Home | History | Annotate | Line # | Download | only in netinet
ip_encap.c revision 1.39
      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  * !!!!!!!
     71  * !!NOTE: dom_maxrtkey assumes USE_RADIX is defined.
     72  * !!!!!!!
     73  */
     74 #define USE_RADIX
     75 
     76 #include <sys/cdefs.h>
     77 __KERNEL_RCSID(0, "$NetBSD: ip_encap.c,v 1.39 2011/07/17 20:54:53 joerg Exp $");
     78 
     79 #include "opt_mrouting.h"
     80 #include "opt_inet.h"
     81 
     82 #include <sys/param.h>
     83 #include <sys/systm.h>
     84 #include <sys/socket.h>
     85 #include <sys/sockio.h>
     86 #include <sys/mbuf.h>
     87 #include <sys/errno.h>
     88 #include <sys/protosw.h>
     89 #include <sys/queue.h>
     90 
     91 #include <net/if.h>
     92 #include <net/route.h>
     93 
     94 #include <netinet/in.h>
     95 #include <netinet/in_systm.h>
     96 #include <netinet/ip.h>
     97 #include <netinet/ip_var.h>
     98 #include <netinet/ip_encap.h>
     99 #ifdef MROUTING
    100 #include <netinet/ip_mroute.h>
    101 #endif /* MROUTING */
    102 
    103 #ifdef INET6
    104 #include <netinet/ip6.h>
    105 #include <netinet6/ip6_var.h>
    106 #include <netinet6/ip6protosw.h>
    107 #include <netinet6/in6_var.h>
    108 #include <netinet6/in6_pcb.h>
    109 #include <netinet/icmp6.h>
    110 #endif
    111 
    112 #include <net/net_osdep.h>
    113 
    114 enum direction { INBOUND, OUTBOUND };
    115 
    116 #ifdef INET
    117 static struct encaptab *encap4_lookup(struct mbuf *, int, int, enum direction);
    118 #endif
    119 #ifdef INET6
    120 static struct encaptab *encap6_lookup(struct mbuf *, int, int, enum direction);
    121 #endif
    122 static int encap_add(struct encaptab *);
    123 static int encap_remove(struct encaptab *);
    124 static int encap_afcheck(int, const struct sockaddr *, const struct sockaddr *);
    125 #ifdef USE_RADIX
    126 static struct radix_node_head *encap_rnh(int);
    127 static int mask_matchlen(const struct sockaddr *);
    128 #endif
    129 #ifndef USE_RADIX
    130 static int mask_match(const struct encaptab *, const struct sockaddr *,
    131 		const struct sockaddr *);
    132 #endif
    133 static void encap_fillarg(struct mbuf *, const struct encaptab *);
    134 
    135 LIST_HEAD(, encaptab) encaptab = LIST_HEAD_INITIALIZER(&encaptab);
    136 
    137 #ifdef USE_RADIX
    138 extern int max_keylen;	/* radix.c */
    139 struct radix_node_head *encap_head[2];	/* 0 for AF_INET, 1 for AF_INET6 */
    140 #endif
    141 
    142 void
    143 encap_init(void)
    144 {
    145 	static int initialized = 0;
    146 
    147 	if (initialized)
    148 		return;
    149 	initialized++;
    150 #if 0
    151 	/*
    152 	 * we cannot use LIST_INIT() here, since drivers may want to call
    153 	 * encap_attach(), on driver attach.  encap_init() will be called
    154 	 * on AF_INET{,6} initialization, which happens after driver
    155 	 * initialization - using LIST_INIT() here can nuke encap_attach()
    156 	 * from drivers.
    157 	 */
    158 	LIST_INIT(&encaptab);
    159 #endif
    160 
    161 #ifdef USE_RADIX
    162 	/*
    163 	 * initialize radix lookup table when the radix subsystem is inited.
    164 	 */
    165 	rn_delayedinit((void *)&encap_head[0],
    166 	    sizeof(struct sockaddr_pack) << 3);
    167 #ifdef INET6
    168 	rn_delayedinit((void *)&encap_head[1],
    169 	    sizeof(struct sockaddr_pack) << 3);
    170 #endif
    171 #endif
    172 }
    173 
    174 #ifdef INET
    175 static struct encaptab *
    176 encap4_lookup(struct mbuf *m, int off, int proto, enum direction dir)
    177 {
    178 	struct ip *ip;
    179 	struct ip_pack4 pack;
    180 	struct encaptab *ep, *match;
    181 	int prio, matchprio;
    182 #ifdef USE_RADIX
    183 	struct radix_node_head *rnh = encap_rnh(AF_INET);
    184 	struct radix_node *rn;
    185 #endif
    186 
    187 #ifdef DIAGNOSTIC
    188 	if (m->m_len < sizeof(*ip))
    189 		panic("encap4_lookup");
    190 #endif
    191 	ip = mtod(m, struct ip *);
    192 
    193 	memset(&pack, 0, sizeof(pack));
    194 	pack.p.sp_len = sizeof(pack);
    195 	pack.mine.sin_family = pack.yours.sin_family = AF_INET;
    196 	pack.mine.sin_len = pack.yours.sin_len = sizeof(struct sockaddr_in);
    197 	if (dir == INBOUND) {
    198 		pack.mine.sin_addr = ip->ip_dst;
    199 		pack.yours.sin_addr = ip->ip_src;
    200 	} else {
    201 		pack.mine.sin_addr = ip->ip_src;
    202 		pack.yours.sin_addr = ip->ip_dst;
    203 	}
    204 
    205 	match = NULL;
    206 	matchprio = 0;
    207 
    208 #ifdef USE_RADIX
    209 	rn = rnh->rnh_matchaddr((void *)&pack, rnh);
    210 	if (rn && (rn->rn_flags & RNF_ROOT) == 0) {
    211 		match = (struct encaptab *)rn;
    212 		matchprio = mask_matchlen(match->srcmask) +
    213 		    mask_matchlen(match->dstmask);
    214 	}
    215 #endif
    216 
    217 	LIST_FOREACH(ep, &encaptab, chain) {
    218 		if (ep->af != AF_INET)
    219 			continue;
    220 		if (ep->proto >= 0 && ep->proto != proto)
    221 			continue;
    222 		if (ep->func)
    223 			prio = (*ep->func)(m, off, proto, ep->arg);
    224 		else {
    225 #ifdef USE_RADIX
    226 			continue;
    227 #else
    228 			prio = mask_match(ep, (struct sockaddr *)&pack.mine,
    229 			    (struct sockaddr *)&pack.yours);
    230 #endif
    231 		}
    232 
    233 		/*
    234 		 * We prioritize the matches by using bit length of the
    235 		 * matches.  mask_match() and user-supplied matching function
    236 		 * should return the bit length of the matches (for example,
    237 		 * if both src/dst are matched for IPv4, 64 should be returned).
    238 		 * 0 or negative return value means "it did not match".
    239 		 *
    240 		 * The question is, since we have two "mask" portion, we
    241 		 * cannot really define total order between entries.
    242 		 * For example, which of these should be preferred?
    243 		 * mask_match() returns 48 (32 + 16) for both of them.
    244 		 *	src=3ffe::/16, dst=3ffe:501::/32
    245 		 *	src=3ffe:501::/32, dst=3ffe::/16
    246 		 *
    247 		 * We need to loop through all the possible candidates
    248 		 * to get the best match - the search takes O(n) for
    249 		 * n attachments (i.e. interfaces).
    250 		 *
    251 		 * For radix-based lookup, I guess source takes precedence.
    252 		 * See rn_{refines,lexobetter} for the correct answer.
    253 		 */
    254 		if (prio <= 0)
    255 			continue;
    256 		if (prio > matchprio) {
    257 			matchprio = prio;
    258 			match = ep;
    259 		}
    260 	}
    261 
    262 	return match;
    263 #undef s
    264 #undef d
    265 }
    266 
    267 void
    268 encap4_input(struct mbuf *m, ...)
    269 {
    270 	int off, proto;
    271 	va_list ap;
    272 	const struct protosw *psw;
    273 	struct encaptab *match;
    274 
    275 	va_start(ap, m);
    276 	off = va_arg(ap, int);
    277 	proto = va_arg(ap, int);
    278 	va_end(ap);
    279 
    280 	match = encap4_lookup(m, off, proto, INBOUND);
    281 
    282 	if (match) {
    283 		/* found a match, "match" has the best one */
    284 		psw = match->psw;
    285 		if (psw && psw->pr_input) {
    286 			encap_fillarg(m, match);
    287 			(*psw->pr_input)(m, off, proto);
    288 		} else
    289 			m_freem(m);
    290 		return;
    291 	}
    292 
    293 	/* last resort: inject to raw socket */
    294 	rip_input(m, off, proto);
    295 }
    296 #endif
    297 
    298 #ifdef INET6
    299 static struct encaptab *
    300 encap6_lookup(struct mbuf *m, int off, int proto, enum direction dir)
    301 {
    302 	struct ip6_hdr *ip6;
    303 	struct ip_pack6 pack;
    304 	int prio, matchprio;
    305 	struct encaptab *ep, *match;
    306 #ifdef USE_RADIX
    307 	struct radix_node_head *rnh = encap_rnh(AF_INET6);
    308 	struct radix_node *rn;
    309 #endif
    310 
    311 #ifdef DIAGNOSTIC
    312 	if (m->m_len < sizeof(*ip6))
    313 		panic("encap6_lookup");
    314 #endif
    315 	ip6 = mtod(m, struct ip6_hdr *);
    316 
    317 	memset(&pack, 0, sizeof(pack));
    318 	pack.p.sp_len = sizeof(pack);
    319 	pack.mine.sin6_family = pack.yours.sin6_family = AF_INET6;
    320 	pack.mine.sin6_len = pack.yours.sin6_len = sizeof(struct sockaddr_in6);
    321 	if (dir == INBOUND) {
    322 		pack.mine.sin6_addr = ip6->ip6_dst;
    323 		pack.yours.sin6_addr = ip6->ip6_src;
    324 	} else {
    325 		pack.mine.sin6_addr = ip6->ip6_src;
    326 		pack.yours.sin6_addr = ip6->ip6_dst;
    327 	}
    328 
    329 	match = NULL;
    330 	matchprio = 0;
    331 
    332 #ifdef USE_RADIX
    333 	rn = rnh->rnh_matchaddr((void *)&pack, rnh);
    334 	if (rn && (rn->rn_flags & RNF_ROOT) == 0) {
    335 		match = (struct encaptab *)rn;
    336 		matchprio = mask_matchlen(match->srcmask) +
    337 		    mask_matchlen(match->dstmask);
    338 	}
    339 #endif
    340 
    341 	LIST_FOREACH(ep, &encaptab, chain) {
    342 		if (ep->af != AF_INET6)
    343 			continue;
    344 		if (ep->proto >= 0 && ep->proto != proto)
    345 			continue;
    346 		if (ep->func)
    347 			prio = (*ep->func)(m, off, proto, ep->arg);
    348 		else {
    349 #ifdef USE_RADIX
    350 			continue;
    351 #else
    352 			prio = mask_match(ep, (struct sockaddr *)&pack.mine,
    353 			    (struct sockaddr *)&pack.yours);
    354 #endif
    355 		}
    356 
    357 		/* see encap4_lookup() for issues here */
    358 		if (prio <= 0)
    359 			continue;
    360 		if (prio > matchprio) {
    361 			matchprio = prio;
    362 			match = ep;
    363 		}
    364 	}
    365 
    366 	return match;
    367 #undef s
    368 #undef d
    369 }
    370 
    371 int
    372 encap6_input(struct mbuf **mp, int *offp, int proto)
    373 {
    374 	struct mbuf *m = *mp;
    375 	const struct ip6protosw *psw;
    376 	struct encaptab *match;
    377 
    378 	match = encap6_lookup(m, *offp, proto, INBOUND);
    379 
    380 	if (match) {
    381 		/* found a match */
    382 		psw = (const struct ip6protosw *)match->psw;
    383 		if (psw && psw->pr_input) {
    384 			encap_fillarg(m, match);
    385 			return (*psw->pr_input)(mp, offp, proto);
    386 		} else {
    387 			m_freem(m);
    388 			return IPPROTO_DONE;
    389 		}
    390 	}
    391 
    392 	/* last resort: inject to raw socket */
    393 	return rip6_input(mp, offp, proto);
    394 }
    395 #endif
    396 
    397 static int
    398 encap_add(struct encaptab *ep)
    399 {
    400 #ifdef USE_RADIX
    401 	struct radix_node_head *rnh = encap_rnh(ep->af);
    402 #endif
    403 	int error = 0;
    404 
    405 	LIST_INSERT_HEAD(&encaptab, ep, chain);
    406 #ifdef USE_RADIX
    407 	if (!ep->func && rnh) {
    408 		if (!rnh->rnh_addaddr((void *)ep->addrpack,
    409 		    (void *)ep->maskpack, rnh, ep->nodes)) {
    410 			error = EEXIST;
    411 			goto fail;
    412 		}
    413 	}
    414 #endif
    415 	return error;
    416 
    417  fail:
    418 	LIST_REMOVE(ep, chain);
    419 	return error;
    420 }
    421 
    422 static int
    423 encap_remove(struct encaptab *ep)
    424 {
    425 #ifdef USE_RADIX
    426 	struct radix_node_head *rnh = encap_rnh(ep->af);
    427 #endif
    428 	int error = 0;
    429 
    430 	LIST_REMOVE(ep, chain);
    431 #ifdef USE_RADIX
    432 	if (!ep->func && rnh) {
    433 		if (!rnh->rnh_deladdr((void *)ep->addrpack,
    434 		    (void *)ep->maskpack, rnh))
    435 			error = ESRCH;
    436 	}
    437 #endif
    438 	return error;
    439 }
    440 
    441 static int
    442 encap_afcheck(int af, const struct sockaddr *sp, const struct sockaddr *dp)
    443 {
    444 	if (sp && dp) {
    445 		if (sp->sa_len != dp->sa_len)
    446 			return EINVAL;
    447 		if (af != sp->sa_family || af != dp->sa_family)
    448 			return EINVAL;
    449 	} else if (!sp && !dp)
    450 		;
    451 	else
    452 		return EINVAL;
    453 
    454 	switch (af) {
    455 	case AF_INET:
    456 		if (sp && sp->sa_len != sizeof(struct sockaddr_in))
    457 			return EINVAL;
    458 		if (dp && dp->sa_len != sizeof(struct sockaddr_in))
    459 			return EINVAL;
    460 		break;
    461 #ifdef INET6
    462 	case AF_INET6:
    463 		if (sp && sp->sa_len != sizeof(struct sockaddr_in6))
    464 			return EINVAL;
    465 		if (dp && dp->sa_len != sizeof(struct sockaddr_in6))
    466 			return EINVAL;
    467 		break;
    468 #endif
    469 	default:
    470 		return EAFNOSUPPORT;
    471 	}
    472 
    473 	return 0;
    474 }
    475 
    476 /*
    477  * sp (src ptr) is always my side, and dp (dst ptr) is always remote side.
    478  * length of mask (sm and dm) is assumed to be same as sp/dp.
    479  * Return value will be necessary as input (cookie) for encap_detach().
    480  */
    481 const struct encaptab *
    482 encap_attach(int af, int proto,
    483     const struct sockaddr *sp, const struct sockaddr *sm,
    484     const struct sockaddr *dp, const struct sockaddr *dm,
    485     const struct protosw *psw, void *arg)
    486 {
    487 	struct encaptab *ep;
    488 	int error;
    489 	int s;
    490 	size_t l;
    491 	struct ip_pack4 *pack4;
    492 #ifdef INET6
    493 	struct ip_pack6 *pack6;
    494 #endif
    495 
    496 	s = splsoftnet();
    497 	/* sanity check on args */
    498 	error = encap_afcheck(af, sp, dp);
    499 	if (error)
    500 		goto fail;
    501 
    502 	/* check if anyone have already attached with exactly same config */
    503 	LIST_FOREACH(ep, &encaptab, chain) {
    504 		if (ep->af != af)
    505 			continue;
    506 		if (ep->proto != proto)
    507 			continue;
    508 		if (ep->func)
    509 			continue;
    510 #ifdef DIAGNOSTIC
    511 		if (!ep->src || !ep->dst || !ep->srcmask || !ep->dstmask)
    512 			panic("null pointers in encaptab");
    513 #endif
    514 		if (ep->src->sa_len != sp->sa_len ||
    515 		    memcmp(ep->src, sp, sp->sa_len) != 0 ||
    516 		    memcmp(ep->srcmask, sm, sp->sa_len) != 0)
    517 			continue;
    518 		if (ep->dst->sa_len != dp->sa_len ||
    519 		    memcmp(ep->dst, dp, dp->sa_len) != 0 ||
    520 		    memcmp(ep->dstmask, dm, dp->sa_len) != 0)
    521 			continue;
    522 
    523 		error = EEXIST;
    524 		goto fail;
    525 	}
    526 
    527 	switch (af) {
    528 	case AF_INET:
    529 		l = sizeof(*pack4);
    530 		break;
    531 #ifdef INET6
    532 	case AF_INET6:
    533 		l = sizeof(*pack6);
    534 		break;
    535 #endif
    536 	default:
    537 		goto fail;
    538 	}
    539 
    540 	/* M_NETADDR ok? */
    541 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT|M_ZERO);
    542 	if (ep == NULL) {
    543 		error = ENOBUFS;
    544 		goto fail;
    545 	}
    546 	ep->addrpack = malloc(l, M_NETADDR, M_NOWAIT|M_ZERO);
    547 	if (ep->addrpack == NULL) {
    548 		error = ENOBUFS;
    549 		goto gc;
    550 	}
    551 	ep->maskpack = malloc(l, M_NETADDR, M_NOWAIT|M_ZERO);
    552 	if (ep->maskpack == NULL) {
    553 		error = ENOBUFS;
    554 		goto gc;
    555 	}
    556 
    557 	ep->af = af;
    558 	ep->proto = proto;
    559 	ep->addrpack->sa_len = l & 0xff;
    560 	ep->maskpack->sa_len = l & 0xff;
    561 	switch (af) {
    562 	case AF_INET:
    563 		pack4 = (struct ip_pack4 *)ep->addrpack;
    564 		ep->src = (struct sockaddr *)&pack4->mine;
    565 		ep->dst = (struct sockaddr *)&pack4->yours;
    566 		pack4 = (struct ip_pack4 *)ep->maskpack;
    567 		ep->srcmask = (struct sockaddr *)&pack4->mine;
    568 		ep->dstmask = (struct sockaddr *)&pack4->yours;
    569 		break;
    570 #ifdef INET6
    571 	case AF_INET6:
    572 		pack6 = (struct ip_pack6 *)ep->addrpack;
    573 		ep->src = (struct sockaddr *)&pack6->mine;
    574 		ep->dst = (struct sockaddr *)&pack6->yours;
    575 		pack6 = (struct ip_pack6 *)ep->maskpack;
    576 		ep->srcmask = (struct sockaddr *)&pack6->mine;
    577 		ep->dstmask = (struct sockaddr *)&pack6->yours;
    578 		break;
    579 #endif
    580 	}
    581 
    582 	memcpy(ep->src, sp, sp->sa_len);
    583 	memcpy(ep->srcmask, sm, sp->sa_len);
    584 	memcpy(ep->dst, dp, dp->sa_len);
    585 	memcpy(ep->dstmask, dm, dp->sa_len);
    586 	ep->psw = psw;
    587 	ep->arg = arg;
    588 
    589 	error = encap_add(ep);
    590 	if (error)
    591 		goto gc;
    592 
    593 	error = 0;
    594 	splx(s);
    595 	return ep;
    596 
    597 gc:
    598 	if (ep->addrpack)
    599 		free(ep->addrpack, M_NETADDR);
    600 	if (ep->maskpack)
    601 		free(ep->maskpack, M_NETADDR);
    602 	if (ep)
    603 		free(ep, M_NETADDR);
    604 fail:
    605 	splx(s);
    606 	return NULL;
    607 }
    608 
    609 const struct encaptab *
    610 encap_attach_func(int af, int proto,
    611     int (*func)(struct mbuf *, int, int, void *),
    612     const struct protosw *psw, void *arg)
    613 {
    614 	struct encaptab *ep;
    615 	int error;
    616 	int s;
    617 
    618 	s = splsoftnet();
    619 	/* sanity check on args */
    620 	if (!func) {
    621 		error = EINVAL;
    622 		goto fail;
    623 	}
    624 
    625 	error = encap_afcheck(af, NULL, NULL);
    626 	if (error)
    627 		goto fail;
    628 
    629 	ep = malloc(sizeof(*ep), M_NETADDR, M_NOWAIT);	/*XXX*/
    630 	if (ep == NULL) {
    631 		error = ENOBUFS;
    632 		goto fail;
    633 	}
    634 	memset(ep, 0, sizeof(*ep));
    635 
    636 	ep->af = af;
    637 	ep->proto = proto;
    638 	ep->func = func;
    639 	ep->psw = psw;
    640 	ep->arg = arg;
    641 
    642 	error = encap_add(ep);
    643 	if (error)
    644 		goto fail;
    645 
    646 	error = 0;
    647 	splx(s);
    648 	return ep;
    649 
    650 fail:
    651 	splx(s);
    652 	return NULL;
    653 }
    654 
    655 /* XXX encap4_ctlinput() is necessary if we set DF=1 on outer IPv4 header */
    656 
    657 #ifdef INET6
    658 void *
    659 encap6_ctlinput(int cmd, const struct sockaddr *sa, void *d0)
    660 {
    661 	void *d = d0;
    662 	struct ip6_hdr *ip6;
    663 	struct mbuf *m;
    664 	int off;
    665 	struct ip6ctlparam *ip6cp = NULL;
    666 	int nxt;
    667 	struct encaptab *ep;
    668 	const struct ip6protosw *psw;
    669 
    670 	if (sa->sa_family != AF_INET6 ||
    671 	    sa->sa_len != sizeof(struct sockaddr_in6))
    672 		return NULL;
    673 
    674 	if ((unsigned)cmd >= PRC_NCMDS)
    675 		return NULL;
    676 	if (cmd == PRC_HOSTDEAD)
    677 		d = NULL;
    678 	else if (cmd == PRC_MSGSIZE)
    679 		; /* special code is present, see below */
    680 	else if (inet6ctlerrmap[cmd] == 0)
    681 		return NULL;
    682 
    683 	/* if the parameter is from icmp6, decode it. */
    684 	if (d != NULL) {
    685 		ip6cp = (struct ip6ctlparam *)d;
    686 		m = ip6cp->ip6c_m;
    687 		ip6 = ip6cp->ip6c_ip6;
    688 		off = ip6cp->ip6c_off;
    689 		nxt = ip6cp->ip6c_nxt;
    690 
    691 		if (ip6 && cmd == PRC_MSGSIZE) {
    692 			int valid = 0;
    693 			struct encaptab *match;
    694 
    695 			/*
    696 		 	* Check to see if we have a valid encap configuration.
    697 		 	*/
    698 			match = encap6_lookup(m, off, nxt, OUTBOUND);
    699 			if (match)
    700 				valid++;
    701 
    702 			/*
    703 		 	* Depending on the value of "valid" and routing table
    704 		 	* size (mtudisc_{hi,lo}wat), we will:
    705 		 	* - recalcurate the new MTU and create the
    706 		 	*   corresponding routing entry, or
    707 		 	* - ignore the MTU change notification.
    708 		 	*/
    709 			icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
    710 		}
    711 	} else {
    712 		m = NULL;
    713 		ip6 = NULL;
    714 		nxt = -1;
    715 	}
    716 
    717 	/* inform all listeners */
    718 	LIST_FOREACH(ep, &encaptab, chain) {
    719 		if (ep->af != AF_INET6)
    720 			continue;
    721 		if (ep->proto >= 0 && ep->proto != nxt)
    722 			continue;
    723 
    724 		/* should optimize by looking at address pairs */
    725 
    726 		/* XXX need to pass ep->arg or ep itself to listeners */
    727 		psw = (const struct ip6protosw *)ep->psw;
    728 		if (psw && psw->pr_ctlinput)
    729 			(*psw->pr_ctlinput)(cmd, sa, d);
    730 	}
    731 
    732 	rip6_ctlinput(cmd, sa, d0);
    733 	return NULL;
    734 }
    735 #endif
    736 
    737 int
    738 encap_detach(const struct encaptab *cookie)
    739 {
    740 	const struct encaptab *ep = cookie;
    741 	struct encaptab *p;
    742 	int error;
    743 
    744 	LIST_FOREACH(p, &encaptab, chain) {
    745 		if (p == ep) {
    746 			error = encap_remove(p);
    747 			if (error)
    748 				return error;
    749 			if (!ep->func) {
    750 				free(p->addrpack, M_NETADDR);
    751 				free(p->maskpack, M_NETADDR);
    752 			}
    753 			free(p, M_NETADDR);	/*XXX*/
    754 			return 0;
    755 		}
    756 	}
    757 
    758 	return ENOENT;
    759 }
    760 
    761 #ifdef USE_RADIX
    762 static struct radix_node_head *
    763 encap_rnh(int af)
    764 {
    765 
    766 	switch (af) {
    767 	case AF_INET:
    768 		return encap_head[0];
    769 #ifdef INET6
    770 	case AF_INET6:
    771 		return encap_head[1];
    772 #endif
    773 	default:
    774 		return NULL;
    775 	}
    776 }
    777 
    778 static int
    779 mask_matchlen(const struct sockaddr *sa)
    780 {
    781 	const char *p, *ep;
    782 	int l;
    783 
    784 	p = (const char *)sa;
    785 	ep = p + sa->sa_len;
    786 	p += 2;	/* sa_len + sa_family */
    787 
    788 	l = 0;
    789 	while (p < ep) {
    790 		l += (*p ? 8 : 0);	/* estimate */
    791 		p++;
    792 	}
    793 	return l;
    794 }
    795 #endif
    796 
    797 #ifndef USE_RADIX
    798 static int
    799 mask_match(const struct encaptab *ep,
    800 	   const struct sockaddr *sp,
    801 	   const struct sockaddr *dp)
    802 {
    803 	struct sockaddr_storage s;
    804 	struct sockaddr_storage d;
    805 	int i;
    806 	const u_int8_t *p, *q;
    807 	u_int8_t *r;
    808 	int matchlen;
    809 
    810 #ifdef DIAGNOSTIC
    811 	if (ep->func)
    812 		panic("wrong encaptab passed to mask_match");
    813 #endif
    814 	if (sp->sa_len > sizeof(s) || dp->sa_len > sizeof(d))
    815 		return 0;
    816 	if (sp->sa_family != ep->af || dp->sa_family != ep->af)
    817 		return 0;
    818 	if (sp->sa_len != ep->src->sa_len || dp->sa_len != ep->dst->sa_len)
    819 		return 0;
    820 
    821 	matchlen = 0;
    822 
    823 	p = (const u_int8_t *)sp;
    824 	q = (const u_int8_t *)ep->srcmask;
    825 	r = (u_int8_t *)&s;
    826 	for (i = 0 ; i < sp->sa_len; i++) {
    827 		r[i] = p[i] & q[i];
    828 		/* XXX estimate */
    829 		matchlen += (q[i] ? 8 : 0);
    830 	}
    831 
    832 	p = (const u_int8_t *)dp;
    833 	q = (const u_int8_t *)ep->dstmask;
    834 	r = (u_int8_t *)&d;
    835 	for (i = 0 ; i < dp->sa_len; i++) {
    836 		r[i] = p[i] & q[i];
    837 		/* XXX rough estimate */
    838 		matchlen += (q[i] ? 8 : 0);
    839 	}
    840 
    841 	/* need to overwrite len/family portion as we don't compare them */
    842 	s.ss_len = sp->sa_len;
    843 	s.ss_family = sp->sa_family;
    844 	d.ss_len = dp->sa_len;
    845 	d.ss_family = dp->sa_family;
    846 
    847 	if (memcmp(&s, ep->src, ep->src->sa_len) == 0 &&
    848 	    memcmp(&d, ep->dst, ep->dst->sa_len) == 0) {
    849 		return matchlen;
    850 	} else
    851 		return 0;
    852 }
    853 #endif
    854 
    855 static void
    856 encap_fillarg(struct mbuf *m, const struct encaptab *ep)
    857 {
    858 	struct m_tag *mtag;
    859 
    860 	mtag = m_tag_get(PACKET_TAG_ENCAP, sizeof(void *), M_NOWAIT);
    861 	if (mtag) {
    862 		*(void **)(mtag + 1) = ep->arg;
    863 		m_tag_prepend(m, mtag);
    864 	}
    865 }
    866 
    867 void *
    868 encap_getarg(struct mbuf *m)
    869 {
    870 	void *p;
    871 	struct m_tag *mtag;
    872 
    873 	p = NULL;
    874 	mtag = m_tag_find(m, PACKET_TAG_ENCAP, NULL);
    875 	if (mtag != NULL) {
    876 		p = *(void **)(mtag + 1);
    877 		m_tag_delete(m, mtag);
    878 	}
    879 	return p;
    880 }
    881