xform_ipip.c revision 111119 
1/*	$FreeBSD: head/sys/netipsec/xform_ipip.c 111119 2003-02-19 05:47:46Z imp $	*/
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_random_ip_id.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/route.h>
56#include <net/netisr.h>
57
58#include <netinet/in.h>
59#include <netinet/in_systm.h>
60#include <netinet/in_var.h>
61#include <netinet/ip.h>
62#include <netinet/ip_ecn.h>
63#include <netinet/ip_var.h>
64#include <netinet/ip_encap.h>
65#include <netinet/ipprotosw.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 */
93int	ipip_allow = 0;
94struct	ipipstat ipipstat;
95
96SYSCTL_DECL(_net_inet_ipip);
97SYSCTL_INT(_net_inet_ipip, OID_AUTO,
98	ipip_allow,	CTLFLAG_RW,	&ipip_allow,	0, "");
99SYSCTL_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
105static 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 */
111int
112ip4_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(("ip4_input6: dropped due to policy\n"));
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 */
132void
133ip4_input(struct mbuf *m, ...)
134{
135	va_list ap;
136	int iphlen;
137
138#if 0
139	/* If we do not accept IP-in-IP explicitly, drop.  */
140	if (!ipip_allow && (m->m_flags & M_IPSEC) == 0) {
141		DPRINTF(("ip4_input: dropped due to policy\n"));
142		ipipstat.ipips_pdrops++;
143		m_freem(m);
144		return;
145	}
146#endif
147	va_start(ap, m);
148	iphlen = va_arg(ap, int);
149	va_end(ap);
150
151	_ipip_input(m, iphlen, NULL);
152}
153#endif /* INET */
154
155/*
156 * ipip_input gets called when we receive an IP{46} encapsulated packet,
157 * either because we got it at a real interface, or because AH or ESP
158 * were being used in tunnel mode (in which case the rcvif element will
159 * contain the address of the encX interface associated with the tunnel.
160 */
161
162static void
163_ipip_input(struct mbuf *m, int iphlen, struct ifnet *gifp)
164{
165	register struct sockaddr_in *sin;
166	register struct ifnet *ifp;
167	register struct ifaddr *ifa;
168	struct ifqueue *ifq = NULL;
169	struct ip *ipo;
170#ifdef INET6
171	register struct sockaddr_in6 *sin6;
172	struct ip6_hdr *ip6 = NULL;
173	u_int8_t itos;
174#endif
175	u_int8_t nxt;
176	int isr;
177	u_int8_t otos;
178	u_int8_t v;
179	int hlen;
180
181	ipipstat.ipips_ipackets++;
182
183	m_copydata(m, 0, 1, &v);
184
185	switch (v >> 4) {
186#ifdef INET
187        case 4:
188		hlen = sizeof(struct ip);
189		break;
190#endif /* INET */
191#ifdef INET6
192        case 6:
193		hlen = sizeof(struct ip6_hdr);
194		break;
195#endif
196        default:
197		ipipstat.ipips_family++;
198		m_freem(m);
199		return /* EAFNOSUPPORT */;
200	}
201
202	/* Bring the IP header in the first mbuf, if not there already */
203	if (m->m_len < hlen) {
204		if ((m = m_pullup(m, hlen)) == NULL) {
205			DPRINTF(("ipip_input: m_pullup (1) failed\n"));
206			ipipstat.ipips_hdrops++;
207			return;
208		}
209	}
210
211	ipo = mtod(m, struct ip *);
212
213#ifdef MROUTING
214	if (ipo->ip_v == IPVERSION && ipo->ip_p == IPPROTO_IPV4) {
215		if (IN_MULTICAST(((struct ip *)((char *) ipo + iphlen))->ip_dst.s_addr)) {
216			ipip_mroute_input (m, iphlen);
217			return;
218		}
219	}
220#endif /* MROUTING */
221
222	/* Keep outer ecn field. */
223	switch (v >> 4) {
224#ifdef INET
225	case 4:
226		otos = ipo->ip_tos;
227		break;
228#endif /* INET */
229#ifdef INET6
230	case 6:
231		otos = (ntohl(mtod(m, struct ip6_hdr *)->ip6_flow) >> 20) & 0xff;
232		break;
233#endif
234	default:
235		panic("ipip_input: unknown ip version %u (outer)", v>>4);
236	}
237
238	/* Remove outer IP header */
239	m_adj(m, iphlen);
240
241	/* Sanity check */
242	if (m->m_pkthdr.len < sizeof(struct ip))  {
243		ipipstat.ipips_hdrops++;
244		m_freem(m);
245		return;
246	}
247
248	m_copydata(m, 0, 1, &v);
249
250	switch (v >> 4) {
251#ifdef INET
252        case 4:
253		hlen = sizeof(struct ip);
254		break;
255#endif /* INET */
256
257#ifdef INET6
258        case 6:
259		hlen = sizeof(struct ip6_hdr);
260		break;
261#endif
262	default:
263		ipipstat.ipips_family++;
264		m_freem(m);
265		return; /* EAFNOSUPPORT */
266	}
267
268	/*
269	 * Bring the inner IP header in the first mbuf, if not there already.
270	 */
271	if (m->m_len < hlen) {
272		if ((m = m_pullup(m, hlen)) == NULL) {
273			DPRINTF(("ipip_input: m_pullup (2) failed\n"));
274			ipipstat.ipips_hdrops++;
275			return;
276		}
277	}
278
279	/*
280	 * RFC 1853 specifies that the inner TTL should not be touched on
281	 * decapsulation. There's no reason this comment should be here, but
282	 * this is as good as any a position.
283	 */
284
285	/* Some sanity checks in the inner IP header */
286	switch (v >> 4) {
287#ifdef INET
288    	case 4:
289                ipo = mtod(m, struct ip *);
290                nxt = ipo->ip_p;
291		ip_ecn_egress(ip4_ipsec_ecn, &otos, &ipo->ip_tos);
292                break;
293#endif /* INET */
294#ifdef INET6
295    	case 6:
296                ip6 = (struct ip6_hdr *) ipo;
297                nxt = ip6->ip6_nxt;
298		itos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
299		ip_ecn_egress(ip6_ipsec_ecn, &otos, &itos);
300		ip6->ip6_flow &= ~htonl(0xff << 20);
301		ip6->ip6_flow |= htonl((u_int32_t) itos << 20);
302                break;
303#endif
304	default:
305		panic("ipip_input: unknown ip version %u (inner)", v>>4);
306	}
307
308	/* Check for local address spoofing. */
309	if ((m->m_pkthdr.rcvif == NULL ||
310	    !(m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK)) &&
311	    ipip_allow != 2) {
312	    	IFNET_RLOCK();
313		for (ifp = ifnet.tqh_first; ifp != 0;
314		     ifp = ifp->if_list.tqe_next) {
315			for (ifa = ifp->if_addrlist.tqh_first; ifa != 0;
316			     ifa = ifa->ifa_list.tqe_next) {
317#ifdef INET
318				if (ipo) {
319					if (ifa->ifa_addr->sa_family !=
320					    AF_INET)
321						continue;
322
323					sin = (struct sockaddr_in *) ifa->ifa_addr;
324
325					if (sin->sin_addr.s_addr ==
326					    ipo->ip_src.s_addr)	{
327						ipipstat.ipips_spoof++;
328						m_freem(m);
329						IFNET_RUNLOCK();
330						return;
331					}
332				}
333#endif /* INET */
334
335#ifdef INET6
336				if (ip6) {
337					if (ifa->ifa_addr->sa_family !=
338					    AF_INET6)
339						continue;
340
341					sin6 = (struct sockaddr_in6 *) ifa->ifa_addr;
342
343					if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_src)) {
344						ipipstat.ipips_spoof++;
345						m_freem(m);
346						IFNET_RUNLOCK();
347						return;
348					}
349
350				}
351#endif /* INET6 */
352			}
353		}
354		IFNET_RUNLOCK();
355	}
356
357	/* Statistics */
358	ipipstat.ipips_ibytes += m->m_pkthdr.len - iphlen;
359
360	/*
361	 * Interface pointer stays the same; if no IPsec processing has
362	 * been done (or will be done), this will point to a normal
363	 * interface. Otherwise, it'll point to an enc interface, which
364	 * will allow a packet filter to distinguish between secure and
365	 * untrusted packets.
366	 */
367
368	switch (v >> 4) {
369#ifdef INET
370	case 4:
371		ifq = &ipintrq;
372		isr = NETISR_IP;
373		break;
374#endif
375#ifdef INET6
376	case 6:
377		ifq = &ip6intrq;
378		isr = NETISR_IPV6;
379		break;
380#endif
381	default:
382		panic("ipip_input: should never reach here");
383	}
384
385	if (!IF_HANDOFF(ifq, m, NULL)) {
386		ipipstat.ipips_qfull++;
387
388		DPRINTF(("ipip_input: packet dropped because of full queue\n"));
389	} else {
390		schednetisr(isr);
391	}
392}
393
394int
395ipip_output(
396	struct mbuf *m,
397	struct ipsecrequest *isr,
398	struct mbuf **mp,
399	int skip,
400	int protoff
401)
402{
403	struct secasvar *sav;
404	u_int8_t tp, otos;
405	struct secasindex *saidx;
406	int error;
407#ifdef INET
408	u_int8_t itos;
409	struct ip *ipo;
410#endif /* INET */
411#ifdef INET6
412	struct ip6_hdr *ip6, *ip6o;
413#endif /* INET6 */
414
415#if 0
416	SPLASSERT(net, "ipip_output");
417#endif
418
419	sav = isr->sav;
420	KASSERT(sav != NULL, ("ipip_output: null SA"));
421	KASSERT(sav->sah != NULL, ("ipip_output: null SAH"));
422
423	/* XXX Deal with empty TDB source/destination addresses. */
424
425	m_copydata(m, 0, 1, &tp);
426	tp = (tp >> 4) & 0xff;  /* Get the IP version number. */
427
428	saidx = &sav->sah->saidx;
429	switch (saidx->dst.sa.sa_family) {
430#ifdef INET
431	case AF_INET:
432		if (saidx->src.sa.sa_family != AF_INET ||
433		    saidx->src.sin.sin_addr.s_addr == INADDR_ANY ||
434		    saidx->dst.sin.sin_addr.s_addr == INADDR_ANY) {
435			DPRINTF(("ipip_output: unspecified tunnel endpoint "
436			    "address in SA %s/%08lx\n",
437			    ipsec_address(&saidx->dst),
438			    (u_long) ntohl(sav->spi)));
439			ipipstat.ipips_unspec++;
440			error = EINVAL;
441			goto bad;
442		}
443
444		M_PREPEND(m, sizeof(struct ip), M_DONTWAIT);
445		if (m == 0) {
446			DPRINTF(("ipip_output: M_PREPEND failed\n"));
447			ipipstat.ipips_hdrops++;
448			error = ENOBUFS;
449			goto bad;
450		}
451
452		ipo = mtod(m, struct ip *);
453
454		ipo->ip_v = IPVERSION;
455		ipo->ip_hl = 5;
456		ipo->ip_len = htons(m->m_pkthdr.len);
457		ipo->ip_ttl = ip_defttl;
458		ipo->ip_sum = 0;
459		ipo->ip_src = saidx->src.sin.sin_addr;
460		ipo->ip_dst = saidx->dst.sin.sin_addr;
461
462#ifdef RANDOM_IP_ID
463		ipo->ip_id = ip_randomid();
464#else
465		ipo->ip_id = htons(ip_id++);
466#endif
467
468		/* If the inner protocol is IP... */
469		if (tp == IPVERSION) {
470			/* Save ECN notification */
471			m_copydata(m, sizeof(struct ip) +
472			    offsetof(struct ip, ip_tos),
473			    sizeof(u_int8_t), (caddr_t) &itos);
474
475			ipo->ip_p = IPPROTO_IPIP;
476
477			/*
478			 * We should be keeping tunnel soft-state and
479			 * send back ICMPs if needed.
480			 */
481			m_copydata(m, sizeof(struct ip) +
482			    offsetof(struct ip, ip_off),
483			    sizeof(u_int16_t), (caddr_t) &ipo->ip_off);
484			ipo->ip_off = ntohs(ipo->ip_off);
485			ipo->ip_off &= ~(IP_DF | IP_MF | IP_OFFMASK);
486			ipo->ip_off = htons(ipo->ip_off);
487		}
488#ifdef INET6
489		else if (tp == (IPV6_VERSION >> 4)) {
490			u_int32_t itos32;
491
492			/* Save ECN notification. */
493			m_copydata(m, sizeof(struct ip) +
494			    offsetof(struct ip6_hdr, ip6_flow),
495			    sizeof(u_int32_t), (caddr_t) &itos32);
496			itos = ntohl(itos32) >> 20;
497			ipo->ip_p = IPPROTO_IPV6;
498			ipo->ip_off = 0;
499		}
500#endif /* INET6 */
501		else {
502			goto nofamily;
503		}
504
505		otos = 0;
506		ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
507		ipo->ip_tos = otos;
508		break;
509#endif /* INET */
510
511#ifdef INET6
512	case AF_INET6:
513		if (IN6_IS_ADDR_UNSPECIFIED(&saidx->dst.sin6.sin6_addr) ||
514		    saidx->src.sa.sa_family != AF_INET6 ||
515		    IN6_IS_ADDR_UNSPECIFIED(&saidx->src.sin6.sin6_addr)) {
516			DPRINTF(("ipip_output: unspecified tunnel endpoint "
517			    "address in SA %s/%08lx\n",
518			    ipsec_address(&saidx->dst),
519			    (u_long) ntohl(sav->spi)));
520			ipipstat.ipips_unspec++;
521			error = ENOBUFS;
522			goto bad;
523		}
524
525		/* scoped address handling */
526		ip6 = mtod(m, struct ip6_hdr *);
527		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
528			ip6->ip6_src.s6_addr16[1] = 0;
529		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
530			ip6->ip6_dst.s6_addr16[1] = 0;
531
532		M_PREPEND(m, sizeof(struct ip6_hdr), M_DONTWAIT);
533		if (m == 0) {
534			DPRINTF(("ipip_output: M_PREPEND failed\n"));
535			ipipstat.ipips_hdrops++;
536			*mp = NULL;
537			error = ENOBUFS;
538			goto bad;
539		}
540
541		/* Initialize IPv6 header */
542		ip6o = mtod(m, struct ip6_hdr *);
543		ip6o->ip6_flow = 0;
544		ip6o->ip6_vfc &= ~IPV6_VERSION_MASK;
545		ip6o->ip6_vfc |= IPV6_VERSION;
546		ip6o->ip6_plen = htons(m->m_pkthdr.len);
547		ip6o->ip6_hlim = ip_defttl;
548		ip6o->ip6_dst = saidx->dst.sin6.sin6_addr;
549		ip6o->ip6_src = saidx->src.sin6.sin6_addr;
550
551#ifdef INET
552		if (tp == IPVERSION) {
553			/* Save ECN notification */
554			m_copydata(m, sizeof(struct ip6_hdr) +
555			    offsetof(struct ip, ip_tos), sizeof(u_int8_t),
556			    (caddr_t) &itos);
557
558			/* This is really IPVERSION. */
559			ip6o->ip6_nxt = IPPROTO_IPIP;
560		} else
561#endif /* INET */
562			if (tp == (IPV6_VERSION >> 4)) {
563				u_int32_t itos32;
564
565				/* Save ECN notification. */
566				m_copydata(m, sizeof(struct ip6_hdr) +
567				    offsetof(struct ip6_hdr, ip6_flow),
568				    sizeof(u_int32_t), (caddr_t) &itos32);
569				itos = ntohl(itos32) >> 20;
570
571				ip6o->ip6_nxt = IPPROTO_IPV6;
572			} else {
573				goto nofamily;
574			}
575
576		otos = 0;
577		ip_ecn_ingress(ECN_ALLOWED, &otos, &itos);
578		ip6o->ip6_flow |= htonl((u_int32_t) otos << 20);
579		break;
580#endif /* INET6 */
581
582	default:
583nofamily:
584		DPRINTF(("ipip_output: unsupported protocol family %u\n",
585		    saidx->dst.sa.sa_family));
586		ipipstat.ipips_family++;
587		error = EAFNOSUPPORT;		/* XXX diffs from openbsd */
588		goto bad;
589	}
590
591	ipipstat.ipips_opackets++;
592	*mp = m;
593
594#ifdef INET
595	if (saidx->dst.sa.sa_family == AF_INET) {
596#if 0
597		if (sav->tdb_xform->xf_type == XF_IP4)
598			tdb->tdb_cur_bytes +=
599			    m->m_pkthdr.len - sizeof(struct ip);
600#endif
601		ipipstat.ipips_obytes += m->m_pkthdr.len - sizeof(struct ip);
602	}
603#endif /* INET */
604
605#ifdef INET6
606	if (saidx->dst.sa.sa_family == AF_INET6) {
607#if 0
608		if (sav->tdb_xform->xf_type == XF_IP4)
609			tdb->tdb_cur_bytes +=
610			    m->m_pkthdr.len - sizeof(struct ip6_hdr);
611#endif
612		ipipstat.ipips_obytes +=
613		    m->m_pkthdr.len - sizeof(struct ip6_hdr);
614	}
615#endif /* INET6 */
616
617	return 0;
618bad:
619	if (m)
620		m_freem(m), *mp = NULL;
621	return (error);
622}
623
624#ifdef FAST_IPSEC
625static int
626ipe4_init(struct secasvar *sav, struct xformsw *xsp)
627{
628	sav->tdb_xform = xsp;
629	return 0;
630}
631
632static int
633ipe4_zeroize(struct secasvar *sav)
634{
635	sav->tdb_xform = NULL;
636	return 0;
637}
638
639static int
640ipe4_input(struct mbuf *m, struct secasvar *sav, int skip, int protoff)
641{
642	/* This is a rather serious mistake, so no conditional printing. */
643	printf("ipe4_input: should never be called\n");
644	if (m)
645		m_freem(m);
646	return EOPNOTSUPP;
647}
648
649static struct xformsw ipe4_xformsw = {
650	XF_IP4,		0,		"IPv4 Simple Encapsulation",
651	ipe4_init,	ipe4_zeroize,	ipe4_input,	ipip_output,
652};
653
654extern struct domain inetdomain;
655static struct ipprotosw ipe4_protosw[] = {
656{ SOCK_RAW,	&inetdomain,	IPPROTO_IPV4,	PR_ATOMIC|PR_ADDR|PR_LASTHDR,
657  (pr_in_input_t*) ip4_input,
658		0, 		0,		rip_ctloutput,
659  0,
660  0,		0,		0,		0,
661  &rip_usrreqs
662},
663#ifdef INET6
664{ SOCK_RAW,	&inetdomain,	IPPROTO_IPV6,	PR_ATOMIC|PR_ADDR|PR_LASTHDR,
665  (pr_in_input_t*) ip4_input,
666		0,	 	0,		rip_ctloutput,
667  0,
668  0,		0,		0,		0,
669  &rip_usrreqs
670}
671#endif
672};
673
674/*
675 * Check the encapsulated packet to see if we want it
676 */
677static int
678ipe4_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
679{
680	/*
681	 * Only take packets coming from IPSEC tunnels; the rest
682	 * must be handled by the gif tunnel code.  Note that we
683	 * also return a minimum priority when we want the packet
684	 * so any explicit gif tunnels take precedence.
685	 */
686	return ((m->m_flags & M_IPSEC) != 0 ? 1 : 0);
687}
688
689static void
690ipe4_attach(void)
691{
692	xform_register(&ipe4_xformsw);
693	/* attach to encapsulation framework */
694	/* XXX save return cookie for detach on module remove */
695	(void) encap_attach_func(AF_INET, -1,
696		ipe4_encapcheck, (struct protosw*) &ipe4_protosw[0], NULL);
697#ifdef INET6
698	(void) encap_attach_func(AF_INET6, -1,
699		ipe4_encapcheck, (struct protosw*) &ipe4_protosw[1], NULL);
700#endif
701}
702SYSINIT(ipe4_xform_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ipe4_attach, NULL);
703#endif	/* FAST_IPSEC */
704