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"
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>
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, int off)
135 /* If we do not accept IP-in-IP explicitly, drop. */
136 if (!ipip_allow && (m->m_flags & M_IPSEC) == 0) {
137 DPRINTF(("%s: dropped due to policy\n", __func__));
138 ipipstat.ipips_pdrops++;
143 _ipip_input(m, off, NULL);
148 * ipip_input gets called when we receive an IP{46} encapsulated packet,
149 * either because we got it at a real interface, or because AH or ESP
150 * were being used in tunnel mode (in which case the rcvif element will
151 * contain the address of the encX interface associated with the tunnel.
155 _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp)
157 register struct sockaddr_in *sin;
158 register struct ifnet *ifp;
159 register struct ifaddr *ifa;
162 register struct sockaddr_in6 *sin6;
163 struct ip6_hdr *ip6 = NULL;
172 ipipstat.ipips_ipackets++;
174 m_copydata(m, 0, 1, &v);
179 hlen = sizeof(struct ip);
184 hlen = sizeof(struct ip6_hdr);
188 ipipstat.ipips_family++;
190 return /* EAFNOSUPPORT */;
193 /* Bring the IP header in the first mbuf, if not there already */
194 if (m->m_len < hlen) {
195 if ((m = m_pullup(m, hlen)) == NULL) {
196 DPRINTF(("%s: m_pullup (1) failed\n", __func__));
197 ipipstat.ipips_hdrops++;
202 ipo = mtod(m, struct ip *);
205 if (ipo->ip_v == IPVERSION && ipo->ip_p == IPPROTO_IPV4) {
206 if (IN_MULTICAST(((struct ip *)((char *) ipo + iphlen))->ip_dst.s_addr)) {
207 ipip_mroute_input (m, iphlen);
211 #endif /* MROUTING */
213 /* Keep outer ecn field. */
222 otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
226 panic("ipip_input: unknown ip version %u (outer)", v>>4);
229 /* Remove outer IP header */
233 if (m->m_pkthdr.len < sizeof(struct ip)) {
234 ipipstat.ipips_hdrops++;
239 m_copydata(m, 0, 1, &v);
244 hlen = sizeof(struct ip);
250 hlen = sizeof(struct ip6_hdr);
254 ipipstat.ipips_family++;
256 return; /* EAFNOSUPPORT */
260 * Bring the inner IP header in the first mbuf, if not there already.
262 if (m->m_len < hlen) {
263 if ((m = m_pullup(m, hlen)) == NULL) {
264 DPRINTF(("%s: m_pullup (2) failed\n", __func__));
265 ipipstat.ipips_hdrops++;
271 * RFC 1853 specifies that the inner TTL should not be touched on
272 * decapsulation. There's no reason this comment should be here, but
273 * this is as good as any a position.
276 /* Some sanity checks in the inner IP header */
280 ipo = mtod(m, struct ip *);
282 ip_ecn_egress(ip4_ipsec_ecn, &otos, &ipo->ip_tos);
287 ip6 = (struct ip6_hdr *) ipo;
289 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
290 ip_ecn_egress(ip6_ipsec_ecn, &otos, &itos);
291 ip6->ip6_flow &= ~htonl(0xff << 20);
292 ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
296 panic("ipip_input: unknown ip version %u (inner)", v>>4);
299 /* Check for local address spoofing. */
300 if ((m->m_pkthdr.rcvif == NULL ||
301 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
304 TAILQ_FOREACH(ifp, &ifnet, if_link) {
305 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
308 if (ifa->ifa_addr->sa_family !=
312 sin = (struct sockaddr_in *) ifa->ifa_addr;
314 if (sin->sin_addr.s_addr ==
315 ipo->ip_src.s_addr) {
316 ipipstat.ipips_spoof++;
326 if (ifa->ifa_addr->sa_family !=
330 sin6 = (struct sockaddr_in6 *) ifa->ifa_addr;
332 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_src)) {
333 ipipstat.ipips_spoof++;
347 ipipstat.ipips_ibytes += m->m_pkthdr.len - iphlen;
350 /* pass the mbuf to enc0 for packet filtering */
351 if (ipsec_filter(&m, 1) != 0)
356 * Interface pointer stays the same; if no IPsec processing has
357 * been done (or will be done), this will point to a normal
358 * interface. Otherwise, it'll point to an enc interface, which
359 * will allow a packet filter to distinguish between secure and
375 panic("%s: bogus ip version %u", __func__, v>>4);
378 if (netisr_queue(isr, m)) { /* (0) on success. */
379 ipipstat.ipips_qfull++;
380 DPRINTF(("%s: packet dropped because of full queue\n",
388 struct ipsecrequest *isr,
394 struct secasvar *sav;
396 struct secasindex *saidx;
403 struct ip6_hdr *ip6, *ip6o;
406 IPSEC_SPLASSERT_SOFTNET(__func__);
409 IPSEC_ASSERT(sav != NULL, ("null SA"));
410 IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
412 /* XXX Deal with empty TDB source/destination addresses. */
414 m_copydata(m, 0, 1, &tp);
415 tp = (tp >> 4) & 0xff; /* Get the IP version number. */
417 saidx = &sav->sah->saidx;
418 switch (saidx->dst.sa.sa_family) {
421 if (saidx->src.sa.sa_family != AF_INET ||
422 saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
423 saidx->dst.sin.sin_addr.s_addr == INADDR_ANY) {
424 DPRINTF(("%s: unspecified tunnel endpoint "
425 "address in SA %s/%08lx\n", __func__,
426 ipsec_address(&saidx->dst),
427 (u_long) ntohl(sav->spi)));
428 ipipstat.ipips_unspec++;
433 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
435 DPRINTF(("%s: M_PREPEND failed\n", __func__));
436 ipipstat.ipips_hdrops++;
441 ipo = mtod(m, struct ip *);
443 ipo->ip_v = IPVERSION;
445 ipo->ip_len = htons(m->m_pkthdr.len);
446 ipo->ip_ttl = ip_defttl;
448 ipo->ip_src = saidx->src.sin.sin_addr;
449 ipo->ip_dst = saidx->dst.sin.sin_addr;
451 ipo->ip_id = ip_newid();
453 /* If the inner protocol is IP... */
454 if (tp == IPVERSION) {
455 /* Save ECN notification */
456 m_copydata(m, sizeof(struct ip) +
457 offsetof(struct ip, ip_tos),
458 sizeof(u_int8_t), (caddr_t) &itos);
460 ipo->ip_p = IPPROTO_IPIP;
463 * We should be keeping tunnel soft-state and
464 * send back ICMPs if needed.
466 m_copydata(m, sizeof(struct ip) +
467 offsetof(struct ip, ip_off),
468 sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
469 ipo->ip_off = ntohs(ipo->ip_off);
470 ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
471 ipo->ip_off = htons(ipo->ip_off);
474 else if (tp == (IPV6_VERSION >> 4)) {
477 /* Save ECN notification. */
478 m_copydata(m, sizeof(struct ip) +
479 offsetof(struct ip6_hdr, ip6_flow),
480 sizeof(u_int32_t), (caddr_t) &itos32);
481 itos = ntohl(itos32) >> 20;
482 ipo->ip_p = IPPROTO_IPV6;
491 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
498 if (IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr) ||
499 saidx->src.sa.sa_family != AF_INET6 ||
500 IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr)) {
501 DPRINTF(("%s: unspecified tunnel endpoint "
502 "address in SA %s/%08lx\n", __func__,
503 ipsec_address(&saidx->dst),
504 (u_long) ntohl(sav->spi)));
505 ipipstat.ipips_unspec++;
510 /* scoped address handling */
511 ip6 = mtod(m, struct ip6_hdr *);
512 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
513 ip6->ip6_src.s6_addr16[1] = 0;
514 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
515 ip6->ip6_dst.s6_addr16[1] = 0;
517 M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
519 DPRINTF(("%s: M_PREPEND failed\n", __func__));
520 ipipstat.ipips_hdrops++;
525 /* Initialize IPv6 header */
526 ip6o = mtod(m, struct ip6_hdr *);
528 ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
529 ip6o->ip6_vfc |= IPV6_VERSION;
530 ip6o->ip6_plen = htons(m->m_pkthdr.len);
531 ip6o->ip6_hlim = ip_defttl;
532 ip6o->ip6_dst = saidx->dst.sin6.sin6_addr;
533 ip6o->ip6_src = saidx->src.sin6.sin6_addr;
536 if (tp == IPVERSION) {
537 /* Save ECN notification */
538 m_copydata(m, sizeof(struct ip6_hdr) +
539 offsetof(struct ip, ip_tos), sizeof(u_int8_t),
542 /* This is really IPVERSION. */
543 ip6o->ip6_nxt = IPPROTO_IPIP;
546 if (tp == (IPV6_VERSION >> 4)) {
549 /* Save ECN notification. */
550 m_copydata(m, sizeof(struct ip6_hdr) +
551 offsetof(struct ip6_hdr, ip6_flow),
552 sizeof(u_int32_t), (caddr_t) &itos32);
553 itos = ntohl(itos32) >> 20;
555 ip6o->ip6_nxt = IPPROTO_IPV6;
561 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
562 ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
568 DPRINTF(("%s: unsupported protocol family %u\n", __func__,
569 saidx->dst.sa.sa_family));
570 ipipstat.ipips_family++;
571 error = EAFNOSUPPORT; /* XXX diffs from openbsd */
575 ipipstat.ipips_opackets++;
579 if (saidx->dst.sa.sa_family == AF_INET) {
581 if (sav->tdb_xform->xf_type == XF_IP4)
582 tdb->tdb_cur_bytes +=
583 m->m_pkthdr.len - sizeof(struct ip);
585 ipipstat.ipips_obytes += m->m_pkthdr.len - sizeof(struct ip);
590 if (saidx->dst.sa.sa_family == AF_INET6) {
592 if (sav->tdb_xform->xf_type == XF_IP4)
593 tdb->tdb_cur_bytes +=
594 m->m_pkthdr.len - sizeof(struct ip6_hdr);
596 ipipstat.ipips_obytes +=
597 m->m_pkthdr.len - sizeof(struct ip6_hdr);
611 ipe4_init(struct secasvar *sav, struct xformsw *xsp)
613 sav->tdb_xform = xsp;
618 ipe4_zeroize(struct secasvar *sav)
620 sav->tdb_xform = NULL;
625 ipe4_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff)
627 /* This is a rather serious mistake, so no conditional printing. */
628 printf("%s: should never be called\n", __func__);
634 static struct xformsw ipe4_xformsw = {
635 XF_IP4, 0, "IPv4 Simple Encapsulation",
636 ipe4_init, ipe4_zeroize, ipe4_input, ipip_output,
639 extern struct domain inetdomain;
640 static struct protosw ipe4_protosw =
641 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR|PR_LASTHDR,
649 static struct ip6protosw ipe6_protosw =
650 { SOCK_RAW, &inetdomain, IPPROTO_IPV6, PR_ATOMIC|PR_ADDR|PR_LASTHDR,
660 * Check the encapsulated packet to see if we want it
663 ipe4_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
666 * Only take packets coming from IPSEC tunnels; the rest
667 * must be handled by the gif tunnel code. Note that we
668 * also return a minimum priority when we want the packet
669 * so any explicit gif tunnels take precedence.
671 return ((m->m_flags & M_IPSEC) != 0 ? 1 : 0);
677 xform_register(&ipe4_xformsw);
678 /* attach to encapsulation framework */
679 /* XXX save return cookie for detach on module remove */
680 (void) encap_attach_func(AF_INET, -1,
681 ipe4_encapcheck, &ipe4_protosw, NULL);
683 (void) encap_attach_func(AF_INET6, -1,
684 ipe4_encapcheck, (struct protosw *)&ipe6_protosw, NULL);
687 SYSINIT(ipe4_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipe4_attach, NULL);