xref: /freebsd/sys/netipsec/ipsec.c (revision aa0a1e58)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ipsec.c,v 1.103 2001/05/24 07:14:18 sakane Exp $	*/
3 
4 /*-
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * IPsec controller part.
35  */
36 
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39 #include "opt_ipsec.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/malloc.h>
44 #include <sys/mbuf.h>
45 #include <sys/domain.h>
46 #include <sys/priv.h>
47 #include <sys/protosw.h>
48 #include <sys/socket.h>
49 #include <sys/socketvar.h>
50 #include <sys/errno.h>
51 #include <sys/time.h>
52 #include <sys/kernel.h>
53 #include <sys/syslog.h>
54 #include <sys/sysctl.h>
55 #include <sys/proc.h>
56 
57 #include <net/if.h>
58 #include <net/route.h>
59 #include <net/vnet.h>
60 
61 #include <netinet/in.h>
62 #include <netinet/in_systm.h>
63 #include <netinet/ip.h>
64 #include <netinet/ip_var.h>
65 #include <netinet/in_var.h>
66 #include <netinet/udp.h>
67 #include <netinet/udp_var.h>
68 #include <netinet/tcp.h>
69 #include <netinet/udp.h>
70 
71 #include <netinet/ip6.h>
72 #ifdef INET6
73 #include <netinet6/ip6_var.h>
74 #endif
75 #include <netinet/in_pcb.h>
76 #ifdef INET6
77 #include <netinet/icmp6.h>
78 #endif
79 
80 #include <sys/types.h>
81 #include <netipsec/ipsec.h>
82 #ifdef INET6
83 #include <netipsec/ipsec6.h>
84 #endif
85 #include <netipsec/ah_var.h>
86 #include <netipsec/esp_var.h>
87 #include <netipsec/ipcomp.h>		/*XXX*/
88 #include <netipsec/ipcomp_var.h>
89 
90 #include <netipsec/key.h>
91 #include <netipsec/keydb.h>
92 #include <netipsec/key_debug.h>
93 
94 #include <netipsec/xform.h>
95 
96 #include <machine/in_cksum.h>
97 
98 #include <opencrypto/cryptodev.h>
99 
100 #ifdef IPSEC_DEBUG
101 VNET_DEFINE(int, ipsec_debug) = 1;
102 #else
103 VNET_DEFINE(int, ipsec_debug) = 0;
104 #endif
105 
106 /* NB: name changed so netstat doesn't use it. */
107 VNET_DEFINE(struct ipsecstat, ipsec4stat);
108 VNET_DEFINE(int, ip4_ah_offsetmask) = 0;	/* maybe IP_DF? */
109 /* DF bit on encap. 0: clear 1: set 2: copy */
110 VNET_DEFINE(int, ip4_ipsec_dfbit) = 0;
111 VNET_DEFINE(int, ip4_esp_trans_deflev) = IPSEC_LEVEL_USE;
112 VNET_DEFINE(int, ip4_esp_net_deflev) = IPSEC_LEVEL_USE;
113 VNET_DEFINE(int, ip4_ah_trans_deflev) = IPSEC_LEVEL_USE;
114 VNET_DEFINE(int, ip4_ah_net_deflev) = IPSEC_LEVEL_USE;
115 VNET_DEFINE(struct secpolicy, ip4_def_policy);
116 /* ECN ignore(-1)/forbidden(0)/allowed(1) */
117 VNET_DEFINE(int, ip4_ipsec_ecn) = 0;
118 VNET_DEFINE(int, ip4_esp_randpad) = -1;
119 
120 /*
121  * Crypto support requirements:
122  *
123  *  1	require hardware support
124  * -1	require software support
125  *  0	take anything
126  */
127 VNET_DEFINE(int, crypto_support) = CRYPTOCAP_F_HARDWARE | CRYPTOCAP_F_SOFTWARE;
128 
129 FEATURE(ipsec, "Internet Protocol Security (IPsec)");
130 #ifdef IPSEC_NAT_T
131 FEATURE(ipsec_natt, "UDP Encapsulation of IPsec ESP Packets ('NAT-T')");
132 #endif
133 
134 SYSCTL_DECL(_net_inet_ipsec);
135 
136 /* net.inet.ipsec */
137 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_POLICY, def_policy,
138 	CTLFLAG_RW, &VNET_NAME(ip4_def_policy).policy, 0,
139 	"IPsec default policy.");
140 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_TRANSLEV, esp_trans_deflev,
141 	CTLFLAG_RW, &VNET_NAME(ip4_esp_trans_deflev), 0,
142 	"Default ESP transport mode level");
143 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_ESP_NETLEV, esp_net_deflev,
144 	CTLFLAG_RW, &VNET_NAME(ip4_esp_net_deflev), 0,
145 	"Default ESP tunnel mode level.");
146 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_TRANSLEV, ah_trans_deflev,
147 	CTLFLAG_RW, &VNET_NAME(ip4_ah_trans_deflev), 0,
148 	"AH transfer mode default level.");
149 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEF_AH_NETLEV, ah_net_deflev,
150 	CTLFLAG_RW, &VNET_NAME(ip4_ah_net_deflev), 0,
151 	"AH tunnel mode default level.");
152 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_AH_CLEARTOS, ah_cleartos,
153 	CTLFLAG_RW, &VNET_NAME(ah_cleartos), 0,
154 	"If set clear type-of-service field when doing AH computation.");
155 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_AH_OFFSETMASK, ah_offsetmask,
156 	CTLFLAG_RW, &VNET_NAME(ip4_ah_offsetmask), 0,
157 	"If not set clear offset field mask when doing AH computation.");
158 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DFBIT, dfbit,
159 	CTLFLAG_RW, &VNET_NAME(ip4_ipsec_dfbit), 0,
160 	"Do not fragment bit on encap.");
161 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_ECN, ecn,
162 	CTLFLAG_RW, &VNET_NAME(ip4_ipsec_ecn), 0,
163 	"Explicit Congestion Notification handling.");
164 SYSCTL_VNET_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
165 	CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
166 	"Enable IPsec debugging output when set.");
167 SYSCTL_VNET_INT(_net_inet_ipsec, OID_AUTO, crypto_support,
168 	CTLFLAG_RW, &VNET_NAME(crypto_support), 0,
169 	"Crypto driver selection.");
170 SYSCTL_VNET_STRUCT(_net_inet_ipsec, OID_AUTO, ipsecstats,
171 	CTLFLAG_RD, &VNET_NAME(ipsec4stat), ipsecstat,
172 	"IPsec IPv4 statistics.");
173 
174 #ifdef REGRESSION
175 /*
176  * When set to 1, IPsec will send packets with the same sequence number.
177  * This allows to verify if the other side has proper replay attacks detection.
178  */
179 VNET_DEFINE(int, ipsec_replay) = 0;
180 SYSCTL_VNET_INT(_net_inet_ipsec, OID_AUTO, test_replay,
181 	CTLFLAG_RW, &VNET_NAME(ipsec_replay), 0,
182 	"Emulate replay attack");
183 /*
184  * When set 1, IPsec will send packets with corrupted HMAC.
185  * This allows to verify if the other side properly detects modified packets.
186  */
187 VNET_DEFINE(int, ipsec_integrity) = 0;
188 SYSCTL_VNET_INT(_net_inet_ipsec, OID_AUTO, test_integrity,
189 	CTLFLAG_RW, &VNET_NAME(ipsec_integrity), 0,
190 	"Emulate man-in-the-middle attack");
191 #endif
192 
193 #ifdef INET6
194 VNET_DEFINE(struct ipsecstat, ipsec6stat);
195 VNET_DEFINE(int, ip6_esp_trans_deflev) = IPSEC_LEVEL_USE;
196 VNET_DEFINE(int, ip6_esp_net_deflev) = IPSEC_LEVEL_USE;
197 VNET_DEFINE(int, ip6_ah_trans_deflev) = IPSEC_LEVEL_USE;
198 VNET_DEFINE(int, ip6_ah_net_deflev) = IPSEC_LEVEL_USE;
199 VNET_DEFINE(int, ip6_ipsec_ecn) = 0;	/* ECN ignore(-1)/forbidden(0)/allowed(1) */
200 
201 SYSCTL_DECL(_net_inet6_ipsec6);
202 
203 /* net.inet6.ipsec6 */
204 #ifdef COMPAT_KAME
205 SYSCTL_OID(_net_inet6_ipsec6, IPSECCTL_STATS, stats, CTLFLAG_RD,
206     0, 0, compat_ipsecstats_sysctl, "S", "IPsec IPv6 statistics.");
207 #endif /* COMPAT_KAME */
208 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_POLICY, def_policy, CTLFLAG_RW,
209 	&VNET_NAME(ip4_def_policy).policy, 0,
210 	"IPsec default policy.");
211 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_TRANSLEV,
212 	esp_trans_deflev, CTLFLAG_RW, &VNET_NAME(ip6_esp_trans_deflev),	0,
213 	"Default ESP transport mode level.");
214 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_ESP_NETLEV,
215 	esp_net_deflev, CTLFLAG_RW, &VNET_NAME(ip6_esp_net_deflev),	0,
216 	"Default ESP tunnel mode level.");
217 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_TRANSLEV,
218 	ah_trans_deflev, CTLFLAG_RW, &VNET_NAME(ip6_ah_trans_deflev),	0,
219 	"AH transfer mode default level.");
220 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEF_AH_NETLEV,
221 	ah_net_deflev, CTLFLAG_RW, &VNET_NAME(ip6_ah_net_deflev),	0,
222 	"AH tunnel mode default level.");
223 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_ECN,
224 	ecn, CTLFLAG_RW, &VNET_NAME(ip6_ipsec_ecn),	0,
225 	"Explicit Congestion Notification handling.");
226 SYSCTL_VNET_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug, CTLFLAG_RW,
227 	&VNET_NAME(ipsec_debug), 0,
228 	"Enable IPsec debugging output when set.");
229 SYSCTL_VNET_STRUCT(_net_inet6_ipsec6, IPSECCTL_STATS,
230 	ipsecstats, CTLFLAG_RD, &VNET_NAME(ipsec6stat), ipsecstat,
231 	"IPsec IPv6 statistics.");
232 #endif /* INET6 */
233 
234 static int ipsec_setspidx_inpcb __P((struct mbuf *, struct inpcb *));
235 static int ipsec_setspidx __P((struct mbuf *, struct secpolicyindex *, int));
236 static void ipsec4_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
237 static int ipsec4_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
238 #ifdef INET6
239 static void ipsec6_get_ulp __P((struct mbuf *m, struct secpolicyindex *, int));
240 static int ipsec6_setspidx_ipaddr __P((struct mbuf *, struct secpolicyindex *));
241 #endif
242 static void ipsec_delpcbpolicy __P((struct inpcbpolicy *));
243 static struct secpolicy *ipsec_deepcopy_policy __P((struct secpolicy *src));
244 static void vshiftl __P((unsigned char *, int, int));
245 
246 MALLOC_DEFINE(M_IPSEC_INPCB, "inpcbpolicy", "inpcb-resident ipsec policy");
247 
248 /*
249  * Return a held reference to the default SP.
250  */
251 static struct secpolicy *
252 key_allocsp_default(const char* where, int tag)
253 {
254 	struct secpolicy *sp;
255 
256 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
257 		printf("DP key_allocsp_default from %s:%u\n", where, tag));
258 
259 	sp = &V_ip4_def_policy;
260 	if (sp->policy != IPSEC_POLICY_DISCARD &&
261 	    sp->policy != IPSEC_POLICY_NONE) {
262 		ipseclog((LOG_INFO, "fixed system default policy: %d->%d\n",
263 		    sp->policy, IPSEC_POLICY_NONE));
264 		sp->policy = IPSEC_POLICY_NONE;
265 	}
266 	key_addref(sp);
267 
268 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
269 		printf("DP key_allocsp_default returns SP:%p (%u)\n",
270 			sp, sp->refcnt));
271 	return (sp);
272 }
273 #define	KEY_ALLOCSP_DEFAULT() \
274 	key_allocsp_default(__FILE__, __LINE__)
275 
276 /*
277  * For OUTBOUND packet having a socket. Searching SPD for packet,
278  * and return a pointer to SP.
279  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
280  *		0	: bypass
281  *		EACCES	: discard packet.
282  *		ENOENT	: ipsec_acquire() in progress, maybe.
283  *		others	: error occured.
284  *	others:	a pointer to SP
285  *
286  * NOTE: IPv6 mapped adddress concern is implemented here.
287  */
288 struct secpolicy *
289 ipsec_getpolicy(struct tdb_ident *tdbi, u_int dir)
290 {
291 	struct secpolicy *sp;
292 
293 	IPSEC_ASSERT(tdbi != NULL, ("null tdbi"));
294 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
295 		("invalid direction %u", dir));
296 
297 	sp = KEY_ALLOCSP2(tdbi->spi, &tdbi->dst, tdbi->proto, dir);
298 	if (sp == NULL)			/*XXX????*/
299 		sp = KEY_ALLOCSP_DEFAULT();
300 	IPSEC_ASSERT(sp != NULL, ("null SP"));
301 	return (sp);
302 }
303 
304 /*
305  * For OUTBOUND packet having a socket. Searching SPD for packet,
306  * and return a pointer to SP.
307  * OUT:	NULL:	no apropreate SP found, the following value is set to error.
308  *		0	: bypass
309  *		EACCES	: discard packet.
310  *		ENOENT	: ipsec_acquire() in progress, maybe.
311  *		others	: error occured.
312  *	others:	a pointer to SP
313  *
314  * NOTE: IPv6 mapped adddress concern is implemented here.
315  */
316 static struct secpolicy *
317 ipsec_getpolicybysock(struct mbuf *m, u_int dir, struct inpcb *inp, int *error)
318 {
319 	struct inpcbpolicy *pcbsp;
320 	struct secpolicy *currsp = NULL;	/* Policy on socket. */
321 	struct secpolicy *sp;
322 
323 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
324 	IPSEC_ASSERT(inp != NULL, ("null inpcb"));
325 	IPSEC_ASSERT(error != NULL, ("null error"));
326 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
327 		("invalid direction %u", dir));
328 
329 	/* Set spidx in pcb. */
330 	*error = ipsec_setspidx_inpcb(m, inp);
331 	if (*error)
332 		return (NULL);
333 
334 	pcbsp = inp->inp_sp;
335 	IPSEC_ASSERT(pcbsp != NULL, ("null pcbsp"));
336 	switch (dir) {
337 	case IPSEC_DIR_INBOUND:
338 		currsp = pcbsp->sp_in;
339 		break;
340 	case IPSEC_DIR_OUTBOUND:
341 		currsp = pcbsp->sp_out;
342 		break;
343 	}
344 	IPSEC_ASSERT(currsp != NULL, ("null currsp"));
345 
346 	if (pcbsp->priv) {			/* When privilieged socket. */
347 		switch (currsp->policy) {
348 		case IPSEC_POLICY_BYPASS:
349 		case IPSEC_POLICY_IPSEC:
350 			key_addref(currsp);
351 			sp = currsp;
352 			break;
353 
354 		case IPSEC_POLICY_ENTRUST:
355 			/* Look for a policy in SPD. */
356 			sp = KEY_ALLOCSP(&currsp->spidx, dir);
357 			if (sp == NULL)		/* No SP found. */
358 				sp = KEY_ALLOCSP_DEFAULT();
359 			break;
360 
361 		default:
362 			ipseclog((LOG_ERR, "%s: Invalid policy for PCB %d\n",
363 				__func__, currsp->policy));
364 			*error = EINVAL;
365 			return (NULL);
366 		}
367 	} else {				/* Unpriv, SPD has policy. */
368 		sp = KEY_ALLOCSP(&currsp->spidx, dir);
369 		if (sp == NULL) {		/* No SP found. */
370 			switch (currsp->policy) {
371 			case IPSEC_POLICY_BYPASS:
372 				ipseclog((LOG_ERR, "%s: Illegal policy for "
373 					"non-priviliged defined %d\n",
374 					__func__, currsp->policy));
375 				*error = EINVAL;
376 				return (NULL);
377 
378 			case IPSEC_POLICY_ENTRUST:
379 				sp = KEY_ALLOCSP_DEFAULT();
380 				break;
381 
382 			case IPSEC_POLICY_IPSEC:
383 				key_addref(currsp);
384 				sp = currsp;
385 				break;
386 
387 			default:
388 				ipseclog((LOG_ERR, "%s: Invalid policy for "
389 					"PCB %d\n", __func__, currsp->policy));
390 				*error = EINVAL;
391 				return (NULL);
392 			}
393 		}
394 	}
395 	IPSEC_ASSERT(sp != NULL,
396 		("null SP (priv %u policy %u", pcbsp->priv, currsp->policy));
397 	KEYDEBUG(KEYDEBUG_IPSEC_STAMP,
398 		printf("DP %s (priv %u policy %u) allocate SP:%p (refcnt %u)\n",
399 			__func__, pcbsp->priv, currsp->policy, sp, sp->refcnt));
400 	return (sp);
401 }
402 
403 /*
404  * For FORWADING packet or OUTBOUND without a socket. Searching SPD for packet,
405  * and return a pointer to SP.
406  * OUT:	positive: a pointer to the entry for security policy leaf matched.
407  *	NULL:	no apropreate SP found, the following value is set to error.
408  *		0	: bypass
409  *		EACCES	: discard packet.
410  *		ENOENT	: ipsec_acquire() in progress, maybe.
411  *		others	: error occured.
412  */
413 struct secpolicy *
414 ipsec_getpolicybyaddr(struct mbuf *m, u_int dir, int flag, int *error)
415 {
416 	struct secpolicyindex spidx;
417 	struct secpolicy *sp;
418 
419 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
420 	IPSEC_ASSERT(error != NULL, ("null error"));
421 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
422 		("invalid direction %u", dir));
423 
424 	sp = NULL;
425 	if (key_havesp(dir)) {
426 		/* Make an index to look for a policy. */
427 		*error = ipsec_setspidx(m, &spidx,
428 					(flag & IP_FORWARDING) ? 0 : 1);
429 		if (*error != 0) {
430 			DPRINTF(("%s: setpidx failed, dir %u flag %u\n",
431 				__func__, dir, flag));
432 			return (NULL);
433 		}
434 		spidx.dir = dir;
435 
436 		sp = KEY_ALLOCSP(&spidx, dir);
437 	}
438 	if (sp == NULL)			/* No SP found, use system default. */
439 		sp = KEY_ALLOCSP_DEFAULT();
440 	IPSEC_ASSERT(sp != NULL, ("null SP"));
441 	return (sp);
442 }
443 
444 struct secpolicy *
445 ipsec4_checkpolicy(struct mbuf *m, u_int dir, u_int flag, int *error,
446     struct inpcb *inp)
447 {
448 	struct secpolicy *sp;
449 
450 	*error = 0;
451 	if (inp == NULL)
452 		sp = ipsec_getpolicybyaddr(m, dir, flag, error);
453 	else
454 		sp = ipsec_getpolicybysock(m, dir, inp, error);
455 	if (sp == NULL) {
456 		IPSEC_ASSERT(*error != 0, ("getpolicy failed w/o error"));
457 		V_ipsec4stat.ips_out_inval++;
458 		return (NULL);
459 	}
460 	IPSEC_ASSERT(*error == 0, ("sp w/ error set to %u", *error));
461 	switch (sp->policy) {
462 	case IPSEC_POLICY_ENTRUST:
463 	default:
464 		printf("%s: invalid policy %u\n", __func__, sp->policy);
465 		/* FALLTHROUGH */
466 	case IPSEC_POLICY_DISCARD:
467 		V_ipsec4stat.ips_out_polvio++;
468 		*error = -EINVAL;	/* Packet is discarded by caller. */
469 		break;
470 	case IPSEC_POLICY_BYPASS:
471 	case IPSEC_POLICY_NONE:
472 		KEY_FREESP(&sp);
473 		sp = NULL;		/* NB: force NULL result. */
474 		break;
475 	case IPSEC_POLICY_IPSEC:
476 		if (sp->req == NULL)	/* Acquire a SA. */
477 			*error = key_spdacquire(sp);
478 		break;
479 	}
480 	if (*error != 0) {
481 		KEY_FREESP(&sp);
482 		sp = NULL;
483 	}
484 	return (sp);
485 }
486 
487 static int
488 ipsec_setspidx_inpcb(struct mbuf *m, struct inpcb *inp)
489 {
490 	int error;
491 
492 	IPSEC_ASSERT(inp != NULL, ("null inp"));
493 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
494 	IPSEC_ASSERT(inp->inp_sp->sp_out != NULL && inp->inp_sp->sp_in != NULL,
495 		("null sp_in || sp_out"));
496 
497 	error = ipsec_setspidx(m, &inp->inp_sp->sp_in->spidx, 1);
498 	if (error == 0) {
499 		inp->inp_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
500 		inp->inp_sp->sp_out->spidx = inp->inp_sp->sp_in->spidx;
501 		inp->inp_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
502 	} else {
503 		bzero(&inp->inp_sp->sp_in->spidx,
504 			sizeof (inp->inp_sp->sp_in->spidx));
505 		bzero(&inp->inp_sp->sp_out->spidx,
506 			sizeof (inp->inp_sp->sp_in->spidx));
507 	}
508 	return (error);
509 }
510 
511 /*
512  * Configure security policy index (src/dst/proto/sport/dport)
513  * by looking at the content of mbuf.
514  * The caller is responsible for error recovery (like clearing up spidx).
515  */
516 static int
517 ipsec_setspidx(struct mbuf *m, struct secpolicyindex *spidx, int needport)
518 {
519 	struct ip *ip = NULL;
520 	struct ip ipbuf;
521 	u_int v;
522 	struct mbuf *n;
523 	int len;
524 	int error;
525 
526 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
527 
528 	/*
529 	 * Validate m->m_pkthdr.len.  We see incorrect length if we
530 	 * mistakenly call this function with inconsistent mbuf chain
531 	 * (like 4.4BSD tcp/udp processing).  XXX Should we panic here?
532 	 */
533 	len = 0;
534 	for (n = m; n; n = n->m_next)
535 		len += n->m_len;
536 	if (m->m_pkthdr.len != len) {
537 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
538 			printf("%s: pkthdr len(%d) mismatch (%d), ignored.\n",
539 				__func__, len, m->m_pkthdr.len));
540 		return (EINVAL);
541 	}
542 
543 	if (m->m_pkthdr.len < sizeof(struct ip)) {
544 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
545 			printf("%s: pkthdr len(%d) too small (v4), ignored.\n",
546 			    __func__, m->m_pkthdr.len));
547 		return (EINVAL);
548 	}
549 
550 	if (m->m_len >= sizeof(*ip))
551 		ip = mtod(m, struct ip *);
552 	else {
553 		m_copydata(m, 0, sizeof(ipbuf), (caddr_t)&ipbuf);
554 		ip = &ipbuf;
555 	}
556 #ifdef _IP_VHL
557 	v = _IP_VHL_V(ip->ip_vhl);
558 #else
559 	v = ip->ip_v;
560 #endif
561 	switch (v) {
562 	case 4:
563 		error = ipsec4_setspidx_ipaddr(m, spidx);
564 		if (error)
565 			return (error);
566 		ipsec4_get_ulp(m, spidx, needport);
567 		return (0);
568 #ifdef INET6
569 	case 6:
570 		if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) {
571 			KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
572 				printf("%s: pkthdr len(%d) too small (v6), "
573 				"ignored\n", __func__, m->m_pkthdr.len));
574 			return (EINVAL);
575 		}
576 		error = ipsec6_setspidx_ipaddr(m, spidx);
577 		if (error)
578 			return (error);
579 		ipsec6_get_ulp(m, spidx, needport);
580 		return (0);
581 #endif
582 	default:
583 		KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
584 			printf("%s: " "unknown IP version %u, ignored.\n",
585 				__func__, v));
586 		return (EINVAL);
587 	}
588 }
589 
590 static void
591 ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
592 {
593 	u_int8_t nxt;
594 	int off;
595 
596 	/* Sanity check. */
597 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
598 	IPSEC_ASSERT(m->m_pkthdr.len >= sizeof(struct ip),("packet too short"));
599 
600 	/* NB: ip_input() flips it into host endian. XXX Need more checking. */
601 	if (m->m_len >= sizeof (struct ip)) {
602 		struct ip *ip = mtod(m, struct ip *);
603 		if (ip->ip_off & (IP_MF | IP_OFFMASK))
604 			goto done;
605 #ifdef _IP_VHL
606 		off = _IP_VHL_HL(ip->ip_vhl) << 2;
607 #else
608 		off = ip->ip_hl << 2;
609 #endif
610 		nxt = ip->ip_p;
611 	} else {
612 		struct ip ih;
613 
614 		m_copydata(m, 0, sizeof (struct ip), (caddr_t) &ih);
615 		if (ih.ip_off & (IP_MF | IP_OFFMASK))
616 			goto done;
617 #ifdef _IP_VHL
618 		off = _IP_VHL_HL(ih.ip_vhl) << 2;
619 #else
620 		off = ih.ip_hl << 2;
621 #endif
622 		nxt = ih.ip_p;
623 	}
624 
625 	while (off < m->m_pkthdr.len) {
626 		struct ip6_ext ip6e;
627 		struct tcphdr th;
628 		struct udphdr uh;
629 
630 		switch (nxt) {
631 		case IPPROTO_TCP:
632 			spidx->ul_proto = nxt;
633 			if (!needport)
634 				goto done_proto;
635 			if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
636 				goto done;
637 			m_copydata(m, off, sizeof (th), (caddr_t) &th);
638 			spidx->src.sin.sin_port = th.th_sport;
639 			spidx->dst.sin.sin_port = th.th_dport;
640 			return;
641 		case IPPROTO_UDP:
642 			spidx->ul_proto = nxt;
643 			if (!needport)
644 				goto done_proto;
645 			if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
646 				goto done;
647 			m_copydata(m, off, sizeof (uh), (caddr_t) &uh);
648 			spidx->src.sin.sin_port = uh.uh_sport;
649 			spidx->dst.sin.sin_port = uh.uh_dport;
650 			return;
651 		case IPPROTO_AH:
652 			if (off + sizeof(ip6e) > m->m_pkthdr.len)
653 				goto done;
654 			/* XXX Sigh, this works but is totally bogus. */
655 			m_copydata(m, off, sizeof(ip6e), (caddr_t) &ip6e);
656 			off += (ip6e.ip6e_len + 2) << 2;
657 			nxt = ip6e.ip6e_nxt;
658 			break;
659 		case IPPROTO_ICMP:
660 		default:
661 			/* XXX Intermediate headers??? */
662 			spidx->ul_proto = nxt;
663 			goto done_proto;
664 		}
665 	}
666 done:
667 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
668 done_proto:
669 	spidx->src.sin.sin_port = IPSEC_PORT_ANY;
670 	spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
671 }
672 
673 /* Assumes that m is sane. */
674 static int
675 ipsec4_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
676 {
677 	static const struct sockaddr_in template = {
678 		sizeof (struct sockaddr_in),
679 		AF_INET,
680 		0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 }
681 	};
682 
683 	spidx->src.sin = template;
684 	spidx->dst.sin = template;
685 
686 	if (m->m_len < sizeof (struct ip)) {
687 		m_copydata(m, offsetof(struct ip, ip_src),
688 			   sizeof (struct  in_addr),
689 			   (caddr_t) &spidx->src.sin.sin_addr);
690 		m_copydata(m, offsetof(struct ip, ip_dst),
691 			   sizeof (struct  in_addr),
692 			   (caddr_t) &spidx->dst.sin.sin_addr);
693 	} else {
694 		struct ip *ip = mtod(m, struct ip *);
695 		spidx->src.sin.sin_addr = ip->ip_src;
696 		spidx->dst.sin.sin_addr = ip->ip_dst;
697 	}
698 
699 	spidx->prefs = sizeof(struct in_addr) << 3;
700 	spidx->prefd = sizeof(struct in_addr) << 3;
701 
702 	return (0);
703 }
704 
705 #ifdef INET6
706 static void
707 ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
708 {
709 	int off, nxt;
710 	struct tcphdr th;
711 	struct udphdr uh;
712 	struct icmp6_hdr ih;
713 
714 	/* Sanity check. */
715 	if (m == NULL)
716 		panic("%s: NULL pointer was passed.\n", __func__);
717 
718 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
719 		printf("%s:\n", __func__); kdebug_mbuf(m));
720 
721 	/* Set default. */
722 	spidx->ul_proto = IPSEC_ULPROTO_ANY;
723 	((struct sockaddr_in6 *)&spidx->src)->sin6_port = IPSEC_PORT_ANY;
724 	((struct sockaddr_in6 *)&spidx->dst)->sin6_port = IPSEC_PORT_ANY;
725 
726 	nxt = -1;
727 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
728 	if (off < 0 || m->m_pkthdr.len < off)
729 		return;
730 
731 	switch (nxt) {
732 	case IPPROTO_TCP:
733 		spidx->ul_proto = nxt;
734 		if (!needport)
735 			break;
736 		if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
737 			break;
738 		m_copydata(m, off, sizeof(th), (caddr_t)&th);
739 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = th.th_sport;
740 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = th.th_dport;
741 		break;
742 	case IPPROTO_UDP:
743 		spidx->ul_proto = nxt;
744 		if (!needport)
745 			break;
746 		if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
747 			break;
748 		m_copydata(m, off, sizeof(uh), (caddr_t)&uh);
749 		((struct sockaddr_in6 *)&spidx->src)->sin6_port = uh.uh_sport;
750 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port = uh.uh_dport;
751 		break;
752 	case IPPROTO_ICMPV6:
753 		spidx->ul_proto = nxt;
754 		if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len)
755 			break;
756 		m_copydata(m, off, sizeof(ih), (caddr_t)&ih);
757 		((struct sockaddr_in6 *)&spidx->src)->sin6_port =
758 		    htons((uint16_t)ih.icmp6_type);
759 		((struct sockaddr_in6 *)&spidx->dst)->sin6_port =
760 		    htons((uint16_t)ih.icmp6_code);
761 		break;
762 	default:
763 		/* XXX Intermediate headers??? */
764 		spidx->ul_proto = nxt;
765 		break;
766 	}
767 }
768 
769 /* Assumes that m is sane. */
770 static int
771 ipsec6_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
772 {
773 	struct ip6_hdr *ip6 = NULL;
774 	struct ip6_hdr ip6buf;
775 	struct sockaddr_in6 *sin6;
776 
777 	if (m->m_len >= sizeof(*ip6))
778 		ip6 = mtod(m, struct ip6_hdr *);
779 	else {
780 		m_copydata(m, 0, sizeof(ip6buf), (caddr_t)&ip6buf);
781 		ip6 = &ip6buf;
782 	}
783 
784 	sin6 = (struct sockaddr_in6 *)&spidx->src;
785 	bzero(sin6, sizeof(*sin6));
786 	sin6->sin6_family = AF_INET6;
787 	sin6->sin6_len = sizeof(struct sockaddr_in6);
788 	bcopy(&ip6->ip6_src, &sin6->sin6_addr, sizeof(ip6->ip6_src));
789 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
790 		sin6->sin6_addr.s6_addr16[1] = 0;
791 		sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]);
792 	}
793 	spidx->prefs = sizeof(struct in6_addr) << 3;
794 
795 	sin6 = (struct sockaddr_in6 *)&spidx->dst;
796 	bzero(sin6, sizeof(*sin6));
797 	sin6->sin6_family = AF_INET6;
798 	sin6->sin6_len = sizeof(struct sockaddr_in6);
799 	bcopy(&ip6->ip6_dst, &sin6->sin6_addr, sizeof(ip6->ip6_dst));
800 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
801 		sin6->sin6_addr.s6_addr16[1] = 0;
802 		sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]);
803 	}
804 	spidx->prefd = sizeof(struct in6_addr) << 3;
805 
806 	return (0);
807 }
808 #endif
809 
810 static void
811 ipsec_delpcbpolicy(struct inpcbpolicy *p)
812 {
813 
814 	free(p, M_IPSEC_INPCB);
815 }
816 
817 /* Initialize policy in PCB. */
818 int
819 ipsec_init_policy(struct socket *so, struct inpcbpolicy **pcb_sp)
820 {
821 	struct inpcbpolicy *new;
822 
823 	/* Sanity check. */
824 	if (so == NULL || pcb_sp == NULL)
825 		panic("%s: NULL pointer was passed.\n", __func__);
826 
827 	new = (struct inpcbpolicy *) malloc(sizeof(struct inpcbpolicy),
828 					    M_IPSEC_INPCB, M_NOWAIT|M_ZERO);
829 	if (new == NULL) {
830 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
831 		return (ENOBUFS);
832 	}
833 
834 	new->priv = IPSEC_IS_PRIVILEGED_SO(so);
835 
836 	if ((new->sp_in = KEY_NEWSP()) == NULL) {
837 		ipsec_delpcbpolicy(new);
838 		return (ENOBUFS);
839 	}
840 	new->sp_in->state = IPSEC_SPSTATE_ALIVE;
841 	new->sp_in->policy = IPSEC_POLICY_ENTRUST;
842 
843 	if ((new->sp_out = KEY_NEWSP()) == NULL) {
844 		KEY_FREESP(&new->sp_in);
845 		ipsec_delpcbpolicy(new);
846 		return (ENOBUFS);
847 	}
848 	new->sp_out->state = IPSEC_SPSTATE_ALIVE;
849 	new->sp_out->policy = IPSEC_POLICY_ENTRUST;
850 
851 	*pcb_sp = new;
852 
853 	return (0);
854 }
855 
856 /* Copy old IPsec policy into new. */
857 int
858 ipsec_copy_policy(struct inpcbpolicy *old, struct inpcbpolicy *new)
859 {
860 	struct secpolicy *sp;
861 
862 	sp = ipsec_deepcopy_policy(old->sp_in);
863 	if (sp) {
864 		KEY_FREESP(&new->sp_in);
865 		new->sp_in = sp;
866 	} else
867 		return (ENOBUFS);
868 
869 	sp = ipsec_deepcopy_policy(old->sp_out);
870 	if (sp) {
871 		KEY_FREESP(&new->sp_out);
872 		new->sp_out = sp;
873 	} else
874 		return (ENOBUFS);
875 
876 	new->priv = old->priv;
877 
878 	return (0);
879 }
880 
881 struct ipsecrequest *
882 ipsec_newisr(void)
883 {
884 	struct ipsecrequest *p;
885 
886 	p = malloc(sizeof(struct ipsecrequest), M_IPSEC_SR, M_NOWAIT|M_ZERO);
887 	if (p != NULL)
888 		IPSECREQUEST_LOCK_INIT(p);
889 	return (p);
890 }
891 
892 void
893 ipsec_delisr(struct ipsecrequest *p)
894 {
895 
896 	IPSECREQUEST_LOCK_DESTROY(p);
897 	free(p, M_IPSEC_SR);
898 }
899 
900 /* Deep-copy a policy in PCB. */
901 static struct secpolicy *
902 ipsec_deepcopy_policy(struct secpolicy *src)
903 {
904 	struct ipsecrequest *newchain = NULL;
905 	struct ipsecrequest *p;
906 	struct ipsecrequest **q;
907 	struct ipsecrequest *r;
908 	struct secpolicy *dst;
909 
910 	if (src == NULL)
911 		return (NULL);
912 	dst = KEY_NEWSP();
913 	if (dst == NULL)
914 		return (NULL);
915 
916 	/*
917 	 * Deep-copy IPsec request chain.  This is required since struct
918 	 * ipsecrequest is not reference counted.
919 	 */
920 	q = &newchain;
921 	for (p = src->req; p; p = p->next) {
922 		*q = ipsec_newisr();
923 		if (*q == NULL)
924 			goto fail;
925 		(*q)->saidx.proto = p->saidx.proto;
926 		(*q)->saidx.mode = p->saidx.mode;
927 		(*q)->level = p->level;
928 		(*q)->saidx.reqid = p->saidx.reqid;
929 
930 		bcopy(&p->saidx.src, &(*q)->saidx.src, sizeof((*q)->saidx.src));
931 		bcopy(&p->saidx.dst, &(*q)->saidx.dst, sizeof((*q)->saidx.dst));
932 
933 		(*q)->sp = dst;
934 
935 		q = &((*q)->next);
936 	}
937 
938 	dst->req = newchain;
939 	dst->state = src->state;
940 	dst->policy = src->policy;
941 	/* Do not touch the refcnt fields. */
942 
943 	return (dst);
944 
945 fail:
946 	for (p = newchain; p; p = r) {
947 		r = p->next;
948 		ipsec_delisr(p);
949 		p = NULL;
950 	}
951 	return (NULL);
952 }
953 
954 /* Set policy and IPsec request if present. */
955 static int
956 ipsec_set_policy_internal(struct secpolicy **pcb_sp, int optname,
957     caddr_t request, size_t len, struct ucred *cred)
958 {
959 	struct sadb_x_policy *xpl;
960 	struct secpolicy *newsp = NULL;
961 	int error;
962 
963 	/* Sanity check. */
964 	if (pcb_sp == NULL || *pcb_sp == NULL || request == NULL)
965 		return (EINVAL);
966 	if (len < sizeof(*xpl))
967 		return (EINVAL);
968 	xpl = (struct sadb_x_policy *)request;
969 
970 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
971 		printf("%s: passed policy\n", __func__);
972 		kdebug_sadb_x_policy((struct sadb_ext *)xpl));
973 
974 	/* Check policy type. */
975 	/* ipsec_set_policy_internal() accepts IPSEC, ENTRUST and BYPASS. */
976 	if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD
977 	 || xpl->sadb_x_policy_type == IPSEC_POLICY_NONE)
978 		return (EINVAL);
979 
980 	/* Check privileged socket. */
981 	if (cred != NULL && xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
982 		error = priv_check_cred(cred, PRIV_NETINET_IPSEC, 0);
983 		if (error)
984 			return (EACCES);
985 	}
986 
987 	/* Allocating new SP entry. */
988 	if ((newsp = key_msg2sp(xpl, len, &error)) == NULL)
989 		return (error);
990 
991 	newsp->state = IPSEC_SPSTATE_ALIVE;
992 
993 	/* Clear old SP and set new SP. */
994 	KEY_FREESP(pcb_sp);
995 	*pcb_sp = newsp;
996 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
997 		printf("%s: new policy\n", __func__);
998 		kdebug_secpolicy(newsp));
999 
1000 	return (0);
1001 }
1002 
1003 int
1004 ipsec_set_policy(struct inpcb *inp, int optname, caddr_t request,
1005     size_t len, struct ucred *cred)
1006 {
1007 	struct sadb_x_policy *xpl;
1008 	struct secpolicy **pcb_sp;
1009 
1010 	/* Sanity check. */
1011 	if (inp == NULL || request == NULL)
1012 		return (EINVAL);
1013 	if (len < sizeof(*xpl))
1014 		return (EINVAL);
1015 	xpl = (struct sadb_x_policy *)request;
1016 
1017 	/* Select direction. */
1018 	switch (xpl->sadb_x_policy_dir) {
1019 	case IPSEC_DIR_INBOUND:
1020 		pcb_sp = &inp->inp_sp->sp_in;
1021 		break;
1022 	case IPSEC_DIR_OUTBOUND:
1023 		pcb_sp = &inp->inp_sp->sp_out;
1024 		break;
1025 	default:
1026 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1027 			xpl->sadb_x_policy_dir));
1028 		return (EINVAL);
1029 	}
1030 
1031 	return (ipsec_set_policy_internal(pcb_sp, optname, request, len, cred));
1032 }
1033 
1034 int
1035 ipsec_get_policy(struct inpcb *inp, caddr_t request, size_t len,
1036     struct mbuf **mp)
1037 {
1038 	struct sadb_x_policy *xpl;
1039 	struct secpolicy *pcb_sp;
1040 
1041 	/* Sanity check. */
1042 	if (inp == NULL || request == NULL || mp == NULL)
1043 		return (EINVAL);
1044 	IPSEC_ASSERT(inp->inp_sp != NULL, ("null inp_sp"));
1045 	if (len < sizeof(*xpl))
1046 		return (EINVAL);
1047 	xpl = (struct sadb_x_policy *)request;
1048 
1049 	/* Select direction. */
1050 	switch (xpl->sadb_x_policy_dir) {
1051 	case IPSEC_DIR_INBOUND:
1052 		pcb_sp = inp->inp_sp->sp_in;
1053 		break;
1054 	case IPSEC_DIR_OUTBOUND:
1055 		pcb_sp = inp->inp_sp->sp_out;
1056 		break;
1057 	default:
1058 		ipseclog((LOG_ERR, "%s: invalid direction=%u\n", __func__,
1059 			xpl->sadb_x_policy_dir));
1060 		return (EINVAL);
1061 	}
1062 
1063 	/* Sanity check. Should be an IPSEC_ASSERT. */
1064 	if (pcb_sp == NULL)
1065 		return (EINVAL);
1066 
1067 	*mp = key_sp2msg(pcb_sp);
1068 	if (!*mp) {
1069 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
1070 		return (ENOBUFS);
1071 	}
1072 
1073 	(*mp)->m_type = MT_DATA;
1074 	KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1075 		printf("%s:\n", __func__); kdebug_mbuf(*mp));
1076 
1077 	return (0);
1078 }
1079 
1080 /* Delete policy in PCB. */
1081 int
1082 ipsec_delete_pcbpolicy(struct inpcb *inp)
1083 {
1084 	IPSEC_ASSERT(inp != NULL, ("null inp"));
1085 
1086 	if (inp->inp_sp == NULL)
1087 		return (0);
1088 
1089 	if (inp->inp_sp->sp_in != NULL)
1090 		KEY_FREESP(&inp->inp_sp->sp_in);
1091 
1092 	if (inp->inp_sp->sp_out != NULL)
1093 		KEY_FREESP(&inp->inp_sp->sp_out);
1094 
1095 	ipsec_delpcbpolicy(inp->inp_sp);
1096 	inp->inp_sp = NULL;
1097 
1098 	return (0);
1099 }
1100 
1101 /*
1102  * Return current level.
1103  * Either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE are always returned.
1104  */
1105 u_int
1106 ipsec_get_reqlevel(struct ipsecrequest *isr)
1107 {
1108 	u_int level = 0;
1109 	u_int esp_trans_deflev, esp_net_deflev;
1110 	u_int ah_trans_deflev, ah_net_deflev;
1111 
1112 	IPSEC_ASSERT(isr != NULL && isr->sp != NULL, ("null argument"));
1113 	IPSEC_ASSERT(isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family,
1114 		("af family mismatch, src %u, dst %u",
1115 		 isr->sp->spidx.src.sa.sa_family,
1116 		 isr->sp->spidx.dst.sa.sa_family));
1117 
1118 /* XXX Note that we have ipseclog() expanded here - code sync issue. */
1119 #define IPSEC_CHECK_DEFAULT(lev) \
1120 	(((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE	      \
1121 			&& (lev) != IPSEC_LEVEL_UNIQUE)			      \
1122 		? (V_ipsec_debug						      \
1123 			? log(LOG_INFO, "fixed system default level " #lev ":%d->%d\n",\
1124 				(lev), IPSEC_LEVEL_REQUIRE)		      \
1125 			: 0),						      \
1126 			(lev) = IPSEC_LEVEL_REQUIRE,			      \
1127 			(lev)						      \
1128 		: (lev))
1129 
1130 	/* Set default level. */
1131 	switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) {
1132 #ifdef INET
1133 	case AF_INET:
1134 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_trans_deflev);
1135 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_esp_net_deflev);
1136 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_trans_deflev);
1137 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip4_ah_net_deflev);
1138 		break;
1139 #endif
1140 #ifdef INET6
1141 	case AF_INET6:
1142 		esp_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_trans_deflev);
1143 		esp_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_esp_net_deflev);
1144 		ah_trans_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_trans_deflev);
1145 		ah_net_deflev = IPSEC_CHECK_DEFAULT(V_ip6_ah_net_deflev);
1146 		break;
1147 #endif /* INET6 */
1148 	default:
1149 		panic("%s: unknown af %u",
1150 			__func__, isr->sp->spidx.src.sa.sa_family);
1151 	}
1152 
1153 #undef IPSEC_CHECK_DEFAULT
1154 
1155 	/* Set level. */
1156 	switch (isr->level) {
1157 	case IPSEC_LEVEL_DEFAULT:
1158 		switch (isr->saidx.proto) {
1159 		case IPPROTO_ESP:
1160 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1161 				level = esp_net_deflev;
1162 			else
1163 				level = esp_trans_deflev;
1164 			break;
1165 		case IPPROTO_AH:
1166 			if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1167 				level = ah_net_deflev;
1168 			else
1169 				level = ah_trans_deflev;
1170 			break;
1171 		case IPPROTO_IPCOMP:
1172 			/*
1173 			 * We don't really care, as IPcomp document says that
1174 			 * we shouldn't compress small packets.
1175 			 */
1176 			level = IPSEC_LEVEL_USE;
1177 			break;
1178 		default:
1179 			panic("%s: Illegal protocol defined %u\n", __func__,
1180 				isr->saidx.proto);
1181 		}
1182 		break;
1183 
1184 	case IPSEC_LEVEL_USE:
1185 	case IPSEC_LEVEL_REQUIRE:
1186 		level = isr->level;
1187 		break;
1188 	case IPSEC_LEVEL_UNIQUE:
1189 		level = IPSEC_LEVEL_REQUIRE;
1190 		break;
1191 
1192 	default:
1193 		panic("%s: Illegal IPsec level %u\n", __func__, isr->level);
1194 	}
1195 
1196 	return (level);
1197 }
1198 
1199 /*
1200  * Check security policy requirements against the actual
1201  * packet contents.  Return one if the packet should be
1202  * reject as "invalid"; otherwiser return zero to have the
1203  * packet treated as "valid".
1204  *
1205  * OUT:
1206  *	0: valid
1207  *	1: invalid
1208  */
1209 int
1210 ipsec_in_reject(struct secpolicy *sp, struct mbuf *m)
1211 {
1212 	struct ipsecrequest *isr;
1213 	int need_auth;
1214 
1215 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1216 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1217 
1218 	/* Check policy. */
1219 	switch (sp->policy) {
1220 	case IPSEC_POLICY_DISCARD:
1221 		return (1);
1222 	case IPSEC_POLICY_BYPASS:
1223 	case IPSEC_POLICY_NONE:
1224 		return (0);
1225 	}
1226 
1227 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1228 		("invalid policy %u", sp->policy));
1229 
1230 	/* XXX Should compare policy against IPsec header history. */
1231 
1232 	need_auth = 0;
1233 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1234 		if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE)
1235 			continue;
1236 		switch (isr->saidx.proto) {
1237 		case IPPROTO_ESP:
1238 			if ((m->m_flags & M_DECRYPTED) == 0) {
1239 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1240 				    printf("%s: ESP m_flags:%x\n", __func__,
1241 					    m->m_flags));
1242 				return (1);
1243 			}
1244 
1245 			if (!need_auth &&
1246 			    isr->sav != NULL &&
1247 			    isr->sav->tdb_authalgxform != NULL &&
1248 			    (m->m_flags & M_AUTHIPDGM) == 0) {
1249 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1250 				    printf("%s: ESP/AH m_flags:%x\n", __func__,
1251 					    m->m_flags));
1252 				return (1);
1253 			}
1254 			break;
1255 		case IPPROTO_AH:
1256 			need_auth = 1;
1257 			if ((m->m_flags & M_AUTHIPHDR) == 0) {
1258 				KEYDEBUG(KEYDEBUG_IPSEC_DUMP,
1259 				    printf("%s: AH m_flags:%x\n", __func__,
1260 					    m->m_flags));
1261 				return (1);
1262 			}
1263 			break;
1264 		case IPPROTO_IPCOMP:
1265 			/*
1266 			 * We don't really care, as IPcomp document
1267 			 * says that we shouldn't compress small
1268 			 * packets.  IPComp policy should always be
1269 			 * treated as being in "use" level.
1270 			 */
1271 			break;
1272 		}
1273 	}
1274 	return (0);		/* Valid. */
1275 }
1276 
1277 static int
1278 ipsec46_in_reject(struct mbuf *m, struct inpcb *inp)
1279 {
1280 	struct secpolicy *sp;
1281 	int error;
1282 	int result;
1283 
1284 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1285 
1286 	/*
1287 	 * Get SP for this packet.
1288 	 * When we are called from ip_forward(), we call
1289 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1290 	 */
1291 	if (inp == NULL)
1292 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND, IP_FORWARDING, &error);
1293 	else
1294 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND, inp, &error);
1295 
1296 	if (sp != NULL) {
1297 		result = ipsec_in_reject(sp, m);
1298 		KEY_FREESP(&sp);
1299 	} else {
1300 		result = 0;	/* XXX Should be panic?
1301 				 * -> No, there may be error. */
1302 	}
1303 	return (result);
1304 }
1305 
1306 /*
1307  * Check AH/ESP integrity.
1308  * This function is called from tcp_input(), udp_input(),
1309  * and {ah,esp}4_input for tunnel mode.
1310  */
1311 int
1312 ipsec4_in_reject(struct mbuf *m, struct inpcb *inp)
1313 {
1314 	int result;
1315 
1316 	result = ipsec46_in_reject(m, inp);
1317 	if (result)
1318 		V_ipsec4stat.ips_in_polvio++;
1319 
1320 	return (result);
1321 }
1322 
1323 #ifdef INET6
1324 /*
1325  * Check AH/ESP integrity.
1326  * This function is called from tcp6_input(), udp6_input(),
1327  * and {ah,esp}6_input for tunnel mode.
1328  */
1329 int
1330 ipsec6_in_reject(struct mbuf *m, struct inpcb *inp)
1331 {
1332 	int result;
1333 
1334 	result = ipsec46_in_reject(m, inp);
1335 	if (result)
1336 		V_ipsec6stat.ips_in_polvio++;
1337 
1338 	return (result);
1339 }
1340 #endif
1341 
1342 /*
1343  * Compute the byte size to be occupied by IPsec header.
1344  * In case it is tunnelled, it includes the size of outer IP header.
1345  * NOTE: SP passed is freed in this function.
1346  */
1347 static size_t
1348 ipsec_hdrsiz_internal(struct secpolicy *sp)
1349 {
1350 	struct ipsecrequest *isr;
1351 	size_t size;
1352 
1353 	KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1354 		printf("%s: using SP\n", __func__); kdebug_secpolicy(sp));
1355 
1356 	switch (sp->policy) {
1357 	case IPSEC_POLICY_DISCARD:
1358 	case IPSEC_POLICY_BYPASS:
1359 	case IPSEC_POLICY_NONE:
1360 		return (0);
1361 	}
1362 
1363 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
1364 		("invalid policy %u", sp->policy));
1365 
1366 	size = 0;
1367 	for (isr = sp->req; isr != NULL; isr = isr->next) {
1368 		size_t clen = 0;
1369 
1370 		switch (isr->saidx.proto) {
1371 		case IPPROTO_ESP:
1372 			clen = esp_hdrsiz(isr->sav);
1373 			break;
1374 		case IPPROTO_AH:
1375 			clen = ah_hdrsiz(isr->sav);
1376 			break;
1377 		case IPPROTO_IPCOMP:
1378 			clen = sizeof(struct ipcomp);
1379 			break;
1380 		}
1381 
1382 		if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1383 			switch (isr->saidx.dst.sa.sa_family) {
1384 			case AF_INET:
1385 				clen += sizeof(struct ip);
1386 				break;
1387 #ifdef INET6
1388 			case AF_INET6:
1389 				clen += sizeof(struct ip6_hdr);
1390 				break;
1391 #endif
1392 			default:
1393 				ipseclog((LOG_ERR, "%s: unknown AF %d in "
1394 				    "IPsec tunnel SA\n", __func__,
1395 				    ((struct sockaddr *)&isr->saidx.dst)->sa_family));
1396 				break;
1397 			}
1398 		}
1399 		size += clen;
1400 	}
1401 
1402 	return (size);
1403 }
1404 
1405 /*
1406  * This function is called from ipsec_hdrsiz_tcp(), ip_ipsec_mtu(),
1407  * disabled ip6_ipsec_mtu() and ip6_forward().
1408  */
1409 size_t
1410 ipsec_hdrsiz(struct mbuf *m, u_int dir, struct inpcb *inp)
1411 {
1412 	struct secpolicy *sp;
1413 	int error;
1414 	size_t size;
1415 
1416 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1417 
1418 	/* Get SP for this packet.
1419 	 * When we are called from ip_forward(), we call
1420 	 * ipsec_getpolicybyaddr() with IP_FORWARDING flag.
1421 	 */
1422 	if (inp == NULL)
1423 		sp = ipsec_getpolicybyaddr(m, dir, IP_FORWARDING, &error);
1424 	else
1425 		sp = ipsec_getpolicybysock(m, dir, inp, &error);
1426 
1427 	if (sp != NULL) {
1428 		size = ipsec_hdrsiz_internal(sp);
1429 		KEYDEBUG(KEYDEBUG_IPSEC_DATA,
1430 			printf("%s: size:%lu.\n", __func__,
1431 				(unsigned long)size));
1432 
1433 		KEY_FREESP(&sp);
1434 	} else {
1435 		size = 0;	/* XXX Should be panic?
1436 				 * -> No, we are called w/o knowing if
1437 				 *    IPsec processing is needed. */
1438 	}
1439 	return (size);
1440 }
1441 
1442 /*
1443  * Check the variable replay window.
1444  * ipsec_chkreplay() performs replay check before ICV verification.
1445  * ipsec_updatereplay() updates replay bitmap.  This must be called after
1446  * ICV verification (it also performs replay check, which is usually done
1447  * beforehand).
1448  * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1449  *
1450  * Based on RFC 2401.
1451  */
1452 int
1453 ipsec_chkreplay(u_int32_t seq, struct secasvar *sav)
1454 {
1455 	const struct secreplay *replay;
1456 	u_int32_t diff;
1457 	int fr;
1458 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1459 	int frlast;		/* Constant: last frame. */
1460 
1461 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1462 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1463 
1464 	replay = sav->replay;
1465 
1466 	if (replay->wsize == 0)
1467 		return (1);	/* No need to check replay. */
1468 
1469 	/* Constant. */
1470 	frlast = replay->wsize - 1;
1471 	wsizeb = replay->wsize << 3;
1472 
1473 	/* Sequence number of 0 is invalid. */
1474 	if (seq == 0)
1475 		return (0);
1476 
1477 	/* First time is always okay. */
1478 	if (replay->count == 0)
1479 		return (1);
1480 
1481 	if (seq > replay->lastseq) {
1482 		/* Larger sequences are okay. */
1483 		return (1);
1484 	} else {
1485 		/* seq is equal or less than lastseq. */
1486 		diff = replay->lastseq - seq;
1487 
1488 		/* Over range to check, i.e. too old or wrapped. */
1489 		if (diff >= wsizeb)
1490 			return (0);
1491 
1492 		fr = frlast - diff / 8;
1493 
1494 		/* This packet already seen? */
1495 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1496 			return (0);
1497 
1498 		/* Out of order but good. */
1499 		return (1);
1500 	}
1501 }
1502 
1503 /*
1504  * Check replay counter whether to update or not.
1505  * OUT:	0:	OK
1506  *	1:	NG
1507  */
1508 int
1509 ipsec_updatereplay(u_int32_t seq, struct secasvar *sav)
1510 {
1511 	struct secreplay *replay;
1512 	u_int32_t diff;
1513 	int fr;
1514 	u_int32_t wsizeb;	/* Constant: bits of window size. */
1515 	int frlast;		/* Constant: last frame. */
1516 
1517 	IPSEC_ASSERT(sav != NULL, ("Null SA"));
1518 	IPSEC_ASSERT(sav->replay != NULL, ("Null replay state"));
1519 
1520 	replay = sav->replay;
1521 
1522 	if (replay->wsize == 0)
1523 		goto ok;	/* No need to check replay. */
1524 
1525 	/* Constant. */
1526 	frlast = replay->wsize - 1;
1527 	wsizeb = replay->wsize << 3;
1528 
1529 	/* Sequence number of 0 is invalid. */
1530 	if (seq == 0)
1531 		return (1);
1532 
1533 	/* First time. */
1534 	if (replay->count == 0) {
1535 		replay->lastseq = seq;
1536 		bzero(replay->bitmap, replay->wsize);
1537 		(replay->bitmap)[frlast] = 1;
1538 		goto ok;
1539 	}
1540 
1541 	if (seq > replay->lastseq) {
1542 		/* seq is larger than lastseq. */
1543 		diff = seq - replay->lastseq;
1544 
1545 		/* New larger sequence number. */
1546 		if (diff < wsizeb) {
1547 			/* In window. */
1548 			/* Set bit for this packet. */
1549 			vshiftl(replay->bitmap, diff, replay->wsize);
1550 			(replay->bitmap)[frlast] |= 1;
1551 		} else {
1552 			/* This packet has a "way larger". */
1553 			bzero(replay->bitmap, replay->wsize);
1554 			(replay->bitmap)[frlast] = 1;
1555 		}
1556 		replay->lastseq = seq;
1557 
1558 		/* Larger is good. */
1559 	} else {
1560 		/* seq is equal or less than lastseq. */
1561 		diff = replay->lastseq - seq;
1562 
1563 		/* Over range to check, i.e. too old or wrapped. */
1564 		if (diff >= wsizeb)
1565 			return (1);
1566 
1567 		fr = frlast - diff / 8;
1568 
1569 		/* This packet already seen? */
1570 		if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1571 			return (1);
1572 
1573 		/* Mark as seen. */
1574 		(replay->bitmap)[fr] |= (1 << (diff % 8));
1575 
1576 		/* Out of order but good. */
1577 	}
1578 
1579 ok:
1580 	if (replay->count == ~0) {
1581 
1582 		/* Set overflow flag. */
1583 		replay->overflow++;
1584 
1585 		/* Don't increment, no more packets accepted. */
1586 		if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0)
1587 			return (1);
1588 
1589 		ipseclog((LOG_WARNING, "%s: replay counter made %d cycle. %s\n",
1590 		    __func__, replay->overflow, ipsec_logsastr(sav)));
1591 	}
1592 
1593 	replay->count++;
1594 
1595 	return (0);
1596 }
1597 
1598 /*
1599  * Shift variable length buffer to left.
1600  * IN:	bitmap: pointer to the buffer
1601  * 	nbit:	the number of to shift.
1602  *	wsize:	buffer size (bytes).
1603  */
1604 static void
1605 vshiftl(unsigned char *bitmap, int nbit, int wsize)
1606 {
1607 	int s, j, i;
1608 	unsigned char over;
1609 
1610 	for (j = 0; j < nbit; j += 8) {
1611 		s = (nbit - j < 8) ? (nbit - j): 8;
1612 		bitmap[0] <<= s;
1613 		for (i = 1; i < wsize; i++) {
1614 			over = (bitmap[i] >> (8 - s));
1615 			bitmap[i] <<= s;
1616 			bitmap[i-1] |= over;
1617 		}
1618 	}
1619 }
1620 
1621 #ifdef INET
1622 /* Return a printable string for the IPv4 address. */
1623 static char *
1624 inet_ntoa4(struct in_addr ina)
1625 {
1626 	static char buf[4][4 * sizeof "123" + 4];
1627 	unsigned char *ucp = (unsigned char *) &ina;
1628 	static int i = 3;
1629 
1630 	/* XXX-BZ Returns static buffer. */
1631 	i = (i + 1) % 4;
1632 	sprintf(buf[i], "%d.%d.%d.%d", ucp[0] & 0xff, ucp[1] & 0xff,
1633 	    ucp[2] & 0xff, ucp[3] & 0xff);
1634 	return (buf[i]);
1635 }
1636 #endif
1637 
1638 /* Return a printable string for the address. */
1639 char *
1640 ipsec_address(union sockaddr_union* sa)
1641 {
1642 #ifdef INET6
1643 	char ip6buf[INET6_ADDRSTRLEN];
1644 #endif
1645 
1646 	switch (sa->sa.sa_family) {
1647 #ifdef INET
1648 	case AF_INET:
1649 		return (inet_ntoa4(sa->sin.sin_addr));
1650 #endif /* INET */
1651 #ifdef INET6
1652 	case AF_INET6:
1653 		return (ip6_sprintf(ip6buf, &sa->sin6.sin6_addr));
1654 #endif /* INET6 */
1655 	default:
1656 		return ("(unknown address family)");
1657 	}
1658 }
1659 
1660 const char *
1661 ipsec_logsastr(struct secasvar *sav)
1662 {
1663 	static char buf[256];
1664 	char *p;
1665 	struct secasindex *saidx = &sav->sah->saidx;
1666 
1667 	IPSEC_ASSERT(saidx->src.sa.sa_family == saidx->dst.sa.sa_family,
1668 		("address family mismatch"));
1669 
1670 	p = buf;
1671 	snprintf(buf, sizeof(buf), "SA(SPI=%u ", (u_int32_t)ntohl(sav->spi));
1672 	while (p && *p)
1673 		p++;
1674 	/* NB: only use ipsec_address on one address at a time. */
1675 	snprintf(p, sizeof (buf) - (p - buf), "src=%s ",
1676 		ipsec_address(&saidx->src));
1677 	while (p && *p)
1678 		p++;
1679 	snprintf(p, sizeof (buf) - (p - buf), "dst=%s)",
1680 		ipsec_address(&saidx->dst));
1681 
1682 	return (buf);
1683 }
1684 
1685 void
1686 ipsec_dumpmbuf(struct mbuf *m)
1687 {
1688 	int totlen;
1689 	int i;
1690 	u_char *p;
1691 
1692 	totlen = 0;
1693 	printf("---\n");
1694 	while (m) {
1695 		p = mtod(m, u_char *);
1696 		for (i = 0; i < m->m_len; i++) {
1697 			printf("%02x ", p[i]);
1698 			totlen++;
1699 			if (totlen % 16 == 0)
1700 				printf("\n");
1701 		}
1702 		m = m->m_next;
1703 	}
1704 	if (totlen % 16 != 0)
1705 		printf("\n");
1706 	printf("---\n");
1707 }
1708 
1709 static void
1710 ipsec_init(const void *unused __unused)
1711 {
1712 
1713 	SECPOLICY_LOCK_INIT(&V_ip4_def_policy);
1714 	V_ip4_def_policy.refcnt = 1;			/* NB: disallow free. */
1715 }
1716 VNET_SYSINIT(ipsec_init, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, ipsec_init,
1717     NULL);
1718 
1719 
1720 /* XXX This stuff doesn't belong here... */
1721 
1722 static	struct xformsw* xforms = NULL;
1723 
1724 /*
1725  * Register a transform; typically at system startup.
1726  */
1727 void
1728 xform_register(struct xformsw* xsp)
1729 {
1730 
1731 	xsp->xf_next = xforms;
1732 	xforms = xsp;
1733 }
1734 
1735 /*
1736  * Initialize transform support in an sav.
1737  */
1738 int
1739 xform_init(struct secasvar *sav, int xftype)
1740 {
1741 	struct xformsw *xsp;
1742 
1743 	if (sav->tdb_xform != NULL)	/* Previously initialized. */
1744 		return (0);
1745 	for (xsp = xforms; xsp; xsp = xsp->xf_next)
1746 		if (xsp->xf_type == xftype)
1747 			return ((*xsp->xf_init)(sav, xsp));
1748 	return (EINVAL);
1749 }
1750