1 /* $FreeBSD$ */
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_enc.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/pfil.h>
56 #include <net/route.h>
57 #include <net/netisr.h>
58
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>
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(("%s: dropped due to policy\n", __func__));
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, int off)
134 {
135 #if 0
136 /* If we do not accept IP-in-IP explicitly, drop. */
137 if (!ipip_allow && (m->m_flags & M_IPSEC) == 0) {
138 DPRINTF(("%s: dropped due to policy\n", __func__));
139 ipipstat.ipips_pdrops++;
140 m_freem(m);
141 return;
142 }
143 #endif
144 _ipip_input(m, off, NULL);
145 }
146 #endif /* INET */
147
148 /*
149 * ipip_input gets called when we receive an IP{46} encapsulated packet,
150 * either because we got it at a real interface, or because AH or ESP
151 * were being used in tunnel mode (in which case the rcvif element will
152 * contain the address of the encX interface associated with the tunnel.
153 */
154
155 static void
156 _ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp)
157 {
158 register struct sockaddr_in *sin;
159 register struct ifnet *ifp;
160 register struct ifaddr *ifa;
161 struct ip *ipo;
162 #ifdef INET6
163 register struct sockaddr_in6 *sin6;
164 struct ip6_hdr *ip6 = NULL;
165 u_int8_t itos;
166 #endif
167 u_int8_t nxt;
168 int isr;
169 u_int8_t otos;
170 u_int8_t v;
171 int hlen;
172
173 ipipstat.ipips_ipackets++;
174
175 m_copydata(m, 0, 1, &v);
176
177 switch (v >> 4) {
178 #ifdef INET
179 case 4:
180 hlen = sizeof(struct ip);
181 break;
182 #endif /* INET */
183 #ifdef INET6
184 case 6:
185 hlen = sizeof(struct ip6_hdr);
186 break;
187 #endif
188 default:
189 ipipstat.ipips_family++;
190 m_freem(m);
191 return /* EAFNOSUPPORT */;
192 }
193
194 /* Bring the IP header in the first mbuf, if not there already */
195 if (m->m_len < hlen) {
196 if ((m = m_pullup(m, hlen)) == NULL) {
197 DPRINTF(("%s: m_pullup (1) failed\n", __func__));
198 ipipstat.ipips_hdrops++;
199 return;
200 }
201 }
202
203 ipo = mtod(m, struct ip *);
204
205 #ifdef MROUTING
206 if (ipo->ip_v == IPVERSION && ipo->ip_p == IPPROTO_IPV4) {
207 if (IN_MULTICAST(((struct ip *)((char *) ipo + iphlen))->ip_dst.s_addr)) {
208 ipip_mroute_input (m, iphlen);
209 return;
210 }
211 }
212 #endif /* MROUTING */
213
214 /* Keep outer ecn field. */
215 switch (v >> 4) {
216 #ifdef INET
217 case 4:
218 otos = ipo->ip_tos;
219 break;
220 #endif /* INET */
221 #ifdef INET6
222 case 6:
223 otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
224 break;
225 #endif
226 default:
227 panic("ipip_input: unknown ip version %u (outer)", v>>4);
228 }
229
230 /* Remove outer IP header */
231 m_adj(m, iphlen);
232
233 /* Sanity check */
234 if (m->m_pkthdr.len < sizeof(struct ip)) {
235 ipipstat.ipips_hdrops++;
236 m_freem(m);
237 return;
238 }
239
240 m_copydata(m, 0, 1, &v);
241
242 switch (v >> 4) {
243 #ifdef INET
244 case 4:
245 hlen = sizeof(struct ip);
246 break;
247 #endif /* INET */
248
249 #ifdef INET6
250 case 6:
251 hlen = sizeof(struct ip6_hdr);
252 break;
253 #endif
254 default:
255 ipipstat.ipips_family++;
256 m_freem(m);
257 return; /* EAFNOSUPPORT */
258 }
259
260 /*
261 * Bring the inner IP header in the first mbuf, if not there already.
262 */
263 if (m->m_len < hlen) {
264 if ((m = m_pullup(m, hlen)) == NULL) {
265 DPRINTF(("%s: m_pullup (2) failed\n", __func__));
266 ipipstat.ipips_hdrops++;
267 return;
268 }
269 }
270
271 /*
272 * RFC 1853 specifies that the inner TTL should not be touched on
273 * decapsulation. There's no reason this comment should be here, but
274 * this is as good as any a position.
275 */
276
277 /* Some sanity checks in the inner IP header */
278 switch (v >> 4) {
279 #ifdef INET
280 case 4:
281 ipo = mtod(m, struct ip *);
282 nxt = ipo->ip_p;
283 ip_ecn_egress(ip4_ipsec_ecn, &otos, &ipo->ip_tos);
284 break;
285 #endif /* INET */
286 #ifdef INET6
287 case 6:
288 ip6 = (struct ip6_hdr *) ipo;
289 nxt = ip6->ip6_nxt;
290 itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
291 ip_ecn_egress(ip6_ipsec_ecn, &otos, &itos);
292 ip6->ip6_flow &= ~htonl(0xff << 20);
293 ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
294 break;
295 #endif
296 default:
297 panic("ipip_input: unknown ip version %u (inner)", v>>4);
298 }
299
300 /* Check for local address spoofing. */
301 if ((m->m_pkthdr.rcvif == NULL ||
302 !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
303 ipip_allow != 2) {
304 IFNET_RLOCK();
305 TAILQ_FOREACH(ifp, &ifnet, if_link) {
306 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
307 #ifdef INET
308 if (ipo) {
309 if (ifa->ifa_addr->sa_family !=
310 AF_INET)
311 continue;
312
313 sin = (struct sockaddr_in *) ifa->ifa_addr;
314
315 if (sin->sin_addr.s_addr ==
316 ipo->ip_src.s_addr) {
317 ipipstat.ipips_spoof++;
318 m_freem(m);
319 IFNET_RUNLOCK();
320 return;
321 }
322 }
323 #endif /* INET */
324
325 #ifdef INET6
326 if (ip6) {
327 if (ifa->ifa_addr->sa_family !=
328 AF_INET6)
329 continue;
330
331 sin6 = (struct sockaddr_in6 *) ifa->ifa_addr;
332
333 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_src)) {
334 ipipstat.ipips_spoof++;
335 m_freem(m);
336 IFNET_RUNLOCK();
337 return;
338 }
339
340 }
341 #endif /* INET6 */
342 }
343 }
344 IFNET_RUNLOCK();
345 }
346
347 /* Statistics */
348 ipipstat.ipips_ibytes += m->m_pkthdr.len - iphlen;
349
350 #ifdef DEV_ENC
351 /* pass the mbuf to enc0 for packet filtering */
352 if (ipsec_filter(&m, PFIL_IN) != 0)
353 return;
354 #endif
355
356 /*
357 * Interface pointer stays the same; if no IPsec processing has
358 * been done (or will be done), this will point to a normal
359 * interface. Otherwise, it'll point to an enc interface, which
360 * will allow a packet filter to distinguish between secure and
361 * untrusted packets.
362 */
363
364 switch (v >> 4) {
365 #ifdef INET
366 case 4:
367 isr = NETISR_IP;
368 break;
369 #endif
370 #ifdef INET6
371 case 6:
372 isr = NETISR_IPV6;
373 break;
374 #endif
375 default:
376 panic("%s: bogus ip version %u", __func__, v>>4);
377 }
378
379 if (netisr_queue(isr, m)) { /* (0) on success. */
380 ipipstat.ipips_qfull++;
381 DPRINTF(("%s: packet dropped because of full queue\n",
382 __func__));
383 }
384 }
385
386 int
387 ipip_output(
388 struct mbuf *m,
389 struct ipsecrequest *isr,
390 struct mbuf **mp,
391 int skip,
392 int protoff
393 )
394 {
395 struct secasvar *sav;
396 u_int8_t tp, otos;
397 struct secasindex *saidx;
398 int error;
399 #ifdef INET
400 u_int8_t itos;
401 struct ip *ipo;
402 #endif /* INET */
403 #ifdef INET6
404 struct ip6_hdr *ip6, *ip6o;
405 #endif /* INET6 */
406
407 IPSEC_SPLASSERT_SOFTNET(__func__);
408
409 sav = isr->sav;
410 IPSEC_ASSERT(sav != NULL, ("null SA"));
411 IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
412
413 /* XXX Deal with empty TDB source/destination addresses. */
414
415 m_copydata(m, 0, 1, &tp);
416 tp = (tp >> 4) & 0xff; /* Get the IP version number. */
417
418 saidx = &sav->sah->saidx;
419 switch (saidx->dst.sa.sa_family) {
420 #ifdef INET
421 case AF_INET:
422 if (saidx->src.sa.sa_family != AF_INET ||
423 saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
424 saidx->dst.sin.sin_addr.s_addr == INADDR_ANY) {
425 DPRINTF(("%s: unspecified tunnel endpoint "
426 "address in SA %s/%08lx\n", __func__,
427 ipsec_address(&saidx->dst),
428 (u_long) ntohl(sav->spi)));
429 ipipstat.ipips_unspec++;
430 error = EINVAL;
431 goto bad;
432 }
433
434 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
435 if (m == 0) {
436 DPRINTF(("%s: M_PREPEND failed\n", __func__));
437 ipipstat.ipips_hdrops++;
438 error = ENOBUFS;
439 goto bad;
440 }
441
442 ipo = mtod(m, struct ip *);
443
444 ipo->ip_v = IPVERSION;
445 ipo->ip_hl = 5;
446 ipo->ip_len = htons(m->m_pkthdr.len);
447 ipo->ip_ttl = ip_defttl;
448 ipo->ip_sum = 0;
449 ipo->ip_src = saidx->src.sin.sin_addr;
450 ipo->ip_dst = saidx->dst.sin.sin_addr;
451
452 ipo->ip_id = ip_newid();
453
454 /* If the inner protocol is IP... */
455 if (tp == IPVERSION) {
456 /* Save ECN notification */
457 m_copydata(m, sizeof(struct ip) +
458 offsetof(struct ip, ip_tos),
459 sizeof(u_int8_t), (caddr_t) &itos);
460
461 ipo->ip_p = IPPROTO_IPIP;
462
463 /*
464 * We should be keeping tunnel soft-state and
465 * send back ICMPs if needed.
466 */
467 m_copydata(m, sizeof(struct ip) +
468 offsetof(struct ip, ip_off),
469 sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
470 ipo->ip_off = ntohs(ipo->ip_off);
471 ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
472 ipo->ip_off = htons(ipo->ip_off);
473 }
474 #ifdef INET6
475 else if (tp == (IPV6_VERSION >> 4)) {
476 u_int32_t itos32;
477
478 /* Save ECN notification. */
479 m_copydata(m, sizeof(struct ip) +
480 offsetof(struct ip6_hdr, ip6_flow),
481 sizeof(u_int32_t), (caddr_t) &itos32);
482 itos = ntohl(itos32) >> 20;
483 ipo->ip_p = IPPROTO_IPV6;
484 ipo->ip_off = 0;
485 }
486 #endif /* INET6 */
487 else {
488 goto nofamily;
489 }
490
491 otos = 0;
492 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
493 ipo->ip_tos = otos;
494 break;
495 #endif /* INET */
496
497 #ifdef INET6
498 case AF_INET6:
499 if (IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr) ||
500 saidx->src.sa.sa_family != AF_INET6 ||
501 IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr)) {
502 DPRINTF(("%s: unspecified tunnel endpoint "
503 "address in SA %s/%08lx\n", __func__,
504 ipsec_address(&saidx->dst),
505 (u_long) ntohl(sav->spi)));
506 ipipstat.ipips_unspec++;
507 error = ENOBUFS;
508 goto bad;
509 }
510
511 /* scoped address handling */
512 ip6 = mtod(m, struct ip6_hdr *);
513 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
514 ip6->ip6_src.s6_addr16[1] = 0;
515 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
516 ip6->ip6_dst.s6_addr16[1] = 0;
517
518 M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
519 if (m == 0) {
520 DPRINTF(("%s: M_PREPEND failed\n", __func__));
521 ipipstat.ipips_hdrops++;
522 error = ENOBUFS;
523 goto bad;
524 }
525
526 /* Initialize IPv6 header */
527 ip6o = mtod(m, struct ip6_hdr *);
528 ip6o->ip6_flow = 0;
529 ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
530 ip6o->ip6_vfc |= IPV6_VERSION;
531 ip6o->ip6_plen = htons(m->m_pkthdr.len);
532 ip6o->ip6_hlim = ip_defttl;
533 ip6o->ip6_dst = saidx->dst.sin6.sin6_addr;
534 ip6o->ip6_src = saidx->src.sin6.sin6_addr;
535
536 #ifdef INET
537 if (tp == IPVERSION) {
538 /* Save ECN notification */
539 m_copydata(m, sizeof(struct ip6_hdr) +
540 offsetof(struct ip, ip_tos), sizeof(u_int8_t),
541 (caddr_t) &itos);
542
543 /* This is really IPVERSION. */
544 ip6o->ip6_nxt = IPPROTO_IPIP;
545 } else
546 #endif /* INET */
547 if (tp == (IPV6_VERSION >> 4)) {
548 u_int32_t itos32;
549
550 /* Save ECN notification. */
551 m_copydata(m, sizeof(struct ip6_hdr) +
552 offsetof(struct ip6_hdr, ip6_flow),
553 sizeof(u_int32_t), (caddr_t) &itos32);
554 itos = ntohl(itos32) >> 20;
555
556 ip6o->ip6_nxt = IPPROTO_IPV6;
557 } else {
558 goto nofamily;
559 }
560
561 otos = 0;
562 ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
563 ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
564 break;
565 #endif /* INET6 */
566
567 default:
568 nofamily:
569 DPRINTF(("%s: unsupported protocol family %u\n", __func__,
570 saidx->dst.sa.sa_family));
571 ipipstat.ipips_family++;
572 error = EAFNOSUPPORT; /* XXX diffs from openbsd */
573 goto bad;
574 }
575
576 ipipstat.ipips_opackets++;
577 *mp = m;
578
579 #ifdef INET
580 if (saidx->dst.sa.sa_family == AF_INET) {
581 #if 0
582 if (sav->tdb_xform->xf_type == XF_IP4)
583 tdb->tdb_cur_bytes +=
584 m->m_pkthdr.len - sizeof(struct ip);
585 #endif
586 ipipstat.ipips_obytes += m->m_pkthdr.len - sizeof(struct ip);
587 }
588 #endif /* INET */
589
590 #ifdef INET6
591 if (saidx->dst.sa.sa_family == AF_INET6) {
592 #if 0
593 if (sav->tdb_xform->xf_type == XF_IP4)
594 tdb->tdb_cur_bytes +=
595 m->m_pkthdr.len - sizeof(struct ip6_hdr);
596 #endif
597 ipipstat.ipips_obytes +=
598 m->m_pkthdr.len - sizeof(struct ip6_hdr);
599 }
600 #endif /* INET6 */
601
602 return 0;
603 bad:
604 if (m)
605 m_freem(m);
606 *mp = NULL;
607 return (error);
608 }
609
610 #ifdef FAST_IPSEC
611 static int
612 ipe4_init(struct secasvar *sav, struct xformsw *xsp)
613 {
614 sav->tdb_xform = xsp;
615 return 0;
616 }
617
618 static int
619 ipe4_zeroize(struct secasvar *sav)
620 {
621 sav->tdb_xform = NULL;
622 return 0;
623 }
624
625 static int
626 ipe4_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff)
627 {
628 /* This is a rather serious mistake, so no conditional printing. */
629 printf("%s: should never be called\n", __func__);
630 if (m)
631 m_freem(m);
632 return EOPNOTSUPP;
633 }
634
635 static struct xformsw ipe4_xformsw = {
636 XF_IP4, 0, "IPv4 Simple Encapsulation",
637 ipe4_init, ipe4_zeroize, ipe4_input, ipip_output,
638 };
639
640 extern struct domain inetdomain;
641 static struct protosw ipe4_protosw =
642 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR|PR_LASTHDR,
643 ip4_input,
644 0, 0, rip_ctloutput,
645 0,
646 0, 0, 0, 0,
647 &rip_usrreqs
648 };
649 #ifdef INET6
650 static struct ip6protosw ipe6_protosw =
651 { SOCK_RAW, &inetdomain, IPPROTO_IPV6, PR_ATOMIC|PR_ADDR|PR_LASTHDR,
652 ip4_input6,
653 0, 0, rip_ctloutput,
654 0,
655 0, 0, 0, 0,
656 &rip_usrreqs
657 };
658 #endif
659
660 /*
661 * Check the encapsulated packet to see if we want it
662 */
663 static int
664 ipe4_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
665 {
666 /*
667 * Only take packets coming from IPSEC tunnels; the rest
668 * must be handled by the gif tunnel code. Note that we
669 * also return a minimum priority when we want the packet
670 * so any explicit gif tunnels take precedence.
671 */
672 return ((m->m_flags & M_IPSEC) != 0 ? 1 : 0);
673 }
674
675 static void
676 ipe4_attach(void)
677 {
678 xform_register(&ipe4_xformsw);
679 /* attach to encapsulation framework */
680 /* XXX save return cookie for detach on module remove */
681 (void) encap_attach_func(AF_INET, -1,
682 ipe4_encapcheck, &ipe4_protosw, NULL);
683 #ifdef INET6
684 (void) encap_attach_func(AF_INET6, -1,
685 ipe4_encapcheck, (struct protosw *)&ipe6_protosw, NULL);
686 #endif
687 }
688 SYSINIT(ipe4_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipe4_attach, NULL);
689 #endif /* FAST_IPSEC */
Cache object: 87ce53afa8ec9defcb0d01a280b8a19f
|