xref: /netbsd/sys/netinet/ip_encap.c (revision bf9ec67e)
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 1.8 2002/03/04 13:24:12 sommerfeld 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 #ifdef MROUTING
111 #include <netinet/ip_mroute.h>
112 #endif
113 
114 #include <net/net_osdep.h>
115 
116 /* to lookup a pair of address using radix tree */
117 struct sockaddr_pack {
118 	u_int8_t sp_len;
119 	u_int8_t sp_family;	/* not really used */
120 	/* followed by variable-length data */
121 } __attribute__((__packed__));
122 
123 struct pack4 {
124 	struct sockaddr_pack p;
125 	struct sockaddr_in mine;
126 	struct sockaddr_in yours;
127 } __attribute__((__packed__));
128 struct pack6 {
129 	struct sockaddr_pack p;
130 	struct sockaddr_in6 mine;
131 	struct sockaddr_in6 yours;
132 } __attribute__((__packed__));
133 
134 enum direction { INBOUND, OUTBOUND };
135 
136 #ifdef INET
137 static struct encaptab *encap4_lookup __P((struct mbuf *, int, int,
138 	enum direction));
139 #endif
140 #ifdef INET6
141 static struct encaptab *encap6_lookup __P((struct mbuf *, int, int,
142 	enum direction));
143 #endif
144 static int encap_add __P((struct encaptab *));
145 static int encap_remove __P((struct encaptab *));
146 static int encap_afcheck __P((int, const struct sockaddr *, const struct sockaddr *));
147 #ifdef USE_RADIX
148 static struct radix_node_head *encap_rnh __P((int));
149 static int mask_matchlen __P((const struct sockaddr *));
150 #endif
151 #ifndef USE_RADIX
152 static int mask_match __P((const struct encaptab *, const struct sockaddr *,
153 		const struct sockaddr *));
154 #endif
155 static void encap_fillarg __P((struct mbuf *, const struct encaptab *));
156 
157 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
158 
159 #ifdef USE_RADIX
160 extern int max_keylen;	/* radix.c */
161 struct radix_node_head *encap_head[2];	/* 0 for AF_INET, 1 for AF_INET6 */
162 #endif
163 
164 void
165 encap_init()
166 {
167 	static int initialized = 0;
168 
169 	if (initialized)
170 		return;
171 	initialized++;
172 #if 0
173 	/*
174 	 * we cannot use LIST_INIT() here, since drivers may want to call
175 	 * encap_attach(), on driver attach.  encap_init() will be called
176 	 * on AF_INET{,6} initialization, which happens after driver
177 	 * initialization - using LIST_INIT() here can nuke encap_attach()
178 	 * from drivers.
179 	 */
180 	LIST_INIT(&encaptab);
181 #endif
182 
183 #ifdef USE_RADIX
184 	/*
185 	 * initialize radix lookup table.
186 	 * max_keylen initialization should happen before the call to rn_init().
187 	 */
188 	rn_inithead((void **)&encap_head[0], sizeof(struct sockaddr_pack) << 3);
189 	if (sizeof(struct pack4) > max_keylen)
190 		max_keylen = sizeof(struct pack4);
191 #ifdef INET6
192 	rn_inithead((void **)&encap_head[1], sizeof(struct sockaddr_pack) << 3);
193 	if (sizeof(struct pack6) > max_keylen)
194 		max_keylen = sizeof(struct pack6);
195 #endif
196 #endif
197 }
198 
199 #ifdef INET
200 static struct encaptab *
201 encap4_lookup(m, off, proto, dir)
202 	struct mbuf *m;
203 	int off;
204 	int proto;
205 	enum direction dir;
206 {
207 	struct ip *ip;
208 	struct pack4 pack;
209 	struct encaptab *ep, *match;
210 	int prio, matchprio;
211 #ifdef USE_RADIX
212 	struct radix_node_head *rnh = encap_rnh(AF_INET);
213 	struct radix_node *rn;
214 #endif
215 
216 #ifdef DIAGNOSTIC
217 	if (m->m_len < sizeof(*ip))
218 		panic("encap4_lookup");
219 #endif
220 	ip = mtod(m, struct ip *);
221 
222 	bzero(&pack, sizeof(pack));
223 	pack.p.sp_len = sizeof(pack);
224 	pack.mine.sin_family = pack.yours.sin_family = AF_INET;
225 	pack.mine.sin_len = pack.yours.sin_len = sizeof(struct sockaddr_in);
226 	if (dir == INBOUND) {
227 		pack.mine.sin_addr = ip->ip_dst;
228 		pack.yours.sin_addr = ip->ip_src;
229 	} else {
230 		pack.mine.sin_addr = ip->ip_src;
231 		pack.yours.sin_addr = ip->ip_dst;
232 	}
233 
234 	match = NULL;
235 	matchprio = 0;
236 
237 #ifdef USE_RADIX
238 	rn = rnh->rnh_matchaddr((caddr_t)&pack, rnh);
239 	if (rn && (rn->rn_flags & RNF_ROOT) == 0) {
240 		match = (struct encaptab *)rn;
241 		matchprio = mask_matchlen(match->srcmask) +
242 		    mask_matchlen(match->dstmask);
243 	}
244 #endif
245 
246 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
247 		if (ep->af != AF_INET)
248 			continue;
249 		if (ep->proto >= 0 && ep->proto != proto)
250 			continue;
251 		if (ep->func)
252 			prio = (*ep->func)(m, off, proto, ep->arg);
253 		else {
254 #ifdef USE_RADIX
255 			continue;
256 #else
257 			prio = mask_match(ep, (struct sockaddr *)&pack.mine,
258 			    (struct sockaddr *)&pack.yours);
259 #endif
260 		}
261 
262 		/*
263 		 * We prioritize the matches by using bit length of the
264 		 * matches.  mask_match() and user-supplied matching function
265 		 * should return the bit length of the matches (for example,
266 		 * if both src/dst are matched for IPv4, 64 should be returned).
267 		 * 0 or negative return value means "it did not match".
268 		 *
269 		 * The question is, since we have two "mask" portion, we
270 		 * cannot really define total order between entries.
271 		 * For example, which of these should be preferred?
272 		 * mask_match() returns 48 (32 + 16) for both of them.
273 		 *	src=3ffe::/16, dst=3ffe:501::/32
274 		 *	src=3ffe:501::/32, dst=3ffe::/16
275 		 *
276 		 * We need to loop through all the possible candidates
277 		 * to get the best match - the search takes O(n) for
278 		 * n attachments (i.e. interfaces).
279 		 *
280 		 * For radix-based lookup, I guess source takes precedence.
281 		 * See rn_{refines,lexobetter} for the correct answer.
282 		 */
283 		if (prio <= 0)
284 			continue;
285 		if (prio > matchprio) {
286 			matchprio = prio;
287 			match = ep;
288 		}
289 	}
290 
291 	return match;
292 #undef s
293 #undef d
294 }
295 
296 void
297 #if __STDC__
298 encap4_input(struct mbuf *m, ...)
299 #else
300 encap4_input(m, va_alist)
301 	struct mbuf *m;
302 	va_dcl
303 #endif
304 {
305 	int off, proto;
306 	va_list ap;
307 	const struct protosw *psw;
308 	struct encaptab *match;
309 
310 	va_start(ap, m);
311 	off = va_arg(ap, int);
312 	proto = va_arg(ap, int);
313 	va_end(ap);
314 
315 	match = encap4_lookup(m, off, proto, INBOUND);
316 
317 	if (match) {
318 		/* found a match, "match" has the best one */
319 		psw = match->psw;
320 		if (psw && psw->pr_input) {
321 			encap_fillarg(m, match);
322 			(*psw->pr_input)(m, off, proto);
323 		} else
324 			m_freem(m);
325 		return;
326 	}
327 
328 	/* last resort: inject to raw socket */
329 	rip_input(m, off, proto);
330 }
331 #endif
332 
333 #ifdef INET6
334 static struct encaptab *
335 encap6_lookup(m, off, proto, dir)
336 	struct mbuf *m;
337 	int off;
338 	int proto;
339 	enum direction dir;
340 {
341 	struct ip6_hdr *ip6;
342 	struct pack6 pack;
343 	int prio, matchprio;
344 	struct encaptab *ep, *match;
345 #ifdef USE_RADIX
346 	struct radix_node_head *rnh = encap_rnh(AF_INET6);
347 	struct radix_node *rn;
348 #endif
349 
350 #ifdef DIAGNOSTIC
351 	if (m->m_len < sizeof(*ip6))
352 		panic("encap6_lookup");
353 #endif
354 	ip6 = mtod(m, struct ip6_hdr *);
355 
356 	bzero(&pack, sizeof(pack));
357 	pack.p.sp_len = sizeof(pack);
358 	pack.mine.sin6_family = pack.yours.sin6_family = AF_INET6;
359 	pack.mine.sin6_len = pack.yours.sin6_len = sizeof(struct sockaddr_in6);
360 	if (dir == INBOUND) {
361 		pack.mine.sin6_addr = ip6->ip6_dst;
362 		pack.yours.sin6_addr = ip6->ip6_src;
363 	} else {
364 		pack.mine.sin6_addr = ip6->ip6_src;
365 		pack.yours.sin6_addr = ip6->ip6_dst;
366 	}
367 
368 	match = NULL;
369 	matchprio = 0;
370 
371 #ifdef USE_RADIX
372 	rn = rnh->rnh_matchaddr((caddr_t)&pack, rnh);
373 	if (rn && (rn->rn_flags & RNF_ROOT) == 0) {
374 		match = (struct encaptab *)rn;
375 		matchprio = mask_matchlen(match->srcmask) +
376 		    mask_matchlen(match->dstmask);
377 	}
378 #endif
379 
380 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
381 		if (ep->af != AF_INET6)
382 			continue;
383 		if (ep->proto >= 0 && ep->proto != proto)
384 			continue;
385 		if (ep->func)
386 			prio = (*ep->func)(m, off, proto, ep->arg);
387 		else {
388 #ifdef USE_RADIX
389 			continue;
390 #else
391 			prio = mask_match(ep, (struct sockaddr *)&pack.mine,
392 			    (struct sockaddr *)&pack.yours);
393 #endif
394 		}
395 
396 		/* see encap4_lookup() for issues here */
397 		if (prio <= 0)
398 			continue;
399 		if (prio > matchprio) {
400 			matchprio = prio;
401 			match = ep;
402 		}
403 	}
404 
405 	return match;
406 #undef s
407 #undef d
408 }
409 
410 int
411 encap6_input(mp, offp, proto)
412 	struct mbuf **mp;
413 	int *offp;
414 	int proto;
415 {
416 	struct mbuf *m = *mp;
417 	const struct ip6protosw *psw;
418 	struct encaptab *match;
419 
420 	match = encap6_lookup(m, *offp, proto, INBOUND);
421 
422 	if (match) {
423 		/* found a match */
424 		psw = (const struct ip6protosw *)match->psw;
425 		if (psw && psw->pr_input) {
426 			encap_fillarg(m, match);
427 			return (*psw->pr_input)(mp, offp, proto);
428 		} else {
429 			m_freem(m);
430 			return IPPROTO_DONE;
431 		}
432 	}
433 
434 	/* last resort: inject to raw socket */
435 	return rip6_input(mp, offp, proto);
436 }
437 #endif
438 
439 static int
440 encap_add(ep)
441 	struct encaptab *ep;
442 {
443 #ifdef USE_RADIX
444 	struct radix_node_head *rnh = encap_rnh(ep->af);
445 #endif
446 	int error = 0;
447 
448 	LIST_INSERT_HEAD(&encaptab, ep, chain);
449 #ifdef USE_RADIX
450 	if (!ep->func && rnh) {
451 		if (!rnh->rnh_addaddr((caddr_t)ep->addrpack,
452 		    (caddr_t)ep->maskpack, rnh, ep->nodes)) {
453 			error = EEXIST;
454 			goto fail;
455 		}
456 	}
457 #endif
458 	return error;
459 
460  fail:
461 	LIST_REMOVE(ep, chain);
462 	return error;
463 }
464 
465 static int
466 encap_remove(ep)
467 	struct encaptab *ep;
468 {
469 #ifdef USE_RADIX
470 	struct radix_node_head *rnh = encap_rnh(ep->af);
471 #endif
472 	int error = 0;
473 
474 	LIST_REMOVE(ep, chain);
475 #ifdef USE_RADIX
476 	if (!ep->func && rnh) {
477 		if (!rnh->rnh_deladdr((caddr_t)ep->addrpack,
478 		    (caddr_t)ep->maskpack, rnh))
479 			error = ESRCH;
480 	}
481 #endif
482 	return error;
483 }
484 
485 static int
486 encap_afcheck(af, sp, dp)
487 	int af;
488 	const struct sockaddr *sp;
489 	const struct sockaddr *dp;
490 {
491 	if (sp && dp) {
492 		if (sp->sa_len != dp->sa_len)
493 			return EINVAL;
494 		if (af != sp->sa_family || af != dp->sa_family)
495 			return EINVAL;
496 	} else if (!sp && !dp)
497 		;
498 	else
499 		return EINVAL;
500 
501 	switch (af) {
502 	case AF_INET:
503 		if (sp && sp->sa_len != sizeof(struct sockaddr_in))
504 			return EINVAL;
505 		if (dp && dp->sa_len != sizeof(struct sockaddr_in))
506 			return EINVAL;
507 		break;
508 #ifdef INET6
509 	case AF_INET6:
510 		if (sp && sp->sa_len != sizeof(struct sockaddr_in6))
511 			return EINVAL;
512 		if (dp && dp->sa_len != sizeof(struct sockaddr_in6))
513 			return EINVAL;
514 		break;
515 #endif
516 	default:
517 		return EAFNOSUPPORT;
518 	}
519 
520 	return 0;
521 }
522 
523 /*
524  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
525  * length of mask (sm and dm) is assumed to be same as sp/dp.
526  * Return value will be necessary as input (cookie) for encap_detach().
527  */
528 const struct encaptab *
529 encap_attach(af, proto, sp, sm, dp, dm, psw, arg)
530 	int af;
531 	int proto;
532 	const struct sockaddr *sp, *sm;
533 	const struct sockaddr *dp, *dm;
534 	const struct protosw *psw;
535 	void *arg;
536 {
537 	struct encaptab *ep;
538 	int error;
539 	int s;
540 	size_t l;
541 	struct pack4 *pack4;
542 #ifdef INET6
543 	struct pack6 *pack6;
544 #endif
545 
546 	s = splsoftnet();
547 	/* sanity check on args */
548 	error = encap_afcheck(af, sp, dp);
549 	if (error)
550 		goto fail;
551 
552 	/* check if anyone have already attached with exactly same config */
553 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
554 		if (ep->af != af)
555 			continue;
556 		if (ep->proto != proto)
557 			continue;
558 		if (ep->func)
559 			continue;
560 #ifdef DIAGNOSTIC
561 		if (!ep->src || !ep->dst || !ep->srcmask || !ep->dstmask)
562 			panic("null pointers in encaptab");
563 #endif
564 		if (ep->src->sa_len != sp->sa_len ||
565 		    bcmp(ep->src, sp, sp->sa_len) != 0 ||
566 		    bcmp(ep->srcmask, sm, sp->sa_len) != 0)
567 			continue;
568 		if (ep->dst->sa_len != dp->sa_len ||
569 		    bcmp(ep->dst, dp, dp->sa_len) != 0 ||
570 		    bcmp(ep->dstmask, dm, dp->sa_len) != 0)
571 			continue;
572 
573 		error = EEXIST;
574 		goto fail;
575 	}
576 
577 	switch (af) {
578 	case AF_INET:
579 		l = sizeof(*pack4);
580 		break;
581 #ifdef INET6
582 	case AF_INET6:
583 		l = sizeof(*pack6);
584 		break;
585 #endif
586 	default:
587 		goto fail;
588 	}
589 
590 #ifdef DIAGNOSTIC
591 	/* if l exceeds the value sa_len can possibly express, it's wrong. */
592 	if (l > (1 << (8 * sizeof(ep->addrpack->sa_len)))) {
593 		error = EINVAL;
594 		goto fail;
595 	}
596 #endif
597 
598 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/* M_NETADDR ok? */
599 	if (ep == NULL) {
600 		error = ENOBUFS;
601 		goto fail;
602 	}
603 	bzero(ep, sizeof(*ep));
604 	ep->addrpack = malloc(l, M_NETADDR, M_NOWAIT);
605 	if (ep->addrpack == NULL) {
606 		error = ENOBUFS;
607 		goto gc;
608 	}
609 	ep->maskpack = malloc(l, M_NETADDR, M_NOWAIT);
610 	if (ep->maskpack == NULL) {
611 		error = ENOBUFS;
612 		goto gc;
613 	}
614 
615 	ep->af = af;
616 	ep->proto = proto;
617 	ep->addrpack->sa_len = l & 0xff;
618 	ep->maskpack->sa_len = l & 0xff;
619 	switch (af) {
620 	case AF_INET:
621 		pack4 = (struct pack4 *)ep->addrpack;
622 		ep->src = (struct sockaddr *)&pack4->mine;
623 		ep->dst = (struct sockaddr *)&pack4->yours;
624 		pack4 = (struct pack4 *)ep->maskpack;
625 		ep->srcmask = (struct sockaddr *)&pack4->mine;
626 		ep->dstmask = (struct sockaddr *)&pack4->yours;
627 		break;
628 #ifdef INET6
629 	case AF_INET6:
630 		pack6 = (struct pack6 *)ep->addrpack;
631 		ep->src = (struct sockaddr *)&pack6->mine;
632 		ep->dst = (struct sockaddr *)&pack6->yours;
633 		pack6 = (struct pack6 *)ep->maskpack;
634 		ep->srcmask = (struct sockaddr *)&pack6->mine;
635 		ep->dstmask = (struct sockaddr *)&pack6->yours;
636 		break;
637 #endif
638 	}
639 
640 	bcopy(sp, ep->src, sp->sa_len);
641 	bcopy(sm, ep->srcmask, sp->sa_len);
642 	bcopy(dp, ep->dst, dp->sa_len);
643 	bcopy(dm, ep->dstmask, dp->sa_len);
644 	ep->psw = psw;
645 	ep->arg = arg;
646 
647 	error = encap_add(ep);
648 	if (error)
649 		goto gc;
650 
651 	error = 0;
652 	splx(s);
653 	return ep;
654 
655 gc:
656 	if (ep->addrpack)
657 		free(ep->addrpack, M_NETADDR);
658 	if (ep->maskpack)
659 		free(ep->maskpack, M_NETADDR);
660 	if (ep)
661 		free(ep, M_NETADDR);
662 fail:
663 	splx(s);
664 	return NULL;
665 }
666 
667 const struct encaptab *
668 encap_attach_func(af, proto, func, psw, arg)
669 	int af;
670 	int proto;
671 	int (*func) __P((const struct mbuf *, int, int, void *));
672 	const struct protosw *psw;
673 	void *arg;
674 {
675 	struct encaptab *ep;
676 	int error;
677 	int s;
678 
679 	s = splsoftnet();
680 	/* sanity check on args */
681 	if (!func) {
682 		error = EINVAL;
683 		goto fail;
684 	}
685 
686 	error = encap_afcheck(af, NULL, NULL);
687 	if (error)
688 		goto fail;
689 
690 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
691 	if (ep == NULL) {
692 		error = ENOBUFS;
693 		goto fail;
694 	}
695 	bzero(ep, sizeof(*ep));
696 
697 	ep->af = af;
698 	ep->proto = proto;
699 	ep->func = func;
700 	ep->psw = psw;
701 	ep->arg = arg;
702 
703 	error = encap_add(ep);
704 	if (error)
705 		goto fail;
706 
707 	error = 0;
708 	splx(s);
709 	return ep;
710 
711 fail:
712 	splx(s);
713 	return NULL;
714 }
715 
716 /* XXX encap4_ctlinput() is necessary if we set DF=1 on outer IPv4 header */
717 
718 #ifdef INET6
719 void
720 encap6_ctlinput(cmd, sa, d0)
721 	int cmd;
722 	struct sockaddr *sa;
723 	void *d0;
724 {
725 	void *d = d0;
726 	struct ip6_hdr *ip6;
727 	struct mbuf *m;
728 	int off;
729 	struct ip6ctlparam *ip6cp = NULL;
730 	const struct sockaddr_in6 *sa6_src = NULL;
731 	void *cmdarg;
732 	int nxt;
733 	struct encaptab *ep;
734 	const struct ip6protosw *psw;
735 
736 	if (sa->sa_family != AF_INET6 ||
737 	    sa->sa_len != sizeof(struct sockaddr_in6))
738 		return;
739 
740 	if ((unsigned)cmd >= PRC_NCMDS)
741 		return;
742 	if (cmd == PRC_HOSTDEAD)
743 		d = NULL;
744 	else if (cmd == PRC_MSGSIZE)
745 		; /* special code is present, see below */
746 	else if (inet6ctlerrmap[cmd] == 0)
747 		return;
748 
749 	/* if the parameter is from icmp6, decode it. */
750 	if (d != NULL) {
751 		ip6cp = (struct ip6ctlparam *)d;
752 		m = ip6cp->ip6c_m;
753 		ip6 = ip6cp->ip6c_ip6;
754 		off = ip6cp->ip6c_off;
755 		cmdarg = ip6cp->ip6c_cmdarg;
756 		sa6_src = ip6cp->ip6c_src;
757 		nxt = ip6cp->ip6c_nxt;
758 	} else {
759 		m = NULL;
760 		ip6 = NULL;
761 		cmdarg = NULL;
762 		sa6_src = &sa6_any;
763 		nxt = -1;
764 	}
765 
766 	if (ip6 && cmd == PRC_MSGSIZE) {
767 		int valid = 0;
768 		struct encaptab *match;
769 
770 		/*
771 		 * Check to see if we have a valid encap configuration.
772 		 */
773 		match = encap6_lookup(m, off, nxt, OUTBOUND);
774 
775 		if (match)
776 			valid++;
777 
778 		/*
779 		 * Depending on the value of "valid" and routing table
780 		 * size (mtudisc_{hi,lo}wat), we will:
781 		 * - recalcurate the new MTU and create the
782 		 *   corresponding routing entry, or
783 		 * - ignore the MTU change notification.
784 		 */
785 		icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
786 	}
787 
788 	/* inform all listeners */
789 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
790 		if (ep->af != AF_INET6)
791 			continue;
792 		if (ep->proto >= 0 && ep->proto != nxt)
793 			continue;
794 
795 		/* should optimize by looking at address pairs */
796 
797 		/* XXX need to pass ep->arg or ep itself to listeners */
798 		psw = (const struct ip6protosw *)ep->psw;
799 		if (psw && psw->pr_ctlinput)
800 			(*psw->pr_ctlinput)(cmd, sa, d);
801 	}
802 
803 	rip6_ctlinput(cmd, sa, d0);
804 }
805 #endif
806 
807 int
808 encap_detach(cookie)
809 	const struct encaptab *cookie;
810 {
811 	const struct encaptab *ep = cookie;
812 	struct encaptab *p;
813 	int error;
814 
815 	for (p = LIST_FIRST(&encaptab); p; p = LIST_NEXT(p, chain)) {
816 		if (p == ep) {
817 			error = encap_remove(p);
818 			if (error)
819 				return error;
820 			if (!ep->func) {
821 				free(p->addrpack, M_NETADDR);
822 				free(p->maskpack, M_NETADDR);
823 			}
824 			free(p, M_NETADDR);	/*XXX*/
825 			return 0;
826 		}
827 	}
828 
829 	return ENOENT;
830 }
831 
832 #ifdef USE_RADIX
833 static struct radix_node_head *
834 encap_rnh(af)
835 	int af;
836 {
837 
838 	switch (af) {
839 	case AF_INET:
840 		return encap_head[0];
841 #ifdef INET6
842 	case AF_INET6:
843 		return encap_head[1];
844 #endif
845 	default:
846 		return NULL;
847 	}
848 }
849 
850 static int
851 mask_matchlen(sa)
852 	const struct sockaddr *sa;
853 {
854 	const char *p, *ep;
855 	int l;
856 
857 	p = (const char *)sa;
858 	ep = p + sa->sa_len;
859 	p += 2;	/* sa_len + sa_family */
860 
861 	l = 0;
862 	while (p < ep) {
863 		l += (*p ? 8 : 0);	/* estimate */
864 		p++;
865 	}
866 	return l;
867 }
868 #endif
869 
870 #ifndef USE_RADIX
871 static int
872 mask_match(ep, sp, dp)
873 	const struct encaptab *ep;
874 	const struct sockaddr *sp;
875 	const struct sockaddr *dp;
876 {
877 	struct sockaddr_storage s;
878 	struct sockaddr_storage d;
879 	int i;
880 	const u_int8_t *p, *q;
881 	u_int8_t *r;
882 	int matchlen;
883 
884 #ifdef DIAGNOSTIC
885 	if (ep->func)
886 		panic("wrong encaptab passed to mask_match");
887 #endif
888 	if (sp->sa_len > sizeof(s) || dp->sa_len > sizeof(d))
889 		return 0;
890 	if (sp->sa_family != ep->af || dp->sa_family != ep->af)
891 		return 0;
892 	if (sp->sa_len != ep->src->sa_len || dp->sa_len != ep->dst->sa_len)
893 		return 0;
894 
895 	matchlen = 0;
896 
897 	p = (const u_int8_t *)sp;
898 	q = (const u_int8_t *)ep->srcmask;
899 	r = (u_int8_t *)&s;
900 	for (i = 0 ; i < sp->sa_len; i++) {
901 		r[i] = p[i] & q[i];
902 		/* XXX estimate */
903 		matchlen += (q[i] ? 8 : 0);
904 	}
905 
906 	p = (const u_int8_t *)dp;
907 	q = (const u_int8_t *)ep->dstmask;
908 	r = (u_int8_t *)&d;
909 	for (i = 0 ; i < dp->sa_len; i++) {
910 		r[i] = p[i] & q[i];
911 		/* XXX rough estimate */
912 		matchlen += (q[i] ? 8 : 0);
913 	}
914 
915 	/* need to overwrite len/family portion as we don't compare them */
916 	s.ss_len = sp->sa_len;
917 	s.ss_family = sp->sa_family;
918 	d.ss_len = dp->sa_len;
919 	d.ss_family = dp->sa_family;
920 
921 	if (bcmp(&s, ep->src, ep->src->sa_len) == 0 &&
922 	    bcmp(&d, ep->dst, ep->dst->sa_len) == 0) {
923 		return matchlen;
924 	} else
925 		return 0;
926 }
927 #endif
928 
929 static void
930 encap_fillarg(m, ep)
931 	struct mbuf *m;
932 	const struct encaptab *ep;
933 {
934 	struct mbuf *n;
935 
936 	n = m_aux_add(m, AF_INET, IPPROTO_IPV4);
937 	if (n) {
938 		*mtod(n, void **) = ep->arg;
939 		n->m_len = sizeof(void *);
940 	}
941 }
942 
943 void *
944 encap_getarg(m)
945 	struct mbuf *m;
946 {
947 	void *p;
948 	struct mbuf *n;
949 
950 	p = NULL;
951 	n = m_aux_find(m, AF_INET, IPPROTO_IPV4);
952 	if (n) {
953 		if (n->m_len == sizeof(void *))
954 			p = *mtod(n, void **);
955 		m_aux_delete(m, n);
956 	}
957 	return p;
958 }
959