Home | History | Annotate | Line # | Download | only in netipsec
ipsecif.c revision 1.10.2.2
      1 /*	$NetBSD: ipsecif.c,v 1.10.2.2 2020/04/08 14:08:58 martin Exp $  */
      2 
      3 /*
      4  * Copyright (c) 2017 Internet Initiative Japan Inc.
      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  *
     16  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     17  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     18  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     19  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     20  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     21  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     22  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     23  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     24  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     25  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     26  * POSSIBILITY OF SUCH DAMAGE.
     27  */
     28 
     29 #include <sys/cdefs.h>
     30 __KERNEL_RCSID(0, "$NetBSD: ipsecif.c,v 1.10.2.2 2020/04/08 14:08:58 martin Exp $");
     31 
     32 #ifdef _KERNEL_OPT
     33 #include "opt_inet.h"
     34 #include "opt_ipsec.h"
     35 #endif
     36 
     37 #include <sys/param.h>
     38 #include <sys/systm.h>
     39 #include <sys/socket.h>
     40 #include <sys/sockio.h>
     41 #include <sys/mbuf.h>
     42 #include <sys/errno.h>
     43 #include <sys/ioctl.h>
     44 #include <sys/syslog.h>
     45 #include <sys/kernel.h>
     46 
     47 #include <net/if.h>
     48 #include <net/route.h>
     49 
     50 #include <netinet/in.h>
     51 #include <netinet/in_systm.h>
     52 #include <netinet/ip.h>
     53 #include <netinet/ip_var.h>
     54 #include <netinet/in_var.h>
     55 #include <netinet/ip_encap.h>
     56 #include <netinet/ip_ecn.h>
     57 #include <netinet/ip_private.h>
     58 #include <netinet/udp.h>
     59 
     60 #ifdef INET6
     61 #include <netinet/ip6.h>
     62 #include <netinet6/ip6_var.h>
     63 #include <netinet6/ip6_private.h>
     64 #include <netinet6/in6_var.h>
     65 #include <netinet6/ip6protosw.h> /* for struct ip6ctlparam */
     66 #include <netinet/ip_ecn.h>
     67 #endif
     68 
     69 #include <netipsec/key.h>
     70 #include <netipsec/ipsecif.h>
     71 
     72 #include <net/if_ipsec.h>
     73 
     74 static int ipsecif_set_natt_ports(struct ipsec_variant *, struct mbuf *);
     75 static void ipsecif4_input(struct mbuf *, int, int, void *);
     76 static int ipsecif4_output(struct ipsec_variant *, int, struct mbuf *);
     77 static int ipsecif4_filter4(const struct ip *, struct ipsec_variant *,
     78 	struct ifnet *);
     79 
     80 #ifdef INET6
     81 static int ipsecif6_input(struct mbuf **, int *, int, void *);
     82 static int ipsecif6_output(struct ipsec_variant *, int, struct mbuf *);
     83 static int ipsecif6_filter6(const struct ip6_hdr *, struct ipsec_variant *,
     84 	struct ifnet *);
     85 #endif
     86 
     87 static int ip_ipsec_ttl = IPSEC_TTL;
     88 static int ip_ipsec_copy_tos = 0;
     89 #ifdef INET6
     90 static int ip6_ipsec_hlim = IPSEC_HLIM;
     91 static int ip6_ipsec_pmtu = 0; /* XXX: per interface configuration?? */
     92 static int ip6_ipsec_copy_tos = 0;
     93 #endif
     94 
     95 static const struct encapsw ipsecif4_encapsw = {
     96 	.encapsw4 = {
     97 		.pr_input = ipsecif4_input,
     98 		.pr_ctlinput = NULL,
     99 	}
    100 };
    101 
    102 #ifdef INET6
    103 static const struct encapsw ipsecif6_encapsw;
    104 #endif
    105 
    106 static int
    107 ipsecif_set_natt_ports(struct ipsec_variant *var, struct mbuf *m)
    108 {
    109 
    110 	KASSERT(if_ipsec_heldref_variant(var));
    111 
    112 	if (var->iv_sport || var->iv_dport) {
    113 		struct m_tag *mtag;
    114 
    115 		mtag = m_tag_get(PACKET_TAG_IPSEC_NAT_T_PORTS,
    116 		    sizeof(uint16_t) + sizeof(uint16_t), M_DONTWAIT);
    117 		if (mtag) {
    118 			uint16_t *natt_port;
    119 
    120 			natt_port = (uint16_t *)(mtag + 1);
    121 			natt_port[0] = var->iv_dport;
    122 			natt_port[1] = var->iv_sport;
    123 			m_tag_prepend(m, mtag);
    124 		} else {
    125 			return ENOBUFS;
    126 		}
    127 	}
    128 
    129 	return 0;
    130 }
    131 
    132 static struct mbuf *
    133 ipsecif4_prepend_hdr(struct ipsec_variant *var, struct mbuf *m,
    134     uint8_t proto, uint8_t tos)
    135 {
    136 	struct ip *ip;
    137 	struct sockaddr_in *src, *dst;
    138 
    139 	src = satosin(var->iv_psrc);
    140 	dst = satosin(var->iv_pdst);
    141 
    142 	if (in_nullhost(src->sin_addr) || in_nullhost(src->sin_addr) ||
    143 	    src->sin_addr.s_addr == INADDR_BROADCAST ||
    144 	    dst->sin_addr.s_addr == INADDR_BROADCAST) {
    145 		m_freem(m);
    146 		return NULL;
    147 	}
    148 	m->m_flags &= ~M_BCAST;
    149 
    150 	if (IN_MULTICAST(src->sin_addr.s_addr) ||
    151 	    IN_MULTICAST(dst->sin_addr.s_addr)) {
    152 		m_freem(m);
    153 		return NULL;
    154 	}
    155 
    156 	M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
    157 	if (m && M_UNWRITABLE(m, sizeof(struct ip)))
    158 		m = m_pullup(m, sizeof(struct ip));
    159 	if (m == NULL)
    160 		return NULL;
    161 
    162 	ip = mtod(m, struct ip *);
    163 	ip->ip_v = IPVERSION;
    164 	ip->ip_off = htons(0);
    165 	if (m->m_pkthdr.len < IP_MINFRAGSIZE)
    166 		ip->ip_id = 0;
    167 	else
    168 		ip->ip_id = ip_newid(NULL);
    169 	ip->ip_hl = sizeof(*ip) >> 2;
    170 	if (ip_ipsec_copy_tos)
    171 		ip->ip_tos = tos;
    172 	else
    173 		ip->ip_tos = 0;
    174 	ip->ip_sum = 0;
    175 	ip->ip_src = src->sin_addr;
    176 	ip->ip_dst = dst->sin_addr;
    177 	ip->ip_p = proto;
    178 	ip->ip_ttl = ip_ipsec_ttl;
    179 	ip->ip_len = htons(m->m_pkthdr.len);
    180 #ifndef IPSEC_TX_TOS_CLEAR
    181 	struct ifnet *ifp = &var->iv_softc->ipsec_if;
    182 	if (ifp->if_flags & IFF_ECN)
    183 		ip_ecn_ingress(ECN_ALLOWED, &ip->ip_tos, &tos);
    184 	else
    185 		ip_ecn_ingress(ECN_NOCARE, &ip->ip_tos, &tos);
    186 #endif
    187 
    188 	return m;
    189 }
    190 
    191 static int
    192 ipsecif4_needfrag(struct mbuf *m, struct ipsecrequest *isr)
    193 {
    194 	struct ip ip0;
    195 	struct ip *ip;
    196 	int mtu;
    197 	struct secasvar *sav;
    198 
    199 	sav = key_lookup_sa_bysaidx(&isr->saidx);
    200 	if (sav == NULL)
    201 		return 0;
    202 
    203 	if (!(sav->natt_type & UDP_ENCAP_ESPINUDP)) {
    204 		mtu = 0;
    205 		goto out;
    206 	}
    207 
    208 	if (m->m_len < sizeof(struct ip)) {
    209 		m_copydata(m, 0, sizeof(ip0), &ip0);
    210 		ip = &ip0;
    211 	} else {
    212 		ip = mtod(m, struct ip *);
    213 	}
    214 	mtu = sav->esp_frag;
    215 	if (ntohs(ip->ip_len) <= mtu)
    216 		mtu = 0;
    217 
    218 out:
    219 	KEY_SA_UNREF(&sav);
    220 	return mtu;
    221 }
    222 
    223 static struct mbuf *
    224 ipsecif4_flowinfo(struct mbuf *m, int family, int *proto0, u_int8_t *tos0)
    225 {
    226 	const struct ip *ip;
    227 	int proto;
    228 	int tos;
    229 
    230 	KASSERT(proto0 != NULL);
    231 	KASSERT(tos0 != NULL);
    232 
    233 	switch (family) {
    234 	case AF_INET:
    235 		proto = IPPROTO_IPV4;
    236 		if (m->m_len < sizeof(*ip)) {
    237 			m = m_pullup(m, sizeof(*ip));
    238 			if (m == NULL) {
    239 				*tos0 = 0;
    240 				*proto0 = 0;
    241 				return NULL;
    242 			}
    243 		}
    244 		ip = mtod(m, const struct ip *);
    245 		tos = ip->ip_tos;
    246 		/* TODO: support ALTQ for innner packet */
    247 		break;
    248 #ifdef INET6
    249 	case AF_INET6: {
    250 		const struct ip6_hdr *ip6;
    251 		proto = IPPROTO_IPV6;
    252 		if (m->m_len < sizeof(*ip6)) {
    253 			m = m_pullup(m, sizeof(*ip6));
    254 			if (m == NULL) {
    255 				*tos0 = 0;
    256 				*proto0 = 0;
    257 				return NULL;
    258 			}
    259 		}
    260 		ip6 = mtod(m, const struct ip6_hdr *);
    261 		tos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
    262 		/* TODO: support ALTQ for innner packet */
    263 		break;
    264 	}
    265 #endif /* INET6 */
    266 	default:
    267 		*tos0 = 0;
    268 		*proto0 = 0;
    269 		return NULL;
    270 	}
    271 
    272 	*proto0 = proto;
    273 	*tos0 = tos;
    274 	return m;
    275 }
    276 
    277 static int
    278 ipsecif4_fragout(struct ipsec_variant *var, int family, struct mbuf *m, int mtu)
    279 {
    280 	struct ifnet *ifp = &var->iv_softc->ipsec_if;
    281 	struct mbuf *next;
    282 	struct m_tag *mtag;
    283 	int error;
    284 
    285 	KASSERT(if_ipsec_heldref_variant(var));
    286 
    287 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_NAT_T_PORTS);
    288 	if (mtag)
    289 		m_tag_delete(m, mtag);
    290 
    291 	/* consider new IP header prepended in ipsecif4_output() */
    292 	if (mtu <= sizeof(struct ip)) {
    293 		m_freem(m);
    294 		return ENETUNREACH;
    295 	}
    296 	m->m_pkthdr.csum_flags |= M_CSUM_IPv4;
    297 	error = ip_fragment(m, ifp, mtu - sizeof(struct ip));
    298 	if (error)
    299 		return error;
    300 
    301 	for (error = 0; m; m = next) {
    302 		next = m->m_nextpkt;
    303 		m->m_nextpkt = NULL;
    304 		if (error) {
    305 			m_freem(m);
    306 			continue;
    307 		}
    308 
    309 		error = ipsecif4_output(var, family, m);
    310 	}
    311 	if (error == 0)
    312 		IP_STATINC(IP_STAT_FRAGMENTED);
    313 
    314 	return error;
    315 }
    316 
    317 int
    318 ipsecif4_encap_func(struct mbuf *m, struct ip *ip, struct ipsec_variant *var)
    319 {
    320 	struct m_tag *mtag;
    321 	struct sockaddr_in *src, *dst;
    322 	u_int16_t src_port = 0;
    323 	u_int16_t dst_port = 0;
    324 
    325 	KASSERT(var != NULL);
    326 
    327 	src = satosin(var->iv_psrc);
    328 	dst = satosin(var->iv_pdst);
    329 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_NAT_T_PORTS);
    330 	if (mtag) {
    331 		u_int16_t *ports;
    332 
    333 		ports = (u_int16_t *)(mtag + 1);
    334 		src_port = ports[0];
    335 		dst_port = ports[1];
    336 	}
    337 
    338 	/* address match */
    339 	if (src->sin_addr.s_addr != ip->ip_dst.s_addr ||
    340 	    dst->sin_addr.s_addr != ip->ip_src.s_addr)
    341 		return 0;
    342 
    343 	/* UDP encap? */
    344 	if (mtag == NULL && var->iv_sport == 0 && var->iv_dport == 0)
    345 		goto match;
    346 
    347 	/* port match */
    348 	if (src_port != var->iv_dport ||
    349 	    dst_port != var->iv_sport) {
    350 #ifdef DEBUG
    351 		printf("%s: port mismatch: pkt(%u, %u), if(%u, %u)\n",
    352 		    __func__, ntohs(src_port), ntohs(dst_port),
    353 		    ntohs(var->iv_sport), ntohs(var->iv_dport));
    354 #endif
    355 		return 0;
    356 	}
    357 
    358 match:
    359 	/*
    360 	 * hide NAT-T information from encapsulated traffics.
    361 	 * they don't know about IPsec.
    362 	 */
    363 	if (mtag)
    364 		m_tag_delete(m, mtag);
    365 	return sizeof(src->sin_addr) + sizeof(dst->sin_addr);
    366 }
    367 
    368 static int
    369 ipsecif4_output(struct ipsec_variant *var, int family, struct mbuf *m)
    370 {
    371 	struct secpolicy *sp = NULL;
    372 	u_int8_t tos;
    373 	int proto;
    374 	int error;
    375 	int mtu;
    376 	u_long sa_mtu = 0;
    377 
    378 	KASSERT(if_ipsec_heldref_variant(var));
    379 	KASSERT(if_ipsec_variant_is_configured(var));
    380 	KASSERT(var->iv_psrc->sa_family == AF_INET);
    381 	KASSERT(var->iv_pdst->sa_family == AF_INET);
    382 
    383 	switch (family) {
    384 	case AF_INET:
    385 		sp = IV_SP_OUT(var);
    386 		break;
    387 	case AF_INET6:
    388 		sp = IV_SP_OUT6(var);
    389 		break;
    390 	default:
    391 		m_freem(m);
    392 		return EAFNOSUPPORT;
    393 	}
    394 	KASSERT(sp != NULL);
    395 	/*
    396 	 * The SPs in ipsec_variant are prevented from freed by
    397 	 * ipsec_variant->iv_psref. So, KEY_SP_REF() is unnecessary here.
    398 	 */
    399 
    400 	KASSERT(sp->policy != IPSEC_POLICY_NONE);
    401 	KASSERT(sp->policy != IPSEC_POLICY_ENTRUST);
    402 	KASSERT(sp->policy != IPSEC_POLICY_BYPASS);
    403 	if (sp->policy != IPSEC_POLICY_IPSEC) {
    404 		m_freem(m);
    405 		error = ENETUNREACH;
    406 		goto done;
    407 	}
    408 
    409 	/* get flowinfo */
    410 	m = ipsecif4_flowinfo(m, family, &proto, &tos);
    411 	if (m == NULL) {
    412 		error = ENETUNREACH;
    413 		goto done;
    414 	}
    415 
    416 	/* prepend new IP header */
    417 	m = ipsecif4_prepend_hdr(var, m, proto, tos);
    418 	if (m == NULL) {
    419 		error = ENETUNREACH;
    420 		goto done;
    421 	}
    422 
    423 	/*
    424 	 * Normal netipsec's NAT-T fragmentation is done in ip_output().
    425 	 * See "natt_frag" processing.
    426 	 * However, ipsec(4) interface's one is not done in the same way,
    427 	 * so we must do NAT-T fragmentation by own code.
    428 	 */
    429 	/* NAT-T ESP fragmentation */
    430 	mtu = ipsecif4_needfrag(m, sp->req);
    431 	if (mtu > 0)
    432 		return ipsecif4_fragout(var, family, m, mtu);
    433 
    434 	/* set NAT-T ports */
    435 	error = ipsecif_set_natt_ports(var, m);
    436 	if (error) {
    437 		m_freem(m);
    438 		goto done;
    439 	}
    440 
    441 	/* IPsec output */
    442 	IP_STATINC(IP_STAT_LOCALOUT);
    443 	error = ipsec4_process_packet(m, sp->req, &sa_mtu);
    444 	if (error == ENOENT)
    445 		error = 0;
    446 	/*
    447 	 * frangmentation is already done in ipsecif4_fragout(),
    448 	 * so ipsec4_process_packet() must not do fragmentation here.
    449 	 */
    450 	KASSERT(sa_mtu == 0);
    451 
    452 done:
    453 	return error;
    454 }
    455 
    456 #ifdef INET6
    457 int
    458 ipsecif6_encap_func(struct mbuf *m, struct ip6_hdr *ip6, struct ipsec_variant *var)
    459 {
    460 	struct m_tag *mtag;
    461 	struct sockaddr_in6 *src, *dst;
    462 	u_int16_t src_port = 0;
    463 	u_int16_t dst_port = 0;
    464 
    465 	KASSERT(var != NULL);
    466 
    467 	src = satosin6(var->iv_psrc);
    468 	dst = satosin6(var->iv_pdst);
    469 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_NAT_T_PORTS);
    470 	if (mtag) {
    471 		u_int16_t *ports;
    472 
    473 		ports = (u_int16_t *)(mtag + 1);
    474 		src_port = ports[0];
    475 		dst_port = ports[1];
    476 	}
    477 
    478 	/* address match */
    479 	if (!IN6_ARE_ADDR_EQUAL(&src->sin6_addr, &ip6->ip6_dst) ||
    480 	    !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_src))
    481 		return 0;
    482 
    483 	/* UDP encap? */
    484 	if (mtag == NULL && var->iv_sport == 0 && var->iv_dport == 0)
    485 		goto match;
    486 
    487 	/* port match */
    488 	if (src_port != var->iv_dport ||
    489 	    dst_port != var->iv_sport) {
    490 #ifdef DEBUG
    491 		printf("%s: port mismatch: pkt(%u, %u), if(%u, %u)\n",
    492 		    __func__, ntohs(src_port), ntohs(dst_port),
    493 		    ntohs(var->iv_sport), ntohs(var->iv_dport));
    494 #endif
    495 		return 0;
    496 	}
    497 
    498 match:
    499 	/*
    500 	 * hide NAT-T information from encapsulated traffics.
    501 	 * they don't know about IPsec.
    502 	 */
    503 	if (mtag)
    504 		m_tag_delete(m, mtag);
    505 	return sizeof(src->sin6_addr) + sizeof(dst->sin6_addr);
    506 }
    507 
    508 static int
    509 ipsecif6_output(struct ipsec_variant *var, int family, struct mbuf *m)
    510 {
    511 	struct ifnet *ifp = &var->iv_softc->ipsec_if;
    512 	struct ipsec_softc *sc = ifp->if_softc;
    513 	struct ipsec_ro *iro;
    514 	struct rtentry *rt;
    515 	struct sockaddr_in6 *sin6_src;
    516 	struct sockaddr_in6 *sin6_dst;
    517 	struct ip6_hdr *ip6;
    518 	int proto, error;
    519 	u_int8_t itos, otos;
    520 	union {
    521 		struct sockaddr		dst;
    522 		struct sockaddr_in6	dst6;
    523 	} u;
    524 
    525 	KASSERT(if_ipsec_heldref_variant(var));
    526 	KASSERT(if_ipsec_variant_is_configured(var));
    527 
    528 	sin6_src = satosin6(var->iv_psrc);
    529 	sin6_dst = satosin6(var->iv_pdst);
    530 
    531 	KASSERT(sin6_src->sin6_family == AF_INET6);
    532 	KASSERT(sin6_dst->sin6_family == AF_INET6);
    533 
    534 	switch (family) {
    535 #ifdef INET
    536 	case AF_INET:
    537 	    {
    538 		struct ip *ip;
    539 
    540 		proto = IPPROTO_IPV4;
    541 		if (m->m_len < sizeof(*ip)) {
    542 			m = m_pullup(m, sizeof(*ip));
    543 			if (m == NULL)
    544 				return ENOBUFS;
    545 		}
    546 		ip = mtod(m, struct ip *);
    547 		itos = ip->ip_tos;
    548 		/* TODO: support ALTQ for innner packet */
    549 		break;
    550 	    }
    551 #endif /* INET */
    552 	case AF_INET6:
    553 	    {
    554 		struct ip6_hdr *xip6;
    555 		proto = IPPROTO_IPV6;
    556 		if (m->m_len < sizeof(*xip6)) {
    557 			m = m_pullup(m, sizeof(*xip6));
    558 			if (m == NULL)
    559 				return ENOBUFS;
    560 		}
    561 		xip6 = mtod(m, struct ip6_hdr *);
    562 		itos = (ntohl(xip6->ip6_flow) >> 20) & 0xff;
    563 		/* TODO: support ALTQ for innner packet */
    564 		break;
    565 	    }
    566 	default:
    567 		m_freem(m);
    568 		return EAFNOSUPPORT;
    569 	}
    570 
    571 	/* prepend new IP header */
    572 	M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
    573 	if (m && M_UNWRITABLE(m, sizeof(struct ip6_hdr)))
    574 		m = m_pullup(m, sizeof(struct ip6_hdr));
    575 	if (m == NULL)
    576 		return ENOBUFS;
    577 
    578 	ip6 = mtod(m, struct ip6_hdr *);
    579 	ip6->ip6_flow	= 0;
    580 	ip6->ip6_vfc	&= ~IPV6_VERSION_MASK;
    581 	ip6->ip6_vfc	|= IPV6_VERSION;
    582 #if 0	/* ip6->ip6_plen will be filled by ip6_output */
    583 	ip6->ip6_plen	= htons((u_short)m->m_pkthdr.len - sizeof(*ip6));
    584 #endif
    585 	ip6->ip6_nxt	= proto;
    586 	ip6->ip6_hlim	= ip6_ipsec_hlim;
    587 	ip6->ip6_src	= sin6_src->sin6_addr;
    588 	/* bidirectional configured tunnel mode */
    589 	if (!IN6_IS_ADDR_UNSPECIFIED(&sin6_dst->sin6_addr)) {
    590 		ip6->ip6_dst = sin6_dst->sin6_addr;
    591 	} else  {
    592 		m_freem(m);
    593 		return ENETUNREACH;
    594 	}
    595 #ifndef IPSEC_TX_TOS_CLEAR
    596 	if (!ip6_ipsec_copy_tos)
    597 		otos = 0;
    598 
    599 	if (ifp->if_flags & IFF_ECN)
    600 		ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
    601 	else
    602 		ip_ecn_ingress(ECN_NOCARE, &otos, &itos);
    603 #else
    604 	if (ip6_ipsec_copy_tos)
    605 		otos = itos;
    606 	else
    607 		otos = 0;
    608 #endif
    609 	ip6->ip6_flow &= ~ntohl(0xff00000);
    610 	ip6->ip6_flow |= htonl((u_int32_t)otos << 20);
    611 
    612 	sockaddr_in6_init(&u.dst6, &sin6_dst->sin6_addr, 0, 0, 0);
    613 
    614 	iro = percpu_getref(sc->ipsec_ro_percpu);
    615 	mutex_enter(iro->ir_lock);
    616 	if ((rt = rtcache_lookup(&iro->ir_ro, &u.dst)) == NULL) {
    617 		mutex_exit(iro->ir_lock);
    618 		percpu_putref(sc->ipsec_ro_percpu);
    619 		m_freem(m);
    620 		return ENETUNREACH;
    621 	}
    622 
    623 	if (rt->rt_ifp == ifp) {
    624 		rtcache_unref(rt, &iro->ir_ro);
    625 		rtcache_free(&iro->ir_ro);
    626 		mutex_exit(iro->ir_lock);
    627 		percpu_putref(sc->ipsec_ro_percpu);
    628 		m_freem(m);
    629 		return ENETUNREACH;
    630 	}
    631 	rtcache_unref(rt, &iro->ir_ro);
    632 
    633 	/* set NAT-T ports */
    634 	error = ipsecif_set_natt_ports(var, m);
    635 	if (error) {
    636 		m_freem(m);
    637 		goto out;
    638 	}
    639 
    640 	/*
    641 	 * force fragmentation to minimum MTU, to avoid path MTU discovery.
    642 	 * it is too painful to ask for resend of inner packet, to achieve
    643 	 * path MTU discovery for encapsulated packets.
    644 	 */
    645 	error = ip6_output(m, 0, &iro->ir_ro,
    646 	    ip6_ipsec_pmtu ? 0 : IPV6_MINMTU, 0, NULL, NULL);
    647 
    648 out:
    649 	if (error)
    650 		rtcache_free(&iro->ir_ro);
    651 	mutex_exit(iro->ir_lock);
    652 	percpu_putref(sc->ipsec_ro_percpu);
    653 
    654 	return error;
    655 }
    656 #endif /* INET6 */
    657 
    658 static void
    659 ipsecif4_input(struct mbuf *m, int off, int proto, void *eparg)
    660 {
    661 	struct ifnet *ipsecp;
    662 	struct ipsec_softc *sc = eparg;
    663 	struct ipsec_variant *var;
    664 	const struct ip *ip;
    665 	int af;
    666 #ifndef IPSEC_TX_TOS_CLEAR
    667 	u_int8_t otos;
    668 #endif
    669 	struct psref psref_rcvif;
    670 	struct psref psref_var;
    671 	struct ifnet *rcvif;
    672 
    673 	KASSERT(sc != NULL);
    674 
    675 	ipsecp = &sc->ipsec_if;
    676 	if ((ipsecp->if_flags & IFF_UP) == 0) {
    677 		m_freem(m);
    678 		ip_statinc(IP_STAT_NOIPSEC);
    679 		return;
    680 	}
    681 
    682 	var = if_ipsec_getref_variant(sc, &psref_var);
    683 	if (if_ipsec_variant_is_unconfigured(var)) {
    684 		if_ipsec_putref_variant(var, &psref_var);
    685 		m_freem(m);
    686 		ip_statinc(IP_STAT_NOIPSEC);
    687 		return;
    688 	}
    689 
    690 	ip = mtod(m, const struct ip *);
    691 
    692 	rcvif = m_get_rcvif_psref(m, &psref_rcvif);
    693 	if (rcvif == NULL || !ipsecif4_filter4(ip, var, rcvif)) {
    694 		m_put_rcvif_psref(rcvif, &psref_rcvif);
    695 		if_ipsec_putref_variant(var, &psref_var);
    696 		m_freem(m);
    697 		ip_statinc(IP_STAT_NOIPSEC);
    698 		return;
    699 	}
    700 	m_put_rcvif_psref(rcvif, &psref_rcvif);
    701 	if_ipsec_putref_variant(var, &psref_var);
    702 #ifndef IPSEC_TX_TOS_CLEAR
    703 	otos = ip->ip_tos;
    704 #endif
    705 	m_adj(m, off);
    706 
    707 	switch (proto) {
    708 	case IPPROTO_IPV4:
    709 	    {
    710 		struct ip *xip;
    711 		af = AF_INET;
    712 		if (M_UNWRITABLE(m, sizeof(*xip))) {
    713 			m = m_pullup(m, sizeof(*xip));
    714 			if (m == NULL)
    715 				return;
    716 		}
    717 		xip = mtod(m, struct ip *);
    718 #ifndef IPSEC_TX_TOS_CLEAR
    719 		if (ipsecp->if_flags & IFF_ECN)
    720 			ip_ecn_egress(ECN_ALLOWED, &otos, &xip->ip_tos);
    721 		else
    722 			ip_ecn_egress(ECN_NOCARE, &otos, &xip->ip_tos);
    723 #endif
    724 		break;
    725 	    }
    726 #ifdef INET6
    727 	case IPPROTO_IPV6:
    728 	    {
    729 		struct ip6_hdr *ip6;
    730 		u_int8_t itos;
    731 		af = AF_INET6;
    732 		if (M_UNWRITABLE(m, sizeof(*ip6))) {
    733 			m = m_pullup(m, sizeof(*ip6));
    734 			if (m == NULL)
    735 				return;
    736 		}
    737 		ip6 = mtod(m, struct ip6_hdr *);
    738 		itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
    739 #ifndef IPSEC_TX_TOS_CLEAR
    740 		if (ipsecp->if_flags & IFF_ECN)
    741 			ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
    742 		else
    743 			ip_ecn_egress(ECN_NOCARE, &otos, &itos);
    744 #endif
    745 		ip6->ip6_flow &= ~htonl(0xff << 20);
    746 		ip6->ip6_flow |= htonl((u_int32_t)itos << 20);
    747 		break;
    748 	    }
    749 #endif /* INET6 */
    750 	default:
    751 		ip_statinc(IP_STAT_NOIPSEC);
    752 		m_freem(m);
    753 		return;
    754 	}
    755 	if_ipsec_input(m, af, ipsecp);
    756 
    757 	return;
    758 }
    759 
    760 /*
    761  * validate and filter the packet
    762  */
    763 static int
    764 ipsecif4_filter4(const struct ip *ip, struct ipsec_variant *var,
    765     struct ifnet *ifp)
    766 {
    767 	struct sockaddr_in *src, *dst;
    768 
    769 	src = satosin(var->iv_psrc);
    770 	dst = satosin(var->iv_pdst);
    771 
    772 	return in_tunnel_validate(ip, src->sin_addr, dst->sin_addr);
    773 }
    774 
    775 #ifdef INET6
    776 static int
    777 ipsecif6_input(struct mbuf **mp, int *offp, int proto, void *eparg)
    778 {
    779 	struct mbuf *m = *mp;
    780 	struct ifnet *ipsecp;
    781 	struct ipsec_softc *sc = eparg;
    782 	struct ipsec_variant *var;
    783 	struct ip6_hdr *ip6;
    784 	int af = 0;
    785 #ifndef IPSEC_TX_TOS_CLEAR
    786 	u_int32_t otos;
    787 #endif
    788 	struct psref psref_rcvif;
    789 	struct psref psref_var;
    790 	struct ifnet *rcvif;
    791 
    792 	KASSERT(eparg != NULL);
    793 
    794 	ipsecp = &sc->ipsec_if;
    795 	if ((ipsecp->if_flags & IFF_UP) == 0) {
    796 		m_freem(m);
    797 		IP6_STATINC(IP6_STAT_NOIPSEC);
    798 		return IPPROTO_DONE;
    799 	}
    800 
    801 	var = if_ipsec_getref_variant(sc, &psref_var);
    802 	if (if_ipsec_variant_is_unconfigured(var)) {
    803 		if_ipsec_putref_variant(var, &psref_var);
    804 		m_freem(m);
    805 		IP6_STATINC(IP6_STAT_NOIPSEC);
    806 		return IPPROTO_DONE;
    807 	}
    808 
    809 	ip6 = mtod(m, struct ip6_hdr *);
    810 
    811 	rcvif = m_get_rcvif_psref(m, &psref_rcvif);
    812 	if (rcvif == NULL || !ipsecif6_filter6(ip6, var, rcvif)) {
    813 		m_put_rcvif_psref(rcvif, &psref_rcvif);
    814 		if_ipsec_putref_variant(var, &psref_var);
    815 		m_freem(m);
    816 		IP6_STATINC(IP6_STAT_NOIPSEC);
    817 		return IPPROTO_DONE;
    818 	}
    819 	m_put_rcvif_psref(rcvif, &psref_rcvif);
    820 	if_ipsec_putref_variant(var, &psref_var);
    821 
    822 #ifndef IPSEC_TX_TOS_CLEAR
    823 	otos = ip6->ip6_flow;
    824 #endif
    825 	m_adj(m, *offp);
    826 
    827 	switch (proto) {
    828 #ifdef INET
    829 	case IPPROTO_IPV4:
    830 	    {
    831 		af = AF_INET;
    832 #ifndef IPSEC_TX_TOS_CLEAR
    833 		struct ip *ip;
    834 		u_int8_t otos8;
    835 		otos8 = (ntohl(otos) >> 20) & 0xff;
    836 
    837 		if (M_UNWRITABLE(m, sizeof(*ip))) {
    838 			m = m_pullup(m, sizeof(*ip));
    839 			if (m == NULL)
    840 				return IPPROTO_DONE;
    841 		}
    842 		ip = mtod(m, struct ip *);
    843 		if (ipsecp->if_flags & IFF_ECN)
    844 			ip_ecn_egress(ECN_ALLOWED, &otos8, &ip->ip_tos);
    845 		else
    846 			ip_ecn_egress(ECN_NOCARE, &otos8, &ip->ip_tos);
    847 #endif
    848 		break;
    849 	    }
    850 #endif /* INET */
    851 	case IPPROTO_IPV6:
    852 	    {
    853 		af = AF_INET6;
    854 #ifndef IPSEC_TX_TOS_CLEAR
    855 		struct ip6_hdr *xip6;
    856 
    857 		if (M_UNWRITABLE(m, sizeof(*xip6))) {
    858 			m = m_pullup(m, sizeof(*xip6));
    859 			if (m == NULL)
    860 				return IPPROTO_DONE;
    861 		}
    862 		xip6 = mtod(m, struct ip6_hdr *);
    863 		if (ipsecp->if_flags & IFF_ECN)
    864 			ip6_ecn_egress(ECN_ALLOWED, &otos, &xip6->ip6_flow);
    865 		else
    866 			ip6_ecn_egress(ECN_NOCARE, &otos, &xip6->ip6_flow);
    867 		break;
    868 #endif
    869 	    }
    870 	default:
    871 		IP6_STATINC(IP6_STAT_NOIPSEC);
    872 		m_freem(m);
    873 		return IPPROTO_DONE;
    874 	}
    875 
    876 	if_ipsec_input(m, af, ipsecp);
    877 	return IPPROTO_DONE;
    878 }
    879 
    880 /*
    881  * validate and filter the packet.
    882  */
    883 static int
    884 ipsecif6_filter6(const struct ip6_hdr *ip6, struct ipsec_variant *var,
    885     struct ifnet *ifp)
    886 {
    887 	struct sockaddr_in6 *src, *dst;
    888 
    889 	src = satosin6(var->iv_psrc);
    890 	dst = satosin6(var->iv_pdst);
    891 
    892 	return in6_tunnel_validate(ip6, &src->sin6_addr, &dst->sin6_addr);
    893 }
    894 #endif /* INET6 */
    895 
    896 int
    897 ipsecif4_attach(struct ipsec_variant *var)
    898 {
    899 	struct ipsec_softc *sc = var->iv_softc;
    900 
    901 	KASSERT(if_ipsec_variant_is_configured(var));
    902 
    903 	if (var->iv_encap_cookie4 != NULL)
    904 		return EALREADY;
    905 	var->iv_encap_cookie4 = encap_attach_func(AF_INET, -1, if_ipsec_encap_func,
    906 	    &ipsecif4_encapsw, sc);
    907 	if (var->iv_encap_cookie4 == NULL)
    908 		return EEXIST;
    909 
    910 	var->iv_output = ipsecif4_output;
    911 	return 0;
    912 }
    913 
    914 int
    915 ipsecif4_detach(struct ipsec_variant *var)
    916 {
    917 	int error;
    918 
    919 	if (var->iv_encap_cookie4 == NULL)
    920 		return 0;
    921 
    922 	var->iv_output = NULL;
    923 	error = encap_detach(var->iv_encap_cookie4);
    924 	if (error == 0)
    925 		var->iv_encap_cookie4 = NULL;
    926 
    927 	return error;
    928 }
    929 
    930 #ifdef INET6
    931 int
    932 ipsecif6_attach(struct ipsec_variant *var)
    933 {
    934 	struct ipsec_softc *sc = var->iv_softc;
    935 
    936 	KASSERT(if_ipsec_variant_is_configured(var));
    937 	KASSERT(var->iv_encap_cookie6 == NULL);
    938 
    939 	var->iv_encap_cookie6 = encap_attach_func(AF_INET6, -1, if_ipsec_encap_func,
    940 	    &ipsecif6_encapsw, sc);
    941 	if (var->iv_encap_cookie6 == NULL)
    942 		return EEXIST;
    943 
    944 	var->iv_output = ipsecif6_output;
    945 	return 0;
    946 }
    947 
    948 static void
    949 ipsecif6_rtcache_free_pc(void *p, void *arg __unused, struct cpu_info *ci __unused)
    950 {
    951 	struct ipsec_ro *iro = p;
    952 
    953 	mutex_enter(iro->ir_lock);
    954 	rtcache_free(&iro->ir_ro);
    955 	mutex_exit(iro->ir_lock);
    956 }
    957 
    958 int
    959 ipsecif6_detach(struct ipsec_variant *var)
    960 {
    961 	struct ipsec_softc *sc = var->iv_softc;
    962 	int error;
    963 
    964 	KASSERT(var->iv_encap_cookie6 != NULL);
    965 
    966 	percpu_foreach(sc->ipsec_ro_percpu, ipsecif6_rtcache_free_pc, NULL);
    967 
    968 	var->iv_output = NULL;
    969 	error = encap_detach(var->iv_encap_cookie6);
    970 	if (error == 0)
    971 		var->iv_encap_cookie6 = NULL;
    972 	return error;
    973 }
    974 
    975 void *
    976 ipsecif6_ctlinput(int cmd, const struct sockaddr *sa, void *d, void *eparg)
    977 {
    978 	struct ipsec_softc *sc = eparg;
    979 	struct ip6ctlparam *ip6cp = NULL;
    980 	struct ip6_hdr *ip6;
    981 	const struct sockaddr_in6 *dst6;
    982 	struct ipsec_ro *iro;
    983 
    984 	if (sa->sa_family != AF_INET6 ||
    985 	    sa->sa_len != sizeof(struct sockaddr_in6))
    986 		return NULL;
    987 
    988 	if ((unsigned)cmd >= PRC_NCMDS)
    989 		return NULL;
    990 	if (cmd == PRC_HOSTDEAD)
    991 		d = NULL;
    992 	else if (inet6ctlerrmap[cmd] == 0)
    993 		return NULL;
    994 
    995 	/* if the parameter is from icmp6, decode it. */
    996 	if (d != NULL) {
    997 		ip6cp = (struct ip6ctlparam *)d;
    998 		ip6 = ip6cp->ip6c_ip6;
    999 	} else {
   1000 		ip6 = NULL;
   1001 	}
   1002 
   1003 	if (!ip6)
   1004 		return NULL;
   1005 
   1006 	iro = percpu_getref(sc->ipsec_ro_percpu);
   1007 	mutex_enter(iro->ir_lock);
   1008 	dst6 = satocsin6(rtcache_getdst(&iro->ir_ro));
   1009 	/* XXX scope */
   1010 	if (dst6 == NULL)
   1011 		;
   1012 	else if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst6->sin6_addr))
   1013 		/* flush route cache */
   1014 		rtcache_free(&iro->ir_ro);
   1015 
   1016 	mutex_exit(iro->ir_lock);
   1017 	percpu_putref(sc->ipsec_ro_percpu);
   1018 
   1019 	return NULL;
   1020 }
   1021 
   1022 ENCAP_PR_WRAP_CTLINPUT(ipsecif6_ctlinput)
   1023 #define	ipsecif6_ctlinput	ipsecif6_ctlinput_wrapper
   1024 
   1025 static const struct encapsw ipsecif6_encapsw = {
   1026 	.encapsw6 = {
   1027 		.pr_input = ipsecif6_input,
   1028 		.pr_ctlinput = ipsecif6_ctlinput,
   1029 	}
   1030 };
   1031 #endif /* INET6 */
   1032