xref: /freebsd/sys/netipsec/ipsec.c (revision 129d3046ef0427d3b22b78a71f3494854d817fba)
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 static int ipsec_iattach(const void *);
107 
108 #ifdef VIMAGE_GLOBALS
109 /* NB: name changed so netstat doesn't use it. */
110 struct ipsecstat ipsec4stat;
111 struct secpolicy ip4_def_policy;
112 int ipsec_debug;
113 int ip4_ah_offsetmask;
114 int ip4_ipsec_dfbit;
115 int ip4_esp_trans_deflev;
116 int ip4_esp_net_deflev;
117 int ip4_ah_trans_deflev;
118 int ip4_ah_net_deflev;
119 int ip4_ipsec_ecn;
120 int ip4_esp_randpad;
121 /*
122  * Crypto support requirements:
123  *
124  *  1	require hardware support
125  * -1	require software support
126  *  0	take anything
127  */
128 int	crypto_support;
129 #endif /* VIMAGE_GLOBALS */
130 
131 SYSCTL_DECL(_net_inet_ipsec);
132 
133 /* net.inet.ipsec */
134 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DEF_POLICY,
135 	def_policy, CTLFLAG_RW, ip4_def_policy.policy,  0,
136 	"IPsec default policy.");
137 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV,
138 	esp_trans_deflev, CTLFLAG_RW, ip4_esp_trans_deflev,	0,
139 	"Default ESP transport mode level");
140 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV,
141 	esp_net_deflev, CTLFLAG_RW, ip4_esp_net_deflev,	0,
142 	"Default ESP tunnel mode level.");
143 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV,
144 	ah_trans_deflev, CTLFLAG_RW, ip4_ah_trans_deflev,	0,
145 	"AH transfer mode default level.");
146 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV,
147 	ah_net_deflev, CTLFLAG_RW, ip4_ah_net_deflev,	0,
148 	"AH tunnel mode default level.");
149 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_AH_CLEARTOS,
150 	ah_cleartos, CTLFLAG_RW,	ah_cleartos,	0,
151 	"If set clear type-of-service field when doing AH computation.");
152 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_AH_OFFSETMASK,
153 	ah_offsetmask, CTLFLAG_RW,	ip4_ah_offsetmask,	0,
154 	"If not set clear offset field mask when doing AH computation.");
155 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DFBIT,
156 	dfbit, CTLFLAG_RW,	ip4_ipsec_dfbit,	0,
157 	"Do not fragment bit on encap.");
158 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_ECN,
159 	ecn, CTLFLAG_RW,	ip4_ipsec_ecn,	0,
160 	"Explicit Congestion Notification handling.");
161 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, IPSECCTL_DEBUG,
162 	debug, CTLFLAG_RW,	ipsec_debug,	0,
163 	"Enable IPsec debugging output when set.");
164 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, OID_AUTO,
165 	crypto_support,	CTLFLAG_RW,	crypto_support,0,
166 	"Crypto driver selection.");
167 SYSCTL_V_STRUCT(V_NET, vnet_ipsec, _net_inet_ipsec, OID_AUTO,
168 	ipsecstats,	CTLFLAG_RD,	ipsec4stat, ipsecstat,
169 	"IPsec IPv4 statistics.");
170 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet_ipsec, OID_AUTO,
171 	filtertunnel, CTLFLAG_RW, ip4_ipsec_filtertunnel,  0,
172 	"If set filter packets from an IPsec tunnel.");
173 
174 #ifdef REGRESSION
175 #ifdef VIMAGE_GLOBALS
176 int ipsec_replay;
177 int ipsec_integrity;
178 #endif
179 /*
180  * When set to 1, IPsec will send packets with the same sequence number.
181  * This allows to verify if the other side has proper replay attacks detection.
182  */
183 SYSCTL_V_INT(V_NET, vnet_ipsec,_net_inet_ipsec, OID_AUTO, test_replay,
184 	CTLFLAG_RW, ipsec_replay, 0, "Emulate replay attack");
185 /*
186  * When set 1, IPsec will send packets with corrupted HMAC.
187  * This allows to verify if the other side properly detects modified packets.
188  */
189 SYSCTL_V_INT(V_NET, vnet_ipsec,_net_inet_ipsec, OID_AUTO, test_integrity,
190 	CTLFLAG_RW, ipsec_integrity, 0, "Emulate man-in-the-middle attack");
191 #endif
192 
193 #ifdef INET6
194 #ifdef VIMAGE_GLOBALS
195 struct ipsecstat ipsec6stat;
196 int ip6_esp_trans_deflev;
197 int ip6_esp_net_deflev;
198 int ip6_ah_trans_deflev;
199 int ip6_ah_net_deflev;
200 int ip6_ipsec_ecn;
201 #endif
202 
203 SYSCTL_DECL(_net_inet6_ipsec6);
204 
205 /* net.inet6.ipsec6 */
206 #ifdef COMPAT_KAME
207 SYSCTL_OID(_net_inet6_ipsec6, IPSECCTL_STATS, stats, CTLFLAG_RD,
208     0, 0, compat_ipsecstats_sysctl, "S", "IPsec IPv6 statistics.");
209 #endif /* COMPAT_KAME */
210 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_DEF_POLICY,
211 	def_policy, CTLFLAG_RW,	ip4_def_policy.policy,	0,
212 	"IPsec default policy.");
213 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV,
214 	esp_trans_deflev, CTLFLAG_RW, ip6_esp_trans_deflev,	0,
215 	"Default ESP transport mode level.");
216 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV,
217 	esp_net_deflev, CTLFLAG_RW, ip6_esp_net_deflev,	0,
218 	"Default ESP tunnel mode level.");
219 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV,
220 	ah_trans_deflev, CTLFLAG_RW, ip6_ah_trans_deflev,	0,
221 	"AH transfer mode default level.");
222 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV,
223 	ah_net_deflev, CTLFLAG_RW, ip6_ah_net_deflev,	0,
224 	"AH tunnel mode default level.");
225 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_ECN,
226 	ecn, CTLFLAG_RW, ip6_ipsec_ecn,	0,
227 	"Explicit Congestion Notification handling.");
228 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_DEBUG,
229 	debug, CTLFLAG_RW,	ipsec_debug,	0,
230 	"Enable IPsec debugging output when set.");
231 SYSCTL_V_STRUCT(V_NET, vnet_ipsec, _net_inet6_ipsec6, IPSECCTL_STATS,
232 	ipsecstats, CTLFLAG_RD, ipsec6stat, ipsecstat,
233 	"IPsec IPv6 statistics.");
234 SYSCTL_V_INT(V_NET, vnet_ipsec, _net_inet6_ipsec6, OID_AUTO,
235 	filtertunnel, CTLFLAG_RW, ip6_ipsec6_filtertunnel,  0,
236 	"If set filter packets from an IPsec tunnel.");
237 #endif /* INET6 */
238 
239 static int ipsec_setspidx_inpcb __P((struct mbuf *, struct inpcb *));
240 static int ipsec_setspidx __P((struct mbuf *, struct secpolicyindex *, int));
241 static void ipsec4_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
242 static int ipsec4_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
243 #ifdef INET6
244 static void ipsec6_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
245 static int ipsec6_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
246 #endif
247 static void ipsec_delpcbpolicy __P((struct inpcbpolicy *));
248 static struct secpolicy *ipsec_deepcopy_policy __P((struct secpolicy *src));
249 static void vshiftl __P((unsigned char *, int, int));
250 
251 MALLOC_DEFINE(M_IPSEC_INPCB, "inpcbpolicy", "inpcb-resident ipsec policy");
252 
253 #ifndef VIMAGE_GLOBALS
254 static const vnet_modinfo_t vnet_ipsec_modinfo = {
255 	.vmi_id		= VNET_MOD_IPSEC,
256 	.vmi_name	= "ipsec",
257 	.vmi_size	= sizeof(struct vnet_ipsec),
258 	.vmi_dependson	= VNET_MOD_INET,	/* XXX revisit - INET6 ? */
259 	.vmi_iattach	= ipsec_iattach
260 };
261 #endif /* !VIMAGE_GLOBALS */
262 
263 void
264 ipsec_init(void)
265 {
266 	INIT_VNET_IPSEC(curvnet);
267 
268 #ifdef IPSEC_DEBUG
269 	V_ipsec_debug = 1;
270 #else
271 	V_ipsec_debug = 0;
272 #endif
273 
274 	V_ip4_ah_offsetmask = 0;	/* maybe IP_DF? */
275 	V_ip4_ipsec_dfbit = 0;	/* DF bit on encap. 0: clear 1: set 2: copy */
276 	V_ip4_esp_trans_deflev = IPSEC_LEVEL_USE;
277 	V_ip4_esp_net_deflev = IPSEC_LEVEL_USE;
278 	V_ip4_ah_trans_deflev = IPSEC_LEVEL_USE;
279 	V_ip4_ah_net_deflev = IPSEC_LEVEL_USE;
280 	V_ip4_ipsec_ecn = 0;	/* ECN ignore(-1)/forbidden(0)/allowed(1) */
281 	V_ip4_esp_randpad = -1;
282 #ifdef IPSEC_FILTERTUNNEL
283 	V_ip4_ipsec_filtertunnel = 1;
284 #else
285 	V_ip4_ipsec_filtertunnel = 0;
286 #endif
287 
288 	V_crypto_support = CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE;
289 
290 #ifdef REGRESSION
291 	V_ipsec_replay = 0;
292 	V_ipsec_integrity = 0;
293 #endif
294 
295 #ifdef INET6
296 	V_ip6_esp_trans_deflev = IPSEC_LEVEL_USE;
297 	V_ip6_esp_net_deflev = IPSEC_LEVEL_USE;
298 	V_ip6_ah_trans_deflev = IPSEC_LEVEL_USE;
299 	V_ip6_ah_net_deflev = IPSEC_LEVEL_USE;
300 	V_ip6_ipsec_ecn = 0;	/* ECN ignore(-1)/forbidden(0)/allowed(1) */
301 #ifdef IPSEC_FILTERTUNNEL
302 	V_ip6_ipsec6_filtertunnel = 1;
303 #else
304 	V_ip6_ipsec6_filtertunnel = 0;
305 #endif
306 #endif
307 }
308 
309 /*
310  * Return a held reference to the default SP.
311  */
312 static struct secpolicy *
313 key_allocsp_default(const char* where, int tag)
314 {
315 	INIT_VNET_IPSEC(curvnet);
316 	struct secpolicy *sp;
317 
318 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
319 		printf("DP key_allocsp_default from %s:%u\n", where, tag));
320 
321 	sp = &V_ip4_def_policy;
322 	if (sp->policy != IPSEC_POLICY_DISCARD &&
323 	    sp->policy != IPSEC_POLICY_NONE) {
324 		ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n",
325 		    sp->policy, IPSEC_POLICY_NONE));
326 		sp->policy = IPSEC_POLICY_NONE;
327 	}
328 	key_addref(sp);
329 
330 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
331 		printf("DP key_allocsp_default returns SP:%p (%u)\n",
332 			sp, sp->refcnt));
333 	return (sp);
334 }
335 #define	KEY_ALLOCSP_DEFAULT() \
336 	key_allocsp_default(__FILE__, __LINE__)
337 
338 /*
339  * For OUTBOUND packet having a socket. Searching SPD for packet,
340  * and return a pointer to SP.
341  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
342  *		0	: bypass
343  *		EACCES	: discard packet.
344  *		ENOENT	: ipsec_acquire() in progress, maybe.
345  *		others	: error occured.
346  *	others:	a pointer to SP
347  *
348  * NOTE: IPv6 mapped adddress concern is implemented here.
349  */
350 struct secpolicy *
351 ipsec_getpolicy(struct tdb_ident *tdbi, u_int dir)
352 {
353 	struct secpolicy *sp;
354 
355 	IPSEC_ASSERT(tdbi != NULL, ("null tdbi"));
356 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
357 		("invalid direction %u", dir));
358 
359 	sp = KEY_ALLOCSP2(tdbi->spi, &tdbi->dst, tdbi->proto, dir);
360 	if (sp == NULL)			/*XXX????*/
361 		sp = KEY_ALLOCSP_DEFAULT();
362 	IPSEC_ASSERT(sp != NULL, ("null SP"));
363 	return (sp);
364 }
365 
366 /*
367  * For OUTBOUND packet having a socket. Searching SPD for packet,
368  * and return a pointer to SP.
369  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
370  *		0	: bypass
371  *		EACCES	: discard packet.
372  *		ENOENT	: ipsec_acquire() in progress, maybe.
373  *		others	: error occured.
374  *	others:	a pointer to SP
375  *
376  * NOTE: IPv6 mapped adddress concern is implemented here.
377  */
378 static struct secpolicy *
379 ipsec_getpolicybysock(struct mbuf *m, u_int dir, struct inpcb *inp, int *error)
380 {
381 	INIT_VNET_IPSEC(curvnet);
382 	struct inpcbpolicy *pcbsp;
383 	struct secpolicy *currsp = NULL;	/* Policy on socket. */
384 	struct secpolicy *sp;
385 
386 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
387 	IPSEC_ASSERT(inp != NULL, ("null inpcb"));
388 	IPSEC_ASSERT(error != NULL, ("null error"));
389 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
390 		("invalid direction %u", dir));
391 
392 	/* Set spidx in pcb. */
393 	*error = ipsec_setspidx_inpcb(m, inp);
394 	if (*error)
395 		return (NULL);
396 
397 	pcbsp = inp->inp_sp;
398 	IPSEC_ASSERT(pcbsp != NULL, ("null pcbsp"));
399 	switch (dir) {
400 	case IPSEC_DIR_INBOUND:
401 		currsp = pcbsp->sp_in;
402 		break;
403 	case IPSEC_DIR_OUTBOUND:
404 		currsp = pcbsp->sp_out;
405 		break;
406 	}
407 	IPSEC_ASSERT(currsp != NULL, ("null currsp"));
408 
409 	if (pcbsp->priv) {			/* When privilieged socket. */
410 		switch (currsp->policy) {
411 		case IPSEC_POLICY_BYPASS:
412 		case IPSEC_POLICY_IPSEC:
413 			key_addref(currsp);
414 			sp = currsp;
415 			break;
416 
417 		case IPSEC_POLICY_ENTRUST:
418 			/* Look for a policy in SPD. */
419 			sp = KEY_ALLOCSP(&currsp->spidx, dir);
420 			if (sp == NULL)		/* No SP found. */
421 				sp = KEY_ALLOCSP_DEFAULT();
422 			break;
423 
424 		default:
425 			ipseclog((LOG_ERR, "%s: Invalid policy for PCB %d\n",
426 				__func__, currsp->policy));
427 			*error = EINVAL;
428 			return (NULL);
429 		}
430 	} else {				/* Unpriv, SPD has policy. */
431 		sp = KEY_ALLOCSP(&currsp->spidx, dir);
432 		if (sp == NULL) {		/* No SP found. */
433 			switch (currsp->policy) {
434 			case IPSEC_POLICY_BYPASS:
435 				ipseclog((LOG_ERR, "%s: Illegal policy for "
436 					"non-priviliged defined %d\n",
437 					__func__, currsp->policy));
438 				*error = EINVAL;
439 				return (NULL);
440 
441 			case IPSEC_POLICY_ENTRUST:
442 				sp = KEY_ALLOCSP_DEFAULT();
443 				break;
444 
445 			case IPSEC_POLICY_IPSEC:
446 				key_addref(currsp);
447 				sp = currsp;
448 				break;
449 
450 			default:
451 				ipseclog((LOG_ERR, "%s: Invalid policy for "
452 					"PCB %d\n", __func__, currsp->policy));
453 				*error = EINVAL;
454 				return (NULL);
455 			}
456 		}
457 	}
458 	IPSEC_ASSERT(sp != NULL,
459 		("null SP (priv %u policy %u", pcbsp->priv, currsp->policy));
460 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
461 		printf("DP %s (priv %u policy %u) allocate SP:%p (refcnt %u)\n",
462 			__func__, pcbsp->priv, currsp->policy, sp, sp->refcnt));
463 	return (sp);
464 }
465 
466 /*
467  * For FORWADING packet or OUTBOUND without a socket. Searching SPD for packet,
468  * and return a pointer to SP.
469  * OUT:	positive: a pointer to the entry for security policy leaf matched.
470  *	NULL:	no apropreate SP found, the following value is set to error.
471  *		0	: bypass
472  *		EACCES	: discard packet.
473  *		ENOENT	: ipsec_acquire() in progress, maybe.
474  *		others	: error occured.
475  */
476 struct secpolicy *
477 ipsec_getpolicybyaddr(struct mbuf *m, u_int dir, int flag, int *error)
478 {
479 	INIT_VNET_IPSEC(curvnet);
480 	struct secpolicyindex spidx;
481 	struct secpolicy *sp;
482 
483 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
484 	IPSEC_ASSERT(error != NULL, ("null error"));
485 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
486 		("invalid direction %u", dir));
487 
488 	sp = NULL;
489 	if (key_havesp(dir)) {
490 		/* Make an index to look for a policy. */
491 		*error = ipsec_setspidx(m, &spidx,
492 					(flag & IP_FORWARDING) ? 0 : 1);
493 		if (*error != 0) {
494 			DPRINTF(("%s: setpidx failed, dir %u flag %u\n",
495 				__func__, dir, flag));
496 			return (NULL);
497 		}
498 		spidx.dir = dir;
499 
500 		sp = KEY_ALLOCSP(&spidx, dir);
501 	}
502 	if (sp == NULL)			/* No SP found, use system default. */
503 		sp = KEY_ALLOCSP_DEFAULT();
504 	IPSEC_ASSERT(sp != NULL, ("null SP"));
505 	return (sp);
506 }
507 
508 struct secpolicy *
509 ipsec4_checkpolicy(struct mbuf *m, u_int dir, u_int flag, int *error,
510     struct inpcb *inp)
511 {
512 	INIT_VNET_IPSEC(curvnet);
513 	struct secpolicy *sp;
514 
515 	*error = 0;
516 	if (inp == NULL)
517 		sp = ipsec_getpolicybyaddr(m, dir, flag, error);
518 	else
519 		sp = ipsec_getpolicybysock(m, dir, inp, error);
520 	if (sp == NULL) {
521 		IPSEC_ASSERT(*error != 0, ("getpolicy failed w/o error"));
522 		V_ipsec4stat.ips_out_inval++;
523 		return (NULL);
524 	}
525 	IPSEC_ASSERT(*error == 0, ("sp w/ error set to %u", *error));
526 	switch (sp->policy) {
527 	case IPSEC_POLICY_ENTRUST:
528 	default:
529 		printf("%s: invalid policy %u\n", __func__, sp->policy);
530 		/* FALLTHROUGH */
531 	case IPSEC_POLICY_DISCARD:
532 		V_ipsec4stat.ips_out_polvio++;
533 		*error = -EINVAL;	/* Packet is discarded by caller. */
534 		break;
535 	case IPSEC_POLICY_BYPASS:
536 	case IPSEC_POLICY_NONE:
537 		KEY_FREESP(&sp);
538 		sp = NULL;		/* NB: force NULL result. */
539 		break;
540 	case IPSEC_POLICY_IPSEC:
541 		if (sp->req == NULL)	/* Acquire a SA. */
542 			*error = key_spdacquire(sp);
543 		break;
544 	}
545 	if (*error != 0) {
546 		KEY_FREESP(&sp);
547 		sp = NULL;
548 	}
549 	return (sp);
550 }
551 
552 static int
553 ipsec_setspidx_inpcb(struct mbuf *m, struct inpcb *inp)
554 {
555 	int error;
556 
557 	IPSEC_ASSERT(inp != NULL, ("null inp"));
558 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
559 	IPSEC_ASSERT(inp->inp_sp->sp_out != NULL && inp->inp_sp->sp_in != NULL,
560 		("null sp_in || sp_out"));
561 
562 	error = ipsec_setspidx(m, &inp->inp_sp->sp_in->spidx, 1);
563 	if (error == 0) {
564 		inp->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
565 		inp->inp_sp->sp_out->spidx = inp->inp_sp->sp_in->spidx;
566 		inp->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
567 	} else {
568 		bzero(&inp->inp_sp->sp_in->spidx,
569 			sizeof (inp->inp_sp->sp_in->spidx));
570 		bzero(&inp->inp_sp->sp_out->spidx,
571 			sizeof (inp->inp_sp->sp_in->spidx));
572 	}
573 	return (error);
574 }
575 
576 /*
577  * Configure security policy index (src/dst/proto/sport/dport)
578  * by looking at the content of mbuf.
579  * The caller is responsible for error recovery (like clearing up spidx).
580  */
581 static int
582 ipsec_setspidx(struct mbuf *m, struct secpolicyindex *spidx, int needport)
583 {
584 	INIT_VNET_IPSEC(curvnet);
585 	struct ip *ip = NULL;
586 	struct ip ipbuf;
587 	u_int v;
588 	struct mbuf *n;
589 	int len;
590 	int error;
591 
592 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
593 
594 	/*
595 	 * Validate m->m_pkthdr.len.  We see incorrect length if we
596 	 * mistakenly call this function with inconsistent mbuf chain
597 	 * (like 4.4BSD tcp/udp processing).  XXX Should we panic here?
598 	 */
599 	len = 0;
600 	for (n = m; n; n = n->m_next)
601 		len += n->m_len;
602 	if (m->m_pkthdr.len != len) {
603 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
604 			printf("%s: pkthdr len(%d) mismatch (%d), ignored.\n",
605 				__func__, len, m->m_pkthdr.len));
606 		return (EINVAL);
607 	}
608 
609 	if (m->m_pkthdr.len < sizeof(struct ip)) {
610 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
611 			printf("%s: pkthdr len(%d) too small (v4), ignored.\n",
612 			    __func__, m->m_pkthdr.len));
613 		return (EINVAL);
614 	}
615 
616 	if (m->m_len >= sizeof(*ip))
617 		ip = mtod(m, struct ip *);
618 	else {
619 		m_copydata(m, 0, sizeof(ipbuf), (caddr_t)&ipbuf);
620 		ip = &ipbuf;
621 	}
622 #ifdef _IP_VHL
623 	v = _IP_VHL_V(ip->ip_vhl);
624 #else
625 	v = ip->ip_v;
626 #endif
627 	switch (v) {
628 	case 4:
629 		error = ipsec4_setspidx_ipaddr(m, spidx);
630 		if (error)
631 			return (error);
632 		ipsec4_get_ulp(m, spidx, needport);
633 		return (0);
634 #ifdef INET6
635 	case 6:
636 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) {
637 			KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
638 				printf("%s: pkthdr len(%d) too small (v6), "
639 				"ignored\n", __func__, m->m_pkthdr.len));
640 			return (EINVAL);
641 		}
642 		error = ipsec6_setspidx_ipaddr(m, spidx);
643 		if (error)
644 			return (error);
645 		ipsec6_get_ulp(m, spidx, needport);
646 		return (0);
647 #endif
648 	default:
649 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
650 			printf("%s: " "unknown IP version %u, ignored.\n",
651 				__func__, v));
652 		return (EINVAL);
653 	}
654 }
655 
656 static void
657 ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
658 {
659 	u_int8_t nxt;
660 	int off;
661 
662 	/* Sanity check. */
663 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
664 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),("packet too short"));
665 
666 	/* NB: ip_input() flips it into host endian. XXX Need more checking. */
667 	if (m->m_len < sizeof (struct ip)) {
668 		struct ip *ip = mtod(m, struct ip *);
669 		if (ip->ip_off & (IP_MF | IP_OFFMASK))
670 			goto done;
671 #ifdef _IP_VHL
672 		off = _IP_VHL_HL(ip->ip_vhl) << 2;
673 #else
674 		off = ip->ip_hl << 2;
675 #endif
676 		nxt = ip->ip_p;
677 	} else {
678 		struct ip ih;
679 
680 		m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih);
681 		if (ih.ip_off & (IP_MF | IP_OFFMASK))
682 			goto done;
683 #ifdef _IP_VHL
684 		off = _IP_VHL_HL(ih.ip_vhl) << 2;
685 #else
686 		off = ih.ip_hl << 2;
687 #endif
688 		nxt = ih.ip_p;
689 	}
690 
691 	while (off < m->m_pkthdr.len) {
692 		struct ip6_ext ip6e;
693 		struct tcphdr th;
694 		struct udphdr uh;
695 
696 		switch (nxt) {
697 		case IPPROTO_TCP:
698 			spidx->ul_proto = nxt;
699 			if (!needport)
700 				goto done_proto;
701 			if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
702 				goto done;
703 			m_copydata(m, off, sizeof (th), (caddr_t) &th);
704 			spidx->src.sin.sin_port = th.th_sport;
705 			spidx->dst.sin.sin_port = th.th_dport;
706 			return;
707 		case IPPROTO_UDP:
708 			spidx->ul_proto = nxt;
709 			if (!needport)
710 				goto done_proto;
711 			if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
712 				goto done;
713 			m_copydata(m, off, sizeof (uh), (caddr_t) &uh);
714 			spidx->src.sin.sin_port = uh.uh_sport;
715 			spidx->dst.sin.sin_port = uh.uh_dport;
716 			return;
717 		case IPPROTO_AH:
718 			if (off + sizeof(ip6e) > m->m_pkthdr.len)
719 				goto done;
720 			/* XXX Sigh, this works but is totally bogus. */
721 			m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e);
722 			off += (ip6e.ip6e_len + 2) << 2;
723 			nxt = ip6e.ip6e_nxt;
724 			break;
725 		case IPPROTO_ICMP:
726 		default:
727 			/* XXX Intermediate headers??? */
728 			spidx->ul_proto = nxt;
729 			goto done_proto;
730 		}
731 	}
732 done:
733 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
734 done_proto:
735 	spidx->src.sin.sin_port = IPSEC_PORT_ANY;
736 	spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
737 }
738 
739 /* Assumes that m is sane. */
740 static int
741 ipsec4_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
742 {
743 	static const struct sockaddr_in template = {
744 		sizeof (struct sockaddr_in),
745 		AF_INET,
746 		0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 }
747 	};
748 
749 	spidx->src.sin = template;
750 	spidx->dst.sin = template;
751 
752 	if (m->m_len < sizeof (struct ip)) {
753 		m_copydata(m, offsetof(struct ip, ip_src),
754 			   sizeof (struct  in_addr),
755 			   (caddr_t) &spidx->src.sin.sin_addr);
756 		m_copydata(m, offsetof(struct ip, ip_dst),
757 			   sizeof (struct  in_addr),
758 			   (caddr_t) &spidx->dst.sin.sin_addr);
759 	} else {
760 		struct ip *ip = mtod(m, struct ip *);
761 		spidx->src.sin.sin_addr = ip->ip_src;
762 		spidx->dst.sin.sin_addr = ip->ip_dst;
763 	}
764 
765 	spidx->prefs = sizeof(struct in_addr) << 3;
766 	spidx->prefd = sizeof(struct in_addr) << 3;
767 
768 	return (0);
769 }
770 
771 #ifdef INET6
772 static void
773 ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
774 {
775 	INIT_VNET_IPSEC(curvnet);
776 	int off, nxt;
777 	struct tcphdr th;
778 	struct udphdr uh;
779 	struct icmp6_hdr ih;
780 
781 	/* Sanity check. */
782 	if (m == NULL)
783 		panic("%s: NULL pointer was passed.\n", __func__);
784 
785 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
786 		printf("%s:\n", __func__); kdebug_mbuf(m));
787 
788 	/* Set default. */
789 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
790 	((struct sockaddr_in6 *)&spidx->src)->sin6_port = IPSEC_PORT_ANY;
791 	((struct sockaddr_in6 *)&spidx->dst)->sin6_port = IPSEC_PORT_ANY;
792 
793 	nxt = -1;
794 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
795 	if (off < 0 || m->m_pkthdr.len < off)
796 		return;
797 
798 	switch (nxt) {
799 	case IPPROTO_TCP:
800 		spidx->ul_proto = nxt;
801 		if (!needport)
802 			break;
803 		if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
804 			break;
805 		m_copydata(m, off, sizeof(th), (caddr_t)&th);
806 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = th.th_sport;
807 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = th.th_dport;
808 		break;
809 	case IPPROTO_UDP:
810 		spidx->ul_proto = nxt;
811 		if (!needport)
812 			break;
813 		if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
814 			break;
815 		m_copydata(m, off, sizeof(uh), (caddr_t)&uh);
816 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = uh.uh_sport;
817 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = uh.uh_dport;
818 		break;
819 	case IPPROTO_ICMPV6:
820 		spidx->ul_proto = nxt;
821 		if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len)
822 			break;
823 		m_copydata(m, off, sizeof(ih), (caddr_t)&ih);
824 		((struct sockaddr_in6 *)&spidx->src)->sin6_port =
825 		    htons((uint16_t)ih.icmp6_type);
826 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port =
827 		    htons((uint16_t)ih.icmp6_code);
828 		break;
829 	default:
830 		/* XXX Intermediate headers??? */
831 		spidx->ul_proto = nxt;
832 		break;
833 	}
834 }
835 
836 /* Assumes that m is sane. */
837 static int
838 ipsec6_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
839 {
840 	struct ip6_hdr *ip6 = NULL;
841 	struct ip6_hdr ip6buf;
842 	struct sockaddr_in6 *sin6;
843 
844 	if (m->m_len >= sizeof(*ip6))
845 		ip6 = mtod(m, struct ip6_hdr *);
846 	else {
847 		m_copydata(m, 0, sizeof(ip6buf), (caddr_t)&ip6buf);
848 		ip6 = &ip6buf;
849 	}
850 
851 	sin6 = (struct sockaddr_in6 *)&spidx->src;
852 	bzero(sin6, sizeof(*sin6));
853 	sin6->sin6_family = AF_INET6;
854 	sin6->sin6_len = sizeof(struct sockaddr_in6);
855 	bcopy(&ip6->ip6_src, &sin6->sin6_addr, sizeof(ip6->ip6_src));
856 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
857 		sin6->sin6_addr.s6_addr16[1] = 0;
858 		sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]);
859 	}
860 	spidx->prefs = sizeof(struct in6_addr) << 3;
861 
862 	sin6 = (struct sockaddr_in6 *)&spidx->dst;
863 	bzero(sin6, sizeof(*sin6));
864 	sin6->sin6_family = AF_INET6;
865 	sin6->sin6_len = sizeof(struct sockaddr_in6);
866 	bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(ip6->ip6_dst));
867 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
868 		sin6->sin6_addr.s6_addr16[1] = 0;
869 		sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]);
870 	}
871 	spidx->prefd = sizeof(struct in6_addr) << 3;
872 
873 	return (0);
874 }
875 #endif
876 
877 static void
878 ipsec_delpcbpolicy(struct inpcbpolicy *p)
879 {
880 
881 	free(p, M_IPSEC_INPCB);
882 }
883 
884 /* Initialize policy in PCB. */
885 int
886 ipsec_init_policy(struct socket *so, struct inpcbpolicy **pcb_sp)
887 {
888 	INIT_VNET_IPSEC(curvnet);
889 	struct inpcbpolicy *new;
890 
891 	/* Sanity check. */
892 	if (so == NULL || pcb_sp == NULL)
893 		panic("%s: NULL pointer was passed.\n", __func__);
894 
895 	new = (struct inpcbpolicy *) malloc(sizeof(struct inpcbpolicy),
896 					    M_IPSEC_INPCB, M_NOWAIT|M_ZERO);
897 	if (new == NULL) {
898 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
899 		return (ENOBUFS);
900 	}
901 
902 	new->priv = IPSEC_IS_PRIVILEGED_SO(so);
903 
904 	if ((new->sp_in = KEY_NEWSP()) == NULL) {
905 		ipsec_delpcbpolicy(new);
906 		return (ENOBUFS);
907 	}
908 	new->sp_in->state = IPSEC_SPSTATE_ALIVE;
909 	new->sp_in->policy = IPSEC_POLICY_ENTRUST;
910 
911 	if ((new->sp_out = KEY_NEWSP()) == NULL) {
912 		KEY_FREESP(&new->sp_in);
913 		ipsec_delpcbpolicy(new);
914 		return (ENOBUFS);
915 	}
916 	new->sp_out->state = IPSEC_SPSTATE_ALIVE;
917 	new->sp_out->policy = IPSEC_POLICY_ENTRUST;
918 
919 	*pcb_sp = new;
920 
921 	return (0);
922 }
923 
924 /* Copy old IPsec policy into new. */
925 int
926 ipsec_copy_policy(struct inpcbpolicy *old, struct inpcbpolicy *new)
927 {
928 	struct secpolicy *sp;
929 
930 	sp = ipsec_deepcopy_policy(old->sp_in);
931 	if (sp) {
932 		KEY_FREESP(&new->sp_in);
933 		new->sp_in = sp;
934 	} else
935 		return (ENOBUFS);
936 
937 	sp = ipsec_deepcopy_policy(old->sp_out);
938 	if (sp) {
939 		KEY_FREESP(&new->sp_out);
940 		new->sp_out = sp;
941 	} else
942 		return (ENOBUFS);
943 
944 	new->priv = old->priv;
945 
946 	return (0);
947 }
948 
949 struct ipsecrequest *
950 ipsec_newisr(void)
951 {
952 	struct ipsecrequest *p;
953 
954 	p = malloc(sizeof(struct ipsecrequest), M_IPSEC_SR, M_NOWAIT|M_ZERO);
955 	if (p != NULL)
956 		IPSECREQUEST_LOCK_INIT(p);
957 	return (p);
958 }
959 
960 void
961 ipsec_delisr(struct ipsecrequest *p)
962 {
963 
964 	IPSECREQUEST_LOCK_DESTROY(p);
965 	free(p, M_IPSEC_SR);
966 }
967 
968 /* Deep-copy a policy in PCB. */
969 static struct secpolicy *
970 ipsec_deepcopy_policy(struct secpolicy *src)
971 {
972 	struct ipsecrequest *newchain = NULL;
973 	struct ipsecrequest *p;
974 	struct ipsecrequest **q;
975 	struct ipsecrequest *r;
976 	struct secpolicy *dst;
977 
978 	if (src == NULL)
979 		return (NULL);
980 	dst = KEY_NEWSP();
981 	if (dst == NULL)
982 		return (NULL);
983 
984 	/*
985 	 * Deep-copy IPsec request chain.  This is required since struct
986 	 * ipsecrequest is not reference counted.
987 	 */
988 	q = &newchain;
989 	for (p = src->req; p; p = p->next) {
990 		*q = ipsec_newisr();
991 		if (*q == NULL)
992 			goto fail;
993 		(*q)->saidx.proto = p->saidx.proto;
994 		(*q)->saidx.mode = p->saidx.mode;
995 		(*q)->level = p->level;
996 		(*q)->saidx.reqid = p->saidx.reqid;
997 
998 		bcopy(&p->saidx.src, &(*q)->saidx.src, sizeof((*q)->saidx.src));
999 		bcopy(&p->saidx.dst, &(*q)->saidx.dst, sizeof((*q)->saidx.dst));
1000 
1001 		(*q)->sp = dst;
1002 
1003 		q = &((*q)->next);
1004 	}
1005 
1006 	dst->req = newchain;
1007 	dst->state = src->state;
1008 	dst->policy = src->policy;
1009 	/* Do not touch the refcnt fields. */
1010 
1011 	return (dst);
1012 
1013 fail:
1014 	for (p = newchain; p; p = r) {
1015 		r = p->next;
1016 		ipsec_delisr(p);
1017 		p = NULL;
1018 	}
1019 	return (NULL);
1020 }
1021 
1022 /* Set policy and IPsec request if present. */
1023 static int
1024 ipsec_set_policy_internal(struct secpolicy **pcb_sp, int optname,
1025     caddr_t request, size_t len, struct ucred *cred)
1026 {
1027 	INIT_VNET_IPSEC(curvnet);
1028 	struct sadb_x_policy *xpl;
1029 	struct secpolicy *newsp = NULL;
1030 	int error;
1031 
1032 	/* Sanity check. */
1033 	if (pcb_sp == NULL || *pcb_sp == NULL || request == NULL)
1034 		return (EINVAL);
1035 	if (len < sizeof(*xpl))
1036 		return (EINVAL);
1037 	xpl = (struct sadb_x_policy *)request;
1038 
1039 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1040 		printf("%s: passed policy\n", __func__);
1041 		kdebug_sadb_x_policy((struct sadb_ext *)xpl));
1042 
1043 	/* Check policy type. */
1044 	/* ipsec_set_policy_internal() accepts IPSEC, ENTRUST and BYPASS. */
1045 	if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD
1046 	 || xpl->sadb_x_policy_type == IPSEC_POLICY_NONE)
1047 		return (EINVAL);
1048 
1049 	/* Check privileged socket. */
1050 	if (cred != NULL && xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
1051 		error = priv_check_cred(cred, PRIV_NETINET_IPSEC, 0);
1052 		if (error)
1053 			return (EACCES);
1054 	}
1055 
1056 	/* Allocating new SP entry. */
1057 	if ((newsp = key_msg2sp(xpl, len, &error)) == NULL)
1058 		return (error);
1059 
1060 	newsp->state = IPSEC_SPSTATE_ALIVE;
1061 
1062 	/* Clear old SP and set new SP. */
1063 	KEY_FREESP(pcb_sp);
1064 	*pcb_sp = newsp;
1065 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1066 		printf("%s: new policy\n", __func__);
1067 		kdebug_secpolicy(newsp));
1068 
1069 	return (0);
1070 }
1071 
1072 int
1073 ipsec_set_policy(struct inpcb *inp, int optname, caddr_t request,
1074     size_t len, struct ucred *cred)
1075 {
1076 	INIT_VNET_IPSEC(curvnet);
1077 	struct sadb_x_policy *xpl;
1078 	struct secpolicy **pcb_sp;
1079 
1080 	/* Sanity check. */
1081 	if (inp == NULL || request == NULL)
1082 		return (EINVAL);
1083 	if (len < sizeof(*xpl))
1084 		return (EINVAL);
1085 	xpl = (struct sadb_x_policy *)request;
1086 
1087 	/* Select direction. */
1088 	switch (xpl->sadb_x_policy_dir) {
1089 	case IPSEC_DIR_INBOUND:
1090 		pcb_sp = &inp->inp_sp->sp_in;
1091 		break;
1092 	case IPSEC_DIR_OUTBOUND:
1093 		pcb_sp = &inp->inp_sp->sp_out;
1094 		break;
1095 	default:
1096 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1097 			xpl->sadb_x_policy_dir));
1098 		return (EINVAL);
1099 	}
1100 
1101 	return (ipsec_set_policy_internal(pcb_sp, optname, request, len, cred));
1102 }
1103 
1104 int
1105 ipsec_get_policy(struct inpcb *inp, caddr_t request, size_t len,
1106     struct mbuf **mp)
1107 {
1108 	INIT_VNET_IPSEC(curvnet);
1109 	struct sadb_x_policy *xpl;
1110 	struct secpolicy *pcb_sp;
1111 
1112 	/* Sanity check. */
1113 	if (inp == NULL || request == NULL || mp == NULL)
1114 		return (EINVAL);
1115 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
1116 	if (len < sizeof(*xpl))
1117 		return (EINVAL);
1118 	xpl = (struct sadb_x_policy *)request;
1119 
1120 	/* Select direction. */
1121 	switch (xpl->sadb_x_policy_dir) {
1122 	case IPSEC_DIR_INBOUND:
1123 		pcb_sp = inp->inp_sp->sp_in;
1124 		break;
1125 	case IPSEC_DIR_OUTBOUND:
1126 		pcb_sp = inp->inp_sp->sp_out;
1127 		break;
1128 	default:
1129 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1130 			xpl->sadb_x_policy_dir));
1131 		return (EINVAL);
1132 	}
1133 
1134 	/* Sanity check. Should be an IPSEC_ASSERT. */
1135 	if (pcb_sp == NULL)
1136 		return (EINVAL);
1137 
1138 	*mp = key_sp2msg(pcb_sp);
1139 	if (!*mp) {
1140 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
1141 		return (ENOBUFS);
1142 	}
1143 
1144 	(*mp)->m_type = MT_DATA;
1145 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1146 		printf("%s:\n", __func__); kdebug_mbuf(*mp));
1147 
1148 	return (0);
1149 }
1150 
1151 /* Delete policy in PCB. */
1152 int
1153 ipsec_delete_pcbpolicy(struct inpcb *inp)
1154 {
1155 	IPSEC_ASSERT(inp != NULL, ("null inp"));
1156 
1157 	if (inp->inp_sp == NULL)
1158 		return (0);
1159 
1160 	if (inp->inp_sp->sp_in != NULL)
1161 		KEY_FREESP(&inp->inp_sp->sp_in);
1162 
1163 	if (inp->inp_sp->sp_out != NULL)
1164 		KEY_FREESP(&inp->inp_sp->sp_out);
1165 
1166 	ipsec_delpcbpolicy(inp->inp_sp);
1167 	inp->inp_sp = NULL;
1168 
1169 	return (0);
1170 }
1171 
1172 /*
1173  * Return current level.
1174  * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
1175  */
1176 u_int
1177 ipsec_get_reqlevel(struct ipsecrequest *isr)
1178 {
1179 	INIT_VNET_IPSEC(curvnet);
1180 	u_int level = 0;
1181 	u_int esp_trans_deflev, esp_net_deflev;
1182 	u_int ah_trans_deflev, ah_net_deflev;
1183 
1184 	IPSEC_ASSERT(isr != NULL && isr->sp != NULL, ("null argument"));
1185 	IPSEC_ASSERT(isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family,
1186 		("af family mismatch, src %u, dst %u",
1187 		 isr->sp->spidx.src.sa.sa_family,
1188 		 isr->sp->spidx.dst.sa.sa_family));
1189 
1190 /* XXX Note that we have ipseclog() expanded here - code sync issue. */
1191 #define IPSEC_CHECK_DEFAULT(lev) \
1192 	(((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE	      \
1193 			&& (lev) != IPSEC_LEVEL_UNIQUE)			      \
1194 		? (V_ipsec_debug						      \
1195 			? log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
1196 				(lev), IPSEC_LEVEL_REQUIRE)		      \
1197 			: 0),						      \
1198 			(lev) = IPSEC_LEVEL_REQUIRE,			      \
1199 			(lev)						      \
1200 		: (lev))
1201 
1202 	/* Set default level. */
1203 	switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) {
1204 #ifdef INET
1205 	case AF_INET:
1206 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev);
1207 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev);
1208 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev);
1209 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev);
1210 		break;
1211 #endif
1212 #ifdef INET6
1213 	case AF_INET6:
1214 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev);
1215 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev);
1216 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev);
1217 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev);
1218 		break;
1219 #endif /* INET6 */
1220 	default:
1221 		panic("%s: unknown af %u",
1222 			__func__, isr->sp->spidx.src.sa.sa_family);
1223 	}
1224 
1225 #undef IPSEC_CHECK_DEFAULT
1226 
1227 	/* Set level. */
1228 	switch (isr->level) {
1229 	case IPSEC_LEVEL_DEFAULT:
1230 		switch (isr->saidx.proto) {
1231 		case IPPROTO_ESP:
1232 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1233 				level = esp_net_deflev;
1234 			else
1235 				level = esp_trans_deflev;
1236 			break;
1237 		case IPPROTO_AH:
1238 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1239 				level = ah_net_deflev;
1240 			else
1241 				level = ah_trans_deflev;
1242 			break;
1243 		case IPPROTO_IPCOMP:
1244 			/*
1245 			 * We don't really care, as IPcomp document says that
1246 			 * we shouldn't compress small packets.
1247 			 */
1248 			level = IPSEC_LEVEL_USE;
1249 			break;
1250 		default:
1251 			panic("%s: Illegal protocol defined %u\n", __func__,
1252 				isr->saidx.proto);
1253 		}
1254 		break;
1255 
1256 	case IPSEC_LEVEL_USE:
1257 	case IPSEC_LEVEL_REQUIRE:
1258 		level = isr->level;
1259 		break;
1260 	case IPSEC_LEVEL_UNIQUE:
1261 		level = IPSEC_LEVEL_REQUIRE;
1262 		break;
1263 
1264 	default:
1265 		panic("%s: Illegal IPsec level %u\n", __func__, isr->level);
1266 	}
1267 
1268 	return (level);
1269 }
1270 
1271 /*
1272  * Check security policy requirements against the actual
1273  * packet contents.  Return one if the packet should be
1274  * reject as "invalid"; otherwiser return zero to have the
1275  * packet treated as "valid".
1276  *
1277  * OUT:
1278  *	0: valid
1279  *	1: invalid
1280  */
1281 int
1282 ipsec_in_reject(struct secpolicy *sp, struct mbuf *m)
1283 {
1284 	INIT_VNET_IPSEC(curvnet);
1285 	struct ipsecrequest *isr;
1286 	int need_auth;
1287 
1288 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1289 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1290 
1291 	/* Check policy. */
1292 	switch (sp->policy) {
1293 	case IPSEC_POLICY_DISCARD:
1294 		return (1);
1295 	case IPSEC_POLICY_BYPASS:
1296 	case IPSEC_POLICY_NONE:
1297 		return (0);
1298 	}
1299 
1300 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1301 		("invalid policy %u", sp->policy));
1302 
1303 	/* XXX Should compare policy against IPsec header history. */
1304 
1305 	need_auth = 0;
1306 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1307 		if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE)
1308 			continue;
1309 		switch (isr->saidx.proto) {
1310 		case IPPROTO_ESP:
1311 			if ((m->m_flags & M_DECRYPTED) == 0) {
1312 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1313 				    printf("%s: ESP m_flags:%x\n", __func__,
1314 					    m->m_flags));
1315 				return (1);
1316 			}
1317 
1318 			if (!need_auth &&
1319 			    isr->sav != NULL &&
1320 			    isr->sav->tdb_authalgxform != NULL &&
1321 			    (m->m_flags & M_AUTHIPDGM) == 0) {
1322 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1323 				    printf("%s: ESP/AH m_flags:%x\n", __func__,
1324 					    m->m_flags));
1325 				return (1);
1326 			}
1327 			break;
1328 		case IPPROTO_AH:
1329 			need_auth = 1;
1330 			if ((m->m_flags & M_AUTHIPHDR) == 0) {
1331 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1332 				    printf("%s: AH m_flags:%x\n", __func__,
1333 					    m->m_flags));
1334 				return (1);
1335 			}
1336 			break;
1337 		case IPPROTO_IPCOMP:
1338 			/*
1339 			 * We don't really care, as IPcomp document
1340 			 * says that we shouldn't compress small
1341 			 * packets.  IPComp policy should always be
1342 			 * treated as being in "use" level.
1343 			 */
1344 			break;
1345 		}
1346 	}
1347 	return (0);		/* Valid. */
1348 }
1349 
1350 static int
1351 ipsec46_in_reject(struct mbuf *m, struct inpcb *inp)
1352 {
1353 	struct secpolicy *sp;
1354 	int error;
1355 	int result;
1356 
1357 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1358 
1359 	/*
1360 	 * Get SP for this packet.
1361 	 * When we are called from ip_forward(), we call
1362 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1363 	 */
1364 	if (inp == NULL)
1365 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, IP_FORWARDING, &error);
1366 	else
1367 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND, inp, &error);
1368 
1369 	if (sp != NULL) {
1370 		result = ipsec_in_reject(sp, m);
1371 		KEY_FREESP(&sp);
1372 	} else {
1373 		result = 0;	/* XXX Should be panic?
1374 				 * -> No, there may be error. */
1375 	}
1376 	return (result);
1377 }
1378 
1379 /*
1380  * Check AH/ESP integrity.
1381  * This function is called from tcp_input(), udp_input(),
1382  * and {ah,esp}4_input for tunnel mode.
1383  */
1384 int
1385 ipsec4_in_reject(struct mbuf *m, struct inpcb *inp)
1386 {
1387 	INIT_VNET_IPSEC(curvnet);
1388 	int result;
1389 
1390 	result = ipsec46_in_reject(m, inp);
1391 	if (result)
1392 		V_ipsec4stat.ips_in_polvio++;
1393 
1394 	return (result);
1395 }
1396 
1397 #ifdef INET6
1398 /*
1399  * Check AH/ESP integrity.
1400  * This function is called from tcp6_input(), udp6_input(),
1401  * and {ah,esp}6_input for tunnel mode.
1402  */
1403 int
1404 ipsec6_in_reject(struct mbuf *m, struct inpcb *inp)
1405 {
1406 	INIT_VNET_IPSEC(curvnet);
1407 	int result;
1408 
1409 	result = ipsec46_in_reject(m, inp);
1410 	if (result)
1411 		V_ipsec6stat.ips_in_polvio++;
1412 
1413 	return (result);
1414 }
1415 #endif
1416 
1417 /*
1418  * Compute the byte size to be occupied by IPsec header.
1419  * In case it is tunnelled, it includes the size of outer IP header.
1420  * NOTE: SP passed is freed in this function.
1421  */
1422 static size_t
1423 ipsec_hdrsiz_internal(struct secpolicy *sp)
1424 {
1425 	INIT_VNET_IPSEC(curvnet);
1426 	struct ipsecrequest *isr;
1427 	size_t size;
1428 
1429 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1430 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1431 
1432 	switch (sp->policy) {
1433 	case IPSEC_POLICY_DISCARD:
1434 	case IPSEC_POLICY_BYPASS:
1435 	case IPSEC_POLICY_NONE:
1436 		return (0);
1437 	}
1438 
1439 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1440 		("invalid policy %u", sp->policy));
1441 
1442 	size = 0;
1443 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1444 		size_t clen = 0;
1445 
1446 		switch (isr->saidx.proto) {
1447 		case IPPROTO_ESP:
1448 			clen = esp_hdrsiz(isr->sav);
1449 			break;
1450 		case IPPROTO_AH:
1451 			clen = ah_hdrsiz(isr->sav);
1452 			break;
1453 		case IPPROTO_IPCOMP:
1454 			clen = sizeof(struct ipcomp);
1455 			break;
1456 		}
1457 
1458 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1459 			switch (isr->saidx.dst.sa.sa_family) {
1460 			case AF_INET:
1461 				clen += sizeof(struct ip);
1462 				break;
1463 #ifdef INET6
1464 			case AF_INET6:
1465 				clen += sizeof(struct ip6_hdr);
1466 				break;
1467 #endif
1468 			default:
1469 				ipseclog((LOG_ERR, "%s: unknown AF %d in "
1470 				    "IPsec tunnel SA\n", __func__,
1471 				    ((struct sockaddr *)&isr->saidx.dst)->sa_family));
1472 				break;
1473 			}
1474 		}
1475 		size += clen;
1476 	}
1477 
1478 	return (size);
1479 }
1480 
1481 /*
1482  * This function is called from ipsec_hdrsiz_tcp(), ip_ipsec_mtu(),
1483  * disabled ip6_ipsec_mtu() and ip6_forward().
1484  */
1485 size_t
1486 ipsec_hdrsiz(struct mbuf *m, u_int dir, struct inpcb *inp)
1487 {
1488 	INIT_VNET_IPSEC(curvnet);
1489 	struct secpolicy *sp;
1490 	int error;
1491 	size_t size;
1492 
1493 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
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, dir, IP_FORWARDING, &error);
1501 	else
1502 		sp = ipsec_getpolicybysock(m, dir, inp, &error);
1503 
1504 	if (sp != NULL) {
1505 		size = ipsec_hdrsiz_internal(sp);
1506 		KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1507 			printf("%s: size:%lu.\n", __func__,
1508 				(unsigned long)size));
1509 
1510 		KEY_FREESP(&sp);
1511 	} else {
1512 		size = 0;	/* XXX Should be panic?
1513 				 * -> No, we are called w/o knowing if
1514 				 *    IPsec processing is needed. */
1515 	}
1516 	return (size);
1517 }
1518 
1519 /*
1520  * Check the variable replay window.
1521  * ipsec_chkreplay() performs replay check before ICV verification.
1522  * ipsec_updatereplay() updates replay bitmap.  This must be called after
1523  * ICV verification (it also performs replay check, which is usually done
1524  * beforehand).
1525  * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1526  *
1527  * Based on RFC 2401.
1528  */
1529 int
1530 ipsec_chkreplay(u_int32_t seq, struct secasvar *sav)
1531 {
1532 	const struct secreplay *replay;
1533 	u_int32_t diff;
1534 	int fr;
1535 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1536 	int frlast;		/* Constant: last frame. */
1537 
1538 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1539 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1540 
1541 	replay = sav->replay;
1542 
1543 	if (replay->wsize == 0)
1544 		return (1);	/* No need to check replay. */
1545 
1546 	/* Constant. */
1547 	frlast = replay->wsize - 1;
1548 	wsizeb = replay->wsize << 3;
1549 
1550 	/* Sequence number of 0 is invalid. */
1551 	if (seq == 0)
1552 		return (0);
1553 
1554 	/* First time is always okay. */
1555 	if (replay->count == 0)
1556 		return (1);
1557 
1558 	if (seq > replay->lastseq) {
1559 		/* Larger sequences are okay. */
1560 		return (1);
1561 	} else {
1562 		/* seq is equal or less than lastseq. */
1563 		diff = replay->lastseq - seq;
1564 
1565 		/* Over range to check, i.e. too old or wrapped. */
1566 		if (diff >= wsizeb)
1567 			return (0);
1568 
1569 		fr = frlast - diff / 8;
1570 
1571 		/* This packet already seen? */
1572 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1573 			return (0);
1574 
1575 		/* Out of order but good. */
1576 		return (1);
1577 	}
1578 }
1579 
1580 /*
1581  * Check replay counter whether to update or not.
1582  * OUT:	0:	OK
1583  *	1:	NG
1584  */
1585 int
1586 ipsec_updatereplay(u_int32_t seq, struct secasvar *sav)
1587 {
1588 	INIT_VNET_IPSEC(curvnet);
1589 	struct secreplay *replay;
1590 	u_int32_t diff;
1591 	int fr;
1592 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1593 	int frlast;		/* Constant: last frame. */
1594 
1595 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1596 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1597 
1598 	replay = sav->replay;
1599 
1600 	if (replay->wsize == 0)
1601 		goto ok;	/* No need to check replay. */
1602 
1603 	/* Constant. */
1604 	frlast = replay->wsize - 1;
1605 	wsizeb = replay->wsize << 3;
1606 
1607 	/* Sequence number of 0 is invalid. */
1608 	if (seq == 0)
1609 		return (1);
1610 
1611 	/* First time. */
1612 	if (replay->count == 0) {
1613 		replay->lastseq = seq;
1614 		bzero(replay->bitmap, replay->wsize);
1615 		(replay->bitmap)[frlast] = 1;
1616 		goto ok;
1617 	}
1618 
1619 	if (seq > replay->lastseq) {
1620 		/* seq is larger than lastseq. */
1621 		diff = seq - replay->lastseq;
1622 
1623 		/* New larger sequence number. */
1624 		if (diff < wsizeb) {
1625 			/* In window. */
1626 			/* Set bit for this packet. */
1627 			vshiftl(replay->bitmap, diff, replay->wsize);
1628 			(replay->bitmap)[frlast] |= 1;
1629 		} else {
1630 			/* This packet has a "way larger". */
1631 			bzero(replay->bitmap, replay->wsize);
1632 			(replay->bitmap)[frlast] = 1;
1633 		}
1634 		replay->lastseq = seq;
1635 
1636 		/* Larger is good. */
1637 	} else {
1638 		/* seq is equal or less than lastseq. */
1639 		diff = replay->lastseq - seq;
1640 
1641 		/* Over range to check, i.e. too old or wrapped. */
1642 		if (diff >= wsizeb)
1643 			return (1);
1644 
1645 		fr = frlast - diff / 8;
1646 
1647 		/* This packet already seen? */
1648 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1649 			return (1);
1650 
1651 		/* Mark as seen. */
1652 		(replay->bitmap)[fr] |= (1 << (diff % 8));
1653 
1654 		/* Out of order but good. */
1655 	}
1656 
1657 ok:
1658 	if (replay->count == ~0) {
1659 
1660 		/* Set overflow flag. */
1661 		replay->overflow++;
1662 
1663 		/* Don't increment, no more packets accepted. */
1664 		if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0)
1665 			return (1);
1666 
1667 		ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n",
1668 		    __func__, replay->overflow, ipsec_logsastr(sav)));
1669 	}
1670 
1671 	replay->count++;
1672 
1673 	return (0);
1674 }
1675 
1676 /*
1677  * Shift variable length buffer to left.
1678  * IN:	bitmap: pointer to the buffer
1679  * 	nbit:	the number of to shift.
1680  *	wsize:	buffer size (bytes).
1681  */
1682 static void
1683 vshiftl(unsigned char *bitmap, int nbit, int wsize)
1684 {
1685 	int s, j, i;
1686 	unsigned char over;
1687 
1688 	for (j = 0; j < nbit; j += 8) {
1689 		s = (nbit - j < 8) ? (nbit - j): 8;
1690 		bitmap[0] <<= s;
1691 		for (i = 1; i < wsize; i++) {
1692 			over = (bitmap[i] >> (8 - s));
1693 			bitmap[i] <<= s;
1694 			bitmap[i-1] |= over;
1695 		}
1696 	}
1697 }
1698 
1699 /* Return a printable string for the IPv4 address. */
1700 static char *
1701 inet_ntoa4(struct in_addr ina)
1702 {
1703 	static char buf[4][4 * sizeof "123" + 4];
1704 	unsigned char *ucp = (unsigned char *) &ina;
1705 	static int i = 3;
1706 
1707 	/* XXX-BZ Returns static buffer. */
1708 	i = (i + 1) % 4;
1709 	sprintf(buf[i], "%d.%d.%d.%d", ucp[0] & 0xff, ucp[1] & 0xff,
1710 	    ucp[2] & 0xff, ucp[3] & 0xff);
1711 	return (buf[i]);
1712 }
1713 
1714 /* Return a printable string for the address. */
1715 char *
1716 ipsec_address(union sockaddr_union* sa)
1717 {
1718 #ifdef INET6
1719 	char ip6buf[INET6_ADDRSTRLEN];
1720 #endif
1721 
1722 	switch (sa->sa.sa_family) {
1723 #ifdef INET
1724 	case AF_INET:
1725 		return (inet_ntoa4(sa->sin.sin_addr));
1726 #endif /* INET */
1727 #ifdef INET6
1728 	case AF_INET6:
1729 		return (ip6_sprintf(ip6buf, &sa->sin6.sin6_addr));
1730 #endif /* INET6 */
1731 	default:
1732 		return ("(unknown address family)");
1733 	}
1734 }
1735 
1736 const char *
1737 ipsec_logsastr(struct secasvar *sav)
1738 {
1739 	static char buf[256];
1740 	char *p;
1741 	struct secasindex *saidx = &sav->sah->saidx;
1742 
1743 	IPSEC_ASSERT(saidx->src.sa.sa_family == saidx->dst.sa.sa_family,
1744 		("address family mismatch"));
1745 
1746 	p = buf;
1747 	snprintf(buf, sizeof(buf), "SA(SPI=%u ", (u_int32_t)ntohl(sav->spi));
1748 	while (p && *p)
1749 		p++;
1750 	/* NB: only use ipsec_address on one address at a time. */
1751 	snprintf(p, sizeof (buf) - (p - buf), "src=%s ",
1752 		ipsec_address(&saidx->src));
1753 	while (p && *p)
1754 		p++;
1755 	snprintf(p, sizeof (buf) - (p - buf), "dst=%s)",
1756 		ipsec_address(&saidx->dst));
1757 
1758 	return (buf);
1759 }
1760 
1761 void
1762 ipsec_dumpmbuf(struct mbuf *m)
1763 {
1764 	int totlen;
1765 	int i;
1766 	u_char *p;
1767 
1768 	totlen = 0;
1769 	printf("---\n");
1770 	while (m) {
1771 		p = mtod(m, u_char *);
1772 		for (i = 0; i < m->m_len; i++) {
1773 			printf("%02x ", p[i]);
1774 			totlen++;
1775 			if (totlen % 16 == 0)
1776 				printf("\n");
1777 		}
1778 		m = m->m_next;
1779 	}
1780 	if (totlen % 16 != 0)
1781 		printf("\n");
1782 	printf("---\n");
1783 }
1784 
1785 static void
1786 ipsec_attach(void)
1787 {
1788 
1789 #ifndef VIMAGE_GLOBALS
1790 	vnet_mod_register(&vnet_ipsec_modinfo);
1791 #else
1792 	ipsec_iattach(NULL);
1793 #endif
1794 
1795 }
1796 
1797 static int
1798 ipsec_iattach(const void *unused __unused)
1799 {
1800 	INIT_VNET_IPSEC(curvnet);
1801 
1802 	SECPOLICY_LOCK_INIT(&V_ip4_def_policy);
1803 	V_ip4_def_policy.refcnt = 1;			/* NB: disallow free. */
1804 
1805 	return (0);
1806 }
1807 SYSINIT(ipsec, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, ipsec_attach, NULL);
1808 
1809 
1810 /* XXX This stuff doesn't belong here... */
1811 
1812 static	struct xformsw* xforms = NULL;
1813 
1814 /*
1815  * Register a transform; typically at system startup.
1816  */
1817 void
1818 xform_register(struct xformsw* xsp)
1819 {
1820 
1821 	xsp->xf_next = xforms;
1822 	xforms = xsp;
1823 }
1824 
1825 /*
1826  * Initialize transform support in an sav.
1827  */
1828 int
1829 xform_init(struct secasvar *sav, int xftype)
1830 {
1831 	struct xformsw *xsp;
1832 
1833 	if (sav->tdb_xform != NULL)	/* Previously initialized. */
1834 		return (0);
1835 	for (xsp = xforms; xsp; xsp = xsp->xf_next)
1836 		if (xsp->xf_type == xftype)
1837 			return ((*xsp->xf_init)(sav, xsp));
1838 	return (EINVAL);
1839 }
1840