xref: /freebsd/sys/netipsec/ipsec.c (revision f05cddf9)
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/route.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_DEFINE(struct ipsecstat, ipsec4stat);
108 VNET_DEFINE(int, ip4_ah_offsetmask) = 0;	/* maybe IP_DF? */
109 /* DF bit on encap. 0: clear 1: set 2: copy */
110 VNET_DEFINE(int, ip4_ipsec_dfbit) = 0;
111 VNET_DEFINE(int, ip4_esp_trans_deflev) = IPSEC_LEVEL_USE;
112 VNET_DEFINE(int, ip4_esp_net_deflev) = IPSEC_LEVEL_USE;
113 VNET_DEFINE(int, ip4_ah_trans_deflev) = IPSEC_LEVEL_USE;
114 VNET_DEFINE(int, ip4_ah_net_deflev) = IPSEC_LEVEL_USE;
115 VNET_DEFINE(struct secpolicy, ip4_def_policy);
116 /* ECN ignore(-1)/forbidden(0)/allowed(1) */
117 VNET_DEFINE(int, ip4_ipsec_ecn) = 0;
118 VNET_DEFINE(int, ip4_esp_randpad) = -1;
119 
120 /*
121  * Crypto support requirements:
122  *
123  *  1	require hardware support
124  * -1	require software support
125  *  0	take anything
126  */
127 VNET_DEFINE(int, crypto_support) = CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE;
128 
129 FEATURE(ipsec, "Internet Protocol Security (IPsec)");
130 #ifdef IPSEC_NAT_T
131 FEATURE(ipsec_natt, "UDP Encapsulation of IPsec ESP Packets ('NAT-T')");
132 #endif
133 
134 SYSCTL_DECL(_net_inet_ipsec);
135 
136 /* net.inet.ipsec */
137 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_POLICY, def_policy,
138 	CTLFLAG_RW, &VNET_NAME(ip4_def_policy).policy, 0,
139 	"IPsec default policy.");
140 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
141 	CTLFLAG_RW, &VNET_NAME(ip4_esp_trans_deflev), 0,
142 	"Default ESP transport mode level");
143 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
144 	CTLFLAG_RW, &VNET_NAME(ip4_esp_net_deflev), 0,
145 	"Default ESP tunnel mode level.");
146 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
147 	CTLFLAG_RW, &VNET_NAME(ip4_ah_trans_deflev), 0,
148 	"AH transfer mode default level.");
149 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
150 	CTLFLAG_RW, &VNET_NAME(ip4_ah_net_deflev), 0,
151 	"AH tunnel mode default level.");
152 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_AH_CLEARTOS, ah_cleartos,
153 	CTLFLAG_RW, &VNET_NAME(ah_cleartos), 0,
154 	"If set clear type-of-service field when doing AH computation.");
155 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_AH_OFFSETMASK, ah_offsetmask,
156 	CTLFLAG_RW, &VNET_NAME(ip4_ah_offsetmask), 0,
157 	"If not set clear offset field mask when doing AH computation.");
158 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DFBIT, dfbit,
159 	CTLFLAG_RW, &VNET_NAME(ip4_ipsec_dfbit), 0,
160 	"Do not fragment bit on encap.");
161 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_ECN, ecn,
162 	CTLFLAG_RW, &VNET_NAME(ip4_ipsec_ecn), 0,
163 	"Explicit Congestion Notification handling.");
164 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
165 	CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
166 	"Enable IPsec debugging output when set.");
167 SYSCTL_VNET_INT(_net_inet_ipsec, OID_AUTO, crypto_support,
168 	CTLFLAG_RW, &VNET_NAME(crypto_support), 0,
169 	"Crypto driver selection.");
170 SYSCTL_VNET_STRUCT(_net_inet_ipsec, OID_AUTO, ipsecstats,
171 	CTLFLAG_RD, &VNET_NAME(ipsec4stat), ipsecstat,
172 	"IPsec IPv4 statistics.");
173 
174 #ifdef REGRESSION
175 /*
176  * When set to 1, IPsec will send packets with the same sequence number.
177  * This allows to verify if the other side has proper replay attacks detection.
178  */
179 VNET_DEFINE(int, ipsec_replay) = 0;
180 SYSCTL_VNET_INT(_net_inet_ipsec, OID_AUTO, test_replay,
181 	CTLFLAG_RW, &VNET_NAME(ipsec_replay), 0,
182 	"Emulate replay attack");
183 /*
184  * When set 1, IPsec will send packets with corrupted HMAC.
185  * This allows to verify if the other side properly detects modified packets.
186  */
187 VNET_DEFINE(int, ipsec_integrity) = 0;
188 SYSCTL_VNET_INT(_net_inet_ipsec, OID_AUTO, test_integrity,
189 	CTLFLAG_RW, &VNET_NAME(ipsec_integrity), 0,
190 	"Emulate man-in-the-middle attack");
191 #endif
192 
193 #ifdef INET6
194 VNET_DEFINE(struct ipsecstat, ipsec6stat);
195 VNET_DEFINE(int, ip6_esp_trans_deflev) = IPSEC_LEVEL_USE;
196 VNET_DEFINE(int, ip6_esp_net_deflev) = IPSEC_LEVEL_USE;
197 VNET_DEFINE(int, ip6_ah_trans_deflev) = IPSEC_LEVEL_USE;
198 VNET_DEFINE(int, ip6_ah_net_deflev) = IPSEC_LEVEL_USE;
199 VNET_DEFINE(int, ip6_ipsec_ecn) = 0;	/* ECN ignore(-1)/forbidden(0)/allowed(1) */
200 
201 SYSCTL_DECL(_net_inet6_ipsec6);
202 
203 /* net.inet6.ipsec6 */
204 #ifdef COMPAT_KAME
205 SYSCTL_OID(_net_inet6_ipsec6, IPSECCTL_STATS, stats, CTLFLAG_RD,
206     0, 0, compat_ipsecstats_sysctl, "S", "IPsec IPv6 statistics.");
207 #endif /* COMPAT_KAME */
208 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_POLICY, def_policy, CTLFLAG_RW,
209 	&VNET_NAME(ip4_def_policy).policy, 0,
210 	"IPsec default policy.");
211 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV,
212 	esp_trans_deflev, CTLFLAG_RW, &VNET_NAME(ip6_esp_trans_deflev),	0,
213 	"Default ESP transport mode level.");
214 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV,
215 	esp_net_deflev, CTLFLAG_RW, &VNET_NAME(ip6_esp_net_deflev),	0,
216 	"Default ESP tunnel mode level.");
217 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV,
218 	ah_trans_deflev, CTLFLAG_RW, &VNET_NAME(ip6_ah_trans_deflev),	0,
219 	"AH transfer mode default level.");
220 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV,
221 	ah_net_deflev, CTLFLAG_RW, &VNET_NAME(ip6_ah_net_deflev),	0,
222 	"AH tunnel mode default level.");
223 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_ECN,
224 	ecn, CTLFLAG_RW, &VNET_NAME(ip6_ipsec_ecn),	0,
225 	"Explicit Congestion Notification handling.");
226 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug, CTLFLAG_RW,
227 	&VNET_NAME(ipsec_debug), 0,
228 	"Enable IPsec debugging output when set.");
229 SYSCTL_VNET_STRUCT(_net_inet6_ipsec6, IPSECCTL_STATS,
230 	ipsecstats, CTLFLAG_RD, &VNET_NAME(ipsec6stat), ipsecstat,
231 	"IPsec IPv6 statistics.");
232 #endif /* INET6 */
233 
234 static int ipsec_setspidx_inpcb __P((struct mbuf *, struct inpcb *));
235 static int ipsec_setspidx __P((struct mbuf *, struct secpolicyindex *, int));
236 static void ipsec4_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
237 static int ipsec4_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
238 #ifdef INET6
239 static void ipsec6_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
240 static int ipsec6_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
241 #endif
242 static void ipsec_delpcbpolicy __P((struct inpcbpolicy *));
243 static struct secpolicy *ipsec_deepcopy_policy __P((struct secpolicy *src));
244 static void vshiftl __P((unsigned char *, int, int));
245 
246 MALLOC_DEFINE(M_IPSEC_INPCB, "inpcbpolicy", "inpcb-resident ipsec policy");
247 
248 /*
249  * Return a held reference to the default SP.
250  */
251 static struct secpolicy *
252 key_allocsp_default(const char* where, int tag)
253 {
254 	struct secpolicy *sp;
255 
256 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
257 		printf("DP key_allocsp_default from %s:%u\n", where, tag));
258 
259 	sp = &V_ip4_def_policy;
260 	if (sp->policy != IPSEC_POLICY_DISCARD &&
261 	    sp->policy != IPSEC_POLICY_NONE) {
262 		ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n",
263 		    sp->policy, IPSEC_POLICY_NONE));
264 		sp->policy = IPSEC_POLICY_NONE;
265 	}
266 	key_addref(sp);
267 
268 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
269 		printf("DP key_allocsp_default returns SP:%p (%u)\n",
270 			sp, sp->refcnt));
271 	return (sp);
272 }
273 #define	KEY_ALLOCSP_DEFAULT() \
274 	key_allocsp_default(__FILE__, __LINE__)
275 
276 /*
277  * For OUTBOUND packet having a socket. Searching SPD for packet,
278  * and return a pointer to SP.
279  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
280  *		0	: bypass
281  *		EACCES	: discard packet.
282  *		ENOENT	: ipsec_acquire() in progress, maybe.
283  *		others	: error occured.
284  *	others:	a pointer to SP
285  *
286  * NOTE: IPv6 mapped adddress concern is implemented here.
287  */
288 struct secpolicy *
289 ipsec_getpolicy(struct tdb_ident *tdbi, u_int dir)
290 {
291 	struct secpolicy *sp;
292 
293 	IPSEC_ASSERT(tdbi != NULL, ("null tdbi"));
294 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
295 		("invalid direction %u", dir));
296 
297 	sp = KEY_ALLOCSP2(tdbi->spi, &tdbi->dst, tdbi->proto, dir);
298 	if (sp == NULL)			/*XXX????*/
299 		sp = KEY_ALLOCSP_DEFAULT();
300 	IPSEC_ASSERT(sp != NULL, ("null SP"));
301 	return (sp);
302 }
303 
304 /*
305  * For OUTBOUND packet having a socket. Searching SPD for packet,
306  * and return a pointer to SP.
307  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
308  *		0	: bypass
309  *		EACCES	: discard packet.
310  *		ENOENT	: ipsec_acquire() in progress, maybe.
311  *		others	: error occured.
312  *	others:	a pointer to SP
313  *
314  * NOTE: IPv6 mapped adddress concern is implemented here.
315  */
316 static struct secpolicy *
317 ipsec_getpolicybysock(struct mbuf *m, u_int dir, struct inpcb *inp, int *error)
318 {
319 	struct inpcbpolicy *pcbsp;
320 	struct secpolicy *currsp = NULL;	/* Policy on socket. */
321 	struct secpolicy *sp;
322 
323 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
324 	IPSEC_ASSERT(inp != NULL, ("null inpcb"));
325 	IPSEC_ASSERT(error != NULL, ("null error"));
326 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
327 		("invalid direction %u", dir));
328 
329 	/* Set spidx in pcb. */
330 	*error = ipsec_setspidx_inpcb(m, inp);
331 	if (*error)
332 		return (NULL);
333 
334 	pcbsp = inp->inp_sp;
335 	IPSEC_ASSERT(pcbsp != NULL, ("null pcbsp"));
336 	switch (dir) {
337 	case IPSEC_DIR_INBOUND:
338 		currsp = pcbsp->sp_in;
339 		break;
340 	case IPSEC_DIR_OUTBOUND:
341 		currsp = pcbsp->sp_out;
342 		break;
343 	}
344 	IPSEC_ASSERT(currsp != NULL, ("null currsp"));
345 
346 	if (pcbsp->priv) {			/* When privilieged socket. */
347 		switch (currsp->policy) {
348 		case IPSEC_POLICY_BYPASS:
349 		case IPSEC_POLICY_IPSEC:
350 			key_addref(currsp);
351 			sp = currsp;
352 			break;
353 
354 		case IPSEC_POLICY_ENTRUST:
355 			/* Look for a policy in SPD. */
356 			sp = KEY_ALLOCSP(&currsp->spidx, dir);
357 			if (sp == NULL)		/* No SP found. */
358 				sp = KEY_ALLOCSP_DEFAULT();
359 			break;
360 
361 		default:
362 			ipseclog((LOG_ERR, "%s: Invalid policy for PCB %d\n",
363 				__func__, currsp->policy));
364 			*error = EINVAL;
365 			return (NULL);
366 		}
367 	} else {				/* Unpriv, SPD has policy. */
368 		sp = KEY_ALLOCSP(&currsp->spidx, dir);
369 		if (sp == NULL) {		/* No SP found. */
370 			switch (currsp->policy) {
371 			case IPSEC_POLICY_BYPASS:
372 				ipseclog((LOG_ERR, "%s: Illegal policy for "
373 					"non-priviliged defined %d\n",
374 					__func__, currsp->policy));
375 				*error = EINVAL;
376 				return (NULL);
377 
378 			case IPSEC_POLICY_ENTRUST:
379 				sp = KEY_ALLOCSP_DEFAULT();
380 				break;
381 
382 			case IPSEC_POLICY_IPSEC:
383 				key_addref(currsp);
384 				sp = currsp;
385 				break;
386 
387 			default:
388 				ipseclog((LOG_ERR, "%s: Invalid policy for "
389 					"PCB %d\n", __func__, currsp->policy));
390 				*error = EINVAL;
391 				return (NULL);
392 			}
393 		}
394 	}
395 	IPSEC_ASSERT(sp != NULL,
396 		("null SP (priv %u policy %u", pcbsp->priv, currsp->policy));
397 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
398 		printf("DP %s (priv %u policy %u) allocate SP:%p (refcnt %u)\n",
399 			__func__, pcbsp->priv, currsp->policy, sp, sp->refcnt));
400 	return (sp);
401 }
402 
403 /*
404  * For FORWADING packet or OUTBOUND without a socket. Searching SPD for packet,
405  * and return a pointer to SP.
406  * OUT:	positive: a pointer to the entry for security policy leaf matched.
407  *	NULL:	no apropreate SP found, the following value is set to error.
408  *		0	: bypass
409  *		EACCES	: discard packet.
410  *		ENOENT	: ipsec_acquire() in progress, maybe.
411  *		others	: error occured.
412  */
413 struct secpolicy *
414 ipsec_getpolicybyaddr(struct mbuf *m, u_int dir, int flag, int *error)
415 {
416 	struct secpolicyindex spidx;
417 	struct secpolicy *sp;
418 
419 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
420 	IPSEC_ASSERT(error != NULL, ("null error"));
421 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
422 		("invalid direction %u", dir));
423 
424 	sp = NULL;
425 	if (key_havesp(dir)) {
426 		/* Make an index to look for a policy. */
427 		*error = ipsec_setspidx(m, &spidx,
428 					(flag & IP_FORWARDING) ? 0 : 1);
429 		if (*error != 0) {
430 			DPRINTF(("%s: setpidx failed, dir %u flag %u\n",
431 				__func__, dir, flag));
432 			return (NULL);
433 		}
434 		spidx.dir = dir;
435 
436 		sp = KEY_ALLOCSP(&spidx, dir);
437 	}
438 	if (sp == NULL)			/* No SP found, use system default. */
439 		sp = KEY_ALLOCSP_DEFAULT();
440 	IPSEC_ASSERT(sp != NULL, ("null SP"));
441 	return (sp);
442 }
443 
444 struct secpolicy *
445 ipsec4_checkpolicy(struct mbuf *m, u_int dir, u_int flag, int *error,
446     struct inpcb *inp)
447 {
448 	struct secpolicy *sp;
449 
450 	*error = 0;
451 	if (inp == NULL)
452 		sp = ipsec_getpolicybyaddr(m, dir, flag, error);
453 	else
454 		sp = ipsec_getpolicybysock(m, dir, inp, error);
455 	if (sp == NULL) {
456 		IPSEC_ASSERT(*error != 0, ("getpolicy failed w/o error"));
457 		IPSECSTAT_INC(ips_out_inval);
458 		return (NULL);
459 	}
460 	IPSEC_ASSERT(*error == 0, ("sp w/ error set to %u", *error));
461 	switch (sp->policy) {
462 	case IPSEC_POLICY_ENTRUST:
463 	default:
464 		printf("%s: invalid policy %u\n", __func__, sp->policy);
465 		/* FALLTHROUGH */
466 	case IPSEC_POLICY_DISCARD:
467 		IPSECSTAT_INC(ips_out_polvio);
468 		*error = -EINVAL;	/* Packet is discarded by caller. */
469 		break;
470 	case IPSEC_POLICY_BYPASS:
471 	case IPSEC_POLICY_NONE:
472 		KEY_FREESP(&sp);
473 		sp = NULL;		/* NB: force NULL result. */
474 		break;
475 	case IPSEC_POLICY_IPSEC:
476 		if (sp->req == NULL)	/* Acquire a SA. */
477 			*error = key_spdacquire(sp);
478 		break;
479 	}
480 	if (*error != 0) {
481 		KEY_FREESP(&sp);
482 		sp = NULL;
483 	}
484 	return (sp);
485 }
486 
487 static int
488 ipsec_setspidx_inpcb(struct mbuf *m, struct inpcb *inp)
489 {
490 	int error;
491 
492 	IPSEC_ASSERT(inp != NULL, ("null inp"));
493 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
494 	IPSEC_ASSERT(inp->inp_sp->sp_out != NULL && inp->inp_sp->sp_in != NULL,
495 		("null sp_in || sp_out"));
496 
497 	error = ipsec_setspidx(m, &inp->inp_sp->sp_in->spidx, 1);
498 	if (error == 0) {
499 		inp->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
500 		inp->inp_sp->sp_out->spidx = inp->inp_sp->sp_in->spidx;
501 		inp->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
502 	} else {
503 		bzero(&inp->inp_sp->sp_in->spidx,
504 			sizeof (inp->inp_sp->sp_in->spidx));
505 		bzero(&inp->inp_sp->sp_out->spidx,
506 			sizeof (inp->inp_sp->sp_in->spidx));
507 	}
508 	return (error);
509 }
510 
511 /*
512  * Configure security policy index (src/dst/proto/sport/dport)
513  * by looking at the content of mbuf.
514  * The caller is responsible for error recovery (like clearing up spidx).
515  */
516 static int
517 ipsec_setspidx(struct mbuf *m, struct secpolicyindex *spidx, int needport)
518 {
519 	struct ip *ip = NULL;
520 	struct ip ipbuf;
521 	u_int v;
522 	struct mbuf *n;
523 	int len;
524 	int error;
525 
526 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
527 
528 	/*
529 	 * Validate m->m_pkthdr.len.  We see incorrect length if we
530 	 * mistakenly call this function with inconsistent mbuf chain
531 	 * (like 4.4BSD tcp/udp processing).  XXX Should we panic here?
532 	 */
533 	len = 0;
534 	for (n = m; n; n = n->m_next)
535 		len += n->m_len;
536 	if (m->m_pkthdr.len != len) {
537 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
538 			printf("%s: pkthdr len(%d) mismatch (%d), ignored.\n",
539 				__func__, len, m->m_pkthdr.len));
540 		return (EINVAL);
541 	}
542 
543 	if (m->m_pkthdr.len < sizeof(struct ip)) {
544 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
545 			printf("%s: pkthdr len(%d) too small (v4), ignored.\n",
546 			    __func__, m->m_pkthdr.len));
547 		return (EINVAL);
548 	}
549 
550 	if (m->m_len >= sizeof(*ip))
551 		ip = mtod(m, struct ip *);
552 	else {
553 		m_copydata(m, 0, sizeof(ipbuf), (caddr_t)&ipbuf);
554 		ip = &ipbuf;
555 	}
556 #ifdef _IP_VHL
557 	v = _IP_VHL_V(ip->ip_vhl);
558 #else
559 	v = ip->ip_v;
560 #endif
561 	switch (v) {
562 	case 4:
563 		error = ipsec4_setspidx_ipaddr(m, spidx);
564 		if (error)
565 			return (error);
566 		ipsec4_get_ulp(m, spidx, needport);
567 		return (0);
568 #ifdef INET6
569 	case 6:
570 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) {
571 			KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
572 				printf("%s: pkthdr len(%d) too small (v6), "
573 				"ignored\n", __func__, m->m_pkthdr.len));
574 			return (EINVAL);
575 		}
576 		error = ipsec6_setspidx_ipaddr(m, spidx);
577 		if (error)
578 			return (error);
579 		ipsec6_get_ulp(m, spidx, needport);
580 		return (0);
581 #endif
582 	default:
583 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
584 			printf("%s: " "unknown IP version %u, ignored.\n",
585 				__func__, v));
586 		return (EINVAL);
587 	}
588 }
589 
590 static void
591 ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
592 {
593 	u_int8_t nxt;
594 	int off;
595 
596 	/* Sanity check. */
597 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
598 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),("packet too short"));
599 
600 	if (m->m_len >= sizeof (struct ip)) {
601 		struct ip *ip = mtod(m, struct ip *);
602 		if (ip->ip_off & htons(IP_MF | IP_OFFMASK))
603 			goto done;
604 #ifdef _IP_VHL
605 		off = _IP_VHL_HL(ip->ip_vhl) << 2;
606 #else
607 		off = ip->ip_hl << 2;
608 #endif
609 		nxt = ip->ip_p;
610 	} else {
611 		struct ip ih;
612 
613 		m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih);
614 		if (ih.ip_off & htons(IP_MF | IP_OFFMASK))
615 			goto done;
616 #ifdef _IP_VHL
617 		off = _IP_VHL_HL(ih.ip_vhl) << 2;
618 #else
619 		off = ih.ip_hl << 2;
620 #endif
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->state = IPSEC_SPSTATE_ALIVE;
840 	new->sp_in->policy = IPSEC_POLICY_ENTRUST;
841 
842 	if ((new->sp_out = KEY_NEWSP()) == NULL) {
843 		KEY_FREESP(&new->sp_in);
844 		ipsec_delpcbpolicy(new);
845 		return (ENOBUFS);
846 	}
847 	new->sp_out->state = IPSEC_SPSTATE_ALIVE;
848 	new->sp_out->policy = IPSEC_POLICY_ENTRUST;
849 
850 	*pcb_sp = new;
851 
852 	return (0);
853 }
854 
855 /* Copy old IPsec policy into new. */
856 int
857 ipsec_copy_policy(struct inpcbpolicy *old, struct inpcbpolicy *new)
858 {
859 	struct secpolicy *sp;
860 
861 	sp = ipsec_deepcopy_policy(old->sp_in);
862 	if (sp) {
863 		KEY_FREESP(&new->sp_in);
864 		new->sp_in = sp;
865 	} else
866 		return (ENOBUFS);
867 
868 	sp = ipsec_deepcopy_policy(old->sp_out);
869 	if (sp) {
870 		KEY_FREESP(&new->sp_out);
871 		new->sp_out = sp;
872 	} else
873 		return (ENOBUFS);
874 
875 	new->priv = old->priv;
876 
877 	return (0);
878 }
879 
880 struct ipsecrequest *
881 ipsec_newisr(void)
882 {
883 	struct ipsecrequest *p;
884 
885 	p = malloc(sizeof(struct ipsecrequest), M_IPSEC_SR, M_NOWAIT|M_ZERO);
886 	if (p != NULL)
887 		IPSECREQUEST_LOCK_INIT(p);
888 	return (p);
889 }
890 
891 void
892 ipsec_delisr(struct ipsecrequest *p)
893 {
894 
895 	IPSECREQUEST_LOCK_DESTROY(p);
896 	free(p, M_IPSEC_SR);
897 }
898 
899 /* Deep-copy a policy in PCB. */
900 static struct secpolicy *
901 ipsec_deepcopy_policy(struct secpolicy *src)
902 {
903 	struct ipsecrequest *newchain = NULL;
904 	struct ipsecrequest *p;
905 	struct ipsecrequest **q;
906 	struct ipsecrequest *r;
907 	struct secpolicy *dst;
908 
909 	if (src == NULL)
910 		return (NULL);
911 	dst = KEY_NEWSP();
912 	if (dst == NULL)
913 		return (NULL);
914 
915 	/*
916 	 * Deep-copy IPsec request chain.  This is required since struct
917 	 * ipsecrequest is not reference counted.
918 	 */
919 	q = &newchain;
920 	for (p = src->req; p; p = p->next) {
921 		*q = ipsec_newisr();
922 		if (*q == NULL)
923 			goto fail;
924 		(*q)->saidx.proto = p->saidx.proto;
925 		(*q)->saidx.mode = p->saidx.mode;
926 		(*q)->level = p->level;
927 		(*q)->saidx.reqid = p->saidx.reqid;
928 
929 		bcopy(&p->saidx.src, &(*q)->saidx.src, sizeof((*q)->saidx.src));
930 		bcopy(&p->saidx.dst, &(*q)->saidx.dst, sizeof((*q)->saidx.dst));
931 
932 		(*q)->sp = dst;
933 
934 		q = &((*q)->next);
935 	}
936 
937 	dst->req = newchain;
938 	dst->state = src->state;
939 	dst->policy = src->policy;
940 	/* Do not touch the refcnt fields. */
941 
942 	return (dst);
943 
944 fail:
945 	for (p = newchain; p; p = r) {
946 		r = p->next;
947 		ipsec_delisr(p);
948 		p = NULL;
949 	}
950 	return (NULL);
951 }
952 
953 /* Set policy and IPsec request if present. */
954 static int
955 ipsec_set_policy_internal(struct secpolicy **pcb_sp, int optname,
956     caddr_t request, size_t len, struct ucred *cred)
957 {
958 	struct sadb_x_policy *xpl;
959 	struct secpolicy *newsp = NULL;
960 	int error;
961 
962 	/* Sanity check. */
963 	if (pcb_sp == NULL || *pcb_sp == NULL || request == NULL)
964 		return (EINVAL);
965 	if (len < sizeof(*xpl))
966 		return (EINVAL);
967 	xpl = (struct sadb_x_policy *)request;
968 
969 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
970 		printf("%s: passed policy\n", __func__);
971 		kdebug_sadb_x_policy((struct sadb_ext *)xpl));
972 
973 	/* Check policy type. */
974 	/* ipsec_set_policy_internal() accepts IPSEC, ENTRUST and BYPASS. */
975 	if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD
976 	 || xpl->sadb_x_policy_type == IPSEC_POLICY_NONE)
977 		return (EINVAL);
978 
979 	/* Check privileged socket. */
980 	if (cred != NULL && xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
981 		error = priv_check_cred(cred, PRIV_NETINET_IPSEC, 0);
982 		if (error)
983 			return (EACCES);
984 	}
985 
986 	/* Allocating new SP entry. */
987 	if ((newsp = key_msg2sp(xpl, len, &error)) == NULL)
988 		return (error);
989 
990 	newsp->state = IPSEC_SPSTATE_ALIVE;
991 
992 	/* Clear old SP and set new SP. */
993 	KEY_FREESP(pcb_sp);
994 	*pcb_sp = newsp;
995 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
996 		printf("%s: new policy\n", __func__);
997 		kdebug_secpolicy(newsp));
998 
999 	return (0);
1000 }
1001 
1002 int
1003 ipsec_set_policy(struct inpcb *inp, int optname, caddr_t request,
1004     size_t len, struct ucred *cred)
1005 {
1006 	struct sadb_x_policy *xpl;
1007 	struct secpolicy **pcb_sp;
1008 
1009 	/* Sanity check. */
1010 	if (inp == NULL || request == NULL)
1011 		return (EINVAL);
1012 	if (len < sizeof(*xpl))
1013 		return (EINVAL);
1014 	xpl = (struct sadb_x_policy *)request;
1015 
1016 	/* Select direction. */
1017 	switch (xpl->sadb_x_policy_dir) {
1018 	case IPSEC_DIR_INBOUND:
1019 		pcb_sp = &inp->inp_sp->sp_in;
1020 		break;
1021 	case IPSEC_DIR_OUTBOUND:
1022 		pcb_sp = &inp->inp_sp->sp_out;
1023 		break;
1024 	default:
1025 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1026 			xpl->sadb_x_policy_dir));
1027 		return (EINVAL);
1028 	}
1029 
1030 	return (ipsec_set_policy_internal(pcb_sp, optname, request, len, cred));
1031 }
1032 
1033 int
1034 ipsec_get_policy(struct inpcb *inp, caddr_t request, size_t len,
1035     struct mbuf **mp)
1036 {
1037 	struct sadb_x_policy *xpl;
1038 	struct secpolicy *pcb_sp;
1039 
1040 	/* Sanity check. */
1041 	if (inp == NULL || request == NULL || mp == NULL)
1042 		return (EINVAL);
1043 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
1044 	if (len < sizeof(*xpl))
1045 		return (EINVAL);
1046 	xpl = (struct sadb_x_policy *)request;
1047 
1048 	/* Select direction. */
1049 	switch (xpl->sadb_x_policy_dir) {
1050 	case IPSEC_DIR_INBOUND:
1051 		pcb_sp = inp->inp_sp->sp_in;
1052 		break;
1053 	case IPSEC_DIR_OUTBOUND:
1054 		pcb_sp = inp->inp_sp->sp_out;
1055 		break;
1056 	default:
1057 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1058 			xpl->sadb_x_policy_dir));
1059 		return (EINVAL);
1060 	}
1061 
1062 	/* Sanity check. Should be an IPSEC_ASSERT. */
1063 	if (pcb_sp == NULL)
1064 		return (EINVAL);
1065 
1066 	*mp = key_sp2msg(pcb_sp);
1067 	if (!*mp) {
1068 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
1069 		return (ENOBUFS);
1070 	}
1071 
1072 	(*mp)->m_type = MT_DATA;
1073 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1074 		printf("%s:\n", __func__); kdebug_mbuf(*mp));
1075 
1076 	return (0);
1077 }
1078 
1079 /* Delete policy in PCB. */
1080 int
1081 ipsec_delete_pcbpolicy(struct inpcb *inp)
1082 {
1083 	IPSEC_ASSERT(inp != NULL, ("null inp"));
1084 
1085 	if (inp->inp_sp == NULL)
1086 		return (0);
1087 
1088 	if (inp->inp_sp->sp_in != NULL)
1089 		KEY_FREESP(&inp->inp_sp->sp_in);
1090 
1091 	if (inp->inp_sp->sp_out != NULL)
1092 		KEY_FREESP(&inp->inp_sp->sp_out);
1093 
1094 	ipsec_delpcbpolicy(inp->inp_sp);
1095 	inp->inp_sp = NULL;
1096 
1097 	return (0);
1098 }
1099 
1100 /*
1101  * Return current level.
1102  * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
1103  */
1104 u_int
1105 ipsec_get_reqlevel(struct ipsecrequest *isr)
1106 {
1107 	u_int level = 0;
1108 	u_int esp_trans_deflev, esp_net_deflev;
1109 	u_int ah_trans_deflev, ah_net_deflev;
1110 
1111 	IPSEC_ASSERT(isr != NULL && isr->sp != NULL, ("null argument"));
1112 	IPSEC_ASSERT(isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family,
1113 		("af family mismatch, src %u, dst %u",
1114 		 isr->sp->spidx.src.sa.sa_family,
1115 		 isr->sp->spidx.dst.sa.sa_family));
1116 
1117 /* XXX Note that we have ipseclog() expanded here - code sync issue. */
1118 #define IPSEC_CHECK_DEFAULT(lev) \
1119 	(((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE	      \
1120 			&& (lev) != IPSEC_LEVEL_UNIQUE)			      \
1121 		? (V_ipsec_debug						      \
1122 			? log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
1123 				(lev), IPSEC_LEVEL_REQUIRE)		      \
1124 			: 0),						      \
1125 			(lev) = IPSEC_LEVEL_REQUIRE,			      \
1126 			(lev)						      \
1127 		: (lev))
1128 
1129 	/* Set default level. */
1130 	switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) {
1131 #ifdef INET
1132 	case AF_INET:
1133 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev);
1134 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev);
1135 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev);
1136 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev);
1137 		break;
1138 #endif
1139 #ifdef INET6
1140 	case AF_INET6:
1141 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev);
1142 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev);
1143 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev);
1144 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev);
1145 		break;
1146 #endif /* INET6 */
1147 	default:
1148 		panic("%s: unknown af %u",
1149 			__func__, isr->sp->spidx.src.sa.sa_family);
1150 	}
1151 
1152 #undef IPSEC_CHECK_DEFAULT
1153 
1154 	/* Set level. */
1155 	switch (isr->level) {
1156 	case IPSEC_LEVEL_DEFAULT:
1157 		switch (isr->saidx.proto) {
1158 		case IPPROTO_ESP:
1159 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1160 				level = esp_net_deflev;
1161 			else
1162 				level = esp_trans_deflev;
1163 			break;
1164 		case IPPROTO_AH:
1165 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1166 				level = ah_net_deflev;
1167 			else
1168 				level = ah_trans_deflev;
1169 			break;
1170 		case IPPROTO_IPCOMP:
1171 			/*
1172 			 * We don't really care, as IPcomp document says that
1173 			 * we shouldn't compress small packets.
1174 			 */
1175 			level = IPSEC_LEVEL_USE;
1176 			break;
1177 		default:
1178 			panic("%s: Illegal protocol defined %u\n", __func__,
1179 				isr->saidx.proto);
1180 		}
1181 		break;
1182 
1183 	case IPSEC_LEVEL_USE:
1184 	case IPSEC_LEVEL_REQUIRE:
1185 		level = isr->level;
1186 		break;
1187 	case IPSEC_LEVEL_UNIQUE:
1188 		level = IPSEC_LEVEL_REQUIRE;
1189 		break;
1190 
1191 	default:
1192 		panic("%s: Illegal IPsec level %u\n", __func__, isr->level);
1193 	}
1194 
1195 	return (level);
1196 }
1197 
1198 /*
1199  * Check security policy requirements against the actual
1200  * packet contents.  Return one if the packet should be
1201  * reject as "invalid"; otherwiser return zero to have the
1202  * packet treated as "valid".
1203  *
1204  * OUT:
1205  *	0: valid
1206  *	1: invalid
1207  */
1208 int
1209 ipsec_in_reject(struct secpolicy *sp, struct mbuf *m)
1210 {
1211 	struct ipsecrequest *isr;
1212 	int need_auth;
1213 
1214 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1215 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1216 
1217 	/* Check policy. */
1218 	switch (sp->policy) {
1219 	case IPSEC_POLICY_DISCARD:
1220 		return (1);
1221 	case IPSEC_POLICY_BYPASS:
1222 	case IPSEC_POLICY_NONE:
1223 		return (0);
1224 	}
1225 
1226 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1227 		("invalid policy %u", sp->policy));
1228 
1229 	/* XXX Should compare policy against IPsec header history. */
1230 
1231 	need_auth = 0;
1232 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1233 		if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE)
1234 			continue;
1235 		switch (isr->saidx.proto) {
1236 		case IPPROTO_ESP:
1237 			if ((m->m_flags & M_DECRYPTED) == 0) {
1238 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1239 				    printf("%s: ESP m_flags:%x\n", __func__,
1240 					    m->m_flags));
1241 				return (1);
1242 			}
1243 
1244 			if (!need_auth &&
1245 			    isr->sav != NULL &&
1246 			    isr->sav->tdb_authalgxform != NULL &&
1247 			    (m->m_flags & M_AUTHIPDGM) == 0) {
1248 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1249 				    printf("%s: ESP/AH m_flags:%x\n", __func__,
1250 					    m->m_flags));
1251 				return (1);
1252 			}
1253 			break;
1254 		case IPPROTO_AH:
1255 			need_auth = 1;
1256 			if ((m->m_flags & M_AUTHIPHDR) == 0) {
1257 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1258 				    printf("%s: AH m_flags:%x\n", __func__,
1259 					    m->m_flags));
1260 				return (1);
1261 			}
1262 			break;
1263 		case IPPROTO_IPCOMP:
1264 			/*
1265 			 * We don't really care, as IPcomp document
1266 			 * says that we shouldn't compress small
1267 			 * packets.  IPComp policy should always be
1268 			 * treated as being in "use" level.
1269 			 */
1270 			break;
1271 		}
1272 	}
1273 	return (0);		/* Valid. */
1274 }
1275 
1276 static int
1277 ipsec46_in_reject(struct mbuf *m, struct inpcb *inp)
1278 {
1279 	struct secpolicy *sp;
1280 	int error;
1281 	int result;
1282 
1283 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1284 
1285 	/*
1286 	 * Get SP for this packet.
1287 	 * When we are called from ip_forward(), we call
1288 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1289 	 */
1290 	if (inp == NULL)
1291 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, IP_FORWARDING, &error);
1292 	else
1293 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND, inp, &error);
1294 
1295 	if (sp != NULL) {
1296 		result = ipsec_in_reject(sp, m);
1297 		KEY_FREESP(&sp);
1298 	} else {
1299 		result = 0;	/* XXX Should be panic?
1300 				 * -> No, there may be error. */
1301 	}
1302 	return (result);
1303 }
1304 
1305 /*
1306  * Check AH/ESP integrity.
1307  * This function is called from tcp_input(), udp_input(),
1308  * and {ah,esp}4_input for tunnel mode.
1309  */
1310 int
1311 ipsec4_in_reject(struct mbuf *m, struct inpcb *inp)
1312 {
1313 	int result;
1314 
1315 	result = ipsec46_in_reject(m, inp);
1316 	if (result)
1317 		IPSECSTAT_INC(ips_in_polvio);
1318 
1319 	return (result);
1320 }
1321 
1322 #ifdef INET6
1323 /*
1324  * Check AH/ESP integrity.
1325  * This function is called from tcp6_input(), udp6_input(),
1326  * and {ah,esp}6_input for tunnel mode.
1327  */
1328 int
1329 ipsec6_in_reject(struct mbuf *m, struct inpcb *inp)
1330 {
1331 	int result;
1332 
1333 	result = ipsec46_in_reject(m, inp);
1334 	if (result)
1335 		IPSEC6STAT_INC(ips_in_polvio);
1336 
1337 	return (result);
1338 }
1339 #endif
1340 
1341 /*
1342  * Compute the byte size to be occupied by IPsec header.
1343  * In case it is tunnelled, it includes the size of outer IP header.
1344  * NOTE: SP passed is freed in this function.
1345  */
1346 static size_t
1347 ipsec_hdrsiz_internal(struct secpolicy *sp)
1348 {
1349 	struct ipsecrequest *isr;
1350 	size_t size;
1351 
1352 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1353 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1354 
1355 	switch (sp->policy) {
1356 	case IPSEC_POLICY_DISCARD:
1357 	case IPSEC_POLICY_BYPASS:
1358 	case IPSEC_POLICY_NONE:
1359 		return (0);
1360 	}
1361 
1362 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1363 		("invalid policy %u", sp->policy));
1364 
1365 	size = 0;
1366 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1367 		size_t clen = 0;
1368 
1369 		switch (isr->saidx.proto) {
1370 		case IPPROTO_ESP:
1371 			clen = esp_hdrsiz(isr->sav);
1372 			break;
1373 		case IPPROTO_AH:
1374 			clen = ah_hdrsiz(isr->sav);
1375 			break;
1376 		case IPPROTO_IPCOMP:
1377 			clen = sizeof(struct ipcomp);
1378 			break;
1379 		}
1380 
1381 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1382 			switch (isr->saidx.dst.sa.sa_family) {
1383 			case AF_INET:
1384 				clen += sizeof(struct ip);
1385 				break;
1386 #ifdef INET6
1387 			case AF_INET6:
1388 				clen += sizeof(struct ip6_hdr);
1389 				break;
1390 #endif
1391 			default:
1392 				ipseclog((LOG_ERR, "%s: unknown AF %d in "
1393 				    "IPsec tunnel SA\n", __func__,
1394 				    ((struct sockaddr *)&isr->saidx.dst)->sa_family));
1395 				break;
1396 			}
1397 		}
1398 		size += clen;
1399 	}
1400 
1401 	return (size);
1402 }
1403 
1404 /*
1405  * This function is called from ipsec_hdrsiz_tcp(), ip_ipsec_mtu(),
1406  * disabled ip6_ipsec_mtu() and ip6_forward().
1407  */
1408 size_t
1409 ipsec_hdrsiz(struct mbuf *m, u_int dir, struct inpcb *inp)
1410 {
1411 	struct secpolicy *sp;
1412 	int error;
1413 	size_t size;
1414 
1415 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1416 
1417 	/* Get SP for this packet.
1418 	 * When we are called from ip_forward(), we call
1419 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1420 	 */
1421 	if (inp == NULL)
1422 		sp = ipsec_getpolicybyaddr(m, dir, IP_FORWARDING, &error);
1423 	else
1424 		sp = ipsec_getpolicybysock(m, dir, inp, &error);
1425 
1426 	if (sp != NULL) {
1427 		size = ipsec_hdrsiz_internal(sp);
1428 		KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1429 			printf("%s: size:%lu.\n", __func__,
1430 				(unsigned long)size));
1431 
1432 		KEY_FREESP(&sp);
1433 	} else {
1434 		size = 0;	/* XXX Should be panic?
1435 				 * -> No, we are called w/o knowing if
1436 				 *    IPsec processing is needed. */
1437 	}
1438 	return (size);
1439 }
1440 
1441 /*
1442  * Check the variable replay window.
1443  * ipsec_chkreplay() performs replay check before ICV verification.
1444  * ipsec_updatereplay() updates replay bitmap.  This must be called after
1445  * ICV verification (it also performs replay check, which is usually done
1446  * beforehand).
1447  * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1448  *
1449  * Based on RFC 2401.
1450  */
1451 int
1452 ipsec_chkreplay(u_int32_t seq, struct secasvar *sav)
1453 {
1454 	const struct secreplay *replay;
1455 	u_int32_t diff;
1456 	int fr;
1457 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1458 	int frlast;		/* Constant: last frame. */
1459 
1460 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1461 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1462 
1463 	replay = sav->replay;
1464 
1465 	if (replay->wsize == 0)
1466 		return (1);	/* No need to check replay. */
1467 
1468 	/* Constant. */
1469 	frlast = replay->wsize - 1;
1470 	wsizeb = replay->wsize << 3;
1471 
1472 	/* Sequence number of 0 is invalid. */
1473 	if (seq == 0)
1474 		return (0);
1475 
1476 	/* First time is always okay. */
1477 	if (replay->count == 0)
1478 		return (1);
1479 
1480 	if (seq > replay->lastseq) {
1481 		/* Larger sequences are okay. */
1482 		return (1);
1483 	} else {
1484 		/* seq is equal or less than lastseq. */
1485 		diff = replay->lastseq - seq;
1486 
1487 		/* Over range to check, i.e. too old or wrapped. */
1488 		if (diff >= wsizeb)
1489 			return (0);
1490 
1491 		fr = frlast - diff / 8;
1492 
1493 		/* This packet already seen? */
1494 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1495 			return (0);
1496 
1497 		/* Out of order but good. */
1498 		return (1);
1499 	}
1500 }
1501 
1502 /*
1503  * Check replay counter whether to update or not.
1504  * OUT:	0:	OK
1505  *	1:	NG
1506  */
1507 int
1508 ipsec_updatereplay(u_int32_t seq, struct secasvar *sav)
1509 {
1510 	struct secreplay *replay;
1511 	u_int32_t diff;
1512 	int fr;
1513 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1514 	int frlast;		/* Constant: last frame. */
1515 
1516 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1517 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1518 
1519 	replay = sav->replay;
1520 
1521 	if (replay->wsize == 0)
1522 		goto ok;	/* No need to check replay. */
1523 
1524 	/* Constant. */
1525 	frlast = replay->wsize - 1;
1526 	wsizeb = replay->wsize << 3;
1527 
1528 	/* Sequence number of 0 is invalid. */
1529 	if (seq == 0)
1530 		return (1);
1531 
1532 	/* First time. */
1533 	if (replay->count == 0) {
1534 		replay->lastseq = seq;
1535 		bzero(replay->bitmap, replay->wsize);
1536 		(replay->bitmap)[frlast] = 1;
1537 		goto ok;
1538 	}
1539 
1540 	if (seq > replay->lastseq) {
1541 		/* seq is larger than lastseq. */
1542 		diff = seq - replay->lastseq;
1543 
1544 		/* New larger sequence number. */
1545 		if (diff < wsizeb) {
1546 			/* In window. */
1547 			/* Set bit for this packet. */
1548 			vshiftl(replay->bitmap, diff, replay->wsize);
1549 			(replay->bitmap)[frlast] |= 1;
1550 		} else {
1551 			/* This packet has a "way larger". */
1552 			bzero(replay->bitmap, replay->wsize);
1553 			(replay->bitmap)[frlast] = 1;
1554 		}
1555 		replay->lastseq = seq;
1556 
1557 		/* Larger is good. */
1558 	} else {
1559 		/* seq is equal or less than lastseq. */
1560 		diff = replay->lastseq - seq;
1561 
1562 		/* Over range to check, i.e. too old or wrapped. */
1563 		if (diff >= wsizeb)
1564 			return (1);
1565 
1566 		fr = frlast - diff / 8;
1567 
1568 		/* This packet already seen? */
1569 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1570 			return (1);
1571 
1572 		/* Mark as seen. */
1573 		(replay->bitmap)[fr] |= (1 << (diff % 8));
1574 
1575 		/* Out of order but good. */
1576 	}
1577 
1578 ok:
1579 	if (replay->count == ~0) {
1580 
1581 		/* Set overflow flag. */
1582 		replay->overflow++;
1583 
1584 		/* Don't increment, no more packets accepted. */
1585 		if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0)
1586 			return (1);
1587 
1588 		ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n",
1589 		    __func__, replay->overflow, ipsec_logsastr(sav)));
1590 	}
1591 
1592 	replay->count++;
1593 
1594 	return (0);
1595 }
1596 
1597 /*
1598  * Shift variable length buffer to left.
1599  * IN:	bitmap: pointer to the buffer
1600  * 	nbit:	the number of to shift.
1601  *	wsize:	buffer size (bytes).
1602  */
1603 static void
1604 vshiftl(unsigned char *bitmap, int nbit, int wsize)
1605 {
1606 	int s, j, i;
1607 	unsigned char over;
1608 
1609 	for (j = 0; j < nbit; j += 8) {
1610 		s = (nbit - j < 8) ? (nbit - j): 8;
1611 		bitmap[0] <<= s;
1612 		for (i = 1; i < wsize; i++) {
1613 			over = (bitmap[i] >> (8 - s));
1614 			bitmap[i] <<= s;
1615 			bitmap[i-1] |= over;
1616 		}
1617 	}
1618 }
1619 
1620 #ifdef INET
1621 /* Return a printable string for the IPv4 address. */
1622 static char *
1623 inet_ntoa4(struct in_addr ina)
1624 {
1625 	static char buf[4][4 * sizeof "123" + 4];
1626 	unsigned char *ucp = (unsigned char *) &ina;
1627 	static int i = 3;
1628 
1629 	/* XXX-BZ Returns static buffer. */
1630 	i = (i + 1) % 4;
1631 	sprintf(buf[i], "%d.%d.%d.%d", ucp[0] & 0xff, ucp[1] & 0xff,
1632 	    ucp[2] & 0xff, ucp[3] & 0xff);
1633 	return (buf[i]);
1634 }
1635 #endif
1636 
1637 /* Return a printable string for the address. */
1638 char *
1639 ipsec_address(union sockaddr_union* sa)
1640 {
1641 #ifdef INET6
1642 	char ip6buf[INET6_ADDRSTRLEN];
1643 #endif
1644 
1645 	switch (sa->sa.sa_family) {
1646 #ifdef INET
1647 	case AF_INET:
1648 		return (inet_ntoa4(sa->sin.sin_addr));
1649 #endif /* INET */
1650 #ifdef INET6
1651 	case AF_INET6:
1652 		return (ip6_sprintf(ip6buf, &sa->sin6.sin6_addr));
1653 #endif /* INET6 */
1654 	default:
1655 		return ("(unknown address family)");
1656 	}
1657 }
1658 
1659 const char *
1660 ipsec_logsastr(struct secasvar *sav)
1661 {
1662 	static char buf[256];
1663 	char *p;
1664 	struct secasindex *saidx = &sav->sah->saidx;
1665 
1666 	IPSEC_ASSERT(saidx->src.sa.sa_family == saidx->dst.sa.sa_family,
1667 		("address family mismatch"));
1668 
1669 	p = buf;
1670 	snprintf(buf, sizeof(buf), "SA(SPI=%u ", (u_int32_t)ntohl(sav->spi));
1671 	while (p && *p)
1672 		p++;
1673 	/* NB: only use ipsec_address on one address at a time. */
1674 	snprintf(p, sizeof (buf) - (p - buf), "src=%s ",
1675 		ipsec_address(&saidx->src));
1676 	while (p && *p)
1677 		p++;
1678 	snprintf(p, sizeof (buf) - (p - buf), "dst=%s)",
1679 		ipsec_address(&saidx->dst));
1680 
1681 	return (buf);
1682 }
1683 
1684 void
1685 ipsec_dumpmbuf(struct mbuf *m)
1686 {
1687 	int totlen;
1688 	int i;
1689 	u_char *p;
1690 
1691 	totlen = 0;
1692 	printf("---\n");
1693 	while (m) {
1694 		p = mtod(m, u_char *);
1695 		for (i = 0; i < m->m_len; i++) {
1696 			printf("%02x ", p[i]);
1697 			totlen++;
1698 			if (totlen % 16 == 0)
1699 				printf("\n");
1700 		}
1701 		m = m->m_next;
1702 	}
1703 	if (totlen % 16 != 0)
1704 		printf("\n");
1705 	printf("---\n");
1706 }
1707 
1708 static void
1709 ipsec_init(const void *unused __unused)
1710 {
1711 
1712 	SECPOLICY_LOCK_INIT(&V_ip4_def_policy);
1713 	V_ip4_def_policy.refcnt = 1;			/* NB: disallow free. */
1714 }
1715 VNET_SYSINIT(ipsec_init, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, ipsec_init,
1716     NULL);
1717 
1718 
1719 /* XXX This stuff doesn't belong here... */
1720 
1721 static	struct xformsw* xforms = NULL;
1722 
1723 /*
1724  * Register a transform; typically at system startup.
1725  */
1726 void
1727 xform_register(struct xformsw* xsp)
1728 {
1729 
1730 	xsp->xf_next = xforms;
1731 	xforms = xsp;
1732 }
1733 
1734 /*
1735  * Initialize transform support in an sav.
1736  */
1737 int
1738 xform_init(struct secasvar *sav, int xftype)
1739 {
1740 	struct xformsw *xsp;
1741 
1742 	if (sav->tdb_xform != NULL)	/* Previously initialized. */
1743 		return (0);
1744 	for (xsp = xforms; xsp; xsp = xsp->xf_next)
1745 		if (xsp->xf_type == xftype)
1746 			return ((*xsp->xf_init)(sav, xsp));
1747 	return (EINVAL);
1748 }
1749