xref: /openbsd-src/sys/netinet/ip_ipip.c (revision 2badd5e3f47d2d4252969cd98d7042b4e701b5ac)
1 /*	$OpenBSD: ip_ipip.c,v 1.21 2001/08/19 06:31:56 angelos 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 #include <net/bpf.h>
52 
53 #include <netinet/in.h>
54 #include <netinet/in_systm.h>
55 #include <netinet/ip.h>
56 #include <netinet/in_pcb.h>
57 #include <netinet/in_var.h>
58 #include <netinet/ip_var.h>
59 #include <netinet/ip_ecn.h>
60 
61 #ifdef MROUTING
62 #include <netinet/ip_mroute.h>
63 #endif
64 
65 #include <netinet/ip_ipsp.h>
66 #include <netinet/ip_ipip.h>
67 
68 #include "bpfilter.h"
69 
70 #ifdef ENCDEBUG
71 #define DPRINTF(x)	if (encdebug) printf x
72 #else
73 #define DPRINTF(x)
74 #endif
75 
76 /*
77  * We can control the acceptance of IP4 packets by altering the sysctl
78  * net.inet.ipip.allow value.  Zero means drop them, all else is acceptance.
79  */
80 int ipip_allow = 0;
81 
82 struct ipipstat ipipstat;
83 
84 #ifdef INET6
85 /*
86  * Really only a wrapper for ipip_input(), for use with IPv6.
87  */
88 int
89 ip4_input6(struct mbuf **m, int *offp, int proto)
90 {
91 	/* If we do not accept IPv4 explicitly, drop.  */
92 	if (!ipip_allow && ((*m)->m_flags & (M_AUTH|M_CONF)) == 0) {
93 		DPRINTF(("ip4_input6(): dropped due to policy\n"));
94 		ipipstat.ipips_pdrops++;
95 		m_freem(*m);
96 		return IPPROTO_DONE;
97 	}
98 
99 	ipip_input(*m, *offp, NULL);
100 	return IPPROTO_DONE;
101 }
102 #endif /* INET6 */
103 
104 #ifdef INET
105 /*
106  * Really only a wrapper for ipip_input(), for use with IPv4.
107  */
108 void
109 ip4_input(struct mbuf *m, ...)
110 {
111 	va_list ap;
112 	int iphlen;
113 
114 	/* If we do not accept IPv4 explicitly, drop.  */
115 	if (!ipip_allow && (m->m_flags & (M_AUTH|M_CONF)) == 0) {
116 		DPRINTF(("ip4_input(): dropped due to policy\n"));
117 		ipipstat.ipips_pdrops++;
118 		m_freem(m);
119 		return;
120 	}
121 
122 	va_start(ap, m);
123 	iphlen = va_arg(ap, int);
124 	va_end(ap);
125 
126 	ipip_input(m, iphlen, NULL);
127 }
128 #endif /* INET */
129 
130 /*
131  * ipip_input gets called when we receive an IP{46} encapsulated packet,
132  * either because we got it at a real interface, or because AH or ESP
133  * were being used in tunnel mode (in which case the rcvif element will
134  * contain the address of the encX interface associated with the tunnel.
135  */
136 
137 void
138 ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp)
139 {
140 	register struct sockaddr_in *sin;
141 	register struct ifnet *ifp;
142 	register struct ifaddr *ifa;
143 	struct ifqueue *ifq = NULL;
144 	struct ip *ipo;
145 #ifdef INET6
146 	register struct sockaddr_in6 *sin6;
147 	struct ip6_hdr *ip6 = NULL;
148 	u_int8_t itos;
149 #endif
150 	u_int8_t nxt;
151 	int isr;
152 	u_int8_t otos;
153 	u_int8_t v;
154 	int hlen, s;
155 
156 	ipipstat.ipips_ipackets++;
157 
158 	m_copydata(m, 0, 1, &v);
159 
160 	switch (v >> 4) {
161 #ifdef INET
162         case 4:
163 		hlen = sizeof(struct ip);
164 		break;
165 #endif /* INET */
166 
167 #ifdef INET6
168         case 6:
169 		hlen = sizeof(struct ip6_hdr);
170 		break;
171 #endif
172         default:
173 		m_freem(m);
174 		return /* EAFNOSUPPORT */;
175 	}
176 
177 	/* Bring the IP header in the first mbuf, if not there already */
178 	if (m->m_len < hlen) {
179 		if ((m = m_pullup(m, hlen)) == NULL) {
180 			DPRINTF(("ipip_input(): m_pullup() failed\n"));
181 			ipipstat.ipips_hdrops++;
182 			return;
183 		}
184 	}
185 
186 	ipo = mtod(m, struct ip *);
187 
188 #ifdef MROUTING
189 	if (ipo->ip_v == IPVERSION && ipo->ip_p == IPPROTO_IPV4) {
190 		if (IN_MULTICAST(((struct ip *)((char *) ipo + iphlen))->ip_dst.s_addr)) {
191 			ipip_mroute_input (m, iphlen);
192 			return;
193 		}
194 	}
195 #endif /* MROUTING */
196 
197 	/* Keep outer ecn field. */
198 #ifdef INET
199 	if ((v >> 4) == 4)
200 		otos = ipo->ip_tos;
201 #endif /* INET */
202 
203 #ifdef INET6
204 	if ((v >> 4) == 6)
205 		otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
206 #endif
207 
208 	/* Remove outer IP header */
209 	m_adj(m, iphlen);
210 
211 	m_copydata(m, 0, 1, &v);
212 
213 	switch (v >> 4) {
214 #ifdef INET
215         case 4:
216 		hlen = sizeof(struct ip);
217 		break;
218 #endif /* INET */
219 
220 #ifdef INET6
221         case 6:
222 		hlen = sizeof(struct ip6_hdr);
223 		break;
224 #endif
225 	}
226 
227 	/*
228 	 * Bring the inner IP header in the first mbuf, if not there already.
229 	 */
230 	if (m->m_len < hlen) {
231 		if ((m = m_pullup(m, hlen)) == NULL) {
232 			DPRINTF(("ipip_input(): m_pullup() failed\n"));
233 			ipipstat.ipips_hdrops++;
234 			return;
235 		}
236 	}
237 
238 	/*
239 	 * RFC 1853 specifies that the inner TTL should not be touched on
240 	 * decapsulation. There's no reason this comment should be here, but
241 	 * this is as good as any a position.
242 	 */
243 
244 	/* Some sanity checks in the inner IPv4 header */
245 	switch (v >> 4) {
246 #ifdef INET
247     	case 4:
248                 ipo = mtod(m, struct ip *);
249                 nxt = ipo->ip_p;
250 		ip_ecn_egress(ECN_ALLOWED, &otos, &ipo->ip_tos);
251                 break;
252 #endif /* INET */
253 
254 #ifdef INET6
255     	case 6:
256                 ip6 = (struct ip6_hdr *) ipo;
257                 nxt = ip6->ip6_nxt;
258 		itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
259 		ip_ecn_egress(ECN_ALLOWED, &otos, &itos);
260 		ip6->ip6_flow &= ~htonl(0xff << 20);
261 		ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
262                 break;
263 #endif
264 	}
265 
266 	/* Check for local address spoofing. */
267 	if ((m->m_pkthdr.rcvif == NULL ||
268 	    !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
269 	    ipip_allow != 2) {
270 		for (ifp = ifnet.tqh_first; ifp != 0;
271 		     ifp = ifp->if_list.tqe_next) {
272 			for (ifa = ifp->if_addrlist.tqh_first; ifa != 0;
273 			     ifa = ifa->ifa_list.tqe_next) {
274 #ifdef INET
275 				if (ipo) {
276 					if (ifa->ifa_addr->sa_family !=
277 					    AF_INET)
278 						continue;
279 
280 					sin = (struct sockaddr_in *) ifa->ifa_addr;
281 
282 					if (sin->sin_addr.s_addr ==
283 					    ipo->ip_src.s_addr)	{
284 						ipipstat.ipips_spoof++;
285 						m_freem(m);
286 						return;
287 					}
288 				}
289 #endif /* INET */
290 
291 #ifdef INET6
292 				if (ip6) {
293 					if (ifa->ifa_addr->sa_family !=
294 					    AF_INET6)
295 						continue;
296 
297 					sin6 = (struct sockaddr_in6 *) ifa->ifa_addr;
298 
299 					if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_src)) {
300 						ipipstat.ipips_spoof++;
301 						m_freem(m);
302 						return;
303 					}
304 
305 				}
306 #endif /* INET6 */
307 			}
308 		}
309 	}
310 
311 	/* Statistics */
312 	ipipstat.ipips_ibytes += m->m_pkthdr.len - iphlen;
313 
314 	/*
315 	 * Interface pointer stays the same; if no IPsec processing has
316 	 * been done (or will be done), this will point to a normal
317 	 * interface. Otherwise, it'll point to an enc interface, which
318 	 * will allow a packet filter to distinguish between secure and
319 	 * untrusted packets.
320 	 */
321 
322 #ifdef INET
323 	if (ipo) {
324 		ifq = &ipintrq;
325 		isr = NETISR_IP;
326 	}
327 #endif /* INET */
328 
329 #ifdef INET6
330 	if (ip6) {
331 		ifq = &ip6intrq;
332 		isr = NETISR_IPV6;
333 	}
334 #endif /* INET6 */
335 
336 #if NBPFILTER > 0
337 	if (gifp && gifp->if_bpf) {
338 		struct mbuf m0;
339 		u_int af;
340 
341 		if (ipo)
342 			af = AF_INET;
343 		else
344 			af = AF_INET6;
345 
346 		m0.m_next = m;
347 		m0.m_len = 4;
348 		m0.m_data = (char *)&af;
349 
350 		bpf_mtap(gifp->if_bpf, &m0);
351 	}
352 #endif
353 
354 	s = splimp();			/* isn't it already? */
355 	if (IF_QFULL(ifq)) {
356 		IF_DROP(ifq);
357 		m_freem(m);
358 		ipipstat.ipips_qfull++;
359 
360 		splx(s);
361 
362 		DPRINTF(("ipip_input(): packet dropped because of full "
363 		    "queue\n"));
364 		return;
365 	}
366 
367 	IF_ENQUEUE(ifq, m);
368 	schednetisr(isr);
369 	splx(s);
370 	return;
371 }
372 
373 int
374 ipip_output(struct mbuf *m, struct tdb *tdb, struct mbuf **mp, int skip,
375     int protoff)
376 {
377 	u_int8_t tp, otos;
378 
379 #ifdef INET
380 	u_int8_t itos;
381 	struct ip *ipo;
382 #endif /* INET */
383 
384 #ifdef INET6
385 	struct ip6_hdr *ip6, *ip6o;
386 #endif /* INET6 */
387 
388 	/* XXX Deal with empty TDB source/destination addresses. */
389 
390 	m_copydata(m, 0, 1, &tp);
391 	tp = (tp >> 4) & 0xff;  /* Get the IP version number. */
392 
393 	switch (tdb->tdb_dst.sa.sa_family) {
394 #ifdef INET
395 	case AF_INET:
396 		if (tdb->tdb_src.sa.sa_family != AF_INET ||
397 		    tdb->tdb_src.sin.sin_addr.s_addr == INADDR_ANY ||
398 		    tdb->tdb_dst.sin.sin_addr.s_addr == INADDR_ANY) {
399 
400 			DPRINTF(("ipip_output(): unspecified tunnel endpoind "
401 			    "address in SA %s/%08x\n",
402 			    ipsp_address(tdb->tdb_dst), ntohl(tdb->tdb_spi)));
403 
404 			ipipstat.ipips_unspec++;
405 			m_freem(m);
406 			*mp = NULL;
407 			return EINVAL;
408 		}
409 
410 		M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
411 		if (m == 0) {
412 			DPRINTF(("ipip_output(): M_PREPEND failed\n"));
413 			ipipstat.ipips_hdrops++;
414 			*mp = NULL;
415 			return ENOBUFS;
416 		}
417 
418 		ipo = mtod(m, struct ip *);
419 
420 		ipo->ip_v = IPVERSION;
421 		ipo->ip_hl = 5;
422 		ipo->ip_len = htons(m->m_pkthdr.len);
423 		ipo->ip_ttl = ip_defttl;
424 		ipo->ip_sum = 0;
425 		ipo->ip_src = tdb->tdb_src.sin.sin_addr;
426 		ipo->ip_dst = tdb->tdb_dst.sin.sin_addr;
427 
428 		/*
429 		 * We do the htons() to prevent snoopers from determining our
430 		 * endianness.
431 		 */
432 		ipo->ip_id = htons(ip_randomid());
433 
434 		/* If the inner protocol is IP... */
435 		if (tp == IPVERSION) {
436 			/* Save ECN notification */
437 			m_copydata(m, sizeof(struct ip) +
438 			    offsetof(struct ip, ip_tos),
439 			    sizeof(u_int8_t), (caddr_t) &itos);
440 
441 			ipo->ip_p = IPPROTO_IPIP;
442 
443 			/*
444 			 * We should be keeping tunnel soft-state and
445 			 * send back ICMPs if needed.
446 			 */
447 			m_copydata(m, sizeof(struct ip) +
448 			    offsetof(struct ip, ip_off),
449 			    sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
450 			NTOHS(ipo->ip_off);
451 			ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
452 			HTONS(ipo->ip_off);
453 		}
454 #ifdef INET6
455 		else if (tp == (IPV6_VERSION >> 4)) {
456 			u_int32_t itos32;
457 
458 			/* Save ECN notification. */
459 			m_copydata(m, sizeof(struct ip) +
460 			    offsetof(struct ip6_hdr, ip6_flow),
461 			    sizeof(u_int32_t), (caddr_t) &itos32);
462 			itos = ntohl(itos32) >> 20;
463 			ipo->ip_p = IPPROTO_IPV6;
464 			ipo->ip_off = 0;
465 		}
466 #endif /* INET6 */
467 		else {
468 			m_freem(m);
469 			*mp = NULL;
470 			return EAFNOSUPPORT;
471 		}
472 
473 		otos = 0;
474 		ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
475 		ipo->ip_tos = otos;
476 		break;
477 #endif /* INET */
478 
479 #ifdef INET6
480 	case AF_INET6:
481 		if (IN6_IS_ADDR_UNSPECIFIED(&tdb->tdb_dst.sin6.sin6_addr) ||
482 		    tdb->tdb_src.sa.sa_family != AF_INET6 ||
483 		    IN6_IS_ADDR_UNSPECIFIED(&tdb->tdb_src.sin6.sin6_addr)) {
484 
485 			DPRINTF(("ipip_output(): unspecified tunnel endpoind "
486 			    "address in SA %s/%08x\n",
487 			    ipsp_address(tdb->tdb_dst), ntohl(tdb->tdb_spi)));
488 
489 			ipipstat.ipips_unspec++;
490 			m_freem(m);
491 			*mp = NULL;
492 			return ENOBUFS;
493 		}
494 
495 		/* scoped address handling */
496 		ip6 = mtod(m, struct ip6_hdr *);
497 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
498 			ip6->ip6_src.s6_addr16[1] = 0;
499 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
500 			ip6->ip6_dst.s6_addr16[1] = 0;
501 
502 		M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
503 		if (m == 0) {
504 			DPRINTF(("ipip_output(): M_PREPEND failed\n"));
505 			ipipstat.ipips_hdrops++;
506 			*mp = NULL;
507 			return ENOBUFS;
508 		}
509 
510 		/* Initialize IPv6 header */
511 		ip6o = mtod(m, struct ip6_hdr *);
512 		ip6o->ip6_flow = 0;
513 		ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
514 		ip6o->ip6_vfc |= IPV6_VERSION;
515 		ip6o->ip6_plen = htons(m->m_pkthdr.len);
516 		ip6o->ip6_hlim = ip_defttl;
517 		ip6o->ip6_dst = tdb->tdb_dst.sin6.sin6_addr;
518 		ip6o->ip6_src = tdb->tdb_src.sin6.sin6_addr;
519 
520 #ifdef INET
521 		if (tp == IPVERSION) {
522 			/* Save ECN notification */
523 			m_copydata(m, sizeof(struct ip6_hdr) +
524 			    offsetof(struct ip, ip_tos), sizeof(u_int8_t),
525 			    (caddr_t) &itos);
526 
527 			/* This is really IPVERSION. */
528 			ip6o->ip6_nxt = IPPROTO_IPIP;
529 		}
530 		else
531 #endif /* INET */
532 			if (tp == (IPV6_VERSION >> 4)) {
533 				u_int32_t itos32;
534 
535 				/* Save ECN notification. */
536 				m_copydata(m, sizeof(struct ip6_hdr) +
537 				    offsetof(struct ip6_hdr, ip6_flow),
538 				    sizeof(u_int32_t), (caddr_t) &itos32);
539 				itos = ntohl(itos32) >> 20;
540 
541 				ip6o->ip6_nxt = IPPROTO_IPV6;
542 			} else {
543 				m_freem(m);
544 				*mp = NULL;
545 				return EAFNOSUPPORT;
546 			}
547 
548 		otos = 0;
549 		ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
550 		ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
551 		break;
552 #endif /* INET6 */
553 
554 	default:
555 		DPRINTF(("ipip_output(): unsupported protocol family %d\n",
556 		    tdb->tdb_dst.sa.sa_family));
557 		m_freem(m);
558 		*mp = NULL;
559 		ipipstat.ipips_family++;
560 		return ENOBUFS;
561 	}
562 
563 	ipipstat.ipips_opackets++;
564 	*mp = m;
565 
566 #ifdef INET
567 	if (tdb->tdb_dst.sa.sa_family == AF_INET) {
568 		if (tdb->tdb_xform->xf_type == XF_IP4)
569 			tdb->tdb_cur_bytes +=
570 			    m->m_pkthdr.len - sizeof(struct ip);
571 
572 		ipipstat.ipips_obytes += m->m_pkthdr.len - sizeof(struct ip);
573 	}
574 #endif /* INET */
575 
576 #ifdef INET6
577 	if (tdb->tdb_dst.sa.sa_family == AF_INET6) {
578 		if (tdb->tdb_xform->xf_type == XF_IP4)
579 			tdb->tdb_cur_bytes +=
580 			    m->m_pkthdr.len - sizeof(struct ip6_hdr);
581 
582 		ipipstat.ipips_obytes +=
583 		    m->m_pkthdr.len - sizeof(struct ip6_hdr);
584 	}
585 #endif /* INET6 */
586 
587 	return 0;
588 }
589 
590 #ifdef IPSEC
591 int
592 ipe4_attach()
593 {
594 	return 0;
595 }
596 
597 int
598 ipe4_init(struct tdb *tdbp, struct xformsw *xsp, struct ipsecinit *ii)
599 {
600 	tdbp->tdb_xform = xsp;
601 	return 0;
602 }
603 
604 int
605 ipe4_zeroize(struct tdb *tdbp)
606 {
607 	return 0;
608 }
609 
610 void
611 ipe4_input(struct mbuf *m, ...)
612 {
613 	/* This is a rather serious mistake, so no conditional printing. */
614 	printf("ipe4_input(): should never be called\n");
615 	if (m)
616 		m_freem(m);
617 }
618 #endif	/* IPSEC */
619 
620 int
621 ipip_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
622     size_t newlen)
623 {
624 	/* All sysctl names at this level are terminal. */
625 	if (namelen != 1)
626 		return (ENOTDIR);
627 
628 	switch (name[0]) {
629 	case IPIPCTL_ALLOW:
630 		return (sysctl_int(oldp, oldlenp, newp, newlen, &ipip_allow));
631 	default:
632 		return (ENOPROTOOPT);
633 	}
634 	/* NOTREACHED */
635 }
636