2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2002, 2003 Sam Leffler, Errno Consulting
5 * Copyright (c) 2016 Andrey V. Elsukov <ae@FreeBSD.org>
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * IPsec output processing.
36 #include "opt_inet6.h"
37 #include "opt_ipsec.h"
40 #include <sys/param.h>
41 #include <sys/systm.h>
43 #include <sys/domain.h>
44 #include <sys/protosw.h>
45 #include <sys/socket.h>
46 #include <sys/errno.h>
47 #include <sys/hhook.h>
48 #include <sys/syslog.h>
51 #include <net/if_enc.h>
52 #include <net/if_var.h>
55 #include <netinet/in.h>
56 #include <netinet/in_systm.h>
57 #include <netinet/ip.h>
58 #include <netinet/ip_var.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip_ecn.h>
62 #include <netinet6/ip6_ecn.h>
65 #include <netinet/ip6.h>
67 #include <netinet6/ip6_var.h>
68 #include <netinet6/scope6_var.h>
70 #include <netinet/in_pcb.h>
72 #include <netinet/icmp6.h>
74 #if defined(SCTP) || defined(SCTP_SUPPORT)
75 #include <netinet/sctp_crc32.h>
78 #include <netinet/udp.h>
79 #include <netipsec/ah.h>
80 #include <netipsec/esp.h>
81 #include <netipsec/ipsec.h>
83 #include <netipsec/ipsec6.h>
85 #include <netipsec/ah_var.h>
86 #include <netipsec/esp_var.h>
87 #include <netipsec/ipcomp_var.h>
89 #include <netipsec/xform.h>
91 #include <netipsec/key.h>
92 #include <netipsec/keydb.h>
93 #include <netipsec/key_debug.h>
95 #include <machine/in_cksum.h>
97 #define IPSEC_OSTAT_INC(proto, name) do { \
98 if ((proto) == IPPROTO_ESP) \
99 ESPSTAT_INC(esps_##name); \
100 else if ((proto) == IPPROTO_AH)\
101 AHSTAT_INC(ahs_##name); \
103 IPCOMPSTAT_INC(ipcomps_##name); \
106 static int ipsec_encap(struct mbuf **mp, struct secasindex *saidx);
109 static struct secasvar *
110 ipsec4_allocsa(struct mbuf *m, struct secpolicy *sp, u_int *pidx, int *error)
112 struct secasindex *saidx, tmpsaidx;
113 struct ipsecrequest *isr;
114 struct sockaddr_in *sin;
115 struct secasvar *sav;
119 * Check system global policy controls.
122 isr = sp->req[*pidx];
123 if ((isr->saidx.proto == IPPROTO_ESP && !V_esp_enable) ||
124 (isr->saidx.proto == IPPROTO_AH && !V_ah_enable) ||
125 (isr->saidx.proto == IPPROTO_IPCOMP && !V_ipcomp_enable)) {
126 DPRINTF(("%s: IPsec outbound packet dropped due"
127 " to policy (check your sysctls)\n", __func__));
128 IPSEC_OSTAT_INC(isr->saidx.proto, pdrops);
129 *error = EHOSTUNREACH;
133 * Craft SA index to search for proper SA. Note that
134 * we only initialize unspecified SA peers for transport
135 * mode; for tunnel mode they must already be filled in.
137 if (isr->saidx.mode == IPSEC_MODE_TRANSPORT) {
140 ip = mtod(m, struct ip *);
141 if (saidx->src.sa.sa_len == 0) {
142 sin = &saidx->src.sin;
143 sin->sin_len = sizeof(*sin);
144 sin->sin_family = AF_INET;
145 sin->sin_port = IPSEC_PORT_ANY;
146 sin->sin_addr = ip->ip_src;
148 if (saidx->dst.sa.sa_len == 0) {
149 sin = &saidx->dst.sin;
150 sin->sin_len = sizeof(*sin);
151 sin->sin_family = AF_INET;
152 sin->sin_port = IPSEC_PORT_ANY;
153 sin->sin_addr = ip->ip_dst;
156 saidx = &sp->req[*pidx]->saidx;
158 * Lookup SA and validate it.
160 sav = key_allocsa_policy(sp, saidx, error);
162 IPSECSTAT_INC(ips_out_nosa);
165 if (ipsec_get_reqlevel(sp, *pidx) != IPSEC_LEVEL_REQUIRE) {
167 * We have no SA and policy that doesn't require
168 * this IPsec transform, thus we can continue w/o
169 * IPsec processing, i.e. return EJUSTRETURN.
170 * But first check if there is some bundled transform.
172 if (sp->tcount > ++(*pidx))
174 *error = EJUSTRETURN;
178 IPSEC_ASSERT(sav->tdb_xform != NULL, ("SA with NULL tdb_xform"));
183 * IPsec output logic for IPv4.
186 ipsec4_perform_request(struct mbuf *m, struct secpolicy *sp,
187 struct inpcb *inp, u_int idx)
189 struct ipsec_ctx_data ctx;
190 union sockaddr_union *dst;
191 struct secasvar *sav;
195 IPSEC_ASSERT(idx < sp->tcount, ("Wrong IPsec request index %d", idx));
198 * We hold the reference to SP. Content of SP couldn't be changed.
199 * Craft secasindex and do lookup for suitable SA.
200 * Then do encapsulation if needed and call xform's output.
201 * We need to store SP in the xform callback parameters.
202 * In xform callback we will extract SP and it can be used to
203 * determine next transform. At the end of transform we can
204 * release reference to SP.
206 sav = ipsec4_allocsa(m, sp, &idx, &error);
208 if (error == EJUSTRETURN) { /* No IPsec required */
215 * XXXAE: most likely ip_sum at this point is wrong.
217 IPSEC_INIT_CTX(&ctx, &m, inp, sav, AF_INET, IPSEC_ENC_BEFORE);
218 if ((error = ipsec_run_hhooks(&ctx, HHOOK_TYPE_IPSEC_OUT)) != 0)
221 ip = mtod(m, struct ip *);
222 dst = &sav->sah->saidx.dst;
223 /* Do the appropriate encapsulation, if necessary */
224 if (sp->req[idx]->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
225 dst->sa.sa_family != AF_INET || /* PF mismatch */
226 (dst->sa.sa_family == AF_INET && /* Proxy */
227 dst->sin.sin_addr.s_addr != INADDR_ANY &&
228 dst->sin.sin_addr.s_addr != ip->ip_dst.s_addr)) {
229 /* Fix IPv4 header checksum and length */
230 ip->ip_len = htons(m->m_pkthdr.len);
232 ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
233 error = ipsec_encap(&m, &sav->sah->saidx);
235 DPRINTF(("%s: encapsulation for SPI 0x%08x failed "
236 "with error %d\n", __func__, ntohl(sav->spi),
238 /* XXXAE: IPSEC_OSTAT_INC(tunnel); */
244 IPSEC_INIT_CTX(&ctx, &m, inp, sav, dst->sa.sa_family, IPSEC_ENC_AFTER);
245 if ((error = ipsec_run_hhooks(&ctx, HHOOK_TYPE_IPSEC_OUT)) != 0)
249 * Dispatch to the appropriate IPsec transform logic. The
250 * packet will be returned for transmission after crypto
251 * processing, etc. are completed.
253 * NB: m & sav are ``passed to caller'' who's responsible for
254 * reclaiming their resources.
256 switch(dst->sa.sa_family) {
258 ip = mtod(m, struct ip *);
260 off = offsetof(struct ip, ip_p);
264 i = sizeof(struct ip6_hdr);
265 off = offsetof(struct ip6_hdr, ip6_nxt);
269 DPRINTF(("%s: unsupported protocol family %u\n",
270 __func__, dst->sa.sa_family));
271 error = EPFNOSUPPORT;
272 IPSEC_OSTAT_INC(sav->sah->saidx.proto, nopf);
275 error = (*sav->tdb_xform->xf_output)(m, sp, sav, idx, i, off);
278 IPSECSTAT_INC(ips_out_inval);
288 ipsec4_process_packet(struct mbuf *m, struct secpolicy *sp,
292 return (ipsec4_perform_request(m, sp, inp, 0));
296 ipsec4_common_output(struct mbuf *m, struct inpcb *inp, int forwarding)
298 struct secpolicy *sp;
301 /* Lookup for the corresponding outbound security policy */
302 sp = ipsec4_checkpolicy(m, inp, &error, !forwarding);
304 if (error == -EINVAL) {
305 /* Discarded by policy. */
309 return (0); /* No IPsec required. */
313 * Usually we have to have tunnel mode IPsec security policy
314 * when we are forwarding a packet. Otherwise we could not handle
315 * encrypted replies, because they are not destined for us. But
316 * some users are doing source address translation for forwarded
317 * packets, and thus, even if they are forwarded, the replies will
322 * Do delayed checksums now because we send before
323 * this is done in the normal processing path.
325 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
327 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
329 #if defined(SCTP) || defined(SCTP_SUPPORT)
330 if (m->m_pkthdr.csum_flags & CSUM_SCTP) {
333 ip = mtod(m, struct ip *);
334 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
335 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
339 /* NB: callee frees mbuf and releases reference to SP */
340 error = ipsec4_process_packet(m, sp, inp);
341 if (error == EJUSTRETURN) {
343 * We had a SP with a level of 'use' and no SA. We
344 * will just continue to process the packet without
345 * IPsec processing and return without error.
350 return (EINPROGRESS); /* consumed by IPsec */
355 * IPSEC_OUTPUT() method implementation for IPv4.
356 * 0 - no IPsec handling needed
357 * other values - mbuf consumed by IPsec.
360 ipsec4_output(struct mbuf *m, struct inpcb *inp)
364 * If the packet is resubmitted to ip_output (e.g. after
365 * AH, ESP, etc. processing), there will be a tag to bypass
366 * the lookup and related policy checking.
368 if (m_tag_find(m, PACKET_TAG_IPSEC_OUT_DONE, NULL) != NULL)
371 return (ipsec4_common_output(m, inp, 0));
375 * IPSEC_FORWARD() method implementation for IPv4.
376 * 0 - no IPsec handling needed
377 * other values - mbuf consumed by IPsec.
380 ipsec4_forward(struct mbuf *m)
384 * Check if this packet has an active inbound SP and needs to be
385 * dropped instead of forwarded.
387 if (ipsec4_in_reject(m, NULL) != 0) {
391 return (ipsec4_common_output(m, NULL, 1));
397 in6_sa_equal_addrwithscope(const struct sockaddr_in6 *sa,
398 const struct in6_addr *ia)
402 if (IN6_IS_SCOPE_LINKLOCAL(&sa->sin6_addr)) {
403 memcpy(&ia2, &sa->sin6_addr, sizeof(ia2));
404 ia2.s6_addr16[1] = htons(sa->sin6_scope_id);
405 return (IN6_ARE_ADDR_EQUAL(ia, &ia2));
407 return (IN6_ARE_ADDR_EQUAL(&sa->sin6_addr, ia));
410 static struct secasvar *
411 ipsec6_allocsa(struct mbuf *m, struct secpolicy *sp, u_int *pidx, int *error)
413 struct secasindex *saidx, tmpsaidx;
414 struct ipsecrequest *isr;
415 struct sockaddr_in6 *sin6;
416 struct secasvar *sav;
420 * Check system global policy controls.
423 isr = sp->req[*pidx];
424 if ((isr->saidx.proto == IPPROTO_ESP && !V_esp_enable) ||
425 (isr->saidx.proto == IPPROTO_AH && !V_ah_enable) ||
426 (isr->saidx.proto == IPPROTO_IPCOMP && !V_ipcomp_enable)) {
427 DPRINTF(("%s: IPsec outbound packet dropped due"
428 " to policy (check your sysctls)\n", __func__));
429 IPSEC_OSTAT_INC(isr->saidx.proto, pdrops);
430 *error = EHOSTUNREACH;
434 * Craft SA index to search for proper SA. Note that
435 * we only fillin unspecified SA peers for transport
436 * mode; for tunnel mode they must already be filled in.
438 if (isr->saidx.mode == IPSEC_MODE_TRANSPORT) {
441 ip6 = mtod(m, struct ip6_hdr *);
442 if (saidx->src.sin6.sin6_len == 0) {
443 sin6 = (struct sockaddr_in6 *)&saidx->src;
444 sin6->sin6_len = sizeof(*sin6);
445 sin6->sin6_family = AF_INET6;
446 sin6->sin6_port = IPSEC_PORT_ANY;
447 sin6->sin6_addr = ip6->ip6_src;
448 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
449 /* fix scope id for comparing SPD */
450 sin6->sin6_addr.s6_addr16[1] = 0;
451 sin6->sin6_scope_id =
452 ntohs(ip6->ip6_src.s6_addr16[1]);
455 if (saidx->dst.sin6.sin6_len == 0) {
456 sin6 = (struct sockaddr_in6 *)&saidx->dst;
457 sin6->sin6_len = sizeof(*sin6);
458 sin6->sin6_family = AF_INET6;
459 sin6->sin6_port = IPSEC_PORT_ANY;
460 sin6->sin6_addr = ip6->ip6_dst;
461 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
462 /* fix scope id for comparing SPD */
463 sin6->sin6_addr.s6_addr16[1] = 0;
464 sin6->sin6_scope_id =
465 ntohs(ip6->ip6_dst.s6_addr16[1]);
469 saidx = &sp->req[*pidx]->saidx;
471 * Lookup SA and validate it.
473 sav = key_allocsa_policy(sp, saidx, error);
475 IPSEC6STAT_INC(ips_out_nosa);
478 if (ipsec_get_reqlevel(sp, *pidx) != IPSEC_LEVEL_REQUIRE) {
480 * We have no SA and policy that doesn't require
481 * this IPsec transform, thus we can continue w/o
482 * IPsec processing, i.e. return EJUSTRETURN.
483 * But first check if there is some bundled transform.
485 if (sp->tcount > ++(*pidx))
487 *error = EJUSTRETURN;
491 IPSEC_ASSERT(sav->tdb_xform != NULL, ("SA with NULL tdb_xform"));
496 * IPsec output logic for IPv6.
499 ipsec6_perform_request(struct mbuf *m, struct secpolicy *sp,
500 struct inpcb *inp, u_int idx)
502 struct ipsec_ctx_data ctx;
503 union sockaddr_union *dst;
504 struct secasvar *sav;
508 IPSEC_ASSERT(idx < sp->tcount, ("Wrong IPsec request index %d", idx));
510 sav = ipsec6_allocsa(m, sp, &idx, &error);
512 if (error == EJUSTRETURN) { /* No IPsec required */
519 /* Fix IP length in case if it is not set yet. */
520 ip6 = mtod(m, struct ip6_hdr *);
521 ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(*ip6));
523 IPSEC_INIT_CTX(&ctx, &m, inp, sav, AF_INET6, IPSEC_ENC_BEFORE);
524 if ((error = ipsec_run_hhooks(&ctx, HHOOK_TYPE_IPSEC_OUT)) != 0)
527 ip6 = mtod(m, struct ip6_hdr *); /* pfil can change mbuf */
528 dst = &sav->sah->saidx.dst;
530 /* Do the appropriate encapsulation, if necessary */
531 if (sp->req[idx]->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
532 dst->sa.sa_family != AF_INET6 || /* PF mismatch */
533 ((dst->sa.sa_family == AF_INET6) &&
534 (!IN6_IS_ADDR_UNSPECIFIED(&dst->sin6.sin6_addr)) &&
535 (!in6_sa_equal_addrwithscope(&dst->sin6, &ip6->ip6_dst)))) {
536 if (m->m_pkthdr.len - sizeof(*ip6) > IPV6_MAXPACKET) {
537 /* No jumbogram support. */
538 error = ENXIO; /*XXX*/
541 error = ipsec_encap(&m, &sav->sah->saidx);
543 DPRINTF(("%s: encapsulation for SPI 0x%08x failed "
544 "with error %d\n", __func__, ntohl(sav->spi),
546 /* XXXAE: IPSEC_OSTAT_INC(tunnel); */
552 IPSEC_INIT_CTX(&ctx, &m, inp, sav, dst->sa.sa_family, IPSEC_ENC_AFTER);
553 if ((error = ipsec_run_hhooks(&ctx, HHOOK_TYPE_IPSEC_OUT)) != 0)
556 switch(dst->sa.sa_family) {
561 ip = mtod(m, struct ip *);
563 off = offsetof(struct ip, ip_p);
568 i = sizeof(struct ip6_hdr);
569 off = offsetof(struct ip6_hdr, ip6_nxt);
572 DPRINTF(("%s: unsupported protocol family %u\n",
573 __func__, dst->sa.sa_family));
574 error = EPFNOSUPPORT;
575 IPSEC_OSTAT_INC(sav->sah->saidx.proto, nopf);
578 error = (*sav->tdb_xform->xf_output)(m, sp, sav, idx, i, off);
581 IPSEC6STAT_INC(ips_out_inval);
591 ipsec6_process_packet(struct mbuf *m, struct secpolicy *sp,
595 return (ipsec6_perform_request(m, sp, inp, 0));
599 ipsec6_common_output(struct mbuf *m, struct inpcb *inp, int forwarding)
601 struct secpolicy *sp;
604 /* Lookup for the corresponding outbound security policy */
605 sp = ipsec6_checkpolicy(m, inp, &error, !forwarding);
607 if (error == -EINVAL) {
608 /* Discarded by policy. */
612 return (0); /* No IPsec required. */
617 * Do delayed checksums now because we send before
618 * this is done in the normal processing path.
620 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
621 in6_delayed_cksum(m, m->m_pkthdr.len -
622 sizeof(struct ip6_hdr), sizeof(struct ip6_hdr));
623 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
625 #if defined(SCTP) || defined(SCTP_SUPPORT)
626 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
627 sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
628 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
632 /* NB: callee frees mbuf and releases reference to SP */
633 error = ipsec6_process_packet(m, sp, inp);
634 if (error == EJUSTRETURN) {
636 * We had a SP with a level of 'use' and no SA. We
637 * will just continue to process the packet without
638 * IPsec processing and return without error.
643 return (EINPROGRESS); /* consumed by IPsec */
648 * IPSEC_OUTPUT() method implementation for IPv6.
649 * 0 - no IPsec handling needed
650 * other values - mbuf consumed by IPsec.
653 ipsec6_output(struct mbuf *m, struct inpcb *inp)
657 * If the packet is resubmitted to ip_output (e.g. after
658 * AH, ESP, etc. processing), there will be a tag to bypass
659 * the lookup and related policy checking.
661 if (m_tag_find(m, PACKET_TAG_IPSEC_OUT_DONE, NULL) != NULL)
664 return (ipsec6_common_output(m, inp, 0));
668 * IPSEC_FORWARD() method implementation for IPv6.
669 * 0 - no IPsec handling needed
670 * other values - mbuf consumed by IPsec.
673 ipsec6_forward(struct mbuf *m)
677 * Check if this packet has an active inbound SP and needs to be
678 * dropped instead of forwarded.
680 if (ipsec6_in_reject(m, NULL) != 0) {
684 return (ipsec6_common_output(m, NULL, 1));
689 ipsec_process_done(struct mbuf *m, struct secpolicy *sp, struct secasvar *sav,
692 struct epoch_tracker et;
693 struct xform_history *xh;
694 struct secasindex *saidx;
698 saidx = &sav->sah->saidx;
699 switch (saidx->dst.sa.sa_family) {
702 /* Fix the header length, for AH processing. */
703 mtod(m, struct ip *)->ip_len = htons(m->m_pkthdr.len);
708 /* Fix the header length, for AH processing. */
709 if (m->m_pkthdr.len < sizeof (struct ip6_hdr)) {
713 if (m->m_pkthdr.len - sizeof (struct ip6_hdr) > IPV6_MAXPACKET) {
714 /* No jumbogram support. */
718 mtod(m, struct ip6_hdr *)->ip6_plen =
719 htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
723 DPRINTF(("%s: unknown protocol family %u\n", __func__,
724 saidx->dst.sa.sa_family));
730 * Add a record of what we've done to the packet.
732 mtag = m_tag_get(PACKET_TAG_IPSEC_OUT_DONE, sizeof(*xh), M_NOWAIT);
734 DPRINTF(("%s: could not get packet tag\n", __func__));
739 xh = (struct xform_history *)(mtag + 1);
740 xh->dst = saidx->dst;
741 xh->proto = saidx->proto;
742 xh->mode = saidx->mode;
744 m_tag_prepend(m, mtag);
746 key_sa_recordxfer(sav, m); /* record data transfer */
749 * If there's another (bundled) SA to apply, do so.
750 * Note that this puts a burden on the kernel stack size.
751 * If this is a problem we'll need to introduce a queue
752 * to set the packet on so we can unwind the stack before
753 * doing further processing.
755 if (++idx < sp->tcount) {
756 switch (saidx->dst.sa.sa_family) {
760 IPSECSTAT_INC(ips_out_bundlesa);
761 return (ipsec4_perform_request(m, sp, NULL, idx));
767 IPSEC6STAT_INC(ips_out_bundlesa);
768 return (ipsec6_perform_request(m, sp, NULL, idx));
772 DPRINTF(("%s: unknown protocol family %u\n", __func__,
773 saidx->dst.sa.sa_family));
774 error = EPFNOSUPPORT;
779 key_freesp(&sp), sp = NULL; /* Release reference to SP */
782 * Do UDP encapsulation if SA requires it.
784 if (sav->natt != NULL) {
785 error = udp_ipsec_output(m, sav);
791 * We're done with IPsec processing, transmit the packet using the
792 * appropriate network protocol (IP or IPv6).
795 switch (saidx->dst.sa.sa_family) {
799 error = ip_output(m, NULL, NULL, IP_RAWOUTPUT, NULL, NULL);
805 error = ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
809 panic("ipsec_process_done");
822 * ipsec_prepend() is optimized version of M_PREPEND().
823 * ipsec_encap() is called by IPsec output routine for tunnel mode SA.
824 * It is expected that after IP encapsulation some IPsec transform will
825 * be performed. Each IPsec transform inserts its variable length header
826 * just after outer IP header using m_makespace(). If given mbuf has not
827 * enough free space at the beginning, we allocate new mbuf and reserve
828 * some space at the beginning and at the end.
829 * This helps avoid allocating of new mbuf and data copying in m_makespace(),
830 * we place outer header in the middle of mbuf's data with reserved leading
831 * and trailing space:
832 * [ LEADINGSPACE ][ Outer IP header ][ TRAILINGSPACE ]
833 * LEADINGSPACE will be used to add ethernet header, TRAILINGSPACE will
834 * be used to inject AH/ESP/IPCOMP header.
836 #define IPSEC_TRAILINGSPACE (sizeof(struct udphdr) +/* NAT-T */ \
837 max(sizeof(struct newesp) + EALG_MAX_BLOCK_LEN, /* ESP + IV */ \
838 sizeof(struct newah) + HASH_MAX_LEN /* AH + ICV */))
840 ipsec_prepend(struct mbuf *m, int len, int how)
845 IPSEC_ASSERT(len < MHLEN, ("wrong length"));
846 if (M_LEADINGSPACE(m) >= len) {
847 /* No need to allocate new mbuf. */
850 m->m_pkthdr.len += len;
853 n = m_gethdr(how, m->m_type);
860 if (len + IPSEC_TRAILINGSPACE < M_SIZE(n))
861 m_align(n, len + IPSEC_TRAILINGSPACE);
863 n->m_pkthdr.len += len;
868 ipsec_encap(struct mbuf **mp, struct secasindex *saidx)
877 ip = mtod(*mp, struct ip *);
881 proto = IPPROTO_IPIP;
883 * Collect IP_DF state from the inner header
884 * and honor system-wide control of how to handle it.
886 switch (V_ip4_ipsec_dfbit) {
887 case 0: /* clear in outer header */
888 case 1: /* set in outer header */
889 setdf = V_ip4_ipsec_dfbit;
891 default:/* propagate to outer header */
892 setdf = (ip->ip_off & htons(IP_DF)) != 0;
898 case (IPV6_VERSION >> 4):
899 proto = IPPROTO_IPV6;
900 ip6 = mtod(*mp, struct ip6_hdr *);
901 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
902 setdf = V_ip4_ipsec_dfbit ? 1: 0;
903 /* scoped address handling */
904 in6_clearscope(&ip6->ip6_src);
905 in6_clearscope(&ip6->ip6_dst);
909 return (EAFNOSUPPORT);
911 switch (saidx->dst.sa.sa_family) {
914 if (saidx->src.sa.sa_family != AF_INET ||
915 saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
916 saidx->dst.sin.sin_addr.s_addr == INADDR_ANY)
918 *mp = ipsec_prepend(*mp, sizeof(struct ip), M_NOWAIT);
921 ip = mtod(*mp, struct ip *);
922 ip->ip_v = IPVERSION;
923 ip->ip_hl = sizeof(struct ip) >> 2;
925 ip->ip_len = htons((*mp)->m_pkthdr.len);
926 ip->ip_ttl = V_ip_defttl;
928 ip->ip_off = setdf ? htons(IP_DF): 0;
929 ip->ip_src = saidx->src.sin.sin_addr;
930 ip->ip_dst = saidx->dst.sin.sin_addr;
931 ip_ecn_ingress(V_ip4_ipsec_ecn, &ip->ip_tos, &itos);
937 if (saidx->src.sa.sa_family != AF_INET6 ||
938 IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr) ||
939 IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr))
941 *mp = ipsec_prepend(*mp, sizeof(struct ip6_hdr), M_NOWAIT);
944 ip6 = mtod(*mp, struct ip6_hdr *);
946 ip6->ip6_vfc = IPV6_VERSION;
947 ip6->ip6_hlim = V_ip6_defhlim;
948 ip6->ip6_nxt = proto;
949 ip6->ip6_dst = saidx->dst.sin6.sin6_addr;
950 /* For link-local address embed scope zone id */
951 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
952 ip6->ip6_dst.s6_addr16[1] =
953 htons(saidx->dst.sin6.sin6_scope_id & 0xffff);
954 ip6->ip6_src = saidx->src.sin6.sin6_addr;
955 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
956 ip6->ip6_src.s6_addr16[1] =
957 htons(saidx->src.sin6.sin6_scope_id & 0xffff);
958 ip6->ip6_plen = htons((*mp)->m_pkthdr.len - sizeof(*ip6));
959 ip_ecn_ingress(V_ip6_ipsec_ecn, &proto, &itos);
960 ip6->ip6_flow |= htonl((uint32_t)proto << 20);
964 return (EAFNOSUPPORT);
966 (*mp)->m_flags &= ~(M_BCAST | M_MCAST);