xref: /openbsd-src/sys/netinet/ip_ipip.c (revision b2ea75c1b17e1a9a339660e7ed45cd24946b230e)
1 /*	$OpenBSD: ip_ipip.c,v 1.20 2001/07/04 23:14:55 espie Exp $ */
2 /*
3  * The authors of this code are John Ioannidis (ji@tla.org),
4  * Angelos D. Keromytis (kermit@csd.uch.gr) and
5  * Niels Provos (provos@physnet.uni-hamburg.de).
6  *
7  * The original version of this code was written by John Ioannidis
8  * for BSD/OS in Athens, Greece, in November 1995.
9  *
10  * Ported to OpenBSD and NetBSD, with additional transforms, in December 1996,
11  * by Angelos D. Keromytis.
12  *
13  * Additional transforms and features in 1997 and 1998 by Angelos D. Keromytis
14  * and Niels Provos.
15  *
16  * Additional features in 1999 by Angelos D. Keromytis.
17  *
18  * Copyright (C) 1995, 1996, 1997, 1998, 1999 by John Ioannidis,
19  * Angelos D. Keromytis and Niels Provos.
20  * Copyright (c) 2001, Angelos D. Keromytis.
21  *
22  * Permission to use, copy, and modify this software with or without fee
23  * is hereby granted, provided that this entire notice is included in
24  * all copies of any software which is or includes a copy or
25  * modification of this software.
26  * You may use this code under the GNU public license if you so wish. Please
27  * contribute changes back to the authors under this freer than GPL license
28  * so that we may further the use of strong encryption without limitations to
29  * all.
30  *
31  * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR
32  * IMPLIED WARRANTY. IN PARTICULAR, NONE OF THE AUTHORS MAKES ANY
33  * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE
34  * MERCHANTABILITY OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR
35  * PURPOSE.
36  */
37 
38 /*
39  * IP-inside-IP processing
40  */
41 
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/mbuf.h>
45 #include <sys/socket.h>
46 #include <sys/sysctl.h>
47 
48 #include <net/if.h>
49 #include <net/route.h>
50 #include <net/netisr.h>
51 
52 #include <netinet/in.h>
53 #include <netinet/in_systm.h>
54 #include <netinet/ip.h>
55 #include <netinet/in_pcb.h>
56 #include <netinet/in_var.h>
57 #include <netinet/ip_var.h>
58 #include <netinet/ip_ecn.h>
59 
60 #ifdef MROUTING
61 #include <netinet/ip_mroute.h>
62 #endif
63 
64 #include <netinet/ip_ipsp.h>
65 #include <netinet/ip_ipip.h>
66 
67 #ifdef ENCDEBUG
68 #define DPRINTF(x)	if (encdebug) printf x
69 #else
70 #define DPRINTF(x)
71 #endif
72 
73 /*
74  * We can control the acceptance of IP4 packets by altering the sysctl
75  * net.inet.ipip.allow value.  Zero means drop them, all else is acceptance.
76  */
77 int ipip_allow = 0;
78 
79 struct ipipstat ipipstat;
80 
81 #ifdef INET6
82 /*
83  * Really only a wrapper for ipip_input(), for use with IPv6.
84  */
85 int
86 ip4_input6(struct mbuf **m, int *offp, int proto)
87 {
88 	/* If we do not accept IPv4 explicitly, drop.  */
89 	if (!ipip_allow && ((*m)->m_flags & (M_AUTH|M_CONF)) == 0) {
90 		DPRINTF(("ip4_input6(): dropped due to policy\n"));
91 		ipipstat.ipips_pdrops++;
92 		m_freem(*m);
93 		return IPPROTO_DONE;
94 	}
95 
96 	ipip_input(*m, *offp);
97 	return IPPROTO_DONE;
98 }
99 #endif /* INET6 */
100 
101 #ifdef INET
102 /*
103  * Really only a wrapper for ipip_input(), for use with IPv4.
104  */
105 void
106 ip4_input(struct mbuf *m, ...)
107 {
108 	va_list ap;
109 	int iphlen;
110 
111 	/* If we do not accept IPv4 explicitly, drop.  */
112 	if (!ipip_allow && (m->m_flags & (M_AUTH|M_CONF)) == 0) {
113 		DPRINTF(("ip4_input(): dropped due to policy\n"));
114 		ipipstat.ipips_pdrops++;
115 		m_freem(m);
116 		return;
117 	}
118 
119 	va_start(ap, m);
120 	iphlen = va_arg(ap, int);
121 	va_end(ap);
122 
123 	ipip_input(m, iphlen);
124 }
125 #endif /* INET */
126 
127 /*
128  * ipip_input gets called when we receive an IP{46} encapsulated packet,
129  * either because we got it at a real interface, or because AH or ESP
130  * were being used in tunnel mode (in which case the rcvif element will
131  * contain the address of the encX interface associated with the tunnel.
132  */
133 
134 void
135 ipip_input(struct mbuf *m, int iphlen)
136 {
137 	register struct sockaddr_in *sin;
138 	register struct ifnet *ifp;
139 	register struct ifaddr *ifa;
140 	struct ifqueue *ifq = NULL;
141 	struct ip *ipo;
142 #ifdef INET6
143 	register struct sockaddr_in6 *sin6;
144 	struct ip6_hdr *ip6 = NULL;
145 	u_int8_t itos;
146 #endif
147 	u_int8_t nxt;
148 	int isr;
149 	u_int8_t otos;
150 	u_int8_t v;
151 	int hlen, s;
152 
153 	ipipstat.ipips_ipackets++;
154 
155 	m_copydata(m, 0, 1, &v);
156 
157 	switch (v >> 4) {
158 #ifdef INET
159         case 4:
160 		hlen = sizeof(struct ip);
161 		break;
162 #endif /* INET */
163 
164 #ifdef INET6
165         case 6:
166 		hlen = sizeof(struct ip6_hdr);
167 		break;
168 #endif
169         default:
170 		m_freem(m);
171 		return /* EAFNOSUPPORT */;
172 	}
173 
174 	/* Bring the IP header in the first mbuf, if not there already */
175 	if (m->m_len < hlen) {
176 		if ((m = m_pullup(m, hlen)) == NULL) {
177 			DPRINTF(("ipip_input(): m_pullup() failed\n"));
178 			ipipstat.ipips_hdrops++;
179 			return;
180 		}
181 	}
182 
183 	ipo = mtod(m, struct ip *);
184 
185 #ifdef MROUTING
186 	if (ipo->ip_v == IPVERSION && ipo->ip_p == IPPROTO_IPV4) {
187 		if (IN_MULTICAST(((struct ip *)((char *) ipo + iphlen))->ip_dst.s_addr)) {
188 			ipip_mroute_input (m, iphlen);
189 			return;
190 		}
191 	}
192 #endif /* MROUTING */
193 
194 	/* Keep outer ecn field. */
195 #ifdef INET
196 	if ((v >> 4) == 4)
197 		otos = ipo->ip_tos;
198 #endif /* INET */
199 
200 #ifdef INET6
201 	if ((v >> 4) == 6)
202 		otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
203 #endif
204 
205 	/* Remove outer IP header */
206 	m_adj(m, iphlen);
207 
208 	m_copydata(m, 0, 1, &v);
209 
210 	switch (v >> 4) {
211 #ifdef INET
212         case 4:
213 		hlen = sizeof(struct ip);
214 		break;
215 #endif /* INET */
216 
217 #ifdef INET6
218         case 6:
219 		hlen = sizeof(struct ip6_hdr);
220 		break;
221 #endif
222 
223         default:
224 		m_freem(m);
225 		return /* EAFNOSUPPORT */;
226 	}
227 
228 	/*
229 	 * Bring the inner IP header in the first mbuf, if not there already.
230 	 */
231 	if (m->m_len < hlen) {
232 		if ((m = m_pullup(m, hlen)) == NULL) {
233 			DPRINTF(("ipip_input(): m_pullup() failed\n"));
234 			ipipstat.ipips_hdrops++;
235 			return;
236 		}
237 	}
238 
239 	/*
240 	 * RFC 1853 specifies that the inner TTL should not be touched on
241 	 * decapsulation. There's no reason this comment should be here, but
242 	 * this is as good as any a position.
243 	 */
244 
245 	/* Some sanity checks in the inner IPv4 header */
246 	switch (v >> 4) {
247 #ifdef INET
248     	case 4:
249                 ipo = mtod(m, struct ip *);
250                 nxt = ipo->ip_p;
251 		ip_ecn_egress(ECN_ALLOWED, &otos, &ipo->ip_tos);
252                 break;
253 #endif /* INET */
254 
255 #ifdef INET6
256     	case 6:
257                 ip6 = (struct ip6_hdr *) ipo;
258                 nxt = ip6->ip6_nxt;
259 		itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
260 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
261 		ip6->ip6_flow &= ~htonl(0xff << 20);
262 		ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
263                 break;
264 #endif
265 	}
266 
267 	/* Check for local address spoofing. */
268 	if ((m->m_pkthdr.rcvif == NULL ||
269 	    !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
270 	    ipip_allow != 2) {
271 		for (ifp = ifnet.tqh_first; ifp != 0;
272 		     ifp = ifp->if_list.tqe_next) {
273 			for (ifa = ifp->if_addrlist.tqh_first; ifa != 0;
274 			     ifa = ifa->ifa_list.tqe_next) {
275 #ifdef INET
276 				if (ipo) {
277 					if (ifa->ifa_addr->sa_family !=
278 					    AF_INET)
279 						continue;
280 
281 					sin = (struct sockaddr_in *) ifa->ifa_addr;
282 
283 					if (sin->sin_addr.s_addr ==
284 					    ipo->ip_src.s_addr)	{
285 						ipipstat.ipips_spoof++;
286 						m_freem(m);
287 						return;
288 					}
289 				}
290 #endif /* INET */
291 
292 #ifdef INET6
293 				if (ip6) {
294 					if (ifa->ifa_addr->sa_family !=
295 					    AF_INET6)
296 						continue;
297 
298 					sin6 = (struct sockaddr_in6 *) ifa->ifa_addr;
299 
300 					if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_src)) {
301 						ipipstat.ipips_spoof++;
302 						m_freem(m);
303 						return;
304 					}
305 
306 				}
307 #endif /* INET6 */
308 			}
309 		}
310 	}
311 
312 	/* Statistics */
313 	ipipstat.ipips_ibytes += m->m_pkthdr.len - iphlen;
314 
315 	/*
316 	 * Interface pointer stays the same; if no IPsec processing has
317 	 * been done (or will be done), this will point to a normal
318 	 * interface. Otherwise, it'll point to an enc interface, which
319 	 * will allow a packet filter to distinguish between secure and
320 	 * untrusted packets.
321 	 */
322 
323 #ifdef INET
324 	if (ipo) {
325 		ifq = &ipintrq;
326 		isr = NETISR_IP;
327 	}
328 #endif /* INET */
329 
330 #ifdef INET6
331 	if (ip6) {
332 		ifq = &ip6intrq;
333 		isr = NETISR_IPV6;
334 	}
335 #endif /* INET6 */
336 
337 	s = splimp();			/* isn't it already? */
338 	if (IF_QFULL(ifq)) {
339 		IF_DROP(ifq);
340 		m_freem(m);
341 		ipipstat.ipips_qfull++;
342 
343 		splx(s);
344 
345 		DPRINTF(("ipip_input(): packet dropped because of full "
346 		    "queue\n"));
347 		return;
348 	}
349 
350 	IF_ENQUEUE(ifq, m);
351 	schednetisr(isr);
352 	splx(s);
353 	return;
354 }
355 
356 int
357 ipip_output(struct mbuf *m, struct tdb *tdb, struct mbuf **mp, int skip,
358     int protoff)
359 {
360 	u_int8_t tp, otos;
361 
362 #ifdef INET
363 	u_int8_t itos;
364 	struct ip *ipo;
365 #endif /* INET */
366 
367 #ifdef INET6
368 	struct ip6_hdr *ip6, *ip6o;
369 #endif /* INET6 */
370 
371 	/* XXX Deal with empty TDB source/destination addresses. */
372 
373 	m_copydata(m, 0, 1, &tp);
374 	tp = (tp >> 4) & 0xff;  /* Get the IP version number. */
375 
376 	switch (tdb->tdb_dst.sa.sa_family) {
377 #ifdef INET
378 	case AF_INET:
379 		if (tdb->tdb_src.sa.sa_family != AF_INET ||
380 		    tdb->tdb_src.sin.sin_addr.s_addr == INADDR_ANY ||
381 		    tdb->tdb_dst.sin.sin_addr.s_addr == INADDR_ANY) {
382 
383 			DPRINTF(("ipip_output(): unspecified tunnel endpoind "
384 			    "address in SA %s/%08x\n",
385 			    ipsp_address(tdb->tdb_dst), ntohl(tdb->tdb_spi)));
386 
387 			ipipstat.ipips_unspec++;
388 			m_freem(m);
389 			*mp = NULL;
390 			return EINVAL;
391 		}
392 
393 		M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
394 		if (m == 0) {
395 			DPRINTF(("ipip_output(): M_PREPEND failed\n"));
396 			ipipstat.ipips_hdrops++;
397 			*mp = NULL;
398 			return ENOBUFS;
399 		}
400 
401 		ipo = mtod(m, struct ip *);
402 
403 		ipo->ip_v = IPVERSION;
404 		ipo->ip_hl = 5;
405 		ipo->ip_len = htons(m->m_pkthdr.len);
406 		ipo->ip_ttl = ip_defttl;
407 		ipo->ip_sum = 0;
408 		ipo->ip_src = tdb->tdb_src.sin.sin_addr;
409 		ipo->ip_dst = tdb->tdb_dst.sin.sin_addr;
410 
411 		/*
412 		 * We do the htons() to prevent snoopers from determining our
413 		 * endianness.
414 		 */
415 		ipo->ip_id = htons(ip_randomid());
416 
417 		/* If the inner protocol is IP... */
418 		if (tp == IPVERSION) {
419 			/* Save ECN notification */
420 			m_copydata(m, sizeof(struct ip) +
421 			    offsetof(struct ip, ip_tos),
422 			    sizeof(u_int8_t), (caddr_t) &itos);
423 
424 			ipo->ip_p = IPPROTO_IPIP;
425 
426 			/*
427 			 * We should be keeping tunnel soft-state and
428 			 * send back ICMPs if needed.
429 			 */
430 			m_copydata(m, sizeof(struct ip) +
431 			    offsetof(struct ip, ip_off),
432 			    sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
433 			NTOHS(ipo->ip_off);
434 			ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
435 			HTONS(ipo->ip_off);
436 		}
437 #ifdef INET6
438 		else if (tp == (IPV6_VERSION >> 4)) {
439 			u_int32_t itos32;
440 
441 			/* Save ECN notification. */
442 			m_copydata(m, sizeof(struct ip) +
443 			    offsetof(struct ip6_hdr, ip6_flow),
444 			    sizeof(u_int32_t), (caddr_t) &itos32);
445 			itos = ntohl(itos32) >> 20;
446 			ipo->ip_p = IPPROTO_IPV6;
447 			ipo->ip_off = 0;
448 		}
449 #endif /* INET6 */
450 		else {
451 			m_freem(m);
452 			*mp = NULL;
453 			return EAFNOSUPPORT;
454 		}
455 
456 		otos = 0;
457 		ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
458 		ipo->ip_tos = otos;
459 		break;
460 #endif /* INET */
461 
462 #ifdef INET6
463 	case AF_INET6:
464 		if (IN6_IS_ADDR_UNSPECIFIED(&tdb->tdb_dst.sin6.sin6_addr) ||
465 		    tdb->tdb_src.sa.sa_family != AF_INET6 ||
466 		    IN6_IS_ADDR_UNSPECIFIED(&tdb->tdb_src.sin6.sin6_addr)) {
467 
468 			DPRINTF(("ipip_output(): unspecified tunnel endpoind "
469 			    "address in SA %s/%08x\n",
470 			    ipsp_address(tdb->tdb_dst), ntohl(tdb->tdb_spi)));
471 
472 			ipipstat.ipips_unspec++;
473 			m_freem(m);
474 			*mp = NULL;
475 			return ENOBUFS;
476 		}
477 
478 		/* scoped address handling */
479 		ip6 = mtod(m, struct ip6_hdr *);
480 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
481 			ip6->ip6_src.s6_addr16[1] = 0;
482 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
483 			ip6->ip6_dst.s6_addr16[1] = 0;
484 
485 		M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
486 		if (m == 0) {
487 			DPRINTF(("ipip_output(): M_PREPEND failed\n"));
488 			ipipstat.ipips_hdrops++;
489 			*mp = NULL;
490 			return ENOBUFS;
491 		}
492 
493 		/* Initialize IPv6 header */
494 		ip6o = mtod(m, struct ip6_hdr *);
495 		ip6o->ip6_flow = 0;
496 		ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
497 		ip6o->ip6_vfc |= IPV6_VERSION;
498 		ip6o->ip6_plen = htons(m->m_pkthdr.len);
499 		ip6o->ip6_hlim = ip_defttl;
500 		ip6o->ip6_dst = tdb->tdb_dst.sin6.sin6_addr;
501 		ip6o->ip6_src = tdb->tdb_src.sin6.sin6_addr;
502 
503 #ifdef INET
504 		if (tp == IPVERSION) {
505 			/* Save ECN notification */
506 			m_copydata(m, sizeof(struct ip6_hdr) +
507 			    offsetof(struct ip, ip_tos), sizeof(u_int8_t),
508 			    (caddr_t) &itos);
509 
510 			/* This is really IPVERSION. */
511 			ip6o->ip6_nxt = IPPROTO_IPIP;
512 		}
513 		else
514 #endif /* INET */
515 			if (tp == (IPV6_VERSION >> 4)) {
516 				u_int32_t itos32;
517 
518 				/* Save ECN notification. */
519 				m_copydata(m, sizeof(struct ip6_hdr) +
520 				    offsetof(struct ip6_hdr, ip6_flow),
521 				    sizeof(u_int32_t), (caddr_t) &itos32);
522 				itos = ntohl(itos32) >> 20;
523 
524 				ip6o->ip6_nxt = IPPROTO_IPV6;
525 			} else {
526 				m_freem(m);
527 				*mp = NULL;
528 				return EAFNOSUPPORT;
529 			}
530 
531 		otos = 0;
532 		ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
533 		ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
534 		break;
535 #endif /* INET6 */
536 
537 	default:
538 		DPRINTF(("ipip_output(): unsupported protocol family %d\n",
539 		    tdb->tdb_dst.sa.sa_family));
540 		m_freem(m);
541 		*mp = NULL;
542 		ipipstat.ipips_family++;
543 		return ENOBUFS;
544 	}
545 
546 	ipipstat.ipips_opackets++;
547 	*mp = m;
548 
549 #ifdef INET
550 	if (tdb->tdb_dst.sa.sa_family == AF_INET) {
551 		if (tdb->tdb_xform->xf_type == XF_IP4)
552 			tdb->tdb_cur_bytes +=
553 			    m->m_pkthdr.len - sizeof(struct ip);
554 
555 		ipipstat.ipips_obytes += m->m_pkthdr.len - sizeof(struct ip);
556 	}
557 #endif /* INET */
558 
559 #ifdef INET6
560 	if (tdb->tdb_dst.sa.sa_family == AF_INET6) {
561 		if (tdb->tdb_xform->xf_type == XF_IP4)
562 			tdb->tdb_cur_bytes +=
563 			    m->m_pkthdr.len - sizeof(struct ip6_hdr);
564 
565 		ipipstat.ipips_obytes +=
566 		    m->m_pkthdr.len - sizeof(struct ip6_hdr);
567 	}
568 #endif /* INET6 */
569 
570 	return 0;
571 }
572 
573 #ifdef IPSEC
574 int
575 ipe4_attach()
576 {
577 	return 0;
578 }
579 
580 int
581 ipe4_init(struct tdb *tdbp, struct xformsw *xsp, struct ipsecinit *ii)
582 {
583 	tdbp->tdb_xform = xsp;
584 	return 0;
585 }
586 
587 int
588 ipe4_zeroize(struct tdb *tdbp)
589 {
590 	return 0;
591 }
592 
593 void
594 ipe4_input(struct mbuf *m, ...)
595 {
596 	/* This is a rather serious mistake, so no conditional printing. */
597 	printf("ipe4_input(): should never be called\n");
598 	if (m)
599 		m_freem(m);
600 }
601 #endif	/* IPSEC */
602 
603 int
604 ipip_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
605     size_t newlen)
606 {
607 	/* All sysctl names at this level are terminal. */
608 	if (namelen != 1)
609 		return (ENOTDIR);
610 
611 	switch (name[0]) {
612 	case IPIPCTL_ALLOW:
613 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ipip_allow));
614 	default:
615 		return (ENOPROTOOPT);
616 	}
617 	/* NOTREACHED */
618 }
619