1/*
2 * Copyright (c) 2000-2018 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/*
30 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
31 * All rights reserved.
32 *
33 * Redistribution and use in source and binary forms, with or without
34 * modification, are permitted provided that the following conditions
35 * are met:
36 * 1. Redistributions of source code must retain the above copyright
37 * notice, this list of conditions and the following disclaimer.
38 * 2. Redistributions in binary form must reproduce the above copyright
39 * notice, this list of conditions and the following disclaimer in the
40 * documentation and/or other materials provided with the distribution.
41 * 3. Neither the name of the project nor the names of its contributors
42 * may be used to endorse or promote products derived from this software
43 * without specific prior written permission.
44 *
45 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
46 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
47 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
48 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
49 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
50 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
51 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
52 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
53 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
54 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
55 * SUCH DAMAGE.
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
94#include <sys/param.h>
95#include <sys/systm.h>
96#include <sys/malloc.h>
97#include <sys/mbuf.h>
98#include <sys/protosw.h>
99#include <sys/socket.h>
100#include <sys/socketvar.h>
101#include <sys/errno.h>
102#include <sys/time.h>
103#include <sys/proc.h>
104#include <sys/sysctl.h>
105#include <sys/kauth.h>
106#include <sys/priv.h>
107#include <kern/locks.h>
108
109#include <net/if.h>
110#include <net/if_types.h>
111#include <net/route.h>
112
113#include <netinet/in.h>
114#include <netinet/in_var.h>
115#include <netinet/in_systm.h>
116#include <netinet/ip.h>
117#include <netinet/in_pcb.h>
118#include <netinet6/in6_var.h>
119#include <netinet/ip6.h>
120#include <netinet6/in6_pcb.h>
121#include <netinet6/ip6_var.h>
122#include <netinet6/scope6_var.h>
123#include <netinet6/nd6.h>
124
125#include <net/net_osdep.h>
126
127#include "loop.h"
128
129SYSCTL_DECL(_net_inet6_ip6);
130
131static int ip6_select_srcif_debug = 0;
132SYSCTL_INT(_net_inet6_ip6, OID_AUTO, select_srcif_debug,
133 CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_select_srcif_debug, 0,
134 "log source interface selection debug info");
135
136static int ip6_select_srcaddr_debug = 0;
137SYSCTL_INT(_net_inet6_ip6, OID_AUTO, select_srcaddr_debug,
138 CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_select_srcaddr_debug, 0,
139 "log source address selection debug info");
140
141static int ip6_select_src_expensive_secondary_if = 0;
142SYSCTL_INT(_net_inet6_ip6, OID_AUTO, select_src_expensive_secondary_if,
143 CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_select_src_expensive_secondary_if, 0,
144 "allow source interface selection to use expensive secondaries");
145
146static int ip6_select_src_strong_end = 1;
147SYSCTL_INT(_net_inet6_ip6, OID_AUTO, select_src_strong_end,
148 CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_select_src_strong_end, 0,
149 "limit source address selection to outgoing interface");
150
151#define ADDR_LABEL_NOTAPP (-1)
152struct in6_addrpolicy defaultaddrpolicy;
153
154int ip6_prefer_tempaddr = 1;
155#ifdef ENABLE_ADDRSEL
156extern lck_mtx_t *addrsel_mutex;
157#define ADDRSEL_LOCK() lck_mtx_lock(addrsel_mutex)
158#define ADDRSEL_UNLOCK() lck_mtx_unlock(addrsel_mutex)
159#else
160#define ADDRSEL_LOCK()
161#define ADDRSEL_UNLOCK()
162#endif
163
164static int selectroute(struct sockaddr_in6 *, struct sockaddr_in6 *,
165 struct ip6_pktopts *, struct ip6_moptions *, struct in6_ifaddr **,
166 struct route_in6 *, struct ifnet **, struct rtentry **, int, int,
167 struct ip6_out_args *ip6oa);
168static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
169 struct ip6_moptions *, struct route_in6 *ro,
170 struct ip6_out_args *, struct ifnet **);
171static void init_policy_queue(void);
172static int add_addrsel_policyent(const struct in6_addrpolicy *);
173#ifdef ENABLE_ADDRSEL
174static int delete_addrsel_policyent(const struct in6_addrpolicy *);
175#endif
176static int walk_addrsel_policy(int (*)(const struct in6_addrpolicy *, void *),
177 void *);
178static int dump_addrsel_policyent(const struct in6_addrpolicy *, void *);
179static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
180void addrsel_policy_init(void);
181
182#define SASEL_DO_DBG(inp) \
183 (ip6_select_srcaddr_debug && (inp) != NULL && \
184 (inp)->inp_socket != NULL && \
185 ((inp)->inp_socket->so_options & SO_DEBUG))
186
187#define SASEL_LOG(fmt, ...) \
188do { \
189 if (srcsel_debug) \
190 printf("%s:%d " fmt "\n",\
191 __FUNCTION__, __LINE__, ##__VA_ARGS__); \
192} while (0); \
193
194/*
195 * Return an IPv6 address, which is the most appropriate for a given
196 * destination and user specified options.
197 * If necessary, this function lookups the routing table and returns
198 * an entry to the caller for later use.
199 */
200#define REPLACE(r) do {\
201 SASEL_LOG("REPLACE r %d ia %s ifp1 %s\n", \
202 (r), s_src, ifp1->if_xname); \
203 srcrule = (r); \
204 goto replace; \
205} while (0)
206
207#define NEXTSRC(r) do {\
208 SASEL_LOG("NEXTSRC r %d ia %s ifp1 %s\n", \
209 (r), s_src, ifp1->if_xname); \
210 goto next; /* XXX: we can't use 'continue' here */ \
211} while (0)
212
213#define BREAK(r) do { \
214 SASEL_LOG("BREAK r %d ia %s ifp1 %s\n", \
215 (r), s_src, ifp1->if_xname); \
216 srcrule = (r); \
217 goto out; /* XXX: we can't use 'break' here */ \
218} while (0)
219
220
221struct ifaddr *
222in6_selectsrc_core_ifa(struct sockaddr_in6 *addr, struct ifnet *ifp, int srcsel_debug) {
223 int err = 0;
224 struct ifnet *src_ifp = NULL;
225 struct in6_addr src_storage = {};
226 struct in6_addr *in6 = NULL;
227 struct ifaddr *ifa = NULL;
228
229 if((in6 = in6_selectsrc_core(addr,
230 (ip6_prefer_tempaddr ? IPV6_SRCSEL_HINT_PREFER_TMPADDR : 0),
231 ifp, 0, &src_storage, &src_ifp, &err, &ifa)) == NULL) {
232 if (err == 0)
233 err = EADDRNOTAVAIL;
234 VERIFY(src_ifp == NULL);
235 if (ifa != NULL) {
236 IFA_REMREF(ifa);
237 ifa = NULL;
238 }
239 goto done;
240 }
241
242 if (src_ifp != ifp) {
243 if (err == 0)
244 err = ENETUNREACH;
245 if (ifa != NULL) {
246 IFA_REMREF(ifa);
247 ifa = NULL;
248 }
249 goto done;
250 }
251
252 VERIFY(ifa != NULL);
253 ifnet_lock_shared(ifp);
254 if ((ifa->ifa_debug & IFD_DETACHING) != 0) {
255 err = EHOSTUNREACH;
256 ifnet_lock_done(ifp);
257 if (ifa != NULL) {
258 IFA_REMREF(ifa);
259 ifa = NULL;
260 }
261 goto done;
262 }
263 ifnet_lock_done(ifp);
264
265done:
266 SASEL_LOG("Returned with error: %d", err);
267 if (src_ifp != NULL)
268 ifnet_release(src_ifp);
269 return (ifa);
270}
271
272struct in6_addr *
273in6_selectsrc_core(struct sockaddr_in6 *dstsock, uint32_t hint_mask,
274 struct ifnet *ifp, int srcsel_debug, struct in6_addr *src_storage,
275 struct ifnet **sifp, int *errorp, struct ifaddr **ifapp)
276{
277 u_int32_t odstzone;
278 int bestrule = IP6S_SRCRULE_0;
279 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
280 struct in6_addr dst;
281 struct in6_ifaddr *ia = NULL, *ia_best = NULL;
282 char s_src[MAX_IPv6_STR_LEN] = {0};
283 char s_dst[MAX_IPv6_STR_LEN] = {0};
284 const struct in6_addr *tmp = NULL;
285 int dst_scope = -1, best_scope = -1, best_matchlen = -1;
286 uint64_t secs = net_uptime();
287 VERIFY(dstsock != NULL);
288 VERIFY(src_storage != NULL);
289 VERIFY(ifp != NULL);
290
291 if (sifp != NULL)
292 *sifp = NULL;
293
294 if (ifapp != NULL)
295 *ifapp = NULL;
296
297 dst = dstsock->sin6_addr; /* make a copy for local operation */
298
299 if (srcsel_debug) {
300 (void) inet_ntop(AF_INET6, &dst, s_dst, sizeof (s_src));
301
302 tmp = &in6addr_any;
303 (void) inet_ntop(AF_INET6, tmp, s_src, sizeof (s_src));
304 printf("%s out src %s dst %s ifp %s\n",
305 __func__, s_src, s_dst, ifp->if_xname);
306 }
307
308 *errorp = in6_setscope(&dst, ifp, &odstzone);
309 if (*errorp != 0) {
310 src_storage = NULL;
311 goto done;
312 }
313
314 lck_rw_lock_shared(&in6_ifaddr_rwlock);
315 for (ia = in6_ifaddrs; ia; ia = ia->ia_next) {
316 int new_scope = -1, new_matchlen = -1;
317 struct in6_addrpolicy *new_policy = NULL;
318 u_int32_t srczone = 0, osrczone, dstzone;
319 struct in6_addr src;
320 struct ifnet *ifp1 = ia->ia_ifp;
321 int srcrule;
322
323 if (srcsel_debug)
324 (void) inet_ntop(AF_INET6, &ia->ia_addr.sin6_addr,
325 s_src, sizeof (s_src));
326
327 IFA_LOCK(&ia->ia_ifa);
328
329 /*
330 * Simply skip addresses reserved for CLAT46
331 */
332 if (ia->ia6_flags & IN6_IFF_CLAT46) {
333 SASEL_LOG("NEXT ia %s address on ifp1 %s skipped as it is "
334 "reserved for CLAT46", s_src, ifp1->if_xname);
335 goto next;
336 }
337
338 /*
339 * XXX By default we are strong end system and will
340 * limit candidate set of source address to the ones
341 * configured on the outgoing interface.
342 */
343 if (ip6_select_src_strong_end &&
344 ifp1 != ifp) {
345 SASEL_LOG("NEXT ia %s ifp1 %s address is not on outgoing "
346 "interface \n", s_src, ifp1->if_xname);
347 goto next;
348 }
349
350 /*
351 * We'll never take an address that breaks the scope zone
352 * of the destination. We also skip an address if its zone
353 * does not contain the outgoing interface.
354 * XXX: we should probably use sin6_scope_id here.
355 */
356 if (in6_setscope(&dst, ifp1, &dstzone) ||
357 odstzone != dstzone) {
358 SASEL_LOG("NEXT ia %s ifp1 %s odstzone %d != dstzone %d\n",
359 s_src, ifp1->if_xname, odstzone, dstzone);
360 goto next;
361 }
362 src = ia->ia_addr.sin6_addr;
363 if (in6_setscope(&src, ifp, &osrczone) ||
364 in6_setscope(&src, ifp1, &srczone) ||
365 osrczone != srczone) {
366 SASEL_LOG("NEXT ia %s ifp1 %s osrczone %d != srczone %d\n",
367 s_src, ifp1->if_xname, osrczone, srczone);
368 goto next;
369 }
370 /* avoid unusable addresses */
371 if ((ia->ia6_flags &
372 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
373 SASEL_LOG("NEXT ia %s ifp1 %s ia6_flags 0x%x\n",
374 s_src, ifp1->if_xname, ia->ia6_flags);
375 goto next;
376 }
377 if (!ip6_use_deprecated && IFA6_IS_DEPRECATED(ia, secs)) {
378 SASEL_LOG("NEXT ia %s ifp1 %s IFA6_IS_DEPRECATED\n",
379 s_src, ifp1->if_xname);
380 goto next;
381 }
382 if (!nd6_optimistic_dad &&
383 (ia->ia6_flags & IN6_IFF_OPTIMISTIC) != 0) {
384 SASEL_LOG("NEXT ia %s ifp1 %s IN6_IFF_OPTIMISTIC\n",
385 s_src, ifp1->if_xname);
386 goto next;
387 }
388 /* Rule 1: Prefer same address */
389 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr))
390 BREAK(IP6S_SRCRULE_1); /* there should be no better candidate */
391
392 if (ia_best == NULL)
393 REPLACE(IP6S_SRCRULE_0);
394
395 /* Rule 2: Prefer appropriate scope */
396 if (dst_scope < 0)
397 dst_scope = in6_addrscope(&dst);
398 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
399 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
400 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
401 REPLACE(IP6S_SRCRULE_2);
402 NEXTSRC(IP6S_SRCRULE_2);
403 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
404 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
405 NEXTSRC(IP6S_SRCRULE_2);
406 REPLACE(IP6S_SRCRULE_2);
407 }
408
409 /*
410 * Rule 3: Avoid deprecated addresses. Note that the case of
411 * !ip6_use_deprecated is already rejected above.
412 */
413 if (!IFA6_IS_DEPRECATED(ia_best, secs) &&
414 IFA6_IS_DEPRECATED(ia, secs))
415 NEXTSRC(IP6S_SRCRULE_3);
416 if (IFA6_IS_DEPRECATED(ia_best, secs) &&
417 !IFA6_IS_DEPRECATED(ia, secs))
418 REPLACE(IP6S_SRCRULE_3);
419
420 /*
421 * RFC 4429 says that optimistic addresses are equivalent to
422 * deprecated addresses, so avoid them here.
423 */
424 if ((ia_best->ia6_flags & IN6_IFF_OPTIMISTIC) == 0 &&
425 (ia->ia6_flags & IN6_IFF_OPTIMISTIC) != 0)
426 NEXTSRC(IP6S_SRCRULE_3);
427 if ((ia_best->ia6_flags & IN6_IFF_OPTIMISTIC) != 0 &&
428 (ia->ia6_flags & IN6_IFF_OPTIMISTIC) == 0)
429 REPLACE(IP6S_SRCRULE_3);
430
431 /* Rule 4: Prefer home addresses */
432 /*
433 * XXX: This is a TODO. We should probably merge the MIP6
434 * case above.
435 */
436
437 /* Rule 5: Prefer outgoing interface */
438 /*
439 * XXX By default we are strong end with source address
440 * selection. That means all address selection candidate
441 * addresses will be the ones hosted on the outgoing interface
442 * making the following check redundant.
443 */
444 if (ip6_select_src_strong_end == 0) {
445 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
446 NEXTSRC(IP6S_SRCRULE_5);
447 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
448 REPLACE(IP6S_SRCRULE_5);
449 }
450
451 /*
452 * Rule 6: Prefer matching label
453 * Note that best_policy should be non-NULL here.
454 */
455 if (dst_policy == NULL)
456 dst_policy = in6_addrsel_lookup_policy(dstsock);
457 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
458 new_policy = in6_addrsel_lookup_policy(&ia->ia_addr);
459 if (dst_policy->label == best_policy->label &&
460 dst_policy->label != new_policy->label)
461 NEXTSRC(IP6S_SRCRULE_6);
462 if (dst_policy->label != best_policy->label &&
463 dst_policy->label == new_policy->label)
464 REPLACE(IP6S_SRCRULE_6);
465 }
466
467 /*
468 * Rule 7: Prefer temporary addresses.
469 * We allow users to reverse the logic by configuring
470 * a sysctl variable, so that transparency conscious users can
471 * always prefer stable addresses.
472 */
473 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
474 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
475 if (hint_mask & IPV6_SRCSEL_HINT_PREFER_TMPADDR)
476 REPLACE(IP6S_SRCRULE_7);
477 else
478 NEXTSRC(IP6S_SRCRULE_7);
479 }
480 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
481 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
482 if (hint_mask & IPV6_SRCSEL_HINT_PREFER_TMPADDR)
483 NEXTSRC(IP6S_SRCRULE_7);
484 else
485 REPLACE(IP6S_SRCRULE_7);
486 }
487
488 /*
489 * Rule 7x: prefer addresses on alive interfaces.
490 * This is a KAME specific rule.
491 */
492 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
493 !(ia->ia_ifp->if_flags & IFF_UP))
494 NEXTSRC(IP6S_SRCRULE_7x);
495 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
496 (ia->ia_ifp->if_flags & IFF_UP))
497 REPLACE(IP6S_SRCRULE_7x);
498
499 /*
500 * Rule 8: Use longest matching prefix.
501 */
502 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
503 if (best_matchlen < new_matchlen)
504 REPLACE(IP6S_SRCRULE_8);
505 if (new_matchlen < best_matchlen)
506 NEXTSRC(IP6S_SRCRULE_8);
507
508 /*
509 * Last resort: just keep the current candidate.
510 * Or, do we need more rules?
511 */
512 if (ifp1 != ifp && (ifp1->if_eflags & IFEF_EXPENSIVE) &&
513 ip6_select_src_expensive_secondary_if == 0) {
514 SASEL_LOG("NEXT ia %s ifp1 %s IFEF_EXPENSIVE\n",
515 s_src, ifp1->if_xname);
516 ip6stat.ip6s_sources_skip_expensive_secondary_if++;
517 goto next;
518 }
519 SASEL_LOG("NEXT ia %s ifp1 %s last resort\n",
520 s_src, ifp1->if_xname);
521 IFA_UNLOCK(&ia->ia_ifa);
522 continue;
523
524replace:
525 /*
526 * Ignore addresses on secondary interfaces that are marked
527 * expensive
528 */
529 if (ifp1 != ifp && (ifp1->if_eflags & IFEF_EXPENSIVE) &&
530 ip6_select_src_expensive_secondary_if == 0) {
531 SASEL_LOG("NEXT ia %s ifp1 %s IFEF_EXPENSIVE\n",
532 s_src, ifp1->if_xname);
533 ip6stat.ip6s_sources_skip_expensive_secondary_if++;
534 goto next;
535 }
536 bestrule = srcrule;
537 best_scope = (new_scope >= 0 ? new_scope :
538 in6_addrscope(&ia->ia_addr.sin6_addr));
539 best_policy = (new_policy ? new_policy :
540 in6_addrsel_lookup_policy(&ia->ia_addr));
541 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
542 in6_matchlen(&ia->ia_addr.sin6_addr, &dst));
543 SASEL_LOG("NEXT ia %s ifp1 %s best_scope %d new_scope %d dst_scope %d\n",
544 s_src, ifp1->if_xname, best_scope, new_scope, dst_scope);
545 IFA_ADDREF_LOCKED(&ia->ia_ifa); /* for ia_best */
546 IFA_UNLOCK(&ia->ia_ifa);
547 if (ia_best != NULL)
548 IFA_REMREF(&ia_best->ia_ifa);
549 ia_best = ia;
550 continue;
551
552next:
553 IFA_UNLOCK(&ia->ia_ifa);
554 continue;
555
556out:
557 IFA_ADDREF_LOCKED(&ia->ia_ifa); /* for ia_best */
558 IFA_UNLOCK(&ia->ia_ifa);
559 if (ia_best != NULL)
560 IFA_REMREF(&ia_best->ia_ifa);
561 ia_best = ia;
562 break;
563 }
564
565 lck_rw_done(&in6_ifaddr_rwlock);
566
567 if ((ia = ia_best) == NULL) {
568 if (*errorp == 0)
569 *errorp = EADDRNOTAVAIL;
570 src_storage = NULL;
571 goto done;
572 }
573
574 if (sifp != NULL) {
575 *sifp = ia->ia_ifa.ifa_ifp;
576 ifnet_reference(*sifp);
577 }
578
579 IFA_LOCK_SPIN(&ia->ia_ifa);
580 if (bestrule < IP6S_SRCRULE_COUNT)
581 ip6stat.ip6s_sources_rule[bestrule]++;
582 *src_storage = satosin6(&ia->ia_addr)->sin6_addr;
583 IFA_UNLOCK(&ia->ia_ifa);
584
585 if (ifapp != NULL)
586 *ifapp = &ia->ia_ifa;
587 else
588 IFA_REMREF(&ia->ia_ifa);
589
590done:
591 if (srcsel_debug) {
592 (void) inet_ntop(AF_INET6, &dst, s_dst, sizeof (s_src));
593
594 tmp = (src_storage != NULL) ? src_storage : &in6addr_any;
595 (void) inet_ntop(AF_INET6, tmp, s_src, sizeof (s_src));
596
597 printf("%s out src %s dst %s dst_scope %d best_scope %d\n",
598 __func__, s_src, s_dst, dst_scope, best_scope);
599 }
600
601 return (src_storage);
602}
603
604/*
605 * Regardless of error, it will return an ifp with a reference held if the
606 * caller provides a non-NULL ifpp. The caller is responsible for checking
607 * if the returned ifp is valid and release its reference at all times.
608 */
609struct in6_addr *
610in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
611 struct inpcb *inp, struct route_in6 *ro,
612 struct ifnet **ifpp, struct in6_addr *src_storage, unsigned int ifscope,
613 int *errorp)
614{
615 struct ifnet *ifp = NULL;
616 struct in6_pktinfo *pi = NULL;
617 struct ip6_moptions *mopts;
618 struct ip6_out_args ip6oa;
619 boolean_t inp_debug = FALSE;
620 uint32_t hint_mask = 0;
621 int prefer_tempaddr = 0;
622 struct ifnet *sifp = NULL;
623
624 bzero(&ip6oa, sizeof(ip6oa));
625 ip6oa.ip6oa_boundif = ifscope;
626 ip6oa.ip6oa_flags = IP6OAF_SELECT_SRCIF;
627 ip6oa.ip6oa_sotc = SO_TC_UNSPEC;
628 ip6oa.ip6oa_netsvctype = _NET_SERVICE_TYPE_UNSPEC;
629
630 *errorp = 0;
631 if (ifpp != NULL)
632 *ifpp = NULL;
633
634 if (inp != NULL) {
635 inp_debug = SASEL_DO_DBG(inp);
636 mopts = inp->in6p_moptions;
637 if (INP_NO_CELLULAR(inp))
638 ip6oa.ip6oa_flags |= IP6OAF_NO_CELLULAR;
639 if (INP_NO_EXPENSIVE(inp))
640 ip6oa.ip6oa_flags |= IP6OAF_NO_EXPENSIVE;
641 if (INP_AWDL_UNRESTRICTED(inp))
642 ip6oa.ip6oa_flags |= IP6OAF_AWDL_UNRESTRICTED;
643 if (INP_INTCOPROC_ALLOWED(inp))
644 ip6oa.ip6oa_flags |= IP6OAF_INTCOPROC_ALLOWED;
645 } else {
646 mopts = NULL;
647 /* Allow the kernel to retransmit packets. */
648 ip6oa.ip6oa_flags |= IP6OAF_INTCOPROC_ALLOWED |
649 IP6OAF_AWDL_UNRESTRICTED;
650 }
651
652 if (ip6oa.ip6oa_boundif != IFSCOPE_NONE)
653 ip6oa.ip6oa_flags |= IP6OAF_BOUND_IF;
654
655 /*
656 * If the source address is explicitly specified by the caller,
657 * check if the requested source address is indeed a unicast address
658 * assigned to the node, and can be used as the packet's source
659 * address. If everything is okay, use the address as source.
660 */
661 if (opts && (pi = opts->ip6po_pktinfo) &&
662 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
663 struct sockaddr_in6 srcsock;
664 struct in6_ifaddr *ia6;
665
666 /* get the outgoing interface */
667 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ip6oa,
668 &ifp)) != 0) {
669 src_storage = NULL;
670 goto done;
671 }
672
673 /*
674 * determine the appropriate zone id of the source based on
675 * the zone of the destination and the outgoing interface.
676 * If the specified address is ambiguous wrt the scope zone,
677 * the interface must be specified; otherwise, ifa_ifwithaddr()
678 * will fail matching the address.
679 */
680 bzero(&srcsock, sizeof (srcsock));
681 srcsock.sin6_family = AF_INET6;
682 srcsock.sin6_len = sizeof (srcsock);
683 srcsock.sin6_addr = pi->ipi6_addr;
684 if (ifp != NULL) {
685 *errorp = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
686 if (*errorp != 0) {
687 src_storage = NULL;
688 goto done;
689 }
690 }
691 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)
692 (&srcsock));
693 if (ia6 == NULL) {
694 *errorp = EADDRNOTAVAIL;
695 src_storage = NULL;
696 goto done;
697 }
698 IFA_LOCK_SPIN(&ia6->ia_ifa);
699 if ((ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY | IN6_IFF_CLAT46)) ||
700 (inp && inp_restricted_send(inp, ia6->ia_ifa.ifa_ifp))) {
701 IFA_UNLOCK(&ia6->ia_ifa);
702 IFA_REMREF(&ia6->ia_ifa);
703 *errorp = EHOSTUNREACH;
704 src_storage = NULL;
705 goto done;
706 }
707
708 *src_storage = satosin6(&ia6->ia_addr)->sin6_addr;
709 IFA_UNLOCK(&ia6->ia_ifa);
710 IFA_REMREF(&ia6->ia_ifa);
711 goto done;
712 }
713
714 /*
715 * Otherwise, if the socket has already bound the source, just use it.
716 */
717 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
718 src_storage = &inp->in6p_laddr;
719 goto done;
720 }
721
722 /*
723 * If the address is not specified, choose the best one based on
724 * the outgoing interface and the destination address.
725 */
726 /* get the outgoing interface */
727 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ip6oa,
728 &ifp)) != 0) {
729 src_storage = NULL;
730 goto done;
731 }
732
733 VERIFY(ifp != NULL);
734
735 if (opts == NULL ||
736 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
737 prefer_tempaddr = ip6_prefer_tempaddr;
738 } else if (opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_NOTPREFER) {
739 prefer_tempaddr = 0;
740 } else
741 prefer_tempaddr = 1;
742
743 if (prefer_tempaddr)
744 hint_mask |= IPV6_SRCSEL_HINT_PREFER_TMPADDR;
745
746 if (in6_selectsrc_core(dstsock, hint_mask, ifp, inp_debug, src_storage,
747 &sifp, errorp, NULL) == NULL) {
748 src_storage = NULL;
749 goto done;
750 }
751
752 VERIFY(sifp != NULL);
753
754 if (inp && inp_restricted_send(inp, sifp)) {
755 src_storage = NULL;
756 *errorp = EHOSTUNREACH;
757 ifnet_release(sifp);
758 goto done;
759 } else {
760 ifnet_release(sifp);
761 }
762
763done:
764 if (ifpp != NULL) {
765 /* if ifp is non-NULL, refcnt held in in6_selectif() */
766 *ifpp = ifp;
767 } else if (ifp != NULL) {
768 ifnet_release(ifp);
769 }
770 return (src_storage);
771}
772
773/*
774 * Given a source IPv6 address (and route, if available), determine the best
775 * interface to send the packet from. Checking for (and updating) the
776 * ROF_SRCIF_SELECTED flag in the pcb-supplied route placeholder is done
777 * without any locks, based on the assumption that in the event this is
778 * called from ip6_output(), the output operation is single-threaded per-pcb,
779 * i.e. for any given pcb there can only be one thread performing output at
780 * the IPv6 layer.
781 *
782 * This routine is analogous to in_selectsrcif() for IPv4. Regardless of
783 * error, it will return an ifp with a reference held if the caller provides
784 * a non-NULL retifp. The caller is responsible for checking if the
785 * returned ifp is valid and release its reference at all times.
786 *
787 * clone - meaningful only for bsdi and freebsd
788 */
789static int
790selectroute(struct sockaddr_in6 *srcsock, struct sockaddr_in6 *dstsock,
791 struct ip6_pktopts *opts, struct ip6_moptions *mopts,
792 struct in6_ifaddr **retsrcia, struct route_in6 *ro,
793 struct ifnet **retifp, struct rtentry **retrt, int clone,
794 int norouteok, struct ip6_out_args *ip6oa)
795{
796 int error = 0;
797 struct ifnet *ifp = NULL, *ifp0 = NULL;
798 struct route_in6 *route = NULL;
799 struct sockaddr_in6 *sin6_next;
800 struct in6_pktinfo *pi = NULL;
801 struct in6_addr *dst = &dstsock->sin6_addr;
802 struct ifaddr *ifa = NULL;
803 char s_src[MAX_IPv6_STR_LEN], s_dst[MAX_IPv6_STR_LEN];
804 boolean_t select_srcif, proxied_ifa = FALSE, local_dst = FALSE;
805 unsigned int ifscope = ((ip6oa != NULL) ?
806 ip6oa->ip6oa_boundif : IFSCOPE_NONE);
807
808 if (retifp != NULL)
809 *retifp = NULL;
810
811 if (retrt != NULL)
812 *retrt = NULL;
813
814 if (ip6_select_srcif_debug) {
815 struct in6_addr src;
816 src = (srcsock != NULL) ? srcsock->sin6_addr : in6addr_any;
817 (void) inet_ntop(AF_INET6, &src, s_src, sizeof (s_src));
818 (void) inet_ntop(AF_INET6, dst, s_dst, sizeof (s_dst));
819 }
820
821 /*
822 * If the destination address is UNSPECIFIED addr, bail out.
823 */
824 if (IN6_IS_ADDR_UNSPECIFIED(dst)) {
825 error = EHOSTUNREACH;
826 goto done;
827 }
828
829 /*
830 * Perform source interface selection only if Scoped Routing
831 * is enabled and a source address that isn't unspecified.
832 */
833 select_srcif = (srcsock != NULL &&
834 !IN6_IS_ADDR_UNSPECIFIED(&srcsock->sin6_addr));
835
836 if (ip6_select_srcif_debug) {
837 printf("%s src %s dst %s ifscope %d select_srcif %d\n",
838 __func__, s_src, s_dst, ifscope, select_srcif);
839 }
840
841 /* If the caller specified the outgoing interface explicitly, use it */
842 if (opts != NULL && (pi = opts->ip6po_pktinfo) != NULL &&
843 pi->ipi6_ifindex != 0) {
844 /*
845 * If IPV6_PKTINFO takes precedence over IPV6_BOUND_IF.
846 */
847 ifscope = pi->ipi6_ifindex;
848 ifnet_head_lock_shared();
849 /* ifp may be NULL if detached or out of range */
850 ifp = ifp0 =
851 ((ifscope <= if_index) ? ifindex2ifnet[ifscope] : NULL);
852 ifnet_head_done();
853 if (norouteok || retrt == NULL || IN6_IS_ADDR_MULTICAST(dst)) {
854 /*
855 * We do not have to check or get the route for
856 * multicast. If the caller didn't ask/care for
857 * the route and we have no interface to use,
858 * it's an error.
859 */
860 if (ifp == NULL)
861 error = EHOSTUNREACH;
862 goto done;
863 } else {
864 goto getsrcif;
865 }
866 }
867
868 /*
869 * If the destination address is a multicast address and the outgoing
870 * interface for the address is specified by the caller, use it.
871 */
872 if (IN6_IS_ADDR_MULTICAST(dst) && mopts != NULL) {
873 IM6O_LOCK(mopts);
874 if ((ifp = ifp0 = mopts->im6o_multicast_ifp) != NULL) {
875 IM6O_UNLOCK(mopts);
876 goto done; /* we do not need a route for multicast. */
877 }
878 IM6O_UNLOCK(mopts);
879 }
880
881getsrcif:
882 /*
883 * If the outgoing interface was not set via IPV6_BOUND_IF or
884 * IPV6_PKTINFO, use the scope ID in the destination address.
885 */
886 if (ifscope == IFSCOPE_NONE)
887 ifscope = dstsock->sin6_scope_id;
888
889 /*
890 * Perform source interface selection; the source IPv6 address
891 * must belong to one of the addresses of the interface used
892 * by the route. For performance reasons, do this only if
893 * there is no route, or if the routing table has changed,
894 * or if we haven't done source interface selection on this
895 * route (for this PCB instance) before.
896 */
897 if (!select_srcif) {
898 goto getroute;
899 } else if (!ROUTE_UNUSABLE(ro) && ro->ro_srcia != NULL &&
900 (ro->ro_flags & ROF_SRCIF_SELECTED)) {
901 if (ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK)
902 local_dst = TRUE;
903 ifa = ro->ro_srcia;
904 IFA_ADDREF(ifa); /* for caller */
905 goto getroute;
906 }
907
908 /*
909 * Given the source IPv6 address, find a suitable source interface
910 * to use for transmission; if a scope ID has been specified,
911 * optimize the search by looking at the addresses only for that
912 * interface. This is still suboptimal, however, as we need to
913 * traverse the per-interface list.
914 */
915 if (ifscope != IFSCOPE_NONE || (ro != NULL && ro->ro_rt != NULL)) {
916 unsigned int scope = ifscope;
917 struct ifnet *rt_ifp;
918
919 rt_ifp = (ro->ro_rt != NULL) ? ro->ro_rt->rt_ifp : NULL;
920
921 /*
922 * If no scope is specified and the route is stale (pointing
923 * to a defunct interface) use the current primary interface;
924 * this happens when switching between interfaces configured
925 * with the same IPv6 address. Otherwise pick up the scope
926 * information from the route; the ULP may have looked up a
927 * correct route and we just need to verify it here and mark
928 * it with the ROF_SRCIF_SELECTED flag below.
929 */
930 if (scope == IFSCOPE_NONE) {
931 scope = rt_ifp->if_index;
932 if (scope != get_primary_ifscope(AF_INET6) &&
933 ROUTE_UNUSABLE(ro))
934 scope = get_primary_ifscope(AF_INET6);
935 }
936
937 ifa = (struct ifaddr *)
938 ifa_foraddr6_scoped(&srcsock->sin6_addr, scope);
939
940 /*
941 * If we are forwarding and proxying prefix(es), see if the
942 * source address is one of ours and is a proxied address;
943 * if so, use it.
944 */
945 if (ifa == NULL && ip6_forwarding && nd6_prproxy) {
946 ifa = (struct ifaddr *)
947 ifa_foraddr6(&srcsock->sin6_addr);
948 if (ifa != NULL && !(proxied_ifa =
949 nd6_prproxy_ifaddr((struct in6_ifaddr *)ifa))) {
950 IFA_REMREF(ifa);
951 ifa = NULL;
952 }
953 }
954
955 if (ip6_select_srcif_debug && ifa != NULL) {
956 if (ro->ro_rt != NULL) {
957 printf("%s %s->%s ifscope %d->%d ifa_if %s "
958 "ro_if %s\n",
959 __func__,
960 s_src, s_dst, ifscope,
961 scope, if_name(ifa->ifa_ifp),
962 if_name(rt_ifp));
963 } else {
964 printf("%s %s->%s ifscope %d->%d ifa_if %s\n",
965 __func__,
966 s_src, s_dst, ifscope, scope,
967 if_name(ifa->ifa_ifp));
968 }
969 }
970 }
971
972 /*
973 * Slow path; search for an interface having the corresponding source
974 * IPv6 address if the scope was not specified by the caller, and:
975 *
976 * 1) There currently isn't any route, or,
977 * 2) The interface used by the route does not own that source
978 * IPv6 address; in this case, the route will get blown away
979 * and we'll do a more specific scoped search using the newly
980 * found interface.
981 */
982 if (ifa == NULL && ifscope == IFSCOPE_NONE) {
983 struct ifaddr *ifadst;
984
985 /* Check if the destination address is one of ours */
986 ifadst = (struct ifaddr *)ifa_foraddr6(&dstsock->sin6_addr);
987 if (ifadst != NULL) {
988 local_dst = TRUE;
989 IFA_REMREF(ifadst);
990 }
991
992 ifa = (struct ifaddr *)ifa_foraddr6(&srcsock->sin6_addr);
993
994 if (ip6_select_srcif_debug && ifa != NULL) {
995 printf("%s %s->%s ifscope %d ifa_if %s\n",
996 __func__,
997 s_src, s_dst, ifscope, if_name(ifa->ifa_ifp));
998 } else if (ip6_select_srcif_debug) {
999 printf("%s %s->%s ifscope %d ifa_if NULL\n",
1000 __func__,
1001 s_src, s_dst, ifscope);
1002 }
1003 }
1004
1005getroute:
1006 if (ifa != NULL && !proxied_ifa && !local_dst)
1007 ifscope = ifa->ifa_ifp->if_index;
1008
1009 /*
1010 * If the next hop address for the packet is specified by the caller,
1011 * use it as the gateway.
1012 */
1013 if (opts != NULL && opts->ip6po_nexthop != NULL) {
1014 struct route_in6 *ron;
1015
1016 sin6_next = satosin6(opts->ip6po_nexthop);
1017
1018 /* at this moment, we only support AF_INET6 next hops */
1019 if (sin6_next->sin6_family != AF_INET6) {
1020 error = EAFNOSUPPORT; /* or should we proceed? */
1021 goto done;
1022 }
1023
1024 /*
1025 * If the next hop is an IPv6 address, then the node identified
1026 * by that address must be a neighbor of the sending host.
1027 */
1028 ron = &opts->ip6po_nextroute;
1029 if (ron->ro_rt != NULL)
1030 RT_LOCK(ron->ro_rt);
1031 if (ROUTE_UNUSABLE(ron) || (ron->ro_rt != NULL &&
1032 (!(ron->ro_rt->rt_flags & RTF_LLINFO) ||
1033 (select_srcif && (ifa == NULL ||
1034 (ifa->ifa_ifp != ron->ro_rt->rt_ifp && !proxied_ifa))))) ||
1035 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr,
1036 &sin6_next->sin6_addr)) {
1037 if (ron->ro_rt != NULL)
1038 RT_UNLOCK(ron->ro_rt);
1039
1040 ROUTE_RELEASE(ron);
1041 *satosin6(&ron->ro_dst) = *sin6_next;
1042 }
1043 if (ron->ro_rt == NULL) {
1044 rtalloc_scoped((struct route *)ron, ifscope);
1045 if (ron->ro_rt != NULL)
1046 RT_LOCK(ron->ro_rt);
1047 if (ROUTE_UNUSABLE(ron) ||
1048 !(ron->ro_rt->rt_flags & RTF_LLINFO) ||
1049 !IN6_ARE_ADDR_EQUAL(&satosin6(rt_key(ron->ro_rt))->
1050 sin6_addr, &sin6_next->sin6_addr)) {
1051 if (ron->ro_rt != NULL)
1052 RT_UNLOCK(ron->ro_rt);
1053
1054 ROUTE_RELEASE(ron);
1055 error = EHOSTUNREACH;
1056 goto done;
1057 }
1058 }
1059 route = ron;
1060 ifp = ifp0 = ron->ro_rt->rt_ifp;
1061
1062 /*
1063 * When cloning is required, try to allocate a route to the
1064 * destination so that the caller can store path MTU
1065 * information.
1066 */
1067 if (!clone) {
1068 if (select_srcif) {
1069 /* Keep the route locked */
1070 goto validateroute;
1071 }
1072 RT_UNLOCK(ron->ro_rt);
1073 goto done;
1074 }
1075 RT_UNLOCK(ron->ro_rt);
1076 }
1077
1078 /*
1079 * Use a cached route if it exists and is valid, else try to allocate
1080 * a new one. Note that we should check the address family of the
1081 * cached destination, in case of sharing the cache with IPv4.
1082 */
1083 if (ro == NULL)
1084 goto done;
1085 if (ro->ro_rt != NULL)
1086 RT_LOCK_SPIN(ro->ro_rt);
1087 if (ROUTE_UNUSABLE(ro) || (ro->ro_rt != NULL &&
1088 (satosin6(&ro->ro_dst)->sin6_family != AF_INET6 ||
1089 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, dst) ||
1090 (select_srcif && (ifa == NULL ||
1091 (ifa->ifa_ifp != ro->ro_rt->rt_ifp && !proxied_ifa)))))) {
1092 if (ro->ro_rt != NULL)
1093 RT_UNLOCK(ro->ro_rt);
1094
1095 ROUTE_RELEASE(ro);
1096 }
1097 if (ro->ro_rt == NULL) {
1098 struct sockaddr_in6 *sa6;
1099
1100 if (ro->ro_rt != NULL)
1101 RT_UNLOCK(ro->ro_rt);
1102 /* No route yet, so try to acquire one */
1103 bzero(&ro->ro_dst, sizeof (struct sockaddr_in6));
1104 sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
1105 sa6->sin6_family = AF_INET6;
1106 sa6->sin6_len = sizeof (struct sockaddr_in6);
1107 sa6->sin6_addr = *dst;
1108 if (IN6_IS_ADDR_MULTICAST(dst)) {
1109 ro->ro_rt = rtalloc1_scoped(
1110 &((struct route *)ro)->ro_dst, 0, 0, ifscope);
1111 } else {
1112 rtalloc_scoped((struct route *)ro, ifscope);
1113 }
1114 if (ro->ro_rt != NULL)
1115 RT_LOCK_SPIN(ro->ro_rt);
1116 }
1117
1118 /*
1119 * Do not care about the result if we have the nexthop
1120 * explicitly specified (in case we're asked to clone.)
1121 */
1122 if (opts != NULL && opts->ip6po_nexthop != NULL) {
1123 if (ro->ro_rt != NULL)
1124 RT_UNLOCK(ro->ro_rt);
1125 goto done;
1126 }
1127
1128 if (ro->ro_rt != NULL) {
1129 RT_LOCK_ASSERT_HELD(ro->ro_rt);
1130 ifp = ifp0 = ro->ro_rt->rt_ifp;
1131 } else {
1132 error = EHOSTUNREACH;
1133 }
1134 route = ro;
1135
1136validateroute:
1137 if (select_srcif) {
1138 boolean_t has_route = (route != NULL && route->ro_rt != NULL);
1139 boolean_t srcif_selected = FALSE;
1140
1141 if (has_route)
1142 RT_LOCK_ASSERT_HELD(route->ro_rt);
1143 /*
1144 * If there is a non-loopback route with the wrong interface,
1145 * or if there is no interface configured with such an address,
1146 * blow it away. Except for local/loopback, we look for one
1147 * with a matching interface scope/index.
1148 */
1149 if (has_route && (ifa == NULL ||
1150 (ifa->ifa_ifp != ifp && ifp != lo_ifp) ||
1151 !(route->ro_rt->rt_flags & RTF_UP))) {
1152 /*
1153 * If the destination address belongs to a proxied
1154 * prefix, relax the requirement and allow the packet
1155 * to come out of the proxy interface with the source
1156 * address of the real interface.
1157 */
1158 if (ifa != NULL && proxied_ifa &&
1159 (route->ro_rt->rt_flags & (RTF_UP|RTF_PROXY)) ==
1160 (RTF_UP|RTF_PROXY)) {
1161 srcif_selected = TRUE;
1162 } else {
1163 if (ip6_select_srcif_debug) {
1164 if (ifa != NULL) {
1165 printf("%s->%s ifscope %d "
1166 "ro_if %s != ifa_if %s "
1167 "(cached route cleared)\n",
1168 s_src, s_dst,
1169 ifscope, if_name(ifp),
1170 if_name(ifa->ifa_ifp));
1171 } else {
1172 printf("%s->%s ifscope %d "
1173 "ro_if %s (no ifa_if "
1174 "found)\n", s_src, s_dst,
1175 ifscope, if_name(ifp));
1176 }
1177 }
1178 RT_UNLOCK(route->ro_rt);
1179 ROUTE_RELEASE(route);
1180 error = EHOSTUNREACH;
1181 /* Undo the settings done above */
1182 route = NULL;
1183 ifp = NULL; /* ditch ifp; keep ifp0 */
1184 has_route = FALSE;
1185 }
1186 } else if (has_route) {
1187 srcif_selected = TRUE;
1188 }
1189
1190 if (srcif_selected) {
1191 VERIFY(has_route);
1192 if (ifa != route->ro_srcia ||
1193 !(route->ro_flags & ROF_SRCIF_SELECTED)) {
1194 RT_CONVERT_LOCK(route->ro_rt);
1195 if (ifa != NULL)
1196 IFA_ADDREF(ifa); /* for route_in6 */
1197 if (route->ro_srcia != NULL)
1198 IFA_REMREF(route->ro_srcia);
1199 route->ro_srcia = ifa;
1200 route->ro_flags |= ROF_SRCIF_SELECTED;
1201 RT_GENID_SYNC(route->ro_rt);
1202 }
1203 RT_UNLOCK(route->ro_rt);
1204 }
1205 } else {
1206 if (ro->ro_rt != NULL)
1207 RT_UNLOCK(ro->ro_rt);
1208 if (ifp != NULL && opts != NULL &&
1209 opts->ip6po_pktinfo != NULL &&
1210 opts->ip6po_pktinfo->ipi6_ifindex != 0) {
1211 /*
1212 * Check if the outgoing interface conflicts with the
1213 * interface specified by ipi6_ifindex (if specified).
1214 * Note that loopback interface is always okay.
1215 * (this may happen when we are sending a packet to
1216 * one of our own addresses.)
1217 */
1218 if (!(ifp->if_flags & IFF_LOOPBACK) && ifp->if_index !=
1219 opts->ip6po_pktinfo->ipi6_ifindex) {
1220 error = EHOSTUNREACH;
1221 goto done;
1222 }
1223 }
1224 }
1225
1226done:
1227 /*
1228 * Check for interface restrictions.
1229 */
1230#define CHECK_RESTRICTIONS(_ip6oa, _ifp) \
1231 ((((_ip6oa)->ip6oa_flags & IP6OAF_NO_CELLULAR) && \
1232 IFNET_IS_CELLULAR(_ifp)) || \
1233 (((_ip6oa)->ip6oa_flags & IP6OAF_NO_EXPENSIVE) && \
1234 IFNET_IS_EXPENSIVE(_ifp)) || \
1235 (!((_ip6oa)->ip6oa_flags & IP6OAF_INTCOPROC_ALLOWED) && \
1236 IFNET_IS_INTCOPROC(_ifp)) || \
1237 (!((_ip6oa)->ip6oa_flags & IP6OAF_AWDL_UNRESTRICTED) && \
1238 IFNET_IS_AWDL_RESTRICTED(_ifp)))
1239
1240 if (error == 0 && ip6oa != NULL &&
1241 ((ifp && CHECK_RESTRICTIONS(ip6oa, ifp)) ||
1242 (route && route->ro_rt &&
1243 CHECK_RESTRICTIONS(ip6oa, route->ro_rt->rt_ifp)))) {
1244 if (route != NULL && route->ro_rt != NULL) {
1245 ROUTE_RELEASE(route);
1246 route = NULL;
1247 }
1248 ifp = NULL; /* ditch ifp; keep ifp0 */
1249 error = EHOSTUNREACH;
1250 ip6oa->ip6oa_retflags |= IP6OARF_IFDENIED;
1251 }
1252#undef CHECK_RESTRICTIONS
1253
1254 /*
1255 * If the interface is disabled for IPv6, then ENETDOWN error.
1256 */
1257 if (error == 0 &&
1258 ifp != NULL && (ifp->if_eflags & IFEF_IPV6_DISABLED)) {
1259 error = ENETDOWN;
1260 }
1261
1262 if (ifp == NULL && (route == NULL || route->ro_rt == NULL)) {
1263 /*
1264 * This can happen if the caller did not pass a cached route
1265 * nor any other hints. We treat this case an error.
1266 */
1267 error = EHOSTUNREACH;
1268 }
1269 if (error == EHOSTUNREACH || error == ENETDOWN)
1270 ip6stat.ip6s_noroute++;
1271
1272 /*
1273 * We'll return ifp regardless of error, so pick it up from ifp0
1274 * in case it was nullified above. Caller is responsible for
1275 * releasing the ifp if it is non-NULL.
1276 */
1277 ifp = ifp0;
1278 if (retifp != NULL) {
1279 if (ifp != NULL)
1280 ifnet_reference(ifp); /* for caller */
1281 *retifp = ifp;
1282 }
1283
1284 if (retsrcia != NULL) {
1285 if (ifa != NULL)
1286 IFA_ADDREF(ifa); /* for caller */
1287 *retsrcia = (struct in6_ifaddr *)ifa;
1288 }
1289
1290 if (error == 0) {
1291 if (retrt != NULL && route != NULL)
1292 *retrt = route->ro_rt; /* ro_rt may be NULL */
1293 }
1294 if (ip6_select_srcif_debug) {
1295 printf("%s %s->%s ifscope %d ifa_if %s ro_if %s (error=%d)\n",
1296 __func__,
1297 s_src, s_dst, ifscope,
1298 (ifa != NULL) ? if_name(ifa->ifa_ifp) : "NONE",
1299 (ifp != NULL) ? if_name(ifp) : "NONE", error);
1300 }
1301
1302 if (ifa != NULL)
1303 IFA_REMREF(ifa);
1304
1305 return (error);
1306}
1307
1308/*
1309 * Regardless of error, it will return an ifp with a reference held if the
1310 * caller provides a non-NULL retifp. The caller is responsible for checking
1311 * if the returned ifp is valid and release its reference at all times.
1312 */
1313int
1314in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
1315 struct ip6_moptions *mopts, struct route_in6 *ro,
1316 struct ip6_out_args *ip6oa, struct ifnet **retifp)
1317{
1318 int err = 0;
1319 struct route_in6 sro;
1320 struct rtentry *rt = NULL;
1321
1322 if (ro == NULL) {
1323 bzero(&sro, sizeof (sro));
1324 ro = &sro;
1325 }
1326
1327 if ((err = selectroute(NULL, dstsock, opts, mopts, NULL, ro, retifp,
1328 &rt, 0, 1, ip6oa)) != 0)
1329 goto done;
1330
1331 /*
1332 * do not use a rejected or black hole route.
1333 * XXX: this check should be done in the L2 output routine.
1334 * However, if we skipped this check here, we'd see the following
1335 * scenario:
1336 * - install a rejected route for a scoped address prefix
1337 * (like fe80::/10)
1338 * - send a packet to a destination that matches the scoped prefix,
1339 * with ambiguity about the scope zone.
1340 * - pick the outgoing interface from the route, and disambiguate the
1341 * scope zone with the interface.
1342 * - ip6_output() would try to get another route with the "new"
1343 * destination, which may be valid.
1344 * - we'd see no error on output.
1345 * Although this may not be very harmful, it should still be confusing.
1346 * We thus reject the case here.
1347 */
1348 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
1349 err = ((rt->rt_flags & RTF_HOST) ? EHOSTUNREACH : ENETUNREACH);
1350 goto done;
1351 }
1352
1353 /*
1354 * Adjust the "outgoing" interface. If we're going to loop the packet
1355 * back to ourselves, the ifp would be the loopback interface.
1356 * However, we'd rather know the interface associated to the
1357 * destination address (which should probably be one of our own
1358 * addresses.)
1359 */
1360 if (rt != NULL && rt->rt_ifa != NULL && rt->rt_ifa->ifa_ifp != NULL &&
1361 retifp != NULL) {
1362 ifnet_reference(rt->rt_ifa->ifa_ifp);
1363 if (*retifp != NULL)
1364 ifnet_release(*retifp);
1365 *retifp = rt->rt_ifa->ifa_ifp;
1366 }
1367
1368done:
1369 if (ro == &sro) {
1370 VERIFY(rt == NULL || rt == ro->ro_rt);
1371 ROUTE_RELEASE(ro);
1372 }
1373
1374 /*
1375 * retifp might point to a valid ifp with a reference held;
1376 * caller is responsible for releasing it if non-NULL.
1377 */
1378 return (err);
1379}
1380
1381/*
1382 * Regardless of error, it will return an ifp with a reference held if the
1383 * caller provides a non-NULL retifp. The caller is responsible for checking
1384 * if the returned ifp is valid and release its reference at all times.
1385 *
1386 * clone - meaningful only for bsdi and freebsd
1387 */
1388int
1389in6_selectroute(struct sockaddr_in6 *srcsock, struct sockaddr_in6 *dstsock,
1390 struct ip6_pktopts *opts, struct ip6_moptions *mopts,
1391 struct in6_ifaddr **retsrcia, struct route_in6 *ro, struct ifnet **retifp,
1392 struct rtentry **retrt, int clone, struct ip6_out_args *ip6oa)
1393{
1394
1395 return (selectroute(srcsock, dstsock, opts, mopts, retsrcia, ro, retifp,
1396 retrt, clone, 0, ip6oa));
1397}
1398
1399/*
1400 * Default hop limit selection. The precedence is as follows:
1401 * 1. Hoplimit value specified via ioctl.
1402 * 2. (If the outgoing interface is detected) the current
1403 * hop limit of the interface specified by router advertisement.
1404 * 3. The system default hoplimit.
1405 */
1406int
1407in6_selecthlim(struct in6pcb *in6p, struct ifnet *ifp)
1408{
1409 if (in6p && in6p->in6p_hops >= 0) {
1410 return (in6p->in6p_hops);
1411 } else if (NULL != ifp) {
1412 u_int8_t chlim;
1413 struct nd_ifinfo *ndi = ND_IFINFO(ifp);
1414 if (ndi && ndi->initialized) {
1415 /* access chlim without lock, for performance */
1416 chlim = ndi->chlim;
1417 } else {
1418 chlim = ip6_defhlim;
1419 }
1420 return (chlim);
1421 }
1422
1423 return (ip6_defhlim);
1424}
1425
1426/*
1427 * XXX: this is borrowed from in6_pcbbind(). If possible, we should
1428 * share this function by all *bsd*...
1429 */
1430int
1431in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct proc *p,
1432 int locked)
1433{
1434 struct socket *so = inp->inp_socket;
1435 u_int16_t lport = 0, first, last, *lastport;
1436 int count, error = 0, wild = 0;
1437 boolean_t counting_down;
1438 bool found;
1439 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1440 kauth_cred_t cred;
1441#pragma unused(laddr)
1442 if (!locked) { /* Make sure we don't run into a deadlock: 4052373 */
1443 if (!lck_rw_try_lock_exclusive(pcbinfo->ipi_lock)) {
1444 socket_unlock(inp->inp_socket, 0);
1445 lck_rw_lock_exclusive(pcbinfo->ipi_lock);
1446 socket_lock(inp->inp_socket, 0);
1447 }
1448
1449 /*
1450 * Check if a local port was assigned to the inp while
1451 * this thread was waiting for the pcbinfo lock
1452 */
1453 if (inp->inp_lport != 0) {
1454 VERIFY(inp->inp_flags2 & INP2_INHASHLIST);
1455 lck_rw_done(pcbinfo->ipi_lock);
1456
1457 /*
1458 * It is not an error if another thread allocated
1459 * a port
1460 */
1461 return (0);
1462 }
1463 }
1464
1465 /* XXX: this is redundant when called from in6_pcbbind */
1466 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
1467 wild = INPLOOKUP_WILDCARD;
1468
1469 if (inp->inp_flags & INP_HIGHPORT) {
1470 first = ipport_hifirstauto; /* sysctl */
1471 last = ipport_hilastauto;
1472 lastport = &pcbinfo->ipi_lasthi;
1473 } else if (inp->inp_flags & INP_LOWPORT) {
1474 cred = kauth_cred_proc_ref(p);
1475 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0);
1476 kauth_cred_unref(&cred);
1477 if (error != 0) {
1478 if (!locked)
1479 lck_rw_done(pcbinfo->ipi_lock);
1480 return (error);
1481 }
1482 first = ipport_lowfirstauto; /* 1023 */
1483 last = ipport_lowlastauto; /* 600 */
1484 lastport = &pcbinfo->ipi_lastlow;
1485 } else {
1486 first = ipport_firstauto; /* sysctl */
1487 last = ipport_lastauto;
1488 lastport = &pcbinfo->ipi_lastport;
1489 }
1490 /*
1491 * Simple check to ensure all ports are not used up causing
1492 * a deadlock here.
1493 */
1494 found = false;
1495 if (first > last) {
1496 /* counting down */
1497 count = first - last;
1498 counting_down = TRUE;
1499 } else {
1500 /* counting up */
1501 count = last - first;
1502 counting_down = FALSE;
1503 }
1504 do {
1505 if (count-- < 0) { /* completely used? */
1506 /*
1507 * Undo any address bind that may have
1508 * occurred above.
1509 */
1510 inp->in6p_laddr = in6addr_any;
1511 inp->in6p_last_outifp = NULL;
1512 if (!locked)
1513 lck_rw_done(pcbinfo->ipi_lock);
1514 return (EAGAIN);
1515 }
1516 if (counting_down) {
1517 --*lastport;
1518 if (*lastport > first || *lastport < last) {
1519 *lastport = first;
1520 }
1521 } else {
1522 ++*lastport;
1523 if (*lastport < first || *lastport > last)
1524 *lastport = first;
1525 }
1526 lport = htons(*lastport);
1527 found = (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr,
1528 lport, wild) == NULL);
1529 } while (!found);
1530
1531 inp->inp_lport = lport;
1532 inp->inp_flags |= INP_ANONPORT;
1533
1534 if (in_pcbinshash(inp, 1) != 0) {
1535 inp->in6p_laddr = in6addr_any;
1536 inp->in6p_last_outifp = NULL;
1537
1538 inp->inp_lport = 0;
1539 inp->inp_flags &= ~INP_ANONPORT;
1540 if (!locked)
1541 lck_rw_done(pcbinfo->ipi_lock);
1542 return (EAGAIN);
1543 }
1544
1545 if (!locked)
1546 lck_rw_done(pcbinfo->ipi_lock);
1547 return (0);
1548}
1549
1550/*
1551 * The followings are implementation of the policy table using a
1552 * simple tail queue.
1553 * XXX such details should be hidden.
1554 * XXX implementation using binary tree should be more efficient.
1555 */
1556struct addrsel_policyent {
1557 TAILQ_ENTRY(addrsel_policyent) ape_entry;
1558 struct in6_addrpolicy ape_policy;
1559};
1560
1561TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1562
1563struct addrsel_policyhead addrsel_policytab;
1564
1565static void
1566init_policy_queue(void)
1567{
1568 TAILQ_INIT(&addrsel_policytab);
1569}
1570
1571void
1572addrsel_policy_init(void)
1573{
1574 /*
1575 * Default address selection policy based on RFC 6724.
1576 */
1577 static const struct in6_addrpolicy defaddrsel[] = {
1578 /* Loopback -- prefix=::1/128, precedence=50, label=0 */
1579 {
1580 .addr = {
1581 .sin6_family = AF_INET6,
1582 .sin6_addr = IN6ADDR_LOOPBACK_INIT,
1583 .sin6_len = sizeof (struct sockaddr_in6)
1584 },
1585 .addrmask = {
1586 .sin6_family = AF_INET6,
1587 .sin6_addr = IN6MASK128,
1588 .sin6_len = sizeof (struct sockaddr_in6)
1589 },
1590 .preced = 50,
1591 .label = 0
1592 },
1593
1594 /* Unspecified -- prefix=::/0, precedence=40, label=1 */
1595 {
1596 .addr = {
1597 .sin6_family = AF_INET6,
1598 .sin6_addr = IN6ADDR_ANY_INIT,
1599 .sin6_len = sizeof (struct sockaddr_in6)
1600 },
1601 .addrmask = {
1602 .sin6_family = AF_INET6,
1603 .sin6_addr = IN6MASK0,
1604 .sin6_len = sizeof (struct sockaddr_in6)
1605 },
1606 .preced = 40,
1607 .label = 1
1608 },
1609
1610 /* IPv4 Mapped -- prefix=::ffff:0:0/96, precedence=35, label=4 */
1611 {
1612 .addr = {
1613 .sin6_family = AF_INET6,
1614 .sin6_addr = IN6ADDR_V4MAPPED_INIT,
1615 .sin6_len = sizeof (struct sockaddr_in6)
1616 },
1617 .addrmask = {
1618 .sin6_family = AF_INET6,
1619 .sin6_addr = IN6MASK96,
1620 .sin6_len = sizeof (struct sockaddr_in6)
1621 },
1622 .preced = 35,
1623 .label = 4
1624 },
1625
1626 /* 6to4 -- prefix=2002::/16, precedence=30, label=2 */
1627 {
1628 .addr = {
1629 .sin6_family = AF_INET6,
1630 .sin6_addr = {{{ 0x20, 0x02 }}},
1631 .sin6_len = sizeof (struct sockaddr_in6)
1632 },
1633 .addrmask = {
1634 .sin6_family = AF_INET6,
1635 .sin6_addr = IN6MASK16,
1636 .sin6_len = sizeof (struct sockaddr_in6)
1637 },
1638 .preced = 30,
1639 .label = 2
1640 },
1641
1642 /* Teredo -- prefix=2001::/32, precedence=5, label=5 */
1643 {
1644 .addr = {
1645 .sin6_family = AF_INET6,
1646 .sin6_addr = {{{ 0x20, 0x01 }}},
1647 .sin6_len = sizeof (struct sockaddr_in6)
1648 },
1649 .addrmask = {
1650 .sin6_family = AF_INET6,
1651 .sin6_addr = IN6MASK32,
1652 .sin6_len = sizeof (struct sockaddr_in6)
1653 },
1654 .preced = 5,
1655 .label = 5
1656 },
1657
1658 /* Unique Local (ULA) -- prefix=fc00::/7, precedence=3, label=13 */
1659 {
1660 .addr = {
1661 .sin6_family = AF_INET6,
1662 .sin6_addr = {{{ 0xfc }}},
1663 .sin6_len = sizeof (struct sockaddr_in6)
1664 },
1665 .addrmask = {
1666 .sin6_family = AF_INET6,
1667 .sin6_addr = IN6MASK7,
1668 .sin6_len = sizeof (struct sockaddr_in6)
1669 },
1670 .preced = 3,
1671 .label = 13
1672 },
1673
1674 /* IPv4 Compatible -- prefix=::/96, precedence=1, label=3 */
1675 {
1676 .addr = {
1677 .sin6_family = AF_INET6,
1678 .sin6_addr = IN6ADDR_ANY_INIT,
1679 .sin6_len = sizeof (struct sockaddr_in6)
1680 },
1681 .addrmask = {
1682 .sin6_family = AF_INET6,
1683 .sin6_addr = IN6MASK96,
1684 .sin6_len = sizeof (struct sockaddr_in6)
1685 },
1686 .preced = 1,
1687 .label = 3
1688 },
1689
1690 /* Site-local (deprecated) -- prefix=fec0::/10, precedence=1, label=11 */
1691 {
1692 .addr = {
1693 .sin6_family = AF_INET6,
1694 .sin6_addr = {{{ 0xfe, 0xc0 }}},
1695 .sin6_len = sizeof (struct sockaddr_in6)
1696 },
1697 .addrmask = {
1698 .sin6_family = AF_INET6,
1699 .sin6_addr = IN6MASK16,
1700 .sin6_len = sizeof (struct sockaddr_in6)
1701 },
1702 .preced = 1,
1703 .label = 11
1704 },
1705
1706 /* 6bone (deprecated) -- prefix=3ffe::/16, precedence=1, label=12 */
1707 {
1708 .addr = {
1709 .sin6_family = AF_INET6,
1710 .sin6_addr = {{{ 0x3f, 0xfe }}},
1711 .sin6_len = sizeof (struct sockaddr_in6)
1712 },
1713 .addrmask = {
1714 .sin6_family = AF_INET6,
1715 .sin6_addr = IN6MASK16,
1716 .sin6_len = sizeof (struct sockaddr_in6)
1717 },
1718 .preced = 1,
1719 .label = 12
1720 },
1721 };
1722 int i;
1723
1724 init_policy_queue();
1725
1726 /* initialize the "last resort" policy */
1727 bzero(&defaultaddrpolicy, sizeof (defaultaddrpolicy));
1728 defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
1729
1730 for (i = 0; i < sizeof (defaddrsel) / sizeof (defaddrsel[0]); i++)
1731 add_addrsel_policyent(&defaddrsel[i]);
1732
1733}
1734
1735struct in6_addrpolicy *
1736in6_addrsel_lookup_policy(struct sockaddr_in6 *key)
1737{
1738 struct in6_addrpolicy *match = NULL;
1739
1740 ADDRSEL_LOCK();
1741 match = match_addrsel_policy(key);
1742
1743 if (match == NULL)
1744 match = &defaultaddrpolicy;
1745 else
1746 match->use++;
1747 ADDRSEL_UNLOCK();
1748
1749 return (match);
1750}
1751
1752static struct in6_addrpolicy *
1753match_addrsel_policy(struct sockaddr_in6 *key)
1754{
1755 struct addrsel_policyent *pent;
1756 struct in6_addrpolicy *bestpol = NULL, *pol;
1757 int matchlen, bestmatchlen = -1;
1758 u_char *mp, *ep, *k, *p, m;
1759
1760 TAILQ_FOREACH(pent, &addrsel_policytab, ape_entry) {
1761 matchlen = 0;
1762
1763 pol = &pent->ape_policy;
1764 mp = (u_char *)&pol->addrmask.sin6_addr;
1765 ep = mp + 16; /* XXX: scope field? */
1766 k = (u_char *)&key->sin6_addr;
1767 p = (u_char *)&pol->addr.sin6_addr;
1768 for (; mp < ep && *mp; mp++, k++, p++) {
1769 m = *mp;
1770 if ((*k & m) != *p)
1771 goto next; /* not match */
1772 if (m == 0xff) /* short cut for a typical case */
1773 matchlen += 8;
1774 else {
1775 while (m >= 0x80) {
1776 matchlen++;
1777 m <<= 1;
1778 }
1779 }
1780 }
1781
1782 /* matched. check if this is better than the current best. */
1783 if (bestpol == NULL ||
1784 matchlen > bestmatchlen) {
1785 bestpol = pol;
1786 bestmatchlen = matchlen;
1787 }
1788
1789 next:
1790 continue;
1791 }
1792
1793 return (bestpol);
1794}
1795
1796static int
1797add_addrsel_policyent(const struct in6_addrpolicy *newpolicy)
1798{
1799 struct addrsel_policyent *new, *pol;
1800
1801 MALLOC(new, struct addrsel_policyent *, sizeof (*new), M_IFADDR,
1802 M_WAITOK);
1803
1804 ADDRSEL_LOCK();
1805
1806 /* duplication check */
1807 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
1808 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1809 &pol->ape_policy.addr.sin6_addr) &&
1810 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1811 &pol->ape_policy.addrmask.sin6_addr)) {
1812 ADDRSEL_UNLOCK();
1813 FREE(new, M_IFADDR);
1814 return (EEXIST); /* or override it? */
1815 }
1816 }
1817
1818 bzero(new, sizeof (*new));
1819
1820 /* XXX: should validate entry */
1821 new->ape_policy = *newpolicy;
1822
1823 TAILQ_INSERT_TAIL(&addrsel_policytab, new, ape_entry);
1824 ADDRSEL_UNLOCK();
1825
1826 return (0);
1827}
1828#ifdef ENABLE_ADDRSEL
1829static int
1830delete_addrsel_policyent(const struct in6_addrpolicy *key)
1831{
1832 struct addrsel_policyent *pol;
1833
1834
1835 ADDRSEL_LOCK();
1836
1837 /* search for the entry in the table */
1838 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
1839 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1840 &pol->ape_policy.addr.sin6_addr) &&
1841 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1842 &pol->ape_policy.addrmask.sin6_addr)) {
1843 break;
1844 }
1845 }
1846 if (pol == NULL) {
1847 ADDRSEL_UNLOCK();
1848 return (ESRCH);
1849 }
1850
1851 TAILQ_REMOVE(&addrsel_policytab, pol, ape_entry);
1852 FREE(pol, M_IFADDR);
1853 pol = NULL;
1854 ADDRSEL_UNLOCK();
1855
1856 return (0);
1857}
1858#endif /* ENABLE_ADDRSEL */
1859
1860int
1861walk_addrsel_policy(int (*callback)(const struct in6_addrpolicy *, void *),
1862 void *w)
1863{
1864 struct addrsel_policyent *pol;
1865 int error = 0;
1866
1867 ADDRSEL_LOCK();
1868 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
1869 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1870 ADDRSEL_UNLOCK();
1871 return (error);
1872 }
1873 }
1874 ADDRSEL_UNLOCK();
1875 return (error);
1876}
1877/*
1878 * Subroutines to manage the address selection policy table via sysctl.
1879 */
1880struct walkarg {
1881 struct sysctl_req *w_req;
1882};
1883
1884
1885static int
1886dump_addrsel_policyent(const struct in6_addrpolicy *pol, void *arg)
1887{
1888 int error = 0;
1889 struct walkarg *w = arg;
1890
1891 error = SYSCTL_OUT(w->w_req, pol, sizeof (*pol));
1892
1893 return (error);
1894}
1895
1896static int
1897in6_src_sysctl SYSCTL_HANDLER_ARGS
1898{
1899#pragma unused(oidp, arg1, arg2)
1900struct walkarg w;
1901
1902 if (req->newptr)
1903 return (EPERM);
1904 bzero(&w, sizeof (w));
1905 w.w_req = req;
1906
1907 return (walk_addrsel_policy(dump_addrsel_policyent, &w));
1908}
1909
1910
1911SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
1912 CTLFLAG_RD | CTLFLAG_LOCKED, in6_src_sysctl, "");
1913int
1914in6_src_ioctl(u_long cmd, caddr_t data)
1915{
1916 int i;
1917 struct in6_addrpolicy ent0;
1918
1919 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
1920 return (EOPNOTSUPP); /* check for safety */
1921
1922 bcopy(data, &ent0, sizeof (ent0));
1923
1924 if (ent0.label == ADDR_LABEL_NOTAPP)
1925 return (EINVAL);
1926 /* check if the prefix mask is consecutive. */
1927 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
1928 return (EINVAL);
1929 /* clear trailing garbages (if any) of the prefix address. */
1930 for (i = 0; i < 4; i++) {
1931 ent0.addr.sin6_addr.s6_addr32[i] &=
1932 ent0.addrmask.sin6_addr.s6_addr32[i];
1933 }
1934 ent0.use = 0;
1935
1936 switch (cmd) {
1937 case SIOCAADDRCTL_POLICY:
1938#ifdef ENABLE_ADDRSEL
1939 return (add_addrsel_policyent(&ent0));
1940#else
1941 return (ENOTSUP);
1942#endif
1943 case SIOCDADDRCTL_POLICY:
1944#ifdef ENABLE_ADDRSEL
1945 return (delete_addrsel_policyent(&ent0));
1946#else
1947 return (ENOTSUP);
1948#endif
1949 }
1950
1951 return (0); /* XXX: compromise compilers */
1952}
1953
1954/*
1955 * generate kernel-internal form (scopeid embedded into s6_addr16[1]).
1956 * If the address scope of is link-local, embed the interface index in the
1957 * address. The routine determines our precedence
1958 * between advanced API scope/interface specification and basic API
1959 * specification.
1960 *
1961 * this function should be nuked in the future, when we get rid of
1962 * embedded scopeid thing.
1963 *
1964 * XXX actually, it is over-specification to return ifp against sin6_scope_id.
1965 * there can be multiple interfaces that belong to a particular scope zone
1966 * (in specification, we have 1:N mapping between a scope zone and interfaces).
1967 * we may want to change the function to return something other than ifp.
1968 */
1969int
1970in6_embedscope(struct in6_addr *in6, const struct sockaddr_in6 *sin6,
1971 struct in6pcb *in6p, struct ifnet **ifpp, struct ip6_pktopts *opt)
1972{
1973 struct ifnet *ifp = NULL;
1974 u_int32_t scopeid;
1975 struct ip6_pktopts *optp = NULL;
1976
1977 *in6 = sin6->sin6_addr;
1978 scopeid = sin6->sin6_scope_id;
1979 if (ifpp != NULL)
1980 *ifpp = NULL;
1981
1982 /*
1983 * don't try to read sin6->sin6_addr beyond here, since the caller may
1984 * ask us to overwrite existing sockaddr_in6
1985 */
1986
1987#ifdef ENABLE_DEFAULT_SCOPE
1988 if (scopeid == 0)
1989 scopeid = scope6_addr2default(in6);
1990#endif
1991
1992 if (IN6_IS_SCOPE_LINKLOCAL(in6) || IN6_IS_ADDR_MC_INTFACELOCAL(in6)) {
1993 struct in6_pktinfo *pi;
1994 struct ifnet *im6o_multicast_ifp = NULL;
1995
1996 if (in6p != NULL && IN6_IS_ADDR_MULTICAST(in6) &&
1997 in6p->in6p_moptions != NULL) {
1998 IM6O_LOCK(in6p->in6p_moptions);
1999 im6o_multicast_ifp =
2000 in6p->in6p_moptions->im6o_multicast_ifp;
2001 IM6O_UNLOCK(in6p->in6p_moptions);
2002 }
2003
2004 if (opt != NULL)
2005 optp = opt;
2006 else if (in6p != NULL)
2007 optp = in6p->in6p_outputopts;
2008 /*
2009 * KAME assumption: link id == interface id
2010 */
2011 if (in6p != NULL && optp != NULL &&
2012 (pi = optp->ip6po_pktinfo) != NULL &&
2013 pi->ipi6_ifindex != 0) {
2014 /* ifp is needed here if only we're returning it */
2015 if (ifpp != NULL) {
2016 ifnet_head_lock_shared();
2017 ifp = ifindex2ifnet[pi->ipi6_ifindex];
2018 ifnet_head_done();
2019 }
2020 in6->s6_addr16[1] = htons(pi->ipi6_ifindex);
2021 } else if (in6p != NULL && IN6_IS_ADDR_MULTICAST(in6) &&
2022 in6p->in6p_moptions != NULL && im6o_multicast_ifp != NULL) {
2023 ifp = im6o_multicast_ifp;
2024 in6->s6_addr16[1] = htons(ifp->if_index);
2025 } else if (scopeid != 0) {
2026 /*
2027 * Since scopeid is unsigned, we only have to check it
2028 * against if_index (ifnet_head_lock not needed since
2029 * if_index is an ever-increasing integer.)
2030 */
2031 if (if_index < scopeid)
2032 return (ENXIO); /* XXX EINVAL? */
2033
2034 /* ifp is needed here only if we're returning it */
2035 if (ifpp != NULL) {
2036 ifnet_head_lock_shared();
2037 ifp = ifindex2ifnet[scopeid];
2038 ifnet_head_done();
2039 }
2040 /* XXX assignment to 16bit from 32bit variable */
2041 in6->s6_addr16[1] = htons(scopeid & 0xffff);
2042 }
2043
2044 if (ifpp != NULL) {
2045 if (ifp != NULL)
2046 ifnet_reference(ifp); /* for caller */
2047 *ifpp = ifp;
2048 }
2049 }
2050
2051 return (0);
2052}
2053
2054/*
2055 * generate standard sockaddr_in6 from embedded form.
2056 * touches sin6_addr and sin6_scope_id only.
2057 *
2058 * this function should be nuked in the future, when we get rid of
2059 * embedded scopeid thing.
2060 */
2061int
2062in6_recoverscope(
2063 struct sockaddr_in6 *sin6,
2064 const struct in6_addr *in6,
2065 struct ifnet *ifp)
2066{
2067 u_int32_t scopeid;
2068
2069 sin6->sin6_addr = *in6;
2070
2071 /*
2072 * don't try to read *in6 beyond here, since the caller may
2073 * ask us to overwrite existing sockaddr_in6
2074 */
2075
2076 sin6->sin6_scope_id = 0;
2077 if (IN6_IS_SCOPE_LINKLOCAL(in6) || IN6_IS_ADDR_MC_INTFACELOCAL(in6)) {
2078 /*
2079 * KAME assumption: link id == interface id
2080 */
2081 scopeid = ntohs(sin6->sin6_addr.s6_addr16[1]);
2082 if (scopeid) {
2083 /*
2084 * sanity check
2085 *
2086 * Since scopeid is unsigned, we only have to check it
2087 * against if_index
2088 */
2089 if (if_index < scopeid)
2090 return (ENXIO);
2091 if (ifp && ifp->if_index != scopeid)
2092 return (ENXIO);
2093 sin6->sin6_addr.s6_addr16[1] = 0;
2094 sin6->sin6_scope_id = scopeid;
2095 }
2096 }
2097
2098 return (0);
2099}
2100