xref: /xnu-8020.121.3/bsd/netinet/tcp_usrreq.c (revision fdd8201d7b966f0c3ea610489d29bd841d358941)
1 /*
2  * Copyright (c) 2000-2021 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 errno_t tcp_fill_info_for_info_tuple(struct info_tuple *, struct tcp_info *);
124 
125 int tcp_sysctl_info(struct sysctl_oid *, void *, int, struct sysctl_req *);
126 static void tcp_connection_fill_info(struct tcpcb *tp,
127     struct tcp_connection_info *tci);
128 static int tcp_get_mpkl_send_info(struct mbuf *, struct so_mpkl_send_info *);
129 
130 /*
131  * TCP protocol interface to socket abstraction.
132  */
133 static int      tcp_attach(struct socket *, struct proc *);
134 static int      tcp_connect(struct tcpcb *, struct sockaddr *, struct proc *);
135 static int      tcp6_connect(struct tcpcb *, struct sockaddr *, struct proc *);
136 static int      tcp6_usr_connect(struct socket *, struct sockaddr *,
137     struct proc *);
138 static struct tcpcb *tcp_disconnect(struct tcpcb *);
139 static struct tcpcb *tcp_usrclosed(struct tcpcb *);
140 extern void tcp_sbrcv_trim(struct tcpcb *tp, struct sockbuf *sb);
141 
142 #if TCPDEBUG
143 #define TCPDEBUG0       int ostate = 0
144 #define TCPDEBUG1()     ostate = tp ? tp->t_state : 0
145 #define TCPDEBUG2(req)  if (tp && (so->so_options & SO_DEBUG)) \
146 	                        tcp_trace(TA_USER, ostate, tp, 0, 0, req)
147 #else
148 #define TCPDEBUG0
149 #define TCPDEBUG1()
150 #define TCPDEBUG2(req)
151 #endif
152 
153 SYSCTL_PROC(_net_inet_tcp, OID_AUTO, info,
154     CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_ANYBODY | CTLFLAG_KERN,
155     0, 0, tcp_sysctl_info, "S", "TCP info per tuple");
156 
157 /*
158  * TCP attaches to socket via pru_attach(), reserving space,
159  * and an internet control block.
160  *
161  * Returns:	0			Success
162  *		EISCONN
163  *	tcp_attach:ENOBUFS
164  *	tcp_attach:ENOMEM
165  *	tcp_attach:???			[IPSEC specific]
166  */
167 static int
tcp_usr_attach(struct socket * so,__unused int proto,struct proc * p)168 tcp_usr_attach(struct socket *so, __unused int proto, struct proc *p)
169 {
170 	int error;
171 	struct inpcb *inp = sotoinpcb(so);
172 	struct tcpcb *tp = 0;
173 	TCPDEBUG0;
174 
175 	TCPDEBUG1();
176 	if (inp) {
177 		error = EISCONN;
178 		goto out;
179 	}
180 
181 	error = tcp_attach(so, p);
182 	if (error) {
183 		goto out;
184 	}
185 
186 	if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
187 		so->so_linger = (short)(TCP_LINGERTIME * hz);
188 	}
189 	tp = sototcpcb(so);
190 out:
191 	TCPDEBUG2(PRU_ATTACH);
192 	return error;
193 }
194 
195 /*
196  * pru_detach() detaches the TCP protocol from the socket.
197  * If the protocol state is non-embryonic, then can't
198  * do this directly: have to initiate a pru_disconnect(),
199  * which may finish later; embryonic TCB's can just
200  * be discarded here.
201  */
202 static int
tcp_usr_detach(struct socket * so)203 tcp_usr_detach(struct socket *so)
204 {
205 	int error = 0;
206 	struct inpcb *inp = sotoinpcb(so);
207 	struct tcpcb *tp;
208 	TCPDEBUG0;
209 
210 	if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) {
211 		return EINVAL;  /* XXX */
212 	}
213 	socket_lock_assert_owned(so);
214 	tp = intotcpcb(inp);
215 	/* In case we got disconnected from the peer */
216 	if (tp == NULL) {
217 		goto out;
218 	}
219 	TCPDEBUG1();
220 
221 	calculate_tcp_clock();
222 
223 	tp = tcp_disconnect(tp);
224 out:
225 	TCPDEBUG2(PRU_DETACH);
226 	return error;
227 }
228 
229 #if NECP
230 #define COMMON_START_ALLOW_FLOW_DIVERT(allow)  TCPDEBUG0;               \
231 do {                                                                    \
232 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD)          \
233 	        return (EINVAL);                                        \
234 	if (!(allow) && necp_socket_should_use_flow_divert(inp))        \
235 	        return (EPROTOTYPE);                                    \
236 	tp = intotcpcb(inp);                                            \
237 	TCPDEBUG1();                                                    \
238 	calculate_tcp_clock();                                          \
239 } while (0)
240 #else /* NECP */
241 #define COMMON_START_ALLOW_FLOW_DIVERT(allow)  TCPDEBUG0;               \
242 do {                                                                    \
243 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD)          \
244 	        return (EINVAL);                                        \
245 	tp = intotcpcb(inp);                                            \
246 	TCPDEBUG1();                                                    \
247 	calculate_tcp_clock();                                          \
248 } while (0)
249 #endif /* !NECP */
250 
251 #define COMMON_START() COMMON_START_ALLOW_FLOW_DIVERT(false)
252 #define COMMON_END(req) out: TCPDEBUG2(req); return error; goto out
253 
254 
255 /*
256  * Give the socket an address.
257  *
258  * Returns:	0			Success
259  *		EINVAL			Invalid argument [COMMON_START]
260  *		EAFNOSUPPORT		Address family not supported
261  *	in_pcbbind:EADDRNOTAVAIL	Address not available.
262  *	in_pcbbind:EINVAL		Invalid argument
263  *	in_pcbbind:EAFNOSUPPORT		Address family not supported [notdef]
264  *	in_pcbbind:EACCES		Permission denied
265  *	in_pcbbind:EADDRINUSE		Address in use
266  *	in_pcbbind:EAGAIN		Resource unavailable, try again
267  *	in_pcbbind:EPERM		Operation not permitted
268  */
269 static int
tcp_usr_bind(struct socket * so,struct sockaddr * nam,struct proc * p)270 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
271 {
272 	int error = 0;
273 	struct inpcb *inp = sotoinpcb(so);
274 	struct tcpcb *tp;
275 	struct sockaddr_in *sinp;
276 
277 	COMMON_START_ALLOW_FLOW_DIVERT(true);
278 
279 	if (nam->sa_family != 0 && nam->sa_family != AF_INET) {
280 		error = EAFNOSUPPORT;
281 		goto out;
282 	}
283 
284 	/*
285 	 * Must check for multicast and broadcast addresses and disallow binding
286 	 * to them.
287 	 */
288 	sinp = (struct sockaddr_in *)(void *)nam;
289 	if (sinp->sin_family == AF_INET &&
290 	    (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)) ||
291 	    sinp->sin_addr.s_addr == INADDR_BROADCAST)) {
292 		error = EAFNOSUPPORT;
293 		goto out;
294 	}
295 	error = in_pcbbind(inp, nam, p);
296 	if (error) {
297 		goto out;
298 	}
299 
300 #if NECP
301 	/* Update NECP client with bind result if not in middle of connect */
302 	if ((inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS) &&
303 	    !uuid_is_null(inp->necp_client_uuid)) {
304 		socket_unlock(so, 0);
305 		necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
306 		socket_lock(so, 0);
307 	}
308 #endif /* NECP */
309 
310 	COMMON_END(PRU_BIND);
311 }
312 
313 static int
tcp6_usr_bind(struct socket * so,struct sockaddr * nam,struct proc * p)314 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
315 {
316 	int error = 0;
317 	struct inpcb *inp = sotoinpcb(so);
318 	struct tcpcb *tp;
319 	struct sockaddr_in6 *sin6p;
320 
321 	COMMON_START_ALLOW_FLOW_DIVERT(true);
322 
323 	if (nam->sa_family != 0 && nam->sa_family != AF_INET6) {
324 		error = EAFNOSUPPORT;
325 		goto out;
326 	}
327 
328 	/*
329 	 * Must check for multicast and broadcast addresses and disallow binding
330 	 * to them.
331 	 */
332 	sin6p = (struct sockaddr_in6 *)(void *)nam;
333 	if (sin6p->sin6_family == AF_INET6 &&
334 	    (IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr) ||
335 	    ((IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr) ||
336 	    IN6_IS_ADDR_V4COMPAT(&sin6p->sin6_addr)) &&
337 	    (IN_MULTICAST(ntohl(sin6p->sin6_addr.s6_addr32[3])) ||
338 	    sin6p->sin6_addr.s6_addr32[3] == INADDR_BROADCAST)))) {
339 		error = EAFNOSUPPORT;
340 		goto out;
341 	}
342 	inp->inp_vflag &= ~INP_IPV4;
343 	inp->inp_vflag |= INP_IPV6;
344 	if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
345 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr)) {
346 			inp->inp_vflag |= INP_IPV4;
347 		} else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
348 			struct sockaddr_in sin;
349 			const uint8_t old_flags = inp->inp_vflag;
350 
351 			in6_sin6_2_sin(&sin, sin6p);
352 			inp->inp_vflag |= INP_IPV4;
353 			inp->inp_vflag &= ~INP_IPV6;
354 
355 			error = in_pcbbind(inp, (struct sockaddr *)&sin, p);
356 			if (error != 0) {
357 				inp->inp_vflag = old_flags;
358 			}
359 			goto out;
360 		}
361 	}
362 	error = in6_pcbbind(inp, nam, p);
363 	if (error) {
364 		goto out;
365 	}
366 	COMMON_END(PRU_BIND);
367 }
368 
369 /*
370  * Prepare to accept connections.
371  *
372  * Returns:	0			Success
373  *		EINVAL [COMMON_START]
374  *	in_pcbbind:EADDRNOTAVAIL	Address not available.
375  *	in_pcbbind:EINVAL		Invalid argument
376  *	in_pcbbind:EAFNOSUPPORT		Address family not supported [notdef]
377  *	in_pcbbind:EACCES		Permission denied
378  *	in_pcbbind:EADDRINUSE		Address in use
379  *	in_pcbbind:EAGAIN		Resource unavailable, try again
380  *	in_pcbbind:EPERM		Operation not permitted
381  */
382 static int
tcp_usr_listen(struct socket * so,struct proc * p)383 tcp_usr_listen(struct socket *so, struct proc *p)
384 {
385 	int error = 0;
386 	struct inpcb *inp = sotoinpcb(so);
387 	struct tcpcb *tp;
388 
389 	COMMON_START_ALLOW_FLOW_DIVERT(true);
390 	if (inp->inp_lport == 0) {
391 		error = in_pcbbind(inp, NULL, p);
392 	}
393 	if (error == 0) {
394 		tp->t_state = TCPS_LISTEN;
395 	}
396 	TCP_LOG_LISTEN(tp, error);
397 	COMMON_END(PRU_LISTEN);
398 }
399 
400 static int
tcp6_usr_listen(struct socket * so,struct proc * p)401 tcp6_usr_listen(struct socket *so, struct proc *p)
402 {
403 	int error = 0;
404 	struct inpcb *inp = sotoinpcb(so);
405 	struct tcpcb *tp;
406 
407 	COMMON_START_ALLOW_FLOW_DIVERT(true);
408 	if (inp->inp_lport == 0) {
409 		inp->inp_vflag &= ~INP_IPV4;
410 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
411 			inp->inp_vflag |= INP_IPV4;
412 		}
413 		error = in6_pcbbind(inp, NULL, p);
414 	}
415 	if (error == 0) {
416 		tp->t_state = TCPS_LISTEN;
417 	}
418 	TCP_LOG_LISTEN(tp, error);
419 	COMMON_END(PRU_LISTEN);
420 }
421 
422 static int
tcp_connect_complete(struct socket * so)423 tcp_connect_complete(struct socket *so)
424 {
425 	struct tcpcb *tp = sototcpcb(so);
426 	struct inpcb *inp = sotoinpcb(so);
427 	int error = 0;
428 
429 	/* TFO delays the tcp_output until later, when the app calls write() */
430 	if (so->so_flags1 & SOF1_PRECONNECT_DATA) {
431 		if (!necp_socket_is_allowed_to_send_recv(sotoinpcb(so), NULL, 0, NULL, NULL, NULL, NULL)) {
432 			TCP_LOG_DROP_NECP(NULL, NULL, tp, true);
433 			return EHOSTUNREACH;
434 		}
435 
436 		/* Initialize enough state so that we can actually send data */
437 		tcp_mss(tp, -1, IFSCOPE_NONE);
438 		tp->snd_wnd = tp->t_maxseg;
439 		tp->max_sndwnd = tp->snd_wnd;
440 	} else {
441 		error = tcp_output(tp);
442 	}
443 
444 #if NECP
445 	/* Update NECP client with connected five-tuple */
446 	if (error == 0 && !uuid_is_null(inp->necp_client_uuid)) {
447 		socket_unlock(so, 0);
448 		necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
449 		socket_lock(so, 0);
450 	}
451 #endif /* NECP */
452 
453 	return error;
454 }
455 
456 /*
457  * Initiate connection to peer.
458  * Create a template for use in transmissions on this connection.
459  * Enter SYN_SENT state, and mark socket as connecting.
460  * Start keep-alive timer, and seed output sequence space.
461  * Send initial segment on connection.
462  */
463 static int
tcp_usr_connect(struct socket * so,struct sockaddr * nam,struct proc * p)464 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
465 {
466 	int error = 0;
467 	struct inpcb *inp = sotoinpcb(so);
468 	struct tcpcb *tp;
469 	struct sockaddr_in *sinp;
470 
471 	TCPDEBUG0;
472 	if (inp == NULL) {
473 		return EINVAL;
474 	} else if (inp->inp_state == INPCB_STATE_DEAD) {
475 		if (so->so_error) {
476 			error = so->so_error;
477 			so->so_error = 0;
478 			return error;
479 		} else {
480 			return EINVAL;
481 		}
482 	}
483 #if NECP
484 #if CONTENT_FILTER
485 	error = cfil_sock_attach(so, NULL, nam, CFS_CONNECTION_DIR_OUT);
486 	if (error != 0) {
487 		return error;
488 	}
489 #endif /* CONTENT_FILTER */
490 #if FLOW_DIVERT
491 	if (necp_socket_should_use_flow_divert(inp)) {
492 		error = flow_divert_pcb_init(so);
493 		if (error == 0) {
494 			error = flow_divert_connect_out(so, nam, p);
495 		}
496 		return error;
497 	}
498 #endif /* FLOW_DIVERT */
499 #endif /* NECP */
500 	tp = intotcpcb(inp);
501 	TCPDEBUG1();
502 
503 	calculate_tcp_clock();
504 
505 	if (nam->sa_family != 0 && nam->sa_family != AF_INET) {
506 		error = EAFNOSUPPORT;
507 		goto out;
508 	}
509 	/*
510 	 * Must disallow TCP ``connections'' to multicast and broadcast addresses.
511 	 */
512 	sinp = (struct sockaddr_in *)(void *)nam;
513 	if (sinp->sin_family == AF_INET &&
514 	    (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr)) ||
515 	    sinp->sin_addr.s_addr == INADDR_BROADCAST)) {
516 		error = EAFNOSUPPORT;
517 		goto out;
518 	}
519 
520 	if ((error = tcp_connect(tp, nam, p)) != 0) {
521 		TCP_LOG_CONNECT(tp, true, error);
522 		goto out;
523 	}
524 
525 	error = tcp_connect_complete(so);
526 
527 	TCP_LOG_CONNECT(tp, true, error);
528 
529 	COMMON_END(PRU_CONNECT);
530 }
531 
532 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)533 tcp_usr_connectx_common(struct socket *so, int af,
534     struct sockaddr *src, struct sockaddr *dst,
535     struct proc *p, uint32_t ifscope, sae_associd_t aid, sae_connid_t *pcid,
536     uint32_t flags, void *arg, uint32_t arglen, struct uio *auio,
537     user_ssize_t *bytes_written)
538 {
539 #pragma unused(aid, flags, arg, arglen)
540 	struct inpcb *inp = sotoinpcb(so);
541 	int error = 0;
542 	user_ssize_t datalen = 0;
543 
544 	if (inp == NULL) {
545 		return EINVAL;
546 	}
547 
548 	VERIFY(dst != NULL);
549 
550 	ASSERT(!(inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS));
551 	inp->inp_flags2 |= INP2_CONNECT_IN_PROGRESS;
552 
553 #if NECP
554 	inp_update_necp_policy(inp, src, dst, ifscope);
555 #endif /* NECP */
556 
557 	if ((so->so_flags1 & SOF1_DATA_IDEMPOTENT) &&
558 	    (tcp_fastopen & TCP_FASTOPEN_CLIENT)) {
559 		sototcpcb(so)->t_flagsext |= TF_FASTOPEN;
560 	}
561 
562 	/* bind socket to the specified interface, if requested */
563 	if (ifscope != IFSCOPE_NONE &&
564 	    (error = inp_bindif(inp, ifscope, NULL)) != 0) {
565 		goto done;
566 	}
567 
568 	/* if source address and/or port is specified, bind to it */
569 	if (src != NULL) {
570 		error = sobindlock(so, src, 0); /* already locked */
571 		if (error != 0) {
572 			goto done;
573 		}
574 	}
575 
576 	switch (af) {
577 	case AF_INET:
578 		error = tcp_usr_connect(so, dst, p);
579 		break;
580 	case AF_INET6:
581 		error = tcp6_usr_connect(so, dst, p);
582 		break;
583 	default:
584 		VERIFY(0);
585 		/* NOTREACHED */
586 	}
587 
588 	if (error != 0) {
589 		goto done;
590 	}
591 
592 	/* if there is data, copy it */
593 	if (auio != NULL) {
594 		socket_unlock(so, 0);
595 
596 		VERIFY(bytes_written != NULL);
597 
598 		datalen = uio_resid(auio);
599 		error = so->so_proto->pr_usrreqs->pru_sosend(so, NULL,
600 		    (uio_t)auio, NULL, NULL, 0);
601 		socket_lock(so, 0);
602 
603 		if (error == 0 || error == EWOULDBLOCK) {
604 			*bytes_written = datalen - uio_resid(auio);
605 		}
606 
607 		/*
608 		 * sosend returns EWOULDBLOCK if it's a non-blocking
609 		 * socket or a timeout occured (this allows to return
610 		 * the amount of queued data through sendit()).
611 		 *
612 		 * However, connectx() returns EINPROGRESS in case of a
613 		 * blocking socket. So we change the return value here.
614 		 */
615 		if (error == EWOULDBLOCK) {
616 			error = EINPROGRESS;
617 		}
618 	}
619 
620 	if (error == 0 && pcid != NULL) {
621 		*pcid = 1; /* there is only one connection in regular TCP */
622 	}
623 done:
624 	if (error && error != EINPROGRESS) {
625 		so->so_flags1 &= ~SOF1_PRECONNECT_DATA;
626 	}
627 
628 	inp->inp_flags2 &= ~INP2_CONNECT_IN_PROGRESS;
629 	return error;
630 }
631 
632 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)633 tcp_usr_connectx(struct socket *so, struct sockaddr *src,
634     struct sockaddr *dst, struct proc *p, uint32_t ifscope,
635     sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
636     uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written)
637 {
638 	return tcp_usr_connectx_common(so, AF_INET, src, dst, p, ifscope, aid,
639 	           pcid, flags, arg, arglen, uio, bytes_written);
640 }
641 
642 static int
tcp6_usr_connect(struct socket * so,struct sockaddr * nam,struct proc * p)643 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
644 {
645 	int error = 0;
646 	struct inpcb *inp = sotoinpcb(so);
647 	struct tcpcb *tp;
648 	struct sockaddr_in6 *sin6p;
649 
650 	TCPDEBUG0;
651 	if (inp == NULL) {
652 		return EINVAL;
653 	} else if (inp->inp_state == INPCB_STATE_DEAD) {
654 		if (so->so_error) {
655 			error = so->so_error;
656 			so->so_error = 0;
657 			return error;
658 		} else {
659 			return EINVAL;
660 		}
661 	}
662 #if NECP
663 #if CONTENT_FILTER
664 	error = cfil_sock_attach(so, NULL, nam, CFS_CONNECTION_DIR_OUT);
665 	if (error != 0) {
666 		return error;
667 	}
668 #endif /* CONTENT_FILTER */
669 #if FLOW_DIVERT
670 	if (necp_socket_should_use_flow_divert(inp)) {
671 		error = flow_divert_pcb_init(so);
672 		if (error == 0) {
673 			error = flow_divert_connect_out(so, nam, p);
674 		}
675 		return error;
676 	}
677 #endif /* FLOW_DIVERT */
678 #endif /* NECP */
679 
680 	tp = intotcpcb(inp);
681 	TCPDEBUG1();
682 
683 	calculate_tcp_clock();
684 
685 	if (nam->sa_family != 0 && nam->sa_family != AF_INET6) {
686 		error = EAFNOSUPPORT;
687 		goto out;
688 	}
689 
690 	/*
691 	 * Must disallow TCP ``connections'' to multicast and broadcast addresses
692 	 */
693 	sin6p = (struct sockaddr_in6 *)(void *)nam;
694 	if (sin6p->sin6_family == AF_INET6 &&
695 	    IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
696 		error = EAFNOSUPPORT;
697 		goto out;
698 	}
699 
700 	if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
701 		struct sockaddr_in sin;
702 
703 		if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) {
704 			error = EINVAL;
705 			goto out;
706 		}
707 
708 		in6_sin6_2_sin(&sin, sin6p);
709 		/*
710 		 * Must disallow TCP ``connections'' to multicast and broadcast addresses.
711 		 */
712 		if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr)) ||
713 		    sin.sin_addr.s_addr == INADDR_BROADCAST) {
714 			error = EAFNOSUPPORT;
715 			goto out;
716 		}
717 		inp->inp_vflag |= INP_IPV4;
718 		inp->inp_vflag &= ~INP_IPV6;
719 		if ((error = tcp_connect(tp, (struct sockaddr *)&sin, p)) != 0) {
720 			TCP_LOG_CONNECT(tp, true, error);
721 			goto out;
722 		}
723 
724 		error = tcp_connect_complete(so);
725 		goto out;
726 	} else if (IN6_IS_ADDR_V4COMPAT(&sin6p->sin6_addr)) {
727 		/*
728 		 * Must disallow TCP ``connections'' to multicast and broadcast addresses.
729 		 */
730 		if (IN_MULTICAST(ntohl(sin6p->sin6_addr.s6_addr32[3])) ||
731 		    sin6p->sin6_addr.s6_addr32[3] == INADDR_BROADCAST) {
732 			error = EAFNOSUPPORT;
733 			goto out;
734 		}
735 	}
736 
737 	inp->inp_vflag &= ~INP_IPV4;
738 	inp->inp_vflag |= INP_IPV6;
739 	if ((error = tcp6_connect(tp, nam, p)) != 0) {
740 		TCP_LOG_CONNECT(tp, true, error);
741 		goto out;
742 	}
743 
744 	error = tcp_connect_complete(so);
745 
746 	TCP_LOG_CONNECT(tp, true, error);
747 
748 	COMMON_END(PRU_CONNECT);
749 }
750 
751 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)752 tcp6_usr_connectx(struct socket *so, struct sockaddr*src,
753     struct sockaddr *dst, struct proc *p, uint32_t ifscope,
754     sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
755     uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written)
756 {
757 	return tcp_usr_connectx_common(so, AF_INET6, src, dst, p, ifscope, aid,
758 	           pcid, flags, arg, arglen, uio, bytes_written);
759 }
760 
761 /*
762  * Initiate disconnect from peer.
763  * If connection never passed embryonic stage, just drop;
764  * else if don't need to let data drain, then can just drop anyways,
765  * else have to begin TCP shutdown process: mark socket disconnecting,
766  * drain unread data, state switch to reflect user close, and
767  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
768  * when peer sends FIN and acks ours.
769  *
770  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
771  */
772 static int
tcp_usr_disconnect(struct socket * so)773 tcp_usr_disconnect(struct socket *so)
774 {
775 	int error = 0;
776 	struct inpcb *inp = sotoinpcb(so);
777 	struct tcpcb *tp;
778 
779 	socket_lock_assert_owned(so);
780 	COMMON_START();
781 	/* In case we got disconnected from the peer */
782 	if (tp == NULL) {
783 		goto out;
784 	}
785 	tp = tcp_disconnect(tp);
786 	COMMON_END(PRU_DISCONNECT);
787 }
788 
789 /*
790  * User-protocol pru_disconnectx callback.
791  */
792 static int
tcp_usr_disconnectx(struct socket * so,sae_associd_t aid,sae_connid_t cid)793 tcp_usr_disconnectx(struct socket *so, sae_associd_t aid, sae_connid_t cid)
794 {
795 #pragma unused(cid)
796 	if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL) {
797 		return EINVAL;
798 	}
799 
800 	return tcp_usr_disconnect(so);
801 }
802 
803 /*
804  * Accept a connection.  Essentially all the work is
805  * done at higher levels; just return the address
806  * of the peer, storing through addr.
807  */
808 static int
tcp_usr_accept(struct socket * so,struct sockaddr ** nam)809 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
810 {
811 	int error = 0;
812 	struct inpcb *inp = sotoinpcb(so);
813 	struct tcpcb *tp = NULL;
814 	TCPDEBUG0;
815 
816 	in_getpeeraddr(so, nam);
817 
818 	if (so->so_state & SS_ISDISCONNECTED) {
819 		error = ECONNABORTED;
820 		goto out;
821 	}
822 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
823 		return EINVAL;
824 	}
825 #if NECP
826 	else if (necp_socket_should_use_flow_divert(inp)) {
827 		return EPROTOTYPE;
828 	}
829 
830 #endif /* NECP */
831 
832 	tp = intotcpcb(inp);
833 	TCPDEBUG1();
834 
835 	TCP_LOG_ACCEPT(tp, 0);
836 
837 	calculate_tcp_clock();
838 
839 	COMMON_END(PRU_ACCEPT);
840 }
841 
842 static int
tcp6_usr_accept(struct socket * so,struct sockaddr ** nam)843 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
844 {
845 	int error = 0;
846 	struct inpcb *inp = sotoinpcb(so);
847 	struct tcpcb *tp = NULL;
848 	TCPDEBUG0;
849 
850 	if (so->so_state & SS_ISDISCONNECTED) {
851 		error = ECONNABORTED;
852 		goto out;
853 	}
854 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
855 		return EINVAL;
856 	}
857 #if NECP
858 	else if (necp_socket_should_use_flow_divert(inp)) {
859 		return EPROTOTYPE;
860 	}
861 
862 #endif /* NECP */
863 
864 	tp = intotcpcb(inp);
865 	TCPDEBUG1();
866 
867 	TCP_LOG_ACCEPT(tp, 0);
868 
869 	calculate_tcp_clock();
870 
871 	in6_mapped_peeraddr(so, nam);
872 	COMMON_END(PRU_ACCEPT);
873 }
874 
875 /*
876  * Mark the connection as being incapable of further output.
877  *
878  * Returns:	0			Success
879  *		EINVAL [COMMON_START]
880  *	tcp_output:EADDRNOTAVAIL
881  *	tcp_output:ENOBUFS
882  *	tcp_output:EMSGSIZE
883  *	tcp_output:EHOSTUNREACH
884  *	tcp_output:ENETUNREACH
885  *	tcp_output:ENETDOWN
886  *	tcp_output:ENOMEM
887  *	tcp_output:EACCES
888  *	tcp_output:EMSGSIZE
889  *	tcp_output:ENOBUFS
890  *	tcp_output:???			[ignorable: mostly IPSEC/firewall/DLIL]
891  */
892 static int
tcp_usr_shutdown(struct socket * so)893 tcp_usr_shutdown(struct socket *so)
894 {
895 	int error = 0;
896 	struct inpcb *inp = sotoinpcb(so);
897 	struct tcpcb *tp;
898 
899 	TCPDEBUG0;
900 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
901 		return EINVAL;
902 	}
903 
904 	socantsendmore(so);
905 
906 	/*
907 	 * In case we got disconnected from the peer, or if this is
908 	 * a socket that is to be flow-diverted (but not yet).
909 	 */
910 	tp = intotcpcb(inp);
911 	TCPDEBUG1();
912 
913 	if (tp == NULL
914 #if NECP
915 	    || (necp_socket_should_use_flow_divert(inp))
916 #endif /* NECP */
917 	    ) {
918 		if (tp != NULL) {
919 			error = EPROTOTYPE;
920 		}
921 		goto out;
922 	}
923 
924 	calculate_tcp_clock();
925 
926 	tp = tcp_usrclosed(tp);
927 #if MPTCP
928 	/* A reset has been sent but socket exists, do not send FIN */
929 	if ((so->so_flags & SOF_MP_SUBFLOW) &&
930 	    (tp) && (tp->t_mpflags & TMPF_RESET)) {
931 		goto out;
932 	}
933 #endif
934 #if CONTENT_FILTER
935 	/* Don't send a FIN yet */
936 	if (tp && !(so->so_state & SS_ISDISCONNECTED) &&
937 	    cfil_sock_data_pending(&so->so_snd)) {
938 		goto out;
939 	}
940 #endif /* CONTENT_FILTER */
941 	if (tp) {
942 		error = tcp_output(tp);
943 	}
944 	COMMON_END(PRU_SHUTDOWN);
945 }
946 
947 /*
948  * After a receive, possibly send window update to peer.
949  */
950 static int
tcp_usr_rcvd(struct socket * so,int flags)951 tcp_usr_rcvd(struct socket *so, int flags)
952 {
953 	int error = 0;
954 	struct inpcb *inp = sotoinpcb(so);
955 	struct tcpcb *tp;
956 
957 	COMMON_START();
958 	/* In case we got disconnected from the peer */
959 	if (tp == NULL) {
960 		goto out;
961 	}
962 	tcp_sbrcv_trim(tp, &so->so_rcv);
963 
964 	if ((flags & MSG_WAITALL) && SEQ_LT(tp->last_ack_sent, tp->rcv_nxt)) {
965 		tp->t_flags |= TF_ACKNOW;
966 	}
967 
968 	/*
969 	 * This tcp_output is solely there to trigger window-updates.
970 	 * However, we really do not want these window-updates while we
971 	 * are still in SYN_SENT or SYN_RECEIVED.
972 	 */
973 	if (TCPS_HAVEESTABLISHED(tp->t_state)) {
974 		tcp_output(tp);
975 	}
976 
977 #if CONTENT_FILTER
978 	cfil_sock_buf_update(&so->so_rcv);
979 #endif /* CONTENT_FILTER */
980 
981 	COMMON_END(PRU_RCVD);
982 }
983 
984 __attribute__((noinline))
985 static int
tcp_send_implied_connect(struct tcpcb * tp,struct sockaddr * nam,struct proc * p,int isipv6)986 tcp_send_implied_connect(struct tcpcb *tp, struct sockaddr *nam, struct proc *p, int isipv6)
987 {
988 	int error = 0;
989 
990 	if (isipv6) {
991 		error = tcp6_connect(tp, nam, p);
992 	} else {
993 		error = tcp_connect(tp, nam, p);
994 	}
995 	if (error != 0) {
996 		goto out;
997 	}
998 	/*
999 	 * initialize window to default value, and
1000 	 * initialize maxseg/maxopd using peer's cached
1001 	 * MSS.
1002 	 */
1003 	tp->snd_wnd = TTCP_CLIENT_SND_WND;
1004 	tp->max_sndwnd = tp->snd_wnd;
1005 	tcp_mss(tp, -1, IFSCOPE_NONE);
1006 out:
1007 	TCP_LOG_CONNECT(tp, true, error);
1008 
1009 	return error;
1010 }
1011 
1012 __attribute__((noinline))
1013 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)1014 mpkl_tcp_send(struct socket *so, struct tcpcb *tp, uint32_t mpkl_seq, uint32_t mpkl_len,
1015     struct so_mpkl_send_info *mpkl_send_info)
1016 {
1017 	struct inpcb *inp = tp->t_inpcb;
1018 
1019 	if (inp == NULL) {
1020 		return;
1021 	}
1022 
1023 	if ((inp->inp_last_outifp != NULL &&
1024 	    (inp->inp_last_outifp->if_xflags & IFXF_MPK_LOG)) ||
1025 	    (inp->inp_boundifp != NULL &&
1026 	    (inp->inp_boundifp->if_xflags & IFXF_MPK_LOG))) {
1027 		MPKL_TCP_SEND(tcp_mpkl_log_object,
1028 		    mpkl_send_info->mpkl_proto,
1029 		    mpkl_send_info->mpkl_uuid,
1030 		    ntohs(inp->inp_lport),
1031 		    ntohs(inp->inp_fport),
1032 		    mpkl_seq,
1033 		    mpkl_len,
1034 		    so->last_pid,
1035 		    so->so_log_seqn++);
1036 	}
1037 }
1038 
1039 /*
1040  * Do a send by putting data in output queue and updating urgent
1041  * marker if URG set.  Possibly send more data.  Unlike the other
1042  * pru_*() routines, the mbuf chains are our responsibility.  We
1043  * must either enqueue them or free them.  The other pru_* routines
1044  * generally are caller-frees.
1045  *
1046  * Returns:	0			Success
1047  *		ECONNRESET
1048  *		EINVAL
1049  *		ENOBUFS
1050  *	tcp_connect:EADDRINUSE		Address in use
1051  *	tcp_connect:EADDRNOTAVAIL	Address not available.
1052  *	tcp_connect:EINVAL		Invalid argument
1053  *	tcp_connect:EAFNOSUPPORT	Address family not supported [notdef]
1054  *	tcp_connect:EACCES		Permission denied
1055  *	tcp_connect:EAGAIN		Resource unavailable, try again
1056  *	tcp_connect:EPERM		Operation not permitted
1057  *	tcp_output:EADDRNOTAVAIL
1058  *	tcp_output:ENOBUFS
1059  *	tcp_output:EMSGSIZE
1060  *	tcp_output:EHOSTUNREACH
1061  *	tcp_output:ENETUNREACH
1062  *	tcp_output:ENETDOWN
1063  *	tcp_output:ENOMEM
1064  *	tcp_output:EACCES
1065  *	tcp_output:EMSGSIZE
1066  *	tcp_output:ENOBUFS
1067  *	tcp_output:???			[ignorable: mostly IPSEC/firewall/DLIL]
1068  *	tcp6_connect:???		[IPV6 only]
1069  */
1070 static int
tcp_usr_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct proc * p)1071 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
1072     struct sockaddr *nam, struct mbuf *control, struct proc *p)
1073 {
1074 	int error = 0;
1075 	struct inpcb *inp = sotoinpcb(so);
1076 	struct tcpcb *tp;
1077 	uint32_t mpkl_len = 0; /* length of mbuf chain */
1078 	uint32_t mpkl_seq = 0; /* sequence number where new data is added */
1079 	struct so_mpkl_send_info mpkl_send_info = {};
1080 
1081 	int isipv6;
1082 	TCPDEBUG0;
1083 
1084 	if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD
1085 #if NECP
1086 	    || (necp_socket_should_use_flow_divert(inp))
1087 #endif /* NECP */
1088 	    ) {
1089 		/*
1090 		 * OOPS! we lost a race, the TCP session got reset after
1091 		 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a
1092 		 * network interrupt in the non-splnet() section of sosend().
1093 		 */
1094 		if (m != NULL) {
1095 			m_freem(m);
1096 		}
1097 		if (control != NULL) {
1098 			m_freem(control);
1099 			control = NULL;
1100 		}
1101 
1102 		if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) {
1103 			error = ECONNRESET;     /* XXX EPIPE? */
1104 		} else {
1105 			error = EPROTOTYPE;
1106 		}
1107 		tp = NULL;
1108 		TCPDEBUG1();
1109 		goto out;
1110 	}
1111 	isipv6 = nam && nam->sa_family == AF_INET6;
1112 	tp = intotcpcb(inp);
1113 	TCPDEBUG1();
1114 
1115 	calculate_tcp_clock();
1116 
1117 	if (net_mpklog_enabled) {
1118 		mpkl_seq = tp->snd_una + so->so_snd.sb_cc;
1119 		if (m) {
1120 			mpkl_len = m_length(m);
1121 		}
1122 		if (so->so_flags1 & SOF1_MPKL_SEND_INFO) {
1123 			uuid_copy(mpkl_send_info.mpkl_uuid, so->so_mpkl_send_uuid);
1124 			mpkl_send_info.mpkl_proto = so->so_mpkl_send_proto;
1125 		}
1126 	}
1127 
1128 	if (control != NULL) {
1129 		if (control->m_len > 0 && net_mpklog_enabled) {
1130 			error = tcp_get_mpkl_send_info(control, &mpkl_send_info);
1131 			/*
1132 			 * Intepretation of the returned code:
1133 			 *  0: client wants us to use value passed in SCM_MPKL_SEND_INFO
1134 			 *  1: SCM_MPKL_SEND_INFO was not present
1135 			 *  other: failure
1136 			 */
1137 			if (error != 0 && error != ENOMSG) {
1138 				m_freem(control);
1139 				if (m != NULL) {
1140 					m_freem(m);
1141 				}
1142 				control = NULL;
1143 				m = NULL;
1144 				goto out;
1145 			}
1146 		}
1147 		/*
1148 		 * Silently drop unsupported ancillary data messages
1149 		 */
1150 		m_freem(control);
1151 		control = NULL;
1152 	}
1153 
1154 	/* MPTCP sublow socket buffers must not be compressed */
1155 	VERIFY(!(so->so_flags & SOF_MP_SUBFLOW) ||
1156 	    (so->so_snd.sb_flags & SB_NOCOMPRESS));
1157 
1158 	if (!(flags & PRUS_OOB) || (so->so_flags1 & SOF1_PRECONNECT_DATA)) {
1159 		sbappendstream(&so->so_snd, m);
1160 
1161 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1162 			/*
1163 			 * Do implied connect if not yet connected,
1164 			 */
1165 			error = tcp_send_implied_connect(tp, nam, p, isipv6);
1166 			if (error != 0) {
1167 				goto out;
1168 			}
1169 			/* The sequence number of the data is past the SYN */
1170 			mpkl_seq = tp->iss + 1;
1171 		}
1172 
1173 		if (flags & PRUS_EOF) {
1174 			/*
1175 			 * Close the send side of the connection after
1176 			 * the data is sent.
1177 			 */
1178 			socantsendmore(so);
1179 			tp = tcp_usrclosed(tp);
1180 		}
1181 		if (tp != NULL) {
1182 			if (flags & PRUS_MORETOCOME) {
1183 				tp->t_flags |= TF_MORETOCOME;
1184 			}
1185 			error = tcp_output(tp);
1186 			if (flags & PRUS_MORETOCOME) {
1187 				tp->t_flags &= ~TF_MORETOCOME;
1188 			}
1189 		}
1190 	} else {
1191 		if (sbspace(&so->so_snd) == 0) {
1192 			/* if no space is left in sockbuf,
1193 			 * do not try to squeeze in OOB traffic */
1194 			m_freem(m);
1195 			error = ENOBUFS;
1196 			goto out;
1197 		}
1198 		/*
1199 		 * According to RFC961 (Assigned Protocols),
1200 		 * the urgent pointer points to the last octet
1201 		 * of urgent data.  We continue, however,
1202 		 * to consider it to indicate the first octet
1203 		 * of data past the urgent section.
1204 		 * Otherwise, snd_up should be one lower.
1205 		 */
1206 		sbappendstream(&so->so_snd, m);
1207 		if (nam && tp->t_state < TCPS_SYN_SENT) {
1208 			/*
1209 			 * Do implied connect if not yet connected,
1210 			 * initialize window to default value, and
1211 			 * initialize maxseg/maxopd using peer's cached
1212 			 * MSS.
1213 			 */
1214 			error = tcp_send_implied_connect(tp, nam, p, isipv6);
1215 			if (error != 0) {
1216 				goto out;
1217 			}
1218 		}
1219 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
1220 		tp->t_flagsext |= TF_FORCE;
1221 		error = tcp_output(tp);
1222 		tp->t_flagsext &= ~TF_FORCE;
1223 	}
1224 
1225 	if (net_mpklog_enabled) {
1226 		mpkl_tcp_send(so, tp, mpkl_seq, mpkl_len, &mpkl_send_info);
1227 	}
1228 
1229 	/*
1230 	 * We wait for the socket to successfully connect before returning.
1231 	 * This allows us to signal a timeout to the application.
1232 	 */
1233 	if (so->so_state & SS_ISCONNECTING) {
1234 		if (so->so_state & SS_NBIO) {
1235 			error = EWOULDBLOCK;
1236 		} else {
1237 			error = sbwait(&so->so_snd);
1238 		}
1239 	}
1240 
1241 	COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
1242 	    ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1243 }
1244 
1245 /*
1246  * Abort the TCP.
1247  */
1248 static int
tcp_usr_abort(struct socket * so)1249 tcp_usr_abort(struct socket *so)
1250 {
1251 	int error = 0;
1252 	struct inpcb *inp = sotoinpcb(so);
1253 	struct tcpcb *tp;
1254 
1255 	COMMON_START();
1256 	/* In case we got disconnected from the peer */
1257 	if (tp == NULL) {
1258 		goto out;
1259 	}
1260 	tp = tcp_drop(tp, ECONNABORTED);
1261 	VERIFY(so->so_usecount > 0);
1262 	so->so_usecount--;
1263 	COMMON_END(PRU_ABORT);
1264 }
1265 
1266 /*
1267  * Receive out-of-band data.
1268  *
1269  * Returns:	0			Success
1270  *		EINVAL [COMMON_START]
1271  *		EINVAL
1272  *		EWOULDBLOCK
1273  */
1274 static int
tcp_usr_rcvoob(struct socket * so,struct mbuf * m,int flags)1275 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1276 {
1277 	int error = 0;
1278 	struct inpcb *inp = sotoinpcb(so);
1279 	struct tcpcb *tp;
1280 
1281 	COMMON_START();
1282 	if ((so->so_oobmark == 0 &&
1283 	    (so->so_state & SS_RCVATMARK) == 0) ||
1284 	    so->so_options & SO_OOBINLINE ||
1285 	    tp->t_oobflags & TCPOOB_HADDATA) {
1286 		error = EINVAL;
1287 		goto out;
1288 	}
1289 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1290 		error = EWOULDBLOCK;
1291 		goto out;
1292 	}
1293 	m->m_len = 1;
1294 	*mtod(m, caddr_t) = tp->t_iobc;
1295 	so->so_state &= ~SS_RCVATMARK;
1296 	if ((flags & MSG_PEEK) == 0) {
1297 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1298 	}
1299 	COMMON_END(PRU_RCVOOB);
1300 }
1301 
1302 static int
tcp_usr_preconnect(struct socket * so)1303 tcp_usr_preconnect(struct socket *so)
1304 {
1305 	struct inpcb *inp = sotoinpcb(so);
1306 	int error = 0;
1307 
1308 #if NECP
1309 	if (necp_socket_should_use_flow_divert(inp)) {
1310 		/* May happen, if in tcp_usr_connect we did not had a chance
1311 		 * to set the usrreqs (due to some error). So, let's get out
1312 		 * of here.
1313 		 */
1314 		goto out;
1315 	}
1316 #endif /* NECP */
1317 
1318 	error = tcp_output(sototcpcb(so));
1319 
1320 	soclearfastopen(so);
1321 
1322 	COMMON_END(PRU_PRECONNECT);
1323 }
1324 
1325 /* xxx - should be const */
1326 struct pr_usrreqs tcp_usrreqs = {
1327 	.pru_abort =            tcp_usr_abort,
1328 	.pru_accept =           tcp_usr_accept,
1329 	.pru_attach =           tcp_usr_attach,
1330 	.pru_bind =             tcp_usr_bind,
1331 	.pru_connect =          tcp_usr_connect,
1332 	.pru_connectx =         tcp_usr_connectx,
1333 	.pru_control =          in_control,
1334 	.pru_detach =           tcp_usr_detach,
1335 	.pru_disconnect =       tcp_usr_disconnect,
1336 	.pru_disconnectx =      tcp_usr_disconnectx,
1337 	.pru_listen =           tcp_usr_listen,
1338 	.pru_peeraddr =         in_getpeeraddr,
1339 	.pru_rcvd =             tcp_usr_rcvd,
1340 	.pru_rcvoob =           tcp_usr_rcvoob,
1341 	.pru_send =             tcp_usr_send,
1342 	.pru_shutdown =         tcp_usr_shutdown,
1343 	.pru_sockaddr =         in_getsockaddr,
1344 	.pru_sosend =           sosend,
1345 	.pru_soreceive =        soreceive,
1346 	.pru_preconnect =       tcp_usr_preconnect,
1347 };
1348 
1349 struct pr_usrreqs tcp6_usrreqs = {
1350 	.pru_abort =            tcp_usr_abort,
1351 	.pru_accept =           tcp6_usr_accept,
1352 	.pru_attach =           tcp_usr_attach,
1353 	.pru_bind =             tcp6_usr_bind,
1354 	.pru_connect =          tcp6_usr_connect,
1355 	.pru_connectx =         tcp6_usr_connectx,
1356 	.pru_control =          in6_control,
1357 	.pru_detach =           tcp_usr_detach,
1358 	.pru_disconnect =       tcp_usr_disconnect,
1359 	.pru_disconnectx =      tcp_usr_disconnectx,
1360 	.pru_listen =           tcp6_usr_listen,
1361 	.pru_peeraddr =         in6_mapped_peeraddr,
1362 	.pru_rcvd =             tcp_usr_rcvd,
1363 	.pru_rcvoob =           tcp_usr_rcvoob,
1364 	.pru_send =             tcp_usr_send,
1365 	.pru_shutdown =         tcp_usr_shutdown,
1366 	.pru_sockaddr =         in6_mapped_sockaddr,
1367 	.pru_sosend =           sosend,
1368 	.pru_soreceive =        soreceive,
1369 	.pru_preconnect =       tcp_usr_preconnect,
1370 };
1371 
1372 /*
1373  * Common subroutine to open a TCP connection to remote host specified
1374  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
1375  * port number if needed.  Call in_pcbladdr to do the routing and to choose
1376  * a local host address (interface).  If there is an existing incarnation
1377  * of the same connection in TIME-WAIT state and if the remote host was
1378  * sending CC options and if the connection duration was < MSL, then
1379  * truncate the previous TIME-WAIT state and proceed.
1380  * Initialize connection parameters and enter SYN-SENT state.
1381  *
1382  * Returns:	0			Success
1383  *		EADDRINUSE
1384  *		EINVAL
1385  *	in_pcbbind:EADDRNOTAVAIL	Address not available.
1386  *	in_pcbbind:EINVAL		Invalid argument
1387  *	in_pcbbind:EAFNOSUPPORT		Address family not supported [notdef]
1388  *	in_pcbbind:EACCES		Permission denied
1389  *	in_pcbbind:EADDRINUSE		Address in use
1390  *	in_pcbbind:EAGAIN		Resource unavailable, try again
1391  *	in_pcbbind:EPERM		Operation not permitted
1392  *	in_pcbladdr:EINVAL		Invalid argument
1393  *	in_pcbladdr:EAFNOSUPPORT	Address family not supported
1394  *	in_pcbladdr:EADDRNOTAVAIL	Address not available
1395  */
1396 static int
tcp_connect(struct tcpcb * tp,struct sockaddr * nam,struct proc * p)1397 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct proc *p)
1398 {
1399 	struct inpcb *inp = tp->t_inpcb, *oinp;
1400 	struct socket *so = inp->inp_socket;
1401 	struct tcpcb *otp;
1402 	struct sockaddr_in *sin = (struct sockaddr_in *)(void *)nam;
1403 	struct in_addr laddr;
1404 	int error = 0;
1405 	struct ifnet *outif = NULL;
1406 
1407 	if (inp->inp_lport == 0) {
1408 		error = in_pcbbind(inp, NULL, p);
1409 		if (error) {
1410 			goto done;
1411 		}
1412 	}
1413 
1414 	/*
1415 	 * Cannot simply call in_pcbconnect, because there might be an
1416 	 * earlier incarnation of this same connection still in
1417 	 * TIME_WAIT state, creating an ADDRINUSE error.
1418 	 */
1419 	error = in_pcbladdr(inp, nam, &laddr, IFSCOPE_NONE, &outif, 0);
1420 	if (error) {
1421 		goto done;
1422 	}
1423 
1424 	socket_unlock(inp->inp_socket, 0);
1425 	oinp = in_pcblookup_hash(inp->inp_pcbinfo,
1426 	    sin->sin_addr, sin->sin_port,
1427 	    inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr : laddr,
1428 	    inp->inp_lport, 0, NULL);
1429 
1430 	socket_lock(inp->inp_socket, 0);
1431 	if (oinp) {
1432 		if (oinp != inp) { /* 4143933: avoid deadlock if inp == oinp */
1433 			socket_lock(oinp->inp_socket, 1);
1434 		}
1435 		if (in_pcb_checkstate(oinp, WNT_RELEASE, 1) == WNT_STOPUSING) {
1436 			if (oinp != inp) {
1437 				socket_unlock(oinp->inp_socket, 1);
1438 			}
1439 			goto skip_oinp;
1440 		}
1441 
1442 		if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
1443 		    otp->t_state == TCPS_TIME_WAIT &&
1444 		    ((int)(tcp_now - otp->t_starttime)) < tcp_msl &&
1445 		    (otp->t_flags & TF_RCVD_CC)) {
1446 			otp = tcp_close(otp);
1447 		} else {
1448 			printf("tcp_connect: inp=0x%llx err=EADDRINUSE\n",
1449 			    (uint64_t)VM_KERNEL_ADDRPERM(inp));
1450 			if (oinp != inp) {
1451 				socket_unlock(oinp->inp_socket, 1);
1452 			}
1453 			error = EADDRINUSE;
1454 			goto done;
1455 		}
1456 		if (oinp != inp) {
1457 			socket_unlock(oinp->inp_socket, 1);
1458 		}
1459 	}
1460 skip_oinp:
1461 	if ((inp->inp_laddr.s_addr == INADDR_ANY ? laddr.s_addr :
1462 	    inp->inp_laddr.s_addr) == sin->sin_addr.s_addr &&
1463 	    inp->inp_lport == sin->sin_port) {
1464 		error = EINVAL;
1465 		goto done;
1466 	}
1467 #if SKYWALK
1468 	if (!NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1469 		error = netns_reserve_in(&inp->inp_netns_token,
1470 		    inp->inp_laddr.s_addr != INADDR_ANY ?
1471 		    inp->inp_laddr : laddr,
1472 		    IPPROTO_TCP, inp->inp_lport, NETNS_BSD, NULL);
1473 		if (error) {
1474 			goto done;
1475 		}
1476 	}
1477 #endif /* SKYWALK */
1478 	if (!lck_rw_try_lock_exclusive(&inp->inp_pcbinfo->ipi_lock)) {
1479 		/*lock inversion issue, mostly with udp multicast packets */
1480 		socket_unlock(inp->inp_socket, 0);
1481 		lck_rw_lock_exclusive(&inp->inp_pcbinfo->ipi_lock);
1482 		socket_lock(inp->inp_socket, 0);
1483 	}
1484 	if (inp->inp_laddr.s_addr == INADDR_ANY) {
1485 		inp->inp_laddr = laddr;
1486 		/* no reference needed */
1487 		inp->inp_last_outifp = outif;
1488 #if SKYWALK
1489 		if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1490 			netns_set_ifnet(&inp->inp_netns_token, inp->inp_last_outifp);
1491 		}
1492 #endif /* SKYWALK */
1493 
1494 		inp->inp_flags |= INP_INADDR_ANY;
1495 	}
1496 	inp->inp_faddr = sin->sin_addr;
1497 	inp->inp_fport = sin->sin_port;
1498 	in_pcbrehash(inp);
1499 	lck_rw_done(&inp->inp_pcbinfo->ipi_lock);
1500 
1501 	if (inp->inp_flowhash == 0) {
1502 		inp->inp_flowhash = inp_calc_flowhash(inp);
1503 	}
1504 
1505 	tcp_set_max_rwinscale(tp, so);
1506 
1507 	soisconnecting(so);
1508 	tcpstat.tcps_connattempt++;
1509 	tp->t_state = TCPS_SYN_SENT;
1510 	tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp, TCP_CONN_KEEPINIT(tp));
1511 	tp->iss = tcp_new_isn(tp);
1512 	tcp_sendseqinit(tp);
1513 	tp->t_connect_time = tcp_now;
1514 	if (nstat_collect) {
1515 		nstat_route_connect_attempt(inp->inp_route.ro_rt);
1516 	}
1517 
1518 	tcp_add_fsw_flow(tp, outif);
1519 
1520 done:
1521 	if (outif != NULL) {
1522 		ifnet_release(outif);
1523 	}
1524 
1525 	return error;
1526 }
1527 
1528 static int
tcp6_connect(struct tcpcb * tp,struct sockaddr * nam,struct proc * p)1529 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct proc *p)
1530 {
1531 	struct inpcb *inp = tp->t_inpcb, *oinp;
1532 	struct socket *so = inp->inp_socket;
1533 	struct tcpcb *otp;
1534 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)(void *)nam;
1535 	struct in6_addr addr6;
1536 	int error = 0;
1537 	struct ifnet *outif = NULL;
1538 
1539 	if (inp->inp_lport == 0) {
1540 		error = in6_pcbbind(inp, NULL, p);
1541 		if (error) {
1542 			goto done;
1543 		}
1544 	}
1545 
1546 	/*
1547 	 * Cannot simply call in_pcbconnect, because there might be an
1548 	 * earlier incarnation of this same connection still in
1549 	 * TIME_WAIT state, creating an ADDRINUSE error.
1550 	 *
1551 	 * in6_pcbladdr() might return an ifp with its reference held
1552 	 * even in the error case, so make sure that it's released
1553 	 * whenever it's non-NULL.
1554 	 */
1555 	error = in6_pcbladdr(inp, nam, &addr6, &outif);
1556 	if (error) {
1557 		goto done;
1558 	}
1559 	socket_unlock(inp->inp_socket, 0);
1560 
1561 	uint32_t lifscope = IFSCOPE_NONE;
1562 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1563 		lifscope = inp->inp_lifscope;
1564 	} else if (sin6->sin6_scope_id != IFSCOPE_NONE) {
1565 		lifscope = sin6->sin6_scope_id;
1566 	} else if (outif != NULL) {
1567 		lifscope = outif->if_index;
1568 	}
1569 	oinp = in6_pcblookup_hash(inp->inp_pcbinfo,
1570 	    &sin6->sin6_addr, sin6->sin6_port, sin6->sin6_scope_id,
1571 	    IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
1572 	    ? &addr6
1573 	    : &inp->in6p_laddr,
1574 	    inp->inp_lport, lifscope, 0, NULL);
1575 	socket_lock(inp->inp_socket, 0);
1576 	if (oinp) {
1577 		if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
1578 		    otp->t_state == TCPS_TIME_WAIT &&
1579 		    ((int)(tcp_now - otp->t_starttime)) < tcp_msl &&
1580 		    (otp->t_flags & TF_RCVD_CC)) {
1581 			otp = tcp_close(otp);
1582 		} else {
1583 			error = EADDRINUSE;
1584 			goto done;
1585 		}
1586 	}
1587 #if SKYWALK
1588 	if (!NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1589 		error = netns_reserve_in6(&inp->inp_netns_token,
1590 		    IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ?
1591 		    addr6 : inp->in6p_laddr,
1592 		    IPPROTO_TCP, inp->inp_lport, NETNS_BSD, NULL);
1593 		if (error) {
1594 			goto done;
1595 		}
1596 	}
1597 #endif /* SKYWALK */
1598 	if (!lck_rw_try_lock_exclusive(&inp->inp_pcbinfo->ipi_lock)) {
1599 		/*lock inversion issue, mostly with udp multicast packets */
1600 		socket_unlock(inp->inp_socket, 0);
1601 		lck_rw_lock_exclusive(&inp->inp_pcbinfo->ipi_lock);
1602 		socket_lock(inp->inp_socket, 0);
1603 	}
1604 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1605 		inp->in6p_laddr = addr6;
1606 		inp->in6p_last_outifp = outif;  /* no reference needed */
1607 		inp->inp_lifscope = lifscope;
1608 		in6_verify_ifscope(&inp->in6p_laddr, inp->inp_lifscope);
1609 #if SKYWALK
1610 		if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
1611 			netns_set_ifnet(&inp->inp_netns_token, inp->in6p_last_outifp);
1612 		}
1613 #endif /* SKYWALK */
1614 		inp->in6p_flags |= INP_IN6ADDR_ANY;
1615 	}
1616 	inp->in6p_faddr = sin6->sin6_addr;
1617 	inp->inp_fport = sin6->sin6_port;
1618 	inp->inp_fifscope = sin6->sin6_scope_id;
1619 	in6_verify_ifscope(&inp->in6p_faddr, inp->inp_fifscope);
1620 	if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != 0) {
1621 		inp->inp_flow = sin6->sin6_flowinfo;
1622 	}
1623 	in_pcbrehash(inp);
1624 	lck_rw_done(&inp->inp_pcbinfo->ipi_lock);
1625 
1626 	if (inp->inp_flowhash == 0) {
1627 		inp->inp_flowhash = inp_calc_flowhash(inp);
1628 	}
1629 	/* update flowinfo - RFC 6437 */
1630 	if (inp->inp_flow == 0 && inp->in6p_flags & IN6P_AUTOFLOWLABEL) {
1631 		inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
1632 		inp->inp_flow |=
1633 		    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
1634 	}
1635 
1636 	tcp_set_max_rwinscale(tp, so);
1637 
1638 	soisconnecting(so);
1639 	tcpstat.tcps_connattempt++;
1640 	tp->t_state = TCPS_SYN_SENT;
1641 	tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
1642 	    TCP_CONN_KEEPINIT(tp));
1643 	tp->iss = tcp_new_isn(tp);
1644 	tcp_sendseqinit(tp);
1645 	tp->t_connect_time = tcp_now;
1646 	if (nstat_collect) {
1647 		nstat_route_connect_attempt(inp->inp_route.ro_rt);
1648 	}
1649 
1650 	tcp_add_fsw_flow(tp, outif);
1651 
1652 done:
1653 	if (outif != NULL) {
1654 		ifnet_release(outif);
1655 	}
1656 
1657 	return error;
1658 }
1659 
1660 /*
1661  * Export TCP internal state information via a struct tcp_info
1662  */
1663 void
tcp_fill_info(struct tcpcb * tp,struct tcp_info * ti)1664 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1665 {
1666 	struct inpcb *inp = tp->t_inpcb;
1667 
1668 	bzero(ti, sizeof(*ti));
1669 
1670 	ti->tcpi_state = (uint8_t)tp->t_state;
1671 	ti->tcpi_flowhash = inp != NULL ? inp->inp_flowhash: 0;
1672 
1673 	if (TSTMP_SUPPORTED(tp)) {
1674 		ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1675 	}
1676 	if (SACK_ENABLED(tp)) {
1677 		ti->tcpi_options |= TCPI_OPT_SACK;
1678 	}
1679 	if (TCP_WINDOW_SCALE_ENABLED(tp)) {
1680 		ti->tcpi_options |= TCPI_OPT_WSCALE;
1681 		ti->tcpi_snd_wscale = tp->snd_scale;
1682 		ti->tcpi_rcv_wscale = tp->rcv_scale;
1683 	}
1684 	if (TCP_ECN_ENABLED(tp)) {
1685 		ti->tcpi_options |= TCPI_OPT_ECN;
1686 	}
1687 
1688 	/* Are we in retranmission episode */
1689 	if (IN_FASTRECOVERY(tp) || tp->t_rxtshift > 0) {
1690 		ti->tcpi_flags |= TCPI_FLAG_LOSSRECOVERY;
1691 	}
1692 
1693 	if (tp->t_flags & TF_STREAMING_ON) {
1694 		ti->tcpi_flags |= TCPI_FLAG_STREAMING_ON;
1695 	}
1696 
1697 	ti->tcpi_rto = tp->t_timer[TCPT_REXMT] ? tp->t_rxtcur : 0;
1698 	ti->tcpi_snd_mss = tp->t_maxseg;
1699 	ti->tcpi_rcv_mss = tp->t_maxseg;
1700 
1701 	ti->tcpi_rttcur = tp->t_rttcur;
1702 	ti->tcpi_srtt = tp->t_srtt >> TCP_RTT_SHIFT;
1703 	ti->tcpi_rcv_srtt = tp->rcv_srtt >> TCP_RTT_SHIFT;
1704 	ti->tcpi_rttvar = tp->t_rttvar >> TCP_RTTVAR_SHIFT;
1705 	ti->tcpi_rttbest = tp->t_rttbest >> TCP_RTT_SHIFT;
1706 
1707 	ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1708 	ti->tcpi_snd_cwnd = tp->snd_cwnd;
1709 	if (inp != NULL && inp->inp_socket != NULL) {
1710 		ti->tcpi_snd_sbbytes = inp->inp_socket->so_snd.sb_cc;
1711 	}
1712 
1713 	ti->tcpi_rcv_space = tp->rcv_adv > tp->rcv_nxt ?
1714 	    tp->rcv_adv - tp->rcv_nxt : 0;
1715 
1716 	ti->tcpi_snd_wnd = tp->snd_wnd;
1717 	ti->tcpi_snd_nxt = tp->snd_nxt;
1718 	ti->tcpi_rcv_nxt = tp->rcv_nxt;
1719 
1720 	/* convert bytes/msec to bits/sec */
1721 	if ((tp->t_flagsext & TF_MEASURESNDBW) != 0 &&
1722 	    tp->t_bwmeas != NULL) {
1723 		ti->tcpi_snd_bw = (tp->t_bwmeas->bw_sndbw * 8000);
1724 	}
1725 
1726 	ti->tcpi_txpackets = inp != NULL ? inp->inp_stat->txpackets : 0;
1727 	ti->tcpi_txbytes = inp != NULL ? inp->inp_stat->txbytes : 0;
1728 	ti->tcpi_txretransmitbytes = tp->t_stat.txretransmitbytes;
1729 	ti->tcpi_txretransmitpackets = tp->t_stat.rxmitpkts;
1730 	ti->tcpi_txunacked = tp->snd_max - tp->snd_una;
1731 
1732 	ti->tcpi_rxpackets = inp != NULL ? inp->inp_stat->rxpackets : 0;
1733 	ti->tcpi_rxbytes = inp != NULL ? inp->inp_stat->rxbytes : 0;
1734 	ti->tcpi_rxduplicatebytes = tp->t_stat.rxduplicatebytes;
1735 	ti->tcpi_rxoutoforderbytes = tp->t_stat.rxoutoforderbytes;
1736 
1737 	if (tp->t_state > TCPS_LISTEN) {
1738 		ti->tcpi_synrexmits = (uint8_t)tp->t_stat.rxmitsyns;
1739 	}
1740 	if (inp != NULL) {
1741 		ti->tcpi_cell_rxpackets = inp->inp_cstat->rxpackets;
1742 		ti->tcpi_cell_rxbytes = inp->inp_cstat->rxbytes;
1743 		ti->tcpi_cell_txpackets = inp->inp_cstat->txpackets;
1744 		ti->tcpi_cell_txbytes = inp->inp_cstat->txbytes;
1745 
1746 		ti->tcpi_wifi_rxpackets = inp->inp_wstat->rxpackets;
1747 		ti->tcpi_wifi_rxbytes = inp->inp_wstat->rxbytes;
1748 		ti->tcpi_wifi_txpackets = inp->inp_wstat->txpackets;
1749 		ti->tcpi_wifi_txbytes = inp->inp_wstat->txbytes;
1750 
1751 		ti->tcpi_wired_rxpackets = inp->inp_Wstat->rxpackets;
1752 		ti->tcpi_wired_rxbytes = inp->inp_Wstat->rxbytes;
1753 		ti->tcpi_wired_txpackets = inp->inp_Wstat->txpackets;
1754 		ti->tcpi_wired_txbytes = inp->inp_Wstat->txbytes;
1755 	}
1756 	tcp_get_connectivity_status(tp, &ti->tcpi_connstatus);
1757 
1758 	ti->tcpi_tfo_syn_data_rcv = !!(tp->t_tfo_stats & TFO_S_SYNDATA_RCV);
1759 	ti->tcpi_tfo_cookie_req_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIEREQ_RECV);
1760 	ti->tcpi_tfo_cookie_sent = !!(tp->t_tfo_stats & TFO_S_COOKIE_SENT);
1761 	ti->tcpi_tfo_cookie_invalid = !!(tp->t_tfo_stats & TFO_S_COOKIE_INVALID);
1762 
1763 	ti->tcpi_tfo_cookie_req = !!(tp->t_tfo_stats & TFO_S_COOKIE_REQ);
1764 	ti->tcpi_tfo_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIE_RCV);
1765 	ti->tcpi_tfo_syn_data_sent = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_SENT);
1766 	ti->tcpi_tfo_syn_data_acked = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_ACKED);
1767 	ti->tcpi_tfo_syn_loss = !!(tp->t_tfo_stats & TFO_S_SYN_LOSS);
1768 	ti->tcpi_tfo_cookie_wrong = !!(tp->t_tfo_stats & TFO_S_COOKIE_WRONG);
1769 	ti->tcpi_tfo_no_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_NO_COOKIE_RCV);
1770 	ti->tcpi_tfo_heuristics_disable = !!(tp->t_tfo_stats & TFO_S_HEURISTICS_DISABLE);
1771 	ti->tcpi_tfo_send_blackhole = !!(tp->t_tfo_stats & TFO_S_SEND_BLACKHOLE);
1772 	ti->tcpi_tfo_recv_blackhole = !!(tp->t_tfo_stats & TFO_S_RECV_BLACKHOLE);
1773 	ti->tcpi_tfo_onebyte_proxy = !!(tp->t_tfo_stats & TFO_S_ONE_BYTE_PROXY);
1774 
1775 	ti->tcpi_ecn_client_setup = !!(tp->ecn_flags & TE_SETUPSENT);
1776 	ti->tcpi_ecn_server_setup = !!(tp->ecn_flags & TE_SETUPRECEIVED);
1777 	ti->tcpi_ecn_success = (tp->ecn_flags & TE_ECN_ON) == TE_ECN_ON ? 1 : 0;
1778 	ti->tcpi_ecn_lost_syn = !!(tp->ecn_flags & TE_LOST_SYN);
1779 	ti->tcpi_ecn_lost_synack = !!(tp->ecn_flags & TE_LOST_SYNACK);
1780 
1781 	ti->tcpi_local_peer = !!(tp->t_flags & TF_LOCAL);
1782 
1783 	if (inp != NULL && inp->inp_last_outifp != NULL) {
1784 		ti->tcpi_last_outif = inp->inp_last_outifp->if_index;
1785 
1786 		if (IFNET_IS_CELLULAR(inp->inp_last_outifp)) {
1787 			ti->tcpi_if_cell = 1;
1788 		}
1789 		if (IFNET_IS_WIFI(inp->inp_last_outifp)) {
1790 			ti->tcpi_if_wifi = 1;
1791 		}
1792 		if (IFNET_IS_WIRED(inp->inp_last_outifp)) {
1793 			ti->tcpi_if_wired = 1;
1794 		}
1795 		if (IFNET_IS_WIFI_INFRA(inp->inp_last_outifp)) {
1796 			ti->tcpi_if_wifi_infra = 1;
1797 		}
1798 		if (inp->inp_last_outifp->if_eflags & IFEF_AWDL) {
1799 			ti->tcpi_if_wifi_awdl = 1;
1800 		}
1801 	}
1802 	if (tp->tcp_cc_index == TCP_CC_ALGO_BACKGROUND_INDEX) {
1803 		ti->tcpi_snd_background = 1;
1804 	}
1805 	if (tcp_recv_bg == 1 || (inp != NULL && inp->inp_socket != NULL &&
1806 	    IS_TCP_RECV_BG(inp->inp_socket))) {
1807 		ti->tcpi_rcv_background = 1;
1808 	}
1809 
1810 	ti->tcpi_ecn_recv_ce = tp->t_ecn_recv_ce;
1811 	ti->tcpi_ecn_recv_cwr = tp->t_ecn_recv_cwr;
1812 
1813 	ti->tcpi_rcvoopack = tp->t_rcvoopack;
1814 	ti->tcpi_pawsdrop = tp->t_pawsdrop;
1815 	ti->tcpi_sack_recovery_episode = tp->t_sack_recovery_episode;
1816 	ti->tcpi_reordered_pkts = tp->t_reordered_pkts;
1817 	ti->tcpi_dsack_sent = tp->t_dsack_sent;
1818 	ti->tcpi_dsack_recvd = tp->t_dsack_recvd;
1819 }
1820 
1821 __private_extern__ errno_t
tcp_fill_info_for_info_tuple(struct info_tuple * itpl,struct tcp_info * ti)1822 tcp_fill_info_for_info_tuple(struct info_tuple *itpl, struct tcp_info *ti)
1823 {
1824 	struct inpcbinfo *pcbinfo = NULL;
1825 	struct inpcb *inp = NULL;
1826 	struct socket *so;
1827 	struct tcpcb *tp;
1828 
1829 	if (itpl->itpl_proto == IPPROTO_TCP) {
1830 		pcbinfo = &tcbinfo;
1831 	} else {
1832 		return EINVAL;
1833 	}
1834 
1835 	if (itpl->itpl_local_sa.sa_family == AF_INET &&
1836 	    itpl->itpl_remote_sa.sa_family == AF_INET) {
1837 		inp = in_pcblookup_hash(pcbinfo,
1838 		    itpl->itpl_remote_sin.sin_addr,
1839 		    itpl->itpl_remote_sin.sin_port,
1840 		    itpl->itpl_local_sin.sin_addr,
1841 		    itpl->itpl_local_sin.sin_port,
1842 		    0, NULL);
1843 	} else if (itpl->itpl_local_sa.sa_family == AF_INET6 &&
1844 	    itpl->itpl_remote_sa.sa_family == AF_INET6) {
1845 		struct in6_addr ina6_local;
1846 		struct in6_addr ina6_remote;
1847 
1848 		ina6_local = itpl->itpl_local_sin6.sin6_addr;
1849 		if (in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&ina6_local) &&
1850 		    itpl->itpl_local_sin6.sin6_scope_id) {
1851 			ina6_local.s6_addr16[1] = htons((uint16_t)itpl->itpl_local_sin6.sin6_scope_id);
1852 		}
1853 
1854 		ina6_remote = itpl->itpl_remote_sin6.sin6_addr;
1855 		if (in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&ina6_remote) &&
1856 		    itpl->itpl_remote_sin6.sin6_scope_id) {
1857 			ina6_remote.s6_addr16[1] = htons((uint16_t)itpl->itpl_remote_sin6.sin6_scope_id);
1858 		}
1859 
1860 		inp = in6_pcblookup_hash(pcbinfo,
1861 		    &ina6_remote,
1862 		    itpl->itpl_remote_sin6.sin6_port,
1863 		    itpl->itpl_remote_sin6.sin6_scope_id,
1864 		    &ina6_local,
1865 		    itpl->itpl_local_sin6.sin6_port,
1866 		    itpl->itpl_local_sin6.sin6_scope_id,
1867 		    0, NULL);
1868 	} else {
1869 		return EINVAL;
1870 	}
1871 
1872 	if (inp != NULL) {
1873 		if ((so = inp->inp_socket) == NULL) {
1874 			return ENOENT;
1875 		}
1876 		socket_lock(so, 0);
1877 		if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
1878 			socket_unlock(so, 0);
1879 			return ENOENT;
1880 		}
1881 		tp = intotcpcb(inp);
1882 
1883 		tcp_fill_info(tp, ti);
1884 		socket_unlock(so, 0);
1885 
1886 		return 0;
1887 	}
1888 #if SKYWALK
1889 	else {
1890 		/* if no pcb found, check for flowswitch for uTCP flow */
1891 		int error;
1892 		struct nexus_mib_filter nmf = {
1893 			.nmf_type = NXMIB_FLOW,
1894 			.nmf_bitmap = NXMIB_FILTER_INFO_TUPLE,
1895 			.nmf_info_tuple = *itpl,
1896 		};
1897 		struct sk_stats_flow sf;
1898 		size_t len = sizeof(sf);
1899 		error = kernel_sysctlbyname(SK_STATS_FLOW, &sf, &len, &nmf, sizeof(nmf));
1900 		if (error != 0) {
1901 			printf("kernel_sysctlbyname err %d\n", error);
1902 			return error;
1903 		}
1904 		if (len != sizeof(sf)) {
1905 			printf("kernel_sysctlbyname invalid len %zu\n", len);
1906 			return ENOENT;
1907 		}
1908 
1909 		/*
1910 		 * This is what flow tracker can offer right now, which is good
1911 		 * for mDNS TCP keep alive offload.
1912 		 */
1913 		ti->tcpi_snd_nxt = sf.sf_lseq;
1914 		ti->tcpi_rcv_nxt = sf.sf_rseq;
1915 		ti->tcpi_rcv_space = (uint32_t)(sf.sf_lmax_win << sf.sf_lwscale);
1916 		ti->tcpi_rcv_wscale = sf.sf_lwscale;
1917 		ti->tcpi_last_outif = (int32_t)sf.sf_if_index;
1918 
1919 		return 0;
1920 	}
1921 #endif /* SKYWALK */
1922 
1923 	return ENOENT;
1924 }
1925 
1926 static void
tcp_connection_fill_info(struct tcpcb * tp,struct tcp_connection_info * tci)1927 tcp_connection_fill_info(struct tcpcb *tp, struct tcp_connection_info *tci)
1928 {
1929 	struct inpcb *inp = tp->t_inpcb;
1930 
1931 	bzero(tci, sizeof(*tci));
1932 	tci->tcpi_state = (uint8_t)tp->t_state;
1933 
1934 	if (TSTMP_SUPPORTED(tp)) {
1935 		tci->tcpi_options |= TCPCI_OPT_TIMESTAMPS;
1936 	}
1937 	if (SACK_ENABLED(tp)) {
1938 		tci->tcpi_options |= TCPCI_OPT_SACK;
1939 	}
1940 	if (TCP_WINDOW_SCALE_ENABLED(tp)) {
1941 		tci->tcpi_options |= TCPCI_OPT_WSCALE;
1942 		tci->tcpi_snd_wscale = tp->snd_scale;
1943 		tci->tcpi_rcv_wscale = tp->rcv_scale;
1944 	}
1945 	if (TCP_ECN_ENABLED(tp)) {
1946 		tci->tcpi_options |= TCPCI_OPT_ECN;
1947 	}
1948 	if (IN_FASTRECOVERY(tp) || tp->t_rxtshift > 0) {
1949 		tci->tcpi_flags |= TCPCI_FLAG_LOSSRECOVERY;
1950 	}
1951 	if (tp->t_flagsext & TF_PKTS_REORDERED) {
1952 		tci->tcpi_flags |= TCPCI_FLAG_REORDERING_DETECTED;
1953 	}
1954 	tci->tcpi_rto = tp->t_timer[TCPT_REXMT] > 0 ? tp->t_rxtcur : 0;
1955 	tci->tcpi_maxseg = tp->t_maxseg;
1956 	tci->tcpi_snd_ssthresh = tp->snd_ssthresh;
1957 	tci->tcpi_snd_cwnd = tp->snd_cwnd;
1958 	tci->tcpi_snd_wnd = tp->snd_wnd;
1959 	if (inp != NULL && inp->inp_socket != NULL) {
1960 		tci->tcpi_snd_sbbytes = inp->inp_socket->so_snd.sb_cc;
1961 	}
1962 	tci->tcpi_rcv_wnd = tp->rcv_adv > tp->rcv_nxt ? tp->rcv_adv - tp->rcv_nxt : 0;
1963 	tci->tcpi_rttcur = tp->t_rttcur;
1964 	tci->tcpi_srtt = (tp->t_srtt >> TCP_RTT_SHIFT);
1965 	tci->tcpi_rttvar = (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1966 	tci->tcpi_txpackets = inp != NULL ? inp->inp_stat->txpackets : 0;
1967 	tci->tcpi_txbytes = inp != NULL ? inp->inp_stat->txbytes : 0;
1968 	tci->tcpi_txretransmitbytes = tp->t_stat.txretransmitbytes;
1969 	tci->tcpi_txretransmitpackets = tp->t_stat.rxmitpkts;
1970 	tci->tcpi_rxpackets = inp != NULL ? inp->inp_stat->rxpackets : 0;
1971 	tci->tcpi_rxbytes = inp != NULL ? inp->inp_stat->rxbytes : 0;
1972 	tci->tcpi_rxoutoforderbytes = tp->t_stat.rxoutoforderbytes;
1973 
1974 	tci->tcpi_tfo_syn_data_rcv = !!(tp->t_tfo_stats & TFO_S_SYNDATA_RCV);
1975 	tci->tcpi_tfo_cookie_req_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIEREQ_RECV);
1976 	tci->tcpi_tfo_cookie_sent = !!(tp->t_tfo_stats & TFO_S_COOKIE_SENT);
1977 	tci->tcpi_tfo_cookie_invalid = !!(tp->t_tfo_stats & TFO_S_COOKIE_INVALID);
1978 	tci->tcpi_tfo_cookie_req = !!(tp->t_tfo_stats & TFO_S_COOKIE_REQ);
1979 	tci->tcpi_tfo_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIE_RCV);
1980 	tci->tcpi_tfo_syn_data_sent = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_SENT);
1981 	tci->tcpi_tfo_syn_data_acked = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_ACKED);
1982 	tci->tcpi_tfo_syn_loss = !!(tp->t_tfo_stats & TFO_S_SYN_LOSS);
1983 	tci->tcpi_tfo_cookie_wrong = !!(tp->t_tfo_stats & TFO_S_COOKIE_WRONG);
1984 	tci->tcpi_tfo_no_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_NO_COOKIE_RCV);
1985 	tci->tcpi_tfo_heuristics_disable = !!(tp->t_tfo_stats & TFO_S_HEURISTICS_DISABLE);
1986 	tci->tcpi_tfo_send_blackhole = !!(tp->t_tfo_stats & TFO_S_SEND_BLACKHOLE);
1987 	tci->tcpi_tfo_recv_blackhole = !!(tp->t_tfo_stats & TFO_S_RECV_BLACKHOLE);
1988 	tci->tcpi_tfo_onebyte_proxy = !!(tp->t_tfo_stats & TFO_S_ONE_BYTE_PROXY);
1989 }
1990 
1991 
1992 __private_extern__ int
tcp_sysctl_info(__unused struct sysctl_oid * oidp,__unused void * arg1,__unused int arg2,struct sysctl_req * req)1993 tcp_sysctl_info(__unused struct sysctl_oid *oidp, __unused void *arg1, __unused int arg2, struct sysctl_req *req)
1994 {
1995 	int error;
1996 	struct tcp_info ti = {};
1997 	struct info_tuple itpl;
1998 
1999 	if (req->newptr == USER_ADDR_NULL) {
2000 		return EINVAL;
2001 	}
2002 	if (req->newlen < sizeof(struct info_tuple)) {
2003 		return EINVAL;
2004 	}
2005 	error = SYSCTL_IN(req, &itpl, sizeof(struct info_tuple));
2006 	if (error != 0) {
2007 		return error;
2008 	}
2009 	error = tcp_fill_info_for_info_tuple(&itpl, &ti);
2010 	if (error != 0) {
2011 		return error;
2012 	}
2013 	error = SYSCTL_OUT(req, &ti, sizeof(struct tcp_info));
2014 	if (error != 0) {
2015 		return error;
2016 	}
2017 
2018 	return 0;
2019 }
2020 
2021 static int
tcp_lookup_peer_pid_locked(struct socket * so,pid_t * out_pid)2022 tcp_lookup_peer_pid_locked(struct socket *so, pid_t *out_pid)
2023 {
2024 	int error = EHOSTUNREACH;
2025 	*out_pid = -1;
2026 	if ((so->so_state & SS_ISCONNECTED) == 0) {
2027 		return ENOTCONN;
2028 	}
2029 
2030 	struct inpcb    *inp = (struct inpcb*)so->so_pcb;
2031 	uint16_t                lport = inp->inp_lport;
2032 	uint16_t                fport = inp->inp_fport;
2033 	uint32_t                                fifscope = inp->inp_fifscope;
2034 	uint32_t                                lifscope = inp->inp_lifscope;
2035 
2036 	struct inpcb    *finp = NULL;
2037 	struct  in6_addr laddr6, faddr6;
2038 	struct in_addr laddr4, faddr4;
2039 
2040 	if (inp->inp_vflag & INP_IPV6) {
2041 		laddr6 = inp->in6p_laddr;
2042 		faddr6 = inp->in6p_faddr;
2043 	} else if (inp->inp_vflag & INP_IPV4) {
2044 		laddr4 = inp->inp_laddr;
2045 		faddr4 = inp->inp_faddr;
2046 	}
2047 
2048 	socket_unlock(so, 0);
2049 	if (inp->inp_vflag & INP_IPV6) {
2050 		finp = in6_pcblookup_hash(&tcbinfo, &laddr6, lport, lifscope, &faddr6, fport, fifscope, 0, NULL);
2051 	} else if (inp->inp_vflag & INP_IPV4) {
2052 		finp = in_pcblookup_hash(&tcbinfo, laddr4, lport, faddr4, fport, 0, NULL);
2053 	}
2054 
2055 	if (finp) {
2056 		*out_pid = finp->inp_socket->last_pid;
2057 		error = 0;
2058 		in_pcb_checkstate(finp, WNT_RELEASE, 0);
2059 	}
2060 	socket_lock(so, 0);
2061 
2062 	return error;
2063 }
2064 
2065 void
tcp_getconninfo(struct socket * so,struct conninfo_tcp * tcp_ci)2066 tcp_getconninfo(struct socket *so, struct conninfo_tcp *tcp_ci)
2067 {
2068 	(void) tcp_lookup_peer_pid_locked(so, &tcp_ci->tcpci_peer_pid);
2069 	tcp_fill_info(sototcpcb(so), &tcp_ci->tcpci_tcp_info);
2070 }
2071 
2072 void
tcp_clear_keep_alive_offload(struct socket * so)2073 tcp_clear_keep_alive_offload(struct socket *so)
2074 {
2075 	struct inpcb *inp;
2076 	struct ifnet *ifp;
2077 
2078 	inp = sotoinpcb(so);
2079 	if (inp == NULL) {
2080 		return;
2081 	}
2082 
2083 	if ((inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD) == 0) {
2084 		return;
2085 	}
2086 
2087 	ifp = inp->inp_boundifp != NULL ? inp->inp_boundifp :
2088 	    inp->inp_last_outifp;
2089 	if (ifp == NULL) {
2090 		panic("%s: so %p inp %p ifp NULL",
2091 		    __func__, so, inp);
2092 	}
2093 
2094 	ifnet_lock_exclusive(ifp);
2095 
2096 	if (ifp->if_tcp_kao_cnt == 0) {
2097 		panic("%s: so %p inp %p ifp %p if_tcp_kao_cnt == 0",
2098 		    __func__, so, inp, ifp);
2099 	}
2100 	ifp->if_tcp_kao_cnt--;
2101 	inp->inp_flags2 &= ~INP2_KEEPALIVE_OFFLOAD;
2102 
2103 	ifnet_lock_done(ifp);
2104 }
2105 
2106 static int
tcp_set_keep_alive_offload(struct socket * so,struct proc * proc)2107 tcp_set_keep_alive_offload(struct socket *so, struct proc *proc)
2108 {
2109 	int error = 0;
2110 	struct inpcb *inp;
2111 	struct ifnet *ifp;
2112 
2113 	inp = sotoinpcb(so);
2114 	if (inp == NULL) {
2115 		return ECONNRESET;
2116 	}
2117 	if ((inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD) != 0) {
2118 		return 0;
2119 	}
2120 
2121 	ifp = inp->inp_boundifp != NULL ? inp->inp_boundifp :
2122 	    inp->inp_last_outifp;
2123 	if (ifp == NULL) {
2124 		error = ENXIO;
2125 		os_log_info(OS_LOG_DEFAULT,
2126 		    "%s: error %d for proc %s[%u] out ifp is not set\n",
2127 		    __func__, error,
2128 		    proc != NULL ? proc->p_comm : "kernel",
2129 		    proc != NULL ? proc_getpid(proc) : 0);
2130 		return ENXIO;
2131 	}
2132 
2133 	error = if_get_tcp_kao_max(ifp);
2134 	if (error != 0) {
2135 		return error;
2136 	}
2137 
2138 	ifnet_lock_exclusive(ifp);
2139 	if (ifp->if_tcp_kao_cnt < ifp->if_tcp_kao_max) {
2140 		ifp->if_tcp_kao_cnt++;
2141 		inp->inp_flags2 |= INP2_KEEPALIVE_OFFLOAD;
2142 	} else {
2143 		error = ETOOMANYREFS;
2144 		os_log_info(OS_LOG_DEFAULT,
2145 		    "%s: error %d for proc %s[%u] if_tcp_kao_max %u\n",
2146 		    __func__, error,
2147 		    proc != NULL ? proc->p_comm : "kernel",
2148 		    proc != NULL ? proc_getpid(proc) : 0,
2149 		    ifp->if_tcp_kao_max);
2150 	}
2151 	ifnet_lock_done(ifp);
2152 
2153 	return error;
2154 }
2155 
2156 /*
2157  * The new sockopt interface makes it possible for us to block in the
2158  * copyin/out step (if we take a page fault).  Taking a page fault at
2159  * splnet() is probably a Bad Thing.  (Since sockets and pcbs both now
2160  * use TSM, there probably isn't any need for this function to run at
2161  * splnet() any more.  This needs more examination.)
2162  */
2163 int
tcp_ctloutput(struct socket * so,struct sockopt * sopt)2164 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
2165 {
2166 	int     error = 0, opt = 0, optval = 0;
2167 	struct  inpcb *inp;
2168 	struct  tcpcb *tp;
2169 
2170 	inp = sotoinpcb(so);
2171 	if (inp == NULL) {
2172 		return ECONNRESET;
2173 	}
2174 	/* Allow <SOL_SOCKET,SO_FLUSH/SO_TRAFFIC_MGT_BACKGROUND> at this level */
2175 	if (sopt->sopt_level != IPPROTO_TCP &&
2176 	    !(sopt->sopt_level == SOL_SOCKET && (sopt->sopt_name == SO_FLUSH ||
2177 	    sopt->sopt_name == SO_TRAFFIC_MGT_BACKGROUND))) {
2178 		if (SOCK_CHECK_DOM(so, PF_INET6)) {
2179 			error = ip6_ctloutput(so, sopt);
2180 		} else {
2181 			error = ip_ctloutput(so, sopt);
2182 		}
2183 		return error;
2184 	}
2185 	tp = intotcpcb(inp);
2186 	if (tp == NULL) {
2187 		return ECONNRESET;
2188 	}
2189 
2190 	calculate_tcp_clock();
2191 
2192 	switch (sopt->sopt_dir) {
2193 	case SOPT_SET:
2194 		switch (sopt->sopt_name) {
2195 		case TCP_NODELAY:
2196 		case TCP_NOOPT:
2197 		case TCP_NOPUSH:
2198 			error = sooptcopyin(sopt, &optval, sizeof optval,
2199 			    sizeof optval);
2200 			if (error) {
2201 				break;
2202 			}
2203 
2204 			switch (sopt->sopt_name) {
2205 			case TCP_NODELAY:
2206 				opt = TF_NODELAY;
2207 				break;
2208 			case TCP_NOOPT:
2209 				opt = TF_NOOPT;
2210 				break;
2211 			case TCP_NOPUSH:
2212 				opt = TF_NOPUSH;
2213 				break;
2214 			default:
2215 				opt = 0; /* dead code to fool gcc */
2216 				break;
2217 			}
2218 
2219 			if (optval) {
2220 				tp->t_flags |= opt;
2221 			} else {
2222 				tp->t_flags &= ~opt;
2223 			}
2224 			break;
2225 		case TCP_RXT_FINDROP:
2226 		case TCP_NOTIMEWAIT:
2227 			error = sooptcopyin(sopt, &optval, sizeof optval,
2228 			    sizeof optval);
2229 			if (error) {
2230 				break;
2231 			}
2232 			switch (sopt->sopt_name) {
2233 			case TCP_RXT_FINDROP:
2234 				opt = TF_RXTFINDROP;
2235 				break;
2236 			case TCP_NOTIMEWAIT:
2237 				opt = TF_NOTIMEWAIT;
2238 				break;
2239 			default:
2240 				opt = 0;
2241 				break;
2242 			}
2243 			if (optval) {
2244 				tp->t_flagsext |= opt;
2245 			} else {
2246 				tp->t_flagsext &= ~opt;
2247 			}
2248 			break;
2249 		case TCP_MEASURE_SND_BW:
2250 			error = sooptcopyin(sopt, &optval, sizeof optval,
2251 			    sizeof optval);
2252 			if (error) {
2253 				break;
2254 			}
2255 			opt = TF_MEASURESNDBW;
2256 			if (optval) {
2257 				if (tp->t_bwmeas == NULL) {
2258 					tp->t_bwmeas = tcp_bwmeas_alloc(tp);
2259 					if (tp->t_bwmeas == NULL) {
2260 						error = ENOMEM;
2261 						break;
2262 					}
2263 				}
2264 				tp->t_flagsext |= opt;
2265 			} else {
2266 				tp->t_flagsext &= ~opt;
2267 				/* Reset snd bw measurement state */
2268 				tp->t_flagsext &= ~(TF_BWMEAS_INPROGRESS);
2269 				if (tp->t_bwmeas != NULL) {
2270 					tcp_bwmeas_free(tp);
2271 				}
2272 			}
2273 			break;
2274 		case TCP_MEASURE_BW_BURST: {
2275 			struct tcp_measure_bw_burst in;
2276 			uint32_t minpkts, maxpkts;
2277 			bzero(&in, sizeof(in));
2278 
2279 			error = sooptcopyin(sopt, &in, sizeof(in),
2280 			    sizeof(in));
2281 			if (error) {
2282 				break;
2283 			}
2284 			if ((tp->t_flagsext & TF_MEASURESNDBW) == 0 ||
2285 			    tp->t_bwmeas == NULL) {
2286 				error = EINVAL;
2287 				break;
2288 			}
2289 			minpkts = (in.min_burst_size != 0) ? in.min_burst_size :
2290 			    tp->t_bwmeas->bw_minsizepkts;
2291 			maxpkts = (in.max_burst_size != 0) ? in.max_burst_size :
2292 			    tp->t_bwmeas->bw_maxsizepkts;
2293 			if (minpkts > maxpkts) {
2294 				error = EINVAL;
2295 				break;
2296 			}
2297 			tp->t_bwmeas->bw_minsizepkts = minpkts;
2298 			tp->t_bwmeas->bw_maxsizepkts = maxpkts;
2299 			tp->t_bwmeas->bw_minsize = (minpkts * tp->t_maxseg);
2300 			tp->t_bwmeas->bw_maxsize = (maxpkts * tp->t_maxseg);
2301 			break;
2302 		}
2303 		case TCP_MAXSEG:
2304 			error = sooptcopyin(sopt, &optval, sizeof optval,
2305 			    sizeof optval);
2306 			if (error) {
2307 				break;
2308 			}
2309 
2310 			if (optval > 0 && optval <= tp->t_maxseg &&
2311 			    optval + 40 >= tcp_minmss) {
2312 				tp->t_maxseg = optval;
2313 			} else {
2314 				error = EINVAL;
2315 			}
2316 			break;
2317 
2318 		case TCP_KEEPALIVE:
2319 			error = sooptcopyin(sopt, &optval, sizeof optval,
2320 			    sizeof optval);
2321 			if (error) {
2322 				break;
2323 			}
2324 			if (optval < 0 || optval > UINT32_MAX / TCP_RETRANSHZ) {
2325 				error = EINVAL;
2326 			} else {
2327 				tp->t_keepidle = optval * TCP_RETRANSHZ;
2328 				/* reset the timer to new value */
2329 				tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
2330 				    TCP_CONN_KEEPIDLE(tp));
2331 				tcp_check_timer_state(tp);
2332 			}
2333 			break;
2334 
2335 		case TCP_CONNECTIONTIMEOUT:
2336 			error = sooptcopyin(sopt, &optval, sizeof optval,
2337 			    sizeof optval);
2338 			if (error) {
2339 				break;
2340 			}
2341 			if (optval < 0 || optval > UINT32_MAX / TCP_RETRANSHZ) {
2342 				error = EINVAL;
2343 			} else {
2344 				tp->t_keepinit = optval * TCP_RETRANSHZ;
2345 				if (tp->t_state == TCPS_SYN_RECEIVED ||
2346 				    tp->t_state == TCPS_SYN_SENT) {
2347 					tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
2348 					    TCP_CONN_KEEPINIT(tp));
2349 					tcp_check_timer_state(tp);
2350 				}
2351 			}
2352 			break;
2353 
2354 		case TCP_KEEPINTVL:
2355 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2356 			    sizeof(optval));
2357 			if (error) {
2358 				break;
2359 			}
2360 			if (optval < 0 || optval > UINT32_MAX / TCP_RETRANSHZ) {
2361 				error = EINVAL;
2362 			} else {
2363 				tp->t_keepintvl = optval * TCP_RETRANSHZ;
2364 				if (tp->t_state == TCPS_FIN_WAIT_2 &&
2365 				    TCP_CONN_MAXIDLE(tp) > 0) {
2366 					tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp,
2367 					    TCP_CONN_MAXIDLE(tp));
2368 					tcp_check_timer_state(tp);
2369 				}
2370 			}
2371 			break;
2372 
2373 		case TCP_KEEPCNT:
2374 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2375 			    sizeof(optval));
2376 			if (error) {
2377 				break;
2378 			}
2379 			if (optval < 0 || optval > INT32_MAX) {
2380 				error = EINVAL;
2381 			} else {
2382 				tp->t_keepcnt = optval;
2383 				if (tp->t_state == TCPS_FIN_WAIT_2 &&
2384 				    TCP_CONN_MAXIDLE(tp) > 0) {
2385 					tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp,
2386 					    TCP_CONN_MAXIDLE(tp));
2387 					tcp_check_timer_state(tp);
2388 				}
2389 			}
2390 			break;
2391 
2392 		case TCP_KEEPALIVE_OFFLOAD:
2393 			if ((error = priv_check_cred(kauth_cred_get(),
2394 			    PRIV_NETINET_TCP_KA_OFFLOAD, 0)) != 0) {
2395 				break;
2396 			}
2397 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2398 			    sizeof(optval));
2399 			if (error) {
2400 				break;
2401 			}
2402 			if (optval < 0 || optval > INT32_MAX) {
2403 				error = EINVAL;
2404 				break;
2405 			}
2406 			if (optval != 0) {
2407 				error = tcp_set_keep_alive_offload(so,
2408 				    sopt->sopt_p);
2409 			} else {
2410 				tcp_clear_keep_alive_offload(so);
2411 			}
2412 			break;
2413 
2414 		case PERSIST_TIMEOUT:
2415 			error = sooptcopyin(sopt, &optval, sizeof optval,
2416 			    sizeof optval);
2417 			if (error) {
2418 				break;
2419 			}
2420 			if (optval < 0) {
2421 				error = EINVAL;
2422 			} else {
2423 				tp->t_persist_timeout = optval * TCP_RETRANSHZ;
2424 			}
2425 			break;
2426 		case TCP_RXT_CONNDROPTIME:
2427 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2428 			    sizeof(optval));
2429 			if (error) {
2430 				break;
2431 			}
2432 			if (optval < 0) {
2433 				error = EINVAL;
2434 			} else {
2435 				tp->t_rxt_conndroptime = optval * TCP_RETRANSHZ;
2436 			}
2437 			break;
2438 		case TCP_NOTSENT_LOWAT:
2439 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2440 			    sizeof(optval));
2441 			if (error) {
2442 				break;
2443 			}
2444 			if (optval < 0) {
2445 				error = EINVAL;
2446 				break;
2447 			} else {
2448 				if (optval == 0) {
2449 					so->so_flags &= ~(SOF_NOTSENT_LOWAT);
2450 					tp->t_notsent_lowat = 0;
2451 				} else {
2452 					so->so_flags |= SOF_NOTSENT_LOWAT;
2453 					tp->t_notsent_lowat = optval;
2454 				}
2455 			}
2456 			break;
2457 		case TCP_ADAPTIVE_READ_TIMEOUT:
2458 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2459 			    sizeof(optval));
2460 			if (error) {
2461 				break;
2462 			}
2463 			if (optval < 0 ||
2464 			    optval > TCP_ADAPTIVE_TIMEOUT_MAX) {
2465 				error = EINVAL;
2466 				break;
2467 			} else if (optval == 0) {
2468 				tp->t_adaptive_rtimo = 0;
2469 				tcp_keepalive_reset(tp);
2470 
2471 				if (tp->t_mpsub) {
2472 					mptcp_reset_keepalive(tp);
2473 				}
2474 			} else {
2475 				tp->t_adaptive_rtimo = (uint8_t)optval;
2476 			}
2477 			break;
2478 		case TCP_ADAPTIVE_WRITE_TIMEOUT:
2479 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2480 			    sizeof(optval));
2481 			if (error) {
2482 				break;
2483 			}
2484 			if (optval < 0 ||
2485 			    optval > TCP_ADAPTIVE_TIMEOUT_MAX) {
2486 				error = EINVAL;
2487 				break;
2488 			} else {
2489 				tp->t_adaptive_wtimo = (uint8_t)optval;
2490 			}
2491 			break;
2492 		case TCP_SENDMOREACKS:
2493 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2494 			    sizeof(optval));
2495 			if (error) {
2496 				break;
2497 			}
2498 			if (optval < 0 || optval > 1) {
2499 				error = EINVAL;
2500 			} else if (optval == 0) {
2501 				tp->t_flagsext &= ~(TF_NOSTRETCHACK);
2502 			} else {
2503 				tp->t_flagsext |= TF_NOSTRETCHACK;
2504 			}
2505 			break;
2506 		case TCP_DISABLE_BLACKHOLE_DETECTION:
2507 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2508 			    sizeof(optval));
2509 			if (error) {
2510 				break;
2511 			}
2512 			if (optval < 0 || optval > 1) {
2513 				error = EINVAL;
2514 			} else if (optval == 0) {
2515 				tp->t_flagsext &= ~TF_NOBLACKHOLE_DETECTION;
2516 			} else {
2517 				tp->t_flagsext |= TF_NOBLACKHOLE_DETECTION;
2518 				if ((tp->t_flags & TF_BLACKHOLE) &&
2519 				    tp->t_pmtud_saved_maxopd > 0) {
2520 					tcp_pmtud_revert_segment_size(tp);
2521 				}
2522 			}
2523 			break;
2524 		case TCP_FASTOPEN:
2525 			if (!(tcp_fastopen & TCP_FASTOPEN_SERVER)) {
2526 				error = ENOTSUP;
2527 				break;
2528 			}
2529 
2530 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2531 			    sizeof(optval));
2532 			if (error) {
2533 				break;
2534 			}
2535 			if (optval < 0 || optval > 1) {
2536 				error = EINVAL;
2537 				break;
2538 			}
2539 			if (tp->t_state != TCPS_LISTEN) {
2540 				error =  EINVAL;
2541 				break;
2542 			}
2543 			if (optval) {
2544 				tp->t_flagsext |= TF_FASTOPEN;
2545 			} else {
2546 				tcp_disable_tfo(tp);
2547 			}
2548 			break;
2549 		case TCP_FASTOPEN_FORCE_HEURISTICS:
2550 
2551 			break;
2552 		case TCP_FASTOPEN_FORCE_ENABLE:
2553 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2554 			    sizeof(optval));
2555 
2556 			if (error) {
2557 				break;
2558 			}
2559 			if (optval < 0 || optval > 1) {
2560 				error = EINVAL;
2561 				break;
2562 			}
2563 
2564 			if (tp->t_state != TCPS_CLOSED) {
2565 				error =  EINVAL;
2566 				break;
2567 			}
2568 			if (optval) {
2569 				tp->t_flagsext |= TF_FASTOPEN_FORCE_ENABLE;
2570 			} else {
2571 				tp->t_flagsext &= ~TF_FASTOPEN_FORCE_ENABLE;
2572 			}
2573 
2574 			break;
2575 		case TCP_ENABLE_ECN:
2576 			error = sooptcopyin(sopt, &optval, sizeof optval,
2577 			    sizeof optval);
2578 			if (error) {
2579 				break;
2580 			}
2581 			if (optval) {
2582 				tp->ecn_flags |= TE_ECN_MODE_ENABLE;
2583 				tp->ecn_flags &= ~TE_ECN_MODE_DISABLE;
2584 			} else {
2585 				tp->ecn_flags &= ~TE_ECN_MODE_ENABLE;
2586 				tp->ecn_flags |= TE_ECN_MODE_DISABLE;
2587 			}
2588 			break;
2589 		case TCP_ECN_MODE:
2590 			error = sooptcopyin(sopt, &optval, sizeof optval,
2591 			    sizeof optval);
2592 			if (error) {
2593 				break;
2594 			}
2595 			if (optval == ECN_MODE_DEFAULT) {
2596 				tp->ecn_flags &= ~TE_ECN_MODE_ENABLE;
2597 				tp->ecn_flags &= ~TE_ECN_MODE_DISABLE;
2598 			} else if (optval == ECN_MODE_ENABLE) {
2599 				tp->ecn_flags |= TE_ECN_MODE_ENABLE;
2600 				tp->ecn_flags &= ~TE_ECN_MODE_DISABLE;
2601 			} else if (optval == ECN_MODE_DISABLE) {
2602 				tp->ecn_flags &= ~TE_ECN_MODE_ENABLE;
2603 				tp->ecn_flags |= TE_ECN_MODE_DISABLE;
2604 			} else {
2605 				error = EINVAL;
2606 			}
2607 			break;
2608 		case TCP_NOTIFY_ACKNOWLEDGEMENT:
2609 			error = sooptcopyin(sopt, &optval,
2610 			    sizeof(optval), sizeof(optval));
2611 			if (error) {
2612 				break;
2613 			}
2614 			if (optval <= 0) {
2615 				error = EINVAL;
2616 				break;
2617 			}
2618 			if (tp->t_notify_ack_count >= TCP_MAX_NOTIFY_ACK) {
2619 				error = ETOOMANYREFS;
2620 				break;
2621 			}
2622 
2623 			/*
2624 			 * validate that the given marker id is not
2625 			 * a duplicate to avoid ambiguity
2626 			 */
2627 			if ((error = tcp_notify_ack_id_valid(tp, so,
2628 			    optval)) != 0) {
2629 				break;
2630 			}
2631 			error = tcp_add_notify_ack_marker(tp, optval);
2632 			break;
2633 		case SO_FLUSH:
2634 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
2635 			    sizeof(optval))) != 0) {
2636 				break;
2637 			}
2638 
2639 			error = inp_flush(inp, optval);
2640 			break;
2641 
2642 		case SO_TRAFFIC_MGT_BACKGROUND:
2643 			if ((error = sooptcopyin(sopt, &optval, sizeof(optval),
2644 			    sizeof(optval))) != 0) {
2645 				break;
2646 			}
2647 
2648 			if (optval) {
2649 				socket_set_traffic_mgt_flags_locked(so,
2650 				    TRAFFIC_MGT_SO_BACKGROUND);
2651 			} else {
2652 				socket_clear_traffic_mgt_flags_locked(so,
2653 				    TRAFFIC_MGT_SO_BACKGROUND);
2654 			}
2655 			break;
2656 		case TCP_RXT_MINIMUM_TIMEOUT:
2657 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2658 			    sizeof(optval));
2659 			if (error) {
2660 				break;
2661 			}
2662 			if (optval < 0) {
2663 				error = EINVAL;
2664 				break;
2665 			}
2666 			if (optval == 0) {
2667 				tp->t_rxt_minimum_timeout = 0;
2668 			} else {
2669 				tp->t_rxt_minimum_timeout = min(optval,
2670 				    TCP_RXT_MINIMUM_TIMEOUT_LIMIT);
2671 				/* convert to milliseconds */
2672 				tp->t_rxt_minimum_timeout *= TCP_RETRANSHZ;
2673 			}
2674 			break;
2675 		default:
2676 			error = ENOPROTOOPT;
2677 			break;
2678 		}
2679 		break;
2680 
2681 	case SOPT_GET:
2682 		switch (sopt->sopt_name) {
2683 		case TCP_NODELAY:
2684 			optval = tp->t_flags & TF_NODELAY;
2685 			break;
2686 		case TCP_MAXSEG:
2687 			optval = tp->t_maxseg;
2688 			break;
2689 		case TCP_KEEPALIVE:
2690 			if (tp->t_keepidle > 0) {
2691 				optval = tp->t_keepidle / TCP_RETRANSHZ;
2692 			} else {
2693 				optval = tcp_keepidle  / TCP_RETRANSHZ;
2694 			}
2695 			break;
2696 		case TCP_KEEPINTVL:
2697 			if (tp->t_keepintvl > 0) {
2698 				optval = tp->t_keepintvl / TCP_RETRANSHZ;
2699 			} else {
2700 				optval = tcp_keepintvl / TCP_RETRANSHZ;
2701 			}
2702 			break;
2703 		case TCP_KEEPCNT:
2704 			if (tp->t_keepcnt > 0) {
2705 				optval = tp->t_keepcnt;
2706 			} else {
2707 				optval = tcp_keepcnt;
2708 			}
2709 			break;
2710 		case TCP_KEEPALIVE_OFFLOAD:
2711 			optval = !!(inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD);
2712 			break;
2713 		case TCP_NOOPT:
2714 			optval = tp->t_flags & TF_NOOPT;
2715 			break;
2716 		case TCP_NOPUSH:
2717 			optval = tp->t_flags & TF_NOPUSH;
2718 			break;
2719 		case TCP_ENABLE_ECN:
2720 			optval = (tp->ecn_flags & TE_ECN_MODE_ENABLE) ? 1 : 0;
2721 			break;
2722 		case TCP_ECN_MODE:
2723 			if (tp->ecn_flags & TE_ECN_MODE_ENABLE) {
2724 				optval = ECN_MODE_ENABLE;
2725 			} else if (tp->ecn_flags & TE_ECN_MODE_DISABLE) {
2726 				optval = ECN_MODE_DISABLE;
2727 			} else {
2728 				optval = ECN_MODE_DEFAULT;
2729 			}
2730 			break;
2731 		case TCP_CONNECTIONTIMEOUT:
2732 			optval = tp->t_keepinit / TCP_RETRANSHZ;
2733 			break;
2734 		case PERSIST_TIMEOUT:
2735 			optval = tp->t_persist_timeout / TCP_RETRANSHZ;
2736 			break;
2737 		case TCP_RXT_CONNDROPTIME:
2738 			optval = tp->t_rxt_conndroptime / TCP_RETRANSHZ;
2739 			break;
2740 		case TCP_RXT_FINDROP:
2741 			optval = tp->t_flagsext & TF_RXTFINDROP;
2742 			break;
2743 		case TCP_NOTIMEWAIT:
2744 			optval = (tp->t_flagsext & TF_NOTIMEWAIT) ? 1 : 0;
2745 			break;
2746 		case TCP_FASTOPEN:
2747 			if (tp->t_state != TCPS_LISTEN ||
2748 			    !(tcp_fastopen & TCP_FASTOPEN_SERVER)) {
2749 				error = ENOTSUP;
2750 				break;
2751 			}
2752 			optval = tfo_enabled(tp);
2753 			break;
2754 		case TCP_FASTOPEN_FORCE_HEURISTICS:
2755 			optval = 0;
2756 			break;
2757 		case TCP_FASTOPEN_FORCE_ENABLE:
2758 			optval = (tp->t_flagsext & TF_FASTOPEN_FORCE_ENABLE) ? 1 : 0;
2759 			break;
2760 		case TCP_MEASURE_SND_BW:
2761 			optval = tp->t_flagsext & TF_MEASURESNDBW;
2762 			break;
2763 		case TCP_INFO: {
2764 			struct tcp_info ti;
2765 
2766 			tcp_fill_info(tp, &ti);
2767 			error = sooptcopyout(sopt, &ti, sizeof(struct tcp_info));
2768 			goto done;
2769 			/* NOT REACHED */
2770 		}
2771 		case TCP_CONNECTION_INFO: {
2772 			struct tcp_connection_info tci;
2773 			tcp_connection_fill_info(tp, &tci);
2774 			error = sooptcopyout(sopt, &tci,
2775 			    sizeof(struct tcp_connection_info));
2776 			goto done;
2777 		}
2778 		case TCP_MEASURE_BW_BURST: {
2779 			struct tcp_measure_bw_burst out = {};
2780 			if ((tp->t_flagsext & TF_MEASURESNDBW) == 0 ||
2781 			    tp->t_bwmeas == NULL) {
2782 				error = EINVAL;
2783 				break;
2784 			}
2785 			out.min_burst_size = tp->t_bwmeas->bw_minsizepkts;
2786 			out.max_burst_size = tp->t_bwmeas->bw_maxsizepkts;
2787 			error = sooptcopyout(sopt, &out, sizeof(out));
2788 			goto done;
2789 		}
2790 		case TCP_NOTSENT_LOWAT:
2791 			if ((so->so_flags & SOF_NOTSENT_LOWAT) != 0) {
2792 				optval = tp->t_notsent_lowat;
2793 			} else {
2794 				optval = 0;
2795 			}
2796 			break;
2797 		case TCP_SENDMOREACKS:
2798 			if (tp->t_flagsext & TF_NOSTRETCHACK) {
2799 				optval = 1;
2800 			} else {
2801 				optval = 0;
2802 			}
2803 			break;
2804 		case TCP_DISABLE_BLACKHOLE_DETECTION:
2805 			if (tp->t_flagsext & TF_NOBLACKHOLE_DETECTION) {
2806 				optval = 1;
2807 			} else {
2808 				optval = 0;
2809 			}
2810 			break;
2811 		case TCP_PEER_PID: {
2812 			pid_t   pid;
2813 			error = tcp_lookup_peer_pid_locked(so, &pid);
2814 			if (error == 0) {
2815 				error = sooptcopyout(sopt, &pid, sizeof(pid));
2816 			}
2817 			goto done;
2818 		}
2819 		case TCP_ADAPTIVE_READ_TIMEOUT:
2820 			optval = tp->t_adaptive_rtimo;
2821 			break;
2822 		case TCP_ADAPTIVE_WRITE_TIMEOUT:
2823 			optval = tp->t_adaptive_wtimo;
2824 			break;
2825 		case SO_TRAFFIC_MGT_BACKGROUND:
2826 			optval = (so->so_flags1 &
2827 			    SOF1_TRAFFIC_MGT_SO_BACKGROUND) ? 1 : 0;
2828 			break;
2829 		case TCP_NOTIFY_ACKNOWLEDGEMENT: {
2830 			struct tcp_notify_ack_complete retid;
2831 
2832 			if (sopt->sopt_valsize != sizeof(retid)) {
2833 				error = EINVAL;
2834 				break;
2835 			}
2836 			bzero(&retid, sizeof(retid));
2837 			tcp_get_notify_ack_count(tp, &retid);
2838 			if (retid.notify_complete_count > 0) {
2839 				tcp_get_notify_ack_ids(tp, &retid);
2840 			}
2841 
2842 			error = sooptcopyout(sopt, &retid, sizeof(retid));
2843 			goto done;
2844 		}
2845 		case TCP_RXT_MINIMUM_TIMEOUT:
2846 			optval = tp->t_rxt_minimum_timeout / TCP_RETRANSHZ;
2847 			break;
2848 		default:
2849 			error = ENOPROTOOPT;
2850 			break;
2851 		}
2852 		if (error == 0) {
2853 			error = sooptcopyout(sopt, &optval, sizeof optval);
2854 		}
2855 		break;
2856 	}
2857 done:
2858 	return error;
2859 }
2860 
2861 /*
2862  * tcp_sendspace and tcp_recvspace are the default send and receive window
2863  * sizes, respectively.  These are obsolescent (this information should
2864  * be set by the route).
2865  */
2866 u_int32_t       tcp_sendspace = 1448 * 256;
2867 u_int32_t       tcp_recvspace = 1448 * 384;
2868 
2869 /* During attach, the size of socket buffer allocated is limited to
2870  * sb_max in sbreserve. Disallow setting the tcp send and recv space
2871  * to be more than sb_max because that will cause tcp_attach to fail
2872  * (see radar 5713060)
2873  */
2874 static int
sysctl_tcp_sospace(struct sysctl_oid * oidp,__unused void * arg1,int arg2,struct sysctl_req * req)2875 sysctl_tcp_sospace(struct sysctl_oid *oidp, __unused void *arg1,
2876     int arg2, struct sysctl_req *req)
2877 {
2878 #pragma unused(arg2)
2879 	u_int32_t new_value = 0, *space_p = NULL;
2880 	int changed = 0, error = 0;
2881 	u_quad_t sb_effective_max = (sb_max / (MSIZE + MCLBYTES)) * MCLBYTES;
2882 
2883 	switch (oidp->oid_number) {
2884 	case TCPCTL_SENDSPACE:
2885 		space_p = &tcp_sendspace;
2886 		break;
2887 	case TCPCTL_RECVSPACE:
2888 		space_p = &tcp_recvspace;
2889 		break;
2890 	default:
2891 		return EINVAL;
2892 	}
2893 	error = sysctl_io_number(req, *space_p, sizeof(u_int32_t),
2894 	    &new_value, &changed);
2895 	if (changed) {
2896 		if (new_value > 0 && new_value <= sb_effective_max) {
2897 			*space_p = new_value;
2898 			SYSCTL_SKMEM_UPDATE_AT_OFFSET(arg2, new_value);
2899 		} else {
2900 			error = ERANGE;
2901 		}
2902 	}
2903 	return error;
2904 }
2905 
2906 #if SYSCTL_SKMEM
2907 SYSCTL_PROC(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace,
2908     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_sendspace,
2909     offsetof(skmem_sysctl, tcp.sendspace), sysctl_tcp_sospace,
2910     "IU", "Maximum outgoing TCP datagram size");
2911 SYSCTL_PROC(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace,
2912     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_recvspace,
2913     offsetof(skmem_sysctl, tcp.recvspace), sysctl_tcp_sospace,
2914     "IU", "Maximum incoming TCP datagram size");
2915 #else /* SYSCTL_SKMEM */
2916 SYSCTL_PROC(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
2917     &tcp_sendspace, 0, &sysctl_tcp_sospace, "IU", "Maximum outgoing TCP datagram size");
2918 SYSCTL_PROC(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
2919     &tcp_recvspace, 0, &sysctl_tcp_sospace, "IU", "Maximum incoming TCP datagram size");
2920 #endif /* SYSCTL_SKMEM */
2921 
2922 /*
2923  * Attach TCP protocol to socket, allocating
2924  * internet protocol control block, tcp control block,
2925  * bufer space, and entering LISTEN state if to accept connections.
2926  *
2927  * Returns:	0			Success
2928  *	in_pcballoc:ENOBUFS
2929  *	in_pcballoc:ENOMEM
2930  *	in_pcballoc:???			[IPSEC specific]
2931  *	soreserve:ENOBUFS
2932  */
2933 static int
tcp_attach(struct socket * so,struct proc * p)2934 tcp_attach(struct socket *so, struct proc *p)
2935 {
2936 	struct tcpcb *tp;
2937 	struct inpcb *inp;
2938 	int error;
2939 	int isipv6 = SOCK_CHECK_DOM(so, PF_INET6) != 0;
2940 
2941 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
2942 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
2943 		if (error) {
2944 			return error;
2945 		}
2946 	}
2947 
2948 	error = in_pcballoc(so, &tcbinfo, p);
2949 	if (error) {
2950 		return error;
2951 	}
2952 
2953 	inp = sotoinpcb(so);
2954 
2955 	if (so->so_snd.sb_preconn_hiwat == 0) {
2956 		soreserve_preconnect(so, 2048);
2957 	}
2958 
2959 	if ((so->so_rcv.sb_flags & SB_USRSIZE) == 0) {
2960 		so->so_rcv.sb_flags |= SB_AUTOSIZE;
2961 	}
2962 	if ((so->so_snd.sb_flags & SB_USRSIZE) == 0) {
2963 		so->so_snd.sb_flags |= SB_AUTOSIZE;
2964 	}
2965 
2966 	if (isipv6) {
2967 		inp->inp_vflag |= INP_IPV6;
2968 		inp->in6p_hops = -1;    /* use kernel default */
2969 	} else {
2970 		inp->inp_vflag |= INP_IPV4;
2971 	}
2972 	tp = tcp_newtcpcb(inp);
2973 	if (tp == NULL) {
2974 		int nofd = so->so_state & SS_NOFDREF;   /* XXX */
2975 
2976 		so->so_state &= ~SS_NOFDREF;    /* don't free the socket yet */
2977 		if (isipv6) {
2978 			in6_pcbdetach(inp);
2979 		} else {
2980 			in_pcbdetach(inp);
2981 		}
2982 		so->so_state |= nofd;
2983 		return ENOBUFS;
2984 	}
2985 	if (nstat_collect) {
2986 		nstat_tcp_new_pcb(inp);
2987 	}
2988 	tp->t_state = TCPS_CLOSED;
2989 	return 0;
2990 }
2991 
2992 /*
2993  * Initiate (or continue) disconnect.
2994  * If embryonic state, just send reset (once).
2995  * If in ``let data drain'' option and linger null, just drop.
2996  * Otherwise (hard), mark socket disconnecting and drop
2997  * current input data; switch states based on user close, and
2998  * send segment to peer (with FIN).
2999  */
3000 static struct tcpcb *
tcp_disconnect(struct tcpcb * tp)3001 tcp_disconnect(struct tcpcb *tp)
3002 {
3003 	struct socket *so = tp->t_inpcb->inp_socket;
3004 
3005 	if (so->so_rcv.sb_cc != 0 || tp->t_reassqlen != 0) {
3006 		return tcp_drop(tp, 0);
3007 	}
3008 
3009 	if (tp->t_state < TCPS_ESTABLISHED) {
3010 		tp = tcp_close(tp);
3011 	} else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
3012 		tp = tcp_drop(tp, 0);
3013 	} else {
3014 		soisdisconnecting(so);
3015 		sbflush(&so->so_rcv);
3016 		tp = tcp_usrclosed(tp);
3017 #if MPTCP
3018 		/* A reset has been sent but socket exists, do not send FIN */
3019 		if ((so->so_flags & SOF_MP_SUBFLOW) &&
3020 		    (tp) && (tp->t_mpflags & TMPF_RESET)) {
3021 			return tp;
3022 		}
3023 #endif
3024 		if (tp) {
3025 			(void) tcp_output(tp);
3026 		}
3027 	}
3028 	return tp;
3029 }
3030 
3031 /*
3032  * User issued close, and wish to trail through shutdown states:
3033  * if never received SYN, just forget it.  If got a SYN from peer,
3034  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
3035  * If already got a FIN from peer, then almost done; go to LAST_ACK
3036  * state.  In all other cases, have already sent FIN to peer (e.g.
3037  * after PRU_SHUTDOWN), and just have to play tedious game waiting
3038  * for peer to send FIN or not respond to keep-alives, etc.
3039  * We can let the user exit from the close as soon as the FIN is acked.
3040  */
3041 static struct tcpcb *
tcp_usrclosed(struct tcpcb * tp)3042 tcp_usrclosed(struct tcpcb *tp)
3043 {
3044 	switch (tp->t_state) {
3045 	case TCPS_CLOSED:
3046 	case TCPS_LISTEN:
3047 	case TCPS_SYN_SENT:
3048 		tp = tcp_close(tp);
3049 		break;
3050 
3051 	case TCPS_SYN_RECEIVED:
3052 		tp->t_flags |= TF_NEEDFIN;
3053 		break;
3054 
3055 	case TCPS_ESTABLISHED:
3056 		DTRACE_TCP4(state__change, void, NULL,
3057 		    struct inpcb *, tp->t_inpcb,
3058 		    struct tcpcb *, tp,
3059 		    int32_t, TCPS_FIN_WAIT_1);
3060 		tp->t_state = TCPS_FIN_WAIT_1;
3061 		TCP_LOG_CONNECTION_SUMMARY(tp);
3062 		break;
3063 
3064 	case TCPS_CLOSE_WAIT:
3065 		DTRACE_TCP4(state__change, void, NULL,
3066 		    struct inpcb *, tp->t_inpcb,
3067 		    struct tcpcb *, tp,
3068 		    int32_t, TCPS_LAST_ACK);
3069 		tp->t_state = TCPS_LAST_ACK;
3070 		TCP_LOG_CONNECTION_SUMMARY(tp);
3071 		break;
3072 	}
3073 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
3074 		soisdisconnected(tp->t_inpcb->inp_socket);
3075 		/* To prevent the connection hanging in FIN_WAIT_2 forever. */
3076 		if (tp->t_state == TCPS_FIN_WAIT_2) {
3077 			tcp_set_finwait_timeout(tp);
3078 		}
3079 	}
3080 	return tp;
3081 }
3082 
3083 void
tcp_in_cksum_stats(u_int32_t len)3084 tcp_in_cksum_stats(u_int32_t len)
3085 {
3086 	tcpstat.tcps_rcv_swcsum++;
3087 	tcpstat.tcps_rcv_swcsum_bytes += len;
3088 }
3089 
3090 void
tcp_out_cksum_stats(u_int32_t len)3091 tcp_out_cksum_stats(u_int32_t len)
3092 {
3093 	tcpstat.tcps_snd_swcsum++;
3094 	tcpstat.tcps_snd_swcsum_bytes += len;
3095 }
3096 
3097 void
tcp_in6_cksum_stats(u_int32_t len)3098 tcp_in6_cksum_stats(u_int32_t len)
3099 {
3100 	tcpstat.tcps_rcv6_swcsum++;
3101 	tcpstat.tcps_rcv6_swcsum_bytes += len;
3102 }
3103 
3104 void
tcp_out6_cksum_stats(u_int32_t len)3105 tcp_out6_cksum_stats(u_int32_t len)
3106 {
3107 	tcpstat.tcps_snd6_swcsum++;
3108 	tcpstat.tcps_snd6_swcsum_bytes += len;
3109 }
3110 
3111 int
tcp_get_mpkl_send_info(struct mbuf * control,struct so_mpkl_send_info * mpkl_send_info)3112 tcp_get_mpkl_send_info(struct mbuf *control,
3113     struct so_mpkl_send_info *mpkl_send_info)
3114 {
3115 	struct cmsghdr *cm;
3116 
3117 	if (control == NULL || mpkl_send_info == NULL) {
3118 		return EINVAL;
3119 	}
3120 
3121 	for (cm = M_FIRST_CMSGHDR(control); cm;
3122 	    cm = M_NXT_CMSGHDR(control, cm)) {
3123 		if (cm->cmsg_len < sizeof(struct cmsghdr) ||
3124 		    cm->cmsg_len > control->m_len) {
3125 			return EINVAL;
3126 		}
3127 		if (cm->cmsg_level != SOL_SOCKET ||
3128 		    cm->cmsg_type != SCM_MPKL_SEND_INFO) {
3129 			continue;
3130 		}
3131 		if (cm->cmsg_len != CMSG_LEN(sizeof(struct so_mpkl_send_info))) {
3132 			return EINVAL;
3133 		}
3134 		memcpy(mpkl_send_info, CMSG_DATA(cm),
3135 		    sizeof(struct so_mpkl_send_info));
3136 		return 0;
3137 	}
3138 	return ENOMSG;
3139 }
3140