ipsec_output.c revision 177175
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: head/sys/netipsec/ipsec_output.c 177175 2008-03-14 16:38:11Z bz $
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
50#include <netinet/in.h>
51#include <netinet/in_systm.h>
52#include <netinet/ip.h>
53#include <netinet/ip_var.h>
54#include <netinet/in_var.h>
55#include <netinet/ip_ecn.h>
56#ifdef INET6
57#include <netinet6/ip6_ecn.h>
58#endif
59
60#include <netinet/ip6.h>
61#ifdef INET6
62#include <netinet6/ip6_var.h>
63#endif
64#include <netinet/in_pcb.h>
65#ifdef INET6
66#include <netinet/icmp6.h>
67#endif
68
69#include <netipsec/ipsec.h>
70#ifdef INET6
71#include <netipsec/ipsec6.h>
72#endif
73#include <netipsec/ah_var.h>
74#include <netipsec/esp_var.h>
75#include <netipsec/ipcomp_var.h>
76
77#include <netipsec/xform.h>
78
79#include <netipsec/key.h>
80#include <netipsec/keydb.h>
81#include <netipsec/key_debug.h>
82
83#include <machine/in_cksum.h>
84
85int
86ipsec_process_done(struct mbuf *m, struct ipsecrequest *isr)
87{
88	struct tdb_ident *tdbi;
89	struct m_tag *mtag;
90	struct secasvar *sav;
91	struct secasindex *saidx;
92	int error;
93
94	IPSEC_SPLASSERT_SOFTNET(__func__);
95
96	IPSEC_ASSERT(m != NULL, ("null mbuf"));
97	IPSEC_ASSERT(isr != NULL, ("null ISR"));
98	sav = isr->sav;
99	IPSEC_ASSERT(sav != NULL, ("null SA"));
100	IPSEC_ASSERT(sav->sah != NULL, ("null SAH"));
101
102	saidx = &sav->sah->saidx;
103	switch (saidx->dst.sa.sa_family) {
104#ifdef INET
105	case AF_INET:
106		/* Fix the header length, for AH processing. */
107		mtod(m, struct ip *)->ip_len = htons(m->m_pkthdr.len);
108		break;
109#endif /* INET */
110#ifdef INET6
111	case AF_INET6:
112		/* Fix the header length, for AH processing. */
113		if (m->m_pkthdr.len < sizeof (struct ip6_hdr)) {
114			error = ENXIO;
115			goto bad;
116		}
117		if (m->m_pkthdr.len - sizeof (struct ip6_hdr) > IPV6_MAXPACKET) {
118			/* No jumbogram support. */
119			error = ENXIO;	/*?*/
120			goto bad;
121		}
122		mtod(m, struct ip6_hdr *)->ip6_plen =
123			htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
124		break;
125#endif /* INET6 */
126	default:
127		DPRINTF(("%s: unknown protocol family %u\n", __func__,
128		    saidx->dst.sa.sa_family));
129		error = ENXIO;
130		goto bad;
131	}
132
133	/*
134	 * Add a record of what we've done or what needs to be done to the
135	 * packet.
136	 */
137	mtag = m_tag_get(PACKET_TAG_IPSEC_OUT_DONE,
138			sizeof(struct tdb_ident), M_NOWAIT);
139	if (mtag == NULL) {
140		DPRINTF(("%s: could not get packet tag\n", __func__));
141		error = ENOMEM;
142		goto bad;
143	}
144
145	tdbi = (struct tdb_ident *)(mtag + 1);
146	tdbi->dst = saidx->dst;
147	tdbi->proto = saidx->proto;
148	tdbi->spi = sav->spi;
149	m_tag_prepend(m, mtag);
150
151	/*
152	 * If there's another (bundled) SA to apply, do so.
153	 * Note that this puts a burden on the kernel stack size.
154	 * If this is a problem we'll need to introduce a queue
155	 * to set the packet on so we can unwind the stack before
156	 * doing further processing.
157	 */
158	if (isr->next) {
159		ipsec4stat.ips_out_bundlesa++;
160		return ipsec4_process_packet(m, isr->next, 0, 0);
161	}
162	key_sa_recordxfer(sav, m);		/* record data transfer */
163
164	/*
165	 * We're done with IPsec processing, transmit the packet using the
166	 * appropriate network protocol (IP or IPv6). SPD lookup will be
167	 * performed again there.
168	 */
169	switch (saidx->dst.sa.sa_family) {
170#ifdef INET
171	struct ip *ip;
172	case AF_INET:
173		ip = mtod(m, struct ip *);
174		ip->ip_len = ntohs(ip->ip_len);
175		ip->ip_off = ntohs(ip->ip_off);
176
177		return ip_output(m, NULL, NULL, IP_RAWOUTPUT, NULL, NULL);
178#endif /* INET */
179#ifdef INET6
180	case AF_INET6:
181		/*
182		 * We don't need massage, IPv6 header fields are always in
183		 * net endian.
184		 */
185		return ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
186#endif /* INET6 */
187	}
188	panic("ipsec_process_done");
189bad:
190	m_freem(m);
191	KEY_FREESAV(&sav);
192	return (error);
193}
194
195static struct ipsecrequest *
196ipsec_nextisr(
197	struct mbuf *m,
198	struct ipsecrequest *isr,
199	int af,
200	struct secasindex *saidx,
201	int *error
202)
203{
204#define IPSEC_OSTAT(x,y,z) (isr->saidx.proto == IPPROTO_ESP ? (x)++ : \
205			    isr->saidx.proto == IPPROTO_AH ? (y)++ : (z)++)
206	struct secasvar *sav;
207
208	IPSEC_SPLASSERT_SOFTNET(__func__);
209	IPSECREQUEST_LOCK_ASSERT(isr);
210
211	IPSEC_ASSERT(af == AF_INET || af == AF_INET6,
212		("invalid address family %u", af));
213again:
214	/*
215	 * Craft SA index to search for proper SA.  Note that
216	 * we only fillin unspecified SA peers for transport
217	 * mode; for tunnel mode they must already be filled in.
218	 */
219	*saidx = isr->saidx;
220	if (isr->saidx.mode == IPSEC_MODE_TRANSPORT) {
221		/* Fillin unspecified SA peers only for transport mode */
222		if (af == AF_INET) {
223			struct sockaddr_in *sin;
224			struct ip *ip = mtod(m, struct ip *);
225
226			if (saidx->src.sa.sa_len == 0) {
227				sin = &saidx->src.sin;
228				sin->sin_len = sizeof(*sin);
229				sin->sin_family = AF_INET;
230				sin->sin_port = IPSEC_PORT_ANY;
231				sin->sin_addr = ip->ip_src;
232			}
233			if (saidx->dst.sa.sa_len == 0) {
234				sin = &saidx->dst.sin;
235				sin->sin_len = sizeof(*sin);
236				sin->sin_family = AF_INET;
237				sin->sin_port = IPSEC_PORT_ANY;
238				sin->sin_addr = ip->ip_dst;
239			}
240		} else {
241			struct sockaddr_in6 *sin6;
242			struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
243
244			if (saidx->src.sin6.sin6_len == 0) {
245				sin6 = (struct sockaddr_in6 *)&saidx->src;
246				sin6->sin6_len = sizeof(*sin6);
247				sin6->sin6_family = AF_INET6;
248				sin6->sin6_port = IPSEC_PORT_ANY;
249				sin6->sin6_addr = ip6->ip6_src;
250				if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
251					/* fix scope id for comparing SPD */
252					sin6->sin6_addr.s6_addr16[1] = 0;
253					sin6->sin6_scope_id =
254					    ntohs(ip6->ip6_src.s6_addr16[1]);
255				}
256			}
257			if (saidx->dst.sin6.sin6_len == 0) {
258				sin6 = (struct sockaddr_in6 *)&saidx->dst;
259				sin6->sin6_len = sizeof(*sin6);
260				sin6->sin6_family = AF_INET6;
261				sin6->sin6_port = IPSEC_PORT_ANY;
262				sin6->sin6_addr = ip6->ip6_dst;
263				if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
264					/* fix scope id for comparing SPD */
265					sin6->sin6_addr.s6_addr16[1] = 0;
266					sin6->sin6_scope_id =
267					    ntohs(ip6->ip6_dst.s6_addr16[1]);
268				}
269			}
270		}
271	}
272
273	/*
274	 * Lookup SA and validate it.
275	 */
276	*error = key_checkrequest(isr, saidx);
277	if (*error != 0) {
278		/*
279		 * IPsec processing is required, but no SA found.
280		 * I assume that key_acquire() had been called
281		 * to get/establish the SA. Here I discard
282		 * this packet because it is responsibility for
283		 * upper layer to retransmit the packet.
284		 */
285		ipsec4stat.ips_out_nosa++;
286		goto bad;
287	}
288	sav = isr->sav;
289	if (sav == NULL) {
290		IPSEC_ASSERT(ipsec_get_reqlevel(isr) == IPSEC_LEVEL_USE,
291			("no SA found, but required; level %u",
292			ipsec_get_reqlevel(isr)));
293		IPSECREQUEST_UNLOCK(isr);
294		isr = isr->next;
295		/*
296		 * If isr is NULL, we found a 'use' policy w/o SA.
297		 * Return w/o error and w/o isr so we can drop out
298		 * and continue w/o IPsec processing.
299		 */
300		if (isr == NULL)
301			return isr;
302		IPSECREQUEST_LOCK(isr);
303		goto again;
304	}
305
306	/*
307	 * Check system global policy controls.
308	 */
309	if ((isr->saidx.proto == IPPROTO_ESP && !esp_enable) ||
310	    (isr->saidx.proto == IPPROTO_AH && !ah_enable) ||
311	    (isr->saidx.proto == IPPROTO_IPCOMP && !ipcomp_enable)) {
312		DPRINTF(("%s: IPsec outbound packet dropped due"
313			" to policy (check your sysctls)\n", __func__));
314		IPSEC_OSTAT(espstat.esps_pdrops, ahstat.ahs_pdrops,
315		    ipcompstat.ipcomps_pdrops);
316		*error = EHOSTUNREACH;
317		goto bad;
318	}
319
320	/*
321	 * Sanity check the SA contents for the caller
322	 * before they invoke the xform output method.
323	 */
324	if (sav->tdb_xform == NULL) {
325		DPRINTF(("%s: no transform for SA\n", __func__));
326		IPSEC_OSTAT(espstat.esps_noxform, ahstat.ahs_noxform,
327		    ipcompstat.ipcomps_noxform);
328		*error = EHOSTUNREACH;
329		goto bad;
330	}
331	return isr;
332bad:
333	IPSEC_ASSERT(*error != 0, ("error return w/ no error code"));
334	IPSECREQUEST_UNLOCK(isr);
335	return NULL;
336#undef IPSEC_OSTAT
337}
338
339#ifdef INET
340/*
341 * IPsec output logic for IPv4.
342 */
343int
344ipsec4_process_packet(
345	struct mbuf *m,
346	struct ipsecrequest *isr,
347	int flags,
348	int tunalready)
349{
350	struct secasindex saidx;
351	struct secasvar *sav;
352	struct ip *ip;
353	int error, i, off;
354
355	IPSEC_ASSERT(m != NULL, ("null mbuf"));
356	IPSEC_ASSERT(isr != NULL, ("null isr"));
357
358	IPSECREQUEST_LOCK(isr);		/* insure SA contents don't change */
359
360	isr = ipsec_nextisr(m, isr, AF_INET, &saidx, &error);
361	if (isr == NULL) {
362		if (error != 0)
363			goto bad;
364		return EJUSTRETURN;
365	}
366
367	sav = isr->sav;
368
369#ifdef DEV_ENC
370	/* pass the mbuf to enc0 for bpf processing */
371	ipsec_bpf(m, sav, AF_INET, ENC_OUT|ENC_BEFORE);
372	/* pass the mbuf to enc0 for packet filtering */
373	if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0)
374		goto bad;
375#endif
376
377	if (!tunalready) {
378		union sockaddr_union *dst = &sav->sah->saidx.dst;
379		int setdf;
380
381		/*
382		 * Collect IP_DF state from the outer header.
383		 */
384		if (dst->sa.sa_family == AF_INET) {
385			if (m->m_len < sizeof (struct ip) &&
386			    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
387				error = ENOBUFS;
388				goto bad;
389			}
390			ip = mtod(m, struct ip *);
391			/* Honor system-wide control of how to handle IP_DF */
392			switch (ip4_ipsec_dfbit) {
393			case 0:			/* clear in outer header */
394			case 1:			/* set in outer header */
395				setdf = ip4_ipsec_dfbit;
396				break;
397			default:		/* propagate to outer header */
398				setdf = ntohs(ip->ip_off & IP_DF);
399				break;
400			}
401		} else {
402			ip = NULL;		/* keep compiler happy */
403			setdf = 0;
404		}
405		/* Do the appropriate encapsulation, if necessary */
406		if (isr->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
407		    dst->sa.sa_family != AF_INET ||	    /* PF mismatch */
408#if 0
409		    (sav->flags & SADB_X_SAFLAGS_TUNNEL) || /* Tunnel requ'd */
410		    sav->tdb_xform->xf_type == XF_IP4 ||    /* ditto */
411#endif
412		    (dst->sa.sa_family == AF_INET &&	    /* Proxy */
413		     dst->sin.sin_addr.s_addr != INADDR_ANY &&
414		     dst->sin.sin_addr.s_addr != ip->ip_dst.s_addr)) {
415			struct mbuf *mp;
416
417			/* Fix IPv4 header checksum and length */
418			if (m->m_len < sizeof (struct ip) &&
419			    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
420				error = ENOBUFS;
421				goto bad;
422			}
423			ip = mtod(m, struct ip *);
424			ip->ip_len = htons(m->m_pkthdr.len);
425			ip->ip_sum = 0;
426#ifdef _IP_VHL
427			if (ip->ip_vhl == IP_VHL_BORING)
428				ip->ip_sum = in_cksum_hdr(ip);
429			else
430				ip->ip_sum = in_cksum(m,
431					_IP_VHL_HL(ip->ip_vhl) << 2);
432#else
433			ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
434#endif
435
436			/* Encapsulate the packet */
437			error = ipip_output(m, isr, &mp, 0, 0);
438			if (mp == NULL && !error) {
439				/* Should never happen. */
440				DPRINTF(("%s: ipip_output returns no mbuf and "
441					"no error!", __func__));
442				error = EFAULT;
443			}
444			if (error) {
445				if (mp) {
446					/* XXX: Should never happen! */
447					m_freem(mp);
448				}
449				m = NULL; /* ipip_output() already freed it */
450				goto bad;
451			}
452			m = mp, mp = NULL;
453			/*
454			 * ipip_output clears IP_DF in the new header.  If
455			 * we need to propagate IP_DF from the outer header,
456			 * then we have to do it here.
457			 *
458			 * XXX shouldn't assume what ipip_output does.
459			 */
460			if (dst->sa.sa_family == AF_INET && setdf) {
461				if (m->m_len < sizeof (struct ip) &&
462				    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
463					error = ENOBUFS;
464					goto bad;
465				}
466				ip = mtod(m, struct ip *);
467				ip->ip_off = ntohs(ip->ip_off);
468				ip->ip_off |= IP_DF;
469				ip->ip_off = htons(ip->ip_off);
470			}
471		}
472	}
473
474#ifdef DEV_ENC
475	/* pass the mbuf to enc0 for bpf processing */
476	ipsec_bpf(m, sav, AF_INET, ENC_OUT|ENC_AFTER);
477	/* pass the mbuf to enc0 for packet filtering */
478	if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0)
479		goto bad;
480#endif
481
482	/*
483	 * Dispatch to the appropriate IPsec transform logic.  The
484	 * packet will be returned for transmission after crypto
485	 * processing, etc. are completed.  For encapsulation we
486	 * bypass this call because of the explicit call done above
487	 * (necessary to deal with IP_DF handling for IPv4).
488	 *
489	 * NB: m & sav are ``passed to caller'' who's reponsible for
490	 *     for reclaiming their resources.
491	 */
492	if (sav->tdb_xform->xf_type != XF_IP4) {
493		ip = mtod(m, struct ip *);
494		i = ip->ip_hl << 2;
495		off = offsetof(struct ip, ip_p);
496		error = (*sav->tdb_xform->xf_output)(m, isr, NULL, i, off);
497	} else {
498		error = ipsec_process_done(m, isr);
499	}
500	IPSECREQUEST_UNLOCK(isr);
501	return error;
502bad:
503	if (isr)
504		IPSECREQUEST_UNLOCK(isr);
505	if (m)
506		m_freem(m);
507	return error;
508}
509#endif
510
511#ifdef INET6
512/*
513 * Chop IP6 header from the payload.
514 */
515static struct mbuf *
516ipsec6_splithdr(struct mbuf *m)
517{
518	struct mbuf *mh;
519	struct ip6_hdr *ip6;
520	int hlen;
521
522	IPSEC_ASSERT(m->m_len >= sizeof (struct ip6_hdr),
523		("first mbuf too short, len %u", m->m_len));
524	ip6 = mtod(m, struct ip6_hdr *);
525	hlen = sizeof(struct ip6_hdr);
526	if (m->m_len > hlen) {
527		MGETHDR(mh, M_DONTWAIT, MT_DATA);
528		if (!mh) {
529			m_freem(m);
530			return NULL;
531		}
532		M_MOVE_PKTHDR(mh, m);
533		MH_ALIGN(mh, hlen);
534		m->m_len -= hlen;
535		m->m_data += hlen;
536		mh->m_next = m;
537		m = mh;
538		m->m_len = hlen;
539		bcopy((caddr_t)ip6, mtod(m, caddr_t), hlen);
540	} else if (m->m_len < hlen) {
541		m = m_pullup(m, hlen);
542		if (!m)
543			return NULL;
544	}
545	return m;
546}
547
548/*
549 * IPsec output logic for IPv6, transport mode.
550 */
551int
552ipsec6_output_trans(
553	struct ipsec_output_state *state,
554	u_char *nexthdrp,
555	struct mbuf *mprev,
556	struct secpolicy *sp,
557	int flags,
558	int *tun)
559{
560	struct ipsecrequest *isr;
561	struct secasindex saidx;
562	int error = 0;
563	struct mbuf *m;
564
565	IPSEC_ASSERT(state != NULL, ("null state"));
566	IPSEC_ASSERT(state->m != NULL, ("null m"));
567	IPSEC_ASSERT(nexthdrp != NULL, ("null nexthdrp"));
568	IPSEC_ASSERT(mprev != NULL, ("null mprev"));
569	IPSEC_ASSERT(sp != NULL, ("null sp"));
570	IPSEC_ASSERT(tun != NULL, ("null tun"));
571
572	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
573		printf("%s: applied SP\n", __func__);
574		kdebug_secpolicy(sp));
575
576	isr = sp->req;
577	if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
578		/* the rest will be handled by ipsec6_output_tunnel() */
579		*tun = 1;		/* need tunnel-mode processing */
580		return 0;
581	}
582
583	*tun = 0;
584	m = state->m;
585
586	IPSECREQUEST_LOCK(isr);		/* insure SA contents don't change */
587	isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
588	if (isr == NULL) {
589		if (error != 0) {
590#ifdef notdef
591			/* XXX should notification be done for all errors ? */
592			/*
593			 * Notify the fact that the packet is discarded
594			 * to ourselves. I believe this is better than
595			 * just silently discarding. (jinmei@kame.net)
596			 * XXX: should we restrict the error to TCP packets?
597			 * XXX: should we directly notify sockets via
598			 *      pfctlinputs?
599			 */
600			icmp6_error(m, ICMP6_DST_UNREACH,
601				    ICMP6_DST_UNREACH_ADMIN, 0);
602			m = NULL;	/* NB: icmp6_error frees mbuf */
603#endif
604			goto bad;
605		}
606		return EJUSTRETURN;
607	}
608
609	error = (*isr->sav->tdb_xform->xf_output)(m, isr, NULL,
610						  sizeof (struct ip6_hdr),
611						  offsetof(struct ip6_hdr,
612							   ip6_nxt));
613	IPSECREQUEST_UNLOCK(isr);
614	return error;
615bad:
616	if (isr)
617		IPSECREQUEST_UNLOCK(isr);
618	if (m)
619		m_freem(m);
620	state->m = NULL;
621	return error;
622}
623
624static int
625ipsec6_encapsulate(struct mbuf *m, struct secasvar *sav)
626{
627	struct ip6_hdr *oip6;
628	struct ip6_hdr *ip6;
629	size_t plen;
630
631	/* can't tunnel between different AFs */
632	if (sav->sah->saidx.src.sa.sa_family != AF_INET6 ||
633	    sav->sah->saidx.dst.sa.sa_family != AF_INET6) {
634		m_freem(m);
635		return EINVAL;
636	}
637	IPSEC_ASSERT(m->m_len == sizeof (struct ip6_hdr),
638		("mbuf wrong size; len %u", m->m_len));
639
640
641	/*
642	 * grow the mbuf to accomodate the new IPv6 header.
643	 */
644	plen = m->m_pkthdr.len;
645	if (M_LEADINGSPACE(m->m_next) < sizeof(struct ip6_hdr)) {
646		struct mbuf *n;
647		MGET(n, M_DONTWAIT, MT_DATA);
648		if (!n) {
649			m_freem(m);
650			return ENOBUFS;
651		}
652		n->m_len = sizeof(struct ip6_hdr);
653		n->m_next = m->m_next;
654		m->m_next = n;
655		m->m_pkthdr.len += sizeof(struct ip6_hdr);
656		oip6 = mtod(n, struct ip6_hdr *);
657	} else {
658		m->m_next->m_len += sizeof(struct ip6_hdr);
659		m->m_next->m_data -= sizeof(struct ip6_hdr);
660		m->m_pkthdr.len += sizeof(struct ip6_hdr);
661		oip6 = mtod(m->m_next, struct ip6_hdr *);
662	}
663	ip6 = mtod(m, struct ip6_hdr *);
664	bcopy((caddr_t)ip6, (caddr_t)oip6, sizeof(struct ip6_hdr));
665
666	/* Fake link-local scope-class addresses */
667	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
668		oip6->ip6_src.s6_addr16[1] = 0;
669	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
670		oip6->ip6_dst.s6_addr16[1] = 0;
671
672	/* construct new IPv6 header. see RFC 2401 5.1.2.2 */
673	/* ECN consideration. */
674	ip6_ecn_ingress(ip6_ipsec_ecn, &ip6->ip6_flow, &oip6->ip6_flow);
675	if (plen < IPV6_MAXPACKET - sizeof(struct ip6_hdr))
676		ip6->ip6_plen = htons(plen);
677	else {
678		/* ip6->ip6_plen will be updated in ip6_output() */
679	}
680	ip6->ip6_nxt = IPPROTO_IPV6;
681	ip6->ip6_src = sav->sah->saidx.src.sin6.sin6_addr;
682	ip6->ip6_dst = sav->sah->saidx.dst.sin6.sin6_addr;
683	ip6->ip6_hlim = IPV6_DEFHLIM;
684
685	/* XXX Should ip6_src be updated later ? */
686
687	return 0;
688}
689
690/*
691 * IPsec output logic for IPv6, tunnel mode.
692 */
693int
694ipsec6_output_tunnel(struct ipsec_output_state *state, struct secpolicy *sp, int flags)
695{
696	struct ip6_hdr *ip6;
697	struct ipsecrequest *isr;
698	struct secasindex saidx;
699	int error;
700	struct sockaddr_in6* dst6;
701	struct mbuf *m;
702
703	IPSEC_ASSERT(state != NULL, ("null state"));
704	IPSEC_ASSERT(state->m != NULL, ("null m"));
705	IPSEC_ASSERT(sp != NULL, ("null sp"));
706
707	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
708		printf("%s: applied SP\n", __func__);
709		kdebug_secpolicy(sp));
710
711	m = state->m;
712	/*
713	 * transport mode ipsec (before the 1st tunnel mode) is already
714	 * processed by ipsec6_output_trans().
715	 */
716	for (isr = sp->req; isr; isr = isr->next) {
717		if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
718			break;
719	}
720
721	IPSECREQUEST_LOCK(isr);		/* insure SA contents don't change */
722	isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
723	if (isr == NULL) {
724		if (error != 0)
725			goto bad;
726		return EJUSTRETURN;
727	}
728
729#ifdef DEV_ENC
730	/* pass the mbuf to enc0 for bpf processing */
731	ipsec_bpf(m, isr->sav, AF_INET6, ENC_OUT|ENC_BEFORE);
732	/* pass the mbuf to enc0 for packet filtering */
733	if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0)
734		goto bad;
735#endif
736
737	/*
738	 * There may be the case that SA status will be changed when
739	 * we are refering to one. So calling splsoftnet().
740	 */
741	if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
742		/*
743		 * build IPsec tunnel.
744		 */
745		/* XXX should be processed with other familiy */
746		if (isr->sav->sah->saidx.src.sa.sa_family != AF_INET6) {
747			ipseclog((LOG_ERR, "%s: family mismatched between "
748			    "inner and outer, spi=%u\n", __func__,
749			    ntohl(isr->sav->spi)));
750			ipsec6stat.ips_out_inval++;
751			error = EAFNOSUPPORT;
752			goto bad;
753		}
754
755		m = ipsec6_splithdr(m);
756		if (!m) {
757			ipsec6stat.ips_out_nomem++;
758			error = ENOMEM;
759			goto bad;
760		}
761		error = ipsec6_encapsulate(m, isr->sav);
762		if (error) {
763			m = NULL;
764			goto bad;
765		}
766		ip6 = mtod(m, struct ip6_hdr *);
767
768		state->ro = &isr->sav->sah->sa_route;
769		state->dst = (struct sockaddr *)&state->ro->ro_dst;
770		dst6 = (struct sockaddr_in6 *)state->dst;
771		if (state->ro->ro_rt
772		 && ((state->ro->ro_rt->rt_flags & RTF_UP) == 0
773		  || !IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr, &ip6->ip6_dst))) {
774			RTFREE(state->ro->ro_rt);
775			state->ro->ro_rt = NULL;
776		}
777		if (state->ro->ro_rt == 0) {
778			bzero(dst6, sizeof(*dst6));
779			dst6->sin6_family = AF_INET6;
780			dst6->sin6_len = sizeof(*dst6);
781			dst6->sin6_addr = ip6->ip6_dst;
782			rtalloc(state->ro);
783		}
784		if (state->ro->ro_rt == 0) {
785			ip6stat.ip6s_noroute++;
786			ipsec6stat.ips_out_noroute++;
787			error = EHOSTUNREACH;
788			goto bad;
789		}
790
791		/* adjust state->dst if tunnel endpoint is offlink */
792		if (state->ro->ro_rt->rt_flags & RTF_GATEWAY) {
793			state->dst = (struct sockaddr *)state->ro->ro_rt->rt_gateway;
794			dst6 = (struct sockaddr_in6 *)state->dst;
795		}
796	}
797
798	m = ipsec6_splithdr(m);
799	if (!m) {
800		ipsec6stat.ips_out_nomem++;
801		error = ENOMEM;
802		goto bad;
803	}
804	ip6 = mtod(m, struct ip6_hdr *);
805
806#ifdef DEV_ENC
807	/* pass the mbuf to enc0 for bpf processing */
808	ipsec_bpf(m, isr->sav, AF_INET6, ENC_OUT|ENC_AFTER);
809	/* pass the mbuf to enc0 for packet filtering */
810	if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0)
811		goto bad;
812#endif
813
814	error = (*isr->sav->tdb_xform->xf_output)(m, isr, NULL,
815		sizeof (struct ip6_hdr),
816		offsetof(struct ip6_hdr, ip6_nxt));
817	IPSECREQUEST_UNLOCK(isr);
818	return error;
819bad:
820	if (isr)
821		IPSECREQUEST_UNLOCK(isr);
822	if (m)
823		m_freem(m);
824	state->m = NULL;
825	return error;
826}
827#endif /*INET6*/
828