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 /*
144 * in6_pcblookup_local_and_cleanup does everything
145 * in6_pcblookup_local does but it checks for a socket
146 * that's going away. Since we know that the lock is
147 * held read+write when this function is called, we
148 * can safely dispose of this socket like the slow
149 * timer would usually do and return NULL. This is
150 * great for bind.
151 */
152 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)153 in6_pcblookup_local_and_cleanup(struct inpcbinfo *pcbinfo,
154 struct in6_addr *laddr, u_int lport_arg, uint32_t ifscope, int wild_okay)
155 {
156 struct inpcb *inp;
157
158 /* Perform normal lookup */
159 inp = in6_pcblookup_local(pcbinfo, laddr, lport_arg, ifscope, wild_okay);
160
161 /* Check if we found a match but it's waiting to be disposed */
162 if (inp != NULL && inp->inp_wantcnt == WNT_STOPUSING) {
163 struct socket *so = inp->inp_socket;
164
165 socket_lock(so, 0);
166
167 if (so->so_usecount == 0) {
168 if (inp->inp_state != INPCB_STATE_DEAD) {
169 in6_pcbdetach(inp);
170 }
171 in_pcbdispose(inp); /* will unlock & destroy */
172 inp = NULL;
173 } else {
174 socket_unlock(so, 0);
175 }
176 }
177
178 return inp;
179 }
180
181 /*
182 * Bind an INPCB to an address and/or port. This routine should not alter
183 * the caller-supplied local address "nam".
184 */
185 int
in6_pcbbind(struct inpcb * inp,struct sockaddr * nam,struct proc * p)186 in6_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct proc *p)
187 {
188 struct socket *so = inp->inp_socket;
189 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
190 u_short lport = 0;
191 int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
192 struct ifnet *outif = NULL;
193 struct sockaddr_in6 sin6;
194 uint32_t lifscope = IFSCOPE_NONE;
195 #if XNU_TARGET_OS_OSX
196 int error;
197 kauth_cred_t cred;
198 #endif /* XNU_TARGET_OS_OSX */
199
200 if (TAILQ_EMPTY(&in6_ifaddrhead)) { /* XXX broken! */
201 return EADDRNOTAVAIL;
202 }
203 if (!(so->so_options & (SO_REUSEADDR | SO_REUSEPORT))) {
204 wild = 1;
205 }
206
207 in_pcb_check_management_entitled(inp);
208
209 socket_unlock(so, 0); /* keep reference */
210 lck_rw_lock_exclusive(&pcbinfo->ipi_lock);
211 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
212 /* another thread completed the bind */
213 lck_rw_done(&pcbinfo->ipi_lock);
214 socket_lock(so, 0);
215 return EINVAL;
216 }
217
218 bzero(&sin6, sizeof(sin6));
219 if (nam != NULL) {
220 if (nam->sa_len != sizeof(struct sockaddr_in6)) {
221 lck_rw_done(&pcbinfo->ipi_lock);
222 socket_lock(so, 0);
223 return EINVAL;
224 }
225 /*
226 * family check.
227 */
228 if (nam->sa_family != AF_INET6) {
229 lck_rw_done(&pcbinfo->ipi_lock);
230 socket_lock(so, 0);
231 return EAFNOSUPPORT;
232 }
233 lport = SIN6(nam)->sin6_port;
234
235 *(&sin6) = *SIN6(nam);
236
237 /* KAME hack: embed scopeid */
238 if (in6_embedscope(&sin6.sin6_addr, &sin6, inp, NULL,
239 NULL, &lifscope) != 0) {
240 lck_rw_done(&pcbinfo->ipi_lock);
241 socket_lock(so, 0);
242 return EINVAL;
243 }
244
245 /* Sanitize local copy for address searches */
246 sin6.sin6_flowinfo = 0;
247 sin6.sin6_port = 0;
248 if (in6_embedded_scope) {
249 sin6.sin6_scope_id = 0;
250 }
251
252 if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
253 /*
254 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
255 * allow compepte duplication of binding if
256 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
257 * and a multicast address is bound on both
258 * new and duplicated sockets.
259 */
260 if (so->so_options & SO_REUSEADDR) {
261 reuseport = SO_REUSEADDR | SO_REUSEPORT;
262 }
263 } else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr)) {
264 struct ifaddr *ifa;
265
266 ifa = ifa_ifwithaddr(SA(&sin6));
267 if (ifa == NULL) {
268 lck_rw_done(&pcbinfo->ipi_lock);
269 socket_lock(so, 0);
270 return EADDRNOTAVAIL;
271 } else {
272 /*
273 * XXX: bind to an anycast address might
274 * accidentally cause sending a packet with
275 * anycast source address. We should allow
276 * to bind to a deprecated address, since
277 * the application dare to use it.
278 */
279 IFA_LOCK_SPIN(ifa);
280 if (((struct in6_ifaddr *)ifa)->ia6_flags &
281 (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY |
282 IN6_IFF_DETACHED | IN6_IFF_CLAT46)) {
283 IFA_UNLOCK(ifa);
284 IFA_REMREF(ifa);
285 lck_rw_done(&pcbinfo->ipi_lock);
286 socket_lock(so, 0);
287 return EADDRNOTAVAIL;
288 }
289 /*
290 * Opportunistically determine the outbound
291 * interface that may be used; this may not
292 * hold true if we end up using a route
293 * going over a different interface, e.g.
294 * when sending to a local address. This
295 * will get updated again after sending.
296 */
297 outif = ifa->ifa_ifp;
298 IFA_UNLOCK(ifa);
299 IFA_REMREF(ifa);
300 }
301 }
302
303 #if SKYWALK
304 if (inp->inp_flags2 & INP2_EXTERNAL_PORT) {
305 // Extract the external flow info
306 struct ns_flow_info nfi = {};
307 int netns_error = necp_client_get_netns_flow_info(inp->necp_client_uuid,
308 &nfi);
309 if (netns_error != 0) {
310 lck_rw_done(&pcbinfo->ipi_lock);
311 socket_lock(so, 0);
312 return netns_error;
313 }
314
315 // Extract the reserved port
316 u_int16_t reserved_lport = 0;
317 if (nfi.nfi_laddr.sa.sa_family == AF_INET) {
318 reserved_lport = nfi.nfi_laddr.sin.sin_port;
319 } else if (nfi.nfi_laddr.sa.sa_family == AF_INET6) {
320 reserved_lport = nfi.nfi_laddr.sin6.sin6_port;
321 } else {
322 lck_rw_done(&pcbinfo->ipi_lock);
323 socket_lock(so, 0);
324 return EINVAL;
325 }
326
327 // Validate or use the reserved port
328 if (lport == 0) {
329 lport = reserved_lport;
330 } else if (lport != reserved_lport) {
331 lck_rw_done(&pcbinfo->ipi_lock);
332 socket_lock(so, 0);
333 return EINVAL;
334 }
335 }
336
337 /* Do not allow reserving a UDP port if remaining UDP port count is below 4096 */
338 if (SOCK_PROTO(so) == IPPROTO_UDP && !allow_udp_port_exhaustion) {
339 uint32_t current_reservations = 0;
340 current_reservations = netns_lookup_reservations_count_in6(inp->in6p_laddr, IPPROTO_UDP);
341 if (USHRT_MAX - UDP_RANDOM_PORT_RESERVE < current_reservations) {
342 log(LOG_ERR, "UDP port not available, less than 4096 UDP ports left");
343 lck_rw_done(&pcbinfo->ipi_lock);
344 socket_lock(so, 0);
345 return EADDRNOTAVAIL;
346 }
347 }
348
349 #endif /* SKYWALK */
350
351 if (lport != 0) {
352 struct inpcb *t;
353 uid_t u;
354
355 #if XNU_TARGET_OS_OSX
356 if (ntohs(lport) < IPV6PORT_RESERVED &&
357 !IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr) &&
358 !(inp->inp_flags2 & INP2_EXTERNAL_PORT)) {
359 cred = kauth_cred_proc_ref(p);
360 error = priv_check_cred(cred,
361 PRIV_NETINET_RESERVEDPORT, 0);
362 kauth_cred_unref(&cred);
363 if (error != 0) {
364 lck_rw_done(&pcbinfo->ipi_lock);
365 socket_lock(so, 0);
366 return EACCES;
367 }
368 }
369 #endif /* XNU_TARGET_OS_OSX */
370 /*
371 * Check wether the process is allowed to bind to a restricted port
372 */
373 if (!current_task_can_use_restricted_in_port(lport,
374 (uint8_t)SOCK_PROTO(so), PORT_FLAGS_BSD)) {
375 lck_rw_done(&pcbinfo->ipi_lock);
376 socket_lock(so, 0);
377 return EADDRINUSE;
378 }
379
380 if (!IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr) &&
381 (u = kauth_cred_getuid(so->so_cred)) != 0) {
382 t = in6_pcblookup_local_and_cleanup(pcbinfo,
383 &sin6.sin6_addr, lport, sin6.sin6_scope_id,
384 INPLOOKUP_WILDCARD);
385 if (t != NULL &&
386 (!IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr) ||
387 !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
388 !(t->inp_socket->so_options & SO_REUSEPORT)) &&
389 (u != kauth_cred_getuid(t->inp_socket->so_cred)) &&
390 !(t->inp_socket->so_flags & SOF_REUSESHAREUID) &&
391 (!(t->inp_flags2 & INP2_EXTERNAL_PORT) ||
392 !(inp->inp_flags2 & INP2_EXTERNAL_PORT) ||
393 uuid_compare(t->necp_client_uuid, inp->necp_client_uuid) != 0)) {
394 lck_rw_done(&pcbinfo->ipi_lock);
395 socket_lock(so, 0);
396 return EADDRINUSE;
397 }
398 if (!(inp->inp_flags & IN6P_IPV6_V6ONLY) &&
399 IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr)) {
400 struct sockaddr_in sin;
401
402 in6_sin6_2_sin(&sin, &sin6);
403 t = in_pcblookup_local_and_cleanup(
404 pcbinfo, sin.sin_addr, lport,
405 INPLOOKUP_WILDCARD);
406 if (t != NULL &&
407 !(t->inp_socket->so_options & SO_REUSEPORT) &&
408 (kauth_cred_getuid(so->so_cred) !=
409 kauth_cred_getuid(t->inp_socket->so_cred)) &&
410 (t->inp_laddr.s_addr != INADDR_ANY ||
411 SOCK_DOM(so) == SOCK_DOM(t->inp_socket)) &&
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 return EADDRINUSE;
418 }
419
420 #if SKYWALK
421 VERIFY(!NETNS_TOKEN_VALID(
422 &inp->inp_wildcard_netns_token));
423 if ((SOCK_PROTO(so) == IPPROTO_TCP ||
424 SOCK_PROTO(so) == IPPROTO_UDP) &&
425 !(inp->inp_flags2 & INP2_EXTERNAL_PORT)) {
426 if (netns_reserve_in(&inp->
427 inp_wildcard_netns_token,
428 sin.sin_addr,
429 (uint8_t)SOCK_PROTO(so), lport,
430 NETNS_BSD, NULL) != 0) {
431 lck_rw_done(&pcbinfo->ipi_lock);
432 socket_lock(so, 0);
433 return EADDRINUSE;
434 }
435 }
436 #endif /* SKYWALK */
437 }
438 }
439 t = in6_pcblookup_local_and_cleanup(pcbinfo,
440 &sin6.sin6_addr, lport, sin6.sin6_scope_id, wild);
441 if (t != NULL &&
442 (reuseport & t->inp_socket->so_options) == 0 &&
443 (!(t->inp_flags2 & INP2_EXTERNAL_PORT) ||
444 !(inp->inp_flags2 & INP2_EXTERNAL_PORT) ||
445 uuid_compare(t->necp_client_uuid, inp->necp_client_uuid) != 0)) {
446 #if SKYWALK
447 netns_release(&inp->inp_wildcard_netns_token);
448 #endif /* SKYWALK */
449 lck_rw_done(&pcbinfo->ipi_lock);
450 socket_lock(so, 0);
451 return EADDRINUSE;
452 }
453 if (!(inp->inp_flags & IN6P_IPV6_V6ONLY) &&
454 IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr)) {
455 struct sockaddr_in sin;
456
457 in6_sin6_2_sin(&sin, &sin6);
458 t = in_pcblookup_local_and_cleanup(pcbinfo,
459 sin.sin_addr, lport, wild);
460 if (t != NULL && (reuseport &
461 t->inp_socket->so_options) == 0 &&
462 (t->inp_laddr.s_addr != INADDR_ANY ||
463 SOCK_DOM(so) == SOCK_DOM(t->inp_socket)) &&
464 (!(t->inp_flags2 & INP2_EXTERNAL_PORT) ||
465 !(inp->inp_flags2 & INP2_EXTERNAL_PORT) ||
466 uuid_compare(t->necp_client_uuid, inp->necp_client_uuid) != 0)) {
467 #if SKYWALK
468 netns_release(&inp->inp_wildcard_netns_token);
469 #endif /* SKYWALK */
470 lck_rw_done(&pcbinfo->ipi_lock);
471 socket_lock(so, 0);
472 return EADDRINUSE;
473 }
474 #if SKYWALK
475 if ((SOCK_PROTO(so) == IPPROTO_TCP ||
476 SOCK_PROTO(so) == IPPROTO_UDP) &&
477 !(inp->inp_flags2 & INP2_EXTERNAL_PORT) &&
478 (!NETNS_TOKEN_VALID(
479 &inp->inp_wildcard_netns_token))) {
480 if (netns_reserve_in(&inp->
481 inp_wildcard_netns_token,
482 sin.sin_addr,
483 (uint8_t)SOCK_PROTO(so), lport,
484 NETNS_BSD, NULL) != 0) {
485 lck_rw_done(&pcbinfo->ipi_lock);
486 socket_lock(so, 0);
487 return EADDRINUSE;
488 }
489 }
490 #endif /* SKYWALK */
491 }
492 #if SKYWALK
493 if ((SOCK_PROTO(so) == IPPROTO_TCP ||
494 SOCK_PROTO(so) == IPPROTO_UDP) &&
495 !(inp->inp_flags2 & INP2_EXTERNAL_PORT)) {
496 if (netns_reserve_in6(&inp->inp_netns_token,
497 sin6.sin6_addr, (uint8_t)SOCK_PROTO(so), lport,
498 NETNS_BSD, NULL) != 0) {
499 netns_release(&inp->inp_wildcard_netns_token);
500 lck_rw_done(&pcbinfo->ipi_lock);
501 socket_lock(so, 0);
502 return EADDRINUSE;
503 }
504 }
505 #endif /* SKYWALK */
506 }
507 }
508
509 socket_lock(so, 0);
510 /*
511 * We unlocked socket's protocol lock for a long time.
512 * The socket might have been dropped/defuncted.
513 * Checking if world has changed since.
514 */
515 if (inp->inp_state == INPCB_STATE_DEAD) {
516 #if SKYWALK
517 netns_release(&inp->inp_netns_token);
518 netns_release(&inp->inp_wildcard_netns_token);
519 #endif /* SKYWALK */
520 lck_rw_done(&pcbinfo->ipi_lock);
521 return ECONNABORTED;
522 }
523
524 /* check if the socket got bound when the lock was released */
525 if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
526 #if SKYWALK
527 netns_release(&inp->inp_netns_token);
528 netns_release(&inp->inp_wildcard_netns_token);
529 #endif /* SKYWALK */
530 lck_rw_done(&pcbinfo->ipi_lock);
531 return EINVAL;
532 }
533
534 if (!IN6_IS_ADDR_UNSPECIFIED(&sin6.sin6_addr)) {
535 inp->in6p_laddr = sin6.sin6_addr;
536 inp->in6p_last_outifp = outif;
537 inp->inp_lifscope = lifscope;
538 in6_verify_ifscope(&inp->in6p_laddr, lifscope);
539 #if SKYWALK
540 if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
541 netns_set_ifnet(&inp->inp_netns_token,
542 inp->in6p_last_outifp);
543 }
544 #endif /* SKYWALK */
545 }
546
547 if (lport == 0) {
548 int e;
549 if ((e = in6_pcbsetport(&inp->in6p_laddr, inp, p, 1)) != 0) {
550 /* Undo any address bind from above. */
551 #if SKYWALK
552 netns_release(&inp->inp_netns_token);
553 netns_release(&inp->inp_wildcard_netns_token);
554 #endif /* SKYWALK */
555 inp->in6p_laddr = in6addr_any;
556 inp->in6p_last_outifp = NULL;
557 inp->inp_lifscope = IFSCOPE_NONE;
558 lck_rw_done(&pcbinfo->ipi_lock);
559 return e;
560 }
561 } else {
562 inp->inp_lport = lport;
563 if (in_pcbinshash(inp, 1) != 0) {
564 #if SKYWALK
565 netns_release(&inp->inp_netns_token);
566 netns_release(&inp->inp_wildcard_netns_token);
567 #endif /* SKYWALK */
568 inp->in6p_laddr = in6addr_any;
569 inp->inp_lifscope = IFSCOPE_NONE;
570 inp->inp_lport = 0;
571 inp->in6p_last_outifp = NULL;
572 lck_rw_done(&pcbinfo->ipi_lock);
573 return EAGAIN;
574 }
575 }
576 lck_rw_done(&pcbinfo->ipi_lock);
577 sflt_notify(so, sock_evt_bound, NULL);
578 return 0;
579 }
580
581 /*
582 * Transform old in6_pcbconnect() into an inner subroutine for new
583 * in6_pcbconnect(); do some validity-checking on the remote address
584 * (in "nam") and then determine local host address (i.e., which
585 * interface) to use to access that remote host.
586 *
587 * This routine may alter the caller-supplied remote address "nam".
588 *
589 * This routine might return an ifp with a reference held if the caller
590 * provides a non-NULL outif, even in the error case. The caller is
591 * responsible for releasing its reference.
592 */
593 int
in6_pcbladdr(struct inpcb * inp,struct sockaddr * nam,struct in6_addr * plocal_addr6,struct ifnet ** outif)594 in6_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
595 struct in6_addr *plocal_addr6, struct ifnet **outif)
596 {
597 struct in6_addr *addr6 = NULL;
598 struct in6_addr src_storage;
599 int error = 0;
600 unsigned int ifscope;
601
602 if (outif != NULL) {
603 *outif = NULL;
604 }
605 if (nam->sa_len != sizeof(struct sockaddr_in6)) {
606 return EINVAL;
607 }
608 if (SIN6(nam)->sin6_family != AF_INET6) {
609 return EAFNOSUPPORT;
610 }
611 if (SIN6(nam)->sin6_port == 0) {
612 return EADDRNOTAVAIL;
613 }
614
615 /* KAME hack: embed scopeid */
616 if (in6_embedscope(&SIN6(nam)->sin6_addr, SIN6(nam), inp, NULL, NULL, IN6_NULL_IF_EMBEDDED_SCOPE(&SIN6(nam)->sin6_scope_id)) != 0) {
617 return EINVAL;
618 }
619
620 in_pcb_check_management_entitled(inp);
621
622 if (!TAILQ_EMPTY(&in6_ifaddrhead)) {
623 /*
624 * If the destination address is UNSPECIFIED addr,
625 * use the loopback addr, e.g ::1.
626 */
627 if (IN6_IS_ADDR_UNSPECIFIED(&SIN6(nam)->sin6_addr)) {
628 SIN6(nam)->sin6_addr = in6addr_loopback;
629 }
630 }
631
632 ifscope = (inp->inp_flags & INP_BOUND_IF) ?
633 inp->inp_boundifp->if_index : IFSCOPE_NONE;
634
635 /*
636 * XXX: in6_selectsrc might replace the bound local address
637 * with the address specified by setsockopt(IPV6_PKTINFO).
638 * Is it the intended behavior?
639 *
640 * in6_selectsrc() might return outif with its reference held
641 * even in the error case; caller always needs to release it
642 * if non-NULL.
643 */
644 addr6 = in6_selectsrc(SIN6(nam), inp->in6p_outputopts, inp,
645 &inp->in6p_route, outif, &src_storage, ifscope, &error);
646
647 if (outif != NULL) {
648 struct rtentry *rt = inp->in6p_route.ro_rt;
649 /*
650 * If in6_selectsrc() returns a route, it should be one
651 * which points to the same ifp as outif. Just in case
652 * it isn't, use the one from the route for consistency.
653 * Otherwise if there is no route, leave outif alone as
654 * it could still be useful to the caller.
655 */
656 if (rt != NULL && rt->rt_ifp != *outif) {
657 ifnet_reference(rt->rt_ifp); /* for caller */
658 if (*outif != NULL) {
659 ifnet_release(*outif);
660 }
661 *outif = rt->rt_ifp;
662 }
663 }
664
665 if (addr6 == NULL) {
666 if (outif != NULL && (*outif) != NULL &&
667 inp_restricted_send(inp, *outif)) {
668 soevent(inp->inp_socket,
669 (SO_FILT_HINT_LOCKED | SO_FILT_HINT_IFDENIED));
670 error = EHOSTUNREACH;
671 }
672 if (error == 0) {
673 error = EADDRNOTAVAIL;
674 }
675 return error;
676 }
677
678 *plocal_addr6 = *addr6;
679 /*
680 * Don't do pcblookup call here; return interface in
681 * plocal_addr6 and exit to caller, that will do the lookup.
682 */
683 return 0;
684 }
685
686 /*
687 * Outer subroutine:
688 * Connect from a socket to a specified address.
689 * Both address and port must be specified in argument sin.
690 * If don't have a local address for this socket yet,
691 * then pick one.
692 */
693 int
in6_pcbconnect(struct inpcb * inp,struct sockaddr * nam,struct proc * p)694 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct proc *p)
695 {
696 struct in6_addr addr6;
697 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)(void *)nam;
698 struct inpcb *pcb;
699 int error = 0;
700 struct ifnet *outif = NULL;
701 struct socket *so = inp->inp_socket;
702
703 #if CONTENT_FILTER
704 so->so_state_change_cnt++;
705 #endif
706
707 if (so->so_proto->pr_protocol == IPPROTO_UDP &&
708 sin6->sin6_port == htons(53) && !(so->so_flags1 & SOF1_DNS_COUNTED)) {
709 so->so_flags1 |= SOF1_DNS_COUNTED;
710 INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet_dgram_dns);
711 }
712
713 /*
714 * Call inner routine, to assign local interface address.
715 * in6_pcbladdr() may automatically fill in sin6_scope_id.
716 *
717 * in6_pcbladdr() might return an ifp with its reference held
718 * even in the error case, so make sure that it's released
719 * whenever it's non-NULL.
720 */
721 if ((error = in6_pcbladdr(inp, nam, &addr6, &outif)) != 0) {
722 if (outif != NULL && inp_restricted_send(inp, outif)) {
723 soevent(so,
724 (SO_FILT_HINT_LOCKED | SO_FILT_HINT_IFDENIED));
725 }
726 goto done;
727 }
728 socket_unlock(so, 0);
729
730 uint32_t lifscope;
731 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
732 lifscope = inp->inp_lifscope;
733 } else if (outif != NULL) {
734 lifscope = in6_addr2scopeid(outif, &addr6);
735 } else {
736 lifscope = sin6->sin6_scope_id;
737 }
738
739 pcb = in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr,
740 sin6->sin6_port, sin6->sin6_scope_id, IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ?
741 &addr6 : &inp->in6p_laddr, inp->inp_lport, lifscope, 0, NULL);
742 socket_lock(so, 0);
743 if (pcb != NULL) {
744 in_pcb_checkstate(pcb, WNT_RELEASE, pcb == inp ? 1 : 0);
745 error = EADDRINUSE;
746 goto done;
747 }
748 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
749 if (inp->inp_lport == 0) {
750 error = in6_pcbbind(inp, NULL, p);
751 if (error) {
752 goto done;
753 }
754 }
755 inp->in6p_laddr = addr6;
756 inp->in6p_last_outifp = outif; /* no reference needed */
757 if (IN6_IS_SCOPE_EMBED(&inp->in6p_laddr) &&
758 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sin6->sin6_addr)) {
759 inp->inp_lifscope = sin6->sin6_scope_id;
760 } else {
761 inp->inp_lifscope = lifscope;
762 }
763 in6_verify_ifscope(&inp->in6p_laddr, inp->inp_lifscope);
764 #if SKYWALK
765 if (NETNS_TOKEN_VALID(&inp->inp_netns_token)) {
766 netns_set_ifnet(&inp->inp_netns_token,
767 inp->in6p_last_outifp);
768 }
769 #endif /* SKYWALK */
770 inp->in6p_flags |= INP_IN6ADDR_ANY;
771 }
772 if (!lck_rw_try_lock_exclusive(&inp->inp_pcbinfo->ipi_lock)) {
773 /* lock inversion issue, mostly with udp multicast packets */
774 socket_unlock(so, 0);
775 lck_rw_lock_exclusive(&inp->inp_pcbinfo->ipi_lock);
776 socket_lock(so, 0);
777 }
778 inp->in6p_faddr = sin6->sin6_addr;
779 inp->inp_fport = sin6->sin6_port;
780 inp->inp_fifscope = sin6->sin6_scope_id;
781 in6_verify_ifscope(&inp->in6p_faddr, inp->inp_fifscope);
782 if (nstat_collect && SOCK_PROTO(so) == IPPROTO_UDP) {
783 nstat_pcb_invalidate_cache(inp);
784 }
785 in_pcbrehash(inp);
786 lck_rw_done(&inp->inp_pcbinfo->ipi_lock);
787
788 done:
789 if (outif != NULL) {
790 ifnet_release(outif);
791 }
792
793 return error;
794 }
795
796 void
in6_pcbdisconnect(struct inpcb * inp)797 in6_pcbdisconnect(struct inpcb *inp)
798 {
799 struct socket *so = inp->inp_socket;
800
801 #if CONTENT_FILTER
802 if (so) {
803 so->so_state_change_cnt++;
804 }
805 #endif
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 if (nstat_collect && SOCK_PROTO(so) == IPPROTO_UDP) {
814 nstat_pcb_cache(inp);
815 }
816 bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
817 inp->inp_fport = 0;
818 /* clear flowinfo - RFC 6437 */
819 inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
820 in_pcbrehash(inp);
821 lck_rw_done(&inp->inp_pcbinfo->ipi_lock);
822 /*
823 * A multipath subflow socket would have its SS_NOFDREF set by default,
824 * so check for SOF_MP_SUBFLOW socket flag before detaching the PCB;
825 * when the socket is closed for real, SOF_MP_SUBFLOW would be cleared.
826 */
827 if (!(so->so_flags & SOF_MP_SUBFLOW) && (so->so_state & SS_NOFDREF)) {
828 in6_pcbdetach(inp);
829 }
830 }
831
832 void
in6_pcbdetach(struct inpcb * inp)833 in6_pcbdetach(struct inpcb *inp)
834 {
835 struct socket *so = inp->inp_socket;
836
837 if (so->so_pcb == NULL) {
838 /* PCB has been disposed */
839 panic("%s: inp=%p so=%p proto=%d so_pcb is null!", __func__,
840 inp, so, SOCK_PROTO(so));
841 /* NOTREACHED */
842 }
843
844 #if IPSEC
845 if (inp->in6p_sp != NULL) {
846 (void) ipsec6_delete_pcbpolicy(inp);
847 }
848 #endif /* IPSEC */
849
850 if (inp->inp_stat != NULL && SOCK_PROTO(so) == IPPROTO_UDP) {
851 if (inp->inp_stat->rxpackets == 0 && inp->inp_stat->txpackets == 0) {
852 INC_ATOMIC_INT64_LIM(net_api_stats.nas_socket_inet6_dgram_no_data);
853 }
854 }
855
856 /*
857 * Let NetworkStatistics know this PCB is going away
858 * before we detach it.
859 */
860 if (nstat_collect &&
861 (SOCK_PROTO(so) == IPPROTO_TCP || SOCK_PROTO(so) == IPPROTO_UDP)) {
862 nstat_pcb_detach(inp);
863 }
864 /* mark socket state as dead */
865 if (in_pcb_checkstate(inp, WNT_STOPUSING, 1) != WNT_STOPUSING) {
866 panic("%s: so=%p proto=%d couldn't set to STOPUSING",
867 __func__, so, SOCK_PROTO(so));
868 /* NOTREACHED */
869 }
870
871 if (!(so->so_flags & SOF_PCBCLEARING)) {
872 struct ip_moptions *imo;
873 struct ip6_moptions *im6o;
874
875 inp->inp_vflag = 0;
876 if (inp->in6p_options != NULL) {
877 m_freem(inp->in6p_options);
878 inp->in6p_options = NULL;
879 }
880 ip6_freepcbopts(inp->in6p_outputopts);
881 inp->in6p_outputopts = NULL;
882 ROUTE_RELEASE(&inp->in6p_route);
883 /* free IPv4 related resources in case of mapped addr */
884 if (inp->inp_options != NULL) {
885 (void) m_free(inp->inp_options);
886 inp->inp_options = NULL;
887 }
888 im6o = inp->in6p_moptions;
889 inp->in6p_moptions = NULL;
890 if (im6o != NULL) {
891 IM6O_REMREF(im6o);
892 }
893 imo = inp->inp_moptions;
894 inp->inp_moptions = NULL;
895 if (imo != NULL) {
896 IMO_REMREF(imo);
897 }
898
899 sofreelastref(so, 0);
900 inp->inp_state = INPCB_STATE_DEAD;
901 /* makes sure we're not called twice from so_close */
902 so->so_flags |= SOF_PCBCLEARING;
903
904 inpcb_gc_sched(inp->inp_pcbinfo, INPCB_TIMER_FAST);
905 }
906 }
907
908 struct sockaddr *
in6_sockaddr(in_port_t port,struct in6_addr * addr_p,uint32_t ifscope)909 in6_sockaddr(in_port_t port, struct in6_addr *addr_p, uint32_t ifscope)
910 {
911 struct sockaddr_in6 *sin6;
912
913 sin6 = (struct sockaddr_in6 *)alloc_sockaddr(sizeof(*sin6),
914 Z_WAITOK | Z_NOFAIL);
915
916 sin6->sin6_family = AF_INET6;
917 sin6->sin6_port = port;
918 sin6->sin6_addr = *addr_p;
919
920 /* would be good to use sa6_recoverscope(), except for locking */
921 if (IN6_IS_SCOPE_EMBED(&sin6->sin6_addr)) {
922 sin6->sin6_scope_id = ifscope;
923 if (in6_embedded_scope) {
924 in6_verify_ifscope(&sin6->sin6_addr, ifscope);
925 sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
926 }
927 } else {
928 sin6->sin6_scope_id = 0; /* XXX */
929 }
930 if (in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
931 sin6->sin6_addr.s6_addr16[1] = 0;
932 }
933
934 return (struct sockaddr *)sin6;
935 }
936
937 void
in6_sockaddr_s(in_port_t port,struct in6_addr * addr_p,struct sockaddr_in6 * sin6,uint32_t ifscope)938 in6_sockaddr_s(in_port_t port, struct in6_addr *addr_p,
939 struct sockaddr_in6 *sin6, uint32_t ifscope)
940 {
941 bzero(sin6, sizeof(*sin6));
942 sin6->sin6_family = AF_INET6;
943 sin6->sin6_len = sizeof(*sin6);
944 sin6->sin6_port = port;
945 sin6->sin6_addr = *addr_p;
946
947 /* would be good to use sa6_recoverscope(), except for locking */
948 if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
949 sin6->sin6_scope_id = ifscope;
950 if (in6_embedded_scope) {
951 in6_verify_ifscope(&sin6->sin6_addr, ifscope);
952 sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
953 }
954 } else {
955 sin6->sin6_scope_id = 0; /* XXX */
956 }
957 if (in6_embedded_scope && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
958 sin6->sin6_addr.s6_addr16[1] = 0;
959 }
960 }
961
962 /*
963 * The calling convention of in6_getsockaddr() and in6_getpeeraddr() was
964 * modified to match the pru_sockaddr() and pru_peeraddr() entry points
965 * in struct pr_usrreqs, so that protocols can just reference then directly
966 * without the need for a wrapper function.
967 */
968 int
in6_getsockaddr(struct socket * so,struct sockaddr ** nam)969 in6_getsockaddr(struct socket *so, struct sockaddr **nam)
970 {
971 struct inpcb *inp;
972 struct in6_addr addr;
973 in_port_t port;
974
975 if ((inp = sotoinpcb(so)) == NULL) {
976 return EINVAL;
977 }
978
979 port = inp->inp_lport;
980 addr = inp->in6p_laddr;
981
982 *nam = in6_sockaddr(port, &addr, inp->inp_lifscope);
983 if (*nam == NULL) {
984 return ENOBUFS;
985 }
986 return 0;
987 }
988
989 int
in6_getsockaddr_s(struct socket * so,struct sockaddr_in6 * ss)990 in6_getsockaddr_s(struct socket *so, struct sockaddr_in6 *ss)
991 {
992 struct inpcb *inp;
993 struct in6_addr addr;
994 in_port_t port;
995
996 VERIFY(ss != NULL);
997 bzero(ss, sizeof(*ss));
998
999 if ((inp = sotoinpcb(so)) == NULL) {
1000 return EINVAL;
1001 }
1002
1003 port = inp->inp_lport;
1004 addr = inp->in6p_laddr;
1005
1006 in6_sockaddr_s(port, &addr, ss, inp->inp_lifscope);
1007 return 0;
1008 }
1009
1010 int
in6_getpeeraddr(struct socket * so,struct sockaddr ** nam)1011 in6_getpeeraddr(struct socket *so, struct sockaddr **nam)
1012 {
1013 struct inpcb *inp;
1014 struct in6_addr addr;
1015 in_port_t port;
1016
1017 if ((inp = sotoinpcb(so)) == NULL) {
1018 return EINVAL;
1019 }
1020
1021 port = inp->inp_fport;
1022 addr = inp->in6p_faddr;
1023
1024 *nam = in6_sockaddr(port, &addr, inp->inp_fifscope);
1025 if (*nam == NULL) {
1026 return ENOBUFS;
1027 }
1028 return 0;
1029 }
1030
1031 int
in6_mapped_sockaddr(struct socket * so,struct sockaddr ** nam)1032 in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam)
1033 {
1034 struct inpcb *inp = sotoinpcb(so);
1035 int error;
1036
1037 if (inp == NULL) {
1038 return EINVAL;
1039 }
1040 if (inp->inp_vflag & INP_IPV4) {
1041 error = in_getsockaddr(so, nam);
1042 if (error == 0) {
1043 error = in6_sin_2_v4mapsin6_in_sock(nam);
1044 }
1045 } else {
1046 /* scope issues will be handled in in6_getsockaddr(). */
1047 error = in6_getsockaddr(so, nam);
1048 }
1049 return error;
1050 }
1051
1052 int
in6_mapped_peeraddr(struct socket * so,struct sockaddr ** nam)1053 in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam)
1054 {
1055 struct inpcb *inp = sotoinpcb(so);
1056 int error;
1057
1058 if (inp == NULL) {
1059 return EINVAL;
1060 }
1061 if (inp->inp_vflag & INP_IPV4) {
1062 error = in_getpeeraddr(so, nam);
1063 if (error == 0) {
1064 error = in6_sin_2_v4mapsin6_in_sock(nam);
1065 }
1066 } else {
1067 /* scope issues will be handled in in6_getpeeraddr(). */
1068 error = in6_getpeeraddr(so, nam);
1069 }
1070 return error;
1071 }
1072
1073 /*
1074 * Pass some notification to all connections of a protocol
1075 * associated with address dst. The local address and/or port numbers
1076 * may be specified to limit the search. The "usual action" will be
1077 * taken, depending on the ctlinput cmd. The caller must filter any
1078 * cmds that are uninteresting (e.g., no error in the map).
1079 * Call the protocol specific routine (if any) to report
1080 * any errors for each matching socket.
1081 */
1082 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))1083 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst, u_int fport_arg,
1084 const struct sockaddr *src, u_int lport_arg, int cmd, void *cmdarg,
1085 void (*notify)(struct inpcb *, int))
1086 {
1087 struct inpcbhead *head = pcbinfo->ipi_listhead;
1088 struct inpcb *inp, *ninp;
1089 struct sockaddr_in6 sa6_src, *sa6_dst;
1090 uint16_t fport = (uint16_t)fport_arg, lport = (uint16_t)lport_arg;
1091 u_int32_t flowinfo;
1092 int errno;
1093
1094 if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6) {
1095 return;
1096 }
1097
1098 sa6_dst = (struct sockaddr_in6 *)(void *)dst;
1099 if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr)) {
1100 return;
1101 }
1102
1103 /*
1104 * note that src can be NULL when we get notify by local fragmentation.
1105 */
1106 sa6_src = (src == NULL) ?
1107 sa6_any : *(struct sockaddr_in6 *)(uintptr_t)(size_t)src;
1108 flowinfo = sa6_src.sin6_flowinfo;
1109
1110 /*
1111 * Redirects go to all references to the destination,
1112 * and use in6_rtchange to invalidate the route cache.
1113 * Dead host indications: also use in6_rtchange to invalidate
1114 * the cache, and deliver the error to all the sockets.
1115 * Otherwise, if we have knowledge of the local port and address,
1116 * deliver only to that socket.
1117 */
1118 if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
1119 fport = 0;
1120 lport = 0;
1121 bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
1122
1123 if (cmd != PRC_HOSTDEAD) {
1124 notify = in6_rtchange;
1125 }
1126 }
1127 errno = inet6ctlerrmap[cmd];
1128 lck_rw_lock_shared(&pcbinfo->ipi_lock);
1129 for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
1130 ninp = LIST_NEXT(inp, inp_list);
1131
1132 if (!(inp->inp_vflag & INP_IPV6)) {
1133 continue;
1134 }
1135
1136 /*
1137 * If the error designates a new path MTU for a destination
1138 * and the application (associated with this socket) wanted to
1139 * know the value, notify. Note that we notify for all
1140 * disconnected sockets if the corresponding application
1141 * wanted. This is because some UDP applications keep sending
1142 * sockets disconnected.
1143 * XXX: should we avoid to notify the value to TCP sockets?
1144 */
1145 if (cmd == PRC_MSGSIZE && cmdarg != NULL) {
1146 socket_lock(inp->inp_socket, 1);
1147 ip6_notify_pmtu(inp, (struct sockaddr_in6 *)(void *)dst,
1148 (u_int32_t *)cmdarg);
1149 socket_unlock(inp->inp_socket, 1);
1150 }
1151
1152 /*
1153 * Detect if we should notify the error. If no source and
1154 * destination ports are specifed, but non-zero flowinfo and
1155 * local address match, notify the error. This is the case
1156 * when the error is delivered with an encrypted buffer
1157 * by ESP. Otherwise, just compare addresses and ports
1158 * as usual.
1159 */
1160 if (lport == 0 && fport == 0 && flowinfo &&
1161 inp->inp_socket != NULL &&
1162 flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) &&
1163 in6_are_addr_equal_scoped(&inp->in6p_laddr, &sa6_src.sin6_addr, inp->inp_lifscope, sa6_src.sin6_scope_id)) {
1164 goto do_notify;
1165 } else if (!in6_are_addr_equal_scoped(&inp->in6p_faddr, &sa6_dst->sin6_addr,
1166 inp->inp_fifscope, sa6_dst->sin6_scope_id) || inp->inp_socket == NULL ||
1167 (lport && inp->inp_lport != lport) ||
1168 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
1169 !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)) {
1170 continue;
1171 }
1172
1173 do_notify:
1174 if (notify) {
1175 if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) ==
1176 WNT_STOPUSING) {
1177 continue;
1178 }
1179 socket_lock(inp->inp_socket, 1);
1180 (*notify)(inp, errno);
1181 (void) in_pcb_checkstate(inp, WNT_RELEASE, 1);
1182 socket_unlock(inp->inp_socket, 1);
1183 }
1184 }
1185 lck_rw_done(&pcbinfo->ipi_lock);
1186 }
1187
1188 /*
1189 * Lookup a PCB based on the local address and port.
1190 */
1191 struct inpcb *
in6_pcblookup_local(struct inpcbinfo * pcbinfo,struct in6_addr * laddr,u_int lport_arg,uint32_t ifscope,int wild_okay)1192 in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr,
1193 u_int lport_arg, uint32_t ifscope, int wild_okay)
1194 {
1195 struct inpcb *inp;
1196 int matchwild = 3, wildcard;
1197 uint16_t lport = (uint16_t)lport_arg;
1198 struct inpcbporthead *porthash;
1199 struct inpcb *match = NULL;
1200 struct inpcbport *phd;
1201
1202 if (!wild_okay) {
1203 struct inpcbhead *head;
1204 /*
1205 * Look for an unconnected (wildcard foreign addr) PCB that
1206 * matches the local address and port we're looking for.
1207 */
1208 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0,
1209 pcbinfo->ipi_hashmask)];
1210 LIST_FOREACH(inp, head, inp_hash) {
1211 if (!(inp->inp_vflag & INP_IPV6)) {
1212 continue;
1213 }
1214 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1215 in6_are_addr_equal_scoped(&inp->in6p_laddr, laddr, inp->inp_lifscope, ifscope) &&
1216 inp->inp_lport == lport) {
1217 /*
1218 * Found.
1219 */
1220 return inp;
1221 }
1222 }
1223 /*
1224 * Not found.
1225 */
1226 return NULL;
1227 }
1228 /*
1229 * Best fit PCB lookup.
1230 *
1231 * First see if this local port is in use by looking on the
1232 * port hash list.
1233 */
1234 porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
1235 pcbinfo->ipi_porthashmask)];
1236 LIST_FOREACH(phd, porthash, phd_hash) {
1237 if (phd->phd_port == lport) {
1238 break;
1239 }
1240 }
1241 if (phd != NULL) {
1242 /*
1243 * Port is in use by one or more PCBs. Look for best
1244 * fit.
1245 */
1246 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
1247 wildcard = 0;
1248 if (!(inp->inp_vflag & INP_IPV6)) {
1249 continue;
1250 }
1251 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
1252 wildcard++;
1253 }
1254 if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1255 if (IN6_IS_ADDR_UNSPECIFIED(laddr)) {
1256 wildcard++;
1257 } else if (!in6_are_addr_equal_scoped(
1258 &inp->in6p_laddr, laddr, inp->inp_lifscope, ifscope)) {
1259 continue;
1260 }
1261 } else {
1262 if (!IN6_IS_ADDR_UNSPECIFIED(laddr)) {
1263 wildcard++;
1264 }
1265 }
1266 if (wildcard < matchwild) {
1267 match = inp;
1268 matchwild = wildcard;
1269 if (matchwild == 0) {
1270 break;
1271 }
1272 }
1273 }
1274 }
1275 return match;
1276 }
1277
1278 /*
1279 * Check for alternatives when higher level complains
1280 * about service problems. For now, invalidate cached
1281 * routing information. If the route was created dynamically
1282 * (by a redirect), time to try a default gateway again.
1283 */
1284 void
in6_losing(struct inpcb * in6p)1285 in6_losing(struct inpcb *in6p)
1286 {
1287 struct rtentry *rt;
1288
1289 if ((rt = in6p->in6p_route.ro_rt) != NULL) {
1290 RT_LOCK(rt);
1291 if (rt->rt_flags & RTF_DYNAMIC) {
1292 /*
1293 * Prevent another thread from modifying rt_key,
1294 * rt_gateway via rt_setgate() after the rt_lock
1295 * is dropped by marking the route as defunct.
1296 */
1297 rt->rt_flags |= RTF_CONDEMNED;
1298 RT_UNLOCK(rt);
1299 (void) rtrequest(RTM_DELETE, rt_key(rt),
1300 rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
1301 } else {
1302 RT_UNLOCK(rt);
1303 }
1304 /*
1305 * A new route can be allocated
1306 * the next time output is attempted.
1307 */
1308 }
1309 ROUTE_RELEASE(&in6p->in6p_route);
1310 }
1311
1312 /*
1313 * After a routing change, flush old routing
1314 * and allocate a (hopefully) better one.
1315 */
1316 void
in6_rtchange(struct inpcb * inp,int errno)1317 in6_rtchange(struct inpcb *inp, int errno)
1318 {
1319 #pragma unused(errno)
1320 /*
1321 * A new route can be allocated the next time
1322 * output is attempted.
1323 */
1324 ROUTE_RELEASE(&inp->in6p_route);
1325 }
1326
1327 /*
1328 * Check if PCB exists hash list. Also returns uid and gid of socket
1329 */
1330 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)1331 in6_pcblookup_hash_exists(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1332 u_int fport_arg, uint32_t fifscope, struct in6_addr *laddr, u_int lport_arg, uint32_t lifscope, int wildcard,
1333 uid_t *uid, gid_t *gid, struct ifnet *ifp, bool relaxed)
1334 {
1335 struct inpcbhead *head;
1336 struct inpcb *inp;
1337 uint16_t fport = (uint16_t)fport_arg, lport = (uint16_t)lport_arg;
1338 int found;
1339
1340 *uid = UID_MAX;
1341 *gid = GID_MAX;
1342
1343 lck_rw_lock_shared(&pcbinfo->ipi_lock);
1344
1345 /*
1346 * First look for an exact match.
1347 */
1348 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr->s6_addr32[3] /* XXX */,
1349 lport, fport, pcbinfo->ipi_hashmask)];
1350 LIST_FOREACH(inp, head, inp_hash) {
1351 if (!(inp->inp_vflag & INP_IPV6)) {
1352 continue;
1353 }
1354
1355 if (inp_restricted_recv(inp, ifp)) {
1356 continue;
1357 }
1358
1359 #if NECP
1360 if (!necp_socket_is_allowed_to_recv_on_interface(inp, ifp)) {
1361 continue;
1362 }
1363 #endif /* NECP */
1364
1365 if (((in6_are_addr_equal_scoped(&inp->in6p_faddr, faddr, inp->inp_fifscope, fifscope) &&
1366 in6_are_addr_equal_scoped(&inp->in6p_laddr, laddr, inp->inp_lifscope, lifscope)) ||
1367 (relaxed &&
1368 IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
1369 IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr))) &&
1370 inp->inp_fport == fport &&
1371 inp->inp_lport == lport) {
1372 if ((found = (inp->inp_socket != NULL))) {
1373 /*
1374 * Found. Check if pcb is still valid
1375 */
1376 *uid = kauth_cred_getuid(
1377 inp->inp_socket->so_cred);
1378 *gid = kauth_cred_getgid(
1379 inp->inp_socket->so_cred);
1380 }
1381 lck_rw_done(&pcbinfo->ipi_lock);
1382 return found;
1383 }
1384 }
1385 if (wildcard) {
1386 struct inpcb *local_wild = NULL;
1387
1388 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0,
1389 pcbinfo->ipi_hashmask)];
1390 LIST_FOREACH(inp, head, inp_hash) {
1391 if (!(inp->inp_vflag & INP_IPV6)) {
1392 continue;
1393 }
1394
1395 if (inp_restricted_recv(inp, ifp)) {
1396 continue;
1397 }
1398
1399 #if NECP
1400 if (!necp_socket_is_allowed_to_recv_on_interface(inp, ifp)) {
1401 continue;
1402 }
1403 #endif /* NECP */
1404
1405 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1406 inp->inp_lport == lport) {
1407 if (in6_are_addr_equal_scoped(&inp->in6p_laddr,
1408 laddr, inp->inp_lifscope, lifscope) ||
1409 (relaxed && IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr))) {
1410 found = (inp->inp_socket != NULL);
1411 if (found) {
1412 *uid = kauth_cred_getuid(
1413 inp->inp_socket->so_cred);
1414 *gid = kauth_cred_getgid(
1415 inp->inp_socket->so_cred);
1416 }
1417 lck_rw_done(&pcbinfo->ipi_lock);
1418 return found;
1419 } else if (IN6_IS_ADDR_UNSPECIFIED(
1420 &inp->in6p_laddr)) {
1421 local_wild = inp;
1422 }
1423 }
1424 }
1425 if (local_wild) {
1426 if ((found = (local_wild->inp_socket != NULL))) {
1427 *uid = kauth_cred_getuid(
1428 local_wild->inp_socket->so_cred);
1429 *gid = kauth_cred_getgid(
1430 local_wild->inp_socket->so_cred);
1431 }
1432 lck_rw_done(&pcbinfo->ipi_lock);
1433 return found;
1434 }
1435 }
1436
1437 /*
1438 * Not found.
1439 */
1440 lck_rw_done(&pcbinfo->ipi_lock);
1441 return 0;
1442 }
1443
1444 /*
1445 * Lookup PCB in hash list.
1446 */
1447 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)1448 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1449 u_int fport_arg, uint32_t fifscope, struct in6_addr *laddr, u_int lport_arg, uint32_t lifscope, int wildcard,
1450 struct ifnet *ifp)
1451 {
1452 struct inpcbhead *head;
1453 struct inpcb *inp;
1454 uint16_t fport = (uint16_t)fport_arg, lport = (uint16_t)lport_arg;
1455
1456 lck_rw_lock_shared(&pcbinfo->ipi_lock);
1457
1458 /*
1459 * First look for an exact match.
1460 */
1461 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(faddr->s6_addr32[3] /* XXX */,
1462 lport, fport, pcbinfo->ipi_hashmask)];
1463 LIST_FOREACH(inp, head, inp_hash) {
1464 if (!(inp->inp_vflag & INP_IPV6)) {
1465 continue;
1466 }
1467
1468 if (inp_restricted_recv(inp, ifp)) {
1469 continue;
1470 }
1471
1472 #if NECP
1473 if (!necp_socket_is_allowed_to_recv_on_interface(inp, ifp)) {
1474 continue;
1475 }
1476 #endif /* NECP */
1477
1478 if (in6_are_addr_equal_scoped(&inp->in6p_faddr, faddr, inp->inp_fifscope, fifscope) &&
1479 in6_are_addr_equal_scoped(&inp->in6p_laddr, laddr, inp->inp_lifscope, lifscope) &&
1480 inp->inp_fport == fport &&
1481 inp->inp_lport == lport) {
1482 /*
1483 * Found. Check if pcb is still valid
1484 */
1485 if (in_pcb_checkstate(inp, WNT_ACQUIRE, 0) !=
1486 WNT_STOPUSING) {
1487 lck_rw_done(&pcbinfo->ipi_lock);
1488 return inp;
1489 } else {
1490 /* it's there but dead, say it isn't found */
1491 lck_rw_done(&pcbinfo->ipi_lock);
1492 return NULL;
1493 }
1494 }
1495 }
1496 if (wildcard) {
1497 struct inpcb *local_wild = NULL;
1498
1499 head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport, 0,
1500 pcbinfo->ipi_hashmask)];
1501 LIST_FOREACH(inp, head, inp_hash) {
1502 if (!(inp->inp_vflag & INP_IPV6)) {
1503 continue;
1504 }
1505
1506 if (inp_restricted_recv(inp, ifp)) {
1507 continue;
1508 }
1509
1510 #if NECP
1511 if (!necp_socket_is_allowed_to_recv_on_interface(inp, ifp)) {
1512 continue;
1513 }
1514 #endif /* NECP */
1515
1516 if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1517 inp->inp_lport == lport) {
1518 if (in6_are_addr_equal_scoped(&inp->in6p_laddr,
1519 laddr, inp->inp_lifscope, lifscope)) {
1520 if (in_pcb_checkstate(inp, WNT_ACQUIRE,
1521 0) != WNT_STOPUSING) {
1522 lck_rw_done(&pcbinfo->ipi_lock);
1523 return inp;
1524 } else {
1525 /* dead; say it isn't found */
1526 lck_rw_done(&pcbinfo->ipi_lock);
1527 return NULL;
1528 }
1529 } else if (IN6_IS_ADDR_UNSPECIFIED(
1530 &inp->in6p_laddr)) {
1531 local_wild = inp;
1532 }
1533 }
1534 }
1535 if (local_wild && in_pcb_checkstate(local_wild,
1536 WNT_ACQUIRE, 0) != WNT_STOPUSING) {
1537 lck_rw_done(&pcbinfo->ipi_lock);
1538 return local_wild;
1539 } else {
1540 lck_rw_done(&pcbinfo->ipi_lock);
1541 return NULL;
1542 }
1543 }
1544
1545 /*
1546 * Not found.
1547 */
1548 lck_rw_done(&pcbinfo->ipi_lock);
1549 return NULL;
1550 }
1551
1552 void
init_sin6(struct sockaddr_in6 * sin6,struct mbuf * m)1553 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m)
1554 {
1555 struct ip6_hdr *ip;
1556
1557 ip = mtod(m, struct ip6_hdr *);
1558 bzero(sin6, sizeof(*sin6));
1559 sin6->sin6_len = sizeof(*sin6);
1560 sin6->sin6_family = AF_INET6;
1561 sin6->sin6_addr = ip->ip6_src;
1562 if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr)) {
1563 if (in6_embedded_scope) {
1564 sin6->sin6_addr.s6_addr16[1] = 0;
1565 }
1566 if ((m->m_pkthdr.pkt_flags & (PKTF_LOOP | PKTF_IFAINFO)) ==
1567 (PKTF_LOOP | PKTF_IFAINFO)) {
1568 sin6->sin6_scope_id = m->m_pkthdr.src_ifindex;
1569 } else if (m->m_pkthdr.rcvif != NULL) {
1570 sin6->sin6_scope_id = m->m_pkthdr.rcvif->if_index;
1571 }
1572 }
1573 }
1574
1575 /*
1576 * The following routines implement this scheme:
1577 *
1578 * Callers of ip6_output() that intend to cache the route in the inpcb pass
1579 * a local copy of the struct route to ip6_output(). Using a local copy of
1580 * the cached route significantly simplifies things as IP no longer has to
1581 * worry about having exclusive access to the passed in struct route, since
1582 * it's defined in the caller's stack; in essence, this allows for a lock-
1583 * less operation when updating the struct route at the IP level and below,
1584 * whenever necessary. The scheme works as follows:
1585 *
1586 * Prior to dropping the socket's lock and calling ip6_output(), the caller
1587 * copies the struct route from the inpcb into its stack, and adds a reference
1588 * to the cached route entry, if there was any. The socket's lock is then
1589 * dropped and ip6_output() is called with a pointer to the copy of struct
1590 * route defined on the stack (not to the one in the inpcb.)
1591 *
1592 * Upon returning from ip6_output(), the caller then acquires the socket's
1593 * lock and synchronizes the cache; if there is no route cached in the inpcb,
1594 * it copies the local copy of struct route (which may or may not contain any
1595 * route) back into the cache; otherwise, if the inpcb has a route cached in
1596 * it, the one in the local copy will be freed, if there's any. Trashing the
1597 * cached route in the inpcb can be avoided because ip6_output() is single-
1598 * threaded per-PCB (i.e. multiple transmits on a PCB are always serialized
1599 * by the socket/transport layer.)
1600 */
1601 void
in6p_route_copyout(struct inpcb * inp,struct route_in6 * dst)1602 in6p_route_copyout(struct inpcb *inp, struct route_in6 *dst)
1603 {
1604 struct route_in6 *src = &inp->in6p_route;
1605
1606 socket_lock_assert_owned(inp->inp_socket);
1607
1608 /* Minor sanity check */
1609 if (src->ro_rt != NULL && rt_key(src->ro_rt)->sa_family != AF_INET6) {
1610 panic("%s: wrong or corrupted route: %p", __func__, src);
1611 }
1612
1613 route_copyout((struct route *)dst, (struct route *)src, sizeof(*dst));
1614 }
1615
1616 void
in6p_route_copyin(struct inpcb * inp,struct route_in6 * src)1617 in6p_route_copyin(struct inpcb *inp, struct route_in6 *src)
1618 {
1619 struct route_in6 *dst = &inp->in6p_route;
1620
1621 socket_lock_assert_owned(inp->inp_socket);
1622
1623 /* Minor sanity check */
1624 if (src->ro_rt != NULL && rt_key(src->ro_rt)->sa_family != AF_INET6) {
1625 panic("%s: wrong or corrupted route: %p", __func__, src);
1626 }
1627
1628 route_copyin((struct route *)src, (struct route *)dst, sizeof(*src));
1629 }
1630