Home | History | Annotate | Line # | Download | only in netinet
ip_encap.c revision 1.52
      1 /*	$NetBSD: ip_encap.c,v 1.52 2016/02/26 07:35:17 knakahara 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 
     69 #include <sys/cdefs.h>
     70 __KERNEL_RCSID(0, "$NetBSD: ip_encap.c,v 1.52 2016/02/26 07:35:17 knakahara Exp $");
     71 
     72 #ifdef _KERNEL_OPT
     73 #include "opt_mrouting.h"
     74 #include "opt_inet.h"
     75 #endif
     76 
     77 #include <sys/param.h>
     78 #include <sys/systm.h>
     79 #include <sys/socket.h>
     80 #include <sys/sockio.h>
     81 #include <sys/mbuf.h>
     82 #include <sys/errno.h>
     83 #include <sys/queue.h>
     84 #include <sys/kmem.h>
     85 
     86 #include <net/if.h>
     87 #include <net/route.h>
     88 
     89 #include <netinet/in.h>
     90 #include <netinet/in_systm.h>
     91 #include <netinet/ip.h>
     92 #include <netinet/ip_var.h>
     93 #include <netinet/ip_encap.h>
     94 #ifdef MROUTING
     95 #include <netinet/ip_mroute.h>
     96 #endif /* MROUTING */
     97 
     98 #ifdef INET6
     99 #include <netinet/ip6.h>
    100 #include <netinet6/ip6_var.h>
    101 #include <netinet6/ip6protosw.h> /* for struct ip6ctlparam */
    102 #include <netinet6/in6_var.h>
    103 #include <netinet6/in6_pcb.h>
    104 #include <netinet/icmp6.h>
    105 #endif
    106 
    107 #include <net/net_osdep.h>
    108 
    109 enum direction { INBOUND, OUTBOUND };
    110 
    111 #ifdef INET
    112 static struct encaptab *encap4_lookup(struct mbuf *, int, int, enum direction);
    113 #endif
    114 #ifdef INET6
    115 static struct encaptab *encap6_lookup(struct mbuf *, int, int, enum direction);
    116 #endif
    117 static int encap_add(struct encaptab *);
    118 static int encap_remove(struct encaptab *);
    119 static int encap_afcheck(int, const struct sockaddr *, const struct sockaddr *);
    120 static struct radix_node_head *encap_rnh(int);
    121 static int mask_matchlen(const struct sockaddr *);
    122 static void encap_fillarg(struct mbuf *, const struct encaptab *);
    123 
    124 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
    125 
    126 struct radix_node_head *encap_head[2];	/* 0 for AF_INET, 1 for AF_INET6 */
    127 
    128 void
    129 encap_init(void)
    130 {
    131 	static int initialized = 0;
    132 
    133 	if (initialized)
    134 		return;
    135 	initialized++;
    136 #if 0
    137 	/*
    138 	 * we cannot use LIST_INIT() here, since drivers may want to call
    139 	 * encap_attach(), on driver attach.  encap_init() will be called
    140 	 * on AF_INET{,6} initialization, which happens after driver
    141 	 * initialization - using LIST_INIT() here can nuke encap_attach()
    142 	 * from drivers.
    143 	 */
    144 	LIST_INIT(&encaptab);
    145 #endif
    146 
    147 	/*
    148 	 * initialize radix lookup table when the radix subsystem is inited.
    149 	 */
    150 	rn_delayedinit((void *)&encap_head[0],
    151 	    sizeof(struct sockaddr_pack) << 3);
    152 #ifdef INET6
    153 	rn_delayedinit((void *)&encap_head[1],
    154 	    sizeof(struct sockaddr_pack) << 3);
    155 #endif
    156 }
    157 
    158 #ifdef INET
    159 static struct encaptab *
    160 encap4_lookup(struct mbuf *m, int off, int proto, enum direction dir)
    161 {
    162 	struct ip *ip;
    163 	struct ip_pack4 pack;
    164 	struct encaptab *ep, *match;
    165 	int prio, matchprio;
    166 	struct radix_node_head *rnh = encap_rnh(AF_INET);
    167 	struct radix_node *rn;
    168 
    169 	KASSERT(m->m_len >= sizeof(*ip));
    170 
    171 	ip = mtod(m, struct ip *);
    172 
    173 	memset(&pack, 0, sizeof(pack));
    174 	pack.p.sp_len = sizeof(pack);
    175 	pack.mine.sin_family = pack.yours.sin_family = AF_INET;
    176 	pack.mine.sin_len = pack.yours.sin_len = sizeof(struct sockaddr_in);
    177 	if (dir == INBOUND) {
    178 		pack.mine.sin_addr = ip->ip_dst;
    179 		pack.yours.sin_addr = ip->ip_src;
    180 	} else {
    181 		pack.mine.sin_addr = ip->ip_src;
    182 		pack.yours.sin_addr = ip->ip_dst;
    183 	}
    184 
    185 	match = NULL;
    186 	matchprio = 0;
    187 
    188 	rn = rnh->rnh_matchaddr((void *)&pack, rnh);
    189 	if (rn && (rn->rn_flags & RNF_ROOT) == 0) {
    190 		match = (struct encaptab *)rn;
    191 		matchprio = mask_matchlen(match->srcmask) +
    192 		    mask_matchlen(match->dstmask);
    193 	}
    194 
    195 	LIST_FOREACH(ep, &encaptab, chain) {
    196 		if (ep->af != AF_INET)
    197 			continue;
    198 		if (ep->proto >= 0 && ep->proto != proto)
    199 			continue;
    200 		if (ep->func)
    201 			prio = (*ep->func)(m, off, proto, ep->arg);
    202 		else
    203 			continue;
    204 
    205 		/*
    206 		 * We prioritize the matches by using bit length of the
    207 		 * matches.  mask_match() and user-supplied matching function
    208 		 * should return the bit length of the matches (for example,
    209 		 * if both src/dst are matched for IPv4, 64 should be returned).
    210 		 * 0 or negative return value means "it did not match".
    211 		 *
    212 		 * The question is, since we have two "mask" portion, we
    213 		 * cannot really define total order between entries.
    214 		 * For example, which of these should be preferred?
    215 		 * mask_match() returns 48 (32 + 16) for both of them.
    216 		 *	src=3ffe::/16, dst=3ffe:501::/32
    217 		 *	src=3ffe:501::/32, dst=3ffe::/16
    218 		 *
    219 		 * We need to loop through all the possible candidates
    220 		 * to get the best match - the search takes O(n) for
    221 		 * n attachments (i.e. interfaces).
    222 		 *
    223 		 * For radix-based lookup, I guess source takes precedence.
    224 		 * See rn_{refines,lexobetter} for the correct answer.
    225 		 */
    226 		if (prio <= 0)
    227 			continue;
    228 		if (prio > matchprio) {
    229 			matchprio = prio;
    230 			match = ep;
    231 		}
    232 	}
    233 
    234 	return match;
    235 }
    236 
    237 void
    238 encap4_input(struct mbuf *m, ...)
    239 {
    240 	int off, proto;
    241 	va_list ap;
    242 	const struct encapsw *esw;
    243 	struct encaptab *match;
    244 
    245 	va_start(ap, m);
    246 	off = va_arg(ap, int);
    247 	proto = va_arg(ap, int);
    248 	va_end(ap);
    249 
    250 	match = encap4_lookup(m, off, proto, INBOUND);
    251 
    252 	if (match) {
    253 		/* found a match, "match" has the best one */
    254 		esw = match->esw;
    255 		if (esw && esw->encapsw4.pr_input) {
    256 			encap_fillarg(m, match);
    257 			(*esw->encapsw4.pr_input)(m, off, proto);
    258 		} else
    259 			m_freem(m);
    260 		return;
    261 	}
    262 
    263 	/* last resort: inject to raw socket */
    264 	rip_input(m, off, proto);
    265 }
    266 #endif
    267 
    268 #ifdef INET6
    269 static struct encaptab *
    270 encap6_lookup(struct mbuf *m, int off, int proto, enum direction dir)
    271 {
    272 	struct ip6_hdr *ip6;
    273 	struct ip_pack6 pack;
    274 	int prio, matchprio;
    275 	struct encaptab *ep, *match;
    276 	struct radix_node_head *rnh = encap_rnh(AF_INET6);
    277 	struct radix_node *rn;
    278 
    279 	KASSERT(m->m_len >= sizeof(*ip6));
    280 
    281 	ip6 = mtod(m, struct ip6_hdr *);
    282 
    283 	memset(&pack, 0, sizeof(pack));
    284 	pack.p.sp_len = sizeof(pack);
    285 	pack.mine.sin6_family = pack.yours.sin6_family = AF_INET6;
    286 	pack.mine.sin6_len = pack.yours.sin6_len = sizeof(struct sockaddr_in6);
    287 	if (dir == INBOUND) {
    288 		pack.mine.sin6_addr = ip6->ip6_dst;
    289 		pack.yours.sin6_addr = ip6->ip6_src;
    290 	} else {
    291 		pack.mine.sin6_addr = ip6->ip6_src;
    292 		pack.yours.sin6_addr = ip6->ip6_dst;
    293 	}
    294 
    295 	match = NULL;
    296 	matchprio = 0;
    297 
    298 	rn = rnh->rnh_matchaddr((void *)&pack, rnh);
    299 	if (rn && (rn->rn_flags & RNF_ROOT) == 0) {
    300 		match = (struct encaptab *)rn;
    301 		matchprio = mask_matchlen(match->srcmask) +
    302 		    mask_matchlen(match->dstmask);
    303 	}
    304 
    305 	LIST_FOREACH(ep, &encaptab, chain) {
    306 		if (ep->af != AF_INET6)
    307 			continue;
    308 		if (ep->proto >= 0 && ep->proto != proto)
    309 			continue;
    310 		if (ep->func)
    311 			prio = (*ep->func)(m, off, proto, ep->arg);
    312 		else
    313 			continue;
    314 
    315 		/* see encap4_lookup() for issues here */
    316 		if (prio <= 0)
    317 			continue;
    318 		if (prio > matchprio) {
    319 			matchprio = prio;
    320 			match = ep;
    321 		}
    322 	}
    323 
    324 	return match;
    325 }
    326 
    327 int
    328 encap6_input(struct mbuf **mp, int *offp, int proto)
    329 {
    330 	struct mbuf *m = *mp;
    331 	const struct encapsw *esw;
    332 	struct encaptab *match;
    333 
    334 	match = encap6_lookup(m, *offp, proto, INBOUND);
    335 
    336 	if (match) {
    337 		/* found a match */
    338 		esw = match->esw;
    339 		if (esw && esw->encapsw6.pr_input) {
    340 			encap_fillarg(m, match);
    341 			return (*esw->encapsw6.pr_input)(mp, offp, proto);
    342 		} else {
    343 			m_freem(m);
    344 			return IPPROTO_DONE;
    345 		}
    346 	}
    347 
    348 	/* last resort: inject to raw socket */
    349 	return rip6_input(mp, offp, proto);
    350 }
    351 #endif
    352 
    353 static int
    354 encap_add(struct encaptab *ep)
    355 {
    356 	struct radix_node_head *rnh = encap_rnh(ep->af);
    357 	int error = 0;
    358 
    359 	LIST_INSERT_HEAD(&encaptab, ep, chain);
    360 	if (!ep->func && rnh) {
    361 		if (!rnh->rnh_addaddr((void *)ep->addrpack,
    362 		    (void *)ep->maskpack, rnh, ep->nodes)) {
    363 			error = EEXIST;
    364 			goto fail;
    365 		}
    366 	}
    367 	return error;
    368 
    369  fail:
    370 	LIST_REMOVE(ep, chain);
    371 	return error;
    372 }
    373 
    374 static int
    375 encap_remove(struct encaptab *ep)
    376 {
    377 	struct radix_node_head *rnh = encap_rnh(ep->af);
    378 	int error = 0;
    379 
    380 	LIST_REMOVE(ep, chain);
    381 	if (!ep->func && rnh) {
    382 		if (!rnh->rnh_deladdr((void *)ep->addrpack,
    383 		    (void *)ep->maskpack, rnh))
    384 			error = ESRCH;
    385 	}
    386 	return error;
    387 }
    388 
    389 static int
    390 encap_afcheck(int af, const struct sockaddr *sp, const struct sockaddr *dp)
    391 {
    392 	if (sp && dp) {
    393 		if (sp->sa_len != dp->sa_len)
    394 			return EINVAL;
    395 		if (af != sp->sa_family || af != dp->sa_family)
    396 			return EINVAL;
    397 	} else if (!sp && !dp)
    398 		;
    399 	else
    400 		return EINVAL;
    401 
    402 	switch (af) {
    403 	case AF_INET:
    404 		if (sp && sp->sa_len != sizeof(struct sockaddr_in))
    405 			return EINVAL;
    406 		if (dp && dp->sa_len != sizeof(struct sockaddr_in))
    407 			return EINVAL;
    408 		break;
    409 #ifdef INET6
    410 	case AF_INET6:
    411 		if (sp && sp->sa_len != sizeof(struct sockaddr_in6))
    412 			return EINVAL;
    413 		if (dp && dp->sa_len != sizeof(struct sockaddr_in6))
    414 			return EINVAL;
    415 		break;
    416 #endif
    417 	default:
    418 		return EAFNOSUPPORT;
    419 	}
    420 
    421 	return 0;
    422 }
    423 
    424 /*
    425  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
    426  * length of mask (sm and dm) is assumed to be same as sp/dp.
    427  * Return value will be necessary as input (cookie) for encap_detach().
    428  */
    429 const struct encaptab *
    430 encap_attach(int af, int proto,
    431     const struct sockaddr *sp, const struct sockaddr *sm,
    432     const struct sockaddr *dp, const struct sockaddr *dm,
    433     const struct encapsw *esw, void *arg)
    434 {
    435 	struct encaptab *ep;
    436 	int error;
    437 	int s;
    438 	size_t l;
    439 	struct ip_pack4 *pack4;
    440 #ifdef INET6
    441 	struct ip_pack6 *pack6;
    442 #endif
    443 
    444 	s = splsoftnet();
    445 	/* sanity check on args */
    446 	error = encap_afcheck(af, sp, dp);
    447 	if (error)
    448 		goto fail;
    449 
    450 	/* check if anyone have already attached with exactly same config */
    451 	LIST_FOREACH(ep, &encaptab, chain) {
    452 		if (ep->af != af)
    453 			continue;
    454 		if (ep->proto != proto)
    455 			continue;
    456 		if (ep->func)
    457 			continue;
    458 
    459 		KASSERT(ep->src != NULL);
    460 		KASSERT(ep->dst != NULL);
    461 		KASSERT(ep->srcmask != NULL);
    462 		KASSERT(ep->dstmask != NULL);
    463 
    464 		if (ep->src->sa_len != sp->sa_len ||
    465 		    memcmp(ep->src, sp, sp->sa_len) != 0 ||
    466 		    memcmp(ep->srcmask, sm, sp->sa_len) != 0)
    467 			continue;
    468 		if (ep->dst->sa_len != dp->sa_len ||
    469 		    memcmp(ep->dst, dp, dp->sa_len) != 0 ||
    470 		    memcmp(ep->dstmask, dm, dp->sa_len) != 0)
    471 			continue;
    472 
    473 		error = EEXIST;
    474 		goto fail;
    475 	}
    476 
    477 	switch (af) {
    478 	case AF_INET:
    479 		l = sizeof(*pack4);
    480 		break;
    481 #ifdef INET6
    482 	case AF_INET6:
    483 		l = sizeof(*pack6);
    484 		break;
    485 #endif
    486 	default:
    487 		goto fail;
    488 	}
    489 
    490 	/* M_NETADDR ok? */
    491 	ep = kmem_zalloc(sizeof(*ep), KM_NOSLEEP);
    492 	if (ep == NULL) {
    493 		error = ENOBUFS;
    494 		goto fail;
    495 	}
    496 	ep->addrpack = kmem_zalloc(l, KM_NOSLEEP);
    497 	if (ep->addrpack == NULL) {
    498 		error = ENOBUFS;
    499 		goto gc;
    500 	}
    501 	ep->maskpack = kmem_zalloc(l, KM_NOSLEEP);
    502 	if (ep->maskpack == NULL) {
    503 		error = ENOBUFS;
    504 		goto gc;
    505 	}
    506 
    507 	ep->af = af;
    508 	ep->proto = proto;
    509 	ep->addrpack->sa_len = l & 0xff;
    510 	ep->maskpack->sa_len = l & 0xff;
    511 	switch (af) {
    512 	case AF_INET:
    513 		pack4 = (struct ip_pack4 *)ep->addrpack;
    514 		ep->src = (struct sockaddr *)&pack4->mine;
    515 		ep->dst = (struct sockaddr *)&pack4->yours;
    516 		pack4 = (struct ip_pack4 *)ep->maskpack;
    517 		ep->srcmask = (struct sockaddr *)&pack4->mine;
    518 		ep->dstmask = (struct sockaddr *)&pack4->yours;
    519 		break;
    520 #ifdef INET6
    521 	case AF_INET6:
    522 		pack6 = (struct ip_pack6 *)ep->addrpack;
    523 		ep->src = (struct sockaddr *)&pack6->mine;
    524 		ep->dst = (struct sockaddr *)&pack6->yours;
    525 		pack6 = (struct ip_pack6 *)ep->maskpack;
    526 		ep->srcmask = (struct sockaddr *)&pack6->mine;
    527 		ep->dstmask = (struct sockaddr *)&pack6->yours;
    528 		break;
    529 #endif
    530 	}
    531 
    532 	memcpy(ep->src, sp, sp->sa_len);
    533 	memcpy(ep->srcmask, sm, sp->sa_len);
    534 	memcpy(ep->dst, dp, dp->sa_len);
    535 	memcpy(ep->dstmask, dm, dp->sa_len);
    536 	ep->esw = esw;
    537 	ep->arg = arg;
    538 
    539 	error = encap_add(ep);
    540 	if (error)
    541 		goto gc;
    542 
    543 	error = 0;
    544 	splx(s);
    545 	return ep;
    546 
    547 gc:
    548 	if (ep->addrpack)
    549 		kmem_free(ep->addrpack, l);
    550 	if (ep->maskpack)
    551 		kmem_free(ep->maskpack, l);
    552 	if (ep)
    553 		kmem_free(ep, sizeof(*ep));
    554 fail:
    555 	splx(s);
    556 	return NULL;
    557 }
    558 
    559 const struct encaptab *
    560 encap_attach_func(int af, int proto,
    561     int (*func)(struct mbuf *, int, int, void *),
    562     const struct encapsw *esw, void *arg)
    563 {
    564 	struct encaptab *ep;
    565 	int error;
    566 	int s;
    567 
    568 	s = splsoftnet();
    569 	/* sanity check on args */
    570 	if (!func) {
    571 		error = EINVAL;
    572 		goto fail;
    573 	}
    574 
    575 	error = encap_afcheck(af, NULL, NULL);
    576 	if (error)
    577 		goto fail;
    578 
    579 	ep = kmem_alloc(sizeof(*ep), KM_NOSLEEP);	/*XXX*/
    580 	if (ep == NULL) {
    581 		error = ENOBUFS;
    582 		goto fail;
    583 	}
    584 	memset(ep, 0, sizeof(*ep));
    585 
    586 	ep->af = af;
    587 	ep->proto = proto;
    588 	ep->func = func;
    589 	ep->esw = esw;
    590 	ep->arg = arg;
    591 
    592 	error = encap_add(ep);
    593 	if (error)
    594 		goto fail;
    595 
    596 	error = 0;
    597 	splx(s);
    598 	return ep;
    599 
    600 fail:
    601 	splx(s);
    602 	return NULL;
    603 }
    604 
    605 /* XXX encap4_ctlinput() is necessary if we set DF=1 on outer IPv4 header */
    606 
    607 #ifdef INET6
    608 void *
    609 encap6_ctlinput(int cmd, const struct sockaddr *sa, void *d0)
    610 {
    611 	void *d = d0;
    612 	struct ip6_hdr *ip6;
    613 	struct mbuf *m;
    614 	int off;
    615 	struct ip6ctlparam *ip6cp = NULL;
    616 	int nxt;
    617 	struct encaptab *ep;
    618 	const struct encapsw *esw;
    619 
    620 	if (sa->sa_family != AF_INET6 ||
    621 	    sa->sa_len != sizeof(struct sockaddr_in6))
    622 		return NULL;
    623 
    624 	if ((unsigned)cmd >= PRC_NCMDS)
    625 		return NULL;
    626 	if (cmd == PRC_HOSTDEAD)
    627 		d = NULL;
    628 	else if (cmd == PRC_MSGSIZE)
    629 		; /* special code is present, see below */
    630 	else if (inet6ctlerrmap[cmd] == 0)
    631 		return NULL;
    632 
    633 	/* if the parameter is from icmp6, decode it. */
    634 	if (d != NULL) {
    635 		ip6cp = (struct ip6ctlparam *)d;
    636 		m = ip6cp->ip6c_m;
    637 		ip6 = ip6cp->ip6c_ip6;
    638 		off = ip6cp->ip6c_off;
    639 		nxt = ip6cp->ip6c_nxt;
    640 
    641 		if (ip6 && cmd == PRC_MSGSIZE) {
    642 			int valid = 0;
    643 			struct encaptab *match;
    644 
    645 			/*
    646 		 	* Check to see if we have a valid encap configuration.
    647 		 	*/
    648 			match = encap6_lookup(m, off, nxt, OUTBOUND);
    649 			if (match)
    650 				valid++;
    651 
    652 			/*
    653 		 	* Depending on the value of "valid" and routing table
    654 		 	* size (mtudisc_{hi,lo}wat), we will:
    655 		 	* - recalcurate the new MTU and create the
    656 		 	*   corresponding routing entry, or
    657 		 	* - ignore the MTU change notification.
    658 		 	*/
    659 			icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
    660 		}
    661 	} else {
    662 		m = NULL;
    663 		ip6 = NULL;
    664 		nxt = -1;
    665 	}
    666 
    667 	/* inform all listeners */
    668 	LIST_FOREACH(ep, &encaptab, chain) {
    669 		if (ep->af != AF_INET6)
    670 			continue;
    671 		if (ep->proto >= 0 && ep->proto != nxt)
    672 			continue;
    673 
    674 		/* should optimize by looking at address pairs */
    675 
    676 		/* XXX need to pass ep->arg or ep itself to listeners */
    677 		esw = ep->esw;
    678 		if (esw && esw->encapsw6.pr_ctlinput) {
    679 			(*esw->encapsw6.pr_ctlinput)(cmd, sa, d, ep->arg);
    680 		}
    681 	}
    682 
    683 	rip6_ctlinput(cmd, sa, d0);
    684 	return NULL;
    685 }
    686 #endif
    687 
    688 int
    689 encap_detach(const struct encaptab *cookie)
    690 {
    691 	const struct encaptab *ep = cookie;
    692 	struct encaptab *p, *np;
    693 	int error;
    694 
    695 	LIST_FOREACH_SAFE(p, &encaptab, chain, np) {
    696 		if (p == ep) {
    697 			error = encap_remove(p);
    698 			if (error)
    699 				return error;
    700 			if (!ep->func) {
    701 				kmem_free(p->addrpack, ep->addrpack->sa_len);
    702 				kmem_free(p->maskpack, ep->maskpack->sa_len);
    703 			}
    704 			kmem_free(p, sizeof(*p));	/*XXX*/
    705 			return 0;
    706 		}
    707 	}
    708 
    709 	return ENOENT;
    710 }
    711 
    712 static struct radix_node_head *
    713 encap_rnh(int af)
    714 {
    715 
    716 	switch (af) {
    717 	case AF_INET:
    718 		return encap_head[0];
    719 #ifdef INET6
    720 	case AF_INET6:
    721 		return encap_head[1];
    722 #endif
    723 	default:
    724 		return NULL;
    725 	}
    726 }
    727 
    728 static int
    729 mask_matchlen(const struct sockaddr *sa)
    730 {
    731 	const char *p, *ep;
    732 	int l;
    733 
    734 	p = (const char *)sa;
    735 	ep = p + sa->sa_len;
    736 	p += 2;	/* sa_len + sa_family */
    737 
    738 	l = 0;
    739 	while (p < ep) {
    740 		l += (*p ? 8 : 0);	/* estimate */
    741 		p++;
    742 	}
    743 	return l;
    744 }
    745 
    746 static void
    747 encap_fillarg(struct mbuf *m, const struct encaptab *ep)
    748 {
    749 	struct m_tag *mtag;
    750 
    751 	mtag = m_tag_get(PACKET_TAG_ENCAP, sizeof(void *), M_NOWAIT);
    752 	if (mtag) {
    753 		*(void **)(mtag + 1) = ep->arg;
    754 		m_tag_prepend(m, mtag);
    755 	}
    756 }
    757 
    758 void *
    759 encap_getarg(struct mbuf *m)
    760 {
    761 	void *p;
    762 	struct m_tag *mtag;
    763 
    764 	p = NULL;
    765 	mtag = m_tag_find(m, PACKET_TAG_ENCAP, NULL);
    766 	if (mtag != NULL) {
    767 		p = *(void **)(mtag + 1);
    768 		m_tag_delete(m, mtag);
    769 	}
    770 	return p;
    771 }
    772