xref: /openbsd/sys/netinet/ip_ipip.c (revision 5dea098c)
1 /*	$OpenBSD: ip_ipip.c,v 1.101 2024/02/11 01:27:45 bluhm 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 "bpfilter.h"
43 #include "gif.h"
44 #include "pf.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/sysctl.h>
51 
52 #include <net/if.h>
53 #include <net/if_types.h>
54 #include <net/if_var.h>
55 #include <net/route.h>
56 #include <net/netisr.h>
57 #include <net/bpf.h>
58 
59 #include <netinet/in.h>
60 #include <netinet/ip.h>
61 #include <netinet/in_pcb.h>
62 #include <netinet/ip_var.h>
63 #include <netinet6/ip6_var.h>
64 #include <netinet/ip_ecn.h>
65 #include <netinet/ip_ipip.h>
66 
67 #ifdef MROUTING
68 #include <netinet/ip_mroute.h>
69 #endif
70 
71 #if NPF > 0
72 #include <net/pfvar.h>
73 #endif
74 
75 #ifdef ENCDEBUG
76 #define DPRINTF(fmt, args...)						\
77 	do {								\
78 		if (encdebug)						\
79 			printf("%s: " fmt "\n", __func__, ## args);	\
80 	} while (0)
81 #else
82 #define DPRINTF(fmt, args...)						\
83 	do { } while (0)
84 #endif
85 
86 /*
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.
89  */
90 int ipip_allow = 0;
91 
92 struct cpumem *ipipcounters;
93 
94 void
95 ipip_init(void)
96 {
97 	ipipcounters = counters_alloc(ipips_ncounters);
98 }
99 
100 /*
101  * Really only a wrapper for ipip_input_if(), for use with pr_input.
102  */
103 int
104 ipip_input(struct mbuf **mp, int *offp, int nxt, int af)
105 {
106 	struct ifnet *ifp;
107 
108 	/* If we do not accept IP-in-IP explicitly, drop.  */
109 	if (!ipip_allow && ((*mp)->m_flags & (M_AUTH|M_CONF)) == 0) {
110 		DPRINTF("dropped due to policy");
111 		ipipstat_inc(ipips_pdrops);
112 		m_freemp(mp);
113 		return IPPROTO_DONE;
114 	}
115 
116 	ifp = if_get((*mp)->m_pkthdr.ph_ifidx);
117 	if (ifp == NULL) {
118 		m_freemp(mp);
119 		return IPPROTO_DONE;
120 	}
121 	nxt = ipip_input_if(mp, offp, nxt, af, ifp);
122 	if_put(ifp);
123 
124 	return nxt;
125 }
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 ph_ifidx element
131  * will contain the index of the encX interface associated with the
132  * tunnel.
133  */
134 
135 int
136 ipip_input_if(struct mbuf **mp, int *offp, int proto, int oaf,
137     struct ifnet *ifp)
138 {
139 	struct mbuf *m = *mp;
140 	struct sockaddr_in *sin;
141 	struct ip *ip;
142 #ifdef INET6
143 	struct sockaddr_in6 *sin6;
144 	struct ip6_hdr *ip6;
145 #endif
146 	int mode, hlen;
147 	u_int8_t itos, otos;
148 	sa_family_t iaf;
149 
150 	ipipstat_inc(ipips_ipackets);
151 
152 	switch (oaf) {
153 	case AF_INET:
154 		hlen = sizeof(struct ip);
155 		break;
156 #ifdef INET6
157 	case AF_INET6:
158 		hlen = sizeof(struct ip6_hdr);
159 		break;
160 #endif
161 	default:
162 		unhandled_af(oaf);
163 	}
164 
165 	/* Bring the IP header in the first mbuf, if not there already */
166 	if (m->m_len < hlen) {
167 		if ((m = *mp = m_pullup(m, hlen)) == NULL) {
168 			DPRINTF("m_pullup() failed");
169 			ipipstat_inc(ipips_hdrops);
170 			goto bad;
171 		}
172 	}
173 
174 	/* Keep outer ecn field. */
175 	switch (oaf) {
176 	case AF_INET:
177 		ip = mtod(m, struct ip *);
178 		otos = ip->ip_tos;
179 		break;
180 #ifdef INET6
181 	case AF_INET6:
182 		ip6 = mtod(m, struct ip6_hdr *);
183 		otos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
184 		break;
185 #endif
186 	}
187 
188 	/* Remove outer IP header */
189 	KASSERT(*offp > 0);
190 	m_adj(m, *offp);
191 	*offp = 0;
192 	ip = NULL;
193 #ifdef INET6
194 	ip6 = NULL;
195 #endif
196 
197 	switch (proto) {
198 	case IPPROTO_IPV4:
199 		hlen = sizeof(struct ip);
200 		break;
201 
202 #ifdef INET6
203 	case IPPROTO_IPV6:
204 		hlen = sizeof(struct ip6_hdr);
205 		break;
206 #endif
207 	default:
208 		ipipstat_inc(ipips_family);
209 		goto bad;
210 	}
211 
212 	/* Sanity check */
213 	if (m->m_pkthdr.len < hlen) {
214 		ipipstat_inc(ipips_hdrops);
215 		goto bad;
216 	}
217 
218 	/*
219 	 * Bring the inner header into the first mbuf, if not there already.
220 	 */
221 	if (m->m_len < hlen) {
222 		if ((m = *mp = m_pullup(m, hlen)) == NULL) {
223 			DPRINTF("m_pullup() failed");
224 			ipipstat_inc(ipips_hdrops);
225 			goto bad;
226 		}
227 	}
228 
229 	/*
230 	 * RFC 1853 specifies that the inner TTL should not be touched on
231 	 * decapsulation. There's no reason this comment should be here, but
232 	 * this is as good as any a position.
233 	 */
234 
235 	/* Some sanity checks in the inner IP header */
236 	switch (proto) {
237 	case IPPROTO_IPV4:
238 		iaf = AF_INET;
239 		ip = mtod(m, struct ip *);
240 		hlen = ip->ip_hl << 2;
241 		if (m->m_pkthdr.len < hlen) {
242 			ipipstat_inc(ipips_hdrops);
243 			goto bad;
244 		}
245 		itos = ip->ip_tos;
246 		mode = m->m_flags & (M_AUTH|M_CONF) ?
247 		    ECN_ALLOWED_IPSEC : ECN_ALLOWED;
248 		if (!ip_ecn_egress(mode, &otos, &itos)) {
249 			DPRINTF("ip_ecn_egress() failed");
250 			ipipstat_inc(ipips_pdrops);
251 			goto bad;
252 		}
253 		/* re-calculate the checksum if ip_tos was changed */
254 		if (itos != ip->ip_tos)
255 			ip_tos_patch(ip, itos);
256 		break;
257 #ifdef INET6
258 	case IPPROTO_IPV6:
259 		iaf = AF_INET6;
260 		ip6 = mtod(m, struct ip6_hdr *);
261 		itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
262 		if (!ip_ecn_egress(ECN_ALLOWED, &otos, &itos)) {
263 			DPRINTF("ip_ecn_egress() failed");
264 			ipipstat_inc(ipips_pdrops);
265 			goto bad;
266 		}
267 		ip6->ip6_flow &= ~htonl(0xff << 20);
268 		ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
269 		break;
270 #endif
271 	}
272 
273 	/* Check for local address spoofing. */
274 	if (!(ifp->if_flags & IFF_LOOPBACK) && ipip_allow != 2) {
275 		struct sockaddr_storage ss;
276 		struct rtentry *rt;
277 
278 		memset(&ss, 0, sizeof(ss));
279 
280 		if (ip) {
281 			sin = (struct sockaddr_in *)&ss;
282 			sin->sin_family = AF_INET;
283 			sin->sin_len = sizeof(*sin);
284 			sin->sin_addr = ip->ip_src;
285 #ifdef INET6
286 		} else if (ip6) {
287 			sin6 = (struct sockaddr_in6 *)&ss;
288 			sin6->sin6_family = AF_INET6;
289 			sin6->sin6_len = sizeof(*sin6);
290 			sin6->sin6_addr = ip6->ip6_src;
291 #endif /* INET6 */
292 		}
293 		rt = rtalloc(sstosa(&ss), 0, m->m_pkthdr.ph_rtableid);
294 		if ((rt != NULL) && (rt->rt_flags & RTF_LOCAL)) {
295 			ipipstat_inc(ipips_spoof);
296 			rtfree(rt);
297 			goto bad;
298 		}
299 		rtfree(rt);
300 	}
301 
302 	/* Statistics */
303 	ipipstat_add(ipips_ibytes, m->m_pkthdr.len - hlen);
304 
305 #if NBPFILTER > 0 && NGIF > 0
306 	if (ifp->if_type == IFT_GIF && ifp->if_bpf != NULL)
307 		bpf_mtap_af(ifp->if_bpf, iaf, m, BPF_DIRECTION_IN);
308 #endif
309 #if NPF > 0
310 	pf_pkt_addr_changed(m);
311 #endif
312 
313 	/*
314 	 * Interface pointer stays the same; if no IPsec processing has
315 	 * been done (or will be done), this will point to a normal
316 	 * interface. Otherwise, it'll point to an enc interface, which
317 	 * will allow a packet filter to distinguish between secure and
318 	 * untrusted packets.
319 	 */
320 
321 	switch (proto) {
322 	case IPPROTO_IPV4:
323 		return ip_input_if(mp, offp, proto, oaf, ifp);
324 #ifdef INET6
325 	case IPPROTO_IPV6:
326 		return ip6_input_if(mp, offp, proto, oaf, ifp);
327 #endif
328 	}
329  bad:
330 	m_freemp(mp);
331 	return IPPROTO_DONE;
332 }
333 
334 int
335 ipip_output(struct mbuf **mp, struct tdb *tdb)
336 {
337 	struct mbuf *m = *mp;
338 	u_int8_t tp, otos, itos;
339 	u_int64_t obytes;
340 	struct ip *ipo;
341 #ifdef INET6
342 	struct ip6_hdr *ip6, *ip6o;
343 #endif /* INET6 */
344 #ifdef ENCDEBUG
345 	char buf[INET6_ADDRSTRLEN];
346 #endif
347 	int error;
348 
349 	/* XXX Deal with empty TDB source/destination addresses. */
350 
351 	m_copydata(m, 0, 1, &tp);
352 	tp = (tp >> 4) & 0xff;  /* Get the IP version number. */
353 
354 	switch (tdb->tdb_dst.sa.sa_family) {
355 	case AF_INET:
356 		if (tdb->tdb_src.sa.sa_family != AF_INET ||
357 		    tdb->tdb_src.sin.sin_addr.s_addr == INADDR_ANY ||
358 		    tdb->tdb_dst.sin.sin_addr.s_addr == INADDR_ANY) {
359 
360 			DPRINTF("unspecified tunnel endpoint address "
361 			    "in SA %s/%08x",
362 			    ipsp_address(&tdb->tdb_dst, buf, sizeof(buf)),
363 			    ntohl(tdb->tdb_spi));
364 
365 			ipipstat_inc(ipips_unspec);
366 			error = EINVAL;
367 			goto drop;
368 		}
369 
370 		M_PREPEND(*mp, sizeof(struct ip), M_DONTWAIT);
371 		if (*mp == NULL) {
372 			DPRINTF("M_PREPEND failed");
373 			ipipstat_inc(ipips_hdrops);
374 			error = ENOBUFS;
375 			goto drop;
376 		}
377 		m = *mp;
378 
379 		ipo = mtod(m, struct ip *);
380 
381 		ipo->ip_v = IPVERSION;
382 		ipo->ip_hl = 5;
383 		ipo->ip_len = htons(m->m_pkthdr.len);
384 		ipo->ip_ttl = ip_defttl;
385 		ipo->ip_sum = 0;
386 		ipo->ip_src = tdb->tdb_src.sin.sin_addr;
387 		ipo->ip_dst = tdb->tdb_dst.sin.sin_addr;
388 
389 		/*
390 		 * We do the htons() to prevent snoopers from determining our
391 		 * endianness.
392 		 */
393 		ipo->ip_id = htons(ip_randomid());
394 
395 		/* If the inner protocol is IP... */
396 		if (tp == IPVERSION) {
397 			/* Save ECN notification */
398 			m_copydata(m, sizeof(struct ip) +
399 			    offsetof(struct ip, ip_tos),
400 			    sizeof(u_int8_t), (caddr_t) &itos);
401 
402 			ipo->ip_p = IPPROTO_IPIP;
403 
404 			/*
405 			 * We should be keeping tunnel soft-state and
406 			 * send back ICMPs if needed.
407 			 */
408 			m_copydata(m, sizeof(struct ip) +
409 			    offsetof(struct ip, ip_off),
410 			    sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
411 			ipo->ip_off = ntohs(ipo->ip_off);
412 			ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
413 			ipo->ip_off = htons(ipo->ip_off);
414 		}
415 #ifdef INET6
416 		else if (tp == (IPV6_VERSION >> 4)) {
417 			u_int32_t itos32;
418 
419 			/* Save ECN notification. */
420 			m_copydata(m, sizeof(struct ip) +
421 			    offsetof(struct ip6_hdr, ip6_flow),
422 			    sizeof(u_int32_t), (caddr_t) &itos32);
423 			itos = ntohl(itos32) >> 20;
424 			ipo->ip_p = IPPROTO_IPV6;
425 			ipo->ip_off = 0;
426 		}
427 #endif /* INET6 */
428 		else {
429 			ipipstat_inc(ipips_family);
430 			error = EAFNOSUPPORT;
431 			goto drop;
432 		}
433 
434 		otos = 0;
435 		ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
436 		ipo->ip_tos = otos;
437 
438 		obytes = m->m_pkthdr.len - sizeof(struct ip);
439 		if (tdb->tdb_xform->xf_type == XF_IP4)
440 			tdb->tdb_cur_bytes += obytes;
441 		break;
442 
443 #ifdef INET6
444 	case AF_INET6:
445 		if (IN6_IS_ADDR_UNSPECIFIED(&tdb->tdb_dst.sin6.sin6_addr) ||
446 		    tdb->tdb_src.sa.sa_family != AF_INET6 ||
447 		    IN6_IS_ADDR_UNSPECIFIED(&tdb->tdb_src.sin6.sin6_addr)) {
448 
449 			DPRINTF("unspecified tunnel endpoint address "
450 			    "in SA %s/%08x",
451 			    ipsp_address(&tdb->tdb_dst, buf, sizeof(buf)),
452 			    ntohl(tdb->tdb_spi));
453 
454 			ipipstat_inc(ipips_unspec);
455 			error = EINVAL;
456 			goto drop;
457 		}
458 
459 		/* If the inner protocol is IPv6, clear link local scope */
460 		if (tp == (IPV6_VERSION >> 4)) {
461 			/* scoped address handling */
462 			ip6 = mtod(m, struct ip6_hdr *);
463 			if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
464 				ip6->ip6_src.s6_addr16[1] = 0;
465 			if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
466 				ip6->ip6_dst.s6_addr16[1] = 0;
467 		}
468 
469 		M_PREPEND(*mp, sizeof(struct ip6_hdr), M_DONTWAIT);
470 		if (*mp == NULL) {
471 			DPRINTF("M_PREPEND failed");
472 			ipipstat_inc(ipips_hdrops);
473 			error = ENOBUFS;
474 			goto drop;
475 		}
476 		m = *mp;
477 
478 		/* Initialize IPv6 header */
479 		ip6o = mtod(m, struct ip6_hdr *);
480 		ip6o->ip6_flow = 0;
481 		ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
482 		ip6o->ip6_vfc |= IPV6_VERSION;
483 		ip6o->ip6_plen = htons(m->m_pkthdr.len - sizeof(*ip6o));
484 		ip6o->ip6_hlim = ip_defttl;
485 		in6_embedscope(&ip6o->ip6_src, &tdb->tdb_src.sin6, NULL, NULL);
486 		in6_embedscope(&ip6o->ip6_dst, &tdb->tdb_dst.sin6, NULL, NULL);
487 
488 		if (tp == IPVERSION) {
489 			/* Save ECN notification */
490 			m_copydata(m, sizeof(struct ip6_hdr) +
491 			    offsetof(struct ip, ip_tos), sizeof(u_int8_t),
492 			    (caddr_t) &itos);
493 
494 			/* This is really IPVERSION. */
495 			ip6o->ip6_nxt = IPPROTO_IPIP;
496 		}
497 		else
498 			if (tp == (IPV6_VERSION >> 4)) {
499 				u_int32_t itos32;
500 
501 				/* Save ECN notification. */
502 				m_copydata(m, sizeof(struct ip6_hdr) +
503 				    offsetof(struct ip6_hdr, ip6_flow),
504 				    sizeof(u_int32_t), (caddr_t) &itos32);
505 				itos = ntohl(itos32) >> 20;
506 
507 				ip6o->ip6_nxt = IPPROTO_IPV6;
508 			} else {
509 				ipipstat_inc(ipips_family);
510 				error = EAFNOSUPPORT;
511 				goto drop;
512 			}
513 
514 		otos = 0;
515 		ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
516 		ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
517 
518 		obytes = m->m_pkthdr.len - sizeof(struct ip6_hdr);
519 		if (tdb->tdb_xform->xf_type == XF_IP4)
520 			tdb->tdb_cur_bytes += obytes;
521 		break;
522 #endif /* INET6 */
523 
524 	default:
525 		DPRINTF("unsupported protocol family %d",
526 		    tdb->tdb_dst.sa.sa_family);
527 		ipipstat_inc(ipips_family);
528 		error = EPFNOSUPPORT;
529 		goto drop;
530 	}
531 
532 	ipipstat_pkt(ipips_opackets, ipips_obytes, obytes);
533 	return 0;
534 
535  drop:
536 	m_freemp(mp);
537 	return error;
538 }
539 
540 #ifdef IPSEC
541 int
542 ipe4_attach(void)
543 {
544 	return 0;
545 }
546 
547 int
548 ipe4_init(struct tdb *tdbp, const struct xformsw *xsp, struct ipsecinit *ii)
549 {
550 	tdbp->tdb_xform = xsp;
551 	return 0;
552 }
553 
554 int
555 ipe4_zeroize(struct tdb *tdbp)
556 {
557 	return 0;
558 }
559 
560 int
561 ipe4_input(struct mbuf **mp, struct tdb *tdb, int hlen, int proto)
562 {
563 	/* This is a rather serious mistake, so no conditional printing. */
564 	printf("%s: should never be called\n", __func__);
565 	m_freemp(mp);
566 	return EINVAL;
567 }
568 #endif	/* IPSEC */
569 
570 int
571 ipip_sysctl_ipipstat(void *oldp, size_t *oldlenp, void *newp)
572 {
573 	struct ipipstat ipipstat;
574 
575 	CTASSERT(sizeof(ipipstat) == (ipips_ncounters * sizeof(uint64_t)));
576 	memset(&ipipstat, 0, sizeof ipipstat);
577 	counters_read(ipipcounters, (uint64_t *)&ipipstat, ipips_ncounters,
578 	    NULL);
579 	return (sysctl_rdstruct(oldp, oldlenp, newp,
580 	    &ipipstat, sizeof(ipipstat)));
581 }
582 
583 int
584 ipip_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
585     size_t newlen)
586 {
587 	int error;
588 
589 	/* All sysctl names at this level are terminal. */
590 	if (namelen != 1)
591 		return (ENOTDIR);
592 
593 	switch (name[0]) {
594 	case IPIPCTL_ALLOW:
595 		NET_LOCK();
596 		error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
597 		    &ipip_allow, 0, 2);
598 		NET_UNLOCK();
599 		return (error);
600 	case IPIPCTL_STATS:
601 		return (ipip_sysctl_ipipstat(oldp, oldlenp, newp));
602 	default:
603 		return (ENOPROTOOPT);
604 	}
605 	/* NOTREACHED */
606 }
607