xref: /netbsd-src/sys/netinet/ip_encap.c (revision 4b896b232495b7a9b8b94a1cf1e21873296d53b8)
1 /*	$KAME: ip_encap.c,v 1.73 2001/10/02 08:30:58 itojun Exp $	*/
2 
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 /*
32  * My grandfather said that there's a devil inside tunnelling technology...
33  *
34  * We have surprisingly many protocols that want packets with IP protocol
35  * #4 or #41.  Here's a list of protocols that want protocol #41:
36  *	RFC1933 configured tunnel
37  *	RFC1933 automatic tunnel
38  *	RFC2401 IPsec tunnel
39  *	RFC2473 IPv6 generic packet tunnelling
40  *	RFC2529 6over4 tunnel
41  *	RFC3056 6to4 tunnel
42  *	isatap tunnel
43  *	mobile-ip6 (uses RFC2473)
44  * Here's a list of protocol that want protocol #4:
45  *	RFC1853 IPv4-in-IPv4 tunnelling
46  *	RFC2003 IPv4 encapsulation within IPv4
47  *	RFC2344 reverse tunnelling for mobile-ip4
48  *	RFC2401 IPsec tunnel
49  * Well, what can I say.  They impose different en/decapsulation mechanism
50  * from each other, so they need separate protocol handler.  The only one
51  * we can easily determine by protocol # is IPsec, which always has
52  * AH/ESP/IPComp header right after outer IP header.
53  *
54  * So, clearly good old protosw does not work for protocol #4 and #41.
55  * The code will let you match protocol via src/dst address pair.
56  */
57 /* XXX is M_NETADDR correct? */
58 
59 /*
60  * With USE_RADIX the code will use radix table for tunnel lookup, for
61  * tunnels registered with encap_attach() with a addr/mask pair.
62  * Faster on machines with thousands of tunnel registerations (= interfaces).
63  *
64  * The code assumes that radix table code can handle non-continuous netmask,
65  * as it will pass radix table memory region with (src + dst) sockaddr pair.
66  *
67  * FreeBSD is excluded here as they make max_keylen a static variable, and
68  * thus forbid definition of radix table other than proper domains.
69  */
70 #define USE_RADIX
71 
72 #include <sys/cdefs.h>
73 __KERNEL_RCSID(0, "$NetBSD: ip_encap.c,v 1.18 2004/04/26 01:31:56 matt Exp $");
74 
75 #include "opt_mrouting.h"
76 #include "opt_inet.h"
77 
78 #include <sys/param.h>
79 #include <sys/systm.h>
80 #include <sys/socket.h>
81 #include <sys/sockio.h>
82 #include <sys/mbuf.h>
83 #include <sys/errno.h>
84 #include <sys/protosw.h>
85 #include <sys/queue.h>
86 
87 #include <net/if.h>
88 #include <net/route.h>
89 
90 #include <netinet/in.h>
91 #include <netinet/in_systm.h>
92 #include <netinet/ip.h>
93 #include <netinet/ip_var.h>
94 #include <netinet/ip_encap.h>
95 #ifdef MROUTING
96 #include <netinet/ip_mroute.h>
97 #endif /* MROUTING */
98 
99 #ifdef INET6
100 #include <netinet/ip6.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet6/ip6protosw.h>
103 #include <netinet6/in6_var.h>
104 #include <netinet6/in6_pcb.h>
105 #include <netinet/icmp6.h>
106 #endif
107 
108 #include <machine/stdarg.h>
109 
110 #include <net/net_osdep.h>
111 
112 /* to lookup a pair of address using radix tree */
113 struct sockaddr_pack {
114 	u_int8_t sp_len;
115 	u_int8_t sp_family;	/* not really used */
116 	/* followed by variable-length data */
117 };
118 
119 struct pack4 {
120 	struct sockaddr_pack p;
121 	struct sockaddr_in mine;
122 	struct sockaddr_in yours;
123 };
124 struct pack6 {
125 	struct sockaddr_pack p;
126 	struct sockaddr_in6 mine;
127 	struct sockaddr_in6 yours;
128 };
129 
130 enum direction { INBOUND, OUTBOUND };
131 
132 #ifdef INET
133 static struct encaptab *encap4_lookup __P((struct mbuf *, int, int,
134 	enum direction));
135 #endif
136 #ifdef INET6
137 static struct encaptab *encap6_lookup __P((struct mbuf *, int, int,
138 	enum direction));
139 #endif
140 static int encap_add __P((struct encaptab *));
141 static int encap_remove __P((struct encaptab *));
142 static int encap_afcheck __P((int, const struct sockaddr *, const struct sockaddr *));
143 #ifdef USE_RADIX
144 static struct radix_node_head *encap_rnh __P((int));
145 static int mask_matchlen __P((const struct sockaddr *));
146 #endif
147 #ifndef USE_RADIX
148 static int mask_match __P((const struct encaptab *, const struct sockaddr *,
149 		const struct sockaddr *));
150 #endif
151 static void encap_fillarg __P((struct mbuf *, const struct encaptab *));
152 
153 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
154 
155 #ifdef USE_RADIX
156 extern int max_keylen;	/* radix.c */
157 struct radix_node_head *encap_head[2];	/* 0 for AF_INET, 1 for AF_INET6 */
158 #endif
159 
160 void
161 encap_init()
162 {
163 	static int initialized = 0;
164 
165 	if (initialized)
166 		return;
167 	initialized++;
168 #if 0
169 	/*
170 	 * we cannot use LIST_INIT() here, since drivers may want to call
171 	 * encap_attach(), on driver attach.  encap_init() will be called
172 	 * on AF_INET{,6} initialization, which happens after driver
173 	 * initialization - using LIST_INIT() here can nuke encap_attach()
174 	 * from drivers.
175 	 */
176 	LIST_INIT(&encaptab);
177 #endif
178 
179 #ifdef USE_RADIX
180 	/*
181 	 * initialize radix lookup table.
182 	 * max_keylen initialization should happen before the call to rn_init().
183 	 */
184 	rn_inithead((void *)&encap_head[0], sizeof(struct sockaddr_pack) << 3);
185 	if (sizeof(struct pack4) > max_keylen)
186 		max_keylen = sizeof(struct pack4);
187 #ifdef INET6
188 	rn_inithead((void *)&encap_head[1], sizeof(struct sockaddr_pack) << 3);
189 	if (sizeof(struct pack6) > max_keylen)
190 		max_keylen = sizeof(struct pack6);
191 #endif
192 #endif
193 }
194 
195 #ifdef INET
196 static struct encaptab *
197 encap4_lookup(m, off, proto, dir)
198 	struct mbuf *m;
199 	int off;
200 	int proto;
201 	enum direction dir;
202 {
203 	struct ip *ip;
204 	struct pack4 pack;
205 	struct encaptab *ep, *match;
206 	int prio, matchprio;
207 #ifdef USE_RADIX
208 	struct radix_node_head *rnh = encap_rnh(AF_INET);
209 	struct radix_node *rn;
210 #endif
211 
212 #ifdef DIAGNOSTIC
213 	if (m->m_len < sizeof(*ip))
214 		panic("encap4_lookup");
215 #endif
216 	ip = mtod(m, struct ip *);
217 
218 	bzero(&pack, sizeof(pack));
219 	pack.p.sp_len = sizeof(pack);
220 	pack.mine.sin_family = pack.yours.sin_family = AF_INET;
221 	pack.mine.sin_len = pack.yours.sin_len = sizeof(struct sockaddr_in);
222 	if (dir == INBOUND) {
223 		pack.mine.sin_addr = ip->ip_dst;
224 		pack.yours.sin_addr = ip->ip_src;
225 	} else {
226 		pack.mine.sin_addr = ip->ip_src;
227 		pack.yours.sin_addr = ip->ip_dst;
228 	}
229 
230 	match = NULL;
231 	matchprio = 0;
232 
233 #ifdef USE_RADIX
234 	rn = rnh->rnh_matchaddr((caddr_t)&pack, rnh);
235 	if (rn && (rn->rn_flags & RNF_ROOT) == 0) {
236 		match = (struct encaptab *)rn;
237 		matchprio = mask_matchlen(match->srcmask) +
238 		    mask_matchlen(match->dstmask);
239 	}
240 #endif
241 
242 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
243 		if (ep->af != AF_INET)
244 			continue;
245 		if (ep->proto >= 0 && ep->proto != proto)
246 			continue;
247 		if (ep->func)
248 			prio = (*ep->func)(m, off, proto, ep->arg);
249 		else {
250 #ifdef USE_RADIX
251 			continue;
252 #else
253 			prio = mask_match(ep, (struct sockaddr *)&pack.mine,
254 			    (struct sockaddr *)&pack.yours);
255 #endif
256 		}
257 
258 		/*
259 		 * We prioritize the matches by using bit length of the
260 		 * matches.  mask_match() and user-supplied matching function
261 		 * should return the bit length of the matches (for example,
262 		 * if both src/dst are matched for IPv4, 64 should be returned).
263 		 * 0 or negative return value means "it did not match".
264 		 *
265 		 * The question is, since we have two "mask" portion, we
266 		 * cannot really define total order between entries.
267 		 * For example, which of these should be preferred?
268 		 * mask_match() returns 48 (32 + 16) for both of them.
269 		 *	src=3ffe::/16, dst=3ffe:501::/32
270 		 *	src=3ffe:501::/32, dst=3ffe::/16
271 		 *
272 		 * We need to loop through all the possible candidates
273 		 * to get the best match - the search takes O(n) for
274 		 * n attachments (i.e. interfaces).
275 		 *
276 		 * For radix-based lookup, I guess source takes precedence.
277 		 * See rn_{refines,lexobetter} for the correct answer.
278 		 */
279 		if (prio <= 0)
280 			continue;
281 		if (prio > matchprio) {
282 			matchprio = prio;
283 			match = ep;
284 		}
285 	}
286 
287 	return match;
288 #undef s
289 #undef d
290 }
291 
292 void
293 encap4_input(struct mbuf *m, ...)
294 {
295 	int off, proto;
296 	va_list ap;
297 	const struct protosw *psw;
298 	struct encaptab *match;
299 
300 	va_start(ap, m);
301 	off = va_arg(ap, int);
302 	proto = va_arg(ap, int);
303 	va_end(ap);
304 
305 	match = encap4_lookup(m, off, proto, INBOUND);
306 
307 	if (match) {
308 		/* found a match, "match" has the best one */
309 		psw = match->psw;
310 		if (psw && psw->pr_input) {
311 			encap_fillarg(m, match);
312 			(*psw->pr_input)(m, off, proto);
313 		} else
314 			m_freem(m);
315 		return;
316 	}
317 
318 	/* last resort: inject to raw socket */
319 	rip_input(m, off, proto);
320 }
321 #endif
322 
323 #ifdef INET6
324 static struct encaptab *
325 encap6_lookup(m, off, proto, dir)
326 	struct mbuf *m;
327 	int off;
328 	int proto;
329 	enum direction dir;
330 {
331 	struct ip6_hdr *ip6;
332 	struct pack6 pack;
333 	int prio, matchprio;
334 	struct encaptab *ep, *match;
335 #ifdef USE_RADIX
336 	struct radix_node_head *rnh = encap_rnh(AF_INET6);
337 	struct radix_node *rn;
338 #endif
339 
340 #ifdef DIAGNOSTIC
341 	if (m->m_len < sizeof(*ip6))
342 		panic("encap6_lookup");
343 #endif
344 	ip6 = mtod(m, struct ip6_hdr *);
345 
346 	bzero(&pack, sizeof(pack));
347 	pack.p.sp_len = sizeof(pack);
348 	pack.mine.sin6_family = pack.yours.sin6_family = AF_INET6;
349 	pack.mine.sin6_len = pack.yours.sin6_len = sizeof(struct sockaddr_in6);
350 	if (dir == INBOUND) {
351 		pack.mine.sin6_addr = ip6->ip6_dst;
352 		pack.yours.sin6_addr = ip6->ip6_src;
353 	} else {
354 		pack.mine.sin6_addr = ip6->ip6_src;
355 		pack.yours.sin6_addr = ip6->ip6_dst;
356 	}
357 
358 	match = NULL;
359 	matchprio = 0;
360 
361 #ifdef USE_RADIX
362 	rn = rnh->rnh_matchaddr((caddr_t)&pack, rnh);
363 	if (rn && (rn->rn_flags & RNF_ROOT) == 0) {
364 		match = (struct encaptab *)rn;
365 		matchprio = mask_matchlen(match->srcmask) +
366 		    mask_matchlen(match->dstmask);
367 	}
368 #endif
369 
370 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
371 		if (ep->af != AF_INET6)
372 			continue;
373 		if (ep->proto >= 0 && ep->proto != proto)
374 			continue;
375 		if (ep->func)
376 			prio = (*ep->func)(m, off, proto, ep->arg);
377 		else {
378 #ifdef USE_RADIX
379 			continue;
380 #else
381 			prio = mask_match(ep, (struct sockaddr *)&pack.mine,
382 			    (struct sockaddr *)&pack.yours);
383 #endif
384 		}
385 
386 		/* see encap4_lookup() for issues here */
387 		if (prio <= 0)
388 			continue;
389 		if (prio > matchprio) {
390 			matchprio = prio;
391 			match = ep;
392 		}
393 	}
394 
395 	return match;
396 #undef s
397 #undef d
398 }
399 
400 int
401 encap6_input(mp, offp, proto)
402 	struct mbuf **mp;
403 	int *offp;
404 	int proto;
405 {
406 	struct mbuf *m = *mp;
407 	const struct ip6protosw *psw;
408 	struct encaptab *match;
409 
410 	match = encap6_lookup(m, *offp, proto, INBOUND);
411 
412 	if (match) {
413 		/* found a match */
414 		psw = (const struct ip6protosw *)match->psw;
415 		if (psw && psw->pr_input) {
416 			encap_fillarg(m, match);
417 			return (*psw->pr_input)(mp, offp, proto);
418 		} else {
419 			m_freem(m);
420 			return IPPROTO_DONE;
421 		}
422 	}
423 
424 	/* last resort: inject to raw socket */
425 	return rip6_input(mp, offp, proto);
426 }
427 #endif
428 
429 static int
430 encap_add(ep)
431 	struct encaptab *ep;
432 {
433 #ifdef USE_RADIX
434 	struct radix_node_head *rnh = encap_rnh(ep->af);
435 #endif
436 	int error = 0;
437 
438 	LIST_INSERT_HEAD(&encaptab, ep, chain);
439 #ifdef USE_RADIX
440 	if (!ep->func && rnh) {
441 		if (!rnh->rnh_addaddr((caddr_t)ep->addrpack,
442 		    (caddr_t)ep->maskpack, rnh, ep->nodes)) {
443 			error = EEXIST;
444 			goto fail;
445 		}
446 	}
447 #endif
448 	return error;
449 
450  fail:
451 	LIST_REMOVE(ep, chain);
452 	return error;
453 }
454 
455 static int
456 encap_remove(ep)
457 	struct encaptab *ep;
458 {
459 #ifdef USE_RADIX
460 	struct radix_node_head *rnh = encap_rnh(ep->af);
461 #endif
462 	int error = 0;
463 
464 	LIST_REMOVE(ep, chain);
465 #ifdef USE_RADIX
466 	if (!ep->func && rnh) {
467 		if (!rnh->rnh_deladdr((caddr_t)ep->addrpack,
468 		    (caddr_t)ep->maskpack, rnh))
469 			error = ESRCH;
470 	}
471 #endif
472 	return error;
473 }
474 
475 static int
476 encap_afcheck(af, sp, dp)
477 	int af;
478 	const struct sockaddr *sp;
479 	const struct sockaddr *dp;
480 {
481 	if (sp && dp) {
482 		if (sp->sa_len != dp->sa_len)
483 			return EINVAL;
484 		if (af != sp->sa_family || af != dp->sa_family)
485 			return EINVAL;
486 	} else if (!sp && !dp)
487 		;
488 	else
489 		return EINVAL;
490 
491 	switch (af) {
492 	case AF_INET:
493 		if (sp && sp->sa_len != sizeof(struct sockaddr_in))
494 			return EINVAL;
495 		if (dp && dp->sa_len != sizeof(struct sockaddr_in))
496 			return EINVAL;
497 		break;
498 #ifdef INET6
499 	case AF_INET6:
500 		if (sp && sp->sa_len != sizeof(struct sockaddr_in6))
501 			return EINVAL;
502 		if (dp && dp->sa_len != sizeof(struct sockaddr_in6))
503 			return EINVAL;
504 		break;
505 #endif
506 	default:
507 		return EAFNOSUPPORT;
508 	}
509 
510 	return 0;
511 }
512 
513 /*
514  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
515  * length of mask (sm and dm) is assumed to be same as sp/dp.
516  * Return value will be necessary as input (cookie) for encap_detach().
517  */
518 const struct encaptab *
519 encap_attach(af, proto, sp, sm, dp, dm, psw, arg)
520 	int af;
521 	int proto;
522 	const struct sockaddr *sp, *sm;
523 	const struct sockaddr *dp, *dm;
524 	const struct protosw *psw;
525 	void *arg;
526 {
527 	struct encaptab *ep;
528 	int error;
529 	int s;
530 	size_t l;
531 	struct pack4 *pack4;
532 #ifdef INET6
533 	struct pack6 *pack6;
534 #endif
535 
536 	s = splsoftnet();
537 	/* sanity check on args */
538 	error = encap_afcheck(af, sp, dp);
539 	if (error)
540 		goto fail;
541 
542 	/* check if anyone have already attached with exactly same config */
543 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
544 		if (ep->af != af)
545 			continue;
546 		if (ep->proto != proto)
547 			continue;
548 		if (ep->func)
549 			continue;
550 #ifdef DIAGNOSTIC
551 		if (!ep->src || !ep->dst || !ep->srcmask || !ep->dstmask)
552 			panic("null pointers in encaptab");
553 #endif
554 		if (ep->src->sa_len != sp->sa_len ||
555 		    bcmp(ep->src, sp, sp->sa_len) != 0 ||
556 		    bcmp(ep->srcmask, sm, sp->sa_len) != 0)
557 			continue;
558 		if (ep->dst->sa_len != dp->sa_len ||
559 		    bcmp(ep->dst, dp, dp->sa_len) != 0 ||
560 		    bcmp(ep->dstmask, dm, dp->sa_len) != 0)
561 			continue;
562 
563 		error = EEXIST;
564 		goto fail;
565 	}
566 
567 	switch (af) {
568 	case AF_INET:
569 		l = sizeof(*pack4);
570 		break;
571 #ifdef INET6
572 	case AF_INET6:
573 		l = sizeof(*pack6);
574 		break;
575 #endif
576 	default:
577 		goto fail;
578 	}
579 
580 #ifdef DIAGNOSTIC
581 	/* if l exceeds the value sa_len can possibly express, it's wrong. */
582 	if (l > (1 << (8 * sizeof(ep->addrpack->sa_len)))) {
583 		error = EINVAL;
584 		goto fail;
585 	}
586 #endif
587 
588 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/* M_NETADDR ok? */
589 	if (ep == NULL) {
590 		error = ENOBUFS;
591 		goto fail;
592 	}
593 	bzero(ep, sizeof(*ep));
594 	ep->addrpack = malloc(l, M_NETADDR, M_NOWAIT);
595 	if (ep->addrpack == NULL) {
596 		error = ENOBUFS;
597 		goto gc;
598 	}
599 	ep->maskpack = malloc(l, M_NETADDR, M_NOWAIT);
600 	if (ep->maskpack == NULL) {
601 		error = ENOBUFS;
602 		goto gc;
603 	}
604 
605 	ep->af = af;
606 	ep->proto = proto;
607 	ep->addrpack->sa_len = l & 0xff;
608 	ep->maskpack->sa_len = l & 0xff;
609 	switch (af) {
610 	case AF_INET:
611 		pack4 = (struct pack4 *)ep->addrpack;
612 		ep->src = (struct sockaddr *)&pack4->mine;
613 		ep->dst = (struct sockaddr *)&pack4->yours;
614 		pack4 = (struct pack4 *)ep->maskpack;
615 		ep->srcmask = (struct sockaddr *)&pack4->mine;
616 		ep->dstmask = (struct sockaddr *)&pack4->yours;
617 		break;
618 #ifdef INET6
619 	case AF_INET6:
620 		pack6 = (struct pack6 *)ep->addrpack;
621 		ep->src = (struct sockaddr *)&pack6->mine;
622 		ep->dst = (struct sockaddr *)&pack6->yours;
623 		pack6 = (struct pack6 *)ep->maskpack;
624 		ep->srcmask = (struct sockaddr *)&pack6->mine;
625 		ep->dstmask = (struct sockaddr *)&pack6->yours;
626 		break;
627 #endif
628 	}
629 
630 	bcopy(sp, ep->src, sp->sa_len);
631 	bcopy(sm, ep->srcmask, sp->sa_len);
632 	bcopy(dp, ep->dst, dp->sa_len);
633 	bcopy(dm, ep->dstmask, dp->sa_len);
634 	ep->psw = psw;
635 	ep->arg = arg;
636 
637 	error = encap_add(ep);
638 	if (error)
639 		goto gc;
640 
641 	error = 0;
642 	splx(s);
643 	return ep;
644 
645 gc:
646 	if (ep->addrpack)
647 		free(ep->addrpack, M_NETADDR);
648 	if (ep->maskpack)
649 		free(ep->maskpack, M_NETADDR);
650 	if (ep)
651 		free(ep, M_NETADDR);
652 fail:
653 	splx(s);
654 	return NULL;
655 }
656 
657 const struct encaptab *
658 encap_attach_func(af, proto, func, psw, arg)
659 	int af;
660 	int proto;
661 	int (*func) __P((const struct mbuf *, int, int, void *));
662 	const struct protosw *psw;
663 	void *arg;
664 {
665 	struct encaptab *ep;
666 	int error;
667 	int s;
668 
669 	s = splsoftnet();
670 	/* sanity check on args */
671 	if (!func) {
672 		error = EINVAL;
673 		goto fail;
674 	}
675 
676 	error = encap_afcheck(af, NULL, NULL);
677 	if (error)
678 		goto fail;
679 
680 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
681 	if (ep == NULL) {
682 		error = ENOBUFS;
683 		goto fail;
684 	}
685 	bzero(ep, sizeof(*ep));
686 
687 	ep->af = af;
688 	ep->proto = proto;
689 	ep->func = func;
690 	ep->psw = psw;
691 	ep->arg = arg;
692 
693 	error = encap_add(ep);
694 	if (error)
695 		goto fail;
696 
697 	error = 0;
698 	splx(s);
699 	return ep;
700 
701 fail:
702 	splx(s);
703 	return NULL;
704 }
705 
706 /* XXX encap4_ctlinput() is necessary if we set DF=1 on outer IPv4 header */
707 
708 #ifdef INET6
709 void
710 encap6_ctlinput(cmd, sa, d0)
711 	int cmd;
712 	struct sockaddr *sa;
713 	void *d0;
714 {
715 	void *d = d0;
716 	struct ip6_hdr *ip6;
717 	struct mbuf *m;
718 	int off;
719 	struct ip6ctlparam *ip6cp = NULL;
720 	int nxt;
721 	struct encaptab *ep;
722 	const struct ip6protosw *psw;
723 
724 	if (sa->sa_family != AF_INET6 ||
725 	    sa->sa_len != sizeof(struct sockaddr_in6))
726 		return;
727 
728 	if ((unsigned)cmd >= PRC_NCMDS)
729 		return;
730 	if (cmd == PRC_HOSTDEAD)
731 		d = NULL;
732 	else if (cmd == PRC_MSGSIZE)
733 		; /* special code is present, see below */
734 	else if (inet6ctlerrmap[cmd] == 0)
735 		return;
736 
737 	/* if the parameter is from icmp6, decode it. */
738 	if (d != NULL) {
739 		ip6cp = (struct ip6ctlparam *)d;
740 		m = ip6cp->ip6c_m;
741 		ip6 = ip6cp->ip6c_ip6;
742 		off = ip6cp->ip6c_off;
743 		nxt = ip6cp->ip6c_nxt;
744 
745 		if (ip6 && cmd == PRC_MSGSIZE) {
746 			int valid = 0;
747 			struct encaptab *match;
748 
749 			/*
750 		 	* Check to see if we have a valid encap configuration.
751 		 	*/
752 			match = encap6_lookup(m, off, nxt, OUTBOUND);
753 			if (match)
754 				valid++;
755 
756 			/*
757 		 	* Depending on the value of "valid" and routing table
758 		 	* size (mtudisc_{hi,lo}wat), we will:
759 		 	* - recalcurate the new MTU and create the
760 		 	*   corresponding routing entry, or
761 		 	* - ignore the MTU change notification.
762 		 	*/
763 			icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
764 		}
765 	} else {
766 		m = NULL;
767 		ip6 = NULL;
768 		nxt = -1;
769 	}
770 
771 	/* inform all listeners */
772 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
773 		if (ep->af != AF_INET6)
774 			continue;
775 		if (ep->proto >= 0 && ep->proto != nxt)
776 			continue;
777 
778 		/* should optimize by looking at address pairs */
779 
780 		/* XXX need to pass ep->arg or ep itself to listeners */
781 		psw = (const struct ip6protosw *)ep->psw;
782 		if (psw && psw->pr_ctlinput)
783 			(*psw->pr_ctlinput)(cmd, sa, d);
784 	}
785 
786 	rip6_ctlinput(cmd, sa, d0);
787 }
788 #endif
789 
790 int
791 encap_detach(cookie)
792 	const struct encaptab *cookie;
793 {
794 	const struct encaptab *ep = cookie;
795 	struct encaptab *p;
796 	int error;
797 
798 	for (p = LIST_FIRST(&encaptab); p; p = LIST_NEXT(p, chain)) {
799 		if (p == ep) {
800 			error = encap_remove(p);
801 			if (error)
802 				return error;
803 			if (!ep->func) {
804 				free(p->addrpack, M_NETADDR);
805 				free(p->maskpack, M_NETADDR);
806 			}
807 			free(p, M_NETADDR);	/*XXX*/
808 			return 0;
809 		}
810 	}
811 
812 	return ENOENT;
813 }
814 
815 #ifdef USE_RADIX
816 static struct radix_node_head *
817 encap_rnh(af)
818 	int af;
819 {
820 
821 	switch (af) {
822 	case AF_INET:
823 		return encap_head[0];
824 #ifdef INET6
825 	case AF_INET6:
826 		return encap_head[1];
827 #endif
828 	default:
829 		return NULL;
830 	}
831 }
832 
833 static int
834 mask_matchlen(sa)
835 	const struct sockaddr *sa;
836 {
837 	const char *p, *ep;
838 	int l;
839 
840 	p = (const char *)sa;
841 	ep = p + sa->sa_len;
842 	p += 2;	/* sa_len + sa_family */
843 
844 	l = 0;
845 	while (p < ep) {
846 		l += (*p ? 8 : 0);	/* estimate */
847 		p++;
848 	}
849 	return l;
850 }
851 #endif
852 
853 #ifndef USE_RADIX
854 static int
855 mask_match(ep, sp, dp)
856 	const struct encaptab *ep;
857 	const struct sockaddr *sp;
858 	const struct sockaddr *dp;
859 {
860 	struct sockaddr_storage s;
861 	struct sockaddr_storage d;
862 	int i;
863 	const u_int8_t *p, *q;
864 	u_int8_t *r;
865 	int matchlen;
866 
867 #ifdef DIAGNOSTIC
868 	if (ep->func)
869 		panic("wrong encaptab passed to mask_match");
870 #endif
871 	if (sp->sa_len > sizeof(s) || dp->sa_len > sizeof(d))
872 		return 0;
873 	if (sp->sa_family != ep->af || dp->sa_family != ep->af)
874 		return 0;
875 	if (sp->sa_len != ep->src->sa_len || dp->sa_len != ep->dst->sa_len)
876 		return 0;
877 
878 	matchlen = 0;
879 
880 	p = (const u_int8_t *)sp;
881 	q = (const u_int8_t *)ep->srcmask;
882 	r = (u_int8_t *)&s;
883 	for (i = 0 ; i < sp->sa_len; i++) {
884 		r[i] = p[i] & q[i];
885 		/* XXX estimate */
886 		matchlen += (q[i] ? 8 : 0);
887 	}
888 
889 	p = (const u_int8_t *)dp;
890 	q = (const u_int8_t *)ep->dstmask;
891 	r = (u_int8_t *)&d;
892 	for (i = 0 ; i < dp->sa_len; i++) {
893 		r[i] = p[i] & q[i];
894 		/* XXX rough estimate */
895 		matchlen += (q[i] ? 8 : 0);
896 	}
897 
898 	/* need to overwrite len/family portion as we don't compare them */
899 	s.ss_len = sp->sa_len;
900 	s.ss_family = sp->sa_family;
901 	d.ss_len = dp->sa_len;
902 	d.ss_family = dp->sa_family;
903 
904 	if (bcmp(&s, ep->src, ep->src->sa_len) == 0 &&
905 	    bcmp(&d, ep->dst, ep->dst->sa_len) == 0) {
906 		return matchlen;
907 	} else
908 		return 0;
909 }
910 #endif
911 
912 static void
913 encap_fillarg(m, ep)
914 	struct mbuf *m;
915 	const struct encaptab *ep;
916 {
917 	struct m_tag *mtag;
918 
919 	mtag = m_tag_get(PACKET_TAG_ENCAP, sizeof(void *), M_NOWAIT);
920 	if (mtag) {
921 		*(void **)(mtag + 1) = ep->arg;
922 		m_tag_prepend(m, mtag);
923 	}
924 }
925 
926 void *
927 encap_getarg(m)
928 	struct mbuf *m;
929 {
930 	void *p;
931 	struct m_tag *mtag;
932 
933 	p = NULL;
934 	mtag = m_tag_find(m, PACKET_TAG_ENCAP, NULL);
935 	if (mtag != NULL) {
936 		p = *(void **)(mtag + 1);
937 		m_tag_delete(m, mtag);
938 	}
939 	return p;
940 }
941