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>
90 * We can control the acceptance of IP4 packets by altering the sysctl
91 * net.inet.ipip.allow value. Zero means drop them, all else is acceptance.
94 struct ipipstat ipipstat;
96 SYSCTL_DECL(_net_inet_ipip);
97 SYSCTL_INT(_net_inet_ipip, OID_AUTO,
98 ipip_allow, CTLFLAG_RW, &ipip_allow, 0, "");
99 SYSCTL_STRUCT(_net_inet_ipip, IPSECCTL_STATS,
100 stats, CTLFLAG_RD, &ipipstat, ipipstat, "");
103 #define M_IPSEC (M_AUTHIPHDR|M_AUTHIPDGM|M_DECRYPTED)
105 static void _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp);
109 * Really only a wrapper for ipip_input(), for use with IPv6.
112 ip4_input6(struct mbuf **m, int *offp, int proto)
115 /* If we do not accept IP-in-IP explicitly, drop. */
116 if (!ipip_allow && ((*m)->m_flags & M_IPSEC) == 0) {
117 DPRINTF(("%s: dropped due to policy\n", __func__));
118 ipipstat.ipips_pdrops++;
123 _ipip_input(*m, *offp, NULL);
130 * Really only a wrapper for ipip_input(), for use with IPv4.
133 ip4_input(struct mbuf *m, ...)
139 /* If we do not accept IP-in-IP explicitly, drop. */
140 if (!ipip_allow && (m->m_flags & M_IPSEC) == 0) {
141 DPRINTF(("%s: dropped due to policy\n", __func__));
142 ipipstat.ipips_pdrops++;
148 iphlen = va_arg(ap, int);
151 _ipip_input(m, iphlen, NULL);
156 * ipip_input gets called when we receive an IP{46} encapsulated packet,
157 * either because we got it at a real interface, or because AH or ESP
158 * were being used in tunnel mode (in which case the rcvif element will
159 * contain the address of the encX interface associated with the tunnel.
163 _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp)
165 register struct sockaddr_in *sin;
166 register struct ifnet *ifp;
167 register struct ifaddr *ifa;
170 register struct sockaddr_in6 *sin6;
171 struct ip6_hdr *ip6 = NULL;
180 ipipstat.ipips_ipackets++;
182 m_copydata(m, 0, 1, &v);
187 hlen = sizeof(struct ip);
192 hlen = sizeof(struct ip6_hdr);
196 ipipstat.ipips_family++;
198 return /* EAFNOSUPPORT */;
201 /* Bring the IP header in the first mbuf, if not there already */
202 if (m->m_len < hlen) {
203 if ((m = m_pullup(m, hlen)) == NULL) {
204 DPRINTF(("%s: m_pullup (1) failed\n", __func__));
205 ipipstat.ipips_hdrops++;
210 ipo = mtod(m, struct ip *);
213 if (ipo->ip_v == IPVERSION && ipo->ip_p == IPPROTO_IPV4) {
214 if (IN_MULTICAST(((struct ip *)((char *) ipo + iphlen))->ip_dst.s_addr)) {
215 ipip_mroute_input (m, iphlen);
219 #endif /* MROUTING */
221 /* Keep outer ecn field. */
230 otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
234 panic("ipip_input: unknown ip version %u (outer)", v>>4);
237 /* Remove outer IP header */
241 if (m->m_pkthdr.len < sizeof(struct ip)) {
242 ipipstat.ipips_hdrops++;
247 m_copydata(m, 0, 1, &v);
252 hlen = sizeof(struct ip);
258 hlen = sizeof(struct ip6_hdr);
262 ipipstat.ipips_family++;
264 return; /* EAFNOSUPPORT */
268 * Bring the inner IP header in the first mbuf, if not there already.
270 if (m->m_len < hlen) {
271 if ((m = m_pullup(m, hlen)) == NULL) {
272 DPRINTF(("%s: m_pullup (2) failed\n", __func__));
273 ipipstat.ipips_hdrops++;
279 * RFC 1853 specifies that the inner TTL should not be touched on
280 * decapsulation. There's no reason this comment should be here, but
281 * this is as good as any a position.
284 /* Some sanity checks in the inner IP header */
288 ipo = mtod(m, struct ip *);
290 ip_ecn_egress(ip4_ipsec_ecn, &otos, &ipo->ip_tos);
295 ip6 = (struct ip6_hdr *) ipo;
297 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
298 ip_ecn_egress(ip6_ipsec_ecn, &otos, &itos);
299 ip6->ip6_flow &= ~htonl(0xff << 20);
300 ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
304 panic("ipip_input: unknown ip version %u (inner)", v>>4);
307 /* Check for local address spoofing. */
308 if ((m->m_pkthdr.rcvif == NULL ||
309 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
312 for (ifp = ifnet.tqh_first; ifp != 0;
313 ifp = ifp->if_list.tqe_next) {
314 for (ifa = ifp->if_addrlist.tqh_first; ifa != 0;
315 ifa = ifa->ifa_list.tqe_next) {
318 if (ifa->ifa_addr->sa_family !=
322 sin = (struct sockaddr_in *) ifa->ifa_addr;
324 if (sin->sin_addr.s_addr ==
325 ipo->ip_src.s_addr) {
326 ipipstat.ipips_spoof++;
336 if (ifa->ifa_addr->sa_family !=
340 sin6 = (struct sockaddr_in6 *) ifa->ifa_addr;
342 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_src)) {
343 ipipstat.ipips_spoof++;
357 ipipstat.ipips_ibytes += m->m_pkthdr.len - iphlen;
360 * Interface pointer stays the same; if no IPsec processing has
361 * been done (or will be done), this will point to a normal
362 * interface. Otherwise, it'll point to an enc interface, which
363 * will allow a packet filter to distinguish between secure and
379 panic("%s: bogus ip version %u", __func__, v>>4);
382 if (!netisr_queue(isr, m)) {
383 ipipstat.ipips_qfull++;
384 DPRINTF(("%s: packet dropped because of full queue\n",
392 struct ipsecrequest *isr,
398 struct secasvar *sav;
400 struct secasindex *saidx;
407 struct ip6_hdr *ip6, *ip6o;
410 IPSEC_SPLASSERT_SOFTNET(__func__);
413 IPSEC_ASSERT(sav != NULL, ("null SA"));
414 IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
416 /* XXX Deal with empty TDB source/destination addresses. */
418 m_copydata(m, 0, 1, &tp);
419 tp = (tp >> 4) & 0xff; /* Get the IP version number. */
421 saidx = &sav->sah->saidx;
422 switch (saidx->dst.sa.sa_family) {
425 if (saidx->src.sa.sa_family != AF_INET ||
426 saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
427 saidx->dst.sin.sin_addr.s_addr == INADDR_ANY) {
428 DPRINTF(("%s: unspecified tunnel endpoint "
429 "address in SA %s/%08lx\n", __func__,
430 ipsec_address(&saidx->dst),
431 (u_long) ntohl(sav->spi)));
432 ipipstat.ipips_unspec++;
437 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
439 DPRINTF(("%s: M_PREPEND failed\n", __func__));
440 ipipstat.ipips_hdrops++;
445 ipo = mtod(m, struct ip *);
447 ipo->ip_v = IPVERSION;
449 ipo->ip_len = htons(m->m_pkthdr.len);
450 ipo->ip_ttl = ip_defttl;
452 ipo->ip_src = saidx->src.sin.sin_addr;
453 ipo->ip_dst = saidx->dst.sin.sin_addr;
456 ipo->ip_id = ip_randomid();
458 ipo->ip_id = htons(ip_id++);
461 /* If the inner protocol is IP... */
462 if (tp == IPVERSION) {
463 /* Save ECN notification */
464 m_copydata(m, sizeof(struct ip) +
465 offsetof(struct ip, ip_tos),
466 sizeof(u_int8_t), (caddr_t) &itos);
468 ipo->ip_p = IPPROTO_IPIP;
471 * We should be keeping tunnel soft-state and
472 * send back ICMPs if needed.
474 m_copydata(m, sizeof(struct ip) +
475 offsetof(struct ip, ip_off),
476 sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
477 ipo->ip_off = ntohs(ipo->ip_off);
478 ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
479 ipo->ip_off = htons(ipo->ip_off);
482 else if (tp == (IPV6_VERSION >> 4)) {
485 /* Save ECN notification. */
486 m_copydata(m, sizeof(struct ip) +
487 offsetof(struct ip6_hdr, ip6_flow),
488 sizeof(u_int32_t), (caddr_t) &itos32);
489 itos = ntohl(itos32) >> 20;
490 ipo->ip_p = IPPROTO_IPV6;
499 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
506 if (IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr) ||
507 saidx->src.sa.sa_family != AF_INET6 ||
508 IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr)) {
509 DPRINTF(("%s: unspecified tunnel endpoint "
510 "address in SA %s/%08lx\n", __func__,
511 ipsec_address(&saidx->dst),
512 (u_long) ntohl(sav->spi)));
513 ipipstat.ipips_unspec++;
518 /* scoped address handling */
519 ip6 = mtod(m, struct ip6_hdr *);
520 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
521 ip6->ip6_src.s6_addr16[1] = 0;
522 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
523 ip6->ip6_dst.s6_addr16[1] = 0;
525 M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
527 DPRINTF(("%s: M_PREPEND failed\n", __func__));
528 ipipstat.ipips_hdrops++;
534 /* Initialize IPv6 header */
535 ip6o = mtod(m, struct ip6_hdr *);
537 ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
538 ip6o->ip6_vfc |= IPV6_VERSION;
539 ip6o->ip6_plen = htons(m->m_pkthdr.len);
540 ip6o->ip6_hlim = ip_defttl;
541 ip6o->ip6_dst = saidx->dst.sin6.sin6_addr;
542 ip6o->ip6_src = saidx->src.sin6.sin6_addr;
545 if (tp == IPVERSION) {
546 /* Save ECN notification */
547 m_copydata(m, sizeof(struct ip6_hdr) +
548 offsetof(struct ip, ip_tos), sizeof(u_int8_t),
551 /* This is really IPVERSION. */
552 ip6o->ip6_nxt = IPPROTO_IPIP;
555 if (tp == (IPV6_VERSION >> 4)) {
558 /* Save ECN notification. */
559 m_copydata(m, sizeof(struct ip6_hdr) +
560 offsetof(struct ip6_hdr, ip6_flow),
561 sizeof(u_int32_t), (caddr_t) &itos32);
562 itos = ntohl(itos32) >> 20;
564 ip6o->ip6_nxt = IPPROTO_IPV6;
570 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
571 ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
577 DPRINTF(("%s: unsupported protocol family %u\n", __func__,
578 saidx->dst.sa.sa_family));
579 ipipstat.ipips_family++;
580 error = EAFNOSUPPORT; /* XXX diffs from openbsd */
584 ipipstat.ipips_opackets++;
588 if (saidx->dst.sa.sa_family == AF_INET) {
590 if (sav->tdb_xform->xf_type == XF_IP4)
591 tdb->tdb_cur_bytes +=
592 m->m_pkthdr.len - sizeof(struct ip);
594 ipipstat.ipips_obytes += m->m_pkthdr.len - sizeof(struct ip);
599 if (saidx->dst.sa.sa_family == AF_INET6) {
601 if (sav->tdb_xform->xf_type == XF_IP4)
602 tdb->tdb_cur_bytes +=
603 m->m_pkthdr.len - sizeof(struct ip6_hdr);
605 ipipstat.ipips_obytes +=
606 m->m_pkthdr.len - sizeof(struct ip6_hdr);
613 m_freem(m), *mp = NULL;
619 ipe4_init(struct secasvar *sav, struct xformsw *xsp)
621 sav->tdb_xform = xsp;
626 ipe4_zeroize(struct secasvar *sav)
628 sav->tdb_xform = NULL;
633 ipe4_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff)
635 /* This is a rather serious mistake, so no conditional printing. */
636 printf("%s: should never be called\n", __func__);
642 static struct xformsw ipe4_xformsw = {
643 XF_IP4, 0, "IPv4 Simple Encapsulation",
644 ipe4_init, ipe4_zeroize, ipe4_input, ipip_output,
647 extern struct domain inetdomain;
648 static struct ipprotosw ipe4_protosw[] = {
649 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR|PR_LASTHDR,
650 (pr_in_input_t*) ip4_input,
657 { SOCK_RAW, &inetdomain, IPPROTO_IPV6, PR_ATOMIC|PR_ADDR|PR_LASTHDR,
658 (pr_in_input_t*) ip4_input,
668 * Check the encapsulated packet to see if we want it
671 ipe4_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
674 * Only take packets coming from IPSEC tunnels; the rest
675 * must be handled by the gif tunnel code. Note that we
676 * also return a minimum priority when we want the packet
677 * so any explicit gif tunnels take precedence.
679 return ((m->m_flags & M_IPSEC) != 0 ? 1 : 0);
685 xform_register(&ipe4_xformsw);
686 /* attach to encapsulation framework */
687 /* XXX save return cookie for detach on module remove */
688 (void) encap_attach_func(AF_INET, -1,
689 ipe4_encapcheck, (struct protosw*) &ipe4_protosw[0], NULL);
691 (void) encap_attach_func(AF_INET6, -1,
692 ipe4_encapcheck, (struct protosw*) &ipe4_protosw[1], NULL);
695 SYSINIT(ipe4_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipe4_attach, NULL);
696 #endif /* FAST_IPSEC */