xref: /netbsd-src/sys/netipsec/ipsec.c (revision 7fa608457b817eca6e0977b37f758ae064f3c99c)
1 /*	$NetBSD: ipsec.c,v 1.34 2007/10/28 15:48:23 adrianp 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 1.34 2007/10/28 15:48:23 adrianp 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 
108 /*
109  * When set to 1, IPsec will send packets with the same sequence number.
110  * This allows to verify if the other side has proper replay attacks detection.
111  */
112 int ipsec_replay = 0;
113 
114 /*
115  * When set 1, IPsec will send packets with corrupted HMAC.
116  * This allows to verify if the other side properly detects modified packets.
117  */
118 int ipsec_integrity = 0;
119 #else
120 int ipsec_debug = 0;
121 #endif
122 
123 /* NB: name changed so netstat doesn't use it */
124 struct newipsecstat newipsecstat;
125 int ip4_ah_offsetmask = 0;	/* maybe IP_DF? */
126 int ip4_ipsec_dfbit = 2;	/* DF bit on encap. 0: clear 1: set 2: copy */
127 int ip4_esp_trans_deflev = IPSEC_LEVEL_USE;
128 int ip4_esp_net_deflev = IPSEC_LEVEL_USE;
129 int ip4_ah_trans_deflev = IPSEC_LEVEL_USE;
130 int ip4_ah_net_deflev = IPSEC_LEVEL_USE;
131 struct secpolicy ip4_def_policy;
132 int ip4_ipsec_ecn = 0;		/* ECN ignore(-1)/forbidden(0)/allowed(1) */
133 int ip4_esp_randpad = -1;
134 
135 #ifdef __NetBSD__
136 u_int ipsec_spdgen = 1;		/* SPD generation # */
137 
138 static struct secpolicy *ipsec_checkpcbcache (struct mbuf *,
139 	struct inpcbpolicy *, int);
140 static int ipsec_fillpcbcache (struct inpcbpolicy *, struct mbuf *,
141 	struct secpolicy *, int);
142 static int ipsec_invalpcbcache (struct inpcbpolicy *, int);
143 #endif /* __NetBSD__ */
144 
145 /*
146  * Crypto support requirements:
147  *
148  *  1	require hardware support
149  * -1	require software support
150  *  0	take anything
151  */
152 int	crypto_support = 0;
153 
154 static struct secpolicy *ipsec_getpolicybysock(struct mbuf *, u_int,
155 	PCB_T *, int *);
156 
157 #ifdef __FreeBSD__
158 SYSCTL_DECL(_net_inet_ipsec);
159 
160 /* net.inet.ipsec */
161 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_POLICY,
162 	def_policy, CTLFLAG_RW,	&ip4_def_policy.policy,	0, "");
163 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
164 	CTLFLAG_RW, &ip4_esp_trans_deflev,	0, "");
165 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
166 	CTLFLAG_RW, &ip4_esp_net_deflev,	0, "");
167 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
168 	CTLFLAG_RW, &ip4_ah_trans_deflev,	0, "");
169 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
170 	CTLFLAG_RW, &ip4_ah_net_deflev,	0, "");
171 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_CLEARTOS,
172 	ah_cleartos, CTLFLAG_RW,	&ip4_ah_cleartos,	0, "");
173 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_OFFSETMASK,
174 	ah_offsetmask, CTLFLAG_RW,	&ip4_ah_offsetmask,	0, "");
175 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DFBIT,
176 	dfbit, CTLFLAG_RW,	&ip4_ipsec_dfbit,	0, "");
177 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_ECN,
178 	ecn, CTLFLAG_RW,	&ip4_ipsec_ecn,	0, "");
179 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG,
180 	debug, CTLFLAG_RW,	&ipsec_debug,	0, "");
181 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_ESP_RANDPAD,
182 	esp_randpad, CTLFLAG_RW,	&ip4_esp_randpad,	0, "");
183 SYSCTL_INT(_net_inet_ipsec, OID_AUTO,
184 	crypto_support,	CTLFLAG_RW,	&crypto_support,0, "");
185 SYSCTL_STRUCT(_net_inet_ipsec, OID_AUTO,
186 	ipsecstats,	CTLFLAG_RD,	&newipsecstat,	newipsecstat, "");
187 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_replay, CTLFLAG_RW, &ipsec_replay, 0,
188 	"Emulate replay attack");
189 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_integrity, CTLFLAG_RW,
190 	&ipsec_integrity, 0, "Emulate man-in-the-middle attack");
191 #endif /* __FreeBSD__ */
192 
193 #ifdef INET6
194 int ip6_esp_trans_deflev = IPSEC_LEVEL_USE;
195 int ip6_esp_net_deflev = IPSEC_LEVEL_USE;
196 int ip6_ah_trans_deflev = IPSEC_LEVEL_USE;
197 int ip6_ah_net_deflev = IPSEC_LEVEL_USE;
198 struct secpolicy ip6_def_policy;
199 int ip6_ipsec_ecn = 0;		/* ECN ignore(-1)/forbidden(0)/allowed(1) */
200 int ip6_esp_randpad = -1;
201 
202 
203 #ifdef __FreeBSD__
204 SYSCTL_DECL(_net_inet6_ipsec6);
205 
206 /* net.inet6.ipsec6 */
207 #ifdef COMPAT_KAME
208 SYSCTL_OID(_net_inet6_ipsec6, IPSECCTL_STATS, stats, CTLFLAG_RD,
209 	0,0, compat_ipsecstats_sysctl, "S", "");
210 #endif /* COMPAT_KAME */
211 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_POLICY,
212 	def_policy, CTLFLAG_RW,	&ip4_def_policy.policy,	0, "");
213 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
214 	CTLFLAG_RW, &ip6_esp_trans_deflev,	0, "");
215 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
216 	CTLFLAG_RW, &ip6_esp_net_deflev,	0, "");
217 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
218 	CTLFLAG_RW, &ip6_ah_trans_deflev,	0, "");
219 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
220 	CTLFLAG_RW, &ip6_ah_net_deflev,	0, "");
221 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_ECN,
222 	ecn, CTLFLAG_RW,	&ip6_ipsec_ecn,	0, "");
223 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG,
224 	debug, CTLFLAG_RW,	&ipsec_debug,	0, "");
225 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_ESP_RANDPAD,
226 	esp_randpad, CTLFLAG_RW,	&ip6_esp_randpad,	0, "");
227 #endif /* INET6 */
228 #endif /* __FreeBSD__ */
229 
230 static int ipsec4_setspidx_inpcb (struct mbuf *, struct inpcb *);
231 #ifdef INET6
232 static int ipsec6_setspidx_in6pcb (struct mbuf *, struct in6pcb *);
233 #endif
234 static int ipsec_setspidx (struct mbuf *, struct secpolicyindex *, int);
235 static void ipsec4_get_ulp (struct mbuf *m, struct secpolicyindex *, int);
236 static int ipsec4_setspidx_ipaddr (struct mbuf *, struct secpolicyindex *);
237 #ifdef INET6
238 static void ipsec6_get_ulp (struct mbuf *m, struct secpolicyindex *, int);
239 static int ipsec6_setspidx_ipaddr (struct mbuf *, struct secpolicyindex *);
240 #endif
241 static void ipsec_delpcbpolicy (struct inpcbpolicy *);
242 static struct secpolicy *ipsec_deepcopy_policy (struct secpolicy *);
243 static int ipsec_set_policy (struct secpolicy **,int , void *, size_t , int );
244 static int ipsec_get_policy (struct secpolicy *, struct mbuf **);
245 static void vshiftl (unsigned char *, int, int);
246 static size_t ipsec_hdrsiz (struct secpolicy *);
247 
248 #ifdef __NetBSD__
249 /*
250  * Try to validate and use cached policy on a PCB.
251  */
252 static struct secpolicy *
253 ipsec_checkpcbcache(struct mbuf *m, struct inpcbpolicy *pcbsp, int dir)
254 {
255 	struct secpolicyindex spidx;
256 
257 	switch (dir) {
258 	case IPSEC_DIR_INBOUND:
259 	case IPSEC_DIR_OUTBOUND:
260 	case IPSEC_DIR_ANY:
261 		break;
262 	default:
263 		return NULL;
264 	}
265 #ifdef DIAGNOSTIC
266 	if (pcbsp == NULL) {
267 		printf("ipsec_checkpcbcache: NULL pcbsp\n");
268 		/* XXX panic? */
269 		return NULL;
270 	}
271 #endif
272 
273 #ifdef DIAGNOSTIC
274 	if (dir >= sizeof(pcbsp->sp_cache)/sizeof(pcbsp->sp_cache[0]))
275 		panic("dir too big in ipsec_checkpcbcache");
276 #endif
277 	/* SPD table change invalidate all the caches. */
278 	if (ipsec_spdgen != pcbsp->sp_cache[dir].cachegen) {
279 		ipsec_invalpcbcache(pcbsp, dir);
280 		return NULL;
281 	}
282 	if (!pcbsp->sp_cache[dir].cachesp)
283 		return NULL;
284 	if (pcbsp->sp_cache[dir].cachesp->state != IPSEC_SPSTATE_ALIVE) {
285 		ipsec_invalpcbcache(pcbsp, dir);
286 		return NULL;
287 	}
288 	if ((pcbsp->sp_cacheflags & IPSEC_PCBSP_CONNECTED) == 0) {
289 		if (!pcbsp->sp_cache[dir].cachesp)
290 			return NULL;
291 		if (ipsec_setspidx(m, &spidx, 1) != 0)
292 			return NULL;
293 
294 		/*
295 		 * We have to make an exact match here since the cached rule
296 		 * might have lower priority than a rule that would otherwise
297 		 * have matched the packet.
298 		 */
299 
300 		if (bcmp(&pcbsp->sp_cache[dir].cacheidx, &spidx, sizeof(spidx)))
301 			return NULL;
302 
303 	} else {
304 		/*
305 		 * The pcb is connected, and the L4 code is sure that:
306 		 * - outgoing side uses inp_[lf]addr
307 		 * - incoming side looks up policy after inpcb lookup
308 		 * and address pair is know to be stable.  We do not need
309 		 * to generate spidx again, nor check the address match again.
310 		 *
311 		 * For IPv4/v6 SOCK_STREAM sockets, this assumptions holds
312 		 * and there are calls to ipsec_pcbconn() from in_pcbconnect().
313 		 */
314 	}
315 
316 	pcbsp->sp_cache[dir].cachesp->lastused = time_second;
317 	pcbsp->sp_cache[dir].cachesp->refcnt++;
318 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
319 		printf("DP ipsec_checkpcbcache cause refcnt++:%d SP:%p\n",
320 		pcbsp->sp_cache[dir].cachesp->refcnt,
321 		pcbsp->sp_cache[dir].cachesp));
322 	return pcbsp->sp_cache[dir].cachesp;
323 }
324 
325 static int
326 ipsec_fillpcbcache(struct inpcbpolicy *pcbsp, struct mbuf *m,
327 	struct secpolicy *sp, int dir)
328 {
329 
330 	switch (dir) {
331 	case IPSEC_DIR_INBOUND:
332 	case IPSEC_DIR_OUTBOUND:
333 		break;
334 	default:
335 		return EINVAL;
336 	}
337 #ifdef DIAGNOSTIC
338 	if (dir >= sizeof(pcbsp->sp_cache)/sizeof(pcbsp->sp_cache[0]))
339 		panic("dir too big in ipsec_fillpcbcache");
340 #endif
341 
342 	if (pcbsp->sp_cache[dir].cachesp)
343 		KEY_FREESP(&pcbsp->sp_cache[dir].cachesp);
344 	pcbsp->sp_cache[dir].cachesp = NULL;
345 	pcbsp->sp_cache[dir].cachehint = IPSEC_PCBHINT_MAYBE;
346 	if (ipsec_setspidx(m, &pcbsp->sp_cache[dir].cacheidx, 1) != 0) {
347 		return EINVAL;
348 	}
349 	pcbsp->sp_cache[dir].cachesp = sp;
350 	if (pcbsp->sp_cache[dir].cachesp) {
351 		pcbsp->sp_cache[dir].cachesp->refcnt++;
352 		KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
353 			printf("DP ipsec_fillpcbcache cause refcnt++:%d SP:%p\n",
354 			pcbsp->sp_cache[dir].cachesp->refcnt,
355 			pcbsp->sp_cache[dir].cachesp));
356 
357 		/*
358 		 * If the PCB is connected, we can remember a hint to
359 		 * possibly short-circuit IPsec processing in other places.
360 		 */
361 		if (pcbsp->sp_cacheflags & IPSEC_PCBSP_CONNECTED) {
362 			switch (pcbsp->sp_cache[dir].cachesp->policy) {
363 			case IPSEC_POLICY_NONE:
364 			case IPSEC_POLICY_BYPASS:
365 				pcbsp->sp_cache[dir].cachehint =
366 					IPSEC_PCBHINT_NO;
367 				break;
368 			default:
369 				pcbsp->sp_cache[dir].cachehint =
370 					IPSEC_PCBHINT_YES;
371 			}
372 		}
373 	}
374 	pcbsp->sp_cache[dir].cachegen = ipsec_spdgen;
375 
376 	return 0;
377 }
378 
379 static int
380 ipsec_invalpcbcache(struct inpcbpolicy *pcbsp, int dir)
381 {
382 	int i;
383 
384 	for (i = IPSEC_DIR_INBOUND; i <= IPSEC_DIR_OUTBOUND; i++) {
385 		if (dir != IPSEC_DIR_ANY && i != dir)
386 			continue;
387 		if (pcbsp->sp_cache[i].cachesp)
388 			KEY_FREESP(&pcbsp->sp_cache[i].cachesp);
389 		pcbsp->sp_cache[i].cachesp = NULL;
390 		pcbsp->sp_cache[i].cachehint = IPSEC_PCBHINT_MAYBE;
391 		pcbsp->sp_cache[i].cachegen = 0;
392 		bzero(&pcbsp->sp_cache[i].cacheidx,
393 			  sizeof(pcbsp->sp_cache[i].cacheidx));
394 	}
395 	return 0;
396 }
397 
398 void
399 ipsec_pcbconn(struct inpcbpolicy *pcbsp)
400 {
401 
402 	pcbsp->sp_cacheflags |= IPSEC_PCBSP_CONNECTED;
403 	ipsec_invalpcbcache(pcbsp, IPSEC_DIR_ANY);
404 }
405 
406 void
407 ipsec_pcbdisconn(struct inpcbpolicy *pcbsp)
408 {
409 
410 	pcbsp->sp_cacheflags &= ~IPSEC_PCBSP_CONNECTED;
411 	ipsec_invalpcbcache(pcbsp, IPSEC_DIR_ANY);
412 }
413 
414 void
415 ipsec_invalpcbcacheall(void)
416 {
417 
418 	if (ipsec_spdgen == UINT_MAX)
419 		ipsec_spdgen = 1;
420 	else
421 		ipsec_spdgen++;
422 }
423 #endif /* __NetBSD__ */
424 
425 /*
426  * Return a held reference to the default SP.
427  */
428 static struct secpolicy *
429 key_allocsp_default(int af, const char* where, int tag)
430 {
431 	struct secpolicy *sp;
432 
433 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
434 		printf("DP key_allocsp_default from %s:%u\n", where, tag));
435 
436     switch(af) {
437         case AF_INET:
438 	        sp = &ip4_def_policy;
439             break;
440 #ifdef INET6
441         case AF_INET6:
442             sp = &ip6_def_policy;
443             break;
444 #endif
445         default:
446 	        KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
447 		    printf("key_allocsp_default : unexpected protocol family %u\n",
448                    af));
449             return NULL;
450     }
451 
452 	if (sp->policy != IPSEC_POLICY_DISCARD &&
453 		sp->policy != IPSEC_POLICY_NONE) {
454 		ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n",
455 			sp->policy, IPSEC_POLICY_NONE));
456 		sp->policy = IPSEC_POLICY_NONE;
457 	}
458 	sp->refcnt++;
459 
460 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
461 		printf("DP key_allocsp_default returns SP:%p (%u)\n",
462 			sp, sp->refcnt));
463 	return sp;
464 }
465 #define	KEY_ALLOCSP_DEFAULT(af) \
466 	key_allocsp_default((af),__FILE__, __LINE__)
467 
468 /*
469  * For OUTBOUND packet having a socket. Searching SPD for packet,
470  * and return a pointer to SP.
471  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
472  *		0	: bypass
473  *		EACCES	: discard packet.
474  *		ENOENT	: ipsec_acquire() in progress, maybe.
475  *		others	: error occurred.
476  *	others:	a pointer to SP
477  *
478  * NOTE: IPv6 mapped address concern is implemented here.
479  */
480 struct secpolicy *
481 ipsec_getpolicy(struct tdb_ident *tdbi, u_int dir)
482 {
483 	struct secpolicy *sp;
484 
485 	IPSEC_ASSERT(tdbi != NULL, ("ipsec_getpolicy: null tdbi"));
486 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
487 		("ipsec_getpolicy: invalid direction %u", dir));
488 
489 	sp = KEY_ALLOCSP2(tdbi->spi, &tdbi->dst, tdbi->proto, dir);
490 	if (sp == NULL)			/*XXX????*/
491 		sp = KEY_ALLOCSP_DEFAULT(tdbi->dst.sa.sa_family);
492 	IPSEC_ASSERT(sp != NULL, ("ipsec_getpolicy: null SP"));
493 	return sp;
494 }
495 
496 /*
497  * For OUTBOUND packet having a socket. Searching SPD for packet,
498  * and return a pointer to SP.
499  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
500  *		0	: bypass
501  *		EACCES	: discard packet.
502  *		ENOENT	: ipsec_acquire() in progress, maybe.
503  *		others	: error occurred.
504  *	others:	a pointer to SP
505  *
506  * NOTE: IPv6 mapped address concern is implemented here.
507  */
508 static struct secpolicy *
509 ipsec_getpolicybysock(struct mbuf *m, u_int dir, PCB_T *inp, int *error)
510 {
511 	struct inpcbpolicy *pcbsp = NULL;
512 	struct secpolicy *currsp = NULL;	/* policy on socket */
513 	struct secpolicy *sp;
514 	int af;
515 
516 	IPSEC_ASSERT(m != NULL, ("ipsec_getpolicybysock: null mbuf"));
517 	IPSEC_ASSERT(inp != NULL, ("ipsec_getpolicybysock: null inpcb"));
518 	IPSEC_ASSERT(error != NULL, ("ipsec_getpolicybysock: null error"));
519 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
520 		("ipsec_getpolicybysock: invalid direction %u", dir));
521 
522 	IPSEC_ASSERT(PCB_SOCKET(inp) != NULL,
523 		("ipsec_getppolicybysock: null socket\n"));
524 
525 	/* XXX FIXME inpcb/in6pcb  vs socket*/
526 	af = PCB_FAMILY(inp);
527 	IPSEC_ASSERT(af == AF_INET || af == AF_INET6,
528 		("ipsec_getpolicybysock: unexpected protocol family %u", af));
529 
530 #ifdef __NetBSD__
531 	IPSEC_ASSERT(inp->inph_sp != NULL, ("null PCB policy cache"));
532 	/* If we have a cached entry, and if it is still valid, use it. */
533 	ipsecstat.ips_spdcache_lookup++;
534 	currsp = ipsec_checkpcbcache(m, /*inpcb_hdr*/inp->inph_sp, dir);
535 	if (currsp) {
536 		*error = 0;
537 		return currsp;
538 	}
539 	ipsecstat.ips_spdcache_miss++;
540 #endif /* __NetBSD__ */
541 
542 	switch (af) {
543 	case AF_INET: {
544 		struct inpcb *in4p = PCB_TO_IN4PCB(inp);
545 		/* set spidx in pcb */
546 		*error = ipsec4_setspidx_inpcb(m, in4p);
547 		pcbsp = in4p->inp_sp;
548 		break;
549 		}
550 
551 #if defined(INET6)
552 	case AF_INET6: {
553 		struct in6pcb *in6p = PCB_TO_IN6PCB(inp);
554 		/* set spidx in pcb */
555 		*error = ipsec6_setspidx_in6pcb(m, in6p);
556 		pcbsp = in6p->in6p_sp;
557 		break;
558 		}
559 #endif
560 	default:
561 		*error = EPFNOSUPPORT;
562 		break;
563 	}
564 	if (*error)
565 		return NULL;
566 
567 	IPSEC_ASSERT(pcbsp != NULL, ("ipsec_getpolicybysock: null pcbsp"));
568 	switch (dir) {
569 	case IPSEC_DIR_INBOUND:
570 		currsp = pcbsp->sp_in;
571 		break;
572 	case IPSEC_DIR_OUTBOUND:
573 		currsp = pcbsp->sp_out;
574 		break;
575 	}
576 	IPSEC_ASSERT(currsp != NULL, ("ipsec_getpolicybysock: null currsp"));
577 
578 	if (pcbsp->priv) {			/* when privilieged socket */
579 		switch (currsp->policy) {
580 		case IPSEC_POLICY_BYPASS:
581 		case IPSEC_POLICY_IPSEC:
582 			currsp->refcnt++;
583 			sp = currsp;
584 			break;
585 
586 		case IPSEC_POLICY_ENTRUST:
587 			/* look for a policy in SPD */
588 			sp = KEY_ALLOCSP(&currsp->spidx, dir);
589 			if (sp == NULL)		/* no SP found */
590 				sp = KEY_ALLOCSP_DEFAULT(af);
591 			break;
592 
593 		default:
594 			ipseclog((LOG_ERR, "ipsec_getpolicybysock: "
595 				  "Invalid policy for PCB %d\n", currsp->policy));
596 			*error = EINVAL;
597 			return NULL;
598 		}
599 	} else {				/* unpriv, SPD has policy */
600 		sp = KEY_ALLOCSP(&currsp->spidx, dir);
601 		if (sp == NULL) {		/* no SP found */
602 			switch (currsp->policy) {
603 			case IPSEC_POLICY_BYPASS:
604 				ipseclog((LOG_ERR, "ipsec_getpolicybysock: "
605 					   "Illegal policy for non-priviliged defined %d\n",
606 					currsp->policy));
607 				*error = EINVAL;
608 				return NULL;
609 
610 			case IPSEC_POLICY_ENTRUST:
611 				sp = KEY_ALLOCSP_DEFAULT(af);
612 				break;
613 
614 			case IPSEC_POLICY_IPSEC:
615 				currsp->refcnt++;
616 				sp = currsp;
617 				break;
618 
619 			default:
620 				ipseclog((LOG_ERR, "ipsec_getpolicybysock: "
621 				   "Invalid policy for PCB %d\n", currsp->policy));
622 				*error = EINVAL;
623 				return NULL;
624 			}
625 		}
626 	}
627 	IPSEC_ASSERT(sp != NULL,
628 		("ipsec_getpolicybysock: null SP (priv %u policy %u",
629 		 pcbsp->priv, currsp->policy));
630 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
631 		printf("DP ipsec_getpolicybysock (priv %u policy %u) allocates "
632 			   "SP:%p (refcnt %u)\n", pcbsp->priv, currsp->policy,
633 			   sp, sp->refcnt));
634 #ifdef __NetBSD__
635 	ipsec_fillpcbcache(pcbsp, m, sp, dir);
636 #endif /* __NetBSD__ */
637 	return sp;
638 }
639 
640 /*
641  * For FORWADING packet or OUTBOUND without a socket. Searching SPD for packet,
642  * and return a pointer to SP.
643  * OUT:	positive: a pointer to the entry for security policy leaf matched.
644  *	NULL:	no apropreate SP found, the following value is set to error.
645  *		0	: bypass
646  *		EACCES	: discard packet.
647  *		ENOENT	: ipsec_acquire() in progress, maybe.
648  *		others	: error occurred.
649  */
650 struct secpolicy *
651 ipsec_getpolicybyaddr(struct mbuf *m, u_int dir, int flag, int *error)
652 {
653 	struct secpolicyindex spidx;
654 	struct secpolicy *sp;
655 
656 	IPSEC_ASSERT(m != NULL, ("ipsec_getpolicybyaddr: null mbuf"));
657 	IPSEC_ASSERT(error != NULL, ("ipsec_getpolicybyaddr: null error"));
658 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
659 		("ipsec4_getpolicybaddr: invalid direction %u", dir));
660 
661 	sp = NULL;
662 
663 	/* Make an index to look for a policy. */
664 	*error = ipsec_setspidx(m, &spidx, (flag & IP_FORWARDING) ? 0 : 1);
665 	if (*error != 0) {
666 		DPRINTF(("ipsec_getpolicybyaddr: setpidx failed,"
667 			" dir %u flag %u\n", dir, flag));
668 		bzero(&spidx, sizeof (spidx));
669 		return NULL;
670 	}
671 
672 	spidx.dir = dir;
673 
674 	if (key_havesp(dir)) {
675 		sp = KEY_ALLOCSP(&spidx, dir);
676 	}
677 
678 	if (sp == NULL)			/* no SP found, use system default */
679 		sp = KEY_ALLOCSP_DEFAULT(spidx.dst.sa.sa_family);
680 	IPSEC_ASSERT(sp != NULL, ("ipsec_getpolicybyaddr: null SP"));
681 	return sp;
682 }
683 
684 struct secpolicy *
685 ipsec4_checkpolicy(struct mbuf *m, u_int dir, u_int flag, int *error,
686 		   struct inpcb *inp)
687 {
688 	struct secpolicy *sp;
689 
690 	*error = 0;
691 
692 
693 	/* XXX KAME IPv6 calls us with non-null inp but bogus inp_socket? */
694 	if (inp == NULL || inp->inp_socket == NULL) {
695 		sp = ipsec_getpolicybyaddr(m, dir, flag, error);
696 	} else
697 		sp = ipsec_getpolicybysock(m, dir, IN4PCB_TO_PCB(inp), error);
698 	if (sp == NULL) {
699 		IPSEC_ASSERT(*error != 0,
700 			("ipsec4_checkpolicy: getpolicy failed w/o error"));
701 		newipsecstat.ips_out_inval++;
702 		return NULL;
703 	}
704 	IPSEC_ASSERT(*error == 0,
705 		("ipsec4_checkpolicy: sp w/ error set to %u", *error));
706 	switch (sp->policy) {
707 	case IPSEC_POLICY_ENTRUST:
708 	default:
709 		printf("ipsec4_checkpolicy: invalid policy %u\n", sp->policy);
710 		/* fall thru... */
711 	case IPSEC_POLICY_DISCARD:
712 		newipsecstat.ips_out_polvio++;
713 		*error = -EINVAL;	/* packet is discarded by caller */
714 		break;
715 	case IPSEC_POLICY_BYPASS:
716 	case IPSEC_POLICY_NONE:
717 		KEY_FREESP(&sp);
718 		sp = NULL;		/* NB: force NULL result */
719 		break;
720 	case IPSEC_POLICY_IPSEC:
721 		if (sp->req == NULL)	/* acquire an SA */
722 			*error = key_spdacquire(sp);
723 		break;
724 	}
725 	if (*error != 0) {
726 		KEY_FREESP(&sp);
727 		sp = NULL;
728 	}
729 	DPRINTF(("ipsecpol: done, sp %p error %d, \n", sp, *error));
730 	return sp;
731 }
732 
733 #ifdef INET6
734 struct secpolicy *
735 ipsec6_checkpolicy(struct mbuf *m, u_int dir, u_int flag, int *error,
736 	 	   struct in6pcb *in6p)
737 {
738 	struct secpolicy *sp;
739 
740 	*error = 0;
741 
742 
743 	/* XXX KAME IPv6 calls us with non-null inp but bogus inp_socket? */
744 	if (in6p == NULL || in6p->in6p_socket == NULL) {
745 		sp = ipsec_getpolicybyaddr(m, dir, flag, error);
746 	} else
747 		sp = ipsec_getpolicybysock(m, dir, IN6PCB_TO_PCB(in6p), error);
748 	if (sp == NULL) {
749 		IPSEC_ASSERT(*error != 0,
750 			("ipsec6_checkpolicy: getpolicy failed w/o error"));
751 		newipsecstat.ips_out_inval++;
752 		return NULL;
753 	}
754 	IPSEC_ASSERT(*error == 0,
755 		("ipsec6_checkpolicy: sp w/ error set to %u", *error));
756 	switch (sp->policy) {
757 	case IPSEC_POLICY_ENTRUST:
758 	default:
759 		printf("ipsec6_checkpolicy: invalid policy %u\n", sp->policy);
760 		/* fall thru... */
761 	case IPSEC_POLICY_DISCARD:
762 		newipsecstat.ips_out_polvio++;
763 		*error = -EINVAL;   /* packet is discarded by caller */
764 		break;
765 	case IPSEC_POLICY_BYPASS:
766 	case IPSEC_POLICY_NONE:
767 		KEY_FREESP(&sp);
768 		sp = NULL;	  /* NB: force NULL result */
769 		break;
770 	case IPSEC_POLICY_IPSEC:
771 		if (sp->req == NULL)	/* acquire an SA */
772 			*error = key_spdacquire(sp);
773 		break;
774 	}
775 	if (*error != 0) {
776 		KEY_FREESP(&sp);
777 		sp = NULL;
778 	}
779 	DPRINTF(("ipsecpol: done, sp %p error %d, \n", sp, *error));
780 	return sp;
781 }
782 #endif /* INET6 */
783 
784 static int
785 ipsec4_setspidx_inpcb(struct mbuf *m ,struct inpcb *pcb)
786 {
787 	int error;
788 
789 	IPSEC_ASSERT(pcb != NULL, ("ipsec4_setspidx_inpcb: null pcb"));
790 	IPSEC_ASSERT(pcb->inp_sp != NULL, ("ipsec4_setspidx_inpcb: null inp_sp"));
791 	IPSEC_ASSERT(pcb->inp_sp->sp_out != NULL && pcb->inp_sp->sp_in != NULL,
792 		("ipsec4_setspidx_inpcb: null sp_in || sp_out"));
793 
794 	error = ipsec_setspidx(m, &pcb->inp_sp->sp_in->spidx, 1);
795 	if (error == 0) {
796 		pcb->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
797 		pcb->inp_sp->sp_out->spidx = pcb->inp_sp->sp_in->spidx;
798 		pcb->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
799 	} else {
800 		bzero(&pcb->inp_sp->sp_in->spidx,
801 			sizeof (pcb->inp_sp->sp_in->spidx));
802 		bzero(&pcb->inp_sp->sp_out->spidx,
803 			sizeof (pcb->inp_sp->sp_in->spidx));
804 	}
805 	return error;
806 }
807 
808 #ifdef INET6
809 static int
810 ipsec6_setspidx_in6pcb(struct mbuf *m, struct in6pcb *pcb)
811 {
812 	struct secpolicyindex *spidx;
813 	int error;
814 
815 	IPSEC_ASSERT(pcb != NULL, ("ipsec6_setspidx_in6pcb: null pcb"));
816 	IPSEC_ASSERT(pcb->in6p_sp != NULL, ("ipsec6_setspidx_in6pcb: null inp_sp"));
817 	IPSEC_ASSERT(pcb->in6p_sp->sp_out != NULL && pcb->in6p_sp->sp_in != NULL,
818 		("ipsec6_setspidx_in6pcb: null sp_in || sp_out"));
819 
820 	bzero(&pcb->in6p_sp->sp_in->spidx, sizeof(*spidx));
821 	bzero(&pcb->in6p_sp->sp_out->spidx, sizeof(*spidx));
822 
823 	spidx = &pcb->in6p_sp->sp_in->spidx;
824 	error = ipsec_setspidx(m, spidx, 1);
825 	if (error)
826 		goto bad;
827 	spidx->dir = IPSEC_DIR_INBOUND;
828 
829 	spidx = &pcb->in6p_sp->sp_out->spidx;
830 	error = ipsec_setspidx(m, spidx, 1);
831 	if (error)
832 		goto bad;
833 	spidx->dir = IPSEC_DIR_OUTBOUND;
834 
835 	return 0;
836 
837 bad:
838 	bzero(&pcb->in6p_sp->sp_in->spidx, sizeof(*spidx));
839 	bzero(&pcb->in6p_sp->sp_out->spidx, sizeof(*spidx));
840 	return error;
841 }
842 #endif
843 
844 /*
845  * configure security policy index (src/dst/proto/sport/dport)
846  * by looking at the content of mbuf.
847  * the caller is responsible for error recovery (like clearing up spidx).
848  */
849 static int
850 ipsec_setspidx(struct mbuf *m, struct secpolicyindex *spidx, int needport)
851 {
852 	struct ip *ip = NULL;
853 	struct ip ipbuf;
854 	u_int v;
855 	struct mbuf *n;
856 	int len;
857 	int error;
858 
859 	IPSEC_ASSERT(m != NULL, ("ipsec_setspidx: null mbuf"));
860 
861 	/*
862 	 * validate m->m_pkthdr.len.  we see incorrect length if we
863 	 * mistakenly call this function with inconsistent mbuf chain
864 	 * (like 4.4BSD tcp/udp processing).  XXX should we panic here?
865 	 */
866 	len = 0;
867 	for (n = m; n; n = n->m_next)
868 		len += n->m_len;
869 	if (m->m_pkthdr.len != len) {
870 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
871 			printf("ipsec_setspidx: "
872 				   "total of m_len(%d) != pkthdr.len(%d), "
873 				   "ignored.\n",
874 				len, m->m_pkthdr.len));
875 		return EINVAL;
876 	}
877 
878 	if (m->m_pkthdr.len < sizeof(struct ip)) {
879 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
880 			printf("ipsec_setspidx: "
881 				"pkthdr.len(%d) < sizeof(struct ip), ignored.\n",
882 				m->m_pkthdr.len));
883 		return EINVAL;
884 	}
885 
886 	if (m->m_len >= sizeof(*ip))
887 		ip = mtod(m, struct ip *);
888 	else {
889 		m_copydata(m, 0, sizeof(ipbuf), &ipbuf);
890 		ip = &ipbuf;
891 	}
892 #ifdef _IP_VHL
893 	v = _IP_VHL_V(ip->ip_vhl);
894 #else
895 	v = ip->ip_v;
896 #endif
897 	switch (v) {
898 	case 4:
899 		error = ipsec4_setspidx_ipaddr(m, spidx);
900 		if (error)
901 			return error;
902 		ipsec4_get_ulp(m, spidx, needport);
903 		return 0;
904 #ifdef INET6
905 	case 6:
906 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) {
907 			KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
908 				printf("ipsec_setspidx: "
909 					"pkthdr.len(%d) < sizeof(struct ip6_hdr), "
910 					"ignored.\n", m->m_pkthdr.len));
911 			return EINVAL;
912 		}
913 		error = ipsec6_setspidx_ipaddr(m, spidx);
914 		if (error)
915 			return error;
916 		ipsec6_get_ulp(m, spidx, needport);
917 		return 0;
918 #endif
919 	default:
920 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
921 			printf("ipsec_setspidx: "
922 				"unknown IP version %u, ignored.\n", v));
923 		return EINVAL;
924 	}
925 }
926 
927 static void
928 ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
929 {
930 	u_int8_t nxt;
931 	int off;
932 
933 	/* sanity check */
934 	IPSEC_ASSERT(m != NULL, ("ipsec4_get_ulp: null mbuf"));
935 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),
936 		("ipsec4_get_ulp: packet too short"));
937 
938 	/* NB: ip_input() flips it into host endian XXX need more checking */
939 	if (m->m_len >= sizeof(struct ip)) {
940 		struct ip *ip = mtod(m, struct ip *);
941 		if (ip->ip_off & IP_OFF_CONVERT(IP_MF | IP_OFFMASK))
942 			goto done;
943 #ifdef _IP_VHL
944 		off = _IP_VHL_HL(ip->ip_vhl) << 2;
945 #else
946 		off = ip->ip_hl << 2;
947 #endif
948 		nxt = ip->ip_p;
949 	} else {
950 		struct ip ih;
951 
952 		m_copydata(m, 0, sizeof (struct ip), &ih);
953 		if (ih.ip_off & IP_OFF_CONVERT(IP_MF | IP_OFFMASK))
954 			goto done;
955 #ifdef _IP_VHL
956 		off = _IP_VHL_HL(ih.ip_vhl) << 2;
957 #else
958 		off = ih.ip_hl << 2;
959 #endif
960 		nxt = ih.ip_p;
961 	}
962 
963 	while (off < m->m_pkthdr.len) {
964 		struct ip6_ext ip6e;
965 		struct tcphdr th;
966 		struct udphdr uh;
967 
968 		switch (nxt) {
969 		case IPPROTO_TCP:
970 			spidx->ul_proto = nxt;
971 			if (!needport)
972 				goto done_proto;
973 			if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
974 				goto done;
975 			m_copydata(m, off, sizeof (th), &th);
976 			spidx->src.sin.sin_port = th.th_sport;
977 			spidx->dst.sin.sin_port = th.th_dport;
978 			return;
979 		case IPPROTO_UDP:
980 			spidx->ul_proto = nxt;
981 			if (!needport)
982 				goto done_proto;
983 			if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
984 				goto done;
985 			m_copydata(m, off, sizeof (uh), &uh);
986 			spidx->src.sin.sin_port = uh.uh_sport;
987 			spidx->dst.sin.sin_port = uh.uh_dport;
988 			return;
989 		case IPPROTO_AH:
990 			if (m->m_pkthdr.len > off + sizeof(ip6e))
991 				goto done;
992 			/* XXX sigh, this works but is totally bogus */
993 			m_copydata(m, off, sizeof(ip6e), &ip6e);
994 			off += (ip6e.ip6e_len + 2) << 2;
995 			nxt = ip6e.ip6e_nxt;
996 			break;
997 		case IPPROTO_ICMP:
998 		default:
999 			/* XXX intermediate headers??? */
1000 			spidx->ul_proto = nxt;
1001 			goto done_proto;
1002 		}
1003 	}
1004 done:
1005 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
1006 done_proto:
1007 	spidx->src.sin.sin_port = IPSEC_PORT_ANY;
1008 	spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
1009 }
1010 
1011 /* assumes that m is sane */
1012 static int
1013 ipsec4_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
1014 {
1015 	static const struct sockaddr_in template = {
1016 		sizeof (struct sockaddr_in),
1017 		AF_INET,
1018 		0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 }
1019 	};
1020 
1021 	spidx->src.sin = template;
1022 	spidx->dst.sin = template;
1023 
1024 	if (m->m_len < sizeof (struct ip)) {
1025 		m_copydata(m, offsetof(struct ip, ip_src),
1026 			   sizeof (struct  in_addr),
1027 			   &spidx->src.sin.sin_addr);
1028 		m_copydata(m, offsetof(struct ip, ip_dst),
1029 			   sizeof (struct  in_addr),
1030 			   &spidx->dst.sin.sin_addr);
1031 	} else {
1032 		struct ip *ip = mtod(m, struct ip *);
1033 		spidx->src.sin.sin_addr = ip->ip_src;
1034 		spidx->dst.sin.sin_addr = ip->ip_dst;
1035 	}
1036 
1037 	spidx->prefs = sizeof(struct in_addr) << 3;
1038 	spidx->prefd = sizeof(struct in_addr) << 3;
1039 
1040 	return 0;
1041 }
1042 
1043 #ifdef INET6
1044 static void
1045 ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *spidx,
1046 	       int needport)
1047 {
1048 	int off, nxt;
1049 	struct tcphdr th;
1050 	struct udphdr uh;
1051 
1052 	/* sanity check */
1053 	if (m == NULL)
1054 		panic("ipsec6_get_ulp: NULL pointer was passed");
1055 
1056 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1057 		printf("ipsec6_get_ulp:\n"); kdebug_mbuf(m));
1058 
1059 	/* set default */
1060 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
1061 	((struct sockaddr_in6 *)&spidx->src)->sin6_port = IPSEC_PORT_ANY;
1062 	((struct sockaddr_in6 *)&spidx->dst)->sin6_port = IPSEC_PORT_ANY;
1063 
1064 	nxt = -1;
1065 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
1066 	if (off < 0 || m->m_pkthdr.len < off)
1067 		return;
1068 
1069 	switch (nxt) {
1070 	case IPPROTO_TCP:
1071 		spidx->ul_proto = nxt;
1072 		if (!needport)
1073 			break;
1074 		if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
1075 			break;
1076 		m_copydata(m, off, sizeof(th), &th);
1077 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = th.th_sport;
1078 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = th.th_dport;
1079 		break;
1080 	case IPPROTO_UDP:
1081 		spidx->ul_proto = nxt;
1082 		if (!needport)
1083 			break;
1084 		if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
1085 			break;
1086 		m_copydata(m, off, sizeof(uh), &uh);
1087 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = uh.uh_sport;
1088 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = uh.uh_dport;
1089 		break;
1090 	case IPPROTO_ICMPV6:
1091 	default:
1092 		/* XXX intermediate headers??? */
1093 		spidx->ul_proto = nxt;
1094 		break;
1095 	}
1096 }
1097 
1098 /* assumes that m is sane */
1099 static int
1100 ipsec6_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
1101 {
1102 	struct ip6_hdr *ip6 = NULL;
1103 	struct ip6_hdr ip6buf;
1104 	struct sockaddr_in6 *sin6;
1105 
1106 	if (m->m_len >= sizeof(*ip6))
1107 		ip6 = mtod(m, struct ip6_hdr *);
1108 	else {
1109 		m_copydata(m, 0, sizeof(ip6buf), &ip6buf);
1110 		ip6 = &ip6buf;
1111 	}
1112 
1113 	sin6 = (struct sockaddr_in6 *)&spidx->src;
1114 	bzero(sin6, sizeof(*sin6));
1115 	sin6->sin6_family = AF_INET6;
1116 	sin6->sin6_len = sizeof(struct sockaddr_in6);
1117 	bcopy(&ip6->ip6_src, &sin6->sin6_addr, sizeof(ip6->ip6_src));
1118 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
1119 		sin6->sin6_addr.s6_addr16[1] = 0;
1120 		sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]);
1121 	}
1122 	spidx->prefs = sizeof(struct in6_addr) << 3;
1123 
1124 	sin6 = (struct sockaddr_in6 *)&spidx->dst;
1125 	bzero(sin6, sizeof(*sin6));
1126 	sin6->sin6_family = AF_INET6;
1127 	sin6->sin6_len = sizeof(struct sockaddr_in6);
1128 	bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(ip6->ip6_dst));
1129 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
1130 		sin6->sin6_addr.s6_addr16[1] = 0;
1131 		sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]);
1132 	}
1133 	spidx->prefd = sizeof(struct in6_addr) << 3;
1134 
1135 	return 0;
1136 }
1137 #endif
1138 
1139 static void
1140 ipsec_delpcbpolicy(struct inpcbpolicy *p)
1141 {
1142 	free(p, M_SECA);
1143 }
1144 
1145 /* initialize policy in PCB */
1146 int
1147 ipsec_init_policy(struct socket *so, struct inpcbpolicy **pcb_sp)
1148 {
1149 	struct inpcbpolicy *new;
1150 
1151 	/* sanity check. */
1152 	if (so == NULL || pcb_sp == NULL)
1153 		panic("ipsec_init_policy: NULL pointer was passed");
1154 
1155 	new = (struct inpcbpolicy *) malloc(sizeof(struct inpcbpolicy),
1156 						M_SECA, M_NOWAIT|M_ZERO);
1157 	if (new == NULL) {
1158 		ipseclog((LOG_DEBUG, "ipsec_init_policy: No more memory.\n"));
1159 		return ENOBUFS;
1160 	}
1161 
1162 	if (IPSEC_PRIVILEGED_SO(so))
1163 		new->priv = 1;
1164 	else
1165 		new->priv = 0;
1166 
1167 	if ((new->sp_in = KEY_NEWSP()) == NULL) {
1168 		ipsec_delpcbpolicy(new);
1169 		return ENOBUFS;
1170 	}
1171 	new->sp_in->state = IPSEC_SPSTATE_ALIVE;
1172 	new->sp_in->policy = IPSEC_POLICY_ENTRUST;
1173 
1174 	if ((new->sp_out = KEY_NEWSP()) == NULL) {
1175 		KEY_FREESP(&new->sp_in);
1176 		ipsec_delpcbpolicy(new);
1177 		return ENOBUFS;
1178 	}
1179 	new->sp_out->state = IPSEC_SPSTATE_ALIVE;
1180 	new->sp_out->policy = IPSEC_POLICY_ENTRUST;
1181 
1182 	*pcb_sp = new;
1183 
1184 	return 0;
1185 }
1186 
1187 /* copy old ipsec policy into new */
1188 int
1189 ipsec_copy_policy(struct inpcbpolicy *old, struct inpcbpolicy *new)
1190 {
1191 	struct secpolicy *sp;
1192 
1193 	sp = ipsec_deepcopy_policy(old->sp_in);
1194 	if (sp) {
1195 		KEY_FREESP(&new->sp_in);
1196 		new->sp_in = sp;
1197 	} else
1198 		return ENOBUFS;
1199 
1200 	sp = ipsec_deepcopy_policy(old->sp_out);
1201 	if (sp) {
1202 		KEY_FREESP(&new->sp_out);
1203 		new->sp_out = sp;
1204 	} else
1205 		return ENOBUFS;
1206 
1207 	new->priv = old->priv;
1208 
1209 	return 0;
1210 }
1211 
1212 /* deep-copy a policy in PCB */
1213 static struct secpolicy *
1214 ipsec_deepcopy_policy(struct secpolicy *src)
1215 {
1216 	struct ipsecrequest *newchain = NULL;
1217 	struct ipsecrequest *p;
1218 	struct ipsecrequest **q;
1219 	struct ipsecrequest *r;
1220 	struct secpolicy *dst;
1221 
1222 	if (src == NULL)
1223 		return NULL;
1224 	dst = KEY_NEWSP();
1225 	if (dst == NULL)
1226 		return NULL;
1227 
1228 	/*
1229 	 * deep-copy IPsec request chain.  This is required since struct
1230 	 * ipsecrequest is not reference counted.
1231 	 */
1232 	q = &newchain;
1233 	for (p = src->req; p; p = p->next) {
1234 		*q = (struct ipsecrequest *)malloc(sizeof(struct ipsecrequest),
1235 			M_SECA, M_NOWAIT);
1236 		if (*q == NULL)
1237 			goto fail;
1238 		bzero(*q, sizeof(**q));
1239 		(*q)->next = NULL;
1240 
1241 		(*q)->saidx.proto = p->saidx.proto;
1242 		(*q)->saidx.mode = p->saidx.mode;
1243 		(*q)->level = p->level;
1244 		(*q)->saidx.reqid = p->saidx.reqid;
1245 
1246 		bcopy(&p->saidx.src, &(*q)->saidx.src, sizeof((*q)->saidx.src));
1247 		bcopy(&p->saidx.dst, &(*q)->saidx.dst, sizeof((*q)->saidx.dst));
1248 
1249 		(*q)->sav = NULL;
1250 		(*q)->sp = dst;
1251 
1252 		q = &((*q)->next);
1253 	}
1254 
1255 	dst->req = newchain;
1256 	dst->state = src->state;
1257 	dst->policy = src->policy;
1258 	/* do not touch the refcnt fields */
1259 
1260 	return dst;
1261 
1262 fail:
1263 	for (p = newchain; p; p = r) {
1264 		r = p->next;
1265 		free(p, M_SECA);
1266 		p = NULL;
1267 	}
1268 	return NULL;
1269 }
1270 
1271 /* set policy and ipsec request if present. */
1272 static int
1273 ipsec_set_policy(
1274 	struct secpolicy **pcb_sp,
1275 	int optname,
1276 	void *request,
1277 	size_t len,
1278 	int priv
1279 )
1280 {
1281 	struct sadb_x_policy *xpl;
1282 	struct secpolicy *newsp = NULL;
1283 	int error;
1284 
1285 	/* sanity check. */
1286 	if (pcb_sp == NULL || *pcb_sp == NULL || request == NULL)
1287 		return EINVAL;
1288 	if (len < sizeof(*xpl))
1289 		return EINVAL;
1290 	xpl = (struct sadb_x_policy *)request;
1291 
1292 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1293 		printf("ipsec_set_policy: passed policy\n");
1294 		kdebug_sadb_x_policy((struct sadb_ext *)xpl));
1295 
1296 	/* check policy type */
1297 	/* ipsec_set_policy() accepts IPSEC, ENTRUST and BYPASS. */
1298 	if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD
1299 	 || xpl->sadb_x_policy_type == IPSEC_POLICY_NONE)
1300 		return EINVAL;
1301 
1302 	/* check privileged socket */
1303 	if (priv == 0 && xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS)
1304 		return EACCES;
1305 
1306 	/* allocation new SP entry */
1307 	if ((newsp = key_msg2sp(xpl, len, &error)) == NULL)
1308 		return error;
1309 
1310 	newsp->state = IPSEC_SPSTATE_ALIVE;
1311 
1312 	/* clear old SP and set new SP */
1313 	KEY_FREESP(pcb_sp);
1314 	*pcb_sp = newsp;
1315 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1316 		printf("ipsec_set_policy: new policy\n");
1317 		kdebug_secpolicy(newsp));
1318 
1319 	return 0;
1320 }
1321 
1322 static int
1323 ipsec_get_policy(struct secpolicy *pcb_sp, struct mbuf **mp)
1324 {
1325 
1326 	/* sanity check. */
1327 	if (pcb_sp == NULL || mp == NULL)
1328 		return EINVAL;
1329 
1330 	*mp = key_sp2msg(pcb_sp);
1331 	if (!*mp) {
1332 		ipseclog((LOG_DEBUG, "ipsec_get_policy: No more memory.\n"));
1333 		return ENOBUFS;
1334 	}
1335 
1336 	(*mp)->m_type = MT_DATA;
1337 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1338 		printf("ipsec_get_policy:\n");
1339 		kdebug_mbuf(*mp));
1340 
1341 	return 0;
1342 }
1343 
1344 int
1345 ipsec4_set_policy(struct inpcb *inp, int optname ,void *request,
1346 		  size_t len, int priv)
1347 {
1348 	struct sadb_x_policy *xpl;
1349 	struct secpolicy **pcb_sp;
1350 
1351 	/* sanity check. */
1352 	if (inp == NULL || request == NULL)
1353 		return EINVAL;
1354 	if (len < sizeof(*xpl))
1355 		return EINVAL;
1356 	xpl = (struct sadb_x_policy *)request;
1357 
1358 	IPSEC_ASSERT(inp->inp_sp != NULL,
1359 			 ("ipsec4_set_policy(): null inp->in_sp"));
1360 
1361 	/* select direction */
1362 	switch (xpl->sadb_x_policy_dir) {
1363 	case IPSEC_DIR_INBOUND:
1364 		pcb_sp = &inp->inp_sp->sp_in;
1365 		break;
1366 	case IPSEC_DIR_OUTBOUND:
1367 		pcb_sp = &inp->inp_sp->sp_out;
1368 		break;
1369 	default:
1370 		ipseclog((LOG_ERR, "ipsec4_set_policy: invalid direction=%u\n",
1371 			xpl->sadb_x_policy_dir));
1372 		return EINVAL;
1373 	}
1374 
1375 	return ipsec_set_policy(pcb_sp, optname, request, len, priv);
1376 }
1377 
1378 int
1379 ipsec4_get_policy(struct inpcb *inp, void *request, size_t len,
1380 		  struct mbuf **mp)
1381 {
1382 	struct sadb_x_policy *xpl;
1383 	struct secpolicy *pcb_sp;
1384 
1385 	/* sanity check. */
1386 	if (inp == NULL || request == NULL || mp == NULL)
1387 		return EINVAL;
1388 	IPSEC_ASSERT(inp->inp_sp != NULL, ("ipsec4_get_policy: null inp_sp"));
1389 	if (len < sizeof(*xpl))
1390 		return EINVAL;
1391 	xpl = (struct sadb_x_policy *)request;
1392 
1393 	/* select direction */
1394 	switch (xpl->sadb_x_policy_dir) {
1395 	case IPSEC_DIR_INBOUND:
1396 		pcb_sp = inp->inp_sp->sp_in;
1397 		break;
1398 	case IPSEC_DIR_OUTBOUND:
1399 		pcb_sp = inp->inp_sp->sp_out;
1400 		break;
1401 	default:
1402 		ipseclog((LOG_ERR, "ipsec4_set_policy: invalid direction=%u\n",
1403 			xpl->sadb_x_policy_dir));
1404 		return EINVAL;
1405 	}
1406 
1407 	return ipsec_get_policy(pcb_sp, mp);
1408 }
1409 
1410 /* delete policy in PCB */
1411 int
1412 ipsec4_delete_pcbpolicy(struct inpcb *inp)
1413 {
1414 	IPSEC_ASSERT(inp != NULL, ("ipsec4_delete_pcbpolicy: null inp"));
1415 
1416 	if (inp->inp_sp == NULL)
1417 		return 0;
1418 
1419 	if (inp->inp_sp->sp_in != NULL)
1420 		KEY_FREESP(&inp->inp_sp->sp_in);
1421 
1422 	if (inp->inp_sp->sp_out != NULL)
1423 		KEY_FREESP(&inp->inp_sp->sp_out);
1424 
1425 	ipsec_delpcbpolicy(inp->inp_sp);
1426 	inp->inp_sp = NULL;
1427 
1428 	return 0;
1429 }
1430 
1431 #ifdef INET6
1432 int
1433 ipsec6_set_policy(struct in6pcb *in6p, int optname, void *request,
1434 		  size_t len, int priv)
1435 {
1436 	struct sadb_x_policy *xpl;
1437 	struct secpolicy **pcb_sp;
1438 
1439 	/* sanity check. */
1440 	if (in6p == NULL || request == NULL)
1441 		return EINVAL;
1442 	if (len < sizeof(*xpl))
1443 		return EINVAL;
1444 	xpl = (struct sadb_x_policy *)request;
1445 
1446 	/* select direction */
1447 	switch (xpl->sadb_x_policy_dir) {
1448 	case IPSEC_DIR_INBOUND:
1449 		pcb_sp = &in6p->in6p_sp->sp_in;
1450 		break;
1451 	case IPSEC_DIR_OUTBOUND:
1452 		pcb_sp = &in6p->in6p_sp->sp_out;
1453 		break;
1454 	default:
1455 		ipseclog((LOG_ERR, "ipsec6_set_policy: invalid direction=%u\n",
1456 			xpl->sadb_x_policy_dir));
1457 		return EINVAL;
1458 	}
1459 
1460 	return ipsec_set_policy(pcb_sp, optname, request, len, priv);
1461 }
1462 
1463 int
1464 ipsec6_get_policy(struct in6pcb *in6p, void *request, size_t len,
1465 		  struct mbuf **mp)
1466 {
1467 	struct sadb_x_policy *xpl;
1468 	struct secpolicy *pcb_sp;
1469 
1470 	/* sanity check. */
1471 	if (in6p == NULL || request == NULL || mp == NULL)
1472 		return EINVAL;
1473 	IPSEC_ASSERT(in6p->in6p_sp != NULL, ("ipsec6_get_policy: null in6p_sp"));
1474 	if (len < sizeof(*xpl))
1475 		return EINVAL;
1476 	xpl = (struct sadb_x_policy *)request;
1477 
1478 	/* select direction */
1479 	switch (xpl->sadb_x_policy_dir) {
1480 	case IPSEC_DIR_INBOUND:
1481 		pcb_sp = in6p->in6p_sp->sp_in;
1482 		break;
1483 	case IPSEC_DIR_OUTBOUND:
1484 		pcb_sp = in6p->in6p_sp->sp_out;
1485 		break;
1486 	default:
1487 		ipseclog((LOG_ERR, "ipsec6_set_policy: invalid direction=%u\n",
1488 			xpl->sadb_x_policy_dir));
1489 		return EINVAL;
1490 	}
1491 
1492 	return ipsec_get_policy(pcb_sp, mp);
1493 }
1494 
1495 int
1496 ipsec6_delete_pcbpolicy(struct in6pcb *in6p)
1497 {
1498 	IPSEC_ASSERT(in6p != NULL, ("ipsec6_delete_pcbpolicy: null in6p"));
1499 
1500 	if (in6p->in6p_sp == NULL)
1501 		return 0;
1502 
1503 	if (in6p->in6p_sp->sp_in != NULL)
1504 		KEY_FREESP(&in6p->in6p_sp->sp_in);
1505 
1506 	if (in6p->in6p_sp->sp_out != NULL)
1507 		KEY_FREESP(&in6p->in6p_sp->sp_out);
1508 
1509 	ipsec_delpcbpolicy(in6p->in6p_sp);
1510 	in6p->in6p_sp = NULL;
1511 
1512 	return 0;
1513 }
1514 #endif
1515 
1516 /*
1517  * return current level.
1518  * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
1519  */
1520 u_int
1521 ipsec_get_reqlevel(struct ipsecrequest *isr)
1522 {
1523 	u_int level = 0;
1524 	u_int esp_trans_deflev, esp_net_deflev;
1525 	u_int ah_trans_deflev, ah_net_deflev;
1526 
1527 	IPSEC_ASSERT(isr != NULL && isr->sp != NULL,
1528 		("ipsec_get_reqlevel: null argument"));
1529 	IPSEC_ASSERT(isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family,
1530 		("ipsec_get_reqlevel: af family mismatch, src %u, dst %u",
1531 		 isr->sp->spidx.src.sa.sa_family,
1532 		 isr->sp->spidx.dst.sa.sa_family));
1533 
1534 /* XXX note that we have ipseclog() expanded here - code sync issue */
1535 #define IPSEC_CHECK_DEFAULT(lev) \
1536 	(((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE		  \
1537 			&& (lev) != IPSEC_LEVEL_UNIQUE)				  \
1538 		? (ipsec_debug							  \
1539 			? log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
1540 				(lev), IPSEC_LEVEL_REQUIRE)			  \
1541 			: 0),							  \
1542 			(lev) = IPSEC_LEVEL_REQUIRE,				  \
1543 			(lev)							  \
1544 		: (lev))
1545 
1546 	/* set default level */
1547 	switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) {
1548 #ifdef INET
1549 	case AF_INET:
1550 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(ip4_esp_trans_deflev);
1551 		esp_net_deflev = IPSEC_CHECK_DEFAULT(ip4_esp_net_deflev);
1552 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(ip4_ah_trans_deflev);
1553 		ah_net_deflev = IPSEC_CHECK_DEFAULT(ip4_ah_net_deflev);
1554 		break;
1555 #endif
1556 #ifdef INET6
1557 	case AF_INET6:
1558 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(ip6_esp_trans_deflev);
1559 		esp_net_deflev = IPSEC_CHECK_DEFAULT(ip6_esp_net_deflev);
1560 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(ip6_ah_trans_deflev);
1561 		ah_net_deflev = IPSEC_CHECK_DEFAULT(ip6_ah_net_deflev);
1562 		break;
1563 #endif /* INET6 */
1564 	default:
1565 		panic("key_get_reqlevel: unknown af %u",
1566 			isr->sp->spidx.src.sa.sa_family);
1567 	}
1568 
1569 #undef IPSEC_CHECK_DEFAULT
1570 
1571 	/* set level */
1572 	switch (isr->level) {
1573 	case IPSEC_LEVEL_DEFAULT:
1574 		switch (isr->saidx.proto) {
1575 		case IPPROTO_ESP:
1576 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1577 				level = esp_net_deflev;
1578 			else
1579 				level = esp_trans_deflev;
1580 			break;
1581 		case IPPROTO_AH:
1582 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1583 				level = ah_net_deflev;
1584 			else
1585 				level = ah_trans_deflev;
1586 			break;
1587 		case IPPROTO_IPCOMP:
1588 			/*
1589 			 * we don't really care, as IPcomp document says that
1590 			 * we shouldn't compress small packets
1591 			 */
1592 			level = IPSEC_LEVEL_USE;
1593 			break;
1594 		default:
1595 			panic("ipsec_get_reqlevel: Illegal protocol defined %u",
1596 				isr->saidx.proto);
1597 		}
1598 		break;
1599 
1600 	case IPSEC_LEVEL_USE:
1601 	case IPSEC_LEVEL_REQUIRE:
1602 		level = isr->level;
1603 		break;
1604 	case IPSEC_LEVEL_UNIQUE:
1605 		level = IPSEC_LEVEL_REQUIRE;
1606 		break;
1607 
1608 	default:
1609 		panic("ipsec_get_reqlevel: Illegal IPsec level %u",
1610 			isr->level);
1611 	}
1612 
1613 	return level;
1614 }
1615 
1616 /*
1617  * Check security policy requirements against the actual
1618  * packet contents.  Return one if the packet should be
1619  * reject as "invalid"; otherwiser return zero to have the
1620  * packet treated as "valid".
1621  *
1622  * OUT:
1623  *	0: valid
1624  *	1: invalid
1625  */
1626 int
1627 ipsec_in_reject(struct secpolicy *sp, struct mbuf *m)
1628 {
1629 	struct ipsecrequest *isr;
1630 	int need_auth;
1631 
1632 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1633 		printf("ipsec_in_reject: using SP\n");
1634 		kdebug_secpolicy(sp));
1635 
1636 	/* check policy */
1637 	switch (sp->policy) {
1638 	case IPSEC_POLICY_DISCARD:
1639 		return 1;
1640 	case IPSEC_POLICY_BYPASS:
1641 	case IPSEC_POLICY_NONE:
1642 		return 0;
1643 	}
1644 
1645 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1646 		("ipsec_in_reject: invalid policy %u", sp->policy));
1647 
1648 	/* XXX should compare policy against ipsec header history */
1649 
1650 	need_auth = 0;
1651 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1652 		if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE)
1653 			continue;
1654 		switch (isr->saidx.proto) {
1655 		case IPPROTO_ESP:
1656 			if ((m->m_flags & M_DECRYPTED) == 0) {
1657 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1658 					printf("ipsec_in_reject: ESP m_flags:%x\n",
1659 						m->m_flags));
1660 				return 1;
1661 			}
1662 
1663 			if (!need_auth &&
1664 				isr->sav != NULL &&
1665 				isr->sav->tdb_authalgxform != NULL &&
1666 				(m->m_flags & M_AUTHIPDGM) == 0) {
1667 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1668 					printf("ipsec_in_reject: ESP/AH m_flags:%x\n",
1669 						m->m_flags));
1670 				return 1;
1671 			}
1672 			break;
1673 		case IPPROTO_AH:
1674 			need_auth = 1;
1675 			if ((m->m_flags & M_AUTHIPHDR) == 0) {
1676 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1677 					printf("ipsec_in_reject: AH m_flags:%x\n",
1678 						m->m_flags));
1679 				return 1;
1680 			}
1681 			break;
1682 		case IPPROTO_IPCOMP:
1683 			/*
1684 			 * we don't really care, as IPcomp document
1685 			 * says that we shouldn't compress small
1686 			 * packets, IPComp policy should always be
1687 			 * treated as being in "use" level.
1688 			 */
1689 			break;
1690 		}
1691 	}
1692 	return 0;		/* valid */
1693 }
1694 
1695 /*
1696  * Check AH/ESP integrity.
1697  * This function is called from tcp_input(), udp_input(),
1698  * and {ah,esp}4_input for tunnel mode
1699  */
1700 int
1701 ipsec4_in_reject(struct mbuf *m, struct inpcb *inp)
1702 {
1703 	struct secpolicy *sp;
1704 	int error;
1705 	int result;
1706 
1707 	IPSEC_ASSERT(m != NULL, ("ipsec4_in_reject_so: null mbuf"));
1708 
1709 	/* get SP for this packet.
1710 	 * When we are called from ip_forward(), we call
1711 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1712 	 */
1713 	if (inp == NULL)
1714 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, IP_FORWARDING, &error);
1715 	else
1716 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND,
1717 					   IN4PCB_TO_PCB(inp), &error);
1718 
1719 	if (sp != NULL) {
1720 		result = ipsec_in_reject(sp, m);
1721 		if (result)
1722 			newipsecstat.ips_in_polvio++;
1723 		KEY_FREESP(&sp);
1724 	} else {
1725 		result = 0;	/* XXX should be panic ?
1726 				 * -> No, there may be error. */
1727 	}
1728 	return result;
1729 }
1730 
1731 
1732 #ifdef INET6
1733 /*
1734  * Check AH/ESP integrity.
1735  * This function is called from tcp6_input(), udp6_input(),
1736  * and {ah,esp}6_input for tunnel mode
1737  */
1738 int
1739 ipsec6_in_reject(struct mbuf *m, struct in6pcb *in6p)
1740 {
1741 	struct secpolicy *sp = NULL;
1742 	int error;
1743 	int result;
1744 
1745 	/* sanity check */
1746 	if (m == NULL)
1747 		return 0;	/* XXX should be panic ? */
1748 
1749 	/* get SP for this packet.
1750 	 * When we are called from ip_forward(), we call
1751 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1752 	 */
1753 	if (in6p == NULL)
1754 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, IP_FORWARDING, &error);
1755 	else
1756 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND,
1757 			IN6PCB_TO_PCB(in6p),
1758 			&error);
1759 
1760 	if (sp != NULL) {
1761 		result = ipsec_in_reject(sp, m);
1762 		if (result)
1763 			newipsecstat.ips_in_polvio++;
1764 		KEY_FREESP(&sp);
1765 	} else {
1766 		result = 0;
1767 	}
1768 	return result;
1769 }
1770 #endif
1771 
1772 /*
1773  * compute the byte size to be occupied by IPsec header.
1774  * in case it is tunneled, it includes the size of outer IP header.
1775  * NOTE: SP passed is free in this function.
1776  */
1777 static size_t
1778 ipsec_hdrsiz(struct secpolicy *sp)
1779 {
1780 	struct ipsecrequest *isr;
1781 	size_t siz;
1782 
1783 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1784 		printf("ipsec_hdrsiz: using SP\n");
1785 		kdebug_secpolicy(sp));
1786 
1787 	switch (sp->policy) {
1788 	case IPSEC_POLICY_DISCARD:
1789 	case IPSEC_POLICY_BYPASS:
1790 	case IPSEC_POLICY_NONE:
1791 		return 0;
1792 	}
1793 
1794 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1795 		("ipsec_hdrsiz: invalid policy %u", sp->policy));
1796 
1797 	siz = 0;
1798 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1799 		size_t clen = 0;
1800 
1801 		switch (isr->saidx.proto) {
1802 		case IPPROTO_ESP:
1803 			clen = esp_hdrsiz(isr->sav);
1804 			break;
1805 		case IPPROTO_AH:
1806 			clen = ah_hdrsiz(isr->sav);
1807 			break;
1808 		case IPPROTO_IPCOMP:
1809 			clen = sizeof(struct ipcomp);
1810 			break;
1811 		}
1812 
1813 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1814 			switch (isr->saidx.dst.sa.sa_family) {
1815 			case AF_INET:
1816 				clen += sizeof(struct ip);
1817 				break;
1818 #ifdef INET6
1819 			case AF_INET6:
1820 				clen += sizeof(struct ip6_hdr);
1821 				break;
1822 #endif
1823 			default:
1824 				ipseclog((LOG_ERR, "ipsec_hdrsiz: "
1825 					"unknown AF %d in IPsec tunnel SA\n",
1826 					((struct sockaddr *)&isr->saidx.dst)->sa_family));
1827 				break;
1828 			}
1829 		}
1830 		siz += clen;
1831 	}
1832 
1833 	return siz;
1834 }
1835 
1836 /* This function is called from ip_forward() and ipsec4_hdrsize_tcp(). */
1837 size_t
1838 ipsec4_hdrsiz(struct mbuf *m, u_int dir, struct inpcb *inp)
1839 {
1840 	struct secpolicy *sp;
1841 	int error;
1842 	size_t size;
1843 
1844 	IPSEC_ASSERT(m != NULL, ("ipsec4_hdrsiz: null mbuf"));
1845 	IPSEC_ASSERT(inp == NULL || inp->inp_socket != NULL,
1846 		("ipsec4_hdrsize: socket w/o inpcb"));
1847 
1848 	/* get SP for this packet.
1849 	 * When we are called from ip_forward(), we call
1850 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1851 	 */
1852 	if (inp == NULL)
1853 		sp = ipsec_getpolicybyaddr(m, dir, IP_FORWARDING, &error);
1854 	else
1855 		sp = ipsec_getpolicybysock(m, dir,
1856 					   IN4PCB_TO_PCB(inp), &error);
1857 
1858 	if (sp != NULL) {
1859 		size = ipsec_hdrsiz(sp);
1860 		KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1861 			printf("ipsec4_hdrsiz: size:%lu.\n",
1862 				(unsigned long)size));
1863 
1864 		KEY_FREESP(&sp);
1865 	} else {
1866 		size = 0;	/* XXX should be panic ? */
1867 	}
1868 	return size;
1869 }
1870 
1871 #ifdef INET6
1872 /* This function is called from ipsec6_hdrsize_tcp(),
1873  * and maybe from ip6_forward.()
1874  */
1875 size_t
1876 ipsec6_hdrsiz(struct mbuf *m, u_int dir, struct in6pcb *in6p)
1877 {
1878 	struct secpolicy *sp;
1879 	int error;
1880 	size_t size;
1881 
1882 	IPSEC_ASSERT(m != NULL, ("ipsec6_hdrsiz: null mbuf"));
1883 	IPSEC_ASSERT(in6p == NULL || in6p->in6p_socket != NULL,
1884 		("ipsec6_hdrsize: socket w/o inpcb"));
1885 
1886 	/* get SP for this packet */
1887 	/* XXX Is it right to call with IP_FORWARDING. */
1888 	if (in6p == NULL)
1889 		sp = ipsec_getpolicybyaddr(m, dir, IP_FORWARDING, &error);
1890 	else
1891 		sp = ipsec_getpolicybysock(m, dir,
1892 			IN6PCB_TO_PCB(in6p),
1893 			&error);
1894 
1895 	if (sp == NULL)
1896 		return 0;
1897 	size = ipsec_hdrsiz(sp);
1898 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1899 		printf("ipsec6_hdrsiz: size:%lu.\n", (unsigned long)size));
1900 	KEY_FREESP(&sp);
1901 
1902 	return size;
1903 }
1904 #endif /*INET6*/
1905 
1906 /*
1907  * Check the variable replay window.
1908  * ipsec_chkreplay() performs replay check before ICV verification.
1909  * ipsec_updatereplay() updates replay bitmap.  This must be called after
1910  * ICV verification (it also performs replay check, which is usually done
1911  * beforehand).
1912  * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1913  *
1914  * based on RFC 2401.
1915  */
1916 int
1917 ipsec_chkreplay(u_int32_t seq, struct secasvar *sav)
1918 {
1919 	const struct secreplay *replay;
1920 	u_int32_t diff;
1921 	int fr;
1922 	u_int32_t wsizeb;	/* constant: bits of window size */
1923 	int frlast;		/* constant: last frame */
1924 
1925 	IPSEC_SPLASSERT_SOFTNET("ipsec_chkreplay");
1926 
1927 	IPSEC_ASSERT(sav != NULL, ("ipsec_chkreplay: Null SA"));
1928 	IPSEC_ASSERT(sav->replay != NULL, ("ipsec_chkreplay: Null replay state"));
1929 
1930 	replay = sav->replay;
1931 
1932 	if (replay->wsize == 0)
1933 		return 1;	/* no need to check replay. */
1934 
1935 	/* constant */
1936 	frlast = replay->wsize - 1;
1937 	wsizeb = replay->wsize << 3;
1938 
1939 	/* sequence number of 0 is invalid */
1940 	if (seq == 0)
1941 		return 0;
1942 
1943 	/* first time is always okay */
1944 	if (replay->count == 0)
1945 		return 1;
1946 
1947 	if (seq > replay->lastseq) {
1948 		/* larger sequences are okay */
1949 		return 1;
1950 	} else {
1951 		/* seq is equal or less than lastseq. */
1952 		diff = replay->lastseq - seq;
1953 
1954 		/* over range to check, i.e. too old or wrapped */
1955 		if (diff >= wsizeb)
1956 			return 0;
1957 
1958 		fr = frlast - diff / 8;
1959 
1960 		/* this packet already seen ? */
1961 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1962 			return 0;
1963 
1964 		/* out of order but good */
1965 		return 1;
1966 	}
1967 }
1968 
1969 /*
1970  * check replay counter whether to update or not.
1971  * OUT:	0:	OK
1972  *	1:	NG
1973  */
1974 int
1975 ipsec_updatereplay(u_int32_t seq, struct secasvar *sav)
1976 {
1977 	struct secreplay *replay;
1978 	u_int32_t diff;
1979 	int fr;
1980 	u_int32_t wsizeb;	/* constant: bits of window size */
1981 	int frlast;		/* constant: last frame */
1982 
1983 	IPSEC_SPLASSERT_SOFTNET("ipsec_updatereplay");
1984 
1985 	IPSEC_ASSERT(sav != NULL, ("ipsec_updatereplay: Null SA"));
1986 	IPSEC_ASSERT(sav->replay != NULL, ("ipsec_updatereplay: Null replay state"));
1987 
1988 	replay = sav->replay;
1989 
1990 	if (replay->wsize == 0)
1991 		goto ok;	/* no need to check replay. */
1992 
1993 	/* constant */
1994 	frlast = replay->wsize - 1;
1995 	wsizeb = replay->wsize << 3;
1996 
1997 	/* sequence number of 0 is invalid */
1998 	if (seq == 0)
1999 		return 1;
2000 
2001 	/* first time */
2002 	if (replay->count == 0) {
2003 		replay->lastseq = seq;
2004 		bzero(replay->bitmap, replay->wsize);
2005 		(replay->bitmap)[frlast] = 1;
2006 		goto ok;
2007 	}
2008 
2009 	if (seq > replay->lastseq) {
2010 		/* seq is larger than lastseq. */
2011 		diff = seq - replay->lastseq;
2012 
2013 		/* new larger sequence number */
2014 		if (diff < wsizeb) {
2015 			/* In window */
2016 			/* set bit for this packet */
2017 			vshiftl(replay->bitmap, diff, replay->wsize);
2018 			(replay->bitmap)[frlast] |= 1;
2019 		} else {
2020 			/* this packet has a "way larger" */
2021 			bzero(replay->bitmap, replay->wsize);
2022 			(replay->bitmap)[frlast] = 1;
2023 		}
2024 		replay->lastseq = seq;
2025 
2026 		/* larger is good */
2027 	} else {
2028 		/* seq is equal or less than lastseq. */
2029 		diff = replay->lastseq - seq;
2030 
2031 		/* over range to check, i.e. too old or wrapped */
2032 		if (diff >= wsizeb)
2033 			return 1;
2034 
2035 		fr = frlast - diff / 8;
2036 
2037 		/* this packet already seen ? */
2038 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
2039 			return 1;
2040 
2041 		/* mark as seen */
2042 		(replay->bitmap)[fr] |= (1 << (diff % 8));
2043 
2044 		/* out of order but good */
2045 	}
2046 
2047 ok:
2048 	if (replay->count == ~0) {
2049 
2050 		/* set overflow flag */
2051 		replay->overflow++;
2052 
2053 		/* don't increment, no more packets accepted */
2054 		if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0)
2055 			return 1;
2056 
2057 		ipseclog((LOG_WARNING, "replay counter made %d cycle. %s\n",
2058 			replay->overflow, ipsec_logsastr(sav)));
2059 	}
2060 
2061 	replay->count++;
2062 
2063 	return 0;
2064 }
2065 
2066 /*
2067  * shift variable length bunffer to left.
2068  * IN:	bitmap: pointer to the buffer
2069  * 	nbit:	the number of to shift.
2070  *	wsize:	buffer size (bytes).
2071  */
2072 static void
2073 vshiftl(unsigned char *bitmap, int nbit, int wsize)
2074 {
2075 	int s, j, i;
2076 	unsigned char over;
2077 
2078 	for (j = 0; j < nbit; j += 8) {
2079 		s = (nbit - j < 8) ? (nbit - j): 8;
2080 		bitmap[0] <<= s;
2081 		for (i = 1; i < wsize; i++) {
2082 			over = (bitmap[i] >> (8 - s));
2083 			bitmap[i] <<= s;
2084 			bitmap[i-1] |= over;
2085 		}
2086 	}
2087 
2088 	return;
2089 }
2090 
2091 /* Return a printable string for the IPv4 address. */
2092 static char *
2093 inet_ntoa4(struct in_addr ina)
2094 {
2095 	static char buf[4][4 * sizeof "123" + 4];
2096 	unsigned char *ucp = (unsigned char *) &ina;
2097 	static int i = 3;
2098 
2099 	i = (i + 1) % 4;
2100 	snprintf(buf[i], sizeof(buf[i]), "%d.%d.%d.%d",
2101 		ucp[0] & 0xff, ucp[1] & 0xff, ucp[2] & 0xff, ucp[3] & 0xff);
2102 	return (buf[i]);
2103 }
2104 
2105 /* Return a printable string for the address. */
2106 const char *
2107 ipsec_address(union sockaddr_union* sa)
2108 {
2109 	switch (sa->sa.sa_family) {
2110 #if INET
2111 	case AF_INET:
2112 		return inet_ntoa4(sa->sin.sin_addr);
2113 #endif /* INET */
2114 
2115 #if INET6
2116 	case AF_INET6:
2117 		return ip6_sprintf(&sa->sin6.sin6_addr);
2118 #endif /* INET6 */
2119 
2120 	default:
2121 		return "(unknown address family)";
2122 	}
2123 }
2124 
2125 const char *
2126 ipsec_logsastr(struct secasvar *sav)
2127 {
2128 	static char buf[256];
2129 	char *p;
2130 	struct secasindex *saidx = &sav->sah->saidx;
2131 
2132 	IPSEC_ASSERT(saidx->src.sa.sa_family == saidx->dst.sa.sa_family,
2133 		("ipsec_logsastr: address family mismatch"));
2134 
2135 	p = buf;
2136 	snprintf(buf, sizeof(buf), "SA(SPI=%u ", (u_int32_t)ntohl(sav->spi));
2137 	while (p && *p)
2138 		p++;
2139 	/* NB: only use ipsec_address on one address at a time */
2140 	snprintf(p, sizeof (buf) - (p - buf), "src=%s ",
2141 		ipsec_address(&saidx->src));
2142 	while (p && *p)
2143 		p++;
2144 	snprintf(p, sizeof (buf) - (p - buf), "dst=%s)",
2145 		ipsec_address(&saidx->dst));
2146 
2147 	return buf;
2148 }
2149 
2150 void
2151 ipsec_dumpmbuf(struct mbuf *m)
2152 {
2153 	int totlen;
2154 	int i;
2155 	u_char *p;
2156 
2157 	totlen = 0;
2158 	printf("---\n");
2159 	while (m) {
2160 		p = mtod(m, u_char *);
2161 		for (i = 0; i < m->m_len; i++) {
2162 			printf("%02x ", p[i]);
2163 			totlen++;
2164 			if (totlen % 16 == 0)
2165 				printf("\n");
2166 		}
2167 		m = m->m_next;
2168 	}
2169 	if (totlen % 16 != 0)
2170 		printf("\n");
2171 	printf("---\n");
2172 }
2173 
2174 #ifdef INET6
2175 struct secpolicy *
2176 ipsec6_check_policy(struct mbuf * m, const struct socket * so,
2177 		    int flags, int * needipsecp, int * errorp)
2178 {
2179 	struct in6pcb *in6p = NULL;
2180 	struct m_tag *mtag;
2181 	struct secpolicy *sp = NULL;
2182 	struct tdb_ident *tdbi;
2183 	int s;
2184 	int error = 0;
2185 	int needipsec = 0;
2186 
2187 	if (so != NULL && so->so_proto->pr_domain->dom_family == AF_INET6)
2188 		in6p = sotoin6pcb(so);
2189 
2190 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
2191 	s = splsoftnet();
2192 	if (mtag != NULL) {
2193 		tdbi = (struct tdb_ident *)(mtag + 1);
2194 		sp = ipsec_getpolicy(tdbi, IPSEC_DIR_OUTBOUND);
2195 		if (sp == NULL)
2196 			error = -EINVAL;	/* force silent drop */
2197 		m_tag_delete(m, mtag);
2198 	} else {
2199 		if (in6p != NULL &&
2200 			IPSEC_PCB_SKIP_IPSEC(in6p->in6p_sp, IPSEC_DIR_OUTBOUND))
2201 			goto skippolicycheck;
2202 		sp = ipsec6_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, &error,in6p);
2203 	}
2204 
2205 	/*
2206 	 * There are four return cases:
2207 	 *	sp != NULL			apply IPsec policy
2208 	 *	sp == NULL, error == 0		no IPsec handling needed
2209 	 *	sp == NULL, error == -EINVAL  discard packet w/o error
2210 	 *	sp == NULL, error != 0		discard packet, report error
2211  	 */
2212 
2213 	if (sp == NULL) {
2214 		splx(s);
2215 
2216 		if (error != 0) {
2217 			needipsec = 0;
2218 		} else {
2219 			/* No IPsec processing for this packet. */
2220 			needipsec = 0;
2221 		}
2222 	} else {
2223 		/* Loop detection, check if ipsec processing already done */
2224 		IPSEC_ASSERT(sp->req != NULL, ("ip6_output: no ipsec request"));
2225 		for (mtag = m_tag_first(m); mtag != NULL;
2226 			 mtag = m_tag_next(m, mtag)) {
2227 #ifdef MTAG_ABI_COMPAT
2228 			if (mtag->m_tag_cookie != MTAG_ABI_COMPAT)
2229 				continue;
2230 #endif
2231 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
2232 				mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
2233 				continue;
2234 			/*
2235 			 * Check if policy has an SA associated with it.
2236 			 * This can happen when an SP has yet to acquire
2237 			 * an SA; e.g. on first reference.  If it occurs,
2238 			 * then we let ipsec4_process_packet do its thing.
2239 			 */
2240 			if (sp->req->sav == NULL)
2241 				break;
2242 			tdbi = (struct tdb_ident *)(mtag + 1);
2243 			if (tdbi->spi == sp->req->sav->spi &&
2244 				tdbi->proto == sp->req->sav->sah->saidx.proto &&
2245 				bcmp(&tdbi->dst, &sp->req->sav->sah->saidx.dst,
2246 				 sizeof (union sockaddr_union)) == 0) {
2247 				/*
2248 				 * No IPsec processing is needed, free
2249 				 * reference to SP.
2250 				 *
2251 				 * NB: null pointer to avoid free at
2252 				 *	 done: below.
2253 				 */
2254 				KEY_FREESP(&sp), sp = NULL;
2255 				needipsec = 0;
2256 				splx(s);
2257 				goto skippolicycheck;
2258 			}
2259 		}
2260 		splx(s);
2261 		needipsec = 1;
2262 	}
2263 skippolicycheck:;
2264 
2265 	*errorp = error;
2266 	*needipsecp = needipsec;
2267 	return sp;
2268 }
2269 #endif
2270 
2271 
2272 
2273 /* XXX this stuff doesn't belong here... */
2274 
2275 static	struct xformsw* xforms = NULL;
2276 
2277 /*
2278  * Register a transform; typically at system startup.
2279  */
2280 void
2281 xform_register(struct xformsw* xsp)
2282 {
2283 	xsp->xf_next = xforms;
2284 	xforms = xsp;
2285 }
2286 
2287 /*
2288  * Initialize transform support in an sav.
2289  */
2290 int
2291 xform_init(struct secasvar *sav, int xftype)
2292 {
2293 	struct xformsw *xsp;
2294 
2295 	if (sav->tdb_xform != NULL)	/* previously initialized */
2296 		return 0;
2297 	for (xsp = xforms; xsp; xsp = xsp->xf_next)
2298 		if (xsp->xf_type == xftype)
2299 			return (*xsp->xf_init)(sav, xsp);
2300 
2301 	DPRINTF(("xform_init: no match for xform type %d\n", xftype));
2302 	return EINVAL;
2303 }
2304 
2305 #ifdef __NetBSD__
2306 void
2307 ipsec_attach(void)
2308 {
2309 	printf("initializing IPsec...");
2310 	ah_attach();
2311 	esp_attach();
2312 	ipcomp_attach();
2313 	ipe4_attach();
2314 #ifdef TCP_SIGNATURE
2315 	tcpsignature_attach();
2316 #endif
2317 	printf(" done\n");
2318 }
2319 #endif	/* __NetBSD__ */
2320