xref: /netbsd-src/sys/netinet6/in6_src.c (revision 7dec41f044f3905aa6d3ea2553b711107f0a6cd6)
1 /*	$NetBSD: in6_src.c,v 1.92 2023/08/03 04:24:55 ozaki-r Exp $	*/
2 /*	$KAME: in6_src.c,v 1.159 2005/10/19 01:40:32 t-momose Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
66  */
67 
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: in6_src.c,v 1.92 2023/08/03 04:24:55 ozaki-r Exp $");
70 
71 #ifdef _KERNEL_OPT
72 #include "opt_inet.h"
73 #endif
74 
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/protosw.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/ioctl.h>
83 #include <sys/errno.h>
84 #include <sys/time.h>
85 #include <sys/kernel.h>
86 #include <sys/proc.h>
87 #include <sys/kauth.h>
88 
89 #include <net/if.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 
93 #include <netinet/in.h>
94 #include <netinet/in_var.h>
95 #include <netinet/in_systm.h>
96 #include <netinet/ip.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet/portalgo.h>
99 #include <netinet6/in6_var.h>
100 #include <netinet/ip6.h>
101 #include <netinet6/in6_pcb.h>
102 #include <netinet6/ip6_var.h>
103 #include <netinet6/ip6_private.h>
104 #include <netinet6/nd6.h>
105 #include <netinet6/scope6_var.h>
106 
107 #ifdef MIP6
108 #include <netinet6/mip6.h>
109 #include <netinet6/mip6_var.h>
110 #include "mip.h"
111 #if NMIP > 0
112 #include <net/if_mip.h>
113 #endif /* NMIP > 0 */
114 #endif /* MIP6 */
115 
116 #include <netinet/tcp_vtw.h>
117 
118 #define ADDR_LABEL_NOTAPP (-1)
119 struct in6_addrpolicy defaultaddrpolicy;
120 
121 int ip6_prefer_tempaddr = 0;
122 
123 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
124 	struct ip6_moptions *, struct route *, struct ifnet **, struct psref *);
125 
126 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
127 
128 static void init_policy_queue(void);
129 static int add_addrsel_policyent(struct in6_addrpolicy *);
130 static int delete_addrsel_policyent(struct in6_addrpolicy *);
131 static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *),
132 				    void *);
133 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
134 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
135 
136 #define	IFA6_IS_VALIDATED(ia) \
137 	(((ia)->ia6_flags & (IN6_IFF_TENTATIVE | IN6_IFF_DETACHED)) == 0)
138 
139 /*
140  * Return an IPv6 address, which is the most appropriate for a given
141  * destination and user specified options.
142  * If necessary, this function lookups the routing table and returns
143  * an entry to the caller for later use.
144  */
145 #if 0				/* disabled ad-hoc */
146 #define REPLACE(r) do {\
147 	char _buf1[INET6_ADDRSTRLEN], _buf2[INET6_ADDRSTRLEN]; \
148 	if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
149 		sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
150 		ip6stat.ip6s_sources_rule[(r)]++; \
151 	printf("%s: replace %s with %s by %d\n", __func__, ia_best ? \
152 	    IN6_PRINT(_buf1, &ia_best->ia_addr.sin6_addr) : "none", \
153 	    IN6_PRINT(_buf2, &ia->ia_addr.sin6_addr), (r)); \
154 	goto replace; \
155 } while(/*CONSTCOND*/0)
156 #define NEXT(r) do {\
157 	if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
158 		sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
159 		ip6stat.ip6s_sources_rule[(r)]++; \
160 	printf("%s: keep %s against %s by %d\n", ia_best ? \
161 	    IN6_PRINT(_buf1, &ia_best->ia_addr.sin6_addr) : "none", \
162 	    IN6_PRINT(_buf2, &ia->ia_addr.sin6_addr), (r)); \
163 	goto next; 		/* XXX: we can't use 'continue' here */ \
164 } while(/*CONSTCOND*/0)
165 #define BREAK(r) do { \
166 	if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
167 		sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
168 		ip6stat.ip6s_sources_rule[(r)]++; \
169 	goto out; 		/* XXX: we can't use 'break' here */ \
170 } while(/*CONSTCOND*/0)
171 #else
172 #define REPLACE(r) goto replace
173 #define NEXT(r) goto next
174 #define BREAK(r) goto out
175 #endif
176 
177 /*
178  * Called inside pserialize critical section. Don't sleep/block.
179  */
180 static struct in6_ifaddr *
in6_select_best_ia(struct sockaddr_in6 * dstsock,struct in6_addr * dst,const struct ifnet * ifp,const struct ip6_pktopts * opts,const u_int32_t odstzone)181 in6_select_best_ia(struct sockaddr_in6 *dstsock, struct in6_addr *dst,
182     const struct ifnet *ifp, const struct ip6_pktopts *opts,
183     const u_int32_t odstzone)
184 {
185 	struct in6_ifaddr *ia, *ia_best = NULL;
186 	int dst_scope = -1, best_scope = -1, best_matchlen = -1;
187 	struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
188 
189 	IN6_ADDRLIST_READER_FOREACH(ia) {
190 		int new_scope = -1, new_matchlen = -1;
191 		struct in6_addrpolicy *new_policy = NULL;
192 		u_int32_t srczone, osrczone, dstzone;
193 		struct in6_addr src;
194 		struct ifnet *ifp1 = ia->ia_ifp;
195 		int prefer_tempaddr;
196 
197 		/*
198 		 * We'll never take an address that breaks the scope zone
199 		 * of the destination.  We also skip an address if its zone
200 		 * does not contain the outgoing interface.
201 		 * XXX: we should probably use sin6_scope_id here.
202 		 */
203 		if (in6_setscope(dst, ifp1, &dstzone) ||
204 		    odstzone != dstzone) {
205 			continue;
206 		}
207 		src = ia->ia_addr.sin6_addr;
208 
209 		/* Skip the scope test in impossible cases */
210 		if (!(ifp->if_flags & IFF_LOOPBACK) &&
211 		    IN6_IS_ADDR_LOOPBACK(&src))
212 			continue;
213 
214 		if (in6_setscope(&src, ifp, &osrczone) ||
215 		    in6_setscope(&src, ifp1, &srczone) ||
216 		    osrczone != srczone) {
217 			continue;
218 		}
219 
220 		/* avoid unusable addresses */
221 		if ((ia->ia6_flags & (IN6_IFF_DUPLICATED | IN6_IFF_ANYCAST)))
222 			continue;
223 		if (!ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
224 			continue;
225 
226 #if defined(MIP6) && NMIP > 0
227 		/* avoid unusable home addresses. */
228 		if ((ia->ia6_flags & IN6_IFF_HOME) &&
229 		    !mip6_ifa6_is_addr_valid_hoa(ia))
230 			continue;
231 #endif /* MIP6 && NMIP > 0 */
232 
233 		/* Rule 1: Prefer same address */
234 		if (IN6_ARE_ADDR_EQUAL(dst, &ia->ia_addr.sin6_addr)) {
235 			ia_best = ia;
236 			BREAK(1); /* there should be no better candidate */
237 		}
238 
239 		if (ia_best == NULL)
240 			REPLACE(1);
241 
242 		/* Rule 2: Prefer appropriate scope */
243 		if (dst_scope < 0)
244 			dst_scope = in6_addrscope(dst);
245 		new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
246 		if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
247 			if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
248 				REPLACE(2);
249 			NEXT(2);
250 		} else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
251 			if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
252 				NEXT(2);
253 			REPLACE(2);
254 		}
255 
256 		/*
257 		 * Rule 3: Avoid deprecated addresses.  Note that the case of
258 		 * !ip6_use_deprecated is already rejected above.
259 		 * Treat unvalidated addresses as deprecated here.
260 		 */
261 		if (IFA6_IS_VALIDATED(ia_best) && !IFA6_IS_VALIDATED(ia))
262 			NEXT(3);
263 		if (!IFA6_IS_VALIDATED(ia_best) && IFA6_IS_VALIDATED(ia))
264 			REPLACE(3);
265 		if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
266 			NEXT(3);
267 		if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
268 			REPLACE(3);
269 
270 		/* Rule 4: Prefer home addresses */
271 #if defined(MIP6) && NMIP > 0
272 		if (!MIP6_IS_MN)
273 			goto skip_rule4;
274 
275 		if ((ia_best->ia6_flags & IN6_IFF_HOME) == 0 &&
276 		    (ia->ia6_flags & IN6_IFF_HOME) == 0) {
277 			/* both address are not home addresses. */
278 			goto skip_rule4;
279 		}
280 
281 		/*
282 		 * If SA is simultaneously a home address and care-of
283 		 * address and SB is not, then prefer SA. Similarly,
284 		 * if SB is simultaneously a home address and care-of
285 		 * address and SA is not, then prefer SB.
286 		 */
287 		if (((ia_best->ia6_flags & IN6_IFF_HOME) != 0 &&
288 			ia_best->ia_ifp->if_type != IFT_MIP)
289 		    &&
290 		    ((ia->ia6_flags & IN6_IFF_HOME) != 0 &&
291 			ia->ia_ifp->if_type == IFT_MIP))
292 			NEXT(4);
293 		if (((ia_best->ia6_flags & IN6_IFF_HOME) != 0 &&
294 			ia_best->ia_ifp->if_type == IFT_MIP)
295 		    &&
296 		    ((ia->ia6_flags & IN6_IFF_HOME) != 0 &&
297 			ia->ia_ifp->if_type != IFT_MIP))
298 			REPLACE(4);
299 		if (ip6po_usecoa == 0) {
300 			/*
301 			 * If SA is just a home address and SB is just
302 			 * a care-of address, then prefer
303 			 * SA. Similarly, if SB is just a home address
304 			 * and SA is just a care-of address, then
305 			 * prefer SB.
306 			 */
307 			if ((ia_best->ia6_flags & IN6_IFF_HOME) != 0 &&
308 			    (ia->ia6_flags & IN6_IFF_HOME) == 0) {
309 				NEXT(4);
310 			}
311 			if ((ia_best->ia6_flags & IN6_IFF_HOME) == 0 &&
312 			    (ia->ia6_flags & IN6_IFF_HOME) != 0) {
313 				REPLACE(4);
314 			}
315 		} else {
316 			/*
317 			 * a sender don't want to use a home address
318 			 * because:
319 			 *
320 			 * 1) we cannot use.  (ex. NS or NA to global
321 			 * addresses.)
322 			 *
323 			 * 2) a user specified not to use.
324 			 * (ex. mip6control -u)
325 			 */
326 			if ((ia_best->ia6_flags & IN6_IFF_HOME) == 0 &&
327 			    (ia->ia6_flags & IN6_IFF_HOME) != 0) {
328 				/* XXX breaks stat */
329 				NEXT(0);
330 			}
331 			if ((ia_best->ia6_flags & IN6_IFF_HOME) != 0 &&
332 			    (ia->ia6_flags & IN6_IFF_HOME) == 0) {
333 				/* XXX breaks stat */
334 				REPLACE(0);
335 			}
336 		}
337 	skip_rule4:
338 #endif /* MIP6 && NMIP > 0 */
339 
340 		/* Rule 5: Prefer outgoing interface */
341 		if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
342 			NEXT(5);
343 		if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
344 			REPLACE(5);
345 
346 		/*
347 		 * Rule 6: Prefer matching label
348 		 * Note that best_policy should be non-NULL here.
349 		 */
350 		if (dst_policy == NULL)
351 			dst_policy = lookup_addrsel_policy(dstsock);
352 		if (dst_policy->label != ADDR_LABEL_NOTAPP) {
353 			new_policy = lookup_addrsel_policy(&ia->ia_addr);
354 			if (dst_policy->label == best_policy->label &&
355 			    dst_policy->label != new_policy->label)
356 				NEXT(6);
357 			if (dst_policy->label != best_policy->label &&
358 			    dst_policy->label == new_policy->label)
359 				REPLACE(6);
360 		}
361 
362 		/*
363 		 * Rule 7: Prefer public addresses.
364 		 * We allow users to reverse the logic by configuring
365 		 * a sysctl variable, so that privacy conscious users can
366 		 * always prefer temporary addresses.
367 		 */
368 		if (opts == NULL ||
369 		    opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
370 			prefer_tempaddr = ip6_prefer_tempaddr;
371 		} else if (opts->ip6po_prefer_tempaddr ==
372 		    IP6PO_TEMPADDR_NOTPREFER) {
373 			prefer_tempaddr = 0;
374 		} else
375 			prefer_tempaddr = 1;
376 		if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
377 		    (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
378 			if (prefer_tempaddr)
379 				REPLACE(7);
380 			else
381 				NEXT(7);
382 		}
383 		if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
384 		    !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
385 			if (prefer_tempaddr)
386 				NEXT(7);
387 			else
388 				REPLACE(7);
389 		}
390 
391 		/*
392 		 * Rule 8: prefer addresses on alive interfaces.
393 		 * This is a KAME specific rule.
394 		 */
395 		if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
396 		    !(ia->ia_ifp->if_flags & IFF_UP))
397 			NEXT(8);
398 		if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
399 		    (ia->ia_ifp->if_flags & IFF_UP))
400 			REPLACE(8);
401 
402 		/*
403 		 * Rule 9: prefer addresses on "preferred" interfaces.
404 		 * This is a KAME specific rule.
405 		 */
406 #ifdef notyet			/* until introducing address selection */
407 #define NDI_BEST ND_IFINFO(ia_best->ia_ifp)
408 #define NDI_NEW  ND_IFINFO(ia->ia_ifp)
409 		if ((NDI_BEST->flags & ND6_IFF_PREFER_SOURCE) &&
410 		    !(NDI_NEW->flags & ND6_IFF_PREFER_SOURCE))
411 			NEXT(9);
412 		if (!(NDI_BEST->flags & ND6_IFF_PREFER_SOURCE) &&
413 		    (NDI_NEW->flags & ND6_IFF_PREFER_SOURCE))
414 			REPLACE(9);
415 #undef NDI_BEST
416 #undef NDI_NEW
417 #endif
418 
419 		/*
420 		 * Rule 14: Use longest matching prefix.
421 		 * Note: in the address selection draft, this rule is
422 		 * documented as "Rule 8".  However, since it is also
423 		 * documented that this rule can be overridden, we assign
424 		 * a large number so that it is easy to assign smaller numbers
425 		 * to more preferred rules.
426 		 */
427 		new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, dst);
428 		if (best_matchlen < new_matchlen)
429 			REPLACE(14);
430 		if (new_matchlen < best_matchlen)
431 			NEXT(14);
432 
433 		/* Rule 15 is reserved. */
434 
435 		/*
436 		 * Last resort: just keep the current candidate.
437 		 * Or, do we need more rules?
438 		 */
439 		continue;
440 
441 	  replace:
442 		ia_best = ia;
443 		best_scope = (new_scope >= 0 ? new_scope :
444 			      in6_addrscope(&ia_best->ia_addr.sin6_addr));
445 		best_policy = (new_policy ? new_policy :
446 			       lookup_addrsel_policy(&ia_best->ia_addr));
447 		best_matchlen = (new_matchlen >= 0 ? new_matchlen :
448 				 in6_matchlen(&ia_best->ia_addr.sin6_addr,
449 					      dst));
450 
451 	  next:
452 		continue;
453 
454 	  out:
455 		break;
456 	}
457 
458 	return ia_best;
459 }
460 #undef REPLACE
461 #undef BREAK
462 #undef NEXT
463 
464 int
in6_selectsrc(struct sockaddr_in6 * dstsock,struct ip6_pktopts * opts,struct ip6_moptions * mopts,struct route * ro,struct in6_addr * laddr,struct ifnet ** ifpp,struct psref * psref,struct in6_addr * ret_ia6)465 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
466 	struct ip6_moptions *mopts, struct route *ro, struct in6_addr *laddr,
467 	struct ifnet **ifpp, struct psref *psref, struct in6_addr *ret_ia6)
468 {
469 	struct in6_addr dst;
470 	struct ifnet *ifp = NULL;
471 	struct in6_ifaddr *ia = NULL;
472 	struct in6_pktinfo *pi = NULL;
473 	u_int32_t odstzone;
474 	int error = 0, iferror;
475 #if defined(MIP6) && NMIP > 0
476 	u_int8_t ip6po_usecoa = 0;
477 #endif /* MIP6 && NMIP > 0 */
478 	struct psref local_psref;
479 	int bound = curlwp_bind();
480 #define PSREF (psref == NULL) ? &local_psref : psref
481 	int s;
482 
483 	KASSERT((ifpp != NULL && psref != NULL) ||
484 	        (ifpp == NULL && psref == NULL));
485 
486 	dst = dstsock->sin6_addr; /* make a copy for local operation */
487 	if (ifpp)
488 		*ifpp = NULL;
489 
490 	/*
491 	 * Try to determine the outgoing interface for the given destination.
492 	 * We do this regardless of whether the socket is bound, since the
493 	 * caller may need this information as a side effect of the call
494 	 * to this function (e.g., for identifying the appropriate scope zone
495 	 * ID).
496 	 */
497 	iferror = in6_selectif(dstsock, opts, mopts, ro, &ifp, PSREF);
498 	if (ifpp != NULL)
499 		*ifpp = ifp;
500 
501 	/*
502 	 * If the source address is explicitly specified by the caller,
503 	 * check if the requested source address is indeed a unicast address
504 	 * assigned to the node, and can be used as the packet's source
505 	 * address.  If everything is okay, use the address as source.
506 	 */
507 	if (opts && (pi = opts->ip6po_pktinfo) &&
508 	    !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
509 		struct sockaddr_in6 srcsock;
510 		struct in6_ifaddr *ia6;
511 		int _s;
512 		struct ifaddr *ifa;
513 
514 		/*
515 		 * Determine the appropriate zone id of the source based on
516 		 * the zone of the destination and the outgoing interface.
517 		 * If the specified address is ambiguous wrt the scope zone,
518 		 * the interface must be specified; otherwise, ifa_ifwithaddr()
519 		 * will fail matching the address.
520 		 */
521 		memset(&srcsock, 0, sizeof(srcsock));
522 		srcsock.sin6_family = AF_INET6;
523 		srcsock.sin6_len = sizeof(srcsock);
524 		srcsock.sin6_addr = pi->ipi6_addr;
525 		if (ifp) {
526 			error = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
527 			if (error != 0)
528 				goto exit;
529 		}
530 
531 		_s = pserialize_read_enter();
532 		ifa = ifa_ifwithaddr(sin6tosa(&srcsock));
533 		if ((ia6 = ifatoia6(ifa)) == NULL ||
534 		    ia6->ia6_flags &
535 		    (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY)) {
536 			pserialize_read_exit(_s);
537 			error = EADDRNOTAVAIL;
538 			goto exit;
539 		}
540 		pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
541 		if (ifpp)
542 			*ifpp = ifp;
543 		*ret_ia6 = ia6->ia_addr.sin6_addr;
544 		pserialize_read_exit(_s);
545 		goto exit;
546 	}
547 
548 	/*
549 	 * If the socket has already bound the source, just use it.  We don't
550 	 * care at the moment whether in6_selectif() succeeded above, even
551 	 * though it would eventually cause an error.
552 	 */
553 	if (laddr && !IN6_IS_ADDR_UNSPECIFIED(laddr)) {
554 		*ret_ia6 = *laddr;
555 		goto exit;
556 	}
557 
558 	/*
559 	 * The outgoing interface is crucial in the general selection procedure
560 	 * below.  If it is not known at this point, we fail.
561 	 */
562 	if (ifp == NULL) {
563 		error = iferror;
564 		goto exit;
565 	}
566 
567 	/*
568 	 * If the address is not yet determined, choose the best one based on
569 	 * the outgoing interface and the destination address.
570 	 */
571 
572 #if defined(MIP6) && NMIP > 0
573 	/*
574 	 * a caller can specify IP6PO_USECOA to not to use a home
575 	 * address.  for example, the case that the neighbour
576 	 * unreachability detection to the global address.
577 	 */
578 	if (opts != NULL &&
579 	    (opts->ip6po_flags & IP6PO_USECOA) != 0) {
580 		ip6po_usecoa = 1;
581 	}
582 #endif /* MIP6 && NMIP > 0 */
583 
584 	error = in6_setscope(&dst, ifp, &odstzone);
585 	if (error != 0)
586 		goto exit;
587 
588 	s = pserialize_read_enter();
589 
590 	ia = in6_select_best_ia(dstsock, &dst, ifp, opts, odstzone);
591 	if (ia == NULL) {
592 		pserialize_read_exit(s);
593 		error = EADDRNOTAVAIL;
594 		goto exit;
595 	}
596 	*ret_ia6 = ia->ia_addr.sin6_addr;
597 
598 	pserialize_read_exit(s);
599 exit:
600 	if (ifpp == NULL)
601 		if_put(ifp, PSREF);
602 	curlwp_bindx(bound);
603 	return error;
604 #undef PSREF
605 }
606 
607 int
in6_selectroute(struct sockaddr_in6 * dstsock,struct ip6_pktopts * opts,struct route ** ro,struct rtentry ** retrt,bool count_discard)608 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
609     struct route **ro, struct rtentry **retrt, bool count_discard)
610 {
611 	int error = 0;
612 	struct rtentry *rt = NULL;
613 	union {
614 		struct sockaddr		dst;
615 		struct sockaddr_in	dst4;
616 		struct sockaddr_in6	dst6;
617 	} u;
618 
619 	KASSERT(ro != NULL);
620 	KASSERT(*ro != NULL);
621 	KASSERT(retrt != NULL);
622 
623 #if 0
624 	if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
625 	    dstsock->sin6_addr.s6_addr32[1] == 0 &&
626 	    !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
627 		char ip6buf[INET6_ADDRSTRLEN];
628 		printf("%s: strange destination %s\n", __func__,
629 		       IN6_PRINT(ip6buf, &dstsock->sin6_addr));
630 	} else {
631 		char ip6buf[INET6_ADDRSTRLEN];
632 		printf("%s: destination = %s%%%d\n", __func__,
633 		       IN6_PRINT(ip6buf, &dstsock->sin6_addr),
634 		       dstsock->sin6_scope_id); /* for debug */
635 	}
636 #endif
637 
638 	/*
639 	 * If the next hop address for the packet is specified by the caller,
640 	 * use it as the gateway.
641 	 */
642 	if (opts && opts->ip6po_nexthop) {
643 		struct route *ron;
644 		struct sockaddr_in6 *sin6_next;
645 
646 		sin6_next = satosin6(opts->ip6po_nexthop);
647 
648 		/* at this moment, we only support AF_INET6 next hops */
649 		if (sin6_next->sin6_family != AF_INET6) {
650 			IP6_STATINC(IP6_STAT_ODROPPED);
651 			error = EAFNOSUPPORT; /* or should we proceed? */
652 			goto done;
653 		}
654 
655 		/*
656 		 * If the next hop is an IPv6 address, then the node identified
657 		 * by that address must be a neighbor of the sending host.
658 		 */
659 		ron = &opts->ip6po_nextroute;
660 		rt = rtcache_lookup(ron, sin6tosa(sin6_next));
661 		if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) != 0 ||
662 		    !nd6_is_addr_neighbor(sin6_next, rt->rt_ifp)) {
663 			if (rt != NULL) {
664 				if (count_discard)
665 					in6_ifstat_inc(rt->rt_ifp,
666 					    ifs6_out_discard);
667 				rtcache_unref(rt, ron);
668 				rt = NULL;
669 			}
670 			rtcache_free(ron);
671 			error = EHOSTUNREACH;
672 			goto done;
673 		}
674 		*ro = ron;
675 
676 		goto done;
677 	}
678 
679 	/*
680 	 * Use a cached route if it exists and is valid, else try to allocate
681 	 * a new one.  Note that we should check the address family of the
682 	 * cached destination, in case of sharing the cache with IPv4.
683 	 *
684 	 * for V4 mapped addresses we want to pick up the v4 route
685 	 * see PR kern/56348
686 	 */
687 	if (IN6_IS_ADDR_V4MAPPED(&dstsock->sin6_addr)) {
688 		in6_sin6_2_sin(&u.dst4, dstsock);
689 	} else {
690 		u.dst6 = *dstsock;
691 		u.dst6.sin6_scope_id = 0;
692 	}
693 
694 	rt = rtcache_lookup1(*ro, &u.dst, 1);
695 
696 	if (rt == NULL)
697 		error = EHOSTUNREACH;
698 
699 	/*
700 	 * Check if the outgoing interface conflicts with
701 	 * the interface specified by ipi6_ifindex (if specified).
702 	 * Note that loopback interface is always okay.
703 	 * (this may happen when we are sending a packet to one of
704 	 *  our own addresses.)
705 	 */
706 	if (opts && opts->ip6po_pktinfo && opts->ip6po_pktinfo->ipi6_ifindex) {
707 		if (rt != NULL && !(rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
708 		    rt->rt_ifp->if_index != opts->ip6po_pktinfo->ipi6_ifindex) {
709 			if (count_discard)
710 				in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
711 			error = EHOSTUNREACH;
712 			rtcache_unref(rt, *ro);
713 			rt = NULL;
714 		}
715 	}
716 
717 done:
718 	if (error == EHOSTUNREACH)
719 		IP6_STATINC(IP6_STAT_NOROUTE);
720 	*retrt = rt;
721 	return error;
722 }
723 
724 static int
in6_selectif(struct sockaddr_in6 * dstsock,struct ip6_pktopts * opts,struct ip6_moptions * mopts,struct route * ro,struct ifnet ** retifp,struct psref * psref)725 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
726 	struct ip6_moptions *mopts, struct route *ro, struct ifnet **retifp,
727 	struct psref *psref)
728 {
729 	int error = 0;
730 	struct rtentry *rt = NULL;
731 	struct in6_addr *dst;
732 	struct in6_pktinfo *pi = NULL;
733 
734 	KASSERT(retifp != NULL);
735 	*retifp = NULL;
736 	dst = &dstsock->sin6_addr;
737 
738 	/* If the caller specify the outgoing interface explicitly, use it. */
739 	if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
740 		/* XXX boundary check is assumed to be already done. */
741 		*retifp = if_get_byindex(pi->ipi6_ifindex, psref);
742 		if (*retifp != NULL)
743 			return 0;
744 		goto getroute;
745 	}
746 
747 	/*
748 	 * If the destination address is a multicast address and the outgoing
749 	 * interface for the address is specified by the caller, use it.
750 	 */
751 	if (IN6_IS_ADDR_MULTICAST(dst) && mopts != NULL) {
752 		*retifp = if_get_byindex(mopts->im6o_multicast_if_index, psref);
753 		if (*retifp != NULL)
754 			return 0; /* we do not need a route for multicast. */
755 	}
756 
757 getroute:
758 	error = in6_selectroute(dstsock, opts, &ro, &rt, false);
759 	if (error != 0)
760 		return error;
761 
762 	*retifp = if_get_byindex(rt->rt_ifp->if_index, psref);
763 
764 	/*
765 	 * do not use a rejected or black hole route.
766 	 * XXX: this check should be done in the L2 output routine.
767 	 * However, if we skipped this check here, we'd see the following
768 	 * scenario:
769 	 * - install a rejected route for a scoped address prefix
770 	 *   (like fe80::/10)
771 	 * - send a packet to a destination that matches the scoped prefix,
772 	 *   with ambiguity about the scope zone.
773 	 * - pick the outgoing interface from the route, and disambiguate the
774 	 *   scope zone with the interface.
775 	 * - ip6_output() would try to get another route with the "new"
776 	 *   destination, which may be valid.
777 	 * - we'd see no error on output.
778 	 * Although this may not be very harmful, it should still be confusing.
779 	 * We thus reject the case here.
780 	 */
781 	if ((rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
782 		error = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
783 		/* XXX: ifp can be returned with psref even if error */
784 		goto out;
785 	}
786 
787 	/*
788 	 * Adjust the "outgoing" interface.  If we're going to loop the packet
789 	 * back to ourselves, the ifp would be the loopback interface.
790 	 * However, we'd rather know the interface associated to the
791 	 * destination address (which should probably be one of our own
792 	 * addresses.)
793 	 */
794 	if (rt->rt_ifa->ifa_ifp != *retifp &&
795 	    !if_is_deactivated(rt->rt_ifa->ifa_ifp)) {
796 		if_put(*retifp, psref);
797 		*retifp = rt->rt_ifa->ifa_ifp;
798 		if_acquire(*retifp, psref);
799 	}
800 out:
801 	rtcache_unref(rt, ro);
802 	return error;
803 }
804 
805 /*
806  * Default hop limit selection. The precedence is as follows:
807  * 1. Hoplimit value specified via ioctl.
808  * 2. (If the outgoing interface is detected) the current
809  *     hop limit of the interface specified by router advertisement.
810  * 3. The system default hoplimit.
811 */
812 int
in6pcb_selecthlim(struct inpcb * inp,struct ifnet * ifp)813 in6pcb_selecthlim(struct inpcb *inp, struct ifnet *ifp)
814 {
815 	if (inp && in6p_hops6(inp) >= 0)
816 		return in6p_hops6(inp);
817 	else if (ifp)
818 		return (ND_IFINFO(ifp)->chlim);
819 	else
820 		return (ip6_defhlim);
821 }
822 
823 int
in6pcb_selecthlim_rt(struct inpcb * inp)824 in6pcb_selecthlim_rt(struct inpcb *inp)
825 {
826 	struct rtentry *rt;
827 
828 	if (inp == NULL)
829 		return in6pcb_selecthlim(inp, NULL);
830 
831 	rt = rtcache_validate(&inp->inp_route);
832 	if (rt != NULL) {
833 		int ret = in6pcb_selecthlim(inp, rt->rt_ifp);
834 		rtcache_unref(rt, &inp->inp_route);
835 		return ret;
836 	} else
837 		return in6pcb_selecthlim(inp, NULL);
838 }
839 
840 /*
841  * Find an empty port and set it to the specified PCB.
842  */
843 int
in6pcb_set_port(struct sockaddr_in6 * sin6,struct inpcb * inp,struct lwp * l)844 in6pcb_set_port(struct sockaddr_in6 *sin6, struct inpcb *inp, struct lwp *l)
845 {
846 	struct socket *so = inp->inp_socket;
847 	struct inpcbtable *table = inp->inp_table;
848 	u_int16_t lport, *lastport;
849 	enum kauth_network_req req;
850 	int error = 0;
851 
852 	if (inp->inp_flags & IN6P_LOWPORT) {
853 #ifndef IPNOPRIVPORTS
854 		req = KAUTH_REQ_NETWORK_BIND_PRIVPORT;
855 #else
856 		req = KAUTH_REQ_NETWORK_BIND_PORT;
857 #endif
858 		lastport = &table->inpt_lastlow;
859 	} else {
860 		req = KAUTH_REQ_NETWORK_BIND_PORT;
861 
862 		lastport = &table->inpt_lastport;
863 	}
864 
865 	/* XXX-kauth: KAUTH_REQ_NETWORK_BIND_AUTOASSIGN_{,PRIV}PORT */
866 	error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_BIND, req, so,
867 	    sin6, NULL);
868 	if (error)
869 		return (EACCES);
870 
871        /*
872         * Use RFC6056 randomized port selection
873         */
874 	error = portalgo_randport(&lport, inp, l->l_cred);
875 	if (error)
876 		return error;
877 
878 	inp->inp_flags |= IN6P_ANONPORT;
879 	*lastport = lport;
880 	inp->inp_lport = htons(lport);
881 	in6pcb_set_state(inp, INP_BOUND);
882 	return (0);		/* success */
883 }
884 
885 void
addrsel_policy_init(void)886 addrsel_policy_init(void)
887 {
888 	init_policy_queue();
889 
890 	/* initialize the "last resort" policy */
891 	memset(&defaultaddrpolicy, 0, sizeof(defaultaddrpolicy));
892 	defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
893 }
894 
895 /*
896  * XXX: NOMPSAFE if a policy is set
897  */
898 static struct in6_addrpolicy *
lookup_addrsel_policy(struct sockaddr_in6 * key)899 lookup_addrsel_policy(struct sockaddr_in6 *key)
900 {
901 	struct in6_addrpolicy *match = NULL;
902 
903 	match = match_addrsel_policy(key);
904 
905 	if (match == NULL)
906 		match = &defaultaddrpolicy;
907 	else
908 		match->use++;
909 
910 	return (match);
911 }
912 
913 /*
914  * Subroutines to manage the address selection policy table via sysctl.
915  */
916 struct sel_walkarg {
917 	size_t	w_total;
918 	size_t	w_given;
919 	void *	w_where;
920 	void *w_limit;
921 };
922 
923 int sysctl_net_inet6_addrctlpolicy(SYSCTLFN_ARGS);
924 int
sysctl_net_inet6_addrctlpolicy(SYSCTLFN_ARGS)925 sysctl_net_inet6_addrctlpolicy(SYSCTLFN_ARGS)
926 {
927 	int error = 0;
928 	int s;
929 
930 	s = splsoftnet();
931 
932 	if (newp) {
933 		error = EPERM;
934 		goto end;
935 	}
936 	if (oldp && oldlenp == NULL) {
937 		error = EINVAL;
938 		goto end;
939 	}
940 	if (oldp || oldlenp) {
941 		struct sel_walkarg w;
942 		size_t oldlen = *oldlenp;
943 
944 		memset(&w, 0, sizeof(w));
945 		w.w_given = oldlen;
946 		w.w_where = oldp;
947 		if (oldp)
948 			w.w_limit = (char *)oldp + oldlen;
949 
950 		error = walk_addrsel_policy(dump_addrsel_policyent, &w);
951 
952 		*oldlenp = w.w_total;
953 		if (oldp && w.w_total > oldlen && error == 0)
954 			error = ENOMEM;
955 	}
956 
957   end:
958 	splx(s);
959 
960 	return (error);
961 }
962 
963 int
in6_src_ioctl(u_long cmd,void * data)964 in6_src_ioctl(u_long cmd, void *data)
965 {
966 	int i;
967 	struct in6_addrpolicy ent0;
968 
969 	if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
970 		return (EOPNOTSUPP); /* check for safety */
971 
972 	ent0 = *(struct in6_addrpolicy *)data;
973 
974 	if (ent0.label == ADDR_LABEL_NOTAPP)
975 		return (EINVAL);
976 	/* check if the prefix mask is consecutive. */
977 	if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
978 		return (EINVAL);
979 	/* clear trailing garbages (if any) of the prefix address. */
980 	for (i = 0; i < 4; i++) {
981 		ent0.addr.sin6_addr.s6_addr32[i] &=
982 			ent0.addrmask.sin6_addr.s6_addr32[i];
983 	}
984 	ent0.use = 0;
985 
986 	switch (cmd) {
987 	case SIOCAADDRCTL_POLICY:
988 		return (add_addrsel_policyent(&ent0));
989 	case SIOCDADDRCTL_POLICY:
990 		return (delete_addrsel_policyent(&ent0));
991 	}
992 
993 	return (0);		/* XXX: compromise compilers */
994 }
995 
996 /*
997  * The followings are implementation of the policy table using a
998  * simple tail queue.
999  * XXX such details should be hidden.
1000  * XXX implementation using binary tree should be more efficient.
1001  */
1002 struct addrsel_policyent {
1003 	TAILQ_ENTRY(addrsel_policyent) ape_entry;
1004 	struct in6_addrpolicy ape_policy;
1005 };
1006 
1007 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1008 
1009 struct addrsel_policyhead addrsel_policytab;
1010 
1011 static void
init_policy_queue(void)1012 init_policy_queue(void)
1013 {
1014 	TAILQ_INIT(&addrsel_policytab);
1015 }
1016 
1017 static int
add_addrsel_policyent(struct in6_addrpolicy * newpolicy)1018 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1019 {
1020 	struct addrsel_policyent *newpol, *pol;
1021 
1022 	/* duplication check */
1023 	TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
1024 		if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1025 		    &pol->ape_policy.addr.sin6_addr) &&
1026 		    IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1027 		    &pol->ape_policy.addrmask.sin6_addr)) {
1028 			return (EEXIST);	/* or override it? */
1029 		}
1030 	}
1031 
1032 	newpol = malloc(sizeof(*newpol), M_IFADDR, M_WAITOK|M_ZERO);
1033 
1034 	/* XXX: should validate entry */
1035 	newpol->ape_policy = *newpolicy;
1036 
1037 	TAILQ_INSERT_TAIL(&addrsel_policytab, newpol, ape_entry);
1038 
1039 	return (0);
1040 }
1041 
1042 static int
delete_addrsel_policyent(struct in6_addrpolicy * key)1043 delete_addrsel_policyent(struct in6_addrpolicy *key)
1044 {
1045 	struct addrsel_policyent *pol;
1046 
1047 	/* search for the entry in the table */
1048 	for (pol = TAILQ_FIRST(&addrsel_policytab); pol;
1049 	     pol = TAILQ_NEXT(pol, ape_entry)) {
1050 		if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1051 		    &pol->ape_policy.addr.sin6_addr) &&
1052 		    IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1053 		    &pol->ape_policy.addrmask.sin6_addr)) {
1054 			break;
1055 		}
1056 	}
1057 	if (pol == NULL) {
1058 		return (ESRCH);
1059 	}
1060 
1061 	TAILQ_REMOVE(&addrsel_policytab, pol, ape_entry);
1062 
1063 	return (0);
1064 }
1065 
1066 static int
walk_addrsel_policy(int (* callback)(struct in6_addrpolicy *,void *),void * w)1067 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1068 {
1069 	struct addrsel_policyent *pol;
1070 	int error = 0;
1071 
1072 	TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
1073 		if ((error = (*callback)(&pol->ape_policy, w)) != 0)
1074 			return error;
1075 	}
1076 
1077 	return error;
1078 }
1079 
1080 static int
dump_addrsel_policyent(struct in6_addrpolicy * pol,void * arg)1081 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1082 {
1083 	int error = 0;
1084 	struct sel_walkarg *w = arg;
1085 
1086 	if (w->w_where && (char *)w->w_where + sizeof(*pol) <= (char *)w->w_limit) {
1087 		if ((error = copyout(pol, w->w_where, sizeof(*pol))) != 0)
1088 			return error;
1089 		w->w_where = (char *)w->w_where + sizeof(*pol);
1090 	}
1091 	w->w_total += 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 	for (pent = TAILQ_FIRST(&addrsel_policytab); pent;
1105 	     pent = TAILQ_NEXT(pent, ape_entry)) {
1106 		matchlen = 0;
1107 
1108 		pol = &pent->ape_policy;
1109 		mp = (u_char *)&pol->addrmask.sin6_addr;
1110 		ep = mp + 16;	/* XXX: scope field? */
1111 		k = (u_char *)&key->sin6_addr;
1112 		p = (u_char *)&pol->addr.sin6_addr;
1113 		for (; mp < ep && *mp; mp++, k++, p++) {
1114 			m = *mp;
1115 			if ((*k & m) != *p)
1116 				goto next; /* not match */
1117 			if (m == 0xff) /* short cut for a typical case */
1118 				matchlen += 8;
1119 			else {
1120 				while (m >= 0x80) {
1121 					matchlen++;
1122 					m <<= 1;
1123 				}
1124 			}
1125 		}
1126 
1127 		/* matched.  check if this is better than the current best. */
1128 		if (bestpol == NULL ||
1129 		    matchlen > bestmatchlen) {
1130 			bestpol = pol;
1131 			bestmatchlen = matchlen;
1132 		}
1133 
1134 	  next:
1135 		continue;
1136 	}
1137 
1138 	return (bestpol);
1139 }
1140