1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 *
31 * $KAME: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $
32 */
33
34 /*-
35 * Copyright (c) 1982, 1986, 1991, 1993
36 * The Regents of the University of California. All rights reserved.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. Neither the name of the University nor the names of its contributors
47 * may be used to endorse or promote products derived from this software
48 * without specific prior written permission.
49 *
50 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60 * SUCH DAMAGE.
61 */
62
63 #include "opt_inet.h"
64 #include "opt_inet6.h"
65 #include <sys/param.h>
66 #include <sys/systm.h>
67 #include <sys/lock.h>
68 #include <sys/malloc.h>
69 #include <sys/mbuf.h>
70 #include <sys/priv.h>
71 #include <sys/protosw.h>
72 #include <sys/socket.h>
73 #include <sys/socketvar.h>
74 #include <sys/sockio.h>
75 #include <sys/sysctl.h>
76 #include <sys/errno.h>
77 #include <sys/time.h>
78 #include <sys/jail.h>
79 #include <sys/kernel.h>
80 #include <sys/rmlock.h>
81 #include <sys/sx.h>
82
83 #include <net/if.h>
84 #include <net/if_var.h>
85 #include <net/if_dl.h>
86 #include <net/if_private.h>
87 #include <net/route.h>
88 #include <net/route/nhop.h>
89 #include <net/if_llatbl.h>
90
91 #include <netinet/in.h>
92 #include <netinet/in_var.h>
93 #include <netinet/in_systm.h>
94 #include <netinet/ip.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet/ip_var.h>
97 #include <netinet/udp.h>
98 #include <netinet/udp_var.h>
99
100 #include <netinet6/in6_var.h>
101 #include <netinet/ip6.h>
102 #include <netinet6/in6_fib.h>
103 #include <netinet6/in6_pcb.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet6/scope6_var.h>
106 #include <netinet6/nd6.h>
107
108 static struct mtx addrsel_lock;
109 #define ADDRSEL_LOCK_INIT() mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
110 #define ADDRSEL_LOCK() mtx_lock(&addrsel_lock)
111 #define ADDRSEL_UNLOCK() mtx_unlock(&addrsel_lock)
112 #define ADDRSEL_LOCK_ASSERT() mtx_assert(&addrsel_lock, MA_OWNED)
113
114 static struct sx addrsel_sxlock;
115 #define ADDRSEL_SXLOCK_INIT() sx_init(&addrsel_sxlock, "addrsel_sxlock")
116 #define ADDRSEL_SLOCK() sx_slock(&addrsel_sxlock)
117 #define ADDRSEL_SUNLOCK() sx_sunlock(&addrsel_sxlock)
118 #define ADDRSEL_XLOCK() sx_xlock(&addrsel_sxlock)
119 #define ADDRSEL_XUNLOCK() sx_xunlock(&addrsel_sxlock)
120
121 #define ADDR_LABEL_NOTAPP (-1)
122 VNET_DEFINE_STATIC(struct in6_addrpolicy, defaultaddrpolicy);
123 #define V_defaultaddrpolicy VNET(defaultaddrpolicy)
124
125 VNET_DEFINE(int, ip6_prefer_tempaddr) = 0;
126
127 static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *,
128 struct ip6_moptions *, struct route_in6 *, struct ifnet **,
129 struct nhop_object **, int, u_int, uint32_t);
130 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
131 struct ip6_moptions *, struct ifnet **,
132 struct ifnet *, u_int);
133 static int in6_selectsrc(uint32_t, struct sockaddr_in6 *,
134 struct ip6_pktopts *, struct ip6_moptions *, struct inpcb *,
135 struct ucred *, struct ifnet **, struct in6_addr *);
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_INC2(ip6s_sources_rule, (r)); \
155 /* { \
156 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
157 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)); \
158 } */ \
159 goto replace; \
160 } while(0)
161 #define NEXT(r) do {\
162 /* { \
163 char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
164 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)); \
165 } */ \
166 goto next; /* XXX: we can't use 'continue' here */ \
167 } while(0)
168 #define BREAK(r) do { \
169 IP6STAT_INC2(ip6s_sources_rule, (r)); \
170 goto out; /* XXX: we can't use 'break' here */ \
171 } while(0)
172
173 static int
in6_selectsrc(uint32_t fibnum,struct sockaddr_in6 * dstsock,struct ip6_pktopts * opts,struct ip6_moptions * mopts,struct inpcb * inp,struct ucred * cred,struct ifnet ** ifpp,struct in6_addr * srcp)174 in6_selectsrc(uint32_t fibnum, struct sockaddr_in6 *dstsock,
175 struct ip6_pktopts *opts, struct ip6_moptions *mopts, struct inpcb *inp,
176 struct ucred *cred, struct ifnet **ifpp, struct in6_addr *srcp)
177 {
178 struct rm_priotracker in6_ifa_tracker;
179 struct in6_addr dst, tmp;
180 struct ifnet *ifp = NULL, *oifp = NULL;
181 struct in6_ifaddr *ia = NULL, *ia_best = NULL;
182 struct in6_pktinfo *pi = NULL;
183 int dst_scope = -1, best_scope = -1, best_matchlen = -1;
184 struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
185 u_int32_t odstzone;
186 int prefer_tempaddr;
187 int error;
188
189 NET_EPOCH_ASSERT();
190 KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__));
191
192 dst = dstsock->sin6_addr; /* make a copy for local operation */
193 if (ifpp) {
194 /*
195 * Save a possibly passed in ifp for in6_selectsrc. Only
196 * neighbor discovery code should use this feature, where
197 * we may know the interface but not the FIB number holding
198 * the connected subnet in case someone deleted it from the
199 * default FIB and we need to check the interface.
200 */
201 if (*ifpp != NULL)
202 oifp = *ifpp;
203 *ifpp = NULL;
204 }
205
206 /*
207 * If the source address is explicitly specified by the caller,
208 * check if the requested source address is indeed a unicast address
209 * assigned to the node, and can be used as the packet's source
210 * address. If everything is okay, use the address as source.
211 */
212 if (opts && (pi = opts->ip6po_pktinfo) &&
213 !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
214 /* get the outgoing interface */
215 if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp,
216 fibnum))
217 != 0)
218 return (error);
219
220 /*
221 * determine the appropriate zone id of the source based on
222 * the zone of the destination and the outgoing interface.
223 * If the specified address is ambiguous wrt the scope zone,
224 * the interface must be specified; otherwise, ifa_ifwithaddr()
225 * will fail matching the address.
226 */
227 tmp = pi->ipi6_addr;
228 if (ifp) {
229 error = in6_setscope(&tmp, ifp, &odstzone);
230 if (error)
231 return (error);
232 }
233 if (cred != NULL && (error = prison_local_ip6(cred,
234 &tmp, (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0)
235 return (error);
236
237 /*
238 * If IPV6_BINDANY socket option is set, we allow to specify
239 * non local addresses as source address in IPV6_PKTINFO
240 * ancillary data.
241 */
242 if ((inp->inp_flags & INP_BINDANY) == 0) {
243 ia = in6ifa_ifwithaddr(&tmp, 0 /* XXX */, false);
244 if (ia == NULL || (ia->ia6_flags & (IN6_IFF_ANYCAST |
245 IN6_IFF_NOTREADY)))
246 return (EADDRNOTAVAIL);
247 bcopy(&ia->ia_addr.sin6_addr, srcp, sizeof(*srcp));
248 } else
249 bcopy(&tmp, srcp, sizeof(*srcp));
250 pi->ipi6_addr = tmp; /* XXX: this overrides pi */
251 if (ifpp)
252 *ifpp = ifp;
253 return (0);
254 }
255
256 /*
257 * Otherwise, if the socket has already bound the source, just use it.
258 */
259 if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
260 if (cred != NULL &&
261 (error = prison_local_ip6(cred, &inp->in6p_laddr,
262 ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
263 return (error);
264 bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp));
265 return (0);
266 }
267
268 /*
269 * Bypass source address selection and use the primary jail IP
270 * if requested.
271 */
272 if (cred != NULL && !prison_saddrsel_ip6(cred, srcp))
273 return (0);
274
275 /*
276 * If the address is not specified, choose the best one based on
277 * the outgoing interface and the destination address.
278 */
279 /* get the outgoing interface */
280 if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp,
281 (inp != NULL) ? inp->inp_inc.inc_fibnum : fibnum)) != 0)
282 return (error);
283
284 #ifdef DIAGNOSTIC
285 if (ifp == NULL) /* this should not happen */
286 panic("in6_selectsrc: NULL ifp");
287 #endif
288 error = in6_setscope(&dst, ifp, &odstzone);
289 if (error)
290 return (error);
291
292 IN6_IFADDR_RLOCK(&in6_ifa_tracker);
293 CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
294 int new_scope = -1, new_matchlen = -1;
295 struct in6_addrpolicy *new_policy = NULL;
296 u_int32_t srczone, osrczone, dstzone;
297 struct in6_addr src;
298 struct ifnet *ifp1 = ia->ia_ifp;
299
300 /*
301 * We'll never take an address that breaks the scope zone
302 * of the destination. We also skip an address if its zone
303 * does not contain the outgoing interface.
304 * XXX: we should probably use sin6_scope_id here.
305 */
306 if (in6_setscope(&dst, ifp1, &dstzone) ||
307 odstzone != dstzone) {
308 continue;
309 }
310 src = ia->ia_addr.sin6_addr;
311 if (in6_setscope(&src, ifp, &osrczone) ||
312 in6_setscope(&src, ifp1, &srczone) ||
313 osrczone != srczone) {
314 continue;
315 }
316
317 /* avoid unusable addresses */
318 if ((ia->ia6_flags &
319 (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
320 continue;
321 }
322 if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
323 continue;
324
325 /* If jailed only take addresses of the jail into account. */
326 if (cred != NULL &&
327 prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0)
328 continue;
329
330 /* Rule 1: Prefer same address */
331 if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
332 ia_best = ia;
333 BREAK(1); /* there should be no better candidate */
334 }
335
336 if (ia_best == NULL)
337 REPLACE(0);
338
339 /* Rule 2: Prefer appropriate scope */
340 if (dst_scope < 0)
341 dst_scope = in6_addrscope(&dst);
342 new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
343 if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
344 if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
345 REPLACE(2);
346 NEXT(2);
347 } else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
348 if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
349 NEXT(2);
350 REPLACE(2);
351 }
352
353 /*
354 * Rule 3: Avoid deprecated addresses. Note that the case of
355 * !ip6_use_deprecated is already rejected above.
356 */
357 if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
358 NEXT(3);
359 if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
360 REPLACE(3);
361
362 /* Rule 4: Prefer home addresses */
363 /*
364 * XXX: This is a TODO. We should probably merge the MIP6
365 * case above.
366 */
367
368 /* Rule 5: Prefer outgoing interface */
369 if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) {
370 if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
371 NEXT(5);
372 if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
373 REPLACE(5);
374 }
375
376 /*
377 * Rule 6: Prefer matching label
378 * Note that best_policy should be non-NULL here.
379 */
380 if (dst_policy == NULL)
381 dst_policy = lookup_addrsel_policy(dstsock);
382 if (dst_policy->label != ADDR_LABEL_NOTAPP) {
383 new_policy = lookup_addrsel_policy(&ia->ia_addr);
384 if (dst_policy->label == best_policy->label &&
385 dst_policy->label != new_policy->label)
386 NEXT(6);
387 if (dst_policy->label != best_policy->label &&
388 dst_policy->label == new_policy->label)
389 REPLACE(6);
390 }
391
392 /*
393 * Rule 7: Prefer public addresses.
394 * We allow users to reverse the logic by configuring
395 * a sysctl variable, so that privacy conscious users can
396 * always prefer temporary addresses.
397 */
398 if (opts == NULL ||
399 opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
400 prefer_tempaddr = V_ip6_prefer_tempaddr;
401 } else if (opts->ip6po_prefer_tempaddr ==
402 IP6PO_TEMPADDR_NOTPREFER) {
403 prefer_tempaddr = 0;
404 } else
405 prefer_tempaddr = 1;
406 if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
407 (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
408 if (prefer_tempaddr)
409 REPLACE(7);
410 else
411 NEXT(7);
412 }
413 if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
414 !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
415 if (prefer_tempaddr)
416 NEXT(7);
417 else
418 REPLACE(7);
419 }
420
421 /*
422 * Rule 8: prefer addresses on alive interfaces.
423 * This is a KAME specific rule.
424 */
425 if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
426 !(ia->ia_ifp->if_flags & IFF_UP))
427 NEXT(8);
428 if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
429 (ia->ia_ifp->if_flags & IFF_UP))
430 REPLACE(8);
431
432 /*
433 * Rule 9: prefer address with better virtual status.
434 */
435 if (ifa_preferred(&ia_best->ia_ifa, &ia->ia_ifa))
436 REPLACE(9);
437 if (ifa_preferred(&ia->ia_ifa, &ia_best->ia_ifa))
438 NEXT(9);
439
440 /*
441 * Rule 10: prefer address with `prefer_source' flag.
442 */
443 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0 &&
444 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0)
445 REPLACE(10);
446 if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0 &&
447 (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0)
448 NEXT(10);
449
450 /*
451 * Rule 14: Use longest matching prefix.
452 * Note: in the address selection draft, this rule is
453 * documented as "Rule 8". However, since it is also
454 * documented that this rule can be overridden, we assign
455 * a large number so that it is easy to assign smaller numbers
456 * to more preferred rules.
457 */
458 new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
459 if (best_matchlen < new_matchlen)
460 REPLACE(14);
461 if (new_matchlen < best_matchlen)
462 NEXT(14);
463
464 /* Rule 15 is reserved. */
465
466 /*
467 * Last resort: just keep the current candidate.
468 * Or, do we need more rules?
469 */
470 continue;
471
472 replace:
473 ia_best = ia;
474 best_scope = (new_scope >= 0 ? new_scope :
475 in6_addrscope(&ia_best->ia_addr.sin6_addr));
476 best_policy = (new_policy ? new_policy :
477 lookup_addrsel_policy(&ia_best->ia_addr));
478 best_matchlen = (new_matchlen >= 0 ? new_matchlen :
479 in6_matchlen(&ia_best->ia_addr.sin6_addr,
480 &dst));
481
482 next:
483 continue;
484
485 out:
486 break;
487 }
488
489 if ((ia = ia_best) == NULL) {
490 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
491 IP6STAT_INC(ip6s_sources_none);
492 return (EADDRNOTAVAIL);
493 }
494
495 /*
496 * At this point at least one of the addresses belonged to the jail
497 * but it could still be, that we want to further restrict it, e.g.
498 * theoratically IN6_IS_ADDR_LOOPBACK.
499 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore.
500 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should
501 * let all others previously selected pass.
502 * Use tmp to not change ::1 on lo0 to the primary jail address.
503 */
504 tmp = ia->ia_addr.sin6_addr;
505 if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL &&
506 (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) {
507 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
508 IP6STAT_INC(ip6s_sources_none);
509 return (EADDRNOTAVAIL);
510 }
511
512 if (ifpp)
513 *ifpp = ifp;
514
515 bcopy(&tmp, srcp, sizeof(*srcp));
516 if (ia->ia_ifp == ifp)
517 IP6STAT_INC2(ip6s_sources_sameif, best_scope);
518 else
519 IP6STAT_INC2(ip6s_sources_otherif, best_scope);
520 if (dst_scope == best_scope)
521 IP6STAT_INC2(ip6s_sources_samescope, best_scope);
522 else
523 IP6STAT_INC2(ip6s_sources_otherscope, best_scope);
524 if (IFA6_IS_DEPRECATED(ia))
525 IP6STAT_INC2(ip6s_sources_deprecated, best_scope);
526 IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
527 return (0);
528 }
529
530 /*
531 * Select source address based on @inp, @dstsock and @opts.
532 * Stores selected address to @srcp. If @scope_ambiguous is set,
533 * embed scope from selected outgoing interface. If @hlim pointer
534 * is provided, stores calculated hop limit there.
535 * Returns 0 on success.
536 */
537 int
in6_selectsrc_socket(struct sockaddr_in6 * dstsock,struct ip6_pktopts * opts,struct inpcb * inp,struct ucred * cred,int scope_ambiguous,struct in6_addr * srcp,int * hlim)538 in6_selectsrc_socket(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
539 struct inpcb *inp, struct ucred *cred, int scope_ambiguous,
540 struct in6_addr *srcp, int *hlim)
541 {
542 struct ifnet *retifp;
543 uint32_t fibnum;
544 int error;
545
546 INP_LOCK_ASSERT(inp);
547
548 fibnum = inp->inp_inc.inc_fibnum;
549 retifp = NULL;
550
551 error = in6_selectsrc(fibnum, dstsock, opts, inp->in6p_moptions,
552 inp, cred, &retifp, srcp);
553 if (error != 0)
554 return (error);
555
556 if (hlim != NULL)
557 *hlim = in6_selecthlim(inp, retifp);
558
559 if (retifp == NULL || scope_ambiguous == 0)
560 return (0);
561
562 /*
563 * Application should provide a proper zone ID or the use of
564 * default zone IDs should be enabled. Unfortunately, some
565 * applications do not behave as it should, so we need a
566 * workaround. Even if an appropriate ID is not determined
567 * (when it's required), if we can determine the outgoing
568 * interface. determine the zone ID based on the interface.
569 */
570 error = in6_setscope(&dstsock->sin6_addr, retifp, NULL);
571
572 return (error);
573 }
574
575 /*
576 * Select source address based on @fibnum, @dst and @scopeid.
577 * Stores selected address to @srcp.
578 * Returns 0 on success.
579 *
580 * Used by non-socket based consumers
581 */
582 int
in6_selectsrc_addr(uint32_t fibnum,const struct in6_addr * dst,uint32_t scopeid,struct ifnet * ifp,struct in6_addr * srcp,int * hlim)583 in6_selectsrc_addr(uint32_t fibnum, const struct in6_addr *dst,
584 uint32_t scopeid, struct ifnet *ifp, struct in6_addr *srcp,
585 int *hlim)
586 {
587 struct ifnet *retifp;
588 struct sockaddr_in6 dst_sa;
589 int error;
590
591 retifp = ifp;
592 bzero(&dst_sa, sizeof(dst_sa));
593 dst_sa.sin6_family = AF_INET6;
594 dst_sa.sin6_len = sizeof(dst_sa);
595 dst_sa.sin6_addr = *dst;
596 dst_sa.sin6_scope_id = scopeid;
597 sa6_embedscope(&dst_sa, 0);
598
599 error = in6_selectsrc(fibnum, &dst_sa, NULL, NULL,
600 NULL, NULL, &retifp, srcp);
601 if (hlim != NULL)
602 *hlim = in6_selecthlim(NULL, retifp);
603
604 return (error);
605 }
606
607 /*
608 * Select source address based on @fibnum, @dst and @mopts.
609 * Stores selected address to @srcp.
610 * Returns 0 on success.
611 *
612 * Used by non-socket based consumers (ND code mostly)
613 */
614 int
in6_selectsrc_nbr(uint32_t fibnum,const struct in6_addr * dst,struct ip6_moptions * mopts,struct ifnet * ifp,struct in6_addr * srcp)615 in6_selectsrc_nbr(uint32_t fibnum, const struct in6_addr *dst,
616 struct ip6_moptions *mopts, struct ifnet *ifp, struct in6_addr *srcp)
617 {
618 struct sockaddr_in6 dst_sa;
619 struct ifnet *retifp;
620 int error;
621
622 retifp = ifp;
623 bzero(&dst_sa, sizeof(dst_sa));
624 dst_sa.sin6_family = AF_INET6;
625 dst_sa.sin6_len = sizeof(dst_sa);
626 dst_sa.sin6_addr = *dst;
627 dst_sa.sin6_scope_id = ntohs(in6_getscope(dst));
628 sa6_embedscope(&dst_sa, 0);
629
630 error = in6_selectsrc(fibnum, &dst_sa, NULL, mopts,
631 NULL, NULL, &retifp, srcp);
632 return (error);
633 }
634
635 static struct nhop_object *
cache_route(uint32_t fibnum,const struct sockaddr_in6 * dst,struct route_in6 * ro,uint32_t flowid)636 cache_route(uint32_t fibnum, const struct sockaddr_in6 *dst, struct route_in6 *ro,
637 uint32_t flowid)
638 {
639 /*
640 * Use a cached route if it exists and is valid, else try to allocate
641 * a new one. Note that we should check the address family of the
642 * cached destination, in case of sharing the cache with IPv4.
643 * Assumes that 'struct route_in6' is exclusively locked.
644 */
645 if (ro->ro_nh != NULL && (
646 !NH_IS_VALID(ro->ro_nh) || ro->ro_dst.sin6_family != AF_INET6 ||
647 !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &dst->sin6_addr)))
648 RO_NHFREE(ro);
649
650 if (ro->ro_nh == NULL) {
651 ro->ro_dst = *dst;
652
653 const struct in6_addr *paddr;
654 struct in6_addr unscoped_addr;
655 uint32_t scopeid = 0;
656 if (IN6_IS_SCOPE_LINKLOCAL(&dst->sin6_addr)) {
657 in6_splitscope(&dst->sin6_addr, &unscoped_addr, &scopeid);
658 paddr = &unscoped_addr;
659 } else
660 paddr = &dst->sin6_addr;
661 ro->ro_nh = fib6_lookup(fibnum, paddr, scopeid, NHR_REF, flowid);
662 }
663 return (ro->ro_nh);
664 }
665
666 static struct nhop_object *
lookup_route(uint32_t fibnum,struct sockaddr_in6 * dst,struct route_in6 * ro,struct ip6_pktopts * opts,uint32_t flowid)667 lookup_route(uint32_t fibnum, struct sockaddr_in6 *dst, struct route_in6 *ro,
668 struct ip6_pktopts *opts, uint32_t flowid)
669 {
670 struct nhop_object *nh = NULL;
671
672 /*
673 * If the next hop address for the packet is specified by the caller,
674 * use it as the gateway.
675 */
676 if (opts && opts->ip6po_nexthop) {
677 struct route_in6 *ron = &opts->ip6po_nextroute;
678 struct sockaddr_in6 *sin6_next = satosin6(opts->ip6po_nexthop);
679
680 nh = cache_route(fibnum, sin6_next, ron, flowid);
681
682 /*
683 * The node identified by that address must be a
684 * neighbor of the sending host.
685 */
686 if (nh != NULL && (nh->nh_flags & NHF_GATEWAY) != 0)
687 nh = NULL;
688 } else if (ro != NULL) {
689 nh = cache_route(fibnum, dst, ro, flowid);
690 if (nh == NULL)
691 return (NULL);
692
693 /*
694 * Check if the outgoing interface conflicts with
695 * the interface specified by ipi6_ifindex (if specified).
696 */
697 struct in6_pktinfo *pi;
698 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
699 if (nh->nh_aifp->if_index != pi->ipi6_ifindex)
700 nh = NULL;
701 }
702 }
703
704 return (nh);
705 }
706
707 /*
708 * Finds outgoing nexthop or the outgoing interface for the
709 * @dstsock.
710 * Return 0 on success and stores the lookup result in @retnh and @retifp
711 */
712 static int
selectroute(struct sockaddr_in6 * dstsock,struct ip6_pktopts * opts,struct ip6_moptions * mopts,struct route_in6 * ro,struct ifnet ** retifp,struct nhop_object ** retnh,int norouteok,u_int fibnum,uint32_t flowid)713 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
714 struct ip6_moptions *mopts, struct route_in6 *ro,
715 struct ifnet **retifp, struct nhop_object **retnh, int norouteok,
716 u_int fibnum, uint32_t flowid)
717 {
718 int error = 0;
719 struct ifnet *ifp = NULL;
720 struct in6_pktinfo *pi = NULL;
721 struct in6_addr *dst = &dstsock->sin6_addr;
722
723 /* If the caller specify the outgoing interface explicitly, use it. */
724 if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
725 /* XXX boundary check is assumed to be already done. */
726 ifp = ifnet_byindex(pi->ipi6_ifindex);
727 if (ifp != NULL && (norouteok || IN6_IS_ADDR_MULTICAST(dst))) {
728 /*
729 * we do not have to check or get the route for
730 * multicast.
731 */
732 goto done;
733 } else
734 goto getroute;
735 }
736 /*
737 * If the destination address is a multicast address and the outgoing
738 * interface for the address is specified by the caller, use it.
739 */
740 if (IN6_IS_ADDR_MULTICAST(dst) &&
741 mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
742 goto done; /* we do not need a route for multicast. */
743 }
744 /*
745 * If destination address is LLA or link- or node-local multicast,
746 * use it's embedded scope zone id to determine outgoing interface.
747 */
748 if (IN6_IS_ADDR_MC_LINKLOCAL(dst) ||
749 IN6_IS_ADDR_MC_NODELOCAL(dst)) {
750 uint32_t zoneid = ntohs(in6_getscope(dst));
751 if (zoneid > 0) {
752 ifp = in6_getlinkifnet(zoneid);
753 goto done;
754 }
755 }
756
757 getroute:;
758 struct nhop_object *nh = lookup_route(fibnum, dstsock, ro, opts, flowid);
759 if (nh != NULL) {
760 *retifp = nh->nh_aifp;
761 error = 0;
762 } else {
763 *retifp = NULL;
764 IP6STAT_INC(ip6s_noroute);
765 error = EHOSTUNREACH;
766 }
767 *retnh = nh;
768 return (error);
769
770 done:
771 if (ifp == NULL) {
772 /*
773 * This can happen if the caller did not pass a cached route
774 * nor any other hints. We treat this case an error.
775 */
776 error = EHOSTUNREACH;
777 }
778 if (error == EHOSTUNREACH)
779 IP6STAT_INC(ip6s_noroute);
780
781 *retifp = ifp;
782 *retnh = NULL;
783
784 return (error);
785 }
786
787 static int
in6_selectif(struct sockaddr_in6 * dstsock,struct ip6_pktopts * opts,struct ip6_moptions * mopts,struct ifnet ** retifp,struct ifnet * oifp,u_int fibnum)788 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
789 struct ip6_moptions *mopts, struct ifnet **retifp,
790 struct ifnet *oifp, u_int fibnum)
791 {
792 int error;
793 struct route_in6 sro;
794 struct nhop_object *nh = NULL;
795 uint16_t nh_flags;
796
797 KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__));
798
799 bzero(&sro, sizeof(sro));
800 nh_flags = 0;
801
802 error = selectroute(dstsock, opts, mopts, &sro, retifp, &nh, 1, fibnum, 0);
803
804 if (nh != NULL)
805 nh_flags = nh->nh_flags;
806 if (nh != NULL && nh == sro.ro_nh)
807 NH_FREE(nh);
808
809 if (error != 0) {
810 /* Help ND. See oifp comment in in6_selectsrc(). */
811 if (oifp != NULL && fibnum == RT_DEFAULT_FIB) {
812 *retifp = oifp;
813 error = 0;
814 }
815 return (error);
816 }
817
818 /*
819 * do not use a rejected or black hole route.
820 * XXX: this check should be done in the L2 output routine.
821 * However, if we skipped this check here, we'd see the following
822 * scenario:
823 * - install a rejected route for a scoped address prefix
824 * (like fe80::/10)
825 * - send a packet to a destination that matches the scoped prefix,
826 * with ambiguity about the scope zone.
827 * - pick the outgoing interface from the route, and disambiguate the
828 * scope zone with the interface.
829 * - ip6_output() would try to get another route with the "new"
830 * destination, which may be valid.
831 * - we'd see no error on output.
832 * Although this may not be very harmful, it should still be confusing.
833 * We thus reject the case here.
834 */
835
836 if (nh_flags & (NHF_REJECT | NHF_BLACKHOLE)) {
837 error = (nh_flags & NHF_HOST ? EHOSTUNREACH : ENETUNREACH);
838 return (error);
839 }
840
841 return (0);
842 }
843
844 /* Public wrapper function to selectroute(). */
845 int
in6_selectroute(struct sockaddr_in6 * dstsock,struct ip6_pktopts * opts,struct ip6_moptions * mopts,struct route_in6 * ro,struct ifnet ** retifp,struct nhop_object ** retnh,u_int fibnum,uint32_t flowid)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 nhop_object **retnh, u_int fibnum, uint32_t flowid)
849 {
850 MPASS(retifp != NULL);
851 MPASS(retnh != NULL);
852
853 return (selectroute(dstsock, opts, mopts, ro, retifp,
854 retnh, 0, fibnum, flowid));
855 }
856
857 /*
858 * Default hop limit selection. The precedence is as follows:
859 * 1. Hoplimit value specified via ioctl.
860 * 2. (If the outgoing interface is detected) the current
861 * hop limit of the interface specified by router advertisement.
862 * 3. The system default hoplimit.
863 */
864 int
in6_selecthlim(struct inpcb * inp,struct ifnet * ifp)865 in6_selecthlim(struct inpcb *inp, struct ifnet *ifp)
866 {
867
868 if (inp && inp->in6p_hops >= 0)
869 return (inp->in6p_hops);
870 else if (ifp)
871 return (ND_IFINFO(ifp)->chlim);
872 else if (inp && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
873 struct nhop_object *nh;
874 struct in6_addr dst;
875 uint32_t fibnum, scopeid;
876 int hlim;
877
878 fibnum = inp->inp_inc.inc_fibnum;
879 in6_splitscope(&inp->in6p_faddr, &dst, &scopeid);
880 nh = fib6_lookup(fibnum, &dst, scopeid, 0, 0);
881 if (nh != NULL) {
882 hlim = ND_IFINFO(nh->nh_ifp)->chlim;
883 return (hlim);
884 }
885 }
886 return (V_ip6_defhlim);
887 }
888
889 void
addrsel_policy_init(void)890 addrsel_policy_init(void)
891 {
892
893 init_policy_queue();
894
895 /* initialize the "last resort" policy */
896 bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
897 V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
898
899 if (!IS_DEFAULT_VNET(curvnet))
900 return;
901
902 ADDRSEL_LOCK_INIT();
903 ADDRSEL_SXLOCK_INIT();
904 }
905
906 static struct in6_addrpolicy *
lookup_addrsel_policy(struct sockaddr_in6 * key)907 lookup_addrsel_policy(struct sockaddr_in6 *key)
908 {
909 struct in6_addrpolicy *match = NULL;
910
911 ADDRSEL_LOCK();
912 match = match_addrsel_policy(key);
913
914 if (match == NULL)
915 match = &V_defaultaddrpolicy;
916 else
917 match->use++;
918 ADDRSEL_UNLOCK();
919
920 return (match);
921 }
922
923 /*
924 * Subroutines to manage the address selection policy table via sysctl.
925 */
926 struct walkarg {
927 struct sysctl_req *w_req;
928 };
929
930 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
931 SYSCTL_DECL(_net_inet6_ip6);
932 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
933 CTLFLAG_RD | CTLFLAG_MPSAFE, in6_src_sysctl,
934 "");
935
936 static int
in6_src_sysctl(SYSCTL_HANDLER_ARGS)937 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
938 {
939 struct walkarg w;
940
941 if (req->newptr)
942 return EPERM;
943
944 bzero(&w, sizeof(w));
945 w.w_req = req;
946
947 return (walk_addrsel_policy(dump_addrsel_policyent, &w));
948 }
949
950 int
in6_src_ioctl(u_long cmd,caddr_t data)951 in6_src_ioctl(u_long cmd, caddr_t data)
952 {
953 struct in6_addrpolicy ent0;
954
955 if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
956 return (EOPNOTSUPP); /* check for safety */
957
958 ent0 = *(struct in6_addrpolicy *)data;
959
960 if (ent0.label == ADDR_LABEL_NOTAPP)
961 return (EINVAL);
962 /* check if the prefix mask is consecutive. */
963 if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
964 return (EINVAL);
965 /* clear trailing garbages (if any) of the prefix address. */
966 IN6_MASK_ADDR(&ent0.addr.sin6_addr, &ent0.addrmask.sin6_addr);
967 ent0.use = 0;
968
969 switch (cmd) {
970 case SIOCAADDRCTL_POLICY:
971 return (add_addrsel_policyent(&ent0));
972 case SIOCDADDRCTL_POLICY:
973 return (delete_addrsel_policyent(&ent0));
974 }
975
976 return (0); /* XXX: compromise compilers */
977 }
978
979 /*
980 * The followings are implementation of the policy table using a
981 * simple tail queue.
982 * XXX such details should be hidden.
983 * XXX implementation using binary tree should be more efficient.
984 */
985 struct addrsel_policyent {
986 TAILQ_ENTRY(addrsel_policyent) ape_entry;
987 struct in6_addrpolicy ape_policy;
988 };
989
990 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
991
992 VNET_DEFINE_STATIC(struct addrsel_policyhead, addrsel_policytab);
993 #define V_addrsel_policytab VNET(addrsel_policytab)
994
995 static void
init_policy_queue(void)996 init_policy_queue(void)
997 {
998
999 TAILQ_INIT(&V_addrsel_policytab);
1000 }
1001
1002 static int
add_addrsel_policyent(struct in6_addrpolicy * newpolicy)1003 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1004 {
1005 struct addrsel_policyent *new, *pol;
1006
1007 new = malloc(sizeof(*new), M_IFADDR,
1008 M_WAITOK);
1009 ADDRSEL_XLOCK();
1010 ADDRSEL_LOCK();
1011
1012 /* duplication check */
1013 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1014 if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1015 &pol->ape_policy.addr.sin6_addr) &&
1016 IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1017 &pol->ape_policy.addrmask.sin6_addr)) {
1018 ADDRSEL_UNLOCK();
1019 ADDRSEL_XUNLOCK();
1020 free(new, M_IFADDR);
1021 return (EEXIST); /* or override it? */
1022 }
1023 }
1024
1025 bzero(new, sizeof(*new));
1026
1027 /* XXX: should validate entry */
1028 new->ape_policy = *newpolicy;
1029
1030 TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1031 ADDRSEL_UNLOCK();
1032 ADDRSEL_XUNLOCK();
1033
1034 return (0);
1035 }
1036
1037 static int
delete_addrsel_policyent(struct in6_addrpolicy * key)1038 delete_addrsel_policyent(struct in6_addrpolicy *key)
1039 {
1040 struct addrsel_policyent *pol;
1041
1042 ADDRSEL_XLOCK();
1043 ADDRSEL_LOCK();
1044
1045 /* search for the entry in the table */
1046 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1047 if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1048 &pol->ape_policy.addr.sin6_addr) &&
1049 IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1050 &pol->ape_policy.addrmask.sin6_addr)) {
1051 break;
1052 }
1053 }
1054 if (pol == NULL) {
1055 ADDRSEL_UNLOCK();
1056 ADDRSEL_XUNLOCK();
1057 return (ESRCH);
1058 }
1059
1060 TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1061 ADDRSEL_UNLOCK();
1062 ADDRSEL_XUNLOCK();
1063 free(pol, M_IFADDR);
1064
1065 return (0);
1066 }
1067
1068 static int
walk_addrsel_policy(int (* callback)(struct in6_addrpolicy *,void *),void * w)1069 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1070 {
1071 struct addrsel_policyent *pol;
1072 int error = 0;
1073
1074 ADDRSEL_SLOCK();
1075 TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1076 if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1077 ADDRSEL_SUNLOCK();
1078 return (error);
1079 }
1080 }
1081 ADDRSEL_SUNLOCK();
1082 return (error);
1083 }
1084
1085 static int
dump_addrsel_policyent(struct in6_addrpolicy * pol,void * arg)1086 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1087 {
1088 int error = 0;
1089 struct walkarg *w = arg;
1090
1091 error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1092
1093 return (error);
1094 }
1095
1096 static struct in6_addrpolicy *
match_addrsel_policy(struct sockaddr_in6 * key)1097 match_addrsel_policy(struct sockaddr_in6 *key)
1098 {
1099 struct addrsel_policyent *pent;
1100 struct in6_addrpolicy *bestpol = NULL, *pol;
1101 int matchlen, bestmatchlen = -1;
1102 u_char *mp, *ep, *k, *p, m;
1103
1104 TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1105 matchlen = 0;
1106
1107 pol = &pent->ape_policy;
1108 mp = (u_char *)&pol->addrmask.sin6_addr;
1109 ep = mp + 16; /* XXX: scope field? */
1110 k = (u_char *)&key->sin6_addr;
1111 p = (u_char *)&pol->addr.sin6_addr;
1112 for (; mp < ep && *mp; mp++, k++, p++) {
1113 m = *mp;
1114 if ((*k & m) != *p)
1115 goto next; /* not match */
1116 if (m == 0xff) /* short cut for a typical case */
1117 matchlen += 8;
1118 else {
1119 while (m >= 0x80) {
1120 matchlen++;
1121 m <<= 1;
1122 }
1123 }
1124 }
1125
1126 /* matched. check if this is better than the current best. */
1127 if (bestpol == NULL ||
1128 matchlen > bestmatchlen) {
1129 bestpol = pol;
1130 bestmatchlen = matchlen;
1131 }
1132
1133 next:
1134 continue;
1135 }
1136
1137 return (bestpol);
1138 }
1139