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/netisr.h>
59 #include <netinet/in.h>
60 #include <netinet/in_systm.h>
61 #include <netinet/in_var.h>
62 #include <netinet/ip.h>
63 #include <netinet/ip_ecn.h>
64 #include <netinet/ip_var.h>
65 #include <netinet/ip_encap.h>
67 #include <netipsec/ipsec.h>
68 #include <netipsec/xform.h>
70 #include <netipsec/ipip_var.h>
73 #include <netinet/ip6.h>
74 #include <netipsec/ipsec6.h>
75 #include <netinet6/ip6_ecn.h>
76 #include <netinet6/in6_var.h>
77 #include <netinet6/ip6protosw.h>
80 #include <netipsec/key.h>
81 #include <netipsec/key_debug.h>
83 #include <machine/stdarg.h>
86 * We can control the acceptance of IP4 packets by altering the sysctl
87 * net.inet.ipip.allow value. Zero means drop them, all else is acceptance.
89 VNET_DEFINE(int, ipip_allow) = 0;
90 VNET_PCPUSTAT_DEFINE(struct ipipstat, ipipstat);
91 VNET_PCPUSTAT_SYSINIT(ipipstat);
94 VNET_PCPUSTAT_SYSUNINIT(ipipstat);
97 SYSCTL_DECL(_net_inet_ipip);
98 SYSCTL_VNET_INT(_net_inet_ipip, OID_AUTO,
99 ipip_allow, CTLFLAG_RW, &VNET_NAME(ipip_allow), 0, "");
100 SYSCTL_VNET_PCPUSTAT(_net_inet_ipip, IPSECCTL_STATS, stats,
101 struct ipipstat, ipipstat,
102 "IPIP statistics (struct ipipstat, netipsec/ipip_var.h)");
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 (!V_ipip_allow && ((*m)->m_flags & M_IPSEC) == 0) {
119 DPRINTF(("%s: dropped due to policy\n", __func__));
120 IPIPSTAT_INC(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, int off)
138 /* If we do not accept IP-in-IP explicitly, drop. */
139 if (!V_ipip_allow && (m->m_flags & M_IPSEC) == 0) {
140 DPRINTF(("%s: dropped due to policy\n", __func__));
141 IPIPSTAT_INC(ipips_pdrops);
146 _ipip_input(m, off, NULL);
151 * ipip_input gets called when we receive an IP{46} encapsulated packet,
152 * either because we got it at a real interface, or because AH or ESP
153 * were being used in tunnel mode (in which case the rcvif element will
154 * contain the address of the encX interface associated with the tunnel.
158 _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp)
162 struct ip6_hdr *ip6 = NULL;
170 IPIPSTAT_INC(ipips_ipackets);
172 m_copydata(m, 0, 1, &v);
177 hlen = sizeof(struct ip);
182 hlen = sizeof(struct ip6_hdr);
186 IPIPSTAT_INC(ipips_family);
188 return /* EAFNOSUPPORT */;
191 /* Bring the IP header in the first mbuf, if not there already */
192 if (m->m_len < hlen) {
193 if ((m = m_pullup(m, hlen)) == NULL) {
194 DPRINTF(("%s: m_pullup (1) failed\n", __func__));
195 IPIPSTAT_INC(ipips_hdrops);
199 ipo = mtod(m, struct ip *);
201 /* Keep outer ecn field. */
210 otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
214 panic("ipip_input: unknown ip version %u (outer)", v>>4);
217 /* Remove outer IP header */
221 if (m->m_pkthdr.len < sizeof(struct ip)) {
222 IPIPSTAT_INC(ipips_hdrops);
227 m_copydata(m, 0, 1, &v);
232 hlen = sizeof(struct ip);
238 hlen = sizeof(struct ip6_hdr);
242 IPIPSTAT_INC(ipips_family);
244 return; /* EAFNOSUPPORT */
248 * Bring the inner IP header in the first mbuf, if not there already.
250 if (m->m_len < hlen) {
251 if ((m = m_pullup(m, hlen)) == NULL) {
252 DPRINTF(("%s: m_pullup (2) failed\n", __func__));
253 IPIPSTAT_INC(ipips_hdrops);
259 * RFC 1853 specifies that the inner TTL should not be touched on
260 * decapsulation. There's no reason this comment should be here, but
261 * this is as good as any a position.
264 /* Some sanity checks in the inner IP header */
268 ipo = mtod(m, struct ip *);
269 ip_ecn_egress(V_ip4_ipsec_ecn, &otos, &ipo->ip_tos);
274 ip6 = (struct ip6_hdr *) ipo;
275 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
276 ip_ecn_egress(V_ip6_ipsec_ecn, &otos, &itos);
277 ip6->ip6_flow &= ~htonl(0xff << 20);
278 ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
282 panic("ipip_input: unknown ip version %u (inner)", v>>4);
285 /* Check for local address spoofing. */
286 if ((m->m_pkthdr.rcvif == NULL ||
287 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
290 if ((v >> 4) == IPVERSION &&
291 in_localip(ipo->ip_src) != 0) {
292 IPIPSTAT_INC(ipips_spoof);
298 if ((v & IPV6_VERSION_MASK) == IPV6_VERSION &&
299 in6_localip(&ip6->ip6_src) != 0) {
300 IPIPSTAT_INC(ipips_spoof);
308 IPIPSTAT_ADD(ipips_ibytes, m->m_pkthdr.len - iphlen);
311 * Interface pointer stays the same; if no IPsec processing has
312 * been done (or will be done), this will point to a normal
313 * interface. Otherwise, it'll point to an enc interface, which
314 * will allow a packet filter to distinguish between secure and
330 panic("%s: bogus ip version %u", __func__, v>>4);
333 if (netisr_queue(isr, m)) { /* (0) on success. */
334 IPIPSTAT_INC(ipips_qfull);
335 DPRINTF(("%s: packet dropped because of full queue\n",
343 struct ipsecrequest *isr,
349 struct secasvar *sav;
351 struct secasindex *saidx;
353 #if defined(INET) || defined(INET6)
360 struct ip6_hdr *ip6, *ip6o;
364 IPSEC_ASSERT(sav != NULL, ("null SA"));
365 IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
367 /* XXX Deal with empty TDB source/destination addresses. */
369 m_copydata(m, 0, 1, &tp);
370 tp = (tp >> 4) & 0xff; /* Get the IP version number. */
372 saidx = &sav->sah->saidx;
373 switch (saidx->dst.sa.sa_family) {
376 if (saidx->src.sa.sa_family != AF_INET ||
377 saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
378 saidx->dst.sin.sin_addr.s_addr == INADDR_ANY) {
379 DPRINTF(("%s: unspecified tunnel endpoint "
380 "address in SA %s/%08lx\n", __func__,
381 ipsec_address(&saidx->dst),
382 (u_long) ntohl(sav->spi)));
383 IPIPSTAT_INC(ipips_unspec);
388 M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
390 DPRINTF(("%s: M_PREPEND failed\n", __func__));
391 IPIPSTAT_INC(ipips_hdrops);
396 ipo = mtod(m, struct ip *);
398 ipo->ip_v = IPVERSION;
400 ipo->ip_len = htons(m->m_pkthdr.len);
401 ipo->ip_ttl = V_ip_defttl;
403 ipo->ip_src = saidx->src.sin.sin_addr;
404 ipo->ip_dst = saidx->dst.sin.sin_addr;
406 ipo->ip_id = ip_newid();
408 /* If the inner protocol is IP... */
411 /* Save ECN notification */
412 m_copydata(m, sizeof(struct ip) +
413 offsetof(struct ip, ip_tos),
414 sizeof(u_int8_t), (caddr_t) &itos);
416 ipo->ip_p = IPPROTO_IPIP;
419 * We should be keeping tunnel soft-state and
420 * send back ICMPs if needed.
422 m_copydata(m, sizeof(struct ip) +
423 offsetof(struct ip, ip_off),
424 sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
425 ipo->ip_off = ntohs(ipo->ip_off);
426 ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
427 ipo->ip_off = htons(ipo->ip_off);
430 case (IPV6_VERSION >> 4):
434 /* Save ECN notification. */
435 m_copydata(m, sizeof(struct ip) +
436 offsetof(struct ip6_hdr, ip6_flow),
437 sizeof(u_int32_t), (caddr_t) &itos32);
438 itos = ntohl(itos32) >> 20;
439 ipo->ip_p = IPPROTO_IPV6;
449 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
456 if (IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr) ||
457 saidx->src.sa.sa_family != AF_INET6 ||
458 IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr)) {
459 DPRINTF(("%s: unspecified tunnel endpoint "
460 "address in SA %s/%08lx\n", __func__,
461 ipsec_address(&saidx->dst),
462 (u_long) ntohl(sav->spi)));
463 IPIPSTAT_INC(ipips_unspec);
468 /* scoped address handling */
469 ip6 = mtod(m, struct ip6_hdr *);
470 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
471 ip6->ip6_src.s6_addr16[1] = 0;
472 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
473 ip6->ip6_dst.s6_addr16[1] = 0;
475 M_PREPEND(m, sizeof(struct ip6_hdr), M_NOWAIT);
477 DPRINTF(("%s: M_PREPEND failed\n", __func__));
478 IPIPSTAT_INC(ipips_hdrops);
483 /* Initialize IPv6 header */
484 ip6o = mtod(m, struct ip6_hdr *);
486 ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
487 ip6o->ip6_vfc |= IPV6_VERSION;
488 ip6o->ip6_hlim = IPV6_DEFHLIM;
489 ip6o->ip6_dst = saidx->dst.sin6.sin6_addr;
490 ip6o->ip6_src = saidx->src.sin6.sin6_addr;
491 ip6o->ip6_plen = htons(m->m_pkthdr.len - sizeof(*ip6));
496 /* Save ECN notification */
497 m_copydata(m, sizeof(struct ip6_hdr) +
498 offsetof(struct ip, ip_tos), sizeof(u_int8_t),
501 /* This is really IPVERSION. */
502 ip6o->ip6_nxt = IPPROTO_IPIP;
505 case (IPV6_VERSION >> 4):
509 /* Save ECN notification. */
510 m_copydata(m, sizeof(struct ip6_hdr) +
511 offsetof(struct ip6_hdr, ip6_flow),
512 sizeof(u_int32_t), (caddr_t) &itos32);
513 itos = ntohl(itos32) >> 20;
515 ip6o->ip6_nxt = IPPROTO_IPV6;
523 ip_ecn_ingress(V_ip6_ipsec_ecn, &otos, &itos);
524 ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
530 DPRINTF(("%s: unsupported protocol family %u\n", __func__,
531 saidx->dst.sa.sa_family));
532 IPIPSTAT_INC(ipips_family);
533 error = EAFNOSUPPORT; /* XXX diffs from openbsd */
537 IPIPSTAT_INC(ipips_opackets);
541 if (saidx->dst.sa.sa_family == AF_INET) {
543 if (sav->tdb_xform->xf_type == XF_IP4)
544 tdb->tdb_cur_bytes +=
545 m->m_pkthdr.len - sizeof(struct ip);
547 IPIPSTAT_ADD(ipips_obytes,
548 m->m_pkthdr.len - sizeof(struct ip));
553 if (saidx->dst.sa.sa_family == AF_INET6) {
555 if (sav->tdb_xform->xf_type == XF_IP4)
556 tdb->tdb_cur_bytes +=
557 m->m_pkthdr.len - sizeof(struct ip6_hdr);
559 IPIPSTAT_ADD(ipips_obytes,
560 m->m_pkthdr.len - sizeof(struct ip6_hdr));
573 #if defined(INET) || defined(INET6)
575 ipe4_init(struct secasvar *sav, struct xformsw *xsp)
577 sav->tdb_xform = xsp;
582 ipe4_zeroize(struct secasvar *sav)
584 sav->tdb_xform = NULL;
589 ipe4_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff)
591 /* This is a rather serious mistake, so no conditional printing. */
592 printf("%s: should never be called\n", __func__);
598 static struct xformsw ipe4_xformsw = {
599 XF_IP4, 0, "IPv4 Simple Encapsulation",
600 ipe4_init, ipe4_zeroize, ipe4_input, ipip_output,
603 extern struct domain inetdomain;
604 #endif /* INET || INET6 */
606 static struct protosw ipe4_protosw = {
608 .pr_domain = &inetdomain,
609 .pr_protocol = IPPROTO_IPV4,
610 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
611 .pr_input = ip4_input,
612 .pr_ctloutput = rip_ctloutput,
613 .pr_usrreqs = &rip_usrreqs
616 #if defined(INET6) && defined(INET)
617 static struct ip6protosw ipe6_protosw = {
619 .pr_domain = &inetdomain,
620 .pr_protocol = IPPROTO_IPV6,
621 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
622 .pr_input = ip4_input6,
623 .pr_ctloutput = rip_ctloutput,
624 .pr_usrreqs = &rip_usrreqs
626 #endif /* INET6 && INET */
630 * Check the encapsulated packet to see if we want it
633 ipe4_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
636 * Only take packets coming from IPSEC tunnels; the rest
637 * must be handled by the gif tunnel code. Note that we
638 * also return a minimum priority when we want the packet
639 * so any explicit gif tunnels take precedence.
641 return ((m->m_flags & M_IPSEC) != 0 ? 1 : 0);
649 xform_register(&ipe4_xformsw);
650 /* attach to encapsulation framework */
651 /* XXX save return cookie for detach on module remove */
653 (void) encap_attach_func(AF_INET, -1,
654 ipe4_encapcheck, &ipe4_protosw, NULL);
656 #if defined(INET6) && defined(INET)
657 (void) encap_attach_func(AF_INET6, -1,
658 ipe4_encapcheck, (struct protosw *)&ipe6_protosw, NULL);
661 SYSINIT(ipe4_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipe4_attach, NULL);