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/if_var.h>
57 #include <net/route.h>
58 #include <net/netisr.h>
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/in_var.h>
64 #include <netinet/ip.h>
65 #include <netinet/ip_ecn.h>
66 #include <netinet/ip_var.h>
67 #include <netinet/ip_encap.h>
69 #include <netinet/ip_mroute.h>
72 #include <netipsec/ipsec.h>
73 #include <netipsec/xform.h>
75 #include <netipsec/ipip_var.h>
78 #include <netinet/ip6.h>
79 #include <netipsec/ipsec6.h>
80 #include <netinet6/ip6_ecn.h>
81 #include <netinet6/in6_var.h>
82 #include <netinet6/ip6protosw.h>
85 #include <netipsec/key.h>
86 #include <netipsec/key_debug.h>
88 #include <machine/stdarg.h>
91 * We can control the acceptance of IP4 packets by altering the sysctl
92 * net.inet.ipip.allow value. Zero means drop them, all else is acceptance.
94 VNET_DEFINE(int, ipip_allow) = 0;
95 VNET_PCPUSTAT_DEFINE(struct ipipstat, ipipstat);
96 VNET_PCPUSTAT_SYSINIT(ipipstat);
99 VNET_PCPUSTAT_SYSUNINIT(ipipstat);
102 SYSCTL_DECL(_net_inet_ipip);
103 SYSCTL_VNET_INT(_net_inet_ipip, OID_AUTO,
104 ipip_allow, CTLFLAG_RW, &VNET_NAME(ipip_allow), 0, "");
105 SYSCTL_VNET_PCPUSTAT(_net_inet_ipip, IPSECCTL_STATS, stats,
106 struct ipipstat, ipipstat,
107 "IPIP statistics (struct ipipstat, netipsec/ipip_var.h)");
110 #define M_IPSEC (M_AUTHIPHDR|M_AUTHIPDGM|M_DECRYPTED)
112 static void _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp);
116 * Really only a wrapper for ipip_input(), for use with IPv6.
119 ip4_input6(struct mbuf **m, int *offp, int proto)
122 /* If we do not accept IP-in-IP explicitly, drop. */
123 if (!V_ipip_allow && ((*m)->m_flags & M_IPSEC) == 0) {
124 DPRINTF(("%s: dropped due to policy\n", __func__));
125 IPIPSTAT_INC(ipips_pdrops);
130 _ipip_input(*m, *offp, NULL);
137 * Really only a wrapper for ipip_input(), for use with IPv4.
140 ip4_input(struct mbuf *m, int off)
143 /* If we do not accept IP-in-IP explicitly, drop. */
144 if (!V_ipip_allow && (m->m_flags & M_IPSEC) == 0) {
145 DPRINTF(("%s: dropped due to policy\n", __func__));
146 IPIPSTAT_INC(ipips_pdrops);
151 _ipip_input(m, off, 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)
166 register struct sockaddr_in *sin;
168 register struct ifnet *ifp;
169 register struct ifaddr *ifa;
172 register struct sockaddr_in6 *sin6;
173 struct ip6_hdr *ip6 = NULL;
182 IPIPSTAT_INC(ipips_ipackets);
184 m_copydata(m, 0, 1, &v);
189 hlen = sizeof(struct ip);
194 hlen = sizeof(struct ip6_hdr);
198 IPIPSTAT_INC(ipips_family);
200 return /* EAFNOSUPPORT */;
203 /* Bring the IP header in the first mbuf, if not there already */
204 if (m->m_len < hlen) {
205 if ((m = m_pullup(m, hlen)) == NULL) {
206 DPRINTF(("%s: m_pullup (1) failed\n", __func__));
207 IPIPSTAT_INC(ipips_hdrops);
212 ipo = mtod(m, struct ip *);
215 if (ipo->ip_v == IPVERSION && ipo->ip_p == IPPROTO_IPV4) {
216 if (IN_MULTICAST(((struct ip *)((char *) ipo + iphlen))->ip_dst.s_addr)) {
217 ipip_mroute_input (m, iphlen);
221 #endif /* MROUTING */
223 /* Keep outer ecn field. */
232 otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
236 panic("ipip_input: unknown ip version %u (outer)", v>>4);
239 /* Remove outer IP header */
243 if (m->m_pkthdr.len < sizeof(struct ip)) {
244 IPIPSTAT_INC(ipips_hdrops);
249 m_copydata(m, 0, 1, &v);
254 hlen = sizeof(struct ip);
260 hlen = sizeof(struct ip6_hdr);
264 IPIPSTAT_INC(ipips_family);
266 return; /* EAFNOSUPPORT */
270 * Bring the inner IP header in the first mbuf, if not there already.
272 if (m->m_len < hlen) {
273 if ((m = m_pullup(m, hlen)) == NULL) {
274 DPRINTF(("%s: m_pullup (2) failed\n", __func__));
275 IPIPSTAT_INC(ipips_hdrops);
281 * RFC 1853 specifies that the inner TTL should not be touched on
282 * decapsulation. There's no reason this comment should be here, but
283 * this is as good as any a position.
286 /* Some sanity checks in the inner IP header */
290 ipo = mtod(m, struct ip *);
292 ip_ecn_egress(V_ip4_ipsec_ecn, &otos, &ipo->ip_tos);
297 ip6 = (struct ip6_hdr *) ipo;
299 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
300 ip_ecn_egress(V_ip6_ipsec_ecn, &otos, &itos);
301 ip6->ip6_flow &= ~htonl(0xff << 20);
302 ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
306 panic("ipip_input: unknown ip version %u (inner)", v>>4);
309 /* Check for local address spoofing. */
310 if ((m->m_pkthdr.rcvif == NULL ||
311 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
313 IFNET_RLOCK_NOSLEEP();
314 TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
315 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
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_INC(ipips_spoof);
328 IFNET_RUNLOCK_NOSLEEP();
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_INC(ipips_spoof);
345 IFNET_RUNLOCK_NOSLEEP();
353 IFNET_RUNLOCK_NOSLEEP();
357 IPIPSTAT_ADD(ipips_ibytes, m->m_pkthdr.len - iphlen);
363 ipsec_bpf(m, NULL, AF_INET, ENC_IN|ENC_AFTER);
368 ipsec_bpf(m, NULL, AF_INET6, ENC_IN|ENC_AFTER);
372 panic("%s: bogus ip version %u", __func__, v>>4);
374 /* pass the mbuf to enc0 for packet filtering */
375 if (ipsec_filter(&m, PFIL_IN, ENC_IN|ENC_AFTER) != 0)
380 * Interface pointer stays the same; if no IPsec processing has
381 * been done (or will be done), this will point to a normal
382 * interface. Otherwise, it'll point to an enc interface, which
383 * will allow a packet filter to distinguish between secure and
399 panic("%s: bogus ip version %u", __func__, v>>4);
402 if (netisr_queue(isr, m)) { /* (0) on success. */
403 IPIPSTAT_INC(ipips_qfull);
404 DPRINTF(("%s: packet dropped because of full queue\n",
412 struct ipsecrequest *isr,
418 struct secasvar *sav;
420 struct secasindex *saidx;
422 #if defined(INET) || defined(INET6)
429 struct ip6_hdr *ip6, *ip6o;
433 IPSEC_ASSERT(sav != NULL, ("null SA"));
434 IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
436 /* XXX Deal with empty TDB source/destination addresses. */
438 m_copydata(m, 0, 1, &tp);
439 tp = (tp >> 4) & 0xff; /* Get the IP version number. */
441 saidx = &sav->sah->saidx;
442 switch (saidx->dst.sa.sa_family) {
445 if (saidx->src.sa.sa_family != AF_INET ||
446 saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
447 saidx->dst.sin.sin_addr.s_addr == INADDR_ANY) {
448 DPRINTF(("%s: unspecified tunnel endpoint "
449 "address in SA %s/%08lx\n", __func__,
450 ipsec_address(&saidx->dst),
451 (u_long) ntohl(sav->spi)));
452 IPIPSTAT_INC(ipips_unspec);
457 M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
459 DPRINTF(("%s: M_PREPEND failed\n", __func__));
460 IPIPSTAT_INC(ipips_hdrops);
465 ipo = mtod(m, struct ip *);
467 ipo->ip_v = IPVERSION;
469 ipo->ip_len = htons(m->m_pkthdr.len);
470 ipo->ip_ttl = V_ip_defttl;
472 ipo->ip_src = saidx->src.sin.sin_addr;
473 ipo->ip_dst = saidx->dst.sin.sin_addr;
475 ipo->ip_id = ip_newid();
477 /* If the inner protocol is IP... */
480 /* Save ECN notification */
481 m_copydata(m, sizeof(struct ip) +
482 offsetof(struct ip, ip_tos),
483 sizeof(u_int8_t), (caddr_t) &itos);
485 ipo->ip_p = IPPROTO_IPIP;
488 * We should be keeping tunnel soft-state and
489 * send back ICMPs if needed.
491 m_copydata(m, sizeof(struct ip) +
492 offsetof(struct ip, ip_off),
493 sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
494 ipo->ip_off = ntohs(ipo->ip_off);
495 ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
496 ipo->ip_off = htons(ipo->ip_off);
499 case (IPV6_VERSION >> 4):
503 /* Save ECN notification. */
504 m_copydata(m, sizeof(struct ip) +
505 offsetof(struct ip6_hdr, ip6_flow),
506 sizeof(u_int32_t), (caddr_t) &itos32);
507 itos = ntohl(itos32) >> 20;
508 ipo->ip_p = IPPROTO_IPV6;
518 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
525 if (IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr) ||
526 saidx->src.sa.sa_family != AF_INET6 ||
527 IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr)) {
528 DPRINTF(("%s: unspecified tunnel endpoint "
529 "address in SA %s/%08lx\n", __func__,
530 ipsec_address(&saidx->dst),
531 (u_long) ntohl(sav->spi)));
532 IPIPSTAT_INC(ipips_unspec);
537 /* scoped address handling */
538 ip6 = mtod(m, struct ip6_hdr *);
539 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
540 ip6->ip6_src.s6_addr16[1] = 0;
541 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
542 ip6->ip6_dst.s6_addr16[1] = 0;
544 M_PREPEND(m, sizeof(struct ip6_hdr), M_NOWAIT);
546 DPRINTF(("%s: M_PREPEND failed\n", __func__));
547 IPIPSTAT_INC(ipips_hdrops);
552 /* Initialize IPv6 header */
553 ip6o = mtod(m, struct ip6_hdr *);
555 ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
556 ip6o->ip6_vfc |= IPV6_VERSION;
557 ip6o->ip6_plen = htons(m->m_pkthdr.len);
558 ip6o->ip6_hlim = V_ip_defttl;
559 ip6o->ip6_dst = saidx->dst.sin6.sin6_addr;
560 ip6o->ip6_src = saidx->src.sin6.sin6_addr;
565 /* Save ECN notification */
566 m_copydata(m, sizeof(struct ip6_hdr) +
567 offsetof(struct ip, ip_tos), sizeof(u_int8_t),
570 /* This is really IPVERSION. */
571 ip6o->ip6_nxt = IPPROTO_IPIP;
574 case (IPV6_VERSION >> 4):
578 /* Save ECN notification. */
579 m_copydata(m, sizeof(struct ip6_hdr) +
580 offsetof(struct ip6_hdr, ip6_flow),
581 sizeof(u_int32_t), (caddr_t) &itos32);
582 itos = ntohl(itos32) >> 20;
584 ip6o->ip6_nxt = IPPROTO_IPV6;
592 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
593 ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
599 DPRINTF(("%s: unsupported protocol family %u\n", __func__,
600 saidx->dst.sa.sa_family));
601 IPIPSTAT_INC(ipips_family);
602 error = EAFNOSUPPORT; /* XXX diffs from openbsd */
606 IPIPSTAT_INC(ipips_opackets);
610 if (saidx->dst.sa.sa_family == AF_INET) {
612 if (sav->tdb_xform->xf_type == XF_IP4)
613 tdb->tdb_cur_bytes +=
614 m->m_pkthdr.len - sizeof(struct ip);
616 IPIPSTAT_ADD(ipips_obytes,
617 m->m_pkthdr.len - sizeof(struct ip));
622 if (saidx->dst.sa.sa_family == AF_INET6) {
624 if (sav->tdb_xform->xf_type == XF_IP4)
625 tdb->tdb_cur_bytes +=
626 m->m_pkthdr.len - sizeof(struct ip6_hdr);
628 IPIPSTAT_ADD(ipips_obytes,
629 m->m_pkthdr.len - sizeof(struct ip6_hdr));
642 #if defined(INET) || defined(INET6)
644 ipe4_init(struct secasvar *sav, struct xformsw *xsp)
646 sav->tdb_xform = xsp;
651 ipe4_zeroize(struct secasvar *sav)
653 sav->tdb_xform = NULL;
658 ipe4_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff)
660 /* This is a rather serious mistake, so no conditional printing. */
661 printf("%s: should never be called\n", __func__);
667 static struct xformsw ipe4_xformsw = {
668 XF_IP4, 0, "IPv4 Simple Encapsulation",
669 ipe4_init, ipe4_zeroize, ipe4_input, ipip_output,
672 extern struct domain inetdomain;
673 #endif /* INET || INET6 */
675 static struct protosw ipe4_protosw = {
677 .pr_domain = &inetdomain,
678 .pr_protocol = IPPROTO_IPV4,
679 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
680 .pr_input = ip4_input,
681 .pr_ctloutput = rip_ctloutput,
682 .pr_usrreqs = &rip_usrreqs
685 #if defined(INET6) && defined(INET)
686 static struct ip6protosw ipe6_protosw = {
688 .pr_domain = &inetdomain,
689 .pr_protocol = IPPROTO_IPV6,
690 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
691 .pr_input = ip4_input6,
692 .pr_ctloutput = rip_ctloutput,
693 .pr_usrreqs = &rip_usrreqs
695 #endif /* INET6 && INET */
699 * Check the encapsulated packet to see if we want it
702 ipe4_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
705 * Only take packets coming from IPSEC tunnels; the rest
706 * must be handled by the gif tunnel code. Note that we
707 * also return a minimum priority when we want the packet
708 * so any explicit gif tunnels take precedence.
710 return ((m->m_flags & M_IPSEC) != 0 ? 1 : 0);
718 xform_register(&ipe4_xformsw);
719 /* attach to encapsulation framework */
720 /* XXX save return cookie for detach on module remove */
722 (void) encap_attach_func(AF_INET, -1,
723 ipe4_encapcheck, &ipe4_protosw, NULL);
725 #if defined(INET6) && defined(INET)
726 (void) encap_attach_func(AF_INET6, -1,
727 ipe4_encapcheck, (struct protosw *)&ipe6_protosw, NULL);
730 SYSINIT(ipe4_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipe4_attach, NULL);