xref: /freebsd/sys/netinet/tcp_usrreq.c (revision 08aba0aec7b7f676ccc3f7886f59f277d668d5b4)
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 	if ((error = tcp6_connect(tp, nam, td)) != 0)
697 		goto out;
698 #ifdef TCP_OFFLOAD
699 	if (registered_toedevs > 0 &&
700 	    (so->so_options & SO_NO_OFFLOAD) == 0 &&
701 	    (error = tcp_offload_connect(so, nam)) == 0)
702 		goto out;
703 #endif
704 	tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
705 	NET_EPOCH_ENTER(et);
706 	error = tcp_output(tp);
707 #ifdef INET
708 out_in_epoch:
709 #endif
710 	NET_EPOCH_EXIT(et);
711 out:
712 	KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()"
713 	    ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error));
714 	/*
715 	 * If the implicit bind in the connect call fails, restore
716 	 * the flags we modified.
717 	 */
718 	if (error != 0 && inp->inp_lport == 0) {
719 		inp->inp_vflag = vflagsav;
720 		inp->inp_inc.inc_flags = incflagsav;
721 	}
722 
723 	TCPDEBUG2(PRU_CONNECT);
724 	TCP_PROBE2(debug__user, tp, PRU_CONNECT);
725 	INP_WUNLOCK(inp);
726 	return (error);
727 }
728 #endif /* INET6 */
729 
730 /*
731  * Initiate disconnect from peer.
732  * If connection never passed embryonic stage, just drop;
733  * else if don't need to let data drain, then can just drop anyways,
734  * else have to begin TCP shutdown process: mark socket disconnecting,
735  * drain unread data, state switch to reflect user close, and
736  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
737  * when peer sends FIN and acks ours.
738  *
739  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
740  */
741 static int
742 tcp_usr_disconnect(struct socket *so)
743 {
744 	struct inpcb *inp;
745 	struct tcpcb *tp = NULL;
746 	struct epoch_tracker et;
747 	int error = 0;
748 
749 	TCPDEBUG0;
750 	NET_EPOCH_ENTER(et);
751 	inp = sotoinpcb(so);
752 	KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
753 	INP_WLOCK(inp);
754 	if (inp->inp_flags & INP_TIMEWAIT)
755 		goto out;
756 	if (inp->inp_flags & INP_DROPPED) {
757 		error = ECONNRESET;
758 		goto out;
759 	}
760 	tp = intotcpcb(inp);
761 	TCPDEBUG1();
762 	tcp_disconnect(tp);
763 out:
764 	TCPDEBUG2(PRU_DISCONNECT);
765 	TCP_PROBE2(debug__user, tp, PRU_DISCONNECT);
766 	INP_WUNLOCK(inp);
767 	NET_EPOCH_EXIT(et);
768 	return (error);
769 }
770 
771 #ifdef INET
772 /*
773  * Accept a connection.  Essentially all the work is done at higher levels;
774  * just return the address of the peer, storing through addr.
775  */
776 static int
777 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
778 {
779 	int error = 0;
780 	struct inpcb *inp = NULL;
781 #ifdef KDTRACE_HOOKS
782 	struct tcpcb *tp = NULL;
783 #endif
784 	struct in_addr addr;
785 	in_port_t port = 0;
786 	TCPDEBUG0;
787 
788 	if (so->so_state & SS_ISDISCONNECTED)
789 		return (ECONNABORTED);
790 
791 	inp = sotoinpcb(so);
792 	KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
793 	INP_WLOCK(inp);
794 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
795 		error = ECONNABORTED;
796 		goto out;
797 	}
798 #ifdef KDTRACE_HOOKS
799 	tp = intotcpcb(inp);
800 #endif
801 	TCPDEBUG1();
802 
803 	/*
804 	 * We inline in_getpeeraddr and COMMON_END here, so that we can
805 	 * copy the data of interest and defer the malloc until after we
806 	 * release the lock.
807 	 */
808 	port = inp->inp_fport;
809 	addr = inp->inp_faddr;
810 
811 out:
812 	TCPDEBUG2(PRU_ACCEPT);
813 	TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
814 	INP_WUNLOCK(inp);
815 	if (error == 0)
816 		*nam = in_sockaddr(port, &addr);
817 	return error;
818 }
819 #endif /* INET */
820 
821 #ifdef INET6
822 static int
823 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
824 {
825 	struct inpcb *inp = NULL;
826 	int error = 0;
827 #ifdef KDTRACE_HOOKS
828 	struct tcpcb *tp = NULL;
829 #endif
830 	struct in_addr addr;
831 	struct in6_addr addr6;
832 	struct epoch_tracker et;
833 	in_port_t port = 0;
834 	int v4 = 0;
835 	TCPDEBUG0;
836 
837 	if (so->so_state & SS_ISDISCONNECTED)
838 		return (ECONNABORTED);
839 
840 	inp = sotoinpcb(so);
841 	KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
842 	NET_EPOCH_ENTER(et);
843 	INP_WLOCK(inp);
844 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
845 		error = ECONNABORTED;
846 		goto out;
847 	}
848 #ifdef KDTRACE_HOOKS
849 	tp = intotcpcb(inp);
850 #endif
851 	TCPDEBUG1();
852 
853 	/*
854 	 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
855 	 * copy the data of interest and defer the malloc until after we
856 	 * release the lock.
857 	 */
858 	if (inp->inp_vflag & INP_IPV4) {
859 		v4 = 1;
860 		port = inp->inp_fport;
861 		addr = inp->inp_faddr;
862 	} else {
863 		port = inp->inp_fport;
864 		addr6 = inp->in6p_faddr;
865 	}
866 
867 out:
868 	TCPDEBUG2(PRU_ACCEPT);
869 	TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
870 	INP_WUNLOCK(inp);
871 	NET_EPOCH_EXIT(et);
872 	if (error == 0) {
873 		if (v4)
874 			*nam = in6_v4mapsin6_sockaddr(port, &addr);
875 		else
876 			*nam = in6_sockaddr(port, &addr6);
877 	}
878 	return error;
879 }
880 #endif /* INET6 */
881 
882 /*
883  * Mark the connection as being incapable of further output.
884  */
885 static int
886 tcp_usr_shutdown(struct socket *so)
887 {
888 	int error = 0;
889 	struct inpcb *inp;
890 	struct tcpcb *tp = NULL;
891 	struct epoch_tracker et;
892 
893 	TCPDEBUG0;
894 	inp = sotoinpcb(so);
895 	KASSERT(inp != NULL, ("inp == NULL"));
896 	INP_WLOCK(inp);
897 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
898 		INP_WUNLOCK(inp);
899 		return (ECONNRESET);
900 	}
901 	tp = intotcpcb(inp);
902 	NET_EPOCH_ENTER(et);
903 	TCPDEBUG1();
904 	socantsendmore(so);
905 	tcp_usrclosed(tp);
906 	if (!(inp->inp_flags & INP_DROPPED))
907 		error = tcp_output_nodrop(tp);
908 	TCPDEBUG2(PRU_SHUTDOWN);
909 	TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN);
910 	error = tcp_unlock_or_drop(tp, error);
911 	NET_EPOCH_EXIT(et);
912 
913 	return (error);
914 }
915 
916 /*
917  * After a receive, possibly send window update to peer.
918  */
919 static int
920 tcp_usr_rcvd(struct socket *so, int flags)
921 {
922 	struct epoch_tracker et;
923 	struct inpcb *inp;
924 	struct tcpcb *tp = NULL;
925 	int outrv = 0, error = 0;
926 
927 	TCPDEBUG0;
928 	inp = sotoinpcb(so);
929 	KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
930 	INP_WLOCK(inp);
931 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
932 		INP_WUNLOCK(inp);
933 		return (ECONNRESET);
934 	}
935 	tp = intotcpcb(inp);
936 	NET_EPOCH_ENTER(et);
937 	TCPDEBUG1();
938 	/*
939 	 * For passively-created TFO connections, don't attempt a window
940 	 * update while still in SYN_RECEIVED as this may trigger an early
941 	 * SYN|ACK.  It is preferable to have the SYN|ACK be sent along with
942 	 * application response data, or failing that, when the DELACK timer
943 	 * expires.
944 	 */
945 	if (IS_FASTOPEN(tp->t_flags) &&
946 	    (tp->t_state == TCPS_SYN_RECEIVED))
947 		goto out;
948 #ifdef TCP_OFFLOAD
949 	if (tp->t_flags & TF_TOE)
950 		tcp_offload_rcvd(tp);
951 	else
952 #endif
953 		outrv = tcp_output_nodrop(tp);
954 out:
955 	TCPDEBUG2(PRU_RCVD);
956 	TCP_PROBE2(debug__user, tp, PRU_RCVD);
957 	(void) tcp_unlock_or_drop(tp, outrv);
958 	NET_EPOCH_EXIT(et);
959 	return (error);
960 }
961 
962 /*
963  * Do a send by putting data in output queue and updating urgent
964  * marker if URG set.  Possibly send more data.  Unlike the other
965  * pru_*() routines, the mbuf chains are our responsibility.  We
966  * must either enqueue them or free them.  The other pru_* routines
967  * generally are caller-frees.
968  */
969 static int
970 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
971     struct sockaddr *nam, struct mbuf *control, struct thread *td)
972 {
973 	struct epoch_tracker et;
974 	int error = 0;
975 	struct inpcb *inp;
976 	struct tcpcb *tp = NULL;
977 #ifdef INET
978 #ifdef INET6
979 	struct sockaddr_in sin;
980 #endif
981 	struct sockaddr_in *sinp;
982 #endif
983 #ifdef INET6
984 	int isipv6;
985 #endif
986 	u_int8_t incflagsav;
987 	u_char vflagsav;
988 	bool restoreflags;
989 	TCPDEBUG0;
990 
991 	if (control != NULL) {
992 		/* TCP doesn't do control messages (rights, creds, etc) */
993 		if (control->m_len) {
994 			m_freem(control);
995 			return (EINVAL);
996 		}
997 		m_freem(control);	/* empty control, just free it */
998 	}
999 
1000 	inp = sotoinpcb(so);
1001 	KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
1002 	INP_WLOCK(inp);
1003 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1004 		if (m != NULL && (flags & PRUS_NOTREADY) == 0)
1005 			m_freem(m);
1006 		INP_WUNLOCK(inp);
1007 		return (ECONNRESET);
1008 	}
1009 
1010 	vflagsav = inp->inp_vflag;
1011 	incflagsav = inp->inp_inc.inc_flags;
1012 	restoreflags = false;
1013 	tp = intotcpcb(inp);
1014 
1015 	NET_EPOCH_ENTER(et);
1016 	if ((flags & PRUS_OOB) != 0 &&
1017 	    (error = tcp_pru_options_support(tp, PRUS_OOB)) != 0)
1018 		goto out;
1019 
1020 	TCPDEBUG1();
1021 	if (nam != NULL && tp->t_state < TCPS_SYN_SENT) {
1022 		if (tp->t_state == TCPS_LISTEN) {
1023 			error = EINVAL;
1024 			goto out;
1025 		}
1026 		switch (nam->sa_family) {
1027 #ifdef INET
1028 		case AF_INET:
1029 			sinp = (struct sockaddr_in *)nam;
1030 			if (sinp->sin_len != sizeof(struct sockaddr_in)) {
1031 				error = EINVAL;
1032 				goto out;
1033 			}
1034 			if ((inp->inp_vflag & INP_IPV6) != 0) {
1035 				error = EAFNOSUPPORT;
1036 				goto out;
1037 			}
1038 			if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
1039 				error = EAFNOSUPPORT;
1040 				goto out;
1041 			}
1042 			if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) {
1043 				error = EACCES;
1044 				goto out;
1045 			}
1046 			if ((error = prison_remote_ip4(td->td_ucred,
1047 			    &sinp->sin_addr)))
1048 				goto out;
1049 #ifdef INET6
1050 			isipv6 = 0;
1051 #endif
1052 			break;
1053 #endif /* INET */
1054 #ifdef INET6
1055 		case AF_INET6:
1056 		{
1057 			struct sockaddr_in6 *sin6;
1058 
1059 			sin6 = (struct sockaddr_in6 *)nam;
1060 			if (sin6->sin6_len != sizeof(*sin6)) {
1061 				error = EINVAL;
1062 				goto out;
1063 			}
1064 			if ((inp->inp_vflag & INP_IPV6PROTO) == 0) {
1065 				error = EAFNOSUPPORT;
1066 				goto out;
1067 			}
1068 			if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
1069 				error = EAFNOSUPPORT;
1070 				goto out;
1071 			}
1072 			if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1073 #ifdef INET
1074 				if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1075 					error = EINVAL;
1076 					goto out;
1077 				}
1078 				if ((inp->inp_vflag & INP_IPV4) == 0) {
1079 					error = EAFNOSUPPORT;
1080 					goto out;
1081 				}
1082 				restoreflags = true;
1083 				inp->inp_vflag &= ~INP_IPV6;
1084 				sinp = &sin;
1085 				in6_sin6_2_sin(sinp, sin6);
1086 				if (IN_MULTICAST(
1087 				    ntohl(sinp->sin_addr.s_addr))) {
1088 					error = EAFNOSUPPORT;
1089 					goto out;
1090 				}
1091 				if ((error = prison_remote_ip4(td->td_ucred,
1092 				    &sinp->sin_addr)))
1093 					goto out;
1094 				isipv6 = 0;
1095 #else /* !INET */
1096 				error = EAFNOSUPPORT;
1097 				goto out;
1098 #endif /* INET */
1099 			} else {
1100 				if ((inp->inp_vflag & INP_IPV6) == 0) {
1101 					error = EAFNOSUPPORT;
1102 					goto out;
1103 				}
1104 				restoreflags = true;
1105 				inp->inp_vflag &= ~INP_IPV4;
1106 				inp->inp_inc.inc_flags |= INC_ISIPV6;
1107 				if ((error = prison_remote_ip6(td->td_ucred,
1108 				    &sin6->sin6_addr)))
1109 					goto out;
1110 				isipv6 = 1;
1111 			}
1112 			break;
1113 		}
1114 #endif /* INET6 */
1115 		default:
1116 			error = EAFNOSUPPORT;
1117 			goto out;
1118 		}
1119 	}
1120 	if (!(flags & PRUS_OOB)) {
1121 		sbappendstream(&so->so_snd, m, flags);
1122 		m = NULL;
1123 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1124 			KASSERT(tp->t_state == TCPS_CLOSED,
1125 			    ("%s: tp %p is listening", __func__, tp));
1126 
1127 			/*
1128 			 * Do implied connect if not yet connected,
1129 			 * initialize window to default value, and
1130 			 * initialize maxseg using peer's cached MSS.
1131 			 */
1132 #ifdef INET6
1133 			if (isipv6)
1134 				error = tcp6_connect(tp, nam, td);
1135 #endif /* INET6 */
1136 #if defined(INET6) && defined(INET)
1137 			else
1138 #endif
1139 #ifdef INET
1140 				error = tcp_connect(tp,
1141 				    (struct sockaddr *)sinp, td);
1142 #endif
1143 			/*
1144 			 * The bind operation in tcp_connect succeeded. We
1145 			 * no longer want to restore the flags if later
1146 			 * operations fail.
1147 			 */
1148 			if (error == 0 || inp->inp_lport != 0)
1149 				restoreflags = false;
1150 
1151 			if (error) {
1152 				/* m is freed if PRUS_NOTREADY is unset. */
1153 				sbflush(&so->so_snd);
1154 				goto out;
1155 			}
1156 			if (IS_FASTOPEN(tp->t_flags))
1157 				tcp_fastopen_connect(tp);
1158 			else {
1159 				tp->snd_wnd = TTCP_CLIENT_SND_WND;
1160 				tcp_mss(tp, -1);
1161 			}
1162 		}
1163 		if (flags & PRUS_EOF) {
1164 			/*
1165 			 * Close the send side of the connection after
1166 			 * the data is sent.
1167 			 */
1168 			socantsendmore(so);
1169 			tcp_usrclosed(tp);
1170 		}
1171 		if (TCPS_HAVEESTABLISHED(tp->t_state) &&
1172 		    ((tp->t_flags2 & TF2_FBYTES_COMPLETE) == 0) &&
1173 		    (tp->t_fbyte_out == 0) &&
1174 		    (so->so_snd.sb_ccc > 0)) {
1175 			tp->t_fbyte_out = ticks;
1176 			if (tp->t_fbyte_out == 0)
1177 				tp->t_fbyte_out = 1;
1178 			if (tp->t_fbyte_out && tp->t_fbyte_in)
1179 				tp->t_flags2 |= TF2_FBYTES_COMPLETE;
1180 		}
1181 		if (!(inp->inp_flags & INP_DROPPED) &&
1182 		    !(flags & PRUS_NOTREADY)) {
1183 			if (flags & PRUS_MORETOCOME)
1184 				tp->t_flags |= TF_MORETOCOME;
1185 			error = tcp_output_nodrop(tp);
1186 			if (flags & PRUS_MORETOCOME)
1187 				tp->t_flags &= ~TF_MORETOCOME;
1188 		}
1189 	} else {
1190 		/*
1191 		 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
1192 		 */
1193 		SOCKBUF_LOCK(&so->so_snd);
1194 		if (sbspace(&so->so_snd) < -512) {
1195 			SOCKBUF_UNLOCK(&so->so_snd);
1196 			error = ENOBUFS;
1197 			goto out;
1198 		}
1199 		/*
1200 		 * According to RFC961 (Assigned Protocols),
1201 		 * the urgent pointer points to the last octet
1202 		 * of urgent data.  We continue, however,
1203 		 * to consider it to indicate the first octet
1204 		 * of data past the urgent section.
1205 		 * Otherwise, snd_up should be one lower.
1206 		 */
1207 		sbappendstream_locked(&so->so_snd, m, flags);
1208 		SOCKBUF_UNLOCK(&so->so_snd);
1209 		m = NULL;
1210 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1211 			/*
1212 			 * Do implied connect if not yet connected,
1213 			 * initialize window to default value, and
1214 			 * initialize maxseg using peer's cached MSS.
1215 			 */
1216 
1217 			/*
1218 			 * Not going to contemplate SYN|URG
1219 			 */
1220 			if (IS_FASTOPEN(tp->t_flags))
1221 				tp->t_flags &= ~TF_FASTOPEN;
1222 #ifdef INET6
1223 			if (isipv6)
1224 				error = tcp6_connect(tp, nam, td);
1225 #endif /* INET6 */
1226 #if defined(INET6) && defined(INET)
1227 			else
1228 #endif
1229 #ifdef INET
1230 				error = tcp_connect(tp,
1231 				    (struct sockaddr *)sinp, td);
1232 #endif
1233 			/*
1234 			 * The bind operation in tcp_connect succeeded. We
1235 			 * no longer want to restore the flags if later
1236 			 * operations fail.
1237 			 */
1238 			if (error == 0 || inp->inp_lport != 0)
1239 				restoreflags = false;
1240 
1241 			if (error != 0) {
1242 				/* m is freed if PRUS_NOTREADY is unset. */
1243 				sbflush(&so->so_snd);
1244 				goto out;
1245 			}
1246 			tp->snd_wnd = TTCP_CLIENT_SND_WND;
1247 			tcp_mss(tp, -1);
1248 		}
1249 		tp->snd_up = tp->snd_una + sbavail(&so->so_snd);
1250 		if ((flags & PRUS_NOTREADY) == 0) {
1251 			tp->t_flags |= TF_FORCEDATA;
1252 			error = tcp_output_nodrop(tp);
1253 			tp->t_flags &= ~TF_FORCEDATA;
1254 		}
1255 	}
1256 	TCP_LOG_EVENT(tp, NULL,
1257 	    &inp->inp_socket->so_rcv,
1258 	    &inp->inp_socket->so_snd,
1259 	    TCP_LOG_USERSEND, error,
1260 	    0, NULL, false);
1261 
1262 out:
1263 	/*
1264 	 * In case of PRUS_NOTREADY, the caller or tcp_usr_ready() is
1265 	 * responsible for freeing memory.
1266 	 */
1267 	if (m != NULL && (flags & PRUS_NOTREADY) == 0)
1268 		m_freem(m);
1269 
1270 	/*
1271 	 * If the request was unsuccessful and we changed flags,
1272 	 * restore the original flags.
1273 	 */
1274 	if (error != 0 && restoreflags) {
1275 		inp->inp_vflag = vflagsav;
1276 		inp->inp_inc.inc_flags = incflagsav;
1277 	}
1278 	TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
1279 		  ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1280 	TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB :
1281 		   ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1282 	error = tcp_unlock_or_drop(tp, error);
1283 	NET_EPOCH_EXIT(et);
1284 	return (error);
1285 }
1286 
1287 static int
1288 tcp_usr_ready(struct socket *so, struct mbuf *m, int count)
1289 {
1290 	struct epoch_tracker et;
1291 	struct inpcb *inp;
1292 	struct tcpcb *tp;
1293 	int error;
1294 
1295 	inp = sotoinpcb(so);
1296 	INP_WLOCK(inp);
1297 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1298 		INP_WUNLOCK(inp);
1299 		mb_free_notready(m, count);
1300 		return (ECONNRESET);
1301 	}
1302 	tp = intotcpcb(inp);
1303 
1304 	SOCKBUF_LOCK(&so->so_snd);
1305 	error = sbready(&so->so_snd, m, count);
1306 	SOCKBUF_UNLOCK(&so->so_snd);
1307 	if (error) {
1308 		INP_WUNLOCK(inp);
1309 		return (error);
1310 	}
1311 	NET_EPOCH_ENTER(et);
1312 	error = tcp_output_unlock(tp);
1313 	NET_EPOCH_EXIT(et);
1314 
1315 	return (error);
1316 }
1317 
1318 /*
1319  * Abort the TCP.  Drop the connection abruptly.
1320  */
1321 static void
1322 tcp_usr_abort(struct socket *so)
1323 {
1324 	struct inpcb *inp;
1325 	struct tcpcb *tp = NULL;
1326 	struct epoch_tracker et;
1327 	TCPDEBUG0;
1328 
1329 	inp = sotoinpcb(so);
1330 	KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
1331 
1332 	NET_EPOCH_ENTER(et);
1333 	INP_WLOCK(inp);
1334 	KASSERT(inp->inp_socket != NULL,
1335 	    ("tcp_usr_abort: inp_socket == NULL"));
1336 
1337 	/*
1338 	 * If we still have full TCP state, and we're not dropped, drop.
1339 	 */
1340 	if (!(inp->inp_flags & INP_TIMEWAIT) &&
1341 	    !(inp->inp_flags & INP_DROPPED)) {
1342 		tp = intotcpcb(inp);
1343 		TCPDEBUG1();
1344 		tp = tcp_drop(tp, ECONNABORTED);
1345 		if (tp == NULL)
1346 			goto dropped;
1347 		TCPDEBUG2(PRU_ABORT);
1348 		TCP_PROBE2(debug__user, tp, PRU_ABORT);
1349 	}
1350 	if (!(inp->inp_flags & INP_DROPPED)) {
1351 		soref(so);
1352 		inp->inp_flags |= INP_SOCKREF;
1353 	}
1354 	INP_WUNLOCK(inp);
1355 dropped:
1356 	NET_EPOCH_EXIT(et);
1357 }
1358 
1359 /*
1360  * TCP socket is closed.  Start friendly disconnect.
1361  */
1362 static void
1363 tcp_usr_close(struct socket *so)
1364 {
1365 	struct inpcb *inp;
1366 	struct tcpcb *tp = NULL;
1367 	struct epoch_tracker et;
1368 	TCPDEBUG0;
1369 
1370 	inp = sotoinpcb(so);
1371 	KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
1372 
1373 	NET_EPOCH_ENTER(et);
1374 	INP_WLOCK(inp);
1375 	KASSERT(inp->inp_socket != NULL,
1376 	    ("tcp_usr_close: inp_socket == NULL"));
1377 
1378 	/*
1379 	 * If we still have full TCP state, and we're not dropped, initiate
1380 	 * a disconnect.
1381 	 */
1382 	if (!(inp->inp_flags & INP_TIMEWAIT) &&
1383 	    !(inp->inp_flags & INP_DROPPED)) {
1384 		tp = intotcpcb(inp);
1385 		tp->t_flags |= TF_CLOSED;
1386 		TCPDEBUG1();
1387 		tcp_disconnect(tp);
1388 		TCPDEBUG2(PRU_CLOSE);
1389 		TCP_PROBE2(debug__user, tp, PRU_CLOSE);
1390 	}
1391 	if (!(inp->inp_flags & INP_DROPPED)) {
1392 		soref(so);
1393 		inp->inp_flags |= INP_SOCKREF;
1394 	}
1395 	INP_WUNLOCK(inp);
1396 	NET_EPOCH_EXIT(et);
1397 }
1398 
1399 static int
1400 tcp_pru_options_support(struct tcpcb *tp, int flags)
1401 {
1402 	/*
1403 	 * If the specific TCP stack has a pru_options
1404 	 * specified then it does not always support
1405 	 * all the PRU_XX options and we must ask it.
1406 	 * If the function is not specified then all
1407 	 * of the PRU_XX options are supported.
1408 	 */
1409 	int ret = 0;
1410 
1411 	if (tp->t_fb->tfb_pru_options) {
1412 		ret = (*tp->t_fb->tfb_pru_options)(tp, flags);
1413 	}
1414 	return (ret);
1415 }
1416 
1417 /*
1418  * Receive out-of-band data.
1419  */
1420 static int
1421 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1422 {
1423 	int error = 0;
1424 	struct inpcb *inp;
1425 	struct tcpcb *tp = NULL;
1426 
1427 	TCPDEBUG0;
1428 	inp = sotoinpcb(so);
1429 	KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1430 	INP_WLOCK(inp);
1431 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1432 		error = ECONNRESET;
1433 		goto out;
1434 	}
1435 	tp = intotcpcb(inp);
1436 	error = tcp_pru_options_support(tp, PRUS_OOB);
1437 	if (error) {
1438 		goto out;
1439 	}
1440 	TCPDEBUG1();
1441 	if ((so->so_oobmark == 0 &&
1442 	     (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1443 	    so->so_options & SO_OOBINLINE ||
1444 	    tp->t_oobflags & TCPOOB_HADDATA) {
1445 		error = EINVAL;
1446 		goto out;
1447 	}
1448 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1449 		error = EWOULDBLOCK;
1450 		goto out;
1451 	}
1452 	m->m_len = 1;
1453 	*mtod(m, caddr_t) = tp->t_iobc;
1454 	if ((flags & MSG_PEEK) == 0)
1455 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1456 
1457 out:
1458 	TCPDEBUG2(PRU_RCVOOB);
1459 	TCP_PROBE2(debug__user, tp, PRU_RCVOOB);
1460 	INP_WUNLOCK(inp);
1461 	return (error);
1462 }
1463 
1464 #ifdef INET
1465 struct protosw tcp_protosw = {
1466 	.pr_type =		SOCK_STREAM,
1467 	.pr_protocol =		IPPROTO_TCP,
1468 	.pr_flags =		PR_CONNREQUIRED | PR_IMPLOPCL | PR_WANTRCVD |
1469 				    PR_CAPATTACH,
1470 	.pr_ctloutput =		tcp_ctloutput,
1471 	.pr_abort =		tcp_usr_abort,
1472 	.pr_accept =		tcp_usr_accept,
1473 	.pr_attach =		tcp_usr_attach,
1474 	.pr_bind =		tcp_usr_bind,
1475 	.pr_connect =		tcp_usr_connect,
1476 	.pr_control =		in_control,
1477 	.pr_detach =		tcp_usr_detach,
1478 	.pr_disconnect =	tcp_usr_disconnect,
1479 	.pr_listen =		tcp_usr_listen,
1480 	.pr_peeraddr =		in_getpeeraddr,
1481 	.pr_rcvd =		tcp_usr_rcvd,
1482 	.pr_rcvoob =		tcp_usr_rcvoob,
1483 	.pr_send =		tcp_usr_send,
1484 	.pr_ready =		tcp_usr_ready,
1485 	.pr_shutdown =		tcp_usr_shutdown,
1486 	.pr_sockaddr =		in_getsockaddr,
1487 	.pr_sosetlabel =	in_pcbsosetlabel,
1488 	.pr_close =		tcp_usr_close,
1489 };
1490 #endif /* INET */
1491 
1492 #ifdef INET6
1493 struct protosw tcp6_protosw = {
1494 	.pr_type =		SOCK_STREAM,
1495 	.pr_protocol =		IPPROTO_TCP,
1496 	.pr_flags =		PR_CONNREQUIRED | PR_IMPLOPCL |PR_WANTRCVD |
1497 				    PR_CAPATTACH,
1498 	.pr_ctloutput =		tcp_ctloutput,
1499 	.pr_abort =		tcp_usr_abort,
1500 	.pr_accept =		tcp6_usr_accept,
1501 	.pr_attach =		tcp_usr_attach,
1502 	.pr_bind =		tcp6_usr_bind,
1503 	.pr_connect =		tcp6_usr_connect,
1504 	.pr_control =		in6_control,
1505 	.pr_detach =		tcp_usr_detach,
1506 	.pr_disconnect =	tcp_usr_disconnect,
1507 	.pr_listen =		tcp6_usr_listen,
1508 	.pr_peeraddr =		in6_mapped_peeraddr,
1509 	.pr_rcvd =		tcp_usr_rcvd,
1510 	.pr_rcvoob =		tcp_usr_rcvoob,
1511 	.pr_send =		tcp_usr_send,
1512 	.pr_ready =		tcp_usr_ready,
1513 	.pr_shutdown =		tcp_usr_shutdown,
1514 	.pr_sockaddr =		in6_mapped_sockaddr,
1515 	.pr_sosetlabel =	in_pcbsosetlabel,
1516 	.pr_close =		tcp_usr_close,
1517 };
1518 #endif /* INET6 */
1519 
1520 #ifdef INET
1521 /*
1522  * Common subroutine to open a TCP connection to remote host specified
1523  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
1524  * port number if needed.  Call in_pcbconnect_setup to do the routing and
1525  * to choose a local host address (interface).  If there is an existing
1526  * incarnation of the same connection in TIME-WAIT state and if the remote
1527  * host was sending CC options and if the connection duration was < MSL, then
1528  * truncate the previous TIME-WAIT state and proceed.
1529  * Initialize connection parameters and enter SYN-SENT state.
1530  */
1531 static int
1532 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1533 {
1534 	struct inpcb *inp = tp->t_inpcb, *oinp;
1535 	struct socket *so = inp->inp_socket;
1536 	struct in_addr laddr;
1537 	u_short lport;
1538 	int error;
1539 
1540 	NET_EPOCH_ASSERT();
1541 	INP_WLOCK_ASSERT(inp);
1542 	INP_HASH_WLOCK(&V_tcbinfo);
1543 
1544 	if (V_tcp_require_unique_port && inp->inp_lport == 0) {
1545 		error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1546 		if (error)
1547 			goto out;
1548 	}
1549 
1550 	/*
1551 	 * Cannot simply call in_pcbconnect, because there might be an
1552 	 * earlier incarnation of this same connection still in
1553 	 * TIME_WAIT state, creating an ADDRINUSE error.
1554 	 */
1555 	laddr = inp->inp_laddr;
1556 	lport = inp->inp_lport;
1557 	error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
1558 	    &inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
1559 	if (error && oinp == NULL)
1560 		goto out;
1561 	if (oinp) {
1562 		error = EADDRINUSE;
1563 		goto out;
1564 	}
1565 	/* Handle initial bind if it hadn't been done in advance. */
1566 	if (inp->inp_lport == 0) {
1567 		inp->inp_lport = lport;
1568 		if (in_pcbinshash(inp) != 0) {
1569 			inp->inp_lport = 0;
1570 			error = EAGAIN;
1571 			goto out;
1572 		}
1573 	}
1574 	inp->inp_laddr = laddr;
1575 	in_pcbrehash(inp);
1576 	INP_HASH_WUNLOCK(&V_tcbinfo);
1577 
1578 	/*
1579 	 * Compute window scaling to request:
1580 	 * Scale to fit into sweet spot.  See tcp_syncache.c.
1581 	 * XXX: This should move to tcp_output().
1582 	 */
1583 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1584 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1585 		tp->request_r_scale++;
1586 
1587 	soisconnecting(so);
1588 	TCPSTAT_INC(tcps_connattempt);
1589 	tcp_state_change(tp, TCPS_SYN_SENT);
1590 	tp->iss = tcp_new_isn(&inp->inp_inc);
1591 	if (tp->t_flags & TF_REQ_TSTMP)
1592 		tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
1593 	tcp_sendseqinit(tp);
1594 
1595 	return 0;
1596 
1597 out:
1598 	INP_HASH_WUNLOCK(&V_tcbinfo);
1599 	return (error);
1600 }
1601 #endif /* INET */
1602 
1603 #ifdef INET6
1604 static int
1605 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
1606 {
1607 	struct inpcb *inp = tp->t_inpcb;
1608 	int error;
1609 
1610 	INP_WLOCK_ASSERT(inp);
1611 	INP_HASH_WLOCK(&V_tcbinfo);
1612 
1613 	if (V_tcp_require_unique_port && inp->inp_lport == 0) {
1614 		error = in6_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
1615 		if (error)
1616 			goto out;
1617 	}
1618 	error = in6_pcbconnect(inp, nam, td->td_ucred);
1619 	if (error != 0)
1620 		goto out;
1621 	INP_HASH_WUNLOCK(&V_tcbinfo);
1622 
1623 	/* Compute window scaling to request.  */
1624 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1625 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1626 		tp->request_r_scale++;
1627 
1628 	soisconnecting(inp->inp_socket);
1629 	TCPSTAT_INC(tcps_connattempt);
1630 	tcp_state_change(tp, TCPS_SYN_SENT);
1631 	tp->iss = tcp_new_isn(&inp->inp_inc);
1632 	if (tp->t_flags & TF_REQ_TSTMP)
1633 		tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
1634 	tcp_sendseqinit(tp);
1635 
1636 	return 0;
1637 
1638 out:
1639 	INP_HASH_WUNLOCK(&V_tcbinfo);
1640 	return error;
1641 }
1642 #endif /* INET6 */
1643 
1644 /*
1645  * Export TCP internal state information via a struct tcp_info, based on the
1646  * Linux 2.6 API.  Not ABI compatible as our constants are mapped differently
1647  * (TCP state machine, etc).  We export all information using FreeBSD-native
1648  * constants -- for example, the numeric values for tcpi_state will differ
1649  * from Linux.
1650  */
1651 static void
1652 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1653 {
1654 
1655 	INP_WLOCK_ASSERT(tp->t_inpcb);
1656 	bzero(ti, sizeof(*ti));
1657 
1658 	ti->tcpi_state = tp->t_state;
1659 	if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1660 		ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1661 	if (tp->t_flags & TF_SACK_PERMIT)
1662 		ti->tcpi_options |= TCPI_OPT_SACK;
1663 	if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1664 		ti->tcpi_options |= TCPI_OPT_WSCALE;
1665 		ti->tcpi_snd_wscale = tp->snd_scale;
1666 		ti->tcpi_rcv_wscale = tp->rcv_scale;
1667 	}
1668 	if (tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT))
1669 		ti->tcpi_options |= TCPI_OPT_ECN;
1670 
1671 	ti->tcpi_rto = tp->t_rxtcur * tick;
1672 	ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick;
1673 	ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1674 	ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1675 
1676 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1677 	ti->tcpi_snd_cwnd = tp->snd_cwnd;
1678 
1679 	/*
1680 	 * FreeBSD-specific extension fields for tcp_info.
1681 	 */
1682 	ti->tcpi_rcv_space = tp->rcv_wnd;
1683 	ti->tcpi_rcv_nxt = tp->rcv_nxt;
1684 	ti->tcpi_snd_wnd = tp->snd_wnd;
1685 	ti->tcpi_snd_bwnd = 0;		/* Unused, kept for compat. */
1686 	ti->tcpi_snd_nxt = tp->snd_nxt;
1687 	ti->tcpi_snd_mss = tp->t_maxseg;
1688 	ti->tcpi_rcv_mss = tp->t_maxseg;
1689 	ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1690 	ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1691 	ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1692 #ifdef TCP_OFFLOAD
1693 	if (tp->t_flags & TF_TOE) {
1694 		ti->tcpi_options |= TCPI_OPT_TOE;
1695 		tcp_offload_tcp_info(tp, ti);
1696 	}
1697 #endif
1698 }
1699 
1700 /*
1701  * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1702  * socket option arguments.  When it re-acquires the lock after the copy, it
1703  * has to revalidate that the connection is still valid for the socket
1704  * option.
1705  */
1706 #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do {			\
1707 	INP_WLOCK(inp);							\
1708 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {		\
1709 		INP_WUNLOCK(inp);					\
1710 		cleanup;						\
1711 		return (ECONNRESET);					\
1712 	}								\
1713 	tp = intotcpcb(inp);						\
1714 } while(0)
1715 #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */)
1716 
1717 int
1718 tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt)
1719 {
1720 	struct socket *so = inp->inp_socket;
1721 	struct tcpcb *tp = intotcpcb(inp);
1722 	int error = 0;
1723 
1724 	MPASS(sopt->sopt_dir == SOPT_SET);
1725 	INP_WLOCK_ASSERT(inp);
1726 	KASSERT((inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) == 0,
1727 	    ("inp_flags == %x", inp->inp_flags));
1728 	KASSERT(so != NULL, ("inp_socket == NULL"));
1729 
1730 	if (sopt->sopt_level != IPPROTO_TCP) {
1731 		INP_WUNLOCK(inp);
1732 #ifdef INET6
1733 		if (inp->inp_vflag & INP_IPV6PROTO)
1734 			error = ip6_ctloutput(so, sopt);
1735 #endif
1736 #if defined(INET6) && defined(INET)
1737 		else
1738 #endif
1739 #ifdef INET
1740 			error = ip_ctloutput(so, sopt);
1741 #endif
1742 		/*
1743 		 * When an IP-level socket option affects TCP, pass control
1744 		 * down to stack tfb_tcp_ctloutput, otherwise return what
1745 		 * IP level returned.
1746 		 */
1747 		switch (sopt->sopt_level) {
1748 #ifdef INET6
1749 		case IPPROTO_IPV6:
1750 			if ((inp->inp_vflag & INP_IPV6PROTO) == 0)
1751 				return (error);
1752 			switch (sopt->sopt_name) {
1753 			case IPV6_TCLASS:
1754 				/* Notify tcp stacks that care (e.g. RACK). */
1755 				break;
1756 			case IPV6_USE_MIN_MTU:
1757 				/* Update t_maxseg accordingly. */
1758 				break;
1759 			default:
1760 				return (error);
1761 			}
1762 			break;
1763 #endif
1764 #ifdef INET
1765 		case IPPROTO_IP:
1766 			switch (sopt->sopt_name) {
1767 			case IP_TOS:
1768 				inp->inp_ip_tos &= ~IPTOS_ECN_MASK;
1769 				break;
1770 			case IP_TTL:
1771 				/* Notify tcp stacks that care (e.g. RACK). */
1772 				break;
1773 			default:
1774 				return (error);
1775 			}
1776 			break;
1777 #endif
1778 		default:
1779 			return (error);
1780 		}
1781 		INP_WLOCK(inp);
1782 		if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1783 			INP_WUNLOCK(inp);
1784 			return (ECONNRESET);
1785 		}
1786 	} else if (sopt->sopt_name == TCP_FUNCTION_BLK) {
1787 		/*
1788 		 * Protect the TCP option TCP_FUNCTION_BLK so
1789 		 * that a sub-function can *never* overwrite this.
1790 		 */
1791 		struct tcp_function_set fsn;
1792 		struct tcp_function_block *blk;
1793 
1794 		INP_WUNLOCK(inp);
1795 		error = sooptcopyin(sopt, &fsn, sizeof fsn, sizeof fsn);
1796 		if (error)
1797 			return (error);
1798 
1799 		INP_WLOCK(inp);
1800 		if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1801 			INP_WUNLOCK(inp);
1802 			return (ECONNRESET);
1803 		}
1804 		tp = intotcpcb(inp);
1805 
1806 		blk = find_and_ref_tcp_functions(&fsn);
1807 		if (blk == NULL) {
1808 			INP_WUNLOCK(inp);
1809 			return (ENOENT);
1810 		}
1811 		if (tp->t_fb == blk) {
1812 			/* You already have this */
1813 			refcount_release(&blk->tfb_refcnt);
1814 			INP_WUNLOCK(inp);
1815 			return (0);
1816 		}
1817 		if (tp->t_state != TCPS_CLOSED) {
1818 			/*
1819 			 * The user has advanced the state
1820 			 * past the initial point, we may not
1821 			 * be able to switch.
1822 			 */
1823 			if (blk->tfb_tcp_handoff_ok != NULL) {
1824 				/*
1825 				 * Does the stack provide a
1826 				 * query mechanism, if so it may
1827 				 * still be possible?
1828 				 */
1829 				error = (*blk->tfb_tcp_handoff_ok)(tp);
1830 			} else
1831 				error = EINVAL;
1832 			if (error) {
1833 				refcount_release(&blk->tfb_refcnt);
1834 				INP_WUNLOCK(inp);
1835 				return(error);
1836 			}
1837 		}
1838 		if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) {
1839 			refcount_release(&blk->tfb_refcnt);
1840 			INP_WUNLOCK(inp);
1841 			return (ENOENT);
1842 		}
1843 		/*
1844 		 * Release the old refcnt, the
1845 		 * lookup acquired a ref on the
1846 		 * new one already.
1847 		 */
1848 		if (tp->t_fb->tfb_tcp_fb_fini) {
1849 			struct epoch_tracker et;
1850 			/*
1851 			 * Tell the stack to cleanup with 0 i.e.
1852 			 * the tcb is not going away.
1853 			 */
1854 			NET_EPOCH_ENTER(et);
1855 			(*tp->t_fb->tfb_tcp_fb_fini)(tp, 0);
1856 			NET_EPOCH_EXIT(et);
1857 		}
1858 #ifdef TCPHPTS
1859 		/* Assure that we are not on any hpts */
1860 		tcp_hpts_remove(tp->t_inpcb);
1861 #endif
1862 		if (blk->tfb_tcp_fb_init) {
1863 			error = (*blk->tfb_tcp_fb_init)(tp);
1864 			if (error) {
1865 				refcount_release(&blk->tfb_refcnt);
1866 				if (tp->t_fb->tfb_tcp_fb_init) {
1867 					if((*tp->t_fb->tfb_tcp_fb_init)(tp) != 0)  {
1868 						/* Fall back failed, drop the connection */
1869 						INP_WUNLOCK(inp);
1870 						soabort(so);
1871 						return (error);
1872 					}
1873 				}
1874 				goto err_out;
1875 			}
1876 		}
1877 		refcount_release(&tp->t_fb->tfb_refcnt);
1878 		tp->t_fb = blk;
1879 #ifdef TCP_OFFLOAD
1880 		if (tp->t_flags & TF_TOE) {
1881 			tcp_offload_ctloutput(tp, sopt->sopt_dir,
1882 			     sopt->sopt_name);
1883 		}
1884 #endif
1885 err_out:
1886 		INP_WUNLOCK(inp);
1887 		return (error);
1888 	}
1889 
1890 	/* Pass in the INP locked, callee must unlock it. */
1891 	return (tp->t_fb->tfb_tcp_ctloutput(inp, sopt));
1892 }
1893 
1894 static int
1895 tcp_ctloutput_get(struct inpcb *inp, struct sockopt *sopt)
1896 {
1897 	struct socket *so = inp->inp_socket;
1898 	struct tcpcb *tp = intotcpcb(inp);
1899 	int error = 0;
1900 
1901 	MPASS(sopt->sopt_dir == SOPT_GET);
1902 	INP_WLOCK_ASSERT(inp);
1903 	KASSERT((inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) == 0,
1904 	    ("inp_flags == %x", inp->inp_flags));
1905 	KASSERT(so != NULL, ("inp_socket == NULL"));
1906 
1907 	if (sopt->sopt_level != IPPROTO_TCP) {
1908 		INP_WUNLOCK(inp);
1909 #ifdef INET6
1910 		if (inp->inp_vflag & INP_IPV6PROTO)
1911 			error = ip6_ctloutput(so, sopt);
1912 #endif /* INET6 */
1913 #if defined(INET6) && defined(INET)
1914 		else
1915 #endif
1916 #ifdef INET
1917 			error = ip_ctloutput(so, sopt);
1918 #endif
1919 		return (error);
1920 	}
1921 	if (((sopt->sopt_name == TCP_FUNCTION_BLK) ||
1922 	     (sopt->sopt_name == TCP_FUNCTION_ALIAS))) {
1923 		struct tcp_function_set fsn;
1924 
1925 		if (sopt->sopt_name == TCP_FUNCTION_ALIAS) {
1926 			memset(&fsn, 0, sizeof(fsn));
1927 			find_tcp_function_alias(tp->t_fb, &fsn);
1928 		} else {
1929 			strncpy(fsn.function_set_name,
1930 			    tp->t_fb->tfb_tcp_block_name,
1931 			    TCP_FUNCTION_NAME_LEN_MAX);
1932 			fsn.function_set_name[TCP_FUNCTION_NAME_LEN_MAX - 1] = '\0';
1933 		}
1934 		fsn.pcbcnt = tp->t_fb->tfb_refcnt;
1935 		INP_WUNLOCK(inp);
1936 		error = sooptcopyout(sopt, &fsn, sizeof fsn);
1937 		return (error);
1938 	}
1939 
1940 	/* Pass in the INP locked, callee must unlock it. */
1941 	return (tp->t_fb->tfb_tcp_ctloutput(inp, sopt));
1942 }
1943 
1944 int
1945 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1946 {
1947 	struct	inpcb *inp;
1948 
1949 	inp = sotoinpcb(so);
1950 	KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1951 
1952 	INP_WLOCK(inp);
1953 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
1954 		INP_WUNLOCK(inp);
1955 		return (ECONNRESET);
1956 	}
1957 	if (sopt->sopt_dir == SOPT_SET)
1958 		return (tcp_ctloutput_set(inp, sopt));
1959 	else if (sopt->sopt_dir == SOPT_GET)
1960 		return (tcp_ctloutput_get(inp, sopt));
1961 	else
1962 		panic("%s: sopt_dir $%d", __func__, sopt->sopt_dir);
1963 }
1964 
1965 /*
1966  * If this assert becomes untrue, we need to change the size of the buf
1967  * variable in tcp_default_ctloutput().
1968  */
1969 #ifdef CTASSERT
1970 CTASSERT(TCP_CA_NAME_MAX <= TCP_LOG_ID_LEN);
1971 CTASSERT(TCP_LOG_REASON_LEN <= TCP_LOG_ID_LEN);
1972 #endif
1973 
1974 #ifdef KERN_TLS
1975 static int
1976 copyin_tls_enable(struct sockopt *sopt, struct tls_enable *tls)
1977 {
1978 	struct tls_enable_v0 tls_v0;
1979 	int error;
1980 
1981 	if (sopt->sopt_valsize == sizeof(tls_v0)) {
1982 		error = sooptcopyin(sopt, &tls_v0, sizeof(tls_v0),
1983 		    sizeof(tls_v0));
1984 		if (error)
1985 			return (error);
1986 		memset(tls, 0, sizeof(*tls));
1987 		tls->cipher_key = tls_v0.cipher_key;
1988 		tls->iv = tls_v0.iv;
1989 		tls->auth_key = tls_v0.auth_key;
1990 		tls->cipher_algorithm = tls_v0.cipher_algorithm;
1991 		tls->cipher_key_len = tls_v0.cipher_key_len;
1992 		tls->iv_len = tls_v0.iv_len;
1993 		tls->auth_algorithm = tls_v0.auth_algorithm;
1994 		tls->auth_key_len = tls_v0.auth_key_len;
1995 		tls->flags = tls_v0.flags;
1996 		tls->tls_vmajor = tls_v0.tls_vmajor;
1997 		tls->tls_vminor = tls_v0.tls_vminor;
1998 		return (0);
1999 	}
2000 
2001 	return (sooptcopyin(sopt, tls, sizeof(*tls), sizeof(*tls)));
2002 }
2003 #endif
2004 
2005 extern struct cc_algo newreno_cc_algo;
2006 
2007 static int
2008 tcp_set_cc_mod(struct inpcb *inp, struct sockopt *sopt)
2009 {
2010 	struct cc_algo *algo;
2011 	void *ptr = NULL;
2012 	struct tcpcb *tp;
2013 	struct cc_var cc_mem;
2014 	char	buf[TCP_CA_NAME_MAX];
2015 	size_t mem_sz;
2016 	int error;
2017 
2018 	INP_WUNLOCK(inp);
2019 	error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1);
2020 	if (error)
2021 		return(error);
2022 	buf[sopt->sopt_valsize] = '\0';
2023 	CC_LIST_RLOCK();
2024 	STAILQ_FOREACH(algo, &cc_list, entries) {
2025 		if (strncmp(buf, algo->name,
2026 			    TCP_CA_NAME_MAX) == 0) {
2027 			if (algo->flags & CC_MODULE_BEING_REMOVED) {
2028 				/* We can't "see" modules being unloaded */
2029 				continue;
2030 			}
2031 			break;
2032 		}
2033 	}
2034 	if (algo == NULL) {
2035 		CC_LIST_RUNLOCK();
2036 		return(ESRCH);
2037 	}
2038 	/*
2039 	 * With a reference the algorithm cannot be removed
2040 	 * so we hold a reference through the change process.
2041 	 */
2042 	cc_refer(algo);
2043 	CC_LIST_RUNLOCK();
2044 	if (algo->cb_init != NULL) {
2045 		/* We can now pre-get the memory for the CC */
2046 		mem_sz = (*algo->cc_data_sz)();
2047 		if (mem_sz == 0) {
2048 			goto no_mem_needed;
2049 		}
2050 		ptr = malloc(mem_sz, M_CC_MEM, M_WAITOK);
2051 	} else {
2052 no_mem_needed:
2053 		mem_sz = 0;
2054 		ptr = NULL;
2055 	}
2056 	/*
2057 	 * Make sure its all clean and zero and also get
2058 	 * back the inplock.
2059 	 */
2060 	memset(&cc_mem, 0, sizeof(cc_mem));
2061 	INP_WLOCK(inp);
2062 	if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
2063 		INP_WUNLOCK(inp);
2064 		if (ptr)
2065 			free(ptr, M_CC_MEM);
2066 		/* Release our temp reference */
2067 		CC_LIST_RLOCK();
2068 		cc_release(algo);
2069 		CC_LIST_RUNLOCK();
2070 		return (ECONNRESET);
2071 	}
2072 	tp = intotcpcb(inp);
2073 	if (ptr != NULL)
2074 		memset(ptr, 0, mem_sz);
2075 	cc_mem.ccvc.tcp = tp;
2076 	/*
2077 	 * We once again hold a write lock over the tcb so it's
2078 	 * safe to do these things without ordering concerns.
2079 	 * Note here we init into stack memory.
2080 	 */
2081 	if (algo->cb_init != NULL)
2082 		error = algo->cb_init(&cc_mem, ptr);
2083 	else
2084 		error = 0;
2085 	/*
2086 	 * The CC algorithms, when given their memory
2087 	 * should not fail we could in theory have a
2088 	 * KASSERT here.
2089 	 */
2090 	if (error == 0) {
2091 		/*
2092 		 * Touchdown, lets go ahead and move the
2093 		 * connection to the new CC module by
2094 		 * copying in the cc_mem after we call
2095 		 * the old ones cleanup (if any).
2096 		 */
2097 		if (CC_ALGO(tp)->cb_destroy != NULL)
2098 			CC_ALGO(tp)->cb_destroy(tp->ccv);
2099 		/* Detach the old CC from the tcpcb  */
2100 		cc_detach(tp);
2101 		/* Copy in our temp memory that was inited */
2102 		memcpy(tp->ccv, &cc_mem, sizeof(struct cc_var));
2103 		/* Now attach the new, which takes a reference */
2104 		cc_attach(tp, algo);
2105 		/* Ok now are we where we have gotten past any conn_init? */
2106 		if (TCPS_HAVEESTABLISHED(tp->t_state) && (CC_ALGO(tp)->conn_init != NULL)) {
2107 			/* Yep run the connection init for the new CC */
2108 			CC_ALGO(tp)->conn_init(tp->ccv);
2109 		}
2110 	} else if (ptr)
2111 		free(ptr, M_CC_MEM);
2112 	INP_WUNLOCK(inp);
2113 	/* Now lets release our temp reference */
2114 	CC_LIST_RLOCK();
2115 	cc_release(algo);
2116 	CC_LIST_RUNLOCK();
2117 	return (error);
2118 }
2119 
2120 int
2121 tcp_default_ctloutput(struct inpcb *inp, struct sockopt *sopt)
2122 {
2123 	struct tcpcb *tp = intotcpcb(inp);
2124 	int	error, opt, optval;
2125 	u_int	ui;
2126 	struct	tcp_info ti;
2127 #ifdef KERN_TLS
2128 	struct tls_enable tls;
2129 	struct socket *so = inp->inp_socket;
2130 #endif
2131 	char	*pbuf, buf[TCP_LOG_ID_LEN];
2132 #ifdef STATS
2133 	struct statsblob *sbp;
2134 #endif
2135 	size_t	len;
2136 
2137 	INP_WLOCK_ASSERT(inp);
2138 	KASSERT((inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) == 0,
2139 	    ("inp_flags == %x", inp->inp_flags));
2140 	KASSERT(inp->inp_socket != NULL, ("inp_socket == NULL"));
2141 
2142 	switch (sopt->sopt_level) {
2143 #ifdef INET6
2144 	case IPPROTO_IPV6:
2145 		MPASS(inp->inp_vflag & INP_IPV6PROTO);
2146 		switch (sopt->sopt_name) {
2147 		case IPV6_USE_MIN_MTU:
2148 			tcp6_use_min_mtu(tp);
2149 			/* FALLTHROUGH */
2150 		}
2151 		INP_WUNLOCK(inp);
2152 		return (0);
2153 #endif
2154 #ifdef INET
2155 	case IPPROTO_IP:
2156 		INP_WUNLOCK(inp);
2157 		return (0);
2158 #endif
2159 	}
2160 
2161 	/*
2162 	 * For TCP_CCALGOOPT forward the control to CC module, for both
2163 	 * SOPT_SET and SOPT_GET.
2164 	 */
2165 	switch (sopt->sopt_name) {
2166 	case TCP_CCALGOOPT:
2167 		INP_WUNLOCK(inp);
2168 		if (sopt->sopt_valsize > CC_ALGOOPT_LIMIT)
2169 			return (EINVAL);
2170 		pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO);
2171 		error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize,
2172 		    sopt->sopt_valsize);
2173 		if (error) {
2174 			free(pbuf, M_TEMP);
2175 			return (error);
2176 		}
2177 		INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP));
2178 		if (CC_ALGO(tp)->ctl_output != NULL)
2179 			error = CC_ALGO(tp)->ctl_output(tp->ccv, sopt, pbuf);
2180 		else
2181 			error = ENOENT;
2182 		INP_WUNLOCK(inp);
2183 		if (error == 0 && sopt->sopt_dir == SOPT_GET)
2184 			error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize);
2185 		free(pbuf, M_TEMP);
2186 		return (error);
2187 	}
2188 
2189 	switch (sopt->sopt_dir) {
2190 	case SOPT_SET:
2191 		switch (sopt->sopt_name) {
2192 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
2193 		case TCP_MD5SIG:
2194 			INP_WUNLOCK(inp);
2195 			if (!TCPMD5_ENABLED())
2196 				return (ENOPROTOOPT);
2197 			error = TCPMD5_PCBCTL(inp, sopt);
2198 			if (error)
2199 				return (error);
2200 			INP_WLOCK_RECHECK(inp);
2201 			goto unlock_and_done;
2202 #endif /* IPSEC */
2203 
2204 		case TCP_NODELAY:
2205 		case TCP_NOOPT:
2206 		case TCP_LRD:
2207 			INP_WUNLOCK(inp);
2208 			error = sooptcopyin(sopt, &optval, sizeof optval,
2209 			    sizeof optval);
2210 			if (error)
2211 				return (error);
2212 
2213 			INP_WLOCK_RECHECK(inp);
2214 			switch (sopt->sopt_name) {
2215 			case TCP_NODELAY:
2216 				opt = TF_NODELAY;
2217 				break;
2218 			case TCP_NOOPT:
2219 				opt = TF_NOOPT;
2220 				break;
2221 			case TCP_LRD:
2222 				opt = TF_LRD;
2223 				break;
2224 			default:
2225 				opt = 0; /* dead code to fool gcc */
2226 				break;
2227 			}
2228 
2229 			if (optval)
2230 				tp->t_flags |= opt;
2231 			else
2232 				tp->t_flags &= ~opt;
2233 unlock_and_done:
2234 #ifdef TCP_OFFLOAD
2235 			if (tp->t_flags & TF_TOE) {
2236 				tcp_offload_ctloutput(tp, sopt->sopt_dir,
2237 				    sopt->sopt_name);
2238 			}
2239 #endif
2240 			INP_WUNLOCK(inp);
2241 			break;
2242 
2243 		case TCP_NOPUSH:
2244 			INP_WUNLOCK(inp);
2245 			error = sooptcopyin(sopt, &optval, sizeof optval,
2246 			    sizeof optval);
2247 			if (error)
2248 				return (error);
2249 
2250 			INP_WLOCK_RECHECK(inp);
2251 			if (optval)
2252 				tp->t_flags |= TF_NOPUSH;
2253 			else if (tp->t_flags & TF_NOPUSH) {
2254 				tp->t_flags &= ~TF_NOPUSH;
2255 				if (TCPS_HAVEESTABLISHED(tp->t_state)) {
2256 					struct epoch_tracker et;
2257 
2258 					NET_EPOCH_ENTER(et);
2259 					error = tcp_output_nodrop(tp);
2260 					NET_EPOCH_EXIT(et);
2261 				}
2262 			}
2263 			goto unlock_and_done;
2264 
2265 		case TCP_REMOTE_UDP_ENCAPS_PORT:
2266 			INP_WUNLOCK(inp);
2267 			error = sooptcopyin(sopt, &optval, sizeof optval,
2268 			    sizeof optval);
2269 			if (error)
2270 				return (error);
2271 			if ((optval < TCP_TUNNELING_PORT_MIN) ||
2272 			    (optval > TCP_TUNNELING_PORT_MAX)) {
2273 				/* Its got to be in range */
2274 				return (EINVAL);
2275 			}
2276 			if ((V_tcp_udp_tunneling_port == 0) && (optval != 0)) {
2277 				/* You have to have enabled a UDP tunneling port first */
2278 				return (EINVAL);
2279 			}
2280 			INP_WLOCK_RECHECK(inp);
2281 			if (tp->t_state != TCPS_CLOSED) {
2282 				/* You can't change after you are connected */
2283 				error = EINVAL;
2284 			} else {
2285 				/* Ok we are all good set the port */
2286 				tp->t_port = htons(optval);
2287 			}
2288 			goto unlock_and_done;
2289 
2290 		case TCP_MAXSEG:
2291 			INP_WUNLOCK(inp);
2292 			error = sooptcopyin(sopt, &optval, sizeof optval,
2293 			    sizeof optval);
2294 			if (error)
2295 				return (error);
2296 
2297 			INP_WLOCK_RECHECK(inp);
2298 			if (optval > 0 && optval <= tp->t_maxseg &&
2299 			    optval + 40 >= V_tcp_minmss)
2300 				tp->t_maxseg = optval;
2301 			else
2302 				error = EINVAL;
2303 			goto unlock_and_done;
2304 
2305 		case TCP_INFO:
2306 			INP_WUNLOCK(inp);
2307 			error = EINVAL;
2308 			break;
2309 
2310 		case TCP_STATS:
2311 			INP_WUNLOCK(inp);
2312 #ifdef STATS
2313 			error = sooptcopyin(sopt, &optval, sizeof optval,
2314 			    sizeof optval);
2315 			if (error)
2316 				return (error);
2317 
2318 			if (optval > 0)
2319 				sbp = stats_blob_alloc(
2320 				    V_tcp_perconn_stats_dflt_tpl, 0);
2321 			else
2322 				sbp = NULL;
2323 
2324 			INP_WLOCK_RECHECK(inp);
2325 			if ((tp->t_stats != NULL && sbp == NULL) ||
2326 			    (tp->t_stats == NULL && sbp != NULL)) {
2327 				struct statsblob *t = tp->t_stats;
2328 				tp->t_stats = sbp;
2329 				sbp = t;
2330 			}
2331 			INP_WUNLOCK(inp);
2332 
2333 			stats_blob_destroy(sbp);
2334 #else
2335 			return (EOPNOTSUPP);
2336 #endif /* !STATS */
2337 			break;
2338 
2339 		case TCP_CONGESTION:
2340 			error = tcp_set_cc_mod(inp, sopt);
2341 			break;
2342 
2343 		case TCP_REUSPORT_LB_NUMA:
2344 			INP_WUNLOCK(inp);
2345 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2346 			    sizeof(optval));
2347 			INP_WLOCK_RECHECK(inp);
2348 			if (!error)
2349 				error = in_pcblbgroup_numa(inp, optval);
2350 			INP_WUNLOCK(inp);
2351 			break;
2352 
2353 #ifdef KERN_TLS
2354 		case TCP_TXTLS_ENABLE:
2355 			INP_WUNLOCK(inp);
2356 			error = copyin_tls_enable(sopt, &tls);
2357 			if (error)
2358 				break;
2359 			error = ktls_enable_tx(so, &tls);
2360 			break;
2361 		case TCP_TXTLS_MODE:
2362 			INP_WUNLOCK(inp);
2363 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2364 			if (error)
2365 				return (error);
2366 
2367 			INP_WLOCK_RECHECK(inp);
2368 			error = ktls_set_tx_mode(so, ui);
2369 			INP_WUNLOCK(inp);
2370 			break;
2371 		case TCP_RXTLS_ENABLE:
2372 			INP_WUNLOCK(inp);
2373 			error = sooptcopyin(sopt, &tls, sizeof(tls),
2374 			    sizeof(tls));
2375 			if (error)
2376 				break;
2377 			error = ktls_enable_rx(so, &tls);
2378 			break;
2379 #endif
2380 
2381 		case TCP_KEEPIDLE:
2382 		case TCP_KEEPINTVL:
2383 		case TCP_KEEPINIT:
2384 			INP_WUNLOCK(inp);
2385 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2386 			if (error)
2387 				return (error);
2388 
2389 			if (ui > (UINT_MAX / hz)) {
2390 				error = EINVAL;
2391 				break;
2392 			}
2393 			ui *= hz;
2394 
2395 			INP_WLOCK_RECHECK(inp);
2396 			switch (sopt->sopt_name) {
2397 			case TCP_KEEPIDLE:
2398 				tp->t_keepidle = ui;
2399 				/*
2400 				 * XXX: better check current remaining
2401 				 * timeout and "merge" it with new value.
2402 				 */
2403 				if ((tp->t_state > TCPS_LISTEN) &&
2404 				    (tp->t_state <= TCPS_CLOSING))
2405 					tcp_timer_activate(tp, TT_KEEP,
2406 					    TP_KEEPIDLE(tp));
2407 				break;
2408 			case TCP_KEEPINTVL:
2409 				tp->t_keepintvl = ui;
2410 				if ((tp->t_state == TCPS_FIN_WAIT_2) &&
2411 				    (TP_MAXIDLE(tp) > 0))
2412 					tcp_timer_activate(tp, TT_2MSL,
2413 					    TP_MAXIDLE(tp));
2414 				break;
2415 			case TCP_KEEPINIT:
2416 				tp->t_keepinit = ui;
2417 				if (tp->t_state == TCPS_SYN_RECEIVED ||
2418 				    tp->t_state == TCPS_SYN_SENT)
2419 					tcp_timer_activate(tp, TT_KEEP,
2420 					    TP_KEEPINIT(tp));
2421 				break;
2422 			}
2423 			goto unlock_and_done;
2424 
2425 		case TCP_KEEPCNT:
2426 			INP_WUNLOCK(inp);
2427 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2428 			if (error)
2429 				return (error);
2430 
2431 			INP_WLOCK_RECHECK(inp);
2432 			tp->t_keepcnt = ui;
2433 			if ((tp->t_state == TCPS_FIN_WAIT_2) &&
2434 			    (TP_MAXIDLE(tp) > 0))
2435 				tcp_timer_activate(tp, TT_2MSL,
2436 				    TP_MAXIDLE(tp));
2437 			goto unlock_and_done;
2438 
2439 #ifdef TCPPCAP
2440 		case TCP_PCAP_OUT:
2441 		case TCP_PCAP_IN:
2442 			INP_WUNLOCK(inp);
2443 			error = sooptcopyin(sopt, &optval, sizeof optval,
2444 			    sizeof optval);
2445 			if (error)
2446 				return (error);
2447 
2448 			INP_WLOCK_RECHECK(inp);
2449 			if (optval >= 0)
2450 				tcp_pcap_set_sock_max(TCP_PCAP_OUT ?
2451 					&(tp->t_outpkts) : &(tp->t_inpkts),
2452 					optval);
2453 			else
2454 				error = EINVAL;
2455 			goto unlock_and_done;
2456 #endif
2457 
2458 		case TCP_FASTOPEN: {
2459 			struct tcp_fastopen tfo_optval;
2460 
2461 			INP_WUNLOCK(inp);
2462 			if (!V_tcp_fastopen_client_enable &&
2463 			    !V_tcp_fastopen_server_enable)
2464 				return (EPERM);
2465 
2466 			error = sooptcopyin(sopt, &tfo_optval,
2467 				    sizeof(tfo_optval), sizeof(int));
2468 			if (error)
2469 				return (error);
2470 
2471 			INP_WLOCK_RECHECK(inp);
2472 			if ((tp->t_state != TCPS_CLOSED) &&
2473 			    (tp->t_state != TCPS_LISTEN)) {
2474 				error = EINVAL;
2475 				goto unlock_and_done;
2476 			}
2477 			if (tfo_optval.enable) {
2478 				if (tp->t_state == TCPS_LISTEN) {
2479 					if (!V_tcp_fastopen_server_enable) {
2480 						error = EPERM;
2481 						goto unlock_and_done;
2482 					}
2483 
2484 					if (tp->t_tfo_pending == NULL)
2485 						tp->t_tfo_pending =
2486 						    tcp_fastopen_alloc_counter();
2487 				} else {
2488 					/*
2489 					 * If a pre-shared key was provided,
2490 					 * stash it in the client cookie
2491 					 * field of the tcpcb for use during
2492 					 * connect.
2493 					 */
2494 					if (sopt->sopt_valsize ==
2495 					    sizeof(tfo_optval)) {
2496 						memcpy(tp->t_tfo_cookie.client,
2497 						       tfo_optval.psk,
2498 						       TCP_FASTOPEN_PSK_LEN);
2499 						tp->t_tfo_client_cookie_len =
2500 						    TCP_FASTOPEN_PSK_LEN;
2501 					}
2502 				}
2503 				tp->t_flags |= TF_FASTOPEN;
2504 			} else
2505 				tp->t_flags &= ~TF_FASTOPEN;
2506 			goto unlock_and_done;
2507 		}
2508 
2509 #ifdef TCP_BLACKBOX
2510 		case TCP_LOG:
2511 			INP_WUNLOCK(inp);
2512 			error = sooptcopyin(sopt, &optval, sizeof optval,
2513 			    sizeof optval);
2514 			if (error)
2515 				return (error);
2516 
2517 			INP_WLOCK_RECHECK(inp);
2518 			error = tcp_log_state_change(tp, optval);
2519 			goto unlock_and_done;
2520 
2521 		case TCP_LOGBUF:
2522 			INP_WUNLOCK(inp);
2523 			error = EINVAL;
2524 			break;
2525 
2526 		case TCP_LOGID:
2527 			INP_WUNLOCK(inp);
2528 			error = sooptcopyin(sopt, buf, TCP_LOG_ID_LEN - 1, 0);
2529 			if (error)
2530 				break;
2531 			buf[sopt->sopt_valsize] = '\0';
2532 			INP_WLOCK_RECHECK(inp);
2533 			error = tcp_log_set_id(tp, buf);
2534 			/* tcp_log_set_id() unlocks the INP. */
2535 			break;
2536 
2537 		case TCP_LOGDUMP:
2538 		case TCP_LOGDUMPID:
2539 			INP_WUNLOCK(inp);
2540 			error =
2541 			    sooptcopyin(sopt, buf, TCP_LOG_REASON_LEN - 1, 0);
2542 			if (error)
2543 				break;
2544 			buf[sopt->sopt_valsize] = '\0';
2545 			INP_WLOCK_RECHECK(inp);
2546 			if (sopt->sopt_name == TCP_LOGDUMP) {
2547 				error = tcp_log_dump_tp_logbuf(tp, buf,
2548 				    M_WAITOK, true);
2549 				INP_WUNLOCK(inp);
2550 			} else {
2551 				tcp_log_dump_tp_bucket_logbufs(tp, buf);
2552 				/*
2553 				 * tcp_log_dump_tp_bucket_logbufs() drops the
2554 				 * INP lock.
2555 				 */
2556 			}
2557 			break;
2558 #endif
2559 
2560 		default:
2561 			INP_WUNLOCK(inp);
2562 			error = ENOPROTOOPT;
2563 			break;
2564 		}
2565 		break;
2566 
2567 	case SOPT_GET:
2568 		tp = intotcpcb(inp);
2569 		switch (sopt->sopt_name) {
2570 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
2571 		case TCP_MD5SIG:
2572 			INP_WUNLOCK(inp);
2573 			if (!TCPMD5_ENABLED())
2574 				return (ENOPROTOOPT);
2575 			error = TCPMD5_PCBCTL(inp, sopt);
2576 			break;
2577 #endif
2578 
2579 		case TCP_NODELAY:
2580 			optval = tp->t_flags & TF_NODELAY;
2581 			INP_WUNLOCK(inp);
2582 			error = sooptcopyout(sopt, &optval, sizeof optval);
2583 			break;
2584 		case TCP_MAXSEG:
2585 			optval = tp->t_maxseg;
2586 			INP_WUNLOCK(inp);
2587 			error = sooptcopyout(sopt, &optval, sizeof optval);
2588 			break;
2589 		case TCP_REMOTE_UDP_ENCAPS_PORT:
2590 			optval = ntohs(tp->t_port);
2591 			INP_WUNLOCK(inp);
2592 			error = sooptcopyout(sopt, &optval, sizeof optval);
2593 			break;
2594 		case TCP_NOOPT:
2595 			optval = tp->t_flags & TF_NOOPT;
2596 			INP_WUNLOCK(inp);
2597 			error = sooptcopyout(sopt, &optval, sizeof optval);
2598 			break;
2599 		case TCP_NOPUSH:
2600 			optval = tp->t_flags & TF_NOPUSH;
2601 			INP_WUNLOCK(inp);
2602 			error = sooptcopyout(sopt, &optval, sizeof optval);
2603 			break;
2604 		case TCP_INFO:
2605 			tcp_fill_info(tp, &ti);
2606 			INP_WUNLOCK(inp);
2607 			error = sooptcopyout(sopt, &ti, sizeof ti);
2608 			break;
2609 		case TCP_STATS:
2610 			{
2611 #ifdef STATS
2612 			int nheld;
2613 			TYPEOF_MEMBER(struct statsblob, flags) sbflags = 0;
2614 
2615 			error = 0;
2616 			socklen_t outsbsz = sopt->sopt_valsize;
2617 			if (tp->t_stats == NULL)
2618 				error = ENOENT;
2619 			else if (outsbsz >= tp->t_stats->cursz)
2620 				outsbsz = tp->t_stats->cursz;
2621 			else if (outsbsz >= sizeof(struct statsblob))
2622 				outsbsz = sizeof(struct statsblob);
2623 			else
2624 				error = EINVAL;
2625 			INP_WUNLOCK(inp);
2626 			if (error)
2627 				break;
2628 
2629 			sbp = sopt->sopt_val;
2630 			nheld = atop(round_page(((vm_offset_t)sbp) +
2631 			    (vm_size_t)outsbsz) - trunc_page((vm_offset_t)sbp));
2632 			vm_page_t ma[nheld];
2633 			if (vm_fault_quick_hold_pages(
2634 			    &curproc->p_vmspace->vm_map, (vm_offset_t)sbp,
2635 			    outsbsz, VM_PROT_READ | VM_PROT_WRITE, ma,
2636 			    nheld) < 0) {
2637 				error = EFAULT;
2638 				break;
2639 			}
2640 
2641 			if ((error = copyin_nofault(&(sbp->flags), &sbflags,
2642 			    SIZEOF_MEMBER(struct statsblob, flags))))
2643 				goto unhold;
2644 
2645 			INP_WLOCK_RECHECK(inp);
2646 			error = stats_blob_snapshot(&sbp, outsbsz, tp->t_stats,
2647 			    sbflags | SB_CLONE_USRDSTNOFAULT);
2648 			INP_WUNLOCK(inp);
2649 			sopt->sopt_valsize = outsbsz;
2650 unhold:
2651 			vm_page_unhold_pages(ma, nheld);
2652 #else
2653 			INP_WUNLOCK(inp);
2654 			error = EOPNOTSUPP;
2655 #endif /* !STATS */
2656 			break;
2657 			}
2658 		case TCP_CONGESTION:
2659 			len = strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
2660 			INP_WUNLOCK(inp);
2661 			error = sooptcopyout(sopt, buf, len + 1);
2662 			break;
2663 		case TCP_KEEPIDLE:
2664 		case TCP_KEEPINTVL:
2665 		case TCP_KEEPINIT:
2666 		case TCP_KEEPCNT:
2667 			switch (sopt->sopt_name) {
2668 			case TCP_KEEPIDLE:
2669 				ui = TP_KEEPIDLE(tp) / hz;
2670 				break;
2671 			case TCP_KEEPINTVL:
2672 				ui = TP_KEEPINTVL(tp) / hz;
2673 				break;
2674 			case TCP_KEEPINIT:
2675 				ui = TP_KEEPINIT(tp) / hz;
2676 				break;
2677 			case TCP_KEEPCNT:
2678 				ui = TP_KEEPCNT(tp);
2679 				break;
2680 			}
2681 			INP_WUNLOCK(inp);
2682 			error = sooptcopyout(sopt, &ui, sizeof(ui));
2683 			break;
2684 #ifdef TCPPCAP
2685 		case TCP_PCAP_OUT:
2686 		case TCP_PCAP_IN:
2687 			optval = tcp_pcap_get_sock_max(TCP_PCAP_OUT ?
2688 					&(tp->t_outpkts) : &(tp->t_inpkts));
2689 			INP_WUNLOCK(inp);
2690 			error = sooptcopyout(sopt, &optval, sizeof optval);
2691 			break;
2692 #endif
2693 		case TCP_FASTOPEN:
2694 			optval = tp->t_flags & TF_FASTOPEN;
2695 			INP_WUNLOCK(inp);
2696 			error = sooptcopyout(sopt, &optval, sizeof optval);
2697 			break;
2698 #ifdef TCP_BLACKBOX
2699 		case TCP_LOG:
2700 			optval = tp->t_logstate;
2701 			INP_WUNLOCK(inp);
2702 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2703 			break;
2704 		case TCP_LOGBUF:
2705 			/* tcp_log_getlogbuf() does INP_WUNLOCK(inp) */
2706 			error = tcp_log_getlogbuf(sopt, tp);
2707 			break;
2708 		case TCP_LOGID:
2709 			len = tcp_log_get_id(tp, buf);
2710 			INP_WUNLOCK(inp);
2711 			error = sooptcopyout(sopt, buf, len + 1);
2712 			break;
2713 		case TCP_LOGDUMP:
2714 		case TCP_LOGDUMPID:
2715 			INP_WUNLOCK(inp);
2716 			error = EINVAL;
2717 			break;
2718 #endif
2719 #ifdef KERN_TLS
2720 		case TCP_TXTLS_MODE:
2721 			error = ktls_get_tx_mode(so, &optval);
2722 			INP_WUNLOCK(inp);
2723 			if (error == 0)
2724 				error = sooptcopyout(sopt, &optval,
2725 				    sizeof(optval));
2726 			break;
2727 		case TCP_RXTLS_MODE:
2728 			error = ktls_get_rx_mode(so, &optval);
2729 			INP_WUNLOCK(inp);
2730 			if (error == 0)
2731 				error = sooptcopyout(sopt, &optval,
2732 				    sizeof(optval));
2733 			break;
2734 #endif
2735 		case TCP_LRD:
2736 			optval = tp->t_flags & TF_LRD;
2737 			INP_WUNLOCK(inp);
2738 			error = sooptcopyout(sopt, &optval, sizeof optval);
2739 			break;
2740 		default:
2741 			INP_WUNLOCK(inp);
2742 			error = ENOPROTOOPT;
2743 			break;
2744 		}
2745 		break;
2746 	}
2747 	return (error);
2748 }
2749 #undef INP_WLOCK_RECHECK
2750 #undef INP_WLOCK_RECHECK_CLEANUP
2751 
2752 /*
2753  * Initiate (or continue) disconnect.
2754  * If embryonic state, just send reset (once).
2755  * If in ``let data drain'' option and linger null, just drop.
2756  * Otherwise (hard), mark socket disconnecting and drop
2757  * current input data; switch states based on user close, and
2758  * send segment to peer (with FIN).
2759  */
2760 static void
2761 tcp_disconnect(struct tcpcb *tp)
2762 {
2763 	struct inpcb *inp = tp->t_inpcb;
2764 	struct socket *so = inp->inp_socket;
2765 
2766 	NET_EPOCH_ASSERT();
2767 	INP_WLOCK_ASSERT(inp);
2768 
2769 	/*
2770 	 * Neither tcp_close() nor tcp_drop() should return NULL, as the
2771 	 * socket is still open.
2772 	 */
2773 	if (tp->t_state < TCPS_ESTABLISHED &&
2774 	    !(tp->t_state > TCPS_LISTEN && IS_FASTOPEN(tp->t_flags))) {
2775 		tp = tcp_close(tp);
2776 		KASSERT(tp != NULL,
2777 		    ("tcp_disconnect: tcp_close() returned NULL"));
2778 	} else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
2779 		tp = tcp_drop(tp, 0);
2780 		KASSERT(tp != NULL,
2781 		    ("tcp_disconnect: tcp_drop() returned NULL"));
2782 	} else {
2783 		soisdisconnecting(so);
2784 		sbflush(&so->so_rcv);
2785 		tcp_usrclosed(tp);
2786 		if (!(inp->inp_flags & INP_DROPPED))
2787 			/* Ignore stack's drop request, we already at it. */
2788 			(void)tcp_output_nodrop(tp);
2789 	}
2790 }
2791 
2792 /*
2793  * User issued close, and wish to trail through shutdown states:
2794  * if never received SYN, just forget it.  If got a SYN from peer,
2795  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
2796  * If already got a FIN from peer, then almost done; go to LAST_ACK
2797  * state.  In all other cases, have already sent FIN to peer (e.g.
2798  * after PRU_SHUTDOWN), and just have to play tedious game waiting
2799  * for peer to send FIN or not respond to keep-alives, etc.
2800  * We can let the user exit from the close as soon as the FIN is acked.
2801  */
2802 static void
2803 tcp_usrclosed(struct tcpcb *tp)
2804 {
2805 
2806 	NET_EPOCH_ASSERT();
2807 	INP_WLOCK_ASSERT(tp->t_inpcb);
2808 
2809 	switch (tp->t_state) {
2810 	case TCPS_LISTEN:
2811 #ifdef TCP_OFFLOAD
2812 		tcp_offload_listen_stop(tp);
2813 #endif
2814 		tcp_state_change(tp, TCPS_CLOSED);
2815 		/* FALLTHROUGH */
2816 	case TCPS_CLOSED:
2817 		tp = tcp_close(tp);
2818 		/*
2819 		 * tcp_close() should never return NULL here as the socket is
2820 		 * still open.
2821 		 */
2822 		KASSERT(tp != NULL,
2823 		    ("tcp_usrclosed: tcp_close() returned NULL"));
2824 		break;
2825 
2826 	case TCPS_SYN_SENT:
2827 	case TCPS_SYN_RECEIVED:
2828 		tp->t_flags |= TF_NEEDFIN;
2829 		break;
2830 
2831 	case TCPS_ESTABLISHED:
2832 		tcp_state_change(tp, TCPS_FIN_WAIT_1);
2833 		break;
2834 
2835 	case TCPS_CLOSE_WAIT:
2836 		tcp_state_change(tp, TCPS_LAST_ACK);
2837 		break;
2838 	}
2839 	if (tp->t_state >= TCPS_FIN_WAIT_2) {
2840 		soisdisconnected(tp->t_inpcb->inp_socket);
2841 		/* Prevent the connection hanging in FIN_WAIT_2 forever. */
2842 		if (tp->t_state == TCPS_FIN_WAIT_2) {
2843 			int timeout;
2844 
2845 			timeout = (tcp_fast_finwait2_recycle) ?
2846 			    tcp_finwait2_timeout : TP_MAXIDLE(tp);
2847 			tcp_timer_activate(tp, TT_2MSL, timeout);
2848 		}
2849 	}
2850 }
2851 
2852 #ifdef DDB
2853 static void
2854 db_print_indent(int indent)
2855 {
2856 	int i;
2857 
2858 	for (i = 0; i < indent; i++)
2859 		db_printf(" ");
2860 }
2861 
2862 static void
2863 db_print_tstate(int t_state)
2864 {
2865 
2866 	switch (t_state) {
2867 	case TCPS_CLOSED:
2868 		db_printf("TCPS_CLOSED");
2869 		return;
2870 
2871 	case TCPS_LISTEN:
2872 		db_printf("TCPS_LISTEN");
2873 		return;
2874 
2875 	case TCPS_SYN_SENT:
2876 		db_printf("TCPS_SYN_SENT");
2877 		return;
2878 
2879 	case TCPS_SYN_RECEIVED:
2880 		db_printf("TCPS_SYN_RECEIVED");
2881 		return;
2882 
2883 	case TCPS_ESTABLISHED:
2884 		db_printf("TCPS_ESTABLISHED");
2885 		return;
2886 
2887 	case TCPS_CLOSE_WAIT:
2888 		db_printf("TCPS_CLOSE_WAIT");
2889 		return;
2890 
2891 	case TCPS_FIN_WAIT_1:
2892 		db_printf("TCPS_FIN_WAIT_1");
2893 		return;
2894 
2895 	case TCPS_CLOSING:
2896 		db_printf("TCPS_CLOSING");
2897 		return;
2898 
2899 	case TCPS_LAST_ACK:
2900 		db_printf("TCPS_LAST_ACK");
2901 		return;
2902 
2903 	case TCPS_FIN_WAIT_2:
2904 		db_printf("TCPS_FIN_WAIT_2");
2905 		return;
2906 
2907 	case TCPS_TIME_WAIT:
2908 		db_printf("TCPS_TIME_WAIT");
2909 		return;
2910 
2911 	default:
2912 		db_printf("unknown");
2913 		return;
2914 	}
2915 }
2916 
2917 static void
2918 db_print_tflags(u_int t_flags)
2919 {
2920 	int comma;
2921 
2922 	comma = 0;
2923 	if (t_flags & TF_ACKNOW) {
2924 		db_printf("%sTF_ACKNOW", comma ? ", " : "");
2925 		comma = 1;
2926 	}
2927 	if (t_flags & TF_DELACK) {
2928 		db_printf("%sTF_DELACK", comma ? ", " : "");
2929 		comma = 1;
2930 	}
2931 	if (t_flags & TF_NODELAY) {
2932 		db_printf("%sTF_NODELAY", comma ? ", " : "");
2933 		comma = 1;
2934 	}
2935 	if (t_flags & TF_NOOPT) {
2936 		db_printf("%sTF_NOOPT", comma ? ", " : "");
2937 		comma = 1;
2938 	}
2939 	if (t_flags & TF_SENTFIN) {
2940 		db_printf("%sTF_SENTFIN", comma ? ", " : "");
2941 		comma = 1;
2942 	}
2943 	if (t_flags & TF_REQ_SCALE) {
2944 		db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
2945 		comma = 1;
2946 	}
2947 	if (t_flags & TF_RCVD_SCALE) {
2948 		db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
2949 		comma = 1;
2950 	}
2951 	if (t_flags & TF_REQ_TSTMP) {
2952 		db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
2953 		comma = 1;
2954 	}
2955 	if (t_flags & TF_RCVD_TSTMP) {
2956 		db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
2957 		comma = 1;
2958 	}
2959 	if (t_flags & TF_SACK_PERMIT) {
2960 		db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
2961 		comma = 1;
2962 	}
2963 	if (t_flags & TF_NEEDSYN) {
2964 		db_printf("%sTF_NEEDSYN", comma ? ", " : "");
2965 		comma = 1;
2966 	}
2967 	if (t_flags & TF_NEEDFIN) {
2968 		db_printf("%sTF_NEEDFIN", comma ? ", " : "");
2969 		comma = 1;
2970 	}
2971 	if (t_flags & TF_NOPUSH) {
2972 		db_printf("%sTF_NOPUSH", comma ? ", " : "");
2973 		comma = 1;
2974 	}
2975 	if (t_flags & TF_PREVVALID) {
2976 		db_printf("%sTF_PREVVALID", comma ? ", " : "");
2977 		comma = 1;
2978 	}
2979 	if (t_flags & TF_MORETOCOME) {
2980 		db_printf("%sTF_MORETOCOME", comma ? ", " : "");
2981 		comma = 1;
2982 	}
2983 	if (t_flags & TF_LQ_OVERFLOW) {
2984 		db_printf("%sTF_LQ_OVERFLOW", comma ? ", " : "");
2985 		comma = 1;
2986 	}
2987 	if (t_flags & TF_LASTIDLE) {
2988 		db_printf("%sTF_LASTIDLE", comma ? ", " : "");
2989 		comma = 1;
2990 	}
2991 	if (t_flags & TF_RXWIN0SENT) {
2992 		db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
2993 		comma = 1;
2994 	}
2995 	if (t_flags & TF_FASTRECOVERY) {
2996 		db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
2997 		comma = 1;
2998 	}
2999 	if (t_flags & TF_CONGRECOVERY) {
3000 		db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
3001 		comma = 1;
3002 	}
3003 	if (t_flags & TF_WASFRECOVERY) {
3004 		db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
3005 		comma = 1;
3006 	}
3007 	if (t_flags & TF_WASCRECOVERY) {
3008 		db_printf("%sTF_WASCRECOVERY", comma ? ", " : "");
3009 		comma = 1;
3010 	}
3011 	if (t_flags & TF_SIGNATURE) {
3012 		db_printf("%sTF_SIGNATURE", comma ? ", " : "");
3013 		comma = 1;
3014 	}
3015 	if (t_flags & TF_FORCEDATA) {
3016 		db_printf("%sTF_FORCEDATA", comma ? ", " : "");
3017 		comma = 1;
3018 	}
3019 	if (t_flags & TF_TSO) {
3020 		db_printf("%sTF_TSO", comma ? ", " : "");
3021 		comma = 1;
3022 	}
3023 	if (t_flags & TF_FASTOPEN) {
3024 		db_printf("%sTF_FASTOPEN", comma ? ", " : "");
3025 		comma = 1;
3026 	}
3027 }
3028 
3029 static void
3030 db_print_tflags2(u_int t_flags2)
3031 {
3032 	int comma;
3033 
3034 	comma = 0;
3035 	if (t_flags2 & TF2_PLPMTU_BLACKHOLE) {
3036 		db_printf("%sTF2_PLPMTU_BLACKHOLE", comma ? ", " : "");
3037 		comma = 1;
3038 	}
3039 	if (t_flags2 & TF2_PLPMTU_PMTUD) {
3040 		db_printf("%sTF2_PLPMTU_PMTUD", comma ? ", " : "");
3041 		comma = 1;
3042 	}
3043 	if (t_flags2 & TF2_PLPMTU_MAXSEGSNT) {
3044 		db_printf("%sTF2_PLPMTU_MAXSEGSNT", comma ? ", " : "");
3045 		comma = 1;
3046 	}
3047 	if (t_flags2 & TF2_LOG_AUTO) {
3048 		db_printf("%sTF2_LOG_AUTO", comma ? ", " : "");
3049 		comma = 1;
3050 	}
3051 	if (t_flags2 & TF2_DROP_AF_DATA) {
3052 		db_printf("%sTF2_DROP_AF_DATA", comma ? ", " : "");
3053 		comma = 1;
3054 	}
3055 	if (t_flags2 & TF2_ECN_PERMIT) {
3056 		db_printf("%sTF2_ECN_PERMIT", comma ? ", " : "");
3057 		comma = 1;
3058 	}
3059 	if (t_flags2 & TF2_ECN_SND_CWR) {
3060 		db_printf("%sTF2_ECN_SND_CWR", comma ? ", " : "");
3061 		comma = 1;
3062 	}
3063 	if (t_flags2 & TF2_ECN_SND_ECE) {
3064 		db_printf("%sTF2_ECN_SND_ECE", comma ? ", " : "");
3065 		comma = 1;
3066 	}
3067 	if (t_flags2 & TF2_ACE_PERMIT) {
3068 		db_printf("%sTF2_ACE_PERMIT", comma ? ", " : "");
3069 		comma = 1;
3070 	}
3071 	if (t_flags2 & TF2_FBYTES_COMPLETE) {
3072 		db_printf("%sTF2_FBYTES_COMPLETE", comma ? ", " : "");
3073 		comma = 1;
3074 	}
3075 }
3076 
3077 static void
3078 db_print_toobflags(char t_oobflags)
3079 {
3080 	int comma;
3081 
3082 	comma = 0;
3083 	if (t_oobflags & TCPOOB_HAVEDATA) {
3084 		db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
3085 		comma = 1;
3086 	}
3087 	if (t_oobflags & TCPOOB_HADDATA) {
3088 		db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
3089 		comma = 1;
3090 	}
3091 }
3092 
3093 static void
3094 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
3095 {
3096 
3097 	db_print_indent(indent);
3098 	db_printf("%s at %p\n", name, tp);
3099 
3100 	indent += 2;
3101 
3102 	db_print_indent(indent);
3103 	db_printf("t_segq first: %p   t_segqlen: %d   t_dupacks: %d\n",
3104 	   TAILQ_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
3105 
3106 	db_print_indent(indent);
3107 	db_printf("tt_rexmt: %p   tt_persist: %p   tt_keep: %p\n",
3108 	    &tp->t_timers->tt_rexmt, &tp->t_timers->tt_persist, &tp->t_timers->tt_keep);
3109 
3110 	db_print_indent(indent);
3111 	db_printf("tt_2msl: %p   tt_delack: %p   t_inpcb: %p\n", &tp->t_timers->tt_2msl,
3112 	    &tp->t_timers->tt_delack, tp->t_inpcb);
3113 
3114 	db_print_indent(indent);
3115 	db_printf("t_state: %d (", tp->t_state);
3116 	db_print_tstate(tp->t_state);
3117 	db_printf(")\n");
3118 
3119 	db_print_indent(indent);
3120 	db_printf("t_flags: 0x%x (", tp->t_flags);
3121 	db_print_tflags(tp->t_flags);
3122 	db_printf(")\n");
3123 
3124 	db_print_indent(indent);
3125 	db_printf("t_flags2: 0x%x (", tp->t_flags2);
3126 	db_print_tflags2(tp->t_flags2);
3127 	db_printf(")\n");
3128 
3129 	db_print_indent(indent);
3130 	db_printf("snd_una: 0x%08x   snd_max: 0x%08x   snd_nxt: x0%08x\n",
3131 	    tp->snd_una, tp->snd_max, tp->snd_nxt);
3132 
3133 	db_print_indent(indent);
3134 	db_printf("snd_up: 0x%08x   snd_wl1: 0x%08x   snd_wl2: 0x%08x\n",
3135 	   tp->snd_up, tp->snd_wl1, tp->snd_wl2);
3136 
3137 	db_print_indent(indent);
3138 	db_printf("iss: 0x%08x   irs: 0x%08x   rcv_nxt: 0x%08x\n",
3139 	    tp->iss, tp->irs, tp->rcv_nxt);
3140 
3141 	db_print_indent(indent);
3142 	db_printf("rcv_adv: 0x%08x   rcv_wnd: %u   rcv_up: 0x%08x\n",
3143 	    tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
3144 
3145 	db_print_indent(indent);
3146 	db_printf("snd_wnd: %u   snd_cwnd: %u\n",
3147 	   tp->snd_wnd, tp->snd_cwnd);
3148 
3149 	db_print_indent(indent);
3150 	db_printf("snd_ssthresh: %u   snd_recover: "
3151 	    "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
3152 
3153 	db_print_indent(indent);
3154 	db_printf("t_rcvtime: %u   t_startime: %u\n",
3155 	    tp->t_rcvtime, tp->t_starttime);
3156 
3157 	db_print_indent(indent);
3158 	db_printf("t_rttime: %u   t_rtsq: 0x%08x\n",
3159 	    tp->t_rtttime, tp->t_rtseq);
3160 
3161 	db_print_indent(indent);
3162 	db_printf("t_rxtcur: %d   t_maxseg: %u   t_srtt: %d\n",
3163 	    tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
3164 
3165 	db_print_indent(indent);
3166 	db_printf("t_rttvar: %d   t_rxtshift: %d   t_rttmin: %u   "
3167 	    "t_rttbest: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin,
3168 	    tp->t_rttbest);
3169 
3170 	db_print_indent(indent);
3171 	db_printf("t_rttupdated: %lu   max_sndwnd: %u   t_softerror: %d\n",
3172 	    tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
3173 
3174 	db_print_indent(indent);
3175 	db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
3176 	db_print_toobflags(tp->t_oobflags);
3177 	db_printf(")   t_iobc: 0x%02x\n", tp->t_iobc);
3178 
3179 	db_print_indent(indent);
3180 	db_printf("snd_scale: %u   rcv_scale: %u   request_r_scale: %u\n",
3181 	    tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
3182 
3183 	db_print_indent(indent);
3184 	db_printf("ts_recent: %u   ts_recent_age: %u\n",
3185 	    tp->ts_recent, tp->ts_recent_age);
3186 
3187 	db_print_indent(indent);
3188 	db_printf("ts_offset: %u   last_ack_sent: 0x%08x   snd_cwnd_prev: "
3189 	    "%u\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
3190 
3191 	db_print_indent(indent);
3192 	db_printf("snd_ssthresh_prev: %u   snd_recover_prev: 0x%08x   "
3193 	    "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
3194 	    tp->snd_recover_prev, tp->t_badrxtwin);
3195 
3196 	db_print_indent(indent);
3197 	db_printf("snd_numholes: %d  snd_holes first: %p\n",
3198 	    tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
3199 
3200 	db_print_indent(indent);
3201 	db_printf("snd_fack: 0x%08x   rcv_numsacks: %d\n",
3202 	    tp->snd_fack, tp->rcv_numsacks);
3203 
3204 	/* Skip sackblks, sackhint. */
3205 
3206 	db_print_indent(indent);
3207 	db_printf("t_rttlow: %d   rfbuf_ts: %u   rfbuf_cnt: %d\n",
3208 	    tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
3209 }
3210 
3211 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
3212 {
3213 	struct tcpcb *tp;
3214 
3215 	if (!have_addr) {
3216 		db_printf("usage: show tcpcb <addr>\n");
3217 		return;
3218 	}
3219 	tp = (struct tcpcb *)addr;
3220 
3221 	db_print_tcpcb(tp, "tcpcb", 0);
3222 }
3223 #endif
3224