1 /* $FreeBSD: releng/5.4/sys/netinet6/in6_src.c 141090 2005-01-31 23:27:04Z imp $ */
2 /* $KAME: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $ */
3
4 /*-
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33 /*-
34 * Copyright (c) 1982, 1986, 1991, 1993
35 * The Regents of the University of California. All rights reserved.
36 *
37 * Redistribution and use in source and binary forms, with or without
38 * modification, are permitted provided that the following conditions
39 * are met:
40 * 1. Redistributions of source code must retain the above copyright
41 * notice, this list of conditions and the following disclaimer.
42 * 2. Redistributions in binary form must reproduce the above copyright
43 * notice, this list of conditions and the following disclaimer in the
44 * documentation and/or other materials provided with the distribution.
45 * 4. Neither the name of the University nor the names of its contributors
46 * may be used to endorse or promote products derived from this software
47 * without specific prior written permission.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 *
61 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94
62 */
63
64 #include "opt_inet.h"
65 #include "opt_inet6.h"
66
67 #include <sys/param.h>
68 #include <sys/systm.h>
69 #include <sys/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/protosw.h>
72 #include <sys/socket.h>
73 #include <sys/socketvar.h>
74 #include <sys/sockio.h>
75 #include <sys/sysctl.h>
76 #include <sys/errno.h>
77 #include <sys/time.h>
78 #include <sys/kernel.h>
79
80 #include <net/if.h>
81 #include <net/route.h>
82
83 #include <netinet/in.h>
84 #include <netinet/in_var.h>
85 #include <netinet/in_systm.h>
86 #include <netinet/ip.h>
87 #include <netinet/in_pcb.h>
88 #include <netinet6/in6_var.h>
89 #include <netinet/ip6.h>
90 #include <netinet6/in6_pcb.h>
91 #include <netinet6/ip6_var.h>
92 #include <netinet6/nd6.h>
93 #ifdef ENABLE_DEFAULT_SCOPE
94 #include <netinet6/scope6_var.h>
95 #endif
96
97 #include <net/net_osdep.h>
98
99 static struct mtx addrsel_lock;
100 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
101 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock)
102 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock)
103 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED)
104
105 #define ADDR_LABEL_NOTAPP (-1)
106 struct in6_addrpolicy defaultaddrpolicy;
107
108 int ip6_prefer_tempaddr = 0;
109
110 static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *,
111 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **));
112
113 static struct in6_addrpolicy *lookup_addrsel_policy __P((struct sockaddr_in6 *));
114
115 static void init_policy_queue __P((void));
116 static int add_addrsel_policyent __P((struct in6_addrpolicy *));
117 static int delete_addrsel_policyent __P((struct in6_addrpolicy *));
118 static int walk_addrsel_policy __P((int (*)(struct in6_addrpolicy *, void *),
119 void *));
120 static int dump_addrsel_policyent __P((struct in6_addrpolicy *, void *));
121 static struct in6_addrpolicy *match_addrsel_policy __P((struct sockaddr_in6 *));
122
123 /*
124 * Return an IPv6 address, which is the most appropriate for a given
125 * destination and user specified options.
126 * If necessary, this function lookups the routing table and returns
127 * an entry to the caller for later use.
128 */
129 #define REPLACE(r) do {\
130 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
131 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
132 ip6stat.ip6s_sources_rule[(r)]++; \
133 /* printf("in6_selectsrc: replace %s with %s by %d\n", ia_best ? ip6_sprintf(&ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(&ia->ia_addr.sin6_addr), (r)); */ \
134 goto replace; \
135 } while(0)
136 #define NEXT(r) do {\
137 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
138 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
139 ip6stat.ip6s_sources_rule[(r)]++; \
140 /* printf("in6_selectsrc: keep %s against %s by %d\n", ia_best ? ip6_sprintf(&ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(&ia->ia_addr.sin6_addr), (r)); */ \
141 goto next; /* XXX: we can't use 'continue' here */ \
142 } while(0)
143 #define BREAK(r) do { \
144 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
145 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
146 ip6stat.ip6s_sources_rule[(r)]++; \
147 goto out; /* XXX: we can't use 'break' here */ \
148 } while(0)
149
150 struct in6_addr *
151 in6_selectsrc(dstsock, opts, mopts, ro, laddr, errorp)
152 struct sockaddr_in6 *dstsock;
153 struct ip6_pktopts *opts;
154 struct ip6_moptions *mopts;
155 struct route_in6 *ro;
156 struct in6_addr *laddr;
157 int *errorp;
158 {
159 struct in6_addr *dst;
160 struct ifnet *ifp = NULL;
161 struct in6_ifaddr *ia = NULL, *ia_best = NULL;
162 struct in6_pktinfo *pi = NULL;
163 int dst_scope = -1, best_scope = -1, best_matchlen = -1;
164 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
165 u_int32_t odstzone;
166 int prefer_tempaddr;
167 struct sockaddr_in6 dstsock0;
168
169 dstsock0 = *dstsock;
170 if (IN6_IS_SCOPE_LINKLOCAL(&dstsock0.sin6_addr) ||
171 IN6_IS_ADDR_MC_INTFACELOCAL(&dstsock0.sin6_addr)) {
172 /* KAME assumption: link id == interface id */
173 if (opts && opts->ip6po_pktinfo &&
174 opts->ip6po_pktinfo->ipi6_ifindex) {
175 ifp = ifnet_byindex(opts->ip6po_pktinfo->ipi6_ifindex);
176 dstsock0.sin6_addr.s6_addr16[1] =
177 htons(opts->ip6po_pktinfo->ipi6_ifindex);
178 } else if (mopts &&
179 IN6_IS_ADDR_MULTICAST(&dstsock0.sin6_addr) &&
180 mopts->im6o_multicast_ifp) {
181 ifp = mopts->im6o_multicast_ifp;
182 dstsock0.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
183 } else if ((*errorp = in6_embedscope(&dstsock0.sin6_addr,
184 &dstsock0, NULL, NULL)) != 0)
185 return (NULL);
186 }
187 dstsock = &dstsock0;
188
189 dst = &dstsock->sin6_addr;
190 *errorp = 0;
191
192 /*
193 * If the source address is explicitly specified by the caller,
194 * check if the requested source address is indeed a unicast address
195 * assigned to the node, and can be used as the packet's source
196 * address. If everything is okay, use the address as source.
197 */
198 if (opts && (pi = opts->ip6po_pktinfo) &&
199 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
200 struct sockaddr_in6 srcsock;
201 struct in6_ifaddr *ia6;
202
203 /* get the outgoing interface */
204 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp))
205 != 0) {
206 return (NULL);
207 }
208
209 /*
210 * determine the appropriate zone id of the source based on
211 * the zone of the destination and the outgoing interface.
212 */
213 bzero(&srcsock, sizeof(srcsock));
214 srcsock.sin6_family = AF_INET6;
215 srcsock.sin6_len = sizeof(srcsock);
216 srcsock.sin6_addr = pi->ipi6_addr;
217 if (ifp) {
218 if (in6_addr2zoneid(ifp, &pi->ipi6_addr,
219 &srcsock.sin6_scope_id)) {
220 *errorp = EINVAL; /* XXX */
221 return (NULL);
222 }
223 }
224 if ((*errorp = in6_embedscope(&srcsock.sin6_addr, &srcsock,
225 NULL, NULL)) != 0) {
226 return (NULL);
227 }
228 srcsock.sin6_scope_id = 0; /* XXX: ifa_ifwithaddr expects 0 */
229 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)(&srcsock));
230 if (ia6 == NULL ||
231 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) {
232 *errorp = EADDRNOTAVAIL;
233 return (NULL);
234 }
235 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
236 return (&ia6->ia_addr.sin6_addr);
237 }
238
239 /*
240 * Otherwise, if the socket has already bound the source, just use it.
241 */
242 if (laddr && !IN6_IS_ADDR_UNSPECIFIED(laddr))
243 return (laddr);
244
245 /*
246 * If the address is not specified, choose the best one based on
247 * the outgoing interface and the destination address.
248 */
249 /* get the outgoing interface */
250 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0)
251 return (NULL);
252
253 #ifdef DIAGNOSTIC
254 if (ifp == NULL) /* this should not happen */
255 panic("in6_selectsrc: NULL ifp");
256 #endif
257 if (in6_addr2zoneid(ifp, dst, &odstzone)) { /* impossible */
258 *errorp = EIO; /* XXX */
259 return (NULL);
260 }
261 for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
262 int new_scope = -1, new_matchlen = -1;
263 struct in6_addrpolicy *new_policy = NULL;
264 u_int32_t srczone, osrczone, dstzone;
265 struct ifnet *ifp1 = ia->ia_ifp;
266
267 /*
268 * We'll never take an address that breaks the scope zone
269 * of the destination. We also skip an address if its zone
270 * does not contain the outgoing interface.
271 * XXX: we should probably use sin6_scope_id here.
272 */
273 if (in6_addr2zoneid(ifp1, dst, &dstzone) ||
274 odstzone != dstzone) {
275 continue;
276 }
277 if (in6_addr2zoneid(ifp, &ia->ia_addr.sin6_addr, &osrczone) ||
278 in6_addr2zoneid(ifp1, &ia->ia_addr.sin6_addr, &srczone) ||
279 osrczone != srczone) {
280 continue;
281 }
282
283 /* avoid unusable addresses */
284 if ((ia->ia6_flags &
285 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
286 continue;
287 }
288 if (!ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
289 continue;
290
291 /* Rule 1: Prefer same address */
292 if (IN6_ARE_ADDR_EQUAL(dst, &ia->ia_addr.sin6_addr)) {
293 ia_best = ia;
294 BREAK(1); /* there should be no better candidate */
295 }
296
297 if (ia_best == NULL)
298 REPLACE(0);
299
300 /* Rule 2: Prefer appropriate scope */
301 if (dst_scope < 0)
302 dst_scope = in6_addrscope(dst);
303 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
304 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
305 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
306 REPLACE(2);
307 NEXT(2);
308 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
309 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
310 NEXT(2);
311 REPLACE(2);
312 }
313
314 /*
315 * Rule 3: Avoid deprecated addresses. Note that the case of
316 * !ip6_use_deprecated is already rejected above.
317 */
318 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
319 NEXT(3);
320 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
321 REPLACE(3);
322
323 /* Rule 4: Prefer home addresses */
324 /*
325 * XXX: This is a TODO. We should probably merge the MIP6
326 * case above.
327 */
328
329 /* Rule 5: Prefer outgoing interface */
330 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
331 NEXT(5);
332 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
333 REPLACE(5);
334
335 /*
336 * Rule 6: Prefer matching label
337 * Note that best_policy should be non-NULL here.
338 */
339 if (dst_policy == NULL)
340 dst_policy = lookup_addrsel_policy(dstsock);
341 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
342 new_policy = lookup_addrsel_policy(&ia->ia_addr);
343 if (dst_policy->label == best_policy->label &&
344 dst_policy->label != new_policy->label)
345 NEXT(6);
346 if (dst_policy->label != best_policy->label &&
347 dst_policy->label == new_policy->label)
348 REPLACE(6);
349 }
350
351 /*
352 * Rule 7: Prefer public addresses.
353 * We allow users to reverse the logic by configuring
354 * a sysctl variable, so that privacy conscious users can
355 * always prefer temporary addresses.
356 */
357 if (opts == NULL ||
358 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
359 prefer_tempaddr = ip6_prefer_tempaddr;
360 } else if (opts->ip6po_prefer_tempaddr ==
361 IP6PO_TEMPADDR_NOTPREFER) {
362 prefer_tempaddr = 0;
363 } else
364 prefer_tempaddr = 1;
365 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
366 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
367 if (prefer_tempaddr)
368 REPLACE(7);
369 else
370 NEXT(7);
371 }
372 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
373 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
374 if (prefer_tempaddr)
375 NEXT(7);
376 else
377 REPLACE(7);
378 }
379
380 /*
381 * Rule 8: prefer addresses on alive interfaces.
382 * This is a KAME specific rule.
383 */
384 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
385 !(ia->ia_ifp->if_flags & IFF_UP))
386 NEXT(8);
387 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
388 (ia->ia_ifp->if_flags & IFF_UP))
389 REPLACE(8);
390
391 /*
392 * Rule 14: Use longest matching prefix.
393 * Note: in the address selection draft, this rule is
394 * documented as "Rule 8". However, since it is also
395 * documented that this rule can be overridden, we assign
396 * a large number so that it is easy to assign smaller numbers
397 * to more preferred rules.
398 */
399 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, dst);
400 if (best_matchlen < new_matchlen)
401 REPLACE(14);
402 if (new_matchlen < best_matchlen)
403 NEXT(14);
404
405 /* Rule 15 is reserved. */
406
407 /*
408 * Last resort: just keep the current candidate.
409 * Or, do we need more rules?
410 */
411 continue;
412
413 replace:
414 ia_best = ia;
415 best_scope = (new_scope >= 0 ? new_scope :
416 in6_addrscope(&ia_best->ia_addr.sin6_addr));
417 best_policy = (new_policy ? new_policy :
418 lookup_addrsel_policy(&ia_best->ia_addr));
419 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
420 in6_matchlen(&ia_best->ia_addr.sin6_addr,
421 dst));
422
423 next:
424 continue;
425
426 out:
427 break;
428 }
429
430 if ((ia = ia_best) == NULL) {
431 *errorp = EADDRNOTAVAIL;
432 return (NULL);
433 }
434
435 return (&ia->ia_addr.sin6_addr);
436 }
437
438 static int
439 in6_selectif(dstsock, opts, mopts, ro, retifp)
440 struct sockaddr_in6 *dstsock;
441 struct ip6_pktopts *opts;
442 struct ip6_moptions *mopts;
443 struct route_in6 *ro;
444 struct ifnet **retifp;
445 {
446 int error;
447 struct route_in6 sro;
448 struct rtentry *rt = NULL;
449
450 if (ro == NULL) {
451 bzero(&sro, sizeof(sro));
452 ro = &sro;
453 }
454
455 if ((error = in6_selectroute(dstsock, opts, mopts, ro, retifp,
456 &rt, 0)) != 0) {
457 if (rt && rt == sro.ro_rt)
458 RTFREE(rt);
459 return (error);
460 }
461
462 /*
463 * do not use a rejected or black hole route.
464 * XXX: this check should be done in the L2 output routine.
465 * However, if we skipped this check here, we'd see the following
466 * scenario:
467 * - install a rejected route for a scoped address prefix
468 * (like fe80::/10)
469 * - send a packet to a destination that matches the scoped prefix,
470 * with ambiguity about the scope zone.
471 * - pick the outgoing interface from the route, and disambiguate the
472 * scope zone with the interface.
473 * - ip6_output() would try to get another route with the "new"
474 * destination, which may be valid.
475 * - we'd see no error on output.
476 * Although this may not be very harmful, it should still be confusing.
477 * We thus reject the case here.
478 */
479 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
480 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
481
482 if (rt && rt == sro.ro_rt)
483 RTFREE(rt);
484 return (flags);
485 }
486
487 /*
488 * Adjust the "outgoing" interface. If we're going to loop the packet
489 * back to ourselves, the ifp would be the loopback interface.
490 * However, we'd rather know the interface associated to the
491 * destination address (which should probably be one of our own
492 * addresses.)
493 */
494 if (rt && rt->rt_ifa && rt->rt_ifa->ifa_ifp)
495 *retifp = rt->rt_ifa->ifa_ifp;
496
497 if (rt && rt == sro.ro_rt)
498 RTFREE(rt);
499 return (0);
500 }
501
502 int
503 in6_selectroute(dstsock, opts, mopts, ro, retifp, retrt, clone)
504 struct sockaddr_in6 *dstsock;
505 struct ip6_pktopts *opts;
506 struct ip6_moptions *mopts;
507 struct route_in6 *ro;
508 struct ifnet **retifp;
509 struct rtentry **retrt;
510 int clone; /* meaningful only for bsdi and freebsd. */
511 {
512 int error = 0;
513 struct ifnet *ifp = NULL;
514 struct rtentry *rt = NULL;
515 struct sockaddr_in6 *sin6_next;
516 struct in6_pktinfo *pi = NULL;
517 struct in6_addr *dst = &dstsock->sin6_addr;
518
519 #if 0
520 if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
521 dstsock->sin6_addr.s6_addr32[1] == 0 &&
522 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
523 printf("in6_selectroute: strange destination %s\n",
524 ip6_sprintf(&dstsock->sin6_addr));
525 } else {
526 printf("in6_selectroute: destination = %s%%%d\n",
527 ip6_sprintf(&dstsock->sin6_addr),
528 dstsock->sin6_scope_id); /* for debug */
529 }
530 #endif
531
532 /* If the caller specify the outgoing interface explicitly, use it. */
533 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
534 /* XXX boundary check is assumed to be already done. */
535 ifp = ifnet_byindex(pi->ipi6_ifindex);
536 if (ifp != NULL &&
537 (retrt == NULL || IN6_IS_ADDR_MULTICAST(dst))) {
538 /*
539 * we do not have to check nor get the route for
540 * multicast.
541 */
542 goto done;
543 } else
544 goto getroute;
545 }
546
547 /*
548 * If the destination address is a multicast address and the outgoing
549 * interface for the address is specified by the caller, use it.
550 */
551 if (IN6_IS_ADDR_MULTICAST(dst) &&
552 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
553 goto done; /* we do not need a route for multicast. */
554 }
555
556 getroute:
557 /*
558 * If the next hop address for the packet is specified by the caller,
559 * use it as the gateway.
560 */
561 if (opts && opts->ip6po_nexthop) {
562 struct route_in6 *ron;
563
564 sin6_next = satosin6(opts->ip6po_nexthop);
565
566 /* at this moment, we only support AF_INET6 next hops */
567 if (sin6_next->sin6_family != AF_INET6) {
568 error = EAFNOSUPPORT; /* or should we proceed? */
569 goto done;
570 }
571
572 /*
573 * If the next hop is an IPv6 address, then the node identified
574 * by that address must be a neighbor of the sending host.
575 */
576 ron = &opts->ip6po_nextroute;
577 if ((ron->ro_rt &&
578 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) !=
579 (RTF_UP | RTF_LLINFO)) ||
580 !SA6_ARE_ADDR_EQUAL(satosin6(&ron->ro_dst), sin6_next)) {
581 if (ron->ro_rt) {
582 RTFREE(ron->ro_rt);
583 ron->ro_rt = NULL;
584 }
585 *satosin6(&ron->ro_dst) = *sin6_next;
586 }
587 if (ron->ro_rt == NULL) {
588 rtalloc((struct route *)ron); /* multi path case? */
589 if (ron->ro_rt == NULL ||
590 !(ron->ro_rt->rt_flags & RTF_LLINFO)) {
591 if (ron->ro_rt) {
592 RTFREE(ron->ro_rt);
593 ron->ro_rt = NULL;
594 }
595 error = EHOSTUNREACH;
596 goto done;
597 }
598 }
599 rt = ron->ro_rt;
600 ifp = rt->rt_ifp;
601
602 /*
603 * When cloning is required, try to allocate a route to the
604 * destination so that the caller can store path MTU
605 * information.
606 */
607 if (!clone)
608 goto done;
609 }
610
611 /*
612 * Use a cached route if it exists and is valid, else try to allocate
613 * a new one. Note that we should check the address family of the
614 * cached destination, in case of sharing the cache with IPv4.
615 */
616 if (ro) {
617 if (ro->ro_rt &&
618 (!(ro->ro_rt->rt_flags & RTF_UP) ||
619 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
620 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
621 dst))) {
622 RTFREE(ro->ro_rt);
623 ro->ro_rt = (struct rtentry *)NULL;
624 }
625 if (ro->ro_rt == (struct rtentry *)NULL) {
626 struct sockaddr_in6 *sa6;
627
628 /* No route yet, so try to acquire one */
629 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
630 sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
631 *sa6 = *dstsock;
632 sa6->sin6_scope_id = 0;
633
634 if (clone) {
635 rtalloc((struct route *)ro);
636 } else {
637 ro->ro_rt = rtalloc1(&((struct route *)ro)
638 ->ro_dst, 0, 0UL);
639 if (ro->ro_rt)
640 RT_UNLOCK(ro->ro_rt);
641 }
642 }
643
644 /*
645 * do not care about the result if we have the nexthop
646 * explicitly specified.
647 */
648 if (opts && opts->ip6po_nexthop)
649 goto done;
650
651 if (ro->ro_rt) {
652 ifp = ro->ro_rt->rt_ifp;
653
654 if (ifp == NULL) { /* can this really happen? */
655 RTFREE(ro->ro_rt);
656 ro->ro_rt = NULL;
657 }
658 }
659 if (ro->ro_rt == NULL)
660 error = EHOSTUNREACH;
661 rt = ro->ro_rt;
662
663 /*
664 * Check if the outgoing interface conflicts with
665 * the interface specified by ipi6_ifindex (if specified).
666 * Note that loopback interface is always okay.
667 * (this may happen when we are sending a packet to one of
668 * our own addresses.)
669 */
670 if (opts && opts->ip6po_pktinfo &&
671 opts->ip6po_pktinfo->ipi6_ifindex) {
672 if (!(ifp->if_flags & IFF_LOOPBACK) &&
673 ifp->if_index !=
674 opts->ip6po_pktinfo->ipi6_ifindex) {
675 error = EHOSTUNREACH;
676 goto done;
677 }
678 }
679 }
680
681 done:
682 if (ifp == NULL && rt == NULL) {
683 /*
684 * This can happen if the caller did not pass a cached route
685 * nor any other hints. We treat this case an error.
686 */
687 error = EHOSTUNREACH;
688 }
689 if (error == EHOSTUNREACH)
690 ip6stat.ip6s_noroute++;
691
692 if (retifp != NULL)
693 *retifp = ifp;
694 if (retrt != NULL)
695 *retrt = rt; /* rt may be NULL */
696
697 return (error);
698 }
699
700 /*
701 * Default hop limit selection. The precedence is as follows:
702 * 1. Hoplimit value specified via ioctl.
703 * 2. (If the outgoing interface is detected) the current
704 * hop limit of the interface specified by router advertisement.
705 * 3. The system default hoplimit.
706 */
707 int
708 in6_selecthlim(in6p, ifp)
709 struct in6pcb *in6p;
710 struct ifnet *ifp;
711 {
712 if (in6p && in6p->in6p_hops >= 0)
713 return (in6p->in6p_hops);
714 else if (ifp)
715 return (ND_IFINFO(ifp)->chlim);
716 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
717 struct route_in6 ro6;
718 struct ifnet *lifp;
719
720 bzero(&ro6, sizeof(ro6));
721 ro6.ro_dst.sin6_family = AF_INET6;
722 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
723 ro6.ro_dst.sin6_addr = in6p->in6p_faddr;
724 rtalloc((struct route *)&ro6);
725 if (ro6.ro_rt) {
726 lifp = ro6.ro_rt->rt_ifp;
727 RTFREE(ro6.ro_rt);
728 if (lifp)
729 return (ND_IFINFO(lifp)->chlim);
730 } else
731 return (ip6_defhlim);
732 }
733 return (ip6_defhlim);
734 }
735
736 /*
737 * XXX: this is borrowed from in6_pcbbind(). If possible, we should
738 * share this function by all *bsd*...
739 */
740 int
741 in6_pcbsetport(laddr, inp, cred)
742 struct in6_addr *laddr;
743 struct inpcb *inp;
744 struct ucred *cred;
745 {
746 struct socket *so = inp->inp_socket;
747 u_int16_t lport = 0, first, last, *lastport;
748 int count, error = 0, wild = 0;
749 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
750
751 /* XXX: this is redundant when called from in6_pcbbind */
752 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
753 wild = INPLOOKUP_WILDCARD;
754
755 inp->inp_flags |= INP_ANONPORT;
756
757 if (inp->inp_flags & INP_HIGHPORT) {
758 first = ipport_hifirstauto; /* sysctl */
759 last = ipport_hilastauto;
760 lastport = &pcbinfo->lasthi;
761 } else if (inp->inp_flags & INP_LOWPORT) {
762 if ((error = suser_cred(cred, 0)))
763 return error;
764 first = ipport_lowfirstauto; /* 1023 */
765 last = ipport_lowlastauto; /* 600 */
766 lastport = &pcbinfo->lastlow;
767 } else {
768 first = ipport_firstauto; /* sysctl */
769 last = ipport_lastauto;
770 lastport = &pcbinfo->lastport;
771 }
772 /*
773 * Simple check to ensure all ports are not used up causing
774 * a deadlock here.
775 *
776 * We split the two cases (up and down) so that the direction
777 * is not being tested on each round of the loop.
778 */
779 if (first > last) {
780 /*
781 * counting down
782 */
783 count = first - last;
784
785 do {
786 if (count-- < 0) { /* completely used? */
787 /*
788 * Undo any address bind that may have
789 * occurred above.
790 */
791 inp->in6p_laddr = in6addr_any;
792 return (EAGAIN);
793 }
794 --*lastport;
795 if (*lastport > first || *lastport < last)
796 *lastport = first;
797 lport = htons(*lastport);
798 } while (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr,
799 lport, wild));
800 } else {
801 /*
802 * counting up
803 */
804 count = last - first;
805
806 do {
807 if (count-- < 0) { /* completely used? */
808 /*
809 * Undo any address bind that may have
810 * occurred above.
811 */
812 inp->in6p_laddr = in6addr_any;
813 return (EAGAIN);
814 }
815 ++*lastport;
816 if (*lastport < first || *lastport > last)
817 *lastport = first;
818 lport = htons(*lastport);
819 } while (in6_pcblookup_local(pcbinfo,
820 &inp->in6p_laddr, lport, wild));
821 }
822
823 inp->inp_lport = lport;
824 if (in_pcbinshash(inp) != 0) {
825 inp->in6p_laddr = in6addr_any;
826 inp->inp_lport = 0;
827 return (EAGAIN);
828 }
829
830 return (0);
831 }
832
833 /*
834 * Generate kernel-internal form (scopeid embedded into s6_addr16[1]).
835 * If the address scope of is link-local, embed the interface index in the
836 * address. The routine determines our precedence
837 * between advanced API scope/interface specification and basic API
838 * specification.
839 *
840 * This function should be nuked in the future, when we get rid of embedded
841 * scopeid thing.
842 *
843 * XXX actually, it is over-specification to return ifp against sin6_scope_id.
844 * there can be multiple interfaces that belong to a particular scope zone
845 * (in specification, we have 1:N mapping between a scope zone and interfaces).
846 * we may want to change the function to return something other than ifp.
847 */
848 int
849 in6_embedscope(in6, sin6, in6p, ifpp)
850 struct in6_addr *in6;
851 const struct sockaddr_in6 *sin6;
852 struct in6pcb *in6p;
853 struct ifnet **ifpp;
854 {
855 struct ifnet *ifp = NULL;
856 u_int32_t zoneid = sin6->sin6_scope_id;
857
858 *in6 = sin6->sin6_addr;
859 if (ifpp)
860 *ifpp = NULL;
861
862 /*
863 * don't try to read sin6->sin6_addr beyond here, since the caller may
864 * ask us to overwrite existing sockaddr_in6
865 */
866
867 #ifdef ENABLE_DEFAULT_SCOPE
868 if (zoneid == 0)
869 zoneid = scope6_addr2default(in6);
870 #endif
871
872 if (IN6_IS_SCOPE_LINKLOCAL(in6) || IN6_IS_ADDR_MC_INTFACELOCAL(in6)) {
873 struct in6_pktinfo *pi;
874
875 /* KAME assumption: link id == interface id */
876 if (in6p && in6p->in6p_outputopts &&
877 (pi = in6p->in6p_outputopts->ip6po_pktinfo) &&
878 pi->ipi6_ifindex) {
879 ifp = ifnet_byindex(pi->ipi6_ifindex);
880 in6->s6_addr16[1] = htons(pi->ipi6_ifindex);
881 } else if (in6p && IN6_IS_ADDR_MULTICAST(in6) &&
882 in6p->in6p_moptions &&
883 in6p->in6p_moptions->im6o_multicast_ifp) {
884 ifp = in6p->in6p_moptions->im6o_multicast_ifp;
885 in6->s6_addr16[1] = htons(ifp->if_index);
886 } else if (zoneid) {
887 if (if_index < zoneid)
888 return (ENXIO); /* XXX EINVAL? */
889 ifp = ifnet_byindex(zoneid);
890
891 /* XXX assignment to 16bit from 32bit variable */
892 in6->s6_addr16[1] = htons(zoneid & 0xffff);
893 }
894
895 if (ifpp)
896 *ifpp = ifp;
897 }
898
899 return 0;
900 }
901
902 /*
903 * generate standard sockaddr_in6 from embedded form.
904 * touches sin6_addr and sin6_scope_id only.
905 *
906 * this function should be nuked in the future, when we get rid of
907 * embedded scopeid thing.
908 */
909 int
910 in6_recoverscope(sin6, in6, ifp)
911 struct sockaddr_in6 *sin6;
912 const struct in6_addr *in6;
913 struct ifnet *ifp;
914 {
915 u_int32_t zoneid;
916
917 sin6->sin6_addr = *in6;
918
919 /*
920 * don't try to read *in6 beyond here, since the caller may
921 * ask us to overwrite existing sockaddr_in6
922 */
923
924 sin6->sin6_scope_id = 0;
925 if (IN6_IS_SCOPE_LINKLOCAL(in6) || IN6_IS_ADDR_MC_INTFACELOCAL(in6)) {
926 /*
927 * KAME assumption: link id == interface id
928 */
929 zoneid = ntohs(sin6->sin6_addr.s6_addr16[1]);
930 if (zoneid) {
931 /* sanity check */
932 if (zoneid < 0 || if_index < zoneid)
933 return ENXIO;
934 if (ifp && ifp->if_index != zoneid)
935 return ENXIO;
936 sin6->sin6_addr.s6_addr16[1] = 0;
937 sin6->sin6_scope_id = zoneid;
938 }
939 }
940
941 return 0;
942 }
943
944 /*
945 * just clear the embedded scope identifier.
946 */
947 void
948 in6_clearscope(addr)
949 struct in6_addr *addr;
950 {
951 if (IN6_IS_SCOPE_LINKLOCAL(addr) || IN6_IS_ADDR_MC_INTFACELOCAL(addr))
952 addr->s6_addr16[1] = 0;
953 }
954
955 void
956 addrsel_policy_init()
957 {
958 ADDRSEL_LOCK_INIT();
959
960 init_policy_queue();
961
962 /* initialize the "last resort" policy */
963 bzero(&defaultaddrpolicy, sizeof(defaultaddrpolicy));
964 defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
965 }
966
967 static struct in6_addrpolicy *
968 lookup_addrsel_policy(key)
969 struct sockaddr_in6 *key;
970 {
971 struct in6_addrpolicy *match = NULL;
972
973 ADDRSEL_LOCK();
974 match = match_addrsel_policy(key);
975
976 if (match == NULL)
977 match = &defaultaddrpolicy;
978 else
979 match->use++;
980 ADDRSEL_UNLOCK();
981
982 return (match);
983 }
984
985 /*
986 * Subroutines to manage the address selection policy table via sysctl.
987 */
988 struct walkarg {
989 struct sysctl_req *w_req;
990 };
991
992 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
993 SYSCTL_DECL(_net_inet6_ip6);
994 SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
995 CTLFLAG_RD, in6_src_sysctl, "");
996
997 static int
998 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
999 {
1000 struct walkarg w;
1001
1002 if (req->newptr)
1003 return EPERM;
1004
1005 bzero(&w, sizeof(w));
1006 w.w_req = req;
1007
1008 return (walk_addrsel_policy(dump_addrsel_policyent, &w));
1009 }
1010
1011 int
1012 in6_src_ioctl(cmd, data)
1013 u_long cmd;
1014 caddr_t data;
1015 {
1016 int i;
1017 struct in6_addrpolicy ent0;
1018
1019 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
1020 return (EOPNOTSUPP); /* check for safety */
1021
1022 ent0 = *(struct in6_addrpolicy *)data;
1023
1024 if (ent0.label == ADDR_LABEL_NOTAPP)
1025 return (EINVAL);
1026 /* check if the prefix mask is consecutive. */
1027 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
1028 return (EINVAL);
1029 /* clear trailing garbages (if any) of the prefix address. */
1030 for (i = 0; i < 4; i++) {
1031 ent0.addr.sin6_addr.s6_addr32[i] &=
1032 ent0.addrmask.sin6_addr.s6_addr32[i];
1033 }
1034 ent0.use = 0;
1035
1036 switch (cmd) {
1037 case SIOCAADDRCTL_POLICY:
1038 return (add_addrsel_policyent(&ent0));
1039 case SIOCDADDRCTL_POLICY:
1040 return (delete_addrsel_policyent(&ent0));
1041 }
1042
1043 return (0); /* XXX: compromise compilers */
1044 }
1045
1046 /*
1047 * The followings are implementation of the policy table using a
1048 * simple tail queue.
1049 * XXX such details should be hidden.
1050 * XXX implementation using binary tree should be more efficient.
1051 */
1052 struct addrsel_policyent {
1053 TAILQ_ENTRY(addrsel_policyent) ape_entry;
1054 struct in6_addrpolicy ape_policy;
1055 };
1056
1057 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1058
1059 struct addrsel_policyhead addrsel_policytab;
1060
1061 static void
1062 init_policy_queue()
1063 {
1064 TAILQ_INIT(&addrsel_policytab);
1065 }
1066
1067 static int
1068 add_addrsel_policyent(newpolicy)
1069 struct in6_addrpolicy *newpolicy;
1070 {
1071 struct addrsel_policyent *new, *pol;
1072
1073 MALLOC(new, struct addrsel_policyent *, sizeof(*new), M_IFADDR,
1074 M_WAITOK);
1075 ADDRSEL_LOCK();
1076
1077 /* duplication check */
1078 for (pol = TAILQ_FIRST(&addrsel_policytab); pol;
1079 pol = TAILQ_NEXT(pol, ape_entry)) {
1080 if (SA6_ARE_ADDR_EQUAL(&newpolicy->addr,
1081 &pol->ape_policy.addr) &&
1082 SA6_ARE_ADDR_EQUAL(&newpolicy->addrmask,
1083 &pol->ape_policy.addrmask)) {
1084 ADDRSEL_UNLOCK();
1085 FREE(new, M_IFADDR);
1086 return (EEXIST); /* or override it? */
1087 }
1088 }
1089
1090 bzero(new, sizeof(*new));
1091
1092 /* XXX: should validate entry */
1093 new->ape_policy = *newpolicy;
1094
1095 TAILQ_INSERT_TAIL(&addrsel_policytab, new, ape_entry);
1096 ADDRSEL_UNLOCK();
1097
1098 return (0);
1099 }
1100
1101 static int
1102 delete_addrsel_policyent(key)
1103 struct in6_addrpolicy *key;
1104 {
1105 struct addrsel_policyent *pol;
1106
1107 ADDRSEL_LOCK();
1108
1109 /* search for the entry in the table */
1110 for (pol = TAILQ_FIRST(&addrsel_policytab); pol;
1111 pol = TAILQ_NEXT(pol, ape_entry)) {
1112 if (SA6_ARE_ADDR_EQUAL(&key->addr, &pol->ape_policy.addr) &&
1113 SA6_ARE_ADDR_EQUAL(&key->addrmask,
1114 &pol->ape_policy.addrmask)) {
1115 break;
1116 }
1117 }
1118 if (pol == NULL) {
1119 ADDRSEL_UNLOCK();
1120 return (ESRCH);
1121 }
1122
1123 TAILQ_REMOVE(&addrsel_policytab, pol, ape_entry);
1124 ADDRSEL_UNLOCK();
1125
1126 return (0);
1127 }
1128
1129 static int
1130 walk_addrsel_policy(callback, w)
1131 int (*callback) __P((struct in6_addrpolicy *, void *));
1132 void *w;
1133 {
1134 struct addrsel_policyent *pol;
1135 int error = 0;
1136
1137 ADDRSEL_LOCK();
1138 for (pol = TAILQ_FIRST(&addrsel_policytab); pol;
1139 pol = TAILQ_NEXT(pol, ape_entry)) {
1140 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1141 ADDRSEL_UNLOCK();
1142 return (error);
1143 }
1144 }
1145 ADDRSEL_UNLOCK();
1146
1147 return (error);
1148 }
1149
1150 static int
1151 dump_addrsel_policyent(pol, arg)
1152 struct in6_addrpolicy *pol;
1153 void *arg;
1154 {
1155 int error = 0;
1156 struct walkarg *w = arg;
1157
1158 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1159
1160 return (error);
1161 }
1162
1163 static struct in6_addrpolicy *
1164 match_addrsel_policy(key)
1165 struct sockaddr_in6 *key;
1166 {
1167 struct addrsel_policyent *pent;
1168 struct in6_addrpolicy *bestpol = NULL, *pol;
1169 int matchlen, bestmatchlen = -1;
1170 u_char *mp, *ep, *k, *p, m;
1171
1172 for (pent = TAILQ_FIRST(&addrsel_policytab); pent;
1173 pent = TAILQ_NEXT(pent, ape_entry)) {
1174 matchlen = 0;
1175
1176 pol = &pent->ape_policy;
1177 mp = (u_char *)&pol->addrmask.sin6_addr;
1178 ep = mp + 16; /* XXX: scope field? */
1179 k = (u_char *)&key->sin6_addr;
1180 p = (u_char *)&pol->addr.sin6_addr;
1181 for (; mp < ep && *mp; mp++, k++, p++) {
1182 m = *mp;
1183 if ((*k & m) != *p)
1184 goto next; /* not match */
1185 if (m == 0xff) /* short cut for a typical case */
1186 matchlen += 8;
1187 else {
1188 while (m >= 0x80) {
1189 matchlen++;
1190 m <<= 1;
1191 }
1192 }
1193 }
1194
1195 /* matched. check if this is better than the current best. */
1196 if (bestpol == NULL ||
1197 matchlen > bestmatchlen) {
1198 bestpol = pol;
1199 bestmatchlen = matchlen;
1200 }
1201
1202 next:
1203 continue;
1204 }
1205
1206 return (bestpol);
1207 }
Cache object: 1b9a49b7b650b2504c22ca7cb2cd0532
|