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>
56 #include <net/route.h>
57 #include <net/netisr.h>
60 #include <netinet/in.h>
61 #include <netinet/in_systm.h>
62 #include <netinet/in_var.h>
63 #include <netinet/ip.h>
64 #include <netinet/ip_ecn.h>
65 #include <netinet/ip_var.h>
66 #include <netinet/ip_encap.h>
68 #include <netipsec/ipsec.h>
69 #include <netipsec/xform.h>
71 #include <netipsec/ipip_var.h>
74 #include <netinet/ip6.h>
75 #include <netipsec/ipsec6.h>
76 #include <netinet6/ip6_ecn.h>
77 #include <netinet6/in6_var.h>
78 #include <netinet6/ip6protosw.h>
81 #include <netipsec/key.h>
82 #include <netipsec/key_debug.h>
84 #include <machine/stdarg.h>
87 * We can control the acceptance of IP4 packets by altering the sysctl
88 * net.inet.ipip.allow value. Zero means drop them, all else is acceptance.
90 VNET_DEFINE(int, ipip_allow) = 0;
91 VNET_DEFINE(struct ipipstat, ipipstat);
93 SYSCTL_DECL(_net_inet_ipip);
94 SYSCTL_VNET_INT(_net_inet_ipip, OID_AUTO,
95 ipip_allow, CTLFLAG_RW, &VNET_NAME(ipip_allow), 0, "");
96 SYSCTL_VNET_STRUCT(_net_inet_ipip, IPSECCTL_STATS,
97 stats, CTLFLAG_RD, &VNET_NAME(ipipstat), ipipstat, "");
100 #define M_IPSEC (M_AUTHIPHDR|M_AUTHIPDGM|M_DECRYPTED)
102 static void _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp);
106 * Really only a wrapper for ipip_input(), for use with IPv6.
109 ip4_input6(struct mbuf **m, int *offp, int proto)
112 /* If we do not accept IP-in-IP explicitly, drop. */
113 if (!V_ipip_allow && ((*m)->m_flags & M_IPSEC) == 0) {
114 DPRINTF(("%s: dropped due to policy\n", __func__));
115 IPIPSTAT_INC(ipips_pdrops);
120 _ipip_input(*m, *offp, NULL);
127 * Really only a wrapper for ipip_input(), for use with IPv4.
130 ip4_input(struct mbuf *m, int off)
133 /* If we do not accept IP-in-IP explicitly, drop. */
134 if (!V_ipip_allow && (m->m_flags & M_IPSEC) == 0) {
135 DPRINTF(("%s: dropped due to policy\n", __func__));
136 IPIPSTAT_INC(ipips_pdrops);
141 _ipip_input(m, off, NULL);
146 * ipip_input gets called when we receive an IP{46} encapsulated packet,
147 * either because we got it at a real interface, or because AH or ESP
148 * were being used in tunnel mode (in which case the rcvif element will
149 * contain the address of the encX interface associated with the tunnel.
153 _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp)
157 struct ip6_hdr *ip6 = NULL;
165 IPIPSTAT_INC(ipips_ipackets);
167 m_copydata(m, 0, 1, &v);
172 hlen = sizeof(struct ip);
177 hlen = sizeof(struct ip6_hdr);
181 IPIPSTAT_INC(ipips_family);
183 return /* EAFNOSUPPORT */;
186 /* Bring the IP header in the first mbuf, if not there already */
187 if (m->m_len < hlen) {
188 if ((m = m_pullup(m, hlen)) == NULL) {
189 DPRINTF(("%s: m_pullup (1) failed\n", __func__));
190 IPIPSTAT_INC(ipips_hdrops);
194 ipo = mtod(m, struct ip *);
196 /* Keep outer ecn field. */
205 otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
209 panic("ipip_input: unknown ip version %u (outer)", v>>4);
212 /* Remove outer IP header */
216 if (m->m_pkthdr.len < sizeof(struct ip)) {
217 IPIPSTAT_INC(ipips_hdrops);
222 m_copydata(m, 0, 1, &v);
227 hlen = sizeof(struct ip);
233 hlen = sizeof(struct ip6_hdr);
237 IPIPSTAT_INC(ipips_family);
239 return; /* EAFNOSUPPORT */
243 * Bring the inner IP header in the first mbuf, if not there already.
245 if (m->m_len < hlen) {
246 if ((m = m_pullup(m, hlen)) == NULL) {
247 DPRINTF(("%s: m_pullup (2) failed\n", __func__));
248 IPIPSTAT_INC(ipips_hdrops);
254 * RFC 1853 specifies that the inner TTL should not be touched on
255 * decapsulation. There's no reason this comment should be here, but
256 * this is as good as any a position.
259 /* Some sanity checks in the inner IP header */
263 ipo = mtod(m, struct ip *);
264 ip_ecn_egress(V_ip4_ipsec_ecn, &otos, &ipo->ip_tos);
269 ip6 = (struct ip6_hdr *) ipo;
270 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
271 ip_ecn_egress(V_ip6_ipsec_ecn, &otos, &itos);
272 ip6->ip6_flow &= ~htonl(0xff << 20);
273 ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
277 panic("ipip_input: unknown ip version %u (inner)", v>>4);
280 /* Check for local address spoofing. */
281 if ((m->m_pkthdr.rcvif == NULL ||
282 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
285 if ((v >> 4) == IPVERSION &&
286 in_localip(ipo->ip_src) != 0) {
287 IPIPSTAT_INC(ipips_spoof);
293 if ((v & IPV6_VERSION_MASK) == IPV6_VERSION &&
294 in6_localip(&ip6->ip6_src) != 0) {
295 IPIPSTAT_INC(ipips_spoof);
303 IPIPSTAT_ADD(ipips_ibytes, m->m_pkthdr.len - iphlen);
309 ipsec_bpf(m, NULL, AF_INET, ENC_IN|ENC_AFTER);
314 ipsec_bpf(m, NULL, AF_INET6, ENC_IN|ENC_AFTER);
318 panic("%s: bogus ip version %u", __func__, v>>4);
320 /* pass the mbuf to enc0 for packet filtering */
321 if (ipsec_filter(&m, PFIL_IN, ENC_IN|ENC_AFTER) != 0)
326 * Interface pointer stays the same; if no IPsec processing has
327 * been done (or will be done), this will point to a normal
328 * interface. Otherwise, it'll point to an enc interface, which
329 * will allow a packet filter to distinguish between secure and
345 panic("%s: bogus ip version %u", __func__, v>>4);
350 if (netisr_queue(isr, m)) { /* (0) on success. */
351 IPIPSTAT_INC(ipips_qfull);
352 DPRINTF(("%s: packet dropped because of full queue\n",
360 struct ipsecrequest *isr,
366 struct secasvar *sav;
368 struct secasindex *saidx;
370 #if defined(INET) || defined(INET6)
377 struct ip6_hdr *ip6, *ip6o;
381 IPSEC_ASSERT(sav != NULL, ("null SA"));
382 IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
384 /* XXX Deal with empty TDB source/destination addresses. */
386 m_copydata(m, 0, 1, &tp);
387 tp = (tp >> 4) & 0xff; /* Get the IP version number. */
389 saidx = &sav->sah->saidx;
390 switch (saidx->dst.sa.sa_family) {
393 if (saidx->src.sa.sa_family != AF_INET ||
394 saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
395 saidx->dst.sin.sin_addr.s_addr == INADDR_ANY) {
396 DPRINTF(("%s: unspecified tunnel endpoint "
397 "address in SA %s/%08lx\n", __func__,
398 ipsec_address(&saidx->dst),
399 (u_long) ntohl(sav->spi)));
400 IPIPSTAT_INC(ipips_unspec);
405 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
407 DPRINTF(("%s: M_PREPEND failed\n", __func__));
408 IPIPSTAT_INC(ipips_hdrops);
413 ipo = mtod(m, struct ip *);
415 ipo->ip_v = IPVERSION;
417 ipo->ip_len = htons(m->m_pkthdr.len);
418 ipo->ip_ttl = V_ip_defttl;
420 ipo->ip_src = saidx->src.sin.sin_addr;
421 ipo->ip_dst = saidx->dst.sin.sin_addr;
423 ipo->ip_id = ip_newid();
425 /* If the inner protocol is IP... */
428 /* Save ECN notification */
429 m_copydata(m, sizeof(struct ip) +
430 offsetof(struct ip, ip_tos),
431 sizeof(u_int8_t), (caddr_t) &itos);
433 ipo->ip_p = IPPROTO_IPIP;
436 * We should be keeping tunnel soft-state and
437 * send back ICMPs if needed.
439 m_copydata(m, sizeof(struct ip) +
440 offsetof(struct ip, ip_off),
441 sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
442 ipo->ip_off = ntohs(ipo->ip_off);
443 ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
444 ipo->ip_off = htons(ipo->ip_off);
447 case (IPV6_VERSION >> 4):
451 /* Save ECN notification. */
452 m_copydata(m, sizeof(struct ip) +
453 offsetof(struct ip6_hdr, ip6_flow),
454 sizeof(u_int32_t), (caddr_t) &itos32);
455 itos = ntohl(itos32) >> 20;
456 ipo->ip_p = IPPROTO_IPV6;
466 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
473 if (IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr) ||
474 saidx->src.sa.sa_family != AF_INET6 ||
475 IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr)) {
476 DPRINTF(("%s: unspecified tunnel endpoint "
477 "address in SA %s/%08lx\n", __func__,
478 ipsec_address(&saidx->dst),
479 (u_long) ntohl(sav->spi)));
480 IPIPSTAT_INC(ipips_unspec);
485 /* scoped address handling */
486 ip6 = mtod(m, struct ip6_hdr *);
487 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
488 ip6->ip6_src.s6_addr16[1] = 0;
489 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
490 ip6->ip6_dst.s6_addr16[1] = 0;
492 M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
494 DPRINTF(("%s: M_PREPEND failed\n", __func__));
495 IPIPSTAT_INC(ipips_hdrops);
500 /* Initialize IPv6 header */
501 ip6o = mtod(m, struct ip6_hdr *);
503 ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
504 ip6o->ip6_vfc |= IPV6_VERSION;
505 ip6o->ip6_plen = htons(m->m_pkthdr.len);
506 ip6o->ip6_hlim = V_ip_defttl;
507 ip6o->ip6_dst = saidx->dst.sin6.sin6_addr;
508 ip6o->ip6_src = saidx->src.sin6.sin6_addr;
513 /* Save ECN notification */
514 m_copydata(m, sizeof(struct ip6_hdr) +
515 offsetof(struct ip, ip_tos), sizeof(u_int8_t),
518 /* This is really IPVERSION. */
519 ip6o->ip6_nxt = IPPROTO_IPIP;
522 case (IPV6_VERSION >> 4):
526 /* Save ECN notification. */
527 m_copydata(m, sizeof(struct ip6_hdr) +
528 offsetof(struct ip6_hdr, ip6_flow),
529 sizeof(u_int32_t), (caddr_t) &itos32);
530 itos = ntohl(itos32) >> 20;
532 ip6o->ip6_nxt = IPPROTO_IPV6;
540 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
541 ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
547 DPRINTF(("%s: unsupported protocol family %u\n", __func__,
548 saidx->dst.sa.sa_family));
549 IPIPSTAT_INC(ipips_family);
550 error = EAFNOSUPPORT; /* XXX diffs from openbsd */
554 IPIPSTAT_INC(ipips_opackets);
558 if (saidx->dst.sa.sa_family == AF_INET) {
560 if (sav->tdb_xform->xf_type == XF_IP4)
561 tdb->tdb_cur_bytes +=
562 m->m_pkthdr.len - sizeof(struct ip);
564 IPIPSTAT_ADD(ipips_obytes,
565 m->m_pkthdr.len - sizeof(struct ip));
570 if (saidx->dst.sa.sa_family == AF_INET6) {
572 if (sav->tdb_xform->xf_type == XF_IP4)
573 tdb->tdb_cur_bytes +=
574 m->m_pkthdr.len - sizeof(struct ip6_hdr);
576 IPIPSTAT_ADD(ipips_obytes,
577 m->m_pkthdr.len - sizeof(struct ip6_hdr));
590 #if defined(INET) || defined(INET6)
592 ipe4_init(struct secasvar *sav, struct xformsw *xsp)
594 sav->tdb_xform = xsp;
599 ipe4_zeroize(struct secasvar *sav)
601 sav->tdb_xform = NULL;
606 ipe4_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff)
608 /* This is a rather serious mistake, so no conditional printing. */
609 printf("%s: should never be called\n", __func__);
615 static struct xformsw ipe4_xformsw = {
616 XF_IP4, 0, "IPv4 Simple Encapsulation",
617 ipe4_init, ipe4_zeroize, ipe4_input, ipip_output,
620 extern struct domain inetdomain;
621 #endif /* INET || INET6 */
623 static struct protosw ipe4_protosw = {
625 .pr_domain = &inetdomain,
626 .pr_protocol = IPPROTO_IPV4,
627 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
628 .pr_input = ip4_input,
629 .pr_ctloutput = rip_ctloutput,
630 .pr_usrreqs = &rip_usrreqs
633 #if defined(INET6) && defined(INET)
634 static struct ip6protosw ipe6_protosw = {
636 .pr_domain = &inetdomain,
637 .pr_protocol = IPPROTO_IPV6,
638 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
639 .pr_input = ip4_input6,
640 .pr_ctloutput = rip_ctloutput,
641 .pr_usrreqs = &rip_usrreqs
643 #endif /* INET6 && INET */
647 * Check the encapsulated packet to see if we want it
650 ipe4_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
653 * Only take packets coming from IPSEC tunnels; the rest
654 * must be handled by the gif tunnel code. Note that we
655 * also return a minimum priority when we want the packet
656 * so any explicit gif tunnels take precedence.
658 return ((m->m_flags & M_IPSEC) != 0 ? 1 : 0);
666 xform_register(&ipe4_xformsw);
667 /* attach to encapsulation framework */
668 /* XXX save return cookie for detach on module remove */
670 (void) encap_attach_func(AF_INET, -1,
671 ipe4_encapcheck, &ipe4_protosw, NULL);
673 #if defined(INET6) && defined(INET)
674 (void) encap_attach_func(AF_INET6, -1,
675 ipe4_encapcheck, (struct protosw *)&ipe6_protosw, NULL);
678 SYSINIT(ipe4_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipe4_attach, NULL);