xref: /freebsd/sys/netipsec/ipsec.c (revision ea825d02749f382c3f7e17f28247f20a48733eab)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ipsec.c,v 1.103 2001/05/24 07:14:18 sakane Exp $	*/
3 
4 /*-
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * IPsec controller part.
35  */
36 
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39 #include "opt_ipsec.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/domain.h>
46 #include <sys/priv.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/errno.h>
51 #include <sys/hhook.h>
52 #include <sys/time.h>
53 #include <sys/kernel.h>
54 #include <sys/syslog.h>
55 #include <sys/sysctl.h>
56 #include <sys/proc.h>
57 
58 #include <net/if.h>
59 #include <net/if_enc.h>
60 #include <net/if_var.h>
61 #include <net/vnet.h>
62 
63 #include <netinet/in.h>
64 #include <netinet/in_systm.h>
65 #include <netinet/ip.h>
66 #include <netinet/ip_var.h>
67 #include <netinet/in_var.h>
68 #include <netinet/udp.h>
69 #include <netinet/udp_var.h>
70 #include <netinet/tcp.h>
71 #include <netinet/udp.h>
72 
73 #include <netinet/ip6.h>
74 #ifdef INET6
75 #include <netinet6/ip6_var.h>
76 #endif
77 #include <netinet/in_pcb.h>
78 #ifdef INET6
79 #include <netinet/icmp6.h>
80 #endif
81 
82 #include <sys/types.h>
83 #include <netipsec/ipsec.h>
84 #ifdef INET6
85 #include <netipsec/ipsec6.h>
86 #endif
87 #include <netipsec/ah_var.h>
88 #include <netipsec/esp_var.h>
89 #include <netipsec/ipcomp.h>		/*XXX*/
90 #include <netipsec/ipcomp_var.h>
91 #include <netipsec/ipsec_support.h>
92 
93 #include <netipsec/key.h>
94 #include <netipsec/keydb.h>
95 #include <netipsec/key_debug.h>
96 
97 #include <netipsec/xform.h>
98 
99 #include <machine/in_cksum.h>
100 
101 #include <opencrypto/cryptodev.h>
102 
103 /* NB: name changed so netstat doesn't use it. */
104 VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec4stat);
105 VNET_PCPUSTAT_SYSINIT(ipsec4stat);
106 
107 #ifdef VIMAGE
108 VNET_PCPUSTAT_SYSUNINIT(ipsec4stat);
109 #endif /* VIMAGE */
110 
111 VNET_DEFINE(int, ip4_ah_offsetmask) = 0;	/* maybe IP_DF? */
112 /* DF bit on encap. 0: clear 1: set 2: copy */
113 VNET_DEFINE(int, ip4_ipsec_dfbit) = 0;
114 VNET_DEFINE(int, ip4_esp_trans_deflev) = IPSEC_LEVEL_USE;
115 VNET_DEFINE(int, ip4_esp_net_deflev) = IPSEC_LEVEL_USE;
116 VNET_DEFINE(int, ip4_ah_trans_deflev) = IPSEC_LEVEL_USE;
117 VNET_DEFINE(int, ip4_ah_net_deflev) = IPSEC_LEVEL_USE;
118 /* ECN ignore(-1)/forbidden(0)/allowed(1) */
119 VNET_DEFINE(int, ip4_ipsec_ecn) = 0;
120 VNET_DEFINE(int, ip4_esp_randpad) = -1;
121 
122 static VNET_DEFINE(int, ip4_filtertunnel) = 0;
123 #define	V_ip4_filtertunnel VNET(ip4_filtertunnel)
124 static VNET_DEFINE(int, check_policy_history) = 0;
125 #define	V_check_policy_history	VNET(check_policy_history)
126 static VNET_DEFINE(struct secpolicy *, def_policy) = NULL;
127 #define	V_def_policy	VNET(def_policy)
128 static int
129 sysctl_def_policy(SYSCTL_HANDLER_ARGS)
130 {
131 	int error, value;
132 
133 	value = V_def_policy->policy;
134 	error = sysctl_handle_int(oidp, &value, 0, req);
135 	if (error == 0) {
136 		if (value != IPSEC_POLICY_DISCARD &&
137 		    value != IPSEC_POLICY_NONE)
138 			return (EINVAL);
139 		V_def_policy->policy = value;
140 	}
141 	return (error);
142 }
143 
144 /*
145  * Crypto support requirements:
146  *
147  *  1	require hardware support
148  * -1	require software support
149  *  0	take anything
150  */
151 VNET_DEFINE(int, crypto_support) = CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE;
152 /*
153  * TCP/UDP checksum handling policy for transport mode NAT-T (RFC3948)
154  *
155  * 0 - auto: incrementally recompute, when checksum delta is known;
156  *     if checksum delta isn't known, reset checksum to zero for UDP,
157  *     and mark csum_flags as valid for TCP.
158  * 1 - fully recompute TCP/UDP checksum.
159  */
160 VNET_DEFINE(int, natt_cksum_policy) = 0;
161 
162 FEATURE(ipsec, "Internet Protocol Security (IPsec)");
163 FEATURE(ipsec_natt, "UDP Encapsulation of IPsec ESP Packets ('NAT-T')");
164 
165 SYSCTL_DECL(_net_inet_ipsec);
166 
167 /* net.inet.ipsec */
168 SYSCTL_PROC(_net_inet_ipsec, IPSECCTL_DEF_POLICY, def_policy,
169 	CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RW, 0, 0, sysctl_def_policy, "I",
170 	"IPsec default policy.");
171 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
172 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_esp_trans_deflev), 0,
173 	"Default ESP transport mode level");
174 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
175 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_esp_net_deflev), 0,
176 	"Default ESP tunnel mode level.");
177 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
178 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_trans_deflev), 0,
179 	"AH transfer mode default level.");
180 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
181 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_net_deflev), 0,
182 	"AH tunnel mode default level.");
183 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_CLEARTOS, ah_cleartos,
184 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ah_cleartos), 0,
185 	"If set, clear type-of-service field when doing AH computation.");
186 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_OFFSETMASK, ah_offsetmask,
187 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_offsetmask), 0,
188 	"If not set, clear offset field mask when doing AH computation.");
189 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DFBIT, dfbit,
190 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_dfbit), 0,
191 	"Do not fragment bit on encap.");
192 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_ECN, ecn,
193 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_ecn), 0,
194 	"Explicit Congestion Notification handling.");
195 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, crypto_support,
196 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(crypto_support), 0,
197 	"Crypto driver selection.");
198 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, check_policy_history,
199 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(check_policy_history), 0,
200 	"Use strict check of inbound packets to security policy compliance.");
201 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, natt_cksum_policy,
202 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(natt_cksum_policy), 0,
203 	"Method to fix TCP/UDP checksum for transport mode IPsec after NAT.");
204 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, filtertunnel,
205 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_filtertunnel), 0,
206 	"If set, filter packets from an IPsec tunnel.");
207 SYSCTL_VNET_PCPUSTAT(_net_inet_ipsec, OID_AUTO, ipsecstats, struct ipsecstat,
208     ipsec4stat, "IPsec IPv4 statistics.");
209 
210 #ifdef REGRESSION
211 /*
212  * When set to 1, IPsec will send packets with the same sequence number.
213  * This allows to verify if the other side has proper replay attacks detection.
214  */
215 VNET_DEFINE(int, ipsec_replay) = 0;
216 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_replay,
217 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_replay), 0,
218 	"Emulate replay attack");
219 /*
220  * When set 1, IPsec will send packets with corrupted HMAC.
221  * This allows to verify if the other side properly detects modified packets.
222  */
223 VNET_DEFINE(int, ipsec_integrity) = 0;
224 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_integrity,
225 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_integrity), 0,
226 	"Emulate man-in-the-middle attack");
227 #endif
228 
229 #ifdef INET6
230 VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec6stat);
231 VNET_PCPUSTAT_SYSINIT(ipsec6stat);
232 
233 #ifdef VIMAGE
234 VNET_PCPUSTAT_SYSUNINIT(ipsec6stat);
235 #endif /* VIMAGE */
236 
237 VNET_DEFINE(int, ip6_esp_trans_deflev) = IPSEC_LEVEL_USE;
238 VNET_DEFINE(int, ip6_esp_net_deflev) = IPSEC_LEVEL_USE;
239 VNET_DEFINE(int, ip6_ah_trans_deflev) = IPSEC_LEVEL_USE;
240 VNET_DEFINE(int, ip6_ah_net_deflev) = IPSEC_LEVEL_USE;
241 VNET_DEFINE(int, ip6_ipsec_ecn) = 0;	/* ECN ignore(-1)/forbidden(0)/allowed(1) */
242 
243 static VNET_DEFINE(int, ip6_filtertunnel) = 0;
244 #define	V_ip6_filtertunnel	VNET(ip6_filtertunnel)
245 
246 SYSCTL_DECL(_net_inet6_ipsec6);
247 
248 /* net.inet6.ipsec6 */
249 SYSCTL_PROC(_net_inet6_ipsec6, IPSECCTL_DEF_POLICY, def_policy,
250 	CTLTYPE_INT | CTLFLAG_VNET | CTLFLAG_RW, 0, 0, sysctl_def_policy, "I",
251 	"IPsec default policy.");
252 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
253 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_esp_trans_deflev), 0,
254 	"Default ESP transport mode level.");
255 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
256 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_esp_net_deflev), 0,
257 	"Default ESP tunnel mode level.");
258 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
259 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ah_trans_deflev), 0,
260 	"AH transfer mode default level.");
261 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
262 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ah_net_deflev), 0,
263 	"AH tunnel mode default level.");
264 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_ECN, ecn,
265 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ipsec_ecn), 0,
266 	"Explicit Congestion Notification handling.");
267 SYSCTL_INT(_net_inet6_ipsec6, OID_AUTO, filtertunnel,
268 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_filtertunnel),  0,
269 	"If set, filter packets from an IPsec tunnel.");
270 SYSCTL_VNET_PCPUSTAT(_net_inet6_ipsec6, IPSECCTL_STATS, ipsecstats,
271     struct ipsecstat, ipsec6stat, "IPsec IPv6 statistics.");
272 #endif /* INET6 */
273 
274 static int ipsec_in_reject(struct secpolicy *, struct inpcb *,
275     const struct mbuf *);
276 
277 #ifdef INET
278 static void ipsec4_get_ulp(const struct mbuf *, struct secpolicyindex *, int);
279 static void ipsec4_setspidx_ipaddr(const struct mbuf *,
280     struct secpolicyindex *);
281 #endif
282 #ifdef INET6
283 static void ipsec6_get_ulp(const struct mbuf *m, struct secpolicyindex *, int);
284 static void ipsec6_setspidx_ipaddr(const struct mbuf *,
285     struct secpolicyindex *);
286 #endif
287 
288 /*
289  * Return a held reference to the default SP.
290  */
291 static struct secpolicy *
292 key_allocsp_default(void)
293 {
294 
295 	key_addref(V_def_policy);
296 	return (V_def_policy);
297 }
298 
299 static void
300 ipsec_invalidate_cache(struct inpcb *inp, u_int dir)
301 {
302 	struct secpolicy *sp;
303 
304 	INP_WLOCK_ASSERT(inp);
305 	if (dir == IPSEC_DIR_OUTBOUND) {
306 		if (inp->inp_sp->flags & INP_INBOUND_POLICY)
307 			return;
308 		sp = inp->inp_sp->sp_in;
309 		inp->inp_sp->sp_in = NULL;
310 	} else {
311 		if (inp->inp_sp->flags & INP_OUTBOUND_POLICY)
312 			return;
313 		sp = inp->inp_sp->sp_out;
314 		inp->inp_sp->sp_out = NULL;
315 	}
316 	if (sp != NULL)
317 		key_freesp(&sp); /* release extra reference */
318 }
319 
320 static void
321 ipsec_cachepolicy(struct inpcb *inp, struct secpolicy *sp, u_int dir)
322 {
323 	uint32_t genid;
324 	int downgrade;
325 
326 	INP_LOCK_ASSERT(inp);
327 
328 	if (dir == IPSEC_DIR_OUTBOUND) {
329 		/* Do we have configured PCB policy? */
330 		if (inp->inp_sp->flags & INP_OUTBOUND_POLICY)
331 			return;
332 		/* Another thread has already set cached policy */
333 		if (inp->inp_sp->sp_out != NULL)
334 			return;
335 		/*
336 		 * Do not cache OUTBOUND policy if PCB isn't connected,
337 		 * i.e. foreign address is INADDR_ANY/UNSPECIFIED.
338 		 */
339 #ifdef INET
340 		if ((inp->inp_vflag & INP_IPV4) != 0 &&
341 		    inp->inp_faddr.s_addr == INADDR_ANY)
342 			return;
343 #endif
344 #ifdef INET6
345 		if ((inp->inp_vflag & INP_IPV6) != 0 &&
346 		    IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
347 			return;
348 #endif
349 	} else {
350 		/* Do we have configured PCB policy? */
351 		if (inp->inp_sp->flags & INP_INBOUND_POLICY)
352 			return;
353 		/* Another thread has already set cached policy */
354 		if (inp->inp_sp->sp_in != NULL)
355 			return;
356 		/*
357 		 * Do not cache INBOUND policy for listen socket,
358 		 * that is bound to INADDR_ANY/UNSPECIFIED address.
359 		 */
360 #ifdef INET
361 		if ((inp->inp_vflag & INP_IPV4) != 0 &&
362 		    inp->inp_faddr.s_addr == INADDR_ANY)
363 			return;
364 #endif
365 #ifdef INET6
366 		if ((inp->inp_vflag & INP_IPV6) != 0 &&
367 		    IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
368 			return;
369 #endif
370 	}
371 	downgrade = 0;
372 	if (!INP_WLOCKED(inp)) {
373 		if ((downgrade = INP_TRY_UPGRADE(inp)) == 0)
374 			return;
375 	}
376 	if (dir == IPSEC_DIR_OUTBOUND)
377 		inp->inp_sp->sp_out = sp;
378 	else
379 		inp->inp_sp->sp_in = sp;
380 	/*
381 	 * SP is already referenced by the lookup code.
382 	 * We take extra reference here to avoid race in the
383 	 * ipsec_getpcbpolicy() function - SP will not be freed in the
384 	 * time between we take SP pointer from the cache and key_addref()
385 	 * call.
386 	 */
387 	key_addref(sp);
388 	genid = key_getspgen();
389 	if (genid != inp->inp_sp->genid) {
390 		ipsec_invalidate_cache(inp, dir);
391 		inp->inp_sp->genid = genid;
392 	}
393 	KEYDBG(IPSEC_STAMP,
394 	    printf("%s: PCB(%p): cached %s SP(%p)\n",
395 	    __func__, inp, dir == IPSEC_DIR_OUTBOUND ? "OUTBOUND":
396 	    "INBOUND", sp));
397 	if (downgrade != 0)
398 		INP_DOWNGRADE(inp);
399 }
400 
401 static struct secpolicy *
402 ipsec_checkpolicy(struct secpolicy *sp, struct inpcb *inp, int *error)
403 {
404 
405 	/* Save found OUTBOUND policy into PCB SP cache. */
406 	if (inp != NULL && inp->inp_sp != NULL && inp->inp_sp->sp_out == NULL)
407 		ipsec_cachepolicy(inp, sp, IPSEC_DIR_OUTBOUND);
408 
409 	switch (sp->policy) {
410 	default:
411 		printf("%s: invalid policy %u\n", __func__, sp->policy);
412 		/* FALLTHROUGH */
413 	case IPSEC_POLICY_DISCARD:
414 		*error = -EINVAL;	/* Packet is discarded by caller. */
415 		/* FALLTHROUGH */
416 	case IPSEC_POLICY_BYPASS:
417 	case IPSEC_POLICY_NONE:
418 		key_freesp(&sp);
419 		sp = NULL;		/* NB: force NULL result. */
420 		break;
421 	case IPSEC_POLICY_IPSEC:
422 		/* XXXAE: handle LARVAL SP */
423 		break;
424 	}
425 	KEYDBG(IPSEC_DUMP,
426 	    printf("%s: get SP(%p), error %d\n", __func__, sp, *error));
427 	return (sp);
428 }
429 
430 static struct secpolicy *
431 ipsec_getpcbpolicy(struct inpcb *inp, u_int dir)
432 {
433 	struct secpolicy *sp;
434 	int flags, downgrade;
435 
436 	if (inp == NULL || inp->inp_sp == NULL)
437 		return (NULL);
438 
439 	INP_LOCK_ASSERT(inp);
440 
441 	flags = inp->inp_sp->flags;
442 	if (dir == IPSEC_DIR_OUTBOUND) {
443 		sp = inp->inp_sp->sp_out;
444 		flags &= INP_OUTBOUND_POLICY;
445 	} else {
446 		sp = inp->inp_sp->sp_in;
447 		flags &= INP_INBOUND_POLICY;
448 	}
449 	/*
450 	 * Check flags. If we have PCB SP, just return it.
451 	 * Otherwise we need to check that cached SP entry isn't stale.
452 	 */
453 	if (flags == 0) {
454 		if (sp == NULL)
455 			return (NULL);
456 		if (inp->inp_sp->genid != key_getspgen()) {
457 			/* Invalidate the cache. */
458 			downgrade = 0;
459 			if (!INP_WLOCKED(inp)) {
460 				if ((downgrade = INP_TRY_UPGRADE(inp)) == 0)
461 					return (NULL);
462 			}
463 			ipsec_invalidate_cache(inp, IPSEC_DIR_OUTBOUND);
464 			ipsec_invalidate_cache(inp, IPSEC_DIR_INBOUND);
465 			if (downgrade != 0)
466 				INP_DOWNGRADE(inp);
467 			return (NULL);
468 		}
469 		KEYDBG(IPSEC_STAMP,
470 		    printf("%s: PCB(%p): cache hit SP(%p)\n",
471 		    __func__, inp, sp));
472 		/* Return referenced cached policy */
473 	}
474 	key_addref(sp);
475 	return (sp);
476 }
477 
478 #ifdef INET
479 static void
480 ipsec4_get_ulp(const struct mbuf *m, struct secpolicyindex *spidx,
481     int needport)
482 {
483 	uint8_t nxt;
484 	int off;
485 
486 	/* Sanity check. */
487 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),
488 	    ("packet too short"));
489 
490 	if (m->m_len >= sizeof (struct ip)) {
491 		const struct ip *ip = mtod(m, const struct ip *);
492 		if (ip->ip_off & htons(IP_MF | IP_OFFMASK))
493 			goto done;
494 		off = ip->ip_hl << 2;
495 		nxt = ip->ip_p;
496 	} else {
497 		struct ip ih;
498 
499 		m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih);
500 		if (ih.ip_off & htons(IP_MF | IP_OFFMASK))
501 			goto done;
502 		off = ih.ip_hl << 2;
503 		nxt = ih.ip_p;
504 	}
505 
506 	while (off < m->m_pkthdr.len) {
507 		struct ip6_ext ip6e;
508 		struct tcphdr th;
509 		struct udphdr uh;
510 
511 		switch (nxt) {
512 		case IPPROTO_TCP:
513 			spidx->ul_proto = nxt;
514 			if (!needport)
515 				goto done_proto;
516 			if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
517 				goto done;
518 			m_copydata(m, off, sizeof (th), (caddr_t) &th);
519 			spidx->src.sin.sin_port = th.th_sport;
520 			spidx->dst.sin.sin_port = th.th_dport;
521 			return;
522 		case IPPROTO_UDP:
523 			spidx->ul_proto = nxt;
524 			if (!needport)
525 				goto done_proto;
526 			if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
527 				goto done;
528 			m_copydata(m, off, sizeof (uh), (caddr_t) &uh);
529 			spidx->src.sin.sin_port = uh.uh_sport;
530 			spidx->dst.sin.sin_port = uh.uh_dport;
531 			return;
532 		case IPPROTO_AH:
533 			if (off + sizeof(ip6e) > m->m_pkthdr.len)
534 				goto done;
535 			/* XXX Sigh, this works but is totally bogus. */
536 			m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e);
537 			off += (ip6e.ip6e_len + 2) << 2;
538 			nxt = ip6e.ip6e_nxt;
539 			break;
540 		case IPPROTO_ICMP:
541 		default:
542 			/* XXX Intermediate headers??? */
543 			spidx->ul_proto = nxt;
544 			goto done_proto;
545 		}
546 	}
547 done:
548 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
549 done_proto:
550 	spidx->src.sin.sin_port = IPSEC_PORT_ANY;
551 	spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
552 	KEYDBG(IPSEC_DUMP,
553 	    printf("%s: ", __func__); kdebug_secpolicyindex(spidx, NULL));
554 }
555 
556 static void
557 ipsec4_setspidx_ipaddr(const struct mbuf *m, struct secpolicyindex *spidx)
558 {
559 
560 	ipsec4_setsockaddrs(m, &spidx->src, &spidx->dst);
561 	spidx->prefs = sizeof(struct in_addr) << 3;
562 	spidx->prefd = sizeof(struct in_addr) << 3;
563 }
564 
565 static struct secpolicy *
566 ipsec4_getpolicy(const struct mbuf *m, struct inpcb *inp, u_int dir,
567     int needport)
568 {
569 	struct secpolicyindex spidx;
570 	struct secpolicy *sp;
571 
572 	sp = ipsec_getpcbpolicy(inp, dir);
573 	if (sp == NULL && key_havesp(dir)) {
574 		/* Make an index to look for a policy. */
575 		ipsec4_setspidx_ipaddr(m, &spidx);
576 		ipsec4_get_ulp(m, &spidx, needport);
577 		spidx.dir = dir;
578 		sp = key_allocsp(&spidx, dir);
579 	}
580 	if (sp == NULL)		/* No SP found, use system default. */
581 		sp = key_allocsp_default();
582 	return (sp);
583 }
584 
585 /*
586  * Check security policy for *OUTBOUND* IPv4 packet.
587  */
588 struct secpolicy *
589 ipsec4_checkpolicy(const struct mbuf *m, struct inpcb *inp, int *error,
590     int needport)
591 {
592 	struct secpolicy *sp;
593 
594 	*error = 0;
595 	sp = ipsec4_getpolicy(m, inp, IPSEC_DIR_OUTBOUND, needport);
596 	if (sp != NULL)
597 		sp = ipsec_checkpolicy(sp, inp, error);
598 	if (sp == NULL) {
599 		switch (*error) {
600 		case 0: /* No IPsec required: BYPASS or NONE */
601 			break;
602 		case -EINVAL:
603 			IPSECSTAT_INC(ips_out_polvio);
604 			break;
605 		default:
606 			IPSECSTAT_INC(ips_out_inval);
607 		}
608 	}
609 	KEYDBG(IPSEC_STAMP,
610 	    printf("%s: using SP(%p), error %d\n", __func__, sp, *error));
611 	if (sp != NULL)
612 		KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
613 	return (sp);
614 }
615 
616 /*
617  * Check IPv4 packet against *INBOUND* security policy.
618  * This function is called from tcp_input(), udp_input(),
619  * rip_input() and sctp_input().
620  */
621 int
622 ipsec4_in_reject(const struct mbuf *m, struct inpcb *inp)
623 {
624 	struct secpolicy *sp;
625 	int result;
626 
627 	sp = ipsec4_getpolicy(m, inp, IPSEC_DIR_INBOUND, 0);
628 	result = ipsec_in_reject(sp, inp, m);
629 	key_freesp(&sp);
630 	if (result != 0)
631 		IPSECSTAT_INC(ips_in_polvio);
632 	return (result);
633 }
634 
635 /*
636  * IPSEC_CAP() method implementation for IPv4.
637  */
638 int
639 ipsec4_capability(struct mbuf *m, u_int cap)
640 {
641 
642 	switch (cap) {
643 	case IPSEC_CAP_BYPASS_FILTER:
644 		/*
645 		 * Bypass packet filtering for packets previously handled
646 		 * by IPsec.
647 		 */
648 		if (!V_ip4_filtertunnel &&
649 		    m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL)
650 			return (1);
651 		return (0);
652 	case IPSEC_CAP_OPERABLE:
653 		/* Do we have active security policies? */
654 		if (key_havesp(IPSEC_DIR_INBOUND) != 0 ||
655 		    key_havesp(IPSEC_DIR_OUTBOUND) != 0)
656 			return (1);
657 		return (0);
658 	};
659 	return (EOPNOTSUPP);
660 }
661 
662 #endif /* INET */
663 
664 #ifdef INET6
665 static void
666 ipsec6_get_ulp(const struct mbuf *m, struct secpolicyindex *spidx,
667     int needport)
668 {
669 	struct tcphdr th;
670 	struct udphdr uh;
671 	struct icmp6_hdr ih;
672 	int off, nxt;
673 
674 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip6_hdr),
675 	    ("packet too short"));
676 
677 	/* Set default. */
678 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
679 	spidx->src.sin6.sin6_port = IPSEC_PORT_ANY;
680 	spidx->dst.sin6.sin6_port = IPSEC_PORT_ANY;
681 
682 	nxt = -1;
683 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
684 	if (off < 0 || m->m_pkthdr.len < off)
685 		return;
686 
687 	switch (nxt) {
688 	case IPPROTO_TCP:
689 		spidx->ul_proto = nxt;
690 		if (!needport)
691 			break;
692 		if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
693 			break;
694 		m_copydata(m, off, sizeof(th), (caddr_t)&th);
695 		spidx->src.sin6.sin6_port = th.th_sport;
696 		spidx->dst.sin6.sin6_port = th.th_dport;
697 		break;
698 	case IPPROTO_UDP:
699 		spidx->ul_proto = nxt;
700 		if (!needport)
701 			break;
702 		if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
703 			break;
704 		m_copydata(m, off, sizeof(uh), (caddr_t)&uh);
705 		spidx->src.sin6.sin6_port = uh.uh_sport;
706 		spidx->dst.sin6.sin6_port = uh.uh_dport;
707 		break;
708 	case IPPROTO_ICMPV6:
709 		spidx->ul_proto = nxt;
710 		if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len)
711 			break;
712 		m_copydata(m, off, sizeof(ih), (caddr_t)&ih);
713 		spidx->src.sin6.sin6_port = htons((uint16_t)ih.icmp6_type);
714 		spidx->dst.sin6.sin6_port = htons((uint16_t)ih.icmp6_code);
715 		break;
716 	default:
717 		/* XXX Intermediate headers??? */
718 		spidx->ul_proto = nxt;
719 		break;
720 	}
721 	KEYDBG(IPSEC_DUMP,
722 	    printf("%s: ", __func__); kdebug_secpolicyindex(spidx, NULL));
723 }
724 
725 static void
726 ipsec6_setspidx_ipaddr(const struct mbuf *m, struct secpolicyindex *spidx)
727 {
728 
729 	ipsec6_setsockaddrs(m, &spidx->src, &spidx->dst);
730 	spidx->prefs = sizeof(struct in6_addr) << 3;
731 	spidx->prefd = sizeof(struct in6_addr) << 3;
732 }
733 
734 static struct secpolicy *
735 ipsec6_getpolicy(const struct mbuf *m, struct inpcb *inp, u_int dir,
736     int needport)
737 {
738 	struct secpolicyindex spidx;
739 	struct secpolicy *sp;
740 
741 	sp = ipsec_getpcbpolicy(inp, dir);
742 	if (sp == NULL && key_havesp(dir)) {
743 		/* Make an index to look for a policy. */
744 		ipsec6_setspidx_ipaddr(m, &spidx);
745 		ipsec6_get_ulp(m, &spidx, needport);
746 		spidx.dir = dir;
747 		sp = key_allocsp(&spidx, dir);
748 	}
749 	if (sp == NULL)		/* No SP found, use system default. */
750 		sp = key_allocsp_default();
751 	return (sp);
752 }
753 
754 /*
755  * Check security policy for *OUTBOUND* IPv6 packet.
756  */
757 struct secpolicy *
758 ipsec6_checkpolicy(const struct mbuf *m, struct inpcb *inp, int *error,
759     int needport)
760 {
761 	struct secpolicy *sp;
762 
763 	*error = 0;
764 	sp = ipsec6_getpolicy(m, inp, IPSEC_DIR_OUTBOUND, needport);
765 	if (sp != NULL)
766 		sp = ipsec_checkpolicy(sp, inp, error);
767 	if (sp == NULL) {
768 		switch (*error) {
769 		case 0: /* No IPsec required: BYPASS or NONE */
770 			break;
771 		case -EINVAL:
772 			IPSEC6STAT_INC(ips_out_polvio);
773 			break;
774 		default:
775 			IPSEC6STAT_INC(ips_out_inval);
776 		}
777 	}
778 	KEYDBG(IPSEC_STAMP,
779 	    printf("%s: using SP(%p), error %d\n", __func__, sp, *error));
780 	if (sp != NULL)
781 		KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
782 	return (sp);
783 }
784 
785 /*
786  * Check IPv6 packet against inbound security policy.
787  * This function is called from tcp6_input(), udp6_input(),
788  * rip6_input() and sctp_input().
789  */
790 int
791 ipsec6_in_reject(const struct mbuf *m, struct inpcb *inp)
792 {
793 	struct secpolicy *sp;
794 	int result;
795 
796 	sp = ipsec6_getpolicy(m, inp, IPSEC_DIR_INBOUND, 0);
797 	result = ipsec_in_reject(sp, inp, m);
798 	key_freesp(&sp);
799 	if (result)
800 		IPSEC6STAT_INC(ips_in_polvio);
801 	return (result);
802 }
803 
804 /*
805  * IPSEC_CAP() method implementation for IPv6.
806  */
807 int
808 ipsec6_capability(struct mbuf *m, u_int cap)
809 {
810 
811 	switch (cap) {
812 	case IPSEC_CAP_BYPASS_FILTER:
813 		/*
814 		 * Bypass packet filtering for packets previously handled
815 		 * by IPsec.
816 		 */
817 		if (!V_ip6_filtertunnel &&
818 		    m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL) != NULL)
819 			return (1);
820 		return (0);
821 	case IPSEC_CAP_OPERABLE:
822 		/* Do we have active security policies? */
823 		if (key_havesp(IPSEC_DIR_INBOUND) != 0 ||
824 		    key_havesp(IPSEC_DIR_OUTBOUND) != 0)
825 			return (1);
826 		return (0);
827 	};
828 	return (EOPNOTSUPP);
829 }
830 #endif /* INET6 */
831 
832 int
833 ipsec_run_hhooks(struct ipsec_ctx_data *ctx, int type)
834 {
835 	int idx;
836 
837 	switch (ctx->af) {
838 #ifdef INET
839 	case AF_INET:
840 		idx = HHOOK_IPSEC_INET;
841 		break;
842 #endif
843 #ifdef INET6
844 	case AF_INET6:
845 		idx = HHOOK_IPSEC_INET6;
846 		break;
847 #endif
848 	default:
849 		return (EPFNOSUPPORT);
850 	}
851 	if (type == HHOOK_TYPE_IPSEC_IN)
852 		HHOOKS_RUN_IF(V_ipsec_hhh_in[idx], ctx, NULL);
853 	else
854 		HHOOKS_RUN_IF(V_ipsec_hhh_out[idx], ctx, NULL);
855 	if (*ctx->mp == NULL)
856 		return (EACCES);
857 	return (0);
858 }
859 
860 /*
861  * Return current level.
862  * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
863  */
864 u_int
865 ipsec_get_reqlevel(struct secpolicy *sp, u_int idx)
866 {
867 	struct ipsecrequest *isr;
868 	u_int esp_trans_deflev, esp_net_deflev;
869 	u_int ah_trans_deflev, ah_net_deflev;
870 	u_int level = 0;
871 
872 	IPSEC_ASSERT(idx < sp->tcount, ("Wrong IPsec request index %d", idx));
873 /* XXX Note that we have ipseclog() expanded here - code sync issue. */
874 #define IPSEC_CHECK_DEFAULT(lev) \
875 	(((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE &&	\
876 	  (lev) != IPSEC_LEVEL_UNIQUE)					\
877 		? (V_ipsec_debug  ?					\
878 		log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
879 		(lev), IPSEC_LEVEL_REQUIRE) : 0),			\
880 		(lev) = IPSEC_LEVEL_REQUIRE, (lev) : (lev))
881 
882 	/*
883 	 * IPsec VTI uses unique security policy with fake spidx filled
884 	 * with zeroes. Just return IPSEC_LEVEL_REQUIRE instead of doing
885 	 * full level lookup for such policies.
886 	 */
887 	if (sp->state == IPSEC_SPSTATE_IFNET) {
888 		IPSEC_ASSERT(sp->req[idx]->level == IPSEC_LEVEL_UNIQUE,
889 		    ("Wrong IPsec request level %d", sp->req[idx]->level));
890 		return (IPSEC_LEVEL_REQUIRE);
891 	}
892 
893 	/* Set default level. */
894 	switch (sp->spidx.src.sa.sa_family) {
895 #ifdef INET
896 	case AF_INET:
897 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev);
898 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev);
899 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev);
900 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev);
901 		break;
902 #endif
903 #ifdef INET6
904 	case AF_INET6:
905 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev);
906 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev);
907 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev);
908 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev);
909 		break;
910 #endif /* INET6 */
911 	default:
912 		panic("%s: unknown af %u",
913 			__func__, sp->spidx.src.sa.sa_family);
914 	}
915 
916 #undef IPSEC_CHECK_DEFAULT
917 
918 	isr = sp->req[idx];
919 	/* Set level. */
920 	switch (isr->level) {
921 	case IPSEC_LEVEL_DEFAULT:
922 		switch (isr->saidx.proto) {
923 		case IPPROTO_ESP:
924 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
925 				level = esp_net_deflev;
926 			else
927 				level = esp_trans_deflev;
928 			break;
929 		case IPPROTO_AH:
930 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
931 				level = ah_net_deflev;
932 			else
933 				level = ah_trans_deflev;
934 			break;
935 		case IPPROTO_IPCOMP:
936 			/*
937 			 * We don't really care, as IPcomp document says that
938 			 * we shouldn't compress small packets.
939 			 */
940 			level = IPSEC_LEVEL_USE;
941 			break;
942 		default:
943 			panic("%s: Illegal protocol defined %u\n", __func__,
944 				isr->saidx.proto);
945 		}
946 		break;
947 
948 	case IPSEC_LEVEL_USE:
949 	case IPSEC_LEVEL_REQUIRE:
950 		level = isr->level;
951 		break;
952 	case IPSEC_LEVEL_UNIQUE:
953 		level = IPSEC_LEVEL_REQUIRE;
954 		break;
955 
956 	default:
957 		panic("%s: Illegal IPsec level %u\n", __func__, isr->level);
958 	}
959 
960 	return (level);
961 }
962 
963 static int
964 ipsec_check_history(const struct mbuf *m, struct secpolicy *sp, u_int idx)
965 {
966 	struct xform_history *xh;
967 	struct m_tag *mtag;
968 
969 	mtag = NULL;
970 	while ((mtag = m_tag_find(__DECONST(struct mbuf *, m),
971 	    PACKET_TAG_IPSEC_IN_DONE, mtag)) != NULL) {
972 		xh = (struct xform_history *)(mtag + 1);
973 		KEYDBG(IPSEC_DATA,
974 		    char buf[IPSEC_ADDRSTRLEN];
975 		    printf("%s: mode %s proto %u dst %s\n", __func__,
976 			kdebug_secasindex_mode(xh->mode), xh->proto,
977 			ipsec_address(&xh->dst, buf, sizeof(buf))));
978 		if (xh->proto != sp->req[idx]->saidx.proto)
979 			continue;
980 		/* If SA had IPSEC_MODE_ANY, consider this as match. */
981 		if (xh->mode != sp->req[idx]->saidx.mode &&
982 		    xh->mode != IPSEC_MODE_ANY)
983 			continue;
984 		/*
985 		 * For transport mode IPsec request doesn't contain
986 		 * addresses. We need to use address from spidx.
987 		 */
988 		if (sp->req[idx]->saidx.mode == IPSEC_MODE_TRANSPORT) {
989 			if (key_sockaddrcmp_withmask(&xh->dst.sa,
990 			    &sp->spidx.dst.sa, sp->spidx.prefd) != 0)
991 				continue;
992 		} else {
993 			if (key_sockaddrcmp(&xh->dst.sa,
994 			    &sp->req[idx]->saidx.dst.sa, 0) != 0)
995 				continue;
996 		}
997 		return (0); /* matched */
998 	}
999 	return (1);
1000 }
1001 
1002 /*
1003  * Check security policy requirements against the actual
1004  * packet contents.  Return one if the packet should be
1005  * reject as "invalid"; otherwiser return zero to have the
1006  * packet treated as "valid".
1007  *
1008  * OUT:
1009  *	0: valid
1010  *	1: invalid
1011  */
1012 static int
1013 ipsec_in_reject(struct secpolicy *sp, struct inpcb *inp, const struct mbuf *m)
1014 {
1015 	int i;
1016 
1017 	KEYDBG(IPSEC_STAMP,
1018 	    printf("%s: PCB(%p): using SP(%p)\n", __func__, inp, sp));
1019 	KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1020 
1021 	if (inp != NULL && inp->inp_sp != NULL && inp->inp_sp->sp_in == NULL)
1022 		ipsec_cachepolicy(inp, sp, IPSEC_DIR_INBOUND);
1023 
1024 	/* Check policy. */
1025 	switch (sp->policy) {
1026 	case IPSEC_POLICY_DISCARD:
1027 		return (1);
1028 	case IPSEC_POLICY_BYPASS:
1029 	case IPSEC_POLICY_NONE:
1030 		return (0);
1031 	}
1032 
1033 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1034 		("invalid policy %u", sp->policy));
1035 
1036 	/*
1037 	 * ipsec[46]_common_input_cb after each transform adds
1038 	 * PACKET_TAG_IPSEC_IN_DONE mbuf tag. It contains SPI, proto, mode
1039 	 * and destination address from saidx. We can compare info from
1040 	 * these tags with requirements in SP.
1041 	 */
1042 	for (i = 0; i < sp->tcount; i++) {
1043 		/*
1044 		 * Do not check IPcomp, since IPcomp document
1045 		 * says that we shouldn't compress small packets.
1046 		 * IPComp policy should always be treated as being
1047 		 * in "use" level.
1048 		 */
1049 		if (sp->req[i]->saidx.proto == IPPROTO_IPCOMP ||
1050 		    ipsec_get_reqlevel(sp, i) != IPSEC_LEVEL_REQUIRE)
1051 			continue;
1052 		if (V_check_policy_history != 0 &&
1053 		    ipsec_check_history(m, sp, i) != 0)
1054 			return (1);
1055 		else switch (sp->req[i]->saidx.proto) {
1056 		case IPPROTO_ESP:
1057 			if ((m->m_flags & M_DECRYPTED) == 0) {
1058 				KEYDBG(IPSEC_DUMP,
1059 				    printf("%s: ESP m_flags:%x\n", __func__,
1060 					    m->m_flags));
1061 				return (1);
1062 			}
1063 			break;
1064 		case IPPROTO_AH:
1065 			if ((m->m_flags & M_AUTHIPHDR) == 0) {
1066 				KEYDBG(IPSEC_DUMP,
1067 				    printf("%s: AH m_flags:%x\n", __func__,
1068 					    m->m_flags));
1069 				return (1);
1070 			}
1071 			break;
1072 		}
1073 	}
1074 	return (0);		/* Valid. */
1075 }
1076 
1077 /*
1078  * Compute the byte size to be occupied by IPsec header.
1079  * In case it is tunnelled, it includes the size of outer IP header.
1080  */
1081 static size_t
1082 ipsec_hdrsiz_internal(struct secpolicy *sp)
1083 {
1084 	size_t size;
1085 	int i;
1086 
1087 	KEYDBG(IPSEC_STAMP, printf("%s: using SP(%p)\n", __func__, sp));
1088 	KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1089 
1090 	switch (sp->policy) {
1091 	case IPSEC_POLICY_DISCARD:
1092 	case IPSEC_POLICY_BYPASS:
1093 	case IPSEC_POLICY_NONE:
1094 		return (0);
1095 	}
1096 
1097 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1098 		("invalid policy %u", sp->policy));
1099 
1100 	/*
1101 	 * XXX: for each transform we need to lookup suitable SA
1102 	 * and use info from SA to calculate headers size.
1103 	 * XXX: for NAT-T we need to cosider UDP header size.
1104 	 */
1105 	size = 0;
1106 	for (i = 0; i < sp->tcount; i++) {
1107 		switch (sp->req[i]->saidx.proto) {
1108 		case IPPROTO_ESP:
1109 			size += esp_hdrsiz(NULL);
1110 			break;
1111 		case IPPROTO_AH:
1112 			size += ah_hdrsiz(NULL);
1113 			break;
1114 		case IPPROTO_IPCOMP:
1115 			size += sizeof(struct ipcomp);
1116 			break;
1117 		}
1118 
1119 		if (sp->req[i]->saidx.mode == IPSEC_MODE_TUNNEL) {
1120 			switch (sp->req[i]->saidx.dst.sa.sa_family) {
1121 #ifdef INET
1122 			case AF_INET:
1123 				size += sizeof(struct ip);
1124 				break;
1125 #endif
1126 #ifdef INET6
1127 			case AF_INET6:
1128 				size += sizeof(struct ip6_hdr);
1129 				break;
1130 #endif
1131 			default:
1132 				ipseclog((LOG_ERR, "%s: unknown AF %d in "
1133 				    "IPsec tunnel SA\n", __func__,
1134 				    sp->req[i]->saidx.dst.sa.sa_family));
1135 				break;
1136 			}
1137 		}
1138 	}
1139 	return (size);
1140 }
1141 
1142 /*
1143  * Compute ESP/AH header size for protocols with PCB, including
1144  * outer IP header. Currently only tcp_output() uses it.
1145  */
1146 size_t
1147 ipsec_hdrsiz_inpcb(struct inpcb *inp)
1148 {
1149 	struct secpolicyindex spidx;
1150 	struct secpolicy *sp;
1151 	size_t sz;
1152 
1153 	sp = ipsec_getpcbpolicy(inp, IPSEC_DIR_OUTBOUND);
1154 	if (sp == NULL && key_havesp(IPSEC_DIR_OUTBOUND)) {
1155 		ipsec_setspidx_inpcb(inp, &spidx, IPSEC_DIR_OUTBOUND);
1156 		sp = key_allocsp(&spidx, IPSEC_DIR_OUTBOUND);
1157 	}
1158 	if (sp == NULL)
1159 		sp = key_allocsp_default();
1160 	sz = ipsec_hdrsiz_internal(sp);
1161 	key_freesp(&sp);
1162 	return (sz);
1163 }
1164 
1165 /*
1166  * Check the variable replay window.
1167  * ipsec_chkreplay() performs replay check before ICV verification.
1168  * ipsec_updatereplay() updates replay bitmap.  This must be called after
1169  * ICV verification (it also performs replay check, which is usually done
1170  * beforehand).
1171  * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1172  *
1173  * Based on RFC 6479. Blocks are 32 bits unsigned integers
1174  */
1175 
1176 #define IPSEC_BITMAP_INDEX_MASK(w)	(w - 1)
1177 #define IPSEC_REDUNDANT_BIT_SHIFTS	5
1178 #define IPSEC_REDUNDANT_BITS		(1 << IPSEC_REDUNDANT_BIT_SHIFTS)
1179 #define IPSEC_BITMAP_LOC_MASK		(IPSEC_REDUNDANT_BITS - 1)
1180 
1181 int
1182 ipsec_chkreplay(uint32_t seq, struct secasvar *sav)
1183 {
1184 	const struct secreplay *replay;
1185 	uint32_t wsizeb;		/* Constant: window size. */
1186 	int index, bit_location;
1187 
1188 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1189 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1190 
1191 	replay = sav->replay;
1192 
1193 	/* No need to check replay if disabled. */
1194 	if (replay->wsize == 0)
1195 		return (1);
1196 
1197 	/* Constant. */
1198 	wsizeb = replay->wsize << 3;
1199 
1200 	/* Sequence number of 0 is invalid. */
1201 	if (seq == 0)
1202 		return (0);
1203 
1204 	/* First time is always okay. */
1205 	if (replay->count == 0)
1206 		return (1);
1207 
1208 	/* Larger sequences are okay. */
1209 	if (seq > replay->lastseq)
1210 		return (1);
1211 
1212 	/* Over range to check, i.e. too old or wrapped. */
1213 	if (replay->lastseq - seq >= wsizeb)
1214 		return (0);
1215 
1216 	/* The sequence is inside the sliding window
1217 	 * now check the bit in the bitmap
1218 	 * bit location only depends on the sequence number
1219 	 */
1220 	bit_location = seq & IPSEC_BITMAP_LOC_MASK;
1221 	index = (seq >> IPSEC_REDUNDANT_BIT_SHIFTS)
1222 		& IPSEC_BITMAP_INDEX_MASK(replay->bitmap_size);
1223 
1224 	/* This packet already seen? */
1225 	if ((replay->bitmap)[index] & (1 << bit_location))
1226 		return (0);
1227 	return (1);
1228 }
1229 
1230 /*
1231  * Check replay counter whether to update or not.
1232  * OUT:	0:	OK
1233  *	1:	NG
1234  */
1235 int
1236 ipsec_updatereplay(uint32_t seq, struct secasvar *sav)
1237 {
1238 	char buf[128];
1239 	struct secreplay *replay;
1240 	uint32_t wsizeb;		/* Constant: window size. */
1241 	int diff, index, bit_location;
1242 
1243 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1244 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1245 
1246 	replay = sav->replay;
1247 
1248 	if (replay->wsize == 0)
1249 		goto ok;	/* No need to check replay. */
1250 
1251 	/* Constant. */
1252 	wsizeb = replay->wsize << 3;
1253 
1254 	/* Sequence number of 0 is invalid. */
1255 	if (seq == 0)
1256 		return (1);
1257 
1258 	/* The packet is too old, no need to update */
1259 	if (wsizeb + seq < replay->lastseq)
1260 		goto ok;
1261 
1262 	/* Now update the bit */
1263 	index = (seq >> IPSEC_REDUNDANT_BIT_SHIFTS);
1264 
1265 	/* First check if the sequence number is in the range */
1266 	if (seq > replay->lastseq) {
1267 		int id;
1268 		int index_cur = replay->lastseq >> IPSEC_REDUNDANT_BIT_SHIFTS;
1269 
1270 		diff = index - index_cur;
1271 		if (diff > replay->bitmap_size) {
1272 			/* something unusual in this case */
1273 			diff = replay->bitmap_size;
1274 		}
1275 
1276 		for (id = 0; id < diff; ++id) {
1277 			replay->bitmap[(id + index_cur + 1)
1278 			& IPSEC_BITMAP_INDEX_MASK(replay->bitmap_size)] = 0;
1279 		}
1280 
1281 		replay->lastseq = seq;
1282 	}
1283 
1284 	index &= IPSEC_BITMAP_INDEX_MASK(replay->bitmap_size);
1285 	bit_location = seq & IPSEC_BITMAP_LOC_MASK;
1286 
1287 	/* this packet has already been received */
1288 	if (replay->bitmap[index] & (1 << bit_location))
1289 		return (1);
1290 
1291 	replay->bitmap[index] |= (1 << bit_location);
1292 
1293 ok:
1294 	if (replay->count == ~0) {
1295 
1296 		/* Set overflow flag. */
1297 		replay->overflow++;
1298 
1299 		/* Don't increment, no more packets accepted. */
1300 		if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0) {
1301 			if (sav->sah->saidx.proto == IPPROTO_AH)
1302 				AHSTAT_INC(ahs_wrap);
1303 			else if (sav->sah->saidx.proto == IPPROTO_ESP)
1304 				ESPSTAT_INC(esps_wrap);
1305 			return (1);
1306 		}
1307 
1308 		ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n",
1309 		    __func__, replay->overflow,
1310 		    ipsec_sa2str(sav, buf, sizeof(buf))));
1311 	}
1312 	return (0);
1313 }
1314 
1315 int
1316 ipsec_updateid(struct secasvar *sav, uint64_t *new, uint64_t *old)
1317 {
1318 	uint64_t tmp;
1319 
1320 	/*
1321 	 * tdb_cryptoid is initialized by xform_init().
1322 	 * Then it can be changed only when some crypto error occurred or
1323 	 * when SA is deleted. We stored used cryptoid in the xform_data
1324 	 * structure. In case when crypto error occurred and crypto
1325 	 * subsystem has reinited the session, it returns new cryptoid
1326 	 * and EAGAIN error code.
1327 	 *
1328 	 * This function will be called when we got EAGAIN from crypto
1329 	 * subsystem.
1330 	 * *new is cryptoid that was returned by crypto subsystem in
1331 	 * the crp_sid.
1332 	 * *old is the original cryptoid that we stored in xform_data.
1333 	 *
1334 	 * For first failed request *old == sav->tdb_cryptoid, then
1335 	 * we update sav->tdb_cryptoid and redo crypto_dispatch().
1336 	 * For next failed request *old != sav->tdb_cryptoid, then
1337 	 * we store cryptoid from first request into the *new variable
1338 	 * and crp_sid from this second session will be returned via
1339 	 * *old pointer, so caller can release second session.
1340 	 *
1341 	 * XXXAE: check this more carefully.
1342 	 */
1343 	KEYDBG(IPSEC_STAMP,
1344 	    printf("%s: SA(%p) moves cryptoid %jd -> %jd\n",
1345 		__func__, sav, (uintmax_t)(*old), (uintmax_t)(*new)));
1346 	KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1347 	SECASVAR_LOCK(sav);
1348 	if (sav->tdb_cryptoid != *old) {
1349 		/* cryptoid was already updated */
1350 		tmp = *new;
1351 		*new = sav->tdb_cryptoid;
1352 		*old = tmp;
1353 		SECASVAR_UNLOCK(sav);
1354 		return (1);
1355 	}
1356 	sav->tdb_cryptoid = *new;
1357 	SECASVAR_UNLOCK(sav);
1358 	return (0);
1359 }
1360 
1361 int
1362 ipsec_initialized(void)
1363 {
1364 
1365 	return (V_def_policy != NULL);
1366 }
1367 
1368 static void
1369 def_policy_init(const void *unused __unused)
1370 {
1371 
1372 	V_def_policy = key_newsp();
1373 	if (V_def_policy != NULL) {
1374 		V_def_policy->policy = IPSEC_POLICY_NONE;
1375 		/* Force INPCB SP cache invalidation */
1376 		key_bumpspgen();
1377 	} else
1378 		printf("%s: failed to initialize default policy\n", __func__);
1379 }
1380 
1381 
1382 static void
1383 def_policy_uninit(const void *unused __unused)
1384 {
1385 
1386 	if (V_def_policy != NULL) {
1387 		key_freesp(&V_def_policy);
1388 		key_bumpspgen();
1389 	}
1390 }
1391 
1392 VNET_SYSINIT(def_policy_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST,
1393     def_policy_init, NULL);
1394 VNET_SYSUNINIT(def_policy_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST,
1395     def_policy_uninit, NULL);
1396