xref: /freebsd/sys/netipsec/ipsec.c (revision 09d325677d53a12c79a43664ff29871e92247629)
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/route.h>
60 #include <net/vnet.h>
61 
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/ip.h>
65 #include <netinet/ip_var.h>
66 #include <netinet/in_var.h>
67 #include <netinet/udp.h>
68 #include <netinet/udp_var.h>
69 #include <netinet/tcp.h>
70 #include <netinet/udp.h>
71 
72 #include <netinet/ip6.h>
73 #ifdef INET6
74 #include <netinet6/ip6_var.h>
75 #endif
76 #include <netinet/in_pcb.h>
77 #ifdef INET6
78 #include <netinet/icmp6.h>
79 #endif
80 
81 #include <sys/types.h>
82 #include <netipsec/ipsec.h>
83 #ifdef INET6
84 #include <netipsec/ipsec6.h>
85 #endif
86 #include <netipsec/ah_var.h>
87 #include <netipsec/esp_var.h>
88 #include <netipsec/ipcomp.h>		/*XXX*/
89 #include <netipsec/ipcomp_var.h>
90 
91 #include <netipsec/key.h>
92 #include <netipsec/keydb.h>
93 #include <netipsec/key_debug.h>
94 
95 #include <netipsec/xform.h>
96 
97 #include <machine/in_cksum.h>
98 
99 #include <opencrypto/cryptodev.h>
100 
101 #ifdef IPSEC_DEBUG
102 VNET_DEFINE(int, ipsec_debug) = 1;
103 #else
104 VNET_DEFINE(int, ipsec_debug) = 0;
105 #endif
106 
107 /* NB: name changed so netstat doesn't use it. */
108 VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec4stat);
109 VNET_PCPUSTAT_SYSINIT(ipsec4stat);
110 
111 #ifdef VIMAGE
112 VNET_PCPUSTAT_SYSUNINIT(ipsec4stat);
113 #endif /* VIMAGE */
114 
115 VNET_DEFINE(int, ip4_ah_offsetmask) = 0;	/* maybe IP_DF? */
116 /* DF bit on encap. 0: clear 1: set 2: copy */
117 VNET_DEFINE(int, ip4_ipsec_dfbit) = 0;
118 VNET_DEFINE(int, ip4_esp_trans_deflev) = IPSEC_LEVEL_USE;
119 VNET_DEFINE(int, ip4_esp_net_deflev) = IPSEC_LEVEL_USE;
120 VNET_DEFINE(int, ip4_ah_trans_deflev) = IPSEC_LEVEL_USE;
121 VNET_DEFINE(int, ip4_ah_net_deflev) = IPSEC_LEVEL_USE;
122 VNET_DEFINE(struct secpolicy, ip4_def_policy);
123 /* ECN ignore(-1)/forbidden(0)/allowed(1) */
124 VNET_DEFINE(int, ip4_ipsec_ecn) = 0;
125 VNET_DEFINE(int, ip4_esp_randpad) = -1;
126 
127 /*
128  * Crypto support requirements:
129  *
130  *  1	require hardware support
131  * -1	require software support
132  *  0	take anything
133  */
134 VNET_DEFINE(int, crypto_support) = CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE;
135 
136 FEATURE(ipsec, "Internet Protocol Security (IPsec)");
137 #ifdef IPSEC_NAT_T
138 FEATURE(ipsec_natt, "UDP Encapsulation of IPsec ESP Packets ('NAT-T')");
139 #endif
140 
141 SYSCTL_DECL(_net_inet_ipsec);
142 
143 /* net.inet.ipsec */
144 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_POLICY, def_policy,
145 	CTLFLAG_RW, &VNET_NAME(ip4_def_policy).policy, 0,
146 	"IPsec default policy.");
147 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
148 	CTLFLAG_RW, &VNET_NAME(ip4_esp_trans_deflev), 0,
149 	"Default ESP transport mode level");
150 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
151 	CTLFLAG_RW, &VNET_NAME(ip4_esp_net_deflev), 0,
152 	"Default ESP tunnel mode level.");
153 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
154 	CTLFLAG_RW, &VNET_NAME(ip4_ah_trans_deflev), 0,
155 	"AH transfer mode default level.");
156 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
157 	CTLFLAG_RW, &VNET_NAME(ip4_ah_net_deflev), 0,
158 	"AH tunnel mode default level.");
159 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_AH_CLEARTOS, ah_cleartos,
160 	CTLFLAG_RW, &VNET_NAME(ah_cleartos), 0,
161 	"If set clear type-of-service field when doing AH computation.");
162 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_AH_OFFSETMASK, ah_offsetmask,
163 	CTLFLAG_RW, &VNET_NAME(ip4_ah_offsetmask), 0,
164 	"If not set clear offset field mask when doing AH computation.");
165 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DFBIT, dfbit,
166 	CTLFLAG_RW, &VNET_NAME(ip4_ipsec_dfbit), 0,
167 	"Do not fragment bit on encap.");
168 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_ECN, ecn,
169 	CTLFLAG_RW, &VNET_NAME(ip4_ipsec_ecn), 0,
170 	"Explicit Congestion Notification handling.");
171 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
172 	CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
173 	"Enable IPsec debugging output when set.");
174 SYSCTL_VNET_INT(_net_inet_ipsec, OID_AUTO, crypto_support,
175 	CTLFLAG_RW, &VNET_NAME(crypto_support), 0,
176 	"Crypto driver selection.");
177 SYSCTL_VNET_PCPUSTAT(_net_inet_ipsec, OID_AUTO, ipsecstats, struct ipsecstat,
178     ipsec4stat, "IPsec IPv4 statistics.");
179 
180 #ifdef REGRESSION
181 /*
182  * When set to 1, IPsec will send packets with the same sequence number.
183  * This allows to verify if the other side has proper replay attacks detection.
184  */
185 VNET_DEFINE(int, ipsec_replay) = 0;
186 SYSCTL_VNET_INT(_net_inet_ipsec, OID_AUTO, test_replay,
187 	CTLFLAG_RW, &VNET_NAME(ipsec_replay), 0,
188 	"Emulate replay attack");
189 /*
190  * When set 1, IPsec will send packets with corrupted HMAC.
191  * This allows to verify if the other side properly detects modified packets.
192  */
193 VNET_DEFINE(int, ipsec_integrity) = 0;
194 SYSCTL_VNET_INT(_net_inet_ipsec, OID_AUTO, test_integrity,
195 	CTLFLAG_RW, &VNET_NAME(ipsec_integrity), 0,
196 	"Emulate man-in-the-middle attack");
197 #endif
198 
199 #ifdef INET6
200 VNET_PCPUSTAT_DEFINE(struct ipsecstat, ipsec6stat);
201 VNET_PCPUSTAT_SYSINIT(ipsec6stat);
202 
203 #ifdef VIMAGE
204 VNET_PCPUSTAT_SYSUNINIT(ipsec6stat);
205 #endif /* VIMAGE */
206 
207 VNET_DEFINE(int, ip6_esp_trans_deflev) = IPSEC_LEVEL_USE;
208 VNET_DEFINE(int, ip6_esp_net_deflev) = IPSEC_LEVEL_USE;
209 VNET_DEFINE(int, ip6_ah_trans_deflev) = IPSEC_LEVEL_USE;
210 VNET_DEFINE(int, ip6_ah_net_deflev) = IPSEC_LEVEL_USE;
211 VNET_DEFINE(int, ip6_ipsec_ecn) = 0;	/* ECN ignore(-1)/forbidden(0)/allowed(1) */
212 
213 SYSCTL_DECL(_net_inet6_ipsec6);
214 
215 /* net.inet6.ipsec6 */
216 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_POLICY, def_policy, CTLFLAG_RW,
217 	&VNET_NAME(ip4_def_policy).policy, 0,
218 	"IPsec default policy.");
219 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV,
220 	esp_trans_deflev, CTLFLAG_RW, &VNET_NAME(ip6_esp_trans_deflev),	0,
221 	"Default ESP transport mode level.");
222 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV,
223 	esp_net_deflev, CTLFLAG_RW, &VNET_NAME(ip6_esp_net_deflev),	0,
224 	"Default ESP tunnel mode level.");
225 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV,
226 	ah_trans_deflev, CTLFLAG_RW, &VNET_NAME(ip6_ah_trans_deflev),	0,
227 	"AH transfer mode default level.");
228 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV,
229 	ah_net_deflev, CTLFLAG_RW, &VNET_NAME(ip6_ah_net_deflev),	0,
230 	"AH tunnel mode default level.");
231 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_ECN,
232 	ecn, CTLFLAG_RW, &VNET_NAME(ip6_ipsec_ecn),	0,
233 	"Explicit Congestion Notification handling.");
234 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug, CTLFLAG_RW,
235 	&VNET_NAME(ipsec_debug), 0,
236 	"Enable IPsec debugging output when set.");
237 SYSCTL_VNET_PCPUSTAT(_net_inet6_ipsec6, IPSECCTL_STATS, ipsecstats,
238     struct ipsecstat, ipsec6stat, "IPsec IPv6 statistics.");
239 #endif /* INET6 */
240 
241 static int ipsec_setspidx_inpcb __P((struct mbuf *, struct inpcb *));
242 static int ipsec_setspidx __P((struct mbuf *, struct secpolicyindex *, int));
243 static void ipsec4_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
244 static int ipsec4_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
245 #ifdef INET6
246 static void ipsec6_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
247 static int ipsec6_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
248 #endif
249 static void ipsec_delpcbpolicy __P((struct inpcbpolicy *));
250 static struct secpolicy *ipsec_deepcopy_policy __P((struct secpolicy *src));
251 static void vshiftl __P((unsigned char *, int, int));
252 
253 MALLOC_DEFINE(M_IPSEC_INPCB, "inpcbpolicy", "inpcb-resident ipsec policy");
254 
255 /*
256  * Return a held reference to the default SP.
257  */
258 static struct secpolicy *
259 key_allocsp_default(const char* where, int tag)
260 {
261 	struct secpolicy *sp;
262 
263 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
264 		printf("DP key_allocsp_default from %s:%u\n", where, tag));
265 
266 	sp = &V_ip4_def_policy;
267 	if (sp->policy != IPSEC_POLICY_DISCARD &&
268 	    sp->policy != IPSEC_POLICY_NONE) {
269 		ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n",
270 		    sp->policy, IPSEC_POLICY_NONE));
271 		sp->policy = IPSEC_POLICY_NONE;
272 	}
273 	key_addref(sp);
274 
275 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
276 		printf("DP key_allocsp_default returns SP:%p (%u)\n",
277 			sp, sp->refcnt));
278 	return (sp);
279 }
280 #define	KEY_ALLOCSP_DEFAULT() \
281 	key_allocsp_default(__FILE__, __LINE__)
282 
283 /*
284  * For OUTBOUND packet having a socket. Searching SPD for packet,
285  * and return a pointer to SP.
286  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
287  *		0	: bypass
288  *		EACCES	: discard packet.
289  *		ENOENT	: ipsec_acquire() in progress, maybe.
290  *		others	: error occured.
291  *	others:	a pointer to SP
292  *
293  * NOTE: IPv6 mapped adddress concern is implemented here.
294  */
295 struct secpolicy *
296 ipsec_getpolicy(struct tdb_ident *tdbi, u_int dir)
297 {
298 	struct secpolicy *sp;
299 
300 	IPSEC_ASSERT(tdbi != NULL, ("null tdbi"));
301 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
302 		("invalid direction %u", dir));
303 
304 	sp = KEY_ALLOCSP2(tdbi->spi, &tdbi->dst, tdbi->proto, dir);
305 	if (sp == NULL)			/*XXX????*/
306 		sp = KEY_ALLOCSP_DEFAULT();
307 	IPSEC_ASSERT(sp != NULL, ("null SP"));
308 	return (sp);
309 }
310 
311 /*
312  * For OUTBOUND packet having a socket. Searching SPD for packet,
313  * and return a pointer to SP.
314  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
315  *		0	: bypass
316  *		EACCES	: discard packet.
317  *		ENOENT	: ipsec_acquire() in progress, maybe.
318  *		others	: error occured.
319  *	others:	a pointer to SP
320  *
321  * NOTE: IPv6 mapped adddress concern is implemented here.
322  */
323 static struct secpolicy *
324 ipsec_getpolicybysock(struct mbuf *m, u_int dir, struct inpcb *inp, int *error)
325 {
326 	struct inpcbpolicy *pcbsp;
327 	struct secpolicy *currsp = NULL;	/* Policy on socket. */
328 	struct secpolicy *sp;
329 
330 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
331 	IPSEC_ASSERT(inp != NULL, ("null inpcb"));
332 	IPSEC_ASSERT(error != NULL, ("null error"));
333 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
334 		("invalid direction %u", dir));
335 
336 	/* Set spidx in pcb. */
337 	*error = ipsec_setspidx_inpcb(m, inp);
338 	if (*error)
339 		return (NULL);
340 
341 	pcbsp = inp->inp_sp;
342 	IPSEC_ASSERT(pcbsp != NULL, ("null pcbsp"));
343 	switch (dir) {
344 	case IPSEC_DIR_INBOUND:
345 		currsp = pcbsp->sp_in;
346 		break;
347 	case IPSEC_DIR_OUTBOUND:
348 		currsp = pcbsp->sp_out;
349 		break;
350 	}
351 	IPSEC_ASSERT(currsp != NULL, ("null currsp"));
352 
353 	if (pcbsp->priv) {			/* When privilieged socket. */
354 		switch (currsp->policy) {
355 		case IPSEC_POLICY_BYPASS:
356 		case IPSEC_POLICY_IPSEC:
357 			key_addref(currsp);
358 			sp = currsp;
359 			break;
360 
361 		case IPSEC_POLICY_ENTRUST:
362 			/* Look for a policy in SPD. */
363 			sp = KEY_ALLOCSP(&currsp->spidx, dir);
364 			if (sp == NULL)		/* No SP found. */
365 				sp = KEY_ALLOCSP_DEFAULT();
366 			break;
367 
368 		default:
369 			ipseclog((LOG_ERR, "%s: Invalid policy for PCB %d\n",
370 				__func__, currsp->policy));
371 			*error = EINVAL;
372 			return (NULL);
373 		}
374 	} else {				/* Unpriv, SPD has policy. */
375 		sp = KEY_ALLOCSP(&currsp->spidx, dir);
376 		if (sp == NULL) {		/* No SP found. */
377 			switch (currsp->policy) {
378 			case IPSEC_POLICY_BYPASS:
379 				ipseclog((LOG_ERR, "%s: Illegal policy for "
380 					"non-priviliged defined %d\n",
381 					__func__, currsp->policy));
382 				*error = EINVAL;
383 				return (NULL);
384 
385 			case IPSEC_POLICY_ENTRUST:
386 				sp = KEY_ALLOCSP_DEFAULT();
387 				break;
388 
389 			case IPSEC_POLICY_IPSEC:
390 				key_addref(currsp);
391 				sp = currsp;
392 				break;
393 
394 			default:
395 				ipseclog((LOG_ERR, "%s: Invalid policy for "
396 					"PCB %d\n", __func__, currsp->policy));
397 				*error = EINVAL;
398 				return (NULL);
399 			}
400 		}
401 	}
402 	IPSEC_ASSERT(sp != NULL,
403 		("null SP (priv %u policy %u", pcbsp->priv, currsp->policy));
404 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
405 		printf("DP %s (priv %u policy %u) allocate SP:%p (refcnt %u)\n",
406 			__func__, pcbsp->priv, currsp->policy, sp, sp->refcnt));
407 	return (sp);
408 }
409 
410 /*
411  * For FORWADING packet or OUTBOUND without a socket. Searching SPD for packet,
412  * and return a pointer to SP.
413  * OUT:	positive: a pointer to the entry for security policy leaf matched.
414  *	NULL:	no apropreate SP found, the following value is set to error.
415  *		0	: bypass
416  *		EACCES	: discard packet.
417  *		ENOENT	: ipsec_acquire() in progress, maybe.
418  *		others	: error occured.
419  */
420 struct secpolicy *
421 ipsec_getpolicybyaddr(struct mbuf *m, u_int dir, int flag, int *error)
422 {
423 	struct secpolicyindex spidx;
424 	struct secpolicy *sp;
425 
426 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
427 	IPSEC_ASSERT(error != NULL, ("null error"));
428 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
429 		("invalid direction %u", dir));
430 
431 	sp = NULL;
432 	if (key_havesp(dir)) {
433 		/* Make an index to look for a policy. */
434 		*error = ipsec_setspidx(m, &spidx,
435 					(flag & IP_FORWARDING) ? 0 : 1);
436 		if (*error != 0) {
437 			DPRINTF(("%s: setpidx failed, dir %u flag %u\n",
438 				__func__, dir, flag));
439 			return (NULL);
440 		}
441 		spidx.dir = dir;
442 
443 		sp = KEY_ALLOCSP(&spidx, dir);
444 	}
445 	if (sp == NULL)			/* No SP found, use system default. */
446 		sp = KEY_ALLOCSP_DEFAULT();
447 	IPSEC_ASSERT(sp != NULL, ("null SP"));
448 	return (sp);
449 }
450 
451 struct secpolicy *
452 ipsec4_checkpolicy(struct mbuf *m, u_int dir, u_int flag, int *error,
453     struct inpcb *inp)
454 {
455 	struct secpolicy *sp;
456 
457 	*error = 0;
458 	if (inp == NULL)
459 		sp = ipsec_getpolicybyaddr(m, dir, flag, error);
460 	else
461 		sp = ipsec_getpolicybysock(m, dir, inp, error);
462 	if (sp == NULL) {
463 		IPSEC_ASSERT(*error != 0, ("getpolicy failed w/o error"));
464 		IPSECSTAT_INC(ips_out_inval);
465 		return (NULL);
466 	}
467 	IPSEC_ASSERT(*error == 0, ("sp w/ error set to %u", *error));
468 	switch (sp->policy) {
469 	case IPSEC_POLICY_ENTRUST:
470 	default:
471 		printf("%s: invalid policy %u\n", __func__, sp->policy);
472 		/* FALLTHROUGH */
473 	case IPSEC_POLICY_DISCARD:
474 		IPSECSTAT_INC(ips_out_polvio);
475 		*error = -EINVAL;	/* Packet is discarded by caller. */
476 		break;
477 	case IPSEC_POLICY_BYPASS:
478 	case IPSEC_POLICY_NONE:
479 		KEY_FREESP(&sp);
480 		sp = NULL;		/* NB: force NULL result. */
481 		break;
482 	case IPSEC_POLICY_IPSEC:
483 		if (sp->req == NULL)	/* Acquire a SA. */
484 			*error = key_spdacquire(sp);
485 		break;
486 	}
487 	if (*error != 0) {
488 		KEY_FREESP(&sp);
489 		sp = NULL;
490 	}
491 	return (sp);
492 }
493 
494 static int
495 ipsec_setspidx_inpcb(struct mbuf *m, struct inpcb *inp)
496 {
497 	int error;
498 
499 	IPSEC_ASSERT(inp != NULL, ("null inp"));
500 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
501 	IPSEC_ASSERT(inp->inp_sp->sp_out != NULL && inp->inp_sp->sp_in != NULL,
502 		("null sp_in || sp_out"));
503 
504 	error = ipsec_setspidx(m, &inp->inp_sp->sp_in->spidx, 1);
505 	if (error == 0) {
506 		inp->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
507 		inp->inp_sp->sp_out->spidx = inp->inp_sp->sp_in->spidx;
508 		inp->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
509 	} else {
510 		bzero(&inp->inp_sp->sp_in->spidx,
511 			sizeof (inp->inp_sp->sp_in->spidx));
512 		bzero(&inp->inp_sp->sp_out->spidx,
513 			sizeof (inp->inp_sp->sp_in->spidx));
514 	}
515 	return (error);
516 }
517 
518 /*
519  * Configure security policy index (src/dst/proto/sport/dport)
520  * by looking at the content of mbuf.
521  * The caller is responsible for error recovery (like clearing up spidx).
522  */
523 static int
524 ipsec_setspidx(struct mbuf *m, struct secpolicyindex *spidx, int needport)
525 {
526 	struct ip *ip = NULL;
527 	struct ip ipbuf;
528 	u_int v;
529 	struct mbuf *n;
530 	int len;
531 	int error;
532 
533 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
534 
535 	/*
536 	 * Validate m->m_pkthdr.len.  We see incorrect length if we
537 	 * mistakenly call this function with inconsistent mbuf chain
538 	 * (like 4.4BSD tcp/udp processing).  XXX Should we panic here?
539 	 */
540 	len = 0;
541 	for (n = m; n; n = n->m_next)
542 		len += n->m_len;
543 	if (m->m_pkthdr.len != len) {
544 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
545 			printf("%s: pkthdr len(%d) mismatch (%d), ignored.\n",
546 				__func__, len, m->m_pkthdr.len));
547 		return (EINVAL);
548 	}
549 
550 	if (m->m_pkthdr.len < sizeof(struct ip)) {
551 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
552 			printf("%s: pkthdr len(%d) too small (v4), ignored.\n",
553 			    __func__, m->m_pkthdr.len));
554 		return (EINVAL);
555 	}
556 
557 	if (m->m_len >= sizeof(*ip))
558 		ip = mtod(m, struct ip *);
559 	else {
560 		m_copydata(m, 0, sizeof(ipbuf), (caddr_t)&ipbuf);
561 		ip = &ipbuf;
562 	}
563 #ifdef _IP_VHL
564 	v = _IP_VHL_V(ip->ip_vhl);
565 #else
566 	v = ip->ip_v;
567 #endif
568 	switch (v) {
569 	case 4:
570 		error = ipsec4_setspidx_ipaddr(m, spidx);
571 		if (error)
572 			return (error);
573 		ipsec4_get_ulp(m, spidx, needport);
574 		return (0);
575 #ifdef INET6
576 	case 6:
577 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) {
578 			KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
579 				printf("%s: pkthdr len(%d) too small (v6), "
580 				"ignored\n", __func__, m->m_pkthdr.len));
581 			return (EINVAL);
582 		}
583 		error = ipsec6_setspidx_ipaddr(m, spidx);
584 		if (error)
585 			return (error);
586 		ipsec6_get_ulp(m, spidx, needport);
587 		return (0);
588 #endif
589 	default:
590 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
591 			printf("%s: " "unknown IP version %u, ignored.\n",
592 				__func__, v));
593 		return (EINVAL);
594 	}
595 }
596 
597 static void
598 ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
599 {
600 	u_int8_t nxt;
601 	int off;
602 
603 	/* Sanity check. */
604 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
605 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),("packet too short"));
606 
607 	if (m->m_len >= sizeof (struct ip)) {
608 		struct ip *ip = mtod(m, struct ip *);
609 		if (ip->ip_off & htons(IP_MF | IP_OFFMASK))
610 			goto done;
611 #ifdef _IP_VHL
612 		off = _IP_VHL_HL(ip->ip_vhl) << 2;
613 #else
614 		off = ip->ip_hl << 2;
615 #endif
616 		nxt = ip->ip_p;
617 	} else {
618 		struct ip ih;
619 
620 		m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih);
621 		if (ih.ip_off & htons(IP_MF | IP_OFFMASK))
622 			goto done;
623 #ifdef _IP_VHL
624 		off = _IP_VHL_HL(ih.ip_vhl) << 2;
625 #else
626 		off = ih.ip_hl << 2;
627 #endif
628 		nxt = ih.ip_p;
629 	}
630 
631 	while (off < m->m_pkthdr.len) {
632 		struct ip6_ext ip6e;
633 		struct tcphdr th;
634 		struct udphdr uh;
635 
636 		switch (nxt) {
637 		case IPPROTO_TCP:
638 			spidx->ul_proto = nxt;
639 			if (!needport)
640 				goto done_proto;
641 			if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
642 				goto done;
643 			m_copydata(m, off, sizeof (th), (caddr_t) &th);
644 			spidx->src.sin.sin_port = th.th_sport;
645 			spidx->dst.sin.sin_port = th.th_dport;
646 			return;
647 		case IPPROTO_UDP:
648 			spidx->ul_proto = nxt;
649 			if (!needport)
650 				goto done_proto;
651 			if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
652 				goto done;
653 			m_copydata(m, off, sizeof (uh), (caddr_t) &uh);
654 			spidx->src.sin.sin_port = uh.uh_sport;
655 			spidx->dst.sin.sin_port = uh.uh_dport;
656 			return;
657 		case IPPROTO_AH:
658 			if (off + sizeof(ip6e) > m->m_pkthdr.len)
659 				goto done;
660 			/* XXX Sigh, this works but is totally bogus. */
661 			m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e);
662 			off += (ip6e.ip6e_len + 2) << 2;
663 			nxt = ip6e.ip6e_nxt;
664 			break;
665 		case IPPROTO_ICMP:
666 		default:
667 			/* XXX Intermediate headers??? */
668 			spidx->ul_proto = nxt;
669 			goto done_proto;
670 		}
671 	}
672 done:
673 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
674 done_proto:
675 	spidx->src.sin.sin_port = IPSEC_PORT_ANY;
676 	spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
677 }
678 
679 /* Assumes that m is sane. */
680 static int
681 ipsec4_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
682 {
683 	static const struct sockaddr_in template = {
684 		sizeof (struct sockaddr_in),
685 		AF_INET,
686 		0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 }
687 	};
688 
689 	spidx->src.sin = template;
690 	spidx->dst.sin = template;
691 
692 	if (m->m_len < sizeof (struct ip)) {
693 		m_copydata(m, offsetof(struct ip, ip_src),
694 			   sizeof (struct  in_addr),
695 			   (caddr_t) &spidx->src.sin.sin_addr);
696 		m_copydata(m, offsetof(struct ip, ip_dst),
697 			   sizeof (struct  in_addr),
698 			   (caddr_t) &spidx->dst.sin.sin_addr);
699 	} else {
700 		struct ip *ip = mtod(m, struct ip *);
701 		spidx->src.sin.sin_addr = ip->ip_src;
702 		spidx->dst.sin.sin_addr = ip->ip_dst;
703 	}
704 
705 	spidx->prefs = sizeof(struct in_addr) << 3;
706 	spidx->prefd = sizeof(struct in_addr) << 3;
707 
708 	return (0);
709 }
710 
711 #ifdef INET6
712 static void
713 ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
714 {
715 	int off, nxt;
716 	struct tcphdr th;
717 	struct udphdr uh;
718 	struct icmp6_hdr ih;
719 
720 	/* Sanity check. */
721 	if (m == NULL)
722 		panic("%s: NULL pointer was passed.\n", __func__);
723 
724 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
725 		printf("%s:\n", __func__); kdebug_mbuf(m));
726 
727 	/* Set default. */
728 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
729 	((struct sockaddr_in6 *)&spidx->src)->sin6_port = IPSEC_PORT_ANY;
730 	((struct sockaddr_in6 *)&spidx->dst)->sin6_port = IPSEC_PORT_ANY;
731 
732 	nxt = -1;
733 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
734 	if (off < 0 || m->m_pkthdr.len < off)
735 		return;
736 
737 	switch (nxt) {
738 	case IPPROTO_TCP:
739 		spidx->ul_proto = nxt;
740 		if (!needport)
741 			break;
742 		if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
743 			break;
744 		m_copydata(m, off, sizeof(th), (caddr_t)&th);
745 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = th.th_sport;
746 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = th.th_dport;
747 		break;
748 	case IPPROTO_UDP:
749 		spidx->ul_proto = nxt;
750 		if (!needport)
751 			break;
752 		if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
753 			break;
754 		m_copydata(m, off, sizeof(uh), (caddr_t)&uh);
755 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = uh.uh_sport;
756 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = uh.uh_dport;
757 		break;
758 	case IPPROTO_ICMPV6:
759 		spidx->ul_proto = nxt;
760 		if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len)
761 			break;
762 		m_copydata(m, off, sizeof(ih), (caddr_t)&ih);
763 		((struct sockaddr_in6 *)&spidx->src)->sin6_port =
764 		    htons((uint16_t)ih.icmp6_type);
765 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port =
766 		    htons((uint16_t)ih.icmp6_code);
767 		break;
768 	default:
769 		/* XXX Intermediate headers??? */
770 		spidx->ul_proto = nxt;
771 		break;
772 	}
773 }
774 
775 /* Assumes that m is sane. */
776 static int
777 ipsec6_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
778 {
779 	struct ip6_hdr *ip6 = NULL;
780 	struct ip6_hdr ip6buf;
781 	struct sockaddr_in6 *sin6;
782 
783 	if (m->m_len >= sizeof(*ip6))
784 		ip6 = mtod(m, struct ip6_hdr *);
785 	else {
786 		m_copydata(m, 0, sizeof(ip6buf), (caddr_t)&ip6buf);
787 		ip6 = &ip6buf;
788 	}
789 
790 	sin6 = (struct sockaddr_in6 *)&spidx->src;
791 	bzero(sin6, sizeof(*sin6));
792 	sin6->sin6_family = AF_INET6;
793 	sin6->sin6_len = sizeof(struct sockaddr_in6);
794 	bcopy(&ip6->ip6_src, &sin6->sin6_addr, sizeof(ip6->ip6_src));
795 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
796 		sin6->sin6_addr.s6_addr16[1] = 0;
797 		sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]);
798 	}
799 	spidx->prefs = sizeof(struct in6_addr) << 3;
800 
801 	sin6 = (struct sockaddr_in6 *)&spidx->dst;
802 	bzero(sin6, sizeof(*sin6));
803 	sin6->sin6_family = AF_INET6;
804 	sin6->sin6_len = sizeof(struct sockaddr_in6);
805 	bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(ip6->ip6_dst));
806 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
807 		sin6->sin6_addr.s6_addr16[1] = 0;
808 		sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]);
809 	}
810 	spidx->prefd = sizeof(struct in6_addr) << 3;
811 
812 	return (0);
813 }
814 #endif
815 
816 static void
817 ipsec_delpcbpolicy(struct inpcbpolicy *p)
818 {
819 
820 	free(p, M_IPSEC_INPCB);
821 }
822 
823 /* Initialize policy in PCB. */
824 int
825 ipsec_init_policy(struct socket *so, struct inpcbpolicy **pcb_sp)
826 {
827 	struct inpcbpolicy *new;
828 
829 	/* Sanity check. */
830 	if (so == NULL || pcb_sp == NULL)
831 		panic("%s: NULL pointer was passed.\n", __func__);
832 
833 	new = (struct inpcbpolicy *) malloc(sizeof(struct inpcbpolicy),
834 					    M_IPSEC_INPCB, M_NOWAIT|M_ZERO);
835 	if (new == NULL) {
836 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
837 		return (ENOBUFS);
838 	}
839 
840 	new->priv = IPSEC_IS_PRIVILEGED_SO(so);
841 
842 	if ((new->sp_in = KEY_NEWSP()) == NULL) {
843 		ipsec_delpcbpolicy(new);
844 		return (ENOBUFS);
845 	}
846 	new->sp_in->state = IPSEC_SPSTATE_ALIVE;
847 	new->sp_in->policy = IPSEC_POLICY_ENTRUST;
848 
849 	if ((new->sp_out = KEY_NEWSP()) == NULL) {
850 		KEY_FREESP(&new->sp_in);
851 		ipsec_delpcbpolicy(new);
852 		return (ENOBUFS);
853 	}
854 	new->sp_out->state = IPSEC_SPSTATE_ALIVE;
855 	new->sp_out->policy = IPSEC_POLICY_ENTRUST;
856 
857 	*pcb_sp = new;
858 
859 	return (0);
860 }
861 
862 /* Copy old IPsec policy into new. */
863 int
864 ipsec_copy_policy(struct inpcbpolicy *old, struct inpcbpolicy *new)
865 {
866 	struct secpolicy *sp;
867 
868 	sp = ipsec_deepcopy_policy(old->sp_in);
869 	if (sp) {
870 		KEY_FREESP(&new->sp_in);
871 		new->sp_in = sp;
872 	} else
873 		return (ENOBUFS);
874 
875 	sp = ipsec_deepcopy_policy(old->sp_out);
876 	if (sp) {
877 		KEY_FREESP(&new->sp_out);
878 		new->sp_out = sp;
879 	} else
880 		return (ENOBUFS);
881 
882 	new->priv = old->priv;
883 
884 	return (0);
885 }
886 
887 struct ipsecrequest *
888 ipsec_newisr(void)
889 {
890 	struct ipsecrequest *p;
891 
892 	p = malloc(sizeof(struct ipsecrequest), M_IPSEC_SR, M_NOWAIT|M_ZERO);
893 	if (p != NULL)
894 		IPSECREQUEST_LOCK_INIT(p);
895 	return (p);
896 }
897 
898 void
899 ipsec_delisr(struct ipsecrequest *p)
900 {
901 
902 	IPSECREQUEST_LOCK_DESTROY(p);
903 	free(p, M_IPSEC_SR);
904 }
905 
906 /* Deep-copy a policy in PCB. */
907 static struct secpolicy *
908 ipsec_deepcopy_policy(struct secpolicy *src)
909 {
910 	struct ipsecrequest *newchain = NULL;
911 	struct ipsecrequest *p;
912 	struct ipsecrequest **q;
913 	struct ipsecrequest *r;
914 	struct secpolicy *dst;
915 
916 	if (src == NULL)
917 		return (NULL);
918 	dst = KEY_NEWSP();
919 	if (dst == NULL)
920 		return (NULL);
921 
922 	/*
923 	 * Deep-copy IPsec request chain.  This is required since struct
924 	 * ipsecrequest is not reference counted.
925 	 */
926 	q = &newchain;
927 	for (p = src->req; p; p = p->next) {
928 		*q = ipsec_newisr();
929 		if (*q == NULL)
930 			goto fail;
931 		(*q)->saidx.proto = p->saidx.proto;
932 		(*q)->saidx.mode = p->saidx.mode;
933 		(*q)->level = p->level;
934 		(*q)->saidx.reqid = p->saidx.reqid;
935 
936 		bcopy(&p->saidx.src, &(*q)->saidx.src, sizeof((*q)->saidx.src));
937 		bcopy(&p->saidx.dst, &(*q)->saidx.dst, sizeof((*q)->saidx.dst));
938 
939 		(*q)->sp = dst;
940 
941 		q = &((*q)->next);
942 	}
943 
944 	dst->req = newchain;
945 	dst->state = src->state;
946 	dst->policy = src->policy;
947 	/* Do not touch the refcnt fields. */
948 
949 	return (dst);
950 
951 fail:
952 	for (p = newchain; p; p = r) {
953 		r = p->next;
954 		ipsec_delisr(p);
955 		p = NULL;
956 	}
957 	return (NULL);
958 }
959 
960 /* Set policy and IPsec request if present. */
961 static int
962 ipsec_set_policy_internal(struct secpolicy **pcb_sp, int optname,
963     caddr_t request, size_t len, struct ucred *cred)
964 {
965 	struct sadb_x_policy *xpl;
966 	struct secpolicy *newsp = NULL;
967 	int error;
968 
969 	/* Sanity check. */
970 	if (pcb_sp == NULL || *pcb_sp == NULL || request == NULL)
971 		return (EINVAL);
972 	if (len < sizeof(*xpl))
973 		return (EINVAL);
974 	xpl = (struct sadb_x_policy *)request;
975 
976 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
977 		printf("%s: passed policy\n", __func__);
978 		kdebug_sadb_x_policy((struct sadb_ext *)xpl));
979 
980 	/* Check policy type. */
981 	/* ipsec_set_policy_internal() accepts IPSEC, ENTRUST and BYPASS. */
982 	if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD
983 	 || xpl->sadb_x_policy_type == IPSEC_POLICY_NONE)
984 		return (EINVAL);
985 
986 	/* Check privileged socket. */
987 	if (cred != NULL && xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
988 		error = priv_check_cred(cred, PRIV_NETINET_IPSEC, 0);
989 		if (error)
990 			return (EACCES);
991 	}
992 
993 	/* Allocating new SP entry. */
994 	if ((newsp = key_msg2sp(xpl, len, &error)) == NULL)
995 		return (error);
996 
997 	newsp->state = IPSEC_SPSTATE_ALIVE;
998 
999 	/* Clear old SP and set new SP. */
1000 	KEY_FREESP(pcb_sp);
1001 	*pcb_sp = newsp;
1002 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1003 		printf("%s: new policy\n", __func__);
1004 		kdebug_secpolicy(newsp));
1005 
1006 	return (0);
1007 }
1008 
1009 int
1010 ipsec_set_policy(struct inpcb *inp, int optname, caddr_t request,
1011     size_t len, struct ucred *cred)
1012 {
1013 	struct sadb_x_policy *xpl;
1014 	struct secpolicy **pcb_sp;
1015 
1016 	/* Sanity check. */
1017 	if (inp == NULL || request == NULL)
1018 		return (EINVAL);
1019 	if (len < sizeof(*xpl))
1020 		return (EINVAL);
1021 	xpl = (struct sadb_x_policy *)request;
1022 
1023 	/* Select direction. */
1024 	switch (xpl->sadb_x_policy_dir) {
1025 	case IPSEC_DIR_INBOUND:
1026 		pcb_sp = &inp->inp_sp->sp_in;
1027 		break;
1028 	case IPSEC_DIR_OUTBOUND:
1029 		pcb_sp = &inp->inp_sp->sp_out;
1030 		break;
1031 	default:
1032 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1033 			xpl->sadb_x_policy_dir));
1034 		return (EINVAL);
1035 	}
1036 
1037 	return (ipsec_set_policy_internal(pcb_sp, optname, request, len, cred));
1038 }
1039 
1040 int
1041 ipsec_get_policy(struct inpcb *inp, caddr_t request, size_t len,
1042     struct mbuf **mp)
1043 {
1044 	struct sadb_x_policy *xpl;
1045 	struct secpolicy *pcb_sp;
1046 
1047 	/* Sanity check. */
1048 	if (inp == NULL || request == NULL || mp == NULL)
1049 		return (EINVAL);
1050 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
1051 	if (len < sizeof(*xpl))
1052 		return (EINVAL);
1053 	xpl = (struct sadb_x_policy *)request;
1054 
1055 	/* Select direction. */
1056 	switch (xpl->sadb_x_policy_dir) {
1057 	case IPSEC_DIR_INBOUND:
1058 		pcb_sp = inp->inp_sp->sp_in;
1059 		break;
1060 	case IPSEC_DIR_OUTBOUND:
1061 		pcb_sp = inp->inp_sp->sp_out;
1062 		break;
1063 	default:
1064 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1065 			xpl->sadb_x_policy_dir));
1066 		return (EINVAL);
1067 	}
1068 
1069 	/* Sanity check. Should be an IPSEC_ASSERT. */
1070 	if (pcb_sp == NULL)
1071 		return (EINVAL);
1072 
1073 	*mp = key_sp2msg(pcb_sp);
1074 	if (!*mp) {
1075 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
1076 		return (ENOBUFS);
1077 	}
1078 
1079 	(*mp)->m_type = MT_DATA;
1080 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1081 		printf("%s:\n", __func__); kdebug_mbuf(*mp));
1082 
1083 	return (0);
1084 }
1085 
1086 /* Delete policy in PCB. */
1087 int
1088 ipsec_delete_pcbpolicy(struct inpcb *inp)
1089 {
1090 	IPSEC_ASSERT(inp != NULL, ("null inp"));
1091 
1092 	if (inp->inp_sp == NULL)
1093 		return (0);
1094 
1095 	if (inp->inp_sp->sp_in != NULL)
1096 		KEY_FREESP(&inp->inp_sp->sp_in);
1097 
1098 	if (inp->inp_sp->sp_out != NULL)
1099 		KEY_FREESP(&inp->inp_sp->sp_out);
1100 
1101 	ipsec_delpcbpolicy(inp->inp_sp);
1102 	inp->inp_sp = NULL;
1103 
1104 	return (0);
1105 }
1106 
1107 /*
1108  * Return current level.
1109  * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
1110  */
1111 u_int
1112 ipsec_get_reqlevel(struct ipsecrequest *isr)
1113 {
1114 	u_int level = 0;
1115 	u_int esp_trans_deflev, esp_net_deflev;
1116 	u_int ah_trans_deflev, ah_net_deflev;
1117 
1118 	IPSEC_ASSERT(isr != NULL && isr->sp != NULL, ("null argument"));
1119 	IPSEC_ASSERT(isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family,
1120 		("af family mismatch, src %u, dst %u",
1121 		 isr->sp->spidx.src.sa.sa_family,
1122 		 isr->sp->spidx.dst.sa.sa_family));
1123 
1124 /* XXX Note that we have ipseclog() expanded here - code sync issue. */
1125 #define IPSEC_CHECK_DEFAULT(lev) \
1126 	(((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE	      \
1127 			&& (lev) != IPSEC_LEVEL_UNIQUE)			      \
1128 		? (V_ipsec_debug						      \
1129 			? log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
1130 				(lev), IPSEC_LEVEL_REQUIRE)		      \
1131 			: 0),						      \
1132 			(lev) = IPSEC_LEVEL_REQUIRE,			      \
1133 			(lev)						      \
1134 		: (lev))
1135 
1136 	/* Set default level. */
1137 	switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) {
1138 #ifdef INET
1139 	case AF_INET:
1140 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev);
1141 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev);
1142 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev);
1143 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev);
1144 		break;
1145 #endif
1146 #ifdef INET6
1147 	case AF_INET6:
1148 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev);
1149 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev);
1150 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev);
1151 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev);
1152 		break;
1153 #endif /* INET6 */
1154 	default:
1155 		panic("%s: unknown af %u",
1156 			__func__, isr->sp->spidx.src.sa.sa_family);
1157 	}
1158 
1159 #undef IPSEC_CHECK_DEFAULT
1160 
1161 	/* Set level. */
1162 	switch (isr->level) {
1163 	case IPSEC_LEVEL_DEFAULT:
1164 		switch (isr->saidx.proto) {
1165 		case IPPROTO_ESP:
1166 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1167 				level = esp_net_deflev;
1168 			else
1169 				level = esp_trans_deflev;
1170 			break;
1171 		case IPPROTO_AH:
1172 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1173 				level = ah_net_deflev;
1174 			else
1175 				level = ah_trans_deflev;
1176 			break;
1177 		case IPPROTO_IPCOMP:
1178 			/*
1179 			 * We don't really care, as IPcomp document says that
1180 			 * we shouldn't compress small packets.
1181 			 */
1182 			level = IPSEC_LEVEL_USE;
1183 			break;
1184 		default:
1185 			panic("%s: Illegal protocol defined %u\n", __func__,
1186 				isr->saidx.proto);
1187 		}
1188 		break;
1189 
1190 	case IPSEC_LEVEL_USE:
1191 	case IPSEC_LEVEL_REQUIRE:
1192 		level = isr->level;
1193 		break;
1194 	case IPSEC_LEVEL_UNIQUE:
1195 		level = IPSEC_LEVEL_REQUIRE;
1196 		break;
1197 
1198 	default:
1199 		panic("%s: Illegal IPsec level %u\n", __func__, isr->level);
1200 	}
1201 
1202 	return (level);
1203 }
1204 
1205 /*
1206  * Check security policy requirements against the actual
1207  * packet contents.  Return one if the packet should be
1208  * reject as "invalid"; otherwiser return zero to have the
1209  * packet treated as "valid".
1210  *
1211  * OUT:
1212  *	0: valid
1213  *	1: invalid
1214  */
1215 int
1216 ipsec_in_reject(struct secpolicy *sp, struct mbuf *m)
1217 {
1218 	struct ipsecrequest *isr;
1219 	int need_auth;
1220 
1221 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1222 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1223 
1224 	/* Check policy. */
1225 	switch (sp->policy) {
1226 	case IPSEC_POLICY_DISCARD:
1227 		return (1);
1228 	case IPSEC_POLICY_BYPASS:
1229 	case IPSEC_POLICY_NONE:
1230 		return (0);
1231 	}
1232 
1233 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1234 		("invalid policy %u", sp->policy));
1235 
1236 	/* XXX Should compare policy against IPsec header history. */
1237 
1238 	need_auth = 0;
1239 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1240 		if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE)
1241 			continue;
1242 		switch (isr->saidx.proto) {
1243 		case IPPROTO_ESP:
1244 			if ((m->m_flags & M_DECRYPTED) == 0) {
1245 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1246 				    printf("%s: ESP m_flags:%x\n", __func__,
1247 					    m->m_flags));
1248 				return (1);
1249 			}
1250 
1251 			if (!need_auth &&
1252 			    isr->sav != NULL &&
1253 			    isr->sav->tdb_authalgxform != NULL &&
1254 			    (m->m_flags & M_AUTHIPDGM) == 0) {
1255 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1256 				    printf("%s: ESP/AH m_flags:%x\n", __func__,
1257 					    m->m_flags));
1258 				return (1);
1259 			}
1260 			break;
1261 		case IPPROTO_AH:
1262 			need_auth = 1;
1263 			if ((m->m_flags & M_AUTHIPHDR) == 0) {
1264 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1265 				    printf("%s: AH m_flags:%x\n", __func__,
1266 					    m->m_flags));
1267 				return (1);
1268 			}
1269 			break;
1270 		case IPPROTO_IPCOMP:
1271 			/*
1272 			 * We don't really care, as IPcomp document
1273 			 * says that we shouldn't compress small
1274 			 * packets.  IPComp policy should always be
1275 			 * treated as being in "use" level.
1276 			 */
1277 			break;
1278 		}
1279 	}
1280 	return (0);		/* Valid. */
1281 }
1282 
1283 static int
1284 ipsec46_in_reject(struct mbuf *m, struct inpcb *inp)
1285 {
1286 	struct secpolicy *sp;
1287 	int error;
1288 	int result;
1289 
1290 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1291 
1292 	/*
1293 	 * Get SP for this packet.
1294 	 * When we are called from ip_forward(), we call
1295 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1296 	 */
1297 	if (inp == NULL)
1298 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, IP_FORWARDING, &error);
1299 	else
1300 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND, inp, &error);
1301 
1302 	if (sp != NULL) {
1303 		result = ipsec_in_reject(sp, m);
1304 		KEY_FREESP(&sp);
1305 	} else {
1306 		result = 0;	/* XXX Should be panic?
1307 				 * -> No, there may be error. */
1308 	}
1309 	return (result);
1310 }
1311 
1312 /*
1313  * Check AH/ESP integrity.
1314  * This function is called from tcp_input(), udp_input(),
1315  * and {ah,esp}4_input for tunnel mode.
1316  */
1317 int
1318 ipsec4_in_reject(struct mbuf *m, struct inpcb *inp)
1319 {
1320 	int result;
1321 
1322 	result = ipsec46_in_reject(m, inp);
1323 	if (result)
1324 		IPSECSTAT_INC(ips_in_polvio);
1325 
1326 	return (result);
1327 }
1328 
1329 #ifdef INET6
1330 /*
1331  * Check AH/ESP integrity.
1332  * This function is called from tcp6_input(), udp6_input(),
1333  * and {ah,esp}6_input for tunnel mode.
1334  */
1335 int
1336 ipsec6_in_reject(struct mbuf *m, struct inpcb *inp)
1337 {
1338 	int result;
1339 
1340 	result = ipsec46_in_reject(m, inp);
1341 	if (result)
1342 		IPSEC6STAT_INC(ips_in_polvio);
1343 
1344 	return (result);
1345 }
1346 #endif
1347 
1348 /*
1349  * Compute the byte size to be occupied by IPsec header.
1350  * In case it is tunnelled, it includes the size of outer IP header.
1351  * NOTE: SP passed is freed in this function.
1352  */
1353 static size_t
1354 ipsec_hdrsiz_internal(struct secpolicy *sp)
1355 {
1356 	struct ipsecrequest *isr;
1357 	size_t size;
1358 
1359 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1360 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1361 
1362 	switch (sp->policy) {
1363 	case IPSEC_POLICY_DISCARD:
1364 	case IPSEC_POLICY_BYPASS:
1365 	case IPSEC_POLICY_NONE:
1366 		return (0);
1367 	}
1368 
1369 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1370 		("invalid policy %u", sp->policy));
1371 
1372 	size = 0;
1373 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1374 		size_t clen = 0;
1375 
1376 		switch (isr->saidx.proto) {
1377 		case IPPROTO_ESP:
1378 			clen = esp_hdrsiz(isr->sav);
1379 			break;
1380 		case IPPROTO_AH:
1381 			clen = ah_hdrsiz(isr->sav);
1382 			break;
1383 		case IPPROTO_IPCOMP:
1384 			clen = sizeof(struct ipcomp);
1385 			break;
1386 		}
1387 
1388 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1389 			switch (isr->saidx.dst.sa.sa_family) {
1390 			case AF_INET:
1391 				clen += sizeof(struct ip);
1392 				break;
1393 #ifdef INET6
1394 			case AF_INET6:
1395 				clen += sizeof(struct ip6_hdr);
1396 				break;
1397 #endif
1398 			default:
1399 				ipseclog((LOG_ERR, "%s: unknown AF %d in "
1400 				    "IPsec tunnel SA\n", __func__,
1401 				    ((struct sockaddr *)&isr->saidx.dst)->sa_family));
1402 				break;
1403 			}
1404 		}
1405 		size += clen;
1406 	}
1407 
1408 	return (size);
1409 }
1410 
1411 /*
1412  * This function is called from ipsec_hdrsiz_tcp(), ip_ipsec_mtu(),
1413  * disabled ip6_ipsec_mtu() and ip6_forward().
1414  */
1415 size_t
1416 ipsec_hdrsiz(struct mbuf *m, u_int dir, struct inpcb *inp)
1417 {
1418 	struct secpolicy *sp;
1419 	int error;
1420 	size_t size;
1421 
1422 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1423 
1424 	/* Get SP for this packet.
1425 	 * When we are called from ip_forward(), we call
1426 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1427 	 */
1428 	if (inp == NULL)
1429 		sp = ipsec_getpolicybyaddr(m, dir, IP_FORWARDING, &error);
1430 	else
1431 		sp = ipsec_getpolicybysock(m, dir, inp, &error);
1432 
1433 	if (sp != NULL) {
1434 		size = ipsec_hdrsiz_internal(sp);
1435 		KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1436 			printf("%s: size:%lu.\n", __func__,
1437 				(unsigned long)size));
1438 
1439 		KEY_FREESP(&sp);
1440 	} else {
1441 		size = 0;	/* XXX Should be panic?
1442 				 * -> No, we are called w/o knowing if
1443 				 *    IPsec processing is needed. */
1444 	}
1445 	return (size);
1446 }
1447 
1448 /*
1449  * Check the variable replay window.
1450  * ipsec_chkreplay() performs replay check before ICV verification.
1451  * ipsec_updatereplay() updates replay bitmap.  This must be called after
1452  * ICV verification (it also performs replay check, which is usually done
1453  * beforehand).
1454  * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1455  *
1456  * Based on RFC 2401.
1457  */
1458 int
1459 ipsec_chkreplay(u_int32_t seq, struct secasvar *sav)
1460 {
1461 	const struct secreplay *replay;
1462 	u_int32_t diff;
1463 	int fr;
1464 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1465 	int frlast;		/* Constant: last frame. */
1466 
1467 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1468 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1469 
1470 	replay = sav->replay;
1471 
1472 	if (replay->wsize == 0)
1473 		return (1);	/* No need to check replay. */
1474 
1475 	/* Constant. */
1476 	frlast = replay->wsize - 1;
1477 	wsizeb = replay->wsize << 3;
1478 
1479 	/* Sequence number of 0 is invalid. */
1480 	if (seq == 0)
1481 		return (0);
1482 
1483 	/* First time is always okay. */
1484 	if (replay->count == 0)
1485 		return (1);
1486 
1487 	if (seq > replay->lastseq) {
1488 		/* Larger sequences are okay. */
1489 		return (1);
1490 	} else {
1491 		/* seq is equal or less than lastseq. */
1492 		diff = replay->lastseq - seq;
1493 
1494 		/* Over range to check, i.e. too old or wrapped. */
1495 		if (diff >= wsizeb)
1496 			return (0);
1497 
1498 		fr = frlast - diff / 8;
1499 
1500 		/* This packet already seen? */
1501 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1502 			return (0);
1503 
1504 		/* Out of order but good. */
1505 		return (1);
1506 	}
1507 }
1508 
1509 /*
1510  * Check replay counter whether to update or not.
1511  * OUT:	0:	OK
1512  *	1:	NG
1513  */
1514 int
1515 ipsec_updatereplay(u_int32_t seq, struct secasvar *sav)
1516 {
1517 	struct secreplay *replay;
1518 	u_int32_t diff;
1519 	int fr;
1520 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1521 	int frlast;		/* Constant: last frame. */
1522 
1523 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1524 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1525 
1526 	replay = sav->replay;
1527 
1528 	if (replay->wsize == 0)
1529 		goto ok;	/* No need to check replay. */
1530 
1531 	/* Constant. */
1532 	frlast = replay->wsize - 1;
1533 	wsizeb = replay->wsize << 3;
1534 
1535 	/* Sequence number of 0 is invalid. */
1536 	if (seq == 0)
1537 		return (1);
1538 
1539 	/* First time. */
1540 	if (replay->count == 0) {
1541 		replay->lastseq = seq;
1542 		bzero(replay->bitmap, replay->wsize);
1543 		(replay->bitmap)[frlast] = 1;
1544 		goto ok;
1545 	}
1546 
1547 	if (seq > replay->lastseq) {
1548 		/* seq is larger than lastseq. */
1549 		diff = seq - replay->lastseq;
1550 
1551 		/* New larger sequence number. */
1552 		if (diff < wsizeb) {
1553 			/* In window. */
1554 			/* Set bit for this packet. */
1555 			vshiftl(replay->bitmap, diff, replay->wsize);
1556 			(replay->bitmap)[frlast] |= 1;
1557 		} else {
1558 			/* This packet has a "way larger". */
1559 			bzero(replay->bitmap, replay->wsize);
1560 			(replay->bitmap)[frlast] = 1;
1561 		}
1562 		replay->lastseq = seq;
1563 
1564 		/* Larger is good. */
1565 	} else {
1566 		/* seq is equal or less than lastseq. */
1567 		diff = replay->lastseq - seq;
1568 
1569 		/* Over range to check, i.e. too old or wrapped. */
1570 		if (diff >= wsizeb)
1571 			return (1);
1572 
1573 		fr = frlast - diff / 8;
1574 
1575 		/* This packet already seen? */
1576 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1577 			return (1);
1578 
1579 		/* Mark as seen. */
1580 		(replay->bitmap)[fr] |= (1 << (diff % 8));
1581 
1582 		/* Out of order but good. */
1583 	}
1584 
1585 ok:
1586 	if (replay->count == ~0) {
1587 
1588 		/* Set overflow flag. */
1589 		replay->overflow++;
1590 
1591 		/* Don't increment, no more packets accepted. */
1592 		if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0)
1593 			return (1);
1594 
1595 		ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n",
1596 		    __func__, replay->overflow, ipsec_logsastr(sav)));
1597 	}
1598 
1599 	replay->count++;
1600 
1601 	return (0);
1602 }
1603 
1604 /*
1605  * Shift variable length buffer to left.
1606  * IN:	bitmap: pointer to the buffer
1607  * 	nbit:	the number of to shift.
1608  *	wsize:	buffer size (bytes).
1609  */
1610 static void
1611 vshiftl(unsigned char *bitmap, int nbit, int wsize)
1612 {
1613 	int s, j, i;
1614 	unsigned char over;
1615 
1616 	for (j = 0; j < nbit; j += 8) {
1617 		s = (nbit - j < 8) ? (nbit - j): 8;
1618 		bitmap[0] <<= s;
1619 		for (i = 1; i < wsize; i++) {
1620 			over = (bitmap[i] >> (8 - s));
1621 			bitmap[i] <<= s;
1622 			bitmap[i-1] |= over;
1623 		}
1624 	}
1625 }
1626 
1627 #ifdef INET
1628 /* Return a printable string for the IPv4 address. */
1629 static char *
1630 inet_ntoa4(struct in_addr ina)
1631 {
1632 	static char buf[4][4 * sizeof "123" + 4];
1633 	unsigned char *ucp = (unsigned char *) &ina;
1634 	static int i = 3;
1635 
1636 	/* XXX-BZ Returns static buffer. */
1637 	i = (i + 1) % 4;
1638 	sprintf(buf[i], "%d.%d.%d.%d", ucp[0] & 0xff, ucp[1] & 0xff,
1639 	    ucp[2] & 0xff, ucp[3] & 0xff);
1640 	return (buf[i]);
1641 }
1642 #endif
1643 
1644 /* Return a printable string for the address. */
1645 char *
1646 ipsec_address(union sockaddr_union* sa)
1647 {
1648 #ifdef INET6
1649 	char ip6buf[INET6_ADDRSTRLEN];
1650 #endif
1651 
1652 	switch (sa->sa.sa_family) {
1653 #ifdef INET
1654 	case AF_INET:
1655 		return (inet_ntoa4(sa->sin.sin_addr));
1656 #endif /* INET */
1657 #ifdef INET6
1658 	case AF_INET6:
1659 		return (ip6_sprintf(ip6buf, &sa->sin6.sin6_addr));
1660 #endif /* INET6 */
1661 	default:
1662 		return ("(unknown address family)");
1663 	}
1664 }
1665 
1666 const char *
1667 ipsec_logsastr(struct secasvar *sav)
1668 {
1669 	static char buf[256];
1670 	char *p;
1671 	struct secasindex *saidx = &sav->sah->saidx;
1672 
1673 	IPSEC_ASSERT(saidx->src.sa.sa_family == saidx->dst.sa.sa_family,
1674 		("address family mismatch"));
1675 
1676 	p = buf;
1677 	snprintf(buf, sizeof(buf), "SA(SPI=%u ", (u_int32_t)ntohl(sav->spi));
1678 	while (p && *p)
1679 		p++;
1680 	/* NB: only use ipsec_address on one address at a time. */
1681 	snprintf(p, sizeof (buf) - (p - buf), "src=%s ",
1682 		ipsec_address(&saidx->src));
1683 	while (p && *p)
1684 		p++;
1685 	snprintf(p, sizeof (buf) - (p - buf), "dst=%s)",
1686 		ipsec_address(&saidx->dst));
1687 
1688 	return (buf);
1689 }
1690 
1691 void
1692 ipsec_dumpmbuf(struct mbuf *m)
1693 {
1694 	int totlen;
1695 	int i;
1696 	u_char *p;
1697 
1698 	totlen = 0;
1699 	printf("---\n");
1700 	while (m) {
1701 		p = mtod(m, u_char *);
1702 		for (i = 0; i < m->m_len; i++) {
1703 			printf("%02x ", p[i]);
1704 			totlen++;
1705 			if (totlen % 16 == 0)
1706 				printf("\n");
1707 		}
1708 		m = m->m_next;
1709 	}
1710 	if (totlen % 16 != 0)
1711 		printf("\n");
1712 	printf("---\n");
1713 }
1714 
1715 static void
1716 ipsec_init(const void *unused __unused)
1717 {
1718 
1719 	SECPOLICY_LOCK_INIT(&V_ip4_def_policy);
1720 	V_ip4_def_policy.refcnt = 1;			/* NB: disallow free. */
1721 }
1722 VNET_SYSINIT(ipsec_init, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, ipsec_init,
1723     NULL);
1724 
1725 
1726 /* XXX This stuff doesn't belong here... */
1727 
1728 static	struct xformsw* xforms = NULL;
1729 
1730 /*
1731  * Register a transform; typically at system startup.
1732  */
1733 void
1734 xform_register(struct xformsw* xsp)
1735 {
1736 
1737 	xsp->xf_next = xforms;
1738 	xforms = xsp;
1739 }
1740 
1741 /*
1742  * Initialize transform support in an sav.
1743  */
1744 int
1745 xform_init(struct secasvar *sav, int xftype)
1746 {
1747 	struct xformsw *xsp;
1748 
1749 	if (sav->tdb_xform != NULL)	/* Previously initialized. */
1750 		return (0);
1751 	for (xsp = xforms; xsp; xsp = xsp->xf_next)
1752 		if (xsp->xf_type == xftype)
1753 			return ((*xsp->xf_init)(sav, xsp));
1754 	return (EINVAL);
1755 }
1756