xref: /netbsd-src/sys/netipsec/xform_ipip.c (revision c38e7cc395b1472a774ff828e46123de44c628e9)
1 /*	$NetBSD: xform_ipip.c,v 1.70 2018/04/29 14:35:35 maxv Exp $	*/
2 /*	$FreeBSD: xform_ipip.c,v 1.3.2.1 2003/01/24 05:11:36 sam Exp $	*/
3 /*	$OpenBSD: ip_ipip.c,v 1.25 2002/06/10 18:04:55 itojun Exp $ */
4 
5 /*
6  * The authors of this code are John Ioannidis (ji@tla.org),
7  * Angelos D. Keromytis (kermit@csd.uch.gr) and
8  * Niels Provos (provos@physnet.uni-hamburg.de).
9  *
10  * The original version of this code was written by John Ioannidis
11  * for BSD/OS in Athens, Greece, in November 1995.
12  *
13  * Ported to OpenBSD and NetBSD, with additional transforms, in December 1996,
14  * by Angelos D. Keromytis.
15  *
16  * Additional transforms and features in 1997 and 1998 by Angelos D. Keromytis
17  * and Niels Provos.
18  *
19  * Additional features in 1999 by Angelos D. Keromytis.
20  *
21  * Copyright (C) 1995, 1996, 1997, 1998, 1999 by John Ioannidis,
22  * Angelos D. Keromytis and Niels Provos.
23  * Copyright (c) 2001, Angelos D. Keromytis.
24  *
25  * Permission to use, copy, and modify this software with or without fee
26  * is hereby granted, provided that this entire notice is included in
27  * all copies of any software which is or includes a copy or
28  * modification of this software.
29  * You may use this code under the GNU public license if you so wish. Please
30  * contribute changes back to the authors under this freer than GPL license
31  * so that we may further the use of strong encryption without limitations to
32  * all.
33  *
34  * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR
35  * IMPLIED WARRANTY. IN PARTICULAR, NONE OF THE AUTHORS MAKES ANY
36  * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE
37  * MERCHANTABILITY OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR
38  * PURPOSE.
39  */
40 
41 #include <sys/cdefs.h>
42 __KERNEL_RCSID(0, "$NetBSD: xform_ipip.c,v 1.70 2018/04/29 14:35:35 maxv Exp $");
43 
44 /*
45  * IP-inside-IP processing
46  */
47 #if defined(_KERNEL_OPT)
48 #include "opt_inet.h"
49 #endif
50 
51 #include <sys/param.h>
52 #include <sys/systm.h>
53 #include <sys/mbuf.h>
54 #include <sys/socket.h>
55 #include <sys/kernel.h>
56 #include <sys/protosw.h>
57 #include <sys/sysctl.h>
58 
59 #include <net/if.h>
60 #include <net/route.h>
61 #include <net/netisr.h>
62 
63 #include <netinet/in.h>
64 #include <netinet/in_systm.h>
65 #include <netinet/in_var.h>
66 #include <netinet/ip.h>
67 #include <netinet/ip_ecn.h>
68 #include <netinet/ip_var.h>
69 #include <netinet/ip_encap.h>
70 
71 #include <netipsec/ipsec.h>
72 #include <netipsec/ipsec_private.h>
73 #include <netipsec/xform.h>
74 
75 #include <netipsec/ipip_var.h>
76 
77 #ifdef INET6
78 #include <netinet/ip6.h>
79 #include <netipsec/ipsec6.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 /* XXX IPCOMP */
88 #define	M_IPSEC	(M_AUTHIPHDR|M_DECRYPTED)
89 
90 int ipip_spoofcheck = 1;
91 percpu_t *ipipstat_percpu;
92 
93 void ipe4_attach(void);
94 
95 static void _ipip_input(struct mbuf *, int);
96 
97 #ifdef INET6
98 int
99 ip4_input6(struct mbuf **m, int *offp, int proto, void *eparg __unused)
100 {
101 	_ipip_input(*m, *offp);
102 	return IPPROTO_DONE;
103 }
104 #endif
105 
106 #ifdef INET
107 void
108 ip4_input(struct mbuf *m, int off, int proto, void *eparg __unused)
109 {
110 	_ipip_input(m, off);
111 }
112 #endif
113 
114 /*
115  * _ipip_input gets called when we receive an IP{46} encapsulated packet,
116  * because AH or ESP were being used in tunnel mode.
117  */
118 static void
119 _ipip_input(struct mbuf *m, int iphlen)
120 {
121 	register struct sockaddr_in *sin;
122 	register struct ifnet *ifp;
123 	register struct ifaddr *ifa;
124 	pktqueue_t *pktq = NULL;
125 	struct ip *ip4 = NULL;
126 #ifdef INET6
127 	register struct sockaddr_in6 *sin6;
128 	struct ip6_hdr *ip6 = NULL;
129 	uint8_t itos;
130 #endif
131 	uint8_t otos;
132 	uint8_t v;
133 	int hlen;
134 
135 	IPIP_STATINC(IPIP_STAT_IPACKETS);
136 
137 	m_copydata(m, 0, 1, &v);
138 
139 	switch (v >> 4) {
140 #ifdef INET
141 	case 4:
142 		hlen = sizeof(struct ip);
143 		break;
144 #endif
145 #ifdef INET6
146 	case 6:
147 		hlen = sizeof(struct ip6_hdr);
148 		break;
149 #endif
150 	default:
151 		DPRINTF(("%s: bad protocol version 0x%x (%u) "
152 		    "for outer header\n", __func__, v, v>>4));
153 		IPIP_STATINC(IPIP_STAT_FAMILY);
154 		m_freem(m);
155 		return;
156 	}
157 
158 	/* Bring the IP header in the first mbuf, if not there already */
159 	if (m->m_len < hlen) {
160 		if ((m = m_pullup(m, hlen)) == NULL) {
161 			DPRINTF(("%s: m_pullup (1) failed\n", __func__));
162 			IPIP_STATINC(IPIP_STAT_HDROPS);
163 			return;
164 		}
165 	}
166 
167 	/* Keep outer ecn field. */
168 	switch (v >> 4) {
169 #ifdef INET
170 	case 4:
171 		otos = mtod(m, struct ip *)->ip_tos;
172 		break;
173 #endif
174 #ifdef INET6
175 	case 6:
176 		otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
177 		break;
178 #endif
179 	default:
180 		panic("%s: impossible (1)", __func__);
181 	}
182 
183 	/* Remove outer IP header */
184 	m_adj(m, iphlen);
185 
186 	/* Sanity check */
187 	if (m->m_pkthdr.len < sizeof(struct ip))  {
188 		IPIP_STATINC(IPIP_STAT_HDROPS);
189 		m_freem(m);
190 		return;
191 	}
192 
193 	m_copydata(m, 0, 1, &v);
194 
195 	switch (v >> 4) {
196 #ifdef INET
197 	case 4:
198 		hlen = sizeof(struct ip);
199 		pktq = ip_pktq;
200 		break;
201 #endif
202 #ifdef INET6
203 	case 6:
204 		hlen = sizeof(struct ip6_hdr);
205 		pktq = ip6_pktq;
206 		break;
207 #endif
208 	default:
209 		DPRINTF(("%s: bad protocol version %#x (%u) "
210 		    "for inner header\n", __func__, v, v >> 4));
211 		IPIP_STATINC(IPIP_STAT_FAMILY);
212 		m_freem(m);
213 		return;
214 	}
215 
216 	/*
217 	 * Bring the inner IP header in the first mbuf, if not there already.
218 	 */
219 	if (m->m_len < hlen) {
220 		if ((m = m_pullup(m, hlen)) == NULL) {
221 			DPRINTF(("%s: m_pullup (2) failed\n", __func__));
222 			IPIP_STATINC(IPIP_STAT_HDROPS);
223 			return;
224 		}
225 	}
226 
227 	/*
228 	 * RFC 1853 specifies that the inner TTL should not be touched on
229 	 * decapsulation. There's no reason this comment should be here, but
230 	 * this is as good as any a position.
231 	 */
232 
233 	/* Some sanity checks in the inner IP header */
234 	switch (v >> 4) {
235 #ifdef INET
236 	case 4:
237 		ip4 = mtod(m, struct ip *);
238 		ip_ecn_egress(ip4_ipsec_ecn, &otos, &ip4->ip_tos);
239 		break;
240 #endif
241 #ifdef INET6
242 	case 6:
243 		ip6 = mtod(m, struct ip6_hdr *);
244 		itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
245 		ip_ecn_egress(ip6_ipsec_ecn, &otos, &itos);
246 		ip6->ip6_flow &= ~htonl(0xff << 20);
247 		ip6->ip6_flow |= htonl((uint32_t)itos << 20);
248 		break;
249 #endif
250 	default:
251 		panic("%s: impossible (2)", __func__);
252 	}
253 
254 	/* Check for local address spoofing. */
255 	if ((m_get_rcvif_NOMPSAFE(m) == NULL ||
256 	    !(m_get_rcvif_NOMPSAFE(m)->if_flags & IFF_LOOPBACK)) &&
257 	    ipip_spoofcheck) {
258 		int s = pserialize_read_enter();
259 		IFNET_READER_FOREACH(ifp) {
260 			IFADDR_READER_FOREACH(ifa, ifp) {
261 #ifdef INET
262 				if (ip4) {
263 					if (ifa->ifa_addr->sa_family !=
264 					    AF_INET)
265 						continue;
266 
267 					sin = (struct sockaddr_in *)ifa->ifa_addr;
268 
269 					if (sin->sin_addr.s_addr ==
270 					    ip4->ip_src.s_addr)	{
271 						pserialize_read_exit(s);
272 						IPIP_STATINC(IPIP_STAT_SPOOF);
273 						m_freem(m);
274 						return;
275 					}
276 				}
277 #endif
278 
279 #ifdef INET6
280 				if (ip6) {
281 					if (ifa->ifa_addr->sa_family !=
282 					    AF_INET6)
283 						continue;
284 
285 					sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
286 
287 					if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_src)) {
288 						pserialize_read_exit(s);
289 						IPIP_STATINC(IPIP_STAT_SPOOF);
290 						m_freem(m);
291 						return;
292 					}
293 
294 				}
295 #endif
296 			}
297 		}
298 		pserialize_read_exit(s);
299 	}
300 
301 	/* Statistics: m->m_pkthdr.len is the length of the inner packet */
302 	IPIP_STATADD(IPIP_STAT_IBYTES, m->m_pkthdr.len);
303 
304 	/*
305 	 * Interface pointer stays the same; if no IPsec processing has
306 	 * been done (or will be done), this will point to a normal
307 	 * interface. Otherwise, it'll point to an enc interface, which
308 	 * will allow a packet filter to distinguish between secure and
309 	 * untrusted packets.
310 	 */
311 
312 	int s = splnet();
313 	if (__predict_false(!pktq_enqueue(pktq, m, 0))) {
314 		IPIP_STATINC(IPIP_STAT_QFULL);
315 		m_freem(m);
316 	}
317 	splx(s);
318 }
319 
320 int
321 ipip_output(struct mbuf *m, const struct ipsecrequest *isr,
322     struct secasvar *sav, struct mbuf **mp, int skip, int protoff)
323 {
324 	char buf[IPSEC_ADDRSTRLEN];
325 	uint8_t tp, otos;
326 	struct secasindex *saidx;
327 	int error, iphlen;
328 #ifdef INET
329 	uint8_t itos;
330 	struct ip *ipo;
331 #endif
332 #ifdef INET6
333 	struct ip6_hdr *ip6, *ip6o;
334 #endif
335 
336 	KASSERT(sav != NULL);
337 
338 	/* XXX Deal with empty TDB source/destination addresses. */
339 
340 	m_copydata(m, 0, 1, &tp);
341 	tp = (tp >> 4) & 0xff;  /* Get the IP version number. */
342 
343 	saidx = &sav->sah->saidx;
344 	switch (saidx->dst.sa.sa_family) {
345 #ifdef INET
346 	case AF_INET:
347 		if (saidx->src.sa.sa_family != AF_INET ||
348 		    saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
349 		    saidx->dst.sin.sin_addr.s_addr == INADDR_ANY) {
350 			DPRINTF(("%s: unspecified tunnel endpoint "
351 			    "address in SA %s/%08lx\n", __func__,
352 			    ipsec_address(&saidx->dst, buf, sizeof(buf)),
353 			    (u_long)ntohl(sav->spi)));
354 			IPIP_STATINC(IPIP_STAT_UNSPEC);
355 			error = EINVAL;
356 			goto bad;
357 		}
358 
359 		M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
360 		if (m == NULL) {
361 			DPRINTF(("%s: M_PREPEND failed\n", __func__));
362 			IPIP_STATINC(IPIP_STAT_HDROPS);
363 			error = ENOBUFS;
364 			goto bad;
365 		}
366 
367 		iphlen = sizeof(struct ip);
368 
369 		ipo = mtod(m, struct ip *);
370 		ipo->ip_v = IPVERSION;
371 		ipo->ip_hl = 5;
372 		ipo->ip_len = htons(m->m_pkthdr.len);
373 		ipo->ip_ttl = ip_defttl;
374 		ipo->ip_sum = 0;
375 		ipo->ip_src = saidx->src.sin.sin_addr;
376 		ipo->ip_dst = saidx->dst.sin.sin_addr;
377 		ipo->ip_id = ip_newid(NULL);
378 
379 		/* If the inner protocol is IP... */
380 		if (tp == IPVERSION) {
381 			/* Save ECN notification */
382 			m_copydata(m, sizeof(struct ip) +
383 			    offsetof(struct ip, ip_tos),
384 			    sizeof(uint8_t), &itos);
385 
386 			ipo->ip_p = IPPROTO_IPIP;
387 
388 			/*
389 			 * We should be keeping tunnel soft-state and
390 			 * send back ICMPs if needed.
391 			 */
392 			m_copydata(m, sizeof(struct ip) +
393 			    offsetof(struct ip, ip_off),
394 			    sizeof(uint16_t), &ipo->ip_off);
395 			ipo->ip_off &= ~ htons(IP_DF | IP_MF | IP_OFFMASK);
396 		}
397 #ifdef INET6
398 		else if (tp == (IPV6_VERSION >> 4)) {
399 			uint32_t itos32;
400 
401 			/* Save ECN notification. */
402 			m_copydata(m, sizeof(struct ip) +
403 			    offsetof(struct ip6_hdr, ip6_flow),
404 			    sizeof(uint32_t), &itos32);
405 			itos = ntohl(itos32) >> 20;
406 			ipo->ip_p = IPPROTO_IPV6;
407 			ipo->ip_off = 0;
408 		}
409 #endif
410 		else {
411 			goto nofamily;
412 		}
413 
414 		otos = 0;
415 		ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
416 		ipo->ip_tos = otos;
417 		break;
418 #endif /* INET */
419 
420 #ifdef INET6
421 	case AF_INET6:
422 		if (IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr) ||
423 		    saidx->src.sa.sa_family != AF_INET6 ||
424 		    IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr)) {
425 			DPRINTF(("%s: unspecified tunnel endpoint "
426 			    "address in SA %s/%08lx\n", __func__,
427 			    ipsec_address(&saidx->dst, buf, sizeof(buf)),
428 			    (u_long)ntohl(sav->spi)));
429 			IPIP_STATINC(IPIP_STAT_UNSPEC);
430 			error = ENOBUFS;
431 			goto bad;
432 		}
433 
434 		if (tp == (IPV6_VERSION >> 4)) {
435 			/* scoped address handling */
436 			ip6 = mtod(m, struct ip6_hdr *);
437 			if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
438 				ip6->ip6_src.s6_addr16[1] = 0;
439 			if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
440 				ip6->ip6_dst.s6_addr16[1] = 0;
441 		}
442 
443 		M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
444 		if (m == NULL) {
445 			DPRINTF(("%s: M_PREPEND failed\n", __func__));
446 			IPIP_STATINC(IPIP_STAT_HDROPS);
447 			error = ENOBUFS;
448 			goto bad;
449 		}
450 
451 		iphlen = sizeof(struct ip6_hdr);
452 
453 		/* Initialize IPv6 header */
454 		ip6o = mtod(m, struct ip6_hdr *);
455 		ip6o->ip6_flow = 0;
456 		ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
457 		ip6o->ip6_vfc |= IPV6_VERSION;
458 		ip6o->ip6_plen = htons(m->m_pkthdr.len - sizeof(*ip6o));
459 		ip6o->ip6_hlim = ip_defttl;
460 		ip6o->ip6_dst = saidx->dst.sin6.sin6_addr;
461 		ip6o->ip6_src = saidx->src.sin6.sin6_addr;
462 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6o->ip6_dst))
463 			ip6o->ip6_dst.s6_addr16[1] = htons(saidx->dst.sin6.sin6_scope_id);
464 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6o->ip6_src))
465 			ip6o->ip6_src.s6_addr16[1] = htons(saidx->src.sin6.sin6_scope_id);
466 
467 #ifdef INET
468 		if (tp == IPVERSION) {
469 			/* Save ECN notification */
470 			m_copydata(m, sizeof(struct ip6_hdr) +
471 			    offsetof(struct ip, ip_tos), sizeof(uint8_t),
472 			    &itos);
473 
474 			/* This is really IPVERSION. */
475 			ip6o->ip6_nxt = IPPROTO_IPIP;
476 		} else
477 #endif
478 		if (tp == (IPV6_VERSION >> 4)) {
479 			uint32_t itos32;
480 
481 			/* Save ECN notification. */
482 			m_copydata(m, sizeof(struct ip6_hdr) +
483 			    offsetof(struct ip6_hdr, ip6_flow),
484 			    sizeof(uint32_t), &itos32);
485 			itos = ntohl(itos32) >> 20;
486 
487 			ip6o->ip6_nxt = IPPROTO_IPV6;
488 		} else {
489 			goto nofamily;
490 		}
491 
492 		otos = 0;
493 		ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
494 		ip6o->ip6_flow |= htonl((uint32_t)otos << 20);
495 		break;
496 #endif /* INET6 */
497 
498 	default:
499 nofamily:
500 		DPRINTF(("%s: unsupported protocol family %u\n", __func__,
501 		    saidx->dst.sa.sa_family));
502 		IPIP_STATINC(IPIP_STAT_FAMILY);
503 		error = EAFNOSUPPORT;
504 		goto bad;
505 	}
506 
507 	IPIP_STATINC(IPIP_STAT_OPACKETS);
508 	IPIP_STATADD(IPIP_STAT_OBYTES, m->m_pkthdr.len - iphlen);
509 
510 	*mp = m;
511 	return 0;
512 
513 bad:
514 	if (m)
515 		m_freem(m);
516 	*mp = NULL;
517 	return error;
518 }
519 
520 static int
521 ipe4_init(struct secasvar *sav, const struct xformsw *xsp)
522 {
523 	sav->tdb_xform = xsp;
524 	return 0;
525 }
526 
527 static int
528 ipe4_zeroize(struct secasvar *sav)
529 {
530 	sav->tdb_xform = NULL;
531 	return 0;
532 }
533 
534 static int
535 ipe4_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff)
536 {
537 	/* This is a rather serious mistake, so no conditional printing. */
538 	printf("%s: should never be called\n", __func__);
539 	if (m)
540 		m_freem(m);
541 	return EOPNOTSUPP;
542 }
543 
544 static struct xformsw ipe4_xformsw = {
545 	.xf_type	= XF_IP4,
546 	.xf_flags	= 0,
547 	.xf_name	= "IPv4 Simple Encapsulation",
548 	.xf_init	= ipe4_init,
549 	.xf_zeroize	= ipe4_zeroize,
550 	.xf_input	= ipe4_input,
551 	.xf_output	= ipip_output,
552 	.xf_next	= NULL,
553 };
554 
555 #ifdef INET
556 static struct encapsw ipe4_encapsw = {
557 	.encapsw4 = {
558 		.pr_input = ip4_input,
559 		.pr_ctlinput = NULL,
560 	}
561 };
562 #endif
563 #ifdef INET6
564 static struct encapsw ipe4_encapsw6 = {
565 	.encapsw6 = {
566 		.pr_input = ip4_input6,
567 		.pr_ctlinput = NULL,
568 	}
569 };
570 #endif
571 
572 /*
573  * Check the encapsulated packet to see if we want it
574  */
575 static int
576 ipe4_encapcheck(struct mbuf *m, int off, int proto, void *arg)
577 {
578 	/*
579 	 * Only take packets coming from IPSEC tunnels; the rest
580 	 * must be handled by the gif tunnel code.  Note that we
581 	 * also return a minimum priority when we want the packet
582 	 * so any explicit gif tunnels take precedence.
583 	 */
584 	return ((m->m_flags & M_IPSEC) != 0 ? 1 : 0);
585 }
586 
587 void
588 ipe4_attach(void)
589 {
590 
591 	ipipstat_percpu = percpu_alloc(sizeof(uint64_t) * IPIP_NSTATS);
592 
593 	xform_register(&ipe4_xformsw);
594 	/* attach to encapsulation framework */
595 	/* XXX save return cookie for detach on module remove */
596 
597 	encapinit();
598 	/* This function is called before ifinit(). Who else gets lock? */
599 	(void)encap_lock_enter();
600 	/* ipe4_encapsw and ipe4_encapsw must be added atomically */
601 #ifdef INET
602 	(void)encap_attach_func(AF_INET, -1, ipe4_encapcheck, &ipe4_encapsw,
603 	    NULL);
604 #endif
605 #ifdef INET6
606 	(void)encap_attach_func(AF_INET6, -1, ipe4_encapcheck, &ipe4_encapsw6,
607 	    NULL);
608 #endif
609 	encap_lock_exit();
610 }
611