xref: /freebsd-14.2/sys/netinet/tcp_usrreq.c (revision f18b9b2b)
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 #include "opt_ddb.h"
42 #include "opt_inet.h"
43 #include "opt_inet6.h"
44 #include "opt_ipsec.h"
45 #include "opt_kern_tls.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/arb.h>
50 #include <sys/limits.h>
51 #include <sys/malloc.h>
52 #include <sys/refcount.h>
53 #include <sys/kernel.h>
54 #include <sys/ktls.h>
55 #include <sys/qmath.h>
56 #include <sys/sysctl.h>
57 #include <sys/mbuf.h>
58 #ifdef INET6
59 #include <sys/domain.h>
60 #endif /* INET6 */
61 #include <sys/socket.h>
62 #include <sys/socketvar.h>
63 #include <sys/protosw.h>
64 #include <sys/proc.h>
65 #include <sys/jail.h>
66 #include <sys/stats.h>
67 
68 #ifdef DDB
69 #include <ddb/ddb.h>
70 #endif
71 
72 #include <net/if.h>
73 #include <net/if_var.h>
74 #include <net/route.h>
75 #include <net/vnet.h>
76 
77 #include <netinet/in.h>
78 #include <netinet/in_kdtrace.h>
79 #include <netinet/in_pcb.h>
80 #include <netinet/in_systm.h>
81 #include <netinet/in_var.h>
82 #include <netinet/ip.h>
83 #include <netinet/ip_var.h>
84 #ifdef INET6
85 #include <netinet/ip6.h>
86 #include <netinet6/in6_pcb.h>
87 #include <netinet6/ip6_var.h>
88 #include <netinet6/scope6_var.h>
89 #endif
90 #include <netinet/tcp.h>
91 #include <netinet/tcp_fsm.h>
92 #include <netinet/tcp_seq.h>
93 #include <netinet/tcp_timer.h>
94 #include <netinet/tcp_var.h>
95 #include <netinet/tcp_log_buf.h>
96 #include <netinet/tcpip.h>
97 #include <netinet/cc/cc.h>
98 #include <netinet/tcp_fastopen.h>
99 #include <netinet/tcp_hpts.h>
100 #ifdef TCPPCAP
101 #include <netinet/tcp_pcap.h>
102 #endif
103 #ifdef TCP_OFFLOAD
104 #include <netinet/tcp_offload.h>
105 #endif
106 #include <netipsec/ipsec_support.h>
107 
108 #include <vm/vm.h>
109 #include <vm/vm_param.h>
110 #include <vm/pmap.h>
111 #include <vm/vm_extern.h>
112 #include <vm/vm_map.h>
113 #include <vm/vm_page.h>
114 
115 /*
116  * TCP protocol interface to socket abstraction.
117  */
118 #ifdef INET
119 static int	tcp_connect(struct tcpcb *, struct sockaddr_in *,
120 		    struct thread *td);
121 #endif /* INET */
122 #ifdef INET6
123 static int	tcp6_connect(struct tcpcb *, struct sockaddr_in6 *,
124 		    struct thread *td);
125 #endif /* INET6 */
126 static void	tcp_disconnect(struct tcpcb *);
127 static void	tcp_usrclosed(struct tcpcb *);
128 static void	tcp_fill_info(const struct tcpcb *, struct tcp_info *);
129 
130 static int	tcp_pru_options_support(struct tcpcb *tp, int flags);
131 
132 static void
tcp_bblog_pru(struct tcpcb * tp,uint32_t pru,int error)133 tcp_bblog_pru(struct tcpcb *tp, uint32_t pru, int error)
134 {
135 	struct tcp_log_buffer *lgb;
136 
137 	KASSERT(tp != NULL, ("tcp_bblog_pru: tp == NULL"));
138 	INP_WLOCK_ASSERT(tptoinpcb(tp));
139 	if (tcp_bblogging_on(tp)) {
140 		lgb = tcp_log_event(tp, NULL, NULL, NULL, TCP_LOG_PRU, error,
141 		    0, NULL, false, NULL, NULL, 0, NULL);
142 	} else {
143 		lgb = NULL;
144 	}
145 	if (lgb != NULL) {
146 		if (error >= 0) {
147 			lgb->tlb_errno = (uint32_t)error;
148 		}
149 		lgb->tlb_flex1 = pru;
150 	}
151 }
152 
153 /*
154  * TCP attaches to socket via pru_attach(), reserving space,
155  * and an internet control block.
156  */
157 static int
tcp_usr_attach(struct socket * so,int proto,struct thread * td)158 tcp_usr_attach(struct socket *so, int proto, struct thread *td)
159 {
160 	struct inpcb *inp;
161 	struct tcpcb *tp = NULL;
162 	int error;
163 
164 	inp = sotoinpcb(so);
165 	KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL"));
166 
167 	error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace);
168 	if (error)
169 		goto out;
170 
171 	so->so_rcv.sb_flags |= SB_AUTOSIZE;
172 	so->so_snd.sb_flags |= SB_AUTOSIZE;
173 	error = in_pcballoc(so, &V_tcbinfo);
174 	if (error)
175 		goto out;
176 	inp = sotoinpcb(so);
177 	tp = tcp_newtcpcb(inp, NULL);
178 	if (tp == NULL) {
179 		error = ENOBUFS;
180 		in_pcbfree(inp);
181 		goto out;
182 	}
183 	tp->t_state = TCPS_CLOSED;
184 	tcp_bblog_pru(tp, PRU_ATTACH, error);
185 	INP_WUNLOCK(inp);
186 	TCPSTATES_INC(TCPS_CLOSED);
187 out:
188 	TCP_PROBE2(debug__user, tp, PRU_ATTACH);
189 	return (error);
190 }
191 
192 /*
193  * tcp_usr_detach is called when the socket layer loses its final reference
194  * to the socket, be it a file descriptor reference, a reference from TCP,
195  * etc.  At this point, there is only one case in which we will keep around
196  * inpcb state: time wait.
197  */
198 static void
tcp_usr_detach(struct socket * so)199 tcp_usr_detach(struct socket *so)
200 {
201 	struct inpcb *inp;
202 	struct tcpcb *tp;
203 
204 	inp = sotoinpcb(so);
205 	KASSERT(inp != NULL, ("%s: inp == NULL", __func__));
206 	INP_WLOCK(inp);
207 	KASSERT(so->so_pcb == inp && inp->inp_socket == so,
208 		("%s: socket %p inp %p mismatch", __func__, so, inp));
209 
210 	tp = intotcpcb(inp);
211 
212 	KASSERT(inp->inp_flags & INP_DROPPED ||
213 	    tp->t_state < TCPS_SYN_SENT,
214 	    ("%s: inp %p not dropped or embryonic", __func__, inp));
215 
216 	tcp_discardcb(tp);
217 	in_pcbfree(inp);
218 }
219 
220 #ifdef INET
221 /*
222  * Give the socket an address.
223  */
224 static int
tcp_usr_bind(struct socket * so,struct sockaddr * nam,struct thread * td)225 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
226 {
227 	int error = 0;
228 	struct inpcb *inp;
229 	struct tcpcb *tp;
230 	struct sockaddr_in *sinp;
231 
232 	inp = sotoinpcb(so);
233 	KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL"));
234 	INP_WLOCK(inp);
235 	if (inp->inp_flags & INP_DROPPED) {
236 		INP_WUNLOCK(inp);
237 		return (EINVAL);
238 	}
239 	tp = intotcpcb(inp);
240 
241 	sinp = (struct sockaddr_in *)nam;
242 	if (nam->sa_family != AF_INET) {
243 		/*
244 		 * Preserve compatibility with old programs.
245 		 */
246 		if (nam->sa_family != AF_UNSPEC ||
247 		    nam->sa_len < offsetof(struct sockaddr_in, sin_zero) ||
248 		    sinp->sin_addr.s_addr != INADDR_ANY) {
249 			error = EAFNOSUPPORT;
250 			goto out;
251 		}
252 		nam->sa_family = AF_INET;
253 	}
254 	if (nam->sa_len != sizeof(*sinp)) {
255 		error = EINVAL;
256 		goto out;
257 	}
258 	/*
259 	 * Must check for multicast addresses and disallow binding
260 	 * to them.
261 	 */
262 	if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
263 		error = EAFNOSUPPORT;
264 		goto out;
265 	}
266 	INP_HASH_WLOCK(&V_tcbinfo);
267 	error = in_pcbbind(inp, sinp, td->td_ucred);
268 	INP_HASH_WUNLOCK(&V_tcbinfo);
269 out:
270 	tcp_bblog_pru(tp, PRU_BIND, error);
271 	TCP_PROBE2(debug__user, tp, PRU_BIND);
272 	INP_WUNLOCK(inp);
273 
274 	return (error);
275 }
276 #endif /* INET */
277 
278 #ifdef INET6
279 static int
tcp6_usr_bind(struct socket * so,struct sockaddr * nam,struct thread * td)280 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
281 {
282 	int error = 0;
283 	struct inpcb *inp;
284 	struct tcpcb *tp;
285 	struct sockaddr_in6 *sin6;
286 	u_char vflagsav;
287 
288 	inp = sotoinpcb(so);
289 	KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL"));
290 	INP_WLOCK(inp);
291 	if (inp->inp_flags & INP_DROPPED) {
292 		INP_WUNLOCK(inp);
293 		return (EINVAL);
294 	}
295 	tp = intotcpcb(inp);
296 
297 	vflagsav = inp->inp_vflag;
298 
299 	sin6 = (struct sockaddr_in6 *)nam;
300 	if (nam->sa_family != AF_INET6) {
301 		error = EAFNOSUPPORT;
302 		goto out;
303 	}
304 	if (nam->sa_len != sizeof(*sin6)) {
305 		error = EINVAL;
306 		goto out;
307 	}
308 	/*
309 	 * Must check for multicast addresses and disallow binding
310 	 * to them.
311 	 */
312 	if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
313 		error = EAFNOSUPPORT;
314 		goto out;
315 	}
316 
317 	INP_HASH_WLOCK(&V_tcbinfo);
318 	inp->inp_vflag &= ~INP_IPV4;
319 	inp->inp_vflag |= INP_IPV6;
320 #ifdef INET
321 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
322 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
323 			inp->inp_vflag |= INP_IPV4;
324 		else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
325 			struct sockaddr_in sin;
326 
327 			in6_sin6_2_sin(&sin, sin6);
328 			if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
329 				error = EAFNOSUPPORT;
330 				INP_HASH_WUNLOCK(&V_tcbinfo);
331 				goto out;
332 			}
333 			inp->inp_vflag |= INP_IPV4;
334 			inp->inp_vflag &= ~INP_IPV6;
335 			error = in_pcbbind(inp, &sin, td->td_ucred);
336 			INP_HASH_WUNLOCK(&V_tcbinfo);
337 			goto out;
338 		}
339 	}
340 #endif
341 	error = in6_pcbbind(inp, sin6, td->td_ucred);
342 	INP_HASH_WUNLOCK(&V_tcbinfo);
343 out:
344 	if (error != 0)
345 		inp->inp_vflag = vflagsav;
346 	tcp_bblog_pru(tp, PRU_BIND, error);
347 	TCP_PROBE2(debug__user, tp, PRU_BIND);
348 	INP_WUNLOCK(inp);
349 	return (error);
350 }
351 #endif /* INET6 */
352 
353 #ifdef INET
354 /*
355  * Prepare to accept connections.
356  */
357 static int
tcp_usr_listen(struct socket * so,int backlog,struct thread * td)358 tcp_usr_listen(struct socket *so, int backlog, struct thread *td)
359 {
360 	int error = 0;
361 	struct inpcb *inp;
362 	struct tcpcb *tp;
363 
364 	inp = sotoinpcb(so);
365 	KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL"));
366 	INP_WLOCK(inp);
367 	if (inp->inp_flags & INP_DROPPED) {
368 		INP_WUNLOCK(inp);
369 		return (EINVAL);
370 	}
371 	tp = intotcpcb(inp);
372 
373 	SOCK_LOCK(so);
374 	error = solisten_proto_check(so);
375 	if (error != 0) {
376 		SOCK_UNLOCK(so);
377 		goto out;
378 	}
379 	if (inp->inp_lport == 0) {
380 		INP_HASH_WLOCK(&V_tcbinfo);
381 		error = in_pcbbind(inp, NULL, td->td_ucred);
382 		INP_HASH_WUNLOCK(&V_tcbinfo);
383 	}
384 	if (error == 0) {
385 		tcp_state_change(tp, TCPS_LISTEN);
386 		solisten_proto(so, backlog);
387 #ifdef TCP_OFFLOAD
388 		if ((so->so_options & SO_NO_OFFLOAD) == 0)
389 			tcp_offload_listen_start(tp);
390 #endif
391 	} else {
392 		solisten_proto_abort(so);
393 	}
394 	SOCK_UNLOCK(so);
395 
396 	if (IS_FASTOPEN(tp->t_flags))
397 		tp->t_tfo_pending = tcp_fastopen_alloc_counter();
398 
399 out:
400 	tcp_bblog_pru(tp, PRU_LISTEN, error);
401 	TCP_PROBE2(debug__user, tp, PRU_LISTEN);
402 	INP_WUNLOCK(inp);
403 	return (error);
404 }
405 #endif /* INET */
406 
407 #ifdef INET6
408 static int
tcp6_usr_listen(struct socket * so,int backlog,struct thread * td)409 tcp6_usr_listen(struct socket *so, int backlog, struct thread *td)
410 {
411 	int error = 0;
412 	struct inpcb *inp;
413 	struct tcpcb *tp;
414 	u_char vflagsav;
415 
416 	inp = sotoinpcb(so);
417 	KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL"));
418 	INP_WLOCK(inp);
419 	if (inp->inp_flags & INP_DROPPED) {
420 		INP_WUNLOCK(inp);
421 		return (EINVAL);
422 	}
423 	tp = intotcpcb(inp);
424 
425 	vflagsav = inp->inp_vflag;
426 
427 	SOCK_LOCK(so);
428 	error = solisten_proto_check(so);
429 	if (error != 0) {
430 		SOCK_UNLOCK(so);
431 		goto out;
432 	}
433 	INP_HASH_WLOCK(&V_tcbinfo);
434 	if (inp->inp_lport == 0) {
435 		inp->inp_vflag &= ~INP_IPV4;
436 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
437 			inp->inp_vflag |= INP_IPV4;
438 		error = in6_pcbbind(inp, NULL, td->td_ucred);
439 	}
440 	INP_HASH_WUNLOCK(&V_tcbinfo);
441 	if (error == 0) {
442 		tcp_state_change(tp, TCPS_LISTEN);
443 		solisten_proto(so, backlog);
444 #ifdef TCP_OFFLOAD
445 		if ((so->so_options & SO_NO_OFFLOAD) == 0)
446 			tcp_offload_listen_start(tp);
447 #endif
448 	} else {
449 		solisten_proto_abort(so);
450 	}
451 	SOCK_UNLOCK(so);
452 
453 	if (IS_FASTOPEN(tp->t_flags))
454 		tp->t_tfo_pending = tcp_fastopen_alloc_counter();
455 
456 	if (error != 0)
457 		inp->inp_vflag = vflagsav;
458 
459 out:
460 	tcp_bblog_pru(tp, PRU_LISTEN, error);
461 	TCP_PROBE2(debug__user, tp, PRU_LISTEN);
462 	INP_WUNLOCK(inp);
463 	return (error);
464 }
465 #endif /* INET6 */
466 
467 #ifdef INET
468 /*
469  * Initiate connection to peer.
470  * Create a template for use in transmissions on this connection.
471  * Enter SYN_SENT state, and mark socket as connecting.
472  * Start keep-alive timer, and seed output sequence space.
473  * Send initial segment on connection.
474  */
475 static int
tcp_usr_connect(struct socket * so,struct sockaddr * nam,struct thread * td)476 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
477 {
478 	struct epoch_tracker et;
479 	int error = 0;
480 	struct inpcb *inp;
481 	struct tcpcb *tp;
482 	struct sockaddr_in *sinp;
483 
484 	inp = sotoinpcb(so);
485 	KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL"));
486 	INP_WLOCK(inp);
487 	if (inp->inp_flags & INP_DROPPED) {
488 		INP_WUNLOCK(inp);
489 		return (ECONNREFUSED);
490 	}
491 	tp = intotcpcb(inp);
492 
493 	sinp = (struct sockaddr_in *)nam;
494 	if (nam->sa_family != AF_INET) {
495 		error = EAFNOSUPPORT;
496 		goto out;
497 	}
498 	if (nam->sa_len != sizeof (*sinp)) {
499 		error = EINVAL;
500 		goto out;
501 	}
502 	/*
503 	 * Must disallow TCP ``connections'' to multicast addresses.
504 	 */
505 	if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
506 		error = EAFNOSUPPORT;
507 		goto out;
508 	}
509 	if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) {
510 		error = EACCES;
511 		goto out;
512 	}
513 	if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0)
514 		goto out;
515 	if (SOLISTENING(so)) {
516 		error = EOPNOTSUPP;
517 		goto out;
518 	}
519 	NET_EPOCH_ENTER(et);
520 	if ((error = tcp_connect(tp, sinp, td)) != 0)
521 		goto out_in_epoch;
522 #ifdef TCP_OFFLOAD
523 	if (registered_toedevs > 0 &&
524 	    (so->so_options & SO_NO_OFFLOAD) == 0 &&
525 	    (error = tcp_offload_connect(so, nam)) == 0)
526 		goto out_in_epoch;
527 #endif
528 	tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
529 	error = tcp_output(tp);
530 	KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()"
531 	    ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error));
532 out_in_epoch:
533 	NET_EPOCH_EXIT(et);
534 out:
535 	tcp_bblog_pru(tp, PRU_CONNECT, error);
536 	TCP_PROBE2(debug__user, tp, PRU_CONNECT);
537 	INP_WUNLOCK(inp);
538 	return (error);
539 }
540 #endif /* INET */
541 
542 #ifdef INET6
543 static int
tcp6_usr_connect(struct socket * so,struct sockaddr * nam,struct thread * td)544 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
545 {
546 	struct epoch_tracker et;
547 	int error = 0;
548 	struct inpcb *inp;
549 	struct tcpcb *tp;
550 	struct sockaddr_in6 *sin6;
551 	u_int8_t incflagsav;
552 	u_char vflagsav;
553 
554 	inp = sotoinpcb(so);
555 	KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL"));
556 	INP_WLOCK(inp);
557 	if (inp->inp_flags & INP_DROPPED) {
558 		INP_WUNLOCK(inp);
559 		return (ECONNREFUSED);
560 	}
561 	tp = intotcpcb(inp);
562 
563 	vflagsav = inp->inp_vflag;
564 	incflagsav = inp->inp_inc.inc_flags;
565 
566 	sin6 = (struct sockaddr_in6 *)nam;
567 	if (nam->sa_family != AF_INET6) {
568 		error = EAFNOSUPPORT;
569 		goto out;
570 	}
571 	if (nam->sa_len != sizeof (*sin6)) {
572 		error = EINVAL;
573 		goto out;
574 	}
575 	/*
576 	 * Must disallow TCP ``connections'' to multicast addresses.
577 	 */
578 	if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
579 		error = EAFNOSUPPORT;
580 		goto out;
581 	}
582 	if (SOLISTENING(so)) {
583 		error = EINVAL;
584 		goto out;
585 	}
586 #ifdef INET
587 	/*
588 	 * XXXRW: Some confusion: V4/V6 flags relate to binding, and
589 	 * therefore probably require the hash lock, which isn't held here.
590 	 * Is this a significant problem?
591 	 */
592 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
593 		struct sockaddr_in sin;
594 
595 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
596 			error = EINVAL;
597 			goto out;
598 		}
599 		if ((inp->inp_vflag & INP_IPV4) == 0) {
600 			error = EAFNOSUPPORT;
601 			goto out;
602 		}
603 
604 		in6_sin6_2_sin(&sin, sin6);
605 		if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
606 			error = EAFNOSUPPORT;
607 			goto out;
608 		}
609 		if (ntohl(sin.sin_addr.s_addr) == INADDR_BROADCAST) {
610 			error = EACCES;
611 			goto out;
612 		}
613 		if ((error = prison_remote_ip4(td->td_ucred,
614 		    &sin.sin_addr)) != 0)
615 			goto out;
616 		inp->inp_vflag |= INP_IPV4;
617 		inp->inp_vflag &= ~INP_IPV6;
618 		NET_EPOCH_ENTER(et);
619 		if ((error = tcp_connect(tp, &sin, td)) != 0)
620 			goto out_in_epoch;
621 #ifdef TCP_OFFLOAD
622 		if (registered_toedevs > 0 &&
623 		    (so->so_options & SO_NO_OFFLOAD) == 0 &&
624 		    (error = tcp_offload_connect(so, nam)) == 0)
625 			goto out_in_epoch;
626 #endif
627 		error = tcp_output(tp);
628 		goto out_in_epoch;
629 	} else {
630 		if ((inp->inp_vflag & INP_IPV6) == 0) {
631 			error = EAFNOSUPPORT;
632 			goto out;
633 		}
634 	}
635 #endif
636 	if ((error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr)) != 0)
637 		goto out;
638 	inp->inp_vflag &= ~INP_IPV4;
639 	inp->inp_vflag |= INP_IPV6;
640 	inp->inp_inc.inc_flags |= INC_ISIPV6;
641 	NET_EPOCH_ENTER(et);
642 	if ((error = tcp6_connect(tp, sin6, td)) != 0)
643 		goto out_in_epoch;
644 #ifdef TCP_OFFLOAD
645 	if (registered_toedevs > 0 &&
646 	    (so->so_options & SO_NO_OFFLOAD) == 0 &&
647 	    (error = tcp_offload_connect(so, nam)) == 0)
648 		goto out_in_epoch;
649 #endif
650 	tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp));
651 	error = tcp_output(tp);
652 out_in_epoch:
653 	NET_EPOCH_EXIT(et);
654 out:
655 	KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()"
656 	    ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error));
657 	/*
658 	 * If the implicit bind in the connect call fails, restore
659 	 * the flags we modified.
660 	 */
661 	if (error != 0 && inp->inp_lport == 0) {
662 		inp->inp_vflag = vflagsav;
663 		inp->inp_inc.inc_flags = incflagsav;
664 	}
665 
666 	tcp_bblog_pru(tp, PRU_CONNECT, error);
667 	TCP_PROBE2(debug__user, tp, PRU_CONNECT);
668 	INP_WUNLOCK(inp);
669 	return (error);
670 }
671 #endif /* INET6 */
672 
673 /*
674  * Initiate disconnect from peer.
675  * If connection never passed embryonic stage, just drop;
676  * else if don't need to let data drain, then can just drop anyways,
677  * else have to begin TCP shutdown process: mark socket disconnecting,
678  * drain unread data, state switch to reflect user close, and
679  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
680  * when peer sends FIN and acks ours.
681  *
682  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
683  */
684 static int
tcp_usr_disconnect(struct socket * so)685 tcp_usr_disconnect(struct socket *so)
686 {
687 	struct inpcb *inp;
688 	struct tcpcb *tp = NULL;
689 	struct epoch_tracker et;
690 	int error = 0;
691 
692 	NET_EPOCH_ENTER(et);
693 	inp = sotoinpcb(so);
694 	KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL"));
695 	INP_WLOCK(inp);
696 	if (inp->inp_flags & INP_DROPPED) {
697 		INP_WUNLOCK(inp);
698 		NET_EPOCH_EXIT(et);
699 		return (ECONNRESET);
700 	}
701 	tp = intotcpcb(inp);
702 
703 	if (tp->t_state == TCPS_TIME_WAIT)
704 		goto out;
705 	tcp_disconnect(tp);
706 out:
707 	tcp_bblog_pru(tp, PRU_DISCONNECT, error);
708 	TCP_PROBE2(debug__user, tp, PRU_DISCONNECT);
709 	INP_WUNLOCK(inp);
710 	NET_EPOCH_EXIT(et);
711 	return (error);
712 }
713 
714 #ifdef INET
715 /*
716  * Accept a connection.  Essentially all the work is done at higher levels;
717  * just return the address of the peer, storing through addr.
718  */
719 static int
tcp_usr_accept(struct socket * so,struct sockaddr ** nam)720 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
721 {
722 	int error = 0;
723 	struct inpcb *inp;
724 	struct tcpcb *tp;
725 	struct in_addr addr;
726 	in_port_t port = 0;
727 
728 	inp = sotoinpcb(so);
729 	KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL"));
730 	INP_WLOCK(inp);
731 	if (inp->inp_flags & INP_DROPPED) {
732 		INP_WUNLOCK(inp);
733 		return (ECONNABORTED);
734 	}
735 	tp = intotcpcb(inp);
736 
737 	if (so->so_state & SS_ISDISCONNECTED) {
738 		error = ECONNABORTED;
739 		goto out;
740 	}
741 	/*
742 	 * We inline in_getpeeraddr and COMMON_END here, so that we can
743 	 * copy the data of interest and defer the malloc until after we
744 	 * release the lock.
745 	 */
746 	port = inp->inp_fport;
747 	addr = inp->inp_faddr;
748 
749 out:
750 	tcp_bblog_pru(tp, PRU_ACCEPT, error);
751 	TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
752 	INP_WUNLOCK(inp);
753 	if (error == 0)
754 		*nam = in_sockaddr(port, &addr);
755 	return error;
756 }
757 #endif /* INET */
758 
759 #ifdef INET6
760 static int
tcp6_usr_accept(struct socket * so,struct sockaddr ** nam)761 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
762 {
763 	struct inpcb *inp;
764 	int error = 0;
765 	struct tcpcb *tp;
766 	struct in_addr addr;
767 	struct in6_addr addr6;
768 	struct epoch_tracker et;
769 	in_port_t port = 0;
770 	int v4 = 0;
771 
772 	inp = sotoinpcb(so);
773 	KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL"));
774 	NET_EPOCH_ENTER(et); /* XXXMT Why is this needed? */
775 	INP_WLOCK(inp);
776 	if (inp->inp_flags & INP_DROPPED) {
777 		INP_WUNLOCK(inp);
778 		NET_EPOCH_EXIT(et);
779 		return (ECONNABORTED);
780 	}
781 	tp = intotcpcb(inp);
782 
783 	if (so->so_state & SS_ISDISCONNECTED) {
784 		error = ECONNABORTED;
785 		goto out;
786 	}
787 	/*
788 	 * We inline in6_mapped_peeraddr and COMMON_END here, so that we can
789 	 * copy the data of interest and defer the malloc until after we
790 	 * release the lock.
791 	 */
792 	if (inp->inp_vflag & INP_IPV4) {
793 		v4 = 1;
794 		port = inp->inp_fport;
795 		addr = inp->inp_faddr;
796 	} else {
797 		port = inp->inp_fport;
798 		addr6 = inp->in6p_faddr;
799 	}
800 
801 out:
802 	tcp_bblog_pru(tp, PRU_ACCEPT, error);
803 	TCP_PROBE2(debug__user, tp, PRU_ACCEPT);
804 	INP_WUNLOCK(inp);
805 	NET_EPOCH_EXIT(et);
806 	if (error == 0) {
807 		if (v4)
808 			*nam = in6_v4mapsin6_sockaddr(port, &addr);
809 		else
810 			*nam = in6_sockaddr(port, &addr6);
811 	}
812 	return error;
813 }
814 #endif /* INET6 */
815 
816 /*
817  * Mark the connection as being incapable of further output.
818  */
819 static int
tcp_usr_shutdown(struct socket * so)820 tcp_usr_shutdown(struct socket *so)
821 {
822 	int error = 0;
823 	struct inpcb *inp;
824 	struct tcpcb *tp;
825 	struct epoch_tracker et;
826 
827 	inp = sotoinpcb(so);
828 	KASSERT(inp != NULL, ("inp == NULL"));
829 	INP_WLOCK(inp);
830 	if (inp->inp_flags & INP_DROPPED) {
831 		INP_WUNLOCK(inp);
832 		return (ECONNRESET);
833 	}
834 	tp = intotcpcb(inp);
835 
836 	NET_EPOCH_ENTER(et);
837 	socantsendmore(so);
838 	tcp_usrclosed(tp);
839 	if (!(inp->inp_flags & INP_DROPPED))
840 		error = tcp_output_nodrop(tp);
841 	tcp_bblog_pru(tp, PRU_SHUTDOWN, error);
842 	TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN);
843 	error = tcp_unlock_or_drop(tp, error);
844 	NET_EPOCH_EXIT(et);
845 
846 	return (error);
847 }
848 
849 /*
850  * After a receive, possibly send window update to peer.
851  */
852 static int
tcp_usr_rcvd(struct socket * so,int flags)853 tcp_usr_rcvd(struct socket *so, int flags)
854 {
855 	struct epoch_tracker et;
856 	struct inpcb *inp;
857 	struct tcpcb *tp;
858 	int outrv = 0, error = 0;
859 
860 	inp = sotoinpcb(so);
861 	KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL"));
862 	INP_WLOCK(inp);
863 	if (inp->inp_flags & INP_DROPPED) {
864 		INP_WUNLOCK(inp);
865 		return (ECONNRESET);
866 	}
867 	tp = intotcpcb(inp);
868 
869 	NET_EPOCH_ENTER(et);
870 	/*
871 	 * For passively-created TFO connections, don't attempt a window
872 	 * update while still in SYN_RECEIVED as this may trigger an early
873 	 * SYN|ACK.  It is preferable to have the SYN|ACK be sent along with
874 	 * application response data, or failing that, when the DELACK timer
875 	 * expires.
876 	 */
877 	if (IS_FASTOPEN(tp->t_flags) &&
878 	    (tp->t_state == TCPS_SYN_RECEIVED))
879 		goto out;
880 #ifdef TCP_OFFLOAD
881 	if (tp->t_flags & TF_TOE)
882 		tcp_offload_rcvd(tp);
883 	else
884 #endif
885 		outrv = tcp_output_nodrop(tp);
886 out:
887 	tcp_bblog_pru(tp, PRU_RCVD, error);
888 	TCP_PROBE2(debug__user, tp, PRU_RCVD);
889 	(void) tcp_unlock_or_drop(tp, outrv);
890 	NET_EPOCH_EXIT(et);
891 	return (error);
892 }
893 
894 /*
895  * Do a send by putting data in output queue and updating urgent
896  * marker if URG set.  Possibly send more data.  Unlike the other
897  * pru_*() routines, the mbuf chains are our responsibility.  We
898  * must either enqueue them or free them.  The other pru_* routines
899  * generally are caller-frees.
900  */
901 static int
tcp_usr_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct thread * td)902 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
903     struct sockaddr *nam, struct mbuf *control, struct thread *td)
904 {
905 	struct epoch_tracker et;
906 	int error = 0;
907 	struct inpcb *inp;
908 	struct tcpcb *tp;
909 #ifdef INET
910 #ifdef INET6
911 	struct sockaddr_in sin;
912 #endif
913 	struct sockaddr_in *sinp;
914 #endif
915 #ifdef INET6
916 	struct sockaddr_in6 *sin6;
917 	int isipv6;
918 #endif
919 	u_int8_t incflagsav;
920 	u_char vflagsav;
921 	bool restoreflags;
922 
923 	inp = sotoinpcb(so);
924 	KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
925 	INP_WLOCK(inp);
926 	if (inp->inp_flags & INP_DROPPED) {
927 		if (m != NULL && (flags & PRUS_NOTREADY) == 0)
928 			m_freem(m);
929 		INP_WUNLOCK(inp);
930 		return (ECONNRESET);
931 	}
932 	tp = intotcpcb(inp);
933 
934 	vflagsav = inp->inp_vflag;
935 	incflagsav = inp->inp_inc.inc_flags;
936 	restoreflags = false;
937 
938 	NET_EPOCH_ENTER(et);
939 	if (control != NULL) {
940 		/* TCP doesn't do control messages (rights, creds, etc) */
941 		if (control->m_len > 0) {
942 			m_freem(control);
943 			error = EINVAL;
944 			goto out;
945 		}
946 		m_freem(control);	/* empty control, just free it */
947 	}
948 
949 	if ((flags & PRUS_OOB) != 0 &&
950 	    (error = tcp_pru_options_support(tp, PRUS_OOB)) != 0)
951 		goto out;
952 
953 	if (nam != NULL && tp->t_state < TCPS_SYN_SENT) {
954 		if (tp->t_state == TCPS_LISTEN) {
955 			error = EINVAL;
956 			goto out;
957 		}
958 		switch (nam->sa_family) {
959 #ifdef INET
960 		case AF_INET:
961 			sinp = (struct sockaddr_in *)nam;
962 			if (sinp->sin_len != sizeof(struct sockaddr_in)) {
963 				error = EINVAL;
964 				goto out;
965 			}
966 			if ((inp->inp_vflag & INP_IPV6) != 0) {
967 				error = EAFNOSUPPORT;
968 				goto out;
969 			}
970 			if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
971 				error = EAFNOSUPPORT;
972 				goto out;
973 			}
974 			if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) {
975 				error = EACCES;
976 				goto out;
977 			}
978 			if ((error = prison_remote_ip4(td->td_ucred,
979 			    &sinp->sin_addr)))
980 				goto out;
981 #ifdef INET6
982 			isipv6 = 0;
983 #endif
984 			break;
985 #endif /* INET */
986 #ifdef INET6
987 		case AF_INET6:
988 			sin6 = (struct sockaddr_in6 *)nam;
989 			if (sin6->sin6_len != sizeof(*sin6)) {
990 				error = EINVAL;
991 				goto out;
992 			}
993 			if ((inp->inp_vflag & INP_IPV6PROTO) == 0) {
994 				error = EAFNOSUPPORT;
995 				goto out;
996 			}
997 			if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
998 				error = EAFNOSUPPORT;
999 				goto out;
1000 			}
1001 			if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
1002 #ifdef INET
1003 				if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
1004 					error = EINVAL;
1005 					goto out;
1006 				}
1007 				if ((inp->inp_vflag & INP_IPV4) == 0) {
1008 					error = EAFNOSUPPORT;
1009 					goto out;
1010 				}
1011 				restoreflags = true;
1012 				inp->inp_vflag &= ~INP_IPV6;
1013 				sinp = &sin;
1014 				in6_sin6_2_sin(sinp, sin6);
1015 				if (IN_MULTICAST(
1016 				    ntohl(sinp->sin_addr.s_addr))) {
1017 					error = EAFNOSUPPORT;
1018 					goto out;
1019 				}
1020 				if ((error = prison_remote_ip4(td->td_ucred,
1021 				    &sinp->sin_addr)))
1022 					goto out;
1023 				isipv6 = 0;
1024 #else /* !INET */
1025 				error = EAFNOSUPPORT;
1026 				goto out;
1027 #endif /* INET */
1028 			} else {
1029 				if ((inp->inp_vflag & INP_IPV6) == 0) {
1030 					error = EAFNOSUPPORT;
1031 					goto out;
1032 				}
1033 				restoreflags = true;
1034 				inp->inp_vflag &= ~INP_IPV4;
1035 				inp->inp_inc.inc_flags |= INC_ISIPV6;
1036 				if ((error = prison_remote_ip6(td->td_ucred,
1037 				    &sin6->sin6_addr)))
1038 					goto out;
1039 				isipv6 = 1;
1040 			}
1041 			break;
1042 #endif /* INET6 */
1043 		default:
1044 			error = EAFNOSUPPORT;
1045 			goto out;
1046 		}
1047 	}
1048 	if (!(flags & PRUS_OOB)) {
1049 		if (tp->t_acktime == 0)
1050 			tp->t_acktime = ticks;
1051 		sbappendstream(&so->so_snd, m, flags);
1052 		m = NULL;
1053 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1054 			KASSERT(tp->t_state == TCPS_CLOSED,
1055 			    ("%s: tp %p is listening", __func__, tp));
1056 
1057 			/*
1058 			 * Do implied connect if not yet connected,
1059 			 * initialize window to default value, and
1060 			 * initialize maxseg using peer's cached MSS.
1061 			 */
1062 #ifdef INET6
1063 			if (isipv6)
1064 				error = tcp6_connect(tp, sin6, td);
1065 #endif /* INET6 */
1066 #if defined(INET6) && defined(INET)
1067 			else
1068 #endif
1069 #ifdef INET
1070 				error = tcp_connect(tp, sinp, td);
1071 #endif
1072 			/*
1073 			 * The bind operation in tcp_connect succeeded. We
1074 			 * no longer want to restore the flags if later
1075 			 * operations fail.
1076 			 */
1077 			if (error == 0 || inp->inp_lport != 0)
1078 				restoreflags = false;
1079 
1080 			if (error) {
1081 				/* m is freed if PRUS_NOTREADY is unset. */
1082 				sbflush(&so->so_snd);
1083 				goto out;
1084 			}
1085 			if (IS_FASTOPEN(tp->t_flags))
1086 				tcp_fastopen_connect(tp);
1087 			else {
1088 				tp->snd_wnd = TTCP_CLIENT_SND_WND;
1089 				tcp_mss(tp, -1);
1090 			}
1091 		}
1092 		if (flags & PRUS_EOF) {
1093 			/*
1094 			 * Close the send side of the connection after
1095 			 * the data is sent.
1096 			 */
1097 			socantsendmore(so);
1098 			tcp_usrclosed(tp);
1099 		}
1100 		if (TCPS_HAVEESTABLISHED(tp->t_state) &&
1101 		    ((tp->t_flags2 & TF2_FBYTES_COMPLETE) == 0) &&
1102 		    (tp->t_fbyte_out == 0) &&
1103 		    (so->so_snd.sb_ccc > 0)) {
1104 			tp->t_fbyte_out = ticks;
1105 			if (tp->t_fbyte_out == 0)
1106 				tp->t_fbyte_out = 1;
1107 			if (tp->t_fbyte_out && tp->t_fbyte_in)
1108 				tp->t_flags2 |= TF2_FBYTES_COMPLETE;
1109 		}
1110 		if (!(inp->inp_flags & INP_DROPPED) &&
1111 		    !(flags & PRUS_NOTREADY)) {
1112 			if (flags & PRUS_MORETOCOME)
1113 				tp->t_flags |= TF_MORETOCOME;
1114 			error = tcp_output_nodrop(tp);
1115 			if (flags & PRUS_MORETOCOME)
1116 				tp->t_flags &= ~TF_MORETOCOME;
1117 		}
1118 	} else {
1119 		/*
1120 		 * XXXRW: PRUS_EOF not implemented with PRUS_OOB?
1121 		 */
1122 		SOCKBUF_LOCK(&so->so_snd);
1123 		if (sbspace(&so->so_snd) < -512) {
1124 			SOCKBUF_UNLOCK(&so->so_snd);
1125 			error = ENOBUFS;
1126 			goto out;
1127 		}
1128 		/*
1129 		 * According to RFC961 (Assigned Protocols),
1130 		 * the urgent pointer points to the last octet
1131 		 * of urgent data.  We continue, however,
1132 		 * to consider it to indicate the first octet
1133 		 * of data past the urgent section.
1134 		 * Otherwise, snd_up should be one lower.
1135 		 */
1136 		if (tp->t_acktime == 0)
1137 			tp->t_acktime = ticks;
1138 		sbappendstream_locked(&so->so_snd, m, flags);
1139 		SOCKBUF_UNLOCK(&so->so_snd);
1140 		m = NULL;
1141 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1142 			/*
1143 			 * Do implied connect if not yet connected,
1144 			 * initialize window to default value, and
1145 			 * initialize maxseg using peer's cached MSS.
1146 			 */
1147 
1148 			/*
1149 			 * Not going to contemplate SYN|URG
1150 			 */
1151 			if (IS_FASTOPEN(tp->t_flags))
1152 				tp->t_flags &= ~TF_FASTOPEN;
1153 #ifdef INET6
1154 			if (isipv6)
1155 				error = tcp6_connect(tp, sin6, td);
1156 #endif /* INET6 */
1157 #if defined(INET6) && defined(INET)
1158 			else
1159 #endif
1160 #ifdef INET
1161 				error = tcp_connect(tp, sinp, td);
1162 #endif
1163 			/*
1164 			 * The bind operation in tcp_connect succeeded. We
1165 			 * no longer want to restore the flags if later
1166 			 * operations fail.
1167 			 */
1168 			if (error == 0 || inp->inp_lport != 0)
1169 				restoreflags = false;
1170 
1171 			if (error != 0) {
1172 				/* m is freed if PRUS_NOTREADY is unset. */
1173 				sbflush(&so->so_snd);
1174 				goto out;
1175 			}
1176 			tp->snd_wnd = TTCP_CLIENT_SND_WND;
1177 			tcp_mss(tp, -1);
1178 		}
1179 		tp->snd_up = tp->snd_una + sbavail(&so->so_snd);
1180 		if ((flags & PRUS_NOTREADY) == 0) {
1181 			tp->t_flags |= TF_FORCEDATA;
1182 			error = tcp_output_nodrop(tp);
1183 			tp->t_flags &= ~TF_FORCEDATA;
1184 		}
1185 	}
1186 	TCP_LOG_EVENT(tp, NULL,
1187 	    &inp->inp_socket->so_rcv,
1188 	    &inp->inp_socket->so_snd,
1189 	    TCP_LOG_USERSEND, error,
1190 	    0, NULL, false);
1191 
1192 out:
1193 	/*
1194 	 * In case of PRUS_NOTREADY, the caller or tcp_usr_ready() is
1195 	 * responsible for freeing memory.
1196 	 */
1197 	if (m != NULL && (flags & PRUS_NOTREADY) == 0)
1198 		m_freem(m);
1199 
1200 	/*
1201 	 * If the request was unsuccessful and we changed flags,
1202 	 * restore the original flags.
1203 	 */
1204 	if (error != 0 && restoreflags) {
1205 		inp->inp_vflag = vflagsav;
1206 		inp->inp_inc.inc_flags = incflagsav;
1207 	}
1208 	tcp_bblog_pru(tp, (flags & PRUS_OOB) ? PRU_SENDOOB :
1209 		      ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND), error);
1210 	TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB :
1211 		   ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1212 	error = tcp_unlock_or_drop(tp, error);
1213 	NET_EPOCH_EXIT(et);
1214 	return (error);
1215 }
1216 
1217 static int
tcp_usr_ready(struct socket * so,struct mbuf * m,int count)1218 tcp_usr_ready(struct socket *so, struct mbuf *m, int count)
1219 {
1220 	struct epoch_tracker et;
1221 	struct inpcb *inp;
1222 	struct tcpcb *tp;
1223 	int error;
1224 
1225 	inp = sotoinpcb(so);
1226 	INP_WLOCK(inp);
1227 	if (inp->inp_flags & INP_DROPPED) {
1228 		INP_WUNLOCK(inp);
1229 		mb_free_notready(m, count);
1230 		return (ECONNRESET);
1231 	}
1232 	tp = intotcpcb(inp);
1233 
1234 	SOCKBUF_LOCK(&so->so_snd);
1235 	error = sbready(&so->so_snd, m, count);
1236 	SOCKBUF_UNLOCK(&so->so_snd);
1237 	if (error) {
1238 		INP_WUNLOCK(inp);
1239 		return (error);
1240 	}
1241 	NET_EPOCH_ENTER(et);
1242 	error = tcp_output_unlock(tp);
1243 	NET_EPOCH_EXIT(et);
1244 
1245 	return (error);
1246 }
1247 
1248 /*
1249  * Abort the TCP.  Drop the connection abruptly.
1250  */
1251 static void
tcp_usr_abort(struct socket * so)1252 tcp_usr_abort(struct socket *so)
1253 {
1254 	struct inpcb *inp;
1255 	struct tcpcb *tp;
1256 	struct epoch_tracker et;
1257 
1258 	inp = sotoinpcb(so);
1259 	KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL"));
1260 
1261 	NET_EPOCH_ENTER(et);
1262 	INP_WLOCK(inp);
1263 	KASSERT(inp->inp_socket != NULL,
1264 	    ("tcp_usr_abort: inp_socket == NULL"));
1265 
1266 	/*
1267 	 * If we still have full TCP state, and we're not dropped, drop.
1268 	 */
1269 	if (!(inp->inp_flags & INP_DROPPED)) {
1270 		tp = intotcpcb(inp);
1271 		tp = tcp_drop(tp, ECONNABORTED);
1272 		if (tp == NULL)
1273 			goto dropped;
1274 		tcp_bblog_pru(tp, PRU_ABORT, 0);
1275 		TCP_PROBE2(debug__user, tp, PRU_ABORT);
1276 	}
1277 	if (!(inp->inp_flags & INP_DROPPED)) {
1278 		soref(so);
1279 		inp->inp_flags |= INP_SOCKREF;
1280 	}
1281 	INP_WUNLOCK(inp);
1282 dropped:
1283 	NET_EPOCH_EXIT(et);
1284 }
1285 
1286 /*
1287  * TCP socket is closed.  Start friendly disconnect.
1288  */
1289 static void
tcp_usr_close(struct socket * so)1290 tcp_usr_close(struct socket *so)
1291 {
1292 	struct inpcb *inp;
1293 	struct tcpcb *tp;
1294 	struct epoch_tracker et;
1295 
1296 	inp = sotoinpcb(so);
1297 	KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL"));
1298 
1299 	NET_EPOCH_ENTER(et);
1300 	INP_WLOCK(inp);
1301 	KASSERT(inp->inp_socket != NULL,
1302 	    ("tcp_usr_close: inp_socket == NULL"));
1303 
1304 	/*
1305 	 * If we are still connected and we're not dropped, initiate
1306 	 * a disconnect.
1307 	 */
1308 	if (!(inp->inp_flags & INP_DROPPED)) {
1309 		tp = intotcpcb(inp);
1310 		if (tp->t_state != TCPS_TIME_WAIT) {
1311 			tp->t_flags |= TF_CLOSED;
1312 			tcp_disconnect(tp);
1313 			tcp_bblog_pru(tp, PRU_CLOSE, 0);
1314 			TCP_PROBE2(debug__user, tp, PRU_CLOSE);
1315 		}
1316 	}
1317 	if (!(inp->inp_flags & INP_DROPPED)) {
1318 		soref(so);
1319 		inp->inp_flags |= INP_SOCKREF;
1320 	}
1321 	INP_WUNLOCK(inp);
1322 	NET_EPOCH_EXIT(et);
1323 }
1324 
1325 static int
tcp_pru_options_support(struct tcpcb * tp,int flags)1326 tcp_pru_options_support(struct tcpcb *tp, int flags)
1327 {
1328 	/*
1329 	 * If the specific TCP stack has a pru_options
1330 	 * specified then it does not always support
1331 	 * all the PRU_XX options and we must ask it.
1332 	 * If the function is not specified then all
1333 	 * of the PRU_XX options are supported.
1334 	 */
1335 	int ret = 0;
1336 
1337 	if (tp->t_fb->tfb_pru_options) {
1338 		ret = (*tp->t_fb->tfb_pru_options)(tp, flags);
1339 	}
1340 	return (ret);
1341 }
1342 
1343 /*
1344  * Receive out-of-band data.
1345  */
1346 static int
tcp_usr_rcvoob(struct socket * so,struct mbuf * m,int flags)1347 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1348 {
1349 	int error = 0;
1350 	struct inpcb *inp;
1351 	struct tcpcb *tp;
1352 
1353 	inp = sotoinpcb(so);
1354 	KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL"));
1355 	INP_WLOCK(inp);
1356 	if (inp->inp_flags & INP_DROPPED) {
1357 		INP_WUNLOCK(inp);
1358 		return (ECONNRESET);
1359 	}
1360 	tp = intotcpcb(inp);
1361 
1362 	error = tcp_pru_options_support(tp, PRUS_OOB);
1363 	if (error) {
1364 		goto out;
1365 	}
1366 	if ((so->so_oobmark == 0 &&
1367 	     (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) ||
1368 	    so->so_options & SO_OOBINLINE ||
1369 	    tp->t_oobflags & TCPOOB_HADDATA) {
1370 		error = EINVAL;
1371 		goto out;
1372 	}
1373 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1374 		error = EWOULDBLOCK;
1375 		goto out;
1376 	}
1377 	m->m_len = 1;
1378 	*mtod(m, caddr_t) = tp->t_iobc;
1379 	if ((flags & MSG_PEEK) == 0)
1380 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1381 
1382 out:
1383 	tcp_bblog_pru(tp, PRU_RCVOOB, error);
1384 	TCP_PROBE2(debug__user, tp, PRU_RCVOOB);
1385 	INP_WUNLOCK(inp);
1386 	return (error);
1387 }
1388 
1389 #ifdef INET
1390 struct protosw tcp_protosw = {
1391 	.pr_type =		SOCK_STREAM,
1392 	.pr_protocol =		IPPROTO_TCP,
1393 	.pr_flags =		PR_CONNREQUIRED | PR_IMPLOPCL | PR_WANTRCVD |
1394 				    PR_CAPATTACH,
1395 	.pr_ctloutput =		tcp_ctloutput,
1396 	.pr_abort =		tcp_usr_abort,
1397 	.pr_accept =		tcp_usr_accept,
1398 	.pr_attach =		tcp_usr_attach,
1399 	.pr_bind =		tcp_usr_bind,
1400 	.pr_connect =		tcp_usr_connect,
1401 	.pr_control =		in_control,
1402 	.pr_detach =		tcp_usr_detach,
1403 	.pr_disconnect =	tcp_usr_disconnect,
1404 	.pr_listen =		tcp_usr_listen,
1405 	.pr_peeraddr =		in_getpeeraddr,
1406 	.pr_rcvd =		tcp_usr_rcvd,
1407 	.pr_rcvoob =		tcp_usr_rcvoob,
1408 	.pr_send =		tcp_usr_send,
1409 	.pr_ready =		tcp_usr_ready,
1410 	.pr_shutdown =		tcp_usr_shutdown,
1411 	.pr_sockaddr =		in_getsockaddr,
1412 	.pr_sosetlabel =	in_pcbsosetlabel,
1413 	.pr_close =		tcp_usr_close,
1414 };
1415 #endif /* INET */
1416 
1417 #ifdef INET6
1418 struct protosw tcp6_protosw = {
1419 	.pr_type =		SOCK_STREAM,
1420 	.pr_protocol =		IPPROTO_TCP,
1421 	.pr_flags =		PR_CONNREQUIRED | PR_IMPLOPCL |PR_WANTRCVD |
1422 				    PR_CAPATTACH,
1423 	.pr_ctloutput =		tcp_ctloutput,
1424 	.pr_abort =		tcp_usr_abort,
1425 	.pr_accept =		tcp6_usr_accept,
1426 	.pr_attach =		tcp_usr_attach,
1427 	.pr_bind =		tcp6_usr_bind,
1428 	.pr_connect =		tcp6_usr_connect,
1429 	.pr_control =		in6_control,
1430 	.pr_detach =		tcp_usr_detach,
1431 	.pr_disconnect =	tcp_usr_disconnect,
1432 	.pr_listen =		tcp6_usr_listen,
1433 	.pr_peeraddr =		in6_mapped_peeraddr,
1434 	.pr_rcvd =		tcp_usr_rcvd,
1435 	.pr_rcvoob =		tcp_usr_rcvoob,
1436 	.pr_send =		tcp_usr_send,
1437 	.pr_ready =		tcp_usr_ready,
1438 	.pr_shutdown =		tcp_usr_shutdown,
1439 	.pr_sockaddr =		in6_mapped_sockaddr,
1440 	.pr_sosetlabel =	in_pcbsosetlabel,
1441 	.pr_close =		tcp_usr_close,
1442 };
1443 #endif /* INET6 */
1444 
1445 #ifdef INET
1446 /*
1447  * Common subroutine to open a TCP connection to remote host specified
1448  * by struct sockaddr_in.  Call in_pcbconnect() to choose local host address
1449  * and assign a local port number and install the inpcb into the hash.
1450  * Initialize connection parameters and enter SYN-SENT state.
1451  */
1452 static int
tcp_connect(struct tcpcb * tp,struct sockaddr_in * sin,struct thread * td)1453 tcp_connect(struct tcpcb *tp, struct sockaddr_in *sin, struct thread *td)
1454 {
1455 	struct inpcb *inp = tptoinpcb(tp);
1456 	struct socket *so = tptosocket(tp);
1457 	int error;
1458 
1459 	NET_EPOCH_ASSERT();
1460 	INP_WLOCK_ASSERT(inp);
1461 
1462 	if (__predict_false((so->so_state &
1463 	    (SS_ISCONNECTING | SS_ISCONNECTED | SS_ISDISCONNECTING |
1464 	    SS_ISDISCONNECTED)) != 0))
1465 		return (EISCONN);
1466 
1467 	INP_HASH_WLOCK(&V_tcbinfo);
1468 	error = in_pcbconnect(inp, sin, td->td_ucred, true);
1469 	INP_HASH_WUNLOCK(&V_tcbinfo);
1470 	if (error != 0)
1471 		return (error);
1472 
1473 	/*
1474 	 * Compute window scaling to request:
1475 	 * Scale to fit into sweet spot.  See tcp_syncache.c.
1476 	 * XXX: This should move to tcp_output().
1477 	 */
1478 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1479 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1480 		tp->request_r_scale++;
1481 
1482 	soisconnecting(so);
1483 	TCPSTAT_INC(tcps_connattempt);
1484 	tcp_state_change(tp, TCPS_SYN_SENT);
1485 	tp->iss = tcp_new_isn(&inp->inp_inc);
1486 	if (tp->t_flags & TF_REQ_TSTMP)
1487 		tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
1488 	tcp_sendseqinit(tp);
1489 
1490 	return (0);
1491 }
1492 #endif /* INET */
1493 
1494 #ifdef INET6
1495 static int
tcp6_connect(struct tcpcb * tp,struct sockaddr_in6 * sin6,struct thread * td)1496 tcp6_connect(struct tcpcb *tp, struct sockaddr_in6 *sin6, struct thread *td)
1497 {
1498 	struct inpcb *inp = tptoinpcb(tp);
1499 	struct socket *so = tptosocket(tp);
1500 	int error;
1501 
1502 	NET_EPOCH_ASSERT();
1503 	INP_WLOCK_ASSERT(inp);
1504 
1505 	if (__predict_false((so->so_state &
1506 	    (SS_ISCONNECTING | SS_ISCONNECTED)) != 0))
1507 		return (EISCONN);
1508 
1509 	INP_HASH_WLOCK(&V_tcbinfo);
1510 	error = in6_pcbconnect(inp, sin6, td->td_ucred, true);
1511 	INP_HASH_WUNLOCK(&V_tcbinfo);
1512 	if (error != 0)
1513 		return (error);
1514 
1515 	/* Compute window scaling to request.  */
1516 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
1517 	    (TCP_MAXWIN << tp->request_r_scale) < sb_max)
1518 		tp->request_r_scale++;
1519 
1520 	soisconnecting(so);
1521 	TCPSTAT_INC(tcps_connattempt);
1522 	tcp_state_change(tp, TCPS_SYN_SENT);
1523 	tp->iss = tcp_new_isn(&inp->inp_inc);
1524 	if (tp->t_flags & TF_REQ_TSTMP)
1525 		tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc);
1526 	tcp_sendseqinit(tp);
1527 
1528 	return (0);
1529 }
1530 #endif /* INET6 */
1531 
1532 /*
1533  * Export TCP internal state information via a struct tcp_info, based on the
1534  * Linux 2.6 API.  Not ABI compatible as our constants are mapped differently
1535  * (TCP state machine, etc).  We export all information using FreeBSD-native
1536  * constants -- for example, the numeric values for tcpi_state will differ
1537  * from Linux.
1538  */
1539 void
tcp_fill_info(const struct tcpcb * tp,struct tcp_info * ti)1540 tcp_fill_info(const struct tcpcb *tp, struct tcp_info *ti)
1541 {
1542 
1543 	INP_LOCK_ASSERT(tptoinpcb(tp));
1544 	bzero(ti, sizeof(*ti));
1545 
1546 	ti->tcpi_state = tp->t_state;
1547 	if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP))
1548 		ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1549 	if (tp->t_flags & TF_SACK_PERMIT)
1550 		ti->tcpi_options |= TCPI_OPT_SACK;
1551 	if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) {
1552 		ti->tcpi_options |= TCPI_OPT_WSCALE;
1553 		ti->tcpi_snd_wscale = tp->snd_scale;
1554 		ti->tcpi_rcv_wscale = tp->rcv_scale;
1555 	}
1556 	switch (tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) {
1557 		case TF2_ECN_PERMIT:
1558 			ti->tcpi_options |= TCPI_OPT_ECN;
1559 			break;
1560 		case TF2_ACE_PERMIT:
1561 			/* FALLTHROUGH */
1562 		case TF2_ECN_PERMIT | TF2_ACE_PERMIT:
1563 			ti->tcpi_options |= TCPI_OPT_ACE;
1564 			break;
1565 		default:
1566 			break;
1567 	}
1568 	if (IS_FASTOPEN(tp->t_flags))
1569 		ti->tcpi_options |= TCPI_OPT_TFO;
1570 
1571 	ti->tcpi_rto = tp->t_rxtcur * tick;
1572 	ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick;
1573 	ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT;
1574 	ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT;
1575 
1576 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1577 	ti->tcpi_snd_cwnd = tp->snd_cwnd;
1578 
1579 	/*
1580 	 * FreeBSD-specific extension fields for tcp_info.
1581 	 */
1582 	ti->tcpi_rcv_space = tp->rcv_wnd;
1583 	ti->tcpi_rcv_nxt = tp->rcv_nxt;
1584 	ti->tcpi_snd_wnd = tp->snd_wnd;
1585 	ti->tcpi_snd_bwnd = 0;		/* Unused, kept for compat. */
1586 	ti->tcpi_snd_nxt = tp->snd_nxt;
1587 	ti->tcpi_snd_mss = tp->t_maxseg;
1588 	ti->tcpi_rcv_mss = tp->t_maxseg;
1589 	ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack;
1590 	ti->tcpi_rcv_ooopack = tp->t_rcvoopack;
1591 	ti->tcpi_snd_zerowin = tp->t_sndzerowin;
1592 	ti->tcpi_snd_una = tp->snd_una;
1593 	ti->tcpi_snd_max = tp->snd_max;
1594 	ti->tcpi_rcv_numsacks = tp->rcv_numsacks;
1595 	ti->tcpi_rcv_adv = tp->rcv_adv;
1596 	ti->tcpi_dupacks = tp->t_dupacks;
1597 #ifdef TCP_OFFLOAD
1598 	if (tp->t_flags & TF_TOE) {
1599 		ti->tcpi_options |= TCPI_OPT_TOE;
1600 		tcp_offload_tcp_info(tp, ti);
1601 	}
1602 #endif
1603 	/*
1604 	 * AccECN related counters.
1605 	 */
1606 	if ((tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) ==
1607 	    (TF2_ECN_PERMIT | TF2_ACE_PERMIT))
1608 		/*
1609 		 * Internal counter starts at 5 for AccECN
1610 		 * but 0 for RFC3168 ECN.
1611 		 */
1612 		ti->tcpi_delivered_ce = tp->t_scep - 5;
1613 	else
1614 		ti->tcpi_delivered_ce = tp->t_scep;
1615 	ti->tcpi_received_ce = tp->t_rcep;
1616 }
1617 
1618 /*
1619  * tcp_ctloutput() must drop the inpcb lock before performing copyin on
1620  * socket option arguments.  When it re-acquires the lock after the copy, it
1621  * has to revalidate that the connection is still valid for the socket
1622  * option.
1623  */
1624 #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do {			\
1625 	INP_WLOCK(inp);							\
1626 	if (inp->inp_flags & INP_DROPPED) {				\
1627 		INP_WUNLOCK(inp);					\
1628 		cleanup;						\
1629 		return (ECONNRESET);					\
1630 	}								\
1631 	tp = intotcpcb(inp);						\
1632 } while(0)
1633 #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */)
1634 
1635 int
tcp_ctloutput_set(struct inpcb * inp,struct sockopt * sopt)1636 tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt)
1637 {
1638 	struct socket *so = inp->inp_socket;
1639 	struct tcpcb *tp = intotcpcb(inp);
1640 	int error = 0;
1641 
1642 	MPASS(sopt->sopt_dir == SOPT_SET);
1643 	INP_WLOCK_ASSERT(inp);
1644 	KASSERT((inp->inp_flags & INP_DROPPED) == 0,
1645 	    ("inp_flags == %x", inp->inp_flags));
1646 	KASSERT(so != NULL, ("inp_socket == NULL"));
1647 
1648 	if (sopt->sopt_level != IPPROTO_TCP) {
1649 		INP_WUNLOCK(inp);
1650 #ifdef INET6
1651 		if (inp->inp_vflag & INP_IPV6PROTO)
1652 			error = ip6_ctloutput(so, sopt);
1653 #endif
1654 #if defined(INET6) && defined(INET)
1655 		else
1656 #endif
1657 #ifdef INET
1658 			error = ip_ctloutput(so, sopt);
1659 #endif
1660 		/*
1661 		 * When an IP-level socket option affects TCP, pass control
1662 		 * down to stack tfb_tcp_ctloutput, otherwise return what
1663 		 * IP level returned.
1664 		 */
1665 		switch (sopt->sopt_level) {
1666 #ifdef INET6
1667 		case IPPROTO_IPV6:
1668 			if ((inp->inp_vflag & INP_IPV6PROTO) == 0)
1669 				return (error);
1670 			switch (sopt->sopt_name) {
1671 			case IPV6_TCLASS:
1672 				/* Notify tcp stacks that care (e.g. RACK). */
1673 				break;
1674 			case IPV6_USE_MIN_MTU:
1675 				/* Update t_maxseg accordingly. */
1676 				break;
1677 			default:
1678 				return (error);
1679 			}
1680 			break;
1681 #endif
1682 #ifdef INET
1683 		case IPPROTO_IP:
1684 			switch (sopt->sopt_name) {
1685 			case IP_TOS:
1686 				inp->inp_ip_tos &= ~IPTOS_ECN_MASK;
1687 				break;
1688 			case IP_TTL:
1689 				/* Notify tcp stacks that care (e.g. RACK). */
1690 				break;
1691 			default:
1692 				return (error);
1693 			}
1694 			break;
1695 #endif
1696 		default:
1697 			return (error);
1698 		}
1699 		INP_WLOCK_RECHECK(inp);
1700 	} else if (sopt->sopt_name == TCP_FUNCTION_BLK) {
1701 		/*
1702 		 * Protect the TCP option TCP_FUNCTION_BLK so
1703 		 * that a sub-function can *never* overwrite this.
1704 		 */
1705 		struct tcp_function_set fsn;
1706 		struct tcp_function_block *blk;
1707 		void *ptr = NULL;
1708 
1709 		INP_WUNLOCK(inp);
1710 		error = sooptcopyin(sopt, &fsn, sizeof fsn, sizeof fsn);
1711 		if (error)
1712 			return (error);
1713 
1714 		INP_WLOCK_RECHECK(inp);
1715 
1716 		blk = find_and_ref_tcp_functions(&fsn);
1717 		if (blk == NULL) {
1718 			INP_WUNLOCK(inp);
1719 			return (ENOENT);
1720 		}
1721 		if (tp->t_fb == blk) {
1722 			/* You already have this */
1723 			refcount_release(&blk->tfb_refcnt);
1724 			INP_WUNLOCK(inp);
1725 			return (0);
1726 		}
1727 		if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) {
1728 			refcount_release(&blk->tfb_refcnt);
1729 			INP_WUNLOCK(inp);
1730 			return (ENOENT);
1731 		}
1732 		error = (*blk->tfb_tcp_handoff_ok)(tp);
1733 		if (error) {
1734 			refcount_release(&blk->tfb_refcnt);
1735 			INP_WUNLOCK(inp);
1736 			return (error);
1737 		}
1738 		/*
1739 		 * Ensure the new stack takes ownership with a
1740 		 * clean slate on peak rate threshold.
1741 		 */
1742 		if (tp->t_fb->tfb_tcp_timer_stop_all != NULL)
1743 			tp->t_fb->tfb_tcp_timer_stop_all(tp);
1744 		if (blk->tfb_tcp_fb_init) {
1745 			error = (*blk->tfb_tcp_fb_init)(tp, &ptr);
1746 			if (error) {
1747 				/*
1748 				 * Release the ref count the lookup
1749 				 * acquired.
1750 				 */
1751 				refcount_release(&blk->tfb_refcnt);
1752 				/*
1753 				 * Now there is a chance that the
1754 				 * init() function mucked with some
1755 				 * things before it failed, such as
1756 				 * hpts or inp_flags2 or timer granularity.
1757 				 * It should not of, but lets give the old
1758 				 * stack a chance to reset to a known good state.
1759 				 */
1760 				if (tp->t_fb->tfb_switch_failed) {
1761 					(*tp->t_fb->tfb_switch_failed)(tp);
1762 				}
1763 			 	goto err_out;
1764 			}
1765 		}
1766 		if (tp->t_fb->tfb_tcp_fb_fini) {
1767 			struct epoch_tracker et;
1768 			/*
1769 			 * Tell the stack to cleanup with 0 i.e.
1770 			 * the tcb is not going away.
1771 			 */
1772 			NET_EPOCH_ENTER(et);
1773 			(*tp->t_fb->tfb_tcp_fb_fini)(tp, 0);
1774 			NET_EPOCH_EXIT(et);
1775 		}
1776 		/*
1777 		 * Release the old refcnt, the
1778 		 * lookup acquired a ref on the
1779 		 * new one already.
1780 		 */
1781 		refcount_release(&tp->t_fb->tfb_refcnt);
1782 		/*
1783 		 * Set in the new stack.
1784 		 */
1785 		tp->t_fb = blk;
1786 		tp->t_fb_ptr = ptr;
1787 #ifdef TCP_OFFLOAD
1788 		if (tp->t_flags & TF_TOE) {
1789 			tcp_offload_ctloutput(tp, sopt->sopt_dir,
1790 			     sopt->sopt_name);
1791 		}
1792 #endif
1793 err_out:
1794 		INP_WUNLOCK(inp);
1795 		return (error);
1796 
1797 	}
1798 
1799 	/* Pass in the INP locked, callee must unlock it. */
1800 	return (tp->t_fb->tfb_tcp_ctloutput(tp, sopt));
1801 }
1802 
1803 static int
tcp_ctloutput_get(struct inpcb * inp,struct sockopt * sopt)1804 tcp_ctloutput_get(struct inpcb *inp, struct sockopt *sopt)
1805 {
1806 	struct socket *so = inp->inp_socket;
1807 	struct tcpcb *tp = intotcpcb(inp);
1808 	int error = 0;
1809 
1810 	MPASS(sopt->sopt_dir == SOPT_GET);
1811 	INP_WLOCK_ASSERT(inp);
1812 	KASSERT((inp->inp_flags & INP_DROPPED) == 0,
1813 	    ("inp_flags == %x", inp->inp_flags));
1814 	KASSERT(so != NULL, ("inp_socket == NULL"));
1815 
1816 	if (sopt->sopt_level != IPPROTO_TCP) {
1817 		INP_WUNLOCK(inp);
1818 #ifdef INET6
1819 		if (inp->inp_vflag & INP_IPV6PROTO)
1820 			error = ip6_ctloutput(so, sopt);
1821 #endif /* INET6 */
1822 #if defined(INET6) && defined(INET)
1823 		else
1824 #endif
1825 #ifdef INET
1826 			error = ip_ctloutput(so, sopt);
1827 #endif
1828 		return (error);
1829 	}
1830 	if (((sopt->sopt_name == TCP_FUNCTION_BLK) ||
1831 	     (sopt->sopt_name == TCP_FUNCTION_ALIAS))) {
1832 		struct tcp_function_set fsn;
1833 
1834 		if (sopt->sopt_name == TCP_FUNCTION_ALIAS) {
1835 			memset(&fsn, 0, sizeof(fsn));
1836 			find_tcp_function_alias(tp->t_fb, &fsn);
1837 		} else {
1838 			strncpy(fsn.function_set_name,
1839 			    tp->t_fb->tfb_tcp_block_name,
1840 			    TCP_FUNCTION_NAME_LEN_MAX);
1841 			fsn.function_set_name[TCP_FUNCTION_NAME_LEN_MAX - 1] = '\0';
1842 		}
1843 		fsn.pcbcnt = tp->t_fb->tfb_refcnt;
1844 		INP_WUNLOCK(inp);
1845 		error = sooptcopyout(sopt, &fsn, sizeof fsn);
1846 		return (error);
1847 	}
1848 
1849 	/* Pass in the INP locked, callee must unlock it. */
1850 	return (tp->t_fb->tfb_tcp_ctloutput(tp, sopt));
1851 }
1852 
1853 int
tcp_ctloutput(struct socket * so,struct sockopt * sopt)1854 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1855 {
1856 	struct	inpcb *inp;
1857 
1858 	inp = sotoinpcb(so);
1859 	KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL"));
1860 
1861 	INP_WLOCK(inp);
1862 	if (inp->inp_flags & INP_DROPPED) {
1863 		INP_WUNLOCK(inp);
1864 		return (ECONNRESET);
1865 	}
1866 	if (sopt->sopt_dir == SOPT_SET)
1867 		return (tcp_ctloutput_set(inp, sopt));
1868 	else if (sopt->sopt_dir == SOPT_GET)
1869 		return (tcp_ctloutput_get(inp, sopt));
1870 	else
1871 		panic("%s: sopt_dir $%d", __func__, sopt->sopt_dir);
1872 }
1873 
1874 /*
1875  * If this assert becomes untrue, we need to change the size of the buf
1876  * variable in tcp_default_ctloutput().
1877  */
1878 #ifdef CTASSERT
1879 CTASSERT(TCP_CA_NAME_MAX <= TCP_LOG_ID_LEN);
1880 CTASSERT(TCP_LOG_REASON_LEN <= TCP_LOG_ID_LEN);
1881 #endif
1882 
1883 #ifdef KERN_TLS
1884 static int
copyin_tls_enable(struct sockopt * sopt,struct tls_enable * tls)1885 copyin_tls_enable(struct sockopt *sopt, struct tls_enable *tls)
1886 {
1887 	struct tls_enable_v0 tls_v0;
1888 	int error;
1889 
1890 	if (sopt->sopt_valsize == sizeof(tls_v0)) {
1891 		error = sooptcopyin(sopt, &tls_v0, sizeof(tls_v0),
1892 		    sizeof(tls_v0));
1893 		if (error)
1894 			return (error);
1895 		memset(tls, 0, sizeof(*tls));
1896 		tls->cipher_key = tls_v0.cipher_key;
1897 		tls->iv = tls_v0.iv;
1898 		tls->auth_key = tls_v0.auth_key;
1899 		tls->cipher_algorithm = tls_v0.cipher_algorithm;
1900 		tls->cipher_key_len = tls_v0.cipher_key_len;
1901 		tls->iv_len = tls_v0.iv_len;
1902 		tls->auth_algorithm = tls_v0.auth_algorithm;
1903 		tls->auth_key_len = tls_v0.auth_key_len;
1904 		tls->flags = tls_v0.flags;
1905 		tls->tls_vmajor = tls_v0.tls_vmajor;
1906 		tls->tls_vminor = tls_v0.tls_vminor;
1907 		return (0);
1908 	}
1909 
1910 	return (sooptcopyin(sopt, tls, sizeof(*tls), sizeof(*tls)));
1911 }
1912 #endif
1913 
1914 extern struct cc_algo newreno_cc_algo;
1915 
1916 static int
tcp_set_cc_mod(struct inpcb * inp,struct sockopt * sopt)1917 tcp_set_cc_mod(struct inpcb *inp, struct sockopt *sopt)
1918 {
1919 	struct cc_algo *algo;
1920 	void *ptr = NULL;
1921 	struct tcpcb *tp;
1922 	struct cc_var cc_mem;
1923 	char	buf[TCP_CA_NAME_MAX];
1924 	size_t mem_sz;
1925 	int error;
1926 
1927 	INP_WUNLOCK(inp);
1928 	error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1);
1929 	if (error)
1930 		return(error);
1931 	buf[sopt->sopt_valsize] = '\0';
1932 	CC_LIST_RLOCK();
1933 	STAILQ_FOREACH(algo, &cc_list, entries) {
1934 		if (strncmp(buf, algo->name,
1935 			    TCP_CA_NAME_MAX) == 0) {
1936 			if (algo->flags & CC_MODULE_BEING_REMOVED) {
1937 				/* We can't "see" modules being unloaded */
1938 				continue;
1939 			}
1940 			break;
1941 		}
1942 	}
1943 	if (algo == NULL) {
1944 		CC_LIST_RUNLOCK();
1945 		return(ESRCH);
1946 	}
1947 	/*
1948 	 * With a reference the algorithm cannot be removed
1949 	 * so we hold a reference through the change process.
1950 	 */
1951 	cc_refer(algo);
1952 	CC_LIST_RUNLOCK();
1953 	if (algo->cb_init != NULL) {
1954 		/* We can now pre-get the memory for the CC */
1955 		mem_sz = (*algo->cc_data_sz)();
1956 		if (mem_sz == 0) {
1957 			goto no_mem_needed;
1958 		}
1959 		ptr = malloc(mem_sz, M_CC_MEM, M_WAITOK);
1960 	} else {
1961 no_mem_needed:
1962 		mem_sz = 0;
1963 		ptr = NULL;
1964 	}
1965 	/*
1966 	 * Make sure its all clean and zero and also get
1967 	 * back the inplock.
1968 	 */
1969 	memset(&cc_mem, 0, sizeof(cc_mem));
1970 	INP_WLOCK(inp);
1971 	if (inp->inp_flags & INP_DROPPED) {
1972 		INP_WUNLOCK(inp);
1973 		if (ptr)
1974 			free(ptr, M_CC_MEM);
1975 		/* Release our temp reference */
1976 		CC_LIST_RLOCK();
1977 		cc_release(algo);
1978 		CC_LIST_RUNLOCK();
1979 		return (ECONNRESET);
1980 	}
1981 	tp = intotcpcb(inp);
1982 	if (ptr != NULL)
1983 		memset(ptr, 0, mem_sz);
1984 	cc_mem.ccvc.tcp = tp;
1985 	/*
1986 	 * We once again hold a write lock over the tcb so it's
1987 	 * safe to do these things without ordering concerns.
1988 	 * Note here we init into stack memory.
1989 	 */
1990 	if (algo->cb_init != NULL)
1991 		error = algo->cb_init(&cc_mem, ptr);
1992 	else
1993 		error = 0;
1994 	/*
1995 	 * The CC algorithms, when given their memory
1996 	 * should not fail we could in theory have a
1997 	 * KASSERT here.
1998 	 */
1999 	if (error == 0) {
2000 		/*
2001 		 * Touchdown, lets go ahead and move the
2002 		 * connection to the new CC module by
2003 		 * copying in the cc_mem after we call
2004 		 * the old ones cleanup (if any).
2005 		 */
2006 		if (CC_ALGO(tp)->cb_destroy != NULL)
2007 			CC_ALGO(tp)->cb_destroy(&tp->t_ccv);
2008 		/* Detach the old CC from the tcpcb  */
2009 		cc_detach(tp);
2010 		/* Copy in our temp memory that was inited */
2011 		memcpy(&tp->t_ccv, &cc_mem, sizeof(struct cc_var));
2012 		/* Now attach the new, which takes a reference */
2013 		cc_attach(tp, algo);
2014 		/* Ok now are we where we have gotten past any conn_init? */
2015 		if (TCPS_HAVEESTABLISHED(tp->t_state) && (CC_ALGO(tp)->conn_init != NULL)) {
2016 			/* Yep run the connection init for the new CC */
2017 			CC_ALGO(tp)->conn_init(&tp->t_ccv);
2018 		}
2019 	} else if (ptr)
2020 		free(ptr, M_CC_MEM);
2021 	INP_WUNLOCK(inp);
2022 	/* Now lets release our temp reference */
2023 	CC_LIST_RLOCK();
2024 	cc_release(algo);
2025 	CC_LIST_RUNLOCK();
2026 	return (error);
2027 }
2028 
2029 int
tcp_default_ctloutput(struct tcpcb * tp,struct sockopt * sopt)2030 tcp_default_ctloutput(struct tcpcb *tp, struct sockopt *sopt)
2031 {
2032 	struct inpcb *inp = tptoinpcb(tp);
2033 	int	error, opt, optval;
2034 	u_int	ui;
2035 	struct	tcp_info ti;
2036 #ifdef KERN_TLS
2037 	struct tls_enable tls;
2038 	struct socket *so = inp->inp_socket;
2039 #endif
2040 	char	*pbuf, buf[TCP_LOG_ID_LEN];
2041 #ifdef STATS
2042 	struct statsblob *sbp;
2043 #endif
2044 	size_t	len;
2045 
2046 	INP_WLOCK_ASSERT(inp);
2047 	KASSERT((inp->inp_flags & INP_DROPPED) == 0,
2048 	    ("inp_flags == %x", inp->inp_flags));
2049 	KASSERT(inp->inp_socket != NULL, ("inp_socket == NULL"));
2050 
2051 	switch (sopt->sopt_level) {
2052 #ifdef INET6
2053 	case IPPROTO_IPV6:
2054 		MPASS(inp->inp_vflag & INP_IPV6PROTO);
2055 		switch (sopt->sopt_name) {
2056 		case IPV6_USE_MIN_MTU:
2057 			tcp6_use_min_mtu(tp);
2058 			/* FALLTHROUGH */
2059 		}
2060 		INP_WUNLOCK(inp);
2061 		return (0);
2062 #endif
2063 #ifdef INET
2064 	case IPPROTO_IP:
2065 		INP_WUNLOCK(inp);
2066 		return (0);
2067 #endif
2068 	}
2069 
2070 	/*
2071 	 * For TCP_CCALGOOPT forward the control to CC module, for both
2072 	 * SOPT_SET and SOPT_GET.
2073 	 */
2074 	switch (sopt->sopt_name) {
2075 	case TCP_CCALGOOPT:
2076 		INP_WUNLOCK(inp);
2077 		if (sopt->sopt_valsize > CC_ALGOOPT_LIMIT)
2078 			return (EINVAL);
2079 		pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO);
2080 		error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize,
2081 		    sopt->sopt_valsize);
2082 		if (error) {
2083 			free(pbuf, M_TEMP);
2084 			return (error);
2085 		}
2086 		INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP));
2087 		if (CC_ALGO(tp)->ctl_output != NULL)
2088 			error = CC_ALGO(tp)->ctl_output(&tp->t_ccv, sopt, pbuf);
2089 		else
2090 			error = ENOENT;
2091 		INP_WUNLOCK(inp);
2092 		if (error == 0 && sopt->sopt_dir == SOPT_GET)
2093 			error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize);
2094 		free(pbuf, M_TEMP);
2095 		return (error);
2096 	}
2097 
2098 	switch (sopt->sopt_dir) {
2099 	case SOPT_SET:
2100 		switch (sopt->sopt_name) {
2101 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
2102 		case TCP_MD5SIG:
2103 			INP_WUNLOCK(inp);
2104 			if (!TCPMD5_ENABLED())
2105 				return (ENOPROTOOPT);
2106 			error = TCPMD5_PCBCTL(inp, sopt);
2107 			if (error)
2108 				return (error);
2109 			INP_WLOCK_RECHECK(inp);
2110 			goto unlock_and_done;
2111 #endif /* IPSEC */
2112 
2113 		case TCP_NODELAY:
2114 		case TCP_NOOPT:
2115 		case TCP_LRD:
2116 			INP_WUNLOCK(inp);
2117 			error = sooptcopyin(sopt, &optval, sizeof optval,
2118 			    sizeof optval);
2119 			if (error)
2120 				return (error);
2121 
2122 			INP_WLOCK_RECHECK(inp);
2123 			switch (sopt->sopt_name) {
2124 			case TCP_NODELAY:
2125 				opt = TF_NODELAY;
2126 				break;
2127 			case TCP_NOOPT:
2128 				opt = TF_NOOPT;
2129 				break;
2130 			case TCP_LRD:
2131 				opt = TF_LRD;
2132 				break;
2133 			default:
2134 				opt = 0; /* dead code to fool gcc */
2135 				break;
2136 			}
2137 
2138 			if (optval)
2139 				tp->t_flags |= opt;
2140 			else
2141 				tp->t_flags &= ~opt;
2142 unlock_and_done:
2143 #ifdef TCP_OFFLOAD
2144 			if (tp->t_flags & TF_TOE) {
2145 				tcp_offload_ctloutput(tp, sopt->sopt_dir,
2146 				    sopt->sopt_name);
2147 			}
2148 #endif
2149 			INP_WUNLOCK(inp);
2150 			break;
2151 
2152 		case TCP_NOPUSH:
2153 			INP_WUNLOCK(inp);
2154 			error = sooptcopyin(sopt, &optval, sizeof optval,
2155 			    sizeof optval);
2156 			if (error)
2157 				return (error);
2158 
2159 			INP_WLOCK_RECHECK(inp);
2160 			if (optval)
2161 				tp->t_flags |= TF_NOPUSH;
2162 			else if (tp->t_flags & TF_NOPUSH) {
2163 				tp->t_flags &= ~TF_NOPUSH;
2164 				if (TCPS_HAVEESTABLISHED(tp->t_state)) {
2165 					struct epoch_tracker et;
2166 
2167 					NET_EPOCH_ENTER(et);
2168 					error = tcp_output_nodrop(tp);
2169 					NET_EPOCH_EXIT(et);
2170 				}
2171 			}
2172 			goto unlock_and_done;
2173 
2174 		case TCP_REMOTE_UDP_ENCAPS_PORT:
2175 			INP_WUNLOCK(inp);
2176 			error = sooptcopyin(sopt, &optval, sizeof optval,
2177 			    sizeof optval);
2178 			if (error)
2179 				return (error);
2180 			if ((optval < TCP_TUNNELING_PORT_MIN) ||
2181 			    (optval > TCP_TUNNELING_PORT_MAX)) {
2182 				/* Its got to be in range */
2183 				return (EINVAL);
2184 			}
2185 			if ((V_tcp_udp_tunneling_port == 0) && (optval != 0)) {
2186 				/* You have to have enabled a UDP tunneling port first */
2187 				return (EINVAL);
2188 			}
2189 			INP_WLOCK_RECHECK(inp);
2190 			if (tp->t_state != TCPS_CLOSED) {
2191 				/* You can't change after you are connected */
2192 				error = EINVAL;
2193 			} else {
2194 				/* Ok we are all good set the port */
2195 				tp->t_port = htons(optval);
2196 			}
2197 			goto unlock_and_done;
2198 
2199 		case TCP_MAXSEG:
2200 			INP_WUNLOCK(inp);
2201 			error = sooptcopyin(sopt, &optval, sizeof optval,
2202 			    sizeof optval);
2203 			if (error)
2204 				return (error);
2205 
2206 			INP_WLOCK_RECHECK(inp);
2207 			if (optval > 0 && optval <= tp->t_maxseg &&
2208 			    optval + 40 >= V_tcp_minmss)
2209 				tp->t_maxseg = optval;
2210 			else
2211 				error = EINVAL;
2212 			goto unlock_and_done;
2213 
2214 		case TCP_INFO:
2215 			INP_WUNLOCK(inp);
2216 			error = EINVAL;
2217 			break;
2218 
2219 		case TCP_STATS:
2220 			INP_WUNLOCK(inp);
2221 #ifdef STATS
2222 			error = sooptcopyin(sopt, &optval, sizeof optval,
2223 			    sizeof optval);
2224 			if (error)
2225 				return (error);
2226 
2227 			if (optval > 0)
2228 				sbp = stats_blob_alloc(
2229 				    V_tcp_perconn_stats_dflt_tpl, 0);
2230 			else
2231 				sbp = NULL;
2232 
2233 			INP_WLOCK_RECHECK(inp);
2234 			if ((tp->t_stats != NULL && sbp == NULL) ||
2235 			    (tp->t_stats == NULL && sbp != NULL)) {
2236 				struct statsblob *t = tp->t_stats;
2237 				tp->t_stats = sbp;
2238 				sbp = t;
2239 			}
2240 			INP_WUNLOCK(inp);
2241 
2242 			stats_blob_destroy(sbp);
2243 #else
2244 			return (EOPNOTSUPP);
2245 #endif /* !STATS */
2246 			break;
2247 
2248 		case TCP_CONGESTION:
2249 			error = tcp_set_cc_mod(inp, sopt);
2250 			break;
2251 
2252 		case TCP_REUSPORT_LB_NUMA:
2253 			INP_WUNLOCK(inp);
2254 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2255 			    sizeof(optval));
2256 			INP_WLOCK_RECHECK(inp);
2257 			if (!error)
2258 				error = in_pcblbgroup_numa(inp, optval);
2259 			INP_WUNLOCK(inp);
2260 			break;
2261 
2262 #ifdef KERN_TLS
2263 		case TCP_TXTLS_ENABLE:
2264 			INP_WUNLOCK(inp);
2265 			error = copyin_tls_enable(sopt, &tls);
2266 			if (error)
2267 				break;
2268 			error = ktls_enable_tx(so, &tls);
2269 			break;
2270 		case TCP_TXTLS_MODE:
2271 			INP_WUNLOCK(inp);
2272 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2273 			if (error)
2274 				return (error);
2275 
2276 			INP_WLOCK_RECHECK(inp);
2277 			error = ktls_set_tx_mode(so, ui);
2278 			INP_WUNLOCK(inp);
2279 			break;
2280 		case TCP_RXTLS_ENABLE:
2281 			INP_WUNLOCK(inp);
2282 			error = sooptcopyin(sopt, &tls, sizeof(tls),
2283 			    sizeof(tls));
2284 			if (error)
2285 				break;
2286 			error = ktls_enable_rx(so, &tls);
2287 			break;
2288 #endif
2289 		case TCP_MAXUNACKTIME:
2290 		case TCP_KEEPIDLE:
2291 		case TCP_KEEPINTVL:
2292 		case TCP_KEEPINIT:
2293 			INP_WUNLOCK(inp);
2294 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2295 			if (error)
2296 				return (error);
2297 
2298 			if (ui > (UINT_MAX / hz)) {
2299 				error = EINVAL;
2300 				break;
2301 			}
2302 			ui *= hz;
2303 
2304 			INP_WLOCK_RECHECK(inp);
2305 			switch (sopt->sopt_name) {
2306 			case TCP_MAXUNACKTIME:
2307 				tp->t_maxunacktime = ui;
2308 				break;
2309 
2310 			case TCP_KEEPIDLE:
2311 				tp->t_keepidle = ui;
2312 				/*
2313 				 * XXX: better check current remaining
2314 				 * timeout and "merge" it with new value.
2315 				 */
2316 				if ((tp->t_state > TCPS_LISTEN) &&
2317 				    (tp->t_state <= TCPS_CLOSING))
2318 					tcp_timer_activate(tp, TT_KEEP,
2319 					    TP_KEEPIDLE(tp));
2320 				break;
2321 			case TCP_KEEPINTVL:
2322 				tp->t_keepintvl = ui;
2323 				if ((tp->t_state == TCPS_FIN_WAIT_2) &&
2324 				    (TP_MAXIDLE(tp) > 0))
2325 					tcp_timer_activate(tp, TT_2MSL,
2326 					    TP_MAXIDLE(tp));
2327 				break;
2328 			case TCP_KEEPINIT:
2329 				tp->t_keepinit = ui;
2330 				if (tp->t_state == TCPS_SYN_RECEIVED ||
2331 				    tp->t_state == TCPS_SYN_SENT)
2332 					tcp_timer_activate(tp, TT_KEEP,
2333 					    TP_KEEPINIT(tp));
2334 				break;
2335 			}
2336 			goto unlock_and_done;
2337 
2338 		case TCP_KEEPCNT:
2339 			INP_WUNLOCK(inp);
2340 			error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui));
2341 			if (error)
2342 				return (error);
2343 
2344 			INP_WLOCK_RECHECK(inp);
2345 			tp->t_keepcnt = ui;
2346 			if ((tp->t_state == TCPS_FIN_WAIT_2) &&
2347 			    (TP_MAXIDLE(tp) > 0))
2348 				tcp_timer_activate(tp, TT_2MSL,
2349 				    TP_MAXIDLE(tp));
2350 			goto unlock_and_done;
2351 
2352 #ifdef TCPPCAP
2353 		case TCP_PCAP_OUT:
2354 		case TCP_PCAP_IN:
2355 			INP_WUNLOCK(inp);
2356 			error = sooptcopyin(sopt, &optval, sizeof optval,
2357 			    sizeof optval);
2358 			if (error)
2359 				return (error);
2360 
2361 			INP_WLOCK_RECHECK(inp);
2362 			if (optval >= 0)
2363 				tcp_pcap_set_sock_max(
2364 					(sopt->sopt_name == TCP_PCAP_OUT) ?
2365 					&(tp->t_outpkts) : &(tp->t_inpkts),
2366 					optval);
2367 			else
2368 				error = EINVAL;
2369 			goto unlock_and_done;
2370 #endif
2371 
2372 		case TCP_FASTOPEN: {
2373 			struct tcp_fastopen tfo_optval;
2374 
2375 			INP_WUNLOCK(inp);
2376 			if (!V_tcp_fastopen_client_enable &&
2377 			    !V_tcp_fastopen_server_enable)
2378 				return (EPERM);
2379 
2380 			error = sooptcopyin(sopt, &tfo_optval,
2381 				    sizeof(tfo_optval), sizeof(int));
2382 			if (error)
2383 				return (error);
2384 
2385 			INP_WLOCK_RECHECK(inp);
2386 			if ((tp->t_state != TCPS_CLOSED) &&
2387 			    (tp->t_state != TCPS_LISTEN)) {
2388 				error = EINVAL;
2389 				goto unlock_and_done;
2390 			}
2391 			if (tfo_optval.enable) {
2392 				if (tp->t_state == TCPS_LISTEN) {
2393 					if (!V_tcp_fastopen_server_enable) {
2394 						error = EPERM;
2395 						goto unlock_and_done;
2396 					}
2397 
2398 					if (tp->t_tfo_pending == NULL)
2399 						tp->t_tfo_pending =
2400 						    tcp_fastopen_alloc_counter();
2401 				} else {
2402 					/*
2403 					 * If a pre-shared key was provided,
2404 					 * stash it in the client cookie
2405 					 * field of the tcpcb for use during
2406 					 * connect.
2407 					 */
2408 					if (sopt->sopt_valsize ==
2409 					    sizeof(tfo_optval)) {
2410 						memcpy(tp->t_tfo_cookie.client,
2411 						       tfo_optval.psk,
2412 						       TCP_FASTOPEN_PSK_LEN);
2413 						tp->t_tfo_client_cookie_len =
2414 						    TCP_FASTOPEN_PSK_LEN;
2415 					}
2416 				}
2417 				tp->t_flags |= TF_FASTOPEN;
2418 			} else
2419 				tp->t_flags &= ~TF_FASTOPEN;
2420 			goto unlock_and_done;
2421 		}
2422 
2423 #ifdef TCP_BLACKBOX
2424 		case TCP_LOG:
2425 			INP_WUNLOCK(inp);
2426 			error = sooptcopyin(sopt, &optval, sizeof optval,
2427 			    sizeof optval);
2428 			if (error)
2429 				return (error);
2430 
2431 			INP_WLOCK_RECHECK(inp);
2432 			error = tcp_log_state_change(tp, optval);
2433 			goto unlock_and_done;
2434 
2435 		case TCP_LOGBUF:
2436 			INP_WUNLOCK(inp);
2437 			error = EINVAL;
2438 			break;
2439 
2440 		case TCP_LOGID:
2441 			INP_WUNLOCK(inp);
2442 			error = sooptcopyin(sopt, buf, TCP_LOG_ID_LEN - 1, 0);
2443 			if (error)
2444 				break;
2445 			buf[sopt->sopt_valsize] = '\0';
2446 			INP_WLOCK_RECHECK(inp);
2447 			error = tcp_log_set_id(tp, buf);
2448 			/* tcp_log_set_id() unlocks the INP. */
2449 			break;
2450 
2451 		case TCP_LOGDUMP:
2452 		case TCP_LOGDUMPID:
2453 			INP_WUNLOCK(inp);
2454 			error =
2455 			    sooptcopyin(sopt, buf, TCP_LOG_REASON_LEN - 1, 0);
2456 			if (error)
2457 				break;
2458 			buf[sopt->sopt_valsize] = '\0';
2459 			INP_WLOCK_RECHECK(inp);
2460 			if (sopt->sopt_name == TCP_LOGDUMP) {
2461 				error = tcp_log_dump_tp_logbuf(tp, buf,
2462 				    M_WAITOK, true);
2463 				INP_WUNLOCK(inp);
2464 			} else {
2465 				tcp_log_dump_tp_bucket_logbufs(tp, buf);
2466 				/*
2467 				 * tcp_log_dump_tp_bucket_logbufs() drops the
2468 				 * INP lock.
2469 				 */
2470 			}
2471 			break;
2472 #endif
2473 
2474 		default:
2475 			INP_WUNLOCK(inp);
2476 			error = ENOPROTOOPT;
2477 			break;
2478 		}
2479 		break;
2480 
2481 	case SOPT_GET:
2482 		tp = intotcpcb(inp);
2483 		switch (sopt->sopt_name) {
2484 #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE)
2485 		case TCP_MD5SIG:
2486 			INP_WUNLOCK(inp);
2487 			if (!TCPMD5_ENABLED())
2488 				return (ENOPROTOOPT);
2489 			error = TCPMD5_PCBCTL(inp, sopt);
2490 			break;
2491 #endif
2492 
2493 		case TCP_NODELAY:
2494 			optval = tp->t_flags & TF_NODELAY;
2495 			INP_WUNLOCK(inp);
2496 			error = sooptcopyout(sopt, &optval, sizeof optval);
2497 			break;
2498 		case TCP_MAXSEG:
2499 			optval = tp->t_maxseg;
2500 			INP_WUNLOCK(inp);
2501 			error = sooptcopyout(sopt, &optval, sizeof optval);
2502 			break;
2503 		case TCP_REMOTE_UDP_ENCAPS_PORT:
2504 			optval = ntohs(tp->t_port);
2505 			INP_WUNLOCK(inp);
2506 			error = sooptcopyout(sopt, &optval, sizeof optval);
2507 			break;
2508 		case TCP_NOOPT:
2509 			optval = tp->t_flags & TF_NOOPT;
2510 			INP_WUNLOCK(inp);
2511 			error = sooptcopyout(sopt, &optval, sizeof optval);
2512 			break;
2513 		case TCP_NOPUSH:
2514 			optval = tp->t_flags & TF_NOPUSH;
2515 			INP_WUNLOCK(inp);
2516 			error = sooptcopyout(sopt, &optval, sizeof optval);
2517 			break;
2518 		case TCP_INFO:
2519 			tcp_fill_info(tp, &ti);
2520 			INP_WUNLOCK(inp);
2521 			error = sooptcopyout(sopt, &ti, sizeof ti);
2522 			break;
2523 		case TCP_STATS:
2524 			{
2525 #ifdef STATS
2526 			int nheld;
2527 			TYPEOF_MEMBER(struct statsblob, flags) sbflags = 0;
2528 
2529 			error = 0;
2530 			socklen_t outsbsz = sopt->sopt_valsize;
2531 			if (tp->t_stats == NULL)
2532 				error = ENOENT;
2533 			else if (outsbsz >= tp->t_stats->cursz)
2534 				outsbsz = tp->t_stats->cursz;
2535 			else if (outsbsz >= sizeof(struct statsblob))
2536 				outsbsz = sizeof(struct statsblob);
2537 			else
2538 				error = EINVAL;
2539 			INP_WUNLOCK(inp);
2540 			if (error)
2541 				break;
2542 
2543 			sbp = sopt->sopt_val;
2544 			nheld = atop(round_page(((vm_offset_t)sbp) +
2545 			    (vm_size_t)outsbsz) - trunc_page((vm_offset_t)sbp));
2546 			vm_page_t ma[nheld];
2547 			if (vm_fault_quick_hold_pages(
2548 			    &curproc->p_vmspace->vm_map, (vm_offset_t)sbp,
2549 			    outsbsz, VM_PROT_READ | VM_PROT_WRITE, ma,
2550 			    nheld) < 0) {
2551 				error = EFAULT;
2552 				break;
2553 			}
2554 
2555 			if ((error = copyin_nofault(&(sbp->flags), &sbflags,
2556 			    SIZEOF_MEMBER(struct statsblob, flags))))
2557 				goto unhold;
2558 
2559 			INP_WLOCK_RECHECK(inp);
2560 			error = stats_blob_snapshot(&sbp, outsbsz, tp->t_stats,
2561 			    sbflags | SB_CLONE_USRDSTNOFAULT);
2562 			INP_WUNLOCK(inp);
2563 			sopt->sopt_valsize = outsbsz;
2564 unhold:
2565 			vm_page_unhold_pages(ma, nheld);
2566 #else
2567 			INP_WUNLOCK(inp);
2568 			error = EOPNOTSUPP;
2569 #endif /* !STATS */
2570 			break;
2571 			}
2572 		case TCP_CONGESTION:
2573 			len = strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX);
2574 			INP_WUNLOCK(inp);
2575 			error = sooptcopyout(sopt, buf, len + 1);
2576 			break;
2577 		case TCP_MAXUNACKTIME:
2578 		case TCP_KEEPIDLE:
2579 		case TCP_KEEPINTVL:
2580 		case TCP_KEEPINIT:
2581 		case TCP_KEEPCNT:
2582 			switch (sopt->sopt_name) {
2583 			case TCP_MAXUNACKTIME:
2584 				ui = TP_MAXUNACKTIME(tp) / hz;
2585 				break;
2586 			case TCP_KEEPIDLE:
2587 				ui = TP_KEEPIDLE(tp) / hz;
2588 				break;
2589 			case TCP_KEEPINTVL:
2590 				ui = TP_KEEPINTVL(tp) / hz;
2591 				break;
2592 			case TCP_KEEPINIT:
2593 				ui = TP_KEEPINIT(tp) / hz;
2594 				break;
2595 			case TCP_KEEPCNT:
2596 				ui = TP_KEEPCNT(tp);
2597 				break;
2598 			}
2599 			INP_WUNLOCK(inp);
2600 			error = sooptcopyout(sopt, &ui, sizeof(ui));
2601 			break;
2602 #ifdef TCPPCAP
2603 		case TCP_PCAP_OUT:
2604 		case TCP_PCAP_IN:
2605 			optval = tcp_pcap_get_sock_max(
2606 					(sopt->sopt_name == TCP_PCAP_OUT) ?
2607 					&(tp->t_outpkts) : &(tp->t_inpkts));
2608 			INP_WUNLOCK(inp);
2609 			error = sooptcopyout(sopt, &optval, sizeof optval);
2610 			break;
2611 #endif
2612 		case TCP_FASTOPEN:
2613 			optval = tp->t_flags & TF_FASTOPEN;
2614 			INP_WUNLOCK(inp);
2615 			error = sooptcopyout(sopt, &optval, sizeof optval);
2616 			break;
2617 #ifdef TCP_BLACKBOX
2618 		case TCP_LOG:
2619 			optval = tcp_get_bblog_state(tp);
2620 			INP_WUNLOCK(inp);
2621 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2622 			break;
2623 		case TCP_LOGBUF:
2624 			/* tcp_log_getlogbuf() does INP_WUNLOCK(inp) */
2625 			error = tcp_log_getlogbuf(sopt, tp);
2626 			break;
2627 		case TCP_LOGID:
2628 			len = tcp_log_get_id(tp, buf);
2629 			INP_WUNLOCK(inp);
2630 			error = sooptcopyout(sopt, buf, len + 1);
2631 			break;
2632 		case TCP_LOGDUMP:
2633 		case TCP_LOGDUMPID:
2634 			INP_WUNLOCK(inp);
2635 			error = EINVAL;
2636 			break;
2637 #endif
2638 #ifdef KERN_TLS
2639 		case TCP_TXTLS_MODE:
2640 			error = ktls_get_tx_mode(so, &optval);
2641 			INP_WUNLOCK(inp);
2642 			if (error == 0)
2643 				error = sooptcopyout(sopt, &optval,
2644 				    sizeof(optval));
2645 			break;
2646 		case TCP_RXTLS_MODE:
2647 			error = ktls_get_rx_mode(so, &optval);
2648 			INP_WUNLOCK(inp);
2649 			if (error == 0)
2650 				error = sooptcopyout(sopt, &optval,
2651 				    sizeof(optval));
2652 			break;
2653 #endif
2654 		case TCP_LRD:
2655 			optval = tp->t_flags & TF_LRD;
2656 			INP_WUNLOCK(inp);
2657 			error = sooptcopyout(sopt, &optval, sizeof optval);
2658 			break;
2659 		default:
2660 			INP_WUNLOCK(inp);
2661 			error = ENOPROTOOPT;
2662 			break;
2663 		}
2664 		break;
2665 	}
2666 	return (error);
2667 }
2668 #undef INP_WLOCK_RECHECK
2669 #undef INP_WLOCK_RECHECK_CLEANUP
2670 
2671 /*
2672  * Initiate (or continue) disconnect.
2673  * If embryonic state, just send reset (once).
2674  * If in ``let data drain'' option and linger null, just drop.
2675  * Otherwise (hard), mark socket disconnecting and drop
2676  * current input data; switch states based on user close, and
2677  * send segment to peer (with FIN).
2678  */
2679 static void
tcp_disconnect(struct tcpcb * tp)2680 tcp_disconnect(struct tcpcb *tp)
2681 {
2682 	struct inpcb *inp = tptoinpcb(tp);
2683 	struct socket *so = tptosocket(tp);
2684 
2685 	NET_EPOCH_ASSERT();
2686 	INP_WLOCK_ASSERT(inp);
2687 
2688 	/*
2689 	 * Neither tcp_close() nor tcp_drop() should return NULL, as the
2690 	 * socket is still open.
2691 	 */
2692 	if (tp->t_state < TCPS_ESTABLISHED &&
2693 	    !(tp->t_state > TCPS_LISTEN && IS_FASTOPEN(tp->t_flags))) {
2694 		tp = tcp_close(tp);
2695 		KASSERT(tp != NULL,
2696 		    ("tcp_disconnect: tcp_close() returned NULL"));
2697 	} else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
2698 		tp = tcp_drop(tp, 0);
2699 		KASSERT(tp != NULL,
2700 		    ("tcp_disconnect: tcp_drop() returned NULL"));
2701 	} else {
2702 		soisdisconnecting(so);
2703 		sbflush(&so->so_rcv);
2704 		tcp_usrclosed(tp);
2705 		if (!(inp->inp_flags & INP_DROPPED))
2706 			/* Ignore stack's drop request, we already at it. */
2707 			(void)tcp_output_nodrop(tp);
2708 	}
2709 }
2710 
2711 /*
2712  * User issued close, and wish to trail through shutdown states:
2713  * if never received SYN, just forget it.  If got a SYN from peer,
2714  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
2715  * If already got a FIN from peer, then almost done; go to LAST_ACK
2716  * state.  In all other cases, have already sent FIN to peer (e.g.
2717  * after PRU_SHUTDOWN), and just have to play tedious game waiting
2718  * for peer to send FIN or not respond to keep-alives, etc.
2719  * We can let the user exit from the close as soon as the FIN is acked.
2720  */
2721 static void
tcp_usrclosed(struct tcpcb * tp)2722 tcp_usrclosed(struct tcpcb *tp)
2723 {
2724 
2725 	NET_EPOCH_ASSERT();
2726 	INP_WLOCK_ASSERT(tptoinpcb(tp));
2727 
2728 	switch (tp->t_state) {
2729 	case TCPS_LISTEN:
2730 #ifdef TCP_OFFLOAD
2731 		tcp_offload_listen_stop(tp);
2732 #endif
2733 		tcp_state_change(tp, TCPS_CLOSED);
2734 		/* FALLTHROUGH */
2735 	case TCPS_CLOSED:
2736 		tp = tcp_close(tp);
2737 		/*
2738 		 * tcp_close() should never return NULL here as the socket is
2739 		 * still open.
2740 		 */
2741 		KASSERT(tp != NULL,
2742 		    ("tcp_usrclosed: tcp_close() returned NULL"));
2743 		break;
2744 
2745 	case TCPS_SYN_SENT:
2746 	case TCPS_SYN_RECEIVED:
2747 		tp->t_flags |= TF_NEEDFIN;
2748 		break;
2749 
2750 	case TCPS_ESTABLISHED:
2751 		tcp_state_change(tp, TCPS_FIN_WAIT_1);
2752 		break;
2753 
2754 	case TCPS_CLOSE_WAIT:
2755 		tcp_state_change(tp, TCPS_LAST_ACK);
2756 		break;
2757 	}
2758 	if (tp->t_acktime == 0)
2759 		tp->t_acktime = ticks;
2760 	if (tp->t_state >= TCPS_FIN_WAIT_2) {
2761 		soisdisconnected(tptosocket(tp));
2762 		/* Prevent the connection hanging in FIN_WAIT_2 forever. */
2763 		if (tp->t_state == TCPS_FIN_WAIT_2) {
2764 			int timeout;
2765 
2766 			timeout = (tcp_fast_finwait2_recycle) ?
2767 			    tcp_finwait2_timeout : TP_MAXIDLE(tp);
2768 			tcp_timer_activate(tp, TT_2MSL, timeout);
2769 		}
2770 	}
2771 }
2772 
2773 #ifdef DDB
2774 static void
db_print_indent(int indent)2775 db_print_indent(int indent)
2776 {
2777 	int i;
2778 
2779 	for (i = 0; i < indent; i++)
2780 		db_printf(" ");
2781 }
2782 
2783 static void
db_print_tstate(int t_state)2784 db_print_tstate(int t_state)
2785 {
2786 
2787 	switch (t_state) {
2788 	case TCPS_CLOSED:
2789 		db_printf("TCPS_CLOSED");
2790 		return;
2791 
2792 	case TCPS_LISTEN:
2793 		db_printf("TCPS_LISTEN");
2794 		return;
2795 
2796 	case TCPS_SYN_SENT:
2797 		db_printf("TCPS_SYN_SENT");
2798 		return;
2799 
2800 	case TCPS_SYN_RECEIVED:
2801 		db_printf("TCPS_SYN_RECEIVED");
2802 		return;
2803 
2804 	case TCPS_ESTABLISHED:
2805 		db_printf("TCPS_ESTABLISHED");
2806 		return;
2807 
2808 	case TCPS_CLOSE_WAIT:
2809 		db_printf("TCPS_CLOSE_WAIT");
2810 		return;
2811 
2812 	case TCPS_FIN_WAIT_1:
2813 		db_printf("TCPS_FIN_WAIT_1");
2814 		return;
2815 
2816 	case TCPS_CLOSING:
2817 		db_printf("TCPS_CLOSING");
2818 		return;
2819 
2820 	case TCPS_LAST_ACK:
2821 		db_printf("TCPS_LAST_ACK");
2822 		return;
2823 
2824 	case TCPS_FIN_WAIT_2:
2825 		db_printf("TCPS_FIN_WAIT_2");
2826 		return;
2827 
2828 	case TCPS_TIME_WAIT:
2829 		db_printf("TCPS_TIME_WAIT");
2830 		return;
2831 
2832 	default:
2833 		db_printf("unknown");
2834 		return;
2835 	}
2836 }
2837 
2838 static void
db_print_tflags(u_int t_flags)2839 db_print_tflags(u_int t_flags)
2840 {
2841 	int comma;
2842 
2843 	comma = 0;
2844 	if (t_flags & TF_ACKNOW) {
2845 		db_printf("%sTF_ACKNOW", comma ? ", " : "");
2846 		comma = 1;
2847 	}
2848 	if (t_flags & TF_DELACK) {
2849 		db_printf("%sTF_DELACK", comma ? ", " : "");
2850 		comma = 1;
2851 	}
2852 	if (t_flags & TF_NODELAY) {
2853 		db_printf("%sTF_NODELAY", comma ? ", " : "");
2854 		comma = 1;
2855 	}
2856 	if (t_flags & TF_NOOPT) {
2857 		db_printf("%sTF_NOOPT", comma ? ", " : "");
2858 		comma = 1;
2859 	}
2860 	if (t_flags & TF_SENTFIN) {
2861 		db_printf("%sTF_SENTFIN", comma ? ", " : "");
2862 		comma = 1;
2863 	}
2864 	if (t_flags & TF_REQ_SCALE) {
2865 		db_printf("%sTF_REQ_SCALE", comma ? ", " : "");
2866 		comma = 1;
2867 	}
2868 	if (t_flags & TF_RCVD_SCALE) {
2869 		db_printf("%sTF_RECVD_SCALE", comma ? ", " : "");
2870 		comma = 1;
2871 	}
2872 	if (t_flags & TF_REQ_TSTMP) {
2873 		db_printf("%sTF_REQ_TSTMP", comma ? ", " : "");
2874 		comma = 1;
2875 	}
2876 	if (t_flags & TF_RCVD_TSTMP) {
2877 		db_printf("%sTF_RCVD_TSTMP", comma ? ", " : "");
2878 		comma = 1;
2879 	}
2880 	if (t_flags & TF_SACK_PERMIT) {
2881 		db_printf("%sTF_SACK_PERMIT", comma ? ", " : "");
2882 		comma = 1;
2883 	}
2884 	if (t_flags & TF_NEEDSYN) {
2885 		db_printf("%sTF_NEEDSYN", comma ? ", " : "");
2886 		comma = 1;
2887 	}
2888 	if (t_flags & TF_NEEDFIN) {
2889 		db_printf("%sTF_NEEDFIN", comma ? ", " : "");
2890 		comma = 1;
2891 	}
2892 	if (t_flags & TF_NOPUSH) {
2893 		db_printf("%sTF_NOPUSH", comma ? ", " : "");
2894 		comma = 1;
2895 	}
2896 	if (t_flags & TF_PREVVALID) {
2897 		db_printf("%sTF_PREVVALID", comma ? ", " : "");
2898 		comma = 1;
2899 	}
2900 	if (t_flags & TF_WAKESOR) {
2901 		db_printf("%sTF_WAKESOR", comma ? ", " : "");
2902 		comma = 1;
2903 	}
2904 	if (t_flags & TF_GPUTINPROG) {
2905 		db_printf("%sTF_GPUTINPROG", comma ? ", " : "");
2906 		comma = 1;
2907 	}
2908 	if (t_flags & TF_MORETOCOME) {
2909 		db_printf("%sTF_MORETOCOME", comma ? ", " : "");
2910 		comma = 1;
2911 	}
2912 	if (t_flags & TF_SONOTCONN) {
2913 		db_printf("%sTF_SONOTCONN", comma ? ", " : "");
2914 		comma = 1;
2915 	}
2916 	if (t_flags & TF_LASTIDLE) {
2917 		db_printf("%sTF_LASTIDLE", comma ? ", " : "");
2918 		comma = 1;
2919 	}
2920 	if (t_flags & TF_RXWIN0SENT) {
2921 		db_printf("%sTF_RXWIN0SENT", comma ? ", " : "");
2922 		comma = 1;
2923 	}
2924 	if (t_flags & TF_FASTRECOVERY) {
2925 		db_printf("%sTF_FASTRECOVERY", comma ? ", " : "");
2926 		comma = 1;
2927 	}
2928 	if (t_flags & TF_WASFRECOVERY) {
2929 		db_printf("%sTF_WASFRECOVERY", comma ? ", " : "");
2930 		comma = 1;
2931 	}
2932 	if (t_flags & TF_SIGNATURE) {
2933 		db_printf("%sTF_SIGNATURE", comma ? ", " : "");
2934 		comma = 1;
2935 	}
2936 	if (t_flags & TF_FORCEDATA) {
2937 		db_printf("%sTF_FORCEDATA", comma ? ", " : "");
2938 		comma = 1;
2939 	}
2940 	if (t_flags & TF_TSO) {
2941 		db_printf("%sTF_TSO", comma ? ", " : "");
2942 		comma = 1;
2943 	}
2944 	if (t_flags & TF_TOE) {
2945 		db_printf("%sTF_TOE", comma ? ", " : "");
2946 		comma = 1;
2947 	}
2948 	if (t_flags & TF_CLOSED) {
2949 		db_printf("%sTF_CLOSED", comma ? ", " : "");
2950 		comma = 1;
2951 	}
2952 	if (t_flags & TF_UNUSED1) {
2953 		db_printf("%sTF_UNUSED1", comma ? ", " : "");
2954 		comma = 1;
2955 	}
2956 	if (t_flags & TF_LRD) {
2957 		db_printf("%sTF_LRD", comma ? ", " : "");
2958 		comma = 1;
2959 	}
2960 	if (t_flags & TF_CONGRECOVERY) {
2961 		db_printf("%sTF_CONGRECOVERY", comma ? ", " : "");
2962 		comma = 1;
2963 	}
2964 	if (t_flags & TF_WASCRECOVERY) {
2965 		db_printf("%sTF_WASCRECOVERY", comma ? ", " : "");
2966 		comma = 1;
2967 	}
2968 	if (t_flags & TF_FASTOPEN) {
2969 		db_printf("%sTF_FASTOPEN", comma ? ", " : "");
2970 		comma = 1;
2971 	}
2972 }
2973 
2974 static void
db_print_tflags2(u_int t_flags2)2975 db_print_tflags2(u_int t_flags2)
2976 {
2977 	int comma;
2978 
2979 	comma = 0;
2980 	if (t_flags2 & TF2_PLPMTU_BLACKHOLE) {
2981 		db_printf("%sTF2_PLPMTU_BLACKHOLE", comma ? ", " : "");
2982 		comma = 1;
2983 	}
2984 	if (t_flags2 & TF2_PLPMTU_PMTUD) {
2985 		db_printf("%sTF2_PLPMTU_PMTUD", comma ? ", " : "");
2986 		comma = 1;
2987 	}
2988 	if (t_flags2 & TF2_PLPMTU_MAXSEGSNT) {
2989 		db_printf("%sTF2_PLPMTU_MAXSEGSNT", comma ? ", " : "");
2990 		comma = 1;
2991 	}
2992 	if (t_flags2 & TF2_LOG_AUTO) {
2993 		db_printf("%sTF2_LOG_AUTO", comma ? ", " : "");
2994 		comma = 1;
2995 	}
2996 	if (t_flags2 & TF2_DROP_AF_DATA) {
2997 		db_printf("%sTF2_DROP_AF_DATA", comma ? ", " : "");
2998 		comma = 1;
2999 	}
3000 	if (t_flags2 & TF2_ECN_PERMIT) {
3001 		db_printf("%sTF2_ECN_PERMIT", comma ? ", " : "");
3002 		comma = 1;
3003 	}
3004 	if (t_flags2 & TF2_ECN_SND_CWR) {
3005 		db_printf("%sTF2_ECN_SND_CWR", comma ? ", " : "");
3006 		comma = 1;
3007 	}
3008 	if (t_flags2 & TF2_ECN_SND_ECE) {
3009 		db_printf("%sTF2_ECN_SND_ECE", comma ? ", " : "");
3010 		comma = 1;
3011 	}
3012 	if (t_flags2 & TF2_ACE_PERMIT) {
3013 		db_printf("%sTF2_ACE_PERMIT", comma ? ", " : "");
3014 		comma = 1;
3015 	}
3016 	if (t_flags2 & TF2_HPTS_CPU_SET) {
3017 		db_printf("%sTF2_HPTS_CPU_SET", comma ? ", " : "");
3018 		comma = 1;
3019 	}
3020 	if (t_flags2 & TF2_FBYTES_COMPLETE) {
3021 		db_printf("%sTF2_FBYTES_COMPLETE", comma ? ", " : "");
3022 		comma = 1;
3023 	}
3024 	if (t_flags2 & TF2_ECN_USE_ECT1) {
3025 		db_printf("%sTF2_ECN_USE_ECT1", comma ? ", " : "");
3026 		comma = 1;
3027 	}
3028 	if (t_flags2 & TF2_TCP_ACCOUNTING) {
3029 		db_printf("%sTF2_TCP_ACCOUNTING", comma ? ", " : "");
3030 		comma = 1;
3031 	}
3032 	if (t_flags2 & TF2_HPTS_CALLS) {
3033 		db_printf("%sTF2_HPTS_CALLS", comma ? ", " : "");
3034 		comma = 1;
3035 	}
3036 	if (t_flags2 & TF2_MBUF_L_ACKS) {
3037 		db_printf("%sTF2_MBUF_L_ACKS", comma ? ", " : "");
3038 		comma = 1;
3039 	}
3040 	if (t_flags2 & TF2_MBUF_ACKCMP) {
3041 		db_printf("%sTF2_MBUF_ACKCMP", comma ? ", " : "");
3042 		comma = 1;
3043 	}
3044 	if (t_flags2 & TF2_SUPPORTS_MBUFQ) {
3045 		db_printf("%sTF2_SUPPORTS_MBUFQ", comma ? ", " : "");
3046 		comma = 1;
3047 	}
3048 	if (t_flags2 & TF2_MBUF_QUEUE_READY) {
3049 		db_printf("%sTF2_MBUF_QUEUE_READY", comma ? ", " : "");
3050 		comma = 1;
3051 	}
3052 	if (t_flags2 & TF2_DONT_SACK_QUEUE) {
3053 		db_printf("%sTF2_DONT_SACK_QUEUE", comma ? ", " : "");
3054 		comma = 1;
3055 	}
3056 	if (t_flags2 & TF2_CANNOT_DO_ECN) {
3057 		db_printf("%sTF2_CANNOT_DO_ECN", comma ? ", " : "");
3058 		comma = 1;
3059 	}
3060 	if (t_flags2 & TF2_NO_ISS_CHECK) {
3061 		db_printf("%sTF2_NO_ISS_CHECK", comma ? ", " : "");
3062 		comma = 1;
3063 	}
3064 }
3065 
3066 static void
db_print_toobflags(char t_oobflags)3067 db_print_toobflags(char t_oobflags)
3068 {
3069 	int comma;
3070 
3071 	comma = 0;
3072 	if (t_oobflags & TCPOOB_HAVEDATA) {
3073 		db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : "");
3074 		comma = 1;
3075 	}
3076 	if (t_oobflags & TCPOOB_HADDATA) {
3077 		db_printf("%sTCPOOB_HADDATA", comma ? ", " : "");
3078 		comma = 1;
3079 	}
3080 }
3081 
3082 static void
db_print_tcpcb(struct tcpcb * tp,const char * name,int indent)3083 db_print_tcpcb(struct tcpcb *tp, const char *name, int indent)
3084 {
3085 
3086 	db_print_indent(indent);
3087 	db_printf("%s at %p\n", name, tp);
3088 
3089 	indent += 2;
3090 
3091 	db_print_indent(indent);
3092 	db_printf("t_segq first: %p   t_segqlen: %d   t_dupacks: %d\n",
3093 	   TAILQ_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks);
3094 
3095 	db_print_indent(indent);
3096 	db_printf("t_callout: %p   t_timers: %p\n",
3097 	    &tp->t_callout, &tp->t_timers);
3098 
3099 	db_print_indent(indent);
3100 	db_printf("t_state: %d (", tp->t_state);
3101 	db_print_tstate(tp->t_state);
3102 	db_printf(")\n");
3103 
3104 	db_print_indent(indent);
3105 	db_printf("t_flags: 0x%x (", tp->t_flags);
3106 	db_print_tflags(tp->t_flags);
3107 	db_printf(")\n");
3108 
3109 	db_print_indent(indent);
3110 	db_printf("t_flags2: 0x%x (", tp->t_flags2);
3111 	db_print_tflags2(tp->t_flags2);
3112 	db_printf(")\n");
3113 
3114 	db_print_indent(indent);
3115 	db_printf("snd_una: 0x%08x   snd_max: 0x%08x   snd_nxt: 0x%08x\n",
3116 	    tp->snd_una, tp->snd_max, tp->snd_nxt);
3117 
3118 	db_print_indent(indent);
3119 	db_printf("snd_up: 0x%08x   snd_wl1: 0x%08x   snd_wl2: 0x%08x\n",
3120 	   tp->snd_up, tp->snd_wl1, tp->snd_wl2);
3121 
3122 	db_print_indent(indent);
3123 	db_printf("iss: 0x%08x   irs: 0x%08x   rcv_nxt: 0x%08x\n",
3124 	    tp->iss, tp->irs, tp->rcv_nxt);
3125 
3126 	db_print_indent(indent);
3127 	db_printf("rcv_adv: 0x%08x   rcv_wnd: %u   rcv_up: 0x%08x\n",
3128 	    tp->rcv_adv, tp->rcv_wnd, tp->rcv_up);
3129 
3130 	db_print_indent(indent);
3131 	db_printf("snd_wnd: %u   snd_cwnd: %u\n",
3132 	   tp->snd_wnd, tp->snd_cwnd);
3133 
3134 	db_print_indent(indent);
3135 	db_printf("snd_ssthresh: %u   snd_recover: "
3136 	    "0x%08x\n", tp->snd_ssthresh, tp->snd_recover);
3137 
3138 	db_print_indent(indent);
3139 	db_printf("t_rcvtime: %u   t_startime: %u\n",
3140 	    tp->t_rcvtime, tp->t_starttime);
3141 
3142 	db_print_indent(indent);
3143 	db_printf("t_rttime: %u   t_rtsq: 0x%08x\n",
3144 	    tp->t_rtttime, tp->t_rtseq);
3145 
3146 	db_print_indent(indent);
3147 	db_printf("t_rxtcur: %d   t_maxseg: %u   t_srtt: %d\n",
3148 	    tp->t_rxtcur, tp->t_maxseg, tp->t_srtt);
3149 
3150 	db_print_indent(indent);
3151 	db_printf("t_rttvar: %d   t_rxtshift: %d   t_rttmin: %u\n",
3152 	    tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin);
3153 
3154 	db_print_indent(indent);
3155 	db_printf("t_rttupdated: %u   max_sndwnd: %u   t_softerror: %d\n",
3156 	    tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror);
3157 
3158 	db_print_indent(indent);
3159 	db_printf("t_oobflags: 0x%x (", tp->t_oobflags);
3160 	db_print_toobflags(tp->t_oobflags);
3161 	db_printf(")   t_iobc: 0x%02x\n", tp->t_iobc);
3162 
3163 	db_print_indent(indent);
3164 	db_printf("snd_scale: %u   rcv_scale: %u   request_r_scale: %u\n",
3165 	    tp->snd_scale, tp->rcv_scale, tp->request_r_scale);
3166 
3167 	db_print_indent(indent);
3168 	db_printf("ts_recent: %u   ts_recent_age: %u\n",
3169 	    tp->ts_recent, tp->ts_recent_age);
3170 
3171 	db_print_indent(indent);
3172 	db_printf("ts_offset: %u   last_ack_sent: 0x%08x   snd_cwnd_prev: "
3173 	    "%u\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev);
3174 
3175 	db_print_indent(indent);
3176 	db_printf("snd_ssthresh_prev: %u   snd_recover_prev: 0x%08x   "
3177 	    "t_badrxtwin: %u\n", tp->snd_ssthresh_prev,
3178 	    tp->snd_recover_prev, tp->t_badrxtwin);
3179 
3180 	db_print_indent(indent);
3181 	db_printf("snd_numholes: %d  snd_holes first: %p\n",
3182 	    tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes));
3183 
3184 	db_print_indent(indent);
3185 	db_printf("snd_fack: 0x%08x   rcv_numsacks: %d\n",
3186 	    tp->snd_fack, tp->rcv_numsacks);
3187 
3188 	/* Skip sackblks, sackhint. */
3189 
3190 	db_print_indent(indent);
3191 	db_printf("t_rttlow: %d   rfbuf_ts: %u   rfbuf_cnt: %d\n",
3192 	    tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt);
3193 }
3194 
DB_SHOW_COMMAND(tcpcb,db_show_tcpcb)3195 DB_SHOW_COMMAND(tcpcb, db_show_tcpcb)
3196 {
3197 	struct tcpcb *tp;
3198 
3199 	if (!have_addr) {
3200 		db_printf("usage: show tcpcb <addr>\n");
3201 		return;
3202 	}
3203 	tp = (struct tcpcb *)addr;
3204 
3205 	db_print_tcpcb(tp, "tcpcb", 0);
3206 }
3207 #endif
3208