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_PCPUSTAT_DEFINE(struct ipipstat, ipipstat);
92 VNET_PCPUSTAT_SYSINIT(ipipstat);
95 VNET_PCPUSTAT_SYSUNINIT(ipipstat);
98 SYSCTL_DECL(_net_inet_ipip);
99 SYSCTL_VNET_INT(_net_inet_ipip, OID_AUTO,
100 ipip_allow, CTLFLAG_RW, &VNET_NAME(ipip_allow), 0, "");
101 SYSCTL_VNET_PCPUSTAT(_net_inet_ipip, IPSECCTL_STATS, stats,
102 struct ipipstat, ipipstat,
103 "IPIP statistics (struct ipipstat, netipsec/ipip_var.h)");
106 #define M_IPSEC (M_AUTHIPHDR|M_AUTHIPDGM|M_DECRYPTED)
108 static void _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp);
112 * Really only a wrapper for ipip_input(), for use with IPv6.
115 ip4_input6(struct mbuf **m, int *offp, int proto)
118 /* If we do not accept IP-in-IP explicitly, drop. */
119 if (!V_ipip_allow && ((*m)->m_flags & M_IPSEC) == 0) {
120 DPRINTF(("%s: dropped due to policy\n", __func__));
121 IPIPSTAT_INC(ipips_pdrops);
126 _ipip_input(*m, *offp, NULL);
133 * Really only a wrapper for ipip_input(), for use with IPv4.
136 ip4_input(struct mbuf *m, int off)
139 /* If we do not accept IP-in-IP explicitly, drop. */
140 if (!V_ipip_allow && (m->m_flags & M_IPSEC) == 0) {
141 DPRINTF(("%s: dropped due to policy\n", __func__));
142 IPIPSTAT_INC(ipips_pdrops);
147 _ipip_input(m, off, NULL);
152 * ipip_input gets called when we receive an IP{46} encapsulated packet,
153 * either because we got it at a real interface, or because AH or ESP
154 * were being used in tunnel mode (in which case the rcvif element will
155 * contain the address of the encX interface associated with the tunnel.
159 _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp)
163 struct ip6_hdr *ip6 = NULL;
171 IPIPSTAT_INC(ipips_ipackets);
173 m_copydata(m, 0, 1, &v);
178 hlen = sizeof(struct ip);
183 hlen = sizeof(struct ip6_hdr);
187 IPIPSTAT_INC(ipips_family);
189 return /* EAFNOSUPPORT */;
192 /* Bring the IP header in the first mbuf, if not there already */
193 if (m->m_len < hlen) {
194 if ((m = m_pullup(m, hlen)) == NULL) {
195 DPRINTF(("%s: m_pullup (1) failed\n", __func__));
196 IPIPSTAT_INC(ipips_hdrops);
200 ipo = mtod(m, struct ip *);
202 /* Keep outer ecn field. */
211 otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
215 panic("ipip_input: unknown ip version %u (outer)", v>>4);
218 /* Remove outer IP header */
222 if (m->m_pkthdr.len < sizeof(struct ip)) {
223 IPIPSTAT_INC(ipips_hdrops);
228 m_copydata(m, 0, 1, &v);
233 hlen = sizeof(struct ip);
239 hlen = sizeof(struct ip6_hdr);
243 IPIPSTAT_INC(ipips_family);
245 return; /* EAFNOSUPPORT */
249 * Bring the inner IP header in the first mbuf, if not there already.
251 if (m->m_len < hlen) {
252 if ((m = m_pullup(m, hlen)) == NULL) {
253 DPRINTF(("%s: m_pullup (2) failed\n", __func__));
254 IPIPSTAT_INC(ipips_hdrops);
260 * RFC 1853 specifies that the inner TTL should not be touched on
261 * decapsulation. There's no reason this comment should be here, but
262 * this is as good as any a position.
265 /* Some sanity checks in the inner IP header */
269 ipo = mtod(m, struct ip *);
270 ip_ecn_egress(V_ip4_ipsec_ecn, &otos, &ipo->ip_tos);
275 ip6 = (struct ip6_hdr *) ipo;
276 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
277 ip_ecn_egress(V_ip6_ipsec_ecn, &otos, &itos);
278 ip6->ip6_flow &= ~htonl(0xff << 20);
279 ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
283 panic("ipip_input: unknown ip version %u (inner)", v>>4);
286 /* Check for local address spoofing. */
287 if ((m->m_pkthdr.rcvif == NULL ||
288 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
291 if ((v >> 4) == IPVERSION &&
292 in_localip(ipo->ip_src) != 0) {
293 IPIPSTAT_INC(ipips_spoof);
299 if ((v & IPV6_VERSION_MASK) == IPV6_VERSION &&
300 in6_localip(&ip6->ip6_src) != 0) {
301 IPIPSTAT_INC(ipips_spoof);
309 IPIPSTAT_ADD(ipips_ibytes, m->m_pkthdr.len - iphlen);
315 ipsec_bpf(m, NULL, AF_INET, ENC_IN|ENC_AFTER);
320 ipsec_bpf(m, NULL, AF_INET6, ENC_IN|ENC_AFTER);
324 panic("%s: bogus ip version %u", __func__, v>>4);
326 /* pass the mbuf to enc0 for packet filtering */
327 if (ipsec_filter(&m, PFIL_IN, ENC_IN|ENC_AFTER) != 0)
332 * Interface pointer stays the same; if no IPsec processing has
333 * been done (or will be done), this will point to a normal
334 * interface. Otherwise, it'll point to an enc interface, which
335 * will allow a packet filter to distinguish between secure and
351 panic("%s: bogus ip version %u", __func__, v>>4);
354 if (netisr_queue(isr, m)) { /* (0) on success. */
355 IPIPSTAT_INC(ipips_qfull);
356 DPRINTF(("%s: packet dropped because of full queue\n",
364 struct ipsecrequest *isr,
370 struct secasvar *sav;
372 struct secasindex *saidx;
374 #if defined(INET) || defined(INET6)
381 struct ip6_hdr *ip6, *ip6o;
385 IPSEC_ASSERT(sav != NULL, ("null SA"));
386 IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
388 /* XXX Deal with empty TDB source/destination addresses. */
390 m_copydata(m, 0, 1, &tp);
391 tp = (tp >> 4) & 0xff; /* Get the IP version number. */
393 saidx = &sav->sah->saidx;
394 switch (saidx->dst.sa.sa_family) {
397 if (saidx->src.sa.sa_family != AF_INET ||
398 saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
399 saidx->dst.sin.sin_addr.s_addr == INADDR_ANY) {
400 DPRINTF(("%s: unspecified tunnel endpoint "
401 "address in SA %s/%08lx\n", __func__,
402 ipsec_address(&saidx->dst),
403 (u_long) ntohl(sav->spi)));
404 IPIPSTAT_INC(ipips_unspec);
409 M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
411 DPRINTF(("%s: M_PREPEND failed\n", __func__));
412 IPIPSTAT_INC(ipips_hdrops);
417 ipo = mtod(m, struct ip *);
419 ipo->ip_v = IPVERSION;
421 ipo->ip_len = htons(m->m_pkthdr.len);
422 ipo->ip_ttl = V_ip_defttl;
424 ipo->ip_src = saidx->src.sin.sin_addr;
425 ipo->ip_dst = saidx->dst.sin.sin_addr;
427 ipo->ip_id = ip_newid();
429 /* If the inner protocol is IP... */
432 /* Save ECN notification */
433 m_copydata(m, sizeof(struct ip) +
434 offsetof(struct ip, ip_tos),
435 sizeof(u_int8_t), (caddr_t) &itos);
437 ipo->ip_p = IPPROTO_IPIP;
440 * We should be keeping tunnel soft-state and
441 * send back ICMPs if needed.
443 m_copydata(m, sizeof(struct ip) +
444 offsetof(struct ip, ip_off),
445 sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
446 ipo->ip_off = ntohs(ipo->ip_off);
447 ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
448 ipo->ip_off = htons(ipo->ip_off);
451 case (IPV6_VERSION >> 4):
455 /* Save ECN notification. */
456 m_copydata(m, sizeof(struct ip) +
457 offsetof(struct ip6_hdr, ip6_flow),
458 sizeof(u_int32_t), (caddr_t) &itos32);
459 itos = ntohl(itos32) >> 20;
460 ipo->ip_p = IPPROTO_IPV6;
470 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
477 if (IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr) ||
478 saidx->src.sa.sa_family != AF_INET6 ||
479 IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr)) {
480 DPRINTF(("%s: unspecified tunnel endpoint "
481 "address in SA %s/%08lx\n", __func__,
482 ipsec_address(&saidx->dst),
483 (u_long) ntohl(sav->spi)));
484 IPIPSTAT_INC(ipips_unspec);
489 /* scoped address handling */
490 ip6 = mtod(m, struct ip6_hdr *);
491 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
492 ip6->ip6_src.s6_addr16[1] = 0;
493 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
494 ip6->ip6_dst.s6_addr16[1] = 0;
496 M_PREPEND(m, sizeof(struct ip6_hdr), M_NOWAIT);
498 DPRINTF(("%s: M_PREPEND failed\n", __func__));
499 IPIPSTAT_INC(ipips_hdrops);
504 /* Initialize IPv6 header */
505 ip6o = mtod(m, struct ip6_hdr *);
507 ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
508 ip6o->ip6_vfc |= IPV6_VERSION;
509 ip6o->ip6_plen = htons(m->m_pkthdr.len);
510 ip6o->ip6_hlim = V_ip_defttl;
511 ip6o->ip6_dst = saidx->dst.sin6.sin6_addr;
512 ip6o->ip6_src = saidx->src.sin6.sin6_addr;
517 /* Save ECN notification */
518 m_copydata(m, sizeof(struct ip6_hdr) +
519 offsetof(struct ip, ip_tos), sizeof(u_int8_t),
522 /* This is really IPVERSION. */
523 ip6o->ip6_nxt = IPPROTO_IPIP;
526 case (IPV6_VERSION >> 4):
530 /* Save ECN notification. */
531 m_copydata(m, sizeof(struct ip6_hdr) +
532 offsetof(struct ip6_hdr, ip6_flow),
533 sizeof(u_int32_t), (caddr_t) &itos32);
534 itos = ntohl(itos32) >> 20;
536 ip6o->ip6_nxt = IPPROTO_IPV6;
544 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
545 ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
551 DPRINTF(("%s: unsupported protocol family %u\n", __func__,
552 saidx->dst.sa.sa_family));
553 IPIPSTAT_INC(ipips_family);
554 error = EAFNOSUPPORT; /* XXX diffs from openbsd */
558 IPIPSTAT_INC(ipips_opackets);
562 if (saidx->dst.sa.sa_family == AF_INET) {
564 if (sav->tdb_xform->xf_type == XF_IP4)
565 tdb->tdb_cur_bytes +=
566 m->m_pkthdr.len - sizeof(struct ip);
568 IPIPSTAT_ADD(ipips_obytes,
569 m->m_pkthdr.len - sizeof(struct ip));
574 if (saidx->dst.sa.sa_family == AF_INET6) {
576 if (sav->tdb_xform->xf_type == XF_IP4)
577 tdb->tdb_cur_bytes +=
578 m->m_pkthdr.len - sizeof(struct ip6_hdr);
580 IPIPSTAT_ADD(ipips_obytes,
581 m->m_pkthdr.len - sizeof(struct ip6_hdr));
594 #if defined(INET) || defined(INET6)
596 ipe4_init(struct secasvar *sav, struct xformsw *xsp)
598 sav->tdb_xform = xsp;
603 ipe4_zeroize(struct secasvar *sav)
605 sav->tdb_xform = NULL;
610 ipe4_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff)
612 /* This is a rather serious mistake, so no conditional printing. */
613 printf("%s: should never be called\n", __func__);
619 static struct xformsw ipe4_xformsw = {
620 XF_IP4, 0, "IPv4 Simple Encapsulation",
621 ipe4_init, ipe4_zeroize, ipe4_input, ipip_output,
624 extern struct domain inetdomain;
625 #endif /* INET || INET6 */
627 static struct protosw ipe4_protosw = {
629 .pr_domain = &inetdomain,
630 .pr_protocol = IPPROTO_IPV4,
631 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
632 .pr_input = ip4_input,
633 .pr_ctloutput = rip_ctloutput,
634 .pr_usrreqs = &rip_usrreqs
637 #if defined(INET6) && defined(INET)
638 static struct ip6protosw ipe6_protosw = {
640 .pr_domain = &inetdomain,
641 .pr_protocol = IPPROTO_IPV6,
642 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
643 .pr_input = ip4_input6,
644 .pr_ctloutput = rip_ctloutput,
645 .pr_usrreqs = &rip_usrreqs
647 #endif /* INET6 && INET */
651 * Check the encapsulated packet to see if we want it
654 ipe4_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
657 * Only take packets coming from IPSEC tunnels; the rest
658 * must be handled by the gif tunnel code. Note that we
659 * also return a minimum priority when we want the packet
660 * so any explicit gif tunnels take precedence.
662 return ((m->m_flags & M_IPSEC) != 0 ? 1 : 0);
670 xform_register(&ipe4_xformsw);
671 /* attach to encapsulation framework */
672 /* XXX save return cookie for detach on module remove */
674 (void) encap_attach_func(AF_INET, -1,
675 ipe4_encapcheck, &ipe4_protosw, NULL);
677 #if defined(INET6) && defined(INET)
678 (void) encap_attach_func(AF_INET6, -1,
679 ipe4_encapcheck, (struct protosw *)&ipe6_protosw, NULL);
682 SYSINIT(ipe4_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipe4_attach, NULL);