xref: /freebsd/sys/netipsec/ipsec.c (revision 9162f64b58d01ec01481d60b6cdc06ffd8e8c7fc)
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 #include <sys/vimage.h>
57 
58 #include <net/if.h>
59 #include <net/route.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 #ifndef VIMAGE
101 #ifndef VIMAGE_GLOBALS
102 struct vnet_ipsec vnet_ipsec_0;
103 #endif
104 #endif
105 
106 #ifdef VIMAGE_GLOBALS
107 /* NB: name changed so netstat doesn't use it. */
108 struct ipsecstat ipsec4stat;
109 struct secpolicy ip4_def_policy;
110 int ipsec_debug;
111 int ip4_ah_offsetmask;
112 int ip4_ipsec_dfbit;
113 int ip4_esp_trans_deflev;
114 int ip4_esp_net_deflev;
115 int ip4_ah_trans_deflev;
116 int ip4_ah_net_deflev;
117 int ip4_ipsec_ecn;
118 int ip4_esp_randpad;
119 /*
120  * Crypto support requirements:
121  *
122  *  1	require hardware support
123  * -1	require software support
124  *  0	take anything
125  */
126 int	crypto_support;
127 #endif /* VIMAGE_GLOBALS */
128 
129 SYSCTL_DECL(_net_inet_ipsec);
130 
131 /* net.inet.ipsec */
132 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DEF_POLICY,
133 	def_policy, CTLFLAG_RW, ip4_def_policy.policy,  0,
134 	"IPsec default policy.");
135 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV,
136 	esp_trans_deflev, CTLFLAG_RW, ip4_esp_trans_deflev,	0,
137 	"Default ESP transport mode level");
138 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV,
139 	esp_net_deflev, CTLFLAG_RW, ip4_esp_net_deflev,	0,
140 	"Default ESP tunnel mode level.");
141 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV,
142 	ah_trans_deflev, CTLFLAG_RW, ip4_ah_trans_deflev,	0,
143 	"AH transfer mode default level.");
144 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV,
145 	ah_net_deflev, CTLFLAG_RW, ip4_ah_net_deflev,	0,
146 	"AH tunnel mode default level.");
147 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_AH_CLEARTOS,
148 	ah_cleartos, CTLFLAG_RW,	ah_cleartos,	0,
149 	"If set clear type-of-service field when doing AH computation.");
150 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_AH_OFFSETMASK,
151 	ah_offsetmask, CTLFLAG_RW,	ip4_ah_offsetmask,	0,
152 	"If not set clear offset field mask when doing AH computation.");
153 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DFBIT,
154 	dfbit, CTLFLAG_RW,	ip4_ipsec_dfbit,	0,
155 	"Do not fragment bit on encap.");
156 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_ECN,
157 	ecn, CTLFLAG_RW,	ip4_ipsec_ecn,	0,
158 	"Explicit Congestion Notification handling.");
159 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DEBUG,
160 	debug, CTLFLAG_RW,	ipsec_debug,	0,
161 	"Enable IPsec debugging output when set.");
162 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, OID_AUTO,
163 	crypto_support,	CTLFLAG_RW,	crypto_support,0,
164 	"Crypto driver selection.");
165 SYSCTL_V_STRUCT(V_NET, vnet_ipsec, _net_inet_ipsec, OID_AUTO,
166 	ipsecstats,	CTLFLAG_RD,	ipsec4stat, ipsecstat,
167 	"IPsec IPv4 statistics.");
168 
169 #ifdef REGRESSION
170 #ifdef VIMAGE_GLOBALS
171 int ipsec_replay;
172 int ipsec_integrity;
173 #endif
174 /*
175  * When set to 1, IPsec will send packets with the same sequence number.
176  * This allows to verify if the other side has proper replay attacks detection.
177  */
178 SYSCTL_V_INT(V_NET, vnet_ipsec,_net_inet_ipsec, OID_AUTO, test_replay,
179 	CTLFLAG_RW, ipsec_replay, 0, "Emulate replay attack");
180 /*
181  * When set 1, IPsec will send packets with corrupted HMAC.
182  * This allows to verify if the other side properly detects modified packets.
183  */
184 SYSCTL_V_INT(V_NET, vnet_ipsec,_net_inet_ipsec, OID_AUTO, test_integrity,
185 	CTLFLAG_RW, ipsec_integrity, 0, "Emulate man-in-the-middle attack");
186 #endif
187 
188 #ifdef INET6
189 #ifdef VIMAGE_GLOBALS
190 struct ipsecstat ipsec6stat;
191 int ip6_esp_trans_deflev;
192 int ip6_esp_net_deflev;
193 int ip6_ah_trans_deflev;
194 int ip6_ah_net_deflev;
195 int ip6_ipsec_ecn;
196 #endif
197 
198 SYSCTL_DECL(_net_inet6_ipsec6);
199 
200 /* net.inet6.ipsec6 */
201 #ifdef COMPAT_KAME
202 SYSCTL_OID(_net_inet6_ipsec6, IPSECCTL_STATS, stats, CTLFLAG_RD,
203     0, 0, compat_ipsecstats_sysctl, "S", "IPsec IPv6 statistics.");
204 #endif /* COMPAT_KAME */
205 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_DEF_POLICY,
206 	def_policy, CTLFLAG_RW,	ip4_def_policy.policy,	0,
207 	"IPsec default policy.");
208 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV,
209 	esp_trans_deflev, CTLFLAG_RW, ip6_esp_trans_deflev,	0,
210 	"Default ESP transport mode level.");
211 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV,
212 	esp_net_deflev, CTLFLAG_RW, ip6_esp_net_deflev,	0,
213 	"Default ESP tunnel mode level.");
214 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV,
215 	ah_trans_deflev, CTLFLAG_RW, ip6_ah_trans_deflev,	0,
216 	"AH transfer mode default level.");
217 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV,
218 	ah_net_deflev, CTLFLAG_RW, ip6_ah_net_deflev,	0,
219 	"AH tunnel mode default level.");
220 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_ECN,
221 	ecn, CTLFLAG_RW, ip6_ipsec_ecn,	0,
222 	"Explicit Congestion Notification handling.");
223 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_DEBUG,
224 	debug, CTLFLAG_RW,	ipsec_debug,	0,
225 	"Enable IPsec debugging output when set.");
226 SYSCTL_V_STRUCT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_STATS,
227 	ipsecstats, CTLFLAG_RD, ipsec6stat, ipsecstat,
228 	"IPsec IPv6 statistics.");
229 #endif /* INET6 */
230 
231 static int ipsec4_setspidx_inpcb __P((struct mbuf *, struct inpcb *));
232 #ifdef INET6
233 static int ipsec6_setspidx_inpcb __P((struct mbuf *, struct inpcb *));
234 #endif
235 static int ipsec_setspidx __P((struct mbuf *, struct secpolicyindex *, int));
236 static void ipsec4_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
237 static int ipsec4_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
238 #ifdef INET6
239 static void ipsec6_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
240 static int ipsec6_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
241 #endif
242 static void ipsec_delpcbpolicy __P((struct inpcbpolicy *));
243 static struct secpolicy *ipsec_deepcopy_policy __P((struct secpolicy *src));
244 static int ipsec_set_policy __P((struct secpolicy **pcb_sp,
245 	int optname, caddr_t request, size_t len, struct ucred *cred));
246 static int ipsec_get_policy __P((struct secpolicy *pcb_sp, struct mbuf **mp));
247 static void vshiftl __P((unsigned char *, int, int));
248 static size_t ipsec_hdrsiz __P((struct secpolicy *));
249 
250 MALLOC_DEFINE(M_IPSEC_INPCB, "inpcbpolicy", "inpcb-resident ipsec policy");
251 
252 void
253 ipsec_init(void)
254 {
255 	INIT_VNET_IPSEC(curvnet);
256 
257 #ifdef IPSEC_DEBUG
258 	V_ipsec_debug = 1;
259 #else
260 	V_ipsec_debug = 0;
261 #endif
262 
263 	V_ip4_ah_offsetmask = 0;	/* maybe IP_DF? */
264 	V_ip4_ipsec_dfbit = 0;	/* DF bit on encap. 0: clear 1: set 2: copy */
265 	V_ip4_esp_trans_deflev = IPSEC_LEVEL_USE;
266 	V_ip4_esp_net_deflev = IPSEC_LEVEL_USE;
267 	V_ip4_ah_trans_deflev = IPSEC_LEVEL_USE;
268 	V_ip4_ah_net_deflev = IPSEC_LEVEL_USE;
269 	V_ip4_ipsec_ecn = 0;	/* ECN ignore(-1)/forbidden(0)/allowed(1) */
270 	V_ip4_esp_randpad = -1;
271 
272 	V_crypto_support = CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE;
273 
274 #ifdef REGRESSION
275 	V_ipsec_replay = 0;
276 	V_ipsec_integrity = 0;
277 #endif
278 
279 #ifdef INET6
280 	V_ip6_esp_trans_deflev = IPSEC_LEVEL_USE;
281 	V_ip6_esp_net_deflev = IPSEC_LEVEL_USE;
282 	V_ip6_ah_trans_deflev = IPSEC_LEVEL_USE;
283 	V_ip6_ah_net_deflev = IPSEC_LEVEL_USE;
284 	V_ip6_ipsec_ecn = 0;	/* ECN ignore(-1)/forbidden(0)/allowed(1) */
285 #endif
286 }
287 
288 /*
289  * Return a held reference to the default SP.
290  */
291 static struct secpolicy *
292 key_allocsp_default(const char* where, int tag)
293 {
294 	INIT_VNET_IPSEC(curvnet);
295 	struct secpolicy *sp;
296 
297 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
298 		printf("DP key_allocsp_default from %s:%u\n", where, tag));
299 
300 	sp = &V_ip4_def_policy;
301 	if (sp->policy != IPSEC_POLICY_DISCARD &&
302 	    sp->policy != IPSEC_POLICY_NONE) {
303 		ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n",
304 		    sp->policy, IPSEC_POLICY_NONE));
305 		sp->policy = IPSEC_POLICY_NONE;
306 	}
307 	key_addref(sp);
308 
309 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
310 		printf("DP key_allocsp_default returns SP:%p (%u)\n",
311 			sp, sp->refcnt));
312 	return (sp);
313 }
314 #define	KEY_ALLOCSP_DEFAULT() \
315 	key_allocsp_default(__FILE__, __LINE__)
316 
317 /*
318  * For OUTBOUND packet having a socket. Searching SPD for packet,
319  * and return a pointer to SP.
320  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
321  *		0	: bypass
322  *		EACCES	: discard packet.
323  *		ENOENT	: ipsec_acquire() in progress, maybe.
324  *		others	: error occured.
325  *	others:	a pointer to SP
326  *
327  * NOTE: IPv6 mapped adddress concern is implemented here.
328  */
329 struct secpolicy *
330 ipsec_getpolicy(struct tdb_ident *tdbi, u_int dir)
331 {
332 	struct secpolicy *sp;
333 
334 	IPSEC_ASSERT(tdbi != NULL, ("null tdbi"));
335 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
336 		("invalid direction %u", dir));
337 
338 	sp = KEY_ALLOCSP2(tdbi->spi, &tdbi->dst, tdbi->proto, dir);
339 	if (sp == NULL)			/*XXX????*/
340 		sp = KEY_ALLOCSP_DEFAULT();
341 	IPSEC_ASSERT(sp != NULL, ("null SP"));
342 	return (sp);
343 }
344 
345 /*
346  * For OUTBOUND packet having a socket. Searching SPD for packet,
347  * and return a pointer to SP.
348  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
349  *		0	: bypass
350  *		EACCES	: discard packet.
351  *		ENOENT	: ipsec_acquire() in progress, maybe.
352  *		others	: error occured.
353  *	others:	a pointer to SP
354  *
355  * NOTE: IPv6 mapped adddress concern is implemented here.
356  */
357 static struct secpolicy *
358 ipsec_getpolicybysock(struct mbuf *m, u_int dir, struct inpcb *inp, int *error)
359 {
360 	INIT_VNET_IPSEC(curvnet);
361 	struct inpcbpolicy *pcbsp = NULL;
362 	struct secpolicy *currsp = NULL;	/* Policy on socket. */
363 	struct secpolicy *sp;
364 
365 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
366 	IPSEC_ASSERT(inp != NULL, ("null inpcb"));
367 	IPSEC_ASSERT(error != NULL, ("null error"));
368 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
369 		("invalid direction %u", dir));
370 
371 	/* Set spidx in pcb. */
372 	if (inp->inp_vflag & INP_IPV6PROTO) {
373 #ifdef INET6
374 		*error = ipsec6_setspidx_inpcb(m, inp);
375 		pcbsp = inp->inp_sp;
376 #else
377 		*error = EINVAL;		/* Should not happen. */
378 #endif
379 	} else {
380 		*error = ipsec4_setspidx_inpcb(m, inp);
381 		pcbsp = inp->inp_sp;
382 	}
383 	if (*error)
384 		return (NULL);
385 
386 	IPSEC_ASSERT(pcbsp != NULL, ("null pcbsp"));
387 	switch (dir) {
388 	case IPSEC_DIR_INBOUND:
389 		currsp = pcbsp->sp_in;
390 		break;
391 	case IPSEC_DIR_OUTBOUND:
392 		currsp = pcbsp->sp_out;
393 		break;
394 	}
395 	IPSEC_ASSERT(currsp != NULL, ("null currsp"));
396 
397 	if (pcbsp->priv) {			/* When privilieged socket. */
398 		switch (currsp->policy) {
399 		case IPSEC_POLICY_BYPASS:
400 		case IPSEC_POLICY_IPSEC:
401 			key_addref(currsp);
402 			sp = currsp;
403 			break;
404 
405 		case IPSEC_POLICY_ENTRUST:
406 			/* Look for a policy in SPD. */
407 			sp = KEY_ALLOCSP(&currsp->spidx, dir);
408 			if (sp == NULL)		/* No SP found. */
409 				sp = KEY_ALLOCSP_DEFAULT();
410 			break;
411 
412 		default:
413 			ipseclog((LOG_ERR, "%s: Invalid policy for PCB %d\n",
414 				__func__, currsp->policy));
415 			*error = EINVAL;
416 			return (NULL);
417 		}
418 	} else {				/* Unpriv, SPD has policy. */
419 		sp = KEY_ALLOCSP(&currsp->spidx, dir);
420 		if (sp == NULL) {		/* No SP found. */
421 			switch (currsp->policy) {
422 			case IPSEC_POLICY_BYPASS:
423 				ipseclog((LOG_ERR, "%s: Illegal policy for "
424 					"non-priviliged defined %d\n",
425 					__func__, currsp->policy));
426 				*error = EINVAL;
427 				return (NULL);
428 
429 			case IPSEC_POLICY_ENTRUST:
430 				sp = KEY_ALLOCSP_DEFAULT();
431 				break;
432 
433 			case IPSEC_POLICY_IPSEC:
434 				key_addref(currsp);
435 				sp = currsp;
436 				break;
437 
438 			default:
439 				ipseclog((LOG_ERR, "%s: Invalid policy for "
440 					"PCB %d\n", __func__, currsp->policy));
441 				*error = EINVAL;
442 				return (NULL);
443 			}
444 		}
445 	}
446 	IPSEC_ASSERT(sp != NULL,
447 		("null SP (priv %u policy %u", pcbsp->priv, currsp->policy));
448 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
449 		printf("DP %s (priv %u policy %u) allocate SP:%p (refcnt %u)\n",
450 			__func__, pcbsp->priv, currsp->policy, sp, sp->refcnt));
451 	return (sp);
452 }
453 
454 /*
455  * For FORWADING packet or OUTBOUND without a socket. Searching SPD for packet,
456  * and return a pointer to SP.
457  * OUT:	positive: a pointer to the entry for security policy leaf matched.
458  *	NULL:	no apropreate SP found, the following value is set to error.
459  *		0	: bypass
460  *		EACCES	: discard packet.
461  *		ENOENT	: ipsec_acquire() in progress, maybe.
462  *		others	: error occured.
463  */
464 struct secpolicy *
465 ipsec_getpolicybyaddr(struct mbuf *m, u_int dir, int flag, int *error)
466 {
467 	INIT_VNET_IPSEC(curvnet);
468 	struct secpolicyindex spidx;
469 	struct secpolicy *sp;
470 
471 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
472 	IPSEC_ASSERT(error != NULL, ("null error"));
473 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
474 		("invalid direction %u", dir));
475 
476 	sp = NULL;
477 	if (key_havesp(dir)) {
478 		/* Make an index to look for a policy. */
479 		*error = ipsec_setspidx(m, &spidx,
480 					(flag & IP_FORWARDING) ? 0 : 1);
481 		if (*error != 0) {
482 			DPRINTF(("%s: setpidx failed, dir %u flag %u\n",
483 				__func__, dir, flag));
484 			return (NULL);
485 		}
486 		spidx.dir = dir;
487 
488 		sp = KEY_ALLOCSP(&spidx, dir);
489 	}
490 	if (sp == NULL)			/* No SP found, use system default. */
491 		sp = KEY_ALLOCSP_DEFAULT();
492 	IPSEC_ASSERT(sp != NULL, ("null SP"));
493 	return (sp);
494 }
495 
496 struct secpolicy *
497 ipsec4_checkpolicy(struct mbuf *m, u_int dir, u_int flag, int *error,
498     struct inpcb *inp)
499 {
500 	INIT_VNET_IPSEC(curvnet);
501 	struct secpolicy *sp;
502 
503 	*error = 0;
504 	if (inp == NULL)
505 		sp = ipsec_getpolicybyaddr(m, dir, flag, error);
506 	else
507 		sp = ipsec_getpolicybysock(m, dir, inp, error);
508 	if (sp == NULL) {
509 		IPSEC_ASSERT(*error != 0, ("getpolicy failed w/o error"));
510 		V_ipsec4stat.ips_out_inval++;
511 		return (NULL);
512 	}
513 	IPSEC_ASSERT(*error == 0, ("sp w/ error set to %u", *error));
514 	switch (sp->policy) {
515 	case IPSEC_POLICY_ENTRUST:
516 	default:
517 		printf("%s: invalid policy %u\n", __func__, sp->policy);
518 		/* FALLTHROUGH */
519 	case IPSEC_POLICY_DISCARD:
520 		V_ipsec4stat.ips_out_polvio++;
521 		*error = -EINVAL;	/* Packet is discarded by caller. */
522 		break;
523 	case IPSEC_POLICY_BYPASS:
524 	case IPSEC_POLICY_NONE:
525 		KEY_FREESP(&sp);
526 		sp = NULL;		/* NB: force NULL result. */
527 		break;
528 	case IPSEC_POLICY_IPSEC:
529 		if (sp->req == NULL)	/* Acquire a SA. */
530 			*error = key_spdacquire(sp);
531 		break;
532 	}
533 	if (*error != 0) {
534 		KEY_FREESP(&sp);
535 		sp = NULL;
536 	}
537 	return (sp);
538 }
539 
540 static int
541 ipsec4_setspidx_inpcb(struct mbuf *m, struct inpcb *inp)
542 {
543 	int error;
544 
545 	IPSEC_ASSERT(inp != NULL, ("null inp"));
546 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
547 	IPSEC_ASSERT(inp->inp_sp->sp_out != NULL && inp->inp_sp->sp_in != NULL,
548 		("null sp_in || sp_out"));
549 
550 	error = ipsec_setspidx(m, &inp->inp_sp->sp_in->spidx, 1);
551 	if (error == 0) {
552 		inp->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
553 		inp->inp_sp->sp_out->spidx = inp->inp_sp->sp_in->spidx;
554 		inp->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
555 	} else {
556 		bzero(&inp->inp_sp->sp_in->spidx,
557 			sizeof (inp->inp_sp->sp_in->spidx));
558 		bzero(&inp->inp_sp->sp_out->spidx,
559 			sizeof (inp->inp_sp->sp_in->spidx));
560 	}
561 	return (error);
562 }
563 
564 #ifdef INET6
565 static int
566 ipsec6_setspidx_inpcb(struct mbuf *m, struct inpcb *inp)
567 {
568 	int error;
569 
570 	IPSEC_ASSERT(inp != NULL, ("null inp"));
571 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
572 	IPSEC_ASSERT(inp->inp_sp->sp_out != NULL && inp->inp_sp->sp_in != NULL,
573 		("null sp_in || sp_out"));
574 
575 	error = ipsec_setspidx(m, &inp->inp_sp->sp_in->spidx, 1);
576 	if (error == 0) {
577 		inp->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
578 		inp->inp_sp->sp_out->spidx = inp->inp_sp->sp_in->spidx;
579 		inp->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
580 	} else {
581 		bzero(&inp->inp_sp->sp_in->spidx,
582 		    sizeof(inp->inp_sp->sp_in->spidx));
583 		bzero(&inp->inp_sp->sp_out->spidx,
584 		    sizeof(inp->inp_sp->sp_in->spidx));
585 	}
586 
587 	return (error);
588 }
589 #endif
590 
591 /*
592  * Configure security policy index (src/dst/proto/sport/dport)
593  * by looking at the content of mbuf.
594  * The caller is responsible for error recovery (like clearing up spidx).
595  */
596 static int
597 ipsec_setspidx(struct mbuf *m, struct secpolicyindex *spidx, int needport)
598 {
599 	INIT_VNET_IPSEC(curvnet);
600 	struct ip *ip = NULL;
601 	struct ip ipbuf;
602 	u_int v;
603 	struct mbuf *n;
604 	int len;
605 	int error;
606 
607 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
608 
609 	/*
610 	 * Validate m->m_pkthdr.len.  We see incorrect length if we
611 	 * mistakenly call this function with inconsistent mbuf chain
612 	 * (like 4.4BSD tcp/udp processing).  XXX Should we panic here?
613 	 */
614 	len = 0;
615 	for (n = m; n; n = n->m_next)
616 		len += n->m_len;
617 	if (m->m_pkthdr.len != len) {
618 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
619 			printf("%s: pkthdr len(%d) mismatch (%d), ignored.\n",
620 				__func__, len, m->m_pkthdr.len));
621 		return (EINVAL);
622 	}
623 
624 	if (m->m_pkthdr.len < sizeof(struct ip)) {
625 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
626 			printf("%s: pkthdr len(%d) too small (v4), ignored.\n",
627 			    __func__, m->m_pkthdr.len));
628 		return (EINVAL);
629 	}
630 
631 	if (m->m_len >= sizeof(*ip))
632 		ip = mtod(m, struct ip *);
633 	else {
634 		m_copydata(m, 0, sizeof(ipbuf), (caddr_t)&ipbuf);
635 		ip = &ipbuf;
636 	}
637 #ifdef _IP_VHL
638 	v = _IP_VHL_V(ip->ip_vhl);
639 #else
640 	v = ip->ip_v;
641 #endif
642 	switch (v) {
643 	case 4:
644 		error = ipsec4_setspidx_ipaddr(m, spidx);
645 		if (error)
646 			return (error);
647 		ipsec4_get_ulp(m, spidx, needport);
648 		return (0);
649 #ifdef INET6
650 	case 6:
651 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) {
652 			KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
653 				printf("%s: pkthdr len(%d) too small (v6), "
654 				"ignored\n", __func__, m->m_pkthdr.len));
655 			return (EINVAL);
656 		}
657 		error = ipsec6_setspidx_ipaddr(m, spidx);
658 		if (error)
659 			return (error);
660 		ipsec6_get_ulp(m, spidx, needport);
661 		return (0);
662 #endif
663 	default:
664 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
665 			printf("%s: " "unknown IP version %u, ignored.\n",
666 				__func__, v));
667 		return (EINVAL);
668 	}
669 }
670 
671 static void
672 ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
673 {
674 	u_int8_t nxt;
675 	int off;
676 
677 	/* Sanity check. */
678 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
679 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),("packet too short"));
680 
681 	/* NB: ip_input() flips it into host endian. XXX Need more checking. */
682 	if (m->m_len < sizeof (struct ip)) {
683 		struct ip *ip = mtod(m, struct ip *);
684 		if (ip->ip_off & (IP_MF | IP_OFFMASK))
685 			goto done;
686 #ifdef _IP_VHL
687 		off = _IP_VHL_HL(ip->ip_vhl) << 2;
688 #else
689 		off = ip->ip_hl << 2;
690 #endif
691 		nxt = ip->ip_p;
692 	} else {
693 		struct ip ih;
694 
695 		m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih);
696 		if (ih.ip_off & (IP_MF | IP_OFFMASK))
697 			goto done;
698 #ifdef _IP_VHL
699 		off = _IP_VHL_HL(ih.ip_vhl) << 2;
700 #else
701 		off = ih.ip_hl << 2;
702 #endif
703 		nxt = ih.ip_p;
704 	}
705 
706 	while (off < m->m_pkthdr.len) {
707 		struct ip6_ext ip6e;
708 		struct tcphdr th;
709 		struct udphdr uh;
710 
711 		switch (nxt) {
712 		case IPPROTO_TCP:
713 			spidx->ul_proto = nxt;
714 			if (!needport)
715 				goto done_proto;
716 			if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
717 				goto done;
718 			m_copydata(m, off, sizeof (th), (caddr_t) &th);
719 			spidx->src.sin.sin_port = th.th_sport;
720 			spidx->dst.sin.sin_port = th.th_dport;
721 			return;
722 		case IPPROTO_UDP:
723 			spidx->ul_proto = nxt;
724 			if (!needport)
725 				goto done_proto;
726 			if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
727 				goto done;
728 			m_copydata(m, off, sizeof (uh), (caddr_t) &uh);
729 			spidx->src.sin.sin_port = uh.uh_sport;
730 			spidx->dst.sin.sin_port = uh.uh_dport;
731 			return;
732 		case IPPROTO_AH:
733 			if (off + sizeof(ip6e) > m->m_pkthdr.len)
734 				goto done;
735 			/* XXX Sigh, this works but is totally bogus. */
736 			m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e);
737 			off += (ip6e.ip6e_len + 2) << 2;
738 			nxt = ip6e.ip6e_nxt;
739 			break;
740 		case IPPROTO_ICMP:
741 		default:
742 			/* XXX Intermediate headers??? */
743 			spidx->ul_proto = nxt;
744 			goto done_proto;
745 		}
746 	}
747 done:
748 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
749 done_proto:
750 	spidx->src.sin.sin_port = IPSEC_PORT_ANY;
751 	spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
752 }
753 
754 /* Assumes that m is sane. */
755 static int
756 ipsec4_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
757 {
758 	static const struct sockaddr_in template = {
759 		sizeof (struct sockaddr_in),
760 		AF_INET,
761 		0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 }
762 	};
763 
764 	spidx->src.sin = template;
765 	spidx->dst.sin = template;
766 
767 	if (m->m_len < sizeof (struct ip)) {
768 		m_copydata(m, offsetof(struct ip, ip_src),
769 			   sizeof (struct  in_addr),
770 			   (caddr_t) &spidx->src.sin.sin_addr);
771 		m_copydata(m, offsetof(struct ip, ip_dst),
772 			   sizeof (struct  in_addr),
773 			   (caddr_t) &spidx->dst.sin.sin_addr);
774 	} else {
775 		struct ip *ip = mtod(m, struct ip *);
776 		spidx->src.sin.sin_addr = ip->ip_src;
777 		spidx->dst.sin.sin_addr = ip->ip_dst;
778 	}
779 
780 	spidx->prefs = sizeof(struct in_addr) << 3;
781 	spidx->prefd = sizeof(struct in_addr) << 3;
782 
783 	return (0);
784 }
785 
786 #ifdef INET6
787 static void
788 ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
789 {
790 	INIT_VNET_IPSEC(curvnet);
791 	int off, nxt;
792 	struct tcphdr th;
793 	struct udphdr uh;
794 	struct icmp6_hdr ih;
795 
796 	/* Sanity check. */
797 	if (m == NULL)
798 		panic("%s: NULL pointer was passed.\n", __func__);
799 
800 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
801 		printf("%s:\n", __func__); kdebug_mbuf(m));
802 
803 	/* Set default. */
804 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
805 	((struct sockaddr_in6 *)&spidx->src)->sin6_port = IPSEC_PORT_ANY;
806 	((struct sockaddr_in6 *)&spidx->dst)->sin6_port = IPSEC_PORT_ANY;
807 
808 	nxt = -1;
809 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
810 	if (off < 0 || m->m_pkthdr.len < off)
811 		return;
812 
813 	switch (nxt) {
814 	case IPPROTO_TCP:
815 		spidx->ul_proto = nxt;
816 		if (!needport)
817 			break;
818 		if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
819 			break;
820 		m_copydata(m, off, sizeof(th), (caddr_t)&th);
821 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = th.th_sport;
822 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = th.th_dport;
823 		break;
824 	case IPPROTO_UDP:
825 		spidx->ul_proto = nxt;
826 		if (!needport)
827 			break;
828 		if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
829 			break;
830 		m_copydata(m, off, sizeof(uh), (caddr_t)&uh);
831 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = uh.uh_sport;
832 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = uh.uh_dport;
833 		break;
834 	case IPPROTO_ICMPV6:
835 		spidx->ul_proto = nxt;
836 		if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len)
837 			break;
838 		m_copydata(m, off, sizeof(ih), (caddr_t)&ih);
839 		((struct sockaddr_in6 *)&spidx->src)->sin6_port =
840 		    htons((uint16_t)ih.icmp6_type);
841 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port =
842 		    htons((uint16_t)ih.icmp6_code);
843 		break;
844 	default:
845 		/* XXX Intermediate headers??? */
846 		spidx->ul_proto = nxt;
847 		break;
848 	}
849 }
850 
851 /* Assumes that m is sane. */
852 static int
853 ipsec6_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
854 {
855 	struct ip6_hdr *ip6 = NULL;
856 	struct ip6_hdr ip6buf;
857 	struct sockaddr_in6 *sin6;
858 
859 	if (m->m_len >= sizeof(*ip6))
860 		ip6 = mtod(m, struct ip6_hdr *);
861 	else {
862 		m_copydata(m, 0, sizeof(ip6buf), (caddr_t)&ip6buf);
863 		ip6 = &ip6buf;
864 	}
865 
866 	sin6 = (struct sockaddr_in6 *)&spidx->src;
867 	bzero(sin6, sizeof(*sin6));
868 	sin6->sin6_family = AF_INET6;
869 	sin6->sin6_len = sizeof(struct sockaddr_in6);
870 	bcopy(&ip6->ip6_src, &sin6->sin6_addr, sizeof(ip6->ip6_src));
871 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
872 		sin6->sin6_addr.s6_addr16[1] = 0;
873 		sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]);
874 	}
875 	spidx->prefs = sizeof(struct in6_addr) << 3;
876 
877 	sin6 = (struct sockaddr_in6 *)&spidx->dst;
878 	bzero(sin6, sizeof(*sin6));
879 	sin6->sin6_family = AF_INET6;
880 	sin6->sin6_len = sizeof(struct sockaddr_in6);
881 	bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(ip6->ip6_dst));
882 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
883 		sin6->sin6_addr.s6_addr16[1] = 0;
884 		sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]);
885 	}
886 	spidx->prefd = sizeof(struct in6_addr) << 3;
887 
888 	return (0);
889 }
890 #endif
891 
892 static void
893 ipsec_delpcbpolicy(struct inpcbpolicy *p)
894 {
895 
896 	free(p, M_IPSEC_INPCB);
897 }
898 
899 /* Initialize policy in PCB. */
900 int
901 ipsec_init_policy(struct socket *so, struct inpcbpolicy **pcb_sp)
902 {
903 	INIT_VNET_IPSEC(curvnet);
904 	struct inpcbpolicy *new;
905 
906 	/* Sanity check. */
907 	if (so == NULL || pcb_sp == NULL)
908 		panic("%s: NULL pointer was passed.\n", __func__);
909 
910 	new = (struct inpcbpolicy *) malloc(sizeof(struct inpcbpolicy),
911 					    M_IPSEC_INPCB, M_NOWAIT|M_ZERO);
912 	if (new == NULL) {
913 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
914 		return (ENOBUFS);
915 	}
916 
917 	new->priv = IPSEC_IS_PRIVILEGED_SO(so);
918 
919 	if ((new->sp_in = KEY_NEWSP()) == NULL) {
920 		ipsec_delpcbpolicy(new);
921 		return (ENOBUFS);
922 	}
923 	new->sp_in->state = IPSEC_SPSTATE_ALIVE;
924 	new->sp_in->policy = IPSEC_POLICY_ENTRUST;
925 
926 	if ((new->sp_out = KEY_NEWSP()) == NULL) {
927 		KEY_FREESP(&new->sp_in);
928 		ipsec_delpcbpolicy(new);
929 		return (ENOBUFS);
930 	}
931 	new->sp_out->state = IPSEC_SPSTATE_ALIVE;
932 	new->sp_out->policy = IPSEC_POLICY_ENTRUST;
933 
934 	*pcb_sp = new;
935 
936 	return (0);
937 }
938 
939 /* Copy old IPsec policy into new. */
940 int
941 ipsec_copy_policy(struct inpcbpolicy *old, struct inpcbpolicy *new)
942 {
943 	struct secpolicy *sp;
944 
945 	sp = ipsec_deepcopy_policy(old->sp_in);
946 	if (sp) {
947 		KEY_FREESP(&new->sp_in);
948 		new->sp_in = sp;
949 	} else
950 		return (ENOBUFS);
951 
952 	sp = ipsec_deepcopy_policy(old->sp_out);
953 	if (sp) {
954 		KEY_FREESP(&new->sp_out);
955 		new->sp_out = sp;
956 	} else
957 		return (ENOBUFS);
958 
959 	new->priv = old->priv;
960 
961 	return (0);
962 }
963 
964 struct ipsecrequest *
965 ipsec_newisr(void)
966 {
967 	struct ipsecrequest *p;
968 
969 	p = malloc(sizeof(struct ipsecrequest), M_IPSEC_SR, M_NOWAIT|M_ZERO);
970 	if (p != NULL)
971 		IPSECREQUEST_LOCK_INIT(p);
972 	return (p);
973 }
974 
975 void
976 ipsec_delisr(struct ipsecrequest *p)
977 {
978 
979 	IPSECREQUEST_LOCK_DESTROY(p);
980 	free(p, M_IPSEC_SR);
981 }
982 
983 /* Deep-copy a policy in PCB. */
984 static struct secpolicy *
985 ipsec_deepcopy_policy(struct secpolicy *src)
986 {
987 	struct ipsecrequest *newchain = NULL;
988 	struct ipsecrequest *p;
989 	struct ipsecrequest **q;
990 	struct ipsecrequest *r;
991 	struct secpolicy *dst;
992 
993 	if (src == NULL)
994 		return (NULL);
995 	dst = KEY_NEWSP();
996 	if (dst == NULL)
997 		return (NULL);
998 
999 	/*
1000 	 * Deep-copy IPsec request chain.  This is required since struct
1001 	 * ipsecrequest is not reference counted.
1002 	 */
1003 	q = &newchain;
1004 	for (p = src->req; p; p = p->next) {
1005 		*q = ipsec_newisr();
1006 		if (*q == NULL)
1007 			goto fail;
1008 		(*q)->saidx.proto = p->saidx.proto;
1009 		(*q)->saidx.mode = p->saidx.mode;
1010 		(*q)->level = p->level;
1011 		(*q)->saidx.reqid = p->saidx.reqid;
1012 
1013 		bcopy(&p->saidx.src, &(*q)->saidx.src, sizeof((*q)->saidx.src));
1014 		bcopy(&p->saidx.dst, &(*q)->saidx.dst, sizeof((*q)->saidx.dst));
1015 
1016 		(*q)->sp = dst;
1017 
1018 		q = &((*q)->next);
1019 	}
1020 
1021 	dst->req = newchain;
1022 	dst->state = src->state;
1023 	dst->policy = src->policy;
1024 	/* Do not touch the refcnt fields. */
1025 
1026 	return (dst);
1027 
1028 fail:
1029 	for (p = newchain; p; p = r) {
1030 		r = p->next;
1031 		ipsec_delisr(p);
1032 		p = NULL;
1033 	}
1034 	return (NULL);
1035 }
1036 
1037 /* Set policy and IPsec request if present. */
1038 static int
1039 ipsec_set_policy(struct secpolicy **pcb_sp, int optname, caddr_t request,
1040     size_t len, struct ucred *cred)
1041 {
1042 	INIT_VNET_IPSEC(curvnet);
1043 	struct sadb_x_policy *xpl;
1044 	struct secpolicy *newsp = NULL;
1045 	int error;
1046 
1047 	/* Sanity check. */
1048 	if (pcb_sp == NULL || *pcb_sp == NULL || request == NULL)
1049 		return (EINVAL);
1050 	if (len < sizeof(*xpl))
1051 		return (EINVAL);
1052 	xpl = (struct sadb_x_policy *)request;
1053 
1054 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1055 		printf("%s: passed policy\n", __func__);
1056 		kdebug_sadb_x_policy((struct sadb_ext *)xpl));
1057 
1058 	/* Check policy type. */
1059 	/* ipsec_set_policy() accepts IPSEC, ENTRUST and BYPASS. */
1060 	if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD
1061 	 || xpl->sadb_x_policy_type == IPSEC_POLICY_NONE)
1062 		return (EINVAL);
1063 
1064 	/* Check privileged socket. */
1065 	if (cred != NULL && xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
1066 		error = priv_check_cred(cred, PRIV_NETINET_IPSEC, 0);
1067 		if (error)
1068 			return (EACCES);
1069 	}
1070 
1071 	/* Allocating new SP entry. */
1072 	if ((newsp = key_msg2sp(xpl, len, &error)) == NULL)
1073 		return (error);
1074 
1075 	newsp->state = IPSEC_SPSTATE_ALIVE;
1076 
1077 	/* Clear old SP and set new SP. */
1078 	KEY_FREESP(pcb_sp);
1079 	*pcb_sp = newsp;
1080 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1081 		printf("%s: new policy\n", __func__);
1082 		kdebug_secpolicy(newsp));
1083 
1084 	return (0);
1085 }
1086 
1087 static int
1088 ipsec_get_policy(struct secpolicy *pcb_sp, struct mbuf **mp)
1089 {
1090 	INIT_VNET_IPSEC(curvnet);
1091 
1092 	/* Sanity check. */
1093 	if (pcb_sp == NULL || mp == NULL)
1094 		return (EINVAL);
1095 
1096 	*mp = key_sp2msg(pcb_sp);
1097 	if (!*mp) {
1098 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
1099 		return (ENOBUFS);
1100 	}
1101 
1102 	(*mp)->m_type = MT_DATA;
1103 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1104 		printf("%s:\n", __func__); kdebug_mbuf(*mp));
1105 
1106 	return (0);
1107 }
1108 
1109 int
1110 ipsec4_set_policy(struct inpcb *inp, int optname, caddr_t request,
1111     size_t len, struct ucred *cred)
1112 {
1113 	INIT_VNET_IPSEC(curvnet);
1114 	struct sadb_x_policy *xpl;
1115 	struct secpolicy **pcb_sp;
1116 
1117 	/* Sanity check. */
1118 	if (inp == NULL || request == NULL)
1119 		return (EINVAL);
1120 	if (len < sizeof(*xpl))
1121 		return (EINVAL);
1122 	xpl = (struct sadb_x_policy *)request;
1123 
1124 	/* Select direction. */
1125 	switch (xpl->sadb_x_policy_dir) {
1126 	case IPSEC_DIR_INBOUND:
1127 		pcb_sp = &inp->inp_sp->sp_in;
1128 		break;
1129 	case IPSEC_DIR_OUTBOUND:
1130 		pcb_sp = &inp->inp_sp->sp_out;
1131 		break;
1132 	default:
1133 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1134 			xpl->sadb_x_policy_dir));
1135 		return (EINVAL);
1136 	}
1137 
1138 	return (ipsec_set_policy(pcb_sp, optname, request, len, cred));
1139 }
1140 
1141 int
1142 ipsec4_get_policy(struct inpcb *inp, caddr_t request, size_t len,
1143     struct mbuf **mp)
1144 {
1145 	INIT_VNET_IPSEC(curvnet);
1146 	struct sadb_x_policy *xpl;
1147 	struct secpolicy *pcb_sp;
1148 
1149 	/* Sanity check. */
1150 	if (inp == NULL || request == NULL || mp == NULL)
1151 		return (EINVAL);
1152 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
1153 	if (len < sizeof(*xpl))
1154 		return (EINVAL);
1155 	xpl = (struct sadb_x_policy *)request;
1156 
1157 	/* Select direction. */
1158 	switch (xpl->sadb_x_policy_dir) {
1159 	case IPSEC_DIR_INBOUND:
1160 		pcb_sp = inp->inp_sp->sp_in;
1161 		break;
1162 	case IPSEC_DIR_OUTBOUND:
1163 		pcb_sp = inp->inp_sp->sp_out;
1164 		break;
1165 	default:
1166 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1167 			xpl->sadb_x_policy_dir));
1168 		return (EINVAL);
1169 	}
1170 
1171 	return (ipsec_get_policy(pcb_sp, mp));
1172 }
1173 
1174 /* Delete policy in PCB. */
1175 int
1176 ipsec_delete_pcbpolicy(struct inpcb *inp)
1177 {
1178 	IPSEC_ASSERT(inp != NULL, ("null inp"));
1179 
1180 	if (inp->inp_sp == NULL)
1181 		return (0);
1182 
1183 	if (inp->inp_sp->sp_in != NULL)
1184 		KEY_FREESP(&inp->inp_sp->sp_in);
1185 
1186 	if (inp->inp_sp->sp_out != NULL)
1187 		KEY_FREESP(&inp->inp_sp->sp_out);
1188 
1189 	ipsec_delpcbpolicy(inp->inp_sp);
1190 	inp->inp_sp = NULL;
1191 
1192 	return (0);
1193 }
1194 
1195 #ifdef INET6
1196 int
1197 ipsec6_set_policy(struct inpcb *inp, int optname, caddr_t request,
1198     size_t len, struct ucred *cred)
1199 {
1200 	INIT_VNET_IPSEC(curvnet);
1201 	struct sadb_x_policy *xpl;
1202 	struct secpolicy **pcb_sp;
1203 
1204 	/* Sanity check. */
1205 	if (inp == NULL || request == NULL)
1206 		return (EINVAL);
1207 	if (len < sizeof(*xpl))
1208 		return (EINVAL);
1209 	xpl = (struct sadb_x_policy *)request;
1210 
1211 	/* Select direction. */
1212 	switch (xpl->sadb_x_policy_dir) {
1213 	case IPSEC_DIR_INBOUND:
1214 		pcb_sp = &inp->inp_sp->sp_in;
1215 		break;
1216 	case IPSEC_DIR_OUTBOUND:
1217 		pcb_sp = &inp->inp_sp->sp_out;
1218 		break;
1219 	default:
1220 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1221 			xpl->sadb_x_policy_dir));
1222 		return (EINVAL);
1223 	}
1224 
1225 	return (ipsec_set_policy(pcb_sp, optname, request, len, cred));
1226 }
1227 
1228 int
1229 ipsec6_get_policy(struct inpcb *inp, caddr_t request, size_t len,
1230     struct mbuf **mp)
1231 {
1232 	INIT_VNET_IPSEC(curvnet);
1233 	struct sadb_x_policy *xpl;
1234 	struct secpolicy *pcb_sp;
1235 
1236 	/* Sanity check. */
1237 	if (inp == NULL || request == NULL || mp == NULL)
1238 		return (EINVAL);
1239 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
1240 	if (len < sizeof(*xpl))
1241 		return (EINVAL);
1242 	xpl = (struct sadb_x_policy *)request;
1243 
1244 	/* Select direction. */
1245 	switch (xpl->sadb_x_policy_dir) {
1246 	case IPSEC_DIR_INBOUND:
1247 		pcb_sp = inp->inp_sp->sp_in;
1248 		break;
1249 	case IPSEC_DIR_OUTBOUND:
1250 		pcb_sp = inp->inp_sp->sp_out;
1251 		break;
1252 	default:
1253 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1254 			xpl->sadb_x_policy_dir));
1255 		return (EINVAL);
1256 	}
1257 
1258 	return (ipsec_get_policy(pcb_sp, mp));
1259 }
1260 #endif
1261 
1262 /*
1263  * Return current level.
1264  * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
1265  */
1266 u_int
1267 ipsec_get_reqlevel(struct ipsecrequest *isr)
1268 {
1269 	INIT_VNET_IPSEC(curvnet);
1270 	u_int level = 0;
1271 	u_int esp_trans_deflev, esp_net_deflev;
1272 	u_int ah_trans_deflev, ah_net_deflev;
1273 
1274 	IPSEC_ASSERT(isr != NULL && isr->sp != NULL, ("null argument"));
1275 	IPSEC_ASSERT(isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family,
1276 		("af family mismatch, src %u, dst %u",
1277 		 isr->sp->spidx.src.sa.sa_family,
1278 		 isr->sp->spidx.dst.sa.sa_family));
1279 
1280 /* XXX Note that we have ipseclog() expanded here - code sync issue. */
1281 #define IPSEC_CHECK_DEFAULT(lev) \
1282 	(((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE	      \
1283 			&& (lev) != IPSEC_LEVEL_UNIQUE)			      \
1284 		? (V_ipsec_debug						      \
1285 			? log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
1286 				(lev), IPSEC_LEVEL_REQUIRE)		      \
1287 			: 0),						      \
1288 			(lev) = IPSEC_LEVEL_REQUIRE,			      \
1289 			(lev)						      \
1290 		: (lev))
1291 
1292 	/* Set default level. */
1293 	switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) {
1294 #ifdef INET
1295 	case AF_INET:
1296 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev);
1297 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev);
1298 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev);
1299 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev);
1300 		break;
1301 #endif
1302 #ifdef INET6
1303 	case AF_INET6:
1304 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev);
1305 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev);
1306 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev);
1307 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev);
1308 		break;
1309 #endif /* INET6 */
1310 	default:
1311 		panic("%s: unknown af %u",
1312 			__func__, isr->sp->spidx.src.sa.sa_family);
1313 	}
1314 
1315 #undef IPSEC_CHECK_DEFAULT
1316 
1317 	/* Set level. */
1318 	switch (isr->level) {
1319 	case IPSEC_LEVEL_DEFAULT:
1320 		switch (isr->saidx.proto) {
1321 		case IPPROTO_ESP:
1322 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1323 				level = esp_net_deflev;
1324 			else
1325 				level = esp_trans_deflev;
1326 			break;
1327 		case IPPROTO_AH:
1328 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1329 				level = ah_net_deflev;
1330 			else
1331 				level = ah_trans_deflev;
1332 			break;
1333 		case IPPROTO_IPCOMP:
1334 			/*
1335 			 * We don't really care, as IPcomp document says that
1336 			 * we shouldn't compress small packets.
1337 			 */
1338 			level = IPSEC_LEVEL_USE;
1339 			break;
1340 		default:
1341 			panic("%s: Illegal protocol defined %u\n", __func__,
1342 				isr->saidx.proto);
1343 		}
1344 		break;
1345 
1346 	case IPSEC_LEVEL_USE:
1347 	case IPSEC_LEVEL_REQUIRE:
1348 		level = isr->level;
1349 		break;
1350 	case IPSEC_LEVEL_UNIQUE:
1351 		level = IPSEC_LEVEL_REQUIRE;
1352 		break;
1353 
1354 	default:
1355 		panic("%s: Illegal IPsec level %u\n", __func__, isr->level);
1356 	}
1357 
1358 	return (level);
1359 }
1360 
1361 /*
1362  * Check security policy requirements against the actual
1363  * packet contents.  Return one if the packet should be
1364  * reject as "invalid"; otherwiser return zero to have the
1365  * packet treated as "valid".
1366  *
1367  * OUT:
1368  *	0: valid
1369  *	1: invalid
1370  */
1371 int
1372 ipsec_in_reject(struct secpolicy *sp, struct mbuf *m)
1373 {
1374 	INIT_VNET_IPSEC(curvnet);
1375 	struct ipsecrequest *isr;
1376 	int need_auth;
1377 
1378 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1379 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1380 
1381 	/* Check policy. */
1382 	switch (sp->policy) {
1383 	case IPSEC_POLICY_DISCARD:
1384 		return (1);
1385 	case IPSEC_POLICY_BYPASS:
1386 	case IPSEC_POLICY_NONE:
1387 		return (0);
1388 	}
1389 
1390 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1391 		("invalid policy %u", sp->policy));
1392 
1393 	/* XXX Should compare policy against IPsec header history. */
1394 
1395 	need_auth = 0;
1396 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1397 		if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE)
1398 			continue;
1399 		switch (isr->saidx.proto) {
1400 		case IPPROTO_ESP:
1401 			if ((m->m_flags & M_DECRYPTED) == 0) {
1402 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1403 				    printf("%s: ESP m_flags:%x\n", __func__,
1404 					    m->m_flags));
1405 				return (1);
1406 			}
1407 
1408 			if (!need_auth &&
1409 			    isr->sav != NULL &&
1410 			    isr->sav->tdb_authalgxform != NULL &&
1411 			    (m->m_flags & M_AUTHIPDGM) == 0) {
1412 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1413 				    printf("%s: ESP/AH m_flags:%x\n", __func__,
1414 					    m->m_flags));
1415 				return (1);
1416 			}
1417 			break;
1418 		case IPPROTO_AH:
1419 			need_auth = 1;
1420 			if ((m->m_flags & M_AUTHIPHDR) == 0) {
1421 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1422 				    printf("%s: AH m_flags:%x\n", __func__,
1423 					    m->m_flags));
1424 				return (1);
1425 			}
1426 			break;
1427 		case IPPROTO_IPCOMP:
1428 			/*
1429 			 * We don't really care, as IPcomp document
1430 			 * says that we shouldn't compress small
1431 			 * packets.  IPComp policy should always be
1432 			 * treated as being in "use" level.
1433 			 */
1434 			break;
1435 		}
1436 	}
1437 	return (0);		/* Valid. */
1438 }
1439 
1440 /*
1441  * Check AH/ESP integrity.
1442  * This function is called from tcp_input(), udp_input(),
1443  * and {ah,esp}4_input for tunnel mode.
1444  */
1445 int
1446 ipsec4_in_reject(struct mbuf *m, struct inpcb *inp)
1447 {
1448 	INIT_VNET_IPSEC(curvnet);
1449 	struct secpolicy *sp;
1450 	int error;
1451 	int result;
1452 
1453 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1454 
1455 	/*
1456 	 * Get SP for this packet.
1457 	 * When we are called from ip_forward(), we call
1458 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1459 	 */
1460 	if (inp == NULL)
1461 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, IP_FORWARDING, &error);
1462 	else
1463 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND, inp, &error);
1464 
1465 	if (sp != NULL) {
1466 		result = ipsec_in_reject(sp, m);
1467 		if (result)
1468 			V_ipsec4stat.ips_in_polvio++;
1469 		KEY_FREESP(&sp);
1470 	} else {
1471 		result = 0;	/* XXX Should be panic?
1472 				 * -> No, there may be error. */
1473 	}
1474 	return (result);
1475 }
1476 
1477 #ifdef INET6
1478 /*
1479  * Check AH/ESP integrity.
1480  * This function is called from tcp6_input(), udp6_input(),
1481  * and {ah,esp}6_input for tunnel mode.
1482  */
1483 int
1484 ipsec6_in_reject(struct mbuf *m, struct inpcb *inp)
1485 {
1486 	INIT_VNET_IPSEC(curvnet);
1487 	struct secpolicy *sp = NULL;
1488 	int error;
1489 	int result;
1490 
1491 	/* Sanity check. */
1492 	if (m == NULL)
1493 		return (0);	/* XXX Should be panic? */
1494 
1495 	/* Get SP for this packet.
1496 	 * When we are called from ip_forward(), we call
1497 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1498 	 */
1499 	if (inp == NULL)
1500 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, IP_FORWARDING, &error);
1501 	else
1502 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND, inp, &error);
1503 
1504 	if (sp != NULL) {
1505 		result = ipsec_in_reject(sp, m);
1506 		if (result)
1507 			V_ipsec6stat.ips_in_polvio++;
1508 		KEY_FREESP(&sp);
1509 	} else {
1510 		result = 0;
1511 	}
1512 	return (result);
1513 }
1514 #endif
1515 
1516 /*
1517  * Compute the byte size to be occupied by IPsec header.
1518  * In case it is tunnelled, it includes the size of outer IP header.
1519  * NOTE: SP passed is freed in this function.
1520  */
1521 static size_t
1522 ipsec_hdrsiz(struct secpolicy *sp)
1523 {
1524 	INIT_VNET_IPSEC(curvnet);
1525 	struct ipsecrequest *isr;
1526 	size_t size;
1527 
1528 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1529 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1530 
1531 	switch (sp->policy) {
1532 	case IPSEC_POLICY_DISCARD:
1533 	case IPSEC_POLICY_BYPASS:
1534 	case IPSEC_POLICY_NONE:
1535 		return (0);
1536 	}
1537 
1538 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1539 		("invalid policy %u", sp->policy));
1540 
1541 	size = 0;
1542 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1543 		size_t clen = 0;
1544 
1545 		switch (isr->saidx.proto) {
1546 		case IPPROTO_ESP:
1547 			clen = esp_hdrsiz(isr->sav);
1548 			break;
1549 		case IPPROTO_AH:
1550 			clen = ah_hdrsiz(isr->sav);
1551 			break;
1552 		case IPPROTO_IPCOMP:
1553 			clen = sizeof(struct ipcomp);
1554 			break;
1555 		}
1556 
1557 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1558 			switch (isr->saidx.dst.sa.sa_family) {
1559 			case AF_INET:
1560 				clen += sizeof(struct ip);
1561 				break;
1562 #ifdef INET6
1563 			case AF_INET6:
1564 				clen += sizeof(struct ip6_hdr);
1565 				break;
1566 #endif
1567 			default:
1568 				ipseclog((LOG_ERR, "%s: unknown AF %d in "
1569 				    "IPsec tunnel SA\n", __func__,
1570 				    ((struct sockaddr *)&isr->saidx.dst)->sa_family));
1571 				break;
1572 			}
1573 		}
1574 		size += clen;
1575 	}
1576 
1577 	return (size);
1578 }
1579 
1580 /* This function is called from ip_forward() and ipsec4_hdrsize_tcp(). */
1581 size_t
1582 ipsec4_hdrsiz(struct mbuf *m, u_int dir, struct inpcb *inp)
1583 {
1584 	INIT_VNET_IPSEC(curvnet);
1585 	struct secpolicy *sp;
1586 	int error;
1587 	size_t size;
1588 
1589 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1590 
1591 	/* Get SP for this packet.
1592 	 * When we are called from ip_forward(), we call
1593 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1594 	 */
1595 	if (inp == NULL)
1596 		sp = ipsec_getpolicybyaddr(m, dir, IP_FORWARDING, &error);
1597 	else
1598 		sp = ipsec_getpolicybysock(m, dir, inp, &error);
1599 
1600 	if (sp != NULL) {
1601 		size = ipsec_hdrsiz(sp);
1602 		KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1603 			printf("%s: size:%lu.\n", __func__,
1604 				(unsigned long)size));
1605 
1606 		KEY_FREESP(&sp);
1607 	} else {
1608 		size = 0;	/* XXX Should be panic?
1609 				 * -> No, we are called w/o knowing if
1610 				 *    IPsec processing is needed. */
1611 	}
1612 	return (size);
1613 }
1614 
1615 #ifdef INET6
1616 /* This function is called from ipsec6_hdrsize_tcp(),
1617  * and maybe from ip6_forward().
1618  */
1619 size_t
1620 ipsec6_hdrsiz(struct mbuf *m, u_int dir, struct inpcb *inp)
1621 {
1622 	INIT_VNET_IPSEC(curvnet);
1623 	struct secpolicy *sp;
1624 	int error;
1625 	size_t size;
1626 
1627 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1628 	IPSEC_ASSERT(inp == NULL || inp->inp_socket != NULL,
1629 		("socket w/o inpcb"));
1630 
1631 	/* Get SP for this packet. */
1632 	/* XXX Is it right to call with IP_FORWARDING. */
1633 	if (inp == NULL)
1634 		sp = ipsec_getpolicybyaddr(m, dir, IP_FORWARDING, &error);
1635 	else
1636 		sp = ipsec_getpolicybysock(m, dir, inp, &error);
1637 
1638 	if (sp == NULL)
1639 		return (0);
1640 	size = ipsec_hdrsiz(sp);
1641 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1642 		printf("%s: size:%lu.\n", __func__, (unsigned long)size));
1643 	KEY_FREESP(&sp);
1644 
1645 	return (size);
1646 }
1647 #endif /*INET6*/
1648 
1649 /*
1650  * Check the variable replay window.
1651  * ipsec_chkreplay() performs replay check before ICV verification.
1652  * ipsec_updatereplay() updates replay bitmap.  This must be called after
1653  * ICV verification (it also performs replay check, which is usually done
1654  * beforehand).
1655  * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1656  *
1657  * Based on RFC 2401.
1658  */
1659 int
1660 ipsec_chkreplay(u_int32_t seq, struct secasvar *sav)
1661 {
1662 	const struct secreplay *replay;
1663 	u_int32_t diff;
1664 	int fr;
1665 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1666 	int frlast;		/* Constant: last frame. */
1667 
1668 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1669 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1670 
1671 	replay = sav->replay;
1672 
1673 	if (replay->wsize == 0)
1674 		return (1);	/* No need to check replay. */
1675 
1676 	/* Constant. */
1677 	frlast = replay->wsize - 1;
1678 	wsizeb = replay->wsize << 3;
1679 
1680 	/* Sequence number of 0 is invalid. */
1681 	if (seq == 0)
1682 		return (0);
1683 
1684 	/* First time is always okay. */
1685 	if (replay->count == 0)
1686 		return (1);
1687 
1688 	if (seq > replay->lastseq) {
1689 		/* Larger sequences are okay. */
1690 		return (1);
1691 	} else {
1692 		/* seq is equal or less than lastseq. */
1693 		diff = replay->lastseq - seq;
1694 
1695 		/* Over range to check, i.e. too old or wrapped. */
1696 		if (diff >= wsizeb)
1697 			return (0);
1698 
1699 		fr = frlast - diff / 8;
1700 
1701 		/* This packet already seen? */
1702 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1703 			return (0);
1704 
1705 		/* Out of order but good. */
1706 		return (1);
1707 	}
1708 }
1709 
1710 /*
1711  * Check replay counter whether to update or not.
1712  * OUT:	0:	OK
1713  *	1:	NG
1714  */
1715 int
1716 ipsec_updatereplay(u_int32_t seq, struct secasvar *sav)
1717 {
1718 	INIT_VNET_IPSEC(curvnet);
1719 	struct secreplay *replay;
1720 	u_int32_t diff;
1721 	int fr;
1722 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1723 	int frlast;		/* Constant: last frame. */
1724 
1725 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1726 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1727 
1728 	replay = sav->replay;
1729 
1730 	if (replay->wsize == 0)
1731 		goto ok;	/* No need to check replay. */
1732 
1733 	/* Constant. */
1734 	frlast = replay->wsize - 1;
1735 	wsizeb = replay->wsize << 3;
1736 
1737 	/* Sequence number of 0 is invalid. */
1738 	if (seq == 0)
1739 		return (1);
1740 
1741 	/* First time. */
1742 	if (replay->count == 0) {
1743 		replay->lastseq = seq;
1744 		bzero(replay->bitmap, replay->wsize);
1745 		(replay->bitmap)[frlast] = 1;
1746 		goto ok;
1747 	}
1748 
1749 	if (seq > replay->lastseq) {
1750 		/* seq is larger than lastseq. */
1751 		diff = seq - replay->lastseq;
1752 
1753 		/* New larger sequence number. */
1754 		if (diff < wsizeb) {
1755 			/* In window. */
1756 			/* Set bit for this packet. */
1757 			vshiftl(replay->bitmap, diff, replay->wsize);
1758 			(replay->bitmap)[frlast] |= 1;
1759 		} else {
1760 			/* This packet has a "way larger". */
1761 			bzero(replay->bitmap, replay->wsize);
1762 			(replay->bitmap)[frlast] = 1;
1763 		}
1764 		replay->lastseq = seq;
1765 
1766 		/* Larger is good. */
1767 	} else {
1768 		/* seq is equal or less than lastseq. */
1769 		diff = replay->lastseq - seq;
1770 
1771 		/* Over range to check, i.e. too old or wrapped. */
1772 		if (diff >= wsizeb)
1773 			return (1);
1774 
1775 		fr = frlast - diff / 8;
1776 
1777 		/* This packet already seen? */
1778 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1779 			return (1);
1780 
1781 		/* Mark as seen. */
1782 		(replay->bitmap)[fr] |= (1 << (diff % 8));
1783 
1784 		/* Out of order but good. */
1785 	}
1786 
1787 ok:
1788 	if (replay->count == ~0) {
1789 
1790 		/* Set overflow flag. */
1791 		replay->overflow++;
1792 
1793 		/* Don't increment, no more packets accepted. */
1794 		if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0)
1795 			return (1);
1796 
1797 		ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n",
1798 		    __func__, replay->overflow, ipsec_logsastr(sav)));
1799 	}
1800 
1801 	replay->count++;
1802 
1803 	return (0);
1804 }
1805 
1806 /*
1807  * Shift variable length buffer to left.
1808  * IN:	bitmap: pointer to the buffer
1809  * 	nbit:	the number of to shift.
1810  *	wsize:	buffer size (bytes).
1811  */
1812 static void
1813 vshiftl(unsigned char *bitmap, int nbit, int wsize)
1814 {
1815 	int s, j, i;
1816 	unsigned char over;
1817 
1818 	for (j = 0; j < nbit; j += 8) {
1819 		s = (nbit - j < 8) ? (nbit - j): 8;
1820 		bitmap[0] <<= s;
1821 		for (i = 1; i < wsize; i++) {
1822 			over = (bitmap[i] >> (8 - s));
1823 			bitmap[i] <<= s;
1824 			bitmap[i-1] |= over;
1825 		}
1826 	}
1827 }
1828 
1829 /* Return a printable string for the IPv4 address. */
1830 static char *
1831 inet_ntoa4(struct in_addr ina)
1832 {
1833 	static char buf[4][4 * sizeof "123" + 4];
1834 	unsigned char *ucp = (unsigned char *) &ina;
1835 	static int i = 3;
1836 
1837 	/* XXX-BZ Returns static buffer. */
1838 	i = (i + 1) % 4;
1839 	sprintf(buf[i], "%d.%d.%d.%d", ucp[0] & 0xff, ucp[1] & 0xff,
1840 	    ucp[2] & 0xff, ucp[3] & 0xff);
1841 	return (buf[i]);
1842 }
1843 
1844 /* Return a printable string for the address. */
1845 char *
1846 ipsec_address(union sockaddr_union* sa)
1847 {
1848 #ifdef INET6
1849 	char ip6buf[INET6_ADDRSTRLEN];
1850 #endif
1851 
1852 	switch (sa->sa.sa_family) {
1853 #ifdef INET
1854 	case AF_INET:
1855 		return (inet_ntoa4(sa->sin.sin_addr));
1856 #endif /* INET */
1857 #ifdef INET6
1858 	case AF_INET6:
1859 		return (ip6_sprintf(ip6buf, &sa->sin6.sin6_addr));
1860 #endif /* INET6 */
1861 	default:
1862 		return ("(unknown address family)");
1863 	}
1864 }
1865 
1866 const char *
1867 ipsec_logsastr(struct secasvar *sav)
1868 {
1869 	static char buf[256];
1870 	char *p;
1871 	struct secasindex *saidx = &sav->sah->saidx;
1872 
1873 	IPSEC_ASSERT(saidx->src.sa.sa_family == saidx->dst.sa.sa_family,
1874 		("address family mismatch"));
1875 
1876 	p = buf;
1877 	snprintf(buf, sizeof(buf), "SA(SPI=%u ", (u_int32_t)ntohl(sav->spi));
1878 	while (p && *p)
1879 		p++;
1880 	/* NB: only use ipsec_address on one address at a time. */
1881 	snprintf(p, sizeof (buf) - (p - buf), "src=%s ",
1882 		ipsec_address(&saidx->src));
1883 	while (p && *p)
1884 		p++;
1885 	snprintf(p, sizeof (buf) - (p - buf), "dst=%s)",
1886 		ipsec_address(&saidx->dst));
1887 
1888 	return (buf);
1889 }
1890 
1891 void
1892 ipsec_dumpmbuf(struct mbuf *m)
1893 {
1894 	int totlen;
1895 	int i;
1896 	u_char *p;
1897 
1898 	totlen = 0;
1899 	printf("---\n");
1900 	while (m) {
1901 		p = mtod(m, u_char *);
1902 		for (i = 0; i < m->m_len; i++) {
1903 			printf("%02x ", p[i]);
1904 			totlen++;
1905 			if (totlen % 16 == 0)
1906 				printf("\n");
1907 		}
1908 		m = m->m_next;
1909 	}
1910 	if (totlen % 16 != 0)
1911 		printf("\n");
1912 	printf("---\n");
1913 }
1914 
1915 static void
1916 ipsec_attach(void)
1917 {
1918 
1919 	SECPOLICY_LOCK_INIT(&V_ip4_def_policy);
1920 	V_ip4_def_policy.refcnt = 1;			/* NB: disallow free. */
1921 }
1922 SYSINIT(ipsec, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, ipsec_attach, NULL);
1923 
1924 
1925 /* XXX This stuff doesn't belong here... */
1926 
1927 static	struct xformsw* xforms = NULL;
1928 
1929 /*
1930  * Register a transform; typically at system startup.
1931  */
1932 void
1933 xform_register(struct xformsw* xsp)
1934 {
1935 
1936 	xsp->xf_next = xforms;
1937 	xforms = xsp;
1938 }
1939 
1940 /*
1941  * Initialize transform support in an sav.
1942  */
1943 int
1944 xform_init(struct secasvar *sav, int xftype)
1945 {
1946 	struct xformsw *xsp;
1947 
1948 	if (sav->tdb_xform != NULL)	/* Previously initialized. */
1949 		return (0);
1950 	for (xsp = xforms; xsp; xsp = xsp->xf_next)
1951 		if (xsp->xf_type == xftype)
1952 			return ((*xsp->xf_init)(sav, xsp));
1953 	return (EINVAL);
1954 }
1955