xref: /netbsd-src/sys/netinet/ip_encap.c (revision d48f14661dda8638fee055ba15d35bdfb29b9fa8)
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 1.28 2006/05/28 11:07:04 liamjfoy 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(struct mbuf *, int, int, enum direction);
134 #endif
135 #ifdef INET6
136 static struct encaptab *encap6_lookup(struct mbuf *, int, int, enum direction);
137 #endif
138 static int encap_add(struct encaptab *);
139 static int encap_remove(struct encaptab *);
140 static int encap_afcheck(int, const struct sockaddr *, const struct sockaddr *);
141 #ifdef USE_RADIX
142 static struct radix_node_head *encap_rnh(int);
143 static int mask_matchlen(const struct sockaddr *);
144 #endif
145 #ifndef USE_RADIX
146 static int mask_match(const struct encaptab *, const struct sockaddr *,
147 		const struct sockaddr *);
148 #endif
149 static void encap_fillarg(struct mbuf *, const struct encaptab *);
150 
151 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
152 
153 #ifdef USE_RADIX
154 extern int max_keylen;	/* radix.c */
155 struct radix_node_head *encap_head[2];	/* 0 for AF_INET, 1 for AF_INET6 */
156 #endif
157 
158 void
159 encap_setkeylen(void)
160 {
161 #ifdef USE_RADIX
162 	if (sizeof(struct pack4) > max_keylen)
163 		max_keylen = sizeof(struct pack4);
164 #ifdef INET6
165 	if (sizeof(struct pack6) > max_keylen)
166 		max_keylen = sizeof(struct pack6);
167 #endif
168 #endif
169 }
170 
171 void
172 encap_init(void)
173 {
174 	static int initialized = 0;
175 
176 	if (initialized)
177 		return;
178 	initialized++;
179 #if 0
180 	/*
181 	 * we cannot use LIST_INIT() here, since drivers may want to call
182 	 * encap_attach(), on driver attach.  encap_init() will be called
183 	 * on AF_INET{,6} initialization, which happens after driver
184 	 * initialization - using LIST_INIT() here can nuke encap_attach()
185 	 * from drivers.
186 	 */
187 	LIST_INIT(&encaptab);
188 #endif
189 
190 #ifdef USE_RADIX
191 	/*
192 	 * initialize radix lookup table.
193 	 * max_keylen initialization happen in the rn_init().
194 	 */
195 	rn_init();
196 	rn_inithead((void *)&encap_head[0], sizeof(struct sockaddr_pack) << 3);
197 #ifdef INET6
198 	rn_inithead((void *)&encap_head[1], sizeof(struct sockaddr_pack) << 3);
199 #endif
200 #endif
201 }
202 
203 #ifdef INET
204 static struct encaptab *
205 encap4_lookup(struct mbuf *m, int off, int proto, 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 encap4_input(struct mbuf *m, ...)
298 {
299 	int off, proto;
300 	va_list ap;
301 	const struct protosw *psw;
302 	struct encaptab *match;
303 
304 	va_start(ap, m);
305 	off = va_arg(ap, int);
306 	proto = va_arg(ap, int);
307 	va_end(ap);
308 
309 	match = encap4_lookup(m, off, proto, INBOUND);
310 
311 	if (match) {
312 		/* found a match, "match" has the best one */
313 		psw = match->psw;
314 		if (psw && psw->pr_input) {
315 			encap_fillarg(m, match);
316 			(*psw->pr_input)(m, off, proto);
317 		} else
318 			m_freem(m);
319 		return;
320 	}
321 
322 	/* last resort: inject to raw socket */
323 	rip_input(m, off, proto);
324 }
325 #endif
326 
327 #ifdef INET6
328 static struct encaptab *
329 encap6_lookup(struct mbuf *m, int off, int proto, 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(struct mbuf **mp, int *offp, int proto)
402 {
403 	struct mbuf *m = *mp;
404 	const struct ip6protosw *psw;
405 	struct encaptab *match;
406 
407 	match = encap6_lookup(m, *offp, proto, INBOUND);
408 
409 	if (match) {
410 		/* found a match */
411 		psw = (const struct ip6protosw *)match->psw;
412 		if (psw && psw->pr_input) {
413 			encap_fillarg(m, match);
414 			return (*psw->pr_input)(mp, offp, proto);
415 		} else {
416 			m_freem(m);
417 			return IPPROTO_DONE;
418 		}
419 	}
420 
421 	/* last resort: inject to raw socket */
422 	return rip6_input(mp, offp, proto);
423 }
424 #endif
425 
426 static int
427 encap_add(struct encaptab *ep)
428 {
429 #ifdef USE_RADIX
430 	struct radix_node_head *rnh = encap_rnh(ep->af);
431 #endif
432 	int error = 0;
433 
434 	LIST_INSERT_HEAD(&encaptab, ep, chain);
435 #ifdef USE_RADIX
436 	if (!ep->func && rnh) {
437 		if (!rnh->rnh_addaddr((caddr_t)ep->addrpack,
438 		    (caddr_t)ep->maskpack, rnh, ep->nodes)) {
439 			error = EEXIST;
440 			goto fail;
441 		}
442 	}
443 #endif
444 	return error;
445 
446  fail:
447 	LIST_REMOVE(ep, chain);
448 	return error;
449 }
450 
451 static int
452 encap_remove(struct encaptab *ep)
453 {
454 #ifdef USE_RADIX
455 	struct radix_node_head *rnh = encap_rnh(ep->af);
456 #endif
457 	int error = 0;
458 
459 	LIST_REMOVE(ep, chain);
460 #ifdef USE_RADIX
461 	if (!ep->func && rnh) {
462 		if (!rnh->rnh_deladdr((caddr_t)ep->addrpack,
463 		    (caddr_t)ep->maskpack, rnh))
464 			error = ESRCH;
465 	}
466 #endif
467 	return error;
468 }
469 
470 static int
471 encap_afcheck(int af, const struct sockaddr *sp, const struct sockaddr *dp)
472 {
473 	if (sp && dp) {
474 		if (sp->sa_len != dp->sa_len)
475 			return EINVAL;
476 		if (af != sp->sa_family || af != dp->sa_family)
477 			return EINVAL;
478 	} else if (!sp && !dp)
479 		;
480 	else
481 		return EINVAL;
482 
483 	switch (af) {
484 	case AF_INET:
485 		if (sp && sp->sa_len != sizeof(struct sockaddr_in))
486 			return EINVAL;
487 		if (dp && dp->sa_len != sizeof(struct sockaddr_in))
488 			return EINVAL;
489 		break;
490 #ifdef INET6
491 	case AF_INET6:
492 		if (sp && sp->sa_len != sizeof(struct sockaddr_in6))
493 			return EINVAL;
494 		if (dp && dp->sa_len != sizeof(struct sockaddr_in6))
495 			return EINVAL;
496 		break;
497 #endif
498 	default:
499 		return EAFNOSUPPORT;
500 	}
501 
502 	return 0;
503 }
504 
505 /*
506  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
507  * length of mask (sm and dm) is assumed to be same as sp/dp.
508  * Return value will be necessary as input (cookie) for encap_detach().
509  */
510 const struct encaptab *
511 encap_attach(int af, int proto,
512     const struct sockaddr *sp, const struct sockaddr *sm,
513     const struct sockaddr *dp, const struct sockaddr *dm,
514     const struct protosw *psw, void *arg)
515 {
516 	struct encaptab *ep;
517 	int error;
518 	int s;
519 	size_t l;
520 	struct pack4 *pack4;
521 #ifdef INET6
522 	struct pack6 *pack6;
523 #endif
524 
525 	s = splsoftnet();
526 	/* sanity check on args */
527 	error = encap_afcheck(af, sp, dp);
528 	if (error)
529 		goto fail;
530 
531 	/* check if anyone have already attached with exactly same config */
532 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
533 		if (ep->af != af)
534 			continue;
535 		if (ep->proto != proto)
536 			continue;
537 		if (ep->func)
538 			continue;
539 #ifdef DIAGNOSTIC
540 		if (!ep->src || !ep->dst || !ep->srcmask || !ep->dstmask)
541 			panic("null pointers in encaptab");
542 #endif
543 		if (ep->src->sa_len != sp->sa_len ||
544 		    bcmp(ep->src, sp, sp->sa_len) != 0 ||
545 		    bcmp(ep->srcmask, sm, sp->sa_len) != 0)
546 			continue;
547 		if (ep->dst->sa_len != dp->sa_len ||
548 		    bcmp(ep->dst, dp, dp->sa_len) != 0 ||
549 		    bcmp(ep->dstmask, dm, dp->sa_len) != 0)
550 			continue;
551 
552 		error = EEXIST;
553 		goto fail;
554 	}
555 
556 	switch (af) {
557 	case AF_INET:
558 		l = sizeof(*pack4);
559 		break;
560 #ifdef INET6
561 	case AF_INET6:
562 		l = sizeof(*pack6);
563 		break;
564 #endif
565 	default:
566 		goto fail;
567 	}
568 
569 	/* M_NETADDR ok? */
570 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT|M_ZERO);
571 	if (ep == NULL) {
572 		error = ENOBUFS;
573 		goto fail;
574 	}
575 	ep->addrpack = malloc(l, M_NETADDR, M_NOWAIT|M_ZERO);
576 	if (ep->addrpack == NULL) {
577 		error = ENOBUFS;
578 		goto gc;
579 	}
580 	ep->maskpack = malloc(l, M_NETADDR, M_NOWAIT|M_ZERO);
581 	if (ep->maskpack == NULL) {
582 		error = ENOBUFS;
583 		goto gc;
584 	}
585 
586 	ep->af = af;
587 	ep->proto = proto;
588 	ep->addrpack->sa_len = l & 0xff;
589 	ep->maskpack->sa_len = l & 0xff;
590 	switch (af) {
591 	case AF_INET:
592 		pack4 = (struct pack4 *)ep->addrpack;
593 		ep->src = (struct sockaddr *)&pack4->mine;
594 		ep->dst = (struct sockaddr *)&pack4->yours;
595 		pack4 = (struct pack4 *)ep->maskpack;
596 		ep->srcmask = (struct sockaddr *)&pack4->mine;
597 		ep->dstmask = (struct sockaddr *)&pack4->yours;
598 		break;
599 #ifdef INET6
600 	case AF_INET6:
601 		pack6 = (struct pack6 *)ep->addrpack;
602 		ep->src = (struct sockaddr *)&pack6->mine;
603 		ep->dst = (struct sockaddr *)&pack6->yours;
604 		pack6 = (struct pack6 *)ep->maskpack;
605 		ep->srcmask = (struct sockaddr *)&pack6->mine;
606 		ep->dstmask = (struct sockaddr *)&pack6->yours;
607 		break;
608 #endif
609 	}
610 
611 	bcopy(sp, ep->src, sp->sa_len);
612 	bcopy(sm, ep->srcmask, sp->sa_len);
613 	bcopy(dp, ep->dst, dp->sa_len);
614 	bcopy(dm, ep->dstmask, dp->sa_len);
615 	ep->psw = psw;
616 	ep->arg = arg;
617 
618 	error = encap_add(ep);
619 	if (error)
620 		goto gc;
621 
622 	error = 0;
623 	splx(s);
624 	return ep;
625 
626 gc:
627 	if (ep->addrpack)
628 		free(ep->addrpack, M_NETADDR);
629 	if (ep->maskpack)
630 		free(ep->maskpack, M_NETADDR);
631 	if (ep)
632 		free(ep, M_NETADDR);
633 fail:
634 	splx(s);
635 	return NULL;
636 }
637 
638 const struct encaptab *
639 encap_attach_func(int af, int proto,
640     int (*func)(struct mbuf *, int, int, void *),
641     const struct protosw *psw, void *arg)
642 {
643 	struct encaptab *ep;
644 	int error;
645 	int s;
646 
647 	s = splsoftnet();
648 	/* sanity check on args */
649 	if (!func) {
650 		error = EINVAL;
651 		goto fail;
652 	}
653 
654 	error = encap_afcheck(af, NULL, NULL);
655 	if (error)
656 		goto fail;
657 
658 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
659 	if (ep == NULL) {
660 		error = ENOBUFS;
661 		goto fail;
662 	}
663 	bzero(ep, sizeof(*ep));
664 
665 	ep->af = af;
666 	ep->proto = proto;
667 	ep->func = func;
668 	ep->psw = psw;
669 	ep->arg = arg;
670 
671 	error = encap_add(ep);
672 	if (error)
673 		goto fail;
674 
675 	error = 0;
676 	splx(s);
677 	return ep;
678 
679 fail:
680 	splx(s);
681 	return NULL;
682 }
683 
684 /* XXX encap4_ctlinput() is necessary if we set DF=1 on outer IPv4 header */
685 
686 #ifdef INET6
687 void
688 encap6_ctlinput(int cmd, struct sockaddr *sa, void *d0)
689 {
690 	void *d = d0;
691 	struct ip6_hdr *ip6;
692 	struct mbuf *m;
693 	int off;
694 	struct ip6ctlparam *ip6cp = NULL;
695 	int nxt;
696 	struct encaptab *ep;
697 	const struct ip6protosw *psw;
698 
699 	if (sa->sa_family != AF_INET6 ||
700 	    sa->sa_len != sizeof(struct sockaddr_in6))
701 		return;
702 
703 	if ((unsigned)cmd >= PRC_NCMDS)
704 		return;
705 	if (cmd == PRC_HOSTDEAD)
706 		d = NULL;
707 	else if (cmd == PRC_MSGSIZE)
708 		; /* special code is present, see below */
709 	else if (inet6ctlerrmap[cmd] == 0)
710 		return;
711 
712 	/* if the parameter is from icmp6, decode it. */
713 	if (d != NULL) {
714 		ip6cp = (struct ip6ctlparam *)d;
715 		m = ip6cp->ip6c_m;
716 		ip6 = ip6cp->ip6c_ip6;
717 		off = ip6cp->ip6c_off;
718 		nxt = ip6cp->ip6c_nxt;
719 
720 		if (ip6 && cmd == PRC_MSGSIZE) {
721 			int valid = 0;
722 			struct encaptab *match;
723 
724 			/*
725 		 	* Check to see if we have a valid encap configuration.
726 		 	*/
727 			match = encap6_lookup(m, off, nxt, OUTBOUND);
728 			if (match)
729 				valid++;
730 
731 			/*
732 		 	* Depending on the value of "valid" and routing table
733 		 	* size (mtudisc_{hi,lo}wat), we will:
734 		 	* - recalcurate the new MTU and create the
735 		 	*   corresponding routing entry, or
736 		 	* - ignore the MTU change notification.
737 		 	*/
738 			icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
739 		}
740 	} else {
741 		m = NULL;
742 		ip6 = NULL;
743 		nxt = -1;
744 	}
745 
746 	/* inform all listeners */
747 	for (ep = LIST_FIRST(&encaptab); ep; ep = LIST_NEXT(ep, chain)) {
748 		if (ep->af != AF_INET6)
749 			continue;
750 		if (ep->proto >= 0 && ep->proto != nxt)
751 			continue;
752 
753 		/* should optimize by looking at address pairs */
754 
755 		/* XXX need to pass ep->arg or ep itself to listeners */
756 		psw = (const struct ip6protosw *)ep->psw;
757 		if (psw && psw->pr_ctlinput)
758 			(*psw->pr_ctlinput)(cmd, sa, d);
759 	}
760 
761 	rip6_ctlinput(cmd, sa, d0);
762 }
763 #endif
764 
765 int
766 encap_detach(const struct encaptab *cookie)
767 {
768 	const struct encaptab *ep = cookie;
769 	struct encaptab *p;
770 	int error;
771 
772 	for (p = LIST_FIRST(&encaptab); p; p = LIST_NEXT(p, chain)) {
773 		if (p == ep) {
774 			error = encap_remove(p);
775 			if (error)
776 				return error;
777 			if (!ep->func) {
778 				free(p->addrpack, M_NETADDR);
779 				free(p->maskpack, M_NETADDR);
780 			}
781 			free(p, M_NETADDR);	/*XXX*/
782 			return 0;
783 		}
784 	}
785 
786 	return ENOENT;
787 }
788 
789 #ifdef USE_RADIX
790 static struct radix_node_head *
791 encap_rnh(int af)
792 {
793 
794 	switch (af) {
795 	case AF_INET:
796 		return encap_head[0];
797 #ifdef INET6
798 	case AF_INET6:
799 		return encap_head[1];
800 #endif
801 	default:
802 		return NULL;
803 	}
804 }
805 
806 static int
807 mask_matchlen(const struct sockaddr *sa)
808 {
809 	const char *p, *ep;
810 	int l;
811 
812 	p = (const char *)sa;
813 	ep = p + sa->sa_len;
814 	p += 2;	/* sa_len + sa_family */
815 
816 	l = 0;
817 	while (p < ep) {
818 		l += (*p ? 8 : 0);	/* estimate */
819 		p++;
820 	}
821 	return l;
822 }
823 #endif
824 
825 #ifndef USE_RADIX
826 static int
827 mask_match(const struct encaptab *ep,
828 	   const struct sockaddr *sp,
829 	   const struct sockaddr *dp)
830 {
831 	struct sockaddr_storage s;
832 	struct sockaddr_storage d;
833 	int i;
834 	const u_int8_t *p, *q;
835 	u_int8_t *r;
836 	int matchlen;
837 
838 #ifdef DIAGNOSTIC
839 	if (ep->func)
840 		panic("wrong encaptab passed to mask_match");
841 #endif
842 	if (sp->sa_len > sizeof(s) || dp->sa_len > sizeof(d))
843 		return 0;
844 	if (sp->sa_family != ep->af || dp->sa_family != ep->af)
845 		return 0;
846 	if (sp->sa_len != ep->src->sa_len || dp->sa_len != ep->dst->sa_len)
847 		return 0;
848 
849 	matchlen = 0;
850 
851 	p = (const u_int8_t *)sp;
852 	q = (const u_int8_t *)ep->srcmask;
853 	r = (u_int8_t *)&s;
854 	for (i = 0 ; i < sp->sa_len; i++) {
855 		r[i] = p[i] & q[i];
856 		/* XXX estimate */
857 		matchlen += (q[i] ? 8 : 0);
858 	}
859 
860 	p = (const u_int8_t *)dp;
861 	q = (const u_int8_t *)ep->dstmask;
862 	r = (u_int8_t *)&d;
863 	for (i = 0 ; i < dp->sa_len; i++) {
864 		r[i] = p[i] & q[i];
865 		/* XXX rough estimate */
866 		matchlen += (q[i] ? 8 : 0);
867 	}
868 
869 	/* need to overwrite len/family portion as we don't compare them */
870 	s.ss_len = sp->sa_len;
871 	s.ss_family = sp->sa_family;
872 	d.ss_len = dp->sa_len;
873 	d.ss_family = dp->sa_family;
874 
875 	if (bcmp(&s, ep->src, ep->src->sa_len) == 0 &&
876 	    bcmp(&d, ep->dst, ep->dst->sa_len) == 0) {
877 		return matchlen;
878 	} else
879 		return 0;
880 }
881 #endif
882 
883 static void
884 encap_fillarg(struct mbuf *m, const struct encaptab *ep)
885 {
886 	struct m_tag *mtag;
887 
888 	mtag = m_tag_get(PACKET_TAG_ENCAP, sizeof(void *), M_NOWAIT);
889 	if (mtag) {
890 		*(void **)(mtag + 1) = ep->arg;
891 		m_tag_prepend(m, mtag);
892 	}
893 }
894 
895 void *
896 encap_getarg(struct mbuf *m)
897 {
898 	void *p;
899 	struct m_tag *mtag;
900 
901 	p = NULL;
902 	mtag = m_tag_find(m, PACKET_TAG_ENCAP, NULL);
903 	if (mtag != NULL) {
904 		p = *(void **)(mtag + 1);
905 		m_tag_delete(m, mtag);
906 	}
907 	return p;
908 }
909