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
26 * $FreeBSD: src/sys/netipsec/ipsec_output.c,v 1.3.2.2 2003/03/28 20:32:53 sam Exp $
30 * IPsec output processing.
33 #include "opt_inet6.h"
34 #include "opt_ipsec.h"
36 #include <sys/param.h>
37 #include <sys/systm.h>
39 #include <sys/domain.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/errno.h>
43 #include <sys/syslog.h>
44 #include <sys/in_cksum.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 <netproto/ipsec/ipsec.h>
70 #include <netproto/ipsec/ipsec6.h>
72 #include <netproto/ipsec/ah_var.h>
73 #include <netproto/ipsec/esp_var.h>
74 #include <netproto/ipsec/ipcomp_var.h>
76 #include <netproto/ipsec/xform.h>
78 #include <netproto/ipsec/key.h>
79 #include <netproto/ipsec/keydb.h>
80 #include <netproto/ipsec/key_debug.h>
83 ipsec_process_done(struct mbuf *m, struct ipsecrequest *isr)
85 struct tdb_ident *tdbi;
88 struct secasindex *saidx;
91 KASSERT(m != NULL, ("ipsec_process_done: null mbuf"));
92 KASSERT(isr != NULL, ("ipsec_process_done: null ISR"));
94 KASSERT(sav != NULL, ("ipsec_process_done: null SA"));
95 KASSERT(sav->sah != NULL, ("ipsec_process_done: null SAH"));
97 saidx = &sav->sah->saidx;
98 switch (saidx->dst.sa.sa_family) {
101 /* Fix the header length, for AH processing. */
102 mtod(m, struct ip *)->ip_len = htons(m->m_pkthdr.len);
107 /* Fix the header length, for AH processing. */
108 if (m->m_pkthdr.len < sizeof (struct ip6_hdr)) {
112 if (m->m_pkthdr.len - sizeof (struct ip6_hdr) > IPV6_MAXPACKET) {
113 /* No jumbogram support. */
117 mtod(m, struct ip6_hdr *)->ip6_plen =
118 htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
122 DPRINTF(("ipsec_process_done: unknown protocol family %u\n",
123 saidx->dst.sa.sa_family));
129 * Add a record of what we've done or what needs to be done to the
132 mtag = m_tag_get(PACKET_TAG_IPSEC_OUT_DONE,
133 sizeof(struct tdb_ident), M_NOWAIT);
135 DPRINTF(("ipsec_process_done: could not get packet tag\n"));
140 tdbi = (struct tdb_ident *)m_tag_data(mtag);
141 tdbi->dst = saidx->dst;
142 tdbi->proto = saidx->proto;
143 tdbi->spi = sav->spi;
144 m_tag_prepend(m, mtag);
147 * If there's another (bundled) SA to apply, do so.
148 * Note that this puts a burden on the kernel stack size.
149 * If this is a problem we'll need to introduce a queue
150 * to set the packet on so we can unwind the stack before
151 * doing further processing.
154 newipsecstat.ips_out_bundlesa++;
155 return ipsec4_process_packet(m, isr->next, 0, 0);
159 * We're done with IPsec processing, transmit the packet using the
160 * appropriate network protocol (IP or IPv6). SPD lookup will be
161 * performed again there.
163 switch (saidx->dst.sa.sa_family) {
167 ip = mtod(m, struct ip *);
168 ip->ip_len = ntohs(ip->ip_len);
169 ip->ip_off = ntohs(ip->ip_off);
171 return ip_output(m, NULL, NULL, IP_RAWOUTPUT, NULL, NULL);
176 * We don't need massage, IPv6 header fields are always in
179 return ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
182 panic("ipsec_process_done");
189 static struct ipsecrequest *
192 struct ipsecrequest *isr,
194 struct secasindex *saidx,
198 #define IPSEC_OSTAT(x,y,z) (isr->saidx.proto == IPPROTO_ESP ? (x)++ : \
199 isr->saidx.proto == IPPROTO_AH ? (y)++ : (z)++)
200 struct secasvar *sav;
202 KASSERT(af == AF_INET || af == AF_INET6,
203 ("ipsec_nextisr: invalid address family %u", af));
206 * Craft SA index to search for proper SA. Note that
207 * we only fillin unspecified SA peers for transport
208 * mode; for tunnel mode they must already be filled in.
211 if (isr->saidx.mode == IPSEC_MODE_TRANSPORT) {
212 /* Fillin unspecified SA peers only for transport mode */
214 struct sockaddr_in *sin;
215 struct ip *ip = mtod(m, struct ip *);
217 if (saidx->src.sa.sa_len == 0) {
218 sin = &saidx->src.sin;
219 sin->sin_len = sizeof(*sin);
220 sin->sin_family = AF_INET;
221 sin->sin_port = IPSEC_PORT_ANY;
222 sin->sin_addr = ip->ip_src;
224 if (saidx->dst.sa.sa_len == 0) {
225 sin = &saidx->dst.sin;
226 sin->sin_len = sizeof(*sin);
227 sin->sin_family = AF_INET;
228 sin->sin_port = IPSEC_PORT_ANY;
229 sin->sin_addr = ip->ip_dst;
232 struct sockaddr_in6 *sin6;
233 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
235 if (saidx->src.sin6.sin6_len == 0) {
236 sin6 = (struct sockaddr_in6 *)&saidx->src;
237 sin6->sin6_len = sizeof(*sin6);
238 sin6->sin6_family = AF_INET6;
239 sin6->sin6_port = IPSEC_PORT_ANY;
240 sin6->sin6_addr = ip6->ip6_src;
241 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
242 /* fix scope id for comparing SPD */
243 sin6->sin6_addr.s6_addr16[1] = 0;
244 sin6->sin6_scope_id =
245 ntohs(ip6->ip6_src.s6_addr16[1]);
248 if (saidx->dst.sin6.sin6_len == 0) {
249 sin6 = (struct sockaddr_in6 *)&saidx->dst;
250 sin6->sin6_len = sizeof(*sin6);
251 sin6->sin6_family = AF_INET6;
252 sin6->sin6_port = IPSEC_PORT_ANY;
253 sin6->sin6_addr = ip6->ip6_dst;
254 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
255 /* fix scope id for comparing SPD */
256 sin6->sin6_addr.s6_addr16[1] = 0;
257 sin6->sin6_scope_id =
258 ntohs(ip6->ip6_dst.s6_addr16[1]);
265 * Lookup SA and validate it.
267 *error = key_checkrequest(isr, saidx);
270 * IPsec processing is required, but no SA found.
271 * I assume that key_acquire() had been called
272 * to get/establish the SA. Here I discard
273 * this packet because it is responsibility for
274 * upper layer to retransmit the packet.
276 newipsecstat.ips_out_nosa++;
280 if (sav == NULL) { /* XXX valid return */
281 KASSERT(ipsec_get_reqlevel(isr) == IPSEC_LEVEL_USE,
282 ("ipsec_nextisr: no SA found, but required; level %u",
283 ipsec_get_reqlevel(isr)));
287 *error = EINVAL; /*XXX*/
294 * Check system global policy controls.
296 if ((isr->saidx.proto == IPPROTO_ESP && !esp_enable) ||
297 (isr->saidx.proto == IPPROTO_AH && !ah_enable) ||
298 (isr->saidx.proto == IPPROTO_IPCOMP && !ipcomp_enable)) {
299 DPRINTF(("ipsec_nextisr: IPsec outbound packet dropped due"
300 " to policy (check your sysctls)\n"));
301 IPSEC_OSTAT(espstat.esps_pdrops, ahstat.ahs_pdrops,
302 ipcompstat.ipcomps_pdrops);
303 *error = EHOSTUNREACH;
308 * Sanity check the SA contents for the caller
309 * before they invoke the xform output method.
311 if (sav->tdb_xform == NULL) {
312 DPRINTF(("ipsec_nextisr: no transform for SA\n"));
313 IPSEC_OSTAT(espstat.esps_noxform, ahstat.ahs_noxform,
314 ipcompstat.ipcomps_noxform);
315 *error = EHOSTUNREACH;
320 KASSERT(*error != 0, ("ipsec_nextisr: error return w/ no error code"));
327 * IPsec output logic for IPv4.
330 ipsec4_process_packet(
332 struct ipsecrequest *isr,
336 struct secasindex saidx;
337 struct secasvar *sav;
341 KASSERT(m != NULL, ("ipsec4_process_packet: null mbuf"));
342 KASSERT(isr != NULL, ("ipsec4_process_packet: null isr"));
346 isr = ipsec_nextisr(m, isr, AF_INET, &saidx, &error);
352 union sockaddr_union *dst = &sav->sah->saidx.dst;
356 * Collect IP_DF state from the outer header.
358 if (dst->sa.sa_family == AF_INET) {
359 if (m->m_len < sizeof (struct ip) &&
360 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
364 ip = mtod(m, struct ip *);
365 /* Honor system-wide control of how to handle IP_DF */
366 switch (ip4_ipsec_dfbit) {
367 case 0: /* clear in outer header */
368 case 1: /* set in outer header */
369 setdf = ip4_ipsec_dfbit;
371 default: /* propagate to outer header */
372 setdf = ntohs(ip->ip_off & IP_DF);
376 ip = NULL; /* keep compiler happy */
379 /* Do the appropriate encapsulation, if necessary */
380 if (isr->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
381 dst->sa.sa_family != AF_INET || /* PF mismatch */
383 (sav->flags & SADB_X_SAFLAGS_TUNNEL) || /* Tunnel requ'd */
384 sav->tdb_xform->xf_type == XF_IP4 || /* ditto */
386 (dst->sa.sa_family == AF_INET && /* Proxy */
387 dst->sin.sin_addr.s_addr != INADDR_ANY &&
388 dst->sin.sin_addr.s_addr != ip->ip_dst.s_addr)) {
391 /* Fix IPv4 header checksum and length */
392 if (m->m_len < sizeof (struct ip) &&
393 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
397 ip = mtod(m, struct ip *);
398 ip->ip_len = htons(m->m_pkthdr.len);
401 if (ip->ip_vhl == IP_VHL_BORING)
402 ip->ip_sum = in_cksum_hdr(ip);
404 ip->ip_sum = in_cksum(m,
405 _IP_VHL_HL(ip->ip_vhl) << 2);
407 ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
410 /* Encapsulate the packet */
411 error = ipip_output(m, isr, &mp, 0, 0);
412 if (mp == NULL && !error) {
413 /* Should never happen. */
414 DPRINTF(("ipsec4_process_packet: ipip_output "
415 "returns no mbuf and no error!"));
425 * ipip_output clears IP_DF in the new header. If
426 * we need to propagate IP_DF from the outer header,
427 * then we have to do it here.
429 * XXX shouldn't assume what ipip_output does.
431 if (dst->sa.sa_family == AF_INET && setdf) {
432 if (m->m_len < sizeof (struct ip) &&
433 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
437 ip = mtod(m, struct ip *);
438 ip->ip_off = ntohs(ip->ip_off);
440 ip->ip_off = htons(ip->ip_off);
446 * Dispatch to the appropriate IPsec transform logic. The
447 * packet will be returned for transmission after crypto
448 * processing, etc. are completed. For encapsulation we
449 * bypass this call because of the explicit call done above
450 * (necessary to deal with IP_DF handling for IPv4).
452 * NB: m & sav are ``passed to caller'' who's reponsible for
453 * for reclaiming their resources.
455 if (sav->tdb_xform->xf_type != XF_IP4) {
456 ip = mtod(m, struct ip *);
458 off = offsetof(struct ip, ip_p);
459 error = (*sav->tdb_xform->xf_output)(m, isr, NULL, i, off);
461 error = ipsec_process_done(m, isr);
475 * Chop IP6 header from the payload.
478 ipsec6_splithdr(struct mbuf *m)
484 KASSERT(m->m_len >= sizeof (struct ip6_hdr),
485 ("ipsec6_splithdr: first mbuf too short, len %u", m->m_len));
486 ip6 = mtod(m, struct ip6_hdr *);
487 hlen = sizeof(struct ip6_hdr);
488 if (m->m_len > hlen) {
489 MGETHDR(mh, M_NOWAIT, MT_HEADER);
494 M_MOVE_PKTHDR(mh, m);
501 bcopy((caddr_t)ip6, mtod(m, caddr_t), hlen);
502 } else if (m->m_len < hlen) {
503 m = m_pullup(m, hlen);
511 * IPsec output logic for IPv6, transport mode.
515 struct ipsec_output_state *state,
518 struct secpolicy *sp,
522 struct ipsecrequest *isr;
523 struct secasindex saidx;
527 KASSERT(state != NULL, ("ipsec6_output: null state"));
528 KASSERT(state->m != NULL, ("ipsec6_output: null m"));
529 KASSERT(nexthdrp != NULL, ("ipsec6_output: null nexthdrp"));
530 KASSERT(mprev != NULL, ("ipsec6_output: null mprev"));
531 KASSERT(sp != NULL, ("ipsec6_output: null sp"));
532 KASSERT(tun != NULL, ("ipsec6_output: null tun"));
534 KEYDEBUG(KEYDEBUG_IPSEC_DATA,
535 kprintf("ipsec6_output_trans: applyed SP\n");
536 kdebug_secpolicy(sp));
539 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
540 /* the rest will be handled by ipsec6_output_tunnel() */
541 *tun = 1; /* need tunnel-mode processing */
548 isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
551 /* XXX should notification be done for all errors ? */
553 * Notify the fact that the packet is discarded
554 * to ourselves. I believe this is better than
555 * just silently discarding. (jinmei@kame.net)
556 * XXX: should we restrict the error to TCP packets?
557 * XXX: should we directly notify sockets via
560 icmp6_error(m, ICMP6_DST_UNREACH,
561 ICMP6_DST_UNREACH_ADMIN, 0);
562 m = NULL; /* NB: icmp6_error frees mbuf */
567 return (*isr->sav->tdb_xform->xf_output)(m, isr, NULL,
568 sizeof (struct ip6_hdr),
569 offsetof(struct ip6_hdr, ip6_nxt));
578 ipsec6_encapsulate(struct mbuf *m, struct secasvar *sav)
580 struct ip6_hdr *oip6;
584 /* can't tunnel between different AFs */
585 if (sav->sah->saidx.src.sa.sa_family != AF_INET6 ||
586 sav->sah->saidx.dst.sa.sa_family != AF_INET6) {
590 KASSERT(m->m_len != sizeof (struct ip6_hdr),
591 ("ipsec6_encapsulate: mbuf wrong size; len %u", m->m_len));
595 * grow the mbuf to accomodate the new IPv6 header.
597 plen = m->m_pkthdr.len;
598 if (M_LEADINGSPACE(m->m_next) < sizeof(struct ip6_hdr)) {
600 MGET(n, M_NOWAIT, MT_DATA);
605 n->m_len = sizeof(struct ip6_hdr);
606 n->m_next = m->m_next;
608 m->m_pkthdr.len += sizeof(struct ip6_hdr);
609 oip6 = mtod(n, struct ip6_hdr *);
611 m->m_next->m_len += sizeof(struct ip6_hdr);
612 m->m_next->m_data -= sizeof(struct ip6_hdr);
613 m->m_pkthdr.len += sizeof(struct ip6_hdr);
614 oip6 = mtod(m->m_next, struct ip6_hdr *);
616 ip6 = mtod(m, struct ip6_hdr *);
617 ovbcopy((caddr_t)ip6, (caddr_t)oip6, sizeof(struct ip6_hdr));
619 /* Fake link-local scope-class addresses */
620 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
621 oip6->ip6_src.s6_addr16[1] = 0;
622 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
623 oip6->ip6_dst.s6_addr16[1] = 0;
625 /* construct new IPv6 header. see RFC 2401 5.1.2.2 */
626 /* ECN consideration. */
627 ip6_ecn_ingress(ip6_ipsec_ecn, &ip6->ip6_flow, &oip6->ip6_flow);
628 if (plen < IPV6_MAXPACKET - sizeof(struct ip6_hdr))
629 ip6->ip6_plen = htons(plen);
631 /* ip6->ip6_plen will be updated in ip6_output() */
633 ip6->ip6_nxt = IPPROTO_IPV6;
634 sav->sah->saidx.src.sin6.sin6_addr = ip6->ip6_src;
635 sav->sah->saidx.dst.sin6.sin6_addr = ip6->ip6_dst;
636 ip6->ip6_hlim = IPV6_DEFHLIM;
638 /* XXX Should ip6_src be updated later ? */
644 * IPsec output logic for IPv6, tunnel mode.
647 ipsec6_output_tunnel(struct ipsec_output_state *state, struct secpolicy *sp, int flags)
650 struct ipsecrequest *isr;
651 struct secasindex saidx;
653 struct sockaddr_in6* dst6;
656 KASSERT(state != NULL, ("ipsec6_output: null state"));
657 KASSERT(state->m != NULL, ("ipsec6_output: null m"));
658 KASSERT(sp != NULL, ("ipsec6_output: null sp"));
660 KEYDEBUG(KEYDEBUG_IPSEC_DATA,
661 kprintf("ipsec6_output_tunnel: applyed SP\n");
662 kdebug_secpolicy(sp));
666 * transport mode ipsec (before the 1st tunnel mode) is already
667 * processed by ipsec6_output_trans().
669 for (isr = sp->req; isr; isr = isr->next) {
670 if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
673 isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
678 * There may be the case that SA status will be changed when
679 * we are refering to one. So calling splsoftnet().
681 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
683 * build IPsec tunnel.
685 /* XXX should be processed with other familiy */
686 if (isr->sav->sah->saidx.src.sa.sa_family != AF_INET6) {
687 ipseclog((LOG_ERR, "ipsec6_output_tunnel: "
688 "family mismatched between inner and outer, spi=%u\n",
689 ntohl(isr->sav->spi)));
690 newipsecstat.ips_out_inval++;
691 error = EAFNOSUPPORT;
695 m = ipsec6_splithdr(m);
697 newipsecstat.ips_out_nomem++;
701 error = ipsec6_encapsulate(m, isr->sav);
706 ip6 = mtod(m, struct ip6_hdr *);
708 state->ro = &isr->sav->sah->sa_route;
709 state->dst = (struct sockaddr *)&state->ro->ro_dst;
710 dst6 = (struct sockaddr_in6 *)state->dst;
712 && ((state->ro->ro_rt->rt_flags & RTF_UP) == 0
713 || !IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr, &ip6->ip6_dst))) {
714 RTFREE(state->ro->ro_rt);
715 state->ro->ro_rt = NULL;
717 if (state->ro->ro_rt == 0) {
718 bzero(dst6, sizeof(*dst6));
719 dst6->sin6_family = AF_INET6;
720 dst6->sin6_len = sizeof(*dst6);
721 dst6->sin6_addr = ip6->ip6_dst;
724 if (state->ro->ro_rt == 0) {
725 ip6stat.ip6s_noroute++;
726 newipsecstat.ips_out_noroute++;
727 error = EHOSTUNREACH;
731 /* adjust state->dst if tunnel endpoint is offlink */
732 if (state->ro->ro_rt->rt_flags & RTF_GATEWAY) {
733 state->dst = (struct sockaddr *)state->ro->ro_rt->rt_gateway;
734 dst6 = (struct sockaddr_in6 *)state->dst;
738 m = ipsec6_splithdr(m);
740 newipsecstat.ips_out_nomem++;
744 ip6 = mtod(m, struct ip6_hdr *);
745 return (*isr->sav->tdb_xform->xf_output)(m, isr, NULL,
746 sizeof (struct ip6_hdr),
747 offsetof(struct ip6_hdr, ip6_nxt));