xref: /netbsd-src/sys/netinet/ip_encap.c (revision a6f3f22f245acb8ee3bbf6871d7dce989204fa97)
1 /*	$NetBSD: ip_encap.c,v 1.46 2015/08/24 22:21:26 pooka Exp $	*/
2 /*	$KAME: ip_encap.c,v 1.73 2001/10/02 08:30:58 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 /*
33  * My grandfather said that there's a devil inside tunnelling technology...
34  *
35  * We have surprisingly many protocols that want packets with IP protocol
36  * #4 or #41.  Here's a list of protocols that want protocol #41:
37  *	RFC1933 configured tunnel
38  *	RFC1933 automatic tunnel
39  *	RFC2401 IPsec tunnel
40  *	RFC2473 IPv6 generic packet tunnelling
41  *	RFC2529 6over4 tunnel
42  *	RFC3056 6to4 tunnel
43  *	isatap tunnel
44  *	mobile-ip6 (uses RFC2473)
45  * Here's a list of protocol that want protocol #4:
46  *	RFC1853 IPv4-in-IPv4 tunnelling
47  *	RFC2003 IPv4 encapsulation within IPv4
48  *	RFC2344 reverse tunnelling for mobile-ip4
49  *	RFC2401 IPsec tunnel
50  * Well, what can I say.  They impose different en/decapsulation mechanism
51  * from each other, so they need separate protocol handler.  The only one
52  * we can easily determine by protocol # is IPsec, which always has
53  * AH/ESP/IPComp header right after outer IP header.
54  *
55  * So, clearly good old protosw does not work for protocol #4 and #41.
56  * The code will let you match protocol via src/dst address pair.
57  */
58 /* XXX is M_NETADDR correct? */
59 
60 /*
61  * The code will use radix table for tunnel lookup, for
62  * tunnels registered with encap_attach() with a addr/mask pair.
63  * Faster on machines with thousands of tunnel registerations (= interfaces).
64  *
65  * The code assumes that radix table code can handle non-continuous netmask,
66  * as it will pass radix table memory region with (src + dst) sockaddr pair.
67  *
68  * FreeBSD is excluded here as they make max_keylen a static variable, and
69  * thus forbid definition of radix table other than proper domains.
70  */
71 
72 #include <sys/cdefs.h>
73 __KERNEL_RCSID(0, "$NetBSD: ip_encap.c,v 1.46 2015/08/24 22:21:26 pooka Exp $");
74 
75 #ifdef _KERNEL_OPT
76 #include "opt_mrouting.h"
77 #include "opt_inet.h"
78 #endif
79 
80 #include <sys/param.h>
81 #include <sys/systm.h>
82 #include <sys/socket.h>
83 #include <sys/sockio.h>
84 #include <sys/mbuf.h>
85 #include <sys/errno.h>
86 #include <sys/protosw.h>
87 #include <sys/queue.h>
88 
89 #include <net/if.h>
90 #include <net/route.h>
91 
92 #include <netinet/in.h>
93 #include <netinet/in_systm.h>
94 #include <netinet/ip.h>
95 #include <netinet/ip_var.h>
96 #include <netinet/ip_encap.h>
97 #ifdef MROUTING
98 #include <netinet/ip_mroute.h>
99 #endif /* MROUTING */
100 
101 #ifdef INET6
102 #include <netinet/ip6.h>
103 #include <netinet6/ip6_var.h>
104 #include <netinet6/ip6protosw.h>
105 #include <netinet6/in6_var.h>
106 #include <netinet6/in6_pcb.h>
107 #include <netinet/icmp6.h>
108 #endif
109 
110 #include <net/net_osdep.h>
111 
112 enum direction { INBOUND, OUTBOUND };
113 
114 #ifdef INET
115 static struct encaptab *encap4_lookup(struct mbuf *, int, int, enum direction);
116 #endif
117 #ifdef INET6
118 static struct encaptab *encap6_lookup(struct mbuf *, int, int, enum direction);
119 #endif
120 static int encap_add(struct encaptab *);
121 static int encap_remove(struct encaptab *);
122 static int encap_afcheck(int, const struct sockaddr *, const struct sockaddr *);
123 static struct radix_node_head *encap_rnh(int);
124 static int mask_matchlen(const struct sockaddr *);
125 static void encap_fillarg(struct mbuf *, const struct encaptab *);
126 
127 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
128 
129 extern int max_keylen;	/* radix.c */
130 struct radix_node_head *encap_head[2];	/* 0 for AF_INET, 1 for AF_INET6 */
131 
132 void
133 encap_init(void)
134 {
135 	static int initialized = 0;
136 
137 	if (initialized)
138 		return;
139 	initialized++;
140 #if 0
141 	/*
142 	 * we cannot use LIST_INIT() here, since drivers may want to call
143 	 * encap_attach(), on driver attach.  encap_init() will be called
144 	 * on AF_INET{,6} initialization, which happens after driver
145 	 * initialization - using LIST_INIT() here can nuke encap_attach()
146 	 * from drivers.
147 	 */
148 	LIST_INIT(&encaptab);
149 #endif
150 
151 	/*
152 	 * initialize radix lookup table when the radix subsystem is inited.
153 	 */
154 	rn_delayedinit((void *)&encap_head[0],
155 	    sizeof(struct sockaddr_pack) << 3);
156 #ifdef INET6
157 	rn_delayedinit((void *)&encap_head[1],
158 	    sizeof(struct sockaddr_pack) << 3);
159 #endif
160 }
161 
162 #ifdef INET
163 static struct encaptab *
164 encap4_lookup(struct mbuf *m, int off, int proto, enum direction dir)
165 {
166 	struct ip *ip;
167 	struct ip_pack4 pack;
168 	struct encaptab *ep, *match;
169 	int prio, matchprio;
170 	struct radix_node_head *rnh = encap_rnh(AF_INET);
171 	struct radix_node *rn;
172 
173 	KASSERT(m->m_len >= sizeof(*ip));
174 
175 	ip = mtod(m, struct ip *);
176 
177 	memset(&pack, 0, sizeof(pack));
178 	pack.p.sp_len = sizeof(pack);
179 	pack.mine.sin_family = pack.yours.sin_family = AF_INET;
180 	pack.mine.sin_len = pack.yours.sin_len = sizeof(struct sockaddr_in);
181 	if (dir == INBOUND) {
182 		pack.mine.sin_addr = ip->ip_dst;
183 		pack.yours.sin_addr = ip->ip_src;
184 	} else {
185 		pack.mine.sin_addr = ip->ip_src;
186 		pack.yours.sin_addr = ip->ip_dst;
187 	}
188 
189 	match = NULL;
190 	matchprio = 0;
191 
192 	rn = rnh->rnh_matchaddr((void *)&pack, rnh);
193 	if (rn && (rn->rn_flags & RNF_ROOT) == 0) {
194 		match = (struct encaptab *)rn;
195 		matchprio = mask_matchlen(match->srcmask) +
196 		    mask_matchlen(match->dstmask);
197 	}
198 
199 	LIST_FOREACH(ep, &encaptab, chain) {
200 		if (ep->af != AF_INET)
201 			continue;
202 		if (ep->proto >= 0 && ep->proto != proto)
203 			continue;
204 		if (ep->func)
205 			prio = (*ep->func)(m, off, proto, ep->arg);
206 		else
207 			continue;
208 
209 		/*
210 		 * We prioritize the matches by using bit length of the
211 		 * matches.  mask_match() and user-supplied matching function
212 		 * should return the bit length of the matches (for example,
213 		 * if both src/dst are matched for IPv4, 64 should be returned).
214 		 * 0 or negative return value means "it did not match".
215 		 *
216 		 * The question is, since we have two "mask" portion, we
217 		 * cannot really define total order between entries.
218 		 * For example, which of these should be preferred?
219 		 * mask_match() returns 48 (32 + 16) for both of them.
220 		 *	src=3ffe::/16, dst=3ffe:501::/32
221 		 *	src=3ffe:501::/32, dst=3ffe::/16
222 		 *
223 		 * We need to loop through all the possible candidates
224 		 * to get the best match - the search takes O(n) for
225 		 * n attachments (i.e. interfaces).
226 		 *
227 		 * For radix-based lookup, I guess source takes precedence.
228 		 * See rn_{refines,lexobetter} for the correct answer.
229 		 */
230 		if (prio <= 0)
231 			continue;
232 		if (prio > matchprio) {
233 			matchprio = prio;
234 			match = ep;
235 		}
236 	}
237 
238 	return match;
239 }
240 
241 void
242 encap4_input(struct mbuf *m, ...)
243 {
244 	int off, proto;
245 	va_list ap;
246 	const struct protosw *psw;
247 	struct encaptab *match;
248 
249 	va_start(ap, m);
250 	off = va_arg(ap, int);
251 	proto = va_arg(ap, int);
252 	va_end(ap);
253 
254 	match = encap4_lookup(m, off, proto, INBOUND);
255 
256 	if (match) {
257 		/* found a match, "match" has the best one */
258 		psw = match->psw;
259 		if (psw && psw->pr_input) {
260 			encap_fillarg(m, match);
261 			(*psw->pr_input)(m, off, proto);
262 		} else
263 			m_freem(m);
264 		return;
265 	}
266 
267 	/* last resort: inject to raw socket */
268 	rip_input(m, off, proto);
269 }
270 #endif
271 
272 #ifdef INET6
273 static struct encaptab *
274 encap6_lookup(struct mbuf *m, int off, int proto, enum direction dir)
275 {
276 	struct ip6_hdr *ip6;
277 	struct ip_pack6 pack;
278 	int prio, matchprio;
279 	struct encaptab *ep, *match;
280 	struct radix_node_head *rnh = encap_rnh(AF_INET6);
281 	struct radix_node *rn;
282 
283 	KASSERT(m->m_len >= sizeof(*ip6));
284 
285 	ip6 = mtod(m, struct ip6_hdr *);
286 
287 	memset(&pack, 0, sizeof(pack));
288 	pack.p.sp_len = sizeof(pack);
289 	pack.mine.sin6_family = pack.yours.sin6_family = AF_INET6;
290 	pack.mine.sin6_len = pack.yours.sin6_len = sizeof(struct sockaddr_in6);
291 	if (dir == INBOUND) {
292 		pack.mine.sin6_addr = ip6->ip6_dst;
293 		pack.yours.sin6_addr = ip6->ip6_src;
294 	} else {
295 		pack.mine.sin6_addr = ip6->ip6_src;
296 		pack.yours.sin6_addr = ip6->ip6_dst;
297 	}
298 
299 	match = NULL;
300 	matchprio = 0;
301 
302 	rn = rnh->rnh_matchaddr((void *)&pack, rnh);
303 	if (rn && (rn->rn_flags & RNF_ROOT) == 0) {
304 		match = (struct encaptab *)rn;
305 		matchprio = mask_matchlen(match->srcmask) +
306 		    mask_matchlen(match->dstmask);
307 	}
308 
309 	LIST_FOREACH(ep, &encaptab, chain) {
310 		if (ep->af != AF_INET6)
311 			continue;
312 		if (ep->proto >= 0 && ep->proto != proto)
313 			continue;
314 		if (ep->func)
315 			prio = (*ep->func)(m, off, proto, ep->arg);
316 		else
317 			continue;
318 
319 		/* see encap4_lookup() for issues here */
320 		if (prio <= 0)
321 			continue;
322 		if (prio > matchprio) {
323 			matchprio = prio;
324 			match = ep;
325 		}
326 	}
327 
328 	return match;
329 }
330 
331 int
332 encap6_input(struct mbuf **mp, int *offp, int proto)
333 {
334 	struct mbuf *m = *mp;
335 	const struct ip6protosw *psw;
336 	struct encaptab *match;
337 
338 	match = encap6_lookup(m, *offp, proto, INBOUND);
339 
340 	if (match) {
341 		/* found a match */
342 		psw = (const struct ip6protosw *)match->psw;
343 		if (psw && psw->pr_input) {
344 			encap_fillarg(m, match);
345 			return (*psw->pr_input)(mp, offp, proto);
346 		} else {
347 			m_freem(m);
348 			return IPPROTO_DONE;
349 		}
350 	}
351 
352 	/* last resort: inject to raw socket */
353 	return rip6_input(mp, offp, proto);
354 }
355 #endif
356 
357 static int
358 encap_add(struct encaptab *ep)
359 {
360 	struct radix_node_head *rnh = encap_rnh(ep->af);
361 	int error = 0;
362 
363 	LIST_INSERT_HEAD(&encaptab, ep, chain);
364 	if (!ep->func && rnh) {
365 		if (!rnh->rnh_addaddr((void *)ep->addrpack,
366 		    (void *)ep->maskpack, rnh, ep->nodes)) {
367 			error = EEXIST;
368 			goto fail;
369 		}
370 	}
371 	return error;
372 
373  fail:
374 	LIST_REMOVE(ep, chain);
375 	return error;
376 }
377 
378 static int
379 encap_remove(struct encaptab *ep)
380 {
381 	struct radix_node_head *rnh = encap_rnh(ep->af);
382 	int error = 0;
383 
384 	LIST_REMOVE(ep, chain);
385 	if (!ep->func && rnh) {
386 		if (!rnh->rnh_deladdr((void *)ep->addrpack,
387 		    (void *)ep->maskpack, rnh))
388 			error = ESRCH;
389 	}
390 	return error;
391 }
392 
393 static int
394 encap_afcheck(int af, const struct sockaddr *sp, const struct sockaddr *dp)
395 {
396 	if (sp && dp) {
397 		if (sp->sa_len != dp->sa_len)
398 			return EINVAL;
399 		if (af != sp->sa_family || af != dp->sa_family)
400 			return EINVAL;
401 	} else if (!sp && !dp)
402 		;
403 	else
404 		return EINVAL;
405 
406 	switch (af) {
407 	case AF_INET:
408 		if (sp && sp->sa_len != sizeof(struct sockaddr_in))
409 			return EINVAL;
410 		if (dp && dp->sa_len != sizeof(struct sockaddr_in))
411 			return EINVAL;
412 		break;
413 #ifdef INET6
414 	case AF_INET6:
415 		if (sp && sp->sa_len != sizeof(struct sockaddr_in6))
416 			return EINVAL;
417 		if (dp && dp->sa_len != sizeof(struct sockaddr_in6))
418 			return EINVAL;
419 		break;
420 #endif
421 	default:
422 		return EAFNOSUPPORT;
423 	}
424 
425 	return 0;
426 }
427 
428 /*
429  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
430  * length of mask (sm and dm) is assumed to be same as sp/dp.
431  * Return value will be necessary as input (cookie) for encap_detach().
432  */
433 const struct encaptab *
434 encap_attach(int af, int proto,
435     const struct sockaddr *sp, const struct sockaddr *sm,
436     const struct sockaddr *dp, const struct sockaddr *dm,
437     const struct protosw *psw, void *arg)
438 {
439 	struct encaptab *ep;
440 	int error;
441 	int s;
442 	size_t l;
443 	struct ip_pack4 *pack4;
444 #ifdef INET6
445 	struct ip_pack6 *pack6;
446 #endif
447 
448 	s = splsoftnet();
449 	/* sanity check on args */
450 	error = encap_afcheck(af, sp, dp);
451 	if (error)
452 		goto fail;
453 
454 	/* check if anyone have already attached with exactly same config */
455 	LIST_FOREACH(ep, &encaptab, chain) {
456 		if (ep->af != af)
457 			continue;
458 		if (ep->proto != proto)
459 			continue;
460 		if (ep->func)
461 			continue;
462 
463 		KASSERT(ep->src != NULL);
464 		KASSERT(ep->dst != NULL);
465 		KASSERT(ep->srcmask != NULL);
466 		KASSERT(ep->dstmask != NULL);
467 
468 		if (ep->src->sa_len != sp->sa_len ||
469 		    memcmp(ep->src, sp, sp->sa_len) != 0 ||
470 		    memcmp(ep->srcmask, sm, sp->sa_len) != 0)
471 			continue;
472 		if (ep->dst->sa_len != dp->sa_len ||
473 		    memcmp(ep->dst, dp, dp->sa_len) != 0 ||
474 		    memcmp(ep->dstmask, dm, dp->sa_len) != 0)
475 			continue;
476 
477 		error = EEXIST;
478 		goto fail;
479 	}
480 
481 	switch (af) {
482 	case AF_INET:
483 		l = sizeof(*pack4);
484 		break;
485 #ifdef INET6
486 	case AF_INET6:
487 		l = sizeof(*pack6);
488 		break;
489 #endif
490 	default:
491 		goto fail;
492 	}
493 
494 	/* M_NETADDR ok? */
495 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT|M_ZERO);
496 	if (ep == NULL) {
497 		error = ENOBUFS;
498 		goto fail;
499 	}
500 	ep->addrpack = malloc(l, M_NETADDR, M_NOWAIT|M_ZERO);
501 	if (ep->addrpack == NULL) {
502 		error = ENOBUFS;
503 		goto gc;
504 	}
505 	ep->maskpack = malloc(l, M_NETADDR, M_NOWAIT|M_ZERO);
506 	if (ep->maskpack == NULL) {
507 		error = ENOBUFS;
508 		goto gc;
509 	}
510 
511 	ep->af = af;
512 	ep->proto = proto;
513 	ep->addrpack->sa_len = l & 0xff;
514 	ep->maskpack->sa_len = l & 0xff;
515 	switch (af) {
516 	case AF_INET:
517 		pack4 = (struct ip_pack4 *)ep->addrpack;
518 		ep->src = (struct sockaddr *)&pack4->mine;
519 		ep->dst = (struct sockaddr *)&pack4->yours;
520 		pack4 = (struct ip_pack4 *)ep->maskpack;
521 		ep->srcmask = (struct sockaddr *)&pack4->mine;
522 		ep->dstmask = (struct sockaddr *)&pack4->yours;
523 		break;
524 #ifdef INET6
525 	case AF_INET6:
526 		pack6 = (struct ip_pack6 *)ep->addrpack;
527 		ep->src = (struct sockaddr *)&pack6->mine;
528 		ep->dst = (struct sockaddr *)&pack6->yours;
529 		pack6 = (struct ip_pack6 *)ep->maskpack;
530 		ep->srcmask = (struct sockaddr *)&pack6->mine;
531 		ep->dstmask = (struct sockaddr *)&pack6->yours;
532 		break;
533 #endif
534 	}
535 
536 	memcpy(ep->src, sp, sp->sa_len);
537 	memcpy(ep->srcmask, sm, sp->sa_len);
538 	memcpy(ep->dst, dp, dp->sa_len);
539 	memcpy(ep->dstmask, dm, dp->sa_len);
540 	ep->psw = psw;
541 	ep->arg = arg;
542 
543 	error = encap_add(ep);
544 	if (error)
545 		goto gc;
546 
547 	error = 0;
548 	splx(s);
549 	return ep;
550 
551 gc:
552 	if (ep->addrpack)
553 		free(ep->addrpack, M_NETADDR);
554 	if (ep->maskpack)
555 		free(ep->maskpack, M_NETADDR);
556 	if (ep)
557 		free(ep, M_NETADDR);
558 fail:
559 	splx(s);
560 	return NULL;
561 }
562 
563 const struct encaptab *
564 encap_attach_func(int af, int proto,
565     int (*func)(struct mbuf *, int, int, void *),
566     const struct protosw *psw, void *arg)
567 {
568 	struct encaptab *ep;
569 	int error;
570 	int s;
571 
572 	s = splsoftnet();
573 	/* sanity check on args */
574 	if (!func) {
575 		error = EINVAL;
576 		goto fail;
577 	}
578 
579 	error = encap_afcheck(af, NULL, NULL);
580 	if (error)
581 		goto fail;
582 
583 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
584 	if (ep == NULL) {
585 		error = ENOBUFS;
586 		goto fail;
587 	}
588 	memset(ep, 0, sizeof(*ep));
589 
590 	ep->af = af;
591 	ep->proto = proto;
592 	ep->func = func;
593 	ep->psw = psw;
594 	ep->arg = arg;
595 
596 	error = encap_add(ep);
597 	if (error)
598 		goto fail;
599 
600 	error = 0;
601 	splx(s);
602 	return ep;
603 
604 fail:
605 	splx(s);
606 	return NULL;
607 }
608 
609 /* XXX encap4_ctlinput() is necessary if we set DF=1 on outer IPv4 header */
610 
611 #ifdef INET6
612 void *
613 encap6_ctlinput(int cmd, const struct sockaddr *sa, void *d0)
614 {
615 	void *d = d0;
616 	struct ip6_hdr *ip6;
617 	struct mbuf *m;
618 	int off;
619 	struct ip6ctlparam *ip6cp = NULL;
620 	int nxt;
621 	struct encaptab *ep;
622 	const struct ip6protosw *psw;
623 
624 	if (sa->sa_family != AF_INET6 ||
625 	    sa->sa_len != sizeof(struct sockaddr_in6))
626 		return NULL;
627 
628 	if ((unsigned)cmd >= PRC_NCMDS)
629 		return NULL;
630 	if (cmd == PRC_HOSTDEAD)
631 		d = NULL;
632 	else if (cmd == PRC_MSGSIZE)
633 		; /* special code is present, see below */
634 	else if (inet6ctlerrmap[cmd] == 0)
635 		return NULL;
636 
637 	/* if the parameter is from icmp6, decode it. */
638 	if (d != NULL) {
639 		ip6cp = (struct ip6ctlparam *)d;
640 		m = ip6cp->ip6c_m;
641 		ip6 = ip6cp->ip6c_ip6;
642 		off = ip6cp->ip6c_off;
643 		nxt = ip6cp->ip6c_nxt;
644 
645 		if (ip6 && cmd == PRC_MSGSIZE) {
646 			int valid = 0;
647 			struct encaptab *match;
648 
649 			/*
650 		 	* Check to see if we have a valid encap configuration.
651 		 	*/
652 			match = encap6_lookup(m, off, nxt, OUTBOUND);
653 			if (match)
654 				valid++;
655 
656 			/*
657 		 	* Depending on the value of "valid" and routing table
658 		 	* size (mtudisc_{hi,lo}wat), we will:
659 		 	* - recalcurate the new MTU and create the
660 		 	*   corresponding routing entry, or
661 		 	* - ignore the MTU change notification.
662 		 	*/
663 			icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
664 		}
665 	} else {
666 		m = NULL;
667 		ip6 = NULL;
668 		nxt = -1;
669 	}
670 
671 	/* inform all listeners */
672 	LIST_FOREACH(ep, &encaptab, chain) {
673 		if (ep->af != AF_INET6)
674 			continue;
675 		if (ep->proto >= 0 && ep->proto != nxt)
676 			continue;
677 
678 		/* should optimize by looking at address pairs */
679 
680 		/* XXX need to pass ep->arg or ep itself to listeners */
681 		psw = (const struct ip6protosw *)ep->psw;
682 		if (psw && psw->pr_ctlinput)
683 			(*psw->pr_ctlinput)(cmd, sa, d);
684 	}
685 
686 	rip6_ctlinput(cmd, sa, d0);
687 	return NULL;
688 }
689 #endif
690 
691 int
692 encap_detach(const struct encaptab *cookie)
693 {
694 	const struct encaptab *ep = cookie;
695 	struct encaptab *p, *np;
696 	int error;
697 
698 	LIST_FOREACH_SAFE(p, &encaptab, chain, np) {
699 		if (p == ep) {
700 			error = encap_remove(p);
701 			if (error)
702 				return error;
703 			if (!ep->func) {
704 				free(p->addrpack, M_NETADDR);
705 				free(p->maskpack, M_NETADDR);
706 			}
707 			free(p, M_NETADDR);	/*XXX*/
708 			return 0;
709 		}
710 	}
711 
712 	return ENOENT;
713 }
714 
715 static struct radix_node_head *
716 encap_rnh(int af)
717 {
718 
719 	switch (af) {
720 	case AF_INET:
721 		return encap_head[0];
722 #ifdef INET6
723 	case AF_INET6:
724 		return encap_head[1];
725 #endif
726 	default:
727 		return NULL;
728 	}
729 }
730 
731 static int
732 mask_matchlen(const struct sockaddr *sa)
733 {
734 	const char *p, *ep;
735 	int l;
736 
737 	p = (const char *)sa;
738 	ep = p + sa->sa_len;
739 	p += 2;	/* sa_len + sa_family */
740 
741 	l = 0;
742 	while (p < ep) {
743 		l += (*p ? 8 : 0);	/* estimate */
744 		p++;
745 	}
746 	return l;
747 }
748 
749 static void
750 encap_fillarg(struct mbuf *m, const struct encaptab *ep)
751 {
752 	struct m_tag *mtag;
753 
754 	mtag = m_tag_get(PACKET_TAG_ENCAP, sizeof(void *), M_NOWAIT);
755 	if (mtag) {
756 		*(void **)(mtag + 1) = ep->arg;
757 		m_tag_prepend(m, mtag);
758 	}
759 }
760 
761 void *
762 encap_getarg(struct mbuf *m)
763 {
764 	void *p;
765 	struct m_tag *mtag;
766 
767 	p = NULL;
768 	mtag = m_tag_find(m, PACKET_TAG_ENCAP, NULL);
769 	if (mtag != NULL) {
770 		p = *(void **)(mtag + 1);
771 		m_tag_delete(m, mtag);
772 	}
773 	return p;
774 }
775