ipsec.c revision 1.151.2.1 1 /* $NetBSD: ipsec.c,v 1.151.2.1 2018/04/07 04:12:20 pgoyette Exp $ */
2 /* $FreeBSD: src/sys/netipsec/ipsec.c,v 1.2.2.2 2003/07/01 01:38:13 sam Exp $ */
3 /* $KAME: ipsec.c,v 1.103 2001/05/24 07:14:18 sakane Exp $ */
4
5 /*
6 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 */
33
34 #include <sys/cdefs.h>
35 __KERNEL_RCSID(0, "$NetBSD: ipsec.c,v 1.151.2.1 2018/04/07 04:12:20 pgoyette Exp $");
36
37 /*
38 * IPsec controller part.
39 */
40
41 #if defined(_KERNEL_OPT)
42 #include "opt_inet.h"
43 #include "opt_ipsec.h"
44 #endif
45
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/mbuf.h>
49 #include <sys/domain.h>
50 #include <sys/protosw.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/errno.h>
54 #include <sys/time.h>
55 #include <sys/kernel.h>
56 #include <sys/syslog.h>
57 #include <sys/sysctl.h>
58 #include <sys/proc.h>
59 #include <sys/kauth.h>
60 #include <sys/cpu.h>
61 #include <sys/kmem.h>
62 #include <sys/pserialize.h>
63
64 #include <net/if.h>
65 #include <net/route.h>
66
67 #include <netinet/in.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/ip.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/in_var.h>
72 #include <netinet/udp.h>
73 #include <netinet/udp_var.h>
74 #include <netinet/tcp.h>
75 #include <netinet/udp.h>
76 #include <netinet/ip_icmp.h>
77 #include <netinet/ip_private.h>
78
79 #include <netinet/ip6.h>
80 #ifdef INET6
81 #include <netinet6/ip6_var.h>
82 #endif
83 #include <netinet/in_pcb.h>
84 #ifdef INET6
85 #include <netinet6/in6_pcb.h>
86 #include <netinet/icmp6.h>
87 #endif
88
89 #include <netipsec/ipsec.h>
90 #include <netipsec/ipsec_var.h>
91 #include <netipsec/ipsec_private.h>
92 #ifdef INET6
93 #include <netipsec/ipsec6.h>
94 #endif
95 #include <netipsec/ah_var.h>
96 #include <netipsec/esp_var.h>
97 #include <netipsec/ipcomp.h> /*XXX*/
98 #include <netipsec/ipcomp_var.h>
99
100 #include <netipsec/key.h>
101 #include <netipsec/keydb.h>
102 #include <netipsec/key_debug.h>
103
104 #include <netipsec/xform.h>
105
106 int ipsec_used = 0;
107 int ipsec_enabled = 1;
108
109 #ifdef IPSEC_DEBUG
110 int ipsec_debug = 1;
111
112 /*
113 * When set to 1, IPsec will send packets with the same sequence number.
114 * This allows to verify if the other side has proper replay attacks detection.
115 */
116 int ipsec_replay = 0;
117
118 /*
119 * When set 1, IPsec will send packets with corrupted HMAC.
120 * This allows to verify if the other side properly detects modified packets.
121 */
122 int ipsec_integrity = 0;
123 #else
124 int ipsec_debug = 0;
125 #endif
126
127 percpu_t *ipsecstat_percpu;
128
129 int ip4_ah_offsetmask = 0; /* maybe IP_DF? */
130 int ip4_ipsec_dfbit = 2; /* DF bit on encap. 0: clear 1: set 2: copy */
131 int ip4_esp_trans_deflev = IPSEC_LEVEL_USE;
132 int ip4_esp_net_deflev = IPSEC_LEVEL_USE;
133 int ip4_ah_trans_deflev = IPSEC_LEVEL_USE;
134 int ip4_ah_net_deflev = IPSEC_LEVEL_USE;
135 struct secpolicy ip4_def_policy;
136 int ip4_ipsec_ecn = 0; /* ECN ignore(-1)/forbidden(0)/allowed(1) */
137
138 u_int ipsec_spdgen = 1; /* SPD generation # */
139
140 static struct secpolicy ipsec_dummy_sp __read_mostly = {
141 .state = IPSEC_SPSTATE_ALIVE,
142 /* If ENTRUST, the dummy SP never be used. See ipsec_getpolicybysock. */
143 .policy = IPSEC_POLICY_ENTRUST,
144 };
145
146 static struct secpolicy *ipsec_checkpcbcache(struct mbuf *,
147 struct inpcbpolicy *, int);
148 static int ipsec_fillpcbcache(struct inpcbpolicy *, struct mbuf *,
149 struct secpolicy *, int);
150 static int ipsec_invalpcbcache(struct inpcbpolicy *, int);
151
152 /*
153 * Crypto support requirements:
154 *
155 * 1 require hardware support
156 * -1 require software support
157 * 0 take anything
158 */
159 int crypto_support = 0;
160
161 static struct secpolicy *ipsec_getpolicybysock(struct mbuf *, u_int,
162 struct inpcb_hdr *, int *);
163
164 #ifdef INET6
165 int ip6_esp_trans_deflev = IPSEC_LEVEL_USE;
166 int ip6_esp_net_deflev = IPSEC_LEVEL_USE;
167 int ip6_ah_trans_deflev = IPSEC_LEVEL_USE;
168 int ip6_ah_net_deflev = IPSEC_LEVEL_USE;
169 struct secpolicy ip6_def_policy;
170 int ip6_ipsec_ecn = 0; /* ECN ignore(-1)/forbidden(0)/allowed(1) */
171 #endif
172
173 static int ipsec_setspidx_inpcb(struct mbuf *, void *);
174 static int ipsec_setspidx(struct mbuf *, struct secpolicyindex *, int);
175 static void ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *, int);
176 static int ipsec4_setspidx_ipaddr(struct mbuf *, struct secpolicyindex *);
177 #ifdef INET6
178 static void ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *, int);
179 static int ipsec6_setspidx_ipaddr(struct mbuf *, struct secpolicyindex *);
180 #endif
181 static void ipsec_delpcbpolicy(struct inpcbpolicy *);
182 static void ipsec_destroy_policy(struct secpolicy *);
183 static int ipsec_sp_reject(const struct secpolicy *, const struct mbuf *);
184 static void vshiftl(unsigned char *, int, int);
185 static size_t ipsec_sp_hdrsiz(const struct secpolicy *, const struct mbuf *);
186
187 /*
188 * Try to validate and use cached policy on a PCB.
189 */
190 static struct secpolicy *
191 ipsec_checkpcbcache(struct mbuf *m, struct inpcbpolicy *pcbsp, int dir)
192 {
193 struct secpolicyindex spidx;
194 struct secpolicy *sp = NULL;
195 int s;
196
197 KASSERT(IPSEC_DIR_IS_VALID(dir));
198 KASSERT(pcbsp != NULL);
199 KASSERT(dir < __arraycount(pcbsp->sp_cache));
200 KASSERT(inph_locked(pcbsp->sp_inph));
201
202 /*
203 * Checking the generation and sp->state and taking a reference to an SP
204 * must be in a critical section of pserialize. See key_unlink_sp.
205 */
206 s = pserialize_read_enter();
207 /* SPD table change invalidate all the caches. */
208 if (ipsec_spdgen != pcbsp->sp_cache[dir].cachegen) {
209 ipsec_invalpcbcache(pcbsp, dir);
210 goto out;
211 }
212 sp = pcbsp->sp_cache[dir].cachesp;
213 if (sp == NULL)
214 goto out;
215 if (sp->state != IPSEC_SPSTATE_ALIVE) {
216 sp = NULL;
217 ipsec_invalpcbcache(pcbsp, dir);
218 goto out;
219 }
220 if ((pcbsp->sp_cacheflags & IPSEC_PCBSP_CONNECTED) == 0) {
221 /* NB: assume ipsec_setspidx never sleep */
222 if (ipsec_setspidx(m, &spidx, 1) != 0) {
223 sp = NULL;
224 goto out;
225 }
226
227 /*
228 * We have to make an exact match here since the cached rule
229 * might have lower priority than a rule that would otherwise
230 * have matched the packet.
231 */
232 if (memcmp(&pcbsp->sp_cache[dir].cacheidx, &spidx,
233 sizeof(spidx))) {
234 sp = NULL;
235 goto out;
236 }
237 } else {
238 /*
239 * The pcb is connected, and the L4 code is sure that:
240 * - outgoing side uses inp_[lf]addr
241 * - incoming side looks up policy after inpcb lookup
242 * and address pair is know to be stable. We do not need
243 * to generate spidx again, nor check the address match again.
244 *
245 * For IPv4/v6 SOCK_STREAM sockets, this assumptions holds
246 * and there are calls to ipsec_pcbconn() from in_pcbconnect().
247 */
248 }
249
250 sp->lastused = time_second;
251 KEY_SP_REF(sp);
252 KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
253 "DP cause refcnt++:%d SP:%p\n",
254 key_sp_refcnt(sp), pcbsp->sp_cache[dir].cachesp);
255 out:
256 pserialize_read_exit(s);
257 return sp;
258 }
259
260 static int
261 ipsec_fillpcbcache(struct inpcbpolicy *pcbsp, struct mbuf *m,
262 struct secpolicy *sp, int dir)
263 {
264
265 KASSERT(IPSEC_DIR_IS_INOROUT(dir));
266 KASSERT(dir < __arraycount(pcbsp->sp_cache));
267 KASSERT(inph_locked(pcbsp->sp_inph));
268
269 pcbsp->sp_cache[dir].cachesp = NULL;
270 pcbsp->sp_cache[dir].cachehint = IPSEC_PCBHINT_UNKNOWN;
271 if (ipsec_setspidx(m, &pcbsp->sp_cache[dir].cacheidx, 1) != 0) {
272 return EINVAL;
273 }
274 pcbsp->sp_cache[dir].cachesp = sp;
275 if (pcbsp->sp_cache[dir].cachesp) {
276 /*
277 * If the PCB is connected, we can remember a hint to
278 * possibly short-circuit IPsec processing in other places.
279 */
280 if (pcbsp->sp_cacheflags & IPSEC_PCBSP_CONNECTED) {
281 switch (pcbsp->sp_cache[dir].cachesp->policy) {
282 case IPSEC_POLICY_NONE:
283 case IPSEC_POLICY_BYPASS:
284 pcbsp->sp_cache[dir].cachehint =
285 IPSEC_PCBHINT_NO;
286 break;
287 default:
288 pcbsp->sp_cache[dir].cachehint =
289 IPSEC_PCBHINT_YES;
290 }
291 }
292 }
293 pcbsp->sp_cache[dir].cachegen = ipsec_spdgen;
294
295 return 0;
296 }
297
298 static int
299 ipsec_invalpcbcache(struct inpcbpolicy *pcbsp, int dir)
300 {
301 int i;
302
303 KASSERT(inph_locked(pcbsp->sp_inph));
304
305 for (i = IPSEC_DIR_INBOUND; i <= IPSEC_DIR_OUTBOUND; i++) {
306 if (dir != IPSEC_DIR_ANY && i != dir)
307 continue;
308 pcbsp->sp_cache[i].cachesp = NULL;
309 pcbsp->sp_cache[i].cachehint = IPSEC_PCBHINT_UNKNOWN;
310 pcbsp->sp_cache[i].cachegen = 0;
311 memset(&pcbsp->sp_cache[i].cacheidx, 0,
312 sizeof(pcbsp->sp_cache[i].cacheidx));
313 }
314 return 0;
315 }
316
317 void
318 ipsec_pcbconn(struct inpcbpolicy *pcbsp)
319 {
320
321 KASSERT(inph_locked(pcbsp->sp_inph));
322
323 pcbsp->sp_cacheflags |= IPSEC_PCBSP_CONNECTED;
324 ipsec_invalpcbcache(pcbsp, IPSEC_DIR_ANY);
325 }
326
327 void
328 ipsec_pcbdisconn(struct inpcbpolicy *pcbsp)
329 {
330
331 KASSERT(inph_locked(pcbsp->sp_inph));
332
333 pcbsp->sp_cacheflags &= ~IPSEC_PCBSP_CONNECTED;
334 ipsec_invalpcbcache(pcbsp, IPSEC_DIR_ANY);
335 }
336
337 void
338 ipsec_invalpcbcacheall(void)
339 {
340
341 if (ipsec_spdgen == UINT_MAX)
342 ipsec_spdgen = 1;
343 else
344 ipsec_spdgen++;
345 }
346
347 /*
348 * Return a held reference to the default SP.
349 */
350 static struct secpolicy *
351 key_get_default_sp(int af, const char *where, int tag)
352 {
353 struct secpolicy *sp;
354
355 KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
356
357 switch(af) {
358 case AF_INET:
359 sp = &ip4_def_policy;
360 break;
361 #ifdef INET6
362 case AF_INET6:
363 sp = &ip6_def_policy;
364 break;
365 #endif
366 default:
367 KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
368 "unexpected protocol family %u\n", af);
369 return NULL;
370 }
371
372 if (sp->policy != IPSEC_POLICY_DISCARD &&
373 sp->policy != IPSEC_POLICY_NONE) {
374 IPSECLOG(LOG_INFO, "fixed system default policy: %d->%d\n",
375 sp->policy, IPSEC_POLICY_NONE);
376 sp->policy = IPSEC_POLICY_NONE;
377 }
378 KEY_SP_REF(sp);
379
380 KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP returns SP:%p (%u)\n",
381 sp, key_sp_refcnt(sp));
382 return sp;
383 }
384
385 #define KEY_GET_DEFAULT_SP(af) \
386 key_get_default_sp((af), __func__, __LINE__)
387
388 /*
389 * For OUTBOUND packet having a socket. Searching SPD for packet,
390 * and return a pointer to SP.
391 * OUT: NULL: no appropriate SP found, the following value is set to error.
392 * 0 : bypass
393 * EACCES : discard packet.
394 * ENOENT : ipsec_acquire() in progress, maybe.
395 * others : error occurred.
396 * others: a pointer to SP
397 *
398 * NOTE: IPv6 mapped address concern is implemented here.
399 */
400 static struct secpolicy *
401 ipsec_getpolicybysock(struct mbuf *m, u_int dir, struct inpcb_hdr *inph,
402 int *error)
403 {
404 struct inpcbpolicy *pcbsp = NULL;
405 struct secpolicy *currsp = NULL; /* policy on socket */
406 struct secpolicy *sp;
407 int af;
408
409 KASSERT(m != NULL);
410 KASSERT(inph != NULL);
411 KASSERT(error != NULL);
412 KASSERTMSG(IPSEC_DIR_IS_INOROUT(dir), "invalid direction %u", dir);
413
414 KASSERT(inph->inph_socket != NULL);
415 KASSERT(inph_locked(inph));
416
417 /* XXX FIXME inpcb/in6pcb vs socket*/
418 af = inph->inph_af;
419 KASSERTMSG(af == AF_INET || af == AF_INET6,
420 "unexpected protocol family %u", af);
421
422 KASSERT(inph->inph_sp != NULL);
423 /* If we have a cached entry, and if it is still valid, use it. */
424 IPSEC_STATINC(IPSEC_STAT_SPDCACHELOOKUP);
425 currsp = ipsec_checkpcbcache(m, inph->inph_sp, dir);
426 if (currsp) {
427 *error = 0;
428 return currsp;
429 }
430 IPSEC_STATINC(IPSEC_STAT_SPDCACHEMISS);
431
432 switch (af) {
433 case AF_INET:
434 #if defined(INET6)
435 case AF_INET6:
436 #endif
437 *error = ipsec_setspidx_inpcb(m, inph);
438 pcbsp = inph->inph_sp;
439 break;
440 default:
441 *error = EPFNOSUPPORT;
442 break;
443 }
444 if (*error)
445 return NULL;
446
447 KASSERT(pcbsp != NULL);
448 switch (dir) {
449 case IPSEC_DIR_INBOUND:
450 currsp = pcbsp->sp_in;
451 break;
452 case IPSEC_DIR_OUTBOUND:
453 currsp = pcbsp->sp_out;
454 break;
455 }
456 KASSERT(currsp != NULL);
457
458 if (pcbsp->priv) { /* when privileged socket */
459 switch (currsp->policy) {
460 case IPSEC_POLICY_BYPASS:
461 case IPSEC_POLICY_IPSEC:
462 KEY_SP_REF(currsp);
463 sp = currsp;
464 break;
465
466 case IPSEC_POLICY_ENTRUST:
467 /* look for a policy in SPD */
468 sp = KEY_LOOKUP_SP_BYSPIDX(&currsp->spidx, dir);
469 if (sp == NULL) /* no SP found */
470 sp = KEY_GET_DEFAULT_SP(af);
471 break;
472
473 default:
474 IPSECLOG(LOG_ERR, "Invalid policy for PCB %d\n",
475 currsp->policy);
476 *error = EINVAL;
477 return NULL;
478 }
479 } else { /* unpriv, SPD has policy */
480 sp = KEY_LOOKUP_SP_BYSPIDX(&currsp->spidx, dir);
481 if (sp == NULL) { /* no SP found */
482 switch (currsp->policy) {
483 case IPSEC_POLICY_BYPASS:
484 IPSECLOG(LOG_ERR, "Illegal policy for "
485 "non-priviliged defined %d\n",
486 currsp->policy);
487 *error = EINVAL;
488 return NULL;
489
490 case IPSEC_POLICY_ENTRUST:
491 sp = KEY_GET_DEFAULT_SP(af);
492 break;
493
494 case IPSEC_POLICY_IPSEC:
495 KEY_SP_REF(currsp);
496 sp = currsp;
497 break;
498
499 default:
500 IPSECLOG(LOG_ERR, "Invalid policy for "
501 "PCB %d\n", currsp->policy);
502 *error = EINVAL;
503 return NULL;
504 }
505 }
506 }
507 KASSERTMSG(sp != NULL, "null SP (priv %u policy %u", pcbsp->priv,
508 currsp->policy);
509 KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
510 "DP (priv %u policy %u) allocates SP:%p (refcnt %u)\n",
511 pcbsp->priv, currsp->policy, sp, key_sp_refcnt(sp));
512 ipsec_fillpcbcache(pcbsp, m, sp, dir);
513 return sp;
514 }
515
516 /*
517 * For FORWARDING packet or OUTBOUND without a socket. Searching SPD for packet,
518 * and return a pointer to SP.
519 * OUT: positive: a pointer to the entry for security policy leaf matched.
520 * NULL: no appropriate SP found, the following value is set to error.
521 * 0 : bypass
522 * EACCES : discard packet.
523 * ENOENT : ipsec_acquire() in progress, maybe.
524 * others : error occurred.
525 */
526 static struct secpolicy *
527 ipsec_getpolicybyaddr(struct mbuf *m, u_int dir, int flag, int *error)
528 {
529 struct secpolicyindex spidx;
530 struct secpolicy *sp;
531
532 KASSERT(m != NULL);
533 KASSERT(error != NULL);
534 KASSERTMSG(IPSEC_DIR_IS_INOROUT(dir), "invalid direction %u", dir);
535
536 sp = NULL;
537
538 /* Make an index to look for a policy. */
539 *error = ipsec_setspidx(m, &spidx, (flag & IP_FORWARDING) ? 0 : 1);
540 if (*error != 0) {
541 IPSECLOG(LOG_DEBUG, "setpidx failed, dir %u flag %u\n", dir, flag);
542 memset(&spidx, 0, sizeof(spidx));
543 return NULL;
544 }
545
546 spidx.dir = dir;
547
548 if (key_havesp(dir)) {
549 sp = KEY_LOOKUP_SP_BYSPIDX(&spidx, dir);
550 }
551
552 if (sp == NULL) /* no SP found, use system default */
553 sp = KEY_GET_DEFAULT_SP(spidx.dst.sa.sa_family);
554 KASSERT(sp != NULL);
555 return sp;
556 }
557
558 static struct secpolicy *
559 ipsec_checkpolicy(struct mbuf *m, u_int dir, u_int flag, int *error,
560 void *inp)
561 {
562 struct secpolicy *sp;
563
564 *error = 0;
565
566 if (inp == NULL) {
567 sp = ipsec_getpolicybyaddr(m, dir, flag, error);
568 } else {
569 struct inpcb_hdr *inph = (struct inpcb_hdr *)inp;
570 KASSERT(inph->inph_socket != NULL);
571 sp = ipsec_getpolicybysock(m, dir, inph, error);
572 }
573 if (sp == NULL) {
574 KASSERTMSG(*error != 0, "getpolicy failed w/o error");
575 IPSEC_STATINC(IPSEC_STAT_OUT_INVAL);
576 return NULL;
577 }
578 KASSERTMSG(*error == 0, "sp w/ error set to %u", *error);
579
580 switch (sp->policy) {
581 case IPSEC_POLICY_ENTRUST:
582 default:
583 printf("%s: invalid policy %u\n", __func__, sp->policy);
584 /* fall thru... */
585 case IPSEC_POLICY_DISCARD:
586 IPSEC_STATINC(IPSEC_STAT_OUT_POLVIO);
587 *error = -EINVAL; /* packet is discarded by caller */
588 break;
589 case IPSEC_POLICY_BYPASS:
590 case IPSEC_POLICY_NONE:
591 KEY_SP_UNREF(&sp);
592 sp = NULL; /* NB: force NULL result */
593 break;
594 case IPSEC_POLICY_IPSEC:
595 KASSERT(sp->req != NULL);
596 break;
597 }
598
599 if (*error != 0) {
600 KEY_SP_UNREF(&sp);
601 sp = NULL;
602 IPSECLOG(LOG_DEBUG, "done, error %d\n", *error);
603 }
604
605 return sp;
606 }
607
608 int
609 ipsec4_output(struct mbuf *m, struct inpcb *inp, int flags,
610 u_long *mtu, bool *natt_frag, bool *done)
611 {
612 struct secpolicy *sp = NULL;
613 u_long _mtu = 0;
614 int error, s;
615
616 /*
617 * Check the security policy (SP) for the packet and, if required,
618 * do IPsec-related processing. There are two cases here; the first
619 * time a packet is sent through it will be untagged and handled by
620 * ipsec_checkpolicy(). If the packet is resubmitted to ip_output
621 * (e.g. after AH, ESP, etc. processing), there will be a tag to
622 * bypass the lookup and related policy checking.
623 */
624 if (ipsec_outdone(m)) {
625 return 0;
626 }
627 s = splsoftnet();
628 if (inp && ipsec_pcb_skip_ipsec(inp->inp_sp, IPSEC_DIR_OUTBOUND)) {
629 splx(s);
630 return 0;
631 }
632 sp = ipsec_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, &error, inp);
633
634 /*
635 * There are four return cases:
636 * sp != NULL apply IPsec policy
637 * sp == NULL, error == 0 no IPsec handling needed
638 * sp == NULL, error == -EINVAL discard packet w/o error
639 * sp == NULL, error != 0 discard packet, report error
640 */
641 if (sp == NULL) {
642 splx(s);
643 if (error) {
644 /*
645 * Hack: -EINVAL is used to signal that a packet
646 * should be silently discarded. This is typically
647 * because we asked key management for an SA and
648 * it was delayed (e.g. kicked up to IKE).
649 */
650 if (error == -EINVAL)
651 error = 0;
652 m_freem(m);
653 *done = true;
654 return error;
655 }
656 /* No IPsec processing for this packet. */
657 return 0;
658 }
659
660 /*
661 * Do delayed checksums now because we send before
662 * this is done in the normal processing path.
663 */
664 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) {
665 in_delayed_cksum(m);
666 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4);
667 }
668
669 error = ipsec4_process_packet(m, sp->req, &_mtu);
670 if (error == 0 && _mtu != 0) {
671 /*
672 * NAT-T ESP fragmentation: do not do IPSec processing
673 * now, we will do it on each fragmented packet.
674 */
675 *mtu = _mtu;
676 *natt_frag = true;
677 KEY_SP_UNREF(&sp);
678 splx(s);
679 return 0;
680 }
681
682 /*
683 * Preserve KAME behaviour: ENOENT can be returned
684 * when an SA acquire is in progress. Don't propagate
685 * this to user-level; it confuses applications.
686 *
687 * XXX this will go away when the SADB is redone.
688 */
689 if (error == ENOENT)
690 error = 0;
691 KEY_SP_UNREF(&sp);
692 splx(s);
693 *done = true;
694 return error;
695 }
696
697 int
698 ipsec4_input(struct mbuf *m, int flags)
699 {
700 struct secpolicy *sp;
701 int error, s;
702
703 s = splsoftnet();
704 error = ipsec_in_reject(m, NULL);
705 splx(s);
706 if (error) {
707 return EINVAL;
708 }
709
710 if (flags == 0) {
711 /* We are done. */
712 return 0;
713 }
714
715 /*
716 * Peek at the outbound SP for this packet to determine if
717 * it is a Fast Forward candidate.
718 */
719 s = splsoftnet();
720 sp = ipsec_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, &error, NULL);
721 if (sp != NULL) {
722 m->m_flags &= ~M_CANFASTFWD;
723 KEY_SP_UNREF(&sp);
724 }
725 splx(s);
726 return 0;
727 }
728
729 int
730 ipsec4_forward(struct mbuf *m, int *destmtu)
731 {
732 /*
733 * If the packet is routed over IPsec tunnel, tell the
734 * originator the tunnel MTU.
735 * tunnel MTU = if MTU - sizeof(IP) - ESP/AH hdrsiz
736 * XXX quickhack!!!
737 */
738 struct secpolicy *sp;
739 size_t ipsechdr;
740 int error;
741
742 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING,
743 &error);
744 if (sp == NULL) {
745 return EINVAL;
746 }
747
748 /* Count IPsec header size. */
749 ipsechdr = ipsec_sp_hdrsiz(sp, m);
750
751 /*
752 * Find the correct route for outer IPv4 header, compute tunnel MTU.
753 */
754 if (sp->req) {
755 struct secasvar *sav;
756
757 sav = ipsec_lookup_sa(sp->req, m);
758 if (sav != NULL) {
759 struct route *ro;
760 struct rtentry *rt;
761
762 ro = &sav->sah->sa_route;
763 rt = rtcache_validate(ro);
764 if (rt && rt->rt_ifp) {
765 *destmtu = rt->rt_rmx.rmx_mtu ?
766 rt->rt_rmx.rmx_mtu : rt->rt_ifp->if_mtu;
767 *destmtu -= ipsechdr;
768 }
769 rtcache_unref(rt, ro);
770 KEY_SA_UNREF(&sav);
771 }
772 }
773 KEY_SP_UNREF(&sp);
774 return 0;
775 }
776
777 static int
778 ipsec_setspidx_inpcb(struct mbuf *m, void *pcb)
779 {
780 struct inpcb_hdr *inph = (struct inpcb_hdr *)pcb;
781 int error;
782
783 KASSERT(inph != NULL);
784 KASSERT(inph->inph_sp != NULL);
785 KASSERT(inph->inph_sp->sp_out != NULL);
786 KASSERT(inph->inph_sp->sp_in != NULL);
787
788 error = ipsec_setspidx(m, &inph->inph_sp->sp_in->spidx, 1);
789 if (error == 0) {
790 inph->inph_sp->sp_in->spidx.dir = IPSEC_DIR_INBOUND;
791 inph->inph_sp->sp_out->spidx = inph->inph_sp->sp_in->spidx;
792 inph->inph_sp->sp_out->spidx.dir = IPSEC_DIR_OUTBOUND;
793 } else {
794 memset(&inph->inph_sp->sp_in->spidx, 0,
795 sizeof(inph->inph_sp->sp_in->spidx));
796 memset(&inph->inph_sp->sp_out->spidx, 0,
797 sizeof(inph->inph_sp->sp_out->spidx));
798 }
799 return error;
800 }
801
802 /*
803 * configure security policy index (src/dst/proto/sport/dport)
804 * by looking at the content of mbuf.
805 * the caller is responsible for error recovery (like clearing up spidx).
806 */
807 static int
808 ipsec_setspidx(struct mbuf *m, struct secpolicyindex *spidx, int needport)
809 {
810 struct ip *ip = NULL;
811 struct ip ipbuf;
812 u_int v;
813 struct mbuf *n;
814 int len;
815 int error;
816
817 KASSERT(m != NULL);
818
819 /*
820 * validate m->m_pkthdr.len. we see incorrect length if we
821 * mistakenly call this function with inconsistent mbuf chain
822 * (like 4.4BSD tcp/udp processing). XXX should we panic here?
823 */
824 len = 0;
825 for (n = m; n; n = n->m_next)
826 len += n->m_len;
827 if (m->m_pkthdr.len != len) {
828 KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_DUMP,
829 "total of m_len(%d) != pkthdr.len(%d), ignored.\n",
830 len, m->m_pkthdr.len);
831 KASSERTMSG(0, "impossible");
832 return EINVAL;
833 }
834
835 if (m->m_pkthdr.len < sizeof(struct ip)) {
836 KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_DUMP,
837 "pkthdr.len(%d) < sizeof(struct ip), ignored.\n",
838 m->m_pkthdr.len);
839 return EINVAL;
840 }
841
842 if (m->m_len >= sizeof(*ip)) {
843 ip = mtod(m, struct ip *);
844 } else {
845 m_copydata(m, 0, sizeof(ipbuf), &ipbuf);
846 ip = &ipbuf;
847 }
848 v = ip->ip_v;
849 switch (v) {
850 case 4:
851 error = ipsec4_setspidx_ipaddr(m, spidx);
852 if (error)
853 return error;
854 ipsec4_get_ulp(m, spidx, needport);
855 return 0;
856 #ifdef INET6
857 case 6:
858 if (m->m_pkthdr.len < sizeof(struct ip6_hdr)) {
859 KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_DUMP,
860 "pkthdr.len(%d) < sizeof(struct ip6_hdr), "
861 "ignored.\n", m->m_pkthdr.len);
862 return EINVAL;
863 }
864 error = ipsec6_setspidx_ipaddr(m, spidx);
865 if (error)
866 return error;
867 ipsec6_get_ulp(m, spidx, needport);
868 return 0;
869 #endif
870 default:
871 KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_DUMP,
872 "unknown IP version %u, ignored.\n", v);
873 return EINVAL;
874 }
875 }
876
877 static void
878 ipsec4_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
879 {
880 u_int8_t nxt;
881 int off;
882
883 KASSERT(m != NULL);
884 KASSERTMSG(m->m_pkthdr.len >= sizeof(struct ip), "packet too short");
885
886 /* NB: ip_input() flips it into host endian XXX need more checking */
887 if (m->m_len >= sizeof(struct ip)) {
888 struct ip *ip = mtod(m, struct ip *);
889 if (ip->ip_off & htons(IP_MF | IP_OFFMASK))
890 goto done;
891 off = ip->ip_hl << 2;
892 nxt = ip->ip_p;
893 } else {
894 struct ip ih;
895
896 m_copydata(m, 0, sizeof(struct ip), &ih);
897 if (ih.ip_off & htons(IP_MF | IP_OFFMASK))
898 goto done;
899 off = ih.ip_hl << 2;
900 nxt = ih.ip_p;
901 }
902
903 while (off < m->m_pkthdr.len) {
904 struct ip6_ext ip6e;
905 struct tcphdr th;
906 struct udphdr uh;
907 struct icmp icmph;
908
909 switch (nxt) {
910 case IPPROTO_TCP:
911 spidx->ul_proto = nxt;
912 if (!needport)
913 goto done_proto;
914 if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
915 goto done;
916 m_copydata(m, off, sizeof(th), &th);
917 spidx->src.sin.sin_port = th.th_sport;
918 spidx->dst.sin.sin_port = th.th_dport;
919 return;
920 case IPPROTO_UDP:
921 spidx->ul_proto = nxt;
922 if (!needport)
923 goto done_proto;
924 if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
925 goto done;
926 m_copydata(m, off, sizeof(uh), &uh);
927 spidx->src.sin.sin_port = uh.uh_sport;
928 spidx->dst.sin.sin_port = uh.uh_dport;
929 return;
930 case IPPROTO_AH:
931 if (off + sizeof(ip6e) > m->m_pkthdr.len)
932 goto done;
933 /* XXX sigh, this works but is totally bogus */
934 m_copydata(m, off, sizeof(ip6e), &ip6e);
935 off += (ip6e.ip6e_len + 2) << 2;
936 nxt = ip6e.ip6e_nxt;
937 break;
938 case IPPROTO_ICMP:
939 spidx->ul_proto = nxt;
940 if (off + sizeof(struct icmp) > m->m_pkthdr.len)
941 goto done;
942 m_copydata(m, off, sizeof(icmph), &icmph);
943 ((struct sockaddr_in *)&spidx->src)->sin_port =
944 htons((uint16_t)icmph.icmp_type);
945 ((struct sockaddr_in *)&spidx->dst)->sin_port =
946 htons((uint16_t)icmph.icmp_code);
947 return;
948 default:
949 /* XXX intermediate headers??? */
950 spidx->ul_proto = nxt;
951 goto done_proto;
952 }
953 }
954 done:
955 spidx->ul_proto = IPSEC_ULPROTO_ANY;
956 done_proto:
957 spidx->src.sin.sin_port = IPSEC_PORT_ANY;
958 spidx->dst.sin.sin_port = IPSEC_PORT_ANY;
959 }
960
961 static int
962 ipsec4_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
963 {
964 static const struct sockaddr_in template = {
965 sizeof(struct sockaddr_in),
966 AF_INET,
967 0, { 0 }, { 0, 0, 0, 0, 0, 0, 0, 0 }
968 };
969
970 spidx->src.sin = template;
971 spidx->dst.sin = template;
972
973 if (m->m_len < sizeof(struct ip)) {
974 m_copydata(m, offsetof(struct ip, ip_src),
975 sizeof(struct in_addr), &spidx->src.sin.sin_addr);
976 m_copydata(m, offsetof(struct ip, ip_dst),
977 sizeof(struct in_addr), &spidx->dst.sin.sin_addr);
978 } else {
979 struct ip *ip = mtod(m, struct ip *);
980 spidx->src.sin.sin_addr = ip->ip_src;
981 spidx->dst.sin.sin_addr = ip->ip_dst;
982 }
983
984 spidx->prefs = sizeof(struct in_addr) << 3;
985 spidx->prefd = sizeof(struct in_addr) << 3;
986
987 return 0;
988 }
989
990 #ifdef INET6
991 static void
992 ipsec6_get_ulp(struct mbuf *m, struct secpolicyindex *spidx, int needport)
993 {
994 int off, nxt;
995 struct tcphdr th;
996 struct udphdr uh;
997 struct icmp6_hdr icmph;
998
999 KASSERT(m != NULL);
1000
1001 if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DUMP)) {
1002 kdebug_mbuf(__func__, m);
1003 }
1004
1005 /* set default */
1006 spidx->ul_proto = IPSEC_ULPROTO_ANY;
1007 ((struct sockaddr_in6 *)&spidx->src)->sin6_port = IPSEC_PORT_ANY;
1008 ((struct sockaddr_in6 *)&spidx->dst)->sin6_port = IPSEC_PORT_ANY;
1009
1010 nxt = -1;
1011 off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
1012 if (off < 0 || m->m_pkthdr.len < off)
1013 return;
1014
1015 switch (nxt) {
1016 case IPPROTO_TCP:
1017 spidx->ul_proto = nxt;
1018 if (!needport)
1019 break;
1020 if (off + sizeof(struct tcphdr) > m->m_pkthdr.len)
1021 break;
1022 m_copydata(m, off, sizeof(th), &th);
1023 ((struct sockaddr_in6 *)&spidx->src)->sin6_port = th.th_sport;
1024 ((struct sockaddr_in6 *)&spidx->dst)->sin6_port = th.th_dport;
1025 break;
1026 case IPPROTO_UDP:
1027 spidx->ul_proto = nxt;
1028 if (!needport)
1029 break;
1030 if (off + sizeof(struct udphdr) > m->m_pkthdr.len)
1031 break;
1032 m_copydata(m, off, sizeof(uh), &uh);
1033 ((struct sockaddr_in6 *)&spidx->src)->sin6_port = uh.uh_sport;
1034 ((struct sockaddr_in6 *)&spidx->dst)->sin6_port = uh.uh_dport;
1035 break;
1036 case IPPROTO_ICMPV6:
1037 spidx->ul_proto = nxt;
1038 if (off + sizeof(struct icmp6_hdr) > m->m_pkthdr.len)
1039 break;
1040 m_copydata(m, off, sizeof(icmph), &icmph);
1041 ((struct sockaddr_in6 *)&spidx->src)->sin6_port =
1042 htons((uint16_t)icmph.icmp6_type);
1043 ((struct sockaddr_in6 *)&spidx->dst)->sin6_port =
1044 htons((uint16_t)icmph.icmp6_code);
1045 break;
1046 default:
1047 /* XXX intermediate headers??? */
1048 spidx->ul_proto = nxt;
1049 break;
1050 }
1051 }
1052
1053 static int
1054 ipsec6_setspidx_ipaddr(struct mbuf *m, struct secpolicyindex *spidx)
1055 {
1056 struct ip6_hdr *ip6 = NULL;
1057 struct ip6_hdr ip6buf;
1058 struct sockaddr_in6 *sin6;
1059
1060 if (m->m_len >= sizeof(*ip6))
1061 ip6 = mtod(m, struct ip6_hdr *);
1062 else {
1063 m_copydata(m, 0, sizeof(ip6buf), &ip6buf);
1064 ip6 = &ip6buf;
1065 }
1066
1067 sin6 = (struct sockaddr_in6 *)&spidx->src;
1068 memset(sin6, 0, sizeof(*sin6));
1069 sin6->sin6_family = AF_INET6;
1070 sin6->sin6_len = sizeof(struct sockaddr_in6);
1071 memcpy(&sin6->sin6_addr, &ip6->ip6_src, sizeof(ip6->ip6_src));
1072 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) {
1073 sin6->sin6_addr.s6_addr16[1] = 0;
1074 sin6->sin6_scope_id = ntohs(ip6->ip6_src.s6_addr16[1]);
1075 }
1076 spidx->prefs = sizeof(struct in6_addr) << 3;
1077
1078 sin6 = (struct sockaddr_in6 *)&spidx->dst;
1079 memset(sin6, 0, sizeof(*sin6));
1080 sin6->sin6_family = AF_INET6;
1081 sin6->sin6_len = sizeof(struct sockaddr_in6);
1082 memcpy(&sin6->sin6_addr, &ip6->ip6_dst, sizeof(ip6->ip6_dst));
1083 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
1084 sin6->sin6_addr.s6_addr16[1] = 0;
1085 sin6->sin6_scope_id = ntohs(ip6->ip6_dst.s6_addr16[1]);
1086 }
1087 spidx->prefd = sizeof(struct in6_addr) << 3;
1088
1089 return 0;
1090 }
1091 #endif
1092
1093 static void
1094 ipsec_delpcbpolicy(struct inpcbpolicy *p)
1095 {
1096
1097 kmem_intr_free(p, sizeof(*p));
1098 }
1099
1100 /* initialize policy in PCB */
1101 int
1102 ipsec_init_policy(struct socket *so, struct inpcbpolicy **policy)
1103 {
1104 struct inpcbpolicy *new;
1105
1106 KASSERT(so != NULL);
1107 KASSERT(policy != NULL);
1108
1109 new = kmem_intr_zalloc(sizeof(*new), KM_NOSLEEP);
1110 if (new == NULL) {
1111 IPSECLOG(LOG_DEBUG, "No more memory.\n");
1112 return ENOBUFS;
1113 }
1114
1115 if (IPSEC_PRIVILEGED_SO(so))
1116 new->priv = 1;
1117 else
1118 new->priv = 0;
1119
1120 /*
1121 * Set dummy SPs. Actual SPs will be allocated later if needed.
1122 */
1123 new->sp_in = &ipsec_dummy_sp;
1124 new->sp_out = &ipsec_dummy_sp;
1125
1126 *policy = new;
1127
1128 return 0;
1129 }
1130
1131 static void
1132 ipsec_destroy_policy(struct secpolicy *sp)
1133 {
1134
1135 if (sp == &ipsec_dummy_sp)
1136 ; /* It's dummy. No need to free it. */
1137 else {
1138 /*
1139 * We cannot destroy here because it can be called in
1140 * softint. So mark the SP as DEAD and let the timer
1141 * destroy it. See key_timehandler_spd.
1142 */
1143 sp->state = IPSEC_SPSTATE_DEAD;
1144 }
1145 }
1146
1147 int
1148 ipsec_set_policy(void *inp, int optname, const void *request, size_t len,
1149 kauth_cred_t cred)
1150 {
1151 struct inpcb_hdr *inph = (struct inpcb_hdr *)inp;
1152 const struct sadb_x_policy *xpl;
1153 struct secpolicy *newsp, *oldsp;
1154 struct secpolicy **policy;
1155 int error;
1156
1157 KASSERT(!cpu_softintr_p());
1158 KASSERT(inph != NULL);
1159 KASSERT(inph_locked(inph));
1160 KASSERT(request != NULL);
1161
1162 if (len < sizeof(*xpl))
1163 return EINVAL;
1164 xpl = (const struct sadb_x_policy *)request;
1165
1166 KASSERT(inph->inph_sp != NULL);
1167
1168 /* select direction */
1169 switch (xpl->sadb_x_policy_dir) {
1170 case IPSEC_DIR_INBOUND:
1171 policy = &inph->inph_sp->sp_in;
1172 break;
1173 case IPSEC_DIR_OUTBOUND:
1174 policy = &inph->inph_sp->sp_out;
1175 break;
1176 default:
1177 IPSECLOG(LOG_ERR, "invalid direction=%u\n",
1178 xpl->sadb_x_policy_dir);
1179 return EINVAL;
1180 }
1181
1182 /* sanity check. */
1183 if (policy == NULL || *policy == NULL)
1184 return EINVAL;
1185
1186 if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DUMP)) {
1187 kdebug_sadb_xpolicy("set passed policy", request);
1188 }
1189
1190 /* check policy type */
1191 /* ipsec_set_policy() accepts IPSEC, ENTRUST and BYPASS. */
1192 if (xpl->sadb_x_policy_type == IPSEC_POLICY_DISCARD ||
1193 xpl->sadb_x_policy_type == IPSEC_POLICY_NONE)
1194 return EINVAL;
1195
1196 /* check privileged socket */
1197 if (xpl->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
1198 error = kauth_authorize_network(cred, KAUTH_NETWORK_IPSEC,
1199 KAUTH_REQ_NETWORK_IPSEC_BYPASS, NULL, NULL, NULL);
1200 if (error)
1201 return error;
1202 }
1203
1204 /* allocation new SP entry */
1205 if ((newsp = key_msg2sp(xpl, len, &error)) == NULL)
1206 return error;
1207
1208 key_init_sp(newsp);
1209 newsp->created = time_uptime;
1210 /* Insert the global list for SPs for sockets */
1211 key_socksplist_add(newsp);
1212
1213 /* clear old SP and set new SP */
1214 oldsp = *policy;
1215 *policy = newsp;
1216 ipsec_destroy_policy(oldsp);
1217
1218 if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DUMP)) {
1219 printf("%s: new policy\n", __func__);
1220 kdebug_secpolicy(newsp);
1221 }
1222
1223 return 0;
1224 }
1225
1226 int
1227 ipsec_get_policy(void *inp, const void *request, size_t len,
1228 struct mbuf **mp)
1229 {
1230 struct inpcb_hdr *inph = (struct inpcb_hdr *)inp;
1231 const struct sadb_x_policy *xpl;
1232 struct secpolicy *policy;
1233
1234 /* sanity check. */
1235 if (inph == NULL || request == NULL || mp == NULL)
1236 return EINVAL;
1237 KASSERT(inph->inph_sp != NULL);
1238 if (len < sizeof(*xpl))
1239 return EINVAL;
1240 xpl = (const struct sadb_x_policy *)request;
1241
1242 /* select direction */
1243 switch (xpl->sadb_x_policy_dir) {
1244 case IPSEC_DIR_INBOUND:
1245 policy = inph->inph_sp->sp_in;
1246 break;
1247 case IPSEC_DIR_OUTBOUND:
1248 policy = inph->inph_sp->sp_out;
1249 break;
1250 default:
1251 IPSECLOG(LOG_ERR, "invalid direction=%u\n",
1252 xpl->sadb_x_policy_dir);
1253 return EINVAL;
1254 }
1255
1256 if (policy == NULL)
1257 return EINVAL;
1258
1259 *mp = key_sp2msg(policy, M_NOWAIT);
1260 if (!*mp) {
1261 IPSECLOG(LOG_DEBUG, "No more memory.\n");
1262 return ENOBUFS;
1263 }
1264
1265 (*mp)->m_type = MT_DATA;
1266 if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DUMP)) {
1267 kdebug_mbuf(__func__, *mp);
1268 }
1269
1270 return 0;
1271 }
1272
1273 int
1274 ipsec_delete_pcbpolicy(void *inp)
1275 {
1276 struct inpcb_hdr *inph = (struct inpcb_hdr *)inp;
1277
1278 KASSERT(inph != NULL);
1279
1280 if (inph->inph_sp == NULL)
1281 return 0;
1282
1283 if (inph->inph_sp->sp_in != NULL)
1284 ipsec_destroy_policy(inph->inph_sp->sp_in);
1285
1286 if (inph->inph_sp->sp_out != NULL)
1287 ipsec_destroy_policy(inph->inph_sp->sp_out);
1288
1289 ipsec_invalpcbcache(inph->inph_sp, IPSEC_DIR_ANY);
1290
1291 ipsec_delpcbpolicy(inph->inph_sp);
1292 inph->inph_sp = NULL;
1293
1294 return 0;
1295 }
1296
1297 /*
1298 * Return the current level (either IPSEC_LEVEL_USE or IPSEC_LEVEL_REQUIRE).
1299 */
1300 u_int
1301 ipsec_get_reqlevel(const struct ipsecrequest *isr)
1302 {
1303 u_int level = 0;
1304 u_int esp_trans_deflev, esp_net_deflev;
1305 u_int ah_trans_deflev, ah_net_deflev;
1306
1307 KASSERT(isr != NULL);
1308 KASSERT(isr->sp != NULL);
1309 KASSERTMSG(
1310 isr->sp->spidx.src.sa.sa_family == isr->sp->spidx.dst.sa.sa_family,
1311 "af family mismatch, src %u, dst %u",
1312 isr->sp->spidx.src.sa.sa_family, isr->sp->spidx.dst.sa.sa_family);
1313
1314 /* XXX note that we have ipseclog() expanded here - code sync issue */
1315 #define IPSEC_CHECK_DEFAULT(lev) \
1316 (((lev) != IPSEC_LEVEL_USE && (lev) != IPSEC_LEVEL_REQUIRE \
1317 && (lev) != IPSEC_LEVEL_UNIQUE) ? \
1318 (ipsec_debug ? log(LOG_INFO, "fixed system default level " #lev \
1319 ":%d->%d\n", (lev), IPSEC_LEVEL_REQUIRE) : (void)0), \
1320 (lev) = IPSEC_LEVEL_REQUIRE, (lev) \
1321 : (lev))
1322
1323 /* set default level */
1324 switch (((struct sockaddr *)&isr->sp->spidx.src)->sa_family) {
1325 #ifdef INET
1326 case AF_INET:
1327 esp_trans_deflev = IPSEC_CHECK_DEFAULT(ip4_esp_trans_deflev);
1328 esp_net_deflev = IPSEC_CHECK_DEFAULT(ip4_esp_net_deflev);
1329 ah_trans_deflev = IPSEC_CHECK_DEFAULT(ip4_ah_trans_deflev);
1330 ah_net_deflev = IPSEC_CHECK_DEFAULT(ip4_ah_net_deflev);
1331 break;
1332 #endif
1333 #ifdef INET6
1334 case AF_INET6:
1335 esp_trans_deflev = IPSEC_CHECK_DEFAULT(ip6_esp_trans_deflev);
1336 esp_net_deflev = IPSEC_CHECK_DEFAULT(ip6_esp_net_deflev);
1337 ah_trans_deflev = IPSEC_CHECK_DEFAULT(ip6_ah_trans_deflev);
1338 ah_net_deflev = IPSEC_CHECK_DEFAULT(ip6_ah_net_deflev);
1339 break;
1340 #endif /* INET6 */
1341 default:
1342 panic("%s: unknown af %u", __func__,
1343 isr->sp->spidx.src.sa.sa_family);
1344 }
1345
1346 #undef IPSEC_CHECK_DEFAULT
1347
1348 /* set level */
1349 switch (isr->level) {
1350 case IPSEC_LEVEL_DEFAULT:
1351 switch (isr->saidx.proto) {
1352 case IPPROTO_ESP:
1353 if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1354 level = esp_net_deflev;
1355 else
1356 level = esp_trans_deflev;
1357 break;
1358 case IPPROTO_AH:
1359 if (isr->saidx.mode == IPSEC_MODE_TUNNEL)
1360 level = ah_net_deflev;
1361 else
1362 level = ah_trans_deflev;
1363 break;
1364 case IPPROTO_IPCOMP:
1365 /*
1366 * we don't really care, as IPcomp document says that
1367 * we shouldn't compress small packets
1368 */
1369 level = IPSEC_LEVEL_USE;
1370 break;
1371 default:
1372 panic("%s: Illegal protocol defined %u", __func__,
1373 isr->saidx.proto);
1374 }
1375 break;
1376
1377 case IPSEC_LEVEL_USE:
1378 case IPSEC_LEVEL_REQUIRE:
1379 level = isr->level;
1380 break;
1381 case IPSEC_LEVEL_UNIQUE:
1382 level = IPSEC_LEVEL_REQUIRE;
1383 break;
1384
1385 default:
1386 panic("%s: Illegal IPsec level %u", __func__, isr->level);
1387 }
1388
1389 return level;
1390 }
1391
1392 /*
1393 * Check security policy requirements against the actual packet contents.
1394 *
1395 * If the SP requires an IPsec packet, and the packet was neither AH nor ESP,
1396 * then kick it.
1397 */
1398 static int
1399 ipsec_sp_reject(const struct secpolicy *sp, const struct mbuf *m)
1400 {
1401 struct ipsecrequest *isr;
1402
1403 if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
1404 printf("%s: using SP\n", __func__);
1405 kdebug_secpolicy(sp);
1406 }
1407
1408 /* check policy */
1409 switch (sp->policy) {
1410 case IPSEC_POLICY_DISCARD:
1411 return 1;
1412 case IPSEC_POLICY_BYPASS:
1413 case IPSEC_POLICY_NONE:
1414 return 0;
1415 }
1416
1417 KASSERTMSG(sp->policy == IPSEC_POLICY_IPSEC,
1418 "invalid policy %u", sp->policy);
1419
1420 /* XXX should compare policy against ipsec header history */
1421
1422 for (isr = sp->req; isr != NULL; isr = isr->next) {
1423 if (ipsec_get_reqlevel(isr) != IPSEC_LEVEL_REQUIRE)
1424 continue;
1425 switch (isr->saidx.proto) {
1426 case IPPROTO_ESP:
1427 if ((m->m_flags & M_DECRYPTED) == 0) {
1428 KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_DUMP,
1429 "ESP m_flags:%x\n", m->m_flags);
1430 return 1;
1431 }
1432 break;
1433 case IPPROTO_AH:
1434 if ((m->m_flags & M_AUTHIPHDR) == 0) {
1435 KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_DUMP,
1436 "AH m_flags:%x\n", m->m_flags);
1437 return 1;
1438 }
1439 break;
1440 case IPPROTO_IPCOMP:
1441 /*
1442 * We don't really care, as IPcomp document
1443 * says that we shouldn't compress small
1444 * packets, IPComp policy should always be
1445 * treated as being in "use" level.
1446 */
1447 break;
1448 }
1449 }
1450
1451 return 0;
1452 }
1453
1454 /*
1455 * Check security policy requirements.
1456 */
1457 int
1458 ipsec_in_reject(struct mbuf *m, void *inp)
1459 {
1460 struct inpcb_hdr *inph = (struct inpcb_hdr *)inp;
1461 struct secpolicy *sp;
1462 int error;
1463 int result;
1464
1465 KASSERT(m != NULL);
1466
1467 if (inph == NULL)
1468 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_INBOUND,
1469 IP_FORWARDING, &error);
1470 else
1471 sp = ipsec_getpolicybysock(m, IPSEC_DIR_INBOUND,
1472 inph, &error);
1473
1474 if (sp != NULL) {
1475 result = ipsec_sp_reject(sp, m);
1476 if (result)
1477 IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
1478 KEY_SP_UNREF(&sp);
1479 } else {
1480 result = 0;
1481 }
1482 return result;
1483 }
1484
1485 /*
1486 * Compute the byte size to be occupied by the IPsec header. If it is
1487 * tunneled, it includes the size of outer IP header.
1488 */
1489 static size_t
1490 ipsec_sp_hdrsiz(const struct secpolicy *sp, const struct mbuf *m)
1491 {
1492 struct ipsecrequest *isr;
1493 size_t siz;
1494
1495 if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
1496 printf("%s: using SP\n", __func__);
1497 kdebug_secpolicy(sp);
1498 }
1499
1500 switch (sp->policy) {
1501 case IPSEC_POLICY_DISCARD:
1502 case IPSEC_POLICY_BYPASS:
1503 case IPSEC_POLICY_NONE:
1504 return 0;
1505 }
1506
1507 KASSERTMSG(sp->policy == IPSEC_POLICY_IPSEC,
1508 "invalid policy %u", sp->policy);
1509
1510 siz = 0;
1511 for (isr = sp->req; isr != NULL; isr = isr->next) {
1512 size_t clen = 0;
1513 struct secasvar *sav;
1514
1515 switch (isr->saidx.proto) {
1516 case IPPROTO_ESP:
1517 sav = ipsec_lookup_sa(isr, m);
1518 if (sav != NULL) {
1519 clen = esp_hdrsiz(sav);
1520 KEY_SA_UNREF(&sav);
1521 } else
1522 clen = esp_hdrsiz(NULL);
1523 break;
1524 case IPPROTO_AH:
1525 sav = ipsec_lookup_sa(isr, m);
1526 if (sav != NULL) {
1527 clen = ah_hdrsiz(sav);
1528 KEY_SA_UNREF(&sav);
1529 } else
1530 clen = ah_hdrsiz(NULL);
1531 break;
1532 case IPPROTO_IPCOMP:
1533 clen = sizeof(struct ipcomp);
1534 break;
1535 }
1536
1537 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1538 switch (isr->saidx.dst.sa.sa_family) {
1539 case AF_INET:
1540 clen += sizeof(struct ip);
1541 break;
1542 #ifdef INET6
1543 case AF_INET6:
1544 clen += sizeof(struct ip6_hdr);
1545 break;
1546 #endif
1547 default:
1548 IPSECLOG(LOG_ERR, "unknown AF %d in "
1549 "IPsec tunnel SA\n",
1550 ((const struct sockaddr *)&isr->saidx.dst)
1551 ->sa_family);
1552 break;
1553 }
1554 }
1555 siz += clen;
1556 }
1557
1558 return siz;
1559 }
1560
1561 size_t
1562 ipsec_hdrsiz(struct mbuf *m, u_int dir, void *inp)
1563 {
1564 struct inpcb_hdr *inph = (struct inpcb_hdr *)inp;
1565 struct secpolicy *sp;
1566 int error;
1567 size_t size;
1568
1569 KASSERT(m != NULL);
1570 KASSERTMSG(inph == NULL || inph->inph_socket != NULL,
1571 "socket w/o inpcb");
1572
1573 if (inph == NULL)
1574 sp = ipsec_getpolicybyaddr(m, dir, IP_FORWARDING, &error);
1575 else
1576 sp = ipsec_getpolicybysock(m, dir, inph, &error);
1577
1578 if (sp != NULL) {
1579 size = ipsec_sp_hdrsiz(sp, m);
1580 KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_DATA, "size:%zu.\n", size);
1581 KEY_SP_UNREF(&sp);
1582 } else {
1583 size = 0;
1584 }
1585
1586 return size;
1587 }
1588
1589 /*
1590 * Check the variable replay window.
1591 * ipsec_chkreplay() performs replay check before ICV verification.
1592 * ipsec_updatereplay() updates replay bitmap. This must be called after
1593 * ICV verification (it also performs replay check, which is usually done
1594 * beforehand).
1595 * 0 (zero) is returned if packet disallowed, 1 if packet permitted.
1596 *
1597 * based on RFC 2401.
1598 */
1599 int
1600 ipsec_chkreplay(u_int32_t seq, const struct secasvar *sav)
1601 {
1602 const struct secreplay *replay;
1603 u_int32_t diff;
1604 int fr;
1605 u_int32_t wsizeb; /* constant: bits of window size */
1606 int frlast; /* constant: last frame */
1607
1608 IPSEC_SPLASSERT_SOFTNET(__func__);
1609
1610 KASSERT(sav != NULL);
1611 KASSERT(sav->replay != NULL);
1612
1613 replay = sav->replay;
1614
1615 if (replay->wsize == 0)
1616 return 1; /* no need to check replay. */
1617
1618 /* constant */
1619 frlast = replay->wsize - 1;
1620 wsizeb = replay->wsize << 3;
1621
1622 /* sequence number of 0 is invalid */
1623 if (seq == 0)
1624 return 0;
1625
1626 /* first time is always okay */
1627 if (replay->count == 0)
1628 return 1;
1629
1630 if (seq > replay->lastseq) {
1631 /* larger sequences are okay */
1632 return 1;
1633 } else {
1634 /* seq is equal or less than lastseq. */
1635 diff = replay->lastseq - seq;
1636
1637 /* over range to check, i.e. too old or wrapped */
1638 if (diff >= wsizeb)
1639 return 0;
1640
1641 fr = frlast - diff / 8;
1642
1643 /* this packet already seen ? */
1644 if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1645 return 0;
1646
1647 /* out of order but good */
1648 return 1;
1649 }
1650 }
1651
1652 /*
1653 * check replay counter whether to update or not.
1654 * OUT: 0: OK
1655 * 1: NG
1656 */
1657 int
1658 ipsec_updatereplay(u_int32_t seq, const struct secasvar *sav)
1659 {
1660 struct secreplay *replay;
1661 u_int32_t diff;
1662 int fr;
1663 u_int32_t wsizeb; /* constant: bits of window size */
1664 int frlast; /* constant: last frame */
1665
1666 IPSEC_SPLASSERT_SOFTNET(__func__);
1667
1668 KASSERT(sav != NULL);
1669 KASSERT(sav->replay != NULL);
1670
1671 replay = sav->replay;
1672
1673 if (replay->wsize == 0)
1674 goto ok; /* no need to check replay. */
1675
1676 /* constant */
1677 frlast = replay->wsize - 1;
1678 wsizeb = replay->wsize << 3;
1679
1680 /* sequence number of 0 is invalid */
1681 if (seq == 0)
1682 return 1;
1683
1684 /* first time */
1685 if (replay->count == 0) {
1686 replay->lastseq = seq;
1687 memset(replay->bitmap, 0, replay->wsize);
1688 (replay->bitmap)[frlast] = 1;
1689 goto ok;
1690 }
1691
1692 if (seq > replay->lastseq) {
1693 /* seq is larger than lastseq. */
1694 diff = seq - replay->lastseq;
1695
1696 /* new larger sequence number */
1697 if (diff < wsizeb) {
1698 /* In window */
1699 /* set bit for this packet */
1700 vshiftl(replay->bitmap, diff, replay->wsize);
1701 (replay->bitmap)[frlast] |= 1;
1702 } else {
1703 /* this packet has a "way larger" */
1704 memset(replay->bitmap, 0, replay->wsize);
1705 (replay->bitmap)[frlast] = 1;
1706 }
1707 replay->lastseq = seq;
1708
1709 /* larger is good */
1710 } else {
1711 /* seq is equal or less than lastseq. */
1712 diff = replay->lastseq - seq;
1713
1714 /* over range to check, i.e. too old or wrapped */
1715 if (diff >= wsizeb)
1716 return 1;
1717
1718 fr = frlast - diff / 8;
1719
1720 /* this packet already seen ? */
1721 if ((replay->bitmap)[fr] & (1 << (diff % 8)))
1722 return 1;
1723
1724 /* mark as seen */
1725 (replay->bitmap)[fr] |= (1 << (diff % 8));
1726
1727 /* out of order but good */
1728 }
1729
1730 ok:
1731 if (replay->count == ~0) {
1732 char buf[IPSEC_LOGSASTRLEN];
1733
1734 /* set overflow flag */
1735 replay->overflow++;
1736
1737 /* don't increment, no more packets accepted */
1738 if ((sav->flags & SADB_X_EXT_CYCSEQ) == 0)
1739 return 1;
1740
1741 IPSECLOG(LOG_WARNING, "replay counter made %d cycle. %s\n",
1742 replay->overflow, ipsec_logsastr(sav, buf, sizeof(buf)));
1743 }
1744
1745 replay->count++;
1746
1747 return 0;
1748 }
1749
1750 /*
1751 * shift variable length buffer to left.
1752 * IN: bitmap: pointer to the buffer
1753 * nbit: the number of to shift.
1754 * wsize: buffer size (bytes).
1755 */
1756 static void
1757 vshiftl(unsigned char *bitmap, int nbit, int wsize)
1758 {
1759 int s, j, i;
1760 unsigned char over;
1761
1762 for (j = 0; j < nbit; j += 8) {
1763 s = (nbit - j < 8) ? (nbit - j): 8;
1764 bitmap[0] <<= s;
1765 for (i = 1; i < wsize; i++) {
1766 over = (bitmap[i] >> (8 - s));
1767 bitmap[i] <<= s;
1768 bitmap[i-1] |= over;
1769 }
1770 }
1771
1772 return;
1773 }
1774
1775 /* Return a printable string for the address. */
1776 const char *
1777 ipsec_address(const union sockaddr_union *sa, char *buf, size_t size)
1778 {
1779 switch (sa->sa.sa_family) {
1780 #if INET
1781 case AF_INET:
1782 in_print(buf, size, &sa->sin.sin_addr);
1783 return buf;
1784 #endif
1785 #if INET6
1786 case AF_INET6:
1787 in6_print(buf, size, &sa->sin6.sin6_addr);
1788 return buf;
1789 #endif
1790 default:
1791 return "(unknown address family)";
1792 }
1793 }
1794
1795 const char *
1796 ipsec_logsastr(const struct secasvar *sav, char *buf, size_t size)
1797 {
1798 const struct secasindex *saidx = &sav->sah->saidx;
1799 char sbuf[IPSEC_ADDRSTRLEN], dbuf[IPSEC_ADDRSTRLEN];
1800
1801 KASSERTMSG(saidx->src.sa.sa_family == saidx->dst.sa.sa_family,
1802 "af family mismatch, src %u, dst %u",
1803 saidx->src.sa.sa_family, saidx->dst.sa.sa_family);
1804
1805 snprintf(buf, size, "SA(SPI=%u src=%s dst=%s)",
1806 (u_int32_t)ntohl(sav->spi),
1807 ipsec_address(&saidx->src, sbuf, sizeof(sbuf)),
1808 ipsec_address(&saidx->dst, dbuf, sizeof(dbuf)));
1809
1810 return buf;
1811 }
1812
1813 void
1814 ipsec_dumpmbuf(struct mbuf *m)
1815 {
1816 int totlen;
1817 int i;
1818 u_char *p;
1819
1820 totlen = 0;
1821 printf("---\n");
1822 while (m) {
1823 p = mtod(m, u_char *);
1824 for (i = 0; i < m->m_len; i++) {
1825 printf("%02x ", p[i]);
1826 totlen++;
1827 if (totlen % 16 == 0)
1828 printf("\n");
1829 }
1830 m = m->m_next;
1831 }
1832 if (totlen % 16 != 0)
1833 printf("\n");
1834 printf("---\n");
1835 }
1836
1837 #ifdef INET6
1838 struct secpolicy *
1839 ipsec6_check_policy(struct mbuf *m, struct in6pcb *in6p, int flags,
1840 int *needipsecp, int *errorp)
1841 {
1842 struct secpolicy *sp = NULL;
1843 int s;
1844 int error = 0;
1845 int needipsec = 0;
1846
1847 if (ipsec_outdone(m)) {
1848 goto skippolicycheck;
1849 }
1850 s = splsoftnet();
1851 if (in6p && ipsec_pcb_skip_ipsec(in6p->in6p_sp, IPSEC_DIR_OUTBOUND)) {
1852 splx(s);
1853 goto skippolicycheck;
1854 }
1855 sp = ipsec_checkpolicy(m, IPSEC_DIR_OUTBOUND, flags, &error, in6p);
1856 splx(s);
1857
1858 /*
1859 * There are four return cases:
1860 * sp != NULL apply IPsec policy
1861 * sp == NULL, error == 0 no IPsec handling needed
1862 * sp == NULL, error == -EINVAL discard packet w/o error
1863 * sp == NULL, error != 0 discard packet, report error
1864 */
1865 if (sp == NULL) {
1866 needipsec = 0;
1867 } else {
1868 needipsec = 1;
1869 }
1870
1871 skippolicycheck:
1872 *errorp = error;
1873 *needipsecp = needipsec;
1874 return sp;
1875 }
1876
1877 int
1878 ipsec6_input(struct mbuf *m)
1879 {
1880 int s, error;
1881
1882 s = splsoftnet();
1883 error = ipsec_in_reject(m, NULL);
1884 splx(s);
1885 if (error) {
1886 return EINVAL;
1887 }
1888
1889 return 0;
1890 }
1891 #endif /* INET6 */
1892
1893 /*
1894 * -----------------------------------------------------------------------------
1895 */
1896
1897 /* XXX this stuff doesn't belong here... */
1898
1899 static struct xformsw *xforms = NULL;
1900
1901 /*
1902 * Register a transform; typically at system startup.
1903 */
1904 void
1905 xform_register(struct xformsw *xsp)
1906 {
1907 xsp->xf_next = xforms;
1908 xforms = xsp;
1909 }
1910
1911 /*
1912 * Initialize transform support in an sav.
1913 */
1914 int
1915 xform_init(struct secasvar *sav, int xftype)
1916 {
1917 struct xformsw *xsp;
1918
1919 if (sav->tdb_xform != NULL) /* previously initialized */
1920 return 0;
1921 for (xsp = xforms; xsp; xsp = xsp->xf_next)
1922 if (xsp->xf_type == xftype)
1923 return (*xsp->xf_init)(sav, xsp);
1924
1925 IPSECLOG(LOG_DEBUG, "no match for xform type %d\n", xftype);
1926 return EINVAL;
1927 }
1928
1929 void
1930 nat_t_ports_get(struct mbuf *m, u_int16_t *dport, u_int16_t *sport)
1931 {
1932 struct m_tag *tag;
1933
1934 if ((tag = m_tag_find(m, PACKET_TAG_IPSEC_NAT_T_PORTS, NULL))) {
1935 *sport = ((u_int16_t *)(tag + 1))[0];
1936 *dport = ((u_int16_t *)(tag + 1))[1];
1937 } else
1938 *sport = *dport = 0;
1939 }
1940
1941 /*
1942 * XXXJRT This should be done as a protosw init call.
1943 */
1944 void
1945 ipsec_attach(void)
1946 {
1947
1948 ipsec_output_init();
1949
1950 ipsecstat_percpu = percpu_alloc(sizeof(uint64_t) * IPSEC_NSTATS);
1951
1952 sysctl_net_inet_ipsec_setup(NULL);
1953 #ifdef INET6
1954 sysctl_net_inet6_ipsec6_setup(NULL);
1955 #endif
1956
1957 ah_attach();
1958 esp_attach();
1959 ipcomp_attach();
1960 ipe4_attach();
1961 #ifdef TCP_SIGNATURE
1962 tcpsignature_attach();
1963 #endif
1964 }
1965