]> git.saurik.com Git - apple/xnu.git/blob - bsd/netinet/tcp_usrreq.c
projects / apple / xnu.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
xnu-4570.31.3.tar.gz
[apple/xnu.git] / bsd / netinet / tcp_usrreq.c
1 /*
2 * Copyright (c) 2000-2017 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 #if INET6
71 #include <sys/domain.h>
72 #endif /* INET6 */
73 #if !CONFIG_EMBEDDED
74 #include <sys/kasl.h>
75 #endif
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/protosw.h>
79 #include <sys/syslog.h>
80
81 #include <net/if.h>
82 #include <net/route.h>
83 #include <net/ntstat.h>
84 #include <net/content_filter.h>
85
86 #include <netinet/in.h>
87 #include <netinet/in_systm.h>
88 #if INET6
89 #include <netinet/ip6.h>
90 #endif
91 #include <netinet/in_pcb.h>
92 #if INET6
93 #include <netinet6/in6_pcb.h>
94 #endif
95 #include <netinet/in_var.h>
96 #include <netinet/ip_var.h>
97 #if INET6
98 #include <netinet6/ip6_var.h>
99 #endif
100 #include <netinet/tcp.h>
101 #include <netinet/tcp_fsm.h>
102 #include <netinet/tcp_seq.h>
103 #include <netinet/tcp_timer.h>
104 #include <netinet/tcp_var.h>
105 #include <netinet/tcpip.h>
106 #include <netinet/tcp_cc.h>
107 #include <mach/sdt.h>
108 #if TCPDEBUG
109 #include <netinet/tcp_debug.h>
110 #endif
111 #if MPTCP
112 #include <netinet/mptcp_var.h>
113 #endif /* MPTCP */
114
115 #if IPSEC
116 #include <netinet6/ipsec.h>
117 #endif /*IPSEC*/
118
119 #if FLOW_DIVERT
120 #include <netinet/flow_divert.h>
121 #endif /* FLOW_DIVERT */
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
129 /*
130 * TCP protocol interface to socket abstraction.
131 */
132 extern char *tcpstates[]; /* XXX ??? */
133
134 static int tcp_attach(struct socket *, struct proc *);
135 static int tcp_connect(struct tcpcb *, struct sockaddr *, struct proc *);
136 #if INET6
137 static int tcp6_connect(struct tcpcb *, struct sockaddr *, struct proc *);
138 static int tcp6_usr_connect(struct socket *, struct sockaddr *,
139 struct proc *);
140 #endif /* INET6 */
141 static struct tcpcb *tcp_disconnect(struct tcpcb *);
142 static struct tcpcb *tcp_usrclosed(struct tcpcb *);
143 extern void tcp_sbrcv_trim(struct tcpcb *tp, struct sockbuf *sb);
144
145 #if TCPDEBUG
146 #define TCPDEBUG0 int ostate = 0
147 #define TCPDEBUG1() ostate = tp ? tp->t_state : 0
148 #define TCPDEBUG2(req) if (tp && (so->so_options & SO_DEBUG)) \
149 tcp_trace(TA_USER, ostate, tp, 0, 0, req)
150 #else
151 #define TCPDEBUG0
152 #define TCPDEBUG1()
153 #define TCPDEBUG2(req)
154 #endif
155
156 SYSCTL_PROC(_net_inet_tcp, OID_AUTO, info,
157 CTLFLAG_RW | CTLFLAG_LOCKED | CTLFLAG_ANYBODY | CTLFLAG_KERN,
158 0 , 0, tcp_sysctl_info, "S", "TCP info per tuple");
159
160 /*
161 * TCP attaches to socket via pru_attach(), reserving space,
162 * and an internet control block.
163 *
164 * Returns: 0 Success
165 * EISCONN
166 * tcp_attach:ENOBUFS
167 * tcp_attach:ENOMEM
168 * tcp_attach:??? [IPSEC specific]
169 */
170 static int
171 tcp_usr_attach(struct socket *so, __unused int proto, struct proc *p)
172 {
173 int error;
174 struct inpcb *inp = sotoinpcb(so);
175 struct tcpcb *tp = 0;
176 TCPDEBUG0;
177
178 TCPDEBUG1();
179 if (inp) {
180 error = EISCONN;
181 goto out;
182 }
183
184 error = tcp_attach(so, p);
185 if (error)
186 goto out;
187
188 if ((so->so_options & SO_LINGER) && so->so_linger == 0)
189 so->so_linger = TCP_LINGERTIME * hz;
190 tp = sototcpcb(so);
191 out:
192 TCPDEBUG2(PRU_ATTACH);
193 return error;
194 }
195
196 /*
197 * pru_detach() detaches the TCP protocol from the socket.
198 * If the protocol state is non-embryonic, then can't
199 * do this directly: have to initiate a pru_disconnect(),
200 * which may finish later; embryonic TCB's can just
201 * be discarded here.
202 */
203 static int
204 tcp_usr_detach(struct socket *so)
205 {
206 int error = 0;
207 struct inpcb *inp = sotoinpcb(so);
208 struct tcpcb *tp;
209 TCPDEBUG0;
210
211 if (inp == 0 || (inp->inp_state == INPCB_STATE_DEAD)) {
212 return EINVAL; /* XXX */
213 }
214 socket_lock_assert_owned(so);
215 tp = intotcpcb(inp);
216 /* In case we got disconnected from the peer */
217 if (tp == NULL)
218 goto out;
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() TCPDEBUG0; \
231 do { \
232 if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD) \
233 return (EINVAL); \
234 if (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() 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_END(req) out: TCPDEBUG2(req); return error; goto out
252
253
254 /*
255 * Give the socket an address.
256 *
257 * Returns: 0 Success
258 * EINVAL Invalid argument [COMMON_START]
259 * EAFNOSUPPORT Address family not supported
260 * in_pcbbind:EADDRNOTAVAIL Address not available.
261 * in_pcbbind:EINVAL Invalid argument
262 * in_pcbbind:EAFNOSUPPORT Address family not supported [notdef]
263 * in_pcbbind:EACCES Permission denied
264 * in_pcbbind:EADDRINUSE Address in use
265 * in_pcbbind:EAGAIN Resource unavailable, try again
266 * in_pcbbind:EPERM Operation not permitted
267 */
268 static int
269 tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
270 {
271 int error = 0;
272 struct inpcb *inp = sotoinpcb(so);
273 struct tcpcb *tp;
274 struct sockaddr_in *sinp;
275
276 COMMON_START();
277
278 if (nam->sa_family != 0 && nam->sa_family != AF_INET) {
279 error = EAFNOSUPPORT;
280 goto out;
281 }
282
283 /*
284 * Must check for multicast addresses and disallow binding
285 * to them.
286 */
287 sinp = (struct sockaddr_in *)(void *)nam;
288 if (sinp->sin_family == AF_INET &&
289 IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
290 error = EAFNOSUPPORT;
291 goto out;
292 }
293 error = in_pcbbind(inp, nam, p);
294 if (error)
295 goto out;
296
297 #if NECP
298 /* Update NECP client with bind result if not in middle of connect */
299 if ((inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS) &&
300 !uuid_is_null(inp->necp_client_uuid)) {
301 socket_unlock(so, 0);
302 necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
303 socket_lock(so, 0);
304 }
305 #endif /* NECP */
306
307 COMMON_END(PRU_BIND);
308
309 }
310
311 #if INET6
312 static int
313 tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
314 {
315 int error = 0;
316 struct inpcb *inp = sotoinpcb(so);
317 struct tcpcb *tp;
318 struct sockaddr_in6 *sin6p;
319
320 COMMON_START();
321
322 if (nam->sa_family != 0 && nam->sa_family != AF_INET6) {
323 error = EAFNOSUPPORT;
324 goto out;
325 }
326
327 /*
328 * Must check for multicast addresses and disallow binding
329 * to them.
330 */
331 sin6p = (struct sockaddr_in6 *)(void *)nam;
332 if (sin6p->sin6_family == AF_INET6 &&
333 IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
334 error = EAFNOSUPPORT;
335 goto out;
336 }
337 inp->inp_vflag &= ~INP_IPV4;
338 inp->inp_vflag |= INP_IPV6;
339 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) {
340 if (IN6_IS_ADDR_UNSPECIFIED(&sin6p->sin6_addr))
341 inp->inp_vflag |= INP_IPV4;
342 else if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
343 struct sockaddr_in sin;
344
345 in6_sin6_2_sin(&sin, sin6p);
346 inp->inp_vflag |= INP_IPV4;
347 inp->inp_vflag &= ~INP_IPV6;
348 error = in_pcbbind(inp, (struct sockaddr *)&sin, p);
349 goto out;
350 }
351 }
352 error = in6_pcbbind(inp, nam, p);
353 if (error)
354 goto out;
355 COMMON_END(PRU_BIND);
356 }
357 #endif /* INET6 */
358
359 /*
360 * Prepare to accept connections.
361 *
362 * Returns: 0 Success
363 * EINVAL [COMMON_START]
364 * in_pcbbind:EADDRNOTAVAIL Address not available.
365 * in_pcbbind:EINVAL Invalid argument
366 * in_pcbbind:EAFNOSUPPORT Address family not supported [notdef]
367 * in_pcbbind:EACCES Permission denied
368 * in_pcbbind:EADDRINUSE Address in use
369 * in_pcbbind:EAGAIN Resource unavailable, try again
370 * in_pcbbind:EPERM Operation not permitted
371 */
372 static int
373 tcp_usr_listen(struct socket *so, struct proc *p)
374 {
375 int error = 0;
376 struct inpcb *inp = sotoinpcb(so);
377 struct tcpcb *tp;
378
379 COMMON_START();
380 if (inp->inp_lport == 0)
381 error = in_pcbbind(inp, NULL, p);
382 if (error == 0)
383 tp->t_state = TCPS_LISTEN;
384 COMMON_END(PRU_LISTEN);
385 }
386
387 #if INET6
388 static int
389 tcp6_usr_listen(struct socket *so, struct proc *p)
390 {
391 int error = 0;
392 struct inpcb *inp = sotoinpcb(so);
393 struct tcpcb *tp;
394
395 COMMON_START();
396 if (inp->inp_lport == 0) {
397 inp->inp_vflag &= ~INP_IPV4;
398 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0)
399 inp->inp_vflag |= INP_IPV4;
400 error = in6_pcbbind(inp, NULL, p);
401 }
402 if (error == 0)
403 tp->t_state = TCPS_LISTEN;
404 COMMON_END(PRU_LISTEN);
405 }
406 #endif /* INET6 */
407
408 static int
409 tcp_connect_complete(struct socket *so)
410 {
411 struct tcpcb *tp = sototcpcb(so);
412 struct inpcb *inp = sotoinpcb(so);
413 int error = 0;
414
415 #if NECP
416 /* Update NECP client with connected five-tuple */
417 if (!uuid_is_null(inp->necp_client_uuid)) {
418 socket_unlock(so, 0);
419 necp_client_assign_from_socket(so->last_pid, inp->necp_client_uuid, inp);
420 socket_lock(so, 0);
421 }
422 #endif /* NECP */
423
424 /* TFO delays the tcp_output until later, when the app calls write() */
425 if (so->so_flags1 & SOF1_PRECONNECT_DATA) {
426 if (!necp_socket_is_allowed_to_send_recv(sotoinpcb(so), NULL, NULL))
427 return (EHOSTUNREACH);
428
429 /* Initialize enough state so that we can actually send data */
430 tcp_mss(tp, -1, IFSCOPE_NONE);
431 tp->snd_wnd = tp->t_maxseg;
432 } else {
433 error = tcp_output(tp);
434 }
435
436 return (error);
437 }
438
439 /*
440 * Initiate connection to peer.
441 * Create a template for use in transmissions on this connection.
442 * Enter SYN_SENT state, and mark socket as connecting.
443 * Start keep-alive timer, and seed output sequence space.
444 * Send initial segment on connection.
445 */
446 static int
447 tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
448 {
449 int error = 0;
450 struct inpcb *inp = sotoinpcb(so);
451 struct tcpcb *tp;
452 struct sockaddr_in *sinp;
453
454 TCPDEBUG0;
455 if (inp == NULL) {
456 return EINVAL;
457 } else if (inp->inp_state == INPCB_STATE_DEAD) {
458 if (so->so_error) {
459 error = so->so_error;
460 so->so_error = 0;
461 return error;
462 } else
463 return EINVAL;
464 }
465 #if NECP
466 #if FLOW_DIVERT
467 else if (necp_socket_should_use_flow_divert(inp)) {
468 uint32_t fd_ctl_unit = necp_socket_get_flow_divert_control_unit(inp);
469 if (fd_ctl_unit > 0) {
470 error = flow_divert_pcb_init(so, fd_ctl_unit);
471 if (error == 0) {
472 error = flow_divert_connect_out(so, nam, p);
473 }
474 } else {
475 error = ENETDOWN;
476 }
477
478 return error;
479 }
480 #endif /* FLOW_DIVERT */
481 #if CONTENT_FILTER
482 error = cfil_sock_attach(so);
483 if (error != 0)
484 return error;
485 #endif /* CONTENT_FILTER */
486 #endif /* NECP */
487 tp = intotcpcb(inp);
488 TCPDEBUG1();
489
490 calculate_tcp_clock();
491
492 if (nam->sa_family != 0 && nam->sa_family != AF_INET) {
493 error = EAFNOSUPPORT;
494 goto out;
495 }
496 /*
497 * Must disallow TCP ``connections'' to multicast addresses.
498 */
499 sinp = (struct sockaddr_in *)(void *)nam;
500 if (sinp->sin_family == AF_INET
501 && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
502 error = EAFNOSUPPORT;
503 goto out;
504 }
505
506 if ((error = tcp_connect(tp, nam, p)) != 0)
507 goto out;
508
509 error = tcp_connect_complete(so);
510
511 COMMON_END(PRU_CONNECT);
512 }
513
514 static int
515 tcp_usr_connectx_common(struct socket *so, int af,
516 struct sockaddr *src, struct sockaddr *dst,
517 struct proc *p, uint32_t ifscope, sae_associd_t aid, sae_connid_t *pcid,
518 uint32_t flags, void *arg, uint32_t arglen, struct uio *auio,
519 user_ssize_t *bytes_written)
520 {
521 #pragma unused(aid, flags, arg, arglen)
522 struct inpcb *inp = sotoinpcb(so);
523 int error = 0;
524 user_ssize_t datalen = 0;
525
526 if (inp == NULL)
527 return (EINVAL);
528
529 VERIFY(dst != NULL);
530
531 ASSERT(!(inp->inp_flags2 & INP2_CONNECT_IN_PROGRESS));
532 inp->inp_flags2 |= INP2_CONNECT_IN_PROGRESS;
533
534 #if NECP
535 inp_update_necp_policy(inp, src, dst, ifscope);
536 #endif /* NECP */
537
538 if ((so->so_flags1 & SOF1_DATA_IDEMPOTENT) &&
539 (tcp_fastopen & TCP_FASTOPEN_CLIENT))
540 sototcpcb(so)->t_flagsext |= TF_FASTOPEN;
541
542 /* bind socket to the specified interface, if requested */
543 if (ifscope != IFSCOPE_NONE &&
544 (error = inp_bindif(inp, ifscope, NULL)) != 0) {
545 goto done;
546 }
547
548 /* if source address and/or port is specified, bind to it */
549 if (src != NULL) {
550 error = sobindlock(so, src, 0); /* already locked */
551 if (error != 0) {
552 goto done;
553 }
554 }
555
556 switch (af) {
557 case AF_INET:
558 error = tcp_usr_connect(so, dst, p);
559 break;
560 #if INET6
561 case AF_INET6:
562 error = tcp6_usr_connect(so, dst, p);
563 break;
564 #endif /* INET6 */
565 default:
566 VERIFY(0);
567 /* NOTREACHED */
568 }
569
570 if (error != 0) {
571 goto done;
572 }
573
574 /* if there is data, copy it */
575 if (auio != NULL) {
576 socket_unlock(so, 0);
577
578 VERIFY(bytes_written != NULL);
579
580 datalen = uio_resid(auio);
581 error = so->so_proto->pr_usrreqs->pru_sosend(so, NULL,
582 (uio_t)auio, NULL, NULL, 0);
583 socket_lock(so, 0);
584
585 if (error == 0 || error == EWOULDBLOCK)
586 *bytes_written = datalen - uio_resid(auio);
587
588 /*
589 * sosend returns EWOULDBLOCK if it's a non-blocking
590 * socket or a timeout occured (this allows to return
591 * the amount of queued data through sendit()).
592 *
593 * However, connectx() returns EINPROGRESS in case of a
594 * blocking socket. So we change the return value here.
595 */
596 if (error == EWOULDBLOCK)
597 error = EINPROGRESS;
598 }
599
600 if (error == 0 && pcid != NULL)
601 *pcid = 1; /* there is only one connection in regular TCP */
602
603 done:
604 inp->inp_flags2 &= ~INP2_CONNECT_IN_PROGRESS;
605 return (error);
606 }
607
608 static int
609 tcp_usr_connectx(struct socket *so, struct sockaddr *src,
610 struct sockaddr *dst, struct proc *p, uint32_t ifscope,
611 sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
612 uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written)
613 {
614 return (tcp_usr_connectx_common(so, AF_INET, src, dst, p, ifscope, aid,
615 pcid, flags, arg, arglen, uio, bytes_written));
616 }
617
618 #if INET6
619 static int
620 tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
621 {
622 int error = 0;
623 struct inpcb *inp = sotoinpcb(so);
624 struct tcpcb *tp;
625 struct sockaddr_in6 *sin6p;
626
627 TCPDEBUG0;
628 if (inp == NULL) {
629 return EINVAL;
630 } else if (inp->inp_state == INPCB_STATE_DEAD) {
631 if (so->so_error) {
632 error = so->so_error;
633 so->so_error = 0;
634 return error;
635 } else
636 return EINVAL;
637 }
638 #if NECP
639 #if FLOW_DIVERT
640 else if (necp_socket_should_use_flow_divert(inp)) {
641 uint32_t fd_ctl_unit = necp_socket_get_flow_divert_control_unit(inp);
642 if (fd_ctl_unit > 0) {
643 error = flow_divert_pcb_init(so, fd_ctl_unit);
644 if (error == 0) {
645 error = flow_divert_connect_out(so, nam, p);
646 }
647 } else {
648 error = ENETDOWN;
649 }
650
651 return error;
652 }
653 #endif /* FLOW_DIVERT */
654 #if CONTENT_FILTER
655 error = cfil_sock_attach(so);
656 if (error != 0)
657 return error;
658 #endif /* CONTENT_FILTER */
659 #endif /* NECP */
660
661 tp = intotcpcb(inp);
662 TCPDEBUG1();
663
664 calculate_tcp_clock();
665
666 if (nam->sa_family != 0 && nam->sa_family != AF_INET6) {
667 error = EAFNOSUPPORT;
668 goto out;
669 }
670
671 /*
672 * Must disallow TCP ``connections'' to multicast addresses.
673 */
674 sin6p = (struct sockaddr_in6 *)(void *)nam;
675 if (sin6p->sin6_family == AF_INET6
676 && IN6_IS_ADDR_MULTICAST(&sin6p->sin6_addr)) {
677 error = EAFNOSUPPORT;
678 goto out;
679 }
680
681 if (IN6_IS_ADDR_V4MAPPED(&sin6p->sin6_addr)) {
682 struct sockaddr_in sin;
683
684 if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)
685 return (EINVAL);
686
687 in6_sin6_2_sin(&sin, sin6p);
688 inp->inp_vflag |= INP_IPV4;
689 inp->inp_vflag &= ~INP_IPV6;
690 if ((error = tcp_connect(tp, (struct sockaddr *)&sin, p)) != 0)
691 goto out;
692
693 error = tcp_connect_complete(so);
694 goto out;
695 }
696 inp->inp_vflag &= ~INP_IPV4;
697 inp->inp_vflag |= INP_IPV6;
698 if ((error = tcp6_connect(tp, nam, p)) != 0)
699 goto out;
700
701 error = tcp_connect_complete(so);
702 COMMON_END(PRU_CONNECT);
703 }
704
705 static int
706 tcp6_usr_connectx(struct socket *so, struct sockaddr*src,
707 struct sockaddr *dst, struct proc *p, uint32_t ifscope,
708 sae_associd_t aid, sae_connid_t *pcid, uint32_t flags, void *arg,
709 uint32_t arglen, struct uio *uio, user_ssize_t *bytes_written)
710 {
711 return (tcp_usr_connectx_common(so, AF_INET6, src, dst, p, ifscope, aid,
712 pcid, flags, arg, arglen, uio, bytes_written));
713 }
714 #endif /* INET6 */
715
716 /*
717 * Initiate disconnect from peer.
718 * If connection never passed embryonic stage, just drop;
719 * else if don't need to let data drain, then can just drop anyways,
720 * else have to begin TCP shutdown process: mark socket disconnecting,
721 * drain unread data, state switch to reflect user close, and
722 * send segment (e.g. FIN) to peer. Socket will be really disconnected
723 * when peer sends FIN and acks ours.
724 *
725 * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
726 */
727 static int
728 tcp_usr_disconnect(struct socket *so)
729 {
730 int error = 0;
731 struct inpcb *inp = sotoinpcb(so);
732 struct tcpcb *tp;
733
734 socket_lock_assert_owned(so);
735 COMMON_START();
736 /* In case we got disconnected from the peer */
737 if (tp == NULL)
738 goto out;
739 tp = tcp_disconnect(tp);
740 COMMON_END(PRU_DISCONNECT);
741 }
742
743 /*
744 * User-protocol pru_disconnectx callback.
745 */
746 static int
747 tcp_usr_disconnectx(struct socket *so, sae_associd_t aid, sae_connid_t cid)
748 {
749 #pragma unused(cid)
750 if (aid != SAE_ASSOCID_ANY && aid != SAE_ASSOCID_ALL)
751 return (EINVAL);
752
753 return (tcp_usr_disconnect(so));
754 }
755
756 /*
757 * Accept a connection. Essentially all the work is
758 * done at higher levels; just return the address
759 * of the peer, storing through addr.
760 */
761 static int
762 tcp_usr_accept(struct socket *so, struct sockaddr **nam)
763 {
764 int error = 0;
765 struct inpcb *inp = sotoinpcb(so);
766 struct tcpcb *tp = NULL;
767 TCPDEBUG0;
768
769 in_getpeeraddr(so, nam);
770
771 if (so->so_state & SS_ISDISCONNECTED) {
772 error = ECONNABORTED;
773 goto out;
774 }
775 if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD)
776 return (EINVAL);
777 #if NECP
778 else if (necp_socket_should_use_flow_divert(inp))
779 return (EPROTOTYPE);
780 #if CONTENT_FILTER
781 error = cfil_sock_attach(so);
782 if (error != 0)
783 return (error);
784 #endif /* CONTENT_FILTER */
785 #endif /* NECP */
786
787 tp = intotcpcb(inp);
788 TCPDEBUG1();
789
790 calculate_tcp_clock();
791
792 COMMON_END(PRU_ACCEPT);
793 }
794
795 #if INET6
796 static int
797 tcp6_usr_accept(struct socket *so, struct sockaddr **nam)
798 {
799 int error = 0;
800 struct inpcb *inp = sotoinpcb(so);
801 struct tcpcb *tp = NULL;
802 TCPDEBUG0;
803
804 if (so->so_state & SS_ISDISCONNECTED) {
805 error = ECONNABORTED;
806 goto out;
807 }
808 if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD)
809 return (EINVAL);
810 #if NECP
811 else if (necp_socket_should_use_flow_divert(inp))
812 return (EPROTOTYPE);
813 #if CONTENT_FILTER
814 error = cfil_sock_attach(so);
815 if (error != 0)
816 return (error);
817 #endif /* CONTENT_FILTER */
818 #endif /* NECP */
819
820 tp = intotcpcb(inp);
821 TCPDEBUG1();
822
823 calculate_tcp_clock();
824
825 in6_mapped_peeraddr(so, nam);
826 COMMON_END(PRU_ACCEPT);
827 }
828 #endif /* INET6 */
829
830 /*
831 * Mark the connection as being incapable of further output.
832 *
833 * Returns: 0 Success
834 * EINVAL [COMMON_START]
835 * tcp_output:EADDRNOTAVAIL
836 * tcp_output:ENOBUFS
837 * tcp_output:EMSGSIZE
838 * tcp_output:EHOSTUNREACH
839 * tcp_output:ENETUNREACH
840 * tcp_output:ENETDOWN
841 * tcp_output:ENOMEM
842 * tcp_output:EACCES
843 * tcp_output:EMSGSIZE
844 * tcp_output:ENOBUFS
845 * tcp_output:??? [ignorable: mostly IPSEC/firewall/DLIL]
846 */
847 static int
848 tcp_usr_shutdown(struct socket *so)
849 {
850 int error = 0;
851 struct inpcb *inp = sotoinpcb(so);
852 struct tcpcb *tp;
853
854 TCPDEBUG0;
855 if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD)
856 return (EINVAL);
857
858 socantsendmore(so);
859
860 /*
861 * In case we got disconnected from the peer, or if this is
862 * a socket that is to be flow-diverted (but not yet).
863 */
864 tp = intotcpcb(inp);
865 TCPDEBUG1();
866
867 if (tp == NULL
868 #if NECP
869 || (necp_socket_should_use_flow_divert(inp))
870 #endif /* NECP */
871 ) {
872 if (tp != NULL)
873 error = EPROTOTYPE;
874 goto out;
875 }
876
877 calculate_tcp_clock();
878
879 tp = tcp_usrclosed(tp);
880 #if MPTCP
881 /* A reset has been sent but socket exists, do not send FIN */
882 if ((so->so_flags & SOF_MP_SUBFLOW) &&
883 (tp) && (tp->t_mpflags & TMPF_RESET)) {
884 goto out;
885 }
886 #endif
887 #if CONTENT_FILTER
888 /* Don't send a FIN yet */
889 if (tp && !(so->so_state & SS_ISDISCONNECTED) &&
890 cfil_sock_data_pending(&so->so_snd))
891 goto out;
892 #endif /* CONTENT_FILTER */
893 if (tp)
894 error = tcp_output(tp);
895 COMMON_END(PRU_SHUTDOWN);
896 }
897
898 /*
899 * After a receive, possibly send window update to peer.
900 */
901 static int
902 tcp_usr_rcvd(struct socket *so, __unused int flags)
903 {
904 int error = 0;
905 struct inpcb *inp = sotoinpcb(so);
906 struct tcpcb *tp;
907
908 COMMON_START();
909 /* In case we got disconnected from the peer */
910 if (tp == NULL)
911 goto out;
912 tcp_sbrcv_trim(tp, &so->so_rcv);
913
914 /*
915 * This tcp_output is solely there to trigger window-updates.
916 * However, we really do not want these window-updates while we
917 * are still in SYN_SENT or SYN_RECEIVED.
918 */
919 if (TCPS_HAVEESTABLISHED(tp->t_state))
920 tcp_output(tp);
921
922 #if CONTENT_FILTER
923 cfil_sock_buf_update(&so->so_rcv);
924 #endif /* CONTENT_FILTER */
925
926 COMMON_END(PRU_RCVD);
927 }
928
929 /*
930 * Do a send by putting data in output queue and updating urgent
931 * marker if URG set. Possibly send more data. Unlike the other
932 * pru_*() routines, the mbuf chains are our responsibility. We
933 * must either enqueue them or free them. The other pru_* routines
934 * generally are caller-frees.
935 *
936 * Returns: 0 Success
937 * ECONNRESET
938 * EINVAL
939 * ENOBUFS
940 * tcp_connect:EADDRINUSE Address in use
941 * tcp_connect:EADDRNOTAVAIL Address not available.
942 * tcp_connect:EINVAL Invalid argument
943 * tcp_connect:EAFNOSUPPORT Address family not supported [notdef]
944 * tcp_connect:EACCES Permission denied
945 * tcp_connect:EAGAIN Resource unavailable, try again
946 * tcp_connect:EPERM Operation not permitted
947 * tcp_output:EADDRNOTAVAIL
948 * tcp_output:ENOBUFS
949 * tcp_output:EMSGSIZE
950 * tcp_output:EHOSTUNREACH
951 * tcp_output:ENETUNREACH
952 * tcp_output:ENETDOWN
953 * tcp_output:ENOMEM
954 * tcp_output:EACCES
955 * tcp_output:EMSGSIZE
956 * tcp_output:ENOBUFS
957 * tcp_output:??? [ignorable: mostly IPSEC/firewall/DLIL]
958 * tcp6_connect:??? [IPV6 only]
959 */
960 static int
961 tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
962 struct sockaddr *nam, struct mbuf *control, struct proc *p)
963 {
964 int error = 0;
965 struct inpcb *inp = sotoinpcb(so);
966 struct tcpcb *tp;
967 uint32_t msgpri = MSG_PRI_DEFAULT;
968 #if INET6
969 int isipv6;
970 #endif
971 TCPDEBUG0;
972
973 if (inp == NULL || inp->inp_state == INPCB_STATE_DEAD
974 #if NECP
975 || (necp_socket_should_use_flow_divert(inp))
976 #endif /* NECP */
977 ) {
978 /*
979 * OOPS! we lost a race, the TCP session got reset after
980 * we checked SS_CANTSENDMORE, eg: while doing uiomove or a
981 * network interrupt in the non-splnet() section of sosend().
982 */
983 if (m != NULL)
984 m_freem(m);
985 if (control != NULL) {
986 m_freem(control);
987 control = NULL;
988 }
989
990 if (inp == NULL)
991 error = ECONNRESET; /* XXX EPIPE? */
992 else
993 error = EPROTOTYPE;
994 tp = NULL;
995 TCPDEBUG1();
996 goto out;
997 }
998 #if INET6
999 isipv6 = nam && nam->sa_family == AF_INET6;
1000 #endif /* INET6 */
1001 tp = intotcpcb(inp);
1002 TCPDEBUG1();
1003
1004 calculate_tcp_clock();
1005
1006 if (control != NULL) {
1007 if (so->so_flags & SOF_ENABLE_MSGS) {
1008 /* Get the msg priority from control mbufs */
1009 error = tcp_get_msg_priority(control, &msgpri);
1010 if (error) {
1011 m_freem(control);
1012 if (m != NULL)
1013 m_freem(m);
1014 control = NULL;
1015 m = NULL;
1016 goto out;
1017 }
1018 m_freem(control);
1019 control = NULL;
1020 } else if (control->m_len) {
1021 /*
1022 * if not unordered, TCP should not have
1023 * control mbufs
1024 */
1025 m_freem(control);
1026 if (m != NULL)
1027 m_freem(m);
1028 control = NULL;
1029 m = NULL;
1030 error = EINVAL;
1031 goto out;
1032 }
1033 }
1034
1035 if (so->so_flags & SOF_ENABLE_MSGS) {
1036 VERIFY(m->m_flags & M_PKTHDR);
1037 m->m_pkthdr.msg_pri = msgpri;
1038 }
1039
1040 /* MPTCP sublow socket buffers must not be compressed */
1041 VERIFY(!(so->so_flags & SOF_MP_SUBFLOW) ||
1042 (so->so_snd.sb_flags & SB_NOCOMPRESS));
1043
1044 if(!(flags & PRUS_OOB) || (so->so_flags1 & SOF1_PRECONNECT_DATA)) {
1045 /* Call msg send if message delivery is enabled */
1046 if (so->so_flags & SOF_ENABLE_MSGS)
1047 sbappendmsg_snd(&so->so_snd, m);
1048 else
1049 sbappendstream(&so->so_snd, m);
1050
1051 if (nam && tp->t_state < TCPS_SYN_SENT) {
1052
1053 /*
1054 * Do implied connect if not yet connected,
1055 * initialize window to default value, and
1056 * initialize maxseg/maxopd using peer's cached
1057 * MSS.
1058 */
1059 #if INET6
1060 if (isipv6)
1061 error = tcp6_connect(tp, nam, p);
1062 else
1063 #endif /* INET6 */
1064 error = tcp_connect(tp, nam, p);
1065 if (error)
1066 goto out;
1067 tp->snd_wnd = TTCP_CLIENT_SND_WND;
1068 tcp_mss(tp, -1, IFSCOPE_NONE);
1069 }
1070
1071 if (flags & PRUS_EOF) {
1072 /*
1073 * Close the send side of the connection after
1074 * the data is sent.
1075 */
1076 socantsendmore(so);
1077 tp = tcp_usrclosed(tp);
1078 }
1079 if (tp != NULL) {
1080 if (flags & PRUS_MORETOCOME)
1081 tp->t_flags |= TF_MORETOCOME;
1082 error = tcp_output(tp);
1083 if (flags & PRUS_MORETOCOME)
1084 tp->t_flags &= ~TF_MORETOCOME;
1085 }
1086 } else {
1087 if (sbspace(&so->so_snd) == 0) {
1088 /* if no space is left in sockbuf,
1089 * do not try to squeeze in OOB traffic */
1090 m_freem(m);
1091 error = ENOBUFS;
1092 goto out;
1093 }
1094 /*
1095 * According to RFC961 (Assigned Protocols),
1096 * the urgent pointer points to the last octet
1097 * of urgent data. We continue, however,
1098 * to consider it to indicate the first octet
1099 * of data past the urgent section.
1100 * Otherwise, snd_up should be one lower.
1101 */
1102 sbappendstream(&so->so_snd, m);
1103 if (nam && tp->t_state < TCPS_SYN_SENT) {
1104 /*
1105 * Do implied connect if not yet connected,
1106 * initialize window to default value, and
1107 * initialize maxseg/maxopd using peer's cached
1108 * MSS.
1109 */
1110 #if INET6
1111 if (isipv6)
1112 error = tcp6_connect(tp, nam, p);
1113 else
1114 #endif /* INET6 */
1115 error = tcp_connect(tp, nam, p);
1116 if (error)
1117 goto out;
1118 tp->snd_wnd = TTCP_CLIENT_SND_WND;
1119 tcp_mss(tp, -1, IFSCOPE_NONE);
1120 }
1121 tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
1122 tp->t_flagsext |= TF_FORCE;
1123 error = tcp_output(tp);
1124 tp->t_flagsext &= ~TF_FORCE;
1125 }
1126
1127
1128 /*
1129 * We wait for the socket to successfully connect before returning.
1130 * This allows us to signal a timeout to the application.
1131 */
1132 if (so->so_state & SS_ISCONNECTING) {
1133 if (so->so_state & SS_NBIO)
1134 error = EWOULDBLOCK;
1135 else
1136 error = sbwait(&so->so_snd);
1137 }
1138
1139 COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
1140 ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
1141 }
1142
1143 /*
1144 * Abort the TCP.
1145 */
1146 static int
1147 tcp_usr_abort(struct socket *so)
1148 {
1149 int error = 0;
1150 struct inpcb *inp = sotoinpcb(so);
1151 struct tcpcb *tp;
1152
1153 COMMON_START();
1154 /* In case we got disconnected from the peer */
1155 if (tp == NULL)
1156 goto out;
1157 tp = tcp_drop(tp, ECONNABORTED);
1158 VERIFY(so->so_usecount > 0);
1159 so->so_usecount--;
1160 COMMON_END(PRU_ABORT);
1161 }
1162
1163 /*
1164 * Receive out-of-band data.
1165 *
1166 * Returns: 0 Success
1167 * EINVAL [COMMON_START]
1168 * EINVAL
1169 * EWOULDBLOCK
1170 */
1171 static int
1172 tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
1173 {
1174 int error = 0;
1175 struct inpcb *inp = sotoinpcb(so);
1176 struct tcpcb *tp;
1177
1178 COMMON_START();
1179 if ((so->so_oobmark == 0 &&
1180 (so->so_state & SS_RCVATMARK) == 0) ||
1181 so->so_options & SO_OOBINLINE ||
1182 tp->t_oobflags & TCPOOB_HADDATA) {
1183 error = EINVAL;
1184 goto out;
1185 }
1186 if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
1187 error = EWOULDBLOCK;
1188 goto out;
1189 }
1190 m->m_len = 1;
1191 *mtod(m, caddr_t) = tp->t_iobc;
1192 so->so_state &= ~SS_RCVATMARK;
1193 if ((flags & MSG_PEEK) == 0)
1194 tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
1195 COMMON_END(PRU_RCVOOB);
1196 }
1197
1198 static int
1199 tcp_usr_preconnect(struct socket *so)
1200 {
1201 struct inpcb *inp = sotoinpcb(so);
1202 int error = 0;
1203
1204 #if NECP
1205 if (necp_socket_should_use_flow_divert(inp)) {
1206 /* May happen, if in tcp_usr_connect we did not had a chance
1207 * to set the usrreqs (due to some error). So, let's get out
1208 * of here.
1209 */
1210 goto out;
1211 }
1212 #endif /* NECP */
1213
1214 error = tcp_output(sototcpcb(so));
1215
1216 soclearfastopen(so);
1217
1218 COMMON_END(PRU_PRECONNECT);
1219 }
1220
1221 /* xxx - should be const */
1222 struct pr_usrreqs tcp_usrreqs = {
1223 .pru_abort = tcp_usr_abort,
1224 .pru_accept = tcp_usr_accept,
1225 .pru_attach = tcp_usr_attach,
1226 .pru_bind = tcp_usr_bind,
1227 .pru_connect = tcp_usr_connect,
1228 .pru_connectx = tcp_usr_connectx,
1229 .pru_control = in_control,
1230 .pru_detach = tcp_usr_detach,
1231 .pru_disconnect = tcp_usr_disconnect,
1232 .pru_disconnectx = tcp_usr_disconnectx,
1233 .pru_listen = tcp_usr_listen,
1234 .pru_peeraddr = in_getpeeraddr,
1235 .pru_rcvd = tcp_usr_rcvd,
1236 .pru_rcvoob = tcp_usr_rcvoob,
1237 .pru_send = tcp_usr_send,
1238 .pru_shutdown = tcp_usr_shutdown,
1239 .pru_sockaddr = in_getsockaddr,
1240 .pru_sosend = sosend,
1241 .pru_soreceive = soreceive,
1242 .pru_preconnect = tcp_usr_preconnect,
1243 };
1244
1245 #if INET6
1246 struct pr_usrreqs tcp6_usrreqs = {
1247 .pru_abort = tcp_usr_abort,
1248 .pru_accept = tcp6_usr_accept,
1249 .pru_attach = tcp_usr_attach,
1250 .pru_bind = tcp6_usr_bind,
1251 .pru_connect = tcp6_usr_connect,
1252 .pru_connectx = tcp6_usr_connectx,
1253 .pru_control = in6_control,
1254 .pru_detach = tcp_usr_detach,
1255 .pru_disconnect = tcp_usr_disconnect,
1256 .pru_disconnectx = tcp_usr_disconnectx,
1257 .pru_listen = tcp6_usr_listen,
1258 .pru_peeraddr = in6_mapped_peeraddr,
1259 .pru_rcvd = tcp_usr_rcvd,
1260 .pru_rcvoob = tcp_usr_rcvoob,
1261 .pru_send = tcp_usr_send,
1262 .pru_shutdown = tcp_usr_shutdown,
1263 .pru_sockaddr = in6_mapped_sockaddr,
1264 .pru_sosend = sosend,
1265 .pru_soreceive = soreceive,
1266 .pru_preconnect = tcp_usr_preconnect,
1267 };
1268 #endif /* INET6 */
1269
1270 /*
1271 * Common subroutine to open a TCP connection to remote host specified
1272 * by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
1273 * port number if needed. Call in_pcbladdr to do the routing and to choose
1274 * a local host address (interface). If there is an existing incarnation
1275 * of the same connection in TIME-WAIT state and if the remote host was
1276 * sending CC options and if the connection duration was < MSL, then
1277 * truncate the previous TIME-WAIT state and proceed.
1278 * Initialize connection parameters and enter SYN-SENT state.
1279 *
1280 * Returns: 0 Success
1281 * EADDRINUSE
1282 * EINVAL
1283 * in_pcbbind:EADDRNOTAVAIL Address not available.
1284 * in_pcbbind:EINVAL Invalid argument
1285 * in_pcbbind:EAFNOSUPPORT Address family not supported [notdef]
1286 * in_pcbbind:EACCES Permission denied
1287 * in_pcbbind:EADDRINUSE Address in use
1288 * in_pcbbind:EAGAIN Resource unavailable, try again
1289 * in_pcbbind:EPERM Operation not permitted
1290 * in_pcbladdr:EINVAL Invalid argument
1291 * in_pcbladdr:EAFNOSUPPORT Address family not supported
1292 * in_pcbladdr:EADDRNOTAVAIL Address not available
1293 */
1294 static int
1295 tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct proc *p)
1296 {
1297 struct inpcb *inp = tp->t_inpcb, *oinp;
1298 struct socket *so = inp->inp_socket;
1299 struct tcpcb *otp;
1300 struct sockaddr_in *sin = (struct sockaddr_in *)(void *)nam;
1301 struct in_addr laddr;
1302 int error = 0;
1303 struct ifnet *outif = NULL;
1304
1305 if (inp->inp_lport == 0) {
1306 error = in_pcbbind(inp, NULL, p);
1307 if (error)
1308 goto done;
1309 }
1310
1311 /*
1312 * Cannot simply call in_pcbconnect, because there might be an
1313 * earlier incarnation of this same connection still in
1314 * TIME_WAIT state, creating an ADDRINUSE error.
1315 */
1316 error = in_pcbladdr(inp, nam, &laddr, IFSCOPE_NONE, &outif, 0);
1317 if (error)
1318 goto done;
1319
1320 socket_unlock(inp->inp_socket, 0);
1321 oinp = in_pcblookup_hash(inp->inp_pcbinfo,
1322 sin->sin_addr, sin->sin_port,
1323 inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr : laddr,
1324 inp->inp_lport, 0, NULL);
1325
1326 socket_lock(inp->inp_socket, 0);
1327 if (oinp) {
1328 if (oinp != inp) /* 4143933: avoid deadlock if inp == oinp */
1329 socket_lock(oinp->inp_socket, 1);
1330 if (in_pcb_checkstate(oinp, WNT_RELEASE, 1) == WNT_STOPUSING) {
1331 if (oinp != inp)
1332 socket_unlock(oinp->inp_socket, 1);
1333 goto skip_oinp;
1334 }
1335
1336 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
1337 otp->t_state == TCPS_TIME_WAIT &&
1338 ((int)(tcp_now - otp->t_starttime)) < tcp_msl &&
1339 (otp->t_flags & TF_RCVD_CC)) {
1340 otp = tcp_close(otp);
1341 } else {
1342 printf("tcp_connect: inp=0x%llx err=EADDRINUSE\n",
1343 (uint64_t)VM_KERNEL_ADDRPERM(inp));
1344 if (oinp != inp)
1345 socket_unlock(oinp->inp_socket, 1);
1346 error = EADDRINUSE;
1347 goto done;
1348 }
1349 if (oinp != inp)
1350 socket_unlock(oinp->inp_socket, 1);
1351 }
1352 skip_oinp:
1353 if ((inp->inp_laddr.s_addr == INADDR_ANY ? laddr.s_addr :
1354 inp->inp_laddr.s_addr) == sin->sin_addr.s_addr &&
1355 inp->inp_lport == sin->sin_port) {
1356 error = EINVAL;
1357 goto done;
1358 }
1359 if (!lck_rw_try_lock_exclusive(inp->inp_pcbinfo->ipi_lock)) {
1360 /*lock inversion issue, mostly with udp multicast packets */
1361 socket_unlock(inp->inp_socket, 0);
1362 lck_rw_lock_exclusive(inp->inp_pcbinfo->ipi_lock);
1363 socket_lock(inp->inp_socket, 0);
1364 }
1365 if (inp->inp_laddr.s_addr == INADDR_ANY) {
1366 inp->inp_laddr = laddr;
1367 /* no reference needed */
1368 inp->inp_last_outifp = outif;
1369
1370 inp->inp_flags |= INP_INADDR_ANY;
1371 }
1372 inp->inp_faddr = sin->sin_addr;
1373 inp->inp_fport = sin->sin_port;
1374 in_pcbrehash(inp);
1375 lck_rw_done(inp->inp_pcbinfo->ipi_lock);
1376
1377 if (inp->inp_flowhash == 0)
1378 inp->inp_flowhash = inp_calc_flowhash(inp);
1379
1380 tcp_set_max_rwinscale(tp, so, TCP_AUTORCVBUF_MAX(outif));
1381
1382 soisconnecting(so);
1383 tcpstat.tcps_connattempt++;
1384 tp->t_state = TCPS_SYN_SENT;
1385 tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp, TCP_CONN_KEEPINIT(tp));
1386 tp->iss = tcp_new_isn(tp);
1387 tcp_sendseqinit(tp);
1388 if (nstat_collect)
1389 nstat_route_connect_attempt(inp->inp_route.ro_rt);
1390
1391 done:
1392 if (outif != NULL)
1393 ifnet_release(outif);
1394
1395 return (error);
1396 }
1397
1398 #if INET6
1399 static int
1400 tcp6_connect(struct tcpcb *tp, struct sockaddr *nam, struct proc *p)
1401 {
1402 struct inpcb *inp = tp->t_inpcb, *oinp;
1403 struct socket *so = inp->inp_socket;
1404 struct tcpcb *otp;
1405 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)(void *)nam;
1406 struct in6_addr addr6;
1407 int error = 0;
1408 struct ifnet *outif = NULL;
1409
1410 if (inp->inp_lport == 0) {
1411 error = in6_pcbbind(inp, NULL, p);
1412 if (error)
1413 goto done;
1414 }
1415
1416 /*
1417 * Cannot simply call in_pcbconnect, because there might be an
1418 * earlier incarnation of this same connection still in
1419 * TIME_WAIT state, creating an ADDRINUSE error.
1420 *
1421 * in6_pcbladdr() might return an ifp with its reference held
1422 * even in the error case, so make sure that it's released
1423 * whenever it's non-NULL.
1424 */
1425 error = in6_pcbladdr(inp, nam, &addr6, &outif);
1426 if (error)
1427 goto done;
1428 socket_unlock(inp->inp_socket, 0);
1429 oinp = in6_pcblookup_hash(inp->inp_pcbinfo,
1430 &sin6->sin6_addr, sin6->sin6_port,
1431 IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
1432 ? &addr6
1433 : &inp->in6p_laddr,
1434 inp->inp_lport, 0, NULL);
1435 socket_lock(inp->inp_socket, 0);
1436 if (oinp) {
1437 if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
1438 otp->t_state == TCPS_TIME_WAIT &&
1439 ((int)(tcp_now - otp->t_starttime)) < tcp_msl &&
1440 (otp->t_flags & TF_RCVD_CC)) {
1441 otp = tcp_close(otp);
1442 } else {
1443 error = EADDRINUSE;
1444 goto done;
1445 }
1446 }
1447 if (!lck_rw_try_lock_exclusive(inp->inp_pcbinfo->ipi_lock)) {
1448 /*lock inversion issue, mostly with udp multicast packets */
1449 socket_unlock(inp->inp_socket, 0);
1450 lck_rw_lock_exclusive(inp->inp_pcbinfo->ipi_lock);
1451 socket_lock(inp->inp_socket, 0);
1452 }
1453 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1454 inp->in6p_laddr = addr6;
1455 inp->in6p_last_outifp = outif; /* no reference needed */
1456 inp->in6p_flags |= INP_IN6ADDR_ANY;
1457 }
1458 inp->in6p_faddr = sin6->sin6_addr;
1459 inp->inp_fport = sin6->sin6_port;
1460 if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != 0)
1461 inp->inp_flow = sin6->sin6_flowinfo;
1462 in_pcbrehash(inp);
1463 lck_rw_done(inp->inp_pcbinfo->ipi_lock);
1464
1465 if (inp->inp_flowhash == 0)
1466 inp->inp_flowhash = inp_calc_flowhash(inp);
1467 /* update flowinfo - RFC 6437 */
1468 if (inp->inp_flow == 0 && inp->in6p_flags & IN6P_AUTOFLOWLABEL) {
1469 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
1470 inp->inp_flow |=
1471 (htonl(inp->inp_flowhash) & IPV6_FLOWLABEL_MASK);
1472 }
1473
1474 tcp_set_max_rwinscale(tp, so, TCP_AUTORCVBUF_MAX(outif));
1475
1476 soisconnecting(so);
1477 tcpstat.tcps_connattempt++;
1478 tp->t_state = TCPS_SYN_SENT;
1479 tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
1480 TCP_CONN_KEEPINIT(tp));
1481 tp->iss = tcp_new_isn(tp);
1482 tcp_sendseqinit(tp);
1483 if (nstat_collect)
1484 nstat_route_connect_attempt(inp->inp_route.ro_rt);
1485
1486 done:
1487 if (outif != NULL)
1488 ifnet_release(outif);
1489
1490 return (error);
1491 }
1492 #endif /* INET6 */
1493
1494 /*
1495 * Export TCP internal state information via a struct tcp_info
1496 */
1497 void
1498 tcp_fill_info(struct tcpcb *tp, struct tcp_info *ti)
1499 {
1500 struct inpcb *inp = tp->t_inpcb;
1501
1502 bzero(ti, sizeof(*ti));
1503
1504 ti->tcpi_state = tp->t_state;
1505 ti->tcpi_flowhash = inp->inp_flowhash;
1506
1507 if (tp->t_state > TCPS_LISTEN) {
1508 if (TSTMP_SUPPORTED(tp))
1509 ti->tcpi_options |= TCPI_OPT_TIMESTAMPS;
1510 if (SACK_ENABLED(tp))
1511 ti->tcpi_options |= TCPI_OPT_SACK;
1512 if (TCP_WINDOW_SCALE_ENABLED(tp)) {
1513 ti->tcpi_options |= TCPI_OPT_WSCALE;
1514 ti->tcpi_snd_wscale = tp->snd_scale;
1515 ti->tcpi_rcv_wscale = tp->rcv_scale;
1516 }
1517 if (TCP_ECN_ENABLED(tp))
1518 ti->tcpi_options |= TCPI_OPT_ECN;
1519
1520 /* Are we in retranmission episode */
1521 if (IN_FASTRECOVERY(tp) || tp->t_rxtshift > 0)
1522 ti->tcpi_flags |= TCPI_FLAG_LOSSRECOVERY;
1523
1524 if (tp->t_flags & TF_STREAMING_ON)
1525 ti->tcpi_flags |= TCPI_FLAG_STREAMING_ON;
1526
1527 ti->tcpi_rto = tp->t_timer[TCPT_REXMT] ? tp->t_rxtcur : 0;
1528 ti->tcpi_snd_mss = tp->t_maxseg;
1529 ti->tcpi_rcv_mss = tp->t_maxseg;
1530
1531 ti->tcpi_rttcur = tp->t_rttcur;
1532 ti->tcpi_srtt = tp->t_srtt >> TCP_RTT_SHIFT;
1533 ti->tcpi_rttvar = tp->t_rttvar >> TCP_RTTVAR_SHIFT;
1534 ti->tcpi_rttbest = tp->t_rttbest >> TCP_RTT_SHIFT;
1535
1536 ti->tcpi_snd_ssthresh = tp->snd_ssthresh;
1537 ti->tcpi_snd_cwnd = tp->snd_cwnd;
1538 ti->tcpi_snd_sbbytes = inp->inp_socket->so_snd.sb_cc;
1539
1540 ti->tcpi_rcv_space = tp->rcv_wnd;
1541
1542 ti->tcpi_snd_wnd = tp->snd_wnd;
1543 ti->tcpi_snd_nxt = tp->snd_nxt;
1544 ti->tcpi_rcv_nxt = tp->rcv_nxt;
1545
1546 /* convert bytes/msec to bits/sec */
1547 if ((tp->t_flagsext & TF_MEASURESNDBW) != 0 &&
1548 tp->t_bwmeas != NULL) {
1549 ti->tcpi_snd_bw = (tp->t_bwmeas->bw_sndbw * 8000);
1550 }
1551
1552 ti->tcpi_last_outif = (tp->t_inpcb->inp_last_outifp == NULL) ? 0 :
1553 tp->t_inpcb->inp_last_outifp->if_index;
1554
1555 //atomic_get_64(ti->tcpi_txbytes, &inp->inp_stat->txbytes);
1556 ti->tcpi_txpackets = inp->inp_stat->txpackets;
1557 ti->tcpi_txbytes = inp->inp_stat->txbytes;
1558 ti->tcpi_txretransmitbytes = tp->t_stat.txretransmitbytes;
1559 ti->tcpi_txretransmitpackets = tp->t_stat.rxmitpkts;
1560 ti->tcpi_txunacked = tp->snd_max - tp->snd_una;
1561
1562 //atomic_get_64(ti->tcpi_rxbytes, &inp->inp_stat->rxbytes);
1563 ti->tcpi_rxpackets = inp->inp_stat->rxpackets;
1564 ti->tcpi_rxbytes = inp->inp_stat->rxbytes;
1565 ti->tcpi_rxduplicatebytes = tp->t_stat.rxduplicatebytes;
1566 ti->tcpi_rxoutoforderbytes = tp->t_stat.rxoutoforderbytes;
1567
1568 if (tp->t_state > TCPS_LISTEN) {
1569 ti->tcpi_synrexmits = tp->t_stat.synrxtshift;
1570 }
1571 ti->tcpi_cell_rxpackets = inp->inp_cstat->rxpackets;
1572 ti->tcpi_cell_rxbytes = inp->inp_cstat->rxbytes;
1573 ti->tcpi_cell_txpackets = inp->inp_cstat->txpackets;
1574 ti->tcpi_cell_txbytes = inp->inp_cstat->txbytes;
1575
1576 ti->tcpi_wifi_rxpackets = inp->inp_wstat->rxpackets;
1577 ti->tcpi_wifi_rxbytes = inp->inp_wstat->rxbytes;
1578 ti->tcpi_wifi_txpackets = inp->inp_wstat->txpackets;
1579 ti->tcpi_wifi_txbytes = inp->inp_wstat->txbytes;
1580
1581 ti->tcpi_wired_rxpackets = inp->inp_Wstat->rxpackets;
1582 ti->tcpi_wired_rxbytes = inp->inp_Wstat->rxbytes;
1583 ti->tcpi_wired_txpackets = inp->inp_Wstat->txpackets;
1584 ti->tcpi_wired_txbytes = inp->inp_Wstat->txbytes;
1585 tcp_get_connectivity_status(tp, &ti->tcpi_connstatus);
1586
1587 ti->tcpi_tfo_syn_data_rcv = !!(tp->t_tfo_stats & TFO_S_SYNDATA_RCV);
1588 ti->tcpi_tfo_cookie_req_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIEREQ_RECV);
1589 ti->tcpi_tfo_cookie_sent = !!(tp->t_tfo_stats & TFO_S_COOKIE_SENT);
1590 ti->tcpi_tfo_cookie_invalid = !!(tp->t_tfo_stats & TFO_S_COOKIE_INVALID);
1591
1592 ti->tcpi_tfo_cookie_req = !!(tp->t_tfo_stats & TFO_S_COOKIE_REQ);
1593 ti->tcpi_tfo_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIE_RCV);
1594 ti->tcpi_tfo_syn_data_sent = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_SENT);
1595 ti->tcpi_tfo_syn_data_acked = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_ACKED);
1596 ti->tcpi_tfo_syn_loss = !!(tp->t_tfo_stats & TFO_S_SYN_LOSS);
1597 ti->tcpi_tfo_cookie_wrong = !!(tp->t_tfo_stats & TFO_S_COOKIE_WRONG);
1598 ti->tcpi_tfo_no_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_NO_COOKIE_RCV);
1599 ti->tcpi_tfo_heuristics_disable = !!(tp->t_tfo_stats & TFO_S_HEURISTICS_DISABLE);
1600 ti->tcpi_tfo_send_blackhole = !!(tp->t_tfo_stats & TFO_S_SEND_BLACKHOLE);
1601 ti->tcpi_tfo_recv_blackhole = !!(tp->t_tfo_stats & TFO_S_RECV_BLACKHOLE);
1602 ti->tcpi_tfo_onebyte_proxy = !!(tp->t_tfo_stats & TFO_S_ONE_BYTE_PROXY);
1603
1604 ti->tcpi_ecn_client_setup = !!(tp->ecn_flags & TE_SETUPSENT);
1605 ti->tcpi_ecn_server_setup = !!(tp->ecn_flags & TE_SETUPRECEIVED);
1606 ti->tcpi_ecn_success = (tp->ecn_flags & TE_ECN_ON) == TE_ECN_ON ? 1 : 0;
1607 ti->tcpi_ecn_lost_syn = !!(tp->ecn_flags & TE_LOST_SYN);
1608 ti->tcpi_ecn_lost_synack = !!(tp->ecn_flags & TE_LOST_SYNACK);
1609
1610 ti->tcpi_local_peer = !!(tp->t_flags & TF_LOCAL);
1611
1612 if (tp->t_inpcb->inp_last_outifp != NULL) {
1613 if (IFNET_IS_CELLULAR(tp->t_inpcb->inp_last_outifp))
1614 ti->tcpi_if_cell = 1;
1615 if (IFNET_IS_WIFI(tp->t_inpcb->inp_last_outifp))
1616 ti->tcpi_if_wifi = 1;
1617 if (IFNET_IS_WIRED(tp->t_inpcb->inp_last_outifp))
1618 ti->tcpi_if_wired = 1;
1619 if (IFNET_IS_WIFI_INFRA(tp->t_inpcb->inp_last_outifp))
1620 ti->tcpi_if_wifi_infra = 1;
1621 if (tp->t_inpcb->inp_last_outifp->if_eflags & IFEF_AWDL)
1622 ti->tcpi_if_wifi_awdl = 1;
1623 }
1624 if (tp->tcp_cc_index == TCP_CC_ALGO_BACKGROUND_INDEX)
1625 ti->tcpi_snd_background = 1;
1626 if (tcp_recv_bg == 1 ||
1627 IS_TCP_RECV_BG(tp->t_inpcb->inp_socket))
1628 ti->tcpi_rcv_background = 1;
1629
1630 ti->tcpi_ecn_recv_ce = tp->t_ecn_recv_ce;
1631 ti->tcpi_ecn_recv_cwr = tp->t_ecn_recv_cwr;
1632
1633 ti->tcpi_rcvoopack = tp->t_rcvoopack;
1634 ti->tcpi_pawsdrop = tp->t_pawsdrop;
1635 ti->tcpi_sack_recovery_episode = tp->t_sack_recovery_episode;
1636 ti->tcpi_reordered_pkts = tp->t_reordered_pkts;
1637 ti->tcpi_dsack_sent = tp->t_dsack_sent;
1638 ti->tcpi_dsack_recvd = tp->t_dsack_recvd;
1639 }
1640 }
1641
1642 __private_extern__ errno_t
1643 tcp_fill_info_for_info_tuple(struct info_tuple *itpl, struct tcp_info *ti)
1644 {
1645 struct inpcbinfo *pcbinfo = NULL;
1646 struct inpcb *inp = NULL;
1647 struct socket *so;
1648 struct tcpcb *tp;
1649
1650 if (itpl->itpl_proto == IPPROTO_TCP)
1651 pcbinfo = &tcbinfo;
1652 else
1653 return EINVAL;
1654
1655 if (itpl->itpl_local_sa.sa_family == AF_INET &&
1656 itpl->itpl_remote_sa.sa_family == AF_INET) {
1657 inp = in_pcblookup_hash(pcbinfo,
1658 itpl->itpl_remote_sin.sin_addr,
1659 itpl->itpl_remote_sin.sin_port,
1660 itpl->itpl_local_sin.sin_addr,
1661 itpl->itpl_local_sin.sin_port,
1662 0, NULL);
1663 } else if (itpl->itpl_local_sa.sa_family == AF_INET6 &&
1664 itpl->itpl_remote_sa.sa_family == AF_INET6) {
1665 struct in6_addr ina6_local;
1666 struct in6_addr ina6_remote;
1667
1668 ina6_local = itpl->itpl_local_sin6.sin6_addr;
1669 if (IN6_IS_SCOPE_LINKLOCAL(&ina6_local) &&
1670 itpl->itpl_local_sin6.sin6_scope_id)
1671 ina6_local.s6_addr16[1] = htons(itpl->itpl_local_sin6.sin6_scope_id);
1672
1673 ina6_remote = itpl->itpl_remote_sin6.sin6_addr;
1674 if (IN6_IS_SCOPE_LINKLOCAL(&ina6_remote) &&
1675 itpl->itpl_remote_sin6.sin6_scope_id)
1676 ina6_remote.s6_addr16[1] = htons(itpl->itpl_remote_sin6.sin6_scope_id);
1677
1678 inp = in6_pcblookup_hash(pcbinfo,
1679 &ina6_remote,
1680 itpl->itpl_remote_sin6.sin6_port,
1681 &ina6_local,
1682 itpl->itpl_local_sin6.sin6_port,
1683 0, NULL);
1684 } else {
1685 return EINVAL;
1686 }
1687 if (inp == NULL || (so = inp->inp_socket) == NULL)
1688 return ENOENT;
1689
1690 socket_lock(so, 0);
1691 if (in_pcb_checkstate(inp, WNT_RELEASE, 1) == WNT_STOPUSING) {
1692 socket_unlock(so, 0);
1693 return ENOENT;
1694 }
1695 tp = intotcpcb(inp);
1696
1697 tcp_fill_info(tp, ti);
1698 socket_unlock(so, 0);
1699
1700 return 0;
1701 }
1702
1703 static void
1704 tcp_connection_fill_info(struct tcpcb *tp, struct tcp_connection_info *tci)
1705 {
1706 struct inpcb *inp = tp->t_inpcb;
1707
1708 bzero(tci, sizeof(*tci));
1709 tci->tcpi_state = tp->t_state;
1710 if (tp->t_state > TCPS_LISTEN) {
1711 if (TSTMP_SUPPORTED(tp))
1712 tci->tcpi_options |= TCPCI_OPT_TIMESTAMPS;
1713 if (SACK_ENABLED(tp))
1714 tci->tcpi_options |= TCPCI_OPT_SACK;
1715 if (TCP_WINDOW_SCALE_ENABLED(tp)) {
1716 tci->tcpi_options |= TCPCI_OPT_WSCALE;
1717 tci->tcpi_snd_wscale = tp->snd_scale;
1718 tci->tcpi_rcv_wscale = tp->rcv_scale;
1719 }
1720 if (TCP_ECN_ENABLED(tp))
1721 tci->tcpi_options |= TCPCI_OPT_ECN;
1722 if (IN_FASTRECOVERY(tp) || tp->t_rxtshift > 0)
1723 tci->tcpi_flags |= TCPCI_FLAG_LOSSRECOVERY;
1724 if (tp->t_flagsext & TF_PKTS_REORDERED)
1725 tci->tcpi_flags |= TCPCI_FLAG_REORDERING_DETECTED;
1726 tci->tcpi_rto = (tp->t_timer[TCPT_REXMT] > 0) ?
1727 tp->t_rxtcur : 0;
1728 tci->tcpi_maxseg = tp->t_maxseg;
1729 tci->tcpi_snd_ssthresh = tp->snd_ssthresh;
1730 tci->tcpi_snd_cwnd = tp->snd_cwnd;
1731 tci->tcpi_snd_wnd = tp->snd_wnd;
1732 tci->tcpi_snd_sbbytes = inp->inp_socket->so_snd.sb_cc;
1733 tci->tcpi_rcv_wnd = tp->rcv_wnd;
1734 tci->tcpi_rttcur = tp->t_rttcur;
1735 tci->tcpi_srtt = (tp->t_srtt >> TCP_RTT_SHIFT);
1736 tci->tcpi_rttvar = (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1737 tci->tcpi_txpackets = inp->inp_stat->txpackets;
1738 tci->tcpi_txbytes = inp->inp_stat->txbytes;
1739 tci->tcpi_txretransmitbytes = tp->t_stat.txretransmitbytes;
1740 tci->tcpi_txretransmitpackets = tp->t_stat.rxmitpkts;
1741 tci->tcpi_rxpackets = inp->inp_stat->rxpackets;
1742 tci->tcpi_rxbytes = inp->inp_stat->rxbytes;
1743 tci->tcpi_rxoutoforderbytes = tp->t_stat.rxoutoforderbytes;
1744
1745 tci->tcpi_tfo_syn_data_rcv = !!(tp->t_tfo_stats & TFO_S_SYNDATA_RCV);
1746 tci->tcpi_tfo_cookie_req_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIEREQ_RECV);
1747 tci->tcpi_tfo_cookie_sent = !!(tp->t_tfo_stats & TFO_S_COOKIE_SENT);
1748 tci->tcpi_tfo_cookie_invalid = !!(tp->t_tfo_stats & TFO_S_COOKIE_INVALID);
1749 tci->tcpi_tfo_cookie_req = !!(tp->t_tfo_stats & TFO_S_COOKIE_REQ);
1750 tci->tcpi_tfo_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_COOKIE_RCV);
1751 tci->tcpi_tfo_syn_data_sent = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_SENT);
1752 tci->tcpi_tfo_syn_data_acked = !!(tp->t_tfo_stats & TFO_S_SYN_DATA_ACKED);
1753 tci->tcpi_tfo_syn_loss = !!(tp->t_tfo_stats & TFO_S_SYN_LOSS);
1754 tci->tcpi_tfo_cookie_wrong = !!(tp->t_tfo_stats & TFO_S_COOKIE_WRONG);
1755 tci->tcpi_tfo_no_cookie_rcv = !!(tp->t_tfo_stats & TFO_S_NO_COOKIE_RCV);
1756 tci->tcpi_tfo_heuristics_disable = !!(tp->t_tfo_stats & TFO_S_HEURISTICS_DISABLE);
1757 tci->tcpi_tfo_send_blackhole = !!(tp->t_tfo_stats & TFO_S_SEND_BLACKHOLE);
1758 tci->tcpi_tfo_recv_blackhole = !!(tp->t_tfo_stats & TFO_S_RECV_BLACKHOLE);
1759 tci->tcpi_tfo_onebyte_proxy = !!(tp->t_tfo_stats & TFO_S_ONE_BYTE_PROXY);
1760 }
1761 }
1762
1763
1764 __private_extern__ int
1765 tcp_sysctl_info(__unused struct sysctl_oid *oidp, __unused void *arg1, __unused int arg2, struct sysctl_req *req)
1766 {
1767 int error;
1768 struct tcp_info ti;
1769 struct info_tuple itpl;
1770 #if !CONFIG_EMBEDDED
1771 proc_t caller = PROC_NULL;
1772 proc_t caller_parent = PROC_NULL;
1773 char command_name[MAXCOMLEN + 1] = "";
1774 char parent_name[MAXCOMLEN + 1] = "";
1775
1776 if ((caller = proc_self()) != PROC_NULL) {
1777 /* get process name */
1778 strlcpy(command_name, caller->p_comm, sizeof(command_name));
1779
1780 /* get parent process name if possible */
1781 if ((caller_parent = proc_find(caller->p_ppid)) != PROC_NULL) {
1782 strlcpy(parent_name, caller_parent->p_comm,
1783 sizeof(parent_name));
1784 proc_rele(caller_parent);
1785 }
1786
1787 if ((escape_str(command_name, strlen(command_name),
1788 sizeof(command_name)) == 0) &&
1789 (escape_str(parent_name, strlen(parent_name),
1790 sizeof(parent_name)) == 0)) {
1791 kern_asl_msg(LOG_DEBUG, "messagetracer",
1792 5,
1793 "com.apple.message.domain",
1794 "com.apple.kernel.tcpstat", /* 1 */
1795 "com.apple.message.signature",
1796 "tcpinfo", /* 2 */
1797 "com.apple.message.signature2", command_name, /* 3 */
1798 "com.apple.message.signature3", parent_name, /* 4 */
1799 "com.apple.message.summarize", "YES", /* 5 */
1800 NULL);
1801 }
1802 }
1803
1804 if (caller != PROC_NULL)
1805 proc_rele(caller);
1806 #endif /* !CONFIG_EMBEDDED */
1807
1808 if (req->newptr == USER_ADDR_NULL) {
1809 return EINVAL;
1810 }
1811 if (req->newlen < sizeof(struct info_tuple)) {
1812 return EINVAL;
1813 }
1814 error = SYSCTL_IN(req, &itpl, sizeof(struct info_tuple));
1815 if (error != 0) {
1816 return error;
1817 }
1818 error = tcp_fill_info_for_info_tuple(&itpl, &ti);
1819 if (error != 0) {
1820 return error;
1821 }
1822 error = SYSCTL_OUT(req, &ti, sizeof(struct tcp_info));
1823 if (error != 0) {
1824 return error;
1825 }
1826
1827 return 0;
1828 }
1829
1830 static int
1831 tcp_lookup_peer_pid_locked(struct socket *so, pid_t *out_pid)
1832 {
1833 int error = EHOSTUNREACH;
1834 *out_pid = -1;
1835 if ((so->so_state & SS_ISCONNECTED) == 0) return ENOTCONN;
1836
1837 struct inpcb *inp = (struct inpcb*)so->so_pcb;
1838 uint16_t lport = inp->inp_lport;
1839 uint16_t fport = inp->inp_fport;
1840 struct inpcb *finp = NULL;
1841
1842 if (inp->inp_vflag & INP_IPV6) {
1843 struct in6_addr laddr6 = inp->in6p_laddr;
1844 struct in6_addr faddr6 = inp->in6p_faddr;
1845 socket_unlock(so, 0);
1846 finp = in6_pcblookup_hash(&tcbinfo, &laddr6, lport, &faddr6, fport, 0, NULL);
1847 socket_lock(so, 0);
1848 } else if (inp->inp_vflag & INP_IPV4) {
1849 struct in_addr laddr4 = inp->inp_laddr;
1850 struct in_addr faddr4 = inp->inp_faddr;
1851 socket_unlock(so, 0);
1852 finp = in_pcblookup_hash(&tcbinfo, laddr4, lport, faddr4, fport, 0, NULL);
1853 socket_lock(so, 0);
1854 }
1855
1856 if (finp) {
1857 *out_pid = finp->inp_socket->last_pid;
1858 error = 0;
1859 /* Avoid deadlock due to same inpcb for loopback socket */
1860 if (inp == finp)
1861 in_pcb_checkstate(finp, WNT_RELEASE, 1);
1862 else
1863 in_pcb_checkstate(finp, WNT_RELEASE, 0);
1864 }
1865
1866 return error;
1867 }
1868
1869 void
1870 tcp_getconninfo(struct socket *so, struct conninfo_tcp *tcp_ci)
1871 {
1872 (void) tcp_lookup_peer_pid_locked(so, &tcp_ci->tcpci_peer_pid);
1873 tcp_fill_info(sototcpcb(so), &tcp_ci->tcpci_tcp_info);
1874 }
1875
1876 /*
1877 * The new sockopt interface makes it possible for us to block in the
1878 * copyin/out step (if we take a page fault). Taking a page fault at
1879 * splnet() is probably a Bad Thing. (Since sockets and pcbs both now
1880 * use TSM, there probably isn't any need for this function to run at
1881 * splnet() any more. This needs more examination.)
1882 */
1883 int
1884 tcp_ctloutput(struct socket *so, struct sockopt *sopt)
1885 {
1886 int error, opt, optval;
1887 struct inpcb *inp;
1888 struct tcpcb *tp;
1889
1890 error = 0;
1891 inp = sotoinpcb(so);
1892 if (inp == NULL) {
1893 return (ECONNRESET);
1894 }
1895 /* Allow <SOL_SOCKET,SO_FLUSH/SO_TRAFFIC_MGT_BACKGROUND> at this level */
1896 if (sopt->sopt_level != IPPROTO_TCP &&
1897 !(sopt->sopt_level == SOL_SOCKET && (sopt->sopt_name == SO_FLUSH ||
1898 sopt->sopt_name == SO_TRAFFIC_MGT_BACKGROUND))) {
1899 #if INET6
1900 if (SOCK_CHECK_DOM(so, PF_INET6))
1901 error = ip6_ctloutput(so, sopt);
1902 else
1903 #endif /* INET6 */
1904 error = ip_ctloutput(so, sopt);
1905 return (error);
1906 }
1907 tp = intotcpcb(inp);
1908 if (tp == NULL) {
1909 return (ECONNRESET);
1910 }
1911
1912 calculate_tcp_clock();
1913
1914 switch (sopt->sopt_dir) {
1915 case SOPT_SET:
1916 switch (sopt->sopt_name) {
1917 case TCP_NODELAY:
1918 case TCP_NOOPT:
1919 case TCP_NOPUSH:
1920 error = sooptcopyin(sopt, &optval, sizeof optval,
1921 sizeof optval);
1922 if (error)
1923 break;
1924
1925 switch (sopt->sopt_name) {
1926 case TCP_NODELAY:
1927 opt = TF_NODELAY;
1928 break;
1929 case TCP_NOOPT:
1930 opt = TF_NOOPT;
1931 break;
1932 case TCP_NOPUSH:
1933 opt = TF_NOPUSH;
1934 break;
1935 default:
1936 opt = 0; /* dead code to fool gcc */
1937 break;
1938 }
1939
1940 if (optval)
1941 tp->t_flags |= opt;
1942 else
1943 tp->t_flags &= ~opt;
1944 break;
1945 case TCP_RXT_FINDROP:
1946 case TCP_NOTIMEWAIT:
1947 error = sooptcopyin(sopt, &optval, sizeof optval,
1948 sizeof optval);
1949 if (error)
1950 break;
1951 switch (sopt->sopt_name) {
1952 case TCP_RXT_FINDROP:
1953 opt = TF_RXTFINDROP;
1954 break;
1955 case TCP_NOTIMEWAIT:
1956 opt = TF_NOTIMEWAIT;
1957 break;
1958 default:
1959 opt = 0;
1960 break;
1961 }
1962 if (optval)
1963 tp->t_flagsext |= opt;
1964 else
1965 tp->t_flagsext &= ~opt;
1966 break;
1967 case TCP_MEASURE_SND_BW:
1968 error = sooptcopyin(sopt, &optval, sizeof optval,
1969 sizeof optval);
1970 if (error)
1971 break;
1972 opt = TF_MEASURESNDBW;
1973 if (optval) {
1974 if (tp->t_bwmeas == NULL) {
1975 tp->t_bwmeas = tcp_bwmeas_alloc(tp);
1976 if (tp->t_bwmeas == NULL) {
1977 error = ENOMEM;
1978 break;
1979 }
1980 }
1981 tp->t_flagsext |= opt;
1982 } else {
1983 tp->t_flagsext &= ~opt;
1984 /* Reset snd bw measurement state */
1985 tp->t_flagsext &= ~(TF_BWMEAS_INPROGRESS);
1986 if (tp->t_bwmeas != NULL) {
1987 tcp_bwmeas_free(tp);
1988 }
1989 }
1990 break;
1991 case TCP_MEASURE_BW_BURST: {
1992 struct tcp_measure_bw_burst in;
1993 uint32_t minpkts, maxpkts;
1994 bzero(&in, sizeof(in));
1995
1996 error = sooptcopyin(sopt, &in, sizeof(in),
1997 sizeof(in));
1998 if (error)
1999 break;
2000 if ((tp->t_flagsext & TF_MEASURESNDBW) == 0 ||
2001 tp->t_bwmeas == NULL) {
2002 error = EINVAL;
2003 break;
2004 }
2005 minpkts = (in.min_burst_size != 0) ? in.min_burst_size :
2006 tp->t_bwmeas->bw_minsizepkts;
2007 maxpkts = (in.max_burst_size != 0) ? in.max_burst_size :
2008 tp->t_bwmeas->bw_maxsizepkts;
2009 if (minpkts > maxpkts) {
2010 error = EINVAL;
2011 break;
2012 }
2013 tp->t_bwmeas->bw_minsizepkts = minpkts;
2014 tp->t_bwmeas->bw_maxsizepkts = maxpkts;
2015 tp->t_bwmeas->bw_minsize = (minpkts * tp->t_maxseg);
2016 tp->t_bwmeas->bw_maxsize = (maxpkts * tp->t_maxseg);
2017 break;
2018 }
2019 case TCP_MAXSEG:
2020 error = sooptcopyin(sopt, &optval, sizeof optval,
2021 sizeof optval);
2022 if (error)
2023 break;
2024
2025 if (optval > 0 && optval <= tp->t_maxseg &&
2026 optval + 40 >= tcp_minmss)
2027 tp->t_maxseg = optval;
2028 else
2029 error = EINVAL;
2030 break;
2031
2032 case TCP_KEEPALIVE:
2033 error = sooptcopyin(sopt, &optval, sizeof optval,
2034 sizeof optval);
2035 if (error)
2036 break;
2037 if (optval < 0 || optval > UINT32_MAX/TCP_RETRANSHZ) {
2038 error = EINVAL;
2039 } else {
2040 tp->t_keepidle = optval * TCP_RETRANSHZ;
2041 /* reset the timer to new value */
2042 tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
2043 TCP_CONN_KEEPIDLE(tp));
2044 tcp_check_timer_state(tp);
2045 }
2046 break;
2047
2048 case TCP_CONNECTIONTIMEOUT:
2049 error = sooptcopyin(sopt, &optval, sizeof optval,
2050 sizeof optval);
2051 if (error)
2052 break;
2053 if (optval < 0 || optval > UINT32_MAX/TCP_RETRANSHZ) {
2054 error = EINVAL;
2055 } else {
2056 tp->t_keepinit = optval * TCP_RETRANSHZ;
2057 if (tp->t_state == TCPS_SYN_RECEIVED ||
2058 tp->t_state == TCPS_SYN_SENT) {
2059 tp->t_timer[TCPT_KEEP] = OFFSET_FROM_START(tp,
2060 TCP_CONN_KEEPINIT(tp));
2061 tcp_check_timer_state(tp);
2062 }
2063 }
2064 break;
2065
2066 case TCP_KEEPINTVL:
2067 error = sooptcopyin(sopt, &optval, sizeof(optval),
2068 sizeof(optval));
2069 if (error)
2070 break;
2071 if (optval < 0 || optval > UINT32_MAX/TCP_RETRANSHZ) {
2072 error = EINVAL;
2073 } else {
2074 tp->t_keepintvl = optval * TCP_RETRANSHZ;
2075 if (tp->t_state == TCPS_FIN_WAIT_2 &&
2076 TCP_CONN_MAXIDLE(tp) > 0) {
2077 tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp,
2078 TCP_CONN_MAXIDLE(tp));
2079 tcp_check_timer_state(tp);
2080 }
2081 }
2082 break;
2083
2084 case TCP_KEEPCNT:
2085 error = sooptcopyin(sopt, &optval, sizeof(optval),
2086 sizeof(optval));
2087 if (error)
2088 break;
2089 if (optval < 0 || optval > INT32_MAX) {
2090 error = EINVAL;
2091 } else {
2092 tp->t_keepcnt = optval;
2093 if (tp->t_state == TCPS_FIN_WAIT_2 &&
2094 TCP_CONN_MAXIDLE(tp) > 0) {
2095 tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp,
2096 TCP_CONN_MAXIDLE(tp));
2097 tcp_check_timer_state(tp);
2098 }
2099 }
2100 break;
2101
2102 case TCP_KEEPALIVE_OFFLOAD:
2103 error = sooptcopyin(sopt, &optval, sizeof(optval),
2104 sizeof(optval));
2105 if (error)
2106 break;
2107 if (optval < 0 || optval > INT32_MAX) {
2108 error = EINVAL;
2109 break;
2110 }
2111 if (optval != 0)
2112 inp->inp_flags2 |= INP2_KEEPALIVE_OFFLOAD;
2113 else
2114 inp->inp_flags2 &= ~INP2_KEEPALIVE_OFFLOAD;
2115 break;
2116
2117 case PERSIST_TIMEOUT:
2118 error = sooptcopyin(sopt, &optval, sizeof optval,
2119 sizeof optval);
2120 if (error)
2121 break;
2122 if (optval < 0)
2123 error = EINVAL;
2124 else
2125 tp->t_persist_timeout = optval * TCP_RETRANSHZ;
2126 break;
2127 case TCP_RXT_CONNDROPTIME:
2128 error = sooptcopyin(sopt, &optval, sizeof(optval),
2129 sizeof(optval));
2130 if (error)
2131 break;
2132 if (optval < 0)
2133 error = EINVAL;
2134 else
2135 tp->t_rxt_conndroptime = optval * TCP_RETRANSHZ;
2136 break;
2137 case TCP_NOTSENT_LOWAT:
2138 error = sooptcopyin(sopt, &optval, sizeof(optval),
2139 sizeof(optval));
2140 if (error)
2141 break;
2142 if (optval < 0) {
2143 error = EINVAL;
2144 break;
2145 } else {
2146 if (optval == 0) {
2147 so->so_flags &= ~(SOF_NOTSENT_LOWAT);
2148 tp->t_notsent_lowat = 0;
2149 } else {
2150 so->so_flags |= SOF_NOTSENT_LOWAT;
2151 tp->t_notsent_lowat = optval;
2152 }
2153 }
2154 break;
2155 case TCP_ADAPTIVE_READ_TIMEOUT:
2156 error = sooptcopyin(sopt, &optval, sizeof (optval),
2157 sizeof(optval));
2158 if (error)
2159 break;
2160 if (optval < 0 ||
2161 optval > TCP_ADAPTIVE_TIMEOUT_MAX) {
2162 error = EINVAL;
2163 break;
2164 } else if (optval == 0) {
2165 tp->t_adaptive_rtimo = 0;
2166 tcp_keepalive_reset(tp);
2167
2168 if (tp->t_mpsub)
2169 mptcp_reset_keepalive(tp);
2170 } else {
2171 tp->t_adaptive_rtimo = optval;
2172 }
2173 break;
2174 case TCP_ADAPTIVE_WRITE_TIMEOUT:
2175 error = sooptcopyin(sopt, &optval, sizeof (optval),
2176 sizeof (optval));
2177 if (error)
2178 break;
2179 if (optval < 0 ||
2180 optval > TCP_ADAPTIVE_TIMEOUT_MAX) {
2181 error = EINVAL;
2182 break;
2183 } else {
2184 tp->t_adaptive_wtimo = optval;
2185 }
2186 break;
2187 case TCP_ENABLE_MSGS:
2188 error = sooptcopyin(sopt, &optval, sizeof(optval),
2189 sizeof(optval));
2190 if (error)
2191 break;
2192 if (optval < 0 || optval > 1) {
2193 error = EINVAL;
2194 } else if (optval == 1) {
2195 /*
2196 * Check if messages option is already
2197 * enabled, if so return.
2198 */
2199 if (so->so_flags & SOF_ENABLE_MSGS) {
2200 VERIFY(so->so_msg_state != NULL);
2201 break;
2202 }
2203
2204 /*
2205 * allocate memory for storing message
2206 * related state
2207 */
2208 VERIFY(so->so_msg_state == NULL);
2209 MALLOC(so->so_msg_state,
2210 struct msg_state *,
2211 sizeof(struct msg_state),
2212 M_TEMP, M_WAITOK | M_ZERO);
2213 if (so->so_msg_state == NULL) {
2214 error = ENOMEM;
2215 break;
2216 }
2217
2218 /* Enable message delivery */
2219 so->so_flags |= SOF_ENABLE_MSGS;
2220 } else {
2221 /*
2222 * Can't disable message delivery on socket
2223 * because of restrictions imposed by
2224 * encoding/decoding
2225 */
2226 error = EINVAL;
2227 }
2228 break;
2229 case TCP_SENDMOREACKS:
2230 error = sooptcopyin(sopt, &optval, sizeof(optval),
2231 sizeof(optval));
2232 if (error)
2233 break;
2234 if (optval < 0 || optval > 1) {
2235 error = EINVAL;
2236 } else if (optval == 0) {
2237 tp->t_flagsext &= ~(TF_NOSTRETCHACK);
2238 } else {
2239 tp->t_flagsext |= TF_NOSTRETCHACK;
2240 }
2241 break;
2242 case TCP_DISABLE_BLACKHOLE_DETECTION:
2243 error = sooptcopyin(sopt, &optval, sizeof(optval),
2244 sizeof(optval));
2245 if (error)
2246 break;
2247 if (optval < 0 || optval > 1) {
2248 error = EINVAL;
2249 } else if (optval == 0) {
2250 tp->t_flagsext &= ~TF_NOBLACKHOLE_DETECTION;
2251 } else {
2252 tp->t_flagsext |= TF_NOBLACKHOLE_DETECTION;
2253 if ((tp->t_flags & TF_BLACKHOLE) &&
2254 tp->t_pmtud_saved_maxopd > 0)
2255 tcp_pmtud_revert_segment_size(tp);
2256 }
2257 break;
2258 case TCP_FASTOPEN:
2259 if (!(tcp_fastopen & TCP_FASTOPEN_SERVER)) {
2260 error = ENOTSUP;
2261 break;
2262 }
2263
2264 error = sooptcopyin(sopt, &optval, sizeof(optval),
2265 sizeof(optval));
2266 if (error)
2267 break;
2268 if (optval < 0 || optval > 1) {
2269 error = EINVAL;
2270 break;
2271 }
2272 if (tp->t_state != TCPS_LISTEN) {
2273 error = EINVAL;
2274 break;
2275 }
2276 if (optval)
2277 tp->t_flagsext |= TF_FASTOPEN;
2278 else
2279 tcp_disable_tfo(tp);
2280 break;
2281 case TCP_FASTOPEN_FORCE_HEURISTICS:
2282 error = sooptcopyin(sopt, &optval, sizeof(optval),
2283 sizeof(optval));
2284
2285 if (error)
2286 break;
2287 if (optval < 0 || optval > 1) {
2288 error = EINVAL;
2289 break;
2290 }
2291
2292 if (tp->t_state != TCPS_CLOSED) {
2293 error = EINVAL;
2294 break;
2295 }
2296 if (optval)
2297 tp->t_flagsext |= TF_FASTOPEN_HEUR;
2298 else
2299 tp->t_flagsext &= ~TF_FASTOPEN_HEUR;
2300
2301 break;
2302 case TCP_ENABLE_ECN:
2303 error = sooptcopyin(sopt, &optval, sizeof optval,
2304 sizeof optval);
2305 if (error)
2306 break;
2307 if (optval) {
2308 tp->ecn_flags |= TE_ECN_MODE_ENABLE;
2309 tp->ecn_flags &= ~TE_ECN_MODE_DISABLE;
2310 } else {
2311 tp->ecn_flags &= ~TE_ECN_MODE_ENABLE;
2312 tp->ecn_flags |= TE_ECN_MODE_DISABLE;
2313 }
2314 break;
2315 case TCP_ECN_MODE:
2316 error = sooptcopyin(sopt, &optval, sizeof optval,
2317 sizeof optval);
2318 if (error)
2319 break;
2320 if (optval == ECN_MODE_DEFAULT) {
2321 tp->ecn_flags &= ~TE_ECN_MODE_ENABLE;
2322 tp->ecn_flags &= ~TE_ECN_MODE_DISABLE;
2323 } else if (optval == ECN_MODE_ENABLE) {
2324 tp->ecn_flags |= TE_ECN_MODE_ENABLE;
2325 tp->ecn_flags &= ~TE_ECN_MODE_DISABLE;
2326 } else if (optval == ECN_MODE_DISABLE) {
2327 tp->ecn_flags &= ~TE_ECN_MODE_ENABLE;
2328 tp->ecn_flags |= TE_ECN_MODE_DISABLE;
2329 } else {
2330 error = EINVAL;
2331 }
2332 break;
2333 case TCP_NOTIFY_ACKNOWLEDGEMENT:
2334 error = sooptcopyin(sopt, &optval,
2335 sizeof(optval), sizeof(optval));
2336 if (error)
2337 break;
2338 if (optval <= 0) {
2339 error = EINVAL;
2340 break;
2341 }
2342 if (tp->t_notify_ack_count >= TCP_MAX_NOTIFY_ACK) {
2343 error = ETOOMANYREFS;
2344 break;
2345 }
2346
2347 /*
2348 * validate that the given marker id is not
2349 * a duplicate to avoid ambiguity
2350 */
2351 if ((error = tcp_notify_ack_id_valid(tp, so,
2352 optval)) != 0) {
2353 break;
2354 }
2355 error = tcp_add_notify_ack_marker(tp, optval);
2356 break;
2357 case SO_FLUSH:
2358 if ((error = sooptcopyin(sopt, &optval, sizeof (optval),
2359 sizeof (optval))) != 0)
2360 break;
2361
2362 error = inp_flush(inp, optval);
2363 break;
2364
2365 case SO_TRAFFIC_MGT_BACKGROUND:
2366 if ((error = sooptcopyin(sopt, &optval, sizeof (optval),
2367 sizeof (optval))) != 0)
2368 break;
2369
2370 if (optval) {
2371 socket_set_traffic_mgt_flags_locked(so,
2372 TRAFFIC_MGT_SO_BACKGROUND);
2373 } else {
2374 socket_clear_traffic_mgt_flags_locked(so,
2375 TRAFFIC_MGT_SO_BACKGROUND);
2376 }
2377 break;
2378 case TCP_RXT_MINIMUM_TIMEOUT:
2379 error = sooptcopyin(sopt, &optval, sizeof(optval),
2380 sizeof(optval));
2381 if (error)
2382 break;
2383 if (optval < 0) {
2384 error = EINVAL;
2385 break;
2386 }
2387 if (optval == 0) {
2388 tp->t_rxt_minimum_timeout = 0;
2389 } else {
2390 tp->t_rxt_minimum_timeout = min(optval,
2391 TCP_RXT_MINIMUM_TIMEOUT_LIMIT);
2392 /* convert to milliseconds */
2393 tp->t_rxt_minimum_timeout *= TCP_RETRANSHZ;
2394 }
2395 break;
2396 default:
2397 error = ENOPROTOOPT;
2398 break;
2399 }
2400 break;
2401
2402 case SOPT_GET:
2403 switch (sopt->sopt_name) {
2404 case TCP_NODELAY:
2405 optval = tp->t_flags & TF_NODELAY;
2406 break;
2407 case TCP_MAXSEG:
2408 optval = tp->t_maxseg;
2409 break;
2410 case TCP_KEEPALIVE:
2411 if (tp->t_keepidle > 0)
2412 optval = tp->t_keepidle / TCP_RETRANSHZ;
2413 else
2414 optval = tcp_keepidle / TCP_RETRANSHZ;
2415 break;
2416 case TCP_KEEPINTVL:
2417 if (tp->t_keepintvl > 0)
2418 optval = tp->t_keepintvl / TCP_RETRANSHZ;
2419 else
2420 optval = tcp_keepintvl / TCP_RETRANSHZ;
2421 break;
2422 case TCP_KEEPCNT:
2423 if (tp->t_keepcnt > 0)
2424 optval = tp->t_keepcnt;
2425 else
2426 optval = tcp_keepcnt;
2427 break;
2428 case TCP_KEEPALIVE_OFFLOAD:
2429 optval = !!(inp->inp_flags2 & INP2_KEEPALIVE_OFFLOAD);
2430 break;
2431 case TCP_NOOPT:
2432 optval = tp->t_flags & TF_NOOPT;
2433 break;
2434 case TCP_NOPUSH:
2435 optval = tp->t_flags & TF_NOPUSH;
2436 break;
2437 case TCP_ENABLE_ECN:
2438 optval = (tp->ecn_flags & TE_ECN_MODE_ENABLE) ? 1 : 0;
2439 break;
2440 case TCP_ECN_MODE:
2441 if (tp->ecn_flags & TE_ECN_MODE_ENABLE)
2442 optval = ECN_MODE_ENABLE;
2443 else if (tp->ecn_flags & TE_ECN_MODE_DISABLE)
2444 optval = ECN_MODE_DISABLE;
2445 else
2446 optval = ECN_MODE_DEFAULT;
2447 break;
2448 case TCP_CONNECTIONTIMEOUT:
2449 optval = tp->t_keepinit / TCP_RETRANSHZ;
2450 break;
2451 case PERSIST_TIMEOUT:
2452 optval = tp->t_persist_timeout / TCP_RETRANSHZ;
2453 break;
2454 case TCP_RXT_CONNDROPTIME:
2455 optval = tp->t_rxt_conndroptime / TCP_RETRANSHZ;
2456 break;
2457 case TCP_RXT_FINDROP:
2458 optval = tp->t_flagsext & TF_RXTFINDROP;
2459 break;
2460 case TCP_NOTIMEWAIT:
2461 optval = (tp->t_flagsext & TF_NOTIMEWAIT) ? 1 : 0;
2462 break;
2463 case TCP_FASTOPEN:
2464 if (tp->t_state != TCPS_LISTEN ||
2465 !(tcp_fastopen & TCP_FASTOPEN_SERVER)) {
2466 error = ENOTSUP;
2467 break;
2468 }
2469 optval = tfo_enabled(tp);
2470 break;
2471 case TCP_FASTOPEN_FORCE_HEURISTICS:
2472 optval = (tp->t_flagsext & TF_FASTOPEN_HEUR) ? 1 : 0;
2473 break;
2474 case TCP_MEASURE_SND_BW:
2475 optval = tp->t_flagsext & TF_MEASURESNDBW;
2476 break;
2477 case TCP_INFO: {
2478 struct tcp_info ti;
2479
2480 tcp_fill_info(tp, &ti);
2481 error = sooptcopyout(sopt, &ti, sizeof(struct tcp_info));
2482 goto done;
2483 /* NOT REACHED */
2484 }
2485 case TCP_CONNECTION_INFO: {
2486 struct tcp_connection_info tci;
2487 tcp_connection_fill_info(tp, &tci);
2488 error = sooptcopyout(sopt, &tci,
2489 sizeof(struct tcp_connection_info));
2490 goto done;
2491 }
2492 case TCP_MEASURE_BW_BURST: {
2493 struct tcp_measure_bw_burst out;
2494 if ((tp->t_flagsext & TF_MEASURESNDBW) == 0 ||
2495 tp->t_bwmeas == NULL) {
2496 error = EINVAL;
2497 break;
2498 }
2499 out.min_burst_size = tp->t_bwmeas->bw_minsizepkts;
2500 out.max_burst_size = tp->t_bwmeas->bw_maxsizepkts;
2501 error = sooptcopyout(sopt, &out, sizeof(out));
2502 goto done;
2503 }
2504 case TCP_NOTSENT_LOWAT:
2505 if ((so->so_flags & SOF_NOTSENT_LOWAT) != 0) {
2506 optval = tp->t_notsent_lowat;
2507 } else {
2508 optval = 0;
2509 }
2510 break;
2511
2512 case TCP_ENABLE_MSGS:
2513 if (so->so_flags & SOF_ENABLE_MSGS) {
2514 optval = 1;
2515 } else {
2516 optval = 0;
2517 }
2518 break;
2519 case TCP_SENDMOREACKS:
2520 if (tp->t_flagsext & TF_NOSTRETCHACK)
2521 optval = 1;
2522 else
2523 optval = 0;
2524 break;
2525 case TCP_DISABLE_BLACKHOLE_DETECTION:
2526 if (tp->t_flagsext & TF_NOBLACKHOLE_DETECTION)
2527 optval = 1;
2528 else
2529 optval = 0;
2530 break;
2531 case TCP_PEER_PID: {
2532 pid_t pid;
2533 error = tcp_lookup_peer_pid_locked(so, &pid);
2534 if (error == 0)
2535 error = sooptcopyout(sopt, &pid, sizeof(pid));
2536 goto done;
2537 }
2538 case TCP_ADAPTIVE_READ_TIMEOUT:
2539 optval = tp->t_adaptive_rtimo;
2540 break;
2541 case TCP_ADAPTIVE_WRITE_TIMEOUT:
2542 optval = tp->t_adaptive_wtimo;
2543 break;
2544 case SO_TRAFFIC_MGT_BACKGROUND:
2545 optval = (so->so_flags1 &
2546 SOF1_TRAFFIC_MGT_SO_BACKGROUND) ? 1 : 0;
2547 break;
2548 case TCP_NOTIFY_ACKNOWLEDGEMENT: {
2549 struct tcp_notify_ack_complete retid;
2550
2551 if (sopt->sopt_valsize != sizeof (retid)) {
2552 error = EINVAL;
2553 break;
2554 }
2555 bzero(&retid, sizeof (retid));
2556 tcp_get_notify_ack_count(tp, &retid);
2557 if (retid.notify_complete_count > 0)
2558 tcp_get_notify_ack_ids(tp, &retid);
2559
2560 error = sooptcopyout(sopt, &retid, sizeof (retid));
2561 goto done;
2562 }
2563 case TCP_RXT_MINIMUM_TIMEOUT:
2564 optval = tp->t_rxt_minimum_timeout / TCP_RETRANSHZ;
2565 break;
2566 default:
2567 error = ENOPROTOOPT;
2568 break;
2569 }
2570 if (error == 0)
2571 error = sooptcopyout(sopt, &optval, sizeof optval);
2572 break;
2573 }
2574 done:
2575 return (error);
2576 }
2577
2578 /*
2579 * tcp_sendspace and tcp_recvspace are the default send and receive window
2580 * sizes, respectively. These are obsolescent (this information should
2581 * be set by the route).
2582 */
2583 u_int32_t tcp_sendspace = 1448*256;
2584 u_int32_t tcp_recvspace = 1448*384;
2585
2586 /* During attach, the size of socket buffer allocated is limited to
2587 * sb_max in sbreserve. Disallow setting the tcp send and recv space
2588 * to be more than sb_max because that will cause tcp_attach to fail
2589 * (see radar 5713060)
2590 */
2591 static int
2592 sysctl_tcp_sospace(struct sysctl_oid *oidp, __unused void *arg1,
2593 int arg2, struct sysctl_req *req)
2594 {
2595 #pragma unused(arg2)
2596 u_int32_t new_value = 0, *space_p = NULL;
2597 int changed = 0, error = 0;
2598 u_quad_t sb_effective_max = (sb_max / (MSIZE+MCLBYTES)) * MCLBYTES;
2599
2600 switch (oidp->oid_number) {
2601 case TCPCTL_SENDSPACE:
2602 space_p = &tcp_sendspace;
2603 break;
2604 case TCPCTL_RECVSPACE:
2605 space_p = &tcp_recvspace;
2606 break;
2607 default:
2608 return EINVAL;
2609 }
2610 error = sysctl_io_number(req, *space_p, sizeof(u_int32_t),
2611 &new_value, &changed);
2612 if (changed) {
2613 if (new_value > 0 && new_value <= sb_effective_max) {
2614 *space_p = new_value;
2615 SYSCTL_SKMEM_UPDATE_AT_OFFSET(arg2, new_value);
2616 } else {
2617 error = ERANGE;
2618 }
2619 }
2620 return error;
2621 }
2622
2623 #if SYSCTL_SKMEM
2624 SYSCTL_PROC(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace,
2625 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_sendspace,
2626 offsetof(skmem_sysctl, tcp.sendspace), sysctl_tcp_sospace,
2627 "IU", "Maximum outgoing TCP datagram size");
2628 SYSCTL_PROC(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace,
2629 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, &tcp_recvspace,
2630 offsetof(skmem_sysctl, tcp.recvspace), sysctl_tcp_sospace,
2631 "IU", "Maximum incoming TCP datagram size");
2632 #else /* SYSCTL_SKMEM */
2633 SYSCTL_PROC(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
2634 &tcp_sendspace , 0, &sysctl_tcp_sospace, "IU", "Maximum outgoing TCP datagram size");
2635 SYSCTL_PROC(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace, CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED,
2636 &tcp_recvspace , 0, &sysctl_tcp_sospace, "IU", "Maximum incoming TCP datagram size");
2637 #endif /* SYSCTL_SKMEM */
2638
2639 /*
2640 * Attach TCP protocol to socket, allocating
2641 * internet protocol control block, tcp control block,
2642 * bufer space, and entering LISTEN state if to accept connections.
2643 *
2644 * Returns: 0 Success
2645 * in_pcballoc:ENOBUFS
2646 * in_pcballoc:ENOMEM
2647 * in_pcballoc:??? [IPSEC specific]
2648 * soreserve:ENOBUFS
2649 */
2650 static int
2651 tcp_attach(struct socket *so, struct proc *p)
2652 {
2653 struct tcpcb *tp;
2654 struct inpcb *inp;
2655 int error;
2656 #if INET6
2657 int isipv6 = SOCK_CHECK_DOM(so, PF_INET6) != 0;
2658 #endif
2659
2660 error = in_pcballoc(so, &tcbinfo, p);
2661 if (error)
2662 return (error);
2663
2664 inp = sotoinpcb(so);
2665
2666 if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
2667 error = soreserve(so, tcp_sendspace, tcp_recvspace);
2668 if (error)
2669 return (error);
2670 }
2671
2672 if (so->so_snd.sb_preconn_hiwat == 0) {
2673 soreserve_preconnect(so, 2048);
2674 }
2675
2676 if ((so->so_rcv.sb_flags & SB_USRSIZE) == 0)
2677 so->so_rcv.sb_flags |= SB_AUTOSIZE;
2678 if ((so->so_snd.sb_flags & SB_USRSIZE) == 0)
2679 so->so_snd.sb_flags |= SB_AUTOSIZE;
2680
2681 #if INET6
2682 if (isipv6) {
2683 inp->inp_vflag |= INP_IPV6;
2684 inp->in6p_hops = -1; /* use kernel default */
2685 }
2686 else
2687 #endif /* INET6 */
2688 inp->inp_vflag |= INP_IPV4;
2689 tp = tcp_newtcpcb(inp);
2690 if (tp == NULL) {
2691 int nofd = so->so_state & SS_NOFDREF; /* XXX */
2692
2693 so->so_state &= ~SS_NOFDREF; /* don't free the socket yet */
2694 #if INET6
2695 if (isipv6)
2696 in6_pcbdetach(inp);
2697 else
2698 #endif /* INET6 */
2699 in_pcbdetach(inp);
2700 so->so_state |= nofd;
2701 return (ENOBUFS);
2702 }
2703 if (nstat_collect)
2704 nstat_tcp_new_pcb(inp);
2705 tp->t_state = TCPS_CLOSED;
2706 return (0);
2707 }
2708
2709 /*
2710 * Initiate (or continue) disconnect.
2711 * If embryonic state, just send reset (once).
2712 * If in ``let data drain'' option and linger null, just drop.
2713 * Otherwise (hard), mark socket disconnecting and drop
2714 * current input data; switch states based on user close, and
2715 * send segment to peer (with FIN).
2716 */
2717 static struct tcpcb *
2718 tcp_disconnect(struct tcpcb *tp)
2719 {
2720 struct socket *so = tp->t_inpcb->inp_socket;
2721
2722 if (so->so_rcv.sb_cc != 0 || tp->t_reassqlen != 0)
2723 return tcp_drop(tp, 0);
2724
2725 if (tp->t_state < TCPS_ESTABLISHED)
2726 tp = tcp_close(tp);
2727 else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
2728 tp = tcp_drop(tp, 0);
2729 else {
2730 soisdisconnecting(so);
2731 sbflush(&so->so_rcv);
2732 tp = tcp_usrclosed(tp);
2733 #if MPTCP
2734 /* A reset has been sent but socket exists, do not send FIN */
2735 if ((so->so_flags & SOF_MP_SUBFLOW) &&
2736 (tp) && (tp->t_mpflags & TMPF_RESET))
2737 return (tp);
2738 #endif
2739 if (tp)
2740 (void) tcp_output(tp);
2741 }
2742 return (tp);
2743 }
2744
2745 /*
2746 * User issued close, and wish to trail through shutdown states:
2747 * if never received SYN, just forget it. If got a SYN from peer,
2748 * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
2749 * If already got a FIN from peer, then almost done; go to LAST_ACK
2750 * state. In all other cases, have already sent FIN to peer (e.g.
2751 * after PRU_SHUTDOWN), and just have to play tedious game waiting
2752 * for peer to send FIN or not respond to keep-alives, etc.
2753 * We can let the user exit from the close as soon as the FIN is acked.
2754 */
2755 static struct tcpcb *
2756 tcp_usrclosed(struct tcpcb *tp)
2757 {
2758 switch (tp->t_state) {
2759
2760 case TCPS_CLOSED:
2761 case TCPS_LISTEN:
2762 tp = tcp_close(tp);
2763 break;
2764
2765 case TCPS_SYN_SENT:
2766 case TCPS_SYN_RECEIVED:
2767 tp->t_flags |= TF_NEEDFIN;
2768 break;
2769
2770 case TCPS_ESTABLISHED:
2771 DTRACE_TCP4(state__change, void, NULL,
2772 struct inpcb *, tp->t_inpcb,
2773 struct tcpcb *, tp,
2774 int32_t, TCPS_FIN_WAIT_1);
2775 tp->t_state = TCPS_FIN_WAIT_1;
2776 break;
2777
2778 case TCPS_CLOSE_WAIT:
2779 DTRACE_TCP4(state__change, void, NULL,
2780 struct inpcb *, tp->t_inpcb,
2781 struct tcpcb *, tp,
2782 int32_t, TCPS_LAST_ACK);
2783 tp->t_state = TCPS_LAST_ACK;
2784 break;
2785 }
2786 if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
2787 soisdisconnected(tp->t_inpcb->inp_socket);
2788 /* To prevent the connection hanging in FIN_WAIT_2 forever. */
2789 if (tp->t_state == TCPS_FIN_WAIT_2)
2790 tp->t_timer[TCPT_2MSL] = OFFSET_FROM_START(tp,
2791 TCP_CONN_MAXIDLE(tp));
2792 }
2793 return (tp);
2794 }
2795
2796 void
2797 tcp_in_cksum_stats(u_int32_t len)
2798 {
2799 tcpstat.tcps_rcv_swcsum++;
2800 tcpstat.tcps_rcv_swcsum_bytes += len;
2801 }
2802
2803 void
2804 tcp_out_cksum_stats(u_int32_t len)
2805 {
2806 tcpstat.tcps_snd_swcsum++;
2807 tcpstat.tcps_snd_swcsum_bytes += len;
2808 }
2809
2810 #if INET6
2811 void
2812 tcp_in6_cksum_stats(u_int32_t len)
2813 {
2814 tcpstat.tcps_rcv6_swcsum++;
2815 tcpstat.tcps_rcv6_swcsum_bytes += len;
2816 }
2817
2818 void
2819 tcp_out6_cksum_stats(u_int32_t len)
2820 {
2821 tcpstat.tcps_snd6_swcsum++;
2822 tcpstat.tcps_snd6_swcsum_bytes += len;
2823 }
2824
2825 /*
2826 * When messages are enabled on a TCP socket, the message priority
2827 * is sent as a control message. This function will extract it.
2828 */
2829 int
2830 tcp_get_msg_priority(struct mbuf *control, uint32_t *msgpri)
2831 {
2832 struct cmsghdr *cm;
2833 if (control == NULL)
2834 return(EINVAL);
2835
2836 for (cm = M_FIRST_CMSGHDR(control); cm;
2837 cm = M_NXT_CMSGHDR(control, cm)) {
2838 if (cm->cmsg_len < sizeof(struct cmsghdr) ||
2839 cm->cmsg_len > control->m_len) {
2840 return (EINVAL);
2841 }
2842 if (cm->cmsg_level == SOL_SOCKET &&
2843 cm->cmsg_type == SCM_MSG_PRIORITY) {
2844 *msgpri = *(unsigned int *)(void *)CMSG_DATA(cm);
2845 break;
2846 }
2847 }
2848
2849 VERIFY(*msgpri >= MSG_PRI_MIN && *msgpri <= MSG_PRI_MAX);
2850 return (0);
2851 }
2852 #endif /* INET6 */
mirror of XNU