xref: /freebsd/sys/netinet/tcp_usrreq.c (revision dda082e0093936517795d386242ff9abcdaeff53)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1986, 1988, 1993
5  *	The Regents of the University of California.
6  * Copyright (c) 2006-2007 Robert N. M. Watson
7  * Copyright (c) 2010-2011 Juniper Networks, Inc.
8  * All rights reserved.
9  *
10  * Portions of this software were developed by Robert N. M. Watson under
11  * contract to Juniper Networks, Inc.
12  *
13  * Redistribution and use in source and binary forms, with or without
14  * modification, are permitted provided that the following conditions
15  * are met:
16  * 1. Redistributions of source code must retain the above copyright
17  *    notice, this list of conditions and the following disclaimer.
18  * 2. Redistributions in binary form must reproduce the above copyright
19  *    notice, this list of conditions and the following disclaimer in the
20  *    documentation and/or other materials provided with the distribution.
21  * 3. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  *
37  *	From: @(#)tcp_usrreq.c	8.2 (Berkeley) 1/3/94
38  */
39 
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include "opt_ddb.h"
44 #include "opt_inet.h"
45 #include "opt_inet6.h"
46 #include "opt_ipsec.h"
47 #include "opt_kern_tls.h"
48 #include "opt_tcpdebug.h"
49 
50 #include <sys/param.h>
51 #include <sys/systm.h>
52 #include <sys/arb.h>
53 #include <sys/limits.h>
54 #include <sys/malloc.h>
55 #include <sys/refcount.h>
56 #include <sys/kernel.h>
57 #include <sys/ktls.h>
58 #include <sys/qmath.h>
59 #include <sys/sysctl.h>
60 #include <sys/mbuf.h>
61 #ifdef INET6
62 #include <sys/domain.h>
63 #endif /* INET6 */
64 #include <sys/socket.h>
65 #include <sys/socketvar.h>
66 #include <sys/protosw.h>
67 #include <sys/proc.h>
68 #include <sys/jail.h>
69 #include <sys/stats.h>
70 
71 #ifdef DDB
72 #include <ddb/ddb.h>
73 #endif
74 
75 #include <net/if.h>
76 #include <net/if_var.h>
77 #include <net/route.h>
78 #include <net/vnet.h>
79 
80 #include <netinet/in.h>
81 #include <netinet/in_kdtrace.h>
82 #include <netinet/in_pcb.h>
83 #include <netinet/in_systm.h>
84 #include <netinet/in_var.h>
85 #include <netinet/ip.h>
86 #include <netinet/ip_var.h>
87 #ifdef INET6
88 #include <netinet/ip6.h>
89 #include <netinet6/in6_pcb.h>
90 #include <netinet6/ip6_var.h>
91 #include <netinet6/scope6_var.h>
92 #endif
93 #include <netinet/tcp.h>
94 #include <netinet/tcp_fsm.h>
95 #include <netinet/tcp_seq.h>
96 #include <netinet/tcp_timer.h>
97 #include <netinet/tcp_var.h>
98 #include <netinet/tcp_log_buf.h>
99 #include <netinet/tcpip.h>
100 #include <netinet/cc/cc.h>
101 #include <netinet/tcp_fastopen.h>
102 #include <netinet/tcp_hpts.h>
103 #ifdef TCPPCAP
104 #include <netinet/tcp_pcap.h>
105 #endif
106 #include <netinet/tcp_debug.h>
107 #ifdef TCP_OFFLOAD
108 #include <netinet/tcp_offload.h>
109 #endif
110 #include <netipsec/ipsec_support.h>
111 
112 #include <vm/vm.h>
113 #include <vm/vm_param.h>
114 #include <vm/pmap.h>
115 #include <vm/vm_extern.h>
116 #include <vm/vm_map.h>
117 #include <vm/vm_page.h>
118 
119 /*
120  * TCP protocol interface to socket abstraction.
121  */
122 #ifdef INET
123 static int	tcp_connect(struct tcpcb *, struct sockaddr *,
124 		    struct thread *td);
125 #endif /* INET */
126 #ifdef INET6
127 static int	tcp6_connect(struct tcpcb *, struct sockaddr *,
128 		    struct thread *td);
129 #endif /* INET6 */
130 static void	tcp_disconnect(struct tcpcb *);
131 static void	tcp_usrclosed(struct tcpcb *);
132 static void	tcp_fill_info(struct tcpcb *, struct tcp_info *);
133 
134 static int	tcp_pru_options_support(struct tcpcb *tp, int flags);
135 
136 #ifdef TCPDEBUG
137 #define	TCPDEBUG0	int ostate = 0
138 #define	TCPDEBUG1()	ostate = tp ? tp->t_state : 0
139 #define	TCPDEBUG2(req)	if (tp && (so->so_options & SO_DEBUG)) \
140 				tcp_trace(TA_USER, ostate, tp, 0, 0, req)
141 #else
142 #define	TCPDEBUG0
143 #define	TCPDEBUG1()
144 #define	TCPDEBUG2(req)
145 #endif
146 
147 /*
148  * tcp_require_unique port requires a globally-unique source port for each
149  * outgoing connection.  The default is to require the 4-tuple to be unique.
150  */
151 VNET_DEFINE(int, tcp_require_unique_port) = 0;
152 SYSCTL_INT(_net_inet_tcp, OID_AUTO, require_unique_port,
153     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(tcp_require_unique_port), 0,
154     "Require globally-unique ephemeral port for outgoing connections");
155 #define	V_tcp_require_unique_port	VNET(tcp_require_unique_port)
156 
157 /*
158  * TCP attaches to socket via pru_attach(), reserving space,
159  * and an internet control block.
160  */
161 static int
162 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
163 {
164 	struct inpcb *inp;
165 	struct tcpcb *tp = NULL;
166 	int error;
167 	TCPDEBUG0;
168 
169 	inp = sotoinpcb(so);
170 	KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
171 	TCPDEBUG1();
172 
173 	error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
174 	if (error)
175 		goto out;
176 
177 	so->so_rcv.sb_flags |= SB_AUTOSIZE;
178 	so->so_snd.sb_flags |= SB_AUTOSIZE;
179 	error = in_pcballoc(so, &V_tcbinfo);
180 	if (error)
181 		goto out;
182 	inp = sotoinpcb(so);
183 	tp = tcp_newtcpcb(inp);
184 	if (tp == NULL) {
185 		error = ENOBUFS;
186 		in_pcbdetach(inp);
187 		in_pcbfree(inp);
188 		goto out;
189 	}
190 	tp->t_state = TCPS_CLOSED;
191 	INP_WUNLOCK(inp);
192 	TCPSTATES_INC(TCPS_CLOSED);
193 out:
194 	TCPDEBUG2(PRU_ATTACH);
195 	TCP_PROBE2(debug__user, tp, PRU_ATTACH);
196 	return (error);
197 }
198 
199 /*
200  * tcp_usr_detach is called when the socket layer loses its final reference
201  * to the socket, be it a file descriptor reference, a reference from TCP,
202  * etc.  At this point, there is only one case in which we will keep around
203  * inpcb state: time wait.
204  */
205 static void
206 tcp_usr_detach(struct socket *so)
207 {
208 	struct inpcb *inp;
209 	struct tcpcb *tp;
210 
211 	inp = sotoinpcb(so);
212 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
213 	INP_WLOCK(inp);
214 	KASSERT(so->so_pcb == inp && inp->inp_socket == so,
215 		("%s: socket %p inp %p mismatch", __func__, so, inp));
216 
217 	tp = intotcpcb(inp);
218 
219 	if (inp->inp_flags & INP_TIMEWAIT) {
220 		/*
221 		 * There are two cases to handle: one in which the time wait
222 		 * state is being discarded (INP_DROPPED), and one in which
223 		 * this connection will remain in timewait.  In the former,
224 		 * it is time to discard all state (except tcptw, which has
225 		 * already been discarded by the timewait close code, which
226 		 * should be further up the call stack somewhere).  In the
227 		 * latter case, we detach from the socket, but leave the pcb
228 		 * present until timewait ends.
229 		 *
230 		 * XXXRW: Would it be cleaner to free the tcptw here?
231 		 *
232 		 * Astute question indeed, from twtcp perspective there are
233 		 * four cases to consider:
234 		 *
235 		 * #1 tcp_usr_detach is called at tcptw creation time by
236 		 *  tcp_twstart, then do not discard the newly created tcptw
237 		 *  and leave inpcb present until timewait ends
238 		 * #2 tcp_usr_detach is called at tcptw creation time by
239 		 *  tcp_twstart, but connection is local and tw will be
240 		 *  discarded immediately
241 		 * #3 tcp_usr_detach is called at timewait end (or reuse) by
242 		 *  tcp_twclose, then the tcptw has already been discarded
243 		 *  (or reused) and inpcb is freed here
244 		 * #4 tcp_usr_detach is called() after timewait ends (or reuse)
245 		 *  (e.g. by soclose), then tcptw has already been discarded
246 		 *  (or reused) and inpcb is freed here
247 		 *
248 		 *  In all three cases the tcptw should not be freed here.
249 		 */
250 		if (inp->inp_flags & INP_DROPPED) {
251 			KASSERT(tp == NULL, ("tcp_detach: INP_TIMEWAIT && "
252 			    "INP_DROPPED && tp != NULL"));
253 			in_pcbdetach(inp);
254 			in_pcbfree(inp);
255 		} else {
256 			in_pcbdetach(inp);
257 			INP_WUNLOCK(inp);
258 		}
259 	} else {
260 		/*
261 		 * If the connection is not in timewait, it must be either
262 		 * dropped or embryonic.
263 		 */
264 		KASSERT(inp->inp_flags & INP_DROPPED ||
265 		    tp->t_state < TCPS_SYN_SENT,
266 		    ("%s: inp %p not dropped or embryonic", __func__, inp));
267 		tcp_discardcb(tp);
268 		in_pcbdetach(inp);
269 		in_pcbfree(inp);
270 	}
271 }
272 
273 #ifdef INET
274 /*
275  * Give the socket an address.
276  */
277 static int
278 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
279 {
280 	int error = 0;
281 	struct inpcb *inp;
282 #ifdef KDTRACE_HOOKS
283 	struct tcpcb *tp = NULL;
284 #endif
285 	struct sockaddr_in *sinp;
286 
287 	sinp = (struct sockaddr_in *)nam;
288 	if (nam->sa_family != AF_INET) {
289 		/*
290 		 * Preserve compatibility with old programs.
291 		 */
292 		if (nam->sa_family != AF_UNSPEC ||
293 		    nam->sa_len < offsetof(struct sockaddr_in, sin_zero) ||
294 		    sinp->sin_addr.s_addr != INADDR_ANY)
295 			return (EAFNOSUPPORT);
296 		nam->sa_family = AF_INET;
297 	}
298 	if (nam->sa_len != sizeof(*sinp))
299 		return (EINVAL);
300 
301 	/*
302 	 * Must check for multicast addresses and disallow binding
303 	 * to them.
304 	 */
305 	if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
306 		return (EAFNOSUPPORT);
307 
308 	TCPDEBUG0;
309 	inp = sotoinpcb(so);
310 	KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
311 	INP_WLOCK(inp);
312 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
313 		error = EINVAL;
314 		goto out;
315 	}
316 #ifdef KDTRACE_HOOKS
317 	tp = intotcpcb(inp);
318 #endif
319 	TCPDEBUG1();
320 	INP_HASH_WLOCK(&V_tcbinfo);
321 	error = in_pcbbind(inp, nam, td->td_ucred);
322 	INP_HASH_WUNLOCK(&V_tcbinfo);
323 out:
324 	TCPDEBUG2(PRU_BIND);
325 	TCP_PROBE2(debug__user, tp, PRU_BIND);
326 	INP_WUNLOCK(inp);
327 
328 	return (error);
329 }
330 #endif /* INET */
331 
332 #ifdef INET6
333 static int
334 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
335 {
336 	int error = 0;
337 	struct inpcb *inp;
338 #ifdef KDTRACE_HOOKS
339 	struct tcpcb *tp = NULL;
340 #endif
341 	struct sockaddr_in6 *sin6;
342 	u_char vflagsav;
343 
344 	sin6 = (struct sockaddr_in6 *)nam;
345 	if (nam->sa_family != AF_INET6)
346 		return (EAFNOSUPPORT);
347 	if (nam->sa_len != sizeof(*sin6))
348 		return (EINVAL);
349 
350 	/*
351 	 * Must check for multicast addresses and disallow binding
352 	 * to them.
353 	 */
354 	if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
355 		return (EAFNOSUPPORT);
356 
357 	TCPDEBUG0;
358 	inp = sotoinpcb(so);
359 	KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
360 	INP_WLOCK(inp);
361 	vflagsav = inp->inp_vflag;
362 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
363 		error = EINVAL;
364 		goto out;
365 	}
366 #ifdef KDTRACE_HOOKS
367 	tp = intotcpcb(inp);
368 #endif
369 	TCPDEBUG1();
370 	INP_HASH_WLOCK(&V_tcbinfo);
371 	inp->inp_vflag &= ~INP_IPV4;
372 	inp->inp_vflag |= INP_IPV6;
373 #ifdef INET
374 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
375 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
376 			inp->inp_vflag |= INP_IPV4;
377 		else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
378 			struct sockaddr_in sin;
379 
380 			in6_sin6_2_sin(&sin, sin6);
381 			if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
382 				error = EAFNOSUPPORT;
383 				INP_HASH_WUNLOCK(&V_tcbinfo);
384 				goto out;
385 			}
386 			inp->inp_vflag |= INP_IPV4;
387 			inp->inp_vflag &= ~INP_IPV6;
388 			error = in_pcbbind(inp, (struct sockaddr *)&sin,
389 			    td->td_ucred);
390 			INP_HASH_WUNLOCK(&V_tcbinfo);
391 			goto out;
392 		}
393 	}
394 #endif
395 	error = in6_pcbbind(inp, nam, td->td_ucred);
396 	INP_HASH_WUNLOCK(&V_tcbinfo);
397 out:
398 	if (error != 0)
399 		inp->inp_vflag = vflagsav;
400 	TCPDEBUG2(PRU_BIND);
401 	TCP_PROBE2(debug__user, tp, PRU_BIND);
402 	INP_WUNLOCK(inp);
403 	return (error);
404 }
405 #endif /* INET6 */
406 
407 #ifdef INET
408 /*
409  * Prepare to accept connections.
410  */
411 static int
412 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
413 {
414 	int error = 0;
415 	struct inpcb *inp;
416 	struct tcpcb *tp = NULL;
417 
418 	TCPDEBUG0;
419 	inp = sotoinpcb(so);
420 	KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
421 	INP_WLOCK(inp);
422 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
423 		error = EINVAL;
424 		goto out;
425 	}
426 	tp = intotcpcb(inp);
427 	TCPDEBUG1();
428 	SOCK_LOCK(so);
429 	error = solisten_proto_check(so);
430 	if (error != 0) {
431 		SOCK_UNLOCK(so);
432 		goto out;
433 	}
434 	if (inp->inp_lport == 0) {
435 		INP_HASH_WLOCK(&V_tcbinfo);
436 		error = in_pcbbind(inp, NULL, td->td_ucred);
437 		INP_HASH_WUNLOCK(&V_tcbinfo);
438 	}
439 	if (error == 0) {
440 		tcp_state_change(tp, TCPS_LISTEN);
441 		solisten_proto(so, backlog);
442 #ifdef TCP_OFFLOAD
443 		if ((so->so_options & SO_NO_OFFLOAD) == 0)
444 			tcp_offload_listen_start(tp);
445 #endif
446 	} else {
447 		solisten_proto_abort(so);
448 	}
449 	SOCK_UNLOCK(so);
450 
451 	if (IS_FASTOPEN(tp->t_flags))
452 		tp->t_tfo_pending = tcp_fastopen_alloc_counter();
453 
454 out:
455 	TCPDEBUG2(PRU_LISTEN);
456 	TCP_PROBE2(debug__user, tp, PRU_LISTEN);
457 	INP_WUNLOCK(inp);
458 	return (error);
459 }
460 #endif /* INET */
461 
462 #ifdef INET6
463 static int
464 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
465 {
466 	int error = 0;
467 	struct inpcb *inp;
468 	struct tcpcb *tp = NULL;
469 	u_char vflagsav;
470 
471 	TCPDEBUG0;
472 	inp = sotoinpcb(so);
473 	KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
474 	INP_WLOCK(inp);
475 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
476 		error = EINVAL;
477 		goto out;
478 	}
479 	vflagsav = inp->inp_vflag;
480 	tp = intotcpcb(inp);
481 	TCPDEBUG1();
482 	SOCK_LOCK(so);
483 	error = solisten_proto_check(so);
484 	if (error != 0) {
485 		SOCK_UNLOCK(so);
486 		goto out;
487 	}
488 	INP_HASH_WLOCK(&V_tcbinfo);
489 	if (inp->inp_lport == 0) {
490 		inp->inp_vflag &= ~INP_IPV4;
491 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
492 			inp->inp_vflag |= INP_IPV4;
493 		error = in6_pcbbind(inp, NULL, td->td_ucred);
494 	}
495 	INP_HASH_WUNLOCK(&V_tcbinfo);
496 	if (error == 0) {
497 		tcp_state_change(tp, TCPS_LISTEN);
498 		solisten_proto(so, backlog);
499 #ifdef TCP_OFFLOAD
500 		if ((so->so_options & SO_NO_OFFLOAD) == 0)
501 			tcp_offload_listen_start(tp);
502 #endif
503 	} else {
504 		solisten_proto_abort(so);
505 	}
506 	SOCK_UNLOCK(so);
507 
508 	if (IS_FASTOPEN(tp->t_flags))
509 		tp->t_tfo_pending = tcp_fastopen_alloc_counter();
510 
511 	if (error != 0)
512 		inp->inp_vflag = vflagsav;
513 
514 out:
515 	TCPDEBUG2(PRU_LISTEN);
516 	TCP_PROBE2(debug__user, tp, PRU_LISTEN);
517 	INP_WUNLOCK(inp);
518 	return (error);
519 }
520 #endif /* INET6 */
521 
522 #ifdef INET
523 /*
524  * Initiate connection to peer.
525  * Create a template for use in transmissions on this connection.
526  * Enter SYN_SENT state, and mark socket as connecting.
527  * Start keep-alive timer, and seed output sequence space.
528  * Send initial segment on connection.
529  */
530 static int
531 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
532 {
533 	struct epoch_tracker et;
534 	int error = 0;
535 	struct inpcb *inp;
536 	struct tcpcb *tp = NULL;
537 	struct sockaddr_in *sinp;
538 
539 	sinp = (struct sockaddr_in *)nam;
540 	if (nam->sa_family != AF_INET)
541 		return (EAFNOSUPPORT);
542 	if (nam->sa_len != sizeof (*sinp))
543 		return (EINVAL);
544 
545 	/*
546 	 * Must disallow TCP ``connections'' to multicast addresses.
547 	 */
548 	if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)))
549 		return (EAFNOSUPPORT);
550 	if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST)
551 		return (EACCES);
552 	if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
553 		return (error);
554 
555 	TCPDEBUG0;
556 	inp = sotoinpcb(so);
557 	KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
558 	INP_WLOCK(inp);
559 	if (inp->inp_flags & INP_TIMEWAIT) {
560 		error = EADDRINUSE;
561 		goto out;
562 	}
563 	if (inp->inp_flags & INP_DROPPED) {
564 		error = ECONNREFUSED;
565 		goto out;
566 	}
567 	if (SOLISTENING(so)) {
568 		error = EOPNOTSUPP;
569 		goto out;
570 	}
571 	tp = intotcpcb(inp);
572 	TCPDEBUG1();
573 	NET_EPOCH_ENTER(et);
574 	if ((error = tcp_connect(tp, nam, td)) != 0)
575 		goto out_in_epoch;
576 #ifdef TCP_OFFLOAD
577 	if (registered_toedevs > 0 &&
578 	    (so->so_options & SO_NO_OFFLOAD) == 0 &&
579 	    (error = tcp_offload_connect(so, nam)) == 0)
580 		goto out_in_epoch;
581 #endif
582 	tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
583 	error = tcp_output(tp);
584 	KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()"
585 	    ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error));
586 out_in_epoch:
587 	NET_EPOCH_EXIT(et);
588 out:
589 	TCPDEBUG2(PRU_CONNECT);
590 	TCP_PROBE2(debug__user, tp, PRU_CONNECT);
591 	INP_WUNLOCK(inp);
592 	return (error);
593 }
594 #endif /* INET */
595 
596 #ifdef INET6
597 static int
598 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
599 {
600 	struct epoch_tracker et;
601 	int error = 0;
602 	struct inpcb *inp;
603 	struct tcpcb *tp = NULL;
604 	struct sockaddr_in6 *sin6;
605 	u_int8_t incflagsav;
606 	u_char vflagsav;
607 
608 	TCPDEBUG0;
609 
610 	sin6 = (struct sockaddr_in6 *)nam;
611 	if (nam->sa_family != AF_INET6)
612 		return (EAFNOSUPPORT);
613 	if (nam->sa_len != sizeof (*sin6))
614 		return (EINVAL);
615 
616 	/*
617 	 * Must disallow TCP ``connections'' to multicast addresses.
618 	 */
619 	if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
620 		return (EAFNOSUPPORT);
621 
622 	inp = sotoinpcb(so);
623 	KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
624 	INP_WLOCK(inp);
625 	vflagsav = inp->inp_vflag;
626 	incflagsav = inp->inp_inc.inc_flags;
627 	if (inp->inp_flags & INP_TIMEWAIT) {
628 		error = EADDRINUSE;
629 		goto out;
630 	}
631 	if (inp->inp_flags & INP_DROPPED) {
632 		error = ECONNREFUSED;
633 		goto out;
634 	}
635 	if (SOLISTENING(so)) {
636 		error = EINVAL;
637 		goto out;
638 	}
639 	tp = intotcpcb(inp);
640 	TCPDEBUG1();
641 #ifdef INET
642 	/*
643 	 * XXXRW: Some confusion: V4/V6 flags relate to binding, and
644 	 * therefore probably require the hash lock, which isn't held here.
645 	 * Is this a significant problem?
646 	 */
647 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
648 		struct sockaddr_in sin;
649 
650 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
651 			error = EINVAL;
652 			goto out;
653 		}
654 		if ((inp->inp_vflag & INP_IPV4) == 0) {
655 			error = EAFNOSUPPORT;
656 			goto out;
657 		}
658 
659 		in6_sin6_2_sin(&sin, sin6);
660 		if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
661 			error = EAFNOSUPPORT;
662 			goto out;
663 		}
664 		if (ntohl(sin.sin_addr.s_addr) == INADDR_BROADCAST) {
665 			error = EACCES;
666 			goto out;
667 		}
668 		if ((error = prison_remote_ip4(td->td_ucred,
669 		    &sin.sin_addr)) != 0)
670 			goto out;
671 		inp->inp_vflag |= INP_IPV4;
672 		inp->inp_vflag &= ~INP_IPV6;
673 		NET_EPOCH_ENTER(et);
674 		if ((error = tcp_connect(tp, (struct sockaddr *)&sin, td)) != 0)
675 			goto out_in_epoch;
676 #ifdef TCP_OFFLOAD
677 		if (registered_toedevs > 0 &&
678 		    (so->so_options & SO_NO_OFFLOAD) == 0 &&
679 		    (error = tcp_offload_connect(so, nam)) == 0)
680 			goto out_in_epoch;
681 #endif
682 		error = tcp_output(tp);
683 		goto out_in_epoch;
684 	} else {
685 		if ((inp->inp_vflag & INP_IPV6) == 0) {
686 			error = EAFNOSUPPORT;
687 			goto out;
688 		}
689 	}
690 #endif
691 	if ((error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr)) != 0)
692 		goto out;
693 	inp->inp_vflag &= ~INP_IPV4;
694 	inp->inp_vflag |= INP_IPV6;
695 	inp->inp_inc.inc_flags |= INC_ISIPV6;
696 	NET_EPOCH_ENTER(et);
697 	if ((error = tcp6_connect(tp, nam, td)) != 0)
698 		goto out_in_epoch;
699 #ifdef TCP_OFFLOAD
700 	if (registered_toedevs > 0 &&
701 	    (so->so_options & SO_NO_OFFLOAD) == 0 &&
702 	    (error = tcp_offload_connect(so, nam)) == 0)
703 		goto out_in_epoch;
704 #endif
705 	tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
706 	error = tcp_output(tp);
707 out_in_epoch:
708 	NET_EPOCH_EXIT(et);
709 out:
710 	KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()"
711 	    ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error));
712 	/*
713 	 * If the implicit bind in the connect call fails, restore
714 	 * the flags we modified.
715 	 */
716 	if (error != 0 && inp->inp_lport == 0) {
717 		inp->inp_vflag = vflagsav;
718 		inp->inp_inc.inc_flags = incflagsav;
719 	}
720 
721 	TCPDEBUG2(PRU_CONNECT);
722 	TCP_PROBE2(debug__user, tp, PRU_CONNECT);
723 	INP_WUNLOCK(inp);
724 	return (error);
725 }
726 #endif /* INET6 */
727 
728 /*
729  * Initiate disconnect from peer.
730  * If connection never passed embryonic stage, just drop;
731  * else if don't need to let data drain, then can just drop anyways,
732  * else have to begin TCP shutdown process: mark socket disconnecting,
733  * drain unread data, state switch to reflect user close, and
734  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
735  * when peer sends FIN and acks ours.
736  *
737  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
738  */
739 static int
740 tcp_usr_disconnect(struct socket *so)
741 {
742 	struct inpcb *inp;
743 	struct tcpcb *tp = NULL;
744 	struct epoch_tracker et;
745 	int error = 0;
746 
747 	TCPDEBUG0;
748 	NET_EPOCH_ENTER(et);
749 	inp = sotoinpcb(so);
750 	KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
751 	INP_WLOCK(inp);
752 	if (inp->inp_flags & INP_TIMEWAIT)
753 		goto out;
754 	if (inp->inp_flags & INP_DROPPED) {
755 		error = ECONNRESET;
756 		goto out;
757 	}
758 	tp = intotcpcb(inp);
759 	TCPDEBUG1();
760 	tcp_disconnect(tp);
761 out:
762 	TCPDEBUG2(PRU_DISCONNECT);
763 	TCP_PROBE2(debug__user, tp, PRU_DISCONNECT);
764 	INP_WUNLOCK(inp);
765 	NET_EPOCH_EXIT(et);
766 	return (error);
767 }
768 
769 #ifdef INET
770 /*
771  * Accept a connection.  Essentially all the work is done at higher levels;
772  * just return the address of the peer, storing through addr.
773  */
774 static int
775 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
776 {
777 	int error = 0;
778 	struct inpcb *inp = NULL;
779 #ifdef KDTRACE_HOOKS
780 	struct tcpcb *tp = NULL;
781 #endif
782 	struct in_addr addr;
783 	in_port_t port = 0;
784 	TCPDEBUG0;
785 
786 	if (so->so_state & SS_ISDISCONNECTED)
787 		return (ECONNABORTED);
788 
789 	inp = sotoinpcb(so);
790 	KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
791 	INP_WLOCK(inp);
792 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
793 		error = ECONNABORTED;
794 		goto out;
795 	}
796 #ifdef KDTRACE_HOOKS
797 	tp = intotcpcb(inp);
798 #endif
799 	TCPDEBUG1();
800 
801 	/*
802 	 * We inline in_getpeeraddr and COMMON_END here, so that we can
803 	 * copy the data of interest and defer the malloc until after we
804 	 * release the lock.
805 	 */
806 	port = inp->inp_fport;
807 	addr = inp->inp_faddr;
808 
809 out:
810 	TCPDEBUG2(PRU_ACCEPT);
811 	TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
812 	INP_WUNLOCK(inp);
813 	if (error == 0)
814 		*nam = in_sockaddr(port, &addr);
815 	return error;
816 }
817 #endif /* INET */
818 
819 #ifdef INET6
820 static int
821 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
822 {
823 	struct inpcb *inp = NULL;
824 	int error = 0;
825 #ifdef KDTRACE_HOOKS
826 	struct tcpcb *tp = NULL;
827 #endif
828 	struct in_addr addr;
829 	struct in6_addr addr6;
830 	struct epoch_tracker et;
831 	in_port_t port = 0;
832 	int v4 = 0;
833 	TCPDEBUG0;
834 
835 	if (so->so_state & SS_ISDISCONNECTED)
836 		return (ECONNABORTED);
837 
838 	inp = sotoinpcb(so);
839 	KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
840 	NET_EPOCH_ENTER(et);
841 	INP_WLOCK(inp);
842 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
843 		error = ECONNABORTED;
844 		goto out;
845 	}
846 #ifdef KDTRACE_HOOKS
847 	tp = intotcpcb(inp);
848 #endif
849 	TCPDEBUG1();
850 
851 	/*
852 	 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
853 	 * copy the data of interest and defer the malloc until after we
854 	 * release the lock.
855 	 */
856 	if (inp->inp_vflag & INP_IPV4) {
857 		v4 = 1;
858 		port = inp->inp_fport;
859 		addr = inp->inp_faddr;
860 	} else {
861 		port = inp->inp_fport;
862 		addr6 = inp->in6p_faddr;
863 	}
864 
865 out:
866 	TCPDEBUG2(PRU_ACCEPT);
867 	TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
868 	INP_WUNLOCK(inp);
869 	NET_EPOCH_EXIT(et);
870 	if (error == 0) {
871 		if (v4)
872 			*nam = in6_v4mapsin6_sockaddr(port, &addr);
873 		else
874 			*nam = in6_sockaddr(port, &addr6);
875 	}
876 	return error;
877 }
878 #endif /* INET6 */
879 
880 /*
881  * Mark the connection as being incapable of further output.
882  */
883 static int
884 tcp_usr_shutdown(struct socket *so)
885 {
886 	int error = 0;
887 	struct inpcb *inp;
888 	struct tcpcb *tp = NULL;
889 	struct epoch_tracker et;
890 
891 	TCPDEBUG0;
892 	inp = sotoinpcb(so);
893 	KASSERT(inp != NULL, ("inp == NULL"));
894 	INP_WLOCK(inp);
895 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
896 		INP_WUNLOCK(inp);
897 		return (ECONNRESET);
898 	}
899 	tp = intotcpcb(inp);
900 	NET_EPOCH_ENTER(et);
901 	TCPDEBUG1();
902 	socantsendmore(so);
903 	tcp_usrclosed(tp);
904 	if (!(inp->inp_flags & INP_DROPPED))
905 		error = tcp_output_nodrop(tp);
906 	TCPDEBUG2(PRU_SHUTDOWN);
907 	TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN);
908 	error = tcp_unlock_or_drop(tp, error);
909 	NET_EPOCH_EXIT(et);
910 
911 	return (error);
912 }
913 
914 /*
915  * After a receive, possibly send window update to peer.
916  */
917 static int
918 tcp_usr_rcvd(struct socket *so, int flags)
919 {
920 	struct epoch_tracker et;
921 	struct inpcb *inp;
922 	struct tcpcb *tp = NULL;
923 	int outrv = 0, error = 0;
924 
925 	TCPDEBUG0;
926 	inp = sotoinpcb(so);
927 	KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
928 	INP_WLOCK(inp);
929 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
930 		INP_WUNLOCK(inp);
931 		return (ECONNRESET);
932 	}
933 	tp = intotcpcb(inp);
934 	NET_EPOCH_ENTER(et);
935 	TCPDEBUG1();
936 	/*
937 	 * For passively-created TFO connections, don't attempt a window
938 	 * update while still in SYN_RECEIVED as this may trigger an early
939 	 * SYN|ACK.  It is preferable to have the SYN|ACK be sent along with
940 	 * application response data, or failing that, when the DELACK timer
941 	 * expires.
942 	 */
943 	if (IS_FASTOPEN(tp->t_flags) &&
944 	    (tp->t_state == TCPS_SYN_RECEIVED))
945 		goto out;
946 #ifdef TCP_OFFLOAD
947 	if (tp->t_flags & TF_TOE)
948 		tcp_offload_rcvd(tp);
949 	else
950 #endif
951 		outrv = tcp_output_nodrop(tp);
952 out:
953 	TCPDEBUG2(PRU_RCVD);
954 	TCP_PROBE2(debug__user, tp, PRU_RCVD);
955 	(void) tcp_unlock_or_drop(tp, outrv);
956 	NET_EPOCH_EXIT(et);
957 	return (error);
958 }
959 
960 /*
961  * Do a send by putting data in output queue and updating urgent
962  * marker if URG set.  Possibly send more data.  Unlike the other
963  * pru_*() routines, the mbuf chains are our responsibility.  We
964  * must either enqueue them or free them.  The other pru_* routines
965  * generally are caller-frees.
966  */
967 static int
968 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
969     struct sockaddr *nam, struct mbuf *control, struct thread *td)
970 {
971 	struct epoch_tracker et;
972 	int error = 0;
973 	struct inpcb *inp;
974 	struct tcpcb *tp = NULL;
975 #ifdef INET
976 #ifdef INET6
977 	struct sockaddr_in sin;
978 #endif
979 	struct sockaddr_in *sinp;
980 #endif
981 #ifdef INET6
982 	int isipv6;
983 #endif
984 	u_int8_t incflagsav;
985 	u_char vflagsav;
986 	bool restoreflags;
987 	TCPDEBUG0;
988 
989 	if (control != NULL) {
990 		/* TCP doesn't do control messages (rights, creds, etc) */
991 		if (control->m_len) {
992 			m_freem(control);
993 			return (EINVAL);
994 		}
995 		m_freem(control);	/* empty control, just free it */
996 	}
997 
998 	inp = sotoinpcb(so);
999 	KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
1000 	INP_WLOCK(inp);
1001 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1002 		if (m != NULL && (flags & PRUS_NOTREADY) == 0)
1003 			m_freem(m);
1004 		INP_WUNLOCK(inp);
1005 		return (ECONNRESET);
1006 	}
1007 
1008 	vflagsav = inp->inp_vflag;
1009 	incflagsav = inp->inp_inc.inc_flags;
1010 	restoreflags = false;
1011 	tp = intotcpcb(inp);
1012 
1013 	NET_EPOCH_ENTER(et);
1014 	if ((flags & PRUS_OOB) != 0 &&
1015 	    (error = tcp_pru_options_support(tp, PRUS_OOB)) != 0)
1016 		goto out;
1017 
1018 	TCPDEBUG1();
1019 	if (nam != NULL && tp->t_state < TCPS_SYN_SENT) {
1020 		if (tp->t_state == TCPS_LISTEN) {
1021 			error = EINVAL;
1022 			goto out;
1023 		}
1024 		switch (nam->sa_family) {
1025 #ifdef INET
1026 		case AF_INET:
1027 			sinp = (struct sockaddr_in *)nam;
1028 			if (sinp->sin_len != sizeof(struct sockaddr_in)) {
1029 				error = EINVAL;
1030 				goto out;
1031 			}
1032 			if ((inp->inp_vflag & INP_IPV6) != 0) {
1033 				error = EAFNOSUPPORT;
1034 				goto out;
1035 			}
1036 			if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
1037 				error = EAFNOSUPPORT;
1038 				goto out;
1039 			}
1040 			if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) {
1041 				error = EACCES;
1042 				goto out;
1043 			}
1044 			if ((error = prison_remote_ip4(td->td_ucred,
1045 			    &sinp->sin_addr)))
1046 				goto out;
1047 #ifdef INET6
1048 			isipv6 = 0;
1049 #endif
1050 			break;
1051 #endif /* INET */
1052 #ifdef INET6
1053 		case AF_INET6:
1054 		{
1055 			struct sockaddr_in6 *sin6;
1056 
1057 			sin6 = (struct sockaddr_in6 *)nam;
1058 			if (sin6->sin6_len != sizeof(*sin6)) {
1059 				error = EINVAL;
1060 				goto out;
1061 			}
1062 			if ((inp->inp_vflag & INP_IPV6PROTO) == 0) {
1063 				error = EAFNOSUPPORT;
1064 				goto out;
1065 			}
1066 			if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
1067 				error = EAFNOSUPPORT;
1068 				goto out;
1069 			}
1070 			if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1071 #ifdef INET
1072 				if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1073 					error = EINVAL;
1074 					goto out;
1075 				}
1076 				if ((inp->inp_vflag & INP_IPV4) == 0) {
1077 					error = EAFNOSUPPORT;
1078 					goto out;
1079 				}
1080 				restoreflags = true;
1081 				inp->inp_vflag &= ~INP_IPV6;
1082 				sinp = &sin;
1083 				in6_sin6_2_sin(sinp, sin6);
1084 				if (IN_MULTICAST(
1085 				    ntohl(sinp->sin_addr.s_addr))) {
1086 					error = EAFNOSUPPORT;
1087 					goto out;
1088 				}
1089 				if ((error = prison_remote_ip4(td->td_ucred,
1090 				    &sinp->sin_addr)))
1091 					goto out;
1092 				isipv6 = 0;
1093 #else /* !INET */
1094 				error = EAFNOSUPPORT;
1095 				goto out;
1096 #endif /* INET */
1097 			} else {
1098 				if ((inp->inp_vflag & INP_IPV6) == 0) {
1099 					error = EAFNOSUPPORT;
1100 					goto out;
1101 				}
1102 				restoreflags = true;
1103 				inp->inp_vflag &= ~INP_IPV4;
1104 				inp->inp_inc.inc_flags |= INC_ISIPV6;
1105 				if ((error = prison_remote_ip6(td->td_ucred,
1106 				    &sin6->sin6_addr)))
1107 					goto out;
1108 				isipv6 = 1;
1109 			}
1110 			break;
1111 		}
1112 #endif /* INET6 */
1113 		default:
1114 			error = EAFNOSUPPORT;
1115 			goto out;
1116 		}
1117 	}
1118 	if (!(flags & PRUS_OOB)) {
1119 		sbappendstream(&so->so_snd, m, flags);
1120 		m = NULL;
1121 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1122 			KASSERT(tp->t_state == TCPS_CLOSED,
1123 			    ("%s: tp %p is listening", __func__, tp));
1124 
1125 			/*
1126 			 * Do implied connect if not yet connected,
1127 			 * initialize window to default value, and
1128 			 * initialize maxseg using peer's cached MSS.
1129 			 */
1130 #ifdef INET6
1131 			if (isipv6)
1132 				error = tcp6_connect(tp, nam, td);
1133 #endif /* INET6 */
1134 #if defined(INET6) && defined(INET)
1135 			else
1136 #endif
1137 #ifdef INET
1138 				error = tcp_connect(tp,
1139 				    (struct sockaddr *)sinp, td);
1140 #endif
1141 			/*
1142 			 * The bind operation in tcp_connect succeeded. We
1143 			 * no longer want to restore the flags if later
1144 			 * operations fail.
1145 			 */
1146 			if (error == 0 || inp->inp_lport != 0)
1147 				restoreflags = false;
1148 
1149 			if (error) {
1150 				/* m is freed if PRUS_NOTREADY is unset. */
1151 				sbflush(&so->so_snd);
1152 				goto out;
1153 			}
1154 			if (IS_FASTOPEN(tp->t_flags))
1155 				tcp_fastopen_connect(tp);
1156 			else {
1157 				tp->snd_wnd = TTCP_CLIENT_SND_WND;
1158 				tcp_mss(tp, -1);
1159 			}
1160 		}
1161 		if (flags & PRUS_EOF) {
1162 			/*
1163 			 * Close the send side of the connection after
1164 			 * the data is sent.
1165 			 */
1166 			socantsendmore(so);
1167 			tcp_usrclosed(tp);
1168 		}
1169 		if (TCPS_HAVEESTABLISHED(tp->t_state) &&
1170 		    ((tp->t_flags2 & TF2_FBYTES_COMPLETE) == 0) &&
1171 		    (tp->t_fbyte_out == 0) &&
1172 		    (so->so_snd.sb_ccc > 0)) {
1173 			tp->t_fbyte_out = ticks;
1174 			if (tp->t_fbyte_out == 0)
1175 				tp->t_fbyte_out = 1;
1176 			if (tp->t_fbyte_out && tp->t_fbyte_in)
1177 				tp->t_flags2 |= TF2_FBYTES_COMPLETE;
1178 		}
1179 		if (!(inp->inp_flags & INP_DROPPED) &&
1180 		    !(flags & PRUS_NOTREADY)) {
1181 			if (flags & PRUS_MORETOCOME)
1182 				tp->t_flags |= TF_MORETOCOME;
1183 			error = tcp_output_nodrop(tp);
1184 			if (flags & PRUS_MORETOCOME)
1185 				tp->t_flags &= ~TF_MORETOCOME;
1186 		}
1187 	} else {
1188 		/*
1189 		 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
1190 		 */
1191 		SOCKBUF_LOCK(&so->so_snd);
1192 		if (sbspace(&so->so_snd) < -512) {
1193 			SOCKBUF_UNLOCK(&so->so_snd);
1194 			error = ENOBUFS;
1195 			goto out;
1196 		}
1197 		/*
1198 		 * According to RFC961 (Assigned Protocols),
1199 		 * the urgent pointer points to the last octet
1200 		 * of urgent data.  We continue, however,
1201 		 * to consider it to indicate the first octet
1202 		 * of data past the urgent section.
1203 		 * Otherwise, snd_up should be one lower.
1204 		 */
1205 		sbappendstream_locked(&so->so_snd, m, flags);
1206 		SOCKBUF_UNLOCK(&so->so_snd);
1207 		m = NULL;
1208 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1209 			/*
1210 			 * Do implied connect if not yet connected,
1211 			 * initialize window to default value, and
1212 			 * initialize maxseg using peer's cached MSS.
1213 			 */
1214 
1215 			/*
1216 			 * Not going to contemplate SYN|URG
1217 			 */
1218 			if (IS_FASTOPEN(tp->t_flags))
1219 				tp->t_flags &= ~TF_FASTOPEN;
1220 #ifdef INET6
1221 			if (isipv6)
1222 				error = tcp6_connect(tp, nam, td);
1223 #endif /* INET6 */
1224 #if defined(INET6) && defined(INET)
1225 			else
1226 #endif
1227 #ifdef INET
1228 				error = tcp_connect(tp,
1229 				    (struct sockaddr *)sinp, td);
1230 #endif
1231 			/*
1232 			 * The bind operation in tcp_connect succeeded. We
1233 			 * no longer want to restore the flags if later
1234 			 * operations fail.
1235 			 */
1236 			if (error == 0 || inp->inp_lport != 0)
1237 				restoreflags = false;
1238 
1239 			if (error != 0) {
1240 				/* m is freed if PRUS_NOTREADY is unset. */
1241 				sbflush(&so->so_snd);
1242 				goto out;
1243 			}
1244 			tp->snd_wnd = TTCP_CLIENT_SND_WND;
1245 			tcp_mss(tp, -1);
1246 		}
1247 		tp->snd_up = tp->snd_una + sbavail(&so->so_snd);
1248 		if ((flags & PRUS_NOTREADY) == 0) {
1249 			tp->t_flags |= TF_FORCEDATA;
1250 			error = tcp_output_nodrop(tp);
1251 			tp->t_flags &= ~TF_FORCEDATA;
1252 		}
1253 	}
1254 	TCP_LOG_EVENT(tp, NULL,
1255 	    &inp->inp_socket->so_rcv,
1256 	    &inp->inp_socket->so_snd,
1257 	    TCP_LOG_USERSEND, error,
1258 	    0, NULL, false);
1259 
1260 out:
1261 	/*
1262 	 * In case of PRUS_NOTREADY, the caller or tcp_usr_ready() is
1263 	 * responsible for freeing memory.
1264 	 */
1265 	if (m != NULL && (flags & PRUS_NOTREADY) == 0)
1266 		m_freem(m);
1267 
1268 	/*
1269 	 * If the request was unsuccessful and we changed flags,
1270 	 * restore the original flags.
1271 	 */
1272 	if (error != 0 && restoreflags) {
1273 		inp->inp_vflag = vflagsav;
1274 		inp->inp_inc.inc_flags = incflagsav;
1275 	}
1276 	TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
1277 		  ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1278 	TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB :
1279 		   ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1280 	error = tcp_unlock_or_drop(tp, error);
1281 	NET_EPOCH_EXIT(et);
1282 	return (error);
1283 }
1284 
1285 static int
1286 tcp_usr_ready(struct socket *so, struct mbuf *m, int count)
1287 {
1288 	struct epoch_tracker et;
1289 	struct inpcb *inp;
1290 	struct tcpcb *tp;
1291 	int error;
1292 
1293 	inp = sotoinpcb(so);
1294 	INP_WLOCK(inp);
1295 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1296 		INP_WUNLOCK(inp);
1297 		mb_free_notready(m, count);
1298 		return (ECONNRESET);
1299 	}
1300 	tp = intotcpcb(inp);
1301 
1302 	SOCKBUF_LOCK(&so->so_snd);
1303 	error = sbready(&so->so_snd, m, count);
1304 	SOCKBUF_UNLOCK(&so->so_snd);
1305 	if (error) {
1306 		INP_WUNLOCK(inp);
1307 		return (error);
1308 	}
1309 	NET_EPOCH_ENTER(et);
1310 	error = tcp_output_unlock(tp);
1311 	NET_EPOCH_EXIT(et);
1312 
1313 	return (error);
1314 }
1315 
1316 /*
1317  * Abort the TCP.  Drop the connection abruptly.
1318  */
1319 static void
1320 tcp_usr_abort(struct socket *so)
1321 {
1322 	struct inpcb *inp;
1323 	struct tcpcb *tp = NULL;
1324 	struct epoch_tracker et;
1325 	TCPDEBUG0;
1326 
1327 	inp = sotoinpcb(so);
1328 	KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
1329 
1330 	NET_EPOCH_ENTER(et);
1331 	INP_WLOCK(inp);
1332 	KASSERT(inp->inp_socket != NULL,
1333 	    ("tcp_usr_abort: inp_socket == NULL"));
1334 
1335 	/*
1336 	 * If we still have full TCP state, and we're not dropped, drop.
1337 	 */
1338 	if (!(inp->inp_flags & INP_TIMEWAIT) &&
1339 	    !(inp->inp_flags & INP_DROPPED)) {
1340 		tp = intotcpcb(inp);
1341 		TCPDEBUG1();
1342 		tp = tcp_drop(tp, ECONNABORTED);
1343 		if (tp == NULL)
1344 			goto dropped;
1345 		TCPDEBUG2(PRU_ABORT);
1346 		TCP_PROBE2(debug__user, tp, PRU_ABORT);
1347 	}
1348 	if (!(inp->inp_flags & INP_DROPPED)) {
1349 		soref(so);
1350 		inp->inp_flags |= INP_SOCKREF;
1351 	}
1352 	INP_WUNLOCK(inp);
1353 dropped:
1354 	NET_EPOCH_EXIT(et);
1355 }
1356 
1357 /*
1358  * TCP socket is closed.  Start friendly disconnect.
1359  */
1360 static void
1361 tcp_usr_close(struct socket *so)
1362 {
1363 	struct inpcb *inp;
1364 	struct tcpcb *tp = NULL;
1365 	struct epoch_tracker et;
1366 	TCPDEBUG0;
1367 
1368 	inp = sotoinpcb(so);
1369 	KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
1370 
1371 	NET_EPOCH_ENTER(et);
1372 	INP_WLOCK(inp);
1373 	KASSERT(inp->inp_socket != NULL,
1374 	    ("tcp_usr_close: inp_socket == NULL"));
1375 
1376 	/*
1377 	 * If we still have full TCP state, and we're not dropped, initiate
1378 	 * a disconnect.
1379 	 */
1380 	if (!(inp->inp_flags & INP_TIMEWAIT) &&
1381 	    !(inp->inp_flags & INP_DROPPED)) {
1382 		tp = intotcpcb(inp);
1383 		tp->t_flags |= TF_CLOSED;
1384 		TCPDEBUG1();
1385 		tcp_disconnect(tp);
1386 		TCPDEBUG2(PRU_CLOSE);
1387 		TCP_PROBE2(debug__user, tp, PRU_CLOSE);
1388 	}
1389 	if (!(inp->inp_flags & INP_DROPPED)) {
1390 		soref(so);
1391 		inp->inp_flags |= INP_SOCKREF;
1392 	}
1393 	INP_WUNLOCK(inp);
1394 	NET_EPOCH_EXIT(et);
1395 }
1396 
1397 static int
1398 tcp_pru_options_support(struct tcpcb *tp, int flags)
1399 {
1400 	/*
1401 	 * If the specific TCP stack has a pru_options
1402 	 * specified then it does not always support
1403 	 * all the PRU_XX options and we must ask it.
1404 	 * If the function is not specified then all
1405 	 * of the PRU_XX options are supported.
1406 	 */
1407 	int ret = 0;
1408 
1409 	if (tp->t_fb->tfb_pru_options) {
1410 		ret = (*tp->t_fb->tfb_pru_options)(tp, flags);
1411 	}
1412 	return (ret);
1413 }
1414 
1415 /*
1416  * Receive out-of-band data.
1417  */
1418 static int
1419 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1420 {
1421 	int error = 0;
1422 	struct inpcb *inp;
1423 	struct tcpcb *tp = NULL;
1424 
1425 	TCPDEBUG0;
1426 	inp = sotoinpcb(so);
1427 	KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1428 	INP_WLOCK(inp);
1429 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1430 		error = ECONNRESET;
1431 		goto out;
1432 	}
1433 	tp = intotcpcb(inp);
1434 	error = tcp_pru_options_support(tp, PRUS_OOB);
1435 	if (error) {
1436 		goto out;
1437 	}
1438 	TCPDEBUG1();
1439 	if ((so->so_oobmark == 0 &&
1440 	     (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1441 	    so->so_options & SO_OOBINLINE ||
1442 	    tp->t_oobflags & TCPOOB_HADDATA) {
1443 		error = EINVAL;
1444 		goto out;
1445 	}
1446 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1447 		error = EWOULDBLOCK;
1448 		goto out;
1449 	}
1450 	m->m_len = 1;
1451 	*mtod(m, caddr_t) = tp->t_iobc;
1452 	if ((flags & MSG_PEEK) == 0)
1453 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1454 
1455 out:
1456 	TCPDEBUG2(PRU_RCVOOB);
1457 	TCP_PROBE2(debug__user, tp, PRU_RCVOOB);
1458 	INP_WUNLOCK(inp);
1459 	return (error);
1460 }
1461 
1462 #ifdef INET
1463 struct protosw tcp_protosw = {
1464 	.pr_type =		SOCK_STREAM,
1465 	.pr_protocol =		IPPROTO_TCP,
1466 	.pr_flags =		PR_CONNREQUIRED | PR_IMPLOPCL | PR_WANTRCVD |
1467 				    PR_CAPATTACH,
1468 	.pr_ctloutput =		tcp_ctloutput,
1469 	.pr_abort =		tcp_usr_abort,
1470 	.pr_accept =		tcp_usr_accept,
1471 	.pr_attach =		tcp_usr_attach,
1472 	.pr_bind =		tcp_usr_bind,
1473 	.pr_connect =		tcp_usr_connect,
1474 	.pr_control =		in_control,
1475 	.pr_detach =		tcp_usr_detach,
1476 	.pr_disconnect =	tcp_usr_disconnect,
1477 	.pr_listen =		tcp_usr_listen,
1478 	.pr_peeraddr =		in_getpeeraddr,
1479 	.pr_rcvd =		tcp_usr_rcvd,
1480 	.pr_rcvoob =		tcp_usr_rcvoob,
1481 	.pr_send =		tcp_usr_send,
1482 	.pr_ready =		tcp_usr_ready,
1483 	.pr_shutdown =		tcp_usr_shutdown,
1484 	.pr_sockaddr =		in_getsockaddr,
1485 	.pr_sosetlabel =	in_pcbsosetlabel,
1486 	.pr_close =		tcp_usr_close,
1487 };
1488 #endif /* INET */
1489 
1490 #ifdef INET6
1491 struct protosw tcp6_protosw = {
1492 	.pr_type =		SOCK_STREAM,
1493 	.pr_protocol =		IPPROTO_TCP,
1494 	.pr_flags =		PR_CONNREQUIRED | PR_IMPLOPCL |PR_WANTRCVD |
1495 				    PR_CAPATTACH,
1496 	.pr_ctloutput =		tcp_ctloutput,
1497 	.pr_abort =		tcp_usr_abort,
1498 	.pr_accept =		tcp6_usr_accept,
1499 	.pr_attach =		tcp_usr_attach,
1500 	.pr_bind =		tcp6_usr_bind,
1501 	.pr_connect =		tcp6_usr_connect,
1502 	.pr_control =		in6_control,
1503 	.pr_detach =		tcp_usr_detach,
1504 	.pr_disconnect =	tcp_usr_disconnect,
1505 	.pr_listen =		tcp6_usr_listen,
1506 	.pr_peeraddr =		in6_mapped_peeraddr,
1507 	.pr_rcvd =		tcp_usr_rcvd,
1508 	.pr_rcvoob =		tcp_usr_rcvoob,
1509 	.pr_send =		tcp_usr_send,
1510 	.pr_ready =		tcp_usr_ready,
1511 	.pr_shutdown =		tcp_usr_shutdown,
1512 	.pr_sockaddr =		in6_mapped_sockaddr,
1513 	.pr_sosetlabel =	in_pcbsosetlabel,
1514 	.pr_close =		tcp_usr_close,
1515 };
1516 #endif /* INET6 */
1517 
1518 #ifdef INET
1519 /*
1520  * Common subroutine to open a TCP connection to remote host specified
1521  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
1522  * port number if needed.  Call in_pcbconnect_setup to do the routing and
1523  * to choose a local host address (interface).  If there is an existing
1524  * incarnation of the same connection in TIME-WAIT state and if the remote
1525  * host was sending CC options and if the connection duration was < MSL, then
1526  * truncate the previous TIME-WAIT state and proceed.
1527  * Initialize connection parameters and enter SYN-SENT state.
1528  */
1529 static int
1530 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1531 {
1532 	struct inpcb *inp = tp->t_inpcb, *oinp;
1533 	struct socket *so = inp->inp_socket;
1534 	struct in_addr laddr;
1535 	u_short lport;
1536 	int error;
1537 
1538 	NET_EPOCH_ASSERT();
1539 	INP_WLOCK_ASSERT(inp);
1540 	INP_HASH_WLOCK(&V_tcbinfo);
1541 
1542 	if (V_tcp_require_unique_port && inp->inp_lport == 0) {
1543 		error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1544 		if (error)
1545 			goto out;
1546 	}
1547 
1548 	/*
1549 	 * Cannot simply call in_pcbconnect, because there might be an
1550 	 * earlier incarnation of this same connection still in
1551 	 * TIME_WAIT state, creating an ADDRINUSE error.
1552 	 */
1553 	laddr = inp->inp_laddr;
1554 	lport = inp->inp_lport;
1555 	error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1556 	    &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1557 	if (error && oinp == NULL)
1558 		goto out;
1559 	if (oinp) {
1560 		error = EADDRINUSE;
1561 		goto out;
1562 	}
1563 	/* Handle initial bind if it hadn't been done in advance. */
1564 	if (inp->inp_lport == 0) {
1565 		inp->inp_lport = lport;
1566 		if (in_pcbinshash(inp) != 0) {
1567 			inp->inp_lport = 0;
1568 			error = EAGAIN;
1569 			goto out;
1570 		}
1571 	}
1572 	inp->inp_laddr = laddr;
1573 	in_pcbrehash(inp);
1574 	INP_HASH_WUNLOCK(&V_tcbinfo);
1575 
1576 	/*
1577 	 * Compute window scaling to request:
1578 	 * Scale to fit into sweet spot.  See tcp_syncache.c.
1579 	 * XXX: This should move to tcp_output().
1580 	 */
1581 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1582 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1583 		tp->request_r_scale++;
1584 
1585 	soisconnecting(so);
1586 	TCPSTAT_INC(tcps_connattempt);
1587 	tcp_state_change(tp, TCPS_SYN_SENT);
1588 	tp->iss = tcp_new_isn(&inp->inp_inc);
1589 	if (tp->t_flags & TF_REQ_TSTMP)
1590 		tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
1591 	tcp_sendseqinit(tp);
1592 
1593 	return 0;
1594 
1595 out:
1596 	INP_HASH_WUNLOCK(&V_tcbinfo);
1597 	return (error);
1598 }
1599 #endif /* INET */
1600 
1601 #ifdef INET6
1602 static int
1603 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1604 {
1605 	struct inpcb *inp = tp->t_inpcb;
1606 	int error;
1607 
1608 	INP_WLOCK_ASSERT(inp);
1609 	INP_HASH_WLOCK(&V_tcbinfo);
1610 
1611 	if (V_tcp_require_unique_port && inp->inp_lport == 0) {
1612 		error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1613 		if (error)
1614 			goto out;
1615 	}
1616 	error = in6_pcbconnect(inp, nam, td->td_ucred);
1617 	if (error != 0)
1618 		goto out;
1619 	INP_HASH_WUNLOCK(&V_tcbinfo);
1620 
1621 	/* Compute window scaling to request.  */
1622 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1623 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1624 		tp->request_r_scale++;
1625 
1626 	soisconnecting(inp->inp_socket);
1627 	TCPSTAT_INC(tcps_connattempt);
1628 	tcp_state_change(tp, TCPS_SYN_SENT);
1629 	tp->iss = tcp_new_isn(&inp->inp_inc);
1630 	if (tp->t_flags & TF_REQ_TSTMP)
1631 		tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
1632 	tcp_sendseqinit(tp);
1633 
1634 	return 0;
1635 
1636 out:
1637 	INP_HASH_WUNLOCK(&V_tcbinfo);
1638 	return error;
1639 }
1640 #endif /* INET6 */
1641 
1642 /*
1643  * Export TCP internal state information via a struct tcp_info, based on the
1644  * Linux 2.6 API.  Not ABI compatible as our constants are mapped differently
1645  * (TCP state machine, etc).  We export all information using FreeBSD-native
1646  * constants -- for example, the numeric values for tcpi_state will differ
1647  * from Linux.
1648  */
1649 static void
1650 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1651 {
1652 
1653 	INP_WLOCK_ASSERT(tp->t_inpcb);
1654 	bzero(ti, sizeof(*ti));
1655 
1656 	ti->tcpi_state = tp->t_state;
1657 	if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1658 		ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1659 	if (tp->t_flags & TF_SACK_PERMIT)
1660 		ti->tcpi_options |= TCPI_OPT_SACK;
1661 	if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1662 		ti->tcpi_options |= TCPI_OPT_WSCALE;
1663 		ti->tcpi_snd_wscale = tp->snd_scale;
1664 		ti->tcpi_rcv_wscale = tp->rcv_scale;
1665 	}
1666 	if (tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT))
1667 		ti->tcpi_options |= TCPI_OPT_ECN;
1668 
1669 	ti->tcpi_rto = tp->t_rxtcur * tick;
1670 	ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick;
1671 	ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1672 	ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1673 
1674 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1675 	ti->tcpi_snd_cwnd = tp->snd_cwnd;
1676 
1677 	/*
1678 	 * FreeBSD-specific extension fields for tcp_info.
1679 	 */
1680 	ti->tcpi_rcv_space = tp->rcv_wnd;
1681 	ti->tcpi_rcv_nxt = tp->rcv_nxt;
1682 	ti->tcpi_snd_wnd = tp->snd_wnd;
1683 	ti->tcpi_snd_bwnd = 0;		/* Unused, kept for compat. */
1684 	ti->tcpi_snd_nxt = tp->snd_nxt;
1685 	ti->tcpi_snd_mss = tp->t_maxseg;
1686 	ti->tcpi_rcv_mss = tp->t_maxseg;
1687 	ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1688 	ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1689 	ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1690 #ifdef TCP_OFFLOAD
1691 	if (tp->t_flags & TF_TOE) {
1692 		ti->tcpi_options |= TCPI_OPT_TOE;
1693 		tcp_offload_tcp_info(tp, ti);
1694 	}
1695 #endif
1696 }
1697 
1698 /*
1699  * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1700  * socket option arguments.  When it re-acquires the lock after the copy, it
1701  * has to revalidate that the connection is still valid for the socket
1702  * option.
1703  */
1704 #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do {			\
1705 	INP_WLOCK(inp);							\
1706 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {		\
1707 		INP_WUNLOCK(inp);					\
1708 		cleanup;						\
1709 		return (ECONNRESET);					\
1710 	}								\
1711 	tp = intotcpcb(inp);						\
1712 } while(0)
1713 #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */)
1714 
1715 int
1716 tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt)
1717 {
1718 	struct socket *so = inp->inp_socket;
1719 	struct tcpcb *tp = intotcpcb(inp);
1720 	int error = 0;
1721 
1722 	MPASS(sopt->sopt_dir == SOPT_SET);
1723 	INP_WLOCK_ASSERT(inp);
1724 	KASSERT((inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) == 0,
1725 	    ("inp_flags == %x", inp->inp_flags));
1726 	KASSERT(so != NULL, ("inp_socket == NULL"));
1727 
1728 	if (sopt->sopt_level != IPPROTO_TCP) {
1729 		INP_WUNLOCK(inp);
1730 #ifdef INET6
1731 		if (inp->inp_vflag & INP_IPV6PROTO)
1732 			error = ip6_ctloutput(so, sopt);
1733 #endif
1734 #if defined(INET6) && defined(INET)
1735 		else
1736 #endif
1737 #ifdef INET
1738 			error = ip_ctloutput(so, sopt);
1739 #endif
1740 		/*
1741 		 * When an IP-level socket option affects TCP, pass control
1742 		 * down to stack tfb_tcp_ctloutput, otherwise return what
1743 		 * IP level returned.
1744 		 */
1745 		switch (sopt->sopt_level) {
1746 #ifdef INET6
1747 		case IPPROTO_IPV6:
1748 			if ((inp->inp_vflag & INP_IPV6PROTO) == 0)
1749 				return (error);
1750 			switch (sopt->sopt_name) {
1751 			case IPV6_TCLASS:
1752 				/* Notify tcp stacks that care (e.g. RACK). */
1753 				break;
1754 			case IPV6_USE_MIN_MTU:
1755 				/* Update t_maxseg accordingly. */
1756 				break;
1757 			default:
1758 				return (error);
1759 			}
1760 			break;
1761 #endif
1762 #ifdef INET
1763 		case IPPROTO_IP:
1764 			switch (sopt->sopt_name) {
1765 			case IP_TOS:
1766 				inp->inp_ip_tos &= ~IPTOS_ECN_MASK;
1767 				break;
1768 			case IP_TTL:
1769 				/* Notify tcp stacks that care (e.g. RACK). */
1770 				break;
1771 			default:
1772 				return (error);
1773 			}
1774 			break;
1775 #endif
1776 		default:
1777 			return (error);
1778 		}
1779 		INP_WLOCK(inp);
1780 		if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1781 			INP_WUNLOCK(inp);
1782 			return (ECONNRESET);
1783 		}
1784 	} else if (sopt->sopt_name == TCP_FUNCTION_BLK) {
1785 		/*
1786 		 * Protect the TCP option TCP_FUNCTION_BLK so
1787 		 * that a sub-function can *never* overwrite this.
1788 		 */
1789 		struct tcp_function_set fsn;
1790 		struct tcp_function_block *blk;
1791 
1792 		INP_WUNLOCK(inp);
1793 		error = sooptcopyin(sopt, &fsn, sizeof fsn, sizeof fsn);
1794 		if (error)
1795 			return (error);
1796 
1797 		INP_WLOCK(inp);
1798 		if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1799 			INP_WUNLOCK(inp);
1800 			return (ECONNRESET);
1801 		}
1802 		tp = intotcpcb(inp);
1803 
1804 		blk = find_and_ref_tcp_functions(&fsn);
1805 		if (blk == NULL) {
1806 			INP_WUNLOCK(inp);
1807 			return (ENOENT);
1808 		}
1809 		if (tp->t_fb == blk) {
1810 			/* You already have this */
1811 			refcount_release(&blk->tfb_refcnt);
1812 			INP_WUNLOCK(inp);
1813 			return (0);
1814 		}
1815 		if (tp->t_state != TCPS_CLOSED) {
1816 			/*
1817 			 * The user has advanced the state
1818 			 * past the initial point, we may not
1819 			 * be able to switch.
1820 			 */
1821 			if (blk->tfb_tcp_handoff_ok != NULL) {
1822 				/*
1823 				 * Does the stack provide a
1824 				 * query mechanism, if so it may
1825 				 * still be possible?
1826 				 */
1827 				error = (*blk->tfb_tcp_handoff_ok)(tp);
1828 			} else
1829 				error = EINVAL;
1830 			if (error) {
1831 				refcount_release(&blk->tfb_refcnt);
1832 				INP_WUNLOCK(inp);
1833 				return(error);
1834 			}
1835 		}
1836 		if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) {
1837 			refcount_release(&blk->tfb_refcnt);
1838 			INP_WUNLOCK(inp);
1839 			return (ENOENT);
1840 		}
1841 		/*
1842 		 * Release the old refcnt, the
1843 		 * lookup acquired a ref on the
1844 		 * new one already.
1845 		 */
1846 		if (tp->t_fb->tfb_tcp_fb_fini) {
1847 			struct epoch_tracker et;
1848 			/*
1849 			 * Tell the stack to cleanup with 0 i.e.
1850 			 * the tcb is not going away.
1851 			 */
1852 			NET_EPOCH_ENTER(et);
1853 			(*tp->t_fb->tfb_tcp_fb_fini)(tp, 0);
1854 			NET_EPOCH_EXIT(et);
1855 		}
1856 #ifdef TCPHPTS
1857 		/* Assure that we are not on any hpts */
1858 		tcp_hpts_remove(tp->t_inpcb);
1859 #endif
1860 		if (blk->tfb_tcp_fb_init) {
1861 			error = (*blk->tfb_tcp_fb_init)(tp);
1862 			if (error) {
1863 				refcount_release(&blk->tfb_refcnt);
1864 				if (tp->t_fb->tfb_tcp_fb_init) {
1865 					if((*tp->t_fb->tfb_tcp_fb_init)(tp) != 0)  {
1866 						/* Fall back failed, drop the connection */
1867 						INP_WUNLOCK(inp);
1868 						soabort(so);
1869 						return (error);
1870 					}
1871 				}
1872 				goto err_out;
1873 			}
1874 		}
1875 		refcount_release(&tp->t_fb->tfb_refcnt);
1876 		tp->t_fb = blk;
1877 #ifdef TCP_OFFLOAD
1878 		if (tp->t_flags & TF_TOE) {
1879 			tcp_offload_ctloutput(tp, sopt->sopt_dir,
1880 			     sopt->sopt_name);
1881 		}
1882 #endif
1883 err_out:
1884 		INP_WUNLOCK(inp);
1885 		return (error);
1886 	}
1887 
1888 	/* Pass in the INP locked, callee must unlock it. */
1889 	return (tp->t_fb->tfb_tcp_ctloutput(inp, sopt));
1890 }
1891 
1892 static int
1893 tcp_ctloutput_get(struct inpcb *inp, struct sockopt *sopt)
1894 {
1895 	struct socket *so = inp->inp_socket;
1896 	struct tcpcb *tp = intotcpcb(inp);
1897 	int error = 0;
1898 
1899 	MPASS(sopt->sopt_dir == SOPT_GET);
1900 	INP_WLOCK_ASSERT(inp);
1901 	KASSERT((inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) == 0,
1902 	    ("inp_flags == %x", inp->inp_flags));
1903 	KASSERT(so != NULL, ("inp_socket == NULL"));
1904 
1905 	if (sopt->sopt_level != IPPROTO_TCP) {
1906 		INP_WUNLOCK(inp);
1907 #ifdef INET6
1908 		if (inp->inp_vflag & INP_IPV6PROTO)
1909 			error = ip6_ctloutput(so, sopt);
1910 #endif /* INET6 */
1911 #if defined(INET6) && defined(INET)
1912 		else
1913 #endif
1914 #ifdef INET
1915 			error = ip_ctloutput(so, sopt);
1916 #endif
1917 		return (error);
1918 	}
1919 	if (((sopt->sopt_name == TCP_FUNCTION_BLK) ||
1920 	     (sopt->sopt_name == TCP_FUNCTION_ALIAS))) {
1921 		struct tcp_function_set fsn;
1922 
1923 		if (sopt->sopt_name == TCP_FUNCTION_ALIAS) {
1924 			memset(&fsn, 0, sizeof(fsn));
1925 			find_tcp_function_alias(tp->t_fb, &fsn);
1926 		} else {
1927 			strncpy(fsn.function_set_name,
1928 			    tp->t_fb->tfb_tcp_block_name,
1929 			    TCP_FUNCTION_NAME_LEN_MAX);
1930 			fsn.function_set_name[TCP_FUNCTION_NAME_LEN_MAX - 1] = '\0';
1931 		}
1932 		fsn.pcbcnt = tp->t_fb->tfb_refcnt;
1933 		INP_WUNLOCK(inp);
1934 		error = sooptcopyout(sopt, &fsn, sizeof fsn);
1935 		return (error);
1936 	}
1937 
1938 	/* Pass in the INP locked, callee must unlock it. */
1939 	return (tp->t_fb->tfb_tcp_ctloutput(inp, sopt));
1940 }
1941 
1942 int
1943 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1944 {
1945 	struct	inpcb *inp;
1946 
1947 	inp = sotoinpcb(so);
1948 	KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1949 
1950 	INP_WLOCK(inp);
1951 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1952 		INP_WUNLOCK(inp);
1953 		return (ECONNRESET);
1954 	}
1955 	if (sopt->sopt_dir == SOPT_SET)
1956 		return (tcp_ctloutput_set(inp, sopt));
1957 	else if (sopt->sopt_dir == SOPT_GET)
1958 		return (tcp_ctloutput_get(inp, sopt));
1959 	else
1960 		panic("%s: sopt_dir $%d", __func__, sopt->sopt_dir);
1961 }
1962 
1963 /*
1964  * If this assert becomes untrue, we need to change the size of the buf
1965  * variable in tcp_default_ctloutput().
1966  */
1967 #ifdef CTASSERT
1968 CTASSERT(TCP_CA_NAME_MAX <= TCP_LOG_ID_LEN);
1969 CTASSERT(TCP_LOG_REASON_LEN <= TCP_LOG_ID_LEN);
1970 #endif
1971 
1972 #ifdef KERN_TLS
1973 static int
1974 copyin_tls_enable(struct sockopt *sopt, struct tls_enable *tls)
1975 {
1976 	struct tls_enable_v0 tls_v0;
1977 	int error;
1978 
1979 	if (sopt->sopt_valsize == sizeof(tls_v0)) {
1980 		error = sooptcopyin(sopt, &tls_v0, sizeof(tls_v0),
1981 		    sizeof(tls_v0));
1982 		if (error)
1983 			return (error);
1984 		memset(tls, 0, sizeof(*tls));
1985 		tls->cipher_key = tls_v0.cipher_key;
1986 		tls->iv = tls_v0.iv;
1987 		tls->auth_key = tls_v0.auth_key;
1988 		tls->cipher_algorithm = tls_v0.cipher_algorithm;
1989 		tls->cipher_key_len = tls_v0.cipher_key_len;
1990 		tls->iv_len = tls_v0.iv_len;
1991 		tls->auth_algorithm = tls_v0.auth_algorithm;
1992 		tls->auth_key_len = tls_v0.auth_key_len;
1993 		tls->flags = tls_v0.flags;
1994 		tls->tls_vmajor = tls_v0.tls_vmajor;
1995 		tls->tls_vminor = tls_v0.tls_vminor;
1996 		return (0);
1997 	}
1998 
1999 	return (sooptcopyin(sopt, tls, sizeof(*tls), sizeof(*tls)));
2000 }
2001 #endif
2002 
2003 extern struct cc_algo newreno_cc_algo;
2004 
2005 static int
2006 tcp_set_cc_mod(struct inpcb *inp, struct sockopt *sopt)
2007 {
2008 	struct cc_algo *algo;
2009 	void *ptr = NULL;
2010 	struct tcpcb *tp;
2011 	struct cc_var cc_mem;
2012 	char	buf[TCP_CA_NAME_MAX];
2013 	size_t mem_sz;
2014 	int error;
2015 
2016 	INP_WUNLOCK(inp);
2017 	error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1);
2018 	if (error)
2019 		return(error);
2020 	buf[sopt->sopt_valsize] = '\0';
2021 	CC_LIST_RLOCK();
2022 	STAILQ_FOREACH(algo, &cc_list, entries) {
2023 		if (strncmp(buf, algo->name,
2024 			    TCP_CA_NAME_MAX) == 0) {
2025 			if (algo->flags & CC_MODULE_BEING_REMOVED) {
2026 				/* We can't "see" modules being unloaded */
2027 				continue;
2028 			}
2029 			break;
2030 		}
2031 	}
2032 	if (algo == NULL) {
2033 		CC_LIST_RUNLOCK();
2034 		return(ESRCH);
2035 	}
2036 	/*
2037 	 * With a reference the algorithm cannot be removed
2038 	 * so we hold a reference through the change process.
2039 	 */
2040 	cc_refer(algo);
2041 	CC_LIST_RUNLOCK();
2042 	if (algo->cb_init != NULL) {
2043 		/* We can now pre-get the memory for the CC */
2044 		mem_sz = (*algo->cc_data_sz)();
2045 		if (mem_sz == 0) {
2046 			goto no_mem_needed;
2047 		}
2048 		ptr = malloc(mem_sz, M_CC_MEM, M_WAITOK);
2049 	} else {
2050 no_mem_needed:
2051 		mem_sz = 0;
2052 		ptr = NULL;
2053 	}
2054 	/*
2055 	 * Make sure its all clean and zero and also get
2056 	 * back the inplock.
2057 	 */
2058 	memset(&cc_mem, 0, sizeof(cc_mem));
2059 	INP_WLOCK(inp);
2060 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
2061 		INP_WUNLOCK(inp);
2062 		if (ptr)
2063 			free(ptr, M_CC_MEM);
2064 		/* Release our temp reference */
2065 		CC_LIST_RLOCK();
2066 		cc_release(algo);
2067 		CC_LIST_RUNLOCK();
2068 		return (ECONNRESET);
2069 	}
2070 	tp = intotcpcb(inp);
2071 	if (ptr != NULL)
2072 		memset(ptr, 0, mem_sz);
2073 	cc_mem.ccvc.tcp = tp;
2074 	/*
2075 	 * We once again hold a write lock over the tcb so it's
2076 	 * safe to do these things without ordering concerns.
2077 	 * Note here we init into stack memory.
2078 	 */
2079 	if (algo->cb_init != NULL)
2080 		error = algo->cb_init(&cc_mem, ptr);
2081 	else
2082 		error = 0;
2083 	/*
2084 	 * The CC algorithms, when given their memory
2085 	 * should not fail we could in theory have a
2086 	 * KASSERT here.
2087 	 */
2088 	if (error == 0) {
2089 		/*
2090 		 * Touchdown, lets go ahead and move the
2091 		 * connection to the new CC module by
2092 		 * copying in the cc_mem after we call
2093 		 * the old ones cleanup (if any).
2094 		 */
2095 		if (CC_ALGO(tp)->cb_destroy != NULL)
2096 			CC_ALGO(tp)->cb_destroy(tp->ccv);
2097 		/* Detach the old CC from the tcpcb  */
2098 		cc_detach(tp);
2099 		/* Copy in our temp memory that was inited */
2100 		memcpy(tp->ccv, &cc_mem, sizeof(struct cc_var));
2101 		/* Now attach the new, which takes a reference */
2102 		cc_attach(tp, algo);
2103 		/* Ok now are we where we have gotten past any conn_init? */
2104 		if (TCPS_HAVEESTABLISHED(tp->t_state) && (CC_ALGO(tp)->conn_init != NULL)) {
2105 			/* Yep run the connection init for the new CC */
2106 			CC_ALGO(tp)->conn_init(tp->ccv);
2107 		}
2108 	} else if (ptr)
2109 		free(ptr, M_CC_MEM);
2110 	INP_WUNLOCK(inp);
2111 	/* Now lets release our temp reference */
2112 	CC_LIST_RLOCK();
2113 	cc_release(algo);
2114 	CC_LIST_RUNLOCK();
2115 	return (error);
2116 }
2117 
2118 int
2119 tcp_default_ctloutput(struct inpcb *inp, struct sockopt *sopt)
2120 {
2121 	struct tcpcb *tp = intotcpcb(inp);
2122 	int	error, opt, optval;
2123 	u_int	ui;
2124 	struct	tcp_info ti;
2125 #ifdef KERN_TLS
2126 	struct tls_enable tls;
2127 	struct socket *so = inp->inp_socket;
2128 #endif
2129 	char	*pbuf, buf[TCP_LOG_ID_LEN];
2130 #ifdef STATS
2131 	struct statsblob *sbp;
2132 #endif
2133 	size_t	len;
2134 
2135 	INP_WLOCK_ASSERT(inp);
2136 	KASSERT((inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) == 0,
2137 	    ("inp_flags == %x", inp->inp_flags));
2138 	KASSERT(inp->inp_socket != NULL, ("inp_socket == NULL"));
2139 
2140 	switch (sopt->sopt_level) {
2141 #ifdef INET6
2142 	case IPPROTO_IPV6:
2143 		MPASS(inp->inp_vflag & INP_IPV6PROTO);
2144 		switch (sopt->sopt_name) {
2145 		case IPV6_USE_MIN_MTU:
2146 			tcp6_use_min_mtu(tp);
2147 			/* FALLTHROUGH */
2148 		}
2149 		INP_WUNLOCK(inp);
2150 		return (0);
2151 #endif
2152 #ifdef INET
2153 	case IPPROTO_IP:
2154 		INP_WUNLOCK(inp);
2155 		return (0);
2156 #endif
2157 	}
2158 
2159 	/*
2160 	 * For TCP_CCALGOOPT forward the control to CC module, for both
2161 	 * SOPT_SET and SOPT_GET.
2162 	 */
2163 	switch (sopt->sopt_name) {
2164 	case TCP_CCALGOOPT:
2165 		INP_WUNLOCK(inp);
2166 		if (sopt->sopt_valsize > CC_ALGOOPT_LIMIT)
2167 			return (EINVAL);
2168 		pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO);
2169 		error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize,
2170 		    sopt->sopt_valsize);
2171 		if (error) {
2172 			free(pbuf, M_TEMP);
2173 			return (error);
2174 		}
2175 		INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP));
2176 		if (CC_ALGO(tp)->ctl_output != NULL)
2177 			error = CC_ALGO(tp)->ctl_output(tp->ccv, sopt, pbuf);
2178 		else
2179 			error = ENOENT;
2180 		INP_WUNLOCK(inp);
2181 		if (error == 0 && sopt->sopt_dir == SOPT_GET)
2182 			error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize);
2183 		free(pbuf, M_TEMP);
2184 		return (error);
2185 	}
2186 
2187 	switch (sopt->sopt_dir) {
2188 	case SOPT_SET:
2189 		switch (sopt->sopt_name) {
2190 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
2191 		case TCP_MD5SIG:
2192 			INP_WUNLOCK(inp);
2193 			if (!TCPMD5_ENABLED())
2194 				return (ENOPROTOOPT);
2195 			error = TCPMD5_PCBCTL(inp, sopt);
2196 			if (error)
2197 				return (error);
2198 			INP_WLOCK_RECHECK(inp);
2199 			goto unlock_and_done;
2200 #endif /* IPSEC */
2201 
2202 		case TCP_NODELAY:
2203 		case TCP_NOOPT:
2204 		case TCP_LRD:
2205 			INP_WUNLOCK(inp);
2206 			error = sooptcopyin(sopt, &optval, sizeof optval,
2207 			    sizeof optval);
2208 			if (error)
2209 				return (error);
2210 
2211 			INP_WLOCK_RECHECK(inp);
2212 			switch (sopt->sopt_name) {
2213 			case TCP_NODELAY:
2214 				opt = TF_NODELAY;
2215 				break;
2216 			case TCP_NOOPT:
2217 				opt = TF_NOOPT;
2218 				break;
2219 			case TCP_LRD:
2220 				opt = TF_LRD;
2221 				break;
2222 			default:
2223 				opt = 0; /* dead code to fool gcc */
2224 				break;
2225 			}
2226 
2227 			if (optval)
2228 				tp->t_flags |= opt;
2229 			else
2230 				tp->t_flags &= ~opt;
2231 unlock_and_done:
2232 #ifdef TCP_OFFLOAD
2233 			if (tp->t_flags & TF_TOE) {
2234 				tcp_offload_ctloutput(tp, sopt->sopt_dir,
2235 				    sopt->sopt_name);
2236 			}
2237 #endif
2238 			INP_WUNLOCK(inp);
2239 			break;
2240 
2241 		case TCP_NOPUSH:
2242 			INP_WUNLOCK(inp);
2243 			error = sooptcopyin(sopt, &optval, sizeof optval,
2244 			    sizeof optval);
2245 			if (error)
2246 				return (error);
2247 
2248 			INP_WLOCK_RECHECK(inp);
2249 			if (optval)
2250 				tp->t_flags |= TF_NOPUSH;
2251 			else if (tp->t_flags & TF_NOPUSH) {
2252 				tp->t_flags &= ~TF_NOPUSH;
2253 				if (TCPS_HAVEESTABLISHED(tp->t_state)) {
2254 					struct epoch_tracker et;
2255 
2256 					NET_EPOCH_ENTER(et);
2257 					error = tcp_output_nodrop(tp);
2258 					NET_EPOCH_EXIT(et);
2259 				}
2260 			}
2261 			goto unlock_and_done;
2262 
2263 		case TCP_REMOTE_UDP_ENCAPS_PORT:
2264 			INP_WUNLOCK(inp);
2265 			error = sooptcopyin(sopt, &optval, sizeof optval,
2266 			    sizeof optval);
2267 			if (error)
2268 				return (error);
2269 			if ((optval < TCP_TUNNELING_PORT_MIN) ||
2270 			    (optval > TCP_TUNNELING_PORT_MAX)) {
2271 				/* Its got to be in range */
2272 				return (EINVAL);
2273 			}
2274 			if ((V_tcp_udp_tunneling_port == 0) && (optval != 0)) {
2275 				/* You have to have enabled a UDP tunneling port first */
2276 				return (EINVAL);
2277 			}
2278 			INP_WLOCK_RECHECK(inp);
2279 			if (tp->t_state != TCPS_CLOSED) {
2280 				/* You can't change after you are connected */
2281 				error = EINVAL;
2282 			} else {
2283 				/* Ok we are all good set the port */
2284 				tp->t_port = htons(optval);
2285 			}
2286 			goto unlock_and_done;
2287 
2288 		case TCP_MAXSEG:
2289 			INP_WUNLOCK(inp);
2290 			error = sooptcopyin(sopt, &optval, sizeof optval,
2291 			    sizeof optval);
2292 			if (error)
2293 				return (error);
2294 
2295 			INP_WLOCK_RECHECK(inp);
2296 			if (optval > 0 && optval <= tp->t_maxseg &&
2297 			    optval + 40 >= V_tcp_minmss)
2298 				tp->t_maxseg = optval;
2299 			else
2300 				error = EINVAL;
2301 			goto unlock_and_done;
2302 
2303 		case TCP_INFO:
2304 			INP_WUNLOCK(inp);
2305 			error = EINVAL;
2306 			break;
2307 
2308 		case TCP_STATS:
2309 			INP_WUNLOCK(inp);
2310 #ifdef STATS
2311 			error = sooptcopyin(sopt, &optval, sizeof optval,
2312 			    sizeof optval);
2313 			if (error)
2314 				return (error);
2315 
2316 			if (optval > 0)
2317 				sbp = stats_blob_alloc(
2318 				    V_tcp_perconn_stats_dflt_tpl, 0);
2319 			else
2320 				sbp = NULL;
2321 
2322 			INP_WLOCK_RECHECK(inp);
2323 			if ((tp->t_stats != NULL && sbp == NULL) ||
2324 			    (tp->t_stats == NULL && sbp != NULL)) {
2325 				struct statsblob *t = tp->t_stats;
2326 				tp->t_stats = sbp;
2327 				sbp = t;
2328 			}
2329 			INP_WUNLOCK(inp);
2330 
2331 			stats_blob_destroy(sbp);
2332 #else
2333 			return (EOPNOTSUPP);
2334 #endif /* !STATS */
2335 			break;
2336 
2337 		case TCP_CONGESTION:
2338 			error = tcp_set_cc_mod(inp, sopt);
2339 			break;
2340 
2341 		case TCP_REUSPORT_LB_NUMA:
2342 			INP_WUNLOCK(inp);
2343 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2344 			    sizeof(optval));
2345 			INP_WLOCK_RECHECK(inp);
2346 			if (!error)
2347 				error = in_pcblbgroup_numa(inp, optval);
2348 			INP_WUNLOCK(inp);
2349 			break;
2350 
2351 #ifdef KERN_TLS
2352 		case TCP_TXTLS_ENABLE:
2353 			INP_WUNLOCK(inp);
2354 			error = copyin_tls_enable(sopt, &tls);
2355 			if (error)
2356 				break;
2357 			error = ktls_enable_tx(so, &tls);
2358 			break;
2359 		case TCP_TXTLS_MODE:
2360 			INP_WUNLOCK(inp);
2361 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2362 			if (error)
2363 				return (error);
2364 
2365 			INP_WLOCK_RECHECK(inp);
2366 			error = ktls_set_tx_mode(so, ui);
2367 			INP_WUNLOCK(inp);
2368 			break;
2369 		case TCP_RXTLS_ENABLE:
2370 			INP_WUNLOCK(inp);
2371 			error = sooptcopyin(sopt, &tls, sizeof(tls),
2372 			    sizeof(tls));
2373 			if (error)
2374 				break;
2375 			error = ktls_enable_rx(so, &tls);
2376 			break;
2377 #endif
2378 
2379 		case TCP_KEEPIDLE:
2380 		case TCP_KEEPINTVL:
2381 		case TCP_KEEPINIT:
2382 			INP_WUNLOCK(inp);
2383 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2384 			if (error)
2385 				return (error);
2386 
2387 			if (ui > (UINT_MAX / hz)) {
2388 				error = EINVAL;
2389 				break;
2390 			}
2391 			ui *= hz;
2392 
2393 			INP_WLOCK_RECHECK(inp);
2394 			switch (sopt->sopt_name) {
2395 			case TCP_KEEPIDLE:
2396 				tp->t_keepidle = ui;
2397 				/*
2398 				 * XXX: better check current remaining
2399 				 * timeout and "merge" it with new value.
2400 				 */
2401 				if ((tp->t_state > TCPS_LISTEN) &&
2402 				    (tp->t_state <= TCPS_CLOSING))
2403 					tcp_timer_activate(tp, TT_KEEP,
2404 					    TP_KEEPIDLE(tp));
2405 				break;
2406 			case TCP_KEEPINTVL:
2407 				tp->t_keepintvl = ui;
2408 				if ((tp->t_state == TCPS_FIN_WAIT_2) &&
2409 				    (TP_MAXIDLE(tp) > 0))
2410 					tcp_timer_activate(tp, TT_2MSL,
2411 					    TP_MAXIDLE(tp));
2412 				break;
2413 			case TCP_KEEPINIT:
2414 				tp->t_keepinit = ui;
2415 				if (tp->t_state == TCPS_SYN_RECEIVED ||
2416 				    tp->t_state == TCPS_SYN_SENT)
2417 					tcp_timer_activate(tp, TT_KEEP,
2418 					    TP_KEEPINIT(tp));
2419 				break;
2420 			}
2421 			goto unlock_and_done;
2422 
2423 		case TCP_KEEPCNT:
2424 			INP_WUNLOCK(inp);
2425 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2426 			if (error)
2427 				return (error);
2428 
2429 			INP_WLOCK_RECHECK(inp);
2430 			tp->t_keepcnt = ui;
2431 			if ((tp->t_state == TCPS_FIN_WAIT_2) &&
2432 			    (TP_MAXIDLE(tp) > 0))
2433 				tcp_timer_activate(tp, TT_2MSL,
2434 				    TP_MAXIDLE(tp));
2435 			goto unlock_and_done;
2436 
2437 #ifdef TCPPCAP
2438 		case TCP_PCAP_OUT:
2439 		case TCP_PCAP_IN:
2440 			INP_WUNLOCK(inp);
2441 			error = sooptcopyin(sopt, &optval, sizeof optval,
2442 			    sizeof optval);
2443 			if (error)
2444 				return (error);
2445 
2446 			INP_WLOCK_RECHECK(inp);
2447 			if (optval >= 0)
2448 				tcp_pcap_set_sock_max(TCP_PCAP_OUT ?
2449 					&(tp->t_outpkts) : &(tp->t_inpkts),
2450 					optval);
2451 			else
2452 				error = EINVAL;
2453 			goto unlock_and_done;
2454 #endif
2455 
2456 		case TCP_FASTOPEN: {
2457 			struct tcp_fastopen tfo_optval;
2458 
2459 			INP_WUNLOCK(inp);
2460 			if (!V_tcp_fastopen_client_enable &&
2461 			    !V_tcp_fastopen_server_enable)
2462 				return (EPERM);
2463 
2464 			error = sooptcopyin(sopt, &tfo_optval,
2465 				    sizeof(tfo_optval), sizeof(int));
2466 			if (error)
2467 				return (error);
2468 
2469 			INP_WLOCK_RECHECK(inp);
2470 			if ((tp->t_state != TCPS_CLOSED) &&
2471 			    (tp->t_state != TCPS_LISTEN)) {
2472 				error = EINVAL;
2473 				goto unlock_and_done;
2474 			}
2475 			if (tfo_optval.enable) {
2476 				if (tp->t_state == TCPS_LISTEN) {
2477 					if (!V_tcp_fastopen_server_enable) {
2478 						error = EPERM;
2479 						goto unlock_and_done;
2480 					}
2481 
2482 					if (tp->t_tfo_pending == NULL)
2483 						tp->t_tfo_pending =
2484 						    tcp_fastopen_alloc_counter();
2485 				} else {
2486 					/*
2487 					 * If a pre-shared key was provided,
2488 					 * stash it in the client cookie
2489 					 * field of the tcpcb for use during
2490 					 * connect.
2491 					 */
2492 					if (sopt->sopt_valsize ==
2493 					    sizeof(tfo_optval)) {
2494 						memcpy(tp->t_tfo_cookie.client,
2495 						       tfo_optval.psk,
2496 						       TCP_FASTOPEN_PSK_LEN);
2497 						tp->t_tfo_client_cookie_len =
2498 						    TCP_FASTOPEN_PSK_LEN;
2499 					}
2500 				}
2501 				tp->t_flags |= TF_FASTOPEN;
2502 			} else
2503 				tp->t_flags &= ~TF_FASTOPEN;
2504 			goto unlock_and_done;
2505 		}
2506 
2507 #ifdef TCP_BLACKBOX
2508 		case TCP_LOG:
2509 			INP_WUNLOCK(inp);
2510 			error = sooptcopyin(sopt, &optval, sizeof optval,
2511 			    sizeof optval);
2512 			if (error)
2513 				return (error);
2514 
2515 			INP_WLOCK_RECHECK(inp);
2516 			error = tcp_log_state_change(tp, optval);
2517 			goto unlock_and_done;
2518 
2519 		case TCP_LOGBUF:
2520 			INP_WUNLOCK(inp);
2521 			error = EINVAL;
2522 			break;
2523 
2524 		case TCP_LOGID:
2525 			INP_WUNLOCK(inp);
2526 			error = sooptcopyin(sopt, buf, TCP_LOG_ID_LEN - 1, 0);
2527 			if (error)
2528 				break;
2529 			buf[sopt->sopt_valsize] = '\0';
2530 			INP_WLOCK_RECHECK(inp);
2531 			error = tcp_log_set_id(tp, buf);
2532 			/* tcp_log_set_id() unlocks the INP. */
2533 			break;
2534 
2535 		case TCP_LOGDUMP:
2536 		case TCP_LOGDUMPID:
2537 			INP_WUNLOCK(inp);
2538 			error =
2539 			    sooptcopyin(sopt, buf, TCP_LOG_REASON_LEN - 1, 0);
2540 			if (error)
2541 				break;
2542 			buf[sopt->sopt_valsize] = '\0';
2543 			INP_WLOCK_RECHECK(inp);
2544 			if (sopt->sopt_name == TCP_LOGDUMP) {
2545 				error = tcp_log_dump_tp_logbuf(tp, buf,
2546 				    M_WAITOK, true);
2547 				INP_WUNLOCK(inp);
2548 			} else {
2549 				tcp_log_dump_tp_bucket_logbufs(tp, buf);
2550 				/*
2551 				 * tcp_log_dump_tp_bucket_logbufs() drops the
2552 				 * INP lock.
2553 				 */
2554 			}
2555 			break;
2556 #endif
2557 
2558 		default:
2559 			INP_WUNLOCK(inp);
2560 			error = ENOPROTOOPT;
2561 			break;
2562 		}
2563 		break;
2564 
2565 	case SOPT_GET:
2566 		tp = intotcpcb(inp);
2567 		switch (sopt->sopt_name) {
2568 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
2569 		case TCP_MD5SIG:
2570 			INP_WUNLOCK(inp);
2571 			if (!TCPMD5_ENABLED())
2572 				return (ENOPROTOOPT);
2573 			error = TCPMD5_PCBCTL(inp, sopt);
2574 			break;
2575 #endif
2576 
2577 		case TCP_NODELAY:
2578 			optval = tp->t_flags & TF_NODELAY;
2579 			INP_WUNLOCK(inp);
2580 			error = sooptcopyout(sopt, &optval, sizeof optval);
2581 			break;
2582 		case TCP_MAXSEG:
2583 			optval = tp->t_maxseg;
2584 			INP_WUNLOCK(inp);
2585 			error = sooptcopyout(sopt, &optval, sizeof optval);
2586 			break;
2587 		case TCP_REMOTE_UDP_ENCAPS_PORT:
2588 			optval = ntohs(tp->t_port);
2589 			INP_WUNLOCK(inp);
2590 			error = sooptcopyout(sopt, &optval, sizeof optval);
2591 			break;
2592 		case TCP_NOOPT:
2593 			optval = tp->t_flags & TF_NOOPT;
2594 			INP_WUNLOCK(inp);
2595 			error = sooptcopyout(sopt, &optval, sizeof optval);
2596 			break;
2597 		case TCP_NOPUSH:
2598 			optval = tp->t_flags & TF_NOPUSH;
2599 			INP_WUNLOCK(inp);
2600 			error = sooptcopyout(sopt, &optval, sizeof optval);
2601 			break;
2602 		case TCP_INFO:
2603 			tcp_fill_info(tp, &ti);
2604 			INP_WUNLOCK(inp);
2605 			error = sooptcopyout(sopt, &ti, sizeof ti);
2606 			break;
2607 		case TCP_STATS:
2608 			{
2609 #ifdef STATS
2610 			int nheld;
2611 			TYPEOF_MEMBER(struct statsblob, flags) sbflags = 0;
2612 
2613 			error = 0;
2614 			socklen_t outsbsz = sopt->sopt_valsize;
2615 			if (tp->t_stats == NULL)
2616 				error = ENOENT;
2617 			else if (outsbsz >= tp->t_stats->cursz)
2618 				outsbsz = tp->t_stats->cursz;
2619 			else if (outsbsz >= sizeof(struct statsblob))
2620 				outsbsz = sizeof(struct statsblob);
2621 			else
2622 				error = EINVAL;
2623 			INP_WUNLOCK(inp);
2624 			if (error)
2625 				break;
2626 
2627 			sbp = sopt->sopt_val;
2628 			nheld = atop(round_page(((vm_offset_t)sbp) +
2629 			    (vm_size_t)outsbsz) - trunc_page((vm_offset_t)sbp));
2630 			vm_page_t ma[nheld];
2631 			if (vm_fault_quick_hold_pages(
2632 			    &curproc->p_vmspace->vm_map, (vm_offset_t)sbp,
2633 			    outsbsz, VM_PROT_READ | VM_PROT_WRITE, ma,
2634 			    nheld) < 0) {
2635 				error = EFAULT;
2636 				break;
2637 			}
2638 
2639 			if ((error = copyin_nofault(&(sbp->flags), &sbflags,
2640 			    SIZEOF_MEMBER(struct statsblob, flags))))
2641 				goto unhold;
2642 
2643 			INP_WLOCK_RECHECK(inp);
2644 			error = stats_blob_snapshot(&sbp, outsbsz, tp->t_stats,
2645 			    sbflags | SB_CLONE_USRDSTNOFAULT);
2646 			INP_WUNLOCK(inp);
2647 			sopt->sopt_valsize = outsbsz;
2648 unhold:
2649 			vm_page_unhold_pages(ma, nheld);
2650 #else
2651 			INP_WUNLOCK(inp);
2652 			error = EOPNOTSUPP;
2653 #endif /* !STATS */
2654 			break;
2655 			}
2656 		case TCP_CONGESTION:
2657 			len = strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
2658 			INP_WUNLOCK(inp);
2659 			error = sooptcopyout(sopt, buf, len + 1);
2660 			break;
2661 		case TCP_KEEPIDLE:
2662 		case TCP_KEEPINTVL:
2663 		case TCP_KEEPINIT:
2664 		case TCP_KEEPCNT:
2665 			switch (sopt->sopt_name) {
2666 			case TCP_KEEPIDLE:
2667 				ui = TP_KEEPIDLE(tp) / hz;
2668 				break;
2669 			case TCP_KEEPINTVL:
2670 				ui = TP_KEEPINTVL(tp) / hz;
2671 				break;
2672 			case TCP_KEEPINIT:
2673 				ui = TP_KEEPINIT(tp) / hz;
2674 				break;
2675 			case TCP_KEEPCNT:
2676 				ui = TP_KEEPCNT(tp);
2677 				break;
2678 			}
2679 			INP_WUNLOCK(inp);
2680 			error = sooptcopyout(sopt, &ui, sizeof(ui));
2681 			break;
2682 #ifdef TCPPCAP
2683 		case TCP_PCAP_OUT:
2684 		case TCP_PCAP_IN:
2685 			optval = tcp_pcap_get_sock_max(TCP_PCAP_OUT ?
2686 					&(tp->t_outpkts) : &(tp->t_inpkts));
2687 			INP_WUNLOCK(inp);
2688 			error = sooptcopyout(sopt, &optval, sizeof optval);
2689 			break;
2690 #endif
2691 		case TCP_FASTOPEN:
2692 			optval = tp->t_flags & TF_FASTOPEN;
2693 			INP_WUNLOCK(inp);
2694 			error = sooptcopyout(sopt, &optval, sizeof optval);
2695 			break;
2696 #ifdef TCP_BLACKBOX
2697 		case TCP_LOG:
2698 			optval = tp->t_logstate;
2699 			INP_WUNLOCK(inp);
2700 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2701 			break;
2702 		case TCP_LOGBUF:
2703 			/* tcp_log_getlogbuf() does INP_WUNLOCK(inp) */
2704 			error = tcp_log_getlogbuf(sopt, tp);
2705 			break;
2706 		case TCP_LOGID:
2707 			len = tcp_log_get_id(tp, buf);
2708 			INP_WUNLOCK(inp);
2709 			error = sooptcopyout(sopt, buf, len + 1);
2710 			break;
2711 		case TCP_LOGDUMP:
2712 		case TCP_LOGDUMPID:
2713 			INP_WUNLOCK(inp);
2714 			error = EINVAL;
2715 			break;
2716 #endif
2717 #ifdef KERN_TLS
2718 		case TCP_TXTLS_MODE:
2719 			error = ktls_get_tx_mode(so, &optval);
2720 			INP_WUNLOCK(inp);
2721 			if (error == 0)
2722 				error = sooptcopyout(sopt, &optval,
2723 				    sizeof(optval));
2724 			break;
2725 		case TCP_RXTLS_MODE:
2726 			error = ktls_get_rx_mode(so, &optval);
2727 			INP_WUNLOCK(inp);
2728 			if (error == 0)
2729 				error = sooptcopyout(sopt, &optval,
2730 				    sizeof(optval));
2731 			break;
2732 #endif
2733 		case TCP_LRD:
2734 			optval = tp->t_flags & TF_LRD;
2735 			INP_WUNLOCK(inp);
2736 			error = sooptcopyout(sopt, &optval, sizeof optval);
2737 			break;
2738 		default:
2739 			INP_WUNLOCK(inp);
2740 			error = ENOPROTOOPT;
2741 			break;
2742 		}
2743 		break;
2744 	}
2745 	return (error);
2746 }
2747 #undef INP_WLOCK_RECHECK
2748 #undef INP_WLOCK_RECHECK_CLEANUP
2749 
2750 /*
2751  * Initiate (or continue) disconnect.
2752  * If embryonic state, just send reset (once).
2753  * If in ``let data drain'' option and linger null, just drop.
2754  * Otherwise (hard), mark socket disconnecting and drop
2755  * current input data; switch states based on user close, and
2756  * send segment to peer (with FIN).
2757  */
2758 static void
2759 tcp_disconnect(struct tcpcb *tp)
2760 {
2761 	struct inpcb *inp = tp->t_inpcb;
2762 	struct socket *so = inp->inp_socket;
2763 
2764 	NET_EPOCH_ASSERT();
2765 	INP_WLOCK_ASSERT(inp);
2766 
2767 	/*
2768 	 * Neither tcp_close() nor tcp_drop() should return NULL, as the
2769 	 * socket is still open.
2770 	 */
2771 	if (tp->t_state < TCPS_ESTABLISHED &&
2772 	    !(tp->t_state > TCPS_LISTEN && IS_FASTOPEN(tp->t_flags))) {
2773 		tp = tcp_close(tp);
2774 		KASSERT(tp != NULL,
2775 		    ("tcp_disconnect: tcp_close() returned NULL"));
2776 	} else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
2777 		tp = tcp_drop(tp, 0);
2778 		KASSERT(tp != NULL,
2779 		    ("tcp_disconnect: tcp_drop() returned NULL"));
2780 	} else {
2781 		soisdisconnecting(so);
2782 		sbflush(&so->so_rcv);
2783 		tcp_usrclosed(tp);
2784 		if (!(inp->inp_flags & INP_DROPPED))
2785 			/* Ignore stack's drop request, we already at it. */
2786 			(void)tcp_output_nodrop(tp);
2787 	}
2788 }
2789 
2790 /*
2791  * User issued close, and wish to trail through shutdown states:
2792  * if never received SYN, just forget it.  If got a SYN from peer,
2793  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
2794  * If already got a FIN from peer, then almost done; go to LAST_ACK
2795  * state.  In all other cases, have already sent FIN to peer (e.g.
2796  * after PRU_SHUTDOWN), and just have to play tedious game waiting
2797  * for peer to send FIN or not respond to keep-alives, etc.
2798  * We can let the user exit from the close as soon as the FIN is acked.
2799  */
2800 static void
2801 tcp_usrclosed(struct tcpcb *tp)
2802 {
2803 
2804 	NET_EPOCH_ASSERT();
2805 	INP_WLOCK_ASSERT(tp->t_inpcb);
2806 
2807 	switch (tp->t_state) {
2808 	case TCPS_LISTEN:
2809 #ifdef TCP_OFFLOAD
2810 		tcp_offload_listen_stop(tp);
2811 #endif
2812 		tcp_state_change(tp, TCPS_CLOSED);
2813 		/* FALLTHROUGH */
2814 	case TCPS_CLOSED:
2815 		tp = tcp_close(tp);
2816 		/*
2817 		 * tcp_close() should never return NULL here as the socket is
2818 		 * still open.
2819 		 */
2820 		KASSERT(tp != NULL,
2821 		    ("tcp_usrclosed: tcp_close() returned NULL"));
2822 		break;
2823 
2824 	case TCPS_SYN_SENT:
2825 	case TCPS_SYN_RECEIVED:
2826 		tp->t_flags |= TF_NEEDFIN;
2827 		break;
2828 
2829 	case TCPS_ESTABLISHED:
2830 		tcp_state_change(tp, TCPS_FIN_WAIT_1);
2831 		break;
2832 
2833 	case TCPS_CLOSE_WAIT:
2834 		tcp_state_change(tp, TCPS_LAST_ACK);
2835 		break;
2836 	}
2837 	if (tp->t_state >= TCPS_FIN_WAIT_2) {
2838 		soisdisconnected(tp->t_inpcb->inp_socket);
2839 		/* Prevent the connection hanging in FIN_WAIT_2 forever. */
2840 		if (tp->t_state == TCPS_FIN_WAIT_2) {
2841 			int timeout;
2842 
2843 			timeout = (tcp_fast_finwait2_recycle) ?
2844 			    tcp_finwait2_timeout : TP_MAXIDLE(tp);
2845 			tcp_timer_activate(tp, TT_2MSL, timeout);
2846 		}
2847 	}
2848 }
2849 
2850 #ifdef DDB
2851 static void
2852 db_print_indent(int indent)
2853 {
2854 	int i;
2855 
2856 	for (i = 0; i < indent; i++)
2857 		db_printf(" ");
2858 }
2859 
2860 static void
2861 db_print_tstate(int t_state)
2862 {
2863 
2864 	switch (t_state) {
2865 	case TCPS_CLOSED:
2866 		db_printf("TCPS_CLOSED");
2867 		return;
2868 
2869 	case TCPS_LISTEN:
2870 		db_printf("TCPS_LISTEN");
2871 		return;
2872 
2873 	case TCPS_SYN_SENT:
2874 		db_printf("TCPS_SYN_SENT");
2875 		return;
2876 
2877 	case TCPS_SYN_RECEIVED:
2878 		db_printf("TCPS_SYN_RECEIVED");
2879 		return;
2880 
2881 	case TCPS_ESTABLISHED:
2882 		db_printf("TCPS_ESTABLISHED");
2883 		return;
2884 
2885 	case TCPS_CLOSE_WAIT:
2886 		db_printf("TCPS_CLOSE_WAIT");
2887 		return;
2888 
2889 	case TCPS_FIN_WAIT_1:
2890 		db_printf("TCPS_FIN_WAIT_1");
2891 		return;
2892 
2893 	case TCPS_CLOSING:
2894 		db_printf("TCPS_CLOSING");
2895 		return;
2896 
2897 	case TCPS_LAST_ACK:
2898 		db_printf("TCPS_LAST_ACK");
2899 		return;
2900 
2901 	case TCPS_FIN_WAIT_2:
2902 		db_printf("TCPS_FIN_WAIT_2");
2903 		return;
2904 
2905 	case TCPS_TIME_WAIT:
2906 		db_printf("TCPS_TIME_WAIT");
2907 		return;
2908 
2909 	default:
2910 		db_printf("unknown");
2911 		return;
2912 	}
2913 }
2914 
2915 static void
2916 db_print_tflags(u_int t_flags)
2917 {
2918 	int comma;
2919 
2920 	comma = 0;
2921 	if (t_flags & TF_ACKNOW) {
2922 		db_printf("%sTF_ACKNOW", comma ? ", " : "");
2923 		comma = 1;
2924 	}
2925 	if (t_flags & TF_DELACK) {
2926 		db_printf("%sTF_DELACK", comma ? ", " : "");
2927 		comma = 1;
2928 	}
2929 	if (t_flags & TF_NODELAY) {
2930 		db_printf("%sTF_NODELAY", comma ? ", " : "");
2931 		comma = 1;
2932 	}
2933 	if (t_flags & TF_NOOPT) {
2934 		db_printf("%sTF_NOOPT", comma ? ", " : "");
2935 		comma = 1;
2936 	}
2937 	if (t_flags & TF_SENTFIN) {
2938 		db_printf("%sTF_SENTFIN", comma ? ", " : "");
2939 		comma = 1;
2940 	}
2941 	if (t_flags & TF_REQ_SCALE) {
2942 		db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
2943 		comma = 1;
2944 	}
2945 	if (t_flags & TF_RCVD_SCALE) {
2946 		db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
2947 		comma = 1;
2948 	}
2949 	if (t_flags & TF_REQ_TSTMP) {
2950 		db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
2951 		comma = 1;
2952 	}
2953 	if (t_flags & TF_RCVD_TSTMP) {
2954 		db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
2955 		comma = 1;
2956 	}
2957 	if (t_flags & TF_SACK_PERMIT) {
2958 		db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
2959 		comma = 1;
2960 	}
2961 	if (t_flags & TF_NEEDSYN) {
2962 		db_printf("%sTF_NEEDSYN", comma ? ", " : "");
2963 		comma = 1;
2964 	}
2965 	if (t_flags & TF_NEEDFIN) {
2966 		db_printf("%sTF_NEEDFIN", comma ? ", " : "");
2967 		comma = 1;
2968 	}
2969 	if (t_flags & TF_NOPUSH) {
2970 		db_printf("%sTF_NOPUSH", comma ? ", " : "");
2971 		comma = 1;
2972 	}
2973 	if (t_flags & TF_PREVVALID) {
2974 		db_printf("%sTF_PREVVALID", comma ? ", " : "");
2975 		comma = 1;
2976 	}
2977 	if (t_flags & TF_MORETOCOME) {
2978 		db_printf("%sTF_MORETOCOME", comma ? ", " : "");
2979 		comma = 1;
2980 	}
2981 	if (t_flags & TF_LQ_OVERFLOW) {
2982 		db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
2983 		comma = 1;
2984 	}
2985 	if (t_flags & TF_LASTIDLE) {
2986 		db_printf("%sTF_LASTIDLE", comma ? ", " : "");
2987 		comma = 1;
2988 	}
2989 	if (t_flags & TF_RXWIN0SENT) {
2990 		db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
2991 		comma = 1;
2992 	}
2993 	if (t_flags & TF_FASTRECOVERY) {
2994 		db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
2995 		comma = 1;
2996 	}
2997 	if (t_flags & TF_CONGRECOVERY) {
2998 		db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
2999 		comma = 1;
3000 	}
3001 	if (t_flags & TF_WASFRECOVERY) {
3002 		db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
3003 		comma = 1;
3004 	}
3005 	if (t_flags & TF_WASCRECOVERY) {
3006 		db_printf("%sTF_WASCRECOVERY", comma ? ", " : "");
3007 		comma = 1;
3008 	}
3009 	if (t_flags & TF_SIGNATURE) {
3010 		db_printf("%sTF_SIGNATURE", comma ? ", " : "");
3011 		comma = 1;
3012 	}
3013 	if (t_flags & TF_FORCEDATA) {
3014 		db_printf("%sTF_FORCEDATA", comma ? ", " : "");
3015 		comma = 1;
3016 	}
3017 	if (t_flags & TF_TSO) {
3018 		db_printf("%sTF_TSO", comma ? ", " : "");
3019 		comma = 1;
3020 	}
3021 	if (t_flags & TF_FASTOPEN) {
3022 		db_printf("%sTF_FASTOPEN", comma ? ", " : "");
3023 		comma = 1;
3024 	}
3025 }
3026 
3027 static void
3028 db_print_tflags2(u_int t_flags2)
3029 {
3030 	int comma;
3031 
3032 	comma = 0;
3033 	if (t_flags2 & TF2_PLPMTU_BLACKHOLE) {
3034 		db_printf("%sTF2_PLPMTU_BLACKHOLE", comma ? ", " : "");
3035 		comma = 1;
3036 	}
3037 	if (t_flags2 & TF2_PLPMTU_PMTUD) {
3038 		db_printf("%sTF2_PLPMTU_PMTUD", comma ? ", " : "");
3039 		comma = 1;
3040 	}
3041 	if (t_flags2 & TF2_PLPMTU_MAXSEGSNT) {
3042 		db_printf("%sTF2_PLPMTU_MAXSEGSNT", comma ? ", " : "");
3043 		comma = 1;
3044 	}
3045 	if (t_flags2 & TF2_LOG_AUTO) {
3046 		db_printf("%sTF2_LOG_AUTO", comma ? ", " : "");
3047 		comma = 1;
3048 	}
3049 	if (t_flags2 & TF2_DROP_AF_DATA) {
3050 		db_printf("%sTF2_DROP_AF_DATA", comma ? ", " : "");
3051 		comma = 1;
3052 	}
3053 	if (t_flags2 & TF2_ECN_PERMIT) {
3054 		db_printf("%sTF2_ECN_PERMIT", comma ? ", " : "");
3055 		comma = 1;
3056 	}
3057 	if (t_flags2 & TF2_ECN_SND_CWR) {
3058 		db_printf("%sTF2_ECN_SND_CWR", comma ? ", " : "");
3059 		comma = 1;
3060 	}
3061 	if (t_flags2 & TF2_ECN_SND_ECE) {
3062 		db_printf("%sTF2_ECN_SND_ECE", comma ? ", " : "");
3063 		comma = 1;
3064 	}
3065 	if (t_flags2 & TF2_ACE_PERMIT) {
3066 		db_printf("%sTF2_ACE_PERMIT", comma ? ", " : "");
3067 		comma = 1;
3068 	}
3069 	if (t_flags2 & TF2_FBYTES_COMPLETE) {
3070 		db_printf("%sTF2_FBYTES_COMPLETE", comma ? ", " : "");
3071 		comma = 1;
3072 	}
3073 }
3074 
3075 static void
3076 db_print_toobflags(char t_oobflags)
3077 {
3078 	int comma;
3079 
3080 	comma = 0;
3081 	if (t_oobflags & TCPOOB_HAVEDATA) {
3082 		db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
3083 		comma = 1;
3084 	}
3085 	if (t_oobflags & TCPOOB_HADDATA) {
3086 		db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
3087 		comma = 1;
3088 	}
3089 }
3090 
3091 static void
3092 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
3093 {
3094 
3095 	db_print_indent(indent);
3096 	db_printf("%s at %p\n", name, tp);
3097 
3098 	indent += 2;
3099 
3100 	db_print_indent(indent);
3101 	db_printf("t_segq first: %p   t_segqlen: %d   t_dupacks: %d\n",
3102 	   TAILQ_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
3103 
3104 	db_print_indent(indent);
3105 	db_printf("tt_rexmt: %p   tt_persist: %p   tt_keep: %p\n",
3106 	    &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
3107 
3108 	db_print_indent(indent);
3109 	db_printf("tt_2msl: %p   tt_delack: %p   t_inpcb: %p\n", &tp->t_timers->tt_2msl,
3110 	    &tp->t_timers->tt_delack, tp->t_inpcb);
3111 
3112 	db_print_indent(indent);
3113 	db_printf("t_state: %d (", tp->t_state);
3114 	db_print_tstate(tp->t_state);
3115 	db_printf(")\n");
3116 
3117 	db_print_indent(indent);
3118 	db_printf("t_flags: 0x%x (", tp->t_flags);
3119 	db_print_tflags(tp->t_flags);
3120 	db_printf(")\n");
3121 
3122 	db_print_indent(indent);
3123 	db_printf("t_flags2: 0x%x (", tp->t_flags2);
3124 	db_print_tflags2(tp->t_flags2);
3125 	db_printf(")\n");
3126 
3127 	db_print_indent(indent);
3128 	db_printf("snd_una: 0x%08x   snd_max: 0x%08x   snd_nxt: x0%08x\n",
3129 	    tp->snd_una, tp->snd_max, tp->snd_nxt);
3130 
3131 	db_print_indent(indent);
3132 	db_printf("snd_up: 0x%08x   snd_wl1: 0x%08x   snd_wl2: 0x%08x\n",
3133 	   tp->snd_up, tp->snd_wl1, tp->snd_wl2);
3134 
3135 	db_print_indent(indent);
3136 	db_printf("iss: 0x%08x   irs: 0x%08x   rcv_nxt: 0x%08x\n",
3137 	    tp->iss, tp->irs, tp->rcv_nxt);
3138 
3139 	db_print_indent(indent);
3140 	db_printf("rcv_adv: 0x%08x   rcv_wnd: %u   rcv_up: 0x%08x\n",
3141 	    tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
3142 
3143 	db_print_indent(indent);
3144 	db_printf("snd_wnd: %u   snd_cwnd: %u\n",
3145 	   tp->snd_wnd, tp->snd_cwnd);
3146 
3147 	db_print_indent(indent);
3148 	db_printf("snd_ssthresh: %u   snd_recover: "
3149 	    "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
3150 
3151 	db_print_indent(indent);
3152 	db_printf("t_rcvtime: %u   t_startime: %u\n",
3153 	    tp->t_rcvtime, tp->t_starttime);
3154 
3155 	db_print_indent(indent);
3156 	db_printf("t_rttime: %u   t_rtsq: 0x%08x\n",
3157 	    tp->t_rtttime, tp->t_rtseq);
3158 
3159 	db_print_indent(indent);
3160 	db_printf("t_rxtcur: %d   t_maxseg: %u   t_srtt: %d\n",
3161 	    tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
3162 
3163 	db_print_indent(indent);
3164 	db_printf("t_rttvar: %d   t_rxtshift: %d   t_rttmin: %u   "
3165 	    "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
3166 	    tp->t_rttbest);
3167 
3168 	db_print_indent(indent);
3169 	db_printf("t_rttupdated: %lu   max_sndwnd: %u   t_softerror: %d\n",
3170 	    tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
3171 
3172 	db_print_indent(indent);
3173 	db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
3174 	db_print_toobflags(tp->t_oobflags);
3175 	db_printf(")   t_iobc: 0x%02x\n", tp->t_iobc);
3176 
3177 	db_print_indent(indent);
3178 	db_printf("snd_scale: %u   rcv_scale: %u   request_r_scale: %u\n",
3179 	    tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
3180 
3181 	db_print_indent(indent);
3182 	db_printf("ts_recent: %u   ts_recent_age: %u\n",
3183 	    tp->ts_recent, tp->ts_recent_age);
3184 
3185 	db_print_indent(indent);
3186 	db_printf("ts_offset: %u   last_ack_sent: 0x%08x   snd_cwnd_prev: "
3187 	    "%u\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
3188 
3189 	db_print_indent(indent);
3190 	db_printf("snd_ssthresh_prev: %u   snd_recover_prev: 0x%08x   "
3191 	    "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
3192 	    tp->snd_recover_prev, tp->t_badrxtwin);
3193 
3194 	db_print_indent(indent);
3195 	db_printf("snd_numholes: %d  snd_holes first: %p\n",
3196 	    tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
3197 
3198 	db_print_indent(indent);
3199 	db_printf("snd_fack: 0x%08x   rcv_numsacks: %d\n",
3200 	    tp->snd_fack, tp->rcv_numsacks);
3201 
3202 	/* Skip sackblks, sackhint. */
3203 
3204 	db_print_indent(indent);
3205 	db_printf("t_rttlow: %d   rfbuf_ts: %u   rfbuf_cnt: %d\n",
3206 	    tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
3207 }
3208 
3209 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
3210 {
3211 	struct tcpcb *tp;
3212 
3213 	if (!have_addr) {
3214 		db_printf("usage: show tcpcb <addr>\n");
3215 		return;
3216 	}
3217 	tp = (struct tcpcb *)addr;
3218 
3219 	db_print_tcpcb(tp, "tcpcb", 0);
3220 }
3221 #endif
3222