1 /* $FreeBSD: src/sys/netipsec/xform_ipip.c,v 1.3.2.1 2003/01/24 05:11:36 sam Exp $ */
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"
44 #include "opt_random_ip_id.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
49 #include <sys/socket.h>
50 #include <sys/kernel.h>
51 #include <sys/protosw.h>
52 #include <sys/sysctl.h>
55 #include <net/route.h>
56 #include <net/netisr.h>
58 #include <netinet/in.h>
59 #include <netinet/in_systm.h>
60 #include <netinet/in_var.h>
61 #include <netinet/ip.h>
62 #include <netinet/ip_ecn.h>
63 #include <netinet/ip_var.h>
64 #include <netinet/ip_encap.h>
65 #include <netinet/ipprotosw.h>
67 #include <netipsec/ipsec.h>
68 #include <netipsec/xform.h>
70 #include <netipsec/ipip_var.h>
73 #include <netinet/ip_mroute.h>
77 #include <netinet/ip6.h>
78 #include <netipsec/ipsec6.h>
79 #include <netinet6/ip6_ecn.h>
80 #include <netinet6/in6_var.h>
81 #include <netinet6/ip6protosw.h>
84 #include <netipsec/key.h>
85 #include <netipsec/key_debug.h>
87 #include <machine/stdarg.h>
89 typedef void pr_in_input_t (struct mbuf *, int, int); /* XXX FIX THIS */
92 * We can control the acceptance of IP4 packets by altering the sysctl
93 * net.inet.ipip.allow value. Zero means drop them, all else is acceptance.
96 struct ipipstat ipipstat;
98 SYSCTL_DECL(_net_inet_ipip);
99 SYSCTL_INT(_net_inet_ipip, OID_AUTO,
100 ipip_allow, CTLFLAG_RW, &ipip_allow, 0, "");
101 SYSCTL_STRUCT(_net_inet_ipip, IPSECCTL_STATS,
102 stats, CTLFLAG_RD, &ipipstat, ipipstat, "");
105 #define M_IPSEC (M_AUTHIPHDR|M_AUTHIPDGM|M_DECRYPTED)
107 static void _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp);
111 * Really only a wrapper for ipip_input(), for use with IPv6.
114 ip4_input6(struct mbuf **m, int *offp, int proto)
117 /* If we do not accept IP-in-IP explicitly, drop. */
118 if (!ipip_allow && ((*m)->m_flags & M_IPSEC) == 0) {
119 DPRINTF(("ip4_input6: dropped due to policy\n"));
120 ipipstat.ipips_pdrops++;
125 _ipip_input(*m, *offp, NULL);
132 * Really only a wrapper for ipip_input(), for use with IPv4.
135 ip4_input(struct mbuf *m, ...)
141 /* If we do not accept IP-in-IP explicitly, drop. */
142 if (!ipip_allow && (m->m_flags & M_IPSEC) == 0) {
143 DPRINTF(("ip4_input: dropped due to policy\n"));
144 ipipstat.ipips_pdrops++;
150 iphlen = va_arg(ap, int);
153 _ipip_input(m, iphlen, NULL);
158 * ipip_input gets called when we receive an IP{46} encapsulated packet,
159 * either because we got it at a real interface, or because AH or ESP
160 * were being used in tunnel mode (in which case the rcvif element will
161 * contain the address of the encX interface associated with the tunnel.
165 _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp)
167 register struct sockaddr_in *sin;
168 register struct ifnet *ifp;
169 register struct ifaddr *ifa;
170 struct ifqueue *ifq = NULL;
173 register struct sockaddr_in6 *sin6;
174 struct ip6_hdr *ip6 = NULL;
183 ipipstat.ipips_ipackets++;
185 m_copydata(m, 0, 1, &v);
190 hlen = sizeof(struct ip);
195 hlen = sizeof(struct ip6_hdr);
199 DPRINTF(("_ipip_input: bad protocol version 0x%x (%u) "
200 "for outer header\n", v, v>>4));
201 ipipstat.ipips_family++;
203 return /* EAFNOSUPPORT */;
206 /* Bring the IP header in the first mbuf, if not there already */
207 if (m->m_len < hlen) {
208 if ((m = m_pullup(m, hlen)) == NULL) {
209 DPRINTF(("ipip_input: m_pullup (1) failed\n"));
210 ipipstat.ipips_hdrops++;
215 ipo = mtod(m, struct ip *);
218 if (ipo->ip_v == IPVERSION && ipo->ip_p == IPPROTO_IPV4) {
219 if (IN_MULTICAST(((struct ip *)((char *) ipo + iphlen))->ip_dst.s_addr)) {
220 ipip_mroute_input (m, iphlen);
224 #endif /* MROUTING */
226 /* Keep outer ecn field. */
235 otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
239 panic("ipip_input: unknown ip version %u (outer)", v>>4);
242 /* Remove outer IP header */
246 if (m->m_pkthdr.len < sizeof(struct ip)) {
247 ipipstat.ipips_hdrops++;
252 m_copydata(m, 0, 1, &v);
257 hlen = sizeof(struct ip);
263 hlen = sizeof(struct ip6_hdr);
267 DPRINTF(("_ipip_input: bad protocol version 0x%x (%u) "
268 "for inner header\n", v, v>>4));
269 ipipstat.ipips_family++;
271 return; /* EAFNOSUPPORT */
275 * Bring the inner IP header in the first mbuf, if not there already.
277 if (m->m_len < hlen) {
278 if ((m = m_pullup(m, hlen)) == NULL) {
279 DPRINTF(("ipip_input: m_pullup (2) failed\n"));
280 ipipstat.ipips_hdrops++;
286 * RFC 1853 specifies that the inner TTL should not be touched on
287 * decapsulation. There's no reason this comment should be here, but
288 * this is as good as any a position.
291 /* Some sanity checks in the inner IP header */
295 ipo = mtod(m, struct ip *);
297 ip_ecn_egress(ip4_ipsec_ecn, &otos, &ipo->ip_tos);
302 ip6 = (struct ip6_hdr *) ipo;
304 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
305 ip_ecn_egress(ip6_ipsec_ecn, &otos, &itos);
306 ip6->ip6_flow &= ~htonl(0xff << 20);
307 ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
311 panic("ipip_input: unknown ip version %u (inner)", v>>4);
314 /* Check for local address spoofing. */
315 if ((m->m_pkthdr.rcvif == NULL ||
316 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
318 for (ifp = ifnet.tqh_first; ifp != 0;
319 ifp = ifp->if_list.tqe_next) {
320 for (ifa = ifp->if_addrlist.tqh_first; ifa != 0;
321 ifa = ifa->ifa_list.tqe_next) {
324 if (ifa->ifa_addr->sa_family !=
328 sin = (struct sockaddr_in *) ifa->ifa_addr;
330 if (sin->sin_addr.s_addr ==
331 ipo->ip_src.s_addr) {
332 ipipstat.ipips_spoof++;
341 if (ifa->ifa_addr->sa_family !=
345 sin6 = (struct sockaddr_in6 *) ifa->ifa_addr;
347 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_src)) {
348 ipipstat.ipips_spoof++;
360 ipipstat.ipips_ibytes += m->m_pkthdr.len - iphlen;
363 * Interface pointer stays the same; if no IPsec processing has
364 * been done (or will be done), this will point to a normal
365 * interface. Otherwise, it'll point to an enc interface, which
366 * will allow a packet filter to distinguish between secure and
384 panic("ipip_input: should never reach here");
387 if (!IF_HANDOFF(ifq, m, NULL)) {
388 ipipstat.ipips_qfull++;
390 DPRINTF(("ipip_input: packet dropped because of full queue\n"));
399 struct ipsecrequest *isr,
405 struct secasvar *sav;
407 struct secasindex *saidx;
414 struct ip6_hdr *ip6, *ip6o;
417 SPLASSERT(net, "ipip_output");
420 KASSERT(sav != NULL, ("ipip_output: null SA"));
421 KASSERT(sav->sah != NULL, ("ipip_output: null SAH"));
423 /* XXX Deal with empty TDB source/destination addresses. */
425 m_copydata(m, 0, 1, &tp);
426 tp = (tp >> 4) & 0xff; /* Get the IP version number. */
428 saidx = &sav->sah->saidx;
429 switch (saidx->dst.sa.sa_family) {
432 if (saidx->src.sa.sa_family != AF_INET ||
433 saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
434 saidx->dst.sin.sin_addr.s_addr == INADDR_ANY) {
435 DPRINTF(("ipip_output: unspecified tunnel endpoint "
436 "address in SA %s/%08lx\n",
437 ipsec_address(&saidx->dst),
438 (u_long) ntohl(sav->spi)));
439 ipipstat.ipips_unspec++;
444 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
446 DPRINTF(("ipip_output: M_PREPEND failed\n"));
447 ipipstat.ipips_hdrops++;
452 ipo = mtod(m, struct ip *);
454 ipo->ip_v = IPVERSION;
456 ipo->ip_len = htons(m->m_pkthdr.len);
457 ipo->ip_ttl = ip_defttl;
459 ipo->ip_src = saidx->src.sin.sin_addr;
460 ipo->ip_dst = saidx->dst.sin.sin_addr;
463 ipo->ip_id = ip_randomid();
465 ipo->ip_id = htons(ip_id++);
468 /* If the inner protocol is IP... */
469 if (tp == IPVERSION) {
470 /* Save ECN notification */
471 m_copydata(m, sizeof(struct ip) +
472 offsetof(struct ip, ip_tos),
473 sizeof(u_int8_t), (caddr_t) &itos);
475 ipo->ip_p = IPPROTO_IPIP;
478 * We should be keeping tunnel soft-state and
479 * send back ICMPs if needed.
481 m_copydata(m, sizeof(struct ip) +
482 offsetof(struct ip, ip_off),
483 sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
484 ipo->ip_off = ntohs(ipo->ip_off);
485 ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
486 ipo->ip_off = htons(ipo->ip_off);
489 else if (tp == (IPV6_VERSION >> 4)) {
492 /* Save ECN notification. */
493 m_copydata(m, sizeof(struct ip) +
494 offsetof(struct ip6_hdr, ip6_flow),
495 sizeof(u_int32_t), (caddr_t) &itos32);
496 itos = ntohl(itos32) >> 20;
497 ipo->ip_p = IPPROTO_IPV6;
506 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
513 if (IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr) ||
514 saidx->src.sa.sa_family != AF_INET6 ||
515 IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr)) {
516 DPRINTF(("ipip_output: unspecified tunnel endpoint "
517 "address in SA %s/%08lx\n",
518 ipsec_address(&saidx->dst),
519 (u_long) ntohl(sav->spi)));
520 ipipstat.ipips_unspec++;
525 /* scoped address handling */
526 ip6 = mtod(m, struct ip6_hdr *);
527 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
528 ip6->ip6_src.s6_addr16[1] = 0;
529 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
530 ip6->ip6_dst.s6_addr16[1] = 0;
532 M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
534 DPRINTF(("ipip_output: M_PREPEND failed\n"));
535 ipipstat.ipips_hdrops++;
541 /* Initialize IPv6 header */
542 ip6o = mtod(m, struct ip6_hdr *);
544 ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
545 ip6o->ip6_vfc |= IPV6_VERSION;
546 ip6o->ip6_plen = htons(m->m_pkthdr.len);
547 ip6o->ip6_hlim = ip_defttl;
548 ip6o->ip6_dst = saidx->dst.sin6.sin6_addr;
549 ip6o->ip6_src = saidx->src.sin6.sin6_addr;
552 if (tp == IPVERSION) {
553 /* Save ECN notification */
554 m_copydata(m, sizeof(struct ip6_hdr) +
555 offsetof(struct ip, ip_tos), sizeof(u_int8_t),
558 /* This is really IPVERSION. */
559 ip6o->ip6_nxt = IPPROTO_IPIP;
562 if (tp == (IPV6_VERSION >> 4)) {
565 /* Save ECN notification. */
566 m_copydata(m, sizeof(struct ip6_hdr) +
567 offsetof(struct ip6_hdr, ip6_flow),
568 sizeof(u_int32_t), (caddr_t) &itos32);
569 itos = ntohl(itos32) >> 20;
571 ip6o->ip6_nxt = IPPROTO_IPV6;
577 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
578 ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
584 DPRINTF(("ipip_output: unsupported protocol family %u\n",
585 saidx->dst.sa.sa_family));
586 ipipstat.ipips_family++;
587 error = EAFNOSUPPORT; /* XXX diffs from openbsd */
591 ipipstat.ipips_opackets++;
595 if (saidx->dst.sa.sa_family == AF_INET) {
597 if (sav->tdb_xform->xf_type == XF_IP4)
598 tdb->tdb_cur_bytes +=
599 m->m_pkthdr.len - sizeof(struct ip);
601 ipipstat.ipips_obytes += m->m_pkthdr.len - sizeof(struct ip);
606 if (saidx->dst.sa.sa_family == AF_INET6) {
608 if (sav->tdb_xform->xf_type == XF_IP4)
609 tdb->tdb_cur_bytes +=
610 m->m_pkthdr.len - sizeof(struct ip6_hdr);
612 ipipstat.ipips_obytes +=
613 m->m_pkthdr.len - sizeof(struct ip6_hdr);
620 m_freem(m), *mp = NULL;
626 ipe4_init(struct secasvar *sav, struct xformsw *xsp)
628 sav->tdb_xform = xsp;
633 ipe4_zeroize(struct secasvar *sav)
635 sav->tdb_xform = NULL;
640 ipe4_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff)
642 /* This is a rather serious mistake, so no conditional printing. */
643 printf("ipe4_input: should never be called\n");
649 static struct xformsw ipe4_xformsw = {
650 XF_IP4, 0, "IPv4 Simple Encapsulation",
651 ipe4_init, ipe4_zeroize, ipe4_input, ipip_output,
654 extern struct domain inetdomain;
655 static struct ipprotosw ipe4_protosw[] = {
656 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR|PR_LASTHDR,
657 (pr_in_input_t*) ip4_input,
664 { SOCK_RAW, &inetdomain, IPPROTO_IPV6, PR_ATOMIC|PR_ADDR|PR_LASTHDR,
665 (pr_in_input_t*) ip4_input,
675 * Check the encapsulated packet to see if we want it
678 ipe4_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
681 * Only take packets coming from IPSEC tunnels; the rest
682 * must be handled by the gif tunnel code. Note that we
683 * also return a minimum priority when we want the packet
684 * so any explicit gif tunnels take precedence.
686 return ((m->m_flags & M_IPSEC) != 0 ? 1 : 0);
692 xform_register(&ipe4_xformsw);
693 /* attach to encapsulation framework */
694 /* XXX save return cookie for detach on module remove */
695 (void) encap_attach_func(AF_INET, -1,
696 ipe4_encapcheck, (struct protosw*) &ipe4_protosw[0], NULL);
698 (void) encap_attach_func(AF_INET6, -1,
699 ipe4_encapcheck, (struct protosw*) &ipe4_protosw[1], NULL);
702 SYSINIT(ipe4_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipe4_attach, NULL);
703 #endif /* FAST_IPSEC */