1 /*
2 * Copyright (c) 2003-2021 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) 1995, 1996, 1997, and 1998 WIDE Project.
30 * 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. Neither the name of the project nor the names of its contributors
41 * may be used to endorse or promote products derived from this software
42 * without specific prior written permission.
43 *
44 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
45 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
46 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
47 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
48 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
49 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
50 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
51 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
52 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
53 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
54 * SUCH DAMAGE.
55 *
56 */
57
58 /*
59 * Copyright (c) 1982, 1986, 1991, 1993
60 * The Regents of the University of California. All rights reserved.
61 *
62 * Redistribution and use in source and binary forms, with or without
63 * modification, are permitted provided that the following conditions
64 * are met:
65 * 1. Redistributions of source code must retain the above copyright
66 * notice, this list of conditions and the following disclaimer.
67 * 2. Redistributions in binary form must reproduce the above copyright
68 * notice, this list of conditions and the following disclaimer in the
69 * documentation and/or other materials provided with the distribution.
70 * 3. All advertising materials mentioning features or use of this software
71 * must display the following acknowledgement:
72 * This product includes software developed by the University of
73 * California, Berkeley and its contributors.
74 * 4. Neither the name of the University nor the names of its contributors
75 * may be used to endorse or promote products derived from this software
76 * without specific prior written permission.
77 *
78 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
79 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
80 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
81 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
82 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
83 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
84 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
85 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
86 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
87 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
88 * SUCH DAMAGE.
89 *
90 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94
91 */
92
93 #include <sys/param.h>
94 #include <sys/systm.h>
95 #include <sys/malloc.h>
96 #include <sys/mbuf.h>
97 #include <sys/domain.h>
98 #include <sys/protosw.h>
99 #include <sys/socket.h>
100 #include <sys/socketvar.h>
101 #include <sys/sockio.h>
102 #include <sys/errno.h>
103 #include <sys/time.h>
104 #include <sys/proc.h>
105 #include <sys/kauth.h>
106 #include <sys/priv.h>
107
108 #include <net/if.h>
109 #include <net/if_types.h>
110 #include <net/route.h>
111 #include <net/ntstat.h>
112 #include <net/restricted_in_port.h>
113
114 #include <netinet/in.h>
115 #include <netinet/in_var.h>
116 #include <netinet/in_systm.h>
117 #include <netinet/ip6.h>
118 #include <netinet/ip_var.h>
119
120 #include <netinet6/ip6_var.h>
121 #include <netinet6/nd6.h>
122 #include <netinet/in_pcb.h>
123 #include <netinet6/in6_pcb.h>
124
125 #include <net/if_types.h>
126 #include <net/if_var.h>
127
128 #include <kern/kern_types.h>
129 #include <kern/zalloc.h>
130
131 #if IPSEC
132 #include <netinet6/ipsec.h>
133 #include <netinet6/ipsec6.h>
134 #include <netinet6/ah.h>
135 #include <netinet6/ah6.h>
136 #include <netkey/key.h>
137 #endif /* IPSEC */
138
139 #if NECP
140 #include <net/necp.h>
141 #endif /* NECP */
142
143 #include <net/sockaddr_utils.h>
144
145 /*
146 * in6_pcblookup_local_and_cleanup does everything
147 * in6_pcblookup_local does but it checks for a socket
148 * that's going away. Since we know that the lock is
149 * held read+write when this function is called, we
150 * can safely dispose of this socket like the slow
151 * timer would usually do and return NULL. This is
152 * great for bind.
153 */
154 static struct inpcb *
in6_pcblookup_local_and_cleanup(struct inpcbinfo * pcbinfo,struct in6_addr * laddr,u_int lport_arg,uint32_t ifscope,int wild_okay)155 in6_pcblookup_local_and_cleanup(struct inpcbinfo *pcbinfo,
156 struct in6_addr *laddr, u_int lport_arg, uint32_t ifscope, int wild_okay)
157 {
158 struct inpcb *__single inp;
159
160 /* Perform normal lookup */
161 inp = in6_pcblookup_local(pcbinfo, laddr, lport_arg, ifscope, wild_okay);
162
163 /* Check if we found a match but it's waiting to be disposed */
164 if (inp != NULL && inp->inp_wantcnt == WNT_STOPUSING) {
165 struct socket *so = inp->inp_socket;
166
167 socket_lock(so, 0);
168
169 if (so->so_usecount == 0) {
170 if (inp->inp_state != INPCB_STATE_DEAD) {
171 in6_pcbdetach(inp);
172 }
173 in_pcbdispose(inp); /* will unlock & destroy */
174 inp = NULL;
175 } else {
176 socket_unlock(so, 0);
177 }
178 }
179
180 return inp;
181 }
182
183 /*
184 * Bind an INPCB to an address and/or port. This routine should not alter
185 * the caller-supplied local address "nam" or remote address "remote".
186 */
187 int
in6_pcbbind(struct inpcb * inp,struct sockaddr * nam,struct sockaddr * remote,struct proc * p)188 in6_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct sockaddr *remote, struct proc *p)
189 {
190 struct socket *__single so = inp->inp_socket;
191 struct inpcbinfo *__single pcbinfo = inp->inp_pcbinfo;
192 u_short lport = 0;
193 int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
194 ifnet_ref_t outif = NULL;
195 struct sockaddr_in6 sin6;
196 uint32_t lifscope = IFSCOPE_NONE;
197 int error = 0;
198 #if XNU_TARGET_OS_OSX
199 kauth_cred_t __single cred;
200 #endif /* XNU_TARGET_OS_OSX */
201
202 if (inp->inp_flags2 & INP2_BIND_IN_PROGRESS) {
203 return EINVAL;
204 }
205 inp->inp_flags2 |= INP2_BIND_IN_PROGRESS;
206
207 if (TAILQ_EMPTY(&in6_ifaddrhead)) { /* XXX broken! */
208 error = EADDRNOTAVAIL;
209 goto done;
210 }
211 if (!(so->so_options & (SO_REUSEADDR | SO_REUSEPORT))) {
212 wild = 1;
213 }
214
215 in_pcb_check_management_entitled(inp);
216 in_pcb_check_ultra_constrained_entitled(inp);
217
218 socket_unlock(so, 0); /* keep reference */
219 lck_rw_lock_exclusive(&pcbinfo->ipi_lock);
220 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
221 /* another thread completed the bind */
222 lck_rw_done(&pcbinfo->ipi_lock);
223 socket_lock(so, 0);
224 error = EINVAL;
225 goto done;
226 }
227
228 SOCKADDR_ZERO(&sin6, sizeof(sin6));
229 if (nam != NULL) {
230 if (nam->sa_len != sizeof(struct sockaddr_in6)) {
231 lck_rw_done(&pcbinfo->ipi_lock);
232 socket_lock(so, 0);
233 error = EINVAL;
234 goto done;
235 }
236 /*
237 * family check.
238 */
239 if (nam->sa_family != AF_INET6) {
240 lck_rw_done(&pcbinfo->ipi_lock);
241 socket_lock(so, 0);
242 error = EAFNOSUPPORT;
243 goto done;
244 }
245 lport = SIN6(nam)->sin6_port;
246
247 *(&sin6) = *SIN6(nam);
248
249 /* KAME hack: embed scopeid */
250 if (in6_embedscope(&sin6.sin6_addr, &sin6, inp, NULL,
251 NULL, &lifscope) != 0) {
252 lck_rw_done(&pcbinfo->ipi_lock);
253 socket_lock(so, 0);
254 error = EINVAL;
255 goto done;
256 }
257
258 /* Sanitize local copy for address searches */
259 sin6.sin6_flowinfo = 0;
260 sin6.sin6_port = 0;
261 if (in6_embedded_scope) {
262 sin6.sin6_scope_id = 0;
263 }
264
265 if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
266 /*
267 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
268 * allow compepte duplication of binding if
269 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
270 * and a multicast address is bound on both
271 * new and duplicated sockets.
272 */
273 if (so->so_options & SO_REUSEADDR) {
274 reuseport = SO_REUSEADDR | SO_REUSEPORT;
275 }
276 } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr)) {
277 struct ifaddr *__single ifa;
278
279 ifa = ifa_ifwithaddr(SA(&sin6));
280 if (ifa == NULL) {
281 lck_rw_done(&pcbinfo->ipi_lock);
282 socket_lock(so, 0);
283 error = EADDRNOTAVAIL;
284 goto done;
285 } else {
286 /*
287 * XXX: bind to an anycast address might
288 * accidentally cause sending a packet with
289 * anycast source address. We should allow
290 * to bind to a deprecated address, since
291 * the application dare to use it.
292 */
293 IFA_LOCK_SPIN(ifa);
294 if ((ifatoia6(ifa))->ia6_flags &
295 (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY |
296 IN6_IFF_DETACHED | IN6_IFF_CLAT46)) {
297 IFA_UNLOCK(ifa);
298 ifa_remref(ifa);
299 lck_rw_done(&pcbinfo->ipi_lock);
300 socket_lock(so, 0);
301 error = EADDRNOTAVAIL;
302 goto done;
303 }
304 /*
305 * Opportunistically determine the outbound
306 * interface that may be used; this may not
307 * hold true if we end up using a route
308 * going over a different interface, e.g.
309 * when sending to a local address. This
310 * will get updated again after sending.
311 */
312 outif = ifa->ifa_ifp;
313 IFA_UNLOCK(ifa);
314 ifa_remref(ifa);
315 }
316 }
317
318 #if SKYWALK
319 if (inp->inp_flags2 & INP2_EXTERNAL_PORT) {
320 // Extract the external flow info
321 struct ns_flow_info nfi = {};
322 int netns_error = necp_client_get_netns_flow_info(inp->necp_client_uuid,
323 &nfi);
324 if (netns_error != 0) {
325 lck_rw_done(&pcbinfo->ipi_lock);
326 socket_lock(so, 0);
327 error = netns_error;
328 goto done;
329 }
330
331 // Extract the reserved port
332 u_int16_t reserved_lport = 0;
333 if (nfi.nfi_laddr.sa.sa_family == AF_INET) {
334 reserved_lport = nfi.nfi_laddr.sin.sin_port;
335 } else if (nfi.nfi_laddr.sa.sa_family == AF_INET6) {
336 reserved_lport = nfi.nfi_laddr.sin6.sin6_port;
337 } else {
338 lck_rw_done(&pcbinfo->ipi_lock);
339 socket_lock(so, 0);
340 error = EINVAL;
341 goto done;
342 }
343
344 // Validate or use the reserved port
345 if (lport == 0) {
346 lport = reserved_lport;
347 } else if (lport != reserved_lport) {
348 lck_rw_done(&pcbinfo->ipi_lock);
349 socket_lock(so, 0);
350 error = EINVAL;
351 goto done;
352 }
353 }
354
355 /* Do not allow reserving a UDP port if remaining UDP port count is below 4096 */
356 if (SOCK_PROTO(so) == IPPROTO_UDP && !allow_udp_port_exhaustion) {
357 uint32_t current_reservations = 0;
358 current_reservations = netns_lookup_reservations_count_in6(inp->in6p_laddr, IPPROTO_UDP);
359 if (USHRT_MAX - UDP_RANDOM_PORT_RESERVE < current_reservations) {
360 log(LOG_ERR, "UDP port not available, less than 4096 UDP ports left");
361 lck_rw_done(&pcbinfo->ipi_lock);
362 socket_lock(so, 0);
363 error = EADDRNOTAVAIL;
364 goto done;
365 }
366 }
367
368 #endif /* SKYWALK */
369
370 if (lport != 0) {
371 struct inpcb *__single t;
372 uid_t u;
373
374 #if XNU_TARGET_OS_OSX
375 if (ntohs(lport) < IPV6PORT_RESERVED &&
376 !IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr) &&
377 !(inp->inp_flags2 & INP2_EXTERNAL_PORT)) {
378 cred = kauth_cred_proc_ref(p);
379 error = priv_check_cred(cred,
380 PRIV_NETINET_RESERVEDPORT, 0);
381 kauth_cred_unref(&cred);
382 if (error != 0) {
383 lck_rw_done(&pcbinfo->ipi_lock);
384 socket_lock(so, 0);
385 error = EACCES;
386 goto done;
387 }
388 }
389 #endif /* XNU_TARGET_OS_OSX */
390 /*
391 * Check wether the process is allowed to bind to a restricted port
392 */
393 if (!current_task_can_use_restricted_in_port(lport,
394 (uint8_t)SOCK_PROTO(so), PORT_FLAGS_BSD)) {
395 lck_rw_done(&pcbinfo->ipi_lock);
396 socket_lock(so, 0);
397 error = EADDRINUSE;
398 goto done;
399 }
400
401 if (!IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr) &&
402 (u = kauth_cred_getuid(so->so_cred)) != 0) {
403 t = in6_pcblookup_local_and_cleanup(pcbinfo,
404 &sin6.sin6_addr, lport, sin6.sin6_scope_id,
405 INPLOOKUP_WILDCARD);
406 if (t != NULL &&
407 (!IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr) ||
408 !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
409 !(t->inp_socket->so_options & SO_REUSEPORT)) &&
410 (u != kauth_cred_getuid(t->inp_socket->so_cred)) &&
411 !(t->inp_socket->so_flags & SOF_REUSESHAREUID) &&
412 (!(t->inp_flags2 & INP2_EXTERNAL_PORT) ||
413 !(inp->inp_flags2 & INP2_EXTERNAL_PORT) ||
414 uuid_compare(t->necp_client_uuid, inp->necp_client_uuid) != 0)) {
415 lck_rw_done(&pcbinfo->ipi_lock);
416 socket_lock(so, 0);
417 error = EADDRINUSE;
418 goto done;
419 }
420 if (!(inp->inp_flags & IN6P_IPV6_V6ONLY) &&
421 IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr)) {
422 struct sockaddr_in sin;
423
424 in6_sin6_2_sin(&sin, &sin6);
425 t = in_pcblookup_local_and_cleanup(
426 pcbinfo, sin.sin_addr, lport,
427 INPLOOKUP_WILDCARD);
428 if (t != NULL &&
429 !(t->inp_socket->so_options & SO_REUSEPORT) &&
430 (kauth_cred_getuid(so->so_cred) !=
431 kauth_cred_getuid(t->inp_socket->so_cred)) &&
432 (t->inp_laddr.s_addr != INADDR_ANY ||
433 SOCK_DOM(so) == SOCK_DOM(t->inp_socket)) &&
434 (!(t->inp_flags2 & INP2_EXTERNAL_PORT) ||
435 !(inp->inp_flags2 & INP2_EXTERNAL_PORT) ||
436 uuid_compare(t->necp_client_uuid, inp->necp_client_uuid) != 0)) {
437 lck_rw_done(&pcbinfo->ipi_lock);
438 socket_lock(so, 0);
439 error = EADDRINUSE;
440 goto done;
441 }
442
443 #if SKYWALK
444 VERIFY(!NETNS_TOKEN_VALID(
445 &inp->inp_wildcard_netns_token));
446 if ((SOCK_PROTO(so) == IPPROTO_TCP ||
447 SOCK_PROTO(so) == IPPROTO_UDP) &&
448 !(inp->inp_flags2 & INP2_EXTERNAL_PORT)) {
449 if (netns_reserve_in(&inp->
450 inp_wildcard_netns_token,
451 sin.sin_addr,
452 (uint8_t)SOCK_PROTO(so), lport,
453 NETNS_BSD, NULL) != 0) {
454 lck_rw_done(&pcbinfo->ipi_lock);
455 socket_lock(so, 0);
456 error = EADDRINUSE;
457 goto done;
458 }
459 }
460 #endif /* SKYWALK */
461 }
462 }
463 t = in6_pcblookup_local_and_cleanup(pcbinfo,
464 &sin6.sin6_addr, lport, sin6.sin6_scope_id, wild);
465 if (t != NULL &&
466 (reuseport & t->inp_socket->so_options) == 0 &&
467 (!(t->inp_flags2 & INP2_EXTERNAL_PORT) ||
468 !(inp->inp_flags2 & INP2_EXTERNAL_PORT) ||
469 uuid_compare(t->necp_client_uuid, inp->necp_client_uuid) != 0)) {
470 #if SKYWALK
471 netns_release(&inp->inp_wildcard_netns_token);
472 #endif /* SKYWALK */
473 lck_rw_done(&pcbinfo->ipi_lock);
474 socket_lock(so, 0);
475 error = EADDRINUSE;
476 goto done;
477 }
478 if (!(inp->inp_flags & IN6P_IPV6_V6ONLY) &&
479 IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr)) {
480 struct sockaddr_in sin;
481
482 in6_sin6_2_sin(&sin, &sin6);
483 t = in_pcblookup_local_and_cleanup(pcbinfo,
484 sin.sin_addr, lport, wild);
485 if (t != NULL && (reuseport &
486 t->inp_socket->so_options) == 0 &&
487 (t->inp_laddr.s_addr != INADDR_ANY ||
488 SOCK_DOM(so) == SOCK_DOM(t->inp_socket)) &&
489 (!(t->inp_flags2 & INP2_EXTERNAL_PORT) ||
490 !(inp->inp_flags2 & INP2_EXTERNAL_PORT) ||
491 uuid_compare(t->necp_client_uuid, inp->necp_client_uuid) != 0)) {
492 #if SKYWALK
493 netns_release(&inp->inp_wildcard_netns_token);
494 #endif /* SKYWALK */
495 lck_rw_done(&pcbinfo->ipi_lock);
496 socket_lock(so, 0);
497 error = EADDRINUSE;
498 goto done;
499 }
500 #if SKYWALK
501 if ((SOCK_PROTO(so) == IPPROTO_TCP ||
502 SOCK_PROTO(so) == IPPROTO_UDP) &&
503 !(inp->inp_flags2 & INP2_EXTERNAL_PORT) &&
504 (!NETNS_TOKEN_VALID(
505 &inp->inp_wildcard_netns_token))) {
506 if (netns_reserve_in(&inp->
507 inp_wildcard_netns_token,
508 sin.sin_addr,
509 (uint8_t)SOCK_PROTO(so), lport,
510 NETNS_BSD, NULL) != 0) {
511 lck_rw_done(&pcbinfo->ipi_lock);
512 socket_lock(so, 0);
513 error = EADDRINUSE;
514 goto done;
515 }
516 }
517 #endif /* SKYWALK */
518 }
519 #if SKYWALK
520 if ((SOCK_PROTO(so) == IPPROTO_TCP ||
521 SOCK_PROTO(so) == IPPROTO_UDP) &&
522 !(inp->inp_flags2 & INP2_EXTERNAL_PORT)) {
523 if (netns_reserve_in6(&inp->inp_netns_token,
524 sin6.sin6_addr, (uint8_t)SOCK_PROTO(so), lport,
525 NETNS_BSD, NULL) != 0) {
526 netns_release(&inp->inp_wildcard_netns_token);
527 lck_rw_done(&pcbinfo->ipi_lock);
528 socket_lock(so, 0);
529 error = EADDRINUSE;
530 goto done;
531 }
532 }
533 #endif /* SKYWALK */
534 }
535 }
536
537 socket_lock(so, 0);
538 /*
539 * We unlocked socket's protocol lock for a long time.
540 * The socket might have been dropped/defuncted.
541 * Checking if world has changed since.
542 */
543 if (inp->inp_state == INPCB_STATE_DEAD) {
544 #if SKYWALK
545 netns_release(&inp->inp_netns_token);
546 netns_release(&inp->inp_wildcard_netns_token);
547 #endif /* SKYWALK */
548 lck_rw_done(&pcbinfo->ipi_lock);
549 error = ECONNABORTED;
550 goto done;
551 }
552
553 /* check if the socket got bound when the lock was released */
554 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
555 #if SKYWALK
556 netns_release(&inp->inp_netns_token);
557 netns_release(&inp->inp_wildcard_netns_token);
558 #endif /* SKYWALK */
559 lck_rw_done(&pcbinfo->ipi_lock);
560 error = EINVAL;
561 goto done;
562 }
563
564 if (!IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr)) {
565 inp->in6p_laddr = sin6.sin6_addr;
566 inp->in6p_last_outifp = outif;
567 inp->inp_lifscope = lifscope;
568 in6_verify_ifscope(&inp->in6p_laddr, lifscope);
569 #if SKYWALK
570 if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
571 netns_set_ifnet(&inp->inp_netns_token,
572 inp->in6p_last_outifp);
573 }
574 #endif /* SKYWALK */
575 }
576
577 if (lport == 0) {
578 int e;
579 if ((e = in6_pcbsetport(&inp->in6p_laddr, remote, inp, p, 1)) != 0) {
580 /* Undo any address bind from above. */
581 #if SKYWALK
582 netns_release(&inp->inp_netns_token);
583 netns_release(&inp->inp_wildcard_netns_token);
584 #endif /* SKYWALK */
585 inp->in6p_laddr = in6addr_any;
586 inp->in6p_last_outifp = NULL;
587 inp->inp_lifscope = IFSCOPE_NONE;
588 lck_rw_done(&pcbinfo->ipi_lock);
589 error = e;
590 goto done;
591 }
592 } else {
593 inp->inp_lport = lport;
594 if (in_pcbinshash(inp, remote, 1) != 0) {
595 #if SKYWALK
596 netns_release(&inp->inp_netns_token);
597 netns_release(&inp->inp_wildcard_netns_token);
598 #endif /* SKYWALK */
599 inp->in6p_laddr = in6addr_any;
600 inp->inp_lifscope = IFSCOPE_NONE;
601 inp->inp_lport = 0;
602 inp->in6p_last_outifp = NULL;
603 lck_rw_done(&pcbinfo->ipi_lock);
604 error = EAGAIN;
605 goto done;
606 }
607 }
608 lck_rw_done(&pcbinfo->ipi_lock);
609 sflt_notify(so, sock_evt_bound, NULL);
610 done:
611 inp->inp_flags2 &= ~INP2_BIND_IN_PROGRESS;
612 return error;
613 }
614
615 /*
616 * Transform old in6_pcbconnect() into an inner subroutine for new
617 * in6_pcbconnect(); do some validity-checking on the remote address
618 * (in "nam") and then determine local host address (i.e., which
619 * interface) to use to access that remote host.
620 *
621 * This routine may alter the caller-supplied remote address "nam".
622 *
623 * This routine might return an ifp with a reference held if the caller
624 * provides a non-NULL outif, even in the error case. The caller is
625 * responsible for releasing its reference.
626 */
627 int
in6_pcbladdr(struct inpcb * inp,struct sockaddr * nam,struct in6_addr * plocal_addr6,struct ifnet ** outif)628 in6_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
629 struct in6_addr *plocal_addr6, struct ifnet **outif)
630 {
631 struct in6_addr *__single addr6 = NULL;
632 struct in6_addr src_storage;
633 int error = 0;
634 unsigned int ifscope;
635
636 if (outif != NULL) {
637 *outif = NULL;
638 }
639 if (nam->sa_len != sizeof(struct sockaddr_in6)) {
640 return EINVAL;
641 }
642 if (SIN6(nam)->sin6_family != AF_INET6) {
643 return EAFNOSUPPORT;
644 }
645 if (SIN6(nam)->sin6_port == 0) {
646 return EADDRNOTAVAIL;
647 }
648
649 /* KAME hack: embed scopeid */
650 if (in6_embedscope(&SIN6(nam)->sin6_addr, SIN6(nam), inp, NULL, NULL, IN6_NULL_IF_EMBEDDED_SCOPE(&SIN6(nam)->sin6_scope_id)) != 0) {
651 return EINVAL;
652 }
653
654 in_pcb_check_management_entitled(inp);
655 in_pcb_check_ultra_constrained_entitled(inp);
656
657 if (!TAILQ_EMPTY(&in6_ifaddrhead)) {
658 /*
659 * If the destination address is UNSPECIFIED addr,
660 * use the loopback addr, e.g ::1.
661 */
662 if (IN6_IS_ADDR_UNSPECIFIED(&SIN6(nam)->sin6_addr)) {
663 SIN6(nam)->sin6_addr = in6addr_loopback;
664 }
665 }
666
667 ifscope = (inp->inp_flags & INP_BOUND_IF) ?
668 inp->inp_boundifp->if_index : IFSCOPE_NONE;
669
670 /*
671 * XXX: in6_selectsrc might replace the bound local address
672 * with the address specified by setsockopt(IPV6_PKTINFO).
673 * Is it the intended behavior?
674 *
675 * in6_selectsrc() might return outif with its reference held
676 * even in the error case; caller always needs to release it
677 * if non-NULL.
678 */
679 addr6 = in6_selectsrc(SIN6(nam), inp->in6p_outputopts, inp,
680 &inp->in6p_route, outif, &src_storage, ifscope, &error);
681
682 if (outif != NULL) {
683 rtentry_ref_t rt = inp->in6p_route.ro_rt;
684 /*
685 * If in6_selectsrc() returns a route, it should be one
686 * which points to the same ifp as outif. Just in case
687 * it isn't, use the one from the route for consistency.
688 * Otherwise if there is no route, leave outif alone as
689 * it could still be useful to the caller.
690 */
691 if (rt != NULL && rt->rt_ifp != *outif) {
692 ifnet_reference(rt->rt_ifp); /* for caller */
693 if (*outif != NULL) {
694 ifnet_release(*outif);
695 }
696 *outif = rt->rt_ifp;
697 }
698 }
699
700 if (addr6 == NULL) {
701 if (outif != NULL && (*outif) != NULL &&
702 inp_restricted_send(inp, *outif)) {
703 soevent(inp->inp_socket,
704 (SO_FILT_HINT_LOCKED | SO_FILT_HINT_IFDENIED));
705 error = EHOSTUNREACH;
706 }
707 if (error == 0) {
708 error = EADDRNOTAVAIL;
709 }
710 return error;
711 }
712
713 *plocal_addr6 = *addr6;
714 /*
715 * Don't do pcblookup call here; return interface in
716 * plocal_addr6 and exit to caller, that will do the lookup.
717 */
718 return 0;
719 }
720
721 /*
722 * Outer subroutine:
723 * Connect from a socket to a specified address.
724 * Both address and port must be specified in argument sin.
725 * If don't have a local address for this socket yet,
726 * then pick one.
727 */
728 int
in6_pcbconnect(struct inpcb * inp,struct sockaddr * nam,struct proc * p)729 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct proc *p)
730 {
731 struct in6_addr addr6;
732 struct sockaddr_in6 *__single sin6 = SIN6(nam);
733 struct inpcb *__single pcb;
734 int error = 0;
735 ifnet_ref_t outif = NULL;
736 struct socket *so = inp->inp_socket;
737
738 #if CONTENT_FILTER
739 so->so_state_change_cnt++;
740 #endif
741
742 if (SOCK_CHECK_PROTO(so, IPPROTO_UDP) &&
743 sin6->sin6_port == htons(53) && !(so->so_flags1 & SOF1_DNS_COUNTED)) {
744 so->so_flags1 |= SOF1_DNS_COUNTED;
745 INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_dns);
746 }
747
748 /*
749 * Call inner routine, to assign local interface address.
750 * in6_pcbladdr() may automatically fill in sin6_scope_id.
751 *
752 * in6_pcbladdr() might return an ifp with its reference held
753 * even in the error case, so make sure that it's released
754 * whenever it's non-NULL.
755 */
756 if ((error = in6_pcbladdr(inp, nam, &addr6, &outif)) != 0) {
757 if (outif != NULL && inp_restricted_send(inp, outif)) {
758 soevent(so,
759 (SO_FILT_HINT_LOCKED | SO_FILT_HINT_IFDENIED));
760 }
761 goto done;
762 }
763 socket_unlock(so, 0);
764
765 uint32_t lifscope;
766 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
767 lifscope = inp->inp_lifscope;
768 } else if (outif != NULL) {
769 lifscope = in6_addr2scopeid(outif, &addr6);
770 } else {
771 lifscope = sin6->sin6_scope_id;
772 }
773
774 pcb = in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr,
775 sin6->sin6_port, sin6->sin6_scope_id, IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ?
776 &addr6 : &inp->in6p_laddr, inp->inp_lport, lifscope, 0, NULL);
777 socket_lock(so, 0);
778 if (pcb != NULL) {
779 in_pcb_checkstate(pcb, WNT_RELEASE, pcb == inp ? 1 : 0);
780 error = EADDRINUSE;
781 goto done;
782 }
783 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
784 if (inp->inp_lport == 0) {
785 error = in6_pcbbind(inp, NULL, nam, p);
786 if (error) {
787 goto done;
788 }
789 }
790 inp->in6p_laddr = addr6;
791 inp->in6p_last_outifp = outif; /* no reference needed */
792 if (IN6_IS_SCOPE_EMBED(&inp->in6p_laddr) &&
793 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sin6->sin6_addr)) {
794 inp->inp_lifscope = sin6->sin6_scope_id;
795 } else {
796 inp->inp_lifscope = lifscope;
797 }
798 in6_verify_ifscope(&inp->in6p_laddr, inp->inp_lifscope);
799 #if SKYWALK
800 if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
801 netns_set_ifnet(&inp->inp_netns_token,
802 inp->in6p_last_outifp);
803 }
804 #endif /* SKYWALK */
805 inp->in6p_flags |= INP_IN6ADDR_ANY;
806 }
807 if (!lck_rw_try_lock_exclusive(&inp->inp_pcbinfo->ipi_lock)) {
808 /* lock inversion issue, mostly with udp multicast packets */
809 socket_unlock(so, 0);
810 lck_rw_lock_exclusive(&inp->inp_pcbinfo->ipi_lock);
811 socket_lock(so, 0);
812 }
813 inp->in6p_faddr = sin6->sin6_addr;
814 inp->inp_fport = sin6->sin6_port;
815 inp->inp_fifscope = sin6->sin6_scope_id;
816 in6_verify_ifscope(&inp->in6p_faddr, inp->inp_fifscope);
817 if (nstat_collect && SOCK_PROTO(so) == IPPROTO_UDP) {
818 nstat_pcb_invalidate_cache(inp);
819 }
820 in_pcbrehash(inp);
821 lck_rw_done(&inp->inp_pcbinfo->ipi_lock);
822
823 done:
824 if (outif != NULL) {
825 ifnet_release(outif);
826 }
827
828 return error;
829 }
830
831 void
in6_pcbdisconnect(struct inpcb * inp)832 in6_pcbdisconnect(struct inpcb *inp)
833 {
834 struct socket *__single so = inp->inp_socket;
835
836 #if CONTENT_FILTER
837 if (so) {
838 so->so_state_change_cnt++;
839 }
840 #endif
841
842 if (!lck_rw_try_lock_exclusive(&inp->inp_pcbinfo->ipi_lock)) {
843 /* lock inversion issue, mostly with udp multicast packets */
844 socket_unlock(so, 0);
845 lck_rw_lock_exclusive(&inp->inp_pcbinfo->ipi_lock);
846 socket_lock(so, 0);
847 }
848 if (nstat_collect && SOCK_PROTO(so) == IPPROTO_UDP) {
849 nstat_pcb_cache(inp);
850 }
851 bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
852 inp->inp_fport = 0;
853 /* clear flowinfo - RFC 6437 */
854 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
855 in_pcbrehash(inp);
856 lck_rw_done(&inp->inp_pcbinfo->ipi_lock);
857 /*
858 * A multipath subflow socket would have its SS_NOFDREF set by default,
859 * so check for SOF_MP_SUBFLOW socket flag before detaching the PCB;
860 * when the socket is closed for real, SOF_MP_SUBFLOW would be cleared.
861 */
862 if (!(so->so_flags & SOF_MP_SUBFLOW) && (so->so_state & SS_NOFDREF)) {
863 in6_pcbdetach(inp);
864 }
865 }
866
867 void
in6_pcbdetach(struct inpcb * inp)868 in6_pcbdetach(struct inpcb *inp)
869 {
870 struct socket *__single so = inp->inp_socket;
871
872 if (so->so_pcb == NULL) {
873 /* PCB has been disposed */
874 panic("%s: inp=%p so=%p proto=%d so_pcb is null!", __func__,
875 inp, so, SOCK_PROTO(so));
876 /* NOTREACHED */
877 }
878
879 #if IPSEC
880 if (inp->in6p_sp != NULL) {
881 (void) ipsec6_delete_pcbpolicy(inp);
882 }
883 #endif /* IPSEC */
884
885 if (inp->inp_stat != NULL && SOCK_PROTO(so) == IPPROTO_UDP) {
886 if (inp->inp_stat->rxpackets == 0 && inp->inp_stat->txpackets == 0) {
887 INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet6_dgram_no_data);
888 }
889 }
890
891 /*
892 * Let NetworkStatistics know this PCB is going away
893 * before we detach it.
894 */
895 if (nstat_collect &&
896 (SOCK_PROTO(so) == IPPROTO_TCP || SOCK_PROTO(so) == IPPROTO_UDP)) {
897 nstat_pcb_detach(inp);
898 }
899 /* mark socket state as dead */
900 if (in_pcb_checkstate(inp, WNT_STOPUSING, 1) != WNT_STOPUSING) {
901 panic("%s: so=%p proto=%d couldn't set to STOPUSING",
902 __func__, so, SOCK_PROTO(so));
903 /* NOTREACHED */
904 }
905
906 #if SKYWALK
907 /* Free up the port in the namespace registrar if not in TIME_WAIT */
908 if (!(inp->inp_flags2 & INP2_TIMEWAIT)) {
909 netns_release(&inp->inp_netns_token);
910 netns_release(&inp->inp_wildcard_netns_token);
911 }
912 #endif /* SKYWALK */
913
914 if (!(so->so_flags & SOF_PCBCLEARING)) {
915 struct ip_moptions *__single imo;
916 struct ip6_moptions *__single im6o;
917
918 inp->inp_vflag = 0;
919 if (inp->in6p_options != NULL) {
920 m_freem(inp->in6p_options);
921 inp->in6p_options = NULL;
922 }
923 ip6_freepcbopts(inp->in6p_outputopts);
924 inp->in6p_outputopts = NULL;
925 ROUTE_RELEASE(&inp->in6p_route);
926 /* free IPv4 related resources in case of mapped addr */
927 if (inp->inp_options != NULL) {
928 (void) m_free(inp->inp_options);
929 inp->inp_options = NULL;
930 }
931 im6o = inp->in6p_moptions;
932 inp->in6p_moptions = NULL;
933 if (im6o != NULL) {
934 IM6O_REMREF(im6o);
935 }
936 imo = inp->inp_moptions;
937 inp->inp_moptions = NULL;
938 if (imo != NULL) {
939 IMO_REMREF(imo);
940 }
941
942 sofreelastref(so, 0);
943 inp->inp_state = INPCB_STATE_DEAD;
944 /* makes sure we're not called twice from so_close */
945 so->so_flags |= SOF_PCBCLEARING;
946
947 inpcb_gc_sched(inp->inp_pcbinfo, INPCB_TIMER_FAST);
948 }
949 }
950
951 struct sockaddr *
in6_sockaddr(in_port_t port,struct in6_addr * addr_p,uint32_t ifscope)952 in6_sockaddr(in_port_t port, struct in6_addr *addr_p, uint32_t ifscope)
953 {
954 struct sockaddr_in6 *__single sin6;
955
956 sin6 = (struct sockaddr_in6 *)(alloc_sockaddr(sizeof(*sin6),
957 Z_WAITOK | Z_NOFAIL));
958
959 sin6->sin6_family = AF_INET6;
960 sin6->sin6_port = port;
961 sin6->sin6_addr = *addr_p;
962
963 /* would be good to use sa6_recoverscope(), except for locking */
964 if (IN6_IS_SCOPE_EMBED(&sin6->sin6_addr)) {
965 sin6->sin6_scope_id = ifscope;
966 if (in6_embedded_scope) {
967 in6_verify_ifscope(&sin6->sin6_addr, ifscope);
968 sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
969 }
970 } else {
971 sin6->sin6_scope_id = 0; /* XXX */
972 }
973 if (in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
974 sin6->sin6_addr.s6_addr16[1] = 0;
975 }
976
977 return SA(sin6);
978 }
979
980 void
in6_sockaddr_s(in_port_t port,struct in6_addr * addr_p,struct sockaddr_in6 * sin6,uint32_t ifscope)981 in6_sockaddr_s(in_port_t port, struct in6_addr *addr_p,
982 struct sockaddr_in6 *sin6, uint32_t ifscope)
983 {
984 SOCKADDR_ZERO(sin6, sizeof(*sin6));
985 sin6->sin6_family = AF_INET6;
986 sin6->sin6_len = sizeof(*sin6);
987 sin6->sin6_port = port;
988 sin6->sin6_addr = *addr_p;
989
990 /* would be good to use sa6_recoverscope(), except for locking */
991 if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
992 sin6->sin6_scope_id = ifscope;
993 if (in6_embedded_scope) {
994 in6_verify_ifscope(&sin6->sin6_addr, ifscope);
995 sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
996 }
997 } else {
998 sin6->sin6_scope_id = 0; /* XXX */
999 }
1000 if (in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
1001 sin6->sin6_addr.s6_addr16[1] = 0;
1002 }
1003 }
1004
1005 /*
1006 * The calling convention of in6_getsockaddr() and in6_getpeeraddr() was
1007 * modified to match the pru_sockaddr() and pru_peeraddr() entry points
1008 * in struct pr_usrreqs, so that protocols can just reference then directly
1009 * without the need for a wrapper function.
1010 */
1011 int
in6_getsockaddr(struct socket * so,struct sockaddr ** nam)1012 in6_getsockaddr(struct socket *so, struct sockaddr **nam)
1013 {
1014 struct inpcb *__single inp;
1015 struct in6_addr addr;
1016 in_port_t port;
1017
1018 if ((inp = sotoinpcb(so)) == NULL) {
1019 return EINVAL;
1020 }
1021
1022 port = inp->inp_lport;
1023 addr = inp->in6p_laddr;
1024
1025 *nam = in6_sockaddr(port, &addr, inp->inp_lifscope);
1026 if (*nam == NULL) {
1027 return ENOBUFS;
1028 }
1029 return 0;
1030 }
1031
1032 int
in6_getsockaddr_s(struct socket * so,struct sockaddr_in6 * ss)1033 in6_getsockaddr_s(struct socket *so, struct sockaddr_in6 *ss)
1034 {
1035 struct inpcb *__single inp;
1036 struct in6_addr addr;
1037 in_port_t port;
1038
1039 VERIFY(ss != NULL);
1040 SOCKADDR_ZERO(ss, sizeof(*ss));
1041
1042 if ((inp = sotoinpcb(so)) == NULL) {
1043 return EINVAL;
1044 }
1045
1046 port = inp->inp_lport;
1047 addr = inp->in6p_laddr;
1048
1049 in6_sockaddr_s(port, &addr, ss, inp->inp_lifscope);
1050 return 0;
1051 }
1052
1053 int
in6_getpeeraddr(struct socket * so,struct sockaddr ** nam)1054 in6_getpeeraddr(struct socket *so, struct sockaddr **nam)
1055 {
1056 struct inpcb *__single inp;
1057 struct in6_addr addr;
1058 in_port_t port;
1059
1060 if ((inp = sotoinpcb(so)) == NULL) {
1061 return EINVAL;
1062 }
1063
1064 port = inp->inp_fport;
1065 addr = inp->in6p_faddr;
1066
1067 *nam = in6_sockaddr(port, &addr, inp->inp_fifscope);
1068 if (*nam == NULL) {
1069 return ENOBUFS;
1070 }
1071 return 0;
1072 }
1073
1074 int
in6_mapped_sockaddr(struct socket * so,struct sockaddr ** nam)1075 in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam)
1076 {
1077 struct inpcb *__single inp = sotoinpcb(so);
1078 int error;
1079
1080 if (inp == NULL) {
1081 return EINVAL;
1082 }
1083 if (inp->inp_vflag & INP_IPV4) {
1084 error = in_getsockaddr(so, nam);
1085 if (error == 0) {
1086 error = in6_sin_2_v4mapsin6_in_sock(nam);
1087 }
1088 } else {
1089 /* scope issues will be handled in in6_getsockaddr(). */
1090 error = in6_getsockaddr(so, nam);
1091 }
1092 return error;
1093 }
1094
1095 int
in6_mapped_peeraddr(struct socket * so,struct sockaddr ** nam)1096 in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam)
1097 {
1098 struct inpcb *inp = sotoinpcb(so);
1099 int error;
1100
1101 if (inp == NULL) {
1102 return EINVAL;
1103 }
1104 if (inp->inp_vflag & INP_IPV4) {
1105 error = in_getpeeraddr(so, nam);
1106 if (error == 0) {
1107 error = in6_sin_2_v4mapsin6_in_sock(nam);
1108 }
1109 } else {
1110 /* scope issues will be handled in in6_getpeeraddr(). */
1111 error = in6_getpeeraddr(so, nam);
1112 }
1113 return error;
1114 }
1115
1116 /*
1117 * Pass some notification to all connections of a protocol
1118 * associated with address dst. The local address and/or port numbers
1119 * may be specified to limit the search. The "usual action" will be
1120 * taken, depending on the ctlinput cmd. The caller must filter any
1121 * cmds that are uninteresting (e.g., no error in the map).
1122 * Call the protocol specific routine (if any) to report
1123 * any errors for each matching socket.
1124 */
1125 void
in6_pcbnotify(struct inpcbinfo * pcbinfo,struct sockaddr * dst,u_int fport_arg,const struct sockaddr * src,u_int lport_arg,int cmd,void * cmdarg,void (* notify)(struct inpcb *,int))1126 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst, u_int fport_arg,
1127 const struct sockaddr *src, u_int lport_arg, int cmd, void *cmdarg,
1128 void (*notify)(struct inpcb *, int))
1129 {
1130 struct inpcbhead *__single head = pcbinfo->ipi_listhead;
1131 struct inpcb *__single inp, *__single ninp;
1132 struct sockaddr_in6 sa6_src;
1133 struct sockaddr_in6 *__single sa6_dst;
1134 uint16_t fport = (uint16_t)fport_arg, lport = (uint16_t)lport_arg;
1135 u_int32_t flowinfo;
1136 int errno;
1137
1138 if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6) {
1139 return;
1140 }
1141
1142 sa6_dst = SIN6(dst);
1143 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr)) {
1144 return;
1145 }
1146
1147 /*
1148 * note that src can be NULL when we get notify by local fragmentation.
1149 */
1150 sa6_src = (src == NULL) ?
1151 sa6_any : *SIN6(src);
1152 flowinfo = sa6_src.sin6_flowinfo;
1153
1154 /*
1155 * Redirects go to all references to the destination,
1156 * and use in6_rtchange to invalidate the route cache.
1157 * Dead host indications: also use in6_rtchange to invalidate
1158 * the cache, and deliver the error to all the sockets.
1159 * Otherwise, if we have knowledge of the local port and address,
1160 * deliver only to that socket.
1161 */
1162 if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
1163 fport = 0;
1164 lport = 0;
1165 bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
1166
1167 if (cmd != PRC_HOSTDEAD) {
1168 notify = in6_rtchange;
1169 }
1170 }
1171 errno = inet6ctlerrmap[cmd];
1172 lck_rw_lock_shared(&pcbinfo->ipi_lock);
1173 for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
1174 ninp = LIST_NEXT(inp, inp_list);
1175
1176 if (!(inp->inp_vflag & INP_IPV6)) {
1177 continue;
1178 }
1179
1180 /*
1181 * If the error designates a new path MTU for a destination
1182 * and the application (associated with this socket) wanted to
1183 * know the value, notify. Note that we notify for all
1184 * disconnected sockets if the corresponding application
1185 * wanted. This is because some UDP applications keep sending
1186 * sockets disconnected.
1187 * XXX: should we avoid to notify the value to TCP sockets?
1188 */
1189 if (cmd == PRC_MSGSIZE && cmdarg != NULL) {
1190 socket_lock(inp->inp_socket, 1);
1191 ip6_notify_pmtu(inp, SIN6(dst),
1192 (u_int32_t *)cmdarg);
1193 socket_unlock(inp->inp_socket, 1);
1194 }
1195
1196 /*
1197 * Detect if we should notify the error. If no source and
1198 * destination ports are specifed, but non-zero flowinfo and
1199 * local address match, notify the error. This is the case
1200 * when the error is delivered with an encrypted buffer
1201 * by ESP. Otherwise, just compare addresses and ports
1202 * as usual.
1203 */
1204 if (lport == 0 && fport == 0 && flowinfo &&
1205 inp->inp_socket != NULL &&
1206 flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) &&
1207 in6_are_addr_equal_scoped(&inp->in6p_laddr, &sa6_src.sin6_addr, inp->inp_lifscope, sa6_src.sin6_scope_id)) {
1208 goto do_notify;
1209 } else if (!in6_are_addr_equal_scoped(&inp->in6p_faddr, &sa6_dst->sin6_addr,
1210 inp->inp_fifscope, sa6_dst->sin6_scope_id) || inp->inp_socket == NULL ||
1211 (lport && inp->inp_lport != lport) ||
1212 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
1213 !in6_are_addr_equal_scoped(&inp->in6p_laddr, &sa6_src.sin6_addr, inp->inp_lifscope, sa6_src.sin6_scope_id)) || (fport && inp->inp_fport != fport)) {
1214 continue;
1215 }
1216
1217 do_notify:
1218 if (notify) {
1219 if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) ==
1220 WNT_STOPUSING) {
1221 continue;
1222 }
1223 socket_lock(inp->inp_socket, 1);
1224 (*notify)(inp, errno);
1225 (void) in_pcb_checkstate(inp, WNT_RELEASE, 1);
1226 socket_unlock(inp->inp_socket, 1);
1227 }
1228 }
1229 lck_rw_done(&pcbinfo->ipi_lock);
1230 }
1231
1232 /*
1233 * Lookup a PCB based on the local address and port.
1234 */
1235 struct inpcb *
in6_pcblookup_local(struct inpcbinfo * pcbinfo,struct in6_addr * laddr,u_int lport_arg,uint32_t ifscope,int wild_okay)1236 in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr,
1237 u_int lport_arg, uint32_t ifscope, int wild_okay)
1238 {
1239 struct inpcb *__single inp;
1240 int matchwild = 3, wildcard;
1241 uint16_t lport = (uint16_t)lport_arg;
1242 struct inpcbporthead *__single porthash;
1243 struct inpcb *__single match = NULL;
1244 struct inpcbport *__single phd;
1245
1246 if (!wild_okay) {
1247 struct inpcbhead *__single head;
1248 /*
1249 * Look for an unconnected (wildcard foreign addr) PCB that
1250 * matches the local address and port we're looking for.
1251 */
1252 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0,
1253 pcbinfo->ipi_hashmask)];
1254 LIST_FOREACH(inp, head, inp_hash) {
1255 if (!(inp->inp_vflag & INP_IPV6)) {
1256 continue;
1257 }
1258 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1259 in6_are_addr_equal_scoped(&inp->in6p_laddr, laddr, inp->inp_lifscope, ifscope) &&
1260 inp->inp_lport == lport) {
1261 /*
1262 * Found.
1263 */
1264 return inp;
1265 }
1266 }
1267 /*
1268 * Not found.
1269 */
1270 return NULL;
1271 }
1272 /*
1273 * Best fit PCB lookup.
1274 *
1275 * First see if this local port is in use by looking on the
1276 * port hash list.
1277 */
1278 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1279 pcbinfo->ipi_porthashmask)];
1280 LIST_FOREACH(phd, porthash, phd_hash) {
1281 if (phd->phd_port == lport) {
1282 break;
1283 }
1284 }
1285
1286 if (phd != NULL) {
1287 /*
1288 * Port is in use by one or more PCBs. Look for best
1289 * fit.
1290 */
1291 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1292 wildcard = 0;
1293 if (!(inp->inp_vflag & INP_IPV6)) {
1294 continue;
1295 }
1296 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1297 wildcard++;
1298 }
1299 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1300 if (IN6_IS_ADDR_UNSPECIFIED(laddr)) {
1301 wildcard++;
1302 } else if (!in6_are_addr_equal_scoped(
1303 &inp->in6p_laddr, laddr, inp->inp_lifscope, ifscope)) {
1304 continue;
1305 }
1306 } else {
1307 if (!IN6_IS_ADDR_UNSPECIFIED(laddr)) {
1308 wildcard++;
1309 }
1310 }
1311 if (wildcard < matchwild) {
1312 match = inp;
1313 matchwild = wildcard;
1314 if (matchwild == 0) {
1315 break;
1316 }
1317 }
1318 }
1319 }
1320 return match;
1321 }
1322
1323 /*
1324 * Check for alternatives when higher level complains
1325 * about service problems. For now, invalidate cached
1326 * routing information. If the route was created dynamically
1327 * (by a redirect), time to try a default gateway again.
1328 */
1329 void
in6_losing(struct inpcb * in6p)1330 in6_losing(struct inpcb *in6p)
1331 {
1332 rtentry_ref_t rt;
1333
1334 if ((rt = in6p->in6p_route.ro_rt) != NULL) {
1335 RT_LOCK(rt);
1336 if (rt->rt_flags & RTF_DYNAMIC) {
1337 /*
1338 * Prevent another thread from modifying rt_key,
1339 * rt_gateway via rt_setgate() after the rt_lock
1340 * is dropped by marking the route as defunct.
1341 */
1342 rt->rt_flags |= RTF_CONDEMNED;
1343 RT_UNLOCK(rt);
1344 (void) rtrequest(RTM_DELETE, rt_key(rt),
1345 rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
1346 } else {
1347 RT_UNLOCK(rt);
1348 }
1349 /*
1350 * A new route can be allocated
1351 * the next time output is attempted.
1352 */
1353 }
1354 ROUTE_RELEASE(&in6p->in6p_route);
1355 }
1356
1357 /*
1358 * After a routing change, flush old routing
1359 * and allocate a (hopefully) better one.
1360 */
1361 void
in6_rtchange(struct inpcb * inp,int errno)1362 in6_rtchange(struct inpcb *inp, int errno)
1363 {
1364 #pragma unused(errno)
1365 /*
1366 * A new route can be allocated the next time
1367 * output is attempted.
1368 */
1369 ROUTE_RELEASE(&inp->in6p_route);
1370 }
1371
1372 /*
1373 * Check if PCB exists hash list. Also returns uid and gid of socket
1374 */
1375 int
in6_pcblookup_hash_exists(struct inpcbinfo * pcbinfo,struct in6_addr * faddr,u_int fport_arg,uint32_t fifscope,struct in6_addr * laddr,u_int lport_arg,uint32_t lifscope,int wildcard,uid_t * uid,gid_t * gid,struct ifnet * ifp,bool relaxed)1376 in6_pcblookup_hash_exists(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1377 u_int fport_arg, uint32_t fifscope, struct in6_addr *laddr, u_int lport_arg, uint32_t lifscope, int wildcard,
1378 uid_t *uid, gid_t *gid, struct ifnet *ifp, bool relaxed)
1379 {
1380 struct inpcbhead *__single head;
1381 struct inpcb *__single inp;
1382 uint16_t fport = (uint16_t)fport_arg, lport = (uint16_t)lport_arg;
1383 int found;
1384
1385 *uid = UID_MAX;
1386 *gid = GID_MAX;
1387
1388 lck_rw_lock_shared(&pcbinfo->ipi_lock);
1389
1390 /*
1391 * First look for an exact match.
1392 */
1393 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr->s6_addr32[3] /* XXX */,
1394 lport, fport, pcbinfo->ipi_hashmask)];
1395 LIST_FOREACH(inp, head, inp_hash) {
1396 if (!(inp->inp_vflag & INP_IPV6)) {
1397 continue;
1398 }
1399
1400 if (inp_restricted_recv(inp, ifp)) {
1401 continue;
1402 }
1403
1404 #if NECP
1405 if (!necp_socket_is_allowed_to_recv_on_interface(inp, ifp)) {
1406 continue;
1407 }
1408 #endif /* NECP */
1409
1410 if (((in6_are_addr_equal_scoped(&inp->in6p_faddr, faddr, inp->inp_fifscope, fifscope) &&
1411 in6_are_addr_equal_scoped(&inp->in6p_laddr, laddr, inp->inp_lifscope, lifscope)) ||
1412 (relaxed &&
1413 IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
1414 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr))) &&
1415 inp->inp_fport == fport &&
1416 inp->inp_lport == lport) {
1417 if ((found = (inp->inp_socket != NULL))) {
1418 /*
1419 * Found. Check if pcb is still valid
1420 */
1421 *uid = kauth_cred_getuid(
1422 inp->inp_socket->so_cred);
1423 *gid = kauth_cred_getgid(
1424 inp->inp_socket->so_cred);
1425 }
1426 lck_rw_done(&pcbinfo->ipi_lock);
1427 return found;
1428 }
1429 }
1430 if (wildcard) {
1431 struct inpcb *__single local_wild = NULL;
1432
1433 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0,
1434 pcbinfo->ipi_hashmask)];
1435 LIST_FOREACH(inp, head, inp_hash) {
1436 if (!(inp->inp_vflag & INP_IPV6)) {
1437 continue;
1438 }
1439
1440 if (inp_restricted_recv(inp, ifp)) {
1441 continue;
1442 }
1443
1444 #if NECP
1445 if (!necp_socket_is_allowed_to_recv_on_interface(inp, ifp)) {
1446 continue;
1447 }
1448 #endif /* NECP */
1449
1450 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1451 inp->inp_lport == lport) {
1452 if (in6_are_addr_equal_scoped(&inp->in6p_laddr,
1453 laddr, inp->inp_lifscope, lifscope) ||
1454 (relaxed && IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr))) {
1455 found = (inp->inp_socket != NULL);
1456 if (found) {
1457 *uid = kauth_cred_getuid(
1458 inp->inp_socket->so_cred);
1459 *gid = kauth_cred_getgid(
1460 inp->inp_socket->so_cred);
1461 }
1462 lck_rw_done(&pcbinfo->ipi_lock);
1463 return found;
1464 } else if (IN6_IS_ADDR_UNSPECIFIED(
1465 &inp->in6p_laddr)) {
1466 local_wild = inp;
1467 }
1468 }
1469 }
1470 if (local_wild) {
1471 if ((found = (local_wild->inp_socket != NULL))) {
1472 *uid = kauth_cred_getuid(
1473 local_wild->inp_socket->so_cred);
1474 *gid = kauth_cred_getgid(
1475 local_wild->inp_socket->so_cred);
1476 }
1477 lck_rw_done(&pcbinfo->ipi_lock);
1478 return found;
1479 }
1480 }
1481
1482 /*
1483 * Not found.
1484 */
1485 lck_rw_done(&pcbinfo->ipi_lock);
1486 return 0;
1487 }
1488
1489 /*
1490 * Lookup PCB in hash list.
1491 */
1492 struct inpcb *
in6_pcblookup_hash(struct inpcbinfo * pcbinfo,struct in6_addr * faddr,u_int fport_arg,uint32_t fifscope,struct in6_addr * laddr,u_int lport_arg,uint32_t lifscope,int wildcard,struct ifnet * ifp)1493 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1494 u_int fport_arg, uint32_t fifscope, struct in6_addr *laddr, u_int lport_arg, uint32_t lifscope, int wildcard,
1495 struct ifnet *ifp)
1496 {
1497 struct inpcbhead *__single head;
1498 struct inpcb *__single inp;
1499 uint16_t fport = (uint16_t)fport_arg, lport = (uint16_t)lport_arg;
1500
1501 lck_rw_lock_shared(&pcbinfo->ipi_lock);
1502
1503 /*
1504 * First look for an exact match.
1505 */
1506 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr->s6_addr32[3] /* XXX */,
1507 lport, fport, pcbinfo->ipi_hashmask)];
1508 LIST_FOREACH(inp, head, inp_hash) {
1509 if (!(inp->inp_vflag & INP_IPV6)) {
1510 continue;
1511 }
1512
1513 if (inp_restricted_recv(inp, ifp)) {
1514 continue;
1515 }
1516
1517 #if NECP
1518 if (!necp_socket_is_allowed_to_recv_on_interface(inp, ifp)) {
1519 continue;
1520 }
1521 #endif /* NECP */
1522
1523 if (in6_are_addr_equal_scoped(&inp->in6p_faddr, faddr, inp->inp_fifscope, fifscope) &&
1524 in6_are_addr_equal_scoped(&inp->in6p_laddr, laddr, inp->inp_lifscope, lifscope) &&
1525 inp->inp_fport == fport &&
1526 inp->inp_lport == lport) {
1527 /*
1528 * Found. Check if pcb is still valid
1529 */
1530 if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) !=
1531 WNT_STOPUSING) {
1532 lck_rw_done(&pcbinfo->ipi_lock);
1533 return inp;
1534 } else {
1535 /* it's there but dead, say it isn't found */
1536 lck_rw_done(&pcbinfo->ipi_lock);
1537 return NULL;
1538 }
1539 }
1540 }
1541 if (wildcard) {
1542 struct inpcb *__single local_wild = NULL;
1543
1544 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0,
1545 pcbinfo->ipi_hashmask)];
1546 LIST_FOREACH(inp, head, inp_hash) {
1547 if (!(inp->inp_vflag & INP_IPV6)) {
1548 continue;
1549 }
1550
1551 if (inp_restricted_recv(inp, ifp)) {
1552 continue;
1553 }
1554
1555 #if NECP
1556 if (!necp_socket_is_allowed_to_recv_on_interface(inp, ifp)) {
1557 continue;
1558 }
1559 #endif /* NECP */
1560
1561 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1562 inp->inp_lport == lport) {
1563 if (in6_are_addr_equal_scoped(&inp->in6p_laddr,
1564 laddr, inp->inp_lifscope, lifscope)) {
1565 if (in_pcb_checkstate(inp, WNT_ACQUIRE,
1566 0) != WNT_STOPUSING) {
1567 lck_rw_done(&pcbinfo->ipi_lock);
1568 return inp;
1569 } else {
1570 /* dead; say it isn't found */
1571 lck_rw_done(&pcbinfo->ipi_lock);
1572 return NULL;
1573 }
1574 } else if (IN6_IS_ADDR_UNSPECIFIED(
1575 &inp->in6p_laddr)) {
1576 local_wild = inp;
1577 }
1578 }
1579 }
1580 if (local_wild && in_pcb_checkstate(local_wild,
1581 WNT_ACQUIRE, 0) != WNT_STOPUSING) {
1582 lck_rw_done(&pcbinfo->ipi_lock);
1583 return local_wild;
1584 } else {
1585 lck_rw_done(&pcbinfo->ipi_lock);
1586 return NULL;
1587 }
1588 }
1589
1590 /*
1591 * Not found.
1592 */
1593 lck_rw_done(&pcbinfo->ipi_lock);
1594 return NULL;
1595 }
1596
1597 void
init_sin6(struct sockaddr_in6 * sin6,struct mbuf * m)1598 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m)
1599 {
1600 struct ip6_hdr *__single ip;
1601
1602 ip = mtod(m, struct ip6_hdr *);
1603 SOCKADDR_ZERO(sin6, sizeof(*sin6));
1604 sin6->sin6_len = sizeof(*sin6);
1605 sin6->sin6_family = AF_INET6;
1606 sin6->sin6_addr = ip->ip6_src;
1607 if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
1608 if (in6_embedded_scope) {
1609 sin6->sin6_addr.s6_addr16[1] = 0;
1610 }
1611 if ((m->m_pkthdr.pkt_flags & (PKTF_LOOP | PKTF_IFAINFO)) ==
1612 (PKTF_LOOP | PKTF_IFAINFO)) {
1613 sin6->sin6_scope_id = m->m_pkthdr.src_ifindex;
1614 } else if (m->m_pkthdr.rcvif != NULL) {
1615 sin6->sin6_scope_id = m->m_pkthdr.rcvif->if_index;
1616 }
1617 }
1618 }
1619
1620 /*
1621 * The following routines implement this scheme:
1622 *
1623 * Callers of ip6_output() that intend to cache the route in the inpcb pass
1624 * a local copy of the struct route to ip6_output(). Using a local copy of
1625 * the cached route significantly simplifies things as IP no longer has to
1626 * worry about having exclusive access to the passed in struct route, since
1627 * it's defined in the caller's stack; in essence, this allows for a lock-
1628 * less operation when updating the struct route at the IP level and below,
1629 * whenever necessary. The scheme works as follows:
1630 *
1631 * Prior to dropping the socket's lock and calling ip6_output(), the caller
1632 * copies the struct route from the inpcb into its stack, and adds a reference
1633 * to the cached route entry, if there was any. The socket's lock is then
1634 * dropped and ip6_output() is called with a pointer to the copy of struct
1635 * route defined on the stack (not to the one in the inpcb.)
1636 *
1637 * Upon returning from ip6_output(), the caller then acquires the socket's
1638 * lock and synchronizes the cache; if there is no route cached in the inpcb,
1639 * it copies the local copy of struct route (which may or may not contain any
1640 * route) back into the cache; otherwise, if the inpcb has a route cached in
1641 * it, the one in the local copy will be freed, if there's any. Trashing the
1642 * cached route in the inpcb can be avoided because ip6_output() is single-
1643 * threaded per-PCB (i.e. multiple transmits on a PCB are always serialized
1644 * by the socket/transport layer.)
1645 */
1646 void
in6p_route_copyout(struct inpcb * inp,struct route_in6 * dst)1647 in6p_route_copyout(struct inpcb *inp, struct route_in6 *dst)
1648 {
1649 struct route_in6 *__single src = &inp->in6p_route;
1650
1651 socket_lock_assert_owned(inp->inp_socket);
1652
1653 /* Minor sanity check */
1654 if (src->ro_rt != NULL && rt_key(src->ro_rt)->sa_family != AF_INET6) {
1655 panic("%s: wrong or corrupted route: %p", __func__, src);
1656 }
1657
1658 route_copyout((struct route *)dst, (struct route *)src, sizeof(*dst));
1659 }
1660
1661 void
in6p_route_copyin(struct inpcb * inp,struct route_in6 * src)1662 in6p_route_copyin(struct inpcb *inp, struct route_in6 *src)
1663 {
1664 struct route_in6 *__single dst = &inp->in6p_route;
1665
1666 socket_lock_assert_owned(inp->inp_socket);
1667
1668 /* Minor sanity check */
1669 if (src->ro_rt != NULL && rt_key(src->ro_rt)->sa_family != AF_INET6) {
1670 panic("%s: wrong or corrupted route: %p", __func__, src);
1671 }
1672
1673 route_copyin((struct route *)src, (struct route *)dst, sizeof(*src));
1674 }
1675