2 * Copyright (c) 2002, 2003 Sam Leffler, Errno Consulting
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * IPsec output processing.
33 #include "opt_inet6.h"
34 #include "opt_ipsec.h"
37 #include <sys/param.h>
38 #include <sys/systm.h>
40 #include <sys/domain.h>
41 #include <sys/protosw.h>
42 #include <sys/socket.h>
43 #include <sys/errno.h>
44 #include <sys/syslog.h>
47 #include <net/if_var.h>
51 #include <netinet/in.h>
52 #include <netinet/in_systm.h>
53 #include <netinet/ip.h>
54 #include <netinet/ip_var.h>
55 #include <netinet/in_var.h>
56 #include <netinet/ip_ecn.h>
58 #include <netinet6/ip6_ecn.h>
61 #include <netinet/ip6.h>
63 #include <netinet6/ip6_var.h>
64 #include <netinet6/scope6_var.h>
66 #include <netinet/in_pcb.h>
68 #include <netinet/icmp6.h>
71 #include <netipsec/ipsec.h>
73 #include <netipsec/ipsec6.h>
75 #include <netipsec/ah_var.h>
76 #include <netipsec/esp_var.h>
77 #include <netipsec/ipcomp_var.h>
79 #include <netipsec/xform.h>
81 #include <netipsec/key.h>
82 #include <netipsec/keydb.h>
83 #include <netipsec/key_debug.h>
85 #include <machine/in_cksum.h>
88 #include <netinet/udp.h>
92 #include <net/if_enc.h>
97 ipsec_process_done(struct mbuf *m, struct ipsecrequest *isr)
99 struct tdb_ident *tdbi;
101 struct secasvar *sav;
102 struct secasindex *saidx;
105 IPSEC_ASSERT(m != NULL, ("null mbuf"));
106 IPSEC_ASSERT(isr != NULL, ("null ISR"));
107 IPSEC_ASSERT(isr->sp != NULL, ("NULL isr->sp"));
109 IPSEC_ASSERT(sav != NULL, ("null SA"));
110 IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
112 saidx = &sav->sah->saidx;
113 switch (saidx->dst.sa.sa_family) {
116 /* Fix the header length, for AH processing. */
117 mtod(m, struct ip *)->ip_len = htons(m->m_pkthdr.len);
122 /* Fix the header length, for AH processing. */
123 if (m->m_pkthdr.len < sizeof (struct ip6_hdr)) {
127 if (m->m_pkthdr.len - sizeof (struct ip6_hdr) > IPV6_MAXPACKET) {
128 /* No jumbogram support. */
132 mtod(m, struct ip6_hdr *)->ip6_plen =
133 htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
137 DPRINTF(("%s: unknown protocol family %u\n", __func__,
138 saidx->dst.sa.sa_family));
144 * Add a record of what we've done or what needs to be done to the
147 mtag = m_tag_get(PACKET_TAG_IPSEC_OUT_DONE,
148 sizeof(struct tdb_ident), M_NOWAIT);
150 DPRINTF(("%s: could not get packet tag\n", __func__));
155 tdbi = (struct tdb_ident *)(mtag + 1);
156 tdbi->dst = saidx->dst;
157 tdbi->proto = saidx->proto;
158 tdbi->spi = sav->spi;
159 m_tag_prepend(m, mtag);
162 * If there's another (bundled) SA to apply, do so.
163 * Note that this puts a burden on the kernel stack size.
164 * If this is a problem we'll need to introduce a queue
165 * to set the packet on so we can unwind the stack before
166 * doing further processing.
168 * If ipsec[46]_process_packet() will successfully queue
169 * the request, we need to take additional reference to SP,
170 * because xform callback will release reference.
173 /* XXX-BZ currently only support same AF bundles. */
174 switch (saidx->dst.sa.sa_family) {
177 IPSECSTAT_INC(ips_out_bundlesa);
179 error = ipsec4_process_packet(m, isr->next);
181 KEY_FREESP(&isr->sp);
189 IPSEC6STAT_INC(ips_out_bundlesa);
191 error = ipsec6_process_packet(m, isr->next);
193 KEY_FREESP(&isr->sp);
199 DPRINTF(("%s: unknown protocol family %u\n", __func__,
200 saidx->dst.sa.sa_family));
205 key_sa_recordxfer(sav, m); /* record data transfer */
208 * We're done with IPsec processing, transmit the packet using the
209 * appropriate network protocol (IP or IPv6).
211 switch (saidx->dst.sa.sa_family) {
216 * If NAT-T is enabled, now that all IPsec processing is done
217 * insert UDP encapsulation header after IP header.
219 if (sav->natt_type) {
220 struct ip *ip = mtod(m, struct ip *);
221 const int hlen = (ip->ip_hl << 2);
226 size = sizeof(struct udphdr);
227 if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE) {
229 * draft-ietf-ipsec-nat-t-ike-0[01].txt and
230 * draft-ietf-ipsec-udp-encaps-(00/)01.txt,
231 * ignoring possible AH mode
232 * non-IKE marker + non-ESP marker
233 * from draft-ietf-ipsec-udp-encaps-00.txt.
235 size += sizeof(u_int64_t);
237 mi = m_makespace(m, hlen, size, &off);
239 DPRINTF(("%s: m_makespace for udphdr failed\n",
245 udp = (struct udphdr *)(mtod(mi, caddr_t) + off);
246 if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE)
247 udp->uh_sport = htons(UDP_ENCAP_ESPINUDP_PORT);
250 KEY_PORTFROMSADDR(&sav->sah->saidx.src);
251 udp->uh_dport = KEY_PORTFROMSADDR(&sav->sah->saidx.dst);
253 udp->uh_ulen = htons(m->m_pkthdr.len - hlen);
254 ip->ip_len = htons(m->m_pkthdr.len);
255 ip->ip_p = IPPROTO_UDP;
257 if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE)
258 *(u_int64_t *)(udp + 1) = 0;
260 #endif /* IPSEC_NAT_T */
262 return ip_output(m, NULL, NULL, IP_RAWOUTPUT, NULL, NULL);
267 * We don't need massage, IPv6 header fields are always in
270 return ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
273 panic("ipsec_process_done");
279 static struct ipsecrequest *
282 struct ipsecrequest *isr,
284 struct secasindex *saidx,
288 #define IPSEC_OSTAT(name) do { \
289 if (isr->saidx.proto == IPPROTO_ESP) \
290 ESPSTAT_INC(esps_##name); \
291 else if (isr->saidx.proto == IPPROTO_AH)\
292 AHSTAT_INC(ahs_##name); \
294 IPCOMPSTAT_INC(ipcomps_##name); \
296 struct secasvar *sav;
298 IPSECREQUEST_LOCK_ASSERT(isr);
300 IPSEC_ASSERT(af == AF_INET || af == AF_INET6,
301 ("invalid address family %u", af));
304 * Craft SA index to search for proper SA. Note that
305 * we only fillin unspecified SA peers for transport
306 * mode; for tunnel mode they must already be filled in.
309 if (isr->saidx.mode == IPSEC_MODE_TRANSPORT) {
310 /* Fillin unspecified SA peers only for transport mode */
312 struct sockaddr_in *sin;
313 struct ip *ip = mtod(m, struct ip *);
315 if (saidx->src.sa.sa_len == 0) {
316 sin = &saidx->src.sin;
317 sin->sin_len = sizeof(*sin);
318 sin->sin_family = AF_INET;
319 sin->sin_port = IPSEC_PORT_ANY;
320 sin->sin_addr = ip->ip_src;
322 if (saidx->dst.sa.sa_len == 0) {
323 sin = &saidx->dst.sin;
324 sin->sin_len = sizeof(*sin);
325 sin->sin_family = AF_INET;
326 sin->sin_port = IPSEC_PORT_ANY;
327 sin->sin_addr = ip->ip_dst;
330 struct sockaddr_in6 *sin6;
331 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
333 if (saidx->src.sin6.sin6_len == 0) {
334 sin6 = (struct sockaddr_in6 *)&saidx->src;
335 sin6->sin6_len = sizeof(*sin6);
336 sin6->sin6_family = AF_INET6;
337 sin6->sin6_port = IPSEC_PORT_ANY;
338 sin6->sin6_addr = ip6->ip6_src;
339 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
340 /* fix scope id for comparing SPD */
341 sin6->sin6_addr.s6_addr16[1] = 0;
342 sin6->sin6_scope_id =
343 ntohs(ip6->ip6_src.s6_addr16[1]);
346 if (saidx->dst.sin6.sin6_len == 0) {
347 sin6 = (struct sockaddr_in6 *)&saidx->dst;
348 sin6->sin6_len = sizeof(*sin6);
349 sin6->sin6_family = AF_INET6;
350 sin6->sin6_port = IPSEC_PORT_ANY;
351 sin6->sin6_addr = ip6->ip6_dst;
352 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
353 /* fix scope id for comparing SPD */
354 sin6->sin6_addr.s6_addr16[1] = 0;
355 sin6->sin6_scope_id =
356 ntohs(ip6->ip6_dst.s6_addr16[1]);
363 * Lookup SA and validate it.
365 *error = key_checkrequest(isr, saidx);
368 * IPsec processing is required, but no SA found.
369 * I assume that key_acquire() had been called
370 * to get/establish the SA. Here I discard
371 * this packet because it is responsibility for
372 * upper layer to retransmit the packet.
376 IPSECSTAT_INC(ips_out_nosa);
380 IPSEC6STAT_INC(ips_out_nosa);
388 IPSEC_ASSERT(ipsec_get_reqlevel(isr) == IPSEC_LEVEL_USE,
389 ("no SA found, but required; level %u",
390 ipsec_get_reqlevel(isr)));
391 IPSECREQUEST_UNLOCK(isr);
394 * If isr is NULL, we found a 'use' policy w/o SA.
395 * Return w/o error and w/o isr so we can drop out
396 * and continue w/o IPsec processing.
400 IPSECREQUEST_LOCK(isr);
405 * Check system global policy controls.
407 if ((isr->saidx.proto == IPPROTO_ESP && !V_esp_enable) ||
408 (isr->saidx.proto == IPPROTO_AH && !V_ah_enable) ||
409 (isr->saidx.proto == IPPROTO_IPCOMP && !V_ipcomp_enable)) {
410 DPRINTF(("%s: IPsec outbound packet dropped due"
411 " to policy (check your sysctls)\n", __func__));
413 *error = EHOSTUNREACH;
418 * Sanity check the SA contents for the caller
419 * before they invoke the xform output method.
421 if (sav->tdb_xform == NULL) {
422 DPRINTF(("%s: no transform for SA\n", __func__));
423 IPSEC_OSTAT(noxform);
424 *error = EHOSTUNREACH;
429 IPSEC_ASSERT(*error != 0, ("error return w/ no error code"));
430 IPSECREQUEST_UNLOCK(isr);
436 ipsec_encap(struct mbuf **mp, struct secasindex *saidx)
445 ip = mtod(*mp, struct ip *);
449 proto = IPPROTO_IPIP;
451 * Collect IP_DF state from the inner header
452 * and honor system-wide control of how to handle it.
454 switch (V_ip4_ipsec_dfbit) {
455 case 0: /* clear in outer header */
456 case 1: /* set in outer header */
457 setdf = V_ip4_ipsec_dfbit;
459 default:/* propagate to outer header */
460 setdf = (ip->ip_off & ntohs(IP_DF)) != 0;
466 case (IPV6_VERSION >> 4):
467 proto = IPPROTO_IPV6;
468 ip6 = mtod(*mp, struct ip6_hdr *);
469 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
470 setdf = V_ip4_ipsec_dfbit ? 1: 0;
471 /* scoped address handling */
472 in6_clearscope(&ip6->ip6_src);
473 in6_clearscope(&ip6->ip6_dst);
477 return (EAFNOSUPPORT);
479 switch (saidx->dst.sa.sa_family) {
482 if (saidx->src.sa.sa_family != AF_INET ||
483 saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
484 saidx->dst.sin.sin_addr.s_addr == INADDR_ANY)
486 M_PREPEND(*mp, sizeof(struct ip), M_NOWAIT);
489 ip = mtod(*mp, struct ip *);
490 ip->ip_v = IPVERSION;
491 ip->ip_hl = sizeof(struct ip) >> 2;
493 ip->ip_len = htons((*mp)->m_pkthdr.len);
494 ip->ip_ttl = V_ip_defttl;
496 ip->ip_off = setdf ? htons(IP_DF): 0;
497 ip->ip_src = saidx->src.sin.sin_addr;
498 ip->ip_dst = saidx->dst.sin.sin_addr;
499 ip_ecn_ingress(V_ip4_ipsec_ecn, &ip->ip_tos, &itos);
505 if (saidx->src.sa.sa_family != AF_INET6 ||
506 IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr) ||
507 IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr))
509 M_PREPEND(*mp, sizeof(struct ip6_hdr), M_NOWAIT);
512 ip6 = mtod(*mp, struct ip6_hdr *);
514 ip6->ip6_vfc = IPV6_VERSION;
515 ip6->ip6_hlim = V_ip6_defhlim;
516 ip6->ip6_nxt = proto;
517 ip6->ip6_dst = saidx->dst.sin6.sin6_addr;
518 /* For link-local address embed scope zone id */
519 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
520 ip6->ip6_dst.s6_addr16[1] =
521 htons(saidx->dst.sin6.sin6_scope_id & 0xffff);
522 ip6->ip6_src = saidx->src.sin6.sin6_addr;
523 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
524 ip6->ip6_src.s6_addr16[1] =
525 htons(saidx->src.sin6.sin6_scope_id & 0xffff);
526 ip6->ip6_plen = htons((*mp)->m_pkthdr.len - sizeof(*ip6));
527 ip_ecn_ingress(V_ip6_ipsec_ecn, &proto, &itos);
528 ip6->ip6_flow |= htonl((uint32_t)proto << 20);
532 return (EAFNOSUPPORT);
539 * IPsec output logic for IPv4.
542 ipsec4_process_packet(struct mbuf *m, struct ipsecrequest *isr)
544 char sbuf[INET6_ADDRSTRLEN], dbuf[INET6_ADDRSTRLEN];
545 union sockaddr_union *dst;
546 struct secasindex saidx;
547 struct secasvar *sav;
551 IPSEC_ASSERT(m != NULL, ("null mbuf"));
552 IPSEC_ASSERT(isr != NULL, ("null isr"));
554 IPSECREQUEST_LOCK(isr); /* insure SA contents don't change */
556 isr = ipsec_nextisr(m, isr, AF_INET, &saidx, &error);
564 if (m->m_len < sizeof(struct ip) &&
565 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
569 ip = mtod(m, struct ip *);
570 dst = &sav->sah->saidx.dst;
572 if_inc_counter(encif, IFCOUNTER_OPACKETS, 1);
573 if_inc_counter(encif, IFCOUNTER_OBYTES, m->m_pkthdr.len);
575 /* pass the mbuf to enc0 for bpf processing */
576 ipsec_bpf(m, sav, AF_INET, ENC_OUT|ENC_BEFORE);
577 /* pass the mbuf to enc0 for packet filtering */
578 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0)
580 ip = mtod(m, struct ip *);
582 /* Do the appropriate encapsulation, if necessary */
583 if (isr->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
584 dst->sa.sa_family != AF_INET || /* PF mismatch */
585 (dst->sa.sa_family == AF_INET && /* Proxy */
586 dst->sin.sin_addr.s_addr != INADDR_ANY &&
587 dst->sin.sin_addr.s_addr != ip->ip_dst.s_addr)) {
588 /* Fix IPv4 header checksum and length */
589 ip->ip_len = htons(m->m_pkthdr.len);
591 ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
592 error = ipsec_encap(&m, &sav->sah->saidx);
594 DPRINTF(("%s: encapsulation for SA %s->%s "
595 "SPI 0x%08x failed with error %d\n", __func__,
596 ipsec_address(&sav->sah->saidx.src, sbuf,
598 ipsec_address(&sav->sah->saidx.dst, dbuf,
599 sizeof(dbuf)), ntohl(sav->spi), error));
604 /* pass the mbuf to enc0 for bpf processing */
605 ipsec_bpf(m, sav, sav->sah->saidx.dst.sa.sa_family, ENC_OUT|ENC_AFTER);
606 /* pass the mbuf to enc0 for packet filtering */
607 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0)
612 * Dispatch to the appropriate IPsec transform logic. The
613 * packet will be returned for transmission after crypto
614 * processing, etc. are completed.
616 * NB: m & sav are ``passed to caller'' who's reponsible for
617 * for reclaiming their resources.
619 switch(dst->sa.sa_family) {
621 ip = mtod(m, struct ip *);
623 off = offsetof(struct ip, ip_p);
627 i = sizeof(struct ip6_hdr);
628 off = offsetof(struct ip6_hdr, ip6_nxt);
632 DPRINTF(("%s: unsupported protocol family %u\n",
633 __func__, dst->sa.sa_family));
634 error = EPFNOSUPPORT;
635 IPSECSTAT_INC(ips_out_inval);
638 error = (*sav->tdb_xform->xf_output)(m, isr, NULL, i, off);
639 IPSECREQUEST_UNLOCK(isr);
643 IPSECREQUEST_UNLOCK(isr);
653 in6_sa_equal_addrwithscope(const struct sockaddr_in6 *sa, const struct in6_addr *ia)
657 memcpy(&ia2, &sa->sin6_addr, sizeof(ia2));
658 if (IN6_IS_SCOPE_LINKLOCAL(&sa->sin6_addr))
659 ia2.s6_addr16[1] = htons(sa->sin6_scope_id);
661 return IN6_ARE_ADDR_EQUAL(ia, &ia2);
665 * IPsec output logic for IPv6.
668 ipsec6_process_packet(struct mbuf *m, struct ipsecrequest *isr)
670 char sbuf[INET6_ADDRSTRLEN], dbuf[INET6_ADDRSTRLEN];
671 struct secasindex saidx;
672 struct secasvar *sav;
675 union sockaddr_union *dst;
677 IPSEC_ASSERT(m != NULL, ("ipsec6_process_packet: null mbuf"));
678 IPSEC_ASSERT(isr != NULL, ("ipsec6_process_packet: null isr"));
680 IPSECREQUEST_LOCK(isr); /* insure SA contents don't change */
682 isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
689 dst = &sav->sah->saidx.dst;
691 ip6 = mtod(m, struct ip6_hdr *);
692 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(*ip6));
694 if_inc_counter(encif, IFCOUNTER_OPACKETS, 1);
695 if_inc_counter(encif, IFCOUNTER_OBYTES, m->m_pkthdr.len);
697 /* pass the mbuf to enc0 for bpf processing */
698 ipsec_bpf(m, isr->sav, AF_INET6, ENC_OUT|ENC_BEFORE);
699 /* pass the mbuf to enc0 for packet filtering */
700 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0)
702 ip6 = mtod(m, struct ip6_hdr *);
705 /* Do the appropriate encapsulation, if necessary */
706 if (isr->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
707 dst->sa.sa_family != AF_INET6 || /* PF mismatch */
708 ((dst->sa.sa_family == AF_INET6) &&
709 (!IN6_IS_ADDR_UNSPECIFIED(&dst->sin6.sin6_addr)) &&
710 (!in6_sa_equal_addrwithscope(&dst->sin6,
712 if (m->m_pkthdr.len - sizeof(*ip6) > IPV6_MAXPACKET) {
713 /* No jumbogram support. */
714 error = ENXIO; /*XXX*/
717 error = ipsec_encap(&m, &sav->sah->saidx);
719 DPRINTF(("%s: encapsulation for SA %s->%s "
720 "SPI 0x%08x failed with error %d\n", __func__,
721 ipsec_address(&sav->sah->saidx.src, sbuf,
723 ipsec_address(&sav->sah->saidx.dst, dbuf,
724 sizeof(dbuf)), ntohl(sav->spi), error));
730 ipsec_bpf(m, isr->sav, dst->sa.sa_family, ENC_OUT|ENC_AFTER);
731 /* pass the mbuf to enc0 for packet filtering */
732 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0)
736 switch(dst->sa.sa_family) {
741 ip = mtod(m, struct ip *);
743 off = offsetof(struct ip, ip_p);
748 i = sizeof(struct ip6_hdr);
749 off = offsetof(struct ip6_hdr, ip6_nxt);
752 DPRINTF(("%s: unsupported protocol family %u\n",
753 __func__, dst->sa.sa_family));
754 error = EPFNOSUPPORT;
755 IPSEC6STAT_INC(ips_out_inval);
758 error = (*sav->tdb_xform->xf_output)(m, isr, NULL, i, off);
759 IPSECREQUEST_UNLOCK(isr);
764 IPSECREQUEST_UNLOCK(isr);