1 /*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * $KAME: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $
30 */
31
32 /*-
33 * Copyright (c) 1982, 1986, 1991, 1993
34 * The Regents of the University of California. All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 * notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 * notice, this list of conditions and the following disclaimer in the
43 * documentation and/or other materials provided with the distribution.
44 * 4. Neither the name of the University nor the names of its contributors
45 * may be used to endorse or promote products derived from this software
46 * without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 * @(#)in_pcb.c 8.2 (Berkeley) 1/4/94
61 */
62
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/malloc.h>
72 #include <sys/mbuf.h>
73 #include <sys/protosw.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/sockio.h>
77 #include <sys/sysctl.h>
78 #include <sys/errno.h>
79 #include <sys/time.h>
80 #include <sys/kernel.h>
81 #include <sys/lock.h>
82 #include <sys/sx.h>
83
84 #include <net/if.h>
85 #include <net/route.h>
86
87 #include <netinet/in.h>
88 #include <netinet/in_var.h>
89 #include <netinet/in_systm.h>
90 #include <netinet/ip.h>
91 #include <netinet/in_pcb.h>
92 #include <netinet6/in6_var.h>
93 #include <netinet/ip6.h>
94 #include <netinet6/in6_pcb.h>
95 #include <netinet6/ip6_var.h>
96 #include <netinet6/scope6_var.h>
97 #include <netinet6/nd6.h>
98
99 #include <net/net_osdep.h>
100
101 static struct mtx addrsel_lock;
102 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
103 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock)
104 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock)
105 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED)
106
107 static struct sx addrsel_sxlock;
108 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock")
109 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock)
110 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock)
111 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock)
112 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock)
113
114 #define ADDR_LABEL_NOTAPP (-1)
115 struct in6_addrpolicy defaultaddrpolicy;
116
117 int ip6_prefer_tempaddr = 0;
118
119 static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *,
120 struct ip6_moptions *, struct route_in6 *, struct ifnet **,
121 struct rtentry **, int, int));
122 static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *,
123 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **));
124
125 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
126
127 static void init_policy_queue(void);
128 static int add_addrsel_policyent(struct in6_addrpolicy *);
129 static int delete_addrsel_policyent(struct in6_addrpolicy *);
130 static int walk_addrsel_policy __P((int (*)(struct in6_addrpolicy *, void *),
131 void *));
132 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
133 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
134
135 /*
136 * Return an IPv6 address, which is the most appropriate for a given
137 * destination and user specified options.
138 * If necessary, this function lookups the routing table and returns
139 * an entry to the caller for later use.
140 */
141 #define REPLACE(r) do {\
142 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
143 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
144 ip6stat.ip6s_sources_rule[(r)]++; \
145 /* 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)); */ \
146 goto replace; \
147 } while(0)
148 #define NEXT(r) do {\
149 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
150 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
151 ip6stat.ip6s_sources_rule[(r)]++; \
152 /* 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)); */ \
153 goto next; /* XXX: we can't use 'continue' here */ \
154 } while(0)
155 #define BREAK(r) do { \
156 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
157 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
158 ip6stat.ip6s_sources_rule[(r)]++; \
159 goto out; /* XXX: we can't use 'break' here */ \
160 } while(0)
161
162 struct in6_addr *
163 in6_selectsrc(dstsock, opts, mopts, ro, laddr, ifpp, errorp)
164 struct sockaddr_in6 *dstsock;
165 struct ip6_pktopts *opts;
166 struct ip6_moptions *mopts;
167 struct route_in6 *ro;
168 struct in6_addr *laddr;
169 struct ifnet **ifpp;
170 int *errorp;
171 {
172 struct in6_addr dst;
173 struct ifnet *ifp = NULL;
174 struct in6_ifaddr *ia = NULL, *ia_best = NULL;
175 struct in6_pktinfo *pi = NULL;
176 int dst_scope = -1, best_scope = -1, best_matchlen = -1;
177 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
178 u_int32_t odstzone;
179 int prefer_tempaddr;
180
181 dst = dstsock->sin6_addr; /* make a copy for local operation */
182 *errorp = 0;
183 if (ifpp)
184 *ifpp = NULL;
185
186 /*
187 * If the source address is explicitly specified by the caller,
188 * check if the requested source address is indeed a unicast address
189 * assigned to the node, and can be used as the packet's source
190 * address. If everything is okay, use the address as source.
191 */
192 if (opts && (pi = opts->ip6po_pktinfo) &&
193 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
194 struct sockaddr_in6 srcsock;
195 struct in6_ifaddr *ia6;
196
197 /* get the outgoing interface */
198 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp))
199 != 0) {
200 return (NULL);
201 }
202
203 /*
204 * determine the appropriate zone id of the source based on
205 * the zone of the destination and the outgoing interface.
206 * If the specified address is ambiguous wrt the scope zone,
207 * the interface must be specified; otherwise, ifa_ifwithaddr()
208 * will fail matching the address.
209 */
210 bzero(&srcsock, sizeof(srcsock));
211 srcsock.sin6_family = AF_INET6;
212 srcsock.sin6_len = sizeof(srcsock);
213 srcsock.sin6_addr = pi->ipi6_addr;
214 if (ifp) {
215 *errorp = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
216 if (*errorp != 0)
217 return (NULL);
218 }
219
220 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)(&srcsock));
221 if (ia6 == NULL ||
222 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) {
223 *errorp = EADDRNOTAVAIL;
224 return (NULL);
225 }
226 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
227 if (ifpp)
228 *ifpp = ifp;
229 return (&ia6->ia_addr.sin6_addr);
230 }
231
232 /*
233 * Otherwise, if the socket has already bound the source, just use it.
234 */
235 if (laddr && !IN6_IS_ADDR_UNSPECIFIED(laddr))
236 return (laddr);
237
238 /*
239 * If the address is not specified, choose the best one based on
240 * the outgoing interface and the destination address.
241 */
242 /* get the outgoing interface */
243 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0)
244 return (NULL);
245
246 #ifdef DIAGNOSTIC
247 if (ifp == NULL) /* this should not happen */
248 panic("in6_selectsrc: NULL ifp");
249 #endif
250 *errorp = in6_setscope(&dst, ifp, &odstzone);
251 if (*errorp != 0)
252 return (NULL);
253
254 for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
255 int new_scope = -1, new_matchlen = -1;
256 struct in6_addrpolicy *new_policy = NULL;
257 u_int32_t srczone, osrczone, dstzone;
258 struct in6_addr src;
259 struct ifnet *ifp1 = ia->ia_ifp;
260
261 /*
262 * We'll never take an address that breaks the scope zone
263 * of the destination. We also skip an address if its zone
264 * does not contain the outgoing interface.
265 * XXX: we should probably use sin6_scope_id here.
266 */
267 if (in6_setscope(&dst, ifp1, &dstzone) ||
268 odstzone != dstzone) {
269 continue;
270 }
271 src = ia->ia_addr.sin6_addr;
272 if (in6_setscope(&src, ifp, &osrczone) ||
273 in6_setscope(&src, ifp1, &srczone) ||
274 osrczone != srczone) {
275 continue;
276 }
277
278 /* avoid unusable addresses */
279 if ((ia->ia6_flags &
280 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
281 continue;
282 }
283 if (!ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
284 continue;
285
286 /* Rule 1: Prefer same address */
287 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
288 ia_best = ia;
289 BREAK(1); /* there should be no better candidate */
290 }
291
292 if (ia_best == NULL)
293 REPLACE(0);
294
295 /* Rule 2: Prefer appropriate scope */
296 if (dst_scope < 0)
297 dst_scope = in6_addrscope(&dst);
298 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
299 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
300 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
301 REPLACE(2);
302 NEXT(2);
303 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
304 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
305 NEXT(2);
306 REPLACE(2);
307 }
308
309 /*
310 * Rule 3: Avoid deprecated addresses. Note that the case of
311 * !ip6_use_deprecated is already rejected above.
312 */
313 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
314 NEXT(3);
315 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
316 REPLACE(3);
317
318 /* Rule 4: Prefer home addresses */
319 /*
320 * XXX: This is a TODO. We should probably merge the MIP6
321 * case above.
322 */
323
324 /* Rule 5: Prefer outgoing interface */
325 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
326 NEXT(5);
327 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
328 REPLACE(5);
329
330 /*
331 * Rule 6: Prefer matching label
332 * Note that best_policy should be non-NULL here.
333 */
334 if (dst_policy == NULL)
335 dst_policy = lookup_addrsel_policy(dstsock);
336 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
337 new_policy = lookup_addrsel_policy(&ia->ia_addr);
338 if (dst_policy->label == best_policy->label &&
339 dst_policy->label != new_policy->label)
340 NEXT(6);
341 if (dst_policy->label != best_policy->label &&
342 dst_policy->label == new_policy->label)
343 REPLACE(6);
344 }
345
346 /*
347 * Rule 7: Prefer public addresses.
348 * We allow users to reverse the logic by configuring
349 * a sysctl variable, so that privacy conscious users can
350 * always prefer temporary addresses.
351 */
352 if (opts == NULL ||
353 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
354 prefer_tempaddr = ip6_prefer_tempaddr;
355 } else if (opts->ip6po_prefer_tempaddr ==
356 IP6PO_TEMPADDR_NOTPREFER) {
357 prefer_tempaddr = 0;
358 } else
359 prefer_tempaddr = 1;
360 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
361 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
362 if (prefer_tempaddr)
363 REPLACE(7);
364 else
365 NEXT(7);
366 }
367 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
368 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
369 if (prefer_tempaddr)
370 NEXT(7);
371 else
372 REPLACE(7);
373 }
374
375 /*
376 * Rule 8: prefer addresses on alive interfaces.
377 * This is a KAME specific rule.
378 */
379 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
380 !(ia->ia_ifp->if_flags & IFF_UP))
381 NEXT(8);
382 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
383 (ia->ia_ifp->if_flags & IFF_UP))
384 REPLACE(8);
385
386 /*
387 * Rule 14: Use longest matching prefix.
388 * Note: in the address selection draft, this rule is
389 * documented as "Rule 8". However, since it is also
390 * documented that this rule can be overridden, we assign
391 * a large number so that it is easy to assign smaller numbers
392 * to more preferred rules.
393 */
394 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
395 if (best_matchlen < new_matchlen)
396 REPLACE(14);
397 if (new_matchlen < best_matchlen)
398 NEXT(14);
399
400 /* Rule 15 is reserved. */
401
402 /*
403 * Last resort: just keep the current candidate.
404 * Or, do we need more rules?
405 */
406 continue;
407
408 replace:
409 ia_best = ia;
410 best_scope = (new_scope >= 0 ? new_scope :
411 in6_addrscope(&ia_best->ia_addr.sin6_addr));
412 best_policy = (new_policy ? new_policy :
413 lookup_addrsel_policy(&ia_best->ia_addr));
414 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
415 in6_matchlen(&ia_best->ia_addr.sin6_addr,
416 &dst));
417
418 next:
419 continue;
420
421 out:
422 break;
423 }
424
425 if ((ia = ia_best) == NULL) {
426 *errorp = EADDRNOTAVAIL;
427 return (NULL);
428 }
429
430 if (ifpp)
431 *ifpp = ifp;
432
433 return (&ia->ia_addr.sin6_addr);
434 }
435
436 static int
437 selectroute(dstsock, opts, mopts, ro, retifp, retrt, clone, norouteok)
438 struct sockaddr_in6 *dstsock;
439 struct ip6_pktopts *opts;
440 struct ip6_moptions *mopts;
441 struct route_in6 *ro;
442 struct ifnet **retifp;
443 struct rtentry **retrt;
444 int clone; /* meaningful only for bsdi and freebsd. */
445 int norouteok;
446 {
447 int error = 0;
448 struct ifnet *ifp = NULL;
449 struct rtentry *rt = NULL;
450 struct sockaddr_in6 *sin6_next;
451 struct in6_pktinfo *pi = NULL;
452 struct in6_addr *dst = &dstsock->sin6_addr;
453
454 #if 0
455 if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
456 dstsock->sin6_addr.s6_addr32[1] == 0 &&
457 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
458 printf("in6_selectroute: strange destination %s\n",
459 ip6_sprintf(&dstsock->sin6_addr));
460 } else {
461 printf("in6_selectroute: destination = %s%%%d\n",
462 ip6_sprintf(&dstsock->sin6_addr),
463 dstsock->sin6_scope_id); /* for debug */
464 }
465 #endif
466
467 /* If the caller specify the outgoing interface explicitly, use it. */
468 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
469 /* XXX boundary check is assumed to be already done. */
470 ifp = ifnet_byindex(pi->ipi6_ifindex);
471 if (ifp != NULL &&
472 (norouteok || retrt == NULL ||
473 IN6_IS_ADDR_MULTICAST(dst))) {
474 /*
475 * we do not have to check or get the route for
476 * multicast.
477 */
478 goto done;
479 } else
480 goto getroute;
481 }
482
483 /*
484 * If the destination address is a multicast address and the outgoing
485 * interface for the address is specified by the caller, use it.
486 */
487 if (IN6_IS_ADDR_MULTICAST(dst) &&
488 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
489 goto done; /* we do not need a route for multicast. */
490 }
491
492 getroute:
493 /*
494 * If the next hop address for the packet is specified by the caller,
495 * use it as the gateway.
496 */
497 if (opts && opts->ip6po_nexthop) {
498 struct route_in6 *ron;
499
500 sin6_next = satosin6(opts->ip6po_nexthop);
501
502 /* at this moment, we only support AF_INET6 next hops */
503 if (sin6_next->sin6_family != AF_INET6) {
504 error = EAFNOSUPPORT; /* or should we proceed? */
505 goto done;
506 }
507
508 /*
509 * If the next hop is an IPv6 address, then the node identified
510 * by that address must be a neighbor of the sending host.
511 */
512 ron = &opts->ip6po_nextroute;
513 if ((ron->ro_rt &&
514 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) !=
515 (RTF_UP | RTF_LLINFO)) ||
516 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr,
517 &sin6_next->sin6_addr)) {
518 if (ron->ro_rt) {
519 RTFREE(ron->ro_rt);
520 ron->ro_rt = NULL;
521 }
522 *satosin6(&ron->ro_dst) = *sin6_next;
523 }
524 if (ron->ro_rt == NULL) {
525 rtalloc((struct route *)ron); /* multi path case? */
526 if (ron->ro_rt == NULL ||
527 !(ron->ro_rt->rt_flags & RTF_LLINFO)) {
528 if (ron->ro_rt) {
529 RTFREE(ron->ro_rt);
530 ron->ro_rt = NULL;
531 }
532 error = EHOSTUNREACH;
533 goto done;
534 }
535 }
536 rt = ron->ro_rt;
537 ifp = rt->rt_ifp;
538
539 /*
540 * When cloning is required, try to allocate a route to the
541 * destination so that the caller can store path MTU
542 * information.
543 */
544 if (!clone)
545 goto done;
546 }
547
548 /*
549 * Use a cached route if it exists and is valid, else try to allocate
550 * a new one. Note that we should check the address family of the
551 * cached destination, in case of sharing the cache with IPv4.
552 */
553 if (ro) {
554 if (ro->ro_rt &&
555 (!(ro->ro_rt->rt_flags & RTF_UP) ||
556 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
557 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
558 dst))) {
559 RTFREE(ro->ro_rt);
560 ro->ro_rt = (struct rtentry *)NULL;
561 }
562 if (ro->ro_rt == (struct rtentry *)NULL) {
563 struct sockaddr_in6 *sa6;
564
565 /* No route yet, so try to acquire one */
566 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
567 sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
568 *sa6 = *dstsock;
569 sa6->sin6_scope_id = 0;
570
571 if (clone) {
572 rtalloc((struct route *)ro);
573 } else {
574 ro->ro_rt = rtalloc1(&((struct route *)ro)
575 ->ro_dst, 0, 0UL);
576 if (ro->ro_rt)
577 RT_UNLOCK(ro->ro_rt);
578 }
579 }
580
581 /*
582 * do not care about the result if we have the nexthop
583 * explicitly specified.
584 */
585 if (opts && opts->ip6po_nexthop)
586 goto done;
587
588 if (ro->ro_rt) {
589 ifp = ro->ro_rt->rt_ifp;
590
591 if (ifp == NULL) { /* can this really happen? */
592 RTFREE(ro->ro_rt);
593 ro->ro_rt = NULL;
594 }
595 }
596 if (ro->ro_rt == NULL)
597 error = EHOSTUNREACH;
598 rt = ro->ro_rt;
599
600 /*
601 * Check if the outgoing interface conflicts with
602 * the interface specified by ipi6_ifindex (if specified).
603 * Note that loopback interface is always okay.
604 * (this may happen when we are sending a packet to one of
605 * our own addresses.)
606 */
607 if (ifp && opts && opts->ip6po_pktinfo &&
608 opts->ip6po_pktinfo->ipi6_ifindex) {
609 if (!(ifp->if_flags & IFF_LOOPBACK) &&
610 ifp->if_index !=
611 opts->ip6po_pktinfo->ipi6_ifindex) {
612 error = EHOSTUNREACH;
613 goto done;
614 }
615 }
616 }
617
618 done:
619 if (ifp == NULL && rt == NULL) {
620 /*
621 * This can happen if the caller did not pass a cached route
622 * nor any other hints. We treat this case an error.
623 */
624 error = EHOSTUNREACH;
625 }
626 if (error == EHOSTUNREACH)
627 ip6stat.ip6s_noroute++;
628
629 if (retifp != NULL)
630 *retifp = ifp;
631 if (retrt != NULL)
632 *retrt = rt; /* rt may be NULL */
633
634 return (error);
635 }
636
637 static int
638 in6_selectif(dstsock, opts, mopts, ro, retifp)
639 struct sockaddr_in6 *dstsock;
640 struct ip6_pktopts *opts;
641 struct ip6_moptions *mopts;
642 struct route_in6 *ro;
643 struct ifnet **retifp;
644 {
645 int error;
646 struct route_in6 sro;
647 struct rtentry *rt = NULL;
648
649 if (ro == NULL) {
650 bzero(&sro, sizeof(sro));
651 ro = &sro;
652 }
653
654 if ((error = selectroute(dstsock, opts, mopts, ro, retifp,
655 &rt, 0, 1)) != 0) {
656 if (rt && rt == sro.ro_rt)
657 RTFREE(rt);
658 return (error);
659 }
660
661 /*
662 * do not use a rejected or black hole route.
663 * XXX: this check should be done in the L2 output routine.
664 * However, if we skipped this check here, we'd see the following
665 * scenario:
666 * - install a rejected route for a scoped address prefix
667 * (like fe80::/10)
668 * - send a packet to a destination that matches the scoped prefix,
669 * with ambiguity about the scope zone.
670 * - pick the outgoing interface from the route, and disambiguate the
671 * scope zone with the interface.
672 * - ip6_output() would try to get another route with the "new"
673 * destination, which may be valid.
674 * - we'd see no error on output.
675 * Although this may not be very harmful, it should still be confusing.
676 * We thus reject the case here.
677 */
678 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
679 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
680
681 if (rt && rt == sro.ro_rt)
682 RTFREE(rt);
683 return (flags);
684 }
685
686 /*
687 * Adjust the "outgoing" interface. If we're going to loop the packet
688 * back to ourselves, the ifp would be the loopback interface.
689 * However, we'd rather know the interface associated to the
690 * destination address (which should probably be one of our own
691 * addresses.)
692 */
693 if (rt && rt->rt_ifa && rt->rt_ifa->ifa_ifp)
694 *retifp = rt->rt_ifa->ifa_ifp;
695
696 if (rt && rt == sro.ro_rt)
697 RTFREE(rt);
698 return (0);
699 }
700
701 int
702 in6_selectroute(dstsock, opts, mopts, ro, retifp, retrt, clone)
703 struct sockaddr_in6 *dstsock;
704 struct ip6_pktopts *opts;
705 struct ip6_moptions *mopts;
706 struct route_in6 *ro;
707 struct ifnet **retifp;
708 struct rtentry **retrt;
709 int clone; /* meaningful only for bsdi and freebsd. */
710 {
711 return (selectroute(dstsock, opts, mopts, ro, retifp,
712 retrt, clone, 0));
713 }
714
715 /*
716 * Default hop limit selection. The precedence is as follows:
717 * 1. Hoplimit value specified via ioctl.
718 * 2. (If the outgoing interface is detected) the current
719 * hop limit of the interface specified by router advertisement.
720 * 3. The system default hoplimit.
721 */
722 int
723 in6_selecthlim(in6p, ifp)
724 struct in6pcb *in6p;
725 struct ifnet *ifp;
726 {
727 if (in6p && in6p->in6p_hops >= 0)
728 return (in6p->in6p_hops);
729 else if (ifp)
730 return (ND_IFINFO(ifp)->chlim);
731 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
732 struct route_in6 ro6;
733 struct ifnet *lifp;
734
735 bzero(&ro6, sizeof(ro6));
736 ro6.ro_dst.sin6_family = AF_INET6;
737 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
738 ro6.ro_dst.sin6_addr = in6p->in6p_faddr;
739 rtalloc((struct route *)&ro6);
740 if (ro6.ro_rt) {
741 lifp = ro6.ro_rt->rt_ifp;
742 RTFREE(ro6.ro_rt);
743 if (lifp)
744 return (ND_IFINFO(lifp)->chlim);
745 } else
746 return (ip6_defhlim);
747 }
748 return (ip6_defhlim);
749 }
750
751 /*
752 * XXX: this is borrowed from in6_pcbbind(). If possible, we should
753 * share this function by all *bsd*...
754 */
755 int
756 in6_pcbsetport(laddr, inp, cred)
757 struct in6_addr *laddr;
758 struct inpcb *inp;
759 struct ucred *cred;
760 {
761 struct socket *so = inp->inp_socket;
762 u_int16_t lport = 0, first, last, *lastport;
763 int count, error = 0, wild = 0;
764 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
765
766 /* XXX: this is redundant when called from in6_pcbbind */
767 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
768 wild = INPLOOKUP_WILDCARD;
769
770 inp->inp_flags |= INP_ANONPORT;
771
772 if (inp->inp_flags & INP_HIGHPORT) {
773 first = ipport_hifirstauto; /* sysctl */
774 last = ipport_hilastauto;
775 lastport = &pcbinfo->lasthi;
776 } else if (inp->inp_flags & INP_LOWPORT) {
777 if ((error = suser_cred(cred, 0)))
778 return error;
779 first = ipport_lowfirstauto; /* 1023 */
780 last = ipport_lowlastauto; /* 600 */
781 lastport = &pcbinfo->lastlow;
782 } else {
783 first = ipport_firstauto; /* sysctl */
784 last = ipport_lastauto;
785 lastport = &pcbinfo->lastport;
786 }
787 /*
788 * Simple check to ensure all ports are not used up causing
789 * a deadlock here.
790 *
791 * We split the two cases (up and down) so that the direction
792 * is not being tested on each round of the loop.
793 */
794 if (first > last) {
795 /*
796 * counting down
797 */
798 count = first - last;
799
800 do {
801 if (count-- < 0) { /* completely used? */
802 /*
803 * Undo any address bind that may have
804 * occurred above.
805 */
806 inp->in6p_laddr = in6addr_any;
807 return (EAGAIN);
808 }
809 --*lastport;
810 if (*lastport > first || *lastport < last)
811 *lastport = first;
812 lport = htons(*lastport);
813 } while (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr,
814 lport, wild));
815 } else {
816 /*
817 * counting up
818 */
819 count = last - first;
820
821 do {
822 if (count-- < 0) { /* completely used? */
823 /*
824 * Undo any address bind that may have
825 * occurred above.
826 */
827 inp->in6p_laddr = in6addr_any;
828 return (EAGAIN);
829 }
830 ++*lastport;
831 if (*lastport < first || *lastport > last)
832 *lastport = first;
833 lport = htons(*lastport);
834 } while (in6_pcblookup_local(pcbinfo,
835 &inp->in6p_laddr, lport, wild));
836 }
837
838 inp->inp_lport = lport;
839 if (in_pcbinshash(inp) != 0) {
840 inp->in6p_laddr = in6addr_any;
841 inp->inp_lport = 0;
842 return (EAGAIN);
843 }
844
845 return (0);
846 }
847
848 void
849 addrsel_policy_init()
850 {
851 ADDRSEL_LOCK_INIT();
852 ADDRSEL_SXLOCK_INIT();
853
854 init_policy_queue();
855
856 /* initialize the "last resort" policy */
857 bzero(&defaultaddrpolicy, sizeof(defaultaddrpolicy));
858 defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
859 }
860
861 static struct in6_addrpolicy *
862 lookup_addrsel_policy(key)
863 struct sockaddr_in6 *key;
864 {
865 struct in6_addrpolicy *match = NULL;
866
867 ADDRSEL_LOCK();
868 match = match_addrsel_policy(key);
869
870 if (match == NULL)
871 match = &defaultaddrpolicy;
872 else
873 match->use++;
874 ADDRSEL_UNLOCK();
875
876 return (match);
877 }
878
879 /*
880 * Subroutines to manage the address selection policy table via sysctl.
881 */
882 struct walkarg {
883 struct sysctl_req *w_req;
884 };
885
886 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
887 SYSCTL_DECL(_net_inet6_ip6);
888 SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
889 CTLFLAG_RD, in6_src_sysctl, "");
890
891 static int
892 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
893 {
894 struct walkarg w;
895
896 if (req->newptr)
897 return EPERM;
898
899 bzero(&w, sizeof(w));
900 w.w_req = req;
901
902 return (walk_addrsel_policy(dump_addrsel_policyent, &w));
903 }
904
905 int
906 in6_src_ioctl(cmd, data)
907 u_long cmd;
908 caddr_t data;
909 {
910 int i;
911 struct in6_addrpolicy ent0;
912
913 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
914 return (EOPNOTSUPP); /* check for safety */
915
916 ent0 = *(struct in6_addrpolicy *)data;
917
918 if (ent0.label == ADDR_LABEL_NOTAPP)
919 return (EINVAL);
920 /* check if the prefix mask is consecutive. */
921 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
922 return (EINVAL);
923 /* clear trailing garbages (if any) of the prefix address. */
924 for (i = 0; i < 4; i++) {
925 ent0.addr.sin6_addr.s6_addr32[i] &=
926 ent0.addrmask.sin6_addr.s6_addr32[i];
927 }
928 ent0.use = 0;
929
930 switch (cmd) {
931 case SIOCAADDRCTL_POLICY:
932 return (add_addrsel_policyent(&ent0));
933 case SIOCDADDRCTL_POLICY:
934 return (delete_addrsel_policyent(&ent0));
935 }
936
937 return (0); /* XXX: compromise compilers */
938 }
939
940 /*
941 * The followings are implementation of the policy table using a
942 * simple tail queue.
943 * XXX such details should be hidden.
944 * XXX implementation using binary tree should be more efficient.
945 */
946 struct addrsel_policyent {
947 TAILQ_ENTRY(addrsel_policyent) ape_entry;
948 struct in6_addrpolicy ape_policy;
949 };
950
951 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
952
953 struct addrsel_policyhead addrsel_policytab;
954
955 static void
956 init_policy_queue()
957 {
958 TAILQ_INIT(&addrsel_policytab);
959 }
960
961 static int
962 add_addrsel_policyent(newpolicy)
963 struct in6_addrpolicy *newpolicy;
964 {
965 struct addrsel_policyent *new, *pol;
966
967 MALLOC(new, struct addrsel_policyent *, sizeof(*new), M_IFADDR,
968 M_WAITOK);
969 ADDRSEL_XLOCK();
970 ADDRSEL_LOCK();
971
972 /* duplication check */
973 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
974 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
975 &pol->ape_policy.addr.sin6_addr) &&
976 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
977 &pol->ape_policy.addrmask.sin6_addr)) {
978 ADDRSEL_UNLOCK();
979 ADDRSEL_XUNLOCK();
980 FREE(new, M_IFADDR);
981 return (EEXIST); /* or override it? */
982 }
983 }
984
985 bzero(new, sizeof(*new));
986
987 /* XXX: should validate entry */
988 new->ape_policy = *newpolicy;
989
990 TAILQ_INSERT_TAIL(&addrsel_policytab, new, ape_entry);
991 ADDRSEL_UNLOCK();
992 ADDRSEL_XUNLOCK();
993
994 return (0);
995 }
996
997 static int
998 delete_addrsel_policyent(key)
999 struct in6_addrpolicy *key;
1000 {
1001 struct addrsel_policyent *pol;
1002
1003 ADDRSEL_XLOCK();
1004 ADDRSEL_LOCK();
1005
1006 /* search for the entry in the table */
1007 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
1008 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1009 &pol->ape_policy.addr.sin6_addr) &&
1010 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1011 &pol->ape_policy.addrmask.sin6_addr)) {
1012 break;
1013 }
1014 }
1015 if (pol == NULL) {
1016 ADDRSEL_UNLOCK();
1017 ADDRSEL_XUNLOCK();
1018 return (ESRCH);
1019 }
1020
1021 TAILQ_REMOVE(&addrsel_policytab, pol, ape_entry);
1022 ADDRSEL_UNLOCK();
1023 ADDRSEL_XUNLOCK();
1024
1025 return (0);
1026 }
1027
1028 static int
1029 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1030 {
1031 struct addrsel_policyent *pol;
1032 int error = 0;
1033
1034 ADDRSEL_SLOCK();
1035 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
1036 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1037 ADDRSEL_SUNLOCK();
1038 return (error);
1039 }
1040 }
1041 ADDRSEL_SUNLOCK();
1042 return (error);
1043 }
1044
1045 static int
1046 dump_addrsel_policyent(pol, arg)
1047 struct in6_addrpolicy *pol;
1048 void *arg;
1049 {
1050 int error = 0;
1051 struct walkarg *w = arg;
1052
1053 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1054
1055 return (error);
1056 }
1057
1058 static struct in6_addrpolicy *
1059 match_addrsel_policy(key)
1060 struct sockaddr_in6 *key;
1061 {
1062 struct addrsel_policyent *pent;
1063 struct in6_addrpolicy *bestpol = NULL, *pol;
1064 int matchlen, bestmatchlen = -1;
1065 u_char *mp, *ep, *k, *p, m;
1066
1067 TAILQ_FOREACH(pent, &addrsel_policytab, ape_entry) {
1068 matchlen = 0;
1069
1070 pol = &pent->ape_policy;
1071 mp = (u_char *)&pol->addrmask.sin6_addr;
1072 ep = mp + 16; /* XXX: scope field? */
1073 k = (u_char *)&key->sin6_addr;
1074 p = (u_char *)&pol->addr.sin6_addr;
1075 for (; mp < ep && *mp; mp++, k++, p++) {
1076 m = *mp;
1077 if ((*k & m) != *p)
1078 goto next; /* not match */
1079 if (m == 0xff) /* short cut for a typical case */
1080 matchlen += 8;
1081 else {
1082 while (m >= 0x80) {
1083 matchlen++;
1084 m <<= 1;
1085 }
1086 }
1087 }
1088
1089 /* matched. check if this is better than the current best. */
1090 if (bestpol == NULL ||
1091 matchlen > bestmatchlen) {
1092 bestpol = pol;
1093 bestmatchlen = matchlen;
1094 }
1095
1096 next:
1097 continue;
1098 }
1099
1100 return (bestpol);
1101 }
Cache object: 120c80f78dd647550901ad02743d0e68
|