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 #include "opt_mpath.h"
69
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/lock.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/priv.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/sockio.h>
80 #include <sys/sysctl.h>
81 #include <sys/errno.h>
82 #include <sys/time.h>
83 #include <sys/jail.h>
84 #include <sys/kernel.h>
85 #include <sys/sx.h>
86
87 #include <net/if.h>
88 #include <net/if_dl.h>
89 #include <net/route.h>
90 #include <net/if_llatbl.h>
91 #ifdef RADIX_MPATH
92 #include <net/radix_mpath.h>
93 #endif
94
95 #include <netinet/in.h>
96 #include <netinet/in_var.h>
97 #include <netinet/in_systm.h>
98 #include <netinet/ip.h>
99 #include <netinet/in_pcb.h>
100 #include <netinet/ip_var.h>
101 #include <netinet/udp.h>
102 #include <netinet/udp_var.h>
103
104 #include <netinet6/in6_var.h>
105 #include <netinet/ip6.h>
106 #include <netinet6/in6_pcb.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/scope6_var.h>
109 #include <netinet6/nd6.h>
110
111 static struct mtx addrsel_lock;
112 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
113 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock)
114 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock)
115 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED)
116
117 static struct sx addrsel_sxlock;
118 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock")
119 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock)
120 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock)
121 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock)
122 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock)
123
124 #define ADDR_LABEL_NOTAPP (-1)
125 static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy);
126 #define V_defaultaddrpolicy VNET(defaultaddrpolicy)
127
128 VNET_DEFINE(int, ip6_prefer_tempaddr) = 0;
129
130 static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *,
131 struct ip6_moptions *, struct route_in6 *, struct ifnet **,
132 struct rtentry **, int, u_int);
133 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
134 struct ip6_moptions *, struct route_in6 *ro, struct ifnet **,
135 struct ifnet *, u_int);
136
137 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
138
139 static void init_policy_queue(void);
140 static int add_addrsel_policyent(struct in6_addrpolicy *);
141 static int delete_addrsel_policyent(struct in6_addrpolicy *);
142 static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *),
143 void *);
144 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
145 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
146
147 /*
148 * Return an IPv6 address, which is the most appropriate for a given
149 * destination and user specified options.
150 * If necessary, this function lookups the routing table and returns
151 * an entry to the caller for later use.
152 */
153 #define REPLACE(r) do {\
154 IP6STAT_INC(ip6s_sources_rule[(r)]); \
155 rule = (r); \
156 /* { \
157 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
158 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)); \
159 } */ \
160 goto replace; \
161 } while(0)
162 #define NEXT(r) do {\
163 /* { \
164 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
165 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)); \
166 } */ \
167 goto next; /* XXX: we can't use 'continue' here */ \
168 } while(0)
169 #define BREAK(r) do { \
170 IP6STAT_INC(ip6s_sources_rule[(r)]); \
171 rule = (r); \
172 goto out; /* XXX: we can't use 'break' here */ \
173 } while(0)
174
175 int
176 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
177 struct inpcb *inp, struct route_in6 *ro, struct ucred *cred,
178 struct ifnet **ifpp, struct in6_addr *srcp)
179 {
180 struct in6_addr dst, tmp;
181 struct ifnet *ifp = NULL, *oifp = NULL;
182 struct in6_ifaddr *ia = NULL, *ia_best = NULL;
183 struct in6_pktinfo *pi = NULL;
184 int dst_scope = -1, best_scope = -1, best_matchlen = -1;
185 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
186 u_int32_t odstzone;
187 int prefer_tempaddr;
188 int error, rule;
189 struct ip6_moptions *mopts;
190
191 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__));
192
193 dst = dstsock->sin6_addr; /* make a copy for local operation */
194 if (ifpp) {
195 /*
196 * Save a possibly passed in ifp for in6_selectsrc. Only
197 * neighbor discovery code should use this feature, where
198 * we may know the interface but not the FIB number holding
199 * the connected subnet in case someone deleted it from the
200 * default FIB and we need to check the interface.
201 */
202 if (*ifpp != NULL)
203 oifp = *ifpp;
204 *ifpp = NULL;
205 }
206
207 if (inp != NULL) {
208 INP_LOCK_ASSERT(inp);
209 mopts = inp->in6p_moptions;
210 } else {
211 mopts = NULL;
212 }
213
214 /*
215 * If the source address is explicitly specified by the caller,
216 * check if the requested source address is indeed a unicast address
217 * assigned to the node, and can be used as the packet's source
218 * address. If everything is okay, use the address as source.
219 */
220 if (opts && (pi = opts->ip6po_pktinfo) &&
221 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
222 struct sockaddr_in6 srcsock;
223 struct in6_ifaddr *ia6;
224
225 /* get the outgoing interface */
226 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp,
227 (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB))
228 != 0)
229 return (error);
230
231 /*
232 * determine the appropriate zone id of the source based on
233 * the zone of the destination and the outgoing interface.
234 * If the specified address is ambiguous wrt the scope zone,
235 * the interface must be specified; otherwise, ifa_ifwithaddr()
236 * will fail matching the address.
237 */
238 bzero(&srcsock, sizeof(srcsock));
239 srcsock.sin6_family = AF_INET6;
240 srcsock.sin6_len = sizeof(srcsock);
241 srcsock.sin6_addr = pi->ipi6_addr;
242 if (ifp) {
243 error = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
244 if (error)
245 return (error);
246 }
247 if (cred != NULL && (error = prison_local_ip6(cred,
248 &srcsock.sin6_addr, (inp != NULL &&
249 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
250 return (error);
251
252 ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(
253 (struct sockaddr *)&srcsock);
254 if (ia6 == NULL ||
255 (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) {
256 if (ia6 != NULL)
257 ifa_free(&ia6->ia_ifa);
258 return (EADDRNOTAVAIL);
259 }
260 pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
261 if (ifpp)
262 *ifpp = ifp;
263 bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp));
264 ifa_free(&ia6->ia_ifa);
265 return (0);
266 }
267
268 /*
269 * Otherwise, if the socket has already bound the source, just use it.
270 */
271 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
272 if (cred != NULL &&
273 (error = prison_local_ip6(cred, &inp->in6p_laddr,
274 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
275 return (error);
276 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp));
277 return (0);
278 }
279
280 /*
281 * Bypass source address selection and use the primary jail IP
282 * if requested.
283 */
284 if (cred != NULL && !prison_saddrsel_ip6(cred, srcp))
285 return (0);
286
287 /*
288 * If the address is not specified, choose the best one based on
289 * the outgoing interface and the destination address.
290 */
291 /* get the outgoing interface */
292 if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp,
293 (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB)) != 0)
294 return (error);
295
296 #ifdef DIAGNOSTIC
297 if (ifp == NULL) /* this should not happen */
298 panic("in6_selectsrc: NULL ifp");
299 #endif
300 error = in6_setscope(&dst, ifp, &odstzone);
301 if (error)
302 return (error);
303
304 rule = 0;
305 IN6_IFADDR_RLOCK();
306 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
307 int new_scope = -1, new_matchlen = -1;
308 struct in6_addrpolicy *new_policy = NULL;
309 u_int32_t srczone, osrczone, dstzone;
310 struct in6_addr src;
311 struct ifnet *ifp1 = ia->ia_ifp;
312
313 /*
314 * We'll never take an address that breaks the scope zone
315 * of the destination. We also skip an address if its zone
316 * does not contain the outgoing interface.
317 * XXX: we should probably use sin6_scope_id here.
318 */
319 if (in6_setscope(&dst, ifp1, &dstzone) ||
320 odstzone != dstzone) {
321 continue;
322 }
323 src = ia->ia_addr.sin6_addr;
324 if (in6_setscope(&src, ifp, &osrczone) ||
325 in6_setscope(&src, ifp1, &srczone) ||
326 osrczone != srczone) {
327 continue;
328 }
329
330 /* avoid unusable addresses */
331 if ((ia->ia6_flags &
332 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
333 continue;
334 }
335 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
336 continue;
337
338 /* If jailed only take addresses of the jail into account. */
339 if (cred != NULL &&
340 prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0)
341 continue;
342
343 /* Rule 1: Prefer same address */
344 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
345 ia_best = ia;
346 BREAK(1); /* there should be no better candidate */
347 }
348
349 if (ia_best == NULL)
350 REPLACE(0);
351
352 /* Rule 2: Prefer appropriate scope */
353 if (dst_scope < 0)
354 dst_scope = in6_addrscope(&dst);
355 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
356 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
357 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
358 REPLACE(2);
359 NEXT(2);
360 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
361 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
362 NEXT(2);
363 REPLACE(2);
364 }
365
366 /*
367 * Rule 3: Avoid deprecated addresses. Note that the case of
368 * !ip6_use_deprecated is already rejected above.
369 */
370 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
371 NEXT(3);
372 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
373 REPLACE(3);
374
375 /* Rule 4: Prefer home addresses */
376 /*
377 * XXX: This is a TODO. We should probably merge the MIP6
378 * case above.
379 */
380
381 /* Rule 5: Prefer outgoing interface */
382 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) {
383 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
384 NEXT(5);
385 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
386 REPLACE(5);
387 }
388
389 /*
390 * Rule 6: Prefer matching label
391 * Note that best_policy should be non-NULL here.
392 */
393 if (dst_policy == NULL)
394 dst_policy = lookup_addrsel_policy(dstsock);
395 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
396 new_policy = lookup_addrsel_policy(&ia->ia_addr);
397 if (dst_policy->label == best_policy->label &&
398 dst_policy->label != new_policy->label)
399 NEXT(6);
400 if (dst_policy->label != best_policy->label &&
401 dst_policy->label == new_policy->label)
402 REPLACE(6);
403 }
404
405 /*
406 * Rule 7: Prefer public addresses.
407 * We allow users to reverse the logic by configuring
408 * a sysctl variable, so that privacy conscious users can
409 * always prefer temporary addresses.
410 */
411 if (opts == NULL ||
412 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
413 prefer_tempaddr = V_ip6_prefer_tempaddr;
414 } else if (opts->ip6po_prefer_tempaddr ==
415 IP6PO_TEMPADDR_NOTPREFER) {
416 prefer_tempaddr = 0;
417 } else
418 prefer_tempaddr = 1;
419 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
420 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
421 if (prefer_tempaddr)
422 REPLACE(7);
423 else
424 NEXT(7);
425 }
426 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
427 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
428 if (prefer_tempaddr)
429 NEXT(7);
430 else
431 REPLACE(7);
432 }
433
434 /*
435 * Rule 8: prefer addresses on alive interfaces.
436 * This is a KAME specific rule.
437 */
438 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
439 !(ia->ia_ifp->if_flags & IFF_UP))
440 NEXT(8);
441 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
442 (ia->ia_ifp->if_flags & IFF_UP))
443 REPLACE(8);
444
445 /*
446 * Rule 10: prefer address with `prefer_source' flag.
447 */
448 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0 &&
449 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0)
450 REPLACE(10);
451 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0 &&
452 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0)
453 NEXT(10);
454
455 /*
456 * Rule 14: Use longest matching prefix.
457 * Note: in the address selection draft, this rule is
458 * documented as "Rule 8". However, since it is also
459 * documented that this rule can be overridden, we assign
460 * a large number so that it is easy to assign smaller numbers
461 * to more preferred rules.
462 */
463 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
464 if (best_matchlen < new_matchlen)
465 REPLACE(14);
466 if (new_matchlen < best_matchlen)
467 NEXT(14);
468
469 /* Rule 15 is reserved. */
470
471 /*
472 * Last resort: just keep the current candidate.
473 * Or, do we need more rules?
474 */
475 continue;
476
477 replace:
478 ia_best = ia;
479 best_scope = (new_scope >= 0 ? new_scope :
480 in6_addrscope(&ia_best->ia_addr.sin6_addr));
481 best_policy = (new_policy ? new_policy :
482 lookup_addrsel_policy(&ia_best->ia_addr));
483 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
484 in6_matchlen(&ia_best->ia_addr.sin6_addr,
485 &dst));
486
487 next:
488 continue;
489
490 out:
491 break;
492 }
493
494 if ((ia = ia_best) == NULL) {
495 IN6_IFADDR_RUNLOCK();
496 IP6STAT_INC(ip6s_sources_none);
497 return (EADDRNOTAVAIL);
498 }
499
500 /*
501 * At this point at least one of the addresses belonged to the jail
502 * but it could still be, that we want to further restrict it, e.g.
503 * theoratically IN6_IS_ADDR_LOOPBACK.
504 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore.
505 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should
506 * let all others previously selected pass.
507 * Use tmp to not change ::1 on lo0 to the primary jail address.
508 */
509 tmp = ia->ia_addr.sin6_addr;
510 if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL &&
511 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) {
512 IN6_IFADDR_RUNLOCK();
513 IP6STAT_INC(ip6s_sources_none);
514 return (EADDRNOTAVAIL);
515 }
516
517 if (ifpp)
518 *ifpp = ifp;
519
520 bcopy(&tmp, srcp, sizeof(*srcp));
521 if (ia->ia_ifp == ifp)
522 IP6STAT_INC(ip6s_sources_sameif[best_scope]);
523 else
524 IP6STAT_INC(ip6s_sources_otherif[best_scope]);
525 if (dst_scope == best_scope)
526 IP6STAT_INC(ip6s_sources_samescope[best_scope]);
527 else
528 IP6STAT_INC(ip6s_sources_otherscope[best_scope]);
529 if (IFA6_IS_DEPRECATED(ia))
530 IP6STAT_INC(ip6s_sources_deprecated[best_scope]);
531 IN6_IFADDR_RUNLOCK();
532 return (0);
533 }
534
535 /*
536 * clone - meaningful only for bsdi and freebsd
537 */
538 static int
539 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
540 struct ip6_moptions *mopts, struct route_in6 *ro,
541 struct ifnet **retifp, struct rtentry **retrt, int norouteok, u_int fibnum)
542 {
543 int error = 0;
544 struct ifnet *ifp = NULL;
545 struct rtentry *rt = NULL;
546 struct sockaddr_in6 *sin6_next;
547 struct in6_pktinfo *pi = NULL;
548 struct in6_addr *dst = &dstsock->sin6_addr;
549 #if 0
550 char ip6buf[INET6_ADDRSTRLEN];
551
552 if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
553 dstsock->sin6_addr.s6_addr32[1] == 0 &&
554 !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
555 printf("in6_selectroute: strange destination %s\n",
556 ip6_sprintf(ip6buf, &dstsock->sin6_addr));
557 } else {
558 printf("in6_selectroute: destination = %s%%%d\n",
559 ip6_sprintf(ip6buf, &dstsock->sin6_addr),
560 dstsock->sin6_scope_id); /* for debug */
561 }
562 #endif
563
564 /* If the caller specify the outgoing interface explicitly, use it. */
565 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
566 /* XXX boundary check is assumed to be already done. */
567 ifp = ifnet_byindex(pi->ipi6_ifindex);
568 if (ifp != NULL &&
569 (norouteok || retrt == NULL ||
570 IN6_IS_ADDR_MULTICAST(dst))) {
571 /*
572 * we do not have to check or get the route for
573 * multicast.
574 */
575 goto done;
576 } else
577 goto getroute;
578 }
579
580 /*
581 * If the destination address is a multicast address and the outgoing
582 * interface for the address is specified by the caller, use it.
583 */
584 if (IN6_IS_ADDR_MULTICAST(dst) &&
585 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
586 goto done; /* we do not need a route for multicast. */
587 }
588
589 getroute:
590 /*
591 * If the next hop address for the packet is specified by the caller,
592 * use it as the gateway.
593 */
594 if (opts && opts->ip6po_nexthop) {
595 struct route_in6 *ron;
596 struct llentry *la;
597
598 sin6_next = satosin6(opts->ip6po_nexthop);
599
600 /* at this moment, we only support AF_INET6 next hops */
601 if (sin6_next->sin6_family != AF_INET6) {
602 error = EAFNOSUPPORT; /* or should we proceed? */
603 goto done;
604 }
605
606 /*
607 * If the next hop is an IPv6 address, then the node identified
608 * by that address must be a neighbor of the sending host.
609 */
610 ron = &opts->ip6po_nextroute;
611 /*
612 * XXX what do we do here?
613 * PLZ to be fixing
614 */
615
616
617 if (ron->ro_rt == NULL) {
618 in6_rtalloc(ron, fibnum); /* multi path case? */
619 if (ron->ro_rt == NULL) {
620 if (ron->ro_rt) {
621 RTFREE(ron->ro_rt);
622 ron->ro_rt = NULL;
623 }
624 error = EHOSTUNREACH;
625 goto done;
626 }
627 }
628
629 rt = ron->ro_rt;
630 ifp = rt->rt_ifp;
631 IF_AFDATA_RLOCK(ifp);
632 la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr);
633 IF_AFDATA_RUNLOCK(ifp);
634 if (la != NULL)
635 LLE_RUNLOCK(la);
636 else {
637 error = EHOSTUNREACH;
638 goto done;
639 }
640 #if 0
641 if ((ron->ro_rt &&
642 (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) !=
643 (RTF_UP | RTF_LLINFO)) ||
644 !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr,
645 &sin6_next->sin6_addr)) {
646 if (ron->ro_rt) {
647 RTFREE(ron->ro_rt);
648 ron->ro_rt = NULL;
649 }
650 *satosin6(&ron->ro_dst) = *sin6_next;
651 }
652 if (ron->ro_rt == NULL) {
653 in6_rtalloc(ron, fibnum); /* multi path case? */
654 if (ron->ro_rt == NULL ||
655 !(ron->ro_rt->rt_flags & RTF_LLINFO)) {
656 if (ron->ro_rt) {
657 RTFREE(ron->ro_rt);
658 ron->ro_rt = NULL;
659 }
660 error = EHOSTUNREACH;
661 goto done;
662 }
663 }
664 #endif
665
666 /*
667 * When cloning is required, try to allocate a route to the
668 * destination so that the caller can store path MTU
669 * information.
670 */
671 goto done;
672 }
673
674 /*
675 * Use a cached route if it exists and is valid, else try to allocate
676 * a new one. Note that we should check the address family of the
677 * cached destination, in case of sharing the cache with IPv4.
678 */
679 if (ro) {
680 if (ro->ro_rt &&
681 (!(ro->ro_rt->rt_flags & RTF_UP) ||
682 ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
683 !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
684 dst))) {
685 RTFREE(ro->ro_rt);
686 ro->ro_rt = (struct rtentry *)NULL;
687 }
688 if (ro->ro_rt == (struct rtentry *)NULL) {
689 struct sockaddr_in6 *sa6;
690
691 /* No route yet, so try to acquire one */
692 bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
693 sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
694 *sa6 = *dstsock;
695 sa6->sin6_scope_id = 0;
696
697 #ifdef RADIX_MPATH
698 rtalloc_mpath_fib((struct route *)ro,
699 ntohl(sa6->sin6_addr.s6_addr32[3]), fibnum);
700 #else
701 ro->ro_rt = in6_rtalloc1((struct sockaddr *)
702 &ro->ro_dst, 0, 0UL, fibnum);
703 if (ro->ro_rt)
704 RT_UNLOCK(ro->ro_rt);
705 #endif
706 }
707
708 /*
709 * do not care about the result if we have the nexthop
710 * explicitly specified.
711 */
712 if (opts && opts->ip6po_nexthop)
713 goto done;
714
715 if (ro->ro_rt) {
716 ifp = ro->ro_rt->rt_ifp;
717
718 if (ifp == NULL) { /* can this really happen? */
719 RTFREE(ro->ro_rt);
720 ro->ro_rt = NULL;
721 }
722 }
723 if (ro->ro_rt == NULL)
724 error = EHOSTUNREACH;
725 rt = ro->ro_rt;
726
727 /*
728 * Check if the outgoing interface conflicts with
729 * the interface specified by ipi6_ifindex (if specified).
730 * Note that loopback interface is always okay.
731 * (this may happen when we are sending a packet to one of
732 * our own addresses.)
733 */
734 if (ifp && opts && opts->ip6po_pktinfo &&
735 opts->ip6po_pktinfo->ipi6_ifindex) {
736 if (!(ifp->if_flags & IFF_LOOPBACK) &&
737 ifp->if_index !=
738 opts->ip6po_pktinfo->ipi6_ifindex) {
739 error = EHOSTUNREACH;
740 goto done;
741 }
742 }
743 }
744
745 done:
746 if (ifp == NULL && rt == NULL) {
747 /*
748 * This can happen if the caller did not pass a cached route
749 * nor any other hints. We treat this case an error.
750 */
751 error = EHOSTUNREACH;
752 }
753 if (error == EHOSTUNREACH)
754 IP6STAT_INC(ip6s_noroute);
755
756 if (retifp != NULL) {
757 *retifp = ifp;
758
759 /*
760 * Adjust the "outgoing" interface. If we're going to loop
761 * the packet back to ourselves, the ifp would be the loopback
762 * interface. However, we'd rather know the interface associated
763 * to the destination address (which should probably be one of
764 * our own addresses.)
765 */
766 if (rt) {
767 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
768 (rt->rt_gateway->sa_family == AF_LINK))
769 *retifp =
770 ifnet_byindex(((struct sockaddr_dl *)
771 rt->rt_gateway)->sdl_index);
772 }
773 }
774
775 if (retrt != NULL)
776 *retrt = rt; /* rt may be NULL */
777
778 return (error);
779 }
780
781 static int
782 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
783 struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp,
784 struct ifnet *oifp, u_int fibnum)
785 {
786 int error;
787 struct route_in6 sro;
788 struct rtentry *rt = NULL;
789
790 KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__));
791
792 if (ro == NULL) {
793 bzero(&sro, sizeof(sro));
794 ro = &sro;
795 }
796
797 if ((error = selectroute(dstsock, opts, mopts, ro, retifp,
798 &rt, 1, fibnum)) != 0) {
799 if (ro == &sro && rt && rt == sro.ro_rt)
800 RTFREE(rt);
801 /* Help ND. See oifp comment in in6_selectsrc(). */
802 if (oifp != NULL && fibnum == RT_DEFAULT_FIB) {
803 *retifp = oifp;
804 error = 0;
805 }
806 return (error);
807 }
808
809 /*
810 * do not use a rejected or black hole route.
811 * XXX: this check should be done in the L2 output routine.
812 * However, if we skipped this check here, we'd see the following
813 * scenario:
814 * - install a rejected route for a scoped address prefix
815 * (like fe80::/10)
816 * - send a packet to a destination that matches the scoped prefix,
817 * with ambiguity about the scope zone.
818 * - pick the outgoing interface from the route, and disambiguate the
819 * scope zone with the interface.
820 * - ip6_output() would try to get another route with the "new"
821 * destination, which may be valid.
822 * - we'd see no error on output.
823 * Although this may not be very harmful, it should still be confusing.
824 * We thus reject the case here.
825 */
826 if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
827 int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
828
829 if (ro == &sro && rt && rt == sro.ro_rt)
830 RTFREE(rt);
831 return (flags);
832 }
833
834 if (ro == &sro && rt && rt == sro.ro_rt)
835 RTFREE(rt);
836 return (0);
837 }
838
839 /*
840 * Public wrapper function to selectroute().
841 *
842 * XXX-BZ in6_selectroute() should and will grow the FIB argument. The
843 * in6_selectroute_fib() function is only there for backward compat on stable.
844 */
845 int
846 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
847 struct ip6_moptions *mopts, struct route_in6 *ro,
848 struct ifnet **retifp, struct rtentry **retrt)
849 {
850
851 return (selectroute(dstsock, opts, mopts, ro, retifp,
852 retrt, 0, RT_DEFAULT_FIB));
853 }
854
855 #ifndef BURN_BRIDGES
856 int
857 in6_selectroute_fib(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
858 struct ip6_moptions *mopts, struct route_in6 *ro,
859 struct ifnet **retifp, struct rtentry **retrt, u_int fibnum)
860 {
861
862 return (selectroute(dstsock, opts, mopts, ro, retifp,
863 retrt, 0, fibnum));
864 }
865 #endif
866
867 /*
868 * Default hop limit selection. The precedence is as follows:
869 * 1. Hoplimit value specified via ioctl.
870 * 2. (If the outgoing interface is detected) the current
871 * hop limit of the interface specified by router advertisement.
872 * 3. The system default hoplimit.
873 */
874 int
875 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp)
876 {
877
878 if (in6p && in6p->in6p_hops >= 0)
879 return (in6p->in6p_hops);
880 else if (ifp)
881 return (ND_IFINFO(ifp)->chlim);
882 else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
883 struct route_in6 ro6;
884 struct ifnet *lifp;
885
886 bzero(&ro6, sizeof(ro6));
887 ro6.ro_dst.sin6_family = AF_INET6;
888 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
889 ro6.ro_dst.sin6_addr = in6p->in6p_faddr;
890 in6_rtalloc(&ro6, in6p->inp_inc.inc_fibnum);
891 if (ro6.ro_rt) {
892 lifp = ro6.ro_rt->rt_ifp;
893 RTFREE(ro6.ro_rt);
894 if (lifp)
895 return (ND_IFINFO(lifp)->chlim);
896 }
897 }
898 return (V_ip6_defhlim);
899 }
900
901 /*
902 * XXX: this is borrowed from in6_pcbbind(). If possible, we should
903 * share this function by all *bsd*...
904 */
905 int
906 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
907 {
908 struct socket *so = inp->inp_socket;
909 u_int16_t lport = 0;
910 int error, lookupflags = 0;
911 #ifdef INVARIANTS
912 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
913 #endif
914
915 INP_WLOCK_ASSERT(inp);
916 INP_HASH_WLOCK_ASSERT(pcbinfo);
917
918 error = prison_local_ip6(cred, laddr,
919 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
920 if (error)
921 return(error);
922
923 /* XXX: this is redundant when called from in6_pcbbind */
924 if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
925 lookupflags = INPLOOKUP_WILDCARD;
926
927 inp->inp_flags |= INP_ANONPORT;
928
929 error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
930 if (error != 0)
931 return (error);
932
933 inp->inp_lport = lport;
934 if (in_pcbinshash(inp) != 0) {
935 inp->in6p_laddr = in6addr_any;
936 inp->inp_lport = 0;
937 return (EAGAIN);
938 }
939
940 return (0);
941 }
942
943 void
944 addrsel_policy_init(void)
945 {
946
947 init_policy_queue();
948
949 /* initialize the "last resort" policy */
950 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
951 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
952
953 if (!IS_DEFAULT_VNET(curvnet))
954 return;
955
956 ADDRSEL_LOCK_INIT();
957 ADDRSEL_SXLOCK_INIT();
958 }
959
960 static struct in6_addrpolicy *
961 lookup_addrsel_policy(struct sockaddr_in6 *key)
962 {
963 struct in6_addrpolicy *match = NULL;
964
965 ADDRSEL_LOCK();
966 match = match_addrsel_policy(key);
967
968 if (match == NULL)
969 match = &V_defaultaddrpolicy;
970 else
971 match->use++;
972 ADDRSEL_UNLOCK();
973
974 return (match);
975 }
976
977 /*
978 * Subroutines to manage the address selection policy table via sysctl.
979 */
980 struct walkarg {
981 struct sysctl_req *w_req;
982 };
983
984 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
985 SYSCTL_DECL(_net_inet6_ip6);
986 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
987 CTLFLAG_RD, in6_src_sysctl, "");
988
989 static int
990 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
991 {
992 struct walkarg w;
993
994 if (req->newptr)
995 return EPERM;
996
997 bzero(&w, sizeof(w));
998 w.w_req = req;
999
1000 return (walk_addrsel_policy(dump_addrsel_policyent, &w));
1001 }
1002
1003 int
1004 in6_src_ioctl(u_long cmd, caddr_t data)
1005 {
1006 int i;
1007 struct in6_addrpolicy ent0;
1008
1009 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
1010 return (EOPNOTSUPP); /* check for safety */
1011
1012 ent0 = *(struct in6_addrpolicy *)data;
1013
1014 if (ent0.label == ADDR_LABEL_NOTAPP)
1015 return (EINVAL);
1016 /* check if the prefix mask is consecutive. */
1017 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
1018 return (EINVAL);
1019 /* clear trailing garbages (if any) of the prefix address. */
1020 for (i = 0; i < 4; i++) {
1021 ent0.addr.sin6_addr.s6_addr32[i] &=
1022 ent0.addrmask.sin6_addr.s6_addr32[i];
1023 }
1024 ent0.use = 0;
1025
1026 switch (cmd) {
1027 case SIOCAADDRCTL_POLICY:
1028 return (add_addrsel_policyent(&ent0));
1029 case SIOCDADDRCTL_POLICY:
1030 return (delete_addrsel_policyent(&ent0));
1031 }
1032
1033 return (0); /* XXX: compromise compilers */
1034 }
1035
1036 /*
1037 * The followings are implementation of the policy table using a
1038 * simple tail queue.
1039 * XXX such details should be hidden.
1040 * XXX implementation using binary tree should be more efficient.
1041 */
1042 struct addrsel_policyent {
1043 TAILQ_ENTRY(addrsel_policyent) ape_entry;
1044 struct in6_addrpolicy ape_policy;
1045 };
1046
1047 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1048
1049 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab);
1050 #define V_addrsel_policytab VNET(addrsel_policytab)
1051
1052 static void
1053 init_policy_queue(void)
1054 {
1055
1056 TAILQ_INIT(&V_addrsel_policytab);
1057 }
1058
1059 static int
1060 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1061 {
1062 struct addrsel_policyent *new, *pol;
1063
1064 new = malloc(sizeof(*new), M_IFADDR,
1065 M_WAITOK);
1066 ADDRSEL_XLOCK();
1067 ADDRSEL_LOCK();
1068
1069 /* duplication check */
1070 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1071 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1072 &pol->ape_policy.addr.sin6_addr) &&
1073 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1074 &pol->ape_policy.addrmask.sin6_addr)) {
1075 ADDRSEL_UNLOCK();
1076 ADDRSEL_XUNLOCK();
1077 free(new, M_IFADDR);
1078 return (EEXIST); /* or override it? */
1079 }
1080 }
1081
1082 bzero(new, sizeof(*new));
1083
1084 /* XXX: should validate entry */
1085 new->ape_policy = *newpolicy;
1086
1087 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1088 ADDRSEL_UNLOCK();
1089 ADDRSEL_XUNLOCK();
1090
1091 return (0);
1092 }
1093
1094 static int
1095 delete_addrsel_policyent(struct in6_addrpolicy *key)
1096 {
1097 struct addrsel_policyent *pol;
1098
1099 ADDRSEL_XLOCK();
1100 ADDRSEL_LOCK();
1101
1102 /* search for the entry in the table */
1103 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1104 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1105 &pol->ape_policy.addr.sin6_addr) &&
1106 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1107 &pol->ape_policy.addrmask.sin6_addr)) {
1108 break;
1109 }
1110 }
1111 if (pol == NULL) {
1112 ADDRSEL_UNLOCK();
1113 ADDRSEL_XUNLOCK();
1114 return (ESRCH);
1115 }
1116
1117 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1118 ADDRSEL_UNLOCK();
1119 ADDRSEL_XUNLOCK();
1120 free(pol, M_IFADDR);
1121
1122 return (0);
1123 }
1124
1125 static int
1126 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *),
1127 void *w)
1128 {
1129 struct addrsel_policyent *pol;
1130 int error = 0;
1131
1132 ADDRSEL_SLOCK();
1133 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1134 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1135 ADDRSEL_SUNLOCK();
1136 return (error);
1137 }
1138 }
1139 ADDRSEL_SUNLOCK();
1140 return (error);
1141 }
1142
1143 static int
1144 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1145 {
1146 int error = 0;
1147 struct walkarg *w = arg;
1148
1149 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1150
1151 return (error);
1152 }
1153
1154 static struct in6_addrpolicy *
1155 match_addrsel_policy(struct sockaddr_in6 *key)
1156 {
1157 struct addrsel_policyent *pent;
1158 struct in6_addrpolicy *bestpol = NULL, *pol;
1159 int matchlen, bestmatchlen = -1;
1160 u_char *mp, *ep, *k, *p, m;
1161
1162 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1163 matchlen = 0;
1164
1165 pol = &pent->ape_policy;
1166 mp = (u_char *)&pol->addrmask.sin6_addr;
1167 ep = mp + 16; /* XXX: scope field? */
1168 k = (u_char *)&key->sin6_addr;
1169 p = (u_char *)&pol->addr.sin6_addr;
1170 for (; mp < ep && *mp; mp++, k++, p++) {
1171 m = *mp;
1172 if ((*k & m) != *p)
1173 goto next; /* not match */
1174 if (m == 0xff) /* short cut for a typical case */
1175 matchlen += 8;
1176 else {
1177 while (m >= 0x80) {
1178 matchlen++;
1179 m <<= 1;
1180 }
1181 }
1182 }
1183
1184 /* matched. check if this is better than the current best. */
1185 if (bestpol == NULL ||
1186 matchlen > bestmatchlen) {
1187 bestpol = pol;
1188 bestmatchlen = matchlen;
1189 }
1190
1191 next:
1192 continue;
1193 }
1194
1195 return (bestpol);
1196 }
Cache object: 29ddf27bc048a6403bf6d3d32521b699
|