xref: /freebsd/sys/netipsec/ipsec.c (revision aa64588d28258aef88cc33b8043112e8856948d0)
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/time.h>
52 #include <sys/kernel.h>
53 #include <sys/syslog.h>
54 #include <sys/sysctl.h>
55 #include <sys/proc.h>
56 
57 #include <net/if.h>
58 #include <net/route.h>
59 #include <net/vnet.h>
60 
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/ip.h>
64 #include <netinet/ip_var.h>
65 #include <netinet/in_var.h>
66 #include <netinet/udp.h>
67 #include <netinet/udp_var.h>
68 #include <netinet/tcp.h>
69 #include <netinet/udp.h>
70 
71 #include <netinet/ip6.h>
72 #ifdef INET6
73 #include <netinet6/ip6_var.h>
74 #endif
75 #include <netinet/in_pcb.h>
76 #ifdef INET6
77 #include <netinet/icmp6.h>
78 #endif
79 
80 #include <sys/types.h>
81 #include <netipsec/ipsec.h>
82 #ifdef INET6
83 #include <netipsec/ipsec6.h>
84 #endif
85 #include <netipsec/ah_var.h>
86 #include <netipsec/esp_var.h>
87 #include <netipsec/ipcomp.h>		/*XXX*/
88 #include <netipsec/ipcomp_var.h>
89 
90 #include <netipsec/key.h>
91 #include <netipsec/keydb.h>
92 #include <netipsec/key_debug.h>
93 
94 #include <netipsec/xform.h>
95 
96 #include <machine/in_cksum.h>
97 
98 #include <opencrypto/cryptodev.h>
99 
100 #ifdef IPSEC_DEBUG
101 VNET_DEFINE(int, ipsec_debug) = 1;
102 #else
103 VNET_DEFINE(int, ipsec_debug) = 0;
104 #endif
105 
106 /* NB: name changed so netstat doesn't use it. */
107 VNET_DEFINE(struct ipsecstat, ipsec4stat);
108 VNET_DEFINE(int, ip4_ah_offsetmask) = 0;	/* maybe IP_DF? */
109 /* DF bit on encap. 0: clear 1: set 2: copy */
110 VNET_DEFINE(int, ip4_ipsec_dfbit) = 0;
111 VNET_DEFINE(int, ip4_esp_trans_deflev) = IPSEC_LEVEL_USE;
112 VNET_DEFINE(int, ip4_esp_net_deflev) = IPSEC_LEVEL_USE;
113 VNET_DEFINE(int, ip4_ah_trans_deflev) = IPSEC_LEVEL_USE;
114 VNET_DEFINE(int, ip4_ah_net_deflev) = IPSEC_LEVEL_USE;
115 VNET_DEFINE(struct secpolicy, ip4_def_policy);
116 /* ECN ignore(-1)/forbidden(0)/allowed(1) */
117 VNET_DEFINE(int, ip4_ipsec_ecn) = 0;
118 VNET_DEFINE(int, ip4_esp_randpad) = -1;
119 
120 /*
121  * Crypto support requirements:
122  *
123  *  1	require hardware support
124  * -1	require software support
125  *  0	take anything
126  */
127 VNET_DEFINE(int, crypto_support) = CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE;
128 
129 SYSCTL_DECL(_net_inet_ipsec);
130 
131 /* net.inet.ipsec */
132 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_POLICY, def_policy,
133 	CTLFLAG_RW, &VNET_NAME(ip4_def_policy).policy, 0,
134 	"IPsec default policy.");
135 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
136 	CTLFLAG_RW, &VNET_NAME(ip4_esp_trans_deflev), 0,
137 	"Default ESP transport mode level");
138 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
139 	CTLFLAG_RW, &VNET_NAME(ip4_esp_net_deflev), 0,
140 	"Default ESP tunnel mode level.");
141 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
142 	CTLFLAG_RW, &VNET_NAME(ip4_ah_trans_deflev), 0,
143 	"AH transfer mode default level.");
144 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
145 	CTLFLAG_RW, &VNET_NAME(ip4_ah_net_deflev), 0,
146 	"AH tunnel mode default level.");
147 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_AH_CLEARTOS, ah_cleartos,
148 	CTLFLAG_RW, &VNET_NAME(ah_cleartos), 0,
149 	"If set clear type-of-service field when doing AH computation.");
150 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_AH_OFFSETMASK, ah_offsetmask,
151 	CTLFLAG_RW, &VNET_NAME(ip4_ah_offsetmask), 0,
152 	"If not set clear offset field mask when doing AH computation.");
153 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DFBIT, dfbit,
154 	CTLFLAG_RW, &VNET_NAME(ip4_ipsec_dfbit), 0,
155 	"Do not fragment bit on encap.");
156 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_ECN, ecn,
157 	CTLFLAG_RW, &VNET_NAME(ip4_ipsec_ecn), 0,
158 	"Explicit Congestion Notification handling.");
159 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
160 	CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
161 	"Enable IPsec debugging output when set.");
162 SYSCTL_VNET_INT(_net_inet_ipsec, OID_AUTO, crypto_support,
163 	CTLFLAG_RW, &VNET_NAME(crypto_support), 0,
164 	"Crypto driver selection.");
165 SYSCTL_VNET_STRUCT(_net_inet_ipsec, OID_AUTO, ipsecstats,
166 	CTLFLAG_RD, &VNET_NAME(ipsec4stat), ipsecstat,
167 	"IPsec IPv4 statistics.");
168 
169 #ifdef REGRESSION
170 /*
171  * When set to 1, IPsec will send packets with the same sequence number.
172  * This allows to verify if the other side has proper replay attacks detection.
173  */
174 VNET_DEFINE(int, ipsec_replay) = 0;
175 SYSCTL_VNET_INT(_net_inet_ipsec, OID_AUTO, test_replay,
176 	CTLFLAG_RW, &VNET_NAME(ipsec_replay), 0,
177 	"Emulate replay attack");
178 /*
179  * When set 1, IPsec will send packets with corrupted HMAC.
180  * This allows to verify if the other side properly detects modified packets.
181  */
182 VNET_DEFINE(int, ipsec_integrity) = 0;
183 SYSCTL_VNET_INT(_net_inet_ipsec, OID_AUTO, test_integrity,
184 	CTLFLAG_RW, &VNET_NAME(ipsec_integrity), 0,
185 	"Emulate man-in-the-middle attack");
186 #endif
187 
188 #ifdef INET6
189 VNET_DEFINE(struct ipsecstat, ipsec6stat);
190 VNET_DEFINE(int, ip6_esp_trans_deflev) = IPSEC_LEVEL_USE;
191 VNET_DEFINE(int, ip6_esp_net_deflev) = IPSEC_LEVEL_USE;
192 VNET_DEFINE(int, ip6_ah_trans_deflev) = IPSEC_LEVEL_USE;
193 VNET_DEFINE(int, ip6_ah_net_deflev) = IPSEC_LEVEL_USE;
194 VNET_DEFINE(int, ip6_ipsec_ecn) = 0;	/* ECN ignore(-1)/forbidden(0)/allowed(1) */
195 
196 SYSCTL_DECL(_net_inet6_ipsec6);
197 
198 /* net.inet6.ipsec6 */
199 #ifdef COMPAT_KAME
200 SYSCTL_OID(_net_inet6_ipsec6, IPSECCTL_STATS, stats, CTLFLAG_RD,
201     0, 0, compat_ipsecstats_sysctl, "S", "IPsec IPv6 statistics.");
202 #endif /* COMPAT_KAME */
203 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_POLICY, def_policy, CTLFLAG_RW,
204 	&VNET_NAME(ip4_def_policy).policy, 0,
205 	"IPsec default policy.");
206 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV,
207 	esp_trans_deflev, CTLFLAG_RW, &VNET_NAME(ip6_esp_trans_deflev),	0,
208 	"Default ESP transport mode level.");
209 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV,
210 	esp_net_deflev, CTLFLAG_RW, &VNET_NAME(ip6_esp_net_deflev),	0,
211 	"Default ESP tunnel mode level.");
212 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV,
213 	ah_trans_deflev, CTLFLAG_RW, &VNET_NAME(ip6_ah_trans_deflev),	0,
214 	"AH transfer mode default level.");
215 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV,
216 	ah_net_deflev, CTLFLAG_RW, &VNET_NAME(ip6_ah_net_deflev),	0,
217 	"AH tunnel mode default level.");
218 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_ECN,
219 	ecn, CTLFLAG_RW, &VNET_NAME(ip6_ipsec_ecn),	0,
220 	"Explicit Congestion Notification handling.");
221 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug, CTLFLAG_RW,
222 	&VNET_NAME(ipsec_debug), 0,
223 	"Enable IPsec debugging output when set.");
224 SYSCTL_VNET_STRUCT(_net_inet6_ipsec6, IPSECCTL_STATS,
225 	ipsecstats, CTLFLAG_RD, &VNET_NAME(ipsec6stat), ipsecstat,
226 	"IPsec IPv6 statistics.");
227 #endif /* INET6 */
228 
229 static int ipsec_setspidx_inpcb __P((struct mbuf *, struct inpcb *));
230 static int ipsec_setspidx __P((struct mbuf *, struct secpolicyindex *, int));
231 static void ipsec4_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
232 static int ipsec4_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
233 #ifdef INET6
234 static void ipsec6_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
235 static int ipsec6_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
236 #endif
237 static void ipsec_delpcbpolicy __P((struct inpcbpolicy *));
238 static struct secpolicy *ipsec_deepcopy_policy __P((struct secpolicy *src));
239 static void vshiftl __P((unsigned char *, int, int));
240 
241 MALLOC_DEFINE(M_IPSEC_INPCB, "inpcbpolicy", "inpcb-resident ipsec policy");
242 
243 /*
244  * Return a held reference to the default SP.
245  */
246 static struct secpolicy *
247 key_allocsp_default(const char* where, int tag)
248 {
249 	struct secpolicy *sp;
250 
251 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
252 		printf("DP key_allocsp_default from %s:%u\n", where, tag));
253 
254 	sp = &V_ip4_def_policy;
255 	if (sp->policy != IPSEC_POLICY_DISCARD &&
256 	    sp->policy != IPSEC_POLICY_NONE) {
257 		ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n",
258 		    sp->policy, IPSEC_POLICY_NONE));
259 		sp->policy = IPSEC_POLICY_NONE;
260 	}
261 	key_addref(sp);
262 
263 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
264 		printf("DP key_allocsp_default returns SP:%p (%u)\n",
265 			sp, sp->refcnt));
266 	return (sp);
267 }
268 #define	KEY_ALLOCSP_DEFAULT() \
269 	key_allocsp_default(__FILE__, __LINE__)
270 
271 /*
272  * For OUTBOUND packet having a socket. Searching SPD for packet,
273  * and return a pointer to SP.
274  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
275  *		0	: bypass
276  *		EACCES	: discard packet.
277  *		ENOENT	: ipsec_acquire() in progress, maybe.
278  *		others	: error occured.
279  *	others:	a pointer to SP
280  *
281  * NOTE: IPv6 mapped adddress concern is implemented here.
282  */
283 struct secpolicy *
284 ipsec_getpolicy(struct tdb_ident *tdbi, u_int dir)
285 {
286 	struct secpolicy *sp;
287 
288 	IPSEC_ASSERT(tdbi != NULL, ("null tdbi"));
289 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
290 		("invalid direction %u", dir));
291 
292 	sp = KEY_ALLOCSP2(tdbi->spi, &tdbi->dst, tdbi->proto, dir);
293 	if (sp == NULL)			/*XXX????*/
294 		sp = KEY_ALLOCSP_DEFAULT();
295 	IPSEC_ASSERT(sp != NULL, ("null SP"));
296 	return (sp);
297 }
298 
299 /*
300  * For OUTBOUND packet having a socket. Searching SPD for packet,
301  * and return a pointer to SP.
302  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
303  *		0	: bypass
304  *		EACCES	: discard packet.
305  *		ENOENT	: ipsec_acquire() in progress, maybe.
306  *		others	: error occured.
307  *	others:	a pointer to SP
308  *
309  * NOTE: IPv6 mapped adddress concern is implemented here.
310  */
311 static struct secpolicy *
312 ipsec_getpolicybysock(struct mbuf *m, u_int dir, struct inpcb *inp, int *error)
313 {
314 	struct inpcbpolicy *pcbsp;
315 	struct secpolicy *currsp = NULL;	/* Policy on socket. */
316 	struct secpolicy *sp;
317 
318 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
319 	IPSEC_ASSERT(inp != NULL, ("null inpcb"));
320 	IPSEC_ASSERT(error != NULL, ("null error"));
321 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
322 		("invalid direction %u", dir));
323 
324 	/* Set spidx in pcb. */
325 	*error = ipsec_setspidx_inpcb(m, inp);
326 	if (*error)
327 		return (NULL);
328 
329 	pcbsp = inp->inp_sp;
330 	IPSEC_ASSERT(pcbsp != NULL, ("null pcbsp"));
331 	switch (dir) {
332 	case IPSEC_DIR_INBOUND:
333 		currsp = pcbsp->sp_in;
334 		break;
335 	case IPSEC_DIR_OUTBOUND:
336 		currsp = pcbsp->sp_out;
337 		break;
338 	}
339 	IPSEC_ASSERT(currsp != NULL, ("null currsp"));
340 
341 	if (pcbsp->priv) {			/* When privilieged socket. */
342 		switch (currsp->policy) {
343 		case IPSEC_POLICY_BYPASS:
344 		case IPSEC_POLICY_IPSEC:
345 			key_addref(currsp);
346 			sp = currsp;
347 			break;
348 
349 		case IPSEC_POLICY_ENTRUST:
350 			/* Look for a policy in SPD. */
351 			sp = KEY_ALLOCSP(&currsp->spidx, dir);
352 			if (sp == NULL)		/* No SP found. */
353 				sp = KEY_ALLOCSP_DEFAULT();
354 			break;
355 
356 		default:
357 			ipseclog((LOG_ERR, "%s: Invalid policy for PCB %d\n",
358 				__func__, currsp->policy));
359 			*error = EINVAL;
360 			return (NULL);
361 		}
362 	} else {				/* Unpriv, SPD has policy. */
363 		sp = KEY_ALLOCSP(&currsp->spidx, dir);
364 		if (sp == NULL) {		/* No SP found. */
365 			switch (currsp->policy) {
366 			case IPSEC_POLICY_BYPASS:
367 				ipseclog((LOG_ERR, "%s: Illegal policy for "
368 					"non-priviliged defined %d\n",
369 					__func__, currsp->policy));
370 				*error = EINVAL;
371 				return (NULL);
372 
373 			case IPSEC_POLICY_ENTRUST:
374 				sp = KEY_ALLOCSP_DEFAULT();
375 				break;
376 
377 			case IPSEC_POLICY_IPSEC:
378 				key_addref(currsp);
379 				sp = currsp;
380 				break;
381 
382 			default:
383 				ipseclog((LOG_ERR, "%s: Invalid policy for "
384 					"PCB %d\n", __func__, currsp->policy));
385 				*error = EINVAL;
386 				return (NULL);
387 			}
388 		}
389 	}
390 	IPSEC_ASSERT(sp != NULL,
391 		("null SP (priv %u policy %u", pcbsp->priv, currsp->policy));
392 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
393 		printf("DP %s (priv %u policy %u) allocate SP:%p (refcnt %u)\n",
394 			__func__, pcbsp->priv, currsp->policy, sp, sp->refcnt));
395 	return (sp);
396 }
397 
398 /*
399  * For FORWADING packet or OUTBOUND without a socket. Searching SPD for packet,
400  * and return a pointer to SP.
401  * OUT:	positive: a pointer to the entry for security policy leaf matched.
402  *	NULL:	no apropreate SP found, the following value is set to error.
403  *		0	: bypass
404  *		EACCES	: discard packet.
405  *		ENOENT	: ipsec_acquire() in progress, maybe.
406  *		others	: error occured.
407  */
408 struct secpolicy *
409 ipsec_getpolicybyaddr(struct mbuf *m, u_int dir, int flag, int *error)
410 {
411 	struct secpolicyindex spidx;
412 	struct secpolicy *sp;
413 
414 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
415 	IPSEC_ASSERT(error != NULL, ("null error"));
416 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
417 		("invalid direction %u", dir));
418 
419 	sp = NULL;
420 	if (key_havesp(dir)) {
421 		/* Make an index to look for a policy. */
422 		*error = ipsec_setspidx(m, &spidx,
423 					(flag & IP_FORWARDING) ? 0 : 1);
424 		if (*error != 0) {
425 			DPRINTF(("%s: setpidx failed, dir %u flag %u\n",
426 				__func__, dir, flag));
427 			return (NULL);
428 		}
429 		spidx.dir = dir;
430 
431 		sp = KEY_ALLOCSP(&spidx, dir);
432 	}
433 	if (sp == NULL)			/* No SP found, use system default. */
434 		sp = KEY_ALLOCSP_DEFAULT();
435 	IPSEC_ASSERT(sp != NULL, ("null SP"));
436 	return (sp);
437 }
438 
439 struct secpolicy *
440 ipsec4_checkpolicy(struct mbuf *m, u_int dir, u_int flag, int *error,
441     struct inpcb *inp)
442 {
443 	struct secpolicy *sp;
444 
445 	*error = 0;
446 	if (inp == NULL)
447 		sp = ipsec_getpolicybyaddr(m, dir, flag, error);
448 	else
449 		sp = ipsec_getpolicybysock(m, dir, inp, error);
450 	if (sp == NULL) {
451 		IPSEC_ASSERT(*error != 0, ("getpolicy failed w/o error"));
452 		V_ipsec4stat.ips_out_inval++;
453 		return (NULL);
454 	}
455 	IPSEC_ASSERT(*error == 0, ("sp w/ error set to %u", *error));
456 	switch (sp->policy) {
457 	case IPSEC_POLICY_ENTRUST:
458 	default:
459 		printf("%s: invalid policy %u\n", __func__, sp->policy);
460 		/* FALLTHROUGH */
461 	case IPSEC_POLICY_DISCARD:
462 		V_ipsec4stat.ips_out_polvio++;
463 		*error = -EINVAL;	/* Packet is discarded by caller. */
464 		break;
465 	case IPSEC_POLICY_BYPASS:
466 	case IPSEC_POLICY_NONE:
467 		KEY_FREESP(&sp);
468 		sp = NULL;		/* NB: force NULL result. */
469 		break;
470 	case IPSEC_POLICY_IPSEC:
471 		if (sp->req == NULL)	/* Acquire a SA. */
472 			*error = key_spdacquire(sp);
473 		break;
474 	}
475 	if (*error != 0) {
476 		KEY_FREESP(&sp);
477 		sp = NULL;
478 	}
479 	return (sp);
480 }
481 
482 static int
483 ipsec_setspidx_inpcb(struct mbuf *m, struct inpcb *inp)
484 {
485 	int error;
486 
487 	IPSEC_ASSERT(inp != NULL, ("null inp"));
488 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
489 	IPSEC_ASSERT(inp->inp_sp->sp_out != NULL && inp->inp_sp->sp_in != NULL,
490 		("null sp_in || sp_out"));
491 
492 	error = ipsec_setspidx(m, &inp->inp_sp->sp_in->spidx, 1);
493 	if (error == 0) {
494 		inp->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
495 		inp->inp_sp->sp_out->spidx = inp->inp_sp->sp_in->spidx;
496 		inp->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
497 	} else {
498 		bzero(&inp->inp_sp->sp_in->spidx,
499 			sizeof (inp->inp_sp->sp_in->spidx));
500 		bzero(&inp->inp_sp->sp_out->spidx,
501 			sizeof (inp->inp_sp->sp_in->spidx));
502 	}
503 	return (error);
504 }
505 
506 /*
507  * Configure security policy index (src/dst/proto/sport/dport)
508  * by looking at the content of mbuf.
509  * The caller is responsible for error recovery (like clearing up spidx).
510  */
511 static int
512 ipsec_setspidx(struct mbuf *m, struct secpolicyindex *spidx, int needport)
513 {
514 	struct ip *ip = NULL;
515 	struct ip ipbuf;
516 	u_int v;
517 	struct mbuf *n;
518 	int len;
519 	int error;
520 
521 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
522 
523 	/*
524 	 * Validate m->m_pkthdr.len.  We see incorrect length if we
525 	 * mistakenly call this function with inconsistent mbuf chain
526 	 * (like 4.4BSD tcp/udp processing).  XXX Should we panic here?
527 	 */
528 	len = 0;
529 	for (n = m; n; n = n->m_next)
530 		len += n->m_len;
531 	if (m->m_pkthdr.len != len) {
532 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
533 			printf("%s: pkthdr len(%d) mismatch (%d), ignored.\n",
534 				__func__, len, m->m_pkthdr.len));
535 		return (EINVAL);
536 	}
537 
538 	if (m->m_pkthdr.len < sizeof(struct ip)) {
539 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
540 			printf("%s: pkthdr len(%d) too small (v4), ignored.\n",
541 			    __func__, m->m_pkthdr.len));
542 		return (EINVAL);
543 	}
544 
545 	if (m->m_len >= sizeof(*ip))
546 		ip = mtod(m, struct ip *);
547 	else {
548 		m_copydata(m, 0, sizeof(ipbuf), (caddr_t)&ipbuf);
549 		ip = &ipbuf;
550 	}
551 #ifdef _IP_VHL
552 	v = _IP_VHL_V(ip->ip_vhl);
553 #else
554 	v = ip->ip_v;
555 #endif
556 	switch (v) {
557 	case 4:
558 		error = ipsec4_setspidx_ipaddr(m, spidx);
559 		if (error)
560 			return (error);
561 		ipsec4_get_ulp(m, spidx, needport);
562 		return (0);
563 #ifdef INET6
564 	case 6:
565 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) {
566 			KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
567 				printf("%s: pkthdr len(%d) too small (v6), "
568 				"ignored\n", __func__, m->m_pkthdr.len));
569 			return (EINVAL);
570 		}
571 		error = ipsec6_setspidx_ipaddr(m, spidx);
572 		if (error)
573 			return (error);
574 		ipsec6_get_ulp(m, spidx, needport);
575 		return (0);
576 #endif
577 	default:
578 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
579 			printf("%s: " "unknown IP version %u, ignored.\n",
580 				__func__, v));
581 		return (EINVAL);
582 	}
583 }
584 
585 static void
586 ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
587 {
588 	u_int8_t nxt;
589 	int off;
590 
591 	/* Sanity check. */
592 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
593 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),("packet too short"));
594 
595 	/* NB: ip_input() flips it into host endian. XXX Need more checking. */
596 	if (m->m_len >= sizeof (struct ip)) {
597 		struct ip *ip = mtod(m, struct ip *);
598 		if (ip->ip_off & (IP_MF | IP_OFFMASK))
599 			goto done;
600 #ifdef _IP_VHL
601 		off = _IP_VHL_HL(ip->ip_vhl) << 2;
602 #else
603 		off = ip->ip_hl << 2;
604 #endif
605 		nxt = ip->ip_p;
606 	} else {
607 		struct ip ih;
608 
609 		m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih);
610 		if (ih.ip_off & (IP_MF | IP_OFFMASK))
611 			goto done;
612 #ifdef _IP_VHL
613 		off = _IP_VHL_HL(ih.ip_vhl) << 2;
614 #else
615 		off = ih.ip_hl << 2;
616 #endif
617 		nxt = ih.ip_p;
618 	}
619 
620 	while (off < m->m_pkthdr.len) {
621 		struct ip6_ext ip6e;
622 		struct tcphdr th;
623 		struct udphdr uh;
624 
625 		switch (nxt) {
626 		case IPPROTO_TCP:
627 			spidx->ul_proto = nxt;
628 			if (!needport)
629 				goto done_proto;
630 			if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
631 				goto done;
632 			m_copydata(m, off, sizeof (th), (caddr_t) &th);
633 			spidx->src.sin.sin_port = th.th_sport;
634 			spidx->dst.sin.sin_port = th.th_dport;
635 			return;
636 		case IPPROTO_UDP:
637 			spidx->ul_proto = nxt;
638 			if (!needport)
639 				goto done_proto;
640 			if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
641 				goto done;
642 			m_copydata(m, off, sizeof (uh), (caddr_t) &uh);
643 			spidx->src.sin.sin_port = uh.uh_sport;
644 			spidx->dst.sin.sin_port = uh.uh_dport;
645 			return;
646 		case IPPROTO_AH:
647 			if (off + sizeof(ip6e) > m->m_pkthdr.len)
648 				goto done;
649 			/* XXX Sigh, this works but is totally bogus. */
650 			m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e);
651 			off += (ip6e.ip6e_len + 2) << 2;
652 			nxt = ip6e.ip6e_nxt;
653 			break;
654 		case IPPROTO_ICMP:
655 		default:
656 			/* XXX Intermediate headers??? */
657 			spidx->ul_proto = nxt;
658 			goto done_proto;
659 		}
660 	}
661 done:
662 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
663 done_proto:
664 	spidx->src.sin.sin_port = IPSEC_PORT_ANY;
665 	spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
666 }
667 
668 /* Assumes that m is sane. */
669 static int
670 ipsec4_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
671 {
672 	static const struct sockaddr_in template = {
673 		sizeof (struct sockaddr_in),
674 		AF_INET,
675 		0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 }
676 	};
677 
678 	spidx->src.sin = template;
679 	spidx->dst.sin = template;
680 
681 	if (m->m_len < sizeof (struct ip)) {
682 		m_copydata(m, offsetof(struct ip, ip_src),
683 			   sizeof (struct  in_addr),
684 			   (caddr_t) &spidx->src.sin.sin_addr);
685 		m_copydata(m, offsetof(struct ip, ip_dst),
686 			   sizeof (struct  in_addr),
687 			   (caddr_t) &spidx->dst.sin.sin_addr);
688 	} else {
689 		struct ip *ip = mtod(m, struct ip *);
690 		spidx->src.sin.sin_addr = ip->ip_src;
691 		spidx->dst.sin.sin_addr = ip->ip_dst;
692 	}
693 
694 	spidx->prefs = sizeof(struct in_addr) << 3;
695 	spidx->prefd = sizeof(struct in_addr) << 3;
696 
697 	return (0);
698 }
699 
700 #ifdef INET6
701 static void
702 ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
703 {
704 	int off, nxt;
705 	struct tcphdr th;
706 	struct udphdr uh;
707 	struct icmp6_hdr ih;
708 
709 	/* Sanity check. */
710 	if (m == NULL)
711 		panic("%s: NULL pointer was passed.\n", __func__);
712 
713 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
714 		printf("%s:\n", __func__); kdebug_mbuf(m));
715 
716 	/* Set default. */
717 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
718 	((struct sockaddr_in6 *)&spidx->src)->sin6_port = IPSEC_PORT_ANY;
719 	((struct sockaddr_in6 *)&spidx->dst)->sin6_port = IPSEC_PORT_ANY;
720 
721 	nxt = -1;
722 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
723 	if (off < 0 || m->m_pkthdr.len < off)
724 		return;
725 
726 	switch (nxt) {
727 	case IPPROTO_TCP:
728 		spidx->ul_proto = nxt;
729 		if (!needport)
730 			break;
731 		if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
732 			break;
733 		m_copydata(m, off, sizeof(th), (caddr_t)&th);
734 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = th.th_sport;
735 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = th.th_dport;
736 		break;
737 	case IPPROTO_UDP:
738 		spidx->ul_proto = nxt;
739 		if (!needport)
740 			break;
741 		if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
742 			break;
743 		m_copydata(m, off, sizeof(uh), (caddr_t)&uh);
744 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = uh.uh_sport;
745 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = uh.uh_dport;
746 		break;
747 	case IPPROTO_ICMPV6:
748 		spidx->ul_proto = nxt;
749 		if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len)
750 			break;
751 		m_copydata(m, off, sizeof(ih), (caddr_t)&ih);
752 		((struct sockaddr_in6 *)&spidx->src)->sin6_port =
753 		    htons((uint16_t)ih.icmp6_type);
754 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port =
755 		    htons((uint16_t)ih.icmp6_code);
756 		break;
757 	default:
758 		/* XXX Intermediate headers??? */
759 		spidx->ul_proto = nxt;
760 		break;
761 	}
762 }
763 
764 /* Assumes that m is sane. */
765 static int
766 ipsec6_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
767 {
768 	struct ip6_hdr *ip6 = NULL;
769 	struct ip6_hdr ip6buf;
770 	struct sockaddr_in6 *sin6;
771 
772 	if (m->m_len >= sizeof(*ip6))
773 		ip6 = mtod(m, struct ip6_hdr *);
774 	else {
775 		m_copydata(m, 0, sizeof(ip6buf), (caddr_t)&ip6buf);
776 		ip6 = &ip6buf;
777 	}
778 
779 	sin6 = (struct sockaddr_in6 *)&spidx->src;
780 	bzero(sin6, sizeof(*sin6));
781 	sin6->sin6_family = AF_INET6;
782 	sin6->sin6_len = sizeof(struct sockaddr_in6);
783 	bcopy(&ip6->ip6_src, &sin6->sin6_addr, sizeof(ip6->ip6_src));
784 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
785 		sin6->sin6_addr.s6_addr16[1] = 0;
786 		sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]);
787 	}
788 	spidx->prefs = sizeof(struct in6_addr) << 3;
789 
790 	sin6 = (struct sockaddr_in6 *)&spidx->dst;
791 	bzero(sin6, sizeof(*sin6));
792 	sin6->sin6_family = AF_INET6;
793 	sin6->sin6_len = sizeof(struct sockaddr_in6);
794 	bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(ip6->ip6_dst));
795 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
796 		sin6->sin6_addr.s6_addr16[1] = 0;
797 		sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]);
798 	}
799 	spidx->prefd = sizeof(struct in6_addr) << 3;
800 
801 	return (0);
802 }
803 #endif
804 
805 static void
806 ipsec_delpcbpolicy(struct inpcbpolicy *p)
807 {
808 
809 	free(p, M_IPSEC_INPCB);
810 }
811 
812 /* Initialize policy in PCB. */
813 int
814 ipsec_init_policy(struct socket *so, struct inpcbpolicy **pcb_sp)
815 {
816 	struct inpcbpolicy *new;
817 
818 	/* Sanity check. */
819 	if (so == NULL || pcb_sp == NULL)
820 		panic("%s: NULL pointer was passed.\n", __func__);
821 
822 	new = (struct inpcbpolicy *) malloc(sizeof(struct inpcbpolicy),
823 					    M_IPSEC_INPCB, M_NOWAIT|M_ZERO);
824 	if (new == NULL) {
825 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
826 		return (ENOBUFS);
827 	}
828 
829 	new->priv = IPSEC_IS_PRIVILEGED_SO(so);
830 
831 	if ((new->sp_in = KEY_NEWSP()) == NULL) {
832 		ipsec_delpcbpolicy(new);
833 		return (ENOBUFS);
834 	}
835 	new->sp_in->state = IPSEC_SPSTATE_ALIVE;
836 	new->sp_in->policy = IPSEC_POLICY_ENTRUST;
837 
838 	if ((new->sp_out = KEY_NEWSP()) == NULL) {
839 		KEY_FREESP(&new->sp_in);
840 		ipsec_delpcbpolicy(new);
841 		return (ENOBUFS);
842 	}
843 	new->sp_out->state = IPSEC_SPSTATE_ALIVE;
844 	new->sp_out->policy = IPSEC_POLICY_ENTRUST;
845 
846 	*pcb_sp = new;
847 
848 	return (0);
849 }
850 
851 /* Copy old IPsec policy into new. */
852 int
853 ipsec_copy_policy(struct inpcbpolicy *old, struct inpcbpolicy *new)
854 {
855 	struct secpolicy *sp;
856 
857 	sp = ipsec_deepcopy_policy(old->sp_in);
858 	if (sp) {
859 		KEY_FREESP(&new->sp_in);
860 		new->sp_in = sp;
861 	} else
862 		return (ENOBUFS);
863 
864 	sp = ipsec_deepcopy_policy(old->sp_out);
865 	if (sp) {
866 		KEY_FREESP(&new->sp_out);
867 		new->sp_out = sp;
868 	} else
869 		return (ENOBUFS);
870 
871 	new->priv = old->priv;
872 
873 	return (0);
874 }
875 
876 struct ipsecrequest *
877 ipsec_newisr(void)
878 {
879 	struct ipsecrequest *p;
880 
881 	p = malloc(sizeof(struct ipsecrequest), M_IPSEC_SR, M_NOWAIT|M_ZERO);
882 	if (p != NULL)
883 		IPSECREQUEST_LOCK_INIT(p);
884 	return (p);
885 }
886 
887 void
888 ipsec_delisr(struct ipsecrequest *p)
889 {
890 
891 	IPSECREQUEST_LOCK_DESTROY(p);
892 	free(p, M_IPSEC_SR);
893 }
894 
895 /* Deep-copy a policy in PCB. */
896 static struct secpolicy *
897 ipsec_deepcopy_policy(struct secpolicy *src)
898 {
899 	struct ipsecrequest *newchain = NULL;
900 	struct ipsecrequest *p;
901 	struct ipsecrequest **q;
902 	struct ipsecrequest *r;
903 	struct secpolicy *dst;
904 
905 	if (src == NULL)
906 		return (NULL);
907 	dst = KEY_NEWSP();
908 	if (dst == NULL)
909 		return (NULL);
910 
911 	/*
912 	 * Deep-copy IPsec request chain.  This is required since struct
913 	 * ipsecrequest is not reference counted.
914 	 */
915 	q = &newchain;
916 	for (p = src->req; p; p = p->next) {
917 		*q = ipsec_newisr();
918 		if (*q == NULL)
919 			goto fail;
920 		(*q)->saidx.proto = p->saidx.proto;
921 		(*q)->saidx.mode = p->saidx.mode;
922 		(*q)->level = p->level;
923 		(*q)->saidx.reqid = p->saidx.reqid;
924 
925 		bcopy(&p->saidx.src, &(*q)->saidx.src, sizeof((*q)->saidx.src));
926 		bcopy(&p->saidx.dst, &(*q)->saidx.dst, sizeof((*q)->saidx.dst));
927 
928 		(*q)->sp = dst;
929 
930 		q = &((*q)->next);
931 	}
932 
933 	dst->req = newchain;
934 	dst->state = src->state;
935 	dst->policy = src->policy;
936 	/* Do not touch the refcnt fields. */
937 
938 	return (dst);
939 
940 fail:
941 	for (p = newchain; p; p = r) {
942 		r = p->next;
943 		ipsec_delisr(p);
944 		p = NULL;
945 	}
946 	return (NULL);
947 }
948 
949 /* Set policy and IPsec request if present. */
950 static int
951 ipsec_set_policy_internal(struct secpolicy **pcb_sp, int optname,
952     caddr_t request, size_t len, struct ucred *cred)
953 {
954 	struct sadb_x_policy *xpl;
955 	struct secpolicy *newsp = NULL;
956 	int error;
957 
958 	/* Sanity check. */
959 	if (pcb_sp == NULL || *pcb_sp == NULL || request == NULL)
960 		return (EINVAL);
961 	if (len < sizeof(*xpl))
962 		return (EINVAL);
963 	xpl = (struct sadb_x_policy *)request;
964 
965 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
966 		printf("%s: passed policy\n", __func__);
967 		kdebug_sadb_x_policy((struct sadb_ext *)xpl));
968 
969 	/* Check policy type. */
970 	/* ipsec_set_policy_internal() accepts IPSEC, ENTRUST and BYPASS. */
971 	if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD
972 	 || xpl->sadb_x_policy_type == IPSEC_POLICY_NONE)
973 		return (EINVAL);
974 
975 	/* Check privileged socket. */
976 	if (cred != NULL && xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
977 		error = priv_check_cred(cred, PRIV_NETINET_IPSEC, 0);
978 		if (error)
979 			return (EACCES);
980 	}
981 
982 	/* Allocating new SP entry. */
983 	if ((newsp = key_msg2sp(xpl, len, &error)) == NULL)
984 		return (error);
985 
986 	newsp->state = IPSEC_SPSTATE_ALIVE;
987 
988 	/* Clear old SP and set new SP. */
989 	KEY_FREESP(pcb_sp);
990 	*pcb_sp = newsp;
991 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
992 		printf("%s: new policy\n", __func__);
993 		kdebug_secpolicy(newsp));
994 
995 	return (0);
996 }
997 
998 int
999 ipsec_set_policy(struct inpcb *inp, int optname, caddr_t request,
1000     size_t len, struct ucred *cred)
1001 {
1002 	struct sadb_x_policy *xpl;
1003 	struct secpolicy **pcb_sp;
1004 
1005 	/* Sanity check. */
1006 	if (inp == NULL || request == NULL)
1007 		return (EINVAL);
1008 	if (len < sizeof(*xpl))
1009 		return (EINVAL);
1010 	xpl = (struct sadb_x_policy *)request;
1011 
1012 	/* Select direction. */
1013 	switch (xpl->sadb_x_policy_dir) {
1014 	case IPSEC_DIR_INBOUND:
1015 		pcb_sp = &inp->inp_sp->sp_in;
1016 		break;
1017 	case IPSEC_DIR_OUTBOUND:
1018 		pcb_sp = &inp->inp_sp->sp_out;
1019 		break;
1020 	default:
1021 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1022 			xpl->sadb_x_policy_dir));
1023 		return (EINVAL);
1024 	}
1025 
1026 	return (ipsec_set_policy_internal(pcb_sp, optname, request, len, cred));
1027 }
1028 
1029 int
1030 ipsec_get_policy(struct inpcb *inp, caddr_t request, size_t len,
1031     struct mbuf **mp)
1032 {
1033 	struct sadb_x_policy *xpl;
1034 	struct secpolicy *pcb_sp;
1035 
1036 	/* Sanity check. */
1037 	if (inp == NULL || request == NULL || mp == NULL)
1038 		return (EINVAL);
1039 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
1040 	if (len < sizeof(*xpl))
1041 		return (EINVAL);
1042 	xpl = (struct sadb_x_policy *)request;
1043 
1044 	/* Select direction. */
1045 	switch (xpl->sadb_x_policy_dir) {
1046 	case IPSEC_DIR_INBOUND:
1047 		pcb_sp = inp->inp_sp->sp_in;
1048 		break;
1049 	case IPSEC_DIR_OUTBOUND:
1050 		pcb_sp = inp->inp_sp->sp_out;
1051 		break;
1052 	default:
1053 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1054 			xpl->sadb_x_policy_dir));
1055 		return (EINVAL);
1056 	}
1057 
1058 	/* Sanity check. Should be an IPSEC_ASSERT. */
1059 	if (pcb_sp == NULL)
1060 		return (EINVAL);
1061 
1062 	*mp = key_sp2msg(pcb_sp);
1063 	if (!*mp) {
1064 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
1065 		return (ENOBUFS);
1066 	}
1067 
1068 	(*mp)->m_type = MT_DATA;
1069 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1070 		printf("%s:\n", __func__); kdebug_mbuf(*mp));
1071 
1072 	return (0);
1073 }
1074 
1075 /* Delete policy in PCB. */
1076 int
1077 ipsec_delete_pcbpolicy(struct inpcb *inp)
1078 {
1079 	IPSEC_ASSERT(inp != NULL, ("null inp"));
1080 
1081 	if (inp->inp_sp == NULL)
1082 		return (0);
1083 
1084 	if (inp->inp_sp->sp_in != NULL)
1085 		KEY_FREESP(&inp->inp_sp->sp_in);
1086 
1087 	if (inp->inp_sp->sp_out != NULL)
1088 		KEY_FREESP(&inp->inp_sp->sp_out);
1089 
1090 	ipsec_delpcbpolicy(inp->inp_sp);
1091 	inp->inp_sp = NULL;
1092 
1093 	return (0);
1094 }
1095 
1096 /*
1097  * Return current level.
1098  * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
1099  */
1100 u_int
1101 ipsec_get_reqlevel(struct ipsecrequest *isr)
1102 {
1103 	u_int level = 0;
1104 	u_int esp_trans_deflev, esp_net_deflev;
1105 	u_int ah_trans_deflev, ah_net_deflev;
1106 
1107 	IPSEC_ASSERT(isr != NULL && isr->sp != NULL, ("null argument"));
1108 	IPSEC_ASSERT(isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family,
1109 		("af family mismatch, src %u, dst %u",
1110 		 isr->sp->spidx.src.sa.sa_family,
1111 		 isr->sp->spidx.dst.sa.sa_family));
1112 
1113 /* XXX Note that we have ipseclog() expanded here - code sync issue. */
1114 #define IPSEC_CHECK_DEFAULT(lev) \
1115 	(((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE	      \
1116 			&& (lev) != IPSEC_LEVEL_UNIQUE)			      \
1117 		? (V_ipsec_debug						      \
1118 			? log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
1119 				(lev), IPSEC_LEVEL_REQUIRE)		      \
1120 			: 0),						      \
1121 			(lev) = IPSEC_LEVEL_REQUIRE,			      \
1122 			(lev)						      \
1123 		: (lev))
1124 
1125 	/* Set default level. */
1126 	switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) {
1127 #ifdef INET
1128 	case AF_INET:
1129 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev);
1130 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev);
1131 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev);
1132 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev);
1133 		break;
1134 #endif
1135 #ifdef INET6
1136 	case AF_INET6:
1137 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev);
1138 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev);
1139 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev);
1140 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev);
1141 		break;
1142 #endif /* INET6 */
1143 	default:
1144 		panic("%s: unknown af %u",
1145 			__func__, isr->sp->spidx.src.sa.sa_family);
1146 	}
1147 
1148 #undef IPSEC_CHECK_DEFAULT
1149 
1150 	/* Set level. */
1151 	switch (isr->level) {
1152 	case IPSEC_LEVEL_DEFAULT:
1153 		switch (isr->saidx.proto) {
1154 		case IPPROTO_ESP:
1155 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1156 				level = esp_net_deflev;
1157 			else
1158 				level = esp_trans_deflev;
1159 			break;
1160 		case IPPROTO_AH:
1161 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1162 				level = ah_net_deflev;
1163 			else
1164 				level = ah_trans_deflev;
1165 			break;
1166 		case IPPROTO_IPCOMP:
1167 			/*
1168 			 * We don't really care, as IPcomp document says that
1169 			 * we shouldn't compress small packets.
1170 			 */
1171 			level = IPSEC_LEVEL_USE;
1172 			break;
1173 		default:
1174 			panic("%s: Illegal protocol defined %u\n", __func__,
1175 				isr->saidx.proto);
1176 		}
1177 		break;
1178 
1179 	case IPSEC_LEVEL_USE:
1180 	case IPSEC_LEVEL_REQUIRE:
1181 		level = isr->level;
1182 		break;
1183 	case IPSEC_LEVEL_UNIQUE:
1184 		level = IPSEC_LEVEL_REQUIRE;
1185 		break;
1186 
1187 	default:
1188 		panic("%s: Illegal IPsec level %u\n", __func__, isr->level);
1189 	}
1190 
1191 	return (level);
1192 }
1193 
1194 /*
1195  * Check security policy requirements against the actual
1196  * packet contents.  Return one if the packet should be
1197  * reject as "invalid"; otherwiser return zero to have the
1198  * packet treated as "valid".
1199  *
1200  * OUT:
1201  *	0: valid
1202  *	1: invalid
1203  */
1204 int
1205 ipsec_in_reject(struct secpolicy *sp, struct mbuf *m)
1206 {
1207 	struct ipsecrequest *isr;
1208 	int need_auth;
1209 
1210 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1211 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1212 
1213 	/* Check policy. */
1214 	switch (sp->policy) {
1215 	case IPSEC_POLICY_DISCARD:
1216 		return (1);
1217 	case IPSEC_POLICY_BYPASS:
1218 	case IPSEC_POLICY_NONE:
1219 		return (0);
1220 	}
1221 
1222 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1223 		("invalid policy %u", sp->policy));
1224 
1225 	/* XXX Should compare policy against IPsec header history. */
1226 
1227 	need_auth = 0;
1228 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1229 		if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE)
1230 			continue;
1231 		switch (isr->saidx.proto) {
1232 		case IPPROTO_ESP:
1233 			if ((m->m_flags & M_DECRYPTED) == 0) {
1234 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1235 				    printf("%s: ESP m_flags:%x\n", __func__,
1236 					    m->m_flags));
1237 				return (1);
1238 			}
1239 
1240 			if (!need_auth &&
1241 			    isr->sav != NULL &&
1242 			    isr->sav->tdb_authalgxform != NULL &&
1243 			    (m->m_flags & M_AUTHIPDGM) == 0) {
1244 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1245 				    printf("%s: ESP/AH m_flags:%x\n", __func__,
1246 					    m->m_flags));
1247 				return (1);
1248 			}
1249 			break;
1250 		case IPPROTO_AH:
1251 			need_auth = 1;
1252 			if ((m->m_flags & M_AUTHIPHDR) == 0) {
1253 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1254 				    printf("%s: AH m_flags:%x\n", __func__,
1255 					    m->m_flags));
1256 				return (1);
1257 			}
1258 			break;
1259 		case IPPROTO_IPCOMP:
1260 			/*
1261 			 * We don't really care, as IPcomp document
1262 			 * says that we shouldn't compress small
1263 			 * packets.  IPComp policy should always be
1264 			 * treated as being in "use" level.
1265 			 */
1266 			break;
1267 		}
1268 	}
1269 	return (0);		/* Valid. */
1270 }
1271 
1272 static int
1273 ipsec46_in_reject(struct mbuf *m, struct inpcb *inp)
1274 {
1275 	struct secpolicy *sp;
1276 	int error;
1277 	int result;
1278 
1279 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1280 
1281 	/*
1282 	 * Get SP for this packet.
1283 	 * When we are called from ip_forward(), we call
1284 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1285 	 */
1286 	if (inp == NULL)
1287 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, IP_FORWARDING, &error);
1288 	else
1289 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND, inp, &error);
1290 
1291 	if (sp != NULL) {
1292 		result = ipsec_in_reject(sp, m);
1293 		KEY_FREESP(&sp);
1294 	} else {
1295 		result = 0;	/* XXX Should be panic?
1296 				 * -> No, there may be error. */
1297 	}
1298 	return (result);
1299 }
1300 
1301 /*
1302  * Check AH/ESP integrity.
1303  * This function is called from tcp_input(), udp_input(),
1304  * and {ah,esp}4_input for tunnel mode.
1305  */
1306 int
1307 ipsec4_in_reject(struct mbuf *m, struct inpcb *inp)
1308 {
1309 	int result;
1310 
1311 	result = ipsec46_in_reject(m, inp);
1312 	if (result)
1313 		V_ipsec4stat.ips_in_polvio++;
1314 
1315 	return (result);
1316 }
1317 
1318 #ifdef INET6
1319 /*
1320  * Check AH/ESP integrity.
1321  * This function is called from tcp6_input(), udp6_input(),
1322  * and {ah,esp}6_input for tunnel mode.
1323  */
1324 int
1325 ipsec6_in_reject(struct mbuf *m, struct inpcb *inp)
1326 {
1327 	int result;
1328 
1329 	result = ipsec46_in_reject(m, inp);
1330 	if (result)
1331 		V_ipsec6stat.ips_in_polvio++;
1332 
1333 	return (result);
1334 }
1335 #endif
1336 
1337 /*
1338  * Compute the byte size to be occupied by IPsec header.
1339  * In case it is tunnelled, it includes the size of outer IP header.
1340  * NOTE: SP passed is freed in this function.
1341  */
1342 static size_t
1343 ipsec_hdrsiz_internal(struct secpolicy *sp)
1344 {
1345 	struct ipsecrequest *isr;
1346 	size_t size;
1347 
1348 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1349 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1350 
1351 	switch (sp->policy) {
1352 	case IPSEC_POLICY_DISCARD:
1353 	case IPSEC_POLICY_BYPASS:
1354 	case IPSEC_POLICY_NONE:
1355 		return (0);
1356 	}
1357 
1358 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1359 		("invalid policy %u", sp->policy));
1360 
1361 	size = 0;
1362 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1363 		size_t clen = 0;
1364 
1365 		switch (isr->saidx.proto) {
1366 		case IPPROTO_ESP:
1367 			clen = esp_hdrsiz(isr->sav);
1368 			break;
1369 		case IPPROTO_AH:
1370 			clen = ah_hdrsiz(isr->sav);
1371 			break;
1372 		case IPPROTO_IPCOMP:
1373 			clen = sizeof(struct ipcomp);
1374 			break;
1375 		}
1376 
1377 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1378 			switch (isr->saidx.dst.sa.sa_family) {
1379 			case AF_INET:
1380 				clen += sizeof(struct ip);
1381 				break;
1382 #ifdef INET6
1383 			case AF_INET6:
1384 				clen += sizeof(struct ip6_hdr);
1385 				break;
1386 #endif
1387 			default:
1388 				ipseclog((LOG_ERR, "%s: unknown AF %d in "
1389 				    "IPsec tunnel SA\n", __func__,
1390 				    ((struct sockaddr *)&isr->saidx.dst)->sa_family));
1391 				break;
1392 			}
1393 		}
1394 		size += clen;
1395 	}
1396 
1397 	return (size);
1398 }
1399 
1400 /*
1401  * This function is called from ipsec_hdrsiz_tcp(), ip_ipsec_mtu(),
1402  * disabled ip6_ipsec_mtu() and ip6_forward().
1403  */
1404 size_t
1405 ipsec_hdrsiz(struct mbuf *m, u_int dir, struct inpcb *inp)
1406 {
1407 	struct secpolicy *sp;
1408 	int error;
1409 	size_t size;
1410 
1411 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1412 
1413 	/* Get SP for this packet.
1414 	 * When we are called from ip_forward(), we call
1415 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1416 	 */
1417 	if (inp == NULL)
1418 		sp = ipsec_getpolicybyaddr(m, dir, IP_FORWARDING, &error);
1419 	else
1420 		sp = ipsec_getpolicybysock(m, dir, inp, &error);
1421 
1422 	if (sp != NULL) {
1423 		size = ipsec_hdrsiz_internal(sp);
1424 		KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1425 			printf("%s: size:%lu.\n", __func__,
1426 				(unsigned long)size));
1427 
1428 		KEY_FREESP(&sp);
1429 	} else {
1430 		size = 0;	/* XXX Should be panic?
1431 				 * -> No, we are called w/o knowing if
1432 				 *    IPsec processing is needed. */
1433 	}
1434 	return (size);
1435 }
1436 
1437 /*
1438  * Check the variable replay window.
1439  * ipsec_chkreplay() performs replay check before ICV verification.
1440  * ipsec_updatereplay() updates replay bitmap.  This must be called after
1441  * ICV verification (it also performs replay check, which is usually done
1442  * beforehand).
1443  * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1444  *
1445  * Based on RFC 2401.
1446  */
1447 int
1448 ipsec_chkreplay(u_int32_t seq, struct secasvar *sav)
1449 {
1450 	const struct secreplay *replay;
1451 	u_int32_t diff;
1452 	int fr;
1453 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1454 	int frlast;		/* Constant: last frame. */
1455 
1456 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1457 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1458 
1459 	replay = sav->replay;
1460 
1461 	if (replay->wsize == 0)
1462 		return (1);	/* No need to check replay. */
1463 
1464 	/* Constant. */
1465 	frlast = replay->wsize - 1;
1466 	wsizeb = replay->wsize << 3;
1467 
1468 	/* Sequence number of 0 is invalid. */
1469 	if (seq == 0)
1470 		return (0);
1471 
1472 	/* First time is always okay. */
1473 	if (replay->count == 0)
1474 		return (1);
1475 
1476 	if (seq > replay->lastseq) {
1477 		/* Larger sequences are okay. */
1478 		return (1);
1479 	} else {
1480 		/* seq is equal or less than lastseq. */
1481 		diff = replay->lastseq - seq;
1482 
1483 		/* Over range to check, i.e. too old or wrapped. */
1484 		if (diff >= wsizeb)
1485 			return (0);
1486 
1487 		fr = frlast - diff / 8;
1488 
1489 		/* This packet already seen? */
1490 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1491 			return (0);
1492 
1493 		/* Out of order but good. */
1494 		return (1);
1495 	}
1496 }
1497 
1498 /*
1499  * Check replay counter whether to update or not.
1500  * OUT:	0:	OK
1501  *	1:	NG
1502  */
1503 int
1504 ipsec_updatereplay(u_int32_t seq, struct secasvar *sav)
1505 {
1506 	struct secreplay *replay;
1507 	u_int32_t diff;
1508 	int fr;
1509 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1510 	int frlast;		/* Constant: last frame. */
1511 
1512 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1513 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1514 
1515 	replay = sav->replay;
1516 
1517 	if (replay->wsize == 0)
1518 		goto ok;	/* No need to check replay. */
1519 
1520 	/* Constant. */
1521 	frlast = replay->wsize - 1;
1522 	wsizeb = replay->wsize << 3;
1523 
1524 	/* Sequence number of 0 is invalid. */
1525 	if (seq == 0)
1526 		return (1);
1527 
1528 	/* First time. */
1529 	if (replay->count == 0) {
1530 		replay->lastseq = seq;
1531 		bzero(replay->bitmap, replay->wsize);
1532 		(replay->bitmap)[frlast] = 1;
1533 		goto ok;
1534 	}
1535 
1536 	if (seq > replay->lastseq) {
1537 		/* seq is larger than lastseq. */
1538 		diff = seq - replay->lastseq;
1539 
1540 		/* New larger sequence number. */
1541 		if (diff < wsizeb) {
1542 			/* In window. */
1543 			/* Set bit for this packet. */
1544 			vshiftl(replay->bitmap, diff, replay->wsize);
1545 			(replay->bitmap)[frlast] |= 1;
1546 		} else {
1547 			/* This packet has a "way larger". */
1548 			bzero(replay->bitmap, replay->wsize);
1549 			(replay->bitmap)[frlast] = 1;
1550 		}
1551 		replay->lastseq = seq;
1552 
1553 		/* Larger is good. */
1554 	} else {
1555 		/* seq is equal or less than lastseq. */
1556 		diff = replay->lastseq - seq;
1557 
1558 		/* Over range to check, i.e. too old or wrapped. */
1559 		if (diff >= wsizeb)
1560 			return (1);
1561 
1562 		fr = frlast - diff / 8;
1563 
1564 		/* This packet already seen? */
1565 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1566 			return (1);
1567 
1568 		/* Mark as seen. */
1569 		(replay->bitmap)[fr] |= (1 << (diff % 8));
1570 
1571 		/* Out of order but good. */
1572 	}
1573 
1574 ok:
1575 	if (replay->count == ~0) {
1576 
1577 		/* Set overflow flag. */
1578 		replay->overflow++;
1579 
1580 		/* Don't increment, no more packets accepted. */
1581 		if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0)
1582 			return (1);
1583 
1584 		ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n",
1585 		    __func__, replay->overflow, ipsec_logsastr(sav)));
1586 	}
1587 
1588 	replay->count++;
1589 
1590 	return (0);
1591 }
1592 
1593 /*
1594  * Shift variable length buffer to left.
1595  * IN:	bitmap: pointer to the buffer
1596  * 	nbit:	the number of to shift.
1597  *	wsize:	buffer size (bytes).
1598  */
1599 static void
1600 vshiftl(unsigned char *bitmap, int nbit, int wsize)
1601 {
1602 	int s, j, i;
1603 	unsigned char over;
1604 
1605 	for (j = 0; j < nbit; j += 8) {
1606 		s = (nbit - j < 8) ? (nbit - j): 8;
1607 		bitmap[0] <<= s;
1608 		for (i = 1; i < wsize; i++) {
1609 			over = (bitmap[i] >> (8 - s));
1610 			bitmap[i] <<= s;
1611 			bitmap[i-1] |= over;
1612 		}
1613 	}
1614 }
1615 
1616 #ifdef INET
1617 /* Return a printable string for the IPv4 address. */
1618 static char *
1619 inet_ntoa4(struct in_addr ina)
1620 {
1621 	static char buf[4][4 * sizeof "123" + 4];
1622 	unsigned char *ucp = (unsigned char *) &ina;
1623 	static int i = 3;
1624 
1625 	/* XXX-BZ Returns static buffer. */
1626 	i = (i + 1) % 4;
1627 	sprintf(buf[i], "%d.%d.%d.%d", ucp[0] & 0xff, ucp[1] & 0xff,
1628 	    ucp[2] & 0xff, ucp[3] & 0xff);
1629 	return (buf[i]);
1630 }
1631 #endif
1632 
1633 /* Return a printable string for the address. */
1634 char *
1635 ipsec_address(union sockaddr_union* sa)
1636 {
1637 #ifdef INET6
1638 	char ip6buf[INET6_ADDRSTRLEN];
1639 #endif
1640 
1641 	switch (sa->sa.sa_family) {
1642 #ifdef INET
1643 	case AF_INET:
1644 		return (inet_ntoa4(sa->sin.sin_addr));
1645 #endif /* INET */
1646 #ifdef INET6
1647 	case AF_INET6:
1648 		return (ip6_sprintf(ip6buf, &sa->sin6.sin6_addr));
1649 #endif /* INET6 */
1650 	default:
1651 		return ("(unknown address family)");
1652 	}
1653 }
1654 
1655 const char *
1656 ipsec_logsastr(struct secasvar *sav)
1657 {
1658 	static char buf[256];
1659 	char *p;
1660 	struct secasindex *saidx = &sav->sah->saidx;
1661 
1662 	IPSEC_ASSERT(saidx->src.sa.sa_family == saidx->dst.sa.sa_family,
1663 		("address family mismatch"));
1664 
1665 	p = buf;
1666 	snprintf(buf, sizeof(buf), "SA(SPI=%u ", (u_int32_t)ntohl(sav->spi));
1667 	while (p && *p)
1668 		p++;
1669 	/* NB: only use ipsec_address on one address at a time. */
1670 	snprintf(p, sizeof (buf) - (p - buf), "src=%s ",
1671 		ipsec_address(&saidx->src));
1672 	while (p && *p)
1673 		p++;
1674 	snprintf(p, sizeof (buf) - (p - buf), "dst=%s)",
1675 		ipsec_address(&saidx->dst));
1676 
1677 	return (buf);
1678 }
1679 
1680 void
1681 ipsec_dumpmbuf(struct mbuf *m)
1682 {
1683 	int totlen;
1684 	int i;
1685 	u_char *p;
1686 
1687 	totlen = 0;
1688 	printf("---\n");
1689 	while (m) {
1690 		p = mtod(m, u_char *);
1691 		for (i = 0; i < m->m_len; i++) {
1692 			printf("%02x ", p[i]);
1693 			totlen++;
1694 			if (totlen % 16 == 0)
1695 				printf("\n");
1696 		}
1697 		m = m->m_next;
1698 	}
1699 	if (totlen % 16 != 0)
1700 		printf("\n");
1701 	printf("---\n");
1702 }
1703 
1704 static void
1705 ipsec_init(const void *unused __unused)
1706 {
1707 
1708 	SECPOLICY_LOCK_INIT(&V_ip4_def_policy);
1709 	V_ip4_def_policy.refcnt = 1;			/* NB: disallow free. */
1710 }
1711 VNET_SYSINIT(ipsec_init, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, ipsec_init,
1712     NULL);
1713 
1714 
1715 /* XXX This stuff doesn't belong here... */
1716 
1717 static	struct xformsw* xforms = NULL;
1718 
1719 /*
1720  * Register a transform; typically at system startup.
1721  */
1722 void
1723 xform_register(struct xformsw* xsp)
1724 {
1725 
1726 	xsp->xf_next = xforms;
1727 	xforms = xsp;
1728 }
1729 
1730 /*
1731  * Initialize transform support in an sav.
1732  */
1733 int
1734 xform_init(struct secasvar *sav, int xftype)
1735 {
1736 	struct xformsw *xsp;
1737 
1738 	if (sav->tdb_xform != NULL)	/* Previously initialized. */
1739 		return (0);
1740 	for (xsp = xforms; xsp; xsp = xsp->xf_next)
1741 		if (xsp->xf_type == xftype)
1742 			return ((*xsp->xf_init)(sav, xsp));
1743 	return (EINVAL);
1744 }
1745