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/lock.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/priv.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/sockio.h>
79 #include <sys/sysctl.h>
80 #include <sys/errno.h>
81 #include <sys/time.h>
82 #include <sys/jail.h>
83 #include <sys/kernel.h>
84 #include <sys/sx.h>
85
86 #include <net/if.h>
87 #include <net/route.h>
88
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/in_systm.h>
92 #include <netinet/ip.h>
93 #include <netinet/in_pcb.h>
94 #include <netinet/ip_var.h>
95 #include <netinet/udp.h>
96 #include <netinet/udp_var.h>
97 #include <netinet6/in6_var.h>
98 #include <netinet/ip6.h>
99 #include <netinet6/in6_pcb.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet6/scope6_var.h>
102 #include <netinet6/nd6.h>
103
104 static struct mtx addrsel_lock;
105 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
106 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock)
107 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock)
108 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED)
109
110 static struct sx addrsel_sxlock;
111 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock")
112 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock)
113 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock)
114 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock)
115 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock)
116
117 #define ADDR_LABEL_NOTAPP (-1)
118 struct in6_addrpolicy defaultaddrpolicy;
119
120 int ip6_prefer_tempaddr = 0;
121
122 static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *,
123 struct ip6_moptions *, struct route_in6 *, struct ifnet **,
124 struct rtentry **, int, int));
125 static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *,
126 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **));
127
128 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
129
130 static void init_policy_queue(void);
131 static int add_addrsel_policyent(struct in6_addrpolicy *);
132 static int delete_addrsel_policyent(struct in6_addrpolicy *);
133 static int walk_addrsel_policy __P((int (*)(struct in6_addrpolicy *, void *),
134 void *));
135 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
136 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
137
138 /*
139 * Return an IPv6 address, which is the most appropriate for a given
140 * destination and user specified options.
141 * If necessary, this function lookups the routing table and returns
142 * an entry to the caller for later use.
143 */
144 #define REPLACE(r) do {\
145 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
146 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
147 ip6stat.ip6s_sources_rule[(r)]++; \
148 /* { \
149 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
150 printf("in6_selectsrc: replace %s with %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \
151 } */ \
152 goto replace; \
153 } while(0)
154 #define NEXT(r) do {\
155 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
156 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
157 ip6stat.ip6s_sources_rule[(r)]++; \
158 /* { \
159 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
160 printf("in6_selectsrc: keep %s against %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \
161 } */ \
162 goto next; /* XXX: we can't use 'continue' here */ \
163 } while(0)
164 #define BREAK(r) do { \
165 if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
166 sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
167 ip6stat.ip6s_sources_rule[(r)]++; \
168 goto out; /* XXX: we can't use 'break' here */ \
169 } while(0)
170
171 struct in6_addr *
172 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
173 struct inpcb *inp, struct route_in6 *ro, struct ucred *cred,
174 struct ifnet **ifpp, int *errorp)
175 {
176 struct in6_addr dst;
177 struct ifnet *ifp = NULL;
178 struct in6_ifaddr *ia = NULL, *ia_best = NULL;
179 struct in6_pktinfo *pi = NULL;
180 int dst_scope = -1, best_scope = -1, best_matchlen = -1;
181 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
182 u_int32_t odstzone;
183 int prefer_tempaddr;
184 struct ip6_moptions *mopts;
185
186 dst = dstsock->sin6_addr; /* make a copy for local operation */
187 *errorp = 0;
188 if (ifpp)
189 *ifpp = NULL;
190
191 if (inp != NULL) {
192 INP_LOCK_ASSERT(inp);
193 mopts = inp->in6p_moptions;
194 } else {
195 mopts = NULL;
196 }
197
198 /*
199 * If the source address is explicitly specified by the caller,
200 * check if the requested source address is indeed a unicast address
201 * assigned to the node, and can be used as the packet's source
202 * address. If everything is okay, use the address as source.
203 */
204 if (opts && (pi = opts->ip6po_pktinfo) &&
205 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
206 struct sockaddr_in6 srcsock;
207 struct in6_ifaddr *ia6;
208
209 /* get the outgoing interface */
210 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp))
211 != 0) {
212 return (NULL);
213 }
214
215 /*
216 * determine the appropriate zone id of the source based on
217 * the zone of the destination and the outgoing interface.
218 * If the specified address is ambiguous wrt the scope zone,
219 * the interface must be specified; otherwise, ifa_ifwithaddr()
220 * will fail matching the address.
221 */
222 bzero(&srcsock, sizeof(srcsock));
223 srcsock.sin6_family = AF_INET6;
224 srcsock.sin6_len = sizeof(srcsock);
225 srcsock.sin6_addr = pi->ipi6_addr;
226 if (ifp) {
227 *errorp = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
228 if (*errorp != 0)
229 return (NULL);
230 }
231 if (cred != NULL && (*errorp = prison_local_ip6(cred,
232 &srcsock.sin6_addr, (inp != NULL &&
233 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
234 return (NULL);
235
236 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)(&srcsock));
237 if (ia6 == NULL ||
238 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) {
239 *errorp = EADDRNOTAVAIL;
240 return (NULL);
241 }
242 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
243 if (ifpp)
244 *ifpp = ifp;
245 return (&ia6->ia_addr.sin6_addr);
246 }
247
248 /*
249 * Otherwise, if the socket has already bound the source, just use it.
250 */
251 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
252 if (cred != NULL &&
253 (*errorp = prison_local_ip6(cred, &inp->in6p_laddr,
254 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
255 return (NULL);
256 return (&inp->in6p_laddr);
257 }
258
259 /*
260 * If the address is not specified, choose the best one based on
261 * the outgoing interface and the destination address.
262 */
263 /* get the outgoing interface */
264 if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0)
265 return (NULL);
266
267 #ifdef DIAGNOSTIC
268 if (ifp == NULL) /* this should not happen */
269 panic("in6_selectsrc: NULL ifp");
270 #endif
271 *errorp = in6_setscope(&dst, ifp, &odstzone);
272 if (*errorp != 0)
273 return (NULL);
274
275 for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
276 int new_scope = -1, new_matchlen = -1;
277 struct in6_addrpolicy *new_policy = NULL;
278 u_int32_t srczone, osrczone, dstzone;
279 struct in6_addr src;
280 struct ifnet *ifp1 = ia->ia_ifp;
281
282 /*
283 * We'll never take an address that breaks the scope zone
284 * of the destination. We also skip an address if its zone
285 * does not contain the outgoing interface.
286 * XXX: we should probably use sin6_scope_id here.
287 */
288 if (in6_setscope(&dst, ifp1, &dstzone) ||
289 odstzone != dstzone) {
290 continue;
291 }
292 src = ia->ia_addr.sin6_addr;
293 if (in6_setscope(&src, ifp, &osrczone) ||
294 in6_setscope(&src, ifp1, &srczone) ||
295 osrczone != srczone) {
296 continue;
297 }
298
299 /* avoid unusable addresses */
300 if ((ia->ia6_flags &
301 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
302 continue;
303 }
304 if (!ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
305 continue;
306
307 if (cred != NULL &&
308 prison_local_ip6(cred, &ia->ia_addr.sin6_addr,
309 (inp != NULL &&
310 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0)
311 continue;
312
313 /* Rule 1: Prefer same address */
314 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
315 ia_best = ia;
316 BREAK(1); /* there should be no better candidate */
317 }
318
319 if (ia_best == NULL)
320 REPLACE(0);
321
322 /* Rule 2: Prefer appropriate scope */
323 if (dst_scope < 0)
324 dst_scope = in6_addrscope(&dst);
325 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
326 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
327 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
328 REPLACE(2);
329 NEXT(2);
330 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
331 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
332 NEXT(2);
333 REPLACE(2);
334 }
335
336 /*
337 * Rule 3: Avoid deprecated addresses. Note that the case of
338 * !ip6_use_deprecated is already rejected above.
339 */
340 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
341 NEXT(3);
342 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
343 REPLACE(3);
344
345 /* Rule 4: Prefer home addresses */
346 /*
347 * XXX: This is a TODO. We should probably merge the MIP6
348 * case above.
349 */
350
351 /* Rule 5: Prefer outgoing interface */
352 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
353 NEXT(5);
354 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
355 REPLACE(5);
356
357 /*
358 * Rule 6: Prefer matching label
359 * Note that best_policy should be non-NULL here.
360 */
361 if (dst_policy == NULL)
362 dst_policy = lookup_addrsel_policy(dstsock);
363 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
364 new_policy = lookup_addrsel_policy(&ia->ia_addr);
365 if (dst_policy->label == best_policy->label &&
366 dst_policy->label != new_policy->label)
367 NEXT(6);
368 if (dst_policy->label != best_policy->label &&
369 dst_policy->label == new_policy->label)
370 REPLACE(6);
371 }
372
373 /*
374 * Rule 7: Prefer public addresses.
375 * We allow users to reverse the logic by configuring
376 * a sysctl variable, so that privacy conscious users can
377 * always prefer temporary addresses.
378 */
379 if (opts == NULL ||
380 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
381 prefer_tempaddr = ip6_prefer_tempaddr;
382 } else if (opts->ip6po_prefer_tempaddr ==
383 IP6PO_TEMPADDR_NOTPREFER) {
384 prefer_tempaddr = 0;
385 } else
386 prefer_tempaddr = 1;
387 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
388 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
389 if (prefer_tempaddr)
390 REPLACE(7);
391 else
392 NEXT(7);
393 }
394 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
395 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
396 if (prefer_tempaddr)
397 NEXT(7);
398 else
399 REPLACE(7);
400 }
401
402 /*
403 * Rule 8: prefer addresses on alive interfaces.
404 * This is a KAME specific rule.
405 */
406 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
407 !(ia->ia_ifp->if_flags & IFF_UP))
408 NEXT(8);
409 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
410 (ia->ia_ifp->if_flags & IFF_UP))
411 REPLACE(8);
412
413 /*
414 * Rule 14: Use longest matching prefix.
415 * Note: in the address selection draft, this rule is
416 * documented as "Rule 8". However, since it is also
417 * documented that this rule can be overridden, we assign
418 * a large number so that it is easy to assign smaller numbers
419 * to more preferred rules.
420 */
421 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
422 if (best_matchlen < new_matchlen)
423 REPLACE(14);
424 if (new_matchlen < best_matchlen)
425 NEXT(14);
426
427 /* Rule 15 is reserved. */
428
429 /*
430 * Last resort: just keep the current candidate.
431 * Or, do we need more rules?
432 */
433 continue;
434
435 replace:
436 ia_best = ia;
437 best_scope = (new_scope >= 0 ? new_scope :
438 in6_addrscope(&ia_best->ia_addr.sin6_addr));
439 best_policy = (new_policy ? new_policy :
440 lookup_addrsel_policy(&ia_best->ia_addr));
441 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
442 in6_matchlen(&ia_best->ia_addr.sin6_addr,
443 &dst));
444
445 next:
446 continue;
447
448 out:
449 break;
450 }
451
452 if ((ia = ia_best) == NULL) {
453 *errorp = EADDRNOTAVAIL;
454 return (NULL);
455 }
456
457 if (ifpp)
458 *ifpp = ifp;
459
460 return (&ia->ia_addr.sin6_addr);
461 }
462
463 /*
464 * clone - meaningful only for bsdi and freebsd
465 */
466 static int
467 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
468 struct ip6_moptions *mopts, struct route_in6 *ro,
469 struct ifnet **retifp, struct rtentry **retrt, int clone,
470 int norouteok)
471 {
472 int error = 0;
473 struct ifnet *ifp = NULL;
474 struct rtentry *rt = NULL;
475 struct sockaddr_in6 *sin6_next;
476 struct in6_pktinfo *pi = NULL;
477 struct in6_addr *dst = &dstsock->sin6_addr;
478 #if 0
479 char ip6buf[INET6_ADDRSTRLEN];
480
481 if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
482 dstsock->sin6_addr.s6_addr32[1] == 0 &&
483 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
484 printf("in6_selectroute: strange destination %s\n",
485 ip6_sprintf(ip6buf, &dstsock->sin6_addr));
486 } else {
487 printf("in6_selectroute: destination = %s%%%d\n",
488 ip6_sprintf(ip6buf, &dstsock->sin6_addr),
489 dstsock->sin6_scope_id); /* for debug */
490 }
491 #endif
492
493 /* If the caller specify the outgoing interface explicitly, use it. */
494 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
495 /* XXX boundary check is assumed to be already done. */
496 ifp = ifnet_byindex(pi->ipi6_ifindex);
497 if (ifp != NULL &&
498 (norouteok || retrt == NULL ||
499 IN6_IS_ADDR_MULTICAST(dst))) {
500 /*
501 * we do not have to check or get the route for
502 * multicast.
503 */
504 goto done;
505 } else
506 goto getroute;
507 }
508
509 /*
510 * If the destination address is a multicast address and the outgoing
511 * interface for the address is specified by the caller, use it.
512 */
513 if (IN6_IS_ADDR_MULTICAST(dst) &&
514 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
515 goto done; /* we do not need a route for multicast. */
516 }
517
518 getroute:
519 /*
520 * If the next hop address for the packet is specified by the caller,
521 * use it as the gateway.
522 */
523 if (opts && opts->ip6po_nexthop) {
524 struct route_in6 *ron;
525
526 sin6_next = satosin6(opts->ip6po_nexthop);
527
528 /* at this moment, we only support AF_INET6 next hops */
529 if (sin6_next->sin6_family != AF_INET6) {
530 error = EAFNOSUPPORT; /* or should we proceed? */
531 goto done;
532 }
533
534 /*
535 * If the next hop is an IPv6 address, then the node identified
536 * by that address must be a neighbor of the sending host.
537 */
538 ron = &opts->ip6po_nextroute;
539 if ((ron->ro_rt &&
540 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) !=
541 (RTF_UP | RTF_LLINFO)) ||
542 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr,
543 &sin6_next->sin6_addr)) {
544 if (ron->ro_rt) {
545 RTFREE(ron->ro_rt);
546 ron->ro_rt = NULL;
547 }
548 *satosin6(&ron->ro_dst) = *sin6_next;
549 }
550 if (ron->ro_rt == NULL) {
551 rtalloc((struct route *)ron); /* multi path case? */
552 if (ron->ro_rt == NULL ||
553 !(ron->ro_rt->rt_flags & RTF_LLINFO)) {
554 if (ron->ro_rt) {
555 RTFREE(ron->ro_rt);
556 ron->ro_rt = NULL;
557 }
558 error = EHOSTUNREACH;
559 goto done;
560 }
561 }
562 rt = ron->ro_rt;
563 ifp = rt->rt_ifp;
564
565 /*
566 * When cloning is required, try to allocate a route to the
567 * destination so that the caller can store path MTU
568 * information.
569 */
570 if (!clone)
571 goto done;
572 }
573
574 /*
575 * Use a cached route if it exists and is valid, else try to allocate
576 * a new one. Note that we should check the address family of the
577 * cached destination, in case of sharing the cache with IPv4.
578 */
579 if (ro) {
580 if (ro->ro_rt &&
581 (!(ro->ro_rt->rt_flags & RTF_UP) ||
582 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
583 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
584 dst))) {
585 RTFREE(ro->ro_rt);
586 ro->ro_rt = (struct rtentry *)NULL;
587 }
588 if (ro->ro_rt == (struct rtentry *)NULL) {
589 struct sockaddr_in6 *sa6;
590
591 /* No route yet, so try to acquire one */
592 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
593 sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
594 *sa6 = *dstsock;
595 sa6->sin6_scope_id = 0;
596
597 if (clone) {
598 rtalloc((struct route *)ro);
599 } else {
600 ro->ro_rt = rtalloc1(&((struct route *)ro)
601 ->ro_dst, 0, 0UL);
602 if (ro->ro_rt)
603 RT_UNLOCK(ro->ro_rt);
604 }
605 }
606
607 /*
608 * do not care about the result if we have the nexthop
609 * explicitly specified.
610 */
611 if (opts && opts->ip6po_nexthop)
612 goto done;
613
614 if (ro->ro_rt) {
615 ifp = ro->ro_rt->rt_ifp;
616
617 if (ifp == NULL) { /* can this really happen? */
618 RTFREE(ro->ro_rt);
619 ro->ro_rt = NULL;
620 }
621 }
622 if (ro->ro_rt == NULL)
623 error = EHOSTUNREACH;
624 rt = ro->ro_rt;
625
626 /*
627 * Check if the outgoing interface conflicts with
628 * the interface specified by ipi6_ifindex (if specified).
629 * Note that loopback interface is always okay.
630 * (this may happen when we are sending a packet to one of
631 * our own addresses.)
632 */
633 if (ifp && opts && opts->ip6po_pktinfo &&
634 opts->ip6po_pktinfo->ipi6_ifindex) {
635 if (!(ifp->if_flags & IFF_LOOPBACK) &&
636 ifp->if_index !=
637 opts->ip6po_pktinfo->ipi6_ifindex) {
638 error = EHOSTUNREACH;
639 goto done;
640 }
641 }
642 }
643
644 done:
645 if (ifp == NULL && rt == NULL) {
646 /*
647 * This can happen if the caller did not pass a cached route
648 * nor any other hints. We treat this case an error.
649 */
650 error = EHOSTUNREACH;
651 }
652 if (error == EHOSTUNREACH)
653 ip6stat.ip6s_noroute++;
654
655 if (retifp != NULL)
656 *retifp = ifp;
657 if (retrt != NULL)
658 *retrt = rt; /* rt may be NULL */
659
660 return (error);
661 }
662
663 static int
664 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
665 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp)
666 {
667 int error;
668 struct route_in6 sro;
669 struct rtentry *rt = NULL;
670
671 if (ro == NULL) {
672 bzero(&sro, sizeof(sro));
673 ro = &sro;
674 }
675
676 if ((error = selectroute(dstsock, opts, mopts, ro, retifp,
677 &rt, 0, 1)) != 0) {
678 if (ro == &sro && rt && rt == sro.ro_rt)
679 RTFREE(rt);
680 return (error);
681 }
682
683 /*
684 * do not use a rejected or black hole route.
685 * XXX: this check should be done in the L2 output routine.
686 * However, if we skipped this check here, we'd see the following
687 * scenario:
688 * - install a rejected route for a scoped address prefix
689 * (like fe80::/10)
690 * - send a packet to a destination that matches the scoped prefix,
691 * with ambiguity about the scope zone.
692 * - pick the outgoing interface from the route, and disambiguate the
693 * scope zone with the interface.
694 * - ip6_output() would try to get another route with the "new"
695 * destination, which may be valid.
696 * - we'd see no error on output.
697 * Although this may not be very harmful, it should still be confusing.
698 * We thus reject the case here.
699 */
700 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
701 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
702
703 if (ro == &sro && rt && rt == sro.ro_rt)
704 RTFREE(rt);
705 return (flags);
706 }
707
708 /*
709 * Adjust the "outgoing" interface. If we're going to loop the packet
710 * back to ourselves, the ifp would be the loopback interface.
711 * However, we'd rather know the interface associated to the
712 * destination address (which should probably be one of our own
713 * addresses.)
714 */
715 if (rt && rt->rt_ifa && rt->rt_ifa->ifa_ifp)
716 *retifp = rt->rt_ifa->ifa_ifp;
717
718 if (ro == &sro && rt && rt == sro.ro_rt)
719 RTFREE(rt);
720 return (0);
721 }
722
723 /*
724 * clone - meaningful only for bsdi and freebsd
725 */
726 int
727 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
728 struct ip6_moptions *mopts, struct route_in6 *ro,
729 struct ifnet **retifp, struct rtentry **retrt, int clone)
730 {
731
732 return (selectroute(dstsock, opts, mopts, ro, retifp,
733 retrt, clone, 0));
734 }
735
736 /*
737 * Default hop limit selection. The precedence is as follows:
738 * 1. Hoplimit value specified via ioctl.
739 * 2. (If the outgoing interface is detected) the current
740 * hop limit of the interface specified by router advertisement.
741 * 3. The system default hoplimit.
742 */
743 int
744 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp)
745 {
746
747 if (in6p && in6p->in6p_hops >= 0)
748 return (in6p->in6p_hops);
749 else if (ifp)
750 return (ND_IFINFO(ifp)->chlim);
751 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
752 struct route_in6 ro6;
753 struct ifnet *lifp;
754
755 bzero(&ro6, sizeof(ro6));
756 ro6.ro_dst.sin6_family = AF_INET6;
757 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
758 ro6.ro_dst.sin6_addr = in6p->in6p_faddr;
759 rtalloc((struct route *)&ro6);
760 if (ro6.ro_rt) {
761 lifp = ro6.ro_rt->rt_ifp;
762 RTFREE(ro6.ro_rt);
763 if (lifp)
764 return (ND_IFINFO(lifp)->chlim);
765 } else
766 return (ip6_defhlim);
767 }
768 return (ip6_defhlim);
769 }
770
771 /*
772 * XXX: this is borrowed from in6_pcbbind(). If possible, we should
773 * share this function by all *bsd*...
774 */
775 int
776 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
777 {
778 struct socket *so = inp->inp_socket;
779 u_int16_t lport = 0, first, last, *lastport;
780 int count, error, wild = 0, dorandom;
781 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
782
783 INP_INFO_WLOCK_ASSERT(pcbinfo);
784 INP_WLOCK_ASSERT(inp);
785
786 error = prison_local_ip6(cred, laddr,
787 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
788 if (error)
789 return(error);
790
791 /* XXX: this is redundant when called from in6_pcbbind */
792 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
793 wild = INPLOOKUP_WILDCARD;
794
795 inp->inp_flags |= INP_ANONPORT;
796
797 if (inp->inp_flags & INP_HIGHPORT) {
798 first = ipport_hifirstauto; /* sysctl */
799 last = ipport_hilastauto;
800 lastport = &pcbinfo->ipi_lasthi;
801 } else if (inp->inp_flags & INP_LOWPORT) {
802 error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0);
803 if (error)
804 return error;
805 first = ipport_lowfirstauto; /* 1023 */
806 last = ipport_lowlastauto; /* 600 */
807 lastport = &pcbinfo->ipi_lastlow;
808 } else {
809 first = ipport_firstauto; /* sysctl */
810 last = ipport_lastauto;
811 lastport = &pcbinfo->ipi_lastport;
812 }
813
814 /*
815 * For UDP, use random port allocation as long as the user
816 * allows it. For TCP (and as of yet unknown) connections,
817 * use random port allocation only if the user allows it AND
818 * ipport_tick() allows it.
819 */
820 if (ipport_randomized &&
821 (!ipport_stoprandom || pcbinfo == &udbinfo))
822 dorandom = 1;
823 else
824 dorandom = 0;
825 /*
826 * It makes no sense to do random port allocation if
827 * we have the only port available.
828 */
829 if (first == last)
830 dorandom = 0;
831 /* Make sure to not include UDP packets in the count. */
832 if (pcbinfo != &udbinfo)
833 ipport_tcpallocs++;
834
835 /*
836 * Instead of having two loops further down counting up or down
837 * make sure that first is always <= last and go with only one
838 * code path implementing all logic.
839 */
840 if (first > last) {
841 u_int16_t aux;
842
843 aux = first;
844 first = last;
845 last = aux;
846 }
847
848 if (dorandom)
849 *lastport = first + (arc4random() % (last - first));
850
851 count = last - first;
852
853 do {
854 if (count-- < 0) { /* completely used? */
855 /* Undo an address bind that may have occurred. */
856 inp->in6p_laddr = in6addr_any;
857 return (EADDRNOTAVAIL);
858 }
859 ++*lastport;
860 if (*lastport < first || *lastport > last)
861 *lastport = first;
862 lport = htons(*lastport);
863 } while (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr,
864 lport, wild, cred));
865
866 inp->inp_lport = lport;
867 if (in_pcbinshash(inp) != 0) {
868 inp->in6p_laddr = in6addr_any;
869 inp->inp_lport = 0;
870 return (EAGAIN);
871 }
872
873 return (0);
874 }
875
876 void
877 addrsel_policy_init(void)
878 {
879 ADDRSEL_LOCK_INIT();
880 ADDRSEL_SXLOCK_INIT();
881
882 init_policy_queue();
883
884 /* initialize the "last resort" policy */
885 bzero(&defaultaddrpolicy, sizeof(defaultaddrpolicy));
886 defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
887 }
888
889 static struct in6_addrpolicy *
890 lookup_addrsel_policy(struct sockaddr_in6 *key)
891 {
892 struct in6_addrpolicy *match = NULL;
893
894 ADDRSEL_LOCK();
895 match = match_addrsel_policy(key);
896
897 if (match == NULL)
898 match = &defaultaddrpolicy;
899 else
900 match->use++;
901 ADDRSEL_UNLOCK();
902
903 return (match);
904 }
905
906 /*
907 * Subroutines to manage the address selection policy table via sysctl.
908 */
909 struct walkarg {
910 struct sysctl_req *w_req;
911 };
912
913 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
914 SYSCTL_DECL(_net_inet6_ip6);
915 SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
916 CTLFLAG_RD, in6_src_sysctl, "");
917
918 static int
919 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
920 {
921 struct walkarg w;
922
923 if (req->newptr)
924 return EPERM;
925
926 bzero(&w, sizeof(w));
927 w.w_req = req;
928
929 return (walk_addrsel_policy(dump_addrsel_policyent, &w));
930 }
931
932 int
933 in6_src_ioctl(u_long cmd, caddr_t data)
934 {
935 int i;
936 struct in6_addrpolicy ent0;
937
938 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
939 return (EOPNOTSUPP); /* check for safety */
940
941 ent0 = *(struct in6_addrpolicy *)data;
942
943 if (ent0.label == ADDR_LABEL_NOTAPP)
944 return (EINVAL);
945 /* check if the prefix mask is consecutive. */
946 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
947 return (EINVAL);
948 /* clear trailing garbages (if any) of the prefix address. */
949 for (i = 0; i < 4; i++) {
950 ent0.addr.sin6_addr.s6_addr32[i] &=
951 ent0.addrmask.sin6_addr.s6_addr32[i];
952 }
953 ent0.use = 0;
954
955 switch (cmd) {
956 case SIOCAADDRCTL_POLICY:
957 return (add_addrsel_policyent(&ent0));
958 case SIOCDADDRCTL_POLICY:
959 return (delete_addrsel_policyent(&ent0));
960 }
961
962 return (0); /* XXX: compromise compilers */
963 }
964
965 /*
966 * The followings are implementation of the policy table using a
967 * simple tail queue.
968 * XXX such details should be hidden.
969 * XXX implementation using binary tree should be more efficient.
970 */
971 struct addrsel_policyent {
972 TAILQ_ENTRY(addrsel_policyent) ape_entry;
973 struct in6_addrpolicy ape_policy;
974 };
975
976 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
977
978 struct addrsel_policyhead addrsel_policytab;
979
980 static void
981 init_policy_queue(void)
982 {
983
984 TAILQ_INIT(&addrsel_policytab);
985 }
986
987 static int
988 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
989 {
990 struct addrsel_policyent *new, *pol;
991
992 MALLOC(new, struct addrsel_policyent *, sizeof(*new), M_IFADDR,
993 M_WAITOK);
994 ADDRSEL_XLOCK();
995 ADDRSEL_LOCK();
996
997 /* duplication check */
998 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
999 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1000 &pol->ape_policy.addr.sin6_addr) &&
1001 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1002 &pol->ape_policy.addrmask.sin6_addr)) {
1003 ADDRSEL_UNLOCK();
1004 ADDRSEL_XUNLOCK();
1005 FREE(new, M_IFADDR);
1006 return (EEXIST); /* or override it? */
1007 }
1008 }
1009
1010 bzero(new, sizeof(*new));
1011
1012 /* XXX: should validate entry */
1013 new->ape_policy = *newpolicy;
1014
1015 TAILQ_INSERT_TAIL(&addrsel_policytab, new, ape_entry);
1016 ADDRSEL_UNLOCK();
1017 ADDRSEL_XUNLOCK();
1018
1019 return (0);
1020 }
1021
1022 static int
1023 delete_addrsel_policyent(struct in6_addrpolicy *key)
1024 {
1025 struct addrsel_policyent *pol;
1026
1027 ADDRSEL_XLOCK();
1028 ADDRSEL_LOCK();
1029
1030 /* search for the entry in the table */
1031 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
1032 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1033 &pol->ape_policy.addr.sin6_addr) &&
1034 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1035 &pol->ape_policy.addrmask.sin6_addr)) {
1036 break;
1037 }
1038 }
1039 if (pol == NULL) {
1040 ADDRSEL_UNLOCK();
1041 ADDRSEL_XUNLOCK();
1042 return (ESRCH);
1043 }
1044
1045 TAILQ_REMOVE(&addrsel_policytab, pol, ape_entry);
1046 ADDRSEL_UNLOCK();
1047 ADDRSEL_XUNLOCK();
1048
1049 return (0);
1050 }
1051
1052 static int
1053 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *),
1054 void *w)
1055 {
1056 struct addrsel_policyent *pol;
1057 int error = 0;
1058
1059 ADDRSEL_SLOCK();
1060 TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
1061 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1062 ADDRSEL_SUNLOCK();
1063 return (error);
1064 }
1065 }
1066 ADDRSEL_SUNLOCK();
1067 return (error);
1068 }
1069
1070 static int
1071 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1072 {
1073 int error = 0;
1074 struct walkarg *w = arg;
1075
1076 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1077
1078 return (error);
1079 }
1080
1081 static struct in6_addrpolicy *
1082 match_addrsel_policy(struct sockaddr_in6 *key)
1083 {
1084 struct addrsel_policyent *pent;
1085 struct in6_addrpolicy *bestpol = NULL, *pol;
1086 int matchlen, bestmatchlen = -1;
1087 u_char *mp, *ep, *k, *p, m;
1088
1089 TAILQ_FOREACH(pent, &addrsel_policytab, ape_entry) {
1090 matchlen = 0;
1091
1092 pol = &pent->ape_policy;
1093 mp = (u_char *)&pol->addrmask.sin6_addr;
1094 ep = mp + 16; /* XXX: scope field? */
1095 k = (u_char *)&key->sin6_addr;
1096 p = (u_char *)&pol->addr.sin6_addr;
1097 for (; mp < ep && *mp; mp++, k++, p++) {
1098 m = *mp;
1099 if ((*k & m) != *p)
1100 goto next; /* not match */
1101 if (m == 0xff) /* short cut for a typical case */
1102 matchlen += 8;
1103 else {
1104 while (m >= 0x80) {
1105 matchlen++;
1106 m <<= 1;
1107 }
1108 }
1109 }
1110
1111 /* matched. check if this is better than the current best. */
1112 if (bestpol == NULL ||
1113 matchlen > bestmatchlen) {
1114 bestpol = pol;
1115 bestmatchlen = matchlen;
1116 }
1117
1118 next:
1119 continue;
1120 }
1121
1122 return (bestpol);
1123 }
Cache object: eb5dc31eb6be39a9dc89adbd67f6ea63
|