1 /* $FreeBSD: releng/5.1/sys/netipsec/xform_ipip.c 111888 2003-03-04 23:19:55Z jlemon $ */
2 /* $OpenBSD: ip_ipip.c,v 1.25 2002/06/10 18:04:55 itojun Exp $ */
3 /*
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).
7 *
8 * The original version of this code was written by John Ioannidis
9 * for BSD/OS in Athens, Greece, in November 1995.
10 *
11 * Ported to OpenBSD and NetBSD, with additional transforms, in December 1996,
12 * by Angelos D. Keromytis.
13 *
14 * Additional transforms and features in 1997 and 1998 by Angelos D. Keromytis
15 * and Niels Provos.
16 *
17 * Additional features in 1999 by Angelos D. Keromytis.
18 *
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.
22 *
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
30 * all.
31 *
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
36 * PURPOSE.
37 */
38
39 /*
40 * IP-inside-IP processing
41 */
42 #include "opt_inet.h"
43 #include "opt_inet6.h"
44 #include "opt_random_ip_id.h"
45
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/mbuf.h>
49 #include <sys/socket.h>
50 #include <sys/kernel.h>
51 #include <sys/protosw.h>
52 #include <sys/sysctl.h>
53
54 #include <net/if.h>
55 #include <net/route.h>
56 #include <net/netisr.h>
57
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>
65 #include <netinet/ipprotosw.h>
66
67 #include <netipsec/ipsec.h>
68 #include <netipsec/xform.h>
69
70 #include <netipsec/ipip_var.h>
71
72 #ifdef MROUTING
73 #include <netinet/ip_mroute.h>
74 #endif
75
76 #ifdef INET6
77 #include <netinet/ip6.h>
78 #include <netipsec/ipsec6.h>
79 #include <netinet6/ip6_ecn.h>
80 #include <netinet6/in6_var.h>
81 #include <netinet6/ip6protosw.h>
82 #endif
83
84 #include <netipsec/key.h>
85 #include <netipsec/key_debug.h>
86
87 #include <machine/stdarg.h>
88
89 /*
90 * We can control the acceptance of IP4 packets by altering the sysctl
91 * net.inet.ipip.allow value. Zero means drop them, all else is acceptance.
92 */
93 int ipip_allow = 0;
94 struct ipipstat ipipstat;
95
96 SYSCTL_DECL(_net_inet_ipip);
97 SYSCTL_INT(_net_inet_ipip, OID_AUTO,
98 ipip_allow, CTLFLAG_RW, &ipip_allow, 0, "");
99 SYSCTL_STRUCT(_net_inet_ipip, IPSECCTL_STATS,
100 stats, CTLFLAG_RD, &ipipstat, ipipstat, "");
101
102 /* XXX IPCOMP */
103 #define M_IPSEC (M_AUTHIPHDR|M_AUTHIPDGM|M_DECRYPTED)
104
105 static void _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp);
106
107 #ifdef INET6
108 /*
109 * Really only a wrapper for ipip_input(), for use with IPv6.
110 */
111 int
112 ip4_input6(struct mbuf **m, int *offp, int proto)
113 {
114 #if 0
115 /* If we do not accept IP-in-IP explicitly, drop. */
116 if (!ipip_allow && ((*m)->m_flags & M_IPSEC) == 0) {
117 DPRINTF(("ip4_input6: dropped due to policy\n"));
118 ipipstat.ipips_pdrops++;
119 m_freem(*m);
120 return IPPROTO_DONE;
121 }
122 #endif
123 _ipip_input(*m, *offp, NULL);
124 return IPPROTO_DONE;
125 }
126 #endif /* INET6 */
127
128 #ifdef INET
129 /*
130 * Really only a wrapper for ipip_input(), for use with IPv4.
131 */
132 void
133 ip4_input(struct mbuf *m, ...)
134 {
135 va_list ap;
136 int iphlen;
137
138 #if 0
139 /* If we do not accept IP-in-IP explicitly, drop. */
140 if (!ipip_allow && (m->m_flags & M_IPSEC) == 0) {
141 DPRINTF(("ip4_input: dropped due to policy\n"));
142 ipipstat.ipips_pdrops++;
143 m_freem(m);
144 return;
145 }
146 #endif
147 va_start(ap, m);
148 iphlen = va_arg(ap, int);
149 va_end(ap);
150
151 _ipip_input(m, iphlen, NULL);
152 }
153 #endif /* INET */
154
155 /*
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.
160 */
161
162 static void
163 _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp)
164 {
165 register struct sockaddr_in *sin;
166 register struct ifnet *ifp;
167 register struct ifaddr *ifa;
168 struct ip *ipo;
169 #ifdef INET6
170 register struct sockaddr_in6 *sin6;
171 struct ip6_hdr *ip6 = NULL;
172 u_int8_t itos;
173 #endif
174 u_int8_t nxt;
175 int isr;
176 u_int8_t otos;
177 u_int8_t v;
178 int hlen;
179
180 ipipstat.ipips_ipackets++;
181
182 m_copydata(m, 0, 1, &v);
183
184 switch (v >> 4) {
185 #ifdef INET
186 case 4:
187 hlen = sizeof(struct ip);
188 break;
189 #endif /* INET */
190 #ifdef INET6
191 case 6:
192 hlen = sizeof(struct ip6_hdr);
193 break;
194 #endif
195 default:
196 ipipstat.ipips_family++;
197 m_freem(m);
198 return /* EAFNOSUPPORT */;
199 }
200
201 /* Bring the IP header in the first mbuf, if not there already */
202 if (m->m_len < hlen) {
203 if ((m = m_pullup(m, hlen)) == NULL) {
204 DPRINTF(("ipip_input: m_pullup (1) failed\n"));
205 ipipstat.ipips_hdrops++;
206 return;
207 }
208 }
209
210 ipo = mtod(m, struct ip *);
211
212 #ifdef MROUTING
213 if (ipo->ip_v == IPVERSION && ipo->ip_p == IPPROTO_IPV4) {
214 if (IN_MULTICAST(((struct ip *)((char *) ipo + iphlen))->ip_dst.s_addr)) {
215 ipip_mroute_input (m, iphlen);
216 return;
217 }
218 }
219 #endif /* MROUTING */
220
221 /* Keep outer ecn field. */
222 switch (v >> 4) {
223 #ifdef INET
224 case 4:
225 otos = ipo->ip_tos;
226 break;
227 #endif /* INET */
228 #ifdef INET6
229 case 6:
230 otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
231 break;
232 #endif
233 default:
234 panic("ipip_input: unknown ip version %u (outer)", v>>4);
235 }
236
237 /* Remove outer IP header */
238 m_adj(m, iphlen);
239
240 /* Sanity check */
241 if (m->m_pkthdr.len < sizeof(struct ip)) {
242 ipipstat.ipips_hdrops++;
243 m_freem(m);
244 return;
245 }
246
247 m_copydata(m, 0, 1, &v);
248
249 switch (v >> 4) {
250 #ifdef INET
251 case 4:
252 hlen = sizeof(struct ip);
253 break;
254 #endif /* INET */
255
256 #ifdef INET6
257 case 6:
258 hlen = sizeof(struct ip6_hdr);
259 break;
260 #endif
261 default:
262 ipipstat.ipips_family++;
263 m_freem(m);
264 return; /* EAFNOSUPPORT */
265 }
266
267 /*
268 * Bring the inner IP header in the first mbuf, if not there already.
269 */
270 if (m->m_len < hlen) {
271 if ((m = m_pullup(m, hlen)) == NULL) {
272 DPRINTF(("ipip_input: m_pullup (2) failed\n"));
273 ipipstat.ipips_hdrops++;
274 return;
275 }
276 }
277
278 /*
279 * RFC 1853 specifies that the inner TTL should not be touched on
280 * decapsulation. There's no reason this comment should be here, but
281 * this is as good as any a position.
282 */
283
284 /* Some sanity checks in the inner IP header */
285 switch (v >> 4) {
286 #ifdef INET
287 case 4:
288 ipo = mtod(m, struct ip *);
289 nxt = ipo->ip_p;
290 ip_ecn_egress(ip4_ipsec_ecn, &otos, &ipo->ip_tos);
291 break;
292 #endif /* INET */
293 #ifdef INET6
294 case 6:
295 ip6 = (struct ip6_hdr *) ipo;
296 nxt = ip6->ip6_nxt;
297 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
298 ip_ecn_egress(ip6_ipsec_ecn, &otos, &itos);
299 ip6->ip6_flow &= ~htonl(0xff << 20);
300 ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
301 break;
302 #endif
303 default:
304 panic("ipip_input: unknown ip version %u (inner)", v>>4);
305 }
306
307 /* Check for local address spoofing. */
308 if ((m->m_pkthdr.rcvif == NULL ||
309 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
310 ipip_allow != 2) {
311 IFNET_RLOCK();
312 for (ifp = ifnet.tqh_first; ifp != 0;
313 ifp = ifp->if_list.tqe_next) {
314 for (ifa = ifp->if_addrlist.tqh_first; ifa != 0;
315 ifa = ifa->ifa_list.tqe_next) {
316 #ifdef INET
317 if (ipo) {
318 if (ifa->ifa_addr->sa_family !=
319 AF_INET)
320 continue;
321
322 sin = (struct sockaddr_in *) ifa->ifa_addr;
323
324 if (sin->sin_addr.s_addr ==
325 ipo->ip_src.s_addr) {
326 ipipstat.ipips_spoof++;
327 m_freem(m);
328 IFNET_RUNLOCK();
329 return;
330 }
331 }
332 #endif /* INET */
333
334 #ifdef INET6
335 if (ip6) {
336 if (ifa->ifa_addr->sa_family !=
337 AF_INET6)
338 continue;
339
340 sin6 = (struct sockaddr_in6 *) ifa->ifa_addr;
341
342 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_src)) {
343 ipipstat.ipips_spoof++;
344 m_freem(m);
345 IFNET_RUNLOCK();
346 return;
347 }
348
349 }
350 #endif /* INET6 */
351 }
352 }
353 IFNET_RUNLOCK();
354 }
355
356 /* Statistics */
357 ipipstat.ipips_ibytes += m->m_pkthdr.len - iphlen;
358
359 /*
360 * Interface pointer stays the same; if no IPsec processing has
361 * been done (or will be done), this will point to a normal
362 * interface. Otherwise, it'll point to an enc interface, which
363 * will allow a packet filter to distinguish between secure and
364 * untrusted packets.
365 */
366
367 switch (v >> 4) {
368 #ifdef INET
369 case 4:
370 isr = NETISR_IP;
371 break;
372 #endif
373 #ifdef INET6
374 case 6:
375 isr = NETISR_IPV6;
376 break;
377 #endif
378 default:
379 panic("ipip_input: should never reach here");
380 }
381
382 if (!netisr_queue(isr, m)) {
383 ipipstat.ipips_qfull++;
384 DPRINTF(("ipip_input: packet dropped because of full queue\n"));
385 }
386 }
387
388 int
389 ipip_output(
390 struct mbuf *m,
391 struct ipsecrequest *isr,
392 struct mbuf **mp,
393 int skip,
394 int protoff
395 )
396 {
397 struct secasvar *sav;
398 u_int8_t tp, otos;
399 struct secasindex *saidx;
400 int error;
401 #ifdef INET
402 u_int8_t itos;
403 struct ip *ipo;
404 #endif /* INET */
405 #ifdef INET6
406 struct ip6_hdr *ip6, *ip6o;
407 #endif /* INET6 */
408
409 #if 0
410 SPLASSERT(net, "ipip_output");
411 #endif
412
413 sav = isr->sav;
414 KASSERT(sav != NULL, ("ipip_output: null SA"));
415 KASSERT(sav->sah != NULL, ("ipip_output: null SAH"));
416
417 /* XXX Deal with empty TDB source/destination addresses. */
418
419 m_copydata(m, 0, 1, &tp);
420 tp = (tp >> 4) & 0xff; /* Get the IP version number. */
421
422 saidx = &sav->sah->saidx;
423 switch (saidx->dst.sa.sa_family) {
424 #ifdef INET
425 case AF_INET:
426 if (saidx->src.sa.sa_family != AF_INET ||
427 saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
428 saidx->dst.sin.sin_addr.s_addr == INADDR_ANY) {
429 DPRINTF(("ipip_output: unspecified tunnel endpoint "
430 "address in SA %s/%08lx\n",
431 ipsec_address(&saidx->dst),
432 (u_long) ntohl(sav->spi)));
433 ipipstat.ipips_unspec++;
434 error = EINVAL;
435 goto bad;
436 }
437
438 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
439 if (m == 0) {
440 DPRINTF(("ipip_output: M_PREPEND failed\n"));
441 ipipstat.ipips_hdrops++;
442 error = ENOBUFS;
443 goto bad;
444 }
445
446 ipo = mtod(m, struct ip *);
447
448 ipo->ip_v = IPVERSION;
449 ipo->ip_hl = 5;
450 ipo->ip_len = htons(m->m_pkthdr.len);
451 ipo->ip_ttl = ip_defttl;
452 ipo->ip_sum = 0;
453 ipo->ip_src = saidx->src.sin.sin_addr;
454 ipo->ip_dst = saidx->dst.sin.sin_addr;
455
456 #ifdef RANDOM_IP_ID
457 ipo->ip_id = ip_randomid();
458 #else
459 ipo->ip_id = htons(ip_id++);
460 #endif
461
462 /* If the inner protocol is IP... */
463 if (tp == IPVERSION) {
464 /* Save ECN notification */
465 m_copydata(m, sizeof(struct ip) +
466 offsetof(struct ip, ip_tos),
467 sizeof(u_int8_t), (caddr_t) &itos);
468
469 ipo->ip_p = IPPROTO_IPIP;
470
471 /*
472 * We should be keeping tunnel soft-state and
473 * send back ICMPs if needed.
474 */
475 m_copydata(m, sizeof(struct ip) +
476 offsetof(struct ip, ip_off),
477 sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
478 ipo->ip_off = ntohs(ipo->ip_off);
479 ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
480 ipo->ip_off = htons(ipo->ip_off);
481 }
482 #ifdef INET6
483 else if (tp == (IPV6_VERSION >> 4)) {
484 u_int32_t itos32;
485
486 /* Save ECN notification. */
487 m_copydata(m, sizeof(struct ip) +
488 offsetof(struct ip6_hdr, ip6_flow),
489 sizeof(u_int32_t), (caddr_t) &itos32);
490 itos = ntohl(itos32) >> 20;
491 ipo->ip_p = IPPROTO_IPV6;
492 ipo->ip_off = 0;
493 }
494 #endif /* INET6 */
495 else {
496 goto nofamily;
497 }
498
499 otos = 0;
500 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
501 ipo->ip_tos = otos;
502 break;
503 #endif /* INET */
504
505 #ifdef INET6
506 case AF_INET6:
507 if (IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr) ||
508 saidx->src.sa.sa_family != AF_INET6 ||
509 IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr)) {
510 DPRINTF(("ipip_output: unspecified tunnel endpoint "
511 "address in SA %s/%08lx\n",
512 ipsec_address(&saidx->dst),
513 (u_long) ntohl(sav->spi)));
514 ipipstat.ipips_unspec++;
515 error = ENOBUFS;
516 goto bad;
517 }
518
519 /* scoped address handling */
520 ip6 = mtod(m, struct ip6_hdr *);
521 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
522 ip6->ip6_src.s6_addr16[1] = 0;
523 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
524 ip6->ip6_dst.s6_addr16[1] = 0;
525
526 M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
527 if (m == 0) {
528 DPRINTF(("ipip_output: M_PREPEND failed\n"));
529 ipipstat.ipips_hdrops++;
530 *mp = NULL;
531 error = ENOBUFS;
532 goto bad;
533 }
534
535 /* Initialize IPv6 header */
536 ip6o = mtod(m, struct ip6_hdr *);
537 ip6o->ip6_flow = 0;
538 ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
539 ip6o->ip6_vfc |= IPV6_VERSION;
540 ip6o->ip6_plen = htons(m->m_pkthdr.len);
541 ip6o->ip6_hlim = ip_defttl;
542 ip6o->ip6_dst = saidx->dst.sin6.sin6_addr;
543 ip6o->ip6_src = saidx->src.sin6.sin6_addr;
544
545 #ifdef INET
546 if (tp == IPVERSION) {
547 /* Save ECN notification */
548 m_copydata(m, sizeof(struct ip6_hdr) +
549 offsetof(struct ip, ip_tos), sizeof(u_int8_t),
550 (caddr_t) &itos);
551
552 /* This is really IPVERSION. */
553 ip6o->ip6_nxt = IPPROTO_IPIP;
554 } else
555 #endif /* INET */
556 if (tp == (IPV6_VERSION >> 4)) {
557 u_int32_t itos32;
558
559 /* Save ECN notification. */
560 m_copydata(m, sizeof(struct ip6_hdr) +
561 offsetof(struct ip6_hdr, ip6_flow),
562 sizeof(u_int32_t), (caddr_t) &itos32);
563 itos = ntohl(itos32) >> 20;
564
565 ip6o->ip6_nxt = IPPROTO_IPV6;
566 } else {
567 goto nofamily;
568 }
569
570 otos = 0;
571 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
572 ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
573 break;
574 #endif /* INET6 */
575
576 default:
577 nofamily:
578 DPRINTF(("ipip_output: unsupported protocol family %u\n",
579 saidx->dst.sa.sa_family));
580 ipipstat.ipips_family++;
581 error = EAFNOSUPPORT; /* XXX diffs from openbsd */
582 goto bad;
583 }
584
585 ipipstat.ipips_opackets++;
586 *mp = m;
587
588 #ifdef INET
589 if (saidx->dst.sa.sa_family == AF_INET) {
590 #if 0
591 if (sav->tdb_xform->xf_type == XF_IP4)
592 tdb->tdb_cur_bytes +=
593 m->m_pkthdr.len - sizeof(struct ip);
594 #endif
595 ipipstat.ipips_obytes += m->m_pkthdr.len - sizeof(struct ip);
596 }
597 #endif /* INET */
598
599 #ifdef INET6
600 if (saidx->dst.sa.sa_family == AF_INET6) {
601 #if 0
602 if (sav->tdb_xform->xf_type == XF_IP4)
603 tdb->tdb_cur_bytes +=
604 m->m_pkthdr.len - sizeof(struct ip6_hdr);
605 #endif
606 ipipstat.ipips_obytes +=
607 m->m_pkthdr.len - sizeof(struct ip6_hdr);
608 }
609 #endif /* INET6 */
610
611 return 0;
612 bad:
613 if (m)
614 m_freem(m), *mp = NULL;
615 return (error);
616 }
617
618 #ifdef FAST_IPSEC
619 static int
620 ipe4_init(struct secasvar *sav, struct xformsw *xsp)
621 {
622 sav->tdb_xform = xsp;
623 return 0;
624 }
625
626 static int
627 ipe4_zeroize(struct secasvar *sav)
628 {
629 sav->tdb_xform = NULL;
630 return 0;
631 }
632
633 static int
634 ipe4_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff)
635 {
636 /* This is a rather serious mistake, so no conditional printing. */
637 printf("ipe4_input: should never be called\n");
638 if (m)
639 m_freem(m);
640 return EOPNOTSUPP;
641 }
642
643 static struct xformsw ipe4_xformsw = {
644 XF_IP4, 0, "IPv4 Simple Encapsulation",
645 ipe4_init, ipe4_zeroize, ipe4_input, ipip_output,
646 };
647
648 extern struct domain inetdomain;
649 static struct ipprotosw ipe4_protosw[] = {
650 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR|PR_LASTHDR,
651 (pr_in_input_t*) ip4_input,
652 0, 0, rip_ctloutput,
653 0,
654 0, 0, 0, 0,
655 &rip_usrreqs
656 },
657 #ifdef INET6
658 { SOCK_RAW, &inetdomain, IPPROTO_IPV6, PR_ATOMIC|PR_ADDR|PR_LASTHDR,
659 (pr_in_input_t*) ip4_input,
660 0, 0, rip_ctloutput,
661 0,
662 0, 0, 0, 0,
663 &rip_usrreqs
664 }
665 #endif
666 };
667
668 /*
669 * Check the encapsulated packet to see if we want it
670 */
671 static int
672 ipe4_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
673 {
674 /*
675 * Only take packets coming from IPSEC tunnels; the rest
676 * must be handled by the gif tunnel code. Note that we
677 * also return a minimum priority when we want the packet
678 * so any explicit gif tunnels take precedence.
679 */
680 return ((m->m_flags & M_IPSEC) != 0 ? 1 : 0);
681 }
682
683 static void
684 ipe4_attach(void)
685 {
686 xform_register(&ipe4_xformsw);
687 /* attach to encapsulation framework */
688 /* XXX save return cookie for detach on module remove */
689 (void) encap_attach_func(AF_INET, -1,
690 ipe4_encapcheck, (struct protosw*) &ipe4_protosw[0], NULL);
691 #ifdef INET6
692 (void) encap_attach_func(AF_INET6, -1,
693 ipe4_encapcheck, (struct protosw*) &ipe4_protosw[1], NULL);
694 #endif
695 }
696 SYSINIT(ipe4_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipe4_attach, NULL);
697 #endif /* FAST_IPSEC */
Cache object: acb50488b8a29b18b5b1123f5f0308d1
|