xref: /netbsd-src/sys/netipsec/ipsec.c (revision df0caa2637da0538ecdf6b878c4d08e684b43d8f)
1 /*	$NetBSD: ipsec.c,v 1.17 2005/06/10 13:22:42 christos Exp $	*/
2 /*	$FreeBSD: /usr/local/www/cvsroot/FreeBSD/src/sys/netipsec/ipsec.c,v 1.2.2.2 2003/07/01 01:38:13 sam Exp $	*/
3 /*	$KAME: ipsec.c,v 1.103 2001/05/24 07:14:18 sakane Exp $	*/
4 
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 #include <sys/cdefs.h>
35 __KERNEL_RCSID(0, "$NetBSD: ipsec.c,v 1.17 2005/06/10 13:22:42 christos Exp $");
36 
37 /*
38  * IPsec controller part.
39  */
40 
41 #include "opt_inet.h"
42 #ifdef __FreeBSD__
43 #include "opt_inet6.h"
44 #endif
45 #include "opt_ipsec.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/malloc.h>
50 #include <sys/mbuf.h>
51 #include <sys/domain.h>
52 #include <sys/protosw.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/errno.h>
56 #include <sys/time.h>
57 #include <sys/kernel.h>
58 #include <sys/syslog.h>
59 #include <sys/sysctl.h>
60 #include <sys/proc.h>
61 
62 #include <net/if.h>
63 #include <net/route.h>
64 
65 #include <netinet/in.h>
66 #include <netinet/in_systm.h>
67 #include <netinet/ip.h>
68 #include <netinet/ip_var.h>
69 #include <netinet/in_var.h>
70 #include <netinet/udp.h>
71 #include <netinet/udp_var.h>
72 #include <netinet/tcp.h>
73 #include <netinet/udp.h>
74 
75 #include <netinet/ip6.h>
76 #ifdef INET6
77 #include <netinet6/ip6_var.h>
78 #endif
79 #include <netinet/in_pcb.h>
80 #ifdef INET6
81 #include <netinet6/in6_pcb.h>
82 #include <netinet/icmp6.h>
83 #endif
84 
85 #include <netipsec/ipsec.h>
86 #include <netipsec/ipsec_var.h>
87 #ifdef INET6
88 #include <netipsec/ipsec6.h>
89 #endif
90 #include <netipsec/ah_var.h>
91 #include <netipsec/esp_var.h>
92 #include <netipsec/ipcomp.h>		/*XXX*/
93 #include <netipsec/ipcomp_var.h>
94 
95 #include <netipsec/key.h>
96 #include <netipsec/keydb.h>
97 #include <netipsec/key_debug.h>
98 
99 #include <netipsec/xform.h>
100 
101 #include <netipsec/ipsec_osdep.h>
102 
103 #include <net/net_osdep.h>
104 
105 #ifdef IPSEC_DEBUG
106 int ipsec_debug = 1;
107 #else
108 int ipsec_debug = 0;
109 #endif
110 
111 /* NB: name changed so netstat doesn't use it */
112 struct newipsecstat newipsecstat;
113 int ip4_ah_offsetmask = 0;	/* maybe IP_DF? */
114 int ip4_ipsec_dfbit = 0;	/* DF bit on encap. 0: clear 1: set 2: copy */
115 int ip4_esp_trans_deflev = IPSEC_LEVEL_USE;
116 int ip4_esp_net_deflev = IPSEC_LEVEL_USE;
117 int ip4_ah_trans_deflev = IPSEC_LEVEL_USE;
118 int ip4_ah_net_deflev = IPSEC_LEVEL_USE;
119 struct secpolicy ip4_def_policy;
120 int ip4_ipsec_ecn = 0;		/* ECN ignore(-1)/forbidden(0)/allowed(1) */
121 int ip4_esp_randpad = -1;
122 
123 #ifdef __NetBSD__
124 u_int ipsec_spdgen = 1;		/* SPD generation # */
125 
126 static struct secpolicy *ipsec_checkpcbcache __P((struct mbuf *,
127 	struct inpcbpolicy *, int));
128 static int ipsec_fillpcbcache __P((struct inpcbpolicy *, struct mbuf *,
129 	struct secpolicy *, int));
130 static int ipsec_invalpcbcache __P((struct inpcbpolicy *, int));
131 #endif /* __NetBSD__ */
132 
133 /*
134  * Crypto support requirements:
135  *
136  *  1	require hardware support
137  * -1	require software support
138  *  0	take anything
139  */
140 int	crypto_support = 0;
141 
142 static struct secpolicy *ipsec_getpolicybysock(struct mbuf *, u_int,
143 	PCB_T *, int *);
144 
145 #ifdef __FreeBSD__
146 SYSCTL_DECL(_net_inet_ipsec);
147 
148 /* net.inet.ipsec */
149 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_POLICY,
150 	def_policy, CTLFLAG_RW,	&ip4_def_policy.policy,	0, "");
151 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
152 	CTLFLAG_RW, &ip4_esp_trans_deflev,	0, "");
153 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
154 	CTLFLAG_RW, &ip4_esp_net_deflev,	0, "");
155 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
156 	CTLFLAG_RW, &ip4_ah_trans_deflev,	0, "");
157 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
158 	CTLFLAG_RW, &ip4_ah_net_deflev,	0, "");
159 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_CLEARTOS,
160 	ah_cleartos, CTLFLAG_RW,	&ah_cleartos,	0, "");
161 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_OFFSETMASK,
162 	ah_offsetmask, CTLFLAG_RW,	&ip4_ah_offsetmask,	0, "");
163 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DFBIT,
164 	dfbit, CTLFLAG_RW,	&ip4_ipsec_dfbit,	0, "");
165 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_ECN,
166 	ecn, CTLFLAG_RW,	&ip4_ipsec_ecn,	0, "");
167 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG,
168 	debug, CTLFLAG_RW,	&ipsec_debug,	0, "");
169 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_ESP_RANDPAD,
170 	esp_randpad, CTLFLAG_RW,	&ip4_esp_randpad,	0, "");
171 SYSCTL_INT(_net_inet_ipsec, OID_AUTO,
172 	crypto_support,	CTLFLAG_RW,	&crypto_support,0, "");
173 SYSCTL_STRUCT(_net_inet_ipsec, OID_AUTO,
174 	ipsecstats,	CTLFLAG_RD,	&newipsecstat,	newipsecstat, "");
175 #endif /* __FreeBSD__ */
176 
177 #ifdef INET6
178 int ip6_esp_trans_deflev = IPSEC_LEVEL_USE;
179 int ip6_esp_net_deflev = IPSEC_LEVEL_USE;
180 int ip6_ah_trans_deflev = IPSEC_LEVEL_USE;
181 int ip6_ah_net_deflev = IPSEC_LEVEL_USE;
182 struct secpolicy ip6_def_policy;
183 int ip6_ipsec_ecn = 0;		/* ECN ignore(-1)/forbidden(0)/allowed(1) */
184 int ip6_esp_randpad = -1;
185 
186 
187 #ifdef __FreeBSD__
188 SYSCTL_DECL(_net_inet6_ipsec6);
189 
190 /* net.inet6.ipsec6 */
191 #ifdef COMPAT_KAME
192 SYSCTL_OID(_net_inet6_ipsec6, IPSECCTL_STATS, stats, CTLFLAG_RD,
193 	0,0, compat_ipsecstats_sysctl, "S", "");
194 #endif /* COMPAT_KAME */
195 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_POLICY,
196 	def_policy, CTLFLAG_RW,	&ip4_def_policy.policy,	0, "");
197 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
198 	CTLFLAG_RW, &ip6_esp_trans_deflev,	0, "");
199 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
200 	CTLFLAG_RW, &ip6_esp_net_deflev,	0, "");
201 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
202 	CTLFLAG_RW, &ip6_ah_trans_deflev,	0, "");
203 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
204 	CTLFLAG_RW, &ip6_ah_net_deflev,	0, "");
205 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_ECN,
206 	ecn, CTLFLAG_RW,	&ip6_ipsec_ecn,	0, "");
207 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG,
208 	debug, CTLFLAG_RW,	&ipsec_debug,	0, "");
209 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_ESP_RANDPAD,
210 	esp_randpad, CTLFLAG_RW,	&ip6_esp_randpad,	0, "");
211 #endif /* INET6 */
212 #endif /* __FreeBSD__ */
213 
214 static int ipsec4_setspidx_inpcb __P((struct mbuf *, struct inpcb *pcb));
215 #ifdef INET6
216 static int ipsec6_setspidx_in6pcb __P((struct mbuf *, struct in6pcb *pcb));
217 #endif
218 static int ipsec_setspidx __P((struct mbuf *, struct secpolicyindex *, int));
219 static void ipsec4_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
220 static int ipsec4_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
221 #ifdef INET6
222 static void ipsec6_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
223 static int ipsec6_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
224 #endif
225 static void ipsec_delpcbpolicy __P((struct inpcbpolicy *));
226 static struct secpolicy *ipsec_deepcopy_policy __P((struct secpolicy *src));
227 static int ipsec_set_policy __P((struct secpolicy **pcb_sp,
228 	int optname, caddr_t request, size_t len, int priv));
229 static int ipsec_get_policy __P((struct secpolicy *pcb_sp, struct mbuf **mp));
230 static void vshiftl __P((unsigned char *, int, int));
231 static size_t ipsec_hdrsiz __P((struct secpolicy *));
232 
233 #ifdef __NetBSD__
234 /*
235  * Try to validate and use cached policy on a PCB.
236  */
237 static struct secpolicy *
238 ipsec_checkpcbcache(struct mbuf *m, struct inpcbpolicy *pcbsp, int dir)
239 {
240 	struct secpolicyindex spidx;
241 
242 	switch (dir) {
243 	case IPSEC_DIR_INBOUND:
244 	case IPSEC_DIR_OUTBOUND:
245 	case IPSEC_DIR_ANY:
246 		break;
247 	default:
248 		return NULL;
249 	}
250 #ifdef DIAGNOSTIC
251 	if (pcbsp == NULL) {
252 		printf("ipsec_checkpcbcache: NULL pcbsp\n");
253 		/* XXX panic? */
254 		return NULL;
255 	}
256 #endif
257 
258 #ifdef DIAGNOSTIC
259 	if (dir >= sizeof(pcbsp->sp_cache)/sizeof(pcbsp->sp_cache[0]))
260 		panic("dir too big in ipsec_checkpcbcache");
261 #endif
262 	/* SPD table change invalidate all the caches. */
263 	if (ipsec_spdgen != pcbsp->sp_cache[dir].cachegen) {
264 		ipsec_invalpcbcache(pcbsp, dir);
265 		return NULL;
266 	}
267 	if (!pcbsp->sp_cache[dir].cachesp)
268 		return NULL;
269 	if (pcbsp->sp_cache[dir].cachesp->state != IPSEC_SPSTATE_ALIVE) {
270 		ipsec_invalpcbcache(pcbsp, dir);
271 		return NULL;
272 	}
273 	if ((pcbsp->sp_cacheflags & IPSEC_PCBSP_CONNECTED) == 0) {
274 		if (!pcbsp->sp_cache[dir].cachesp)
275 			return NULL;
276 		if (ipsec_setspidx(m, &spidx, 1) != 0)
277 			return NULL;
278 		if (bcmp(&pcbsp->sp_cache[dir].cacheidx, &spidx,
279 			 sizeof(spidx))) {
280 			if (!key_cmpspidx_withmask(&pcbsp->sp_cache[dir].cachesp->spidx,
281 			    &spidx))
282 				return NULL;
283 			pcbsp->sp_cache[dir].cacheidx = spidx;
284 		}
285 	} else {
286 		/*
287 		 * The pcb is connected, and the L4 code is sure that:
288 		 * - outgoing side uses inp_[lf]addr
289 		 * - incoming side looks up policy after inpcb lookup
290 		 * and address pair is know to be stable.  We do not need
291 		 * to generate spidx again, nor check the address match again.
292 		 *
293 		 * For IPv4/v6 SOCK_STREAM sockets, this assumptions holds
294 		 * and there are calls to ipsec_pcbconn() from in_pcbconnect().
295 		 */
296 	}
297 
298 	pcbsp->sp_cache[dir].cachesp->lastused = mono_time.tv_sec;
299 	pcbsp->sp_cache[dir].cachesp->refcnt++;
300 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
301 		printf("DP ipsec_checkpcbcache cause refcnt++:%d SP:%p\n",
302 		pcbsp->sp_cache[dir].cachesp->refcnt,
303 		pcbsp->sp_cache[dir].cachesp));
304 	return pcbsp->sp_cache[dir].cachesp;
305 }
306 
307 static int
308 ipsec_fillpcbcache(struct inpcbpolicy *pcbsp, struct mbuf *m,
309     struct secpolicy *sp, int dir)
310 {
311 
312 	switch (dir) {
313 	case IPSEC_DIR_INBOUND:
314 	case IPSEC_DIR_OUTBOUND:
315 		break;
316 	default:
317 		return EINVAL;
318 	}
319 #ifdef DIAGNOSTIC
320 	if (dir >= sizeof(pcbsp->sp_cache)/sizeof(pcbsp->sp_cache[0]))
321 		panic("dir too big in ipsec_fillpcbcache");
322 #endif
323 
324 	if (pcbsp->sp_cache[dir].cachesp)
325 		KEY_FREESP(&pcbsp->sp_cache[dir].cachesp);
326 	pcbsp->sp_cache[dir].cachesp = NULL;
327 	pcbsp->sp_cache[dir].cachehint = IPSEC_PCBHINT_MAYBE;
328 	if (ipsec_setspidx(m, &pcbsp->sp_cache[dir].cacheidx, 1) != 0) {
329 		return EINVAL;
330 	}
331 	pcbsp->sp_cache[dir].cachesp = sp;
332 	if (pcbsp->sp_cache[dir].cachesp) {
333 		pcbsp->sp_cache[dir].cachesp->refcnt++;
334 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
335 			printf("DP ipsec_fillpcbcache cause refcnt++:%d SP:%p\n",
336 			pcbsp->sp_cache[dir].cachesp->refcnt,
337 			pcbsp->sp_cache[dir].cachesp));
338 
339 		/*
340 		 * If the PCB is connected, we can remember a hint to
341 		 * possibly short-circuit IPsec processing in other places.
342 		 */
343 		if (pcbsp->sp_cacheflags & IPSEC_PCBSP_CONNECTED) {
344 			switch (pcbsp->sp_cache[dir].cachesp->policy) {
345 			case IPSEC_POLICY_NONE:
346 			case IPSEC_POLICY_BYPASS:
347 				pcbsp->sp_cache[dir].cachehint =
348 				    IPSEC_PCBHINT_NO;
349 				break;
350 			default:
351 				pcbsp->sp_cache[dir].cachehint =
352 				    IPSEC_PCBHINT_YES;
353 			}
354 		}
355 	}
356 	pcbsp->sp_cache[dir].cachegen = ipsec_spdgen;
357 
358 	return 0;
359 }
360 
361 static int
362 ipsec_invalpcbcache(struct inpcbpolicy *pcbsp, int dir)
363 {
364 	int i;
365 
366 	for (i = IPSEC_DIR_INBOUND; i <= IPSEC_DIR_OUTBOUND; i++) {
367 		if (dir != IPSEC_DIR_ANY && i != dir)
368 			continue;
369 		if (pcbsp->sp_cache[i].cachesp)
370 			KEY_FREESP(&pcbsp->sp_cache[i].cachesp);
371 		pcbsp->sp_cache[i].cachesp = NULL;
372 		pcbsp->sp_cache[i].cachehint = IPSEC_PCBHINT_MAYBE;
373 		pcbsp->sp_cache[i].cachegen = 0;
374 		bzero(&pcbsp->sp_cache[i].cacheidx,
375 		      sizeof(pcbsp->sp_cache[i].cacheidx));
376 	}
377 	return 0;
378 }
379 
380 void
381 ipsec_pcbconn(struct inpcbpolicy *pcbsp)
382 {
383 
384 	pcbsp->sp_cacheflags |= IPSEC_PCBSP_CONNECTED;
385 	ipsec_invalpcbcache(pcbsp, IPSEC_DIR_ANY);
386 }
387 
388 void
389 ipsec_pcbdisconn(struct inpcbpolicy *pcbsp)
390 {
391 
392 	pcbsp->sp_cacheflags &= ~IPSEC_PCBSP_CONNECTED;
393 	ipsec_invalpcbcache(pcbsp, IPSEC_DIR_ANY);
394 }
395 
396 void
397 ipsec_invalpcbcacheall(void)
398 {
399 
400 	if (ipsec_spdgen == UINT_MAX)
401 		ipsec_spdgen = 1;
402 	else
403 		ipsec_spdgen++;
404 }
405 #endif /* __NetBSD__ */
406 
407 /*
408  * Return a held reference to the default SP.
409  */
410 static struct secpolicy *
411 key_allocsp_default(const char* where, int tag)
412 {
413 	struct secpolicy *sp;
414 
415 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
416 		printf("DP key_allocsp_default from %s:%u\n", where, tag));
417 
418 	sp = &ip4_def_policy;
419 	if (sp->policy != IPSEC_POLICY_DISCARD &&
420 	    sp->policy != IPSEC_POLICY_NONE) {
421 		ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n",
422 		    sp->policy, IPSEC_POLICY_NONE));
423 		sp->policy = IPSEC_POLICY_NONE;
424 	}
425 	sp->refcnt++;
426 
427 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
428 		printf("DP key_allocsp_default returns SP:%p (%u)\n",
429 			sp, sp->refcnt));
430 	return sp;
431 }
432 #define	KEY_ALLOCSP_DEFAULT() \
433 	key_allocsp_default(__FILE__, __LINE__)
434 
435 /*
436  * For OUTBOUND packet having a socket. Searching SPD for packet,
437  * and return a pointer to SP.
438  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
439  *		0	: bypass
440  *		EACCES	: discard packet.
441  *		ENOENT	: ipsec_acquire() in progress, maybe.
442  *		others	: error occurred.
443  *	others:	a pointer to SP
444  *
445  * NOTE: IPv6 mapped adddress concern is implemented here.
446  */
447 struct secpolicy *
448 ipsec_getpolicy(struct tdb_ident *tdbi, u_int dir)
449 {
450 	struct secpolicy *sp;
451 
452 	IPSEC_ASSERT(tdbi != NULL, ("ipsec_getpolicy: null tdbi"));
453 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
454 		("ipsec_getpolicy: invalid direction %u", dir));
455 
456 	sp = KEY_ALLOCSP2(tdbi->spi, &tdbi->dst, tdbi->proto, dir);
457 	if (sp == NULL)			/*XXX????*/
458 		sp = KEY_ALLOCSP_DEFAULT();
459 	IPSEC_ASSERT(sp != NULL, ("ipsec_getpolicy: null SP"));
460 	return sp;
461 }
462 
463 /*
464  * For OUTBOUND packet having a socket. Searching SPD for packet,
465  * and return a pointer to SP.
466  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
467  *		0	: bypass
468  *		EACCES	: discard packet.
469  *		ENOENT	: ipsec_acquire() in progress, maybe.
470  *		others	: error occurred.
471  *	others:	a pointer to SP
472  *
473  * NOTE: IPv6 mapped adddress concern is implemented here.
474  */
475 static struct secpolicy *
476 ipsec_getpolicybysock(m, dir, inp, error)
477 	struct mbuf *m;
478 	u_int dir;
479 	PCB_T *inp;
480 	int *error;
481 {
482 	struct inpcbpolicy *pcbsp = NULL;
483 	struct secpolicy *currsp = NULL;	/* policy on socket */
484 	struct secpolicy *sp;
485 	int af;
486 
487 	IPSEC_ASSERT(m != NULL, ("ipsec_getpolicybysock: null mbuf"));
488 	IPSEC_ASSERT(inp != NULL, ("ipsec_getpolicybysock: null inpcb"));
489 	IPSEC_ASSERT(error != NULL, ("ipsec_getpolicybysock: null error"));
490 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
491 		("ipsec_getpolicybysock: invalid direction %u", dir));
492 
493 	IPSEC_ASSERT(PCB_SOCKET(inp) != NULL,
494 	    ("ipsec_getppolicybysock: null socket\n"));
495 
496 	/* XXX FIXME inpcb/in6pcb  vs socket*/
497 	af = PCB_FAMILY(inp);
498 	IPSEC_ASSERT(af == AF_INET || af == AF_INET6,
499 		("ipsec_getpolicybysock: unexpected protocol family %u", af));
500 
501 #ifdef __NetBSD__
502 	IPSEC_ASSERT(inp->inph_sp != NULL, ("null PCB policy cache"));
503 	/* If we have a cached entry, and if it is still valid, use it. */
504 	ipsecstat.ips_spdcache_lookup++;
505 	currsp = ipsec_checkpcbcache(m, /*inpcb_hdr*/inp->inph_sp, dir);
506 	if (currsp) {
507 		*error = 0;
508 		return currsp;
509 	}
510 	ipsecstat.ips_spdcache_miss++;
511 #endif /* __NetBSD__ */
512 
513 	switch (af) {
514 	case AF_INET: {
515 		struct inpcb *in4p = PCB_TO_IN4PCB(inp);
516 		/* set spidx in pcb */
517 		*error = ipsec4_setspidx_inpcb(m, in4p);
518 		pcbsp = in4p->inp_sp;
519 		break;
520 		}
521 
522 #if defined(INET6)
523 	case AF_INET6: {
524 		struct in6pcb *in6p = PCB_TO_IN6PCB(inp);
525 		/* set spidx in pcb */
526 		*error = ipsec6_setspidx_in6pcb(m, in6p);
527 		pcbsp = in6p->in6p_sp;
528 		break;
529 		}
530 #endif
531 	default:
532 		*error = EPFNOSUPPORT;
533 		break;
534 	}
535 	if (*error)
536 		return NULL;
537 
538 	IPSEC_ASSERT(pcbsp != NULL, ("ipsec_getpolicybysock: null pcbsp"));
539 	switch (dir) {
540 	case IPSEC_DIR_INBOUND:
541 		currsp = pcbsp->sp_in;
542 		break;
543 	case IPSEC_DIR_OUTBOUND:
544 		currsp = pcbsp->sp_out;
545 		break;
546 	}
547 	IPSEC_ASSERT(currsp != NULL, ("ipsec_getpolicybysock: null currsp"));
548 
549 	if (pcbsp->priv) {			/* when privilieged socket */
550 		switch (currsp->policy) {
551 		case IPSEC_POLICY_BYPASS:
552 		case IPSEC_POLICY_IPSEC:
553 			currsp->refcnt++;
554 			sp = currsp;
555 			break;
556 
557 		case IPSEC_POLICY_ENTRUST:
558 			/* look for a policy in SPD */
559 			sp = KEY_ALLOCSP(&currsp->spidx, dir);
560 			if (sp == NULL)		/* no SP found */
561 				sp = KEY_ALLOCSP_DEFAULT();
562 			break;
563 
564 		default:
565 			ipseclog((LOG_ERR, "ipsec_getpolicybysock: "
566 			      "Invalid policy for PCB %d\n", currsp->policy));
567 			*error = EINVAL;
568 			return NULL;
569 		}
570 	} else {				/* unpriv, SPD has policy */
571 		sp = KEY_ALLOCSP(&currsp->spidx, dir);
572 		if (sp == NULL) {		/* no SP found */
573 			switch (currsp->policy) {
574 			case IPSEC_POLICY_BYPASS:
575 				ipseclog((LOG_ERR, "ipsec_getpolicybysock: "
576 				       "Illegal policy for non-priviliged defined %d\n",
577 					currsp->policy));
578 				*error = EINVAL;
579 				return NULL;
580 
581 			case IPSEC_POLICY_ENTRUST:
582 				sp = KEY_ALLOCSP_DEFAULT();
583 				break;
584 
585 			case IPSEC_POLICY_IPSEC:
586 				currsp->refcnt++;
587 				sp = currsp;
588 				break;
589 
590 			default:
591 				ipseclog((LOG_ERR, "ipsec_getpolicybysock: "
592 				   "Invalid policy for PCB %d\n", currsp->policy));
593 				*error = EINVAL;
594 				return NULL;
595 			}
596 		}
597 	}
598 	IPSEC_ASSERT(sp != NULL,
599 		("ipsec_getpolicybysock: null SP (priv %u policy %u",
600 		 pcbsp->priv, currsp->policy));
601 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
602 		printf("DP ipsec_getpolicybysock (priv %u policy %u) allocates "
603 		       "SP:%p (refcnt %u)\n", pcbsp->priv, currsp->policy,
604 		       sp, sp->refcnt));
605 #ifdef __NetBSD__
606 	ipsec_fillpcbcache(pcbsp, m, sp, dir);
607 #endif /* __NetBSD__ */
608 	return sp;
609 }
610 
611 /*
612  * For FORWADING packet or OUTBOUND without a socket. Searching SPD for packet,
613  * and return a pointer to SP.
614  * OUT:	positive: a pointer to the entry for security policy leaf matched.
615  *	NULL:	no apropreate SP found, the following value is set to error.
616  *		0	: bypass
617  *		EACCES	: discard packet.
618  *		ENOENT	: ipsec_acquire() in progress, maybe.
619  *		others	: error occurred.
620  */
621 struct secpolicy *
622 ipsec_getpolicybyaddr(m, dir, flag, error)
623 	struct mbuf *m;
624 	u_int dir;
625 	int flag;
626 	int *error;
627 {
628 	struct secpolicyindex spidx;
629 	struct secpolicy *sp;
630 
631 	IPSEC_ASSERT(m != NULL, ("ipsec_getpolicybyaddr: null mbuf"));
632 	IPSEC_ASSERT(error != NULL, ("ipsec_getpolicybyaddr: null error"));
633 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
634 		("ipsec4_getpolicybaddr: invalid direction %u", dir));
635 
636 	sp = NULL;
637 	if (key_havesp(dir)) {
638 		/* Make an index to look for a policy. */
639 		*error = ipsec_setspidx(m, &spidx,
640 					(flag & IP_FORWARDING) ? 0 : 1);
641 		if (*error != 0) {
642 			DPRINTF(("ipsec_getpolicybyaddr: setpidx failed,"
643 				" dir %u flag %u\n", dir, flag));
644 			bzero(&spidx, sizeof (spidx));
645 			return NULL;
646 		}
647 		spidx.dir = dir;
648 
649 		sp = KEY_ALLOCSP(&spidx, dir);
650 	}
651 	if (sp == NULL)			/* no SP found, use system default */
652 		sp = KEY_ALLOCSP_DEFAULT();
653 	IPSEC_ASSERT(sp != NULL, ("ipsec_getpolicybyaddr: null SP"));
654 	return sp;
655 }
656 
657 struct secpolicy *
658 ipsec4_checkpolicy(m, dir, flag, error, inp)
659 	struct mbuf *m;
660 	u_int dir, flag;
661 	int *error;
662 	struct inpcb *inp;
663 {
664 	struct secpolicy *sp;
665 
666 	*error = 0;
667 
668 
669 	/* XXX KAME IPv6 calls us with non-null inp but bogus inp_socket? */
670 	if (inp == NULL || inp->inp_socket == NULL) {
671 		sp = ipsec_getpolicybyaddr(m, dir, flag, error);
672 	} else
673 		sp = ipsec_getpolicybysock(m, dir, IN4PCB_TO_PCB(inp), error);
674 	if (sp == NULL) {
675 		IPSEC_ASSERT(*error != 0,
676 			("ipsec4_checkpolicy: getpolicy failed w/o error"));
677 		newipsecstat.ips_out_inval++;
678 		return NULL;
679 	}
680 	IPSEC_ASSERT(*error == 0,
681 		("ipsec4_checkpolicy: sp w/ error set to %u", *error));
682 	switch (sp->policy) {
683 	case IPSEC_POLICY_ENTRUST:
684 	default:
685 		printf("ipsec4_checkpolicy: invalid policy %u\n", sp->policy);
686 		/* fall thru... */
687 	case IPSEC_POLICY_DISCARD:
688 		newipsecstat.ips_out_polvio++;
689 		*error = -EINVAL;	/* packet is discarded by caller */
690 		break;
691 	case IPSEC_POLICY_BYPASS:
692 	case IPSEC_POLICY_NONE:
693 		KEY_FREESP(&sp);
694 		sp = NULL;		/* NB: force NULL result */
695 		break;
696 	case IPSEC_POLICY_IPSEC:
697 		if (sp->req == NULL)	/* acquire an SA */
698 			*error = key_spdacquire(sp);
699 		break;
700 	}
701 	if (*error != 0) {
702 		KEY_FREESP(&sp);
703 		sp = NULL;
704 	}
705 	DPRINTF(("ipsecpol: done, sp %p error %d, \n", sp, *error));
706 	return sp;
707 }
708 
709 static int
710 ipsec4_setspidx_inpcb(m, pcb)
711 	struct mbuf *m;
712 	struct inpcb *pcb;
713 {
714 	int error;
715 
716 	IPSEC_ASSERT(pcb != NULL, ("ipsec4_setspidx_inpcb: null pcb"));
717 	IPSEC_ASSERT(pcb->inp_sp != NULL, ("ipsec4_setspidx_inpcb: null inp_sp"));
718 	IPSEC_ASSERT(pcb->inp_sp->sp_out != NULL && pcb->inp_sp->sp_in != NULL,
719 		("ipsec4_setspidx_inpcb: null sp_in || sp_out"));
720 
721 	error = ipsec_setspidx(m, &pcb->inp_sp->sp_in->spidx, 1);
722 	if (error == 0) {
723 		pcb->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
724 		pcb->inp_sp->sp_out->spidx = pcb->inp_sp->sp_in->spidx;
725 		pcb->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
726 	} else {
727 		bzero(&pcb->inp_sp->sp_in->spidx,
728 			sizeof (pcb->inp_sp->sp_in->spidx));
729 		bzero(&pcb->inp_sp->sp_out->spidx,
730 			sizeof (pcb->inp_sp->sp_in->spidx));
731 	}
732 	return error;
733 }
734 
735 #ifdef INET6
736 static int
737 ipsec6_setspidx_in6pcb(m, pcb)
738 	struct mbuf *m;
739 	struct in6pcb *pcb;
740 {
741 	struct secpolicyindex *spidx;
742 	int error;
743 
744 	IPSEC_ASSERT(pcb != NULL, ("ipsec6_setspidx_in6pcb: null pcb"));
745 	IPSEC_ASSERT(pcb->in6p_sp != NULL, ("ipsec6_setspidx_in6pcb: null inp_sp"));
746 	IPSEC_ASSERT(pcb->in6p_sp->sp_out != NULL && pcb->in6p_sp->sp_in != NULL,
747 		("ipsec6_setspidx_in6pcb: null sp_in || sp_out"));
748 
749 	bzero(&pcb->in6p_sp->sp_in->spidx, sizeof(*spidx));
750 	bzero(&pcb->in6p_sp->sp_out->spidx, sizeof(*spidx));
751 
752 	spidx = &pcb->in6p_sp->sp_in->spidx;
753 	error = ipsec_setspidx(m, spidx, 1);
754 	if (error)
755 		goto bad;
756 	spidx->dir = IPSEC_DIR_INBOUND;
757 
758 	spidx = &pcb->in6p_sp->sp_out->spidx;
759 	error = ipsec_setspidx(m, spidx, 1);
760 	if (error)
761 		goto bad;
762 	spidx->dir = IPSEC_DIR_OUTBOUND;
763 
764 	return 0;
765 
766 bad:
767 	bzero(&pcb->in6p_sp->sp_in->spidx, sizeof(*spidx));
768 	bzero(&pcb->in6p_sp->sp_out->spidx, sizeof(*spidx));
769 	return error;
770 }
771 #endif
772 
773 /*
774  * configure security policy index (src/dst/proto/sport/dport)
775  * by looking at the content of mbuf.
776  * the caller is responsible for error recovery (like clearing up spidx).
777  */
778 static int
779 ipsec_setspidx(m, spidx, needport)
780 	struct mbuf *m;
781 	struct secpolicyindex *spidx;
782 	int needport;
783 {
784 	struct ip *ip = NULL;
785 	struct ip ipbuf;
786 	u_int v;
787 	struct mbuf *n;
788 	int len;
789 	int error;
790 
791 	IPSEC_ASSERT(m != NULL, ("ipsec_setspidx: null mbuf"));
792 
793 	/*
794 	 * validate m->m_pkthdr.len.  we see incorrect length if we
795 	 * mistakenly call this function with inconsistent mbuf chain
796 	 * (like 4.4BSD tcp/udp processing).  XXX should we panic here?
797 	 */
798 	len = 0;
799 	for (n = m; n; n = n->m_next)
800 		len += n->m_len;
801 	if (m->m_pkthdr.len != len) {
802 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
803 			printf("ipsec_setspidx: "
804 			       "total of m_len(%d) != pkthdr.len(%d), "
805 			       "ignored.\n",
806 				len, m->m_pkthdr.len));
807 		return EINVAL;
808 	}
809 
810 	if (m->m_pkthdr.len < sizeof(struct ip)) {
811 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
812 			printf("ipsec_setspidx: "
813 			    "pkthdr.len(%d) < sizeof(struct ip), ignored.\n",
814 			    m->m_pkthdr.len));
815 		return EINVAL;
816 	}
817 
818 	if (m->m_len >= sizeof(*ip))
819 		ip = mtod(m, struct ip *);
820 	else {
821 		m_copydata(m, 0, sizeof(ipbuf), (caddr_t)&ipbuf);
822 		ip = &ipbuf;
823 	}
824 #ifdef _IP_VHL
825 	v = _IP_VHL_V(ip->ip_vhl);
826 #else
827 	v = ip->ip_v;
828 #endif
829 	switch (v) {
830 	case 4:
831 		error = ipsec4_setspidx_ipaddr(m, spidx);
832 		if (error)
833 			return error;
834 		ipsec4_get_ulp(m, spidx, needport);
835 		return 0;
836 #ifdef INET6
837 	case 6:
838 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) {
839 			KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
840 				printf("ipsec_setspidx: "
841 				    "pkthdr.len(%d) < sizeof(struct ip6_hdr), "
842 				    "ignored.\n", m->m_pkthdr.len));
843 			return EINVAL;
844 		}
845 		error = ipsec6_setspidx_ipaddr(m, spidx);
846 		if (error)
847 			return error;
848 		ipsec6_get_ulp(m, spidx, needport);
849 		return 0;
850 #endif
851 	default:
852 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
853 			printf("ipsec_setspidx: "
854 			    "unknown IP version %u, ignored.\n", v));
855 		return EINVAL;
856 	}
857 }
858 
859 static void
860 ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
861 {
862 	u_int8_t nxt;
863 	int off;
864 
865 	/* sanity check */
866 	IPSEC_ASSERT(m != NULL, ("ipsec4_get_ulp: null mbuf"));
867 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),
868 		("ipsec4_get_ulp: packet too short"));
869 
870 	/* NB: ip_input() flips it into host endian XXX need more checking */
871 	if (m->m_len >= sizeof(struct ip)) {
872 		struct ip *ip = mtod(m, struct ip *);
873 		if (ip->ip_off & (IP_MF | IP_OFFMASK))
874 			goto done;
875 #ifdef _IP_VHL
876 		off = _IP_VHL_HL(ip->ip_vhl) << 2;
877 #else
878 		off = ip->ip_hl << 2;
879 #endif
880 		nxt = ip->ip_p;
881 	} else {
882 		struct ip ih;
883 
884 		m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih);
885 		if (ih.ip_off & (IP_MF | IP_OFFMASK))
886 			goto done;
887 #ifdef _IP_VHL
888 		off = _IP_VHL_HL(ih.ip_vhl) << 2;
889 #else
890 		off = ih.ip_hl << 2;
891 #endif
892 		nxt = ih.ip_p;
893 	}
894 
895 	while (off < m->m_pkthdr.len) {
896 		struct ip6_ext ip6e;
897 		struct tcphdr th;
898 		struct udphdr uh;
899 
900 		switch (nxt) {
901 		case IPPROTO_TCP:
902 			spidx->ul_proto = nxt;
903 			if (!needport)
904 				goto done_proto;
905 			if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
906 				goto done;
907 			m_copydata(m, off, sizeof (th), (caddr_t) &th);
908 			spidx->src.sin.sin_port = th.th_sport;
909 			spidx->dst.sin.sin_port = th.th_dport;
910 			return;
911 		case IPPROTO_UDP:
912 			spidx->ul_proto = nxt;
913 			if (!needport)
914 				goto done_proto;
915 			if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
916 				goto done;
917 			m_copydata(m, off, sizeof (uh), (caddr_t) &uh);
918 			spidx->src.sin.sin_port = uh.uh_sport;
919 			spidx->dst.sin.sin_port = uh.uh_dport;
920 			return;
921 		case IPPROTO_AH:
922 			if (m->m_pkthdr.len > off + sizeof(ip6e))
923 				goto done;
924 			/* XXX sigh, this works but is totally bogus */
925 			m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e);
926 			off += (ip6e.ip6e_len + 2) << 2;
927 			nxt = ip6e.ip6e_nxt;
928 			break;
929 		case IPPROTO_ICMP:
930 		default:
931 			/* XXX intermediate headers??? */
932 			spidx->ul_proto = nxt;
933 			goto done_proto;
934 		}
935 	}
936 done:
937 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
938 done_proto:
939 	spidx->src.sin.sin_port = IPSEC_PORT_ANY;
940 	spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
941 }
942 
943 /* assumes that m is sane */
944 static int
945 ipsec4_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
946 {
947 	static const struct sockaddr_in template = {
948 		sizeof (struct sockaddr_in),
949 		AF_INET,
950 		0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 }
951 	};
952 
953 	spidx->src.sin = template;
954 	spidx->dst.sin = template;
955 
956 	if (m->m_len < sizeof (struct ip)) {
957 		m_copydata(m, offsetof(struct ip, ip_src),
958 			   sizeof (struct  in_addr),
959 			   (caddr_t) &spidx->src.sin.sin_addr);
960 		m_copydata(m, offsetof(struct ip, ip_dst),
961 			   sizeof (struct  in_addr),
962 			   (caddr_t) &spidx->dst.sin.sin_addr);
963 	} else {
964 		struct ip *ip = mtod(m, struct ip *);
965 		spidx->src.sin.sin_addr = ip->ip_src;
966 		spidx->dst.sin.sin_addr = ip->ip_dst;
967 	}
968 
969 	spidx->prefs = sizeof(struct in_addr) << 3;
970 	spidx->prefd = sizeof(struct in_addr) << 3;
971 
972 	return 0;
973 }
974 
975 #ifdef INET6
976 static void
977 ipsec6_get_ulp(m, spidx, needport)
978 	struct mbuf *m;
979 	struct secpolicyindex *spidx;
980 	int needport;
981 {
982 	int off, nxt;
983 	struct tcphdr th;
984 	struct udphdr uh;
985 
986 	/* sanity check */
987 	if (m == NULL)
988 		panic("ipsec6_get_ulp: NULL pointer was passed");
989 
990 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
991 		printf("ipsec6_get_ulp:\n"); kdebug_mbuf(m));
992 
993 	/* set default */
994 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
995 	((struct sockaddr_in6 *)&spidx->src)->sin6_port = IPSEC_PORT_ANY;
996 	((struct sockaddr_in6 *)&spidx->dst)->sin6_port = IPSEC_PORT_ANY;
997 
998 	nxt = -1;
999 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
1000 	if (off < 0 || m->m_pkthdr.len < off)
1001 		return;
1002 
1003 	switch (nxt) {
1004 	case IPPROTO_TCP:
1005 		spidx->ul_proto = nxt;
1006 		if (!needport)
1007 			break;
1008 		if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
1009 			break;
1010 		m_copydata(m, off, sizeof(th), (caddr_t)&th);
1011 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = th.th_sport;
1012 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = th.th_dport;
1013 		break;
1014 	case IPPROTO_UDP:
1015 		spidx->ul_proto = nxt;
1016 		if (!needport)
1017 			break;
1018 		if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
1019 			break;
1020 		m_copydata(m, off, sizeof(uh), (caddr_t)&uh);
1021 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = uh.uh_sport;
1022 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = uh.uh_dport;
1023 		break;
1024 	case IPPROTO_ICMPV6:
1025 	default:
1026 		/* XXX intermediate headers??? */
1027 		spidx->ul_proto = nxt;
1028 		break;
1029 	}
1030 }
1031 
1032 /* assumes that m is sane */
1033 static int
1034 ipsec6_setspidx_ipaddr(m, spidx)
1035 	struct mbuf *m;
1036 	struct secpolicyindex *spidx;
1037 {
1038 	struct ip6_hdr *ip6 = NULL;
1039 	struct ip6_hdr ip6buf;
1040 	struct sockaddr_in6 *sin6;
1041 
1042 	if (m->m_len >= sizeof(*ip6))
1043 		ip6 = mtod(m, struct ip6_hdr *);
1044 	else {
1045 		m_copydata(m, 0, sizeof(ip6buf), (caddr_t)&ip6buf);
1046 		ip6 = &ip6buf;
1047 	}
1048 
1049 	sin6 = (struct sockaddr_in6 *)&spidx->src;
1050 	bzero(sin6, sizeof(*sin6));
1051 	sin6->sin6_family = AF_INET6;
1052 	sin6->sin6_len = sizeof(struct sockaddr_in6);
1053 	bcopy(&ip6->ip6_src, &sin6->sin6_addr, sizeof(ip6->ip6_src));
1054 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
1055 		sin6->sin6_addr.s6_addr16[1] = 0;
1056 		sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]);
1057 	}
1058 	spidx->prefs = sizeof(struct in6_addr) << 3;
1059 
1060 	sin6 = (struct sockaddr_in6 *)&spidx->dst;
1061 	bzero(sin6, sizeof(*sin6));
1062 	sin6->sin6_family = AF_INET6;
1063 	sin6->sin6_len = sizeof(struct sockaddr_in6);
1064 	bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(ip6->ip6_dst));
1065 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
1066 		sin6->sin6_addr.s6_addr16[1] = 0;
1067 		sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]);
1068 	}
1069 	spidx->prefd = sizeof(struct in6_addr) << 3;
1070 
1071 	return 0;
1072 }
1073 #endif
1074 
1075 static void
1076 ipsec_delpcbpolicy(p)
1077 	struct inpcbpolicy *p;
1078 {
1079 	free(p, M_SECA);
1080 }
1081 
1082 /* initialize policy in PCB */
1083 int
1084 ipsec_init_policy(so, pcb_sp)
1085 	struct socket *so;
1086 	struct inpcbpolicy **pcb_sp;
1087 {
1088 	struct inpcbpolicy *new;
1089 
1090 	/* sanity check. */
1091 	if (so == NULL || pcb_sp == NULL)
1092 		panic("ipsec_init_policy: NULL pointer was passed");
1093 
1094 	new = (struct inpcbpolicy *) malloc(sizeof(struct inpcbpolicy),
1095 					    M_SECA, M_NOWAIT|M_ZERO);
1096 	if (new == NULL) {
1097 		ipseclog((LOG_DEBUG, "ipsec_init_policy: No more memory.\n"));
1098 		return ENOBUFS;
1099 	}
1100 
1101 	if (IPSEC_PRIVILEGED_SO(so))
1102 		new->priv = 1;
1103 	else
1104 		new->priv = 0;
1105 
1106 	if ((new->sp_in = KEY_NEWSP()) == NULL) {
1107 		ipsec_delpcbpolicy(new);
1108 		return ENOBUFS;
1109 	}
1110 	new->sp_in->state = IPSEC_SPSTATE_ALIVE;
1111 	new->sp_in->policy = IPSEC_POLICY_ENTRUST;
1112 
1113 	if ((new->sp_out = KEY_NEWSP()) == NULL) {
1114 		KEY_FREESP(&new->sp_in);
1115 		ipsec_delpcbpolicy(new);
1116 		return ENOBUFS;
1117 	}
1118 	new->sp_out->state = IPSEC_SPSTATE_ALIVE;
1119 	new->sp_out->policy = IPSEC_POLICY_ENTRUST;
1120 
1121 	*pcb_sp = new;
1122 
1123 	return 0;
1124 }
1125 
1126 /* copy old ipsec policy into new */
1127 int
1128 ipsec_copy_policy(old, new)
1129 	struct inpcbpolicy *old, *new;
1130 {
1131 	struct secpolicy *sp;
1132 
1133 	sp = ipsec_deepcopy_policy(old->sp_in);
1134 	if (sp) {
1135 		KEY_FREESP(&new->sp_in);
1136 		new->sp_in = sp;
1137 	} else
1138 		return ENOBUFS;
1139 
1140 	sp = ipsec_deepcopy_policy(old->sp_out);
1141 	if (sp) {
1142 		KEY_FREESP(&new->sp_out);
1143 		new->sp_out = sp;
1144 	} else
1145 		return ENOBUFS;
1146 
1147 	new->priv = old->priv;
1148 
1149 	return 0;
1150 }
1151 
1152 /* deep-copy a policy in PCB */
1153 static struct secpolicy *
1154 ipsec_deepcopy_policy(src)
1155 	struct secpolicy *src;
1156 {
1157 	struct ipsecrequest *newchain = NULL;
1158 	struct ipsecrequest *p;
1159 	struct ipsecrequest **q;
1160 	struct ipsecrequest *r;
1161 	struct secpolicy *dst;
1162 
1163 	if (src == NULL)
1164 		return NULL;
1165 	dst = KEY_NEWSP();
1166 	if (dst == NULL)
1167 		return NULL;
1168 
1169 	/*
1170 	 * deep-copy IPsec request chain.  This is required since struct
1171 	 * ipsecrequest is not reference counted.
1172 	 */
1173 	q = &newchain;
1174 	for (p = src->req; p; p = p->next) {
1175 		*q = (struct ipsecrequest *)malloc(sizeof(struct ipsecrequest),
1176 			M_SECA, M_NOWAIT);
1177 		if (*q == NULL)
1178 			goto fail;
1179 		bzero(*q, sizeof(**q));
1180 		(*q)->next = NULL;
1181 
1182 		(*q)->saidx.proto = p->saidx.proto;
1183 		(*q)->saidx.mode = p->saidx.mode;
1184 		(*q)->level = p->level;
1185 		(*q)->saidx.reqid = p->saidx.reqid;
1186 
1187 		bcopy(&p->saidx.src, &(*q)->saidx.src, sizeof((*q)->saidx.src));
1188 		bcopy(&p->saidx.dst, &(*q)->saidx.dst, sizeof((*q)->saidx.dst));
1189 
1190 		(*q)->sav = NULL;
1191 		(*q)->sp = dst;
1192 
1193 		q = &((*q)->next);
1194 	}
1195 
1196 	dst->req = newchain;
1197 	dst->state = src->state;
1198 	dst->policy = src->policy;
1199 	/* do not touch the refcnt fields */
1200 
1201 	return dst;
1202 
1203 fail:
1204 	for (p = newchain; p; p = r) {
1205 		r = p->next;
1206 		free(p, M_SECA);
1207 		p = NULL;
1208 	}
1209 	return NULL;
1210 }
1211 
1212 /* set policy and ipsec request if present. */
1213 static int
1214 ipsec_set_policy(pcb_sp, optname, request, len, priv)
1215 	struct secpolicy **pcb_sp;
1216 	int optname;
1217 	caddr_t request;
1218 	size_t len;
1219 	int priv;
1220 {
1221 	struct sadb_x_policy *xpl;
1222 	struct secpolicy *newsp = NULL;
1223 	int error;
1224 
1225 	/* sanity check. */
1226 	if (pcb_sp == NULL || *pcb_sp == NULL || request == NULL)
1227 		return EINVAL;
1228 	if (len < sizeof(*xpl))
1229 		return EINVAL;
1230 	xpl = (struct sadb_x_policy *)request;
1231 
1232 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1233 		printf("ipsec_set_policy: passed policy\n");
1234 		kdebug_sadb_x_policy((struct sadb_ext *)xpl));
1235 
1236 	/* check policy type */
1237 	/* ipsec_set_policy() accepts IPSEC, ENTRUST and BYPASS. */
1238 	if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD
1239 	 || xpl->sadb_x_policy_type == IPSEC_POLICY_NONE)
1240 		return EINVAL;
1241 
1242 	/* check privileged socket */
1243 	if (priv == 0 && xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS)
1244 		return EACCES;
1245 
1246 	/* allocation new SP entry */
1247 	if ((newsp = key_msg2sp(xpl, len, &error)) == NULL)
1248 		return error;
1249 
1250 	newsp->state = IPSEC_SPSTATE_ALIVE;
1251 
1252 	/* clear old SP and set new SP */
1253 	KEY_FREESP(pcb_sp);
1254 	*pcb_sp = newsp;
1255 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1256 		printf("ipsec_set_policy: new policy\n");
1257 		kdebug_secpolicy(newsp));
1258 
1259 	return 0;
1260 }
1261 
1262 static int
1263 ipsec_get_policy(pcb_sp, mp)
1264 	struct secpolicy *pcb_sp;
1265 	struct mbuf **mp;
1266 {
1267 
1268 	/* sanity check. */
1269 	if (pcb_sp == NULL || mp == NULL)
1270 		return EINVAL;
1271 
1272 	*mp = key_sp2msg(pcb_sp);
1273 	if (!*mp) {
1274 		ipseclog((LOG_DEBUG, "ipsec_get_policy: No more memory.\n"));
1275 		return ENOBUFS;
1276 	}
1277 
1278 	(*mp)->m_type = MT_DATA;
1279 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1280 		printf("ipsec_get_policy:\n");
1281 		kdebug_mbuf(*mp));
1282 
1283 	return 0;
1284 }
1285 
1286 int
1287 ipsec4_set_policy(inp, optname, request, len, priv)
1288 	struct inpcb *inp;
1289 	int optname;
1290 	caddr_t request;
1291 	size_t len;
1292 	int priv;
1293 {
1294 	struct sadb_x_policy *xpl;
1295 	struct secpolicy **pcb_sp;
1296 
1297 	/* sanity check. */
1298 	if (inp == NULL || request == NULL)
1299 		return EINVAL;
1300 	if (len < sizeof(*xpl))
1301 		return EINVAL;
1302 	xpl = (struct sadb_x_policy *)request;
1303 
1304 	IPSEC_ASSERT(inp->inp_sp != NULL,
1305 		     ("ipsec4_set_policy(): null inp->in_sp"));
1306 
1307 	/* select direction */
1308 	switch (xpl->sadb_x_policy_dir) {
1309 	case IPSEC_DIR_INBOUND:
1310 		pcb_sp = &inp->inp_sp->sp_in;
1311 		break;
1312 	case IPSEC_DIR_OUTBOUND:
1313 		pcb_sp = &inp->inp_sp->sp_out;
1314 		break;
1315 	default:
1316 		ipseclog((LOG_ERR, "ipsec4_set_policy: invalid direction=%u\n",
1317 			xpl->sadb_x_policy_dir));
1318 		return EINVAL;
1319 	}
1320 
1321 	return ipsec_set_policy(pcb_sp, optname, request, len, priv);
1322 }
1323 
1324 int
1325 ipsec4_get_policy(inp, request, len, mp)
1326 	struct inpcb *inp;
1327 	caddr_t request;
1328 	size_t len;
1329 	struct mbuf **mp;
1330 {
1331 	struct sadb_x_policy *xpl;
1332 	struct secpolicy *pcb_sp;
1333 
1334 	/* sanity check. */
1335 	if (inp == NULL || request == NULL || mp == NULL)
1336 		return EINVAL;
1337 	IPSEC_ASSERT(inp->inp_sp != NULL, ("ipsec4_get_policy: null inp_sp"));
1338 	if (len < sizeof(*xpl))
1339 		return EINVAL;
1340 	xpl = (struct sadb_x_policy *)request;
1341 
1342 	/* select direction */
1343 	switch (xpl->sadb_x_policy_dir) {
1344 	case IPSEC_DIR_INBOUND:
1345 		pcb_sp = inp->inp_sp->sp_in;
1346 		break;
1347 	case IPSEC_DIR_OUTBOUND:
1348 		pcb_sp = inp->inp_sp->sp_out;
1349 		break;
1350 	default:
1351 		ipseclog((LOG_ERR, "ipsec4_set_policy: invalid direction=%u\n",
1352 			xpl->sadb_x_policy_dir));
1353 		return EINVAL;
1354 	}
1355 
1356 	return ipsec_get_policy(pcb_sp, mp);
1357 }
1358 
1359 /* delete policy in PCB */
1360 int
1361 ipsec4_delete_pcbpolicy(inp)
1362 	struct inpcb *inp;
1363 {
1364 	IPSEC_ASSERT(inp != NULL, ("ipsec4_delete_pcbpolicy: null inp"));
1365 
1366 	if (inp->inp_sp == NULL)
1367 		return 0;
1368 
1369 	if (inp->inp_sp->sp_in != NULL)
1370 		KEY_FREESP(&inp->inp_sp->sp_in);
1371 
1372 	if (inp->inp_sp->sp_out != NULL)
1373 		KEY_FREESP(&inp->inp_sp->sp_out);
1374 
1375 	ipsec_delpcbpolicy(inp->inp_sp);
1376 	inp->inp_sp = NULL;
1377 
1378 	return 0;
1379 }
1380 
1381 #ifdef INET6
1382 int
1383 ipsec6_set_policy(in6p, optname, request, len, priv)
1384 	struct in6pcb *in6p;
1385 	int optname;
1386 	caddr_t request;
1387 	size_t len;
1388 	int priv;
1389 {
1390 	struct sadb_x_policy *xpl;
1391 	struct secpolicy **pcb_sp;
1392 
1393 	/* sanity check. */
1394 	if (in6p == NULL || request == NULL)
1395 		return EINVAL;
1396 	if (len < sizeof(*xpl))
1397 		return EINVAL;
1398 	xpl = (struct sadb_x_policy *)request;
1399 
1400 	/* select direction */
1401 	switch (xpl->sadb_x_policy_dir) {
1402 	case IPSEC_DIR_INBOUND:
1403 		pcb_sp = &in6p->in6p_sp->sp_in;
1404 		break;
1405 	case IPSEC_DIR_OUTBOUND:
1406 		pcb_sp = &in6p->in6p_sp->sp_out;
1407 		break;
1408 	default:
1409 		ipseclog((LOG_ERR, "ipsec6_set_policy: invalid direction=%u\n",
1410 			xpl->sadb_x_policy_dir));
1411 		return EINVAL;
1412 	}
1413 
1414 	return ipsec_set_policy(pcb_sp, optname, request, len, priv);
1415 }
1416 
1417 int
1418 ipsec6_get_policy(in6p, request, len, mp)
1419 	struct in6pcb *in6p;
1420 	caddr_t request;
1421 	size_t len;
1422 	struct mbuf **mp;
1423 {
1424 	struct sadb_x_policy *xpl;
1425 	struct secpolicy *pcb_sp;
1426 
1427 	/* sanity check. */
1428 	if (in6p == NULL || request == NULL || mp == NULL)
1429 		return EINVAL;
1430 	IPSEC_ASSERT(in6p->in6p_sp != NULL, ("ipsec6_get_policy: null in6p_sp"));
1431 	if (len < sizeof(*xpl))
1432 		return EINVAL;
1433 	xpl = (struct sadb_x_policy *)request;
1434 
1435 	/* select direction */
1436 	switch (xpl->sadb_x_policy_dir) {
1437 	case IPSEC_DIR_INBOUND:
1438 		pcb_sp = in6p->in6p_sp->sp_in;
1439 		break;
1440 	case IPSEC_DIR_OUTBOUND:
1441 		pcb_sp = in6p->in6p_sp->sp_out;
1442 		break;
1443 	default:
1444 		ipseclog((LOG_ERR, "ipsec6_set_policy: invalid direction=%u\n",
1445 			xpl->sadb_x_policy_dir));
1446 		return EINVAL;
1447 	}
1448 
1449 	return ipsec_get_policy(pcb_sp, mp);
1450 }
1451 
1452 int
1453 ipsec6_delete_pcbpolicy(in6p)
1454 	struct in6pcb *in6p;
1455 {
1456 	IPSEC_ASSERT(in6p != NULL, ("ipsec6_delete_pcbpolicy: null in6p"));
1457 
1458 	if (in6p->in6p_sp == NULL)
1459 		return 0;
1460 
1461 	if (in6p->in6p_sp->sp_in != NULL)
1462 		KEY_FREESP(&in6p->in6p_sp->sp_in);
1463 
1464 	if (in6p->in6p_sp->sp_out != NULL)
1465 		KEY_FREESP(&in6p->in6p_sp->sp_out);
1466 
1467 	ipsec_delpcbpolicy(in6p->in6p_sp);
1468 	in6p->in6p_sp = NULL;
1469 
1470 	return 0;
1471 }
1472 #endif
1473 
1474 /*
1475  * return current level.
1476  * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
1477  */
1478 u_int
1479 ipsec_get_reqlevel(isr)
1480 	struct ipsecrequest *isr;
1481 {
1482 	u_int level = 0;
1483 	u_int esp_trans_deflev, esp_net_deflev;
1484 	u_int ah_trans_deflev, ah_net_deflev;
1485 
1486 	IPSEC_ASSERT(isr != NULL && isr->sp != NULL,
1487 		("ipsec_get_reqlevel: null argument"));
1488 	IPSEC_ASSERT(isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family,
1489 		("ipsec_get_reqlevel: af family mismatch, src %u, dst %u",
1490 		 isr->sp->spidx.src.sa.sa_family,
1491 		 isr->sp->spidx.dst.sa.sa_family));
1492 
1493 /* XXX note that we have ipseclog() expanded here - code sync issue */
1494 #define IPSEC_CHECK_DEFAULT(lev) \
1495 	(((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE	      \
1496 			&& (lev) != IPSEC_LEVEL_UNIQUE)			      \
1497 		? (ipsec_debug						      \
1498 			? log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
1499 				(lev), IPSEC_LEVEL_REQUIRE)		      \
1500 			: 0),						      \
1501 			(lev) = IPSEC_LEVEL_REQUIRE,			      \
1502 			(lev)						      \
1503 		: (lev))
1504 
1505 	/* set default level */
1506 	switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) {
1507 #ifdef INET
1508 	case AF_INET:
1509 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(ip4_esp_trans_deflev);
1510 		esp_net_deflev = IPSEC_CHECK_DEFAULT(ip4_esp_net_deflev);
1511 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(ip4_ah_trans_deflev);
1512 		ah_net_deflev = IPSEC_CHECK_DEFAULT(ip4_ah_net_deflev);
1513 		break;
1514 #endif
1515 #ifdef INET6
1516 	case AF_INET6:
1517 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(ip6_esp_trans_deflev);
1518 		esp_net_deflev = IPSEC_CHECK_DEFAULT(ip6_esp_net_deflev);
1519 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(ip6_ah_trans_deflev);
1520 		ah_net_deflev = IPSEC_CHECK_DEFAULT(ip6_ah_net_deflev);
1521 		break;
1522 #endif /* INET6 */
1523 	default:
1524 		panic("key_get_reqlevel: unknown af %u",
1525 		    isr->sp->spidx.src.sa.sa_family);
1526 	}
1527 
1528 #undef IPSEC_CHECK_DEFAULT
1529 
1530 	/* set level */
1531 	switch (isr->level) {
1532 	case IPSEC_LEVEL_DEFAULT:
1533 		switch (isr->saidx.proto) {
1534 		case IPPROTO_ESP:
1535 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1536 				level = esp_net_deflev;
1537 			else
1538 				level = esp_trans_deflev;
1539 			break;
1540 		case IPPROTO_AH:
1541 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1542 				level = ah_net_deflev;
1543 			else
1544 				level = ah_trans_deflev;
1545 			break;
1546 		case IPPROTO_IPCOMP:
1547 			/*
1548 			 * we don't really care, as IPcomp document says that
1549 			 * we shouldn't compress small packets
1550 			 */
1551 			level = IPSEC_LEVEL_USE;
1552 			break;
1553 		default:
1554 			panic("ipsec_get_reqlevel: Illegal protocol defined %u",
1555 			    isr->saidx.proto);
1556 		}
1557 		break;
1558 
1559 	case IPSEC_LEVEL_USE:
1560 	case IPSEC_LEVEL_REQUIRE:
1561 		level = isr->level;
1562 		break;
1563 	case IPSEC_LEVEL_UNIQUE:
1564 		level = IPSEC_LEVEL_REQUIRE;
1565 		break;
1566 
1567 	default:
1568 		panic("ipsec_get_reqlevel: Illegal IPsec level %u",
1569 			isr->level);
1570 	}
1571 
1572 	return level;
1573 }
1574 
1575 /*
1576  * Check security policy requirements against the actual
1577  * packet contents.  Return one if the packet should be
1578  * reject as "invalid"; otherwiser return zero to have the
1579  * packet treated as "valid".
1580  *
1581  * OUT:
1582  *	0: valid
1583  *	1: invalid
1584  */
1585 int
1586 ipsec_in_reject(struct secpolicy *sp, struct mbuf *m)
1587 {
1588 	struct ipsecrequest *isr;
1589 	int need_auth;
1590 
1591 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1592 		printf("ipsec_in_reject: using SP\n");
1593 		kdebug_secpolicy(sp));
1594 
1595 	/* check policy */
1596 	switch (sp->policy) {
1597 	case IPSEC_POLICY_DISCARD:
1598 		return 1;
1599 	case IPSEC_POLICY_BYPASS:
1600 	case IPSEC_POLICY_NONE:
1601 		return 0;
1602 	}
1603 
1604 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1605 		("ipsec_in_reject: invalid policy %u", sp->policy));
1606 
1607 	/* XXX should compare policy against ipsec header history */
1608 
1609 	need_auth = 0;
1610 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1611 		if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE)
1612 			continue;
1613 		switch (isr->saidx.proto) {
1614 		case IPPROTO_ESP:
1615 			if ((m->m_flags & M_DECRYPTED) == 0) {
1616 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1617 				    printf("ipsec_in_reject: ESP m_flags:%x\n",
1618 					    m->m_flags));
1619 				return 1;
1620 			}
1621 
1622 			if (!need_auth &&
1623 			    isr->sav != NULL &&
1624 			    isr->sav->tdb_authalgxform != NULL &&
1625 			    (m->m_flags & M_AUTHIPDGM) == 0) {
1626 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1627 				    printf("ipsec_in_reject: ESP/AH m_flags:%x\n",
1628 					    m->m_flags));
1629 				return 1;
1630 			}
1631 			break;
1632 		case IPPROTO_AH:
1633 			need_auth = 1;
1634 			if ((m->m_flags & M_AUTHIPHDR) == 0) {
1635 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1636 				    printf("ipsec_in_reject: AH m_flags:%x\n",
1637 					    m->m_flags));
1638 				return 1;
1639 			}
1640 			break;
1641 		case IPPROTO_IPCOMP:
1642 			/*
1643 			 * we don't really care, as IPcomp document
1644 			 * says that we shouldn't compress small
1645 			 * packets, IPComp policy should always be
1646 			 * treated as being in "use" level.
1647 			 */
1648 			break;
1649 		}
1650 	}
1651 	return 0;		/* valid */
1652 }
1653 
1654 /*
1655  * Check AH/ESP integrity.
1656  * This function is called from tcp_input(), udp_input(),
1657  * and {ah,esp}4_input for tunnel mode
1658  */
1659 int
1660 ipsec4_in_reject(m, inp)
1661 	struct mbuf *m;
1662 	struct inpcb *inp;
1663 {
1664 	struct secpolicy *sp;
1665 	int error;
1666 	int result;
1667 
1668 	IPSEC_ASSERT(m != NULL, ("ipsec4_in_reject_so: null mbuf"));
1669 
1670 	/* get SP for this packet.
1671 	 * When we are called from ip_forward(), we call
1672 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1673 	 */
1674 	if (inp == NULL)
1675 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, IP_FORWARDING, &error);
1676 	else
1677 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND,
1678 					   IN4PCB_TO_PCB(inp), &error);
1679 
1680 	if (sp != NULL) {
1681 		result = ipsec_in_reject(sp, m);
1682 		if (result)
1683 			newipsecstat.ips_in_polvio++;
1684 		KEY_FREESP(&sp);
1685 	} else {
1686 		result = 0;	/* XXX should be panic ?
1687 				 * -> No, there may be error. */
1688 	}
1689 	return result;
1690 }
1691 
1692 
1693 #ifdef INET6
1694 /*
1695  * Check AH/ESP integrity.
1696  * This function is called from tcp6_input(), udp6_input(),
1697  * and {ah,esp}6_input for tunnel mode
1698  */
1699 int
1700 ipsec6_in_reject(m, in6p)
1701 	struct mbuf *m;
1702 	struct in6pcb *in6p;
1703 {
1704 	struct secpolicy *sp = NULL;
1705 	int error;
1706 	int result;
1707 
1708 	/* sanity check */
1709 	if (m == NULL)
1710 		return 0;	/* XXX should be panic ? */
1711 
1712 	/* get SP for this packet.
1713 	 * When we are called from ip_forward(), we call
1714 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1715 	 */
1716 	if (in6p == NULL)
1717 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, IP_FORWARDING, &error);
1718 	else
1719 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND,
1720 			IN6PCB_TO_PCB(in6p),
1721 			&error);
1722 
1723 	if (sp != NULL) {
1724 		result = ipsec_in_reject(sp, m);
1725 		if (result)
1726 			newipsecstat.ips_in_polvio++;
1727 		KEY_FREESP(&sp);
1728 	} else {
1729 		result = 0;
1730 	}
1731 	return result;
1732 }
1733 #endif
1734 
1735 /*
1736  * compute the byte size to be occupied by IPsec header.
1737  * in case it is tunneled, it includes the size of outer IP header.
1738  * NOTE: SP passed is free in this function.
1739  */
1740 static size_t
1741 ipsec_hdrsiz(struct secpolicy *sp)
1742 {
1743 	struct ipsecrequest *isr;
1744 	size_t siz;
1745 
1746 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1747 		printf("ipsec_hdrsiz: using SP\n");
1748 		kdebug_secpolicy(sp));
1749 
1750 	switch (sp->policy) {
1751 	case IPSEC_POLICY_DISCARD:
1752 	case IPSEC_POLICY_BYPASS:
1753 	case IPSEC_POLICY_NONE:
1754 		return 0;
1755 	}
1756 
1757 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1758 		("ipsec_hdrsiz: invalid policy %u", sp->policy));
1759 
1760 	siz = 0;
1761 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1762 		size_t clen = 0;
1763 
1764 		switch (isr->saidx.proto) {
1765 		case IPPROTO_ESP:
1766 			clen = esp_hdrsiz(isr->sav);
1767 			break;
1768 		case IPPROTO_AH:
1769 			clen = ah_hdrsiz(isr->sav);
1770 			break;
1771 		case IPPROTO_IPCOMP:
1772 			clen = sizeof(struct ipcomp);
1773 			break;
1774 		}
1775 
1776 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1777 			switch (isr->saidx.dst.sa.sa_family) {
1778 			case AF_INET:
1779 				clen += sizeof(struct ip);
1780 				break;
1781 #ifdef INET6
1782 			case AF_INET6:
1783 				clen += sizeof(struct ip6_hdr);
1784 				break;
1785 #endif
1786 			default:
1787 				ipseclog((LOG_ERR, "ipsec_hdrsiz: "
1788 				    "unknown AF %d in IPsec tunnel SA\n",
1789 				    ((struct sockaddr *)&isr->saidx.dst)->sa_family));
1790 				break;
1791 			}
1792 		}
1793 		siz += clen;
1794 	}
1795 
1796 	return siz;
1797 }
1798 
1799 /* This function is called from ip_forward() and ipsec4_hdrsize_tcp(). */
1800 size_t
1801 ipsec4_hdrsiz(m, dir, inp)
1802 	struct mbuf *m;
1803 	u_int dir;
1804 	struct inpcb *inp;
1805 {
1806 	struct secpolicy *sp;
1807 	int error;
1808 	size_t size;
1809 
1810 	IPSEC_ASSERT(m != NULL, ("ipsec4_hdrsiz: null mbuf"));
1811 	IPSEC_ASSERT(inp == NULL || inp->inp_socket != NULL,
1812 		("ipsec4_hdrsize: socket w/o inpcb"));
1813 
1814 	/* get SP for this packet.
1815 	 * When we are called from ip_forward(), we call
1816 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1817 	 */
1818 	if (inp == NULL)
1819 		sp = ipsec_getpolicybyaddr(m, dir, IP_FORWARDING, &error);
1820 	else
1821 		sp = ipsec_getpolicybysock(m, dir,
1822 					   IN4PCB_TO_PCB(inp), &error);
1823 
1824 	if (sp != NULL) {
1825 		size = ipsec_hdrsiz(sp);
1826 		KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1827 			printf("ipsec4_hdrsiz: size:%lu.\n",
1828 				(unsigned long)size));
1829 
1830 		KEY_FREESP(&sp);
1831 	} else {
1832 		size = 0;	/* XXX should be panic ? */
1833 	}
1834 	return size;
1835 }
1836 
1837 #ifdef INET6
1838 /* This function is called from ipsec6_hdrsize_tcp(),
1839  * and maybe from ip6_forward.()
1840  */
1841 size_t
1842 ipsec6_hdrsiz(m, dir, in6p)
1843 	struct mbuf *m;
1844 	u_int dir;
1845 	struct in6pcb *in6p;
1846 {
1847 	struct secpolicy *sp;
1848 	int error;
1849 	size_t size;
1850 
1851 	IPSEC_ASSERT(m != NULL, ("ipsec6_hdrsiz: null mbuf"));
1852 	IPSEC_ASSERT(in6p == NULL || in6p->in6p_socket != NULL,
1853 		("ipsec6_hdrsize: socket w/o inpcb"));
1854 
1855 	/* get SP for this packet */
1856 	/* XXX Is it right to call with IP_FORWARDING. */
1857 	if (in6p == NULL)
1858 		sp = ipsec_getpolicybyaddr(m, dir, IP_FORWARDING, &error);
1859 	else
1860 		sp = ipsec_getpolicybysock(m, dir,
1861 			IN6PCB_TO_PCB(in6p),
1862 			&error);
1863 
1864 	if (sp == NULL)
1865 		return 0;
1866 	size = ipsec_hdrsiz(sp);
1867 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1868 		printf("ipsec6_hdrsiz: size:%lu.\n", (unsigned long)size));
1869 	KEY_FREESP(&sp);
1870 
1871 	return size;
1872 }
1873 #endif /*INET6*/
1874 
1875 /*
1876  * Check the variable replay window.
1877  * ipsec_chkreplay() performs replay check before ICV verification.
1878  * ipsec_updatereplay() updates replay bitmap.  This must be called after
1879  * ICV verification (it also performs replay check, which is usually done
1880  * beforehand).
1881  * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1882  *
1883  * based on RFC 2401.
1884  */
1885 int
1886 ipsec_chkreplay(seq, sav)
1887 	u_int32_t seq;
1888 	struct secasvar *sav;
1889 {
1890 	const struct secreplay *replay;
1891 	u_int32_t diff;
1892 	int fr;
1893 	u_int32_t wsizeb;	/* constant: bits of window size */
1894 	int frlast;		/* constant: last frame */
1895 
1896 	IPSEC_SPLASSERT_SOFTNET("ipsec_chkreplay");
1897 
1898 	IPSEC_ASSERT(sav != NULL, ("ipsec_chkreplay: Null SA"));
1899 	IPSEC_ASSERT(sav->replay != NULL, ("ipsec_chkreplay: Null replay state"));
1900 
1901 	replay = sav->replay;
1902 
1903 	if (replay->wsize == 0)
1904 		return 1;	/* no need to check replay. */
1905 
1906 	/* constant */
1907 	frlast = replay->wsize - 1;
1908 	wsizeb = replay->wsize << 3;
1909 
1910 	/* sequence number of 0 is invalid */
1911 	if (seq == 0)
1912 		return 0;
1913 
1914 	/* first time is always okay */
1915 	if (replay->count == 0)
1916 		return 1;
1917 
1918 	if (seq > replay->lastseq) {
1919 		/* larger sequences are okay */
1920 		return 1;
1921 	} else {
1922 		/* seq is equal or less than lastseq. */
1923 		diff = replay->lastseq - seq;
1924 
1925 		/* over range to check, i.e. too old or wrapped */
1926 		if (diff >= wsizeb)
1927 			return 0;
1928 
1929 		fr = frlast - diff / 8;
1930 
1931 		/* this packet already seen ? */
1932 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1933 			return 0;
1934 
1935 		/* out of order but good */
1936 		return 1;
1937 	}
1938 }
1939 
1940 /*
1941  * check replay counter whether to update or not.
1942  * OUT:	0:	OK
1943  *	1:	NG
1944  */
1945 int
1946 ipsec_updatereplay(seq, sav)
1947 	u_int32_t seq;
1948 	struct secasvar *sav;
1949 {
1950 	struct secreplay *replay;
1951 	u_int32_t diff;
1952 	int fr;
1953 	u_int32_t wsizeb;	/* constant: bits of window size */
1954 	int frlast;		/* constant: last frame */
1955 
1956 	IPSEC_SPLASSERT_SOFTNET("ipsec_updatereplay");
1957 
1958 	IPSEC_ASSERT(sav != NULL, ("ipsec_updatereplay: Null SA"));
1959 	IPSEC_ASSERT(sav->replay != NULL, ("ipsec_updatereplay: Null replay state"));
1960 
1961 	replay = sav->replay;
1962 
1963 	if (replay->wsize == 0)
1964 		goto ok;	/* no need to check replay. */
1965 
1966 	/* constant */
1967 	frlast = replay->wsize - 1;
1968 	wsizeb = replay->wsize << 3;
1969 
1970 	/* sequence number of 0 is invalid */
1971 	if (seq == 0)
1972 		return 1;
1973 
1974 	/* first time */
1975 	if (replay->count == 0) {
1976 		replay->lastseq = seq;
1977 		bzero(replay->bitmap, replay->wsize);
1978 		(replay->bitmap)[frlast] = 1;
1979 		goto ok;
1980 	}
1981 
1982 	if (seq > replay->lastseq) {
1983 		/* seq is larger than lastseq. */
1984 		diff = seq - replay->lastseq;
1985 
1986 		/* new larger sequence number */
1987 		if (diff < wsizeb) {
1988 			/* In window */
1989 			/* set bit for this packet */
1990 			vshiftl(replay->bitmap, diff, replay->wsize);
1991 			(replay->bitmap)[frlast] |= 1;
1992 		} else {
1993 			/* this packet has a "way larger" */
1994 			bzero(replay->bitmap, replay->wsize);
1995 			(replay->bitmap)[frlast] = 1;
1996 		}
1997 		replay->lastseq = seq;
1998 
1999 		/* larger is good */
2000 	} else {
2001 		/* seq is equal or less than lastseq. */
2002 		diff = replay->lastseq - seq;
2003 
2004 		/* over range to check, i.e. too old or wrapped */
2005 		if (diff >= wsizeb)
2006 			return 1;
2007 
2008 		fr = frlast - diff / 8;
2009 
2010 		/* this packet already seen ? */
2011 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
2012 			return 1;
2013 
2014 		/* mark as seen */
2015 		(replay->bitmap)[fr] |= (1 << (diff % 8));
2016 
2017 		/* out of order but good */
2018 	}
2019 
2020 ok:
2021 	if (replay->count == ~0) {
2022 
2023 		/* set overflow flag */
2024 		replay->overflow++;
2025 
2026 		/* don't increment, no more packets accepted */
2027 		if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0)
2028 			return 1;
2029 
2030 		ipseclog((LOG_WARNING, "replay counter made %d cycle. %s\n",
2031 		    replay->overflow, ipsec_logsastr(sav)));
2032 	}
2033 
2034 	replay->count++;
2035 
2036 	return 0;
2037 }
2038 
2039 /*
2040  * shift variable length bunffer to left.
2041  * IN:	bitmap: pointer to the buffer
2042  * 	nbit:	the number of to shift.
2043  *	wsize:	buffer size (bytes).
2044  */
2045 static void
2046 vshiftl(bitmap, nbit, wsize)
2047 	unsigned char *bitmap;
2048 	int nbit, wsize;
2049 {
2050 	int s, j, i;
2051 	unsigned char over;
2052 
2053 	for (j = 0; j < nbit; j += 8) {
2054 		s = (nbit - j < 8) ? (nbit - j): 8;
2055 		bitmap[0] <<= s;
2056 		for (i = 1; i < wsize; i++) {
2057 			over = (bitmap[i] >> (8 - s));
2058 			bitmap[i] <<= s;
2059 			bitmap[i-1] |= over;
2060 		}
2061 	}
2062 
2063 	return;
2064 }
2065 
2066 /* Return a printable string for the IPv4 address. */
2067 static char *
2068 inet_ntoa4(struct in_addr ina)
2069 {
2070 	static char buf[4][4 * sizeof "123" + 4];
2071 	unsigned char *ucp = (unsigned char *) &ina;
2072 	static int i = 3;
2073 
2074 	i = (i + 1) % 4;
2075 	snprintf(buf[i], sizeof(buf[i]), "%d.%d.%d.%d",
2076 	    ucp[0] & 0xff, ucp[1] & 0xff, ucp[2] & 0xff, ucp[3] & 0xff);
2077 	return (buf[i]);
2078 }
2079 
2080 /* Return a printable string for the address. */
2081 const char *
2082 ipsec_address(union sockaddr_union* sa)
2083 {
2084 	switch (sa->sa.sa_family) {
2085 #if INET
2086 	case AF_INET:
2087 		return inet_ntoa4(sa->sin.sin_addr);
2088 #endif /* INET */
2089 
2090 #if INET6
2091 	case AF_INET6:
2092 		return ip6_sprintf(&sa->sin6.sin6_addr);
2093 #endif /* INET6 */
2094 
2095 	default:
2096 		return "(unknown address family)";
2097 	}
2098 }
2099 
2100 const char *
2101 ipsec_logsastr(sav)
2102 	struct secasvar *sav;
2103 {
2104 	static char buf[256];
2105 	char *p;
2106 	struct secasindex *saidx = &sav->sah->saidx;
2107 
2108 	IPSEC_ASSERT(saidx->src.sa.sa_family == saidx->dst.sa.sa_family,
2109 		("ipsec_logsastr: address family mismatch"));
2110 
2111 	p = buf;
2112 	snprintf(buf, sizeof(buf), "SA(SPI=%u ", (u_int32_t)ntohl(sav->spi));
2113 	while (p && *p)
2114 		p++;
2115 	/* NB: only use ipsec_address on one address at a time */
2116 	snprintf(p, sizeof (buf) - (p - buf), "src=%s ",
2117 		ipsec_address(&saidx->src));
2118 	while (p && *p)
2119 		p++;
2120 	snprintf(p, sizeof (buf) - (p - buf), "dst=%s)",
2121 		ipsec_address(&saidx->dst));
2122 
2123 	return buf;
2124 }
2125 
2126 void
2127 ipsec_dumpmbuf(m)
2128 	struct mbuf *m;
2129 {
2130 	int totlen;
2131 	int i;
2132 	u_char *p;
2133 
2134 	totlen = 0;
2135 	printf("---\n");
2136 	while (m) {
2137 		p = mtod(m, u_char *);
2138 		for (i = 0; i < m->m_len; i++) {
2139 			printf("%02x ", p[i]);
2140 			totlen++;
2141 			if (totlen % 16 == 0)
2142 				printf("\n");
2143 		}
2144 		m = m->m_next;
2145 	}
2146 	if (totlen % 16 != 0)
2147 		printf("\n");
2148 	printf("---\n");
2149 }
2150 
2151 /* XXX this stuff doesn't belong here... */
2152 
2153 static	struct xformsw* xforms = NULL;
2154 
2155 /*
2156  * Register a transform; typically at system startup.
2157  */
2158 void
2159 xform_register(struct xformsw* xsp)
2160 {
2161 	xsp->xf_next = xforms;
2162 	xforms = xsp;
2163 }
2164 
2165 /*
2166  * Initialize transform support in an sav.
2167  */
2168 int
2169 xform_init(struct secasvar *sav, int xftype)
2170 {
2171 	struct xformsw *xsp;
2172 
2173 	if (sav->tdb_xform != NULL)	/* previously initialized */
2174 		return 0;
2175 	for (xsp = xforms; xsp; xsp = xsp->xf_next)
2176 		if (xsp->xf_type == xftype)
2177 			return (*xsp->xf_init)(sav, xsp);
2178 
2179 	DPRINTF(("xform_init: no match for xform type %d\n", xftype));
2180 	return EINVAL;
2181 }
2182 
2183 #ifdef __NetBSD__
2184 void
2185 ipsec_attach(void)
2186 {
2187 	printf("initializing IPsec...");
2188 	ah_attach();
2189 	esp_attach();
2190 	ipcomp_attach();
2191 	ipe4_attach();
2192 #ifdef TCP_SIGNATURE
2193 	tcpsignature_attach();
2194 #endif
2195 	printf(" done\n");
2196 }
2197 #endif	/* __NetBSD__ */
2198