xref: /freebsd/sys/netipsec/key.c (revision 0957b409)
1 /*	$FreeBSD$	*/
2 /*	$KAME: key.c,v 1.191 2001/06/27 10:46:49 sakane Exp $	*/
3 
4 /*-
5  * SPDX-License-Identifier: BSD-3-Clause
6  *
7  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
8  * All rights reserved.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the project nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 /*
36  * This code is referd to RFC 2367
37  */
38 
39 #include "opt_inet.h"
40 #include "opt_inet6.h"
41 #include "opt_ipsec.h"
42 
43 #include <sys/types.h>
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/fnv_hash.h>
48 #include <sys/lock.h>
49 #include <sys/mutex.h>
50 #include <sys/mbuf.h>
51 #include <sys/domain.h>
52 #include <sys/protosw.h>
53 #include <sys/malloc.h>
54 #include <sys/rmlock.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sysctl.h>
58 #include <sys/errno.h>
59 #include <sys/proc.h>
60 #include <sys/queue.h>
61 #include <sys/refcount.h>
62 #include <sys/syslog.h>
63 
64 #include <vm/uma.h>
65 
66 #include <net/if.h>
67 #include <net/if_var.h>
68 #include <net/vnet.h>
69 #include <net/raw_cb.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_systm.h>
73 #include <netinet/ip.h>
74 #include <netinet/in_var.h>
75 #include <netinet/udp.h>
76 
77 #ifdef INET6
78 #include <netinet/ip6.h>
79 #include <netinet6/in6_var.h>
80 #include <netinet6/ip6_var.h>
81 #endif /* INET6 */
82 
83 #include <net/pfkeyv2.h>
84 #include <netipsec/keydb.h>
85 #include <netipsec/key.h>
86 #include <netipsec/keysock.h>
87 #include <netipsec/key_debug.h>
88 
89 #include <netipsec/ipsec.h>
90 #ifdef INET6
91 #include <netipsec/ipsec6.h>
92 #endif
93 
94 #include <netipsec/xform.h>
95 #include <machine/in_cksum.h>
96 #include <machine/stdarg.h>
97 
98 /* randomness */
99 #include <sys/random.h>
100 
101 #define FULLMASK	0xff
102 #define	_BITS(bytes)	((bytes) << 3)
103 
104 /*
105  * Note on SA reference counting:
106  * - SAs that are not in DEAD state will have (total external reference + 1)
107  *   following value in reference count field.  they cannot be freed and are
108  *   referenced from SA header.
109  * - SAs that are in DEAD state will have (total external reference)
110  *   in reference count field.  they are ready to be freed.  reference from
111  *   SA header will be removed in key_delsav(), when the reference count
112  *   field hits 0 (= no external reference other than from SA header.
113  */
114 
115 VNET_DEFINE(u_int32_t, key_debug_level) = 0;
116 VNET_DEFINE_STATIC(u_int, key_spi_trycnt) = 1000;
117 VNET_DEFINE_STATIC(u_int32_t, key_spi_minval) = 0x100;
118 VNET_DEFINE_STATIC(u_int32_t, key_spi_maxval) = 0x0fffffff;	/* XXX */
119 VNET_DEFINE_STATIC(u_int32_t, policy_id) = 0;
120 /*interval to initialize randseed,1(m)*/
121 VNET_DEFINE_STATIC(u_int, key_int_random) = 60;
122 /* interval to expire acquiring, 30(s)*/
123 VNET_DEFINE_STATIC(u_int, key_larval_lifetime) = 30;
124 /* counter for blocking SADB_ACQUIRE.*/
125 VNET_DEFINE_STATIC(int, key_blockacq_count) = 10;
126 /* lifetime for blocking SADB_ACQUIRE.*/
127 VNET_DEFINE_STATIC(int, key_blockacq_lifetime) = 20;
128 /* preferred old sa rather than new sa.*/
129 VNET_DEFINE_STATIC(int, key_preferred_oldsa) = 1;
130 #define	V_key_spi_trycnt	VNET(key_spi_trycnt)
131 #define	V_key_spi_minval	VNET(key_spi_minval)
132 #define	V_key_spi_maxval	VNET(key_spi_maxval)
133 #define	V_policy_id		VNET(policy_id)
134 #define	V_key_int_random	VNET(key_int_random)
135 #define	V_key_larval_lifetime	VNET(key_larval_lifetime)
136 #define	V_key_blockacq_count	VNET(key_blockacq_count)
137 #define	V_key_blockacq_lifetime	VNET(key_blockacq_lifetime)
138 #define	V_key_preferred_oldsa	VNET(key_preferred_oldsa)
139 
140 VNET_DEFINE_STATIC(u_int32_t, acq_seq) = 0;
141 #define	V_acq_seq		VNET(acq_seq)
142 
143 VNET_DEFINE_STATIC(uint32_t, sp_genid) = 0;
144 #define	V_sp_genid		VNET(sp_genid)
145 
146 /* SPD */
147 TAILQ_HEAD(secpolicy_queue, secpolicy);
148 LIST_HEAD(secpolicy_list, secpolicy);
149 VNET_DEFINE_STATIC(struct secpolicy_queue, sptree[IPSEC_DIR_MAX]);
150 VNET_DEFINE_STATIC(struct secpolicy_queue, sptree_ifnet[IPSEC_DIR_MAX]);
151 static struct rmlock sptree_lock;
152 #define	V_sptree		VNET(sptree)
153 #define	V_sptree_ifnet		VNET(sptree_ifnet)
154 #define	SPTREE_LOCK_INIT()      rm_init(&sptree_lock, "sptree")
155 #define	SPTREE_LOCK_DESTROY()   rm_destroy(&sptree_lock)
156 #define	SPTREE_RLOCK_TRACKER    struct rm_priotracker sptree_tracker
157 #define	SPTREE_RLOCK()          rm_rlock(&sptree_lock, &sptree_tracker)
158 #define	SPTREE_RUNLOCK()        rm_runlock(&sptree_lock, &sptree_tracker)
159 #define	SPTREE_RLOCK_ASSERT()   rm_assert(&sptree_lock, RA_RLOCKED)
160 #define	SPTREE_WLOCK()          rm_wlock(&sptree_lock)
161 #define	SPTREE_WUNLOCK()        rm_wunlock(&sptree_lock)
162 #define	SPTREE_WLOCK_ASSERT()   rm_assert(&sptree_lock, RA_WLOCKED)
163 #define	SPTREE_UNLOCK_ASSERT()  rm_assert(&sptree_lock, RA_UNLOCKED)
164 
165 /* Hash table for lookup SP using unique id */
166 VNET_DEFINE_STATIC(struct secpolicy_list *, sphashtbl);
167 VNET_DEFINE_STATIC(u_long, sphash_mask);
168 #define	V_sphashtbl		VNET(sphashtbl)
169 #define	V_sphash_mask		VNET(sphash_mask)
170 
171 #define	SPHASH_NHASH_LOG2	7
172 #define	SPHASH_NHASH		(1 << SPHASH_NHASH_LOG2)
173 #define	SPHASH_HASHVAL(id)	(key_u32hash(id) & V_sphash_mask)
174 #define	SPHASH_HASH(id)		&V_sphashtbl[SPHASH_HASHVAL(id)]
175 
176 /* SPD cache */
177 struct spdcache_entry {
178    struct secpolicyindex spidx;	/* secpolicyindex */
179    struct secpolicy *sp;	/* cached policy to be used */
180 
181    LIST_ENTRY(spdcache_entry) chain;
182 };
183 LIST_HEAD(spdcache_entry_list, spdcache_entry);
184 
185 #define	SPDCACHE_MAX_ENTRIES_PER_HASH	8
186 
187 VNET_DEFINE_STATIC(u_int, key_spdcache_maxentries) = 0;
188 #define	V_key_spdcache_maxentries	VNET(key_spdcache_maxentries)
189 VNET_DEFINE_STATIC(u_int, key_spdcache_threshold) = 32;
190 #define	V_key_spdcache_threshold	VNET(key_spdcache_threshold)
191 VNET_DEFINE_STATIC(unsigned long, spd_size) = 0;
192 #define	V_spd_size		VNET(spd_size)
193 
194 #define SPDCACHE_ENABLED()	(V_key_spdcache_maxentries != 0)
195 #define SPDCACHE_ACTIVE() \
196 	(SPDCACHE_ENABLED() && V_spd_size >= V_key_spdcache_threshold)
197 
198 VNET_DEFINE_STATIC(struct spdcache_entry_list *, spdcachehashtbl);
199 VNET_DEFINE_STATIC(u_long, spdcachehash_mask);
200 #define	V_spdcachehashtbl	VNET(spdcachehashtbl)
201 #define	V_spdcachehash_mask	VNET(spdcachehash_mask)
202 
203 #define	SPDCACHE_HASHVAL(idx) \
204 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->ul_proto) &  \
205 	    V_spdcachehash_mask)
206 
207 /* Each cache line is protected by a mutex */
208 VNET_DEFINE_STATIC(struct mtx *, spdcache_lock);
209 #define	V_spdcache_lock		VNET(spdcache_lock)
210 
211 #define	SPDCACHE_LOCK_INIT(a) \
212 	mtx_init(&V_spdcache_lock[a], "spdcache", \
213 	    "fast ipsec SPD cache", MTX_DEF|MTX_DUPOK)
214 #define	SPDCACHE_LOCK_DESTROY(a)	mtx_destroy(&V_spdcache_lock[a])
215 #define	SPDCACHE_LOCK(a)		mtx_lock(&V_spdcache_lock[a]);
216 #define	SPDCACHE_UNLOCK(a)		mtx_unlock(&V_spdcache_lock[a]);
217 
218 /* SAD */
219 TAILQ_HEAD(secashead_queue, secashead);
220 LIST_HEAD(secashead_list, secashead);
221 VNET_DEFINE_STATIC(struct secashead_queue, sahtree);
222 static struct rmlock sahtree_lock;
223 #define	V_sahtree		VNET(sahtree)
224 #define	SAHTREE_LOCK_INIT()	rm_init(&sahtree_lock, "sahtree")
225 #define	SAHTREE_LOCK_DESTROY()	rm_destroy(&sahtree_lock)
226 #define	SAHTREE_RLOCK_TRACKER	struct rm_priotracker sahtree_tracker
227 #define	SAHTREE_RLOCK()		rm_rlock(&sahtree_lock, &sahtree_tracker)
228 #define	SAHTREE_RUNLOCK()	rm_runlock(&sahtree_lock, &sahtree_tracker)
229 #define	SAHTREE_RLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_RLOCKED)
230 #define	SAHTREE_WLOCK()		rm_wlock(&sahtree_lock)
231 #define	SAHTREE_WUNLOCK()	rm_wunlock(&sahtree_lock)
232 #define	SAHTREE_WLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_WLOCKED)
233 #define	SAHTREE_UNLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_UNLOCKED)
234 
235 /* Hash table for lookup in SAD using SA addresses */
236 VNET_DEFINE_STATIC(struct secashead_list *, sahaddrhashtbl);
237 VNET_DEFINE_STATIC(u_long, sahaddrhash_mask);
238 #define	V_sahaddrhashtbl	VNET(sahaddrhashtbl)
239 #define	V_sahaddrhash_mask	VNET(sahaddrhash_mask)
240 
241 #define	SAHHASH_NHASH_LOG2	7
242 #define	SAHHASH_NHASH		(1 << SAHHASH_NHASH_LOG2)
243 #define	SAHADDRHASH_HASHVAL(idx)	\
244 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->proto) & \
245 	    V_sahaddrhash_mask)
246 #define	SAHADDRHASH_HASH(saidx)		\
247     &V_sahaddrhashtbl[SAHADDRHASH_HASHVAL(saidx)]
248 
249 /* Hash table for lookup in SAD using SPI */
250 LIST_HEAD(secasvar_list, secasvar);
251 VNET_DEFINE_STATIC(struct secasvar_list *, savhashtbl);
252 VNET_DEFINE_STATIC(u_long, savhash_mask);
253 #define	V_savhashtbl		VNET(savhashtbl)
254 #define	V_savhash_mask		VNET(savhash_mask)
255 #define	SAVHASH_NHASH_LOG2	7
256 #define	SAVHASH_NHASH		(1 << SAVHASH_NHASH_LOG2)
257 #define	SAVHASH_HASHVAL(spi)	(key_u32hash(spi) & V_savhash_mask)
258 #define	SAVHASH_HASH(spi)	&V_savhashtbl[SAVHASH_HASHVAL(spi)]
259 
260 static uint32_t
261 key_addrprotohash(const union sockaddr_union *src,
262     const union sockaddr_union *dst, const uint8_t *proto)
263 {
264 	uint32_t hval;
265 
266 	hval = fnv_32_buf(proto, sizeof(*proto),
267 	    FNV1_32_INIT);
268 	switch (dst->sa.sa_family) {
269 #ifdef INET
270 	case AF_INET:
271 		hval = fnv_32_buf(&src->sin.sin_addr,
272 		    sizeof(in_addr_t), hval);
273 		hval = fnv_32_buf(&dst->sin.sin_addr,
274 		    sizeof(in_addr_t), hval);
275 		break;
276 #endif
277 #ifdef INET6
278 	case AF_INET6:
279 		hval = fnv_32_buf(&src->sin6.sin6_addr,
280 		    sizeof(struct in6_addr), hval);
281 		hval = fnv_32_buf(&dst->sin6.sin6_addr,
282 		    sizeof(struct in6_addr), hval);
283 		break;
284 #endif
285 	default:
286 		hval = 0;
287 		ipseclog((LOG_DEBUG, "%s: unknown address family %d",
288 		    __func__, dst->sa.sa_family));
289 	}
290 	return (hval);
291 }
292 
293 static uint32_t
294 key_u32hash(uint32_t val)
295 {
296 
297 	return (fnv_32_buf(&val, sizeof(val), FNV1_32_INIT));
298 }
299 
300 							/* registed list */
301 VNET_DEFINE_STATIC(LIST_HEAD(_regtree, secreg), regtree[SADB_SATYPE_MAX + 1]);
302 #define	V_regtree		VNET(regtree)
303 static struct mtx regtree_lock;
304 #define	REGTREE_LOCK_INIT() \
305 	mtx_init(&regtree_lock, "regtree", "fast ipsec regtree", MTX_DEF)
306 #define	REGTREE_LOCK_DESTROY()	mtx_destroy(&regtree_lock)
307 #define	REGTREE_LOCK()		mtx_lock(&regtree_lock)
308 #define	REGTREE_UNLOCK()	mtx_unlock(&regtree_lock)
309 #define	REGTREE_LOCK_ASSERT()	mtx_assert(&regtree_lock, MA_OWNED)
310 
311 /* Acquiring list */
312 LIST_HEAD(secacq_list, secacq);
313 VNET_DEFINE_STATIC(struct secacq_list, acqtree);
314 #define	V_acqtree		VNET(acqtree)
315 static struct mtx acq_lock;
316 #define	ACQ_LOCK_INIT() \
317     mtx_init(&acq_lock, "acqtree", "ipsec SA acquiring list", MTX_DEF)
318 #define	ACQ_LOCK_DESTROY()	mtx_destroy(&acq_lock)
319 #define	ACQ_LOCK()		mtx_lock(&acq_lock)
320 #define	ACQ_UNLOCK()		mtx_unlock(&acq_lock)
321 #define	ACQ_LOCK_ASSERT()	mtx_assert(&acq_lock, MA_OWNED)
322 
323 /* Hash table for lookup in ACQ list using SA addresses */
324 VNET_DEFINE_STATIC(struct secacq_list *, acqaddrhashtbl);
325 VNET_DEFINE_STATIC(u_long, acqaddrhash_mask);
326 #define	V_acqaddrhashtbl	VNET(acqaddrhashtbl)
327 #define	V_acqaddrhash_mask	VNET(acqaddrhash_mask)
328 
329 /* Hash table for lookup in ACQ list using SEQ number */
330 VNET_DEFINE_STATIC(struct secacq_list *, acqseqhashtbl);
331 VNET_DEFINE_STATIC(u_long, acqseqhash_mask);
332 #define	V_acqseqhashtbl		VNET(acqseqhashtbl)
333 #define	V_acqseqhash_mask	VNET(acqseqhash_mask)
334 
335 #define	ACQHASH_NHASH_LOG2	7
336 #define	ACQHASH_NHASH		(1 << ACQHASH_NHASH_LOG2)
337 #define	ACQADDRHASH_HASHVAL(idx)	\
338 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->proto) & \
339 	    V_acqaddrhash_mask)
340 #define	ACQSEQHASH_HASHVAL(seq)		\
341     (key_u32hash(seq) & V_acqseqhash_mask)
342 #define	ACQADDRHASH_HASH(saidx)	\
343     &V_acqaddrhashtbl[ACQADDRHASH_HASHVAL(saidx)]
344 #define	ACQSEQHASH_HASH(seq)	\
345     &V_acqseqhashtbl[ACQSEQHASH_HASHVAL(seq)]
346 							/* SP acquiring list */
347 VNET_DEFINE_STATIC(LIST_HEAD(_spacqtree, secspacq), spacqtree);
348 #define	V_spacqtree		VNET(spacqtree)
349 static struct mtx spacq_lock;
350 #define	SPACQ_LOCK_INIT() \
351 	mtx_init(&spacq_lock, "spacqtree", \
352 		"fast ipsec security policy acquire list", MTX_DEF)
353 #define	SPACQ_LOCK_DESTROY()	mtx_destroy(&spacq_lock)
354 #define	SPACQ_LOCK()		mtx_lock(&spacq_lock)
355 #define	SPACQ_UNLOCK()		mtx_unlock(&spacq_lock)
356 #define	SPACQ_LOCK_ASSERT()	mtx_assert(&spacq_lock, MA_OWNED)
357 
358 static const int minsize[] = {
359 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
360 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
361 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
362 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
363 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
364 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_SRC */
365 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_DST */
366 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_PROXY */
367 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_AUTH */
368 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_ENCRYPT */
369 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_SRC */
370 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_DST */
371 	sizeof(struct sadb_sens),	/* SADB_EXT_SENSITIVITY */
372 	sizeof(struct sadb_prop),	/* SADB_EXT_PROPOSAL */
373 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_AUTH */
374 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_ENCRYPT */
375 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
376 	0,				/* SADB_X_EXT_KMPRIVATE */
377 	sizeof(struct sadb_x_policy),	/* SADB_X_EXT_POLICY */
378 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
379 	sizeof(struct sadb_x_nat_t_type),/* SADB_X_EXT_NAT_T_TYPE */
380 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_SPORT */
381 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_DPORT */
382 	sizeof(struct sadb_address),	/* SADB_X_EXT_NAT_T_OAI */
383 	sizeof(struct sadb_address),	/* SADB_X_EXT_NAT_T_OAR */
384 	sizeof(struct sadb_x_nat_t_frag),/* SADB_X_EXT_NAT_T_FRAG */
385 	sizeof(struct sadb_x_sa_replay), /* SADB_X_EXT_SA_REPLAY */
386 	sizeof(struct sadb_address),	/* SADB_X_EXT_NEW_ADDRESS_SRC */
387 	sizeof(struct sadb_address),	/* SADB_X_EXT_NEW_ADDRESS_DST */
388 };
389 _Static_assert(sizeof(minsize)/sizeof(int) == SADB_EXT_MAX + 1, "minsize size mismatch");
390 
391 static const int maxsize[] = {
392 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
393 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
394 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
395 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
396 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
397 	0,				/* SADB_EXT_ADDRESS_SRC */
398 	0,				/* SADB_EXT_ADDRESS_DST */
399 	0,				/* SADB_EXT_ADDRESS_PROXY */
400 	0,				/* SADB_EXT_KEY_AUTH */
401 	0,				/* SADB_EXT_KEY_ENCRYPT */
402 	0,				/* SADB_EXT_IDENTITY_SRC */
403 	0,				/* SADB_EXT_IDENTITY_DST */
404 	0,				/* SADB_EXT_SENSITIVITY */
405 	0,				/* SADB_EXT_PROPOSAL */
406 	0,				/* SADB_EXT_SUPPORTED_AUTH */
407 	0,				/* SADB_EXT_SUPPORTED_ENCRYPT */
408 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
409 	0,				/* SADB_X_EXT_KMPRIVATE */
410 	0,				/* SADB_X_EXT_POLICY */
411 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
412 	sizeof(struct sadb_x_nat_t_type),/* SADB_X_EXT_NAT_T_TYPE */
413 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_SPORT */
414 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_DPORT */
415 	0,				/* SADB_X_EXT_NAT_T_OAI */
416 	0,				/* SADB_X_EXT_NAT_T_OAR */
417 	sizeof(struct sadb_x_nat_t_frag),/* SADB_X_EXT_NAT_T_FRAG */
418 	sizeof(struct sadb_x_sa_replay), /* SADB_X_EXT_SA_REPLAY */
419 	0,				/* SADB_X_EXT_NEW_ADDRESS_SRC */
420 	0,				/* SADB_X_EXT_NEW_ADDRESS_DST */
421 };
422 _Static_assert(sizeof(maxsize)/sizeof(int) == SADB_EXT_MAX + 1, "minsize size mismatch");
423 
424 /*
425  * Internal values for SA flags:
426  * SADB_X_EXT_F_CLONED means that SA was cloned by key_updateaddresses,
427  *	thus we will not free the most of SA content in key_delsav().
428  */
429 #define	SADB_X_EXT_F_CLONED	0x80000000
430 
431 #define	SADB_CHECKLEN(_mhp, _ext)			\
432     ((_mhp)->extlen[(_ext)] < minsize[(_ext)] || (maxsize[(_ext)] != 0 && \
433 	((_mhp)->extlen[(_ext)] > maxsize[(_ext)])))
434 #define	SADB_CHECKHDR(_mhp, _ext)	((_mhp)->ext[(_ext)] == NULL)
435 
436 VNET_DEFINE_STATIC(int, ipsec_esp_keymin) = 256;
437 VNET_DEFINE_STATIC(int, ipsec_esp_auth) = 0;
438 VNET_DEFINE_STATIC(int, ipsec_ah_keymin) = 128;
439 
440 #define	V_ipsec_esp_keymin	VNET(ipsec_esp_keymin)
441 #define	V_ipsec_esp_auth	VNET(ipsec_esp_auth)
442 #define	V_ipsec_ah_keymin	VNET(ipsec_ah_keymin)
443 
444 #ifdef IPSEC_DEBUG
445 VNET_DEFINE(int, ipsec_debug) = 1;
446 #else
447 VNET_DEFINE(int, ipsec_debug) = 0;
448 #endif
449 
450 #ifdef INET
451 SYSCTL_DECL(_net_inet_ipsec);
452 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
453     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
454     "Enable IPsec debugging output when set.");
455 #endif
456 #ifdef INET6
457 SYSCTL_DECL(_net_inet6_ipsec6);
458 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug,
459     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
460     "Enable IPsec debugging output when set.");
461 #endif
462 
463 SYSCTL_DECL(_net_key);
464 SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,
465 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_debug_level), 0, "");
466 
467 /* max count of trial for the decision of spi value */
468 SYSCTL_INT(_net_key, KEYCTL_SPI_TRY, spi_trycnt,
469 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_trycnt), 0, "");
470 
471 /* minimum spi value to allocate automatically. */
472 SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE, spi_minval,
473 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_minval), 0, "");
474 
475 /* maximun spi value to allocate automatically. */
476 SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE, spi_maxval,
477 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_maxval), 0, "");
478 
479 /* interval to initialize randseed */
480 SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT, int_random,
481 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_int_random), 0, "");
482 
483 /* lifetime for larval SA */
484 SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME, larval_lifetime,
485 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_larval_lifetime), 0, "");
486 
487 /* counter for blocking to send SADB_ACQUIRE to IKEd */
488 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT, blockacq_count,
489 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_count), 0, "");
490 
491 /* lifetime for blocking to send SADB_ACQUIRE to IKEd */
492 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME, blockacq_lifetime,
493 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_lifetime), 0, "");
494 
495 /* ESP auth */
496 SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH, esp_auth,
497 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_auth), 0, "");
498 
499 /* minimum ESP key length */
500 SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN, esp_keymin,
501 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_keymin), 0, "");
502 
503 /* minimum AH key length */
504 SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN, ah_keymin,
505 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_ah_keymin), 0, "");
506 
507 /* perfered old SA rather than new SA */
508 SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA, preferred_oldsa,
509 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_preferred_oldsa), 0, "");
510 
511 static SYSCTL_NODE(_net_key, OID_AUTO, spdcache, CTLFLAG_RW, 0, "SPD cache");
512 
513 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, maxentries,
514 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_maxentries), 0,
515 	"Maximum number of entries in the SPD cache"
516 	" (power of 2, 0 to disable)");
517 
518 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, threshold,
519 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_threshold), 0,
520 	"Number of SPs that make the SPD cache active");
521 
522 #define __LIST_CHAINED(elm) \
523 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
524 
525 MALLOC_DEFINE(M_IPSEC_SA, "secasvar", "ipsec security association");
526 MALLOC_DEFINE(M_IPSEC_SAH, "sahead", "ipsec sa head");
527 MALLOC_DEFINE(M_IPSEC_SP, "ipsecpolicy", "ipsec security policy");
528 MALLOC_DEFINE(M_IPSEC_SR, "ipsecrequest", "ipsec security request");
529 MALLOC_DEFINE(M_IPSEC_MISC, "ipsec-misc", "ipsec miscellaneous");
530 MALLOC_DEFINE(M_IPSEC_SAQ, "ipsec-saq", "ipsec sa acquire");
531 MALLOC_DEFINE(M_IPSEC_SAR, "ipsec-reg", "ipsec sa acquire");
532 MALLOC_DEFINE(M_IPSEC_SPDCACHE, "ipsec-spdcache", "ipsec SPD cache");
533 
534 VNET_DEFINE_STATIC(uma_zone_t, key_lft_zone);
535 #define	V_key_lft_zone		VNET(key_lft_zone)
536 
537 /*
538  * set parameters into secpolicyindex buffer.
539  * Must allocate secpolicyindex buffer passed to this function.
540  */
541 #define KEY_SETSECSPIDX(_dir, s, d, ps, pd, ulp, idx) \
542 do { \
543 	bzero((idx), sizeof(struct secpolicyindex));                         \
544 	(idx)->dir = (_dir);                                                 \
545 	(idx)->prefs = (ps);                                                 \
546 	(idx)->prefd = (pd);                                                 \
547 	(idx)->ul_proto = (ulp);                                             \
548 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
549 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
550 } while (0)
551 
552 /*
553  * set parameters into secasindex buffer.
554  * Must allocate secasindex buffer before calling this function.
555  */
556 #define KEY_SETSECASIDX(p, m, r, s, d, idx) \
557 do { \
558 	bzero((idx), sizeof(struct secasindex));                             \
559 	(idx)->proto = (p);                                                  \
560 	(idx)->mode = (m);                                                   \
561 	(idx)->reqid = (r);                                                  \
562 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
563 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
564 	key_porttosaddr(&(idx)->src.sa, 0);				     \
565 	key_porttosaddr(&(idx)->dst.sa, 0);				     \
566 } while (0)
567 
568 /* key statistics */
569 struct _keystat {
570 	u_long getspi_count; /* the avarage of count to try to get new SPI */
571 } keystat;
572 
573 struct sadb_msghdr {
574 	struct sadb_msg *msg;
575 	struct sadb_ext *ext[SADB_EXT_MAX + 1];
576 	int extoff[SADB_EXT_MAX + 1];
577 	int extlen[SADB_EXT_MAX + 1];
578 };
579 
580 static struct supported_ealgs {
581 	int sadb_alg;
582 	const struct enc_xform *xform;
583 } supported_ealgs[] = {
584 	{ SADB_EALG_DESCBC,		&enc_xform_des },
585 	{ SADB_EALG_3DESCBC,		&enc_xform_3des },
586 	{ SADB_X_EALG_AES,		&enc_xform_rijndael128 },
587 	{ SADB_X_EALG_BLOWFISHCBC,	&enc_xform_blf },
588 	{ SADB_X_EALG_CAST128CBC,	&enc_xform_cast5 },
589 	{ SADB_EALG_NULL,		&enc_xform_null },
590 	{ SADB_X_EALG_CAMELLIACBC,	&enc_xform_camellia },
591 	{ SADB_X_EALG_AESCTR,		&enc_xform_aes_icm },
592 	{ SADB_X_EALG_AESGCM16,		&enc_xform_aes_nist_gcm },
593 	{ SADB_X_EALG_AESGMAC,		&enc_xform_aes_nist_gmac },
594 };
595 
596 static struct supported_aalgs {
597 	int sadb_alg;
598 	const struct auth_hash *xform;
599 } supported_aalgs[] = {
600 	{ SADB_X_AALG_NULL,		&auth_hash_null },
601 	{ SADB_AALG_MD5HMAC,		&auth_hash_hmac_md5 },
602 	{ SADB_AALG_SHA1HMAC,		&auth_hash_hmac_sha1 },
603 	{ SADB_X_AALG_RIPEMD160HMAC,	&auth_hash_hmac_ripemd_160 },
604 	{ SADB_X_AALG_MD5,		&auth_hash_key_md5 },
605 	{ SADB_X_AALG_SHA,		&auth_hash_key_sha1 },
606 	{ SADB_X_AALG_SHA2_256,		&auth_hash_hmac_sha2_256 },
607 	{ SADB_X_AALG_SHA2_384,		&auth_hash_hmac_sha2_384 },
608 	{ SADB_X_AALG_SHA2_512,		&auth_hash_hmac_sha2_512 },
609 	{ SADB_X_AALG_AES128GMAC,	&auth_hash_nist_gmac_aes_128 },
610 	{ SADB_X_AALG_AES192GMAC,	&auth_hash_nist_gmac_aes_192 },
611 	{ SADB_X_AALG_AES256GMAC,	&auth_hash_nist_gmac_aes_256 },
612 };
613 
614 static struct supported_calgs {
615 	int sadb_alg;
616 	const struct comp_algo *xform;
617 } supported_calgs[] = {
618 	{ SADB_X_CALG_DEFLATE,		&comp_algo_deflate },
619 };
620 
621 #ifndef IPSEC_DEBUG2
622 static struct callout key_timer;
623 #endif
624 
625 static void key_unlink(struct secpolicy *);
626 static struct secpolicy *key_do_allocsp(struct secpolicyindex *spidx, u_int dir);
627 static struct secpolicy *key_getsp(struct secpolicyindex *);
628 static struct secpolicy *key_getspbyid(u_int32_t);
629 static struct mbuf *key_gather_mbuf(struct mbuf *,
630 	const struct sadb_msghdr *, int, int, ...);
631 static int key_spdadd(struct socket *, struct mbuf *,
632 	const struct sadb_msghdr *);
633 static uint32_t key_getnewspid(void);
634 static int key_spddelete(struct socket *, struct mbuf *,
635 	const struct sadb_msghdr *);
636 static int key_spddelete2(struct socket *, struct mbuf *,
637 	const struct sadb_msghdr *);
638 static int key_spdget(struct socket *, struct mbuf *,
639 	const struct sadb_msghdr *);
640 static int key_spdflush(struct socket *, struct mbuf *,
641 	const struct sadb_msghdr *);
642 static int key_spddump(struct socket *, struct mbuf *,
643 	const struct sadb_msghdr *);
644 static struct mbuf *key_setdumpsp(struct secpolicy *,
645 	u_int8_t, u_int32_t, u_int32_t);
646 static struct mbuf *key_sp2mbuf(struct secpolicy *);
647 static size_t key_getspreqmsglen(struct secpolicy *);
648 static int key_spdexpire(struct secpolicy *);
649 static struct secashead *key_newsah(struct secasindex *);
650 static void key_freesah(struct secashead **);
651 static void key_delsah(struct secashead *);
652 static struct secasvar *key_newsav(const struct sadb_msghdr *,
653     struct secasindex *, uint32_t, int *);
654 static void key_delsav(struct secasvar *);
655 static void key_unlinksav(struct secasvar *);
656 static struct secashead *key_getsah(struct secasindex *);
657 static int key_checkspidup(uint32_t);
658 static struct secasvar *key_getsavbyspi(uint32_t);
659 static int key_setnatt(struct secasvar *, const struct sadb_msghdr *);
660 static int key_setsaval(struct secasvar *, const struct sadb_msghdr *);
661 static int key_updatelifetimes(struct secasvar *, const struct sadb_msghdr *);
662 static int key_updateaddresses(struct socket *, struct mbuf *,
663     const struct sadb_msghdr *, struct secasvar *, struct secasindex *);
664 
665 static struct mbuf *key_setdumpsa(struct secasvar *, u_int8_t,
666 	u_int8_t, u_int32_t, u_int32_t);
667 static struct mbuf *key_setsadbmsg(u_int8_t, u_int16_t, u_int8_t,
668 	u_int32_t, pid_t, u_int16_t);
669 static struct mbuf *key_setsadbsa(struct secasvar *);
670 static struct mbuf *key_setsadbaddr(u_int16_t,
671 	const struct sockaddr *, u_int8_t, u_int16_t);
672 static struct mbuf *key_setsadbxport(u_int16_t, u_int16_t);
673 static struct mbuf *key_setsadbxtype(u_int16_t);
674 static struct mbuf *key_setsadbxsa2(u_int8_t, u_int32_t, u_int32_t);
675 static struct mbuf *key_setsadbxsareplay(u_int32_t);
676 static struct mbuf *key_setsadbxpolicy(u_int16_t, u_int8_t,
677 	u_int32_t, u_int32_t);
678 static struct seckey *key_dup_keymsg(const struct sadb_key *, size_t,
679     struct malloc_type *);
680 static struct seclifetime *key_dup_lifemsg(const struct sadb_lifetime *src,
681     struct malloc_type *);
682 
683 /* flags for key_cmpsaidx() */
684 #define CMP_HEAD	1	/* protocol, addresses. */
685 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
686 #define CMP_REQID	3	/* additionally HEAD, reaid. */
687 #define CMP_EXACTLY	4	/* all elements. */
688 static int key_cmpsaidx(const struct secasindex *,
689     const struct secasindex *, int);
690 static int key_cmpspidx_exactly(struct secpolicyindex *,
691     struct secpolicyindex *);
692 static int key_cmpspidx_withmask(struct secpolicyindex *,
693     struct secpolicyindex *);
694 static int key_bbcmp(const void *, const void *, u_int);
695 static uint8_t key_satype2proto(uint8_t);
696 static uint8_t key_proto2satype(uint8_t);
697 
698 static int key_getspi(struct socket *, struct mbuf *,
699 	const struct sadb_msghdr *);
700 static uint32_t key_do_getnewspi(struct sadb_spirange *, struct secasindex *);
701 static int key_update(struct socket *, struct mbuf *,
702 	const struct sadb_msghdr *);
703 static int key_add(struct socket *, struct mbuf *,
704 	const struct sadb_msghdr *);
705 static int key_setident(struct secashead *, const struct sadb_msghdr *);
706 static struct mbuf *key_getmsgbuf_x1(struct mbuf *,
707 	const struct sadb_msghdr *);
708 static int key_delete(struct socket *, struct mbuf *,
709 	const struct sadb_msghdr *);
710 static int key_delete_all(struct socket *, struct mbuf *,
711 	const struct sadb_msghdr *, struct secasindex *);
712 static int key_get(struct socket *, struct mbuf *,
713 	const struct sadb_msghdr *);
714 
715 static void key_getcomb_setlifetime(struct sadb_comb *);
716 static struct mbuf *key_getcomb_ealg(void);
717 static struct mbuf *key_getcomb_ah(void);
718 static struct mbuf *key_getcomb_ipcomp(void);
719 static struct mbuf *key_getprop(const struct secasindex *);
720 
721 static int key_acquire(const struct secasindex *, struct secpolicy *);
722 static uint32_t key_newacq(const struct secasindex *, int *);
723 static uint32_t key_getacq(const struct secasindex *, int *);
724 static int key_acqdone(const struct secasindex *, uint32_t);
725 static int key_acqreset(uint32_t);
726 static struct secspacq *key_newspacq(struct secpolicyindex *);
727 static struct secspacq *key_getspacq(struct secpolicyindex *);
728 static int key_acquire2(struct socket *, struct mbuf *,
729 	const struct sadb_msghdr *);
730 static int key_register(struct socket *, struct mbuf *,
731 	const struct sadb_msghdr *);
732 static int key_expire(struct secasvar *, int);
733 static int key_flush(struct socket *, struct mbuf *,
734 	const struct sadb_msghdr *);
735 static int key_dump(struct socket *, struct mbuf *,
736 	const struct sadb_msghdr *);
737 static int key_promisc(struct socket *, struct mbuf *,
738 	const struct sadb_msghdr *);
739 static int key_senderror(struct socket *, struct mbuf *, int);
740 static int key_validate_ext(const struct sadb_ext *, int);
741 static int key_align(struct mbuf *, struct sadb_msghdr *);
742 static struct mbuf *key_setlifetime(struct seclifetime *, uint16_t);
743 static struct mbuf *key_setkey(struct seckey *, uint16_t);
744 
745 static void spdcache_init(void);
746 static void spdcache_clear(void);
747 static struct spdcache_entry *spdcache_entry_alloc(
748 	const struct secpolicyindex *spidx,
749 	struct secpolicy *policy);
750 static void spdcache_entry_free(struct spdcache_entry *entry);
751 #ifdef VIMAGE
752 static void spdcache_destroy(void);
753 #endif
754 
755 #define	DBG_IPSEC_INITREF(t, p)	do {				\
756 	refcount_init(&(p)->refcnt, 1);				\
757 	KEYDBG(KEY_STAMP,					\
758 	    printf("%s: Initialize refcnt %s(%p) = %u\n",	\
759 	    __func__, #t, (p), (p)->refcnt));			\
760 } while (0)
761 #define	DBG_IPSEC_ADDREF(t, p)	do {				\
762 	refcount_acquire(&(p)->refcnt);				\
763 	KEYDBG(KEY_STAMP,					\
764 	    printf("%s: Acquire refcnt %s(%p) -> %u\n",		\
765 	    __func__, #t, (p), (p)->refcnt));			\
766 } while (0)
767 #define	DBG_IPSEC_DELREF(t, p)	do {				\
768 	KEYDBG(KEY_STAMP,					\
769 	    printf("%s: Release refcnt %s(%p) -> %u\n",		\
770 	    __func__, #t, (p), (p)->refcnt - 1));		\
771 	refcount_release(&(p)->refcnt);				\
772 } while (0)
773 
774 #define	IPSEC_INITREF(t, p)	refcount_init(&(p)->refcnt, 1)
775 #define	IPSEC_ADDREF(t, p)	refcount_acquire(&(p)->refcnt)
776 #define	IPSEC_DELREF(t, p)	refcount_release(&(p)->refcnt)
777 
778 #define	SP_INITREF(p)	IPSEC_INITREF(SP, p)
779 #define	SP_ADDREF(p)	IPSEC_ADDREF(SP, p)
780 #define	SP_DELREF(p)	IPSEC_DELREF(SP, p)
781 
782 #define	SAH_INITREF(p)	IPSEC_INITREF(SAH, p)
783 #define	SAH_ADDREF(p)	IPSEC_ADDREF(SAH, p)
784 #define	SAH_DELREF(p)	IPSEC_DELREF(SAH, p)
785 
786 #define	SAV_INITREF(p)	IPSEC_INITREF(SAV, p)
787 #define	SAV_ADDREF(p)	IPSEC_ADDREF(SAV, p)
788 #define	SAV_DELREF(p)	IPSEC_DELREF(SAV, p)
789 
790 /*
791  * Update the refcnt while holding the SPTREE lock.
792  */
793 void
794 key_addref(struct secpolicy *sp)
795 {
796 
797 	SP_ADDREF(sp);
798 }
799 
800 /*
801  * Return 0 when there are known to be no SP's for the specified
802  * direction.  Otherwise return 1.  This is used by IPsec code
803  * to optimize performance.
804  */
805 int
806 key_havesp(u_int dir)
807 {
808 
809 	return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ?
810 		TAILQ_FIRST(&V_sptree[dir]) != NULL : 1);
811 }
812 
813 /* %%% IPsec policy management */
814 /*
815  * Return current SPDB generation.
816  */
817 uint32_t
818 key_getspgen(void)
819 {
820 
821 	return (V_sp_genid);
822 }
823 
824 void
825 key_bumpspgen(void)
826 {
827 
828 	V_sp_genid++;
829 }
830 
831 static int
832 key_checksockaddrs(struct sockaddr *src, struct sockaddr *dst)
833 {
834 
835 	/* family match */
836 	if (src->sa_family != dst->sa_family)
837 		return (EINVAL);
838 	/* sa_len match */
839 	if (src->sa_len != dst->sa_len)
840 		return (EINVAL);
841 	switch (src->sa_family) {
842 #ifdef INET
843 	case AF_INET:
844 		if (src->sa_len != sizeof(struct sockaddr_in))
845 			return (EINVAL);
846 		break;
847 #endif
848 #ifdef INET6
849 	case AF_INET6:
850 		if (src->sa_len != sizeof(struct sockaddr_in6))
851 			return (EINVAL);
852 		break;
853 #endif
854 	default:
855 		return (EAFNOSUPPORT);
856 	}
857 	return (0);
858 }
859 
860 struct secpolicy *
861 key_do_allocsp(struct secpolicyindex *spidx, u_int dir)
862 {
863 	SPTREE_RLOCK_TRACKER;
864 	struct secpolicy *sp;
865 
866 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
867 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
868 		("invalid direction %u", dir));
869 
870 	SPTREE_RLOCK();
871 	TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
872 		if (key_cmpspidx_withmask(&sp->spidx, spidx)) {
873 			SP_ADDREF(sp);
874 			break;
875 		}
876 	}
877 	SPTREE_RUNLOCK();
878 	return (sp);
879 }
880 
881 
882 /*
883  * allocating a SP for OUTBOUND or INBOUND packet.
884  * Must call key_freesp() later.
885  * OUT:	NULL:	not found
886  *	others:	found and return the pointer.
887  */
888 struct secpolicy *
889 key_allocsp(struct secpolicyindex *spidx, u_int dir)
890 {
891 	struct spdcache_entry *entry, *lastentry, *tmpentry;
892 	struct secpolicy *sp;
893 	uint32_t hashv;
894 	int nb_entries;
895 
896 	if (!SPDCACHE_ACTIVE()) {
897 		sp = key_do_allocsp(spidx, dir);
898 		goto out;
899 	}
900 
901 	hashv = SPDCACHE_HASHVAL(spidx);
902 	SPDCACHE_LOCK(hashv);
903 	nb_entries = 0;
904 	LIST_FOREACH_SAFE(entry, &V_spdcachehashtbl[hashv], chain, tmpentry) {
905 		/* Removed outdated entries */
906 		if (entry->sp != NULL &&
907 		    entry->sp->state == IPSEC_SPSTATE_DEAD) {
908 			LIST_REMOVE(entry, chain);
909 			spdcache_entry_free(entry);
910 			continue;
911 		}
912 
913 		nb_entries++;
914 		if (!key_cmpspidx_exactly(&entry->spidx, spidx)) {
915 			lastentry = entry;
916 			continue;
917 		}
918 
919 		sp = entry->sp;
920 		if (entry->sp != NULL)
921 			SP_ADDREF(sp);
922 
923 		/* IPSECSTAT_INC(ips_spdcache_hits); */
924 
925 		SPDCACHE_UNLOCK(hashv);
926 		goto out;
927 	}
928 
929 	/* IPSECSTAT_INC(ips_spdcache_misses); */
930 
931 	sp = key_do_allocsp(spidx, dir);
932 	entry = spdcache_entry_alloc(spidx, sp);
933 	if (entry != NULL) {
934 		if (nb_entries >= SPDCACHE_MAX_ENTRIES_PER_HASH) {
935 			LIST_REMOVE(lastentry, chain);
936 			spdcache_entry_free(lastentry);
937 		}
938 
939 		LIST_INSERT_HEAD(&V_spdcachehashtbl[hashv], entry, chain);
940 	}
941 
942 	SPDCACHE_UNLOCK(hashv);
943 
944 out:
945 	if (sp != NULL) {	/* found a SPD entry */
946 		sp->lastused = time_second;
947 		KEYDBG(IPSEC_STAMP,
948 		    printf("%s: return SP(%p)\n", __func__, sp));
949 		KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
950 	} else {
951 		KEYDBG(IPSEC_DATA,
952 		    printf("%s: lookup failed for ", __func__);
953 		    kdebug_secpolicyindex(spidx, NULL));
954 	}
955 	return (sp);
956 }
957 
958 /*
959  * Allocating an SA entry for an *INBOUND* or *OUTBOUND* TCP packet, signed
960  * or should be signed by MD5 signature.
961  * We don't use key_allocsa() for such lookups, because we don't know SPI.
962  * Unlike ESP and AH protocols, SPI isn't transmitted in the TCP header with
963  * signed packet. We use SADB only as storage for password.
964  * OUT:	positive:	corresponding SA for given saidx found.
965  *	NULL:		SA not found
966  */
967 struct secasvar *
968 key_allocsa_tcpmd5(struct secasindex *saidx)
969 {
970 	SAHTREE_RLOCK_TRACKER;
971 	struct secashead *sah;
972 	struct secasvar *sav;
973 
974 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP,
975 	    ("unexpected security protocol %u", saidx->proto));
976 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TCPMD5,
977 	    ("unexpected mode %u", saidx->mode));
978 
979 	SAHTREE_RLOCK();
980 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
981 		KEYDBG(IPSEC_DUMP,
982 		    printf("%s: checking SAH\n", __func__);
983 		    kdebug_secash(sah, "  "));
984 		if (sah->saidx.proto != IPPROTO_TCP)
985 			continue;
986 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
987 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
988 			break;
989 	}
990 	if (sah != NULL) {
991 		if (V_key_preferred_oldsa)
992 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
993 		else
994 			sav = TAILQ_FIRST(&sah->savtree_alive);
995 		if (sav != NULL)
996 			SAV_ADDREF(sav);
997 	} else
998 		sav = NULL;
999 	SAHTREE_RUNLOCK();
1000 
1001 	if (sav != NULL) {
1002 		KEYDBG(IPSEC_STAMP,
1003 		    printf("%s: return SA(%p)\n", __func__, sav));
1004 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1005 	} else {
1006 		KEYDBG(IPSEC_STAMP,
1007 		    printf("%s: SA not found\n", __func__));
1008 		KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1009 	}
1010 	return (sav);
1011 }
1012 
1013 /*
1014  * Allocating an SA entry for an *OUTBOUND* packet.
1015  * OUT:	positive:	corresponding SA for given saidx found.
1016  *	NULL:		SA not found, but will be acquired, check *error
1017  *			for acquiring status.
1018  */
1019 struct secasvar *
1020 key_allocsa_policy(struct secpolicy *sp, const struct secasindex *saidx,
1021     int *error)
1022 {
1023 	SAHTREE_RLOCK_TRACKER;
1024 	struct secashead *sah;
1025 	struct secasvar *sav;
1026 
1027 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
1028 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TRANSPORT ||
1029 		saidx->mode == IPSEC_MODE_TUNNEL,
1030 		("unexpected policy %u", saidx->mode));
1031 
1032 	/*
1033 	 * We check new SA in the IPsec request because a different
1034 	 * SA may be involved each time this request is checked, either
1035 	 * because new SAs are being configured, or this request is
1036 	 * associated with an unconnected datagram socket, or this request
1037 	 * is associated with a system default policy.
1038 	 */
1039 	SAHTREE_RLOCK();
1040 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
1041 		KEYDBG(IPSEC_DUMP,
1042 		    printf("%s: checking SAH\n", __func__);
1043 		    kdebug_secash(sah, "  "));
1044 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID))
1045 			break;
1046 
1047 	}
1048 	if (sah != NULL) {
1049 		/*
1050 		 * Allocate the oldest SA available according to
1051 		 * draft-jenkins-ipsec-rekeying-03.
1052 		 */
1053 		if (V_key_preferred_oldsa)
1054 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1055 		else
1056 			sav = TAILQ_FIRST(&sah->savtree_alive);
1057 		if (sav != NULL)
1058 			SAV_ADDREF(sav);
1059 	} else
1060 		sav = NULL;
1061 	SAHTREE_RUNLOCK();
1062 
1063 	if (sav != NULL) {
1064 		*error = 0;
1065 		KEYDBG(IPSEC_STAMP,
1066 		    printf("%s: chosen SA(%p) for SP(%p)\n", __func__,
1067 			sav, sp));
1068 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1069 		return (sav); /* return referenced SA */
1070 	}
1071 
1072 	/* there is no SA */
1073 	*error = key_acquire(saidx, sp);
1074 	if ((*error) != 0)
1075 		ipseclog((LOG_DEBUG,
1076 		    "%s: error %d returned from key_acquire()\n",
1077 			__func__, *error));
1078 	KEYDBG(IPSEC_STAMP,
1079 	    printf("%s: acquire SA for SP(%p), error %d\n",
1080 		__func__, sp, *error));
1081 	KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1082 	return (NULL);
1083 }
1084 
1085 /*
1086  * allocating a usable SA entry for a *INBOUND* packet.
1087  * Must call key_freesav() later.
1088  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
1089  *	NULL:		not found, or error occurred.
1090  *
1091  * According to RFC 2401 SA is uniquely identified by a triple SPI,
1092  * destination address, and security protocol. But according to RFC 4301,
1093  * SPI by itself suffices to specify an SA.
1094  *
1095  * Note that, however, we do need to keep source address in IPsec SA.
1096  * IKE specification and PF_KEY specification do assume that we
1097  * keep source address in IPsec SA.  We see a tricky situation here.
1098  */
1099 struct secasvar *
1100 key_allocsa(union sockaddr_union *dst, uint8_t proto, uint32_t spi)
1101 {
1102 	SAHTREE_RLOCK_TRACKER;
1103 	struct secasvar *sav;
1104 
1105 	IPSEC_ASSERT(proto == IPPROTO_ESP || proto == IPPROTO_AH ||
1106 	    proto == IPPROTO_IPCOMP, ("unexpected security protocol %u",
1107 	    proto));
1108 
1109 	SAHTREE_RLOCK();
1110 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
1111 		if (sav->spi == spi)
1112 			break;
1113 	}
1114 	/*
1115 	 * We use single SPI namespace for all protocols, so it is
1116 	 * impossible to have SPI duplicates in the SAVHASH.
1117 	 */
1118 	if (sav != NULL) {
1119 		if (sav->state != SADB_SASTATE_LARVAL &&
1120 		    sav->sah->saidx.proto == proto &&
1121 		    key_sockaddrcmp(&dst->sa,
1122 			&sav->sah->saidx.dst.sa, 0) == 0)
1123 			SAV_ADDREF(sav);
1124 		else
1125 			sav = NULL;
1126 	}
1127 	SAHTREE_RUNLOCK();
1128 
1129 	if (sav == NULL) {
1130 		KEYDBG(IPSEC_STAMP,
1131 		    char buf[IPSEC_ADDRSTRLEN];
1132 		    printf("%s: SA not found for spi %u proto %u dst %s\n",
1133 			__func__, ntohl(spi), proto, ipsec_address(dst, buf,
1134 			sizeof(buf))));
1135 	} else {
1136 		KEYDBG(IPSEC_STAMP,
1137 		    printf("%s: return SA(%p)\n", __func__, sav));
1138 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1139 	}
1140 	return (sav);
1141 }
1142 
1143 struct secasvar *
1144 key_allocsa_tunnel(union sockaddr_union *src, union sockaddr_union *dst,
1145     uint8_t proto)
1146 {
1147 	SAHTREE_RLOCK_TRACKER;
1148 	struct secasindex saidx;
1149 	struct secashead *sah;
1150 	struct secasvar *sav;
1151 
1152 	IPSEC_ASSERT(src != NULL, ("null src address"));
1153 	IPSEC_ASSERT(dst != NULL, ("null dst address"));
1154 
1155 	KEY_SETSECASIDX(proto, IPSEC_MODE_TUNNEL, 0, &src->sa,
1156 	    &dst->sa, &saidx);
1157 
1158 	sav = NULL;
1159 	SAHTREE_RLOCK();
1160 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
1161 		if (IPSEC_MODE_TUNNEL != sah->saidx.mode)
1162 			continue;
1163 		if (proto != sah->saidx.proto)
1164 			continue;
1165 		if (key_sockaddrcmp(&src->sa, &sah->saidx.src.sa, 0) != 0)
1166 			continue;
1167 		if (key_sockaddrcmp(&dst->sa, &sah->saidx.dst.sa, 0) != 0)
1168 			continue;
1169 		/* XXXAE: is key_preferred_oldsa reasonably?*/
1170 		if (V_key_preferred_oldsa)
1171 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1172 		else
1173 			sav = TAILQ_FIRST(&sah->savtree_alive);
1174 		if (sav != NULL) {
1175 			SAV_ADDREF(sav);
1176 			break;
1177 		}
1178 	}
1179 	SAHTREE_RUNLOCK();
1180 	KEYDBG(IPSEC_STAMP,
1181 	    printf("%s: return SA(%p)\n", __func__, sav));
1182 	if (sav != NULL)
1183 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1184 	return (sav);
1185 }
1186 
1187 /*
1188  * Must be called after calling key_allocsp().
1189  */
1190 void
1191 key_freesp(struct secpolicy **spp)
1192 {
1193 	struct secpolicy *sp = *spp;
1194 
1195 	IPSEC_ASSERT(sp != NULL, ("null sp"));
1196 	if (SP_DELREF(sp) == 0)
1197 		return;
1198 
1199 	KEYDBG(IPSEC_STAMP,
1200 	    printf("%s: last reference to SP(%p)\n", __func__, sp));
1201 	KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1202 
1203 	*spp = NULL;
1204 	while (sp->tcount > 0)
1205 		ipsec_delisr(sp->req[--sp->tcount]);
1206 	free(sp, M_IPSEC_SP);
1207 }
1208 
1209 static void
1210 key_unlink(struct secpolicy *sp)
1211 {
1212 
1213 	IPSEC_ASSERT(sp->spidx.dir == IPSEC_DIR_INBOUND ||
1214 	    sp->spidx.dir == IPSEC_DIR_OUTBOUND,
1215 	    ("invalid direction %u", sp->spidx.dir));
1216 	SPTREE_UNLOCK_ASSERT();
1217 
1218 	KEYDBG(KEY_STAMP,
1219 	    printf("%s: SP(%p)\n", __func__, sp));
1220 	SPTREE_WLOCK();
1221 	if (sp->state != IPSEC_SPSTATE_ALIVE) {
1222 		/* SP is already unlinked */
1223 		SPTREE_WUNLOCK();
1224 		return;
1225 	}
1226 	sp->state = IPSEC_SPSTATE_DEAD;
1227 	TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
1228 	V_spd_size--;
1229 	LIST_REMOVE(sp, idhash);
1230 	V_sp_genid++;
1231 	SPTREE_WUNLOCK();
1232 	if (SPDCACHE_ENABLED())
1233 		spdcache_clear();
1234 	key_freesp(&sp);
1235 }
1236 
1237 /*
1238  * insert a secpolicy into the SP database. Lower priorities first
1239  */
1240 static void
1241 key_insertsp(struct secpolicy *newsp)
1242 {
1243 	struct secpolicy *sp;
1244 
1245 	SPTREE_WLOCK_ASSERT();
1246 	TAILQ_FOREACH(sp, &V_sptree[newsp->spidx.dir], chain) {
1247 		if (newsp->priority < sp->priority) {
1248 			TAILQ_INSERT_BEFORE(sp, newsp, chain);
1249 			goto done;
1250 		}
1251 	}
1252 	TAILQ_INSERT_TAIL(&V_sptree[newsp->spidx.dir], newsp, chain);
1253 done:
1254 	LIST_INSERT_HEAD(SPHASH_HASH(newsp->id), newsp, idhash);
1255 	newsp->state = IPSEC_SPSTATE_ALIVE;
1256 	V_spd_size++;
1257 	V_sp_genid++;
1258 }
1259 
1260 /*
1261  * Insert a bunch of VTI secpolicies into the SPDB.
1262  * We keep VTI policies in the separate list due to following reasons:
1263  * 1) they should be immutable to user's or some deamon's attempts to
1264  *    delete. The only way delete such policies - destroy or unconfigure
1265  *    corresponding virtual inteface.
1266  * 2) such policies have traffic selector that matches all traffic per
1267  *    address family.
1268  * Since all VTI policies have the same priority, we don't care about
1269  * policies order.
1270  */
1271 int
1272 key_register_ifnet(struct secpolicy **spp, u_int count)
1273 {
1274 	struct mbuf *m;
1275 	u_int i;
1276 
1277 	SPTREE_WLOCK();
1278 	/*
1279 	 * First of try to acquire id for each SP.
1280 	 */
1281 	for (i = 0; i < count; i++) {
1282 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1283 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1284 		    ("invalid direction %u", spp[i]->spidx.dir));
1285 
1286 		if ((spp[i]->id = key_getnewspid()) == 0) {
1287 			SPTREE_WUNLOCK();
1288 			return (EAGAIN);
1289 		}
1290 	}
1291 	for (i = 0; i < count; i++) {
1292 		TAILQ_INSERT_TAIL(&V_sptree_ifnet[spp[i]->spidx.dir],
1293 		    spp[i], chain);
1294 		/*
1295 		 * NOTE: despite the fact that we keep VTI SP in the
1296 		 * separate list, SPHASH contains policies from both
1297 		 * sources. Thus SADB_X_SPDGET will correctly return
1298 		 * SP by id, because it uses SPHASH for lookups.
1299 		 */
1300 		LIST_INSERT_HEAD(SPHASH_HASH(spp[i]->id), spp[i], idhash);
1301 		spp[i]->state = IPSEC_SPSTATE_IFNET;
1302 	}
1303 	SPTREE_WUNLOCK();
1304 	/*
1305 	 * Notify user processes about new SP.
1306 	 */
1307 	for (i = 0; i < count; i++) {
1308 		m = key_setdumpsp(spp[i], SADB_X_SPDADD, 0, 0);
1309 		if (m != NULL)
1310 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1311 	}
1312 	return (0);
1313 }
1314 
1315 void
1316 key_unregister_ifnet(struct secpolicy **spp, u_int count)
1317 {
1318 	struct mbuf *m;
1319 	u_int i;
1320 
1321 	SPTREE_WLOCK();
1322 	for (i = 0; i < count; i++) {
1323 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1324 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1325 		    ("invalid direction %u", spp[i]->spidx.dir));
1326 
1327 		if (spp[i]->state != IPSEC_SPSTATE_IFNET)
1328 			continue;
1329 		spp[i]->state = IPSEC_SPSTATE_DEAD;
1330 		TAILQ_REMOVE(&V_sptree_ifnet[spp[i]->spidx.dir],
1331 		    spp[i], chain);
1332 		V_spd_size--;
1333 		LIST_REMOVE(spp[i], idhash);
1334 	}
1335 	SPTREE_WUNLOCK();
1336 	if (SPDCACHE_ENABLED())
1337 		spdcache_clear();
1338 
1339 	for (i = 0; i < count; i++) {
1340 		m = key_setdumpsp(spp[i], SADB_X_SPDDELETE, 0, 0);
1341 		if (m != NULL)
1342 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1343 	}
1344 }
1345 
1346 /*
1347  * Must be called after calling key_allocsa().
1348  * This function is called by key_freesp() to free some SA allocated
1349  * for a policy.
1350  */
1351 void
1352 key_freesav(struct secasvar **psav)
1353 {
1354 	struct secasvar *sav = *psav;
1355 
1356 	IPSEC_ASSERT(sav != NULL, ("null sav"));
1357 	if (SAV_DELREF(sav) == 0)
1358 		return;
1359 
1360 	KEYDBG(IPSEC_STAMP,
1361 	    printf("%s: last reference to SA(%p)\n", __func__, sav));
1362 
1363 	*psav = NULL;
1364 	key_delsav(sav);
1365 }
1366 
1367 /*
1368  * Unlink SA from SAH and SPI hash under SAHTREE_WLOCK.
1369  * Expect that SA has extra reference due to lookup.
1370  * Release this references, also release SAH reference after unlink.
1371  */
1372 static void
1373 key_unlinksav(struct secasvar *sav)
1374 {
1375 	struct secashead *sah;
1376 
1377 	KEYDBG(KEY_STAMP,
1378 	    printf("%s: SA(%p)\n", __func__, sav));
1379 
1380 	SAHTREE_UNLOCK_ASSERT();
1381 	SAHTREE_WLOCK();
1382 	if (sav->state == SADB_SASTATE_DEAD) {
1383 		/* SA is already unlinked */
1384 		SAHTREE_WUNLOCK();
1385 		return;
1386 	}
1387 	/* Unlink from SAH */
1388 	if (sav->state == SADB_SASTATE_LARVAL)
1389 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
1390 	else
1391 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
1392 	/* Unlink from SPI hash */
1393 	LIST_REMOVE(sav, spihash);
1394 	sav->state = SADB_SASTATE_DEAD;
1395 	sah = sav->sah;
1396 	SAHTREE_WUNLOCK();
1397 	key_freesav(&sav);
1398 	/* Since we are unlinked, release reference to SAH */
1399 	key_freesah(&sah);
1400 }
1401 
1402 /* %%% SPD management */
1403 /*
1404  * search SPD
1405  * OUT:	NULL	: not found
1406  *	others	: found, pointer to a SP.
1407  */
1408 static struct secpolicy *
1409 key_getsp(struct secpolicyindex *spidx)
1410 {
1411 	SPTREE_RLOCK_TRACKER;
1412 	struct secpolicy *sp;
1413 
1414 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
1415 
1416 	SPTREE_RLOCK();
1417 	TAILQ_FOREACH(sp, &V_sptree[spidx->dir], chain) {
1418 		if (key_cmpspidx_exactly(spidx, &sp->spidx)) {
1419 			SP_ADDREF(sp);
1420 			break;
1421 		}
1422 	}
1423 	SPTREE_RUNLOCK();
1424 
1425 	return sp;
1426 }
1427 
1428 /*
1429  * get SP by index.
1430  * OUT:	NULL	: not found
1431  *	others	: found, pointer to referenced SP.
1432  */
1433 static struct secpolicy *
1434 key_getspbyid(uint32_t id)
1435 {
1436 	SPTREE_RLOCK_TRACKER;
1437 	struct secpolicy *sp;
1438 
1439 	SPTREE_RLOCK();
1440 	LIST_FOREACH(sp, SPHASH_HASH(id), idhash) {
1441 		if (sp->id == id) {
1442 			SP_ADDREF(sp);
1443 			break;
1444 		}
1445 	}
1446 	SPTREE_RUNLOCK();
1447 	return (sp);
1448 }
1449 
1450 struct secpolicy *
1451 key_newsp(void)
1452 {
1453 	struct secpolicy *sp;
1454 
1455 	sp = malloc(sizeof(*sp), M_IPSEC_SP, M_NOWAIT | M_ZERO);
1456 	if (sp != NULL)
1457 		SP_INITREF(sp);
1458 	return (sp);
1459 }
1460 
1461 struct ipsecrequest *
1462 ipsec_newisr(void)
1463 {
1464 
1465 	return (malloc(sizeof(struct ipsecrequest), M_IPSEC_SR,
1466 	    M_NOWAIT | M_ZERO));
1467 }
1468 
1469 void
1470 ipsec_delisr(struct ipsecrequest *p)
1471 {
1472 
1473 	free(p, M_IPSEC_SR);
1474 }
1475 
1476 /*
1477  * create secpolicy structure from sadb_x_policy structure.
1478  * NOTE: `state', `secpolicyindex' and 'id' in secpolicy structure
1479  * are not set, so must be set properly later.
1480  */
1481 struct secpolicy *
1482 key_msg2sp(struct sadb_x_policy *xpl0, size_t len, int *error)
1483 {
1484 	struct secpolicy *newsp;
1485 
1486 	IPSEC_ASSERT(xpl0 != NULL, ("null xpl0"));
1487 	IPSEC_ASSERT(len >= sizeof(*xpl0), ("policy too short: %zu", len));
1488 
1489 	if (len != PFKEY_EXTLEN(xpl0)) {
1490 		ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n", __func__));
1491 		*error = EINVAL;
1492 		return NULL;
1493 	}
1494 
1495 	if ((newsp = key_newsp()) == NULL) {
1496 		*error = ENOBUFS;
1497 		return NULL;
1498 	}
1499 
1500 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1501 	newsp->policy = xpl0->sadb_x_policy_type;
1502 	newsp->priority = xpl0->sadb_x_policy_priority;
1503 	newsp->tcount = 0;
1504 
1505 	/* check policy */
1506 	switch (xpl0->sadb_x_policy_type) {
1507 	case IPSEC_POLICY_DISCARD:
1508 	case IPSEC_POLICY_NONE:
1509 	case IPSEC_POLICY_ENTRUST:
1510 	case IPSEC_POLICY_BYPASS:
1511 		break;
1512 
1513 	case IPSEC_POLICY_IPSEC:
1514 	    {
1515 		struct sadb_x_ipsecrequest *xisr;
1516 		struct ipsecrequest *isr;
1517 		int tlen;
1518 
1519 		/* validity check */
1520 		if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1521 			ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n",
1522 				__func__));
1523 			key_freesp(&newsp);
1524 			*error = EINVAL;
1525 			return NULL;
1526 		}
1527 
1528 		tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1529 		xisr = (struct sadb_x_ipsecrequest *)(xpl0 + 1);
1530 
1531 		while (tlen > 0) {
1532 			/* length check */
1533 			if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr) ||
1534 			    xisr->sadb_x_ipsecrequest_len > tlen) {
1535 				ipseclog((LOG_DEBUG, "%s: invalid ipsecrequest "
1536 					"length.\n", __func__));
1537 				key_freesp(&newsp);
1538 				*error = EINVAL;
1539 				return NULL;
1540 			}
1541 
1542 			if (newsp->tcount >= IPSEC_MAXREQ) {
1543 				ipseclog((LOG_DEBUG,
1544 				    "%s: too many ipsecrequests.\n",
1545 				    __func__));
1546 				key_freesp(&newsp);
1547 				*error = EINVAL;
1548 				return (NULL);
1549 			}
1550 
1551 			/* allocate request buffer */
1552 			/* NB: data structure is zero'd */
1553 			isr = ipsec_newisr();
1554 			if (isr == NULL) {
1555 				ipseclog((LOG_DEBUG,
1556 				    "%s: No more memory.\n", __func__));
1557 				key_freesp(&newsp);
1558 				*error = ENOBUFS;
1559 				return NULL;
1560 			}
1561 
1562 			newsp->req[newsp->tcount++] = isr;
1563 
1564 			/* set values */
1565 			switch (xisr->sadb_x_ipsecrequest_proto) {
1566 			case IPPROTO_ESP:
1567 			case IPPROTO_AH:
1568 			case IPPROTO_IPCOMP:
1569 				break;
1570 			default:
1571 				ipseclog((LOG_DEBUG,
1572 				    "%s: invalid proto type=%u\n", __func__,
1573 				    xisr->sadb_x_ipsecrequest_proto));
1574 				key_freesp(&newsp);
1575 				*error = EPROTONOSUPPORT;
1576 				return NULL;
1577 			}
1578 			isr->saidx.proto =
1579 			    (uint8_t)xisr->sadb_x_ipsecrequest_proto;
1580 
1581 			switch (xisr->sadb_x_ipsecrequest_mode) {
1582 			case IPSEC_MODE_TRANSPORT:
1583 			case IPSEC_MODE_TUNNEL:
1584 				break;
1585 			case IPSEC_MODE_ANY:
1586 			default:
1587 				ipseclog((LOG_DEBUG,
1588 				    "%s: invalid mode=%u\n", __func__,
1589 				    xisr->sadb_x_ipsecrequest_mode));
1590 				key_freesp(&newsp);
1591 				*error = EINVAL;
1592 				return NULL;
1593 			}
1594 			isr->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1595 
1596 			switch (xisr->sadb_x_ipsecrequest_level) {
1597 			case IPSEC_LEVEL_DEFAULT:
1598 			case IPSEC_LEVEL_USE:
1599 			case IPSEC_LEVEL_REQUIRE:
1600 				break;
1601 			case IPSEC_LEVEL_UNIQUE:
1602 				/* validity check */
1603 				/*
1604 				 * If range violation of reqid, kernel will
1605 				 * update it, don't refuse it.
1606 				 */
1607 				if (xisr->sadb_x_ipsecrequest_reqid
1608 						> IPSEC_MANUAL_REQID_MAX) {
1609 					ipseclog((LOG_DEBUG,
1610 					    "%s: reqid=%d range "
1611 					    "violation, updated by kernel.\n",
1612 					    __func__,
1613 					    xisr->sadb_x_ipsecrequest_reqid));
1614 					xisr->sadb_x_ipsecrequest_reqid = 0;
1615 				}
1616 
1617 				/* allocate new reqid id if reqid is zero. */
1618 				if (xisr->sadb_x_ipsecrequest_reqid == 0) {
1619 					u_int32_t reqid;
1620 					if ((reqid = key_newreqid()) == 0) {
1621 						key_freesp(&newsp);
1622 						*error = ENOBUFS;
1623 						return NULL;
1624 					}
1625 					isr->saidx.reqid = reqid;
1626 					xisr->sadb_x_ipsecrequest_reqid = reqid;
1627 				} else {
1628 				/* set it for manual keying. */
1629 					isr->saidx.reqid =
1630 					    xisr->sadb_x_ipsecrequest_reqid;
1631 				}
1632 				break;
1633 
1634 			default:
1635 				ipseclog((LOG_DEBUG, "%s: invalid level=%u\n",
1636 					__func__,
1637 					xisr->sadb_x_ipsecrequest_level));
1638 				key_freesp(&newsp);
1639 				*error = EINVAL;
1640 				return NULL;
1641 			}
1642 			isr->level = xisr->sadb_x_ipsecrequest_level;
1643 
1644 			/* set IP addresses if there */
1645 			if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1646 				struct sockaddr *paddr;
1647 
1648 				len = tlen - sizeof(*xisr);
1649 				paddr = (struct sockaddr *)(xisr + 1);
1650 				/* validity check */
1651 				if (len < sizeof(struct sockaddr) ||
1652 				    len < 2 * paddr->sa_len ||
1653 				    paddr->sa_len > sizeof(isr->saidx.src)) {
1654 					ipseclog((LOG_DEBUG, "%s: invalid "
1655 						"request address length.\n",
1656 						__func__));
1657 					key_freesp(&newsp);
1658 					*error = EINVAL;
1659 					return NULL;
1660 				}
1661 				/*
1662 				 * Request length should be enough to keep
1663 				 * source and destination addresses.
1664 				 */
1665 				if (xisr->sadb_x_ipsecrequest_len <
1666 				    sizeof(*xisr) + 2 * paddr->sa_len) {
1667 					ipseclog((LOG_DEBUG, "%s: invalid "
1668 					    "ipsecrequest length.\n",
1669 					    __func__));
1670 					key_freesp(&newsp);
1671 					*error = EINVAL;
1672 					return (NULL);
1673 				}
1674 				bcopy(paddr, &isr->saidx.src, paddr->sa_len);
1675 				paddr = (struct sockaddr *)((caddr_t)paddr +
1676 				    paddr->sa_len);
1677 
1678 				/* validity check */
1679 				if (paddr->sa_len !=
1680 				    isr->saidx.src.sa.sa_len) {
1681 					ipseclog((LOG_DEBUG, "%s: invalid "
1682 						"request address length.\n",
1683 						__func__));
1684 					key_freesp(&newsp);
1685 					*error = EINVAL;
1686 					return NULL;
1687 				}
1688 				/* AF family should match */
1689 				if (paddr->sa_family !=
1690 				    isr->saidx.src.sa.sa_family) {
1691 					ipseclog((LOG_DEBUG, "%s: address "
1692 					    "family doesn't match.\n",
1693 						__func__));
1694 					key_freesp(&newsp);
1695 					*error = EINVAL;
1696 					return (NULL);
1697 				}
1698 				bcopy(paddr, &isr->saidx.dst, paddr->sa_len);
1699 			} else {
1700 				/*
1701 				 * Addresses for TUNNEL mode requests are
1702 				 * mandatory.
1703 				 */
1704 				if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1705 					ipseclog((LOG_DEBUG, "%s: missing "
1706 					    "request addresses.\n", __func__));
1707 					key_freesp(&newsp);
1708 					*error = EINVAL;
1709 					return (NULL);
1710 				}
1711 			}
1712 			tlen -= xisr->sadb_x_ipsecrequest_len;
1713 
1714 			/* validity check */
1715 			if (tlen < 0) {
1716 				ipseclog((LOG_DEBUG, "%s: becoming tlen < 0.\n",
1717 					__func__));
1718 				key_freesp(&newsp);
1719 				*error = EINVAL;
1720 				return NULL;
1721 			}
1722 
1723 			xisr = (struct sadb_x_ipsecrequest *)((caddr_t)xisr
1724 			                 + xisr->sadb_x_ipsecrequest_len);
1725 		}
1726 		/* XXXAE: LARVAL SP */
1727 		if (newsp->tcount < 1) {
1728 			ipseclog((LOG_DEBUG, "%s: valid IPSEC transforms "
1729 			    "not found.\n", __func__));
1730 			key_freesp(&newsp);
1731 			*error = EINVAL;
1732 			return (NULL);
1733 		}
1734 	    }
1735 		break;
1736 	default:
1737 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
1738 		key_freesp(&newsp);
1739 		*error = EINVAL;
1740 		return NULL;
1741 	}
1742 
1743 	*error = 0;
1744 	return (newsp);
1745 }
1746 
1747 uint32_t
1748 key_newreqid(void)
1749 {
1750 	static uint32_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1751 
1752 	if (auto_reqid == ~0)
1753 		auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1754 	else
1755 		auto_reqid++;
1756 
1757 	/* XXX should be unique check */
1758 	return (auto_reqid);
1759 }
1760 
1761 /*
1762  * copy secpolicy struct to sadb_x_policy structure indicated.
1763  */
1764 static struct mbuf *
1765 key_sp2mbuf(struct secpolicy *sp)
1766 {
1767 	struct mbuf *m;
1768 	size_t tlen;
1769 
1770 	tlen = key_getspreqmsglen(sp);
1771 	m = m_get2(tlen, M_NOWAIT, MT_DATA, 0);
1772 	if (m == NULL)
1773 		return (NULL);
1774 	m_align(m, tlen);
1775 	m->m_len = tlen;
1776 	if (key_sp2msg(sp, m->m_data, &tlen) != 0) {
1777 		m_freem(m);
1778 		return (NULL);
1779 	}
1780 	return (m);
1781 }
1782 
1783 int
1784 key_sp2msg(struct secpolicy *sp, void *request, size_t *len)
1785 {
1786 	struct sadb_x_ipsecrequest *xisr;
1787 	struct sadb_x_policy *xpl;
1788 	struct ipsecrequest *isr;
1789 	size_t xlen, ilen;
1790 	caddr_t p;
1791 	int error, i;
1792 
1793 	IPSEC_ASSERT(sp != NULL, ("null policy"));
1794 
1795 	xlen = sizeof(*xpl);
1796 	if (*len < xlen)
1797 		return (EINVAL);
1798 
1799 	error = 0;
1800 	bzero(request, *len);
1801 	xpl = (struct sadb_x_policy *)request;
1802 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
1803 	xpl->sadb_x_policy_type = sp->policy;
1804 	xpl->sadb_x_policy_dir = sp->spidx.dir;
1805 	xpl->sadb_x_policy_id = sp->id;
1806 	xpl->sadb_x_policy_priority = sp->priority;
1807 	switch (sp->state) {
1808 	case IPSEC_SPSTATE_IFNET:
1809 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_IFNET;
1810 		break;
1811 	case IPSEC_SPSTATE_PCB:
1812 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_PCB;
1813 		break;
1814 	default:
1815 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_GLOBAL;
1816 	}
1817 
1818 	/* if is the policy for ipsec ? */
1819 	if (sp->policy == IPSEC_POLICY_IPSEC) {
1820 		p = (caddr_t)xpl + sizeof(*xpl);
1821 		for (i = 0; i < sp->tcount; i++) {
1822 			isr = sp->req[i];
1823 			ilen = PFKEY_ALIGN8(sizeof(*xisr) +
1824 			    isr->saidx.src.sa.sa_len +
1825 			    isr->saidx.dst.sa.sa_len);
1826 			xlen += ilen;
1827 			if (xlen > *len) {
1828 				error = ENOBUFS;
1829 				/* Calculate needed size */
1830 				continue;
1831 			}
1832 			xisr = (struct sadb_x_ipsecrequest *)p;
1833 			xisr->sadb_x_ipsecrequest_len = ilen;
1834 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
1835 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
1836 			xisr->sadb_x_ipsecrequest_level = isr->level;
1837 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
1838 
1839 			p += sizeof(*xisr);
1840 			bcopy(&isr->saidx.src, p, isr->saidx.src.sa.sa_len);
1841 			p += isr->saidx.src.sa.sa_len;
1842 			bcopy(&isr->saidx.dst, p, isr->saidx.dst.sa.sa_len);
1843 			p += isr->saidx.dst.sa.sa_len;
1844 		}
1845 	}
1846 	xpl->sadb_x_policy_len = PFKEY_UNIT64(xlen);
1847 	if (error == 0)
1848 		*len = xlen;
1849 	else
1850 		*len = sizeof(*xpl);
1851 	return (error);
1852 }
1853 
1854 /* m will not be freed nor modified */
1855 static struct mbuf *
1856 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
1857     int ndeep, int nitem, ...)
1858 {
1859 	va_list ap;
1860 	int idx;
1861 	int i;
1862 	struct mbuf *result = NULL, *n;
1863 	int len;
1864 
1865 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1866 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1867 
1868 	va_start(ap, nitem);
1869 	for (i = 0; i < nitem; i++) {
1870 		idx = va_arg(ap, int);
1871 		if (idx < 0 || idx > SADB_EXT_MAX)
1872 			goto fail;
1873 		/* don't attempt to pull empty extension */
1874 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
1875 			continue;
1876 		if (idx != SADB_EXT_RESERVED  &&
1877 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
1878 			continue;
1879 
1880 		if (idx == SADB_EXT_RESERVED) {
1881 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
1882 
1883 			IPSEC_ASSERT(len <= MHLEN, ("header too big %u", len));
1884 
1885 			MGETHDR(n, M_NOWAIT, MT_DATA);
1886 			if (!n)
1887 				goto fail;
1888 			n->m_len = len;
1889 			n->m_next = NULL;
1890 			m_copydata(m, 0, sizeof(struct sadb_msg),
1891 			    mtod(n, caddr_t));
1892 		} else if (i < ndeep) {
1893 			len = mhp->extlen[idx];
1894 			n = m_get2(len, M_NOWAIT, MT_DATA, 0);
1895 			if (n == NULL)
1896 				goto fail;
1897 			m_align(n, len);
1898 			n->m_len = len;
1899 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
1900 			    mtod(n, caddr_t));
1901 		} else {
1902 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
1903 			    M_NOWAIT);
1904 		}
1905 		if (n == NULL)
1906 			goto fail;
1907 
1908 		if (result)
1909 			m_cat(result, n);
1910 		else
1911 			result = n;
1912 	}
1913 	va_end(ap);
1914 
1915 	if ((result->m_flags & M_PKTHDR) != 0) {
1916 		result->m_pkthdr.len = 0;
1917 		for (n = result; n; n = n->m_next)
1918 			result->m_pkthdr.len += n->m_len;
1919 	}
1920 
1921 	return result;
1922 
1923 fail:
1924 	m_freem(result);
1925 	va_end(ap);
1926 	return NULL;
1927 }
1928 
1929 /*
1930  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
1931  * add an entry to SP database, when received
1932  *   <base, address(SD), (lifetime(H),) policy>
1933  * from the user(?).
1934  * Adding to SP database,
1935  * and send
1936  *   <base, address(SD), (lifetime(H),) policy>
1937  * to the socket which was send.
1938  *
1939  * SPDADD set a unique policy entry.
1940  * SPDSETIDX like SPDADD without a part of policy requests.
1941  * SPDUPDATE replace a unique policy entry.
1942  *
1943  * XXXAE: serialize this in PF_KEY to avoid races.
1944  * m will always be freed.
1945  */
1946 static int
1947 key_spdadd(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
1948 {
1949 	struct secpolicyindex spidx;
1950 	struct sadb_address *src0, *dst0;
1951 	struct sadb_x_policy *xpl0, *xpl;
1952 	struct sadb_lifetime *lft = NULL;
1953 	struct secpolicy *newsp;
1954 	int error;
1955 
1956 	IPSEC_ASSERT(so != NULL, ("null socket"));
1957 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1958 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1959 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
1960 
1961 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
1962 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
1963 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
1964 		ipseclog((LOG_DEBUG,
1965 		    "%s: invalid message: missing required header.\n",
1966 		    __func__));
1967 		return key_senderror(so, m, EINVAL);
1968 	}
1969 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
1970 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
1971 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
1972 		ipseclog((LOG_DEBUG,
1973 		    "%s: invalid message: wrong header size.\n", __func__));
1974 		return key_senderror(so, m, EINVAL);
1975 	}
1976 	if (!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD)) {
1977 		if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD)) {
1978 			ipseclog((LOG_DEBUG,
1979 			    "%s: invalid message: wrong header size.\n",
1980 			    __func__));
1981 			return key_senderror(so, m, EINVAL);
1982 		}
1983 		lft = (struct sadb_lifetime *)mhp->ext[SADB_EXT_LIFETIME_HARD];
1984 	}
1985 
1986 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
1987 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
1988 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
1989 
1990 	/* check the direciton */
1991 	switch (xpl0->sadb_x_policy_dir) {
1992 	case IPSEC_DIR_INBOUND:
1993 	case IPSEC_DIR_OUTBOUND:
1994 		break;
1995 	default:
1996 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
1997 		return key_senderror(so, m, EINVAL);
1998 	}
1999 	/* key_spdadd() accepts DISCARD, NONE and IPSEC. */
2000 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2001 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2002 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2003 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2004 		return key_senderror(so, m, EINVAL);
2005 	}
2006 
2007 	/* policy requests are mandatory when action is ipsec. */
2008 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
2009 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
2010 		ipseclog((LOG_DEBUG,
2011 		    "%s: policy requests required.\n", __func__));
2012 		return key_senderror(so, m, EINVAL);
2013 	}
2014 
2015 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
2016 	    (struct sockaddr *)(dst0 + 1));
2017 	if (error != 0 ||
2018 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2019 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2020 		return key_senderror(so, m, error);
2021 	}
2022 	/* make secindex */
2023 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2024 	                src0 + 1,
2025 	                dst0 + 1,
2026 	                src0->sadb_address_prefixlen,
2027 	                dst0->sadb_address_prefixlen,
2028 	                src0->sadb_address_proto,
2029 	                &spidx);
2030 	/* Checking there is SP already or not. */
2031 	newsp = key_getsp(&spidx);
2032 	if (newsp != NULL) {
2033 		if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2034 			KEYDBG(KEY_STAMP,
2035 			    printf("%s: unlink SP(%p) for SPDUPDATE\n",
2036 				__func__, newsp));
2037 			KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
2038 			key_unlink(newsp);
2039 			key_freesp(&newsp);
2040 		} else {
2041 			key_freesp(&newsp);
2042 			ipseclog((LOG_DEBUG, "%s: a SP entry exists already.",
2043 			    __func__));
2044 			return (key_senderror(so, m, EEXIST));
2045 		}
2046 	}
2047 
2048 	/* allocate new SP entry */
2049 	if ((newsp = key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error)) == NULL) {
2050 		return key_senderror(so, m, error);
2051 	}
2052 
2053 	newsp->lastused = newsp->created = time_second;
2054 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
2055 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
2056 	bcopy(&spidx, &newsp->spidx, sizeof(spidx));
2057 
2058 	/* XXXAE: there is race between key_getsp() and key_insertsp() */
2059 	SPTREE_WLOCK();
2060 	if ((newsp->id = key_getnewspid()) == 0) {
2061 		SPTREE_WUNLOCK();
2062 		key_freesp(&newsp);
2063 		return key_senderror(so, m, ENOBUFS);
2064 	}
2065 	key_insertsp(newsp);
2066 	SPTREE_WUNLOCK();
2067 	if (SPDCACHE_ENABLED())
2068 		spdcache_clear();
2069 
2070 	KEYDBG(KEY_STAMP,
2071 	    printf("%s: SP(%p)\n", __func__, newsp));
2072 	KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
2073 
2074     {
2075 	struct mbuf *n, *mpolicy;
2076 	struct sadb_msg *newmsg;
2077 	int off;
2078 
2079 	/* create new sadb_msg to reply. */
2080 	if (lft) {
2081 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
2082 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
2083 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2084 	} else {
2085 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
2086 		    SADB_X_EXT_POLICY,
2087 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2088 	}
2089 	if (!n)
2090 		return key_senderror(so, m, ENOBUFS);
2091 
2092 	if (n->m_len < sizeof(*newmsg)) {
2093 		n = m_pullup(n, sizeof(*newmsg));
2094 		if (!n)
2095 			return key_senderror(so, m, ENOBUFS);
2096 	}
2097 	newmsg = mtod(n, struct sadb_msg *);
2098 	newmsg->sadb_msg_errno = 0;
2099 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2100 
2101 	off = 0;
2102 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
2103 	    sizeof(*xpl), &off);
2104 	if (mpolicy == NULL) {
2105 		/* n is already freed */
2106 		return key_senderror(so, m, ENOBUFS);
2107 	}
2108 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, caddr_t) + off);
2109 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
2110 		m_freem(n);
2111 		return key_senderror(so, m, EINVAL);
2112 	}
2113 	xpl->sadb_x_policy_id = newsp->id;
2114 
2115 	m_freem(m);
2116 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2117     }
2118 }
2119 
2120 /*
2121  * get new policy id.
2122  * OUT:
2123  *	0:	failure.
2124  *	others: success.
2125  */
2126 static uint32_t
2127 key_getnewspid(void)
2128 {
2129 	struct secpolicy *sp;
2130 	uint32_t newid = 0;
2131 	int count = V_key_spi_trycnt;	/* XXX */
2132 
2133 	SPTREE_WLOCK_ASSERT();
2134 	while (count--) {
2135 		if (V_policy_id == ~0) /* overflowed */
2136 			newid = V_policy_id = 1;
2137 		else
2138 			newid = ++V_policy_id;
2139 		LIST_FOREACH(sp, SPHASH_HASH(newid), idhash) {
2140 			if (sp->id == newid)
2141 				break;
2142 		}
2143 		if (sp == NULL)
2144 			break;
2145 	}
2146 	if (count == 0 || newid == 0) {
2147 		ipseclog((LOG_DEBUG, "%s: failed to allocate policy id.\n",
2148 		    __func__));
2149 		return (0);
2150 	}
2151 	return (newid);
2152 }
2153 
2154 /*
2155  * SADB_SPDDELETE processing
2156  * receive
2157  *   <base, address(SD), policy(*)>
2158  * from the user(?), and set SADB_SASTATE_DEAD,
2159  * and send,
2160  *   <base, address(SD), policy(*)>
2161  * to the ikmpd.
2162  * policy(*) including direction of policy.
2163  *
2164  * m will always be freed.
2165  */
2166 static int
2167 key_spddelete(struct socket *so, struct mbuf *m,
2168     const struct sadb_msghdr *mhp)
2169 {
2170 	struct secpolicyindex spidx;
2171 	struct sadb_address *src0, *dst0;
2172 	struct sadb_x_policy *xpl0;
2173 	struct secpolicy *sp;
2174 
2175 	IPSEC_ASSERT(so != NULL, ("null so"));
2176 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2177 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2178 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2179 
2180 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
2181 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
2182 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
2183 		ipseclog((LOG_DEBUG,
2184 		    "%s: invalid message: missing required header.\n",
2185 		    __func__));
2186 		return key_senderror(so, m, EINVAL);
2187 	}
2188 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
2189 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
2190 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2191 		ipseclog((LOG_DEBUG,
2192 		    "%s: invalid message: wrong header size.\n", __func__));
2193 		return key_senderror(so, m, EINVAL);
2194 	}
2195 
2196 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
2197 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
2198 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
2199 
2200 	/* check the direciton */
2201 	switch (xpl0->sadb_x_policy_dir) {
2202 	case IPSEC_DIR_INBOUND:
2203 	case IPSEC_DIR_OUTBOUND:
2204 		break;
2205 	default:
2206 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2207 		return key_senderror(so, m, EINVAL);
2208 	}
2209 	/* Only DISCARD, NONE and IPSEC are allowed */
2210 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2211 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2212 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2213 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2214 		return key_senderror(so, m, EINVAL);
2215 	}
2216 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
2217 	    (struct sockaddr *)(dst0 + 1)) != 0 ||
2218 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2219 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2220 		return key_senderror(so, m, EINVAL);
2221 	}
2222 	/* make secindex */
2223 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2224 	                src0 + 1,
2225 	                dst0 + 1,
2226 	                src0->sadb_address_prefixlen,
2227 	                dst0->sadb_address_prefixlen,
2228 	                src0->sadb_address_proto,
2229 	                &spidx);
2230 
2231 	/* Is there SP in SPD ? */
2232 	if ((sp = key_getsp(&spidx)) == NULL) {
2233 		ipseclog((LOG_DEBUG, "%s: no SP found.\n", __func__));
2234 		return key_senderror(so, m, EINVAL);
2235 	}
2236 
2237 	/* save policy id to buffer to be returned. */
2238 	xpl0->sadb_x_policy_id = sp->id;
2239 
2240 	KEYDBG(KEY_STAMP,
2241 	    printf("%s: SP(%p)\n", __func__, sp));
2242 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2243 	key_unlink(sp);
2244 	key_freesp(&sp);
2245 
2246     {
2247 	struct mbuf *n;
2248 	struct sadb_msg *newmsg;
2249 
2250 	/* create new sadb_msg to reply. */
2251 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2252 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2253 	if (!n)
2254 		return key_senderror(so, m, ENOBUFS);
2255 
2256 	newmsg = mtod(n, struct sadb_msg *);
2257 	newmsg->sadb_msg_errno = 0;
2258 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2259 
2260 	m_freem(m);
2261 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2262     }
2263 }
2264 
2265 /*
2266  * SADB_SPDDELETE2 processing
2267  * receive
2268  *   <base, policy(*)>
2269  * from the user(?), and set SADB_SASTATE_DEAD,
2270  * and send,
2271  *   <base, policy(*)>
2272  * to the ikmpd.
2273  * policy(*) including direction of policy.
2274  *
2275  * m will always be freed.
2276  */
2277 static int
2278 key_spddelete2(struct socket *so, struct mbuf *m,
2279     const struct sadb_msghdr *mhp)
2280 {
2281 	struct secpolicy *sp;
2282 	uint32_t id;
2283 
2284 	IPSEC_ASSERT(so != NULL, ("null socket"));
2285 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2286 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2287 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2288 
2289 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2290 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2291 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2292 		    __func__));
2293 		return key_senderror(so, m, EINVAL);
2294 	}
2295 
2296 	id = ((struct sadb_x_policy *)
2297 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2298 
2299 	/* Is there SP in SPD ? */
2300 	if ((sp = key_getspbyid(id)) == NULL) {
2301 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2302 		    __func__, id));
2303 		return key_senderror(so, m, EINVAL);
2304 	}
2305 
2306 	KEYDBG(KEY_STAMP,
2307 	    printf("%s: SP(%p)\n", __func__, sp));
2308 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2309 	key_unlink(sp);
2310 	if (sp->state != IPSEC_SPSTATE_DEAD) {
2311 		ipseclog((LOG_DEBUG, "%s: failed to delete SP with id %u.\n",
2312 		    __func__, id));
2313 		key_freesp(&sp);
2314 		return (key_senderror(so, m, EACCES));
2315 	}
2316 	key_freesp(&sp);
2317 
2318     {
2319 	struct mbuf *n, *nn;
2320 	struct sadb_msg *newmsg;
2321 	int off, len;
2322 
2323 	/* create new sadb_msg to reply. */
2324 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2325 
2326 	MGETHDR(n, M_NOWAIT, MT_DATA);
2327 	if (n && len > MHLEN) {
2328 		if (!(MCLGET(n, M_NOWAIT))) {
2329 			m_freem(n);
2330 			n = NULL;
2331 		}
2332 	}
2333 	if (!n)
2334 		return key_senderror(so, m, ENOBUFS);
2335 
2336 	n->m_len = len;
2337 	n->m_next = NULL;
2338 	off = 0;
2339 
2340 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
2341 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2342 
2343 	IPSEC_ASSERT(off == len, ("length inconsistency (off %u len %u)",
2344 		off, len));
2345 
2346 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2347 	    mhp->extlen[SADB_X_EXT_POLICY], M_NOWAIT);
2348 	if (!n->m_next) {
2349 		m_freem(n);
2350 		return key_senderror(so, m, ENOBUFS);
2351 	}
2352 
2353 	n->m_pkthdr.len = 0;
2354 	for (nn = n; nn; nn = nn->m_next)
2355 		n->m_pkthdr.len += nn->m_len;
2356 
2357 	newmsg = mtod(n, struct sadb_msg *);
2358 	newmsg->sadb_msg_errno = 0;
2359 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2360 
2361 	m_freem(m);
2362 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2363     }
2364 }
2365 
2366 /*
2367  * SADB_X_SPDGET processing
2368  * receive
2369  *   <base, policy(*)>
2370  * from the user(?),
2371  * and send,
2372  *   <base, address(SD), policy>
2373  * to the ikmpd.
2374  * policy(*) including direction of policy.
2375  *
2376  * m will always be freed.
2377  */
2378 static int
2379 key_spdget(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2380 {
2381 	struct secpolicy *sp;
2382 	struct mbuf *n;
2383 	uint32_t id;
2384 
2385 	IPSEC_ASSERT(so != NULL, ("null socket"));
2386 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2387 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2388 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2389 
2390 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2391 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2392 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2393 		    __func__));
2394 		return key_senderror(so, m, EINVAL);
2395 	}
2396 
2397 	id = ((struct sadb_x_policy *)
2398 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2399 
2400 	/* Is there SP in SPD ? */
2401 	if ((sp = key_getspbyid(id)) == NULL) {
2402 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2403 		    __func__, id));
2404 		return key_senderror(so, m, ENOENT);
2405 	}
2406 
2407 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2408 	    mhp->msg->sadb_msg_pid);
2409 	key_freesp(&sp);
2410 	if (n != NULL) {
2411 		m_freem(m);
2412 		return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2413 	} else
2414 		return key_senderror(so, m, ENOBUFS);
2415 }
2416 
2417 /*
2418  * SADB_X_SPDACQUIRE processing.
2419  * Acquire policy and SA(s) for a *OUTBOUND* packet.
2420  * send
2421  *   <base, policy(*)>
2422  * to KMD, and expect to receive
2423  *   <base> with SADB_X_SPDACQUIRE if error occurred,
2424  * or
2425  *   <base, policy>
2426  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2427  * policy(*) is without policy requests.
2428  *
2429  *    0     : succeed
2430  *    others: error number
2431  */
2432 int
2433 key_spdacquire(struct secpolicy *sp)
2434 {
2435 	struct mbuf *result = NULL, *m;
2436 	struct secspacq *newspacq;
2437 
2438 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2439 	IPSEC_ASSERT(sp->req == NULL, ("policy exists"));
2440 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
2441 		("policy not IPSEC %u", sp->policy));
2442 
2443 	/* Get an entry to check whether sent message or not. */
2444 	newspacq = key_getspacq(&sp->spidx);
2445 	if (newspacq != NULL) {
2446 		if (V_key_blockacq_count < newspacq->count) {
2447 			/* reset counter and do send message. */
2448 			newspacq->count = 0;
2449 		} else {
2450 			/* increment counter and do nothing. */
2451 			newspacq->count++;
2452 			SPACQ_UNLOCK();
2453 			return (0);
2454 		}
2455 		SPACQ_UNLOCK();
2456 	} else {
2457 		/* make new entry for blocking to send SADB_ACQUIRE. */
2458 		newspacq = key_newspacq(&sp->spidx);
2459 		if (newspacq == NULL)
2460 			return ENOBUFS;
2461 	}
2462 
2463 	/* create new sadb_msg to reply. */
2464 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2465 	if (!m)
2466 		return ENOBUFS;
2467 
2468 	result = m;
2469 
2470 	result->m_pkthdr.len = 0;
2471 	for (m = result; m; m = m->m_next)
2472 		result->m_pkthdr.len += m->m_len;
2473 
2474 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2475 	    PFKEY_UNIT64(result->m_pkthdr.len);
2476 
2477 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2478 }
2479 
2480 /*
2481  * SADB_SPDFLUSH processing
2482  * receive
2483  *   <base>
2484  * from the user, and free all entries in secpctree.
2485  * and send,
2486  *   <base>
2487  * to the user.
2488  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2489  *
2490  * m will always be freed.
2491  */
2492 static int
2493 key_spdflush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2494 {
2495 	struct secpolicy_queue drainq;
2496 	struct sadb_msg *newmsg;
2497 	struct secpolicy *sp, *nextsp;
2498 	u_int dir;
2499 
2500 	IPSEC_ASSERT(so != NULL, ("null socket"));
2501 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2502 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2503 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2504 
2505 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2506 		return key_senderror(so, m, EINVAL);
2507 
2508 	TAILQ_INIT(&drainq);
2509 	SPTREE_WLOCK();
2510 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2511 		TAILQ_CONCAT(&drainq, &V_sptree[dir], chain);
2512 	}
2513 	/*
2514 	 * We need to set state to DEAD for each policy to be sure,
2515 	 * that another thread won't try to unlink it.
2516 	 * Also remove SP from sphash.
2517 	 */
2518 	TAILQ_FOREACH(sp, &drainq, chain) {
2519 		sp->state = IPSEC_SPSTATE_DEAD;
2520 		LIST_REMOVE(sp, idhash);
2521 	}
2522 	V_sp_genid++;
2523 	V_spd_size = 0;
2524 	SPTREE_WUNLOCK();
2525 	if (SPDCACHE_ENABLED())
2526 		spdcache_clear();
2527 	sp = TAILQ_FIRST(&drainq);
2528 	while (sp != NULL) {
2529 		nextsp = TAILQ_NEXT(sp, chain);
2530 		key_freesp(&sp);
2531 		sp = nextsp;
2532 	}
2533 
2534 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2535 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
2536 		return key_senderror(so, m, ENOBUFS);
2537 	}
2538 
2539 	if (m->m_next)
2540 		m_freem(m->m_next);
2541 	m->m_next = NULL;
2542 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2543 	newmsg = mtod(m, struct sadb_msg *);
2544 	newmsg->sadb_msg_errno = 0;
2545 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2546 
2547 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2548 }
2549 
2550 static uint8_t
2551 key_satype2scopemask(uint8_t satype)
2552 {
2553 
2554 	if (satype == IPSEC_POLICYSCOPE_ANY)
2555 		return (0xff);
2556 	return (satype);
2557 }
2558 /*
2559  * SADB_SPDDUMP processing
2560  * receive
2561  *   <base>
2562  * from the user, and dump all SP leaves and send,
2563  *   <base> .....
2564  * to the ikmpd.
2565  *
2566  * NOTE:
2567  *   sadb_msg_satype is considered as mask of policy scopes.
2568  *   m will always be freed.
2569  */
2570 static int
2571 key_spddump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2572 {
2573 	SPTREE_RLOCK_TRACKER;
2574 	struct secpolicy *sp;
2575 	struct mbuf *n;
2576 	int cnt;
2577 	u_int dir, scope;
2578 
2579 	IPSEC_ASSERT(so != NULL, ("null socket"));
2580 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2581 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2582 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2583 
2584 	/* search SPD entry and get buffer size. */
2585 	cnt = 0;
2586 	scope = key_satype2scopemask(mhp->msg->sadb_msg_satype);
2587 	SPTREE_RLOCK();
2588 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2589 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2590 			TAILQ_FOREACH(sp, &V_sptree[dir], chain)
2591 				cnt++;
2592 		}
2593 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2594 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain)
2595 				cnt++;
2596 		}
2597 	}
2598 
2599 	if (cnt == 0) {
2600 		SPTREE_RUNLOCK();
2601 		return key_senderror(so, m, ENOENT);
2602 	}
2603 
2604 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2605 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2606 			TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
2607 				--cnt;
2608 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2609 				    mhp->msg->sadb_msg_pid);
2610 
2611 				if (n != NULL)
2612 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2613 			}
2614 		}
2615 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2616 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain) {
2617 				--cnt;
2618 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2619 				    mhp->msg->sadb_msg_pid);
2620 
2621 				if (n != NULL)
2622 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2623 			}
2624 		}
2625 	}
2626 
2627 	SPTREE_RUNLOCK();
2628 	m_freem(m);
2629 	return (0);
2630 }
2631 
2632 static struct mbuf *
2633 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq,
2634     u_int32_t pid)
2635 {
2636 	struct mbuf *result = NULL, *m;
2637 	struct seclifetime lt;
2638 
2639 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid, sp->refcnt);
2640 	if (!m)
2641 		goto fail;
2642 	result = m;
2643 
2644 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2645 	    &sp->spidx.src.sa, sp->spidx.prefs,
2646 	    sp->spidx.ul_proto);
2647 	if (!m)
2648 		goto fail;
2649 	m_cat(result, m);
2650 
2651 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2652 	    &sp->spidx.dst.sa, sp->spidx.prefd,
2653 	    sp->spidx.ul_proto);
2654 	if (!m)
2655 		goto fail;
2656 	m_cat(result, m);
2657 
2658 	m = key_sp2mbuf(sp);
2659 	if (!m)
2660 		goto fail;
2661 	m_cat(result, m);
2662 
2663 	if(sp->lifetime){
2664 		lt.addtime=sp->created;
2665 		lt.usetime= sp->lastused;
2666 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_CURRENT);
2667 		if (!m)
2668 			goto fail;
2669 		m_cat(result, m);
2670 
2671 		lt.addtime=sp->lifetime;
2672 		lt.usetime= sp->validtime;
2673 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_HARD);
2674 		if (!m)
2675 			goto fail;
2676 		m_cat(result, m);
2677 	}
2678 
2679 	if ((result->m_flags & M_PKTHDR) == 0)
2680 		goto fail;
2681 
2682 	if (result->m_len < sizeof(struct sadb_msg)) {
2683 		result = m_pullup(result, sizeof(struct sadb_msg));
2684 		if (result == NULL)
2685 			goto fail;
2686 	}
2687 
2688 	result->m_pkthdr.len = 0;
2689 	for (m = result; m; m = m->m_next)
2690 		result->m_pkthdr.len += m->m_len;
2691 
2692 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2693 	    PFKEY_UNIT64(result->m_pkthdr.len);
2694 
2695 	return result;
2696 
2697 fail:
2698 	m_freem(result);
2699 	return NULL;
2700 }
2701 /*
2702  * get PFKEY message length for security policy and request.
2703  */
2704 static size_t
2705 key_getspreqmsglen(struct secpolicy *sp)
2706 {
2707 	size_t tlen, len;
2708 	int i;
2709 
2710 	tlen = sizeof(struct sadb_x_policy);
2711 	/* if is the policy for ipsec ? */
2712 	if (sp->policy != IPSEC_POLICY_IPSEC)
2713 		return (tlen);
2714 
2715 	/* get length of ipsec requests */
2716 	for (i = 0; i < sp->tcount; i++) {
2717 		len = sizeof(struct sadb_x_ipsecrequest)
2718 			+ sp->req[i]->saidx.src.sa.sa_len
2719 			+ sp->req[i]->saidx.dst.sa.sa_len;
2720 
2721 		tlen += PFKEY_ALIGN8(len);
2722 	}
2723 	return (tlen);
2724 }
2725 
2726 /*
2727  * SADB_SPDEXPIRE processing
2728  * send
2729  *   <base, address(SD), lifetime(CH), policy>
2730  * to KMD by PF_KEY.
2731  *
2732  * OUT:	0	: succeed
2733  *	others	: error number
2734  */
2735 static int
2736 key_spdexpire(struct secpolicy *sp)
2737 {
2738 	struct sadb_lifetime *lt;
2739 	struct mbuf *result = NULL, *m;
2740 	int len, error = -1;
2741 
2742 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2743 
2744 	KEYDBG(KEY_STAMP,
2745 	    printf("%s: SP(%p)\n", __func__, sp));
2746 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2747 
2748 	/* set msg header */
2749 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0);
2750 	if (!m) {
2751 		error = ENOBUFS;
2752 		goto fail;
2753 	}
2754 	result = m;
2755 
2756 	/* create lifetime extension (current and hard) */
2757 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
2758 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
2759 	if (m == NULL) {
2760 		error = ENOBUFS;
2761 		goto fail;
2762 	}
2763 	m_align(m, len);
2764 	m->m_len = len;
2765 	bzero(mtod(m, caddr_t), len);
2766 	lt = mtod(m, struct sadb_lifetime *);
2767 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2768 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
2769 	lt->sadb_lifetime_allocations = 0;
2770 	lt->sadb_lifetime_bytes = 0;
2771 	lt->sadb_lifetime_addtime = sp->created;
2772 	lt->sadb_lifetime_usetime = sp->lastused;
2773 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
2774 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2775 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
2776 	lt->sadb_lifetime_allocations = 0;
2777 	lt->sadb_lifetime_bytes = 0;
2778 	lt->sadb_lifetime_addtime = sp->lifetime;
2779 	lt->sadb_lifetime_usetime = sp->validtime;
2780 	m_cat(result, m);
2781 
2782 	/* set sadb_address for source */
2783 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2784 	    &sp->spidx.src.sa,
2785 	    sp->spidx.prefs, sp->spidx.ul_proto);
2786 	if (!m) {
2787 		error = ENOBUFS;
2788 		goto fail;
2789 	}
2790 	m_cat(result, m);
2791 
2792 	/* set sadb_address for destination */
2793 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2794 	    &sp->spidx.dst.sa,
2795 	    sp->spidx.prefd, sp->spidx.ul_proto);
2796 	if (!m) {
2797 		error = ENOBUFS;
2798 		goto fail;
2799 	}
2800 	m_cat(result, m);
2801 
2802 	/* set secpolicy */
2803 	m = key_sp2mbuf(sp);
2804 	if (!m) {
2805 		error = ENOBUFS;
2806 		goto fail;
2807 	}
2808 	m_cat(result, m);
2809 
2810 	if ((result->m_flags & M_PKTHDR) == 0) {
2811 		error = EINVAL;
2812 		goto fail;
2813 	}
2814 
2815 	if (result->m_len < sizeof(struct sadb_msg)) {
2816 		result = m_pullup(result, sizeof(struct sadb_msg));
2817 		if (result == NULL) {
2818 			error = ENOBUFS;
2819 			goto fail;
2820 		}
2821 	}
2822 
2823 	result->m_pkthdr.len = 0;
2824 	for (m = result; m; m = m->m_next)
2825 		result->m_pkthdr.len += m->m_len;
2826 
2827 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2828 	    PFKEY_UNIT64(result->m_pkthdr.len);
2829 
2830 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
2831 
2832  fail:
2833 	if (result)
2834 		m_freem(result);
2835 	return error;
2836 }
2837 
2838 /* %%% SAD management */
2839 /*
2840  * allocating and initialize new SA head.
2841  * OUT:	NULL	: failure due to the lack of memory.
2842  *	others	: pointer to new SA head.
2843  */
2844 static struct secashead *
2845 key_newsah(struct secasindex *saidx)
2846 {
2847 	struct secashead *sah;
2848 
2849 	sah = malloc(sizeof(struct secashead), M_IPSEC_SAH,
2850 	    M_NOWAIT | M_ZERO);
2851 	if (sah == NULL) {
2852 		PFKEYSTAT_INC(in_nomem);
2853 		return (NULL);
2854 	}
2855 	TAILQ_INIT(&sah->savtree_larval);
2856 	TAILQ_INIT(&sah->savtree_alive);
2857 	sah->saidx = *saidx;
2858 	sah->state = SADB_SASTATE_DEAD;
2859 	SAH_INITREF(sah);
2860 
2861 	KEYDBG(KEY_STAMP,
2862 	    printf("%s: SAH(%p)\n", __func__, sah));
2863 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2864 	return (sah);
2865 }
2866 
2867 static void
2868 key_freesah(struct secashead **psah)
2869 {
2870 	struct secashead *sah = *psah;
2871 
2872 	if (SAH_DELREF(sah) == 0)
2873 		return;
2874 
2875 	KEYDBG(KEY_STAMP,
2876 	    printf("%s: last reference to SAH(%p)\n", __func__, sah));
2877 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2878 
2879 	*psah = NULL;
2880 	key_delsah(sah);
2881 }
2882 
2883 static void
2884 key_delsah(struct secashead *sah)
2885 {
2886 	IPSEC_ASSERT(sah != NULL, ("NULL sah"));
2887 	IPSEC_ASSERT(sah->state == SADB_SASTATE_DEAD,
2888 	    ("Attempt to free non DEAD SAH %p", sah));
2889 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_larval),
2890 	    ("Attempt to free SAH %p with LARVAL SA", sah));
2891 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_alive),
2892 	    ("Attempt to free SAH %p with ALIVE SA", sah));
2893 
2894 	free(sah, M_IPSEC_SAH);
2895 }
2896 
2897 /*
2898  * allocating a new SA for key_add() and key_getspi() call,
2899  * and copy the values of mhp into new buffer.
2900  * When SAD message type is SADB_GETSPI set SA state to LARVAL.
2901  * For SADB_ADD create and initialize SA with MATURE state.
2902  * OUT:	NULL	: fail
2903  *	others	: pointer to new secasvar.
2904  */
2905 static struct secasvar *
2906 key_newsav(const struct sadb_msghdr *mhp, struct secasindex *saidx,
2907     uint32_t spi, int *errp)
2908 {
2909 	struct secashead *sah;
2910 	struct secasvar *sav;
2911 	int isnew;
2912 
2913 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2914 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2915 	IPSEC_ASSERT(mhp->msg->sadb_msg_type == SADB_GETSPI ||
2916 	    mhp->msg->sadb_msg_type == SADB_ADD, ("wrong message type"));
2917 
2918 	sav = NULL;
2919 	sah = NULL;
2920 	/* check SPI value */
2921 	switch (saidx->proto) {
2922 	case IPPROTO_ESP:
2923 	case IPPROTO_AH:
2924 		/*
2925 		 * RFC 4302, 2.4. Security Parameters Index (SPI), SPI values
2926 		 * 1-255 reserved by IANA for future use,
2927 		 * 0 for implementation specific, local use.
2928 		 */
2929 		if (ntohl(spi) <= 255) {
2930 			ipseclog((LOG_DEBUG, "%s: illegal range of SPI %u.\n",
2931 			    __func__, ntohl(spi)));
2932 			*errp = EINVAL;
2933 			goto done;
2934 		}
2935 		break;
2936 	}
2937 
2938 	sav = malloc(sizeof(struct secasvar), M_IPSEC_SA, M_NOWAIT | M_ZERO);
2939 	if (sav == NULL) {
2940 		*errp = ENOBUFS;
2941 		goto done;
2942 	}
2943 	sav->lock = malloc(sizeof(struct mtx), M_IPSEC_MISC,
2944 	    M_NOWAIT | M_ZERO);
2945 	if (sav->lock == NULL) {
2946 		*errp = ENOBUFS;
2947 		goto done;
2948 	}
2949 	mtx_init(sav->lock, "ipsec association", NULL, MTX_DEF);
2950 	sav->lft_c = uma_zalloc_pcpu(V_key_lft_zone, M_NOWAIT);
2951 	if (sav->lft_c == NULL) {
2952 		*errp = ENOBUFS;
2953 		goto done;
2954 	}
2955 	counter_u64_zero(sav->lft_c_allocations);
2956 	counter_u64_zero(sav->lft_c_bytes);
2957 
2958 	sav->spi = spi;
2959 	sav->seq = mhp->msg->sadb_msg_seq;
2960 	sav->state = SADB_SASTATE_LARVAL;
2961 	sav->pid = (pid_t)mhp->msg->sadb_msg_pid;
2962 	SAV_INITREF(sav);
2963 again:
2964 	sah = key_getsah(saidx);
2965 	if (sah == NULL) {
2966 		/* create a new SA index */
2967 		sah = key_newsah(saidx);
2968 		if (sah == NULL) {
2969 			ipseclog((LOG_DEBUG,
2970 			    "%s: No more memory.\n", __func__));
2971 			*errp = ENOBUFS;
2972 			goto done;
2973 		}
2974 		isnew = 1;
2975 	} else
2976 		isnew = 0;
2977 
2978 	sav->sah = sah;
2979 	if (mhp->msg->sadb_msg_type == SADB_GETSPI) {
2980 		sav->created = time_second;
2981 	} else if (sav->state == SADB_SASTATE_LARVAL) {
2982 		/*
2983 		 * Do not call key_setsaval() second time in case
2984 		 * of `goto again`. We will have MATURE state.
2985 		 */
2986 		*errp = key_setsaval(sav, mhp);
2987 		if (*errp != 0)
2988 			goto done;
2989 		sav->state = SADB_SASTATE_MATURE;
2990 	}
2991 
2992 	SAHTREE_WLOCK();
2993 	/*
2994 	 * Check that existing SAH wasn't unlinked.
2995 	 * Since we didn't hold the SAHTREE lock, it is possible,
2996 	 * that callout handler or key_flush() or key_delete() could
2997 	 * unlink this SAH.
2998 	 */
2999 	if (isnew == 0 && sah->state == SADB_SASTATE_DEAD) {
3000 		SAHTREE_WUNLOCK();
3001 		key_freesah(&sah);	/* reference from key_getsah() */
3002 		goto again;
3003 	}
3004 	if (isnew != 0) {
3005 		/*
3006 		 * Add new SAH into SADB.
3007 		 *
3008 		 * XXXAE: we can serialize key_add and key_getspi calls, so
3009 		 * several threads will not fight in the race.
3010 		 * Otherwise we should check under SAHTREE lock, that this
3011 		 * SAH would not added twice.
3012 		 */
3013 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
3014 		/* Add new SAH into hash by addresses */
3015 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
3016 		/* Now we are linked in the chain */
3017 		sah->state = SADB_SASTATE_MATURE;
3018 		/*
3019 		 * SAV references this new SAH.
3020 		 * In case of existing SAH we reuse reference
3021 		 * from key_getsah().
3022 		 */
3023 		SAH_ADDREF(sah);
3024 	}
3025 	/* Link SAV with SAH */
3026 	if (sav->state == SADB_SASTATE_MATURE)
3027 		TAILQ_INSERT_HEAD(&sah->savtree_alive, sav, chain);
3028 	else
3029 		TAILQ_INSERT_HEAD(&sah->savtree_larval, sav, chain);
3030 	/* Add SAV into SPI hash */
3031 	LIST_INSERT_HEAD(SAVHASH_HASH(sav->spi), sav, spihash);
3032 	SAHTREE_WUNLOCK();
3033 	*errp = 0;	/* success */
3034 done:
3035 	if (*errp != 0) {
3036 		if (sav != NULL) {
3037 			if (sav->lock != NULL) {
3038 				mtx_destroy(sav->lock);
3039 				free(sav->lock, M_IPSEC_MISC);
3040 			}
3041 			if (sav->lft_c != NULL)
3042 				uma_zfree_pcpu(V_key_lft_zone, sav->lft_c);
3043 			free(sav, M_IPSEC_SA), sav = NULL;
3044 		}
3045 		if (sah != NULL)
3046 			key_freesah(&sah);
3047 		if (*errp == ENOBUFS) {
3048 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3049 			    __func__));
3050 			PFKEYSTAT_INC(in_nomem);
3051 		}
3052 	}
3053 	return (sav);
3054 }
3055 
3056 /*
3057  * free() SA variable entry.
3058  */
3059 static void
3060 key_cleansav(struct secasvar *sav)
3061 {
3062 
3063 	if (sav->natt != NULL) {
3064 		free(sav->natt, M_IPSEC_MISC);
3065 		sav->natt = NULL;
3066 	}
3067 	if (sav->flags & SADB_X_EXT_F_CLONED)
3068 		return;
3069 	/*
3070 	 * Cleanup xform state.  Note that zeroize'ing causes the
3071 	 * keys to be cleared; otherwise we must do it ourself.
3072 	 */
3073 	if (sav->tdb_xform != NULL) {
3074 		sav->tdb_xform->xf_zeroize(sav);
3075 		sav->tdb_xform = NULL;
3076 	} else {
3077 		if (sav->key_auth != NULL)
3078 			bzero(sav->key_auth->key_data, _KEYLEN(sav->key_auth));
3079 		if (sav->key_enc != NULL)
3080 			bzero(sav->key_enc->key_data, _KEYLEN(sav->key_enc));
3081 	}
3082 	if (sav->key_auth != NULL) {
3083 		if (sav->key_auth->key_data != NULL)
3084 			free(sav->key_auth->key_data, M_IPSEC_MISC);
3085 		free(sav->key_auth, M_IPSEC_MISC);
3086 		sav->key_auth = NULL;
3087 	}
3088 	if (sav->key_enc != NULL) {
3089 		if (sav->key_enc->key_data != NULL)
3090 			free(sav->key_enc->key_data, M_IPSEC_MISC);
3091 		free(sav->key_enc, M_IPSEC_MISC);
3092 		sav->key_enc = NULL;
3093 	}
3094 	if (sav->replay != NULL) {
3095 		if (sav->replay->bitmap != NULL)
3096 			free(sav->replay->bitmap, M_IPSEC_MISC);
3097 		free(sav->replay, M_IPSEC_MISC);
3098 		sav->replay = NULL;
3099 	}
3100 	if (sav->lft_h != NULL) {
3101 		free(sav->lft_h, M_IPSEC_MISC);
3102 		sav->lft_h = NULL;
3103 	}
3104 	if (sav->lft_s != NULL) {
3105 		free(sav->lft_s, M_IPSEC_MISC);
3106 		sav->lft_s = NULL;
3107 	}
3108 }
3109 
3110 /*
3111  * free() SA variable entry.
3112  */
3113 static void
3114 key_delsav(struct secasvar *sav)
3115 {
3116 	IPSEC_ASSERT(sav != NULL, ("null sav"));
3117 	IPSEC_ASSERT(sav->state == SADB_SASTATE_DEAD,
3118 	    ("attempt to free non DEAD SA %p", sav));
3119 	IPSEC_ASSERT(sav->refcnt == 0, ("reference count %u > 0",
3120 	    sav->refcnt));
3121 
3122 	/*
3123 	 * SA must be unlinked from the chain and hashtbl.
3124 	 * If SA was cloned, we leave all fields untouched,
3125 	 * except NAT-T config.
3126 	 */
3127 	key_cleansav(sav);
3128 	if ((sav->flags & SADB_X_EXT_F_CLONED) == 0) {
3129 		mtx_destroy(sav->lock);
3130 		free(sav->lock, M_IPSEC_MISC);
3131 		uma_zfree(V_key_lft_zone, sav->lft_c);
3132 	}
3133 	free(sav, M_IPSEC_SA);
3134 }
3135 
3136 /*
3137  * search SAH.
3138  * OUT:
3139  *	NULL	: not found
3140  *	others	: found, referenced pointer to a SAH.
3141  */
3142 static struct secashead *
3143 key_getsah(struct secasindex *saidx)
3144 {
3145 	SAHTREE_RLOCK_TRACKER;
3146 	struct secashead *sah;
3147 
3148 	SAHTREE_RLOCK();
3149 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
3150 	    if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID) != 0) {
3151 		    SAH_ADDREF(sah);
3152 		    break;
3153 	    }
3154 	}
3155 	SAHTREE_RUNLOCK();
3156 	return (sah);
3157 }
3158 
3159 /*
3160  * Check not to be duplicated SPI.
3161  * OUT:
3162  *	0	: not found
3163  *	1	: found SA with given SPI.
3164  */
3165 static int
3166 key_checkspidup(uint32_t spi)
3167 {
3168 	SAHTREE_RLOCK_TRACKER;
3169 	struct secasvar *sav;
3170 
3171 	/* Assume SPI is in network byte order */
3172 	SAHTREE_RLOCK();
3173 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3174 		if (sav->spi == spi)
3175 			break;
3176 	}
3177 	SAHTREE_RUNLOCK();
3178 	return (sav != NULL);
3179 }
3180 
3181 /*
3182  * Search SA by SPI.
3183  * OUT:
3184  *	NULL	: not found
3185  *	others	: found, referenced pointer to a SA.
3186  */
3187 static struct secasvar *
3188 key_getsavbyspi(uint32_t spi)
3189 {
3190 	SAHTREE_RLOCK_TRACKER;
3191 	struct secasvar *sav;
3192 
3193 	/* Assume SPI is in network byte order */
3194 	SAHTREE_RLOCK();
3195 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3196 		if (sav->spi != spi)
3197 			continue;
3198 		SAV_ADDREF(sav);
3199 		break;
3200 	}
3201 	SAHTREE_RUNLOCK();
3202 	return (sav);
3203 }
3204 
3205 static int
3206 key_updatelifetimes(struct secasvar *sav, const struct sadb_msghdr *mhp)
3207 {
3208 	struct seclifetime *lft_h, *lft_s, *tmp;
3209 
3210 	/* Lifetime extension is optional, check that it is present. */
3211 	if (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3212 	    SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) {
3213 		/*
3214 		 * In case of SADB_UPDATE we may need to change
3215 		 * existing lifetimes.
3216 		 */
3217 		if (sav->state == SADB_SASTATE_MATURE) {
3218 			lft_h = lft_s = NULL;
3219 			goto reset;
3220 		}
3221 		return (0);
3222 	}
3223 	/* Both HARD and SOFT extensions must present */
3224 	if ((SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3225 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
3226 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
3227 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
3228 		ipseclog((LOG_DEBUG,
3229 		    "%s: invalid message: missing required header.\n",
3230 		    __func__));
3231 		return (EINVAL);
3232 	}
3233 	if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD) ||
3234 	    SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_SOFT)) {
3235 		ipseclog((LOG_DEBUG,
3236 		    "%s: invalid message: wrong header size.\n", __func__));
3237 		return (EINVAL);
3238 	}
3239 	lft_h = key_dup_lifemsg((const struct sadb_lifetime *)
3240 	    mhp->ext[SADB_EXT_LIFETIME_HARD], M_IPSEC_MISC);
3241 	if (lft_h == NULL) {
3242 		PFKEYSTAT_INC(in_nomem);
3243 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3244 		return (ENOBUFS);
3245 	}
3246 	lft_s = key_dup_lifemsg((const struct sadb_lifetime *)
3247 	    mhp->ext[SADB_EXT_LIFETIME_SOFT], M_IPSEC_MISC);
3248 	if (lft_s == NULL) {
3249 		PFKEYSTAT_INC(in_nomem);
3250 		free(lft_h, M_IPSEC_MISC);
3251 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3252 		return (ENOBUFS);
3253 	}
3254 reset:
3255 	if (sav->state != SADB_SASTATE_LARVAL) {
3256 		/*
3257 		 * key_update() holds reference to this SA,
3258 		 * so it won't be deleted in meanwhile.
3259 		 */
3260 		SECASVAR_LOCK(sav);
3261 		tmp = sav->lft_h;
3262 		sav->lft_h = lft_h;
3263 		lft_h = tmp;
3264 
3265 		tmp = sav->lft_s;
3266 		sav->lft_s = lft_s;
3267 		lft_s = tmp;
3268 		SECASVAR_UNLOCK(sav);
3269 		if (lft_h != NULL)
3270 			free(lft_h, M_IPSEC_MISC);
3271 		if (lft_s != NULL)
3272 			free(lft_s, M_IPSEC_MISC);
3273 		return (0);
3274 	}
3275 	/* We can update lifetime without holding a lock */
3276 	IPSEC_ASSERT(sav->lft_h == NULL, ("lft_h is already initialized\n"));
3277 	IPSEC_ASSERT(sav->lft_s == NULL, ("lft_s is already initialized\n"));
3278 	sav->lft_h = lft_h;
3279 	sav->lft_s = lft_s;
3280 	return (0);
3281 }
3282 
3283 /*
3284  * copy SA values from PF_KEY message except *SPI, SEQ, PID and TYPE*.
3285  * You must update these if need. Expects only LARVAL SAs.
3286  * OUT:	0:	success.
3287  *	!0:	failure.
3288  */
3289 static int
3290 key_setsaval(struct secasvar *sav, const struct sadb_msghdr *mhp)
3291 {
3292 	const struct sadb_sa *sa0;
3293 	const struct sadb_key *key0;
3294 	uint32_t replay;
3295 	size_t len;
3296 	int error;
3297 
3298 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
3299 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
3300 	IPSEC_ASSERT(sav->state == SADB_SASTATE_LARVAL,
3301 	    ("Attempt to update non LARVAL SA"));
3302 
3303 	/* XXX rewrite */
3304 	error = key_setident(sav->sah, mhp);
3305 	if (error != 0)
3306 		goto fail;
3307 
3308 	/* SA */
3309 	if (!SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
3310 		if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
3311 			error = EINVAL;
3312 			goto fail;
3313 		}
3314 		sa0 = (const struct sadb_sa *)mhp->ext[SADB_EXT_SA];
3315 		sav->alg_auth = sa0->sadb_sa_auth;
3316 		sav->alg_enc = sa0->sadb_sa_encrypt;
3317 		sav->flags = sa0->sadb_sa_flags;
3318 		if ((sav->flags & SADB_KEY_FLAGS_MAX) != sav->flags) {
3319 			ipseclog((LOG_DEBUG,
3320 			    "%s: invalid sa_flags 0x%08x.\n", __func__,
3321 			    sav->flags));
3322 			error = EINVAL;
3323 			goto fail;
3324 		}
3325 
3326 		/* Optional replay window */
3327 		replay = 0;
3328 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0)
3329 			replay = sa0->sadb_sa_replay;
3330 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_SA_REPLAY)) {
3331 			if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA_REPLAY)) {
3332 				error = EINVAL;
3333 				goto fail;
3334 			}
3335 			replay = ((const struct sadb_x_sa_replay *)
3336 			    mhp->ext[SADB_X_EXT_SA_REPLAY])->sadb_x_sa_replay_replay;
3337 
3338 			if (replay > UINT32_MAX - 32) {
3339 				ipseclog((LOG_DEBUG,
3340 				    "%s: replay window too big.\n", __func__));
3341 				error = EINVAL;
3342 				goto fail;
3343 			}
3344 
3345 			replay = (replay + 7) >> 3;
3346 		}
3347 
3348 		sav->replay = malloc(sizeof(struct secreplay), M_IPSEC_MISC,
3349 		    M_NOWAIT | M_ZERO);
3350 		if (sav->replay == NULL) {
3351 			PFKEYSTAT_INC(in_nomem);
3352 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3353 			    __func__));
3354 			error = ENOBUFS;
3355 			goto fail;
3356 		}
3357 
3358 		if (replay != 0) {
3359 			/* number of 32b blocks to be allocated */
3360 			uint32_t bitmap_size;
3361 
3362 			/* RFC 6479:
3363 			 * - the allocated replay window size must be
3364 			 *   a power of two.
3365 			 * - use an extra 32b block as a redundant window.
3366 			 */
3367 			bitmap_size = 1;
3368 			while (replay + 4 > bitmap_size)
3369 				bitmap_size <<= 1;
3370 			bitmap_size = bitmap_size / 4;
3371 
3372 			sav->replay->bitmap = malloc(
3373 			    bitmap_size * sizeof(uint32_t), M_IPSEC_MISC,
3374 			    M_NOWAIT | M_ZERO);
3375 			if (sav->replay->bitmap == NULL) {
3376 				PFKEYSTAT_INC(in_nomem);
3377 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3378 					__func__));
3379 				error = ENOBUFS;
3380 				goto fail;
3381 			}
3382 			sav->replay->bitmap_size = bitmap_size;
3383 			sav->replay->wsize = replay;
3384 		}
3385 	}
3386 
3387 	/* Authentication keys */
3388 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
3389 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH)) {
3390 			error = EINVAL;
3391 			goto fail;
3392 		}
3393 		error = 0;
3394 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_AUTH];
3395 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3396 		switch (mhp->msg->sadb_msg_satype) {
3397 		case SADB_SATYPE_AH:
3398 		case SADB_SATYPE_ESP:
3399 		case SADB_X_SATYPE_TCPSIGNATURE:
3400 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3401 			    sav->alg_auth != SADB_X_AALG_NULL)
3402 				error = EINVAL;
3403 			break;
3404 		case SADB_X_SATYPE_IPCOMP:
3405 		default:
3406 			error = EINVAL;
3407 			break;
3408 		}
3409 		if (error) {
3410 			ipseclog((LOG_DEBUG, "%s: invalid key_auth values.\n",
3411 				__func__));
3412 			goto fail;
3413 		}
3414 
3415 		sav->key_auth = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3416 		if (sav->key_auth == NULL ) {
3417 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3418 				  __func__));
3419 			PFKEYSTAT_INC(in_nomem);
3420 			error = ENOBUFS;
3421 			goto fail;
3422 		}
3423 	}
3424 
3425 	/* Encryption key */
3426 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) {
3427 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT)) {
3428 			error = EINVAL;
3429 			goto fail;
3430 		}
3431 		error = 0;
3432 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_ENCRYPT];
3433 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3434 		switch (mhp->msg->sadb_msg_satype) {
3435 		case SADB_SATYPE_ESP:
3436 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3437 			    sav->alg_enc != SADB_EALG_NULL) {
3438 				error = EINVAL;
3439 				break;
3440 			}
3441 			sav->key_enc = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3442 			if (sav->key_enc == NULL) {
3443 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3444 					__func__));
3445 				PFKEYSTAT_INC(in_nomem);
3446 				error = ENOBUFS;
3447 				goto fail;
3448 			}
3449 			break;
3450 		case SADB_X_SATYPE_IPCOMP:
3451 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3452 				error = EINVAL;
3453 			sav->key_enc = NULL;	/*just in case*/
3454 			break;
3455 		case SADB_SATYPE_AH:
3456 		case SADB_X_SATYPE_TCPSIGNATURE:
3457 		default:
3458 			error = EINVAL;
3459 			break;
3460 		}
3461 		if (error) {
3462 			ipseclog((LOG_DEBUG, "%s: invalid key_enc value.\n",
3463 				__func__));
3464 			goto fail;
3465 		}
3466 	}
3467 
3468 	/* set iv */
3469 	sav->ivlen = 0;
3470 	switch (mhp->msg->sadb_msg_satype) {
3471 	case SADB_SATYPE_AH:
3472 		if (sav->flags & SADB_X_EXT_DERIV) {
3473 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3474 			    "given to AH SA.\n", __func__));
3475 			error = EINVAL;
3476 			goto fail;
3477 		}
3478 		if (sav->alg_enc != SADB_EALG_NONE) {
3479 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3480 			    "mismated.\n", __func__));
3481 			error = EINVAL;
3482 			goto fail;
3483 		}
3484 		error = xform_init(sav, XF_AH);
3485 		break;
3486 	case SADB_SATYPE_ESP:
3487 		if ((sav->flags & (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) ==
3488 		    (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) {
3489 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3490 			    "given to old-esp.\n", __func__));
3491 			error = EINVAL;
3492 			goto fail;
3493 		}
3494 		error = xform_init(sav, XF_ESP);
3495 		break;
3496 	case SADB_X_SATYPE_IPCOMP:
3497 		if (sav->alg_auth != SADB_AALG_NONE) {
3498 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3499 			    "mismated.\n", __func__));
3500 			error = EINVAL;
3501 			goto fail;
3502 		}
3503 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0 &&
3504 		    ntohl(sav->spi) >= 0x10000) {
3505 			ipseclog((LOG_DEBUG, "%s: invalid cpi for IPComp.\n",
3506 			    __func__));
3507 			error = EINVAL;
3508 			goto fail;
3509 		}
3510 		error = xform_init(sav, XF_IPCOMP);
3511 		break;
3512 	case SADB_X_SATYPE_TCPSIGNATURE:
3513 		if (sav->alg_enc != SADB_EALG_NONE) {
3514 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3515 			    "mismated.\n", __func__));
3516 			error = EINVAL;
3517 			goto fail;
3518 		}
3519 		error = xform_init(sav, XF_TCPSIGNATURE);
3520 		break;
3521 	default:
3522 		ipseclog((LOG_DEBUG, "%s: Invalid satype.\n", __func__));
3523 		error = EPROTONOSUPPORT;
3524 		goto fail;
3525 	}
3526 	if (error) {
3527 		ipseclog((LOG_DEBUG, "%s: unable to initialize SA type %u.\n",
3528 		    __func__, mhp->msg->sadb_msg_satype));
3529 		goto fail;
3530 	}
3531 
3532 	/* Handle NAT-T headers */
3533 	error = key_setnatt(sav, mhp);
3534 	if (error != 0)
3535 		goto fail;
3536 
3537 	/* Initialize lifetime for CURRENT */
3538 	sav->firstused = 0;
3539 	sav->created = time_second;
3540 
3541 	/* lifetimes for HARD and SOFT */
3542 	error = key_updatelifetimes(sav, mhp);
3543 	if (error == 0)
3544 		return (0);
3545 fail:
3546 	key_cleansav(sav);
3547 	return (error);
3548 }
3549 
3550 /*
3551  * subroutine for SADB_GET and SADB_DUMP.
3552  */
3553 static struct mbuf *
3554 key_setdumpsa(struct secasvar *sav, uint8_t type, uint8_t satype,
3555     uint32_t seq, uint32_t pid)
3556 {
3557 	struct seclifetime lft_c;
3558 	struct mbuf *result = NULL, *tres = NULL, *m;
3559 	int i, dumporder[] = {
3560 		SADB_EXT_SA, SADB_X_EXT_SA2, SADB_X_EXT_SA_REPLAY,
3561 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3562 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3563 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY,
3564 		SADB_EXT_KEY_AUTH, SADB_EXT_KEY_ENCRYPT,
3565 		SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
3566 		SADB_EXT_SENSITIVITY,
3567 		SADB_X_EXT_NAT_T_TYPE,
3568 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3569 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3570 		SADB_X_EXT_NAT_T_FRAG,
3571 	};
3572 	uint32_t replay_count;
3573 
3574 	m = key_setsadbmsg(type, 0, satype, seq, pid, sav->refcnt);
3575 	if (m == NULL)
3576 		goto fail;
3577 	result = m;
3578 
3579 	for (i = nitems(dumporder) - 1; i >= 0; i--) {
3580 		m = NULL;
3581 		switch (dumporder[i]) {
3582 		case SADB_EXT_SA:
3583 			m = key_setsadbsa(sav);
3584 			if (!m)
3585 				goto fail;
3586 			break;
3587 
3588 		case SADB_X_EXT_SA2:
3589 			SECASVAR_LOCK(sav);
3590 			replay_count = sav->replay ? sav->replay->count : 0;
3591 			SECASVAR_UNLOCK(sav);
3592 			m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
3593 					sav->sah->saidx.reqid);
3594 			if (!m)
3595 				goto fail;
3596 			break;
3597 
3598 		case SADB_X_EXT_SA_REPLAY:
3599 			if (sav->replay == NULL ||
3600 			    sav->replay->wsize <= UINT8_MAX)
3601 				continue;
3602 
3603 			m = key_setsadbxsareplay(sav->replay->wsize);
3604 			if (!m)
3605 				goto fail;
3606 			break;
3607 
3608 		case SADB_EXT_ADDRESS_SRC:
3609 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3610 			    &sav->sah->saidx.src.sa,
3611 			    FULLMASK, IPSEC_ULPROTO_ANY);
3612 			if (!m)
3613 				goto fail;
3614 			break;
3615 
3616 		case SADB_EXT_ADDRESS_DST:
3617 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3618 			    &sav->sah->saidx.dst.sa,
3619 			    FULLMASK, IPSEC_ULPROTO_ANY);
3620 			if (!m)
3621 				goto fail;
3622 			break;
3623 
3624 		case SADB_EXT_KEY_AUTH:
3625 			if (!sav->key_auth)
3626 				continue;
3627 			m = key_setkey(sav->key_auth, SADB_EXT_KEY_AUTH);
3628 			if (!m)
3629 				goto fail;
3630 			break;
3631 
3632 		case SADB_EXT_KEY_ENCRYPT:
3633 			if (!sav->key_enc)
3634 				continue;
3635 			m = key_setkey(sav->key_enc, SADB_EXT_KEY_ENCRYPT);
3636 			if (!m)
3637 				goto fail;
3638 			break;
3639 
3640 		case SADB_EXT_LIFETIME_CURRENT:
3641 			lft_c.addtime = sav->created;
3642 			lft_c.allocations = (uint32_t)counter_u64_fetch(
3643 			    sav->lft_c_allocations);
3644 			lft_c.bytes = counter_u64_fetch(sav->lft_c_bytes);
3645 			lft_c.usetime = sav->firstused;
3646 			m = key_setlifetime(&lft_c, SADB_EXT_LIFETIME_CURRENT);
3647 			if (!m)
3648 				goto fail;
3649 			break;
3650 
3651 		case SADB_EXT_LIFETIME_HARD:
3652 			if (!sav->lft_h)
3653 				continue;
3654 			m = key_setlifetime(sav->lft_h,
3655 					    SADB_EXT_LIFETIME_HARD);
3656 			if (!m)
3657 				goto fail;
3658 			break;
3659 
3660 		case SADB_EXT_LIFETIME_SOFT:
3661 			if (!sav->lft_s)
3662 				continue;
3663 			m = key_setlifetime(sav->lft_s,
3664 					    SADB_EXT_LIFETIME_SOFT);
3665 
3666 			if (!m)
3667 				goto fail;
3668 			break;
3669 
3670 		case SADB_X_EXT_NAT_T_TYPE:
3671 			if (sav->natt == NULL)
3672 				continue;
3673 			m = key_setsadbxtype(UDP_ENCAP_ESPINUDP);
3674 			if (!m)
3675 				goto fail;
3676 			break;
3677 
3678 		case SADB_X_EXT_NAT_T_DPORT:
3679 			if (sav->natt == NULL)
3680 				continue;
3681 			m = key_setsadbxport(sav->natt->dport,
3682 			    SADB_X_EXT_NAT_T_DPORT);
3683 			if (!m)
3684 				goto fail;
3685 			break;
3686 
3687 		case SADB_X_EXT_NAT_T_SPORT:
3688 			if (sav->natt == NULL)
3689 				continue;
3690 			m = key_setsadbxport(sav->natt->sport,
3691 			    SADB_X_EXT_NAT_T_SPORT);
3692 			if (!m)
3693 				goto fail;
3694 			break;
3695 
3696 		case SADB_X_EXT_NAT_T_OAI:
3697 			if (sav->natt == NULL ||
3698 			    (sav->natt->flags & IPSEC_NATT_F_OAI) == 0)
3699 				continue;
3700 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAI,
3701 			    &sav->natt->oai.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3702 			if (!m)
3703 				goto fail;
3704 			break;
3705 		case SADB_X_EXT_NAT_T_OAR:
3706 			if (sav->natt == NULL ||
3707 			    (sav->natt->flags & IPSEC_NATT_F_OAR) == 0)
3708 				continue;
3709 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAR,
3710 			    &sav->natt->oar.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3711 			if (!m)
3712 				goto fail;
3713 			break;
3714 		case SADB_X_EXT_NAT_T_FRAG:
3715 			/* We do not (yet) support those. */
3716 			continue;
3717 
3718 		case SADB_EXT_ADDRESS_PROXY:
3719 		case SADB_EXT_IDENTITY_SRC:
3720 		case SADB_EXT_IDENTITY_DST:
3721 			/* XXX: should we brought from SPD ? */
3722 		case SADB_EXT_SENSITIVITY:
3723 		default:
3724 			continue;
3725 		}
3726 
3727 		if (!m)
3728 			goto fail;
3729 		if (tres)
3730 			m_cat(m, tres);
3731 		tres = m;
3732 	}
3733 
3734 	m_cat(result, tres);
3735 	tres = NULL;
3736 	if (result->m_len < sizeof(struct sadb_msg)) {
3737 		result = m_pullup(result, sizeof(struct sadb_msg));
3738 		if (result == NULL)
3739 			goto fail;
3740 	}
3741 
3742 	result->m_pkthdr.len = 0;
3743 	for (m = result; m; m = m->m_next)
3744 		result->m_pkthdr.len += m->m_len;
3745 
3746 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3747 	    PFKEY_UNIT64(result->m_pkthdr.len);
3748 
3749 	return result;
3750 
3751 fail:
3752 	m_freem(result);
3753 	m_freem(tres);
3754 	return NULL;
3755 }
3756 
3757 /*
3758  * set data into sadb_msg.
3759  */
3760 static struct mbuf *
3761 key_setsadbmsg(u_int8_t type, u_int16_t tlen, u_int8_t satype, u_int32_t seq,
3762     pid_t pid, u_int16_t reserved)
3763 {
3764 	struct mbuf *m;
3765 	struct sadb_msg *p;
3766 	int len;
3767 
3768 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
3769 	if (len > MCLBYTES)
3770 		return NULL;
3771 	MGETHDR(m, M_NOWAIT, MT_DATA);
3772 	if (m && len > MHLEN) {
3773 		if (!(MCLGET(m, M_NOWAIT))) {
3774 			m_freem(m);
3775 			m = NULL;
3776 		}
3777 	}
3778 	if (!m)
3779 		return NULL;
3780 	m->m_pkthdr.len = m->m_len = len;
3781 	m->m_next = NULL;
3782 
3783 	p = mtod(m, struct sadb_msg *);
3784 
3785 	bzero(p, len);
3786 	p->sadb_msg_version = PF_KEY_V2;
3787 	p->sadb_msg_type = type;
3788 	p->sadb_msg_errno = 0;
3789 	p->sadb_msg_satype = satype;
3790 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
3791 	p->sadb_msg_reserved = reserved;
3792 	p->sadb_msg_seq = seq;
3793 	p->sadb_msg_pid = (u_int32_t)pid;
3794 
3795 	return m;
3796 }
3797 
3798 /*
3799  * copy secasvar data into sadb_address.
3800  */
3801 static struct mbuf *
3802 key_setsadbsa(struct secasvar *sav)
3803 {
3804 	struct mbuf *m;
3805 	struct sadb_sa *p;
3806 	int len;
3807 
3808 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
3809 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3810 	if (m == NULL)
3811 		return (NULL);
3812 	m_align(m, len);
3813 	m->m_len = len;
3814 	p = mtod(m, struct sadb_sa *);
3815 	bzero(p, len);
3816 	p->sadb_sa_len = PFKEY_UNIT64(len);
3817 	p->sadb_sa_exttype = SADB_EXT_SA;
3818 	p->sadb_sa_spi = sav->spi;
3819 	p->sadb_sa_replay = sav->replay ?
3820 	    (sav->replay->wsize > UINT8_MAX ? UINT8_MAX :
3821 		sav->replay->wsize): 0;
3822 	p->sadb_sa_state = sav->state;
3823 	p->sadb_sa_auth = sav->alg_auth;
3824 	p->sadb_sa_encrypt = sav->alg_enc;
3825 	p->sadb_sa_flags = sav->flags & SADB_KEY_FLAGS_MAX;
3826 	return (m);
3827 }
3828 
3829 /*
3830  * set data into sadb_address.
3831  */
3832 static struct mbuf *
3833 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
3834     u_int8_t prefixlen, u_int16_t ul_proto)
3835 {
3836 	struct mbuf *m;
3837 	struct sadb_address *p;
3838 	size_t len;
3839 
3840 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
3841 	    PFKEY_ALIGN8(saddr->sa_len);
3842 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3843 	if (m == NULL)
3844 		return (NULL);
3845 	m_align(m, len);
3846 	m->m_len = len;
3847 	p = mtod(m, struct sadb_address *);
3848 
3849 	bzero(p, len);
3850 	p->sadb_address_len = PFKEY_UNIT64(len);
3851 	p->sadb_address_exttype = exttype;
3852 	p->sadb_address_proto = ul_proto;
3853 	if (prefixlen == FULLMASK) {
3854 		switch (saddr->sa_family) {
3855 		case AF_INET:
3856 			prefixlen = sizeof(struct in_addr) << 3;
3857 			break;
3858 		case AF_INET6:
3859 			prefixlen = sizeof(struct in6_addr) << 3;
3860 			break;
3861 		default:
3862 			; /*XXX*/
3863 		}
3864 	}
3865 	p->sadb_address_prefixlen = prefixlen;
3866 	p->sadb_address_reserved = 0;
3867 
3868 	bcopy(saddr,
3869 	    mtod(m, caddr_t) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
3870 	    saddr->sa_len);
3871 
3872 	return m;
3873 }
3874 
3875 /*
3876  * set data into sadb_x_sa2.
3877  */
3878 static struct mbuf *
3879 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int32_t reqid)
3880 {
3881 	struct mbuf *m;
3882 	struct sadb_x_sa2 *p;
3883 	size_t len;
3884 
3885 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
3886 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3887 	if (m == NULL)
3888 		return (NULL);
3889 	m_align(m, len);
3890 	m->m_len = len;
3891 	p = mtod(m, struct sadb_x_sa2 *);
3892 
3893 	bzero(p, len);
3894 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
3895 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
3896 	p->sadb_x_sa2_mode = mode;
3897 	p->sadb_x_sa2_reserved1 = 0;
3898 	p->sadb_x_sa2_reserved2 = 0;
3899 	p->sadb_x_sa2_sequence = seq;
3900 	p->sadb_x_sa2_reqid = reqid;
3901 
3902 	return m;
3903 }
3904 
3905 /*
3906  * Set data into sadb_x_sa_replay.
3907  */
3908 static struct mbuf *
3909 key_setsadbxsareplay(u_int32_t replay)
3910 {
3911 	struct mbuf *m;
3912 	struct sadb_x_sa_replay *p;
3913 	size_t len;
3914 
3915 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa_replay));
3916 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3917 	if (m == NULL)
3918 		return (NULL);
3919 	m_align(m, len);
3920 	m->m_len = len;
3921 	p = mtod(m, struct sadb_x_sa_replay *);
3922 
3923 	bzero(p, len);
3924 	p->sadb_x_sa_replay_len = PFKEY_UNIT64(len);
3925 	p->sadb_x_sa_replay_exttype = SADB_X_EXT_SA_REPLAY;
3926 	p->sadb_x_sa_replay_replay = (replay << 3);
3927 
3928 	return m;
3929 }
3930 
3931 /*
3932  * Set a type in sadb_x_nat_t_type.
3933  */
3934 static struct mbuf *
3935 key_setsadbxtype(u_int16_t type)
3936 {
3937 	struct mbuf *m;
3938 	size_t len;
3939 	struct sadb_x_nat_t_type *p;
3940 
3941 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
3942 
3943 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3944 	if (m == NULL)
3945 		return (NULL);
3946 	m_align(m, len);
3947 	m->m_len = len;
3948 	p = mtod(m, struct sadb_x_nat_t_type *);
3949 
3950 	bzero(p, len);
3951 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
3952 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
3953 	p->sadb_x_nat_t_type_type = type;
3954 
3955 	return (m);
3956 }
3957 /*
3958  * Set a port in sadb_x_nat_t_port.
3959  * In contrast to default RFC 2367 behaviour, port is in network byte order.
3960  */
3961 static struct mbuf *
3962 key_setsadbxport(u_int16_t port, u_int16_t type)
3963 {
3964 	struct mbuf *m;
3965 	size_t len;
3966 	struct sadb_x_nat_t_port *p;
3967 
3968 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
3969 
3970 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3971 	if (m == NULL)
3972 		return (NULL);
3973 	m_align(m, len);
3974 	m->m_len = len;
3975 	p = mtod(m, struct sadb_x_nat_t_port *);
3976 
3977 	bzero(p, len);
3978 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
3979 	p->sadb_x_nat_t_port_exttype = type;
3980 	p->sadb_x_nat_t_port_port = port;
3981 
3982 	return (m);
3983 }
3984 
3985 /*
3986  * Get port from sockaddr. Port is in network byte order.
3987  */
3988 uint16_t
3989 key_portfromsaddr(struct sockaddr *sa)
3990 {
3991 
3992 	switch (sa->sa_family) {
3993 #ifdef INET
3994 	case AF_INET:
3995 		return ((struct sockaddr_in *)sa)->sin_port;
3996 #endif
3997 #ifdef INET6
3998 	case AF_INET6:
3999 		return ((struct sockaddr_in6 *)sa)->sin6_port;
4000 #endif
4001 	}
4002 	return (0);
4003 }
4004 
4005 /*
4006  * Set port in struct sockaddr. Port is in network byte order.
4007  */
4008 void
4009 key_porttosaddr(struct sockaddr *sa, uint16_t port)
4010 {
4011 
4012 	switch (sa->sa_family) {
4013 #ifdef INET
4014 	case AF_INET:
4015 		((struct sockaddr_in *)sa)->sin_port = port;
4016 		break;
4017 #endif
4018 #ifdef INET6
4019 	case AF_INET6:
4020 		((struct sockaddr_in6 *)sa)->sin6_port = port;
4021 		break;
4022 #endif
4023 	default:
4024 		ipseclog((LOG_DEBUG, "%s: unexpected address family %d.\n",
4025 			__func__, sa->sa_family));
4026 		break;
4027 	}
4028 }
4029 
4030 /*
4031  * set data into sadb_x_policy
4032  */
4033 static struct mbuf *
4034 key_setsadbxpolicy(u_int16_t type, u_int8_t dir, u_int32_t id, u_int32_t priority)
4035 {
4036 	struct mbuf *m;
4037 	struct sadb_x_policy *p;
4038 	size_t len;
4039 
4040 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
4041 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4042 	if (m == NULL)
4043 		return (NULL);
4044 	m_align(m, len);
4045 	m->m_len = len;
4046 	p = mtod(m, struct sadb_x_policy *);
4047 
4048 	bzero(p, len);
4049 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
4050 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
4051 	p->sadb_x_policy_type = type;
4052 	p->sadb_x_policy_dir = dir;
4053 	p->sadb_x_policy_id = id;
4054 	p->sadb_x_policy_priority = priority;
4055 
4056 	return m;
4057 }
4058 
4059 /* %%% utilities */
4060 /* Take a key message (sadb_key) from the socket and turn it into one
4061  * of the kernel's key structures (seckey).
4062  *
4063  * IN: pointer to the src
4064  * OUT: NULL no more memory
4065  */
4066 struct seckey *
4067 key_dup_keymsg(const struct sadb_key *src, size_t len,
4068     struct malloc_type *type)
4069 {
4070 	struct seckey *dst;
4071 
4072 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4073 	if (dst != NULL) {
4074 		dst->bits = src->sadb_key_bits;
4075 		dst->key_data = malloc(len, type, M_NOWAIT);
4076 		if (dst->key_data != NULL) {
4077 			bcopy((const char *)(src + 1), dst->key_data, len);
4078 		} else {
4079 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4080 			    __func__));
4081 			free(dst, type);
4082 			dst = NULL;
4083 		}
4084 	} else {
4085 		ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4086 		    __func__));
4087 
4088 	}
4089 	return (dst);
4090 }
4091 
4092 /* Take a lifetime message (sadb_lifetime) passed in on a socket and
4093  * turn it into one of the kernel's lifetime structures (seclifetime).
4094  *
4095  * IN: pointer to the destination, source and malloc type
4096  * OUT: NULL, no more memory
4097  */
4098 
4099 static struct seclifetime *
4100 key_dup_lifemsg(const struct sadb_lifetime *src, struct malloc_type *type)
4101 {
4102 	struct seclifetime *dst;
4103 
4104 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4105 	if (dst == NULL) {
4106 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
4107 		return (NULL);
4108 	}
4109 	dst->allocations = src->sadb_lifetime_allocations;
4110 	dst->bytes = src->sadb_lifetime_bytes;
4111 	dst->addtime = src->sadb_lifetime_addtime;
4112 	dst->usetime = src->sadb_lifetime_usetime;
4113 	return (dst);
4114 }
4115 
4116 /*
4117  * compare two secasindex structure.
4118  * flag can specify to compare 2 saidxes.
4119  * compare two secasindex structure without both mode and reqid.
4120  * don't compare port.
4121  * IN:
4122  *      saidx0: source, it can be in SAD.
4123  *      saidx1: object.
4124  * OUT:
4125  *      1 : equal
4126  *      0 : not equal
4127  */
4128 static int
4129 key_cmpsaidx(const struct secasindex *saidx0, const struct secasindex *saidx1,
4130     int flag)
4131 {
4132 
4133 	/* sanity */
4134 	if (saidx0 == NULL && saidx1 == NULL)
4135 		return 1;
4136 
4137 	if (saidx0 == NULL || saidx1 == NULL)
4138 		return 0;
4139 
4140 	if (saidx0->proto != saidx1->proto)
4141 		return 0;
4142 
4143 	if (flag == CMP_EXACTLY) {
4144 		if (saidx0->mode != saidx1->mode)
4145 			return 0;
4146 		if (saidx0->reqid != saidx1->reqid)
4147 			return 0;
4148 		if (bcmp(&saidx0->src, &saidx1->src,
4149 		    saidx0->src.sa.sa_len) != 0 ||
4150 		    bcmp(&saidx0->dst, &saidx1->dst,
4151 		    saidx0->dst.sa.sa_len) != 0)
4152 			return 0;
4153 	} else {
4154 
4155 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
4156 		if (flag == CMP_MODE_REQID || flag == CMP_REQID) {
4157 			/*
4158 			 * If reqid of SPD is non-zero, unique SA is required.
4159 			 * The result must be of same reqid in this case.
4160 			 */
4161 			if (saidx1->reqid != 0 &&
4162 			    saidx0->reqid != saidx1->reqid)
4163 				return 0;
4164 		}
4165 
4166 		if (flag == CMP_MODE_REQID) {
4167 			if (saidx0->mode != IPSEC_MODE_ANY
4168 			 && saidx0->mode != saidx1->mode)
4169 				return 0;
4170 		}
4171 
4172 		if (key_sockaddrcmp(&saidx0->src.sa, &saidx1->src.sa, 0) != 0)
4173 			return 0;
4174 		if (key_sockaddrcmp(&saidx0->dst.sa, &saidx1->dst.sa, 0) != 0)
4175 			return 0;
4176 	}
4177 
4178 	return 1;
4179 }
4180 
4181 /*
4182  * compare two secindex structure exactly.
4183  * IN:
4184  *	spidx0: source, it is often in SPD.
4185  *	spidx1: object, it is often from PFKEY message.
4186  * OUT:
4187  *	1 : equal
4188  *	0 : not equal
4189  */
4190 static int
4191 key_cmpspidx_exactly(struct secpolicyindex *spidx0,
4192     struct secpolicyindex *spidx1)
4193 {
4194 	/* sanity */
4195 	if (spidx0 == NULL && spidx1 == NULL)
4196 		return 1;
4197 
4198 	if (spidx0 == NULL || spidx1 == NULL)
4199 		return 0;
4200 
4201 	if (spidx0->prefs != spidx1->prefs
4202 	 || spidx0->prefd != spidx1->prefd
4203 	 || spidx0->ul_proto != spidx1->ul_proto
4204 	 || spidx0->dir != spidx1->dir)
4205 		return 0;
4206 
4207 	return key_sockaddrcmp(&spidx0->src.sa, &spidx1->src.sa, 1) == 0 &&
4208 	       key_sockaddrcmp(&spidx0->dst.sa, &spidx1->dst.sa, 1) == 0;
4209 }
4210 
4211 /*
4212  * compare two secindex structure with mask.
4213  * IN:
4214  *	spidx0: source, it is often in SPD.
4215  *	spidx1: object, it is often from IP header.
4216  * OUT:
4217  *	1 : equal
4218  *	0 : not equal
4219  */
4220 static int
4221 key_cmpspidx_withmask(struct secpolicyindex *spidx0,
4222     struct secpolicyindex *spidx1)
4223 {
4224 	/* sanity */
4225 	if (spidx0 == NULL && spidx1 == NULL)
4226 		return 1;
4227 
4228 	if (spidx0 == NULL || spidx1 == NULL)
4229 		return 0;
4230 
4231 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4232 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4233 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4234 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
4235 		return 0;
4236 
4237 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4238 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY
4239 	 && spidx0->ul_proto != spidx1->ul_proto)
4240 		return 0;
4241 
4242 	switch (spidx0->src.sa.sa_family) {
4243 	case AF_INET:
4244 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY
4245 		 && spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
4246 			return 0;
4247 		if (!key_bbcmp(&spidx0->src.sin.sin_addr,
4248 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
4249 			return 0;
4250 		break;
4251 	case AF_INET6:
4252 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY
4253 		 && spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
4254 			return 0;
4255 		/*
4256 		 * scope_id check. if sin6_scope_id is 0, we regard it
4257 		 * as a wildcard scope, which matches any scope zone ID.
4258 		 */
4259 		if (spidx0->src.sin6.sin6_scope_id &&
4260 		    spidx1->src.sin6.sin6_scope_id &&
4261 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
4262 			return 0;
4263 		if (!key_bbcmp(&spidx0->src.sin6.sin6_addr,
4264 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
4265 			return 0;
4266 		break;
4267 	default:
4268 		/* XXX */
4269 		if (bcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
4270 			return 0;
4271 		break;
4272 	}
4273 
4274 	switch (spidx0->dst.sa.sa_family) {
4275 	case AF_INET:
4276 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY
4277 		 && spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
4278 			return 0;
4279 		if (!key_bbcmp(&spidx0->dst.sin.sin_addr,
4280 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
4281 			return 0;
4282 		break;
4283 	case AF_INET6:
4284 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY
4285 		 && spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
4286 			return 0;
4287 		/*
4288 		 * scope_id check. if sin6_scope_id is 0, we regard it
4289 		 * as a wildcard scope, which matches any scope zone ID.
4290 		 */
4291 		if (spidx0->dst.sin6.sin6_scope_id &&
4292 		    spidx1->dst.sin6.sin6_scope_id &&
4293 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
4294 			return 0;
4295 		if (!key_bbcmp(&spidx0->dst.sin6.sin6_addr,
4296 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
4297 			return 0;
4298 		break;
4299 	default:
4300 		/* XXX */
4301 		if (bcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
4302 			return 0;
4303 		break;
4304 	}
4305 
4306 	/* XXX Do we check other field ?  e.g. flowinfo */
4307 
4308 	return 1;
4309 }
4310 
4311 #ifdef satosin
4312 #undef satosin
4313 #endif
4314 #define satosin(s) ((const struct sockaddr_in *)s)
4315 #ifdef satosin6
4316 #undef satosin6
4317 #endif
4318 #define satosin6(s) ((const struct sockaddr_in6 *)s)
4319 /* returns 0 on match */
4320 int
4321 key_sockaddrcmp(const struct sockaddr *sa1, const struct sockaddr *sa2,
4322     int port)
4323 {
4324 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4325 		return 1;
4326 
4327 	switch (sa1->sa_family) {
4328 #ifdef INET
4329 	case AF_INET:
4330 		if (sa1->sa_len != sizeof(struct sockaddr_in))
4331 			return 1;
4332 		if (satosin(sa1)->sin_addr.s_addr !=
4333 		    satosin(sa2)->sin_addr.s_addr) {
4334 			return 1;
4335 		}
4336 		if (port && satosin(sa1)->sin_port != satosin(sa2)->sin_port)
4337 			return 1;
4338 		break;
4339 #endif
4340 #ifdef INET6
4341 	case AF_INET6:
4342 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4343 			return 1;	/*EINVAL*/
4344 		if (satosin6(sa1)->sin6_scope_id !=
4345 		    satosin6(sa2)->sin6_scope_id) {
4346 			return 1;
4347 		}
4348 		if (!IN6_ARE_ADDR_EQUAL(&satosin6(sa1)->sin6_addr,
4349 		    &satosin6(sa2)->sin6_addr)) {
4350 			return 1;
4351 		}
4352 		if (port &&
4353 		    satosin6(sa1)->sin6_port != satosin6(sa2)->sin6_port) {
4354 			return 1;
4355 		}
4356 		break;
4357 #endif
4358 	default:
4359 		if (bcmp(sa1, sa2, sa1->sa_len) != 0)
4360 			return 1;
4361 		break;
4362 	}
4363 
4364 	return 0;
4365 }
4366 
4367 /* returns 0 on match */
4368 int
4369 key_sockaddrcmp_withmask(const struct sockaddr *sa1,
4370     const struct sockaddr *sa2, size_t mask)
4371 {
4372 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4373 		return (1);
4374 
4375 	switch (sa1->sa_family) {
4376 #ifdef INET
4377 	case AF_INET:
4378 		return (!key_bbcmp(&satosin(sa1)->sin_addr,
4379 		    &satosin(sa2)->sin_addr, mask));
4380 #endif
4381 #ifdef INET6
4382 	case AF_INET6:
4383 		if (satosin6(sa1)->sin6_scope_id !=
4384 		    satosin6(sa2)->sin6_scope_id)
4385 			return (1);
4386 		return (!key_bbcmp(&satosin6(sa1)->sin6_addr,
4387 		    &satosin6(sa2)->sin6_addr, mask));
4388 #endif
4389 	}
4390 	return (1);
4391 }
4392 #undef satosin
4393 #undef satosin6
4394 
4395 /*
4396  * compare two buffers with mask.
4397  * IN:
4398  *	addr1: source
4399  *	addr2: object
4400  *	bits:  Number of bits to compare
4401  * OUT:
4402  *	1 : equal
4403  *	0 : not equal
4404  */
4405 static int
4406 key_bbcmp(const void *a1, const void *a2, u_int bits)
4407 {
4408 	const unsigned char *p1 = a1;
4409 	const unsigned char *p2 = a2;
4410 
4411 	/* XXX: This could be considerably faster if we compare a word
4412 	 * at a time, but it is complicated on LSB Endian machines */
4413 
4414 	/* Handle null pointers */
4415 	if (p1 == NULL || p2 == NULL)
4416 		return (p1 == p2);
4417 
4418 	while (bits >= 8) {
4419 		if (*p1++ != *p2++)
4420 			return 0;
4421 		bits -= 8;
4422 	}
4423 
4424 	if (bits > 0) {
4425 		u_int8_t mask = ~((1<<(8-bits))-1);
4426 		if ((*p1 & mask) != (*p2 & mask))
4427 			return 0;
4428 	}
4429 	return 1;	/* Match! */
4430 }
4431 
4432 static void
4433 key_flush_spd(time_t now)
4434 {
4435 	SPTREE_RLOCK_TRACKER;
4436 	struct secpolicy_list drainq;
4437 	struct secpolicy *sp, *nextsp;
4438 	u_int dir;
4439 
4440 	LIST_INIT(&drainq);
4441 	SPTREE_RLOCK();
4442 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4443 		TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
4444 			if (sp->lifetime == 0 && sp->validtime == 0)
4445 				continue;
4446 			if ((sp->lifetime &&
4447 			    now - sp->created > sp->lifetime) ||
4448 			    (sp->validtime &&
4449 			    now - sp->lastused > sp->validtime)) {
4450 				/* Hold extra reference to send SPDEXPIRE */
4451 				SP_ADDREF(sp);
4452 				LIST_INSERT_HEAD(&drainq, sp, drainq);
4453 			}
4454 		}
4455 	}
4456 	SPTREE_RUNLOCK();
4457 	if (LIST_EMPTY(&drainq))
4458 		return;
4459 
4460 	SPTREE_WLOCK();
4461 	sp = LIST_FIRST(&drainq);
4462 	while (sp != NULL) {
4463 		nextsp = LIST_NEXT(sp, drainq);
4464 		/* Check that SP is still linked */
4465 		if (sp->state != IPSEC_SPSTATE_ALIVE) {
4466 			LIST_REMOVE(sp, drainq);
4467 			key_freesp(&sp); /* release extra reference */
4468 			sp = nextsp;
4469 			continue;
4470 		}
4471 		TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
4472 		V_spd_size--;
4473 		LIST_REMOVE(sp, idhash);
4474 		sp->state = IPSEC_SPSTATE_DEAD;
4475 		sp = nextsp;
4476 	}
4477 	V_sp_genid++;
4478 	SPTREE_WUNLOCK();
4479 	if (SPDCACHE_ENABLED())
4480 		spdcache_clear();
4481 
4482 	sp = LIST_FIRST(&drainq);
4483 	while (sp != NULL) {
4484 		nextsp = LIST_NEXT(sp, drainq);
4485 		key_spdexpire(sp);
4486 		key_freesp(&sp); /* release extra reference */
4487 		key_freesp(&sp); /* release last reference */
4488 		sp = nextsp;
4489 	}
4490 }
4491 
4492 static void
4493 key_flush_sad(time_t now)
4494 {
4495 	SAHTREE_RLOCK_TRACKER;
4496 	struct secashead_list emptyq;
4497 	struct secasvar_list drainq, hexpireq, sexpireq, freeq;
4498 	struct secashead *sah, *nextsah;
4499 	struct secasvar *sav, *nextsav;
4500 
4501 	LIST_INIT(&drainq);
4502 	LIST_INIT(&hexpireq);
4503 	LIST_INIT(&sexpireq);
4504 	LIST_INIT(&emptyq);
4505 
4506 	SAHTREE_RLOCK();
4507 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
4508 		/* Check for empty SAH */
4509 		if (TAILQ_EMPTY(&sah->savtree_larval) &&
4510 		    TAILQ_EMPTY(&sah->savtree_alive)) {
4511 			SAH_ADDREF(sah);
4512 			LIST_INSERT_HEAD(&emptyq, sah, drainq);
4513 			continue;
4514 		}
4515 		/* Add all stale LARVAL SAs into drainq */
4516 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
4517 			if (now - sav->created < V_key_larval_lifetime)
4518 				continue;
4519 			SAV_ADDREF(sav);
4520 			LIST_INSERT_HEAD(&drainq, sav, drainq);
4521 		}
4522 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
4523 			/* lifetimes aren't specified */
4524 			if (sav->lft_h == NULL)
4525 				continue;
4526 			SECASVAR_LOCK(sav);
4527 			/*
4528 			 * Check again with lock held, because it may
4529 			 * be updated by SADB_UPDATE.
4530 			 */
4531 			if (sav->lft_h == NULL) {
4532 				SECASVAR_UNLOCK(sav);
4533 				continue;
4534 			}
4535 			/*
4536 			 * RFC 2367:
4537 			 * HARD lifetimes MUST take precedence over SOFT
4538 			 * lifetimes, meaning if the HARD and SOFT lifetimes
4539 			 * are the same, the HARD lifetime will appear on the
4540 			 * EXPIRE message.
4541 			 */
4542 			/* check HARD lifetime */
4543 			if ((sav->lft_h->addtime != 0 &&
4544 			    now - sav->created > sav->lft_h->addtime) ||
4545 			    (sav->lft_h->usetime != 0 && sav->firstused &&
4546 			    now - sav->firstused > sav->lft_h->usetime) ||
4547 			    (sav->lft_h->bytes != 0 && counter_u64_fetch(
4548 			        sav->lft_c_bytes) > sav->lft_h->bytes)) {
4549 				SECASVAR_UNLOCK(sav);
4550 				SAV_ADDREF(sav);
4551 				LIST_INSERT_HEAD(&hexpireq, sav, drainq);
4552 				continue;
4553 			}
4554 			/* check SOFT lifetime (only for MATURE SAs) */
4555 			if (sav->state == SADB_SASTATE_MATURE && (
4556 			    (sav->lft_s->addtime != 0 &&
4557 			    now - sav->created > sav->lft_s->addtime) ||
4558 			    (sav->lft_s->usetime != 0 && sav->firstused &&
4559 			    now - sav->firstused > sav->lft_s->usetime) ||
4560 			    (sav->lft_s->bytes != 0 && counter_u64_fetch(
4561 				sav->lft_c_bytes) > sav->lft_s->bytes))) {
4562 				SECASVAR_UNLOCK(sav);
4563 				SAV_ADDREF(sav);
4564 				LIST_INSERT_HEAD(&sexpireq, sav, drainq);
4565 				continue;
4566 			}
4567 			SECASVAR_UNLOCK(sav);
4568 		}
4569 	}
4570 	SAHTREE_RUNLOCK();
4571 
4572 	if (LIST_EMPTY(&emptyq) && LIST_EMPTY(&drainq) &&
4573 	    LIST_EMPTY(&hexpireq) && LIST_EMPTY(&sexpireq))
4574 		return;
4575 
4576 	LIST_INIT(&freeq);
4577 	SAHTREE_WLOCK();
4578 	/* Unlink stale LARVAL SAs */
4579 	sav = LIST_FIRST(&drainq);
4580 	while (sav != NULL) {
4581 		nextsav = LIST_NEXT(sav, drainq);
4582 		/* Check that SA is still LARVAL */
4583 		if (sav->state != SADB_SASTATE_LARVAL) {
4584 			LIST_REMOVE(sav, drainq);
4585 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4586 			sav = nextsav;
4587 			continue;
4588 		}
4589 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
4590 		LIST_REMOVE(sav, spihash);
4591 		sav->state = SADB_SASTATE_DEAD;
4592 		sav = nextsav;
4593 	}
4594 	/* Unlink all SAs with expired HARD lifetime */
4595 	sav = LIST_FIRST(&hexpireq);
4596 	while (sav != NULL) {
4597 		nextsav = LIST_NEXT(sav, drainq);
4598 		/* Check that SA is not unlinked */
4599 		if (sav->state == SADB_SASTATE_DEAD) {
4600 			LIST_REMOVE(sav, drainq);
4601 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4602 			sav = nextsav;
4603 			continue;
4604 		}
4605 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
4606 		LIST_REMOVE(sav, spihash);
4607 		sav->state = SADB_SASTATE_DEAD;
4608 		sav = nextsav;
4609 	}
4610 	/* Mark all SAs with expired SOFT lifetime as DYING */
4611 	sav = LIST_FIRST(&sexpireq);
4612 	while (sav != NULL) {
4613 		nextsav = LIST_NEXT(sav, drainq);
4614 		/* Check that SA is not unlinked */
4615 		if (sav->state == SADB_SASTATE_DEAD) {
4616 			LIST_REMOVE(sav, drainq);
4617 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4618 			sav = nextsav;
4619 			continue;
4620 		}
4621 		/*
4622 		 * NOTE: this doesn't change SA order in the chain.
4623 		 */
4624 		sav->state = SADB_SASTATE_DYING;
4625 		sav = nextsav;
4626 	}
4627 	/* Unlink empty SAHs */
4628 	sah = LIST_FIRST(&emptyq);
4629 	while (sah != NULL) {
4630 		nextsah = LIST_NEXT(sah, drainq);
4631 		/* Check that SAH is still empty and not unlinked */
4632 		if (sah->state == SADB_SASTATE_DEAD ||
4633 		    !TAILQ_EMPTY(&sah->savtree_larval) ||
4634 		    !TAILQ_EMPTY(&sah->savtree_alive)) {
4635 			LIST_REMOVE(sah, drainq);
4636 			key_freesah(&sah); /* release extra reference */
4637 			sah = nextsah;
4638 			continue;
4639 		}
4640 		TAILQ_REMOVE(&V_sahtree, sah, chain);
4641 		LIST_REMOVE(sah, addrhash);
4642 		sah->state = SADB_SASTATE_DEAD;
4643 		sah = nextsah;
4644 	}
4645 	SAHTREE_WUNLOCK();
4646 
4647 	/* Send SPDEXPIRE messages */
4648 	sav = LIST_FIRST(&hexpireq);
4649 	while (sav != NULL) {
4650 		nextsav = LIST_NEXT(sav, drainq);
4651 		key_expire(sav, 1);
4652 		key_freesah(&sav->sah); /* release reference from SAV */
4653 		key_freesav(&sav); /* release extra reference */
4654 		key_freesav(&sav); /* release last reference */
4655 		sav = nextsav;
4656 	}
4657 	sav = LIST_FIRST(&sexpireq);
4658 	while (sav != NULL) {
4659 		nextsav = LIST_NEXT(sav, drainq);
4660 		key_expire(sav, 0);
4661 		key_freesav(&sav); /* release extra reference */
4662 		sav = nextsav;
4663 	}
4664 	/* Free stale LARVAL SAs */
4665 	sav = LIST_FIRST(&drainq);
4666 	while (sav != NULL) {
4667 		nextsav = LIST_NEXT(sav, drainq);
4668 		key_freesah(&sav->sah); /* release reference from SAV */
4669 		key_freesav(&sav); /* release extra reference */
4670 		key_freesav(&sav); /* release last reference */
4671 		sav = nextsav;
4672 	}
4673 	/* Free SAs that were unlinked/changed by someone else */
4674 	sav = LIST_FIRST(&freeq);
4675 	while (sav != NULL) {
4676 		nextsav = LIST_NEXT(sav, drainq);
4677 		key_freesav(&sav); /* release extra reference */
4678 		sav = nextsav;
4679 	}
4680 	/* Free empty SAH */
4681 	sah = LIST_FIRST(&emptyq);
4682 	while (sah != NULL) {
4683 		nextsah = LIST_NEXT(sah, drainq);
4684 		key_freesah(&sah); /* release extra reference */
4685 		key_freesah(&sah); /* release last reference */
4686 		sah = nextsah;
4687 	}
4688 }
4689 
4690 static void
4691 key_flush_acq(time_t now)
4692 {
4693 	struct secacq *acq, *nextacq;
4694 
4695 	/* ACQ tree */
4696 	ACQ_LOCK();
4697 	acq = LIST_FIRST(&V_acqtree);
4698 	while (acq != NULL) {
4699 		nextacq = LIST_NEXT(acq, chain);
4700 		if (now - acq->created > V_key_blockacq_lifetime) {
4701 			LIST_REMOVE(acq, chain);
4702 			LIST_REMOVE(acq, addrhash);
4703 			LIST_REMOVE(acq, seqhash);
4704 			free(acq, M_IPSEC_SAQ);
4705 		}
4706 		acq = nextacq;
4707 	}
4708 	ACQ_UNLOCK();
4709 }
4710 
4711 static void
4712 key_flush_spacq(time_t now)
4713 {
4714 	struct secspacq *acq, *nextacq;
4715 
4716 	/* SP ACQ tree */
4717 	SPACQ_LOCK();
4718 	for (acq = LIST_FIRST(&V_spacqtree); acq != NULL; acq = nextacq) {
4719 		nextacq = LIST_NEXT(acq, chain);
4720 		if (now - acq->created > V_key_blockacq_lifetime
4721 		 && __LIST_CHAINED(acq)) {
4722 			LIST_REMOVE(acq, chain);
4723 			free(acq, M_IPSEC_SAQ);
4724 		}
4725 	}
4726 	SPACQ_UNLOCK();
4727 }
4728 
4729 /*
4730  * time handler.
4731  * scanning SPD and SAD to check status for each entries,
4732  * and do to remove or to expire.
4733  * XXX: year 2038 problem may remain.
4734  */
4735 static void
4736 key_timehandler(void *arg)
4737 {
4738 	VNET_ITERATOR_DECL(vnet_iter);
4739 	time_t now = time_second;
4740 
4741 	VNET_LIST_RLOCK_NOSLEEP();
4742 	VNET_FOREACH(vnet_iter) {
4743 		CURVNET_SET(vnet_iter);
4744 		key_flush_spd(now);
4745 		key_flush_sad(now);
4746 		key_flush_acq(now);
4747 		key_flush_spacq(now);
4748 		CURVNET_RESTORE();
4749 	}
4750 	VNET_LIST_RUNLOCK_NOSLEEP();
4751 
4752 #ifndef IPSEC_DEBUG2
4753 	/* do exchange to tick time !! */
4754 	callout_schedule(&key_timer, hz);
4755 #endif /* IPSEC_DEBUG2 */
4756 }
4757 
4758 u_long
4759 key_random()
4760 {
4761 	u_long value;
4762 
4763 	key_randomfill(&value, sizeof(value));
4764 	return value;
4765 }
4766 
4767 void
4768 key_randomfill(void *p, size_t l)
4769 {
4770 	size_t n;
4771 	u_long v;
4772 	static int warn = 1;
4773 
4774 	n = 0;
4775 	n = (size_t)read_random(p, (u_int)l);
4776 	/* last resort */
4777 	while (n < l) {
4778 		v = random();
4779 		bcopy(&v, (u_int8_t *)p + n,
4780 		    l - n < sizeof(v) ? l - n : sizeof(v));
4781 		n += sizeof(v);
4782 
4783 		if (warn) {
4784 			printf("WARNING: pseudo-random number generator "
4785 			    "used for IPsec processing\n");
4786 			warn = 0;
4787 		}
4788 	}
4789 }
4790 
4791 /*
4792  * map SADB_SATYPE_* to IPPROTO_*.
4793  * if satype == SADB_SATYPE then satype is mapped to ~0.
4794  * OUT:
4795  *	0: invalid satype.
4796  */
4797 static uint8_t
4798 key_satype2proto(uint8_t satype)
4799 {
4800 	switch (satype) {
4801 	case SADB_SATYPE_UNSPEC:
4802 		return IPSEC_PROTO_ANY;
4803 	case SADB_SATYPE_AH:
4804 		return IPPROTO_AH;
4805 	case SADB_SATYPE_ESP:
4806 		return IPPROTO_ESP;
4807 	case SADB_X_SATYPE_IPCOMP:
4808 		return IPPROTO_IPCOMP;
4809 	case SADB_X_SATYPE_TCPSIGNATURE:
4810 		return IPPROTO_TCP;
4811 	default:
4812 		return 0;
4813 	}
4814 	/* NOTREACHED */
4815 }
4816 
4817 /*
4818  * map IPPROTO_* to SADB_SATYPE_*
4819  * OUT:
4820  *	0: invalid protocol type.
4821  */
4822 static uint8_t
4823 key_proto2satype(uint8_t proto)
4824 {
4825 	switch (proto) {
4826 	case IPPROTO_AH:
4827 		return SADB_SATYPE_AH;
4828 	case IPPROTO_ESP:
4829 		return SADB_SATYPE_ESP;
4830 	case IPPROTO_IPCOMP:
4831 		return SADB_X_SATYPE_IPCOMP;
4832 	case IPPROTO_TCP:
4833 		return SADB_X_SATYPE_TCPSIGNATURE;
4834 	default:
4835 		return 0;
4836 	}
4837 	/* NOTREACHED */
4838 }
4839 
4840 /* %%% PF_KEY */
4841 /*
4842  * SADB_GETSPI processing is to receive
4843  *	<base, (SA2), src address, dst address, (SPI range)>
4844  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
4845  * tree with the status of LARVAL, and send
4846  *	<base, SA(*), address(SD)>
4847  * to the IKMPd.
4848  *
4849  * IN:	mhp: pointer to the pointer to each header.
4850  * OUT:	NULL if fail.
4851  *	other if success, return pointer to the message to send.
4852  */
4853 static int
4854 key_getspi(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
4855 {
4856 	struct secasindex saidx;
4857 	struct sadb_address *src0, *dst0;
4858 	struct secasvar *sav;
4859 	uint32_t reqid, spi;
4860 	int error;
4861 	uint8_t mode, proto;
4862 
4863 	IPSEC_ASSERT(so != NULL, ("null socket"));
4864 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
4865 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
4866 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
4867 
4868 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
4869 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)
4870 #ifdef PFKEY_STRICT_CHECKS
4871 	    || SADB_CHECKHDR(mhp, SADB_EXT_SPIRANGE)
4872 #endif
4873 	    ) {
4874 		ipseclog((LOG_DEBUG,
4875 		    "%s: invalid message: missing required header.\n",
4876 		    __func__));
4877 		error = EINVAL;
4878 		goto fail;
4879 	}
4880 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
4881 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)
4882 #ifdef PFKEY_STRICT_CHECKS
4883 	    || SADB_CHECKLEN(mhp, SADB_EXT_SPIRANGE)
4884 #endif
4885 	    ) {
4886 		ipseclog((LOG_DEBUG,
4887 		    "%s: invalid message: wrong header size.\n", __func__));
4888 		error = EINVAL;
4889 		goto fail;
4890 	}
4891 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
4892 		mode = IPSEC_MODE_ANY;
4893 		reqid = 0;
4894 	} else {
4895 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
4896 			ipseclog((LOG_DEBUG,
4897 			    "%s: invalid message: wrong header size.\n",
4898 			    __func__));
4899 			error = EINVAL;
4900 			goto fail;
4901 		}
4902 		mode = ((struct sadb_x_sa2 *)
4903 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
4904 		reqid = ((struct sadb_x_sa2 *)
4905 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
4906 	}
4907 
4908 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
4909 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
4910 
4911 	/* map satype to proto */
4912 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
4913 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
4914 			__func__));
4915 		error = EINVAL;
4916 		goto fail;
4917 	}
4918 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
4919 	    (struct sockaddr *)(dst0 + 1));
4920 	if (error != 0) {
4921 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
4922 		error = EINVAL;
4923 		goto fail;
4924 	}
4925 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
4926 
4927 	/* SPI allocation */
4928 	spi = key_do_getnewspi(
4929 	    (struct sadb_spirange *)mhp->ext[SADB_EXT_SPIRANGE], &saidx);
4930 	if (spi == 0) {
4931 		/*
4932 		 * Requested SPI or SPI range is not available or
4933 		 * already used.
4934 		 */
4935 		error = EEXIST;
4936 		goto fail;
4937 	}
4938 	sav = key_newsav(mhp, &saidx, spi, &error);
4939 	if (sav == NULL)
4940 		goto fail;
4941 
4942 	if (sav->seq != 0) {
4943 		/*
4944 		 * RFC2367:
4945 		 * If the SADB_GETSPI message is in response to a
4946 		 * kernel-generated SADB_ACQUIRE, the sadb_msg_seq
4947 		 * MUST be the same as the SADB_ACQUIRE message.
4948 		 *
4949 		 * XXXAE: However it doesn't definethe behaviour how to
4950 		 * check this and what to do if it doesn't match.
4951 		 * Also what we should do if it matches?
4952 		 *
4953 		 * We can compare saidx used in SADB_ACQUIRE with saidx
4954 		 * used in SADB_GETSPI, but this probably can break
4955 		 * existing software. For now just warn if it doesn't match.
4956 		 *
4957 		 * XXXAE: anyway it looks useless.
4958 		 */
4959 		key_acqdone(&saidx, sav->seq);
4960 	}
4961 	KEYDBG(KEY_STAMP,
4962 	    printf("%s: SA(%p)\n", __func__, sav));
4963 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
4964 
4965     {
4966 	struct mbuf *n, *nn;
4967 	struct sadb_sa *m_sa;
4968 	struct sadb_msg *newmsg;
4969 	int off, len;
4970 
4971 	/* create new sadb_msg to reply. */
4972 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
4973 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
4974 
4975 	MGETHDR(n, M_NOWAIT, MT_DATA);
4976 	if (len > MHLEN) {
4977 		if (!(MCLGET(n, M_NOWAIT))) {
4978 			m_freem(n);
4979 			n = NULL;
4980 		}
4981 	}
4982 	if (!n) {
4983 		error = ENOBUFS;
4984 		goto fail;
4985 	}
4986 
4987 	n->m_len = len;
4988 	n->m_next = NULL;
4989 	off = 0;
4990 
4991 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
4992 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
4993 
4994 	m_sa = (struct sadb_sa *)(mtod(n, caddr_t) + off);
4995 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
4996 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
4997 	m_sa->sadb_sa_spi = spi; /* SPI is already in network byte order */
4998 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
4999 
5000 	IPSEC_ASSERT(off == len,
5001 		("length inconsistency (off %u len %u)", off, len));
5002 
5003 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
5004 	    SADB_EXT_ADDRESS_DST);
5005 	if (!n->m_next) {
5006 		m_freem(n);
5007 		error = ENOBUFS;
5008 		goto fail;
5009 	}
5010 
5011 	if (n->m_len < sizeof(struct sadb_msg)) {
5012 		n = m_pullup(n, sizeof(struct sadb_msg));
5013 		if (n == NULL)
5014 			return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
5015 	}
5016 
5017 	n->m_pkthdr.len = 0;
5018 	for (nn = n; nn; nn = nn->m_next)
5019 		n->m_pkthdr.len += nn->m_len;
5020 
5021 	newmsg = mtod(n, struct sadb_msg *);
5022 	newmsg->sadb_msg_seq = sav->seq;
5023 	newmsg->sadb_msg_errno = 0;
5024 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
5025 
5026 	m_freem(m);
5027 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
5028     }
5029 
5030 fail:
5031 	return (key_senderror(so, m, error));
5032 }
5033 
5034 /*
5035  * allocating new SPI
5036  * called by key_getspi().
5037  * OUT:
5038  *	0:	failure.
5039  *	others: success, SPI in network byte order.
5040  */
5041 static uint32_t
5042 key_do_getnewspi(struct sadb_spirange *spirange, struct secasindex *saidx)
5043 {
5044 	uint32_t min, max, newspi, t;
5045 	int count = V_key_spi_trycnt;
5046 
5047 	/* set spi range to allocate */
5048 	if (spirange != NULL) {
5049 		min = spirange->sadb_spirange_min;
5050 		max = spirange->sadb_spirange_max;
5051 	} else {
5052 		min = V_key_spi_minval;
5053 		max = V_key_spi_maxval;
5054 	}
5055 	/* IPCOMP needs 2-byte SPI */
5056 	if (saidx->proto == IPPROTO_IPCOMP) {
5057 		if (min >= 0x10000)
5058 			min = 0xffff;
5059 		if (max >= 0x10000)
5060 			max = 0xffff;
5061 		if (min > max) {
5062 			t = min; min = max; max = t;
5063 		}
5064 	}
5065 
5066 	if (min == max) {
5067 		if (!key_checkspidup(htonl(min))) {
5068 			ipseclog((LOG_DEBUG, "%s: SPI %u exists already.\n",
5069 			    __func__, min));
5070 			return 0;
5071 		}
5072 
5073 		count--; /* taking one cost. */
5074 		newspi = min;
5075 	} else {
5076 
5077 		/* init SPI */
5078 		newspi = 0;
5079 
5080 		/* when requesting to allocate spi ranged */
5081 		while (count--) {
5082 			/* generate pseudo-random SPI value ranged. */
5083 			newspi = min + (key_random() % (max - min + 1));
5084 			if (!key_checkspidup(htonl(newspi)))
5085 				break;
5086 		}
5087 
5088 		if (count == 0 || newspi == 0) {
5089 			ipseclog((LOG_DEBUG,
5090 			    "%s: failed to allocate SPI.\n", __func__));
5091 			return 0;
5092 		}
5093 	}
5094 
5095 	/* statistics */
5096 	keystat.getspi_count =
5097 	    (keystat.getspi_count + V_key_spi_trycnt - count) / 2;
5098 
5099 	return (htonl(newspi));
5100 }
5101 
5102 /*
5103  * Find TCP-MD5 SA with corresponding secasindex.
5104  * If not found, return NULL and fill SPI with usable value if needed.
5105  */
5106 static struct secasvar *
5107 key_getsav_tcpmd5(struct secasindex *saidx, uint32_t *spi)
5108 {
5109 	SAHTREE_RLOCK_TRACKER;
5110 	struct secashead *sah;
5111 	struct secasvar *sav;
5112 
5113 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP, ("wrong proto"));
5114 	SAHTREE_RLOCK();
5115 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
5116 		if (sah->saidx.proto != IPPROTO_TCP)
5117 			continue;
5118 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
5119 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
5120 			break;
5121 	}
5122 	if (sah != NULL) {
5123 		if (V_key_preferred_oldsa)
5124 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
5125 		else
5126 			sav = TAILQ_FIRST(&sah->savtree_alive);
5127 		if (sav != NULL) {
5128 			SAV_ADDREF(sav);
5129 			SAHTREE_RUNLOCK();
5130 			return (sav);
5131 		}
5132 	}
5133 	if (spi == NULL) {
5134 		/* No SPI required */
5135 		SAHTREE_RUNLOCK();
5136 		return (NULL);
5137 	}
5138 	/* Check that SPI is unique */
5139 	LIST_FOREACH(sav, SAVHASH_HASH(*spi), spihash) {
5140 		if (sav->spi == *spi)
5141 			break;
5142 	}
5143 	if (sav == NULL) {
5144 		SAHTREE_RUNLOCK();
5145 		/* SPI is already unique */
5146 		return (NULL);
5147 	}
5148 	SAHTREE_RUNLOCK();
5149 	/* XXX: not optimal */
5150 	*spi = key_do_getnewspi(NULL, saidx);
5151 	return (NULL);
5152 }
5153 
5154 static int
5155 key_updateaddresses(struct socket *so, struct mbuf *m,
5156     const struct sadb_msghdr *mhp, struct secasvar *sav,
5157     struct secasindex *saidx)
5158 {
5159 	struct sockaddr *newaddr;
5160 	struct secashead *sah;
5161 	struct secasvar *newsav, *tmp;
5162 	struct mbuf *n;
5163 	int error, isnew;
5164 
5165 	/* Check that we need to change SAH */
5166 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC)) {
5167 		newaddr = (struct sockaddr *)(
5168 		    ((struct sadb_address *)
5169 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_SRC]) + 1);
5170 		bcopy(newaddr, &saidx->src, newaddr->sa_len);
5171 		key_porttosaddr(&saidx->src.sa, 0);
5172 	}
5173 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5174 		newaddr = (struct sockaddr *)(
5175 		    ((struct sadb_address *)
5176 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_DST]) + 1);
5177 		bcopy(newaddr, &saidx->dst, newaddr->sa_len);
5178 		key_porttosaddr(&saidx->dst.sa, 0);
5179 	}
5180 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5181 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5182 		error = key_checksockaddrs(&saidx->src.sa, &saidx->dst.sa);
5183 		if (error != 0) {
5184 			ipseclog((LOG_DEBUG, "%s: invalid new sockaddr.\n",
5185 			    __func__));
5186 			return (error);
5187 		}
5188 
5189 		sah = key_getsah(saidx);
5190 		if (sah == NULL) {
5191 			/* create a new SA index */
5192 			sah = key_newsah(saidx);
5193 			if (sah == NULL) {
5194 				ipseclog((LOG_DEBUG,
5195 				    "%s: No more memory.\n", __func__));
5196 				return (ENOBUFS);
5197 			}
5198 			isnew = 2; /* SAH is new */
5199 		} else
5200 			isnew = 1; /* existing SAH is referenced */
5201 	} else {
5202 		/*
5203 		 * src and dst addresses are still the same.
5204 		 * Do we want to change NAT-T config?
5205 		 */
5206 		if (sav->sah->saidx.proto != IPPROTO_ESP ||
5207 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5208 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5209 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5210 			ipseclog((LOG_DEBUG,
5211 			    "%s: invalid message: missing required header.\n",
5212 			    __func__));
5213 			return (EINVAL);
5214 		}
5215 		/* We hold reference to SA, thus SAH will be referenced too. */
5216 		sah = sav->sah;
5217 		isnew = 0;
5218 	}
5219 
5220 	newsav = malloc(sizeof(struct secasvar), M_IPSEC_SA,
5221 	    M_NOWAIT | M_ZERO);
5222 	if (newsav == NULL) {
5223 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5224 		error = ENOBUFS;
5225 		goto fail;
5226 	}
5227 
5228 	/* Clone SA's content into newsav */
5229 	SAV_INITREF(newsav);
5230 	bcopy(sav, newsav, offsetof(struct secasvar, chain));
5231 	/*
5232 	 * We create new NAT-T config if it is needed.
5233 	 * Old NAT-T config will be freed by key_cleansav() when
5234 	 * last reference to SA will be released.
5235 	 */
5236 	newsav->natt = NULL;
5237 	newsav->sah = sah;
5238 	newsav->state = SADB_SASTATE_MATURE;
5239 	error = key_setnatt(newsav, mhp);
5240 	if (error != 0)
5241 		goto fail;
5242 
5243 	SAHTREE_WLOCK();
5244 	/* Check that SA is still alive */
5245 	if (sav->state == SADB_SASTATE_DEAD) {
5246 		/* SA was unlinked */
5247 		SAHTREE_WUNLOCK();
5248 		error = ESRCH;
5249 		goto fail;
5250 	}
5251 
5252 	/* Unlink SA from SAH and SPI hash */
5253 	IPSEC_ASSERT((sav->flags & SADB_X_EXT_F_CLONED) == 0,
5254 	    ("SA is already cloned"));
5255 	IPSEC_ASSERT(sav->state == SADB_SASTATE_MATURE ||
5256 	    sav->state == SADB_SASTATE_DYING,
5257 	    ("Wrong SA state %u\n", sav->state));
5258 	TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
5259 	LIST_REMOVE(sav, spihash);
5260 	sav->state = SADB_SASTATE_DEAD;
5261 
5262 	/*
5263 	 * Link new SA with SAH. Keep SAs ordered by
5264 	 * create time (newer are first).
5265 	 */
5266 	TAILQ_FOREACH(tmp, &sah->savtree_alive, chain) {
5267 		if (newsav->created > tmp->created) {
5268 			TAILQ_INSERT_BEFORE(tmp, newsav, chain);
5269 			break;
5270 		}
5271 	}
5272 	if (tmp == NULL)
5273 		TAILQ_INSERT_TAIL(&sah->savtree_alive, newsav, chain);
5274 
5275 	/* Add new SA into SPI hash. */
5276 	LIST_INSERT_HEAD(SAVHASH_HASH(newsav->spi), newsav, spihash);
5277 
5278 	/* Add new SAH into SADB. */
5279 	if (isnew == 2) {
5280 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
5281 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
5282 		sah->state = SADB_SASTATE_MATURE;
5283 		SAH_ADDREF(sah); /* newsav references new SAH */
5284 	}
5285 	/*
5286 	 * isnew == 1 -> @sah was referenced by key_getsah().
5287 	 * isnew == 0 -> we use the same @sah, that was used by @sav,
5288 	 *	and we use its reference for @newsav.
5289 	 */
5290 	SECASVAR_LOCK(sav);
5291 	/* XXX: replace cntr with pointer? */
5292 	newsav->cntr = sav->cntr;
5293 	sav->flags |= SADB_X_EXT_F_CLONED;
5294 	SECASVAR_UNLOCK(sav);
5295 
5296 	SAHTREE_WUNLOCK();
5297 
5298 	KEYDBG(KEY_STAMP,
5299 	    printf("%s: SA(%p) cloned into SA(%p)\n",
5300 	    __func__, sav, newsav));
5301 	KEYDBG(KEY_DATA, kdebug_secasv(newsav));
5302 
5303 	key_freesav(&sav); /* release last reference */
5304 
5305 	/* set msg buf from mhp */
5306 	n = key_getmsgbuf_x1(m, mhp);
5307 	if (n == NULL) {
5308 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5309 		return (ENOBUFS);
5310 	}
5311 	m_freem(m);
5312 	key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5313 	return (0);
5314 fail:
5315 	if (isnew != 0)
5316 		key_freesah(&sah);
5317 	if (newsav != NULL) {
5318 		if (newsav->natt != NULL)
5319 			free(newsav->natt, M_IPSEC_MISC);
5320 		free(newsav, M_IPSEC_SA);
5321 	}
5322 	return (error);
5323 }
5324 
5325 /*
5326  * SADB_UPDATE processing
5327  * receive
5328  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5329  *       key(AE), (identity(SD),) (sensitivity)>
5330  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5331  * and send
5332  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5333  *       (identity(SD),) (sensitivity)>
5334  * to the ikmpd.
5335  *
5336  * m will always be freed.
5337  */
5338 static int
5339 key_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5340 {
5341 	struct secasindex saidx;
5342 	struct sadb_address *src0, *dst0;
5343 	struct sadb_sa *sa0;
5344 	struct secasvar *sav;
5345 	uint32_t reqid;
5346 	int error;
5347 	uint8_t mode, proto;
5348 
5349 	IPSEC_ASSERT(so != NULL, ("null socket"));
5350 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5351 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5352 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5353 
5354 	/* map satype to proto */
5355 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5356 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5357 		    __func__));
5358 		return key_senderror(so, m, EINVAL);
5359 	}
5360 
5361 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5362 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5363 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5364 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5365 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5366 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5367 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5368 		ipseclog((LOG_DEBUG,
5369 		    "%s: invalid message: missing required header.\n",
5370 		    __func__));
5371 		return key_senderror(so, m, EINVAL);
5372 	}
5373 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5374 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5375 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5376 		ipseclog((LOG_DEBUG,
5377 		    "%s: invalid message: wrong header size.\n", __func__));
5378 		return key_senderror(so, m, EINVAL);
5379 	}
5380 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5381 		mode = IPSEC_MODE_ANY;
5382 		reqid = 0;
5383 	} else {
5384 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5385 			ipseclog((LOG_DEBUG,
5386 			    "%s: invalid message: wrong header size.\n",
5387 			    __func__));
5388 			return key_senderror(so, m, EINVAL);
5389 		}
5390 		mode = ((struct sadb_x_sa2 *)
5391 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5392 		reqid = ((struct sadb_x_sa2 *)
5393 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5394 	}
5395 
5396 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5397 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5398 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5399 
5400 	/*
5401 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5402 	 * SADB_UPDATE message.
5403 	 */
5404 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5405 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5406 #ifdef PFKEY_STRICT_CHECKS
5407 		return key_senderror(so, m, EINVAL);
5408 #endif
5409 	}
5410 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5411 	    (struct sockaddr *)(dst0 + 1));
5412 	if (error != 0) {
5413 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5414 		return key_senderror(so, m, error);
5415 	}
5416 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5417 	sav = key_getsavbyspi(sa0->sadb_sa_spi);
5418 	if (sav == NULL) {
5419 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u\n",
5420 		    __func__, ntohl(sa0->sadb_sa_spi)));
5421 		return key_senderror(so, m, EINVAL);
5422 	}
5423 	/*
5424 	 * Check that SADB_UPDATE issued by the same process that did
5425 	 * SADB_GETSPI or SADB_ADD.
5426 	 */
5427 	if (sav->pid != mhp->msg->sadb_msg_pid) {
5428 		ipseclog((LOG_DEBUG,
5429 		    "%s: pid mismatched (SPI %u, pid %u vs. %u)\n", __func__,
5430 		    ntohl(sav->spi), sav->pid, mhp->msg->sadb_msg_pid));
5431 		key_freesav(&sav);
5432 		return key_senderror(so, m, EINVAL);
5433 	}
5434 	/* saidx should match with SA. */
5435 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_MODE_REQID) == 0) {
5436 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u",
5437 		    __func__, ntohl(sav->spi)));
5438 		key_freesav(&sav);
5439 		return key_senderror(so, m, ESRCH);
5440 	}
5441 
5442 	if (sav->state == SADB_SASTATE_LARVAL) {
5443 		if ((mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5444 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) ||
5445 		    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5446 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH))) {
5447 			ipseclog((LOG_DEBUG,
5448 			    "%s: invalid message: missing required header.\n",
5449 			    __func__));
5450 			key_freesav(&sav);
5451 			return key_senderror(so, m, EINVAL);
5452 		}
5453 		/*
5454 		 * We can set any values except src, dst and SPI.
5455 		 */
5456 		error = key_setsaval(sav, mhp);
5457 		if (error != 0) {
5458 			key_freesav(&sav);
5459 			return (key_senderror(so, m, error));
5460 		}
5461 		/* Change SA state to MATURE */
5462 		SAHTREE_WLOCK();
5463 		if (sav->state != SADB_SASTATE_LARVAL) {
5464 			/* SA was deleted or another thread made it MATURE. */
5465 			SAHTREE_WUNLOCK();
5466 			key_freesav(&sav);
5467 			return (key_senderror(so, m, ESRCH));
5468 		}
5469 		/*
5470 		 * NOTE: we keep SAs in savtree_alive ordered by created
5471 		 * time. When SA's state changed from LARVAL to MATURE,
5472 		 * we update its created time in key_setsaval() and move
5473 		 * it into head of savtree_alive.
5474 		 */
5475 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
5476 		TAILQ_INSERT_HEAD(&sav->sah->savtree_alive, sav, chain);
5477 		sav->state = SADB_SASTATE_MATURE;
5478 		SAHTREE_WUNLOCK();
5479 	} else {
5480 		/*
5481 		 * For DYING and MATURE SA we can change only state
5482 		 * and lifetimes. Report EINVAL if something else attempted
5483 		 * to change.
5484 		 */
5485 		if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5486 		    !SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
5487 			key_freesav(&sav);
5488 			return (key_senderror(so, m, EINVAL));
5489 		}
5490 		error = key_updatelifetimes(sav, mhp);
5491 		if (error != 0) {
5492 			key_freesav(&sav);
5493 			return (key_senderror(so, m, error));
5494 		}
5495 		/*
5496 		 * This is FreeBSD extension to RFC2367.
5497 		 * IKEd can specify SADB_X_EXT_NEW_ADDRESS_SRC and/or
5498 		 * SADB_X_EXT_NEW_ADDRESS_DST when it wants to change
5499 		 * SA addresses (for example to implement MOBIKE protocol
5500 		 * as described in RFC4555). Also we allow to change
5501 		 * NAT-T config.
5502 		 */
5503 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5504 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST) ||
5505 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5506 		    sav->natt != NULL) {
5507 			error = key_updateaddresses(so, m, mhp, sav, &saidx);
5508 			key_freesav(&sav);
5509 			if (error != 0)
5510 				return (key_senderror(so, m, error));
5511 			return (0);
5512 		}
5513 		/* Check that SA is still alive */
5514 		SAHTREE_WLOCK();
5515 		if (sav->state == SADB_SASTATE_DEAD) {
5516 			/* SA was unlinked */
5517 			SAHTREE_WUNLOCK();
5518 			key_freesav(&sav);
5519 			return (key_senderror(so, m, ESRCH));
5520 		}
5521 		/*
5522 		 * NOTE: there is possible state moving from DYING to MATURE,
5523 		 * but this doesn't change created time, so we won't reorder
5524 		 * this SA.
5525 		 */
5526 		sav->state = SADB_SASTATE_MATURE;
5527 		SAHTREE_WUNLOCK();
5528 	}
5529 	KEYDBG(KEY_STAMP,
5530 	    printf("%s: SA(%p)\n", __func__, sav));
5531 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5532 	key_freesav(&sav);
5533 
5534     {
5535 	struct mbuf *n;
5536 
5537 	/* set msg buf from mhp */
5538 	n = key_getmsgbuf_x1(m, mhp);
5539 	if (n == NULL) {
5540 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5541 		return key_senderror(so, m, ENOBUFS);
5542 	}
5543 
5544 	m_freem(m);
5545 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5546     }
5547 }
5548 
5549 /*
5550  * SADB_ADD processing
5551  * add an entry to SA database, when received
5552  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5553  *       key(AE), (identity(SD),) (sensitivity)>
5554  * from the ikmpd,
5555  * and send
5556  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5557  *       (identity(SD),) (sensitivity)>
5558  * to the ikmpd.
5559  *
5560  * IGNORE identity and sensitivity messages.
5561  *
5562  * m will always be freed.
5563  */
5564 static int
5565 key_add(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5566 {
5567 	struct secasindex saidx;
5568 	struct sadb_address *src0, *dst0;
5569 	struct sadb_sa *sa0;
5570 	struct secasvar *sav;
5571 	uint32_t reqid, spi;
5572 	uint8_t mode, proto;
5573 	int error;
5574 
5575 	IPSEC_ASSERT(so != NULL, ("null socket"));
5576 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5577 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5578 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5579 
5580 	/* map satype to proto */
5581 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5582 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5583 		    __func__));
5584 		return key_senderror(so, m, EINVAL);
5585 	}
5586 
5587 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5588 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5589 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5590 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP && (
5591 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5592 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT))) ||
5593 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH && (
5594 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH) ||
5595 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH))) ||
5596 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5597 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5598 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5599 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5600 		ipseclog((LOG_DEBUG,
5601 		    "%s: invalid message: missing required header.\n",
5602 		    __func__));
5603 		return key_senderror(so, m, EINVAL);
5604 	}
5605 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5606 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5607 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5608 		ipseclog((LOG_DEBUG,
5609 		    "%s: invalid message: wrong header size.\n", __func__));
5610 		return key_senderror(so, m, EINVAL);
5611 	}
5612 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5613 		mode = IPSEC_MODE_ANY;
5614 		reqid = 0;
5615 	} else {
5616 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5617 			ipseclog((LOG_DEBUG,
5618 			    "%s: invalid message: wrong header size.\n",
5619 			    __func__));
5620 			return key_senderror(so, m, EINVAL);
5621 		}
5622 		mode = ((struct sadb_x_sa2 *)
5623 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5624 		reqid = ((struct sadb_x_sa2 *)
5625 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5626 	}
5627 
5628 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5629 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
5630 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
5631 
5632 	/*
5633 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5634 	 * SADB_ADD message.
5635 	 */
5636 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5637 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5638 #ifdef PFKEY_STRICT_CHECKS
5639 		return key_senderror(so, m, EINVAL);
5640 #endif
5641 	}
5642 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5643 	    (struct sockaddr *)(dst0 + 1));
5644 	if (error != 0) {
5645 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5646 		return key_senderror(so, m, error);
5647 	}
5648 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5649 	spi = sa0->sadb_sa_spi;
5650 	/*
5651 	 * For TCP-MD5 SAs we don't use SPI. Check the uniqueness using
5652 	 * secasindex.
5653 	 * XXXAE: IPComp seems also doesn't use SPI.
5654 	 */
5655 	if (proto == IPPROTO_TCP) {
5656 		sav = key_getsav_tcpmd5(&saidx, &spi);
5657 		if (sav == NULL && spi == 0) {
5658 			/* Failed to allocate SPI */
5659 			ipseclog((LOG_DEBUG, "%s: SA already exists.\n",
5660 			    __func__));
5661 			return key_senderror(so, m, EEXIST);
5662 		}
5663 		/* XXX: SPI that we report back can have another value */
5664 	} else {
5665 		/* We can create new SA only if SPI is different. */
5666 		sav = key_getsavbyspi(spi);
5667 	}
5668 	if (sav != NULL) {
5669 		key_freesav(&sav);
5670 		ipseclog((LOG_DEBUG, "%s: SA already exists.\n", __func__));
5671 		return key_senderror(so, m, EEXIST);
5672 	}
5673 
5674 	sav = key_newsav(mhp, &saidx, spi, &error);
5675 	if (sav == NULL)
5676 		return key_senderror(so, m, error);
5677 	KEYDBG(KEY_STAMP,
5678 	    printf("%s: return SA(%p)\n", __func__, sav));
5679 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5680 	/*
5681 	 * If SADB_ADD was in response to SADB_ACQUIRE, we need to schedule
5682 	 * ACQ for deletion.
5683 	 */
5684 	if (sav->seq != 0)
5685 		key_acqdone(&saidx, sav->seq);
5686 
5687     {
5688 	/*
5689 	 * Don't call key_freesav() on error here, as we would like to
5690 	 * keep the SA in the database.
5691 	 */
5692 	struct mbuf *n;
5693 
5694 	/* set msg buf from mhp */
5695 	n = key_getmsgbuf_x1(m, mhp);
5696 	if (n == NULL) {
5697 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5698 		return key_senderror(so, m, ENOBUFS);
5699 	}
5700 
5701 	m_freem(m);
5702 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5703     }
5704 }
5705 
5706 /*
5707  * NAT-T support.
5708  * IKEd may request the use ESP in UDP encapsulation when it detects the
5709  * presence of NAT. It uses NAT-T extension headers for such SAs to specify
5710  * parameters needed for encapsulation and decapsulation. These PF_KEY
5711  * extension headers are not standardized, so this comment addresses our
5712  * implementation.
5713  * SADB_X_EXT_NAT_T_TYPE specifies type of encapsulation, we support only
5714  * UDP_ENCAP_ESPINUDP as described in RFC3948.
5715  * SADB_X_EXT_NAT_T_SPORT/DPORT specifies source and destination ports for
5716  * UDP header. We use these ports in UDP encapsulation procedure, also we
5717  * can check them in UDP decapsulation procedure.
5718  * SADB_X_EXT_NAT_T_OA[IR] specifies original address of initiator or
5719  * responder. These addresses can be used for transport mode to adjust
5720  * checksum after decapsulation and decryption. Since original IP addresses
5721  * used by peer usually different (we detected presence of NAT), TCP/UDP
5722  * pseudo header checksum and IP header checksum was calculated using original
5723  * addresses. After decapsulation and decryption we need to adjust checksum
5724  * to have correct datagram.
5725  *
5726  * We expect presence of NAT-T extension headers only in SADB_ADD and
5727  * SADB_UPDATE messages. We report NAT-T extension headers in replies
5728  * to SADB_ADD, SADB_UPDATE, SADB_GET, and SADB_DUMP messages.
5729  */
5730 static int
5731 key_setnatt(struct secasvar *sav, const struct sadb_msghdr *mhp)
5732 {
5733 	struct sadb_x_nat_t_port *port;
5734 	struct sadb_x_nat_t_type *type;
5735 	struct sadb_address *oai, *oar;
5736 	struct sockaddr *sa;
5737 	uint32_t addr;
5738 	uint16_t cksum;
5739 
5740 	IPSEC_ASSERT(sav->natt == NULL, ("natt is already initialized"));
5741 	/*
5742 	 * Ignore NAT-T headers if sproto isn't ESP.
5743 	 */
5744 	if (sav->sah->saidx.proto != IPPROTO_ESP)
5745 		return (0);
5746 
5747 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) &&
5748 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) &&
5749 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5750 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5751 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5752 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5753 			ipseclog((LOG_DEBUG,
5754 			    "%s: invalid message: wrong header size.\n",
5755 			    __func__));
5756 			return (EINVAL);
5757 		}
5758 	} else
5759 		return (0);
5760 
5761 	type = (struct sadb_x_nat_t_type *)mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5762 	if (type->sadb_x_nat_t_type_type != UDP_ENCAP_ESPINUDP) {
5763 		ipseclog((LOG_DEBUG, "%s: unsupported NAT-T type %u.\n",
5764 		    __func__, type->sadb_x_nat_t_type_type));
5765 		return (EINVAL);
5766 	}
5767 	/*
5768 	 * Allocate storage for NAT-T config.
5769 	 * On error it will be released by key_cleansav().
5770 	 */
5771 	sav->natt = malloc(sizeof(struct secnatt), M_IPSEC_MISC,
5772 	    M_NOWAIT | M_ZERO);
5773 	if (sav->natt == NULL) {
5774 		PFKEYSTAT_INC(in_nomem);
5775 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5776 		return (ENOBUFS);
5777 	}
5778 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5779 	if (port->sadb_x_nat_t_port_port == 0) {
5780 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T sport specified.\n",
5781 		    __func__));
5782 		return (EINVAL);
5783 	}
5784 	sav->natt->sport = port->sadb_x_nat_t_port_port;
5785 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5786 	if (port->sadb_x_nat_t_port_port == 0) {
5787 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T dport specified.\n",
5788 		    __func__));
5789 		return (EINVAL);
5790 	}
5791 	sav->natt->dport = port->sadb_x_nat_t_port_port;
5792 
5793 	/*
5794 	 * SADB_X_EXT_NAT_T_OAI and SADB_X_EXT_NAT_T_OAR are optional
5795 	 * and needed only for transport mode IPsec.
5796 	 * Usually NAT translates only one address, but it is possible,
5797 	 * that both addresses could be translated.
5798 	 * NOTE: Value of SADB_X_EXT_NAT_T_OAI is equal to SADB_X_EXT_NAT_T_OA.
5799 	 */
5800 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAI)) {
5801 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAI)) {
5802 			ipseclog((LOG_DEBUG,
5803 			    "%s: invalid message: wrong header size.\n",
5804 			    __func__));
5805 			return (EINVAL);
5806 		}
5807 		oai = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAI];
5808 	} else
5809 		oai = NULL;
5810 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAR)) {
5811 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAR)) {
5812 			ipseclog((LOG_DEBUG,
5813 			    "%s: invalid message: wrong header size.\n",
5814 			    __func__));
5815 			return (EINVAL);
5816 		}
5817 		oar = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAR];
5818 	} else
5819 		oar = NULL;
5820 
5821 	/* Initialize addresses only for transport mode */
5822 	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL) {
5823 		cksum = 0;
5824 		if (oai != NULL) {
5825 			/* Currently we support only AF_INET */
5826 			sa = (struct sockaddr *)(oai + 1);
5827 			if (sa->sa_family != AF_INET ||
5828 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5829 				ipseclog((LOG_DEBUG,
5830 				    "%s: wrong NAT-OAi header.\n",
5831 				    __func__));
5832 				return (EINVAL);
5833 			}
5834 			/* Ignore address if it the same */
5835 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5836 			    sav->sah->saidx.src.sin.sin_addr.s_addr) {
5837 				bcopy(sa, &sav->natt->oai.sa, sa->sa_len);
5838 				sav->natt->flags |= IPSEC_NATT_F_OAI;
5839 				/* Calculate checksum delta */
5840 				addr = sav->sah->saidx.src.sin.sin_addr.s_addr;
5841 				cksum = in_addword(cksum, ~addr >> 16);
5842 				cksum = in_addword(cksum, ~addr & 0xffff);
5843 				addr = sav->natt->oai.sin.sin_addr.s_addr;
5844 				cksum = in_addword(cksum, addr >> 16);
5845 				cksum = in_addword(cksum, addr & 0xffff);
5846 			}
5847 		}
5848 		if (oar != NULL) {
5849 			/* Currently we support only AF_INET */
5850 			sa = (struct sockaddr *)(oar + 1);
5851 			if (sa->sa_family != AF_INET ||
5852 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5853 				ipseclog((LOG_DEBUG,
5854 				    "%s: wrong NAT-OAr header.\n",
5855 				    __func__));
5856 				return (EINVAL);
5857 			}
5858 			/* Ignore address if it the same */
5859 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5860 			    sav->sah->saidx.dst.sin.sin_addr.s_addr) {
5861 				bcopy(sa, &sav->natt->oar.sa, sa->sa_len);
5862 				sav->natt->flags |= IPSEC_NATT_F_OAR;
5863 				/* Calculate checksum delta */
5864 				addr = sav->sah->saidx.dst.sin.sin_addr.s_addr;
5865 				cksum = in_addword(cksum, ~addr >> 16);
5866 				cksum = in_addword(cksum, ~addr & 0xffff);
5867 				addr = sav->natt->oar.sin.sin_addr.s_addr;
5868 				cksum = in_addword(cksum, addr >> 16);
5869 				cksum = in_addword(cksum, addr & 0xffff);
5870 			}
5871 		}
5872 		sav->natt->cksum = cksum;
5873 	}
5874 	return (0);
5875 }
5876 
5877 static int
5878 key_setident(struct secashead *sah, const struct sadb_msghdr *mhp)
5879 {
5880 	const struct sadb_ident *idsrc, *iddst;
5881 
5882 	IPSEC_ASSERT(sah != NULL, ("null secashead"));
5883 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5884 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5885 
5886 	/* don't make buffer if not there */
5887 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) &&
5888 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5889 		sah->idents = NULL;
5890 		sah->identd = NULL;
5891 		return (0);
5892 	}
5893 
5894 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) ||
5895 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5896 		ipseclog((LOG_DEBUG, "%s: invalid identity.\n", __func__));
5897 		return (EINVAL);
5898 	}
5899 
5900 	idsrc = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_SRC];
5901 	iddst = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_DST];
5902 
5903 	/* validity check */
5904 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
5905 		ipseclog((LOG_DEBUG, "%s: ident type mismatch.\n", __func__));
5906 		return EINVAL;
5907 	}
5908 
5909 	switch (idsrc->sadb_ident_type) {
5910 	case SADB_IDENTTYPE_PREFIX:
5911 	case SADB_IDENTTYPE_FQDN:
5912 	case SADB_IDENTTYPE_USERFQDN:
5913 	default:
5914 		/* XXX do nothing */
5915 		sah->idents = NULL;
5916 		sah->identd = NULL;
5917 	 	return 0;
5918 	}
5919 
5920 	/* make structure */
5921 	sah->idents = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5922 	if (sah->idents == NULL) {
5923 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5924 		return ENOBUFS;
5925 	}
5926 	sah->identd = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5927 	if (sah->identd == NULL) {
5928 		free(sah->idents, M_IPSEC_MISC);
5929 		sah->idents = NULL;
5930 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5931 		return ENOBUFS;
5932 	}
5933 	sah->idents->type = idsrc->sadb_ident_type;
5934 	sah->idents->id = idsrc->sadb_ident_id;
5935 
5936 	sah->identd->type = iddst->sadb_ident_type;
5937 	sah->identd->id = iddst->sadb_ident_id;
5938 
5939 	return 0;
5940 }
5941 
5942 /*
5943  * m will not be freed on return.
5944  * it is caller's responsibility to free the result.
5945  *
5946  * Called from SADB_ADD and SADB_UPDATE. Reply will contain headers
5947  * from the request in defined order.
5948  */
5949 static struct mbuf *
5950 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
5951 {
5952 	struct mbuf *n;
5953 
5954 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5955 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5956 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5957 
5958 	/* create new sadb_msg to reply. */
5959 	n = key_gather_mbuf(m, mhp, 1, 16, SADB_EXT_RESERVED,
5960 	    SADB_EXT_SA, SADB_X_EXT_SA2,
5961 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
5962 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
5963 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
5964 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
5965 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
5966 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NEW_ADDRESS_SRC,
5967 	    SADB_X_EXT_NEW_ADDRESS_DST);
5968 	if (!n)
5969 		return NULL;
5970 
5971 	if (n->m_len < sizeof(struct sadb_msg)) {
5972 		n = m_pullup(n, sizeof(struct sadb_msg));
5973 		if (n == NULL)
5974 			return NULL;
5975 	}
5976 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
5977 	mtod(n, struct sadb_msg *)->sadb_msg_len =
5978 	    PFKEY_UNIT64(n->m_pkthdr.len);
5979 
5980 	return n;
5981 }
5982 
5983 /*
5984  * SADB_DELETE processing
5985  * receive
5986  *   <base, SA(*), address(SD)>
5987  * from the ikmpd, and set SADB_SASTATE_DEAD,
5988  * and send,
5989  *   <base, SA(*), address(SD)>
5990  * to the ikmpd.
5991  *
5992  * m will always be freed.
5993  */
5994 static int
5995 key_delete(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5996 {
5997 	struct secasindex saidx;
5998 	struct sadb_address *src0, *dst0;
5999 	struct secasvar *sav;
6000 	struct sadb_sa *sa0;
6001 	uint8_t proto;
6002 
6003 	IPSEC_ASSERT(so != NULL, ("null socket"));
6004 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6005 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6006 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6007 
6008 	/* map satype to proto */
6009 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6010 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6011 		    __func__));
6012 		return key_senderror(so, m, EINVAL);
6013 	}
6014 
6015 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6016 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
6017 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6018 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6019 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
6020 		    __func__));
6021 		return key_senderror(so, m, EINVAL);
6022 	}
6023 
6024 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
6025 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
6026 
6027 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6028 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6029 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6030 		return (key_senderror(so, m, EINVAL));
6031 	}
6032 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6033 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
6034 		/*
6035 		 * Caller wants us to delete all non-LARVAL SAs
6036 		 * that match the src/dst.  This is used during
6037 		 * IKE INITIAL-CONTACT.
6038 		 * XXXAE: this looks like some extension to RFC2367.
6039 		 */
6040 		ipseclog((LOG_DEBUG, "%s: doing delete all.\n", __func__));
6041 		return (key_delete_all(so, m, mhp, &saidx));
6042 	}
6043 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
6044 		ipseclog((LOG_DEBUG,
6045 		    "%s: invalid message: wrong header size.\n", __func__));
6046 		return (key_senderror(so, m, EINVAL));
6047 	}
6048 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6049 	if (proto == IPPROTO_TCP)
6050 		sav = key_getsav_tcpmd5(&saidx, NULL);
6051 	else
6052 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6053 	if (sav == NULL) {
6054 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u.\n",
6055 		    __func__, ntohl(sa0->sadb_sa_spi)));
6056 		return (key_senderror(so, m, ESRCH));
6057 	}
6058 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6059 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6060 		    __func__, ntohl(sav->spi)));
6061 		key_freesav(&sav);
6062 		return (key_senderror(so, m, ESRCH));
6063 	}
6064 	KEYDBG(KEY_STAMP,
6065 	    printf("%s: SA(%p)\n", __func__, sav));
6066 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
6067 	key_unlinksav(sav);
6068 	key_freesav(&sav);
6069 
6070     {
6071 	struct mbuf *n;
6072 	struct sadb_msg *newmsg;
6073 
6074 	/* create new sadb_msg to reply. */
6075 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
6076 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6077 	if (!n)
6078 		return key_senderror(so, m, ENOBUFS);
6079 
6080 	if (n->m_len < sizeof(struct sadb_msg)) {
6081 		n = m_pullup(n, sizeof(struct sadb_msg));
6082 		if (n == NULL)
6083 			return key_senderror(so, m, ENOBUFS);
6084 	}
6085 	newmsg = mtod(n, struct sadb_msg *);
6086 	newmsg->sadb_msg_errno = 0;
6087 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6088 
6089 	m_freem(m);
6090 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6091     }
6092 }
6093 
6094 /*
6095  * delete all SAs for src/dst.  Called from key_delete().
6096  */
6097 static int
6098 key_delete_all(struct socket *so, struct mbuf *m,
6099     const struct sadb_msghdr *mhp, struct secasindex *saidx)
6100 {
6101 	struct secasvar_queue drainq;
6102 	struct secashead *sah;
6103 	struct secasvar *sav, *nextsav;
6104 
6105 	TAILQ_INIT(&drainq);
6106 	SAHTREE_WLOCK();
6107 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
6108 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_HEAD) == 0)
6109 			continue;
6110 		/* Move all ALIVE SAs into drainq */
6111 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6112 	}
6113 	/* Unlink all queued SAs from SPI hash */
6114 	TAILQ_FOREACH(sav, &drainq, chain) {
6115 		sav->state = SADB_SASTATE_DEAD;
6116 		LIST_REMOVE(sav, spihash);
6117 	}
6118 	SAHTREE_WUNLOCK();
6119 	/* Now we can release reference for all SAs in drainq */
6120 	sav = TAILQ_FIRST(&drainq);
6121 	while (sav != NULL) {
6122 		KEYDBG(KEY_STAMP,
6123 		    printf("%s: SA(%p)\n", __func__, sav));
6124 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6125 		nextsav = TAILQ_NEXT(sav, chain);
6126 		key_freesah(&sav->sah); /* release reference from SAV */
6127 		key_freesav(&sav); /* release last reference */
6128 		sav = nextsav;
6129 	}
6130 
6131     {
6132 	struct mbuf *n;
6133 	struct sadb_msg *newmsg;
6134 
6135 	/* create new sadb_msg to reply. */
6136 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
6137 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6138 	if (!n)
6139 		return key_senderror(so, m, ENOBUFS);
6140 
6141 	if (n->m_len < sizeof(struct sadb_msg)) {
6142 		n = m_pullup(n, sizeof(struct sadb_msg));
6143 		if (n == NULL)
6144 			return key_senderror(so, m, ENOBUFS);
6145 	}
6146 	newmsg = mtod(n, struct sadb_msg *);
6147 	newmsg->sadb_msg_errno = 0;
6148 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6149 
6150 	m_freem(m);
6151 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6152     }
6153 }
6154 
6155 /*
6156  * Delete all alive SAs for corresponding xform.
6157  * Larval SAs have not initialized tdb_xform, so it is safe to leave them
6158  * here when xform disappears.
6159  */
6160 void
6161 key_delete_xform(const struct xformsw *xsp)
6162 {
6163 	struct secasvar_queue drainq;
6164 	struct secashead *sah;
6165 	struct secasvar *sav, *nextsav;
6166 
6167 	TAILQ_INIT(&drainq);
6168 	SAHTREE_WLOCK();
6169 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
6170 		sav = TAILQ_FIRST(&sah->savtree_alive);
6171 		if (sav == NULL)
6172 			continue;
6173 		if (sav->tdb_xform != xsp)
6174 			continue;
6175 		/*
6176 		 * It is supposed that all SAs in the chain are related to
6177 		 * one xform.
6178 		 */
6179 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6180 	}
6181 	/* Unlink all queued SAs from SPI hash */
6182 	TAILQ_FOREACH(sav, &drainq, chain) {
6183 		sav->state = SADB_SASTATE_DEAD;
6184 		LIST_REMOVE(sav, spihash);
6185 	}
6186 	SAHTREE_WUNLOCK();
6187 
6188 	/* Now we can release reference for all SAs in drainq */
6189 	sav = TAILQ_FIRST(&drainq);
6190 	while (sav != NULL) {
6191 		KEYDBG(KEY_STAMP,
6192 		    printf("%s: SA(%p)\n", __func__, sav));
6193 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6194 		nextsav = TAILQ_NEXT(sav, chain);
6195 		key_freesah(&sav->sah); /* release reference from SAV */
6196 		key_freesav(&sav); /* release last reference */
6197 		sav = nextsav;
6198 	}
6199 }
6200 
6201 /*
6202  * SADB_GET processing
6203  * receive
6204  *   <base, SA(*), address(SD)>
6205  * from the ikmpd, and get a SP and a SA to respond,
6206  * and send,
6207  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6208  *       (identity(SD),) (sensitivity)>
6209  * to the ikmpd.
6210  *
6211  * m will always be freed.
6212  */
6213 static int
6214 key_get(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6215 {
6216 	struct secasindex saidx;
6217 	struct sadb_address *src0, *dst0;
6218 	struct sadb_sa *sa0;
6219 	struct secasvar *sav;
6220 	uint8_t proto;
6221 
6222 	IPSEC_ASSERT(so != NULL, ("null socket"));
6223 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6224 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6225 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6226 
6227 	/* map satype to proto */
6228 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6229 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6230 			__func__));
6231 		return key_senderror(so, m, EINVAL);
6232 	}
6233 
6234 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
6235 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6236 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)) {
6237 		ipseclog((LOG_DEBUG,
6238 		    "%s: invalid message: missing required header.\n",
6239 		    __func__));
6240 		return key_senderror(so, m, EINVAL);
6241 	}
6242 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
6243 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6244 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6245 		ipseclog((LOG_DEBUG,
6246 		    "%s: invalid message: wrong header size.\n", __func__));
6247 		return key_senderror(so, m, EINVAL);
6248 	}
6249 
6250 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6251 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
6252 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
6253 
6254 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6255 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6256 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6257 		return key_senderror(so, m, EINVAL);
6258 	}
6259 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6260 
6261 	if (proto == IPPROTO_TCP)
6262 		sav = key_getsav_tcpmd5(&saidx, NULL);
6263 	else
6264 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6265 	if (sav == NULL) {
6266 		ipseclog((LOG_DEBUG, "%s: no SA found.\n", __func__));
6267 		return key_senderror(so, m, ESRCH);
6268 	}
6269 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6270 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6271 		    __func__, ntohl(sa0->sadb_sa_spi)));
6272 		key_freesav(&sav);
6273 		return (key_senderror(so, m, ESRCH));
6274 	}
6275 
6276     {
6277 	struct mbuf *n;
6278 	uint8_t satype;
6279 
6280 	/* map proto to satype */
6281 	if ((satype = key_proto2satype(sav->sah->saidx.proto)) == 0) {
6282 		ipseclog((LOG_DEBUG, "%s: there was invalid proto in SAD.\n",
6283 		    __func__));
6284 		key_freesav(&sav);
6285 		return key_senderror(so, m, EINVAL);
6286 	}
6287 
6288 	/* create new sadb_msg to reply. */
6289 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
6290 	    mhp->msg->sadb_msg_pid);
6291 
6292 	key_freesav(&sav);
6293 	if (!n)
6294 		return key_senderror(so, m, ENOBUFS);
6295 
6296 	m_freem(m);
6297 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
6298     }
6299 }
6300 
6301 /* XXX make it sysctl-configurable? */
6302 static void
6303 key_getcomb_setlifetime(struct sadb_comb *comb)
6304 {
6305 
6306 	comb->sadb_comb_soft_allocations = 1;
6307 	comb->sadb_comb_hard_allocations = 1;
6308 	comb->sadb_comb_soft_bytes = 0;
6309 	comb->sadb_comb_hard_bytes = 0;
6310 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
6311 	comb->sadb_comb_soft_addtime = comb->sadb_comb_soft_addtime * 80 / 100;
6312 	comb->sadb_comb_soft_usetime = 28800;	/* 8 hours */
6313 	comb->sadb_comb_hard_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
6314 }
6315 
6316 /*
6317  * XXX reorder combinations by preference
6318  * XXX no idea if the user wants ESP authentication or not
6319  */
6320 static struct mbuf *
6321 key_getcomb_ealg(void)
6322 {
6323 	struct sadb_comb *comb;
6324 	const struct enc_xform *algo;
6325 	struct mbuf *result = NULL, *m, *n;
6326 	int encmin;
6327 	int i, off, o;
6328 	int totlen;
6329 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6330 
6331 	m = NULL;
6332 	for (i = 1; i <= SADB_EALG_MAX; i++) {
6333 		algo = enc_algorithm_lookup(i);
6334 		if (algo == NULL)
6335 			continue;
6336 
6337 		/* discard algorithms with key size smaller than system min */
6338 		if (_BITS(algo->maxkey) < V_ipsec_esp_keymin)
6339 			continue;
6340 		if (_BITS(algo->minkey) < V_ipsec_esp_keymin)
6341 			encmin = V_ipsec_esp_keymin;
6342 		else
6343 			encmin = _BITS(algo->minkey);
6344 
6345 		if (V_ipsec_esp_auth)
6346 			m = key_getcomb_ah();
6347 		else {
6348 			IPSEC_ASSERT(l <= MLEN,
6349 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6350 			MGET(m, M_NOWAIT, MT_DATA);
6351 			if (m) {
6352 				M_ALIGN(m, l);
6353 				m->m_len = l;
6354 				m->m_next = NULL;
6355 				bzero(mtod(m, caddr_t), m->m_len);
6356 			}
6357 		}
6358 		if (!m)
6359 			goto fail;
6360 
6361 		totlen = 0;
6362 		for (n = m; n; n = n->m_next)
6363 			totlen += n->m_len;
6364 		IPSEC_ASSERT((totlen % l) == 0, ("totlen=%u, l=%u", totlen, l));
6365 
6366 		for (off = 0; off < totlen; off += l) {
6367 			n = m_pulldown(m, off, l, &o);
6368 			if (!n) {
6369 				/* m is already freed */
6370 				goto fail;
6371 			}
6372 			comb = (struct sadb_comb *)(mtod(n, caddr_t) + o);
6373 			bzero(comb, sizeof(*comb));
6374 			key_getcomb_setlifetime(comb);
6375 			comb->sadb_comb_encrypt = i;
6376 			comb->sadb_comb_encrypt_minbits = encmin;
6377 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
6378 		}
6379 
6380 		if (!result)
6381 			result = m;
6382 		else
6383 			m_cat(result, m);
6384 	}
6385 
6386 	return result;
6387 
6388  fail:
6389 	if (result)
6390 		m_freem(result);
6391 	return NULL;
6392 }
6393 
6394 static void
6395 key_getsizes_ah(const struct auth_hash *ah, int alg, u_int16_t* min,
6396     u_int16_t* max)
6397 {
6398 
6399 	*min = *max = ah->hashsize;
6400 	if (ah->keysize == 0) {
6401 		/*
6402 		 * Transform takes arbitrary key size but algorithm
6403 		 * key size is restricted.  Enforce this here.
6404 		 */
6405 		switch (alg) {
6406 		case SADB_X_AALG_MD5:	*min = *max = 16; break;
6407 		case SADB_X_AALG_SHA:	*min = *max = 20; break;
6408 		case SADB_X_AALG_NULL:	*min = 1; *max = 256; break;
6409 		case SADB_X_AALG_SHA2_256: *min = *max = 32; break;
6410 		case SADB_X_AALG_SHA2_384: *min = *max = 48; break;
6411 		case SADB_X_AALG_SHA2_512: *min = *max = 64; break;
6412 		default:
6413 			DPRINTF(("%s: unknown AH algorithm %u\n",
6414 				__func__, alg));
6415 			break;
6416 		}
6417 	}
6418 }
6419 
6420 /*
6421  * XXX reorder combinations by preference
6422  */
6423 static struct mbuf *
6424 key_getcomb_ah()
6425 {
6426 	const struct auth_hash *algo;
6427 	struct sadb_comb *comb;
6428 	struct mbuf *m;
6429 	u_int16_t minkeysize, maxkeysize;
6430 	int i;
6431 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6432 
6433 	m = NULL;
6434 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6435 #if 1
6436 		/* we prefer HMAC algorithms, not old algorithms */
6437 		if (i != SADB_AALG_SHA1HMAC &&
6438 		    i != SADB_AALG_MD5HMAC  &&
6439 		    i != SADB_X_AALG_SHA2_256 &&
6440 		    i != SADB_X_AALG_SHA2_384 &&
6441 		    i != SADB_X_AALG_SHA2_512)
6442 			continue;
6443 #endif
6444 		algo = auth_algorithm_lookup(i);
6445 		if (!algo)
6446 			continue;
6447 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6448 		/* discard algorithms with key size smaller than system min */
6449 		if (_BITS(minkeysize) < V_ipsec_ah_keymin)
6450 			continue;
6451 
6452 		if (!m) {
6453 			IPSEC_ASSERT(l <= MLEN,
6454 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6455 			MGET(m, M_NOWAIT, MT_DATA);
6456 			if (m) {
6457 				M_ALIGN(m, l);
6458 				m->m_len = l;
6459 				m->m_next = NULL;
6460 			}
6461 		} else
6462 			M_PREPEND(m, l, M_NOWAIT);
6463 		if (!m)
6464 			return NULL;
6465 
6466 		comb = mtod(m, struct sadb_comb *);
6467 		bzero(comb, sizeof(*comb));
6468 		key_getcomb_setlifetime(comb);
6469 		comb->sadb_comb_auth = i;
6470 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6471 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6472 	}
6473 
6474 	return m;
6475 }
6476 
6477 /*
6478  * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6479  * XXX reorder combinations by preference
6480  */
6481 static struct mbuf *
6482 key_getcomb_ipcomp()
6483 {
6484 	const struct comp_algo *algo;
6485 	struct sadb_comb *comb;
6486 	struct mbuf *m;
6487 	int i;
6488 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6489 
6490 	m = NULL;
6491 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6492 		algo = comp_algorithm_lookup(i);
6493 		if (!algo)
6494 			continue;
6495 
6496 		if (!m) {
6497 			IPSEC_ASSERT(l <= MLEN,
6498 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6499 			MGET(m, M_NOWAIT, MT_DATA);
6500 			if (m) {
6501 				M_ALIGN(m, l);
6502 				m->m_len = l;
6503 				m->m_next = NULL;
6504 			}
6505 		} else
6506 			M_PREPEND(m, l, M_NOWAIT);
6507 		if (!m)
6508 			return NULL;
6509 
6510 		comb = mtod(m, struct sadb_comb *);
6511 		bzero(comb, sizeof(*comb));
6512 		key_getcomb_setlifetime(comb);
6513 		comb->sadb_comb_encrypt = i;
6514 		/* what should we set into sadb_comb_*_{min,max}bits? */
6515 	}
6516 
6517 	return m;
6518 }
6519 
6520 /*
6521  * XXX no way to pass mode (transport/tunnel) to userland
6522  * XXX replay checking?
6523  * XXX sysctl interface to ipsec_{ah,esp}_keymin
6524  */
6525 static struct mbuf *
6526 key_getprop(const struct secasindex *saidx)
6527 {
6528 	struct sadb_prop *prop;
6529 	struct mbuf *m, *n;
6530 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6531 	int totlen;
6532 
6533 	switch (saidx->proto)  {
6534 	case IPPROTO_ESP:
6535 		m = key_getcomb_ealg();
6536 		break;
6537 	case IPPROTO_AH:
6538 		m = key_getcomb_ah();
6539 		break;
6540 	case IPPROTO_IPCOMP:
6541 		m = key_getcomb_ipcomp();
6542 		break;
6543 	default:
6544 		return NULL;
6545 	}
6546 
6547 	if (!m)
6548 		return NULL;
6549 	M_PREPEND(m, l, M_NOWAIT);
6550 	if (!m)
6551 		return NULL;
6552 
6553 	totlen = 0;
6554 	for (n = m; n; n = n->m_next)
6555 		totlen += n->m_len;
6556 
6557 	prop = mtod(m, struct sadb_prop *);
6558 	bzero(prop, sizeof(*prop));
6559 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6560 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6561 	prop->sadb_prop_replay = 32;	/* XXX */
6562 
6563 	return m;
6564 }
6565 
6566 /*
6567  * SADB_ACQUIRE processing called by key_checkrequest() and key_acquire2().
6568  * send
6569  *   <base, SA, address(SD), (address(P)), x_policy,
6570  *       (identity(SD),) (sensitivity,) proposal>
6571  * to KMD, and expect to receive
6572  *   <base> with SADB_ACQUIRE if error occurred,
6573  * or
6574  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6575  * from KMD by PF_KEY.
6576  *
6577  * XXX x_policy is outside of RFC2367 (KAME extension).
6578  * XXX sensitivity is not supported.
6579  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6580  * see comment for key_getcomb_ipcomp().
6581  *
6582  * OUT:
6583  *    0     : succeed
6584  *    others: error number
6585  */
6586 static int
6587 key_acquire(const struct secasindex *saidx, struct secpolicy *sp)
6588 {
6589 	union sockaddr_union addr;
6590 	struct mbuf *result, *m;
6591 	uint32_t seq;
6592 	int error;
6593 	uint16_t ul_proto;
6594 	uint8_t mask, satype;
6595 
6596 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
6597 	satype = key_proto2satype(saidx->proto);
6598 	IPSEC_ASSERT(satype != 0, ("null satype, protocol %u", saidx->proto));
6599 
6600 	error = -1;
6601 	result = NULL;
6602 	ul_proto = IPSEC_ULPROTO_ANY;
6603 
6604 	/* Get seq number to check whether sending message or not. */
6605 	seq = key_getacq(saidx, &error);
6606 	if (seq == 0)
6607 		return (error);
6608 
6609 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0);
6610 	if (!m) {
6611 		error = ENOBUFS;
6612 		goto fail;
6613 	}
6614 	result = m;
6615 
6616 	/*
6617 	 * set sadb_address for saidx's.
6618 	 *
6619 	 * Note that if sp is supplied, then we're being called from
6620 	 * key_allocsa_policy() and should supply port and protocol
6621 	 * information.
6622 	 * XXXAE: why only TCP and UDP? ICMP and SCTP looks applicable too.
6623 	 * XXXAE: probably we can handle this in the ipsec[46]_allocsa().
6624 	 * XXXAE: it looks like we should save this info in the ACQ entry.
6625 	 */
6626 	if (sp != NULL && (sp->spidx.ul_proto == IPPROTO_TCP ||
6627 	    sp->spidx.ul_proto == IPPROTO_UDP))
6628 		ul_proto = sp->spidx.ul_proto;
6629 
6630 	addr = saidx->src;
6631 	mask = FULLMASK;
6632 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6633 		switch (sp->spidx.src.sa.sa_family) {
6634 		case AF_INET:
6635 			if (sp->spidx.src.sin.sin_port != IPSEC_PORT_ANY) {
6636 				addr.sin.sin_port = sp->spidx.src.sin.sin_port;
6637 				mask = sp->spidx.prefs;
6638 			}
6639 			break;
6640 		case AF_INET6:
6641 			if (sp->spidx.src.sin6.sin6_port != IPSEC_PORT_ANY) {
6642 				addr.sin6.sin6_port =
6643 				    sp->spidx.src.sin6.sin6_port;
6644 				mask = sp->spidx.prefs;
6645 			}
6646 			break;
6647 		default:
6648 			break;
6649 		}
6650 	}
6651 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &addr.sa, mask, ul_proto);
6652 	if (!m) {
6653 		error = ENOBUFS;
6654 		goto fail;
6655 	}
6656 	m_cat(result, m);
6657 
6658 	addr = saidx->dst;
6659 	mask = FULLMASK;
6660 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6661 		switch (sp->spidx.dst.sa.sa_family) {
6662 		case AF_INET:
6663 			if (sp->spidx.dst.sin.sin_port != IPSEC_PORT_ANY) {
6664 				addr.sin.sin_port = sp->spidx.dst.sin.sin_port;
6665 				mask = sp->spidx.prefd;
6666 			}
6667 			break;
6668 		case AF_INET6:
6669 			if (sp->spidx.dst.sin6.sin6_port != IPSEC_PORT_ANY) {
6670 				addr.sin6.sin6_port =
6671 				    sp->spidx.dst.sin6.sin6_port;
6672 				mask = sp->spidx.prefd;
6673 			}
6674 			break;
6675 		default:
6676 			break;
6677 		}
6678 	}
6679 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &addr.sa, mask, ul_proto);
6680 	if (!m) {
6681 		error = ENOBUFS;
6682 		goto fail;
6683 	}
6684 	m_cat(result, m);
6685 
6686 	/* XXX proxy address (optional) */
6687 
6688 	/*
6689 	 * Set sadb_x_policy. This is KAME extension to RFC2367.
6690 	 */
6691 	if (sp != NULL) {
6692 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6693 		    sp->priority);
6694 		if (!m) {
6695 			error = ENOBUFS;
6696 			goto fail;
6697 		}
6698 		m_cat(result, m);
6699 	}
6700 
6701 	/*
6702 	 * Set sadb_x_sa2 extension if saidx->reqid is not zero.
6703 	 * This is FreeBSD extension to RFC2367.
6704 	 */
6705 	if (saidx->reqid != 0) {
6706 		m = key_setsadbxsa2(saidx->mode, 0, saidx->reqid);
6707 		if (m == NULL) {
6708 			error = ENOBUFS;
6709 			goto fail;
6710 		}
6711 		m_cat(result, m);
6712 	}
6713 	/* XXX identity (optional) */
6714 #if 0
6715 	if (idexttype && fqdn) {
6716 		/* create identity extension (FQDN) */
6717 		struct sadb_ident *id;
6718 		int fqdnlen;
6719 
6720 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
6721 		id = (struct sadb_ident *)p;
6722 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6723 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6724 		id->sadb_ident_exttype = idexttype;
6725 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
6726 		bcopy(fqdn, id + 1, fqdnlen);
6727 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
6728 	}
6729 
6730 	if (idexttype) {
6731 		/* create identity extension (USERFQDN) */
6732 		struct sadb_ident *id;
6733 		int userfqdnlen;
6734 
6735 		if (userfqdn) {
6736 			/* +1 for terminating-NUL */
6737 			userfqdnlen = strlen(userfqdn) + 1;
6738 		} else
6739 			userfqdnlen = 0;
6740 		id = (struct sadb_ident *)p;
6741 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6742 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6743 		id->sadb_ident_exttype = idexttype;
6744 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
6745 		/* XXX is it correct? */
6746 		if (curproc && curproc->p_cred)
6747 			id->sadb_ident_id = curproc->p_cred->p_ruid;
6748 		if (userfqdn && userfqdnlen)
6749 			bcopy(userfqdn, id + 1, userfqdnlen);
6750 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
6751 	}
6752 #endif
6753 
6754 	/* XXX sensitivity (optional) */
6755 
6756 	/* create proposal/combination extension */
6757 	m = key_getprop(saidx);
6758 #if 0
6759 	/*
6760 	 * spec conformant: always attach proposal/combination extension,
6761 	 * the problem is that we have no way to attach it for ipcomp,
6762 	 * due to the way sadb_comb is declared in RFC2367.
6763 	 */
6764 	if (!m) {
6765 		error = ENOBUFS;
6766 		goto fail;
6767 	}
6768 	m_cat(result, m);
6769 #else
6770 	/*
6771 	 * outside of spec; make proposal/combination extension optional.
6772 	 */
6773 	if (m)
6774 		m_cat(result, m);
6775 #endif
6776 
6777 	if ((result->m_flags & M_PKTHDR) == 0) {
6778 		error = EINVAL;
6779 		goto fail;
6780 	}
6781 
6782 	if (result->m_len < sizeof(struct sadb_msg)) {
6783 		result = m_pullup(result, sizeof(struct sadb_msg));
6784 		if (result == NULL) {
6785 			error = ENOBUFS;
6786 			goto fail;
6787 		}
6788 	}
6789 
6790 	result->m_pkthdr.len = 0;
6791 	for (m = result; m; m = m->m_next)
6792 		result->m_pkthdr.len += m->m_len;
6793 
6794 	mtod(result, struct sadb_msg *)->sadb_msg_len =
6795 	    PFKEY_UNIT64(result->m_pkthdr.len);
6796 
6797 	KEYDBG(KEY_STAMP,
6798 	    printf("%s: SP(%p)\n", __func__, sp));
6799 	KEYDBG(KEY_DATA, kdebug_secasindex(saidx, NULL));
6800 
6801 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
6802 
6803  fail:
6804 	if (result)
6805 		m_freem(result);
6806 	return error;
6807 }
6808 
6809 static uint32_t
6810 key_newacq(const struct secasindex *saidx, int *perror)
6811 {
6812 	struct secacq *acq;
6813 	uint32_t seq;
6814 
6815 	acq = malloc(sizeof(*acq), M_IPSEC_SAQ, M_NOWAIT | M_ZERO);
6816 	if (acq == NULL) {
6817 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6818 		*perror = ENOBUFS;
6819 		return (0);
6820 	}
6821 
6822 	/* copy secindex */
6823 	bcopy(saidx, &acq->saidx, sizeof(acq->saidx));
6824 	acq->created = time_second;
6825 	acq->count = 0;
6826 
6827 	/* add to acqtree */
6828 	ACQ_LOCK();
6829 	seq = acq->seq = (V_acq_seq == ~0 ? 1 : ++V_acq_seq);
6830 	LIST_INSERT_HEAD(&V_acqtree, acq, chain);
6831 	LIST_INSERT_HEAD(ACQADDRHASH_HASH(saidx), acq, addrhash);
6832 	LIST_INSERT_HEAD(ACQSEQHASH_HASH(seq), acq, seqhash);
6833 	ACQ_UNLOCK();
6834 	*perror = 0;
6835 	return (seq);
6836 }
6837 
6838 static uint32_t
6839 key_getacq(const struct secasindex *saidx, int *perror)
6840 {
6841 	struct secacq *acq;
6842 	uint32_t seq;
6843 
6844 	ACQ_LOCK();
6845 	LIST_FOREACH(acq, ACQADDRHASH_HASH(saidx), addrhash) {
6846 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY)) {
6847 			if (acq->count > V_key_blockacq_count) {
6848 				/*
6849 				 * Reset counter and send message.
6850 				 * Also reset created time to keep ACQ for
6851 				 * this saidx.
6852 				 */
6853 				acq->created = time_second;
6854 				acq->count = 0;
6855 				seq = acq->seq;
6856 			} else {
6857 				/*
6858 				 * Increment counter and do nothing.
6859 				 * We send SADB_ACQUIRE message only
6860 				 * for each V_key_blockacq_count packet.
6861 				 */
6862 				acq->count++;
6863 				seq = 0;
6864 			}
6865 			break;
6866 		}
6867 	}
6868 	ACQ_UNLOCK();
6869 	if (acq != NULL) {
6870 		*perror = 0;
6871 		return (seq);
6872 	}
6873 	/* allocate new  entry */
6874 	return (key_newacq(saidx, perror));
6875 }
6876 
6877 static int
6878 key_acqreset(uint32_t seq)
6879 {
6880 	struct secacq *acq;
6881 
6882 	ACQ_LOCK();
6883 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6884 		if (acq->seq == seq) {
6885 			acq->count = 0;
6886 			acq->created = time_second;
6887 			break;
6888 		}
6889 	}
6890 	ACQ_UNLOCK();
6891 	if (acq == NULL)
6892 		return (ESRCH);
6893 	return (0);
6894 }
6895 /*
6896  * Mark ACQ entry as stale to remove it in key_flush_acq().
6897  * Called after successful SADB_GETSPI message.
6898  */
6899 static int
6900 key_acqdone(const struct secasindex *saidx, uint32_t seq)
6901 {
6902 	struct secacq *acq;
6903 
6904 	ACQ_LOCK();
6905 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6906 		if (acq->seq == seq)
6907 			break;
6908 	}
6909 	if (acq != NULL) {
6910 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY) == 0) {
6911 			ipseclog((LOG_DEBUG,
6912 			    "%s: Mismatched saidx for ACQ %u", __func__, seq));
6913 			acq = NULL;
6914 		} else {
6915 			acq->created = 0;
6916 		}
6917 	} else {
6918 		ipseclog((LOG_DEBUG,
6919 		    "%s: ACQ %u is not found.", __func__, seq));
6920 	}
6921 	ACQ_UNLOCK();
6922 	if (acq == NULL)
6923 		return (ESRCH);
6924 	return (0);
6925 }
6926 
6927 static struct secspacq *
6928 key_newspacq(struct secpolicyindex *spidx)
6929 {
6930 	struct secspacq *acq;
6931 
6932 	/* get new entry */
6933 	acq = malloc(sizeof(struct secspacq), M_IPSEC_SAQ, M_NOWAIT|M_ZERO);
6934 	if (acq == NULL) {
6935 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6936 		return NULL;
6937 	}
6938 
6939 	/* copy secindex */
6940 	bcopy(spidx, &acq->spidx, sizeof(acq->spidx));
6941 	acq->created = time_second;
6942 	acq->count = 0;
6943 
6944 	/* add to spacqtree */
6945 	SPACQ_LOCK();
6946 	LIST_INSERT_HEAD(&V_spacqtree, acq, chain);
6947 	SPACQ_UNLOCK();
6948 
6949 	return acq;
6950 }
6951 
6952 static struct secspacq *
6953 key_getspacq(struct secpolicyindex *spidx)
6954 {
6955 	struct secspacq *acq;
6956 
6957 	SPACQ_LOCK();
6958 	LIST_FOREACH(acq, &V_spacqtree, chain) {
6959 		if (key_cmpspidx_exactly(spidx, &acq->spidx)) {
6960 			/* NB: return holding spacq_lock */
6961 			return acq;
6962 		}
6963 	}
6964 	SPACQ_UNLOCK();
6965 
6966 	return NULL;
6967 }
6968 
6969 /*
6970  * SADB_ACQUIRE processing,
6971  * in first situation, is receiving
6972  *   <base>
6973  * from the ikmpd, and clear sequence of its secasvar entry.
6974  *
6975  * In second situation, is receiving
6976  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6977  * from a user land process, and return
6978  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6979  * to the socket.
6980  *
6981  * m will always be freed.
6982  */
6983 static int
6984 key_acquire2(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6985 {
6986 	SAHTREE_RLOCK_TRACKER;
6987 	struct sadb_address *src0, *dst0;
6988 	struct secasindex saidx;
6989 	struct secashead *sah;
6990 	uint32_t reqid;
6991 	int error;
6992 	uint8_t mode, proto;
6993 
6994 	IPSEC_ASSERT(so != NULL, ("null socket"));
6995 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6996 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6997 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6998 
6999 	/*
7000 	 * Error message from KMd.
7001 	 * We assume that if error was occurred in IKEd, the length of PFKEY
7002 	 * message is equal to the size of sadb_msg structure.
7003 	 * We do not raise error even if error occurred in this function.
7004 	 */
7005 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
7006 		/* check sequence number */
7007 		if (mhp->msg->sadb_msg_seq == 0 ||
7008 		    mhp->msg->sadb_msg_errno == 0) {
7009 			ipseclog((LOG_DEBUG, "%s: must specify sequence "
7010 				"number and errno.\n", __func__));
7011 		} else {
7012 			/*
7013 			 * IKEd reported that error occurred.
7014 			 * XXXAE: what it expects from the kernel?
7015 			 * Probably we should send SADB_ACQUIRE again?
7016 			 * If so, reset ACQ's state.
7017 			 * XXXAE: it looks useless.
7018 			 */
7019 			key_acqreset(mhp->msg->sadb_msg_seq);
7020 		}
7021 		m_freem(m);
7022 		return (0);
7023 	}
7024 
7025 	/*
7026 	 * This message is from user land.
7027 	 */
7028 
7029 	/* map satype to proto */
7030 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7031 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7032 		    __func__));
7033 		return key_senderror(so, m, EINVAL);
7034 	}
7035 
7036 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
7037 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
7038 	    SADB_CHECKHDR(mhp, SADB_EXT_PROPOSAL)) {
7039 		ipseclog((LOG_DEBUG,
7040 		    "%s: invalid message: missing required header.\n",
7041 		    __func__));
7042 		return key_senderror(so, m, EINVAL);
7043 	}
7044 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
7045 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
7046 	    SADB_CHECKLEN(mhp, SADB_EXT_PROPOSAL)) {
7047 		ipseclog((LOG_DEBUG,
7048 		    "%s: invalid message: wrong header size.\n", __func__));
7049 		return key_senderror(so, m, EINVAL);
7050 	}
7051 
7052 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
7053 		mode = IPSEC_MODE_ANY;
7054 		reqid = 0;
7055 	} else {
7056 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
7057 			ipseclog((LOG_DEBUG,
7058 			    "%s: invalid message: wrong header size.\n",
7059 			    __func__));
7060 			return key_senderror(so, m, EINVAL);
7061 		}
7062 		mode = ((struct sadb_x_sa2 *)
7063 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
7064 		reqid = ((struct sadb_x_sa2 *)
7065 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
7066 	}
7067 
7068 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
7069 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
7070 
7071 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
7072 	    (struct sockaddr *)(dst0 + 1));
7073 	if (error != 0) {
7074 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
7075 		return key_senderror(so, m, EINVAL);
7076 	}
7077 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
7078 
7079 	/* get a SA index */
7080 	SAHTREE_RLOCK();
7081 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
7082 		if (key_cmpsaidx(&sah->saidx, &saidx, CMP_MODE_REQID))
7083 			break;
7084 	}
7085 	SAHTREE_RUNLOCK();
7086 	if (sah != NULL) {
7087 		ipseclog((LOG_DEBUG, "%s: a SA exists already.\n", __func__));
7088 		return key_senderror(so, m, EEXIST);
7089 	}
7090 
7091 	error = key_acquire(&saidx, NULL);
7092 	if (error != 0) {
7093 		ipseclog((LOG_DEBUG,
7094 		    "%s: error %d returned from key_acquire()\n",
7095 			__func__, error));
7096 		return key_senderror(so, m, error);
7097 	}
7098 	m_freem(m);
7099 	return (0);
7100 }
7101 
7102 /*
7103  * SADB_REGISTER processing.
7104  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
7105  * receive
7106  *   <base>
7107  * from the ikmpd, and register a socket to send PF_KEY messages,
7108  * and send
7109  *   <base, supported>
7110  * to KMD by PF_KEY.
7111  * If socket is detached, must free from regnode.
7112  *
7113  * m will always be freed.
7114  */
7115 static int
7116 key_register(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7117 {
7118 	struct secreg *reg, *newreg = NULL;
7119 
7120 	IPSEC_ASSERT(so != NULL, ("null socket"));
7121 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7122 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7123 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7124 
7125 	/* check for invalid register message */
7126 	if (mhp->msg->sadb_msg_satype >= sizeof(V_regtree)/sizeof(V_regtree[0]))
7127 		return key_senderror(so, m, EINVAL);
7128 
7129 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
7130 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
7131 		goto setmsg;
7132 
7133 	/* check whether existing or not */
7134 	REGTREE_LOCK();
7135 	LIST_FOREACH(reg, &V_regtree[mhp->msg->sadb_msg_satype], chain) {
7136 		if (reg->so == so) {
7137 			REGTREE_UNLOCK();
7138 			ipseclog((LOG_DEBUG, "%s: socket exists already.\n",
7139 				__func__));
7140 			return key_senderror(so, m, EEXIST);
7141 		}
7142 	}
7143 
7144 	/* create regnode */
7145 	newreg =  malloc(sizeof(struct secreg), M_IPSEC_SAR, M_NOWAIT|M_ZERO);
7146 	if (newreg == NULL) {
7147 		REGTREE_UNLOCK();
7148 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7149 		return key_senderror(so, m, ENOBUFS);
7150 	}
7151 
7152 	newreg->so = so;
7153 	((struct keycb *)sotorawcb(so))->kp_registered++;
7154 
7155 	/* add regnode to regtree. */
7156 	LIST_INSERT_HEAD(&V_regtree[mhp->msg->sadb_msg_satype], newreg, chain);
7157 	REGTREE_UNLOCK();
7158 
7159   setmsg:
7160     {
7161 	struct mbuf *n;
7162 	struct sadb_msg *newmsg;
7163 	struct sadb_supported *sup;
7164 	u_int len, alen, elen;
7165 	int off;
7166 	int i;
7167 	struct sadb_alg *alg;
7168 
7169 	/* create new sadb_msg to reply. */
7170 	alen = 0;
7171 	for (i = 1; i <= SADB_AALG_MAX; i++) {
7172 		if (auth_algorithm_lookup(i))
7173 			alen += sizeof(struct sadb_alg);
7174 	}
7175 	if (alen)
7176 		alen += sizeof(struct sadb_supported);
7177 	elen = 0;
7178 	for (i = 1; i <= SADB_EALG_MAX; i++) {
7179 		if (enc_algorithm_lookup(i))
7180 			elen += sizeof(struct sadb_alg);
7181 	}
7182 	if (elen)
7183 		elen += sizeof(struct sadb_supported);
7184 
7185 	len = sizeof(struct sadb_msg) + alen + elen;
7186 
7187 	if (len > MCLBYTES)
7188 		return key_senderror(so, m, ENOBUFS);
7189 
7190 	MGETHDR(n, M_NOWAIT, MT_DATA);
7191 	if (len > MHLEN) {
7192 		if (!(MCLGET(n, M_NOWAIT))) {
7193 			m_freem(n);
7194 			n = NULL;
7195 		}
7196 	}
7197 	if (!n)
7198 		return key_senderror(so, m, ENOBUFS);
7199 
7200 	n->m_pkthdr.len = n->m_len = len;
7201 	n->m_next = NULL;
7202 	off = 0;
7203 
7204 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
7205 	newmsg = mtod(n, struct sadb_msg *);
7206 	newmsg->sadb_msg_errno = 0;
7207 	newmsg->sadb_msg_len = PFKEY_UNIT64(len);
7208 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7209 
7210 	/* for authentication algorithm */
7211 	if (alen) {
7212 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7213 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7214 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7215 		off += PFKEY_ALIGN8(sizeof(*sup));
7216 
7217 		for (i = 1; i <= SADB_AALG_MAX; i++) {
7218 			const struct auth_hash *aalgo;
7219 			u_int16_t minkeysize, maxkeysize;
7220 
7221 			aalgo = auth_algorithm_lookup(i);
7222 			if (!aalgo)
7223 				continue;
7224 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7225 			alg->sadb_alg_id = i;
7226 			alg->sadb_alg_ivlen = 0;
7227 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7228 			alg->sadb_alg_minbits = _BITS(minkeysize);
7229 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7230 			off += PFKEY_ALIGN8(sizeof(*alg));
7231 		}
7232 	}
7233 
7234 	/* for encryption algorithm */
7235 	if (elen) {
7236 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7237 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7238 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7239 		off += PFKEY_ALIGN8(sizeof(*sup));
7240 
7241 		for (i = 1; i <= SADB_EALG_MAX; i++) {
7242 			const struct enc_xform *ealgo;
7243 
7244 			ealgo = enc_algorithm_lookup(i);
7245 			if (!ealgo)
7246 				continue;
7247 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7248 			alg->sadb_alg_id = i;
7249 			alg->sadb_alg_ivlen = ealgo->ivsize;
7250 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7251 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7252 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7253 		}
7254 	}
7255 
7256 	IPSEC_ASSERT(off == len,
7257 		("length assumption failed (off %u len %u)", off, len));
7258 
7259 	m_freem(m);
7260 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7261     }
7262 }
7263 
7264 /*
7265  * free secreg entry registered.
7266  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7267  */
7268 void
7269 key_freereg(struct socket *so)
7270 {
7271 	struct secreg *reg;
7272 	int i;
7273 
7274 	IPSEC_ASSERT(so != NULL, ("NULL so"));
7275 
7276 	/*
7277 	 * check whether existing or not.
7278 	 * check all type of SA, because there is a potential that
7279 	 * one socket is registered to multiple type of SA.
7280 	 */
7281 	REGTREE_LOCK();
7282 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7283 		LIST_FOREACH(reg, &V_regtree[i], chain) {
7284 			if (reg->so == so && __LIST_CHAINED(reg)) {
7285 				LIST_REMOVE(reg, chain);
7286 				free(reg, M_IPSEC_SAR);
7287 				break;
7288 			}
7289 		}
7290 	}
7291 	REGTREE_UNLOCK();
7292 }
7293 
7294 /*
7295  * SADB_EXPIRE processing
7296  * send
7297  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7298  * to KMD by PF_KEY.
7299  * NOTE: We send only soft lifetime extension.
7300  *
7301  * OUT:	0	: succeed
7302  *	others	: error number
7303  */
7304 static int
7305 key_expire(struct secasvar *sav, int hard)
7306 {
7307 	struct mbuf *result = NULL, *m;
7308 	struct sadb_lifetime *lt;
7309 	uint32_t replay_count;
7310 	int error, len;
7311 	uint8_t satype;
7312 
7313 	IPSEC_ASSERT (sav != NULL, ("null sav"));
7314 	IPSEC_ASSERT (sav->sah != NULL, ("null sa header"));
7315 
7316 	KEYDBG(KEY_STAMP,
7317 	    printf("%s: SA(%p) expired %s lifetime\n", __func__,
7318 		sav, hard ? "hard": "soft"));
7319 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
7320 	/* set msg header */
7321 	satype = key_proto2satype(sav->sah->saidx.proto);
7322 	IPSEC_ASSERT(satype != 0, ("invalid proto, satype %u", satype));
7323 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, sav->refcnt);
7324 	if (!m) {
7325 		error = ENOBUFS;
7326 		goto fail;
7327 	}
7328 	result = m;
7329 
7330 	/* create SA extension */
7331 	m = key_setsadbsa(sav);
7332 	if (!m) {
7333 		error = ENOBUFS;
7334 		goto fail;
7335 	}
7336 	m_cat(result, m);
7337 
7338 	/* create SA extension */
7339 	SECASVAR_LOCK(sav);
7340 	replay_count = sav->replay ? sav->replay->count : 0;
7341 	SECASVAR_UNLOCK(sav);
7342 
7343 	m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
7344 			sav->sah->saidx.reqid);
7345 	if (!m) {
7346 		error = ENOBUFS;
7347 		goto fail;
7348 	}
7349 	m_cat(result, m);
7350 
7351 	if (sav->replay && sav->replay->wsize > UINT8_MAX) {
7352 		m = key_setsadbxsareplay(sav->replay->wsize);
7353 		if (!m) {
7354 			error = ENOBUFS;
7355 			goto fail;
7356 		}
7357 		m_cat(result, m);
7358 	}
7359 
7360 	/* create lifetime extension (current and soft) */
7361 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7362 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
7363 	if (m == NULL) {
7364 		error = ENOBUFS;
7365 		goto fail;
7366 	}
7367 	m_align(m, len);
7368 	m->m_len = len;
7369 	bzero(mtod(m, caddr_t), len);
7370 	lt = mtod(m, struct sadb_lifetime *);
7371 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7372 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7373 	lt->sadb_lifetime_allocations =
7374 	    (uint32_t)counter_u64_fetch(sav->lft_c_allocations);
7375 	lt->sadb_lifetime_bytes =
7376 	    counter_u64_fetch(sav->lft_c_bytes);
7377 	lt->sadb_lifetime_addtime = sav->created;
7378 	lt->sadb_lifetime_usetime = sav->firstused;
7379 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
7380 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7381 	if (hard) {
7382 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
7383 		lt->sadb_lifetime_allocations = sav->lft_h->allocations;
7384 		lt->sadb_lifetime_bytes = sav->lft_h->bytes;
7385 		lt->sadb_lifetime_addtime = sav->lft_h->addtime;
7386 		lt->sadb_lifetime_usetime = sav->lft_h->usetime;
7387 	} else {
7388 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_SOFT;
7389 		lt->sadb_lifetime_allocations = sav->lft_s->allocations;
7390 		lt->sadb_lifetime_bytes = sav->lft_s->bytes;
7391 		lt->sadb_lifetime_addtime = sav->lft_s->addtime;
7392 		lt->sadb_lifetime_usetime = sav->lft_s->usetime;
7393 	}
7394 	m_cat(result, m);
7395 
7396 	/* set sadb_address for source */
7397 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
7398 	    &sav->sah->saidx.src.sa,
7399 	    FULLMASK, IPSEC_ULPROTO_ANY);
7400 	if (!m) {
7401 		error = ENOBUFS;
7402 		goto fail;
7403 	}
7404 	m_cat(result, m);
7405 
7406 	/* set sadb_address for destination */
7407 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
7408 	    &sav->sah->saidx.dst.sa,
7409 	    FULLMASK, IPSEC_ULPROTO_ANY);
7410 	if (!m) {
7411 		error = ENOBUFS;
7412 		goto fail;
7413 	}
7414 	m_cat(result, m);
7415 
7416 	/*
7417 	 * XXX-BZ Handle NAT-T extensions here.
7418 	 * XXXAE: it doesn't seem quite useful. IKEs should not depend on
7419 	 * this information, we report only significant SA fields.
7420 	 */
7421 
7422 	if ((result->m_flags & M_PKTHDR) == 0) {
7423 		error = EINVAL;
7424 		goto fail;
7425 	}
7426 
7427 	if (result->m_len < sizeof(struct sadb_msg)) {
7428 		result = m_pullup(result, sizeof(struct sadb_msg));
7429 		if (result == NULL) {
7430 			error = ENOBUFS;
7431 			goto fail;
7432 		}
7433 	}
7434 
7435 	result->m_pkthdr.len = 0;
7436 	for (m = result; m; m = m->m_next)
7437 		result->m_pkthdr.len += m->m_len;
7438 
7439 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7440 	    PFKEY_UNIT64(result->m_pkthdr.len);
7441 
7442 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7443 
7444  fail:
7445 	if (result)
7446 		m_freem(result);
7447 	return error;
7448 }
7449 
7450 static void
7451 key_freesah_flushed(struct secashead_queue *flushq)
7452 {
7453 	struct secashead *sah, *nextsah;
7454 	struct secasvar *sav, *nextsav;
7455 
7456 	sah = TAILQ_FIRST(flushq);
7457 	while (sah != NULL) {
7458 		sav = TAILQ_FIRST(&sah->savtree_larval);
7459 		while (sav != NULL) {
7460 			nextsav = TAILQ_NEXT(sav, chain);
7461 			TAILQ_REMOVE(&sah->savtree_larval, sav, chain);
7462 			key_freesav(&sav); /* release last reference */
7463 			key_freesah(&sah); /* release reference from SAV */
7464 			sav = nextsav;
7465 		}
7466 		sav = TAILQ_FIRST(&sah->savtree_alive);
7467 		while (sav != NULL) {
7468 			nextsav = TAILQ_NEXT(sav, chain);
7469 			TAILQ_REMOVE(&sah->savtree_alive, sav, chain);
7470 			key_freesav(&sav); /* release last reference */
7471 			key_freesah(&sah); /* release reference from SAV */
7472 			sav = nextsav;
7473 		}
7474 		nextsah = TAILQ_NEXT(sah, chain);
7475 		key_freesah(&sah);	/* release last reference */
7476 		sah = nextsah;
7477 	}
7478 }
7479 
7480 /*
7481  * SADB_FLUSH processing
7482  * receive
7483  *   <base>
7484  * from the ikmpd, and free all entries in secastree.
7485  * and send,
7486  *   <base>
7487  * to the ikmpd.
7488  * NOTE: to do is only marking SADB_SASTATE_DEAD.
7489  *
7490  * m will always be freed.
7491  */
7492 static int
7493 key_flush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7494 {
7495 	struct secashead_queue flushq;
7496 	struct sadb_msg *newmsg;
7497 	struct secashead *sah, *nextsah;
7498 	struct secasvar *sav;
7499 	uint8_t proto;
7500 	int i;
7501 
7502 	IPSEC_ASSERT(so != NULL, ("null socket"));
7503 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7504 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7505 
7506 	/* map satype to proto */
7507 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7508 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7509 			__func__));
7510 		return key_senderror(so, m, EINVAL);
7511 	}
7512 	KEYDBG(KEY_STAMP,
7513 	    printf("%s: proto %u\n", __func__, proto));
7514 
7515 	TAILQ_INIT(&flushq);
7516 	if (proto == IPSEC_PROTO_ANY) {
7517 		/* no SATYPE specified, i.e. flushing all SA. */
7518 		SAHTREE_WLOCK();
7519 		/* Move all SAHs into flushq */
7520 		TAILQ_CONCAT(&flushq, &V_sahtree, chain);
7521 		/* Flush all buckets in SPI hash */
7522 		for (i = 0; i < V_savhash_mask + 1; i++)
7523 			LIST_INIT(&V_savhashtbl[i]);
7524 		/* Flush all buckets in SAHADDRHASH */
7525 		for (i = 0; i < V_sahaddrhash_mask + 1; i++)
7526 			LIST_INIT(&V_sahaddrhashtbl[i]);
7527 		/* Mark all SAHs as unlinked */
7528 		TAILQ_FOREACH(sah, &flushq, chain) {
7529 			sah->state = SADB_SASTATE_DEAD;
7530 			/*
7531 			 * Callout handler makes its job using
7532 			 * RLOCK and drain queues. In case, when this
7533 			 * function will be called just before it
7534 			 * acquires WLOCK, we need to mark SAs as
7535 			 * unlinked to prevent second unlink.
7536 			 */
7537 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7538 				sav->state = SADB_SASTATE_DEAD;
7539 			}
7540 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7541 				sav->state = SADB_SASTATE_DEAD;
7542 			}
7543 		}
7544 		SAHTREE_WUNLOCK();
7545 	} else {
7546 		SAHTREE_WLOCK();
7547 		sah = TAILQ_FIRST(&V_sahtree);
7548 		while (sah != NULL) {
7549 			IPSEC_ASSERT(sah->state != SADB_SASTATE_DEAD,
7550 			    ("DEAD SAH %p in SADB_FLUSH", sah));
7551 			nextsah = TAILQ_NEXT(sah, chain);
7552 			if (sah->saidx.proto != proto) {
7553 				sah = nextsah;
7554 				continue;
7555 			}
7556 			sah->state = SADB_SASTATE_DEAD;
7557 			TAILQ_REMOVE(&V_sahtree, sah, chain);
7558 			LIST_REMOVE(sah, addrhash);
7559 			/* Unlink all SAs from SPI hash */
7560 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7561 				LIST_REMOVE(sav, spihash);
7562 				sav->state = SADB_SASTATE_DEAD;
7563 			}
7564 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7565 				LIST_REMOVE(sav, spihash);
7566 				sav->state = SADB_SASTATE_DEAD;
7567 			}
7568 			/* Add SAH into flushq */
7569 			TAILQ_INSERT_HEAD(&flushq, sah, chain);
7570 			sah = nextsah;
7571 		}
7572 		SAHTREE_WUNLOCK();
7573 	}
7574 
7575 	key_freesah_flushed(&flushq);
7576 	/* Free all queued SAs and SAHs */
7577 	if (m->m_len < sizeof(struct sadb_msg) ||
7578 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7579 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7580 		return key_senderror(so, m, ENOBUFS);
7581 	}
7582 
7583 	if (m->m_next)
7584 		m_freem(m->m_next);
7585 	m->m_next = NULL;
7586 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7587 	newmsg = mtod(m, struct sadb_msg *);
7588 	newmsg->sadb_msg_errno = 0;
7589 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7590 
7591 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7592 }
7593 
7594 /*
7595  * SADB_DUMP processing
7596  * dump all entries including status of DEAD in SAD.
7597  * receive
7598  *   <base>
7599  * from the ikmpd, and dump all secasvar leaves
7600  * and send,
7601  *   <base> .....
7602  * to the ikmpd.
7603  *
7604  * m will always be freed.
7605  */
7606 static int
7607 key_dump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7608 {
7609 	SAHTREE_RLOCK_TRACKER;
7610 	struct secashead *sah;
7611 	struct secasvar *sav;
7612 	struct mbuf *n;
7613 	uint32_t cnt;
7614 	uint8_t proto, satype;
7615 
7616 	IPSEC_ASSERT(so != NULL, ("null socket"));
7617 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7618 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7619 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7620 
7621 	/* map satype to proto */
7622 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7623 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7624 		    __func__));
7625 		return key_senderror(so, m, EINVAL);
7626 	}
7627 
7628 	/* count sav entries to be sent to the userland. */
7629 	cnt = 0;
7630 	SAHTREE_RLOCK();
7631 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7632 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7633 		    proto != sah->saidx.proto)
7634 			continue;
7635 
7636 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain)
7637 			cnt++;
7638 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain)
7639 			cnt++;
7640 	}
7641 
7642 	if (cnt == 0) {
7643 		SAHTREE_RUNLOCK();
7644 		return key_senderror(so, m, ENOENT);
7645 	}
7646 
7647 	/* send this to the userland, one at a time. */
7648 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7649 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7650 		    proto != sah->saidx.proto)
7651 			continue;
7652 
7653 		/* map proto to satype */
7654 		if ((satype = key_proto2satype(sah->saidx.proto)) == 0) {
7655 			SAHTREE_RUNLOCK();
7656 			ipseclog((LOG_DEBUG, "%s: there was invalid proto in "
7657 			    "SAD.\n", __func__));
7658 			return key_senderror(so, m, EINVAL);
7659 		}
7660 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7661 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7662 			    --cnt, mhp->msg->sadb_msg_pid);
7663 			if (n == NULL) {
7664 				SAHTREE_RUNLOCK();
7665 				return key_senderror(so, m, ENOBUFS);
7666 			}
7667 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7668 		}
7669 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7670 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7671 			    --cnt, mhp->msg->sadb_msg_pid);
7672 			if (n == NULL) {
7673 				SAHTREE_RUNLOCK();
7674 				return key_senderror(so, m, ENOBUFS);
7675 			}
7676 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7677 		}
7678 	}
7679 	SAHTREE_RUNLOCK();
7680 	m_freem(m);
7681 	return (0);
7682 }
7683 /*
7684  * SADB_X_PROMISC processing
7685  *
7686  * m will always be freed.
7687  */
7688 static int
7689 key_promisc(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7690 {
7691 	int olen;
7692 
7693 	IPSEC_ASSERT(so != NULL, ("null socket"));
7694 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7695 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7696 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7697 
7698 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7699 
7700 	if (olen < sizeof(struct sadb_msg)) {
7701 #if 1
7702 		return key_senderror(so, m, EINVAL);
7703 #else
7704 		m_freem(m);
7705 		return 0;
7706 #endif
7707 	} else if (olen == sizeof(struct sadb_msg)) {
7708 		/* enable/disable promisc mode */
7709 		struct keycb *kp;
7710 
7711 		if ((kp = (struct keycb *)sotorawcb(so)) == NULL)
7712 			return key_senderror(so, m, EINVAL);
7713 		mhp->msg->sadb_msg_errno = 0;
7714 		switch (mhp->msg->sadb_msg_satype) {
7715 		case 0:
7716 		case 1:
7717 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
7718 			break;
7719 		default:
7720 			return key_senderror(so, m, EINVAL);
7721 		}
7722 
7723 		/* send the original message back to everyone */
7724 		mhp->msg->sadb_msg_errno = 0;
7725 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7726 	} else {
7727 		/* send packet as is */
7728 
7729 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
7730 
7731 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
7732 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7733 	}
7734 }
7735 
7736 static int (*key_typesw[])(struct socket *, struct mbuf *,
7737 		const struct sadb_msghdr *) = {
7738 	NULL,		/* SADB_RESERVED */
7739 	key_getspi,	/* SADB_GETSPI */
7740 	key_update,	/* SADB_UPDATE */
7741 	key_add,	/* SADB_ADD */
7742 	key_delete,	/* SADB_DELETE */
7743 	key_get,	/* SADB_GET */
7744 	key_acquire2,	/* SADB_ACQUIRE */
7745 	key_register,	/* SADB_REGISTER */
7746 	NULL,		/* SADB_EXPIRE */
7747 	key_flush,	/* SADB_FLUSH */
7748 	key_dump,	/* SADB_DUMP */
7749 	key_promisc,	/* SADB_X_PROMISC */
7750 	NULL,		/* SADB_X_PCHANGE */
7751 	key_spdadd,	/* SADB_X_SPDUPDATE */
7752 	key_spdadd,	/* SADB_X_SPDADD */
7753 	key_spddelete,	/* SADB_X_SPDDELETE */
7754 	key_spdget,	/* SADB_X_SPDGET */
7755 	NULL,		/* SADB_X_SPDACQUIRE */
7756 	key_spddump,	/* SADB_X_SPDDUMP */
7757 	key_spdflush,	/* SADB_X_SPDFLUSH */
7758 	key_spdadd,	/* SADB_X_SPDSETIDX */
7759 	NULL,		/* SADB_X_SPDEXPIRE */
7760 	key_spddelete2,	/* SADB_X_SPDDELETE2 */
7761 };
7762 
7763 /*
7764  * parse sadb_msg buffer to process PFKEYv2,
7765  * and create a data to response if needed.
7766  * I think to be dealed with mbuf directly.
7767  * IN:
7768  *     msgp  : pointer to pointer to a received buffer pulluped.
7769  *             This is rewrited to response.
7770  *     so    : pointer to socket.
7771  * OUT:
7772  *    length for buffer to send to user process.
7773  */
7774 int
7775 key_parse(struct mbuf *m, struct socket *so)
7776 {
7777 	struct sadb_msg *msg;
7778 	struct sadb_msghdr mh;
7779 	u_int orglen;
7780 	int error;
7781 	int target;
7782 
7783 	IPSEC_ASSERT(so != NULL, ("null socket"));
7784 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7785 
7786 	if (m->m_len < sizeof(struct sadb_msg)) {
7787 		m = m_pullup(m, sizeof(struct sadb_msg));
7788 		if (!m)
7789 			return ENOBUFS;
7790 	}
7791 	msg = mtod(m, struct sadb_msg *);
7792 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
7793 	target = KEY_SENDUP_ONE;
7794 
7795 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len != orglen) {
7796 		ipseclog((LOG_DEBUG, "%s: invalid message length.\n",__func__));
7797 		PFKEYSTAT_INC(out_invlen);
7798 		error = EINVAL;
7799 		goto senderror;
7800 	}
7801 
7802 	if (msg->sadb_msg_version != PF_KEY_V2) {
7803 		ipseclog((LOG_DEBUG, "%s: PF_KEY version %u is mismatched.\n",
7804 		    __func__, msg->sadb_msg_version));
7805 		PFKEYSTAT_INC(out_invver);
7806 		error = EINVAL;
7807 		goto senderror;
7808 	}
7809 
7810 	if (msg->sadb_msg_type > SADB_MAX) {
7811 		ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7812 		    __func__, msg->sadb_msg_type));
7813 		PFKEYSTAT_INC(out_invmsgtype);
7814 		error = EINVAL;
7815 		goto senderror;
7816 	}
7817 
7818 	/* for old-fashioned code - should be nuked */
7819 	if (m->m_pkthdr.len > MCLBYTES) {
7820 		m_freem(m);
7821 		return ENOBUFS;
7822 	}
7823 	if (m->m_next) {
7824 		struct mbuf *n;
7825 
7826 		MGETHDR(n, M_NOWAIT, MT_DATA);
7827 		if (n && m->m_pkthdr.len > MHLEN) {
7828 			if (!(MCLGET(n, M_NOWAIT))) {
7829 				m_free(n);
7830 				n = NULL;
7831 			}
7832 		}
7833 		if (!n) {
7834 			m_freem(m);
7835 			return ENOBUFS;
7836 		}
7837 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
7838 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
7839 		n->m_next = NULL;
7840 		m_freem(m);
7841 		m = n;
7842 	}
7843 
7844 	/* align the mbuf chain so that extensions are in contiguous region. */
7845 	error = key_align(m, &mh);
7846 	if (error)
7847 		return error;
7848 
7849 	msg = mh.msg;
7850 
7851 	/* We use satype as scope mask for spddump */
7852 	if (msg->sadb_msg_type == SADB_X_SPDDUMP) {
7853 		switch (msg->sadb_msg_satype) {
7854 		case IPSEC_POLICYSCOPE_ANY:
7855 		case IPSEC_POLICYSCOPE_GLOBAL:
7856 		case IPSEC_POLICYSCOPE_IFNET:
7857 		case IPSEC_POLICYSCOPE_PCB:
7858 			break;
7859 		default:
7860 			ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7861 			    __func__, msg->sadb_msg_type));
7862 			PFKEYSTAT_INC(out_invsatype);
7863 			error = EINVAL;
7864 			goto senderror;
7865 		}
7866 	} else {
7867 		switch (msg->sadb_msg_satype) { /* check SA type */
7868 		case SADB_SATYPE_UNSPEC:
7869 			switch (msg->sadb_msg_type) {
7870 			case SADB_GETSPI:
7871 			case SADB_UPDATE:
7872 			case SADB_ADD:
7873 			case SADB_DELETE:
7874 			case SADB_GET:
7875 			case SADB_ACQUIRE:
7876 			case SADB_EXPIRE:
7877 				ipseclog((LOG_DEBUG, "%s: must specify satype "
7878 				    "when msg type=%u.\n", __func__,
7879 				    msg->sadb_msg_type));
7880 				PFKEYSTAT_INC(out_invsatype);
7881 				error = EINVAL;
7882 				goto senderror;
7883 			}
7884 			break;
7885 		case SADB_SATYPE_AH:
7886 		case SADB_SATYPE_ESP:
7887 		case SADB_X_SATYPE_IPCOMP:
7888 		case SADB_X_SATYPE_TCPSIGNATURE:
7889 			switch (msg->sadb_msg_type) {
7890 			case SADB_X_SPDADD:
7891 			case SADB_X_SPDDELETE:
7892 			case SADB_X_SPDGET:
7893 			case SADB_X_SPDFLUSH:
7894 			case SADB_X_SPDSETIDX:
7895 			case SADB_X_SPDUPDATE:
7896 			case SADB_X_SPDDELETE2:
7897 				ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7898 				    __func__, msg->sadb_msg_type));
7899 				PFKEYSTAT_INC(out_invsatype);
7900 				error = EINVAL;
7901 				goto senderror;
7902 			}
7903 			break;
7904 		case SADB_SATYPE_RSVP:
7905 		case SADB_SATYPE_OSPFV2:
7906 		case SADB_SATYPE_RIPV2:
7907 		case SADB_SATYPE_MIP:
7908 			ipseclog((LOG_DEBUG, "%s: type %u isn't supported.\n",
7909 			    __func__, msg->sadb_msg_satype));
7910 			PFKEYSTAT_INC(out_invsatype);
7911 			error = EOPNOTSUPP;
7912 			goto senderror;
7913 		case 1:	/* XXX: What does it do? */
7914 			if (msg->sadb_msg_type == SADB_X_PROMISC)
7915 				break;
7916 			/*FALLTHROUGH*/
7917 		default:
7918 			ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7919 			    __func__, msg->sadb_msg_satype));
7920 			PFKEYSTAT_INC(out_invsatype);
7921 			error = EINVAL;
7922 			goto senderror;
7923 		}
7924 	}
7925 
7926 	/* check field of upper layer protocol and address family */
7927 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL
7928 	 && mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
7929 		struct sadb_address *src0, *dst0;
7930 		u_int plen;
7931 
7932 		src0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_SRC]);
7933 		dst0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_DST]);
7934 
7935 		/* check upper layer protocol */
7936 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
7937 			ipseclog((LOG_DEBUG, "%s: upper layer protocol "
7938 				"mismatched.\n", __func__));
7939 			PFKEYSTAT_INC(out_invaddr);
7940 			error = EINVAL;
7941 			goto senderror;
7942 		}
7943 
7944 		/* check family */
7945 		if (PFKEY_ADDR_SADDR(src0)->sa_family !=
7946 		    PFKEY_ADDR_SADDR(dst0)->sa_family) {
7947 			ipseclog((LOG_DEBUG, "%s: address family mismatched.\n",
7948 				__func__));
7949 			PFKEYSTAT_INC(out_invaddr);
7950 			error = EINVAL;
7951 			goto senderror;
7952 		}
7953 		if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7954 		    PFKEY_ADDR_SADDR(dst0)->sa_len) {
7955 			ipseclog((LOG_DEBUG, "%s: address struct size "
7956 				"mismatched.\n", __func__));
7957 			PFKEYSTAT_INC(out_invaddr);
7958 			error = EINVAL;
7959 			goto senderror;
7960 		}
7961 
7962 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7963 		case AF_INET:
7964 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7965 			    sizeof(struct sockaddr_in)) {
7966 				PFKEYSTAT_INC(out_invaddr);
7967 				error = EINVAL;
7968 				goto senderror;
7969 			}
7970 			break;
7971 		case AF_INET6:
7972 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7973 			    sizeof(struct sockaddr_in6)) {
7974 				PFKEYSTAT_INC(out_invaddr);
7975 				error = EINVAL;
7976 				goto senderror;
7977 			}
7978 			break;
7979 		default:
7980 			ipseclog((LOG_DEBUG, "%s: unsupported address family\n",
7981 				__func__));
7982 			PFKEYSTAT_INC(out_invaddr);
7983 			error = EAFNOSUPPORT;
7984 			goto senderror;
7985 		}
7986 
7987 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7988 		case AF_INET:
7989 			plen = sizeof(struct in_addr) << 3;
7990 			break;
7991 		case AF_INET6:
7992 			plen = sizeof(struct in6_addr) << 3;
7993 			break;
7994 		default:
7995 			plen = 0;	/*fool gcc*/
7996 			break;
7997 		}
7998 
7999 		/* check max prefix length */
8000 		if (src0->sadb_address_prefixlen > plen ||
8001 		    dst0->sadb_address_prefixlen > plen) {
8002 			ipseclog((LOG_DEBUG, "%s: illegal prefixlen.\n",
8003 				__func__));
8004 			PFKEYSTAT_INC(out_invaddr);
8005 			error = EINVAL;
8006 			goto senderror;
8007 		}
8008 
8009 		/*
8010 		 * prefixlen == 0 is valid because there can be a case when
8011 		 * all addresses are matched.
8012 		 */
8013 	}
8014 
8015 	if (msg->sadb_msg_type >= nitems(key_typesw) ||
8016 	    key_typesw[msg->sadb_msg_type] == NULL) {
8017 		PFKEYSTAT_INC(out_invmsgtype);
8018 		error = EINVAL;
8019 		goto senderror;
8020 	}
8021 
8022 	return (*key_typesw[msg->sadb_msg_type])(so, m, &mh);
8023 
8024 senderror:
8025 	msg->sadb_msg_errno = error;
8026 	return key_sendup_mbuf(so, m, target);
8027 }
8028 
8029 static int
8030 key_senderror(struct socket *so, struct mbuf *m, int code)
8031 {
8032 	struct sadb_msg *msg;
8033 
8034 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8035 		("mbuf too small, len %u", m->m_len));
8036 
8037 	msg = mtod(m, struct sadb_msg *);
8038 	msg->sadb_msg_errno = code;
8039 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
8040 }
8041 
8042 /*
8043  * set the pointer to each header into message buffer.
8044  * m will be freed on error.
8045  * XXX larger-than-MCLBYTES extension?
8046  */
8047 static int
8048 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
8049 {
8050 	struct mbuf *n;
8051 	struct sadb_ext *ext;
8052 	size_t off, end;
8053 	int extlen;
8054 	int toff;
8055 
8056 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
8057 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
8058 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8059 		("mbuf too small, len %u", m->m_len));
8060 
8061 	/* initialize */
8062 	bzero(mhp, sizeof(*mhp));
8063 
8064 	mhp->msg = mtod(m, struct sadb_msg *);
8065 	mhp->ext[0] = (struct sadb_ext *)mhp->msg;	/*XXX backward compat */
8066 
8067 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
8068 	extlen = end;	/*just in case extlen is not updated*/
8069 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
8070 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
8071 		if (!n) {
8072 			/* m is already freed */
8073 			return ENOBUFS;
8074 		}
8075 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8076 
8077 		/* set pointer */
8078 		switch (ext->sadb_ext_type) {
8079 		case SADB_EXT_SA:
8080 		case SADB_EXT_ADDRESS_SRC:
8081 		case SADB_EXT_ADDRESS_DST:
8082 		case SADB_EXT_ADDRESS_PROXY:
8083 		case SADB_EXT_LIFETIME_CURRENT:
8084 		case SADB_EXT_LIFETIME_HARD:
8085 		case SADB_EXT_LIFETIME_SOFT:
8086 		case SADB_EXT_KEY_AUTH:
8087 		case SADB_EXT_KEY_ENCRYPT:
8088 		case SADB_EXT_IDENTITY_SRC:
8089 		case SADB_EXT_IDENTITY_DST:
8090 		case SADB_EXT_SENSITIVITY:
8091 		case SADB_EXT_PROPOSAL:
8092 		case SADB_EXT_SUPPORTED_AUTH:
8093 		case SADB_EXT_SUPPORTED_ENCRYPT:
8094 		case SADB_EXT_SPIRANGE:
8095 		case SADB_X_EXT_POLICY:
8096 		case SADB_X_EXT_SA2:
8097 		case SADB_X_EXT_NAT_T_TYPE:
8098 		case SADB_X_EXT_NAT_T_SPORT:
8099 		case SADB_X_EXT_NAT_T_DPORT:
8100 		case SADB_X_EXT_NAT_T_OAI:
8101 		case SADB_X_EXT_NAT_T_OAR:
8102 		case SADB_X_EXT_NAT_T_FRAG:
8103 		case SADB_X_EXT_SA_REPLAY:
8104 		case SADB_X_EXT_NEW_ADDRESS_SRC:
8105 		case SADB_X_EXT_NEW_ADDRESS_DST:
8106 			/* duplicate check */
8107 			/*
8108 			 * XXX Are there duplication payloads of either
8109 			 * KEY_AUTH or KEY_ENCRYPT ?
8110 			 */
8111 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
8112 				ipseclog((LOG_DEBUG, "%s: duplicate ext_type "
8113 					"%u\n", __func__, ext->sadb_ext_type));
8114 				m_freem(m);
8115 				PFKEYSTAT_INC(out_dupext);
8116 				return EINVAL;
8117 			}
8118 			break;
8119 		default:
8120 			ipseclog((LOG_DEBUG, "%s: invalid ext_type %u\n",
8121 				__func__, ext->sadb_ext_type));
8122 			m_freem(m);
8123 			PFKEYSTAT_INC(out_invexttype);
8124 			return EINVAL;
8125 		}
8126 
8127 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
8128 
8129 		if (key_validate_ext(ext, extlen)) {
8130 			m_freem(m);
8131 			PFKEYSTAT_INC(out_invlen);
8132 			return EINVAL;
8133 		}
8134 
8135 		n = m_pulldown(m, off, extlen, &toff);
8136 		if (!n) {
8137 			/* m is already freed */
8138 			return ENOBUFS;
8139 		}
8140 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8141 
8142 		mhp->ext[ext->sadb_ext_type] = ext;
8143 		mhp->extoff[ext->sadb_ext_type] = off;
8144 		mhp->extlen[ext->sadb_ext_type] = extlen;
8145 	}
8146 
8147 	if (off != end) {
8148 		m_freem(m);
8149 		PFKEYSTAT_INC(out_invlen);
8150 		return EINVAL;
8151 	}
8152 
8153 	return 0;
8154 }
8155 
8156 static int
8157 key_validate_ext(const struct sadb_ext *ext, int len)
8158 {
8159 	const struct sockaddr *sa;
8160 	enum { NONE, ADDR } checktype = NONE;
8161 	int baselen = 0;
8162 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
8163 
8164 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
8165 		return EINVAL;
8166 
8167 	/* if it does not match minimum/maximum length, bail */
8168 	if (ext->sadb_ext_type >= nitems(minsize) ||
8169 	    ext->sadb_ext_type >= nitems(maxsize))
8170 		return EINVAL;
8171 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
8172 		return EINVAL;
8173 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
8174 		return EINVAL;
8175 
8176 	/* more checks based on sadb_ext_type XXX need more */
8177 	switch (ext->sadb_ext_type) {
8178 	case SADB_EXT_ADDRESS_SRC:
8179 	case SADB_EXT_ADDRESS_DST:
8180 	case SADB_EXT_ADDRESS_PROXY:
8181 	case SADB_X_EXT_NAT_T_OAI:
8182 	case SADB_X_EXT_NAT_T_OAR:
8183 	case SADB_X_EXT_NEW_ADDRESS_SRC:
8184 	case SADB_X_EXT_NEW_ADDRESS_DST:
8185 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
8186 		checktype = ADDR;
8187 		break;
8188 	case SADB_EXT_IDENTITY_SRC:
8189 	case SADB_EXT_IDENTITY_DST:
8190 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
8191 		    SADB_X_IDENTTYPE_ADDR) {
8192 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
8193 			checktype = ADDR;
8194 		} else
8195 			checktype = NONE;
8196 		break;
8197 	default:
8198 		checktype = NONE;
8199 		break;
8200 	}
8201 
8202 	switch (checktype) {
8203 	case NONE:
8204 		break;
8205 	case ADDR:
8206 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8207 		if (len < baselen + sal)
8208 			return EINVAL;
8209 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8210 			return EINVAL;
8211 		break;
8212 	}
8213 
8214 	return 0;
8215 }
8216 
8217 void
8218 spdcache_init(void)
8219 {
8220 	int i;
8221 
8222 	TUNABLE_INT_FETCH("net.key.spdcache.maxentries",
8223 	    &V_key_spdcache_maxentries);
8224 	TUNABLE_INT_FETCH("net.key.spdcache.threshold",
8225 	    &V_key_spdcache_threshold);
8226 
8227 	if (V_key_spdcache_maxentries) {
8228 		V_key_spdcache_maxentries = MAX(V_key_spdcache_maxentries,
8229 		    SPDCACHE_MAX_ENTRIES_PER_HASH);
8230 		V_spdcachehashtbl = hashinit(V_key_spdcache_maxentries /
8231 		    SPDCACHE_MAX_ENTRIES_PER_HASH,
8232 		    M_IPSEC_SPDCACHE, &V_spdcachehash_mask);
8233 		V_key_spdcache_maxentries = (V_spdcachehash_mask + 1)
8234 		    * SPDCACHE_MAX_ENTRIES_PER_HASH;
8235 
8236 		V_spdcache_lock = malloc(sizeof(struct mtx) *
8237 		    (V_spdcachehash_mask + 1),
8238 		    M_IPSEC_SPDCACHE, M_WAITOK|M_ZERO);
8239 
8240 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8241 			SPDCACHE_LOCK_INIT(i);
8242 	}
8243 }
8244 
8245 struct spdcache_entry *
8246 spdcache_entry_alloc(const struct secpolicyindex *spidx, struct secpolicy *sp)
8247 {
8248 	struct spdcache_entry *entry;
8249 
8250 	entry = malloc(sizeof(struct spdcache_entry),
8251 		    M_IPSEC_SPDCACHE, M_NOWAIT|M_ZERO);
8252 	if (entry == NULL)
8253 		return NULL;
8254 
8255 	if (sp != NULL)
8256 		SP_ADDREF(sp);
8257 
8258 	entry->spidx = *spidx;
8259 	entry->sp = sp;
8260 
8261 	return (entry);
8262 }
8263 
8264 void
8265 spdcache_entry_free(struct spdcache_entry *entry)
8266 {
8267 
8268 	if (entry->sp != NULL)
8269 		key_freesp(&entry->sp);
8270 	free(entry, M_IPSEC_SPDCACHE);
8271 }
8272 
8273 void
8274 spdcache_clear(void)
8275 {
8276 	struct spdcache_entry *entry;
8277 	int i;
8278 
8279 	for (i = 0; i < V_spdcachehash_mask + 1; ++i) {
8280 		SPDCACHE_LOCK(i);
8281 		while (!LIST_EMPTY(&V_spdcachehashtbl[i])) {
8282 			entry = LIST_FIRST(&V_spdcachehashtbl[i]);
8283 			LIST_REMOVE(entry, chain);
8284 			spdcache_entry_free(entry);
8285 		}
8286 		SPDCACHE_UNLOCK(i);
8287 	}
8288 }
8289 
8290 #ifdef VIMAGE
8291 void
8292 spdcache_destroy(void)
8293 {
8294 	int i;
8295 
8296 	if (SPDCACHE_ENABLED()) {
8297 		spdcache_clear();
8298 		hashdestroy(V_spdcachehashtbl, M_IPSEC_SPDCACHE, V_spdcachehash_mask);
8299 
8300 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8301 			SPDCACHE_LOCK_DESTROY(i);
8302 
8303 		free(V_spdcache_lock, M_IPSEC_SPDCACHE);
8304 	}
8305 }
8306 #endif
8307 void
8308 key_init(void)
8309 {
8310 	int i;
8311 
8312 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8313 		TAILQ_INIT(&V_sptree[i]);
8314 		TAILQ_INIT(&V_sptree_ifnet[i]);
8315 	}
8316 
8317 	V_key_lft_zone = uma_zcreate("IPsec SA lft_c",
8318 	    sizeof(uint64_t) * 2, NULL, NULL, NULL, NULL,
8319 	    UMA_ALIGN_PTR, UMA_ZONE_PCPU);
8320 
8321 	TAILQ_INIT(&V_sahtree);
8322 	V_sphashtbl = hashinit(SPHASH_NHASH, M_IPSEC_SP, &V_sphash_mask);
8323 	V_savhashtbl = hashinit(SAVHASH_NHASH, M_IPSEC_SA, &V_savhash_mask);
8324 	V_sahaddrhashtbl = hashinit(SAHHASH_NHASH, M_IPSEC_SAH,
8325 	    &V_sahaddrhash_mask);
8326 	V_acqaddrhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8327 	    &V_acqaddrhash_mask);
8328 	V_acqseqhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8329 	    &V_acqseqhash_mask);
8330 
8331 	spdcache_init();
8332 
8333 	for (i = 0; i <= SADB_SATYPE_MAX; i++)
8334 		LIST_INIT(&V_regtree[i]);
8335 
8336 	LIST_INIT(&V_acqtree);
8337 	LIST_INIT(&V_spacqtree);
8338 
8339 	if (!IS_DEFAULT_VNET(curvnet))
8340 		return;
8341 
8342 	SPTREE_LOCK_INIT();
8343 	REGTREE_LOCK_INIT();
8344 	SAHTREE_LOCK_INIT();
8345 	ACQ_LOCK_INIT();
8346 	SPACQ_LOCK_INIT();
8347 
8348 #ifndef IPSEC_DEBUG2
8349 	callout_init(&key_timer, 1);
8350 	callout_reset(&key_timer, hz, key_timehandler, NULL);
8351 #endif /*IPSEC_DEBUG2*/
8352 
8353 	/* initialize key statistics */
8354 	keystat.getspi_count = 1;
8355 
8356 	if (bootverbose)
8357 		printf("IPsec: Initialized Security Association Processing.\n");
8358 }
8359 
8360 #ifdef VIMAGE
8361 void
8362 key_destroy(void)
8363 {
8364 	struct secashead_queue sahdrainq;
8365 	struct secpolicy_queue drainq;
8366 	struct secpolicy *sp, *nextsp;
8367 	struct secacq *acq, *nextacq;
8368 	struct secspacq *spacq, *nextspacq;
8369 	struct secashead *sah;
8370 	struct secasvar *sav;
8371 	struct secreg *reg;
8372 	int i;
8373 
8374 	/*
8375 	 * XXX: can we just call free() for each object without
8376 	 * walking through safe way with releasing references?
8377 	 */
8378 	TAILQ_INIT(&drainq);
8379 	SPTREE_WLOCK();
8380 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8381 		TAILQ_CONCAT(&drainq, &V_sptree[i], chain);
8382 		TAILQ_CONCAT(&drainq, &V_sptree_ifnet[i], chain);
8383 	}
8384 	for (i = 0; i < V_sphash_mask + 1; i++)
8385 		LIST_INIT(&V_sphashtbl[i]);
8386 	SPTREE_WUNLOCK();
8387 	spdcache_destroy();
8388 
8389 	sp = TAILQ_FIRST(&drainq);
8390 	while (sp != NULL) {
8391 		nextsp = TAILQ_NEXT(sp, chain);
8392 		key_freesp(&sp);
8393 		sp = nextsp;
8394 	}
8395 
8396 	TAILQ_INIT(&sahdrainq);
8397 	SAHTREE_WLOCK();
8398 	TAILQ_CONCAT(&sahdrainq, &V_sahtree, chain);
8399 	for (i = 0; i < V_savhash_mask + 1; i++)
8400 		LIST_INIT(&V_savhashtbl[i]);
8401 	for (i = 0; i < V_sahaddrhash_mask + 1; i++)
8402 		LIST_INIT(&V_sahaddrhashtbl[i]);
8403 	TAILQ_FOREACH(sah, &sahdrainq, chain) {
8404 		sah->state = SADB_SASTATE_DEAD;
8405 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
8406 			sav->state = SADB_SASTATE_DEAD;
8407 		}
8408 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
8409 			sav->state = SADB_SASTATE_DEAD;
8410 		}
8411 	}
8412 	SAHTREE_WUNLOCK();
8413 
8414 	key_freesah_flushed(&sahdrainq);
8415 	hashdestroy(V_sphashtbl, M_IPSEC_SP, V_sphash_mask);
8416 	hashdestroy(V_savhashtbl, M_IPSEC_SA, V_savhash_mask);
8417 	hashdestroy(V_sahaddrhashtbl, M_IPSEC_SAH, V_sahaddrhash_mask);
8418 
8419 	REGTREE_LOCK();
8420 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8421 		LIST_FOREACH(reg, &V_regtree[i], chain) {
8422 			if (__LIST_CHAINED(reg)) {
8423 				LIST_REMOVE(reg, chain);
8424 				free(reg, M_IPSEC_SAR);
8425 				break;
8426 			}
8427 		}
8428 	}
8429 	REGTREE_UNLOCK();
8430 
8431 	ACQ_LOCK();
8432 	acq = LIST_FIRST(&V_acqtree);
8433 	while (acq != NULL) {
8434 		nextacq = LIST_NEXT(acq, chain);
8435 		LIST_REMOVE(acq, chain);
8436 		free(acq, M_IPSEC_SAQ);
8437 		acq = nextacq;
8438 	}
8439 	for (i = 0; i < V_acqaddrhash_mask + 1; i++)
8440 		LIST_INIT(&V_acqaddrhashtbl[i]);
8441 	for (i = 0; i < V_acqseqhash_mask + 1; i++)
8442 		LIST_INIT(&V_acqseqhashtbl[i]);
8443 	ACQ_UNLOCK();
8444 
8445 	SPACQ_LOCK();
8446 	for (spacq = LIST_FIRST(&V_spacqtree); spacq != NULL;
8447 	    spacq = nextspacq) {
8448 		nextspacq = LIST_NEXT(spacq, chain);
8449 		if (__LIST_CHAINED(spacq)) {
8450 			LIST_REMOVE(spacq, chain);
8451 			free(spacq, M_IPSEC_SAQ);
8452 		}
8453 	}
8454 	SPACQ_UNLOCK();
8455 	hashdestroy(V_acqaddrhashtbl, M_IPSEC_SAQ, V_acqaddrhash_mask);
8456 	hashdestroy(V_acqseqhashtbl, M_IPSEC_SAQ, V_acqseqhash_mask);
8457 	uma_zdestroy(V_key_lft_zone);
8458 
8459 	if (!IS_DEFAULT_VNET(curvnet))
8460 		return;
8461 #ifndef IPSEC_DEBUG2
8462 	callout_drain(&key_timer);
8463 #endif
8464 	SPTREE_LOCK_DESTROY();
8465 	REGTREE_LOCK_DESTROY();
8466 	SAHTREE_LOCK_DESTROY();
8467 	ACQ_LOCK_DESTROY();
8468 	SPACQ_LOCK_DESTROY();
8469 }
8470 #endif
8471 
8472 /* record data transfer on SA, and update timestamps */
8473 void
8474 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8475 {
8476 	IPSEC_ASSERT(sav != NULL, ("Null secasvar"));
8477 	IPSEC_ASSERT(m != NULL, ("Null mbuf"));
8478 
8479 	/*
8480 	 * XXX Currently, there is a difference of bytes size
8481 	 * between inbound and outbound processing.
8482 	 */
8483 	counter_u64_add(sav->lft_c_bytes, m->m_pkthdr.len);
8484 
8485 	/*
8486 	 * We use the number of packets as the unit of
8487 	 * allocations.  We increment the variable
8488 	 * whenever {esp,ah}_{in,out}put is called.
8489 	 */
8490 	counter_u64_add(sav->lft_c_allocations, 1);
8491 
8492 	/*
8493 	 * NOTE: We record CURRENT usetime by using wall clock,
8494 	 * in seconds.  HARD and SOFT lifetime are measured by the time
8495 	 * difference (again in seconds) from usetime.
8496 	 *
8497 	 *	usetime
8498 	 *	v     expire   expire
8499 	 * -----+-----+--------+---> t
8500 	 *	<--------------> HARD
8501 	 *	<-----> SOFT
8502 	 */
8503 	if (sav->firstused == 0)
8504 		sav->firstused = time_second;
8505 }
8506 
8507 /*
8508  * Take one of the kernel's security keys and convert it into a PF_KEY
8509  * structure within an mbuf, suitable for sending up to a waiting
8510  * application in user land.
8511  *
8512  * IN:
8513  *    src: A pointer to a kernel security key.
8514  *    exttype: Which type of key this is. Refer to the PF_KEY data structures.
8515  * OUT:
8516  *    a valid mbuf or NULL indicating an error
8517  *
8518  */
8519 
8520 static struct mbuf *
8521 key_setkey(struct seckey *src, uint16_t exttype)
8522 {
8523 	struct mbuf *m;
8524 	struct sadb_key *p;
8525 	int len;
8526 
8527 	if (src == NULL)
8528 		return NULL;
8529 
8530 	len = PFKEY_ALIGN8(sizeof(struct sadb_key) + _KEYLEN(src));
8531 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8532 	if (m == NULL)
8533 		return NULL;
8534 	m_align(m, len);
8535 	m->m_len = len;
8536 	p = mtod(m, struct sadb_key *);
8537 	bzero(p, len);
8538 	p->sadb_key_len = PFKEY_UNIT64(len);
8539 	p->sadb_key_exttype = exttype;
8540 	p->sadb_key_bits = src->bits;
8541 	bcopy(src->key_data, _KEYBUF(p), _KEYLEN(src));
8542 
8543 	return m;
8544 }
8545 
8546 /*
8547  * Take one of the kernel's lifetime data structures and convert it
8548  * into a PF_KEY structure within an mbuf, suitable for sending up to
8549  * a waiting application in user land.
8550  *
8551  * IN:
8552  *    src: A pointer to a kernel lifetime structure.
8553  *    exttype: Which type of lifetime this is. Refer to the PF_KEY
8554  *             data structures for more information.
8555  * OUT:
8556  *    a valid mbuf or NULL indicating an error
8557  *
8558  */
8559 
8560 static struct mbuf *
8561 key_setlifetime(struct seclifetime *src, uint16_t exttype)
8562 {
8563 	struct mbuf *m = NULL;
8564 	struct sadb_lifetime *p;
8565 	int len = PFKEY_ALIGN8(sizeof(struct sadb_lifetime));
8566 
8567 	if (src == NULL)
8568 		return NULL;
8569 
8570 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8571 	if (m == NULL)
8572 		return m;
8573 	m_align(m, len);
8574 	m->m_len = len;
8575 	p = mtod(m, struct sadb_lifetime *);
8576 
8577 	bzero(p, len);
8578 	p->sadb_lifetime_len = PFKEY_UNIT64(len);
8579 	p->sadb_lifetime_exttype = exttype;
8580 	p->sadb_lifetime_allocations = src->allocations;
8581 	p->sadb_lifetime_bytes = src->bytes;
8582 	p->sadb_lifetime_addtime = src->addtime;
8583 	p->sadb_lifetime_usetime = src->usetime;
8584 
8585 	return m;
8586 
8587 }
8588 
8589 const struct enc_xform *
8590 enc_algorithm_lookup(int alg)
8591 {
8592 	int i;
8593 
8594 	for (i = 0; i < nitems(supported_ealgs); i++)
8595 		if (alg == supported_ealgs[i].sadb_alg)
8596 			return (supported_ealgs[i].xform);
8597 	return (NULL);
8598 }
8599 
8600 const struct auth_hash *
8601 auth_algorithm_lookup(int alg)
8602 {
8603 	int i;
8604 
8605 	for (i = 0; i < nitems(supported_aalgs); i++)
8606 		if (alg == supported_aalgs[i].sadb_alg)
8607 			return (supported_aalgs[i].xform);
8608 	return (NULL);
8609 }
8610 
8611 const struct comp_algo *
8612 comp_algorithm_lookup(int alg)
8613 {
8614 	int i;
8615 
8616 	for (i = 0; i < nitems(supported_calgs); i++)
8617 		if (alg == supported_calgs[i].sadb_alg)
8618 			return (supported_calgs[i].xform);
8619 	return (NULL);
8620 }
8621 
8622