xref: /freebsd/sys/netipsec/key.c (revision 983afe3373c427a080f06dccec820b20891be186)
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 static VNET_DEFINE(u_int, key_spi_trycnt) = 1000;
117 static VNET_DEFINE(u_int32_t, key_spi_minval) = 0x100;
118 static VNET_DEFINE(u_int32_t, key_spi_maxval) = 0x0fffffff;	/* XXX */
119 static VNET_DEFINE(u_int32_t, policy_id) = 0;
120 /*interval to initialize randseed,1(m)*/
121 static VNET_DEFINE(u_int, key_int_random) = 60;
122 /* interval to expire acquiring, 30(s)*/
123 static VNET_DEFINE(u_int, key_larval_lifetime) = 30;
124 /* counter for blocking SADB_ACQUIRE.*/
125 static VNET_DEFINE(int, key_blockacq_count) = 10;
126 /* lifetime for blocking SADB_ACQUIRE.*/
127 static VNET_DEFINE(int, key_blockacq_lifetime) = 20;
128 /* preferred old sa rather than new sa.*/
129 static VNET_DEFINE(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 static VNET_DEFINE(u_int32_t, acq_seq) = 0;
141 #define	V_acq_seq		VNET(acq_seq)
142 
143 static VNET_DEFINE(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 static VNET_DEFINE(struct secpolicy_queue, sptree[IPSEC_DIR_MAX]);
150 static VNET_DEFINE(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 static VNET_DEFINE(struct secpolicy_list *, sphashtbl);
167 static VNET_DEFINE(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 static VNET_DEFINE(u_int, key_spdcache_maxentries) = 0;
188 #define	V_key_spdcache_maxentries	VNET(key_spdcache_maxentries)
189 static VNET_DEFINE(u_int, key_spdcache_threshold) = 32;
190 #define	V_key_spdcache_threshold	VNET(key_spdcache_threshold)
191 static VNET_DEFINE(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 static VNET_DEFINE(struct spdcache_entry_list *, spdcachehashtbl);
199 static VNET_DEFINE(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 static VNET_DEFINE(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 static VNET_DEFINE(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 static VNET_DEFINE(struct secashead_list *, sahaddrhashtbl);
237 static VNET_DEFINE(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 static VNET_DEFINE(struct secasvar_list *, savhashtbl);
252 static VNET_DEFINE(u_long, savhash_mask);
253 #define	V_savhashtbl		VNET(savhashtbl)
254 #define	V_savhash_mask		VNET(savhash_mask)
255 #define	SAVHASH_NHASH_LOG2	7
256 #define	SAVHASH_NHASH		(1 << SAVHASH_NHASH_LOG2)
257 #define	SAVHASH_HASHVAL(spi)	(key_u32hash(spi) & V_savhash_mask)
258 #define	SAVHASH_HASH(spi)	&V_savhashtbl[SAVHASH_HASHVAL(spi)]
259 
260 static uint32_t
261 key_addrprotohash(const union sockaddr_union *src,
262     const union sockaddr_union *dst, const uint8_t *proto)
263 {
264 	uint32_t hval;
265 
266 	hval = fnv_32_buf(proto, sizeof(*proto),
267 	    FNV1_32_INIT);
268 	switch (dst->sa.sa_family) {
269 #ifdef INET
270 	case AF_INET:
271 		hval = fnv_32_buf(&src->sin.sin_addr,
272 		    sizeof(in_addr_t), hval);
273 		hval = fnv_32_buf(&dst->sin.sin_addr,
274 		    sizeof(in_addr_t), hval);
275 		break;
276 #endif
277 #ifdef INET6
278 	case AF_INET6:
279 		hval = fnv_32_buf(&src->sin6.sin6_addr,
280 		    sizeof(struct in6_addr), hval);
281 		hval = fnv_32_buf(&dst->sin6.sin6_addr,
282 		    sizeof(struct in6_addr), hval);
283 		break;
284 #endif
285 	default:
286 		hval = 0;
287 		ipseclog((LOG_DEBUG, "%s: unknown address family %d",
288 		    __func__, dst->sa.sa_family));
289 	}
290 	return (hval);
291 }
292 
293 static uint32_t
294 key_u32hash(uint32_t val)
295 {
296 
297 	return (fnv_32_buf(&val, sizeof(val), FNV1_32_INIT));
298 }
299 
300 							/* registed list */
301 static VNET_DEFINE(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 static VNET_DEFINE(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 static VNET_DEFINE(struct secacq_list *, acqaddrhashtbl);
325 static VNET_DEFINE(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 static VNET_DEFINE(struct secacq_list *, acqseqhashtbl);
331 static VNET_DEFINE(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 static VNET_DEFINE(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 static VNET_DEFINE(int, ipsec_esp_keymin) = 256;
437 static VNET_DEFINE(int, ipsec_esp_auth) = 0;
438 static VNET_DEFINE(int, ipsec_ah_keymin) = 128;
439 
440 #define	V_ipsec_esp_keymin	VNET(ipsec_esp_keymin)
441 #define	V_ipsec_esp_auth	VNET(ipsec_esp_auth)
442 #define	V_ipsec_ah_keymin	VNET(ipsec_ah_keymin)
443 
444 #ifdef IPSEC_DEBUG
445 VNET_DEFINE(int, ipsec_debug) = 1;
446 #else
447 VNET_DEFINE(int, ipsec_debug) = 0;
448 #endif
449 
450 #ifdef INET
451 SYSCTL_DECL(_net_inet_ipsec);
452 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
453     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
454     "Enable IPsec debugging output when set.");
455 #endif
456 #ifdef INET6
457 SYSCTL_DECL(_net_inet6_ipsec6);
458 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug,
459     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
460     "Enable IPsec debugging output when set.");
461 #endif
462 
463 SYSCTL_DECL(_net_key);
464 SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,
465 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_debug_level), 0, "");
466 
467 /* max count of trial for the decision of spi value */
468 SYSCTL_INT(_net_key, KEYCTL_SPI_TRY, spi_trycnt,
469 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_trycnt), 0, "");
470 
471 /* minimum spi value to allocate automatically. */
472 SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE, spi_minval,
473 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_minval), 0, "");
474 
475 /* maximun spi value to allocate automatically. */
476 SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE, spi_maxval,
477 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_maxval), 0, "");
478 
479 /* interval to initialize randseed */
480 SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT, int_random,
481 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_int_random), 0, "");
482 
483 /* lifetime for larval SA */
484 SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME, larval_lifetime,
485 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_larval_lifetime), 0, "");
486 
487 /* counter for blocking to send SADB_ACQUIRE to IKEd */
488 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT, blockacq_count,
489 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_count), 0, "");
490 
491 /* lifetime for blocking to send SADB_ACQUIRE to IKEd */
492 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME, blockacq_lifetime,
493 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_lifetime), 0, "");
494 
495 /* ESP auth */
496 SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH, esp_auth,
497 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_auth), 0, "");
498 
499 /* minimum ESP key length */
500 SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN, esp_keymin,
501 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_keymin), 0, "");
502 
503 /* minimum AH key length */
504 SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN, ah_keymin,
505 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_ah_keymin), 0, "");
506 
507 /* perfered old SA rather than new SA */
508 SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA, preferred_oldsa,
509 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_preferred_oldsa), 0, "");
510 
511 static SYSCTL_NODE(_net_key, OID_AUTO, spdcache, CTLFLAG_RW, 0, "SPD cache");
512 
513 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, maxentries,
514 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_maxentries), 0,
515 	"Maximum number of entries in the SPD cache"
516 	" (power of 2, 0 to disable)");
517 
518 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, threshold,
519 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_threshold), 0,
520 	"Number of SPs that make the SPD cache active");
521 
522 #define __LIST_CHAINED(elm) \
523 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
524 
525 MALLOC_DEFINE(M_IPSEC_SA, "secasvar", "ipsec security association");
526 MALLOC_DEFINE(M_IPSEC_SAH, "sahead", "ipsec sa head");
527 MALLOC_DEFINE(M_IPSEC_SP, "ipsecpolicy", "ipsec security policy");
528 MALLOC_DEFINE(M_IPSEC_SR, "ipsecrequest", "ipsec security request");
529 MALLOC_DEFINE(M_IPSEC_MISC, "ipsec-misc", "ipsec miscellaneous");
530 MALLOC_DEFINE(M_IPSEC_SAQ, "ipsec-saq", "ipsec sa acquire");
531 MALLOC_DEFINE(M_IPSEC_SAR, "ipsec-reg", "ipsec sa acquire");
532 MALLOC_DEFINE(M_IPSEC_SPDCACHE, "ipsec-spdcache", "ipsec SPD cache");
533 
534 static VNET_DEFINE(uma_zone_t, key_lft_zone);
535 #define	V_key_lft_zone		VNET(key_lft_zone)
536 
537 static LIST_HEAD(xforms_list, xformsw) xforms = LIST_HEAD_INITIALIZER();
538 static struct mtx xforms_lock;
539 #define	XFORMS_LOCK_INIT()	\
540     mtx_init(&xforms_lock, "xforms_list", "IPsec transforms list", MTX_DEF)
541 #define	XFORMS_LOCK_DESTROY()	mtx_destroy(&xforms_lock)
542 #define	XFORMS_LOCK()		mtx_lock(&xforms_lock)
543 #define	XFORMS_UNLOCK()		mtx_unlock(&xforms_lock)
544 
545 /*
546  * set parameters into secpolicyindex buffer.
547  * Must allocate secpolicyindex buffer passed to this function.
548  */
549 #define KEY_SETSECSPIDX(_dir, s, d, ps, pd, ulp, idx) \
550 do { \
551 	bzero((idx), sizeof(struct secpolicyindex));                         \
552 	(idx)->dir = (_dir);                                                 \
553 	(idx)->prefs = (ps);                                                 \
554 	(idx)->prefd = (pd);                                                 \
555 	(idx)->ul_proto = (ulp);                                             \
556 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
557 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
558 } while (0)
559 
560 /*
561  * set parameters into secasindex buffer.
562  * Must allocate secasindex buffer before calling this function.
563  */
564 #define KEY_SETSECASIDX(p, m, r, s, d, idx) \
565 do { \
566 	bzero((idx), sizeof(struct secasindex));                             \
567 	(idx)->proto = (p);                                                  \
568 	(idx)->mode = (m);                                                   \
569 	(idx)->reqid = (r);                                                  \
570 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
571 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
572 	key_porttosaddr(&(idx)->src.sa, 0);				     \
573 	key_porttosaddr(&(idx)->dst.sa, 0);				     \
574 } while (0)
575 
576 /* key statistics */
577 struct _keystat {
578 	u_long getspi_count; /* the avarage of count to try to get new SPI */
579 } keystat;
580 
581 struct sadb_msghdr {
582 	struct sadb_msg *msg;
583 	struct sadb_ext *ext[SADB_EXT_MAX + 1];
584 	int extoff[SADB_EXT_MAX + 1];
585 	int extlen[SADB_EXT_MAX + 1];
586 };
587 
588 static struct supported_ealgs {
589 	int sadb_alg;
590 	const struct enc_xform *xform;
591 } supported_ealgs[] = {
592 	{ SADB_EALG_DESCBC,		&enc_xform_des },
593 	{ SADB_EALG_3DESCBC,		&enc_xform_3des },
594 	{ SADB_X_EALG_AES,		&enc_xform_rijndael128 },
595 	{ SADB_X_EALG_BLOWFISHCBC,	&enc_xform_blf },
596 	{ SADB_X_EALG_CAST128CBC,	&enc_xform_cast5 },
597 	{ SADB_EALG_NULL,		&enc_xform_null },
598 	{ SADB_X_EALG_CAMELLIACBC,	&enc_xform_camellia },
599 	{ SADB_X_EALG_AESCTR,		&enc_xform_aes_icm },
600 	{ SADB_X_EALG_AESGCM16,		&enc_xform_aes_nist_gcm },
601 	{ SADB_X_EALG_AESGMAC,		&enc_xform_aes_nist_gmac },
602 };
603 
604 static struct supported_aalgs {
605 	int sadb_alg;
606 	const struct auth_hash *xform;
607 } supported_aalgs[] = {
608 	{ SADB_X_AALG_NULL,		&auth_hash_null },
609 	{ SADB_AALG_MD5HMAC,		&auth_hash_hmac_md5 },
610 	{ SADB_AALG_SHA1HMAC,		&auth_hash_hmac_sha1 },
611 	{ SADB_X_AALG_RIPEMD160HMAC,	&auth_hash_hmac_ripemd_160 },
612 	{ SADB_X_AALG_MD5,		&auth_hash_key_md5 },
613 	{ SADB_X_AALG_SHA,		&auth_hash_key_sha1 },
614 	{ SADB_X_AALG_SHA2_256,		&auth_hash_hmac_sha2_256 },
615 	{ SADB_X_AALG_SHA2_384,		&auth_hash_hmac_sha2_384 },
616 	{ SADB_X_AALG_SHA2_512,		&auth_hash_hmac_sha2_512 },
617 	{ SADB_X_AALG_AES128GMAC,	&auth_hash_nist_gmac_aes_128 },
618 	{ SADB_X_AALG_AES192GMAC,	&auth_hash_nist_gmac_aes_192 },
619 	{ SADB_X_AALG_AES256GMAC,	&auth_hash_nist_gmac_aes_256 },
620 };
621 
622 static struct supported_calgs {
623 	int sadb_alg;
624 	const struct comp_algo *xform;
625 } supported_calgs[] = {
626 	{ SADB_X_CALG_DEFLATE,		&comp_algo_deflate },
627 };
628 
629 #ifndef IPSEC_DEBUG2
630 static struct callout key_timer;
631 #endif
632 
633 static void key_unlink(struct secpolicy *);
634 static struct secpolicy *key_do_allocsp(struct secpolicyindex *spidx, u_int dir);
635 static struct secpolicy *key_getsp(struct secpolicyindex *);
636 static struct secpolicy *key_getspbyid(u_int32_t);
637 static struct mbuf *key_gather_mbuf(struct mbuf *,
638 	const struct sadb_msghdr *, int, int, ...);
639 static int key_spdadd(struct socket *, struct mbuf *,
640 	const struct sadb_msghdr *);
641 static uint32_t key_getnewspid(void);
642 static int key_spddelete(struct socket *, struct mbuf *,
643 	const struct sadb_msghdr *);
644 static int key_spddelete2(struct socket *, struct mbuf *,
645 	const struct sadb_msghdr *);
646 static int key_spdget(struct socket *, struct mbuf *,
647 	const struct sadb_msghdr *);
648 static int key_spdflush(struct socket *, struct mbuf *,
649 	const struct sadb_msghdr *);
650 static int key_spddump(struct socket *, struct mbuf *,
651 	const struct sadb_msghdr *);
652 static struct mbuf *key_setdumpsp(struct secpolicy *,
653 	u_int8_t, u_int32_t, u_int32_t);
654 static struct mbuf *key_sp2mbuf(struct secpolicy *);
655 static size_t key_getspreqmsglen(struct secpolicy *);
656 static int key_spdexpire(struct secpolicy *);
657 static struct secashead *key_newsah(struct secasindex *);
658 static void key_freesah(struct secashead **);
659 static void key_delsah(struct secashead *);
660 static struct secasvar *key_newsav(const struct sadb_msghdr *,
661     struct secasindex *, uint32_t, int *);
662 static void key_delsav(struct secasvar *);
663 static void key_unlinksav(struct secasvar *);
664 static struct secashead *key_getsah(struct secasindex *);
665 static int key_checkspidup(uint32_t);
666 static struct secasvar *key_getsavbyspi(uint32_t);
667 static int key_setnatt(struct secasvar *, const struct sadb_msghdr *);
668 static int key_setsaval(struct secasvar *, const struct sadb_msghdr *);
669 static int key_updatelifetimes(struct secasvar *, const struct sadb_msghdr *);
670 static int key_updateaddresses(struct socket *, struct mbuf *,
671     const struct sadb_msghdr *, struct secasvar *, struct secasindex *);
672 
673 static struct mbuf *key_setdumpsa(struct secasvar *, u_int8_t,
674 	u_int8_t, u_int32_t, u_int32_t);
675 static struct mbuf *key_setsadbmsg(u_int8_t, u_int16_t, u_int8_t,
676 	u_int32_t, pid_t, u_int16_t);
677 static struct mbuf *key_setsadbsa(struct secasvar *);
678 static struct mbuf *key_setsadbaddr(u_int16_t,
679 	const struct sockaddr *, u_int8_t, u_int16_t);
680 static struct mbuf *key_setsadbxport(u_int16_t, u_int16_t);
681 static struct mbuf *key_setsadbxtype(u_int16_t);
682 static struct mbuf *key_setsadbxsa2(u_int8_t, u_int32_t, u_int32_t);
683 static struct mbuf *key_setsadbxsareplay(u_int32_t);
684 static struct mbuf *key_setsadbxpolicy(u_int16_t, u_int8_t,
685 	u_int32_t, u_int32_t);
686 static struct seckey *key_dup_keymsg(const struct sadb_key *, size_t,
687     struct malloc_type *);
688 static struct seclifetime *key_dup_lifemsg(const struct sadb_lifetime *src,
689     struct malloc_type *);
690 
691 /* flags for key_cmpsaidx() */
692 #define CMP_HEAD	1	/* protocol, addresses. */
693 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
694 #define CMP_REQID	3	/* additionally HEAD, reaid. */
695 #define CMP_EXACTLY	4	/* all elements. */
696 static int key_cmpsaidx(const struct secasindex *,
697     const struct secasindex *, int);
698 static int key_cmpspidx_exactly(struct secpolicyindex *,
699     struct secpolicyindex *);
700 static int key_cmpspidx_withmask(struct secpolicyindex *,
701     struct secpolicyindex *);
702 static int key_bbcmp(const void *, const void *, u_int);
703 static uint8_t key_satype2proto(uint8_t);
704 static uint8_t key_proto2satype(uint8_t);
705 
706 static int key_getspi(struct socket *, struct mbuf *,
707 	const struct sadb_msghdr *);
708 static uint32_t key_do_getnewspi(struct sadb_spirange *, struct secasindex *);
709 static int key_update(struct socket *, struct mbuf *,
710 	const struct sadb_msghdr *);
711 static int key_add(struct socket *, struct mbuf *,
712 	const struct sadb_msghdr *);
713 static int key_setident(struct secashead *, const struct sadb_msghdr *);
714 static struct mbuf *key_getmsgbuf_x1(struct mbuf *,
715 	const struct sadb_msghdr *);
716 static int key_delete(struct socket *, struct mbuf *,
717 	const struct sadb_msghdr *);
718 static int key_delete_all(struct socket *, struct mbuf *,
719 	const struct sadb_msghdr *, struct secasindex *);
720 static void key_delete_xform(const struct xformsw *);
721 static int key_get(struct socket *, struct mbuf *,
722 	const struct sadb_msghdr *);
723 
724 static void key_getcomb_setlifetime(struct sadb_comb *);
725 static struct mbuf *key_getcomb_ealg(void);
726 static struct mbuf *key_getcomb_ah(void);
727 static struct mbuf *key_getcomb_ipcomp(void);
728 static struct mbuf *key_getprop(const struct secasindex *);
729 
730 static int key_acquire(const struct secasindex *, struct secpolicy *);
731 static uint32_t key_newacq(const struct secasindex *, int *);
732 static uint32_t key_getacq(const struct secasindex *, int *);
733 static int key_acqdone(const struct secasindex *, uint32_t);
734 static int key_acqreset(uint32_t);
735 static struct secspacq *key_newspacq(struct secpolicyindex *);
736 static struct secspacq *key_getspacq(struct secpolicyindex *);
737 static int key_acquire2(struct socket *, struct mbuf *,
738 	const struct sadb_msghdr *);
739 static int key_register(struct socket *, struct mbuf *,
740 	const struct sadb_msghdr *);
741 static int key_expire(struct secasvar *, int);
742 static int key_flush(struct socket *, struct mbuf *,
743 	const struct sadb_msghdr *);
744 static int key_dump(struct socket *, struct mbuf *,
745 	const struct sadb_msghdr *);
746 static int key_promisc(struct socket *, struct mbuf *,
747 	const struct sadb_msghdr *);
748 static int key_senderror(struct socket *, struct mbuf *, int);
749 static int key_validate_ext(const struct sadb_ext *, int);
750 static int key_align(struct mbuf *, struct sadb_msghdr *);
751 static struct mbuf *key_setlifetime(struct seclifetime *, uint16_t);
752 static struct mbuf *key_setkey(struct seckey *, uint16_t);
753 static int xform_init(struct secasvar *, u_short);
754 
755 static void spdcache_init(void);
756 static void spdcache_clear(void);
757 static struct spdcache_entry *spdcache_entry_alloc(
758 	const struct secpolicyindex *spidx,
759 	struct secpolicy *policy);
760 static void spdcache_entry_free(struct spdcache_entry *entry);
761 static void spdcache_destroy(void);
762 
763 
764 #define	DBG_IPSEC_INITREF(t, p)	do {				\
765 	refcount_init(&(p)->refcnt, 1);				\
766 	KEYDBG(KEY_STAMP,					\
767 	    printf("%s: Initialize refcnt %s(%p) = %u\n",	\
768 	    __func__, #t, (p), (p)->refcnt));			\
769 } while (0)
770 #define	DBG_IPSEC_ADDREF(t, p)	do {				\
771 	refcount_acquire(&(p)->refcnt);				\
772 	KEYDBG(KEY_STAMP,					\
773 	    printf("%s: Acquire refcnt %s(%p) -> %u\n",		\
774 	    __func__, #t, (p), (p)->refcnt));			\
775 } while (0)
776 #define	DBG_IPSEC_DELREF(t, p)	do {				\
777 	KEYDBG(KEY_STAMP,					\
778 	    printf("%s: Release refcnt %s(%p) -> %u\n",		\
779 	    __func__, #t, (p), (p)->refcnt - 1));		\
780 	refcount_release(&(p)->refcnt);				\
781 } while (0)
782 
783 #define	IPSEC_INITREF(t, p)	refcount_init(&(p)->refcnt, 1)
784 #define	IPSEC_ADDREF(t, p)	refcount_acquire(&(p)->refcnt)
785 #define	IPSEC_DELREF(t, p)	refcount_release(&(p)->refcnt)
786 
787 #define	SP_INITREF(p)	IPSEC_INITREF(SP, p)
788 #define	SP_ADDREF(p)	IPSEC_ADDREF(SP, p)
789 #define	SP_DELREF(p)	IPSEC_DELREF(SP, p)
790 
791 #define	SAH_INITREF(p)	IPSEC_INITREF(SAH, p)
792 #define	SAH_ADDREF(p)	IPSEC_ADDREF(SAH, p)
793 #define	SAH_DELREF(p)	IPSEC_DELREF(SAH, p)
794 
795 #define	SAV_INITREF(p)	IPSEC_INITREF(SAV, p)
796 #define	SAV_ADDREF(p)	IPSEC_ADDREF(SAV, p)
797 #define	SAV_DELREF(p)	IPSEC_DELREF(SAV, p)
798 
799 /*
800  * Update the refcnt while holding the SPTREE lock.
801  */
802 void
803 key_addref(struct secpolicy *sp)
804 {
805 
806 	SP_ADDREF(sp);
807 }
808 
809 /*
810  * Return 0 when there are known to be no SP's for the specified
811  * direction.  Otherwise return 1.  This is used by IPsec code
812  * to optimize performance.
813  */
814 int
815 key_havesp(u_int dir)
816 {
817 
818 	return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ?
819 		TAILQ_FIRST(&V_sptree[dir]) != NULL : 1);
820 }
821 
822 /* %%% IPsec policy management */
823 /*
824  * Return current SPDB generation.
825  */
826 uint32_t
827 key_getspgen(void)
828 {
829 
830 	return (V_sp_genid);
831 }
832 
833 void
834 key_bumpspgen(void)
835 {
836 
837 	V_sp_genid++;
838 }
839 
840 static int
841 key_checksockaddrs(struct sockaddr *src, struct sockaddr *dst)
842 {
843 
844 	/* family match */
845 	if (src->sa_family != dst->sa_family)
846 		return (EINVAL);
847 	/* sa_len match */
848 	if (src->sa_len != dst->sa_len)
849 		return (EINVAL);
850 	switch (src->sa_family) {
851 #ifdef INET
852 	case AF_INET:
853 		if (src->sa_len != sizeof(struct sockaddr_in))
854 			return (EINVAL);
855 		break;
856 #endif
857 #ifdef INET6
858 	case AF_INET6:
859 		if (src->sa_len != sizeof(struct sockaddr_in6))
860 			return (EINVAL);
861 		break;
862 #endif
863 	default:
864 		return (EAFNOSUPPORT);
865 	}
866 	return (0);
867 }
868 
869 struct secpolicy *
870 key_do_allocsp(struct secpolicyindex *spidx, u_int dir)
871 {
872 	SPTREE_RLOCK_TRACKER;
873 	struct secpolicy *sp;
874 
875 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
876 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
877 		("invalid direction %u", dir));
878 
879 	SPTREE_RLOCK();
880 	TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
881 		if (key_cmpspidx_withmask(&sp->spidx, spidx)) {
882 			SP_ADDREF(sp);
883 			break;
884 		}
885 	}
886 	SPTREE_RUNLOCK();
887 	return (sp);
888 }
889 
890 
891 /*
892  * allocating a SP for OUTBOUND or INBOUND packet.
893  * Must call key_freesp() later.
894  * OUT:	NULL:	not found
895  *	others:	found and return the pointer.
896  */
897 struct secpolicy *
898 key_allocsp(struct secpolicyindex *spidx, u_int dir)
899 {
900 	struct spdcache_entry *entry, *lastentry, *tmpentry;
901 	struct secpolicy *sp;
902 	uint32_t hashv;
903 	int nb_entries;
904 
905 	if (!SPDCACHE_ACTIVE()) {
906 		sp = key_do_allocsp(spidx, dir);
907 		goto out;
908 	}
909 
910 	hashv = SPDCACHE_HASHVAL(spidx);
911 	SPDCACHE_LOCK(hashv);
912 	nb_entries = 0;
913 	LIST_FOREACH_SAFE(entry, &V_spdcachehashtbl[hashv], chain, tmpentry) {
914 		/* Removed outdated entries */
915 		if (entry->sp != NULL &&
916 		    entry->sp->state == IPSEC_SPSTATE_DEAD) {
917 			LIST_REMOVE(entry, chain);
918 			spdcache_entry_free(entry);
919 			continue;
920 		}
921 
922 		nb_entries++;
923 		if (!key_cmpspidx_exactly(&entry->spidx, spidx)) {
924 			lastentry = entry;
925 			continue;
926 		}
927 
928 		sp = entry->sp;
929 		if (entry->sp != NULL)
930 			SP_ADDREF(sp);
931 
932 		IPSECSTAT_INC(ips_spdcache_hits);
933 
934 		SPDCACHE_UNLOCK(hashv);
935 		goto out;
936 	}
937 
938 	IPSECSTAT_INC(ips_spdcache_misses);
939 
940 	sp = key_do_allocsp(spidx, dir);
941 	entry = spdcache_entry_alloc(spidx, sp);
942 	if (entry != NULL) {
943 		if (nb_entries >= SPDCACHE_MAX_ENTRIES_PER_HASH) {
944 			LIST_REMOVE(lastentry, chain);
945 			spdcache_entry_free(lastentry);
946 		}
947 
948 		LIST_INSERT_HEAD(&V_spdcachehashtbl[hashv], entry, chain);
949 	}
950 
951 	SPDCACHE_UNLOCK(hashv);
952 
953 out:
954 	if (sp != NULL) {	/* found a SPD entry */
955 		sp->lastused = time_second;
956 		KEYDBG(IPSEC_STAMP,
957 		    printf("%s: return SP(%p)\n", __func__, sp));
958 		KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
959 	} else {
960 		KEYDBG(IPSEC_DATA,
961 		    printf("%s: lookup failed for ", __func__);
962 		    kdebug_secpolicyindex(spidx, NULL));
963 	}
964 	return (sp);
965 }
966 
967 /*
968  * Allocating an SA entry for an *INBOUND* or *OUTBOUND* TCP packet, signed
969  * or should be signed by MD5 signature.
970  * We don't use key_allocsa() for such lookups, because we don't know SPI.
971  * Unlike ESP and AH protocols, SPI isn't transmitted in the TCP header with
972  * signed packet. We use SADB only as storage for password.
973  * OUT:	positive:	corresponding SA for given saidx found.
974  *	NULL:		SA not found
975  */
976 struct secasvar *
977 key_allocsa_tcpmd5(struct secasindex *saidx)
978 {
979 	SAHTREE_RLOCK_TRACKER;
980 	struct secashead *sah;
981 	struct secasvar *sav;
982 
983 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP,
984 	    ("unexpected security protocol %u", saidx->proto));
985 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TCPMD5,
986 	    ("unexpected mode %u", saidx->mode));
987 
988 	SAHTREE_RLOCK();
989 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
990 		KEYDBG(IPSEC_DUMP,
991 		    printf("%s: checking SAH\n", __func__);
992 		    kdebug_secash(sah, "  "));
993 		if (sah->saidx.proto != IPPROTO_TCP)
994 			continue;
995 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
996 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
997 			break;
998 	}
999 	if (sah != NULL) {
1000 		if (V_key_preferred_oldsa)
1001 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1002 		else
1003 			sav = TAILQ_FIRST(&sah->savtree_alive);
1004 		if (sav != NULL)
1005 			SAV_ADDREF(sav);
1006 	} else
1007 		sav = NULL;
1008 	SAHTREE_RUNLOCK();
1009 
1010 	if (sav != NULL) {
1011 		KEYDBG(IPSEC_STAMP,
1012 		    printf("%s: return SA(%p)\n", __func__, sav));
1013 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1014 	} else {
1015 		KEYDBG(IPSEC_STAMP,
1016 		    printf("%s: SA not found\n", __func__));
1017 		KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1018 	}
1019 	return (sav);
1020 }
1021 
1022 /*
1023  * Allocating an SA entry for an *OUTBOUND* packet.
1024  * OUT:	positive:	corresponding SA for given saidx found.
1025  *	NULL:		SA not found, but will be acquired, check *error
1026  *			for acquiring status.
1027  */
1028 struct secasvar *
1029 key_allocsa_policy(struct secpolicy *sp, const struct secasindex *saidx,
1030     int *error)
1031 {
1032 	SAHTREE_RLOCK_TRACKER;
1033 	struct secashead *sah;
1034 	struct secasvar *sav;
1035 
1036 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
1037 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TRANSPORT ||
1038 		saidx->mode == IPSEC_MODE_TUNNEL,
1039 		("unexpected policy %u", saidx->mode));
1040 
1041 	/*
1042 	 * We check new SA in the IPsec request because a different
1043 	 * SA may be involved each time this request is checked, either
1044 	 * because new SAs are being configured, or this request is
1045 	 * associated with an unconnected datagram socket, or this request
1046 	 * is associated with a system default policy.
1047 	 */
1048 	SAHTREE_RLOCK();
1049 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
1050 		KEYDBG(IPSEC_DUMP,
1051 		    printf("%s: checking SAH\n", __func__);
1052 		    kdebug_secash(sah, "  "));
1053 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID))
1054 			break;
1055 
1056 	}
1057 	if (sah != NULL) {
1058 		/*
1059 		 * Allocate the oldest SA available according to
1060 		 * draft-jenkins-ipsec-rekeying-03.
1061 		 */
1062 		if (V_key_preferred_oldsa)
1063 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1064 		else
1065 			sav = TAILQ_FIRST(&sah->savtree_alive);
1066 		if (sav != NULL)
1067 			SAV_ADDREF(sav);
1068 	} else
1069 		sav = NULL;
1070 	SAHTREE_RUNLOCK();
1071 
1072 	if (sav != NULL) {
1073 		*error = 0;
1074 		KEYDBG(IPSEC_STAMP,
1075 		    printf("%s: chosen SA(%p) for SP(%p)\n", __func__,
1076 			sav, sp));
1077 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1078 		return (sav); /* return referenced SA */
1079 	}
1080 
1081 	/* there is no SA */
1082 	*error = key_acquire(saidx, sp);
1083 	if ((*error) != 0)
1084 		ipseclog((LOG_DEBUG,
1085 		    "%s: error %d returned from key_acquire()\n",
1086 			__func__, *error));
1087 	KEYDBG(IPSEC_STAMP,
1088 	    printf("%s: acquire SA for SP(%p), error %d\n",
1089 		__func__, sp, *error));
1090 	KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1091 	return (NULL);
1092 }
1093 
1094 /*
1095  * allocating a usable SA entry for a *INBOUND* packet.
1096  * Must call key_freesav() later.
1097  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
1098  *	NULL:		not found, or error occurred.
1099  *
1100  * According to RFC 2401 SA is uniquely identified by a triple SPI,
1101  * destination address, and security protocol. But according to RFC 4301,
1102  * SPI by itself suffices to specify an SA.
1103  *
1104  * Note that, however, we do need to keep source address in IPsec SA.
1105  * IKE specification and PF_KEY specification do assume that we
1106  * keep source address in IPsec SA.  We see a tricky situation here.
1107  */
1108 struct secasvar *
1109 key_allocsa(union sockaddr_union *dst, uint8_t proto, uint32_t spi)
1110 {
1111 	SAHTREE_RLOCK_TRACKER;
1112 	struct secasvar *sav;
1113 
1114 	IPSEC_ASSERT(proto == IPPROTO_ESP || proto == IPPROTO_AH ||
1115 	    proto == IPPROTO_IPCOMP, ("unexpected security protocol %u",
1116 	    proto));
1117 
1118 	SAHTREE_RLOCK();
1119 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
1120 		if (sav->spi == spi)
1121 			break;
1122 	}
1123 	/*
1124 	 * We use single SPI namespace for all protocols, so it is
1125 	 * impossible to have SPI duplicates in the SAVHASH.
1126 	 */
1127 	if (sav != NULL) {
1128 		if (sav->state != SADB_SASTATE_LARVAL &&
1129 		    sav->sah->saidx.proto == proto &&
1130 		    key_sockaddrcmp(&dst->sa,
1131 			&sav->sah->saidx.dst.sa, 0) == 0)
1132 			SAV_ADDREF(sav);
1133 		else
1134 			sav = NULL;
1135 	}
1136 	SAHTREE_RUNLOCK();
1137 
1138 	if (sav == NULL) {
1139 		KEYDBG(IPSEC_STAMP,
1140 		    char buf[IPSEC_ADDRSTRLEN];
1141 		    printf("%s: SA not found for spi %u proto %u dst %s\n",
1142 			__func__, ntohl(spi), proto, ipsec_address(dst, buf,
1143 			sizeof(buf))));
1144 	} else {
1145 		KEYDBG(IPSEC_STAMP,
1146 		    printf("%s: return SA(%p)\n", __func__, sav));
1147 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1148 	}
1149 	return (sav);
1150 }
1151 
1152 struct secasvar *
1153 key_allocsa_tunnel(union sockaddr_union *src, union sockaddr_union *dst,
1154     uint8_t proto)
1155 {
1156 	SAHTREE_RLOCK_TRACKER;
1157 	struct secasindex saidx;
1158 	struct secashead *sah;
1159 	struct secasvar *sav;
1160 
1161 	IPSEC_ASSERT(src != NULL, ("null src address"));
1162 	IPSEC_ASSERT(dst != NULL, ("null dst address"));
1163 
1164 	KEY_SETSECASIDX(proto, IPSEC_MODE_TUNNEL, 0, &src->sa,
1165 	    &dst->sa, &saidx);
1166 
1167 	sav = NULL;
1168 	SAHTREE_RLOCK();
1169 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
1170 		if (IPSEC_MODE_TUNNEL != sah->saidx.mode)
1171 			continue;
1172 		if (proto != sah->saidx.proto)
1173 			continue;
1174 		if (key_sockaddrcmp(&src->sa, &sah->saidx.src.sa, 0) != 0)
1175 			continue;
1176 		if (key_sockaddrcmp(&dst->sa, &sah->saidx.dst.sa, 0) != 0)
1177 			continue;
1178 		/* XXXAE: is key_preferred_oldsa reasonably?*/
1179 		if (V_key_preferred_oldsa)
1180 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1181 		else
1182 			sav = TAILQ_FIRST(&sah->savtree_alive);
1183 		if (sav != NULL) {
1184 			SAV_ADDREF(sav);
1185 			break;
1186 		}
1187 	}
1188 	SAHTREE_RUNLOCK();
1189 	KEYDBG(IPSEC_STAMP,
1190 	    printf("%s: return SA(%p)\n", __func__, sav));
1191 	if (sav != NULL)
1192 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1193 	return (sav);
1194 }
1195 
1196 /*
1197  * Must be called after calling key_allocsp().
1198  */
1199 void
1200 key_freesp(struct secpolicy **spp)
1201 {
1202 	struct secpolicy *sp = *spp;
1203 
1204 	IPSEC_ASSERT(sp != NULL, ("null sp"));
1205 	if (SP_DELREF(sp) == 0)
1206 		return;
1207 
1208 	KEYDBG(IPSEC_STAMP,
1209 	    printf("%s: last reference to SP(%p)\n", __func__, sp));
1210 	KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1211 
1212 	*spp = NULL;
1213 	while (sp->tcount > 0)
1214 		ipsec_delisr(sp->req[--sp->tcount]);
1215 	free(sp, M_IPSEC_SP);
1216 }
1217 
1218 static void
1219 key_unlink(struct secpolicy *sp)
1220 {
1221 
1222 	IPSEC_ASSERT(sp->spidx.dir == IPSEC_DIR_INBOUND ||
1223 	    sp->spidx.dir == IPSEC_DIR_OUTBOUND,
1224 	    ("invalid direction %u", sp->spidx.dir));
1225 	SPTREE_UNLOCK_ASSERT();
1226 
1227 	KEYDBG(KEY_STAMP,
1228 	    printf("%s: SP(%p)\n", __func__, sp));
1229 	SPTREE_WLOCK();
1230 	if (sp->state != IPSEC_SPSTATE_ALIVE) {
1231 		/* SP is already unlinked */
1232 		SPTREE_WUNLOCK();
1233 		return;
1234 	}
1235 	sp->state = IPSEC_SPSTATE_DEAD;
1236 	TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
1237 	V_spd_size--;
1238 	LIST_REMOVE(sp, idhash);
1239 	V_sp_genid++;
1240 	SPTREE_WUNLOCK();
1241 	if (SPDCACHE_ENABLED())
1242 		spdcache_clear();
1243 	key_freesp(&sp);
1244 }
1245 
1246 /*
1247  * insert a secpolicy into the SP database. Lower priorities first
1248  */
1249 static void
1250 key_insertsp(struct secpolicy *newsp)
1251 {
1252 	struct secpolicy *sp;
1253 
1254 	SPTREE_WLOCK_ASSERT();
1255 	TAILQ_FOREACH(sp, &V_sptree[newsp->spidx.dir], chain) {
1256 		if (newsp->priority < sp->priority) {
1257 			TAILQ_INSERT_BEFORE(sp, newsp, chain);
1258 			goto done;
1259 		}
1260 	}
1261 	TAILQ_INSERT_TAIL(&V_sptree[newsp->spidx.dir], newsp, chain);
1262 done:
1263 	LIST_INSERT_HEAD(SPHASH_HASH(newsp->id), newsp, idhash);
1264 	newsp->state = IPSEC_SPSTATE_ALIVE;
1265 	V_spd_size++;
1266 	V_sp_genid++;
1267 }
1268 
1269 /*
1270  * Insert a bunch of VTI secpolicies into the SPDB.
1271  * We keep VTI policies in the separate list due to following reasons:
1272  * 1) they should be immutable to user's or some deamon's attempts to
1273  *    delete. The only way delete such policies - destroy or unconfigure
1274  *    corresponding virtual inteface.
1275  * 2) such policies have traffic selector that matches all traffic per
1276  *    address family.
1277  * Since all VTI policies have the same priority, we don't care about
1278  * policies order.
1279  */
1280 int
1281 key_register_ifnet(struct secpolicy **spp, u_int count)
1282 {
1283 	struct mbuf *m;
1284 	u_int i;
1285 
1286 	SPTREE_WLOCK();
1287 	/*
1288 	 * First of try to acquire id for each SP.
1289 	 */
1290 	for (i = 0; i < count; i++) {
1291 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1292 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1293 		    ("invalid direction %u", spp[i]->spidx.dir));
1294 
1295 		if ((spp[i]->id = key_getnewspid()) == 0) {
1296 			SPTREE_WUNLOCK();
1297 			return (EAGAIN);
1298 		}
1299 	}
1300 	for (i = 0; i < count; i++) {
1301 		TAILQ_INSERT_TAIL(&V_sptree_ifnet[spp[i]->spidx.dir],
1302 		    spp[i], chain);
1303 		/*
1304 		 * NOTE: despite the fact that we keep VTI SP in the
1305 		 * separate list, SPHASH contains policies from both
1306 		 * sources. Thus SADB_X_SPDGET will correctly return
1307 		 * SP by id, because it uses SPHASH for lookups.
1308 		 */
1309 		LIST_INSERT_HEAD(SPHASH_HASH(spp[i]->id), spp[i], idhash);
1310 		spp[i]->state = IPSEC_SPSTATE_IFNET;
1311 	}
1312 	SPTREE_WUNLOCK();
1313 	/*
1314 	 * Notify user processes about new SP.
1315 	 */
1316 	for (i = 0; i < count; i++) {
1317 		m = key_setdumpsp(spp[i], SADB_X_SPDADD, 0, 0);
1318 		if (m != NULL)
1319 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1320 	}
1321 	return (0);
1322 }
1323 
1324 void
1325 key_unregister_ifnet(struct secpolicy **spp, u_int count)
1326 {
1327 	struct mbuf *m;
1328 	u_int i;
1329 
1330 	SPTREE_WLOCK();
1331 	for (i = 0; i < count; i++) {
1332 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1333 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1334 		    ("invalid direction %u", spp[i]->spidx.dir));
1335 
1336 		if (spp[i]->state != IPSEC_SPSTATE_IFNET)
1337 			continue;
1338 		spp[i]->state = IPSEC_SPSTATE_DEAD;
1339 		TAILQ_REMOVE(&V_sptree_ifnet[spp[i]->spidx.dir],
1340 		    spp[i], chain);
1341 		V_spd_size--;
1342 		LIST_REMOVE(spp[i], idhash);
1343 	}
1344 	SPTREE_WUNLOCK();
1345 	if (SPDCACHE_ENABLED())
1346 		spdcache_clear();
1347 
1348 	for (i = 0; i < count; i++) {
1349 		m = key_setdumpsp(spp[i], SADB_X_SPDDELETE, 0, 0);
1350 		if (m != NULL)
1351 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1352 	}
1353 }
1354 
1355 /*
1356  * Must be called after calling key_allocsa().
1357  * This function is called by key_freesp() to free some SA allocated
1358  * for a policy.
1359  */
1360 void
1361 key_freesav(struct secasvar **psav)
1362 {
1363 	struct secasvar *sav = *psav;
1364 
1365 	IPSEC_ASSERT(sav != NULL, ("null sav"));
1366 	if (SAV_DELREF(sav) == 0)
1367 		return;
1368 
1369 	KEYDBG(IPSEC_STAMP,
1370 	    printf("%s: last reference to SA(%p)\n", __func__, sav));
1371 
1372 	*psav = NULL;
1373 	key_delsav(sav);
1374 }
1375 
1376 /*
1377  * Unlink SA from SAH and SPI hash under SAHTREE_WLOCK.
1378  * Expect that SA has extra reference due to lookup.
1379  * Release this references, also release SAH reference after unlink.
1380  */
1381 static void
1382 key_unlinksav(struct secasvar *sav)
1383 {
1384 	struct secashead *sah;
1385 
1386 	KEYDBG(KEY_STAMP,
1387 	    printf("%s: SA(%p)\n", __func__, sav));
1388 
1389 	SAHTREE_UNLOCK_ASSERT();
1390 	SAHTREE_WLOCK();
1391 	if (sav->state == SADB_SASTATE_DEAD) {
1392 		/* SA is already unlinked */
1393 		SAHTREE_WUNLOCK();
1394 		return;
1395 	}
1396 	/* Unlink from SAH */
1397 	if (sav->state == SADB_SASTATE_LARVAL)
1398 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
1399 	else
1400 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
1401 	/* Unlink from SPI hash */
1402 	LIST_REMOVE(sav, spihash);
1403 	sav->state = SADB_SASTATE_DEAD;
1404 	sah = sav->sah;
1405 	SAHTREE_WUNLOCK();
1406 	key_freesav(&sav);
1407 	/* Since we are unlinked, release reference to SAH */
1408 	key_freesah(&sah);
1409 }
1410 
1411 /* %%% SPD management */
1412 /*
1413  * search SPD
1414  * OUT:	NULL	: not found
1415  *	others	: found, pointer to a SP.
1416  */
1417 static struct secpolicy *
1418 key_getsp(struct secpolicyindex *spidx)
1419 {
1420 	SPTREE_RLOCK_TRACKER;
1421 	struct secpolicy *sp;
1422 
1423 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
1424 
1425 	SPTREE_RLOCK();
1426 	TAILQ_FOREACH(sp, &V_sptree[spidx->dir], chain) {
1427 		if (key_cmpspidx_exactly(spidx, &sp->spidx)) {
1428 			SP_ADDREF(sp);
1429 			break;
1430 		}
1431 	}
1432 	SPTREE_RUNLOCK();
1433 
1434 	return sp;
1435 }
1436 
1437 /*
1438  * get SP by index.
1439  * OUT:	NULL	: not found
1440  *	others	: found, pointer to referenced SP.
1441  */
1442 static struct secpolicy *
1443 key_getspbyid(uint32_t id)
1444 {
1445 	SPTREE_RLOCK_TRACKER;
1446 	struct secpolicy *sp;
1447 
1448 	SPTREE_RLOCK();
1449 	LIST_FOREACH(sp, SPHASH_HASH(id), idhash) {
1450 		if (sp->id == id) {
1451 			SP_ADDREF(sp);
1452 			break;
1453 		}
1454 	}
1455 	SPTREE_RUNLOCK();
1456 	return (sp);
1457 }
1458 
1459 struct secpolicy *
1460 key_newsp(void)
1461 {
1462 	struct secpolicy *sp;
1463 
1464 	sp = malloc(sizeof(*sp), M_IPSEC_SP, M_NOWAIT | M_ZERO);
1465 	if (sp != NULL)
1466 		SP_INITREF(sp);
1467 	return (sp);
1468 }
1469 
1470 struct ipsecrequest *
1471 ipsec_newisr(void)
1472 {
1473 
1474 	return (malloc(sizeof(struct ipsecrequest), M_IPSEC_SR,
1475 	    M_NOWAIT | M_ZERO));
1476 }
1477 
1478 void
1479 ipsec_delisr(struct ipsecrequest *p)
1480 {
1481 
1482 	free(p, M_IPSEC_SR);
1483 }
1484 
1485 /*
1486  * create secpolicy structure from sadb_x_policy structure.
1487  * NOTE: `state', `secpolicyindex' and 'id' in secpolicy structure
1488  * are not set, so must be set properly later.
1489  */
1490 struct secpolicy *
1491 key_msg2sp(struct sadb_x_policy *xpl0, size_t len, int *error)
1492 {
1493 	struct secpolicy *newsp;
1494 
1495 	IPSEC_ASSERT(xpl0 != NULL, ("null xpl0"));
1496 	IPSEC_ASSERT(len >= sizeof(*xpl0), ("policy too short: %zu", len));
1497 
1498 	if (len != PFKEY_EXTLEN(xpl0)) {
1499 		ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n", __func__));
1500 		*error = EINVAL;
1501 		return NULL;
1502 	}
1503 
1504 	if ((newsp = key_newsp()) == NULL) {
1505 		*error = ENOBUFS;
1506 		return NULL;
1507 	}
1508 
1509 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1510 	newsp->policy = xpl0->sadb_x_policy_type;
1511 	newsp->priority = xpl0->sadb_x_policy_priority;
1512 	newsp->tcount = 0;
1513 
1514 	/* check policy */
1515 	switch (xpl0->sadb_x_policy_type) {
1516 	case IPSEC_POLICY_DISCARD:
1517 	case IPSEC_POLICY_NONE:
1518 	case IPSEC_POLICY_ENTRUST:
1519 	case IPSEC_POLICY_BYPASS:
1520 		break;
1521 
1522 	case IPSEC_POLICY_IPSEC:
1523 	    {
1524 		struct sadb_x_ipsecrequest *xisr;
1525 		struct ipsecrequest *isr;
1526 		int tlen;
1527 
1528 		/* validity check */
1529 		if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1530 			ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n",
1531 				__func__));
1532 			key_freesp(&newsp);
1533 			*error = EINVAL;
1534 			return NULL;
1535 		}
1536 
1537 		tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1538 		xisr = (struct sadb_x_ipsecrequest *)(xpl0 + 1);
1539 
1540 		while (tlen > 0) {
1541 			/* length check */
1542 			if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr) ||
1543 			    xisr->sadb_x_ipsecrequest_len > tlen) {
1544 				ipseclog((LOG_DEBUG, "%s: invalid ipsecrequest "
1545 					"length.\n", __func__));
1546 				key_freesp(&newsp);
1547 				*error = EINVAL;
1548 				return NULL;
1549 			}
1550 
1551 			if (newsp->tcount >= IPSEC_MAXREQ) {
1552 				ipseclog((LOG_DEBUG,
1553 				    "%s: too many ipsecrequests.\n",
1554 				    __func__));
1555 				key_freesp(&newsp);
1556 				*error = EINVAL;
1557 				return (NULL);
1558 			}
1559 
1560 			/* allocate request buffer */
1561 			/* NB: data structure is zero'd */
1562 			isr = ipsec_newisr();
1563 			if (isr == NULL) {
1564 				ipseclog((LOG_DEBUG,
1565 				    "%s: No more memory.\n", __func__));
1566 				key_freesp(&newsp);
1567 				*error = ENOBUFS;
1568 				return NULL;
1569 			}
1570 
1571 			newsp->req[newsp->tcount++] = isr;
1572 
1573 			/* set values */
1574 			switch (xisr->sadb_x_ipsecrequest_proto) {
1575 			case IPPROTO_ESP:
1576 			case IPPROTO_AH:
1577 			case IPPROTO_IPCOMP:
1578 				break;
1579 			default:
1580 				ipseclog((LOG_DEBUG,
1581 				    "%s: invalid proto type=%u\n", __func__,
1582 				    xisr->sadb_x_ipsecrequest_proto));
1583 				key_freesp(&newsp);
1584 				*error = EPROTONOSUPPORT;
1585 				return NULL;
1586 			}
1587 			isr->saidx.proto =
1588 			    (uint8_t)xisr->sadb_x_ipsecrequest_proto;
1589 
1590 			switch (xisr->sadb_x_ipsecrequest_mode) {
1591 			case IPSEC_MODE_TRANSPORT:
1592 			case IPSEC_MODE_TUNNEL:
1593 				break;
1594 			case IPSEC_MODE_ANY:
1595 			default:
1596 				ipseclog((LOG_DEBUG,
1597 				    "%s: invalid mode=%u\n", __func__,
1598 				    xisr->sadb_x_ipsecrequest_mode));
1599 				key_freesp(&newsp);
1600 				*error = EINVAL;
1601 				return NULL;
1602 			}
1603 			isr->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1604 
1605 			switch (xisr->sadb_x_ipsecrequest_level) {
1606 			case IPSEC_LEVEL_DEFAULT:
1607 			case IPSEC_LEVEL_USE:
1608 			case IPSEC_LEVEL_REQUIRE:
1609 				break;
1610 			case IPSEC_LEVEL_UNIQUE:
1611 				/* validity check */
1612 				/*
1613 				 * If range violation of reqid, kernel will
1614 				 * update it, don't refuse it.
1615 				 */
1616 				if (xisr->sadb_x_ipsecrequest_reqid
1617 						> IPSEC_MANUAL_REQID_MAX) {
1618 					ipseclog((LOG_DEBUG,
1619 					    "%s: reqid=%d range "
1620 					    "violation, updated by kernel.\n",
1621 					    __func__,
1622 					    xisr->sadb_x_ipsecrequest_reqid));
1623 					xisr->sadb_x_ipsecrequest_reqid = 0;
1624 				}
1625 
1626 				/* allocate new reqid id if reqid is zero. */
1627 				if (xisr->sadb_x_ipsecrequest_reqid == 0) {
1628 					u_int32_t reqid;
1629 					if ((reqid = key_newreqid()) == 0) {
1630 						key_freesp(&newsp);
1631 						*error = ENOBUFS;
1632 						return NULL;
1633 					}
1634 					isr->saidx.reqid = reqid;
1635 					xisr->sadb_x_ipsecrequest_reqid = reqid;
1636 				} else {
1637 				/* set it for manual keying. */
1638 					isr->saidx.reqid =
1639 					    xisr->sadb_x_ipsecrequest_reqid;
1640 				}
1641 				break;
1642 
1643 			default:
1644 				ipseclog((LOG_DEBUG, "%s: invalid level=%u\n",
1645 					__func__,
1646 					xisr->sadb_x_ipsecrequest_level));
1647 				key_freesp(&newsp);
1648 				*error = EINVAL;
1649 				return NULL;
1650 			}
1651 			isr->level = xisr->sadb_x_ipsecrequest_level;
1652 
1653 			/* set IP addresses if there */
1654 			if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1655 				struct sockaddr *paddr;
1656 
1657 				len = tlen - sizeof(*xisr);
1658 				paddr = (struct sockaddr *)(xisr + 1);
1659 				/* validity check */
1660 				if (len < sizeof(struct sockaddr) ||
1661 				    len < 2 * paddr->sa_len ||
1662 				    paddr->sa_len > sizeof(isr->saidx.src)) {
1663 					ipseclog((LOG_DEBUG, "%s: invalid "
1664 						"request address length.\n",
1665 						__func__));
1666 					key_freesp(&newsp);
1667 					*error = EINVAL;
1668 					return NULL;
1669 				}
1670 				/*
1671 				 * Request length should be enough to keep
1672 				 * source and destination addresses.
1673 				 */
1674 				if (xisr->sadb_x_ipsecrequest_len <
1675 				    sizeof(*xisr) + 2 * paddr->sa_len) {
1676 					ipseclog((LOG_DEBUG, "%s: invalid "
1677 					    "ipsecrequest length.\n",
1678 					    __func__));
1679 					key_freesp(&newsp);
1680 					*error = EINVAL;
1681 					return (NULL);
1682 				}
1683 				bcopy(paddr, &isr->saidx.src, paddr->sa_len);
1684 				paddr = (struct sockaddr *)((caddr_t)paddr +
1685 				    paddr->sa_len);
1686 
1687 				/* validity check */
1688 				if (paddr->sa_len !=
1689 				    isr->saidx.src.sa.sa_len) {
1690 					ipseclog((LOG_DEBUG, "%s: invalid "
1691 						"request address length.\n",
1692 						__func__));
1693 					key_freesp(&newsp);
1694 					*error = EINVAL;
1695 					return NULL;
1696 				}
1697 				/* AF family should match */
1698 				if (paddr->sa_family !=
1699 				    isr->saidx.src.sa.sa_family) {
1700 					ipseclog((LOG_DEBUG, "%s: address "
1701 					    "family doesn't match.\n",
1702 						__func__));
1703 					key_freesp(&newsp);
1704 					*error = EINVAL;
1705 					return (NULL);
1706 				}
1707 				bcopy(paddr, &isr->saidx.dst, paddr->sa_len);
1708 			} else {
1709 				/*
1710 				 * Addresses for TUNNEL mode requests are
1711 				 * mandatory.
1712 				 */
1713 				if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1714 					ipseclog((LOG_DEBUG, "%s: missing "
1715 					    "request addresses.\n", __func__));
1716 					key_freesp(&newsp);
1717 					*error = EINVAL;
1718 					return (NULL);
1719 				}
1720 			}
1721 			tlen -= xisr->sadb_x_ipsecrequest_len;
1722 
1723 			/* validity check */
1724 			if (tlen < 0) {
1725 				ipseclog((LOG_DEBUG, "%s: becoming tlen < 0.\n",
1726 					__func__));
1727 				key_freesp(&newsp);
1728 				*error = EINVAL;
1729 				return NULL;
1730 			}
1731 
1732 			xisr = (struct sadb_x_ipsecrequest *)((caddr_t)xisr
1733 			                 + xisr->sadb_x_ipsecrequest_len);
1734 		}
1735 		/* XXXAE: LARVAL SP */
1736 		if (newsp->tcount < 1) {
1737 			ipseclog((LOG_DEBUG, "%s: valid IPSEC transforms "
1738 			    "not found.\n", __func__));
1739 			key_freesp(&newsp);
1740 			*error = EINVAL;
1741 			return (NULL);
1742 		}
1743 	    }
1744 		break;
1745 	default:
1746 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
1747 		key_freesp(&newsp);
1748 		*error = EINVAL;
1749 		return NULL;
1750 	}
1751 
1752 	*error = 0;
1753 	return (newsp);
1754 }
1755 
1756 uint32_t
1757 key_newreqid(void)
1758 {
1759 	static uint32_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1760 
1761 	if (auto_reqid == ~0)
1762 		auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1763 	else
1764 		auto_reqid++;
1765 
1766 	/* XXX should be unique check */
1767 	return (auto_reqid);
1768 }
1769 
1770 /*
1771  * copy secpolicy struct to sadb_x_policy structure indicated.
1772  */
1773 static struct mbuf *
1774 key_sp2mbuf(struct secpolicy *sp)
1775 {
1776 	struct mbuf *m;
1777 	size_t tlen;
1778 
1779 	tlen = key_getspreqmsglen(sp);
1780 	m = m_get2(tlen, M_NOWAIT, MT_DATA, 0);
1781 	if (m == NULL)
1782 		return (NULL);
1783 	m_align(m, tlen);
1784 	m->m_len = tlen;
1785 	if (key_sp2msg(sp, m->m_data, &tlen) != 0) {
1786 		m_freem(m);
1787 		return (NULL);
1788 	}
1789 	return (m);
1790 }
1791 
1792 int
1793 key_sp2msg(struct secpolicy *sp, void *request, size_t *len)
1794 {
1795 	struct sadb_x_ipsecrequest *xisr;
1796 	struct sadb_x_policy *xpl;
1797 	struct ipsecrequest *isr;
1798 	size_t xlen, ilen;
1799 	caddr_t p;
1800 	int error, i;
1801 
1802 	IPSEC_ASSERT(sp != NULL, ("null policy"));
1803 
1804 	xlen = sizeof(*xpl);
1805 	if (*len < xlen)
1806 		return (EINVAL);
1807 
1808 	error = 0;
1809 	bzero(request, *len);
1810 	xpl = (struct sadb_x_policy *)request;
1811 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
1812 	xpl->sadb_x_policy_type = sp->policy;
1813 	xpl->sadb_x_policy_dir = sp->spidx.dir;
1814 	xpl->sadb_x_policy_id = sp->id;
1815 	xpl->sadb_x_policy_priority = sp->priority;
1816 	switch (sp->state) {
1817 	case IPSEC_SPSTATE_IFNET:
1818 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_IFNET;
1819 		break;
1820 	case IPSEC_SPSTATE_PCB:
1821 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_PCB;
1822 		break;
1823 	default:
1824 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_GLOBAL;
1825 	}
1826 
1827 	/* if is the policy for ipsec ? */
1828 	if (sp->policy == IPSEC_POLICY_IPSEC) {
1829 		p = (caddr_t)xpl + sizeof(*xpl);
1830 		for (i = 0; i < sp->tcount; i++) {
1831 			isr = sp->req[i];
1832 			ilen = PFKEY_ALIGN8(sizeof(*xisr) +
1833 			    isr->saidx.src.sa.sa_len +
1834 			    isr->saidx.dst.sa.sa_len);
1835 			xlen += ilen;
1836 			if (xlen > *len) {
1837 				error = ENOBUFS;
1838 				/* Calculate needed size */
1839 				continue;
1840 			}
1841 			xisr = (struct sadb_x_ipsecrequest *)p;
1842 			xisr->sadb_x_ipsecrequest_len = ilen;
1843 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
1844 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
1845 			xisr->sadb_x_ipsecrequest_level = isr->level;
1846 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
1847 
1848 			p += sizeof(*xisr);
1849 			bcopy(&isr->saidx.src, p, isr->saidx.src.sa.sa_len);
1850 			p += isr->saidx.src.sa.sa_len;
1851 			bcopy(&isr->saidx.dst, p, isr->saidx.dst.sa.sa_len);
1852 			p += isr->saidx.dst.sa.sa_len;
1853 		}
1854 	}
1855 	xpl->sadb_x_policy_len = PFKEY_UNIT64(xlen);
1856 	if (error == 0)
1857 		*len = xlen;
1858 	else
1859 		*len = sizeof(*xpl);
1860 	return (error);
1861 }
1862 
1863 /* m will not be freed nor modified */
1864 static struct mbuf *
1865 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
1866     int ndeep, int nitem, ...)
1867 {
1868 	va_list ap;
1869 	int idx;
1870 	int i;
1871 	struct mbuf *result = NULL, *n;
1872 	int len;
1873 
1874 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1875 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1876 
1877 	va_start(ap, nitem);
1878 	for (i = 0; i < nitem; i++) {
1879 		idx = va_arg(ap, int);
1880 		if (idx < 0 || idx > SADB_EXT_MAX)
1881 			goto fail;
1882 		/* don't attempt to pull empty extension */
1883 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
1884 			continue;
1885 		if (idx != SADB_EXT_RESERVED  &&
1886 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
1887 			continue;
1888 
1889 		if (idx == SADB_EXT_RESERVED) {
1890 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
1891 
1892 			IPSEC_ASSERT(len <= MHLEN, ("header too big %u", len));
1893 
1894 			MGETHDR(n, M_NOWAIT, MT_DATA);
1895 			if (!n)
1896 				goto fail;
1897 			n->m_len = len;
1898 			n->m_next = NULL;
1899 			m_copydata(m, 0, sizeof(struct sadb_msg),
1900 			    mtod(n, caddr_t));
1901 		} else if (i < ndeep) {
1902 			len = mhp->extlen[idx];
1903 			n = m_get2(len, M_NOWAIT, MT_DATA, 0);
1904 			if (n == NULL)
1905 				goto fail;
1906 			m_align(n, len);
1907 			n->m_len = len;
1908 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
1909 			    mtod(n, caddr_t));
1910 		} else {
1911 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
1912 			    M_NOWAIT);
1913 		}
1914 		if (n == NULL)
1915 			goto fail;
1916 
1917 		if (result)
1918 			m_cat(result, n);
1919 		else
1920 			result = n;
1921 	}
1922 	va_end(ap);
1923 
1924 	if ((result->m_flags & M_PKTHDR) != 0) {
1925 		result->m_pkthdr.len = 0;
1926 		for (n = result; n; n = n->m_next)
1927 			result->m_pkthdr.len += n->m_len;
1928 	}
1929 
1930 	return result;
1931 
1932 fail:
1933 	m_freem(result);
1934 	va_end(ap);
1935 	return NULL;
1936 }
1937 
1938 /*
1939  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
1940  * add an entry to SP database, when received
1941  *   <base, address(SD), (lifetime(H),) policy>
1942  * from the user(?).
1943  * Adding to SP database,
1944  * and send
1945  *   <base, address(SD), (lifetime(H),) policy>
1946  * to the socket which was send.
1947  *
1948  * SPDADD set a unique policy entry.
1949  * SPDSETIDX like SPDADD without a part of policy requests.
1950  * SPDUPDATE replace a unique policy entry.
1951  *
1952  * XXXAE: serialize this in PF_KEY to avoid races.
1953  * m will always be freed.
1954  */
1955 static int
1956 key_spdadd(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
1957 {
1958 	struct secpolicyindex spidx;
1959 	struct sadb_address *src0, *dst0;
1960 	struct sadb_x_policy *xpl0, *xpl;
1961 	struct sadb_lifetime *lft = NULL;
1962 	struct secpolicy *newsp;
1963 	int error;
1964 
1965 	IPSEC_ASSERT(so != NULL, ("null socket"));
1966 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1967 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1968 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
1969 
1970 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
1971 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
1972 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
1973 		ipseclog((LOG_DEBUG,
1974 		    "%s: invalid message: missing required header.\n",
1975 		    __func__));
1976 		return key_senderror(so, m, EINVAL);
1977 	}
1978 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
1979 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
1980 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
1981 		ipseclog((LOG_DEBUG,
1982 		    "%s: invalid message: wrong header size.\n", __func__));
1983 		return key_senderror(so, m, EINVAL);
1984 	}
1985 	if (!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD)) {
1986 		if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD)) {
1987 			ipseclog((LOG_DEBUG,
1988 			    "%s: invalid message: wrong header size.\n",
1989 			    __func__));
1990 			return key_senderror(so, m, EINVAL);
1991 		}
1992 		lft = (struct sadb_lifetime *)mhp->ext[SADB_EXT_LIFETIME_HARD];
1993 	}
1994 
1995 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
1996 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
1997 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
1998 
1999 	/* check the direciton */
2000 	switch (xpl0->sadb_x_policy_dir) {
2001 	case IPSEC_DIR_INBOUND:
2002 	case IPSEC_DIR_OUTBOUND:
2003 		break;
2004 	default:
2005 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2006 		return key_senderror(so, m, EINVAL);
2007 	}
2008 	/* key_spdadd() accepts DISCARD, NONE and IPSEC. */
2009 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2010 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2011 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2012 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2013 		return key_senderror(so, m, EINVAL);
2014 	}
2015 
2016 	/* policy requests are mandatory when action is ipsec. */
2017 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
2018 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
2019 		ipseclog((LOG_DEBUG,
2020 		    "%s: policy requests required.\n", __func__));
2021 		return key_senderror(so, m, EINVAL);
2022 	}
2023 
2024 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
2025 	    (struct sockaddr *)(dst0 + 1));
2026 	if (error != 0 ||
2027 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2028 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2029 		return key_senderror(so, m, error);
2030 	}
2031 	/* make secindex */
2032 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2033 	                src0 + 1,
2034 	                dst0 + 1,
2035 	                src0->sadb_address_prefixlen,
2036 	                dst0->sadb_address_prefixlen,
2037 	                src0->sadb_address_proto,
2038 	                &spidx);
2039 	/* Checking there is SP already or not. */
2040 	newsp = key_getsp(&spidx);
2041 	if (newsp != NULL) {
2042 		if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2043 			KEYDBG(KEY_STAMP,
2044 			    printf("%s: unlink SP(%p) for SPDUPDATE\n",
2045 				__func__, newsp));
2046 			KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
2047 			key_unlink(newsp);
2048 			key_freesp(&newsp);
2049 		} else {
2050 			key_freesp(&newsp);
2051 			ipseclog((LOG_DEBUG, "%s: a SP entry exists already.",
2052 			    __func__));
2053 			return (key_senderror(so, m, EEXIST));
2054 		}
2055 	}
2056 
2057 	/* allocate new SP entry */
2058 	if ((newsp = key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error)) == NULL) {
2059 		return key_senderror(so, m, error);
2060 	}
2061 
2062 	newsp->lastused = newsp->created = time_second;
2063 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
2064 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
2065 	bcopy(&spidx, &newsp->spidx, sizeof(spidx));
2066 
2067 	/* XXXAE: there is race between key_getsp() and key_insertsp() */
2068 	SPTREE_WLOCK();
2069 	if ((newsp->id = key_getnewspid()) == 0) {
2070 		SPTREE_WUNLOCK();
2071 		key_freesp(&newsp);
2072 		return key_senderror(so, m, ENOBUFS);
2073 	}
2074 	key_insertsp(newsp);
2075 	SPTREE_WUNLOCK();
2076 	if (SPDCACHE_ENABLED())
2077 		spdcache_clear();
2078 
2079 	KEYDBG(KEY_STAMP,
2080 	    printf("%s: SP(%p)\n", __func__, newsp));
2081 	KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
2082 
2083     {
2084 	struct mbuf *n, *mpolicy;
2085 	struct sadb_msg *newmsg;
2086 	int off;
2087 
2088 	/* create new sadb_msg to reply. */
2089 	if (lft) {
2090 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
2091 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
2092 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2093 	} else {
2094 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
2095 		    SADB_X_EXT_POLICY,
2096 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2097 	}
2098 	if (!n)
2099 		return key_senderror(so, m, ENOBUFS);
2100 
2101 	if (n->m_len < sizeof(*newmsg)) {
2102 		n = m_pullup(n, sizeof(*newmsg));
2103 		if (!n)
2104 			return key_senderror(so, m, ENOBUFS);
2105 	}
2106 	newmsg = mtod(n, struct sadb_msg *);
2107 	newmsg->sadb_msg_errno = 0;
2108 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2109 
2110 	off = 0;
2111 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
2112 	    sizeof(*xpl), &off);
2113 	if (mpolicy == NULL) {
2114 		/* n is already freed */
2115 		return key_senderror(so, m, ENOBUFS);
2116 	}
2117 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, caddr_t) + off);
2118 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
2119 		m_freem(n);
2120 		return key_senderror(so, m, EINVAL);
2121 	}
2122 	xpl->sadb_x_policy_id = newsp->id;
2123 
2124 	m_freem(m);
2125 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2126     }
2127 }
2128 
2129 /*
2130  * get new policy id.
2131  * OUT:
2132  *	0:	failure.
2133  *	others: success.
2134  */
2135 static uint32_t
2136 key_getnewspid(void)
2137 {
2138 	struct secpolicy *sp;
2139 	uint32_t newid = 0;
2140 	int count = V_key_spi_trycnt;	/* XXX */
2141 
2142 	SPTREE_WLOCK_ASSERT();
2143 	while (count--) {
2144 		if (V_policy_id == ~0) /* overflowed */
2145 			newid = V_policy_id = 1;
2146 		else
2147 			newid = ++V_policy_id;
2148 		LIST_FOREACH(sp, SPHASH_HASH(newid), idhash) {
2149 			if (sp->id == newid)
2150 				break;
2151 		}
2152 		if (sp == NULL)
2153 			break;
2154 	}
2155 	if (count == 0 || newid == 0) {
2156 		ipseclog((LOG_DEBUG, "%s: failed to allocate policy id.\n",
2157 		    __func__));
2158 		return (0);
2159 	}
2160 	return (newid);
2161 }
2162 
2163 /*
2164  * SADB_SPDDELETE processing
2165  * receive
2166  *   <base, address(SD), policy(*)>
2167  * from the user(?), and set SADB_SASTATE_DEAD,
2168  * and send,
2169  *   <base, address(SD), policy(*)>
2170  * to the ikmpd.
2171  * policy(*) including direction of policy.
2172  *
2173  * m will always be freed.
2174  */
2175 static int
2176 key_spddelete(struct socket *so, struct mbuf *m,
2177     const struct sadb_msghdr *mhp)
2178 {
2179 	struct secpolicyindex spidx;
2180 	struct sadb_address *src0, *dst0;
2181 	struct sadb_x_policy *xpl0;
2182 	struct secpolicy *sp;
2183 
2184 	IPSEC_ASSERT(so != NULL, ("null so"));
2185 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2186 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2187 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2188 
2189 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
2190 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
2191 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
2192 		ipseclog((LOG_DEBUG,
2193 		    "%s: invalid message: missing required header.\n",
2194 		    __func__));
2195 		return key_senderror(so, m, EINVAL);
2196 	}
2197 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
2198 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
2199 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2200 		ipseclog((LOG_DEBUG,
2201 		    "%s: invalid message: wrong header size.\n", __func__));
2202 		return key_senderror(so, m, EINVAL);
2203 	}
2204 
2205 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
2206 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
2207 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
2208 
2209 	/* check the direciton */
2210 	switch (xpl0->sadb_x_policy_dir) {
2211 	case IPSEC_DIR_INBOUND:
2212 	case IPSEC_DIR_OUTBOUND:
2213 		break;
2214 	default:
2215 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2216 		return key_senderror(so, m, EINVAL);
2217 	}
2218 	/* Only DISCARD, NONE and IPSEC are allowed */
2219 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2220 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2221 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2222 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2223 		return key_senderror(so, m, EINVAL);
2224 	}
2225 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
2226 	    (struct sockaddr *)(dst0 + 1)) != 0 ||
2227 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2228 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2229 		return key_senderror(so, m, EINVAL);
2230 	}
2231 	/* make secindex */
2232 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2233 	                src0 + 1,
2234 	                dst0 + 1,
2235 	                src0->sadb_address_prefixlen,
2236 	                dst0->sadb_address_prefixlen,
2237 	                src0->sadb_address_proto,
2238 	                &spidx);
2239 
2240 	/* Is there SP in SPD ? */
2241 	if ((sp = key_getsp(&spidx)) == NULL) {
2242 		ipseclog((LOG_DEBUG, "%s: no SP found.\n", __func__));
2243 		return key_senderror(so, m, EINVAL);
2244 	}
2245 
2246 	/* save policy id to buffer to be returned. */
2247 	xpl0->sadb_x_policy_id = sp->id;
2248 
2249 	KEYDBG(KEY_STAMP,
2250 	    printf("%s: SP(%p)\n", __func__, sp));
2251 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2252 	key_unlink(sp);
2253 	key_freesp(&sp);
2254 
2255     {
2256 	struct mbuf *n;
2257 	struct sadb_msg *newmsg;
2258 
2259 	/* create new sadb_msg to reply. */
2260 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2261 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2262 	if (!n)
2263 		return key_senderror(so, m, ENOBUFS);
2264 
2265 	newmsg = mtod(n, struct sadb_msg *);
2266 	newmsg->sadb_msg_errno = 0;
2267 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2268 
2269 	m_freem(m);
2270 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2271     }
2272 }
2273 
2274 /*
2275  * SADB_SPDDELETE2 processing
2276  * receive
2277  *   <base, policy(*)>
2278  * from the user(?), and set SADB_SASTATE_DEAD,
2279  * and send,
2280  *   <base, policy(*)>
2281  * to the ikmpd.
2282  * policy(*) including direction of policy.
2283  *
2284  * m will always be freed.
2285  */
2286 static int
2287 key_spddelete2(struct socket *so, struct mbuf *m,
2288     const struct sadb_msghdr *mhp)
2289 {
2290 	struct secpolicy *sp;
2291 	uint32_t id;
2292 
2293 	IPSEC_ASSERT(so != NULL, ("null socket"));
2294 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2295 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2296 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2297 
2298 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2299 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2300 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2301 		    __func__));
2302 		return key_senderror(so, m, EINVAL);
2303 	}
2304 
2305 	id = ((struct sadb_x_policy *)
2306 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2307 
2308 	/* Is there SP in SPD ? */
2309 	if ((sp = key_getspbyid(id)) == NULL) {
2310 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2311 		    __func__, id));
2312 		return key_senderror(so, m, EINVAL);
2313 	}
2314 
2315 	KEYDBG(KEY_STAMP,
2316 	    printf("%s: SP(%p)\n", __func__, sp));
2317 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2318 	key_unlink(sp);
2319 	if (sp->state != IPSEC_SPSTATE_DEAD) {
2320 		ipseclog((LOG_DEBUG, "%s: failed to delete SP with id %u.\n",
2321 		    __func__, id));
2322 		key_freesp(&sp);
2323 		return (key_senderror(so, m, EACCES));
2324 	}
2325 	key_freesp(&sp);
2326 
2327     {
2328 	struct mbuf *n, *nn;
2329 	struct sadb_msg *newmsg;
2330 	int off, len;
2331 
2332 	/* create new sadb_msg to reply. */
2333 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2334 
2335 	MGETHDR(n, M_NOWAIT, MT_DATA);
2336 	if (n && len > MHLEN) {
2337 		if (!(MCLGET(n, M_NOWAIT))) {
2338 			m_freem(n);
2339 			n = NULL;
2340 		}
2341 	}
2342 	if (!n)
2343 		return key_senderror(so, m, ENOBUFS);
2344 
2345 	n->m_len = len;
2346 	n->m_next = NULL;
2347 	off = 0;
2348 
2349 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
2350 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2351 
2352 	IPSEC_ASSERT(off == len, ("length inconsistency (off %u len %u)",
2353 		off, len));
2354 
2355 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2356 	    mhp->extlen[SADB_X_EXT_POLICY], M_NOWAIT);
2357 	if (!n->m_next) {
2358 		m_freem(n);
2359 		return key_senderror(so, m, ENOBUFS);
2360 	}
2361 
2362 	n->m_pkthdr.len = 0;
2363 	for (nn = n; nn; nn = nn->m_next)
2364 		n->m_pkthdr.len += nn->m_len;
2365 
2366 	newmsg = mtod(n, struct sadb_msg *);
2367 	newmsg->sadb_msg_errno = 0;
2368 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2369 
2370 	m_freem(m);
2371 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2372     }
2373 }
2374 
2375 /*
2376  * SADB_X_SPDGET processing
2377  * receive
2378  *   <base, policy(*)>
2379  * from the user(?),
2380  * and send,
2381  *   <base, address(SD), policy>
2382  * to the ikmpd.
2383  * policy(*) including direction of policy.
2384  *
2385  * m will always be freed.
2386  */
2387 static int
2388 key_spdget(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2389 {
2390 	struct secpolicy *sp;
2391 	struct mbuf *n;
2392 	uint32_t id;
2393 
2394 	IPSEC_ASSERT(so != NULL, ("null socket"));
2395 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2396 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2397 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2398 
2399 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2400 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2401 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2402 		    __func__));
2403 		return key_senderror(so, m, EINVAL);
2404 	}
2405 
2406 	id = ((struct sadb_x_policy *)
2407 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2408 
2409 	/* Is there SP in SPD ? */
2410 	if ((sp = key_getspbyid(id)) == NULL) {
2411 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2412 		    __func__, id));
2413 		return key_senderror(so, m, ENOENT);
2414 	}
2415 
2416 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2417 	    mhp->msg->sadb_msg_pid);
2418 	key_freesp(&sp);
2419 	if (n != NULL) {
2420 		m_freem(m);
2421 		return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2422 	} else
2423 		return key_senderror(so, m, ENOBUFS);
2424 }
2425 
2426 /*
2427  * SADB_X_SPDACQUIRE processing.
2428  * Acquire policy and SA(s) for a *OUTBOUND* packet.
2429  * send
2430  *   <base, policy(*)>
2431  * to KMD, and expect to receive
2432  *   <base> with SADB_X_SPDACQUIRE if error occurred,
2433  * or
2434  *   <base, policy>
2435  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2436  * policy(*) is without policy requests.
2437  *
2438  *    0     : succeed
2439  *    others: error number
2440  */
2441 int
2442 key_spdacquire(struct secpolicy *sp)
2443 {
2444 	struct mbuf *result = NULL, *m;
2445 	struct secspacq *newspacq;
2446 
2447 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2448 	IPSEC_ASSERT(sp->req == NULL, ("policy exists"));
2449 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
2450 		("policy not IPSEC %u", sp->policy));
2451 
2452 	/* Get an entry to check whether sent message or not. */
2453 	newspacq = key_getspacq(&sp->spidx);
2454 	if (newspacq != NULL) {
2455 		if (V_key_blockacq_count < newspacq->count) {
2456 			/* reset counter and do send message. */
2457 			newspacq->count = 0;
2458 		} else {
2459 			/* increment counter and do nothing. */
2460 			newspacq->count++;
2461 			SPACQ_UNLOCK();
2462 			return (0);
2463 		}
2464 		SPACQ_UNLOCK();
2465 	} else {
2466 		/* make new entry for blocking to send SADB_ACQUIRE. */
2467 		newspacq = key_newspacq(&sp->spidx);
2468 		if (newspacq == NULL)
2469 			return ENOBUFS;
2470 	}
2471 
2472 	/* create new sadb_msg to reply. */
2473 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2474 	if (!m)
2475 		return ENOBUFS;
2476 
2477 	result = m;
2478 
2479 	result->m_pkthdr.len = 0;
2480 	for (m = result; m; m = m->m_next)
2481 		result->m_pkthdr.len += m->m_len;
2482 
2483 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2484 	    PFKEY_UNIT64(result->m_pkthdr.len);
2485 
2486 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2487 }
2488 
2489 /*
2490  * SADB_SPDFLUSH processing
2491  * receive
2492  *   <base>
2493  * from the user, and free all entries in secpctree.
2494  * and send,
2495  *   <base>
2496  * to the user.
2497  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2498  *
2499  * m will always be freed.
2500  */
2501 static int
2502 key_spdflush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2503 {
2504 	struct secpolicy_queue drainq;
2505 	struct sadb_msg *newmsg;
2506 	struct secpolicy *sp, *nextsp;
2507 	u_int dir;
2508 
2509 	IPSEC_ASSERT(so != NULL, ("null socket"));
2510 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2511 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2512 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2513 
2514 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2515 		return key_senderror(so, m, EINVAL);
2516 
2517 	TAILQ_INIT(&drainq);
2518 	SPTREE_WLOCK();
2519 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2520 		TAILQ_CONCAT(&drainq, &V_sptree[dir], chain);
2521 	}
2522 	/*
2523 	 * We need to set state to DEAD for each policy to be sure,
2524 	 * that another thread won't try to unlink it.
2525 	 * Also remove SP from sphash.
2526 	 */
2527 	TAILQ_FOREACH(sp, &drainq, chain) {
2528 		sp->state = IPSEC_SPSTATE_DEAD;
2529 		LIST_REMOVE(sp, idhash);
2530 	}
2531 	V_sp_genid++;
2532 	V_spd_size = 0;
2533 	SPTREE_WUNLOCK();
2534 	if (SPDCACHE_ENABLED())
2535 		spdcache_clear();
2536 	sp = TAILQ_FIRST(&drainq);
2537 	while (sp != NULL) {
2538 		nextsp = TAILQ_NEXT(sp, chain);
2539 		key_freesp(&sp);
2540 		sp = nextsp;
2541 	}
2542 
2543 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2544 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
2545 		return key_senderror(so, m, ENOBUFS);
2546 	}
2547 
2548 	if (m->m_next)
2549 		m_freem(m->m_next);
2550 	m->m_next = NULL;
2551 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2552 	newmsg = mtod(m, struct sadb_msg *);
2553 	newmsg->sadb_msg_errno = 0;
2554 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2555 
2556 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2557 }
2558 
2559 static uint8_t
2560 key_satype2scopemask(uint8_t satype)
2561 {
2562 
2563 	if (satype == IPSEC_POLICYSCOPE_ANY)
2564 		return (0xff);
2565 	return (satype);
2566 }
2567 /*
2568  * SADB_SPDDUMP processing
2569  * receive
2570  *   <base>
2571  * from the user, and dump all SP leaves and send,
2572  *   <base> .....
2573  * to the ikmpd.
2574  *
2575  * NOTE:
2576  *   sadb_msg_satype is considered as mask of policy scopes.
2577  *   m will always be freed.
2578  */
2579 static int
2580 key_spddump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2581 {
2582 	SPTREE_RLOCK_TRACKER;
2583 	struct secpolicy *sp;
2584 	struct mbuf *n;
2585 	int cnt;
2586 	u_int dir, scope;
2587 
2588 	IPSEC_ASSERT(so != NULL, ("null socket"));
2589 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2590 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2591 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2592 
2593 	/* search SPD entry and get buffer size. */
2594 	cnt = 0;
2595 	scope = key_satype2scopemask(mhp->msg->sadb_msg_satype);
2596 	SPTREE_RLOCK();
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 		}
2602 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2603 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain)
2604 				cnt++;
2605 		}
2606 	}
2607 
2608 	if (cnt == 0) {
2609 		SPTREE_RUNLOCK();
2610 		return key_senderror(so, m, ENOENT);
2611 	}
2612 
2613 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2614 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2615 			TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
2616 				--cnt;
2617 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2618 				    mhp->msg->sadb_msg_pid);
2619 
2620 				if (n != NULL)
2621 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2622 			}
2623 		}
2624 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2625 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain) {
2626 				--cnt;
2627 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2628 				    mhp->msg->sadb_msg_pid);
2629 
2630 				if (n != NULL)
2631 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2632 			}
2633 		}
2634 	}
2635 
2636 	SPTREE_RUNLOCK();
2637 	m_freem(m);
2638 	return (0);
2639 }
2640 
2641 static struct mbuf *
2642 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq,
2643     u_int32_t pid)
2644 {
2645 	struct mbuf *result = NULL, *m;
2646 	struct seclifetime lt;
2647 
2648 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid, sp->refcnt);
2649 	if (!m)
2650 		goto fail;
2651 	result = m;
2652 
2653 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2654 	    &sp->spidx.src.sa, sp->spidx.prefs,
2655 	    sp->spidx.ul_proto);
2656 	if (!m)
2657 		goto fail;
2658 	m_cat(result, m);
2659 
2660 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2661 	    &sp->spidx.dst.sa, sp->spidx.prefd,
2662 	    sp->spidx.ul_proto);
2663 	if (!m)
2664 		goto fail;
2665 	m_cat(result, m);
2666 
2667 	m = key_sp2mbuf(sp);
2668 	if (!m)
2669 		goto fail;
2670 	m_cat(result, m);
2671 
2672 	if(sp->lifetime){
2673 		lt.addtime=sp->created;
2674 		lt.usetime= sp->lastused;
2675 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_CURRENT);
2676 		if (!m)
2677 			goto fail;
2678 		m_cat(result, m);
2679 
2680 		lt.addtime=sp->lifetime;
2681 		lt.usetime= sp->validtime;
2682 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_HARD);
2683 		if (!m)
2684 			goto fail;
2685 		m_cat(result, m);
2686 	}
2687 
2688 	if ((result->m_flags & M_PKTHDR) == 0)
2689 		goto fail;
2690 
2691 	if (result->m_len < sizeof(struct sadb_msg)) {
2692 		result = m_pullup(result, sizeof(struct sadb_msg));
2693 		if (result == NULL)
2694 			goto fail;
2695 	}
2696 
2697 	result->m_pkthdr.len = 0;
2698 	for (m = result; m; m = m->m_next)
2699 		result->m_pkthdr.len += m->m_len;
2700 
2701 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2702 	    PFKEY_UNIT64(result->m_pkthdr.len);
2703 
2704 	return result;
2705 
2706 fail:
2707 	m_freem(result);
2708 	return NULL;
2709 }
2710 /*
2711  * get PFKEY message length for security policy and request.
2712  */
2713 static size_t
2714 key_getspreqmsglen(struct secpolicy *sp)
2715 {
2716 	size_t tlen, len;
2717 	int i;
2718 
2719 	tlen = sizeof(struct sadb_x_policy);
2720 	/* if is the policy for ipsec ? */
2721 	if (sp->policy != IPSEC_POLICY_IPSEC)
2722 		return (tlen);
2723 
2724 	/* get length of ipsec requests */
2725 	for (i = 0; i < sp->tcount; i++) {
2726 		len = sizeof(struct sadb_x_ipsecrequest)
2727 			+ sp->req[i]->saidx.src.sa.sa_len
2728 			+ sp->req[i]->saidx.dst.sa.sa_len;
2729 
2730 		tlen += PFKEY_ALIGN8(len);
2731 	}
2732 	return (tlen);
2733 }
2734 
2735 /*
2736  * SADB_SPDEXPIRE processing
2737  * send
2738  *   <base, address(SD), lifetime(CH), policy>
2739  * to KMD by PF_KEY.
2740  *
2741  * OUT:	0	: succeed
2742  *	others	: error number
2743  */
2744 static int
2745 key_spdexpire(struct secpolicy *sp)
2746 {
2747 	struct sadb_lifetime *lt;
2748 	struct mbuf *result = NULL, *m;
2749 	int len, error = -1;
2750 
2751 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2752 
2753 	KEYDBG(KEY_STAMP,
2754 	    printf("%s: SP(%p)\n", __func__, sp));
2755 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2756 
2757 	/* set msg header */
2758 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0);
2759 	if (!m) {
2760 		error = ENOBUFS;
2761 		goto fail;
2762 	}
2763 	result = m;
2764 
2765 	/* create lifetime extension (current and hard) */
2766 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
2767 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
2768 	if (m == NULL) {
2769 		error = ENOBUFS;
2770 		goto fail;
2771 	}
2772 	m_align(m, len);
2773 	m->m_len = len;
2774 	bzero(mtod(m, caddr_t), len);
2775 	lt = mtod(m, struct sadb_lifetime *);
2776 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2777 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
2778 	lt->sadb_lifetime_allocations = 0;
2779 	lt->sadb_lifetime_bytes = 0;
2780 	lt->sadb_lifetime_addtime = sp->created;
2781 	lt->sadb_lifetime_usetime = sp->lastused;
2782 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
2783 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2784 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
2785 	lt->sadb_lifetime_allocations = 0;
2786 	lt->sadb_lifetime_bytes = 0;
2787 	lt->sadb_lifetime_addtime = sp->lifetime;
2788 	lt->sadb_lifetime_usetime = sp->validtime;
2789 	m_cat(result, m);
2790 
2791 	/* set sadb_address for source */
2792 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2793 	    &sp->spidx.src.sa,
2794 	    sp->spidx.prefs, sp->spidx.ul_proto);
2795 	if (!m) {
2796 		error = ENOBUFS;
2797 		goto fail;
2798 	}
2799 	m_cat(result, m);
2800 
2801 	/* set sadb_address for destination */
2802 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2803 	    &sp->spidx.dst.sa,
2804 	    sp->spidx.prefd, sp->spidx.ul_proto);
2805 	if (!m) {
2806 		error = ENOBUFS;
2807 		goto fail;
2808 	}
2809 	m_cat(result, m);
2810 
2811 	/* set secpolicy */
2812 	m = key_sp2mbuf(sp);
2813 	if (!m) {
2814 		error = ENOBUFS;
2815 		goto fail;
2816 	}
2817 	m_cat(result, m);
2818 
2819 	if ((result->m_flags & M_PKTHDR) == 0) {
2820 		error = EINVAL;
2821 		goto fail;
2822 	}
2823 
2824 	if (result->m_len < sizeof(struct sadb_msg)) {
2825 		result = m_pullup(result, sizeof(struct sadb_msg));
2826 		if (result == NULL) {
2827 			error = ENOBUFS;
2828 			goto fail;
2829 		}
2830 	}
2831 
2832 	result->m_pkthdr.len = 0;
2833 	for (m = result; m; m = m->m_next)
2834 		result->m_pkthdr.len += m->m_len;
2835 
2836 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2837 	    PFKEY_UNIT64(result->m_pkthdr.len);
2838 
2839 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
2840 
2841  fail:
2842 	if (result)
2843 		m_freem(result);
2844 	return error;
2845 }
2846 
2847 /* %%% SAD management */
2848 /*
2849  * allocating and initialize new SA head.
2850  * OUT:	NULL	: failure due to the lack of memory.
2851  *	others	: pointer to new SA head.
2852  */
2853 static struct secashead *
2854 key_newsah(struct secasindex *saidx)
2855 {
2856 	struct secashead *sah;
2857 
2858 	sah = malloc(sizeof(struct secashead), M_IPSEC_SAH,
2859 	    M_NOWAIT | M_ZERO);
2860 	if (sah == NULL) {
2861 		PFKEYSTAT_INC(in_nomem);
2862 		return (NULL);
2863 	}
2864 	TAILQ_INIT(&sah->savtree_larval);
2865 	TAILQ_INIT(&sah->savtree_alive);
2866 	sah->saidx = *saidx;
2867 	sah->state = SADB_SASTATE_DEAD;
2868 	SAH_INITREF(sah);
2869 
2870 	KEYDBG(KEY_STAMP,
2871 	    printf("%s: SAH(%p)\n", __func__, sah));
2872 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2873 	return (sah);
2874 }
2875 
2876 static void
2877 key_freesah(struct secashead **psah)
2878 {
2879 	struct secashead *sah = *psah;
2880 
2881 	if (SAH_DELREF(sah) == 0)
2882 		return;
2883 
2884 	KEYDBG(KEY_STAMP,
2885 	    printf("%s: last reference to SAH(%p)\n", __func__, sah));
2886 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2887 
2888 	*psah = NULL;
2889 	key_delsah(sah);
2890 }
2891 
2892 static void
2893 key_delsah(struct secashead *sah)
2894 {
2895 	IPSEC_ASSERT(sah != NULL, ("NULL sah"));
2896 	IPSEC_ASSERT(sah->state == SADB_SASTATE_DEAD,
2897 	    ("Attempt to free non DEAD SAH %p", sah));
2898 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_larval),
2899 	    ("Attempt to free SAH %p with LARVAL SA", sah));
2900 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_alive),
2901 	    ("Attempt to free SAH %p with ALIVE SA", sah));
2902 
2903 	free(sah, M_IPSEC_SAH);
2904 }
2905 
2906 /*
2907  * allocating a new SA for key_add() and key_getspi() call,
2908  * and copy the values of mhp into new buffer.
2909  * When SAD message type is SADB_GETSPI set SA state to LARVAL.
2910  * For SADB_ADD create and initialize SA with MATURE state.
2911  * OUT:	NULL	: fail
2912  *	others	: pointer to new secasvar.
2913  */
2914 static struct secasvar *
2915 key_newsav(const struct sadb_msghdr *mhp, struct secasindex *saidx,
2916     uint32_t spi, int *errp)
2917 {
2918 	struct secashead *sah;
2919 	struct secasvar *sav;
2920 	int isnew;
2921 
2922 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2923 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2924 	IPSEC_ASSERT(mhp->msg->sadb_msg_type == SADB_GETSPI ||
2925 	    mhp->msg->sadb_msg_type == SADB_ADD, ("wrong message type"));
2926 
2927 	sav = NULL;
2928 	sah = NULL;
2929 	/* check SPI value */
2930 	switch (saidx->proto) {
2931 	case IPPROTO_ESP:
2932 	case IPPROTO_AH:
2933 		/*
2934 		 * RFC 4302, 2.4. Security Parameters Index (SPI), SPI values
2935 		 * 1-255 reserved by IANA for future use,
2936 		 * 0 for implementation specific, local use.
2937 		 */
2938 		if (ntohl(spi) <= 255) {
2939 			ipseclog((LOG_DEBUG, "%s: illegal range of SPI %u.\n",
2940 			    __func__, ntohl(spi)));
2941 			*errp = EINVAL;
2942 			goto done;
2943 		}
2944 		break;
2945 	}
2946 
2947 	sav = malloc(sizeof(struct secasvar), M_IPSEC_SA, M_NOWAIT | M_ZERO);
2948 	if (sav == NULL) {
2949 		*errp = ENOBUFS;
2950 		goto done;
2951 	}
2952 	sav->lock = malloc(sizeof(struct mtx), M_IPSEC_MISC,
2953 	    M_NOWAIT | M_ZERO);
2954 	if (sav->lock == NULL) {
2955 		*errp = ENOBUFS;
2956 		goto done;
2957 	}
2958 	mtx_init(sav->lock, "ipsec association", NULL, MTX_DEF);
2959 	sav->lft_c = uma_zalloc(V_key_lft_zone, M_NOWAIT);
2960 	if (sav->lft_c == NULL) {
2961 		*errp = ENOBUFS;
2962 		goto done;
2963 	}
2964 	counter_u64_zero(sav->lft_c_allocations);
2965 	counter_u64_zero(sav->lft_c_bytes);
2966 
2967 	sav->spi = spi;
2968 	sav->seq = mhp->msg->sadb_msg_seq;
2969 	sav->state = SADB_SASTATE_LARVAL;
2970 	sav->pid = (pid_t)mhp->msg->sadb_msg_pid;
2971 	SAV_INITREF(sav);
2972 again:
2973 	sah = key_getsah(saidx);
2974 	if (sah == NULL) {
2975 		/* create a new SA index */
2976 		sah = key_newsah(saidx);
2977 		if (sah == NULL) {
2978 			ipseclog((LOG_DEBUG,
2979 			    "%s: No more memory.\n", __func__));
2980 			*errp = ENOBUFS;
2981 			goto done;
2982 		}
2983 		isnew = 1;
2984 	} else
2985 		isnew = 0;
2986 
2987 	sav->sah = sah;
2988 	if (mhp->msg->sadb_msg_type == SADB_GETSPI) {
2989 		sav->created = time_second;
2990 	} else if (sav->state == SADB_SASTATE_LARVAL) {
2991 		/*
2992 		 * Do not call key_setsaval() second time in case
2993 		 * of `goto again`. We will have MATURE state.
2994 		 */
2995 		*errp = key_setsaval(sav, mhp);
2996 		if (*errp != 0)
2997 			goto done;
2998 		sav->state = SADB_SASTATE_MATURE;
2999 	}
3000 
3001 	SAHTREE_WLOCK();
3002 	/*
3003 	 * Check that existing SAH wasn't unlinked.
3004 	 * Since we didn't hold the SAHTREE lock, it is possible,
3005 	 * that callout handler or key_flush() or key_delete() could
3006 	 * unlink this SAH.
3007 	 */
3008 	if (isnew == 0 && sah->state == SADB_SASTATE_DEAD) {
3009 		SAHTREE_WUNLOCK();
3010 		key_freesah(&sah);	/* reference from key_getsah() */
3011 		goto again;
3012 	}
3013 	if (isnew != 0) {
3014 		/*
3015 		 * Add new SAH into SADB.
3016 		 *
3017 		 * XXXAE: we can serialize key_add and key_getspi calls, so
3018 		 * several threads will not fight in the race.
3019 		 * Otherwise we should check under SAHTREE lock, that this
3020 		 * SAH would not added twice.
3021 		 */
3022 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
3023 		/* Add new SAH into hash by addresses */
3024 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
3025 		/* Now we are linked in the chain */
3026 		sah->state = SADB_SASTATE_MATURE;
3027 		/*
3028 		 * SAV references this new SAH.
3029 		 * In case of existing SAH we reuse reference
3030 		 * from key_getsah().
3031 		 */
3032 		SAH_ADDREF(sah);
3033 	}
3034 	/* Link SAV with SAH */
3035 	if (sav->state == SADB_SASTATE_MATURE)
3036 		TAILQ_INSERT_HEAD(&sah->savtree_alive, sav, chain);
3037 	else
3038 		TAILQ_INSERT_HEAD(&sah->savtree_larval, sav, chain);
3039 	/* Add SAV into SPI hash */
3040 	LIST_INSERT_HEAD(SAVHASH_HASH(sav->spi), sav, spihash);
3041 	SAHTREE_WUNLOCK();
3042 	*errp = 0;	/* success */
3043 done:
3044 	if (*errp != 0) {
3045 		if (sav != NULL) {
3046 			if (sav->lock != NULL) {
3047 				mtx_destroy(sav->lock);
3048 				free(sav->lock, M_IPSEC_MISC);
3049 			}
3050 			if (sav->lft_c != NULL)
3051 				uma_zfree(V_key_lft_zone, sav->lft_c);
3052 			free(sav, M_IPSEC_SA), sav = NULL;
3053 		}
3054 		if (sah != NULL)
3055 			key_freesah(&sah);
3056 		if (*errp == ENOBUFS) {
3057 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3058 			    __func__));
3059 			PFKEYSTAT_INC(in_nomem);
3060 		}
3061 	}
3062 	return (sav);
3063 }
3064 
3065 /*
3066  * free() SA variable entry.
3067  */
3068 static void
3069 key_cleansav(struct secasvar *sav)
3070 {
3071 
3072 	if (sav->natt != NULL) {
3073 		free(sav->natt, M_IPSEC_MISC);
3074 		sav->natt = NULL;
3075 	}
3076 	if (sav->flags & SADB_X_EXT_F_CLONED)
3077 		return;
3078 	/*
3079 	 * Cleanup xform state.  Note that zeroize'ing causes the
3080 	 * keys to be cleared; otherwise we must do it ourself.
3081 	 */
3082 	if (sav->tdb_xform != NULL) {
3083 		sav->tdb_xform->xf_zeroize(sav);
3084 		sav->tdb_xform = NULL;
3085 	} else {
3086 		if (sav->key_auth != NULL)
3087 			bzero(sav->key_auth->key_data, _KEYLEN(sav->key_auth));
3088 		if (sav->key_enc != NULL)
3089 			bzero(sav->key_enc->key_data, _KEYLEN(sav->key_enc));
3090 	}
3091 	if (sav->key_auth != NULL) {
3092 		if (sav->key_auth->key_data != NULL)
3093 			free(sav->key_auth->key_data, M_IPSEC_MISC);
3094 		free(sav->key_auth, M_IPSEC_MISC);
3095 		sav->key_auth = NULL;
3096 	}
3097 	if (sav->key_enc != NULL) {
3098 		if (sav->key_enc->key_data != NULL)
3099 			free(sav->key_enc->key_data, M_IPSEC_MISC);
3100 		free(sav->key_enc, M_IPSEC_MISC);
3101 		sav->key_enc = NULL;
3102 	}
3103 	if (sav->replay != NULL) {
3104 		if (sav->replay->bitmap != NULL)
3105 			free(sav->replay->bitmap, M_IPSEC_MISC);
3106 		free(sav->replay, M_IPSEC_MISC);
3107 		sav->replay = NULL;
3108 	}
3109 	if (sav->lft_h != NULL) {
3110 		free(sav->lft_h, M_IPSEC_MISC);
3111 		sav->lft_h = NULL;
3112 	}
3113 	if (sav->lft_s != NULL) {
3114 		free(sav->lft_s, M_IPSEC_MISC);
3115 		sav->lft_s = NULL;
3116 	}
3117 }
3118 
3119 /*
3120  * free() SA variable entry.
3121  */
3122 static void
3123 key_delsav(struct secasvar *sav)
3124 {
3125 	IPSEC_ASSERT(sav != NULL, ("null sav"));
3126 	IPSEC_ASSERT(sav->state == SADB_SASTATE_DEAD,
3127 	    ("attempt to free non DEAD SA %p", sav));
3128 	IPSEC_ASSERT(sav->refcnt == 0, ("reference count %u > 0",
3129 	    sav->refcnt));
3130 
3131 	/*
3132 	 * SA must be unlinked from the chain and hashtbl.
3133 	 * If SA was cloned, we leave all fields untouched,
3134 	 * except NAT-T config.
3135 	 */
3136 	key_cleansav(sav);
3137 	if ((sav->flags & SADB_X_EXT_F_CLONED) == 0) {
3138 		mtx_destroy(sav->lock);
3139 		free(sav->lock, M_IPSEC_MISC);
3140 		uma_zfree(V_key_lft_zone, sav->lft_c);
3141 	}
3142 	free(sav, M_IPSEC_SA);
3143 }
3144 
3145 /*
3146  * search SAH.
3147  * OUT:
3148  *	NULL	: not found
3149  *	others	: found, referenced pointer to a SAH.
3150  */
3151 static struct secashead *
3152 key_getsah(struct secasindex *saidx)
3153 {
3154 	SAHTREE_RLOCK_TRACKER;
3155 	struct secashead *sah;
3156 
3157 	SAHTREE_RLOCK();
3158 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
3159 	    if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID) != 0) {
3160 		    SAH_ADDREF(sah);
3161 		    break;
3162 	    }
3163 	}
3164 	SAHTREE_RUNLOCK();
3165 	return (sah);
3166 }
3167 
3168 /*
3169  * Check not to be duplicated SPI.
3170  * OUT:
3171  *	0	: not found
3172  *	1	: found SA with given SPI.
3173  */
3174 static int
3175 key_checkspidup(uint32_t spi)
3176 {
3177 	SAHTREE_RLOCK_TRACKER;
3178 	struct secasvar *sav;
3179 
3180 	/* Assume SPI is in network byte order */
3181 	SAHTREE_RLOCK();
3182 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3183 		if (sav->spi == spi)
3184 			break;
3185 	}
3186 	SAHTREE_RUNLOCK();
3187 	return (sav != NULL);
3188 }
3189 
3190 /*
3191  * Search SA by SPI.
3192  * OUT:
3193  *	NULL	: not found
3194  *	others	: found, referenced pointer to a SA.
3195  */
3196 static struct secasvar *
3197 key_getsavbyspi(uint32_t spi)
3198 {
3199 	SAHTREE_RLOCK_TRACKER;
3200 	struct secasvar *sav;
3201 
3202 	/* Assume SPI is in network byte order */
3203 	SAHTREE_RLOCK();
3204 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3205 		if (sav->spi != spi)
3206 			continue;
3207 		SAV_ADDREF(sav);
3208 		break;
3209 	}
3210 	SAHTREE_RUNLOCK();
3211 	return (sav);
3212 }
3213 
3214 static int
3215 key_updatelifetimes(struct secasvar *sav, const struct sadb_msghdr *mhp)
3216 {
3217 	struct seclifetime *lft_h, *lft_s, *tmp;
3218 
3219 	/* Lifetime extension is optional, check that it is present. */
3220 	if (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3221 	    SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) {
3222 		/*
3223 		 * In case of SADB_UPDATE we may need to change
3224 		 * existing lifetimes.
3225 		 */
3226 		if (sav->state == SADB_SASTATE_MATURE) {
3227 			lft_h = lft_s = NULL;
3228 			goto reset;
3229 		}
3230 		return (0);
3231 	}
3232 	/* Both HARD and SOFT extensions must present */
3233 	if ((SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3234 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
3235 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
3236 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
3237 		ipseclog((LOG_DEBUG,
3238 		    "%s: invalid message: missing required header.\n",
3239 		    __func__));
3240 		return (EINVAL);
3241 	}
3242 	if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD) ||
3243 	    SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_SOFT)) {
3244 		ipseclog((LOG_DEBUG,
3245 		    "%s: invalid message: wrong header size.\n", __func__));
3246 		return (EINVAL);
3247 	}
3248 	lft_h = key_dup_lifemsg((const struct sadb_lifetime *)
3249 	    mhp->ext[SADB_EXT_LIFETIME_HARD], M_IPSEC_MISC);
3250 	if (lft_h == NULL) {
3251 		PFKEYSTAT_INC(in_nomem);
3252 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3253 		return (ENOBUFS);
3254 	}
3255 	lft_s = key_dup_lifemsg((const struct sadb_lifetime *)
3256 	    mhp->ext[SADB_EXT_LIFETIME_SOFT], M_IPSEC_MISC);
3257 	if (lft_s == NULL) {
3258 		PFKEYSTAT_INC(in_nomem);
3259 		free(lft_h, M_IPSEC_MISC);
3260 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3261 		return (ENOBUFS);
3262 	}
3263 reset:
3264 	if (sav->state != SADB_SASTATE_LARVAL) {
3265 		/*
3266 		 * key_update() holds reference to this SA,
3267 		 * so it won't be deleted in meanwhile.
3268 		 */
3269 		SECASVAR_LOCK(sav);
3270 		tmp = sav->lft_h;
3271 		sav->lft_h = lft_h;
3272 		lft_h = tmp;
3273 
3274 		tmp = sav->lft_s;
3275 		sav->lft_s = lft_s;
3276 		lft_s = tmp;
3277 		SECASVAR_UNLOCK(sav);
3278 		if (lft_h != NULL)
3279 			free(lft_h, M_IPSEC_MISC);
3280 		if (lft_s != NULL)
3281 			free(lft_s, M_IPSEC_MISC);
3282 		return (0);
3283 	}
3284 	/* We can update lifetime without holding a lock */
3285 	IPSEC_ASSERT(sav->lft_h == NULL, ("lft_h is already initialized\n"));
3286 	IPSEC_ASSERT(sav->lft_s == NULL, ("lft_s is already initialized\n"));
3287 	sav->lft_h = lft_h;
3288 	sav->lft_s = lft_s;
3289 	return (0);
3290 }
3291 
3292 /*
3293  * copy SA values from PF_KEY message except *SPI, SEQ, PID and TYPE*.
3294  * You must update these if need. Expects only LARVAL SAs.
3295  * OUT:	0:	success.
3296  *	!0:	failure.
3297  */
3298 static int
3299 key_setsaval(struct secasvar *sav, const struct sadb_msghdr *mhp)
3300 {
3301 	const struct sadb_sa *sa0;
3302 	const struct sadb_key *key0;
3303 	uint32_t replay;
3304 	size_t len;
3305 	int error;
3306 
3307 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
3308 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
3309 	IPSEC_ASSERT(sav->state == SADB_SASTATE_LARVAL,
3310 	    ("Attempt to update non LARVAL SA"));
3311 
3312 	/* XXX rewrite */
3313 	error = key_setident(sav->sah, mhp);
3314 	if (error != 0)
3315 		goto fail;
3316 
3317 	/* SA */
3318 	if (!SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
3319 		if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
3320 			error = EINVAL;
3321 			goto fail;
3322 		}
3323 		sa0 = (const struct sadb_sa *)mhp->ext[SADB_EXT_SA];
3324 		sav->alg_auth = sa0->sadb_sa_auth;
3325 		sav->alg_enc = sa0->sadb_sa_encrypt;
3326 		sav->flags = sa0->sadb_sa_flags;
3327 		if ((sav->flags & SADB_KEY_FLAGS_MAX) != sav->flags) {
3328 			ipseclog((LOG_DEBUG,
3329 			    "%s: invalid sa_flags 0x%08x.\n", __func__,
3330 			    sav->flags));
3331 			error = EINVAL;
3332 			goto fail;
3333 		}
3334 
3335 		/* Optional replay window */
3336 		replay = 0;
3337 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0)
3338 			replay = sa0->sadb_sa_replay;
3339 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_SA_REPLAY)) {
3340 			if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA_REPLAY)) {
3341 				error = EINVAL;
3342 				goto fail;
3343 			}
3344 			replay = ((const struct sadb_x_sa_replay *)
3345 			    mhp->ext[SADB_X_EXT_SA_REPLAY])->sadb_x_sa_replay_replay;
3346 
3347 			if (replay > UINT32_MAX - 32) {
3348 				ipseclog((LOG_DEBUG,
3349 				    "%s: replay window too big.\n", __func__));
3350 				error = EINVAL;
3351 				goto fail;
3352 			}
3353 
3354 			replay = (replay + 7) >> 3;
3355 		}
3356 
3357 		sav->replay = malloc(sizeof(struct secreplay), M_IPSEC_MISC,
3358 		    M_NOWAIT | M_ZERO);
3359 		if (sav->replay == NULL) {
3360 			PFKEYSTAT_INC(in_nomem);
3361 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3362 			    __func__));
3363 			error = ENOBUFS;
3364 			goto fail;
3365 		}
3366 
3367 		if (replay != 0) {
3368 			/* number of 32b blocks to be allocated */
3369 			uint32_t bitmap_size;
3370 
3371 			/* RFC 6479:
3372 			 * - the allocated replay window size must be
3373 			 *   a power of two.
3374 			 * - use an extra 32b block as a redundant window.
3375 			 */
3376 			bitmap_size = 1;
3377 			while (replay + 4 > bitmap_size)
3378 				bitmap_size <<= 1;
3379 			bitmap_size = bitmap_size / 4;
3380 
3381 			sav->replay->bitmap = malloc(
3382 			    bitmap_size * sizeof(uint32_t), M_IPSEC_MISC,
3383 			    M_NOWAIT | M_ZERO);
3384 			if (sav->replay->bitmap == NULL) {
3385 				PFKEYSTAT_INC(in_nomem);
3386 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3387 					__func__));
3388 				error = ENOBUFS;
3389 				goto fail;
3390 			}
3391 			sav->replay->bitmap_size = bitmap_size;
3392 			sav->replay->wsize = replay;
3393 		}
3394 	}
3395 
3396 	/* Authentication keys */
3397 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
3398 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH)) {
3399 			error = EINVAL;
3400 			goto fail;
3401 		}
3402 		error = 0;
3403 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_AUTH];
3404 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3405 		switch (mhp->msg->sadb_msg_satype) {
3406 		case SADB_SATYPE_AH:
3407 		case SADB_SATYPE_ESP:
3408 		case SADB_X_SATYPE_TCPSIGNATURE:
3409 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3410 			    sav->alg_auth != SADB_X_AALG_NULL)
3411 				error = EINVAL;
3412 			break;
3413 		case SADB_X_SATYPE_IPCOMP:
3414 		default:
3415 			error = EINVAL;
3416 			break;
3417 		}
3418 		if (error) {
3419 			ipseclog((LOG_DEBUG, "%s: invalid key_auth values.\n",
3420 				__func__));
3421 			goto fail;
3422 		}
3423 
3424 		sav->key_auth = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3425 		if (sav->key_auth == NULL ) {
3426 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3427 				  __func__));
3428 			PFKEYSTAT_INC(in_nomem);
3429 			error = ENOBUFS;
3430 			goto fail;
3431 		}
3432 	}
3433 
3434 	/* Encryption key */
3435 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) {
3436 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT)) {
3437 			error = EINVAL;
3438 			goto fail;
3439 		}
3440 		error = 0;
3441 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_ENCRYPT];
3442 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3443 		switch (mhp->msg->sadb_msg_satype) {
3444 		case SADB_SATYPE_ESP:
3445 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3446 			    sav->alg_enc != SADB_EALG_NULL) {
3447 				error = EINVAL;
3448 				break;
3449 			}
3450 			sav->key_enc = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3451 			if (sav->key_enc == NULL) {
3452 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3453 					__func__));
3454 				PFKEYSTAT_INC(in_nomem);
3455 				error = ENOBUFS;
3456 				goto fail;
3457 			}
3458 			break;
3459 		case SADB_X_SATYPE_IPCOMP:
3460 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3461 				error = EINVAL;
3462 			sav->key_enc = NULL;	/*just in case*/
3463 			break;
3464 		case SADB_SATYPE_AH:
3465 		case SADB_X_SATYPE_TCPSIGNATURE:
3466 		default:
3467 			error = EINVAL;
3468 			break;
3469 		}
3470 		if (error) {
3471 			ipseclog((LOG_DEBUG, "%s: invalid key_enc value.\n",
3472 				__func__));
3473 			goto fail;
3474 		}
3475 	}
3476 
3477 	/* set iv */
3478 	sav->ivlen = 0;
3479 	switch (mhp->msg->sadb_msg_satype) {
3480 	case SADB_SATYPE_AH:
3481 		if (sav->flags & SADB_X_EXT_DERIV) {
3482 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3483 			    "given to AH SA.\n", __func__));
3484 			error = EINVAL;
3485 			goto fail;
3486 		}
3487 		if (sav->alg_enc != SADB_EALG_NONE) {
3488 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3489 			    "mismated.\n", __func__));
3490 			error = EINVAL;
3491 			goto fail;
3492 		}
3493 		error = xform_init(sav, XF_AH);
3494 		break;
3495 	case SADB_SATYPE_ESP:
3496 		if ((sav->flags & (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) ==
3497 		    (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) {
3498 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3499 			    "given to old-esp.\n", __func__));
3500 			error = EINVAL;
3501 			goto fail;
3502 		}
3503 		error = xform_init(sav, XF_ESP);
3504 		break;
3505 	case SADB_X_SATYPE_IPCOMP:
3506 		if (sav->alg_auth != SADB_AALG_NONE) {
3507 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3508 			    "mismated.\n", __func__));
3509 			error = EINVAL;
3510 			goto fail;
3511 		}
3512 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0 &&
3513 		    ntohl(sav->spi) >= 0x10000) {
3514 			ipseclog((LOG_DEBUG, "%s: invalid cpi for IPComp.\n",
3515 			    __func__));
3516 			error = EINVAL;
3517 			goto fail;
3518 		}
3519 		error = xform_init(sav, XF_IPCOMP);
3520 		break;
3521 	case SADB_X_SATYPE_TCPSIGNATURE:
3522 		if (sav->alg_enc != SADB_EALG_NONE) {
3523 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3524 			    "mismated.\n", __func__));
3525 			error = EINVAL;
3526 			goto fail;
3527 		}
3528 		error = xform_init(sav, XF_TCPSIGNATURE);
3529 		break;
3530 	default:
3531 		ipseclog((LOG_DEBUG, "%s: Invalid satype.\n", __func__));
3532 		error = EPROTONOSUPPORT;
3533 		goto fail;
3534 	}
3535 	if (error) {
3536 		ipseclog((LOG_DEBUG, "%s: unable to initialize SA type %u.\n",
3537 		    __func__, mhp->msg->sadb_msg_satype));
3538 		goto fail;
3539 	}
3540 
3541 	/* Handle NAT-T headers */
3542 	error = key_setnatt(sav, mhp);
3543 	if (error != 0)
3544 		goto fail;
3545 
3546 	/* Initialize lifetime for CURRENT */
3547 	sav->firstused = 0;
3548 	sav->created = time_second;
3549 
3550 	/* lifetimes for HARD and SOFT */
3551 	error = key_updatelifetimes(sav, mhp);
3552 	if (error == 0)
3553 		return (0);
3554 fail:
3555 	key_cleansav(sav);
3556 	return (error);
3557 }
3558 
3559 /*
3560  * subroutine for SADB_GET and SADB_DUMP.
3561  */
3562 static struct mbuf *
3563 key_setdumpsa(struct secasvar *sav, uint8_t type, uint8_t satype,
3564     uint32_t seq, uint32_t pid)
3565 {
3566 	struct seclifetime lft_c;
3567 	struct mbuf *result = NULL, *tres = NULL, *m;
3568 	int i, dumporder[] = {
3569 		SADB_EXT_SA, SADB_X_EXT_SA2, SADB_X_EXT_SA_REPLAY,
3570 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3571 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3572 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY,
3573 		SADB_EXT_KEY_AUTH, SADB_EXT_KEY_ENCRYPT,
3574 		SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
3575 		SADB_EXT_SENSITIVITY,
3576 		SADB_X_EXT_NAT_T_TYPE,
3577 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3578 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3579 		SADB_X_EXT_NAT_T_FRAG,
3580 	};
3581 	uint32_t replay_count;
3582 
3583 	m = key_setsadbmsg(type, 0, satype, seq, pid, sav->refcnt);
3584 	if (m == NULL)
3585 		goto fail;
3586 	result = m;
3587 
3588 	for (i = nitems(dumporder) - 1; i >= 0; i--) {
3589 		m = NULL;
3590 		switch (dumporder[i]) {
3591 		case SADB_EXT_SA:
3592 			m = key_setsadbsa(sav);
3593 			if (!m)
3594 				goto fail;
3595 			break;
3596 
3597 		case SADB_X_EXT_SA2:
3598 			SECASVAR_LOCK(sav);
3599 			replay_count = sav->replay ? sav->replay->count : 0;
3600 			SECASVAR_UNLOCK(sav);
3601 			m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
3602 					sav->sah->saidx.reqid);
3603 			if (!m)
3604 				goto fail;
3605 			break;
3606 
3607 		case SADB_X_EXT_SA_REPLAY:
3608 			if (sav->replay == NULL ||
3609 			    sav->replay->wsize <= UINT8_MAX)
3610 				continue;
3611 
3612 			m = key_setsadbxsareplay(sav->replay->wsize);
3613 			if (!m)
3614 				goto fail;
3615 			break;
3616 
3617 		case SADB_EXT_ADDRESS_SRC:
3618 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3619 			    &sav->sah->saidx.src.sa,
3620 			    FULLMASK, IPSEC_ULPROTO_ANY);
3621 			if (!m)
3622 				goto fail;
3623 			break;
3624 
3625 		case SADB_EXT_ADDRESS_DST:
3626 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3627 			    &sav->sah->saidx.dst.sa,
3628 			    FULLMASK, IPSEC_ULPROTO_ANY);
3629 			if (!m)
3630 				goto fail;
3631 			break;
3632 
3633 		case SADB_EXT_KEY_AUTH:
3634 			if (!sav->key_auth)
3635 				continue;
3636 			m = key_setkey(sav->key_auth, SADB_EXT_KEY_AUTH);
3637 			if (!m)
3638 				goto fail;
3639 			break;
3640 
3641 		case SADB_EXT_KEY_ENCRYPT:
3642 			if (!sav->key_enc)
3643 				continue;
3644 			m = key_setkey(sav->key_enc, SADB_EXT_KEY_ENCRYPT);
3645 			if (!m)
3646 				goto fail;
3647 			break;
3648 
3649 		case SADB_EXT_LIFETIME_CURRENT:
3650 			lft_c.addtime = sav->created;
3651 			lft_c.allocations = (uint32_t)counter_u64_fetch(
3652 			    sav->lft_c_allocations);
3653 			lft_c.bytes = counter_u64_fetch(sav->lft_c_bytes);
3654 			lft_c.usetime = sav->firstused;
3655 			m = key_setlifetime(&lft_c, SADB_EXT_LIFETIME_CURRENT);
3656 			if (!m)
3657 				goto fail;
3658 			break;
3659 
3660 		case SADB_EXT_LIFETIME_HARD:
3661 			if (!sav->lft_h)
3662 				continue;
3663 			m = key_setlifetime(sav->lft_h,
3664 					    SADB_EXT_LIFETIME_HARD);
3665 			if (!m)
3666 				goto fail;
3667 			break;
3668 
3669 		case SADB_EXT_LIFETIME_SOFT:
3670 			if (!sav->lft_s)
3671 				continue;
3672 			m = key_setlifetime(sav->lft_s,
3673 					    SADB_EXT_LIFETIME_SOFT);
3674 
3675 			if (!m)
3676 				goto fail;
3677 			break;
3678 
3679 		case SADB_X_EXT_NAT_T_TYPE:
3680 			if (sav->natt == NULL)
3681 				continue;
3682 			m = key_setsadbxtype(UDP_ENCAP_ESPINUDP);
3683 			if (!m)
3684 				goto fail;
3685 			break;
3686 
3687 		case SADB_X_EXT_NAT_T_DPORT:
3688 			if (sav->natt == NULL)
3689 				continue;
3690 			m = key_setsadbxport(sav->natt->dport,
3691 			    SADB_X_EXT_NAT_T_DPORT);
3692 			if (!m)
3693 				goto fail;
3694 			break;
3695 
3696 		case SADB_X_EXT_NAT_T_SPORT:
3697 			if (sav->natt == NULL)
3698 				continue;
3699 			m = key_setsadbxport(sav->natt->sport,
3700 			    SADB_X_EXT_NAT_T_SPORT);
3701 			if (!m)
3702 				goto fail;
3703 			break;
3704 
3705 		case SADB_X_EXT_NAT_T_OAI:
3706 			if (sav->natt == NULL ||
3707 			    (sav->natt->flags & IPSEC_NATT_F_OAI) == 0)
3708 				continue;
3709 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAI,
3710 			    &sav->natt->oai.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3711 			if (!m)
3712 				goto fail;
3713 			break;
3714 		case SADB_X_EXT_NAT_T_OAR:
3715 			if (sav->natt == NULL ||
3716 			    (sav->natt->flags & IPSEC_NATT_F_OAR) == 0)
3717 				continue;
3718 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAR,
3719 			    &sav->natt->oar.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3720 			if (!m)
3721 				goto fail;
3722 			break;
3723 		case SADB_X_EXT_NAT_T_FRAG:
3724 			/* We do not (yet) support those. */
3725 			continue;
3726 
3727 		case SADB_EXT_ADDRESS_PROXY:
3728 		case SADB_EXT_IDENTITY_SRC:
3729 		case SADB_EXT_IDENTITY_DST:
3730 			/* XXX: should we brought from SPD ? */
3731 		case SADB_EXT_SENSITIVITY:
3732 		default:
3733 			continue;
3734 		}
3735 
3736 		if (!m)
3737 			goto fail;
3738 		if (tres)
3739 			m_cat(m, tres);
3740 		tres = m;
3741 	}
3742 
3743 	m_cat(result, tres);
3744 	tres = NULL;
3745 	if (result->m_len < sizeof(struct sadb_msg)) {
3746 		result = m_pullup(result, sizeof(struct sadb_msg));
3747 		if (result == NULL)
3748 			goto fail;
3749 	}
3750 
3751 	result->m_pkthdr.len = 0;
3752 	for (m = result; m; m = m->m_next)
3753 		result->m_pkthdr.len += m->m_len;
3754 
3755 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3756 	    PFKEY_UNIT64(result->m_pkthdr.len);
3757 
3758 	return result;
3759 
3760 fail:
3761 	m_freem(result);
3762 	m_freem(tres);
3763 	return NULL;
3764 }
3765 
3766 /*
3767  * set data into sadb_msg.
3768  */
3769 static struct mbuf *
3770 key_setsadbmsg(u_int8_t type, u_int16_t tlen, u_int8_t satype, u_int32_t seq,
3771     pid_t pid, u_int16_t reserved)
3772 {
3773 	struct mbuf *m;
3774 	struct sadb_msg *p;
3775 	int len;
3776 
3777 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
3778 	if (len > MCLBYTES)
3779 		return NULL;
3780 	MGETHDR(m, M_NOWAIT, MT_DATA);
3781 	if (m && len > MHLEN) {
3782 		if (!(MCLGET(m, M_NOWAIT))) {
3783 			m_freem(m);
3784 			m = NULL;
3785 		}
3786 	}
3787 	if (!m)
3788 		return NULL;
3789 	m->m_pkthdr.len = m->m_len = len;
3790 	m->m_next = NULL;
3791 
3792 	p = mtod(m, struct sadb_msg *);
3793 
3794 	bzero(p, len);
3795 	p->sadb_msg_version = PF_KEY_V2;
3796 	p->sadb_msg_type = type;
3797 	p->sadb_msg_errno = 0;
3798 	p->sadb_msg_satype = satype;
3799 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
3800 	p->sadb_msg_reserved = reserved;
3801 	p->sadb_msg_seq = seq;
3802 	p->sadb_msg_pid = (u_int32_t)pid;
3803 
3804 	return m;
3805 }
3806 
3807 /*
3808  * copy secasvar data into sadb_address.
3809  */
3810 static struct mbuf *
3811 key_setsadbsa(struct secasvar *sav)
3812 {
3813 	struct mbuf *m;
3814 	struct sadb_sa *p;
3815 	int len;
3816 
3817 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
3818 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3819 	if (m == NULL)
3820 		return (NULL);
3821 	m_align(m, len);
3822 	m->m_len = len;
3823 	p = mtod(m, struct sadb_sa *);
3824 	bzero(p, len);
3825 	p->sadb_sa_len = PFKEY_UNIT64(len);
3826 	p->sadb_sa_exttype = SADB_EXT_SA;
3827 	p->sadb_sa_spi = sav->spi;
3828 	p->sadb_sa_replay = sav->replay ?
3829 	    (sav->replay->wsize > UINT8_MAX ? UINT8_MAX :
3830 		sav->replay->wsize): 0;
3831 	p->sadb_sa_state = sav->state;
3832 	p->sadb_sa_auth = sav->alg_auth;
3833 	p->sadb_sa_encrypt = sav->alg_enc;
3834 	p->sadb_sa_flags = sav->flags & SADB_KEY_FLAGS_MAX;
3835 	return (m);
3836 }
3837 
3838 /*
3839  * set data into sadb_address.
3840  */
3841 static struct mbuf *
3842 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
3843     u_int8_t prefixlen, u_int16_t ul_proto)
3844 {
3845 	struct mbuf *m;
3846 	struct sadb_address *p;
3847 	size_t len;
3848 
3849 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
3850 	    PFKEY_ALIGN8(saddr->sa_len);
3851 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3852 	if (m == NULL)
3853 		return (NULL);
3854 	m_align(m, len);
3855 	m->m_len = len;
3856 	p = mtod(m, struct sadb_address *);
3857 
3858 	bzero(p, len);
3859 	p->sadb_address_len = PFKEY_UNIT64(len);
3860 	p->sadb_address_exttype = exttype;
3861 	p->sadb_address_proto = ul_proto;
3862 	if (prefixlen == FULLMASK) {
3863 		switch (saddr->sa_family) {
3864 		case AF_INET:
3865 			prefixlen = sizeof(struct in_addr) << 3;
3866 			break;
3867 		case AF_INET6:
3868 			prefixlen = sizeof(struct in6_addr) << 3;
3869 			break;
3870 		default:
3871 			; /*XXX*/
3872 		}
3873 	}
3874 	p->sadb_address_prefixlen = prefixlen;
3875 	p->sadb_address_reserved = 0;
3876 
3877 	bcopy(saddr,
3878 	    mtod(m, caddr_t) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
3879 	    saddr->sa_len);
3880 
3881 	return m;
3882 }
3883 
3884 /*
3885  * set data into sadb_x_sa2.
3886  */
3887 static struct mbuf *
3888 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int32_t reqid)
3889 {
3890 	struct mbuf *m;
3891 	struct sadb_x_sa2 *p;
3892 	size_t len;
3893 
3894 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
3895 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3896 	if (m == NULL)
3897 		return (NULL);
3898 	m_align(m, len);
3899 	m->m_len = len;
3900 	p = mtod(m, struct sadb_x_sa2 *);
3901 
3902 	bzero(p, len);
3903 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
3904 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
3905 	p->sadb_x_sa2_mode = mode;
3906 	p->sadb_x_sa2_reserved1 = 0;
3907 	p->sadb_x_sa2_reserved2 = 0;
3908 	p->sadb_x_sa2_sequence = seq;
3909 	p->sadb_x_sa2_reqid = reqid;
3910 
3911 	return m;
3912 }
3913 
3914 /*
3915  * Set data into sadb_x_sa_replay.
3916  */
3917 static struct mbuf *
3918 key_setsadbxsareplay(u_int32_t replay)
3919 {
3920 	struct mbuf *m;
3921 	struct sadb_x_sa_replay *p;
3922 	size_t len;
3923 
3924 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa_replay));
3925 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3926 	if (m == NULL)
3927 		return (NULL);
3928 	m_align(m, len);
3929 	m->m_len = len;
3930 	p = mtod(m, struct sadb_x_sa_replay *);
3931 
3932 	bzero(p, len);
3933 	p->sadb_x_sa_replay_len = PFKEY_UNIT64(len);
3934 	p->sadb_x_sa_replay_exttype = SADB_X_EXT_SA_REPLAY;
3935 	p->sadb_x_sa_replay_replay = (replay << 3);
3936 
3937 	return m;
3938 }
3939 
3940 /*
3941  * Set a type in sadb_x_nat_t_type.
3942  */
3943 static struct mbuf *
3944 key_setsadbxtype(u_int16_t type)
3945 {
3946 	struct mbuf *m;
3947 	size_t len;
3948 	struct sadb_x_nat_t_type *p;
3949 
3950 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
3951 
3952 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3953 	if (m == NULL)
3954 		return (NULL);
3955 	m_align(m, len);
3956 	m->m_len = len;
3957 	p = mtod(m, struct sadb_x_nat_t_type *);
3958 
3959 	bzero(p, len);
3960 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
3961 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
3962 	p->sadb_x_nat_t_type_type = type;
3963 
3964 	return (m);
3965 }
3966 /*
3967  * Set a port in sadb_x_nat_t_port.
3968  * In contrast to default RFC 2367 behaviour, port is in network byte order.
3969  */
3970 static struct mbuf *
3971 key_setsadbxport(u_int16_t port, u_int16_t type)
3972 {
3973 	struct mbuf *m;
3974 	size_t len;
3975 	struct sadb_x_nat_t_port *p;
3976 
3977 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
3978 
3979 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3980 	if (m == NULL)
3981 		return (NULL);
3982 	m_align(m, len);
3983 	m->m_len = len;
3984 	p = mtod(m, struct sadb_x_nat_t_port *);
3985 
3986 	bzero(p, len);
3987 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
3988 	p->sadb_x_nat_t_port_exttype = type;
3989 	p->sadb_x_nat_t_port_port = port;
3990 
3991 	return (m);
3992 }
3993 
3994 /*
3995  * Get port from sockaddr. Port is in network byte order.
3996  */
3997 uint16_t
3998 key_portfromsaddr(struct sockaddr *sa)
3999 {
4000 
4001 	switch (sa->sa_family) {
4002 #ifdef INET
4003 	case AF_INET:
4004 		return ((struct sockaddr_in *)sa)->sin_port;
4005 #endif
4006 #ifdef INET6
4007 	case AF_INET6:
4008 		return ((struct sockaddr_in6 *)sa)->sin6_port;
4009 #endif
4010 	}
4011 	return (0);
4012 }
4013 
4014 /*
4015  * Set port in struct sockaddr. Port is in network byte order.
4016  */
4017 void
4018 key_porttosaddr(struct sockaddr *sa, uint16_t port)
4019 {
4020 
4021 	switch (sa->sa_family) {
4022 #ifdef INET
4023 	case AF_INET:
4024 		((struct sockaddr_in *)sa)->sin_port = port;
4025 		break;
4026 #endif
4027 #ifdef INET6
4028 	case AF_INET6:
4029 		((struct sockaddr_in6 *)sa)->sin6_port = port;
4030 		break;
4031 #endif
4032 	default:
4033 		ipseclog((LOG_DEBUG, "%s: unexpected address family %d.\n",
4034 			__func__, sa->sa_family));
4035 		break;
4036 	}
4037 }
4038 
4039 /*
4040  * set data into sadb_x_policy
4041  */
4042 static struct mbuf *
4043 key_setsadbxpolicy(u_int16_t type, u_int8_t dir, u_int32_t id, u_int32_t priority)
4044 {
4045 	struct mbuf *m;
4046 	struct sadb_x_policy *p;
4047 	size_t len;
4048 
4049 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
4050 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4051 	if (m == NULL)
4052 		return (NULL);
4053 	m_align(m, len);
4054 	m->m_len = len;
4055 	p = mtod(m, struct sadb_x_policy *);
4056 
4057 	bzero(p, len);
4058 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
4059 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
4060 	p->sadb_x_policy_type = type;
4061 	p->sadb_x_policy_dir = dir;
4062 	p->sadb_x_policy_id = id;
4063 	p->sadb_x_policy_priority = priority;
4064 
4065 	return m;
4066 }
4067 
4068 /* %%% utilities */
4069 /* Take a key message (sadb_key) from the socket and turn it into one
4070  * of the kernel's key structures (seckey).
4071  *
4072  * IN: pointer to the src
4073  * OUT: NULL no more memory
4074  */
4075 struct seckey *
4076 key_dup_keymsg(const struct sadb_key *src, size_t len,
4077     struct malloc_type *type)
4078 {
4079 	struct seckey *dst;
4080 
4081 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4082 	if (dst != NULL) {
4083 		dst->bits = src->sadb_key_bits;
4084 		dst->key_data = malloc(len, type, M_NOWAIT);
4085 		if (dst->key_data != NULL) {
4086 			bcopy((const char *)(src + 1), dst->key_data, len);
4087 		} else {
4088 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4089 			    __func__));
4090 			free(dst, type);
4091 			dst = NULL;
4092 		}
4093 	} else {
4094 		ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4095 		    __func__));
4096 
4097 	}
4098 	return (dst);
4099 }
4100 
4101 /* Take a lifetime message (sadb_lifetime) passed in on a socket and
4102  * turn it into one of the kernel's lifetime structures (seclifetime).
4103  *
4104  * IN: pointer to the destination, source and malloc type
4105  * OUT: NULL, no more memory
4106  */
4107 
4108 static struct seclifetime *
4109 key_dup_lifemsg(const struct sadb_lifetime *src, struct malloc_type *type)
4110 {
4111 	struct seclifetime *dst;
4112 
4113 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4114 	if (dst == NULL) {
4115 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
4116 		return (NULL);
4117 	}
4118 	dst->allocations = src->sadb_lifetime_allocations;
4119 	dst->bytes = src->sadb_lifetime_bytes;
4120 	dst->addtime = src->sadb_lifetime_addtime;
4121 	dst->usetime = src->sadb_lifetime_usetime;
4122 	return (dst);
4123 }
4124 
4125 /*
4126  * compare two secasindex structure.
4127  * flag can specify to compare 2 saidxes.
4128  * compare two secasindex structure without both mode and reqid.
4129  * don't compare port.
4130  * IN:
4131  *      saidx0: source, it can be in SAD.
4132  *      saidx1: object.
4133  * OUT:
4134  *      1 : equal
4135  *      0 : not equal
4136  */
4137 static int
4138 key_cmpsaidx(const struct secasindex *saidx0, const struct secasindex *saidx1,
4139     int flag)
4140 {
4141 
4142 	/* sanity */
4143 	if (saidx0 == NULL && saidx1 == NULL)
4144 		return 1;
4145 
4146 	if (saidx0 == NULL || saidx1 == NULL)
4147 		return 0;
4148 
4149 	if (saidx0->proto != saidx1->proto)
4150 		return 0;
4151 
4152 	if (flag == CMP_EXACTLY) {
4153 		if (saidx0->mode != saidx1->mode)
4154 			return 0;
4155 		if (saidx0->reqid != saidx1->reqid)
4156 			return 0;
4157 		if (bcmp(&saidx0->src, &saidx1->src,
4158 		    saidx0->src.sa.sa_len) != 0 ||
4159 		    bcmp(&saidx0->dst, &saidx1->dst,
4160 		    saidx0->dst.sa.sa_len) != 0)
4161 			return 0;
4162 	} else {
4163 
4164 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
4165 		if (flag == CMP_MODE_REQID || flag == CMP_REQID) {
4166 			/*
4167 			 * If reqid of SPD is non-zero, unique SA is required.
4168 			 * The result must be of same reqid in this case.
4169 			 */
4170 			if (saidx1->reqid != 0 &&
4171 			    saidx0->reqid != saidx1->reqid)
4172 				return 0;
4173 		}
4174 
4175 		if (flag == CMP_MODE_REQID) {
4176 			if (saidx0->mode != IPSEC_MODE_ANY
4177 			 && saidx0->mode != saidx1->mode)
4178 				return 0;
4179 		}
4180 
4181 		if (key_sockaddrcmp(&saidx0->src.sa, &saidx1->src.sa, 0) != 0)
4182 			return 0;
4183 		if (key_sockaddrcmp(&saidx0->dst.sa, &saidx1->dst.sa, 0) != 0)
4184 			return 0;
4185 	}
4186 
4187 	return 1;
4188 }
4189 
4190 /*
4191  * compare two secindex structure exactly.
4192  * IN:
4193  *	spidx0: source, it is often in SPD.
4194  *	spidx1: object, it is often from PFKEY message.
4195  * OUT:
4196  *	1 : equal
4197  *	0 : not equal
4198  */
4199 static int
4200 key_cmpspidx_exactly(struct secpolicyindex *spidx0,
4201     struct secpolicyindex *spidx1)
4202 {
4203 	/* sanity */
4204 	if (spidx0 == NULL && spidx1 == NULL)
4205 		return 1;
4206 
4207 	if (spidx0 == NULL || spidx1 == NULL)
4208 		return 0;
4209 
4210 	if (spidx0->prefs != spidx1->prefs
4211 	 || spidx0->prefd != spidx1->prefd
4212 	 || spidx0->ul_proto != spidx1->ul_proto
4213 	 || spidx0->dir != spidx1->dir)
4214 		return 0;
4215 
4216 	return key_sockaddrcmp(&spidx0->src.sa, &spidx1->src.sa, 1) == 0 &&
4217 	       key_sockaddrcmp(&spidx0->dst.sa, &spidx1->dst.sa, 1) == 0;
4218 }
4219 
4220 /*
4221  * compare two secindex structure with mask.
4222  * IN:
4223  *	spidx0: source, it is often in SPD.
4224  *	spidx1: object, it is often from IP header.
4225  * OUT:
4226  *	1 : equal
4227  *	0 : not equal
4228  */
4229 static int
4230 key_cmpspidx_withmask(struct secpolicyindex *spidx0,
4231     struct secpolicyindex *spidx1)
4232 {
4233 	/* sanity */
4234 	if (spidx0 == NULL && spidx1 == NULL)
4235 		return 1;
4236 
4237 	if (spidx0 == NULL || spidx1 == NULL)
4238 		return 0;
4239 
4240 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4241 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4242 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4243 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
4244 		return 0;
4245 
4246 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4247 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY
4248 	 && spidx0->ul_proto != spidx1->ul_proto)
4249 		return 0;
4250 
4251 	switch (spidx0->src.sa.sa_family) {
4252 	case AF_INET:
4253 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY
4254 		 && spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
4255 			return 0;
4256 		if (!key_bbcmp(&spidx0->src.sin.sin_addr,
4257 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
4258 			return 0;
4259 		break;
4260 	case AF_INET6:
4261 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY
4262 		 && spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
4263 			return 0;
4264 		/*
4265 		 * scope_id check. if sin6_scope_id is 0, we regard it
4266 		 * as a wildcard scope, which matches any scope zone ID.
4267 		 */
4268 		if (spidx0->src.sin6.sin6_scope_id &&
4269 		    spidx1->src.sin6.sin6_scope_id &&
4270 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
4271 			return 0;
4272 		if (!key_bbcmp(&spidx0->src.sin6.sin6_addr,
4273 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
4274 			return 0;
4275 		break;
4276 	default:
4277 		/* XXX */
4278 		if (bcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
4279 			return 0;
4280 		break;
4281 	}
4282 
4283 	switch (spidx0->dst.sa.sa_family) {
4284 	case AF_INET:
4285 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY
4286 		 && spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
4287 			return 0;
4288 		if (!key_bbcmp(&spidx0->dst.sin.sin_addr,
4289 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
4290 			return 0;
4291 		break;
4292 	case AF_INET6:
4293 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY
4294 		 && spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
4295 			return 0;
4296 		/*
4297 		 * scope_id check. if sin6_scope_id is 0, we regard it
4298 		 * as a wildcard scope, which matches any scope zone ID.
4299 		 */
4300 		if (spidx0->dst.sin6.sin6_scope_id &&
4301 		    spidx1->dst.sin6.sin6_scope_id &&
4302 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
4303 			return 0;
4304 		if (!key_bbcmp(&spidx0->dst.sin6.sin6_addr,
4305 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
4306 			return 0;
4307 		break;
4308 	default:
4309 		/* XXX */
4310 		if (bcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
4311 			return 0;
4312 		break;
4313 	}
4314 
4315 	/* XXX Do we check other field ?  e.g. flowinfo */
4316 
4317 	return 1;
4318 }
4319 
4320 #ifdef satosin
4321 #undef satosin
4322 #endif
4323 #define satosin(s) ((const struct sockaddr_in *)s)
4324 #ifdef satosin6
4325 #undef satosin6
4326 #endif
4327 #define satosin6(s) ((const struct sockaddr_in6 *)s)
4328 /* returns 0 on match */
4329 int
4330 key_sockaddrcmp(const struct sockaddr *sa1, const struct sockaddr *sa2,
4331     int port)
4332 {
4333 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4334 		return 1;
4335 
4336 	switch (sa1->sa_family) {
4337 #ifdef INET
4338 	case AF_INET:
4339 		if (sa1->sa_len != sizeof(struct sockaddr_in))
4340 			return 1;
4341 		if (satosin(sa1)->sin_addr.s_addr !=
4342 		    satosin(sa2)->sin_addr.s_addr) {
4343 			return 1;
4344 		}
4345 		if (port && satosin(sa1)->sin_port != satosin(sa2)->sin_port)
4346 			return 1;
4347 		break;
4348 #endif
4349 #ifdef INET6
4350 	case AF_INET6:
4351 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4352 			return 1;	/*EINVAL*/
4353 		if (satosin6(sa1)->sin6_scope_id !=
4354 		    satosin6(sa2)->sin6_scope_id) {
4355 			return 1;
4356 		}
4357 		if (!IN6_ARE_ADDR_EQUAL(&satosin6(sa1)->sin6_addr,
4358 		    &satosin6(sa2)->sin6_addr)) {
4359 			return 1;
4360 		}
4361 		if (port &&
4362 		    satosin6(sa1)->sin6_port != satosin6(sa2)->sin6_port) {
4363 			return 1;
4364 		}
4365 		break;
4366 #endif
4367 	default:
4368 		if (bcmp(sa1, sa2, sa1->sa_len) != 0)
4369 			return 1;
4370 		break;
4371 	}
4372 
4373 	return 0;
4374 }
4375 
4376 /* returns 0 on match */
4377 int
4378 key_sockaddrcmp_withmask(const struct sockaddr *sa1,
4379     const struct sockaddr *sa2, size_t mask)
4380 {
4381 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4382 		return (1);
4383 
4384 	switch (sa1->sa_family) {
4385 #ifdef INET
4386 	case AF_INET:
4387 		return (!key_bbcmp(&satosin(sa1)->sin_addr,
4388 		    &satosin(sa2)->sin_addr, mask));
4389 #endif
4390 #ifdef INET6
4391 	case AF_INET6:
4392 		if (satosin6(sa1)->sin6_scope_id !=
4393 		    satosin6(sa2)->sin6_scope_id)
4394 			return (1);
4395 		return (!key_bbcmp(&satosin6(sa1)->sin6_addr,
4396 		    &satosin6(sa2)->sin6_addr, mask));
4397 #endif
4398 	}
4399 	return (1);
4400 }
4401 #undef satosin
4402 #undef satosin6
4403 
4404 /*
4405  * compare two buffers with mask.
4406  * IN:
4407  *	addr1: source
4408  *	addr2: object
4409  *	bits:  Number of bits to compare
4410  * OUT:
4411  *	1 : equal
4412  *	0 : not equal
4413  */
4414 static int
4415 key_bbcmp(const void *a1, const void *a2, u_int bits)
4416 {
4417 	const unsigned char *p1 = a1;
4418 	const unsigned char *p2 = a2;
4419 
4420 	/* XXX: This could be considerably faster if we compare a word
4421 	 * at a time, but it is complicated on LSB Endian machines */
4422 
4423 	/* Handle null pointers */
4424 	if (p1 == NULL || p2 == NULL)
4425 		return (p1 == p2);
4426 
4427 	while (bits >= 8) {
4428 		if (*p1++ != *p2++)
4429 			return 0;
4430 		bits -= 8;
4431 	}
4432 
4433 	if (bits > 0) {
4434 		u_int8_t mask = ~((1<<(8-bits))-1);
4435 		if ((*p1 & mask) != (*p2 & mask))
4436 			return 0;
4437 	}
4438 	return 1;	/* Match! */
4439 }
4440 
4441 static void
4442 key_flush_spd(time_t now)
4443 {
4444 	SPTREE_RLOCK_TRACKER;
4445 	struct secpolicy_list drainq;
4446 	struct secpolicy *sp, *nextsp;
4447 	u_int dir;
4448 
4449 	LIST_INIT(&drainq);
4450 	SPTREE_RLOCK();
4451 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4452 		TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
4453 			if (sp->lifetime == 0 && sp->validtime == 0)
4454 				continue;
4455 			if ((sp->lifetime &&
4456 			    now - sp->created > sp->lifetime) ||
4457 			    (sp->validtime &&
4458 			    now - sp->lastused > sp->validtime)) {
4459 				/* Hold extra reference to send SPDEXPIRE */
4460 				SP_ADDREF(sp);
4461 				LIST_INSERT_HEAD(&drainq, sp, drainq);
4462 			}
4463 		}
4464 	}
4465 	SPTREE_RUNLOCK();
4466 	if (LIST_EMPTY(&drainq))
4467 		return;
4468 
4469 	SPTREE_WLOCK();
4470 	sp = LIST_FIRST(&drainq);
4471 	while (sp != NULL) {
4472 		nextsp = LIST_NEXT(sp, drainq);
4473 		/* Check that SP is still linked */
4474 		if (sp->state != IPSEC_SPSTATE_ALIVE) {
4475 			LIST_REMOVE(sp, drainq);
4476 			key_freesp(&sp); /* release extra reference */
4477 			sp = nextsp;
4478 			continue;
4479 		}
4480 		TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
4481 		V_spd_size--;
4482 		LIST_REMOVE(sp, idhash);
4483 		sp->state = IPSEC_SPSTATE_DEAD;
4484 		sp = nextsp;
4485 	}
4486 	V_sp_genid++;
4487 	SPTREE_WUNLOCK();
4488 	if (SPDCACHE_ENABLED())
4489 		spdcache_clear();
4490 
4491 	sp = LIST_FIRST(&drainq);
4492 	while (sp != NULL) {
4493 		nextsp = LIST_NEXT(sp, drainq);
4494 		key_spdexpire(sp);
4495 		key_freesp(&sp); /* release extra reference */
4496 		key_freesp(&sp); /* release last reference */
4497 		sp = nextsp;
4498 	}
4499 }
4500 
4501 static void
4502 key_flush_sad(time_t now)
4503 {
4504 	SAHTREE_RLOCK_TRACKER;
4505 	struct secashead_list emptyq;
4506 	struct secasvar_list drainq, hexpireq, sexpireq, freeq;
4507 	struct secashead *sah, *nextsah;
4508 	struct secasvar *sav, *nextsav;
4509 
4510 	LIST_INIT(&drainq);
4511 	LIST_INIT(&hexpireq);
4512 	LIST_INIT(&sexpireq);
4513 	LIST_INIT(&emptyq);
4514 
4515 	SAHTREE_RLOCK();
4516 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
4517 		/* Check for empty SAH */
4518 		if (TAILQ_EMPTY(&sah->savtree_larval) &&
4519 		    TAILQ_EMPTY(&sah->savtree_alive)) {
4520 			SAH_ADDREF(sah);
4521 			LIST_INSERT_HEAD(&emptyq, sah, drainq);
4522 			continue;
4523 		}
4524 		/* Add all stale LARVAL SAs into drainq */
4525 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
4526 			if (now - sav->created < V_key_larval_lifetime)
4527 				continue;
4528 			SAV_ADDREF(sav);
4529 			LIST_INSERT_HEAD(&drainq, sav, drainq);
4530 		}
4531 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
4532 			/* lifetimes aren't specified */
4533 			if (sav->lft_h == NULL)
4534 				continue;
4535 			SECASVAR_LOCK(sav);
4536 			/*
4537 			 * Check again with lock held, because it may
4538 			 * be updated by SADB_UPDATE.
4539 			 */
4540 			if (sav->lft_h == NULL) {
4541 				SECASVAR_UNLOCK(sav);
4542 				continue;
4543 			}
4544 			/*
4545 			 * RFC 2367:
4546 			 * HARD lifetimes MUST take precedence over SOFT
4547 			 * lifetimes, meaning if the HARD and SOFT lifetimes
4548 			 * are the same, the HARD lifetime will appear on the
4549 			 * EXPIRE message.
4550 			 */
4551 			/* check HARD lifetime */
4552 			if ((sav->lft_h->addtime != 0 &&
4553 			    now - sav->created > sav->lft_h->addtime) ||
4554 			    (sav->lft_h->usetime != 0 && sav->firstused &&
4555 			    now - sav->firstused > sav->lft_h->usetime) ||
4556 			    (sav->lft_h->bytes != 0 && counter_u64_fetch(
4557 			        sav->lft_c_bytes) > sav->lft_h->bytes)) {
4558 				SECASVAR_UNLOCK(sav);
4559 				SAV_ADDREF(sav);
4560 				LIST_INSERT_HEAD(&hexpireq, sav, drainq);
4561 				continue;
4562 			}
4563 			/* check SOFT lifetime (only for MATURE SAs) */
4564 			if (sav->state == SADB_SASTATE_MATURE && (
4565 			    (sav->lft_s->addtime != 0 &&
4566 			    now - sav->created > sav->lft_s->addtime) ||
4567 			    (sav->lft_s->usetime != 0 && sav->firstused &&
4568 			    now - sav->firstused > sav->lft_s->usetime) ||
4569 			    (sav->lft_s->bytes != 0 && counter_u64_fetch(
4570 				sav->lft_c_bytes) > sav->lft_s->bytes))) {
4571 				SECASVAR_UNLOCK(sav);
4572 				SAV_ADDREF(sav);
4573 				LIST_INSERT_HEAD(&sexpireq, sav, drainq);
4574 				continue;
4575 			}
4576 			SECASVAR_UNLOCK(sav);
4577 		}
4578 	}
4579 	SAHTREE_RUNLOCK();
4580 
4581 	if (LIST_EMPTY(&emptyq) && LIST_EMPTY(&drainq) &&
4582 	    LIST_EMPTY(&hexpireq) && LIST_EMPTY(&sexpireq))
4583 		return;
4584 
4585 	LIST_INIT(&freeq);
4586 	SAHTREE_WLOCK();
4587 	/* Unlink stale LARVAL SAs */
4588 	sav = LIST_FIRST(&drainq);
4589 	while (sav != NULL) {
4590 		nextsav = LIST_NEXT(sav, drainq);
4591 		/* Check that SA is still LARVAL */
4592 		if (sav->state != SADB_SASTATE_LARVAL) {
4593 			LIST_REMOVE(sav, drainq);
4594 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4595 			sav = nextsav;
4596 			continue;
4597 		}
4598 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
4599 		LIST_REMOVE(sav, spihash);
4600 		sav->state = SADB_SASTATE_DEAD;
4601 		sav = nextsav;
4602 	}
4603 	/* Unlink all SAs with expired HARD lifetime */
4604 	sav = LIST_FIRST(&hexpireq);
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 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
4615 		LIST_REMOVE(sav, spihash);
4616 		sav->state = SADB_SASTATE_DEAD;
4617 		sav = nextsav;
4618 	}
4619 	/* Mark all SAs with expired SOFT lifetime as DYING */
4620 	sav = LIST_FIRST(&sexpireq);
4621 	while (sav != NULL) {
4622 		nextsav = LIST_NEXT(sav, drainq);
4623 		/* Check that SA is not unlinked */
4624 		if (sav->state == SADB_SASTATE_DEAD) {
4625 			LIST_REMOVE(sav, drainq);
4626 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4627 			sav = nextsav;
4628 			continue;
4629 		}
4630 		/*
4631 		 * NOTE: this doesn't change SA order in the chain.
4632 		 */
4633 		sav->state = SADB_SASTATE_DYING;
4634 		sav = nextsav;
4635 	}
4636 	/* Unlink empty SAHs */
4637 	sah = LIST_FIRST(&emptyq);
4638 	while (sah != NULL) {
4639 		nextsah = LIST_NEXT(sah, drainq);
4640 		/* Check that SAH is still empty and not unlinked */
4641 		if (sah->state == SADB_SASTATE_DEAD ||
4642 		    !TAILQ_EMPTY(&sah->savtree_larval) ||
4643 		    !TAILQ_EMPTY(&sah->savtree_alive)) {
4644 			LIST_REMOVE(sah, drainq);
4645 			key_freesah(&sah); /* release extra reference */
4646 			sah = nextsah;
4647 			continue;
4648 		}
4649 		TAILQ_REMOVE(&V_sahtree, sah, chain);
4650 		LIST_REMOVE(sah, addrhash);
4651 		sah->state = SADB_SASTATE_DEAD;
4652 		sah = nextsah;
4653 	}
4654 	SAHTREE_WUNLOCK();
4655 
4656 	/* Send SPDEXPIRE messages */
4657 	sav = LIST_FIRST(&hexpireq);
4658 	while (sav != NULL) {
4659 		nextsav = LIST_NEXT(sav, drainq);
4660 		key_expire(sav, 1);
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 	sav = LIST_FIRST(&sexpireq);
4667 	while (sav != NULL) {
4668 		nextsav = LIST_NEXT(sav, drainq);
4669 		key_expire(sav, 0);
4670 		key_freesav(&sav); /* release extra reference */
4671 		sav = nextsav;
4672 	}
4673 	/* Free stale LARVAL SAs */
4674 	sav = LIST_FIRST(&drainq);
4675 	while (sav != NULL) {
4676 		nextsav = LIST_NEXT(sav, drainq);
4677 		key_freesah(&sav->sah); /* release reference from SAV */
4678 		key_freesav(&sav); /* release extra reference */
4679 		key_freesav(&sav); /* release last reference */
4680 		sav = nextsav;
4681 	}
4682 	/* Free SAs that were unlinked/changed by someone else */
4683 	sav = LIST_FIRST(&freeq);
4684 	while (sav != NULL) {
4685 		nextsav = LIST_NEXT(sav, drainq);
4686 		key_freesav(&sav); /* release extra reference */
4687 		sav = nextsav;
4688 	}
4689 	/* Free empty SAH */
4690 	sah = LIST_FIRST(&emptyq);
4691 	while (sah != NULL) {
4692 		nextsah = LIST_NEXT(sah, drainq);
4693 		key_freesah(&sah); /* release extra reference */
4694 		key_freesah(&sah); /* release last reference */
4695 		sah = nextsah;
4696 	}
4697 }
4698 
4699 static void
4700 key_flush_acq(time_t now)
4701 {
4702 	struct secacq *acq, *nextacq;
4703 
4704 	/* ACQ tree */
4705 	ACQ_LOCK();
4706 	acq = LIST_FIRST(&V_acqtree);
4707 	while (acq != NULL) {
4708 		nextacq = LIST_NEXT(acq, chain);
4709 		if (now - acq->created > V_key_blockacq_lifetime) {
4710 			LIST_REMOVE(acq, chain);
4711 			LIST_REMOVE(acq, addrhash);
4712 			LIST_REMOVE(acq, seqhash);
4713 			free(acq, M_IPSEC_SAQ);
4714 		}
4715 		acq = nextacq;
4716 	}
4717 	ACQ_UNLOCK();
4718 }
4719 
4720 static void
4721 key_flush_spacq(time_t now)
4722 {
4723 	struct secspacq *acq, *nextacq;
4724 
4725 	/* SP ACQ tree */
4726 	SPACQ_LOCK();
4727 	for (acq = LIST_FIRST(&V_spacqtree); acq != NULL; acq = nextacq) {
4728 		nextacq = LIST_NEXT(acq, chain);
4729 		if (now - acq->created > V_key_blockacq_lifetime
4730 		 && __LIST_CHAINED(acq)) {
4731 			LIST_REMOVE(acq, chain);
4732 			free(acq, M_IPSEC_SAQ);
4733 		}
4734 	}
4735 	SPACQ_UNLOCK();
4736 }
4737 
4738 /*
4739  * time handler.
4740  * scanning SPD and SAD to check status for each entries,
4741  * and do to remove or to expire.
4742  * XXX: year 2038 problem may remain.
4743  */
4744 static void
4745 key_timehandler(void *arg)
4746 {
4747 	VNET_ITERATOR_DECL(vnet_iter);
4748 	time_t now = time_second;
4749 
4750 	VNET_LIST_RLOCK_NOSLEEP();
4751 	VNET_FOREACH(vnet_iter) {
4752 		CURVNET_SET(vnet_iter);
4753 		key_flush_spd(now);
4754 		key_flush_sad(now);
4755 		key_flush_acq(now);
4756 		key_flush_spacq(now);
4757 		CURVNET_RESTORE();
4758 	}
4759 	VNET_LIST_RUNLOCK_NOSLEEP();
4760 
4761 #ifndef IPSEC_DEBUG2
4762 	/* do exchange to tick time !! */
4763 	callout_schedule(&key_timer, hz);
4764 #endif /* IPSEC_DEBUG2 */
4765 }
4766 
4767 u_long
4768 key_random()
4769 {
4770 	u_long value;
4771 
4772 	key_randomfill(&value, sizeof(value));
4773 	return value;
4774 }
4775 
4776 void
4777 key_randomfill(void *p, size_t l)
4778 {
4779 	size_t n;
4780 	u_long v;
4781 	static int warn = 1;
4782 
4783 	n = 0;
4784 	n = (size_t)read_random(p, (u_int)l);
4785 	/* last resort */
4786 	while (n < l) {
4787 		v = random();
4788 		bcopy(&v, (u_int8_t *)p + n,
4789 		    l - n < sizeof(v) ? l - n : sizeof(v));
4790 		n += sizeof(v);
4791 
4792 		if (warn) {
4793 			printf("WARNING: pseudo-random number generator "
4794 			    "used for IPsec processing\n");
4795 			warn = 0;
4796 		}
4797 	}
4798 }
4799 
4800 /*
4801  * map SADB_SATYPE_* to IPPROTO_*.
4802  * if satype == SADB_SATYPE then satype is mapped to ~0.
4803  * OUT:
4804  *	0: invalid satype.
4805  */
4806 static uint8_t
4807 key_satype2proto(uint8_t satype)
4808 {
4809 	switch (satype) {
4810 	case SADB_SATYPE_UNSPEC:
4811 		return IPSEC_PROTO_ANY;
4812 	case SADB_SATYPE_AH:
4813 		return IPPROTO_AH;
4814 	case SADB_SATYPE_ESP:
4815 		return IPPROTO_ESP;
4816 	case SADB_X_SATYPE_IPCOMP:
4817 		return IPPROTO_IPCOMP;
4818 	case SADB_X_SATYPE_TCPSIGNATURE:
4819 		return IPPROTO_TCP;
4820 	default:
4821 		return 0;
4822 	}
4823 	/* NOTREACHED */
4824 }
4825 
4826 /*
4827  * map IPPROTO_* to SADB_SATYPE_*
4828  * OUT:
4829  *	0: invalid protocol type.
4830  */
4831 static uint8_t
4832 key_proto2satype(uint8_t proto)
4833 {
4834 	switch (proto) {
4835 	case IPPROTO_AH:
4836 		return SADB_SATYPE_AH;
4837 	case IPPROTO_ESP:
4838 		return SADB_SATYPE_ESP;
4839 	case IPPROTO_IPCOMP:
4840 		return SADB_X_SATYPE_IPCOMP;
4841 	case IPPROTO_TCP:
4842 		return SADB_X_SATYPE_TCPSIGNATURE;
4843 	default:
4844 		return 0;
4845 	}
4846 	/* NOTREACHED */
4847 }
4848 
4849 /* %%% PF_KEY */
4850 /*
4851  * SADB_GETSPI processing is to receive
4852  *	<base, (SA2), src address, dst address, (SPI range)>
4853  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
4854  * tree with the status of LARVAL, and send
4855  *	<base, SA(*), address(SD)>
4856  * to the IKMPd.
4857  *
4858  * IN:	mhp: pointer to the pointer to each header.
4859  * OUT:	NULL if fail.
4860  *	other if success, return pointer to the message to send.
4861  */
4862 static int
4863 key_getspi(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
4864 {
4865 	struct secasindex saidx;
4866 	struct sadb_address *src0, *dst0;
4867 	struct secasvar *sav;
4868 	uint32_t reqid, spi;
4869 	int error;
4870 	uint8_t mode, proto;
4871 
4872 	IPSEC_ASSERT(so != NULL, ("null socket"));
4873 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
4874 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
4875 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
4876 
4877 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
4878 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)
4879 #ifdef PFKEY_STRICT_CHECKS
4880 	    || SADB_CHECKHDR(mhp, SADB_EXT_SPIRANGE)
4881 #endif
4882 	    ) {
4883 		ipseclog((LOG_DEBUG,
4884 		    "%s: invalid message: missing required header.\n",
4885 		    __func__));
4886 		error = EINVAL;
4887 		goto fail;
4888 	}
4889 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
4890 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)
4891 #ifdef PFKEY_STRICT_CHECKS
4892 	    || SADB_CHECKLEN(mhp, SADB_EXT_SPIRANGE)
4893 #endif
4894 	    ) {
4895 		ipseclog((LOG_DEBUG,
4896 		    "%s: invalid message: wrong header size.\n", __func__));
4897 		error = EINVAL;
4898 		goto fail;
4899 	}
4900 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
4901 		mode = IPSEC_MODE_ANY;
4902 		reqid = 0;
4903 	} else {
4904 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
4905 			ipseclog((LOG_DEBUG,
4906 			    "%s: invalid message: wrong header size.\n",
4907 			    __func__));
4908 			error = EINVAL;
4909 			goto fail;
4910 		}
4911 		mode = ((struct sadb_x_sa2 *)
4912 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
4913 		reqid = ((struct sadb_x_sa2 *)
4914 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
4915 	}
4916 
4917 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
4918 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
4919 
4920 	/* map satype to proto */
4921 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
4922 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
4923 			__func__));
4924 		error = EINVAL;
4925 		goto fail;
4926 	}
4927 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
4928 	    (struct sockaddr *)(dst0 + 1));
4929 	if (error != 0) {
4930 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
4931 		error = EINVAL;
4932 		goto fail;
4933 	}
4934 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
4935 
4936 	/* SPI allocation */
4937 	spi = key_do_getnewspi(
4938 	    (struct sadb_spirange *)mhp->ext[SADB_EXT_SPIRANGE], &saidx);
4939 	if (spi == 0) {
4940 		/*
4941 		 * Requested SPI or SPI range is not available or
4942 		 * already used.
4943 		 */
4944 		error = EEXIST;
4945 		goto fail;
4946 	}
4947 	sav = key_newsav(mhp, &saidx, spi, &error);
4948 	if (sav == NULL)
4949 		goto fail;
4950 
4951 	if (sav->seq != 0) {
4952 		/*
4953 		 * RFC2367:
4954 		 * If the SADB_GETSPI message is in response to a
4955 		 * kernel-generated SADB_ACQUIRE, the sadb_msg_seq
4956 		 * MUST be the same as the SADB_ACQUIRE message.
4957 		 *
4958 		 * XXXAE: However it doesn't definethe behaviour how to
4959 		 * check this and what to do if it doesn't match.
4960 		 * Also what we should do if it matches?
4961 		 *
4962 		 * We can compare saidx used in SADB_ACQUIRE with saidx
4963 		 * used in SADB_GETSPI, but this probably can break
4964 		 * existing software. For now just warn if it doesn't match.
4965 		 *
4966 		 * XXXAE: anyway it looks useless.
4967 		 */
4968 		key_acqdone(&saidx, sav->seq);
4969 	}
4970 	KEYDBG(KEY_STAMP,
4971 	    printf("%s: SA(%p)\n", __func__, sav));
4972 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
4973 
4974     {
4975 	struct mbuf *n, *nn;
4976 	struct sadb_sa *m_sa;
4977 	struct sadb_msg *newmsg;
4978 	int off, len;
4979 
4980 	/* create new sadb_msg to reply. */
4981 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
4982 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
4983 
4984 	MGETHDR(n, M_NOWAIT, MT_DATA);
4985 	if (len > MHLEN) {
4986 		if (!(MCLGET(n, M_NOWAIT))) {
4987 			m_freem(n);
4988 			n = NULL;
4989 		}
4990 	}
4991 	if (!n) {
4992 		error = ENOBUFS;
4993 		goto fail;
4994 	}
4995 
4996 	n->m_len = len;
4997 	n->m_next = NULL;
4998 	off = 0;
4999 
5000 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
5001 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
5002 
5003 	m_sa = (struct sadb_sa *)(mtod(n, caddr_t) + off);
5004 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
5005 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
5006 	m_sa->sadb_sa_spi = spi; /* SPI is already in network byte order */
5007 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
5008 
5009 	IPSEC_ASSERT(off == len,
5010 		("length inconsistency (off %u len %u)", off, len));
5011 
5012 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
5013 	    SADB_EXT_ADDRESS_DST);
5014 	if (!n->m_next) {
5015 		m_freem(n);
5016 		error = ENOBUFS;
5017 		goto fail;
5018 	}
5019 
5020 	if (n->m_len < sizeof(struct sadb_msg)) {
5021 		n = m_pullup(n, sizeof(struct sadb_msg));
5022 		if (n == NULL)
5023 			return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
5024 	}
5025 
5026 	n->m_pkthdr.len = 0;
5027 	for (nn = n; nn; nn = nn->m_next)
5028 		n->m_pkthdr.len += nn->m_len;
5029 
5030 	newmsg = mtod(n, struct sadb_msg *);
5031 	newmsg->sadb_msg_seq = sav->seq;
5032 	newmsg->sadb_msg_errno = 0;
5033 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
5034 
5035 	m_freem(m);
5036 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
5037     }
5038 
5039 fail:
5040 	return (key_senderror(so, m, error));
5041 }
5042 
5043 /*
5044  * allocating new SPI
5045  * called by key_getspi().
5046  * OUT:
5047  *	0:	failure.
5048  *	others: success, SPI in network byte order.
5049  */
5050 static uint32_t
5051 key_do_getnewspi(struct sadb_spirange *spirange, struct secasindex *saidx)
5052 {
5053 	uint32_t min, max, newspi, t;
5054 	int count = V_key_spi_trycnt;
5055 
5056 	/* set spi range to allocate */
5057 	if (spirange != NULL) {
5058 		min = spirange->sadb_spirange_min;
5059 		max = spirange->sadb_spirange_max;
5060 	} else {
5061 		min = V_key_spi_minval;
5062 		max = V_key_spi_maxval;
5063 	}
5064 	/* IPCOMP needs 2-byte SPI */
5065 	if (saidx->proto == IPPROTO_IPCOMP) {
5066 		if (min >= 0x10000)
5067 			min = 0xffff;
5068 		if (max >= 0x10000)
5069 			max = 0xffff;
5070 		if (min > max) {
5071 			t = min; min = max; max = t;
5072 		}
5073 	}
5074 
5075 	if (min == max) {
5076 		if (!key_checkspidup(htonl(min))) {
5077 			ipseclog((LOG_DEBUG, "%s: SPI %u exists already.\n",
5078 			    __func__, min));
5079 			return 0;
5080 		}
5081 
5082 		count--; /* taking one cost. */
5083 		newspi = min;
5084 	} else {
5085 
5086 		/* init SPI */
5087 		newspi = 0;
5088 
5089 		/* when requesting to allocate spi ranged */
5090 		while (count--) {
5091 			/* generate pseudo-random SPI value ranged. */
5092 			newspi = min + (key_random() % (max - min + 1));
5093 			if (!key_checkspidup(htonl(newspi)))
5094 				break;
5095 		}
5096 
5097 		if (count == 0 || newspi == 0) {
5098 			ipseclog((LOG_DEBUG,
5099 			    "%s: failed to allocate SPI.\n", __func__));
5100 			return 0;
5101 		}
5102 	}
5103 
5104 	/* statistics */
5105 	keystat.getspi_count =
5106 	    (keystat.getspi_count + V_key_spi_trycnt - count) / 2;
5107 
5108 	return (htonl(newspi));
5109 }
5110 
5111 /*
5112  * Find TCP-MD5 SA with corresponding secasindex.
5113  * If not found, return NULL and fill SPI with usable value if needed.
5114  */
5115 static struct secasvar *
5116 key_getsav_tcpmd5(struct secasindex *saidx, uint32_t *spi)
5117 {
5118 	SAHTREE_RLOCK_TRACKER;
5119 	struct secashead *sah;
5120 	struct secasvar *sav;
5121 
5122 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP, ("wrong proto"));
5123 	SAHTREE_RLOCK();
5124 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
5125 		if (sah->saidx.proto != IPPROTO_TCP)
5126 			continue;
5127 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
5128 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
5129 			break;
5130 	}
5131 	if (sah != NULL) {
5132 		if (V_key_preferred_oldsa)
5133 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
5134 		else
5135 			sav = TAILQ_FIRST(&sah->savtree_alive);
5136 		if (sav != NULL) {
5137 			SAV_ADDREF(sav);
5138 			SAHTREE_RUNLOCK();
5139 			return (sav);
5140 		}
5141 	}
5142 	if (spi == NULL) {
5143 		/* No SPI required */
5144 		SAHTREE_RUNLOCK();
5145 		return (NULL);
5146 	}
5147 	/* Check that SPI is unique */
5148 	LIST_FOREACH(sav, SAVHASH_HASH(*spi), spihash) {
5149 		if (sav->spi == *spi)
5150 			break;
5151 	}
5152 	if (sav == NULL) {
5153 		SAHTREE_RUNLOCK();
5154 		/* SPI is already unique */
5155 		return (NULL);
5156 	}
5157 	SAHTREE_RUNLOCK();
5158 	/* XXX: not optimal */
5159 	*spi = key_do_getnewspi(NULL, saidx);
5160 	return (NULL);
5161 }
5162 
5163 static int
5164 key_updateaddresses(struct socket *so, struct mbuf *m,
5165     const struct sadb_msghdr *mhp, struct secasvar *sav,
5166     struct secasindex *saidx)
5167 {
5168 	struct sockaddr *newaddr;
5169 	struct secashead *sah;
5170 	struct secasvar *newsav, *tmp;
5171 	struct mbuf *n;
5172 	int error, isnew;
5173 
5174 	/* Check that we need to change SAH */
5175 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC)) {
5176 		newaddr = (struct sockaddr *)(
5177 		    ((struct sadb_address *)
5178 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_SRC]) + 1);
5179 		bcopy(newaddr, &saidx->src, newaddr->sa_len);
5180 		key_porttosaddr(&saidx->src.sa, 0);
5181 	}
5182 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5183 		newaddr = (struct sockaddr *)(
5184 		    ((struct sadb_address *)
5185 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_DST]) + 1);
5186 		bcopy(newaddr, &saidx->dst, newaddr->sa_len);
5187 		key_porttosaddr(&saidx->dst.sa, 0);
5188 	}
5189 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5190 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5191 		error = key_checksockaddrs(&saidx->src.sa, &saidx->dst.sa);
5192 		if (error != 0) {
5193 			ipseclog((LOG_DEBUG, "%s: invalid new sockaddr.\n",
5194 			    __func__));
5195 			return (error);
5196 		}
5197 
5198 		sah = key_getsah(saidx);
5199 		if (sah == NULL) {
5200 			/* create a new SA index */
5201 			sah = key_newsah(saidx);
5202 			if (sah == NULL) {
5203 				ipseclog((LOG_DEBUG,
5204 				    "%s: No more memory.\n", __func__));
5205 				return (ENOBUFS);
5206 			}
5207 			isnew = 2; /* SAH is new */
5208 		} else
5209 			isnew = 1; /* existing SAH is referenced */
5210 	} else {
5211 		/*
5212 		 * src and dst addresses are still the same.
5213 		 * Do we want to change NAT-T config?
5214 		 */
5215 		if (sav->sah->saidx.proto != IPPROTO_ESP ||
5216 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5217 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5218 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5219 			ipseclog((LOG_DEBUG,
5220 			    "%s: invalid message: missing required header.\n",
5221 			    __func__));
5222 			return (EINVAL);
5223 		}
5224 		/* We hold reference to SA, thus SAH will be referenced too. */
5225 		sah = sav->sah;
5226 		isnew = 0;
5227 	}
5228 
5229 	newsav = malloc(sizeof(struct secasvar), M_IPSEC_SA,
5230 	    M_NOWAIT | M_ZERO);
5231 	if (newsav == NULL) {
5232 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5233 		error = ENOBUFS;
5234 		goto fail;
5235 	}
5236 
5237 	/* Clone SA's content into newsav */
5238 	SAV_INITREF(newsav);
5239 	bcopy(sav, newsav, offsetof(struct secasvar, chain));
5240 	/*
5241 	 * We create new NAT-T config if it is needed.
5242 	 * Old NAT-T config will be freed by key_cleansav() when
5243 	 * last reference to SA will be released.
5244 	 */
5245 	newsav->natt = NULL;
5246 	newsav->sah = sah;
5247 	newsav->state = SADB_SASTATE_MATURE;
5248 	error = key_setnatt(newsav, mhp);
5249 	if (error != 0)
5250 		goto fail;
5251 
5252 	SAHTREE_WLOCK();
5253 	/* Check that SA is still alive */
5254 	if (sav->state == SADB_SASTATE_DEAD) {
5255 		/* SA was unlinked */
5256 		SAHTREE_WUNLOCK();
5257 		error = ESRCH;
5258 		goto fail;
5259 	}
5260 
5261 	/* Unlink SA from SAH and SPI hash */
5262 	IPSEC_ASSERT((sav->flags & SADB_X_EXT_F_CLONED) == 0,
5263 	    ("SA is already cloned"));
5264 	IPSEC_ASSERT(sav->state == SADB_SASTATE_MATURE ||
5265 	    sav->state == SADB_SASTATE_DYING,
5266 	    ("Wrong SA state %u\n", sav->state));
5267 	TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
5268 	LIST_REMOVE(sav, spihash);
5269 	sav->state = SADB_SASTATE_DEAD;
5270 
5271 	/*
5272 	 * Link new SA with SAH. Keep SAs ordered by
5273 	 * create time (newer are first).
5274 	 */
5275 	TAILQ_FOREACH(tmp, &sah->savtree_alive, chain) {
5276 		if (newsav->created > tmp->created) {
5277 			TAILQ_INSERT_BEFORE(tmp, newsav, chain);
5278 			break;
5279 		}
5280 	}
5281 	if (tmp == NULL)
5282 		TAILQ_INSERT_TAIL(&sah->savtree_alive, newsav, chain);
5283 
5284 	/* Add new SA into SPI hash. */
5285 	LIST_INSERT_HEAD(SAVHASH_HASH(newsav->spi), newsav, spihash);
5286 
5287 	/* Add new SAH into SADB. */
5288 	if (isnew == 2) {
5289 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
5290 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
5291 		sah->state = SADB_SASTATE_MATURE;
5292 		SAH_ADDREF(sah); /* newsav references new SAH */
5293 	}
5294 	/*
5295 	 * isnew == 1 -> @sah was referenced by key_getsah().
5296 	 * isnew == 0 -> we use the same @sah, that was used by @sav,
5297 	 *	and we use its reference for @newsav.
5298 	 */
5299 	SECASVAR_LOCK(sav);
5300 	/* XXX: replace cntr with pointer? */
5301 	newsav->cntr = sav->cntr;
5302 	sav->flags |= SADB_X_EXT_F_CLONED;
5303 	SECASVAR_UNLOCK(sav);
5304 
5305 	SAHTREE_WUNLOCK();
5306 
5307 	KEYDBG(KEY_STAMP,
5308 	    printf("%s: SA(%p) cloned into SA(%p)\n",
5309 	    __func__, sav, newsav));
5310 	KEYDBG(KEY_DATA, kdebug_secasv(newsav));
5311 
5312 	key_freesav(&sav); /* release last reference */
5313 
5314 	/* set msg buf from mhp */
5315 	n = key_getmsgbuf_x1(m, mhp);
5316 	if (n == NULL) {
5317 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5318 		return (ENOBUFS);
5319 	}
5320 	m_freem(m);
5321 	key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5322 	return (0);
5323 fail:
5324 	if (isnew != 0)
5325 		key_freesah(&sah);
5326 	if (newsav != NULL) {
5327 		if (newsav->natt != NULL)
5328 			free(newsav->natt, M_IPSEC_MISC);
5329 		free(newsav, M_IPSEC_SA);
5330 	}
5331 	return (error);
5332 }
5333 
5334 /*
5335  * SADB_UPDATE processing
5336  * receive
5337  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5338  *       key(AE), (identity(SD),) (sensitivity)>
5339  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5340  * and send
5341  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5342  *       (identity(SD),) (sensitivity)>
5343  * to the ikmpd.
5344  *
5345  * m will always be freed.
5346  */
5347 static int
5348 key_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5349 {
5350 	struct secasindex saidx;
5351 	struct sadb_address *src0, *dst0;
5352 	struct sadb_sa *sa0;
5353 	struct secasvar *sav;
5354 	uint32_t reqid;
5355 	int error;
5356 	uint8_t mode, proto;
5357 
5358 	IPSEC_ASSERT(so != NULL, ("null socket"));
5359 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5360 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5361 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5362 
5363 	/* map satype to proto */
5364 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5365 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5366 		    __func__));
5367 		return key_senderror(so, m, EINVAL);
5368 	}
5369 
5370 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5371 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5372 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5373 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5374 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5375 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5376 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5377 		ipseclog((LOG_DEBUG,
5378 		    "%s: invalid message: missing required header.\n",
5379 		    __func__));
5380 		return key_senderror(so, m, EINVAL);
5381 	}
5382 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5383 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5384 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5385 		ipseclog((LOG_DEBUG,
5386 		    "%s: invalid message: wrong header size.\n", __func__));
5387 		return key_senderror(so, m, EINVAL);
5388 	}
5389 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5390 		mode = IPSEC_MODE_ANY;
5391 		reqid = 0;
5392 	} else {
5393 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5394 			ipseclog((LOG_DEBUG,
5395 			    "%s: invalid message: wrong header size.\n",
5396 			    __func__));
5397 			return key_senderror(so, m, EINVAL);
5398 		}
5399 		mode = ((struct sadb_x_sa2 *)
5400 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5401 		reqid = ((struct sadb_x_sa2 *)
5402 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5403 	}
5404 
5405 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5406 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5407 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5408 
5409 	/*
5410 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5411 	 * SADB_UPDATE message.
5412 	 */
5413 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5414 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5415 #ifdef PFKEY_STRICT_CHECKS
5416 		return key_senderror(so, m, EINVAL);
5417 #endif
5418 	}
5419 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5420 	    (struct sockaddr *)(dst0 + 1));
5421 	if (error != 0) {
5422 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5423 		return key_senderror(so, m, error);
5424 	}
5425 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5426 	sav = key_getsavbyspi(sa0->sadb_sa_spi);
5427 	if (sav == NULL) {
5428 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u\n",
5429 		    __func__, ntohl(sa0->sadb_sa_spi)));
5430 		return key_senderror(so, m, EINVAL);
5431 	}
5432 	/*
5433 	 * Check that SADB_UPDATE issued by the same process that did
5434 	 * SADB_GETSPI or SADB_ADD.
5435 	 */
5436 	if (sav->pid != mhp->msg->sadb_msg_pid) {
5437 		ipseclog((LOG_DEBUG,
5438 		    "%s: pid mismatched (SPI %u, pid %u vs. %u)\n", __func__,
5439 		    ntohl(sav->spi), sav->pid, mhp->msg->sadb_msg_pid));
5440 		key_freesav(&sav);
5441 		return key_senderror(so, m, EINVAL);
5442 	}
5443 	/* saidx should match with SA. */
5444 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_MODE_REQID) == 0) {
5445 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u",
5446 		    __func__, ntohl(sav->spi)));
5447 		key_freesav(&sav);
5448 		return key_senderror(so, m, ESRCH);
5449 	}
5450 
5451 	if (sav->state == SADB_SASTATE_LARVAL) {
5452 		if ((mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5453 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) ||
5454 		    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5455 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH))) {
5456 			ipseclog((LOG_DEBUG,
5457 			    "%s: invalid message: missing required header.\n",
5458 			    __func__));
5459 			key_freesav(&sav);
5460 			return key_senderror(so, m, EINVAL);
5461 		}
5462 		/*
5463 		 * We can set any values except src, dst and SPI.
5464 		 */
5465 		error = key_setsaval(sav, mhp);
5466 		if (error != 0) {
5467 			key_freesav(&sav);
5468 			return (key_senderror(so, m, error));
5469 		}
5470 		/* Change SA state to MATURE */
5471 		SAHTREE_WLOCK();
5472 		if (sav->state != SADB_SASTATE_LARVAL) {
5473 			/* SA was deleted or another thread made it MATURE. */
5474 			SAHTREE_WUNLOCK();
5475 			key_freesav(&sav);
5476 			return (key_senderror(so, m, ESRCH));
5477 		}
5478 		/*
5479 		 * NOTE: we keep SAs in savtree_alive ordered by created
5480 		 * time. When SA's state changed from LARVAL to MATURE,
5481 		 * we update its created time in key_setsaval() and move
5482 		 * it into head of savtree_alive.
5483 		 */
5484 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
5485 		TAILQ_INSERT_HEAD(&sav->sah->savtree_alive, sav, chain);
5486 		sav->state = SADB_SASTATE_MATURE;
5487 		SAHTREE_WUNLOCK();
5488 	} else {
5489 		/*
5490 		 * For DYING and MATURE SA we can change only state
5491 		 * and lifetimes. Report EINVAL if something else attempted
5492 		 * to change.
5493 		 */
5494 		if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5495 		    !SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
5496 			key_freesav(&sav);
5497 			return (key_senderror(so, m, EINVAL));
5498 		}
5499 		error = key_updatelifetimes(sav, mhp);
5500 		if (error != 0) {
5501 			key_freesav(&sav);
5502 			return (key_senderror(so, m, error));
5503 		}
5504 		/*
5505 		 * This is FreeBSD extension to RFC2367.
5506 		 * IKEd can specify SADB_X_EXT_NEW_ADDRESS_SRC and/or
5507 		 * SADB_X_EXT_NEW_ADDRESS_DST when it wants to change
5508 		 * SA addresses (for example to implement MOBIKE protocol
5509 		 * as described in RFC4555). Also we allow to change
5510 		 * NAT-T config.
5511 		 */
5512 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5513 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST) ||
5514 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5515 		    sav->natt != NULL) {
5516 			error = key_updateaddresses(so, m, mhp, sav, &saidx);
5517 			key_freesav(&sav);
5518 			if (error != 0)
5519 				return (key_senderror(so, m, error));
5520 			return (0);
5521 		}
5522 		/* Check that SA is still alive */
5523 		SAHTREE_WLOCK();
5524 		if (sav->state == SADB_SASTATE_DEAD) {
5525 			/* SA was unlinked */
5526 			SAHTREE_WUNLOCK();
5527 			key_freesav(&sav);
5528 			return (key_senderror(so, m, ESRCH));
5529 		}
5530 		/*
5531 		 * NOTE: there is possible state moving from DYING to MATURE,
5532 		 * but this doesn't change created time, so we won't reorder
5533 		 * this SA.
5534 		 */
5535 		sav->state = SADB_SASTATE_MATURE;
5536 		SAHTREE_WUNLOCK();
5537 	}
5538 	KEYDBG(KEY_STAMP,
5539 	    printf("%s: SA(%p)\n", __func__, sav));
5540 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5541 	key_freesav(&sav);
5542 
5543     {
5544 	struct mbuf *n;
5545 
5546 	/* set msg buf from mhp */
5547 	n = key_getmsgbuf_x1(m, mhp);
5548 	if (n == NULL) {
5549 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5550 		return key_senderror(so, m, ENOBUFS);
5551 	}
5552 
5553 	m_freem(m);
5554 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5555     }
5556 }
5557 
5558 /*
5559  * SADB_ADD processing
5560  * add an entry to SA database, when received
5561  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5562  *       key(AE), (identity(SD),) (sensitivity)>
5563  * from the ikmpd,
5564  * and send
5565  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5566  *       (identity(SD),) (sensitivity)>
5567  * to the ikmpd.
5568  *
5569  * IGNORE identity and sensitivity messages.
5570  *
5571  * m will always be freed.
5572  */
5573 static int
5574 key_add(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5575 {
5576 	struct secasindex saidx;
5577 	struct sadb_address *src0, *dst0;
5578 	struct sadb_sa *sa0;
5579 	struct secasvar *sav;
5580 	uint32_t reqid, spi;
5581 	uint8_t mode, proto;
5582 	int error;
5583 
5584 	IPSEC_ASSERT(so != NULL, ("null socket"));
5585 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5586 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5587 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5588 
5589 	/* map satype to proto */
5590 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5591 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5592 		    __func__));
5593 		return key_senderror(so, m, EINVAL);
5594 	}
5595 
5596 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5597 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5598 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5599 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP && (
5600 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5601 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT))) ||
5602 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH && (
5603 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH) ||
5604 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH))) ||
5605 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5606 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5607 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5608 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5609 		ipseclog((LOG_DEBUG,
5610 		    "%s: invalid message: missing required header.\n",
5611 		    __func__));
5612 		return key_senderror(so, m, EINVAL);
5613 	}
5614 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5615 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5616 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5617 		ipseclog((LOG_DEBUG,
5618 		    "%s: invalid message: wrong header size.\n", __func__));
5619 		return key_senderror(so, m, EINVAL);
5620 	}
5621 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5622 		mode = IPSEC_MODE_ANY;
5623 		reqid = 0;
5624 	} else {
5625 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5626 			ipseclog((LOG_DEBUG,
5627 			    "%s: invalid message: wrong header size.\n",
5628 			    __func__));
5629 			return key_senderror(so, m, EINVAL);
5630 		}
5631 		mode = ((struct sadb_x_sa2 *)
5632 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5633 		reqid = ((struct sadb_x_sa2 *)
5634 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5635 	}
5636 
5637 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5638 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
5639 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
5640 
5641 	/*
5642 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5643 	 * SADB_ADD message.
5644 	 */
5645 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5646 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5647 #ifdef PFKEY_STRICT_CHECKS
5648 		return key_senderror(so, m, EINVAL);
5649 #endif
5650 	}
5651 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5652 	    (struct sockaddr *)(dst0 + 1));
5653 	if (error != 0) {
5654 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5655 		return key_senderror(so, m, error);
5656 	}
5657 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5658 	spi = sa0->sadb_sa_spi;
5659 	/*
5660 	 * For TCP-MD5 SAs we don't use SPI. Check the uniqueness using
5661 	 * secasindex.
5662 	 * XXXAE: IPComp seems also doesn't use SPI.
5663 	 */
5664 	if (proto == IPPROTO_TCP) {
5665 		sav = key_getsav_tcpmd5(&saidx, &spi);
5666 		if (sav == NULL && spi == 0) {
5667 			/* Failed to allocate SPI */
5668 			ipseclog((LOG_DEBUG, "%s: SA already exists.\n",
5669 			    __func__));
5670 			return key_senderror(so, m, EEXIST);
5671 		}
5672 		/* XXX: SPI that we report back can have another value */
5673 	} else {
5674 		/* We can create new SA only if SPI is different. */
5675 		sav = key_getsavbyspi(spi);
5676 	}
5677 	if (sav != NULL) {
5678 		key_freesav(&sav);
5679 		ipseclog((LOG_DEBUG, "%s: SA already exists.\n", __func__));
5680 		return key_senderror(so, m, EEXIST);
5681 	}
5682 
5683 	sav = key_newsav(mhp, &saidx, spi, &error);
5684 	if (sav == NULL)
5685 		return key_senderror(so, m, error);
5686 	KEYDBG(KEY_STAMP,
5687 	    printf("%s: return SA(%p)\n", __func__, sav));
5688 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5689 	/*
5690 	 * If SADB_ADD was in response to SADB_ACQUIRE, we need to schedule
5691 	 * ACQ for deletion.
5692 	 */
5693 	if (sav->seq != 0)
5694 		key_acqdone(&saidx, sav->seq);
5695 
5696     {
5697 	/*
5698 	 * Don't call key_freesav() on error here, as we would like to
5699 	 * keep the SA in the database.
5700 	 */
5701 	struct mbuf *n;
5702 
5703 	/* set msg buf from mhp */
5704 	n = key_getmsgbuf_x1(m, mhp);
5705 	if (n == NULL) {
5706 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5707 		return key_senderror(so, m, ENOBUFS);
5708 	}
5709 
5710 	m_freem(m);
5711 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5712     }
5713 }
5714 
5715 /*
5716  * NAT-T support.
5717  * IKEd may request the use ESP in UDP encapsulation when it detects the
5718  * presence of NAT. It uses NAT-T extension headers for such SAs to specify
5719  * parameters needed for encapsulation and decapsulation. These PF_KEY
5720  * extension headers are not standardized, so this comment addresses our
5721  * implementation.
5722  * SADB_X_EXT_NAT_T_TYPE specifies type of encapsulation, we support only
5723  * UDP_ENCAP_ESPINUDP as described in RFC3948.
5724  * SADB_X_EXT_NAT_T_SPORT/DPORT specifies source and destination ports for
5725  * UDP header. We use these ports in UDP encapsulation procedure, also we
5726  * can check them in UDP decapsulation procedure.
5727  * SADB_X_EXT_NAT_T_OA[IR] specifies original address of initiator or
5728  * responder. These addresses can be used for transport mode to adjust
5729  * checksum after decapsulation and decryption. Since original IP addresses
5730  * used by peer usually different (we detected presence of NAT), TCP/UDP
5731  * pseudo header checksum and IP header checksum was calculated using original
5732  * addresses. After decapsulation and decryption we need to adjust checksum
5733  * to have correct datagram.
5734  *
5735  * We expect presence of NAT-T extension headers only in SADB_ADD and
5736  * SADB_UPDATE messages. We report NAT-T extension headers in replies
5737  * to SADB_ADD, SADB_UPDATE, SADB_GET, and SADB_DUMP messages.
5738  */
5739 static int
5740 key_setnatt(struct secasvar *sav, const struct sadb_msghdr *mhp)
5741 {
5742 	struct sadb_x_nat_t_port *port;
5743 	struct sadb_x_nat_t_type *type;
5744 	struct sadb_address *oai, *oar;
5745 	struct sockaddr *sa;
5746 	uint32_t addr;
5747 	uint16_t cksum;
5748 
5749 	IPSEC_ASSERT(sav->natt == NULL, ("natt is already initialized"));
5750 	/*
5751 	 * Ignore NAT-T headers if sproto isn't ESP.
5752 	 */
5753 	if (sav->sah->saidx.proto != IPPROTO_ESP)
5754 		return (0);
5755 
5756 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) &&
5757 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) &&
5758 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5759 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5760 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5761 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5762 			ipseclog((LOG_DEBUG,
5763 			    "%s: invalid message: wrong header size.\n",
5764 			    __func__));
5765 			return (EINVAL);
5766 		}
5767 	} else
5768 		return (0);
5769 
5770 	type = (struct sadb_x_nat_t_type *)mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5771 	if (type->sadb_x_nat_t_type_type != UDP_ENCAP_ESPINUDP) {
5772 		ipseclog((LOG_DEBUG, "%s: unsupported NAT-T type %u.\n",
5773 		    __func__, type->sadb_x_nat_t_type_type));
5774 		return (EINVAL);
5775 	}
5776 	/*
5777 	 * Allocate storage for NAT-T config.
5778 	 * On error it will be released by key_cleansav().
5779 	 */
5780 	sav->natt = malloc(sizeof(struct secnatt), M_IPSEC_MISC,
5781 	    M_NOWAIT | M_ZERO);
5782 	if (sav->natt == NULL) {
5783 		PFKEYSTAT_INC(in_nomem);
5784 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5785 		return (ENOBUFS);
5786 	}
5787 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5788 	if (port->sadb_x_nat_t_port_port == 0) {
5789 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T sport specified.\n",
5790 		    __func__));
5791 		return (EINVAL);
5792 	}
5793 	sav->natt->sport = port->sadb_x_nat_t_port_port;
5794 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5795 	if (port->sadb_x_nat_t_port_port == 0) {
5796 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T dport specified.\n",
5797 		    __func__));
5798 		return (EINVAL);
5799 	}
5800 	sav->natt->dport = port->sadb_x_nat_t_port_port;
5801 
5802 	/*
5803 	 * SADB_X_EXT_NAT_T_OAI and SADB_X_EXT_NAT_T_OAR are optional
5804 	 * and needed only for transport mode IPsec.
5805 	 * Usually NAT translates only one address, but it is possible,
5806 	 * that both addresses could be translated.
5807 	 * NOTE: Value of SADB_X_EXT_NAT_T_OAI is equal to SADB_X_EXT_NAT_T_OA.
5808 	 */
5809 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAI)) {
5810 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAI)) {
5811 			ipseclog((LOG_DEBUG,
5812 			    "%s: invalid message: wrong header size.\n",
5813 			    __func__));
5814 			return (EINVAL);
5815 		}
5816 		oai = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAI];
5817 	} else
5818 		oai = NULL;
5819 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAR)) {
5820 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAR)) {
5821 			ipseclog((LOG_DEBUG,
5822 			    "%s: invalid message: wrong header size.\n",
5823 			    __func__));
5824 			return (EINVAL);
5825 		}
5826 		oar = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAR];
5827 	} else
5828 		oar = NULL;
5829 
5830 	/* Initialize addresses only for transport mode */
5831 	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL) {
5832 		cksum = 0;
5833 		if (oai != NULL) {
5834 			/* Currently we support only AF_INET */
5835 			sa = (struct sockaddr *)(oai + 1);
5836 			if (sa->sa_family != AF_INET ||
5837 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5838 				ipseclog((LOG_DEBUG,
5839 				    "%s: wrong NAT-OAi header.\n",
5840 				    __func__));
5841 				return (EINVAL);
5842 			}
5843 			/* Ignore address if it the same */
5844 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5845 			    sav->sah->saidx.src.sin.sin_addr.s_addr) {
5846 				bcopy(sa, &sav->natt->oai.sa, sa->sa_len);
5847 				sav->natt->flags |= IPSEC_NATT_F_OAI;
5848 				/* Calculate checksum delta */
5849 				addr = sav->sah->saidx.src.sin.sin_addr.s_addr;
5850 				cksum = in_addword(cksum, ~addr >> 16);
5851 				cksum = in_addword(cksum, ~addr & 0xffff);
5852 				addr = sav->natt->oai.sin.sin_addr.s_addr;
5853 				cksum = in_addword(cksum, addr >> 16);
5854 				cksum = in_addword(cksum, addr & 0xffff);
5855 			}
5856 		}
5857 		if (oar != NULL) {
5858 			/* Currently we support only AF_INET */
5859 			sa = (struct sockaddr *)(oar + 1);
5860 			if (sa->sa_family != AF_INET ||
5861 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5862 				ipseclog((LOG_DEBUG,
5863 				    "%s: wrong NAT-OAr header.\n",
5864 				    __func__));
5865 				return (EINVAL);
5866 			}
5867 			/* Ignore address if it the same */
5868 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5869 			    sav->sah->saidx.dst.sin.sin_addr.s_addr) {
5870 				bcopy(sa, &sav->natt->oar.sa, sa->sa_len);
5871 				sav->natt->flags |= IPSEC_NATT_F_OAR;
5872 				/* Calculate checksum delta */
5873 				addr = sav->sah->saidx.dst.sin.sin_addr.s_addr;
5874 				cksum = in_addword(cksum, ~addr >> 16);
5875 				cksum = in_addword(cksum, ~addr & 0xffff);
5876 				addr = sav->natt->oar.sin.sin_addr.s_addr;
5877 				cksum = in_addword(cksum, addr >> 16);
5878 				cksum = in_addword(cksum, addr & 0xffff);
5879 			}
5880 		}
5881 		sav->natt->cksum = cksum;
5882 	}
5883 	return (0);
5884 }
5885 
5886 static int
5887 key_setident(struct secashead *sah, const struct sadb_msghdr *mhp)
5888 {
5889 	const struct sadb_ident *idsrc, *iddst;
5890 
5891 	IPSEC_ASSERT(sah != NULL, ("null secashead"));
5892 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5893 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5894 
5895 	/* don't make buffer if not there */
5896 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) &&
5897 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5898 		sah->idents = NULL;
5899 		sah->identd = NULL;
5900 		return (0);
5901 	}
5902 
5903 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) ||
5904 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5905 		ipseclog((LOG_DEBUG, "%s: invalid identity.\n", __func__));
5906 		return (EINVAL);
5907 	}
5908 
5909 	idsrc = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_SRC];
5910 	iddst = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_DST];
5911 
5912 	/* validity check */
5913 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
5914 		ipseclog((LOG_DEBUG, "%s: ident type mismatch.\n", __func__));
5915 		return EINVAL;
5916 	}
5917 
5918 	switch (idsrc->sadb_ident_type) {
5919 	case SADB_IDENTTYPE_PREFIX:
5920 	case SADB_IDENTTYPE_FQDN:
5921 	case SADB_IDENTTYPE_USERFQDN:
5922 	default:
5923 		/* XXX do nothing */
5924 		sah->idents = NULL;
5925 		sah->identd = NULL;
5926 	 	return 0;
5927 	}
5928 
5929 	/* make structure */
5930 	sah->idents = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5931 	if (sah->idents == NULL) {
5932 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5933 		return ENOBUFS;
5934 	}
5935 	sah->identd = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5936 	if (sah->identd == NULL) {
5937 		free(sah->idents, M_IPSEC_MISC);
5938 		sah->idents = NULL;
5939 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5940 		return ENOBUFS;
5941 	}
5942 	sah->idents->type = idsrc->sadb_ident_type;
5943 	sah->idents->id = idsrc->sadb_ident_id;
5944 
5945 	sah->identd->type = iddst->sadb_ident_type;
5946 	sah->identd->id = iddst->sadb_ident_id;
5947 
5948 	return 0;
5949 }
5950 
5951 /*
5952  * m will not be freed on return.
5953  * it is caller's responsibility to free the result.
5954  *
5955  * Called from SADB_ADD and SADB_UPDATE. Reply will contain headers
5956  * from the request in defined order.
5957  */
5958 static struct mbuf *
5959 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
5960 {
5961 	struct mbuf *n;
5962 
5963 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5964 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5965 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5966 
5967 	/* create new sadb_msg to reply. */
5968 	n = key_gather_mbuf(m, mhp, 1, 16, SADB_EXT_RESERVED,
5969 	    SADB_EXT_SA, SADB_X_EXT_SA2,
5970 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
5971 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
5972 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
5973 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
5974 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
5975 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NEW_ADDRESS_SRC,
5976 	    SADB_X_EXT_NEW_ADDRESS_DST);
5977 	if (!n)
5978 		return NULL;
5979 
5980 	if (n->m_len < sizeof(struct sadb_msg)) {
5981 		n = m_pullup(n, sizeof(struct sadb_msg));
5982 		if (n == NULL)
5983 			return NULL;
5984 	}
5985 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
5986 	mtod(n, struct sadb_msg *)->sadb_msg_len =
5987 	    PFKEY_UNIT64(n->m_pkthdr.len);
5988 
5989 	return n;
5990 }
5991 
5992 /*
5993  * SADB_DELETE processing
5994  * receive
5995  *   <base, SA(*), address(SD)>
5996  * from the ikmpd, and set SADB_SASTATE_DEAD,
5997  * and send,
5998  *   <base, SA(*), address(SD)>
5999  * to the ikmpd.
6000  *
6001  * m will always be freed.
6002  */
6003 static int
6004 key_delete(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6005 {
6006 	struct secasindex saidx;
6007 	struct sadb_address *src0, *dst0;
6008 	struct secasvar *sav;
6009 	struct sadb_sa *sa0;
6010 	uint8_t proto;
6011 
6012 	IPSEC_ASSERT(so != NULL, ("null socket"));
6013 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6014 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6015 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6016 
6017 	/* map satype to proto */
6018 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6019 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6020 		    __func__));
6021 		return key_senderror(so, m, EINVAL);
6022 	}
6023 
6024 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6025 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
6026 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6027 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6028 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
6029 		    __func__));
6030 		return key_senderror(so, m, EINVAL);
6031 	}
6032 
6033 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
6034 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
6035 
6036 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6037 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6038 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6039 		return (key_senderror(so, m, EINVAL));
6040 	}
6041 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6042 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
6043 		/*
6044 		 * Caller wants us to delete all non-LARVAL SAs
6045 		 * that match the src/dst.  This is used during
6046 		 * IKE INITIAL-CONTACT.
6047 		 * XXXAE: this looks like some extension to RFC2367.
6048 		 */
6049 		ipseclog((LOG_DEBUG, "%s: doing delete all.\n", __func__));
6050 		return (key_delete_all(so, m, mhp, &saidx));
6051 	}
6052 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
6053 		ipseclog((LOG_DEBUG,
6054 		    "%s: invalid message: wrong header size.\n", __func__));
6055 		return (key_senderror(so, m, EINVAL));
6056 	}
6057 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6058 	if (proto == IPPROTO_TCP)
6059 		sav = key_getsav_tcpmd5(&saidx, NULL);
6060 	else
6061 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6062 	if (sav == NULL) {
6063 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u.\n",
6064 		    __func__, ntohl(sa0->sadb_sa_spi)));
6065 		return (key_senderror(so, m, ESRCH));
6066 	}
6067 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6068 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6069 		    __func__, ntohl(sav->spi)));
6070 		key_freesav(&sav);
6071 		return (key_senderror(so, m, ESRCH));
6072 	}
6073 	KEYDBG(KEY_STAMP,
6074 	    printf("%s: SA(%p)\n", __func__, sav));
6075 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
6076 	key_unlinksav(sav);
6077 	key_freesav(&sav);
6078 
6079     {
6080 	struct mbuf *n;
6081 	struct sadb_msg *newmsg;
6082 
6083 	/* create new sadb_msg to reply. */
6084 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
6085 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6086 	if (!n)
6087 		return key_senderror(so, m, ENOBUFS);
6088 
6089 	if (n->m_len < sizeof(struct sadb_msg)) {
6090 		n = m_pullup(n, sizeof(struct sadb_msg));
6091 		if (n == NULL)
6092 			return key_senderror(so, m, ENOBUFS);
6093 	}
6094 	newmsg = mtod(n, struct sadb_msg *);
6095 	newmsg->sadb_msg_errno = 0;
6096 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6097 
6098 	m_freem(m);
6099 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6100     }
6101 }
6102 
6103 /*
6104  * delete all SAs for src/dst.  Called from key_delete().
6105  */
6106 static int
6107 key_delete_all(struct socket *so, struct mbuf *m,
6108     const struct sadb_msghdr *mhp, struct secasindex *saidx)
6109 {
6110 	struct secasvar_queue drainq;
6111 	struct secashead *sah;
6112 	struct secasvar *sav, *nextsav;
6113 
6114 	TAILQ_INIT(&drainq);
6115 	SAHTREE_WLOCK();
6116 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
6117 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_HEAD) == 0)
6118 			continue;
6119 		/* Move all ALIVE SAs into drainq */
6120 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6121 	}
6122 	/* Unlink all queued SAs from SPI hash */
6123 	TAILQ_FOREACH(sav, &drainq, chain) {
6124 		sav->state = SADB_SASTATE_DEAD;
6125 		LIST_REMOVE(sav, spihash);
6126 	}
6127 	SAHTREE_WUNLOCK();
6128 	/* Now we can release reference for all SAs in drainq */
6129 	sav = TAILQ_FIRST(&drainq);
6130 	while (sav != NULL) {
6131 		KEYDBG(KEY_STAMP,
6132 		    printf("%s: SA(%p)\n", __func__, sav));
6133 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6134 		nextsav = TAILQ_NEXT(sav, chain);
6135 		key_freesah(&sav->sah); /* release reference from SAV */
6136 		key_freesav(&sav); /* release last reference */
6137 		sav = nextsav;
6138 	}
6139 
6140     {
6141 	struct mbuf *n;
6142 	struct sadb_msg *newmsg;
6143 
6144 	/* create new sadb_msg to reply. */
6145 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
6146 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6147 	if (!n)
6148 		return key_senderror(so, m, ENOBUFS);
6149 
6150 	if (n->m_len < sizeof(struct sadb_msg)) {
6151 		n = m_pullup(n, sizeof(struct sadb_msg));
6152 		if (n == NULL)
6153 			return key_senderror(so, m, ENOBUFS);
6154 	}
6155 	newmsg = mtod(n, struct sadb_msg *);
6156 	newmsg->sadb_msg_errno = 0;
6157 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6158 
6159 	m_freem(m);
6160 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6161     }
6162 }
6163 
6164 /*
6165  * Delete all alive SAs for corresponding xform.
6166  * Larval SAs have not initialized tdb_xform, so it is safe to leave them
6167  * here when xform disappears.
6168  */
6169 static void
6170 key_delete_xform(const struct xformsw *xsp)
6171 {
6172 	struct secasvar_queue drainq;
6173 	struct secashead *sah;
6174 	struct secasvar *sav, *nextsav;
6175 
6176 	TAILQ_INIT(&drainq);
6177 	SAHTREE_WLOCK();
6178 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
6179 		sav = TAILQ_FIRST(&sah->savtree_alive);
6180 		if (sav == NULL)
6181 			continue;
6182 		if (sav->tdb_xform != xsp)
6183 			continue;
6184 		/*
6185 		 * It is supposed that all SAs in the chain are related to
6186 		 * one xform.
6187 		 */
6188 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6189 	}
6190 	/* Unlink all queued SAs from SPI hash */
6191 	TAILQ_FOREACH(sav, &drainq, chain) {
6192 		sav->state = SADB_SASTATE_DEAD;
6193 		LIST_REMOVE(sav, spihash);
6194 	}
6195 	SAHTREE_WUNLOCK();
6196 
6197 	/* Now we can release reference for all SAs in drainq */
6198 	sav = TAILQ_FIRST(&drainq);
6199 	while (sav != NULL) {
6200 		KEYDBG(KEY_STAMP,
6201 		    printf("%s: SA(%p)\n", __func__, sav));
6202 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6203 		nextsav = TAILQ_NEXT(sav, chain);
6204 		key_freesah(&sav->sah); /* release reference from SAV */
6205 		key_freesav(&sav); /* release last reference */
6206 		sav = nextsav;
6207 	}
6208 }
6209 
6210 /*
6211  * SADB_GET processing
6212  * receive
6213  *   <base, SA(*), address(SD)>
6214  * from the ikmpd, and get a SP and a SA to respond,
6215  * and send,
6216  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6217  *       (identity(SD),) (sensitivity)>
6218  * to the ikmpd.
6219  *
6220  * m will always be freed.
6221  */
6222 static int
6223 key_get(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6224 {
6225 	struct secasindex saidx;
6226 	struct sadb_address *src0, *dst0;
6227 	struct sadb_sa *sa0;
6228 	struct secasvar *sav;
6229 	uint8_t proto;
6230 
6231 	IPSEC_ASSERT(so != NULL, ("null socket"));
6232 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6233 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6234 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6235 
6236 	/* map satype to proto */
6237 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6238 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6239 			__func__));
6240 		return key_senderror(so, m, EINVAL);
6241 	}
6242 
6243 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
6244 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6245 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)) {
6246 		ipseclog((LOG_DEBUG,
6247 		    "%s: invalid message: missing required header.\n",
6248 		    __func__));
6249 		return key_senderror(so, m, EINVAL);
6250 	}
6251 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
6252 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6253 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6254 		ipseclog((LOG_DEBUG,
6255 		    "%s: invalid message: wrong header size.\n", __func__));
6256 		return key_senderror(so, m, EINVAL);
6257 	}
6258 
6259 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6260 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
6261 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
6262 
6263 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6264 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6265 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6266 		return key_senderror(so, m, EINVAL);
6267 	}
6268 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6269 
6270 	if (proto == IPPROTO_TCP)
6271 		sav = key_getsav_tcpmd5(&saidx, NULL);
6272 	else
6273 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6274 	if (sav == NULL) {
6275 		ipseclog((LOG_DEBUG, "%s: no SA found.\n", __func__));
6276 		return key_senderror(so, m, ESRCH);
6277 	}
6278 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6279 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6280 		    __func__, ntohl(sa0->sadb_sa_spi)));
6281 		key_freesav(&sav);
6282 		return (key_senderror(so, m, ESRCH));
6283 	}
6284 
6285     {
6286 	struct mbuf *n;
6287 	uint8_t satype;
6288 
6289 	/* map proto to satype */
6290 	if ((satype = key_proto2satype(sav->sah->saidx.proto)) == 0) {
6291 		ipseclog((LOG_DEBUG, "%s: there was invalid proto in SAD.\n",
6292 		    __func__));
6293 		key_freesav(&sav);
6294 		return key_senderror(so, m, EINVAL);
6295 	}
6296 
6297 	/* create new sadb_msg to reply. */
6298 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
6299 	    mhp->msg->sadb_msg_pid);
6300 
6301 	key_freesav(&sav);
6302 	if (!n)
6303 		return key_senderror(so, m, ENOBUFS);
6304 
6305 	m_freem(m);
6306 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
6307     }
6308 }
6309 
6310 /* XXX make it sysctl-configurable? */
6311 static void
6312 key_getcomb_setlifetime(struct sadb_comb *comb)
6313 {
6314 
6315 	comb->sadb_comb_soft_allocations = 1;
6316 	comb->sadb_comb_hard_allocations = 1;
6317 	comb->sadb_comb_soft_bytes = 0;
6318 	comb->sadb_comb_hard_bytes = 0;
6319 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
6320 	comb->sadb_comb_soft_addtime = comb->sadb_comb_soft_addtime * 80 / 100;
6321 	comb->sadb_comb_soft_usetime = 28800;	/* 8 hours */
6322 	comb->sadb_comb_hard_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
6323 }
6324 
6325 /*
6326  * XXX reorder combinations by preference
6327  * XXX no idea if the user wants ESP authentication or not
6328  */
6329 static struct mbuf *
6330 key_getcomb_ealg(void)
6331 {
6332 	struct sadb_comb *comb;
6333 	const struct enc_xform *algo;
6334 	struct mbuf *result = NULL, *m, *n;
6335 	int encmin;
6336 	int i, off, o;
6337 	int totlen;
6338 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6339 
6340 	m = NULL;
6341 	for (i = 1; i <= SADB_EALG_MAX; i++) {
6342 		algo = enc_algorithm_lookup(i);
6343 		if (algo == NULL)
6344 			continue;
6345 
6346 		/* discard algorithms with key size smaller than system min */
6347 		if (_BITS(algo->maxkey) < V_ipsec_esp_keymin)
6348 			continue;
6349 		if (_BITS(algo->minkey) < V_ipsec_esp_keymin)
6350 			encmin = V_ipsec_esp_keymin;
6351 		else
6352 			encmin = _BITS(algo->minkey);
6353 
6354 		if (V_ipsec_esp_auth)
6355 			m = key_getcomb_ah();
6356 		else {
6357 			IPSEC_ASSERT(l <= MLEN,
6358 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6359 			MGET(m, M_NOWAIT, MT_DATA);
6360 			if (m) {
6361 				M_ALIGN(m, l);
6362 				m->m_len = l;
6363 				m->m_next = NULL;
6364 				bzero(mtod(m, caddr_t), m->m_len);
6365 			}
6366 		}
6367 		if (!m)
6368 			goto fail;
6369 
6370 		totlen = 0;
6371 		for (n = m; n; n = n->m_next)
6372 			totlen += n->m_len;
6373 		IPSEC_ASSERT((totlen % l) == 0, ("totlen=%u, l=%u", totlen, l));
6374 
6375 		for (off = 0; off < totlen; off += l) {
6376 			n = m_pulldown(m, off, l, &o);
6377 			if (!n) {
6378 				/* m is already freed */
6379 				goto fail;
6380 			}
6381 			comb = (struct sadb_comb *)(mtod(n, caddr_t) + o);
6382 			bzero(comb, sizeof(*comb));
6383 			key_getcomb_setlifetime(comb);
6384 			comb->sadb_comb_encrypt = i;
6385 			comb->sadb_comb_encrypt_minbits = encmin;
6386 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
6387 		}
6388 
6389 		if (!result)
6390 			result = m;
6391 		else
6392 			m_cat(result, m);
6393 	}
6394 
6395 	return result;
6396 
6397  fail:
6398 	if (result)
6399 		m_freem(result);
6400 	return NULL;
6401 }
6402 
6403 static void
6404 key_getsizes_ah(const struct auth_hash *ah, int alg, u_int16_t* min,
6405     u_int16_t* max)
6406 {
6407 
6408 	*min = *max = ah->hashsize;
6409 	if (ah->keysize == 0) {
6410 		/*
6411 		 * Transform takes arbitrary key size but algorithm
6412 		 * key size is restricted.  Enforce this here.
6413 		 */
6414 		switch (alg) {
6415 		case SADB_X_AALG_MD5:	*min = *max = 16; break;
6416 		case SADB_X_AALG_SHA:	*min = *max = 20; break;
6417 		case SADB_X_AALG_NULL:	*min = 1; *max = 256; break;
6418 		case SADB_X_AALG_SHA2_256: *min = *max = 32; break;
6419 		case SADB_X_AALG_SHA2_384: *min = *max = 48; break;
6420 		case SADB_X_AALG_SHA2_512: *min = *max = 64; break;
6421 		default:
6422 			DPRINTF(("%s: unknown AH algorithm %u\n",
6423 				__func__, alg));
6424 			break;
6425 		}
6426 	}
6427 }
6428 
6429 /*
6430  * XXX reorder combinations by preference
6431  */
6432 static struct mbuf *
6433 key_getcomb_ah()
6434 {
6435 	const struct auth_hash *algo;
6436 	struct sadb_comb *comb;
6437 	struct mbuf *m;
6438 	u_int16_t minkeysize, maxkeysize;
6439 	int i;
6440 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6441 
6442 	m = NULL;
6443 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6444 #if 1
6445 		/* we prefer HMAC algorithms, not old algorithms */
6446 		if (i != SADB_AALG_SHA1HMAC &&
6447 		    i != SADB_AALG_MD5HMAC  &&
6448 		    i != SADB_X_AALG_SHA2_256 &&
6449 		    i != SADB_X_AALG_SHA2_384 &&
6450 		    i != SADB_X_AALG_SHA2_512)
6451 			continue;
6452 #endif
6453 		algo = auth_algorithm_lookup(i);
6454 		if (!algo)
6455 			continue;
6456 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6457 		/* discard algorithms with key size smaller than system min */
6458 		if (_BITS(minkeysize) < V_ipsec_ah_keymin)
6459 			continue;
6460 
6461 		if (!m) {
6462 			IPSEC_ASSERT(l <= MLEN,
6463 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6464 			MGET(m, M_NOWAIT, MT_DATA);
6465 			if (m) {
6466 				M_ALIGN(m, l);
6467 				m->m_len = l;
6468 				m->m_next = NULL;
6469 			}
6470 		} else
6471 			M_PREPEND(m, l, M_NOWAIT);
6472 		if (!m)
6473 			return NULL;
6474 
6475 		comb = mtod(m, struct sadb_comb *);
6476 		bzero(comb, sizeof(*comb));
6477 		key_getcomb_setlifetime(comb);
6478 		comb->sadb_comb_auth = i;
6479 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6480 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6481 	}
6482 
6483 	return m;
6484 }
6485 
6486 /*
6487  * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6488  * XXX reorder combinations by preference
6489  */
6490 static struct mbuf *
6491 key_getcomb_ipcomp()
6492 {
6493 	const struct comp_algo *algo;
6494 	struct sadb_comb *comb;
6495 	struct mbuf *m;
6496 	int i;
6497 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6498 
6499 	m = NULL;
6500 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6501 		algo = comp_algorithm_lookup(i);
6502 		if (!algo)
6503 			continue;
6504 
6505 		if (!m) {
6506 			IPSEC_ASSERT(l <= MLEN,
6507 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6508 			MGET(m, M_NOWAIT, MT_DATA);
6509 			if (m) {
6510 				M_ALIGN(m, l);
6511 				m->m_len = l;
6512 				m->m_next = NULL;
6513 			}
6514 		} else
6515 			M_PREPEND(m, l, M_NOWAIT);
6516 		if (!m)
6517 			return NULL;
6518 
6519 		comb = mtod(m, struct sadb_comb *);
6520 		bzero(comb, sizeof(*comb));
6521 		key_getcomb_setlifetime(comb);
6522 		comb->sadb_comb_encrypt = i;
6523 		/* what should we set into sadb_comb_*_{min,max}bits? */
6524 	}
6525 
6526 	return m;
6527 }
6528 
6529 /*
6530  * XXX no way to pass mode (transport/tunnel) to userland
6531  * XXX replay checking?
6532  * XXX sysctl interface to ipsec_{ah,esp}_keymin
6533  */
6534 static struct mbuf *
6535 key_getprop(const struct secasindex *saidx)
6536 {
6537 	struct sadb_prop *prop;
6538 	struct mbuf *m, *n;
6539 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6540 	int totlen;
6541 
6542 	switch (saidx->proto)  {
6543 	case IPPROTO_ESP:
6544 		m = key_getcomb_ealg();
6545 		break;
6546 	case IPPROTO_AH:
6547 		m = key_getcomb_ah();
6548 		break;
6549 	case IPPROTO_IPCOMP:
6550 		m = key_getcomb_ipcomp();
6551 		break;
6552 	default:
6553 		return NULL;
6554 	}
6555 
6556 	if (!m)
6557 		return NULL;
6558 	M_PREPEND(m, l, M_NOWAIT);
6559 	if (!m)
6560 		return NULL;
6561 
6562 	totlen = 0;
6563 	for (n = m; n; n = n->m_next)
6564 		totlen += n->m_len;
6565 
6566 	prop = mtod(m, struct sadb_prop *);
6567 	bzero(prop, sizeof(*prop));
6568 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6569 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6570 	prop->sadb_prop_replay = 32;	/* XXX */
6571 
6572 	return m;
6573 }
6574 
6575 /*
6576  * SADB_ACQUIRE processing called by key_checkrequest() and key_acquire2().
6577  * send
6578  *   <base, SA, address(SD), (address(P)), x_policy,
6579  *       (identity(SD),) (sensitivity,) proposal>
6580  * to KMD, and expect to receive
6581  *   <base> with SADB_ACQUIRE if error occurred,
6582  * or
6583  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6584  * from KMD by PF_KEY.
6585  *
6586  * XXX x_policy is outside of RFC2367 (KAME extension).
6587  * XXX sensitivity is not supported.
6588  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6589  * see comment for key_getcomb_ipcomp().
6590  *
6591  * OUT:
6592  *    0     : succeed
6593  *    others: error number
6594  */
6595 static int
6596 key_acquire(const struct secasindex *saidx, struct secpolicy *sp)
6597 {
6598 	union sockaddr_union addr;
6599 	struct mbuf *result, *m;
6600 	uint32_t seq;
6601 	int error;
6602 	uint16_t ul_proto;
6603 	uint8_t mask, satype;
6604 
6605 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
6606 	satype = key_proto2satype(saidx->proto);
6607 	IPSEC_ASSERT(satype != 0, ("null satype, protocol %u", saidx->proto));
6608 
6609 	error = -1;
6610 	result = NULL;
6611 	ul_proto = IPSEC_ULPROTO_ANY;
6612 
6613 	/* Get seq number to check whether sending message or not. */
6614 	seq = key_getacq(saidx, &error);
6615 	if (seq == 0)
6616 		return (error);
6617 
6618 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0);
6619 	if (!m) {
6620 		error = ENOBUFS;
6621 		goto fail;
6622 	}
6623 	result = m;
6624 
6625 	/*
6626 	 * set sadb_address for saidx's.
6627 	 *
6628 	 * Note that if sp is supplied, then we're being called from
6629 	 * key_allocsa_policy() and should supply port and protocol
6630 	 * information.
6631 	 * XXXAE: why only TCP and UDP? ICMP and SCTP looks applicable too.
6632 	 * XXXAE: probably we can handle this in the ipsec[46]_allocsa().
6633 	 * XXXAE: it looks like we should save this info in the ACQ entry.
6634 	 */
6635 	if (sp != NULL && (sp->spidx.ul_proto == IPPROTO_TCP ||
6636 	    sp->spidx.ul_proto == IPPROTO_UDP))
6637 		ul_proto = sp->spidx.ul_proto;
6638 
6639 	addr = saidx->src;
6640 	mask = FULLMASK;
6641 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6642 		switch (sp->spidx.src.sa.sa_family) {
6643 		case AF_INET:
6644 			if (sp->spidx.src.sin.sin_port != IPSEC_PORT_ANY) {
6645 				addr.sin.sin_port = sp->spidx.src.sin.sin_port;
6646 				mask = sp->spidx.prefs;
6647 			}
6648 			break;
6649 		case AF_INET6:
6650 			if (sp->spidx.src.sin6.sin6_port != IPSEC_PORT_ANY) {
6651 				addr.sin6.sin6_port =
6652 				    sp->spidx.src.sin6.sin6_port;
6653 				mask = sp->spidx.prefs;
6654 			}
6655 			break;
6656 		default:
6657 			break;
6658 		}
6659 	}
6660 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &addr.sa, mask, ul_proto);
6661 	if (!m) {
6662 		error = ENOBUFS;
6663 		goto fail;
6664 	}
6665 	m_cat(result, m);
6666 
6667 	addr = saidx->dst;
6668 	mask = FULLMASK;
6669 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6670 		switch (sp->spidx.dst.sa.sa_family) {
6671 		case AF_INET:
6672 			if (sp->spidx.dst.sin.sin_port != IPSEC_PORT_ANY) {
6673 				addr.sin.sin_port = sp->spidx.dst.sin.sin_port;
6674 				mask = sp->spidx.prefd;
6675 			}
6676 			break;
6677 		case AF_INET6:
6678 			if (sp->spidx.dst.sin6.sin6_port != IPSEC_PORT_ANY) {
6679 				addr.sin6.sin6_port =
6680 				    sp->spidx.dst.sin6.sin6_port;
6681 				mask = sp->spidx.prefd;
6682 			}
6683 			break;
6684 		default:
6685 			break;
6686 		}
6687 	}
6688 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &addr.sa, mask, ul_proto);
6689 	if (!m) {
6690 		error = ENOBUFS;
6691 		goto fail;
6692 	}
6693 	m_cat(result, m);
6694 
6695 	/* XXX proxy address (optional) */
6696 
6697 	/* set sadb_x_policy */
6698 	if (sp != NULL) {
6699 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6700 		    sp->priority);
6701 		if (!m) {
6702 			error = ENOBUFS;
6703 			goto fail;
6704 		}
6705 		m_cat(result, m);
6706 	}
6707 
6708 	/* XXX identity (optional) */
6709 #if 0
6710 	if (idexttype && fqdn) {
6711 		/* create identity extension (FQDN) */
6712 		struct sadb_ident *id;
6713 		int fqdnlen;
6714 
6715 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
6716 		id = (struct sadb_ident *)p;
6717 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6718 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6719 		id->sadb_ident_exttype = idexttype;
6720 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
6721 		bcopy(fqdn, id + 1, fqdnlen);
6722 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
6723 	}
6724 
6725 	if (idexttype) {
6726 		/* create identity extension (USERFQDN) */
6727 		struct sadb_ident *id;
6728 		int userfqdnlen;
6729 
6730 		if (userfqdn) {
6731 			/* +1 for terminating-NUL */
6732 			userfqdnlen = strlen(userfqdn) + 1;
6733 		} else
6734 			userfqdnlen = 0;
6735 		id = (struct sadb_ident *)p;
6736 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6737 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6738 		id->sadb_ident_exttype = idexttype;
6739 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
6740 		/* XXX is it correct? */
6741 		if (curproc && curproc->p_cred)
6742 			id->sadb_ident_id = curproc->p_cred->p_ruid;
6743 		if (userfqdn && userfqdnlen)
6744 			bcopy(userfqdn, id + 1, userfqdnlen);
6745 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
6746 	}
6747 #endif
6748 
6749 	/* XXX sensitivity (optional) */
6750 
6751 	/* create proposal/combination extension */
6752 	m = key_getprop(saidx);
6753 #if 0
6754 	/*
6755 	 * spec conformant: always attach proposal/combination extension,
6756 	 * the problem is that we have no way to attach it for ipcomp,
6757 	 * due to the way sadb_comb is declared in RFC2367.
6758 	 */
6759 	if (!m) {
6760 		error = ENOBUFS;
6761 		goto fail;
6762 	}
6763 	m_cat(result, m);
6764 #else
6765 	/*
6766 	 * outside of spec; make proposal/combination extension optional.
6767 	 */
6768 	if (m)
6769 		m_cat(result, m);
6770 #endif
6771 
6772 	if ((result->m_flags & M_PKTHDR) == 0) {
6773 		error = EINVAL;
6774 		goto fail;
6775 	}
6776 
6777 	if (result->m_len < sizeof(struct sadb_msg)) {
6778 		result = m_pullup(result, sizeof(struct sadb_msg));
6779 		if (result == NULL) {
6780 			error = ENOBUFS;
6781 			goto fail;
6782 		}
6783 	}
6784 
6785 	result->m_pkthdr.len = 0;
6786 	for (m = result; m; m = m->m_next)
6787 		result->m_pkthdr.len += m->m_len;
6788 
6789 	mtod(result, struct sadb_msg *)->sadb_msg_len =
6790 	    PFKEY_UNIT64(result->m_pkthdr.len);
6791 
6792 	KEYDBG(KEY_STAMP,
6793 	    printf("%s: SP(%p)\n", __func__, sp));
6794 	KEYDBG(KEY_DATA, kdebug_secasindex(saidx, NULL));
6795 
6796 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
6797 
6798  fail:
6799 	if (result)
6800 		m_freem(result);
6801 	return error;
6802 }
6803 
6804 static uint32_t
6805 key_newacq(const struct secasindex *saidx, int *perror)
6806 {
6807 	struct secacq *acq;
6808 	uint32_t seq;
6809 
6810 	acq = malloc(sizeof(*acq), M_IPSEC_SAQ, M_NOWAIT | M_ZERO);
6811 	if (acq == NULL) {
6812 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6813 		*perror = ENOBUFS;
6814 		return (0);
6815 	}
6816 
6817 	/* copy secindex */
6818 	bcopy(saidx, &acq->saidx, sizeof(acq->saidx));
6819 	acq->created = time_second;
6820 	acq->count = 0;
6821 
6822 	/* add to acqtree */
6823 	ACQ_LOCK();
6824 	seq = acq->seq = (V_acq_seq == ~0 ? 1 : ++V_acq_seq);
6825 	LIST_INSERT_HEAD(&V_acqtree, acq, chain);
6826 	LIST_INSERT_HEAD(ACQADDRHASH_HASH(saidx), acq, addrhash);
6827 	LIST_INSERT_HEAD(ACQSEQHASH_HASH(seq), acq, seqhash);
6828 	ACQ_UNLOCK();
6829 	*perror = 0;
6830 	return (seq);
6831 }
6832 
6833 static uint32_t
6834 key_getacq(const struct secasindex *saidx, int *perror)
6835 {
6836 	struct secacq *acq;
6837 	uint32_t seq;
6838 
6839 	ACQ_LOCK();
6840 	LIST_FOREACH(acq, ACQADDRHASH_HASH(saidx), addrhash) {
6841 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY)) {
6842 			if (acq->count > V_key_blockacq_count) {
6843 				/*
6844 				 * Reset counter and send message.
6845 				 * Also reset created time to keep ACQ for
6846 				 * this saidx.
6847 				 */
6848 				acq->created = time_second;
6849 				acq->count = 0;
6850 				seq = acq->seq;
6851 			} else {
6852 				/*
6853 				 * Increment counter and do nothing.
6854 				 * We send SADB_ACQUIRE message only
6855 				 * for each V_key_blockacq_count packet.
6856 				 */
6857 				acq->count++;
6858 				seq = 0;
6859 			}
6860 			break;
6861 		}
6862 	}
6863 	ACQ_UNLOCK();
6864 	if (acq != NULL) {
6865 		*perror = 0;
6866 		return (seq);
6867 	}
6868 	/* allocate new  entry */
6869 	return (key_newacq(saidx, perror));
6870 }
6871 
6872 static int
6873 key_acqreset(uint32_t seq)
6874 {
6875 	struct secacq *acq;
6876 
6877 	ACQ_LOCK();
6878 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6879 		if (acq->seq == seq) {
6880 			acq->count = 0;
6881 			acq->created = time_second;
6882 			break;
6883 		}
6884 	}
6885 	ACQ_UNLOCK();
6886 	if (acq == NULL)
6887 		return (ESRCH);
6888 	return (0);
6889 }
6890 /*
6891  * Mark ACQ entry as stale to remove it in key_flush_acq().
6892  * Called after successful SADB_GETSPI message.
6893  */
6894 static int
6895 key_acqdone(const struct secasindex *saidx, uint32_t seq)
6896 {
6897 	struct secacq *acq;
6898 
6899 	ACQ_LOCK();
6900 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6901 		if (acq->seq == seq)
6902 			break;
6903 	}
6904 	if (acq != NULL) {
6905 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY) == 0) {
6906 			ipseclog((LOG_DEBUG,
6907 			    "%s: Mismatched saidx for ACQ %u", __func__, seq));
6908 			acq = NULL;
6909 		} else {
6910 			acq->created = 0;
6911 		}
6912 	} else {
6913 		ipseclog((LOG_DEBUG,
6914 		    "%s: ACQ %u is not found.", __func__, seq));
6915 	}
6916 	ACQ_UNLOCK();
6917 	if (acq == NULL)
6918 		return (ESRCH);
6919 	return (0);
6920 }
6921 
6922 static struct secspacq *
6923 key_newspacq(struct secpolicyindex *spidx)
6924 {
6925 	struct secspacq *acq;
6926 
6927 	/* get new entry */
6928 	acq = malloc(sizeof(struct secspacq), M_IPSEC_SAQ, M_NOWAIT|M_ZERO);
6929 	if (acq == NULL) {
6930 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6931 		return NULL;
6932 	}
6933 
6934 	/* copy secindex */
6935 	bcopy(spidx, &acq->spidx, sizeof(acq->spidx));
6936 	acq->created = time_second;
6937 	acq->count = 0;
6938 
6939 	/* add to spacqtree */
6940 	SPACQ_LOCK();
6941 	LIST_INSERT_HEAD(&V_spacqtree, acq, chain);
6942 	SPACQ_UNLOCK();
6943 
6944 	return acq;
6945 }
6946 
6947 static struct secspacq *
6948 key_getspacq(struct secpolicyindex *spidx)
6949 {
6950 	struct secspacq *acq;
6951 
6952 	SPACQ_LOCK();
6953 	LIST_FOREACH(acq, &V_spacqtree, chain) {
6954 		if (key_cmpspidx_exactly(spidx, &acq->spidx)) {
6955 			/* NB: return holding spacq_lock */
6956 			return acq;
6957 		}
6958 	}
6959 	SPACQ_UNLOCK();
6960 
6961 	return NULL;
6962 }
6963 
6964 /*
6965  * SADB_ACQUIRE processing,
6966  * in first situation, is receiving
6967  *   <base>
6968  * from the ikmpd, and clear sequence of its secasvar entry.
6969  *
6970  * In second situation, is receiving
6971  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6972  * from a user land process, and return
6973  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6974  * to the socket.
6975  *
6976  * m will always be freed.
6977  */
6978 static int
6979 key_acquire2(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6980 {
6981 	SAHTREE_RLOCK_TRACKER;
6982 	struct sadb_address *src0, *dst0;
6983 	struct secasindex saidx;
6984 	struct secashead *sah;
6985 	uint32_t reqid;
6986 	int error;
6987 	uint8_t mode, proto;
6988 
6989 	IPSEC_ASSERT(so != NULL, ("null socket"));
6990 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6991 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6992 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6993 
6994 	/*
6995 	 * Error message from KMd.
6996 	 * We assume that if error was occurred in IKEd, the length of PFKEY
6997 	 * message is equal to the size of sadb_msg structure.
6998 	 * We do not raise error even if error occurred in this function.
6999 	 */
7000 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
7001 		/* check sequence number */
7002 		if (mhp->msg->sadb_msg_seq == 0 ||
7003 		    mhp->msg->sadb_msg_errno == 0) {
7004 			ipseclog((LOG_DEBUG, "%s: must specify sequence "
7005 				"number and errno.\n", __func__));
7006 		} else {
7007 			/*
7008 			 * IKEd reported that error occurred.
7009 			 * XXXAE: what it expects from the kernel?
7010 			 * Probably we should send SADB_ACQUIRE again?
7011 			 * If so, reset ACQ's state.
7012 			 * XXXAE: it looks useless.
7013 			 */
7014 			key_acqreset(mhp->msg->sadb_msg_seq);
7015 		}
7016 		m_freem(m);
7017 		return (0);
7018 	}
7019 
7020 	/*
7021 	 * This message is from user land.
7022 	 */
7023 
7024 	/* map satype to proto */
7025 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7026 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7027 		    __func__));
7028 		return key_senderror(so, m, EINVAL);
7029 	}
7030 
7031 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
7032 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
7033 	    SADB_CHECKHDR(mhp, SADB_EXT_PROPOSAL)) {
7034 		ipseclog((LOG_DEBUG,
7035 		    "%s: invalid message: missing required header.\n",
7036 		    __func__));
7037 		return key_senderror(so, m, EINVAL);
7038 	}
7039 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
7040 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
7041 	    SADB_CHECKLEN(mhp, SADB_EXT_PROPOSAL)) {
7042 		ipseclog((LOG_DEBUG,
7043 		    "%s: invalid message: wrong header size.\n", __func__));
7044 		return key_senderror(so, m, EINVAL);
7045 	}
7046 
7047 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
7048 		mode = IPSEC_MODE_ANY;
7049 		reqid = 0;
7050 	} else {
7051 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
7052 			ipseclog((LOG_DEBUG,
7053 			    "%s: invalid message: wrong header size.\n",
7054 			    __func__));
7055 			return key_senderror(so, m, EINVAL);
7056 		}
7057 		mode = ((struct sadb_x_sa2 *)
7058 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
7059 		reqid = ((struct sadb_x_sa2 *)
7060 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
7061 	}
7062 
7063 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
7064 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
7065 
7066 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
7067 	    (struct sockaddr *)(dst0 + 1));
7068 	if (error != 0) {
7069 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
7070 		return key_senderror(so, m, EINVAL);
7071 	}
7072 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
7073 
7074 	/* get a SA index */
7075 	SAHTREE_RLOCK();
7076 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
7077 		if (key_cmpsaidx(&sah->saidx, &saidx, CMP_MODE_REQID))
7078 			break;
7079 	}
7080 	SAHTREE_RUNLOCK();
7081 	if (sah != NULL) {
7082 		ipseclog((LOG_DEBUG, "%s: a SA exists already.\n", __func__));
7083 		return key_senderror(so, m, EEXIST);
7084 	}
7085 
7086 	error = key_acquire(&saidx, NULL);
7087 	if (error != 0) {
7088 		ipseclog((LOG_DEBUG,
7089 		    "%s: error %d returned from key_acquire()\n",
7090 			__func__, error));
7091 		return key_senderror(so, m, error);
7092 	}
7093 	m_freem(m);
7094 	return (0);
7095 }
7096 
7097 /*
7098  * SADB_REGISTER processing.
7099  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
7100  * receive
7101  *   <base>
7102  * from the ikmpd, and register a socket to send PF_KEY messages,
7103  * and send
7104  *   <base, supported>
7105  * to KMD by PF_KEY.
7106  * If socket is detached, must free from regnode.
7107  *
7108  * m will always be freed.
7109  */
7110 static int
7111 key_register(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7112 {
7113 	struct secreg *reg, *newreg = NULL;
7114 
7115 	IPSEC_ASSERT(so != NULL, ("null socket"));
7116 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7117 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7118 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7119 
7120 	/* check for invalid register message */
7121 	if (mhp->msg->sadb_msg_satype >= sizeof(V_regtree)/sizeof(V_regtree[0]))
7122 		return key_senderror(so, m, EINVAL);
7123 
7124 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
7125 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
7126 		goto setmsg;
7127 
7128 	/* check whether existing or not */
7129 	REGTREE_LOCK();
7130 	LIST_FOREACH(reg, &V_regtree[mhp->msg->sadb_msg_satype], chain) {
7131 		if (reg->so == so) {
7132 			REGTREE_UNLOCK();
7133 			ipseclog((LOG_DEBUG, "%s: socket exists already.\n",
7134 				__func__));
7135 			return key_senderror(so, m, EEXIST);
7136 		}
7137 	}
7138 
7139 	/* create regnode */
7140 	newreg =  malloc(sizeof(struct secreg), M_IPSEC_SAR, M_NOWAIT|M_ZERO);
7141 	if (newreg == NULL) {
7142 		REGTREE_UNLOCK();
7143 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7144 		return key_senderror(so, m, ENOBUFS);
7145 	}
7146 
7147 	newreg->so = so;
7148 	((struct keycb *)sotorawcb(so))->kp_registered++;
7149 
7150 	/* add regnode to regtree. */
7151 	LIST_INSERT_HEAD(&V_regtree[mhp->msg->sadb_msg_satype], newreg, chain);
7152 	REGTREE_UNLOCK();
7153 
7154   setmsg:
7155     {
7156 	struct mbuf *n;
7157 	struct sadb_msg *newmsg;
7158 	struct sadb_supported *sup;
7159 	u_int len, alen, elen;
7160 	int off;
7161 	int i;
7162 	struct sadb_alg *alg;
7163 
7164 	/* create new sadb_msg to reply. */
7165 	alen = 0;
7166 	for (i = 1; i <= SADB_AALG_MAX; i++) {
7167 		if (auth_algorithm_lookup(i))
7168 			alen += sizeof(struct sadb_alg);
7169 	}
7170 	if (alen)
7171 		alen += sizeof(struct sadb_supported);
7172 	elen = 0;
7173 	for (i = 1; i <= SADB_EALG_MAX; i++) {
7174 		if (enc_algorithm_lookup(i))
7175 			elen += sizeof(struct sadb_alg);
7176 	}
7177 	if (elen)
7178 		elen += sizeof(struct sadb_supported);
7179 
7180 	len = sizeof(struct sadb_msg) + alen + elen;
7181 
7182 	if (len > MCLBYTES)
7183 		return key_senderror(so, m, ENOBUFS);
7184 
7185 	MGETHDR(n, M_NOWAIT, MT_DATA);
7186 	if (len > MHLEN) {
7187 		if (!(MCLGET(n, M_NOWAIT))) {
7188 			m_freem(n);
7189 			n = NULL;
7190 		}
7191 	}
7192 	if (!n)
7193 		return key_senderror(so, m, ENOBUFS);
7194 
7195 	n->m_pkthdr.len = n->m_len = len;
7196 	n->m_next = NULL;
7197 	off = 0;
7198 
7199 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
7200 	newmsg = mtod(n, struct sadb_msg *);
7201 	newmsg->sadb_msg_errno = 0;
7202 	newmsg->sadb_msg_len = PFKEY_UNIT64(len);
7203 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7204 
7205 	/* for authentication algorithm */
7206 	if (alen) {
7207 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7208 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7209 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7210 		off += PFKEY_ALIGN8(sizeof(*sup));
7211 
7212 		for (i = 1; i <= SADB_AALG_MAX; i++) {
7213 			const struct auth_hash *aalgo;
7214 			u_int16_t minkeysize, maxkeysize;
7215 
7216 			aalgo = auth_algorithm_lookup(i);
7217 			if (!aalgo)
7218 				continue;
7219 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7220 			alg->sadb_alg_id = i;
7221 			alg->sadb_alg_ivlen = 0;
7222 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7223 			alg->sadb_alg_minbits = _BITS(minkeysize);
7224 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7225 			off += PFKEY_ALIGN8(sizeof(*alg));
7226 		}
7227 	}
7228 
7229 	/* for encryption algorithm */
7230 	if (elen) {
7231 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7232 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7233 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7234 		off += PFKEY_ALIGN8(sizeof(*sup));
7235 
7236 		for (i = 1; i <= SADB_EALG_MAX; i++) {
7237 			const struct enc_xform *ealgo;
7238 
7239 			ealgo = enc_algorithm_lookup(i);
7240 			if (!ealgo)
7241 				continue;
7242 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7243 			alg->sadb_alg_id = i;
7244 			alg->sadb_alg_ivlen = ealgo->ivsize;
7245 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7246 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7247 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7248 		}
7249 	}
7250 
7251 	IPSEC_ASSERT(off == len,
7252 		("length assumption failed (off %u len %u)", off, len));
7253 
7254 	m_freem(m);
7255 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7256     }
7257 }
7258 
7259 /*
7260  * free secreg entry registered.
7261  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7262  */
7263 void
7264 key_freereg(struct socket *so)
7265 {
7266 	struct secreg *reg;
7267 	int i;
7268 
7269 	IPSEC_ASSERT(so != NULL, ("NULL so"));
7270 
7271 	/*
7272 	 * check whether existing or not.
7273 	 * check all type of SA, because there is a potential that
7274 	 * one socket is registered to multiple type of SA.
7275 	 */
7276 	REGTREE_LOCK();
7277 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7278 		LIST_FOREACH(reg, &V_regtree[i], chain) {
7279 			if (reg->so == so && __LIST_CHAINED(reg)) {
7280 				LIST_REMOVE(reg, chain);
7281 				free(reg, M_IPSEC_SAR);
7282 				break;
7283 			}
7284 		}
7285 	}
7286 	REGTREE_UNLOCK();
7287 }
7288 
7289 /*
7290  * SADB_EXPIRE processing
7291  * send
7292  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7293  * to KMD by PF_KEY.
7294  * NOTE: We send only soft lifetime extension.
7295  *
7296  * OUT:	0	: succeed
7297  *	others	: error number
7298  */
7299 static int
7300 key_expire(struct secasvar *sav, int hard)
7301 {
7302 	struct mbuf *result = NULL, *m;
7303 	struct sadb_lifetime *lt;
7304 	uint32_t replay_count;
7305 	int error, len;
7306 	uint8_t satype;
7307 
7308 	IPSEC_ASSERT (sav != NULL, ("null sav"));
7309 	IPSEC_ASSERT (sav->sah != NULL, ("null sa header"));
7310 
7311 	KEYDBG(KEY_STAMP,
7312 	    printf("%s: SA(%p) expired %s lifetime\n", __func__,
7313 		sav, hard ? "hard": "soft"));
7314 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
7315 	/* set msg header */
7316 	satype = key_proto2satype(sav->sah->saidx.proto);
7317 	IPSEC_ASSERT(satype != 0, ("invalid proto, satype %u", satype));
7318 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, sav->refcnt);
7319 	if (!m) {
7320 		error = ENOBUFS;
7321 		goto fail;
7322 	}
7323 	result = m;
7324 
7325 	/* create SA extension */
7326 	m = key_setsadbsa(sav);
7327 	if (!m) {
7328 		error = ENOBUFS;
7329 		goto fail;
7330 	}
7331 	m_cat(result, m);
7332 
7333 	/* create SA extension */
7334 	SECASVAR_LOCK(sav);
7335 	replay_count = sav->replay ? sav->replay->count : 0;
7336 	SECASVAR_UNLOCK(sav);
7337 
7338 	m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
7339 			sav->sah->saidx.reqid);
7340 	if (!m) {
7341 		error = ENOBUFS;
7342 		goto fail;
7343 	}
7344 	m_cat(result, m);
7345 
7346 	if (sav->replay && sav->replay->wsize > UINT8_MAX) {
7347 		m = key_setsadbxsareplay(sav->replay->wsize);
7348 		if (!m) {
7349 			error = ENOBUFS;
7350 			goto fail;
7351 		}
7352 		m_cat(result, m);
7353 	}
7354 
7355 	/* create lifetime extension (current and soft) */
7356 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7357 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
7358 	if (m == NULL) {
7359 		error = ENOBUFS;
7360 		goto fail;
7361 	}
7362 	m_align(m, len);
7363 	m->m_len = len;
7364 	bzero(mtod(m, caddr_t), len);
7365 	lt = mtod(m, struct sadb_lifetime *);
7366 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7367 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7368 	lt->sadb_lifetime_allocations =
7369 	    (uint32_t)counter_u64_fetch(sav->lft_c_allocations);
7370 	lt->sadb_lifetime_bytes =
7371 	    counter_u64_fetch(sav->lft_c_bytes);
7372 	lt->sadb_lifetime_addtime = sav->created;
7373 	lt->sadb_lifetime_usetime = sav->firstused;
7374 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
7375 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7376 	if (hard) {
7377 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
7378 		lt->sadb_lifetime_allocations = sav->lft_h->allocations;
7379 		lt->sadb_lifetime_bytes = sav->lft_h->bytes;
7380 		lt->sadb_lifetime_addtime = sav->lft_h->addtime;
7381 		lt->sadb_lifetime_usetime = sav->lft_h->usetime;
7382 	} else {
7383 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_SOFT;
7384 		lt->sadb_lifetime_allocations = sav->lft_s->allocations;
7385 		lt->sadb_lifetime_bytes = sav->lft_s->bytes;
7386 		lt->sadb_lifetime_addtime = sav->lft_s->addtime;
7387 		lt->sadb_lifetime_usetime = sav->lft_s->usetime;
7388 	}
7389 	m_cat(result, m);
7390 
7391 	/* set sadb_address for source */
7392 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
7393 	    &sav->sah->saidx.src.sa,
7394 	    FULLMASK, IPSEC_ULPROTO_ANY);
7395 	if (!m) {
7396 		error = ENOBUFS;
7397 		goto fail;
7398 	}
7399 	m_cat(result, m);
7400 
7401 	/* set sadb_address for destination */
7402 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
7403 	    &sav->sah->saidx.dst.sa,
7404 	    FULLMASK, IPSEC_ULPROTO_ANY);
7405 	if (!m) {
7406 		error = ENOBUFS;
7407 		goto fail;
7408 	}
7409 	m_cat(result, m);
7410 
7411 	/*
7412 	 * XXX-BZ Handle NAT-T extensions here.
7413 	 * XXXAE: it doesn't seem quite useful. IKEs should not depend on
7414 	 * this information, we report only significant SA fields.
7415 	 */
7416 
7417 	if ((result->m_flags & M_PKTHDR) == 0) {
7418 		error = EINVAL;
7419 		goto fail;
7420 	}
7421 
7422 	if (result->m_len < sizeof(struct sadb_msg)) {
7423 		result = m_pullup(result, sizeof(struct sadb_msg));
7424 		if (result == NULL) {
7425 			error = ENOBUFS;
7426 			goto fail;
7427 		}
7428 	}
7429 
7430 	result->m_pkthdr.len = 0;
7431 	for (m = result; m; m = m->m_next)
7432 		result->m_pkthdr.len += m->m_len;
7433 
7434 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7435 	    PFKEY_UNIT64(result->m_pkthdr.len);
7436 
7437 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7438 
7439  fail:
7440 	if (result)
7441 		m_freem(result);
7442 	return error;
7443 }
7444 
7445 static void
7446 key_freesah_flushed(struct secashead_queue *flushq)
7447 {
7448 	struct secashead *sah, *nextsah;
7449 	struct secasvar *sav, *nextsav;
7450 
7451 	sah = TAILQ_FIRST(flushq);
7452 	while (sah != NULL) {
7453 		sav = TAILQ_FIRST(&sah->savtree_larval);
7454 		while (sav != NULL) {
7455 			nextsav = TAILQ_NEXT(sav, chain);
7456 			TAILQ_REMOVE(&sah->savtree_larval, sav, chain);
7457 			key_freesav(&sav); /* release last reference */
7458 			key_freesah(&sah); /* release reference from SAV */
7459 			sav = nextsav;
7460 		}
7461 		sav = TAILQ_FIRST(&sah->savtree_alive);
7462 		while (sav != NULL) {
7463 			nextsav = TAILQ_NEXT(sav, chain);
7464 			TAILQ_REMOVE(&sah->savtree_alive, sav, chain);
7465 			key_freesav(&sav); /* release last reference */
7466 			key_freesah(&sah); /* release reference from SAV */
7467 			sav = nextsav;
7468 		}
7469 		nextsah = TAILQ_NEXT(sah, chain);
7470 		key_freesah(&sah);	/* release last reference */
7471 		sah = nextsah;
7472 	}
7473 }
7474 
7475 /*
7476  * SADB_FLUSH processing
7477  * receive
7478  *   <base>
7479  * from the ikmpd, and free all entries in secastree.
7480  * and send,
7481  *   <base>
7482  * to the ikmpd.
7483  * NOTE: to do is only marking SADB_SASTATE_DEAD.
7484  *
7485  * m will always be freed.
7486  */
7487 static int
7488 key_flush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7489 {
7490 	struct secashead_queue flushq;
7491 	struct sadb_msg *newmsg;
7492 	struct secashead *sah, *nextsah;
7493 	struct secasvar *sav;
7494 	uint8_t proto;
7495 	int i;
7496 
7497 	IPSEC_ASSERT(so != NULL, ("null socket"));
7498 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7499 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7500 
7501 	/* map satype to proto */
7502 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7503 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7504 			__func__));
7505 		return key_senderror(so, m, EINVAL);
7506 	}
7507 	KEYDBG(KEY_STAMP,
7508 	    printf("%s: proto %u\n", __func__, proto));
7509 
7510 	TAILQ_INIT(&flushq);
7511 	if (proto == IPSEC_PROTO_ANY) {
7512 		/* no SATYPE specified, i.e. flushing all SA. */
7513 		SAHTREE_WLOCK();
7514 		/* Move all SAHs into flushq */
7515 		TAILQ_CONCAT(&flushq, &V_sahtree, chain);
7516 		/* Flush all buckets in SPI hash */
7517 		for (i = 0; i < V_savhash_mask + 1; i++)
7518 			LIST_INIT(&V_savhashtbl[i]);
7519 		/* Flush all buckets in SAHADDRHASH */
7520 		for (i = 0; i < V_sahaddrhash_mask + 1; i++)
7521 			LIST_INIT(&V_sahaddrhashtbl[i]);
7522 		/* Mark all SAHs as unlinked */
7523 		TAILQ_FOREACH(sah, &flushq, chain) {
7524 			sah->state = SADB_SASTATE_DEAD;
7525 			/*
7526 			 * Callout handler makes its job using
7527 			 * RLOCK and drain queues. In case, when this
7528 			 * function will be called just before it
7529 			 * acquires WLOCK, we need to mark SAs as
7530 			 * unlinked to prevent second unlink.
7531 			 */
7532 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7533 				sav->state = SADB_SASTATE_DEAD;
7534 			}
7535 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7536 				sav->state = SADB_SASTATE_DEAD;
7537 			}
7538 		}
7539 		SAHTREE_WUNLOCK();
7540 	} else {
7541 		SAHTREE_WLOCK();
7542 		sah = TAILQ_FIRST(&V_sahtree);
7543 		while (sah != NULL) {
7544 			IPSEC_ASSERT(sah->state != SADB_SASTATE_DEAD,
7545 			    ("DEAD SAH %p in SADB_FLUSH", sah));
7546 			nextsah = TAILQ_NEXT(sah, chain);
7547 			if (sah->saidx.proto != proto) {
7548 				sah = nextsah;
7549 				continue;
7550 			}
7551 			sah->state = SADB_SASTATE_DEAD;
7552 			TAILQ_REMOVE(&V_sahtree, sah, chain);
7553 			LIST_REMOVE(sah, addrhash);
7554 			/* Unlink all SAs from SPI hash */
7555 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7556 				LIST_REMOVE(sav, spihash);
7557 				sav->state = SADB_SASTATE_DEAD;
7558 			}
7559 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7560 				LIST_REMOVE(sav, spihash);
7561 				sav->state = SADB_SASTATE_DEAD;
7562 			}
7563 			/* Add SAH into flushq */
7564 			TAILQ_INSERT_HEAD(&flushq, sah, chain);
7565 			sah = nextsah;
7566 		}
7567 		SAHTREE_WUNLOCK();
7568 	}
7569 
7570 	key_freesah_flushed(&flushq);
7571 	/* Free all queued SAs and SAHs */
7572 	if (m->m_len < sizeof(struct sadb_msg) ||
7573 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7574 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7575 		return key_senderror(so, m, ENOBUFS);
7576 	}
7577 
7578 	if (m->m_next)
7579 		m_freem(m->m_next);
7580 	m->m_next = NULL;
7581 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7582 	newmsg = mtod(m, struct sadb_msg *);
7583 	newmsg->sadb_msg_errno = 0;
7584 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7585 
7586 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7587 }
7588 
7589 /*
7590  * SADB_DUMP processing
7591  * dump all entries including status of DEAD in SAD.
7592  * receive
7593  *   <base>
7594  * from the ikmpd, and dump all secasvar leaves
7595  * and send,
7596  *   <base> .....
7597  * to the ikmpd.
7598  *
7599  * m will always be freed.
7600  */
7601 static int
7602 key_dump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7603 {
7604 	SAHTREE_RLOCK_TRACKER;
7605 	struct secashead *sah;
7606 	struct secasvar *sav;
7607 	struct mbuf *n;
7608 	uint32_t cnt;
7609 	uint8_t proto, satype;
7610 
7611 	IPSEC_ASSERT(so != NULL, ("null socket"));
7612 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7613 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7614 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7615 
7616 	/* map satype to proto */
7617 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7618 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7619 		    __func__));
7620 		return key_senderror(so, m, EINVAL);
7621 	}
7622 
7623 	/* count sav entries to be sent to the userland. */
7624 	cnt = 0;
7625 	SAHTREE_RLOCK();
7626 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7627 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7628 		    proto != sah->saidx.proto)
7629 			continue;
7630 
7631 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain)
7632 			cnt++;
7633 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain)
7634 			cnt++;
7635 	}
7636 
7637 	if (cnt == 0) {
7638 		SAHTREE_RUNLOCK();
7639 		return key_senderror(so, m, ENOENT);
7640 	}
7641 
7642 	/* send this to the userland, one at a time. */
7643 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7644 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7645 		    proto != sah->saidx.proto)
7646 			continue;
7647 
7648 		/* map proto to satype */
7649 		if ((satype = key_proto2satype(sah->saidx.proto)) == 0) {
7650 			SAHTREE_RUNLOCK();
7651 			ipseclog((LOG_DEBUG, "%s: there was invalid proto in "
7652 			    "SAD.\n", __func__));
7653 			return key_senderror(so, m, EINVAL);
7654 		}
7655 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7656 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7657 			    --cnt, mhp->msg->sadb_msg_pid);
7658 			if (n == NULL) {
7659 				SAHTREE_RUNLOCK();
7660 				return key_senderror(so, m, ENOBUFS);
7661 			}
7662 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7663 		}
7664 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7665 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7666 			    --cnt, mhp->msg->sadb_msg_pid);
7667 			if (n == NULL) {
7668 				SAHTREE_RUNLOCK();
7669 				return key_senderror(so, m, ENOBUFS);
7670 			}
7671 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7672 		}
7673 	}
7674 	SAHTREE_RUNLOCK();
7675 	m_freem(m);
7676 	return (0);
7677 }
7678 /*
7679  * SADB_X_PROMISC processing
7680  *
7681  * m will always be freed.
7682  */
7683 static int
7684 key_promisc(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7685 {
7686 	int olen;
7687 
7688 	IPSEC_ASSERT(so != NULL, ("null socket"));
7689 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7690 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7691 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7692 
7693 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7694 
7695 	if (olen < sizeof(struct sadb_msg)) {
7696 #if 1
7697 		return key_senderror(so, m, EINVAL);
7698 #else
7699 		m_freem(m);
7700 		return 0;
7701 #endif
7702 	} else if (olen == sizeof(struct sadb_msg)) {
7703 		/* enable/disable promisc mode */
7704 		struct keycb *kp;
7705 
7706 		if ((kp = (struct keycb *)sotorawcb(so)) == NULL)
7707 			return key_senderror(so, m, EINVAL);
7708 		mhp->msg->sadb_msg_errno = 0;
7709 		switch (mhp->msg->sadb_msg_satype) {
7710 		case 0:
7711 		case 1:
7712 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
7713 			break;
7714 		default:
7715 			return key_senderror(so, m, EINVAL);
7716 		}
7717 
7718 		/* send the original message back to everyone */
7719 		mhp->msg->sadb_msg_errno = 0;
7720 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7721 	} else {
7722 		/* send packet as is */
7723 
7724 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
7725 
7726 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
7727 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7728 	}
7729 }
7730 
7731 static int (*key_typesw[])(struct socket *, struct mbuf *,
7732 		const struct sadb_msghdr *) = {
7733 	NULL,		/* SADB_RESERVED */
7734 	key_getspi,	/* SADB_GETSPI */
7735 	key_update,	/* SADB_UPDATE */
7736 	key_add,	/* SADB_ADD */
7737 	key_delete,	/* SADB_DELETE */
7738 	key_get,	/* SADB_GET */
7739 	key_acquire2,	/* SADB_ACQUIRE */
7740 	key_register,	/* SADB_REGISTER */
7741 	NULL,		/* SADB_EXPIRE */
7742 	key_flush,	/* SADB_FLUSH */
7743 	key_dump,	/* SADB_DUMP */
7744 	key_promisc,	/* SADB_X_PROMISC */
7745 	NULL,		/* SADB_X_PCHANGE */
7746 	key_spdadd,	/* SADB_X_SPDUPDATE */
7747 	key_spdadd,	/* SADB_X_SPDADD */
7748 	key_spddelete,	/* SADB_X_SPDDELETE */
7749 	key_spdget,	/* SADB_X_SPDGET */
7750 	NULL,		/* SADB_X_SPDACQUIRE */
7751 	key_spddump,	/* SADB_X_SPDDUMP */
7752 	key_spdflush,	/* SADB_X_SPDFLUSH */
7753 	key_spdadd,	/* SADB_X_SPDSETIDX */
7754 	NULL,		/* SADB_X_SPDEXPIRE */
7755 	key_spddelete2,	/* SADB_X_SPDDELETE2 */
7756 };
7757 
7758 /*
7759  * parse sadb_msg buffer to process PFKEYv2,
7760  * and create a data to response if needed.
7761  * I think to be dealed with mbuf directly.
7762  * IN:
7763  *     msgp  : pointer to pointer to a received buffer pulluped.
7764  *             This is rewrited to response.
7765  *     so    : pointer to socket.
7766  * OUT:
7767  *    length for buffer to send to user process.
7768  */
7769 int
7770 key_parse(struct mbuf *m, struct socket *so)
7771 {
7772 	struct sadb_msg *msg;
7773 	struct sadb_msghdr mh;
7774 	u_int orglen;
7775 	int error;
7776 	int target;
7777 
7778 	IPSEC_ASSERT(so != NULL, ("null socket"));
7779 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7780 
7781 	if (m->m_len < sizeof(struct sadb_msg)) {
7782 		m = m_pullup(m, sizeof(struct sadb_msg));
7783 		if (!m)
7784 			return ENOBUFS;
7785 	}
7786 	msg = mtod(m, struct sadb_msg *);
7787 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
7788 	target = KEY_SENDUP_ONE;
7789 
7790 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len != orglen) {
7791 		ipseclog((LOG_DEBUG, "%s: invalid message length.\n",__func__));
7792 		PFKEYSTAT_INC(out_invlen);
7793 		error = EINVAL;
7794 		goto senderror;
7795 	}
7796 
7797 	if (msg->sadb_msg_version != PF_KEY_V2) {
7798 		ipseclog((LOG_DEBUG, "%s: PF_KEY version %u is mismatched.\n",
7799 		    __func__, msg->sadb_msg_version));
7800 		PFKEYSTAT_INC(out_invver);
7801 		error = EINVAL;
7802 		goto senderror;
7803 	}
7804 
7805 	if (msg->sadb_msg_type > SADB_MAX) {
7806 		ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7807 		    __func__, msg->sadb_msg_type));
7808 		PFKEYSTAT_INC(out_invmsgtype);
7809 		error = EINVAL;
7810 		goto senderror;
7811 	}
7812 
7813 	/* for old-fashioned code - should be nuked */
7814 	if (m->m_pkthdr.len > MCLBYTES) {
7815 		m_freem(m);
7816 		return ENOBUFS;
7817 	}
7818 	if (m->m_next) {
7819 		struct mbuf *n;
7820 
7821 		MGETHDR(n, M_NOWAIT, MT_DATA);
7822 		if (n && m->m_pkthdr.len > MHLEN) {
7823 			if (!(MCLGET(n, M_NOWAIT))) {
7824 				m_free(n);
7825 				n = NULL;
7826 			}
7827 		}
7828 		if (!n) {
7829 			m_freem(m);
7830 			return ENOBUFS;
7831 		}
7832 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
7833 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
7834 		n->m_next = NULL;
7835 		m_freem(m);
7836 		m = n;
7837 	}
7838 
7839 	/* align the mbuf chain so that extensions are in contiguous region. */
7840 	error = key_align(m, &mh);
7841 	if (error)
7842 		return error;
7843 
7844 	msg = mh.msg;
7845 
7846 	/* We use satype as scope mask for spddump */
7847 	if (msg->sadb_msg_type == SADB_X_SPDDUMP) {
7848 		switch (msg->sadb_msg_satype) {
7849 		case IPSEC_POLICYSCOPE_ANY:
7850 		case IPSEC_POLICYSCOPE_GLOBAL:
7851 		case IPSEC_POLICYSCOPE_IFNET:
7852 		case IPSEC_POLICYSCOPE_PCB:
7853 			break;
7854 		default:
7855 			ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7856 			    __func__, msg->sadb_msg_type));
7857 			PFKEYSTAT_INC(out_invsatype);
7858 			error = EINVAL;
7859 			goto senderror;
7860 		}
7861 	} else {
7862 		switch (msg->sadb_msg_satype) { /* check SA type */
7863 		case SADB_SATYPE_UNSPEC:
7864 			switch (msg->sadb_msg_type) {
7865 			case SADB_GETSPI:
7866 			case SADB_UPDATE:
7867 			case SADB_ADD:
7868 			case SADB_DELETE:
7869 			case SADB_GET:
7870 			case SADB_ACQUIRE:
7871 			case SADB_EXPIRE:
7872 				ipseclog((LOG_DEBUG, "%s: must specify satype "
7873 				    "when msg type=%u.\n", __func__,
7874 				    msg->sadb_msg_type));
7875 				PFKEYSTAT_INC(out_invsatype);
7876 				error = EINVAL;
7877 				goto senderror;
7878 			}
7879 			break;
7880 		case SADB_SATYPE_AH:
7881 		case SADB_SATYPE_ESP:
7882 		case SADB_X_SATYPE_IPCOMP:
7883 		case SADB_X_SATYPE_TCPSIGNATURE:
7884 			switch (msg->sadb_msg_type) {
7885 			case SADB_X_SPDADD:
7886 			case SADB_X_SPDDELETE:
7887 			case SADB_X_SPDGET:
7888 			case SADB_X_SPDFLUSH:
7889 			case SADB_X_SPDSETIDX:
7890 			case SADB_X_SPDUPDATE:
7891 			case SADB_X_SPDDELETE2:
7892 				ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7893 				    __func__, msg->sadb_msg_type));
7894 				PFKEYSTAT_INC(out_invsatype);
7895 				error = EINVAL;
7896 				goto senderror;
7897 			}
7898 			break;
7899 		case SADB_SATYPE_RSVP:
7900 		case SADB_SATYPE_OSPFV2:
7901 		case SADB_SATYPE_RIPV2:
7902 		case SADB_SATYPE_MIP:
7903 			ipseclog((LOG_DEBUG, "%s: type %u isn't supported.\n",
7904 			    __func__, msg->sadb_msg_satype));
7905 			PFKEYSTAT_INC(out_invsatype);
7906 			error = EOPNOTSUPP;
7907 			goto senderror;
7908 		case 1:	/* XXX: What does it do? */
7909 			if (msg->sadb_msg_type == SADB_X_PROMISC)
7910 				break;
7911 			/*FALLTHROUGH*/
7912 		default:
7913 			ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7914 			    __func__, msg->sadb_msg_satype));
7915 			PFKEYSTAT_INC(out_invsatype);
7916 			error = EINVAL;
7917 			goto senderror;
7918 		}
7919 	}
7920 
7921 	/* check field of upper layer protocol and address family */
7922 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL
7923 	 && mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
7924 		struct sadb_address *src0, *dst0;
7925 		u_int plen;
7926 
7927 		src0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_SRC]);
7928 		dst0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_DST]);
7929 
7930 		/* check upper layer protocol */
7931 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
7932 			ipseclog((LOG_DEBUG, "%s: upper layer protocol "
7933 				"mismatched.\n", __func__));
7934 			PFKEYSTAT_INC(out_invaddr);
7935 			error = EINVAL;
7936 			goto senderror;
7937 		}
7938 
7939 		/* check family */
7940 		if (PFKEY_ADDR_SADDR(src0)->sa_family !=
7941 		    PFKEY_ADDR_SADDR(dst0)->sa_family) {
7942 			ipseclog((LOG_DEBUG, "%s: address family mismatched.\n",
7943 				__func__));
7944 			PFKEYSTAT_INC(out_invaddr);
7945 			error = EINVAL;
7946 			goto senderror;
7947 		}
7948 		if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7949 		    PFKEY_ADDR_SADDR(dst0)->sa_len) {
7950 			ipseclog((LOG_DEBUG, "%s: address struct size "
7951 				"mismatched.\n", __func__));
7952 			PFKEYSTAT_INC(out_invaddr);
7953 			error = EINVAL;
7954 			goto senderror;
7955 		}
7956 
7957 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7958 		case AF_INET:
7959 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7960 			    sizeof(struct sockaddr_in)) {
7961 				PFKEYSTAT_INC(out_invaddr);
7962 				error = EINVAL;
7963 				goto senderror;
7964 			}
7965 			break;
7966 		case AF_INET6:
7967 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7968 			    sizeof(struct sockaddr_in6)) {
7969 				PFKEYSTAT_INC(out_invaddr);
7970 				error = EINVAL;
7971 				goto senderror;
7972 			}
7973 			break;
7974 		default:
7975 			ipseclog((LOG_DEBUG, "%s: unsupported address family\n",
7976 				__func__));
7977 			PFKEYSTAT_INC(out_invaddr);
7978 			error = EAFNOSUPPORT;
7979 			goto senderror;
7980 		}
7981 
7982 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7983 		case AF_INET:
7984 			plen = sizeof(struct in_addr) << 3;
7985 			break;
7986 		case AF_INET6:
7987 			plen = sizeof(struct in6_addr) << 3;
7988 			break;
7989 		default:
7990 			plen = 0;	/*fool gcc*/
7991 			break;
7992 		}
7993 
7994 		/* check max prefix length */
7995 		if (src0->sadb_address_prefixlen > plen ||
7996 		    dst0->sadb_address_prefixlen > plen) {
7997 			ipseclog((LOG_DEBUG, "%s: illegal prefixlen.\n",
7998 				__func__));
7999 			PFKEYSTAT_INC(out_invaddr);
8000 			error = EINVAL;
8001 			goto senderror;
8002 		}
8003 
8004 		/*
8005 		 * prefixlen == 0 is valid because there can be a case when
8006 		 * all addresses are matched.
8007 		 */
8008 	}
8009 
8010 	if (msg->sadb_msg_type >= nitems(key_typesw) ||
8011 	    key_typesw[msg->sadb_msg_type] == NULL) {
8012 		PFKEYSTAT_INC(out_invmsgtype);
8013 		error = EINVAL;
8014 		goto senderror;
8015 	}
8016 
8017 	return (*key_typesw[msg->sadb_msg_type])(so, m, &mh);
8018 
8019 senderror:
8020 	msg->sadb_msg_errno = error;
8021 	return key_sendup_mbuf(so, m, target);
8022 }
8023 
8024 static int
8025 key_senderror(struct socket *so, struct mbuf *m, int code)
8026 {
8027 	struct sadb_msg *msg;
8028 
8029 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8030 		("mbuf too small, len %u", m->m_len));
8031 
8032 	msg = mtod(m, struct sadb_msg *);
8033 	msg->sadb_msg_errno = code;
8034 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
8035 }
8036 
8037 /*
8038  * set the pointer to each header into message buffer.
8039  * m will be freed on error.
8040  * XXX larger-than-MCLBYTES extension?
8041  */
8042 static int
8043 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
8044 {
8045 	struct mbuf *n;
8046 	struct sadb_ext *ext;
8047 	size_t off, end;
8048 	int extlen;
8049 	int toff;
8050 
8051 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
8052 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
8053 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8054 		("mbuf too small, len %u", m->m_len));
8055 
8056 	/* initialize */
8057 	bzero(mhp, sizeof(*mhp));
8058 
8059 	mhp->msg = mtod(m, struct sadb_msg *);
8060 	mhp->ext[0] = (struct sadb_ext *)mhp->msg;	/*XXX backward compat */
8061 
8062 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
8063 	extlen = end;	/*just in case extlen is not updated*/
8064 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
8065 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
8066 		if (!n) {
8067 			/* m is already freed */
8068 			return ENOBUFS;
8069 		}
8070 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8071 
8072 		/* set pointer */
8073 		switch (ext->sadb_ext_type) {
8074 		case SADB_EXT_SA:
8075 		case SADB_EXT_ADDRESS_SRC:
8076 		case SADB_EXT_ADDRESS_DST:
8077 		case SADB_EXT_ADDRESS_PROXY:
8078 		case SADB_EXT_LIFETIME_CURRENT:
8079 		case SADB_EXT_LIFETIME_HARD:
8080 		case SADB_EXT_LIFETIME_SOFT:
8081 		case SADB_EXT_KEY_AUTH:
8082 		case SADB_EXT_KEY_ENCRYPT:
8083 		case SADB_EXT_IDENTITY_SRC:
8084 		case SADB_EXT_IDENTITY_DST:
8085 		case SADB_EXT_SENSITIVITY:
8086 		case SADB_EXT_PROPOSAL:
8087 		case SADB_EXT_SUPPORTED_AUTH:
8088 		case SADB_EXT_SUPPORTED_ENCRYPT:
8089 		case SADB_EXT_SPIRANGE:
8090 		case SADB_X_EXT_POLICY:
8091 		case SADB_X_EXT_SA2:
8092 		case SADB_X_EXT_NAT_T_TYPE:
8093 		case SADB_X_EXT_NAT_T_SPORT:
8094 		case SADB_X_EXT_NAT_T_DPORT:
8095 		case SADB_X_EXT_NAT_T_OAI:
8096 		case SADB_X_EXT_NAT_T_OAR:
8097 		case SADB_X_EXT_NAT_T_FRAG:
8098 		case SADB_X_EXT_SA_REPLAY:
8099 		case SADB_X_EXT_NEW_ADDRESS_SRC:
8100 		case SADB_X_EXT_NEW_ADDRESS_DST:
8101 			/* duplicate check */
8102 			/*
8103 			 * XXX Are there duplication payloads of either
8104 			 * KEY_AUTH or KEY_ENCRYPT ?
8105 			 */
8106 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
8107 				ipseclog((LOG_DEBUG, "%s: duplicate ext_type "
8108 					"%u\n", __func__, ext->sadb_ext_type));
8109 				m_freem(m);
8110 				PFKEYSTAT_INC(out_dupext);
8111 				return EINVAL;
8112 			}
8113 			break;
8114 		default:
8115 			ipseclog((LOG_DEBUG, "%s: invalid ext_type %u\n",
8116 				__func__, ext->sadb_ext_type));
8117 			m_freem(m);
8118 			PFKEYSTAT_INC(out_invexttype);
8119 			return EINVAL;
8120 		}
8121 
8122 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
8123 
8124 		if (key_validate_ext(ext, extlen)) {
8125 			m_freem(m);
8126 			PFKEYSTAT_INC(out_invlen);
8127 			return EINVAL;
8128 		}
8129 
8130 		n = m_pulldown(m, off, extlen, &toff);
8131 		if (!n) {
8132 			/* m is already freed */
8133 			return ENOBUFS;
8134 		}
8135 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8136 
8137 		mhp->ext[ext->sadb_ext_type] = ext;
8138 		mhp->extoff[ext->sadb_ext_type] = off;
8139 		mhp->extlen[ext->sadb_ext_type] = extlen;
8140 	}
8141 
8142 	if (off != end) {
8143 		m_freem(m);
8144 		PFKEYSTAT_INC(out_invlen);
8145 		return EINVAL;
8146 	}
8147 
8148 	return 0;
8149 }
8150 
8151 static int
8152 key_validate_ext(const struct sadb_ext *ext, int len)
8153 {
8154 	const struct sockaddr *sa;
8155 	enum { NONE, ADDR } checktype = NONE;
8156 	int baselen = 0;
8157 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
8158 
8159 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
8160 		return EINVAL;
8161 
8162 	/* if it does not match minimum/maximum length, bail */
8163 	if (ext->sadb_ext_type >= nitems(minsize) ||
8164 	    ext->sadb_ext_type >= nitems(maxsize))
8165 		return EINVAL;
8166 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
8167 		return EINVAL;
8168 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
8169 		return EINVAL;
8170 
8171 	/* more checks based on sadb_ext_type XXX need more */
8172 	switch (ext->sadb_ext_type) {
8173 	case SADB_EXT_ADDRESS_SRC:
8174 	case SADB_EXT_ADDRESS_DST:
8175 	case SADB_EXT_ADDRESS_PROXY:
8176 	case SADB_X_EXT_NAT_T_OAI:
8177 	case SADB_X_EXT_NAT_T_OAR:
8178 	case SADB_X_EXT_NEW_ADDRESS_SRC:
8179 	case SADB_X_EXT_NEW_ADDRESS_DST:
8180 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
8181 		checktype = ADDR;
8182 		break;
8183 	case SADB_EXT_IDENTITY_SRC:
8184 	case SADB_EXT_IDENTITY_DST:
8185 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
8186 		    SADB_X_IDENTTYPE_ADDR) {
8187 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
8188 			checktype = ADDR;
8189 		} else
8190 			checktype = NONE;
8191 		break;
8192 	default:
8193 		checktype = NONE;
8194 		break;
8195 	}
8196 
8197 	switch (checktype) {
8198 	case NONE:
8199 		break;
8200 	case ADDR:
8201 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8202 		if (len < baselen + sal)
8203 			return EINVAL;
8204 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8205 			return EINVAL;
8206 		break;
8207 	}
8208 
8209 	return 0;
8210 }
8211 
8212 void
8213 spdcache_init(void)
8214 {
8215 	int i;
8216 
8217 	TUNABLE_INT_FETCH("net.key.spdcache.maxentries",
8218 	    &V_key_spdcache_maxentries);
8219 	TUNABLE_INT_FETCH("net.key.spdcache.threshold",
8220 	    &V_key_spdcache_threshold);
8221 
8222 	if (V_key_spdcache_maxentries) {
8223 		V_key_spdcache_maxentries = MAX(V_key_spdcache_maxentries,
8224 		    SPDCACHE_MAX_ENTRIES_PER_HASH);
8225 		V_spdcachehashtbl = hashinit(V_key_spdcache_maxentries /
8226 		    SPDCACHE_MAX_ENTRIES_PER_HASH,
8227 		    M_IPSEC_SPDCACHE, &V_spdcachehash_mask);
8228 		V_key_spdcache_maxentries = (V_spdcachehash_mask + 1)
8229 		    * SPDCACHE_MAX_ENTRIES_PER_HASH;
8230 
8231 		V_spdcache_lock = malloc(sizeof(struct mtx) *
8232 		    (V_spdcachehash_mask + 1),
8233 		    M_IPSEC_SPDCACHE, M_WAITOK|M_ZERO);
8234 
8235 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8236 			SPDCACHE_LOCK_INIT(i);
8237 	}
8238 }
8239 
8240 struct spdcache_entry *
8241 spdcache_entry_alloc(const struct secpolicyindex *spidx, struct secpolicy *sp)
8242 {
8243 	struct spdcache_entry *entry;
8244 
8245 	entry = malloc(sizeof(struct spdcache_entry),
8246 		    M_IPSEC_SPDCACHE, M_NOWAIT|M_ZERO);
8247 	if (entry == NULL)
8248 		return NULL;
8249 
8250 	if (sp != NULL)
8251 		SP_ADDREF(sp);
8252 
8253 	entry->spidx = *spidx;
8254 	entry->sp = sp;
8255 
8256 	return (entry);
8257 }
8258 
8259 void
8260 spdcache_entry_free(struct spdcache_entry *entry)
8261 {
8262 
8263 	if (entry->sp != NULL)
8264 		key_freesp(&entry->sp);
8265 	free(entry, M_IPSEC_SPDCACHE);
8266 }
8267 
8268 void
8269 spdcache_clear(void)
8270 {
8271 	struct spdcache_entry *entry;
8272 	int i;
8273 
8274 	for (i = 0; i < V_spdcachehash_mask + 1; ++i) {
8275 		SPDCACHE_LOCK(i);
8276 		while (!LIST_EMPTY(&V_spdcachehashtbl[i])) {
8277 			entry = LIST_FIRST(&V_spdcachehashtbl[i]);
8278 			LIST_REMOVE(entry, chain);
8279 			spdcache_entry_free(entry);
8280 		}
8281 		SPDCACHE_UNLOCK(i);
8282 	}
8283 }
8284 
8285 void
8286 spdcache_destroy(void)
8287 {
8288 	int i;
8289 
8290 	if (SPDCACHE_ENABLED()) {
8291 		spdcache_clear();
8292 		hashdestroy(V_spdcachehashtbl, M_IPSEC_SPDCACHE, V_spdcachehash_mask);
8293 
8294 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8295 			SPDCACHE_LOCK_DESTROY(i);
8296 
8297 		free(V_spdcache_lock, M_IPSEC_SPDCACHE);
8298 	}
8299 }
8300 
8301 void
8302 key_init(void)
8303 {
8304 	int i;
8305 
8306 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8307 		TAILQ_INIT(&V_sptree[i]);
8308 		TAILQ_INIT(&V_sptree_ifnet[i]);
8309 	}
8310 
8311 	V_key_lft_zone = uma_zcreate("IPsec SA lft_c",
8312 	    sizeof(uint64_t) * 2, NULL, NULL, NULL, NULL,
8313 	    UMA_ALIGN_PTR, UMA_ZONE_PCPU);
8314 
8315 	TAILQ_INIT(&V_sahtree);
8316 	V_sphashtbl = hashinit(SPHASH_NHASH, M_IPSEC_SP, &V_sphash_mask);
8317 	V_savhashtbl = hashinit(SAVHASH_NHASH, M_IPSEC_SA, &V_savhash_mask);
8318 	V_sahaddrhashtbl = hashinit(SAHHASH_NHASH, M_IPSEC_SAH,
8319 	    &V_sahaddrhash_mask);
8320 	V_acqaddrhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8321 	    &V_acqaddrhash_mask);
8322 	V_acqseqhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8323 	    &V_acqseqhash_mask);
8324 
8325 	spdcache_init();
8326 
8327 	for (i = 0; i <= SADB_SATYPE_MAX; i++)
8328 		LIST_INIT(&V_regtree[i]);
8329 
8330 	LIST_INIT(&V_acqtree);
8331 	LIST_INIT(&V_spacqtree);
8332 
8333 	if (!IS_DEFAULT_VNET(curvnet))
8334 		return;
8335 
8336 	XFORMS_LOCK_INIT();
8337 	SPTREE_LOCK_INIT();
8338 	REGTREE_LOCK_INIT();
8339 	SAHTREE_LOCK_INIT();
8340 	ACQ_LOCK_INIT();
8341 	SPACQ_LOCK_INIT();
8342 
8343 #ifndef IPSEC_DEBUG2
8344 	callout_init(&key_timer, 1);
8345 	callout_reset(&key_timer, hz, key_timehandler, NULL);
8346 #endif /*IPSEC_DEBUG2*/
8347 
8348 	/* initialize key statistics */
8349 	keystat.getspi_count = 1;
8350 
8351 	if (bootverbose)
8352 		printf("IPsec: Initialized Security Association Processing.\n");
8353 }
8354 
8355 #ifdef VIMAGE
8356 void
8357 key_destroy(void)
8358 {
8359 	struct secashead_queue sahdrainq;
8360 	struct secpolicy_queue drainq;
8361 	struct secpolicy *sp, *nextsp;
8362 	struct secacq *acq, *nextacq;
8363 	struct secspacq *spacq, *nextspacq;
8364 	struct secashead *sah;
8365 	struct secasvar *sav;
8366 	struct secreg *reg;
8367 	int i;
8368 
8369 	/*
8370 	 * XXX: can we just call free() for each object without
8371 	 * walking through safe way with releasing references?
8372 	 */
8373 	TAILQ_INIT(&drainq);
8374 	SPTREE_WLOCK();
8375 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8376 		TAILQ_CONCAT(&drainq, &V_sptree[i], chain);
8377 		TAILQ_CONCAT(&drainq, &V_sptree_ifnet[i], chain);
8378 	}
8379 	for (i = 0; i < V_sphash_mask + 1; i++)
8380 		LIST_INIT(&V_sphashtbl[i]);
8381 	SPTREE_WUNLOCK();
8382 	spdcache_destroy();
8383 
8384 	sp = TAILQ_FIRST(&drainq);
8385 	while (sp != NULL) {
8386 		nextsp = TAILQ_NEXT(sp, chain);
8387 		key_freesp(&sp);
8388 		sp = nextsp;
8389 	}
8390 
8391 	TAILQ_INIT(&sahdrainq);
8392 	SAHTREE_WLOCK();
8393 	TAILQ_CONCAT(&sahdrainq, &V_sahtree, chain);
8394 	for (i = 0; i < V_savhash_mask + 1; i++)
8395 		LIST_INIT(&V_savhashtbl[i]);
8396 	for (i = 0; i < V_sahaddrhash_mask + 1; i++)
8397 		LIST_INIT(&V_sahaddrhashtbl[i]);
8398 	TAILQ_FOREACH(sah, &sahdrainq, chain) {
8399 		sah->state = SADB_SASTATE_DEAD;
8400 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
8401 			sav->state = SADB_SASTATE_DEAD;
8402 		}
8403 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
8404 			sav->state = SADB_SASTATE_DEAD;
8405 		}
8406 	}
8407 	SAHTREE_WUNLOCK();
8408 
8409 	key_freesah_flushed(&sahdrainq);
8410 	hashdestroy(V_sphashtbl, M_IPSEC_SP, V_sphash_mask);
8411 	hashdestroy(V_savhashtbl, M_IPSEC_SA, V_savhash_mask);
8412 	hashdestroy(V_sahaddrhashtbl, M_IPSEC_SAH, V_sahaddrhash_mask);
8413 
8414 	REGTREE_LOCK();
8415 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8416 		LIST_FOREACH(reg, &V_regtree[i], chain) {
8417 			if (__LIST_CHAINED(reg)) {
8418 				LIST_REMOVE(reg, chain);
8419 				free(reg, M_IPSEC_SAR);
8420 				break;
8421 			}
8422 		}
8423 	}
8424 	REGTREE_UNLOCK();
8425 
8426 	ACQ_LOCK();
8427 	acq = LIST_FIRST(&V_acqtree);
8428 	while (acq != NULL) {
8429 		nextacq = LIST_NEXT(acq, chain);
8430 		LIST_REMOVE(acq, chain);
8431 		free(acq, M_IPSEC_SAQ);
8432 		acq = nextacq;
8433 	}
8434 	for (i = 0; i < V_acqaddrhash_mask + 1; i++)
8435 		LIST_INIT(&V_acqaddrhashtbl[i]);
8436 	for (i = 0; i < V_acqseqhash_mask + 1; i++)
8437 		LIST_INIT(&V_acqseqhashtbl[i]);
8438 	ACQ_UNLOCK();
8439 
8440 	SPACQ_LOCK();
8441 	for (spacq = LIST_FIRST(&V_spacqtree); spacq != NULL;
8442 	    spacq = nextspacq) {
8443 		nextspacq = LIST_NEXT(spacq, chain);
8444 		if (__LIST_CHAINED(spacq)) {
8445 			LIST_REMOVE(spacq, chain);
8446 			free(spacq, M_IPSEC_SAQ);
8447 		}
8448 	}
8449 	SPACQ_UNLOCK();
8450 	hashdestroy(V_acqaddrhashtbl, M_IPSEC_SAQ, V_acqaddrhash_mask);
8451 	hashdestroy(V_acqseqhashtbl, M_IPSEC_SAQ, V_acqseqhash_mask);
8452 	uma_zdestroy(V_key_lft_zone);
8453 
8454 	if (!IS_DEFAULT_VNET(curvnet))
8455 		return;
8456 #ifndef IPSEC_DEBUG2
8457 	callout_drain(&key_timer);
8458 #endif
8459 	XFORMS_LOCK_DESTROY();
8460 	SPTREE_LOCK_DESTROY();
8461 	REGTREE_LOCK_DESTROY();
8462 	SAHTREE_LOCK_DESTROY();
8463 	ACQ_LOCK_DESTROY();
8464 	SPACQ_LOCK_DESTROY();
8465 }
8466 #endif
8467 
8468 /* record data transfer on SA, and update timestamps */
8469 void
8470 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8471 {
8472 	IPSEC_ASSERT(sav != NULL, ("Null secasvar"));
8473 	IPSEC_ASSERT(m != NULL, ("Null mbuf"));
8474 
8475 	/*
8476 	 * XXX Currently, there is a difference of bytes size
8477 	 * between inbound and outbound processing.
8478 	 */
8479 	counter_u64_add(sav->lft_c_bytes, m->m_pkthdr.len);
8480 
8481 	/*
8482 	 * We use the number of packets as the unit of
8483 	 * allocations.  We increment the variable
8484 	 * whenever {esp,ah}_{in,out}put is called.
8485 	 */
8486 	counter_u64_add(sav->lft_c_allocations, 1);
8487 
8488 	/*
8489 	 * NOTE: We record CURRENT usetime by using wall clock,
8490 	 * in seconds.  HARD and SOFT lifetime are measured by the time
8491 	 * difference (again in seconds) from usetime.
8492 	 *
8493 	 *	usetime
8494 	 *	v     expire   expire
8495 	 * -----+-----+--------+---> t
8496 	 *	<--------------> HARD
8497 	 *	<-----> SOFT
8498 	 */
8499 	if (sav->firstused == 0)
8500 		sav->firstused = time_second;
8501 }
8502 
8503 /*
8504  * Take one of the kernel's security keys and convert it into a PF_KEY
8505  * structure within an mbuf, suitable for sending up to a waiting
8506  * application in user land.
8507  *
8508  * IN:
8509  *    src: A pointer to a kernel security key.
8510  *    exttype: Which type of key this is. Refer to the PF_KEY data structures.
8511  * OUT:
8512  *    a valid mbuf or NULL indicating an error
8513  *
8514  */
8515 
8516 static struct mbuf *
8517 key_setkey(struct seckey *src, uint16_t exttype)
8518 {
8519 	struct mbuf *m;
8520 	struct sadb_key *p;
8521 	int len;
8522 
8523 	if (src == NULL)
8524 		return NULL;
8525 
8526 	len = PFKEY_ALIGN8(sizeof(struct sadb_key) + _KEYLEN(src));
8527 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8528 	if (m == NULL)
8529 		return NULL;
8530 	m_align(m, len);
8531 	m->m_len = len;
8532 	p = mtod(m, struct sadb_key *);
8533 	bzero(p, len);
8534 	p->sadb_key_len = PFKEY_UNIT64(len);
8535 	p->sadb_key_exttype = exttype;
8536 	p->sadb_key_bits = src->bits;
8537 	bcopy(src->key_data, _KEYBUF(p), _KEYLEN(src));
8538 
8539 	return m;
8540 }
8541 
8542 /*
8543  * Take one of the kernel's lifetime data structures and convert it
8544  * into a PF_KEY structure within an mbuf, suitable for sending up to
8545  * a waiting application in user land.
8546  *
8547  * IN:
8548  *    src: A pointer to a kernel lifetime structure.
8549  *    exttype: Which type of lifetime this is. Refer to the PF_KEY
8550  *             data structures for more information.
8551  * OUT:
8552  *    a valid mbuf or NULL indicating an error
8553  *
8554  */
8555 
8556 static struct mbuf *
8557 key_setlifetime(struct seclifetime *src, uint16_t exttype)
8558 {
8559 	struct mbuf *m = NULL;
8560 	struct sadb_lifetime *p;
8561 	int len = PFKEY_ALIGN8(sizeof(struct sadb_lifetime));
8562 
8563 	if (src == NULL)
8564 		return NULL;
8565 
8566 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8567 	if (m == NULL)
8568 		return m;
8569 	m_align(m, len);
8570 	m->m_len = len;
8571 	p = mtod(m, struct sadb_lifetime *);
8572 
8573 	bzero(p, len);
8574 	p->sadb_lifetime_len = PFKEY_UNIT64(len);
8575 	p->sadb_lifetime_exttype = exttype;
8576 	p->sadb_lifetime_allocations = src->allocations;
8577 	p->sadb_lifetime_bytes = src->bytes;
8578 	p->sadb_lifetime_addtime = src->addtime;
8579 	p->sadb_lifetime_usetime = src->usetime;
8580 
8581 	return m;
8582 
8583 }
8584 
8585 const struct enc_xform *
8586 enc_algorithm_lookup(int alg)
8587 {
8588 	int i;
8589 
8590 	for (i = 0; i < nitems(supported_ealgs); i++)
8591 		if (alg == supported_ealgs[i].sadb_alg)
8592 			return (supported_ealgs[i].xform);
8593 	return (NULL);
8594 }
8595 
8596 const struct auth_hash *
8597 auth_algorithm_lookup(int alg)
8598 {
8599 	int i;
8600 
8601 	for (i = 0; i < nitems(supported_aalgs); i++)
8602 		if (alg == supported_aalgs[i].sadb_alg)
8603 			return (supported_aalgs[i].xform);
8604 	return (NULL);
8605 }
8606 
8607 const struct comp_algo *
8608 comp_algorithm_lookup(int alg)
8609 {
8610 	int i;
8611 
8612 	for (i = 0; i < nitems(supported_calgs); i++)
8613 		if (alg == supported_calgs[i].sadb_alg)
8614 			return (supported_calgs[i].xform);
8615 	return (NULL);
8616 }
8617 
8618 /*
8619  * Register a transform.
8620  */
8621 static int
8622 xform_register(struct xformsw* xsp)
8623 {
8624 	struct xformsw *entry;
8625 
8626 	XFORMS_LOCK();
8627 	LIST_FOREACH(entry, &xforms, chain) {
8628 		if (entry->xf_type == xsp->xf_type) {
8629 			XFORMS_UNLOCK();
8630 			return (EEXIST);
8631 		}
8632 	}
8633 	LIST_INSERT_HEAD(&xforms, xsp, chain);
8634 	XFORMS_UNLOCK();
8635 	return (0);
8636 }
8637 
8638 void
8639 xform_attach(void *data)
8640 {
8641 	struct xformsw *xsp = (struct xformsw *)data;
8642 
8643 	if (xform_register(xsp) != 0)
8644 		printf("%s: failed to register %s xform\n", __func__,
8645 		    xsp->xf_name);
8646 }
8647 
8648 void
8649 xform_detach(void *data)
8650 {
8651 	struct xformsw *xsp = (struct xformsw *)data;
8652 
8653 	XFORMS_LOCK();
8654 	LIST_REMOVE(xsp, chain);
8655 	XFORMS_UNLOCK();
8656 
8657 	/* Delete all SAs related to this xform. */
8658 	key_delete_xform(xsp);
8659 }
8660 
8661 /*
8662  * Initialize transform support in an sav.
8663  */
8664 static int
8665 xform_init(struct secasvar *sav, u_short xftype)
8666 {
8667 	struct xformsw *entry;
8668 	int ret;
8669 
8670 	IPSEC_ASSERT(sav->tdb_xform == NULL,
8671 	    ("tdb_xform is already initialized"));
8672 
8673 	ret = EINVAL;
8674 	XFORMS_LOCK();
8675 	LIST_FOREACH(entry, &xforms, chain) {
8676 	    if (entry->xf_type == xftype) {
8677 		    ret = (*entry->xf_init)(sav, entry);
8678 		    break;
8679 	    }
8680 	}
8681 	XFORMS_UNLOCK();
8682 	return (ret);
8683 }
8684 
8685