xref: /xnu-8792.61.2/bsd/netinet/tcp_usrreq.c (revision 42e220869062b56f8d7d0726fd4c88954f87902c)
1 /*
2  * Copyright (c) 2000-2022 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 #if XNU_TARGET_OS_OSX
72 #include <sys/kasl.h>
73 #endif /* XNU_TARGET_OS_OSX */
74 #include <sys/priv.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/protosw.h>
78 #include <sys/syslog.h>
79 
80 #include <net/if.h>
81 #include <net/route.h>
82 #include <net/ntstat.h>
83 #include <net/content_filter.h>
84 #include <net/multi_layer_pkt_log.h>
85 
86 #include <netinet/in.h>
87 #include <netinet/in_systm.h>
88 #include <netinet/ip6.h>
89 #include <netinet/in_pcb.h>
90 #include <netinet6/in6_pcb.h>
91 #include <netinet/in_var.h>
92 #include <netinet/ip_var.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet/tcp.h>
95 #include <netinet/tcp_fsm.h>
96 #include <netinet/tcp_seq.h>
97 #include <netinet/tcp_timer.h>
98 #include <netinet/tcp_var.h>
99 #include <netinet/tcpip.h>
100 #include <netinet/tcp_cc.h>
101 #include <netinet/tcp_log.h>
102 #include <mach/sdt.h>
103 #if TCPDEBUG
104 #include <netinet/tcp_debug.h>
105 #endif
106 #if MPTCP
107 #include <netinet/mptcp_var.h>
108 #endif /* MPTCP */
109 
110 #if IPSEC
111 #include <netinet6/ipsec.h>
112 #endif /*IPSEC*/
113 
114 #if FLOW_DIVERT
115 #include <netinet/flow_divert.h>
116 #endif /* FLOW_DIVERT */
117 
118 #if SKYWALK
119 #include <libkern/sysctl.h>
120 #include <skywalk/os_stats_private.h>
121 #endif /* SKYWALK */
122 
123 extern char *proc_name_address(void *p);
124 
125 errno_t tcp_fill_info_for_info_tuple(struct info_tuple *, struct tcp_info *);
126 
127 int tcp_sysctl_info(struct sysctl_oid *, void *, int, struct sysctl_req *);
128 static void tcp_connection_fill_info(struct tcpcb *tp,
129     struct tcp_connection_info *tci);
130 static int tcp_get_mpkl_send_info(struct mbuf *, struct so_mpkl_send_info *);
131 
132 /*
133  * TCP protocol interface to socket abstraction.
134  */
135 static int      tcp_attach(struct socket *, struct proc *);
136 static int      tcp_connect(struct tcpcb *, struct sockaddr *, struct proc *);
137 static int      tcp6_connect(struct tcpcb *, struct sockaddr *, struct proc *);
138 static int      tcp6_usr_connect(struct socket *, struct sockaddr *,
139     struct proc *);
140 static struct tcpcb *tcp_disconnect(struct tcpcb *);
141 static struct tcpcb *tcp_usrclosed(struct tcpcb *);
142 extern void tcp_sbrcv_trim(struct tcpcb *tp, struct sockbuf *sb);
143 
144 #if TCPDEBUG
145 #define TCPDEBUG0       int ostate = 0
146 #define TCPDEBUG1()     ostate = tp ? tp->t_state : 0
147 #define TCPDEBUG2(req)  if (tp && (so->so_options & SO_DEBUG)) \
148 	                        tcp_trace(TA_USER, ostate, tp, 0, 0, req)
149 #else
150 #define TCPDEBUG0
151 #define TCPDEBUG1()
152 #define TCPDEBUG2(req)
153 #endif
154 
155 SYSCTL_PROC(_net_inet_tcp, OID_AUTO, info,
156     CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_ANYBODY | CTLFLAG_KERN,
157     0, 0, tcp_sysctl_info, "S", "TCP info per tuple");
158 
159 /*
160  * TCP attaches to socket via pru_attach(), reserving space,
161  * and an internet control block.
162  *
163  * Returns:	0			Success
164  *		EISCONN
165  *	tcp_attach:ENOBUFS
166  *	tcp_attach:ENOMEM
167  *	tcp_attach:???			[IPSEC specific]
168  */
169 static int
tcp_usr_attach(struct socket * so,__unused int proto,struct proc * p)170 tcp_usr_attach(struct socket *so, __unused int proto, struct proc *p)
171 {
172 	int error;
173 	struct inpcb *inp = sotoinpcb(so);
174 	struct tcpcb *tp = 0;
175 	TCPDEBUG0;
176 
177 	TCPDEBUG1();
178 	if (inp) {
179 		error = EISCONN;
180 		goto out;
181 	}
182 
183 	error = tcp_attach(so, p);
184 	if (error) {
185 		goto out;
186 	}
187 
188 	if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
189 		so->so_linger = (short)(TCP_LINGERTIME * hz);
190 	}
191 	tp = sototcpcb(so);
192 out:
193 	TCPDEBUG2(PRU_ATTACH);
194 	return error;
195 }
196 
197 /*
198  * pru_detach() detaches the TCP protocol from the socket.
199  * If the protocol state is non-embryonic, then can't
200  * do this directly: have to initiate a pru_disconnect(),
201  * which may finish later; embryonic TCB's can just
202  * be discarded here.
203  */
204 static int
tcp_usr_detach(struct socket * so)205 tcp_usr_detach(struct socket *so)
206 {
207 	int error = 0;
208 	struct inpcb *inp = sotoinpcb(so);
209 	struct tcpcb *tp;
210 	TCPDEBUG0;
211 
212 	if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) {
213 		return EINVAL;  /* XXX */
214 	}
215 	socket_lock_assert_owned(so);
216 	tp = intotcpcb(inp);
217 	/* In case we got disconnected from the peer */
218 	if (tp == NULL) {
219 		goto out;
220 	}
221 	TCPDEBUG1();
222 
223 	calculate_tcp_clock();
224 
225 	tp = tcp_disconnect(tp);
226 out:
227 	TCPDEBUG2(PRU_DETACH);
228 	return error;
229 }
230 
231 #if NECP
232 #define COMMON_START_ALLOW_FLOW_DIVERT(allow)  TCPDEBUG0;               \
233 do {                                                                    \
234 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD)          \
235 	        return (EINVAL);                                        \
236 	if (!(allow) && necp_socket_should_use_flow_divert(inp))        \
237 	        return (EPROTOTYPE);                                    \
238 	tp = intotcpcb(inp);                                            \
239 	TCPDEBUG1();                                                    \
240 	calculate_tcp_clock();                                          \
241 } while (0)
242 #else /* NECP */
243 #define COMMON_START_ALLOW_FLOW_DIVERT(allow)  TCPDEBUG0;               \
244 do {                                                                    \
245 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD)          \
246 	        return (EINVAL);                                        \
247 	tp = intotcpcb(inp);                                            \
248 	TCPDEBUG1();                                                    \
249 	calculate_tcp_clock();                                          \
250 } while (0)
251 #endif /* !NECP */
252 
253 #define COMMON_START() COMMON_START_ALLOW_FLOW_DIVERT(false)
254 #define COMMON_END(req) out: TCPDEBUG2(req); return error; goto out
255 
256 
257 /*
258  * Give the socket an address.
259  *
260  * Returns:	0			Success
261  *		EINVAL			Invalid argument [COMMON_START]
262  *		EAFNOSUPPORT		Address family not supported
263  *	in_pcbbind:EADDRNOTAVAIL	Address not available.
264  *	in_pcbbind:EINVAL		Invalid argument
265  *	in_pcbbind:EAFNOSUPPORT		Address family not supported [notdef]
266  *	in_pcbbind:EACCES		Permission denied
267  *	in_pcbbind:EADDRINUSE		Address in use
268  *	in_pcbbind:EAGAIN		Resource unavailable, try again
269  *	in_pcbbind:EPERM		Operation not permitted
270  */
271 static int
tcp_usr_bind(struct socket * so,struct sockaddr * nam,struct proc * p)272 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
273 {
274 	int error = 0;
275 	struct inpcb *inp = sotoinpcb(so);
276 	struct tcpcb *tp;
277 	struct sockaddr_in *sinp;
278 
279 	COMMON_START_ALLOW_FLOW_DIVERT(true);
280 
281 	if (nam->sa_family != 0 && nam->sa_family != AF_INET) {
282 		error = EAFNOSUPPORT;
283 		goto out;
284 	}
285 
286 	/*
287 	 * Must check for multicast and broadcast addresses and disallow binding
288 	 * to them.
289 	 */
290 	sinp = (struct sockaddr_in *)(void *)nam;
291 	if (sinp->sin_family == AF_INET &&
292 	    (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)) ||
293 	    sinp->sin_addr.s_addr == INADDR_BROADCAST)) {
294 		error = EAFNOSUPPORT;
295 		goto out;
296 	}
297 	error = in_pcbbind(inp, nam, p);
298 	if (error) {
299 		goto out;
300 	}
301 
302 #if NECP
303 	/* Update NECP client with bind result if not in middle of connect */
304 	if ((inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS) &&
305 	    !uuid_is_null(inp->necp_client_uuid)) {
306 		socket_unlock(so, 0);
307 		necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
308 		socket_lock(so, 0);
309 	}
310 #endif /* NECP */
311 
312 	COMMON_END(PRU_BIND);
313 }
314 
315 static int
tcp6_usr_bind(struct socket * so,struct sockaddr * nam,struct proc * p)316 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
317 {
318 	int error = 0;
319 	struct inpcb *inp = sotoinpcb(so);
320 	const uint8_t old_flags = inp->inp_vflag;
321 	struct tcpcb *tp;
322 	struct sockaddr_in6 *sin6p;
323 
324 	COMMON_START_ALLOW_FLOW_DIVERT(true);
325 
326 	if (nam->sa_family != 0 && nam->sa_family != AF_INET6) {
327 		error = EAFNOSUPPORT;
328 		goto out;
329 	}
330 
331 	/*
332 	 * Must check for multicast and broadcast addresses and disallow binding
333 	 * to them.
334 	 */
335 	sin6p = (struct sockaddr_in6 *)(void *)nam;
336 	if (sin6p->sin6_family == AF_INET6 &&
337 	    (IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr) ||
338 	    ((IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr) ||
339 	    IN6_IS_ADDR_V4COMPAT(&sin6p->sin6_addr)) &&
340 	    (IN_MULTICAST(ntohl(sin6p->sin6_addr.s6_addr32[3])) ||
341 	    sin6p->sin6_addr.s6_addr32[3] == INADDR_BROADCAST)))) {
342 		error = EAFNOSUPPORT;
343 		goto out;
344 	}
345 	inp->inp_vflag &= ~INP_IPV4;
346 	inp->inp_vflag |= INP_IPV6;
347 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
348 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) {
349 			inp->inp_vflag |= INP_IPV4;
350 		} else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
351 			struct sockaddr_in sin;
352 
353 			in6_sin6_2_sin(&sin, sin6p);
354 			inp->inp_vflag |= INP_IPV4;
355 			inp->inp_vflag &= ~INP_IPV6;
356 
357 			error = in_pcbbind(inp, (struct sockaddr *)&sin, p);
358 			if (error != 0) {
359 				inp->inp_vflag = old_flags;
360 				route_clear(&inp->inp_route);
361 			}
362 			goto out;
363 		}
364 	}
365 	error = in6_pcbbind(inp, nam, p);
366 	if (error) {
367 		inp->inp_vflag = old_flags;
368 		route_clear(&inp->inp_route);
369 		goto out;
370 	}
371 	COMMON_END(PRU_BIND);
372 }
373 
374 /*
375  * Prepare to accept connections.
376  *
377  * Returns:	0			Success
378  *		EINVAL [COMMON_START]
379  *	in_pcbbind:EADDRNOTAVAIL	Address not available.
380  *	in_pcbbind:EINVAL		Invalid argument
381  *	in_pcbbind:EAFNOSUPPORT		Address family not supported [notdef]
382  *	in_pcbbind:EACCES		Permission denied
383  *	in_pcbbind:EADDRINUSE		Address in use
384  *	in_pcbbind:EAGAIN		Resource unavailable, try again
385  *	in_pcbbind:EPERM		Operation not permitted
386  */
387 static int
tcp_usr_listen(struct socket * so,struct proc * p)388 tcp_usr_listen(struct socket *so, struct proc *p)
389 {
390 	int error = 0;
391 	struct inpcb *inp = sotoinpcb(so);
392 	struct tcpcb *tp;
393 
394 	COMMON_START_ALLOW_FLOW_DIVERT(true);
395 	if (inp->inp_lport == 0) {
396 		error = in_pcbbind(inp, NULL, p);
397 	}
398 	if (error == 0) {
399 		TCP_LOG_STATE(tp, TCPS_LISTEN);
400 		tp->t_state = TCPS_LISTEN;
401 	}
402 	TCP_LOG_LISTEN(tp, error);
403 	COMMON_END(PRU_LISTEN);
404 }
405 
406 static int
tcp6_usr_listen(struct socket * so,struct proc * p)407 tcp6_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 		inp->inp_vflag &= ~INP_IPV4;
416 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
417 			inp->inp_vflag |= INP_IPV4;
418 		}
419 		error = in6_pcbbind(inp, NULL, p);
420 	}
421 	if (error == 0) {
422 		TCP_LOG_STATE(tp, TCPS_LISTEN);
423 		tp->t_state = TCPS_LISTEN;
424 	}
425 	TCP_LOG_LISTEN(tp, error);
426 	COMMON_END(PRU_LISTEN);
427 }
428 
429 static int
tcp_connect_complete(struct socket * so)430 tcp_connect_complete(struct socket *so)
431 {
432 	struct tcpcb *tp = sototcpcb(so);
433 	struct inpcb *inp = sotoinpcb(so);
434 	int error = 0;
435 
436 	/* TFO delays the tcp_output until later, when the app calls write() */
437 	if (so->so_flags1 & SOF1_PRECONNECT_DATA) {
438 		if (!necp_socket_is_allowed_to_send_recv(sotoinpcb(so), NULL, 0, NULL, NULL, NULL, NULL)) {
439 			TCP_LOG_DROP_NECP(NULL, NULL, tp, true);
440 			return EHOSTUNREACH;
441 		}
442 
443 		/* Initialize enough state so that we can actually send data */
444 		tcp_mss(tp, -1, IFSCOPE_NONE);
445 		tp->snd_wnd = tp->t_maxseg;
446 		tp->max_sndwnd = tp->snd_wnd;
447 	} else {
448 		tp->t_flagsext |= TF_USR_OUTPUT;
449 		error = tcp_output(tp);
450 		tp->t_flagsext &= ~TF_USR_OUTPUT;
451 	}
452 
453 #if NECP
454 	/* Update NECP client with connected five-tuple */
455 	if (error == 0 && !uuid_is_null(inp->necp_client_uuid)) {
456 		socket_unlock(so, 0);
457 		necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
458 		socket_lock(so, 0);
459 	}
460 #endif /* NECP */
461 
462 	return error;
463 }
464 
465 __attribute__((noinline))
466 static void
tcp_log_address_error(int error,struct sockaddr * nam,struct proc * p)467 tcp_log_address_error(int error, struct sockaddr *nam, struct proc *p)
468 {
469 	char buffer[MAX_IPv6_STR_LEN];
470 
471 	if (nam->sa_family == AF_INET6) {
472 		struct sockaddr_in6 *sin6p = (struct sockaddr_in6 *)(void *)nam;
473 
474 		inet_ntop(AF_INET6, &sin6p->sin6_addr, buffer, sizeof(buffer));
475 	} else {
476 		struct sockaddr_in *sinp = (struct sockaddr_in *)(void *)nam;
477 
478 		inet_ntop(AF_INET, &sinp->sin_addr, buffer, sizeof(buffer));
479 	}
480 	if (p == NULL) {
481 		p = current_proc();
482 	}
483 	os_log(OS_LOG_DEFAULT, "connect address error %d for %s process %s:%u",
484 	    error, buffer, proc_name_address(p), proc_pid(p));
485 }
486 
487 /*
488  * Note that connecting to the all-zeros address is OK and is treated as the
489  * loopback address
490  */
491 static int
tcp_usr_connect_common(struct socket * so,struct tcpcb * tp,struct sockaddr * nam,struct proc * p,bool isipv6,bool need_connect_complete)492 tcp_usr_connect_common(struct socket *so, struct tcpcb *tp, struct sockaddr *nam,
493     struct proc *p, bool isipv6, bool need_connect_complete)
494 {
495 	int error = 0;
496 	struct inpcb *inp = sotoinpcb(so);
497 
498 	if (isipv6 == 0) {
499 		struct sockaddr_in *sinp;
500 
501 		if (nam->sa_family != 0 && nam->sa_family != AF_INET) {
502 			error = EAFNOSUPPORT;
503 			goto out;
504 		}
505 		/*
506 		 * Disallow connecting to multicast and broadcast addresses.
507 		 */
508 		sinp = (struct sockaddr_in *)(void *)nam;
509 		if (sinp->sin_family == AF_INET &&
510 		    (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)) ||
511 		    sinp->sin_addr.s_addr == INADDR_BROADCAST)) {
512 			error = EAFNOSUPPORT;
513 			goto out;
514 		}
515 
516 		if ((error = tcp_connect(tp, nam, p)) != 0) {
517 			goto out;
518 		}
519 	} else {
520 		struct sockaddr_in6 *sin6p;
521 
522 		if (nam->sa_family != 0 && nam->sa_family != AF_INET6) {
523 			error = EAFNOSUPPORT;
524 			goto out;
525 		}
526 
527 		/*
528 		 * Disallow connecting to multicast and broadcast addresses.
529 		 */
530 		sin6p = (struct sockaddr_in6 *)(void *)nam;
531 		if (sin6p->sin6_family == AF_INET6 &&
532 		    IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
533 			error = EAFNOSUPPORT;
534 			goto out;
535 		}
536 
537 		if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
538 			struct sockaddr_in sin;
539 
540 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
541 				error = EINVAL;
542 				goto out;
543 			}
544 
545 			in6_sin6_2_sin(&sin, sin6p);
546 			/*
547 			 * Disallow connecting to multicast and broadcast addresses.
548 			 */
549 			if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr)) ||
550 			    sin.sin_addr.s_addr == INADDR_BROADCAST) {
551 				error = EAFNOSUPPORT;
552 				goto out;
553 			}
554 			inp->inp_vflag |= INP_IPV4;
555 			inp->inp_vflag &= ~INP_IPV6;
556 			if ((error = tcp_connect(tp, (struct sockaddr *)&sin, p)) != 0) {
557 				goto out;
558 			}
559 
560 			goto out;
561 		} else if (IN6_IS_ADDR_V4COMPAT(&sin6p->sin6_addr)) {
562 			/*
563 			 * Disallow connecting to multicast and broadcast addresses.
564 			 */
565 			if (IN_MULTICAST(ntohl(sin6p->sin6_addr.s6_addr32[3])) ||
566 			    sin6p->sin6_addr.s6_addr32[3] == INADDR_BROADCAST) {
567 				error = EAFNOSUPPORT;
568 				goto out;
569 			}
570 		}
571 
572 		inp->inp_vflag &= ~INP_IPV4;
573 		inp->inp_vflag |= INP_IPV6;
574 		if ((error = tcp6_connect(tp, nam, p)) != 0) {
575 			goto out;
576 		}
577 	}
578 out:
579 	if (need_connect_complete && error == 0) {
580 		error = tcp_connect_complete(so);
581 	}
582 	TCP_LOG_CONNECT(tp, true, error);
583 	if (error == EAFNOSUPPORT) {
584 		tcp_log_address_error(error, nam, p);
585 	}
586 	return error;
587 }
588 
589 /*
590  * Initiate connection to peer.
591  * Create a template for use in transmissions on this connection.
592  * Enter SYN_SENT state, and mark socket as connecting.
593  * Start keep-alive timer, and seed output sequence space.
594  * Send initial segment on connection.
595  */
596 static int
tcp_usr_connect(struct socket * so,struct sockaddr * nam,struct proc * p)597 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
598 {
599 	int error = 0;
600 	struct inpcb *inp = sotoinpcb(so);
601 	struct tcpcb *tp;
602 
603 	TCPDEBUG0;
604 	if (inp == NULL) {
605 		return EINVAL;
606 	} else if (inp->inp_state == INPCB_STATE_DEAD) {
607 		if (so->so_error) {
608 			error = so->so_error;
609 			so->so_error = 0;
610 			return error;
611 		} else {
612 			return EINVAL;
613 		}
614 	}
615 #if NECP
616 #if CONTENT_FILTER
617 	error = cfil_sock_attach(so, NULL, nam, CFS_CONNECTION_DIR_OUT);
618 	if (error != 0) {
619 		return error;
620 	}
621 #endif /* CONTENT_FILTER */
622 #if FLOW_DIVERT
623 	if (necp_socket_should_use_flow_divert(inp)) {
624 		error = flow_divert_pcb_init(so);
625 		if (error == 0) {
626 			error = flow_divert_connect_out(so, nam, p);
627 		}
628 		return error;
629 	}
630 #endif /* FLOW_DIVERT */
631 #endif /* NECP */
632 	tp = intotcpcb(inp);
633 	TCPDEBUG1();
634 
635 	calculate_tcp_clock();
636 
637 	error = tcp_usr_connect_common(so, tp, nam, p, false, true);
638 	if (error != 0) {
639 		goto out;
640 	}
641 
642 	COMMON_END(PRU_CONNECT);
643 }
644 
645 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)646 tcp_usr_connectx_common(struct socket *so, int af,
647     struct sockaddr *src, struct sockaddr *dst,
648     struct proc *p, uint32_t ifscope, sae_associd_t aid, sae_connid_t *pcid,
649     uint32_t flags, void *arg, uint32_t arglen, struct uio *auio,
650     user_ssize_t *bytes_written)
651 {
652 #pragma unused(aid, flags, arg, arglen)
653 	struct inpcb *inp = sotoinpcb(so);
654 	int error = 0;
655 	user_ssize_t datalen = 0;
656 
657 	if (inp == NULL) {
658 		return EINVAL;
659 	}
660 
661 	VERIFY(dst != NULL);
662 
663 	ASSERT(!(inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS));
664 	inp->inp_flags2 |= INP2_CONNECT_IN_PROGRESS;
665 
666 #if NECP
667 	inp_update_necp_policy(inp, src, dst, ifscope);
668 #endif /* NECP */
669 
670 	if ((so->so_flags1 & SOF1_DATA_IDEMPOTENT) &&
671 	    (tcp_fastopen & TCP_FASTOPEN_CLIENT)) {
672 		sototcpcb(so)->t_flagsext |= TF_FASTOPEN;
673 	}
674 
675 	/* bind socket to the specified interface, if requested */
676 	if (ifscope != IFSCOPE_NONE &&
677 	    (error = inp_bindif(inp, ifscope, NULL)) != 0) {
678 		goto done;
679 	}
680 
681 	/* if source address and/or port is specified, bind to it */
682 	if (src != NULL) {
683 		error = sobindlock(so, src, 0); /* already locked */
684 		if (error != 0) {
685 			goto done;
686 		}
687 	}
688 
689 	switch (af) {
690 	case AF_INET:
691 		error = tcp_usr_connect(so, dst, p);
692 		break;
693 	case AF_INET6:
694 		error = tcp6_usr_connect(so, dst, p);
695 		break;
696 	default:
697 		VERIFY(0);
698 		/* NOTREACHED */
699 	}
700 
701 	if (error != 0) {
702 		goto done;
703 	}
704 
705 	/* if there is data, copy it */
706 	if (auio != NULL) {
707 		socket_unlock(so, 0);
708 
709 		VERIFY(bytes_written != NULL);
710 
711 		datalen = uio_resid(auio);
712 		error = so->so_proto->pr_usrreqs->pru_sosend(so, NULL,
713 		    (uio_t)auio, NULL, NULL, 0);
714 		socket_lock(so, 0);
715 
716 		if (error == 0 || error == EWOULDBLOCK) {
717 			*bytes_written = datalen - uio_resid(auio);
718 		}
719 
720 		/*
721 		 * sosend returns EWOULDBLOCK if it's a non-blocking
722 		 * socket or a timeout occured (this allows to return
723 		 * the amount of queued data through sendit()).
724 		 *
725 		 * However, connectx() returns EINPROGRESS in case of a
726 		 * blocking socket. So we change the return value here.
727 		 */
728 		if (error == EWOULDBLOCK) {
729 			error = EINPROGRESS;
730 		}
731 	}
732 
733 	if (error == 0 && pcid != NULL) {
734 		*pcid = 1; /* there is only one connection in regular TCP */
735 	}
736 done:
737 	if (error && error != EINPROGRESS) {
738 		so->so_flags1 &= ~SOF1_PRECONNECT_DATA;
739 	}
740 
741 	inp->inp_flags2 &= ~INP2_CONNECT_IN_PROGRESS;
742 	return error;
743 }
744 
745 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)746 tcp_usr_connectx(struct socket *so, struct sockaddr *src,
747     struct sockaddr *dst, struct proc *p, uint32_t ifscope,
748     sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
749     uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written)
750 {
751 	return tcp_usr_connectx_common(so, AF_INET, src, dst, p, ifscope, aid,
752 	           pcid, flags, arg, arglen, uio, bytes_written);
753 }
754 
755 static int
tcp6_usr_connect(struct socket * so,struct sockaddr * nam,struct proc * p)756 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
757 {
758 	int error = 0;
759 	struct inpcb *inp = sotoinpcb(so);
760 	struct tcpcb *tp;
761 
762 	TCPDEBUG0;
763 	if (inp == NULL) {
764 		return EINVAL;
765 	} else if (inp->inp_state == INPCB_STATE_DEAD) {
766 		if (so->so_error) {
767 			error = so->so_error;
768 			so->so_error = 0;
769 			return error;
770 		} else {
771 			return EINVAL;
772 		}
773 	}
774 #if NECP
775 #if CONTENT_FILTER
776 	error = cfil_sock_attach(so, NULL, nam, CFS_CONNECTION_DIR_OUT);
777 	if (error != 0) {
778 		return error;
779 	}
780 #endif /* CONTENT_FILTER */
781 #if FLOW_DIVERT
782 	if (necp_socket_should_use_flow_divert(inp)) {
783 		error = flow_divert_pcb_init(so);
784 		if (error == 0) {
785 			error = flow_divert_connect_out(so, nam, p);
786 		}
787 		return error;
788 	}
789 #endif /* FLOW_DIVERT */
790 #endif /* NECP */
791 
792 	tp = intotcpcb(inp);
793 	TCPDEBUG1();
794 
795 	calculate_tcp_clock();
796 
797 	error = tcp_usr_connect_common(so, tp, nam, p, true, true);
798 	if (error != 0) {
799 		route_clear(&inp->inp_route);
800 		goto out;
801 	}
802 
803 	COMMON_END(PRU_CONNECT);
804 }
805 
806 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)807 tcp6_usr_connectx(struct socket *so, struct sockaddr*src,
808     struct sockaddr *dst, struct proc *p, uint32_t ifscope,
809     sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
810     uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written)
811 {
812 	return tcp_usr_connectx_common(so, AF_INET6, src, dst, p, ifscope, aid,
813 	           pcid, flags, arg, arglen, uio, bytes_written);
814 }
815 
816 /*
817  * Initiate disconnect from peer.
818  * If connection never passed embryonic stage, just drop;
819  * else if don't need to let data drain, then can just drop anyways,
820  * else have to begin TCP shutdown process: mark socket disconnecting,
821  * drain unread data, state switch to reflect user close, and
822  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
823  * when peer sends FIN and acks ours.
824  *
825  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
826  */
827 static int
tcp_usr_disconnect(struct socket * so)828 tcp_usr_disconnect(struct socket *so)
829 {
830 	int error = 0;
831 	struct inpcb *inp = sotoinpcb(so);
832 	struct tcpcb *tp;
833 
834 	socket_lock_assert_owned(so);
835 	COMMON_START();
836 	/* In case we got disconnected from the peer */
837 	if (tp == NULL) {
838 		goto out;
839 	}
840 	tp = tcp_disconnect(tp);
841 	COMMON_END(PRU_DISCONNECT);
842 }
843 
844 /*
845  * User-protocol pru_disconnectx callback.
846  */
847 static int
tcp_usr_disconnectx(struct socket * so,sae_associd_t aid,sae_connid_t cid)848 tcp_usr_disconnectx(struct socket *so, sae_associd_t aid, sae_connid_t cid)
849 {
850 #pragma unused(cid)
851 	if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL) {
852 		return EINVAL;
853 	}
854 
855 	return tcp_usr_disconnect(so);
856 }
857 
858 /*
859  * Accept a connection.  Essentially all the work is
860  * done at higher levels; just return the address
861  * of the peer, storing through addr.
862  */
863 static int
tcp_usr_accept(struct socket * so,struct sockaddr ** nam)864 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
865 {
866 	int error = 0;
867 	struct inpcb *inp = sotoinpcb(so);
868 	struct tcpcb *tp = NULL;
869 	TCPDEBUG0;
870 
871 	in_getpeeraddr(so, nam);
872 
873 	if (so->so_state & SS_ISDISCONNECTED) {
874 		error = ECONNABORTED;
875 		goto out;
876 	}
877 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
878 		return EINVAL;
879 	}
880 #if NECP
881 	else if (necp_socket_should_use_flow_divert(inp)) {
882 		return EPROTOTYPE;
883 	}
884 
885 #endif /* NECP */
886 
887 	tp = intotcpcb(inp);
888 	TCPDEBUG1();
889 
890 	TCP_LOG_ACCEPT(tp, 0);
891 
892 	calculate_tcp_clock();
893 
894 	COMMON_END(PRU_ACCEPT);
895 }
896 
897 static int
tcp6_usr_accept(struct socket * so,struct sockaddr ** nam)898 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
899 {
900 	int error = 0;
901 	struct inpcb *inp = sotoinpcb(so);
902 	struct tcpcb *tp = NULL;
903 	TCPDEBUG0;
904 
905 	if (so->so_state & SS_ISDISCONNECTED) {
906 		error = ECONNABORTED;
907 		goto out;
908 	}
909 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
910 		return EINVAL;
911 	}
912 #if NECP
913 	else if (necp_socket_should_use_flow_divert(inp)) {
914 		return EPROTOTYPE;
915 	}
916 
917 #endif /* NECP */
918 
919 	tp = intotcpcb(inp);
920 	TCPDEBUG1();
921 
922 	TCP_LOG_ACCEPT(tp, 0);
923 
924 	calculate_tcp_clock();
925 
926 	in6_mapped_peeraddr(so, nam);
927 	COMMON_END(PRU_ACCEPT);
928 }
929 
930 /*
931  * Mark the connection as being incapable of further output.
932  *
933  * Returns:	0			Success
934  *		EINVAL [COMMON_START]
935  *	tcp_output:EADDRNOTAVAIL
936  *	tcp_output:ENOBUFS
937  *	tcp_output:EMSGSIZE
938  *	tcp_output:EHOSTUNREACH
939  *	tcp_output:ENETUNREACH
940  *	tcp_output:ENETDOWN
941  *	tcp_output:ENOMEM
942  *	tcp_output:EACCES
943  *	tcp_output:EMSGSIZE
944  *	tcp_output:ENOBUFS
945  *	tcp_output:???			[ignorable: mostly IPSEC/firewall/DLIL]
946  */
947 static int
tcp_usr_shutdown(struct socket * so)948 tcp_usr_shutdown(struct socket *so)
949 {
950 	int error = 0;
951 	struct inpcb *inp = sotoinpcb(so);
952 	struct tcpcb *tp;
953 
954 	TCPDEBUG0;
955 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
956 		return EINVAL;
957 	}
958 
959 	socantsendmore(so);
960 
961 	/*
962 	 * In case we got disconnected from the peer, or if this is
963 	 * a socket that is to be flow-diverted (but not yet).
964 	 */
965 	tp = intotcpcb(inp);
966 	TCPDEBUG1();
967 
968 	if (tp == NULL
969 #if NECP
970 	    || (necp_socket_should_use_flow_divert(inp))
971 #endif /* NECP */
972 	    ) {
973 		if (tp != NULL) {
974 			error = EPROTOTYPE;
975 		}
976 		goto out;
977 	}
978 
979 	calculate_tcp_clock();
980 
981 	tp = tcp_usrclosed(tp);
982 #if MPTCP
983 	/* A reset has been sent but socket exists, do not send FIN */
984 	if ((so->so_flags & SOF_MP_SUBFLOW) &&
985 	    (tp) && (tp->t_mpflags & TMPF_RESET)) {
986 		goto out;
987 	}
988 #endif
989 #if CONTENT_FILTER
990 	/* Don't send a FIN yet */
991 	if (tp && !(so->so_state & SS_ISDISCONNECTED) &&
992 	    cfil_sock_data_pending(&so->so_snd)) {
993 		goto out;
994 	}
995 #endif /* CONTENT_FILTER */
996 	if (tp) {
997 		error = tcp_output(tp);
998 	}
999 	COMMON_END(PRU_SHUTDOWN);
1000 }
1001 
1002 /*
1003  * After a receive, possibly send window update to peer.
1004  */
1005 static int
tcp_usr_rcvd(struct socket * so,int flags)1006 tcp_usr_rcvd(struct socket *so, int flags)
1007 {
1008 	int error = 0;
1009 	struct inpcb *inp = sotoinpcb(so);
1010 	struct tcpcb *tp;
1011 
1012 	COMMON_START();
1013 	/* In case we got disconnected from the peer */
1014 	if (tp == NULL) {
1015 		goto out;
1016 	}
1017 	tcp_sbrcv_trim(tp, &so->so_rcv);
1018 
1019 	if ((flags & MSG_WAITALL) && SEQ_LT(tp->last_ack_sent, tp->rcv_nxt)) {
1020 		tp->t_flags |= TF_ACKNOW;
1021 	}
1022 
1023 	/*
1024 	 * This tcp_output is solely there to trigger window-updates.
1025 	 * However, we really do not want these window-updates while we
1026 	 * are still in SYN_SENT or SYN_RECEIVED.
1027 	 */
1028 	if (TCPS_HAVEESTABLISHED(tp->t_state)) {
1029 		tcp_output(tp);
1030 	}
1031 
1032 #if CONTENT_FILTER
1033 	cfil_sock_buf_update(&so->so_rcv);
1034 #endif /* CONTENT_FILTER */
1035 
1036 	COMMON_END(PRU_RCVD);
1037 }
1038 
1039 __attribute__((noinline))
1040 static int
tcp_send_implied_connect(struct socket * so,struct tcpcb * tp,struct sockaddr * nam,struct proc * p,bool isipv6)1041 tcp_send_implied_connect(struct socket *so, struct tcpcb *tp, struct sockaddr *nam,
1042     struct proc *p, bool isipv6)
1043 {
1044 	int error = 0;
1045 
1046 	error = tcp_usr_connect_common(so, tp, nam, p, isipv6, false);
1047 	if (error != 0) {
1048 		goto out;
1049 	}
1050 	/*
1051 	 * initialize window to default value, and
1052 	 * initialize maxseg/maxopd using peer's cached
1053 	 * MSS.
1054 	 */
1055 	tp->snd_wnd = TTCP_CLIENT_SND_WND;
1056 	tp->max_sndwnd = tp->snd_wnd;
1057 	tcp_mss(tp, -1, IFSCOPE_NONE);
1058 out:
1059 	return error;
1060 }
1061 
1062 __attribute__((noinline))
1063 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)1064 mpkl_tcp_send(struct socket *so, struct tcpcb *tp, uint32_t mpkl_seq, uint32_t mpkl_len,
1065     struct so_mpkl_send_info *mpkl_send_info)
1066 {
1067 	struct inpcb *inp = tp->t_inpcb;
1068 
1069 	if (inp == NULL) {
1070 		return;
1071 	}
1072 
1073 	if ((inp->inp_last_outifp != NULL &&
1074 	    (inp->inp_last_outifp->if_xflags & IFXF_MPK_LOG)) ||
1075 	    (inp->inp_boundifp != NULL &&
1076 	    (inp->inp_boundifp->if_xflags & IFXF_MPK_LOG))) {
1077 		MPKL_TCP_SEND(tcp_mpkl_log_object,
1078 		    mpkl_send_info->mpkl_proto,
1079 		    mpkl_send_info->mpkl_uuid,
1080 		    ntohs(inp->inp_lport),
1081 		    ntohs(inp->inp_fport),
1082 		    mpkl_seq,
1083 		    mpkl_len,
1084 		    so->last_pid,
1085 		    so->so_log_seqn++);
1086 	}
1087 }
1088 
1089 /*
1090  * Do a send by putting data in output queue and updating urgent
1091  * marker if URG set.  Possibly send more data.  Unlike the other
1092  * pru_*() routines, the mbuf chains are our responsibility.  We
1093  * must either enqueue them or free them.  The other pru_* routines
1094  * generally are caller-frees.
1095  *
1096  * Returns:	0			Success
1097  *		ECONNRESET
1098  *		EINVAL
1099  *		ENOBUFS
1100  *	tcp_connect:EADDRINUSE		Address in use
1101  *	tcp_connect:EADDRNOTAVAIL	Address not available.
1102  *	tcp_connect:EINVAL		Invalid argument
1103  *	tcp_connect:EAFNOSUPPORT	Address family not supported [notdef]
1104  *	tcp_connect:EACCES		Permission denied
1105  *	tcp_connect:EAGAIN		Resource unavailable, try again
1106  *	tcp_connect:EPERM		Operation not permitted
1107  *	tcp_output:EADDRNOTAVAIL
1108  *	tcp_output:ENOBUFS
1109  *	tcp_output:EMSGSIZE
1110  *	tcp_output:EHOSTUNREACH
1111  *	tcp_output:ENETUNREACH
1112  *	tcp_output:ENETDOWN
1113  *	tcp_output:ENOMEM
1114  *	tcp_output:EACCES
1115  *	tcp_output:EMSGSIZE
1116  *	tcp_output:ENOBUFS
1117  *	tcp_output:???			[ignorable: mostly IPSEC/firewall/DLIL]
1118  *	tcp6_connect:???		[IPV6 only]
1119  */
1120 static int
tcp_usr_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1121 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
1122     struct sockaddr *nam, struct mbuf *control, struct proc *p)
1123 {
1124 	int error = 0;
1125 	struct inpcb *inp = sotoinpcb(so);
1126 	struct tcpcb *tp;
1127 	uint32_t mpkl_len = 0; /* length of mbuf chain */
1128 	uint32_t mpkl_seq = 0; /* sequence number where new data is added */
1129 	struct so_mpkl_send_info mpkl_send_info = {};
1130 	bool isipv6;
1131 
1132 	TCPDEBUG0;
1133 
1134 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD
1135 #if NECP
1136 	    || (necp_socket_should_use_flow_divert(inp))
1137 #endif /* NECP */
1138 	    ) {
1139 		/*
1140 		 * OOPS! we lost a race, the TCP session got reset after
1141 		 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a
1142 		 * network interrupt in the non-splnet() section of sosend().
1143 		 */
1144 		if (m != NULL) {
1145 			m_freem(m);
1146 		}
1147 		if (control != NULL) {
1148 			m_freem(control);
1149 			control = NULL;
1150 		}
1151 
1152 		if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
1153 			error = ECONNRESET;     /* XXX EPIPE? */
1154 		} else {
1155 			error = EPROTOTYPE;
1156 		}
1157 		tp = NULL;
1158 		TCPDEBUG1();
1159 		goto out;
1160 	}
1161 	isipv6 = nam && nam->sa_family == AF_INET6 ? true : false;
1162 	tp = intotcpcb(inp);
1163 	TCPDEBUG1();
1164 
1165 	calculate_tcp_clock();
1166 
1167 	if (net_mpklog_enabled) {
1168 		mpkl_seq = tp->snd_una + so->so_snd.sb_cc;
1169 		if (m) {
1170 			mpkl_len = m_length(m);
1171 		}
1172 		if (so->so_flags1 & SOF1_MPKL_SEND_INFO) {
1173 			uuid_copy(mpkl_send_info.mpkl_uuid, so->so_mpkl_send_uuid);
1174 			mpkl_send_info.mpkl_proto = so->so_mpkl_send_proto;
1175 		}
1176 	}
1177 
1178 	if (control != NULL) {
1179 		if (control->m_len > 0 && net_mpklog_enabled) {
1180 			error = tcp_get_mpkl_send_info(control, &mpkl_send_info);
1181 			/*
1182 			 * Intepretation of the returned code:
1183 			 *  0: client wants us to use value passed in SCM_MPKL_SEND_INFO
1184 			 *  1: SCM_MPKL_SEND_INFO was not present
1185 			 *  other: failure
1186 			 */
1187 			if (error != 0 && error != ENOMSG) {
1188 				m_freem(control);
1189 				if (m != NULL) {
1190 					m_freem(m);
1191 				}
1192 				control = NULL;
1193 				m = NULL;
1194 				goto out;
1195 			}
1196 		}
1197 		/*
1198 		 * Silently drop unsupported ancillary data messages
1199 		 */
1200 		m_freem(control);
1201 		control = NULL;
1202 	}
1203 
1204 	/* MPTCP sublow socket buffers must not be compressed */
1205 	VERIFY(!(so->so_flags & SOF_MP_SUBFLOW) ||
1206 	    (so->so_snd.sb_flags & SB_NOCOMPRESS));
1207 
1208 	if (!(flags & PRUS_OOB) || (so->so_flags1 & SOF1_PRECONNECT_DATA)) {
1209 		sbappendstream(&so->so_snd, m);
1210 
1211 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1212 			/*
1213 			 * Do implied connect if not yet connected,
1214 			 */
1215 			error = tcp_send_implied_connect(so, tp, nam, p, isipv6);
1216 			if (error != 0) {
1217 				goto out;
1218 			}
1219 			/* The sequence number of the data is past the SYN */
1220 			mpkl_seq = tp->iss + 1;
1221 		}
1222 
1223 		if (flags & PRUS_EOF) {
1224 			/*
1225 			 * Close the send side of the connection after
1226 			 * the data is sent.
1227 			 */
1228 			socantsendmore(so);
1229 			tp = tcp_usrclosed(tp);
1230 		}
1231 		if (tp != NULL) {
1232 			if (flags & PRUS_MORETOCOME) {
1233 				tp->t_flags |= TF_MORETOCOME;
1234 			}
1235 			tp->t_flagsext |= TF_USR_OUTPUT;
1236 			error = tcp_output(tp);
1237 			tp->t_flagsext &= ~TF_USR_OUTPUT;
1238 			if (flags & PRUS_MORETOCOME) {
1239 				tp->t_flags &= ~TF_MORETOCOME;
1240 			}
1241 		}
1242 	} else {
1243 		if (sbspace(&so->so_snd) == 0) {
1244 			/* if no space is left in sockbuf,
1245 			 * do not try to squeeze in OOB traffic */
1246 			m_freem(m);
1247 			error = ENOBUFS;
1248 			goto out;
1249 		}
1250 		/*
1251 		 * According to RFC961 (Assigned Protocols),
1252 		 * the urgent pointer points to the last octet
1253 		 * of urgent data.  We continue, however,
1254 		 * to consider it to indicate the first octet
1255 		 * of data past the urgent section.
1256 		 * Otherwise, snd_up should be one lower.
1257 		 */
1258 		sbappendstream(&so->so_snd, m);
1259 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1260 			/*
1261 			 * Do implied connect if not yet connected,
1262 			 * initialize window to default value, and
1263 			 * initialize maxseg/maxopd using peer's cached
1264 			 * MSS.
1265 			 */
1266 			error = tcp_send_implied_connect(so, tp, nam, p, isipv6);
1267 			if (error != 0) {
1268 				goto out;
1269 			}
1270 		}
1271 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
1272 		tp->t_flagsext |= TF_FORCE;
1273 		tp->t_flagsext |= TF_USR_OUTPUT;
1274 		error = tcp_output(tp);
1275 		tp->t_flagsext &= ~TF_USR_OUTPUT;
1276 		tp->t_flagsext &= ~TF_FORCE;
1277 	}
1278 
1279 	if (net_mpklog_enabled) {
1280 		mpkl_tcp_send(so, tp, mpkl_seq, mpkl_len, &mpkl_send_info);
1281 	}
1282 
1283 	/*
1284 	 * We wait for the socket to successfully connect before returning.
1285 	 * This allows us to signal a timeout to the application.
1286 	 */
1287 	if (so->so_state & SS_ISCONNECTING) {
1288 		if (so->so_state & SS_NBIO) {
1289 			error = EWOULDBLOCK;
1290 		} else {
1291 			error = sbwait(&so->so_snd);
1292 		}
1293 	}
1294 
1295 	COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
1296 	    ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1297 }
1298 
1299 /*
1300  * Abort the TCP.
1301  */
1302 static int
tcp_usr_abort(struct socket * so)1303 tcp_usr_abort(struct socket *so)
1304 {
1305 	int error = 0;
1306 	struct inpcb *inp = sotoinpcb(so);
1307 	struct tcpcb *tp;
1308 
1309 	COMMON_START();
1310 	/* In case we got disconnected from the peer */
1311 	if (tp == NULL) {
1312 		goto out;
1313 	}
1314 	tp = tcp_drop(tp, ECONNABORTED);
1315 	VERIFY(so->so_usecount > 0);
1316 	so->so_usecount--;
1317 	COMMON_END(PRU_ABORT);
1318 }
1319 
1320 /*
1321  * Receive out-of-band data.
1322  *
1323  * Returns:	0			Success
1324  *		EINVAL [COMMON_START]
1325  *		EINVAL
1326  *		EWOULDBLOCK
1327  */
1328 static int
tcp_usr_rcvoob(struct socket * so,struct mbuf * m,int flags)1329 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1330 {
1331 	int error = 0;
1332 	struct inpcb *inp = sotoinpcb(so);
1333 	struct tcpcb *tp;
1334 
1335 	COMMON_START();
1336 	if ((so->so_oobmark == 0 &&
1337 	    (so->so_state & SS_RCVATMARK) == 0) ||
1338 	    so->so_options & SO_OOBINLINE ||
1339 	    tp->t_oobflags & TCPOOB_HADDATA) {
1340 		error = EINVAL;
1341 		goto out;
1342 	}
1343 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1344 		error = EWOULDBLOCK;
1345 		goto out;
1346 	}
1347 	m->m_len = 1;
1348 	*mtod(m, caddr_t) = tp->t_iobc;
1349 	so->so_state &= ~SS_RCVATMARK;
1350 	if ((flags & MSG_PEEK) == 0) {
1351 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1352 	}
1353 	COMMON_END(PRU_RCVOOB);
1354 }
1355 
1356 static int
tcp_usr_preconnect(struct socket * so)1357 tcp_usr_preconnect(struct socket *so)
1358 {
1359 	struct inpcb *inp = sotoinpcb(so);
1360 	int error = 0;
1361 
1362 #if NECP
1363 	if (necp_socket_should_use_flow_divert(inp)) {
1364 		/* May happen, if in tcp_usr_connect we did not had a chance
1365 		 * to set the usrreqs (due to some error). So, let's get out
1366 		 * of here.
1367 		 */
1368 		goto out;
1369 	}
1370 #endif /* NECP */
1371 
1372 	error = tcp_output(sototcpcb(so));
1373 
1374 	soclearfastopen(so);
1375 
1376 	COMMON_END(PRU_PRECONNECT);
1377 }
1378 
1379 /* xxx - should be const */
1380 struct pr_usrreqs tcp_usrreqs = {
1381 	.pru_abort =            tcp_usr_abort,
1382 	.pru_accept =           tcp_usr_accept,
1383 	.pru_attach =           tcp_usr_attach,
1384 	.pru_bind =             tcp_usr_bind,
1385 	.pru_connect =          tcp_usr_connect,
1386 	.pru_connectx =         tcp_usr_connectx,
1387 	.pru_control =          in_control,
1388 	.pru_detach =           tcp_usr_detach,
1389 	.pru_disconnect =       tcp_usr_disconnect,
1390 	.pru_disconnectx =      tcp_usr_disconnectx,
1391 	.pru_listen =           tcp_usr_listen,
1392 	.pru_peeraddr =         in_getpeeraddr,
1393 	.pru_rcvd =             tcp_usr_rcvd,
1394 	.pru_rcvoob =           tcp_usr_rcvoob,
1395 	.pru_send =             tcp_usr_send,
1396 	.pru_shutdown =         tcp_usr_shutdown,
1397 	.pru_sockaddr =         in_getsockaddr,
1398 	.pru_sosend =           sosend,
1399 	.pru_soreceive =        soreceive,
1400 	.pru_preconnect =       tcp_usr_preconnect,
1401 };
1402 
1403 struct pr_usrreqs tcp6_usrreqs = {
1404 	.pru_abort =            tcp_usr_abort,
1405 	.pru_accept =           tcp6_usr_accept,
1406 	.pru_attach =           tcp_usr_attach,
1407 	.pru_bind =             tcp6_usr_bind,
1408 	.pru_connect =          tcp6_usr_connect,
1409 	.pru_connectx =         tcp6_usr_connectx,
1410 	.pru_control =          in6_control,
1411 	.pru_detach =           tcp_usr_detach,
1412 	.pru_disconnect =       tcp_usr_disconnect,
1413 	.pru_disconnectx =      tcp_usr_disconnectx,
1414 	.pru_listen =           tcp6_usr_listen,
1415 	.pru_peeraddr =         in6_mapped_peeraddr,
1416 	.pru_rcvd =             tcp_usr_rcvd,
1417 	.pru_rcvoob =           tcp_usr_rcvoob,
1418 	.pru_send =             tcp_usr_send,
1419 	.pru_shutdown =         tcp_usr_shutdown,
1420 	.pru_sockaddr =         in6_mapped_sockaddr,
1421 	.pru_sosend =           sosend,
1422 	.pru_soreceive =        soreceive,
1423 	.pru_preconnect =       tcp_usr_preconnect,
1424 };
1425 
1426 /*
1427  * Common subroutine to open a TCP connection to remote host specified
1428  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
1429  * port number if needed.  Call in_pcbladdr to do the routing and to choose
1430  * a local host address (interface).  If there is an existing incarnation
1431  * of the same connection in TIME-WAIT state and if the remote host was
1432  * sending CC options and if the connection duration was < MSL, then
1433  * truncate the previous TIME-WAIT state and proceed.
1434  * Initialize connection parameters and enter SYN-SENT state.
1435  *
1436  * Returns:	0			Success
1437  *		EADDRINUSE
1438  *		EINVAL
1439  *	in_pcbbind:EADDRNOTAVAIL	Address not available.
1440  *	in_pcbbind:EINVAL		Invalid argument
1441  *	in_pcbbind:EAFNOSUPPORT		Address family not supported [notdef]
1442  *	in_pcbbind:EACCES		Permission denied
1443  *	in_pcbbind:EADDRINUSE		Address in use
1444  *	in_pcbbind:EAGAIN		Resource unavailable, try again
1445  *	in_pcbbind:EPERM		Operation not permitted
1446  *	in_pcbladdr:EINVAL		Invalid argument
1447  *	in_pcbladdr:EAFNOSUPPORT	Address family not supported
1448  *	in_pcbladdr:EADDRNOTAVAIL	Address not available
1449  */
1450 static int
tcp_connect(struct tcpcb * tp,struct sockaddr * nam,struct proc * p)1451 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct proc *p)
1452 {
1453 	struct inpcb *inp = tp->t_inpcb, *oinp;
1454 	struct socket *so = inp->inp_socket;
1455 	struct tcpcb *otp;
1456 	struct sockaddr_in *sin = (struct sockaddr_in *)(void *)nam;
1457 	struct in_addr laddr;
1458 	int error = 0;
1459 	struct ifnet *outif = NULL;
1460 
1461 	if (inp->inp_lport == 0) {
1462 		error = in_pcbbind(inp, NULL, p);
1463 		if (error) {
1464 			goto done;
1465 		}
1466 	}
1467 
1468 	/*
1469 	 * Cannot simply call in_pcbconnect, because there might be an
1470 	 * earlier incarnation of this same connection still in
1471 	 * TIME_WAIT state, creating an ADDRINUSE error.
1472 	 */
1473 	error = in_pcbladdr(inp, nam, &laddr, IFSCOPE_NONE, &outif, 0);
1474 	if (error) {
1475 		goto done;
1476 	}
1477 
1478 	socket_unlock(inp->inp_socket, 0);
1479 	oinp = in_pcblookup_hash(inp->inp_pcbinfo,
1480 	    sin->sin_addr, sin->sin_port,
1481 	    inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr : laddr,
1482 	    inp->inp_lport, 0, NULL);
1483 
1484 	socket_lock(inp->inp_socket, 0);
1485 	if (oinp) {
1486 		if (oinp != inp) { /* 4143933: avoid deadlock if inp == oinp */
1487 			socket_lock(oinp->inp_socket, 1);
1488 		}
1489 		if (in_pcb_checkstate(oinp, WNT_RELEASE, 1) == WNT_STOPUSING) {
1490 			if (oinp != inp) {
1491 				socket_unlock(oinp->inp_socket, 1);
1492 			}
1493 			goto skip_oinp;
1494 		}
1495 
1496 		if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
1497 		    otp->t_state == TCPS_TIME_WAIT &&
1498 		    ((int)(tcp_now - otp->t_starttime)) < tcp_msl &&
1499 		    (otp->t_flags & TF_RCVD_CC)) {
1500 			otp = tcp_close(otp);
1501 		} else {
1502 			printf("tcp_connect: inp=0x%llx err=EADDRINUSE\n",
1503 			    (uint64_t)VM_KERNEL_ADDRPERM(inp));
1504 			if (oinp != inp) {
1505 				socket_unlock(oinp->inp_socket, 1);
1506 			}
1507 			error = EADDRINUSE;
1508 			goto done;
1509 		}
1510 		if (oinp != inp) {
1511 			socket_unlock(oinp->inp_socket, 1);
1512 		}
1513 	}
1514 skip_oinp:
1515 	if ((inp->inp_laddr.s_addr == INADDR_ANY ? laddr.s_addr :
1516 	    inp->inp_laddr.s_addr) == sin->sin_addr.s_addr &&
1517 	    inp->inp_lport == sin->sin_port) {
1518 		error = EINVAL;
1519 		goto done;
1520 	}
1521 #if SKYWALK
1522 	if (!NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1523 		error = netns_reserve_in(&inp->inp_netns_token,
1524 		    inp->inp_laddr.s_addr != INADDR_ANY ?
1525 		    inp->inp_laddr : laddr,
1526 		    IPPROTO_TCP, inp->inp_lport, NETNS_BSD, NULL);
1527 		if (error) {
1528 			goto done;
1529 		}
1530 	}
1531 #endif /* SKYWALK */
1532 	if (!lck_rw_try_lock_exclusive(&inp->inp_pcbinfo->ipi_lock)) {
1533 		/*lock inversion issue, mostly with udp multicast packets */
1534 		socket_unlock(inp->inp_socket, 0);
1535 		lck_rw_lock_exclusive(&inp->inp_pcbinfo->ipi_lock);
1536 		socket_lock(inp->inp_socket, 0);
1537 	}
1538 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
1539 		inp->inp_laddr = laddr;
1540 		/* no reference needed */
1541 		inp->inp_last_outifp = outif;
1542 #if SKYWALK
1543 		if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1544 			netns_set_ifnet(&inp->inp_netns_token, inp->inp_last_outifp);
1545 		}
1546 #endif /* SKYWALK */
1547 
1548 		inp->inp_flags |= INP_INADDR_ANY;
1549 	}
1550 	inp->inp_faddr = sin->sin_addr;
1551 	inp->inp_fport = sin->sin_port;
1552 	in_pcbrehash(inp);
1553 	lck_rw_done(&inp->inp_pcbinfo->ipi_lock);
1554 
1555 	if (inp->inp_flowhash == 0) {
1556 		inp_calc_flowhash(inp);
1557 		ASSERT(inp->inp_flowhash != 0);
1558 	}
1559 
1560 	tcp_set_max_rwinscale(tp, so);
1561 
1562 	soisconnecting(so);
1563 	tcpstat.tcps_connattempt++;
1564 	TCP_LOG_STATE(tp, TCPS_SYN_SENT);
1565 	tp->t_state = TCPS_SYN_SENT;
1566 	tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp, TCP_CONN_KEEPINIT(tp));
1567 	tp->iss = tcp_new_isn(tp);
1568 	tcp_sendseqinit(tp);
1569 	tp->t_connect_time = tcp_now;
1570 	if (nstat_collect) {
1571 		nstat_route_connect_attempt(inp->inp_route.ro_rt);
1572 	}
1573 
1574 	tcp_add_fsw_flow(tp, outif);
1575 
1576 done:
1577 	if (outif != NULL) {
1578 		ifnet_release(outif);
1579 	}
1580 
1581 	return error;
1582 }
1583 
1584 static int
tcp6_connect(struct tcpcb * tp,struct sockaddr * nam,struct proc * p)1585 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct proc *p)
1586 {
1587 	struct inpcb *inp = tp->t_inpcb, *oinp;
1588 	struct socket *so = inp->inp_socket;
1589 	struct tcpcb *otp;
1590 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)(void *)nam;
1591 	struct in6_addr addr6;
1592 	int error = 0;
1593 	struct ifnet *outif = NULL;
1594 
1595 	if (inp->inp_lport == 0) {
1596 		error = in6_pcbbind(inp, NULL, p);
1597 		if (error) {
1598 			goto done;
1599 		}
1600 	}
1601 
1602 	/*
1603 	 * Cannot simply call in_pcbconnect, because there might be an
1604 	 * earlier incarnation of this same connection still in
1605 	 * TIME_WAIT state, creating an ADDRINUSE error.
1606 	 *
1607 	 * in6_pcbladdr() might return an ifp with its reference held
1608 	 * even in the error case, so make sure that it's released
1609 	 * whenever it's non-NULL.
1610 	 */
1611 	error = in6_pcbladdr(inp, nam, &addr6, &outif);
1612 	if (error) {
1613 		goto done;
1614 	}
1615 	socket_unlock(inp->inp_socket, 0);
1616 
1617 	uint32_t lifscope = IFSCOPE_NONE;
1618 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1619 		lifscope = inp->inp_lifscope;
1620 	} else if (sin6->sin6_scope_id != IFSCOPE_NONE) {
1621 		lifscope = sin6->sin6_scope_id;
1622 	} else if (outif != NULL) {
1623 		lifscope = outif->if_index;
1624 	}
1625 	oinp = in6_pcblookup_hash(inp->inp_pcbinfo,
1626 	    &sin6->sin6_addr, sin6->sin6_port, sin6->sin6_scope_id,
1627 	    IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
1628 	    ? &addr6
1629 	    : &inp->in6p_laddr,
1630 	    inp->inp_lport, lifscope, 0, NULL);
1631 	socket_lock(inp->inp_socket, 0);
1632 	if (oinp) {
1633 		if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
1634 		    otp->t_state == TCPS_TIME_WAIT &&
1635 		    ((int)(tcp_now - otp->t_starttime)) < tcp_msl &&
1636 		    (otp->t_flags & TF_RCVD_CC)) {
1637 			otp = tcp_close(otp);
1638 		} else {
1639 			error = EADDRINUSE;
1640 			goto done;
1641 		}
1642 	}
1643 #if SKYWALK
1644 	if (!NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1645 		error = netns_reserve_in6(&inp->inp_netns_token,
1646 		    IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ?
1647 		    addr6 : inp->in6p_laddr,
1648 		    IPPROTO_TCP, inp->inp_lport, NETNS_BSD, NULL);
1649 		if (error) {
1650 			goto done;
1651 		}
1652 	}
1653 #endif /* SKYWALK */
1654 	if (!lck_rw_try_lock_exclusive(&inp->inp_pcbinfo->ipi_lock)) {
1655 		/*lock inversion issue, mostly with udp multicast packets */
1656 		socket_unlock(inp->inp_socket, 0);
1657 		lck_rw_lock_exclusive(&inp->inp_pcbinfo->ipi_lock);
1658 		socket_lock(inp->inp_socket, 0);
1659 	}
1660 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1661 		inp->in6p_laddr = addr6;
1662 		inp->in6p_last_outifp = outif;  /* no reference needed */
1663 		inp->inp_lifscope = lifscope;
1664 		in6_verify_ifscope(&inp->in6p_laddr, inp->inp_lifscope);
1665 #if SKYWALK
1666 		if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1667 			netns_set_ifnet(&inp->inp_netns_token, inp->in6p_last_outifp);
1668 		}
1669 #endif /* SKYWALK */
1670 		inp->in6p_flags |= INP_IN6ADDR_ANY;
1671 	}
1672 	inp->in6p_faddr = sin6->sin6_addr;
1673 	inp->inp_fport = sin6->sin6_port;
1674 	inp->inp_fifscope = sin6->sin6_scope_id;
1675 	in6_verify_ifscope(&inp->in6p_faddr, inp->inp_fifscope);
1676 	if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != 0) {
1677 		inp->inp_flow = sin6->sin6_flowinfo;
1678 	}
1679 	in_pcbrehash(inp);
1680 	lck_rw_done(&inp->inp_pcbinfo->ipi_lock);
1681 
1682 	if (inp->inp_flowhash == 0) {
1683 		inp_calc_flowhash(inp);
1684 		ASSERT(inp->inp_flowhash != 0);
1685 	}
1686 	/* update flowinfo - RFC 6437 */
1687 	if (inp->inp_flow == 0 && inp->in6p_flags & IN6P_AUTOFLOWLABEL) {
1688 		inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
1689 		inp->inp_flow |=
1690 		    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
1691 	}
1692 
1693 	tcp_set_max_rwinscale(tp, so);
1694 
1695 	soisconnecting(so);
1696 	tcpstat.tcps_connattempt++;
1697 	TCP_LOG_STATE(tp, TCPS_SYN_SENT);
1698 	tp->t_state = TCPS_SYN_SENT;
1699 	tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
1700 	    TCP_CONN_KEEPINIT(tp));
1701 	tp->iss = tcp_new_isn(tp);
1702 	tcp_sendseqinit(tp);
1703 	tp->t_connect_time = tcp_now;
1704 	if (nstat_collect) {
1705 		nstat_route_connect_attempt(inp->inp_route.ro_rt);
1706 	}
1707 
1708 	tcp_add_fsw_flow(tp, outif);
1709 
1710 done:
1711 	if (outif != NULL) {
1712 		ifnet_release(outif);
1713 	}
1714 
1715 	return error;
1716 }
1717 
1718 /*
1719  * Export TCP internal state information via a struct tcp_info
1720  */
1721 void
tcp_fill_info(struct tcpcb * tp,struct tcp_info * ti)1722 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1723 {
1724 	struct inpcb *inp = tp->t_inpcb;
1725 
1726 	bzero(ti, sizeof(*ti));
1727 
1728 	ti->tcpi_state = (uint8_t)tp->t_state;
1729 	ti->tcpi_flowhash = inp != NULL ? inp->inp_flowhash: 0;
1730 
1731 	if (TSTMP_SUPPORTED(tp)) {
1732 		ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1733 	}
1734 	if (SACK_ENABLED(tp)) {
1735 		ti->tcpi_options |= TCPI_OPT_SACK;
1736 	}
1737 	if (TCP_WINDOW_SCALE_ENABLED(tp)) {
1738 		ti->tcpi_options |= TCPI_OPT_WSCALE;
1739 		ti->tcpi_snd_wscale = tp->snd_scale;
1740 		ti->tcpi_rcv_wscale = tp->rcv_scale;
1741 	}
1742 	if (TCP_ECN_ENABLED(tp)) {
1743 		ti->tcpi_options |= TCPI_OPT_ECN;
1744 	}
1745 
1746 	/* Are we in retranmission episode */
1747 	if (IN_FASTRECOVERY(tp) || tp->t_rxtshift > 0) {
1748 		ti->tcpi_flags |= TCPI_FLAG_LOSSRECOVERY;
1749 	}
1750 
1751 	if (tp->t_flags & TF_STREAMING_ON) {
1752 		ti->tcpi_flags |= TCPI_FLAG_STREAMING_ON;
1753 	}
1754 
1755 	ti->tcpi_rto = tp->t_timer[TCPT_REXMT] ? tp->t_rxtcur : 0;
1756 	ti->tcpi_snd_mss = tp->t_maxseg;
1757 	ti->tcpi_rcv_mss = tp->t_maxseg;
1758 
1759 	ti->tcpi_rttcur = tp->t_rttcur;
1760 	ti->tcpi_srtt = tp->t_srtt >> TCP_RTT_SHIFT;
1761 	ti->tcpi_rcv_srtt = tp->rcv_srtt >> TCP_RTT_SHIFT;
1762 	ti->tcpi_rttvar = tp->t_rttvar >> TCP_RTTVAR_SHIFT;
1763 	ti->tcpi_rttbest = tp->t_rttbest >> TCP_RTT_SHIFT;
1764 
1765 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1766 	ti->tcpi_snd_cwnd = tp->snd_cwnd;
1767 	if (inp != NULL && inp->inp_socket != NULL) {
1768 		ti->tcpi_snd_sbbytes = inp->inp_socket->so_snd.sb_cc;
1769 	}
1770 
1771 	ti->tcpi_rcv_space = tp->rcv_adv > tp->rcv_nxt ?
1772 	    tp->rcv_adv - tp->rcv_nxt : 0;
1773 
1774 	ti->tcpi_snd_wnd = tp->snd_wnd;
1775 	ti->tcpi_snd_nxt = tp->snd_nxt;
1776 	ti->tcpi_rcv_nxt = tp->rcv_nxt;
1777 
1778 	/* convert bytes/msec to bits/sec */
1779 	if ((tp->t_flagsext & TF_MEASURESNDBW) != 0 &&
1780 	    tp->t_bwmeas != NULL) {
1781 		ti->tcpi_snd_bw = (tp->t_bwmeas->bw_sndbw * 8000);
1782 	}
1783 
1784 	ti->tcpi_txpackets = inp != NULL ? inp->inp_stat->txpackets : 0;
1785 	ti->tcpi_txbytes = inp != NULL ? inp->inp_stat->txbytes : 0;
1786 	ti->tcpi_txretransmitbytes = tp->t_stat.txretransmitbytes;
1787 	ti->tcpi_txretransmitpackets = tp->t_stat.rxmitpkts;
1788 	ti->tcpi_txunacked = tp->snd_max - tp->snd_una;
1789 
1790 	ti->tcpi_rxpackets = inp != NULL ? inp->inp_stat->rxpackets : 0;
1791 	ti->tcpi_rxbytes = inp != NULL ? inp->inp_stat->rxbytes : 0;
1792 	ti->tcpi_rxduplicatebytes = tp->t_stat.rxduplicatebytes;
1793 	ti->tcpi_rxoutoforderbytes = tp->t_stat.rxoutoforderbytes;
1794 
1795 	if (tp->t_state > TCPS_LISTEN) {
1796 		ti->tcpi_synrexmits = (uint8_t)tp->t_stat.rxmitsyns;
1797 	}
1798 	if (inp != NULL) {
1799 		ti->tcpi_cell_rxpackets = inp->inp_cstat->rxpackets;
1800 		ti->tcpi_cell_rxbytes = inp->inp_cstat->rxbytes;
1801 		ti->tcpi_cell_txpackets = inp->inp_cstat->txpackets;
1802 		ti->tcpi_cell_txbytes = inp->inp_cstat->txbytes;
1803 
1804 		ti->tcpi_wifi_rxpackets = inp->inp_wstat->rxpackets;
1805 		ti->tcpi_wifi_rxbytes = inp->inp_wstat->rxbytes;
1806 		ti->tcpi_wifi_txpackets = inp->inp_wstat->txpackets;
1807 		ti->tcpi_wifi_txbytes = inp->inp_wstat->txbytes;
1808 
1809 		ti->tcpi_wired_rxpackets = inp->inp_Wstat->rxpackets;
1810 		ti->tcpi_wired_rxbytes = inp->inp_Wstat->rxbytes;
1811 		ti->tcpi_wired_txpackets = inp->inp_Wstat->txpackets;
1812 		ti->tcpi_wired_txbytes = inp->inp_Wstat->txbytes;
1813 	}
1814 	tcp_get_connectivity_status(tp, &ti->tcpi_connstatus);
1815 
1816 	ti->tcpi_tfo_syn_data_rcv = !!(tp->t_tfo_stats & TFO_S_SYNDATA_RCV);
1817 	ti->tcpi_tfo_cookie_req_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIEREQ_RECV);
1818 	ti->tcpi_tfo_cookie_sent = !!(tp->t_tfo_stats & TFO_S_COOKIE_SENT);
1819 	ti->tcpi_tfo_cookie_invalid = !!(tp->t_tfo_stats & TFO_S_COOKIE_INVALID);
1820 
1821 	ti->tcpi_tfo_cookie_req = !!(tp->t_tfo_stats & TFO_S_COOKIE_REQ);
1822 	ti->tcpi_tfo_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIE_RCV);
1823 	ti->tcpi_tfo_syn_data_sent = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_SENT);
1824 	ti->tcpi_tfo_syn_data_acked = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_ACKED);
1825 	ti->tcpi_tfo_syn_loss = !!(tp->t_tfo_stats & TFO_S_SYN_LOSS);
1826 	ti->tcpi_tfo_cookie_wrong = !!(tp->t_tfo_stats & TFO_S_COOKIE_WRONG);
1827 	ti->tcpi_tfo_no_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_NO_COOKIE_RCV);
1828 	ti->tcpi_tfo_heuristics_disable = !!(tp->t_tfo_stats & TFO_S_HEURISTICS_DISABLE);
1829 	ti->tcpi_tfo_send_blackhole = !!(tp->t_tfo_stats & TFO_S_SEND_BLACKHOLE);
1830 	ti->tcpi_tfo_recv_blackhole = !!(tp->t_tfo_stats & TFO_S_RECV_BLACKHOLE);
1831 	ti->tcpi_tfo_onebyte_proxy = !!(tp->t_tfo_stats & TFO_S_ONE_BYTE_PROXY);
1832 
1833 	ti->tcpi_ecn_client_setup = !!(tp->ecn_flags & (TE_SETUPSENT | TE_ACE_SETUPSENT));
1834 	ti->tcpi_ecn_server_setup = !!(tp->ecn_flags & (TE_SETUPRECEIVED | TE_ACE_SETUPRECEIVED));
1835 	ti->tcpi_ecn_success = (TCP_ECN_ENABLED(tp) || TCP_ACC_ECN_ON(tp)) ? 1 : 0;
1836 	ti->tcpi_ecn_lost_syn = !!(tp->ecn_flags & TE_LOST_SYN);
1837 	ti->tcpi_ecn_lost_synack = !!(tp->ecn_flags & TE_LOST_SYNACK);
1838 
1839 	ti->tcpi_local_peer = !!(tp->t_flags & TF_LOCAL);
1840 
1841 	if (inp != NULL && inp->inp_last_outifp != NULL) {
1842 		ti->tcpi_last_outif = inp->inp_last_outifp->if_index;
1843 
1844 		if (IFNET_IS_CELLULAR(inp->inp_last_outifp)) {
1845 			ti->tcpi_if_cell = 1;
1846 		}
1847 		if (IFNET_IS_WIFI(inp->inp_last_outifp)) {
1848 			ti->tcpi_if_wifi = 1;
1849 		}
1850 		if (IFNET_IS_WIRED(inp->inp_last_outifp)) {
1851 			ti->tcpi_if_wired = 1;
1852 		}
1853 		if (IFNET_IS_WIFI_INFRA(inp->inp_last_outifp)) {
1854 			ti->tcpi_if_wifi_infra = 1;
1855 		}
1856 		if (inp->inp_last_outifp->if_eflags & IFEF_AWDL) {
1857 			ti->tcpi_if_wifi_awdl = 1;
1858 		}
1859 	}
1860 	if (tp->tcp_cc_index == TCP_CC_ALGO_BACKGROUND_INDEX) {
1861 		ti->tcpi_snd_background = 1;
1862 	}
1863 	if (tcp_recv_bg == 1 || (inp != NULL && inp->inp_socket != NULL &&
1864 	    IS_TCP_RECV_BG(inp->inp_socket))) {
1865 		ti->tcpi_rcv_background = 1;
1866 	}
1867 
1868 	ti->tcpi_ecn_recv_ce = tp->t_ecn_recv_ce;
1869 	ti->tcpi_ecn_recv_cwr = tp->t_ecn_recv_cwr;
1870 
1871 	ti->tcpi_rcvoopack = tp->t_rcvoopack;
1872 	ti->tcpi_pawsdrop = tp->t_pawsdrop;
1873 	ti->tcpi_sack_recovery_episode = tp->t_sack_recovery_episode;
1874 	ti->tcpi_reordered_pkts = tp->t_reordered_pkts;
1875 	ti->tcpi_dsack_sent = tp->t_dsack_sent;
1876 	ti->tcpi_dsack_recvd = tp->t_dsack_recvd;
1877 }
1878 
1879 __private_extern__ errno_t
tcp_fill_info_for_info_tuple(struct info_tuple * itpl,struct tcp_info * ti)1880 tcp_fill_info_for_info_tuple(struct info_tuple *itpl, struct tcp_info *ti)
1881 {
1882 	struct inpcbinfo *pcbinfo = NULL;
1883 	struct inpcb *inp = NULL;
1884 	struct socket *so;
1885 	struct tcpcb *tp;
1886 
1887 	if (itpl->itpl_proto == IPPROTO_TCP) {
1888 		pcbinfo = &tcbinfo;
1889 	} else {
1890 		return EINVAL;
1891 	}
1892 
1893 	if (itpl->itpl_local_sa.sa_family == AF_INET &&
1894 	    itpl->itpl_remote_sa.sa_family == AF_INET) {
1895 		inp = in_pcblookup_hash(pcbinfo,
1896 		    itpl->itpl_remote_sin.sin_addr,
1897 		    itpl->itpl_remote_sin.sin_port,
1898 		    itpl->itpl_local_sin.sin_addr,
1899 		    itpl->itpl_local_sin.sin_port,
1900 		    0, NULL);
1901 	} else if (itpl->itpl_local_sa.sa_family == AF_INET6 &&
1902 	    itpl->itpl_remote_sa.sa_family == AF_INET6) {
1903 		struct in6_addr ina6_local;
1904 		struct in6_addr ina6_remote;
1905 
1906 		ina6_local = itpl->itpl_local_sin6.sin6_addr;
1907 		if (in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&ina6_local) &&
1908 		    itpl->itpl_local_sin6.sin6_scope_id) {
1909 			ina6_local.s6_addr16[1] = htons((uint16_t)itpl->itpl_local_sin6.sin6_scope_id);
1910 		}
1911 
1912 		ina6_remote = itpl->itpl_remote_sin6.sin6_addr;
1913 		if (in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&ina6_remote) &&
1914 		    itpl->itpl_remote_sin6.sin6_scope_id) {
1915 			ina6_remote.s6_addr16[1] = htons((uint16_t)itpl->itpl_remote_sin6.sin6_scope_id);
1916 		}
1917 
1918 		inp = in6_pcblookup_hash(pcbinfo,
1919 		    &ina6_remote,
1920 		    itpl->itpl_remote_sin6.sin6_port,
1921 		    itpl->itpl_remote_sin6.sin6_scope_id,
1922 		    &ina6_local,
1923 		    itpl->itpl_local_sin6.sin6_port,
1924 		    itpl->itpl_local_sin6.sin6_scope_id,
1925 		    0, NULL);
1926 	} else {
1927 		return EINVAL;
1928 	}
1929 
1930 	if (inp != NULL) {
1931 		if ((so = inp->inp_socket) == NULL) {
1932 			return ENOENT;
1933 		}
1934 		socket_lock(so, 0);
1935 		if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
1936 			socket_unlock(so, 0);
1937 			return ENOENT;
1938 		}
1939 		tp = intotcpcb(inp);
1940 
1941 		tcp_fill_info(tp, ti);
1942 		socket_unlock(so, 0);
1943 
1944 		return 0;
1945 	}
1946 #if SKYWALK
1947 	else {
1948 		/* if no pcb found, check for flowswitch for uTCP flow */
1949 		int error;
1950 		struct nexus_mib_filter nmf = {
1951 			.nmf_type = NXMIB_FLOW,
1952 			.nmf_bitmap = NXMIB_FILTER_INFO_TUPLE,
1953 			.nmf_info_tuple = *itpl,
1954 		};
1955 		struct sk_stats_flow sf;
1956 		size_t len = sizeof(sf);
1957 		error = kernel_sysctlbyname(SK_STATS_FLOW, &sf, &len, &nmf, sizeof(nmf));
1958 		if (error != 0) {
1959 			printf("kernel_sysctlbyname err %d\n", error);
1960 			return error;
1961 		}
1962 		if (len != sizeof(sf)) {
1963 			printf("kernel_sysctlbyname invalid len %zu\n", len);
1964 			return ENOENT;
1965 		}
1966 
1967 		/*
1968 		 * This is what flow tracker can offer right now, which is good
1969 		 * for mDNS TCP keep alive offload.
1970 		 */
1971 		ti->tcpi_snd_nxt = sf.sf_lseq;
1972 		ti->tcpi_rcv_nxt = sf.sf_rseq;
1973 		ti->tcpi_rcv_space = (uint32_t)(sf.sf_lmax_win << sf.sf_lwscale);
1974 		ti->tcpi_rcv_wscale = sf.sf_lwscale;
1975 		ti->tcpi_last_outif = (int32_t)sf.sf_if_index;
1976 
1977 		return 0;
1978 	}
1979 #endif /* SKYWALK */
1980 
1981 	return ENOENT;
1982 }
1983 
1984 static void
tcp_connection_fill_info(struct tcpcb * tp,struct tcp_connection_info * tci)1985 tcp_connection_fill_info(struct tcpcb *tp, struct tcp_connection_info *tci)
1986 {
1987 	struct inpcb *inp = tp->t_inpcb;
1988 
1989 	bzero(tci, sizeof(*tci));
1990 	tci->tcpi_state = (uint8_t)tp->t_state;
1991 
1992 	if (TSTMP_SUPPORTED(tp)) {
1993 		tci->tcpi_options |= TCPCI_OPT_TIMESTAMPS;
1994 	}
1995 	if (SACK_ENABLED(tp)) {
1996 		tci->tcpi_options |= TCPCI_OPT_SACK;
1997 	}
1998 	if (TCP_WINDOW_SCALE_ENABLED(tp)) {
1999 		tci->tcpi_options |= TCPCI_OPT_WSCALE;
2000 		tci->tcpi_snd_wscale = tp->snd_scale;
2001 		tci->tcpi_rcv_wscale = tp->rcv_scale;
2002 	}
2003 	if (TCP_ECN_ENABLED(tp)) {
2004 		tci->tcpi_options |= TCPCI_OPT_ECN;
2005 	}
2006 	if (IN_FASTRECOVERY(tp) || tp->t_rxtshift > 0) {
2007 		tci->tcpi_flags |= TCPCI_FLAG_LOSSRECOVERY;
2008 	}
2009 	if (tp->t_flagsext & TF_PKTS_REORDERED) {
2010 		tci->tcpi_flags |= TCPCI_FLAG_REORDERING_DETECTED;
2011 	}
2012 	tci->tcpi_rto = tp->t_timer[TCPT_REXMT] > 0 ? tp->t_rxtcur : 0;
2013 	tci->tcpi_maxseg = tp->t_maxseg;
2014 	tci->tcpi_snd_ssthresh = tp->snd_ssthresh;
2015 	tci->tcpi_snd_cwnd = tp->snd_cwnd;
2016 	tci->tcpi_snd_wnd = tp->snd_wnd;
2017 	if (inp != NULL && inp->inp_socket != NULL) {
2018 		tci->tcpi_snd_sbbytes = inp->inp_socket->so_snd.sb_cc;
2019 	}
2020 	tci->tcpi_rcv_wnd = tp->rcv_adv > tp->rcv_nxt ? tp->rcv_adv - tp->rcv_nxt : 0;
2021 	tci->tcpi_rttcur = tp->t_rttcur;
2022 	tci->tcpi_srtt = (tp->t_srtt >> TCP_RTT_SHIFT);
2023 	tci->tcpi_rttvar = (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
2024 	tci->tcpi_txpackets = inp != NULL ? inp->inp_stat->txpackets : 0;
2025 	tci->tcpi_txbytes = inp != NULL ? inp->inp_stat->txbytes : 0;
2026 	tci->tcpi_txretransmitbytes = tp->t_stat.txretransmitbytes;
2027 	tci->tcpi_txretransmitpackets = tp->t_stat.rxmitpkts;
2028 	tci->tcpi_rxpackets = inp != NULL ? inp->inp_stat->rxpackets : 0;
2029 	tci->tcpi_rxbytes = inp != NULL ? inp->inp_stat->rxbytes : 0;
2030 	tci->tcpi_rxoutoforderbytes = tp->t_stat.rxoutoforderbytes;
2031 
2032 	tci->tcpi_tfo_syn_data_rcv = !!(tp->t_tfo_stats & TFO_S_SYNDATA_RCV);
2033 	tci->tcpi_tfo_cookie_req_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIEREQ_RECV);
2034 	tci->tcpi_tfo_cookie_sent = !!(tp->t_tfo_stats & TFO_S_COOKIE_SENT);
2035 	tci->tcpi_tfo_cookie_invalid = !!(tp->t_tfo_stats & TFO_S_COOKIE_INVALID);
2036 	tci->tcpi_tfo_cookie_req = !!(tp->t_tfo_stats & TFO_S_COOKIE_REQ);
2037 	tci->tcpi_tfo_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIE_RCV);
2038 	tci->tcpi_tfo_syn_data_sent = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_SENT);
2039 	tci->tcpi_tfo_syn_data_acked = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_ACKED);
2040 	tci->tcpi_tfo_syn_loss = !!(tp->t_tfo_stats & TFO_S_SYN_LOSS);
2041 	tci->tcpi_tfo_cookie_wrong = !!(tp->t_tfo_stats & TFO_S_COOKIE_WRONG);
2042 	tci->tcpi_tfo_no_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_NO_COOKIE_RCV);
2043 	tci->tcpi_tfo_heuristics_disable = !!(tp->t_tfo_stats & TFO_S_HEURISTICS_DISABLE);
2044 	tci->tcpi_tfo_send_blackhole = !!(tp->t_tfo_stats & TFO_S_SEND_BLACKHOLE);
2045 	tci->tcpi_tfo_recv_blackhole = !!(tp->t_tfo_stats & TFO_S_RECV_BLACKHOLE);
2046 	tci->tcpi_tfo_onebyte_proxy = !!(tp->t_tfo_stats & TFO_S_ONE_BYTE_PROXY);
2047 }
2048 
2049 
2050 __private_extern__ int
tcp_sysctl_info(__unused struct sysctl_oid * oidp,__unused void * arg1,__unused int arg2,struct sysctl_req * req)2051 tcp_sysctl_info(__unused struct sysctl_oid *oidp, __unused void *arg1, __unused int arg2, struct sysctl_req *req)
2052 {
2053 	int error;
2054 	struct tcp_info ti = {};
2055 	struct info_tuple itpl;
2056 
2057 	if (req->newptr == USER_ADDR_NULL) {
2058 		return EINVAL;
2059 	}
2060 	if (req->newlen < sizeof(struct info_tuple)) {
2061 		return EINVAL;
2062 	}
2063 	error = SYSCTL_IN(req, &itpl, sizeof(struct info_tuple));
2064 	if (error != 0) {
2065 		return error;
2066 	}
2067 	error = tcp_fill_info_for_info_tuple(&itpl, &ti);
2068 	if (error != 0) {
2069 		return error;
2070 	}
2071 	error = SYSCTL_OUT(req, &ti, sizeof(struct tcp_info));
2072 	if (error != 0) {
2073 		return error;
2074 	}
2075 
2076 	return 0;
2077 }
2078 
2079 static int
tcp_lookup_peer_pid_locked(struct socket * so,pid_t * out_pid)2080 tcp_lookup_peer_pid_locked(struct socket *so, pid_t *out_pid)
2081 {
2082 	int error = EHOSTUNREACH;
2083 	*out_pid = -1;
2084 	if ((so->so_state & SS_ISCONNECTED) == 0) {
2085 		return ENOTCONN;
2086 	}
2087 
2088 	struct inpcb    *inp = (struct inpcb*)so->so_pcb;
2089 	uint16_t                lport = inp->inp_lport;
2090 	uint16_t                fport = inp->inp_fport;
2091 	uint32_t                                fifscope = inp->inp_fifscope;
2092 	uint32_t                                lifscope = inp->inp_lifscope;
2093 
2094 	struct inpcb    *finp = NULL;
2095 	struct  in6_addr laddr6, faddr6;
2096 	struct in_addr laddr4, faddr4;
2097 
2098 	if (inp->inp_vflag & INP_IPV6) {
2099 		laddr6 = inp->in6p_laddr;
2100 		faddr6 = inp->in6p_faddr;
2101 	} else if (inp->inp_vflag & INP_IPV4) {
2102 		laddr4 = inp->inp_laddr;
2103 		faddr4 = inp->inp_faddr;
2104 	}
2105 
2106 	socket_unlock(so, 0);
2107 	if (inp->inp_vflag & INP_IPV6) {
2108 		finp = in6_pcblookup_hash(&tcbinfo, &laddr6, lport, lifscope, &faddr6, fport, fifscope, 0, NULL);
2109 	} else if (inp->inp_vflag & INP_IPV4) {
2110 		finp = in_pcblookup_hash(&tcbinfo, laddr4, lport, faddr4, fport, 0, NULL);
2111 	}
2112 
2113 	if (finp) {
2114 		*out_pid = finp->inp_socket->last_pid;
2115 		error = 0;
2116 		in_pcb_checkstate(finp, WNT_RELEASE, 0);
2117 	}
2118 	socket_lock(so, 0);
2119 
2120 	return error;
2121 }
2122 
2123 void
tcp_getconninfo(struct socket * so,struct conninfo_tcp * tcp_ci)2124 tcp_getconninfo(struct socket *so, struct conninfo_tcp *tcp_ci)
2125 {
2126 	tcp_fill_info(sototcpcb(so), &tcp_ci->tcpci_tcp_info);
2127 }
2128 
2129 void
tcp_clear_keep_alive_offload(struct socket * so)2130 tcp_clear_keep_alive_offload(struct socket *so)
2131 {
2132 	struct inpcb *inp;
2133 	struct ifnet *ifp;
2134 
2135 	inp = sotoinpcb(so);
2136 	if (inp == NULL) {
2137 		return;
2138 	}
2139 
2140 	if ((inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD) == 0) {
2141 		return;
2142 	}
2143 
2144 	ifp = inp->inp_boundifp != NULL ? inp->inp_boundifp :
2145 	    inp->inp_last_outifp;
2146 	if (ifp == NULL) {
2147 		panic("%s: so %p inp %p ifp NULL",
2148 		    __func__, so, inp);
2149 	}
2150 
2151 	ifnet_lock_exclusive(ifp);
2152 
2153 	if (ifp->if_tcp_kao_cnt == 0) {
2154 		panic("%s: so %p inp %p ifp %p if_tcp_kao_cnt == 0",
2155 		    __func__, so, inp, ifp);
2156 	}
2157 	ifp->if_tcp_kao_cnt--;
2158 	inp->inp_flags2 &= ~INP2_KEEPALIVE_OFFLOAD;
2159 
2160 	ifnet_lock_done(ifp);
2161 }
2162 
2163 static int
tcp_set_keep_alive_offload(struct socket * so,struct proc * proc)2164 tcp_set_keep_alive_offload(struct socket *so, struct proc *proc)
2165 {
2166 	int error = 0;
2167 	struct inpcb *inp;
2168 	struct ifnet *ifp;
2169 
2170 	inp = sotoinpcb(so);
2171 	if (inp == NULL) {
2172 		return ECONNRESET;
2173 	}
2174 	if ((inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD) != 0) {
2175 		return 0;
2176 	}
2177 
2178 	ifp = inp->inp_boundifp != NULL ? inp->inp_boundifp :
2179 	    inp->inp_last_outifp;
2180 	if (ifp == NULL) {
2181 		error = ENXIO;
2182 		os_log_info(OS_LOG_DEFAULT,
2183 		    "%s: error %d for proc %s[%u] out ifp is not set\n",
2184 		    __func__, error,
2185 		    proc != NULL ? proc->p_comm : "kernel",
2186 		    proc != NULL ? proc_getpid(proc) : 0);
2187 		return ENXIO;
2188 	}
2189 
2190 	error = if_get_tcp_kao_max(ifp);
2191 	if (error != 0) {
2192 		return error;
2193 	}
2194 
2195 	ifnet_lock_exclusive(ifp);
2196 	if (ifp->if_tcp_kao_cnt < ifp->if_tcp_kao_max) {
2197 		ifp->if_tcp_kao_cnt++;
2198 		inp->inp_flags2 |= INP2_KEEPALIVE_OFFLOAD;
2199 	} else {
2200 		error = ETOOMANYREFS;
2201 		os_log_info(OS_LOG_DEFAULT,
2202 		    "%s: error %d for proc %s[%u] if_tcp_kao_max %u\n",
2203 		    __func__, error,
2204 		    proc != NULL ? proc->p_comm : "kernel",
2205 		    proc != NULL ? proc_getpid(proc) : 0,
2206 		    ifp->if_tcp_kao_max);
2207 	}
2208 	ifnet_lock_done(ifp);
2209 
2210 	return error;
2211 }
2212 
2213 /*
2214  * The new sockopt interface makes it possible for us to block in the
2215  * copyin/out step (if we take a page fault).  Taking a page fault at
2216  * splnet() is probably a Bad Thing.  (Since sockets and pcbs both now
2217  * use TSM, there probably isn't any need for this function to run at
2218  * splnet() any more.  This needs more examination.)
2219  */
2220 int
tcp_ctloutput(struct socket * so,struct sockopt * sopt)2221 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
2222 {
2223 	int     error = 0, opt = 0, optval = 0;
2224 	struct  inpcb *inp;
2225 	struct  tcpcb *tp;
2226 
2227 	inp = sotoinpcb(so);
2228 	if (inp == NULL) {
2229 		return ECONNRESET;
2230 	}
2231 	/* Allow <SOL_SOCKET,SO_FLUSH/SO_TRAFFIC_MGT_BACKGROUND> at this level */
2232 	if (sopt->sopt_level != IPPROTO_TCP &&
2233 	    !(sopt->sopt_level == SOL_SOCKET && (sopt->sopt_name == SO_FLUSH ||
2234 	    sopt->sopt_name == SO_TRAFFIC_MGT_BACKGROUND))) {
2235 		if (SOCK_CHECK_DOM(so, PF_INET6)) {
2236 			error = ip6_ctloutput(so, sopt);
2237 		} else {
2238 			error = ip_ctloutput(so, sopt);
2239 		}
2240 		return error;
2241 	}
2242 	tp = intotcpcb(inp);
2243 	if (tp == NULL) {
2244 		return ECONNRESET;
2245 	}
2246 
2247 	calculate_tcp_clock();
2248 
2249 	switch (sopt->sopt_dir) {
2250 	case SOPT_SET:
2251 		switch (sopt->sopt_name) {
2252 		case TCP_NODELAY:
2253 		case TCP_NOOPT:
2254 		case TCP_NOPUSH:
2255 			error = sooptcopyin(sopt, &optval, sizeof optval,
2256 			    sizeof optval);
2257 			if (error) {
2258 				break;
2259 			}
2260 
2261 			switch (sopt->sopt_name) {
2262 			case TCP_NODELAY:
2263 				opt = TF_NODELAY;
2264 				break;
2265 			case TCP_NOOPT:
2266 				opt = TF_NOOPT;
2267 				break;
2268 			case TCP_NOPUSH:
2269 				opt = TF_NOPUSH;
2270 				break;
2271 			default:
2272 				opt = 0; /* dead code to fool gcc */
2273 				break;
2274 			}
2275 
2276 			if (optval) {
2277 				tp->t_flags |= opt;
2278 			} else {
2279 				tp->t_flags &= ~opt;
2280 			}
2281 			break;
2282 		case TCP_RXT_FINDROP:
2283 		case TCP_NOTIMEWAIT:
2284 			error = sooptcopyin(sopt, &optval, sizeof optval,
2285 			    sizeof optval);
2286 			if (error) {
2287 				break;
2288 			}
2289 			switch (sopt->sopt_name) {
2290 			case TCP_RXT_FINDROP:
2291 				opt = TF_RXTFINDROP;
2292 				break;
2293 			case TCP_NOTIMEWAIT:
2294 				opt = TF_NOTIMEWAIT;
2295 				break;
2296 			default:
2297 				opt = 0;
2298 				break;
2299 			}
2300 			if (optval) {
2301 				tp->t_flagsext |= opt;
2302 			} else {
2303 				tp->t_flagsext &= ~opt;
2304 			}
2305 			break;
2306 		case TCP_MEASURE_SND_BW:
2307 			error = sooptcopyin(sopt, &optval, sizeof optval,
2308 			    sizeof optval);
2309 			if (error) {
2310 				break;
2311 			}
2312 			opt = TF_MEASURESNDBW;
2313 			if (optval) {
2314 				if (tp->t_bwmeas == NULL) {
2315 					tp->t_bwmeas = tcp_bwmeas_alloc(tp);
2316 					if (tp->t_bwmeas == NULL) {
2317 						error = ENOMEM;
2318 						break;
2319 					}
2320 				}
2321 				tp->t_flagsext |= opt;
2322 			} else {
2323 				tp->t_flagsext &= ~opt;
2324 				/* Reset snd bw measurement state */
2325 				tp->t_flagsext &= ~(TF_BWMEAS_INPROGRESS);
2326 				if (tp->t_bwmeas != NULL) {
2327 					tcp_bwmeas_free(tp);
2328 				}
2329 			}
2330 			break;
2331 		case TCP_MEASURE_BW_BURST: {
2332 			struct tcp_measure_bw_burst in;
2333 			uint32_t minpkts, maxpkts;
2334 			bzero(&in, sizeof(in));
2335 
2336 			error = sooptcopyin(sopt, &in, sizeof(in),
2337 			    sizeof(in));
2338 			if (error) {
2339 				break;
2340 			}
2341 			if ((tp->t_flagsext & TF_MEASURESNDBW) == 0 ||
2342 			    tp->t_bwmeas == NULL) {
2343 				error = EINVAL;
2344 				break;
2345 			}
2346 			minpkts = (in.min_burst_size != 0) ? in.min_burst_size :
2347 			    tp->t_bwmeas->bw_minsizepkts;
2348 			maxpkts = (in.max_burst_size != 0) ? in.max_burst_size :
2349 			    tp->t_bwmeas->bw_maxsizepkts;
2350 			if (minpkts > maxpkts) {
2351 				error = EINVAL;
2352 				break;
2353 			}
2354 			tp->t_bwmeas->bw_minsizepkts = minpkts;
2355 			tp->t_bwmeas->bw_maxsizepkts = maxpkts;
2356 			tp->t_bwmeas->bw_minsize = (minpkts * tp->t_maxseg);
2357 			tp->t_bwmeas->bw_maxsize = (maxpkts * tp->t_maxseg);
2358 			break;
2359 		}
2360 		case TCP_MAXSEG:
2361 			error = sooptcopyin(sopt, &optval, sizeof optval,
2362 			    sizeof optval);
2363 			if (error) {
2364 				break;
2365 			}
2366 
2367 			if (optval > 0 && optval <= tp->t_maxseg &&
2368 			    optval + 40 >= tcp_minmss) {
2369 				tp->t_maxseg = optval;
2370 			} else {
2371 				error = EINVAL;
2372 			}
2373 			break;
2374 
2375 		case TCP_KEEPALIVE:
2376 			error = sooptcopyin(sopt, &optval, sizeof optval,
2377 			    sizeof optval);
2378 			if (error) {
2379 				break;
2380 			}
2381 			if (optval < 0 || optval > UINT32_MAX / TCP_RETRANSHZ) {
2382 				error = EINVAL;
2383 			} else {
2384 				tp->t_keepidle = optval * TCP_RETRANSHZ;
2385 				/* reset the timer to new value */
2386 				tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
2387 				    TCP_CONN_KEEPIDLE(tp));
2388 				tcp_check_timer_state(tp);
2389 			}
2390 			break;
2391 
2392 		case TCP_CONNECTIONTIMEOUT:
2393 			error = sooptcopyin(sopt, &optval, sizeof optval,
2394 			    sizeof optval);
2395 			if (error) {
2396 				break;
2397 			}
2398 			if (optval < 0 || optval > UINT32_MAX / TCP_RETRANSHZ) {
2399 				error = EINVAL;
2400 			} else {
2401 				tp->t_keepinit = optval * TCP_RETRANSHZ;
2402 				if (tp->t_state == TCPS_SYN_RECEIVED ||
2403 				    tp->t_state == TCPS_SYN_SENT) {
2404 					tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
2405 					    TCP_CONN_KEEPINIT(tp));
2406 					tcp_check_timer_state(tp);
2407 				}
2408 			}
2409 			break;
2410 
2411 		case TCP_KEEPINTVL:
2412 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2413 			    sizeof(optval));
2414 			if (error) {
2415 				break;
2416 			}
2417 			if (optval < 0 || optval > UINT32_MAX / TCP_RETRANSHZ) {
2418 				error = EINVAL;
2419 			} else {
2420 				tp->t_keepintvl = optval * TCP_RETRANSHZ;
2421 				if (tp->t_state == TCPS_FIN_WAIT_2 &&
2422 				    TCP_CONN_MAXIDLE(tp) > 0) {
2423 					tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp,
2424 					    TCP_CONN_MAXIDLE(tp));
2425 					tcp_check_timer_state(tp);
2426 				}
2427 			}
2428 			break;
2429 
2430 		case TCP_KEEPCNT:
2431 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2432 			    sizeof(optval));
2433 			if (error) {
2434 				break;
2435 			}
2436 			if (optval < 0 || optval > INT32_MAX) {
2437 				error = EINVAL;
2438 			} else {
2439 				tp->t_keepcnt = optval;
2440 				if (tp->t_state == TCPS_FIN_WAIT_2 &&
2441 				    TCP_CONN_MAXIDLE(tp) > 0) {
2442 					tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp,
2443 					    TCP_CONN_MAXIDLE(tp));
2444 					tcp_check_timer_state(tp);
2445 				}
2446 			}
2447 			break;
2448 
2449 		case TCP_KEEPALIVE_OFFLOAD:
2450 			if ((error = priv_check_cred(kauth_cred_get(),
2451 			    PRIV_NETINET_TCP_KA_OFFLOAD, 0)) != 0) {
2452 				break;
2453 			}
2454 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2455 			    sizeof(optval));
2456 			if (error) {
2457 				break;
2458 			}
2459 			if (optval < 0 || optval > INT32_MAX) {
2460 				error = EINVAL;
2461 				break;
2462 			}
2463 			if (optval != 0) {
2464 				error = tcp_set_keep_alive_offload(so,
2465 				    sopt->sopt_p);
2466 			} else {
2467 				tcp_clear_keep_alive_offload(so);
2468 			}
2469 			break;
2470 
2471 		case PERSIST_TIMEOUT:
2472 			error = sooptcopyin(sopt, &optval, sizeof optval,
2473 			    sizeof optval);
2474 			if (error) {
2475 				break;
2476 			}
2477 			if (optval < 0) {
2478 				error = EINVAL;
2479 			} else {
2480 				tp->t_persist_timeout = optval * TCP_RETRANSHZ;
2481 			}
2482 			break;
2483 		case TCP_RXT_CONNDROPTIME:
2484 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2485 			    sizeof(optval));
2486 			if (error) {
2487 				break;
2488 			}
2489 			if (optval < 0) {
2490 				error = EINVAL;
2491 			} else {
2492 				tp->t_rxt_conndroptime = optval * TCP_RETRANSHZ;
2493 			}
2494 			break;
2495 		case TCP_NOTSENT_LOWAT:
2496 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2497 			    sizeof(optval));
2498 			if (error) {
2499 				break;
2500 			}
2501 			if (optval < 0) {
2502 				error = EINVAL;
2503 				break;
2504 			} else {
2505 				if (optval == 0) {
2506 					so->so_flags &= ~(SOF_NOTSENT_LOWAT);
2507 					tp->t_notsent_lowat = 0;
2508 				} else {
2509 					so->so_flags |= SOF_NOTSENT_LOWAT;
2510 					tp->t_notsent_lowat = optval;
2511 				}
2512 			}
2513 			break;
2514 		case TCP_ADAPTIVE_READ_TIMEOUT:
2515 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2516 			    sizeof(optval));
2517 			if (error) {
2518 				break;
2519 			}
2520 			if (optval < 0 ||
2521 			    optval > TCP_ADAPTIVE_TIMEOUT_MAX) {
2522 				error = EINVAL;
2523 				break;
2524 			} else if (optval == 0) {
2525 				tp->t_adaptive_rtimo = 0;
2526 				tcp_keepalive_reset(tp);
2527 
2528 				if (tp->t_mpsub) {
2529 					mptcp_reset_keepalive(tp);
2530 				}
2531 			} else {
2532 				tp->t_adaptive_rtimo = (uint8_t)optval;
2533 			}
2534 			break;
2535 		case TCP_ADAPTIVE_WRITE_TIMEOUT:
2536 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2537 			    sizeof(optval));
2538 			if (error) {
2539 				break;
2540 			}
2541 			if (optval < 0 ||
2542 			    optval > TCP_ADAPTIVE_TIMEOUT_MAX) {
2543 				error = EINVAL;
2544 				break;
2545 			} else {
2546 				tp->t_adaptive_wtimo = (uint8_t)optval;
2547 			}
2548 			break;
2549 		case TCP_SENDMOREACKS:
2550 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2551 			    sizeof(optval));
2552 			if (error) {
2553 				break;
2554 			}
2555 			if (optval < 0 || optval > 1) {
2556 				error = EINVAL;
2557 			} else if (optval == 0) {
2558 				tp->t_flagsext &= ~(TF_NOSTRETCHACK);
2559 			} else {
2560 				tp->t_flagsext |= TF_NOSTRETCHACK;
2561 			}
2562 			break;
2563 		case TCP_DISABLE_BLACKHOLE_DETECTION:
2564 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2565 			    sizeof(optval));
2566 			if (error) {
2567 				break;
2568 			}
2569 			if (optval < 0 || optval > 1) {
2570 				error = EINVAL;
2571 			} else if (optval == 0) {
2572 				tp->t_flagsext &= ~TF_NOBLACKHOLE_DETECTION;
2573 			} else {
2574 				tp->t_flagsext |= TF_NOBLACKHOLE_DETECTION;
2575 				if ((tp->t_flags & TF_BLACKHOLE) &&
2576 				    tp->t_pmtud_saved_maxopd > 0) {
2577 					tcp_pmtud_revert_segment_size(tp);
2578 				}
2579 			}
2580 			break;
2581 		case TCP_FASTOPEN:
2582 			if (!(tcp_fastopen & TCP_FASTOPEN_SERVER)) {
2583 				error = ENOTSUP;
2584 				break;
2585 			}
2586 
2587 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2588 			    sizeof(optval));
2589 			if (error) {
2590 				break;
2591 			}
2592 			if (optval < 0 || optval > 1) {
2593 				error = EINVAL;
2594 				break;
2595 			}
2596 			if (tp->t_state != TCPS_LISTEN) {
2597 				error =  EINVAL;
2598 				break;
2599 			}
2600 			if (optval) {
2601 				tp->t_flagsext |= TF_FASTOPEN;
2602 			} else {
2603 				tcp_disable_tfo(tp);
2604 			}
2605 			break;
2606 		case TCP_FASTOPEN_FORCE_HEURISTICS:
2607 
2608 			break;
2609 		case TCP_FASTOPEN_FORCE_ENABLE:
2610 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2611 			    sizeof(optval));
2612 
2613 			if (error) {
2614 				break;
2615 			}
2616 			if (optval < 0 || optval > 1) {
2617 				error = EINVAL;
2618 				break;
2619 			}
2620 
2621 			if (tp->t_state != TCPS_CLOSED) {
2622 				error =  EINVAL;
2623 				break;
2624 			}
2625 			if (optval) {
2626 				tp->t_flagsext |= TF_FASTOPEN_FORCE_ENABLE;
2627 			} else {
2628 				tp->t_flagsext &= ~TF_FASTOPEN_FORCE_ENABLE;
2629 			}
2630 
2631 			break;
2632 		case TCP_ENABLE_ECN:
2633 			error = sooptcopyin(sopt, &optval, sizeof optval,
2634 			    sizeof optval);
2635 			if (error) {
2636 				break;
2637 			}
2638 			if (optval) {
2639 				tp->ecn_flags |= TE_ECN_MODE_ENABLE;
2640 				tp->ecn_flags &= ~TE_ECN_MODE_DISABLE;
2641 			} else {
2642 				tp->ecn_flags &= ~TE_ECN_MODE_ENABLE;
2643 				tp->ecn_flags |= TE_ECN_MODE_DISABLE;
2644 			}
2645 			break;
2646 		case TCP_ECN_MODE:
2647 			error = sooptcopyin(sopt, &optval, sizeof optval,
2648 			    sizeof optval);
2649 			if (error) {
2650 				break;
2651 			}
2652 			if (optval == ECN_MODE_DEFAULT) {
2653 				tp->ecn_flags &= ~TE_ECN_MODE_ENABLE;
2654 				tp->ecn_flags &= ~TE_ECN_MODE_DISABLE;
2655 			} else if (optval == ECN_MODE_ENABLE) {
2656 				tp->ecn_flags |= TE_ECN_MODE_ENABLE;
2657 				tp->ecn_flags &= ~TE_ECN_MODE_DISABLE;
2658 			} else if (optval == ECN_MODE_DISABLE) {
2659 				tp->ecn_flags &= ~TE_ECN_MODE_ENABLE;
2660 				tp->ecn_flags |= TE_ECN_MODE_DISABLE;
2661 			} else {
2662 				error = EINVAL;
2663 			}
2664 			break;
2665 		case TCP_NOTIFY_ACKNOWLEDGEMENT:
2666 			error = sooptcopyin(sopt, &optval,
2667 			    sizeof(optval), sizeof(optval));
2668 			if (error) {
2669 				break;
2670 			}
2671 			if (optval <= 0) {
2672 				error = EINVAL;
2673 				break;
2674 			}
2675 			if (tp->t_notify_ack_count >= TCP_MAX_NOTIFY_ACK) {
2676 				error = ETOOMANYREFS;
2677 				break;
2678 			}
2679 
2680 			/*
2681 			 * validate that the given marker id is not
2682 			 * a duplicate to avoid ambiguity
2683 			 */
2684 			if ((error = tcp_notify_ack_id_valid(tp, so,
2685 			    optval)) != 0) {
2686 				break;
2687 			}
2688 			error = tcp_add_notify_ack_marker(tp, optval);
2689 			break;
2690 		case SO_FLUSH:
2691 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
2692 			    sizeof(optval))) != 0) {
2693 				break;
2694 			}
2695 
2696 			error = inp_flush(inp, optval);
2697 			break;
2698 
2699 		case SO_TRAFFIC_MGT_BACKGROUND:
2700 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
2701 			    sizeof(optval))) != 0) {
2702 				break;
2703 			}
2704 
2705 			if (optval) {
2706 				socket_set_traffic_mgt_flags_locked(so,
2707 				    TRAFFIC_MGT_SO_BACKGROUND);
2708 			} else {
2709 				socket_clear_traffic_mgt_flags_locked(so,
2710 				    TRAFFIC_MGT_SO_BACKGROUND);
2711 			}
2712 			break;
2713 		case TCP_RXT_MINIMUM_TIMEOUT:
2714 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2715 			    sizeof(optval));
2716 			if (error) {
2717 				break;
2718 			}
2719 			if (optval < 0) {
2720 				error = EINVAL;
2721 				break;
2722 			}
2723 			if (optval == 0) {
2724 				tp->t_rxt_minimum_timeout = 0;
2725 			} else {
2726 				tp->t_rxt_minimum_timeout = min(optval,
2727 				    TCP_RXT_MINIMUM_TIMEOUT_LIMIT);
2728 				/* convert to milliseconds */
2729 				tp->t_rxt_minimum_timeout *= TCP_RETRANSHZ;
2730 			}
2731 			break;
2732 		default:
2733 			error = ENOPROTOOPT;
2734 			break;
2735 		}
2736 		break;
2737 
2738 	case SOPT_GET:
2739 		switch (sopt->sopt_name) {
2740 		case TCP_NODELAY:
2741 			optval = tp->t_flags & TF_NODELAY;
2742 			break;
2743 		case TCP_MAXSEG:
2744 			optval = tp->t_maxseg;
2745 			break;
2746 		case TCP_KEEPALIVE:
2747 			if (tp->t_keepidle > 0) {
2748 				optval = tp->t_keepidle / TCP_RETRANSHZ;
2749 			} else {
2750 				optval = tcp_keepidle  / TCP_RETRANSHZ;
2751 			}
2752 			break;
2753 		case TCP_KEEPINTVL:
2754 			if (tp->t_keepintvl > 0) {
2755 				optval = tp->t_keepintvl / TCP_RETRANSHZ;
2756 			} else {
2757 				optval = tcp_keepintvl / TCP_RETRANSHZ;
2758 			}
2759 			break;
2760 		case TCP_KEEPCNT:
2761 			if (tp->t_keepcnt > 0) {
2762 				optval = tp->t_keepcnt;
2763 			} else {
2764 				optval = tcp_keepcnt;
2765 			}
2766 			break;
2767 		case TCP_KEEPALIVE_OFFLOAD:
2768 			optval = !!(inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD);
2769 			break;
2770 		case TCP_NOOPT:
2771 			optval = tp->t_flags & TF_NOOPT;
2772 			break;
2773 		case TCP_NOPUSH:
2774 			optval = tp->t_flags & TF_NOPUSH;
2775 			break;
2776 		case TCP_ENABLE_ECN:
2777 			optval = (tp->ecn_flags & TE_ECN_MODE_ENABLE) ? 1 : 0;
2778 			break;
2779 		case TCP_ECN_MODE:
2780 			if (tp->ecn_flags & TE_ECN_MODE_ENABLE) {
2781 				optval = ECN_MODE_ENABLE;
2782 			} else if (tp->ecn_flags & TE_ECN_MODE_DISABLE) {
2783 				optval = ECN_MODE_DISABLE;
2784 			} else {
2785 				optval = ECN_MODE_DEFAULT;
2786 			}
2787 			break;
2788 		case TCP_CONNECTIONTIMEOUT:
2789 			optval = tp->t_keepinit / TCP_RETRANSHZ;
2790 			break;
2791 		case PERSIST_TIMEOUT:
2792 			optval = tp->t_persist_timeout / TCP_RETRANSHZ;
2793 			break;
2794 		case TCP_RXT_CONNDROPTIME:
2795 			optval = tp->t_rxt_conndroptime / TCP_RETRANSHZ;
2796 			break;
2797 		case TCP_RXT_FINDROP:
2798 			optval = tp->t_flagsext & TF_RXTFINDROP;
2799 			break;
2800 		case TCP_NOTIMEWAIT:
2801 			optval = (tp->t_flagsext & TF_NOTIMEWAIT) ? 1 : 0;
2802 			break;
2803 		case TCP_FASTOPEN:
2804 			if (tp->t_state != TCPS_LISTEN ||
2805 			    !(tcp_fastopen & TCP_FASTOPEN_SERVER)) {
2806 				error = ENOTSUP;
2807 				break;
2808 			}
2809 			optval = tfo_enabled(tp);
2810 			break;
2811 		case TCP_FASTOPEN_FORCE_HEURISTICS:
2812 			optval = 0;
2813 			break;
2814 		case TCP_FASTOPEN_FORCE_ENABLE:
2815 			optval = (tp->t_flagsext & TF_FASTOPEN_FORCE_ENABLE) ? 1 : 0;
2816 			break;
2817 		case TCP_MEASURE_SND_BW:
2818 			optval = tp->t_flagsext & TF_MEASURESNDBW;
2819 			break;
2820 		case TCP_INFO: {
2821 			struct tcp_info ti;
2822 
2823 			tcp_fill_info(tp, &ti);
2824 			error = sooptcopyout(sopt, &ti, sizeof(struct tcp_info));
2825 			goto done;
2826 			/* NOT REACHED */
2827 		}
2828 		case TCP_CONNECTION_INFO: {
2829 			struct tcp_connection_info tci;
2830 			tcp_connection_fill_info(tp, &tci);
2831 			error = sooptcopyout(sopt, &tci,
2832 			    sizeof(struct tcp_connection_info));
2833 			goto done;
2834 		}
2835 		case TCP_MEASURE_BW_BURST: {
2836 			struct tcp_measure_bw_burst out = {};
2837 			if ((tp->t_flagsext & TF_MEASURESNDBW) == 0 ||
2838 			    tp->t_bwmeas == NULL) {
2839 				error = EINVAL;
2840 				break;
2841 			}
2842 			out.min_burst_size = tp->t_bwmeas->bw_minsizepkts;
2843 			out.max_burst_size = tp->t_bwmeas->bw_maxsizepkts;
2844 			error = sooptcopyout(sopt, &out, sizeof(out));
2845 			goto done;
2846 		}
2847 		case TCP_NOTSENT_LOWAT:
2848 			if ((so->so_flags & SOF_NOTSENT_LOWAT) != 0) {
2849 				optval = tp->t_notsent_lowat;
2850 			} else {
2851 				optval = 0;
2852 			}
2853 			break;
2854 		case TCP_SENDMOREACKS:
2855 			if (tp->t_flagsext & TF_NOSTRETCHACK) {
2856 				optval = 1;
2857 			} else {
2858 				optval = 0;
2859 			}
2860 			break;
2861 		case TCP_DISABLE_BLACKHOLE_DETECTION:
2862 			if (tp->t_flagsext & TF_NOBLACKHOLE_DETECTION) {
2863 				optval = 1;
2864 			} else {
2865 				optval = 0;
2866 			}
2867 			break;
2868 		case TCP_PEER_PID: {
2869 			pid_t   pid;
2870 			error = tcp_lookup_peer_pid_locked(so, &pid);
2871 			if (error == 0) {
2872 				error = sooptcopyout(sopt, &pid, sizeof(pid));
2873 			}
2874 			goto done;
2875 		}
2876 		case TCP_ADAPTIVE_READ_TIMEOUT:
2877 			optval = tp->t_adaptive_rtimo;
2878 			break;
2879 		case TCP_ADAPTIVE_WRITE_TIMEOUT:
2880 			optval = tp->t_adaptive_wtimo;
2881 			break;
2882 		case SO_TRAFFIC_MGT_BACKGROUND:
2883 			optval = (so->so_flags1 &
2884 			    SOF1_TRAFFIC_MGT_SO_BACKGROUND) ? 1 : 0;
2885 			break;
2886 		case TCP_NOTIFY_ACKNOWLEDGEMENT: {
2887 			struct tcp_notify_ack_complete retid;
2888 
2889 			if (sopt->sopt_valsize != sizeof(retid)) {
2890 				error = EINVAL;
2891 				break;
2892 			}
2893 			bzero(&retid, sizeof(retid));
2894 			tcp_get_notify_ack_count(tp, &retid);
2895 			if (retid.notify_complete_count > 0) {
2896 				tcp_get_notify_ack_ids(tp, &retid);
2897 			}
2898 
2899 			error = sooptcopyout(sopt, &retid, sizeof(retid));
2900 			goto done;
2901 		}
2902 		case TCP_RXT_MINIMUM_TIMEOUT:
2903 			optval = tp->t_rxt_minimum_timeout / TCP_RETRANSHZ;
2904 			break;
2905 		default:
2906 			error = ENOPROTOOPT;
2907 			break;
2908 		}
2909 		if (error == 0) {
2910 			error = sooptcopyout(sopt, &optval, sizeof optval);
2911 		}
2912 		break;
2913 	}
2914 done:
2915 	return error;
2916 }
2917 
2918 /*
2919  * tcp_sendspace and tcp_recvspace are the default send and receive window
2920  * sizes, respectively.  These are obsolescent (this information should
2921  * be set by the route).
2922  */
2923 u_int32_t       tcp_sendspace = 1448 * 256;
2924 u_int32_t       tcp_recvspace = 1448 * 384;
2925 
2926 /* During attach, the size of socket buffer allocated is limited to
2927  * sb_max in sbreserve. Disallow setting the tcp send and recv space
2928  * to be more than sb_max because that will cause tcp_attach to fail
2929  * (see radar 5713060)
2930  */
2931 static int
sysctl_tcp_sospace(struct sysctl_oid * oidp,__unused void * arg1,int arg2,struct sysctl_req * req)2932 sysctl_tcp_sospace(struct sysctl_oid *oidp, __unused void *arg1,
2933     int arg2, struct sysctl_req *req)
2934 {
2935 #pragma unused(arg2)
2936 	u_int32_t new_value = 0, *space_p = NULL;
2937 	int changed = 0, error = 0;
2938 	u_quad_t sb_effective_max = (sb_max / (MSIZE + MCLBYTES)) * MCLBYTES;
2939 
2940 	switch (oidp->oid_number) {
2941 	case TCPCTL_SENDSPACE:
2942 		space_p = &tcp_sendspace;
2943 		break;
2944 	case TCPCTL_RECVSPACE:
2945 		space_p = &tcp_recvspace;
2946 		break;
2947 	default:
2948 		return EINVAL;
2949 	}
2950 	error = sysctl_io_number(req, *space_p, sizeof(u_int32_t),
2951 	    &new_value, &changed);
2952 	if (changed) {
2953 		if (new_value > 0 && new_value <= sb_effective_max) {
2954 			*space_p = new_value;
2955 			SYSCTL_SKMEM_UPDATE_AT_OFFSET(arg2, new_value);
2956 		} else {
2957 			error = ERANGE;
2958 		}
2959 	}
2960 	return error;
2961 }
2962 
2963 #if SYSCTL_SKMEM
2964 SYSCTL_PROC(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace,
2965     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_sendspace,
2966     offsetof(skmem_sysctl, tcp.sendspace), sysctl_tcp_sospace,
2967     "IU", "Maximum outgoing TCP datagram size");
2968 SYSCTL_PROC(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace,
2969     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_recvspace,
2970     offsetof(skmem_sysctl, tcp.recvspace), sysctl_tcp_sospace,
2971     "IU", "Maximum incoming TCP datagram size");
2972 #else /* SYSCTL_SKMEM */
2973 SYSCTL_PROC(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
2974     &tcp_sendspace, 0, &sysctl_tcp_sospace, "IU", "Maximum outgoing TCP datagram size");
2975 SYSCTL_PROC(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
2976     &tcp_recvspace, 0, &sysctl_tcp_sospace, "IU", "Maximum incoming TCP datagram size");
2977 #endif /* SYSCTL_SKMEM */
2978 
2979 /*
2980  * Attach TCP protocol to socket, allocating
2981  * internet protocol control block, tcp control block,
2982  * bufer space, and entering LISTEN state if to accept connections.
2983  *
2984  * Returns:	0			Success
2985  *	in_pcballoc:ENOBUFS
2986  *	in_pcballoc:ENOMEM
2987  *	in_pcballoc:???			[IPSEC specific]
2988  *	soreserve:ENOBUFS
2989  */
2990 static int
tcp_attach(struct socket * so,struct proc * p)2991 tcp_attach(struct socket *so, struct proc *p)
2992 {
2993 	struct tcpcb *tp;
2994 	struct inpcb *inp;
2995 	int error;
2996 	int isipv6 = SOCK_CHECK_DOM(so, PF_INET6) != 0;
2997 
2998 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
2999 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
3000 		if (error) {
3001 			return error;
3002 		}
3003 	}
3004 
3005 	error = in_pcballoc(so, &tcbinfo, p);
3006 	if (error) {
3007 		return error;
3008 	}
3009 
3010 	inp = sotoinpcb(so);
3011 
3012 	if (so->so_snd.sb_preconn_hiwat == 0) {
3013 		soreserve_preconnect(so, 2048);
3014 	}
3015 
3016 	if ((so->so_rcv.sb_flags & SB_USRSIZE) == 0) {
3017 		so->so_rcv.sb_flags |= SB_AUTOSIZE;
3018 	}
3019 	if ((so->so_snd.sb_flags & SB_USRSIZE) == 0) {
3020 		so->so_snd.sb_flags |= SB_AUTOSIZE;
3021 	}
3022 
3023 	if (isipv6) {
3024 		inp->inp_vflag |= INP_IPV6;
3025 		inp->in6p_hops = -1;    /* use kernel default */
3026 	} else {
3027 		inp->inp_vflag |= INP_IPV4;
3028 	}
3029 	tp = tcp_newtcpcb(inp);
3030 	if (tp == NULL) {
3031 		int nofd = so->so_state & SS_NOFDREF;   /* XXX */
3032 
3033 		so->so_state &= ~SS_NOFDREF;    /* don't free the socket yet */
3034 		if (isipv6) {
3035 			in6_pcbdetach(inp);
3036 		} else {
3037 			in_pcbdetach(inp);
3038 		}
3039 		so->so_state |= nofd;
3040 		return ENOBUFS;
3041 	}
3042 	if (nstat_collect) {
3043 		nstat_tcp_new_pcb(inp);
3044 	}
3045 	TCP_LOG_STATE(tp, TCPS_CLOSED);
3046 	tp->t_state = TCPS_CLOSED;
3047 	return 0;
3048 }
3049 
3050 /*
3051  * Initiate (or continue) disconnect.
3052  * If embryonic state, just send reset (once).
3053  * If in ``let data drain'' option and linger null, just drop.
3054  * Otherwise (hard), mark socket disconnecting and drop
3055  * current input data; switch states based on user close, and
3056  * send segment to peer (with FIN).
3057  */
3058 static struct tcpcb *
tcp_disconnect(struct tcpcb * tp)3059 tcp_disconnect(struct tcpcb *tp)
3060 {
3061 	struct socket *so = tp->t_inpcb->inp_socket;
3062 
3063 	if (so->so_rcv.sb_cc != 0 || tp->t_reassqlen != 0) {
3064 		return tcp_drop(tp, 0);
3065 	}
3066 
3067 	if (tp->t_state < TCPS_ESTABLISHED) {
3068 		tp = tcp_close(tp);
3069 	} else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
3070 		tp = tcp_drop(tp, 0);
3071 	} else {
3072 		soisdisconnecting(so);
3073 		sbflush(&so->so_rcv);
3074 		tp = tcp_usrclosed(tp);
3075 #if MPTCP
3076 		/* A reset has been sent but socket exists, do not send FIN */
3077 		if ((so->so_flags & SOF_MP_SUBFLOW) &&
3078 		    (tp) && (tp->t_mpflags & TMPF_RESET)) {
3079 			return tp;
3080 		}
3081 #endif
3082 		if (tp) {
3083 			(void) tcp_output(tp);
3084 		}
3085 	}
3086 	return tp;
3087 }
3088 
3089 /*
3090  * User issued close, and wish to trail through shutdown states:
3091  * if never received SYN, just forget it.  If got a SYN from peer,
3092  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
3093  * If already got a FIN from peer, then almost done; go to LAST_ACK
3094  * state.  In all other cases, have already sent FIN to peer (e.g.
3095  * after PRU_SHUTDOWN), and just have to play tedious game waiting
3096  * for peer to send FIN or not respond to keep-alives, etc.
3097  * We can let the user exit from the close as soon as the FIN is acked.
3098  */
3099 static struct tcpcb *
tcp_usrclosed(struct tcpcb * tp)3100 tcp_usrclosed(struct tcpcb *tp)
3101 {
3102 	switch (tp->t_state) {
3103 	case TCPS_CLOSED:
3104 	case TCPS_LISTEN:
3105 	case TCPS_SYN_SENT:
3106 		tp = tcp_close(tp);
3107 		break;
3108 
3109 	case TCPS_SYN_RECEIVED:
3110 		tp->t_flags |= TF_NEEDFIN;
3111 		break;
3112 
3113 	case TCPS_ESTABLISHED:
3114 		DTRACE_TCP4(state__change, void, NULL,
3115 		    struct inpcb *, tp->t_inpcb,
3116 		    struct tcpcb *, tp,
3117 		    int32_t, TCPS_FIN_WAIT_1);
3118 		TCP_LOG_STATE(tp, TCPS_FIN_WAIT_1);
3119 		tp->t_state = TCPS_FIN_WAIT_1;
3120 		TCP_LOG_CONNECTION_SUMMARY(tp);
3121 		break;
3122 
3123 	case TCPS_CLOSE_WAIT:
3124 		DTRACE_TCP4(state__change, void, NULL,
3125 		    struct inpcb *, tp->t_inpcb,
3126 		    struct tcpcb *, tp,
3127 		    int32_t, TCPS_LAST_ACK);
3128 		TCP_LOG_STATE(tp, TCPS_LAST_ACK);
3129 		tp->t_state = TCPS_LAST_ACK;
3130 		TCP_LOG_CONNECTION_SUMMARY(tp);
3131 		break;
3132 	}
3133 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
3134 		soisdisconnected(tp->t_inpcb->inp_socket);
3135 		/* To prevent the connection hanging in FIN_WAIT_2 forever. */
3136 		if (tp->t_state == TCPS_FIN_WAIT_2) {
3137 			tcp_set_finwait_timeout(tp);
3138 		}
3139 	}
3140 	return tp;
3141 }
3142 
3143 void
tcp_in_cksum_stats(u_int32_t len)3144 tcp_in_cksum_stats(u_int32_t len)
3145 {
3146 	tcpstat.tcps_rcv_swcsum++;
3147 	tcpstat.tcps_rcv_swcsum_bytes += len;
3148 }
3149 
3150 void
tcp_out_cksum_stats(u_int32_t len)3151 tcp_out_cksum_stats(u_int32_t len)
3152 {
3153 	tcpstat.tcps_snd_swcsum++;
3154 	tcpstat.tcps_snd_swcsum_bytes += len;
3155 }
3156 
3157 void
tcp_in6_cksum_stats(u_int32_t len)3158 tcp_in6_cksum_stats(u_int32_t len)
3159 {
3160 	tcpstat.tcps_rcv6_swcsum++;
3161 	tcpstat.tcps_rcv6_swcsum_bytes += len;
3162 }
3163 
3164 void
tcp_out6_cksum_stats(u_int32_t len)3165 tcp_out6_cksum_stats(u_int32_t len)
3166 {
3167 	tcpstat.tcps_snd6_swcsum++;
3168 	tcpstat.tcps_snd6_swcsum_bytes += len;
3169 }
3170 
3171 int
tcp_get_mpkl_send_info(struct mbuf * control,struct so_mpkl_send_info * mpkl_send_info)3172 tcp_get_mpkl_send_info(struct mbuf *control,
3173     struct so_mpkl_send_info *mpkl_send_info)
3174 {
3175 	struct cmsghdr *cm;
3176 
3177 	if (control == NULL || mpkl_send_info == NULL) {
3178 		return EINVAL;
3179 	}
3180 
3181 	for (cm = M_FIRST_CMSGHDR(control); cm;
3182 	    cm = M_NXT_CMSGHDR(control, cm)) {
3183 		if (cm->cmsg_len < sizeof(struct cmsghdr) ||
3184 		    cm->cmsg_len > control->m_len) {
3185 			return EINVAL;
3186 		}
3187 		if (cm->cmsg_level != SOL_SOCKET ||
3188 		    cm->cmsg_type != SCM_MPKL_SEND_INFO) {
3189 			continue;
3190 		}
3191 		if (cm->cmsg_len != CMSG_LEN(sizeof(struct so_mpkl_send_info))) {
3192 			return EINVAL;
3193 		}
3194 		memcpy(mpkl_send_info, CMSG_DATA(cm),
3195 		    sizeof(struct so_mpkl_send_info));
3196 		return 0;
3197 	}
3198 	return ENOMSG;
3199 }
3200 
3201 /*
3202  * tcp socket options.
3203  *
3204  * The switch statement below does nothing at runtime, as it serves as a
3205  * compile time check to ensure that all of the tcp socket options are
3206  * unique.  This works as long as this routine gets updated each time a
3207  * new tcp socket option gets added.
3208  *
3209  * Any failures at compile time indicates duplicated tcp socket option
3210  * values.
3211  */
3212 static __attribute__((unused)) void
tcpsockopt_cassert(void)3213 tcpsockopt_cassert(void)
3214 {
3215 	/*
3216 	 * This is equivalent to _CASSERT() and the compiler wouldn't
3217 	 * generate any instructions, thus for compile time only.
3218 	 */
3219 	switch ((int)0) {
3220 	case 0:
3221 
3222 	/* bsd/netinet/tcp.h */
3223 	case TCP_NODELAY:
3224 	case TCP_MAXSEG:
3225 	case TCP_NOPUSH:
3226 	case TCP_NOOPT:
3227 	case TCP_KEEPALIVE:
3228 	case TCP_CONNECTIONTIMEOUT:
3229 	case PERSIST_TIMEOUT:
3230 	case TCP_RXT_CONNDROPTIME:
3231 	case TCP_RXT_FINDROP:
3232 	case TCP_KEEPINTVL:
3233 	case TCP_KEEPCNT:
3234 	case TCP_SENDMOREACKS:
3235 	case TCP_ENABLE_ECN:
3236 	case TCP_FASTOPEN:
3237 	case TCP_CONNECTION_INFO:
3238 	case TCP_NOTSENT_LOWAT:
3239 
3240 	/* bsd/netinet/tcp_private.h */
3241 	case TCP_INFO:
3242 	case TCP_MEASURE_SND_BW:
3243 	case TCP_MEASURE_BW_BURST:
3244 	case TCP_PEER_PID:
3245 	case TCP_ADAPTIVE_READ_TIMEOUT:
3246 	case TCP_OPTION_UNUSED_0:
3247 	case TCP_ADAPTIVE_WRITE_TIMEOUT:
3248 	case TCP_NOTIMEWAIT:
3249 	case TCP_DISABLE_BLACKHOLE_DETECTION:
3250 	case TCP_ECN_MODE:
3251 	case TCP_KEEPALIVE_OFFLOAD:
3252 		;
3253 	}
3254 }
3255