xref: /freebsd/sys/netipsec/ipsec.c (revision bbb29a3c0f2c4565eff6fda70426807b6ed97f8b)
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 VNET_DEFINE(struct secpolicy, ip4_def_policy);
122 /* ECN ignore(-1)/forbidden(0)/allowed(1) */
123 VNET_DEFINE(int, ip4_ipsec_ecn) = 0;
124 VNET_DEFINE(int, ip4_esp_randpad) = -1;
125 
126 /*
127  * Crypto support requirements:
128  *
129  *  1	require hardware support
130  * -1	require software support
131  *  0	take anything
132  */
133 VNET_DEFINE(int, crypto_support) = CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE;
134 
135 FEATURE(ipsec, "Internet Protocol Security (IPsec)");
136 #ifdef IPSEC_NAT_T
137 FEATURE(ipsec_natt, "UDP Encapsulation of IPsec ESP Packets ('NAT-T')");
138 #endif
139 
140 SYSCTL_DECL(_net_inet_ipsec);
141 
142 /* net.inet.ipsec */
143 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_POLICY, def_policy,
144 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_def_policy).policy, 0,
145 	"IPsec default policy.");
146 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
147 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_esp_trans_deflev), 0,
148 	"Default ESP transport mode level");
149 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
150 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_esp_net_deflev), 0,
151 	"Default ESP tunnel mode level.");
152 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
153 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_trans_deflev), 0,
154 	"AH transfer mode default level.");
155 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
156 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_net_deflev), 0,
157 	"AH tunnel mode default level.");
158 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_CLEARTOS, ah_cleartos,
159 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ah_cleartos), 0,
160 	"If set clear type-of-service field when doing AH computation.");
161 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_AH_OFFSETMASK, ah_offsetmask,
162 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ah_offsetmask), 0,
163 	"If not set clear offset field mask when doing AH computation.");
164 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DFBIT, dfbit,
165 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_dfbit), 0,
166 	"Do not fragment bit on encap.");
167 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_ECN, ecn,
168 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_ipsec_ecn), 0,
169 	"Explicit Congestion Notification handling.");
170 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
171 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
172 	"Enable IPsec debugging output when set.");
173 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, crypto_support,
174 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(crypto_support), 0,
175 	"Crypto driver selection.");
176 SYSCTL_VNET_PCPUSTAT(_net_inet_ipsec, OID_AUTO, ipsecstats, struct ipsecstat,
177     ipsec4stat, "IPsec IPv4 statistics.");
178 
179 #ifdef REGRESSION
180 /*
181  * When set to 1, IPsec will send packets with the same sequence number.
182  * This allows to verify if the other side has proper replay attacks detection.
183  */
184 VNET_DEFINE(int, ipsec_replay) = 0;
185 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_replay,
186 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_replay), 0,
187 	"Emulate replay attack");
188 /*
189  * When set 1, IPsec will send packets with corrupted HMAC.
190  * This allows to verify if the other side properly detects modified packets.
191  */
192 VNET_DEFINE(int, ipsec_integrity) = 0;
193 SYSCTL_INT(_net_inet_ipsec, OID_AUTO, test_integrity,
194 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_integrity), 0,
195 	"Emulate man-in-the-middle attack");
196 #endif
197 
198 #ifdef INET6
199 VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec6stat);
200 VNET_PCPUSTAT_SYSINIT(ipsec6stat);
201 
202 #ifdef VIMAGE
203 VNET_PCPUSTAT_SYSUNINIT(ipsec6stat);
204 #endif /* VIMAGE */
205 
206 VNET_DEFINE(int, ip6_esp_trans_deflev) = IPSEC_LEVEL_USE;
207 VNET_DEFINE(int, ip6_esp_net_deflev) = IPSEC_LEVEL_USE;
208 VNET_DEFINE(int, ip6_ah_trans_deflev) = IPSEC_LEVEL_USE;
209 VNET_DEFINE(int, ip6_ah_net_deflev) = IPSEC_LEVEL_USE;
210 VNET_DEFINE(int, ip6_ipsec_ecn) = 0;	/* ECN ignore(-1)/forbidden(0)/allowed(1) */
211 
212 SYSCTL_DECL(_net_inet6_ipsec6);
213 
214 /* net.inet6.ipsec6 */
215 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_POLICY, def_policy,
216 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip4_def_policy).policy, 0,
217 	"IPsec default policy.");
218 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
219 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_esp_trans_deflev), 0,
220 	"Default ESP transport mode level.");
221 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
222 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_esp_net_deflev), 0,
223 	"Default ESP tunnel mode level.");
224 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
225 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ah_trans_deflev), 0,
226 	"AH transfer mode default level.");
227 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
228 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ah_net_deflev), 0,
229 	"AH tunnel mode default level.");
230 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_ECN, ecn,
231 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip6_ipsec_ecn), 0,
232 	"Explicit Congestion Notification handling.");
233 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug,
234 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
235 	"Enable IPsec debugging output when set.");
236 SYSCTL_VNET_PCPUSTAT(_net_inet6_ipsec6, IPSECCTL_STATS, ipsecstats,
237     struct ipsecstat, ipsec6stat, "IPsec IPv6 statistics.");
238 #endif /* INET6 */
239 
240 static int ipsec_setspidx_inpcb(struct mbuf *, struct inpcb *);
241 static int ipsec_setspidx(struct mbuf *, struct secpolicyindex *, int);
242 static void ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *, int);
243 static int ipsec4_setspidx_ipaddr(struct mbuf *, struct secpolicyindex *);
244 #ifdef INET6
245 static void ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *, int);
246 static int ipsec6_setspidx_ipaddr(struct mbuf *, struct secpolicyindex *);
247 #endif
248 static void ipsec_delpcbpolicy(struct inpcbpolicy *);
249 static struct secpolicy *ipsec_deepcopy_policy(struct secpolicy *src);
250 static void vshiftl(unsigned char *, int, int);
251 
252 MALLOC_DEFINE(M_IPSEC_INPCB, "inpcbpolicy", "inpcb-resident ipsec policy");
253 
254 /*
255  * Return a held reference to the default SP.
256  */
257 static struct secpolicy *
258 key_allocsp_default(const char* where, int tag)
259 {
260 	struct secpolicy *sp;
261 
262 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
263 		printf("DP key_allocsp_default from %s:%u\n", where, tag));
264 
265 	sp = &V_ip4_def_policy;
266 	if (sp->policy != IPSEC_POLICY_DISCARD &&
267 	    sp->policy != IPSEC_POLICY_NONE) {
268 		ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n",
269 		    sp->policy, IPSEC_POLICY_NONE));
270 		sp->policy = IPSEC_POLICY_NONE;
271 	}
272 	key_addref(sp);
273 
274 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
275 		printf("DP key_allocsp_default returns SP:%p (%u)\n",
276 			sp, sp->refcnt));
277 	return (sp);
278 }
279 #define	KEY_ALLOCSP_DEFAULT() \
280 	key_allocsp_default(__FILE__, __LINE__)
281 
282 /*
283  * For OUTBOUND packet having a socket. Searching SPD for packet,
284  * and return a pointer to SP.
285  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
286  *		0	: bypass
287  *		EACCES	: discard packet.
288  *		ENOENT	: ipsec_acquire() in progress, maybe.
289  *		others	: error occured.
290  *	others:	a pointer to SP
291  *
292  * NOTE: IPv6 mapped adddress concern is implemented here.
293  */
294 struct secpolicy *
295 ipsec_getpolicy(struct tdb_ident *tdbi, u_int dir)
296 {
297 	struct secpolicy *sp;
298 
299 	IPSEC_ASSERT(tdbi != NULL, ("null tdbi"));
300 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
301 		("invalid direction %u", dir));
302 
303 	sp = KEY_ALLOCSP2(tdbi->spi, &tdbi->dst, tdbi->proto, dir);
304 	if (sp == NULL)			/*XXX????*/
305 		sp = KEY_ALLOCSP_DEFAULT();
306 	IPSEC_ASSERT(sp != NULL, ("null SP"));
307 	return (sp);
308 }
309 
310 /*
311  * For OUTBOUND packet having a socket. Searching SPD for packet,
312  * and return a pointer to SP.
313  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
314  *		0	: bypass
315  *		EACCES	: discard packet.
316  *		ENOENT	: ipsec_acquire() in progress, maybe.
317  *		others	: error occured.
318  *	others:	a pointer to SP
319  *
320  * NOTE: IPv6 mapped adddress concern is implemented here.
321  */
322 static struct secpolicy *
323 ipsec_getpolicybysock(struct mbuf *m, u_int dir, struct inpcb *inp, int *error)
324 {
325 	struct inpcbpolicy *pcbsp;
326 	struct secpolicy *currsp = NULL;	/* Policy on socket. */
327 	struct secpolicy *sp;
328 
329 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
330 	IPSEC_ASSERT(inp != NULL, ("null inpcb"));
331 	IPSEC_ASSERT(error != NULL, ("null error"));
332 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
333 		("invalid direction %u", dir));
334 
335 	/* Set spidx in pcb. */
336 	*error = ipsec_setspidx_inpcb(m, inp);
337 	if (*error)
338 		return (NULL);
339 
340 	pcbsp = inp->inp_sp;
341 	IPSEC_ASSERT(pcbsp != NULL, ("null pcbsp"));
342 	switch (dir) {
343 	case IPSEC_DIR_INBOUND:
344 		currsp = pcbsp->sp_in;
345 		break;
346 	case IPSEC_DIR_OUTBOUND:
347 		currsp = pcbsp->sp_out;
348 		break;
349 	}
350 	IPSEC_ASSERT(currsp != NULL, ("null currsp"));
351 
352 	if (pcbsp->priv) {			/* When privilieged socket. */
353 		switch (currsp->policy) {
354 		case IPSEC_POLICY_BYPASS:
355 		case IPSEC_POLICY_IPSEC:
356 			key_addref(currsp);
357 			sp = currsp;
358 			break;
359 
360 		case IPSEC_POLICY_ENTRUST:
361 			/* Look for a policy in SPD. */
362 			sp = KEY_ALLOCSP(&currsp->spidx, dir);
363 			if (sp == NULL)		/* No SP found. */
364 				sp = KEY_ALLOCSP_DEFAULT();
365 			break;
366 
367 		default:
368 			ipseclog((LOG_ERR, "%s: Invalid policy for PCB %d\n",
369 				__func__, currsp->policy));
370 			*error = EINVAL;
371 			return (NULL);
372 		}
373 	} else {				/* Unpriv, SPD has policy. */
374 		sp = KEY_ALLOCSP(&currsp->spidx, dir);
375 		if (sp == NULL) {		/* No SP found. */
376 			switch (currsp->policy) {
377 			case IPSEC_POLICY_BYPASS:
378 				ipseclog((LOG_ERR, "%s: Illegal policy for "
379 					"non-priviliged defined %d\n",
380 					__func__, currsp->policy));
381 				*error = EINVAL;
382 				return (NULL);
383 
384 			case IPSEC_POLICY_ENTRUST:
385 				sp = KEY_ALLOCSP_DEFAULT();
386 				break;
387 
388 			case IPSEC_POLICY_IPSEC:
389 				key_addref(currsp);
390 				sp = currsp;
391 				break;
392 
393 			default:
394 				ipseclog((LOG_ERR, "%s: Invalid policy for "
395 					"PCB %d\n", __func__, currsp->policy));
396 				*error = EINVAL;
397 				return (NULL);
398 			}
399 		}
400 	}
401 	IPSEC_ASSERT(sp != NULL,
402 		("null SP (priv %u policy %u", pcbsp->priv, currsp->policy));
403 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
404 		printf("DP %s (priv %u policy %u) allocate SP:%p (refcnt %u)\n",
405 			__func__, pcbsp->priv, currsp->policy, sp, sp->refcnt));
406 	return (sp);
407 }
408 
409 /*
410  * For FORWADING packet or OUTBOUND without a socket. Searching SPD for packet,
411  * and return a pointer to SP.
412  * OUT:	positive: a pointer to the entry for security policy leaf matched.
413  *	NULL:	no apropreate SP found, the following value is set to error.
414  *		0	: bypass
415  *		EACCES	: discard packet.
416  *		ENOENT	: ipsec_acquire() in progress, maybe.
417  *		others	: error occured.
418  */
419 struct secpolicy *
420 ipsec_getpolicybyaddr(struct mbuf *m, u_int dir, int flag, int *error)
421 {
422 	struct secpolicyindex spidx;
423 	struct secpolicy *sp;
424 
425 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
426 	IPSEC_ASSERT(error != NULL, ("null error"));
427 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
428 		("invalid direction %u", dir));
429 
430 	sp = NULL;
431 	if (key_havesp(dir)) {
432 		/* Make an index to look for a policy. */
433 		*error = ipsec_setspidx(m, &spidx,
434 					(flag & IP_FORWARDING) ? 0 : 1);
435 		if (*error != 0) {
436 			DPRINTF(("%s: setpidx failed, dir %u flag %u\n",
437 				__func__, dir, flag));
438 			return (NULL);
439 		}
440 		spidx.dir = dir;
441 
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, u_int flag, int *error,
452     struct inpcb *inp)
453 {
454 	struct secpolicy *sp;
455 
456 	*error = 0;
457 	if (inp == NULL)
458 		sp = ipsec_getpolicybyaddr(m, dir, flag, error);
459 	else
460 		sp = ipsec_getpolicybysock(m, dir, inp, error);
461 	if (sp == NULL) {
462 		IPSEC_ASSERT(*error != 0, ("getpolicy failed w/o error"));
463 		IPSECSTAT_INC(ips_out_inval);
464 		return (NULL);
465 	}
466 	IPSEC_ASSERT(*error == 0, ("sp w/ error set to %u", *error));
467 	switch (sp->policy) {
468 	case IPSEC_POLICY_ENTRUST:
469 	default:
470 		printf("%s: invalid policy %u\n", __func__, sp->policy);
471 		/* FALLTHROUGH */
472 	case IPSEC_POLICY_DISCARD:
473 		IPSECSTAT_INC(ips_out_polvio);
474 		*error = -EINVAL;	/* Packet is discarded by caller. */
475 		break;
476 	case IPSEC_POLICY_BYPASS:
477 	case IPSEC_POLICY_NONE:
478 		KEY_FREESP(&sp);
479 		sp = NULL;		/* NB: force NULL result. */
480 		break;
481 	case IPSEC_POLICY_IPSEC:
482 		if (sp->req == NULL)	/* Acquire a SA. */
483 			*error = key_spdacquire(sp);
484 		break;
485 	}
486 	if (*error != 0) {
487 		KEY_FREESP(&sp);
488 		sp = NULL;
489 	}
490 	return (sp);
491 }
492 
493 static int
494 ipsec_setspidx_inpcb(struct mbuf *m, struct inpcb *inp)
495 {
496 	int error;
497 
498 	IPSEC_ASSERT(inp != NULL, ("null inp"));
499 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
500 	IPSEC_ASSERT(inp->inp_sp->sp_out != NULL && inp->inp_sp->sp_in != NULL,
501 		("null sp_in || sp_out"));
502 
503 	error = ipsec_setspidx(m, &inp->inp_sp->sp_in->spidx, 1);
504 	if (error == 0) {
505 		inp->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
506 		inp->inp_sp->sp_out->spidx = inp->inp_sp->sp_in->spidx;
507 		inp->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
508 	} else {
509 		bzero(&inp->inp_sp->sp_in->spidx,
510 			sizeof (inp->inp_sp->sp_in->spidx));
511 		bzero(&inp->inp_sp->sp_out->spidx,
512 			sizeof (inp->inp_sp->sp_in->spidx));
513 	}
514 	return (error);
515 }
516 
517 /*
518  * Configure security policy index (src/dst/proto/sport/dport)
519  * by looking at the content of mbuf.
520  * The caller is responsible for error recovery (like clearing up spidx).
521  */
522 static int
523 ipsec_setspidx(struct mbuf *m, struct secpolicyindex *spidx, int needport)
524 {
525 	struct ip *ip = NULL;
526 	struct ip ipbuf;
527 	u_int v;
528 	struct mbuf *n;
529 	int len;
530 	int error;
531 
532 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
533 
534 	/*
535 	 * Validate m->m_pkthdr.len.  We see incorrect length if we
536 	 * mistakenly call this function with inconsistent mbuf chain
537 	 * (like 4.4BSD tcp/udp processing).  XXX Should we panic here?
538 	 */
539 	len = 0;
540 	for (n = m; n; n = n->m_next)
541 		len += n->m_len;
542 	if (m->m_pkthdr.len != len) {
543 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
544 			printf("%s: pkthdr len(%d) mismatch (%d), ignored.\n",
545 				__func__, len, m->m_pkthdr.len));
546 		return (EINVAL);
547 	}
548 
549 	if (m->m_pkthdr.len < sizeof(struct ip)) {
550 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
551 			printf("%s: pkthdr len(%d) too small (v4), ignored.\n",
552 			    __func__, m->m_pkthdr.len));
553 		return (EINVAL);
554 	}
555 
556 	if (m->m_len >= sizeof(*ip))
557 		ip = mtod(m, struct ip *);
558 	else {
559 		m_copydata(m, 0, sizeof(ipbuf), (caddr_t)&ipbuf);
560 		ip = &ipbuf;
561 	}
562 	v = ip->ip_v;
563 	switch (v) {
564 	case 4:
565 		error = ipsec4_setspidx_ipaddr(m, spidx);
566 		if (error)
567 			return (error);
568 		ipsec4_get_ulp(m, spidx, needport);
569 		return (0);
570 #ifdef INET6
571 	case 6:
572 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) {
573 			KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
574 				printf("%s: pkthdr len(%d) too small (v6), "
575 				"ignored\n", __func__, m->m_pkthdr.len));
576 			return (EINVAL);
577 		}
578 		error = ipsec6_setspidx_ipaddr(m, spidx);
579 		if (error)
580 			return (error);
581 		ipsec6_get_ulp(m, spidx, needport);
582 		return (0);
583 #endif
584 	default:
585 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
586 			printf("%s: " "unknown IP version %u, ignored.\n",
587 				__func__, v));
588 		return (EINVAL);
589 	}
590 }
591 
592 static void
593 ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
594 {
595 	u_int8_t nxt;
596 	int off;
597 
598 	/* Sanity check. */
599 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
600 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),("packet too short"));
601 
602 	if (m->m_len >= sizeof (struct ip)) {
603 		struct ip *ip = mtod(m, struct ip *);
604 		if (ip->ip_off & htons(IP_MF | IP_OFFMASK))
605 			goto done;
606 		off = ip->ip_hl << 2;
607 		nxt = ip->ip_p;
608 	} else {
609 		struct ip ih;
610 
611 		m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih);
612 		if (ih.ip_off & htons(IP_MF | IP_OFFMASK))
613 			goto done;
614 		off = ih.ip_hl << 2;
615 		nxt = ih.ip_p;
616 	}
617 
618 	while (off < m->m_pkthdr.len) {
619 		struct ip6_ext ip6e;
620 		struct tcphdr th;
621 		struct udphdr uh;
622 
623 		switch (nxt) {
624 		case IPPROTO_TCP:
625 			spidx->ul_proto = nxt;
626 			if (!needport)
627 				goto done_proto;
628 			if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
629 				goto done;
630 			m_copydata(m, off, sizeof (th), (caddr_t) &th);
631 			spidx->src.sin.sin_port = th.th_sport;
632 			spidx->dst.sin.sin_port = th.th_dport;
633 			return;
634 		case IPPROTO_UDP:
635 			spidx->ul_proto = nxt;
636 			if (!needport)
637 				goto done_proto;
638 			if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
639 				goto done;
640 			m_copydata(m, off, sizeof (uh), (caddr_t) &uh);
641 			spidx->src.sin.sin_port = uh.uh_sport;
642 			spidx->dst.sin.sin_port = uh.uh_dport;
643 			return;
644 		case IPPROTO_AH:
645 			if (off + sizeof(ip6e) > m->m_pkthdr.len)
646 				goto done;
647 			/* XXX Sigh, this works but is totally bogus. */
648 			m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e);
649 			off += (ip6e.ip6e_len + 2) << 2;
650 			nxt = ip6e.ip6e_nxt;
651 			break;
652 		case IPPROTO_ICMP:
653 		default:
654 			/* XXX Intermediate headers??? */
655 			spidx->ul_proto = nxt;
656 			goto done_proto;
657 		}
658 	}
659 done:
660 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
661 done_proto:
662 	spidx->src.sin.sin_port = IPSEC_PORT_ANY;
663 	spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
664 }
665 
666 /* Assumes that m is sane. */
667 static int
668 ipsec4_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
669 {
670 	static const struct sockaddr_in template = {
671 		sizeof (struct sockaddr_in),
672 		AF_INET,
673 		0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 }
674 	};
675 
676 	spidx->src.sin = template;
677 	spidx->dst.sin = template;
678 
679 	if (m->m_len < sizeof (struct ip)) {
680 		m_copydata(m, offsetof(struct ip, ip_src),
681 			   sizeof (struct  in_addr),
682 			   (caddr_t) &spidx->src.sin.sin_addr);
683 		m_copydata(m, offsetof(struct ip, ip_dst),
684 			   sizeof (struct  in_addr),
685 			   (caddr_t) &spidx->dst.sin.sin_addr);
686 	} else {
687 		struct ip *ip = mtod(m, struct ip *);
688 		spidx->src.sin.sin_addr = ip->ip_src;
689 		spidx->dst.sin.sin_addr = ip->ip_dst;
690 	}
691 
692 	spidx->prefs = sizeof(struct in_addr) << 3;
693 	spidx->prefd = sizeof(struct in_addr) << 3;
694 
695 	return (0);
696 }
697 
698 #ifdef INET6
699 static void
700 ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
701 {
702 	int off, nxt;
703 	struct tcphdr th;
704 	struct udphdr uh;
705 	struct icmp6_hdr ih;
706 
707 	/* Sanity check. */
708 	if (m == NULL)
709 		panic("%s: NULL pointer was passed.\n", __func__);
710 
711 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
712 		printf("%s:\n", __func__); kdebug_mbuf(m));
713 
714 	/* Set default. */
715 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
716 	((struct sockaddr_in6 *)&spidx->src)->sin6_port = IPSEC_PORT_ANY;
717 	((struct sockaddr_in6 *)&spidx->dst)->sin6_port = IPSEC_PORT_ANY;
718 
719 	nxt = -1;
720 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
721 	if (off < 0 || m->m_pkthdr.len < off)
722 		return;
723 
724 	switch (nxt) {
725 	case IPPROTO_TCP:
726 		spidx->ul_proto = nxt;
727 		if (!needport)
728 			break;
729 		if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
730 			break;
731 		m_copydata(m, off, sizeof(th), (caddr_t)&th);
732 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = th.th_sport;
733 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = th.th_dport;
734 		break;
735 	case IPPROTO_UDP:
736 		spidx->ul_proto = nxt;
737 		if (!needport)
738 			break;
739 		if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
740 			break;
741 		m_copydata(m, off, sizeof(uh), (caddr_t)&uh);
742 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = uh.uh_sport;
743 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = uh.uh_dport;
744 		break;
745 	case IPPROTO_ICMPV6:
746 		spidx->ul_proto = nxt;
747 		if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len)
748 			break;
749 		m_copydata(m, off, sizeof(ih), (caddr_t)&ih);
750 		((struct sockaddr_in6 *)&spidx->src)->sin6_port =
751 		    htons((uint16_t)ih.icmp6_type);
752 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port =
753 		    htons((uint16_t)ih.icmp6_code);
754 		break;
755 	default:
756 		/* XXX Intermediate headers??? */
757 		spidx->ul_proto = nxt;
758 		break;
759 	}
760 }
761 
762 /* Assumes that m is sane. */
763 static int
764 ipsec6_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
765 {
766 	struct ip6_hdr *ip6 = NULL;
767 	struct ip6_hdr ip6buf;
768 	struct sockaddr_in6 *sin6;
769 
770 	if (m->m_len >= sizeof(*ip6))
771 		ip6 = mtod(m, struct ip6_hdr *);
772 	else {
773 		m_copydata(m, 0, sizeof(ip6buf), (caddr_t)&ip6buf);
774 		ip6 = &ip6buf;
775 	}
776 
777 	sin6 = (struct sockaddr_in6 *)&spidx->src;
778 	bzero(sin6, sizeof(*sin6));
779 	sin6->sin6_family = AF_INET6;
780 	sin6->sin6_len = sizeof(struct sockaddr_in6);
781 	bcopy(&ip6->ip6_src, &sin6->sin6_addr, sizeof(ip6->ip6_src));
782 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
783 		sin6->sin6_addr.s6_addr16[1] = 0;
784 		sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]);
785 	}
786 	spidx->prefs = sizeof(struct in6_addr) << 3;
787 
788 	sin6 = (struct sockaddr_in6 *)&spidx->dst;
789 	bzero(sin6, sizeof(*sin6));
790 	sin6->sin6_family = AF_INET6;
791 	sin6->sin6_len = sizeof(struct sockaddr_in6);
792 	bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(ip6->ip6_dst));
793 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
794 		sin6->sin6_addr.s6_addr16[1] = 0;
795 		sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]);
796 	}
797 	spidx->prefd = sizeof(struct in6_addr) << 3;
798 
799 	return (0);
800 }
801 #endif
802 
803 static void
804 ipsec_delpcbpolicy(struct inpcbpolicy *p)
805 {
806 
807 	free(p, M_IPSEC_INPCB);
808 }
809 
810 /* Initialize policy in PCB. */
811 int
812 ipsec_init_policy(struct socket *so, struct inpcbpolicy **pcb_sp)
813 {
814 	struct inpcbpolicy *new;
815 
816 	/* Sanity check. */
817 	if (so == NULL || pcb_sp == NULL)
818 		panic("%s: NULL pointer was passed.\n", __func__);
819 
820 	new = (struct inpcbpolicy *) malloc(sizeof(struct inpcbpolicy),
821 					    M_IPSEC_INPCB, M_NOWAIT|M_ZERO);
822 	if (new == NULL) {
823 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
824 		return (ENOBUFS);
825 	}
826 
827 	new->priv = IPSEC_IS_PRIVILEGED_SO(so);
828 
829 	if ((new->sp_in = KEY_NEWSP()) == NULL) {
830 		ipsec_delpcbpolicy(new);
831 		return (ENOBUFS);
832 	}
833 	new->sp_in->state = IPSEC_SPSTATE_ALIVE;
834 	new->sp_in->policy = IPSEC_POLICY_ENTRUST;
835 
836 	if ((new->sp_out = KEY_NEWSP()) == NULL) {
837 		KEY_FREESP(&new->sp_in);
838 		ipsec_delpcbpolicy(new);
839 		return (ENOBUFS);
840 	}
841 	new->sp_out->state = IPSEC_SPSTATE_ALIVE;
842 	new->sp_out->policy = IPSEC_POLICY_ENTRUST;
843 
844 	*pcb_sp = new;
845 
846 	return (0);
847 }
848 
849 /* Copy old IPsec policy into new. */
850 int
851 ipsec_copy_policy(struct inpcbpolicy *old, struct inpcbpolicy *new)
852 {
853 	struct secpolicy *sp;
854 
855 	sp = ipsec_deepcopy_policy(old->sp_in);
856 	if (sp) {
857 		KEY_FREESP(&new->sp_in);
858 		new->sp_in = sp;
859 	} else
860 		return (ENOBUFS);
861 
862 	sp = ipsec_deepcopy_policy(old->sp_out);
863 	if (sp) {
864 		KEY_FREESP(&new->sp_out);
865 		new->sp_out = sp;
866 	} else
867 		return (ENOBUFS);
868 
869 	new->priv = old->priv;
870 
871 	return (0);
872 }
873 
874 struct ipsecrequest *
875 ipsec_newisr(void)
876 {
877 	struct ipsecrequest *p;
878 
879 	p = malloc(sizeof(struct ipsecrequest), M_IPSEC_SR, M_NOWAIT|M_ZERO);
880 	if (p != NULL)
881 		IPSECREQUEST_LOCK_INIT(p);
882 	return (p);
883 }
884 
885 void
886 ipsec_delisr(struct ipsecrequest *p)
887 {
888 
889 	IPSECREQUEST_LOCK_DESTROY(p);
890 	free(p, M_IPSEC_SR);
891 }
892 
893 /* Deep-copy a policy in PCB. */
894 static struct secpolicy *
895 ipsec_deepcopy_policy(struct secpolicy *src)
896 {
897 	struct ipsecrequest *newchain = NULL;
898 	struct ipsecrequest *p;
899 	struct ipsecrequest **q;
900 	struct ipsecrequest *r;
901 	struct secpolicy *dst;
902 
903 	if (src == NULL)
904 		return (NULL);
905 	dst = KEY_NEWSP();
906 	if (dst == NULL)
907 		return (NULL);
908 
909 	/*
910 	 * Deep-copy IPsec request chain.  This is required since struct
911 	 * ipsecrequest is not reference counted.
912 	 */
913 	q = &newchain;
914 	for (p = src->req; p; p = p->next) {
915 		*q = ipsec_newisr();
916 		if (*q == NULL)
917 			goto fail;
918 		(*q)->saidx.proto = p->saidx.proto;
919 		(*q)->saidx.mode = p->saidx.mode;
920 		(*q)->level = p->level;
921 		(*q)->saidx.reqid = p->saidx.reqid;
922 
923 		bcopy(&p->saidx.src, &(*q)->saidx.src, sizeof((*q)->saidx.src));
924 		bcopy(&p->saidx.dst, &(*q)->saidx.dst, sizeof((*q)->saidx.dst));
925 
926 		(*q)->sp = dst;
927 
928 		q = &((*q)->next);
929 	}
930 
931 	dst->req = newchain;
932 	dst->state = src->state;
933 	dst->policy = src->policy;
934 	/* Do not touch the refcnt fields. */
935 
936 	return (dst);
937 
938 fail:
939 	for (p = newchain; p; p = r) {
940 		r = p->next;
941 		ipsec_delisr(p);
942 		p = NULL;
943 	}
944 	return (NULL);
945 }
946 
947 /* Set policy and IPsec request if present. */
948 static int
949 ipsec_set_policy_internal(struct secpolicy **pcb_sp, int optname,
950     caddr_t request, size_t len, struct ucred *cred)
951 {
952 	struct sadb_x_policy *xpl;
953 	struct secpolicy *newsp = NULL;
954 	int error;
955 
956 	/* Sanity check. */
957 	if (pcb_sp == NULL || *pcb_sp == NULL || request == NULL)
958 		return (EINVAL);
959 	if (len < sizeof(*xpl))
960 		return (EINVAL);
961 	xpl = (struct sadb_x_policy *)request;
962 
963 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
964 		printf("%s: passed policy\n", __func__);
965 		kdebug_sadb_x_policy((struct sadb_ext *)xpl));
966 
967 	/* Check policy type. */
968 	/* ipsec_set_policy_internal() accepts IPSEC, ENTRUST and BYPASS. */
969 	if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD
970 	 || xpl->sadb_x_policy_type == IPSEC_POLICY_NONE)
971 		return (EINVAL);
972 
973 	/* Check privileged socket. */
974 	if (cred != NULL && xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
975 		error = priv_check_cred(cred, PRIV_NETINET_IPSEC, 0);
976 		if (error)
977 			return (EACCES);
978 	}
979 
980 	/* Allocating new SP entry. */
981 	if ((newsp = key_msg2sp(xpl, len, &error)) == NULL)
982 		return (error);
983 
984 	newsp->state = IPSEC_SPSTATE_ALIVE;
985 
986 	/* Clear old SP and set new SP. */
987 	KEY_FREESP(pcb_sp);
988 	*pcb_sp = newsp;
989 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
990 		printf("%s: new policy\n", __func__);
991 		kdebug_secpolicy(newsp));
992 
993 	return (0);
994 }
995 
996 int
997 ipsec_set_policy(struct inpcb *inp, int optname, caddr_t request,
998     size_t len, struct ucred *cred)
999 {
1000 	struct sadb_x_policy *xpl;
1001 	struct secpolicy **pcb_sp;
1002 
1003 	/* Sanity check. */
1004 	if (inp == NULL || request == NULL)
1005 		return (EINVAL);
1006 	if (len < sizeof(*xpl))
1007 		return (EINVAL);
1008 	xpl = (struct sadb_x_policy *)request;
1009 
1010 	/* Select direction. */
1011 	switch (xpl->sadb_x_policy_dir) {
1012 	case IPSEC_DIR_INBOUND:
1013 		pcb_sp = &inp->inp_sp->sp_in;
1014 		break;
1015 	case IPSEC_DIR_OUTBOUND:
1016 		pcb_sp = &inp->inp_sp->sp_out;
1017 		break;
1018 	default:
1019 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1020 			xpl->sadb_x_policy_dir));
1021 		return (EINVAL);
1022 	}
1023 
1024 	return (ipsec_set_policy_internal(pcb_sp, optname, request, len, cred));
1025 }
1026 
1027 int
1028 ipsec_get_policy(struct inpcb *inp, caddr_t request, size_t len,
1029     struct mbuf **mp)
1030 {
1031 	struct sadb_x_policy *xpl;
1032 	struct secpolicy *pcb_sp;
1033 
1034 	/* Sanity check. */
1035 	if (inp == NULL || request == NULL || mp == NULL)
1036 		return (EINVAL);
1037 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
1038 	if (len < sizeof(*xpl))
1039 		return (EINVAL);
1040 	xpl = (struct sadb_x_policy *)request;
1041 
1042 	/* Select direction. */
1043 	switch (xpl->sadb_x_policy_dir) {
1044 	case IPSEC_DIR_INBOUND:
1045 		pcb_sp = inp->inp_sp->sp_in;
1046 		break;
1047 	case IPSEC_DIR_OUTBOUND:
1048 		pcb_sp = inp->inp_sp->sp_out;
1049 		break;
1050 	default:
1051 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1052 			xpl->sadb_x_policy_dir));
1053 		return (EINVAL);
1054 	}
1055 
1056 	/* Sanity check. Should be an IPSEC_ASSERT. */
1057 	if (pcb_sp == NULL)
1058 		return (EINVAL);
1059 
1060 	*mp = key_sp2msg(pcb_sp);
1061 	if (!*mp) {
1062 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
1063 		return (ENOBUFS);
1064 	}
1065 
1066 	(*mp)->m_type = MT_DATA;
1067 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1068 		printf("%s:\n", __func__); kdebug_mbuf(*mp));
1069 
1070 	return (0);
1071 }
1072 
1073 /* Delete policy in PCB. */
1074 int
1075 ipsec_delete_pcbpolicy(struct inpcb *inp)
1076 {
1077 	IPSEC_ASSERT(inp != NULL, ("null inp"));
1078 
1079 	if (inp->inp_sp == NULL)
1080 		return (0);
1081 
1082 	if (inp->inp_sp->sp_in != NULL)
1083 		KEY_FREESP(&inp->inp_sp->sp_in);
1084 
1085 	if (inp->inp_sp->sp_out != NULL)
1086 		KEY_FREESP(&inp->inp_sp->sp_out);
1087 
1088 	ipsec_delpcbpolicy(inp->inp_sp);
1089 	inp->inp_sp = NULL;
1090 
1091 	return (0);
1092 }
1093 
1094 /*
1095  * Return current level.
1096  * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
1097  */
1098 u_int
1099 ipsec_get_reqlevel(struct ipsecrequest *isr)
1100 {
1101 	u_int level = 0;
1102 	u_int esp_trans_deflev, esp_net_deflev;
1103 	u_int ah_trans_deflev, ah_net_deflev;
1104 
1105 	IPSEC_ASSERT(isr != NULL && isr->sp != NULL, ("null argument"));
1106 	IPSEC_ASSERT(isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family,
1107 		("af family mismatch, src %u, dst %u",
1108 		 isr->sp->spidx.src.sa.sa_family,
1109 		 isr->sp->spidx.dst.sa.sa_family));
1110 
1111 /* XXX Note that we have ipseclog() expanded here - code sync issue. */
1112 #define IPSEC_CHECK_DEFAULT(lev) \
1113 	(((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE	      \
1114 			&& (lev) != IPSEC_LEVEL_UNIQUE)			      \
1115 		? (V_ipsec_debug						      \
1116 			? log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
1117 				(lev), IPSEC_LEVEL_REQUIRE)		      \
1118 			: 0),						      \
1119 			(lev) = IPSEC_LEVEL_REQUIRE,			      \
1120 			(lev)						      \
1121 		: (lev))
1122 
1123 	/* Set default level. */
1124 	switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) {
1125 #ifdef INET
1126 	case AF_INET:
1127 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev);
1128 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev);
1129 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev);
1130 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev);
1131 		break;
1132 #endif
1133 #ifdef INET6
1134 	case AF_INET6:
1135 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev);
1136 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev);
1137 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev);
1138 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev);
1139 		break;
1140 #endif /* INET6 */
1141 	default:
1142 		panic("%s: unknown af %u",
1143 			__func__, isr->sp->spidx.src.sa.sa_family);
1144 	}
1145 
1146 #undef IPSEC_CHECK_DEFAULT
1147 
1148 	/* Set level. */
1149 	switch (isr->level) {
1150 	case IPSEC_LEVEL_DEFAULT:
1151 		switch (isr->saidx.proto) {
1152 		case IPPROTO_ESP:
1153 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1154 				level = esp_net_deflev;
1155 			else
1156 				level = esp_trans_deflev;
1157 			break;
1158 		case IPPROTO_AH:
1159 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1160 				level = ah_net_deflev;
1161 			else
1162 				level = ah_trans_deflev;
1163 			break;
1164 		case IPPROTO_IPCOMP:
1165 			/*
1166 			 * We don't really care, as IPcomp document says that
1167 			 * we shouldn't compress small packets.
1168 			 */
1169 			level = IPSEC_LEVEL_USE;
1170 			break;
1171 		default:
1172 			panic("%s: Illegal protocol defined %u\n", __func__,
1173 				isr->saidx.proto);
1174 		}
1175 		break;
1176 
1177 	case IPSEC_LEVEL_USE:
1178 	case IPSEC_LEVEL_REQUIRE:
1179 		level = isr->level;
1180 		break;
1181 	case IPSEC_LEVEL_UNIQUE:
1182 		level = IPSEC_LEVEL_REQUIRE;
1183 		break;
1184 
1185 	default:
1186 		panic("%s: Illegal IPsec level %u\n", __func__, isr->level);
1187 	}
1188 
1189 	return (level);
1190 }
1191 
1192 /*
1193  * Check security policy requirements against the actual
1194  * packet contents.  Return one if the packet should be
1195  * reject as "invalid"; otherwiser return zero to have the
1196  * packet treated as "valid".
1197  *
1198  * OUT:
1199  *	0: valid
1200  *	1: invalid
1201  */
1202 int
1203 ipsec_in_reject(struct secpolicy *sp, struct mbuf *m)
1204 {
1205 	struct ipsecrequest *isr;
1206 	int need_auth;
1207 
1208 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1209 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1210 
1211 	/* Check policy. */
1212 	switch (sp->policy) {
1213 	case IPSEC_POLICY_DISCARD:
1214 		return (1);
1215 	case IPSEC_POLICY_BYPASS:
1216 	case IPSEC_POLICY_NONE:
1217 		return (0);
1218 	}
1219 
1220 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1221 		("invalid policy %u", sp->policy));
1222 
1223 	/* XXX Should compare policy against IPsec header history. */
1224 
1225 	need_auth = 0;
1226 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1227 		if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE)
1228 			continue;
1229 		switch (isr->saidx.proto) {
1230 		case IPPROTO_ESP:
1231 			if ((m->m_flags & M_DECRYPTED) == 0) {
1232 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1233 				    printf("%s: ESP m_flags:%x\n", __func__,
1234 					    m->m_flags));
1235 				return (1);
1236 			}
1237 
1238 			if (!need_auth &&
1239 			    isr->sav != NULL &&
1240 			    isr->sav->tdb_authalgxform != NULL &&
1241 			    (m->m_flags & M_AUTHIPDGM) == 0) {
1242 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1243 				    printf("%s: ESP/AH m_flags:%x\n", __func__,
1244 					    m->m_flags));
1245 				return (1);
1246 			}
1247 			break;
1248 		case IPPROTO_AH:
1249 			need_auth = 1;
1250 			if ((m->m_flags & M_AUTHIPHDR) == 0) {
1251 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1252 				    printf("%s: AH m_flags:%x\n", __func__,
1253 					    m->m_flags));
1254 				return (1);
1255 			}
1256 			break;
1257 		case IPPROTO_IPCOMP:
1258 			/*
1259 			 * We don't really care, as IPcomp document
1260 			 * says that we shouldn't compress small
1261 			 * packets.  IPComp policy should always be
1262 			 * treated as being in "use" level.
1263 			 */
1264 			break;
1265 		}
1266 	}
1267 	return (0);		/* Valid. */
1268 }
1269 
1270 static int
1271 ipsec46_in_reject(struct mbuf *m, struct inpcb *inp)
1272 {
1273 	struct secpolicy *sp;
1274 	int error;
1275 	int result;
1276 
1277 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1278 
1279 	/*
1280 	 * Get SP for this packet.
1281 	 * When we are called from ip_forward(), we call
1282 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1283 	 */
1284 	if (inp == NULL)
1285 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, IP_FORWARDING, &error);
1286 	else
1287 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND, inp, &error);
1288 
1289 	if (sp != NULL) {
1290 		result = ipsec_in_reject(sp, m);
1291 		KEY_FREESP(&sp);
1292 	} else {
1293 		result = 0;	/* XXX Should be panic?
1294 				 * -> No, there may be error. */
1295 	}
1296 	return (result);
1297 }
1298 
1299 /*
1300  * Check AH/ESP integrity.
1301  * This function is called from tcp_input(), udp_input(),
1302  * and {ah,esp}4_input for tunnel mode.
1303  */
1304 int
1305 ipsec4_in_reject(struct mbuf *m, struct inpcb *inp)
1306 {
1307 	int result;
1308 
1309 	result = ipsec46_in_reject(m, inp);
1310 	if (result)
1311 		IPSECSTAT_INC(ips_in_polvio);
1312 
1313 	return (result);
1314 }
1315 
1316 #ifdef INET6
1317 /*
1318  * Check AH/ESP integrity.
1319  * This function is called from tcp6_input(), udp6_input(),
1320  * and {ah,esp}6_input for tunnel mode.
1321  */
1322 int
1323 ipsec6_in_reject(struct mbuf *m, struct inpcb *inp)
1324 {
1325 	int result;
1326 
1327 	result = ipsec46_in_reject(m, inp);
1328 	if (result)
1329 		IPSEC6STAT_INC(ips_in_polvio);
1330 
1331 	return (result);
1332 }
1333 #endif
1334 
1335 /*
1336  * Compute the byte size to be occupied by IPsec header.
1337  * In case it is tunnelled, it includes the size of outer IP header.
1338  * NOTE: SP passed is freed in this function.
1339  */
1340 static size_t
1341 ipsec_hdrsiz_internal(struct secpolicy *sp)
1342 {
1343 	struct ipsecrequest *isr;
1344 	size_t size;
1345 
1346 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1347 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1348 
1349 	switch (sp->policy) {
1350 	case IPSEC_POLICY_DISCARD:
1351 	case IPSEC_POLICY_BYPASS:
1352 	case IPSEC_POLICY_NONE:
1353 		return (0);
1354 	}
1355 
1356 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1357 		("invalid policy %u", sp->policy));
1358 
1359 	size = 0;
1360 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1361 		size_t clen = 0;
1362 
1363 		switch (isr->saidx.proto) {
1364 		case IPPROTO_ESP:
1365 			clen = esp_hdrsiz(isr->sav);
1366 			break;
1367 		case IPPROTO_AH:
1368 			clen = ah_hdrsiz(isr->sav);
1369 			break;
1370 		case IPPROTO_IPCOMP:
1371 			clen = sizeof(struct ipcomp);
1372 			break;
1373 		}
1374 
1375 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1376 			switch (isr->saidx.dst.sa.sa_family) {
1377 			case AF_INET:
1378 				clen += sizeof(struct ip);
1379 				break;
1380 #ifdef INET6
1381 			case AF_INET6:
1382 				clen += sizeof(struct ip6_hdr);
1383 				break;
1384 #endif
1385 			default:
1386 				ipseclog((LOG_ERR, "%s: unknown AF %d in "
1387 				    "IPsec tunnel SA\n", __func__,
1388 				    ((struct sockaddr *)&isr->saidx.dst)->sa_family));
1389 				break;
1390 			}
1391 		}
1392 		size += clen;
1393 	}
1394 
1395 	return (size);
1396 }
1397 
1398 /*
1399  * This function is called from ipsec_hdrsiz_tcp(), ip_ipsec_mtu(),
1400  * disabled ip6_ipsec_mtu() and ip6_forward().
1401  */
1402 size_t
1403 ipsec_hdrsiz(struct mbuf *m, u_int dir, struct inpcb *inp)
1404 {
1405 	struct secpolicy *sp;
1406 	int error;
1407 	size_t size;
1408 
1409 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1410 
1411 	/* Get SP for this packet.
1412 	 * When we are called from ip_forward(), we call
1413 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1414 	 */
1415 	if (inp == NULL)
1416 		sp = ipsec_getpolicybyaddr(m, dir, IP_FORWARDING, &error);
1417 	else
1418 		sp = ipsec_getpolicybysock(m, dir, inp, &error);
1419 
1420 	if (sp != NULL) {
1421 		size = ipsec_hdrsiz_internal(sp);
1422 		KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1423 			printf("%s: size:%lu.\n", __func__,
1424 				(unsigned long)size));
1425 
1426 		KEY_FREESP(&sp);
1427 	} else {
1428 		size = 0;	/* XXX Should be panic?
1429 				 * -> No, we are called w/o knowing if
1430 				 *    IPsec processing is needed. */
1431 	}
1432 	return (size);
1433 }
1434 
1435 /*
1436  * Check the variable replay window.
1437  * ipsec_chkreplay() performs replay check before ICV verification.
1438  * ipsec_updatereplay() updates replay bitmap.  This must be called after
1439  * ICV verification (it also performs replay check, which is usually done
1440  * beforehand).
1441  * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1442  *
1443  * Based on RFC 2401.
1444  */
1445 int
1446 ipsec_chkreplay(u_int32_t seq, struct secasvar *sav)
1447 {
1448 	const struct secreplay *replay;
1449 	u_int32_t diff;
1450 	int fr;
1451 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1452 	int frlast;		/* Constant: last frame. */
1453 
1454 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1455 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1456 
1457 	replay = sav->replay;
1458 
1459 	if (replay->wsize == 0)
1460 		return (1);	/* No need to check replay. */
1461 
1462 	/* Constant. */
1463 	frlast = replay->wsize - 1;
1464 	wsizeb = replay->wsize << 3;
1465 
1466 	/* Sequence number of 0 is invalid. */
1467 	if (seq == 0)
1468 		return (0);
1469 
1470 	/* First time is always okay. */
1471 	if (replay->count == 0)
1472 		return (1);
1473 
1474 	if (seq > replay->lastseq) {
1475 		/* Larger sequences are okay. */
1476 		return (1);
1477 	} else {
1478 		/* seq is equal or less than lastseq. */
1479 		diff = replay->lastseq - seq;
1480 
1481 		/* Over range to check, i.e. too old or wrapped. */
1482 		if (diff >= wsizeb)
1483 			return (0);
1484 
1485 		fr = frlast - diff / 8;
1486 
1487 		/* This packet already seen? */
1488 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1489 			return (0);
1490 
1491 		/* Out of order but good. */
1492 		return (1);
1493 	}
1494 }
1495 
1496 /*
1497  * Check replay counter whether to update or not.
1498  * OUT:	0:	OK
1499  *	1:	NG
1500  */
1501 int
1502 ipsec_updatereplay(u_int32_t seq, struct secasvar *sav)
1503 {
1504 	struct secreplay *replay;
1505 	u_int32_t diff;
1506 	int fr;
1507 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1508 	int frlast;		/* Constant: last frame. */
1509 
1510 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1511 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1512 
1513 	replay = sav->replay;
1514 
1515 	if (replay->wsize == 0)
1516 		goto ok;	/* No need to check replay. */
1517 
1518 	/* Constant. */
1519 	frlast = replay->wsize - 1;
1520 	wsizeb = replay->wsize << 3;
1521 
1522 	/* Sequence number of 0 is invalid. */
1523 	if (seq == 0)
1524 		return (1);
1525 
1526 	/* First time. */
1527 	if (replay->count == 0) {
1528 		replay->lastseq = seq;
1529 		bzero(replay->bitmap, replay->wsize);
1530 		(replay->bitmap)[frlast] = 1;
1531 		goto ok;
1532 	}
1533 
1534 	if (seq > replay->lastseq) {
1535 		/* seq is larger than lastseq. */
1536 		diff = seq - replay->lastseq;
1537 
1538 		/* New larger sequence number. */
1539 		if (diff < wsizeb) {
1540 			/* In window. */
1541 			/* Set bit for this packet. */
1542 			vshiftl(replay->bitmap, diff, replay->wsize);
1543 			(replay->bitmap)[frlast] |= 1;
1544 		} else {
1545 			/* This packet has a "way larger". */
1546 			bzero(replay->bitmap, replay->wsize);
1547 			(replay->bitmap)[frlast] = 1;
1548 		}
1549 		replay->lastseq = seq;
1550 
1551 		/* Larger is good. */
1552 	} else {
1553 		/* seq is equal or less than lastseq. */
1554 		diff = replay->lastseq - seq;
1555 
1556 		/* Over range to check, i.e. too old or wrapped. */
1557 		if (diff >= wsizeb)
1558 			return (1);
1559 
1560 		fr = frlast - diff / 8;
1561 
1562 		/* This packet already seen? */
1563 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1564 			return (1);
1565 
1566 		/* Mark as seen. */
1567 		(replay->bitmap)[fr] |= (1 << (diff % 8));
1568 
1569 		/* Out of order but good. */
1570 	}
1571 
1572 ok:
1573 	if (replay->count == ~0) {
1574 
1575 		/* Set overflow flag. */
1576 		replay->overflow++;
1577 
1578 		/* Don't increment, no more packets accepted. */
1579 		if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0)
1580 			return (1);
1581 
1582 		ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n",
1583 		    __func__, replay->overflow, ipsec_logsastr(sav)));
1584 	}
1585 
1586 	replay->count++;
1587 
1588 	return (0);
1589 }
1590 
1591 /*
1592  * Shift variable length buffer to left.
1593  * IN:	bitmap: pointer to the buffer
1594  * 	nbit:	the number of to shift.
1595  *	wsize:	buffer size (bytes).
1596  */
1597 static void
1598 vshiftl(unsigned char *bitmap, int nbit, int wsize)
1599 {
1600 	int s, j, i;
1601 	unsigned char over;
1602 
1603 	for (j = 0; j < nbit; j += 8) {
1604 		s = (nbit - j < 8) ? (nbit - j): 8;
1605 		bitmap[0] <<= s;
1606 		for (i = 1; i < wsize; i++) {
1607 			over = (bitmap[i] >> (8 - s));
1608 			bitmap[i] <<= s;
1609 			bitmap[i-1] |= over;
1610 		}
1611 	}
1612 }
1613 
1614 #ifdef INET
1615 /* Return a printable string for the IPv4 address. */
1616 static char *
1617 inet_ntoa4(struct in_addr ina)
1618 {
1619 	static char buf[4][4 * sizeof "123" + 4];
1620 	unsigned char *ucp = (unsigned char *) &ina;
1621 	static int i = 3;
1622 
1623 	/* XXX-BZ Returns static buffer. */
1624 	i = (i + 1) % 4;
1625 	sprintf(buf[i], "%d.%d.%d.%d", ucp[0] & 0xff, ucp[1] & 0xff,
1626 	    ucp[2] & 0xff, ucp[3] & 0xff);
1627 	return (buf[i]);
1628 }
1629 #endif
1630 
1631 /* Return a printable string for the address. */
1632 char *
1633 ipsec_address(union sockaddr_union* sa)
1634 {
1635 #ifdef INET6
1636 	char ip6buf[INET6_ADDRSTRLEN];
1637 #endif
1638 
1639 	switch (sa->sa.sa_family) {
1640 #ifdef INET
1641 	case AF_INET:
1642 		return (inet_ntoa4(sa->sin.sin_addr));
1643 #endif /* INET */
1644 #ifdef INET6
1645 	case AF_INET6:
1646 		return (ip6_sprintf(ip6buf, &sa->sin6.sin6_addr));
1647 #endif /* INET6 */
1648 	default:
1649 		return ("(unknown address family)");
1650 	}
1651 }
1652 
1653 const char *
1654 ipsec_logsastr(struct secasvar *sav)
1655 {
1656 	static char buf[256];
1657 	char *p;
1658 	struct secasindex *saidx = &sav->sah->saidx;
1659 
1660 	IPSEC_ASSERT(saidx->src.sa.sa_family == saidx->dst.sa.sa_family,
1661 		("address family mismatch"));
1662 
1663 	p = buf;
1664 	snprintf(buf, sizeof(buf), "SA(SPI=%u ", (u_int32_t)ntohl(sav->spi));
1665 	while (p && *p)
1666 		p++;
1667 	/* NB: only use ipsec_address on one address at a time. */
1668 	snprintf(p, sizeof (buf) - (p - buf), "src=%s ",
1669 		ipsec_address(&saidx->src));
1670 	while (p && *p)
1671 		p++;
1672 	snprintf(p, sizeof (buf) - (p - buf), "dst=%s)",
1673 		ipsec_address(&saidx->dst));
1674 
1675 	return (buf);
1676 }
1677 
1678 void
1679 ipsec_dumpmbuf(struct mbuf *m)
1680 {
1681 	int totlen;
1682 	int i;
1683 	u_char *p;
1684 
1685 	totlen = 0;
1686 	printf("---\n");
1687 	while (m) {
1688 		p = mtod(m, u_char *);
1689 		for (i = 0; i < m->m_len; i++) {
1690 			printf("%02x ", p[i]);
1691 			totlen++;
1692 			if (totlen % 16 == 0)
1693 				printf("\n");
1694 		}
1695 		m = m->m_next;
1696 	}
1697 	if (totlen % 16 != 0)
1698 		printf("\n");
1699 	printf("---\n");
1700 }
1701 
1702 static void
1703 ipsec_init(const void *unused __unused)
1704 {
1705 
1706 	SECPOLICY_LOCK_INIT(&V_ip4_def_policy);
1707 	V_ip4_def_policy.refcnt = 1;			/* NB: disallow free. */
1708 }
1709 VNET_SYSINIT(ipsec_init, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, ipsec_init,
1710     NULL);
1711 
1712 
1713 /* XXX This stuff doesn't belong here... */
1714 
1715 static	struct xformsw* xforms = NULL;
1716 
1717 /*
1718  * Register a transform; typically at system startup.
1719  */
1720 void
1721 xform_register(struct xformsw* xsp)
1722 {
1723 
1724 	xsp->xf_next = xforms;
1725 	xforms = xsp;
1726 }
1727 
1728 /*
1729  * Initialize transform support in an sav.
1730  */
1731 int
1732 xform_init(struct secasvar *sav, int xftype)
1733 {
1734 	struct xformsw *xsp;
1735 
1736 	if (sav->tdb_xform != NULL)	/* Previously initialized. */
1737 		return (0);
1738 	for (xsp = xforms; xsp; xsp = xsp->xf_next)
1739 		if (xsp->xf_type == xftype)
1740 			return ((*xsp->xf_init)(sav, xsp));
1741 	return (EINVAL);
1742 }
1743