xref: /freebsd/sys/netipsec/key.c (revision 94a82666846d62cdff7d78f78d428df35412e50d)
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\n",
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,
512     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
513     "SPD cache");
514 
515 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, maxentries,
516 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_maxentries), 0,
517 	"Maximum number of entries in the SPD cache"
518 	" (power of 2, 0 to disable)");
519 
520 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, threshold,
521 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_threshold), 0,
522 	"Number of SPs that make the SPD cache active");
523 
524 #define __LIST_CHAINED(elm) \
525 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
526 
527 MALLOC_DEFINE(M_IPSEC_SA, "secasvar", "ipsec security association");
528 MALLOC_DEFINE(M_IPSEC_SAH, "sahead", "ipsec sa head");
529 MALLOC_DEFINE(M_IPSEC_SP, "ipsecpolicy", "ipsec security policy");
530 MALLOC_DEFINE(M_IPSEC_SR, "ipsecrequest", "ipsec security request");
531 MALLOC_DEFINE(M_IPSEC_MISC, "ipsec-misc", "ipsec miscellaneous");
532 MALLOC_DEFINE(M_IPSEC_SAQ, "ipsec-saq", "ipsec sa acquire");
533 MALLOC_DEFINE(M_IPSEC_SAR, "ipsec-reg", "ipsec sa acquire");
534 MALLOC_DEFINE(M_IPSEC_SPDCACHE, "ipsec-spdcache", "ipsec SPD cache");
535 
536 VNET_DEFINE_STATIC(uma_zone_t, key_lft_zone);
537 #define	V_key_lft_zone		VNET(key_lft_zone)
538 
539 /*
540  * set parameters into secpolicyindex buffer.
541  * Must allocate secpolicyindex buffer passed to this function.
542  */
543 #define KEY_SETSECSPIDX(_dir, s, d, ps, pd, ulp, idx) \
544 do { \
545 	bzero((idx), sizeof(struct secpolicyindex));                         \
546 	(idx)->dir = (_dir);                                                 \
547 	(idx)->prefs = (ps);                                                 \
548 	(idx)->prefd = (pd);                                                 \
549 	(idx)->ul_proto = (ulp);                                             \
550 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
551 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
552 } while (0)
553 
554 /*
555  * set parameters into secasindex buffer.
556  * Must allocate secasindex buffer before calling this function.
557  */
558 #define KEY_SETSECASIDX(p, m, r, s, d, idx) \
559 do { \
560 	bzero((idx), sizeof(struct secasindex));                             \
561 	(idx)->proto = (p);                                                  \
562 	(idx)->mode = (m);                                                   \
563 	(idx)->reqid = (r);                                                  \
564 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
565 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
566 	key_porttosaddr(&(idx)->src.sa, 0);				     \
567 	key_porttosaddr(&(idx)->dst.sa, 0);				     \
568 } while (0)
569 
570 /* key statistics */
571 struct _keystat {
572 	u_long getspi_count; /* the avarage of count to try to get new SPI */
573 } keystat;
574 
575 struct sadb_msghdr {
576 	struct sadb_msg *msg;
577 	struct sadb_ext *ext[SADB_EXT_MAX + 1];
578 	int extoff[SADB_EXT_MAX + 1];
579 	int extlen[SADB_EXT_MAX + 1];
580 };
581 
582 static struct supported_ealgs {
583 	int sadb_alg;
584 	const struct enc_xform *xform;
585 } supported_ealgs[] = {
586 	{ SADB_X_EALG_AES,		&enc_xform_rijndael128 },
587 	{ SADB_EALG_NULL,		&enc_xform_null },
588 	{ SADB_X_EALG_AESCTR,		&enc_xform_aes_icm },
589 	{ SADB_X_EALG_AESGCM16,		&enc_xform_aes_nist_gcm },
590 	{ SADB_X_EALG_AESGMAC,		&enc_xform_aes_nist_gmac },
591 };
592 
593 static struct supported_aalgs {
594 	int sadb_alg;
595 	const struct auth_hash *xform;
596 } supported_aalgs[] = {
597 	{ SADB_X_AALG_NULL,		&auth_hash_null },
598 	{ SADB_AALG_SHA1HMAC,		&auth_hash_hmac_sha1 },
599 	{ SADB_X_AALG_SHA2_256,		&auth_hash_hmac_sha2_256 },
600 	{ SADB_X_AALG_SHA2_384,		&auth_hash_hmac_sha2_384 },
601 	{ SADB_X_AALG_SHA2_512,		&auth_hash_hmac_sha2_512 },
602 	{ SADB_X_AALG_AES128GMAC,	&auth_hash_nist_gmac_aes_128 },
603 	{ SADB_X_AALG_AES192GMAC,	&auth_hash_nist_gmac_aes_192 },
604 	{ SADB_X_AALG_AES256GMAC,	&auth_hash_nist_gmac_aes_256 },
605 };
606 
607 static struct supported_calgs {
608 	int sadb_alg;
609 	const struct comp_algo *xform;
610 } supported_calgs[] = {
611 	{ SADB_X_CALG_DEFLATE,		&comp_algo_deflate },
612 };
613 
614 #ifndef IPSEC_DEBUG2
615 static struct callout key_timer;
616 #endif
617 
618 static void key_unlink(struct secpolicy *);
619 static struct secpolicy *key_do_allocsp(struct secpolicyindex *spidx, u_int dir);
620 static struct secpolicy *key_getsp(struct secpolicyindex *);
621 static struct secpolicy *key_getspbyid(u_int32_t);
622 static struct mbuf *key_gather_mbuf(struct mbuf *,
623 	const struct sadb_msghdr *, int, int, ...);
624 static int key_spdadd(struct socket *, struct mbuf *,
625 	const struct sadb_msghdr *);
626 static uint32_t key_getnewspid(void);
627 static int key_spddelete(struct socket *, struct mbuf *,
628 	const struct sadb_msghdr *);
629 static int key_spddelete2(struct socket *, struct mbuf *,
630 	const struct sadb_msghdr *);
631 static int key_spdget(struct socket *, struct mbuf *,
632 	const struct sadb_msghdr *);
633 static int key_spdflush(struct socket *, struct mbuf *,
634 	const struct sadb_msghdr *);
635 static int key_spddump(struct socket *, struct mbuf *,
636 	const struct sadb_msghdr *);
637 static struct mbuf *key_setdumpsp(struct secpolicy *,
638 	u_int8_t, u_int32_t, u_int32_t);
639 static struct mbuf *key_sp2mbuf(struct secpolicy *);
640 static size_t key_getspreqmsglen(struct secpolicy *);
641 static int key_spdexpire(struct secpolicy *);
642 static struct secashead *key_newsah(struct secasindex *);
643 static void key_freesah(struct secashead **);
644 static void key_delsah(struct secashead *);
645 static struct secasvar *key_newsav(const struct sadb_msghdr *,
646     struct secasindex *, uint32_t, int *);
647 static void key_delsav(struct secasvar *);
648 static void key_unlinksav(struct secasvar *);
649 static struct secashead *key_getsah(struct secasindex *);
650 static int key_checkspidup(uint32_t);
651 static struct secasvar *key_getsavbyspi(uint32_t);
652 static int key_setnatt(struct secasvar *, const struct sadb_msghdr *);
653 static int key_setsaval(struct secasvar *, const struct sadb_msghdr *);
654 static int key_updatelifetimes(struct secasvar *, const struct sadb_msghdr *);
655 static int key_updateaddresses(struct socket *, struct mbuf *,
656     const struct sadb_msghdr *, struct secasvar *, struct secasindex *);
657 
658 static struct mbuf *key_setdumpsa(struct secasvar *, u_int8_t,
659 	u_int8_t, u_int32_t, u_int32_t);
660 static struct mbuf *key_setsadbmsg(u_int8_t, u_int16_t, u_int8_t,
661 	u_int32_t, pid_t, u_int16_t);
662 static struct mbuf *key_setsadbsa(struct secasvar *);
663 static struct mbuf *key_setsadbaddr(u_int16_t,
664 	const struct sockaddr *, u_int8_t, u_int16_t);
665 static struct mbuf *key_setsadbxport(u_int16_t, u_int16_t);
666 static struct mbuf *key_setsadbxtype(u_int16_t);
667 static struct mbuf *key_setsadbxsa2(u_int8_t, u_int32_t, u_int32_t);
668 static struct mbuf *key_setsadbxsareplay(u_int32_t);
669 static struct mbuf *key_setsadbxpolicy(u_int16_t, u_int8_t,
670 	u_int32_t, u_int32_t);
671 static struct seckey *key_dup_keymsg(const struct sadb_key *, size_t,
672     struct malloc_type *);
673 static struct seclifetime *key_dup_lifemsg(const struct sadb_lifetime *src,
674     struct malloc_type *);
675 
676 /* flags for key_cmpsaidx() */
677 #define CMP_HEAD	1	/* protocol, addresses. */
678 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
679 #define CMP_REQID	3	/* additionally HEAD, reaid. */
680 #define CMP_EXACTLY	4	/* all elements. */
681 static int key_cmpsaidx(const struct secasindex *,
682     const struct secasindex *, int);
683 static int key_cmpspidx_exactly(struct secpolicyindex *,
684     struct secpolicyindex *);
685 static int key_cmpspidx_withmask(struct secpolicyindex *,
686     struct secpolicyindex *);
687 static int key_bbcmp(const void *, const void *, u_int);
688 static uint8_t key_satype2proto(uint8_t);
689 static uint8_t key_proto2satype(uint8_t);
690 
691 static int key_getspi(struct socket *, struct mbuf *,
692 	const struct sadb_msghdr *);
693 static uint32_t key_do_getnewspi(struct sadb_spirange *, struct secasindex *);
694 static int key_update(struct socket *, struct mbuf *,
695 	const struct sadb_msghdr *);
696 static int key_add(struct socket *, struct mbuf *,
697 	const struct sadb_msghdr *);
698 static int key_setident(struct secashead *, const struct sadb_msghdr *);
699 static struct mbuf *key_getmsgbuf_x1(struct mbuf *,
700 	const struct sadb_msghdr *);
701 static int key_delete(struct socket *, struct mbuf *,
702 	const struct sadb_msghdr *);
703 static int key_delete_all(struct socket *, struct mbuf *,
704 	const struct sadb_msghdr *, struct secasindex *);
705 static int key_get(struct socket *, struct mbuf *,
706 	const struct sadb_msghdr *);
707 
708 static void key_getcomb_setlifetime(struct sadb_comb *);
709 static struct mbuf *key_getcomb_ealg(void);
710 static struct mbuf *key_getcomb_ah(void);
711 static struct mbuf *key_getcomb_ipcomp(void);
712 static struct mbuf *key_getprop(const struct secasindex *);
713 
714 static int key_acquire(const struct secasindex *, struct secpolicy *);
715 static uint32_t key_newacq(const struct secasindex *, int *);
716 static uint32_t key_getacq(const struct secasindex *, int *);
717 static int key_acqdone(const struct secasindex *, uint32_t);
718 static int key_acqreset(uint32_t);
719 static struct secspacq *key_newspacq(struct secpolicyindex *);
720 static struct secspacq *key_getspacq(struct secpolicyindex *);
721 static int key_acquire2(struct socket *, struct mbuf *,
722 	const struct sadb_msghdr *);
723 static int key_register(struct socket *, struct mbuf *,
724 	const struct sadb_msghdr *);
725 static int key_expire(struct secasvar *, int);
726 static int key_flush(struct socket *, struct mbuf *,
727 	const struct sadb_msghdr *);
728 static int key_dump(struct socket *, struct mbuf *,
729 	const struct sadb_msghdr *);
730 static int key_promisc(struct socket *, struct mbuf *,
731 	const struct sadb_msghdr *);
732 static int key_senderror(struct socket *, struct mbuf *, int);
733 static int key_validate_ext(const struct sadb_ext *, int);
734 static int key_align(struct mbuf *, struct sadb_msghdr *);
735 static struct mbuf *key_setlifetime(struct seclifetime *, uint16_t);
736 static struct mbuf *key_setkey(struct seckey *, uint16_t);
737 
738 static void spdcache_init(void);
739 static void spdcache_clear(void);
740 static struct spdcache_entry *spdcache_entry_alloc(
741 	const struct secpolicyindex *spidx,
742 	struct secpolicy *policy);
743 static void spdcache_entry_free(struct spdcache_entry *entry);
744 #ifdef VIMAGE
745 static void spdcache_destroy(void);
746 #endif
747 
748 #define	DBG_IPSEC_INITREF(t, p)	do {				\
749 	refcount_init(&(p)->refcnt, 1);				\
750 	KEYDBG(KEY_STAMP,					\
751 	    printf("%s: Initialize refcnt %s(%p) = %u\n",	\
752 	    __func__, #t, (p), (p)->refcnt));			\
753 } while (0)
754 #define	DBG_IPSEC_ADDREF(t, p)	do {				\
755 	refcount_acquire(&(p)->refcnt);				\
756 	KEYDBG(KEY_STAMP,					\
757 	    printf("%s: Acquire refcnt %s(%p) -> %u\n",		\
758 	    __func__, #t, (p), (p)->refcnt));			\
759 } while (0)
760 #define	DBG_IPSEC_DELREF(t, p)	do {				\
761 	KEYDBG(KEY_STAMP,					\
762 	    printf("%s: Release refcnt %s(%p) -> %u\n",		\
763 	    __func__, #t, (p), (p)->refcnt - 1));		\
764 	refcount_release(&(p)->refcnt);				\
765 } while (0)
766 
767 #define	IPSEC_INITREF(t, p)	refcount_init(&(p)->refcnt, 1)
768 #define	IPSEC_ADDREF(t, p)	refcount_acquire(&(p)->refcnt)
769 #define	IPSEC_DELREF(t, p)	refcount_release(&(p)->refcnt)
770 
771 #define	SP_INITREF(p)	IPSEC_INITREF(SP, p)
772 #define	SP_ADDREF(p)	IPSEC_ADDREF(SP, p)
773 #define	SP_DELREF(p)	IPSEC_DELREF(SP, p)
774 
775 #define	SAH_INITREF(p)	IPSEC_INITREF(SAH, p)
776 #define	SAH_ADDREF(p)	IPSEC_ADDREF(SAH, p)
777 #define	SAH_DELREF(p)	IPSEC_DELREF(SAH, p)
778 
779 #define	SAV_INITREF(p)	IPSEC_INITREF(SAV, p)
780 #define	SAV_ADDREF(p)	IPSEC_ADDREF(SAV, p)
781 #define	SAV_DELREF(p)	IPSEC_DELREF(SAV, p)
782 
783 /*
784  * Update the refcnt while holding the SPTREE lock.
785  */
786 void
787 key_addref(struct secpolicy *sp)
788 {
789 
790 	SP_ADDREF(sp);
791 }
792 
793 /*
794  * Return 0 when there are known to be no SP's for the specified
795  * direction.  Otherwise return 1.  This is used by IPsec code
796  * to optimize performance.
797  */
798 int
799 key_havesp(u_int dir)
800 {
801 
802 	return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ?
803 		TAILQ_FIRST(&V_sptree[dir]) != NULL : 1);
804 }
805 
806 /* %%% IPsec policy management */
807 /*
808  * Return current SPDB generation.
809  */
810 uint32_t
811 key_getspgen(void)
812 {
813 
814 	return (V_sp_genid);
815 }
816 
817 void
818 key_bumpspgen(void)
819 {
820 
821 	V_sp_genid++;
822 }
823 
824 static int
825 key_checksockaddrs(struct sockaddr *src, struct sockaddr *dst)
826 {
827 
828 	/* family match */
829 	if (src->sa_family != dst->sa_family)
830 		return (EINVAL);
831 	/* sa_len match */
832 	if (src->sa_len != dst->sa_len)
833 		return (EINVAL);
834 	switch (src->sa_family) {
835 #ifdef INET
836 	case AF_INET:
837 		if (src->sa_len != sizeof(struct sockaddr_in))
838 			return (EINVAL);
839 		break;
840 #endif
841 #ifdef INET6
842 	case AF_INET6:
843 		if (src->sa_len != sizeof(struct sockaddr_in6))
844 			return (EINVAL);
845 		break;
846 #endif
847 	default:
848 		return (EAFNOSUPPORT);
849 	}
850 	return (0);
851 }
852 
853 struct secpolicy *
854 key_do_allocsp(struct secpolicyindex *spidx, u_int dir)
855 {
856 	SPTREE_RLOCK_TRACKER;
857 	struct secpolicy *sp;
858 
859 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
860 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
861 		("invalid direction %u", dir));
862 
863 	SPTREE_RLOCK();
864 	TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
865 		if (key_cmpspidx_withmask(&sp->spidx, spidx)) {
866 			SP_ADDREF(sp);
867 			break;
868 		}
869 	}
870 	SPTREE_RUNLOCK();
871 	return (sp);
872 }
873 
874 
875 /*
876  * allocating a SP for OUTBOUND or INBOUND packet.
877  * Must call key_freesp() later.
878  * OUT:	NULL:	not found
879  *	others:	found and return the pointer.
880  */
881 struct secpolicy *
882 key_allocsp(struct secpolicyindex *spidx, u_int dir)
883 {
884 	struct spdcache_entry *entry, *lastentry, *tmpentry;
885 	struct secpolicy *sp;
886 	uint32_t hashv;
887 	int nb_entries;
888 
889 	if (!SPDCACHE_ACTIVE()) {
890 		sp = key_do_allocsp(spidx, dir);
891 		goto out;
892 	}
893 
894 	hashv = SPDCACHE_HASHVAL(spidx);
895 	SPDCACHE_LOCK(hashv);
896 	nb_entries = 0;
897 	LIST_FOREACH_SAFE(entry, &V_spdcachehashtbl[hashv], chain, tmpentry) {
898 		/* Removed outdated entries */
899 		if (entry->sp != NULL &&
900 		    entry->sp->state == IPSEC_SPSTATE_DEAD) {
901 			LIST_REMOVE(entry, chain);
902 			spdcache_entry_free(entry);
903 			continue;
904 		}
905 
906 		nb_entries++;
907 		if (!key_cmpspidx_exactly(&entry->spidx, spidx)) {
908 			lastentry = entry;
909 			continue;
910 		}
911 
912 		sp = entry->sp;
913 		if (entry->sp != NULL)
914 			SP_ADDREF(sp);
915 
916 		/* IPSECSTAT_INC(ips_spdcache_hits); */
917 
918 		SPDCACHE_UNLOCK(hashv);
919 		goto out;
920 	}
921 
922 	/* IPSECSTAT_INC(ips_spdcache_misses); */
923 
924 	sp = key_do_allocsp(spidx, dir);
925 	entry = spdcache_entry_alloc(spidx, sp);
926 	if (entry != NULL) {
927 		if (nb_entries >= SPDCACHE_MAX_ENTRIES_PER_HASH) {
928 			LIST_REMOVE(lastentry, chain);
929 			spdcache_entry_free(lastentry);
930 		}
931 
932 		LIST_INSERT_HEAD(&V_spdcachehashtbl[hashv], entry, chain);
933 	}
934 
935 	SPDCACHE_UNLOCK(hashv);
936 
937 out:
938 	if (sp != NULL) {	/* found a SPD entry */
939 		sp->lastused = time_second;
940 		KEYDBG(IPSEC_STAMP,
941 		    printf("%s: return SP(%p)\n", __func__, sp));
942 		KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
943 	} else {
944 		KEYDBG(IPSEC_DATA,
945 		    printf("%s: lookup failed for ", __func__);
946 		    kdebug_secpolicyindex(spidx, NULL));
947 	}
948 	return (sp);
949 }
950 
951 /*
952  * Allocating an SA entry for an *INBOUND* or *OUTBOUND* TCP packet, signed
953  * or should be signed by MD5 signature.
954  * We don't use key_allocsa() for such lookups, because we don't know SPI.
955  * Unlike ESP and AH protocols, SPI isn't transmitted in the TCP header with
956  * signed packet. We use SADB only as storage for password.
957  * OUT:	positive:	corresponding SA for given saidx found.
958  *	NULL:		SA not found
959  */
960 struct secasvar *
961 key_allocsa_tcpmd5(struct secasindex *saidx)
962 {
963 	SAHTREE_RLOCK_TRACKER;
964 	struct secashead *sah;
965 	struct secasvar *sav;
966 
967 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP,
968 	    ("unexpected security protocol %u", saidx->proto));
969 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TCPMD5,
970 	    ("unexpected mode %u", saidx->mode));
971 
972 	SAHTREE_RLOCK();
973 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
974 		KEYDBG(IPSEC_DUMP,
975 		    printf("%s: checking SAH\n", __func__);
976 		    kdebug_secash(sah, "  "));
977 		if (sah->saidx.proto != IPPROTO_TCP)
978 			continue;
979 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
980 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
981 			break;
982 	}
983 	if (sah != NULL) {
984 		if (V_key_preferred_oldsa)
985 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
986 		else
987 			sav = TAILQ_FIRST(&sah->savtree_alive);
988 		if (sav != NULL)
989 			SAV_ADDREF(sav);
990 	} else
991 		sav = NULL;
992 	SAHTREE_RUNLOCK();
993 
994 	if (sav != NULL) {
995 		KEYDBG(IPSEC_STAMP,
996 		    printf("%s: return SA(%p)\n", __func__, sav));
997 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
998 	} else {
999 		KEYDBG(IPSEC_STAMP,
1000 		    printf("%s: SA not found\n", __func__));
1001 		KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1002 	}
1003 	return (sav);
1004 }
1005 
1006 /*
1007  * Allocating an SA entry for an *OUTBOUND* packet.
1008  * OUT:	positive:	corresponding SA for given saidx found.
1009  *	NULL:		SA not found, but will be acquired, check *error
1010  *			for acquiring status.
1011  */
1012 struct secasvar *
1013 key_allocsa_policy(struct secpolicy *sp, const struct secasindex *saidx,
1014     int *error)
1015 {
1016 	SAHTREE_RLOCK_TRACKER;
1017 	struct secashead *sah;
1018 	struct secasvar *sav;
1019 
1020 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
1021 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TRANSPORT ||
1022 		saidx->mode == IPSEC_MODE_TUNNEL,
1023 		("unexpected policy %u", saidx->mode));
1024 
1025 	/*
1026 	 * We check new SA in the IPsec request because a different
1027 	 * SA may be involved each time this request is checked, either
1028 	 * because new SAs are being configured, or this request is
1029 	 * associated with an unconnected datagram socket, or this request
1030 	 * is associated with a system default policy.
1031 	 */
1032 	SAHTREE_RLOCK();
1033 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
1034 		KEYDBG(IPSEC_DUMP,
1035 		    printf("%s: checking SAH\n", __func__);
1036 		    kdebug_secash(sah, "  "));
1037 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID))
1038 			break;
1039 
1040 	}
1041 	if (sah != NULL) {
1042 		/*
1043 		 * Allocate the oldest SA available according to
1044 		 * draft-jenkins-ipsec-rekeying-03.
1045 		 */
1046 		if (V_key_preferred_oldsa)
1047 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1048 		else
1049 			sav = TAILQ_FIRST(&sah->savtree_alive);
1050 		if (sav != NULL)
1051 			SAV_ADDREF(sav);
1052 	} else
1053 		sav = NULL;
1054 	SAHTREE_RUNLOCK();
1055 
1056 	if (sav != NULL) {
1057 		*error = 0;
1058 		KEYDBG(IPSEC_STAMP,
1059 		    printf("%s: chosen SA(%p) for SP(%p)\n", __func__,
1060 			sav, sp));
1061 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1062 		return (sav); /* return referenced SA */
1063 	}
1064 
1065 	/* there is no SA */
1066 	*error = key_acquire(saidx, sp);
1067 	if ((*error) != 0)
1068 		ipseclog((LOG_DEBUG,
1069 		    "%s: error %d returned from key_acquire()\n",
1070 			__func__, *error));
1071 	KEYDBG(IPSEC_STAMP,
1072 	    printf("%s: acquire SA for SP(%p), error %d\n",
1073 		__func__, sp, *error));
1074 	KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1075 	return (NULL);
1076 }
1077 
1078 /*
1079  * allocating a usable SA entry for a *INBOUND* packet.
1080  * Must call key_freesav() later.
1081  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
1082  *	NULL:		not found, or error occurred.
1083  *
1084  * According to RFC 2401 SA is uniquely identified by a triple SPI,
1085  * destination address, and security protocol. But according to RFC 4301,
1086  * SPI by itself suffices to specify an SA.
1087  *
1088  * Note that, however, we do need to keep source address in IPsec SA.
1089  * IKE specification and PF_KEY specification do assume that we
1090  * keep source address in IPsec SA.  We see a tricky situation here.
1091  */
1092 struct secasvar *
1093 key_allocsa(union sockaddr_union *dst, uint8_t proto, uint32_t spi)
1094 {
1095 	SAHTREE_RLOCK_TRACKER;
1096 	struct secasvar *sav;
1097 
1098 	IPSEC_ASSERT(proto == IPPROTO_ESP || proto == IPPROTO_AH ||
1099 	    proto == IPPROTO_IPCOMP, ("unexpected security protocol %u",
1100 	    proto));
1101 
1102 	SAHTREE_RLOCK();
1103 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
1104 		if (sav->spi == spi)
1105 			break;
1106 	}
1107 	/*
1108 	 * We use single SPI namespace for all protocols, so it is
1109 	 * impossible to have SPI duplicates in the SAVHASH.
1110 	 */
1111 	if (sav != NULL) {
1112 		if (sav->state != SADB_SASTATE_LARVAL &&
1113 		    sav->sah->saidx.proto == proto &&
1114 		    key_sockaddrcmp(&dst->sa,
1115 			&sav->sah->saidx.dst.sa, 0) == 0)
1116 			SAV_ADDREF(sav);
1117 		else
1118 			sav = NULL;
1119 	}
1120 	SAHTREE_RUNLOCK();
1121 
1122 	if (sav == NULL) {
1123 		KEYDBG(IPSEC_STAMP,
1124 		    char buf[IPSEC_ADDRSTRLEN];
1125 		    printf("%s: SA not found for spi %u proto %u dst %s\n",
1126 			__func__, ntohl(spi), proto, ipsec_address(dst, buf,
1127 			sizeof(buf))));
1128 	} else {
1129 		KEYDBG(IPSEC_STAMP,
1130 		    printf("%s: return SA(%p)\n", __func__, sav));
1131 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1132 	}
1133 	return (sav);
1134 }
1135 
1136 struct secasvar *
1137 key_allocsa_tunnel(union sockaddr_union *src, union sockaddr_union *dst,
1138     uint8_t proto)
1139 {
1140 	SAHTREE_RLOCK_TRACKER;
1141 	struct secasindex saidx;
1142 	struct secashead *sah;
1143 	struct secasvar *sav;
1144 
1145 	IPSEC_ASSERT(src != NULL, ("null src address"));
1146 	IPSEC_ASSERT(dst != NULL, ("null dst address"));
1147 
1148 	KEY_SETSECASIDX(proto, IPSEC_MODE_TUNNEL, 0, &src->sa,
1149 	    &dst->sa, &saidx);
1150 
1151 	sav = NULL;
1152 	SAHTREE_RLOCK();
1153 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
1154 		if (IPSEC_MODE_TUNNEL != sah->saidx.mode)
1155 			continue;
1156 		if (proto != sah->saidx.proto)
1157 			continue;
1158 		if (key_sockaddrcmp(&src->sa, &sah->saidx.src.sa, 0) != 0)
1159 			continue;
1160 		if (key_sockaddrcmp(&dst->sa, &sah->saidx.dst.sa, 0) != 0)
1161 			continue;
1162 		/* XXXAE: is key_preferred_oldsa reasonably?*/
1163 		if (V_key_preferred_oldsa)
1164 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1165 		else
1166 			sav = TAILQ_FIRST(&sah->savtree_alive);
1167 		if (sav != NULL) {
1168 			SAV_ADDREF(sav);
1169 			break;
1170 		}
1171 	}
1172 	SAHTREE_RUNLOCK();
1173 	KEYDBG(IPSEC_STAMP,
1174 	    printf("%s: return SA(%p)\n", __func__, sav));
1175 	if (sav != NULL)
1176 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1177 	return (sav);
1178 }
1179 
1180 /*
1181  * Must be called after calling key_allocsp().
1182  */
1183 void
1184 key_freesp(struct secpolicy **spp)
1185 {
1186 	struct secpolicy *sp = *spp;
1187 
1188 	IPSEC_ASSERT(sp != NULL, ("null sp"));
1189 	if (SP_DELREF(sp) == 0)
1190 		return;
1191 
1192 	KEYDBG(IPSEC_STAMP,
1193 	    printf("%s: last reference to SP(%p)\n", __func__, sp));
1194 	KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1195 
1196 	*spp = NULL;
1197 	while (sp->tcount > 0)
1198 		ipsec_delisr(sp->req[--sp->tcount]);
1199 	free(sp, M_IPSEC_SP);
1200 }
1201 
1202 static void
1203 key_unlink(struct secpolicy *sp)
1204 {
1205 
1206 	IPSEC_ASSERT(sp->spidx.dir == IPSEC_DIR_INBOUND ||
1207 	    sp->spidx.dir == IPSEC_DIR_OUTBOUND,
1208 	    ("invalid direction %u", sp->spidx.dir));
1209 	SPTREE_UNLOCK_ASSERT();
1210 
1211 	KEYDBG(KEY_STAMP,
1212 	    printf("%s: SP(%p)\n", __func__, sp));
1213 	SPTREE_WLOCK();
1214 	if (sp->state != IPSEC_SPSTATE_ALIVE) {
1215 		/* SP is already unlinked */
1216 		SPTREE_WUNLOCK();
1217 		return;
1218 	}
1219 	sp->state = IPSEC_SPSTATE_DEAD;
1220 	TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
1221 	V_spd_size--;
1222 	LIST_REMOVE(sp, idhash);
1223 	V_sp_genid++;
1224 	SPTREE_WUNLOCK();
1225 	if (SPDCACHE_ENABLED())
1226 		spdcache_clear();
1227 	key_freesp(&sp);
1228 }
1229 
1230 /*
1231  * insert a secpolicy into the SP database. Lower priorities first
1232  */
1233 static void
1234 key_insertsp(struct secpolicy *newsp)
1235 {
1236 	struct secpolicy *sp;
1237 
1238 	SPTREE_WLOCK_ASSERT();
1239 	TAILQ_FOREACH(sp, &V_sptree[newsp->spidx.dir], chain) {
1240 		if (newsp->priority < sp->priority) {
1241 			TAILQ_INSERT_BEFORE(sp, newsp, chain);
1242 			goto done;
1243 		}
1244 	}
1245 	TAILQ_INSERT_TAIL(&V_sptree[newsp->spidx.dir], newsp, chain);
1246 done:
1247 	LIST_INSERT_HEAD(SPHASH_HASH(newsp->id), newsp, idhash);
1248 	newsp->state = IPSEC_SPSTATE_ALIVE;
1249 	V_spd_size++;
1250 	V_sp_genid++;
1251 }
1252 
1253 /*
1254  * Insert a bunch of VTI secpolicies into the SPDB.
1255  * We keep VTI policies in the separate list due to following reasons:
1256  * 1) they should be immutable to user's or some deamon's attempts to
1257  *    delete. The only way delete such policies - destroy or unconfigure
1258  *    corresponding virtual inteface.
1259  * 2) such policies have traffic selector that matches all traffic per
1260  *    address family.
1261  * Since all VTI policies have the same priority, we don't care about
1262  * policies order.
1263  */
1264 int
1265 key_register_ifnet(struct secpolicy **spp, u_int count)
1266 {
1267 	struct mbuf *m;
1268 	u_int i;
1269 
1270 	SPTREE_WLOCK();
1271 	/*
1272 	 * First of try to acquire id for each SP.
1273 	 */
1274 	for (i = 0; i < count; i++) {
1275 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1276 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1277 		    ("invalid direction %u", spp[i]->spidx.dir));
1278 
1279 		if ((spp[i]->id = key_getnewspid()) == 0) {
1280 			SPTREE_WUNLOCK();
1281 			return (EAGAIN);
1282 		}
1283 	}
1284 	for (i = 0; i < count; i++) {
1285 		TAILQ_INSERT_TAIL(&V_sptree_ifnet[spp[i]->spidx.dir],
1286 		    spp[i], chain);
1287 		/*
1288 		 * NOTE: despite the fact that we keep VTI SP in the
1289 		 * separate list, SPHASH contains policies from both
1290 		 * sources. Thus SADB_X_SPDGET will correctly return
1291 		 * SP by id, because it uses SPHASH for lookups.
1292 		 */
1293 		LIST_INSERT_HEAD(SPHASH_HASH(spp[i]->id), spp[i], idhash);
1294 		spp[i]->state = IPSEC_SPSTATE_IFNET;
1295 	}
1296 	SPTREE_WUNLOCK();
1297 	/*
1298 	 * Notify user processes about new SP.
1299 	 */
1300 	for (i = 0; i < count; i++) {
1301 		m = key_setdumpsp(spp[i], SADB_X_SPDADD, 0, 0);
1302 		if (m != NULL)
1303 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1304 	}
1305 	return (0);
1306 }
1307 
1308 void
1309 key_unregister_ifnet(struct secpolicy **spp, u_int count)
1310 {
1311 	struct mbuf *m;
1312 	u_int i;
1313 
1314 	SPTREE_WLOCK();
1315 	for (i = 0; i < count; i++) {
1316 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1317 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1318 		    ("invalid direction %u", spp[i]->spidx.dir));
1319 
1320 		if (spp[i]->state != IPSEC_SPSTATE_IFNET)
1321 			continue;
1322 		spp[i]->state = IPSEC_SPSTATE_DEAD;
1323 		TAILQ_REMOVE(&V_sptree_ifnet[spp[i]->spidx.dir],
1324 		    spp[i], chain);
1325 		V_spd_size--;
1326 		LIST_REMOVE(spp[i], idhash);
1327 	}
1328 	SPTREE_WUNLOCK();
1329 	if (SPDCACHE_ENABLED())
1330 		spdcache_clear();
1331 
1332 	for (i = 0; i < count; i++) {
1333 		m = key_setdumpsp(spp[i], SADB_X_SPDDELETE, 0, 0);
1334 		if (m != NULL)
1335 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1336 	}
1337 }
1338 
1339 /*
1340  * Must be called after calling key_allocsa().
1341  * This function is called by key_freesp() to free some SA allocated
1342  * for a policy.
1343  */
1344 void
1345 key_freesav(struct secasvar **psav)
1346 {
1347 	struct secasvar *sav = *psav;
1348 
1349 	IPSEC_ASSERT(sav != NULL, ("null sav"));
1350 	if (SAV_DELREF(sav) == 0)
1351 		return;
1352 
1353 	KEYDBG(IPSEC_STAMP,
1354 	    printf("%s: last reference to SA(%p)\n", __func__, sav));
1355 
1356 	*psav = NULL;
1357 	key_delsav(sav);
1358 }
1359 
1360 /*
1361  * Unlink SA from SAH and SPI hash under SAHTREE_WLOCK.
1362  * Expect that SA has extra reference due to lookup.
1363  * Release this references, also release SAH reference after unlink.
1364  */
1365 static void
1366 key_unlinksav(struct secasvar *sav)
1367 {
1368 	struct secashead *sah;
1369 
1370 	KEYDBG(KEY_STAMP,
1371 	    printf("%s: SA(%p)\n", __func__, sav));
1372 
1373 	SAHTREE_UNLOCK_ASSERT();
1374 	SAHTREE_WLOCK();
1375 	if (sav->state == SADB_SASTATE_DEAD) {
1376 		/* SA is already unlinked */
1377 		SAHTREE_WUNLOCK();
1378 		return;
1379 	}
1380 	/* Unlink from SAH */
1381 	if (sav->state == SADB_SASTATE_LARVAL)
1382 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
1383 	else
1384 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
1385 	/* Unlink from SPI hash */
1386 	LIST_REMOVE(sav, spihash);
1387 	sav->state = SADB_SASTATE_DEAD;
1388 	sah = sav->sah;
1389 	SAHTREE_WUNLOCK();
1390 	key_freesav(&sav);
1391 	/* Since we are unlinked, release reference to SAH */
1392 	key_freesah(&sah);
1393 }
1394 
1395 /* %%% SPD management */
1396 /*
1397  * search SPD
1398  * OUT:	NULL	: not found
1399  *	others	: found, pointer to a SP.
1400  */
1401 static struct secpolicy *
1402 key_getsp(struct secpolicyindex *spidx)
1403 {
1404 	SPTREE_RLOCK_TRACKER;
1405 	struct secpolicy *sp;
1406 
1407 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
1408 
1409 	SPTREE_RLOCK();
1410 	TAILQ_FOREACH(sp, &V_sptree[spidx->dir], chain) {
1411 		if (key_cmpspidx_exactly(spidx, &sp->spidx)) {
1412 			SP_ADDREF(sp);
1413 			break;
1414 		}
1415 	}
1416 	SPTREE_RUNLOCK();
1417 
1418 	return sp;
1419 }
1420 
1421 /*
1422  * get SP by index.
1423  * OUT:	NULL	: not found
1424  *	others	: found, pointer to referenced SP.
1425  */
1426 static struct secpolicy *
1427 key_getspbyid(uint32_t id)
1428 {
1429 	SPTREE_RLOCK_TRACKER;
1430 	struct secpolicy *sp;
1431 
1432 	SPTREE_RLOCK();
1433 	LIST_FOREACH(sp, SPHASH_HASH(id), idhash) {
1434 		if (sp->id == id) {
1435 			SP_ADDREF(sp);
1436 			break;
1437 		}
1438 	}
1439 	SPTREE_RUNLOCK();
1440 	return (sp);
1441 }
1442 
1443 struct secpolicy *
1444 key_newsp(void)
1445 {
1446 	struct secpolicy *sp;
1447 
1448 	sp = malloc(sizeof(*sp), M_IPSEC_SP, M_NOWAIT | M_ZERO);
1449 	if (sp != NULL)
1450 		SP_INITREF(sp);
1451 	return (sp);
1452 }
1453 
1454 struct ipsecrequest *
1455 ipsec_newisr(void)
1456 {
1457 
1458 	return (malloc(sizeof(struct ipsecrequest), M_IPSEC_SR,
1459 	    M_NOWAIT | M_ZERO));
1460 }
1461 
1462 void
1463 ipsec_delisr(struct ipsecrequest *p)
1464 {
1465 
1466 	free(p, M_IPSEC_SR);
1467 }
1468 
1469 /*
1470  * create secpolicy structure from sadb_x_policy structure.
1471  * NOTE: `state', `secpolicyindex' and 'id' in secpolicy structure
1472  * are not set, so must be set properly later.
1473  */
1474 struct secpolicy *
1475 key_msg2sp(struct sadb_x_policy *xpl0, size_t len, int *error)
1476 {
1477 	struct secpolicy *newsp;
1478 
1479 	IPSEC_ASSERT(xpl0 != NULL, ("null xpl0"));
1480 	IPSEC_ASSERT(len >= sizeof(*xpl0), ("policy too short: %zu", len));
1481 
1482 	if (len != PFKEY_EXTLEN(xpl0)) {
1483 		ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n", __func__));
1484 		*error = EINVAL;
1485 		return NULL;
1486 	}
1487 
1488 	if ((newsp = key_newsp()) == NULL) {
1489 		*error = ENOBUFS;
1490 		return NULL;
1491 	}
1492 
1493 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1494 	newsp->policy = xpl0->sadb_x_policy_type;
1495 	newsp->priority = xpl0->sadb_x_policy_priority;
1496 	newsp->tcount = 0;
1497 
1498 	/* check policy */
1499 	switch (xpl0->sadb_x_policy_type) {
1500 	case IPSEC_POLICY_DISCARD:
1501 	case IPSEC_POLICY_NONE:
1502 	case IPSEC_POLICY_ENTRUST:
1503 	case IPSEC_POLICY_BYPASS:
1504 		break;
1505 
1506 	case IPSEC_POLICY_IPSEC:
1507 	    {
1508 		struct sadb_x_ipsecrequest *xisr;
1509 		struct ipsecrequest *isr;
1510 		int tlen;
1511 
1512 		/* validity check */
1513 		if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1514 			ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n",
1515 				__func__));
1516 			key_freesp(&newsp);
1517 			*error = EINVAL;
1518 			return NULL;
1519 		}
1520 
1521 		tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1522 		xisr = (struct sadb_x_ipsecrequest *)(xpl0 + 1);
1523 
1524 		while (tlen > 0) {
1525 			/* length check */
1526 			if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr) ||
1527 			    xisr->sadb_x_ipsecrequest_len > tlen) {
1528 				ipseclog((LOG_DEBUG, "%s: invalid ipsecrequest "
1529 					"length.\n", __func__));
1530 				key_freesp(&newsp);
1531 				*error = EINVAL;
1532 				return NULL;
1533 			}
1534 
1535 			if (newsp->tcount >= IPSEC_MAXREQ) {
1536 				ipseclog((LOG_DEBUG,
1537 				    "%s: too many ipsecrequests.\n",
1538 				    __func__));
1539 				key_freesp(&newsp);
1540 				*error = EINVAL;
1541 				return (NULL);
1542 			}
1543 
1544 			/* allocate request buffer */
1545 			/* NB: data structure is zero'd */
1546 			isr = ipsec_newisr();
1547 			if (isr == NULL) {
1548 				ipseclog((LOG_DEBUG,
1549 				    "%s: No more memory.\n", __func__));
1550 				key_freesp(&newsp);
1551 				*error = ENOBUFS;
1552 				return NULL;
1553 			}
1554 
1555 			newsp->req[newsp->tcount++] = isr;
1556 
1557 			/* set values */
1558 			switch (xisr->sadb_x_ipsecrequest_proto) {
1559 			case IPPROTO_ESP:
1560 			case IPPROTO_AH:
1561 			case IPPROTO_IPCOMP:
1562 				break;
1563 			default:
1564 				ipseclog((LOG_DEBUG,
1565 				    "%s: invalid proto type=%u\n", __func__,
1566 				    xisr->sadb_x_ipsecrequest_proto));
1567 				key_freesp(&newsp);
1568 				*error = EPROTONOSUPPORT;
1569 				return NULL;
1570 			}
1571 			isr->saidx.proto =
1572 			    (uint8_t)xisr->sadb_x_ipsecrequest_proto;
1573 
1574 			switch (xisr->sadb_x_ipsecrequest_mode) {
1575 			case IPSEC_MODE_TRANSPORT:
1576 			case IPSEC_MODE_TUNNEL:
1577 				break;
1578 			case IPSEC_MODE_ANY:
1579 			default:
1580 				ipseclog((LOG_DEBUG,
1581 				    "%s: invalid mode=%u\n", __func__,
1582 				    xisr->sadb_x_ipsecrequest_mode));
1583 				key_freesp(&newsp);
1584 				*error = EINVAL;
1585 				return NULL;
1586 			}
1587 			isr->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1588 
1589 			switch (xisr->sadb_x_ipsecrequest_level) {
1590 			case IPSEC_LEVEL_DEFAULT:
1591 			case IPSEC_LEVEL_USE:
1592 			case IPSEC_LEVEL_REQUIRE:
1593 				break;
1594 			case IPSEC_LEVEL_UNIQUE:
1595 				/* validity check */
1596 				/*
1597 				 * If range violation of reqid, kernel will
1598 				 * update it, don't refuse it.
1599 				 */
1600 				if (xisr->sadb_x_ipsecrequest_reqid
1601 						> IPSEC_MANUAL_REQID_MAX) {
1602 					ipseclog((LOG_DEBUG,
1603 					    "%s: reqid=%d range "
1604 					    "violation, updated by kernel.\n",
1605 					    __func__,
1606 					    xisr->sadb_x_ipsecrequest_reqid));
1607 					xisr->sadb_x_ipsecrequest_reqid = 0;
1608 				}
1609 
1610 				/* allocate new reqid id if reqid is zero. */
1611 				if (xisr->sadb_x_ipsecrequest_reqid == 0) {
1612 					u_int32_t reqid;
1613 					if ((reqid = key_newreqid()) == 0) {
1614 						key_freesp(&newsp);
1615 						*error = ENOBUFS;
1616 						return NULL;
1617 					}
1618 					isr->saidx.reqid = reqid;
1619 					xisr->sadb_x_ipsecrequest_reqid = reqid;
1620 				} else {
1621 				/* set it for manual keying. */
1622 					isr->saidx.reqid =
1623 					    xisr->sadb_x_ipsecrequest_reqid;
1624 				}
1625 				break;
1626 
1627 			default:
1628 				ipseclog((LOG_DEBUG, "%s: invalid level=%u\n",
1629 					__func__,
1630 					xisr->sadb_x_ipsecrequest_level));
1631 				key_freesp(&newsp);
1632 				*error = EINVAL;
1633 				return NULL;
1634 			}
1635 			isr->level = xisr->sadb_x_ipsecrequest_level;
1636 
1637 			/* set IP addresses if there */
1638 			if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1639 				struct sockaddr *paddr;
1640 
1641 				len = tlen - sizeof(*xisr);
1642 				paddr = (struct sockaddr *)(xisr + 1);
1643 				/* validity check */
1644 				if (len < sizeof(struct sockaddr) ||
1645 				    len < 2 * paddr->sa_len ||
1646 				    paddr->sa_len > sizeof(isr->saidx.src)) {
1647 					ipseclog((LOG_DEBUG, "%s: invalid "
1648 						"request address length.\n",
1649 						__func__));
1650 					key_freesp(&newsp);
1651 					*error = EINVAL;
1652 					return NULL;
1653 				}
1654 				/*
1655 				 * Request length should be enough to keep
1656 				 * source and destination addresses.
1657 				 */
1658 				if (xisr->sadb_x_ipsecrequest_len <
1659 				    sizeof(*xisr) + 2 * paddr->sa_len) {
1660 					ipseclog((LOG_DEBUG, "%s: invalid "
1661 					    "ipsecrequest length.\n",
1662 					    __func__));
1663 					key_freesp(&newsp);
1664 					*error = EINVAL;
1665 					return (NULL);
1666 				}
1667 				bcopy(paddr, &isr->saidx.src, paddr->sa_len);
1668 				paddr = (struct sockaddr *)((caddr_t)paddr +
1669 				    paddr->sa_len);
1670 
1671 				/* validity check */
1672 				if (paddr->sa_len !=
1673 				    isr->saidx.src.sa.sa_len) {
1674 					ipseclog((LOG_DEBUG, "%s: invalid "
1675 						"request address length.\n",
1676 						__func__));
1677 					key_freesp(&newsp);
1678 					*error = EINVAL;
1679 					return NULL;
1680 				}
1681 				/* AF family should match */
1682 				if (paddr->sa_family !=
1683 				    isr->saidx.src.sa.sa_family) {
1684 					ipseclog((LOG_DEBUG, "%s: address "
1685 					    "family doesn't match.\n",
1686 						__func__));
1687 					key_freesp(&newsp);
1688 					*error = EINVAL;
1689 					return (NULL);
1690 				}
1691 				bcopy(paddr, &isr->saidx.dst, paddr->sa_len);
1692 			} else {
1693 				/*
1694 				 * Addresses for TUNNEL mode requests are
1695 				 * mandatory.
1696 				 */
1697 				if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1698 					ipseclog((LOG_DEBUG, "%s: missing "
1699 					    "request addresses.\n", __func__));
1700 					key_freesp(&newsp);
1701 					*error = EINVAL;
1702 					return (NULL);
1703 				}
1704 			}
1705 			tlen -= xisr->sadb_x_ipsecrequest_len;
1706 
1707 			/* validity check */
1708 			if (tlen < 0) {
1709 				ipseclog((LOG_DEBUG, "%s: becoming tlen < 0.\n",
1710 					__func__));
1711 				key_freesp(&newsp);
1712 				*error = EINVAL;
1713 				return NULL;
1714 			}
1715 
1716 			xisr = (struct sadb_x_ipsecrequest *)((caddr_t)xisr
1717 			                 + xisr->sadb_x_ipsecrequest_len);
1718 		}
1719 		/* XXXAE: LARVAL SP */
1720 		if (newsp->tcount < 1) {
1721 			ipseclog((LOG_DEBUG, "%s: valid IPSEC transforms "
1722 			    "not found.\n", __func__));
1723 			key_freesp(&newsp);
1724 			*error = EINVAL;
1725 			return (NULL);
1726 		}
1727 	    }
1728 		break;
1729 	default:
1730 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
1731 		key_freesp(&newsp);
1732 		*error = EINVAL;
1733 		return NULL;
1734 	}
1735 
1736 	*error = 0;
1737 	return (newsp);
1738 }
1739 
1740 uint32_t
1741 key_newreqid(void)
1742 {
1743 	static uint32_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1744 
1745 	if (auto_reqid == ~0)
1746 		auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1747 	else
1748 		auto_reqid++;
1749 
1750 	/* XXX should be unique check */
1751 	return (auto_reqid);
1752 }
1753 
1754 /*
1755  * copy secpolicy struct to sadb_x_policy structure indicated.
1756  */
1757 static struct mbuf *
1758 key_sp2mbuf(struct secpolicy *sp)
1759 {
1760 	struct mbuf *m;
1761 	size_t tlen;
1762 
1763 	tlen = key_getspreqmsglen(sp);
1764 	m = m_get2(tlen, M_NOWAIT, MT_DATA, 0);
1765 	if (m == NULL)
1766 		return (NULL);
1767 	m_align(m, tlen);
1768 	m->m_len = tlen;
1769 	if (key_sp2msg(sp, m->m_data, &tlen) != 0) {
1770 		m_freem(m);
1771 		return (NULL);
1772 	}
1773 	return (m);
1774 }
1775 
1776 int
1777 key_sp2msg(struct secpolicy *sp, void *request, size_t *len)
1778 {
1779 	struct sadb_x_ipsecrequest *xisr;
1780 	struct sadb_x_policy *xpl;
1781 	struct ipsecrequest *isr;
1782 	size_t xlen, ilen;
1783 	caddr_t p;
1784 	int error, i;
1785 
1786 	IPSEC_ASSERT(sp != NULL, ("null policy"));
1787 
1788 	xlen = sizeof(*xpl);
1789 	if (*len < xlen)
1790 		return (EINVAL);
1791 
1792 	error = 0;
1793 	bzero(request, *len);
1794 	xpl = (struct sadb_x_policy *)request;
1795 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
1796 	xpl->sadb_x_policy_type = sp->policy;
1797 	xpl->sadb_x_policy_dir = sp->spidx.dir;
1798 	xpl->sadb_x_policy_id = sp->id;
1799 	xpl->sadb_x_policy_priority = sp->priority;
1800 	switch (sp->state) {
1801 	case IPSEC_SPSTATE_IFNET:
1802 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_IFNET;
1803 		break;
1804 	case IPSEC_SPSTATE_PCB:
1805 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_PCB;
1806 		break;
1807 	default:
1808 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_GLOBAL;
1809 	}
1810 
1811 	/* if is the policy for ipsec ? */
1812 	if (sp->policy == IPSEC_POLICY_IPSEC) {
1813 		p = (caddr_t)xpl + sizeof(*xpl);
1814 		for (i = 0; i < sp->tcount; i++) {
1815 			isr = sp->req[i];
1816 			ilen = PFKEY_ALIGN8(sizeof(*xisr) +
1817 			    isr->saidx.src.sa.sa_len +
1818 			    isr->saidx.dst.sa.sa_len);
1819 			xlen += ilen;
1820 			if (xlen > *len) {
1821 				error = ENOBUFS;
1822 				/* Calculate needed size */
1823 				continue;
1824 			}
1825 			xisr = (struct sadb_x_ipsecrequest *)p;
1826 			xisr->sadb_x_ipsecrequest_len = ilen;
1827 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
1828 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
1829 			xisr->sadb_x_ipsecrequest_level = isr->level;
1830 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
1831 
1832 			p += sizeof(*xisr);
1833 			bcopy(&isr->saidx.src, p, isr->saidx.src.sa.sa_len);
1834 			p += isr->saidx.src.sa.sa_len;
1835 			bcopy(&isr->saidx.dst, p, isr->saidx.dst.sa.sa_len);
1836 			p += isr->saidx.dst.sa.sa_len;
1837 		}
1838 	}
1839 	xpl->sadb_x_policy_len = PFKEY_UNIT64(xlen);
1840 	if (error == 0)
1841 		*len = xlen;
1842 	else
1843 		*len = sizeof(*xpl);
1844 	return (error);
1845 }
1846 
1847 /* m will not be freed nor modified */
1848 static struct mbuf *
1849 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
1850     int ndeep, int nitem, ...)
1851 {
1852 	va_list ap;
1853 	int idx;
1854 	int i;
1855 	struct mbuf *result = NULL, *n;
1856 	int len;
1857 
1858 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1859 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1860 
1861 	va_start(ap, nitem);
1862 	for (i = 0; i < nitem; i++) {
1863 		idx = va_arg(ap, int);
1864 		if (idx < 0 || idx > SADB_EXT_MAX)
1865 			goto fail;
1866 		/* don't attempt to pull empty extension */
1867 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
1868 			continue;
1869 		if (idx != SADB_EXT_RESERVED  &&
1870 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
1871 			continue;
1872 
1873 		if (idx == SADB_EXT_RESERVED) {
1874 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
1875 
1876 			IPSEC_ASSERT(len <= MHLEN, ("header too big %u", len));
1877 
1878 			MGETHDR(n, M_NOWAIT, MT_DATA);
1879 			if (!n)
1880 				goto fail;
1881 			n->m_len = len;
1882 			n->m_next = NULL;
1883 			m_copydata(m, 0, sizeof(struct sadb_msg),
1884 			    mtod(n, caddr_t));
1885 		} else if (i < ndeep) {
1886 			len = mhp->extlen[idx];
1887 			n = m_get2(len, M_NOWAIT, MT_DATA, 0);
1888 			if (n == NULL)
1889 				goto fail;
1890 			m_align(n, len);
1891 			n->m_len = len;
1892 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
1893 			    mtod(n, caddr_t));
1894 		} else {
1895 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
1896 			    M_NOWAIT);
1897 		}
1898 		if (n == NULL)
1899 			goto fail;
1900 
1901 		if (result)
1902 			m_cat(result, n);
1903 		else
1904 			result = n;
1905 	}
1906 	va_end(ap);
1907 
1908 	if ((result->m_flags & M_PKTHDR) != 0) {
1909 		result->m_pkthdr.len = 0;
1910 		for (n = result; n; n = n->m_next)
1911 			result->m_pkthdr.len += n->m_len;
1912 	}
1913 
1914 	return result;
1915 
1916 fail:
1917 	m_freem(result);
1918 	va_end(ap);
1919 	return NULL;
1920 }
1921 
1922 /*
1923  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
1924  * add an entry to SP database, when received
1925  *   <base, address(SD), (lifetime(H),) policy>
1926  * from the user(?).
1927  * Adding to SP database,
1928  * and send
1929  *   <base, address(SD), (lifetime(H),) policy>
1930  * to the socket which was send.
1931  *
1932  * SPDADD set a unique policy entry.
1933  * SPDSETIDX like SPDADD without a part of policy requests.
1934  * SPDUPDATE replace a unique policy entry.
1935  *
1936  * XXXAE: serialize this in PF_KEY to avoid races.
1937  * m will always be freed.
1938  */
1939 static int
1940 key_spdadd(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
1941 {
1942 	struct secpolicyindex spidx;
1943 	struct sadb_address *src0, *dst0;
1944 	struct sadb_x_policy *xpl0, *xpl;
1945 	struct sadb_lifetime *lft = NULL;
1946 	struct secpolicy *newsp;
1947 	int error;
1948 
1949 	IPSEC_ASSERT(so != NULL, ("null socket"));
1950 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1951 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1952 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
1953 
1954 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
1955 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
1956 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
1957 		ipseclog((LOG_DEBUG,
1958 		    "%s: invalid message: missing required header.\n",
1959 		    __func__));
1960 		return key_senderror(so, m, EINVAL);
1961 	}
1962 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
1963 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
1964 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
1965 		ipseclog((LOG_DEBUG,
1966 		    "%s: invalid message: wrong header size.\n", __func__));
1967 		return key_senderror(so, m, EINVAL);
1968 	}
1969 	if (!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD)) {
1970 		if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD)) {
1971 			ipseclog((LOG_DEBUG,
1972 			    "%s: invalid message: wrong header size.\n",
1973 			    __func__));
1974 			return key_senderror(so, m, EINVAL);
1975 		}
1976 		lft = (struct sadb_lifetime *)mhp->ext[SADB_EXT_LIFETIME_HARD];
1977 	}
1978 
1979 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
1980 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
1981 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
1982 
1983 	/* check the direciton */
1984 	switch (xpl0->sadb_x_policy_dir) {
1985 	case IPSEC_DIR_INBOUND:
1986 	case IPSEC_DIR_OUTBOUND:
1987 		break;
1988 	default:
1989 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
1990 		return key_senderror(so, m, EINVAL);
1991 	}
1992 	/* key_spdadd() accepts DISCARD, NONE and IPSEC. */
1993 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
1994 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
1995 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
1996 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
1997 		return key_senderror(so, m, EINVAL);
1998 	}
1999 
2000 	/* policy requests are mandatory when action is ipsec. */
2001 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
2002 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
2003 		ipseclog((LOG_DEBUG,
2004 		    "%s: policy requests required.\n", __func__));
2005 		return key_senderror(so, m, EINVAL);
2006 	}
2007 
2008 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
2009 	    (struct sockaddr *)(dst0 + 1));
2010 	if (error != 0 ||
2011 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2012 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2013 		return key_senderror(so, m, error);
2014 	}
2015 	/* make secindex */
2016 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2017 	                src0 + 1,
2018 	                dst0 + 1,
2019 	                src0->sadb_address_prefixlen,
2020 	                dst0->sadb_address_prefixlen,
2021 	                src0->sadb_address_proto,
2022 	                &spidx);
2023 	/* Checking there is SP already or not. */
2024 	newsp = key_getsp(&spidx);
2025 	if (newsp != NULL) {
2026 		if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2027 			KEYDBG(KEY_STAMP,
2028 			    printf("%s: unlink SP(%p) for SPDUPDATE\n",
2029 				__func__, newsp));
2030 			KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
2031 			key_unlink(newsp);
2032 			key_freesp(&newsp);
2033 		} else {
2034 			key_freesp(&newsp);
2035 			ipseclog((LOG_DEBUG,
2036 			    "%s: a SP entry exists already.\n", __func__));
2037 			return (key_senderror(so, m, EEXIST));
2038 		}
2039 	}
2040 
2041 	/* allocate new SP entry */
2042 	if ((newsp = key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error)) == NULL) {
2043 		return key_senderror(so, m, error);
2044 	}
2045 
2046 	newsp->lastused = newsp->created = time_second;
2047 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
2048 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
2049 	bcopy(&spidx, &newsp->spidx, sizeof(spidx));
2050 
2051 	/* XXXAE: there is race between key_getsp() and key_insertsp() */
2052 	SPTREE_WLOCK();
2053 	if ((newsp->id = key_getnewspid()) == 0) {
2054 		SPTREE_WUNLOCK();
2055 		key_freesp(&newsp);
2056 		return key_senderror(so, m, ENOBUFS);
2057 	}
2058 	key_insertsp(newsp);
2059 	SPTREE_WUNLOCK();
2060 	if (SPDCACHE_ENABLED())
2061 		spdcache_clear();
2062 
2063 	KEYDBG(KEY_STAMP,
2064 	    printf("%s: SP(%p)\n", __func__, newsp));
2065 	KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
2066 
2067     {
2068 	struct mbuf *n, *mpolicy;
2069 	struct sadb_msg *newmsg;
2070 	int off;
2071 
2072 	/* create new sadb_msg to reply. */
2073 	if (lft) {
2074 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
2075 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
2076 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2077 	} else {
2078 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
2079 		    SADB_X_EXT_POLICY,
2080 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2081 	}
2082 	if (!n)
2083 		return key_senderror(so, m, ENOBUFS);
2084 
2085 	if (n->m_len < sizeof(*newmsg)) {
2086 		n = m_pullup(n, sizeof(*newmsg));
2087 		if (!n)
2088 			return key_senderror(so, m, ENOBUFS);
2089 	}
2090 	newmsg = mtod(n, struct sadb_msg *);
2091 	newmsg->sadb_msg_errno = 0;
2092 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2093 
2094 	off = 0;
2095 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
2096 	    sizeof(*xpl), &off);
2097 	if (mpolicy == NULL) {
2098 		/* n is already freed */
2099 		return key_senderror(so, m, ENOBUFS);
2100 	}
2101 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, caddr_t) + off);
2102 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
2103 		m_freem(n);
2104 		return key_senderror(so, m, EINVAL);
2105 	}
2106 	xpl->sadb_x_policy_id = newsp->id;
2107 
2108 	m_freem(m);
2109 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2110     }
2111 }
2112 
2113 /*
2114  * get new policy id.
2115  * OUT:
2116  *	0:	failure.
2117  *	others: success.
2118  */
2119 static uint32_t
2120 key_getnewspid(void)
2121 {
2122 	struct secpolicy *sp;
2123 	uint32_t newid = 0;
2124 	int count = V_key_spi_trycnt;	/* XXX */
2125 
2126 	SPTREE_WLOCK_ASSERT();
2127 	while (count--) {
2128 		if (V_policy_id == ~0) /* overflowed */
2129 			newid = V_policy_id = 1;
2130 		else
2131 			newid = ++V_policy_id;
2132 		LIST_FOREACH(sp, SPHASH_HASH(newid), idhash) {
2133 			if (sp->id == newid)
2134 				break;
2135 		}
2136 		if (sp == NULL)
2137 			break;
2138 	}
2139 	if (count == 0 || newid == 0) {
2140 		ipseclog((LOG_DEBUG, "%s: failed to allocate policy id.\n",
2141 		    __func__));
2142 		return (0);
2143 	}
2144 	return (newid);
2145 }
2146 
2147 /*
2148  * SADB_SPDDELETE processing
2149  * receive
2150  *   <base, address(SD), policy(*)>
2151  * from the user(?), and set SADB_SASTATE_DEAD,
2152  * and send,
2153  *   <base, address(SD), policy(*)>
2154  * to the ikmpd.
2155  * policy(*) including direction of policy.
2156  *
2157  * m will always be freed.
2158  */
2159 static int
2160 key_spddelete(struct socket *so, struct mbuf *m,
2161     const struct sadb_msghdr *mhp)
2162 {
2163 	struct secpolicyindex spidx;
2164 	struct sadb_address *src0, *dst0;
2165 	struct sadb_x_policy *xpl0;
2166 	struct secpolicy *sp;
2167 
2168 	IPSEC_ASSERT(so != NULL, ("null so"));
2169 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2170 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2171 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2172 
2173 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
2174 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
2175 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
2176 		ipseclog((LOG_DEBUG,
2177 		    "%s: invalid message: missing required header.\n",
2178 		    __func__));
2179 		return key_senderror(so, m, EINVAL);
2180 	}
2181 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
2182 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
2183 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2184 		ipseclog((LOG_DEBUG,
2185 		    "%s: invalid message: wrong header size.\n", __func__));
2186 		return key_senderror(so, m, EINVAL);
2187 	}
2188 
2189 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
2190 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
2191 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
2192 
2193 	/* check the direciton */
2194 	switch (xpl0->sadb_x_policy_dir) {
2195 	case IPSEC_DIR_INBOUND:
2196 	case IPSEC_DIR_OUTBOUND:
2197 		break;
2198 	default:
2199 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2200 		return key_senderror(so, m, EINVAL);
2201 	}
2202 	/* Only DISCARD, NONE and IPSEC are allowed */
2203 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2204 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2205 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2206 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2207 		return key_senderror(so, m, EINVAL);
2208 	}
2209 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
2210 	    (struct sockaddr *)(dst0 + 1)) != 0 ||
2211 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2212 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2213 		return key_senderror(so, m, EINVAL);
2214 	}
2215 	/* make secindex */
2216 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2217 	                src0 + 1,
2218 	                dst0 + 1,
2219 	                src0->sadb_address_prefixlen,
2220 	                dst0->sadb_address_prefixlen,
2221 	                src0->sadb_address_proto,
2222 	                &spidx);
2223 
2224 	/* Is there SP in SPD ? */
2225 	if ((sp = key_getsp(&spidx)) == NULL) {
2226 		ipseclog((LOG_DEBUG, "%s: no SP found.\n", __func__));
2227 		return key_senderror(so, m, EINVAL);
2228 	}
2229 
2230 	/* save policy id to buffer to be returned. */
2231 	xpl0->sadb_x_policy_id = sp->id;
2232 
2233 	KEYDBG(KEY_STAMP,
2234 	    printf("%s: SP(%p)\n", __func__, sp));
2235 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2236 	key_unlink(sp);
2237 	key_freesp(&sp);
2238 
2239     {
2240 	struct mbuf *n;
2241 	struct sadb_msg *newmsg;
2242 
2243 	/* create new sadb_msg to reply. */
2244 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2245 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2246 	if (!n)
2247 		return key_senderror(so, m, ENOBUFS);
2248 
2249 	newmsg = mtod(n, struct sadb_msg *);
2250 	newmsg->sadb_msg_errno = 0;
2251 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2252 
2253 	m_freem(m);
2254 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2255     }
2256 }
2257 
2258 /*
2259  * SADB_SPDDELETE2 processing
2260  * receive
2261  *   <base, policy(*)>
2262  * from the user(?), and set SADB_SASTATE_DEAD,
2263  * and send,
2264  *   <base, policy(*)>
2265  * to the ikmpd.
2266  * policy(*) including direction of policy.
2267  *
2268  * m will always be freed.
2269  */
2270 static int
2271 key_spddelete2(struct socket *so, struct mbuf *m,
2272     const struct sadb_msghdr *mhp)
2273 {
2274 	struct secpolicy *sp;
2275 	uint32_t id;
2276 
2277 	IPSEC_ASSERT(so != NULL, ("null socket"));
2278 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2279 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2280 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2281 
2282 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2283 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2284 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2285 		    __func__));
2286 		return key_senderror(so, m, EINVAL);
2287 	}
2288 
2289 	id = ((struct sadb_x_policy *)
2290 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2291 
2292 	/* Is there SP in SPD ? */
2293 	if ((sp = key_getspbyid(id)) == NULL) {
2294 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2295 		    __func__, id));
2296 		return key_senderror(so, m, EINVAL);
2297 	}
2298 
2299 	KEYDBG(KEY_STAMP,
2300 	    printf("%s: SP(%p)\n", __func__, sp));
2301 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2302 	key_unlink(sp);
2303 	if (sp->state != IPSEC_SPSTATE_DEAD) {
2304 		ipseclog((LOG_DEBUG, "%s: failed to delete SP with id %u.\n",
2305 		    __func__, id));
2306 		key_freesp(&sp);
2307 		return (key_senderror(so, m, EACCES));
2308 	}
2309 	key_freesp(&sp);
2310 
2311     {
2312 	struct mbuf *n, *nn;
2313 	struct sadb_msg *newmsg;
2314 	int off, len;
2315 
2316 	/* create new sadb_msg to reply. */
2317 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2318 
2319 	MGETHDR(n, M_NOWAIT, MT_DATA);
2320 	if (n && len > MHLEN) {
2321 		if (!(MCLGET(n, M_NOWAIT))) {
2322 			m_freem(n);
2323 			n = NULL;
2324 		}
2325 	}
2326 	if (!n)
2327 		return key_senderror(so, m, ENOBUFS);
2328 
2329 	n->m_len = len;
2330 	n->m_next = NULL;
2331 	off = 0;
2332 
2333 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
2334 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2335 
2336 	IPSEC_ASSERT(off == len, ("length inconsistency (off %u len %u)",
2337 		off, len));
2338 
2339 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2340 	    mhp->extlen[SADB_X_EXT_POLICY], M_NOWAIT);
2341 	if (!n->m_next) {
2342 		m_freem(n);
2343 		return key_senderror(so, m, ENOBUFS);
2344 	}
2345 
2346 	n->m_pkthdr.len = 0;
2347 	for (nn = n; nn; nn = nn->m_next)
2348 		n->m_pkthdr.len += nn->m_len;
2349 
2350 	newmsg = mtod(n, struct sadb_msg *);
2351 	newmsg->sadb_msg_errno = 0;
2352 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2353 
2354 	m_freem(m);
2355 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2356     }
2357 }
2358 
2359 /*
2360  * SADB_X_SPDGET processing
2361  * receive
2362  *   <base, policy(*)>
2363  * from the user(?),
2364  * and send,
2365  *   <base, address(SD), policy>
2366  * to the ikmpd.
2367  * policy(*) including direction of policy.
2368  *
2369  * m will always be freed.
2370  */
2371 static int
2372 key_spdget(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2373 {
2374 	struct secpolicy *sp;
2375 	struct mbuf *n;
2376 	uint32_t id;
2377 
2378 	IPSEC_ASSERT(so != NULL, ("null socket"));
2379 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2380 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2381 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2382 
2383 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2384 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2385 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2386 		    __func__));
2387 		return key_senderror(so, m, EINVAL);
2388 	}
2389 
2390 	id = ((struct sadb_x_policy *)
2391 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2392 
2393 	/* Is there SP in SPD ? */
2394 	if ((sp = key_getspbyid(id)) == NULL) {
2395 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2396 		    __func__, id));
2397 		return key_senderror(so, m, ENOENT);
2398 	}
2399 
2400 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2401 	    mhp->msg->sadb_msg_pid);
2402 	key_freesp(&sp);
2403 	if (n != NULL) {
2404 		m_freem(m);
2405 		return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2406 	} else
2407 		return key_senderror(so, m, ENOBUFS);
2408 }
2409 
2410 /*
2411  * SADB_X_SPDACQUIRE processing.
2412  * Acquire policy and SA(s) for a *OUTBOUND* packet.
2413  * send
2414  *   <base, policy(*)>
2415  * to KMD, and expect to receive
2416  *   <base> with SADB_X_SPDACQUIRE if error occurred,
2417  * or
2418  *   <base, policy>
2419  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2420  * policy(*) is without policy requests.
2421  *
2422  *    0     : succeed
2423  *    others: error number
2424  */
2425 int
2426 key_spdacquire(struct secpolicy *sp)
2427 {
2428 	struct mbuf *result = NULL, *m;
2429 	struct secspacq *newspacq;
2430 
2431 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2432 	IPSEC_ASSERT(sp->req == NULL, ("policy exists"));
2433 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
2434 		("policy not IPSEC %u", sp->policy));
2435 
2436 	/* Get an entry to check whether sent message or not. */
2437 	newspacq = key_getspacq(&sp->spidx);
2438 	if (newspacq != NULL) {
2439 		if (V_key_blockacq_count < newspacq->count) {
2440 			/* reset counter and do send message. */
2441 			newspacq->count = 0;
2442 		} else {
2443 			/* increment counter and do nothing. */
2444 			newspacq->count++;
2445 			SPACQ_UNLOCK();
2446 			return (0);
2447 		}
2448 		SPACQ_UNLOCK();
2449 	} else {
2450 		/* make new entry for blocking to send SADB_ACQUIRE. */
2451 		newspacq = key_newspacq(&sp->spidx);
2452 		if (newspacq == NULL)
2453 			return ENOBUFS;
2454 	}
2455 
2456 	/* create new sadb_msg to reply. */
2457 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2458 	if (!m)
2459 		return ENOBUFS;
2460 
2461 	result = m;
2462 
2463 	result->m_pkthdr.len = 0;
2464 	for (m = result; m; m = m->m_next)
2465 		result->m_pkthdr.len += m->m_len;
2466 
2467 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2468 	    PFKEY_UNIT64(result->m_pkthdr.len);
2469 
2470 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2471 }
2472 
2473 /*
2474  * SADB_SPDFLUSH processing
2475  * receive
2476  *   <base>
2477  * from the user, and free all entries in secpctree.
2478  * and send,
2479  *   <base>
2480  * to the user.
2481  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2482  *
2483  * m will always be freed.
2484  */
2485 static int
2486 key_spdflush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2487 {
2488 	struct secpolicy_queue drainq;
2489 	struct sadb_msg *newmsg;
2490 	struct secpolicy *sp, *nextsp;
2491 	u_int dir;
2492 
2493 	IPSEC_ASSERT(so != NULL, ("null socket"));
2494 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2495 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2496 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2497 
2498 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2499 		return key_senderror(so, m, EINVAL);
2500 
2501 	TAILQ_INIT(&drainq);
2502 	SPTREE_WLOCK();
2503 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2504 		TAILQ_CONCAT(&drainq, &V_sptree[dir], chain);
2505 	}
2506 	/*
2507 	 * We need to set state to DEAD for each policy to be sure,
2508 	 * that another thread won't try to unlink it.
2509 	 * Also remove SP from sphash.
2510 	 */
2511 	TAILQ_FOREACH(sp, &drainq, chain) {
2512 		sp->state = IPSEC_SPSTATE_DEAD;
2513 		LIST_REMOVE(sp, idhash);
2514 	}
2515 	V_sp_genid++;
2516 	V_spd_size = 0;
2517 	SPTREE_WUNLOCK();
2518 	if (SPDCACHE_ENABLED())
2519 		spdcache_clear();
2520 	sp = TAILQ_FIRST(&drainq);
2521 	while (sp != NULL) {
2522 		nextsp = TAILQ_NEXT(sp, chain);
2523 		key_freesp(&sp);
2524 		sp = nextsp;
2525 	}
2526 
2527 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2528 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
2529 		return key_senderror(so, m, ENOBUFS);
2530 	}
2531 
2532 	if (m->m_next)
2533 		m_freem(m->m_next);
2534 	m->m_next = NULL;
2535 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2536 	newmsg = mtod(m, struct sadb_msg *);
2537 	newmsg->sadb_msg_errno = 0;
2538 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2539 
2540 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2541 }
2542 
2543 static uint8_t
2544 key_satype2scopemask(uint8_t satype)
2545 {
2546 
2547 	if (satype == IPSEC_POLICYSCOPE_ANY)
2548 		return (0xff);
2549 	return (satype);
2550 }
2551 /*
2552  * SADB_SPDDUMP processing
2553  * receive
2554  *   <base>
2555  * from the user, and dump all SP leaves and send,
2556  *   <base> .....
2557  * to the ikmpd.
2558  *
2559  * NOTE:
2560  *   sadb_msg_satype is considered as mask of policy scopes.
2561  *   m will always be freed.
2562  */
2563 static int
2564 key_spddump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2565 {
2566 	SPTREE_RLOCK_TRACKER;
2567 	struct secpolicy *sp;
2568 	struct mbuf *n;
2569 	int cnt;
2570 	u_int dir, scope;
2571 
2572 	IPSEC_ASSERT(so != NULL, ("null socket"));
2573 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2574 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2575 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2576 
2577 	/* search SPD entry and get buffer size. */
2578 	cnt = 0;
2579 	scope = key_satype2scopemask(mhp->msg->sadb_msg_satype);
2580 	SPTREE_RLOCK();
2581 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2582 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2583 			TAILQ_FOREACH(sp, &V_sptree[dir], chain)
2584 				cnt++;
2585 		}
2586 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2587 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain)
2588 				cnt++;
2589 		}
2590 	}
2591 
2592 	if (cnt == 0) {
2593 		SPTREE_RUNLOCK();
2594 		return key_senderror(so, m, ENOENT);
2595 	}
2596 
2597 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2598 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2599 			TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
2600 				--cnt;
2601 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2602 				    mhp->msg->sadb_msg_pid);
2603 
2604 				if (n != NULL)
2605 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2606 			}
2607 		}
2608 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2609 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain) {
2610 				--cnt;
2611 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2612 				    mhp->msg->sadb_msg_pid);
2613 
2614 				if (n != NULL)
2615 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2616 			}
2617 		}
2618 	}
2619 
2620 	SPTREE_RUNLOCK();
2621 	m_freem(m);
2622 	return (0);
2623 }
2624 
2625 static struct mbuf *
2626 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq,
2627     u_int32_t pid)
2628 {
2629 	struct mbuf *result = NULL, *m;
2630 	struct seclifetime lt;
2631 
2632 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid, sp->refcnt);
2633 	if (!m)
2634 		goto fail;
2635 	result = m;
2636 
2637 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2638 	    &sp->spidx.src.sa, sp->spidx.prefs,
2639 	    sp->spidx.ul_proto);
2640 	if (!m)
2641 		goto fail;
2642 	m_cat(result, m);
2643 
2644 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2645 	    &sp->spidx.dst.sa, sp->spidx.prefd,
2646 	    sp->spidx.ul_proto);
2647 	if (!m)
2648 		goto fail;
2649 	m_cat(result, m);
2650 
2651 	m = key_sp2mbuf(sp);
2652 	if (!m)
2653 		goto fail;
2654 	m_cat(result, m);
2655 
2656 	if(sp->lifetime){
2657 		lt.addtime=sp->created;
2658 		lt.usetime= sp->lastused;
2659 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_CURRENT);
2660 		if (!m)
2661 			goto fail;
2662 		m_cat(result, m);
2663 
2664 		lt.addtime=sp->lifetime;
2665 		lt.usetime= sp->validtime;
2666 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_HARD);
2667 		if (!m)
2668 			goto fail;
2669 		m_cat(result, m);
2670 	}
2671 
2672 	if ((result->m_flags & M_PKTHDR) == 0)
2673 		goto fail;
2674 
2675 	if (result->m_len < sizeof(struct sadb_msg)) {
2676 		result = m_pullup(result, sizeof(struct sadb_msg));
2677 		if (result == NULL)
2678 			goto fail;
2679 	}
2680 
2681 	result->m_pkthdr.len = 0;
2682 	for (m = result; m; m = m->m_next)
2683 		result->m_pkthdr.len += m->m_len;
2684 
2685 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2686 	    PFKEY_UNIT64(result->m_pkthdr.len);
2687 
2688 	return result;
2689 
2690 fail:
2691 	m_freem(result);
2692 	return NULL;
2693 }
2694 /*
2695  * get PFKEY message length for security policy and request.
2696  */
2697 static size_t
2698 key_getspreqmsglen(struct secpolicy *sp)
2699 {
2700 	size_t tlen, len;
2701 	int i;
2702 
2703 	tlen = sizeof(struct sadb_x_policy);
2704 	/* if is the policy for ipsec ? */
2705 	if (sp->policy != IPSEC_POLICY_IPSEC)
2706 		return (tlen);
2707 
2708 	/* get length of ipsec requests */
2709 	for (i = 0; i < sp->tcount; i++) {
2710 		len = sizeof(struct sadb_x_ipsecrequest)
2711 			+ sp->req[i]->saidx.src.sa.sa_len
2712 			+ sp->req[i]->saidx.dst.sa.sa_len;
2713 
2714 		tlen += PFKEY_ALIGN8(len);
2715 	}
2716 	return (tlen);
2717 }
2718 
2719 /*
2720  * SADB_SPDEXPIRE processing
2721  * send
2722  *   <base, address(SD), lifetime(CH), policy>
2723  * to KMD by PF_KEY.
2724  *
2725  * OUT:	0	: succeed
2726  *	others	: error number
2727  */
2728 static int
2729 key_spdexpire(struct secpolicy *sp)
2730 {
2731 	struct sadb_lifetime *lt;
2732 	struct mbuf *result = NULL, *m;
2733 	int len, error = -1;
2734 
2735 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2736 
2737 	KEYDBG(KEY_STAMP,
2738 	    printf("%s: SP(%p)\n", __func__, sp));
2739 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2740 
2741 	/* set msg header */
2742 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0);
2743 	if (!m) {
2744 		error = ENOBUFS;
2745 		goto fail;
2746 	}
2747 	result = m;
2748 
2749 	/* create lifetime extension (current and hard) */
2750 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
2751 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
2752 	if (m == NULL) {
2753 		error = ENOBUFS;
2754 		goto fail;
2755 	}
2756 	m_align(m, len);
2757 	m->m_len = len;
2758 	bzero(mtod(m, caddr_t), len);
2759 	lt = mtod(m, struct sadb_lifetime *);
2760 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2761 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
2762 	lt->sadb_lifetime_allocations = 0;
2763 	lt->sadb_lifetime_bytes = 0;
2764 	lt->sadb_lifetime_addtime = sp->created;
2765 	lt->sadb_lifetime_usetime = sp->lastused;
2766 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
2767 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2768 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
2769 	lt->sadb_lifetime_allocations = 0;
2770 	lt->sadb_lifetime_bytes = 0;
2771 	lt->sadb_lifetime_addtime = sp->lifetime;
2772 	lt->sadb_lifetime_usetime = sp->validtime;
2773 	m_cat(result, m);
2774 
2775 	/* set sadb_address for source */
2776 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2777 	    &sp->spidx.src.sa,
2778 	    sp->spidx.prefs, sp->spidx.ul_proto);
2779 	if (!m) {
2780 		error = ENOBUFS;
2781 		goto fail;
2782 	}
2783 	m_cat(result, m);
2784 
2785 	/* set sadb_address for destination */
2786 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2787 	    &sp->spidx.dst.sa,
2788 	    sp->spidx.prefd, sp->spidx.ul_proto);
2789 	if (!m) {
2790 		error = ENOBUFS;
2791 		goto fail;
2792 	}
2793 	m_cat(result, m);
2794 
2795 	/* set secpolicy */
2796 	m = key_sp2mbuf(sp);
2797 	if (!m) {
2798 		error = ENOBUFS;
2799 		goto fail;
2800 	}
2801 	m_cat(result, m);
2802 
2803 	if ((result->m_flags & M_PKTHDR) == 0) {
2804 		error = EINVAL;
2805 		goto fail;
2806 	}
2807 
2808 	if (result->m_len < sizeof(struct sadb_msg)) {
2809 		result = m_pullup(result, sizeof(struct sadb_msg));
2810 		if (result == NULL) {
2811 			error = ENOBUFS;
2812 			goto fail;
2813 		}
2814 	}
2815 
2816 	result->m_pkthdr.len = 0;
2817 	for (m = result; m; m = m->m_next)
2818 		result->m_pkthdr.len += m->m_len;
2819 
2820 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2821 	    PFKEY_UNIT64(result->m_pkthdr.len);
2822 
2823 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
2824 
2825  fail:
2826 	if (result)
2827 		m_freem(result);
2828 	return error;
2829 }
2830 
2831 /* %%% SAD management */
2832 /*
2833  * allocating and initialize new SA head.
2834  * OUT:	NULL	: failure due to the lack of memory.
2835  *	others	: pointer to new SA head.
2836  */
2837 static struct secashead *
2838 key_newsah(struct secasindex *saidx)
2839 {
2840 	struct secashead *sah;
2841 
2842 	sah = malloc(sizeof(struct secashead), M_IPSEC_SAH,
2843 	    M_NOWAIT | M_ZERO);
2844 	if (sah == NULL) {
2845 		PFKEYSTAT_INC(in_nomem);
2846 		return (NULL);
2847 	}
2848 	TAILQ_INIT(&sah->savtree_larval);
2849 	TAILQ_INIT(&sah->savtree_alive);
2850 	sah->saidx = *saidx;
2851 	sah->state = SADB_SASTATE_DEAD;
2852 	SAH_INITREF(sah);
2853 
2854 	KEYDBG(KEY_STAMP,
2855 	    printf("%s: SAH(%p)\n", __func__, sah));
2856 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2857 	return (sah);
2858 }
2859 
2860 static void
2861 key_freesah(struct secashead **psah)
2862 {
2863 	struct secashead *sah = *psah;
2864 
2865 	if (SAH_DELREF(sah) == 0)
2866 		return;
2867 
2868 	KEYDBG(KEY_STAMP,
2869 	    printf("%s: last reference to SAH(%p)\n", __func__, sah));
2870 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2871 
2872 	*psah = NULL;
2873 	key_delsah(sah);
2874 }
2875 
2876 static void
2877 key_delsah(struct secashead *sah)
2878 {
2879 	IPSEC_ASSERT(sah != NULL, ("NULL sah"));
2880 	IPSEC_ASSERT(sah->state == SADB_SASTATE_DEAD,
2881 	    ("Attempt to free non DEAD SAH %p", sah));
2882 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_larval),
2883 	    ("Attempt to free SAH %p with LARVAL SA", sah));
2884 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_alive),
2885 	    ("Attempt to free SAH %p with ALIVE SA", sah));
2886 
2887 	free(sah, M_IPSEC_SAH);
2888 }
2889 
2890 /*
2891  * allocating a new SA for key_add() and key_getspi() call,
2892  * and copy the values of mhp into new buffer.
2893  * When SAD message type is SADB_GETSPI set SA state to LARVAL.
2894  * For SADB_ADD create and initialize SA with MATURE state.
2895  * OUT:	NULL	: fail
2896  *	others	: pointer to new secasvar.
2897  */
2898 static struct secasvar *
2899 key_newsav(const struct sadb_msghdr *mhp, struct secasindex *saidx,
2900     uint32_t spi, int *errp)
2901 {
2902 	struct secashead *sah;
2903 	struct secasvar *sav;
2904 	int isnew;
2905 
2906 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2907 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2908 	IPSEC_ASSERT(mhp->msg->sadb_msg_type == SADB_GETSPI ||
2909 	    mhp->msg->sadb_msg_type == SADB_ADD, ("wrong message type"));
2910 
2911 	sav = NULL;
2912 	sah = NULL;
2913 	/* check SPI value */
2914 	switch (saidx->proto) {
2915 	case IPPROTO_ESP:
2916 	case IPPROTO_AH:
2917 		/*
2918 		 * RFC 4302, 2.4. Security Parameters Index (SPI), SPI values
2919 		 * 1-255 reserved by IANA for future use,
2920 		 * 0 for implementation specific, local use.
2921 		 */
2922 		if (ntohl(spi) <= 255) {
2923 			ipseclog((LOG_DEBUG, "%s: illegal range of SPI %u.\n",
2924 			    __func__, ntohl(spi)));
2925 			*errp = EINVAL;
2926 			goto done;
2927 		}
2928 		break;
2929 	}
2930 
2931 	sav = malloc(sizeof(struct secasvar), M_IPSEC_SA, M_NOWAIT | M_ZERO);
2932 	if (sav == NULL) {
2933 		*errp = ENOBUFS;
2934 		goto done;
2935 	}
2936 	sav->lock = malloc(sizeof(struct mtx), M_IPSEC_MISC,
2937 	    M_NOWAIT | M_ZERO);
2938 	if (sav->lock == NULL) {
2939 		*errp = ENOBUFS;
2940 		goto done;
2941 	}
2942 	mtx_init(sav->lock, "ipsec association", NULL, MTX_DEF);
2943 	sav->lft_c = uma_zalloc_pcpu(V_key_lft_zone, M_NOWAIT);
2944 	if (sav->lft_c == NULL) {
2945 		*errp = ENOBUFS;
2946 		goto done;
2947 	}
2948 	counter_u64_zero(sav->lft_c_allocations);
2949 	counter_u64_zero(sav->lft_c_bytes);
2950 
2951 	sav->spi = spi;
2952 	sav->seq = mhp->msg->sadb_msg_seq;
2953 	sav->state = SADB_SASTATE_LARVAL;
2954 	sav->pid = (pid_t)mhp->msg->sadb_msg_pid;
2955 	SAV_INITREF(sav);
2956 again:
2957 	sah = key_getsah(saidx);
2958 	if (sah == NULL) {
2959 		/* create a new SA index */
2960 		sah = key_newsah(saidx);
2961 		if (sah == NULL) {
2962 			ipseclog((LOG_DEBUG,
2963 			    "%s: No more memory.\n", __func__));
2964 			*errp = ENOBUFS;
2965 			goto done;
2966 		}
2967 		isnew = 1;
2968 	} else
2969 		isnew = 0;
2970 
2971 	sav->sah = sah;
2972 	if (mhp->msg->sadb_msg_type == SADB_GETSPI) {
2973 		sav->created = time_second;
2974 	} else if (sav->state == SADB_SASTATE_LARVAL) {
2975 		/*
2976 		 * Do not call key_setsaval() second time in case
2977 		 * of `goto again`. We will have MATURE state.
2978 		 */
2979 		*errp = key_setsaval(sav, mhp);
2980 		if (*errp != 0)
2981 			goto done;
2982 		sav->state = SADB_SASTATE_MATURE;
2983 	}
2984 
2985 	SAHTREE_WLOCK();
2986 	/*
2987 	 * Check that existing SAH wasn't unlinked.
2988 	 * Since we didn't hold the SAHTREE lock, it is possible,
2989 	 * that callout handler or key_flush() or key_delete() could
2990 	 * unlink this SAH.
2991 	 */
2992 	if (isnew == 0 && sah->state == SADB_SASTATE_DEAD) {
2993 		SAHTREE_WUNLOCK();
2994 		key_freesah(&sah);	/* reference from key_getsah() */
2995 		goto again;
2996 	}
2997 	if (isnew != 0) {
2998 		/*
2999 		 * Add new SAH into SADB.
3000 		 *
3001 		 * XXXAE: we can serialize key_add and key_getspi calls, so
3002 		 * several threads will not fight in the race.
3003 		 * Otherwise we should check under SAHTREE lock, that this
3004 		 * SAH would not added twice.
3005 		 */
3006 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
3007 		/* Add new SAH into hash by addresses */
3008 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
3009 		/* Now we are linked in the chain */
3010 		sah->state = SADB_SASTATE_MATURE;
3011 		/*
3012 		 * SAV references this new SAH.
3013 		 * In case of existing SAH we reuse reference
3014 		 * from key_getsah().
3015 		 */
3016 		SAH_ADDREF(sah);
3017 	}
3018 	/* Link SAV with SAH */
3019 	if (sav->state == SADB_SASTATE_MATURE)
3020 		TAILQ_INSERT_HEAD(&sah->savtree_alive, sav, chain);
3021 	else
3022 		TAILQ_INSERT_HEAD(&sah->savtree_larval, sav, chain);
3023 	/* Add SAV into SPI hash */
3024 	LIST_INSERT_HEAD(SAVHASH_HASH(sav->spi), sav, spihash);
3025 	SAHTREE_WUNLOCK();
3026 	*errp = 0;	/* success */
3027 done:
3028 	if (*errp != 0) {
3029 		if (sav != NULL) {
3030 			if (sav->lock != NULL) {
3031 				mtx_destroy(sav->lock);
3032 				free(sav->lock, M_IPSEC_MISC);
3033 			}
3034 			if (sav->lft_c != NULL)
3035 				uma_zfree_pcpu(V_key_lft_zone, sav->lft_c);
3036 			free(sav, M_IPSEC_SA), sav = NULL;
3037 		}
3038 		if (sah != NULL)
3039 			key_freesah(&sah);
3040 		if (*errp == ENOBUFS) {
3041 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3042 			    __func__));
3043 			PFKEYSTAT_INC(in_nomem);
3044 		}
3045 	}
3046 	return (sav);
3047 }
3048 
3049 /*
3050  * free() SA variable entry.
3051  */
3052 static void
3053 key_cleansav(struct secasvar *sav)
3054 {
3055 
3056 	if (sav->natt != NULL) {
3057 		free(sav->natt, M_IPSEC_MISC);
3058 		sav->natt = NULL;
3059 	}
3060 	if (sav->flags & SADB_X_EXT_F_CLONED)
3061 		return;
3062 	/*
3063 	 * Cleanup xform state.  Note that zeroize'ing causes the
3064 	 * keys to be cleared; otherwise we must do it ourself.
3065 	 */
3066 	if (sav->tdb_xform != NULL) {
3067 		sav->tdb_xform->xf_zeroize(sav);
3068 		sav->tdb_xform = NULL;
3069 	} else {
3070 		if (sav->key_auth != NULL)
3071 			bzero(sav->key_auth->key_data, _KEYLEN(sav->key_auth));
3072 		if (sav->key_enc != NULL)
3073 			bzero(sav->key_enc->key_data, _KEYLEN(sav->key_enc));
3074 	}
3075 	if (sav->key_auth != NULL) {
3076 		if (sav->key_auth->key_data != NULL)
3077 			free(sav->key_auth->key_data, M_IPSEC_MISC);
3078 		free(sav->key_auth, M_IPSEC_MISC);
3079 		sav->key_auth = NULL;
3080 	}
3081 	if (sav->key_enc != NULL) {
3082 		if (sav->key_enc->key_data != NULL)
3083 			free(sav->key_enc->key_data, M_IPSEC_MISC);
3084 		free(sav->key_enc, M_IPSEC_MISC);
3085 		sav->key_enc = NULL;
3086 	}
3087 	if (sav->replay != NULL) {
3088 		if (sav->replay->bitmap != NULL)
3089 			free(sav->replay->bitmap, M_IPSEC_MISC);
3090 		free(sav->replay, M_IPSEC_MISC);
3091 		sav->replay = NULL;
3092 	}
3093 	if (sav->lft_h != NULL) {
3094 		free(sav->lft_h, M_IPSEC_MISC);
3095 		sav->lft_h = NULL;
3096 	}
3097 	if (sav->lft_s != NULL) {
3098 		free(sav->lft_s, M_IPSEC_MISC);
3099 		sav->lft_s = NULL;
3100 	}
3101 }
3102 
3103 /*
3104  * free() SA variable entry.
3105  */
3106 static void
3107 key_delsav(struct secasvar *sav)
3108 {
3109 	IPSEC_ASSERT(sav != NULL, ("null sav"));
3110 	IPSEC_ASSERT(sav->state == SADB_SASTATE_DEAD,
3111 	    ("attempt to free non DEAD SA %p", sav));
3112 	IPSEC_ASSERT(sav->refcnt == 0, ("reference count %u > 0",
3113 	    sav->refcnt));
3114 
3115 	/*
3116 	 * SA must be unlinked from the chain and hashtbl.
3117 	 * If SA was cloned, we leave all fields untouched,
3118 	 * except NAT-T config.
3119 	 */
3120 	key_cleansav(sav);
3121 	if ((sav->flags & SADB_X_EXT_F_CLONED) == 0) {
3122 		mtx_destroy(sav->lock);
3123 		free(sav->lock, M_IPSEC_MISC);
3124 		uma_zfree_pcpu(V_key_lft_zone, sav->lft_c);
3125 	}
3126 	free(sav, M_IPSEC_SA);
3127 }
3128 
3129 /*
3130  * search SAH.
3131  * OUT:
3132  *	NULL	: not found
3133  *	others	: found, referenced pointer to a SAH.
3134  */
3135 static struct secashead *
3136 key_getsah(struct secasindex *saidx)
3137 {
3138 	SAHTREE_RLOCK_TRACKER;
3139 	struct secashead *sah;
3140 
3141 	SAHTREE_RLOCK();
3142 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
3143 	    if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID) != 0) {
3144 		    SAH_ADDREF(sah);
3145 		    break;
3146 	    }
3147 	}
3148 	SAHTREE_RUNLOCK();
3149 	return (sah);
3150 }
3151 
3152 /*
3153  * Check not to be duplicated SPI.
3154  * OUT:
3155  *	0	: not found
3156  *	1	: found SA with given SPI.
3157  */
3158 static int
3159 key_checkspidup(uint32_t spi)
3160 {
3161 	SAHTREE_RLOCK_TRACKER;
3162 	struct secasvar *sav;
3163 
3164 	/* Assume SPI is in network byte order */
3165 	SAHTREE_RLOCK();
3166 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3167 		if (sav->spi == spi)
3168 			break;
3169 	}
3170 	SAHTREE_RUNLOCK();
3171 	return (sav != NULL);
3172 }
3173 
3174 /*
3175  * Search SA by SPI.
3176  * OUT:
3177  *	NULL	: not found
3178  *	others	: found, referenced pointer to a SA.
3179  */
3180 static struct secasvar *
3181 key_getsavbyspi(uint32_t spi)
3182 {
3183 	SAHTREE_RLOCK_TRACKER;
3184 	struct secasvar *sav;
3185 
3186 	/* Assume SPI is in network byte order */
3187 	SAHTREE_RLOCK();
3188 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3189 		if (sav->spi != spi)
3190 			continue;
3191 		SAV_ADDREF(sav);
3192 		break;
3193 	}
3194 	SAHTREE_RUNLOCK();
3195 	return (sav);
3196 }
3197 
3198 static int
3199 key_updatelifetimes(struct secasvar *sav, const struct sadb_msghdr *mhp)
3200 {
3201 	struct seclifetime *lft_h, *lft_s, *tmp;
3202 
3203 	/* Lifetime extension is optional, check that it is present. */
3204 	if (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3205 	    SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) {
3206 		/*
3207 		 * In case of SADB_UPDATE we may need to change
3208 		 * existing lifetimes.
3209 		 */
3210 		if (sav->state == SADB_SASTATE_MATURE) {
3211 			lft_h = lft_s = NULL;
3212 			goto reset;
3213 		}
3214 		return (0);
3215 	}
3216 	/* Both HARD and SOFT extensions must present */
3217 	if ((SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3218 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
3219 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
3220 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
3221 		ipseclog((LOG_DEBUG,
3222 		    "%s: invalid message: missing required header.\n",
3223 		    __func__));
3224 		return (EINVAL);
3225 	}
3226 	if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD) ||
3227 	    SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_SOFT)) {
3228 		ipseclog((LOG_DEBUG,
3229 		    "%s: invalid message: wrong header size.\n", __func__));
3230 		return (EINVAL);
3231 	}
3232 	lft_h = key_dup_lifemsg((const struct sadb_lifetime *)
3233 	    mhp->ext[SADB_EXT_LIFETIME_HARD], M_IPSEC_MISC);
3234 	if (lft_h == NULL) {
3235 		PFKEYSTAT_INC(in_nomem);
3236 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3237 		return (ENOBUFS);
3238 	}
3239 	lft_s = key_dup_lifemsg((const struct sadb_lifetime *)
3240 	    mhp->ext[SADB_EXT_LIFETIME_SOFT], M_IPSEC_MISC);
3241 	if (lft_s == NULL) {
3242 		PFKEYSTAT_INC(in_nomem);
3243 		free(lft_h, M_IPSEC_MISC);
3244 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3245 		return (ENOBUFS);
3246 	}
3247 reset:
3248 	if (sav->state != SADB_SASTATE_LARVAL) {
3249 		/*
3250 		 * key_update() holds reference to this SA,
3251 		 * so it won't be deleted in meanwhile.
3252 		 */
3253 		SECASVAR_LOCK(sav);
3254 		tmp = sav->lft_h;
3255 		sav->lft_h = lft_h;
3256 		lft_h = tmp;
3257 
3258 		tmp = sav->lft_s;
3259 		sav->lft_s = lft_s;
3260 		lft_s = tmp;
3261 		SECASVAR_UNLOCK(sav);
3262 		if (lft_h != NULL)
3263 			free(lft_h, M_IPSEC_MISC);
3264 		if (lft_s != NULL)
3265 			free(lft_s, M_IPSEC_MISC);
3266 		return (0);
3267 	}
3268 	/* We can update lifetime without holding a lock */
3269 	IPSEC_ASSERT(sav->lft_h == NULL, ("lft_h is already initialized\n"));
3270 	IPSEC_ASSERT(sav->lft_s == NULL, ("lft_s is already initialized\n"));
3271 	sav->lft_h = lft_h;
3272 	sav->lft_s = lft_s;
3273 	return (0);
3274 }
3275 
3276 /*
3277  * copy SA values from PF_KEY message except *SPI, SEQ, PID and TYPE*.
3278  * You must update these if need. Expects only LARVAL SAs.
3279  * OUT:	0:	success.
3280  *	!0:	failure.
3281  */
3282 static int
3283 key_setsaval(struct secasvar *sav, const struct sadb_msghdr *mhp)
3284 {
3285 	const struct sadb_sa *sa0;
3286 	const struct sadb_key *key0;
3287 	uint32_t replay;
3288 	size_t len;
3289 	int error;
3290 
3291 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
3292 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
3293 	IPSEC_ASSERT(sav->state == SADB_SASTATE_LARVAL,
3294 	    ("Attempt to update non LARVAL SA"));
3295 
3296 	/* XXX rewrite */
3297 	error = key_setident(sav->sah, mhp);
3298 	if (error != 0)
3299 		goto fail;
3300 
3301 	/* SA */
3302 	if (!SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
3303 		if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
3304 			error = EINVAL;
3305 			goto fail;
3306 		}
3307 		sa0 = (const struct sadb_sa *)mhp->ext[SADB_EXT_SA];
3308 		sav->alg_auth = sa0->sadb_sa_auth;
3309 		sav->alg_enc = sa0->sadb_sa_encrypt;
3310 		sav->flags = sa0->sadb_sa_flags;
3311 		if ((sav->flags & SADB_KEY_FLAGS_MAX) != sav->flags) {
3312 			ipseclog((LOG_DEBUG,
3313 			    "%s: invalid sa_flags 0x%08x.\n", __func__,
3314 			    sav->flags));
3315 			error = EINVAL;
3316 			goto fail;
3317 		}
3318 
3319 		/* Optional replay window */
3320 		replay = 0;
3321 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0)
3322 			replay = sa0->sadb_sa_replay;
3323 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_SA_REPLAY)) {
3324 			if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA_REPLAY)) {
3325 				error = EINVAL;
3326 				goto fail;
3327 			}
3328 			replay = ((const struct sadb_x_sa_replay *)
3329 			    mhp->ext[SADB_X_EXT_SA_REPLAY])->sadb_x_sa_replay_replay;
3330 
3331 			if (replay > UINT32_MAX - 32) {
3332 				ipseclog((LOG_DEBUG,
3333 				    "%s: replay window too big.\n", __func__));
3334 				error = EINVAL;
3335 				goto fail;
3336 			}
3337 
3338 			replay = (replay + 7) >> 3;
3339 		}
3340 
3341 		sav->replay = malloc(sizeof(struct secreplay), M_IPSEC_MISC,
3342 		    M_NOWAIT | M_ZERO);
3343 		if (sav->replay == NULL) {
3344 			PFKEYSTAT_INC(in_nomem);
3345 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3346 			    __func__));
3347 			error = ENOBUFS;
3348 			goto fail;
3349 		}
3350 
3351 		if (replay != 0) {
3352 			/* number of 32b blocks to be allocated */
3353 			uint32_t bitmap_size;
3354 
3355 			/* RFC 6479:
3356 			 * - the allocated replay window size must be
3357 			 *   a power of two.
3358 			 * - use an extra 32b block as a redundant window.
3359 			 */
3360 			bitmap_size = 1;
3361 			while (replay + 4 > bitmap_size)
3362 				bitmap_size <<= 1;
3363 			bitmap_size = bitmap_size / 4;
3364 
3365 			sav->replay->bitmap = malloc(
3366 			    bitmap_size * sizeof(uint32_t), M_IPSEC_MISC,
3367 			    M_NOWAIT | M_ZERO);
3368 			if (sav->replay->bitmap == NULL) {
3369 				PFKEYSTAT_INC(in_nomem);
3370 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3371 					__func__));
3372 				error = ENOBUFS;
3373 				goto fail;
3374 			}
3375 			sav->replay->bitmap_size = bitmap_size;
3376 			sav->replay->wsize = replay;
3377 		}
3378 	}
3379 
3380 	/* Authentication keys */
3381 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
3382 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH)) {
3383 			error = EINVAL;
3384 			goto fail;
3385 		}
3386 		error = 0;
3387 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_AUTH];
3388 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3389 		switch (mhp->msg->sadb_msg_satype) {
3390 		case SADB_SATYPE_AH:
3391 		case SADB_SATYPE_ESP:
3392 		case SADB_X_SATYPE_TCPSIGNATURE:
3393 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3394 			    sav->alg_auth != SADB_X_AALG_NULL)
3395 				error = EINVAL;
3396 			break;
3397 		case SADB_X_SATYPE_IPCOMP:
3398 		default:
3399 			error = EINVAL;
3400 			break;
3401 		}
3402 		if (error) {
3403 			ipseclog((LOG_DEBUG, "%s: invalid key_auth values.\n",
3404 				__func__));
3405 			goto fail;
3406 		}
3407 
3408 		sav->key_auth = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3409 		if (sav->key_auth == NULL ) {
3410 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3411 				  __func__));
3412 			PFKEYSTAT_INC(in_nomem);
3413 			error = ENOBUFS;
3414 			goto fail;
3415 		}
3416 	}
3417 
3418 	/* Encryption key */
3419 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) {
3420 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT)) {
3421 			error = EINVAL;
3422 			goto fail;
3423 		}
3424 		error = 0;
3425 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_ENCRYPT];
3426 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3427 		switch (mhp->msg->sadb_msg_satype) {
3428 		case SADB_SATYPE_ESP:
3429 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3430 			    sav->alg_enc != SADB_EALG_NULL) {
3431 				error = EINVAL;
3432 				break;
3433 			}
3434 			sav->key_enc = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3435 			if (sav->key_enc == NULL) {
3436 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3437 					__func__));
3438 				PFKEYSTAT_INC(in_nomem);
3439 				error = ENOBUFS;
3440 				goto fail;
3441 			}
3442 			break;
3443 		case SADB_X_SATYPE_IPCOMP:
3444 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3445 				error = EINVAL;
3446 			sav->key_enc = NULL;	/*just in case*/
3447 			break;
3448 		case SADB_SATYPE_AH:
3449 		case SADB_X_SATYPE_TCPSIGNATURE:
3450 		default:
3451 			error = EINVAL;
3452 			break;
3453 		}
3454 		if (error) {
3455 			ipseclog((LOG_DEBUG, "%s: invalid key_enc value.\n",
3456 				__func__));
3457 			goto fail;
3458 		}
3459 	}
3460 
3461 	/* set iv */
3462 	sav->ivlen = 0;
3463 	switch (mhp->msg->sadb_msg_satype) {
3464 	case SADB_SATYPE_AH:
3465 		if (sav->flags & SADB_X_EXT_DERIV) {
3466 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3467 			    "given to AH SA.\n", __func__));
3468 			error = EINVAL;
3469 			goto fail;
3470 		}
3471 		if (sav->alg_enc != SADB_EALG_NONE) {
3472 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3473 			    "mismated.\n", __func__));
3474 			error = EINVAL;
3475 			goto fail;
3476 		}
3477 		error = xform_init(sav, XF_AH);
3478 		break;
3479 	case SADB_SATYPE_ESP:
3480 		if ((sav->flags & (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) ==
3481 		    (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) {
3482 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3483 			    "given to old-esp.\n", __func__));
3484 			error = EINVAL;
3485 			goto fail;
3486 		}
3487 		error = xform_init(sav, XF_ESP);
3488 		break;
3489 	case SADB_X_SATYPE_IPCOMP:
3490 		if (sav->alg_auth != SADB_AALG_NONE) {
3491 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3492 			    "mismated.\n", __func__));
3493 			error = EINVAL;
3494 			goto fail;
3495 		}
3496 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0 &&
3497 		    ntohl(sav->spi) >= 0x10000) {
3498 			ipseclog((LOG_DEBUG, "%s: invalid cpi for IPComp.\n",
3499 			    __func__));
3500 			error = EINVAL;
3501 			goto fail;
3502 		}
3503 		error = xform_init(sav, XF_IPCOMP);
3504 		break;
3505 	case SADB_X_SATYPE_TCPSIGNATURE:
3506 		if (sav->alg_enc != SADB_EALG_NONE) {
3507 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3508 			    "mismated.\n", __func__));
3509 			error = EINVAL;
3510 			goto fail;
3511 		}
3512 		error = xform_init(sav, XF_TCPSIGNATURE);
3513 		break;
3514 	default:
3515 		ipseclog((LOG_DEBUG, "%s: Invalid satype.\n", __func__));
3516 		error = EPROTONOSUPPORT;
3517 		goto fail;
3518 	}
3519 	if (error) {
3520 		ipseclog((LOG_DEBUG, "%s: unable to initialize SA type %u.\n",
3521 		    __func__, mhp->msg->sadb_msg_satype));
3522 		goto fail;
3523 	}
3524 
3525 	/* Handle NAT-T headers */
3526 	error = key_setnatt(sav, mhp);
3527 	if (error != 0)
3528 		goto fail;
3529 
3530 	/* Initialize lifetime for CURRENT */
3531 	sav->firstused = 0;
3532 	sav->created = time_second;
3533 
3534 	/* lifetimes for HARD and SOFT */
3535 	error = key_updatelifetimes(sav, mhp);
3536 	if (error == 0)
3537 		return (0);
3538 fail:
3539 	key_cleansav(sav);
3540 	return (error);
3541 }
3542 
3543 /*
3544  * subroutine for SADB_GET and SADB_DUMP.
3545  */
3546 static struct mbuf *
3547 key_setdumpsa(struct secasvar *sav, uint8_t type, uint8_t satype,
3548     uint32_t seq, uint32_t pid)
3549 {
3550 	struct seclifetime lft_c;
3551 	struct mbuf *result = NULL, *tres = NULL, *m;
3552 	int i, dumporder[] = {
3553 		SADB_EXT_SA, SADB_X_EXT_SA2, SADB_X_EXT_SA_REPLAY,
3554 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3555 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3556 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY,
3557 		SADB_EXT_KEY_AUTH, SADB_EXT_KEY_ENCRYPT,
3558 		SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
3559 		SADB_EXT_SENSITIVITY,
3560 		SADB_X_EXT_NAT_T_TYPE,
3561 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3562 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3563 		SADB_X_EXT_NAT_T_FRAG,
3564 	};
3565 	uint32_t replay_count;
3566 
3567 	m = key_setsadbmsg(type, 0, satype, seq, pid, sav->refcnt);
3568 	if (m == NULL)
3569 		goto fail;
3570 	result = m;
3571 
3572 	for (i = nitems(dumporder) - 1; i >= 0; i--) {
3573 		m = NULL;
3574 		switch (dumporder[i]) {
3575 		case SADB_EXT_SA:
3576 			m = key_setsadbsa(sav);
3577 			if (!m)
3578 				goto fail;
3579 			break;
3580 
3581 		case SADB_X_EXT_SA2:
3582 			SECASVAR_LOCK(sav);
3583 			replay_count = sav->replay ? sav->replay->count : 0;
3584 			SECASVAR_UNLOCK(sav);
3585 			m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
3586 					sav->sah->saidx.reqid);
3587 			if (!m)
3588 				goto fail;
3589 			break;
3590 
3591 		case SADB_X_EXT_SA_REPLAY:
3592 			if (sav->replay == NULL ||
3593 			    sav->replay->wsize <= UINT8_MAX)
3594 				continue;
3595 
3596 			m = key_setsadbxsareplay(sav->replay->wsize);
3597 			if (!m)
3598 				goto fail;
3599 			break;
3600 
3601 		case SADB_EXT_ADDRESS_SRC:
3602 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3603 			    &sav->sah->saidx.src.sa,
3604 			    FULLMASK, IPSEC_ULPROTO_ANY);
3605 			if (!m)
3606 				goto fail;
3607 			break;
3608 
3609 		case SADB_EXT_ADDRESS_DST:
3610 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3611 			    &sav->sah->saidx.dst.sa,
3612 			    FULLMASK, IPSEC_ULPROTO_ANY);
3613 			if (!m)
3614 				goto fail;
3615 			break;
3616 
3617 		case SADB_EXT_KEY_AUTH:
3618 			if (!sav->key_auth)
3619 				continue;
3620 			m = key_setkey(sav->key_auth, SADB_EXT_KEY_AUTH);
3621 			if (!m)
3622 				goto fail;
3623 			break;
3624 
3625 		case SADB_EXT_KEY_ENCRYPT:
3626 			if (!sav->key_enc)
3627 				continue;
3628 			m = key_setkey(sav->key_enc, SADB_EXT_KEY_ENCRYPT);
3629 			if (!m)
3630 				goto fail;
3631 			break;
3632 
3633 		case SADB_EXT_LIFETIME_CURRENT:
3634 			lft_c.addtime = sav->created;
3635 			lft_c.allocations = (uint32_t)counter_u64_fetch(
3636 			    sav->lft_c_allocations);
3637 			lft_c.bytes = counter_u64_fetch(sav->lft_c_bytes);
3638 			lft_c.usetime = sav->firstused;
3639 			m = key_setlifetime(&lft_c, SADB_EXT_LIFETIME_CURRENT);
3640 			if (!m)
3641 				goto fail;
3642 			break;
3643 
3644 		case SADB_EXT_LIFETIME_HARD:
3645 			if (!sav->lft_h)
3646 				continue;
3647 			m = key_setlifetime(sav->lft_h,
3648 					    SADB_EXT_LIFETIME_HARD);
3649 			if (!m)
3650 				goto fail;
3651 			break;
3652 
3653 		case SADB_EXT_LIFETIME_SOFT:
3654 			if (!sav->lft_s)
3655 				continue;
3656 			m = key_setlifetime(sav->lft_s,
3657 					    SADB_EXT_LIFETIME_SOFT);
3658 
3659 			if (!m)
3660 				goto fail;
3661 			break;
3662 
3663 		case SADB_X_EXT_NAT_T_TYPE:
3664 			if (sav->natt == NULL)
3665 				continue;
3666 			m = key_setsadbxtype(UDP_ENCAP_ESPINUDP);
3667 			if (!m)
3668 				goto fail;
3669 			break;
3670 
3671 		case SADB_X_EXT_NAT_T_DPORT:
3672 			if (sav->natt == NULL)
3673 				continue;
3674 			m = key_setsadbxport(sav->natt->dport,
3675 			    SADB_X_EXT_NAT_T_DPORT);
3676 			if (!m)
3677 				goto fail;
3678 			break;
3679 
3680 		case SADB_X_EXT_NAT_T_SPORT:
3681 			if (sav->natt == NULL)
3682 				continue;
3683 			m = key_setsadbxport(sav->natt->sport,
3684 			    SADB_X_EXT_NAT_T_SPORT);
3685 			if (!m)
3686 				goto fail;
3687 			break;
3688 
3689 		case SADB_X_EXT_NAT_T_OAI:
3690 			if (sav->natt == NULL ||
3691 			    (sav->natt->flags & IPSEC_NATT_F_OAI) == 0)
3692 				continue;
3693 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAI,
3694 			    &sav->natt->oai.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3695 			if (!m)
3696 				goto fail;
3697 			break;
3698 		case SADB_X_EXT_NAT_T_OAR:
3699 			if (sav->natt == NULL ||
3700 			    (sav->natt->flags & IPSEC_NATT_F_OAR) == 0)
3701 				continue;
3702 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAR,
3703 			    &sav->natt->oar.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3704 			if (!m)
3705 				goto fail;
3706 			break;
3707 		case SADB_X_EXT_NAT_T_FRAG:
3708 			/* We do not (yet) support those. */
3709 			continue;
3710 
3711 		case SADB_EXT_ADDRESS_PROXY:
3712 		case SADB_EXT_IDENTITY_SRC:
3713 		case SADB_EXT_IDENTITY_DST:
3714 			/* XXX: should we brought from SPD ? */
3715 		case SADB_EXT_SENSITIVITY:
3716 		default:
3717 			continue;
3718 		}
3719 
3720 		if (!m)
3721 			goto fail;
3722 		if (tres)
3723 			m_cat(m, tres);
3724 		tres = m;
3725 	}
3726 
3727 	m_cat(result, tres);
3728 	tres = NULL;
3729 	if (result->m_len < sizeof(struct sadb_msg)) {
3730 		result = m_pullup(result, sizeof(struct sadb_msg));
3731 		if (result == NULL)
3732 			goto fail;
3733 	}
3734 
3735 	result->m_pkthdr.len = 0;
3736 	for (m = result; m; m = m->m_next)
3737 		result->m_pkthdr.len += m->m_len;
3738 
3739 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3740 	    PFKEY_UNIT64(result->m_pkthdr.len);
3741 
3742 	return result;
3743 
3744 fail:
3745 	m_freem(result);
3746 	m_freem(tres);
3747 	return NULL;
3748 }
3749 
3750 /*
3751  * set data into sadb_msg.
3752  */
3753 static struct mbuf *
3754 key_setsadbmsg(u_int8_t type, u_int16_t tlen, u_int8_t satype, u_int32_t seq,
3755     pid_t pid, u_int16_t reserved)
3756 {
3757 	struct mbuf *m;
3758 	struct sadb_msg *p;
3759 	int len;
3760 
3761 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
3762 	if (len > MCLBYTES)
3763 		return NULL;
3764 	MGETHDR(m, M_NOWAIT, MT_DATA);
3765 	if (m && len > MHLEN) {
3766 		if (!(MCLGET(m, M_NOWAIT))) {
3767 			m_freem(m);
3768 			m = NULL;
3769 		}
3770 	}
3771 	if (!m)
3772 		return NULL;
3773 	m->m_pkthdr.len = m->m_len = len;
3774 	m->m_next = NULL;
3775 
3776 	p = mtod(m, struct sadb_msg *);
3777 
3778 	bzero(p, len);
3779 	p->sadb_msg_version = PF_KEY_V2;
3780 	p->sadb_msg_type = type;
3781 	p->sadb_msg_errno = 0;
3782 	p->sadb_msg_satype = satype;
3783 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
3784 	p->sadb_msg_reserved = reserved;
3785 	p->sadb_msg_seq = seq;
3786 	p->sadb_msg_pid = (u_int32_t)pid;
3787 
3788 	return m;
3789 }
3790 
3791 /*
3792  * copy secasvar data into sadb_address.
3793  */
3794 static struct mbuf *
3795 key_setsadbsa(struct secasvar *sav)
3796 {
3797 	struct mbuf *m;
3798 	struct sadb_sa *p;
3799 	int len;
3800 
3801 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
3802 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3803 	if (m == NULL)
3804 		return (NULL);
3805 	m_align(m, len);
3806 	m->m_len = len;
3807 	p = mtod(m, struct sadb_sa *);
3808 	bzero(p, len);
3809 	p->sadb_sa_len = PFKEY_UNIT64(len);
3810 	p->sadb_sa_exttype = SADB_EXT_SA;
3811 	p->sadb_sa_spi = sav->spi;
3812 	p->sadb_sa_replay = sav->replay ?
3813 	    (sav->replay->wsize > UINT8_MAX ? UINT8_MAX :
3814 		sav->replay->wsize): 0;
3815 	p->sadb_sa_state = sav->state;
3816 	p->sadb_sa_auth = sav->alg_auth;
3817 	p->sadb_sa_encrypt = sav->alg_enc;
3818 	p->sadb_sa_flags = sav->flags & SADB_KEY_FLAGS_MAX;
3819 	return (m);
3820 }
3821 
3822 /*
3823  * set data into sadb_address.
3824  */
3825 static struct mbuf *
3826 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
3827     u_int8_t prefixlen, u_int16_t ul_proto)
3828 {
3829 	struct mbuf *m;
3830 	struct sadb_address *p;
3831 	size_t len;
3832 
3833 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
3834 	    PFKEY_ALIGN8(saddr->sa_len);
3835 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3836 	if (m == NULL)
3837 		return (NULL);
3838 	m_align(m, len);
3839 	m->m_len = len;
3840 	p = mtod(m, struct sadb_address *);
3841 
3842 	bzero(p, len);
3843 	p->sadb_address_len = PFKEY_UNIT64(len);
3844 	p->sadb_address_exttype = exttype;
3845 	p->sadb_address_proto = ul_proto;
3846 	if (prefixlen == FULLMASK) {
3847 		switch (saddr->sa_family) {
3848 		case AF_INET:
3849 			prefixlen = sizeof(struct in_addr) << 3;
3850 			break;
3851 		case AF_INET6:
3852 			prefixlen = sizeof(struct in6_addr) << 3;
3853 			break;
3854 		default:
3855 			; /*XXX*/
3856 		}
3857 	}
3858 	p->sadb_address_prefixlen = prefixlen;
3859 	p->sadb_address_reserved = 0;
3860 
3861 	bcopy(saddr,
3862 	    mtod(m, caddr_t) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
3863 	    saddr->sa_len);
3864 
3865 	return m;
3866 }
3867 
3868 /*
3869  * set data into sadb_x_sa2.
3870  */
3871 static struct mbuf *
3872 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int32_t reqid)
3873 {
3874 	struct mbuf *m;
3875 	struct sadb_x_sa2 *p;
3876 	size_t len;
3877 
3878 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
3879 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3880 	if (m == NULL)
3881 		return (NULL);
3882 	m_align(m, len);
3883 	m->m_len = len;
3884 	p = mtod(m, struct sadb_x_sa2 *);
3885 
3886 	bzero(p, len);
3887 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
3888 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
3889 	p->sadb_x_sa2_mode = mode;
3890 	p->sadb_x_sa2_reserved1 = 0;
3891 	p->sadb_x_sa2_reserved2 = 0;
3892 	p->sadb_x_sa2_sequence = seq;
3893 	p->sadb_x_sa2_reqid = reqid;
3894 
3895 	return m;
3896 }
3897 
3898 /*
3899  * Set data into sadb_x_sa_replay.
3900  */
3901 static struct mbuf *
3902 key_setsadbxsareplay(u_int32_t replay)
3903 {
3904 	struct mbuf *m;
3905 	struct sadb_x_sa_replay *p;
3906 	size_t len;
3907 
3908 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa_replay));
3909 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3910 	if (m == NULL)
3911 		return (NULL);
3912 	m_align(m, len);
3913 	m->m_len = len;
3914 	p = mtod(m, struct sadb_x_sa_replay *);
3915 
3916 	bzero(p, len);
3917 	p->sadb_x_sa_replay_len = PFKEY_UNIT64(len);
3918 	p->sadb_x_sa_replay_exttype = SADB_X_EXT_SA_REPLAY;
3919 	p->sadb_x_sa_replay_replay = (replay << 3);
3920 
3921 	return m;
3922 }
3923 
3924 /*
3925  * Set a type in sadb_x_nat_t_type.
3926  */
3927 static struct mbuf *
3928 key_setsadbxtype(u_int16_t type)
3929 {
3930 	struct mbuf *m;
3931 	size_t len;
3932 	struct sadb_x_nat_t_type *p;
3933 
3934 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
3935 
3936 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3937 	if (m == NULL)
3938 		return (NULL);
3939 	m_align(m, len);
3940 	m->m_len = len;
3941 	p = mtod(m, struct sadb_x_nat_t_type *);
3942 
3943 	bzero(p, len);
3944 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
3945 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
3946 	p->sadb_x_nat_t_type_type = type;
3947 
3948 	return (m);
3949 }
3950 /*
3951  * Set a port in sadb_x_nat_t_port.
3952  * In contrast to default RFC 2367 behaviour, port is in network byte order.
3953  */
3954 static struct mbuf *
3955 key_setsadbxport(u_int16_t port, u_int16_t type)
3956 {
3957 	struct mbuf *m;
3958 	size_t len;
3959 	struct sadb_x_nat_t_port *p;
3960 
3961 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
3962 
3963 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3964 	if (m == NULL)
3965 		return (NULL);
3966 	m_align(m, len);
3967 	m->m_len = len;
3968 	p = mtod(m, struct sadb_x_nat_t_port *);
3969 
3970 	bzero(p, len);
3971 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
3972 	p->sadb_x_nat_t_port_exttype = type;
3973 	p->sadb_x_nat_t_port_port = port;
3974 
3975 	return (m);
3976 }
3977 
3978 /*
3979  * Get port from sockaddr. Port is in network byte order.
3980  */
3981 uint16_t
3982 key_portfromsaddr(struct sockaddr *sa)
3983 {
3984 
3985 	switch (sa->sa_family) {
3986 #ifdef INET
3987 	case AF_INET:
3988 		return ((struct sockaddr_in *)sa)->sin_port;
3989 #endif
3990 #ifdef INET6
3991 	case AF_INET6:
3992 		return ((struct sockaddr_in6 *)sa)->sin6_port;
3993 #endif
3994 	}
3995 	return (0);
3996 }
3997 
3998 /*
3999  * Set port in struct sockaddr. Port is in network byte order.
4000  */
4001 void
4002 key_porttosaddr(struct sockaddr *sa, uint16_t port)
4003 {
4004 
4005 	switch (sa->sa_family) {
4006 #ifdef INET
4007 	case AF_INET:
4008 		((struct sockaddr_in *)sa)->sin_port = port;
4009 		break;
4010 #endif
4011 #ifdef INET6
4012 	case AF_INET6:
4013 		((struct sockaddr_in6 *)sa)->sin6_port = port;
4014 		break;
4015 #endif
4016 	default:
4017 		ipseclog((LOG_DEBUG, "%s: unexpected address family %d.\n",
4018 			__func__, sa->sa_family));
4019 		break;
4020 	}
4021 }
4022 
4023 /*
4024  * set data into sadb_x_policy
4025  */
4026 static struct mbuf *
4027 key_setsadbxpolicy(u_int16_t type, u_int8_t dir, u_int32_t id, u_int32_t priority)
4028 {
4029 	struct mbuf *m;
4030 	struct sadb_x_policy *p;
4031 	size_t len;
4032 
4033 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
4034 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4035 	if (m == NULL)
4036 		return (NULL);
4037 	m_align(m, len);
4038 	m->m_len = len;
4039 	p = mtod(m, struct sadb_x_policy *);
4040 
4041 	bzero(p, len);
4042 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
4043 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
4044 	p->sadb_x_policy_type = type;
4045 	p->sadb_x_policy_dir = dir;
4046 	p->sadb_x_policy_id = id;
4047 	p->sadb_x_policy_priority = priority;
4048 
4049 	return m;
4050 }
4051 
4052 /* %%% utilities */
4053 /* Take a key message (sadb_key) from the socket and turn it into one
4054  * of the kernel's key structures (seckey).
4055  *
4056  * IN: pointer to the src
4057  * OUT: NULL no more memory
4058  */
4059 struct seckey *
4060 key_dup_keymsg(const struct sadb_key *src, size_t len,
4061     struct malloc_type *type)
4062 {
4063 	struct seckey *dst;
4064 
4065 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4066 	if (dst != NULL) {
4067 		dst->bits = src->sadb_key_bits;
4068 		dst->key_data = malloc(len, type, M_NOWAIT);
4069 		if (dst->key_data != NULL) {
4070 			bcopy((const char *)(src + 1), dst->key_data, len);
4071 		} else {
4072 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4073 			    __func__));
4074 			free(dst, type);
4075 			dst = NULL;
4076 		}
4077 	} else {
4078 		ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4079 		    __func__));
4080 
4081 	}
4082 	return (dst);
4083 }
4084 
4085 /* Take a lifetime message (sadb_lifetime) passed in on a socket and
4086  * turn it into one of the kernel's lifetime structures (seclifetime).
4087  *
4088  * IN: pointer to the destination, source and malloc type
4089  * OUT: NULL, no more memory
4090  */
4091 
4092 static struct seclifetime *
4093 key_dup_lifemsg(const struct sadb_lifetime *src, struct malloc_type *type)
4094 {
4095 	struct seclifetime *dst;
4096 
4097 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4098 	if (dst == NULL) {
4099 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
4100 		return (NULL);
4101 	}
4102 	dst->allocations = src->sadb_lifetime_allocations;
4103 	dst->bytes = src->sadb_lifetime_bytes;
4104 	dst->addtime = src->sadb_lifetime_addtime;
4105 	dst->usetime = src->sadb_lifetime_usetime;
4106 	return (dst);
4107 }
4108 
4109 /*
4110  * compare two secasindex structure.
4111  * flag can specify to compare 2 saidxes.
4112  * compare two secasindex structure without both mode and reqid.
4113  * don't compare port.
4114  * IN:
4115  *      saidx0: source, it can be in SAD.
4116  *      saidx1: object.
4117  * OUT:
4118  *      1 : equal
4119  *      0 : not equal
4120  */
4121 static int
4122 key_cmpsaidx(const struct secasindex *saidx0, const struct secasindex *saidx1,
4123     int flag)
4124 {
4125 
4126 	/* sanity */
4127 	if (saidx0 == NULL && saidx1 == NULL)
4128 		return 1;
4129 
4130 	if (saidx0 == NULL || saidx1 == NULL)
4131 		return 0;
4132 
4133 	if (saidx0->proto != saidx1->proto)
4134 		return 0;
4135 
4136 	if (flag == CMP_EXACTLY) {
4137 		if (saidx0->mode != saidx1->mode)
4138 			return 0;
4139 		if (saidx0->reqid != saidx1->reqid)
4140 			return 0;
4141 		if (bcmp(&saidx0->src, &saidx1->src,
4142 		    saidx0->src.sa.sa_len) != 0 ||
4143 		    bcmp(&saidx0->dst, &saidx1->dst,
4144 		    saidx0->dst.sa.sa_len) != 0)
4145 			return 0;
4146 	} else {
4147 
4148 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
4149 		if (flag == CMP_MODE_REQID || flag == CMP_REQID) {
4150 			/*
4151 			 * If reqid of SPD is non-zero, unique SA is required.
4152 			 * The result must be of same reqid in this case.
4153 			 */
4154 			if (saidx1->reqid != 0 &&
4155 			    saidx0->reqid != saidx1->reqid)
4156 				return 0;
4157 		}
4158 
4159 		if (flag == CMP_MODE_REQID) {
4160 			if (saidx0->mode != IPSEC_MODE_ANY
4161 			 && saidx0->mode != saidx1->mode)
4162 				return 0;
4163 		}
4164 
4165 		if (key_sockaddrcmp(&saidx0->src.sa, &saidx1->src.sa, 0) != 0)
4166 			return 0;
4167 		if (key_sockaddrcmp(&saidx0->dst.sa, &saidx1->dst.sa, 0) != 0)
4168 			return 0;
4169 	}
4170 
4171 	return 1;
4172 }
4173 
4174 /*
4175  * compare two secindex structure exactly.
4176  * IN:
4177  *	spidx0: source, it is often in SPD.
4178  *	spidx1: object, it is often from PFKEY message.
4179  * OUT:
4180  *	1 : equal
4181  *	0 : not equal
4182  */
4183 static int
4184 key_cmpspidx_exactly(struct secpolicyindex *spidx0,
4185     struct secpolicyindex *spidx1)
4186 {
4187 	/* sanity */
4188 	if (spidx0 == NULL && spidx1 == NULL)
4189 		return 1;
4190 
4191 	if (spidx0 == NULL || spidx1 == NULL)
4192 		return 0;
4193 
4194 	if (spidx0->prefs != spidx1->prefs
4195 	 || spidx0->prefd != spidx1->prefd
4196 	 || spidx0->ul_proto != spidx1->ul_proto
4197 	 || spidx0->dir != spidx1->dir)
4198 		return 0;
4199 
4200 	return key_sockaddrcmp(&spidx0->src.sa, &spidx1->src.sa, 1) == 0 &&
4201 	       key_sockaddrcmp(&spidx0->dst.sa, &spidx1->dst.sa, 1) == 0;
4202 }
4203 
4204 /*
4205  * compare two secindex structure with mask.
4206  * IN:
4207  *	spidx0: source, it is often in SPD.
4208  *	spidx1: object, it is often from IP header.
4209  * OUT:
4210  *	1 : equal
4211  *	0 : not equal
4212  */
4213 static int
4214 key_cmpspidx_withmask(struct secpolicyindex *spidx0,
4215     struct secpolicyindex *spidx1)
4216 {
4217 	/* sanity */
4218 	if (spidx0 == NULL && spidx1 == NULL)
4219 		return 1;
4220 
4221 	if (spidx0 == NULL || spidx1 == NULL)
4222 		return 0;
4223 
4224 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4225 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4226 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4227 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
4228 		return 0;
4229 
4230 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4231 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY
4232 	 && spidx0->ul_proto != spidx1->ul_proto)
4233 		return 0;
4234 
4235 	switch (spidx0->src.sa.sa_family) {
4236 	case AF_INET:
4237 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY
4238 		 && spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
4239 			return 0;
4240 		if (!key_bbcmp(&spidx0->src.sin.sin_addr,
4241 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
4242 			return 0;
4243 		break;
4244 	case AF_INET6:
4245 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY
4246 		 && spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
4247 			return 0;
4248 		/*
4249 		 * scope_id check. if sin6_scope_id is 0, we regard it
4250 		 * as a wildcard scope, which matches any scope zone ID.
4251 		 */
4252 		if (spidx0->src.sin6.sin6_scope_id &&
4253 		    spidx1->src.sin6.sin6_scope_id &&
4254 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
4255 			return 0;
4256 		if (!key_bbcmp(&spidx0->src.sin6.sin6_addr,
4257 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
4258 			return 0;
4259 		break;
4260 	default:
4261 		/* XXX */
4262 		if (bcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
4263 			return 0;
4264 		break;
4265 	}
4266 
4267 	switch (spidx0->dst.sa.sa_family) {
4268 	case AF_INET:
4269 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY
4270 		 && spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
4271 			return 0;
4272 		if (!key_bbcmp(&spidx0->dst.sin.sin_addr,
4273 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
4274 			return 0;
4275 		break;
4276 	case AF_INET6:
4277 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY
4278 		 && spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
4279 			return 0;
4280 		/*
4281 		 * scope_id check. if sin6_scope_id is 0, we regard it
4282 		 * as a wildcard scope, which matches any scope zone ID.
4283 		 */
4284 		if (spidx0->dst.sin6.sin6_scope_id &&
4285 		    spidx1->dst.sin6.sin6_scope_id &&
4286 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
4287 			return 0;
4288 		if (!key_bbcmp(&spidx0->dst.sin6.sin6_addr,
4289 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
4290 			return 0;
4291 		break;
4292 	default:
4293 		/* XXX */
4294 		if (bcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
4295 			return 0;
4296 		break;
4297 	}
4298 
4299 	/* XXX Do we check other field ?  e.g. flowinfo */
4300 
4301 	return 1;
4302 }
4303 
4304 #ifdef satosin
4305 #undef satosin
4306 #endif
4307 #define satosin(s) ((const struct sockaddr_in *)s)
4308 #ifdef satosin6
4309 #undef satosin6
4310 #endif
4311 #define satosin6(s) ((const struct sockaddr_in6 *)s)
4312 /* returns 0 on match */
4313 int
4314 key_sockaddrcmp(const struct sockaddr *sa1, const struct sockaddr *sa2,
4315     int port)
4316 {
4317 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4318 		return 1;
4319 
4320 	switch (sa1->sa_family) {
4321 #ifdef INET
4322 	case AF_INET:
4323 		if (sa1->sa_len != sizeof(struct sockaddr_in))
4324 			return 1;
4325 		if (satosin(sa1)->sin_addr.s_addr !=
4326 		    satosin(sa2)->sin_addr.s_addr) {
4327 			return 1;
4328 		}
4329 		if (port && satosin(sa1)->sin_port != satosin(sa2)->sin_port)
4330 			return 1;
4331 		break;
4332 #endif
4333 #ifdef INET6
4334 	case AF_INET6:
4335 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4336 			return 1;	/*EINVAL*/
4337 		if (satosin6(sa1)->sin6_scope_id !=
4338 		    satosin6(sa2)->sin6_scope_id) {
4339 			return 1;
4340 		}
4341 		if (!IN6_ARE_ADDR_EQUAL(&satosin6(sa1)->sin6_addr,
4342 		    &satosin6(sa2)->sin6_addr)) {
4343 			return 1;
4344 		}
4345 		if (port &&
4346 		    satosin6(sa1)->sin6_port != satosin6(sa2)->sin6_port) {
4347 			return 1;
4348 		}
4349 		break;
4350 #endif
4351 	default:
4352 		if (bcmp(sa1, sa2, sa1->sa_len) != 0)
4353 			return 1;
4354 		break;
4355 	}
4356 
4357 	return 0;
4358 }
4359 
4360 /* returns 0 on match */
4361 int
4362 key_sockaddrcmp_withmask(const struct sockaddr *sa1,
4363     const struct sockaddr *sa2, size_t mask)
4364 {
4365 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4366 		return (1);
4367 
4368 	switch (sa1->sa_family) {
4369 #ifdef INET
4370 	case AF_INET:
4371 		return (!key_bbcmp(&satosin(sa1)->sin_addr,
4372 		    &satosin(sa2)->sin_addr, mask));
4373 #endif
4374 #ifdef INET6
4375 	case AF_INET6:
4376 		if (satosin6(sa1)->sin6_scope_id !=
4377 		    satosin6(sa2)->sin6_scope_id)
4378 			return (1);
4379 		return (!key_bbcmp(&satosin6(sa1)->sin6_addr,
4380 		    &satosin6(sa2)->sin6_addr, mask));
4381 #endif
4382 	}
4383 	return (1);
4384 }
4385 #undef satosin
4386 #undef satosin6
4387 
4388 /*
4389  * compare two buffers with mask.
4390  * IN:
4391  *	addr1: source
4392  *	addr2: object
4393  *	bits:  Number of bits to compare
4394  * OUT:
4395  *	1 : equal
4396  *	0 : not equal
4397  */
4398 static int
4399 key_bbcmp(const void *a1, const void *a2, u_int bits)
4400 {
4401 	const unsigned char *p1 = a1;
4402 	const unsigned char *p2 = a2;
4403 
4404 	/* XXX: This could be considerably faster if we compare a word
4405 	 * at a time, but it is complicated on LSB Endian machines */
4406 
4407 	/* Handle null pointers */
4408 	if (p1 == NULL || p2 == NULL)
4409 		return (p1 == p2);
4410 
4411 	while (bits >= 8) {
4412 		if (*p1++ != *p2++)
4413 			return 0;
4414 		bits -= 8;
4415 	}
4416 
4417 	if (bits > 0) {
4418 		u_int8_t mask = ~((1<<(8-bits))-1);
4419 		if ((*p1 & mask) != (*p2 & mask))
4420 			return 0;
4421 	}
4422 	return 1;	/* Match! */
4423 }
4424 
4425 static void
4426 key_flush_spd(time_t now)
4427 {
4428 	SPTREE_RLOCK_TRACKER;
4429 	struct secpolicy_list drainq;
4430 	struct secpolicy *sp, *nextsp;
4431 	u_int dir;
4432 
4433 	LIST_INIT(&drainq);
4434 	SPTREE_RLOCK();
4435 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4436 		TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
4437 			if (sp->lifetime == 0 && sp->validtime == 0)
4438 				continue;
4439 			if ((sp->lifetime &&
4440 			    now - sp->created > sp->lifetime) ||
4441 			    (sp->validtime &&
4442 			    now - sp->lastused > sp->validtime)) {
4443 				/* Hold extra reference to send SPDEXPIRE */
4444 				SP_ADDREF(sp);
4445 				LIST_INSERT_HEAD(&drainq, sp, drainq);
4446 			}
4447 		}
4448 	}
4449 	SPTREE_RUNLOCK();
4450 	if (LIST_EMPTY(&drainq))
4451 		return;
4452 
4453 	SPTREE_WLOCK();
4454 	sp = LIST_FIRST(&drainq);
4455 	while (sp != NULL) {
4456 		nextsp = LIST_NEXT(sp, drainq);
4457 		/* Check that SP is still linked */
4458 		if (sp->state != IPSEC_SPSTATE_ALIVE) {
4459 			LIST_REMOVE(sp, drainq);
4460 			key_freesp(&sp); /* release extra reference */
4461 			sp = nextsp;
4462 			continue;
4463 		}
4464 		TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
4465 		V_spd_size--;
4466 		LIST_REMOVE(sp, idhash);
4467 		sp->state = IPSEC_SPSTATE_DEAD;
4468 		sp = nextsp;
4469 	}
4470 	V_sp_genid++;
4471 	SPTREE_WUNLOCK();
4472 	if (SPDCACHE_ENABLED())
4473 		spdcache_clear();
4474 
4475 	sp = LIST_FIRST(&drainq);
4476 	while (sp != NULL) {
4477 		nextsp = LIST_NEXT(sp, drainq);
4478 		key_spdexpire(sp);
4479 		key_freesp(&sp); /* release extra reference */
4480 		key_freesp(&sp); /* release last reference */
4481 		sp = nextsp;
4482 	}
4483 }
4484 
4485 static void
4486 key_flush_sad(time_t now)
4487 {
4488 	SAHTREE_RLOCK_TRACKER;
4489 	struct secashead_list emptyq;
4490 	struct secasvar_list drainq, hexpireq, sexpireq, freeq;
4491 	struct secashead *sah, *nextsah;
4492 	struct secasvar *sav, *nextsav;
4493 
4494 	LIST_INIT(&drainq);
4495 	LIST_INIT(&hexpireq);
4496 	LIST_INIT(&sexpireq);
4497 	LIST_INIT(&emptyq);
4498 
4499 	SAHTREE_RLOCK();
4500 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
4501 		/* Check for empty SAH */
4502 		if (TAILQ_EMPTY(&sah->savtree_larval) &&
4503 		    TAILQ_EMPTY(&sah->savtree_alive)) {
4504 			SAH_ADDREF(sah);
4505 			LIST_INSERT_HEAD(&emptyq, sah, drainq);
4506 			continue;
4507 		}
4508 		/* Add all stale LARVAL SAs into drainq */
4509 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
4510 			if (now - sav->created < V_key_larval_lifetime)
4511 				continue;
4512 			SAV_ADDREF(sav);
4513 			LIST_INSERT_HEAD(&drainq, sav, drainq);
4514 		}
4515 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
4516 			/* lifetimes aren't specified */
4517 			if (sav->lft_h == NULL)
4518 				continue;
4519 			SECASVAR_LOCK(sav);
4520 			/*
4521 			 * Check again with lock held, because it may
4522 			 * be updated by SADB_UPDATE.
4523 			 */
4524 			if (sav->lft_h == NULL) {
4525 				SECASVAR_UNLOCK(sav);
4526 				continue;
4527 			}
4528 			/*
4529 			 * RFC 2367:
4530 			 * HARD lifetimes MUST take precedence over SOFT
4531 			 * lifetimes, meaning if the HARD and SOFT lifetimes
4532 			 * are the same, the HARD lifetime will appear on the
4533 			 * EXPIRE message.
4534 			 */
4535 			/* check HARD lifetime */
4536 			if ((sav->lft_h->addtime != 0 &&
4537 			    now - sav->created > sav->lft_h->addtime) ||
4538 			    (sav->lft_h->usetime != 0 && sav->firstused &&
4539 			    now - sav->firstused > sav->lft_h->usetime) ||
4540 			    (sav->lft_h->bytes != 0 && counter_u64_fetch(
4541 			        sav->lft_c_bytes) > sav->lft_h->bytes)) {
4542 				SECASVAR_UNLOCK(sav);
4543 				SAV_ADDREF(sav);
4544 				LIST_INSERT_HEAD(&hexpireq, sav, drainq);
4545 				continue;
4546 			}
4547 			/* check SOFT lifetime (only for MATURE SAs) */
4548 			if (sav->state == SADB_SASTATE_MATURE && (
4549 			    (sav->lft_s->addtime != 0 &&
4550 			    now - sav->created > sav->lft_s->addtime) ||
4551 			    (sav->lft_s->usetime != 0 && sav->firstused &&
4552 			    now - sav->firstused > sav->lft_s->usetime) ||
4553 			    (sav->lft_s->bytes != 0 && counter_u64_fetch(
4554 				sav->lft_c_bytes) > sav->lft_s->bytes))) {
4555 				SECASVAR_UNLOCK(sav);
4556 				SAV_ADDREF(sav);
4557 				LIST_INSERT_HEAD(&sexpireq, sav, drainq);
4558 				continue;
4559 			}
4560 			SECASVAR_UNLOCK(sav);
4561 		}
4562 	}
4563 	SAHTREE_RUNLOCK();
4564 
4565 	if (LIST_EMPTY(&emptyq) && LIST_EMPTY(&drainq) &&
4566 	    LIST_EMPTY(&hexpireq) && LIST_EMPTY(&sexpireq))
4567 		return;
4568 
4569 	LIST_INIT(&freeq);
4570 	SAHTREE_WLOCK();
4571 	/* Unlink stale LARVAL SAs */
4572 	sav = LIST_FIRST(&drainq);
4573 	while (sav != NULL) {
4574 		nextsav = LIST_NEXT(sav, drainq);
4575 		/* Check that SA is still LARVAL */
4576 		if (sav->state != SADB_SASTATE_LARVAL) {
4577 			LIST_REMOVE(sav, drainq);
4578 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4579 			sav = nextsav;
4580 			continue;
4581 		}
4582 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
4583 		LIST_REMOVE(sav, spihash);
4584 		sav->state = SADB_SASTATE_DEAD;
4585 		sav = nextsav;
4586 	}
4587 	/* Unlink all SAs with expired HARD lifetime */
4588 	sav = LIST_FIRST(&hexpireq);
4589 	while (sav != NULL) {
4590 		nextsav = LIST_NEXT(sav, drainq);
4591 		/* Check that SA is not unlinked */
4592 		if (sav->state == SADB_SASTATE_DEAD) {
4593 			LIST_REMOVE(sav, drainq);
4594 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4595 			sav = nextsav;
4596 			continue;
4597 		}
4598 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
4599 		LIST_REMOVE(sav, spihash);
4600 		sav->state = SADB_SASTATE_DEAD;
4601 		sav = nextsav;
4602 	}
4603 	/* Mark all SAs with expired SOFT lifetime as DYING */
4604 	sav = LIST_FIRST(&sexpireq);
4605 	while (sav != NULL) {
4606 		nextsav = LIST_NEXT(sav, drainq);
4607 		/* Check that SA is not unlinked */
4608 		if (sav->state == SADB_SASTATE_DEAD) {
4609 			LIST_REMOVE(sav, drainq);
4610 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4611 			sav = nextsav;
4612 			continue;
4613 		}
4614 		/*
4615 		 * NOTE: this doesn't change SA order in the chain.
4616 		 */
4617 		sav->state = SADB_SASTATE_DYING;
4618 		sav = nextsav;
4619 	}
4620 	/* Unlink empty SAHs */
4621 	sah = LIST_FIRST(&emptyq);
4622 	while (sah != NULL) {
4623 		nextsah = LIST_NEXT(sah, drainq);
4624 		/* Check that SAH is still empty and not unlinked */
4625 		if (sah->state == SADB_SASTATE_DEAD ||
4626 		    !TAILQ_EMPTY(&sah->savtree_larval) ||
4627 		    !TAILQ_EMPTY(&sah->savtree_alive)) {
4628 			LIST_REMOVE(sah, drainq);
4629 			key_freesah(&sah); /* release extra reference */
4630 			sah = nextsah;
4631 			continue;
4632 		}
4633 		TAILQ_REMOVE(&V_sahtree, sah, chain);
4634 		LIST_REMOVE(sah, addrhash);
4635 		sah->state = SADB_SASTATE_DEAD;
4636 		sah = nextsah;
4637 	}
4638 	SAHTREE_WUNLOCK();
4639 
4640 	/* Send SPDEXPIRE messages */
4641 	sav = LIST_FIRST(&hexpireq);
4642 	while (sav != NULL) {
4643 		nextsav = LIST_NEXT(sav, drainq);
4644 		key_expire(sav, 1);
4645 		key_freesah(&sav->sah); /* release reference from SAV */
4646 		key_freesav(&sav); /* release extra reference */
4647 		key_freesav(&sav); /* release last reference */
4648 		sav = nextsav;
4649 	}
4650 	sav = LIST_FIRST(&sexpireq);
4651 	while (sav != NULL) {
4652 		nextsav = LIST_NEXT(sav, drainq);
4653 		key_expire(sav, 0);
4654 		key_freesav(&sav); /* release extra reference */
4655 		sav = nextsav;
4656 	}
4657 	/* Free stale LARVAL SAs */
4658 	sav = LIST_FIRST(&drainq);
4659 	while (sav != NULL) {
4660 		nextsav = LIST_NEXT(sav, drainq);
4661 		key_freesah(&sav->sah); /* release reference from SAV */
4662 		key_freesav(&sav); /* release extra reference */
4663 		key_freesav(&sav); /* release last reference */
4664 		sav = nextsav;
4665 	}
4666 	/* Free SAs that were unlinked/changed by someone else */
4667 	sav = LIST_FIRST(&freeq);
4668 	while (sav != NULL) {
4669 		nextsav = LIST_NEXT(sav, drainq);
4670 		key_freesav(&sav); /* release extra reference */
4671 		sav = nextsav;
4672 	}
4673 	/* Free empty SAH */
4674 	sah = LIST_FIRST(&emptyq);
4675 	while (sah != NULL) {
4676 		nextsah = LIST_NEXT(sah, drainq);
4677 		key_freesah(&sah); /* release extra reference */
4678 		key_freesah(&sah); /* release last reference */
4679 		sah = nextsah;
4680 	}
4681 }
4682 
4683 static void
4684 key_flush_acq(time_t now)
4685 {
4686 	struct secacq *acq, *nextacq;
4687 
4688 	/* ACQ tree */
4689 	ACQ_LOCK();
4690 	acq = LIST_FIRST(&V_acqtree);
4691 	while (acq != NULL) {
4692 		nextacq = LIST_NEXT(acq, chain);
4693 		if (now - acq->created > V_key_blockacq_lifetime) {
4694 			LIST_REMOVE(acq, chain);
4695 			LIST_REMOVE(acq, addrhash);
4696 			LIST_REMOVE(acq, seqhash);
4697 			free(acq, M_IPSEC_SAQ);
4698 		}
4699 		acq = nextacq;
4700 	}
4701 	ACQ_UNLOCK();
4702 }
4703 
4704 static void
4705 key_flush_spacq(time_t now)
4706 {
4707 	struct secspacq *acq, *nextacq;
4708 
4709 	/* SP ACQ tree */
4710 	SPACQ_LOCK();
4711 	for (acq = LIST_FIRST(&V_spacqtree); acq != NULL; acq = nextacq) {
4712 		nextacq = LIST_NEXT(acq, chain);
4713 		if (now - acq->created > V_key_blockacq_lifetime
4714 		 && __LIST_CHAINED(acq)) {
4715 			LIST_REMOVE(acq, chain);
4716 			free(acq, M_IPSEC_SAQ);
4717 		}
4718 	}
4719 	SPACQ_UNLOCK();
4720 }
4721 
4722 /*
4723  * time handler.
4724  * scanning SPD and SAD to check status for each entries,
4725  * and do to remove or to expire.
4726  * XXX: year 2038 problem may remain.
4727  */
4728 static void
4729 key_timehandler(void *arg)
4730 {
4731 	VNET_ITERATOR_DECL(vnet_iter);
4732 	time_t now = time_second;
4733 
4734 	VNET_LIST_RLOCK_NOSLEEP();
4735 	VNET_FOREACH(vnet_iter) {
4736 		CURVNET_SET(vnet_iter);
4737 		key_flush_spd(now);
4738 		key_flush_sad(now);
4739 		key_flush_acq(now);
4740 		key_flush_spacq(now);
4741 		CURVNET_RESTORE();
4742 	}
4743 	VNET_LIST_RUNLOCK_NOSLEEP();
4744 
4745 #ifndef IPSEC_DEBUG2
4746 	/* do exchange to tick time !! */
4747 	callout_schedule(&key_timer, hz);
4748 #endif /* IPSEC_DEBUG2 */
4749 }
4750 
4751 u_long
4752 key_random()
4753 {
4754 	u_long value;
4755 
4756 	arc4random_buf(&value, sizeof(value));
4757 	return value;
4758 }
4759 
4760 /*
4761  * map SADB_SATYPE_* to IPPROTO_*.
4762  * if satype == SADB_SATYPE then satype is mapped to ~0.
4763  * OUT:
4764  *	0: invalid satype.
4765  */
4766 static uint8_t
4767 key_satype2proto(uint8_t satype)
4768 {
4769 	switch (satype) {
4770 	case SADB_SATYPE_UNSPEC:
4771 		return IPSEC_PROTO_ANY;
4772 	case SADB_SATYPE_AH:
4773 		return IPPROTO_AH;
4774 	case SADB_SATYPE_ESP:
4775 		return IPPROTO_ESP;
4776 	case SADB_X_SATYPE_IPCOMP:
4777 		return IPPROTO_IPCOMP;
4778 	case SADB_X_SATYPE_TCPSIGNATURE:
4779 		return IPPROTO_TCP;
4780 	default:
4781 		return 0;
4782 	}
4783 	/* NOTREACHED */
4784 }
4785 
4786 /*
4787  * map IPPROTO_* to SADB_SATYPE_*
4788  * OUT:
4789  *	0: invalid protocol type.
4790  */
4791 static uint8_t
4792 key_proto2satype(uint8_t proto)
4793 {
4794 	switch (proto) {
4795 	case IPPROTO_AH:
4796 		return SADB_SATYPE_AH;
4797 	case IPPROTO_ESP:
4798 		return SADB_SATYPE_ESP;
4799 	case IPPROTO_IPCOMP:
4800 		return SADB_X_SATYPE_IPCOMP;
4801 	case IPPROTO_TCP:
4802 		return SADB_X_SATYPE_TCPSIGNATURE;
4803 	default:
4804 		return 0;
4805 	}
4806 	/* NOTREACHED */
4807 }
4808 
4809 /* %%% PF_KEY */
4810 /*
4811  * SADB_GETSPI processing is to receive
4812  *	<base, (SA2), src address, dst address, (SPI range)>
4813  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
4814  * tree with the status of LARVAL, and send
4815  *	<base, SA(*), address(SD)>
4816  * to the IKMPd.
4817  *
4818  * IN:	mhp: pointer to the pointer to each header.
4819  * OUT:	NULL if fail.
4820  *	other if success, return pointer to the message to send.
4821  */
4822 static int
4823 key_getspi(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
4824 {
4825 	struct secasindex saidx;
4826 	struct sadb_address *src0, *dst0;
4827 	struct secasvar *sav;
4828 	uint32_t reqid, spi;
4829 	int error;
4830 	uint8_t mode, proto;
4831 
4832 	IPSEC_ASSERT(so != NULL, ("null socket"));
4833 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
4834 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
4835 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
4836 
4837 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
4838 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)
4839 #ifdef PFKEY_STRICT_CHECKS
4840 	    || SADB_CHECKHDR(mhp, SADB_EXT_SPIRANGE)
4841 #endif
4842 	    ) {
4843 		ipseclog((LOG_DEBUG,
4844 		    "%s: invalid message: missing required header.\n",
4845 		    __func__));
4846 		error = EINVAL;
4847 		goto fail;
4848 	}
4849 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
4850 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)
4851 #ifdef PFKEY_STRICT_CHECKS
4852 	    || SADB_CHECKLEN(mhp, SADB_EXT_SPIRANGE)
4853 #endif
4854 	    ) {
4855 		ipseclog((LOG_DEBUG,
4856 		    "%s: invalid message: wrong header size.\n", __func__));
4857 		error = EINVAL;
4858 		goto fail;
4859 	}
4860 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
4861 		mode = IPSEC_MODE_ANY;
4862 		reqid = 0;
4863 	} else {
4864 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
4865 			ipseclog((LOG_DEBUG,
4866 			    "%s: invalid message: wrong header size.\n",
4867 			    __func__));
4868 			error = EINVAL;
4869 			goto fail;
4870 		}
4871 		mode = ((struct sadb_x_sa2 *)
4872 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
4873 		reqid = ((struct sadb_x_sa2 *)
4874 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
4875 	}
4876 
4877 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
4878 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
4879 
4880 	/* map satype to proto */
4881 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
4882 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
4883 			__func__));
4884 		error = EINVAL;
4885 		goto fail;
4886 	}
4887 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
4888 	    (struct sockaddr *)(dst0 + 1));
4889 	if (error != 0) {
4890 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
4891 		error = EINVAL;
4892 		goto fail;
4893 	}
4894 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
4895 
4896 	/* SPI allocation */
4897 	spi = key_do_getnewspi(
4898 	    (struct sadb_spirange *)mhp->ext[SADB_EXT_SPIRANGE], &saidx);
4899 	if (spi == 0) {
4900 		/*
4901 		 * Requested SPI or SPI range is not available or
4902 		 * already used.
4903 		 */
4904 		error = EEXIST;
4905 		goto fail;
4906 	}
4907 	sav = key_newsav(mhp, &saidx, spi, &error);
4908 	if (sav == NULL)
4909 		goto fail;
4910 
4911 	if (sav->seq != 0) {
4912 		/*
4913 		 * RFC2367:
4914 		 * If the SADB_GETSPI message is in response to a
4915 		 * kernel-generated SADB_ACQUIRE, the sadb_msg_seq
4916 		 * MUST be the same as the SADB_ACQUIRE message.
4917 		 *
4918 		 * XXXAE: However it doesn't definethe behaviour how to
4919 		 * check this and what to do if it doesn't match.
4920 		 * Also what we should do if it matches?
4921 		 *
4922 		 * We can compare saidx used in SADB_ACQUIRE with saidx
4923 		 * used in SADB_GETSPI, but this probably can break
4924 		 * existing software. For now just warn if it doesn't match.
4925 		 *
4926 		 * XXXAE: anyway it looks useless.
4927 		 */
4928 		key_acqdone(&saidx, sav->seq);
4929 	}
4930 	KEYDBG(KEY_STAMP,
4931 	    printf("%s: SA(%p)\n", __func__, sav));
4932 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
4933 
4934     {
4935 	struct mbuf *n, *nn;
4936 	struct sadb_sa *m_sa;
4937 	struct sadb_msg *newmsg;
4938 	int off, len;
4939 
4940 	/* create new sadb_msg to reply. */
4941 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
4942 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
4943 
4944 	MGETHDR(n, M_NOWAIT, MT_DATA);
4945 	if (len > MHLEN) {
4946 		if (!(MCLGET(n, M_NOWAIT))) {
4947 			m_freem(n);
4948 			n = NULL;
4949 		}
4950 	}
4951 	if (!n) {
4952 		error = ENOBUFS;
4953 		goto fail;
4954 	}
4955 
4956 	n->m_len = len;
4957 	n->m_next = NULL;
4958 	off = 0;
4959 
4960 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
4961 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
4962 
4963 	m_sa = (struct sadb_sa *)(mtod(n, caddr_t) + off);
4964 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
4965 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
4966 	m_sa->sadb_sa_spi = spi; /* SPI is already in network byte order */
4967 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
4968 
4969 	IPSEC_ASSERT(off == len,
4970 		("length inconsistency (off %u len %u)", off, len));
4971 
4972 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
4973 	    SADB_EXT_ADDRESS_DST);
4974 	if (!n->m_next) {
4975 		m_freem(n);
4976 		error = ENOBUFS;
4977 		goto fail;
4978 	}
4979 
4980 	if (n->m_len < sizeof(struct sadb_msg)) {
4981 		n = m_pullup(n, sizeof(struct sadb_msg));
4982 		if (n == NULL)
4983 			return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
4984 	}
4985 
4986 	n->m_pkthdr.len = 0;
4987 	for (nn = n; nn; nn = nn->m_next)
4988 		n->m_pkthdr.len += nn->m_len;
4989 
4990 	newmsg = mtod(n, struct sadb_msg *);
4991 	newmsg->sadb_msg_seq = sav->seq;
4992 	newmsg->sadb_msg_errno = 0;
4993 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
4994 
4995 	m_freem(m);
4996 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
4997     }
4998 
4999 fail:
5000 	return (key_senderror(so, m, error));
5001 }
5002 
5003 /*
5004  * allocating new SPI
5005  * called by key_getspi().
5006  * OUT:
5007  *	0:	failure.
5008  *	others: success, SPI in network byte order.
5009  */
5010 static uint32_t
5011 key_do_getnewspi(struct sadb_spirange *spirange, struct secasindex *saidx)
5012 {
5013 	uint32_t min, max, newspi, t;
5014 	int count = V_key_spi_trycnt;
5015 
5016 	/* set spi range to allocate */
5017 	if (spirange != NULL) {
5018 		min = spirange->sadb_spirange_min;
5019 		max = spirange->sadb_spirange_max;
5020 	} else {
5021 		min = V_key_spi_minval;
5022 		max = V_key_spi_maxval;
5023 	}
5024 	/* IPCOMP needs 2-byte SPI */
5025 	if (saidx->proto == IPPROTO_IPCOMP) {
5026 		if (min >= 0x10000)
5027 			min = 0xffff;
5028 		if (max >= 0x10000)
5029 			max = 0xffff;
5030 		if (min > max) {
5031 			t = min; min = max; max = t;
5032 		}
5033 	}
5034 
5035 	if (min == max) {
5036 		if (!key_checkspidup(htonl(min))) {
5037 			ipseclog((LOG_DEBUG, "%s: SPI %u exists already.\n",
5038 			    __func__, min));
5039 			return 0;
5040 		}
5041 
5042 		count--; /* taking one cost. */
5043 		newspi = min;
5044 	} else {
5045 
5046 		/* init SPI */
5047 		newspi = 0;
5048 
5049 		/* when requesting to allocate spi ranged */
5050 		while (count--) {
5051 			/* generate pseudo-random SPI value ranged. */
5052 			newspi = min + (key_random() % (max - min + 1));
5053 			if (!key_checkspidup(htonl(newspi)))
5054 				break;
5055 		}
5056 
5057 		if (count == 0 || newspi == 0) {
5058 			ipseclog((LOG_DEBUG,
5059 			    "%s: failed to allocate SPI.\n", __func__));
5060 			return 0;
5061 		}
5062 	}
5063 
5064 	/* statistics */
5065 	keystat.getspi_count =
5066 	    (keystat.getspi_count + V_key_spi_trycnt - count) / 2;
5067 
5068 	return (htonl(newspi));
5069 }
5070 
5071 /*
5072  * Find TCP-MD5 SA with corresponding secasindex.
5073  * If not found, return NULL and fill SPI with usable value if needed.
5074  */
5075 static struct secasvar *
5076 key_getsav_tcpmd5(struct secasindex *saidx, uint32_t *spi)
5077 {
5078 	SAHTREE_RLOCK_TRACKER;
5079 	struct secashead *sah;
5080 	struct secasvar *sav;
5081 
5082 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP, ("wrong proto"));
5083 	SAHTREE_RLOCK();
5084 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
5085 		if (sah->saidx.proto != IPPROTO_TCP)
5086 			continue;
5087 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
5088 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
5089 			break;
5090 	}
5091 	if (sah != NULL) {
5092 		if (V_key_preferred_oldsa)
5093 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
5094 		else
5095 			sav = TAILQ_FIRST(&sah->savtree_alive);
5096 		if (sav != NULL) {
5097 			SAV_ADDREF(sav);
5098 			SAHTREE_RUNLOCK();
5099 			return (sav);
5100 		}
5101 	}
5102 	if (spi == NULL) {
5103 		/* No SPI required */
5104 		SAHTREE_RUNLOCK();
5105 		return (NULL);
5106 	}
5107 	/* Check that SPI is unique */
5108 	LIST_FOREACH(sav, SAVHASH_HASH(*spi), spihash) {
5109 		if (sav->spi == *spi)
5110 			break;
5111 	}
5112 	if (sav == NULL) {
5113 		SAHTREE_RUNLOCK();
5114 		/* SPI is already unique */
5115 		return (NULL);
5116 	}
5117 	SAHTREE_RUNLOCK();
5118 	/* XXX: not optimal */
5119 	*spi = key_do_getnewspi(NULL, saidx);
5120 	return (NULL);
5121 }
5122 
5123 static int
5124 key_updateaddresses(struct socket *so, struct mbuf *m,
5125     const struct sadb_msghdr *mhp, struct secasvar *sav,
5126     struct secasindex *saidx)
5127 {
5128 	struct sockaddr *newaddr;
5129 	struct secashead *sah;
5130 	struct secasvar *newsav, *tmp;
5131 	struct mbuf *n;
5132 	int error, isnew;
5133 
5134 	/* Check that we need to change SAH */
5135 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC)) {
5136 		newaddr = (struct sockaddr *)(
5137 		    ((struct sadb_address *)
5138 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_SRC]) + 1);
5139 		bcopy(newaddr, &saidx->src, newaddr->sa_len);
5140 		key_porttosaddr(&saidx->src.sa, 0);
5141 	}
5142 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5143 		newaddr = (struct sockaddr *)(
5144 		    ((struct sadb_address *)
5145 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_DST]) + 1);
5146 		bcopy(newaddr, &saidx->dst, newaddr->sa_len);
5147 		key_porttosaddr(&saidx->dst.sa, 0);
5148 	}
5149 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5150 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5151 		error = key_checksockaddrs(&saidx->src.sa, &saidx->dst.sa);
5152 		if (error != 0) {
5153 			ipseclog((LOG_DEBUG, "%s: invalid new sockaddr.\n",
5154 			    __func__));
5155 			return (error);
5156 		}
5157 
5158 		sah = key_getsah(saidx);
5159 		if (sah == NULL) {
5160 			/* create a new SA index */
5161 			sah = key_newsah(saidx);
5162 			if (sah == NULL) {
5163 				ipseclog((LOG_DEBUG,
5164 				    "%s: No more memory.\n", __func__));
5165 				return (ENOBUFS);
5166 			}
5167 			isnew = 2; /* SAH is new */
5168 		} else
5169 			isnew = 1; /* existing SAH is referenced */
5170 	} else {
5171 		/*
5172 		 * src and dst addresses are still the same.
5173 		 * Do we want to change NAT-T config?
5174 		 */
5175 		if (sav->sah->saidx.proto != IPPROTO_ESP ||
5176 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5177 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5178 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5179 			ipseclog((LOG_DEBUG,
5180 			    "%s: invalid message: missing required header.\n",
5181 			    __func__));
5182 			return (EINVAL);
5183 		}
5184 		/* We hold reference to SA, thus SAH will be referenced too. */
5185 		sah = sav->sah;
5186 		isnew = 0;
5187 	}
5188 
5189 	newsav = malloc(sizeof(struct secasvar), M_IPSEC_SA,
5190 	    M_NOWAIT | M_ZERO);
5191 	if (newsav == NULL) {
5192 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5193 		error = ENOBUFS;
5194 		goto fail;
5195 	}
5196 
5197 	/* Clone SA's content into newsav */
5198 	SAV_INITREF(newsav);
5199 	bcopy(sav, newsav, offsetof(struct secasvar, chain));
5200 	/*
5201 	 * We create new NAT-T config if it is needed.
5202 	 * Old NAT-T config will be freed by key_cleansav() when
5203 	 * last reference to SA will be released.
5204 	 */
5205 	newsav->natt = NULL;
5206 	newsav->sah = sah;
5207 	newsav->state = SADB_SASTATE_MATURE;
5208 	error = key_setnatt(newsav, mhp);
5209 	if (error != 0)
5210 		goto fail;
5211 
5212 	SAHTREE_WLOCK();
5213 	/* Check that SA is still alive */
5214 	if (sav->state == SADB_SASTATE_DEAD) {
5215 		/* SA was unlinked */
5216 		SAHTREE_WUNLOCK();
5217 		error = ESRCH;
5218 		goto fail;
5219 	}
5220 
5221 	/* Unlink SA from SAH and SPI hash */
5222 	IPSEC_ASSERT((sav->flags & SADB_X_EXT_F_CLONED) == 0,
5223 	    ("SA is already cloned"));
5224 	IPSEC_ASSERT(sav->state == SADB_SASTATE_MATURE ||
5225 	    sav->state == SADB_SASTATE_DYING,
5226 	    ("Wrong SA state %u\n", sav->state));
5227 	TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
5228 	LIST_REMOVE(sav, spihash);
5229 	sav->state = SADB_SASTATE_DEAD;
5230 
5231 	/*
5232 	 * Link new SA with SAH. Keep SAs ordered by
5233 	 * create time (newer are first).
5234 	 */
5235 	TAILQ_FOREACH(tmp, &sah->savtree_alive, chain) {
5236 		if (newsav->created > tmp->created) {
5237 			TAILQ_INSERT_BEFORE(tmp, newsav, chain);
5238 			break;
5239 		}
5240 	}
5241 	if (tmp == NULL)
5242 		TAILQ_INSERT_TAIL(&sah->savtree_alive, newsav, chain);
5243 
5244 	/* Add new SA into SPI hash. */
5245 	LIST_INSERT_HEAD(SAVHASH_HASH(newsav->spi), newsav, spihash);
5246 
5247 	/* Add new SAH into SADB. */
5248 	if (isnew == 2) {
5249 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
5250 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
5251 		sah->state = SADB_SASTATE_MATURE;
5252 		SAH_ADDREF(sah); /* newsav references new SAH */
5253 	}
5254 	/*
5255 	 * isnew == 1 -> @sah was referenced by key_getsah().
5256 	 * isnew == 0 -> we use the same @sah, that was used by @sav,
5257 	 *	and we use its reference for @newsav.
5258 	 */
5259 	SECASVAR_LOCK(sav);
5260 	/* XXX: replace cntr with pointer? */
5261 	newsav->cntr = sav->cntr;
5262 	sav->flags |= SADB_X_EXT_F_CLONED;
5263 	SECASVAR_UNLOCK(sav);
5264 
5265 	SAHTREE_WUNLOCK();
5266 
5267 	KEYDBG(KEY_STAMP,
5268 	    printf("%s: SA(%p) cloned into SA(%p)\n",
5269 	    __func__, sav, newsav));
5270 	KEYDBG(KEY_DATA, kdebug_secasv(newsav));
5271 
5272 	key_freesav(&sav); /* release last reference */
5273 
5274 	/* set msg buf from mhp */
5275 	n = key_getmsgbuf_x1(m, mhp);
5276 	if (n == NULL) {
5277 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5278 		return (ENOBUFS);
5279 	}
5280 	m_freem(m);
5281 	key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5282 	return (0);
5283 fail:
5284 	if (isnew != 0)
5285 		key_freesah(&sah);
5286 	if (newsav != NULL) {
5287 		if (newsav->natt != NULL)
5288 			free(newsav->natt, M_IPSEC_MISC);
5289 		free(newsav, M_IPSEC_SA);
5290 	}
5291 	return (error);
5292 }
5293 
5294 /*
5295  * SADB_UPDATE processing
5296  * receive
5297  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5298  *       key(AE), (identity(SD),) (sensitivity)>
5299  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5300  * and send
5301  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5302  *       (identity(SD),) (sensitivity)>
5303  * to the ikmpd.
5304  *
5305  * m will always be freed.
5306  */
5307 static int
5308 key_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5309 {
5310 	struct secasindex saidx;
5311 	struct sadb_address *src0, *dst0;
5312 	struct sadb_sa *sa0;
5313 	struct secasvar *sav;
5314 	uint32_t reqid;
5315 	int error;
5316 	uint8_t mode, proto;
5317 
5318 	IPSEC_ASSERT(so != NULL, ("null socket"));
5319 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5320 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5321 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5322 
5323 	/* map satype to proto */
5324 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5325 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5326 		    __func__));
5327 		return key_senderror(so, m, EINVAL);
5328 	}
5329 
5330 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5331 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5332 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5333 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5334 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5335 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5336 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5337 		ipseclog((LOG_DEBUG,
5338 		    "%s: invalid message: missing required header.\n",
5339 		    __func__));
5340 		return key_senderror(so, m, EINVAL);
5341 	}
5342 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5343 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5344 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5345 		ipseclog((LOG_DEBUG,
5346 		    "%s: invalid message: wrong header size.\n", __func__));
5347 		return key_senderror(so, m, EINVAL);
5348 	}
5349 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5350 		mode = IPSEC_MODE_ANY;
5351 		reqid = 0;
5352 	} else {
5353 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5354 			ipseclog((LOG_DEBUG,
5355 			    "%s: invalid message: wrong header size.\n",
5356 			    __func__));
5357 			return key_senderror(so, m, EINVAL);
5358 		}
5359 		mode = ((struct sadb_x_sa2 *)
5360 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5361 		reqid = ((struct sadb_x_sa2 *)
5362 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5363 	}
5364 
5365 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5366 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5367 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5368 
5369 	/*
5370 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5371 	 * SADB_UPDATE message.
5372 	 */
5373 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5374 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5375 #ifdef PFKEY_STRICT_CHECKS
5376 		return key_senderror(so, m, EINVAL);
5377 #endif
5378 	}
5379 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5380 	    (struct sockaddr *)(dst0 + 1));
5381 	if (error != 0) {
5382 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5383 		return key_senderror(so, m, error);
5384 	}
5385 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5386 	sav = key_getsavbyspi(sa0->sadb_sa_spi);
5387 	if (sav == NULL) {
5388 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u\n",
5389 		    __func__, ntohl(sa0->sadb_sa_spi)));
5390 		return key_senderror(so, m, EINVAL);
5391 	}
5392 	/*
5393 	 * Check that SADB_UPDATE issued by the same process that did
5394 	 * SADB_GETSPI or SADB_ADD.
5395 	 */
5396 	if (sav->pid != mhp->msg->sadb_msg_pid) {
5397 		ipseclog((LOG_DEBUG,
5398 		    "%s: pid mismatched (SPI %u, pid %u vs. %u)\n", __func__,
5399 		    ntohl(sav->spi), sav->pid, mhp->msg->sadb_msg_pid));
5400 		key_freesav(&sav);
5401 		return key_senderror(so, m, EINVAL);
5402 	}
5403 	/* saidx should match with SA. */
5404 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_MODE_REQID) == 0) {
5405 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u\n",
5406 		    __func__, ntohl(sav->spi)));
5407 		key_freesav(&sav);
5408 		return key_senderror(so, m, ESRCH);
5409 	}
5410 
5411 	if (sav->state == SADB_SASTATE_LARVAL) {
5412 		if ((mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5413 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) ||
5414 		    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5415 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH))) {
5416 			ipseclog((LOG_DEBUG,
5417 			    "%s: invalid message: missing required header.\n",
5418 			    __func__));
5419 			key_freesav(&sav);
5420 			return key_senderror(so, m, EINVAL);
5421 		}
5422 		/*
5423 		 * We can set any values except src, dst and SPI.
5424 		 */
5425 		error = key_setsaval(sav, mhp);
5426 		if (error != 0) {
5427 			key_freesav(&sav);
5428 			return (key_senderror(so, m, error));
5429 		}
5430 		/* Change SA state to MATURE */
5431 		SAHTREE_WLOCK();
5432 		if (sav->state != SADB_SASTATE_LARVAL) {
5433 			/* SA was deleted or another thread made it MATURE. */
5434 			SAHTREE_WUNLOCK();
5435 			key_freesav(&sav);
5436 			return (key_senderror(so, m, ESRCH));
5437 		}
5438 		/*
5439 		 * NOTE: we keep SAs in savtree_alive ordered by created
5440 		 * time. When SA's state changed from LARVAL to MATURE,
5441 		 * we update its created time in key_setsaval() and move
5442 		 * it into head of savtree_alive.
5443 		 */
5444 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
5445 		TAILQ_INSERT_HEAD(&sav->sah->savtree_alive, sav, chain);
5446 		sav->state = SADB_SASTATE_MATURE;
5447 		SAHTREE_WUNLOCK();
5448 	} else {
5449 		/*
5450 		 * For DYING and MATURE SA we can change only state
5451 		 * and lifetimes. Report EINVAL if something else attempted
5452 		 * to change.
5453 		 */
5454 		if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5455 		    !SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
5456 			key_freesav(&sav);
5457 			return (key_senderror(so, m, EINVAL));
5458 		}
5459 		error = key_updatelifetimes(sav, mhp);
5460 		if (error != 0) {
5461 			key_freesav(&sav);
5462 			return (key_senderror(so, m, error));
5463 		}
5464 		/*
5465 		 * This is FreeBSD extension to RFC2367.
5466 		 * IKEd can specify SADB_X_EXT_NEW_ADDRESS_SRC and/or
5467 		 * SADB_X_EXT_NEW_ADDRESS_DST when it wants to change
5468 		 * SA addresses (for example to implement MOBIKE protocol
5469 		 * as described in RFC4555). Also we allow to change
5470 		 * NAT-T config.
5471 		 */
5472 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5473 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST) ||
5474 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5475 		    sav->natt != NULL) {
5476 			error = key_updateaddresses(so, m, mhp, sav, &saidx);
5477 			key_freesav(&sav);
5478 			if (error != 0)
5479 				return (key_senderror(so, m, error));
5480 			return (0);
5481 		}
5482 		/* Check that SA is still alive */
5483 		SAHTREE_WLOCK();
5484 		if (sav->state == SADB_SASTATE_DEAD) {
5485 			/* SA was unlinked */
5486 			SAHTREE_WUNLOCK();
5487 			key_freesav(&sav);
5488 			return (key_senderror(so, m, ESRCH));
5489 		}
5490 		/*
5491 		 * NOTE: there is possible state moving from DYING to MATURE,
5492 		 * but this doesn't change created time, so we won't reorder
5493 		 * this SA.
5494 		 */
5495 		sav->state = SADB_SASTATE_MATURE;
5496 		SAHTREE_WUNLOCK();
5497 	}
5498 	KEYDBG(KEY_STAMP,
5499 	    printf("%s: SA(%p)\n", __func__, sav));
5500 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5501 	key_freesav(&sav);
5502 
5503     {
5504 	struct mbuf *n;
5505 
5506 	/* set msg buf from mhp */
5507 	n = key_getmsgbuf_x1(m, mhp);
5508 	if (n == NULL) {
5509 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5510 		return key_senderror(so, m, ENOBUFS);
5511 	}
5512 
5513 	m_freem(m);
5514 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5515     }
5516 }
5517 
5518 /*
5519  * SADB_ADD processing
5520  * add an entry to SA database, when received
5521  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5522  *       key(AE), (identity(SD),) (sensitivity)>
5523  * from the ikmpd,
5524  * and send
5525  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5526  *       (identity(SD),) (sensitivity)>
5527  * to the ikmpd.
5528  *
5529  * IGNORE identity and sensitivity messages.
5530  *
5531  * m will always be freed.
5532  */
5533 static int
5534 key_add(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5535 {
5536 	struct secasindex saidx;
5537 	struct sadb_address *src0, *dst0;
5538 	struct sadb_sa *sa0;
5539 	struct secasvar *sav;
5540 	uint32_t reqid, spi;
5541 	uint8_t mode, proto;
5542 	int error;
5543 
5544 	IPSEC_ASSERT(so != NULL, ("null socket"));
5545 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5546 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5547 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5548 
5549 	/* map satype to proto */
5550 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5551 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5552 		    __func__));
5553 		return key_senderror(so, m, EINVAL);
5554 	}
5555 
5556 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5557 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5558 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5559 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP && (
5560 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5561 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT))) ||
5562 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH && (
5563 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH) ||
5564 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH))) ||
5565 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5566 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5567 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5568 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5569 		ipseclog((LOG_DEBUG,
5570 		    "%s: invalid message: missing required header.\n",
5571 		    __func__));
5572 		return key_senderror(so, m, EINVAL);
5573 	}
5574 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5575 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5576 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5577 		ipseclog((LOG_DEBUG,
5578 		    "%s: invalid message: wrong header size.\n", __func__));
5579 		return key_senderror(so, m, EINVAL);
5580 	}
5581 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5582 		mode = IPSEC_MODE_ANY;
5583 		reqid = 0;
5584 	} else {
5585 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5586 			ipseclog((LOG_DEBUG,
5587 			    "%s: invalid message: wrong header size.\n",
5588 			    __func__));
5589 			return key_senderror(so, m, EINVAL);
5590 		}
5591 		mode = ((struct sadb_x_sa2 *)
5592 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5593 		reqid = ((struct sadb_x_sa2 *)
5594 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5595 	}
5596 
5597 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5598 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
5599 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
5600 
5601 	/*
5602 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5603 	 * SADB_ADD message.
5604 	 */
5605 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5606 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5607 #ifdef PFKEY_STRICT_CHECKS
5608 		return key_senderror(so, m, EINVAL);
5609 #endif
5610 	}
5611 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5612 	    (struct sockaddr *)(dst0 + 1));
5613 	if (error != 0) {
5614 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5615 		return key_senderror(so, m, error);
5616 	}
5617 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5618 	spi = sa0->sadb_sa_spi;
5619 	/*
5620 	 * For TCP-MD5 SAs we don't use SPI. Check the uniqueness using
5621 	 * secasindex.
5622 	 * XXXAE: IPComp seems also doesn't use SPI.
5623 	 */
5624 	if (proto == IPPROTO_TCP) {
5625 		sav = key_getsav_tcpmd5(&saidx, &spi);
5626 		if (sav == NULL && spi == 0) {
5627 			/* Failed to allocate SPI */
5628 			ipseclog((LOG_DEBUG, "%s: SA already exists.\n",
5629 			    __func__));
5630 			return key_senderror(so, m, EEXIST);
5631 		}
5632 		/* XXX: SPI that we report back can have another value */
5633 	} else {
5634 		/* We can create new SA only if SPI is different. */
5635 		sav = key_getsavbyspi(spi);
5636 	}
5637 	if (sav != NULL) {
5638 		key_freesav(&sav);
5639 		ipseclog((LOG_DEBUG, "%s: SA already exists.\n", __func__));
5640 		return key_senderror(so, m, EEXIST);
5641 	}
5642 
5643 	sav = key_newsav(mhp, &saidx, spi, &error);
5644 	if (sav == NULL)
5645 		return key_senderror(so, m, error);
5646 	KEYDBG(KEY_STAMP,
5647 	    printf("%s: return SA(%p)\n", __func__, sav));
5648 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5649 	/*
5650 	 * If SADB_ADD was in response to SADB_ACQUIRE, we need to schedule
5651 	 * ACQ for deletion.
5652 	 */
5653 	if (sav->seq != 0)
5654 		key_acqdone(&saidx, sav->seq);
5655 
5656     {
5657 	/*
5658 	 * Don't call key_freesav() on error here, as we would like to
5659 	 * keep the SA in the database.
5660 	 */
5661 	struct mbuf *n;
5662 
5663 	/* set msg buf from mhp */
5664 	n = key_getmsgbuf_x1(m, mhp);
5665 	if (n == NULL) {
5666 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5667 		return key_senderror(so, m, ENOBUFS);
5668 	}
5669 
5670 	m_freem(m);
5671 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5672     }
5673 }
5674 
5675 /*
5676  * NAT-T support.
5677  * IKEd may request the use ESP in UDP encapsulation when it detects the
5678  * presence of NAT. It uses NAT-T extension headers for such SAs to specify
5679  * parameters needed for encapsulation and decapsulation. These PF_KEY
5680  * extension headers are not standardized, so this comment addresses our
5681  * implementation.
5682  * SADB_X_EXT_NAT_T_TYPE specifies type of encapsulation, we support only
5683  * UDP_ENCAP_ESPINUDP as described in RFC3948.
5684  * SADB_X_EXT_NAT_T_SPORT/DPORT specifies source and destination ports for
5685  * UDP header. We use these ports in UDP encapsulation procedure, also we
5686  * can check them in UDP decapsulation procedure.
5687  * SADB_X_EXT_NAT_T_OA[IR] specifies original address of initiator or
5688  * responder. These addresses can be used for transport mode to adjust
5689  * checksum after decapsulation and decryption. Since original IP addresses
5690  * used by peer usually different (we detected presence of NAT), TCP/UDP
5691  * pseudo header checksum and IP header checksum was calculated using original
5692  * addresses. After decapsulation and decryption we need to adjust checksum
5693  * to have correct datagram.
5694  *
5695  * We expect presence of NAT-T extension headers only in SADB_ADD and
5696  * SADB_UPDATE messages. We report NAT-T extension headers in replies
5697  * to SADB_ADD, SADB_UPDATE, SADB_GET, and SADB_DUMP messages.
5698  */
5699 static int
5700 key_setnatt(struct secasvar *sav, const struct sadb_msghdr *mhp)
5701 {
5702 	struct sadb_x_nat_t_port *port;
5703 	struct sadb_x_nat_t_type *type;
5704 	struct sadb_address *oai, *oar;
5705 	struct sockaddr *sa;
5706 	uint32_t addr;
5707 	uint16_t cksum;
5708 
5709 	IPSEC_ASSERT(sav->natt == NULL, ("natt is already initialized"));
5710 	/*
5711 	 * Ignore NAT-T headers if sproto isn't ESP.
5712 	 */
5713 	if (sav->sah->saidx.proto != IPPROTO_ESP)
5714 		return (0);
5715 
5716 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) &&
5717 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) &&
5718 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5719 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5720 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5721 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5722 			ipseclog((LOG_DEBUG,
5723 			    "%s: invalid message: wrong header size.\n",
5724 			    __func__));
5725 			return (EINVAL);
5726 		}
5727 	} else
5728 		return (0);
5729 
5730 	type = (struct sadb_x_nat_t_type *)mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5731 	if (type->sadb_x_nat_t_type_type != UDP_ENCAP_ESPINUDP) {
5732 		ipseclog((LOG_DEBUG, "%s: unsupported NAT-T type %u.\n",
5733 		    __func__, type->sadb_x_nat_t_type_type));
5734 		return (EINVAL);
5735 	}
5736 	/*
5737 	 * Allocate storage for NAT-T config.
5738 	 * On error it will be released by key_cleansav().
5739 	 */
5740 	sav->natt = malloc(sizeof(struct secnatt), M_IPSEC_MISC,
5741 	    M_NOWAIT | M_ZERO);
5742 	if (sav->natt == NULL) {
5743 		PFKEYSTAT_INC(in_nomem);
5744 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5745 		return (ENOBUFS);
5746 	}
5747 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5748 	if (port->sadb_x_nat_t_port_port == 0) {
5749 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T sport specified.\n",
5750 		    __func__));
5751 		return (EINVAL);
5752 	}
5753 	sav->natt->sport = port->sadb_x_nat_t_port_port;
5754 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5755 	if (port->sadb_x_nat_t_port_port == 0) {
5756 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T dport specified.\n",
5757 		    __func__));
5758 		return (EINVAL);
5759 	}
5760 	sav->natt->dport = port->sadb_x_nat_t_port_port;
5761 
5762 	/*
5763 	 * SADB_X_EXT_NAT_T_OAI and SADB_X_EXT_NAT_T_OAR are optional
5764 	 * and needed only for transport mode IPsec.
5765 	 * Usually NAT translates only one address, but it is possible,
5766 	 * that both addresses could be translated.
5767 	 * NOTE: Value of SADB_X_EXT_NAT_T_OAI is equal to SADB_X_EXT_NAT_T_OA.
5768 	 */
5769 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAI)) {
5770 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAI)) {
5771 			ipseclog((LOG_DEBUG,
5772 			    "%s: invalid message: wrong header size.\n",
5773 			    __func__));
5774 			return (EINVAL);
5775 		}
5776 		oai = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAI];
5777 	} else
5778 		oai = NULL;
5779 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAR)) {
5780 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAR)) {
5781 			ipseclog((LOG_DEBUG,
5782 			    "%s: invalid message: wrong header size.\n",
5783 			    __func__));
5784 			return (EINVAL);
5785 		}
5786 		oar = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAR];
5787 	} else
5788 		oar = NULL;
5789 
5790 	/* Initialize addresses only for transport mode */
5791 	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL) {
5792 		cksum = 0;
5793 		if (oai != NULL) {
5794 			/* Currently we support only AF_INET */
5795 			sa = (struct sockaddr *)(oai + 1);
5796 			if (sa->sa_family != AF_INET ||
5797 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5798 				ipseclog((LOG_DEBUG,
5799 				    "%s: wrong NAT-OAi header.\n",
5800 				    __func__));
5801 				return (EINVAL);
5802 			}
5803 			/* Ignore address if it the same */
5804 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5805 			    sav->sah->saidx.src.sin.sin_addr.s_addr) {
5806 				bcopy(sa, &sav->natt->oai.sa, sa->sa_len);
5807 				sav->natt->flags |= IPSEC_NATT_F_OAI;
5808 				/* Calculate checksum delta */
5809 				addr = sav->sah->saidx.src.sin.sin_addr.s_addr;
5810 				cksum = in_addword(cksum, ~addr >> 16);
5811 				cksum = in_addword(cksum, ~addr & 0xffff);
5812 				addr = sav->natt->oai.sin.sin_addr.s_addr;
5813 				cksum = in_addword(cksum, addr >> 16);
5814 				cksum = in_addword(cksum, addr & 0xffff);
5815 			}
5816 		}
5817 		if (oar != NULL) {
5818 			/* Currently we support only AF_INET */
5819 			sa = (struct sockaddr *)(oar + 1);
5820 			if (sa->sa_family != AF_INET ||
5821 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5822 				ipseclog((LOG_DEBUG,
5823 				    "%s: wrong NAT-OAr header.\n",
5824 				    __func__));
5825 				return (EINVAL);
5826 			}
5827 			/* Ignore address if it the same */
5828 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5829 			    sav->sah->saidx.dst.sin.sin_addr.s_addr) {
5830 				bcopy(sa, &sav->natt->oar.sa, sa->sa_len);
5831 				sav->natt->flags |= IPSEC_NATT_F_OAR;
5832 				/* Calculate checksum delta */
5833 				addr = sav->sah->saidx.dst.sin.sin_addr.s_addr;
5834 				cksum = in_addword(cksum, ~addr >> 16);
5835 				cksum = in_addword(cksum, ~addr & 0xffff);
5836 				addr = sav->natt->oar.sin.sin_addr.s_addr;
5837 				cksum = in_addword(cksum, addr >> 16);
5838 				cksum = in_addword(cksum, addr & 0xffff);
5839 			}
5840 		}
5841 		sav->natt->cksum = cksum;
5842 	}
5843 	return (0);
5844 }
5845 
5846 static int
5847 key_setident(struct secashead *sah, const struct sadb_msghdr *mhp)
5848 {
5849 	const struct sadb_ident *idsrc, *iddst;
5850 
5851 	IPSEC_ASSERT(sah != NULL, ("null secashead"));
5852 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5853 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5854 
5855 	/* don't make buffer if not there */
5856 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) &&
5857 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5858 		sah->idents = NULL;
5859 		sah->identd = NULL;
5860 		return (0);
5861 	}
5862 
5863 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) ||
5864 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5865 		ipseclog((LOG_DEBUG, "%s: invalid identity.\n", __func__));
5866 		return (EINVAL);
5867 	}
5868 
5869 	idsrc = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_SRC];
5870 	iddst = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_DST];
5871 
5872 	/* validity check */
5873 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
5874 		ipseclog((LOG_DEBUG, "%s: ident type mismatch.\n", __func__));
5875 		return EINVAL;
5876 	}
5877 
5878 	switch (idsrc->sadb_ident_type) {
5879 	case SADB_IDENTTYPE_PREFIX:
5880 	case SADB_IDENTTYPE_FQDN:
5881 	case SADB_IDENTTYPE_USERFQDN:
5882 	default:
5883 		/* XXX do nothing */
5884 		sah->idents = NULL;
5885 		sah->identd = NULL;
5886 	 	return 0;
5887 	}
5888 
5889 	/* make structure */
5890 	sah->idents = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5891 	if (sah->idents == NULL) {
5892 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5893 		return ENOBUFS;
5894 	}
5895 	sah->identd = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5896 	if (sah->identd == NULL) {
5897 		free(sah->idents, M_IPSEC_MISC);
5898 		sah->idents = NULL;
5899 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5900 		return ENOBUFS;
5901 	}
5902 	sah->idents->type = idsrc->sadb_ident_type;
5903 	sah->idents->id = idsrc->sadb_ident_id;
5904 
5905 	sah->identd->type = iddst->sadb_ident_type;
5906 	sah->identd->id = iddst->sadb_ident_id;
5907 
5908 	return 0;
5909 }
5910 
5911 /*
5912  * m will not be freed on return.
5913  * it is caller's responsibility to free the result.
5914  *
5915  * Called from SADB_ADD and SADB_UPDATE. Reply will contain headers
5916  * from the request in defined order.
5917  */
5918 static struct mbuf *
5919 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
5920 {
5921 	struct mbuf *n;
5922 
5923 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5924 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5925 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5926 
5927 	/* create new sadb_msg to reply. */
5928 	n = key_gather_mbuf(m, mhp, 1, 16, SADB_EXT_RESERVED,
5929 	    SADB_EXT_SA, SADB_X_EXT_SA2,
5930 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
5931 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
5932 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
5933 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
5934 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
5935 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NEW_ADDRESS_SRC,
5936 	    SADB_X_EXT_NEW_ADDRESS_DST);
5937 	if (!n)
5938 		return NULL;
5939 
5940 	if (n->m_len < sizeof(struct sadb_msg)) {
5941 		n = m_pullup(n, sizeof(struct sadb_msg));
5942 		if (n == NULL)
5943 			return NULL;
5944 	}
5945 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
5946 	mtod(n, struct sadb_msg *)->sadb_msg_len =
5947 	    PFKEY_UNIT64(n->m_pkthdr.len);
5948 
5949 	return n;
5950 }
5951 
5952 /*
5953  * SADB_DELETE processing
5954  * receive
5955  *   <base, SA(*), address(SD)>
5956  * from the ikmpd, and set SADB_SASTATE_DEAD,
5957  * and send,
5958  *   <base, SA(*), address(SD)>
5959  * to the ikmpd.
5960  *
5961  * m will always be freed.
5962  */
5963 static int
5964 key_delete(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5965 {
5966 	struct secasindex saidx;
5967 	struct sadb_address *src0, *dst0;
5968 	struct secasvar *sav;
5969 	struct sadb_sa *sa0;
5970 	uint8_t proto;
5971 
5972 	IPSEC_ASSERT(so != NULL, ("null socket"));
5973 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5974 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5975 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5976 
5977 	/* map satype to proto */
5978 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5979 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5980 		    __func__));
5981 		return key_senderror(so, m, EINVAL);
5982 	}
5983 
5984 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5985 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5986 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5987 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5988 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
5989 		    __func__));
5990 		return key_senderror(so, m, EINVAL);
5991 	}
5992 
5993 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5994 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5995 
5996 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
5997 	    (struct sockaddr *)(dst0 + 1)) != 0) {
5998 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5999 		return (key_senderror(so, m, EINVAL));
6000 	}
6001 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6002 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
6003 		/*
6004 		 * Caller wants us to delete all non-LARVAL SAs
6005 		 * that match the src/dst.  This is used during
6006 		 * IKE INITIAL-CONTACT.
6007 		 * XXXAE: this looks like some extension to RFC2367.
6008 		 */
6009 		ipseclog((LOG_DEBUG, "%s: doing delete all.\n", __func__));
6010 		return (key_delete_all(so, m, mhp, &saidx));
6011 	}
6012 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
6013 		ipseclog((LOG_DEBUG,
6014 		    "%s: invalid message: wrong header size.\n", __func__));
6015 		return (key_senderror(so, m, EINVAL));
6016 	}
6017 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6018 	if (proto == IPPROTO_TCP)
6019 		sav = key_getsav_tcpmd5(&saidx, NULL);
6020 	else
6021 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6022 	if (sav == NULL) {
6023 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u.\n",
6024 		    __func__, ntohl(sa0->sadb_sa_spi)));
6025 		return (key_senderror(so, m, ESRCH));
6026 	}
6027 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6028 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6029 		    __func__, ntohl(sav->spi)));
6030 		key_freesav(&sav);
6031 		return (key_senderror(so, m, ESRCH));
6032 	}
6033 	KEYDBG(KEY_STAMP,
6034 	    printf("%s: SA(%p)\n", __func__, sav));
6035 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
6036 	key_unlinksav(sav);
6037 	key_freesav(&sav);
6038 
6039     {
6040 	struct mbuf *n;
6041 	struct sadb_msg *newmsg;
6042 
6043 	/* create new sadb_msg to reply. */
6044 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
6045 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6046 	if (!n)
6047 		return key_senderror(so, m, ENOBUFS);
6048 
6049 	if (n->m_len < sizeof(struct sadb_msg)) {
6050 		n = m_pullup(n, sizeof(struct sadb_msg));
6051 		if (n == NULL)
6052 			return key_senderror(so, m, ENOBUFS);
6053 	}
6054 	newmsg = mtod(n, struct sadb_msg *);
6055 	newmsg->sadb_msg_errno = 0;
6056 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6057 
6058 	m_freem(m);
6059 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6060     }
6061 }
6062 
6063 /*
6064  * delete all SAs for src/dst.  Called from key_delete().
6065  */
6066 static int
6067 key_delete_all(struct socket *so, struct mbuf *m,
6068     const struct sadb_msghdr *mhp, struct secasindex *saidx)
6069 {
6070 	struct secasvar_queue drainq;
6071 	struct secashead *sah;
6072 	struct secasvar *sav, *nextsav;
6073 
6074 	TAILQ_INIT(&drainq);
6075 	SAHTREE_WLOCK();
6076 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
6077 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_HEAD) == 0)
6078 			continue;
6079 		/* Move all ALIVE SAs into drainq */
6080 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6081 	}
6082 	/* Unlink all queued SAs from SPI hash */
6083 	TAILQ_FOREACH(sav, &drainq, chain) {
6084 		sav->state = SADB_SASTATE_DEAD;
6085 		LIST_REMOVE(sav, spihash);
6086 	}
6087 	SAHTREE_WUNLOCK();
6088 	/* Now we can release reference for all SAs in drainq */
6089 	sav = TAILQ_FIRST(&drainq);
6090 	while (sav != NULL) {
6091 		KEYDBG(KEY_STAMP,
6092 		    printf("%s: SA(%p)\n", __func__, sav));
6093 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6094 		nextsav = TAILQ_NEXT(sav, chain);
6095 		key_freesah(&sav->sah); /* release reference from SAV */
6096 		key_freesav(&sav); /* release last reference */
6097 		sav = nextsav;
6098 	}
6099 
6100     {
6101 	struct mbuf *n;
6102 	struct sadb_msg *newmsg;
6103 
6104 	/* create new sadb_msg to reply. */
6105 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
6106 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6107 	if (!n)
6108 		return key_senderror(so, m, ENOBUFS);
6109 
6110 	if (n->m_len < sizeof(struct sadb_msg)) {
6111 		n = m_pullup(n, sizeof(struct sadb_msg));
6112 		if (n == NULL)
6113 			return key_senderror(so, m, ENOBUFS);
6114 	}
6115 	newmsg = mtod(n, struct sadb_msg *);
6116 	newmsg->sadb_msg_errno = 0;
6117 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6118 
6119 	m_freem(m);
6120 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6121     }
6122 }
6123 
6124 /*
6125  * Delete all alive SAs for corresponding xform.
6126  * Larval SAs have not initialized tdb_xform, so it is safe to leave them
6127  * here when xform disappears.
6128  */
6129 void
6130 key_delete_xform(const struct xformsw *xsp)
6131 {
6132 	struct secasvar_queue drainq;
6133 	struct secashead *sah;
6134 	struct secasvar *sav, *nextsav;
6135 
6136 	TAILQ_INIT(&drainq);
6137 	SAHTREE_WLOCK();
6138 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
6139 		sav = TAILQ_FIRST(&sah->savtree_alive);
6140 		if (sav == NULL)
6141 			continue;
6142 		if (sav->tdb_xform != xsp)
6143 			continue;
6144 		/*
6145 		 * It is supposed that all SAs in the chain are related to
6146 		 * one xform.
6147 		 */
6148 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6149 	}
6150 	/* Unlink all queued SAs from SPI hash */
6151 	TAILQ_FOREACH(sav, &drainq, chain) {
6152 		sav->state = SADB_SASTATE_DEAD;
6153 		LIST_REMOVE(sav, spihash);
6154 	}
6155 	SAHTREE_WUNLOCK();
6156 
6157 	/* Now we can release reference for all SAs in drainq */
6158 	sav = TAILQ_FIRST(&drainq);
6159 	while (sav != NULL) {
6160 		KEYDBG(KEY_STAMP,
6161 		    printf("%s: SA(%p)\n", __func__, sav));
6162 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6163 		nextsav = TAILQ_NEXT(sav, chain);
6164 		key_freesah(&sav->sah); /* release reference from SAV */
6165 		key_freesav(&sav); /* release last reference */
6166 		sav = nextsav;
6167 	}
6168 }
6169 
6170 /*
6171  * SADB_GET processing
6172  * receive
6173  *   <base, SA(*), address(SD)>
6174  * from the ikmpd, and get a SP and a SA to respond,
6175  * and send,
6176  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6177  *       (identity(SD),) (sensitivity)>
6178  * to the ikmpd.
6179  *
6180  * m will always be freed.
6181  */
6182 static int
6183 key_get(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6184 {
6185 	struct secasindex saidx;
6186 	struct sadb_address *src0, *dst0;
6187 	struct sadb_sa *sa0;
6188 	struct secasvar *sav;
6189 	uint8_t proto;
6190 
6191 	IPSEC_ASSERT(so != NULL, ("null socket"));
6192 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6193 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6194 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6195 
6196 	/* map satype to proto */
6197 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6198 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6199 			__func__));
6200 		return key_senderror(so, m, EINVAL);
6201 	}
6202 
6203 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
6204 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6205 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)) {
6206 		ipseclog((LOG_DEBUG,
6207 		    "%s: invalid message: missing required header.\n",
6208 		    __func__));
6209 		return key_senderror(so, m, EINVAL);
6210 	}
6211 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
6212 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6213 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6214 		ipseclog((LOG_DEBUG,
6215 		    "%s: invalid message: wrong header size.\n", __func__));
6216 		return key_senderror(so, m, EINVAL);
6217 	}
6218 
6219 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6220 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
6221 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
6222 
6223 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6224 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6225 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6226 		return key_senderror(so, m, EINVAL);
6227 	}
6228 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6229 
6230 	if (proto == IPPROTO_TCP)
6231 		sav = key_getsav_tcpmd5(&saidx, NULL);
6232 	else
6233 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6234 	if (sav == NULL) {
6235 		ipseclog((LOG_DEBUG, "%s: no SA found.\n", __func__));
6236 		return key_senderror(so, m, ESRCH);
6237 	}
6238 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6239 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6240 		    __func__, ntohl(sa0->sadb_sa_spi)));
6241 		key_freesav(&sav);
6242 		return (key_senderror(so, m, ESRCH));
6243 	}
6244 
6245     {
6246 	struct mbuf *n;
6247 	uint8_t satype;
6248 
6249 	/* map proto to satype */
6250 	if ((satype = key_proto2satype(sav->sah->saidx.proto)) == 0) {
6251 		ipseclog((LOG_DEBUG, "%s: there was invalid proto in SAD.\n",
6252 		    __func__));
6253 		key_freesav(&sav);
6254 		return key_senderror(so, m, EINVAL);
6255 	}
6256 
6257 	/* create new sadb_msg to reply. */
6258 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
6259 	    mhp->msg->sadb_msg_pid);
6260 
6261 	key_freesav(&sav);
6262 	if (!n)
6263 		return key_senderror(so, m, ENOBUFS);
6264 
6265 	m_freem(m);
6266 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
6267     }
6268 }
6269 
6270 /* XXX make it sysctl-configurable? */
6271 static void
6272 key_getcomb_setlifetime(struct sadb_comb *comb)
6273 {
6274 
6275 	comb->sadb_comb_soft_allocations = 1;
6276 	comb->sadb_comb_hard_allocations = 1;
6277 	comb->sadb_comb_soft_bytes = 0;
6278 	comb->sadb_comb_hard_bytes = 0;
6279 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
6280 	comb->sadb_comb_soft_addtime = comb->sadb_comb_soft_addtime * 80 / 100;
6281 	comb->sadb_comb_soft_usetime = 28800;	/* 8 hours */
6282 	comb->sadb_comb_hard_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
6283 }
6284 
6285 /*
6286  * XXX reorder combinations by preference
6287  * XXX no idea if the user wants ESP authentication or not
6288  */
6289 static struct mbuf *
6290 key_getcomb_ealg(void)
6291 {
6292 	struct sadb_comb *comb;
6293 	const struct enc_xform *algo;
6294 	struct mbuf *result = NULL, *m, *n;
6295 	int encmin;
6296 	int i, off, o;
6297 	int totlen;
6298 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6299 
6300 	m = NULL;
6301 	for (i = 1; i <= SADB_EALG_MAX; i++) {
6302 		algo = enc_algorithm_lookup(i);
6303 		if (algo == NULL)
6304 			continue;
6305 
6306 		/* discard algorithms with key size smaller than system min */
6307 		if (_BITS(algo->maxkey) < V_ipsec_esp_keymin)
6308 			continue;
6309 		if (_BITS(algo->minkey) < V_ipsec_esp_keymin)
6310 			encmin = V_ipsec_esp_keymin;
6311 		else
6312 			encmin = _BITS(algo->minkey);
6313 
6314 		if (V_ipsec_esp_auth)
6315 			m = key_getcomb_ah();
6316 		else {
6317 			IPSEC_ASSERT(l <= MLEN,
6318 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6319 			MGET(m, M_NOWAIT, MT_DATA);
6320 			if (m) {
6321 				M_ALIGN(m, l);
6322 				m->m_len = l;
6323 				m->m_next = NULL;
6324 				bzero(mtod(m, caddr_t), m->m_len);
6325 			}
6326 		}
6327 		if (!m)
6328 			goto fail;
6329 
6330 		totlen = 0;
6331 		for (n = m; n; n = n->m_next)
6332 			totlen += n->m_len;
6333 		IPSEC_ASSERT((totlen % l) == 0, ("totlen=%u, l=%u", totlen, l));
6334 
6335 		for (off = 0; off < totlen; off += l) {
6336 			n = m_pulldown(m, off, l, &o);
6337 			if (!n) {
6338 				/* m is already freed */
6339 				goto fail;
6340 			}
6341 			comb = (struct sadb_comb *)(mtod(n, caddr_t) + o);
6342 			bzero(comb, sizeof(*comb));
6343 			key_getcomb_setlifetime(comb);
6344 			comb->sadb_comb_encrypt = i;
6345 			comb->sadb_comb_encrypt_minbits = encmin;
6346 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
6347 		}
6348 
6349 		if (!result)
6350 			result = m;
6351 		else
6352 			m_cat(result, m);
6353 	}
6354 
6355 	return result;
6356 
6357  fail:
6358 	if (result)
6359 		m_freem(result);
6360 	return NULL;
6361 }
6362 
6363 static void
6364 key_getsizes_ah(const struct auth_hash *ah, int alg, u_int16_t* min,
6365     u_int16_t* max)
6366 {
6367 
6368 	*min = *max = ah->hashsize;
6369 	if (ah->keysize == 0) {
6370 		/*
6371 		 * Transform takes arbitrary key size but algorithm
6372 		 * key size is restricted.  Enforce this here.
6373 		 */
6374 		switch (alg) {
6375 		case SADB_X_AALG_NULL:	*min = 1; *max = 256; break;
6376 		case SADB_X_AALG_SHA2_256: *min = *max = 32; break;
6377 		case SADB_X_AALG_SHA2_384: *min = *max = 48; break;
6378 		case SADB_X_AALG_SHA2_512: *min = *max = 64; break;
6379 		default:
6380 			DPRINTF(("%s: unknown AH algorithm %u\n",
6381 				__func__, alg));
6382 			break;
6383 		}
6384 	}
6385 }
6386 
6387 /*
6388  * XXX reorder combinations by preference
6389  */
6390 static struct mbuf *
6391 key_getcomb_ah()
6392 {
6393 	const struct auth_hash *algo;
6394 	struct sadb_comb *comb;
6395 	struct mbuf *m;
6396 	u_int16_t minkeysize, maxkeysize;
6397 	int i;
6398 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6399 
6400 	m = NULL;
6401 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6402 #if 1
6403 		/* we prefer HMAC algorithms, not old algorithms */
6404 		if (i != SADB_AALG_SHA1HMAC &&
6405 		    i != SADB_X_AALG_SHA2_256 &&
6406 		    i != SADB_X_AALG_SHA2_384 &&
6407 		    i != SADB_X_AALG_SHA2_512)
6408 			continue;
6409 #endif
6410 		algo = auth_algorithm_lookup(i);
6411 		if (!algo)
6412 			continue;
6413 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6414 		/* discard algorithms with key size smaller than system min */
6415 		if (_BITS(minkeysize) < V_ipsec_ah_keymin)
6416 			continue;
6417 
6418 		if (!m) {
6419 			IPSEC_ASSERT(l <= MLEN,
6420 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6421 			MGET(m, M_NOWAIT, MT_DATA);
6422 			if (m) {
6423 				M_ALIGN(m, l);
6424 				m->m_len = l;
6425 				m->m_next = NULL;
6426 			}
6427 		} else
6428 			M_PREPEND(m, l, M_NOWAIT);
6429 		if (!m)
6430 			return NULL;
6431 
6432 		comb = mtod(m, struct sadb_comb *);
6433 		bzero(comb, sizeof(*comb));
6434 		key_getcomb_setlifetime(comb);
6435 		comb->sadb_comb_auth = i;
6436 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6437 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6438 	}
6439 
6440 	return m;
6441 }
6442 
6443 /*
6444  * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6445  * XXX reorder combinations by preference
6446  */
6447 static struct mbuf *
6448 key_getcomb_ipcomp()
6449 {
6450 	const struct comp_algo *algo;
6451 	struct sadb_comb *comb;
6452 	struct mbuf *m;
6453 	int i;
6454 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6455 
6456 	m = NULL;
6457 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6458 		algo = comp_algorithm_lookup(i);
6459 		if (!algo)
6460 			continue;
6461 
6462 		if (!m) {
6463 			IPSEC_ASSERT(l <= MLEN,
6464 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6465 			MGET(m, M_NOWAIT, MT_DATA);
6466 			if (m) {
6467 				M_ALIGN(m, l);
6468 				m->m_len = l;
6469 				m->m_next = NULL;
6470 			}
6471 		} else
6472 			M_PREPEND(m, l, M_NOWAIT);
6473 		if (!m)
6474 			return NULL;
6475 
6476 		comb = mtod(m, struct sadb_comb *);
6477 		bzero(comb, sizeof(*comb));
6478 		key_getcomb_setlifetime(comb);
6479 		comb->sadb_comb_encrypt = i;
6480 		/* what should we set into sadb_comb_*_{min,max}bits? */
6481 	}
6482 
6483 	return m;
6484 }
6485 
6486 /*
6487  * XXX no way to pass mode (transport/tunnel) to userland
6488  * XXX replay checking?
6489  * XXX sysctl interface to ipsec_{ah,esp}_keymin
6490  */
6491 static struct mbuf *
6492 key_getprop(const struct secasindex *saidx)
6493 {
6494 	struct sadb_prop *prop;
6495 	struct mbuf *m, *n;
6496 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6497 	int totlen;
6498 
6499 	switch (saidx->proto)  {
6500 	case IPPROTO_ESP:
6501 		m = key_getcomb_ealg();
6502 		break;
6503 	case IPPROTO_AH:
6504 		m = key_getcomb_ah();
6505 		break;
6506 	case IPPROTO_IPCOMP:
6507 		m = key_getcomb_ipcomp();
6508 		break;
6509 	default:
6510 		return NULL;
6511 	}
6512 
6513 	if (!m)
6514 		return NULL;
6515 	M_PREPEND(m, l, M_NOWAIT);
6516 	if (!m)
6517 		return NULL;
6518 
6519 	totlen = 0;
6520 	for (n = m; n; n = n->m_next)
6521 		totlen += n->m_len;
6522 
6523 	prop = mtod(m, struct sadb_prop *);
6524 	bzero(prop, sizeof(*prop));
6525 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6526 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6527 	prop->sadb_prop_replay = 32;	/* XXX */
6528 
6529 	return m;
6530 }
6531 
6532 /*
6533  * SADB_ACQUIRE processing called by key_checkrequest() and key_acquire2().
6534  * send
6535  *   <base, SA, address(SD), (address(P)), x_policy,
6536  *       (identity(SD),) (sensitivity,) proposal>
6537  * to KMD, and expect to receive
6538  *   <base> with SADB_ACQUIRE if error occurred,
6539  * or
6540  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6541  * from KMD by PF_KEY.
6542  *
6543  * XXX x_policy is outside of RFC2367 (KAME extension).
6544  * XXX sensitivity is not supported.
6545  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6546  * see comment for key_getcomb_ipcomp().
6547  *
6548  * OUT:
6549  *    0     : succeed
6550  *    others: error number
6551  */
6552 static int
6553 key_acquire(const struct secasindex *saidx, struct secpolicy *sp)
6554 {
6555 	union sockaddr_union addr;
6556 	struct mbuf *result, *m;
6557 	uint32_t seq;
6558 	int error;
6559 	uint16_t ul_proto;
6560 	uint8_t mask, satype;
6561 
6562 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
6563 	satype = key_proto2satype(saidx->proto);
6564 	IPSEC_ASSERT(satype != 0, ("null satype, protocol %u", saidx->proto));
6565 
6566 	error = -1;
6567 	result = NULL;
6568 	ul_proto = IPSEC_ULPROTO_ANY;
6569 
6570 	/* Get seq number to check whether sending message or not. */
6571 	seq = key_getacq(saidx, &error);
6572 	if (seq == 0)
6573 		return (error);
6574 
6575 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0);
6576 	if (!m) {
6577 		error = ENOBUFS;
6578 		goto fail;
6579 	}
6580 	result = m;
6581 
6582 	/*
6583 	 * set sadb_address for saidx's.
6584 	 *
6585 	 * Note that if sp is supplied, then we're being called from
6586 	 * key_allocsa_policy() and should supply port and protocol
6587 	 * information.
6588 	 * XXXAE: why only TCP and UDP? ICMP and SCTP looks applicable too.
6589 	 * XXXAE: probably we can handle this in the ipsec[46]_allocsa().
6590 	 * XXXAE: it looks like we should save this info in the ACQ entry.
6591 	 */
6592 	if (sp != NULL && (sp->spidx.ul_proto == IPPROTO_TCP ||
6593 	    sp->spidx.ul_proto == IPPROTO_UDP))
6594 		ul_proto = sp->spidx.ul_proto;
6595 
6596 	addr = saidx->src;
6597 	mask = FULLMASK;
6598 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6599 		switch (sp->spidx.src.sa.sa_family) {
6600 		case AF_INET:
6601 			if (sp->spidx.src.sin.sin_port != IPSEC_PORT_ANY) {
6602 				addr.sin.sin_port = sp->spidx.src.sin.sin_port;
6603 				mask = sp->spidx.prefs;
6604 			}
6605 			break;
6606 		case AF_INET6:
6607 			if (sp->spidx.src.sin6.sin6_port != IPSEC_PORT_ANY) {
6608 				addr.sin6.sin6_port =
6609 				    sp->spidx.src.sin6.sin6_port;
6610 				mask = sp->spidx.prefs;
6611 			}
6612 			break;
6613 		default:
6614 			break;
6615 		}
6616 	}
6617 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &addr.sa, mask, ul_proto);
6618 	if (!m) {
6619 		error = ENOBUFS;
6620 		goto fail;
6621 	}
6622 	m_cat(result, m);
6623 
6624 	addr = saidx->dst;
6625 	mask = FULLMASK;
6626 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6627 		switch (sp->spidx.dst.sa.sa_family) {
6628 		case AF_INET:
6629 			if (sp->spidx.dst.sin.sin_port != IPSEC_PORT_ANY) {
6630 				addr.sin.sin_port = sp->spidx.dst.sin.sin_port;
6631 				mask = sp->spidx.prefd;
6632 			}
6633 			break;
6634 		case AF_INET6:
6635 			if (sp->spidx.dst.sin6.sin6_port != IPSEC_PORT_ANY) {
6636 				addr.sin6.sin6_port =
6637 				    sp->spidx.dst.sin6.sin6_port;
6638 				mask = sp->spidx.prefd;
6639 			}
6640 			break;
6641 		default:
6642 			break;
6643 		}
6644 	}
6645 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &addr.sa, mask, ul_proto);
6646 	if (!m) {
6647 		error = ENOBUFS;
6648 		goto fail;
6649 	}
6650 	m_cat(result, m);
6651 
6652 	/* XXX proxy address (optional) */
6653 
6654 	/*
6655 	 * Set sadb_x_policy. This is KAME extension to RFC2367.
6656 	 */
6657 	if (sp != NULL) {
6658 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6659 		    sp->priority);
6660 		if (!m) {
6661 			error = ENOBUFS;
6662 			goto fail;
6663 		}
6664 		m_cat(result, m);
6665 	}
6666 
6667 	/*
6668 	 * Set sadb_x_sa2 extension if saidx->reqid is not zero.
6669 	 * This is FreeBSD extension to RFC2367.
6670 	 */
6671 	if (saidx->reqid != 0) {
6672 		m = key_setsadbxsa2(saidx->mode, 0, saidx->reqid);
6673 		if (m == NULL) {
6674 			error = ENOBUFS;
6675 			goto fail;
6676 		}
6677 		m_cat(result, m);
6678 	}
6679 	/* XXX identity (optional) */
6680 #if 0
6681 	if (idexttype && fqdn) {
6682 		/* create identity extension (FQDN) */
6683 		struct sadb_ident *id;
6684 		int fqdnlen;
6685 
6686 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
6687 		id = (struct sadb_ident *)p;
6688 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6689 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6690 		id->sadb_ident_exttype = idexttype;
6691 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
6692 		bcopy(fqdn, id + 1, fqdnlen);
6693 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
6694 	}
6695 
6696 	if (idexttype) {
6697 		/* create identity extension (USERFQDN) */
6698 		struct sadb_ident *id;
6699 		int userfqdnlen;
6700 
6701 		if (userfqdn) {
6702 			/* +1 for terminating-NUL */
6703 			userfqdnlen = strlen(userfqdn) + 1;
6704 		} else
6705 			userfqdnlen = 0;
6706 		id = (struct sadb_ident *)p;
6707 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6708 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6709 		id->sadb_ident_exttype = idexttype;
6710 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
6711 		/* XXX is it correct? */
6712 		if (curproc && curproc->p_cred)
6713 			id->sadb_ident_id = curproc->p_cred->p_ruid;
6714 		if (userfqdn && userfqdnlen)
6715 			bcopy(userfqdn, id + 1, userfqdnlen);
6716 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
6717 	}
6718 #endif
6719 
6720 	/* XXX sensitivity (optional) */
6721 
6722 	/* create proposal/combination extension */
6723 	m = key_getprop(saidx);
6724 #if 0
6725 	/*
6726 	 * spec conformant: always attach proposal/combination extension,
6727 	 * the problem is that we have no way to attach it for ipcomp,
6728 	 * due to the way sadb_comb is declared in RFC2367.
6729 	 */
6730 	if (!m) {
6731 		error = ENOBUFS;
6732 		goto fail;
6733 	}
6734 	m_cat(result, m);
6735 #else
6736 	/*
6737 	 * outside of spec; make proposal/combination extension optional.
6738 	 */
6739 	if (m)
6740 		m_cat(result, m);
6741 #endif
6742 
6743 	if ((result->m_flags & M_PKTHDR) == 0) {
6744 		error = EINVAL;
6745 		goto fail;
6746 	}
6747 
6748 	if (result->m_len < sizeof(struct sadb_msg)) {
6749 		result = m_pullup(result, sizeof(struct sadb_msg));
6750 		if (result == NULL) {
6751 			error = ENOBUFS;
6752 			goto fail;
6753 		}
6754 	}
6755 
6756 	result->m_pkthdr.len = 0;
6757 	for (m = result; m; m = m->m_next)
6758 		result->m_pkthdr.len += m->m_len;
6759 
6760 	mtod(result, struct sadb_msg *)->sadb_msg_len =
6761 	    PFKEY_UNIT64(result->m_pkthdr.len);
6762 
6763 	KEYDBG(KEY_STAMP,
6764 	    printf("%s: SP(%p)\n", __func__, sp));
6765 	KEYDBG(KEY_DATA, kdebug_secasindex(saidx, NULL));
6766 
6767 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
6768 
6769  fail:
6770 	if (result)
6771 		m_freem(result);
6772 	return error;
6773 }
6774 
6775 static uint32_t
6776 key_newacq(const struct secasindex *saidx, int *perror)
6777 {
6778 	struct secacq *acq;
6779 	uint32_t seq;
6780 
6781 	acq = malloc(sizeof(*acq), M_IPSEC_SAQ, M_NOWAIT | M_ZERO);
6782 	if (acq == NULL) {
6783 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6784 		*perror = ENOBUFS;
6785 		return (0);
6786 	}
6787 
6788 	/* copy secindex */
6789 	bcopy(saidx, &acq->saidx, sizeof(acq->saidx));
6790 	acq->created = time_second;
6791 	acq->count = 0;
6792 
6793 	/* add to acqtree */
6794 	ACQ_LOCK();
6795 	seq = acq->seq = (V_acq_seq == ~0 ? 1 : ++V_acq_seq);
6796 	LIST_INSERT_HEAD(&V_acqtree, acq, chain);
6797 	LIST_INSERT_HEAD(ACQADDRHASH_HASH(saidx), acq, addrhash);
6798 	LIST_INSERT_HEAD(ACQSEQHASH_HASH(seq), acq, seqhash);
6799 	ACQ_UNLOCK();
6800 	*perror = 0;
6801 	return (seq);
6802 }
6803 
6804 static uint32_t
6805 key_getacq(const struct secasindex *saidx, int *perror)
6806 {
6807 	struct secacq *acq;
6808 	uint32_t seq;
6809 
6810 	ACQ_LOCK();
6811 	LIST_FOREACH(acq, ACQADDRHASH_HASH(saidx), addrhash) {
6812 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY)) {
6813 			if (acq->count > V_key_blockacq_count) {
6814 				/*
6815 				 * Reset counter and send message.
6816 				 * Also reset created time to keep ACQ for
6817 				 * this saidx.
6818 				 */
6819 				acq->created = time_second;
6820 				acq->count = 0;
6821 				seq = acq->seq;
6822 			} else {
6823 				/*
6824 				 * Increment counter and do nothing.
6825 				 * We send SADB_ACQUIRE message only
6826 				 * for each V_key_blockacq_count packet.
6827 				 */
6828 				acq->count++;
6829 				seq = 0;
6830 			}
6831 			break;
6832 		}
6833 	}
6834 	ACQ_UNLOCK();
6835 	if (acq != NULL) {
6836 		*perror = 0;
6837 		return (seq);
6838 	}
6839 	/* allocate new  entry */
6840 	return (key_newacq(saidx, perror));
6841 }
6842 
6843 static int
6844 key_acqreset(uint32_t seq)
6845 {
6846 	struct secacq *acq;
6847 
6848 	ACQ_LOCK();
6849 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6850 		if (acq->seq == seq) {
6851 			acq->count = 0;
6852 			acq->created = time_second;
6853 			break;
6854 		}
6855 	}
6856 	ACQ_UNLOCK();
6857 	if (acq == NULL)
6858 		return (ESRCH);
6859 	return (0);
6860 }
6861 /*
6862  * Mark ACQ entry as stale to remove it in key_flush_acq().
6863  * Called after successful SADB_GETSPI message.
6864  */
6865 static int
6866 key_acqdone(const struct secasindex *saidx, uint32_t seq)
6867 {
6868 	struct secacq *acq;
6869 
6870 	ACQ_LOCK();
6871 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6872 		if (acq->seq == seq)
6873 			break;
6874 	}
6875 	if (acq != NULL) {
6876 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY) == 0) {
6877 			ipseclog((LOG_DEBUG,
6878 			    "%s: Mismatched saidx for ACQ %u\n", __func__, seq));
6879 			acq = NULL;
6880 		} else {
6881 			acq->created = 0;
6882 		}
6883 	} else {
6884 		ipseclog((LOG_DEBUG,
6885 		    "%s: ACQ %u is not found.\n", __func__, seq));
6886 	}
6887 	ACQ_UNLOCK();
6888 	if (acq == NULL)
6889 		return (ESRCH);
6890 	return (0);
6891 }
6892 
6893 static struct secspacq *
6894 key_newspacq(struct secpolicyindex *spidx)
6895 {
6896 	struct secspacq *acq;
6897 
6898 	/* get new entry */
6899 	acq = malloc(sizeof(struct secspacq), M_IPSEC_SAQ, M_NOWAIT|M_ZERO);
6900 	if (acq == NULL) {
6901 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6902 		return NULL;
6903 	}
6904 
6905 	/* copy secindex */
6906 	bcopy(spidx, &acq->spidx, sizeof(acq->spidx));
6907 	acq->created = time_second;
6908 	acq->count = 0;
6909 
6910 	/* add to spacqtree */
6911 	SPACQ_LOCK();
6912 	LIST_INSERT_HEAD(&V_spacqtree, acq, chain);
6913 	SPACQ_UNLOCK();
6914 
6915 	return acq;
6916 }
6917 
6918 static struct secspacq *
6919 key_getspacq(struct secpolicyindex *spidx)
6920 {
6921 	struct secspacq *acq;
6922 
6923 	SPACQ_LOCK();
6924 	LIST_FOREACH(acq, &V_spacqtree, chain) {
6925 		if (key_cmpspidx_exactly(spidx, &acq->spidx)) {
6926 			/* NB: return holding spacq_lock */
6927 			return acq;
6928 		}
6929 	}
6930 	SPACQ_UNLOCK();
6931 
6932 	return NULL;
6933 }
6934 
6935 /*
6936  * SADB_ACQUIRE processing,
6937  * in first situation, is receiving
6938  *   <base>
6939  * from the ikmpd, and clear sequence of its secasvar entry.
6940  *
6941  * In second situation, is receiving
6942  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6943  * from a user land process, and return
6944  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6945  * to the socket.
6946  *
6947  * m will always be freed.
6948  */
6949 static int
6950 key_acquire2(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6951 {
6952 	SAHTREE_RLOCK_TRACKER;
6953 	struct sadb_address *src0, *dst0;
6954 	struct secasindex saidx;
6955 	struct secashead *sah;
6956 	uint32_t reqid;
6957 	int error;
6958 	uint8_t mode, proto;
6959 
6960 	IPSEC_ASSERT(so != NULL, ("null socket"));
6961 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6962 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6963 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6964 
6965 	/*
6966 	 * Error message from KMd.
6967 	 * We assume that if error was occurred in IKEd, the length of PFKEY
6968 	 * message is equal to the size of sadb_msg structure.
6969 	 * We do not raise error even if error occurred in this function.
6970 	 */
6971 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
6972 		/* check sequence number */
6973 		if (mhp->msg->sadb_msg_seq == 0 ||
6974 		    mhp->msg->sadb_msg_errno == 0) {
6975 			ipseclog((LOG_DEBUG, "%s: must specify sequence "
6976 				"number and errno.\n", __func__));
6977 		} else {
6978 			/*
6979 			 * IKEd reported that error occurred.
6980 			 * XXXAE: what it expects from the kernel?
6981 			 * Probably we should send SADB_ACQUIRE again?
6982 			 * If so, reset ACQ's state.
6983 			 * XXXAE: it looks useless.
6984 			 */
6985 			key_acqreset(mhp->msg->sadb_msg_seq);
6986 		}
6987 		m_freem(m);
6988 		return (0);
6989 	}
6990 
6991 	/*
6992 	 * This message is from user land.
6993 	 */
6994 
6995 	/* map satype to proto */
6996 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6997 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6998 		    __func__));
6999 		return key_senderror(so, m, EINVAL);
7000 	}
7001 
7002 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
7003 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
7004 	    SADB_CHECKHDR(mhp, SADB_EXT_PROPOSAL)) {
7005 		ipseclog((LOG_DEBUG,
7006 		    "%s: invalid message: missing required header.\n",
7007 		    __func__));
7008 		return key_senderror(so, m, EINVAL);
7009 	}
7010 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
7011 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
7012 	    SADB_CHECKLEN(mhp, SADB_EXT_PROPOSAL)) {
7013 		ipseclog((LOG_DEBUG,
7014 		    "%s: invalid message: wrong header size.\n", __func__));
7015 		return key_senderror(so, m, EINVAL);
7016 	}
7017 
7018 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
7019 		mode = IPSEC_MODE_ANY;
7020 		reqid = 0;
7021 	} else {
7022 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
7023 			ipseclog((LOG_DEBUG,
7024 			    "%s: invalid message: wrong header size.\n",
7025 			    __func__));
7026 			return key_senderror(so, m, EINVAL);
7027 		}
7028 		mode = ((struct sadb_x_sa2 *)
7029 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
7030 		reqid = ((struct sadb_x_sa2 *)
7031 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
7032 	}
7033 
7034 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
7035 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
7036 
7037 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
7038 	    (struct sockaddr *)(dst0 + 1));
7039 	if (error != 0) {
7040 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
7041 		return key_senderror(so, m, EINVAL);
7042 	}
7043 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
7044 
7045 	/* get a SA index */
7046 	SAHTREE_RLOCK();
7047 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
7048 		if (key_cmpsaidx(&sah->saidx, &saidx, CMP_MODE_REQID))
7049 			break;
7050 	}
7051 	SAHTREE_RUNLOCK();
7052 	if (sah != NULL) {
7053 		ipseclog((LOG_DEBUG, "%s: a SA exists already.\n", __func__));
7054 		return key_senderror(so, m, EEXIST);
7055 	}
7056 
7057 	error = key_acquire(&saidx, NULL);
7058 	if (error != 0) {
7059 		ipseclog((LOG_DEBUG,
7060 		    "%s: error %d returned from key_acquire()\n",
7061 			__func__, error));
7062 		return key_senderror(so, m, error);
7063 	}
7064 	m_freem(m);
7065 	return (0);
7066 }
7067 
7068 /*
7069  * SADB_REGISTER processing.
7070  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
7071  * receive
7072  *   <base>
7073  * from the ikmpd, and register a socket to send PF_KEY messages,
7074  * and send
7075  *   <base, supported>
7076  * to KMD by PF_KEY.
7077  * If socket is detached, must free from regnode.
7078  *
7079  * m will always be freed.
7080  */
7081 static int
7082 key_register(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7083 {
7084 	struct secreg *reg, *newreg = NULL;
7085 
7086 	IPSEC_ASSERT(so != NULL, ("null socket"));
7087 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7088 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7089 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7090 
7091 	/* check for invalid register message */
7092 	if (mhp->msg->sadb_msg_satype >= sizeof(V_regtree)/sizeof(V_regtree[0]))
7093 		return key_senderror(so, m, EINVAL);
7094 
7095 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
7096 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
7097 		goto setmsg;
7098 
7099 	/* check whether existing or not */
7100 	REGTREE_LOCK();
7101 	LIST_FOREACH(reg, &V_regtree[mhp->msg->sadb_msg_satype], chain) {
7102 		if (reg->so == so) {
7103 			REGTREE_UNLOCK();
7104 			ipseclog((LOG_DEBUG, "%s: socket exists already.\n",
7105 				__func__));
7106 			return key_senderror(so, m, EEXIST);
7107 		}
7108 	}
7109 
7110 	/* create regnode */
7111 	newreg =  malloc(sizeof(struct secreg), M_IPSEC_SAR, M_NOWAIT|M_ZERO);
7112 	if (newreg == NULL) {
7113 		REGTREE_UNLOCK();
7114 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7115 		return key_senderror(so, m, ENOBUFS);
7116 	}
7117 
7118 	newreg->so = so;
7119 	((struct keycb *)sotorawcb(so))->kp_registered++;
7120 
7121 	/* add regnode to regtree. */
7122 	LIST_INSERT_HEAD(&V_regtree[mhp->msg->sadb_msg_satype], newreg, chain);
7123 	REGTREE_UNLOCK();
7124 
7125   setmsg:
7126     {
7127 	struct mbuf *n;
7128 	struct sadb_msg *newmsg;
7129 	struct sadb_supported *sup;
7130 	u_int len, alen, elen;
7131 	int off;
7132 	int i;
7133 	struct sadb_alg *alg;
7134 
7135 	/* create new sadb_msg to reply. */
7136 	alen = 0;
7137 	for (i = 1; i <= SADB_AALG_MAX; i++) {
7138 		if (auth_algorithm_lookup(i))
7139 			alen += sizeof(struct sadb_alg);
7140 	}
7141 	if (alen)
7142 		alen += sizeof(struct sadb_supported);
7143 	elen = 0;
7144 	for (i = 1; i <= SADB_EALG_MAX; i++) {
7145 		if (enc_algorithm_lookup(i))
7146 			elen += sizeof(struct sadb_alg);
7147 	}
7148 	if (elen)
7149 		elen += sizeof(struct sadb_supported);
7150 
7151 	len = sizeof(struct sadb_msg) + alen + elen;
7152 
7153 	if (len > MCLBYTES)
7154 		return key_senderror(so, m, ENOBUFS);
7155 
7156 	MGETHDR(n, M_NOWAIT, MT_DATA);
7157 	if (n != NULL && len > MHLEN) {
7158 		if (!(MCLGET(n, M_NOWAIT))) {
7159 			m_freem(n);
7160 			n = NULL;
7161 		}
7162 	}
7163 	if (!n)
7164 		return key_senderror(so, m, ENOBUFS);
7165 
7166 	n->m_pkthdr.len = n->m_len = len;
7167 	n->m_next = NULL;
7168 	off = 0;
7169 
7170 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
7171 	newmsg = mtod(n, struct sadb_msg *);
7172 	newmsg->sadb_msg_errno = 0;
7173 	newmsg->sadb_msg_len = PFKEY_UNIT64(len);
7174 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7175 
7176 	/* for authentication algorithm */
7177 	if (alen) {
7178 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7179 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7180 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7181 		off += PFKEY_ALIGN8(sizeof(*sup));
7182 
7183 		for (i = 1; i <= SADB_AALG_MAX; i++) {
7184 			const struct auth_hash *aalgo;
7185 			u_int16_t minkeysize, maxkeysize;
7186 
7187 			aalgo = auth_algorithm_lookup(i);
7188 			if (!aalgo)
7189 				continue;
7190 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7191 			alg->sadb_alg_id = i;
7192 			alg->sadb_alg_ivlen = 0;
7193 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7194 			alg->sadb_alg_minbits = _BITS(minkeysize);
7195 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7196 			off += PFKEY_ALIGN8(sizeof(*alg));
7197 		}
7198 	}
7199 
7200 	/* for encryption algorithm */
7201 	if (elen) {
7202 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7203 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7204 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7205 		off += PFKEY_ALIGN8(sizeof(*sup));
7206 
7207 		for (i = 1; i <= SADB_EALG_MAX; i++) {
7208 			const struct enc_xform *ealgo;
7209 
7210 			ealgo = enc_algorithm_lookup(i);
7211 			if (!ealgo)
7212 				continue;
7213 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7214 			alg->sadb_alg_id = i;
7215 			alg->sadb_alg_ivlen = ealgo->ivsize;
7216 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7217 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7218 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7219 		}
7220 	}
7221 
7222 	IPSEC_ASSERT(off == len,
7223 		("length assumption failed (off %u len %u)", off, len));
7224 
7225 	m_freem(m);
7226 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7227     }
7228 }
7229 
7230 /*
7231  * free secreg entry registered.
7232  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7233  */
7234 void
7235 key_freereg(struct socket *so)
7236 {
7237 	struct secreg *reg;
7238 	int i;
7239 
7240 	IPSEC_ASSERT(so != NULL, ("NULL so"));
7241 
7242 	/*
7243 	 * check whether existing or not.
7244 	 * check all type of SA, because there is a potential that
7245 	 * one socket is registered to multiple type of SA.
7246 	 */
7247 	REGTREE_LOCK();
7248 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7249 		LIST_FOREACH(reg, &V_regtree[i], chain) {
7250 			if (reg->so == so && __LIST_CHAINED(reg)) {
7251 				LIST_REMOVE(reg, chain);
7252 				free(reg, M_IPSEC_SAR);
7253 				break;
7254 			}
7255 		}
7256 	}
7257 	REGTREE_UNLOCK();
7258 }
7259 
7260 /*
7261  * SADB_EXPIRE processing
7262  * send
7263  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7264  * to KMD by PF_KEY.
7265  * NOTE: We send only soft lifetime extension.
7266  *
7267  * OUT:	0	: succeed
7268  *	others	: error number
7269  */
7270 static int
7271 key_expire(struct secasvar *sav, int hard)
7272 {
7273 	struct mbuf *result = NULL, *m;
7274 	struct sadb_lifetime *lt;
7275 	uint32_t replay_count;
7276 	int error, len;
7277 	uint8_t satype;
7278 
7279 	IPSEC_ASSERT (sav != NULL, ("null sav"));
7280 	IPSEC_ASSERT (sav->sah != NULL, ("null sa header"));
7281 
7282 	KEYDBG(KEY_STAMP,
7283 	    printf("%s: SA(%p) expired %s lifetime\n", __func__,
7284 		sav, hard ? "hard": "soft"));
7285 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
7286 	/* set msg header */
7287 	satype = key_proto2satype(sav->sah->saidx.proto);
7288 	IPSEC_ASSERT(satype != 0, ("invalid proto, satype %u", satype));
7289 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, sav->refcnt);
7290 	if (!m) {
7291 		error = ENOBUFS;
7292 		goto fail;
7293 	}
7294 	result = m;
7295 
7296 	/* create SA extension */
7297 	m = key_setsadbsa(sav);
7298 	if (!m) {
7299 		error = ENOBUFS;
7300 		goto fail;
7301 	}
7302 	m_cat(result, m);
7303 
7304 	/* create SA extension */
7305 	SECASVAR_LOCK(sav);
7306 	replay_count = sav->replay ? sav->replay->count : 0;
7307 	SECASVAR_UNLOCK(sav);
7308 
7309 	m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
7310 			sav->sah->saidx.reqid);
7311 	if (!m) {
7312 		error = ENOBUFS;
7313 		goto fail;
7314 	}
7315 	m_cat(result, m);
7316 
7317 	if (sav->replay && sav->replay->wsize > UINT8_MAX) {
7318 		m = key_setsadbxsareplay(sav->replay->wsize);
7319 		if (!m) {
7320 			error = ENOBUFS;
7321 			goto fail;
7322 		}
7323 		m_cat(result, m);
7324 	}
7325 
7326 	/* create lifetime extension (current and soft) */
7327 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7328 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
7329 	if (m == NULL) {
7330 		error = ENOBUFS;
7331 		goto fail;
7332 	}
7333 	m_align(m, len);
7334 	m->m_len = len;
7335 	bzero(mtod(m, caddr_t), len);
7336 	lt = mtod(m, struct sadb_lifetime *);
7337 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7338 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7339 	lt->sadb_lifetime_allocations =
7340 	    (uint32_t)counter_u64_fetch(sav->lft_c_allocations);
7341 	lt->sadb_lifetime_bytes =
7342 	    counter_u64_fetch(sav->lft_c_bytes);
7343 	lt->sadb_lifetime_addtime = sav->created;
7344 	lt->sadb_lifetime_usetime = sav->firstused;
7345 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
7346 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7347 	if (hard) {
7348 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
7349 		lt->sadb_lifetime_allocations = sav->lft_h->allocations;
7350 		lt->sadb_lifetime_bytes = sav->lft_h->bytes;
7351 		lt->sadb_lifetime_addtime = sav->lft_h->addtime;
7352 		lt->sadb_lifetime_usetime = sav->lft_h->usetime;
7353 	} else {
7354 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_SOFT;
7355 		lt->sadb_lifetime_allocations = sav->lft_s->allocations;
7356 		lt->sadb_lifetime_bytes = sav->lft_s->bytes;
7357 		lt->sadb_lifetime_addtime = sav->lft_s->addtime;
7358 		lt->sadb_lifetime_usetime = sav->lft_s->usetime;
7359 	}
7360 	m_cat(result, m);
7361 
7362 	/* set sadb_address for source */
7363 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
7364 	    &sav->sah->saidx.src.sa,
7365 	    FULLMASK, IPSEC_ULPROTO_ANY);
7366 	if (!m) {
7367 		error = ENOBUFS;
7368 		goto fail;
7369 	}
7370 	m_cat(result, m);
7371 
7372 	/* set sadb_address for destination */
7373 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
7374 	    &sav->sah->saidx.dst.sa,
7375 	    FULLMASK, IPSEC_ULPROTO_ANY);
7376 	if (!m) {
7377 		error = ENOBUFS;
7378 		goto fail;
7379 	}
7380 	m_cat(result, m);
7381 
7382 	/*
7383 	 * XXX-BZ Handle NAT-T extensions here.
7384 	 * XXXAE: it doesn't seem quite useful. IKEs should not depend on
7385 	 * this information, we report only significant SA fields.
7386 	 */
7387 
7388 	if ((result->m_flags & M_PKTHDR) == 0) {
7389 		error = EINVAL;
7390 		goto fail;
7391 	}
7392 
7393 	if (result->m_len < sizeof(struct sadb_msg)) {
7394 		result = m_pullup(result, sizeof(struct sadb_msg));
7395 		if (result == NULL) {
7396 			error = ENOBUFS;
7397 			goto fail;
7398 		}
7399 	}
7400 
7401 	result->m_pkthdr.len = 0;
7402 	for (m = result; m; m = m->m_next)
7403 		result->m_pkthdr.len += m->m_len;
7404 
7405 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7406 	    PFKEY_UNIT64(result->m_pkthdr.len);
7407 
7408 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7409 
7410  fail:
7411 	if (result)
7412 		m_freem(result);
7413 	return error;
7414 }
7415 
7416 static void
7417 key_freesah_flushed(struct secashead_queue *flushq)
7418 {
7419 	struct secashead *sah, *nextsah;
7420 	struct secasvar *sav, *nextsav;
7421 
7422 	sah = TAILQ_FIRST(flushq);
7423 	while (sah != NULL) {
7424 		sav = TAILQ_FIRST(&sah->savtree_larval);
7425 		while (sav != NULL) {
7426 			nextsav = TAILQ_NEXT(sav, chain);
7427 			TAILQ_REMOVE(&sah->savtree_larval, sav, chain);
7428 			key_freesav(&sav); /* release last reference */
7429 			key_freesah(&sah); /* release reference from SAV */
7430 			sav = nextsav;
7431 		}
7432 		sav = TAILQ_FIRST(&sah->savtree_alive);
7433 		while (sav != NULL) {
7434 			nextsav = TAILQ_NEXT(sav, chain);
7435 			TAILQ_REMOVE(&sah->savtree_alive, sav, chain);
7436 			key_freesav(&sav); /* release last reference */
7437 			key_freesah(&sah); /* release reference from SAV */
7438 			sav = nextsav;
7439 		}
7440 		nextsah = TAILQ_NEXT(sah, chain);
7441 		key_freesah(&sah);	/* release last reference */
7442 		sah = nextsah;
7443 	}
7444 }
7445 
7446 /*
7447  * SADB_FLUSH processing
7448  * receive
7449  *   <base>
7450  * from the ikmpd, and free all entries in secastree.
7451  * and send,
7452  *   <base>
7453  * to the ikmpd.
7454  * NOTE: to do is only marking SADB_SASTATE_DEAD.
7455  *
7456  * m will always be freed.
7457  */
7458 static int
7459 key_flush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7460 {
7461 	struct secashead_queue flushq;
7462 	struct sadb_msg *newmsg;
7463 	struct secashead *sah, *nextsah;
7464 	struct secasvar *sav;
7465 	uint8_t proto;
7466 	int i;
7467 
7468 	IPSEC_ASSERT(so != NULL, ("null socket"));
7469 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7470 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7471 
7472 	/* map satype to proto */
7473 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7474 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7475 			__func__));
7476 		return key_senderror(so, m, EINVAL);
7477 	}
7478 	KEYDBG(KEY_STAMP,
7479 	    printf("%s: proto %u\n", __func__, proto));
7480 
7481 	TAILQ_INIT(&flushq);
7482 	if (proto == IPSEC_PROTO_ANY) {
7483 		/* no SATYPE specified, i.e. flushing all SA. */
7484 		SAHTREE_WLOCK();
7485 		/* Move all SAHs into flushq */
7486 		TAILQ_CONCAT(&flushq, &V_sahtree, chain);
7487 		/* Flush all buckets in SPI hash */
7488 		for (i = 0; i < V_savhash_mask + 1; i++)
7489 			LIST_INIT(&V_savhashtbl[i]);
7490 		/* Flush all buckets in SAHADDRHASH */
7491 		for (i = 0; i < V_sahaddrhash_mask + 1; i++)
7492 			LIST_INIT(&V_sahaddrhashtbl[i]);
7493 		/* Mark all SAHs as unlinked */
7494 		TAILQ_FOREACH(sah, &flushq, chain) {
7495 			sah->state = SADB_SASTATE_DEAD;
7496 			/*
7497 			 * Callout handler makes its job using
7498 			 * RLOCK and drain queues. In case, when this
7499 			 * function will be called just before it
7500 			 * acquires WLOCK, we need to mark SAs as
7501 			 * unlinked to prevent second unlink.
7502 			 */
7503 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7504 				sav->state = SADB_SASTATE_DEAD;
7505 			}
7506 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7507 				sav->state = SADB_SASTATE_DEAD;
7508 			}
7509 		}
7510 		SAHTREE_WUNLOCK();
7511 	} else {
7512 		SAHTREE_WLOCK();
7513 		sah = TAILQ_FIRST(&V_sahtree);
7514 		while (sah != NULL) {
7515 			IPSEC_ASSERT(sah->state != SADB_SASTATE_DEAD,
7516 			    ("DEAD SAH %p in SADB_FLUSH", sah));
7517 			nextsah = TAILQ_NEXT(sah, chain);
7518 			if (sah->saidx.proto != proto) {
7519 				sah = nextsah;
7520 				continue;
7521 			}
7522 			sah->state = SADB_SASTATE_DEAD;
7523 			TAILQ_REMOVE(&V_sahtree, sah, chain);
7524 			LIST_REMOVE(sah, addrhash);
7525 			/* Unlink all SAs from SPI hash */
7526 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7527 				LIST_REMOVE(sav, spihash);
7528 				sav->state = SADB_SASTATE_DEAD;
7529 			}
7530 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7531 				LIST_REMOVE(sav, spihash);
7532 				sav->state = SADB_SASTATE_DEAD;
7533 			}
7534 			/* Add SAH into flushq */
7535 			TAILQ_INSERT_HEAD(&flushq, sah, chain);
7536 			sah = nextsah;
7537 		}
7538 		SAHTREE_WUNLOCK();
7539 	}
7540 
7541 	key_freesah_flushed(&flushq);
7542 	/* Free all queued SAs and SAHs */
7543 	if (m->m_len < sizeof(struct sadb_msg) ||
7544 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7545 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7546 		return key_senderror(so, m, ENOBUFS);
7547 	}
7548 
7549 	if (m->m_next)
7550 		m_freem(m->m_next);
7551 	m->m_next = NULL;
7552 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7553 	newmsg = mtod(m, struct sadb_msg *);
7554 	newmsg->sadb_msg_errno = 0;
7555 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7556 
7557 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7558 }
7559 
7560 /*
7561  * SADB_DUMP processing
7562  * dump all entries including status of DEAD in SAD.
7563  * receive
7564  *   <base>
7565  * from the ikmpd, and dump all secasvar leaves
7566  * and send,
7567  *   <base> .....
7568  * to the ikmpd.
7569  *
7570  * m will always be freed.
7571  */
7572 static int
7573 key_dump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7574 {
7575 	SAHTREE_RLOCK_TRACKER;
7576 	struct secashead *sah;
7577 	struct secasvar *sav;
7578 	struct mbuf *n;
7579 	uint32_t cnt;
7580 	uint8_t proto, satype;
7581 
7582 	IPSEC_ASSERT(so != NULL, ("null socket"));
7583 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7584 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7585 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7586 
7587 	/* map satype to proto */
7588 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7589 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7590 		    __func__));
7591 		return key_senderror(so, m, EINVAL);
7592 	}
7593 
7594 	/* count sav entries to be sent to the userland. */
7595 	cnt = 0;
7596 	SAHTREE_RLOCK();
7597 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7598 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7599 		    proto != sah->saidx.proto)
7600 			continue;
7601 
7602 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain)
7603 			cnt++;
7604 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain)
7605 			cnt++;
7606 	}
7607 
7608 	if (cnt == 0) {
7609 		SAHTREE_RUNLOCK();
7610 		return key_senderror(so, m, ENOENT);
7611 	}
7612 
7613 	/* send this to the userland, one at a time. */
7614 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7615 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7616 		    proto != sah->saidx.proto)
7617 			continue;
7618 
7619 		/* map proto to satype */
7620 		if ((satype = key_proto2satype(sah->saidx.proto)) == 0) {
7621 			SAHTREE_RUNLOCK();
7622 			ipseclog((LOG_DEBUG, "%s: there was invalid proto in "
7623 			    "SAD.\n", __func__));
7624 			return key_senderror(so, m, EINVAL);
7625 		}
7626 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7627 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7628 			    --cnt, mhp->msg->sadb_msg_pid);
7629 			if (n == NULL) {
7630 				SAHTREE_RUNLOCK();
7631 				return key_senderror(so, m, ENOBUFS);
7632 			}
7633 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7634 		}
7635 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7636 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7637 			    --cnt, mhp->msg->sadb_msg_pid);
7638 			if (n == NULL) {
7639 				SAHTREE_RUNLOCK();
7640 				return key_senderror(so, m, ENOBUFS);
7641 			}
7642 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7643 		}
7644 	}
7645 	SAHTREE_RUNLOCK();
7646 	m_freem(m);
7647 	return (0);
7648 }
7649 /*
7650  * SADB_X_PROMISC processing
7651  *
7652  * m will always be freed.
7653  */
7654 static int
7655 key_promisc(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7656 {
7657 	int olen;
7658 
7659 	IPSEC_ASSERT(so != NULL, ("null socket"));
7660 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7661 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7662 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7663 
7664 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7665 
7666 	if (olen < sizeof(struct sadb_msg)) {
7667 #if 1
7668 		return key_senderror(so, m, EINVAL);
7669 #else
7670 		m_freem(m);
7671 		return 0;
7672 #endif
7673 	} else if (olen == sizeof(struct sadb_msg)) {
7674 		/* enable/disable promisc mode */
7675 		struct keycb *kp;
7676 
7677 		if ((kp = (struct keycb *)sotorawcb(so)) == NULL)
7678 			return key_senderror(so, m, EINVAL);
7679 		mhp->msg->sadb_msg_errno = 0;
7680 		switch (mhp->msg->sadb_msg_satype) {
7681 		case 0:
7682 		case 1:
7683 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
7684 			break;
7685 		default:
7686 			return key_senderror(so, m, EINVAL);
7687 		}
7688 
7689 		/* send the original message back to everyone */
7690 		mhp->msg->sadb_msg_errno = 0;
7691 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7692 	} else {
7693 		/* send packet as is */
7694 
7695 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
7696 
7697 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
7698 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7699 	}
7700 }
7701 
7702 static int (*key_typesw[])(struct socket *, struct mbuf *,
7703 		const struct sadb_msghdr *) = {
7704 	NULL,		/* SADB_RESERVED */
7705 	key_getspi,	/* SADB_GETSPI */
7706 	key_update,	/* SADB_UPDATE */
7707 	key_add,	/* SADB_ADD */
7708 	key_delete,	/* SADB_DELETE */
7709 	key_get,	/* SADB_GET */
7710 	key_acquire2,	/* SADB_ACQUIRE */
7711 	key_register,	/* SADB_REGISTER */
7712 	NULL,		/* SADB_EXPIRE */
7713 	key_flush,	/* SADB_FLUSH */
7714 	key_dump,	/* SADB_DUMP */
7715 	key_promisc,	/* SADB_X_PROMISC */
7716 	NULL,		/* SADB_X_PCHANGE */
7717 	key_spdadd,	/* SADB_X_SPDUPDATE */
7718 	key_spdadd,	/* SADB_X_SPDADD */
7719 	key_spddelete,	/* SADB_X_SPDDELETE */
7720 	key_spdget,	/* SADB_X_SPDGET */
7721 	NULL,		/* SADB_X_SPDACQUIRE */
7722 	key_spddump,	/* SADB_X_SPDDUMP */
7723 	key_spdflush,	/* SADB_X_SPDFLUSH */
7724 	key_spdadd,	/* SADB_X_SPDSETIDX */
7725 	NULL,		/* SADB_X_SPDEXPIRE */
7726 	key_spddelete2,	/* SADB_X_SPDDELETE2 */
7727 };
7728 
7729 /*
7730  * parse sadb_msg buffer to process PFKEYv2,
7731  * and create a data to response if needed.
7732  * I think to be dealed with mbuf directly.
7733  * IN:
7734  *     msgp  : pointer to pointer to a received buffer pulluped.
7735  *             This is rewrited to response.
7736  *     so    : pointer to socket.
7737  * OUT:
7738  *    length for buffer to send to user process.
7739  */
7740 int
7741 key_parse(struct mbuf *m, struct socket *so)
7742 {
7743 	struct sadb_msg *msg;
7744 	struct sadb_msghdr mh;
7745 	u_int orglen;
7746 	int error;
7747 	int target;
7748 
7749 	IPSEC_ASSERT(so != NULL, ("null socket"));
7750 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7751 
7752 	if (m->m_len < sizeof(struct sadb_msg)) {
7753 		m = m_pullup(m, sizeof(struct sadb_msg));
7754 		if (!m)
7755 			return ENOBUFS;
7756 	}
7757 	msg = mtod(m, struct sadb_msg *);
7758 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
7759 	target = KEY_SENDUP_ONE;
7760 
7761 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len != orglen) {
7762 		ipseclog((LOG_DEBUG, "%s: invalid message length.\n",__func__));
7763 		PFKEYSTAT_INC(out_invlen);
7764 		error = EINVAL;
7765 		goto senderror;
7766 	}
7767 
7768 	if (msg->sadb_msg_version != PF_KEY_V2) {
7769 		ipseclog((LOG_DEBUG, "%s: PF_KEY version %u is mismatched.\n",
7770 		    __func__, msg->sadb_msg_version));
7771 		PFKEYSTAT_INC(out_invver);
7772 		error = EINVAL;
7773 		goto senderror;
7774 	}
7775 
7776 	if (msg->sadb_msg_type > SADB_MAX) {
7777 		ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7778 		    __func__, msg->sadb_msg_type));
7779 		PFKEYSTAT_INC(out_invmsgtype);
7780 		error = EINVAL;
7781 		goto senderror;
7782 	}
7783 
7784 	/* for old-fashioned code - should be nuked */
7785 	if (m->m_pkthdr.len > MCLBYTES) {
7786 		m_freem(m);
7787 		return ENOBUFS;
7788 	}
7789 	if (m->m_next) {
7790 		struct mbuf *n;
7791 
7792 		MGETHDR(n, M_NOWAIT, MT_DATA);
7793 		if (n && m->m_pkthdr.len > MHLEN) {
7794 			if (!(MCLGET(n, M_NOWAIT))) {
7795 				m_free(n);
7796 				n = NULL;
7797 			}
7798 		}
7799 		if (!n) {
7800 			m_freem(m);
7801 			return ENOBUFS;
7802 		}
7803 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
7804 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
7805 		n->m_next = NULL;
7806 		m_freem(m);
7807 		m = n;
7808 	}
7809 
7810 	/* align the mbuf chain so that extensions are in contiguous region. */
7811 	error = key_align(m, &mh);
7812 	if (error)
7813 		return error;
7814 
7815 	msg = mh.msg;
7816 
7817 	/* We use satype as scope mask for spddump */
7818 	if (msg->sadb_msg_type == SADB_X_SPDDUMP) {
7819 		switch (msg->sadb_msg_satype) {
7820 		case IPSEC_POLICYSCOPE_ANY:
7821 		case IPSEC_POLICYSCOPE_GLOBAL:
7822 		case IPSEC_POLICYSCOPE_IFNET:
7823 		case IPSEC_POLICYSCOPE_PCB:
7824 			break;
7825 		default:
7826 			ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7827 			    __func__, msg->sadb_msg_type));
7828 			PFKEYSTAT_INC(out_invsatype);
7829 			error = EINVAL;
7830 			goto senderror;
7831 		}
7832 	} else {
7833 		switch (msg->sadb_msg_satype) { /* check SA type */
7834 		case SADB_SATYPE_UNSPEC:
7835 			switch (msg->sadb_msg_type) {
7836 			case SADB_GETSPI:
7837 			case SADB_UPDATE:
7838 			case SADB_ADD:
7839 			case SADB_DELETE:
7840 			case SADB_GET:
7841 			case SADB_ACQUIRE:
7842 			case SADB_EXPIRE:
7843 				ipseclog((LOG_DEBUG, "%s: must specify satype "
7844 				    "when msg type=%u.\n", __func__,
7845 				    msg->sadb_msg_type));
7846 				PFKEYSTAT_INC(out_invsatype);
7847 				error = EINVAL;
7848 				goto senderror;
7849 			}
7850 			break;
7851 		case SADB_SATYPE_AH:
7852 		case SADB_SATYPE_ESP:
7853 		case SADB_X_SATYPE_IPCOMP:
7854 		case SADB_X_SATYPE_TCPSIGNATURE:
7855 			switch (msg->sadb_msg_type) {
7856 			case SADB_X_SPDADD:
7857 			case SADB_X_SPDDELETE:
7858 			case SADB_X_SPDGET:
7859 			case SADB_X_SPDFLUSH:
7860 			case SADB_X_SPDSETIDX:
7861 			case SADB_X_SPDUPDATE:
7862 			case SADB_X_SPDDELETE2:
7863 				ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7864 				    __func__, msg->sadb_msg_type));
7865 				PFKEYSTAT_INC(out_invsatype);
7866 				error = EINVAL;
7867 				goto senderror;
7868 			}
7869 			break;
7870 		case SADB_SATYPE_RSVP:
7871 		case SADB_SATYPE_OSPFV2:
7872 		case SADB_SATYPE_RIPV2:
7873 		case SADB_SATYPE_MIP:
7874 			ipseclog((LOG_DEBUG, "%s: type %u isn't supported.\n",
7875 			    __func__, msg->sadb_msg_satype));
7876 			PFKEYSTAT_INC(out_invsatype);
7877 			error = EOPNOTSUPP;
7878 			goto senderror;
7879 		case 1:	/* XXX: What does it do? */
7880 			if (msg->sadb_msg_type == SADB_X_PROMISC)
7881 				break;
7882 			/*FALLTHROUGH*/
7883 		default:
7884 			ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7885 			    __func__, msg->sadb_msg_satype));
7886 			PFKEYSTAT_INC(out_invsatype);
7887 			error = EINVAL;
7888 			goto senderror;
7889 		}
7890 	}
7891 
7892 	/* check field of upper layer protocol and address family */
7893 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL
7894 	 && mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
7895 		struct sadb_address *src0, *dst0;
7896 		u_int plen;
7897 
7898 		src0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_SRC]);
7899 		dst0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_DST]);
7900 
7901 		/* check upper layer protocol */
7902 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
7903 			ipseclog((LOG_DEBUG, "%s: upper layer protocol "
7904 				"mismatched.\n", __func__));
7905 			PFKEYSTAT_INC(out_invaddr);
7906 			error = EINVAL;
7907 			goto senderror;
7908 		}
7909 
7910 		/* check family */
7911 		if (PFKEY_ADDR_SADDR(src0)->sa_family !=
7912 		    PFKEY_ADDR_SADDR(dst0)->sa_family) {
7913 			ipseclog((LOG_DEBUG, "%s: address family mismatched.\n",
7914 				__func__));
7915 			PFKEYSTAT_INC(out_invaddr);
7916 			error = EINVAL;
7917 			goto senderror;
7918 		}
7919 		if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7920 		    PFKEY_ADDR_SADDR(dst0)->sa_len) {
7921 			ipseclog((LOG_DEBUG, "%s: address struct size "
7922 				"mismatched.\n", __func__));
7923 			PFKEYSTAT_INC(out_invaddr);
7924 			error = EINVAL;
7925 			goto senderror;
7926 		}
7927 
7928 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7929 		case AF_INET:
7930 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7931 			    sizeof(struct sockaddr_in)) {
7932 				PFKEYSTAT_INC(out_invaddr);
7933 				error = EINVAL;
7934 				goto senderror;
7935 			}
7936 			break;
7937 		case AF_INET6:
7938 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7939 			    sizeof(struct sockaddr_in6)) {
7940 				PFKEYSTAT_INC(out_invaddr);
7941 				error = EINVAL;
7942 				goto senderror;
7943 			}
7944 			break;
7945 		default:
7946 			ipseclog((LOG_DEBUG, "%s: unsupported address family\n",
7947 				__func__));
7948 			PFKEYSTAT_INC(out_invaddr);
7949 			error = EAFNOSUPPORT;
7950 			goto senderror;
7951 		}
7952 
7953 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7954 		case AF_INET:
7955 			plen = sizeof(struct in_addr) << 3;
7956 			break;
7957 		case AF_INET6:
7958 			plen = sizeof(struct in6_addr) << 3;
7959 			break;
7960 		default:
7961 			plen = 0;	/*fool gcc*/
7962 			break;
7963 		}
7964 
7965 		/* check max prefix length */
7966 		if (src0->sadb_address_prefixlen > plen ||
7967 		    dst0->sadb_address_prefixlen > plen) {
7968 			ipseclog((LOG_DEBUG, "%s: illegal prefixlen.\n",
7969 				__func__));
7970 			PFKEYSTAT_INC(out_invaddr);
7971 			error = EINVAL;
7972 			goto senderror;
7973 		}
7974 
7975 		/*
7976 		 * prefixlen == 0 is valid because there can be a case when
7977 		 * all addresses are matched.
7978 		 */
7979 	}
7980 
7981 	if (msg->sadb_msg_type >= nitems(key_typesw) ||
7982 	    key_typesw[msg->sadb_msg_type] == NULL) {
7983 		PFKEYSTAT_INC(out_invmsgtype);
7984 		error = EINVAL;
7985 		goto senderror;
7986 	}
7987 
7988 	return (*key_typesw[msg->sadb_msg_type])(so, m, &mh);
7989 
7990 senderror:
7991 	msg->sadb_msg_errno = error;
7992 	return key_sendup_mbuf(so, m, target);
7993 }
7994 
7995 static int
7996 key_senderror(struct socket *so, struct mbuf *m, int code)
7997 {
7998 	struct sadb_msg *msg;
7999 
8000 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8001 		("mbuf too small, len %u", m->m_len));
8002 
8003 	msg = mtod(m, struct sadb_msg *);
8004 	msg->sadb_msg_errno = code;
8005 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
8006 }
8007 
8008 /*
8009  * set the pointer to each header into message buffer.
8010  * m will be freed on error.
8011  * XXX larger-than-MCLBYTES extension?
8012  */
8013 static int
8014 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
8015 {
8016 	struct mbuf *n;
8017 	struct sadb_ext *ext;
8018 	size_t off, end;
8019 	int extlen;
8020 	int toff;
8021 
8022 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
8023 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
8024 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8025 		("mbuf too small, len %u", m->m_len));
8026 
8027 	/* initialize */
8028 	bzero(mhp, sizeof(*mhp));
8029 
8030 	mhp->msg = mtod(m, struct sadb_msg *);
8031 	mhp->ext[0] = (struct sadb_ext *)mhp->msg;	/*XXX backward compat */
8032 
8033 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
8034 	extlen = end;	/*just in case extlen is not updated*/
8035 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
8036 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
8037 		if (!n) {
8038 			/* m is already freed */
8039 			return ENOBUFS;
8040 		}
8041 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8042 
8043 		/* set pointer */
8044 		switch (ext->sadb_ext_type) {
8045 		case SADB_EXT_SA:
8046 		case SADB_EXT_ADDRESS_SRC:
8047 		case SADB_EXT_ADDRESS_DST:
8048 		case SADB_EXT_ADDRESS_PROXY:
8049 		case SADB_EXT_LIFETIME_CURRENT:
8050 		case SADB_EXT_LIFETIME_HARD:
8051 		case SADB_EXT_LIFETIME_SOFT:
8052 		case SADB_EXT_KEY_AUTH:
8053 		case SADB_EXT_KEY_ENCRYPT:
8054 		case SADB_EXT_IDENTITY_SRC:
8055 		case SADB_EXT_IDENTITY_DST:
8056 		case SADB_EXT_SENSITIVITY:
8057 		case SADB_EXT_PROPOSAL:
8058 		case SADB_EXT_SUPPORTED_AUTH:
8059 		case SADB_EXT_SUPPORTED_ENCRYPT:
8060 		case SADB_EXT_SPIRANGE:
8061 		case SADB_X_EXT_POLICY:
8062 		case SADB_X_EXT_SA2:
8063 		case SADB_X_EXT_NAT_T_TYPE:
8064 		case SADB_X_EXT_NAT_T_SPORT:
8065 		case SADB_X_EXT_NAT_T_DPORT:
8066 		case SADB_X_EXT_NAT_T_OAI:
8067 		case SADB_X_EXT_NAT_T_OAR:
8068 		case SADB_X_EXT_NAT_T_FRAG:
8069 		case SADB_X_EXT_SA_REPLAY:
8070 		case SADB_X_EXT_NEW_ADDRESS_SRC:
8071 		case SADB_X_EXT_NEW_ADDRESS_DST:
8072 			/* duplicate check */
8073 			/*
8074 			 * XXX Are there duplication payloads of either
8075 			 * KEY_AUTH or KEY_ENCRYPT ?
8076 			 */
8077 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
8078 				ipseclog((LOG_DEBUG, "%s: duplicate ext_type "
8079 					"%u\n", __func__, ext->sadb_ext_type));
8080 				m_freem(m);
8081 				PFKEYSTAT_INC(out_dupext);
8082 				return EINVAL;
8083 			}
8084 			break;
8085 		default:
8086 			ipseclog((LOG_DEBUG, "%s: invalid ext_type %u\n",
8087 				__func__, ext->sadb_ext_type));
8088 			m_freem(m);
8089 			PFKEYSTAT_INC(out_invexttype);
8090 			return EINVAL;
8091 		}
8092 
8093 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
8094 
8095 		if (key_validate_ext(ext, extlen)) {
8096 			m_freem(m);
8097 			PFKEYSTAT_INC(out_invlen);
8098 			return EINVAL;
8099 		}
8100 
8101 		n = m_pulldown(m, off, extlen, &toff);
8102 		if (!n) {
8103 			/* m is already freed */
8104 			return ENOBUFS;
8105 		}
8106 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8107 
8108 		mhp->ext[ext->sadb_ext_type] = ext;
8109 		mhp->extoff[ext->sadb_ext_type] = off;
8110 		mhp->extlen[ext->sadb_ext_type] = extlen;
8111 	}
8112 
8113 	if (off != end) {
8114 		m_freem(m);
8115 		PFKEYSTAT_INC(out_invlen);
8116 		return EINVAL;
8117 	}
8118 
8119 	return 0;
8120 }
8121 
8122 static int
8123 key_validate_ext(const struct sadb_ext *ext, int len)
8124 {
8125 	const struct sockaddr *sa;
8126 	enum { NONE, ADDR } checktype = NONE;
8127 	int baselen = 0;
8128 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
8129 
8130 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
8131 		return EINVAL;
8132 
8133 	/* if it does not match minimum/maximum length, bail */
8134 	if (ext->sadb_ext_type >= nitems(minsize) ||
8135 	    ext->sadb_ext_type >= nitems(maxsize))
8136 		return EINVAL;
8137 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
8138 		return EINVAL;
8139 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
8140 		return EINVAL;
8141 
8142 	/* more checks based on sadb_ext_type XXX need more */
8143 	switch (ext->sadb_ext_type) {
8144 	case SADB_EXT_ADDRESS_SRC:
8145 	case SADB_EXT_ADDRESS_DST:
8146 	case SADB_EXT_ADDRESS_PROXY:
8147 	case SADB_X_EXT_NAT_T_OAI:
8148 	case SADB_X_EXT_NAT_T_OAR:
8149 	case SADB_X_EXT_NEW_ADDRESS_SRC:
8150 	case SADB_X_EXT_NEW_ADDRESS_DST:
8151 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
8152 		checktype = ADDR;
8153 		break;
8154 	case SADB_EXT_IDENTITY_SRC:
8155 	case SADB_EXT_IDENTITY_DST:
8156 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
8157 		    SADB_X_IDENTTYPE_ADDR) {
8158 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
8159 			checktype = ADDR;
8160 		} else
8161 			checktype = NONE;
8162 		break;
8163 	default:
8164 		checktype = NONE;
8165 		break;
8166 	}
8167 
8168 	switch (checktype) {
8169 	case NONE:
8170 		break;
8171 	case ADDR:
8172 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8173 		if (len < baselen + sal)
8174 			return EINVAL;
8175 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8176 			return EINVAL;
8177 		break;
8178 	}
8179 
8180 	return 0;
8181 }
8182 
8183 void
8184 spdcache_init(void)
8185 {
8186 	int i;
8187 
8188 	TUNABLE_INT_FETCH("net.key.spdcache.maxentries",
8189 	    &V_key_spdcache_maxentries);
8190 	TUNABLE_INT_FETCH("net.key.spdcache.threshold",
8191 	    &V_key_spdcache_threshold);
8192 
8193 	if (V_key_spdcache_maxentries) {
8194 		V_key_spdcache_maxentries = MAX(V_key_spdcache_maxentries,
8195 		    SPDCACHE_MAX_ENTRIES_PER_HASH);
8196 		V_spdcachehashtbl = hashinit(V_key_spdcache_maxentries /
8197 		    SPDCACHE_MAX_ENTRIES_PER_HASH,
8198 		    M_IPSEC_SPDCACHE, &V_spdcachehash_mask);
8199 		V_key_spdcache_maxentries = (V_spdcachehash_mask + 1)
8200 		    * SPDCACHE_MAX_ENTRIES_PER_HASH;
8201 
8202 		V_spdcache_lock = malloc(sizeof(struct mtx) *
8203 		    (V_spdcachehash_mask + 1),
8204 		    M_IPSEC_SPDCACHE, M_WAITOK|M_ZERO);
8205 
8206 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8207 			SPDCACHE_LOCK_INIT(i);
8208 	}
8209 }
8210 
8211 struct spdcache_entry *
8212 spdcache_entry_alloc(const struct secpolicyindex *spidx, struct secpolicy *sp)
8213 {
8214 	struct spdcache_entry *entry;
8215 
8216 	entry = malloc(sizeof(struct spdcache_entry),
8217 		    M_IPSEC_SPDCACHE, M_NOWAIT|M_ZERO);
8218 	if (entry == NULL)
8219 		return NULL;
8220 
8221 	if (sp != NULL)
8222 		SP_ADDREF(sp);
8223 
8224 	entry->spidx = *spidx;
8225 	entry->sp = sp;
8226 
8227 	return (entry);
8228 }
8229 
8230 void
8231 spdcache_entry_free(struct spdcache_entry *entry)
8232 {
8233 
8234 	if (entry->sp != NULL)
8235 		key_freesp(&entry->sp);
8236 	free(entry, M_IPSEC_SPDCACHE);
8237 }
8238 
8239 void
8240 spdcache_clear(void)
8241 {
8242 	struct spdcache_entry *entry;
8243 	int i;
8244 
8245 	for (i = 0; i < V_spdcachehash_mask + 1; ++i) {
8246 		SPDCACHE_LOCK(i);
8247 		while (!LIST_EMPTY(&V_spdcachehashtbl[i])) {
8248 			entry = LIST_FIRST(&V_spdcachehashtbl[i]);
8249 			LIST_REMOVE(entry, chain);
8250 			spdcache_entry_free(entry);
8251 		}
8252 		SPDCACHE_UNLOCK(i);
8253 	}
8254 }
8255 
8256 #ifdef VIMAGE
8257 void
8258 spdcache_destroy(void)
8259 {
8260 	int i;
8261 
8262 	if (SPDCACHE_ENABLED()) {
8263 		spdcache_clear();
8264 		hashdestroy(V_spdcachehashtbl, M_IPSEC_SPDCACHE, V_spdcachehash_mask);
8265 
8266 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8267 			SPDCACHE_LOCK_DESTROY(i);
8268 
8269 		free(V_spdcache_lock, M_IPSEC_SPDCACHE);
8270 	}
8271 }
8272 #endif
8273 void
8274 key_init(void)
8275 {
8276 	int i;
8277 
8278 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8279 		TAILQ_INIT(&V_sptree[i]);
8280 		TAILQ_INIT(&V_sptree_ifnet[i]);
8281 	}
8282 
8283 	V_key_lft_zone = uma_zcreate("IPsec SA lft_c",
8284 	    sizeof(uint64_t) * 2, NULL, NULL, NULL, NULL,
8285 	    UMA_ALIGN_PTR, UMA_ZONE_PCPU);
8286 
8287 	TAILQ_INIT(&V_sahtree);
8288 	V_sphashtbl = hashinit(SPHASH_NHASH, M_IPSEC_SP, &V_sphash_mask);
8289 	V_savhashtbl = hashinit(SAVHASH_NHASH, M_IPSEC_SA, &V_savhash_mask);
8290 	V_sahaddrhashtbl = hashinit(SAHHASH_NHASH, M_IPSEC_SAH,
8291 	    &V_sahaddrhash_mask);
8292 	V_acqaddrhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8293 	    &V_acqaddrhash_mask);
8294 	V_acqseqhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8295 	    &V_acqseqhash_mask);
8296 
8297 	spdcache_init();
8298 
8299 	for (i = 0; i <= SADB_SATYPE_MAX; i++)
8300 		LIST_INIT(&V_regtree[i]);
8301 
8302 	LIST_INIT(&V_acqtree);
8303 	LIST_INIT(&V_spacqtree);
8304 
8305 	if (!IS_DEFAULT_VNET(curvnet))
8306 		return;
8307 
8308 	SPTREE_LOCK_INIT();
8309 	REGTREE_LOCK_INIT();
8310 	SAHTREE_LOCK_INIT();
8311 	ACQ_LOCK_INIT();
8312 	SPACQ_LOCK_INIT();
8313 
8314 #ifndef IPSEC_DEBUG2
8315 	callout_init(&key_timer, 1);
8316 	callout_reset(&key_timer, hz, key_timehandler, NULL);
8317 #endif /*IPSEC_DEBUG2*/
8318 
8319 	/* initialize key statistics */
8320 	keystat.getspi_count = 1;
8321 
8322 	if (bootverbose)
8323 		printf("IPsec: Initialized Security Association Processing.\n");
8324 }
8325 
8326 #ifdef VIMAGE
8327 void
8328 key_destroy(void)
8329 {
8330 	struct secashead_queue sahdrainq;
8331 	struct secpolicy_queue drainq;
8332 	struct secpolicy *sp, *nextsp;
8333 	struct secacq *acq, *nextacq;
8334 	struct secspacq *spacq, *nextspacq;
8335 	struct secashead *sah;
8336 	struct secasvar *sav;
8337 	struct secreg *reg;
8338 	int i;
8339 
8340 	/*
8341 	 * XXX: can we just call free() for each object without
8342 	 * walking through safe way with releasing references?
8343 	 */
8344 	TAILQ_INIT(&drainq);
8345 	SPTREE_WLOCK();
8346 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8347 		TAILQ_CONCAT(&drainq, &V_sptree[i], chain);
8348 		TAILQ_CONCAT(&drainq, &V_sptree_ifnet[i], chain);
8349 	}
8350 	for (i = 0; i < V_sphash_mask + 1; i++)
8351 		LIST_INIT(&V_sphashtbl[i]);
8352 	SPTREE_WUNLOCK();
8353 	spdcache_destroy();
8354 
8355 	sp = TAILQ_FIRST(&drainq);
8356 	while (sp != NULL) {
8357 		nextsp = TAILQ_NEXT(sp, chain);
8358 		key_freesp(&sp);
8359 		sp = nextsp;
8360 	}
8361 
8362 	TAILQ_INIT(&sahdrainq);
8363 	SAHTREE_WLOCK();
8364 	TAILQ_CONCAT(&sahdrainq, &V_sahtree, chain);
8365 	for (i = 0; i < V_savhash_mask + 1; i++)
8366 		LIST_INIT(&V_savhashtbl[i]);
8367 	for (i = 0; i < V_sahaddrhash_mask + 1; i++)
8368 		LIST_INIT(&V_sahaddrhashtbl[i]);
8369 	TAILQ_FOREACH(sah, &sahdrainq, chain) {
8370 		sah->state = SADB_SASTATE_DEAD;
8371 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
8372 			sav->state = SADB_SASTATE_DEAD;
8373 		}
8374 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
8375 			sav->state = SADB_SASTATE_DEAD;
8376 		}
8377 	}
8378 	SAHTREE_WUNLOCK();
8379 
8380 	key_freesah_flushed(&sahdrainq);
8381 	hashdestroy(V_sphashtbl, M_IPSEC_SP, V_sphash_mask);
8382 	hashdestroy(V_savhashtbl, M_IPSEC_SA, V_savhash_mask);
8383 	hashdestroy(V_sahaddrhashtbl, M_IPSEC_SAH, V_sahaddrhash_mask);
8384 
8385 	REGTREE_LOCK();
8386 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8387 		LIST_FOREACH(reg, &V_regtree[i], chain) {
8388 			if (__LIST_CHAINED(reg)) {
8389 				LIST_REMOVE(reg, chain);
8390 				free(reg, M_IPSEC_SAR);
8391 				break;
8392 			}
8393 		}
8394 	}
8395 	REGTREE_UNLOCK();
8396 
8397 	ACQ_LOCK();
8398 	acq = LIST_FIRST(&V_acqtree);
8399 	while (acq != NULL) {
8400 		nextacq = LIST_NEXT(acq, chain);
8401 		LIST_REMOVE(acq, chain);
8402 		free(acq, M_IPSEC_SAQ);
8403 		acq = nextacq;
8404 	}
8405 	for (i = 0; i < V_acqaddrhash_mask + 1; i++)
8406 		LIST_INIT(&V_acqaddrhashtbl[i]);
8407 	for (i = 0; i < V_acqseqhash_mask + 1; i++)
8408 		LIST_INIT(&V_acqseqhashtbl[i]);
8409 	ACQ_UNLOCK();
8410 
8411 	SPACQ_LOCK();
8412 	for (spacq = LIST_FIRST(&V_spacqtree); spacq != NULL;
8413 	    spacq = nextspacq) {
8414 		nextspacq = LIST_NEXT(spacq, chain);
8415 		if (__LIST_CHAINED(spacq)) {
8416 			LIST_REMOVE(spacq, chain);
8417 			free(spacq, M_IPSEC_SAQ);
8418 		}
8419 	}
8420 	SPACQ_UNLOCK();
8421 	hashdestroy(V_acqaddrhashtbl, M_IPSEC_SAQ, V_acqaddrhash_mask);
8422 	hashdestroy(V_acqseqhashtbl, M_IPSEC_SAQ, V_acqseqhash_mask);
8423 	uma_zdestroy(V_key_lft_zone);
8424 
8425 	if (!IS_DEFAULT_VNET(curvnet))
8426 		return;
8427 #ifndef IPSEC_DEBUG2
8428 	callout_drain(&key_timer);
8429 #endif
8430 	SPTREE_LOCK_DESTROY();
8431 	REGTREE_LOCK_DESTROY();
8432 	SAHTREE_LOCK_DESTROY();
8433 	ACQ_LOCK_DESTROY();
8434 	SPACQ_LOCK_DESTROY();
8435 }
8436 #endif
8437 
8438 /* record data transfer on SA, and update timestamps */
8439 void
8440 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8441 {
8442 	IPSEC_ASSERT(sav != NULL, ("Null secasvar"));
8443 	IPSEC_ASSERT(m != NULL, ("Null mbuf"));
8444 
8445 	/*
8446 	 * XXX Currently, there is a difference of bytes size
8447 	 * between inbound and outbound processing.
8448 	 */
8449 	counter_u64_add(sav->lft_c_bytes, m->m_pkthdr.len);
8450 
8451 	/*
8452 	 * We use the number of packets as the unit of
8453 	 * allocations.  We increment the variable
8454 	 * whenever {esp,ah}_{in,out}put is called.
8455 	 */
8456 	counter_u64_add(sav->lft_c_allocations, 1);
8457 
8458 	/*
8459 	 * NOTE: We record CURRENT usetime by using wall clock,
8460 	 * in seconds.  HARD and SOFT lifetime are measured by the time
8461 	 * difference (again in seconds) from usetime.
8462 	 *
8463 	 *	usetime
8464 	 *	v     expire   expire
8465 	 * -----+-----+--------+---> t
8466 	 *	<--------------> HARD
8467 	 *	<-----> SOFT
8468 	 */
8469 	if (sav->firstused == 0)
8470 		sav->firstused = time_second;
8471 }
8472 
8473 /*
8474  * Take one of the kernel's security keys and convert it into a PF_KEY
8475  * structure within an mbuf, suitable for sending up to a waiting
8476  * application in user land.
8477  *
8478  * IN:
8479  *    src: A pointer to a kernel security key.
8480  *    exttype: Which type of key this is. Refer to the PF_KEY data structures.
8481  * OUT:
8482  *    a valid mbuf or NULL indicating an error
8483  *
8484  */
8485 
8486 static struct mbuf *
8487 key_setkey(struct seckey *src, uint16_t exttype)
8488 {
8489 	struct mbuf *m;
8490 	struct sadb_key *p;
8491 	int len;
8492 
8493 	if (src == NULL)
8494 		return NULL;
8495 
8496 	len = PFKEY_ALIGN8(sizeof(struct sadb_key) + _KEYLEN(src));
8497 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8498 	if (m == NULL)
8499 		return NULL;
8500 	m_align(m, len);
8501 	m->m_len = len;
8502 	p = mtod(m, struct sadb_key *);
8503 	bzero(p, len);
8504 	p->sadb_key_len = PFKEY_UNIT64(len);
8505 	p->sadb_key_exttype = exttype;
8506 	p->sadb_key_bits = src->bits;
8507 	bcopy(src->key_data, _KEYBUF(p), _KEYLEN(src));
8508 
8509 	return m;
8510 }
8511 
8512 /*
8513  * Take one of the kernel's lifetime data structures and convert it
8514  * into a PF_KEY structure within an mbuf, suitable for sending up to
8515  * a waiting application in user land.
8516  *
8517  * IN:
8518  *    src: A pointer to a kernel lifetime structure.
8519  *    exttype: Which type of lifetime this is. Refer to the PF_KEY
8520  *             data structures for more information.
8521  * OUT:
8522  *    a valid mbuf or NULL indicating an error
8523  *
8524  */
8525 
8526 static struct mbuf *
8527 key_setlifetime(struct seclifetime *src, uint16_t exttype)
8528 {
8529 	struct mbuf *m = NULL;
8530 	struct sadb_lifetime *p;
8531 	int len = PFKEY_ALIGN8(sizeof(struct sadb_lifetime));
8532 
8533 	if (src == NULL)
8534 		return NULL;
8535 
8536 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8537 	if (m == NULL)
8538 		return m;
8539 	m_align(m, len);
8540 	m->m_len = len;
8541 	p = mtod(m, struct sadb_lifetime *);
8542 
8543 	bzero(p, len);
8544 	p->sadb_lifetime_len = PFKEY_UNIT64(len);
8545 	p->sadb_lifetime_exttype = exttype;
8546 	p->sadb_lifetime_allocations = src->allocations;
8547 	p->sadb_lifetime_bytes = src->bytes;
8548 	p->sadb_lifetime_addtime = src->addtime;
8549 	p->sadb_lifetime_usetime = src->usetime;
8550 
8551 	return m;
8552 
8553 }
8554 
8555 const struct enc_xform *
8556 enc_algorithm_lookup(int alg)
8557 {
8558 	int i;
8559 
8560 	for (i = 0; i < nitems(supported_ealgs); i++)
8561 		if (alg == supported_ealgs[i].sadb_alg)
8562 			return (supported_ealgs[i].xform);
8563 	return (NULL);
8564 }
8565 
8566 const struct auth_hash *
8567 auth_algorithm_lookup(int alg)
8568 {
8569 	int i;
8570 
8571 	for (i = 0; i < nitems(supported_aalgs); i++)
8572 		if (alg == supported_aalgs[i].sadb_alg)
8573 			return (supported_aalgs[i].xform);
8574 	return (NULL);
8575 }
8576 
8577 const struct comp_algo *
8578 comp_algorithm_lookup(int alg)
8579 {
8580 	int i;
8581 
8582 	for (i = 0; i < nitems(supported_calgs); i++)
8583 		if (alg == supported_calgs[i].sadb_alg)
8584 			return (supported_calgs[i].xform);
8585 	return (NULL);
8586 }
8587 
8588