xref: /netbsd-src/sys/netinet/ip_encap.c (revision 0dd5877adce57db949b16ae963e5a6831cccdfb6)
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 1.7 2001/12/21 06:30:43 itojun 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 "ipip.h"
111 #if NIPIP > 0
112 # include <netinet/ip_ipip.h>
113 #else
114 # ifdef MROUTING
115 #  include <netinet/ip_mroute.h>
116 # endif
117 #endif
118 
119 #include <net/net_osdep.h>
120 
121 /* to lookup a pair of address using radix tree */
122 struct sockaddr_pack {
123 	u_int8_t sp_len;
124 	u_int8_t sp_family;	/* not really used */
125 	/* followed by variable-length data */
126 } __attribute__((__packed__));
127 
128 struct pack4 {
129 	struct sockaddr_pack p;
130 	struct sockaddr_in mine;
131 	struct sockaddr_in yours;
132 } __attribute__((__packed__));
133 struct pack6 {
134 	struct sockaddr_pack p;
135 	struct sockaddr_in6 mine;
136 	struct sockaddr_in6 yours;
137 } __attribute__((__packed__));
138 
139 enum direction { INBOUND, OUTBOUND };
140 
141 #ifdef INET
142 static struct encaptab *encap4_lookup __P((struct mbuf *, int, int,
143 	enum direction));
144 #endif
145 #ifdef INET6
146 static struct encaptab *encap6_lookup __P((struct mbuf *, int, int,
147 	enum direction));
148 #endif
149 static int encap_add __P((struct encaptab *));
150 static int encap_remove __P((struct encaptab *));
151 static int encap_afcheck __P((int, const struct sockaddr *, const struct sockaddr *));
152 #ifdef USE_RADIX
153 static struct radix_node_head *encap_rnh __P((int));
154 static int mask_matchlen __P((const struct sockaddr *));
155 #endif
156 #ifndef USE_RADIX
157 static int mask_match __P((const struct encaptab *, const struct sockaddr *,
158 		const struct sockaddr *));
159 #endif
160 static void encap_fillarg __P((struct mbuf *, const struct encaptab *));
161 
162 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
163 
164 #ifdef USE_RADIX
165 extern int max_keylen;	/* radix.c */
166 struct radix_node_head *encap_head[2];	/* 0 for AF_INET, 1 for AF_INET6 */
167 #endif
168 
169 void
170 encap_init()
171 {
172 	static int initialized = 0;
173 
174 	if (initialized)
175 		return;
176 	initialized++;
177 #if 0
178 	/*
179 	 * we cannot use LIST_INIT() here, since drivers may want to call
180 	 * encap_attach(), on driver attach.  encap_init() will be called
181 	 * on AF_INET{,6} initialization, which happens after driver
182 	 * initialization - using LIST_INIT() here can nuke encap_attach()
183 	 * from drivers.
184 	 */
185 	LIST_INIT(&encaptab);
186 #endif
187 
188 #ifdef USE_RADIX
189 	/*
190 	 * initialize radix lookup table.
191 	 * max_keylen initialization should happen before the call to rn_init().
192 	 */
193 	rn_inithead((void **)&encap_head[0], sizeof(struct sockaddr_pack) << 3);
194 	if (sizeof(struct pack4) > max_keylen)
195 		max_keylen = sizeof(struct pack4);
196 #ifdef INET6
197 	rn_inithead((void **)&encap_head[1], sizeof(struct sockaddr_pack) << 3);
198 	if (sizeof(struct pack6) > max_keylen)
199 		max_keylen = sizeof(struct pack6);
200 #endif
201 #endif
202 }
203 
204 #ifdef INET
205 static struct encaptab *
206 encap4_lookup(m, off, proto, dir)
207 	struct mbuf *m;
208 	int off;
209 	int proto;
210 	enum direction dir;
211 {
212 	struct ip *ip;
213 	struct pack4 pack;
214 	struct encaptab *ep, *match;
215 	int prio, matchprio;
216 #ifdef USE_RADIX
217 	struct radix_node_head *rnh = encap_rnh(AF_INET);
218 	struct radix_node *rn;
219 #endif
220 
221 #ifdef DIAGNOSTIC
222 	if (m->m_len < sizeof(*ip))
223 		panic("encap4_lookup");
224 #endif
225 	ip = mtod(m, struct ip *);
226 
227 	bzero(&pack, sizeof(pack));
228 	pack.p.sp_len = sizeof(pack);
229 	pack.mine.sin_family = pack.yours.sin_family = AF_INET;
230 	pack.mine.sin_len = pack.yours.sin_len = sizeof(struct sockaddr_in);
231 	if (dir == INBOUND) {
232 		pack.mine.sin_addr = ip->ip_dst;
233 		pack.yours.sin_addr = ip->ip_src;
234 	} else {
235 		pack.mine.sin_addr = ip->ip_src;
236 		pack.yours.sin_addr = ip->ip_dst;
237 	}
238 
239 	match = NULL;
240 	matchprio = 0;
241 
242 #ifdef USE_RADIX
243 	rn = rnh->rnh_matchaddr((caddr_t)&pack, rnh);
244 	if (rn && (rn->rn_flags & RNF_ROOT) == 0) {
245 		match = (struct encaptab *)rn;
246 		matchprio = mask_matchlen(match->srcmask) +
247 		    mask_matchlen(match->dstmask);
248 	}
249 #endif
250 
251 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
252 		if (ep->af != AF_INET)
253 			continue;
254 		if (ep->proto >= 0 && ep->proto != proto)
255 			continue;
256 		if (ep->func)
257 			prio = (*ep->func)(m, off, proto, ep->arg);
258 		else {
259 #ifdef USE_RADIX
260 			continue;
261 #else
262 			prio = mask_match(ep, (struct sockaddr *)&pack.mine,
263 			    (struct sockaddr *)&pack.yours);
264 #endif
265 		}
266 
267 		/*
268 		 * We prioritize the matches by using bit length of the
269 		 * matches.  mask_match() and user-supplied matching function
270 		 * should return the bit length of the matches (for example,
271 		 * if both src/dst are matched for IPv4, 64 should be returned).
272 		 * 0 or negative return value means "it did not match".
273 		 *
274 		 * The question is, since we have two "mask" portion, we
275 		 * cannot really define total order between entries.
276 		 * For example, which of these should be preferred?
277 		 * mask_match() returns 48 (32 + 16) for both of them.
278 		 *	src=3ffe::/16, dst=3ffe:501::/32
279 		 *	src=3ffe:501::/32, dst=3ffe::/16
280 		 *
281 		 * We need to loop through all the possible candidates
282 		 * to get the best match - the search takes O(n) for
283 		 * n attachments (i.e. interfaces).
284 		 *
285 		 * For radix-based lookup, I guess source takes precedence.
286 		 * See rn_{refines,lexobetter} for the correct answer.
287 		 */
288 		if (prio <= 0)
289 			continue;
290 		if (prio > matchprio) {
291 			matchprio = prio;
292 			match = ep;
293 		}
294 	}
295 
296 	return match;
297 #undef s
298 #undef d
299 }
300 
301 void
302 #if __STDC__
303 encap4_input(struct mbuf *m, ...)
304 #else
305 encap4_input(m, va_alist)
306 	struct mbuf *m;
307 	va_dcl
308 #endif
309 {
310 	int off, proto;
311 	va_list ap;
312 	const struct protosw *psw;
313 	struct encaptab *match;
314 
315 	va_start(ap, m);
316 	off = va_arg(ap, int);
317 	proto = va_arg(ap, int);
318 	va_end(ap);
319 
320 	match = encap4_lookup(m, off, proto, INBOUND);
321 
322 	if (match) {
323 		/* found a match, "match" has the best one */
324 		psw = match->psw;
325 		if (psw && psw->pr_input) {
326 			encap_fillarg(m, match);
327 			(*psw->pr_input)(m, off, proto);
328 		} else
329 			m_freem(m);
330 		return;
331 	}
332 
333 	/* last resort: inject to raw socket */
334 	rip_input(m, off, proto);
335 }
336 #endif
337 
338 #ifdef INET6
339 static struct encaptab *
340 encap6_lookup(m, off, proto, dir)
341 	struct mbuf *m;
342 	int off;
343 	int proto;
344 	enum direction dir;
345 {
346 	struct ip6_hdr *ip6;
347 	struct pack6 pack;
348 	int prio, matchprio;
349 	struct encaptab *ep, *match;
350 #ifdef USE_RADIX
351 	struct radix_node_head *rnh = encap_rnh(AF_INET6);
352 	struct radix_node *rn;
353 #endif
354 
355 #ifdef DIAGNOSTIC
356 	if (m->m_len < sizeof(*ip6))
357 		panic("encap6_lookup");
358 #endif
359 	ip6 = mtod(m, struct ip6_hdr *);
360 
361 	bzero(&pack, sizeof(pack));
362 	pack.p.sp_len = sizeof(pack);
363 	pack.mine.sin6_family = pack.yours.sin6_family = AF_INET6;
364 	pack.mine.sin6_len = pack.yours.sin6_len = sizeof(struct sockaddr_in6);
365 	if (dir == INBOUND) {
366 		pack.mine.sin6_addr = ip6->ip6_dst;
367 		pack.yours.sin6_addr = ip6->ip6_src;
368 	} else {
369 		pack.mine.sin6_addr = ip6->ip6_src;
370 		pack.yours.sin6_addr = ip6->ip6_dst;
371 	}
372 
373 	match = NULL;
374 	matchprio = 0;
375 
376 #ifdef USE_RADIX
377 	rn = rnh->rnh_matchaddr((caddr_t)&pack, rnh);
378 	if (rn && (rn->rn_flags & RNF_ROOT) == 0) {
379 		match = (struct encaptab *)rn;
380 		matchprio = mask_matchlen(match->srcmask) +
381 		    mask_matchlen(match->dstmask);
382 	}
383 #endif
384 
385 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
386 		if (ep->af != AF_INET6)
387 			continue;
388 		if (ep->proto >= 0 && ep->proto != proto)
389 			continue;
390 		if (ep->func)
391 			prio = (*ep->func)(m, off, proto, ep->arg);
392 		else {
393 #ifdef USE_RADIX
394 			continue;
395 #else
396 			prio = mask_match(ep, (struct sockaddr *)&pack.mine,
397 			    (struct sockaddr *)&pack.yours);
398 #endif
399 		}
400 
401 		/* see encap4_lookup() for issues here */
402 		if (prio <= 0)
403 			continue;
404 		if (prio > matchprio) {
405 			matchprio = prio;
406 			match = ep;
407 		}
408 	}
409 
410 	return match;
411 #undef s
412 #undef d
413 }
414 
415 int
416 encap6_input(mp, offp, proto)
417 	struct mbuf **mp;
418 	int *offp;
419 	int proto;
420 {
421 	struct mbuf *m = *mp;
422 	const struct ip6protosw *psw;
423 	struct encaptab *match;
424 
425 	match = encap6_lookup(m, *offp, proto, INBOUND);
426 
427 	if (match) {
428 		/* found a match */
429 		psw = (const struct ip6protosw *)match->psw;
430 		if (psw && psw->pr_input) {
431 			encap_fillarg(m, match);
432 			return (*psw->pr_input)(mp, offp, proto);
433 		} else {
434 			m_freem(m);
435 			return IPPROTO_DONE;
436 		}
437 	}
438 
439 	/* last resort: inject to raw socket */
440 	return rip6_input(mp, offp, proto);
441 }
442 #endif
443 
444 static int
445 encap_add(ep)
446 	struct encaptab *ep;
447 {
448 #ifdef USE_RADIX
449 	struct radix_node_head *rnh = encap_rnh(ep->af);
450 #endif
451 	int error = 0;
452 
453 	LIST_INSERT_HEAD(&encaptab, ep, chain);
454 #ifdef USE_RADIX
455 	if (!ep->func && rnh) {
456 		if (!rnh->rnh_addaddr((caddr_t)ep->addrpack,
457 		    (caddr_t)ep->maskpack, rnh, ep->nodes)) {
458 			error = EEXIST;
459 			goto fail;
460 		}
461 	}
462 #endif
463 	return error;
464 
465  fail:
466 	LIST_REMOVE(ep, chain);
467 	return error;
468 }
469 
470 static int
471 encap_remove(ep)
472 	struct encaptab *ep;
473 {
474 #ifdef USE_RADIX
475 	struct radix_node_head *rnh = encap_rnh(ep->af);
476 #endif
477 	int error = 0;
478 
479 	LIST_REMOVE(ep, chain);
480 #ifdef USE_RADIX
481 	if (!ep->func && rnh) {
482 		if (!rnh->rnh_deladdr((caddr_t)ep->addrpack,
483 		    (caddr_t)ep->maskpack, rnh))
484 			error = ESRCH;
485 	}
486 #endif
487 	return error;
488 }
489 
490 static int
491 encap_afcheck(af, sp, dp)
492 	int af;
493 	const struct sockaddr *sp;
494 	const struct sockaddr *dp;
495 {
496 	if (sp && dp) {
497 		if (sp->sa_len != dp->sa_len)
498 			return EINVAL;
499 		if (af != sp->sa_family || af != dp->sa_family)
500 			return EINVAL;
501 	} else if (!sp && !dp)
502 		;
503 	else
504 		return EINVAL;
505 
506 	switch (af) {
507 	case AF_INET:
508 		if (sp && sp->sa_len != sizeof(struct sockaddr_in))
509 			return EINVAL;
510 		if (dp && dp->sa_len != sizeof(struct sockaddr_in))
511 			return EINVAL;
512 		break;
513 #ifdef INET6
514 	case AF_INET6:
515 		if (sp && sp->sa_len != sizeof(struct sockaddr_in6))
516 			return EINVAL;
517 		if (dp && dp->sa_len != sizeof(struct sockaddr_in6))
518 			return EINVAL;
519 		break;
520 #endif
521 	default:
522 		return EAFNOSUPPORT;
523 	}
524 
525 	return 0;
526 }
527 
528 /*
529  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
530  * length of mask (sm and dm) is assumed to be same as sp/dp.
531  * Return value will be necessary as input (cookie) for encap_detach().
532  */
533 const struct encaptab *
534 encap_attach(af, proto, sp, sm, dp, dm, psw, arg)
535 	int af;
536 	int proto;
537 	const struct sockaddr *sp, *sm;
538 	const struct sockaddr *dp, *dm;
539 	const struct protosw *psw;
540 	void *arg;
541 {
542 	struct encaptab *ep;
543 	int error;
544 	int s;
545 	size_t l;
546 	struct pack4 *pack4;
547 #ifdef INET6
548 	struct pack6 *pack6;
549 #endif
550 
551 	s = splsoftnet();
552 	/* sanity check on args */
553 	error = encap_afcheck(af, sp, dp);
554 	if (error)
555 		goto fail;
556 
557 	/* check if anyone have already attached with exactly same config */
558 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
559 		if (ep->af != af)
560 			continue;
561 		if (ep->proto != proto)
562 			continue;
563 		if (ep->func)
564 			continue;
565 #ifdef DIAGNOSTIC
566 		if (!ep->src || !ep->dst || !ep->srcmask || !ep->dstmask)
567 			panic("null pointers in encaptab");
568 #endif
569 		if (ep->src->sa_len != sp->sa_len ||
570 		    bcmp(ep->src, sp, sp->sa_len) != 0 ||
571 		    bcmp(ep->srcmask, sm, sp->sa_len) != 0)
572 			continue;
573 		if (ep->dst->sa_len != dp->sa_len ||
574 		    bcmp(ep->dst, dp, dp->sa_len) != 0 ||
575 		    bcmp(ep->dstmask, dm, dp->sa_len) != 0)
576 			continue;
577 
578 		error = EEXIST;
579 		goto fail;
580 	}
581 
582 	switch (af) {
583 	case AF_INET:
584 		l = sizeof(*pack4);
585 		break;
586 #ifdef INET6
587 	case AF_INET6:
588 		l = sizeof(*pack6);
589 		break;
590 #endif
591 	default:
592 		goto fail;
593 	}
594 
595 #ifdef DIAGNOSTIC
596 	/* if l exceeds the value sa_len can possibly express, it's wrong. */
597 	if (l > (1 << (8 * sizeof(ep->addrpack->sa_len)))) {
598 		error = EINVAL;
599 		goto fail;
600 	}
601 #endif
602 
603 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/* M_NETADDR ok? */
604 	if (ep == NULL) {
605 		error = ENOBUFS;
606 		goto fail;
607 	}
608 	bzero(ep, sizeof(*ep));
609 	ep->addrpack = malloc(l, M_NETADDR, M_NOWAIT);
610 	if (ep->addrpack == NULL) {
611 		error = ENOBUFS;
612 		goto gc;
613 	}
614 	ep->maskpack = malloc(l, M_NETADDR, M_NOWAIT);
615 	if (ep->maskpack == NULL) {
616 		error = ENOBUFS;
617 		goto gc;
618 	}
619 
620 	ep->af = af;
621 	ep->proto = proto;
622 	ep->addrpack->sa_len = l & 0xff;
623 	ep->maskpack->sa_len = l & 0xff;
624 	switch (af) {
625 	case AF_INET:
626 		pack4 = (struct pack4 *)ep->addrpack;
627 		ep->src = (struct sockaddr *)&pack4->mine;
628 		ep->dst = (struct sockaddr *)&pack4->yours;
629 		pack4 = (struct pack4 *)ep->maskpack;
630 		ep->srcmask = (struct sockaddr *)&pack4->mine;
631 		ep->dstmask = (struct sockaddr *)&pack4->yours;
632 		break;
633 #ifdef INET6
634 	case AF_INET6:
635 		pack6 = (struct pack6 *)ep->addrpack;
636 		ep->src = (struct sockaddr *)&pack6->mine;
637 		ep->dst = (struct sockaddr *)&pack6->yours;
638 		pack6 = (struct pack6 *)ep->maskpack;
639 		ep->srcmask = (struct sockaddr *)&pack6->mine;
640 		ep->dstmask = (struct sockaddr *)&pack6->yours;
641 		break;
642 #endif
643 	}
644 
645 	bcopy(sp, ep->src, sp->sa_len);
646 	bcopy(sm, ep->srcmask, sp->sa_len);
647 	bcopy(dp, ep->dst, dp->sa_len);
648 	bcopy(dm, ep->dstmask, dp->sa_len);
649 	ep->psw = psw;
650 	ep->arg = arg;
651 
652 	error = encap_add(ep);
653 	if (error)
654 		goto gc;
655 
656 	error = 0;
657 	splx(s);
658 	return ep;
659 
660 gc:
661 	if (ep->addrpack)
662 		free(ep->addrpack, M_NETADDR);
663 	if (ep->maskpack)
664 		free(ep->maskpack, M_NETADDR);
665 	if (ep)
666 		free(ep, M_NETADDR);
667 fail:
668 	splx(s);
669 	return NULL;
670 }
671 
672 const struct encaptab *
673 encap_attach_func(af, proto, func, psw, arg)
674 	int af;
675 	int proto;
676 	int (*func) __P((const struct mbuf *, int, int, void *));
677 	const struct protosw *psw;
678 	void *arg;
679 {
680 	struct encaptab *ep;
681 	int error;
682 	int s;
683 
684 	s = splsoftnet();
685 	/* sanity check on args */
686 	if (!func) {
687 		error = EINVAL;
688 		goto fail;
689 	}
690 
691 	error = encap_afcheck(af, NULL, NULL);
692 	if (error)
693 		goto fail;
694 
695 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
696 	if (ep == NULL) {
697 		error = ENOBUFS;
698 		goto fail;
699 	}
700 	bzero(ep, sizeof(*ep));
701 
702 	ep->af = af;
703 	ep->proto = proto;
704 	ep->func = func;
705 	ep->psw = psw;
706 	ep->arg = arg;
707 
708 	error = encap_add(ep);
709 	if (error)
710 		goto fail;
711 
712 	error = 0;
713 	splx(s);
714 	return ep;
715 
716 fail:
717 	splx(s);
718 	return NULL;
719 }
720 
721 /* XXX encap4_ctlinput() is necessary if we set DF=1 on outer IPv4 header */
722 
723 #ifdef INET6
724 void
725 encap6_ctlinput(cmd, sa, d0)
726 	int cmd;
727 	struct sockaddr *sa;
728 	void *d0;
729 {
730 	void *d = d0;
731 	struct ip6_hdr *ip6;
732 	struct mbuf *m;
733 	int off;
734 	struct ip6ctlparam *ip6cp = NULL;
735 	const struct sockaddr_in6 *sa6_src = NULL;
736 	void *cmdarg;
737 	int nxt;
738 	struct encaptab *ep;
739 	const struct ip6protosw *psw;
740 
741 	if (sa->sa_family != AF_INET6 ||
742 	    sa->sa_len != sizeof(struct sockaddr_in6))
743 		return;
744 
745 	if ((unsigned)cmd >= PRC_NCMDS)
746 		return;
747 	if (cmd == PRC_HOSTDEAD)
748 		d = NULL;
749 	else if (cmd == PRC_MSGSIZE)
750 		; /* special code is present, see below */
751 	else if (inet6ctlerrmap[cmd] == 0)
752 		return;
753 
754 	/* if the parameter is from icmp6, decode it. */
755 	if (d != NULL) {
756 		ip6cp = (struct ip6ctlparam *)d;
757 		m = ip6cp->ip6c_m;
758 		ip6 = ip6cp->ip6c_ip6;
759 		off = ip6cp->ip6c_off;
760 		cmdarg = ip6cp->ip6c_cmdarg;
761 		sa6_src = ip6cp->ip6c_src;
762 		nxt = ip6cp->ip6c_nxt;
763 	} else {
764 		m = NULL;
765 		ip6 = NULL;
766 		cmdarg = NULL;
767 		sa6_src = &sa6_any;
768 		nxt = -1;
769 	}
770 
771 	if (ip6 && cmd == PRC_MSGSIZE) {
772 		int valid = 0;
773 		struct encaptab *match;
774 
775 		/*
776 		 * Check to see if we have a valid encap configuration.
777 		 */
778 		match = encap6_lookup(m, off, nxt, OUTBOUND);
779 
780 		if (match)
781 			valid++;
782 
783 		/*
784 		 * Depending on the value of "valid" and routing table
785 		 * size (mtudisc_{hi,lo}wat), we will:
786 		 * - recalcurate the new MTU and create the
787 		 *   corresponding routing entry, or
788 		 * - ignore the MTU change notification.
789 		 */
790 		icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
791 	}
792 
793 	/* inform all listeners */
794 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
795 		if (ep->af != AF_INET6)
796 			continue;
797 		if (ep->proto >= 0 && ep->proto != nxt)
798 			continue;
799 
800 		/* should optimize by looking at address pairs */
801 
802 		/* XXX need to pass ep->arg or ep itself to listeners */
803 		psw = (const struct ip6protosw *)ep->psw;
804 		if (psw && psw->pr_ctlinput)
805 			(*psw->pr_ctlinput)(cmd, sa, d);
806 	}
807 
808 	rip6_ctlinput(cmd, sa, d0);
809 }
810 #endif
811 
812 int
813 encap_detach(cookie)
814 	const struct encaptab *cookie;
815 {
816 	const struct encaptab *ep = cookie;
817 	struct encaptab *p;
818 	int error;
819 
820 	for (p = LIST_FIRST(&encaptab); p; p = LIST_NEXT(p, chain)) {
821 		if (p == ep) {
822 			error = encap_remove(p);
823 			if (error)
824 				return error;
825 			if (!ep->func) {
826 				free(p->addrpack, M_NETADDR);
827 				free(p->maskpack, M_NETADDR);
828 			}
829 			free(p, M_NETADDR);	/*XXX*/
830 			return 0;
831 		}
832 	}
833 
834 	return ENOENT;
835 }
836 
837 #ifdef USE_RADIX
838 static struct radix_node_head *
839 encap_rnh(af)
840 	int af;
841 {
842 
843 	switch (af) {
844 	case AF_INET:
845 		return encap_head[0];
846 #ifdef INET6
847 	case AF_INET6:
848 		return encap_head[1];
849 #endif
850 	default:
851 		return NULL;
852 	}
853 }
854 
855 static int
856 mask_matchlen(sa)
857 	const struct sockaddr *sa;
858 {
859 	const char *p, *ep;
860 	int l;
861 
862 	p = (const char *)sa;
863 	ep = p + sa->sa_len;
864 	p += 2;	/* sa_len + sa_family */
865 
866 	l = 0;
867 	while (p < ep) {
868 		l += (*p ? 8 : 0);	/* estimate */
869 		p++;
870 	}
871 	return l;
872 }
873 #endif
874 
875 #ifndef USE_RADIX
876 static int
877 mask_match(ep, sp, dp)
878 	const struct encaptab *ep;
879 	const struct sockaddr *sp;
880 	const struct sockaddr *dp;
881 {
882 	struct sockaddr_storage s;
883 	struct sockaddr_storage d;
884 	int i;
885 	const u_int8_t *p, *q;
886 	u_int8_t *r;
887 	int matchlen;
888 
889 #ifdef DIAGNOSTIC
890 	if (ep->func)
891 		panic("wrong encaptab passed to mask_match");
892 #endif
893 	if (sp->sa_len > sizeof(s) || dp->sa_len > sizeof(d))
894 		return 0;
895 	if (sp->sa_family != ep->af || dp->sa_family != ep->af)
896 		return 0;
897 	if (sp->sa_len != ep->src->sa_len || dp->sa_len != ep->dst->sa_len)
898 		return 0;
899 
900 	matchlen = 0;
901 
902 	p = (const u_int8_t *)sp;
903 	q = (const u_int8_t *)ep->srcmask;
904 	r = (u_int8_t *)&s;
905 	for (i = 0 ; i < sp->sa_len; i++) {
906 		r[i] = p[i] & q[i];
907 		/* XXX estimate */
908 		matchlen += (q[i] ? 8 : 0);
909 	}
910 
911 	p = (const u_int8_t *)dp;
912 	q = (const u_int8_t *)ep->dstmask;
913 	r = (u_int8_t *)&d;
914 	for (i = 0 ; i < dp->sa_len; i++) {
915 		r[i] = p[i] & q[i];
916 		/* XXX rough estimate */
917 		matchlen += (q[i] ? 8 : 0);
918 	}
919 
920 	/* need to overwrite len/family portion as we don't compare them */
921 	s.ss_len = sp->sa_len;
922 	s.ss_family = sp->sa_family;
923 	d.ss_len = dp->sa_len;
924 	d.ss_family = dp->sa_family;
925 
926 	if (bcmp(&s, ep->src, ep->src->sa_len) == 0 &&
927 	    bcmp(&d, ep->dst, ep->dst->sa_len) == 0) {
928 		return matchlen;
929 	} else
930 		return 0;
931 }
932 #endif
933 
934 static void
935 encap_fillarg(m, ep)
936 	struct mbuf *m;
937 	const struct encaptab *ep;
938 {
939 	struct mbuf *n;
940 
941 	n = m_aux_add(m, AF_INET, IPPROTO_IPV4);
942 	if (n) {
943 		*mtod(n, void **) = ep->arg;
944 		n->m_len = sizeof(void *);
945 	}
946 }
947 
948 void *
949 encap_getarg(m)
950 	struct mbuf *m;
951 {
952 	void *p;
953 	struct mbuf *n;
954 
955 	p = NULL;
956 	n = m_aux_find(m, AF_INET, IPPROTO_IPV4);
957 	if (n) {
958 		if (n->m_len == sizeof(void *))
959 			p = *mtod(n, void **);
960 		m_aux_delete(m, n);
961 	}
962 	return p;
963 }
964