xref: /openbsd-src/sys/netinet/ipsec_input.c (revision 824adb5411e4389b29bae28eba5c2c2bbd147f34)
1 /*	$OpenBSD: ipsec_input.c,v 1.182 2021/10/05 11:45:26 bluhm Exp $	*/
2 /*
3  * The authors of this code are John Ioannidis (ji@tla.org),
4  * Angelos D. Keromytis (kermit@csd.uch.gr) and
5  * Niels Provos (provos@physnet.uni-hamburg.de).
6  *
7  * This code was written by John Ioannidis for BSD/OS in Athens, Greece,
8  * in November 1995.
9  *
10  * Ported to OpenBSD and NetBSD, with additional transforms, in December 1996,
11  * by Angelos D. Keromytis.
12  *
13  * Additional transforms and features in 1997 and 1998 by Angelos D. Keromytis
14  * and Niels Provos.
15  *
16  * Additional features in 1999 by Angelos D. Keromytis.
17  *
18  * Copyright (C) 1995, 1996, 1997, 1998, 1999 by John Ioannidis,
19  * Angelos D. Keromytis and Niels Provos.
20  * Copyright (c) 2001, Angelos D. Keromytis.
21  *
22  * Permission to use, copy, and modify this software with or without fee
23  * is hereby granted, provided that this entire notice is included in
24  * all copies of any software which is or includes a copy or
25  * modification of this software.
26  * You may use this code under the GNU public license if you so wish. Please
27  * contribute changes back to the authors under this freer than GPL license
28  * so that we may further the use of strong encryption without limitations to
29  * all.
30  *
31  * THIS SOFTWARE IS BEING PROVIDED "AS IS", WITHOUT ANY EXPRESS OR
32  * IMPLIED WARRANTY. IN PARTICULAR, NONE OF THE AUTHORS MAKES ANY
33  * REPRESENTATION OR WARRANTY OF ANY KIND CONCERNING THE
34  * MERCHANTABILITY OF THIS SOFTWARE OR ITS FITNESS FOR ANY PARTICULAR
35  * PURPOSE.
36  */
37 
38 #include "pf.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/protosw.h>
43 #include <sys/mbuf.h>
44 #include <sys/socket.h>
45 #include <sys/sysctl.h>
46 #include <sys/kernel.h>
47 #include <sys/timeout.h>
48 
49 #include <net/if.h>
50 #include <net/if_var.h>
51 #include <net/netisr.h>
52 #include <net/bpf.h>
53 #include <net/route.h>
54 
55 #include <netinet/in.h>
56 #include <netinet/ip.h>
57 #include <netinet/ip_var.h>
58 #include <netinet/ip_icmp.h>
59 #include <netinet/tcp.h>
60 #include <netinet/udp.h>
61 
62 #if NPF > 0
63 #include <net/pfvar.h>
64 #endif
65 
66 #ifdef INET6
67 #include <netinet6/in6_var.h>
68 #include <netinet/ip6.h>
69 #include <netinet6/ip6_var.h>
70 #include <netinet6/ip6protosw.h>
71 #endif /* INET6 */
72 
73 #include <netinet/ip_ipsp.h>
74 #include <netinet/ip_esp.h>
75 #include <netinet/ip_ah.h>
76 #include <netinet/ip_ipcomp.h>
77 
78 #include <net/if_enc.h>
79 
80 #include <crypto/cryptodev.h>
81 #include <crypto/xform.h>
82 
83 #include "bpfilter.h"
84 
85 void ipsec_common_ctlinput(u_int, int, struct sockaddr *, void *, int);
86 
87 #ifdef ENCDEBUG
88 #define DPRINTF(fmt, args...)						\
89 	do {								\
90 		if (encdebug)						\
91 			printf("%s: " fmt "\n", __func__, ## args);	\
92 	} while (0)
93 #else
94 #define DPRINTF(fmt, args...)						\
95 	do { } while (0)
96 #endif
97 
98 /* sysctl variables */
99 int encdebug = 0;
100 int ipsec_keep_invalid = IPSEC_DEFAULT_EMBRYONIC_SA_TIMEOUT;
101 int ipsec_require_pfs = IPSEC_DEFAULT_PFS;
102 int ipsec_soft_allocations = IPSEC_DEFAULT_SOFT_ALLOCATIONS;
103 int ipsec_exp_allocations = IPSEC_DEFAULT_EXP_ALLOCATIONS;
104 int ipsec_soft_bytes = IPSEC_DEFAULT_SOFT_BYTES;
105 int ipsec_exp_bytes = IPSEC_DEFAULT_EXP_BYTES;
106 int ipsec_soft_timeout = IPSEC_DEFAULT_SOFT_TIMEOUT;
107 int ipsec_exp_timeout = IPSEC_DEFAULT_EXP_TIMEOUT;
108 int ipsec_soft_first_use = IPSEC_DEFAULT_SOFT_FIRST_USE;
109 int ipsec_exp_first_use = IPSEC_DEFAULT_EXP_FIRST_USE;
110 int ipsec_expire_acquire = IPSEC_DEFAULT_EXPIRE_ACQUIRE;
111 
112 int esp_enable = 1;
113 int ah_enable = 1;
114 int ipcomp_enable = 0;
115 
116 const struct sysctl_bounded_args espctl_vars[] = {
117 	{ESPCTL_ENABLE, &esp_enable, 0, 1},
118 	{ESPCTL_UDPENCAP_ENABLE, &udpencap_enable, 0, 1},
119 	{ESPCTL_UDPENCAP_PORT, &udpencap_port, 0, 65535},
120 };
121 const struct sysctl_bounded_args ahctl_vars[] = {
122 	{AHCTL_ENABLE, &ah_enable, 0, 1},
123 };
124 const struct sysctl_bounded_args ipcompctl_vars[] = {
125 	{IPCOMPCTL_ENABLE, &ipcomp_enable, 0, 1},
126 };
127 
128 struct cpumem *espcounters;
129 struct cpumem *ahcounters;
130 struct cpumem *ipcompcounters;
131 struct cpumem *ipseccounters;
132 
133 char ipsec_def_enc[20];
134 char ipsec_def_auth[20];
135 char ipsec_def_comp[20];
136 
137 const struct sysctl_bounded_args ipsecctl_vars[] = {
138 	{ IPSEC_ENCDEBUG, &encdebug, 0, 1 },
139 	{ IPSEC_EXPIRE_ACQUIRE, &ipsec_expire_acquire, 0, INT_MAX },
140 	{ IPSEC_EMBRYONIC_SA_TIMEOUT, &ipsec_keep_invalid, 0, INT_MAX },
141 	{ IPSEC_REQUIRE_PFS, &ipsec_require_pfs, 0, 1 },
142 	{ IPSEC_SOFT_ALLOCATIONS, &ipsec_soft_allocations, 0, INT_MAX },
143 	{ IPSEC_ALLOCATIONS, &ipsec_exp_allocations, 0, INT_MAX },
144 	{ IPSEC_SOFT_BYTES, &ipsec_soft_bytes, 0, INT_MAX },
145 	{ IPSEC_BYTES, &ipsec_exp_bytes, 0, INT_MAX },
146 	{ IPSEC_TIMEOUT, &ipsec_exp_timeout, 0, INT_MAX },
147 	{ IPSEC_SOFT_TIMEOUT, &ipsec_soft_timeout,0, INT_MAX },
148 	{ IPSEC_SOFT_FIRSTUSE, &ipsec_soft_first_use, 0, INT_MAX },
149 	{ IPSEC_FIRSTUSE, &ipsec_exp_first_use, 0, INT_MAX },
150 };
151 
152 int esp_sysctl_espstat(void *, size_t *, void *);
153 int ah_sysctl_ahstat(void *, size_t *, void *);
154 int ipcomp_sysctl_ipcompstat(void *, size_t *, void *);
155 int ipsec_sysctl_ipsecstat(void *, size_t *, void *);
156 
157 void
158 ipsec_init(void)
159 {
160 	espcounters = counters_alloc(esps_ncounters);
161 	ahcounters = counters_alloc(ahs_ncounters);
162 	ipcompcounters = counters_alloc(ipcomps_ncounters);
163 	ipseccounters = counters_alloc(ipsec_ncounters);
164 
165 	strlcpy(ipsec_def_enc, IPSEC_DEFAULT_DEF_ENC, sizeof(ipsec_def_enc));
166 	strlcpy(ipsec_def_auth, IPSEC_DEFAULT_DEF_AUTH, sizeof(ipsec_def_auth));
167 	strlcpy(ipsec_def_comp, IPSEC_DEFAULT_DEF_COMP, sizeof(ipsec_def_comp));
168 
169 	ipsp_init();
170 }
171 
172 /*
173  * ipsec_common_input() gets called when we receive an IPsec-protected packet
174  * in IPv4 or IPv6. All it does is find the right TDB and call the appropriate
175  * transform. The callback takes care of further processing (like ingress
176  * filtering).
177  */
178 int
179 ipsec_common_input(struct mbuf *m, int skip, int protoff, int af, int sproto,
180     int udpencap)
181 {
182 #define IPSEC_ISTAT(x,y,z) do {			\
183 	if (sproto == IPPROTO_ESP)		\
184 		espstat_inc(x);			\
185 	else if (sproto == IPPROTO_AH)		\
186 		ahstat_inc(y);			\
187 	else					\
188 		ipcompstat_inc(z);		\
189 } while (0)
190 
191 	union sockaddr_union dst_address;
192 	struct tdb *tdbp = NULL;
193 	struct ifnet *encif;
194 	u_int32_t spi;
195 	u_int16_t cpi;
196 	int error;
197 #ifdef ENCDEBUG
198 	char buf[INET6_ADDRSTRLEN];
199 #endif
200 
201 	NET_ASSERT_LOCKED();
202 
203 	ipsecstat_pkt(ipsec_ipackets, ipsec_ibytes, m->m_pkthdr.len);
204 	IPSEC_ISTAT(esps_input, ahs_input, ipcomps_input);
205 
206 	if (m == NULL) {
207 		DPRINTF("NULL packet received");
208 		IPSEC_ISTAT(esps_hdrops, ahs_hdrops, ipcomps_hdrops);
209 		return EINVAL;
210 	}
211 
212 	if ((sproto == IPPROTO_IPCOMP) && (m->m_flags & M_COMP)) {
213 		DPRINTF("repeated decompression");
214 		ipcompstat_inc(ipcomps_pdrops);
215 		error = EINVAL;
216 		goto drop;
217 	}
218 
219 	if (m->m_pkthdr.len - skip < 2 * sizeof(u_int32_t)) {
220 		DPRINTF("packet too small");
221 		IPSEC_ISTAT(esps_hdrops, ahs_hdrops, ipcomps_hdrops);
222 		error = EINVAL;
223 		goto drop;
224 	}
225 
226 	/* Retrieve the SPI from the relevant IPsec header */
227 	switch (sproto) {
228 	case IPPROTO_ESP:
229 		m_copydata(m, skip, sizeof(u_int32_t), (caddr_t) &spi);
230 		break;
231 	case IPPROTO_AH:
232 		m_copydata(m, skip + sizeof(u_int32_t), sizeof(u_int32_t),
233 		    (caddr_t) &spi);
234 		break;
235 	case IPPROTO_IPCOMP:
236 		m_copydata(m, skip + sizeof(u_int16_t), sizeof(u_int16_t),
237 		    (caddr_t) &cpi);
238 		spi = ntohl(htons(cpi));
239 		break;
240 	default:
241 		panic("%s: unknown/unsupported security protocol %d",
242 		    __func__, sproto);
243 	}
244 
245 	/*
246 	 * Find tunnel control block and (indirectly) call the appropriate
247 	 * kernel crypto routine. The resulting mbuf chain is a valid
248 	 * IP packet ready to go through input processing.
249 	 */
250 
251 	memset(&dst_address, 0, sizeof(dst_address));
252 	dst_address.sa.sa_family = af;
253 
254 	switch (af) {
255 	case AF_INET:
256 		dst_address.sin.sin_len = sizeof(struct sockaddr_in);
257 		m_copydata(m, offsetof(struct ip, ip_dst),
258 		    sizeof(struct in_addr),
259 		    (caddr_t) &(dst_address.sin.sin_addr));
260 		break;
261 
262 #ifdef INET6
263 	case AF_INET6:
264 		dst_address.sin6.sin6_len = sizeof(struct sockaddr_in6);
265 		m_copydata(m, offsetof(struct ip6_hdr, ip6_dst),
266 		    sizeof(struct in6_addr),
267 		    (caddr_t) &(dst_address.sin6.sin6_addr));
268 		in6_recoverscope(&dst_address.sin6,
269 		    &dst_address.sin6.sin6_addr);
270 		break;
271 #endif /* INET6 */
272 
273 	default:
274 		DPRINTF("unsupported protocol family %d", af);
275 		IPSEC_ISTAT(esps_nopf, ahs_nopf, ipcomps_nopf);
276 		error = EPFNOSUPPORT;
277 		goto drop;
278 	}
279 
280 	tdbp = gettdb(rtable_l2(m->m_pkthdr.ph_rtableid),
281 	    spi, &dst_address, sproto);
282 	if (tdbp == NULL) {
283 		DPRINTF("could not find SA for packet to %s, spi %08x",
284 		    ipsp_address(&dst_address, buf, sizeof(buf)), ntohl(spi));
285 		IPSEC_ISTAT(esps_notdb, ahs_notdb, ipcomps_notdb);
286 		error = ENOENT;
287 		goto drop;
288 	}
289 
290 	if (tdbp->tdb_flags & TDBF_INVALID) {
291 		DPRINTF("attempted to use invalid SA %s/%08x/%u",
292 		    ipsp_address(&dst_address, buf, sizeof(buf)),
293 		    ntohl(spi), tdbp->tdb_sproto);
294 		IPSEC_ISTAT(esps_invalid, ahs_invalid, ipcomps_invalid);
295 		error = EINVAL;
296 		goto drop;
297 	}
298 
299 	if (udpencap && !(tdbp->tdb_flags & TDBF_UDPENCAP)) {
300 		DPRINTF("attempted to use non-udpencap SA %s/%08x/%u",
301 		    ipsp_address(&dst_address, buf, sizeof(buf)),
302 		    ntohl(spi), tdbp->tdb_sproto);
303 		espstat_inc(esps_udpinval);
304 		error = EINVAL;
305 		goto drop;
306 	}
307 
308 	if (!udpencap && (tdbp->tdb_flags & TDBF_UDPENCAP)) {
309 		DPRINTF("attempted to use udpencap SA %s/%08x/%u",
310 		    ipsp_address(&dst_address, buf, sizeof(buf)),
311 		    ntohl(spi), tdbp->tdb_sproto);
312 		espstat_inc(esps_udpneeded);
313 		error = EINVAL;
314 		goto drop;
315 	}
316 
317 	if (tdbp->tdb_xform == NULL) {
318 		DPRINTF("attempted to use uninitialized SA %s/%08x/%u",
319 		    ipsp_address(&dst_address, buf, sizeof(buf)),
320 		    ntohl(spi), tdbp->tdb_sproto);
321 		IPSEC_ISTAT(esps_noxform, ahs_noxform, ipcomps_noxform);
322 		error = ENXIO;
323 		goto drop;
324 	}
325 
326 	if (sproto != IPPROTO_IPCOMP) {
327 		if ((encif = enc_getif(tdbp->tdb_rdomain_post,
328 		    tdbp->tdb_tap)) == NULL) {
329 			DPRINTF("no enc%u interface for SA %s/%08x/%u",
330 			    tdbp->tdb_tap,
331 			    ipsp_address(&dst_address, buf, sizeof(buf)),
332 			    ntohl(spi), tdbp->tdb_sproto);
333 			IPSEC_ISTAT(esps_pdrops, ahs_pdrops, ipcomps_pdrops);
334 			error = EACCES;
335 			goto drop;
336 		}
337 
338 		/* XXX This conflicts with the scoped nature of IPv6 */
339 		m->m_pkthdr.ph_ifidx = encif->if_index;
340 	}
341 
342 	/* Register first use, setup expiration timer. */
343 	if (tdbp->tdb_first_use == 0) {
344 		tdbp->tdb_first_use = gettime();
345 		if (tdbp->tdb_flags & TDBF_FIRSTUSE)
346 			timeout_add_sec(&tdbp->tdb_first_tmo,
347 			    tdbp->tdb_exp_first_use);
348 		if (tdbp->tdb_flags & TDBF_SOFT_FIRSTUSE)
349 			timeout_add_sec(&tdbp->tdb_sfirst_tmo,
350 			    tdbp->tdb_soft_first_use);
351 	}
352 
353 	tdbp->tdb_ipackets++;
354 	tdbp->tdb_ibytes += m->m_pkthdr.len;
355 
356 	/*
357 	 * Call appropriate transform and return -- callback takes care of
358 	 * everything else.
359 	 */
360 	error = (*(tdbp->tdb_xform->xf_input))(m, tdbp, skip, protoff);
361 	if (error) {
362 		ipsecstat_inc(ipsec_idrops);
363 		tdbp->tdb_idrops++;
364 	}
365 	return error;
366 
367  drop:
368 	ipsecstat_inc(ipsec_idrops);
369 	if (tdbp != NULL)
370 		tdbp->tdb_idrops++;
371 	m_freem(m);
372 	return error;
373 }
374 
375 void
376 ipsec_input_cb(struct cryptop *crp)
377 {
378 	struct tdb_crypto *tc = (struct tdb_crypto *) crp->crp_opaque;
379 	struct mbuf *m = (struct mbuf *) crp->crp_buf;
380 	struct tdb *tdb = NULL;
381 	int clen, error;
382 
383 	NET_ASSERT_LOCKED();
384 
385 	if (m == NULL) {
386 		DPRINTF("bogus returned buffer from crypto");
387 		ipsecstat_inc(ipsec_crypto);
388 		goto drop;
389 	}
390 
391 	tdb = gettdb(tc->tc_rdomain, tc->tc_spi, &tc->tc_dst, tc->tc_proto);
392 	if (tdb == NULL) {
393 		DPRINTF("TDB is expired while in crypto");
394 		ipsecstat_inc(ipsec_notdb);
395 		goto drop;
396 	}
397 
398 	/* Check for crypto errors */
399 	if (crp->crp_etype) {
400 		if (crp->crp_etype == EAGAIN) {
401 			/* Reset the session ID */
402 			if (tdb->tdb_cryptoid != 0)
403 				tdb->tdb_cryptoid = crp->crp_sid;
404 			error = crypto_dispatch(crp);
405 			if (error) {
406 				DPRINTF("crypto dispatch error %d", error);
407 				goto drop;
408 			}
409 			return;
410 		}
411 		DPRINTF("crypto error %d", crp->crp_etype);
412 		ipsecstat_inc(ipsec_noxform);
413 		goto drop;
414 	}
415 
416 	/* Length of data after processing */
417 	clen = crp->crp_olen;
418 
419 	/* Release the crypto descriptors */
420 	crypto_freereq(crp);
421 
422 	switch (tdb->tdb_sproto) {
423 	case IPPROTO_ESP:
424 		error = esp_input_cb(tdb, tc, m, clen);
425 		break;
426 	case IPPROTO_AH:
427 		error = ah_input_cb(tdb, tc, m, clen);
428 		break;
429 	case IPPROTO_IPCOMP:
430 		error = ipcomp_input_cb(tdb, tc, m, clen);
431 		break;
432 	default:
433 		panic("%s: unknown/unsupported security protocol %d",
434 		    __func__, tdb->tdb_sproto);
435 	}
436 
437 	if (error) {
438 		ipsecstat_inc(ipsec_idrops);
439 		tdb->tdb_idrops++;
440 	}
441 	return;
442 
443  drop:
444 	ipsecstat_inc(ipsec_idrops);
445 	if (tdb != NULL)
446 		tdb->tdb_idrops++;
447 	free(tc, M_XDATA, 0);
448 	m_freem(m);
449 	crypto_freereq(crp);
450 }
451 
452 /*
453  * IPsec input callback, called by the transform callback. Takes care of
454  * filtering and other sanity checks on the processed packet.
455  */
456 int
457 ipsec_common_input_cb(struct mbuf *m, struct tdb *tdbp, int skip, int protoff)
458 {
459 	int af, sproto;
460 	u_int8_t prot;
461 
462 #if NBPFILTER > 0
463 	struct ifnet *encif;
464 #endif
465 
466 	struct ip *ip, ipn;
467 
468 #ifdef INET6
469 	struct ip6_hdr *ip6, ip6n;
470 #endif /* INET6 */
471 	struct m_tag *mtag;
472 	struct tdb_ident *tdbi;
473 
474 #ifdef ENCDEBUG
475 	char buf[INET6_ADDRSTRLEN];
476 #endif
477 
478 	af = tdbp->tdb_dst.sa.sa_family;
479 	sproto = tdbp->tdb_sproto;
480 
481 	tdbp->tdb_last_used = gettime();
482 
483 	/* Sanity check */
484 	if (m == NULL) {
485 		/* The called routine will print a message if necessary */
486 		IPSEC_ISTAT(esps_badkcr, ahs_badkcr, ipcomps_badkcr);
487 		return -1;
488 	}
489 
490 	/* Fix IPv4 header */
491 	if (af == AF_INET) {
492 		if ((m->m_len < skip) && ((m = m_pullup(m, skip)) == NULL)) {
493 			DPRINTF("processing failed for SA %s/%08x",
494 			    ipsp_address(&tdbp->tdb_dst, buf, sizeof(buf)),
495 			    ntohl(tdbp->tdb_spi));
496 			IPSEC_ISTAT(esps_hdrops, ahs_hdrops, ipcomps_hdrops);
497 			return -1;
498 		}
499 
500 		ip = mtod(m, struct ip *);
501 		ip->ip_len = htons(m->m_pkthdr.len);
502 		ip->ip_sum = 0;
503 		ip->ip_sum = in_cksum(m, ip->ip_hl << 2);
504 		prot = ip->ip_p;
505 
506 		/* IP-in-IP encapsulation */
507 		if (prot == IPPROTO_IPIP) {
508 			if (m->m_pkthdr.len - skip < sizeof(struct ip)) {
509 				m_freem(m);
510 				IPSEC_ISTAT(esps_hdrops, ahs_hdrops,
511 				    ipcomps_hdrops);
512 				return -1;
513 			}
514 			/* ipn will now contain the inner IPv4 header */
515 			m_copydata(m, skip, sizeof(struct ip),
516 			    (caddr_t) &ipn);
517 		}
518 
519 #ifdef INET6
520 		/* IPv6-in-IP encapsulation. */
521 		if (prot == IPPROTO_IPV6) {
522 			if (m->m_pkthdr.len - skip < sizeof(struct ip6_hdr)) {
523 				m_freem(m);
524 				IPSEC_ISTAT(esps_hdrops, ahs_hdrops,
525 				    ipcomps_hdrops);
526 				return -1;
527 			}
528 			/* ip6n will now contain the inner IPv6 header. */
529 			m_copydata(m, skip, sizeof(struct ip6_hdr),
530 			    (caddr_t) &ip6n);
531 		}
532 #endif /* INET6 */
533 	}
534 
535 #ifdef INET6
536 	/* Fix IPv6 header */
537 	if (af == AF_INET6)
538 	{
539 		if (m->m_len < sizeof(struct ip6_hdr) &&
540 		    (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
541 
542 			DPRINTF("processing failed for SA %s/%08x",
543 			    ipsp_address(&tdbp->tdb_dst, buf, sizeof(buf)),
544 			    ntohl(tdbp->tdb_spi));
545 			IPSEC_ISTAT(esps_hdrops, ahs_hdrops, ipcomps_hdrops);
546 			return -1;
547 		}
548 
549 		ip6 = mtod(m, struct ip6_hdr *);
550 		ip6->ip6_plen = htons(m->m_pkthdr.len - skip);
551 
552 		/* Save protocol */
553 		m_copydata(m, protoff, 1, (caddr_t) &prot);
554 
555 		/* IP-in-IP encapsulation */
556 		if (prot == IPPROTO_IPIP) {
557 			if (m->m_pkthdr.len - skip < sizeof(struct ip)) {
558 				m_freem(m);
559 				IPSEC_ISTAT(esps_hdrops, ahs_hdrops,
560 				    ipcomps_hdrops);
561 				return -1;
562 			}
563 			/* ipn will now contain the inner IPv4 header */
564 			m_copydata(m, skip, sizeof(struct ip), (caddr_t) &ipn);
565 		}
566 
567 		/* IPv6-in-IP encapsulation */
568 		if (prot == IPPROTO_IPV6) {
569 			if (m->m_pkthdr.len - skip < sizeof(struct ip6_hdr)) {
570 				m_freem(m);
571 				IPSEC_ISTAT(esps_hdrops, ahs_hdrops,
572 				    ipcomps_hdrops);
573 				return -1;
574 			}
575 			/* ip6n will now contain the inner IPv6 header. */
576 			m_copydata(m, skip, sizeof(struct ip6_hdr),
577 			    (caddr_t) &ip6n);
578 		}
579 	}
580 #endif /* INET6 */
581 
582 	/*
583 	 * Fix TCP/UDP checksum of UDP encapsulated transport mode ESP packet.
584 	 * (RFC3948 3.1.2)
585 	 */
586 	if ((af == AF_INET || af == AF_INET6) &&
587 	    (tdbp->tdb_flags & TDBF_UDPENCAP) &&
588 	    (tdbp->tdb_flags & TDBF_TUNNELING) == 0) {
589 		u_int16_t cksum;
590 
591 		switch (prot) {
592 		case IPPROTO_UDP:
593 			if (m->m_pkthdr.len < skip + sizeof(struct udphdr)) {
594 				m_freem(m);
595 				IPSEC_ISTAT(esps_hdrops, ahs_hdrops,
596 				    ipcomps_hdrops);
597 				return -1;
598 			}
599 			cksum = 0;
600 			m_copyback(m, skip + offsetof(struct udphdr, uh_sum),
601 			    sizeof(cksum), &cksum, M_NOWAIT);
602 #ifdef INET6
603 			if (af == AF_INET6) {
604 				cksum = in6_cksum(m, IPPROTO_UDP, skip,
605 				    m->m_pkthdr.len - skip);
606 				m_copyback(m, skip + offsetof(struct udphdr,
607 				    uh_sum), sizeof(cksum), &cksum, M_NOWAIT);
608 			}
609 #endif
610 			break;
611 		case IPPROTO_TCP:
612 			if (m->m_pkthdr.len < skip + sizeof(struct tcphdr)) {
613 				m_freem(m);
614 				IPSEC_ISTAT(esps_hdrops, ahs_hdrops,
615 				    ipcomps_hdrops);
616 				return -1;
617 			}
618 			cksum = 0;
619 			m_copyback(m, skip + offsetof(struct tcphdr, th_sum),
620 			    sizeof(cksum), &cksum, M_NOWAIT);
621 			if (af == AF_INET)
622 				cksum = in4_cksum(m, IPPROTO_TCP, skip,
623 				    m->m_pkthdr.len - skip);
624 #ifdef INET6
625 			else if (af == AF_INET6)
626 				cksum = in6_cksum(m, IPPROTO_TCP, skip,
627 				    m->m_pkthdr.len - skip);
628 #endif
629 			m_copyback(m, skip + offsetof(struct tcphdr, th_sum),
630 			    sizeof(cksum), &cksum, M_NOWAIT);
631 			break;
632 		}
633 	}
634 
635 	/*
636 	 * Record what we've done to the packet (under what SA it was
637 	 * processed).
638 	 */
639 	if (tdbp->tdb_sproto != IPPROTO_IPCOMP) {
640 		mtag = m_tag_get(PACKET_TAG_IPSEC_IN_DONE,
641 		    sizeof(struct tdb_ident), M_NOWAIT);
642 		if (mtag == NULL) {
643 			m_freem(m);
644 			DPRINTF("failed to get tag");
645 			IPSEC_ISTAT(esps_hdrops, ahs_hdrops, ipcomps_hdrops);
646 			return -1;
647 		}
648 
649 		tdbi = (struct tdb_ident *)(mtag + 1);
650 		tdbi->dst = tdbp->tdb_dst;
651 		tdbi->proto = tdbp->tdb_sproto;
652 		tdbi->spi = tdbp->tdb_spi;
653 		tdbi->rdomain = tdbp->tdb_rdomain;
654 
655 		m_tag_prepend(m, mtag);
656 	}
657 
658 	switch (sproto) {
659 	case IPPROTO_ESP:
660 		/* Packet is confidential ? */
661 		if (tdbp->tdb_encalgxform)
662 			m->m_flags |= M_CONF;
663 
664 		/* Check if we had authenticated ESP. */
665 		if (tdbp->tdb_authalgxform)
666 			m->m_flags |= M_AUTH;
667 		break;
668 	case IPPROTO_AH:
669 		m->m_flags |= M_AUTH;
670 		break;
671 	case IPPROTO_IPCOMP:
672 		m->m_flags |= M_COMP;
673 		break;
674 	default:
675 		panic("%s: unknown/unsupported security protocol %d",
676 		    __func__, sproto);
677 	}
678 
679 #if NPF > 0
680 	/* Add pf tag if requested. */
681 	pf_tag_packet(m, tdbp->tdb_tag, -1);
682 	pf_pkt_addr_changed(m);
683 #endif
684 	if (tdbp->tdb_rdomain != tdbp->tdb_rdomain_post)
685 		m->m_pkthdr.ph_rtableid = tdbp->tdb_rdomain_post;
686 
687 	if (tdbp->tdb_flags & TDBF_TUNNELING)
688 		m->m_flags |= M_TUNNEL;
689 
690 	ipsecstat_add(ipsec_idecompbytes, m->m_pkthdr.len);
691 	tdbp->tdb_idecompbytes += m->m_pkthdr.len;
692 
693 #if NBPFILTER > 0
694 	if ((encif = enc_getif(tdbp->tdb_rdomain_post, tdbp->tdb_tap)) != NULL) {
695 		encif->if_ipackets++;
696 		encif->if_ibytes += m->m_pkthdr.len;
697 
698 		if (encif->if_bpf) {
699 			struct enchdr hdr;
700 
701 			hdr.af = af;
702 			hdr.spi = tdbp->tdb_spi;
703 			hdr.flags = m->m_flags & (M_AUTH|M_CONF);
704 
705 			bpf_mtap_hdr(encif->if_bpf, (char *)&hdr,
706 			    ENC_HDRLEN, m, BPF_DIRECTION_IN);
707 		}
708 	}
709 #endif
710 
711 #if NPF > 0
712 	/*
713 	 * The ip_deliver() shortcut avoids running through ip_input() with the
714 	 * same IP header twice.  Packets in transport mode have to be be
715 	 * passed to pf explicitly.  In tunnel mode the inner IP header will
716 	 * run through ip_input() and pf anyway.
717 	 */
718 	if ((tdbp->tdb_flags & TDBF_TUNNELING) == 0) {
719 		struct ifnet *ifp;
720 
721 		/* This is the enc0 interface unless for ipcomp. */
722 		if ((ifp = if_get(m->m_pkthdr.ph_ifidx)) == NULL) {
723 			m_freem(m);
724 			return -1;
725 		}
726 		if (pf_test(af, PF_IN, ifp, &m) != PF_PASS) {
727 			if_put(ifp);
728 			m_freem(m);
729 			return -1;
730 		}
731 		if_put(ifp);
732 		if (m == NULL)
733 			return -1;
734 	}
735 #endif
736 	/* Call the appropriate IPsec transform callback. */
737 	ip_deliver(&m, &skip, prot, af);
738 	return 0;
739 #undef IPSEC_ISTAT
740 }
741 
742 int
743 ipsec_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
744     size_t newlen)
745 {
746 	int error;
747 
748 	switch (name[0]) {
749 	case IPCTL_IPSEC_ENC_ALGORITHM:
750 		NET_LOCK();
751 		error = sysctl_tstring(oldp, oldlenp, newp, newlen,
752 		    ipsec_def_enc, sizeof(ipsec_def_enc));
753 		NET_UNLOCK();
754 		return (error);
755 	case IPCTL_IPSEC_AUTH_ALGORITHM:
756 		NET_LOCK();
757 		error = sysctl_tstring(oldp, oldlenp, newp, newlen,
758 		    ipsec_def_auth, sizeof(ipsec_def_auth));
759 		NET_UNLOCK();
760 		return (error);
761 	case IPCTL_IPSEC_IPCOMP_ALGORITHM:
762 		NET_LOCK();
763 		error = sysctl_tstring(oldp, oldlenp, newp, newlen,
764 		    ipsec_def_comp, sizeof(ipsec_def_comp));
765 		NET_UNLOCK();
766 		return (error);
767 	case IPCTL_IPSEC_STATS:
768 		return (ipsec_sysctl_ipsecstat(oldp, oldlenp, newp));
769 	default:
770 		NET_LOCK();
771 		error = sysctl_bounded_arr(ipsecctl_vars, nitems(ipsecctl_vars),
772 		    name, namelen, oldp, oldlenp, newp, newlen);
773 		NET_UNLOCK();
774 		return (error);
775 	}
776 }
777 
778 int
779 esp_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
780     size_t newlen)
781 {
782 	int error;
783 
784 	/* All sysctl names at this level are terminal. */
785 	if (namelen != 1)
786 		return (ENOTDIR);
787 
788 	switch (name[0]) {
789 	case ESPCTL_STATS:
790 		return (esp_sysctl_espstat(oldp, oldlenp, newp));
791 	default:
792 		NET_LOCK();
793 		error = sysctl_bounded_arr(espctl_vars, nitems(espctl_vars),
794 		    name, namelen, oldp, oldlenp, newp, newlen);
795 		NET_UNLOCK();
796 		return (error);
797 	}
798 }
799 
800 int
801 esp_sysctl_espstat(void *oldp, size_t *oldlenp, void *newp)
802 {
803 	struct espstat espstat;
804 
805 	CTASSERT(sizeof(espstat) == (esps_ncounters * sizeof(uint64_t)));
806 	memset(&espstat, 0, sizeof espstat);
807 	counters_read(espcounters, (uint64_t *)&espstat, esps_ncounters);
808 	return (sysctl_rdstruct(oldp, oldlenp, newp, &espstat,
809 	    sizeof(espstat)));
810 }
811 
812 int
813 ah_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
814     size_t newlen)
815 {
816 	int error;
817 
818 	/* All sysctl names at this level are terminal. */
819 	if (namelen != 1)
820 		return (ENOTDIR);
821 
822 	switch (name[0]) {
823 	case AHCTL_STATS:
824 		return ah_sysctl_ahstat(oldp, oldlenp, newp);
825 	default:
826 		NET_LOCK();
827 		error = sysctl_bounded_arr(ahctl_vars, nitems(ahctl_vars), name,
828 		    namelen, oldp, oldlenp, newp, newlen);
829 		NET_UNLOCK();
830 		return (error);
831 	}
832 }
833 
834 int
835 ah_sysctl_ahstat(void *oldp, size_t *oldlenp, void *newp)
836 {
837 	struct ahstat ahstat;
838 
839 	CTASSERT(sizeof(ahstat) == (ahs_ncounters * sizeof(uint64_t)));
840 	memset(&ahstat, 0, sizeof ahstat);
841 	counters_read(ahcounters, (uint64_t *)&ahstat, ahs_ncounters);
842 	return (sysctl_rdstruct(oldp, oldlenp, newp, &ahstat, sizeof(ahstat)));
843 }
844 
845 int
846 ipcomp_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
847     size_t newlen)
848 {
849 	int error;
850 
851 	/* All sysctl names at this level are terminal. */
852 	if (namelen != 1)
853 		return (ENOTDIR);
854 
855 	switch (name[0]) {
856 	case IPCOMPCTL_STATS:
857 		return ipcomp_sysctl_ipcompstat(oldp, oldlenp, newp);
858 	default:
859 		NET_LOCK();
860 		error = sysctl_bounded_arr(ipcompctl_vars,
861 		    nitems(ipcompctl_vars), name, namelen, oldp, oldlenp,
862 		    newp, newlen);
863 		NET_UNLOCK();
864 		return (error);
865 	}
866 }
867 
868 int
869 ipcomp_sysctl_ipcompstat(void *oldp, size_t *oldlenp, void *newp)
870 {
871 	struct ipcompstat ipcompstat;
872 
873 	CTASSERT(sizeof(ipcompstat) == (ipcomps_ncounters * sizeof(uint64_t)));
874 	memset(&ipcompstat, 0, sizeof ipcompstat);
875 	counters_read(ipcompcounters, (uint64_t *)&ipcompstat,
876 	    ipcomps_ncounters);
877 	return (sysctl_rdstruct(oldp, oldlenp, newp, &ipcompstat,
878 	    sizeof(ipcompstat)));
879 }
880 
881 int
882 ipsec_sysctl_ipsecstat(void *oldp, size_t *oldlenp, void *newp)
883 {
884 	struct ipsecstat ipsecstat;
885 
886 	CTASSERT(sizeof(ipsecstat) == (ipsec_ncounters * sizeof(uint64_t)));
887 	memset(&ipsecstat, 0, sizeof ipsecstat);
888 	counters_read(ipseccounters, (uint64_t *)&ipsecstat, ipsec_ncounters);
889 	return (sysctl_rdstruct(oldp, oldlenp, newp, &ipsecstat,
890 	    sizeof(ipsecstat)));
891 }
892 
893 /* IPv4 AH wrapper. */
894 int
895 ah4_input(struct mbuf **mp, int *offp, int proto, int af)
896 {
897 	if (
898 #if NPF > 0
899 	    ((*mp)->m_pkthdr.pf.flags & PF_TAG_DIVERTED) ||
900 #endif
901 	    !ah_enable)
902 		return rip_input(mp, offp, proto, af);
903 
904 	ipsec_common_input(*mp, *offp, offsetof(struct ip, ip_p), AF_INET,
905 	    proto, 0);
906 	return IPPROTO_DONE;
907 }
908 
909 void
910 ah4_ctlinput(int cmd, struct sockaddr *sa, u_int rdomain, void *v)
911 {
912 	if (sa->sa_family != AF_INET ||
913 	    sa->sa_len != sizeof(struct sockaddr_in))
914 		return;
915 
916 	ipsec_common_ctlinput(rdomain, cmd, sa, v, IPPROTO_AH);
917 }
918 
919 /* IPv4 ESP wrapper. */
920 int
921 esp4_input(struct mbuf **mp, int *offp, int proto, int af)
922 {
923 	if (
924 #if NPF > 0
925 	    ((*mp)->m_pkthdr.pf.flags & PF_TAG_DIVERTED) ||
926 #endif
927 	    !esp_enable)
928 		return rip_input(mp, offp, proto, af);
929 
930 	ipsec_common_input(*mp, *offp, offsetof(struct ip, ip_p), AF_INET,
931 	    proto, 0);
932 	return IPPROTO_DONE;
933 }
934 
935 /* IPv4 IPCOMP wrapper */
936 int
937 ipcomp4_input(struct mbuf **mp, int *offp, int proto, int af)
938 {
939 	if (
940 #if NPF > 0
941 	    ((*mp)->m_pkthdr.pf.flags & PF_TAG_DIVERTED) ||
942 #endif
943 	    !ipcomp_enable)
944 		return rip_input(mp, offp, proto, af);
945 
946 	ipsec_common_input(*mp, *offp, offsetof(struct ip, ip_p), AF_INET,
947 	    proto, 0);
948 	return IPPROTO_DONE;
949 }
950 
951 void
952 ipsec_set_mtu(struct tdb *tdbp, u_int32_t mtu)
953 {
954 	ssize_t adjust;
955 
956 	NET_ASSERT_LOCKED();
957 
958 	/* Walk the chain backwards to the first tdb */
959 	for (; tdbp != NULL; tdbp = tdbp->tdb_inext) {
960 		if (tdbp->tdb_flags & TDBF_INVALID ||
961 		    (adjust = ipsec_hdrsz(tdbp)) == -1)
962 			return;
963 
964 		mtu -= adjust;
965 
966 		/* Store adjusted MTU in tdb */
967 		tdbp->tdb_mtu = mtu;
968 		tdbp->tdb_mtutimeout = gettime() + ip_mtudisc_timeout;
969 		DPRINTF("spi %08x mtu %d adjust %ld",
970 		    ntohl(tdbp->tdb_spi), tdbp->tdb_mtu, adjust);
971 	}
972 }
973 
974 void
975 ipsec_common_ctlinput(u_int rdomain, int cmd, struct sockaddr *sa,
976     void *v, int proto)
977 {
978 	struct ip *ip = v;
979 
980 	if (cmd == PRC_MSGSIZE && ip && ip_mtudisc && ip->ip_v == 4) {
981 		struct tdb *tdbp;
982 		struct sockaddr_in dst;
983 		struct icmp *icp;
984 		int hlen = ip->ip_hl << 2;
985 		u_int32_t spi, mtu;
986 
987 		/* Find the right MTU. */
988 		icp = (struct icmp *)((caddr_t) ip -
989 		    offsetof(struct icmp, icmp_ip));
990 		mtu = ntohs(icp->icmp_nextmtu);
991 
992 		/*
993 		 * Ignore the packet, if we do not receive a MTU
994 		 * or the MTU is too small to be acceptable.
995 		 */
996 		if (mtu < 296)
997 			return;
998 
999 		memset(&dst, 0, sizeof(struct sockaddr_in));
1000 		dst.sin_family = AF_INET;
1001 		dst.sin_len = sizeof(struct sockaddr_in);
1002 		dst.sin_addr.s_addr = ip->ip_dst.s_addr;
1003 
1004 		memcpy(&spi, (caddr_t)ip + hlen, sizeof(u_int32_t));
1005 
1006 		tdbp = gettdb_rev(rdomain, spi, (union sockaddr_union *)&dst,
1007 		    proto);
1008 		ipsec_set_mtu(tdbp, mtu);
1009 	}
1010 }
1011 
1012 void
1013 udpencap_ctlinput(int cmd, struct sockaddr *sa, u_int rdomain, void *v)
1014 {
1015 	struct ip *ip = v;
1016 	struct tdb *tdbp;
1017 	struct icmp *icp;
1018 	u_int32_t mtu;
1019 	struct sockaddr_in dst, src;
1020 	union sockaddr_union *su_dst, *su_src;
1021 
1022 	NET_ASSERT_LOCKED();
1023 
1024 	icp = (struct icmp *)((caddr_t) ip - offsetof(struct icmp, icmp_ip));
1025 	mtu = ntohs(icp->icmp_nextmtu);
1026 
1027 	/*
1028 	 * Ignore the packet, if we do not receive a MTU
1029 	 * or the MTU is too small to be acceptable.
1030 	 */
1031 	if (mtu < 296)
1032 		return;
1033 
1034 	memset(&dst, 0, sizeof(dst));
1035 	dst.sin_family = AF_INET;
1036 	dst.sin_len = sizeof(struct sockaddr_in);
1037 	dst.sin_addr.s_addr = ip->ip_dst.s_addr;
1038 	su_dst = (union sockaddr_union *)&dst;
1039 	memset(&src, 0, sizeof(src));
1040 	src.sin_family = AF_INET;
1041 	src.sin_len = sizeof(struct sockaddr_in);
1042 	src.sin_addr.s_addr = ip->ip_src.s_addr;
1043 	su_src = (union sockaddr_union *)&src;
1044 
1045 	tdbp = gettdbbysrcdst_rev(rdomain, 0, su_src, su_dst,
1046 	    IPPROTO_ESP);
1047 
1048 	for (; tdbp != NULL; tdbp = tdbp->tdb_snext) {
1049 		if (tdbp->tdb_sproto == IPPROTO_ESP &&
1050 		    ((tdbp->tdb_flags & (TDBF_INVALID|TDBF_UDPENCAP)) ==
1051 		    TDBF_UDPENCAP) &&
1052 		    !memcmp(&tdbp->tdb_dst, &dst, su_dst->sa.sa_len) &&
1053 		    !memcmp(&tdbp->tdb_src, &src, su_src->sa.sa_len)) {
1054 			ipsec_set_mtu(tdbp, mtu);
1055 		}
1056 	}
1057 }
1058 
1059 void
1060 esp4_ctlinput(int cmd, struct sockaddr *sa, u_int rdomain, void *v)
1061 {
1062 	if (sa->sa_family != AF_INET ||
1063 	    sa->sa_len != sizeof(struct sockaddr_in))
1064 		return;
1065 
1066 	ipsec_common_ctlinput(rdomain, cmd, sa, v, IPPROTO_ESP);
1067 }
1068 
1069 #ifdef INET6
1070 /* IPv6 AH wrapper. */
1071 int
1072 ah6_input(struct mbuf **mp, int *offp, int proto, int af)
1073 {
1074 	int l = 0;
1075 	int protoff, nxt;
1076 	struct ip6_ext ip6e;
1077 
1078 	if (
1079 #if NPF > 0
1080 	    ((*mp)->m_pkthdr.pf.flags & PF_TAG_DIVERTED) ||
1081 #endif
1082 	    !ah_enable)
1083 		return rip6_input(mp, offp, proto, af);
1084 
1085 	if (*offp < sizeof(struct ip6_hdr)) {
1086 		DPRINTF("bad offset");
1087 		ahstat_inc(ahs_hdrops);
1088 		m_freemp(mp);
1089 		return IPPROTO_DONE;
1090 	} else if (*offp == sizeof(struct ip6_hdr)) {
1091 		protoff = offsetof(struct ip6_hdr, ip6_nxt);
1092 	} else {
1093 		/* Chase down the header chain... */
1094 		protoff = sizeof(struct ip6_hdr);
1095 		nxt = (mtod(*mp, struct ip6_hdr *))->ip6_nxt;
1096 
1097 		do {
1098 			protoff += l;
1099 			m_copydata(*mp, protoff, sizeof(ip6e),
1100 			    (caddr_t) &ip6e);
1101 
1102 			if (nxt == IPPROTO_AH)
1103 				l = (ip6e.ip6e_len + 2) << 2;
1104 			else
1105 				l = (ip6e.ip6e_len + 1) << 3;
1106 #ifdef DIAGNOSTIC
1107 			if (l <= 0)
1108 				panic("ah6_input: l went zero or negative");
1109 #endif
1110 
1111 			nxt = ip6e.ip6e_nxt;
1112 		} while (protoff + l < *offp);
1113 
1114 		/* Malformed packet check */
1115 		if (protoff + l != *offp) {
1116 			DPRINTF("bad packet header chain");
1117 			ahstat_inc(ahs_hdrops);
1118 			m_freemp(mp);
1119 			return IPPROTO_DONE;
1120 		}
1121 		protoff += offsetof(struct ip6_ext, ip6e_nxt);
1122 	}
1123 	ipsec_common_input(*mp, *offp, protoff, AF_INET6, proto, 0);
1124 	return IPPROTO_DONE;
1125 }
1126 
1127 /* IPv6 ESP wrapper. */
1128 int
1129 esp6_input(struct mbuf **mp, int *offp, int proto, int af)
1130 {
1131 	int l = 0;
1132 	int protoff, nxt;
1133 	struct ip6_ext ip6e;
1134 
1135 	if (
1136 #if NPF > 0
1137 	    ((*mp)->m_pkthdr.pf.flags & PF_TAG_DIVERTED) ||
1138 #endif
1139 	    !esp_enable)
1140 		return rip6_input(mp, offp, proto, af);
1141 
1142 	if (*offp < sizeof(struct ip6_hdr)) {
1143 		DPRINTF("bad offset");
1144 		espstat_inc(esps_hdrops);
1145 		m_freemp(mp);
1146 		return IPPROTO_DONE;
1147 	} else if (*offp == sizeof(struct ip6_hdr)) {
1148 		protoff = offsetof(struct ip6_hdr, ip6_nxt);
1149 	} else {
1150 		/* Chase down the header chain... */
1151 		protoff = sizeof(struct ip6_hdr);
1152 		nxt = (mtod(*mp, struct ip6_hdr *))->ip6_nxt;
1153 
1154 		do {
1155 			protoff += l;
1156 			m_copydata(*mp, protoff, sizeof(ip6e),
1157 			    (caddr_t) &ip6e);
1158 
1159 			if (nxt == IPPROTO_AH)
1160 				l = (ip6e.ip6e_len + 2) << 2;
1161 			else
1162 				l = (ip6e.ip6e_len + 1) << 3;
1163 #ifdef DIAGNOSTIC
1164 			if (l <= 0)
1165 				panic("esp6_input: l went zero or negative");
1166 #endif
1167 
1168 			nxt = ip6e.ip6e_nxt;
1169 		} while (protoff + l < *offp);
1170 
1171 		/* Malformed packet check */
1172 		if (protoff + l != *offp) {
1173 			DPRINTF("bad packet header chain");
1174 			espstat_inc(esps_hdrops);
1175 			m_freemp(mp);
1176 			return IPPROTO_DONE;
1177 		}
1178 		protoff += offsetof(struct ip6_ext, ip6e_nxt);
1179 	}
1180 	ipsec_common_input(*mp, *offp, protoff, AF_INET6, proto, 0);
1181 	return IPPROTO_DONE;
1182 
1183 }
1184 
1185 /* IPv6 IPcomp wrapper */
1186 int
1187 ipcomp6_input(struct mbuf **mp, int *offp, int proto, int af)
1188 {
1189 	int l = 0;
1190 	int protoff, nxt;
1191 	struct ip6_ext ip6e;
1192 
1193 	if (
1194 #if NPF > 0
1195 	    ((*mp)->m_pkthdr.pf.flags & PF_TAG_DIVERTED) ||
1196 #endif
1197 	    !ipcomp_enable)
1198 		return rip6_input(mp, offp, proto, af);
1199 
1200 	if (*offp < sizeof(struct ip6_hdr)) {
1201 		DPRINTF("bad offset");
1202 		ipcompstat_inc(ipcomps_hdrops);
1203 		m_freemp(mp);
1204 		return IPPROTO_DONE;
1205 	} else if (*offp == sizeof(struct ip6_hdr)) {
1206 		protoff = offsetof(struct ip6_hdr, ip6_nxt);
1207 	} else {
1208 		/* Chase down the header chain... */
1209 		protoff = sizeof(struct ip6_hdr);
1210 		nxt = (mtod(*mp, struct ip6_hdr *))->ip6_nxt;
1211 
1212 		do {
1213 			protoff += l;
1214 			m_copydata(*mp, protoff, sizeof(ip6e),
1215 			    (caddr_t) &ip6e);
1216 			if (nxt == IPPROTO_AH)
1217 				l = (ip6e.ip6e_len + 2) << 2;
1218 			else
1219 				l = (ip6e.ip6e_len + 1) << 3;
1220 #ifdef DIAGNOSTIC
1221 			if (l <= 0)
1222 				panic("l went zero or negative");
1223 #endif
1224 
1225 			nxt = ip6e.ip6e_nxt;
1226 		} while (protoff + l < *offp);
1227 
1228 		/* Malformed packet check */
1229 		if (protoff + l != *offp) {
1230 			DPRINTF("bad packet header chain");
1231 			ipcompstat_inc(ipcomps_hdrops);
1232 			m_freemp(mp);
1233 			return IPPROTO_DONE;
1234 		}
1235 
1236 		protoff += offsetof(struct ip6_ext, ip6e_nxt);
1237 	}
1238 	ipsec_common_input(*mp, *offp, protoff, AF_INET6, proto, 0);
1239 	return IPPROTO_DONE;
1240 }
1241 #endif /* INET6 */
1242 
1243 int
1244 ipsec_forward_check(struct mbuf *m, int hlen, int af)
1245 {
1246 	struct tdb *tdb;
1247 	struct tdb_ident *tdbi;
1248 	struct m_tag *mtag;
1249 	int error = 0;
1250 
1251 	/*
1252 	 * IPsec policy check for forwarded packets. Look at
1253 	 * inner-most IPsec SA used.
1254 	 */
1255 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
1256 	if (mtag != NULL) {
1257 		tdbi = (struct tdb_ident *)(mtag + 1);
1258 		tdb = gettdb(tdbi->rdomain, tdbi->spi, &tdbi->dst, tdbi->proto);
1259 	} else
1260 		tdb = NULL;
1261 	ipsp_spd_lookup(m, af, hlen, &error, IPSP_DIRECTION_IN, tdb, NULL, 0);
1262 
1263 	return error;
1264 }
1265 
1266 int
1267 ipsec_local_check(struct mbuf *m, int hlen, int proto, int af)
1268 {
1269 	struct tdb *tdb;
1270 	struct tdb_ident *tdbi;
1271 	struct m_tag *mtag;
1272 	int error = 0;
1273 
1274 	/*
1275 	 * If it's a protected packet for us, skip the policy check.
1276 	 * That's because we really only care about the properties of
1277 	 * the protected packet, and not the intermediate versions.
1278 	 * While this is not the most paranoid setting, it allows
1279 	 * some flexibility in handling nested tunnels (in setting up
1280 	 * the policies).
1281 	 */
1282 	if ((proto == IPPROTO_ESP) || (proto == IPPROTO_AH) ||
1283 	    (proto == IPPROTO_IPCOMP))
1284 		return 0;
1285 
1286 	/*
1287 	 * If the protected packet was tunneled, then we need to
1288 	 * verify the protected packet's information, not the
1289 	 * external headers. Thus, skip the policy lookup for the
1290 	 * external packet, and keep the IPsec information linked on
1291 	 * the packet header (the encapsulation routines know how
1292 	 * to deal with that).
1293 	 */
1294 	if ((proto == IPPROTO_IPV4) || (proto == IPPROTO_IPV6))
1295 		return 0;
1296 
1297 	/*
1298 	 * When processing IPv6 header chains, do not look at the
1299 	 * outer header.  The inner protocol is relevant and will
1300 	 * be checked by the local delivery loop later.
1301 	 */
1302 	if ((af == AF_INET6) && ((proto == IPPROTO_DSTOPTS) ||
1303 	    (proto == IPPROTO_ROUTING) || (proto == IPPROTO_FRAGMENT)))
1304 		return 0;
1305 
1306 	/*
1307 	 * If the protected packet is TCP or UDP, we'll do the
1308 	 * policy check in the respective input routine, so we can
1309 	 * check for bypass sockets.
1310 	 */
1311 	if ((proto == IPPROTO_TCP) || (proto == IPPROTO_UDP))
1312 		return 0;
1313 
1314 	/*
1315 	 * IPsec policy check for local-delivery packets. Look at the
1316 	 * inner-most SA that protected the packet. This is in fact
1317 	 * a bit too restrictive (it could end up causing packets to
1318 	 * be dropped that semantically follow the policy, e.g., in
1319 	 * certain SA-bundle configurations); but the alternative is
1320 	 * very complicated (and requires keeping track of what
1321 	 * kinds of tunneling headers have been seen in-between the
1322 	 * IPsec headers), and I don't think we lose much functionality
1323 	 * that's needed in the real world (who uses bundles anyway ?).
1324 	 */
1325 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
1326 	if (mtag) {
1327 		tdbi = (struct tdb_ident *)(mtag + 1);
1328 		tdb = gettdb(tdbi->rdomain, tdbi->spi, &tdbi->dst,
1329 		    tdbi->proto);
1330 	} else
1331 		tdb = NULL;
1332 	ipsp_spd_lookup(m, af, hlen, &error, IPSP_DIRECTION_IN,
1333 	    tdb, NULL, 0);
1334 
1335 	return error;
1336 }
1337