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 IPSECSTAT_INC(ips_out_bundlesa);
168 /* XXX-BZ currently only support same AF bundles. */
169 switch (saidx->dst.sa.sa_family) {
172 return ipsec4_process_packet(m, isr->next, 0, 0);
179 return ipsec6_process_packet(m, isr->next);
184 DPRINTF(("%s: unknown protocol family %u\n", __func__,
185 saidx->dst.sa.sa_family));
190 key_sa_recordxfer(sav, m); /* record data transfer */
193 * We're done with IPsec processing, transmit the packet using the
194 * appropriate network protocol (IP or IPv6). SPD lookup will be
195 * performed again there.
197 switch (saidx->dst.sa.sa_family) {
202 * If NAT-T is enabled, now that all IPsec processing is done
203 * insert UDP encapsulation header after IP header.
205 if (sav->natt_type) {
206 struct ip *ip = mtod(m, struct ip *);
207 const int hlen = (ip->ip_hl << 2);
212 size = sizeof(struct udphdr);
213 if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE) {
215 * draft-ietf-ipsec-nat-t-ike-0[01].txt and
216 * draft-ietf-ipsec-udp-encaps-(00/)01.txt,
217 * ignoring possible AH mode
218 * non-IKE marker + non-ESP marker
219 * from draft-ietf-ipsec-udp-encaps-00.txt.
221 size += sizeof(u_int64_t);
223 mi = m_makespace(m, hlen, size, &off);
225 DPRINTF(("%s: m_makespace for udphdr failed\n",
231 udp = (struct udphdr *)(mtod(mi, caddr_t) + off);
232 if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE)
233 udp->uh_sport = htons(UDP_ENCAP_ESPINUDP_PORT);
236 KEY_PORTFROMSADDR(&sav->sah->saidx.src);
237 udp->uh_dport = KEY_PORTFROMSADDR(&sav->sah->saidx.dst);
239 udp->uh_ulen = htons(m->m_pkthdr.len - hlen);
240 ip->ip_len = htons(m->m_pkthdr.len);
241 ip->ip_p = IPPROTO_UDP;
243 if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE)
244 *(u_int64_t *)(udp + 1) = 0;
246 #endif /* IPSEC_NAT_T */
248 return ip_output(m, NULL, NULL, IP_RAWOUTPUT, NULL, NULL);
253 * We don't need massage, IPv6 header fields are always in
256 return ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
259 panic("ipsec_process_done");
265 static struct ipsecrequest *
268 struct ipsecrequest *isr,
270 struct secasindex *saidx,
274 #define IPSEC_OSTAT(name) do { \
275 if (isr->saidx.proto == IPPROTO_ESP) \
276 ESPSTAT_INC(esps_##name); \
277 else if (isr->saidx.proto == IPPROTO_AH)\
278 AHSTAT_INC(ahs_##name); \
280 IPCOMPSTAT_INC(ipcomps_##name); \
282 struct secasvar *sav;
284 IPSECREQUEST_LOCK_ASSERT(isr);
286 IPSEC_ASSERT(af == AF_INET || af == AF_INET6,
287 ("invalid address family %u", af));
290 * Craft SA index to search for proper SA. Note that
291 * we only fillin unspecified SA peers for transport
292 * mode; for tunnel mode they must already be filled in.
295 if (isr->saidx.mode == IPSEC_MODE_TRANSPORT) {
296 /* Fillin unspecified SA peers only for transport mode */
298 struct sockaddr_in *sin;
299 struct ip *ip = mtod(m, struct ip *);
301 if (saidx->src.sa.sa_len == 0) {
302 sin = &saidx->src.sin;
303 sin->sin_len = sizeof(*sin);
304 sin->sin_family = AF_INET;
305 sin->sin_port = IPSEC_PORT_ANY;
306 sin->sin_addr = ip->ip_src;
308 if (saidx->dst.sa.sa_len == 0) {
309 sin = &saidx->dst.sin;
310 sin->sin_len = sizeof(*sin);
311 sin->sin_family = AF_INET;
312 sin->sin_port = IPSEC_PORT_ANY;
313 sin->sin_addr = ip->ip_dst;
316 struct sockaddr_in6 *sin6;
317 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
319 if (saidx->src.sin6.sin6_len == 0) {
320 sin6 = (struct sockaddr_in6 *)&saidx->src;
321 sin6->sin6_len = sizeof(*sin6);
322 sin6->sin6_family = AF_INET6;
323 sin6->sin6_port = IPSEC_PORT_ANY;
324 sin6->sin6_addr = ip6->ip6_src;
325 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
326 /* fix scope id for comparing SPD */
327 sin6->sin6_addr.s6_addr16[1] = 0;
328 sin6->sin6_scope_id =
329 ntohs(ip6->ip6_src.s6_addr16[1]);
332 if (saidx->dst.sin6.sin6_len == 0) {
333 sin6 = (struct sockaddr_in6 *)&saidx->dst;
334 sin6->sin6_len = sizeof(*sin6);
335 sin6->sin6_family = AF_INET6;
336 sin6->sin6_port = IPSEC_PORT_ANY;
337 sin6->sin6_addr = ip6->ip6_dst;
338 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
339 /* fix scope id for comparing SPD */
340 sin6->sin6_addr.s6_addr16[1] = 0;
341 sin6->sin6_scope_id =
342 ntohs(ip6->ip6_dst.s6_addr16[1]);
349 * Lookup SA and validate it.
351 *error = key_checkrequest(isr, saidx);
354 * IPsec processing is required, but no SA found.
355 * I assume that key_acquire() had been called
356 * to get/establish the SA. Here I discard
357 * this packet because it is responsibility for
358 * upper layer to retransmit the packet.
360 IPSECSTAT_INC(ips_out_nosa);
365 IPSEC_ASSERT(ipsec_get_reqlevel(isr) == IPSEC_LEVEL_USE,
366 ("no SA found, but required; level %u",
367 ipsec_get_reqlevel(isr)));
368 IPSECREQUEST_UNLOCK(isr);
371 * If isr is NULL, we found a 'use' policy w/o SA.
372 * Return w/o error and w/o isr so we can drop out
373 * and continue w/o IPsec processing.
377 IPSECREQUEST_LOCK(isr);
382 * Check system global policy controls.
384 if ((isr->saidx.proto == IPPROTO_ESP && !V_esp_enable) ||
385 (isr->saidx.proto == IPPROTO_AH && !V_ah_enable) ||
386 (isr->saidx.proto == IPPROTO_IPCOMP && !V_ipcomp_enable)) {
387 DPRINTF(("%s: IPsec outbound packet dropped due"
388 " to policy (check your sysctls)\n", __func__));
390 *error = EHOSTUNREACH;
395 * Sanity check the SA contents for the caller
396 * before they invoke the xform output method.
398 if (sav->tdb_xform == NULL) {
399 DPRINTF(("%s: no transform for SA\n", __func__));
400 IPSEC_OSTAT(noxform);
401 *error = EHOSTUNREACH;
406 IPSEC_ASSERT(*error != 0, ("error return w/ no error code"));
407 IPSECREQUEST_UNLOCK(isr);
414 * IPsec output logic for IPv4.
417 ipsec4_process_packet(
419 struct ipsecrequest *isr,
423 struct secasindex saidx;
424 struct secasvar *sav;
428 IPSEC_ASSERT(m != NULL, ("null mbuf"));
429 IPSEC_ASSERT(isr != NULL, ("null isr"));
431 IPSECREQUEST_LOCK(isr); /* insure SA contents don't change */
433 isr = ipsec_nextisr(m, isr, AF_INET, &saidx, &error);
443 encif->if_opackets++;
444 encif->if_obytes += m->m_pkthdr.len;
446 /* pass the mbuf to enc0 for bpf processing */
447 ipsec_bpf(m, sav, AF_INET, ENC_OUT|ENC_BEFORE);
448 /* pass the mbuf to enc0 for packet filtering */
449 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0)
454 union sockaddr_union *dst = &sav->sah->saidx.dst;
458 * Collect IP_DF state from the outer header.
460 if (dst->sa.sa_family == AF_INET) {
461 if (m->m_len < sizeof (struct ip) &&
462 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
466 ip = mtod(m, struct ip *);
467 /* Honor system-wide control of how to handle IP_DF */
468 switch (V_ip4_ipsec_dfbit) {
469 case 0: /* clear in outer header */
470 case 1: /* set in outer header */
471 setdf = V_ip4_ipsec_dfbit;
473 default: /* propagate to outer header */
474 setdf = ntohs(ip->ip_off & IP_DF);
478 ip = NULL; /* keep compiler happy */
481 /* Do the appropriate encapsulation, if necessary */
482 if (isr->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
483 dst->sa.sa_family != AF_INET || /* PF mismatch */
485 (sav->flags & SADB_X_SAFLAGS_TUNNEL) || /* Tunnel requ'd */
486 sav->tdb_xform->xf_type == XF_IP4 || /* ditto */
488 (dst->sa.sa_family == AF_INET && /* Proxy */
489 dst->sin.sin_addr.s_addr != INADDR_ANY &&
490 dst->sin.sin_addr.s_addr != ip->ip_dst.s_addr)) {
493 /* Fix IPv4 header checksum and length */
494 if (m->m_len < sizeof (struct ip) &&
495 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
499 ip = mtod(m, struct ip *);
500 if (ip->ip_v == IPVERSION) {
501 ip->ip_len = htons(m->m_pkthdr.len);
503 ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
506 /* Encapsulate the packet */
507 error = ipip_output(m, isr, &mp, 0, 0);
508 if (mp == NULL && !error) {
509 /* Should never happen. */
510 DPRINTF(("%s: ipip_output returns no mbuf and "
511 "no error!", __func__));
516 /* XXX: Should never happen! */
519 m = NULL; /* ipip_output() already freed it */
524 * ipip_output clears IP_DF in the new header. If
525 * we need to propagate IP_DF from the outer header,
526 * then we have to do it here.
528 * XXX shouldn't assume what ipip_output does.
530 if (dst->sa.sa_family == AF_INET && setdf) {
531 if (m->m_len < sizeof (struct ip) &&
532 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
536 ip = mtod(m, struct ip *);
537 ip->ip_off = ntohs(ip->ip_off);
539 ip->ip_off = htons(ip->ip_off);
545 /* pass the mbuf to enc0 for bpf processing */
546 ipsec_bpf(m, sav, sav->sah->saidx.dst.sa.sa_family, ENC_OUT|ENC_AFTER);
547 /* pass the mbuf to enc0 for packet filtering */
548 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0)
553 * Dispatch to the appropriate IPsec transform logic. The
554 * packet will be returned for transmission after crypto
555 * processing, etc. are completed. For encapsulation we
556 * bypass this call because of the explicit call done above
557 * (necessary to deal with IP_DF handling for IPv4).
559 * NB: m & sav are ``passed to caller'' who's reponsible for
560 * for reclaiming their resources.
562 if (sav->tdb_xform->xf_type != XF_IP4) {
563 union sockaddr_union *dst = &sav->sah->saidx.dst;
564 switch(dst->sa.sa_family) {
566 ip = mtod(m, struct ip *);
568 off = offsetof(struct ip, ip_p);
572 i = sizeof(struct ip6_hdr);
573 off = offsetof(struct ip6_hdr, ip6_nxt);
577 DPRINTF(("%s: unsupported protocol family %u\n",
578 __func__, dst->sa.sa_family));
579 error = EPFNOSUPPORT;
580 IPSECSTAT_INC(ips_out_inval);
583 error = (*sav->tdb_xform->xf_output)(m, isr, NULL, i, off);
585 error = ipsec_process_done(m, isr);
587 IPSECREQUEST_UNLOCK(isr);
591 IPSECREQUEST_UNLOCK(isr);
601 in6_sa_equal_addrwithscope(const struct sockaddr_in6 *sa, const struct in6_addr *ia)
605 memcpy(&ia2, &sa->sin6_addr, sizeof(ia2));
606 if (IN6_IS_SCOPE_LINKLOCAL(&sa->sin6_addr))
607 ia2.s6_addr16[1] = htons(sa->sin6_scope_id);
609 return IN6_ARE_ADDR_EQUAL(ia, &ia2);
613 * IPsec output logic for IPv6.
616 ipsec6_process_packet(
618 struct ipsecrequest *isr
621 struct secasindex saidx;
622 struct secasvar *sav;
625 union sockaddr_union *dst;
627 IPSEC_ASSERT(m != NULL, ("ipsec6_process_packet: null mbuf"));
628 IPSEC_ASSERT(isr != NULL, ("ipsec6_process_packet: null isr"));
630 IPSECREQUEST_LOCK(isr); /* insure SA contents don't change */
632 isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
640 dst = &sav->sah->saidx.dst;
643 encif->if_opackets++;
644 encif->if_obytes += m->m_pkthdr.len;
646 /* pass the mbuf to enc0 for bpf processing */
647 ipsec_bpf(m, isr->sav, AF_INET6, ENC_OUT|ENC_BEFORE);
648 /* pass the mbuf to enc0 for packet filtering */
649 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0)
653 ip6 = mtod(m, struct ip6_hdr *); /* XXX */
655 /* Do the appropriate encapsulation, if necessary */
656 if (isr->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
657 dst->sa.sa_family != AF_INET6 || /* PF mismatch */
658 ((dst->sa.sa_family == AF_INET6) &&
659 (!IN6_IS_ADDR_UNSPECIFIED(&dst->sin6.sin6_addr)) &&
660 (!in6_sa_equal_addrwithscope(&dst->sin6,
664 /* Fix IPv6 header payload length. */
665 if (m->m_len < sizeof(struct ip6_hdr))
666 if ((m = m_pullup(m,sizeof(struct ip6_hdr))) == NULL) {
671 if (m->m_pkthdr.len - sizeof(*ip6) > IPV6_MAXPACKET) {
672 /* No jumbogram support. */
673 error = ENXIO; /*XXX*/
677 ip6 = mtod(m, struct ip6_hdr *);
678 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(*ip6));
680 /* Encapsulate the packet */
681 error = ipip_output(m, isr, &mp, 0, 0);
682 if (mp == NULL && !error) {
683 /* Should never happen. */
684 DPRINTF(("ipsec6_process_packet: ipip_output "
685 "returns no mbuf and no error!"));
692 /* XXX: Should never happen! */
695 m = NULL; /* ipip_output() already freed it */
704 ipsec_bpf(m, isr->sav, dst->sa.sa_family, ENC_OUT|ENC_AFTER);
705 /* pass the mbuf to enc0 for packet filtering */
706 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0)
710 switch(dst->sa.sa_family) {
715 ip = mtod(m, struct ip *);
717 off = offsetof(struct ip, ip_p);
722 i = sizeof(struct ip6_hdr);
723 off = offsetof(struct ip6_hdr, ip6_nxt);
726 DPRINTF(("%s: unsupported protocol family %u\n",
727 __func__, dst->sa.sa_family));
728 error = EPFNOSUPPORT;
729 IPSEC6STAT_INC(ips_out_inval);
732 error = (*sav->tdb_xform->xf_output)(m, isr, NULL, i, off);
733 IPSECREQUEST_UNLOCK(isr);
738 IPSECREQUEST_UNLOCK(isr);