xref: /xnu-10063.101.15/bsd/netinet/tcp_usrreq.c (revision 94d3b452840153a99b38a3a9659680b2a006908e) !
1 /*
2  * Copyright (c) 2000-2023 Apple Inc. All rights reserved.
3  *
4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5  *
6  * This file contains Original Code and/or Modifications of Original Code
7  * as defined in and that are subject to the Apple Public Source License
8  * Version 2.0 (the 'License'). You may not use this file except in
9  * compliance with the License. The rights granted to you under the License
10  * may not be used to create, or enable the creation or redistribution of,
11  * unlawful or unlicensed copies of an Apple operating system, or to
12  * circumvent, violate, or enable the circumvention or violation of, any
13  * terms of an Apple operating system software license agreement.
14  *
15  * Please obtain a copy of the License at
16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
17  *
18  * The Original Code and all software distributed under the License are
19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23  * Please see the License for the specific language governing rights and
24  * limitations under the License.
25  *
26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27  */
28 /*
29  * Copyright (c) 1982, 1986, 1988, 1993
30  *	The Regents of the University of California.  All rights reserved.
31  *
32  * Redistribution and use in source and binary forms, with or without
33  * modification, are permitted provided that the following conditions
34  * are met:
35  * 1. Redistributions of source code must retain the above copyright
36  *    notice, this list of conditions and the following disclaimer.
37  * 2. Redistributions in binary form must reproduce the above copyright
38  *    notice, this list of conditions and the following disclaimer in the
39  *    documentation and/or other materials provided with the distribution.
40  * 3. All advertising materials mentioning features or use of this software
41  *    must display the following acknowledgement:
42  *	This product includes software developed by the University of
43  *	California, Berkeley and its contributors.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	From: @(#)tcp_usrreq.c	8.2 (Berkeley) 1/3/94
61  * $FreeBSD: src/sys/netinet/tcp_usrreq.c,v 1.51.2.9 2001/08/22 00:59:12 silby Exp $
62  */
63 
64 
65 #include <sys/param.h>
66 #include <sys/systm.h>
67 #include <sys/kernel.h>
68 #include <sys/sysctl.h>
69 #include <sys/mbuf.h>
70 #include <sys/domain.h>
71 #include <sys/priv.h>
72 #include <sys/socket.h>
73 #include <sys/socketvar.h>
74 #include <sys/protosw.h>
75 #include <sys/syslog.h>
76 
77 #include <net/if.h>
78 #include <net/route.h>
79 #include <net/ntstat.h>
80 #include <net/content_filter.h>
81 #include <net/multi_layer_pkt_log.h>
82 
83 #include <netinet/in.h>
84 #include <netinet/in_systm.h>
85 #include <netinet/ip6.h>
86 #include <netinet/in_pcb.h>
87 #include <netinet6/in6_pcb.h>
88 #include <netinet/in_var.h>
89 #include <netinet/ip_var.h>
90 #include <netinet6/ip6_var.h>
91 #include <netinet/tcp.h>
92 #include <netinet/tcp_fsm.h>
93 #include <netinet/tcp_seq.h>
94 #include <netinet/tcp_timer.h>
95 #include <netinet/tcp_var.h>
96 #include <netinet/tcpip.h>
97 #include <netinet/tcp_cc.h>
98 #include <netinet/tcp_log.h>
99 #include <mach/sdt.h>
100 #if TCPDEBUG
101 #include <netinet/tcp_debug.h>
102 #endif
103 #if MPTCP
104 #include <netinet/mptcp_var.h>
105 #endif /* MPTCP */
106 
107 #if IPSEC
108 #include <netinet6/ipsec.h>
109 #endif /*IPSEC*/
110 
111 #if FLOW_DIVERT
112 #include <netinet/flow_divert.h>
113 #endif /* FLOW_DIVERT */
114 
115 #if SKYWALK
116 #include <libkern/sysctl.h>
117 #include <skywalk/os_stats_private.h>
118 #endif /* SKYWALK */
119 
120 #include <net/sockaddr_utils.h>
121 
122 extern char *proc_name_address(void *p);
123 
124 errno_t tcp_fill_info_for_info_tuple(struct info_tuple *, struct tcp_info *);
125 
126 static int tcp_sysctl_info(struct sysctl_oid *, void *, int, struct sysctl_req *);
127 static void tcp_connection_fill_info(struct tcpcb *tp,
128     struct tcp_connection_info *tci);
129 static int tcp_get_mpkl_send_info(struct mbuf *, struct so_mpkl_send_info *);
130 
131 /*
132  * TCP protocol interface to socket abstraction.
133  */
134 static int      tcp_attach(struct socket *, struct proc *);
135 static int      tcp_connect(struct tcpcb *, struct sockaddr *, struct proc *);
136 static int      tcp6_connect(struct tcpcb *, struct sockaddr *, struct proc *);
137 static int      tcp6_usr_connect(struct socket *, struct sockaddr *,
138     struct proc *);
139 static struct tcpcb *tcp_disconnect(struct tcpcb *);
140 static struct tcpcb *tcp_usrclosed(struct tcpcb *);
141 extern void tcp_sbrcv_trim(struct tcpcb *tp, struct sockbuf *sb);
142 
143 #if TCPDEBUG
144 #define TCPDEBUG0       int ostate = 0
145 #define TCPDEBUG1()     ostate = tp ? tp->t_state : 0
146 #define TCPDEBUG2(req)  if (tp && (so->so_options & SO_DEBUG)) \
147 	                        tcp_trace(TA_USER, ostate, tp, 0, 0, req)
148 #else
149 #define TCPDEBUG0
150 #define TCPDEBUG1()
151 #define TCPDEBUG2(req)
152 #endif
153 
154 SYSCTL_PROC(_net_inet_tcp, OID_AUTO, info,
155     CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_ANYBODY | CTLFLAG_KERN,
156     0, 0, tcp_sysctl_info, "S", "TCP info per tuple");
157 
158 int faster_mcopy = 1;
159 SYSCTL_INT(_net_inet_tcp, OID_AUTO, faster_mcopy,
160     CTLFLAG_RW | CTLFLAG_LOCKED, &faster_mcopy, 1,
161     "Speed up m_copym");
162 
163 /*
164  * TCP attaches to socket via pru_attach(), reserving space,
165  * and an internet control block.
166  *
167  * Returns:	0			Success
168  *		EISCONN
169  *	tcp_attach:ENOBUFS
170  *	tcp_attach:ENOMEM
171  *	tcp_attach:???			[IPSEC specific]
172  */
173 static int
tcp_usr_attach(struct socket * so,__unused int proto,struct proc * p)174 tcp_usr_attach(struct socket *so, __unused int proto, struct proc *p)
175 {
176 	int error;
177 	struct inpcb *inp = sotoinpcb(so);
178 	struct tcpcb *tp = 0;
179 	TCPDEBUG0;
180 
181 	TCPDEBUG1();
182 	if (inp) {
183 		error = EISCONN;
184 		goto out;
185 	}
186 
187 	error = tcp_attach(so, p);
188 	if (error) {
189 		goto out;
190 	}
191 
192 	if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
193 		so->so_linger = (short)(TCP_LINGERTIME * hz);
194 	}
195 	if (faster_mcopy) {
196 		so->so_snd.sb_flags |= SB_SENDHEAD;
197 		so->so_snd.sb_sendhead = NULL;
198 		so->so_snd.sb_sendoff = 0;
199 	}
200 
201 	tp = sototcpcb(so);
202 out:
203 	TCPDEBUG2(PRU_ATTACH);
204 	return error;
205 }
206 
207 /*
208  * pru_detach() detaches the TCP protocol from the socket.
209  * If the protocol state is non-embryonic, then can't
210  * do this directly: have to initiate a pru_disconnect(),
211  * which may finish later; embryonic TCB's can just
212  * be discarded here.
213  */
214 static int
tcp_usr_detach(struct socket * so)215 tcp_usr_detach(struct socket *so)
216 {
217 	int error = 0;
218 	struct inpcb *inp = sotoinpcb(so);
219 	struct tcpcb *tp;
220 	TCPDEBUG0;
221 
222 	if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) {
223 		return EINVAL;  /* XXX */
224 	}
225 	socket_lock_assert_owned(so);
226 	tp = intotcpcb(inp);
227 	/* In case we got disconnected from the peer */
228 	if (tp == NULL) {
229 		goto out;
230 	}
231 	TCPDEBUG1();
232 
233 	calculate_tcp_clock();
234 
235 	tp = tcp_disconnect(tp);
236 out:
237 	TCPDEBUG2(PRU_DETACH);
238 	return error;
239 }
240 
241 #if NECP
242 #define COMMON_START_ALLOW_FLOW_DIVERT(allow)  TCPDEBUG0;               \
243 do {                                                                    \
244 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD)          \
245 	        return (EINVAL);                                        \
246 	if (!(allow) && necp_socket_should_use_flow_divert(inp))        \
247 	        return (EPROTOTYPE);                                    \
248 	tp = intotcpcb(inp);                                            \
249 	TCPDEBUG1();                                                    \
250 	calculate_tcp_clock();                                          \
251 } while (0)
252 #else /* NECP */
253 #define COMMON_START_ALLOW_FLOW_DIVERT(allow)  TCPDEBUG0;               \
254 do {                                                                    \
255 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD)          \
256 	        return (EINVAL);                                        \
257 	tp = intotcpcb(inp);                                            \
258 	TCPDEBUG1();                                                    \
259 	calculate_tcp_clock();                                          \
260 } while (0)
261 #endif /* !NECP */
262 
263 #define COMMON_START() COMMON_START_ALLOW_FLOW_DIVERT(false)
264 #define COMMON_END(req) out: TCPDEBUG2(req); return error; goto out
265 
266 
267 /*
268  * Give the socket an address.
269  *
270  * Returns:	0			Success
271  *		EINVAL			Invalid argument [COMMON_START]
272  *		EAFNOSUPPORT		Address family not supported
273  *	in_pcbbind:EADDRNOTAVAIL	Address not available.
274  *	in_pcbbind:EINVAL		Invalid argument
275  *	in_pcbbind:EAFNOSUPPORT		Address family not supported [notdef]
276  *	in_pcbbind:EACCES		Permission denied
277  *	in_pcbbind:EADDRINUSE		Address in use
278  *	in_pcbbind:EAGAIN		Resource unavailable, try again
279  *	in_pcbbind:EPERM		Operation not permitted
280  */
281 static int
tcp_usr_bind(struct socket * so,struct sockaddr * nam,struct proc * p)282 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
283 {
284 	int error = 0;
285 	struct inpcb *inp = sotoinpcb(so);
286 	struct tcpcb *tp;
287 	struct sockaddr_in *sinp;
288 
289 	COMMON_START_ALLOW_FLOW_DIVERT(true);
290 
291 	if (nam->sa_family != 0 && nam->sa_family != AF_INET) {
292 		error = EAFNOSUPPORT;
293 		goto out;
294 	}
295 	/*
296 	 * Must check for multicast and broadcast addresses and disallow binding
297 	 * to them.
298 	 */
299 	sinp = SIN(nam);
300 	if (sinp->sin_family == AF_INET &&
301 	    (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)) ||
302 	    sinp->sin_addr.s_addr == INADDR_BROADCAST)) {
303 		error = EAFNOSUPPORT;
304 		goto out;
305 	}
306 
307 	error = in_pcbbind(inp, nam, p);
308 	if (error) {
309 		goto out;
310 	}
311 
312 #if NECP
313 	/* Update NECP client with bind result if not in middle of connect */
314 	if ((inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS) &&
315 	    !uuid_is_null(inp->necp_client_uuid)) {
316 		socket_unlock(so, 0);
317 		necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
318 		socket_lock(so, 0);
319 	}
320 #endif /* NECP */
321 
322 	COMMON_END(PRU_BIND);
323 }
324 
325 static int
tcp6_usr_bind(struct socket * so,struct sockaddr * nam,struct proc * p)326 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
327 {
328 	int error = 0;
329 	struct inpcb *inp = sotoinpcb(so);
330 	const uint8_t old_flags = inp->inp_vflag;
331 	struct tcpcb *tp;
332 	struct sockaddr_in6 *sin6p;
333 
334 	COMMON_START_ALLOW_FLOW_DIVERT(true);
335 
336 	if (nam->sa_family != 0 && nam->sa_family != AF_INET6) {
337 		error = EAFNOSUPPORT;
338 		goto out;
339 	}
340 	/*
341 	 * Must check for multicast and broadcast addresses and disallow binding
342 	 * to them.
343 	 */
344 	sin6p = SIN6(nam);
345 	if (sin6p->sin6_family == AF_INET6 &&
346 	    (IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr) ||
347 	    ((IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr) ||
348 	    IN6_IS_ADDR_V4COMPAT(&sin6p->sin6_addr)) &&
349 	    (IN_MULTICAST(ntohl(sin6p->sin6_addr.s6_addr32[3])) ||
350 	    sin6p->sin6_addr.s6_addr32[3] == INADDR_BROADCAST)))) {
351 		error = EAFNOSUPPORT;
352 		goto out;
353 	}
354 
355 	/*
356 	 * Another thread won the binding race so do not change inp_vflag
357 	 */
358 	if (inp->inp_flags2 & INP2_BIND_IN_PROGRESS) {
359 		error = EINVAL;
360 		goto out;
361 	}
362 
363 	inp->inp_vflag &= ~INP_IPV4;
364 	inp->inp_vflag |= INP_IPV6;
365 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
366 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) {
367 			inp->inp_vflag |= INP_IPV4;
368 		} else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
369 			struct sockaddr_in sin;
370 
371 			in6_sin6_2_sin(&sin, sin6p);
372 			inp->inp_vflag |= INP_IPV4;
373 			inp->inp_vflag &= ~INP_IPV6;
374 
375 			error = in_pcbbind(inp, SA(&sin), p);
376 			if (error != 0) {
377 				inp->inp_vflag = old_flags;
378 				route_clear(&inp->inp_route);
379 			}
380 			goto out;
381 		}
382 	}
383 	error = in6_pcbbind(inp, nam, p);
384 	if (error) {
385 		inp->inp_vflag = old_flags;
386 		route_clear(&inp->inp_route);
387 		goto out;
388 	}
389 
390 	COMMON_END(PRU_BIND);
391 }
392 
393 /*
394  * Prepare to accept connections.
395  *
396  * Returns:	0			Success
397  *		EINVAL [COMMON_START]
398  *	in_pcbbind:EADDRNOTAVAIL	Address not available.
399  *	in_pcbbind:EINVAL		Invalid argument
400  *	in_pcbbind:EAFNOSUPPORT		Address family not supported [notdef]
401  *	in_pcbbind:EACCES		Permission denied
402  *	in_pcbbind:EADDRINUSE		Address in use
403  *	in_pcbbind:EAGAIN		Resource unavailable, try again
404  *	in_pcbbind:EPERM		Operation not permitted
405  */
406 static int
tcp_usr_listen(struct socket * so,struct proc * p)407 tcp_usr_listen(struct socket *so, struct proc *p)
408 {
409 	int error = 0;
410 	struct inpcb *inp = sotoinpcb(so);
411 	struct tcpcb *tp;
412 
413 	COMMON_START_ALLOW_FLOW_DIVERT(true);
414 	if (inp->inp_lport == 0) {
415 		error = in_pcbbind(inp, NULL, p);
416 	}
417 	if (error == 0) {
418 		TCP_LOG_STATE(tp, TCPS_LISTEN);
419 		tp->t_state = TCPS_LISTEN;
420 		if (nstat_collect) {
421 			nstat_pcb_event(inp, NSTAT_EVENT_SRC_FLOW_STATE_LISTEN);
422 		}
423 	}
424 	TCP_LOG_LISTEN(tp, error);
425 	COMMON_END(PRU_LISTEN);
426 }
427 
428 static int
tcp6_usr_listen(struct socket * so,struct proc * p)429 tcp6_usr_listen(struct socket *so, struct proc *p)
430 {
431 	int error = 0;
432 	struct inpcb *inp = sotoinpcb(so);
433 	struct tcpcb *tp;
434 
435 	COMMON_START_ALLOW_FLOW_DIVERT(true);
436 	if (inp->inp_lport == 0) {
437 		inp->inp_vflag &= ~INP_IPV4;
438 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
439 			inp->inp_vflag |= INP_IPV4;
440 		}
441 		error = in6_pcbbind(inp, NULL, p);
442 	}
443 	if (error == 0) {
444 		TCP_LOG_STATE(tp, TCPS_LISTEN);
445 		tp->t_state = TCPS_LISTEN;
446 		if (nstat_collect) {
447 			nstat_pcb_event(inp, NSTAT_EVENT_SRC_FLOW_STATE_LISTEN);
448 		}
449 	}
450 	TCP_LOG_LISTEN(tp, error);
451 	COMMON_END(PRU_LISTEN);
452 }
453 
454 static int
tcp_connect_complete(struct socket * so)455 tcp_connect_complete(struct socket *so)
456 {
457 	struct tcpcb *tp = sototcpcb(so);
458 	struct inpcb *inp = sotoinpcb(so);
459 	int error = 0;
460 
461 	/* TFO delays the tcp_output until later, when the app calls write() */
462 	if (so->so_flags1 & SOF1_PRECONNECT_DATA) {
463 		if (!necp_socket_is_allowed_to_send_recv(sotoinpcb(so), NULL, 0, NULL, NULL, NULL, NULL)) {
464 			TCP_LOG_DROP_NECP(NULL, NULL, tp, true);
465 			return EHOSTUNREACH;
466 		}
467 
468 		/* Initialize enough state so that we can actually send data */
469 		tcp_mss(tp, -1, IFSCOPE_NONE);
470 		tp->snd_wnd = tp->t_maxseg;
471 		tp->max_sndwnd = tp->snd_wnd;
472 	} else {
473 		tp->t_flagsext |= TF_USR_OUTPUT;
474 		error = tcp_output(tp);
475 		tp->t_flagsext &= ~TF_USR_OUTPUT;
476 	}
477 
478 #if NECP
479 	/* Update NECP client with connected five-tuple */
480 	if (error == 0 && !uuid_is_null(inp->necp_client_uuid)) {
481 		socket_unlock(so, 0);
482 		necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
483 		socket_lock(so, 0);
484 	}
485 #endif /* NECP */
486 
487 	return error;
488 }
489 
490 __attribute__((noinline))
491 static void
tcp_log_address_error(int error,struct sockaddr * nam,struct proc * p)492 tcp_log_address_error(int error, struct sockaddr *nam, struct proc *p)
493 {
494 	char buffer[MAX_IPv6_STR_LEN];
495 
496 	if (nam->sa_family == AF_INET6) {
497 		struct sockaddr_in6 *sin6p = SIN6(nam);
498 
499 		inet_ntop(AF_INET6, &sin6p->sin6_addr, buffer, sizeof(buffer));
500 	} else {
501 		struct sockaddr_in *sinp = SIN(nam);
502 
503 		inet_ntop(AF_INET, &sinp->sin_addr, buffer, sizeof(buffer));
504 	}
505 	if (p == NULL) {
506 		p = current_proc();
507 	}
508 	os_log(OS_LOG_DEFAULT, "connect address error %d for %s process %s:%u",
509 	    error, buffer, proc_name_address(p), proc_pid(p));
510 }
511 
512 /*
513  * Note that connecting to the all-zeros address is OK and is treated as the
514  * loopback address
515  */
516 static int
tcp_usr_connect_common(struct socket * so,struct tcpcb * tp,struct sockaddr * nam,struct proc * p,bool isipv6,bool need_connect_complete)517 tcp_usr_connect_common(struct socket *so, struct tcpcb *tp, struct sockaddr *nam,
518     struct proc *p, bool isipv6, bool need_connect_complete)
519 {
520 	int error = 0;
521 	struct inpcb *inp = sotoinpcb(so);
522 
523 	if (isipv6 == 0) {
524 		struct sockaddr_in *sinp;
525 
526 		if (nam->sa_family != 0 && nam->sa_family != AF_INET) {
527 			error = EAFNOSUPPORT;
528 			goto out;
529 		}
530 		/*
531 		 * Disallow connecting to multicast and broadcast addresses.
532 		 */
533 		sinp = SIN(nam);
534 		if (sinp->sin_family == AF_INET &&
535 		    (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)) ||
536 		    sinp->sin_addr.s_addr == INADDR_BROADCAST)) {
537 			error = EAFNOSUPPORT;
538 			goto out;
539 		}
540 
541 		if ((error = tcp_connect(tp, nam, p)) != 0) {
542 			goto out;
543 		}
544 	} else {
545 		struct sockaddr_in6 *sin6p;
546 
547 		if (nam->sa_family != 0 && nam->sa_family != AF_INET6) {
548 			error = EAFNOSUPPORT;
549 			goto out;
550 		}
551 
552 		/*
553 		 * Disallow connecting to multicast and broadcast addresses.
554 		 */
555 		sin6p = SIN6(nam);
556 		if (sin6p->sin6_family == AF_INET6 &&
557 		    IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
558 			error = EAFNOSUPPORT;
559 			goto out;
560 		}
561 
562 		if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
563 			struct sockaddr_in sin;
564 
565 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
566 				error = EINVAL;
567 				goto out;
568 			}
569 
570 			in6_sin6_2_sin(&sin, sin6p);
571 			/*
572 			 * Disallow connecting to multicast and broadcast addresses.
573 			 */
574 			if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr)) ||
575 			    sin.sin_addr.s_addr == INADDR_BROADCAST) {
576 				error = EAFNOSUPPORT;
577 				goto out;
578 			}
579 			inp->inp_vflag |= INP_IPV4;
580 			inp->inp_vflag &= ~INP_IPV6;
581 			if ((error = tcp_connect(tp, SA(&sin), p)) != 0) {
582 				goto out;
583 			}
584 
585 			goto out;
586 		} else if (IN6_IS_ADDR_V4COMPAT(&sin6p->sin6_addr)) {
587 			/*
588 			 * Disallow connecting to multicast and broadcast addresses.
589 			 */
590 			if (IN_MULTICAST(ntohl(sin6p->sin6_addr.s6_addr32[3])) ||
591 			    sin6p->sin6_addr.s6_addr32[3] == INADDR_BROADCAST) {
592 				error = EAFNOSUPPORT;
593 				goto out;
594 			}
595 		}
596 
597 		inp->inp_vflag &= ~INP_IPV4;
598 		inp->inp_vflag |= INP_IPV6;
599 		if ((error = tcp6_connect(tp, nam, p)) != 0) {
600 			goto out;
601 		}
602 	}
603 out:
604 	if (need_connect_complete && error == 0) {
605 		error = tcp_connect_complete(so);
606 	}
607 	TCP_LOG_CONNECT(tp, true, error);
608 	if (error == EAFNOSUPPORT) {
609 		tcp_log_address_error(error, nam, p);
610 	}
611 	return error;
612 }
613 
614 /*
615  * Initiate connection to peer.
616  * Create a template for use in transmissions on this connection.
617  * Enter SYN_SENT state, and mark socket as connecting.
618  * Start keep-alive timer, and seed output sequence space.
619  * Send initial segment on connection.
620  */
621 static int
tcp_usr_connect(struct socket * so,struct sockaddr * nam,struct proc * p)622 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
623 {
624 	int error = 0;
625 	struct inpcb *inp = sotoinpcb(so);
626 	struct tcpcb *tp;
627 
628 	TCPDEBUG0;
629 	if (inp == NULL) {
630 		return EINVAL;
631 	} else if (inp->inp_state == INPCB_STATE_DEAD) {
632 		if (so->so_error) {
633 			error = so->so_error;
634 			so->so_error = 0;
635 			return error;
636 		} else {
637 			return EINVAL;
638 		}
639 	}
640 #if NECP
641 #if CONTENT_FILTER
642 	error = cfil_sock_attach(so, NULL, nam, CFS_CONNECTION_DIR_OUT);
643 	if (error != 0) {
644 		return error;
645 	}
646 #endif /* CONTENT_FILTER */
647 #if FLOW_DIVERT
648 	if (necp_socket_should_use_flow_divert(inp)) {
649 		error = flow_divert_pcb_init(so);
650 		if (error == 0) {
651 			error = flow_divert_connect_out(so, nam, p);
652 		}
653 		return error;
654 	} else {
655 		so->so_flags1 |= SOF1_FLOW_DIVERT_SKIP;
656 	}
657 #endif /* FLOW_DIVERT */
658 #endif /* NECP */
659 	tp = intotcpcb(inp);
660 	TCPDEBUG1();
661 
662 	calculate_tcp_clock();
663 
664 	error = tcp_usr_connect_common(so, tp, nam, p, false, true);
665 	if (error != 0) {
666 		goto out;
667 	}
668 
669 	COMMON_END(PRU_CONNECT);
670 }
671 
672 static int
tcp_usr_connectx_common(struct socket * so,int af,struct sockaddr * src,struct sockaddr * dst,struct proc * p,uint32_t ifscope,sae_associd_t aid,sae_connid_t * pcid,uint32_t flags,void * arg,uint32_t arglen,struct uio * auio,user_ssize_t * bytes_written)673 tcp_usr_connectx_common(struct socket *so, int af,
674     struct sockaddr *src, struct sockaddr *dst,
675     struct proc *p, uint32_t ifscope, sae_associd_t aid, sae_connid_t *pcid,
676     uint32_t flags, void *arg, uint32_t arglen, struct uio *auio,
677     user_ssize_t *bytes_written)
678 {
679 #pragma unused(aid, flags, arg, arglen)
680 	struct inpcb *inp = sotoinpcb(so);
681 	int error = 0;
682 	user_ssize_t datalen = 0;
683 
684 	if (inp == NULL) {
685 		return EINVAL;
686 	}
687 
688 	VERIFY(dst != NULL);
689 
690 	ASSERT(!(inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS));
691 	inp->inp_flags2 |= INP2_CONNECT_IN_PROGRESS;
692 
693 #if NECP
694 	inp_update_necp_policy(inp, src, dst, ifscope);
695 #endif /* NECP */
696 
697 	if ((so->so_flags1 & SOF1_DATA_IDEMPOTENT) &&
698 	    (tcp_fastopen & TCP_FASTOPEN_CLIENT)) {
699 		sototcpcb(so)->t_flagsext |= TF_FASTOPEN;
700 	}
701 
702 	/* bind socket to the specified interface, if requested */
703 	if (ifscope != IFSCOPE_NONE &&
704 	    (error = inp_bindif(inp, ifscope, NULL)) != 0) {
705 		goto done;
706 	}
707 
708 	/* if source address and/or port is specified, bind to it */
709 	if (src != NULL) {
710 		error = sobindlock(so, src, 0); /* already locked */
711 		if (error != 0) {
712 			goto done;
713 		}
714 	}
715 
716 	switch (af) {
717 	case AF_INET:
718 		error = tcp_usr_connect(so, dst, p);
719 		break;
720 	case AF_INET6:
721 		error = tcp6_usr_connect(so, dst, p);
722 		break;
723 	default:
724 		VERIFY(0);
725 		/* NOTREACHED */
726 	}
727 
728 	if (error != 0) {
729 		goto done;
730 	}
731 
732 	/* if there is data, copy it */
733 	if (auio != NULL) {
734 		socket_unlock(so, 0);
735 
736 		VERIFY(bytes_written != NULL);
737 
738 		datalen = uio_resid(auio);
739 		error = so->so_proto->pr_usrreqs->pru_sosend(so, NULL,
740 		    (uio_t)auio, NULL, NULL, 0);
741 		socket_lock(so, 0);
742 
743 		if (error == 0 || error == EWOULDBLOCK) {
744 			*bytes_written = datalen - uio_resid(auio);
745 		}
746 
747 		/*
748 		 * sosend returns EWOULDBLOCK if it's a non-blocking
749 		 * socket or a timeout occured (this allows to return
750 		 * the amount of queued data through sendit()).
751 		 *
752 		 * However, connectx() returns EINPROGRESS in case of a
753 		 * blocking socket. So we change the return value here.
754 		 */
755 		if (error == EWOULDBLOCK) {
756 			error = EINPROGRESS;
757 		}
758 	}
759 
760 	if (error == 0 && pcid != NULL) {
761 		*pcid = 1; /* there is only one connection in regular TCP */
762 	}
763 done:
764 	if (error && error != EINPROGRESS) {
765 		so->so_flags1 &= ~SOF1_PRECONNECT_DATA;
766 	}
767 
768 	inp->inp_flags2 &= ~INP2_CONNECT_IN_PROGRESS;
769 	return error;
770 }
771 
772 static int
tcp_usr_connectx(struct socket * so,struct sockaddr * src,struct sockaddr * dst,struct proc * p,uint32_t ifscope,sae_associd_t aid,sae_connid_t * pcid,uint32_t flags,void * arg,uint32_t arglen,struct uio * uio,user_ssize_t * bytes_written)773 tcp_usr_connectx(struct socket *so, struct sockaddr *src,
774     struct sockaddr *dst, struct proc *p, uint32_t ifscope,
775     sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
776     uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written)
777 {
778 	return tcp_usr_connectx_common(so, AF_INET, src, dst, p, ifscope, aid,
779 	           pcid, flags, arg, arglen, uio, bytes_written);
780 }
781 
782 static int
tcp6_usr_connect(struct socket * so,struct sockaddr * nam,struct proc * p)783 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
784 {
785 	int error = 0;
786 	struct inpcb *inp = sotoinpcb(so);
787 	struct tcpcb *tp;
788 
789 	TCPDEBUG0;
790 	if (inp == NULL) {
791 		return EINVAL;
792 	} else if (inp->inp_state == INPCB_STATE_DEAD) {
793 		if (so->so_error) {
794 			error = so->so_error;
795 			so->so_error = 0;
796 			return error;
797 		} else {
798 			return EINVAL;
799 		}
800 	}
801 #if NECP
802 #if CONTENT_FILTER
803 	error = cfil_sock_attach(so, NULL, nam, CFS_CONNECTION_DIR_OUT);
804 	if (error != 0) {
805 		return error;
806 	}
807 #endif /* CONTENT_FILTER */
808 #if FLOW_DIVERT
809 	if (necp_socket_should_use_flow_divert(inp)) {
810 		error = flow_divert_pcb_init(so);
811 		if (error == 0) {
812 			error = flow_divert_connect_out(so, nam, p);
813 		}
814 		return error;
815 	} else {
816 		so->so_flags1 |= SOF1_FLOW_DIVERT_SKIP;
817 	}
818 #endif /* FLOW_DIVERT */
819 #endif /* NECP */
820 
821 	tp = intotcpcb(inp);
822 	TCPDEBUG1();
823 
824 	calculate_tcp_clock();
825 
826 	error = tcp_usr_connect_common(so, tp, nam, p, true, true);
827 	if (error != 0) {
828 		route_clear(&inp->inp_route);
829 		goto out;
830 	}
831 
832 	COMMON_END(PRU_CONNECT);
833 }
834 
835 static int
tcp6_usr_connectx(struct socket * so,struct sockaddr * src,struct sockaddr * dst,struct proc * p,uint32_t ifscope,sae_associd_t aid,sae_connid_t * pcid,uint32_t flags,void * arg,uint32_t arglen,struct uio * uio,user_ssize_t * bytes_written)836 tcp6_usr_connectx(struct socket *so, struct sockaddr*src,
837     struct sockaddr *dst, struct proc *p, uint32_t ifscope,
838     sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
839     uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written)
840 {
841 	return tcp_usr_connectx_common(so, AF_INET6, src, dst, p, ifscope, aid,
842 	           pcid, flags, arg, arglen, uio, bytes_written);
843 }
844 
845 /*
846  * Initiate disconnect from peer.
847  * If connection never passed embryonic stage, just drop;
848  * else if don't need to let data drain, then can just drop anyways,
849  * else have to begin TCP shutdown process: mark socket disconnecting,
850  * drain unread data, state switch to reflect user close, and
851  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
852  * when peer sends FIN and acks ours.
853  *
854  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
855  */
856 static int
tcp_usr_disconnect(struct socket * so)857 tcp_usr_disconnect(struct socket *so)
858 {
859 	int error = 0;
860 	struct inpcb *inp = sotoinpcb(so);
861 	struct tcpcb *tp;
862 
863 	socket_lock_assert_owned(so);
864 	COMMON_START();
865 	/* In case we got disconnected from the peer */
866 	if (tp == NULL) {
867 		goto out;
868 	}
869 	tp = tcp_disconnect(tp);
870 	COMMON_END(PRU_DISCONNECT);
871 }
872 
873 /*
874  * User-protocol pru_disconnectx callback.
875  */
876 static int
tcp_usr_disconnectx(struct socket * so,sae_associd_t aid,sae_connid_t cid)877 tcp_usr_disconnectx(struct socket *so, sae_associd_t aid, sae_connid_t cid)
878 {
879 #pragma unused(cid)
880 	if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL) {
881 		return EINVAL;
882 	}
883 
884 	return tcp_usr_disconnect(so);
885 }
886 
887 /*
888  * Accept a connection.  Essentially all the work is
889  * done at higher levels; just return the address
890  * of the peer, storing through addr.
891  */
892 static int
tcp_usr_accept(struct socket * so,struct sockaddr ** nam)893 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
894 {
895 	int error = 0;
896 	struct inpcb *inp = sotoinpcb(so);
897 	struct tcpcb *tp = NULL;
898 	TCPDEBUG0;
899 
900 	in_getpeeraddr(so, nam);
901 
902 	if (so->so_state & SS_ISDISCONNECTED) {
903 		error = ECONNABORTED;
904 		goto out;
905 	}
906 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
907 		return EINVAL;
908 	}
909 #if NECP
910 	else if (necp_socket_should_use_flow_divert(inp)) {
911 		return EPROTOTYPE;
912 	}
913 
914 #endif /* NECP */
915 
916 	tp = intotcpcb(inp);
917 	TCPDEBUG1();
918 
919 	TCP_LOG_ACCEPT(tp, 0);
920 
921 	calculate_tcp_clock();
922 
923 	COMMON_END(PRU_ACCEPT);
924 }
925 
926 static int
tcp6_usr_accept(struct socket * so,struct sockaddr ** nam)927 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
928 {
929 	int error = 0;
930 	struct inpcb *inp = sotoinpcb(so);
931 	struct tcpcb *tp = NULL;
932 	TCPDEBUG0;
933 
934 	if (so->so_state & SS_ISDISCONNECTED) {
935 		error = ECONNABORTED;
936 		goto out;
937 	}
938 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
939 		return EINVAL;
940 	}
941 #if NECP
942 	else if (necp_socket_should_use_flow_divert(inp)) {
943 		return EPROTOTYPE;
944 	}
945 
946 #endif /* NECP */
947 
948 	tp = intotcpcb(inp);
949 	TCPDEBUG1();
950 
951 	TCP_LOG_ACCEPT(tp, 0);
952 
953 	calculate_tcp_clock();
954 
955 	in6_mapped_peeraddr(so, nam);
956 	COMMON_END(PRU_ACCEPT);
957 }
958 
959 /*
960  * Mark the connection as being incapable of further output.
961  *
962  * Returns:	0			Success
963  *		EINVAL [COMMON_START]
964  *	tcp_output:EADDRNOTAVAIL
965  *	tcp_output:ENOBUFS
966  *	tcp_output:EMSGSIZE
967  *	tcp_output:EHOSTUNREACH
968  *	tcp_output:ENETUNREACH
969  *	tcp_output:ENETDOWN
970  *	tcp_output:ENOMEM
971  *	tcp_output:EACCES
972  *	tcp_output:EMSGSIZE
973  *	tcp_output:ENOBUFS
974  *	tcp_output:???			[ignorable: mostly IPSEC/firewall/DLIL]
975  */
976 static int
tcp_usr_shutdown(struct socket * so)977 tcp_usr_shutdown(struct socket *so)
978 {
979 	int error = 0;
980 	struct inpcb *inp = sotoinpcb(so);
981 	struct tcpcb *tp;
982 
983 	TCPDEBUG0;
984 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
985 		return EINVAL;
986 	}
987 
988 	socantsendmore(so);
989 
990 	/*
991 	 * In case we got disconnected from the peer, or if this is
992 	 * a socket that is to be flow-diverted (but not yet).
993 	 */
994 	tp = intotcpcb(inp);
995 	TCPDEBUG1();
996 
997 	if (tp == NULL
998 #if NECP
999 	    || (necp_socket_should_use_flow_divert(inp))
1000 #endif /* NECP */
1001 	    ) {
1002 		if (tp != NULL) {
1003 			error = EPROTOTYPE;
1004 		}
1005 		goto out;
1006 	}
1007 
1008 	calculate_tcp_clock();
1009 
1010 	tp = tcp_usrclosed(tp);
1011 #if MPTCP
1012 	/* A reset has been sent but socket exists, do not send FIN */
1013 	if ((so->so_flags & SOF_MP_SUBFLOW) &&
1014 	    (tp) && (tp->t_mpflags & TMPF_RESET)) {
1015 		goto out;
1016 	}
1017 #endif
1018 #if CONTENT_FILTER
1019 	/* Don't send a FIN yet */
1020 	if (tp && !(so->so_state & SS_ISDISCONNECTED) &&
1021 	    cfil_sock_data_pending(&so->so_snd)) {
1022 		goto out;
1023 	}
1024 #endif /* CONTENT_FILTER */
1025 	if (tp) {
1026 		error = tcp_output(tp);
1027 	}
1028 	COMMON_END(PRU_SHUTDOWN);
1029 }
1030 
1031 /*
1032  * After a receive, possibly send window update to peer.
1033  */
1034 static int
tcp_usr_rcvd(struct socket * so,int flags)1035 tcp_usr_rcvd(struct socket *so, int flags)
1036 {
1037 	int error = 0;
1038 	struct inpcb *inp = sotoinpcb(so);
1039 	struct tcpcb *tp;
1040 
1041 	COMMON_START();
1042 	/* In case we got disconnected from the peer */
1043 	if (tp == NULL) {
1044 		goto out;
1045 	}
1046 	tcp_sbrcv_trim(tp, &so->so_rcv);
1047 
1048 	if ((flags & MSG_WAITALL) && SEQ_LT(tp->last_ack_sent, tp->rcv_nxt)) {
1049 		tp->t_flags |= TF_ACKNOW;
1050 	}
1051 
1052 	/*
1053 	 * This tcp_output is solely there to trigger window-updates.
1054 	 * However, we really do not want these window-updates while we
1055 	 * are still in SYN_SENT or SYN_RECEIVED.
1056 	 */
1057 	if (TCPS_HAVEESTABLISHED(tp->t_state)) {
1058 		tcp_output(tp);
1059 	}
1060 
1061 #if CONTENT_FILTER
1062 	cfil_sock_buf_update(&so->so_rcv);
1063 #endif /* CONTENT_FILTER */
1064 
1065 	COMMON_END(PRU_RCVD);
1066 }
1067 
1068 __attribute__((noinline))
1069 static int
tcp_send_implied_connect(struct socket * so,struct tcpcb * tp,struct sockaddr * nam,struct proc * p,bool isipv6)1070 tcp_send_implied_connect(struct socket *so, struct tcpcb *tp, struct sockaddr *nam,
1071     struct proc *p, bool isipv6)
1072 {
1073 	int error = 0;
1074 
1075 	error = tcp_usr_connect_common(so, tp, nam, p, isipv6, false);
1076 	if (error != 0) {
1077 		goto out;
1078 	}
1079 	/*
1080 	 * initialize window to default value, and
1081 	 * initialize maxseg/maxopd using peer's cached
1082 	 * MSS.
1083 	 */
1084 	tp->snd_wnd = TTCP_CLIENT_SND_WND;
1085 	tp->max_sndwnd = tp->snd_wnd;
1086 	tcp_mss(tp, -1, IFSCOPE_NONE);
1087 out:
1088 	return error;
1089 }
1090 
1091 __attribute__((noinline))
1092 static void
mpkl_tcp_send(struct socket * so,struct tcpcb * tp,uint32_t mpkl_seq,uint32_t mpkl_len,struct so_mpkl_send_info * mpkl_send_info)1093 mpkl_tcp_send(struct socket *so, struct tcpcb *tp, uint32_t mpkl_seq, uint32_t mpkl_len,
1094     struct so_mpkl_send_info *mpkl_send_info)
1095 {
1096 	struct inpcb *inp = tp->t_inpcb;
1097 
1098 	if (inp == NULL) {
1099 		return;
1100 	}
1101 
1102 	if ((inp->inp_last_outifp != NULL &&
1103 	    (inp->inp_last_outifp->if_xflags & IFXF_MPK_LOG)) ||
1104 	    (inp->inp_boundifp != NULL &&
1105 	    (inp->inp_boundifp->if_xflags & IFXF_MPK_LOG))) {
1106 		MPKL_TCP_SEND(tcp_mpkl_log_object,
1107 		    mpkl_send_info->mpkl_proto,
1108 		    mpkl_send_info->mpkl_uuid,
1109 		    ntohs(inp->inp_lport),
1110 		    ntohs(inp->inp_fport),
1111 		    mpkl_seq,
1112 		    mpkl_len,
1113 		    so->last_pid,
1114 		    so->so_log_seqn++);
1115 	}
1116 }
1117 
1118 /*
1119  * Do a send by putting data in output queue and updating urgent
1120  * marker if URG set.  Possibly send more data.  Unlike the other
1121  * pru_*() routines, the mbuf chains are our responsibility.  We
1122  * must either enqueue them or free them.  The other pru_* routines
1123  * generally are caller-frees.
1124  *
1125  * Returns:	0			Success
1126  *		ECONNRESET
1127  *		EINVAL
1128  *		ENOBUFS
1129  *	tcp_connect:EADDRINUSE		Address in use
1130  *	tcp_connect:EADDRNOTAVAIL	Address not available.
1131  *	tcp_connect:EINVAL		Invalid argument
1132  *	tcp_connect:EAFNOSUPPORT	Address family not supported [notdef]
1133  *	tcp_connect:EACCES		Permission denied
1134  *	tcp_connect:EAGAIN		Resource unavailable, try again
1135  *	tcp_connect:EPERM		Operation not permitted
1136  *	tcp_output:EADDRNOTAVAIL
1137  *	tcp_output:ENOBUFS
1138  *	tcp_output:EMSGSIZE
1139  *	tcp_output:EHOSTUNREACH
1140  *	tcp_output:ENETUNREACH
1141  *	tcp_output:ENETDOWN
1142  *	tcp_output:ENOMEM
1143  *	tcp_output:EACCES
1144  *	tcp_output:EMSGSIZE
1145  *	tcp_output:ENOBUFS
1146  *	tcp_output:???			[ignorable: mostly IPSEC/firewall/DLIL]
1147  *	tcp6_connect:???		[IPV6 only]
1148  */
1149 static int
tcp_usr_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1150 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
1151     struct sockaddr *nam, struct mbuf *control, struct proc *p)
1152 {
1153 	int error = 0;
1154 	struct inpcb *inp = sotoinpcb(so);
1155 	struct tcpcb *tp;
1156 	uint32_t mpkl_len = 0; /* length of mbuf chain */
1157 	uint32_t mpkl_seq = 0; /* sequence number where new data is added */
1158 	struct so_mpkl_send_info mpkl_send_info = {};
1159 	bool isipv6;
1160 
1161 	TCPDEBUG0;
1162 
1163 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD
1164 #if NECP
1165 	    || (necp_socket_should_use_flow_divert(inp))
1166 #endif /* NECP */
1167 	    ) {
1168 		/*
1169 		 * OOPS! we lost a race, the TCP session got reset after
1170 		 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a
1171 		 * network interrupt in the non-splnet() section of sosend().
1172 		 */
1173 		if (m != NULL) {
1174 			m_freem(m);
1175 		}
1176 		if (control != NULL) {
1177 			m_freem(control);
1178 			control = NULL;
1179 		}
1180 
1181 		if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
1182 			error = ECONNRESET;     /* XXX EPIPE? */
1183 		} else {
1184 			error = EPROTOTYPE;
1185 		}
1186 		tp = NULL;
1187 		TCPDEBUG1();
1188 		goto out;
1189 	}
1190 	isipv6 = nam && nam->sa_family == AF_INET6 ? true : false;
1191 	tp = intotcpcb(inp);
1192 	TCPDEBUG1();
1193 
1194 	calculate_tcp_clock();
1195 
1196 	if (net_mpklog_enabled) {
1197 		mpkl_seq = tp->snd_una + so->so_snd.sb_cc;
1198 		if (m) {
1199 			mpkl_len = m_length(m);
1200 		}
1201 		if (so->so_flags1 & SOF1_MPKL_SEND_INFO) {
1202 			uuid_copy(mpkl_send_info.mpkl_uuid, so->so_mpkl_send_uuid);
1203 			mpkl_send_info.mpkl_proto = so->so_mpkl_send_proto;
1204 		}
1205 	}
1206 
1207 	if (control != NULL) {
1208 		if (control->m_len > 0 && net_mpklog_enabled) {
1209 			error = tcp_get_mpkl_send_info(control, &mpkl_send_info);
1210 			/*
1211 			 * Intepretation of the returned code:
1212 			 *  0: client wants us to use value passed in SCM_MPKL_SEND_INFO
1213 			 *  1: SCM_MPKL_SEND_INFO was not present
1214 			 *  other: failure
1215 			 */
1216 			if (error != 0 && error != ENOMSG) {
1217 				m_freem(control);
1218 				if (m != NULL) {
1219 					m_freem(m);
1220 				}
1221 				control = NULL;
1222 				m = NULL;
1223 				goto out;
1224 			}
1225 		}
1226 		/*
1227 		 * Silently drop unsupported ancillary data messages
1228 		 */
1229 		m_freem(control);
1230 		control = NULL;
1231 	}
1232 
1233 	/* MPTCP sublow socket buffers must not be compressed */
1234 	VERIFY(!(so->so_flags & SOF_MP_SUBFLOW) ||
1235 	    (so->so_snd.sb_flags & SB_NOCOMPRESS));
1236 
1237 	if (!(flags & PRUS_OOB) || (so->so_flags1 & SOF1_PRECONNECT_DATA)) {
1238 		sbappendstream(&so->so_snd, m);
1239 
1240 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1241 			/*
1242 			 * Do implied connect if not yet connected,
1243 			 */
1244 			error = tcp_send_implied_connect(so, tp, nam, p, isipv6);
1245 			if (error != 0) {
1246 				goto out;
1247 			}
1248 			/* The sequence number of the data is past the SYN */
1249 			mpkl_seq = tp->iss + 1;
1250 		}
1251 
1252 		if (flags & PRUS_EOF) {
1253 			/*
1254 			 * Close the send side of the connection after
1255 			 * the data is sent.
1256 			 */
1257 			socantsendmore(so);
1258 			tp = tcp_usrclosed(tp);
1259 		}
1260 		if (tp != NULL) {
1261 			if (flags & PRUS_MORETOCOME) {
1262 				tp->t_flags |= TF_MORETOCOME;
1263 			}
1264 			tp->t_flagsext |= TF_USR_OUTPUT;
1265 			error = tcp_output(tp);
1266 			tp->t_flagsext &= ~TF_USR_OUTPUT;
1267 			if (flags & PRUS_MORETOCOME) {
1268 				tp->t_flags &= ~TF_MORETOCOME;
1269 			}
1270 		}
1271 	} else {
1272 		if (sbspace(&so->so_snd) == 0) {
1273 			/* if no space is left in sockbuf,
1274 			 * do not try to squeeze in OOB traffic */
1275 			m_freem(m);
1276 			error = ENOBUFS;
1277 			goto out;
1278 		}
1279 		/*
1280 		 * According to RFC961 (Assigned Protocols),
1281 		 * the urgent pointer points to the last octet
1282 		 * of urgent data.  We continue, however,
1283 		 * to consider it to indicate the first octet
1284 		 * of data past the urgent section.
1285 		 * Otherwise, snd_up should be one lower.
1286 		 */
1287 		sbappendstream(&so->so_snd, m);
1288 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1289 			/*
1290 			 * Do implied connect if not yet connected,
1291 			 * initialize window to default value, and
1292 			 * initialize maxseg/maxopd using peer's cached
1293 			 * MSS.
1294 			 */
1295 			error = tcp_send_implied_connect(so, tp, nam, p, isipv6);
1296 			if (error != 0) {
1297 				goto out;
1298 			}
1299 		}
1300 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
1301 		tp->t_flagsext |= TF_FORCE;
1302 		tp->t_flagsext |= TF_USR_OUTPUT;
1303 		error = tcp_output(tp);
1304 		tp->t_flagsext &= ~TF_USR_OUTPUT;
1305 		tp->t_flagsext &= ~TF_FORCE;
1306 	}
1307 
1308 	if (net_mpklog_enabled) {
1309 		mpkl_tcp_send(so, tp, mpkl_seq, mpkl_len, &mpkl_send_info);
1310 	}
1311 
1312 	/*
1313 	 * We wait for the socket to successfully connect before returning.
1314 	 * This allows us to signal a timeout to the application.
1315 	 */
1316 	if (so->so_state & SS_ISCONNECTING) {
1317 		if (so->so_state & SS_NBIO) {
1318 			error = EWOULDBLOCK;
1319 		} else {
1320 			error = sbwait(&so->so_snd);
1321 		}
1322 	}
1323 
1324 	COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
1325 	    ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1326 }
1327 
1328 /*
1329  * Abort the TCP.
1330  */
1331 static int
tcp_usr_abort(struct socket * so)1332 tcp_usr_abort(struct socket *so)
1333 {
1334 	int error = 0;
1335 	struct inpcb *inp = sotoinpcb(so);
1336 	struct tcpcb *tp;
1337 
1338 	COMMON_START();
1339 	/* In case we got disconnected from the peer */
1340 	if (tp == NULL) {
1341 		goto out;
1342 	}
1343 	tp = tcp_drop(tp, ECONNABORTED);
1344 	VERIFY(so->so_usecount > 0);
1345 	so->so_usecount--;
1346 	COMMON_END(PRU_ABORT);
1347 }
1348 
1349 /*
1350  * Receive out-of-band data.
1351  *
1352  * Returns:	0			Success
1353  *		EINVAL [COMMON_START]
1354  *		EINVAL
1355  *		EWOULDBLOCK
1356  */
1357 static int
tcp_usr_rcvoob(struct socket * so,struct mbuf * m,int flags)1358 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1359 {
1360 	int error = 0;
1361 	struct inpcb *inp = sotoinpcb(so);
1362 	struct tcpcb *tp;
1363 
1364 	COMMON_START();
1365 	if ((so->so_oobmark == 0 &&
1366 	    (so->so_state & SS_RCVATMARK) == 0) ||
1367 	    so->so_options & SO_OOBINLINE ||
1368 	    tp->t_oobflags & TCPOOB_HADDATA) {
1369 		error = EINVAL;
1370 		goto out;
1371 	}
1372 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1373 		error = EWOULDBLOCK;
1374 		goto out;
1375 	}
1376 	m->m_len = 1;
1377 	*mtod(m, caddr_t) = tp->t_iobc;
1378 	so->so_state &= ~SS_RCVATMARK;
1379 	if ((flags & MSG_PEEK) == 0) {
1380 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1381 	}
1382 	COMMON_END(PRU_RCVOOB);
1383 }
1384 
1385 static int
tcp_usr_preconnect(struct socket * so)1386 tcp_usr_preconnect(struct socket *so)
1387 {
1388 	struct inpcb *inp = sotoinpcb(so);
1389 	int error = 0;
1390 
1391 #if NECP
1392 	if (necp_socket_should_use_flow_divert(inp)) {
1393 		/* May happen, if in tcp_usr_connect we did not had a chance
1394 		 * to set the usrreqs (due to some error). So, let's get out
1395 		 * of here.
1396 		 */
1397 		goto out;
1398 	}
1399 #endif /* NECP */
1400 
1401 	error = tcp_output(sototcpcb(so));
1402 
1403 	soclearfastopen(so);
1404 
1405 	COMMON_END(PRU_PRECONNECT);
1406 }
1407 
1408 /* xxx - should be const */
1409 struct pr_usrreqs tcp_usrreqs = {
1410 	.pru_abort =            tcp_usr_abort,
1411 	.pru_accept =           tcp_usr_accept,
1412 	.pru_attach =           tcp_usr_attach,
1413 	.pru_bind =             tcp_usr_bind,
1414 	.pru_connect =          tcp_usr_connect,
1415 	.pru_connectx =         tcp_usr_connectx,
1416 	.pru_control =          in_control,
1417 	.pru_detach =           tcp_usr_detach,
1418 	.pru_disconnect =       tcp_usr_disconnect,
1419 	.pru_disconnectx =      tcp_usr_disconnectx,
1420 	.pru_listen =           tcp_usr_listen,
1421 	.pru_peeraddr =         in_getpeeraddr,
1422 	.pru_rcvd =             tcp_usr_rcvd,
1423 	.pru_rcvoob =           tcp_usr_rcvoob,
1424 	.pru_send =             tcp_usr_send,
1425 	.pru_shutdown =         tcp_usr_shutdown,
1426 	.pru_sockaddr =         in_getsockaddr,
1427 	.pru_sosend =           sosend,
1428 	.pru_soreceive =        soreceive,
1429 	.pru_preconnect =       tcp_usr_preconnect,
1430 };
1431 
1432 struct pr_usrreqs tcp6_usrreqs = {
1433 	.pru_abort =            tcp_usr_abort,
1434 	.pru_accept =           tcp6_usr_accept,
1435 	.pru_attach =           tcp_usr_attach,
1436 	.pru_bind =             tcp6_usr_bind,
1437 	.pru_connect =          tcp6_usr_connect,
1438 	.pru_connectx =         tcp6_usr_connectx,
1439 	.pru_control =          in6_control,
1440 	.pru_detach =           tcp_usr_detach,
1441 	.pru_disconnect =       tcp_usr_disconnect,
1442 	.pru_disconnectx =      tcp_usr_disconnectx,
1443 	.pru_listen =           tcp6_usr_listen,
1444 	.pru_peeraddr =         in6_mapped_peeraddr,
1445 	.pru_rcvd =             tcp_usr_rcvd,
1446 	.pru_rcvoob =           tcp_usr_rcvoob,
1447 	.pru_send =             tcp_usr_send,
1448 	.pru_shutdown =         tcp_usr_shutdown,
1449 	.pru_sockaddr =         in6_mapped_sockaddr,
1450 	.pru_sosend =           sosend,
1451 	.pru_soreceive =        soreceive,
1452 	.pru_preconnect =       tcp_usr_preconnect,
1453 };
1454 
1455 /*
1456  * Common subroutine to open a TCP connection to remote host specified
1457  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
1458  * port number if needed.  Call in_pcbladdr to do the routing and to choose
1459  * a local host address (interface).  If there is an existing incarnation
1460  * of the same connection in TIME-WAIT state and if the remote host was
1461  * sending CC options and if the connection duration was < MSL, then
1462  * truncate the previous TIME-WAIT state and proceed.
1463  * Initialize connection parameters and enter SYN-SENT state.
1464  *
1465  * Returns:	0			Success
1466  *		EADDRINUSE
1467  *		EINVAL
1468  *	in_pcbbind:EADDRNOTAVAIL	Address not available.
1469  *	in_pcbbind:EINVAL		Invalid argument
1470  *	in_pcbbind:EAFNOSUPPORT		Address family not supported [notdef]
1471  *	in_pcbbind:EACCES		Permission denied
1472  *	in_pcbbind:EADDRINUSE		Address in use
1473  *	in_pcbbind:EAGAIN		Resource unavailable, try again
1474  *	in_pcbbind:EPERM		Operation not permitted
1475  *	in_pcbladdr:EINVAL		Invalid argument
1476  *	in_pcbladdr:EAFNOSUPPORT	Address family not supported
1477  *	in_pcbladdr:EADDRNOTAVAIL	Address not available
1478  */
1479 static int
tcp_connect(struct tcpcb * tp,struct sockaddr * nam,struct proc * p)1480 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct proc *p)
1481 {
1482 	struct inpcb *inp = tp->t_inpcb, *oinp;
1483 	struct socket *so = inp->inp_socket;
1484 	struct tcpcb *otp;
1485 	struct sockaddr_in *sin = SIN(nam);
1486 	struct in_addr laddr;
1487 	int error = 0;
1488 	struct ifnet *outif = NULL;
1489 
1490 	if (inp->inp_lport == 0) {
1491 		error = in_pcbbind(inp, NULL, p);
1492 		if (error) {
1493 			goto done;
1494 		}
1495 	}
1496 
1497 	/*
1498 	 * Cannot simply call in_pcbconnect, because there might be an
1499 	 * earlier incarnation of this same connection still in
1500 	 * TIME_WAIT state, creating an ADDRINUSE error.
1501 	 */
1502 	error = in_pcbladdr(inp, nam, &laddr, IFSCOPE_NONE, &outif, 0);
1503 	if (error) {
1504 		goto done;
1505 	}
1506 
1507 	socket_unlock(inp->inp_socket, 0);
1508 	oinp = in_pcblookup_hash(inp->inp_pcbinfo,
1509 	    sin->sin_addr, sin->sin_port,
1510 	    inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr : laddr,
1511 	    inp->inp_lport, 0, NULL);
1512 
1513 	socket_lock(inp->inp_socket, 0);
1514 	if (oinp) {
1515 		if (oinp != inp) { /* 4143933: avoid deadlock if inp == oinp */
1516 			socket_lock(oinp->inp_socket, 1);
1517 		}
1518 		if (in_pcb_checkstate(oinp, WNT_RELEASE, 1) == WNT_STOPUSING) {
1519 			if (oinp != inp) {
1520 				socket_unlock(oinp->inp_socket, 1);
1521 			}
1522 			goto skip_oinp;
1523 		}
1524 
1525 		if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
1526 		    otp->t_state == TCPS_TIME_WAIT &&
1527 		    ((int)(tcp_now - otp->t_starttime)) < tcp_msl &&
1528 		    (otp->t_flags & TF_RCVD_CC)) {
1529 			otp = tcp_close(otp);
1530 		} else {
1531 			printf("tcp_connect: inp=0x%llx err=EADDRINUSE\n",
1532 			    (uint64_t)VM_KERNEL_ADDRPERM(inp));
1533 			if (oinp != inp) {
1534 				socket_unlock(oinp->inp_socket, 1);
1535 			}
1536 			error = EADDRINUSE;
1537 			goto done;
1538 		}
1539 		if (oinp != inp) {
1540 			socket_unlock(oinp->inp_socket, 1);
1541 		}
1542 	}
1543 skip_oinp:
1544 	if ((inp->inp_laddr.s_addr == INADDR_ANY ? laddr.s_addr :
1545 	    inp->inp_laddr.s_addr) == sin->sin_addr.s_addr &&
1546 	    inp->inp_lport == sin->sin_port) {
1547 		error = EINVAL;
1548 		goto done;
1549 	}
1550 #if SKYWALK
1551 	if (!NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1552 		error = netns_reserve_in(&inp->inp_netns_token,
1553 		    inp->inp_laddr.s_addr != INADDR_ANY ?
1554 		    inp->inp_laddr : laddr,
1555 		    IPPROTO_TCP, inp->inp_lport, NETNS_BSD, NULL);
1556 		if (error) {
1557 			goto done;
1558 		}
1559 	}
1560 #endif /* SKYWALK */
1561 	if (!lck_rw_try_lock_exclusive(&inp->inp_pcbinfo->ipi_lock)) {
1562 		/*lock inversion issue, mostly with udp multicast packets */
1563 		socket_unlock(inp->inp_socket, 0);
1564 		lck_rw_lock_exclusive(&inp->inp_pcbinfo->ipi_lock);
1565 		socket_lock(inp->inp_socket, 0);
1566 	}
1567 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
1568 		inp->inp_laddr = laddr;
1569 		/* no reference needed */
1570 		inp->inp_last_outifp = outif;
1571 #if SKYWALK
1572 		if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1573 			netns_set_ifnet(&inp->inp_netns_token, inp->inp_last_outifp);
1574 		}
1575 #endif /* SKYWALK */
1576 
1577 		inp->inp_flags |= INP_INADDR_ANY;
1578 	}
1579 	inp->inp_faddr = sin->sin_addr;
1580 	inp->inp_fport = sin->sin_port;
1581 	in_pcbrehash(inp);
1582 	lck_rw_done(&inp->inp_pcbinfo->ipi_lock);
1583 
1584 	if (inp->inp_flowhash == 0) {
1585 		inp_calc_flowhash(inp);
1586 		ASSERT(inp->inp_flowhash != 0);
1587 	}
1588 
1589 	tcp_set_max_rwinscale(tp, so);
1590 
1591 	soisconnecting(so);
1592 	tcpstat.tcps_connattempt++;
1593 	TCP_LOG_STATE(tp, TCPS_SYN_SENT);
1594 	tp->t_state = TCPS_SYN_SENT;
1595 	tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp, TCP_CONN_KEEPINIT(tp));
1596 	tp->iss = tcp_new_isn(tp);
1597 	tcp_sendseqinit(tp);
1598 	tp->t_connect_time = tcp_now;
1599 	if (nstat_collect) {
1600 		nstat_pcb_event(inp, NSTAT_EVENT_SRC_FLOW_STATE_OUTBOUND);
1601 		nstat_route_connect_attempt(inp->inp_route.ro_rt);
1602 	}
1603 
1604 	tcp_add_fsw_flow(tp, outif);
1605 
1606 done:
1607 	if (outif != NULL) {
1608 		ifnet_release(outif);
1609 	}
1610 
1611 	return error;
1612 }
1613 
1614 static int
tcp6_connect(struct tcpcb * tp,struct sockaddr * nam,struct proc * p)1615 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct proc *p)
1616 {
1617 	struct inpcb *inp = tp->t_inpcb, *oinp;
1618 	struct socket *so = inp->inp_socket;
1619 	struct tcpcb *otp;
1620 	struct sockaddr_in6 *sin6 = SIN6(nam);
1621 	struct in6_addr addr6;
1622 	int error = 0;
1623 	struct ifnet *outif = NULL;
1624 
1625 	if (inp->inp_lport == 0) {
1626 		error = in6_pcbbind(inp, NULL, p);
1627 		if (error) {
1628 			goto done;
1629 		}
1630 	}
1631 
1632 	/*
1633 	 * Cannot simply call in_pcbconnect, because there might be an
1634 	 * earlier incarnation of this same connection still in
1635 	 * TIME_WAIT state, creating an ADDRINUSE error.
1636 	 *
1637 	 * in6_pcbladdr() might return an ifp with its reference held
1638 	 * even in the error case, so make sure that it's released
1639 	 * whenever it's non-NULL.
1640 	 */
1641 	error = in6_pcbladdr(inp, nam, &addr6, &outif);
1642 	if (error) {
1643 		goto done;
1644 	}
1645 	socket_unlock(inp->inp_socket, 0);
1646 
1647 	uint32_t lifscope = IFSCOPE_NONE;
1648 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1649 		lifscope = inp->inp_lifscope;
1650 	} else if (sin6->sin6_scope_id != IFSCOPE_NONE) {
1651 		lifscope = sin6->sin6_scope_id;
1652 	} else if (outif != NULL) {
1653 		lifscope = outif->if_index;
1654 	}
1655 	oinp = in6_pcblookup_hash(inp->inp_pcbinfo,
1656 	    &sin6->sin6_addr, sin6->sin6_port, sin6->sin6_scope_id,
1657 	    IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
1658 	    ? &addr6
1659 	    : &inp->in6p_laddr,
1660 	    inp->inp_lport, lifscope, 0, NULL);
1661 	socket_lock(inp->inp_socket, 0);
1662 	if (oinp) {
1663 		if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
1664 		    otp->t_state == TCPS_TIME_WAIT &&
1665 		    ((int)(tcp_now - otp->t_starttime)) < tcp_msl &&
1666 		    (otp->t_flags & TF_RCVD_CC)) {
1667 			otp = tcp_close(otp);
1668 		} else {
1669 			error = EADDRINUSE;
1670 			goto done;
1671 		}
1672 	}
1673 #if SKYWALK
1674 	if (!NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1675 		error = netns_reserve_in6(&inp->inp_netns_token,
1676 		    IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ?
1677 		    addr6 : inp->in6p_laddr,
1678 		    IPPROTO_TCP, inp->inp_lport, NETNS_BSD, NULL);
1679 		if (error) {
1680 			goto done;
1681 		}
1682 	}
1683 #endif /* SKYWALK */
1684 	if (!lck_rw_try_lock_exclusive(&inp->inp_pcbinfo->ipi_lock)) {
1685 		/*lock inversion issue, mostly with udp multicast packets */
1686 		socket_unlock(inp->inp_socket, 0);
1687 		lck_rw_lock_exclusive(&inp->inp_pcbinfo->ipi_lock);
1688 		socket_lock(inp->inp_socket, 0);
1689 	}
1690 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1691 		inp->in6p_laddr = addr6;
1692 		inp->in6p_last_outifp = outif;  /* no reference needed */
1693 		inp->inp_lifscope = lifscope;
1694 		in6_verify_ifscope(&inp->in6p_laddr, inp->inp_lifscope);
1695 #if SKYWALK
1696 		if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1697 			netns_set_ifnet(&inp->inp_netns_token, inp->in6p_last_outifp);
1698 		}
1699 #endif /* SKYWALK */
1700 		inp->in6p_flags |= INP_IN6ADDR_ANY;
1701 	}
1702 	inp->in6p_faddr = sin6->sin6_addr;
1703 	inp->inp_fport = sin6->sin6_port;
1704 	inp->inp_fifscope = sin6->sin6_scope_id;
1705 	in6_verify_ifscope(&inp->in6p_faddr, inp->inp_fifscope);
1706 	if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != 0) {
1707 		inp->inp_flow = sin6->sin6_flowinfo;
1708 	}
1709 	in_pcbrehash(inp);
1710 	lck_rw_done(&inp->inp_pcbinfo->ipi_lock);
1711 
1712 	if (inp->inp_flowhash == 0) {
1713 		inp_calc_flowhash(inp);
1714 		ASSERT(inp->inp_flowhash != 0);
1715 	}
1716 	/* update flowinfo - RFC 6437 */
1717 	if (inp->inp_flow == 0 && inp->in6p_flags & IN6P_AUTOFLOWLABEL) {
1718 		inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
1719 		inp->inp_flow |=
1720 		    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
1721 	}
1722 
1723 	tcp_set_max_rwinscale(tp, so);
1724 
1725 	soisconnecting(so);
1726 	tcpstat.tcps_connattempt++;
1727 	TCP_LOG_STATE(tp, TCPS_SYN_SENT);
1728 	tp->t_state = TCPS_SYN_SENT;
1729 	tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
1730 	    TCP_CONN_KEEPINIT(tp));
1731 	tp->iss = tcp_new_isn(tp);
1732 	tcp_sendseqinit(tp);
1733 	tp->t_connect_time = tcp_now;
1734 	if (nstat_collect) {
1735 		nstat_pcb_event(inp, NSTAT_EVENT_SRC_FLOW_STATE_OUTBOUND);
1736 		nstat_route_connect_attempt(inp->inp_route.ro_rt);
1737 	}
1738 
1739 	tcp_add_fsw_flow(tp, outif);
1740 
1741 done:
1742 	if (outif != NULL) {
1743 		ifnet_release(outif);
1744 	}
1745 
1746 	return error;
1747 }
1748 
1749 /*
1750  * Export TCP internal state information via a struct tcp_info
1751  */
1752 void
tcp_fill_info(struct tcpcb * tp,struct tcp_info * ti)1753 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1754 {
1755 	struct inpcb *inp = tp->t_inpcb;
1756 
1757 	bzero(ti, sizeof(*ti));
1758 
1759 	ti->tcpi_state = (uint8_t)tp->t_state;
1760 	ti->tcpi_flowhash = inp != NULL ? inp->inp_flowhash: 0;
1761 
1762 	if (TSTMP_SUPPORTED(tp)) {
1763 		ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1764 	}
1765 	if (SACK_ENABLED(tp)) {
1766 		ti->tcpi_options |= TCPI_OPT_SACK;
1767 	}
1768 	if (TCP_WINDOW_SCALE_ENABLED(tp)) {
1769 		ti->tcpi_options |= TCPI_OPT_WSCALE;
1770 		ti->tcpi_snd_wscale = tp->snd_scale;
1771 		ti->tcpi_rcv_wscale = tp->rcv_scale;
1772 	}
1773 	if (TCP_ECN_ENABLED(tp)) {
1774 		ti->tcpi_options |= TCPI_OPT_ECN;
1775 	}
1776 
1777 	/* Are we in retranmission episode */
1778 	if (IN_FASTRECOVERY(tp) || tp->t_rxtshift > 0) {
1779 		ti->tcpi_flags |= TCPI_FLAG_LOSSRECOVERY;
1780 	}
1781 
1782 	if (tp->t_flags & TF_STREAMING_ON) {
1783 		ti->tcpi_flags |= TCPI_FLAG_STREAMING_ON;
1784 	}
1785 
1786 	ti->tcpi_rto = tp->t_timer[TCPT_REXMT] ? tp->t_rxtcur : 0;
1787 	ti->tcpi_snd_mss = tp->t_maxseg;
1788 	ti->tcpi_rcv_mss = tp->t_maxseg;
1789 
1790 	ti->tcpi_rttcur = tp->t_rttcur;
1791 	ti->tcpi_srtt = tp->t_srtt >> TCP_RTT_SHIFT;
1792 	ti->tcpi_rcv_srtt = tp->rcv_srtt >> TCP_RTT_SHIFT;
1793 	ti->tcpi_rttvar = tp->t_rttvar >> TCP_RTTVAR_SHIFT;
1794 	ti->tcpi_rttbest = tp->t_rttbest >> TCP_RTT_SHIFT;
1795 
1796 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1797 	ti->tcpi_snd_cwnd = tp->snd_cwnd;
1798 	if (inp != NULL && inp->inp_socket != NULL) {
1799 		ti->tcpi_snd_sbbytes = inp->inp_socket->so_snd.sb_cc;
1800 	}
1801 
1802 	ti->tcpi_rcv_space = tp->rcv_adv > tp->rcv_nxt ?
1803 	    tp->rcv_adv - tp->rcv_nxt : 0;
1804 
1805 	ti->tcpi_snd_wnd = tp->snd_wnd;
1806 	ti->tcpi_snd_nxt = tp->snd_nxt;
1807 	ti->tcpi_rcv_nxt = tp->rcv_nxt;
1808 
1809 	/* convert bytes/msec to bits/sec */
1810 	if ((tp->t_flagsext & TF_MEASURESNDBW) != 0 &&
1811 	    tp->t_bwmeas != NULL) {
1812 		ti->tcpi_snd_bw = (tp->t_bwmeas->bw_sndbw * 8000);
1813 	}
1814 
1815 	ti->tcpi_txpackets = inp != NULL ? inp->inp_stat->txpackets : 0;
1816 	ti->tcpi_txbytes = inp != NULL ? inp->inp_stat->txbytes : 0;
1817 	ti->tcpi_txretransmitbytes = tp->t_stat.txretransmitbytes;
1818 	ti->tcpi_txretransmitpackets = tp->t_stat.rxmitpkts;
1819 	ti->tcpi_txunacked = tp->snd_max - tp->snd_una;
1820 
1821 	ti->tcpi_rxpackets = inp != NULL ? inp->inp_stat->rxpackets : 0;
1822 	ti->tcpi_rxbytes = inp != NULL ? inp->inp_stat->rxbytes : 0;
1823 	ti->tcpi_rxduplicatebytes = tp->t_stat.rxduplicatebytes;
1824 	ti->tcpi_rxoutoforderbytes = tp->t_stat.rxoutoforderbytes;
1825 
1826 	if (tp->t_state > TCPS_LISTEN) {
1827 		ti->tcpi_synrexmits = (uint8_t)tp->t_stat.rxmitsyns;
1828 	}
1829 	if (inp != NULL) {
1830 		ti->tcpi_cell_rxpackets = inp->inp_cstat->rxpackets;
1831 		ti->tcpi_cell_rxbytes = inp->inp_cstat->rxbytes;
1832 		ti->tcpi_cell_txpackets = inp->inp_cstat->txpackets;
1833 		ti->tcpi_cell_txbytes = inp->inp_cstat->txbytes;
1834 
1835 		ti->tcpi_wifi_rxpackets = inp->inp_wstat->rxpackets;
1836 		ti->tcpi_wifi_rxbytes = inp->inp_wstat->rxbytes;
1837 		ti->tcpi_wifi_txpackets = inp->inp_wstat->txpackets;
1838 		ti->tcpi_wifi_txbytes = inp->inp_wstat->txbytes;
1839 
1840 		ti->tcpi_wired_rxpackets = inp->inp_Wstat->rxpackets;
1841 		ti->tcpi_wired_rxbytes = inp->inp_Wstat->rxbytes;
1842 		ti->tcpi_wired_txpackets = inp->inp_Wstat->txpackets;
1843 		ti->tcpi_wired_txbytes = inp->inp_Wstat->txbytes;
1844 	}
1845 	tcp_get_connectivity_status(tp, &ti->tcpi_connstatus);
1846 
1847 	ti->tcpi_tfo_syn_data_rcv = !!(tp->t_tfo_stats & TFO_S_SYNDATA_RCV);
1848 	ti->tcpi_tfo_cookie_req_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIEREQ_RECV);
1849 	ti->tcpi_tfo_cookie_sent = !!(tp->t_tfo_stats & TFO_S_COOKIE_SENT);
1850 	ti->tcpi_tfo_cookie_invalid = !!(tp->t_tfo_stats & TFO_S_COOKIE_INVALID);
1851 
1852 	ti->tcpi_tfo_cookie_req = !!(tp->t_tfo_stats & TFO_S_COOKIE_REQ);
1853 	ti->tcpi_tfo_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIE_RCV);
1854 	ti->tcpi_tfo_syn_data_sent = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_SENT);
1855 	ti->tcpi_tfo_syn_data_acked = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_ACKED);
1856 	ti->tcpi_tfo_syn_loss = !!(tp->t_tfo_stats & TFO_S_SYN_LOSS);
1857 	ti->tcpi_tfo_cookie_wrong = !!(tp->t_tfo_stats & TFO_S_COOKIE_WRONG);
1858 	ti->tcpi_tfo_no_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_NO_COOKIE_RCV);
1859 	ti->tcpi_tfo_heuristics_disable = !!(tp->t_tfo_stats & TFO_S_HEURISTICS_DISABLE);
1860 	ti->tcpi_tfo_send_blackhole = !!(tp->t_tfo_stats & TFO_S_SEND_BLACKHOLE);
1861 	ti->tcpi_tfo_recv_blackhole = !!(tp->t_tfo_stats & TFO_S_RECV_BLACKHOLE);
1862 	ti->tcpi_tfo_onebyte_proxy = !!(tp->t_tfo_stats & TFO_S_ONE_BYTE_PROXY);
1863 
1864 	ti->tcpi_ecn_client_setup = !!(tp->ecn_flags & (TE_SETUPSENT | TE_ACE_SETUPSENT));
1865 	ti->tcpi_ecn_server_setup = !!(tp->ecn_flags & (TE_SETUPRECEIVED | TE_ACE_SETUPRECEIVED));
1866 	ti->tcpi_ecn_success = (TCP_ECN_ENABLED(tp) || TCP_ACC_ECN_ON(tp)) ? 1 : 0;
1867 	ti->tcpi_ecn_lost_syn = !!(tp->ecn_flags & TE_LOST_SYN);
1868 	ti->tcpi_ecn_lost_synack = !!(tp->ecn_flags & TE_LOST_SYNACK);
1869 
1870 	ti->tcpi_local_peer = !!(tp->t_flags & TF_LOCAL);
1871 
1872 	if (inp != NULL && inp->inp_last_outifp != NULL) {
1873 		ti->tcpi_last_outif = inp->inp_last_outifp->if_index;
1874 
1875 		if (IFNET_IS_CELLULAR(inp->inp_last_outifp)) {
1876 			ti->tcpi_if_cell = 1;
1877 		}
1878 		if (IFNET_IS_WIFI(inp->inp_last_outifp)) {
1879 			ti->tcpi_if_wifi = 1;
1880 		}
1881 		if (IFNET_IS_WIRED(inp->inp_last_outifp)) {
1882 			ti->tcpi_if_wired = 1;
1883 		}
1884 		if (IFNET_IS_WIFI_INFRA(inp->inp_last_outifp)) {
1885 			ti->tcpi_if_wifi_infra = 1;
1886 		}
1887 		if (inp->inp_last_outifp->if_eflags & IFEF_AWDL) {
1888 			ti->tcpi_if_wifi_awdl = 1;
1889 		}
1890 	}
1891 	if (tp->tcp_cc_index == TCP_CC_ALGO_BACKGROUND_INDEX) {
1892 		ti->tcpi_snd_background = 1;
1893 	}
1894 	if (tcp_recv_bg == 1 || (inp != NULL && inp->inp_socket != NULL &&
1895 	    IS_TCP_RECV_BG(inp->inp_socket))) {
1896 		ti->tcpi_rcv_background = 1;
1897 	}
1898 
1899 	ti->tcpi_ecn_recv_ce = tp->t_ecn_recv_ce;
1900 	ti->tcpi_ecn_recv_cwr = tp->t_ecn_recv_cwr;
1901 
1902 	ti->tcpi_rcvoopack = tp->t_rcvoopack;
1903 	ti->tcpi_pawsdrop = tp->t_pawsdrop;
1904 	ti->tcpi_sack_recovery_episode = tp->t_sack_recovery_episode;
1905 	ti->tcpi_reordered_pkts = tp->t_reordered_pkts;
1906 	ti->tcpi_dsack_sent = tp->t_dsack_sent;
1907 	ti->tcpi_dsack_recvd = tp->t_dsack_recvd;
1908 
1909 	ti->tcpi_client_accecn_state = tp->t_client_accecn_state;
1910 	ti->tcpi_server_accecn_state = tp->t_server_accecn_state;
1911 	ti->tcpi_ecn_capable_packets_sent = tp->t_ecn_capable_packets_sent;
1912 	ti->tcpi_ecn_capable_packets_acked = tp->t_ecn_capable_packets_acked;
1913 	ti->tcpi_ecn_capable_packets_marked = tp->t_ecn_capable_packets_marked;
1914 	ti->tcpi_ecn_capable_packets_lost = tp->t_ecn_capable_packets_lost;
1915 
1916 	ti->tcpi_flow_control_total_time = inp->inp_fadv_total_time;
1917 	ti->tcpi_rcvwnd_limited_total_time = tp->t_rcvwnd_limited_total_time;
1918 }
1919 
1920 __private_extern__ errno_t
tcp_fill_info_for_info_tuple(struct info_tuple * itpl,struct tcp_info * ti)1921 tcp_fill_info_for_info_tuple(struct info_tuple *itpl, struct tcp_info *ti)
1922 {
1923 	struct inpcbinfo *pcbinfo = NULL;
1924 	struct inpcb *inp = NULL;
1925 	struct socket *so;
1926 	struct tcpcb *tp;
1927 
1928 	if (itpl->itpl_proto == IPPROTO_TCP) {
1929 		pcbinfo = &tcbinfo;
1930 	} else {
1931 		return EINVAL;
1932 	}
1933 
1934 	if (itpl->itpl_local_sa.sa_family == AF_INET &&
1935 	    itpl->itpl_remote_sa.sa_family == AF_INET) {
1936 		inp = in_pcblookup_hash(pcbinfo,
1937 		    itpl->itpl_remote_sin.sin_addr,
1938 		    itpl->itpl_remote_sin.sin_port,
1939 		    itpl->itpl_local_sin.sin_addr,
1940 		    itpl->itpl_local_sin.sin_port,
1941 		    0, NULL);
1942 	} else if (itpl->itpl_local_sa.sa_family == AF_INET6 &&
1943 	    itpl->itpl_remote_sa.sa_family == AF_INET6) {
1944 		struct in6_addr ina6_local;
1945 		struct in6_addr ina6_remote;
1946 
1947 		ina6_local = itpl->itpl_local_sin6.sin6_addr;
1948 		if (in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&ina6_local) &&
1949 		    itpl->itpl_local_sin6.sin6_scope_id) {
1950 			ina6_local.s6_addr16[1] = htons((uint16_t)itpl->itpl_local_sin6.sin6_scope_id);
1951 		}
1952 
1953 		ina6_remote = itpl->itpl_remote_sin6.sin6_addr;
1954 		if (in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&ina6_remote) &&
1955 		    itpl->itpl_remote_sin6.sin6_scope_id) {
1956 			ina6_remote.s6_addr16[1] = htons((uint16_t)itpl->itpl_remote_sin6.sin6_scope_id);
1957 		}
1958 
1959 		inp = in6_pcblookup_hash(pcbinfo,
1960 		    &ina6_remote,
1961 		    itpl->itpl_remote_sin6.sin6_port,
1962 		    itpl->itpl_remote_sin6.sin6_scope_id,
1963 		    &ina6_local,
1964 		    itpl->itpl_local_sin6.sin6_port,
1965 		    itpl->itpl_local_sin6.sin6_scope_id,
1966 		    0, NULL);
1967 	} else {
1968 		return EINVAL;
1969 	}
1970 
1971 	if (inp != NULL) {
1972 		if ((so = inp->inp_socket) == NULL) {
1973 			return ENOENT;
1974 		}
1975 		socket_lock(so, 0);
1976 		if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
1977 			socket_unlock(so, 0);
1978 			return ENOENT;
1979 		}
1980 		tp = intotcpcb(inp);
1981 
1982 		tcp_fill_info(tp, ti);
1983 		socket_unlock(so, 0);
1984 
1985 		return 0;
1986 	}
1987 #if SKYWALK
1988 	else {
1989 		/* if no pcb found, check for flowswitch for uTCP flow */
1990 		int error;
1991 		struct nexus_mib_filter nmf = {
1992 			.nmf_type = NXMIB_FLOW,
1993 			.nmf_bitmap = NXMIB_FILTER_INFO_TUPLE,
1994 			.nmf_info_tuple = *itpl,
1995 		};
1996 		struct sk_stats_flow sf;
1997 		size_t len = sizeof(sf);
1998 		error = kernel_sysctlbyname(SK_STATS_FLOW, &sf, &len, &nmf, sizeof(nmf));
1999 		if (error != 0) {
2000 			printf("kernel_sysctlbyname err %d\n", error);
2001 			return error;
2002 		}
2003 		if (len != sizeof(sf)) {
2004 			printf("kernel_sysctlbyname invalid len %zu\n", len);
2005 			return ENOENT;
2006 		}
2007 
2008 		/*
2009 		 * This is what flow tracker can offer right now, which is good
2010 		 * for mDNS TCP keep alive offload.
2011 		 */
2012 		ti->tcpi_snd_nxt = sf.sf_lseq;
2013 		ti->tcpi_rcv_nxt = sf.sf_rseq;
2014 		ti->tcpi_rcv_space = (uint32_t)(sf.sf_lmax_win << sf.sf_lwscale);
2015 		ti->tcpi_rcv_wscale = sf.sf_lwscale;
2016 		ti->tcpi_last_outif = (int32_t)sf.sf_if_index;
2017 
2018 		return 0;
2019 	}
2020 #endif /* SKYWALK */
2021 
2022 	return ENOENT;
2023 }
2024 
2025 static void
tcp_connection_fill_info(struct tcpcb * tp,struct tcp_connection_info * tci)2026 tcp_connection_fill_info(struct tcpcb *tp, struct tcp_connection_info *tci)
2027 {
2028 	struct inpcb *inp = tp->t_inpcb;
2029 
2030 	bzero(tci, sizeof(*tci));
2031 	tci->tcpi_state = (uint8_t)tp->t_state;
2032 
2033 	if (TSTMP_SUPPORTED(tp)) {
2034 		tci->tcpi_options |= TCPCI_OPT_TIMESTAMPS;
2035 	}
2036 	if (SACK_ENABLED(tp)) {
2037 		tci->tcpi_options |= TCPCI_OPT_SACK;
2038 	}
2039 	if (TCP_WINDOW_SCALE_ENABLED(tp)) {
2040 		tci->tcpi_options |= TCPCI_OPT_WSCALE;
2041 		tci->tcpi_snd_wscale = tp->snd_scale;
2042 		tci->tcpi_rcv_wscale = tp->rcv_scale;
2043 	}
2044 	if (TCP_ECN_ENABLED(tp)) {
2045 		tci->tcpi_options |= TCPCI_OPT_ECN;
2046 	}
2047 	if (IN_FASTRECOVERY(tp) || tp->t_rxtshift > 0) {
2048 		tci->tcpi_flags |= TCPCI_FLAG_LOSSRECOVERY;
2049 	}
2050 	if (tp->t_flagsext & TF_PKTS_REORDERED) {
2051 		tci->tcpi_flags |= TCPCI_FLAG_REORDERING_DETECTED;
2052 	}
2053 	tci->tcpi_rto = tp->t_timer[TCPT_REXMT] > 0 ? tp->t_rxtcur : 0;
2054 	tci->tcpi_maxseg = tp->t_maxseg;
2055 	tci->tcpi_snd_ssthresh = tp->snd_ssthresh;
2056 	tci->tcpi_snd_cwnd = tp->snd_cwnd;
2057 	tci->tcpi_snd_wnd = tp->snd_wnd;
2058 	if (inp != NULL && inp->inp_socket != NULL) {
2059 		tci->tcpi_snd_sbbytes = inp->inp_socket->so_snd.sb_cc;
2060 	}
2061 	tci->tcpi_rcv_wnd = tp->rcv_adv > tp->rcv_nxt ? tp->rcv_adv - tp->rcv_nxt : 0;
2062 	tci->tcpi_rttcur = tp->t_rttcur;
2063 	tci->tcpi_srtt = (tp->t_srtt >> TCP_RTT_SHIFT);
2064 	tci->tcpi_rttvar = (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
2065 	tci->tcpi_txpackets = inp != NULL ? inp->inp_stat->txpackets : 0;
2066 	tci->tcpi_txbytes = inp != NULL ? inp->inp_stat->txbytes : 0;
2067 	tci->tcpi_txretransmitbytes = tp->t_stat.txretransmitbytes;
2068 	tci->tcpi_txretransmitpackets = tp->t_stat.rxmitpkts;
2069 	tci->tcpi_rxpackets = inp != NULL ? inp->inp_stat->rxpackets : 0;
2070 	tci->tcpi_rxbytes = inp != NULL ? inp->inp_stat->rxbytes : 0;
2071 	tci->tcpi_rxoutoforderbytes = tp->t_stat.rxoutoforderbytes;
2072 
2073 	tci->tcpi_tfo_syn_data_rcv = !!(tp->t_tfo_stats & TFO_S_SYNDATA_RCV);
2074 	tci->tcpi_tfo_cookie_req_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIEREQ_RECV);
2075 	tci->tcpi_tfo_cookie_sent = !!(tp->t_tfo_stats & TFO_S_COOKIE_SENT);
2076 	tci->tcpi_tfo_cookie_invalid = !!(tp->t_tfo_stats & TFO_S_COOKIE_INVALID);
2077 	tci->tcpi_tfo_cookie_req = !!(tp->t_tfo_stats & TFO_S_COOKIE_REQ);
2078 	tci->tcpi_tfo_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIE_RCV);
2079 	tci->tcpi_tfo_syn_data_sent = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_SENT);
2080 	tci->tcpi_tfo_syn_data_acked = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_ACKED);
2081 	tci->tcpi_tfo_syn_loss = !!(tp->t_tfo_stats & TFO_S_SYN_LOSS);
2082 	tci->tcpi_tfo_cookie_wrong = !!(tp->t_tfo_stats & TFO_S_COOKIE_WRONG);
2083 	tci->tcpi_tfo_no_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_NO_COOKIE_RCV);
2084 	tci->tcpi_tfo_heuristics_disable = !!(tp->t_tfo_stats & TFO_S_HEURISTICS_DISABLE);
2085 	tci->tcpi_tfo_send_blackhole = !!(tp->t_tfo_stats & TFO_S_SEND_BLACKHOLE);
2086 	tci->tcpi_tfo_recv_blackhole = !!(tp->t_tfo_stats & TFO_S_RECV_BLACKHOLE);
2087 	tci->tcpi_tfo_onebyte_proxy = !!(tp->t_tfo_stats & TFO_S_ONE_BYTE_PROXY);
2088 }
2089 
2090 
2091 __private_extern__ int
tcp_sysctl_info(__unused struct sysctl_oid * oidp,__unused void * arg1,__unused int arg2,struct sysctl_req * req)2092 tcp_sysctl_info(__unused struct sysctl_oid *oidp, __unused void *arg1, __unused int arg2, struct sysctl_req *req)
2093 {
2094 	int error;
2095 	struct tcp_info ti = {};
2096 	struct info_tuple itpl;
2097 
2098 	if (req->newptr == USER_ADDR_NULL) {
2099 		return EINVAL;
2100 	}
2101 	if (req->newlen < sizeof(struct info_tuple)) {
2102 		return EINVAL;
2103 	}
2104 	error = SYSCTL_IN(req, &itpl, sizeof(struct info_tuple));
2105 	if (error != 0) {
2106 		return error;
2107 	}
2108 	error = tcp_fill_info_for_info_tuple(&itpl, &ti);
2109 	if (error != 0) {
2110 		return error;
2111 	}
2112 	error = SYSCTL_OUT(req, &ti, sizeof(struct tcp_info));
2113 	if (error != 0) {
2114 		return error;
2115 	}
2116 
2117 	return 0;
2118 }
2119 
2120 static int
tcp_lookup_peer_pid_locked(struct socket * so,pid_t * out_pid)2121 tcp_lookup_peer_pid_locked(struct socket *so, pid_t *out_pid)
2122 {
2123 	int error = EHOSTUNREACH;
2124 	*out_pid = -1;
2125 	if ((so->so_state & SS_ISCONNECTED) == 0) {
2126 		return ENOTCONN;
2127 	}
2128 
2129 	struct inpcb    *inp = (struct inpcb*)so->so_pcb;
2130 	uint16_t                lport = inp->inp_lport;
2131 	uint16_t                fport = inp->inp_fport;
2132 	uint32_t                                fifscope = inp->inp_fifscope;
2133 	uint32_t                                lifscope = inp->inp_lifscope;
2134 
2135 	struct inpcb    *finp = NULL;
2136 	struct  in6_addr laddr6, faddr6;
2137 	struct in_addr laddr4, faddr4;
2138 
2139 	if (inp->inp_vflag & INP_IPV6) {
2140 		laddr6 = inp->in6p_laddr;
2141 		faddr6 = inp->in6p_faddr;
2142 	} else if (inp->inp_vflag & INP_IPV4) {
2143 		laddr4 = inp->inp_laddr;
2144 		faddr4 = inp->inp_faddr;
2145 	}
2146 
2147 	socket_unlock(so, 0);
2148 	if (inp->inp_vflag & INP_IPV6) {
2149 		finp = in6_pcblookup_hash(&tcbinfo, &laddr6, lport, lifscope, &faddr6, fport, fifscope, 0, NULL);
2150 	} else if (inp->inp_vflag & INP_IPV4) {
2151 		finp = in_pcblookup_hash(&tcbinfo, laddr4, lport, faddr4, fport, 0, NULL);
2152 	}
2153 
2154 	if (finp) {
2155 		*out_pid = finp->inp_socket->last_pid;
2156 		error = 0;
2157 		in_pcb_checkstate(finp, WNT_RELEASE, 0);
2158 	}
2159 	socket_lock(so, 0);
2160 
2161 	return error;
2162 }
2163 
2164 void
tcp_getconninfo(struct socket * so,struct conninfo_tcp * tcp_ci)2165 tcp_getconninfo(struct socket *so, struct conninfo_tcp *tcp_ci)
2166 {
2167 	tcp_fill_info(sototcpcb(so), &tcp_ci->tcpci_tcp_info);
2168 }
2169 
2170 void
tcp_clear_keep_alive_offload(struct socket * so)2171 tcp_clear_keep_alive_offload(struct socket *so)
2172 {
2173 	struct inpcb *inp;
2174 	struct ifnet *ifp;
2175 
2176 	inp = sotoinpcb(so);
2177 	if (inp == NULL) {
2178 		return;
2179 	}
2180 
2181 	if ((inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD) == 0) {
2182 		return;
2183 	}
2184 
2185 	ifp = inp->inp_boundifp != NULL ? inp->inp_boundifp :
2186 	    inp->inp_last_outifp;
2187 	if (ifp == NULL) {
2188 		panic("%s: so %p inp %p ifp NULL",
2189 		    __func__, so, inp);
2190 	}
2191 
2192 	ifnet_lock_exclusive(ifp);
2193 
2194 	if (ifp->if_tcp_kao_cnt == 0) {
2195 		panic("%s: so %p inp %p ifp %p if_tcp_kao_cnt == 0",
2196 		    __func__, so, inp, ifp);
2197 	}
2198 	ifp->if_tcp_kao_cnt--;
2199 	inp->inp_flags2 &= ~INP2_KEEPALIVE_OFFLOAD;
2200 
2201 	ifnet_lock_done(ifp);
2202 }
2203 
2204 static int
tcp_set_keep_alive_offload(struct socket * so,struct proc * proc)2205 tcp_set_keep_alive_offload(struct socket *so, struct proc *proc)
2206 {
2207 	int error = 0;
2208 	struct inpcb *inp;
2209 	struct ifnet *ifp;
2210 
2211 	inp = sotoinpcb(so);
2212 	if (inp == NULL) {
2213 		return ECONNRESET;
2214 	}
2215 	if ((inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD) != 0) {
2216 		return 0;
2217 	}
2218 
2219 	ifp = inp->inp_boundifp != NULL ? inp->inp_boundifp :
2220 	    inp->inp_last_outifp;
2221 	if (ifp == NULL) {
2222 		error = ENXIO;
2223 		os_log_info(OS_LOG_DEFAULT,
2224 		    "%s: error %d for proc %s[%u] out ifp is not set\n",
2225 		    __func__, error,
2226 		    proc != NULL ? proc->p_comm : "kernel",
2227 		    proc != NULL ? proc_getpid(proc) : 0);
2228 		return ENXIO;
2229 	}
2230 
2231 	error = if_get_tcp_kao_max(ifp);
2232 	if (error != 0) {
2233 		return error;
2234 	}
2235 
2236 	ifnet_lock_exclusive(ifp);
2237 	if (ifp->if_tcp_kao_cnt < ifp->if_tcp_kao_max) {
2238 		ifp->if_tcp_kao_cnt++;
2239 		inp->inp_flags2 |= INP2_KEEPALIVE_OFFLOAD;
2240 	} else {
2241 		error = ETOOMANYREFS;
2242 		os_log_info(OS_LOG_DEFAULT,
2243 		    "%s: error %d for proc %s[%u] if_tcp_kao_max %u\n",
2244 		    __func__, error,
2245 		    proc != NULL ? proc->p_comm : "kernel",
2246 		    proc != NULL ? proc_getpid(proc) : 0,
2247 		    ifp->if_tcp_kao_max);
2248 	}
2249 	ifnet_lock_done(ifp);
2250 
2251 	return error;
2252 }
2253 
2254 /*
2255  * The new sockopt interface makes it possible for us to block in the
2256  * copyin/out step (if we take a page fault).  Taking a page fault at
2257  * splnet() is probably a Bad Thing.  (Since sockets and pcbs both now
2258  * use TSM, there probably isn't any need for this function to run at
2259  * splnet() any more.  This needs more examination.)
2260  */
2261 int
tcp_ctloutput(struct socket * so,struct sockopt * sopt)2262 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
2263 {
2264 	int     error = 0, opt = 0, optval = 0;
2265 	struct  inpcb *inp;
2266 	struct  tcpcb *tp;
2267 
2268 	inp = sotoinpcb(so);
2269 	if (inp == NULL) {
2270 		return ECONNRESET;
2271 	}
2272 	/* Allow <SOL_SOCKET,SO_FLUSH/SO_TRAFFIC_MGT_BACKGROUND> at this level */
2273 	if (sopt->sopt_level != IPPROTO_TCP &&
2274 	    !(sopt->sopt_level == SOL_SOCKET && (sopt->sopt_name == SO_FLUSH ||
2275 	    sopt->sopt_name == SO_TRAFFIC_MGT_BACKGROUND))) {
2276 		if (SOCK_CHECK_DOM(so, PF_INET6)) {
2277 			error = ip6_ctloutput(so, sopt);
2278 		} else {
2279 			error = ip_ctloutput(so, sopt);
2280 		}
2281 		return error;
2282 	}
2283 	tp = intotcpcb(inp);
2284 	if (tp == NULL) {
2285 		return ECONNRESET;
2286 	}
2287 
2288 	calculate_tcp_clock();
2289 
2290 	switch (sopt->sopt_dir) {
2291 	case SOPT_SET:
2292 		switch (sopt->sopt_name) {
2293 		case TCP_NODELAY:
2294 		case TCP_NOOPT:
2295 		case TCP_NOPUSH:
2296 			error = sooptcopyin(sopt, &optval, sizeof optval,
2297 			    sizeof optval);
2298 			if (error) {
2299 				break;
2300 			}
2301 
2302 			switch (sopt->sopt_name) {
2303 			case TCP_NODELAY:
2304 				opt = TF_NODELAY;
2305 				break;
2306 			case TCP_NOOPT:
2307 				opt = TF_NOOPT;
2308 				break;
2309 			case TCP_NOPUSH:
2310 				opt = TF_NOPUSH;
2311 				break;
2312 			default:
2313 				opt = 0; /* dead code to fool gcc */
2314 				break;
2315 			}
2316 
2317 			if (optval) {
2318 				tp->t_flags |= opt;
2319 			} else {
2320 				tp->t_flags &= ~opt;
2321 			}
2322 			break;
2323 		case TCP_RXT_FINDROP:
2324 		case TCP_NOTIMEWAIT:
2325 			error = sooptcopyin(sopt, &optval, sizeof optval,
2326 			    sizeof optval);
2327 			if (error) {
2328 				break;
2329 			}
2330 			switch (sopt->sopt_name) {
2331 			case TCP_RXT_FINDROP:
2332 				opt = TF_RXTFINDROP;
2333 				break;
2334 			case TCP_NOTIMEWAIT:
2335 				opt = TF_NOTIMEWAIT;
2336 				break;
2337 			default:
2338 				opt = 0;
2339 				break;
2340 			}
2341 			if (optval) {
2342 				tp->t_flagsext |= opt;
2343 			} else {
2344 				tp->t_flagsext &= ~opt;
2345 			}
2346 			break;
2347 		case TCP_MEASURE_SND_BW:
2348 			error = sooptcopyin(sopt, &optval, sizeof optval,
2349 			    sizeof optval);
2350 			if (error) {
2351 				break;
2352 			}
2353 			opt = TF_MEASURESNDBW;
2354 			if (optval) {
2355 				if (tp->t_bwmeas == NULL) {
2356 					tp->t_bwmeas = tcp_bwmeas_alloc(tp);
2357 					if (tp->t_bwmeas == NULL) {
2358 						error = ENOMEM;
2359 						break;
2360 					}
2361 				}
2362 				tp->t_flagsext |= opt;
2363 			} else {
2364 				tp->t_flagsext &= ~opt;
2365 				/* Reset snd bw measurement state */
2366 				tp->t_flagsext &= ~(TF_BWMEAS_INPROGRESS);
2367 				if (tp->t_bwmeas != NULL) {
2368 					tcp_bwmeas_free(tp);
2369 				}
2370 			}
2371 			break;
2372 		case TCP_MEASURE_BW_BURST: {
2373 			struct tcp_measure_bw_burst in;
2374 			uint32_t minpkts, maxpkts;
2375 			bzero(&in, sizeof(in));
2376 
2377 			error = sooptcopyin(sopt, &in, sizeof(in),
2378 			    sizeof(in));
2379 			if (error) {
2380 				break;
2381 			}
2382 			if ((tp->t_flagsext & TF_MEASURESNDBW) == 0 ||
2383 			    tp->t_bwmeas == NULL) {
2384 				error = EINVAL;
2385 				break;
2386 			}
2387 			minpkts = (in.min_burst_size != 0) ? in.min_burst_size :
2388 			    tp->t_bwmeas->bw_minsizepkts;
2389 			maxpkts = (in.max_burst_size != 0) ? in.max_burst_size :
2390 			    tp->t_bwmeas->bw_maxsizepkts;
2391 			if (minpkts > maxpkts) {
2392 				error = EINVAL;
2393 				break;
2394 			}
2395 			tp->t_bwmeas->bw_minsizepkts = minpkts;
2396 			tp->t_bwmeas->bw_maxsizepkts = maxpkts;
2397 			tp->t_bwmeas->bw_minsize = (minpkts * tp->t_maxseg);
2398 			tp->t_bwmeas->bw_maxsize = (maxpkts * tp->t_maxseg);
2399 			break;
2400 		}
2401 		case TCP_MAXSEG:
2402 			error = sooptcopyin(sopt, &optval, sizeof optval,
2403 			    sizeof optval);
2404 			if (error) {
2405 				break;
2406 			}
2407 
2408 			if (optval > 0 && optval <= tp->t_maxseg &&
2409 			    optval + 40 >= tcp_minmss) {
2410 				tp->t_maxseg = optval;
2411 			} else {
2412 				error = EINVAL;
2413 			}
2414 			break;
2415 
2416 		case TCP_KEEPALIVE:
2417 			error = sooptcopyin(sopt, &optval, sizeof optval,
2418 			    sizeof optval);
2419 			if (error) {
2420 				break;
2421 			}
2422 			if (optval < 0 || optval > UINT32_MAX / TCP_RETRANSHZ) {
2423 				error = EINVAL;
2424 			} else {
2425 				tp->t_keepidle = optval * TCP_RETRANSHZ;
2426 				/* reset the timer to new value */
2427 				if (TCPS_HAVEESTABLISHED(tp->t_state)) {
2428 					tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
2429 					    TCP_CONN_KEEPIDLE(tp));
2430 					tcp_check_timer_state(tp);
2431 				}
2432 			}
2433 			break;
2434 
2435 		case TCP_CONNECTIONTIMEOUT:
2436 			error = sooptcopyin(sopt, &optval, sizeof optval,
2437 			    sizeof optval);
2438 			if (error) {
2439 				break;
2440 			}
2441 			if (optval < 0 || optval > UINT32_MAX / TCP_RETRANSHZ) {
2442 				error = EINVAL;
2443 			} else {
2444 				tp->t_keepinit = optval * TCP_RETRANSHZ;
2445 				if (tp->t_state == TCPS_SYN_RECEIVED ||
2446 				    tp->t_state == TCPS_SYN_SENT) {
2447 					tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
2448 					    TCP_CONN_KEEPINIT(tp));
2449 					tcp_check_timer_state(tp);
2450 				}
2451 			}
2452 			break;
2453 
2454 		case TCP_KEEPINTVL:
2455 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2456 			    sizeof(optval));
2457 			if (error) {
2458 				break;
2459 			}
2460 			if (optval < 0 || optval > UINT32_MAX / TCP_RETRANSHZ) {
2461 				error = EINVAL;
2462 			} else {
2463 				tp->t_keepintvl = optval * TCP_RETRANSHZ;
2464 				if (tp->t_state == TCPS_FIN_WAIT_2 &&
2465 				    TCP_CONN_MAXIDLE(tp) > 0) {
2466 					tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp,
2467 					    TCP_CONN_MAXIDLE(tp));
2468 					tcp_check_timer_state(tp);
2469 				}
2470 			}
2471 			break;
2472 
2473 		case TCP_KEEPCNT:
2474 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2475 			    sizeof(optval));
2476 			if (error) {
2477 				break;
2478 			}
2479 			if (optval < 0 || optval > INT32_MAX) {
2480 				error = EINVAL;
2481 			} else {
2482 				tp->t_keepcnt = optval;
2483 				if (tp->t_state == TCPS_FIN_WAIT_2 &&
2484 				    TCP_CONN_MAXIDLE(tp) > 0) {
2485 					tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp,
2486 					    TCP_CONN_MAXIDLE(tp));
2487 					tcp_check_timer_state(tp);
2488 				}
2489 			}
2490 			break;
2491 
2492 		case TCP_KEEPALIVE_OFFLOAD:
2493 			if ((error = priv_check_cred(kauth_cred_get(),
2494 			    PRIV_NETINET_TCP_KA_OFFLOAD, 0)) != 0) {
2495 				break;
2496 			}
2497 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2498 			    sizeof(optval));
2499 			if (error) {
2500 				break;
2501 			}
2502 			if (optval < 0 || optval > INT32_MAX) {
2503 				error = EINVAL;
2504 				break;
2505 			}
2506 			if (optval != 0) {
2507 				error = tcp_set_keep_alive_offload(so,
2508 				    sopt->sopt_p);
2509 			} else {
2510 				tcp_clear_keep_alive_offload(so);
2511 			}
2512 			break;
2513 
2514 		case PERSIST_TIMEOUT:
2515 			error = sooptcopyin(sopt, &optval, sizeof optval,
2516 			    sizeof optval);
2517 			if (error) {
2518 				break;
2519 			}
2520 			if (optval < 0) {
2521 				error = EINVAL;
2522 			} else {
2523 				tp->t_persist_timeout = optval * TCP_RETRANSHZ;
2524 			}
2525 			break;
2526 		case TCP_RXT_CONNDROPTIME:
2527 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2528 			    sizeof(optval));
2529 			if (error) {
2530 				break;
2531 			}
2532 			if (optval < 0) {
2533 				error = EINVAL;
2534 			} else {
2535 				tp->t_rxt_conndroptime = optval * TCP_RETRANSHZ;
2536 			}
2537 			break;
2538 		case TCP_NOTSENT_LOWAT:
2539 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2540 			    sizeof(optval));
2541 			if (error) {
2542 				break;
2543 			}
2544 			if (optval < 0) {
2545 				error = EINVAL;
2546 				break;
2547 			} else {
2548 				if (optval == 0) {
2549 					so->so_flags &= ~(SOF_NOTSENT_LOWAT);
2550 					tp->t_notsent_lowat = 0;
2551 				} else {
2552 					so->so_flags |= SOF_NOTSENT_LOWAT;
2553 					tp->t_notsent_lowat = optval;
2554 				}
2555 			}
2556 			break;
2557 		case TCP_ADAPTIVE_READ_TIMEOUT:
2558 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2559 			    sizeof(optval));
2560 			if (error) {
2561 				break;
2562 			}
2563 			if (optval < 0 ||
2564 			    optval > TCP_ADAPTIVE_TIMEOUT_MAX) {
2565 				error = EINVAL;
2566 				break;
2567 			} else if (optval == 0) {
2568 				tp->t_adaptive_rtimo = 0;
2569 				tcp_keepalive_reset(tp);
2570 
2571 				if (tp->t_mpsub) {
2572 					mptcp_reset_keepalive(tp);
2573 				}
2574 			} else {
2575 				tp->t_adaptive_rtimo = (uint8_t)optval;
2576 			}
2577 			break;
2578 		case TCP_ADAPTIVE_WRITE_TIMEOUT:
2579 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2580 			    sizeof(optval));
2581 			if (error) {
2582 				break;
2583 			}
2584 			if (optval < 0 ||
2585 			    optval > TCP_ADAPTIVE_TIMEOUT_MAX) {
2586 				error = EINVAL;
2587 				break;
2588 			} else {
2589 				tp->t_adaptive_wtimo = (uint8_t)optval;
2590 			}
2591 			break;
2592 		case TCP_SENDMOREACKS:
2593 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2594 			    sizeof(optval));
2595 			if (error) {
2596 				break;
2597 			}
2598 			if (optval < 0 || optval > 1) {
2599 				error = EINVAL;
2600 			} else if (optval == 0) {
2601 				tp->t_flagsext &= ~(TF_NOSTRETCHACK);
2602 			} else {
2603 				tp->t_flagsext |= TF_NOSTRETCHACK;
2604 			}
2605 			break;
2606 		case TCP_DISABLE_BLACKHOLE_DETECTION:
2607 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2608 			    sizeof(optval));
2609 			if (error) {
2610 				break;
2611 			}
2612 			if (optval < 0 || optval > 1) {
2613 				error = EINVAL;
2614 			} else if (optval == 0) {
2615 				tp->t_flagsext &= ~TF_NOBLACKHOLE_DETECTION;
2616 			} else {
2617 				tp->t_flagsext |= TF_NOBLACKHOLE_DETECTION;
2618 				if ((tp->t_flags & TF_BLACKHOLE) &&
2619 				    tp->t_pmtud_saved_maxopd > 0) {
2620 					tcp_pmtud_revert_segment_size(tp);
2621 				}
2622 			}
2623 			break;
2624 		case TCP_FASTOPEN:
2625 			if (!(tcp_fastopen & TCP_FASTOPEN_SERVER)) {
2626 				error = ENOTSUP;
2627 				break;
2628 			}
2629 
2630 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2631 			    sizeof(optval));
2632 			if (error) {
2633 				break;
2634 			}
2635 			if (optval < 0 || optval > 1) {
2636 				error = EINVAL;
2637 				break;
2638 			}
2639 			if (tp->t_state != TCPS_LISTEN) {
2640 				error =  EINVAL;
2641 				break;
2642 			}
2643 			if (optval) {
2644 				tp->t_flagsext |= TF_FASTOPEN;
2645 			} else {
2646 				tcp_disable_tfo(tp);
2647 			}
2648 			break;
2649 		case TCP_FASTOPEN_FORCE_HEURISTICS:
2650 
2651 			break;
2652 		case TCP_FASTOPEN_FORCE_ENABLE:
2653 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2654 			    sizeof(optval));
2655 
2656 			if (error) {
2657 				break;
2658 			}
2659 			if (optval < 0 || optval > 1) {
2660 				error = EINVAL;
2661 				break;
2662 			}
2663 
2664 			if (tp->t_state != TCPS_CLOSED) {
2665 				error =  EINVAL;
2666 				break;
2667 			}
2668 			if (optval) {
2669 				tp->t_flagsext |= TF_FASTOPEN_FORCE_ENABLE;
2670 			} else {
2671 				tp->t_flagsext &= ~TF_FASTOPEN_FORCE_ENABLE;
2672 			}
2673 
2674 			break;
2675 		case TCP_ENABLE_ECN:
2676 			error = sooptcopyin(sopt, &optval, sizeof optval,
2677 			    sizeof optval);
2678 			if (error) {
2679 				break;
2680 			}
2681 			if (optval) {
2682 				tp->ecn_flags |= TE_ECN_MODE_ENABLE;
2683 				tp->ecn_flags &= ~TE_ECN_MODE_DISABLE;
2684 			} else {
2685 				tp->ecn_flags &= ~TE_ECN_MODE_ENABLE;
2686 				tp->ecn_flags |= TE_ECN_MODE_DISABLE;
2687 			}
2688 			break;
2689 		case TCP_ECN_MODE:
2690 			error = sooptcopyin(sopt, &optval, sizeof optval,
2691 			    sizeof optval);
2692 			if (error) {
2693 				break;
2694 			}
2695 			if (optval == ECN_MODE_DEFAULT) {
2696 				tp->ecn_flags &= ~TE_ECN_MODE_ENABLE;
2697 				tp->ecn_flags &= ~TE_ECN_MODE_DISABLE;
2698 			} else if (optval == ECN_MODE_ENABLE) {
2699 				tp->ecn_flags |= TE_ECN_MODE_ENABLE;
2700 				tp->ecn_flags &= ~TE_ECN_MODE_DISABLE;
2701 			} else if (optval == ECN_MODE_DISABLE) {
2702 				tp->ecn_flags &= ~TE_ECN_MODE_ENABLE;
2703 				tp->ecn_flags |= TE_ECN_MODE_DISABLE;
2704 			} else {
2705 				error = EINVAL;
2706 			}
2707 			break;
2708 		case TCP_ENABLE_L4S:
2709 			error = sooptcopyin(sopt, &optval, sizeof optval,
2710 			    sizeof optval);
2711 			if (error) {
2712 				break;
2713 			}
2714 			if (optval < 0 || optval > 1) {
2715 				error = EINVAL;
2716 				break;
2717 			}
2718 			if (tp->t_state != TCPS_CLOSED) {
2719 				error =  EINVAL;
2720 				break;
2721 			}
2722 			if (optval == 1) {
2723 				tp->t_flagsext |= TF_L4S_ENABLED;
2724 				tp->t_flagsext &= ~TF_L4S_DISABLED;
2725 			} else {
2726 				tp->t_flagsext &= ~TF_L4S_ENABLED;
2727 				tp->t_flagsext |= TF_L4S_DISABLED;
2728 			}
2729 			break;
2730 		case TCP_NOTIFY_ACKNOWLEDGEMENT:
2731 			error = sooptcopyin(sopt, &optval,
2732 			    sizeof(optval), sizeof(optval));
2733 			if (error) {
2734 				break;
2735 			}
2736 			if (optval <= 0) {
2737 				error = EINVAL;
2738 				break;
2739 			}
2740 			if (tp->t_notify_ack_count >= TCP_MAX_NOTIFY_ACK) {
2741 				error = ETOOMANYREFS;
2742 				break;
2743 			}
2744 
2745 			/*
2746 			 * validate that the given marker id is not
2747 			 * a duplicate to avoid ambiguity
2748 			 */
2749 			if ((error = tcp_notify_ack_id_valid(tp, so,
2750 			    optval)) != 0) {
2751 				break;
2752 			}
2753 			error = tcp_add_notify_ack_marker(tp, optval);
2754 			break;
2755 		case SO_FLUSH:
2756 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
2757 			    sizeof(optval))) != 0) {
2758 				break;
2759 			}
2760 
2761 			error = inp_flush(inp, optval);
2762 			break;
2763 
2764 		case SO_TRAFFIC_MGT_BACKGROUND:
2765 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
2766 			    sizeof(optval))) != 0) {
2767 				break;
2768 			}
2769 
2770 			if (optval) {
2771 				socket_set_traffic_mgt_flags_locked(so,
2772 				    TRAFFIC_MGT_SO_BACKGROUND);
2773 			} else {
2774 				socket_clear_traffic_mgt_flags_locked(so,
2775 				    TRAFFIC_MGT_SO_BACKGROUND);
2776 			}
2777 			break;
2778 		case TCP_RXT_MINIMUM_TIMEOUT:
2779 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2780 			    sizeof(optval));
2781 			if (error) {
2782 				break;
2783 			}
2784 			if (optval < 0) {
2785 				error = EINVAL;
2786 				break;
2787 			}
2788 			if (optval == 0) {
2789 				tp->t_rxt_minimum_timeout = 0;
2790 			} else {
2791 				tp->t_rxt_minimum_timeout = min(optval,
2792 				    TCP_RXT_MINIMUM_TIMEOUT_LIMIT);
2793 				/* convert to milliseconds */
2794 				tp->t_rxt_minimum_timeout *= TCP_RETRANSHZ;
2795 			}
2796 			break;
2797 		default:
2798 			error = ENOPROTOOPT;
2799 			break;
2800 		}
2801 		break;
2802 
2803 	case SOPT_GET:
2804 		switch (sopt->sopt_name) {
2805 		case TCP_NODELAY:
2806 			optval = tp->t_flags & TF_NODELAY;
2807 			break;
2808 		case TCP_MAXSEG:
2809 			optval = tp->t_maxseg;
2810 			break;
2811 		case TCP_KEEPALIVE:
2812 			if (tp->t_keepidle > 0) {
2813 				optval = tp->t_keepidle / TCP_RETRANSHZ;
2814 			} else {
2815 				optval = tcp_keepidle  / TCP_RETRANSHZ;
2816 			}
2817 			break;
2818 		case TCP_KEEPINTVL:
2819 			if (tp->t_keepintvl > 0) {
2820 				optval = tp->t_keepintvl / TCP_RETRANSHZ;
2821 			} else {
2822 				optval = tcp_keepintvl / TCP_RETRANSHZ;
2823 			}
2824 			break;
2825 		case TCP_KEEPCNT:
2826 			if (tp->t_keepcnt > 0) {
2827 				optval = tp->t_keepcnt;
2828 			} else {
2829 				optval = tcp_keepcnt;
2830 			}
2831 			break;
2832 		case TCP_KEEPALIVE_OFFLOAD:
2833 			optval = !!(inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD);
2834 			break;
2835 		case TCP_NOOPT:
2836 			optval = tp->t_flags & TF_NOOPT;
2837 			break;
2838 		case TCP_NOPUSH:
2839 			optval = tp->t_flags & TF_NOPUSH;
2840 			break;
2841 		case TCP_ENABLE_ECN:
2842 			optval = (tp->ecn_flags & TE_ECN_MODE_ENABLE) ? 1 : 0;
2843 			break;
2844 		case TCP_ECN_MODE:
2845 			if (tp->ecn_flags & TE_ECN_MODE_ENABLE) {
2846 				optval = ECN_MODE_ENABLE;
2847 			} else if (tp->ecn_flags & TE_ECN_MODE_DISABLE) {
2848 				optval = ECN_MODE_DISABLE;
2849 			} else {
2850 				optval = ECN_MODE_DEFAULT;
2851 			}
2852 			break;
2853 		case TCP_ENABLE_L4S:
2854 			optval = (tp->t_flagsext & TF_L4S_ENABLED) ? 1 : 0;
2855 			break;
2856 		case TCP_CONNECTIONTIMEOUT:
2857 			optval = tp->t_keepinit / TCP_RETRANSHZ;
2858 			break;
2859 		case PERSIST_TIMEOUT:
2860 			optval = tp->t_persist_timeout / TCP_RETRANSHZ;
2861 			break;
2862 		case TCP_RXT_CONNDROPTIME:
2863 			optval = tp->t_rxt_conndroptime / TCP_RETRANSHZ;
2864 			break;
2865 		case TCP_RXT_FINDROP:
2866 			optval = tp->t_flagsext & TF_RXTFINDROP;
2867 			break;
2868 		case TCP_NOTIMEWAIT:
2869 			optval = (tp->t_flagsext & TF_NOTIMEWAIT) ? 1 : 0;
2870 			break;
2871 		case TCP_FASTOPEN:
2872 			if (tp->t_state != TCPS_LISTEN ||
2873 			    !(tcp_fastopen & TCP_FASTOPEN_SERVER)) {
2874 				error = ENOTSUP;
2875 				break;
2876 			}
2877 			optval = tfo_enabled(tp);
2878 			break;
2879 		case TCP_FASTOPEN_FORCE_HEURISTICS:
2880 			optval = 0;
2881 			break;
2882 		case TCP_FASTOPEN_FORCE_ENABLE:
2883 			optval = (tp->t_flagsext & TF_FASTOPEN_FORCE_ENABLE) ? 1 : 0;
2884 			break;
2885 		case TCP_MEASURE_SND_BW:
2886 			optval = tp->t_flagsext & TF_MEASURESNDBW;
2887 			break;
2888 		case TCP_INFO: {
2889 			struct tcp_info ti;
2890 
2891 			tcp_fill_info(tp, &ti);
2892 			error = sooptcopyout(sopt, &ti, sizeof(struct tcp_info));
2893 			goto done;
2894 			/* NOT REACHED */
2895 		}
2896 		case TCP_CONNECTION_INFO: {
2897 			struct tcp_connection_info tci;
2898 			tcp_connection_fill_info(tp, &tci);
2899 			error = sooptcopyout(sopt, &tci,
2900 			    sizeof(struct tcp_connection_info));
2901 			goto done;
2902 		}
2903 		case TCP_MEASURE_BW_BURST: {
2904 			struct tcp_measure_bw_burst out = {};
2905 			if ((tp->t_flagsext & TF_MEASURESNDBW) == 0 ||
2906 			    tp->t_bwmeas == NULL) {
2907 				error = EINVAL;
2908 				break;
2909 			}
2910 			out.min_burst_size = tp->t_bwmeas->bw_minsizepkts;
2911 			out.max_burst_size = tp->t_bwmeas->bw_maxsizepkts;
2912 			error = sooptcopyout(sopt, &out, sizeof(out));
2913 			goto done;
2914 		}
2915 		case TCP_NOTSENT_LOWAT:
2916 			if ((so->so_flags & SOF_NOTSENT_LOWAT) != 0) {
2917 				optval = tp->t_notsent_lowat;
2918 			} else {
2919 				optval = 0;
2920 			}
2921 			break;
2922 		case TCP_SENDMOREACKS:
2923 			if (tp->t_flagsext & TF_NOSTRETCHACK) {
2924 				optval = 1;
2925 			} else {
2926 				optval = 0;
2927 			}
2928 			break;
2929 		case TCP_DISABLE_BLACKHOLE_DETECTION:
2930 			if (tp->t_flagsext & TF_NOBLACKHOLE_DETECTION) {
2931 				optval = 1;
2932 			} else {
2933 				optval = 0;
2934 			}
2935 			break;
2936 		case TCP_PEER_PID: {
2937 			pid_t   pid;
2938 			error = tcp_lookup_peer_pid_locked(so, &pid);
2939 			if (error == 0) {
2940 				error = sooptcopyout(sopt, &pid, sizeof(pid));
2941 			}
2942 			goto done;
2943 		}
2944 		case TCP_ADAPTIVE_READ_TIMEOUT:
2945 			optval = tp->t_adaptive_rtimo;
2946 			break;
2947 		case TCP_ADAPTIVE_WRITE_TIMEOUT:
2948 			optval = tp->t_adaptive_wtimo;
2949 			break;
2950 		case SO_TRAFFIC_MGT_BACKGROUND:
2951 			optval = (so->so_flags1 &
2952 			    SOF1_TRAFFIC_MGT_SO_BACKGROUND) ? 1 : 0;
2953 			break;
2954 		case TCP_NOTIFY_ACKNOWLEDGEMENT: {
2955 			struct tcp_notify_ack_complete retid;
2956 
2957 			if (sopt->sopt_valsize != sizeof(retid)) {
2958 				error = EINVAL;
2959 				break;
2960 			}
2961 			bzero(&retid, sizeof(retid));
2962 			tcp_get_notify_ack_count(tp, &retid);
2963 			if (retid.notify_complete_count > 0) {
2964 				tcp_get_notify_ack_ids(tp, &retid);
2965 			}
2966 
2967 			error = sooptcopyout(sopt, &retid, sizeof(retid));
2968 			goto done;
2969 		}
2970 		case TCP_RXT_MINIMUM_TIMEOUT:
2971 			optval = tp->t_rxt_minimum_timeout / TCP_RETRANSHZ;
2972 			break;
2973 		default:
2974 			error = ENOPROTOOPT;
2975 			break;
2976 		}
2977 		if (error == 0) {
2978 			error = sooptcopyout(sopt, &optval, sizeof optval);
2979 		}
2980 		break;
2981 	}
2982 done:
2983 	return error;
2984 }
2985 
2986 /*
2987  * tcp_sendspace and tcp_recvspace are the default send and receive window
2988  * sizes, respectively.  These are obsolescent (this information should
2989  * be set by the route).
2990  */
2991 uint32_t       tcp_sendspace = 1448 * 256;
2992 uint32_t       tcp_recvspace = 1448 * 384;
2993 
2994 /* During attach, the size of socket buffer allocated is limited to
2995  * sb_max in sbreserve. Disallow setting the tcp send and recv space
2996  * to be more than sb_max because that will cause tcp_attach to fail
2997  * (see radar 5713060)
2998  */
2999 static int
sysctl_tcp_sospace(struct sysctl_oid * oidp,__unused void * arg1,int arg2,struct sysctl_req * req)3000 sysctl_tcp_sospace(struct sysctl_oid *oidp, __unused void *arg1,
3001     int arg2, struct sysctl_req *req)
3002 {
3003 #pragma unused(arg2)
3004 	u_int32_t new_value = 0, *space_p = NULL;
3005 	int changed = 0, error = 0;
3006 
3007 	switch (oidp->oid_number) {
3008 	case TCPCTL_SENDSPACE:
3009 		space_p = &tcp_sendspace;
3010 		break;
3011 	case TCPCTL_RECVSPACE:
3012 		space_p = &tcp_recvspace;
3013 		break;
3014 	default:
3015 		return EINVAL;
3016 	}
3017 	error = sysctl_io_number(req, *space_p, sizeof(u_int32_t),
3018 	    &new_value, &changed);
3019 	if (changed) {
3020 		if (new_value > 0 && new_value <= sb_max) {
3021 			*space_p = new_value;
3022 			SYSCTL_SKMEM_UPDATE_AT_OFFSET(arg2, new_value);
3023 		} else {
3024 			error = ERANGE;
3025 		}
3026 	}
3027 	return error;
3028 }
3029 
3030 #if SYSCTL_SKMEM
3031 SYSCTL_PROC(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace,
3032     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_KERN, &tcp_sendspace,
3033     offsetof(skmem_sysctl, tcp.sendspace), sysctl_tcp_sospace,
3034     "IU", "Maximum outgoing TCP datagram size");
3035 SYSCTL_PROC(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace,
3036     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_KERN, &tcp_recvspace,
3037     offsetof(skmem_sysctl, tcp.recvspace), sysctl_tcp_sospace,
3038     "IU", "Maximum incoming TCP datagram size");
3039 #else /* SYSCTL_SKMEM */
3040 SYSCTL_PROC(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace,
3041     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_KERN,
3042     &tcp_sendspace, 0, &sysctl_tcp_sospace, "IU", "Maximum outgoing TCP datagram size");
3043 SYSCTL_PROC(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace,
3044     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_KERN,
3045     &tcp_recvspace, 0, &sysctl_tcp_sospace, "IU", "Maximum incoming TCP datagram size");
3046 #endif /* SYSCTL_SKMEM */
3047 
3048 /*
3049  * Attach TCP protocol to socket, allocating
3050  * internet protocol control block, tcp control block,
3051  * bufer space, and entering LISTEN state if to accept connections.
3052  *
3053  * Returns:	0			Success
3054  *	in_pcballoc:ENOBUFS
3055  *	in_pcballoc:ENOMEM
3056  *	in_pcballoc:???			[IPSEC specific]
3057  *	soreserve:ENOBUFS
3058  */
3059 static int
tcp_attach(struct socket * so,struct proc * p)3060 tcp_attach(struct socket *so, struct proc *p)
3061 {
3062 	struct tcpcb *tp;
3063 	struct inpcb *inp;
3064 	int error;
3065 	int isipv6 = SOCK_CHECK_DOM(so, PF_INET6) != 0;
3066 
3067 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
3068 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
3069 		if (error) {
3070 			return error;
3071 		}
3072 	}
3073 
3074 	error = in_pcballoc(so, &tcbinfo, p);
3075 	if (error) {
3076 		return error;
3077 	}
3078 
3079 	inp = sotoinpcb(so);
3080 
3081 	if (so->so_snd.sb_preconn_hiwat == 0) {
3082 		soreserve_preconnect(so, 2048);
3083 	}
3084 
3085 	if ((so->so_rcv.sb_flags & SB_USRSIZE) == 0) {
3086 		so->so_rcv.sb_flags |= SB_AUTOSIZE;
3087 	}
3088 	if ((so->so_snd.sb_flags & SB_USRSIZE) == 0) {
3089 		so->so_snd.sb_flags |= SB_AUTOSIZE;
3090 	}
3091 
3092 	if (isipv6) {
3093 		inp->inp_vflag |= INP_IPV6;
3094 		inp->in6p_hops = -1;    /* use kernel default */
3095 	} else {
3096 		inp->inp_vflag |= INP_IPV4;
3097 	}
3098 	tp = tcp_newtcpcb(inp);
3099 	if (tp == NULL) {
3100 		short nofd = so->so_state & SS_NOFDREF;   /* XXX */
3101 
3102 		so->so_state &= ~SS_NOFDREF;    /* don't free the socket yet */
3103 		if (isipv6) {
3104 			in6_pcbdetach(inp);
3105 		} else {
3106 			in_pcbdetach(inp);
3107 		}
3108 		so->so_state |= nofd;
3109 		return ENOBUFS;
3110 	}
3111 	if (nstat_collect) {
3112 		nstat_tcp_new_pcb(inp);
3113 	}
3114 	TCP_LOG_STATE(tp, TCPS_CLOSED);
3115 	tp->t_state = TCPS_CLOSED;
3116 	return 0;
3117 }
3118 
3119 /*
3120  * Initiate (or continue) disconnect.
3121  * If embryonic state, just send reset (once).
3122  * If in ``let data drain'' option and linger null, just drop.
3123  * Otherwise (hard), mark socket disconnecting and drop
3124  * current input data; switch states based on user close, and
3125  * send segment to peer (with FIN).
3126  */
3127 static struct tcpcb *
tcp_disconnect(struct tcpcb * tp)3128 tcp_disconnect(struct tcpcb *tp)
3129 {
3130 	struct socket *so = tp->t_inpcb->inp_socket;
3131 
3132 	if (so->so_rcv.sb_cc != 0 || tp->t_reassqlen != 0 ||
3133 	    so->so_flags1 & SOF1_DEFUNCTINPROG) {
3134 		return tcp_drop(tp, 0);
3135 	}
3136 
3137 	if (tp->t_state < TCPS_ESTABLISHED) {
3138 		tp = tcp_close(tp);
3139 	} else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
3140 		tp = tcp_drop(tp, 0);
3141 	} else {
3142 		soisdisconnecting(so);
3143 		sbflush(&so->so_rcv);
3144 		tp = tcp_usrclosed(tp);
3145 #if MPTCP
3146 		/* A reset has been sent but socket exists, do not send FIN */
3147 		if ((so->so_flags & SOF_MP_SUBFLOW) &&
3148 		    (tp) && (tp->t_mpflags & TMPF_RESET)) {
3149 			return tp;
3150 		}
3151 #endif
3152 		if (tp) {
3153 			(void) tcp_output(tp);
3154 		}
3155 	}
3156 	return tp;
3157 }
3158 
3159 /*
3160  * User issued close, and wish to trail through shutdown states:
3161  * if never received SYN, just forget it.  If got a SYN from peer,
3162  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
3163  * If already got a FIN from peer, then almost done; go to LAST_ACK
3164  * state.  In all other cases, have already sent FIN to peer (e.g.
3165  * after PRU_SHUTDOWN), and just have to play tedious game waiting
3166  * for peer to send FIN or not respond to keep-alives, etc.
3167  * We can let the user exit from the close as soon as the FIN is acked.
3168  */
3169 static struct tcpcb *
tcp_usrclosed(struct tcpcb * tp)3170 tcp_usrclosed(struct tcpcb *tp)
3171 {
3172 	switch (tp->t_state) {
3173 	case TCPS_CLOSED:
3174 	case TCPS_LISTEN:
3175 	case TCPS_SYN_SENT:
3176 		tp = tcp_close(tp);
3177 		break;
3178 
3179 	case TCPS_SYN_RECEIVED:
3180 		tp->t_flags |= TF_NEEDFIN;
3181 		break;
3182 
3183 	case TCPS_ESTABLISHED:
3184 		DTRACE_TCP4(state__change, void, NULL,
3185 		    struct inpcb *, tp->t_inpcb,
3186 		    struct tcpcb *, tp,
3187 		    int32_t, TCPS_FIN_WAIT_1);
3188 		TCP_LOG_STATE(tp, TCPS_FIN_WAIT_1);
3189 		tp->t_state = TCPS_FIN_WAIT_1;
3190 		TCP_LOG_CONNECTION_SUMMARY(tp);
3191 		break;
3192 
3193 	case TCPS_CLOSE_WAIT:
3194 		DTRACE_TCP4(state__change, void, NULL,
3195 		    struct inpcb *, tp->t_inpcb,
3196 		    struct tcpcb *, tp,
3197 		    int32_t, TCPS_LAST_ACK);
3198 		TCP_LOG_STATE(tp, TCPS_LAST_ACK);
3199 		tp->t_state = TCPS_LAST_ACK;
3200 		TCP_LOG_CONNECTION_SUMMARY(tp);
3201 		break;
3202 	}
3203 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
3204 		soisdisconnected(tp->t_inpcb->inp_socket);
3205 		/* To prevent the connection hanging in FIN_WAIT_2 forever. */
3206 		if (tp->t_state == TCPS_FIN_WAIT_2) {
3207 			tcp_set_finwait_timeout(tp);
3208 		}
3209 	}
3210 	return tp;
3211 }
3212 
3213 void
tcp_in_cksum_stats(u_int32_t len)3214 tcp_in_cksum_stats(u_int32_t len)
3215 {
3216 	tcpstat.tcps_rcv_swcsum++;
3217 	tcpstat.tcps_rcv_swcsum_bytes += len;
3218 }
3219 
3220 void
tcp_out_cksum_stats(u_int32_t len)3221 tcp_out_cksum_stats(u_int32_t len)
3222 {
3223 	tcpstat.tcps_snd_swcsum++;
3224 	tcpstat.tcps_snd_swcsum_bytes += len;
3225 }
3226 
3227 void
tcp_in6_cksum_stats(u_int32_t len)3228 tcp_in6_cksum_stats(u_int32_t len)
3229 {
3230 	tcpstat.tcps_rcv6_swcsum++;
3231 	tcpstat.tcps_rcv6_swcsum_bytes += len;
3232 }
3233 
3234 void
tcp_out6_cksum_stats(u_int32_t len)3235 tcp_out6_cksum_stats(u_int32_t len)
3236 {
3237 	tcpstat.tcps_snd6_swcsum++;
3238 	tcpstat.tcps_snd6_swcsum_bytes += len;
3239 }
3240 
3241 int
tcp_get_mpkl_send_info(struct mbuf * control,struct so_mpkl_send_info * mpkl_send_info)3242 tcp_get_mpkl_send_info(struct mbuf *control,
3243     struct so_mpkl_send_info *mpkl_send_info)
3244 {
3245 	struct cmsghdr *cm;
3246 
3247 	if (control == NULL || mpkl_send_info == NULL) {
3248 		return EINVAL;
3249 	}
3250 
3251 	for (cm = M_FIRST_CMSGHDR(control); cm;
3252 	    cm = M_NXT_CMSGHDR(control, cm)) {
3253 		if (cm->cmsg_len < sizeof(struct cmsghdr) ||
3254 		    cm->cmsg_len > control->m_len) {
3255 			return EINVAL;
3256 		}
3257 		if (cm->cmsg_level != SOL_SOCKET ||
3258 		    cm->cmsg_type != SCM_MPKL_SEND_INFO) {
3259 			continue;
3260 		}
3261 		if (cm->cmsg_len != CMSG_LEN(sizeof(struct so_mpkl_send_info))) {
3262 			return EINVAL;
3263 		}
3264 		memcpy(mpkl_send_info, CMSG_DATA(cm),
3265 		    sizeof(struct so_mpkl_send_info));
3266 		return 0;
3267 	}
3268 	return ENOMSG;
3269 }
3270 
3271 /*
3272  * tcp socket options.
3273  *
3274  * The switch statement below does nothing at runtime, as it serves as a
3275  * compile time check to ensure that all of the tcp socket options are
3276  * unique.  This works as long as this routine gets updated each time a
3277  * new tcp socket option gets added.
3278  *
3279  * Any failures at compile time indicates duplicated tcp socket option
3280  * values.
3281  */
3282 static __attribute__((unused)) void
tcpsockopt_cassert(void)3283 tcpsockopt_cassert(void)
3284 {
3285 	/*
3286 	 * This is equivalent to _CASSERT() and the compiler wouldn't
3287 	 * generate any instructions, thus for compile time only.
3288 	 */
3289 	switch ((int)0) {
3290 	case 0:
3291 
3292 	/* bsd/netinet/tcp.h */
3293 	case TCP_NODELAY:
3294 	case TCP_MAXSEG:
3295 	case TCP_NOPUSH:
3296 	case TCP_NOOPT:
3297 	case TCP_KEEPALIVE:
3298 	case TCP_CONNECTIONTIMEOUT:
3299 	case PERSIST_TIMEOUT:
3300 	case TCP_RXT_CONNDROPTIME:
3301 	case TCP_RXT_FINDROP:
3302 	case TCP_KEEPINTVL:
3303 	case TCP_KEEPCNT:
3304 	case TCP_SENDMOREACKS:
3305 	case TCP_ENABLE_ECN:
3306 	case TCP_FASTOPEN:
3307 	case TCP_CONNECTION_INFO:
3308 	case TCP_NOTSENT_LOWAT:
3309 
3310 	/* bsd/netinet/tcp_private.h */
3311 	case TCP_INFO:
3312 	case TCP_MEASURE_SND_BW:
3313 	case TCP_MEASURE_BW_BURST:
3314 	case TCP_PEER_PID:
3315 	case TCP_ADAPTIVE_READ_TIMEOUT:
3316 	case TCP_OPTION_UNUSED_0:
3317 	case TCP_ADAPTIVE_WRITE_TIMEOUT:
3318 	case TCP_NOTIMEWAIT:
3319 	case TCP_DISABLE_BLACKHOLE_DETECTION:
3320 	case TCP_ECN_MODE:
3321 	case TCP_KEEPALIVE_OFFLOAD:
3322 		;
3323 	}
3324 }
3325