xref: /freebsd/sys/netipsec/ipsec_output.c (revision 891b8ed4672a213bbe6f3f10522eeadb34d01b76)
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 
95 int
96 ipsec_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 		V_ipsec4stat.ips_out_bundlesa++;
168 		return ipsec4_process_packet(m, isr->next, 0, 0);
169 	}
170 	key_sa_recordxfer(sav, m);		/* record data transfer */
171 
172 	/*
173 	 * We're done with IPsec processing, transmit the packet using the
174 	 * appropriate network protocol (IP or IPv6). SPD lookup will be
175 	 * performed again there.
176 	 */
177 	switch (saidx->dst.sa.sa_family) {
178 #ifdef INET
179 	struct ip *ip;
180 	case AF_INET:
181 		ip = mtod(m, struct ip *);
182 		ip->ip_len = ntohs(ip->ip_len);
183 		ip->ip_off = ntohs(ip->ip_off);
184 
185 #ifdef IPSEC_NAT_T
186 		/*
187 		 * If NAT-T is enabled, now that all IPsec processing is done
188 		 * insert UDP encapsulation header after IP header.
189 		 */
190 		if (sav->natt_type) {
191 #ifdef _IP_VHL
192 			const int hlen = IP_VHL_HL(ip->ip_vhl);
193 #else
194 			const int hlen = (ip->ip_hl << 2);
195 #endif
196 			int size, off;
197 			struct mbuf *mi;
198 			struct udphdr *udp;
199 
200 			size = sizeof(struct udphdr);
201 			if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE) {
202 				/*
203 				 * draft-ietf-ipsec-nat-t-ike-0[01].txt and
204 				 * draft-ietf-ipsec-udp-encaps-(00/)01.txt,
205 				 * ignoring possible AH mode
206 				 * non-IKE marker + non-ESP marker
207 				 * from draft-ietf-ipsec-udp-encaps-00.txt.
208 				 */
209 				size += sizeof(u_int64_t);
210 			}
211 			mi = m_makespace(m, hlen, size, &off);
212 			if (mi == NULL) {
213 				DPRINTF(("%s: m_makespace for udphdr failed\n",
214 				    __func__));
215 				error = ENOBUFS;
216 				goto bad;
217 			}
218 
219 			udp = (struct udphdr *)(mtod(mi, caddr_t) + off);
220 			if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE)
221 				udp->uh_sport = htons(UDP_ENCAP_ESPINUDP_PORT);
222 			else
223 				udp->uh_sport =
224 					KEY_PORTFROMSADDR(&sav->sah->saidx.src);
225 			udp->uh_dport = KEY_PORTFROMSADDR(&sav->sah->saidx.dst);
226 			udp->uh_sum = 0;
227 			udp->uh_ulen = htons(m->m_pkthdr.len - hlen);
228 			ip->ip_len = m->m_pkthdr.len;
229 			ip->ip_p = IPPROTO_UDP;
230 
231 			if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE)
232 				*(u_int64_t *)(udp + 1) = 0;
233 		}
234 #endif /* IPSEC_NAT_T */
235 
236 		return ip_output(m, NULL, NULL, IP_RAWOUTPUT, NULL, NULL);
237 #endif /* INET */
238 #ifdef INET6
239 	case AF_INET6:
240 		/*
241 		 * We don't need massage, IPv6 header fields are always in
242 		 * net endian.
243 		 */
244 		return ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
245 #endif /* INET6 */
246 	}
247 	panic("ipsec_process_done");
248 bad:
249 	m_freem(m);
250 	return (error);
251 }
252 
253 static struct ipsecrequest *
254 ipsec_nextisr(
255 	struct mbuf *m,
256 	struct ipsecrequest *isr,
257 	int af,
258 	struct secasindex *saidx,
259 	int *error
260 )
261 {
262 #define IPSEC_OSTAT(x,y,z) (isr->saidx.proto == IPPROTO_ESP ? (x)++ : \
263 			    isr->saidx.proto == IPPROTO_AH ? (y)++ : (z)++)
264 	struct secasvar *sav;
265 
266 	IPSECREQUEST_LOCK_ASSERT(isr);
267 
268 	IPSEC_ASSERT(af == AF_INET || af == AF_INET6,
269 		("invalid address family %u", af));
270 again:
271 	/*
272 	 * Craft SA index to search for proper SA.  Note that
273 	 * we only fillin unspecified SA peers for transport
274 	 * mode; for tunnel mode they must already be filled in.
275 	 */
276 	*saidx = isr->saidx;
277 	if (isr->saidx.mode == IPSEC_MODE_TRANSPORT) {
278 		/* Fillin unspecified SA peers only for transport mode */
279 		if (af == AF_INET) {
280 			struct sockaddr_in *sin;
281 			struct ip *ip = mtod(m, struct ip *);
282 
283 			if (saidx->src.sa.sa_len == 0) {
284 				sin = &saidx->src.sin;
285 				sin->sin_len = sizeof(*sin);
286 				sin->sin_family = AF_INET;
287 				sin->sin_port = IPSEC_PORT_ANY;
288 				sin->sin_addr = ip->ip_src;
289 			}
290 			if (saidx->dst.sa.sa_len == 0) {
291 				sin = &saidx->dst.sin;
292 				sin->sin_len = sizeof(*sin);
293 				sin->sin_family = AF_INET;
294 				sin->sin_port = IPSEC_PORT_ANY;
295 				sin->sin_addr = ip->ip_dst;
296 			}
297 		} else {
298 			struct sockaddr_in6 *sin6;
299 			struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
300 
301 			if (saidx->src.sin6.sin6_len == 0) {
302 				sin6 = (struct sockaddr_in6 *)&saidx->src;
303 				sin6->sin6_len = sizeof(*sin6);
304 				sin6->sin6_family = AF_INET6;
305 				sin6->sin6_port = IPSEC_PORT_ANY;
306 				sin6->sin6_addr = ip6->ip6_src;
307 				if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
308 					/* fix scope id for comparing SPD */
309 					sin6->sin6_addr.s6_addr16[1] = 0;
310 					sin6->sin6_scope_id =
311 					    ntohs(ip6->ip6_src.s6_addr16[1]);
312 				}
313 			}
314 			if (saidx->dst.sin6.sin6_len == 0) {
315 				sin6 = (struct sockaddr_in6 *)&saidx->dst;
316 				sin6->sin6_len = sizeof(*sin6);
317 				sin6->sin6_family = AF_INET6;
318 				sin6->sin6_port = IPSEC_PORT_ANY;
319 				sin6->sin6_addr = ip6->ip6_dst;
320 				if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
321 					/* fix scope id for comparing SPD */
322 					sin6->sin6_addr.s6_addr16[1] = 0;
323 					sin6->sin6_scope_id =
324 					    ntohs(ip6->ip6_dst.s6_addr16[1]);
325 				}
326 			}
327 		}
328 	}
329 
330 	/*
331 	 * Lookup SA and validate it.
332 	 */
333 	*error = key_checkrequest(isr, saidx);
334 	if (*error != 0) {
335 		/*
336 		 * IPsec processing is required, but no SA found.
337 		 * I assume that key_acquire() had been called
338 		 * to get/establish the SA. Here I discard
339 		 * this packet because it is responsibility for
340 		 * upper layer to retransmit the packet.
341 		 */
342 		V_ipsec4stat.ips_out_nosa++;
343 		goto bad;
344 	}
345 	sav = isr->sav;
346 	if (sav == NULL) {
347 		IPSEC_ASSERT(ipsec_get_reqlevel(isr) == IPSEC_LEVEL_USE,
348 			("no SA found, but required; level %u",
349 			ipsec_get_reqlevel(isr)));
350 		IPSECREQUEST_UNLOCK(isr);
351 		isr = isr->next;
352 		/*
353 		 * If isr is NULL, we found a 'use' policy w/o SA.
354 		 * Return w/o error and w/o isr so we can drop out
355 		 * and continue w/o IPsec processing.
356 		 */
357 		if (isr == NULL)
358 			return isr;
359 		IPSECREQUEST_LOCK(isr);
360 		goto again;
361 	}
362 
363 	/*
364 	 * Check system global policy controls.
365 	 */
366 	if ((isr->saidx.proto == IPPROTO_ESP && !V_esp_enable) ||
367 	    (isr->saidx.proto == IPPROTO_AH && !V_ah_enable) ||
368 	    (isr->saidx.proto == IPPROTO_IPCOMP && !V_ipcomp_enable)) {
369 		DPRINTF(("%s: IPsec outbound packet dropped due"
370 			" to policy (check your sysctls)\n", __func__));
371 		IPSEC_OSTAT(V_espstat.esps_pdrops, V_ahstat.ahs_pdrops,
372 		    V_ipcompstat.ipcomps_pdrops);
373 		*error = EHOSTUNREACH;
374 		goto bad;
375 	}
376 
377 	/*
378 	 * Sanity check the SA contents for the caller
379 	 * before they invoke the xform output method.
380 	 */
381 	if (sav->tdb_xform == NULL) {
382 		DPRINTF(("%s: no transform for SA\n", __func__));
383 		IPSEC_OSTAT(V_espstat.esps_noxform, V_ahstat.ahs_noxform,
384 		    V_ipcompstat.ipcomps_noxform);
385 		*error = EHOSTUNREACH;
386 		goto bad;
387 	}
388 	return isr;
389 bad:
390 	IPSEC_ASSERT(*error != 0, ("error return w/ no error code"));
391 	IPSECREQUEST_UNLOCK(isr);
392 	return NULL;
393 #undef IPSEC_OSTAT
394 }
395 
396 #ifdef INET
397 /*
398  * IPsec output logic for IPv4.
399  */
400 int
401 ipsec4_process_packet(
402 	struct mbuf *m,
403 	struct ipsecrequest *isr,
404 	int flags,
405 	int tunalready)
406 {
407 	struct secasindex saidx;
408 	struct secasvar *sav;
409 	struct ip *ip;
410 	int error, i, off;
411 
412 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
413 	IPSEC_ASSERT(isr != NULL, ("null isr"));
414 
415 	IPSECREQUEST_LOCK(isr);		/* insure SA contents don't change */
416 
417 	isr = ipsec_nextisr(m, isr, AF_INET, &saidx, &error);
418 	if (isr == NULL) {
419 		if (error != 0)
420 			goto bad;
421 		return EJUSTRETURN;
422 	}
423 
424 	sav = isr->sav;
425 
426 #ifdef DEV_ENC
427 	encif->if_opackets++;
428 	encif->if_obytes += m->m_pkthdr.len;
429 
430 	/* pass the mbuf to enc0 for bpf processing */
431 	ipsec_bpf(m, sav, AF_INET, ENC_OUT|ENC_BEFORE);
432 	/* pass the mbuf to enc0 for packet filtering */
433 	if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0)
434 		goto bad;
435 #endif
436 
437 	if (!tunalready) {
438 		union sockaddr_union *dst = &sav->sah->saidx.dst;
439 		int setdf;
440 
441 		/*
442 		 * Collect IP_DF state from the outer header.
443 		 */
444 		if (dst->sa.sa_family == AF_INET) {
445 			if (m->m_len < sizeof (struct ip) &&
446 			    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
447 				error = ENOBUFS;
448 				goto bad;
449 			}
450 			ip = mtod(m, struct ip *);
451 			/* Honor system-wide control of how to handle IP_DF */
452 			switch (V_ip4_ipsec_dfbit) {
453 			case 0:			/* clear in outer header */
454 			case 1:			/* set in outer header */
455 				setdf = V_ip4_ipsec_dfbit;
456 				break;
457 			default:		/* propagate to outer header */
458 				setdf = ntohs(ip->ip_off & IP_DF);
459 				break;
460 			}
461 		} else {
462 			ip = NULL;		/* keep compiler happy */
463 			setdf = 0;
464 		}
465 		/* Do the appropriate encapsulation, if necessary */
466 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */
467 		    dst->sa.sa_family != AF_INET ||	    /* PF mismatch */
468 #if 0
469 		    (sav->flags & SADB_X_SAFLAGS_TUNNEL) || /* Tunnel requ'd */
470 		    sav->tdb_xform->xf_type == XF_IP4 ||    /* ditto */
471 #endif
472 		    (dst->sa.sa_family == AF_INET &&	    /* Proxy */
473 		     dst->sin.sin_addr.s_addr != INADDR_ANY &&
474 		     dst->sin.sin_addr.s_addr != ip->ip_dst.s_addr)) {
475 			struct mbuf *mp;
476 
477 			/* Fix IPv4 header checksum and length */
478 			if (m->m_len < sizeof (struct ip) &&
479 			    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
480 				error = ENOBUFS;
481 				goto bad;
482 			}
483 			ip = mtod(m, struct ip *);
484 			ip->ip_len = htons(m->m_pkthdr.len);
485 			ip->ip_sum = 0;
486 #ifdef _IP_VHL
487 			if (ip->ip_vhl == IP_VHL_BORING)
488 				ip->ip_sum = in_cksum_hdr(ip);
489 			else
490 				ip->ip_sum = in_cksum(m,
491 					_IP_VHL_HL(ip->ip_vhl) << 2);
492 #else
493 			ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
494 #endif
495 
496 			/* Encapsulate the packet */
497 			error = ipip_output(m, isr, &mp, 0, 0);
498 			if (mp == NULL && !error) {
499 				/* Should never happen. */
500 				DPRINTF(("%s: ipip_output returns no mbuf and "
501 					"no error!", __func__));
502 				error = EFAULT;
503 			}
504 			if (error) {
505 				if (mp) {
506 					/* XXX: Should never happen! */
507 					m_freem(mp);
508 				}
509 				m = NULL; /* ipip_output() already freed it */
510 				goto bad;
511 			}
512 			m = mp, mp = NULL;
513 			/*
514 			 * ipip_output clears IP_DF in the new header.  If
515 			 * we need to propagate IP_DF from the outer header,
516 			 * then we have to do it here.
517 			 *
518 			 * XXX shouldn't assume what ipip_output does.
519 			 */
520 			if (dst->sa.sa_family == AF_INET && setdf) {
521 				if (m->m_len < sizeof (struct ip) &&
522 				    (m = m_pullup(m, sizeof (struct ip))) == NULL) {
523 					error = ENOBUFS;
524 					goto bad;
525 				}
526 				ip = mtod(m, struct ip *);
527 				ip->ip_off = ntohs(ip->ip_off);
528 				ip->ip_off |= IP_DF;
529 				ip->ip_off = htons(ip->ip_off);
530 			}
531 		}
532 	}
533 
534 #ifdef DEV_ENC
535 	/* pass the mbuf to enc0 for bpf processing */
536 	ipsec_bpf(m, sav, AF_INET, ENC_OUT|ENC_AFTER);
537 	/* pass the mbuf to enc0 for packet filtering */
538 	if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0)
539 		goto bad;
540 #endif
541 
542 	/*
543 	 * Dispatch to the appropriate IPsec transform logic.  The
544 	 * packet will be returned for transmission after crypto
545 	 * processing, etc. are completed.  For encapsulation we
546 	 * bypass this call because of the explicit call done above
547 	 * (necessary to deal with IP_DF handling for IPv4).
548 	 *
549 	 * NB: m & sav are ``passed to caller'' who's reponsible for
550 	 *     for reclaiming their resources.
551 	 */
552 	if (sav->tdb_xform->xf_type != XF_IP4) {
553 		ip = mtod(m, struct ip *);
554 		i = ip->ip_hl << 2;
555 		off = offsetof(struct ip, ip_p);
556 		error = (*sav->tdb_xform->xf_output)(m, isr, NULL, i, off);
557 	} else {
558 		error = ipsec_process_done(m, isr);
559 	}
560 	IPSECREQUEST_UNLOCK(isr);
561 	return error;
562 bad:
563 	if (isr)
564 		IPSECREQUEST_UNLOCK(isr);
565 	if (m)
566 		m_freem(m);
567 	return error;
568 }
569 #endif
570 
571 #ifdef INET6
572 /*
573  * Chop IP6 header from the payload.
574  */
575 static struct mbuf *
576 ipsec6_splithdr(struct mbuf *m)
577 {
578 	struct mbuf *mh;
579 	struct ip6_hdr *ip6;
580 	int hlen;
581 
582 	IPSEC_ASSERT(m->m_len >= sizeof (struct ip6_hdr),
583 		("first mbuf too short, len %u", m->m_len));
584 	ip6 = mtod(m, struct ip6_hdr *);
585 	hlen = sizeof(struct ip6_hdr);
586 	if (m->m_len > hlen) {
587 		MGETHDR(mh, M_DONTWAIT, MT_DATA);
588 		if (!mh) {
589 			m_freem(m);
590 			return NULL;
591 		}
592 		M_MOVE_PKTHDR(mh, m);
593 		MH_ALIGN(mh, hlen);
594 		m->m_len -= hlen;
595 		m->m_data += hlen;
596 		mh->m_next = m;
597 		m = mh;
598 		m->m_len = hlen;
599 		bcopy((caddr_t)ip6, mtod(m, caddr_t), hlen);
600 	} else if (m->m_len < hlen) {
601 		m = m_pullup(m, hlen);
602 		if (!m)
603 			return NULL;
604 	}
605 	return m;
606 }
607 
608 /*
609  * IPsec output logic for IPv6, transport mode.
610  */
611 int
612 ipsec6_output_trans(
613 	struct ipsec_output_state *state,
614 	u_char *nexthdrp,
615 	struct mbuf *mprev,
616 	struct secpolicy *sp,
617 	int flags,
618 	int *tun)
619 {
620 	struct ipsecrequest *isr;
621 	struct secasindex saidx;
622 	int error = 0;
623 	struct mbuf *m;
624 
625 	IPSEC_ASSERT(state != NULL, ("null state"));
626 	IPSEC_ASSERT(state->m != NULL, ("null m"));
627 	IPSEC_ASSERT(nexthdrp != NULL, ("null nexthdrp"));
628 	IPSEC_ASSERT(mprev != NULL, ("null mprev"));
629 	IPSEC_ASSERT(sp != NULL, ("null sp"));
630 	IPSEC_ASSERT(tun != NULL, ("null tun"));
631 
632 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
633 		printf("%s: applied SP\n", __func__);
634 		kdebug_secpolicy(sp));
635 
636 	isr = sp->req;
637 	if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
638 		/* the rest will be handled by ipsec6_output_tunnel() */
639 		*tun = 1;		/* need tunnel-mode processing */
640 		return 0;
641 	}
642 
643 	*tun = 0;
644 	m = state->m;
645 
646 	IPSECREQUEST_LOCK(isr);		/* insure SA contents don't change */
647 	isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
648 	if (isr == NULL) {
649 		if (error != 0) {
650 #ifdef notdef
651 			/* XXX should notification be done for all errors ? */
652 			/*
653 			 * Notify the fact that the packet is discarded
654 			 * to ourselves. I believe this is better than
655 			 * just silently discarding. (jinmei@kame.net)
656 			 * XXX: should we restrict the error to TCP packets?
657 			 * XXX: should we directly notify sockets via
658 			 *      pfctlinputs?
659 			 */
660 			icmp6_error(m, ICMP6_DST_UNREACH,
661 				    ICMP6_DST_UNREACH_ADMIN, 0);
662 			m = NULL;	/* NB: icmp6_error frees mbuf */
663 #endif
664 			goto bad;
665 		}
666 		return EJUSTRETURN;
667 	}
668 
669 	error = (*isr->sav->tdb_xform->xf_output)(m, isr, NULL,
670 						  sizeof (struct ip6_hdr),
671 						  offsetof(struct ip6_hdr,
672 							   ip6_nxt));
673 	IPSECREQUEST_UNLOCK(isr);
674 	return error;
675 bad:
676 	if (isr)
677 		IPSECREQUEST_UNLOCK(isr);
678 	if (m)
679 		m_freem(m);
680 	state->m = NULL;
681 	return error;
682 }
683 
684 static int
685 ipsec6_encapsulate(struct mbuf *m, struct secasvar *sav)
686 {
687 	struct ip6_hdr *oip6;
688 	struct ip6_hdr *ip6;
689 	size_t plen;
690 
691 	/* can't tunnel between different AFs */
692 	if (sav->sah->saidx.src.sa.sa_family != AF_INET6 ||
693 	    sav->sah->saidx.dst.sa.sa_family != AF_INET6) {
694 		m_freem(m);
695 		return EINVAL;
696 	}
697 	IPSEC_ASSERT(m->m_len == sizeof (struct ip6_hdr),
698 		("mbuf wrong size; len %u", m->m_len));
699 
700 
701 	/*
702 	 * grow the mbuf to accomodate the new IPv6 header.
703 	 */
704 	plen = m->m_pkthdr.len;
705 	if (M_LEADINGSPACE(m->m_next) < sizeof(struct ip6_hdr)) {
706 		struct mbuf *n;
707 		MGET(n, M_DONTWAIT, MT_DATA);
708 		if (!n) {
709 			m_freem(m);
710 			return ENOBUFS;
711 		}
712 		n->m_len = sizeof(struct ip6_hdr);
713 		n->m_next = m->m_next;
714 		m->m_next = n;
715 		m->m_pkthdr.len += sizeof(struct ip6_hdr);
716 		oip6 = mtod(n, struct ip6_hdr *);
717 	} else {
718 		m->m_next->m_len += sizeof(struct ip6_hdr);
719 		m->m_next->m_data -= sizeof(struct ip6_hdr);
720 		m->m_pkthdr.len += sizeof(struct ip6_hdr);
721 		oip6 = mtod(m->m_next, struct ip6_hdr *);
722 	}
723 	ip6 = mtod(m, struct ip6_hdr *);
724 	bcopy((caddr_t)ip6, (caddr_t)oip6, sizeof(struct ip6_hdr));
725 
726 	/* Fake link-local scope-class addresses */
727 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
728 		oip6->ip6_src.s6_addr16[1] = 0;
729 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
730 		oip6->ip6_dst.s6_addr16[1] = 0;
731 
732 	/* construct new IPv6 header. see RFC 2401 5.1.2.2 */
733 	/* ECN consideration. */
734 	ip6_ecn_ingress(V_ip6_ipsec_ecn, &ip6->ip6_flow, &oip6->ip6_flow);
735 	if (plen < IPV6_MAXPACKET - sizeof(struct ip6_hdr))
736 		ip6->ip6_plen = htons(plen);
737 	else {
738 		/* ip6->ip6_plen will be updated in ip6_output() */
739 	}
740 	ip6->ip6_nxt = IPPROTO_IPV6;
741 	ip6->ip6_src = sav->sah->saidx.src.sin6.sin6_addr;
742 	ip6->ip6_dst = sav->sah->saidx.dst.sin6.sin6_addr;
743 	ip6->ip6_hlim = IPV6_DEFHLIM;
744 
745 	/* XXX Should ip6_src be updated later ? */
746 
747 	return 0;
748 }
749 
750 /*
751  * IPsec output logic for IPv6, tunnel mode.
752  */
753 int
754 ipsec6_output_tunnel(struct ipsec_output_state *state, struct secpolicy *sp, int flags)
755 {
756 	struct ip6_hdr *ip6;
757 	struct ipsecrequest *isr;
758 	struct secasindex saidx;
759 	int error;
760 	struct sockaddr_in6 *dst6;
761 	struct mbuf *m;
762 
763 	IPSEC_ASSERT(state != NULL, ("null state"));
764 	IPSEC_ASSERT(state->m != NULL, ("null m"));
765 	IPSEC_ASSERT(sp != NULL, ("null sp"));
766 
767 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
768 		printf("%s: applied SP\n", __func__);
769 		kdebug_secpolicy(sp));
770 
771 	m = state->m;
772 	/*
773 	 * transport mode ipsec (before the 1st tunnel mode) is already
774 	 * processed by ipsec6_output_trans().
775 	 */
776 	for (isr = sp->req; isr; isr = isr->next) {
777 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
778 			break;
779 	}
780 
781 	IPSECREQUEST_LOCK(isr);		/* insure SA contents don't change */
782 	isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error);
783 	if (isr == NULL) {
784 		if (error != 0)
785 			goto bad;
786 		return EJUSTRETURN;
787 	}
788 
789 #ifdef DEV_ENC
790 	encif->if_opackets++;
791 	encif->if_obytes += m->m_pkthdr.len;
792 
793 	/* pass the mbuf to enc0 for bpf processing */
794 	ipsec_bpf(m, isr->sav, AF_INET6, ENC_OUT|ENC_BEFORE);
795 	/* pass the mbuf to enc0 for packet filtering */
796 	if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0)
797 		goto bad;
798 #endif
799 
800 	/*
801 	 * There may be the case that SA status will be changed when
802 	 * we are refering to one. So calling splsoftnet().
803 	 */
804 	if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
805 		/*
806 		 * build IPsec tunnel.
807 		 */
808 		/* XXX should be processed with other familiy */
809 		if (isr->sav->sah->saidx.src.sa.sa_family != AF_INET6) {
810 			ipseclog((LOG_ERR, "%s: family mismatched between "
811 			    "inner and outer, spi=%u\n", __func__,
812 			    ntohl(isr->sav->spi)));
813 			V_ipsec6stat.ips_out_inval++;
814 			error = EAFNOSUPPORT;
815 			goto bad;
816 		}
817 
818 		m = ipsec6_splithdr(m);
819 		if (!m) {
820 			V_ipsec6stat.ips_out_nomem++;
821 			error = ENOMEM;
822 			goto bad;
823 		}
824 		error = ipsec6_encapsulate(m, isr->sav);
825 		if (error) {
826 			m = NULL;
827 			goto bad;
828 		}
829 		ip6 = mtod(m, struct ip6_hdr *);
830 
831 		state->ro =
832 		    (struct route *)&isr->sav->sah->route_cache.sin6_route;
833 		state->dst = (struct sockaddr *)&state->ro->ro_dst;
834 		dst6 = (struct sockaddr_in6 *)state->dst;
835 		if (state->ro->ro_rt
836 		 && ((state->ro->ro_rt->rt_flags & RTF_UP) == 0
837 		  || !IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr, &ip6->ip6_dst))) {
838 			RTFREE(state->ro->ro_rt);
839 			state->ro->ro_rt = NULL;
840 		}
841 		if (state->ro->ro_rt == NULL) {
842 			bzero(dst6, sizeof(*dst6));
843 			dst6->sin6_family = AF_INET6;
844 			dst6->sin6_len = sizeof(*dst6);
845 			dst6->sin6_addr = ip6->ip6_dst;
846 			rtalloc(state->ro);
847 		}
848 		if (state->ro->ro_rt == NULL) {
849 			V_ip6stat.ip6s_noroute++;
850 			V_ipsec6stat.ips_out_noroute++;
851 			error = EHOSTUNREACH;
852 			goto bad;
853 		}
854 
855 		/* adjust state->dst if tunnel endpoint is offlink */
856 		if (state->ro->ro_rt->rt_flags & RTF_GATEWAY)
857 			state->dst = (struct sockaddr *)state->ro->ro_rt->rt_gateway;
858 	}
859 
860 	m = ipsec6_splithdr(m);
861 	if (!m) {
862 		V_ipsec6stat.ips_out_nomem++;
863 		error = ENOMEM;
864 		goto bad;
865 	}
866 	ip6 = mtod(m, struct ip6_hdr *);
867 
868 #ifdef DEV_ENC
869 	/* pass the mbuf to enc0 for bpf processing */
870 	ipsec_bpf(m, isr->sav, AF_INET6, ENC_OUT|ENC_AFTER);
871 	/* pass the mbuf to enc0 for packet filtering */
872 	if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0)
873 		goto bad;
874 #endif
875 
876 	error = (*isr->sav->tdb_xform->xf_output)(m, isr, NULL,
877 		sizeof (struct ip6_hdr),
878 		offsetof(struct ip6_hdr, ip6_nxt));
879 	IPSECREQUEST_UNLOCK(isr);
880 	return error;
881 bad:
882 	if (isr)
883 		IPSECREQUEST_UNLOCK(isr);
884 	if (m)
885 		m_freem(m);
886 	state->m = NULL;
887 	return error;
888 }
889 #endif /*INET6*/
890