xref: /freebsd/sys/netipsec/ipsec.c (revision ec0e626bafb335b30c499d06066997f54b10c092)
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_setspidx_inpcb(struct mbuf *, struct inpcb *);
242 static int ipsec_setspidx(struct mbuf *, struct secpolicyindex *, int);
243 static void ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *, int);
244 static int ipsec4_setspidx_ipaddr(struct mbuf *, struct secpolicyindex *);
245 #ifdef INET6
246 static void ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *, int);
247 static int ipsec6_setspidx_ipaddr(struct mbuf *, struct secpolicyindex *);
248 #endif
249 static void ipsec_delpcbpolicy(struct inpcbpolicy *);
250 static struct secpolicy *ipsec_deepcopy_policy(struct secpolicy *src);
251 static void vshiftl(unsigned char *, int, int);
252 
253 MALLOC_DEFINE(M_IPSEC_INPCB, "inpcbpolicy", "inpcb-resident ipsec policy");
254 
255 /*
256  * Return a held reference to the default SP.
257  */
258 static struct secpolicy *
259 key_allocsp_default(const char* where, int tag)
260 {
261 	struct secpolicy *sp;
262 
263 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
264 		printf("DP key_allocsp_default from %s:%u\n", where, tag));
265 
266 	sp = &V_def_policy;
267 	if (sp->policy != IPSEC_POLICY_DISCARD &&
268 	    sp->policy != IPSEC_POLICY_NONE) {
269 		ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n",
270 		    sp->policy, IPSEC_POLICY_NONE));
271 		sp->policy = IPSEC_POLICY_NONE;
272 	}
273 	key_addref(sp);
274 
275 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
276 		printf("DP key_allocsp_default returns SP:%p (%u)\n",
277 			sp, sp->refcnt));
278 	return (sp);
279 }
280 #define	KEY_ALLOCSP_DEFAULT() \
281 	key_allocsp_default(__FILE__, __LINE__)
282 
283 /*
284  * For OUTBOUND packet having a socket. Searching SPD for packet,
285  * and return a pointer to SP.
286  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
287  *		0	: bypass
288  *		EACCES	: discard packet.
289  *		ENOENT	: ipsec_acquire() in progress, maybe.
290  *		others	: error occured.
291  *	others:	a pointer to SP
292  *
293  * NOTE: IPv6 mapped adddress concern is implemented here.
294  */
295 struct secpolicy *
296 ipsec_getpolicy(struct tdb_ident *tdbi, u_int dir)
297 {
298 	struct secpolicy *sp;
299 
300 	IPSEC_ASSERT(tdbi != NULL, ("null tdbi"));
301 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
302 		("invalid direction %u", dir));
303 
304 	sp = KEY_ALLOCSP2(tdbi->spi, &tdbi->dst, tdbi->proto, dir);
305 	if (sp == NULL)			/*XXX????*/
306 		sp = KEY_ALLOCSP_DEFAULT();
307 	IPSEC_ASSERT(sp != NULL, ("null SP"));
308 	return (sp);
309 }
310 
311 /*
312  * For OUTBOUND packet having a socket. Searching SPD for packet,
313  * and return a pointer to SP.
314  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
315  *		0	: bypass
316  *		EACCES	: discard packet.
317  *		ENOENT	: ipsec_acquire() in progress, maybe.
318  *		others	: error occured.
319  *	others:	a pointer to SP
320  *
321  * NOTE: IPv6 mapped adddress concern is implemented here.
322  */
323 static struct secpolicy *
324 ipsec_getpolicybysock(struct mbuf *m, u_int dir, struct inpcb *inp, int *error)
325 {
326 	struct inpcbpolicy *pcbsp;
327 	struct secpolicy *currsp = NULL;	/* Policy on socket. */
328 	struct secpolicy *sp;
329 
330 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
331 	IPSEC_ASSERT(inp != NULL, ("null inpcb"));
332 	IPSEC_ASSERT(error != NULL, ("null error"));
333 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
334 		("invalid direction %u", dir));
335 
336 	/* Set spidx in pcb. */
337 	*error = ipsec_setspidx_inpcb(m, inp);
338 	if (*error)
339 		return (NULL);
340 
341 	pcbsp = inp->inp_sp;
342 	IPSEC_ASSERT(pcbsp != NULL, ("null pcbsp"));
343 	switch (dir) {
344 	case IPSEC_DIR_INBOUND:
345 		currsp = pcbsp->sp_in;
346 		break;
347 	case IPSEC_DIR_OUTBOUND:
348 		currsp = pcbsp->sp_out;
349 		break;
350 	}
351 	IPSEC_ASSERT(currsp != NULL, ("null currsp"));
352 
353 	if (pcbsp->priv) {			/* When privilieged socket. */
354 		switch (currsp->policy) {
355 		case IPSEC_POLICY_BYPASS:
356 		case IPSEC_POLICY_IPSEC:
357 			key_addref(currsp);
358 			sp = currsp;
359 			break;
360 
361 		case IPSEC_POLICY_ENTRUST:
362 			/* Look for a policy in SPD. */
363 			sp = KEY_ALLOCSP(&currsp->spidx, dir);
364 			if (sp == NULL)		/* No SP found. */
365 				sp = KEY_ALLOCSP_DEFAULT();
366 			break;
367 
368 		default:
369 			ipseclog((LOG_ERR, "%s: Invalid policy for PCB %d\n",
370 				__func__, currsp->policy));
371 			*error = EINVAL;
372 			return (NULL);
373 		}
374 	} else {				/* Unpriv, SPD has policy. */
375 		sp = KEY_ALLOCSP(&currsp->spidx, dir);
376 		if (sp == NULL) {		/* No SP found. */
377 			switch (currsp->policy) {
378 			case IPSEC_POLICY_BYPASS:
379 				ipseclog((LOG_ERR, "%s: Illegal policy for "
380 					"non-priviliged defined %d\n",
381 					__func__, currsp->policy));
382 				*error = EINVAL;
383 				return (NULL);
384 
385 			case IPSEC_POLICY_ENTRUST:
386 				sp = KEY_ALLOCSP_DEFAULT();
387 				break;
388 
389 			case IPSEC_POLICY_IPSEC:
390 				key_addref(currsp);
391 				sp = currsp;
392 				break;
393 
394 			default:
395 				ipseclog((LOG_ERR, "%s: Invalid policy for "
396 					"PCB %d\n", __func__, currsp->policy));
397 				*error = EINVAL;
398 				return (NULL);
399 			}
400 		}
401 	}
402 	IPSEC_ASSERT(sp != NULL,
403 		("null SP (priv %u policy %u", pcbsp->priv, currsp->policy));
404 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
405 		printf("DP %s (priv %u policy %u) allocate SP:%p (refcnt %u)\n",
406 			__func__, pcbsp->priv, currsp->policy, sp, sp->refcnt));
407 	return (sp);
408 }
409 
410 /*
411  * For FORWADING packet or OUTBOUND without a socket. Searching SPD for packet,
412  * and return a pointer to SP.
413  * OUT:	positive: a pointer to the entry for security policy leaf matched.
414  *	NULL:	no apropreate SP found, the following value is set to error.
415  *		0	: bypass
416  *		EACCES	: discard packet.
417  *		ENOENT	: ipsec_acquire() in progress, maybe.
418  *		others	: error occured.
419  */
420 struct secpolicy *
421 ipsec_getpolicybyaddr(struct mbuf *m, u_int dir, int *error)
422 {
423 	struct secpolicyindex spidx;
424 	struct secpolicy *sp;
425 
426 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
427 	IPSEC_ASSERT(error != NULL, ("null error"));
428 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
429 		("invalid direction %u", dir));
430 
431 	sp = NULL;
432 	*error = 0;
433 	if (key_havesp(dir)) {
434 		/* Make an index to look for a policy. */
435 		*error = ipsec_setspidx(m, &spidx, 0);
436 		if (*error != 0) {
437 			DPRINTF(("%s: setpidx failed, dir %u\n",
438 				__func__, dir));
439 			return (NULL);
440 		}
441 		spidx.dir = dir;
442 		sp = KEY_ALLOCSP(&spidx, dir);
443 	}
444 	if (sp == NULL)			/* No SP found, use system default. */
445 		sp = KEY_ALLOCSP_DEFAULT();
446 	IPSEC_ASSERT(sp != NULL, ("null SP"));
447 	return (sp);
448 }
449 
450 struct secpolicy *
451 ipsec4_checkpolicy(struct mbuf *m, u_int dir, int *error, struct inpcb *inp)
452 {
453 	struct secpolicy *sp;
454 
455 	*error = 0;
456 	if (inp == NULL)
457 		sp = ipsec_getpolicybyaddr(m, dir, error);
458 	else
459 		sp = ipsec_getpolicybysock(m, dir, inp, error);
460 	if (sp == NULL) {
461 		IPSEC_ASSERT(*error != 0, ("getpolicy failed w/o error"));
462 		IPSECSTAT_INC(ips_out_inval);
463 		return (NULL);
464 	}
465 	IPSEC_ASSERT(*error == 0, ("sp w/ error set to %u", *error));
466 	switch (sp->policy) {
467 	case IPSEC_POLICY_ENTRUST:
468 	default:
469 		printf("%s: invalid policy %u\n", __func__, sp->policy);
470 		/* FALLTHROUGH */
471 	case IPSEC_POLICY_DISCARD:
472 		IPSECSTAT_INC(ips_out_polvio);
473 		*error = -EINVAL;	/* Packet is discarded by caller. */
474 		break;
475 	case IPSEC_POLICY_BYPASS:
476 	case IPSEC_POLICY_NONE:
477 		KEY_FREESP(&sp);
478 		sp = NULL;		/* NB: force NULL result. */
479 		break;
480 	case IPSEC_POLICY_IPSEC:
481 		if (sp->req == NULL)	/* Acquire a SA. */
482 			*error = key_spdacquire(sp);
483 		break;
484 	}
485 	if (*error != 0) {
486 		KEY_FREESP(&sp);
487 		sp = NULL;
488 	}
489 	return (sp);
490 }
491 
492 static int
493 ipsec_setspidx_inpcb(struct mbuf *m, struct inpcb *inp)
494 {
495 	int error;
496 
497 	IPSEC_ASSERT(inp != NULL, ("null inp"));
498 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
499 	IPSEC_ASSERT(inp->inp_sp->sp_out != NULL && inp->inp_sp->sp_in != NULL,
500 		("null sp_in || sp_out"));
501 
502 	error = ipsec_setspidx(m, &inp->inp_sp->sp_in->spidx, 1);
503 	if (error == 0) {
504 		inp->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
505 		inp->inp_sp->sp_out->spidx = inp->inp_sp->sp_in->spidx;
506 		inp->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
507 	} else {
508 		bzero(&inp->inp_sp->sp_in->spidx,
509 			sizeof (inp->inp_sp->sp_in->spidx));
510 		bzero(&inp->inp_sp->sp_out->spidx,
511 			sizeof (inp->inp_sp->sp_in->spidx));
512 	}
513 	return (error);
514 }
515 
516 /*
517  * Configure security policy index (src/dst/proto/sport/dport)
518  * by looking at the content of mbuf.
519  * The caller is responsible for error recovery (like clearing up spidx).
520  */
521 static int
522 ipsec_setspidx(struct mbuf *m, struct secpolicyindex *spidx, int needport)
523 {
524 	struct ip *ip = NULL;
525 	struct ip ipbuf;
526 	u_int v;
527 	struct mbuf *n;
528 	int len;
529 	int error;
530 
531 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
532 
533 	/*
534 	 * Validate m->m_pkthdr.len.  We see incorrect length if we
535 	 * mistakenly call this function with inconsistent mbuf chain
536 	 * (like 4.4BSD tcp/udp processing).  XXX Should we panic here?
537 	 */
538 	len = 0;
539 	for (n = m; n; n = n->m_next)
540 		len += n->m_len;
541 	if (m->m_pkthdr.len != len) {
542 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
543 			printf("%s: pkthdr len(%d) mismatch (%d), ignored.\n",
544 				__func__, len, m->m_pkthdr.len));
545 		return (EINVAL);
546 	}
547 
548 	if (m->m_pkthdr.len < sizeof(struct ip)) {
549 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
550 			printf("%s: pkthdr len(%d) too small (v4), ignored.\n",
551 			    __func__, m->m_pkthdr.len));
552 		return (EINVAL);
553 	}
554 
555 	if (m->m_len >= sizeof(*ip))
556 		ip = mtod(m, struct ip *);
557 	else {
558 		m_copydata(m, 0, sizeof(ipbuf), (caddr_t)&ipbuf);
559 		ip = &ipbuf;
560 	}
561 	v = ip->ip_v;
562 	switch (v) {
563 	case 4:
564 		error = ipsec4_setspidx_ipaddr(m, spidx);
565 		if (error)
566 			return (error);
567 		ipsec4_get_ulp(m, spidx, needport);
568 		return (0);
569 #ifdef INET6
570 	case 6:
571 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) {
572 			KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
573 				printf("%s: pkthdr len(%d) too small (v6), "
574 				"ignored\n", __func__, m->m_pkthdr.len));
575 			return (EINVAL);
576 		}
577 		error = ipsec6_setspidx_ipaddr(m, spidx);
578 		if (error)
579 			return (error);
580 		ipsec6_get_ulp(m, spidx, needport);
581 		return (0);
582 #endif
583 	default:
584 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
585 			printf("%s: " "unknown IP version %u, ignored.\n",
586 				__func__, v));
587 		return (EINVAL);
588 	}
589 }
590 
591 static void
592 ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
593 {
594 	u_int8_t nxt;
595 	int off;
596 
597 	/* Sanity check. */
598 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
599 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),("packet too short"));
600 
601 	if (m->m_len >= sizeof (struct ip)) {
602 		struct ip *ip = mtod(m, struct ip *);
603 		if (ip->ip_off & htons(IP_MF | IP_OFFMASK))
604 			goto done;
605 		off = ip->ip_hl << 2;
606 		nxt = ip->ip_p;
607 	} else {
608 		struct ip ih;
609 
610 		m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih);
611 		if (ih.ip_off & htons(IP_MF | IP_OFFMASK))
612 			goto done;
613 		off = ih.ip_hl << 2;
614 		nxt = ih.ip_p;
615 	}
616 
617 	while (off < m->m_pkthdr.len) {
618 		struct ip6_ext ip6e;
619 		struct tcphdr th;
620 		struct udphdr uh;
621 
622 		switch (nxt) {
623 		case IPPROTO_TCP:
624 			spidx->ul_proto = nxt;
625 			if (!needport)
626 				goto done_proto;
627 			if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
628 				goto done;
629 			m_copydata(m, off, sizeof (th), (caddr_t) &th);
630 			spidx->src.sin.sin_port = th.th_sport;
631 			spidx->dst.sin.sin_port = th.th_dport;
632 			return;
633 		case IPPROTO_UDP:
634 			spidx->ul_proto = nxt;
635 			if (!needport)
636 				goto done_proto;
637 			if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
638 				goto done;
639 			m_copydata(m, off, sizeof (uh), (caddr_t) &uh);
640 			spidx->src.sin.sin_port = uh.uh_sport;
641 			spidx->dst.sin.sin_port = uh.uh_dport;
642 			return;
643 		case IPPROTO_AH:
644 			if (off + sizeof(ip6e) > m->m_pkthdr.len)
645 				goto done;
646 			/* XXX Sigh, this works but is totally bogus. */
647 			m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e);
648 			off += (ip6e.ip6e_len + 2) << 2;
649 			nxt = ip6e.ip6e_nxt;
650 			break;
651 		case IPPROTO_ICMP:
652 		default:
653 			/* XXX Intermediate headers??? */
654 			spidx->ul_proto = nxt;
655 			goto done_proto;
656 		}
657 	}
658 done:
659 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
660 done_proto:
661 	spidx->src.sin.sin_port = IPSEC_PORT_ANY;
662 	spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
663 }
664 
665 /* Assumes that m is sane. */
666 static int
667 ipsec4_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
668 {
669 	static const struct sockaddr_in template = {
670 		sizeof (struct sockaddr_in),
671 		AF_INET,
672 		0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 }
673 	};
674 
675 	spidx->src.sin = template;
676 	spidx->dst.sin = template;
677 
678 	if (m->m_len < sizeof (struct ip)) {
679 		m_copydata(m, offsetof(struct ip, ip_src),
680 			   sizeof (struct  in_addr),
681 			   (caddr_t) &spidx->src.sin.sin_addr);
682 		m_copydata(m, offsetof(struct ip, ip_dst),
683 			   sizeof (struct  in_addr),
684 			   (caddr_t) &spidx->dst.sin.sin_addr);
685 	} else {
686 		struct ip *ip = mtod(m, struct ip *);
687 		spidx->src.sin.sin_addr = ip->ip_src;
688 		spidx->dst.sin.sin_addr = ip->ip_dst;
689 	}
690 
691 	spidx->prefs = sizeof(struct in_addr) << 3;
692 	spidx->prefd = sizeof(struct in_addr) << 3;
693 
694 	return (0);
695 }
696 
697 #ifdef INET6
698 static void
699 ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
700 {
701 	int off, nxt;
702 	struct tcphdr th;
703 	struct udphdr uh;
704 	struct icmp6_hdr ih;
705 
706 	/* Sanity check. */
707 	if (m == NULL)
708 		panic("%s: NULL pointer was passed.\n", __func__);
709 
710 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
711 		printf("%s:\n", __func__); kdebug_mbuf(m));
712 
713 	/* Set default. */
714 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
715 	((struct sockaddr_in6 *)&spidx->src)->sin6_port = IPSEC_PORT_ANY;
716 	((struct sockaddr_in6 *)&spidx->dst)->sin6_port = IPSEC_PORT_ANY;
717 
718 	nxt = -1;
719 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
720 	if (off < 0 || m->m_pkthdr.len < off)
721 		return;
722 
723 	switch (nxt) {
724 	case IPPROTO_TCP:
725 		spidx->ul_proto = nxt;
726 		if (!needport)
727 			break;
728 		if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
729 			break;
730 		m_copydata(m, off, sizeof(th), (caddr_t)&th);
731 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = th.th_sport;
732 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = th.th_dport;
733 		break;
734 	case IPPROTO_UDP:
735 		spidx->ul_proto = nxt;
736 		if (!needport)
737 			break;
738 		if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
739 			break;
740 		m_copydata(m, off, sizeof(uh), (caddr_t)&uh);
741 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = uh.uh_sport;
742 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = uh.uh_dport;
743 		break;
744 	case IPPROTO_ICMPV6:
745 		spidx->ul_proto = nxt;
746 		if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len)
747 			break;
748 		m_copydata(m, off, sizeof(ih), (caddr_t)&ih);
749 		((struct sockaddr_in6 *)&spidx->src)->sin6_port =
750 		    htons((uint16_t)ih.icmp6_type);
751 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port =
752 		    htons((uint16_t)ih.icmp6_code);
753 		break;
754 	default:
755 		/* XXX Intermediate headers??? */
756 		spidx->ul_proto = nxt;
757 		break;
758 	}
759 }
760 
761 /* Assumes that m is sane. */
762 static int
763 ipsec6_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
764 {
765 	struct ip6_hdr *ip6 = NULL;
766 	struct ip6_hdr ip6buf;
767 	struct sockaddr_in6 *sin6;
768 
769 	if (m->m_len >= sizeof(*ip6))
770 		ip6 = mtod(m, struct ip6_hdr *);
771 	else {
772 		m_copydata(m, 0, sizeof(ip6buf), (caddr_t)&ip6buf);
773 		ip6 = &ip6buf;
774 	}
775 
776 	sin6 = (struct sockaddr_in6 *)&spidx->src;
777 	bzero(sin6, sizeof(*sin6));
778 	sin6->sin6_family = AF_INET6;
779 	sin6->sin6_len = sizeof(struct sockaddr_in6);
780 	bcopy(&ip6->ip6_src, &sin6->sin6_addr, sizeof(ip6->ip6_src));
781 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
782 		sin6->sin6_addr.s6_addr16[1] = 0;
783 		sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]);
784 	}
785 	spidx->prefs = sizeof(struct in6_addr) << 3;
786 
787 	sin6 = (struct sockaddr_in6 *)&spidx->dst;
788 	bzero(sin6, sizeof(*sin6));
789 	sin6->sin6_family = AF_INET6;
790 	sin6->sin6_len = sizeof(struct sockaddr_in6);
791 	bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(ip6->ip6_dst));
792 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
793 		sin6->sin6_addr.s6_addr16[1] = 0;
794 		sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]);
795 	}
796 	spidx->prefd = sizeof(struct in6_addr) << 3;
797 
798 	return (0);
799 }
800 #endif
801 
802 static void
803 ipsec_delpcbpolicy(struct inpcbpolicy *p)
804 {
805 
806 	free(p, M_IPSEC_INPCB);
807 }
808 
809 /* Initialize policy in PCB. */
810 int
811 ipsec_init_policy(struct socket *so, struct inpcbpolicy **pcb_sp)
812 {
813 	struct inpcbpolicy *new;
814 
815 	/* Sanity check. */
816 	if (so == NULL || pcb_sp == NULL)
817 		panic("%s: NULL pointer was passed.\n", __func__);
818 
819 	new = (struct inpcbpolicy *) malloc(sizeof(struct inpcbpolicy),
820 					    M_IPSEC_INPCB, M_NOWAIT|M_ZERO);
821 	if (new == NULL) {
822 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
823 		return (ENOBUFS);
824 	}
825 
826 	new->priv = IPSEC_IS_PRIVILEGED_SO(so);
827 
828 	if ((new->sp_in = KEY_NEWSP()) == NULL) {
829 		ipsec_delpcbpolicy(new);
830 		return (ENOBUFS);
831 	}
832 	new->sp_in->policy = IPSEC_POLICY_ENTRUST;
833 	if ((new->sp_out = KEY_NEWSP()) == NULL) {
834 		KEY_FREESP(&new->sp_in);
835 		ipsec_delpcbpolicy(new);
836 		return (ENOBUFS);
837 	}
838 	new->sp_out->policy = IPSEC_POLICY_ENTRUST;
839 	*pcb_sp = new;
840 
841 	return (0);
842 }
843 
844 /* Copy old IPsec policy into new. */
845 int
846 ipsec_copy_policy(struct inpcbpolicy *old, struct inpcbpolicy *new)
847 {
848 	struct secpolicy *sp;
849 
850 	sp = ipsec_deepcopy_policy(old->sp_in);
851 	if (sp) {
852 		KEY_FREESP(&new->sp_in);
853 		new->sp_in = sp;
854 	} else
855 		return (ENOBUFS);
856 
857 	sp = ipsec_deepcopy_policy(old->sp_out);
858 	if (sp) {
859 		KEY_FREESP(&new->sp_out);
860 		new->sp_out = sp;
861 	} else
862 		return (ENOBUFS);
863 
864 	new->priv = old->priv;
865 
866 	return (0);
867 }
868 
869 struct ipsecrequest *
870 ipsec_newisr(void)
871 {
872 	struct ipsecrequest *p;
873 
874 	p = malloc(sizeof(struct ipsecrequest), M_IPSEC_SR, M_NOWAIT|M_ZERO);
875 	if (p != NULL)
876 		IPSECREQUEST_LOCK_INIT(p);
877 	return (p);
878 }
879 
880 void
881 ipsec_delisr(struct ipsecrequest *p)
882 {
883 
884 	IPSECREQUEST_LOCK_DESTROY(p);
885 	free(p, M_IPSEC_SR);
886 }
887 
888 /* Deep-copy a policy in PCB. */
889 static struct secpolicy *
890 ipsec_deepcopy_policy(struct secpolicy *src)
891 {
892 	struct ipsecrequest *newchain = NULL;
893 	struct ipsecrequest *p;
894 	struct ipsecrequest **q;
895 	struct ipsecrequest *r;
896 	struct secpolicy *dst;
897 
898 	if (src == NULL)
899 		return (NULL);
900 	dst = KEY_NEWSP();
901 	if (dst == NULL)
902 		return (NULL);
903 
904 	/*
905 	 * Deep-copy IPsec request chain.  This is required since struct
906 	 * ipsecrequest is not reference counted.
907 	 */
908 	q = &newchain;
909 	for (p = src->req; p; p = p->next) {
910 		*q = ipsec_newisr();
911 		if (*q == NULL)
912 			goto fail;
913 		(*q)->saidx.proto = p->saidx.proto;
914 		(*q)->saidx.mode = p->saidx.mode;
915 		(*q)->level = p->level;
916 		(*q)->saidx.reqid = p->saidx.reqid;
917 
918 		bcopy(&p->saidx.src, &(*q)->saidx.src, sizeof((*q)->saidx.src));
919 		bcopy(&p->saidx.dst, &(*q)->saidx.dst, sizeof((*q)->saidx.dst));
920 
921 		(*q)->sp = dst;
922 
923 		q = &((*q)->next);
924 	}
925 
926 	dst->req = newchain;
927 	dst->policy = src->policy;
928 	/* Do not touch the refcnt fields. */
929 
930 	return (dst);
931 
932 fail:
933 	for (p = newchain; p; p = r) {
934 		r = p->next;
935 		ipsec_delisr(p);
936 		p = NULL;
937 	}
938 	return (NULL);
939 }
940 
941 /* Set policy and IPsec request if present. */
942 static int
943 ipsec_set_policy_internal(struct secpolicy **pcb_sp, int optname,
944     caddr_t request, size_t len, struct ucred *cred)
945 {
946 	struct sadb_x_policy *xpl;
947 	struct secpolicy *newsp = NULL;
948 	int error;
949 
950 	/* Sanity check. */
951 	if (pcb_sp == NULL || *pcb_sp == NULL || request == NULL)
952 		return (EINVAL);
953 	if (len < sizeof(*xpl))
954 		return (EINVAL);
955 	xpl = (struct sadb_x_policy *)request;
956 
957 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
958 		printf("%s: passed policy\n", __func__);
959 		kdebug_sadb_x_policy((struct sadb_ext *)xpl));
960 
961 	/* Check policy type. */
962 	/* ipsec_set_policy_internal() accepts IPSEC, ENTRUST and BYPASS. */
963 	if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD
964 	 || xpl->sadb_x_policy_type == IPSEC_POLICY_NONE)
965 		return (EINVAL);
966 
967 	/* Check privileged socket. */
968 	if (cred != NULL && xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
969 		error = priv_check_cred(cred, PRIV_NETINET_IPSEC, 0);
970 		if (error)
971 			return (EACCES);
972 	}
973 
974 	/* Allocating new SP entry. */
975 	if ((newsp = key_msg2sp(xpl, len, &error)) == NULL)
976 		return (error);
977 
978 	/* Clear old SP and set new SP. */
979 	KEY_FREESP(pcb_sp);
980 	*pcb_sp = newsp;
981 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
982 		printf("%s: new policy\n", __func__);
983 		kdebug_secpolicy(newsp));
984 
985 	return (0);
986 }
987 
988 int
989 ipsec_set_policy(struct inpcb *inp, int optname, caddr_t request,
990     size_t len, struct ucred *cred)
991 {
992 	struct sadb_x_policy *xpl;
993 	struct secpolicy **pcb_sp;
994 
995 	/* Sanity check. */
996 	if (inp == NULL || request == NULL)
997 		return (EINVAL);
998 	if (len < sizeof(*xpl))
999 		return (EINVAL);
1000 	xpl = (struct sadb_x_policy *)request;
1001 
1002 	/* Select direction. */
1003 	switch (xpl->sadb_x_policy_dir) {
1004 	case IPSEC_DIR_INBOUND:
1005 		pcb_sp = &inp->inp_sp->sp_in;
1006 		break;
1007 	case IPSEC_DIR_OUTBOUND:
1008 		pcb_sp = &inp->inp_sp->sp_out;
1009 		break;
1010 	default:
1011 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1012 			xpl->sadb_x_policy_dir));
1013 		return (EINVAL);
1014 	}
1015 
1016 	return (ipsec_set_policy_internal(pcb_sp, optname, request, len, cred));
1017 }
1018 
1019 int
1020 ipsec_get_policy(struct inpcb *inp, caddr_t request, size_t len,
1021     struct mbuf **mp)
1022 {
1023 	struct sadb_x_policy *xpl;
1024 	struct secpolicy *pcb_sp;
1025 
1026 	/* Sanity check. */
1027 	if (inp == NULL || request == NULL || mp == NULL)
1028 		return (EINVAL);
1029 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
1030 	if (len < sizeof(*xpl))
1031 		return (EINVAL);
1032 	xpl = (struct sadb_x_policy *)request;
1033 
1034 	/* Select direction. */
1035 	switch (xpl->sadb_x_policy_dir) {
1036 	case IPSEC_DIR_INBOUND:
1037 		pcb_sp = inp->inp_sp->sp_in;
1038 		break;
1039 	case IPSEC_DIR_OUTBOUND:
1040 		pcb_sp = inp->inp_sp->sp_out;
1041 		break;
1042 	default:
1043 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1044 			xpl->sadb_x_policy_dir));
1045 		return (EINVAL);
1046 	}
1047 
1048 	/* Sanity check. Should be an IPSEC_ASSERT. */
1049 	if (pcb_sp == NULL)
1050 		return (EINVAL);
1051 
1052 	*mp = key_sp2msg(pcb_sp);
1053 	if (!*mp) {
1054 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
1055 		return (ENOBUFS);
1056 	}
1057 
1058 	(*mp)->m_type = MT_DATA;
1059 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1060 		printf("%s:\n", __func__); kdebug_mbuf(*mp));
1061 
1062 	return (0);
1063 }
1064 
1065 /* Delete policy in PCB. */
1066 int
1067 ipsec_delete_pcbpolicy(struct inpcb *inp)
1068 {
1069 	IPSEC_ASSERT(inp != NULL, ("null inp"));
1070 
1071 	if (inp->inp_sp == NULL)
1072 		return (0);
1073 
1074 	if (inp->inp_sp->sp_in != NULL)
1075 		KEY_FREESP(&inp->inp_sp->sp_in);
1076 
1077 	if (inp->inp_sp->sp_out != NULL)
1078 		KEY_FREESP(&inp->inp_sp->sp_out);
1079 
1080 	ipsec_delpcbpolicy(inp->inp_sp);
1081 	inp->inp_sp = NULL;
1082 
1083 	return (0);
1084 }
1085 
1086 /*
1087  * Return current level.
1088  * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
1089  */
1090 u_int
1091 ipsec_get_reqlevel(struct ipsecrequest *isr)
1092 {
1093 	u_int level = 0;
1094 	u_int esp_trans_deflev, esp_net_deflev;
1095 	u_int ah_trans_deflev, ah_net_deflev;
1096 
1097 	IPSEC_ASSERT(isr != NULL && isr->sp != NULL, ("null argument"));
1098 	IPSEC_ASSERT(isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family,
1099 		("af family mismatch, src %u, dst %u",
1100 		 isr->sp->spidx.src.sa.sa_family,
1101 		 isr->sp->spidx.dst.sa.sa_family));
1102 
1103 /* XXX Note that we have ipseclog() expanded here - code sync issue. */
1104 #define IPSEC_CHECK_DEFAULT(lev) \
1105 	(((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE	      \
1106 			&& (lev) != IPSEC_LEVEL_UNIQUE)			      \
1107 		? (V_ipsec_debug						      \
1108 			? log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
1109 				(lev), IPSEC_LEVEL_REQUIRE)		      \
1110 			: 0),						      \
1111 			(lev) = IPSEC_LEVEL_REQUIRE,			      \
1112 			(lev)						      \
1113 		: (lev))
1114 
1115 	/* Set default level. */
1116 	switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) {
1117 #ifdef INET
1118 	case AF_INET:
1119 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev);
1120 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev);
1121 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev);
1122 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev);
1123 		break;
1124 #endif
1125 #ifdef INET6
1126 	case AF_INET6:
1127 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev);
1128 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev);
1129 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev);
1130 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev);
1131 		break;
1132 #endif /* INET6 */
1133 	default:
1134 		panic("%s: unknown af %u",
1135 			__func__, isr->sp->spidx.src.sa.sa_family);
1136 	}
1137 
1138 #undef IPSEC_CHECK_DEFAULT
1139 
1140 	/* Set level. */
1141 	switch (isr->level) {
1142 	case IPSEC_LEVEL_DEFAULT:
1143 		switch (isr->saidx.proto) {
1144 		case IPPROTO_ESP:
1145 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1146 				level = esp_net_deflev;
1147 			else
1148 				level = esp_trans_deflev;
1149 			break;
1150 		case IPPROTO_AH:
1151 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1152 				level = ah_net_deflev;
1153 			else
1154 				level = ah_trans_deflev;
1155 			break;
1156 		case IPPROTO_IPCOMP:
1157 			/*
1158 			 * We don't really care, as IPcomp document says that
1159 			 * we shouldn't compress small packets.
1160 			 */
1161 			level = IPSEC_LEVEL_USE;
1162 			break;
1163 		default:
1164 			panic("%s: Illegal protocol defined %u\n", __func__,
1165 				isr->saidx.proto);
1166 		}
1167 		break;
1168 
1169 	case IPSEC_LEVEL_USE:
1170 	case IPSEC_LEVEL_REQUIRE:
1171 		level = isr->level;
1172 		break;
1173 	case IPSEC_LEVEL_UNIQUE:
1174 		level = IPSEC_LEVEL_REQUIRE;
1175 		break;
1176 
1177 	default:
1178 		panic("%s: Illegal IPsec level %u\n", __func__, isr->level);
1179 	}
1180 
1181 	return (level);
1182 }
1183 
1184 /*
1185  * Check security policy requirements against the actual
1186  * packet contents.  Return one if the packet should be
1187  * reject as "invalid"; otherwiser return zero to have the
1188  * packet treated as "valid".
1189  *
1190  * OUT:
1191  *	0: valid
1192  *	1: invalid
1193  */
1194 int
1195 ipsec_in_reject(struct secpolicy *sp, struct mbuf *m)
1196 {
1197 	struct ipsecrequest *isr;
1198 	int need_auth;
1199 
1200 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1201 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1202 
1203 	/* Check policy. */
1204 	switch (sp->policy) {
1205 	case IPSEC_POLICY_DISCARD:
1206 		return (1);
1207 	case IPSEC_POLICY_BYPASS:
1208 	case IPSEC_POLICY_NONE:
1209 		return (0);
1210 	}
1211 
1212 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1213 		("invalid policy %u", sp->policy));
1214 
1215 	/* XXX Should compare policy against IPsec header history. */
1216 
1217 	need_auth = 0;
1218 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1219 		if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE)
1220 			continue;
1221 		switch (isr->saidx.proto) {
1222 		case IPPROTO_ESP:
1223 			if ((m->m_flags & M_DECRYPTED) == 0) {
1224 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1225 				    printf("%s: ESP m_flags:%x\n", __func__,
1226 					    m->m_flags));
1227 				return (1);
1228 			}
1229 
1230 			if (!need_auth &&
1231 			    isr->sav != NULL &&
1232 			    isr->sav->tdb_authalgxform != NULL &&
1233 			    (m->m_flags & M_AUTHIPDGM) == 0) {
1234 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1235 				    printf("%s: ESP/AH m_flags:%x\n", __func__,
1236 					    m->m_flags));
1237 				return (1);
1238 			}
1239 			break;
1240 		case IPPROTO_AH:
1241 			need_auth = 1;
1242 			if ((m->m_flags & M_AUTHIPHDR) == 0) {
1243 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1244 				    printf("%s: AH m_flags:%x\n", __func__,
1245 					    m->m_flags));
1246 				return (1);
1247 			}
1248 			break;
1249 		case IPPROTO_IPCOMP:
1250 			/*
1251 			 * We don't really care, as IPcomp document
1252 			 * says that we shouldn't compress small
1253 			 * packets.  IPComp policy should always be
1254 			 * treated as being in "use" level.
1255 			 */
1256 			break;
1257 		}
1258 	}
1259 	return (0);		/* Valid. */
1260 }
1261 
1262 /*
1263  * Non zero return value means security policy DISCARD or policy violation.
1264  */
1265 static int
1266 ipsec46_in_reject(struct mbuf *m, struct inpcb *inp)
1267 {
1268 	struct secpolicy *sp;
1269 	int error;
1270 	int result;
1271 
1272 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1273 
1274 	/* Get SP for this packet. */
1275 	if (inp == NULL)
1276 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, &error);
1277 	else
1278 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND, inp, &error);
1279 
1280 	if (sp != NULL) {
1281 		result = ipsec_in_reject(sp, m);
1282 		KEY_FREESP(&sp);
1283 	} else {
1284 		result = 1;	/* treat errors as policy violation */
1285 	}
1286 	return (result);
1287 }
1288 
1289 /*
1290  * Check AH/ESP integrity.
1291  * This function is called from tcp_input(), udp_input(),
1292  * and {ah,esp}4_input for tunnel mode.
1293  */
1294 int
1295 ipsec4_in_reject(struct mbuf *m, struct inpcb *inp)
1296 {
1297 	int result;
1298 
1299 	result = ipsec46_in_reject(m, inp);
1300 	if (result)
1301 		IPSECSTAT_INC(ips_in_polvio);
1302 
1303 	return (result);
1304 }
1305 
1306 #ifdef INET6
1307 /*
1308  * Check AH/ESP integrity.
1309  * This function is called from tcp6_input(), udp6_input(),
1310  * and {ah,esp}6_input for tunnel mode.
1311  */
1312 int
1313 ipsec6_in_reject(struct mbuf *m, struct inpcb *inp)
1314 {
1315 	int result;
1316 
1317 	result = ipsec46_in_reject(m, inp);
1318 	if (result)
1319 		IPSEC6STAT_INC(ips_in_polvio);
1320 
1321 	return (result);
1322 }
1323 #endif
1324 
1325 /*
1326  * Compute the byte size to be occupied by IPsec header.
1327  * In case it is tunnelled, it includes the size of outer IP header.
1328  * NOTE: SP passed is freed in this function.
1329  */
1330 static size_t
1331 ipsec_hdrsiz_internal(struct secpolicy *sp)
1332 {
1333 	struct ipsecrequest *isr;
1334 	size_t size;
1335 
1336 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1337 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1338 
1339 	switch (sp->policy) {
1340 	case IPSEC_POLICY_DISCARD:
1341 	case IPSEC_POLICY_BYPASS:
1342 	case IPSEC_POLICY_NONE:
1343 		return (0);
1344 	}
1345 
1346 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1347 		("invalid policy %u", sp->policy));
1348 
1349 	size = 0;
1350 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1351 		size_t clen = 0;
1352 
1353 		switch (isr->saidx.proto) {
1354 		case IPPROTO_ESP:
1355 			clen = esp_hdrsiz(isr->sav);
1356 			break;
1357 		case IPPROTO_AH:
1358 			clen = ah_hdrsiz(isr->sav);
1359 			break;
1360 		case IPPROTO_IPCOMP:
1361 			clen = sizeof(struct ipcomp);
1362 			break;
1363 		}
1364 
1365 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1366 			switch (isr->saidx.dst.sa.sa_family) {
1367 			case AF_INET:
1368 				clen += sizeof(struct ip);
1369 				break;
1370 #ifdef INET6
1371 			case AF_INET6:
1372 				clen += sizeof(struct ip6_hdr);
1373 				break;
1374 #endif
1375 			default:
1376 				ipseclog((LOG_ERR, "%s: unknown AF %d in "
1377 				    "IPsec tunnel SA\n", __func__,
1378 				    ((struct sockaddr *)&isr->saidx.dst)->sa_family));
1379 				break;
1380 			}
1381 		}
1382 		size += clen;
1383 	}
1384 
1385 	return (size);
1386 }
1387 
1388 /*
1389  * This function is called from ipsec_hdrsiz_tcp(), ip_ipsec_mtu(),
1390  * disabled ip6_ipsec_mtu() and ip6_forward().
1391  */
1392 size_t
1393 ipsec_hdrsiz(struct mbuf *m, u_int dir, struct inpcb *inp)
1394 {
1395 	struct secpolicy *sp;
1396 	int error;
1397 	size_t size;
1398 
1399 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1400 
1401 	/* Get SP for this packet. */
1402 	if (inp == NULL)
1403 		sp = ipsec_getpolicybyaddr(m, dir, &error);
1404 	else
1405 		sp = ipsec_getpolicybysock(m, dir, inp, &error);
1406 
1407 	if (sp != NULL) {
1408 		size = ipsec_hdrsiz_internal(sp);
1409 		KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1410 			printf("%s: size:%lu.\n", __func__,
1411 				(unsigned long)size));
1412 
1413 		KEY_FREESP(&sp);
1414 	} else {
1415 		size = 0;	/* XXX Should be panic?
1416 				 * -> No, we are called w/o knowing if
1417 				 *    IPsec processing is needed. */
1418 	}
1419 	return (size);
1420 }
1421 
1422 /*
1423  * Check the variable replay window.
1424  * ipsec_chkreplay() performs replay check before ICV verification.
1425  * ipsec_updatereplay() updates replay bitmap.  This must be called after
1426  * ICV verification (it also performs replay check, which is usually done
1427  * beforehand).
1428  * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1429  *
1430  * Based on RFC 2401.
1431  */
1432 int
1433 ipsec_chkreplay(u_int32_t seq, struct secasvar *sav)
1434 {
1435 	const struct secreplay *replay;
1436 	u_int32_t diff;
1437 	int fr;
1438 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1439 	int frlast;		/* Constant: last frame. */
1440 
1441 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1442 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1443 
1444 	replay = sav->replay;
1445 
1446 	if (replay->wsize == 0)
1447 		return (1);	/* No need to check replay. */
1448 
1449 	/* Constant. */
1450 	frlast = replay->wsize - 1;
1451 	wsizeb = replay->wsize << 3;
1452 
1453 	/* Sequence number of 0 is invalid. */
1454 	if (seq == 0)
1455 		return (0);
1456 
1457 	/* First time is always okay. */
1458 	if (replay->count == 0)
1459 		return (1);
1460 
1461 	if (seq > replay->lastseq) {
1462 		/* Larger sequences are okay. */
1463 		return (1);
1464 	} else {
1465 		/* seq is equal or less than lastseq. */
1466 		diff = replay->lastseq - seq;
1467 
1468 		/* Over range to check, i.e. too old or wrapped. */
1469 		if (diff >= wsizeb)
1470 			return (0);
1471 
1472 		fr = frlast - diff / 8;
1473 
1474 		/* This packet already seen? */
1475 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1476 			return (0);
1477 
1478 		/* Out of order but good. */
1479 		return (1);
1480 	}
1481 }
1482 
1483 /*
1484  * Check replay counter whether to update or not.
1485  * OUT:	0:	OK
1486  *	1:	NG
1487  */
1488 int
1489 ipsec_updatereplay(u_int32_t seq, struct secasvar *sav)
1490 {
1491 	struct secreplay *replay;
1492 	u_int32_t diff;
1493 	int fr;
1494 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1495 	int frlast;		/* Constant: last frame. */
1496 
1497 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1498 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1499 
1500 	replay = sav->replay;
1501 
1502 	if (replay->wsize == 0)
1503 		goto ok;	/* No need to check replay. */
1504 
1505 	/* Constant. */
1506 	frlast = replay->wsize - 1;
1507 	wsizeb = replay->wsize << 3;
1508 
1509 	/* Sequence number of 0 is invalid. */
1510 	if (seq == 0)
1511 		return (1);
1512 
1513 	/* First time. */
1514 	if (replay->count == 0) {
1515 		replay->lastseq = seq;
1516 		bzero(replay->bitmap, replay->wsize);
1517 		(replay->bitmap)[frlast] = 1;
1518 		goto ok;
1519 	}
1520 
1521 	if (seq > replay->lastseq) {
1522 		/* seq is larger than lastseq. */
1523 		diff = seq - replay->lastseq;
1524 
1525 		/* New larger sequence number. */
1526 		if (diff < wsizeb) {
1527 			/* In window. */
1528 			/* Set bit for this packet. */
1529 			vshiftl(replay->bitmap, diff, replay->wsize);
1530 			(replay->bitmap)[frlast] |= 1;
1531 		} else {
1532 			/* This packet has a "way larger". */
1533 			bzero(replay->bitmap, replay->wsize);
1534 			(replay->bitmap)[frlast] = 1;
1535 		}
1536 		replay->lastseq = seq;
1537 
1538 		/* Larger is good. */
1539 	} else {
1540 		/* seq is equal or less than lastseq. */
1541 		diff = replay->lastseq - seq;
1542 
1543 		/* Over range to check, i.e. too old or wrapped. */
1544 		if (diff >= wsizeb)
1545 			return (1);
1546 
1547 		fr = frlast - diff / 8;
1548 
1549 		/* This packet already seen? */
1550 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1551 			return (1);
1552 
1553 		/* Mark as seen. */
1554 		(replay->bitmap)[fr] |= (1 << (diff % 8));
1555 
1556 		/* Out of order but good. */
1557 	}
1558 
1559 ok:
1560 	if (replay->count == ~0) {
1561 
1562 		/* Set overflow flag. */
1563 		replay->overflow++;
1564 
1565 		/* Don't increment, no more packets accepted. */
1566 		if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0)
1567 			return (1);
1568 
1569 		ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n",
1570 		    __func__, replay->overflow, ipsec_logsastr(sav)));
1571 	}
1572 
1573 	replay->count++;
1574 
1575 	return (0);
1576 }
1577 
1578 /*
1579  * Shift variable length buffer to left.
1580  * IN:	bitmap: pointer to the buffer
1581  * 	nbit:	the number of to shift.
1582  *	wsize:	buffer size (bytes).
1583  */
1584 static void
1585 vshiftl(unsigned char *bitmap, int nbit, int wsize)
1586 {
1587 	int s, j, i;
1588 	unsigned char over;
1589 
1590 	for (j = 0; j < nbit; j += 8) {
1591 		s = (nbit - j < 8) ? (nbit - j): 8;
1592 		bitmap[0] <<= s;
1593 		for (i = 1; i < wsize; i++) {
1594 			over = (bitmap[i] >> (8 - s));
1595 			bitmap[i] <<= s;
1596 			bitmap[i-1] |= over;
1597 		}
1598 	}
1599 }
1600 
1601 #ifdef INET
1602 /* Return a printable string for the IPv4 address. */
1603 static char *
1604 inet_ntoa4(struct in_addr ina)
1605 {
1606 	static char buf[4][4 * sizeof "123" + 4];
1607 	unsigned char *ucp = (unsigned char *) &ina;
1608 	static int i = 3;
1609 
1610 	/* XXX-BZ Returns static buffer. */
1611 	i = (i + 1) % 4;
1612 	sprintf(buf[i], "%d.%d.%d.%d", ucp[0] & 0xff, ucp[1] & 0xff,
1613 	    ucp[2] & 0xff, ucp[3] & 0xff);
1614 	return (buf[i]);
1615 }
1616 #endif
1617 
1618 /* Return a printable string for the address. */
1619 char *
1620 ipsec_address(union sockaddr_union* sa)
1621 {
1622 #ifdef INET6
1623 	char ip6buf[INET6_ADDRSTRLEN];
1624 #endif
1625 
1626 	switch (sa->sa.sa_family) {
1627 #ifdef INET
1628 	case AF_INET:
1629 		return (inet_ntoa4(sa->sin.sin_addr));
1630 #endif /* INET */
1631 #ifdef INET6
1632 	case AF_INET6:
1633 		return (ip6_sprintf(ip6buf, &sa->sin6.sin6_addr));
1634 #endif /* INET6 */
1635 	default:
1636 		return ("(unknown address family)");
1637 	}
1638 }
1639 
1640 const char *
1641 ipsec_logsastr(struct secasvar *sav)
1642 {
1643 	static char buf[256];
1644 	char *p;
1645 	struct secasindex *saidx = &sav->sah->saidx;
1646 
1647 	IPSEC_ASSERT(saidx->src.sa.sa_family == saidx->dst.sa.sa_family,
1648 		("address family mismatch"));
1649 
1650 	p = buf;
1651 	snprintf(buf, sizeof(buf), "SA(SPI=%u ", (u_int32_t)ntohl(sav->spi));
1652 	while (p && *p)
1653 		p++;
1654 	/* NB: only use ipsec_address on one address at a time. */
1655 	snprintf(p, sizeof (buf) - (p - buf), "src=%s ",
1656 		ipsec_address(&saidx->src));
1657 	while (p && *p)
1658 		p++;
1659 	snprintf(p, sizeof (buf) - (p - buf), "dst=%s)",
1660 		ipsec_address(&saidx->dst));
1661 
1662 	return (buf);
1663 }
1664 
1665 void
1666 ipsec_dumpmbuf(struct mbuf *m)
1667 {
1668 	int totlen;
1669 	int i;
1670 	u_char *p;
1671 
1672 	totlen = 0;
1673 	printf("---\n");
1674 	while (m) {
1675 		p = mtod(m, u_char *);
1676 		for (i = 0; i < m->m_len; i++) {
1677 			printf("%02x ", p[i]);
1678 			totlen++;
1679 			if (totlen % 16 == 0)
1680 				printf("\n");
1681 		}
1682 		m = m->m_next;
1683 	}
1684 	if (totlen % 16 != 0)
1685 		printf("\n");
1686 	printf("---\n");
1687 }
1688 
1689 static void
1690 def_policy_init(const void *unused __unused)
1691 {
1692 
1693 	bzero(&V_def_policy, sizeof(struct secpolicy));
1694 	V_def_policy.policy = IPSEC_POLICY_NONE;
1695 	V_def_policy.refcnt = 1;
1696 }
1697 VNET_SYSINIT(def_policy_init, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY,
1698     def_policy_init, NULL);
1699 
1700 
1701 /* XXX This stuff doesn't belong here... */
1702 
1703 static	struct xformsw* xforms = NULL;
1704 
1705 /*
1706  * Register a transform; typically at system startup.
1707  */
1708 void
1709 xform_register(struct xformsw* xsp)
1710 {
1711 
1712 	xsp->xf_next = xforms;
1713 	xforms = xsp;
1714 }
1715 
1716 /*
1717  * Initialize transform support in an sav.
1718  */
1719 int
1720 xform_init(struct secasvar *sav, int xftype)
1721 {
1722 	struct xformsw *xsp;
1723 
1724 	if (sav->tdb_xform != NULL)	/* Previously initialized. */
1725 		return (0);
1726 	for (xsp = xforms; xsp; xsp = xsp->xf_next)
1727 		if (xsp->xf_type == xftype)
1728 			return ((*xsp->xf_init)(sav, xsp));
1729 	return (EINVAL);
1730 }
1731