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/route.h>
49 #include <netinet/in.h>
50 #include <netinet/in_systm.h>
51 #include <netinet/ip.h>
52 #include <netinet/ip_var.h>
53 #include <netinet/in_var.h>
54 #include <netinet/ip_ecn.h>
56 #include <netinet6/ip6_ecn.h>
59 #include <netinet/ip6.h>
61 #include <netinet6/ip6_var.h>
63 #include <netinet/in_pcb.h>
65 #include <netinet/icmp6.h>
68 #include <netipsec/ipsec.h>
70 #include <netipsec/ipsec6.h>
72 #include <netipsec/ah_var.h>
73 #include <netipsec/esp_var.h>
74 #include <netipsec/ipcomp_var.h>
76 #include <netipsec/xform.h>
78 #include <netipsec/key.h>
79 #include <netipsec/keydb.h>
80 #include <netipsec/key_debug.h>
82 #include <machine/in_cksum.h>
85 ipsec_process_done(struct mbuf *m, struct ipsecrequest *isr)
87 struct tdb_ident *tdbi;
90 struct secasindex *saidx;
93 IPSEC_SPLASSERT_SOFTNET(__func__);
95 IPSEC_ASSERT(m != NULL, ("null mbuf"));
96 IPSEC_ASSERT(isr != NULL, ("null ISR"));
98 IPSEC_ASSERT(sav != NULL, ("null SA"));
99 IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
101 saidx = &sav->sah->saidx;
102 switch (saidx->dst.sa.sa_family) {
105 /* Fix the header length, for AH processing. */
106 mtod(m, struct ip *)->ip_len = htons(m->m_pkthdr.len);
111 /* Fix the header length, for AH processing. */
112 if (m->m_pkthdr.len < sizeof (struct ip6_hdr)) {
116 if (m->m_pkthdr.len - sizeof (struct ip6_hdr) > IPV6_MAXPACKET) {
117 /* No jumbogram support. */
121 mtod(m, struct ip6_hdr *)->ip6_plen =
122 htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
126 DPRINTF(("%s: unknown protocol family %u\n", __func__,
127 saidx->dst.sa.sa_family));
133 * Add a record of what we've done or what needs to be done to the
136 mtag = m_tag_get(PACKET_TAG_IPSEC_OUT_DONE,
137 sizeof(struct tdb_ident), M_NOWAIT);
139 DPRINTF(("%s: could not get packet tag\n", __func__));
144 tdbi = (struct tdb_ident *)(mtag + 1);
145 tdbi->dst = saidx->dst;
146 tdbi->proto = saidx->proto;
147 tdbi->spi = sav->spi;
148 m_tag_prepend(m, mtag);
151 * If there's another (bundled) SA to apply, do so.
152 * Note that this puts a burden on the kernel stack size.
153 * If this is a problem we'll need to introduce a queue
154 * to set the packet on so we can unwind the stack before
155 * doing further processing.
158 newipsecstat.ips_out_bundlesa++;
159 return ipsec4_process_packet(m, isr->next, 0, 0);
161 key_sa_recordxfer(sav, m); /* record data transfer */
164 * We're done with IPsec processing, transmit the packet using the
165 * appropriate network protocol (IP or IPv6). SPD lookup will be
166 * performed again there.
168 switch (saidx->dst.sa.sa_family) {
172 ip = mtod(m, struct ip *);
173 ip->ip_len = ntohs(ip->ip_len);
174 ip->ip_off = ntohs(ip->ip_off);
176 return ip_output(m, NULL, NULL, IP_RAWOUTPUT, NULL, NULL);
181 * We don't need massage, IPv6 header fields are always in
184 return ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
187 panic("ipsec_process_done");
194 static struct ipsecrequest *
197 struct ipsecrequest *isr,
199 struct secasindex *saidx,
203 #define IPSEC_OSTAT(x,y,z) (isr->saidx.proto == IPPROTO_ESP ? (x)++ : \
204 isr->saidx.proto == IPPROTO_AH ? (y)++ : (z)++)
205 struct secasvar *sav;
207 IPSEC_SPLASSERT_SOFTNET(__func__);
208 IPSECREQUEST_LOCK_ASSERT(isr);
210 IPSEC_ASSERT(af == AF_INET || af == AF_INET6,
211 ("invalid address family %u", af));
214 * Craft SA index to search for proper SA. Note that
215 * we only fillin unspecified SA peers for transport
216 * mode; for tunnel mode they must already be filled in.
219 if (isr->saidx.mode == IPSEC_MODE_TRANSPORT) {
220 /* Fillin unspecified SA peers only for transport mode */
222 struct sockaddr_in *sin;
223 struct ip *ip = mtod(m, struct ip *);
225 if (saidx->src.sa.sa_len == 0) {
226 sin = &saidx->src.sin;
227 sin->sin_len = sizeof(*sin);
228 sin->sin_family = AF_INET;
229 sin->sin_port = IPSEC_PORT_ANY;
230 sin->sin_addr = ip->ip_src;
232 if (saidx->dst.sa.sa_len == 0) {
233 sin = &saidx->dst.sin;
234 sin->sin_len = sizeof(*sin);
235 sin->sin_family = AF_INET;
236 sin->sin_port = IPSEC_PORT_ANY;
237 sin->sin_addr = ip->ip_dst;
240 struct sockaddr_in6 *sin6;
241 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
243 if (saidx->src.sin6.sin6_len == 0) {
244 sin6 = (struct sockaddr_in6 *)&saidx->src;
245 sin6->sin6_len = sizeof(*sin6);
246 sin6->sin6_family = AF_INET6;
247 sin6->sin6_port = IPSEC_PORT_ANY;
248 sin6->sin6_addr = ip6->ip6_src;
249 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
250 /* fix scope id for comparing SPD */
251 sin6->sin6_addr.s6_addr16[1] = 0;
252 sin6->sin6_scope_id =
253 ntohs(ip6->ip6_src.s6_addr16[1]);
256 if (saidx->dst.sin6.sin6_len == 0) {
257 sin6 = (struct sockaddr_in6 *)&saidx->dst;
258 sin6->sin6_len = sizeof(*sin6);
259 sin6->sin6_family = AF_INET6;
260 sin6->sin6_port = IPSEC_PORT_ANY;
261 sin6->sin6_addr = ip6->ip6_dst;
262 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
263 /* fix scope id for comparing SPD */
264 sin6->sin6_addr.s6_addr16[1] = 0;
265 sin6->sin6_scope_id =
266 ntohs(ip6->ip6_dst.s6_addr16[1]);
273 * Lookup SA and validate it.
275 *error = key_checkrequest(isr, saidx);
278 * IPsec processing is required, but no SA found.
279 * I assume that key_acquire() had been called
280 * to get/establish the SA. Here I discard
281 * this packet because it is responsibility for
282 * upper layer to retransmit the packet.
284 newipsecstat.ips_out_nosa++;
288 if (sav == NULL) { /* XXX valid return */
289 IPSEC_ASSERT(ipsec_get_reqlevel(isr) == IPSEC_LEVEL_USE,
290 ("no SA found, but required; level %u",
291 ipsec_get_reqlevel(isr)));
292 IPSECREQUEST_UNLOCK(isr);
296 *error = EINVAL; /*XXX*/
299 IPSECREQUEST_LOCK(isr);
304 * Check system global policy controls.
306 if ((isr->saidx.proto == IPPROTO_ESP && !esp_enable) ||
307 (isr->saidx.proto == IPPROTO_AH && !ah_enable) ||
308 (isr->saidx.proto == IPPROTO_IPCOMP && !ipcomp_enable)) {
309 DPRINTF(("%s: IPsec outbound packet dropped due"
310 " to policy (check your sysctls)\n", __func__));
311 IPSEC_OSTAT(espstat.esps_pdrops, ahstat.ahs_pdrops,
312 ipcompstat.ipcomps_pdrops);
313 *error = EHOSTUNREACH;
318 * Sanity check the SA contents for the caller
319 * before they invoke the xform output method.
321 if (sav->tdb_xform == NULL) {
322 DPRINTF(("%s: no transform for SA\n", __func__));
323 IPSEC_OSTAT(espstat.esps_noxform, ahstat.ahs_noxform,
324 ipcompstat.ipcomps_noxform);
325 *error = EHOSTUNREACH;
330 IPSEC_ASSERT(*error != 0, ("error return w/ no error code"));
331 IPSECREQUEST_UNLOCK(isr);
338 * IPsec output logic for IPv4.
341 ipsec4_process_packet(
343 struct ipsecrequest *isr,
347 struct secasindex saidx;
348 struct secasvar *sav;
352 IPSEC_ASSERT(m != NULL, ("null mbuf"));
353 IPSEC_ASSERT(isr != NULL, ("null isr"));
355 IPSECREQUEST_LOCK(isr); /* insure SA contents don't change */
357 isr = ipsec_nextisr(m, isr, AF_INET, &saidx, &error);
364 /* pass the mbuf to enc0 for packet filtering */
365 if ((error = ipsec_filter(&m, 2)) != 0)
370 union sockaddr_union *dst = &sav->sah->saidx.dst;
374 * Collect IP_DF state from the outer header.
376 if (dst->sa.sa_family == AF_INET) {
377 if (m->m_len < sizeof (struct ip) &&
378 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
382 ip = mtod(m, struct ip *);
383 /* Honor system-wide control of how to handle IP_DF */
384 switch (ip4_ipsec_dfbit) {
385 case 0: /* clear in outer header */
386 case 1: /* set in outer header */
387 setdf = ip4_ipsec_dfbit;
389 default: /* propagate to outer header */
390 setdf = ntohs(ip->ip_off & IP_DF);
394 ip = NULL; /* keep compiler happy */
397 /* Do the appropriate encapsulation, if necessary */
398 if (isr->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
399 dst->sa.sa_family != AF_INET || /* PF mismatch */
401 (sav->flags & SADB_X_SAFLAGS_TUNNEL) || /* Tunnel requ'd */
402 sav->tdb_xform->xf_type == XF_IP4 || /* ditto */
404 (dst->sa.sa_family == AF_INET && /* Proxy */
405 dst->sin.sin_addr.s_addr != INADDR_ANY &&
406 dst->sin.sin_addr.s_addr != ip->ip_dst.s_addr)) {
409 /* Fix IPv4 header checksum and length */
410 if (m->m_len < sizeof (struct ip) &&
411 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
415 ip = mtod(m, struct ip *);
416 ip->ip_len = htons(m->m_pkthdr.len);
419 if (ip->ip_vhl == IP_VHL_BORING)
420 ip->ip_sum = in_cksum_hdr(ip);
422 ip->ip_sum = in_cksum(m,
423 _IP_VHL_HL(ip->ip_vhl) << 2);
425 ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
428 /* Encapsulate the packet */
429 error = ipip_output(m, isr, &mp, 0, 0);
430 if (mp == NULL && !error) {
431 /* Should never happen. */
432 DPRINTF(("%s: ipip_output returns no mbuf and "
433 "no error!", __func__));
438 /* XXX: Should never happen! */
441 m = NULL; /* ipip_output() already freed it */
446 * ipip_output clears IP_DF in the new header. If
447 * we need to propagate IP_DF from the outer header,
448 * then we have to do it here.
450 * XXX shouldn't assume what ipip_output does.
452 if (dst->sa.sa_family == AF_INET && setdf) {
453 if (m->m_len < sizeof (struct ip) &&
454 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
458 ip = mtod(m, struct ip *);
459 ip->ip_off = ntohs(ip->ip_off);
461 ip->ip_off = htons(ip->ip_off);
467 /* pass the mbuf to enc0 for bpf processing */
468 ipsec_bpf(m, sav, AF_INET);
472 * Dispatch to the appropriate IPsec transform logic. The
473 * packet will be returned for transmission after crypto
474 * processing, etc. are completed. For encapsulation we
475 * bypass this call because of the explicit call done above
476 * (necessary to deal with IP_DF handling for IPv4).
478 * NB: m & sav are ``passed to caller'' who's reponsible for
479 * for reclaiming their resources.
481 if (sav->tdb_xform->xf_type != XF_IP4) {
482 ip = mtod(m, struct ip *);
484 off = offsetof(struct ip, ip_p);
485 error = (*sav->tdb_xform->xf_output)(m, isr, NULL, i, off);
487 error = ipsec_process_done(m, isr);
489 IPSECREQUEST_UNLOCK(isr);
493 IPSECREQUEST_UNLOCK(isr);
502 * Chop IP6 header from the payload.
505 ipsec6_splithdr(struct mbuf *m)
511 IPSEC_ASSERT(m->m_len >= sizeof (struct ip6_hdr),
512 ("first mbuf too short, len %u", m->m_len));
513 ip6 = mtod(m, struct ip6_hdr *);
514 hlen = sizeof(struct ip6_hdr);
515 if (m->m_len > hlen) {
516 MGETHDR(mh, M_DONTWAIT, MT_HEADER);
521 M_MOVE_PKTHDR(mh, m);
528 bcopy((caddr_t)ip6, mtod(m, caddr_t), hlen);
529 } else if (m->m_len < hlen) {
530 m = m_pullup(m, hlen);
538 * IPsec output logic for IPv6, transport mode.
542 struct ipsec_output_state *state,
545 struct secpolicy *sp,
549 struct ipsecrequest *isr;
550 struct secasindex saidx;
554 IPSEC_ASSERT(state != NULL, ("null state"));
555 IPSEC_ASSERT(state->m != NULL, ("null m"));
556 IPSEC_ASSERT(nexthdrp != NULL, ("null nexthdrp"));
557 IPSEC_ASSERT(mprev != NULL, ("null mprev"));
558 IPSEC_ASSERT(sp != NULL, ("null sp"));
559 IPSEC_ASSERT(tun != NULL, ("null tun"));
561 KEYDEBUG(KEYDEBUG_IPSEC_DATA,
562 printf("%s: applyed SP\n", __func__);
563 kdebug_secpolicy(sp));
566 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
567 /* the rest will be handled by ipsec6_output_tunnel() */
568 *tun = 1; /* need tunnel-mode processing */
575 isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
578 /* XXX should notification be done for all errors ? */
580 * Notify the fact that the packet is discarded
581 * to ourselves. I believe this is better than
582 * just silently discarding. (jinmei@kame.net)
583 * XXX: should we restrict the error to TCP packets?
584 * XXX: should we directly notify sockets via
587 icmp6_error(m, ICMP6_DST_UNREACH,
588 ICMP6_DST_UNREACH_ADMIN, 0);
589 m = NULL; /* NB: icmp6_error frees mbuf */
594 return (*isr->sav->tdb_xform->xf_output)(m, isr, NULL,
595 sizeof (struct ip6_hdr),
596 offsetof(struct ip6_hdr, ip6_nxt));
605 ipsec6_encapsulate(struct mbuf *m, struct secasvar *sav)
607 struct ip6_hdr *oip6;
611 /* can't tunnel between different AFs */
612 if (sav->sah->saidx.src.sa.sa_family != AF_INET6 ||
613 sav->sah->saidx.dst.sa.sa_family != AF_INET6) {
617 IPSEC_ASSERT(m->m_len != sizeof (struct ip6_hdr),
618 ("mbuf wrong size; len %u", m->m_len));
622 * grow the mbuf to accomodate the new IPv6 header.
624 plen = m->m_pkthdr.len;
625 if (M_LEADINGSPACE(m->m_next) < sizeof(struct ip6_hdr)) {
627 MGET(n, M_DONTWAIT, MT_DATA);
632 n->m_len = sizeof(struct ip6_hdr);
633 n->m_next = m->m_next;
635 m->m_pkthdr.len += sizeof(struct ip6_hdr);
636 oip6 = mtod(n, struct ip6_hdr *);
638 m->m_next->m_len += sizeof(struct ip6_hdr);
639 m->m_next->m_data -= sizeof(struct ip6_hdr);
640 m->m_pkthdr.len += sizeof(struct ip6_hdr);
641 oip6 = mtod(m->m_next, struct ip6_hdr *);
643 ip6 = mtod(m, struct ip6_hdr *);
644 bcopy((caddr_t)ip6, (caddr_t)oip6, sizeof(struct ip6_hdr));
646 /* Fake link-local scope-class addresses */
647 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
648 oip6->ip6_src.s6_addr16[1] = 0;
649 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
650 oip6->ip6_dst.s6_addr16[1] = 0;
652 /* construct new IPv6 header. see RFC 2401 5.1.2.2 */
653 /* ECN consideration. */
654 ip6_ecn_ingress(ip6_ipsec_ecn, &ip6->ip6_flow, &oip6->ip6_flow);
655 if (plen < IPV6_MAXPACKET - sizeof(struct ip6_hdr))
656 ip6->ip6_plen = htons(plen);
658 /* ip6->ip6_plen will be updated in ip6_output() */
660 ip6->ip6_nxt = IPPROTO_IPV6;
661 sav->sah->saidx.src.sin6.sin6_addr = ip6->ip6_src;
662 sav->sah->saidx.dst.sin6.sin6_addr = ip6->ip6_dst;
663 ip6->ip6_hlim = IPV6_DEFHLIM;
665 /* XXX Should ip6_src be updated later ? */
671 * IPsec output logic for IPv6, tunnel mode.
674 ipsec6_output_tunnel(struct ipsec_output_state *state, struct secpolicy *sp, int flags)
677 struct ipsecrequest *isr;
678 struct secasindex saidx;
680 struct sockaddr_in6* dst6;
683 IPSEC_ASSERT(state != NULL, ("null state"));
684 IPSEC_ASSERT(state->m != NULL, ("null m"));
685 IPSEC_ASSERT(sp != NULL, ("null sp"));
687 KEYDEBUG(KEYDEBUG_IPSEC_DATA,
688 printf("%s: applyed SP\n", __func__);
689 kdebug_secpolicy(sp));
693 * transport mode ipsec (before the 1st tunnel mode) is already
694 * processed by ipsec6_output_trans().
696 for (isr = sp->req; isr; isr = isr->next) {
697 if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
700 isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
705 * There may be the case that SA status will be changed when
706 * we are refering to one. So calling splsoftnet().
708 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
710 * build IPsec tunnel.
712 /* XXX should be processed with other familiy */
713 if (isr->sav->sah->saidx.src.sa.sa_family != AF_INET6) {
714 ipseclog((LOG_ERR, "%s: family mismatched between "
715 "inner and outer, spi=%u\n", __func__,
716 ntohl(isr->sav->spi)));
717 newipsecstat.ips_out_inval++;
718 error = EAFNOSUPPORT;
722 m = ipsec6_splithdr(m);
724 newipsecstat.ips_out_nomem++;
728 error = ipsec6_encapsulate(m, isr->sav);
733 ip6 = mtod(m, struct ip6_hdr *);
735 state->ro = &isr->sav->sah->sa_route;
736 state->dst = (struct sockaddr *)&state->ro->ro_dst;
737 dst6 = (struct sockaddr_in6 *)state->dst;
739 && ((state->ro->ro_rt->rt_flags & RTF_UP) == 0
740 || !IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr, &ip6->ip6_dst))) {
741 RTFREE(state->ro->ro_rt);
742 state->ro->ro_rt = NULL;
744 if (state->ro->ro_rt == 0) {
745 bzero(dst6, sizeof(*dst6));
746 dst6->sin6_family = AF_INET6;
747 dst6->sin6_len = sizeof(*dst6);
748 dst6->sin6_addr = ip6->ip6_dst;
751 if (state->ro->ro_rt == 0) {
752 ip6stat.ip6s_noroute++;
753 newipsecstat.ips_out_noroute++;
754 error = EHOSTUNREACH;
758 /* adjust state->dst if tunnel endpoint is offlink */
759 if (state->ro->ro_rt->rt_flags & RTF_GATEWAY) {
760 state->dst = (struct sockaddr *)state->ro->ro_rt->rt_gateway;
761 dst6 = (struct sockaddr_in6 *)state->dst;
765 m = ipsec6_splithdr(m);
767 newipsecstat.ips_out_nomem++;
771 ip6 = mtod(m, struct ip6_hdr *);
772 return (*isr->sav->tdb_xform->xf_output)(m, isr, NULL,
773 sizeof (struct ip6_hdr),
774 offsetof(struct ip6_hdr, ip6_nxt));