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1223 {
1224 	struct inpcb *inp = tp->t_inpcb;
1225 	int error;
1226 
1227 	INP_WLOCK_ASSERT(inp);
1228 	INP_HASH_WLOCK(&V_tcbinfo);
1229 
1230 	if (inp->inp_lport == 0) {
1231 		error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1232 		if (error)
1233 			goto out;
1234 	}
1235 	error = in6_pcbconnect(inp, nam, td->td_ucred);
1236 	if (error != 0)
1237 		goto out;
1238 	INP_HASH_WUNLOCK(&V_tcbinfo);
1239 
1240 	/* Compute window scaling to request.  */
1241 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1242 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1243 		tp->request_r_scale++;
1244 
1245 	soisconnecting(inp->inp_socket);
1246 	TCPSTAT_INC(tcps_connattempt);
1247 	tcp_state_change(tp, TCPS_SYN_SENT);
1248 	tp->iss = tcp_new_isn(tp);
1249 	tcp_sendseqinit(tp);
1250 
1251 	return 0;
1252 
1253 out:
1254 	INP_HASH_WUNLOCK(&V_tcbinfo);
1255 	return error;
1256 }
1257 #endif /* INET6 */
1258 
1259 /*
1260  * Export TCP internal state information via a struct tcp_info, based on the
1261  * Linux 2.6 API.  Not ABI compatible as our constants are mapped differently
1262  * (TCP state machine, etc).  We export all information using FreeBSD-native
1263  * constants -- for example, the numeric values for tcpi_state will differ
1264  * from Linux.
1265  */
1266 static void
1267 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1268 {
1269 
1270 	INP_WLOCK_ASSERT(tp->t_inpcb);
1271 	bzero(ti, sizeof(*ti));
1272 
1273 	ti->tcpi_state = tp->t_state;
1274 	if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1275 		ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1276 	if (tp->t_flags & TF_SACK_PERMIT)
1277 		ti->tcpi_options |= TCPI_OPT_SACK;
1278 	if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1279 		ti->tcpi_options |= TCPI_OPT_WSCALE;
1280 		ti->tcpi_snd_wscale = tp->snd_scale;
1281 		ti->tcpi_rcv_wscale = tp->rcv_scale;
1282 	}
1283 
1284 	ti->tcpi_rto = tp->t_rxtcur * tick;
1285 	ti->tcpi_last_data_recv = (long)(ticks - (int)tp->t_rcvtime) * tick;
1286 	ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1287 	ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1288 
1289 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1290 	ti->tcpi_snd_cwnd = tp->snd_cwnd;
1291 
1292 	/*
1293 	 * FreeBSD-specific extension fields for tcp_info.
1294 	 */
1295 	ti->tcpi_rcv_space = tp->rcv_wnd;
1296 	ti->tcpi_rcv_nxt = tp->rcv_nxt;
1297 	ti->tcpi_snd_wnd = tp->snd_wnd;
1298 	ti->tcpi_snd_bwnd = 0;		/* Unused, kept for compat. */
1299 	ti->tcpi_snd_nxt = tp->snd_nxt;
1300 	ti->tcpi_snd_mss = tp->t_maxseg;
1301 	ti->tcpi_rcv_mss = tp->t_maxseg;
1302 	if (tp->t_flags & TF_TOE)
1303 		ti->tcpi_options |= TCPI_OPT_TOE;
1304 	ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1305 	ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1306 	ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1307 }
1308 
1309 /*
1310  * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1311  * socket option arguments.  When it re-acquires the lock after the copy, it
1312  * has to revalidate that the connection is still valid for the socket
1313  * option.
1314  */
1315 #define INP_WLOCK_RECHECK(inp) do {					\
1316 	INP_WLOCK(inp);							\
1317 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {		\
1318 		INP_WUNLOCK(inp);					\
1319 		return (ECONNRESET);					\
1320 	}								\
1321 	tp = intotcpcb(inp);						\
1322 } while(0)
1323 
1324 int
1325 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1326 {
1327 	int	error, opt, optval;
1328 	u_int	ui;
1329 	struct	inpcb *inp;
1330 	struct	tcpcb *tp;
1331 	struct	tcp_info ti;
1332 	char buf[TCP_CA_NAME_MAX];
1333 	struct cc_algo *algo;
1334 
1335 	error = 0;
1336 	inp = sotoinpcb(so);
1337 	KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1338 	INP_WLOCK(inp);
1339 	if (sopt->sopt_level != IPPROTO_TCP) {
1340 #ifdef INET6
1341 		if (inp->inp_vflag & INP_IPV6PROTO) {
1342 			INP_WUNLOCK(inp);
1343 			error = ip6_ctloutput(so, sopt);
1344 		}
1345 #endif /* INET6 */
1346 #if defined(INET6) && defined(INET)
1347 		else
1348 #endif
1349 #ifdef INET
1350 		{
1351 			INP_WUNLOCK(inp);
1352 			error = ip_ctloutput(so, sopt);
1353 		}
1354 #endif
1355 		return (error);
1356 	}
1357 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1358 		INP_WUNLOCK(inp);
1359 		return (ECONNRESET);
1360 	}
1361 
1362 	switch (sopt->sopt_dir) {
1363 	case SOPT_SET:
1364 		switch (sopt->sopt_name) {
1365 #ifdef TCP_SIGNATURE
1366 		case TCP_MD5SIG:
1367 			INP_WUNLOCK(inp);
1368 			error = sooptcopyin(sopt, &optval, sizeof optval,
1369 			    sizeof optval);
1370 			if (error)
1371 				return (error);
1372 
1373 			INP_WLOCK_RECHECK(inp);
1374 			if (optval > 0)
1375 				tp->t_flags |= TF_SIGNATURE;
1376 			else
1377 				tp->t_flags &= ~TF_SIGNATURE;
1378 			goto unlock_and_done;
1379 #endif /* TCP_SIGNATURE */
1380 
1381 		case TCP_NODELAY:
1382 		case TCP_NOOPT:
1383 			INP_WUNLOCK(inp);
1384 			error = sooptcopyin(sopt, &optval, sizeof optval,
1385 			    sizeof optval);
1386 			if (error)
1387 				return (error);
1388 
1389 			INP_WLOCK_RECHECK(inp);
1390 			switch (sopt->sopt_name) {
1391 			case TCP_NODELAY:
1392 				opt = TF_NODELAY;
1393 				break;
1394 			case TCP_NOOPT:
1395 				opt = TF_NOOPT;
1396 				break;
1397 			default:
1398 				opt = 0; /* dead code to fool gcc */
1399 				break;
1400 			}
1401 
1402 			if (optval)
1403 				tp->t_flags |= opt;
1404 			else
1405 				tp->t_flags &= ~opt;
1406 unlock_and_done:
1407 #ifdef TCP_OFFLOAD
1408 			if (tp->t_flags & TF_TOE) {
1409 				tcp_offload_ctloutput(tp, sopt->sopt_dir,
1410 				    sopt->sopt_name);
1411 			}
1412 #endif
1413 			INP_WUNLOCK(inp);
1414 			break;
1415 
1416 		case TCP_NOPUSH:
1417 			INP_WUNLOCK(inp);
1418 			error = sooptcopyin(sopt, &optval, sizeof optval,
1419 			    sizeof optval);
1420 			if (error)
1421 				return (error);
1422 
1423 			INP_WLOCK_RECHECK(inp);
1424 			if (optval)
1425 				tp->t_flags |= TF_NOPUSH;
1426 			else if (tp->t_flags & TF_NOPUSH) {
1427 				tp->t_flags &= ~TF_NOPUSH;
1428 				if (TCPS_HAVEESTABLISHED(tp->t_state))
1429 					error = tcp_output(tp);
1430 			}
1431 			goto unlock_and_done;
1432 
1433 		case TCP_MAXSEG:
1434 			INP_WUNLOCK(inp);
1435 			error = sooptcopyin(sopt, &optval, sizeof optval,
1436 			    sizeof optval);
1437 			if (error)
1438 				return (error);
1439 
1440 			INP_WLOCK_RECHECK(inp);
1441 			if (optval > 0 && optval <= tp->t_maxseg &&
1442 			    optval + 40 >= V_tcp_minmss)
1443 				tp->t_maxseg = optval;
1444 			else
1445 				error = EINVAL;
1446 			goto unlock_and_done;
1447 
1448 		case TCP_INFO:
1449 			INP_WUNLOCK(inp);
1450 			error = EINVAL;
1451 			break;
1452 
1453 		case TCP_CONGESTION:
1454 			INP_WUNLOCK(inp);
1455 			bzero(buf, sizeof(buf));
1456 			error = sooptcopyin(sopt, &buf, sizeof(buf), 1);
1457 			if (error)
1458 				break;
1459 			INP_WLOCK_RECHECK(inp);
1460 			/*
1461 			 * Return EINVAL if we can't find the requested cc algo.
1462 			 */
1463 			error = EINVAL;
1464 			CC_LIST_RLOCK();
1465 			STAILQ_FOREACH(algo, &cc_list, entries) {
1466 				if (strncmp(buf, algo->name, TCP_CA_NAME_MAX)
1467 				    == 0) {
1468 					/* We've found the requested algo. */
1469 					error = 0;
1470 					/*
1471 					 * We hold a write lock over the tcb
1472 					 * so it's safe to do these things
1473 					 * without ordering concerns.
1474 					 */
1475 					if (CC_ALGO(tp)->cb_destroy != NULL)
1476 						CC_ALGO(tp)->cb_destroy(tp->ccv);
1477 					CC_ALGO(tp) = algo;
1478 					/*
1479 					 * If something goes pear shaped
1480 					 * initialising the new algo,
1481 					 * fall back to newreno (which
1482 					 * does not require initialisation).
1483 					 */
1484 					if (algo->cb_init != NULL)
1485 						if (algo->cb_init(tp->ccv) > 0) {
1486 							CC_ALGO(tp) = &newreno_cc_algo;
1487 							/*
1488 							 * The only reason init
1489 							 * should fail is
1490 							 * because of malloc.
1491 							 */
1492 							error = ENOMEM;
1493 						}
1494 					break; /* Break the STAILQ_FOREACH. */
1495 				}
1496 			}
1497 			CC_LIST_RUNLOCK();
1498 			goto unlock_and_done;
1499 
1500 		case TCP_KEEPIDLE:
1501 		case TCP_KEEPINTVL:
1502 		case TCP_KEEPINIT:
1503 			INP_WUNLOCK(inp);
1504 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1505 			if (error)
1506 				return (error);
1507 
1508 			if (ui > (UINT_MAX / hz)) {
1509 				error = EINVAL;
1510 				break;
1511 			}
1512 			ui *= hz;
1513 
1514 			INP_WLOCK_RECHECK(inp);
1515 			switch (sopt->sopt_name) {
1516 			case TCP_KEEPIDLE:
1517 				tp->t_keepidle = ui;
1518 				/*
1519 				 * XXX: better check current remaining
1520 				 * timeout and "merge" it with new value.
1521 				 */
1522 				if ((tp->t_state > TCPS_LISTEN) &&
1523 				    (tp->t_state <= TCPS_CLOSING))
1524 					tcp_timer_activate(tp, TT_KEEP,
1525 					    TP_KEEPIDLE(tp));
1526 				break;
1527 			case TCP_KEEPINTVL:
1528 				tp->t_keepintvl = ui;
1529 				if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1530 				    (TP_MAXIDLE(tp) > 0))
1531 					tcp_timer_activate(tp, TT_2MSL,
1532 					    TP_MAXIDLE(tp));
1533 				break;
1534 			case TCP_KEEPINIT:
1535 				tp->t_keepinit = ui;
1536 				if (tp->t_state == TCPS_SYN_RECEIVED ||
1537 				    tp->t_state == TCPS_SYN_SENT)
1538 					tcp_timer_activate(tp, TT_KEEP,
1539 					    TP_KEEPINIT(tp));
1540 				break;
1541 			}
1542 			goto unlock_and_done;
1543 
1544 		case TCP_KEEPCNT:
1545 			INP_WUNLOCK(inp);
1546 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
1547 			if (error)
1548 				return (error);
1549 
1550 			INP_WLOCK_RECHECK(inp);
1551 			tp->t_keepcnt = ui;
1552 			if ((tp->t_state == TCPS_FIN_WAIT_2) &&
1553 			    (TP_MAXIDLE(tp) > 0))
1554 				tcp_timer_activate(tp, TT_2MSL,
1555 				    TP_MAXIDLE(tp));
1556 			goto unlock_and_done;
1557 
1558 		default:
1559 			INP_WUNLOCK(inp);
1560 			error = ENOPROTOOPT;
1561 			break;
1562 		}
1563 		break;
1564 
1565 	case SOPT_GET:
1566 		tp = intotcpcb(inp);
1567 		switch (sopt->sopt_name) {
1568 #ifdef TCP_SIGNATURE
1569 		case TCP_MD5SIG:
1570 			optval = (tp->t_flags & TF_SIGNATURE) ? 1 : 0;
1571 			INP_WUNLOCK(inp);
1572 			error = sooptcopyout(sopt, &optval, sizeof optval);
1573 			break;
1574 #endif
1575 
1576 		case TCP_NODELAY:
1577 			optval = tp->t_flags & TF_NODELAY;
1578 			INP_WUNLOCK(inp);
1579 			error = sooptcopyout(sopt, &optval, sizeof optval);
1580 			break;
1581 		case TCP_MAXSEG:
1582 			optval = tp->t_maxseg;
1583 			INP_WUNLOCK(inp);
1584 			error = sooptcopyout(sopt, &optval, sizeof optval);
1585 			break;
1586 		case TCP_NOOPT:
1587 			optval = tp->t_flags & TF_NOOPT;
1588 			INP_WUNLOCK(inp);
1589 			error = sooptcopyout(sopt, &optval, sizeof optval);
1590 			break;
1591 		case TCP_NOPUSH:
1592 			optval = tp->t_flags & TF_NOPUSH;
1593 			INP_WUNLOCK(inp);
1594 			error = sooptcopyout(sopt, &optval, sizeof optval);
1595 			break;
1596 		case TCP_INFO:
1597 			tcp_fill_info(tp, &ti);
1598 			INP_WUNLOCK(inp);
1599 			error = sooptcopyout(sopt, &ti, sizeof ti);
1600 			break;
1601 		case TCP_CONGESTION:
1602 			bzero(buf, sizeof(buf));
1603 			strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
1604 			INP_WUNLOCK(inp);
1605 			error = sooptcopyout(sopt, buf, TCP_CA_NAME_MAX);
1606 			break;
1607 		case TCP_KEEPIDLE:
1608 		case TCP_KEEPINTVL:
1609 		case TCP_KEEPINIT:
1610 		case TCP_KEEPCNT:
1611 			switch (sopt->sopt_name) {
1612 			case TCP_KEEPIDLE:
1613 				ui = tp->t_keepidle / hz;
1614 				break;
1615 			case TCP_KEEPINTVL:
1616 				ui = tp->t_keepintvl / hz;
1617 				break;
1618 			case TCP_KEEPINIT:
1619 				ui = tp->t_keepinit / hz;
1620 				break;
1621 			case TCP_KEEPCNT:
1622 				ui = tp->t_keepcnt;
1623 				break;
1624 			}
1625 			INP_WUNLOCK(inp);
1626 			error = sooptcopyout(sopt, &ui, sizeof(ui));
1627 			break;
1628 		default:
1629 			INP_WUNLOCK(inp);
1630 			error = ENOPROTOOPT;
1631 			break;
1632 		}
1633 		break;
1634 	}
1635 	return (error);
1636 }
1637 #undef INP_WLOCK_RECHECK
1638 
1639 /*
1640  * Attach TCP protocol to socket, allocating
1641  * internet protocol control block, tcp control block,
1642  * bufer space, and entering LISTEN state if to accept connections.
1643  */
1644 static int
1645 tcp_attach(struct socket *so)
1646 {
1647 	struct tcpcb *tp;
1648 	struct inpcb *inp;
1649 	int error;
1650 
1651 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
1652 		error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
1653 		if (error)
1654 			return (error);
1655 	}
1656 	so->so_rcv.sb_flags |= SB_AUTOSIZE;
1657 	so->so_snd.sb_flags |= SB_AUTOSIZE;
1658 	INP_INFO_WLOCK(&V_tcbinfo);
1659 	error = in_pcballoc(so, &V_tcbinfo);
1660 	if (error) {
1661 		INP_INFO_WUNLOCK(&V_tcbinfo);
1662 		return (error);
1663 	}
1664 	inp = sotoinpcb(so);
1665 #ifdef INET6
1666 	if (inp->inp_vflag & INP_IPV6PROTO) {
1667 		inp->inp_vflag |= INP_IPV6;
1668 		inp->in6p_hops = -1;	/* use kernel default */
1669 	}
1670 	else
1671 #endif
1672 	inp->inp_vflag |= INP_IPV4;
1673 	tp = tcp_newtcpcb(inp);
1674 	if (tp == NULL) {
1675 		in_pcbdetach(inp);
1676 		in_pcbfree(inp);
1677 		INP_INFO_WUNLOCK(&V_tcbinfo);
1678 		return (ENOBUFS);
1679 	}
1680 	tp->t_state = TCPS_CLOSED;
1681 	INP_WUNLOCK(inp);
1682 	INP_INFO_WUNLOCK(&V_tcbinfo);
1683 	return (0);
1684 }
1685 
1686 /*
1687  * Initiate (or continue) disconnect.
1688  * If embryonic state, just send reset (once).
1689  * If in ``let data drain'' option and linger null, just drop.
1690  * Otherwise (hard), mark socket disconnecting and drop
1691  * current input data; switch states based on user close, and
1692  * send segment to peer (with FIN).
1693  */
1694 static void
1695 tcp_disconnect(struct tcpcb *tp)
1696 {
1697 	struct inpcb *inp = tp->t_inpcb;
1698 	struct socket *so = inp->inp_socket;
1699 
1700 	INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1701 	INP_WLOCK_ASSERT(inp);
1702 
1703 	/*
1704 	 * Neither tcp_close() nor tcp_drop() should return NULL, as the
1705 	 * socket is still open.
1706 	 */
1707 	if (tp->t_state < TCPS_ESTABLISHED) {
1708 		tp = tcp_close(tp);
1709 		KASSERT(tp != NULL,
1710 		    ("tcp_disconnect: tcp_close() returned NULL"));
1711 	} else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
1712 		tp = tcp_drop(tp, 0);
1713 		KASSERT(tp != NULL,
1714 		    ("tcp_disconnect: tcp_drop() returned NULL"));
1715 	} else {
1716 		soisdisconnecting(so);
1717 		sbflush(&so->so_rcv);
1718 		tcp_usrclosed(tp);
1719 		if (!(inp->inp_flags & INP_DROPPED))
1720 			tcp_output(tp);
1721 	}
1722 }
1723 
1724 /*
1725  * User issued close, and wish to trail through shutdown states:
1726  * if never received SYN, just forget it.  If got a SYN from peer,
1727  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
1728  * If already got a FIN from peer, then almost done; go to LAST_ACK
1729  * state.  In all other cases, have already sent FIN to peer (e.g.
1730  * after PRU_SHUTDOWN), and just have to play tedious game waiting
1731  * for peer to send FIN or not respond to keep-alives, etc.
1732  * We can let the user exit from the close as soon as the FIN is acked.
1733  */
1734 static void
1735 tcp_usrclosed(struct tcpcb *tp)
1736 {
1737 
1738 	INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
1739 	INP_WLOCK_ASSERT(tp->t_inpcb);
1740 
1741 	switch (tp->t_state) {
1742 	case TCPS_LISTEN:
1743 #ifdef TCP_OFFLOAD
1744 		tcp_offload_listen_stop(tp);
1745 #endif
1746 		/* FALLTHROUGH */
1747 	case TCPS_CLOSED:
1748 		tcp_state_change(tp, TCPS_CLOSED);
1749 		tp = tcp_close(tp);
1750 		/*
1751 		 * tcp_close() should never return NULL here as the socket is
1752 		 * still open.
1753 		 */
1754 		KASSERT(tp != NULL,
1755 		    ("tcp_usrclosed: tcp_close() returned NULL"));
1756 		break;
1757 
1758 	case TCPS_SYN_SENT:
1759 	case TCPS_SYN_RECEIVED:
1760 		tp->t_flags |= TF_NEEDFIN;
1761 		break;
1762 
1763 	case TCPS_ESTABLISHED:
1764 		tcp_state_change(tp, TCPS_FIN_WAIT_1);
1765 		break;
1766 
1767 	case TCPS_CLOSE_WAIT:
1768 		tcp_state_change(tp, TCPS_LAST_ACK);
1769 		break;
1770 	}
1771 	if (tp->t_state >= TCPS_FIN_WAIT_2) {
1772 		soisdisconnected(tp->t_inpcb->inp_socket);
1773 		/* Prevent the connection hanging in FIN_WAIT_2 forever. */
1774 		if (tp->t_state == TCPS_FIN_WAIT_2) {
1775 			int timeout;
1776 
1777 			timeout = (tcp_fast_finwait2_recycle) ?
1778 			    tcp_finwait2_timeout : TP_MAXIDLE(tp);
1779 			tcp_timer_activate(tp, TT_2MSL, timeout);
1780 		}
1781 	}
1782 }
1783 
1784 #ifdef DDB
1785 static void
1786 db_print_indent(int indent)
1787 {
1788 	int i;
1789 
1790 	for (i = 0; i < indent; i++)
1791 		db_printf(" ");
1792 }
1793 
1794 static void
1795 db_print_tstate(int t_state)
1796 {
1797 
1798 	switch (t_state) {
1799 	case TCPS_CLOSED:
1800 		db_printf("TCPS_CLOSED");
1801 		return;
1802 
1803 	case TCPS_LISTEN:
1804 		db_printf("TCPS_LISTEN");
1805 		return;
1806 
1807 	case TCPS_SYN_SENT:
1808 		db_printf("TCPS_SYN_SENT");
1809 		return;
1810 
1811 	case TCPS_SYN_RECEIVED:
1812 		db_printf("TCPS_SYN_RECEIVED");
1813 		return;
1814 
1815 	case TCPS_ESTABLISHED:
1816 		db_printf("TCPS_ESTABLISHED");
1817 		return;
1818 
1819 	case TCPS_CLOSE_WAIT:
1820 		db_printf("TCPS_CLOSE_WAIT");
1821 		return;
1822 
1823 	case TCPS_FIN_WAIT_1:
1824 		db_printf("TCPS_FIN_WAIT_1");
1825 		return;
1826 
1827 	case TCPS_CLOSING:
1828 		db_printf("TCPS_CLOSING");
1829 		return;
1830 
1831 	case TCPS_LAST_ACK:
1832 		db_printf("TCPS_LAST_ACK");
1833 		return;
1834 
1835 	case TCPS_FIN_WAIT_2:
1836 		db_printf("TCPS_FIN_WAIT_2");
1837 		return;
1838 
1839 	case TCPS_TIME_WAIT:
1840 		db_printf("TCPS_TIME_WAIT");
1841 		return;
1842 
1843 	default:
1844 		db_printf("unknown");
1845 		return;
1846 	}
1847 }
1848 
1849 static void
1850 db_print_tflags(u_int t_flags)
1851 {
1852 	int comma;
1853 
1854 	comma = 0;
1855 	if (t_flags & TF_ACKNOW) {
1856 		db_printf("%sTF_ACKNOW", comma ? ", " : "");
1857 		comma = 1;
1858 	}
1859 	if (t_flags & TF_DELACK) {
1860 		db_printf("%sTF_DELACK", comma ? ", " : "");
1861 		comma = 1;
1862 	}
1863 	if (t_flags & TF_NODELAY) {
1864 		db_printf("%sTF_NODELAY", comma ? ", " : "");
1865 		comma = 1;
1866 	}
1867 	if (t_flags & TF_NOOPT) {
1868 		db_printf("%sTF_NOOPT", comma ? ", " : "");
1869 		comma = 1;
1870 	}
1871 	if (t_flags & TF_SENTFIN) {
1872 		db_printf("%sTF_SENTFIN", comma ? ", " : "");
1873 		comma = 1;
1874 	}
1875 	if (t_flags & TF_REQ_SCALE) {
1876 		db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
1877 		comma = 1;
1878 	}
1879 	if (t_flags & TF_RCVD_SCALE) {
1880 		db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
1881 		comma = 1;
1882 	}
1883 	if (t_flags & TF_REQ_TSTMP) {
1884 		db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
1885 		comma = 1;
1886 	}
1887 	if (t_flags & TF_RCVD_TSTMP) {
1888 		db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
1889 		comma = 1;
1890 	}
1891 	if (t_flags & TF_SACK_PERMIT) {
1892 		db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
1893 		comma = 1;
1894 	}
1895 	if (t_flags & TF_NEEDSYN) {
1896 		db_printf("%sTF_NEEDSYN", comma ? ", " : "");
1897 		comma = 1;
1898 	}
1899 	if (t_flags & TF_NEEDFIN) {
1900 		db_printf("%sTF_NEEDFIN", comma ? ", " : "");
1901 		comma = 1;
1902 	}
1903 	if (t_flags & TF_NOPUSH) {
1904 		db_printf("%sTF_NOPUSH", comma ? ", " : "");
1905 		comma = 1;
1906 	}
1907 	if (t_flags & TF_MORETOCOME) {
1908 		db_printf("%sTF_MORETOCOME", comma ? ", " : "");
1909 		comma = 1;
1910 	}
1911 	if (t_flags & TF_LQ_OVERFLOW) {
1912 		db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
1913 		comma = 1;
1914 	}
1915 	if (t_flags & TF_LASTIDLE) {
1916 		db_printf("%sTF_LASTIDLE", comma ? ", " : "");
1917 		comma = 1;
1918 	}
1919 	if (t_flags & TF_RXWIN0SENT) {
1920 		db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
1921 		comma = 1;
1922 	}
1923 	if (t_flags & TF_FASTRECOVERY) {
1924 		db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
1925 		comma = 1;
1926 	}
1927 	if (t_flags & TF_CONGRECOVERY) {
1928 		db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
1929 		comma = 1;
1930 	}
1931 	if (t_flags & TF_WASFRECOVERY) {
1932 		db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
1933 		comma = 1;
1934 	}
1935 	if (t_flags & TF_SIGNATURE) {
1936 		db_printf("%sTF_SIGNATURE", comma ? ", " : "");
1937 		comma = 1;
1938 	}
1939 	if (t_flags & TF_FORCEDATA) {
1940 		db_printf("%sTF_FORCEDATA", comma ? ", " : "");
1941 		comma = 1;
1942 	}
1943 	if (t_flags & TF_TSO) {
1944 		db_printf("%sTF_TSO", comma ? ", " : "");
1945 		comma = 1;
1946 	}
1947 	if (t_flags & TF_ECN_PERMIT) {
1948 		db_printf("%sTF_ECN_PERMIT", comma ? ", " : "");
1949 		comma = 1;
1950 	}
1951 }
1952 
1953 static void
1954 db_print_toobflags(char t_oobflags)
1955 {
1956 	int comma;
1957 
1958 	comma = 0;
1959 	if (t_oobflags & TCPOOB_HAVEDATA) {
1960 		db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
1961 		comma = 1;
1962 	}
1963 	if (t_oobflags & TCPOOB_HADDATA) {
1964 		db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
1965 		comma = 1;
1966 	}
1967 }
1968 
1969 static void
1970 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
1971 {
1972 
1973 	db_print_indent(indent);
1974 	db_printf("%s at %p\n", name, tp);
1975 
1976 	indent += 2;
1977 
1978 	db_print_indent(indent);
1979 	db_printf("t_segq first: %p   t_segqlen: %d   t_dupacks: %d\n",
1980 	   tp->t_segq, tp->t_segqlen, tp->t_dupacks);
1981 
1982 	db_print_indent(indent);
1983 	db_printf("tt_rexmt: %p   tt_persist: %p   tt_keep: %p\n",
1984 	    &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
1985 
1986 	db_print_indent(indent);
1987 	db_printf("tt_2msl: %p   tt_delack: %p   t_inpcb: %p\n", &tp->t_timers->tt_2msl,
1988 	    &tp->t_timers->tt_delack, tp->t_inpcb);
1989 
1990 	db_print_indent(indent);
1991 	db_printf("t_state: %d (", tp->t_state);
1992 	db_print_tstate(tp->t_state);
1993 	db_printf(")\n");
1994 
1995 	db_print_indent(indent);
1996 	db_printf("t_flags: 0x%x (", tp->t_flags);
1997 	db_print_tflags(tp->t_flags);
1998 	db_printf(")\n");
1999 
2000 	db_print_indent(indent);
2001 	db_printf("snd_una: 0x%08x   snd_max: 0x%08x   snd_nxt: x0%08x\n",
2002 	    tp->snd_una, tp->snd_max, tp->snd_nxt);
2003 
2004 	db_print_indent(indent);
2005 	db_printf("snd_up: 0x%08x   snd_wl1: 0x%08x   snd_wl2: 0x%08x\n",
2006 	   tp->snd_up, tp->snd_wl1, tp->snd_wl2);
2007 
2008 	db_print_indent(indent);
2009 	db_printf("iss: 0x%08x   irs: 0x%08x   rcv_nxt: 0x%08x\n",
2010 	    tp->iss, tp->irs, tp->rcv_nxt);
2011 
2012 	db_print_indent(indent);
2013 	db_printf("rcv_adv: 0x%08x   rcv_wnd: %lu   rcv_up: 0x%08x\n",
2014 	    tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
2015 
2016 	db_print_indent(indent);
2017 	db_printf("snd_wnd: %lu   snd_cwnd: %lu\n",
2018 	   tp->snd_wnd, tp->snd_cwnd);
2019 
2020 	db_print_indent(indent);
2021 	db_printf("snd_ssthresh: %lu   snd_recover: "
2022 	    "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
2023 
2024 	db_print_indent(indent);
2025 	db_printf("t_maxopd: %u   t_rcvtime: %u   t_startime: %u\n",
2026 	    tp->t_maxopd, tp->t_rcvtime, tp->t_starttime);
2027 
2028 	db_print_indent(indent);
2029 	db_printf("t_rttime: %u   t_rtsq: 0x%08x\n",
2030 	    tp->t_rtttime, tp->t_rtseq);
2031 
2032 	db_print_indent(indent);
2033 	db_printf("t_rxtcur: %d   t_maxseg: %u   t_srtt: %d\n",
2034 	    tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
2035 
2036 	db_print_indent(indent);
2037 	db_printf("t_rttvar: %d   t_rxtshift: %d   t_rttmin: %u   "
2038 	    "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
2039 	    tp->t_rttbest);
2040 
2041 	db_print_indent(indent);
2042 	db_printf("t_rttupdated: %lu   max_sndwnd: %lu   t_softerror: %d\n",
2043 	    tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
2044 
2045 	db_print_indent(indent);
2046 	db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
2047 	db_print_toobflags(tp->t_oobflags);
2048 	db_printf(")   t_iobc: 0x%02x\n", tp->t_iobc);
2049 
2050 	db_print_indent(indent);
2051 	db_printf("snd_scale: %u   rcv_scale: %u   request_r_scale: %u\n",
2052 	    tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
2053 
2054 	db_print_indent(indent);
2055 	db_printf("ts_recent: %u   ts_recent_age: %u\n",
2056 	    tp->ts_recent, tp->ts_recent_age);
2057 
2058 	db_print_indent(indent);
2059 	db_printf("ts_offset: %u   last_ack_sent: 0x%08x   snd_cwnd_prev: "
2060 	    "%lu\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
2061 
2062 	db_print_indent(indent);
2063 	db_printf("snd_ssthresh_prev: %lu   snd_recover_prev: 0x%08x   "
2064 	    "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
2065 	    tp->snd_recover_prev, tp->t_badrxtwin);
2066 
2067 	db_print_indent(indent);
2068 	db_printf("snd_numholes: %d  snd_holes first: %p\n",
2069 	    tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
2070 
2071 	db_print_indent(indent);
2072 	db_printf("snd_fack: 0x%08x   rcv_numsacks: %d   sack_newdata: "
2073 	    "0x%08x\n", tp->snd_fack, tp->rcv_numsacks, tp->sack_newdata);
2074 
2075 	/* Skip sackblks, sackhint. */
2076 
2077 	db_print_indent(indent);
2078 	db_printf("t_rttlow: %d   rfbuf_ts: %u   rfbuf_cnt: %d\n",
2079 	    tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
2080 }
2081 
2082 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
2083 {
2084 	struct tcpcb *tp;
2085 
2086 	if (!have_addr) {
2087 		db_printf("usage: show tcpcb <addr>\n");
2088 		return;
2089 	}
2090 	tp = (struct tcpcb *)addr;
2091 
2092 	db_print_tcpcb(tp, "tcpcb", 0);
2093 }
2094 #endif
2095