1/*-
2 * Copyright (c) 2002, 2003 Sam Leffler, Errno Consulting
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 *
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
17 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
24 * SUCH DAMAGE.
25 *
26 * $FreeBSD$
27 */
28
29/*
30 * IPsec output processing.
31 */
32#include "opt_inet.h"
33#include "opt_inet6.h"
34#include "opt_ipsec.h"
35#include "opt_enc.h"
36
37#include <sys/param.h>
38#include <sys/systm.h>
39#include <sys/mbuf.h>
40#include <sys/domain.h>
41#include <sys/protosw.h>
42#include <sys/socket.h>
43#include <sys/errno.h>
44#include <sys/syslog.h>
45
46#include <net/if.h>
47#include <net/pfil.h>
48#include <net/route.h>
49#include <net/vnet.h>
50
51#include <netinet/in.h>
52#include <netinet/in_systm.h>
53#include <netinet/ip.h>
54#include <netinet/ip_var.h>
55#include <netinet/in_var.h>
56#include <netinet/ip_ecn.h>
57#ifdef INET6
58#include <netinet6/ip6_ecn.h>
59#endif
60
61#include <netinet/ip6.h>
62#ifdef INET6
63#include <netinet6/ip6_var.h>
64#endif
65#include <netinet/in_pcb.h>
66#ifdef INET6
67#include <netinet/icmp6.h>
68#endif
69
70#include <netipsec/ipsec.h>
71#ifdef INET6
72#include <netipsec/ipsec6.h>
73#endif
74#include <netipsec/ah_var.h>
75#include <netipsec/esp_var.h>
76#include <netipsec/ipcomp_var.h>
77
78#include <netipsec/xform.h>
79
80#include <netipsec/key.h>
81#include <netipsec/keydb.h>
82#include <netipsec/key_debug.h>
83
84#include <machine/in_cksum.h>
85
86#ifdef IPSEC_NAT_T
87#include <netinet/udp.h>
88#endif
89
90#ifdef DEV_ENC
91#include <net/if_enc.h>
92#endif
93
94
95int
96ipsec_process_done(struct mbuf *m, struct ipsecrequest *isr)
97{
98	struct tdb_ident *tdbi;
99	struct m_tag *mtag;
100	struct secasvar *sav;
101	struct secasindex *saidx;
102	int error;
103
104	IPSEC_ASSERT(m != NULL, ("null mbuf"));
105	IPSEC_ASSERT(isr != NULL, ("null ISR"));
106	sav = isr->sav;
107	IPSEC_ASSERT(sav != NULL, ("null SA"));
108	IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
109
110	saidx = &sav->sah->saidx;
111	switch (saidx->dst.sa.sa_family) {
112#ifdef INET
113	case AF_INET:
114		/* Fix the header length, for AH processing. */
115		mtod(m, struct ip *)->ip_len = htons(m->m_pkthdr.len);
116		break;
117#endif /* INET */
118#ifdef INET6
119	case AF_INET6:
120		/* Fix the header length, for AH processing. */
121		if (m->m_pkthdr.len < sizeof (struct ip6_hdr)) {
122			error = ENXIO;
123			goto bad;
124		}
125		if (m->m_pkthdr.len - sizeof (struct ip6_hdr) > IPV6_MAXPACKET) {
126			/* No jumbogram support. */
127			error = ENXIO;	/*?*/
128			goto bad;
129		}
130		mtod(m, struct ip6_hdr *)->ip6_plen =
131			htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
132		break;
133#endif /* INET6 */
134	default:
135		DPRINTF(("%s: unknown protocol family %u\n", __func__,
136		    saidx->dst.sa.sa_family));
137		error = ENXIO;
138		goto bad;
139	}
140
141	/*
142	 * Add a record of what we've done or what needs to be done to the
143	 * packet.
144	 */
145	mtag = m_tag_get(PACKET_TAG_IPSEC_OUT_DONE,
146			sizeof(struct tdb_ident), M_NOWAIT);
147	if (mtag == NULL) {
148		DPRINTF(("%s: could not get packet tag\n", __func__));
149		error = ENOMEM;
150		goto bad;
151	}
152
153	tdbi = (struct tdb_ident *)(mtag + 1);
154	tdbi->dst = saidx->dst;
155	tdbi->proto = saidx->proto;
156	tdbi->spi = sav->spi;
157	m_tag_prepend(m, mtag);
158
159	/*
160	 * If there's another (bundled) SA to apply, do so.
161	 * Note that this puts a burden on the kernel stack size.
162	 * If this is a problem we'll need to introduce a queue
163	 * to set the packet on so we can unwind the stack before
164	 * doing further processing.
165	 */
166	if (isr->next) {
167		IPSECSTAT_INC(ips_out_bundlesa);
168		/* XXX-BZ currently only support same AF bundles. */
169		switch (saidx->dst.sa.sa_family) {
170#ifdef INET
171		case AF_INET:
172			return ipsec4_process_packet(m, isr->next, 0, 0);
173			/* NOTREACHED */
174#endif
175#ifdef notyet
176#ifdef INET6
177		case AF_INET6:
178			/* XXX */
179			ipsec6_output_trans()
180			ipsec6_output_tunnel()
181			/* NOTREACHED */
182#endif /* INET6 */
183#endif
184		default:
185			DPRINTF(("%s: unknown protocol family %u\n", __func__,
186			    saidx->dst.sa.sa_family));
187			error = ENXIO;
188			goto bad;
189		}
190	}
191	key_sa_recordxfer(sav, m);		/* record data transfer */
192
193	m_addr_changed(m);
194
195	/*
196	 * We're done with IPsec processing, transmit the packet using the
197	 * appropriate network protocol (IP or IPv6). SPD lookup will be
198	 * performed again there.
199	 */
200	switch (saidx->dst.sa.sa_family) {
201#ifdef INET
202	struct ip *ip;
203	case AF_INET:
204		ip = mtod(m, struct ip *);
205		ip->ip_len = ntohs(ip->ip_len);
206		ip->ip_off = ntohs(ip->ip_off);
207
208#ifdef IPSEC_NAT_T
209		/*
210		 * If NAT-T is enabled, now that all IPsec processing is done
211		 * insert UDP encapsulation header after IP header.
212		 */
213		if (sav->natt_type) {
214			const int hlen = (ip->ip_hl << 2);
215			int size, off;
216			struct mbuf *mi;
217			struct udphdr *udp;
218
219			size = sizeof(struct udphdr);
220			if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE) {
221				/*
222				 * draft-ietf-ipsec-nat-t-ike-0[01].txt and
223				 * draft-ietf-ipsec-udp-encaps-(00/)01.txt,
224				 * ignoring possible AH mode
225				 * non-IKE marker + non-ESP marker
226				 * from draft-ietf-ipsec-udp-encaps-00.txt.
227				 */
228				size += sizeof(u_int64_t);
229			}
230			mi = m_makespace(m, hlen, size, &off);
231			if (mi == NULL) {
232				DPRINTF(("%s: m_makespace for udphdr failed\n",
233				    __func__));
234				error = ENOBUFS;
235				goto bad;
236			}
237
238			udp = (struct udphdr *)(mtod(mi, caddr_t) + off);
239			if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE)
240				udp->uh_sport = htons(UDP_ENCAP_ESPINUDP_PORT);
241			else
242				udp->uh_sport =
243					KEY_PORTFROMSADDR(&sav->sah->saidx.src);
244			udp->uh_dport = KEY_PORTFROMSADDR(&sav->sah->saidx.dst);
245			udp->uh_sum = 0;
246			udp->uh_ulen = htons(m->m_pkthdr.len - hlen);
247			ip->ip_len = m->m_pkthdr.len;
248			ip->ip_p = IPPROTO_UDP;
249
250			if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE)
251				*(u_int64_t *)(udp + 1) = 0;
252		}
253#endif /* IPSEC_NAT_T */
254
255		return ip_output(m, NULL, NULL, IP_RAWOUTPUT, NULL, NULL);
256#endif /* INET */
257#ifdef INET6
258	case AF_INET6:
259		/*
260		 * We don't need massage, IPv6 header fields are always in
261		 * net endian.
262		 */
263		return ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
264#endif /* INET6 */
265	}
266	panic("ipsec_process_done");
267bad:
268	m_freem(m);
269	return (error);
270}
271
272static struct ipsecrequest *
273ipsec_nextisr(
274	struct mbuf *m,
275	struct ipsecrequest *isr,
276	int af,
277	struct secasindex *saidx,
278	int *error
279)
280{
281#define	IPSEC_OSTAT(name)	do {		\
282	if (isr->saidx.proto == IPPROTO_ESP)	\
283		ESPSTAT_INC(esps_##name);	\
284	else if (isr->saidx.proto == IPPROTO_AH)\
285		AHSTAT_INC(ahs_##name);		\
286	else					\
287		IPCOMPSTAT_INC(ipcomps_##name);	\
288} while (0)
289	struct secasvar *sav;
290
291	IPSECREQUEST_LOCK_ASSERT(isr);
292
293	IPSEC_ASSERT(af == AF_INET || af == AF_INET6,
294		("invalid address family %u", af));
295again:
296	/*
297	 * Craft SA index to search for proper SA.  Note that
298	 * we only fillin unspecified SA peers for transport
299	 * mode; for tunnel mode they must already be filled in.
300	 */
301	*saidx = isr->saidx;
302	if (isr->saidx.mode == IPSEC_MODE_TRANSPORT) {
303		/* Fillin unspecified SA peers only for transport mode */
304		if (af == AF_INET) {
305			struct sockaddr_in *sin;
306			struct ip *ip = mtod(m, struct ip *);
307
308			if (saidx->src.sa.sa_len == 0) {
309				sin = &saidx->src.sin;
310				sin->sin_len = sizeof(*sin);
311				sin->sin_family = AF_INET;
312				sin->sin_port = IPSEC_PORT_ANY;
313				sin->sin_addr = ip->ip_src;
314			}
315			if (saidx->dst.sa.sa_len == 0) {
316				sin = &saidx->dst.sin;
317				sin->sin_len = sizeof(*sin);
318				sin->sin_family = AF_INET;
319				sin->sin_port = IPSEC_PORT_ANY;
320				sin->sin_addr = ip->ip_dst;
321			}
322		} else {
323			struct sockaddr_in6 *sin6;
324			struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
325
326			if (saidx->src.sin6.sin6_len == 0) {
327				sin6 = (struct sockaddr_in6 *)&saidx->src;
328				sin6->sin6_len = sizeof(*sin6);
329				sin6->sin6_family = AF_INET6;
330				sin6->sin6_port = IPSEC_PORT_ANY;
331				sin6->sin6_addr = ip6->ip6_src;
332				if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
333					/* fix scope id for comparing SPD */
334					sin6->sin6_addr.s6_addr16[1] = 0;
335					sin6->sin6_scope_id =
336					    ntohs(ip6->ip6_src.s6_addr16[1]);
337				}
338			}
339			if (saidx->dst.sin6.sin6_len == 0) {
340				sin6 = (struct sockaddr_in6 *)&saidx->dst;
341				sin6->sin6_len = sizeof(*sin6);
342				sin6->sin6_family = AF_INET6;
343				sin6->sin6_port = IPSEC_PORT_ANY;
344				sin6->sin6_addr = ip6->ip6_dst;
345				if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
346					/* fix scope id for comparing SPD */
347					sin6->sin6_addr.s6_addr16[1] = 0;
348					sin6->sin6_scope_id =
349					    ntohs(ip6->ip6_dst.s6_addr16[1]);
350				}
351			}
352		}
353	}
354
355	/*
356	 * Lookup SA and validate it.
357	 */
358	*error = key_checkrequest(isr, saidx);
359	if (*error != 0) {
360		/*
361		 * IPsec processing is required, but no SA found.
362		 * I assume that key_acquire() had been called
363		 * to get/establish the SA. Here I discard
364		 * this packet because it is responsibility for
365		 * upper layer to retransmit the packet.
366		 */
367		IPSECSTAT_INC(ips_out_nosa);
368		goto bad;
369	}
370	sav = isr->sav;
371	if (sav == NULL) {
372		IPSEC_ASSERT(ipsec_get_reqlevel(isr) == IPSEC_LEVEL_USE,
373			("no SA found, but required; level %u",
374			ipsec_get_reqlevel(isr)));
375		IPSECREQUEST_UNLOCK(isr);
376		isr = isr->next;
377		/*
378		 * If isr is NULL, we found a 'use' policy w/o SA.
379		 * Return w/o error and w/o isr so we can drop out
380		 * and continue w/o IPsec processing.
381		 */
382		if (isr == NULL)
383			return isr;
384		IPSECREQUEST_LOCK(isr);
385		goto again;
386	}
387
388	/*
389	 * Check system global policy controls.
390	 */
391	if ((isr->saidx.proto == IPPROTO_ESP && !V_esp_enable) ||
392	    (isr->saidx.proto == IPPROTO_AH && !V_ah_enable) ||
393	    (isr->saidx.proto == IPPROTO_IPCOMP && !V_ipcomp_enable)) {
394		DPRINTF(("%s: IPsec outbound packet dropped due"
395			" to policy (check your sysctls)\n", __func__));
396		IPSEC_OSTAT(pdrops);
397		*error = EHOSTUNREACH;
398		goto bad;
399	}
400
401	/*
402	 * Sanity check the SA contents for the caller
403	 * before they invoke the xform output method.
404	 */
405	if (sav->tdb_xform == NULL) {
406		DPRINTF(("%s: no transform for SA\n", __func__));
407		IPSEC_OSTAT(noxform);
408		*error = EHOSTUNREACH;
409		goto bad;
410	}
411	return isr;
412bad:
413	IPSEC_ASSERT(*error != 0, ("error return w/ no error code"));
414	IPSECREQUEST_UNLOCK(isr);
415	return NULL;
416#undef IPSEC_OSTAT
417}
418
419#ifdef INET
420/*
421 * IPsec output logic for IPv4.
422 */
423int
424ipsec4_process_packet(
425	struct mbuf *m,
426	struct ipsecrequest *isr,
427	int flags,
428	int tunalready)
429{
430	struct secasindex saidx;
431	struct secasvar *sav;
432	struct ip *ip;
433	int error, i, off;
434
435	IPSEC_ASSERT(m != NULL, ("null mbuf"));
436	IPSEC_ASSERT(isr != NULL, ("null isr"));
437
438	IPSECREQUEST_LOCK(isr);		/* insure SA contents don't change */
439
440	isr = ipsec_nextisr(m, isr, AF_INET, &saidx, &error);
441	if (isr == NULL) {
442		if (error != 0)
443			goto bad;
444		return EJUSTRETURN;
445	}
446
447	sav = isr->sav;
448
449#ifdef DEV_ENC
450	encif->if_opackets++;
451	encif->if_obytes += m->m_pkthdr.len;
452
453	/* pass the mbuf to enc0 for bpf processing */
454	ipsec_bpf(m, sav, AF_INET, ENC_OUT|ENC_BEFORE);
455	/* pass the mbuf to enc0 for packet filtering */
456	if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0)
457		goto bad;
458#endif
459
460	if (!tunalready) {
461		union sockaddr_union *dst = &sav->sah->saidx.dst;
462		int setdf;
463
464		/*
465		 * Collect IP_DF state from the outer header.
466		 */
467		if (dst->sa.sa_family == AF_INET) {
468			if (m->m_len < sizeof (struct ip) &&
469			    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
470				error = ENOBUFS;
471				goto bad;
472			}
473			ip = mtod(m, struct ip *);
474			/* Honor system-wide control of how to handle IP_DF */
475			switch (V_ip4_ipsec_dfbit) {
476			case 0:			/* clear in outer header */
477			case 1:			/* set in outer header */
478				setdf = V_ip4_ipsec_dfbit;
479				break;
480			default:		/* propagate to outer header */
481				setdf = ntohs(ip->ip_off & IP_DF);
482				break;
483			}
484		} else {
485			ip = NULL;		/* keep compiler happy */
486			setdf = 0;
487		}
488		/* Do the appropriate encapsulation, if necessary */
489		if (isr->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
490		    dst->sa.sa_family != AF_INET ||	    /* PF mismatch */
491#if 0
492		    (sav->flags & SADB_X_SAFLAGS_TUNNEL) || /* Tunnel requ'd */
493		    sav->tdb_xform->xf_type == XF_IP4 ||    /* ditto */
494#endif
495		    (dst->sa.sa_family == AF_INET &&	    /* Proxy */
496		     dst->sin.sin_addr.s_addr != INADDR_ANY &&
497		     dst->sin.sin_addr.s_addr != ip->ip_dst.s_addr)) {
498			struct mbuf *mp;
499
500			/* Fix IPv4 header checksum and length */
501			if (m->m_len < sizeof (struct ip) &&
502			    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
503				error = ENOBUFS;
504				goto bad;
505			}
506			ip = mtod(m, struct ip *);
507			ip->ip_len = htons(m->m_pkthdr.len);
508			ip->ip_sum = 0;
509			ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
510
511			/* Encapsulate the packet */
512			error = ipip_output(m, isr, &mp, 0, 0);
513			if (mp == NULL && !error) {
514				/* Should never happen. */
515				DPRINTF(("%s: ipip_output returns no mbuf and "
516					"no error!", __func__));
517				error = EFAULT;
518			}
519			if (error) {
520				if (mp) {
521					/* XXX: Should never happen! */
522					m_freem(mp);
523				}
524				m = NULL; /* ipip_output() already freed it */
525				goto bad;
526			}
527			m = mp, mp = NULL;
528			/*
529			 * ipip_output clears IP_DF in the new header.  If
530			 * we need to propagate IP_DF from the outer header,
531			 * then we have to do it here.
532			 *
533			 * XXX shouldn't assume what ipip_output does.
534			 */
535			if (dst->sa.sa_family == AF_INET && setdf) {
536				if (m->m_len < sizeof (struct ip) &&
537				    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
538					error = ENOBUFS;
539					goto bad;
540				}
541				ip = mtod(m, struct ip *);
542				ip->ip_off = ntohs(ip->ip_off);
543				ip->ip_off |= IP_DF;
544				ip->ip_off = htons(ip->ip_off);
545			}
546		}
547	}
548
549#ifdef DEV_ENC
550	/* pass the mbuf to enc0 for bpf processing */
551	ipsec_bpf(m, sav, AF_INET, ENC_OUT|ENC_AFTER);
552	/* pass the mbuf to enc0 for packet filtering */
553	if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0)
554		goto bad;
555#endif
556
557	/*
558	 * Dispatch to the appropriate IPsec transform logic.  The
559	 * packet will be returned for transmission after crypto
560	 * processing, etc. are completed.  For encapsulation we
561	 * bypass this call because of the explicit call done above
562	 * (necessary to deal with IP_DF handling for IPv4).
563	 *
564	 * NB: m & sav are ``passed to caller'' who's reponsible for
565	 *     for reclaiming their resources.
566	 */
567	if (sav->tdb_xform->xf_type != XF_IP4) {
568		ip = mtod(m, struct ip *);
569		i = ip->ip_hl << 2;
570		off = offsetof(struct ip, ip_p);
571		error = (*sav->tdb_xform->xf_output)(m, isr, NULL, i, off);
572	} else {
573		error = ipsec_process_done(m, isr);
574	}
575	IPSECREQUEST_UNLOCK(isr);
576	return error;
577bad:
578	if (isr)
579		IPSECREQUEST_UNLOCK(isr);
580	if (m)
581		m_freem(m);
582	return error;
583}
584#endif
585
586#ifdef INET6
587/*
588 * Chop IP6 header from the payload.
589 */
590static struct mbuf *
591ipsec6_splithdr(struct mbuf *m)
592{
593	struct mbuf *mh;
594	struct ip6_hdr *ip6;
595	int hlen;
596
597	IPSEC_ASSERT(m->m_len >= sizeof (struct ip6_hdr),
598		("first mbuf too short, len %u", m->m_len));
599	ip6 = mtod(m, struct ip6_hdr *);
600	hlen = sizeof(struct ip6_hdr);
601	if (m->m_len > hlen) {
602		MGETHDR(mh, M_DONTWAIT, MT_DATA);
603		if (!mh) {
604			m_freem(m);
605			return NULL;
606		}
607		M_MOVE_PKTHDR(mh, m);
608		MH_ALIGN(mh, hlen);
609		m->m_len -= hlen;
610		m->m_data += hlen;
611		mh->m_next = m;
612		m = mh;
613		m->m_len = hlen;
614		bcopy((caddr_t)ip6, mtod(m, caddr_t), hlen);
615	} else if (m->m_len < hlen) {
616		m = m_pullup(m, hlen);
617		if (!m)
618			return NULL;
619	}
620	return m;
621}
622
623/*
624 * IPsec output logic for IPv6, transport mode.
625 */
626int
627ipsec6_output_trans(
628	struct ipsec_output_state *state,
629	u_char *nexthdrp,
630	struct mbuf *mprev,
631	struct secpolicy *sp,
632	int flags,
633	int *tun)
634{
635	struct ipsecrequest *isr;
636	struct secasindex saidx;
637	int error = 0;
638	struct mbuf *m;
639
640	IPSEC_ASSERT(state != NULL, ("null state"));
641	IPSEC_ASSERT(state->m != NULL, ("null m"));
642	IPSEC_ASSERT(nexthdrp != NULL, ("null nexthdrp"));
643	IPSEC_ASSERT(mprev != NULL, ("null mprev"));
644	IPSEC_ASSERT(sp != NULL, ("null sp"));
645	IPSEC_ASSERT(tun != NULL, ("null tun"));
646
647	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
648		printf("%s: applied SP\n", __func__);
649		kdebug_secpolicy(sp));
650
651	isr = sp->req;
652	if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
653		/* the rest will be handled by ipsec6_output_tunnel() */
654		*tun = 1;		/* need tunnel-mode processing */
655		return 0;
656	}
657
658	*tun = 0;
659	m = state->m;
660
661	IPSECREQUEST_LOCK(isr);		/* insure SA contents don't change */
662	isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
663	if (isr == NULL) {
664		if (error != 0) {
665#ifdef notdef
666			/* XXX should notification be done for all errors ? */
667			/*
668			 * Notify the fact that the packet is discarded
669			 * to ourselves. I believe this is better than
670			 * just silently discarding. (jinmei@kame.net)
671			 * XXX: should we restrict the error to TCP packets?
672			 * XXX: should we directly notify sockets via
673			 *      pfctlinputs?
674			 */
675			icmp6_error(m, ICMP6_DST_UNREACH,
676				    ICMP6_DST_UNREACH_ADMIN, 0);
677			m = NULL;	/* NB: icmp6_error frees mbuf */
678#endif
679			goto bad;
680		}
681		return EJUSTRETURN;
682	}
683
684	error = (*isr->sav->tdb_xform->xf_output)(m, isr, NULL,
685						  sizeof (struct ip6_hdr),
686						  offsetof(struct ip6_hdr,
687							   ip6_nxt));
688	IPSECREQUEST_UNLOCK(isr);
689	return error;
690bad:
691	if (isr)
692		IPSECREQUEST_UNLOCK(isr);
693	if (m)
694		m_freem(m);
695	state->m = NULL;
696	return error;
697}
698
699static int
700ipsec6_encapsulate(struct mbuf *m, struct secasvar *sav)
701{
702	struct ip6_hdr *oip6;
703	struct ip6_hdr *ip6;
704	size_t plen;
705
706	/* can't tunnel between different AFs */
707	if (sav->sah->saidx.src.sa.sa_family != AF_INET6 ||
708	    sav->sah->saidx.dst.sa.sa_family != AF_INET6) {
709		m_freem(m);
710		return EINVAL;
711	}
712	IPSEC_ASSERT(m->m_len == sizeof (struct ip6_hdr),
713		("mbuf wrong size; len %u", m->m_len));
714
715
716	/*
717	 * grow the mbuf to accomodate the new IPv6 header.
718	 */
719	plen = m->m_pkthdr.len;
720	if (M_LEADINGSPACE(m->m_next) < sizeof(struct ip6_hdr)) {
721		struct mbuf *n;
722		MGET(n, M_DONTWAIT, MT_DATA);
723		if (!n) {
724			m_freem(m);
725			return ENOBUFS;
726		}
727		n->m_len = sizeof(struct ip6_hdr);
728		n->m_next = m->m_next;
729		m->m_next = n;
730		m->m_pkthdr.len += sizeof(struct ip6_hdr);
731		oip6 = mtod(n, struct ip6_hdr *);
732	} else {
733		m->m_next->m_len += sizeof(struct ip6_hdr);
734		m->m_next->m_data -= sizeof(struct ip6_hdr);
735		m->m_pkthdr.len += sizeof(struct ip6_hdr);
736		oip6 = mtod(m->m_next, struct ip6_hdr *);
737	}
738	ip6 = mtod(m, struct ip6_hdr *);
739	bcopy((caddr_t)ip6, (caddr_t)oip6, sizeof(struct ip6_hdr));
740
741	/* Fake link-local scope-class addresses */
742	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
743		oip6->ip6_src.s6_addr16[1] = 0;
744	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
745		oip6->ip6_dst.s6_addr16[1] = 0;
746
747	/* construct new IPv6 header. see RFC 2401 5.1.2.2 */
748	/* ECN consideration. */
749	ip6_ecn_ingress(V_ip6_ipsec_ecn, &ip6->ip6_flow, &oip6->ip6_flow);
750	if (plen < IPV6_MAXPACKET - sizeof(struct ip6_hdr))
751		ip6->ip6_plen = htons(plen);
752	else {
753		/* ip6->ip6_plen will be updated in ip6_output() */
754	}
755	ip6->ip6_nxt = IPPROTO_IPV6;
756	ip6->ip6_src = sav->sah->saidx.src.sin6.sin6_addr;
757	ip6->ip6_dst = sav->sah->saidx.dst.sin6.sin6_addr;
758	ip6->ip6_hlim = IPV6_DEFHLIM;
759
760	/* XXX Should ip6_src be updated later ? */
761
762	return 0;
763}
764
765/*
766 * IPsec output logic for IPv6, tunnel mode.
767 */
768int
769ipsec6_output_tunnel(struct ipsec_output_state *state, struct secpolicy *sp, int flags)
770{
771	struct ip6_hdr *ip6;
772	struct ipsecrequest *isr;
773	struct secasindex saidx;
774	int error;
775	struct sockaddr_in6 *dst6;
776	struct mbuf *m;
777
778	IPSEC_ASSERT(state != NULL, ("null state"));
779	IPSEC_ASSERT(state->m != NULL, ("null m"));
780	IPSEC_ASSERT(sp != NULL, ("null sp"));
781
782	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
783		printf("%s: applied SP\n", __func__);
784		kdebug_secpolicy(sp));
785
786	m = state->m;
787	/*
788	 * transport mode ipsec (before the 1st tunnel mode) is already
789	 * processed by ipsec6_output_trans().
790	 */
791	for (isr = sp->req; isr; isr = isr->next) {
792		if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
793			break;
794	}
795
796	IPSECREQUEST_LOCK(isr);		/* insure SA contents don't change */
797	isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
798	if (isr == NULL) {
799		if (error != 0)
800			goto bad;
801		return EJUSTRETURN;
802	}
803
804#ifdef DEV_ENC
805	encif->if_opackets++;
806	encif->if_obytes += m->m_pkthdr.len;
807
808	/* pass the mbuf to enc0 for bpf processing */
809	ipsec_bpf(m, isr->sav, AF_INET6, ENC_OUT|ENC_BEFORE);
810	/* pass the mbuf to enc0 for packet filtering */
811	if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0)
812		goto bad;
813#endif
814
815	/*
816	 * There may be the case that SA status will be changed when
817	 * we are refering to one. So calling splsoftnet().
818	 */
819	if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
820		/*
821		 * build IPsec tunnel.
822		 */
823		/* XXX should be processed with other familiy */
824		if (isr->sav->sah->saidx.src.sa.sa_family != AF_INET6) {
825			ipseclog((LOG_ERR, "%s: family mismatched between "
826			    "inner and outer, spi=%u\n", __func__,
827			    ntohl(isr->sav->spi)));
828			IPSEC6STAT_INC(ips_out_inval);
829			error = EAFNOSUPPORT;
830			goto bad;
831		}
832
833		m = ipsec6_splithdr(m);
834		if (!m) {
835			IPSEC6STAT_INC(ips_out_nomem);
836			error = ENOMEM;
837			goto bad;
838		}
839		error = ipsec6_encapsulate(m, isr->sav);
840		if (error) {
841			m = NULL;
842			goto bad;
843		}
844		ip6 = mtod(m, struct ip6_hdr *);
845
846		state->ro =
847		    (struct route *)&isr->sav->sah->route_cache.sin6_route;
848		state->dst = (struct sockaddr *)&state->ro->ro_dst;
849		dst6 = (struct sockaddr_in6 *)state->dst;
850		if (state->ro->ro_rt
851		 && ((state->ro->ro_rt->rt_flags & RTF_UP) == 0
852		  || !IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr, &ip6->ip6_dst))) {
853			RTFREE(state->ro->ro_rt);
854			state->ro->ro_rt = NULL;
855		}
856		if (state->ro->ro_rt == NULL) {
857			bzero(dst6, sizeof(*dst6));
858			dst6->sin6_family = AF_INET6;
859			dst6->sin6_len = sizeof(*dst6);
860			dst6->sin6_addr = ip6->ip6_dst;
861			rtalloc_ign_fib(state->ro, 0UL, M_GETFIB(m));
862		}
863		if (state->ro->ro_rt == NULL) {
864			IP6STAT_INC(ip6s_noroute);
865			IPSEC6STAT_INC(ips_out_noroute);
866			error = EHOSTUNREACH;
867			goto bad;
868		}
869
870		/* adjust state->dst if tunnel endpoint is offlink */
871		if (state->ro->ro_rt->rt_flags & RTF_GATEWAY)
872			state->dst = (struct sockaddr *)state->ro->ro_rt->rt_gateway;
873	}
874
875	m = ipsec6_splithdr(m);
876	if (!m) {
877		IPSEC6STAT_INC(ips_out_nomem);
878		error = ENOMEM;
879		goto bad;
880	}
881	ip6 = mtod(m, struct ip6_hdr *);
882
883#ifdef DEV_ENC
884	/* pass the mbuf to enc0 for bpf processing */
885	ipsec_bpf(m, isr->sav, AF_INET6, ENC_OUT|ENC_AFTER);
886	/* pass the mbuf to enc0 for packet filtering */
887	if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0)
888		goto bad;
889#endif
890
891	error = (*isr->sav->tdb_xform->xf_output)(m, isr, NULL,
892		sizeof (struct ip6_hdr),
893		offsetof(struct ip6_hdr, ip6_nxt));
894	IPSECREQUEST_UNLOCK(isr);
895	return error;
896bad:
897	if (isr)
898		IPSECREQUEST_UNLOCK(isr);
899	if (m)
900		m_freem(m);
901	state->m = NULL;
902	return error;
903}
904#endif /*INET6*/
905