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>
48 #include <net/route.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>
65 #include <netinet/in_pcb.h>
67 #include <netinet/icmp6.h>
70 #include <netipsec/ipsec.h>
72 #include <netipsec/ipsec6.h>
74 #include <netipsec/ah_var.h>
75 #include <netipsec/esp_var.h>
76 #include <netipsec/ipcomp_var.h>
78 #include <netipsec/xform.h>
80 #include <netipsec/key.h>
81 #include <netipsec/keydb.h>
82 #include <netipsec/key_debug.h>
84 #include <machine/in_cksum.h>
87 #include <netinet/udp.h>
91 #include <net/if_enc.h>
96 ipsec_process_done(struct mbuf *m, struct ipsecrequest *isr)
98 struct tdb_ident *tdbi;
100 struct secasvar *sav;
101 struct secasindex *saidx;
104 IPSEC_ASSERT(m != NULL, ("null mbuf"));
105 IPSEC_ASSERT(isr != NULL, ("null ISR"));
107 IPSEC_ASSERT(sav != NULL, ("null SA"));
108 IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
110 saidx = &sav->sah->saidx;
111 switch (saidx->dst.sa.sa_family) {
114 /* Fix the header length, for AH processing. */
115 mtod(m, struct ip *)->ip_len = htons(m->m_pkthdr.len);
120 /* Fix the header length, for AH processing. */
121 if (m->m_pkthdr.len < sizeof (struct ip6_hdr)) {
125 if (m->m_pkthdr.len - sizeof (struct ip6_hdr) > IPV6_MAXPACKET) {
126 /* No jumbogram support. */
130 mtod(m, struct ip6_hdr *)->ip6_plen =
131 htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
135 DPRINTF(("%s: unknown protocol family %u\n", __func__,
136 saidx->dst.sa.sa_family));
142 * Add a record of what we've done or what needs to be done to the
145 mtag = m_tag_get(PACKET_TAG_IPSEC_OUT_DONE,
146 sizeof(struct tdb_ident), M_NOWAIT);
148 DPRINTF(("%s: could not get packet tag\n", __func__));
153 tdbi = (struct tdb_ident *)(mtag + 1);
154 tdbi->dst = saidx->dst;
155 tdbi->proto = saidx->proto;
156 tdbi->spi = sav->spi;
157 m_tag_prepend(m, mtag);
160 * If there's another (bundled) SA to apply, do so.
161 * Note that this puts a burden on the kernel stack size.
162 * If this is a problem we'll need to introduce a queue
163 * to set the packet on so we can unwind the stack before
164 * doing further processing.
167 /* XXX-BZ currently only support same AF bundles. */
168 switch (saidx->dst.sa.sa_family) {
171 IPSECSTAT_INC(ips_out_bundlesa);
172 return ipsec4_process_packet(m, isr->next, 0, 0);
179 IPSEC6STAT_INC(ips_out_bundlesa);
180 return ipsec6_process_packet(m, isr->next);
185 DPRINTF(("%s: unknown protocol family %u\n", __func__,
186 saidx->dst.sa.sa_family));
191 key_sa_recordxfer(sav, m); /* record data transfer */
194 * We're done with IPsec processing, transmit the packet using the
195 * appropriate network protocol (IP or IPv6). SPD lookup will be
196 * performed again there.
198 switch (saidx->dst.sa.sa_family) {
203 * If NAT-T is enabled, now that all IPsec processing is done
204 * insert UDP encapsulation header after IP header.
206 if (sav->natt_type) {
207 struct ip *ip = mtod(m, struct ip *);
208 const int hlen = (ip->ip_hl << 2);
213 size = sizeof(struct udphdr);
214 if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE) {
216 * draft-ietf-ipsec-nat-t-ike-0[01].txt and
217 * draft-ietf-ipsec-udp-encaps-(00/)01.txt,
218 * ignoring possible AH mode
219 * non-IKE marker + non-ESP marker
220 * from draft-ietf-ipsec-udp-encaps-00.txt.
222 size += sizeof(u_int64_t);
224 mi = m_makespace(m, hlen, size, &off);
226 DPRINTF(("%s: m_makespace for udphdr failed\n",
232 udp = (struct udphdr *)(mtod(mi, caddr_t) + off);
233 if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE)
234 udp->uh_sport = htons(UDP_ENCAP_ESPINUDP_PORT);
237 KEY_PORTFROMSADDR(&sav->sah->saidx.src);
238 udp->uh_dport = KEY_PORTFROMSADDR(&sav->sah->saidx.dst);
240 udp->uh_ulen = htons(m->m_pkthdr.len - hlen);
241 ip->ip_len = htons(m->m_pkthdr.len);
242 ip->ip_p = IPPROTO_UDP;
244 if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE)
245 *(u_int64_t *)(udp + 1) = 0;
247 #endif /* IPSEC_NAT_T */
249 return ip_output(m, NULL, NULL, IP_RAWOUTPUT, NULL, NULL);
254 * We don't need massage, IPv6 header fields are always in
257 return ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
260 panic("ipsec_process_done");
266 static struct ipsecrequest *
269 struct ipsecrequest *isr,
271 struct secasindex *saidx,
275 #define IPSEC_OSTAT(name) do { \
276 if (isr->saidx.proto == IPPROTO_ESP) \
277 ESPSTAT_INC(esps_##name); \
278 else if (isr->saidx.proto == IPPROTO_AH)\
279 AHSTAT_INC(ahs_##name); \
281 IPCOMPSTAT_INC(ipcomps_##name); \
283 struct secasvar *sav;
285 IPSECREQUEST_LOCK_ASSERT(isr);
287 IPSEC_ASSERT(af == AF_INET || af == AF_INET6,
288 ("invalid address family %u", af));
291 * Craft SA index to search for proper SA. Note that
292 * we only fillin unspecified SA peers for transport
293 * mode; for tunnel mode they must already be filled in.
296 if (isr->saidx.mode == IPSEC_MODE_TRANSPORT) {
297 /* Fillin unspecified SA peers only for transport mode */
299 struct sockaddr_in *sin;
300 struct ip *ip = mtod(m, struct ip *);
302 if (saidx->src.sa.sa_len == 0) {
303 sin = &saidx->src.sin;
304 sin->sin_len = sizeof(*sin);
305 sin->sin_family = AF_INET;
306 sin->sin_port = IPSEC_PORT_ANY;
307 sin->sin_addr = ip->ip_src;
309 if (saidx->dst.sa.sa_len == 0) {
310 sin = &saidx->dst.sin;
311 sin->sin_len = sizeof(*sin);
312 sin->sin_family = AF_INET;
313 sin->sin_port = IPSEC_PORT_ANY;
314 sin->sin_addr = ip->ip_dst;
317 struct sockaddr_in6 *sin6;
318 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
320 if (saidx->src.sin6.sin6_len == 0) {
321 sin6 = (struct sockaddr_in6 *)&saidx->src;
322 sin6->sin6_len = sizeof(*sin6);
323 sin6->sin6_family = AF_INET6;
324 sin6->sin6_port = IPSEC_PORT_ANY;
325 sin6->sin6_addr = ip6->ip6_src;
326 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
327 /* fix scope id for comparing SPD */
328 sin6->sin6_addr.s6_addr16[1] = 0;
329 sin6->sin6_scope_id =
330 ntohs(ip6->ip6_src.s6_addr16[1]);
333 if (saidx->dst.sin6.sin6_len == 0) {
334 sin6 = (struct sockaddr_in6 *)&saidx->dst;
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_dst;
339 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
340 /* fix scope id for comparing SPD */
341 sin6->sin6_addr.s6_addr16[1] = 0;
342 sin6->sin6_scope_id =
343 ntohs(ip6->ip6_dst.s6_addr16[1]);
350 * Lookup SA and validate it.
352 *error = key_checkrequest(isr, saidx);
355 * IPsec processing is required, but no SA found.
356 * I assume that key_acquire() had been called
357 * to get/establish the SA. Here I discard
358 * this packet because it is responsibility for
359 * upper layer to retransmit the packet.
363 IPSECSTAT_INC(ips_out_nosa);
367 IPSEC6STAT_INC(ips_out_nosa);
375 IPSEC_ASSERT(ipsec_get_reqlevel(isr) == IPSEC_LEVEL_USE,
376 ("no SA found, but required; level %u",
377 ipsec_get_reqlevel(isr)));
378 IPSECREQUEST_UNLOCK(isr);
381 * If isr is NULL, we found a 'use' policy w/o SA.
382 * Return w/o error and w/o isr so we can drop out
383 * and continue w/o IPsec processing.
387 IPSECREQUEST_LOCK(isr);
392 * Check system global policy controls.
394 if ((isr->saidx.proto == IPPROTO_ESP && !V_esp_enable) ||
395 (isr->saidx.proto == IPPROTO_AH && !V_ah_enable) ||
396 (isr->saidx.proto == IPPROTO_IPCOMP && !V_ipcomp_enable)) {
397 DPRINTF(("%s: IPsec outbound packet dropped due"
398 " to policy (check your sysctls)\n", __func__));
400 *error = EHOSTUNREACH;
405 * Sanity check the SA contents for the caller
406 * before they invoke the xform output method.
408 if (sav->tdb_xform == NULL) {
409 DPRINTF(("%s: no transform for SA\n", __func__));
410 IPSEC_OSTAT(noxform);
411 *error = EHOSTUNREACH;
416 IPSEC_ASSERT(*error != 0, ("error return w/ no error code"));
417 IPSECREQUEST_UNLOCK(isr);
424 * IPsec output logic for IPv4.
427 ipsec4_process_packet(
429 struct ipsecrequest *isr,
433 struct secasindex saidx;
434 struct secasvar *sav;
438 IPSEC_ASSERT(m != NULL, ("null mbuf"));
439 IPSEC_ASSERT(isr != NULL, ("null isr"));
441 IPSECREQUEST_LOCK(isr); /* insure SA contents don't change */
443 isr = ipsec_nextisr(m, isr, AF_INET, &saidx, &error);
453 encif->if_opackets++;
454 encif->if_obytes += m->m_pkthdr.len;
456 /* pass the mbuf to enc0 for bpf processing */
457 ipsec_bpf(m, sav, AF_INET, ENC_OUT|ENC_BEFORE);
458 /* pass the mbuf to enc0 for packet filtering */
459 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0)
464 union sockaddr_union *dst = &sav->sah->saidx.dst;
468 * Collect IP_DF state from the outer header.
470 if (dst->sa.sa_family == AF_INET) {
471 if (m->m_len < sizeof (struct ip) &&
472 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
476 ip = mtod(m, struct ip *);
477 /* Honor system-wide control of how to handle IP_DF */
478 switch (V_ip4_ipsec_dfbit) {
479 case 0: /* clear in outer header */
480 case 1: /* set in outer header */
481 setdf = V_ip4_ipsec_dfbit;
483 default: /* propagate to outer header */
484 setdf = ntohs(ip->ip_off & IP_DF);
488 ip = NULL; /* keep compiler happy */
491 /* Do the appropriate encapsulation, if necessary */
492 if (isr->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
493 dst->sa.sa_family != AF_INET || /* PF mismatch */
495 (sav->flags & SADB_X_SAFLAGS_TUNNEL) || /* Tunnel requ'd */
496 sav->tdb_xform->xf_type == XF_IP4 || /* ditto */
498 (dst->sa.sa_family == AF_INET && /* Proxy */
499 dst->sin.sin_addr.s_addr != INADDR_ANY &&
500 dst->sin.sin_addr.s_addr != ip->ip_dst.s_addr)) {
503 /* Fix IPv4 header checksum and length */
504 if (m->m_len < sizeof (struct ip) &&
505 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
509 ip = mtod(m, struct ip *);
510 if (ip->ip_v == IPVERSION) {
511 ip->ip_len = htons(m->m_pkthdr.len);
513 ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
516 /* Encapsulate the packet */
517 error = ipip_output(m, isr, &mp, 0, 0);
518 if (mp == NULL && !error) {
519 /* Should never happen. */
520 DPRINTF(("%s: ipip_output returns no mbuf and "
521 "no error!", __func__));
526 /* XXX: Should never happen! */
529 m = NULL; /* ipip_output() already freed it */
534 * ipip_output clears IP_DF in the new header. If
535 * we need to propagate IP_DF from the outer header,
536 * then we have to do it here.
538 * XXX shouldn't assume what ipip_output does.
540 if (dst->sa.sa_family == AF_INET && setdf) {
541 if (m->m_len < sizeof (struct ip) &&
542 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
546 ip = mtod(m, struct ip *);
547 ip->ip_off = ntohs(ip->ip_off);
549 ip->ip_off = htons(ip->ip_off);
555 /* pass the mbuf to enc0 for bpf processing */
556 ipsec_bpf(m, sav, sav->sah->saidx.dst.sa.sa_family, ENC_OUT|ENC_AFTER);
557 /* pass the mbuf to enc0 for packet filtering */
558 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0)
563 * Dispatch to the appropriate IPsec transform logic. The
564 * packet will be returned for transmission after crypto
565 * processing, etc. are completed. For encapsulation we
566 * bypass this call because of the explicit call done above
567 * (necessary to deal with IP_DF handling for IPv4).
569 * NB: m & sav are ``passed to caller'' who's reponsible for
570 * for reclaiming their resources.
572 if (sav->tdb_xform->xf_type != XF_IP4) {
573 union sockaddr_union *dst = &sav->sah->saidx.dst;
574 switch(dst->sa.sa_family) {
576 ip = mtod(m, struct ip *);
578 off = offsetof(struct ip, ip_p);
582 i = sizeof(struct ip6_hdr);
583 off = offsetof(struct ip6_hdr, ip6_nxt);
587 DPRINTF(("%s: unsupported protocol family %u\n",
588 __func__, dst->sa.sa_family));
589 error = EPFNOSUPPORT;
590 IPSECSTAT_INC(ips_out_inval);
593 error = (*sav->tdb_xform->xf_output)(m, isr, NULL, i, off);
595 error = ipsec_process_done(m, isr);
597 IPSECREQUEST_UNLOCK(isr);
601 IPSECREQUEST_UNLOCK(isr);
611 in6_sa_equal_addrwithscope(const struct sockaddr_in6 *sa, const struct in6_addr *ia)
615 memcpy(&ia2, &sa->sin6_addr, sizeof(ia2));
616 if (IN6_IS_SCOPE_LINKLOCAL(&sa->sin6_addr))
617 ia2.s6_addr16[1] = htons(sa->sin6_scope_id);
619 return IN6_ARE_ADDR_EQUAL(ia, &ia2);
623 * IPsec output logic for IPv6.
626 ipsec6_process_packet(
628 struct ipsecrequest *isr
631 struct secasindex saidx;
632 struct secasvar *sav;
635 union sockaddr_union *dst;
637 IPSEC_ASSERT(m != NULL, ("ipsec6_process_packet: null mbuf"));
638 IPSEC_ASSERT(isr != NULL, ("ipsec6_process_packet: null isr"));
640 IPSECREQUEST_LOCK(isr); /* insure SA contents don't change */
642 isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
650 dst = &sav->sah->saidx.dst;
652 ip6 = mtod(m, struct ip6_hdr *);
653 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(*ip6));
655 encif->if_opackets++;
656 encif->if_obytes += m->m_pkthdr.len;
658 /* pass the mbuf to enc0 for bpf processing */
659 ipsec_bpf(m, isr->sav, AF_INET6, ENC_OUT|ENC_BEFORE);
660 /* pass the mbuf to enc0 for packet filtering */
661 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0)
665 /* Do the appropriate encapsulation, if necessary */
666 if (isr->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
667 dst->sa.sa_family != AF_INET6 || /* PF mismatch */
668 ((dst->sa.sa_family == AF_INET6) &&
669 (!IN6_IS_ADDR_UNSPECIFIED(&dst->sin6.sin6_addr)) &&
670 (!in6_sa_equal_addrwithscope(&dst->sin6,
674 /* Fix IPv6 header payload length. */
675 if (m->m_len < sizeof(struct ip6_hdr))
676 if ((m = m_pullup(m,sizeof(struct ip6_hdr))) == NULL) {
681 if (m->m_pkthdr.len - sizeof(*ip6) > IPV6_MAXPACKET) {
682 /* No jumbogram support. */
683 error = ENXIO; /*XXX*/
687 /* Encapsulate the packet */
688 error = ipip_output(m, isr, &mp, 0, 0);
689 if (mp == NULL && !error) {
690 /* Should never happen. */
691 DPRINTF(("ipsec6_process_packet: ipip_output "
692 "returns no mbuf and no error!"));
699 /* XXX: Should never happen! */
702 m = NULL; /* ipip_output() already freed it */
711 ipsec_bpf(m, isr->sav, dst->sa.sa_family, ENC_OUT|ENC_AFTER);
712 /* pass the mbuf to enc0 for packet filtering */
713 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0)
717 switch(dst->sa.sa_family) {
722 ip = mtod(m, struct ip *);
724 off = offsetof(struct ip, ip_p);
729 i = sizeof(struct ip6_hdr);
730 off = offsetof(struct ip6_hdr, ip6_nxt);
733 DPRINTF(("%s: unsupported protocol family %u\n",
734 __func__, dst->sa.sa_family));
735 error = EPFNOSUPPORT;
736 IPSEC6STAT_INC(ips_out_inval);
739 error = (*sav->tdb_xform->xf_output)(m, isr, NULL, i, off);
740 IPSECREQUEST_UNLOCK(isr);
745 IPSECREQUEST_UNLOCK(isr);