2 /* $OpenBSD: ip_ipip.c,v 1.25 2002/06/10 18:04:55 itojun Exp $ */
4 * The authors of this code are John Ioannidis (ji@tla.org),
5 * Angelos D. Keromytis (kermit@csd.uch.gr) and
6 * Niels Provos (provos@physnet.uni-hamburg.de).
8 * The original version of this code was written by John Ioannidis
9 * for BSD/OS in Athens, Greece, in November 1995.
11 * Ported to OpenBSD and NetBSD, with additional transforms, in December 1996,
12 * by Angelos D. Keromytis.
14 * Additional transforms and features in 1997 and 1998 by Angelos D. Keromytis
17 * Additional features in 1999 by Angelos D. Keromytis.
19 * Copyright (C) 1995, 1996, 1997, 1998, 1999 by John Ioannidis,
20 * Angelos D. Keromytis and Niels Provos.
21 * Copyright (c) 2001, Angelos D. Keromytis.
23 * Permission to use, copy, and modify this software with or without fee
24 * is hereby granted, provided that this entire notice is included in
25 * all copies of any software which is or includes a copy or
26 * modification of this software.
27 * You may use this code under the GNU public license if you so wish. Please
28 * contribute changes back to the authors under this freer than GPL license
29 * so that we may further the use of strong encryption without limitations to
32 * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR
33 * IMPLIED WARRANTY. IN PARTICULAR, NONE OF THE AUTHORS MAKES ANY
34 * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE
35 * MERCHANTABILITY OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR
40 * IP-inside-IP processing
43 #include "opt_inet6.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
48 #include <sys/socket.h>
49 #include <sys/kernel.h>
50 #include <sys/protosw.h>
51 #include <sys/sysctl.h>
54 #include <net/route.h>
55 #include <net/netisr.h>
57 #include <netinet/in.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/in_var.h>
60 #include <netinet/ip.h>
61 #include <netinet/ip_ecn.h>
62 #include <netinet/ip_var.h>
63 #include <netinet/ip_encap.h>
64 #include <netinet/ipprotosw.h>
66 #include <netipsec/ipsec.h>
67 #include <netipsec/xform.h>
69 #include <netipsec/ipip_var.h>
72 #include <netinet/ip_mroute.h>
76 #include <netinet/ip6.h>
77 #include <netipsec/ipsec6.h>
78 #include <netinet6/ip6_ecn.h>
79 #include <netinet6/in6_var.h>
80 #include <netinet6/ip6protosw.h>
83 #include <netipsec/key.h>
84 #include <netipsec/key_debug.h>
86 #include <machine/stdarg.h>
89 * We can control the acceptance of IP4 packets by altering the sysctl
90 * net.inet.ipip.allow value. Zero means drop them, all else is acceptance.
93 struct ipipstat ipipstat;
95 SYSCTL_DECL(_net_inet_ipip);
96 SYSCTL_INT(_net_inet_ipip, OID_AUTO,
97 ipip_allow, CTLFLAG_RW, &ipip_allow, 0, "");
98 SYSCTL_STRUCT(_net_inet_ipip, IPSECCTL_STATS,
99 stats, CTLFLAG_RD, &ipipstat, ipipstat, "");
102 #define M_IPSEC (M_AUTHIPHDR|M_AUTHIPDGM|M_DECRYPTED)
104 static void _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp);
108 * Really only a wrapper for ipip_input(), for use with IPv6.
111 ip4_input6(struct mbuf **m, int *offp, int proto)
114 /* If we do not accept IP-in-IP explicitly, drop. */
115 if (!ipip_allow && ((*m)->m_flags & M_IPSEC) == 0) {
116 DPRINTF(("%s: dropped due to policy\n", __func__));
117 ipipstat.ipips_pdrops++;
122 _ipip_input(*m, *offp, NULL);
129 * Really only a wrapper for ipip_input(), for use with IPv4.
132 ip4_input(struct mbuf *m, ...)
138 /* If we do not accept IP-in-IP explicitly, drop. */
139 if (!ipip_allow && (m->m_flags & M_IPSEC) == 0) {
140 DPRINTF(("%s: dropped due to policy\n", __func__));
141 ipipstat.ipips_pdrops++;
147 iphlen = va_arg(ap, int);
150 _ipip_input(m, iphlen, NULL);
155 * ipip_input gets called when we receive an IP{46} encapsulated packet,
156 * either because we got it at a real interface, or because AH or ESP
157 * were being used in tunnel mode (in which case the rcvif element will
158 * contain the address of the encX interface associated with the tunnel.
162 _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp)
164 register struct sockaddr_in *sin;
165 register struct ifnet *ifp;
166 register struct ifaddr *ifa;
169 register struct sockaddr_in6 *sin6;
170 struct ip6_hdr *ip6 = NULL;
179 ipipstat.ipips_ipackets++;
181 m_copydata(m, 0, 1, &v);
186 hlen = sizeof(struct ip);
191 hlen = sizeof(struct ip6_hdr);
195 ipipstat.ipips_family++;
197 return /* EAFNOSUPPORT */;
200 /* Bring the IP header in the first mbuf, if not there already */
201 if (m->m_len < hlen) {
202 if ((m = m_pullup(m, hlen)) == NULL) {
203 DPRINTF(("%s: m_pullup (1) failed\n", __func__));
204 ipipstat.ipips_hdrops++;
209 ipo = mtod(m, struct ip *);
212 if (ipo->ip_v == IPVERSION && ipo->ip_p == IPPROTO_IPV4) {
213 if (IN_MULTICAST(((struct ip *)((char *) ipo + iphlen))->ip_dst.s_addr)) {
214 ipip_mroute_input (m, iphlen);
218 #endif /* MROUTING */
220 /* Keep outer ecn field. */
229 otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
233 panic("ipip_input: unknown ip version %u (outer)", v>>4);
236 /* Remove outer IP header */
240 if (m->m_pkthdr.len < sizeof(struct ip)) {
241 ipipstat.ipips_hdrops++;
246 m_copydata(m, 0, 1, &v);
251 hlen = sizeof(struct ip);
257 hlen = sizeof(struct ip6_hdr);
261 ipipstat.ipips_family++;
263 return; /* EAFNOSUPPORT */
267 * Bring the inner IP header in the first mbuf, if not there already.
269 if (m->m_len < hlen) {
270 if ((m = m_pullup(m, hlen)) == NULL) {
271 DPRINTF(("%s: m_pullup (2) failed\n", __func__));
272 ipipstat.ipips_hdrops++;
278 * RFC 1853 specifies that the inner TTL should not be touched on
279 * decapsulation. There's no reason this comment should be here, but
280 * this is as good as any a position.
283 /* Some sanity checks in the inner IP header */
287 ipo = mtod(m, struct ip *);
289 ip_ecn_egress(ip4_ipsec_ecn, &otos, &ipo->ip_tos);
294 ip6 = (struct ip6_hdr *) ipo;
296 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
297 ip_ecn_egress(ip6_ipsec_ecn, &otos, &itos);
298 ip6->ip6_flow &= ~htonl(0xff << 20);
299 ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
303 panic("ipip_input: unknown ip version %u (inner)", v>>4);
306 /* Check for local address spoofing. */
307 if ((m->m_pkthdr.rcvif == NULL ||
308 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
311 TAILQ_FOREACH(ifp, &ifnet, if_link) {
312 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
315 if (ifa->ifa_addr->sa_family !=
319 sin = (struct sockaddr_in *) ifa->ifa_addr;
321 if (sin->sin_addr.s_addr ==
322 ipo->ip_src.s_addr) {
323 ipipstat.ipips_spoof++;
333 if (ifa->ifa_addr->sa_family !=
337 sin6 = (struct sockaddr_in6 *) ifa->ifa_addr;
339 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_src)) {
340 ipipstat.ipips_spoof++;
354 ipipstat.ipips_ibytes += m->m_pkthdr.len - iphlen;
357 * Interface pointer stays the same; if no IPsec processing has
358 * been done (or will be done), this will point to a normal
359 * interface. Otherwise, it'll point to an enc interface, which
360 * will allow a packet filter to distinguish between secure and
376 panic("%s: bogus ip version %u", __func__, v>>4);
379 if (netisr_queue(isr, m)) { /* (0) on success. */
380 ipipstat.ipips_qfull++;
381 DPRINTF(("%s: packet dropped because of full queue\n",
389 struct ipsecrequest *isr,
395 struct secasvar *sav;
397 struct secasindex *saidx;
404 struct ip6_hdr *ip6, *ip6o;
407 IPSEC_SPLASSERT_SOFTNET(__func__);
410 IPSEC_ASSERT(sav != NULL, ("null SA"));
411 IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
413 /* XXX Deal with empty TDB source/destination addresses. */
415 m_copydata(m, 0, 1, &tp);
416 tp = (tp >> 4) & 0xff; /* Get the IP version number. */
418 saidx = &sav->sah->saidx;
419 switch (saidx->dst.sa.sa_family) {
422 if (saidx->src.sa.sa_family != AF_INET ||
423 saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
424 saidx->dst.sin.sin_addr.s_addr == INADDR_ANY) {
425 DPRINTF(("%s: unspecified tunnel endpoint "
426 "address in SA %s/%08lx\n", __func__,
427 ipsec_address(&saidx->dst),
428 (u_long) ntohl(sav->spi)));
429 ipipstat.ipips_unspec++;
434 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
436 DPRINTF(("%s: M_PREPEND failed\n", __func__));
437 ipipstat.ipips_hdrops++;
442 ipo = mtod(m, struct ip *);
444 ipo->ip_v = IPVERSION;
446 ipo->ip_len = htons(m->m_pkthdr.len);
447 ipo->ip_ttl = ip_defttl;
449 ipo->ip_src = saidx->src.sin.sin_addr;
450 ipo->ip_dst = saidx->dst.sin.sin_addr;
452 ipo->ip_id = ip_newid();
454 /* If the inner protocol is IP... */
455 if (tp == IPVERSION) {
456 /* Save ECN notification */
457 m_copydata(m, sizeof(struct ip) +
458 offsetof(struct ip, ip_tos),
459 sizeof(u_int8_t), (caddr_t) &itos);
461 ipo->ip_p = IPPROTO_IPIP;
464 * We should be keeping tunnel soft-state and
465 * send back ICMPs if needed.
467 m_copydata(m, sizeof(struct ip) +
468 offsetof(struct ip, ip_off),
469 sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
470 ipo->ip_off = ntohs(ipo->ip_off);
471 ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
472 ipo->ip_off = htons(ipo->ip_off);
475 else if (tp == (IPV6_VERSION >> 4)) {
478 /* Save ECN notification. */
479 m_copydata(m, sizeof(struct ip) +
480 offsetof(struct ip6_hdr, ip6_flow),
481 sizeof(u_int32_t), (caddr_t) &itos32);
482 itos = ntohl(itos32) >> 20;
483 ipo->ip_p = IPPROTO_IPV6;
492 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
499 if (IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr) ||
500 saidx->src.sa.sa_family != AF_INET6 ||
501 IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr)) {
502 DPRINTF(("%s: unspecified tunnel endpoint "
503 "address in SA %s/%08lx\n", __func__,
504 ipsec_address(&saidx->dst),
505 (u_long) ntohl(sav->spi)));
506 ipipstat.ipips_unspec++;
511 /* scoped address handling */
512 ip6 = mtod(m, struct ip6_hdr *);
513 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
514 ip6->ip6_src.s6_addr16[1] = 0;
515 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
516 ip6->ip6_dst.s6_addr16[1] = 0;
518 M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
520 DPRINTF(("%s: M_PREPEND failed\n", __func__));
521 ipipstat.ipips_hdrops++;
526 /* Initialize IPv6 header */
527 ip6o = mtod(m, struct ip6_hdr *);
529 ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
530 ip6o->ip6_vfc |= IPV6_VERSION;
531 ip6o->ip6_plen = htons(m->m_pkthdr.len);
532 ip6o->ip6_hlim = ip_defttl;
533 ip6o->ip6_dst = saidx->dst.sin6.sin6_addr;
534 ip6o->ip6_src = saidx->src.sin6.sin6_addr;
537 if (tp == IPVERSION) {
538 /* Save ECN notification */
539 m_copydata(m, sizeof(struct ip6_hdr) +
540 offsetof(struct ip, ip_tos), sizeof(u_int8_t),
543 /* This is really IPVERSION. */
544 ip6o->ip6_nxt = IPPROTO_IPIP;
547 if (tp == (IPV6_VERSION >> 4)) {
550 /* Save ECN notification. */
551 m_copydata(m, sizeof(struct ip6_hdr) +
552 offsetof(struct ip6_hdr, ip6_flow),
553 sizeof(u_int32_t), (caddr_t) &itos32);
554 itos = ntohl(itos32) >> 20;
556 ip6o->ip6_nxt = IPPROTO_IPV6;
562 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
563 ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
569 DPRINTF(("%s: unsupported protocol family %u\n", __func__,
570 saidx->dst.sa.sa_family));
571 ipipstat.ipips_family++;
572 error = EAFNOSUPPORT; /* XXX diffs from openbsd */
576 ipipstat.ipips_opackets++;
580 if (saidx->dst.sa.sa_family == AF_INET) {
582 if (sav->tdb_xform->xf_type == XF_IP4)
583 tdb->tdb_cur_bytes +=
584 m->m_pkthdr.len - sizeof(struct ip);
586 ipipstat.ipips_obytes += m->m_pkthdr.len - sizeof(struct ip);
591 if (saidx->dst.sa.sa_family == AF_INET6) {
593 if (sav->tdb_xform->xf_type == XF_IP4)
594 tdb->tdb_cur_bytes +=
595 m->m_pkthdr.len - sizeof(struct ip6_hdr);
597 ipipstat.ipips_obytes +=
598 m->m_pkthdr.len - sizeof(struct ip6_hdr);
612 ipe4_init(struct secasvar *sav, struct xformsw *xsp)
614 sav->tdb_xform = xsp;
619 ipe4_zeroize(struct secasvar *sav)
621 sav->tdb_xform = NULL;
626 ipe4_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff)
628 /* This is a rather serious mistake, so no conditional printing. */
629 printf("%s: should never be called\n", __func__);
635 static struct xformsw ipe4_xformsw = {
636 XF_IP4, 0, "IPv4 Simple Encapsulation",
637 ipe4_init, ipe4_zeroize, ipe4_input, ipip_output,
640 extern struct domain inetdomain;
641 static struct ipprotosw ipe4_protosw[] = {
642 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR|PR_LASTHDR,
643 (pr_in_input_t*) ip4_input,
650 { SOCK_RAW, &inetdomain, IPPROTO_IPV6, PR_ATOMIC|PR_ADDR|PR_LASTHDR,
651 (pr_in_input_t*) ip4_input,
661 * Check the encapsulated packet to see if we want it
664 ipe4_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
667 * Only take packets coming from IPSEC tunnels; the rest
668 * must be handled by the gif tunnel code. Note that we
669 * also return a minimum priority when we want the packet
670 * so any explicit gif tunnels take precedence.
672 return ((m->m_flags & M_IPSEC) != 0 ? 1 : 0);
678 xform_register(&ipe4_xformsw);
679 /* attach to encapsulation framework */
680 /* XXX save return cookie for detach on module remove */
681 (void) encap_attach_func(AF_INET, -1,
682 ipe4_encapcheck, (struct protosw*) &ipe4_protosw[0], NULL);
684 (void) encap_attach_func(AF_INET6, -1,
685 ipe4_encapcheck, (struct protosw*) &ipe4_protosw[1], NULL);
688 SYSINIT(ipe4_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipe4_attach, NULL);
689 #endif /* FAST_IPSEC */