xref: /freebsd/sys/netipsec/ipsec.c (revision 79b015331632b33f22bc8d8fef136c2672f2241f)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ipsec.c,v 1.103 2001/05/24 07:14:18 sakane Exp $	*/
3 
4 /*-
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * IPsec controller part.
35  */
36 
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39 #include "opt_ipsec.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/domain.h>
46 #include <sys/priv.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/errno.h>
51 #include <sys/time.h>
52 #include <sys/kernel.h>
53 #include <sys/syslog.h>
54 #include <sys/sysctl.h>
55 #include <sys/proc.h>
56 
57 #include <net/if.h>
58 #include <net/if_var.h>
59 #include <net/vnet.h>
60 
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/ip.h>
64 #include <netinet/ip_var.h>
65 #include <netinet/in_var.h>
66 #include <netinet/udp.h>
67 #include <netinet/udp_var.h>
68 #include <netinet/tcp.h>
69 #include <netinet/udp.h>
70 
71 #include <netinet/ip6.h>
72 #ifdef INET6
73 #include <netinet6/ip6_var.h>
74 #endif
75 #include <netinet/in_pcb.h>
76 #ifdef INET6
77 #include <netinet/icmp6.h>
78 #endif
79 
80 #include <sys/types.h>
81 #include <netipsec/ipsec.h>
82 #ifdef INET6
83 #include <netipsec/ipsec6.h>
84 #endif
85 #include <netipsec/ah_var.h>
86 #include <netipsec/esp_var.h>
87 #include <netipsec/ipcomp.h>		/*XXX*/
88 #include <netipsec/ipcomp_var.h>
89 
90 #include <netipsec/key.h>
91 #include <netipsec/keydb.h>
92 #include <netipsec/key_debug.h>
93 
94 #include <netipsec/xform.h>
95 
96 #include <machine/in_cksum.h>
97 
98 #include <opencrypto/cryptodev.h>
99 
100 #ifdef IPSEC_DEBUG
101 VNET_DEFINE(int, ipsec_debug) = 1;
102 #else
103 VNET_DEFINE(int, ipsec_debug) = 0;
104 #endif
105 
106 /* NB: name changed so netstat doesn't use it. */
107 VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec4stat);
108 VNET_PCPUSTAT_SYSINIT(ipsec4stat);
109 
110 #ifdef VIMAGE
111 VNET_PCPUSTAT_SYSUNINIT(ipsec4stat);
112 #endif /* VIMAGE */
113 
114 VNET_DEFINE(int, ip4_ah_offsetmask) = 0;	/* maybe IP_DF? */
115 /* DF bit on encap. 0: clear 1: set 2: copy */
116 VNET_DEFINE(int, ip4_ipsec_dfbit) = 0;
117 VNET_DEFINE(int, ip4_esp_trans_deflev) = IPSEC_LEVEL_USE;
118 VNET_DEFINE(int, ip4_esp_net_deflev) = IPSEC_LEVEL_USE;
119 VNET_DEFINE(int, ip4_ah_trans_deflev) = IPSEC_LEVEL_USE;
120 VNET_DEFINE(int, ip4_ah_net_deflev) = IPSEC_LEVEL_USE;
121 /* ECN ignore(-1)/forbidden(0)/allowed(1) */
122 VNET_DEFINE(int, ip4_ipsec_ecn) = 0;
123 VNET_DEFINE(int, ip4_esp_randpad) = -1;
124 
125 static VNET_DEFINE(struct secpolicy, def_policy);
126 #define	V_def_policy	VNET(def_policy)
127 /*
128  * Crypto support requirements:
129  *
130  *  1	require hardware support
131  * -1	require software support
132  *  0	take anything
133  */
134 VNET_DEFINE(int, crypto_support) = CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE;
135 
136 FEATURE(ipsec, "Internet Protocol Security (IPsec)");
137 #ifdef IPSEC_NAT_T
138 FEATURE(ipsec_natt, "UDP Encapsulation of IPsec ESP Packets ('NAT-T')");
139 #endif
140 
141 SYSCTL_DECL(_net_inet_ipsec);
142 
143 /* net.inet.ipsec */
144 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_POLICY, def_policy,
145 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(def_policy).policy, 0,
146 	"IPsec default policy.");
147 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
148 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_esp_trans_deflev), 0,
149 	"Default ESP transport mode level");
150 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
151 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_esp_net_deflev), 0,
152 	"Default ESP tunnel mode level.");
153 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
154 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_trans_deflev), 0,
155 	"AH transfer mode default level.");
156 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
157 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_net_deflev), 0,
158 	"AH tunnel mode default level.");
159 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_CLEARTOS, ah_cleartos,
160 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ah_cleartos), 0,
161 	"If set clear type-of-service field when doing AH computation.");
162 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_OFFSETMASK, ah_offsetmask,
163 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_offsetmask), 0,
164 	"If not set clear offset field mask when doing AH computation.");
165 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DFBIT, dfbit,
166 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_dfbit), 0,
167 	"Do not fragment bit on encap.");
168 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_ECN, ecn,
169 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_ecn), 0,
170 	"Explicit Congestion Notification handling.");
171 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
172 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
173 	"Enable IPsec debugging output when set.");
174 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, crypto_support,
175 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(crypto_support), 0,
176 	"Crypto driver selection.");
177 SYSCTL_VNET_PCPUSTAT(_net_inet_ipsec, OID_AUTO, ipsecstats, struct ipsecstat,
178     ipsec4stat, "IPsec IPv4 statistics.");
179 
180 #ifdef REGRESSION
181 /*
182  * When set to 1, IPsec will send packets with the same sequence number.
183  * This allows to verify if the other side has proper replay attacks detection.
184  */
185 VNET_DEFINE(int, ipsec_replay) = 0;
186 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_replay,
187 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_replay), 0,
188 	"Emulate replay attack");
189 /*
190  * When set 1, IPsec will send packets with corrupted HMAC.
191  * This allows to verify if the other side properly detects modified packets.
192  */
193 VNET_DEFINE(int, ipsec_integrity) = 0;
194 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_integrity,
195 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_integrity), 0,
196 	"Emulate man-in-the-middle attack");
197 #endif
198 
199 #ifdef INET6
200 VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec6stat);
201 VNET_PCPUSTAT_SYSINIT(ipsec6stat);
202 
203 #ifdef VIMAGE
204 VNET_PCPUSTAT_SYSUNINIT(ipsec6stat);
205 #endif /* VIMAGE */
206 
207 VNET_DEFINE(int, ip6_esp_trans_deflev) = IPSEC_LEVEL_USE;
208 VNET_DEFINE(int, ip6_esp_net_deflev) = IPSEC_LEVEL_USE;
209 VNET_DEFINE(int, ip6_ah_trans_deflev) = IPSEC_LEVEL_USE;
210 VNET_DEFINE(int, ip6_ah_net_deflev) = IPSEC_LEVEL_USE;
211 VNET_DEFINE(int, ip6_ipsec_ecn) = 0;	/* ECN ignore(-1)/forbidden(0)/allowed(1) */
212 
213 SYSCTL_DECL(_net_inet6_ipsec6);
214 
215 /* net.inet6.ipsec6 */
216 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_POLICY, def_policy,
217 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(def_policy).policy, 0,
218 	"IPsec default policy.");
219 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
220 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_esp_trans_deflev), 0,
221 	"Default ESP transport mode level.");
222 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
223 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_esp_net_deflev), 0,
224 	"Default ESP tunnel mode level.");
225 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
226 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ah_trans_deflev), 0,
227 	"AH transfer mode default level.");
228 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
229 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ah_net_deflev), 0,
230 	"AH tunnel mode default level.");
231 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_ECN, ecn,
232 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ipsec_ecn), 0,
233 	"Explicit Congestion Notification handling.");
234 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug,
235 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
236 	"Enable IPsec debugging output when set.");
237 SYSCTL_VNET_PCPUSTAT(_net_inet6_ipsec6, IPSECCTL_STATS, ipsecstats,
238     struct ipsecstat, ipsec6stat, "IPsec IPv6 statistics.");
239 #endif /* INET6 */
240 
241 static int ipsec_in_reject(struct secpolicy *, struct mbuf *);
242 static int ipsec_setspidx_inpcb(struct mbuf *, struct inpcb *);
243 static int ipsec_setspidx(struct mbuf *, struct secpolicyindex *, int);
244 static void ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *, int);
245 static int ipsec4_setspidx_ipaddr(struct mbuf *, struct secpolicyindex *);
246 #ifdef INET6
247 static void ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *, int);
248 static int ipsec6_setspidx_ipaddr(struct mbuf *, struct secpolicyindex *);
249 #endif
250 static void ipsec_delpcbpolicy(struct inpcbpolicy *);
251 static struct secpolicy *ipsec_deepcopy_policy(struct secpolicy *src);
252 static void vshiftl(unsigned char *, int, int);
253 
254 MALLOC_DEFINE(M_IPSEC_INPCB, "inpcbpolicy", "inpcb-resident ipsec policy");
255 
256 /*
257  * Return a held reference to the default SP.
258  */
259 static struct secpolicy *
260 key_allocsp_default(const char* where, int tag)
261 {
262 	struct secpolicy *sp;
263 
264 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
265 		printf("DP key_allocsp_default from %s:%u\n", where, tag));
266 
267 	sp = &V_def_policy;
268 	if (sp->policy != IPSEC_POLICY_DISCARD &&
269 	    sp->policy != IPSEC_POLICY_NONE) {
270 		ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n",
271 		    sp->policy, IPSEC_POLICY_NONE));
272 		sp->policy = IPSEC_POLICY_NONE;
273 	}
274 	key_addref(sp);
275 
276 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
277 		printf("DP key_allocsp_default returns SP:%p (%u)\n",
278 			sp, sp->refcnt));
279 	return (sp);
280 }
281 #define	KEY_ALLOCSP_DEFAULT() \
282 	key_allocsp_default(__FILE__, __LINE__)
283 
284 /*
285  * For OUTBOUND packet having a socket. Searching SPD for packet,
286  * and return a pointer to SP.
287  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
288  *		0	: bypass
289  *		EACCES	: discard packet.
290  *		ENOENT	: ipsec_acquire() in progress, maybe.
291  *		others	: error occured.
292  *	others:	a pointer to SP
293  *
294  * NOTE: IPv6 mapped adddress concern is implemented here.
295  */
296 struct secpolicy *
297 ipsec_getpolicy(struct tdb_ident *tdbi, u_int dir)
298 {
299 	struct secpolicy *sp;
300 
301 	IPSEC_ASSERT(tdbi != NULL, ("null tdbi"));
302 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
303 		("invalid direction %u", dir));
304 
305 	sp = KEY_ALLOCSP2(tdbi->spi, &tdbi->dst, tdbi->proto, dir);
306 	if (sp == NULL)			/*XXX????*/
307 		sp = KEY_ALLOCSP_DEFAULT();
308 	IPSEC_ASSERT(sp != NULL, ("null SP"));
309 	return (sp);
310 }
311 
312 /*
313  * For OUTBOUND packet having a socket. Searching SPD for packet,
314  * and return a pointer to SP.
315  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
316  *		0	: bypass
317  *		EACCES	: discard packet.
318  *		ENOENT	: ipsec_acquire() in progress, maybe.
319  *		others	: error occured.
320  *	others:	a pointer to SP
321  *
322  * NOTE: IPv6 mapped adddress concern is implemented here.
323  */
324 static struct secpolicy *
325 ipsec_getpolicybysock(struct mbuf *m, u_int dir, struct inpcb *inp, int *error)
326 {
327 	struct inpcbpolicy *pcbsp;
328 	struct secpolicy *currsp = NULL;	/* Policy on socket. */
329 	struct secpolicy *sp;
330 
331 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
332 	IPSEC_ASSERT(inp != NULL, ("null inpcb"));
333 	IPSEC_ASSERT(error != NULL, ("null error"));
334 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
335 		("invalid direction %u", dir));
336 
337 	if (!key_havesp(dir)) {
338 		/* No SP found, use system default. */
339 		sp = KEY_ALLOCSP_DEFAULT();
340 		return (sp);
341 	}
342 
343 	/* Set spidx in pcb. */
344 	*error = ipsec_setspidx_inpcb(m, inp);
345 	if (*error)
346 		return (NULL);
347 
348 	pcbsp = inp->inp_sp;
349 	IPSEC_ASSERT(pcbsp != NULL, ("null pcbsp"));
350 	switch (dir) {
351 	case IPSEC_DIR_INBOUND:
352 		currsp = pcbsp->sp_in;
353 		break;
354 	case IPSEC_DIR_OUTBOUND:
355 		currsp = pcbsp->sp_out;
356 		break;
357 	}
358 	IPSEC_ASSERT(currsp != NULL, ("null currsp"));
359 
360 	if (pcbsp->priv) {			/* When privilieged socket. */
361 		switch (currsp->policy) {
362 		case IPSEC_POLICY_BYPASS:
363 		case IPSEC_POLICY_IPSEC:
364 			key_addref(currsp);
365 			sp = currsp;
366 			break;
367 
368 		case IPSEC_POLICY_ENTRUST:
369 			/* Look for a policy in SPD. */
370 			sp = KEY_ALLOCSP(&currsp->spidx, dir);
371 			if (sp == NULL)		/* No SP found. */
372 				sp = KEY_ALLOCSP_DEFAULT();
373 			break;
374 
375 		default:
376 			ipseclog((LOG_ERR, "%s: Invalid policy for PCB %d\n",
377 				__func__, currsp->policy));
378 			*error = EINVAL;
379 			return (NULL);
380 		}
381 	} else {				/* Unpriv, SPD has policy. */
382 		sp = KEY_ALLOCSP(&currsp->spidx, dir);
383 		if (sp == NULL) {		/* No SP found. */
384 			switch (currsp->policy) {
385 			case IPSEC_POLICY_BYPASS:
386 				ipseclog((LOG_ERR, "%s: Illegal policy for "
387 					"non-priviliged defined %d\n",
388 					__func__, currsp->policy));
389 				*error = EINVAL;
390 				return (NULL);
391 
392 			case IPSEC_POLICY_ENTRUST:
393 				sp = KEY_ALLOCSP_DEFAULT();
394 				break;
395 
396 			case IPSEC_POLICY_IPSEC:
397 				key_addref(currsp);
398 				sp = currsp;
399 				break;
400 
401 			default:
402 				ipseclog((LOG_ERR, "%s: Invalid policy for "
403 					"PCB %d\n", __func__, currsp->policy));
404 				*error = EINVAL;
405 				return (NULL);
406 			}
407 		}
408 	}
409 	IPSEC_ASSERT(sp != NULL,
410 		("null SP (priv %u policy %u", pcbsp->priv, currsp->policy));
411 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
412 		printf("DP %s (priv %u policy %u) allocate SP:%p (refcnt %u)\n",
413 			__func__, pcbsp->priv, currsp->policy, sp, sp->refcnt));
414 	return (sp);
415 }
416 
417 /*
418  * For FORWADING packet or OUTBOUND without a socket. Searching SPD for packet,
419  * and return a pointer to SP.
420  * OUT:	positive: a pointer to the entry for security policy leaf matched.
421  *	NULL:	no apropreate SP found, the following value is set to error.
422  *		0	: bypass
423  *		EACCES	: discard packet.
424  *		ENOENT	: ipsec_acquire() in progress, maybe.
425  *		others	: error occured.
426  */
427 struct secpolicy *
428 ipsec_getpolicybyaddr(struct mbuf *m, u_int dir, int *error)
429 {
430 	struct secpolicyindex spidx;
431 	struct secpolicy *sp;
432 
433 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
434 	IPSEC_ASSERT(error != NULL, ("null error"));
435 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
436 		("invalid direction %u", dir));
437 
438 	sp = NULL;
439 	*error = 0;
440 	if (key_havesp(dir)) {
441 		/* Make an index to look for a policy. */
442 		*error = ipsec_setspidx(m, &spidx, 0);
443 		if (*error != 0) {
444 			DPRINTF(("%s: setpidx failed, dir %u\n",
445 				__func__, dir));
446 			return (NULL);
447 		}
448 		spidx.dir = dir;
449 		sp = KEY_ALLOCSP(&spidx, dir);
450 	}
451 	if (sp == NULL)			/* No SP found, use system default. */
452 		sp = KEY_ALLOCSP_DEFAULT();
453 	IPSEC_ASSERT(sp != NULL, ("null SP"));
454 	return (sp);
455 }
456 
457 struct secpolicy *
458 ipsec4_checkpolicy(struct mbuf *m, u_int dir, int *error, struct inpcb *inp)
459 {
460 	struct secpolicy *sp;
461 
462 	*error = 0;
463 	if (inp == NULL)
464 		sp = ipsec_getpolicybyaddr(m, dir, error);
465 	else
466 		sp = ipsec_getpolicybysock(m, dir, inp, error);
467 	if (sp == NULL) {
468 		IPSEC_ASSERT(*error != 0, ("getpolicy failed w/o error"));
469 		IPSECSTAT_INC(ips_out_inval);
470 		return (NULL);
471 	}
472 	IPSEC_ASSERT(*error == 0, ("sp w/ error set to %u", *error));
473 	switch (sp->policy) {
474 	case IPSEC_POLICY_ENTRUST:
475 	default:
476 		printf("%s: invalid policy %u\n", __func__, sp->policy);
477 		/* FALLTHROUGH */
478 	case IPSEC_POLICY_DISCARD:
479 		IPSECSTAT_INC(ips_out_polvio);
480 		*error = -EINVAL;	/* Packet is discarded by caller. */
481 		break;
482 	case IPSEC_POLICY_BYPASS:
483 	case IPSEC_POLICY_NONE:
484 		KEY_FREESP(&sp);
485 		sp = NULL;		/* NB: force NULL result. */
486 		break;
487 	case IPSEC_POLICY_IPSEC:
488 		if (sp->req == NULL)	/* Acquire a SA. */
489 			*error = key_spdacquire(sp);
490 		break;
491 	}
492 	if (*error != 0) {
493 		KEY_FREESP(&sp);
494 		sp = NULL;
495 	}
496 	return (sp);
497 }
498 
499 static int
500 ipsec_setspidx_inpcb(struct mbuf *m, struct inpcb *inp)
501 {
502 	int error;
503 
504 	IPSEC_ASSERT(inp != NULL, ("null inp"));
505 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
506 	IPSEC_ASSERT(inp->inp_sp->sp_out != NULL && inp->inp_sp->sp_in != NULL,
507 		("null sp_in || sp_out"));
508 
509 	error = ipsec_setspidx(m, &inp->inp_sp->sp_in->spidx, 1);
510 	if (error == 0) {
511 		inp->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
512 		inp->inp_sp->sp_out->spidx = inp->inp_sp->sp_in->spidx;
513 		inp->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
514 	} else {
515 		bzero(&inp->inp_sp->sp_in->spidx,
516 			sizeof (inp->inp_sp->sp_in->spidx));
517 		bzero(&inp->inp_sp->sp_out->spidx,
518 			sizeof (inp->inp_sp->sp_in->spidx));
519 	}
520 	return (error);
521 }
522 
523 /*
524  * Configure security policy index (src/dst/proto/sport/dport)
525  * by looking at the content of mbuf.
526  * The caller is responsible for error recovery (like clearing up spidx).
527  */
528 static int
529 ipsec_setspidx(struct mbuf *m, struct secpolicyindex *spidx, int needport)
530 {
531 	struct ip *ip = NULL;
532 	struct ip ipbuf;
533 	u_int v;
534 	struct mbuf *n;
535 	int len;
536 	int error;
537 
538 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
539 
540 	/*
541 	 * Validate m->m_pkthdr.len.  We see incorrect length if we
542 	 * mistakenly call this function with inconsistent mbuf chain
543 	 * (like 4.4BSD tcp/udp processing).  XXX Should we panic here?
544 	 */
545 	len = 0;
546 	for (n = m; n; n = n->m_next)
547 		len += n->m_len;
548 	if (m->m_pkthdr.len != len) {
549 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
550 			printf("%s: pkthdr len(%d) mismatch (%d), ignored.\n",
551 				__func__, len, m->m_pkthdr.len));
552 		return (EINVAL);
553 	}
554 
555 	if (m->m_pkthdr.len < sizeof(struct ip)) {
556 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
557 			printf("%s: pkthdr len(%d) too small (v4), ignored.\n",
558 			    __func__, m->m_pkthdr.len));
559 		return (EINVAL);
560 	}
561 
562 	if (m->m_len >= sizeof(*ip))
563 		ip = mtod(m, struct ip *);
564 	else {
565 		m_copydata(m, 0, sizeof(ipbuf), (caddr_t)&ipbuf);
566 		ip = &ipbuf;
567 	}
568 	v = ip->ip_v;
569 	switch (v) {
570 	case 4:
571 		error = ipsec4_setspidx_ipaddr(m, spidx);
572 		if (error)
573 			return (error);
574 		ipsec4_get_ulp(m, spidx, needport);
575 		return (0);
576 #ifdef INET6
577 	case 6:
578 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) {
579 			KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
580 				printf("%s: pkthdr len(%d) too small (v6), "
581 				"ignored\n", __func__, m->m_pkthdr.len));
582 			return (EINVAL);
583 		}
584 		error = ipsec6_setspidx_ipaddr(m, spidx);
585 		if (error)
586 			return (error);
587 		ipsec6_get_ulp(m, spidx, needport);
588 		return (0);
589 #endif
590 	default:
591 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
592 			printf("%s: " "unknown IP version %u, ignored.\n",
593 				__func__, v));
594 		return (EINVAL);
595 	}
596 }
597 
598 static void
599 ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
600 {
601 	u_int8_t nxt;
602 	int off;
603 
604 	/* Sanity check. */
605 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
606 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),("packet too short"));
607 
608 	if (m->m_len >= sizeof (struct ip)) {
609 		struct ip *ip = mtod(m, struct ip *);
610 		if (ip->ip_off & htons(IP_MF | IP_OFFMASK))
611 			goto done;
612 		off = ip->ip_hl << 2;
613 		nxt = ip->ip_p;
614 	} else {
615 		struct ip ih;
616 
617 		m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih);
618 		if (ih.ip_off & htons(IP_MF | IP_OFFMASK))
619 			goto done;
620 		off = ih.ip_hl << 2;
621 		nxt = ih.ip_p;
622 	}
623 
624 	while (off < m->m_pkthdr.len) {
625 		struct ip6_ext ip6e;
626 		struct tcphdr th;
627 		struct udphdr uh;
628 
629 		switch (nxt) {
630 		case IPPROTO_TCP:
631 			spidx->ul_proto = nxt;
632 			if (!needport)
633 				goto done_proto;
634 			if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
635 				goto done;
636 			m_copydata(m, off, sizeof (th), (caddr_t) &th);
637 			spidx->src.sin.sin_port = th.th_sport;
638 			spidx->dst.sin.sin_port = th.th_dport;
639 			return;
640 		case IPPROTO_UDP:
641 			spidx->ul_proto = nxt;
642 			if (!needport)
643 				goto done_proto;
644 			if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
645 				goto done;
646 			m_copydata(m, off, sizeof (uh), (caddr_t) &uh);
647 			spidx->src.sin.sin_port = uh.uh_sport;
648 			spidx->dst.sin.sin_port = uh.uh_dport;
649 			return;
650 		case IPPROTO_AH:
651 			if (off + sizeof(ip6e) > m->m_pkthdr.len)
652 				goto done;
653 			/* XXX Sigh, this works but is totally bogus. */
654 			m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e);
655 			off += (ip6e.ip6e_len + 2) << 2;
656 			nxt = ip6e.ip6e_nxt;
657 			break;
658 		case IPPROTO_ICMP:
659 		default:
660 			/* XXX Intermediate headers??? */
661 			spidx->ul_proto = nxt;
662 			goto done_proto;
663 		}
664 	}
665 done:
666 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
667 done_proto:
668 	spidx->src.sin.sin_port = IPSEC_PORT_ANY;
669 	spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
670 }
671 
672 /* Assumes that m is sane. */
673 static int
674 ipsec4_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
675 {
676 	static const struct sockaddr_in template = {
677 		sizeof (struct sockaddr_in),
678 		AF_INET,
679 		0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 }
680 	};
681 
682 	spidx->src.sin = template;
683 	spidx->dst.sin = template;
684 
685 	if (m->m_len < sizeof (struct ip)) {
686 		m_copydata(m, offsetof(struct ip, ip_src),
687 			   sizeof (struct  in_addr),
688 			   (caddr_t) &spidx->src.sin.sin_addr);
689 		m_copydata(m, offsetof(struct ip, ip_dst),
690 			   sizeof (struct  in_addr),
691 			   (caddr_t) &spidx->dst.sin.sin_addr);
692 	} else {
693 		struct ip *ip = mtod(m, struct ip *);
694 		spidx->src.sin.sin_addr = ip->ip_src;
695 		spidx->dst.sin.sin_addr = ip->ip_dst;
696 	}
697 
698 	spidx->prefs = sizeof(struct in_addr) << 3;
699 	spidx->prefd = sizeof(struct in_addr) << 3;
700 
701 	return (0);
702 }
703 
704 #ifdef INET6
705 static void
706 ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
707 {
708 	int off, nxt;
709 	struct tcphdr th;
710 	struct udphdr uh;
711 	struct icmp6_hdr ih;
712 
713 	/* Sanity check. */
714 	if (m == NULL)
715 		panic("%s: NULL pointer was passed.\n", __func__);
716 
717 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
718 		printf("%s:\n", __func__); kdebug_mbuf(m));
719 
720 	/* Set default. */
721 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
722 	((struct sockaddr_in6 *)&spidx->src)->sin6_port = IPSEC_PORT_ANY;
723 	((struct sockaddr_in6 *)&spidx->dst)->sin6_port = IPSEC_PORT_ANY;
724 
725 	nxt = -1;
726 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
727 	if (off < 0 || m->m_pkthdr.len < off)
728 		return;
729 
730 	switch (nxt) {
731 	case IPPROTO_TCP:
732 		spidx->ul_proto = nxt;
733 		if (!needport)
734 			break;
735 		if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
736 			break;
737 		m_copydata(m, off, sizeof(th), (caddr_t)&th);
738 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = th.th_sport;
739 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = th.th_dport;
740 		break;
741 	case IPPROTO_UDP:
742 		spidx->ul_proto = nxt;
743 		if (!needport)
744 			break;
745 		if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
746 			break;
747 		m_copydata(m, off, sizeof(uh), (caddr_t)&uh);
748 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = uh.uh_sport;
749 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = uh.uh_dport;
750 		break;
751 	case IPPROTO_ICMPV6:
752 		spidx->ul_proto = nxt;
753 		if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len)
754 			break;
755 		m_copydata(m, off, sizeof(ih), (caddr_t)&ih);
756 		((struct sockaddr_in6 *)&spidx->src)->sin6_port =
757 		    htons((uint16_t)ih.icmp6_type);
758 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port =
759 		    htons((uint16_t)ih.icmp6_code);
760 		break;
761 	default:
762 		/* XXX Intermediate headers??? */
763 		spidx->ul_proto = nxt;
764 		break;
765 	}
766 }
767 
768 /* Assumes that m is sane. */
769 static int
770 ipsec6_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
771 {
772 	struct ip6_hdr *ip6 = NULL;
773 	struct ip6_hdr ip6buf;
774 	struct sockaddr_in6 *sin6;
775 
776 	if (m->m_len >= sizeof(*ip6))
777 		ip6 = mtod(m, struct ip6_hdr *);
778 	else {
779 		m_copydata(m, 0, sizeof(ip6buf), (caddr_t)&ip6buf);
780 		ip6 = &ip6buf;
781 	}
782 
783 	sin6 = (struct sockaddr_in6 *)&spidx->src;
784 	bzero(sin6, sizeof(*sin6));
785 	sin6->sin6_family = AF_INET6;
786 	sin6->sin6_len = sizeof(struct sockaddr_in6);
787 	bcopy(&ip6->ip6_src, &sin6->sin6_addr, sizeof(ip6->ip6_src));
788 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
789 		sin6->sin6_addr.s6_addr16[1] = 0;
790 		sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]);
791 	}
792 	spidx->prefs = sizeof(struct in6_addr) << 3;
793 
794 	sin6 = (struct sockaddr_in6 *)&spidx->dst;
795 	bzero(sin6, sizeof(*sin6));
796 	sin6->sin6_family = AF_INET6;
797 	sin6->sin6_len = sizeof(struct sockaddr_in6);
798 	bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(ip6->ip6_dst));
799 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
800 		sin6->sin6_addr.s6_addr16[1] = 0;
801 		sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]);
802 	}
803 	spidx->prefd = sizeof(struct in6_addr) << 3;
804 
805 	return (0);
806 }
807 #endif
808 
809 static void
810 ipsec_delpcbpolicy(struct inpcbpolicy *p)
811 {
812 
813 	free(p, M_IPSEC_INPCB);
814 }
815 
816 /* Initialize policy in PCB. */
817 int
818 ipsec_init_policy(struct socket *so, struct inpcbpolicy **pcb_sp)
819 {
820 	struct inpcbpolicy *new;
821 
822 	/* Sanity check. */
823 	if (so == NULL || pcb_sp == NULL)
824 		panic("%s: NULL pointer was passed.\n", __func__);
825 
826 	new = (struct inpcbpolicy *) malloc(sizeof(struct inpcbpolicy),
827 					    M_IPSEC_INPCB, M_NOWAIT|M_ZERO);
828 	if (new == NULL) {
829 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
830 		return (ENOBUFS);
831 	}
832 
833 	new->priv = IPSEC_IS_PRIVILEGED_SO(so);
834 
835 	if ((new->sp_in = KEY_NEWSP()) == NULL) {
836 		ipsec_delpcbpolicy(new);
837 		return (ENOBUFS);
838 	}
839 	new->sp_in->policy = IPSEC_POLICY_ENTRUST;
840 	if ((new->sp_out = KEY_NEWSP()) == NULL) {
841 		KEY_FREESP(&new->sp_in);
842 		ipsec_delpcbpolicy(new);
843 		return (ENOBUFS);
844 	}
845 	new->sp_out->policy = IPSEC_POLICY_ENTRUST;
846 	*pcb_sp = new;
847 
848 	return (0);
849 }
850 
851 /* Copy old IPsec policy into new. */
852 int
853 ipsec_copy_policy(struct inpcbpolicy *old, struct inpcbpolicy *new)
854 {
855 	struct secpolicy *sp;
856 
857 	sp = ipsec_deepcopy_policy(old->sp_in);
858 	if (sp) {
859 		KEY_FREESP(&new->sp_in);
860 		new->sp_in = sp;
861 	} else
862 		return (ENOBUFS);
863 
864 	sp = ipsec_deepcopy_policy(old->sp_out);
865 	if (sp) {
866 		KEY_FREESP(&new->sp_out);
867 		new->sp_out = sp;
868 	} else
869 		return (ENOBUFS);
870 
871 	new->priv = old->priv;
872 
873 	return (0);
874 }
875 
876 struct ipsecrequest *
877 ipsec_newisr(void)
878 {
879 	struct ipsecrequest *p;
880 
881 	p = malloc(sizeof(struct ipsecrequest), M_IPSEC_SR, M_NOWAIT|M_ZERO);
882 	if (p != NULL)
883 		IPSECREQUEST_LOCK_INIT(p);
884 	return (p);
885 }
886 
887 void
888 ipsec_delisr(struct ipsecrequest *p)
889 {
890 
891 	IPSECREQUEST_LOCK_DESTROY(p);
892 	free(p, M_IPSEC_SR);
893 }
894 
895 /* Deep-copy a policy in PCB. */
896 static struct secpolicy *
897 ipsec_deepcopy_policy(struct secpolicy *src)
898 {
899 	struct ipsecrequest *newchain = NULL;
900 	struct ipsecrequest *p;
901 	struct ipsecrequest **q;
902 	struct ipsecrequest *r;
903 	struct secpolicy *dst;
904 
905 	if (src == NULL)
906 		return (NULL);
907 	dst = KEY_NEWSP();
908 	if (dst == NULL)
909 		return (NULL);
910 
911 	/*
912 	 * Deep-copy IPsec request chain.  This is required since struct
913 	 * ipsecrequest is not reference counted.
914 	 */
915 	q = &newchain;
916 	for (p = src->req; p; p = p->next) {
917 		*q = ipsec_newisr();
918 		if (*q == NULL)
919 			goto fail;
920 		(*q)->saidx.proto = p->saidx.proto;
921 		(*q)->saidx.mode = p->saidx.mode;
922 		(*q)->level = p->level;
923 		(*q)->saidx.reqid = p->saidx.reqid;
924 
925 		bcopy(&p->saidx.src, &(*q)->saidx.src, sizeof((*q)->saidx.src));
926 		bcopy(&p->saidx.dst, &(*q)->saidx.dst, sizeof((*q)->saidx.dst));
927 
928 		(*q)->sp = dst;
929 
930 		q = &((*q)->next);
931 	}
932 
933 	dst->req = newchain;
934 	dst->policy = src->policy;
935 	/* Do not touch the refcnt fields. */
936 
937 	return (dst);
938 
939 fail:
940 	for (p = newchain; p; p = r) {
941 		r = p->next;
942 		ipsec_delisr(p);
943 		p = NULL;
944 	}
945 	return (NULL);
946 }
947 
948 /* Set policy and IPsec request if present. */
949 static int
950 ipsec_set_policy_internal(struct secpolicy **pcb_sp, int optname,
951     caddr_t request, size_t len, struct ucred *cred)
952 {
953 	struct sadb_x_policy *xpl;
954 	struct secpolicy *newsp = NULL;
955 	int error;
956 
957 	/* Sanity check. */
958 	if (pcb_sp == NULL || *pcb_sp == NULL || request == NULL)
959 		return (EINVAL);
960 	if (len < sizeof(*xpl))
961 		return (EINVAL);
962 	xpl = (struct sadb_x_policy *)request;
963 
964 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
965 		printf("%s: passed policy\n", __func__);
966 		kdebug_sadb_x_policy((struct sadb_ext *)xpl));
967 
968 	/* Check policy type. */
969 	/* ipsec_set_policy_internal() accepts IPSEC, ENTRUST and BYPASS. */
970 	if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD
971 	 || xpl->sadb_x_policy_type == IPSEC_POLICY_NONE)
972 		return (EINVAL);
973 
974 	/* Check privileged socket. */
975 	if (cred != NULL && xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
976 		error = priv_check_cred(cred, PRIV_NETINET_IPSEC, 0);
977 		if (error)
978 			return (EACCES);
979 	}
980 
981 	/* Allocating new SP entry. */
982 	if ((newsp = key_msg2sp(xpl, len, &error)) == NULL)
983 		return (error);
984 
985 	/* Clear old SP and set new SP. */
986 	KEY_FREESP(pcb_sp);
987 	*pcb_sp = newsp;
988 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
989 		printf("%s: new policy\n", __func__);
990 		kdebug_secpolicy(newsp));
991 
992 	return (0);
993 }
994 
995 int
996 ipsec_set_policy(struct inpcb *inp, int optname, caddr_t request,
997     size_t len, struct ucred *cred)
998 {
999 	struct sadb_x_policy *xpl;
1000 	struct secpolicy **pcb_sp;
1001 
1002 	/* Sanity check. */
1003 	if (inp == NULL || request == NULL)
1004 		return (EINVAL);
1005 	if (len < sizeof(*xpl))
1006 		return (EINVAL);
1007 	xpl = (struct sadb_x_policy *)request;
1008 
1009 	/* Select direction. */
1010 	switch (xpl->sadb_x_policy_dir) {
1011 	case IPSEC_DIR_INBOUND:
1012 		pcb_sp = &inp->inp_sp->sp_in;
1013 		break;
1014 	case IPSEC_DIR_OUTBOUND:
1015 		pcb_sp = &inp->inp_sp->sp_out;
1016 		break;
1017 	default:
1018 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1019 			xpl->sadb_x_policy_dir));
1020 		return (EINVAL);
1021 	}
1022 
1023 	return (ipsec_set_policy_internal(pcb_sp, optname, request, len, cred));
1024 }
1025 
1026 int
1027 ipsec_get_policy(struct inpcb *inp, caddr_t request, size_t len,
1028     struct mbuf **mp)
1029 {
1030 	struct sadb_x_policy *xpl;
1031 	struct secpolicy *pcb_sp;
1032 
1033 	/* Sanity check. */
1034 	if (inp == NULL || request == NULL || mp == NULL)
1035 		return (EINVAL);
1036 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
1037 	if (len < sizeof(*xpl))
1038 		return (EINVAL);
1039 	xpl = (struct sadb_x_policy *)request;
1040 
1041 	/* Select direction. */
1042 	switch (xpl->sadb_x_policy_dir) {
1043 	case IPSEC_DIR_INBOUND:
1044 		pcb_sp = inp->inp_sp->sp_in;
1045 		break;
1046 	case IPSEC_DIR_OUTBOUND:
1047 		pcb_sp = inp->inp_sp->sp_out;
1048 		break;
1049 	default:
1050 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1051 			xpl->sadb_x_policy_dir));
1052 		return (EINVAL);
1053 	}
1054 
1055 	/* Sanity check. Should be an IPSEC_ASSERT. */
1056 	if (pcb_sp == NULL)
1057 		return (EINVAL);
1058 
1059 	*mp = key_sp2msg(pcb_sp);
1060 	if (!*mp) {
1061 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
1062 		return (ENOBUFS);
1063 	}
1064 
1065 	(*mp)->m_type = MT_DATA;
1066 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1067 		printf("%s:\n", __func__); kdebug_mbuf(*mp));
1068 
1069 	return (0);
1070 }
1071 
1072 /* Delete policy in PCB. */
1073 int
1074 ipsec_delete_pcbpolicy(struct inpcb *inp)
1075 {
1076 	IPSEC_ASSERT(inp != NULL, ("null inp"));
1077 
1078 	if (inp->inp_sp == NULL)
1079 		return (0);
1080 
1081 	if (inp->inp_sp->sp_in != NULL)
1082 		KEY_FREESP(&inp->inp_sp->sp_in);
1083 
1084 	if (inp->inp_sp->sp_out != NULL)
1085 		KEY_FREESP(&inp->inp_sp->sp_out);
1086 
1087 	ipsec_delpcbpolicy(inp->inp_sp);
1088 	inp->inp_sp = NULL;
1089 
1090 	return (0);
1091 }
1092 
1093 /*
1094  * Return current level.
1095  * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
1096  */
1097 u_int
1098 ipsec_get_reqlevel(struct ipsecrequest *isr)
1099 {
1100 	u_int level = 0;
1101 	u_int esp_trans_deflev, esp_net_deflev;
1102 	u_int ah_trans_deflev, ah_net_deflev;
1103 
1104 	IPSEC_ASSERT(isr != NULL && isr->sp != NULL, ("null argument"));
1105 	IPSEC_ASSERT(isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family,
1106 		("af family mismatch, src %u, dst %u",
1107 		 isr->sp->spidx.src.sa.sa_family,
1108 		 isr->sp->spidx.dst.sa.sa_family));
1109 
1110 /* XXX Note that we have ipseclog() expanded here - code sync issue. */
1111 #define IPSEC_CHECK_DEFAULT(lev) \
1112 	(((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE	      \
1113 			&& (lev) != IPSEC_LEVEL_UNIQUE)			      \
1114 		? (V_ipsec_debug						      \
1115 			? log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
1116 				(lev), IPSEC_LEVEL_REQUIRE)		      \
1117 			: 0),						      \
1118 			(lev) = IPSEC_LEVEL_REQUIRE,			      \
1119 			(lev)						      \
1120 		: (lev))
1121 
1122 	/* Set default level. */
1123 	switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) {
1124 #ifdef INET
1125 	case AF_INET:
1126 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev);
1127 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev);
1128 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev);
1129 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev);
1130 		break;
1131 #endif
1132 #ifdef INET6
1133 	case AF_INET6:
1134 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev);
1135 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev);
1136 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev);
1137 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev);
1138 		break;
1139 #endif /* INET6 */
1140 	default:
1141 		panic("%s: unknown af %u",
1142 			__func__, isr->sp->spidx.src.sa.sa_family);
1143 	}
1144 
1145 #undef IPSEC_CHECK_DEFAULT
1146 
1147 	/* Set level. */
1148 	switch (isr->level) {
1149 	case IPSEC_LEVEL_DEFAULT:
1150 		switch (isr->saidx.proto) {
1151 		case IPPROTO_ESP:
1152 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1153 				level = esp_net_deflev;
1154 			else
1155 				level = esp_trans_deflev;
1156 			break;
1157 		case IPPROTO_AH:
1158 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1159 				level = ah_net_deflev;
1160 			else
1161 				level = ah_trans_deflev;
1162 			break;
1163 		case IPPROTO_IPCOMP:
1164 			/*
1165 			 * We don't really care, as IPcomp document says that
1166 			 * we shouldn't compress small packets.
1167 			 */
1168 			level = IPSEC_LEVEL_USE;
1169 			break;
1170 		default:
1171 			panic("%s: Illegal protocol defined %u\n", __func__,
1172 				isr->saidx.proto);
1173 		}
1174 		break;
1175 
1176 	case IPSEC_LEVEL_USE:
1177 	case IPSEC_LEVEL_REQUIRE:
1178 		level = isr->level;
1179 		break;
1180 	case IPSEC_LEVEL_UNIQUE:
1181 		level = IPSEC_LEVEL_REQUIRE;
1182 		break;
1183 
1184 	default:
1185 		panic("%s: Illegal IPsec level %u\n", __func__, isr->level);
1186 	}
1187 
1188 	return (level);
1189 }
1190 
1191 /*
1192  * Check security policy requirements against the actual
1193  * packet contents.  Return one if the packet should be
1194  * reject as "invalid"; otherwiser return zero to have the
1195  * packet treated as "valid".
1196  *
1197  * OUT:
1198  *	0: valid
1199  *	1: invalid
1200  */
1201 static int
1202 ipsec_in_reject(struct secpolicy *sp, struct mbuf *m)
1203 {
1204 	struct ipsecrequest *isr;
1205 	int need_auth;
1206 
1207 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1208 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1209 
1210 	/* Check policy. */
1211 	switch (sp->policy) {
1212 	case IPSEC_POLICY_DISCARD:
1213 		return (1);
1214 	case IPSEC_POLICY_BYPASS:
1215 	case IPSEC_POLICY_NONE:
1216 		return (0);
1217 	}
1218 
1219 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1220 		("invalid policy %u", sp->policy));
1221 
1222 	/* XXX Should compare policy against IPsec header history. */
1223 
1224 	need_auth = 0;
1225 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1226 		if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE)
1227 			continue;
1228 		switch (isr->saidx.proto) {
1229 		case IPPROTO_ESP:
1230 			if ((m->m_flags & M_DECRYPTED) == 0) {
1231 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1232 				    printf("%s: ESP m_flags:%x\n", __func__,
1233 					    m->m_flags));
1234 				return (1);
1235 			}
1236 
1237 			if (!need_auth &&
1238 			    isr->sav != NULL &&
1239 			    isr->sav->tdb_authalgxform != NULL &&
1240 			    (m->m_flags & M_AUTHIPDGM) == 0) {
1241 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1242 				    printf("%s: ESP/AH m_flags:%x\n", __func__,
1243 					    m->m_flags));
1244 				return (1);
1245 			}
1246 			break;
1247 		case IPPROTO_AH:
1248 			need_auth = 1;
1249 			if ((m->m_flags & M_AUTHIPHDR) == 0) {
1250 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1251 				    printf("%s: AH m_flags:%x\n", __func__,
1252 					    m->m_flags));
1253 				return (1);
1254 			}
1255 			break;
1256 		case IPPROTO_IPCOMP:
1257 			/*
1258 			 * We don't really care, as IPcomp document
1259 			 * says that we shouldn't compress small
1260 			 * packets.  IPComp policy should always be
1261 			 * treated as being in "use" level.
1262 			 */
1263 			break;
1264 		}
1265 	}
1266 	return (0);		/* Valid. */
1267 }
1268 
1269 /*
1270  * Non zero return value means security policy DISCARD or policy violation.
1271  */
1272 static int
1273 ipsec46_in_reject(struct mbuf *m, struct inpcb *inp)
1274 {
1275 	struct secpolicy *sp;
1276 	int error;
1277 	int result;
1278 
1279 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1280 
1281 	/* Get SP for this packet. */
1282 	if (inp == NULL)
1283 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, &error);
1284 	else
1285 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND, inp, &error);
1286 
1287 	if (sp != NULL) {
1288 		result = ipsec_in_reject(sp, m);
1289 		KEY_FREESP(&sp);
1290 	} else {
1291 		result = 1;	/* treat errors as policy violation */
1292 	}
1293 	return (result);
1294 }
1295 
1296 /*
1297  * Check AH/ESP integrity.
1298  * This function is called from tcp_input(), udp_input(),
1299  * and {ah,esp}4_input for tunnel mode.
1300  */
1301 int
1302 ipsec4_in_reject(struct mbuf *m, struct inpcb *inp)
1303 {
1304 	int result;
1305 
1306 	result = ipsec46_in_reject(m, inp);
1307 	if (result)
1308 		IPSECSTAT_INC(ips_in_polvio);
1309 
1310 	return (result);
1311 }
1312 
1313 #ifdef INET6
1314 /*
1315  * Check AH/ESP integrity.
1316  * This function is called from tcp6_input(), udp6_input(),
1317  * and {ah,esp}6_input for tunnel mode.
1318  */
1319 int
1320 ipsec6_in_reject(struct mbuf *m, struct inpcb *inp)
1321 {
1322 	int result;
1323 
1324 	result = ipsec46_in_reject(m, inp);
1325 	if (result)
1326 		IPSEC6STAT_INC(ips_in_polvio);
1327 
1328 	return (result);
1329 }
1330 #endif
1331 
1332 /*
1333  * Compute the byte size to be occupied by IPsec header.
1334  * In case it is tunnelled, it includes the size of outer IP header.
1335  * NOTE: SP passed is freed in this function.
1336  */
1337 static size_t
1338 ipsec_hdrsiz_internal(struct secpolicy *sp)
1339 {
1340 	struct ipsecrequest *isr;
1341 	size_t size;
1342 
1343 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1344 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1345 
1346 	switch (sp->policy) {
1347 	case IPSEC_POLICY_DISCARD:
1348 	case IPSEC_POLICY_BYPASS:
1349 	case IPSEC_POLICY_NONE:
1350 		return (0);
1351 	}
1352 
1353 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1354 		("invalid policy %u", sp->policy));
1355 
1356 	size = 0;
1357 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1358 		size_t clen = 0;
1359 
1360 		switch (isr->saidx.proto) {
1361 		case IPPROTO_ESP:
1362 			clen = esp_hdrsiz(isr->sav);
1363 			break;
1364 		case IPPROTO_AH:
1365 			clen = ah_hdrsiz(isr->sav);
1366 			break;
1367 		case IPPROTO_IPCOMP:
1368 			clen = sizeof(struct ipcomp);
1369 			break;
1370 		}
1371 
1372 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1373 			switch (isr->saidx.dst.sa.sa_family) {
1374 			case AF_INET:
1375 				clen += sizeof(struct ip);
1376 				break;
1377 #ifdef INET6
1378 			case AF_INET6:
1379 				clen += sizeof(struct ip6_hdr);
1380 				break;
1381 #endif
1382 			default:
1383 				ipseclog((LOG_ERR, "%s: unknown AF %d in "
1384 				    "IPsec tunnel SA\n", __func__,
1385 				    ((struct sockaddr *)&isr->saidx.dst)->sa_family));
1386 				break;
1387 			}
1388 		}
1389 		size += clen;
1390 	}
1391 
1392 	return (size);
1393 }
1394 
1395 /*
1396  * This function is called from ipsec_hdrsiz_tcp(), ip_ipsec_mtu(),
1397  * disabled ip6_ipsec_mtu() and ip6_forward().
1398  */
1399 size_t
1400 ipsec_hdrsiz(struct mbuf *m, u_int dir, struct inpcb *inp)
1401 {
1402 	struct secpolicy *sp;
1403 	int error;
1404 	size_t size;
1405 
1406 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1407 
1408 	/* Get SP for this packet. */
1409 	if (inp == NULL)
1410 		sp = ipsec_getpolicybyaddr(m, dir, &error);
1411 	else
1412 		sp = ipsec_getpolicybysock(m, dir, inp, &error);
1413 
1414 	if (sp != NULL) {
1415 		size = ipsec_hdrsiz_internal(sp);
1416 		KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1417 			printf("%s: size:%lu.\n", __func__,
1418 				(unsigned long)size));
1419 
1420 		KEY_FREESP(&sp);
1421 	} else {
1422 		size = 0;	/* XXX Should be panic?
1423 				 * -> No, we are called w/o knowing if
1424 				 *    IPsec processing is needed. */
1425 	}
1426 	return (size);
1427 }
1428 
1429 /*
1430  * Check the variable replay window.
1431  * ipsec_chkreplay() performs replay check before ICV verification.
1432  * ipsec_updatereplay() updates replay bitmap.  This must be called after
1433  * ICV verification (it also performs replay check, which is usually done
1434  * beforehand).
1435  * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1436  *
1437  * Based on RFC 2401.
1438  */
1439 int
1440 ipsec_chkreplay(u_int32_t seq, struct secasvar *sav)
1441 {
1442 	const struct secreplay *replay;
1443 	u_int32_t diff;
1444 	int fr;
1445 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1446 	int frlast;		/* Constant: last frame. */
1447 
1448 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1449 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1450 
1451 	replay = sav->replay;
1452 
1453 	if (replay->wsize == 0)
1454 		return (1);	/* No need to check replay. */
1455 
1456 	/* Constant. */
1457 	frlast = replay->wsize - 1;
1458 	wsizeb = replay->wsize << 3;
1459 
1460 	/* Sequence number of 0 is invalid. */
1461 	if (seq == 0)
1462 		return (0);
1463 
1464 	/* First time is always okay. */
1465 	if (replay->count == 0)
1466 		return (1);
1467 
1468 	if (seq > replay->lastseq) {
1469 		/* Larger sequences are okay. */
1470 		return (1);
1471 	} else {
1472 		/* seq is equal or less than lastseq. */
1473 		diff = replay->lastseq - seq;
1474 
1475 		/* Over range to check, i.e. too old or wrapped. */
1476 		if (diff >= wsizeb)
1477 			return (0);
1478 
1479 		fr = frlast - diff / 8;
1480 
1481 		/* This packet already seen? */
1482 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1483 			return (0);
1484 
1485 		/* Out of order but good. */
1486 		return (1);
1487 	}
1488 }
1489 
1490 /*
1491  * Check replay counter whether to update or not.
1492  * OUT:	0:	OK
1493  *	1:	NG
1494  */
1495 int
1496 ipsec_updatereplay(u_int32_t seq, struct secasvar *sav)
1497 {
1498 	char buf[128];
1499 	struct secreplay *replay;
1500 	u_int32_t diff;
1501 	int fr;
1502 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1503 	int frlast;		/* Constant: last frame. */
1504 
1505 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1506 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1507 
1508 	replay = sav->replay;
1509 
1510 	if (replay->wsize == 0)
1511 		goto ok;	/* No need to check replay. */
1512 
1513 	/* Constant. */
1514 	frlast = replay->wsize - 1;
1515 	wsizeb = replay->wsize << 3;
1516 
1517 	/* Sequence number of 0 is invalid. */
1518 	if (seq == 0)
1519 		return (1);
1520 
1521 	/* First time. */
1522 	if (replay->count == 0) {
1523 		replay->lastseq = seq;
1524 		bzero(replay->bitmap, replay->wsize);
1525 		(replay->bitmap)[frlast] = 1;
1526 		goto ok;
1527 	}
1528 
1529 	if (seq > replay->lastseq) {
1530 		/* seq is larger than lastseq. */
1531 		diff = seq - replay->lastseq;
1532 
1533 		/* New larger sequence number. */
1534 		if (diff < wsizeb) {
1535 			/* In window. */
1536 			/* Set bit for this packet. */
1537 			vshiftl(replay->bitmap, diff, replay->wsize);
1538 			(replay->bitmap)[frlast] |= 1;
1539 		} else {
1540 			/* This packet has a "way larger". */
1541 			bzero(replay->bitmap, replay->wsize);
1542 			(replay->bitmap)[frlast] = 1;
1543 		}
1544 		replay->lastseq = seq;
1545 
1546 		/* Larger is good. */
1547 	} else {
1548 		/* seq is equal or less than lastseq. */
1549 		diff = replay->lastseq - seq;
1550 
1551 		/* Over range to check, i.e. too old or wrapped. */
1552 		if (diff >= wsizeb)
1553 			return (1);
1554 
1555 		fr = frlast - diff / 8;
1556 
1557 		/* This packet already seen? */
1558 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1559 			return (1);
1560 
1561 		/* Mark as seen. */
1562 		(replay->bitmap)[fr] |= (1 << (diff % 8));
1563 
1564 		/* Out of order but good. */
1565 	}
1566 
1567 ok:
1568 	if (replay->count == ~0) {
1569 
1570 		/* Set overflow flag. */
1571 		replay->overflow++;
1572 
1573 		/* Don't increment, no more packets accepted. */
1574 		if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0)
1575 			return (1);
1576 
1577 		ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n",
1578 		    __func__, replay->overflow,
1579 		    ipsec_logsastr(sav, buf, sizeof(buf))));
1580 	}
1581 
1582 	replay->count++;
1583 
1584 	return (0);
1585 }
1586 
1587 /*
1588  * Shift variable length buffer to left.
1589  * IN:	bitmap: pointer to the buffer
1590  * 	nbit:	the number of to shift.
1591  *	wsize:	buffer size (bytes).
1592  */
1593 static void
1594 vshiftl(unsigned char *bitmap, int nbit, int wsize)
1595 {
1596 	int s, j, i;
1597 	unsigned char over;
1598 
1599 	for (j = 0; j < nbit; j += 8) {
1600 		s = (nbit - j < 8) ? (nbit - j): 8;
1601 		bitmap[0] <<= s;
1602 		for (i = 1; i < wsize; i++) {
1603 			over = (bitmap[i] >> (8 - s));
1604 			bitmap[i] <<= s;
1605 			bitmap[i-1] |= over;
1606 		}
1607 	}
1608 }
1609 
1610 /* Return a printable string for the address. */
1611 char*
1612 ipsec_address(union sockaddr_union* sa, char *buf, socklen_t size)
1613 {
1614 
1615 	switch (sa->sa.sa_family) {
1616 #ifdef INET
1617 	case AF_INET:
1618 		return (inet_ntop(AF_INET, &sa->sin.sin_addr, buf, size));
1619 #endif /* INET */
1620 #ifdef INET6
1621 	case AF_INET6:
1622 		return (inet_ntop(AF_INET6, &sa->sin6.sin6_addr, buf, size));
1623 #endif /* INET6 */
1624 	default:
1625 		return ("(unknown address family)");
1626 	}
1627 }
1628 
1629 char *
1630 ipsec_logsastr(struct secasvar *sav, char *buf, size_t size)
1631 {
1632 	char sbuf[INET6_ADDRSTRLEN], dbuf[INET6_ADDRSTRLEN];
1633 
1634 	IPSEC_ASSERT(sav->sah->saidx.src.sa.sa_family ==
1635 	    sav->sah->saidx.dst.sa.sa_family, ("address family mismatch"));
1636 
1637 	snprintf(buf, size, "SA(SPI=%08lx src=%s dst=%s)",
1638 	    (u_long)ntohl(sav->spi),
1639 	    ipsec_address(&sav->sah->saidx.src, sbuf, sizeof(sbuf)),
1640 	    ipsec_address(&sav->sah->saidx.dst, dbuf, sizeof(dbuf)));
1641 	return (buf);
1642 }
1643 
1644 void
1645 ipsec_dumpmbuf(struct mbuf *m)
1646 {
1647 	int totlen;
1648 	int i;
1649 	u_char *p;
1650 
1651 	totlen = 0;
1652 	printf("---\n");
1653 	while (m) {
1654 		p = mtod(m, u_char *);
1655 		for (i = 0; i < m->m_len; i++) {
1656 			printf("%02x ", p[i]);
1657 			totlen++;
1658 			if (totlen % 16 == 0)
1659 				printf("\n");
1660 		}
1661 		m = m->m_next;
1662 	}
1663 	if (totlen % 16 != 0)
1664 		printf("\n");
1665 	printf("---\n");
1666 }
1667 
1668 static void
1669 def_policy_init(const void *unused __unused)
1670 {
1671 
1672 	bzero(&V_def_policy, sizeof(struct secpolicy));
1673 	V_def_policy.policy = IPSEC_POLICY_NONE;
1674 	V_def_policy.refcnt = 1;
1675 }
1676 VNET_SYSINIT(def_policy_init, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY,
1677     def_policy_init, NULL);
1678 
1679 
1680 /* XXX This stuff doesn't belong here... */
1681 
1682 static	struct xformsw* xforms = NULL;
1683 
1684 /*
1685  * Register a transform; typically at system startup.
1686  */
1687 void
1688 xform_register(struct xformsw* xsp)
1689 {
1690 
1691 	xsp->xf_next = xforms;
1692 	xforms = xsp;
1693 }
1694 
1695 /*
1696  * Initialize transform support in an sav.
1697  */
1698 int
1699 xform_init(struct secasvar *sav, int xftype)
1700 {
1701 	struct xformsw *xsp;
1702 
1703 	if (sav->tdb_xform != NULL)	/* Previously initialized. */
1704 		return (0);
1705 	for (xsp = xforms; xsp; xsp = xsp->xf_next)
1706 		if (xsp->xf_type == xftype)
1707 			return ((*xsp->xf_init)(sav, xsp));
1708 	return (EINVAL);
1709 }
1710