xref: /freebsd/sys/netipsec/key.c (revision 573bd1fcf532eae35ac30d6aa2c6ff4985a60fe8)
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 
70 #include <netinet/in.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/ip.h>
73 #include <netinet/in_var.h>
74 #include <netinet/udp.h>
75 
76 #ifdef INET6
77 #include <netinet/ip6.h>
78 #include <netinet6/in6_var.h>
79 #include <netinet6/ip6_var.h>
80 #endif /* INET6 */
81 
82 #include <net/pfkeyv2.h>
83 #include <netipsec/keydb.h>
84 #include <netipsec/key.h>
85 #include <netipsec/keysock.h>
86 #include <netipsec/key_debug.h>
87 
88 #include <netipsec/ipsec.h>
89 #ifdef INET6
90 #include <netipsec/ipsec6.h>
91 #endif
92 
93 #include <netipsec/xform.h>
94 #include <machine/in_cksum.h>
95 #include <machine/stdarg.h>
96 
97 /* randomness */
98 #include <sys/random.h>
99 
100 #define FULLMASK	0xff
101 #define	_BITS(bytes)	((bytes) << 3)
102 
103 #define	UINT32_80PCT	0xcccccccc
104 /*
105  * Note on SA reference counting:
106  * - SAs that are not in DEAD state will have (total external reference + 1)
107  *   following value in reference count field.  they cannot be freed and are
108  *   referenced from SA header.
109  * - SAs that are in DEAD state will have (total external reference)
110  *   in reference count field.  they are ready to be freed.  reference from
111  *   SA header will be removed in key_delsav(), when the reference count
112  *   field hits 0 (= no external reference other than from SA header.
113  */
114 
115 VNET_DEFINE(u_int32_t, key_debug_level) = 0;
116 VNET_DEFINE_STATIC(u_int, key_spi_trycnt) = 1000;
117 VNET_DEFINE_STATIC(u_int32_t, key_spi_minval) = 0x100;
118 VNET_DEFINE_STATIC(u_int32_t, key_spi_maxval) = 0x0fffffff;	/* XXX */
119 VNET_DEFINE_STATIC(u_int32_t, policy_id) = 0;
120 /*interval to initialize randseed,1(m)*/
121 VNET_DEFINE_STATIC(u_int, key_int_random) = 60;
122 /* interval to expire acquiring, 30(s)*/
123 VNET_DEFINE_STATIC(u_int, key_larval_lifetime) = 30;
124 /* counter for blocking SADB_ACQUIRE.*/
125 VNET_DEFINE_STATIC(int, key_blockacq_count) = 10;
126 /* lifetime for blocking SADB_ACQUIRE.*/
127 VNET_DEFINE_STATIC(int, key_blockacq_lifetime) = 20;
128 /* preferred old sa rather than new sa.*/
129 VNET_DEFINE_STATIC(int, key_preferred_oldsa) = 1;
130 #define	V_key_spi_trycnt	VNET(key_spi_trycnt)
131 #define	V_key_spi_minval	VNET(key_spi_minval)
132 #define	V_key_spi_maxval	VNET(key_spi_maxval)
133 #define	V_policy_id		VNET(policy_id)
134 #define	V_key_int_random	VNET(key_int_random)
135 #define	V_key_larval_lifetime	VNET(key_larval_lifetime)
136 #define	V_key_blockacq_count	VNET(key_blockacq_count)
137 #define	V_key_blockacq_lifetime	VNET(key_blockacq_lifetime)
138 #define	V_key_preferred_oldsa	VNET(key_preferred_oldsa)
139 
140 VNET_DEFINE_STATIC(u_int32_t, acq_seq) = 0;
141 #define	V_acq_seq		VNET(acq_seq)
142 
143 VNET_DEFINE_STATIC(uint32_t, sp_genid) = 0;
144 #define	V_sp_genid		VNET(sp_genid)
145 
146 /* SPD */
147 TAILQ_HEAD(secpolicy_queue, secpolicy);
148 LIST_HEAD(secpolicy_list, secpolicy);
149 VNET_DEFINE_STATIC(struct secpolicy_queue, sptree[IPSEC_DIR_MAX]);
150 VNET_DEFINE_STATIC(struct secpolicy_queue, sptree_ifnet[IPSEC_DIR_MAX]);
151 static struct rmlock sptree_lock;
152 #define	V_sptree		VNET(sptree)
153 #define	V_sptree_ifnet		VNET(sptree_ifnet)
154 #define	SPTREE_LOCK_INIT()      rm_init(&sptree_lock, "sptree")
155 #define	SPTREE_LOCK_DESTROY()   rm_destroy(&sptree_lock)
156 #define	SPTREE_RLOCK_TRACKER    struct rm_priotracker sptree_tracker
157 #define	SPTREE_RLOCK()          rm_rlock(&sptree_lock, &sptree_tracker)
158 #define	SPTREE_RUNLOCK()        rm_runlock(&sptree_lock, &sptree_tracker)
159 #define	SPTREE_RLOCK_ASSERT()   rm_assert(&sptree_lock, RA_RLOCKED)
160 #define	SPTREE_WLOCK()          rm_wlock(&sptree_lock)
161 #define	SPTREE_WUNLOCK()        rm_wunlock(&sptree_lock)
162 #define	SPTREE_WLOCK_ASSERT()   rm_assert(&sptree_lock, RA_WLOCKED)
163 #define	SPTREE_UNLOCK_ASSERT()  rm_assert(&sptree_lock, RA_UNLOCKED)
164 
165 /* Hash table for lookup SP using unique id */
166 VNET_DEFINE_STATIC(struct secpolicy_list *, sphashtbl);
167 VNET_DEFINE_STATIC(u_long, sphash_mask);
168 #define	V_sphashtbl		VNET(sphashtbl)
169 #define	V_sphash_mask		VNET(sphash_mask)
170 
171 #define	SPHASH_NHASH_LOG2	7
172 #define	SPHASH_NHASH		(1 << SPHASH_NHASH_LOG2)
173 #define	SPHASH_HASHVAL(id)	(key_u32hash(id) & V_sphash_mask)
174 #define	SPHASH_HASH(id)		&V_sphashtbl[SPHASH_HASHVAL(id)]
175 
176 /* SPD cache */
177 struct spdcache_entry {
178    struct secpolicyindex spidx;	/* secpolicyindex */
179    struct secpolicy *sp;	/* cached policy to be used */
180 
181    LIST_ENTRY(spdcache_entry) chain;
182 };
183 LIST_HEAD(spdcache_entry_list, spdcache_entry);
184 
185 #define	SPDCACHE_MAX_ENTRIES_PER_HASH	8
186 
187 VNET_DEFINE_STATIC(u_int, key_spdcache_maxentries) = 0;
188 #define	V_key_spdcache_maxentries	VNET(key_spdcache_maxentries)
189 VNET_DEFINE_STATIC(u_int, key_spdcache_threshold) = 32;
190 #define	V_key_spdcache_threshold	VNET(key_spdcache_threshold)
191 VNET_DEFINE_STATIC(unsigned long, spd_size) = 0;
192 #define	V_spd_size		VNET(spd_size)
193 
194 #define SPDCACHE_ENABLED()	(V_key_spdcache_maxentries != 0)
195 #define SPDCACHE_ACTIVE() \
196 	(SPDCACHE_ENABLED() && V_spd_size >= V_key_spdcache_threshold)
197 
198 VNET_DEFINE_STATIC(struct spdcache_entry_list *, spdcachehashtbl);
199 VNET_DEFINE_STATIC(u_long, spdcachehash_mask);
200 #define	V_spdcachehashtbl	VNET(spdcachehashtbl)
201 #define	V_spdcachehash_mask	VNET(spdcachehash_mask)
202 
203 #define	SPDCACHE_HASHVAL(idx) \
204 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->ul_proto) &  \
205 	    V_spdcachehash_mask)
206 
207 /* Each cache line is protected by a mutex */
208 VNET_DEFINE_STATIC(struct mtx *, spdcache_lock);
209 #define	V_spdcache_lock		VNET(spdcache_lock)
210 
211 #define	SPDCACHE_LOCK_INIT(a) \
212 	mtx_init(&V_spdcache_lock[a], "spdcache", \
213 	    "fast ipsec SPD cache", MTX_DEF|MTX_DUPOK)
214 #define	SPDCACHE_LOCK_DESTROY(a)	mtx_destroy(&V_spdcache_lock[a])
215 #define	SPDCACHE_LOCK(a)		mtx_lock(&V_spdcache_lock[a]);
216 #define	SPDCACHE_UNLOCK(a)		mtx_unlock(&V_spdcache_lock[a]);
217 
218 static struct sx spi_alloc_lock;
219 #define	SPI_ALLOC_LOCK_INIT()		sx_init(&spi_alloc_lock, "spialloc")
220 #define	SPI_ALLOC_LOCK_DESTROY()	sx_destroy(&spi_alloc_lock)
221 #define	SPI_ALLOC_LOCK()          	sx_xlock(&spi_alloc_lock)
222 #define	SPI_ALLOC_UNLOCK()        	sx_unlock(&spi_alloc_lock)
223 #define	SPI_ALLOC_LOCK_ASSERT()   	sx_assert(&spi_alloc_lock, SA_XLOCKED)
224 
225 /* SAD */
226 TAILQ_HEAD(secashead_queue, secashead);
227 LIST_HEAD(secashead_list, secashead);
228 VNET_DEFINE_STATIC(struct secashead_queue, sahtree);
229 static struct rmlock sahtree_lock;
230 #define	V_sahtree		VNET(sahtree)
231 #define	SAHTREE_LOCK_INIT()	rm_init(&sahtree_lock, "sahtree")
232 #define	SAHTREE_LOCK_DESTROY()	rm_destroy(&sahtree_lock)
233 #define	SAHTREE_RLOCK_TRACKER	struct rm_priotracker sahtree_tracker
234 #define	SAHTREE_RLOCK()		rm_rlock(&sahtree_lock, &sahtree_tracker)
235 #define	SAHTREE_RUNLOCK()	rm_runlock(&sahtree_lock, &sahtree_tracker)
236 #define	SAHTREE_RLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_RLOCKED)
237 #define	SAHTREE_WLOCK()		rm_wlock(&sahtree_lock)
238 #define	SAHTREE_WUNLOCK()	rm_wunlock(&sahtree_lock)
239 #define	SAHTREE_WLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_WLOCKED)
240 #define	SAHTREE_UNLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_UNLOCKED)
241 
242 /* Hash table for lookup in SAD using SA addresses */
243 VNET_DEFINE_STATIC(struct secashead_list *, sahaddrhashtbl);
244 VNET_DEFINE_STATIC(u_long, sahaddrhash_mask);
245 #define	V_sahaddrhashtbl	VNET(sahaddrhashtbl)
246 #define	V_sahaddrhash_mask	VNET(sahaddrhash_mask)
247 
248 #define	SAHHASH_NHASH_LOG2	7
249 #define	SAHHASH_NHASH		(1 << SAHHASH_NHASH_LOG2)
250 #define	SAHADDRHASH_HASHVAL(idx)	\
251 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->proto) & \
252 	    V_sahaddrhash_mask)
253 #define	SAHADDRHASH_HASH(saidx)		\
254     &V_sahaddrhashtbl[SAHADDRHASH_HASHVAL(saidx)]
255 
256 /* Hash table for lookup in SAD using SPI */
257 LIST_HEAD(secasvar_list, secasvar);
258 VNET_DEFINE_STATIC(struct secasvar_list *, savhashtbl);
259 VNET_DEFINE_STATIC(u_long, savhash_mask);
260 #define	V_savhashtbl		VNET(savhashtbl)
261 #define	V_savhash_mask		VNET(savhash_mask)
262 #define	SAVHASH_NHASH_LOG2	7
263 #define	SAVHASH_NHASH		(1 << SAVHASH_NHASH_LOG2)
264 #define	SAVHASH_HASHVAL(spi)	(key_u32hash(spi) & V_savhash_mask)
265 #define	SAVHASH_HASH(spi)	&V_savhashtbl[SAVHASH_HASHVAL(spi)]
266 
267 static uint32_t
268 key_addrprotohash(const union sockaddr_union *src,
269     const union sockaddr_union *dst, const uint8_t *proto)
270 {
271 	uint32_t hval;
272 
273 	hval = fnv_32_buf(proto, sizeof(*proto),
274 	    FNV1_32_INIT);
275 	switch (dst->sa.sa_family) {
276 #ifdef INET
277 	case AF_INET:
278 		hval = fnv_32_buf(&src->sin.sin_addr,
279 		    sizeof(in_addr_t), hval);
280 		hval = fnv_32_buf(&dst->sin.sin_addr,
281 		    sizeof(in_addr_t), hval);
282 		break;
283 #endif
284 #ifdef INET6
285 	case AF_INET6:
286 		hval = fnv_32_buf(&src->sin6.sin6_addr,
287 		    sizeof(struct in6_addr), hval);
288 		hval = fnv_32_buf(&dst->sin6.sin6_addr,
289 		    sizeof(struct in6_addr), hval);
290 		break;
291 #endif
292 	default:
293 		hval = 0;
294 		ipseclog((LOG_DEBUG, "%s: unknown address family %d\n",
295 		    __func__, dst->sa.sa_family));
296 	}
297 	return (hval);
298 }
299 
300 static uint32_t
301 key_u32hash(uint32_t val)
302 {
303 
304 	return (fnv_32_buf(&val, sizeof(val), FNV1_32_INIT));
305 }
306 
307 							/* registed list */
308 VNET_DEFINE_STATIC(LIST_HEAD(_regtree, secreg), regtree[SADB_SATYPE_MAX + 1]);
309 #define	V_regtree		VNET(regtree)
310 static struct mtx regtree_lock;
311 #define	REGTREE_LOCK_INIT() \
312 	mtx_init(&regtree_lock, "regtree", "fast ipsec regtree", MTX_DEF)
313 #define	REGTREE_LOCK_DESTROY()	mtx_destroy(&regtree_lock)
314 #define	REGTREE_LOCK()		mtx_lock(&regtree_lock)
315 #define	REGTREE_UNLOCK()	mtx_unlock(&regtree_lock)
316 #define	REGTREE_LOCK_ASSERT()	mtx_assert(&regtree_lock, MA_OWNED)
317 
318 /* Acquiring list */
319 LIST_HEAD(secacq_list, secacq);
320 VNET_DEFINE_STATIC(struct secacq_list, acqtree);
321 #define	V_acqtree		VNET(acqtree)
322 static struct mtx acq_lock;
323 #define	ACQ_LOCK_INIT() \
324     mtx_init(&acq_lock, "acqtree", "ipsec SA acquiring list", MTX_DEF)
325 #define	ACQ_LOCK_DESTROY()	mtx_destroy(&acq_lock)
326 #define	ACQ_LOCK()		mtx_lock(&acq_lock)
327 #define	ACQ_UNLOCK()		mtx_unlock(&acq_lock)
328 #define	ACQ_LOCK_ASSERT()	mtx_assert(&acq_lock, MA_OWNED)
329 
330 /* Hash table for lookup in ACQ list using SA addresses */
331 VNET_DEFINE_STATIC(struct secacq_list *, acqaddrhashtbl);
332 VNET_DEFINE_STATIC(u_long, acqaddrhash_mask);
333 #define	V_acqaddrhashtbl	VNET(acqaddrhashtbl)
334 #define	V_acqaddrhash_mask	VNET(acqaddrhash_mask)
335 
336 /* Hash table for lookup in ACQ list using SEQ number */
337 VNET_DEFINE_STATIC(struct secacq_list *, acqseqhashtbl);
338 VNET_DEFINE_STATIC(u_long, acqseqhash_mask);
339 #define	V_acqseqhashtbl		VNET(acqseqhashtbl)
340 #define	V_acqseqhash_mask	VNET(acqseqhash_mask)
341 
342 #define	ACQHASH_NHASH_LOG2	7
343 #define	ACQHASH_NHASH		(1 << ACQHASH_NHASH_LOG2)
344 #define	ACQADDRHASH_HASHVAL(idx)	\
345 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->proto) & \
346 	    V_acqaddrhash_mask)
347 #define	ACQSEQHASH_HASHVAL(seq)		\
348     (key_u32hash(seq) & V_acqseqhash_mask)
349 #define	ACQADDRHASH_HASH(saidx)	\
350     &V_acqaddrhashtbl[ACQADDRHASH_HASHVAL(saidx)]
351 #define	ACQSEQHASH_HASH(seq)	\
352     &V_acqseqhashtbl[ACQSEQHASH_HASHVAL(seq)]
353 							/* SP acquiring list */
354 VNET_DEFINE_STATIC(LIST_HEAD(_spacqtree, secspacq), spacqtree);
355 #define	V_spacqtree		VNET(spacqtree)
356 static struct mtx spacq_lock;
357 #define	SPACQ_LOCK_INIT() \
358 	mtx_init(&spacq_lock, "spacqtree", \
359 		"fast ipsec security policy acquire list", MTX_DEF)
360 #define	SPACQ_LOCK_DESTROY()	mtx_destroy(&spacq_lock)
361 #define	SPACQ_LOCK()		mtx_lock(&spacq_lock)
362 #define	SPACQ_UNLOCK()		mtx_unlock(&spacq_lock)
363 #define	SPACQ_LOCK_ASSERT()	mtx_assert(&spacq_lock, MA_OWNED)
364 
365 static const int minsize[] = {
366 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
367 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
368 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
369 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
370 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
371 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_SRC */
372 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_DST */
373 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_PROXY */
374 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_AUTH */
375 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_ENCRYPT */
376 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_SRC */
377 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_DST */
378 	sizeof(struct sadb_sens),	/* SADB_EXT_SENSITIVITY */
379 	sizeof(struct sadb_prop),	/* SADB_EXT_PROPOSAL */
380 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_AUTH */
381 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_ENCRYPT */
382 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
383 	0,				/* SADB_X_EXT_KMPRIVATE */
384 	sizeof(struct sadb_x_policy),	/* SADB_X_EXT_POLICY */
385 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
386 	sizeof(struct sadb_x_nat_t_type),/* SADB_X_EXT_NAT_T_TYPE */
387 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_SPORT */
388 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_DPORT */
389 	sizeof(struct sadb_address),	/* SADB_X_EXT_NAT_T_OAI */
390 	sizeof(struct sadb_address),	/* SADB_X_EXT_NAT_T_OAR */
391 	sizeof(struct sadb_x_nat_t_frag),/* SADB_X_EXT_NAT_T_FRAG */
392 	sizeof(struct sadb_x_sa_replay), /* SADB_X_EXT_SA_REPLAY */
393 	sizeof(struct sadb_address),	/* SADB_X_EXT_NEW_ADDRESS_SRC */
394 	sizeof(struct sadb_address),	/* SADB_X_EXT_NEW_ADDRESS_DST */
395 };
396 _Static_assert(sizeof(minsize)/sizeof(int) == SADB_EXT_MAX + 1, "minsize size mismatch");
397 
398 static const int maxsize[] = {
399 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
400 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
401 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
402 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
403 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
404 	0,				/* SADB_EXT_ADDRESS_SRC */
405 	0,				/* SADB_EXT_ADDRESS_DST */
406 	0,				/* SADB_EXT_ADDRESS_PROXY */
407 	0,				/* SADB_EXT_KEY_AUTH */
408 	0,				/* SADB_EXT_KEY_ENCRYPT */
409 	0,				/* SADB_EXT_IDENTITY_SRC */
410 	0,				/* SADB_EXT_IDENTITY_DST */
411 	0,				/* SADB_EXT_SENSITIVITY */
412 	0,				/* SADB_EXT_PROPOSAL */
413 	0,				/* SADB_EXT_SUPPORTED_AUTH */
414 	0,				/* SADB_EXT_SUPPORTED_ENCRYPT */
415 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
416 	0,				/* SADB_X_EXT_KMPRIVATE */
417 	0,				/* SADB_X_EXT_POLICY */
418 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
419 	sizeof(struct sadb_x_nat_t_type),/* SADB_X_EXT_NAT_T_TYPE */
420 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_SPORT */
421 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_DPORT */
422 	0,				/* SADB_X_EXT_NAT_T_OAI */
423 	0,				/* SADB_X_EXT_NAT_T_OAR */
424 	sizeof(struct sadb_x_nat_t_frag),/* SADB_X_EXT_NAT_T_FRAG */
425 	sizeof(struct sadb_x_sa_replay), /* SADB_X_EXT_SA_REPLAY */
426 	0,				/* SADB_X_EXT_NEW_ADDRESS_SRC */
427 	0,				/* SADB_X_EXT_NEW_ADDRESS_DST */
428 };
429 _Static_assert(sizeof(maxsize)/sizeof(int) == SADB_EXT_MAX + 1, "minsize size mismatch");
430 
431 /*
432  * Internal values for SA flags:
433  * SADB_X_EXT_F_CLONED means that SA was cloned by key_updateaddresses,
434  *	thus we will not free the most of SA content in key_delsav().
435  */
436 #define	SADB_X_EXT_F_CLONED	0x80000000
437 
438 #define	SADB_CHECKLEN(_mhp, _ext)			\
439     ((_mhp)->extlen[(_ext)] < minsize[(_ext)] || (maxsize[(_ext)] != 0 && \
440 	((_mhp)->extlen[(_ext)] > maxsize[(_ext)])))
441 #define	SADB_CHECKHDR(_mhp, _ext)	((_mhp)->ext[(_ext)] == NULL)
442 
443 VNET_DEFINE_STATIC(int, ipsec_esp_keymin) = 256;
444 VNET_DEFINE_STATIC(int, ipsec_esp_auth) = 0;
445 VNET_DEFINE_STATIC(int, ipsec_ah_keymin) = 128;
446 
447 #define	V_ipsec_esp_keymin	VNET(ipsec_esp_keymin)
448 #define	V_ipsec_esp_auth	VNET(ipsec_esp_auth)
449 #define	V_ipsec_ah_keymin	VNET(ipsec_ah_keymin)
450 
451 #ifdef IPSEC_DEBUG
452 VNET_DEFINE(int, ipsec_debug) = 1;
453 #else
454 VNET_DEFINE(int, ipsec_debug) = 0;
455 #endif
456 
457 #ifdef INET
458 SYSCTL_DECL(_net_inet_ipsec);
459 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
460     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
461     "Enable IPsec debugging output when set.");
462 #endif
463 #ifdef INET6
464 SYSCTL_DECL(_net_inet6_ipsec6);
465 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug,
466     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
467     "Enable IPsec debugging output when set.");
468 #endif
469 
470 SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,
471 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_debug_level), 0, "");
472 
473 /* max count of trial for the decision of spi value */
474 SYSCTL_INT(_net_key, KEYCTL_SPI_TRY, spi_trycnt,
475 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_trycnt), 0, "");
476 
477 /* minimum spi value to allocate automatically. */
478 SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE, spi_minval,
479 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_minval), 0, "");
480 
481 /* maximun spi value to allocate automatically. */
482 SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE, spi_maxval,
483 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_maxval), 0, "");
484 
485 /* interval to initialize randseed */
486 SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT, int_random,
487 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_int_random), 0, "");
488 
489 /* lifetime for larval SA */
490 SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME, larval_lifetime,
491 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_larval_lifetime), 0, "");
492 
493 /* counter for blocking to send SADB_ACQUIRE to IKEd */
494 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT, blockacq_count,
495 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_count), 0, "");
496 
497 /* lifetime for blocking to send SADB_ACQUIRE to IKEd */
498 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME, blockacq_lifetime,
499 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_lifetime), 0, "");
500 
501 /* ESP auth */
502 SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH, esp_auth,
503 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_auth), 0, "");
504 
505 /* minimum ESP key length */
506 SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN, esp_keymin,
507 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_keymin), 0, "");
508 
509 /* minimum AH key length */
510 SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN, ah_keymin,
511 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_ah_keymin), 0, "");
512 
513 /* perfered old SA rather than new SA */
514 SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA, preferred_oldsa,
515 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_preferred_oldsa), 0, "");
516 
517 SYSCTL_NODE(_net_key, OID_AUTO, spdcache, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
518     "SPD cache");
519 
520 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, maxentries,
521 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_maxentries), 0,
522 	"Maximum number of entries in the SPD cache"
523 	" (power of 2, 0 to disable)");
524 
525 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, threshold,
526 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_threshold), 0,
527 	"Number of SPs that make the SPD cache active");
528 
529 #define __LIST_CHAINED(elm) \
530 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
531 
532 MALLOC_DEFINE(M_IPSEC_SA, "secasvar", "ipsec security association");
533 MALLOC_DEFINE(M_IPSEC_SAH, "sahead", "ipsec sa head");
534 MALLOC_DEFINE(M_IPSEC_SP, "ipsecpolicy", "ipsec security policy");
535 MALLOC_DEFINE(M_IPSEC_SR, "ipsecrequest", "ipsec security request");
536 MALLOC_DEFINE(M_IPSEC_MISC, "ipsec-misc", "ipsec miscellaneous");
537 MALLOC_DEFINE(M_IPSEC_SAQ, "ipsec-saq", "ipsec sa acquire");
538 MALLOC_DEFINE(M_IPSEC_SAR, "ipsec-reg", "ipsec sa acquire");
539 MALLOC_DEFINE(M_IPSEC_SPDCACHE, "ipsec-spdcache", "ipsec SPD cache");
540 
541 static uma_zone_t __read_mostly ipsec_key_lft_zone;
542 
543 /*
544  * set parameters into secpolicyindex buffer.
545  * Must allocate secpolicyindex buffer passed to this function.
546  */
547 #define KEY_SETSECSPIDX(_dir, s, d, ps, pd, ulp, idx) \
548 do { \
549 	bzero((idx), sizeof(struct secpolicyindex));                         \
550 	(idx)->dir = (_dir);                                                 \
551 	(idx)->prefs = (ps);                                                 \
552 	(idx)->prefd = (pd);                                                 \
553 	(idx)->ul_proto = (ulp);                                             \
554 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
555 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
556 } while (0)
557 
558 /*
559  * set parameters into secasindex buffer.
560  * Must allocate secasindex buffer before calling this function.
561  */
562 #define KEY_SETSECASIDX(p, m, r, s, d, idx) \
563 do { \
564 	bzero((idx), sizeof(struct secasindex));                             \
565 	(idx)->proto = (p);                                                  \
566 	(idx)->mode = (m);                                                   \
567 	(idx)->reqid = (r);                                                  \
568 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
569 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
570 	key_porttosaddr(&(idx)->src.sa, 0);				     \
571 	key_porttosaddr(&(idx)->dst.sa, 0);				     \
572 } while (0)
573 
574 /* key statistics */
575 struct _keystat {
576 	u_long getspi_count; /* the avarage of count to try to get new SPI */
577 } keystat;
578 
579 struct sadb_msghdr {
580 	struct sadb_msg *msg;
581 	struct sadb_ext *ext[SADB_EXT_MAX + 1];
582 	int extoff[SADB_EXT_MAX + 1];
583 	int extlen[SADB_EXT_MAX + 1];
584 };
585 
586 static struct supported_ealgs {
587 	int sadb_alg;
588 	const struct enc_xform *xform;
589 } supported_ealgs[] = {
590 	{ SADB_X_EALG_AES,		&enc_xform_aes_cbc },
591 	{ SADB_EALG_NULL,		&enc_xform_null },
592 	{ SADB_X_EALG_AESCTR,		&enc_xform_aes_icm },
593 	{ SADB_X_EALG_AESGCM16,		&enc_xform_aes_nist_gcm },
594 	{ SADB_X_EALG_AESGMAC,		&enc_xform_aes_nist_gmac },
595 	{ SADB_X_EALG_CHACHA20POLY1305,	&enc_xform_chacha20_poly1305 },
596 };
597 
598 static struct supported_aalgs {
599 	int sadb_alg;
600 	const struct auth_hash *xform;
601 } supported_aalgs[] = {
602 	{ SADB_X_AALG_NULL,		&auth_hash_null },
603 	{ SADB_AALG_SHA1HMAC,		&auth_hash_hmac_sha1 },
604 	{ SADB_X_AALG_SHA2_256,		&auth_hash_hmac_sha2_256 },
605 	{ SADB_X_AALG_SHA2_384,		&auth_hash_hmac_sha2_384 },
606 	{ SADB_X_AALG_SHA2_512,		&auth_hash_hmac_sha2_512 },
607 	{ SADB_X_AALG_AES128GMAC,	&auth_hash_nist_gmac_aes_128 },
608 	{ SADB_X_AALG_AES192GMAC,	&auth_hash_nist_gmac_aes_192 },
609 	{ SADB_X_AALG_AES256GMAC,	&auth_hash_nist_gmac_aes_256 },
610 	{ SADB_X_AALG_CHACHA20POLY1305,	&auth_hash_poly1305 },
611 };
612 
613 static struct supported_calgs {
614 	int sadb_alg;
615 	const struct comp_algo *xform;
616 } supported_calgs[] = {
617 	{ SADB_X_CALG_DEFLATE,		&comp_algo_deflate },
618 };
619 
620 #ifndef IPSEC_DEBUG2
621 static struct callout key_timer;
622 #endif
623 
624 static void key_unlink(struct secpolicy *);
625 static void key_detach(struct secpolicy *);
626 static struct secpolicy *key_do_allocsp(struct secpolicyindex *spidx, u_int dir);
627 static struct secpolicy *key_getsp(struct secpolicyindex *);
628 static struct secpolicy *key_getspbyid(u_int32_t);
629 static struct mbuf *key_gather_mbuf(struct mbuf *,
630 	const struct sadb_msghdr *, int, int, ...);
631 static int key_spdadd(struct socket *, struct mbuf *,
632 	const struct sadb_msghdr *);
633 static uint32_t key_getnewspid(void);
634 static int key_spddelete(struct socket *, struct mbuf *,
635 	const struct sadb_msghdr *);
636 static int key_spddelete2(struct socket *, struct mbuf *,
637 	const struct sadb_msghdr *);
638 static int key_spdget(struct socket *, struct mbuf *,
639 	const struct sadb_msghdr *);
640 static int key_spdflush(struct socket *, struct mbuf *,
641 	const struct sadb_msghdr *);
642 static int key_spddump(struct socket *, struct mbuf *,
643 	const struct sadb_msghdr *);
644 static struct mbuf *key_setdumpsp(struct secpolicy *,
645 	u_int8_t, u_int32_t, u_int32_t);
646 static struct mbuf *key_sp2mbuf(struct secpolicy *);
647 static size_t key_getspreqmsglen(struct secpolicy *);
648 static int key_spdexpire(struct secpolicy *);
649 static struct secashead *key_newsah(struct secasindex *);
650 static void key_freesah(struct secashead **);
651 static void key_delsah(struct secashead *);
652 static struct secasvar *key_newsav(const struct sadb_msghdr *,
653     struct secasindex *, uint32_t, int *);
654 static void key_delsav(struct secasvar *);
655 static void key_unlinksav(struct secasvar *);
656 static struct secashead *key_getsah(struct secasindex *);
657 static int key_checkspidup(uint32_t);
658 static struct secasvar *key_getsavbyspi(uint32_t);
659 static int key_setnatt(struct secasvar *, const struct sadb_msghdr *);
660 static int key_setsaval(struct secasvar *, const struct sadb_msghdr *);
661 static int key_updatelifetimes(struct secasvar *, const struct sadb_msghdr *);
662 static int key_updateaddresses(struct socket *, struct mbuf *,
663     const struct sadb_msghdr *, struct secasvar *, struct secasindex *);
664 
665 static struct mbuf *key_setdumpsa(struct secasvar *, u_int8_t,
666 	u_int8_t, u_int32_t, u_int32_t);
667 static struct mbuf *key_setsadbmsg(u_int8_t, u_int16_t, u_int8_t,
668 	u_int32_t, pid_t, u_int16_t);
669 static struct mbuf *key_setsadbsa(struct secasvar *);
670 static struct mbuf *key_setsadbaddr(u_int16_t,
671 	const struct sockaddr *, u_int8_t, u_int16_t);
672 static struct mbuf *key_setsadbxport(u_int16_t, u_int16_t);
673 static struct mbuf *key_setsadbxtype(u_int16_t);
674 static struct mbuf *key_setsadbxsa2(u_int8_t, u_int32_t, u_int32_t);
675 static struct mbuf *key_setsadbxsareplay(u_int32_t);
676 static struct mbuf *key_setsadbxpolicy(u_int16_t, u_int8_t,
677 	u_int32_t, u_int32_t);
678 static struct seckey *key_dup_keymsg(const struct sadb_key *, size_t,
679     struct malloc_type *);
680 static struct seclifetime *key_dup_lifemsg(const struct sadb_lifetime *src,
681     struct malloc_type *);
682 
683 /* flags for key_cmpsaidx() */
684 #define CMP_HEAD	1	/* protocol, addresses. */
685 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
686 #define CMP_REQID	3	/* additionally HEAD, reaid. */
687 #define CMP_EXACTLY	4	/* all elements. */
688 static int key_cmpsaidx(const struct secasindex *,
689     const struct secasindex *, int);
690 static int key_cmpspidx_exactly(struct secpolicyindex *,
691     struct secpolicyindex *);
692 static int key_cmpspidx_withmask(struct secpolicyindex *,
693     struct secpolicyindex *);
694 static int key_bbcmp(const void *, const void *, u_int);
695 static uint8_t key_satype2proto(uint8_t);
696 static uint8_t key_proto2satype(uint8_t);
697 
698 static int key_getspi(struct socket *, struct mbuf *,
699 	const struct sadb_msghdr *);
700 static uint32_t key_do_getnewspi(struct sadb_spirange *, struct secasindex *);
701 static int key_update(struct socket *, struct mbuf *,
702 	const struct sadb_msghdr *);
703 static int key_add(struct socket *, struct mbuf *,
704 	const struct sadb_msghdr *);
705 static int key_setident(struct secashead *, const struct sadb_msghdr *);
706 static struct mbuf *key_getmsgbuf_x1(struct mbuf *,
707 	const struct sadb_msghdr *);
708 static int key_delete(struct socket *, struct mbuf *,
709 	const struct sadb_msghdr *);
710 static int key_delete_all(struct socket *, struct mbuf *,
711 	const struct sadb_msghdr *, struct secasindex *);
712 static int key_get(struct socket *, struct mbuf *,
713 	const struct sadb_msghdr *);
714 
715 static void key_getcomb_setlifetime(struct sadb_comb *);
716 static struct mbuf *key_getcomb_ealg(void);
717 static struct mbuf *key_getcomb_ah(void);
718 static struct mbuf *key_getcomb_ipcomp(void);
719 static struct mbuf *key_getprop(const struct secasindex *);
720 
721 static int key_acquire(const struct secasindex *, struct secpolicy *);
722 static uint32_t key_newacq(const struct secasindex *, int *);
723 static uint32_t key_getacq(const struct secasindex *, int *);
724 static int key_acqdone(const struct secasindex *, uint32_t);
725 static int key_acqreset(uint32_t);
726 static struct secspacq *key_newspacq(struct secpolicyindex *);
727 static struct secspacq *key_getspacq(struct secpolicyindex *);
728 static int key_acquire2(struct socket *, struct mbuf *,
729 	const struct sadb_msghdr *);
730 static int key_register(struct socket *, struct mbuf *,
731 	const struct sadb_msghdr *);
732 static int key_expire(struct secasvar *, int);
733 static int key_flush(struct socket *, struct mbuf *,
734 	const struct sadb_msghdr *);
735 static int key_dump(struct socket *, struct mbuf *,
736 	const struct sadb_msghdr *);
737 static int key_promisc(struct socket *, struct mbuf *,
738 	const struct sadb_msghdr *);
739 static int key_senderror(struct socket *, struct mbuf *, int);
740 static int key_validate_ext(const struct sadb_ext *, int);
741 static int key_align(struct mbuf *, struct sadb_msghdr *);
742 static struct mbuf *key_setlifetime(struct seclifetime *, uint16_t);
743 static struct mbuf *key_setkey(struct seckey *, uint16_t);
744 
745 static void spdcache_init(void);
746 static void spdcache_clear(void);
747 static struct spdcache_entry *spdcache_entry_alloc(
748 	const struct secpolicyindex *spidx,
749 	struct secpolicy *policy);
750 static void spdcache_entry_free(struct spdcache_entry *entry);
751 #ifdef VIMAGE
752 static void spdcache_destroy(void);
753 #endif
754 
755 #define	DBG_IPSEC_INITREF(t, p)	do {				\
756 	refcount_init(&(p)->refcnt, 1);				\
757 	KEYDBG(KEY_STAMP,					\
758 	    printf("%s: Initialize refcnt %s(%p) = %u\n",	\
759 	    __func__, #t, (p), (p)->refcnt));			\
760 } while (0)
761 #define	DBG_IPSEC_ADDREF(t, p)	do {				\
762 	refcount_acquire(&(p)->refcnt);				\
763 	KEYDBG(KEY_STAMP,					\
764 	    printf("%s: Acquire refcnt %s(%p) -> %u\n",		\
765 	    __func__, #t, (p), (p)->refcnt));			\
766 } while (0)
767 #define	DBG_IPSEC_DELREF(t, p)	do {				\
768 	KEYDBG(KEY_STAMP,					\
769 	    printf("%s: Release refcnt %s(%p) -> %u\n",		\
770 	    __func__, #t, (p), (p)->refcnt - 1));		\
771 	refcount_release(&(p)->refcnt);				\
772 } while (0)
773 
774 #define	IPSEC_INITREF(t, p)	refcount_init(&(p)->refcnt, 1)
775 #define	IPSEC_ADDREF(t, p)	refcount_acquire(&(p)->refcnt)
776 #define	IPSEC_DELREF(t, p)	refcount_release(&(p)->refcnt)
777 
778 #define	SP_INITREF(p)	IPSEC_INITREF(SP, p)
779 #define	SP_ADDREF(p)	IPSEC_ADDREF(SP, p)
780 #define	SP_DELREF(p)	IPSEC_DELREF(SP, p)
781 
782 #define	SAH_INITREF(p)	IPSEC_INITREF(SAH, p)
783 #define	SAH_ADDREF(p)	IPSEC_ADDREF(SAH, p)
784 #define	SAH_DELREF(p)	IPSEC_DELREF(SAH, p)
785 
786 #define	SAV_INITREF(p)	IPSEC_INITREF(SAV, p)
787 #define	SAV_ADDREF(p)	IPSEC_ADDREF(SAV, p)
788 #define	SAV_DELREF(p)	IPSEC_DELREF(SAV, p)
789 
790 /*
791  * Update the refcnt while holding the SPTREE lock.
792  */
793 void
794 key_addref(struct secpolicy *sp)
795 {
796 
797 	SP_ADDREF(sp);
798 }
799 
800 /*
801  * Return 0 when there are known to be no SP's for the specified
802  * direction.  Otherwise return 1.  This is used by IPsec code
803  * to optimize performance.
804  */
805 int
806 key_havesp(u_int dir)
807 {
808 
809 	return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ?
810 		TAILQ_FIRST(&V_sptree[dir]) != NULL : 1);
811 }
812 
813 /* %%% IPsec policy management */
814 /*
815  * Return current SPDB generation.
816  */
817 uint32_t
818 key_getspgen(void)
819 {
820 
821 	return (V_sp_genid);
822 }
823 
824 void
825 key_bumpspgen(void)
826 {
827 
828 	V_sp_genid++;
829 }
830 
831 static int
832 key_checksockaddrs(struct sockaddr *src, struct sockaddr *dst)
833 {
834 
835 	/* family match */
836 	if (src->sa_family != dst->sa_family)
837 		return (EINVAL);
838 	/* sa_len match */
839 	if (src->sa_len != dst->sa_len)
840 		return (EINVAL);
841 	switch (src->sa_family) {
842 #ifdef INET
843 	case AF_INET:
844 		if (src->sa_len != sizeof(struct sockaddr_in))
845 			return (EINVAL);
846 		break;
847 #endif
848 #ifdef INET6
849 	case AF_INET6:
850 		if (src->sa_len != sizeof(struct sockaddr_in6))
851 			return (EINVAL);
852 		break;
853 #endif
854 	default:
855 		return (EAFNOSUPPORT);
856 	}
857 	return (0);
858 }
859 
860 struct secpolicy *
861 key_do_allocsp(struct secpolicyindex *spidx, u_int dir)
862 {
863 	SPTREE_RLOCK_TRACKER;
864 	struct secpolicy *sp;
865 
866 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
867 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
868 		("invalid direction %u", dir));
869 
870 	SPTREE_RLOCK();
871 	TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
872 		if (key_cmpspidx_withmask(&sp->spidx, spidx)) {
873 			SP_ADDREF(sp);
874 			break;
875 		}
876 	}
877 	SPTREE_RUNLOCK();
878 	return (sp);
879 }
880 
881 /*
882  * allocating a SP for OUTBOUND or INBOUND packet.
883  * Must call key_freesp() later.
884  * OUT:	NULL:	not found
885  *	others:	found and return the pointer.
886  */
887 struct secpolicy *
888 key_allocsp(struct secpolicyindex *spidx, u_int dir)
889 {
890 	struct spdcache_entry *entry, *lastentry, *tmpentry;
891 	struct secpolicy *sp;
892 	uint32_t hashv;
893 	int nb_entries;
894 
895 	if (!SPDCACHE_ACTIVE()) {
896 		sp = key_do_allocsp(spidx, dir);
897 		goto out;
898 	}
899 
900 	hashv = SPDCACHE_HASHVAL(spidx);
901 	SPDCACHE_LOCK(hashv);
902 	nb_entries = 0;
903 	LIST_FOREACH_SAFE(entry, &V_spdcachehashtbl[hashv], chain, tmpentry) {
904 		/* Removed outdated entries */
905 		if (entry->sp != NULL &&
906 		    entry->sp->state == IPSEC_SPSTATE_DEAD) {
907 			LIST_REMOVE(entry, chain);
908 			spdcache_entry_free(entry);
909 			continue;
910 		}
911 
912 		nb_entries++;
913 		if (!key_cmpspidx_exactly(&entry->spidx, spidx)) {
914 			lastentry = entry;
915 			continue;
916 		}
917 
918 		sp = entry->sp;
919 		if (entry->sp != NULL)
920 			SP_ADDREF(sp);
921 
922 		/* IPSECSTAT_INC(ips_spdcache_hits); */
923 
924 		SPDCACHE_UNLOCK(hashv);
925 		goto out;
926 	}
927 
928 	/* IPSECSTAT_INC(ips_spdcache_misses); */
929 
930 	sp = key_do_allocsp(spidx, dir);
931 	entry = spdcache_entry_alloc(spidx, sp);
932 	if (entry != NULL) {
933 		if (nb_entries >= SPDCACHE_MAX_ENTRIES_PER_HASH) {
934 			LIST_REMOVE(lastentry, chain);
935 			spdcache_entry_free(lastentry);
936 		}
937 
938 		LIST_INSERT_HEAD(&V_spdcachehashtbl[hashv], entry, chain);
939 	}
940 
941 	SPDCACHE_UNLOCK(hashv);
942 
943 out:
944 	if (sp != NULL) {	/* found a SPD entry */
945 		sp->lastused = time_second;
946 		KEYDBG(IPSEC_STAMP,
947 		    printf("%s: return SP(%p)\n", __func__, sp));
948 		KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
949 	} else {
950 		KEYDBG(IPSEC_DATA,
951 		    printf("%s: lookup failed for ", __func__);
952 		    kdebug_secpolicyindex(spidx, NULL));
953 	}
954 	return (sp);
955 }
956 
957 /*
958  * Allocating an SA entry for an *INBOUND* or *OUTBOUND* TCP packet, signed
959  * or should be signed by MD5 signature.
960  * We don't use key_allocsa() for such lookups, because we don't know SPI.
961  * Unlike ESP and AH protocols, SPI isn't transmitted in the TCP header with
962  * signed packet. We use SADB only as storage for password.
963  * OUT:	positive:	corresponding SA for given saidx found.
964  *	NULL:		SA not found
965  */
966 struct secasvar *
967 key_allocsa_tcpmd5(struct secasindex *saidx)
968 {
969 	SAHTREE_RLOCK_TRACKER;
970 	struct secashead *sah;
971 	struct secasvar *sav;
972 
973 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP,
974 	    ("unexpected security protocol %u", saidx->proto));
975 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TCPMD5,
976 	    ("unexpected mode %u", saidx->mode));
977 
978 	SAHTREE_RLOCK();
979 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
980 		KEYDBG(IPSEC_DUMP,
981 		    printf("%s: checking SAH\n", __func__);
982 		    kdebug_secash(sah, "  "));
983 		if (sah->saidx.proto != IPPROTO_TCP)
984 			continue;
985 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
986 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
987 			break;
988 	}
989 	if (sah != NULL) {
990 		if (V_key_preferred_oldsa)
991 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
992 		else
993 			sav = TAILQ_FIRST(&sah->savtree_alive);
994 		if (sav != NULL)
995 			SAV_ADDREF(sav);
996 	} else
997 		sav = NULL;
998 	SAHTREE_RUNLOCK();
999 
1000 	if (sav != NULL) {
1001 		KEYDBG(IPSEC_STAMP,
1002 		    printf("%s: return SA(%p)\n", __func__, sav));
1003 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1004 	} else {
1005 		KEYDBG(IPSEC_STAMP,
1006 		    printf("%s: SA not found\n", __func__));
1007 		KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1008 	}
1009 	return (sav);
1010 }
1011 
1012 /*
1013  * Allocating an SA entry for an *OUTBOUND* packet.
1014  * OUT:	positive:	corresponding SA for given saidx found.
1015  *	NULL:		SA not found, but will be acquired, check *error
1016  *			for acquiring status.
1017  */
1018 struct secasvar *
1019 key_allocsa_policy(struct secpolicy *sp, const struct secasindex *saidx,
1020     int *error)
1021 {
1022 	SAHTREE_RLOCK_TRACKER;
1023 	struct secashead *sah;
1024 	struct secasvar *sav;
1025 
1026 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
1027 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TRANSPORT ||
1028 		saidx->mode == IPSEC_MODE_TUNNEL,
1029 		("unexpected policy %u", saidx->mode));
1030 
1031 	/*
1032 	 * We check new SA in the IPsec request because a different
1033 	 * SA may be involved each time this request is checked, either
1034 	 * because new SAs are being configured, or this request is
1035 	 * associated with an unconnected datagram socket, or this request
1036 	 * is associated with a system default policy.
1037 	 */
1038 	SAHTREE_RLOCK();
1039 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
1040 		KEYDBG(IPSEC_DUMP,
1041 		    printf("%s: checking SAH\n", __func__);
1042 		    kdebug_secash(sah, "  "));
1043 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID))
1044 			break;
1045 	}
1046 	if (sah != NULL) {
1047 		/*
1048 		 * Allocate the oldest SA available according to
1049 		 * draft-jenkins-ipsec-rekeying-03.
1050 		 */
1051 		if (V_key_preferred_oldsa)
1052 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1053 		else
1054 			sav = TAILQ_FIRST(&sah->savtree_alive);
1055 		if (sav != NULL)
1056 			SAV_ADDREF(sav);
1057 	} else
1058 		sav = NULL;
1059 	SAHTREE_RUNLOCK();
1060 
1061 	if (sav != NULL) {
1062 		*error = 0;
1063 		KEYDBG(IPSEC_STAMP,
1064 		    printf("%s: chosen SA(%p) for SP(%p)\n", __func__,
1065 			sav, sp));
1066 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1067 		return (sav); /* return referenced SA */
1068 	}
1069 
1070 	/* there is no SA */
1071 	*error = key_acquire(saidx, sp);
1072 	if ((*error) != 0)
1073 		ipseclog((LOG_DEBUG,
1074 		    "%s: error %d returned from key_acquire()\n",
1075 			__func__, *error));
1076 	KEYDBG(IPSEC_STAMP,
1077 	    printf("%s: acquire SA for SP(%p), error %d\n",
1078 		__func__, sp, *error));
1079 	KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1080 	return (NULL);
1081 }
1082 
1083 /*
1084  * allocating a usable SA entry for a *INBOUND* packet.
1085  * Must call key_freesav() later.
1086  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
1087  *	NULL:		not found, or error occurred.
1088  *
1089  * According to RFC 2401 SA is uniquely identified by a triple SPI,
1090  * destination address, and security protocol. But according to RFC 4301,
1091  * SPI by itself suffices to specify an SA.
1092  *
1093  * Note that, however, we do need to keep source address in IPsec SA.
1094  * IKE specification and PF_KEY specification do assume that we
1095  * keep source address in IPsec SA.  We see a tricky situation here.
1096  */
1097 struct secasvar *
1098 key_allocsa(union sockaddr_union *dst, uint8_t proto, uint32_t spi)
1099 {
1100 	SAHTREE_RLOCK_TRACKER;
1101 	struct secasvar *sav;
1102 
1103 	IPSEC_ASSERT(proto == IPPROTO_ESP || proto == IPPROTO_AH ||
1104 	    proto == IPPROTO_IPCOMP, ("unexpected security protocol %u",
1105 	    proto));
1106 
1107 	SAHTREE_RLOCK();
1108 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
1109 		if (sav->spi == spi)
1110 			break;
1111 	}
1112 	/*
1113 	 * We use single SPI namespace for all protocols, so it is
1114 	 * impossible to have SPI duplicates in the SAVHASH.
1115 	 */
1116 	if (sav != NULL) {
1117 		if (sav->state != SADB_SASTATE_LARVAL &&
1118 		    sav->sah->saidx.proto == proto &&
1119 		    key_sockaddrcmp(&dst->sa,
1120 			&sav->sah->saidx.dst.sa, 0) == 0)
1121 			SAV_ADDREF(sav);
1122 		else
1123 			sav = NULL;
1124 	}
1125 	SAHTREE_RUNLOCK();
1126 
1127 	if (sav == NULL) {
1128 		KEYDBG(IPSEC_STAMP,
1129 		    char buf[IPSEC_ADDRSTRLEN];
1130 		    printf("%s: SA not found for spi %u proto %u dst %s\n",
1131 			__func__, ntohl(spi), proto, ipsec_address(dst, buf,
1132 			sizeof(buf))));
1133 	} else {
1134 		KEYDBG(IPSEC_STAMP,
1135 		    printf("%s: return SA(%p)\n", __func__, sav));
1136 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1137 	}
1138 	return (sav);
1139 }
1140 
1141 struct secasvar *
1142 key_allocsa_tunnel(union sockaddr_union *src, union sockaddr_union *dst,
1143     uint8_t proto)
1144 {
1145 	SAHTREE_RLOCK_TRACKER;
1146 	struct secasindex saidx;
1147 	struct secashead *sah;
1148 	struct secasvar *sav;
1149 
1150 	IPSEC_ASSERT(src != NULL, ("null src address"));
1151 	IPSEC_ASSERT(dst != NULL, ("null dst address"));
1152 
1153 	KEY_SETSECASIDX(proto, IPSEC_MODE_TUNNEL, 0, &src->sa,
1154 	    &dst->sa, &saidx);
1155 
1156 	sav = NULL;
1157 	SAHTREE_RLOCK();
1158 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
1159 		if (IPSEC_MODE_TUNNEL != sah->saidx.mode)
1160 			continue;
1161 		if (proto != sah->saidx.proto)
1162 			continue;
1163 		if (key_sockaddrcmp(&src->sa, &sah->saidx.src.sa, 0) != 0)
1164 			continue;
1165 		if (key_sockaddrcmp(&dst->sa, &sah->saidx.dst.sa, 0) != 0)
1166 			continue;
1167 		/* XXXAE: is key_preferred_oldsa reasonably?*/
1168 		if (V_key_preferred_oldsa)
1169 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1170 		else
1171 			sav = TAILQ_FIRST(&sah->savtree_alive);
1172 		if (sav != NULL) {
1173 			SAV_ADDREF(sav);
1174 			break;
1175 		}
1176 	}
1177 	SAHTREE_RUNLOCK();
1178 	KEYDBG(IPSEC_STAMP,
1179 	    printf("%s: return SA(%p)\n", __func__, sav));
1180 	if (sav != NULL)
1181 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1182 	return (sav);
1183 }
1184 
1185 /*
1186  * Must be called after calling key_allocsp().
1187  */
1188 void
1189 key_freesp(struct secpolicy **spp)
1190 {
1191 	struct secpolicy *sp = *spp;
1192 
1193 	IPSEC_ASSERT(sp != NULL, ("null sp"));
1194 	if (SP_DELREF(sp) == 0)
1195 		return;
1196 
1197 	KEYDBG(IPSEC_STAMP,
1198 	    printf("%s: last reference to SP(%p)\n", __func__, sp));
1199 	KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1200 
1201 	*spp = NULL;
1202 	while (sp->tcount > 0)
1203 		ipsec_delisr(sp->req[--sp->tcount]);
1204 	free(sp, M_IPSEC_SP);
1205 }
1206 
1207 static void
1208 key_unlink(struct secpolicy *sp)
1209 {
1210 	SPTREE_WLOCK();
1211 	key_detach(sp);
1212 	SPTREE_WUNLOCK();
1213 	if (SPDCACHE_ENABLED())
1214 		spdcache_clear();
1215 	key_freesp(&sp);
1216 }
1217 
1218 static void
1219 key_detach(struct secpolicy *sp)
1220 {
1221 	IPSEC_ASSERT(sp->spidx.dir == IPSEC_DIR_INBOUND ||
1222 	    sp->spidx.dir == IPSEC_DIR_OUTBOUND,
1223 	    ("invalid direction %u", sp->spidx.dir));
1224 	SPTREE_WLOCK_ASSERT();
1225 
1226 	KEYDBG(KEY_STAMP,
1227 	    printf("%s: SP(%p)\n", __func__, sp));
1228 	if (sp->state != IPSEC_SPSTATE_ALIVE) {
1229 		/* SP is already unlinked */
1230 		return;
1231 	}
1232 	sp->state = IPSEC_SPSTATE_DEAD;
1233 	TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
1234 	V_spd_size--;
1235 	LIST_REMOVE(sp, idhash);
1236 	V_sp_genid++;
1237 }
1238 
1239 /*
1240  * insert a secpolicy into the SP database. Lower priorities first
1241  */
1242 static void
1243 key_insertsp(struct secpolicy *newsp)
1244 {
1245 	struct secpolicy *sp;
1246 
1247 	SPTREE_WLOCK_ASSERT();
1248 	TAILQ_FOREACH(sp, &V_sptree[newsp->spidx.dir], chain) {
1249 		if (newsp->priority < sp->priority) {
1250 			TAILQ_INSERT_BEFORE(sp, newsp, chain);
1251 			goto done;
1252 		}
1253 	}
1254 	TAILQ_INSERT_TAIL(&V_sptree[newsp->spidx.dir], newsp, chain);
1255 done:
1256 	LIST_INSERT_HEAD(SPHASH_HASH(newsp->id), newsp, idhash);
1257 	newsp->state = IPSEC_SPSTATE_ALIVE;
1258 	V_spd_size++;
1259 	V_sp_genid++;
1260 }
1261 
1262 /*
1263  * Insert a bunch of VTI secpolicies into the SPDB.
1264  * We keep VTI policies in the separate list due to following reasons:
1265  * 1) they should be immutable to user's or some deamon's attempts to
1266  *    delete. The only way delete such policies - destroy or unconfigure
1267  *    corresponding virtual inteface.
1268  * 2) such policies have traffic selector that matches all traffic per
1269  *    address family.
1270  * Since all VTI policies have the same priority, we don't care about
1271  * policies order.
1272  */
1273 int
1274 key_register_ifnet(struct secpolicy **spp, u_int count)
1275 {
1276 	struct mbuf *m;
1277 	u_int i;
1278 
1279 	SPTREE_WLOCK();
1280 	/*
1281 	 * First of try to acquire id for each SP.
1282 	 */
1283 	for (i = 0; i < count; i++) {
1284 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1285 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1286 		    ("invalid direction %u", spp[i]->spidx.dir));
1287 
1288 		if ((spp[i]->id = key_getnewspid()) == 0) {
1289 			SPTREE_WUNLOCK();
1290 			return (EAGAIN);
1291 		}
1292 	}
1293 	for (i = 0; i < count; i++) {
1294 		TAILQ_INSERT_TAIL(&V_sptree_ifnet[spp[i]->spidx.dir],
1295 		    spp[i], chain);
1296 		/*
1297 		 * NOTE: despite the fact that we keep VTI SP in the
1298 		 * separate list, SPHASH contains policies from both
1299 		 * sources. Thus SADB_X_SPDGET will correctly return
1300 		 * SP by id, because it uses SPHASH for lookups.
1301 		 */
1302 		LIST_INSERT_HEAD(SPHASH_HASH(spp[i]->id), spp[i], idhash);
1303 		spp[i]->state = IPSEC_SPSTATE_IFNET;
1304 	}
1305 	SPTREE_WUNLOCK();
1306 	/*
1307 	 * Notify user processes about new SP.
1308 	 */
1309 	for (i = 0; i < count; i++) {
1310 		m = key_setdumpsp(spp[i], SADB_X_SPDADD, 0, 0);
1311 		if (m != NULL)
1312 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1313 	}
1314 	return (0);
1315 }
1316 
1317 void
1318 key_unregister_ifnet(struct secpolicy **spp, u_int count)
1319 {
1320 	struct mbuf *m;
1321 	u_int i;
1322 
1323 	SPTREE_WLOCK();
1324 	for (i = 0; i < count; i++) {
1325 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1326 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1327 		    ("invalid direction %u", spp[i]->spidx.dir));
1328 
1329 		if (spp[i]->state != IPSEC_SPSTATE_IFNET)
1330 			continue;
1331 		spp[i]->state = IPSEC_SPSTATE_DEAD;
1332 		TAILQ_REMOVE(&V_sptree_ifnet[spp[i]->spidx.dir],
1333 		    spp[i], chain);
1334 		V_spd_size--;
1335 		LIST_REMOVE(spp[i], idhash);
1336 	}
1337 	SPTREE_WUNLOCK();
1338 	if (SPDCACHE_ENABLED())
1339 		spdcache_clear();
1340 
1341 	for (i = 0; i < count; i++) {
1342 		m = key_setdumpsp(spp[i], SADB_X_SPDDELETE, 0, 0);
1343 		if (m != NULL)
1344 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1345 	}
1346 }
1347 
1348 /*
1349  * Must be called after calling key_allocsa().
1350  * This function is called by key_freesp() to free some SA allocated
1351  * for a policy.
1352  */
1353 void
1354 key_freesav(struct secasvar **psav)
1355 {
1356 	struct secasvar *sav = *psav;
1357 
1358 	IPSEC_ASSERT(sav != NULL, ("null sav"));
1359 	CURVNET_ASSERT_SET();
1360 	if (SAV_DELREF(sav) == 0)
1361 		return;
1362 
1363 	KEYDBG(IPSEC_STAMP,
1364 	    printf("%s: last reference to SA(%p)\n", __func__, sav));
1365 
1366 	*psav = NULL;
1367 	key_delsav(sav);
1368 }
1369 
1370 /*
1371  * Unlink SA from SAH and SPI hash under SAHTREE_WLOCK.
1372  * Expect that SA has extra reference due to lookup.
1373  * Release this references, also release SAH reference after unlink.
1374  */
1375 static void
1376 key_unlinksav(struct secasvar *sav)
1377 {
1378 	struct secashead *sah;
1379 
1380 	KEYDBG(KEY_STAMP,
1381 	    printf("%s: SA(%p)\n", __func__, sav));
1382 
1383 	CURVNET_ASSERT_SET();
1384 	SAHTREE_UNLOCK_ASSERT();
1385 	SAHTREE_WLOCK();
1386 	if (sav->state == SADB_SASTATE_DEAD) {
1387 		/* SA is already unlinked */
1388 		SAHTREE_WUNLOCK();
1389 		return;
1390 	}
1391 	/* Unlink from SAH */
1392 	if (sav->state == SADB_SASTATE_LARVAL)
1393 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
1394 	else
1395 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
1396 	/* Unlink from SPI hash */
1397 	LIST_REMOVE(sav, spihash);
1398 	sav->state = SADB_SASTATE_DEAD;
1399 	sah = sav->sah;
1400 	SAHTREE_WUNLOCK();
1401 	key_freesav(&sav);
1402 	/* Since we are unlinked, release reference to SAH */
1403 	key_freesah(&sah);
1404 }
1405 
1406 /* %%% SPD management */
1407 /*
1408  * search SPD
1409  * OUT:	NULL	: not found
1410  *	others	: found, pointer to a SP.
1411  */
1412 static struct secpolicy *
1413 key_getsp(struct secpolicyindex *spidx)
1414 {
1415 	SPTREE_RLOCK_TRACKER;
1416 	struct secpolicy *sp;
1417 
1418 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
1419 
1420 	SPTREE_RLOCK();
1421 	TAILQ_FOREACH(sp, &V_sptree[spidx->dir], chain) {
1422 		if (key_cmpspidx_exactly(spidx, &sp->spidx)) {
1423 			SP_ADDREF(sp);
1424 			break;
1425 		}
1426 	}
1427 	SPTREE_RUNLOCK();
1428 
1429 	return sp;
1430 }
1431 
1432 /*
1433  * get SP by index.
1434  * OUT:	NULL	: not found
1435  *	others	: found, pointer to referenced SP.
1436  */
1437 static struct secpolicy *
1438 key_getspbyid(uint32_t id)
1439 {
1440 	SPTREE_RLOCK_TRACKER;
1441 	struct secpolicy *sp;
1442 
1443 	SPTREE_RLOCK();
1444 	LIST_FOREACH(sp, SPHASH_HASH(id), idhash) {
1445 		if (sp->id == id) {
1446 			SP_ADDREF(sp);
1447 			break;
1448 		}
1449 	}
1450 	SPTREE_RUNLOCK();
1451 	return (sp);
1452 }
1453 
1454 struct secpolicy *
1455 key_newsp(void)
1456 {
1457 	struct secpolicy *sp;
1458 
1459 	sp = malloc(sizeof(*sp), M_IPSEC_SP, M_NOWAIT | M_ZERO);
1460 	if (sp != NULL)
1461 		SP_INITREF(sp);
1462 	return (sp);
1463 }
1464 
1465 struct ipsecrequest *
1466 ipsec_newisr(void)
1467 {
1468 
1469 	return (malloc(sizeof(struct ipsecrequest), M_IPSEC_SR,
1470 	    M_NOWAIT | M_ZERO));
1471 }
1472 
1473 void
1474 ipsec_delisr(struct ipsecrequest *p)
1475 {
1476 
1477 	free(p, M_IPSEC_SR);
1478 }
1479 
1480 /*
1481  * create secpolicy structure from sadb_x_policy structure.
1482  * NOTE: `state', `secpolicyindex' and 'id' in secpolicy structure
1483  * are not set, so must be set properly later.
1484  */
1485 struct secpolicy *
1486 key_msg2sp(struct sadb_x_policy *xpl0, size_t len, int *error)
1487 {
1488 	struct secpolicy *newsp;
1489 
1490 	IPSEC_ASSERT(xpl0 != NULL, ("null xpl0"));
1491 	IPSEC_ASSERT(len >= sizeof(*xpl0), ("policy too short: %zu", len));
1492 
1493 	if (len != PFKEY_EXTLEN(xpl0)) {
1494 		ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n", __func__));
1495 		*error = EINVAL;
1496 		return NULL;
1497 	}
1498 
1499 	if ((newsp = key_newsp()) == NULL) {
1500 		*error = ENOBUFS;
1501 		return NULL;
1502 	}
1503 
1504 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1505 	newsp->policy = xpl0->sadb_x_policy_type;
1506 	newsp->priority = xpl0->sadb_x_policy_priority;
1507 	newsp->tcount = 0;
1508 
1509 	/* check policy */
1510 	switch (xpl0->sadb_x_policy_type) {
1511 	case IPSEC_POLICY_DISCARD:
1512 	case IPSEC_POLICY_NONE:
1513 	case IPSEC_POLICY_ENTRUST:
1514 	case IPSEC_POLICY_BYPASS:
1515 		break;
1516 
1517 	case IPSEC_POLICY_IPSEC:
1518 	    {
1519 		struct sadb_x_ipsecrequest *xisr;
1520 		struct ipsecrequest *isr;
1521 		int tlen;
1522 
1523 		/* validity check */
1524 		if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1525 			ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n",
1526 				__func__));
1527 			key_freesp(&newsp);
1528 			*error = EINVAL;
1529 			return NULL;
1530 		}
1531 
1532 		tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1533 		xisr = (struct sadb_x_ipsecrequest *)(xpl0 + 1);
1534 
1535 		while (tlen > 0) {
1536 			/* length check */
1537 			if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr) ||
1538 			    xisr->sadb_x_ipsecrequest_len > tlen) {
1539 				ipseclog((LOG_DEBUG, "%s: invalid ipsecrequest "
1540 					"length.\n", __func__));
1541 				key_freesp(&newsp);
1542 				*error = EINVAL;
1543 				return NULL;
1544 			}
1545 
1546 			if (newsp->tcount >= IPSEC_MAXREQ) {
1547 				ipseclog((LOG_DEBUG,
1548 				    "%s: too many ipsecrequests.\n",
1549 				    __func__));
1550 				key_freesp(&newsp);
1551 				*error = EINVAL;
1552 				return (NULL);
1553 			}
1554 
1555 			/* allocate request buffer */
1556 			/* NB: data structure is zero'd */
1557 			isr = ipsec_newisr();
1558 			if (isr == NULL) {
1559 				ipseclog((LOG_DEBUG,
1560 				    "%s: No more memory.\n", __func__));
1561 				key_freesp(&newsp);
1562 				*error = ENOBUFS;
1563 				return NULL;
1564 			}
1565 
1566 			newsp->req[newsp->tcount++] = isr;
1567 
1568 			/* set values */
1569 			switch (xisr->sadb_x_ipsecrequest_proto) {
1570 			case IPPROTO_ESP:
1571 			case IPPROTO_AH:
1572 			case IPPROTO_IPCOMP:
1573 				break;
1574 			default:
1575 				ipseclog((LOG_DEBUG,
1576 				    "%s: invalid proto type=%u\n", __func__,
1577 				    xisr->sadb_x_ipsecrequest_proto));
1578 				key_freesp(&newsp);
1579 				*error = EPROTONOSUPPORT;
1580 				return NULL;
1581 			}
1582 			isr->saidx.proto =
1583 			    (uint8_t)xisr->sadb_x_ipsecrequest_proto;
1584 
1585 			switch (xisr->sadb_x_ipsecrequest_mode) {
1586 			case IPSEC_MODE_TRANSPORT:
1587 			case IPSEC_MODE_TUNNEL:
1588 				break;
1589 			case IPSEC_MODE_ANY:
1590 			default:
1591 				ipseclog((LOG_DEBUG,
1592 				    "%s: invalid mode=%u\n", __func__,
1593 				    xisr->sadb_x_ipsecrequest_mode));
1594 				key_freesp(&newsp);
1595 				*error = EINVAL;
1596 				return NULL;
1597 			}
1598 			isr->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1599 
1600 			switch (xisr->sadb_x_ipsecrequest_level) {
1601 			case IPSEC_LEVEL_DEFAULT:
1602 			case IPSEC_LEVEL_USE:
1603 			case IPSEC_LEVEL_REQUIRE:
1604 				break;
1605 			case IPSEC_LEVEL_UNIQUE:
1606 				/* validity check */
1607 				/*
1608 				 * If range violation of reqid, kernel will
1609 				 * update it, don't refuse it.
1610 				 */
1611 				if (xisr->sadb_x_ipsecrequest_reqid
1612 						> IPSEC_MANUAL_REQID_MAX) {
1613 					ipseclog((LOG_DEBUG,
1614 					    "%s: reqid=%d range "
1615 					    "violation, updated by kernel.\n",
1616 					    __func__,
1617 					    xisr->sadb_x_ipsecrequest_reqid));
1618 					xisr->sadb_x_ipsecrequest_reqid = 0;
1619 				}
1620 
1621 				/* allocate new reqid id if reqid is zero. */
1622 				if (xisr->sadb_x_ipsecrequest_reqid == 0) {
1623 					u_int32_t reqid;
1624 					if ((reqid = key_newreqid()) == 0) {
1625 						key_freesp(&newsp);
1626 						*error = ENOBUFS;
1627 						return NULL;
1628 					}
1629 					isr->saidx.reqid = reqid;
1630 					xisr->sadb_x_ipsecrequest_reqid = reqid;
1631 				} else {
1632 				/* set it for manual keying. */
1633 					isr->saidx.reqid =
1634 					    xisr->sadb_x_ipsecrequest_reqid;
1635 				}
1636 				break;
1637 
1638 			default:
1639 				ipseclog((LOG_DEBUG, "%s: invalid level=%u\n",
1640 					__func__,
1641 					xisr->sadb_x_ipsecrequest_level));
1642 				key_freesp(&newsp);
1643 				*error = EINVAL;
1644 				return NULL;
1645 			}
1646 			isr->level = xisr->sadb_x_ipsecrequest_level;
1647 
1648 			/* set IP addresses if there */
1649 			if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1650 				struct sockaddr *paddr;
1651 
1652 				len = tlen - sizeof(*xisr);
1653 				paddr = (struct sockaddr *)(xisr + 1);
1654 				/* validity check */
1655 				if (len < sizeof(struct sockaddr) ||
1656 				    len < 2 * paddr->sa_len ||
1657 				    paddr->sa_len > sizeof(isr->saidx.src)) {
1658 					ipseclog((LOG_DEBUG, "%s: invalid "
1659 						"request address length.\n",
1660 						__func__));
1661 					key_freesp(&newsp);
1662 					*error = EINVAL;
1663 					return NULL;
1664 				}
1665 				/*
1666 				 * Request length should be enough to keep
1667 				 * source and destination addresses.
1668 				 */
1669 				if (xisr->sadb_x_ipsecrequest_len <
1670 				    sizeof(*xisr) + 2 * paddr->sa_len) {
1671 					ipseclog((LOG_DEBUG, "%s: invalid "
1672 					    "ipsecrequest length.\n",
1673 					    __func__));
1674 					key_freesp(&newsp);
1675 					*error = EINVAL;
1676 					return (NULL);
1677 				}
1678 				bcopy(paddr, &isr->saidx.src, paddr->sa_len);
1679 				paddr = (struct sockaddr *)((caddr_t)paddr +
1680 				    paddr->sa_len);
1681 
1682 				/* validity check */
1683 				if (paddr->sa_len !=
1684 				    isr->saidx.src.sa.sa_len) {
1685 					ipseclog((LOG_DEBUG, "%s: invalid "
1686 						"request address length.\n",
1687 						__func__));
1688 					key_freesp(&newsp);
1689 					*error = EINVAL;
1690 					return NULL;
1691 				}
1692 				/* AF family should match */
1693 				if (paddr->sa_family !=
1694 				    isr->saidx.src.sa.sa_family) {
1695 					ipseclog((LOG_DEBUG, "%s: address "
1696 					    "family doesn't match.\n",
1697 						__func__));
1698 					key_freesp(&newsp);
1699 					*error = EINVAL;
1700 					return (NULL);
1701 				}
1702 				bcopy(paddr, &isr->saidx.dst, paddr->sa_len);
1703 			} else {
1704 				/*
1705 				 * Addresses for TUNNEL mode requests are
1706 				 * mandatory.
1707 				 */
1708 				if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1709 					ipseclog((LOG_DEBUG, "%s: missing "
1710 					    "request addresses.\n", __func__));
1711 					key_freesp(&newsp);
1712 					*error = EINVAL;
1713 					return (NULL);
1714 				}
1715 			}
1716 			tlen -= xisr->sadb_x_ipsecrequest_len;
1717 
1718 			/* validity check */
1719 			if (tlen < 0) {
1720 				ipseclog((LOG_DEBUG, "%s: becoming tlen < 0.\n",
1721 					__func__));
1722 				key_freesp(&newsp);
1723 				*error = EINVAL;
1724 				return NULL;
1725 			}
1726 
1727 			xisr = (struct sadb_x_ipsecrequest *)((caddr_t)xisr
1728 			                 + xisr->sadb_x_ipsecrequest_len);
1729 		}
1730 		/* XXXAE: LARVAL SP */
1731 		if (newsp->tcount < 1) {
1732 			ipseclog((LOG_DEBUG, "%s: valid IPSEC transforms "
1733 			    "not found.\n", __func__));
1734 			key_freesp(&newsp);
1735 			*error = EINVAL;
1736 			return (NULL);
1737 		}
1738 	    }
1739 		break;
1740 	default:
1741 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
1742 		key_freesp(&newsp);
1743 		*error = EINVAL;
1744 		return NULL;
1745 	}
1746 
1747 	*error = 0;
1748 	return (newsp);
1749 }
1750 
1751 uint32_t
1752 key_newreqid(void)
1753 {
1754 	static uint32_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1755 
1756 	if (auto_reqid == ~0)
1757 		auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1758 	else
1759 		auto_reqid++;
1760 
1761 	/* XXX should be unique check */
1762 	return (auto_reqid);
1763 }
1764 
1765 /*
1766  * copy secpolicy struct to sadb_x_policy structure indicated.
1767  */
1768 static struct mbuf *
1769 key_sp2mbuf(struct secpolicy *sp)
1770 {
1771 	struct mbuf *m;
1772 	size_t tlen;
1773 
1774 	tlen = key_getspreqmsglen(sp);
1775 	m = m_get2(tlen, M_NOWAIT, MT_DATA, 0);
1776 	if (m == NULL)
1777 		return (NULL);
1778 	m_align(m, tlen);
1779 	m->m_len = tlen;
1780 	if (key_sp2msg(sp, m->m_data, &tlen) != 0) {
1781 		m_freem(m);
1782 		return (NULL);
1783 	}
1784 	return (m);
1785 }
1786 
1787 int
1788 key_sp2msg(struct secpolicy *sp, void *request, size_t *len)
1789 {
1790 	struct sadb_x_ipsecrequest *xisr;
1791 	struct sadb_x_policy *xpl;
1792 	struct ipsecrequest *isr;
1793 	size_t xlen, ilen;
1794 	caddr_t p;
1795 	int error, i;
1796 
1797 	IPSEC_ASSERT(sp != NULL, ("null policy"));
1798 
1799 	xlen = sizeof(*xpl);
1800 	if (*len < xlen)
1801 		return (EINVAL);
1802 
1803 	error = 0;
1804 	bzero(request, *len);
1805 	xpl = (struct sadb_x_policy *)request;
1806 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
1807 	xpl->sadb_x_policy_type = sp->policy;
1808 	xpl->sadb_x_policy_dir = sp->spidx.dir;
1809 	xpl->sadb_x_policy_id = sp->id;
1810 	xpl->sadb_x_policy_priority = sp->priority;
1811 	switch (sp->state) {
1812 	case IPSEC_SPSTATE_IFNET:
1813 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_IFNET;
1814 		break;
1815 	case IPSEC_SPSTATE_PCB:
1816 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_PCB;
1817 		break;
1818 	default:
1819 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_GLOBAL;
1820 	}
1821 
1822 	/* if is the policy for ipsec ? */
1823 	if (sp->policy == IPSEC_POLICY_IPSEC) {
1824 		p = (caddr_t)xpl + sizeof(*xpl);
1825 		for (i = 0; i < sp->tcount; i++) {
1826 			isr = sp->req[i];
1827 			ilen = PFKEY_ALIGN8(sizeof(*xisr) +
1828 			    isr->saidx.src.sa.sa_len +
1829 			    isr->saidx.dst.sa.sa_len);
1830 			xlen += ilen;
1831 			if (xlen > *len) {
1832 				error = ENOBUFS;
1833 				/* Calculate needed size */
1834 				continue;
1835 			}
1836 			xisr = (struct sadb_x_ipsecrequest *)p;
1837 			xisr->sadb_x_ipsecrequest_len = ilen;
1838 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
1839 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
1840 			xisr->sadb_x_ipsecrequest_level = isr->level;
1841 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
1842 
1843 			p += sizeof(*xisr);
1844 			bcopy(&isr->saidx.src, p, isr->saidx.src.sa.sa_len);
1845 			p += isr->saidx.src.sa.sa_len;
1846 			bcopy(&isr->saidx.dst, p, isr->saidx.dst.sa.sa_len);
1847 			p += isr->saidx.dst.sa.sa_len;
1848 		}
1849 	}
1850 	xpl->sadb_x_policy_len = PFKEY_UNIT64(xlen);
1851 	if (error == 0)
1852 		*len = xlen;
1853 	else
1854 		*len = sizeof(*xpl);
1855 	return (error);
1856 }
1857 
1858 /* m will not be freed nor modified */
1859 static struct mbuf *
1860 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
1861     int ndeep, int nitem, ...)
1862 {
1863 	va_list ap;
1864 	int idx;
1865 	int i;
1866 	struct mbuf *result = NULL, *n;
1867 	int len;
1868 
1869 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1870 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1871 
1872 	va_start(ap, nitem);
1873 	for (i = 0; i < nitem; i++) {
1874 		idx = va_arg(ap, int);
1875 		if (idx < 0 || idx > SADB_EXT_MAX)
1876 			goto fail;
1877 		/* don't attempt to pull empty extension */
1878 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
1879 			continue;
1880 		if (idx != SADB_EXT_RESERVED  &&
1881 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
1882 			continue;
1883 
1884 		if (idx == SADB_EXT_RESERVED) {
1885 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
1886 
1887 			IPSEC_ASSERT(len <= MHLEN, ("header too big %u", len));
1888 
1889 			MGETHDR(n, M_NOWAIT, MT_DATA);
1890 			if (!n)
1891 				goto fail;
1892 			n->m_len = len;
1893 			n->m_next = NULL;
1894 			m_copydata(m, 0, sizeof(struct sadb_msg),
1895 			    mtod(n, caddr_t));
1896 		} else if (i < ndeep) {
1897 			len = mhp->extlen[idx];
1898 			n = m_get2(len, M_NOWAIT, MT_DATA, 0);
1899 			if (n == NULL)
1900 				goto fail;
1901 			m_align(n, len);
1902 			n->m_len = len;
1903 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
1904 			    mtod(n, caddr_t));
1905 		} else {
1906 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
1907 			    M_NOWAIT);
1908 		}
1909 		if (n == NULL)
1910 			goto fail;
1911 
1912 		if (result)
1913 			m_cat(result, n);
1914 		else
1915 			result = n;
1916 	}
1917 	va_end(ap);
1918 
1919 	if ((result->m_flags & M_PKTHDR) != 0) {
1920 		result->m_pkthdr.len = 0;
1921 		for (n = result; n; n = n->m_next)
1922 			result->m_pkthdr.len += n->m_len;
1923 	}
1924 
1925 	return result;
1926 
1927 fail:
1928 	m_freem(result);
1929 	va_end(ap);
1930 	return NULL;
1931 }
1932 
1933 /*
1934  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
1935  * add an entry to SP database, when received
1936  *   <base, address(SD), (lifetime(H),) policy>
1937  * from the user(?).
1938  * Adding to SP database,
1939  * and send
1940  *   <base, address(SD), (lifetime(H),) policy>
1941  * to the socket which was send.
1942  *
1943  * SPDADD set a unique policy entry.
1944  * SPDSETIDX like SPDADD without a part of policy requests.
1945  * SPDUPDATE replace a unique policy entry.
1946  *
1947  * XXXAE: serialize this in PF_KEY to avoid races.
1948  * m will always be freed.
1949  */
1950 static int
1951 key_spdadd(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
1952 {
1953 	struct secpolicyindex spidx;
1954 	struct sadb_address *src0, *dst0;
1955 	struct sadb_x_policy *xpl0, *xpl;
1956 	struct sadb_lifetime *lft = NULL;
1957 	struct secpolicy *newsp, *oldsp;
1958 	int error;
1959 
1960 	IPSEC_ASSERT(so != NULL, ("null socket"));
1961 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1962 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1963 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
1964 
1965 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
1966 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
1967 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
1968 		ipseclog((LOG_DEBUG,
1969 		    "%s: invalid message: missing required header.\n",
1970 		    __func__));
1971 		return key_senderror(so, m, EINVAL);
1972 	}
1973 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
1974 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
1975 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
1976 		ipseclog((LOG_DEBUG,
1977 		    "%s: invalid message: wrong header size.\n", __func__));
1978 		return key_senderror(so, m, EINVAL);
1979 	}
1980 	if (!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD)) {
1981 		if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD)) {
1982 			ipseclog((LOG_DEBUG,
1983 			    "%s: invalid message: wrong header size.\n",
1984 			    __func__));
1985 			return key_senderror(so, m, EINVAL);
1986 		}
1987 		lft = (struct sadb_lifetime *)mhp->ext[SADB_EXT_LIFETIME_HARD];
1988 	}
1989 
1990 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
1991 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
1992 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
1993 
1994 	/* check the direciton */
1995 	switch (xpl0->sadb_x_policy_dir) {
1996 	case IPSEC_DIR_INBOUND:
1997 	case IPSEC_DIR_OUTBOUND:
1998 		break;
1999 	default:
2000 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2001 		return key_senderror(so, m, EINVAL);
2002 	}
2003 	/* key_spdadd() accepts DISCARD, NONE and IPSEC. */
2004 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2005 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2006 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2007 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2008 		return key_senderror(so, m, EINVAL);
2009 	}
2010 
2011 	/* policy requests are mandatory when action is ipsec. */
2012 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
2013 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
2014 		ipseclog((LOG_DEBUG,
2015 		    "%s: policy requests required.\n", __func__));
2016 		return key_senderror(so, m, EINVAL);
2017 	}
2018 
2019 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
2020 	    (struct sockaddr *)(dst0 + 1));
2021 	if (error != 0 ||
2022 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2023 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2024 		return key_senderror(so, m, error);
2025 	}
2026 	/* make secindex */
2027 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2028 	                src0 + 1,
2029 	                dst0 + 1,
2030 	                src0->sadb_address_prefixlen,
2031 	                dst0->sadb_address_prefixlen,
2032 	                src0->sadb_address_proto,
2033 	                &spidx);
2034 	/* Checking there is SP already or not. */
2035 	oldsp = key_getsp(&spidx);
2036 	if (oldsp != NULL) {
2037 		if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2038 			KEYDBG(KEY_STAMP,
2039 			    printf("%s: unlink SP(%p) for SPDUPDATE\n",
2040 				__func__, oldsp));
2041 			KEYDBG(KEY_DATA, kdebug_secpolicy(oldsp));
2042 		} else {
2043 			key_freesp(&oldsp);
2044 			ipseclog((LOG_DEBUG,
2045 			    "%s: a SP entry exists already.\n", __func__));
2046 			return (key_senderror(so, m, EEXIST));
2047 		}
2048 	}
2049 
2050 	/* allocate new SP entry */
2051 	if ((newsp = key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error)) == NULL) {
2052 		if (oldsp != NULL) {
2053 			key_unlink(oldsp);
2054 			key_freesp(&oldsp); /* second for our reference */
2055 		}
2056 		return key_senderror(so, m, error);
2057 	}
2058 
2059 	newsp->lastused = newsp->created = time_second;
2060 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
2061 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
2062 	bcopy(&spidx, &newsp->spidx, sizeof(spidx));
2063 
2064 	SPTREE_WLOCK();
2065 	if ((newsp->id = key_getnewspid()) == 0) {
2066 		if (oldsp != NULL)
2067 			key_detach(oldsp);
2068 		SPTREE_WUNLOCK();
2069 		if (oldsp != NULL) {
2070 			key_freesp(&oldsp); /* first for key_detach */
2071 			IPSEC_ASSERT(oldsp != NULL, ("null oldsp: refcount bug"));
2072 			key_freesp(&oldsp); /* second for our reference */
2073 			if (SPDCACHE_ENABLED()) /* refresh cache because of key_detach */
2074 				spdcache_clear();
2075 		}
2076 		key_freesp(&newsp);
2077 		return key_senderror(so, m, ENOBUFS);
2078 	}
2079 	if (oldsp != NULL)
2080 		key_detach(oldsp);
2081 	key_insertsp(newsp);
2082 	SPTREE_WUNLOCK();
2083 	if (oldsp != NULL) {
2084 		key_freesp(&oldsp); /* first for key_detach */
2085 		IPSEC_ASSERT(oldsp != NULL, ("null oldsp: refcount bug"));
2086 		key_freesp(&oldsp); /* second for our reference */
2087 	}
2088 	if (SPDCACHE_ENABLED())
2089 		spdcache_clear();
2090 	KEYDBG(KEY_STAMP,
2091 	    printf("%s: SP(%p)\n", __func__, newsp));
2092 	KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
2093 
2094     {
2095 	struct mbuf *n, *mpolicy;
2096 	struct sadb_msg *newmsg;
2097 	int off;
2098 
2099 	/* create new sadb_msg to reply. */
2100 	if (lft) {
2101 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
2102 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
2103 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2104 	} else {
2105 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
2106 		    SADB_X_EXT_POLICY,
2107 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2108 	}
2109 	if (!n)
2110 		return key_senderror(so, m, ENOBUFS);
2111 
2112 	if (n->m_len < sizeof(*newmsg)) {
2113 		n = m_pullup(n, sizeof(*newmsg));
2114 		if (!n)
2115 			return key_senderror(so, m, ENOBUFS);
2116 	}
2117 	newmsg = mtod(n, struct sadb_msg *);
2118 	newmsg->sadb_msg_errno = 0;
2119 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2120 
2121 	off = 0;
2122 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
2123 	    sizeof(*xpl), &off);
2124 	if (mpolicy == NULL) {
2125 		/* n is already freed */
2126 		return key_senderror(so, m, ENOBUFS);
2127 	}
2128 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, caddr_t) + off);
2129 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
2130 		m_freem(n);
2131 		return key_senderror(so, m, EINVAL);
2132 	}
2133 	xpl->sadb_x_policy_id = newsp->id;
2134 
2135 	m_freem(m);
2136 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2137     }
2138 }
2139 
2140 /*
2141  * get new policy id.
2142  * OUT:
2143  *	0:	failure.
2144  *	others: success.
2145  */
2146 static uint32_t
2147 key_getnewspid(void)
2148 {
2149 	struct secpolicy *sp;
2150 	uint32_t newid = 0;
2151 	int tries, limit;
2152 
2153 	SPTREE_WLOCK_ASSERT();
2154 
2155 	limit = atomic_load_int(&V_key_spi_trycnt);
2156 	for (tries = 0; tries < limit; tries++) {
2157 		if (V_policy_id == ~0) /* overflowed */
2158 			newid = V_policy_id = 1;
2159 		else
2160 			newid = ++V_policy_id;
2161 		LIST_FOREACH(sp, SPHASH_HASH(newid), idhash) {
2162 			if (sp->id == newid)
2163 				break;
2164 		}
2165 		if (sp == NULL)
2166 			break;
2167 	}
2168 	if (tries == limit || newid == 0) {
2169 		ipseclog((LOG_DEBUG, "%s: failed to allocate policy id.\n",
2170 		    __func__));
2171 		return (0);
2172 	}
2173 	return (newid);
2174 }
2175 
2176 /*
2177  * SADB_SPDDELETE processing
2178  * receive
2179  *   <base, address(SD), policy(*)>
2180  * from the user(?), and set SADB_SASTATE_DEAD,
2181  * and send,
2182  *   <base, address(SD), policy(*)>
2183  * to the ikmpd.
2184  * policy(*) including direction of policy.
2185  *
2186  * m will always be freed.
2187  */
2188 static int
2189 key_spddelete(struct socket *so, struct mbuf *m,
2190     const struct sadb_msghdr *mhp)
2191 {
2192 	struct secpolicyindex spidx;
2193 	struct sadb_address *src0, *dst0;
2194 	struct sadb_x_policy *xpl0;
2195 	struct secpolicy *sp;
2196 
2197 	IPSEC_ASSERT(so != NULL, ("null so"));
2198 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2199 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2200 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2201 
2202 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
2203 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
2204 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
2205 		ipseclog((LOG_DEBUG,
2206 		    "%s: invalid message: missing required header.\n",
2207 		    __func__));
2208 		return key_senderror(so, m, EINVAL);
2209 	}
2210 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
2211 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
2212 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2213 		ipseclog((LOG_DEBUG,
2214 		    "%s: invalid message: wrong header size.\n", __func__));
2215 		return key_senderror(so, m, EINVAL);
2216 	}
2217 
2218 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
2219 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
2220 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
2221 
2222 	/* check the direciton */
2223 	switch (xpl0->sadb_x_policy_dir) {
2224 	case IPSEC_DIR_INBOUND:
2225 	case IPSEC_DIR_OUTBOUND:
2226 		break;
2227 	default:
2228 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2229 		return key_senderror(so, m, EINVAL);
2230 	}
2231 	/* Only DISCARD, NONE and IPSEC are allowed */
2232 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2233 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2234 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2235 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2236 		return key_senderror(so, m, EINVAL);
2237 	}
2238 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
2239 	    (struct sockaddr *)(dst0 + 1)) != 0 ||
2240 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2241 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2242 		return key_senderror(so, m, EINVAL);
2243 	}
2244 	/* make secindex */
2245 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2246 	                src0 + 1,
2247 	                dst0 + 1,
2248 	                src0->sadb_address_prefixlen,
2249 	                dst0->sadb_address_prefixlen,
2250 	                src0->sadb_address_proto,
2251 	                &spidx);
2252 
2253 	/* Is there SP in SPD ? */
2254 	if ((sp = key_getsp(&spidx)) == NULL) {
2255 		ipseclog((LOG_DEBUG, "%s: no SP found.\n", __func__));
2256 		return key_senderror(so, m, EINVAL);
2257 	}
2258 
2259 	/* save policy id to buffer to be returned. */
2260 	xpl0->sadb_x_policy_id = sp->id;
2261 
2262 	KEYDBG(KEY_STAMP,
2263 	    printf("%s: SP(%p)\n", __func__, sp));
2264 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2265 	key_unlink(sp);
2266 	key_freesp(&sp);
2267 
2268     {
2269 	struct mbuf *n;
2270 	struct sadb_msg *newmsg;
2271 
2272 	/* create new sadb_msg to reply. */
2273 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2274 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2275 	if (!n)
2276 		return key_senderror(so, m, ENOBUFS);
2277 
2278 	newmsg = mtod(n, struct sadb_msg *);
2279 	newmsg->sadb_msg_errno = 0;
2280 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2281 
2282 	m_freem(m);
2283 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2284     }
2285 }
2286 
2287 /*
2288  * SADB_SPDDELETE2 processing
2289  * receive
2290  *   <base, policy(*)>
2291  * from the user(?), and set SADB_SASTATE_DEAD,
2292  * and send,
2293  *   <base, policy(*)>
2294  * to the ikmpd.
2295  * policy(*) including direction of policy.
2296  *
2297  * m will always be freed.
2298  */
2299 static int
2300 key_spddelete2(struct socket *so, struct mbuf *m,
2301     const struct sadb_msghdr *mhp)
2302 {
2303 	struct secpolicy *sp;
2304 	uint32_t id;
2305 
2306 	IPSEC_ASSERT(so != NULL, ("null socket"));
2307 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2308 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2309 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2310 
2311 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2312 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2313 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2314 		    __func__));
2315 		return key_senderror(so, m, EINVAL);
2316 	}
2317 
2318 	id = ((struct sadb_x_policy *)
2319 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2320 
2321 	/* Is there SP in SPD ? */
2322 	if ((sp = key_getspbyid(id)) == NULL) {
2323 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2324 		    __func__, id));
2325 		return key_senderror(so, m, EINVAL);
2326 	}
2327 
2328 	KEYDBG(KEY_STAMP,
2329 	    printf("%s: SP(%p)\n", __func__, sp));
2330 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2331 	key_unlink(sp);
2332 	if (sp->state != IPSEC_SPSTATE_DEAD) {
2333 		ipseclog((LOG_DEBUG, "%s: failed to delete SP with id %u.\n",
2334 		    __func__, id));
2335 		key_freesp(&sp);
2336 		return (key_senderror(so, m, EACCES));
2337 	}
2338 	key_freesp(&sp);
2339 
2340     {
2341 	struct mbuf *n, *nn;
2342 	struct sadb_msg *newmsg;
2343 	int off, len;
2344 
2345 	/* create new sadb_msg to reply. */
2346 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2347 
2348 	MGETHDR(n, M_NOWAIT, MT_DATA);
2349 	if (n && len > MHLEN) {
2350 		if (!(MCLGET(n, M_NOWAIT))) {
2351 			m_freem(n);
2352 			n = NULL;
2353 		}
2354 	}
2355 	if (!n)
2356 		return key_senderror(so, m, ENOBUFS);
2357 
2358 	n->m_len = len;
2359 	n->m_next = NULL;
2360 	off = 0;
2361 
2362 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
2363 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2364 
2365 	IPSEC_ASSERT(off == len, ("length inconsistency (off %u len %u)",
2366 		off, len));
2367 
2368 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2369 	    mhp->extlen[SADB_X_EXT_POLICY], M_NOWAIT);
2370 	if (!n->m_next) {
2371 		m_freem(n);
2372 		return key_senderror(so, m, ENOBUFS);
2373 	}
2374 
2375 	n->m_pkthdr.len = 0;
2376 	for (nn = n; nn; nn = nn->m_next)
2377 		n->m_pkthdr.len += nn->m_len;
2378 
2379 	newmsg = mtod(n, struct sadb_msg *);
2380 	newmsg->sadb_msg_errno = 0;
2381 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2382 
2383 	m_freem(m);
2384 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2385     }
2386 }
2387 
2388 /*
2389  * SADB_X_SPDGET processing
2390  * receive
2391  *   <base, policy(*)>
2392  * from the user(?),
2393  * and send,
2394  *   <base, address(SD), policy>
2395  * to the ikmpd.
2396  * policy(*) including direction of policy.
2397  *
2398  * m will always be freed.
2399  */
2400 static int
2401 key_spdget(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2402 {
2403 	struct secpolicy *sp;
2404 	struct mbuf *n;
2405 	uint32_t id;
2406 
2407 	IPSEC_ASSERT(so != NULL, ("null socket"));
2408 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2409 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2410 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2411 
2412 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2413 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2414 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2415 		    __func__));
2416 		return key_senderror(so, m, EINVAL);
2417 	}
2418 
2419 	id = ((struct sadb_x_policy *)
2420 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2421 
2422 	/* Is there SP in SPD ? */
2423 	if ((sp = key_getspbyid(id)) == NULL) {
2424 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2425 		    __func__, id));
2426 		return key_senderror(so, m, ENOENT);
2427 	}
2428 
2429 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2430 	    mhp->msg->sadb_msg_pid);
2431 	key_freesp(&sp);
2432 	if (n != NULL) {
2433 		m_freem(m);
2434 		return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2435 	} else
2436 		return key_senderror(so, m, ENOBUFS);
2437 }
2438 
2439 /*
2440  * SADB_X_SPDACQUIRE processing.
2441  * Acquire policy and SA(s) for a *OUTBOUND* packet.
2442  * send
2443  *   <base, policy(*)>
2444  * to KMD, and expect to receive
2445  *   <base> with SADB_X_SPDACQUIRE if error occurred,
2446  * or
2447  *   <base, policy>
2448  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2449  * policy(*) is without policy requests.
2450  *
2451  *    0     : succeed
2452  *    others: error number
2453  */
2454 int
2455 key_spdacquire(struct secpolicy *sp)
2456 {
2457 	struct mbuf *result = NULL, *m;
2458 	struct secspacq *newspacq;
2459 
2460 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2461 	IPSEC_ASSERT(sp->req == NULL, ("policy exists"));
2462 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
2463 		("policy not IPSEC %u", sp->policy));
2464 
2465 	/* Get an entry to check whether sent message or not. */
2466 	newspacq = key_getspacq(&sp->spidx);
2467 	if (newspacq != NULL) {
2468 		if (V_key_blockacq_count < newspacq->count) {
2469 			/* reset counter and do send message. */
2470 			newspacq->count = 0;
2471 		} else {
2472 			/* increment counter and do nothing. */
2473 			newspacq->count++;
2474 			SPACQ_UNLOCK();
2475 			return (0);
2476 		}
2477 		SPACQ_UNLOCK();
2478 	} else {
2479 		/* make new entry for blocking to send SADB_ACQUIRE. */
2480 		newspacq = key_newspacq(&sp->spidx);
2481 		if (newspacq == NULL)
2482 			return ENOBUFS;
2483 	}
2484 
2485 	/* create new sadb_msg to reply. */
2486 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2487 	if (!m)
2488 		return ENOBUFS;
2489 
2490 	result = m;
2491 
2492 	result->m_pkthdr.len = 0;
2493 	for (m = result; m; m = m->m_next)
2494 		result->m_pkthdr.len += m->m_len;
2495 
2496 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2497 	    PFKEY_UNIT64(result->m_pkthdr.len);
2498 
2499 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2500 }
2501 
2502 /*
2503  * SADB_SPDFLUSH processing
2504  * receive
2505  *   <base>
2506  * from the user, and free all entries in secpctree.
2507  * and send,
2508  *   <base>
2509  * to the user.
2510  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2511  *
2512  * m will always be freed.
2513  */
2514 static int
2515 key_spdflush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2516 {
2517 	struct secpolicy_queue drainq;
2518 	struct sadb_msg *newmsg;
2519 	struct secpolicy *sp, *nextsp;
2520 	u_int dir;
2521 
2522 	IPSEC_ASSERT(so != NULL, ("null socket"));
2523 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2524 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2525 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2526 
2527 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2528 		return key_senderror(so, m, EINVAL);
2529 
2530 	TAILQ_INIT(&drainq);
2531 	SPTREE_WLOCK();
2532 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2533 		TAILQ_CONCAT(&drainq, &V_sptree[dir], chain);
2534 	}
2535 	/*
2536 	 * We need to set state to DEAD for each policy to be sure,
2537 	 * that another thread won't try to unlink it.
2538 	 * Also remove SP from sphash.
2539 	 */
2540 	TAILQ_FOREACH(sp, &drainq, chain) {
2541 		sp->state = IPSEC_SPSTATE_DEAD;
2542 		LIST_REMOVE(sp, idhash);
2543 	}
2544 	V_sp_genid++;
2545 	V_spd_size = 0;
2546 	SPTREE_WUNLOCK();
2547 	if (SPDCACHE_ENABLED())
2548 		spdcache_clear();
2549 	sp = TAILQ_FIRST(&drainq);
2550 	while (sp != NULL) {
2551 		nextsp = TAILQ_NEXT(sp, chain);
2552 		key_freesp(&sp);
2553 		sp = nextsp;
2554 	}
2555 
2556 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2557 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
2558 		return key_senderror(so, m, ENOBUFS);
2559 	}
2560 
2561 	if (m->m_next)
2562 		m_freem(m->m_next);
2563 	m->m_next = NULL;
2564 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2565 	newmsg = mtod(m, struct sadb_msg *);
2566 	newmsg->sadb_msg_errno = 0;
2567 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2568 
2569 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2570 }
2571 
2572 static uint8_t
2573 key_satype2scopemask(uint8_t satype)
2574 {
2575 
2576 	if (satype == IPSEC_POLICYSCOPE_ANY)
2577 		return (0xff);
2578 	return (satype);
2579 }
2580 /*
2581  * SADB_SPDDUMP processing
2582  * receive
2583  *   <base>
2584  * from the user, and dump all SP leaves and send,
2585  *   <base> .....
2586  * to the ikmpd.
2587  *
2588  * NOTE:
2589  *   sadb_msg_satype is considered as mask of policy scopes.
2590  *   m will always be freed.
2591  */
2592 static int
2593 key_spddump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2594 {
2595 	SPTREE_RLOCK_TRACKER;
2596 	struct secpolicy *sp;
2597 	struct mbuf *n;
2598 	int cnt;
2599 	u_int dir, scope;
2600 
2601 	IPSEC_ASSERT(so != NULL, ("null socket"));
2602 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2603 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2604 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2605 
2606 	/* search SPD entry and get buffer size. */
2607 	cnt = 0;
2608 	scope = key_satype2scopemask(mhp->msg->sadb_msg_satype);
2609 	SPTREE_RLOCK();
2610 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2611 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2612 			TAILQ_FOREACH(sp, &V_sptree[dir], chain)
2613 				cnt++;
2614 		}
2615 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2616 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain)
2617 				cnt++;
2618 		}
2619 	}
2620 
2621 	if (cnt == 0) {
2622 		SPTREE_RUNLOCK();
2623 		return key_senderror(so, m, ENOENT);
2624 	}
2625 
2626 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2627 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2628 			TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
2629 				--cnt;
2630 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2631 				    mhp->msg->sadb_msg_pid);
2632 
2633 				if (n != NULL)
2634 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2635 			}
2636 		}
2637 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2638 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain) {
2639 				--cnt;
2640 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2641 				    mhp->msg->sadb_msg_pid);
2642 
2643 				if (n != NULL)
2644 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2645 			}
2646 		}
2647 	}
2648 
2649 	SPTREE_RUNLOCK();
2650 	m_freem(m);
2651 	return (0);
2652 }
2653 
2654 static struct mbuf *
2655 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq,
2656     u_int32_t pid)
2657 {
2658 	struct mbuf *result = NULL, *m;
2659 	struct seclifetime lt;
2660 
2661 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid, sp->refcnt);
2662 	if (!m)
2663 		goto fail;
2664 	result = m;
2665 
2666 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2667 	    &sp->spidx.src.sa, sp->spidx.prefs,
2668 	    sp->spidx.ul_proto);
2669 	if (!m)
2670 		goto fail;
2671 	m_cat(result, m);
2672 
2673 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2674 	    &sp->spidx.dst.sa, sp->spidx.prefd,
2675 	    sp->spidx.ul_proto);
2676 	if (!m)
2677 		goto fail;
2678 	m_cat(result, m);
2679 
2680 	m = key_sp2mbuf(sp);
2681 	if (!m)
2682 		goto fail;
2683 	m_cat(result, m);
2684 
2685 	if(sp->lifetime){
2686 		lt.addtime=sp->created;
2687 		lt.usetime= sp->lastused;
2688 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_CURRENT);
2689 		if (!m)
2690 			goto fail;
2691 		m_cat(result, m);
2692 
2693 		lt.addtime=sp->lifetime;
2694 		lt.usetime= sp->validtime;
2695 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_HARD);
2696 		if (!m)
2697 			goto fail;
2698 		m_cat(result, m);
2699 	}
2700 
2701 	if ((result->m_flags & M_PKTHDR) == 0)
2702 		goto fail;
2703 
2704 	if (result->m_len < sizeof(struct sadb_msg)) {
2705 		result = m_pullup(result, sizeof(struct sadb_msg));
2706 		if (result == NULL)
2707 			goto fail;
2708 	}
2709 
2710 	result->m_pkthdr.len = 0;
2711 	for (m = result; m; m = m->m_next)
2712 		result->m_pkthdr.len += m->m_len;
2713 
2714 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2715 	    PFKEY_UNIT64(result->m_pkthdr.len);
2716 
2717 	return result;
2718 
2719 fail:
2720 	m_freem(result);
2721 	return NULL;
2722 }
2723 /*
2724  * get PFKEY message length for security policy and request.
2725  */
2726 static size_t
2727 key_getspreqmsglen(struct secpolicy *sp)
2728 {
2729 	size_t tlen, len;
2730 	int i;
2731 
2732 	tlen = sizeof(struct sadb_x_policy);
2733 	/* if is the policy for ipsec ? */
2734 	if (sp->policy != IPSEC_POLICY_IPSEC)
2735 		return (tlen);
2736 
2737 	/* get length of ipsec requests */
2738 	for (i = 0; i < sp->tcount; i++) {
2739 		len = sizeof(struct sadb_x_ipsecrequest)
2740 			+ sp->req[i]->saidx.src.sa.sa_len
2741 			+ sp->req[i]->saidx.dst.sa.sa_len;
2742 
2743 		tlen += PFKEY_ALIGN8(len);
2744 	}
2745 	return (tlen);
2746 }
2747 
2748 /*
2749  * SADB_SPDEXPIRE processing
2750  * send
2751  *   <base, address(SD), lifetime(CH), policy>
2752  * to KMD by PF_KEY.
2753  *
2754  * OUT:	0	: succeed
2755  *	others	: error number
2756  */
2757 static int
2758 key_spdexpire(struct secpolicy *sp)
2759 {
2760 	struct sadb_lifetime *lt;
2761 	struct mbuf *result = NULL, *m;
2762 	int len, error = -1;
2763 
2764 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2765 
2766 	KEYDBG(KEY_STAMP,
2767 	    printf("%s: SP(%p)\n", __func__, sp));
2768 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2769 
2770 	/* set msg header */
2771 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0);
2772 	if (!m) {
2773 		error = ENOBUFS;
2774 		goto fail;
2775 	}
2776 	result = m;
2777 
2778 	/* create lifetime extension (current and hard) */
2779 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
2780 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
2781 	if (m == NULL) {
2782 		error = ENOBUFS;
2783 		goto fail;
2784 	}
2785 	m_align(m, len);
2786 	m->m_len = len;
2787 	bzero(mtod(m, caddr_t), len);
2788 	lt = mtod(m, struct sadb_lifetime *);
2789 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2790 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
2791 	lt->sadb_lifetime_allocations = 0;
2792 	lt->sadb_lifetime_bytes = 0;
2793 	lt->sadb_lifetime_addtime = sp->created;
2794 	lt->sadb_lifetime_usetime = sp->lastused;
2795 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
2796 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2797 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
2798 	lt->sadb_lifetime_allocations = 0;
2799 	lt->sadb_lifetime_bytes = 0;
2800 	lt->sadb_lifetime_addtime = sp->lifetime;
2801 	lt->sadb_lifetime_usetime = sp->validtime;
2802 	m_cat(result, m);
2803 
2804 	/* set sadb_address for source */
2805 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2806 	    &sp->spidx.src.sa,
2807 	    sp->spidx.prefs, sp->spidx.ul_proto);
2808 	if (!m) {
2809 		error = ENOBUFS;
2810 		goto fail;
2811 	}
2812 	m_cat(result, m);
2813 
2814 	/* set sadb_address for destination */
2815 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2816 	    &sp->spidx.dst.sa,
2817 	    sp->spidx.prefd, sp->spidx.ul_proto);
2818 	if (!m) {
2819 		error = ENOBUFS;
2820 		goto fail;
2821 	}
2822 	m_cat(result, m);
2823 
2824 	/* set secpolicy */
2825 	m = key_sp2mbuf(sp);
2826 	if (!m) {
2827 		error = ENOBUFS;
2828 		goto fail;
2829 	}
2830 	m_cat(result, m);
2831 
2832 	if ((result->m_flags & M_PKTHDR) == 0) {
2833 		error = EINVAL;
2834 		goto fail;
2835 	}
2836 
2837 	if (result->m_len < sizeof(struct sadb_msg)) {
2838 		result = m_pullup(result, sizeof(struct sadb_msg));
2839 		if (result == NULL) {
2840 			error = ENOBUFS;
2841 			goto fail;
2842 		}
2843 	}
2844 
2845 	result->m_pkthdr.len = 0;
2846 	for (m = result; m; m = m->m_next)
2847 		result->m_pkthdr.len += m->m_len;
2848 
2849 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2850 	    PFKEY_UNIT64(result->m_pkthdr.len);
2851 
2852 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
2853 
2854  fail:
2855 	if (result)
2856 		m_freem(result);
2857 	return error;
2858 }
2859 
2860 /* %%% SAD management */
2861 /*
2862  * allocating and initialize new SA head.
2863  * OUT:	NULL	: failure due to the lack of memory.
2864  *	others	: pointer to new SA head.
2865  */
2866 static struct secashead *
2867 key_newsah(struct secasindex *saidx)
2868 {
2869 	struct secashead *sah;
2870 
2871 	sah = malloc(sizeof(struct secashead), M_IPSEC_SAH,
2872 	    M_NOWAIT | M_ZERO);
2873 	if (sah == NULL) {
2874 		PFKEYSTAT_INC(in_nomem);
2875 		return (NULL);
2876 	}
2877 	TAILQ_INIT(&sah->savtree_larval);
2878 	TAILQ_INIT(&sah->savtree_alive);
2879 	sah->saidx = *saidx;
2880 	sah->state = SADB_SASTATE_DEAD;
2881 	SAH_INITREF(sah);
2882 
2883 	KEYDBG(KEY_STAMP,
2884 	    printf("%s: SAH(%p)\n", __func__, sah));
2885 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2886 	return (sah);
2887 }
2888 
2889 static void
2890 key_freesah(struct secashead **psah)
2891 {
2892 	struct secashead *sah = *psah;
2893 
2894 	CURVNET_ASSERT_SET();
2895 
2896 	if (SAH_DELREF(sah) == 0)
2897 		return;
2898 
2899 	KEYDBG(KEY_STAMP,
2900 	    printf("%s: last reference to SAH(%p)\n", __func__, sah));
2901 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2902 
2903 	*psah = NULL;
2904 	key_delsah(sah);
2905 }
2906 
2907 static void
2908 key_delsah(struct secashead *sah)
2909 {
2910 	IPSEC_ASSERT(sah != NULL, ("NULL sah"));
2911 	IPSEC_ASSERT(sah->state == SADB_SASTATE_DEAD,
2912 	    ("Attempt to free non DEAD SAH %p", sah));
2913 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_larval),
2914 	    ("Attempt to free SAH %p with LARVAL SA", sah));
2915 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_alive),
2916 	    ("Attempt to free SAH %p with ALIVE SA", sah));
2917 
2918 	free(sah, M_IPSEC_SAH);
2919 }
2920 
2921 /*
2922  * allocating a new SA for key_add() and key_getspi() call,
2923  * and copy the values of mhp into new buffer.
2924  * When SAD message type is SADB_GETSPI set SA state to LARVAL.
2925  * For SADB_ADD create and initialize SA with MATURE state.
2926  * OUT:	NULL	: fail
2927  *	others	: pointer to new secasvar.
2928  */
2929 static struct secasvar *
2930 key_newsav(const struct sadb_msghdr *mhp, struct secasindex *saidx,
2931     uint32_t spi, int *errp)
2932 {
2933 	struct secashead *sah;
2934 	struct secasvar *sav;
2935 	int isnew;
2936 
2937 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2938 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2939 	IPSEC_ASSERT(mhp->msg->sadb_msg_type == SADB_GETSPI ||
2940 	    mhp->msg->sadb_msg_type == SADB_ADD, ("wrong message type"));
2941 
2942 	sav = NULL;
2943 	sah = NULL;
2944 	/* check SPI value */
2945 	switch (saidx->proto) {
2946 	case IPPROTO_ESP:
2947 	case IPPROTO_AH:
2948 		/*
2949 		 * RFC 4302, 2.4. Security Parameters Index (SPI), SPI values
2950 		 * 1-255 reserved by IANA for future use,
2951 		 * 0 for implementation specific, local use.
2952 		 */
2953 		if (ntohl(spi) <= 255) {
2954 			ipseclog((LOG_DEBUG, "%s: illegal range of SPI %u.\n",
2955 			    __func__, ntohl(spi)));
2956 			*errp = EINVAL;
2957 			goto done;
2958 		}
2959 		break;
2960 	}
2961 
2962 	sav = malloc(sizeof(struct secasvar), M_IPSEC_SA, M_NOWAIT | M_ZERO);
2963 	if (sav == NULL) {
2964 		*errp = ENOBUFS;
2965 		goto done;
2966 	}
2967 	sav->lock = malloc_aligned(max(sizeof(struct rmlock),
2968 	    CACHE_LINE_SIZE), CACHE_LINE_SIZE, M_IPSEC_MISC,
2969 	    M_NOWAIT | M_ZERO);
2970 	if (sav->lock == NULL) {
2971 		*errp = ENOBUFS;
2972 		goto done;
2973 	}
2974 	rm_init(sav->lock, "ipsec association");
2975 	sav->lft_c = uma_zalloc_pcpu(ipsec_key_lft_zone, M_NOWAIT | M_ZERO);
2976 	if (sav->lft_c == NULL) {
2977 		*errp = ENOBUFS;
2978 		goto done;
2979 	}
2980 
2981 	sav->spi = spi;
2982 	sav->seq = mhp->msg->sadb_msg_seq;
2983 	sav->state = SADB_SASTATE_LARVAL;
2984 	sav->pid = (pid_t)mhp->msg->sadb_msg_pid;
2985 	SAV_INITREF(sav);
2986 again:
2987 	sah = key_getsah(saidx);
2988 	if (sah == NULL) {
2989 		/* create a new SA index */
2990 		sah = key_newsah(saidx);
2991 		if (sah == NULL) {
2992 			ipseclog((LOG_DEBUG,
2993 			    "%s: No more memory.\n", __func__));
2994 			*errp = ENOBUFS;
2995 			goto done;
2996 		}
2997 		isnew = 1;
2998 	} else
2999 		isnew = 0;
3000 
3001 	sav->sah = sah;
3002 	if (mhp->msg->sadb_msg_type == SADB_GETSPI) {
3003 		sav->created = time_second;
3004 	} else if (sav->state == SADB_SASTATE_LARVAL) {
3005 		/*
3006 		 * Do not call key_setsaval() second time in case
3007 		 * of `goto again`. We will have MATURE state.
3008 		 */
3009 		*errp = key_setsaval(sav, mhp);
3010 		if (*errp != 0)
3011 			goto done;
3012 		sav->state = SADB_SASTATE_MATURE;
3013 	}
3014 
3015 	SAHTREE_WLOCK();
3016 	/*
3017 	 * Check that existing SAH wasn't unlinked.
3018 	 * Since we didn't hold the SAHTREE lock, it is possible,
3019 	 * that callout handler or key_flush() or key_delete() could
3020 	 * unlink this SAH.
3021 	 */
3022 	if (isnew == 0 && sah->state == SADB_SASTATE_DEAD) {
3023 		SAHTREE_WUNLOCK();
3024 		key_freesah(&sah);	/* reference from key_getsah() */
3025 		goto again;
3026 	}
3027 	if (isnew != 0) {
3028 		/*
3029 		 * Add new SAH into SADB.
3030 		 *
3031 		 * XXXAE: we can serialize key_add and key_getspi calls, so
3032 		 * several threads will not fight in the race.
3033 		 * Otherwise we should check under SAHTREE lock, that this
3034 		 * SAH would not added twice.
3035 		 */
3036 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
3037 		/* Add new SAH into hash by addresses */
3038 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
3039 		/* Now we are linked in the chain */
3040 		sah->state = SADB_SASTATE_MATURE;
3041 		/*
3042 		 * SAV references this new SAH.
3043 		 * In case of existing SAH we reuse reference
3044 		 * from key_getsah().
3045 		 */
3046 		SAH_ADDREF(sah);
3047 	}
3048 	/* Link SAV with SAH */
3049 	if (sav->state == SADB_SASTATE_MATURE)
3050 		TAILQ_INSERT_HEAD(&sah->savtree_alive, sav, chain);
3051 	else
3052 		TAILQ_INSERT_HEAD(&sah->savtree_larval, sav, chain);
3053 	/* Add SAV into SPI hash */
3054 	LIST_INSERT_HEAD(SAVHASH_HASH(sav->spi), sav, spihash);
3055 	SAHTREE_WUNLOCK();
3056 	*errp = 0;	/* success */
3057 done:
3058 	if (*errp != 0) {
3059 		if (sav != NULL) {
3060 			if (sav->lock != NULL) {
3061 				rm_destroy(sav->lock);
3062 				free(sav->lock, M_IPSEC_MISC);
3063 			}
3064 			if (sav->lft_c != NULL)
3065 				uma_zfree_pcpu(ipsec_key_lft_zone, sav->lft_c);
3066 			free(sav, M_IPSEC_SA), sav = NULL;
3067 		}
3068 		if (sah != NULL)
3069 			key_freesah(&sah);
3070 		if (*errp == ENOBUFS) {
3071 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3072 			    __func__));
3073 			PFKEYSTAT_INC(in_nomem);
3074 		}
3075 	}
3076 	return (sav);
3077 }
3078 
3079 /*
3080  * free() SA variable entry.
3081  */
3082 static void
3083 key_cleansav(struct secasvar *sav)
3084 {
3085 
3086 	if (sav->natt != NULL) {
3087 		free(sav->natt, M_IPSEC_MISC);
3088 		sav->natt = NULL;
3089 	}
3090 	if (sav->flags & SADB_X_EXT_F_CLONED)
3091 		return;
3092 	if (sav->tdb_xform != NULL) {
3093 		sav->tdb_xform->xf_cleanup(sav);
3094 		sav->tdb_xform = NULL;
3095 	}
3096 	if (sav->key_auth != NULL) {
3097 		zfree(sav->key_auth->key_data, M_IPSEC_MISC);
3098 		free(sav->key_auth, M_IPSEC_MISC);
3099 		sav->key_auth = NULL;
3100 	}
3101 	if (sav->key_enc != NULL) {
3102 		zfree(sav->key_enc->key_data, M_IPSEC_MISC);
3103 		free(sav->key_enc, M_IPSEC_MISC);
3104 		sav->key_enc = NULL;
3105 	}
3106 	if (sav->replay != NULL) {
3107 		mtx_destroy(&sav->replay->lock);
3108 		if (sav->replay->bitmap != NULL)
3109 			free(sav->replay->bitmap, M_IPSEC_MISC);
3110 		free(sav->replay, M_IPSEC_MISC);
3111 		sav->replay = NULL;
3112 	}
3113 	if (sav->lft_h != NULL) {
3114 		free(sav->lft_h, M_IPSEC_MISC);
3115 		sav->lft_h = NULL;
3116 	}
3117 	if (sav->lft_s != NULL) {
3118 		free(sav->lft_s, M_IPSEC_MISC);
3119 		sav->lft_s = NULL;
3120 	}
3121 }
3122 
3123 /*
3124  * free() SA variable entry.
3125  */
3126 static void
3127 key_delsav(struct secasvar *sav)
3128 {
3129 	IPSEC_ASSERT(sav != NULL, ("null sav"));
3130 	IPSEC_ASSERT(sav->state == SADB_SASTATE_DEAD,
3131 	    ("attempt to free non DEAD SA %p", sav));
3132 	IPSEC_ASSERT(sav->refcnt == 0, ("reference count %u > 0",
3133 	    sav->refcnt));
3134 
3135 	/*
3136 	 * SA must be unlinked from the chain and hashtbl.
3137 	 * If SA was cloned, we leave all fields untouched,
3138 	 * except NAT-T config.
3139 	 */
3140 	key_cleansav(sav);
3141 	if ((sav->flags & SADB_X_EXT_F_CLONED) == 0) {
3142 		rm_destroy(sav->lock);
3143 		free(sav->lock, M_IPSEC_MISC);
3144 		uma_zfree_pcpu(ipsec_key_lft_zone, sav->lft_c);
3145 	}
3146 	free(sav, M_IPSEC_SA);
3147 }
3148 
3149 /*
3150  * search SAH.
3151  * OUT:
3152  *	NULL	: not found
3153  *	others	: found, referenced pointer to a SAH.
3154  */
3155 static struct secashead *
3156 key_getsah(struct secasindex *saidx)
3157 {
3158 	SAHTREE_RLOCK_TRACKER;
3159 	struct secashead *sah;
3160 
3161 	SAHTREE_RLOCK();
3162 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
3163 	    if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID) != 0) {
3164 		    SAH_ADDREF(sah);
3165 		    break;
3166 	    }
3167 	}
3168 	SAHTREE_RUNLOCK();
3169 	return (sah);
3170 }
3171 
3172 /*
3173  * Check not to be duplicated SPI.
3174  * OUT:
3175  *	0	: not found
3176  *	1	: found SA with given SPI.
3177  */
3178 static int
3179 key_checkspidup(uint32_t spi)
3180 {
3181 	SAHTREE_RLOCK_TRACKER;
3182 	struct secasvar *sav;
3183 
3184 	/* Assume SPI is in network byte order */
3185 	SAHTREE_RLOCK();
3186 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3187 		if (sav->spi == spi)
3188 			break;
3189 	}
3190 	SAHTREE_RUNLOCK();
3191 	return (sav != NULL);
3192 }
3193 
3194 /*
3195  * Search SA by SPI.
3196  * OUT:
3197  *	NULL	: not found
3198  *	others	: found, referenced pointer to a SA.
3199  */
3200 static struct secasvar *
3201 key_getsavbyspi(uint32_t spi)
3202 {
3203 	SAHTREE_RLOCK_TRACKER;
3204 	struct secasvar *sav;
3205 
3206 	/* Assume SPI is in network byte order */
3207 	SAHTREE_RLOCK();
3208 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3209 		if (sav->spi != spi)
3210 			continue;
3211 		SAV_ADDREF(sav);
3212 		break;
3213 	}
3214 	SAHTREE_RUNLOCK();
3215 	return (sav);
3216 }
3217 
3218 static int
3219 key_updatelifetimes(struct secasvar *sav, const struct sadb_msghdr *mhp)
3220 {
3221 	struct seclifetime *lft_h, *lft_s, *tmp;
3222 
3223 	/* Lifetime extension is optional, check that it is present. */
3224 	if (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3225 	    SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) {
3226 		/*
3227 		 * In case of SADB_UPDATE we may need to change
3228 		 * existing lifetimes.
3229 		 */
3230 		if (sav->state == SADB_SASTATE_MATURE) {
3231 			lft_h = lft_s = NULL;
3232 			goto reset;
3233 		}
3234 		return (0);
3235 	}
3236 	/* Both HARD and SOFT extensions must present */
3237 	if ((SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3238 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
3239 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
3240 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
3241 		ipseclog((LOG_DEBUG,
3242 		    "%s: invalid message: missing required header.\n",
3243 		    __func__));
3244 		return (EINVAL);
3245 	}
3246 	if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD) ||
3247 	    SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_SOFT)) {
3248 		ipseclog((LOG_DEBUG,
3249 		    "%s: invalid message: wrong header size.\n", __func__));
3250 		return (EINVAL);
3251 	}
3252 	lft_h = key_dup_lifemsg((const struct sadb_lifetime *)
3253 	    mhp->ext[SADB_EXT_LIFETIME_HARD], M_IPSEC_MISC);
3254 	if (lft_h == NULL) {
3255 		PFKEYSTAT_INC(in_nomem);
3256 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3257 		return (ENOBUFS);
3258 	}
3259 	lft_s = key_dup_lifemsg((const struct sadb_lifetime *)
3260 	    mhp->ext[SADB_EXT_LIFETIME_SOFT], M_IPSEC_MISC);
3261 	if (lft_s == NULL) {
3262 		PFKEYSTAT_INC(in_nomem);
3263 		free(lft_h, M_IPSEC_MISC);
3264 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3265 		return (ENOBUFS);
3266 	}
3267 reset:
3268 	if (sav->state != SADB_SASTATE_LARVAL) {
3269 		/*
3270 		 * key_update() holds reference to this SA,
3271 		 * so it won't be deleted in meanwhile.
3272 		 */
3273 		SECASVAR_WLOCK(sav);
3274 		tmp = sav->lft_h;
3275 		sav->lft_h = lft_h;
3276 		lft_h = tmp;
3277 
3278 		tmp = sav->lft_s;
3279 		sav->lft_s = lft_s;
3280 		lft_s = tmp;
3281 		SECASVAR_WUNLOCK(sav);
3282 		if (lft_h != NULL)
3283 			free(lft_h, M_IPSEC_MISC);
3284 		if (lft_s != NULL)
3285 			free(lft_s, M_IPSEC_MISC);
3286 		return (0);
3287 	}
3288 	/* We can update lifetime without holding a lock */
3289 	IPSEC_ASSERT(sav->lft_h == NULL, ("lft_h is already initialized\n"));
3290 	IPSEC_ASSERT(sav->lft_s == NULL, ("lft_s is already initialized\n"));
3291 	sav->lft_h = lft_h;
3292 	sav->lft_s = lft_s;
3293 	return (0);
3294 }
3295 
3296 /*
3297  * copy SA values from PF_KEY message except *SPI, SEQ, PID and TYPE*.
3298  * You must update these if need. Expects only LARVAL SAs.
3299  * OUT:	0:	success.
3300  *	!0:	failure.
3301  */
3302 static int
3303 key_setsaval(struct secasvar *sav, const struct sadb_msghdr *mhp)
3304 {
3305 	const struct sadb_sa *sa0;
3306 	const struct sadb_key *key0;
3307 	uint32_t replay;
3308 	size_t len;
3309 	int error;
3310 
3311 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
3312 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
3313 	IPSEC_ASSERT(sav->state == SADB_SASTATE_LARVAL,
3314 	    ("Attempt to update non LARVAL SA"));
3315 
3316 	/* XXX rewrite */
3317 	error = key_setident(sav->sah, mhp);
3318 	if (error != 0)
3319 		goto fail;
3320 
3321 	/* SA */
3322 	if (!SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
3323 		if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
3324 			error = EINVAL;
3325 			goto fail;
3326 		}
3327 		sa0 = (const struct sadb_sa *)mhp->ext[SADB_EXT_SA];
3328 		sav->alg_auth = sa0->sadb_sa_auth;
3329 		sav->alg_enc = sa0->sadb_sa_encrypt;
3330 		sav->flags = sa0->sadb_sa_flags;
3331 		if ((sav->flags & SADB_KEY_FLAGS_MAX) != sav->flags) {
3332 			ipseclog((LOG_DEBUG,
3333 			    "%s: invalid sa_flags 0x%08x.\n", __func__,
3334 			    sav->flags));
3335 			error = EINVAL;
3336 			goto fail;
3337 		}
3338 
3339 		/* Optional replay window */
3340 		replay = 0;
3341 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0)
3342 			replay = sa0->sadb_sa_replay;
3343 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_SA_REPLAY)) {
3344 			if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA_REPLAY)) {
3345 				error = EINVAL;
3346 				goto fail;
3347 			}
3348 			replay = ((const struct sadb_x_sa_replay *)
3349 			    mhp->ext[SADB_X_EXT_SA_REPLAY])->sadb_x_sa_replay_replay;
3350 
3351 			if (replay > UINT32_MAX - 32) {
3352 				ipseclog((LOG_DEBUG,
3353 				    "%s: replay window too big.\n", __func__));
3354 				error = EINVAL;
3355 				goto fail;
3356 			}
3357 
3358 			replay = (replay + 7) >> 3;
3359 		}
3360 
3361 		sav->replay = malloc(sizeof(struct secreplay), M_IPSEC_MISC,
3362 		    M_NOWAIT | M_ZERO);
3363 		if (sav->replay == NULL) {
3364 			PFKEYSTAT_INC(in_nomem);
3365 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3366 			    __func__));
3367 			error = ENOBUFS;
3368 			goto fail;
3369 		}
3370 		mtx_init(&sav->replay->lock, "ipsec replay", NULL, MTX_DEF);
3371 
3372 		if (replay != 0) {
3373 			/* number of 32b blocks to be allocated */
3374 			uint32_t bitmap_size;
3375 
3376 			/* RFC 6479:
3377 			 * - the allocated replay window size must be
3378 			 *   a power of two.
3379 			 * - use an extra 32b block as a redundant window.
3380 			 */
3381 			bitmap_size = 1;
3382 			while (replay + 4 > bitmap_size)
3383 				bitmap_size <<= 1;
3384 			bitmap_size = bitmap_size / 4;
3385 
3386 			sav->replay->bitmap = malloc(
3387 			    bitmap_size * sizeof(uint32_t), M_IPSEC_MISC,
3388 			    M_NOWAIT | M_ZERO);
3389 			if (sav->replay->bitmap == NULL) {
3390 				PFKEYSTAT_INC(in_nomem);
3391 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3392 					__func__));
3393 				error = ENOBUFS;
3394 				goto fail;
3395 			}
3396 			sav->replay->bitmap_size = bitmap_size;
3397 			sav->replay->wsize = replay;
3398 		}
3399 	}
3400 
3401 	/* Authentication keys */
3402 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
3403 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH)) {
3404 			error = EINVAL;
3405 			goto fail;
3406 		}
3407 		error = 0;
3408 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_AUTH];
3409 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3410 		switch (mhp->msg->sadb_msg_satype) {
3411 		case SADB_SATYPE_AH:
3412 		case SADB_SATYPE_ESP:
3413 		case SADB_X_SATYPE_TCPSIGNATURE:
3414 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3415 			    sav->alg_auth != SADB_X_AALG_NULL)
3416 				error = EINVAL;
3417 			break;
3418 		case SADB_X_SATYPE_IPCOMP:
3419 		default:
3420 			error = EINVAL;
3421 			break;
3422 		}
3423 		if (error) {
3424 			ipseclog((LOG_DEBUG, "%s: invalid key_auth values.\n",
3425 				__func__));
3426 			goto fail;
3427 		}
3428 
3429 		sav->key_auth = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3430 		if (sav->key_auth == NULL ) {
3431 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3432 				  __func__));
3433 			PFKEYSTAT_INC(in_nomem);
3434 			error = ENOBUFS;
3435 			goto fail;
3436 		}
3437 	}
3438 
3439 	/* Encryption key */
3440 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) {
3441 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT)) {
3442 			error = EINVAL;
3443 			goto fail;
3444 		}
3445 		error = 0;
3446 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_ENCRYPT];
3447 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3448 		switch (mhp->msg->sadb_msg_satype) {
3449 		case SADB_SATYPE_ESP:
3450 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3451 			    sav->alg_enc != SADB_EALG_NULL) {
3452 				error = EINVAL;
3453 				break;
3454 			}
3455 			sav->key_enc = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3456 			if (sav->key_enc == NULL) {
3457 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3458 					__func__));
3459 				PFKEYSTAT_INC(in_nomem);
3460 				error = ENOBUFS;
3461 				goto fail;
3462 			}
3463 			break;
3464 		case SADB_X_SATYPE_IPCOMP:
3465 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3466 				error = EINVAL;
3467 			sav->key_enc = NULL;	/*just in case*/
3468 			break;
3469 		case SADB_SATYPE_AH:
3470 		case SADB_X_SATYPE_TCPSIGNATURE:
3471 		default:
3472 			error = EINVAL;
3473 			break;
3474 		}
3475 		if (error) {
3476 			ipseclog((LOG_DEBUG, "%s: invalid key_enc value.\n",
3477 				__func__));
3478 			goto fail;
3479 		}
3480 	}
3481 
3482 	/* set iv */
3483 	sav->ivlen = 0;
3484 	switch (mhp->msg->sadb_msg_satype) {
3485 	case SADB_SATYPE_AH:
3486 		if (sav->flags & SADB_X_EXT_DERIV) {
3487 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3488 			    "given to AH SA.\n", __func__));
3489 			error = EINVAL;
3490 			goto fail;
3491 		}
3492 		if (sav->alg_enc != SADB_EALG_NONE) {
3493 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3494 			    "mismated.\n", __func__));
3495 			error = EINVAL;
3496 			goto fail;
3497 		}
3498 		error = xform_init(sav, XF_AH);
3499 		break;
3500 	case SADB_SATYPE_ESP:
3501 		if ((sav->flags & (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) ==
3502 		    (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) {
3503 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3504 			    "given to old-esp.\n", __func__));
3505 			error = EINVAL;
3506 			goto fail;
3507 		}
3508 		error = xform_init(sav, XF_ESP);
3509 		break;
3510 	case SADB_X_SATYPE_IPCOMP:
3511 		if (sav->alg_auth != SADB_AALG_NONE) {
3512 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3513 			    "mismated.\n", __func__));
3514 			error = EINVAL;
3515 			goto fail;
3516 		}
3517 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0 &&
3518 		    ntohl(sav->spi) >= 0x10000) {
3519 			ipseclog((LOG_DEBUG, "%s: invalid cpi for IPComp.\n",
3520 			    __func__));
3521 			error = EINVAL;
3522 			goto fail;
3523 		}
3524 		error = xform_init(sav, XF_IPCOMP);
3525 		break;
3526 	case SADB_X_SATYPE_TCPSIGNATURE:
3527 		if (sav->alg_enc != SADB_EALG_NONE) {
3528 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3529 			    "mismated.\n", __func__));
3530 			error = EINVAL;
3531 			goto fail;
3532 		}
3533 		error = xform_init(sav, XF_TCPSIGNATURE);
3534 		break;
3535 	default:
3536 		ipseclog((LOG_DEBUG, "%s: Invalid satype.\n", __func__));
3537 		error = EPROTONOSUPPORT;
3538 		goto fail;
3539 	}
3540 	if (error) {
3541 		ipseclog((LOG_DEBUG, "%s: unable to initialize SA type %u.\n",
3542 		    __func__, mhp->msg->sadb_msg_satype));
3543 		goto fail;
3544 	}
3545 
3546 	/* Handle NAT-T headers */
3547 	error = key_setnatt(sav, mhp);
3548 	if (error != 0)
3549 		goto fail;
3550 
3551 	/* Initialize lifetime for CURRENT */
3552 	sav->firstused = 0;
3553 	sav->created = time_second;
3554 
3555 	/* lifetimes for HARD and SOFT */
3556 	error = key_updatelifetimes(sav, mhp);
3557 	if (error == 0)
3558 		return (0);
3559 fail:
3560 	key_cleansav(sav);
3561 	return (error);
3562 }
3563 
3564 /*
3565  * subroutine for SADB_GET and SADB_DUMP.
3566  */
3567 static struct mbuf *
3568 key_setdumpsa(struct secasvar *sav, uint8_t type, uint8_t satype,
3569     uint32_t seq, uint32_t pid)
3570 {
3571 	struct seclifetime lft_c;
3572 	struct mbuf *result = NULL, *tres = NULL, *m;
3573 	int i, dumporder[] = {
3574 		SADB_EXT_SA, SADB_X_EXT_SA2, SADB_X_EXT_SA_REPLAY,
3575 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3576 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3577 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY,
3578 		SADB_EXT_KEY_AUTH, SADB_EXT_KEY_ENCRYPT,
3579 		SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
3580 		SADB_EXT_SENSITIVITY,
3581 		SADB_X_EXT_NAT_T_TYPE,
3582 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3583 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3584 		SADB_X_EXT_NAT_T_FRAG,
3585 	};
3586 	uint32_t replay_count;
3587 
3588 	SECASVAR_RLOCK_TRACKER;
3589 
3590 	m = key_setsadbmsg(type, 0, satype, seq, pid, sav->refcnt);
3591 	if (m == NULL)
3592 		goto fail;
3593 	result = m;
3594 
3595 	for (i = nitems(dumporder) - 1; i >= 0; i--) {
3596 		m = NULL;
3597 		switch (dumporder[i]) {
3598 		case SADB_EXT_SA:
3599 			m = key_setsadbsa(sav);
3600 			if (!m)
3601 				goto fail;
3602 			break;
3603 
3604 		case SADB_X_EXT_SA2: {
3605 			SECASVAR_RLOCK(sav);
3606 			replay_count = sav->replay ? sav->replay->count : 0;
3607 			SECASVAR_RUNLOCK(sav);
3608 			m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
3609 					sav->sah->saidx.reqid);
3610 			if (!m)
3611 				goto fail;
3612 			break;
3613 		}
3614 		case SADB_X_EXT_SA_REPLAY:
3615 			if (sav->replay == NULL ||
3616 			    sav->replay->wsize <= UINT8_MAX)
3617 				continue;
3618 
3619 			m = key_setsadbxsareplay(sav->replay->wsize);
3620 			if (!m)
3621 				goto fail;
3622 			break;
3623 
3624 		case SADB_EXT_ADDRESS_SRC:
3625 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3626 			    &sav->sah->saidx.src.sa,
3627 			    FULLMASK, IPSEC_ULPROTO_ANY);
3628 			if (!m)
3629 				goto fail;
3630 			break;
3631 
3632 		case SADB_EXT_ADDRESS_DST:
3633 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3634 			    &sav->sah->saidx.dst.sa,
3635 			    FULLMASK, IPSEC_ULPROTO_ANY);
3636 			if (!m)
3637 				goto fail;
3638 			break;
3639 
3640 		case SADB_EXT_KEY_AUTH:
3641 			if (!sav->key_auth)
3642 				continue;
3643 			m = key_setkey(sav->key_auth, SADB_EXT_KEY_AUTH);
3644 			if (!m)
3645 				goto fail;
3646 			break;
3647 
3648 		case SADB_EXT_KEY_ENCRYPT:
3649 			if (!sav->key_enc)
3650 				continue;
3651 			m = key_setkey(sav->key_enc, SADB_EXT_KEY_ENCRYPT);
3652 			if (!m)
3653 				goto fail;
3654 			break;
3655 
3656 		case SADB_EXT_LIFETIME_CURRENT:
3657 			lft_c.addtime = sav->created;
3658 			lft_c.allocations = (uint32_t)counter_u64_fetch(
3659 			    sav->lft_c_allocations);
3660 			lft_c.bytes = counter_u64_fetch(sav->lft_c_bytes);
3661 			lft_c.usetime = sav->firstused;
3662 			m = key_setlifetime(&lft_c, SADB_EXT_LIFETIME_CURRENT);
3663 			if (!m)
3664 				goto fail;
3665 			break;
3666 
3667 		case SADB_EXT_LIFETIME_HARD:
3668 			if (!sav->lft_h)
3669 				continue;
3670 			m = key_setlifetime(sav->lft_h,
3671 					    SADB_EXT_LIFETIME_HARD);
3672 			if (!m)
3673 				goto fail;
3674 			break;
3675 
3676 		case SADB_EXT_LIFETIME_SOFT:
3677 			if (!sav->lft_s)
3678 				continue;
3679 			m = key_setlifetime(sav->lft_s,
3680 					    SADB_EXT_LIFETIME_SOFT);
3681 
3682 			if (!m)
3683 				goto fail;
3684 			break;
3685 
3686 		case SADB_X_EXT_NAT_T_TYPE:
3687 			if (sav->natt == NULL)
3688 				continue;
3689 			m = key_setsadbxtype(UDP_ENCAP_ESPINUDP);
3690 			if (!m)
3691 				goto fail;
3692 			break;
3693 
3694 		case SADB_X_EXT_NAT_T_DPORT:
3695 			if (sav->natt == NULL)
3696 				continue;
3697 			m = key_setsadbxport(sav->natt->dport,
3698 			    SADB_X_EXT_NAT_T_DPORT);
3699 			if (!m)
3700 				goto fail;
3701 			break;
3702 
3703 		case SADB_X_EXT_NAT_T_SPORT:
3704 			if (sav->natt == NULL)
3705 				continue;
3706 			m = key_setsadbxport(sav->natt->sport,
3707 			    SADB_X_EXT_NAT_T_SPORT);
3708 			if (!m)
3709 				goto fail;
3710 			break;
3711 
3712 		case SADB_X_EXT_NAT_T_OAI:
3713 			if (sav->natt == NULL ||
3714 			    (sav->natt->flags & IPSEC_NATT_F_OAI) == 0)
3715 				continue;
3716 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAI,
3717 			    &sav->natt->oai.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3718 			if (!m)
3719 				goto fail;
3720 			break;
3721 		case SADB_X_EXT_NAT_T_OAR:
3722 			if (sav->natt == NULL ||
3723 			    (sav->natt->flags & IPSEC_NATT_F_OAR) == 0)
3724 				continue;
3725 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAR,
3726 			    &sav->natt->oar.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3727 			if (!m)
3728 				goto fail;
3729 			break;
3730 		case SADB_X_EXT_NAT_T_FRAG:
3731 			/* We do not (yet) support those. */
3732 			continue;
3733 
3734 		case SADB_EXT_ADDRESS_PROXY:
3735 		case SADB_EXT_IDENTITY_SRC:
3736 		case SADB_EXT_IDENTITY_DST:
3737 			/* XXX: should we brought from SPD ? */
3738 		case SADB_EXT_SENSITIVITY:
3739 		default:
3740 			continue;
3741 		}
3742 
3743 		if (!m)
3744 			goto fail;
3745 		if (tres)
3746 			m_cat(m, tres);
3747 		tres = m;
3748 	}
3749 
3750 	m_cat(result, tres);
3751 	tres = NULL;
3752 	if (result->m_len < sizeof(struct sadb_msg)) {
3753 		result = m_pullup(result, sizeof(struct sadb_msg));
3754 		if (result == NULL)
3755 			goto fail;
3756 	}
3757 
3758 	result->m_pkthdr.len = 0;
3759 	for (m = result; m; m = m->m_next)
3760 		result->m_pkthdr.len += m->m_len;
3761 
3762 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3763 	    PFKEY_UNIT64(result->m_pkthdr.len);
3764 
3765 	return result;
3766 
3767 fail:
3768 	m_freem(result);
3769 	m_freem(tres);
3770 	return NULL;
3771 }
3772 
3773 /*
3774  * set data into sadb_msg.
3775  */
3776 static struct mbuf *
3777 key_setsadbmsg(u_int8_t type, u_int16_t tlen, u_int8_t satype, u_int32_t seq,
3778     pid_t pid, u_int16_t reserved)
3779 {
3780 	struct mbuf *m;
3781 	struct sadb_msg *p;
3782 	int len;
3783 
3784 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
3785 	if (len > MCLBYTES)
3786 		return NULL;
3787 	MGETHDR(m, M_NOWAIT, MT_DATA);
3788 	if (m && len > MHLEN) {
3789 		if (!(MCLGET(m, M_NOWAIT))) {
3790 			m_freem(m);
3791 			m = NULL;
3792 		}
3793 	}
3794 	if (!m)
3795 		return NULL;
3796 	m->m_pkthdr.len = m->m_len = len;
3797 	m->m_next = NULL;
3798 
3799 	p = mtod(m, struct sadb_msg *);
3800 
3801 	bzero(p, len);
3802 	p->sadb_msg_version = PF_KEY_V2;
3803 	p->sadb_msg_type = type;
3804 	p->sadb_msg_errno = 0;
3805 	p->sadb_msg_satype = satype;
3806 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
3807 	p->sadb_msg_reserved = reserved;
3808 	p->sadb_msg_seq = seq;
3809 	p->sadb_msg_pid = (u_int32_t)pid;
3810 
3811 	return m;
3812 }
3813 
3814 /*
3815  * copy secasvar data into sadb_address.
3816  */
3817 static struct mbuf *
3818 key_setsadbsa(struct secasvar *sav)
3819 {
3820 	struct mbuf *m;
3821 	struct sadb_sa *p;
3822 	int len;
3823 
3824 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
3825 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3826 	if (m == NULL)
3827 		return (NULL);
3828 	m_align(m, len);
3829 	m->m_len = len;
3830 	p = mtod(m, struct sadb_sa *);
3831 	bzero(p, len);
3832 	p->sadb_sa_len = PFKEY_UNIT64(len);
3833 	p->sadb_sa_exttype = SADB_EXT_SA;
3834 	p->sadb_sa_spi = sav->spi;
3835 	p->sadb_sa_replay = sav->replay ?
3836 	    (sav->replay->wsize > UINT8_MAX ? UINT8_MAX :
3837 		sav->replay->wsize): 0;
3838 	p->sadb_sa_state = sav->state;
3839 	p->sadb_sa_auth = sav->alg_auth;
3840 	p->sadb_sa_encrypt = sav->alg_enc;
3841 	p->sadb_sa_flags = sav->flags & SADB_KEY_FLAGS_MAX;
3842 	return (m);
3843 }
3844 
3845 /*
3846  * set data into sadb_address.
3847  */
3848 static struct mbuf *
3849 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
3850     u_int8_t prefixlen, u_int16_t ul_proto)
3851 {
3852 	struct mbuf *m;
3853 	struct sadb_address *p;
3854 	size_t len;
3855 
3856 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
3857 	    PFKEY_ALIGN8(saddr->sa_len);
3858 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3859 	if (m == NULL)
3860 		return (NULL);
3861 	m_align(m, len);
3862 	m->m_len = len;
3863 	p = mtod(m, struct sadb_address *);
3864 
3865 	bzero(p, len);
3866 	p->sadb_address_len = PFKEY_UNIT64(len);
3867 	p->sadb_address_exttype = exttype;
3868 	p->sadb_address_proto = ul_proto;
3869 	if (prefixlen == FULLMASK) {
3870 		switch (saddr->sa_family) {
3871 		case AF_INET:
3872 			prefixlen = sizeof(struct in_addr) << 3;
3873 			break;
3874 		case AF_INET6:
3875 			prefixlen = sizeof(struct in6_addr) << 3;
3876 			break;
3877 		default:
3878 			; /*XXX*/
3879 		}
3880 	}
3881 	p->sadb_address_prefixlen = prefixlen;
3882 	p->sadb_address_reserved = 0;
3883 
3884 	bcopy(saddr,
3885 	    mtod(m, caddr_t) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
3886 	    saddr->sa_len);
3887 
3888 	return m;
3889 }
3890 
3891 /*
3892  * set data into sadb_x_sa2.
3893  */
3894 static struct mbuf *
3895 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int32_t reqid)
3896 {
3897 	struct mbuf *m;
3898 	struct sadb_x_sa2 *p;
3899 	size_t len;
3900 
3901 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
3902 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3903 	if (m == NULL)
3904 		return (NULL);
3905 	m_align(m, len);
3906 	m->m_len = len;
3907 	p = mtod(m, struct sadb_x_sa2 *);
3908 
3909 	bzero(p, len);
3910 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
3911 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
3912 	p->sadb_x_sa2_mode = mode;
3913 	p->sadb_x_sa2_reserved1 = 0;
3914 	p->sadb_x_sa2_reserved2 = 0;
3915 	p->sadb_x_sa2_sequence = seq;
3916 	p->sadb_x_sa2_reqid = reqid;
3917 
3918 	return m;
3919 }
3920 
3921 /*
3922  * Set data into sadb_x_sa_replay.
3923  */
3924 static struct mbuf *
3925 key_setsadbxsareplay(u_int32_t replay)
3926 {
3927 	struct mbuf *m;
3928 	struct sadb_x_sa_replay *p;
3929 	size_t len;
3930 
3931 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa_replay));
3932 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3933 	if (m == NULL)
3934 		return (NULL);
3935 	m_align(m, len);
3936 	m->m_len = len;
3937 	p = mtod(m, struct sadb_x_sa_replay *);
3938 
3939 	bzero(p, len);
3940 	p->sadb_x_sa_replay_len = PFKEY_UNIT64(len);
3941 	p->sadb_x_sa_replay_exttype = SADB_X_EXT_SA_REPLAY;
3942 	p->sadb_x_sa_replay_replay = (replay << 3);
3943 
3944 	return m;
3945 }
3946 
3947 /*
3948  * Set a type in sadb_x_nat_t_type.
3949  */
3950 static struct mbuf *
3951 key_setsadbxtype(u_int16_t type)
3952 {
3953 	struct mbuf *m;
3954 	size_t len;
3955 	struct sadb_x_nat_t_type *p;
3956 
3957 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
3958 
3959 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3960 	if (m == NULL)
3961 		return (NULL);
3962 	m_align(m, len);
3963 	m->m_len = len;
3964 	p = mtod(m, struct sadb_x_nat_t_type *);
3965 
3966 	bzero(p, len);
3967 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
3968 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
3969 	p->sadb_x_nat_t_type_type = type;
3970 
3971 	return (m);
3972 }
3973 /*
3974  * Set a port in sadb_x_nat_t_port.
3975  * In contrast to default RFC 2367 behaviour, port is in network byte order.
3976  */
3977 static struct mbuf *
3978 key_setsadbxport(u_int16_t port, u_int16_t type)
3979 {
3980 	struct mbuf *m;
3981 	size_t len;
3982 	struct sadb_x_nat_t_port *p;
3983 
3984 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
3985 
3986 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3987 	if (m == NULL)
3988 		return (NULL);
3989 	m_align(m, len);
3990 	m->m_len = len;
3991 	p = mtod(m, struct sadb_x_nat_t_port *);
3992 
3993 	bzero(p, len);
3994 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
3995 	p->sadb_x_nat_t_port_exttype = type;
3996 	p->sadb_x_nat_t_port_port = port;
3997 
3998 	return (m);
3999 }
4000 
4001 /*
4002  * Get port from sockaddr. Port is in network byte order.
4003  */
4004 uint16_t
4005 key_portfromsaddr(struct sockaddr *sa)
4006 {
4007 
4008 	switch (sa->sa_family) {
4009 #ifdef INET
4010 	case AF_INET:
4011 		return ((struct sockaddr_in *)sa)->sin_port;
4012 #endif
4013 #ifdef INET6
4014 	case AF_INET6:
4015 		return ((struct sockaddr_in6 *)sa)->sin6_port;
4016 #endif
4017 	}
4018 	return (0);
4019 }
4020 
4021 /*
4022  * Set port in struct sockaddr. Port is in network byte order.
4023  */
4024 void
4025 key_porttosaddr(struct sockaddr *sa, uint16_t port)
4026 {
4027 
4028 	switch (sa->sa_family) {
4029 #ifdef INET
4030 	case AF_INET:
4031 		((struct sockaddr_in *)sa)->sin_port = port;
4032 		break;
4033 #endif
4034 #ifdef INET6
4035 	case AF_INET6:
4036 		((struct sockaddr_in6 *)sa)->sin6_port = port;
4037 		break;
4038 #endif
4039 	default:
4040 		ipseclog((LOG_DEBUG, "%s: unexpected address family %d.\n",
4041 			__func__, sa->sa_family));
4042 		break;
4043 	}
4044 }
4045 
4046 /*
4047  * set data into sadb_x_policy
4048  */
4049 static struct mbuf *
4050 key_setsadbxpolicy(u_int16_t type, u_int8_t dir, u_int32_t id, u_int32_t priority)
4051 {
4052 	struct mbuf *m;
4053 	struct sadb_x_policy *p;
4054 	size_t len;
4055 
4056 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
4057 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4058 	if (m == NULL)
4059 		return (NULL);
4060 	m_align(m, len);
4061 	m->m_len = len;
4062 	p = mtod(m, struct sadb_x_policy *);
4063 
4064 	bzero(p, len);
4065 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
4066 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
4067 	p->sadb_x_policy_type = type;
4068 	p->sadb_x_policy_dir = dir;
4069 	p->sadb_x_policy_id = id;
4070 	p->sadb_x_policy_priority = priority;
4071 
4072 	return m;
4073 }
4074 
4075 /* %%% utilities */
4076 /* Take a key message (sadb_key) from the socket and turn it into one
4077  * of the kernel's key structures (seckey).
4078  *
4079  * IN: pointer to the src
4080  * OUT: NULL no more memory
4081  */
4082 struct seckey *
4083 key_dup_keymsg(const struct sadb_key *src, size_t len,
4084     struct malloc_type *type)
4085 {
4086 	struct seckey *dst;
4087 
4088 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4089 	if (dst != NULL) {
4090 		dst->bits = src->sadb_key_bits;
4091 		dst->key_data = malloc(len, type, M_NOWAIT);
4092 		if (dst->key_data != NULL) {
4093 			bcopy((const char *)(src + 1), dst->key_data, len);
4094 		} else {
4095 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4096 			    __func__));
4097 			free(dst, type);
4098 			dst = NULL;
4099 		}
4100 	} else {
4101 		ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4102 		    __func__));
4103 	}
4104 	return (dst);
4105 }
4106 
4107 /* Take a lifetime message (sadb_lifetime) passed in on a socket and
4108  * turn it into one of the kernel's lifetime structures (seclifetime).
4109  *
4110  * IN: pointer to the destination, source and malloc type
4111  * OUT: NULL, no more memory
4112  */
4113 
4114 static struct seclifetime *
4115 key_dup_lifemsg(const struct sadb_lifetime *src, struct malloc_type *type)
4116 {
4117 	struct seclifetime *dst;
4118 
4119 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4120 	if (dst == NULL) {
4121 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
4122 		return (NULL);
4123 	}
4124 	dst->allocations = src->sadb_lifetime_allocations;
4125 	dst->bytes = src->sadb_lifetime_bytes;
4126 	dst->addtime = src->sadb_lifetime_addtime;
4127 	dst->usetime = src->sadb_lifetime_usetime;
4128 	return (dst);
4129 }
4130 
4131 /*
4132  * compare two secasindex structure.
4133  * flag can specify to compare 2 saidxes.
4134  * compare two secasindex structure without both mode and reqid.
4135  * don't compare port.
4136  * IN:
4137  *      saidx0: source, it can be in SAD.
4138  *      saidx1: object.
4139  * OUT:
4140  *      1 : equal
4141  *      0 : not equal
4142  */
4143 static int
4144 key_cmpsaidx(const struct secasindex *saidx0, const struct secasindex *saidx1,
4145     int flag)
4146 {
4147 
4148 	/* sanity */
4149 	if (saidx0 == NULL && saidx1 == NULL)
4150 		return 1;
4151 
4152 	if (saidx0 == NULL || saidx1 == NULL)
4153 		return 0;
4154 
4155 	if (saidx0->proto != saidx1->proto)
4156 		return 0;
4157 
4158 	if (flag == CMP_EXACTLY) {
4159 		if (saidx0->mode != saidx1->mode)
4160 			return 0;
4161 		if (saidx0->reqid != saidx1->reqid)
4162 			return 0;
4163 		if (bcmp(&saidx0->src, &saidx1->src,
4164 		    saidx0->src.sa.sa_len) != 0 ||
4165 		    bcmp(&saidx0->dst, &saidx1->dst,
4166 		    saidx0->dst.sa.sa_len) != 0)
4167 			return 0;
4168 	} else {
4169 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
4170 		if (flag == CMP_MODE_REQID || flag == CMP_REQID) {
4171 			/*
4172 			 * If reqid of SPD is non-zero, unique SA is required.
4173 			 * The result must be of same reqid in this case.
4174 			 */
4175 			if (saidx1->reqid != 0 &&
4176 			    saidx0->reqid != saidx1->reqid)
4177 				return 0;
4178 		}
4179 
4180 		if (flag == CMP_MODE_REQID) {
4181 			if (saidx0->mode != IPSEC_MODE_ANY
4182 			 && saidx0->mode != saidx1->mode)
4183 				return 0;
4184 		}
4185 
4186 		if (key_sockaddrcmp(&saidx0->src.sa, &saidx1->src.sa, 0) != 0)
4187 			return 0;
4188 		if (key_sockaddrcmp(&saidx0->dst.sa, &saidx1->dst.sa, 0) != 0)
4189 			return 0;
4190 	}
4191 
4192 	return 1;
4193 }
4194 
4195 /*
4196  * compare two secindex structure exactly.
4197  * IN:
4198  *	spidx0: source, it is often in SPD.
4199  *	spidx1: object, it is often from PFKEY message.
4200  * OUT:
4201  *	1 : equal
4202  *	0 : not equal
4203  */
4204 static int
4205 key_cmpspidx_exactly(struct secpolicyindex *spidx0,
4206     struct secpolicyindex *spidx1)
4207 {
4208 	/* sanity */
4209 	if (spidx0 == NULL && spidx1 == NULL)
4210 		return 1;
4211 
4212 	if (spidx0 == NULL || spidx1 == NULL)
4213 		return 0;
4214 
4215 	if (spidx0->prefs != spidx1->prefs
4216 	 || spidx0->prefd != spidx1->prefd
4217 	 || spidx0->ul_proto != spidx1->ul_proto
4218 	 || spidx0->dir != spidx1->dir)
4219 		return 0;
4220 
4221 	return key_sockaddrcmp(&spidx0->src.sa, &spidx1->src.sa, 1) == 0 &&
4222 	       key_sockaddrcmp(&spidx0->dst.sa, &spidx1->dst.sa, 1) == 0;
4223 }
4224 
4225 /*
4226  * compare two secindex structure with mask.
4227  * IN:
4228  *	spidx0: source, it is often in SPD.
4229  *	spidx1: object, it is often from IP header.
4230  * OUT:
4231  *	1 : equal
4232  *	0 : not equal
4233  */
4234 static int
4235 key_cmpspidx_withmask(struct secpolicyindex *spidx0,
4236     struct secpolicyindex *spidx1)
4237 {
4238 	/* sanity */
4239 	if (spidx0 == NULL && spidx1 == NULL)
4240 		return 1;
4241 
4242 	if (spidx0 == NULL || spidx1 == NULL)
4243 		return 0;
4244 
4245 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4246 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4247 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4248 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
4249 		return 0;
4250 
4251 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4252 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY
4253 	 && spidx0->ul_proto != spidx1->ul_proto)
4254 		return 0;
4255 
4256 	switch (spidx0->src.sa.sa_family) {
4257 	case AF_INET:
4258 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY
4259 		 && spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
4260 			return 0;
4261 		if (!key_bbcmp(&spidx0->src.sin.sin_addr,
4262 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
4263 			return 0;
4264 		break;
4265 	case AF_INET6:
4266 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY
4267 		 && spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
4268 			return 0;
4269 		/*
4270 		 * scope_id check. if sin6_scope_id is 0, we regard it
4271 		 * as a wildcard scope, which matches any scope zone ID.
4272 		 */
4273 		if (spidx0->src.sin6.sin6_scope_id &&
4274 		    spidx1->src.sin6.sin6_scope_id &&
4275 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
4276 			return 0;
4277 		if (!key_bbcmp(&spidx0->src.sin6.sin6_addr,
4278 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
4279 			return 0;
4280 		break;
4281 	default:
4282 		/* XXX */
4283 		if (bcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
4284 			return 0;
4285 		break;
4286 	}
4287 
4288 	switch (spidx0->dst.sa.sa_family) {
4289 	case AF_INET:
4290 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY
4291 		 && spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
4292 			return 0;
4293 		if (!key_bbcmp(&spidx0->dst.sin.sin_addr,
4294 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
4295 			return 0;
4296 		break;
4297 	case AF_INET6:
4298 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY
4299 		 && spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
4300 			return 0;
4301 		/*
4302 		 * scope_id check. if sin6_scope_id is 0, we regard it
4303 		 * as a wildcard scope, which matches any scope zone ID.
4304 		 */
4305 		if (spidx0->dst.sin6.sin6_scope_id &&
4306 		    spidx1->dst.sin6.sin6_scope_id &&
4307 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
4308 			return 0;
4309 		if (!key_bbcmp(&spidx0->dst.sin6.sin6_addr,
4310 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
4311 			return 0;
4312 		break;
4313 	default:
4314 		/* XXX */
4315 		if (bcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
4316 			return 0;
4317 		break;
4318 	}
4319 
4320 	/* XXX Do we check other field ?  e.g. flowinfo */
4321 
4322 	return 1;
4323 }
4324 
4325 #ifdef satosin
4326 #undef satosin
4327 #endif
4328 #define satosin(s) ((const struct sockaddr_in *)s)
4329 #ifdef satosin6
4330 #undef satosin6
4331 #endif
4332 #define satosin6(s) ((const struct sockaddr_in6 *)s)
4333 /* returns 0 on match */
4334 int
4335 key_sockaddrcmp(const struct sockaddr *sa1, const struct sockaddr *sa2,
4336     int port)
4337 {
4338 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4339 		return 1;
4340 
4341 	switch (sa1->sa_family) {
4342 #ifdef INET
4343 	case AF_INET:
4344 		if (sa1->sa_len != sizeof(struct sockaddr_in))
4345 			return 1;
4346 		if (satosin(sa1)->sin_addr.s_addr !=
4347 		    satosin(sa2)->sin_addr.s_addr) {
4348 			return 1;
4349 		}
4350 		if (port && satosin(sa1)->sin_port != satosin(sa2)->sin_port)
4351 			return 1;
4352 		break;
4353 #endif
4354 #ifdef INET6
4355 	case AF_INET6:
4356 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4357 			return 1;	/*EINVAL*/
4358 		if (satosin6(sa1)->sin6_scope_id !=
4359 		    satosin6(sa2)->sin6_scope_id) {
4360 			return 1;
4361 		}
4362 		if (!IN6_ARE_ADDR_EQUAL(&satosin6(sa1)->sin6_addr,
4363 		    &satosin6(sa2)->sin6_addr)) {
4364 			return 1;
4365 		}
4366 		if (port &&
4367 		    satosin6(sa1)->sin6_port != satosin6(sa2)->sin6_port) {
4368 			return 1;
4369 		}
4370 		break;
4371 #endif
4372 	default:
4373 		if (bcmp(sa1, sa2, sa1->sa_len) != 0)
4374 			return 1;
4375 		break;
4376 	}
4377 
4378 	return 0;
4379 }
4380 
4381 /* returns 0 on match */
4382 int
4383 key_sockaddrcmp_withmask(const struct sockaddr *sa1,
4384     const struct sockaddr *sa2, size_t mask)
4385 {
4386 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4387 		return (1);
4388 
4389 	switch (sa1->sa_family) {
4390 #ifdef INET
4391 	case AF_INET:
4392 		return (!key_bbcmp(&satosin(sa1)->sin_addr,
4393 		    &satosin(sa2)->sin_addr, mask));
4394 #endif
4395 #ifdef INET6
4396 	case AF_INET6:
4397 		if (satosin6(sa1)->sin6_scope_id !=
4398 		    satosin6(sa2)->sin6_scope_id)
4399 			return (1);
4400 		return (!key_bbcmp(&satosin6(sa1)->sin6_addr,
4401 		    &satosin6(sa2)->sin6_addr, mask));
4402 #endif
4403 	}
4404 	return (1);
4405 }
4406 #undef satosin
4407 #undef satosin6
4408 
4409 /*
4410  * compare two buffers with mask.
4411  * IN:
4412  *	addr1: source
4413  *	addr2: object
4414  *	bits:  Number of bits to compare
4415  * OUT:
4416  *	1 : equal
4417  *	0 : not equal
4418  */
4419 static int
4420 key_bbcmp(const void *a1, const void *a2, u_int bits)
4421 {
4422 	const unsigned char *p1 = a1;
4423 	const unsigned char *p2 = a2;
4424 
4425 	/* XXX: This could be considerably faster if we compare a word
4426 	 * at a time, but it is complicated on LSB Endian machines */
4427 
4428 	/* Handle null pointers */
4429 	if (p1 == NULL || p2 == NULL)
4430 		return (p1 == p2);
4431 
4432 	while (bits >= 8) {
4433 		if (*p1++ != *p2++)
4434 			return 0;
4435 		bits -= 8;
4436 	}
4437 
4438 	if (bits > 0) {
4439 		u_int8_t mask = ~((1<<(8-bits))-1);
4440 		if ((*p1 & mask) != (*p2 & mask))
4441 			return 0;
4442 	}
4443 	return 1;	/* Match! */
4444 }
4445 
4446 static void
4447 key_flush_spd(time_t now)
4448 {
4449 	SPTREE_RLOCK_TRACKER;
4450 	struct secpolicy_list drainq;
4451 	struct secpolicy *sp, *nextsp;
4452 	u_int dir;
4453 
4454 	LIST_INIT(&drainq);
4455 	SPTREE_RLOCK();
4456 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4457 		TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
4458 			if (sp->lifetime == 0 && sp->validtime == 0)
4459 				continue;
4460 			if ((sp->lifetime &&
4461 			    now - sp->created > sp->lifetime) ||
4462 			    (sp->validtime &&
4463 			    now - sp->lastused > sp->validtime)) {
4464 				/* Hold extra reference to send SPDEXPIRE */
4465 				SP_ADDREF(sp);
4466 				LIST_INSERT_HEAD(&drainq, sp, drainq);
4467 			}
4468 		}
4469 	}
4470 	SPTREE_RUNLOCK();
4471 	if (LIST_EMPTY(&drainq))
4472 		return;
4473 
4474 	SPTREE_WLOCK();
4475 	sp = LIST_FIRST(&drainq);
4476 	while (sp != NULL) {
4477 		nextsp = LIST_NEXT(sp, drainq);
4478 		/* Check that SP is still linked */
4479 		if (sp->state != IPSEC_SPSTATE_ALIVE) {
4480 			LIST_REMOVE(sp, drainq);
4481 			key_freesp(&sp); /* release extra reference */
4482 			sp = nextsp;
4483 			continue;
4484 		}
4485 		TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
4486 		V_spd_size--;
4487 		LIST_REMOVE(sp, idhash);
4488 		sp->state = IPSEC_SPSTATE_DEAD;
4489 		sp = nextsp;
4490 	}
4491 	V_sp_genid++;
4492 	SPTREE_WUNLOCK();
4493 	if (SPDCACHE_ENABLED())
4494 		spdcache_clear();
4495 
4496 	sp = LIST_FIRST(&drainq);
4497 	while (sp != NULL) {
4498 		nextsp = LIST_NEXT(sp, drainq);
4499 		key_spdexpire(sp);
4500 		key_freesp(&sp); /* release extra reference */
4501 		key_freesp(&sp); /* release last reference */
4502 		sp = nextsp;
4503 	}
4504 }
4505 
4506 static void
4507 key_flush_sad(time_t now)
4508 {
4509 	SAHTREE_RLOCK_TRACKER;
4510 	struct secashead_list emptyq;
4511 	struct secasvar_list drainq, hexpireq, sexpireq, freeq;
4512 	struct secashead *sah, *nextsah;
4513 	struct secasvar *sav, *nextsav;
4514 
4515 	SECASVAR_RLOCK_TRACKER;
4516 
4517 	LIST_INIT(&drainq);
4518 	LIST_INIT(&hexpireq);
4519 	LIST_INIT(&sexpireq);
4520 	LIST_INIT(&emptyq);
4521 
4522 	SAHTREE_RLOCK();
4523 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
4524 		/* Check for empty SAH */
4525 		if (TAILQ_EMPTY(&sah->savtree_larval) &&
4526 		    TAILQ_EMPTY(&sah->savtree_alive)) {
4527 			SAH_ADDREF(sah);
4528 			LIST_INSERT_HEAD(&emptyq, sah, drainq);
4529 			continue;
4530 		}
4531 		/* Add all stale LARVAL SAs into drainq */
4532 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
4533 			if (now - sav->created < V_key_larval_lifetime)
4534 				continue;
4535 			SAV_ADDREF(sav);
4536 			LIST_INSERT_HEAD(&drainq, sav, drainq);
4537 		}
4538 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
4539 			/* lifetimes aren't specified */
4540 			if (sav->lft_h == NULL)
4541 				continue;
4542 			SECASVAR_RLOCK(sav);
4543 			/*
4544 			 * Check again with lock held, because it may
4545 			 * be updated by SADB_UPDATE.
4546 			 */
4547 			if (sav->lft_h == NULL) {
4548 				SECASVAR_RUNLOCK(sav);
4549 				continue;
4550 			}
4551 			/*
4552 			 * RFC 2367:
4553 			 * HARD lifetimes MUST take precedence over SOFT
4554 			 * lifetimes, meaning if the HARD and SOFT lifetimes
4555 			 * are the same, the HARD lifetime will appear on the
4556 			 * EXPIRE message.
4557 			 */
4558 			/* check HARD lifetime */
4559 			if ((sav->lft_h->addtime != 0 &&
4560 			    now - sav->created > sav->lft_h->addtime) ||
4561 			    (sav->lft_h->usetime != 0 && sav->firstused &&
4562 			    now - sav->firstused > sav->lft_h->usetime) ||
4563 			    (sav->lft_h->bytes != 0 && counter_u64_fetch(
4564 			        sav->lft_c_bytes) > sav->lft_h->bytes)) {
4565 				SECASVAR_RUNLOCK(sav);
4566 				SAV_ADDREF(sav);
4567 				LIST_INSERT_HEAD(&hexpireq, sav, drainq);
4568 				continue;
4569 			}
4570 			/* check SOFT lifetime (only for MATURE SAs) */
4571 			if (sav->state == SADB_SASTATE_MATURE && (
4572 			    (sav->lft_s->addtime != 0 &&
4573 			    now - sav->created > sav->lft_s->addtime) ||
4574 			    (sav->lft_s->usetime != 0 && sav->firstused &&
4575 			    now - sav->firstused > sav->lft_s->usetime) ||
4576 			    (sav->lft_s->bytes != 0 && counter_u64_fetch(
4577 				sav->lft_c_bytes) > sav->lft_s->bytes) ||
4578 			    (!(sav->flags & SADB_X_SAFLAGS_ESN) &&
4579 			    (sav->replay != NULL) && (
4580 			    (sav->replay->count > UINT32_80PCT) ||
4581 			    (sav->replay->last > UINT32_80PCT))))) {
4582 				SECASVAR_RUNLOCK(sav);
4583 				SAV_ADDREF(sav);
4584 				LIST_INSERT_HEAD(&sexpireq, sav, drainq);
4585 				continue;
4586 			}
4587 			SECASVAR_RUNLOCK(sav);
4588 		}
4589 	}
4590 	SAHTREE_RUNLOCK();
4591 
4592 	if (LIST_EMPTY(&emptyq) && LIST_EMPTY(&drainq) &&
4593 	    LIST_EMPTY(&hexpireq) && LIST_EMPTY(&sexpireq))
4594 		return;
4595 
4596 	LIST_INIT(&freeq);
4597 	SAHTREE_WLOCK();
4598 	/* Unlink stale LARVAL SAs */
4599 	sav = LIST_FIRST(&drainq);
4600 	while (sav != NULL) {
4601 		nextsav = LIST_NEXT(sav, drainq);
4602 		/* Check that SA is still LARVAL */
4603 		if (sav->state != SADB_SASTATE_LARVAL) {
4604 			LIST_REMOVE(sav, drainq);
4605 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4606 			sav = nextsav;
4607 			continue;
4608 		}
4609 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
4610 		LIST_REMOVE(sav, spihash);
4611 		sav->state = SADB_SASTATE_DEAD;
4612 		sav = nextsav;
4613 	}
4614 	/* Unlink all SAs with expired HARD lifetime */
4615 	sav = LIST_FIRST(&hexpireq);
4616 	while (sav != NULL) {
4617 		nextsav = LIST_NEXT(sav, drainq);
4618 		/* Check that SA is not unlinked */
4619 		if (sav->state == SADB_SASTATE_DEAD) {
4620 			LIST_REMOVE(sav, drainq);
4621 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4622 			sav = nextsav;
4623 			continue;
4624 		}
4625 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
4626 		LIST_REMOVE(sav, spihash);
4627 		sav->state = SADB_SASTATE_DEAD;
4628 		sav = nextsav;
4629 	}
4630 	/* Mark all SAs with expired SOFT lifetime as DYING */
4631 	sav = LIST_FIRST(&sexpireq);
4632 	while (sav != NULL) {
4633 		nextsav = LIST_NEXT(sav, drainq);
4634 		/* Check that SA is not unlinked */
4635 		if (sav->state == SADB_SASTATE_DEAD) {
4636 			LIST_REMOVE(sav, drainq);
4637 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4638 			sav = nextsav;
4639 			continue;
4640 		}
4641 		/*
4642 		 * NOTE: this doesn't change SA order in the chain.
4643 		 */
4644 		sav->state = SADB_SASTATE_DYING;
4645 		sav = nextsav;
4646 	}
4647 	/* Unlink empty SAHs */
4648 	sah = LIST_FIRST(&emptyq);
4649 	while (sah != NULL) {
4650 		nextsah = LIST_NEXT(sah, drainq);
4651 		/* Check that SAH is still empty and not unlinked */
4652 		if (sah->state == SADB_SASTATE_DEAD ||
4653 		    !TAILQ_EMPTY(&sah->savtree_larval) ||
4654 		    !TAILQ_EMPTY(&sah->savtree_alive)) {
4655 			LIST_REMOVE(sah, drainq);
4656 			key_freesah(&sah); /* release extra reference */
4657 			sah = nextsah;
4658 			continue;
4659 		}
4660 		TAILQ_REMOVE(&V_sahtree, sah, chain);
4661 		LIST_REMOVE(sah, addrhash);
4662 		sah->state = SADB_SASTATE_DEAD;
4663 		sah = nextsah;
4664 	}
4665 	SAHTREE_WUNLOCK();
4666 
4667 	/* Send SPDEXPIRE messages */
4668 	sav = LIST_FIRST(&hexpireq);
4669 	while (sav != NULL) {
4670 		nextsav = LIST_NEXT(sav, drainq);
4671 		key_expire(sav, 1);
4672 		key_freesah(&sav->sah); /* release reference from SAV */
4673 		key_freesav(&sav); /* release extra reference */
4674 		key_freesav(&sav); /* release last reference */
4675 		sav = nextsav;
4676 	}
4677 	sav = LIST_FIRST(&sexpireq);
4678 	while (sav != NULL) {
4679 		nextsav = LIST_NEXT(sav, drainq);
4680 		key_expire(sav, 0);
4681 		key_freesav(&sav); /* release extra reference */
4682 		sav = nextsav;
4683 	}
4684 	/* Free stale LARVAL SAs */
4685 	sav = LIST_FIRST(&drainq);
4686 	while (sav != NULL) {
4687 		nextsav = LIST_NEXT(sav, drainq);
4688 		key_freesah(&sav->sah); /* release reference from SAV */
4689 		key_freesav(&sav); /* release extra reference */
4690 		key_freesav(&sav); /* release last reference */
4691 		sav = nextsav;
4692 	}
4693 	/* Free SAs that were unlinked/changed by someone else */
4694 	sav = LIST_FIRST(&freeq);
4695 	while (sav != NULL) {
4696 		nextsav = LIST_NEXT(sav, drainq);
4697 		key_freesav(&sav); /* release extra reference */
4698 		sav = nextsav;
4699 	}
4700 	/* Free empty SAH */
4701 	sah = LIST_FIRST(&emptyq);
4702 	while (sah != NULL) {
4703 		nextsah = LIST_NEXT(sah, drainq);
4704 		key_freesah(&sah); /* release extra reference */
4705 		key_freesah(&sah); /* release last reference */
4706 		sah = nextsah;
4707 	}
4708 }
4709 
4710 static void
4711 key_flush_acq(time_t now)
4712 {
4713 	struct secacq *acq, *nextacq;
4714 
4715 	/* ACQ tree */
4716 	ACQ_LOCK();
4717 	acq = LIST_FIRST(&V_acqtree);
4718 	while (acq != NULL) {
4719 		nextacq = LIST_NEXT(acq, chain);
4720 		if (now - acq->created > V_key_blockacq_lifetime) {
4721 			LIST_REMOVE(acq, chain);
4722 			LIST_REMOVE(acq, addrhash);
4723 			LIST_REMOVE(acq, seqhash);
4724 			free(acq, M_IPSEC_SAQ);
4725 		}
4726 		acq = nextacq;
4727 	}
4728 	ACQ_UNLOCK();
4729 }
4730 
4731 static void
4732 key_flush_spacq(time_t now)
4733 {
4734 	struct secspacq *acq, *nextacq;
4735 
4736 	/* SP ACQ tree */
4737 	SPACQ_LOCK();
4738 	for (acq = LIST_FIRST(&V_spacqtree); acq != NULL; acq = nextacq) {
4739 		nextacq = LIST_NEXT(acq, chain);
4740 		if (now - acq->created > V_key_blockacq_lifetime
4741 		 && __LIST_CHAINED(acq)) {
4742 			LIST_REMOVE(acq, chain);
4743 			free(acq, M_IPSEC_SAQ);
4744 		}
4745 	}
4746 	SPACQ_UNLOCK();
4747 }
4748 
4749 /*
4750  * time handler.
4751  * scanning SPD and SAD to check status for each entries,
4752  * and do to remove or to expire.
4753  * XXX: year 2038 problem may remain.
4754  */
4755 static void
4756 key_timehandler(void *arg)
4757 {
4758 	VNET_ITERATOR_DECL(vnet_iter);
4759 	time_t now = time_second;
4760 
4761 	VNET_LIST_RLOCK_NOSLEEP();
4762 	VNET_FOREACH(vnet_iter) {
4763 		CURVNET_SET(vnet_iter);
4764 		key_flush_spd(now);
4765 		key_flush_sad(now);
4766 		key_flush_acq(now);
4767 		key_flush_spacq(now);
4768 		CURVNET_RESTORE();
4769 	}
4770 	VNET_LIST_RUNLOCK_NOSLEEP();
4771 
4772 #ifndef IPSEC_DEBUG2
4773 	/* do exchange to tick time !! */
4774 	callout_schedule(&key_timer, hz);
4775 #endif /* IPSEC_DEBUG2 */
4776 }
4777 
4778 u_long
4779 key_random(void)
4780 {
4781 	u_long value;
4782 
4783 	arc4random_buf(&value, sizeof(value));
4784 	return value;
4785 }
4786 
4787 /*
4788  * map SADB_SATYPE_* to IPPROTO_*.
4789  * if satype == SADB_SATYPE then satype is mapped to ~0.
4790  * OUT:
4791  *	0: invalid satype.
4792  */
4793 static uint8_t
4794 key_satype2proto(uint8_t satype)
4795 {
4796 	switch (satype) {
4797 	case SADB_SATYPE_UNSPEC:
4798 		return IPSEC_PROTO_ANY;
4799 	case SADB_SATYPE_AH:
4800 		return IPPROTO_AH;
4801 	case SADB_SATYPE_ESP:
4802 		return IPPROTO_ESP;
4803 	case SADB_X_SATYPE_IPCOMP:
4804 		return IPPROTO_IPCOMP;
4805 	case SADB_X_SATYPE_TCPSIGNATURE:
4806 		return IPPROTO_TCP;
4807 	default:
4808 		return 0;
4809 	}
4810 	/* NOTREACHED */
4811 }
4812 
4813 /*
4814  * map IPPROTO_* to SADB_SATYPE_*
4815  * OUT:
4816  *	0: invalid protocol type.
4817  */
4818 static uint8_t
4819 key_proto2satype(uint8_t proto)
4820 {
4821 	switch (proto) {
4822 	case IPPROTO_AH:
4823 		return SADB_SATYPE_AH;
4824 	case IPPROTO_ESP:
4825 		return SADB_SATYPE_ESP;
4826 	case IPPROTO_IPCOMP:
4827 		return SADB_X_SATYPE_IPCOMP;
4828 	case IPPROTO_TCP:
4829 		return SADB_X_SATYPE_TCPSIGNATURE;
4830 	default:
4831 		return 0;
4832 	}
4833 	/* NOTREACHED */
4834 }
4835 
4836 /* %%% PF_KEY */
4837 /*
4838  * SADB_GETSPI processing is to receive
4839  *	<base, (SA2), src address, dst address, (SPI range)>
4840  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
4841  * tree with the status of LARVAL, and send
4842  *	<base, SA(*), address(SD)>
4843  * to the IKMPd.
4844  *
4845  * IN:	mhp: pointer to the pointer to each header.
4846  * OUT:	NULL if fail.
4847  *	other if success, return pointer to the message to send.
4848  */
4849 static int
4850 key_getspi(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
4851 {
4852 	struct secasindex saidx;
4853 	struct sadb_address *src0, *dst0;
4854 	struct secasvar *sav;
4855 	uint32_t reqid, spi;
4856 	int error;
4857 	uint8_t mode, proto;
4858 
4859 	IPSEC_ASSERT(so != NULL, ("null socket"));
4860 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
4861 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
4862 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
4863 
4864 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
4865 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)
4866 #ifdef PFKEY_STRICT_CHECKS
4867 	    || SADB_CHECKHDR(mhp, SADB_EXT_SPIRANGE)
4868 #endif
4869 	    ) {
4870 		ipseclog((LOG_DEBUG,
4871 		    "%s: invalid message: missing required header.\n",
4872 		    __func__));
4873 		error = EINVAL;
4874 		goto fail;
4875 	}
4876 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
4877 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)
4878 #ifdef PFKEY_STRICT_CHECKS
4879 	    || SADB_CHECKLEN(mhp, SADB_EXT_SPIRANGE)
4880 #endif
4881 	    ) {
4882 		ipseclog((LOG_DEBUG,
4883 		    "%s: invalid message: wrong header size.\n", __func__));
4884 		error = EINVAL;
4885 		goto fail;
4886 	}
4887 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
4888 		mode = IPSEC_MODE_ANY;
4889 		reqid = 0;
4890 	} else {
4891 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
4892 			ipseclog((LOG_DEBUG,
4893 			    "%s: invalid message: wrong header size.\n",
4894 			    __func__));
4895 			error = EINVAL;
4896 			goto fail;
4897 		}
4898 		mode = ((struct sadb_x_sa2 *)
4899 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
4900 		reqid = ((struct sadb_x_sa2 *)
4901 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
4902 	}
4903 
4904 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
4905 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
4906 
4907 	/* map satype to proto */
4908 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
4909 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
4910 			__func__));
4911 		error = EINVAL;
4912 		goto fail;
4913 	}
4914 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
4915 	    (struct sockaddr *)(dst0 + 1));
4916 	if (error != 0) {
4917 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
4918 		error = EINVAL;
4919 		goto fail;
4920 	}
4921 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
4922 
4923 	/* SPI allocation */
4924 	SPI_ALLOC_LOCK();
4925 	spi = key_do_getnewspi(
4926 	    (struct sadb_spirange *)mhp->ext[SADB_EXT_SPIRANGE], &saidx);
4927 	if (spi == 0) {
4928 		/*
4929 		 * Requested SPI or SPI range is not available or
4930 		 * already used.
4931 		 */
4932 		SPI_ALLOC_UNLOCK();
4933 		error = EEXIST;
4934 		goto fail;
4935 	}
4936 	sav = key_newsav(mhp, &saidx, spi, &error);
4937 	SPI_ALLOC_UNLOCK();
4938 	if (sav == NULL)
4939 		goto fail;
4940 
4941 	if (sav->seq != 0) {
4942 		/*
4943 		 * RFC2367:
4944 		 * If the SADB_GETSPI message is in response to a
4945 		 * kernel-generated SADB_ACQUIRE, the sadb_msg_seq
4946 		 * MUST be the same as the SADB_ACQUIRE message.
4947 		 *
4948 		 * XXXAE: However it doesn't definethe behaviour how to
4949 		 * check this and what to do if it doesn't match.
4950 		 * Also what we should do if it matches?
4951 		 *
4952 		 * We can compare saidx used in SADB_ACQUIRE with saidx
4953 		 * used in SADB_GETSPI, but this probably can break
4954 		 * existing software. For now just warn if it doesn't match.
4955 		 *
4956 		 * XXXAE: anyway it looks useless.
4957 		 */
4958 		key_acqdone(&saidx, sav->seq);
4959 	}
4960 	KEYDBG(KEY_STAMP,
4961 	    printf("%s: SA(%p)\n", __func__, sav));
4962 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
4963 
4964     {
4965 	struct mbuf *n, *nn;
4966 	struct sadb_sa *m_sa;
4967 	struct sadb_msg *newmsg;
4968 	int off, len;
4969 
4970 	/* create new sadb_msg to reply. */
4971 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
4972 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
4973 
4974 	MGETHDR(n, M_NOWAIT, MT_DATA);
4975 	if (len > MHLEN) {
4976 		if (!(MCLGET(n, M_NOWAIT))) {
4977 			m_freem(n);
4978 			n = NULL;
4979 		}
4980 	}
4981 	if (!n) {
4982 		error = ENOBUFS;
4983 		goto fail;
4984 	}
4985 
4986 	n->m_len = len;
4987 	n->m_next = NULL;
4988 	off = 0;
4989 
4990 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
4991 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
4992 
4993 	m_sa = (struct sadb_sa *)(mtod(n, caddr_t) + off);
4994 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
4995 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
4996 	m_sa->sadb_sa_spi = spi; /* SPI is already in network byte order */
4997 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
4998 
4999 	IPSEC_ASSERT(off == len,
5000 		("length inconsistency (off %u len %u)", off, len));
5001 
5002 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
5003 	    SADB_EXT_ADDRESS_DST);
5004 	if (!n->m_next) {
5005 		m_freem(n);
5006 		error = ENOBUFS;
5007 		goto fail;
5008 	}
5009 
5010 	if (n->m_len < sizeof(struct sadb_msg)) {
5011 		n = m_pullup(n, sizeof(struct sadb_msg));
5012 		if (n == NULL)
5013 			return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
5014 	}
5015 
5016 	n->m_pkthdr.len = 0;
5017 	for (nn = n; nn; nn = nn->m_next)
5018 		n->m_pkthdr.len += nn->m_len;
5019 
5020 	newmsg = mtod(n, struct sadb_msg *);
5021 	newmsg->sadb_msg_seq = sav->seq;
5022 	newmsg->sadb_msg_errno = 0;
5023 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
5024 
5025 	m_freem(m);
5026 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
5027     }
5028 
5029 fail:
5030 	return (key_senderror(so, m, error));
5031 }
5032 
5033 /*
5034  * allocating new SPI
5035  * called by key_getspi().
5036  * OUT:
5037  *	0:	failure.
5038  *	others: success, SPI in network byte order.
5039  */
5040 static uint32_t
5041 key_do_getnewspi(struct sadb_spirange *spirange, struct secasindex *saidx)
5042 {
5043 	uint32_t min, max, newspi, t;
5044 	int tries, limit;
5045 
5046 	SPI_ALLOC_LOCK_ASSERT();
5047 
5048 	/* set spi range to allocate */
5049 	if (spirange != NULL) {
5050 		min = spirange->sadb_spirange_min;
5051 		max = spirange->sadb_spirange_max;
5052 	} else {
5053 		min = V_key_spi_minval;
5054 		max = V_key_spi_maxval;
5055 	}
5056 	/* IPCOMP needs 2-byte SPI */
5057 	if (saidx->proto == IPPROTO_IPCOMP) {
5058 		if (min >= 0x10000)
5059 			min = 0xffff;
5060 		if (max >= 0x10000)
5061 			max = 0xffff;
5062 		if (min > max) {
5063 			t = min; min = max; max = t;
5064 		}
5065 	}
5066 
5067 	if (min == max) {
5068 		if (key_checkspidup(htonl(min))) {
5069 			ipseclog((LOG_DEBUG, "%s: SPI %u exists already.\n",
5070 			    __func__, min));
5071 			return 0;
5072 		}
5073 
5074 		tries = 1;
5075 		newspi = min;
5076 	} else {
5077 		/* init SPI */
5078 		newspi = 0;
5079 
5080 		limit = atomic_load_int(&V_key_spi_trycnt);
5081 		/* when requesting to allocate spi ranged */
5082 		for (tries = 0; tries < limit; tries++) {
5083 			/* generate pseudo-random SPI value ranged. */
5084 			newspi = min + (key_random() % (max - min + 1));
5085 			if (!key_checkspidup(htonl(newspi)))
5086 				break;
5087 		}
5088 
5089 		if (tries == limit || newspi == 0) {
5090 			ipseclog((LOG_DEBUG,
5091 			    "%s: failed to allocate SPI.\n", __func__));
5092 			return 0;
5093 		}
5094 	}
5095 
5096 	/* statistics */
5097 	keystat.getspi_count =
5098 	    (keystat.getspi_count + tries) / 2;
5099 
5100 	return (htonl(newspi));
5101 }
5102 
5103 /*
5104  * Find TCP-MD5 SA with corresponding secasindex.
5105  * If not found, return NULL and fill SPI with usable value if needed.
5106  */
5107 static struct secasvar *
5108 key_getsav_tcpmd5(struct secasindex *saidx, uint32_t *spi)
5109 {
5110 	SAHTREE_RLOCK_TRACKER;
5111 	struct secashead *sah;
5112 	struct secasvar *sav;
5113 
5114 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP, ("wrong proto"));
5115 	SAHTREE_RLOCK();
5116 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
5117 		if (sah->saidx.proto != IPPROTO_TCP)
5118 			continue;
5119 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
5120 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
5121 			break;
5122 	}
5123 	if (sah != NULL) {
5124 		if (V_key_preferred_oldsa)
5125 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
5126 		else
5127 			sav = TAILQ_FIRST(&sah->savtree_alive);
5128 		if (sav != NULL) {
5129 			SAV_ADDREF(sav);
5130 			SAHTREE_RUNLOCK();
5131 			return (sav);
5132 		}
5133 	}
5134 	if (spi == NULL) {
5135 		/* No SPI required */
5136 		SAHTREE_RUNLOCK();
5137 		return (NULL);
5138 	}
5139 	/* Check that SPI is unique */
5140 	LIST_FOREACH(sav, SAVHASH_HASH(*spi), spihash) {
5141 		if (sav->spi == *spi)
5142 			break;
5143 	}
5144 	if (sav == NULL) {
5145 		SAHTREE_RUNLOCK();
5146 		/* SPI is already unique */
5147 		return (NULL);
5148 	}
5149 	SAHTREE_RUNLOCK();
5150 	/* XXX: not optimal */
5151 	*spi = key_do_getnewspi(NULL, saidx);
5152 	return (NULL);
5153 }
5154 
5155 static int
5156 key_updateaddresses(struct socket *so, struct mbuf *m,
5157     const struct sadb_msghdr *mhp, struct secasvar *sav,
5158     struct secasindex *saidx)
5159 {
5160 	struct sockaddr *newaddr;
5161 	struct secashead *sah;
5162 	struct secasvar *newsav, *tmp;
5163 	struct mbuf *n;
5164 	int error, isnew;
5165 
5166 	/* Check that we need to change SAH */
5167 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC)) {
5168 		newaddr = (struct sockaddr *)(
5169 		    ((struct sadb_address *)
5170 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_SRC]) + 1);
5171 		bcopy(newaddr, &saidx->src, newaddr->sa_len);
5172 		key_porttosaddr(&saidx->src.sa, 0);
5173 	}
5174 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5175 		newaddr = (struct sockaddr *)(
5176 		    ((struct sadb_address *)
5177 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_DST]) + 1);
5178 		bcopy(newaddr, &saidx->dst, newaddr->sa_len);
5179 		key_porttosaddr(&saidx->dst.sa, 0);
5180 	}
5181 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5182 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5183 		error = key_checksockaddrs(&saidx->src.sa, &saidx->dst.sa);
5184 		if (error != 0) {
5185 			ipseclog((LOG_DEBUG, "%s: invalid new sockaddr.\n",
5186 			    __func__));
5187 			return (error);
5188 		}
5189 
5190 		sah = key_getsah(saidx);
5191 		if (sah == NULL) {
5192 			/* create a new SA index */
5193 			sah = key_newsah(saidx);
5194 			if (sah == NULL) {
5195 				ipseclog((LOG_DEBUG,
5196 				    "%s: No more memory.\n", __func__));
5197 				return (ENOBUFS);
5198 			}
5199 			isnew = 2; /* SAH is new */
5200 		} else
5201 			isnew = 1; /* existing SAH is referenced */
5202 	} else {
5203 		/*
5204 		 * src and dst addresses are still the same.
5205 		 * Do we want to change NAT-T config?
5206 		 */
5207 		if (sav->sah->saidx.proto != IPPROTO_ESP ||
5208 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5209 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5210 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5211 			ipseclog((LOG_DEBUG,
5212 			    "%s: invalid message: missing required header.\n",
5213 			    __func__));
5214 			return (EINVAL);
5215 		}
5216 		/* We hold reference to SA, thus SAH will be referenced too. */
5217 		sah = sav->sah;
5218 		isnew = 0;
5219 	}
5220 
5221 	newsav = malloc(sizeof(struct secasvar), M_IPSEC_SA,
5222 	    M_NOWAIT | M_ZERO);
5223 	if (newsav == NULL) {
5224 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5225 		error = ENOBUFS;
5226 		goto fail;
5227 	}
5228 
5229 	/* Clone SA's content into newsav */
5230 	SAV_INITREF(newsav);
5231 	bcopy(sav, newsav, offsetof(struct secasvar, chain));
5232 	/*
5233 	 * We create new NAT-T config if it is needed.
5234 	 * Old NAT-T config will be freed by key_cleansav() when
5235 	 * last reference to SA will be released.
5236 	 */
5237 	newsav->natt = NULL;
5238 	newsav->sah = sah;
5239 	newsav->state = SADB_SASTATE_MATURE;
5240 	error = key_setnatt(newsav, mhp);
5241 	if (error != 0)
5242 		goto fail;
5243 
5244 	SAHTREE_WLOCK();
5245 	/* Check that SA is still alive */
5246 	if (sav->state == SADB_SASTATE_DEAD) {
5247 		/* SA was unlinked */
5248 		SAHTREE_WUNLOCK();
5249 		error = ESRCH;
5250 		goto fail;
5251 	}
5252 
5253 	/* Unlink SA from SAH and SPI hash */
5254 	IPSEC_ASSERT((sav->flags & SADB_X_EXT_F_CLONED) == 0,
5255 	    ("SA is already cloned"));
5256 	IPSEC_ASSERT(sav->state == SADB_SASTATE_MATURE ||
5257 	    sav->state == SADB_SASTATE_DYING,
5258 	    ("Wrong SA state %u\n", sav->state));
5259 	TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
5260 	LIST_REMOVE(sav, spihash);
5261 	sav->state = SADB_SASTATE_DEAD;
5262 
5263 	/*
5264 	 * Link new SA with SAH. Keep SAs ordered by
5265 	 * create time (newer are first).
5266 	 */
5267 	TAILQ_FOREACH(tmp, &sah->savtree_alive, chain) {
5268 		if (newsav->created > tmp->created) {
5269 			TAILQ_INSERT_BEFORE(tmp, newsav, chain);
5270 			break;
5271 		}
5272 	}
5273 	if (tmp == NULL)
5274 		TAILQ_INSERT_TAIL(&sah->savtree_alive, newsav, chain);
5275 
5276 	/* Add new SA into SPI hash. */
5277 	LIST_INSERT_HEAD(SAVHASH_HASH(newsav->spi), newsav, spihash);
5278 
5279 	/* Add new SAH into SADB. */
5280 	if (isnew == 2) {
5281 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
5282 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
5283 		sah->state = SADB_SASTATE_MATURE;
5284 		SAH_ADDREF(sah); /* newsav references new SAH */
5285 	}
5286 	/*
5287 	 * isnew == 1 -> @sah was referenced by key_getsah().
5288 	 * isnew == 0 -> we use the same @sah, that was used by @sav,
5289 	 *	and we use its reference for @newsav.
5290 	 */
5291 	SECASVAR_WLOCK(sav);
5292 	/* XXX: replace cntr with pointer? */
5293 	newsav->cntr = sav->cntr;
5294 	sav->flags |= SADB_X_EXT_F_CLONED;
5295 	SECASVAR_WUNLOCK(sav);
5296 
5297 	SAHTREE_WUNLOCK();
5298 
5299 	KEYDBG(KEY_STAMP,
5300 	    printf("%s: SA(%p) cloned into SA(%p)\n",
5301 	    __func__, sav, newsav));
5302 	KEYDBG(KEY_DATA, kdebug_secasv(newsav));
5303 
5304 	key_freesav(&sav); /* release last reference */
5305 
5306 	/* set msg buf from mhp */
5307 	n = key_getmsgbuf_x1(m, mhp);
5308 	if (n == NULL) {
5309 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5310 		return (ENOBUFS);
5311 	}
5312 	m_freem(m);
5313 	key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5314 	return (0);
5315 fail:
5316 	if (isnew != 0)
5317 		key_freesah(&sah);
5318 	if (newsav != NULL) {
5319 		if (newsav->natt != NULL)
5320 			free(newsav->natt, M_IPSEC_MISC);
5321 		free(newsav, M_IPSEC_SA);
5322 	}
5323 	return (error);
5324 }
5325 
5326 /*
5327  * SADB_UPDATE processing
5328  * receive
5329  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5330  *       key(AE), (identity(SD),) (sensitivity)>
5331  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5332  * and send
5333  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5334  *       (identity(SD),) (sensitivity)>
5335  * to the ikmpd.
5336  *
5337  * m will always be freed.
5338  */
5339 static int
5340 key_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5341 {
5342 	struct secasindex saidx;
5343 	struct sadb_address *src0, *dst0;
5344 	struct sadb_sa *sa0;
5345 	struct secasvar *sav;
5346 	uint32_t reqid;
5347 	int error;
5348 	uint8_t mode, proto;
5349 
5350 	IPSEC_ASSERT(so != NULL, ("null socket"));
5351 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5352 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5353 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5354 
5355 	/* map satype to proto */
5356 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5357 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5358 		    __func__));
5359 		return key_senderror(so, m, EINVAL);
5360 	}
5361 
5362 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5363 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5364 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5365 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5366 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5367 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5368 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5369 		ipseclog((LOG_DEBUG,
5370 		    "%s: invalid message: missing required header.\n",
5371 		    __func__));
5372 		return key_senderror(so, m, EINVAL);
5373 	}
5374 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5375 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5376 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5377 		ipseclog((LOG_DEBUG,
5378 		    "%s: invalid message: wrong header size.\n", __func__));
5379 		return key_senderror(so, m, EINVAL);
5380 	}
5381 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5382 		mode = IPSEC_MODE_ANY;
5383 		reqid = 0;
5384 	} else {
5385 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5386 			ipseclog((LOG_DEBUG,
5387 			    "%s: invalid message: wrong header size.\n",
5388 			    __func__));
5389 			return key_senderror(so, m, EINVAL);
5390 		}
5391 		mode = ((struct sadb_x_sa2 *)
5392 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5393 		reqid = ((struct sadb_x_sa2 *)
5394 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5395 	}
5396 
5397 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5398 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5399 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5400 
5401 	/*
5402 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5403 	 * SADB_UPDATE message.
5404 	 */
5405 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5406 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5407 #ifdef PFKEY_STRICT_CHECKS
5408 		return key_senderror(so, m, EINVAL);
5409 #endif
5410 	}
5411 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5412 	    (struct sockaddr *)(dst0 + 1));
5413 	if (error != 0) {
5414 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5415 		return key_senderror(so, m, error);
5416 	}
5417 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5418 	sav = key_getsavbyspi(sa0->sadb_sa_spi);
5419 	if (sav == NULL) {
5420 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u\n",
5421 		    __func__, ntohl(sa0->sadb_sa_spi)));
5422 		return key_senderror(so, m, EINVAL);
5423 	}
5424 	/*
5425 	 * Check that SADB_UPDATE issued by the same process that did
5426 	 * SADB_GETSPI or SADB_ADD.
5427 	 */
5428 	if (sav->pid != mhp->msg->sadb_msg_pid) {
5429 		ipseclog((LOG_DEBUG,
5430 		    "%s: pid mismatched (SPI %u, pid %u vs. %u)\n", __func__,
5431 		    ntohl(sav->spi), sav->pid, mhp->msg->sadb_msg_pid));
5432 		key_freesav(&sav);
5433 		return key_senderror(so, m, EINVAL);
5434 	}
5435 	/* saidx should match with SA. */
5436 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_MODE_REQID) == 0) {
5437 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u\n",
5438 		    __func__, ntohl(sav->spi)));
5439 		key_freesav(&sav);
5440 		return key_senderror(so, m, ESRCH);
5441 	}
5442 
5443 	if (sav->state == SADB_SASTATE_LARVAL) {
5444 		if ((mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5445 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) ||
5446 		    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5447 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH))) {
5448 			ipseclog((LOG_DEBUG,
5449 			    "%s: invalid message: missing required header.\n",
5450 			    __func__));
5451 			key_freesav(&sav);
5452 			return key_senderror(so, m, EINVAL);
5453 		}
5454 		/*
5455 		 * We can set any values except src, dst and SPI.
5456 		 */
5457 		error = key_setsaval(sav, mhp);
5458 		if (error != 0) {
5459 			key_freesav(&sav);
5460 			return (key_senderror(so, m, error));
5461 		}
5462 		/* Change SA state to MATURE */
5463 		SAHTREE_WLOCK();
5464 		if (sav->state != SADB_SASTATE_LARVAL) {
5465 			/* SA was deleted or another thread made it MATURE. */
5466 			SAHTREE_WUNLOCK();
5467 			key_freesav(&sav);
5468 			return (key_senderror(so, m, ESRCH));
5469 		}
5470 		/*
5471 		 * NOTE: we keep SAs in savtree_alive ordered by created
5472 		 * time. When SA's state changed from LARVAL to MATURE,
5473 		 * we update its created time in key_setsaval() and move
5474 		 * it into head of savtree_alive.
5475 		 */
5476 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
5477 		TAILQ_INSERT_HEAD(&sav->sah->savtree_alive, sav, chain);
5478 		sav->state = SADB_SASTATE_MATURE;
5479 		SAHTREE_WUNLOCK();
5480 	} else {
5481 		/*
5482 		 * For DYING and MATURE SA we can change only state
5483 		 * and lifetimes. Report EINVAL if something else attempted
5484 		 * to change.
5485 		 */
5486 		if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5487 		    !SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
5488 			key_freesav(&sav);
5489 			return (key_senderror(so, m, EINVAL));
5490 		}
5491 		error = key_updatelifetimes(sav, mhp);
5492 		if (error != 0) {
5493 			key_freesav(&sav);
5494 			return (key_senderror(so, m, error));
5495 		}
5496 		/*
5497 		 * This is FreeBSD extension to RFC2367.
5498 		 * IKEd can specify SADB_X_EXT_NEW_ADDRESS_SRC and/or
5499 		 * SADB_X_EXT_NEW_ADDRESS_DST when it wants to change
5500 		 * SA addresses (for example to implement MOBIKE protocol
5501 		 * as described in RFC4555). Also we allow to change
5502 		 * NAT-T config.
5503 		 */
5504 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5505 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST) ||
5506 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5507 		    sav->natt != NULL) {
5508 			error = key_updateaddresses(so, m, mhp, sav, &saidx);
5509 			key_freesav(&sav);
5510 			if (error != 0)
5511 				return (key_senderror(so, m, error));
5512 			return (0);
5513 		}
5514 		/* Check that SA is still alive */
5515 		SAHTREE_WLOCK();
5516 		if (sav->state == SADB_SASTATE_DEAD) {
5517 			/* SA was unlinked */
5518 			SAHTREE_WUNLOCK();
5519 			key_freesav(&sav);
5520 			return (key_senderror(so, m, ESRCH));
5521 		}
5522 		/*
5523 		 * NOTE: there is possible state moving from DYING to MATURE,
5524 		 * but this doesn't change created time, so we won't reorder
5525 		 * this SA.
5526 		 */
5527 		sav->state = SADB_SASTATE_MATURE;
5528 		SAHTREE_WUNLOCK();
5529 	}
5530 	KEYDBG(KEY_STAMP,
5531 	    printf("%s: SA(%p)\n", __func__, sav));
5532 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5533 	key_freesav(&sav);
5534 
5535     {
5536 	struct mbuf *n;
5537 
5538 	/* set msg buf from mhp */
5539 	n = key_getmsgbuf_x1(m, mhp);
5540 	if (n == NULL) {
5541 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5542 		return key_senderror(so, m, ENOBUFS);
5543 	}
5544 
5545 	m_freem(m);
5546 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5547     }
5548 }
5549 
5550 /*
5551  * SADB_ADD processing
5552  * add an entry to SA database, when received
5553  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5554  *       key(AE), (identity(SD),) (sensitivity)>
5555  * from the ikmpd,
5556  * and send
5557  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5558  *       (identity(SD),) (sensitivity)>
5559  * to the ikmpd.
5560  *
5561  * IGNORE identity and sensitivity messages.
5562  *
5563  * m will always be freed.
5564  */
5565 static int
5566 key_add(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5567 {
5568 	struct secasindex saidx;
5569 	struct sadb_address *src0, *dst0;
5570 	struct sadb_sa *sa0;
5571 	struct secasvar *sav;
5572 	uint32_t reqid, spi;
5573 	uint8_t mode, proto;
5574 	int error;
5575 
5576 	IPSEC_ASSERT(so != NULL, ("null socket"));
5577 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5578 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5579 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5580 
5581 	/* map satype to proto */
5582 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5583 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5584 		    __func__));
5585 		return key_senderror(so, m, EINVAL);
5586 	}
5587 
5588 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5589 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5590 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5591 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP && (
5592 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5593 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT))) ||
5594 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH && (
5595 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH) ||
5596 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH))) ||
5597 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5598 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5599 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5600 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5601 		ipseclog((LOG_DEBUG,
5602 		    "%s: invalid message: missing required header.\n",
5603 		    __func__));
5604 		return key_senderror(so, m, EINVAL);
5605 	}
5606 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5607 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5608 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5609 		ipseclog((LOG_DEBUG,
5610 		    "%s: invalid message: wrong header size.\n", __func__));
5611 		return key_senderror(so, m, EINVAL);
5612 	}
5613 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5614 		mode = IPSEC_MODE_ANY;
5615 		reqid = 0;
5616 	} else {
5617 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5618 			ipseclog((LOG_DEBUG,
5619 			    "%s: invalid message: wrong header size.\n",
5620 			    __func__));
5621 			return key_senderror(so, m, EINVAL);
5622 		}
5623 		mode = ((struct sadb_x_sa2 *)
5624 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5625 		reqid = ((struct sadb_x_sa2 *)
5626 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5627 	}
5628 
5629 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5630 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
5631 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
5632 
5633 	/*
5634 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5635 	 * SADB_ADD message.
5636 	 */
5637 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5638 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5639 #ifdef PFKEY_STRICT_CHECKS
5640 		return key_senderror(so, m, EINVAL);
5641 #endif
5642 	}
5643 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5644 	    (struct sockaddr *)(dst0 + 1));
5645 	if (error != 0) {
5646 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5647 		return key_senderror(so, m, error);
5648 	}
5649 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5650 	spi = sa0->sadb_sa_spi;
5651 	/*
5652 	 * For TCP-MD5 SAs we don't use SPI. Check the uniqueness using
5653 	 * secasindex.
5654 	 * XXXAE: IPComp seems also doesn't use SPI.
5655 	 */
5656 	SPI_ALLOC_LOCK();
5657 	if (proto == IPPROTO_TCP) {
5658 		sav = key_getsav_tcpmd5(&saidx, &spi);
5659 		if (sav == NULL && spi == 0) {
5660 			SPI_ALLOC_UNLOCK();
5661 			/* Failed to allocate SPI */
5662 			ipseclog((LOG_DEBUG, "%s: SA already exists.\n",
5663 			    __func__));
5664 			return key_senderror(so, m, EEXIST);
5665 		}
5666 		/* XXX: SPI that we report back can have another value */
5667 	} else {
5668 		/* We can create new SA only if SPI is different. */
5669 		sav = key_getsavbyspi(spi);
5670 	}
5671 	if (sav != NULL) {
5672 		SPI_ALLOC_UNLOCK();
5673 		key_freesav(&sav);
5674 		ipseclog((LOG_DEBUG, "%s: SA already exists.\n", __func__));
5675 		return key_senderror(so, m, EEXIST);
5676 	}
5677 
5678 	sav = key_newsav(mhp, &saidx, spi, &error);
5679 	SPI_ALLOC_UNLOCK();
5680 	if (sav == NULL)
5681 		return key_senderror(so, m, error);
5682 	KEYDBG(KEY_STAMP,
5683 	    printf("%s: return SA(%p)\n", __func__, sav));
5684 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5685 	/*
5686 	 * If SADB_ADD was in response to SADB_ACQUIRE, we need to schedule
5687 	 * ACQ for deletion.
5688 	 */
5689 	if (sav->seq != 0)
5690 		key_acqdone(&saidx, sav->seq);
5691 
5692     {
5693 	/*
5694 	 * Don't call key_freesav() on error here, as we would like to
5695 	 * keep the SA in the database.
5696 	 */
5697 	struct mbuf *n;
5698 
5699 	/* set msg buf from mhp */
5700 	n = key_getmsgbuf_x1(m, mhp);
5701 	if (n == NULL) {
5702 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5703 		return key_senderror(so, m, ENOBUFS);
5704 	}
5705 
5706 	m_freem(m);
5707 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5708     }
5709 }
5710 
5711 /*
5712  * NAT-T support.
5713  * IKEd may request the use ESP in UDP encapsulation when it detects the
5714  * presence of NAT. It uses NAT-T extension headers for such SAs to specify
5715  * parameters needed for encapsulation and decapsulation. These PF_KEY
5716  * extension headers are not standardized, so this comment addresses our
5717  * implementation.
5718  * SADB_X_EXT_NAT_T_TYPE specifies type of encapsulation, we support only
5719  * UDP_ENCAP_ESPINUDP as described in RFC3948.
5720  * SADB_X_EXT_NAT_T_SPORT/DPORT specifies source and destination ports for
5721  * UDP header. We use these ports in UDP encapsulation procedure, also we
5722  * can check them in UDP decapsulation procedure.
5723  * SADB_X_EXT_NAT_T_OA[IR] specifies original address of initiator or
5724  * responder. These addresses can be used for transport mode to adjust
5725  * checksum after decapsulation and decryption. Since original IP addresses
5726  * used by peer usually different (we detected presence of NAT), TCP/UDP
5727  * pseudo header checksum and IP header checksum was calculated using original
5728  * addresses. After decapsulation and decryption we need to adjust checksum
5729  * to have correct datagram.
5730  *
5731  * We expect presence of NAT-T extension headers only in SADB_ADD and
5732  * SADB_UPDATE messages. We report NAT-T extension headers in replies
5733  * to SADB_ADD, SADB_UPDATE, SADB_GET, and SADB_DUMP messages.
5734  */
5735 static int
5736 key_setnatt(struct secasvar *sav, const struct sadb_msghdr *mhp)
5737 {
5738 	struct sadb_x_nat_t_port *port;
5739 	struct sadb_x_nat_t_type *type;
5740 	struct sadb_address *oai, *oar;
5741 	struct sockaddr *sa;
5742 	uint32_t addr;
5743 	uint16_t cksum;
5744 
5745 	IPSEC_ASSERT(sav->natt == NULL, ("natt is already initialized"));
5746 	/*
5747 	 * Ignore NAT-T headers if sproto isn't ESP.
5748 	 */
5749 	if (sav->sah->saidx.proto != IPPROTO_ESP)
5750 		return (0);
5751 
5752 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) &&
5753 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) &&
5754 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5755 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5756 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5757 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5758 			ipseclog((LOG_DEBUG,
5759 			    "%s: invalid message: wrong header size.\n",
5760 			    __func__));
5761 			return (EINVAL);
5762 		}
5763 	} else
5764 		return (0);
5765 
5766 	type = (struct sadb_x_nat_t_type *)mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5767 	if (type->sadb_x_nat_t_type_type != UDP_ENCAP_ESPINUDP) {
5768 		ipseclog((LOG_DEBUG, "%s: unsupported NAT-T type %u.\n",
5769 		    __func__, type->sadb_x_nat_t_type_type));
5770 		return (EINVAL);
5771 	}
5772 	/*
5773 	 * Allocate storage for NAT-T config.
5774 	 * On error it will be released by key_cleansav().
5775 	 */
5776 	sav->natt = malloc(sizeof(struct secnatt), M_IPSEC_MISC,
5777 	    M_NOWAIT | M_ZERO);
5778 	if (sav->natt == NULL) {
5779 		PFKEYSTAT_INC(in_nomem);
5780 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5781 		return (ENOBUFS);
5782 	}
5783 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5784 	if (port->sadb_x_nat_t_port_port == 0) {
5785 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T sport specified.\n",
5786 		    __func__));
5787 		return (EINVAL);
5788 	}
5789 	sav->natt->sport = port->sadb_x_nat_t_port_port;
5790 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5791 	if (port->sadb_x_nat_t_port_port == 0) {
5792 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T dport specified.\n",
5793 		    __func__));
5794 		return (EINVAL);
5795 	}
5796 	sav->natt->dport = port->sadb_x_nat_t_port_port;
5797 
5798 	/*
5799 	 * SADB_X_EXT_NAT_T_OAI and SADB_X_EXT_NAT_T_OAR are optional
5800 	 * and needed only for transport mode IPsec.
5801 	 * Usually NAT translates only one address, but it is possible,
5802 	 * that both addresses could be translated.
5803 	 * NOTE: Value of SADB_X_EXT_NAT_T_OAI is equal to SADB_X_EXT_NAT_T_OA.
5804 	 */
5805 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAI)) {
5806 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAI)) {
5807 			ipseclog((LOG_DEBUG,
5808 			    "%s: invalid message: wrong header size.\n",
5809 			    __func__));
5810 			return (EINVAL);
5811 		}
5812 		oai = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAI];
5813 	} else
5814 		oai = NULL;
5815 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAR)) {
5816 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAR)) {
5817 			ipseclog((LOG_DEBUG,
5818 			    "%s: invalid message: wrong header size.\n",
5819 			    __func__));
5820 			return (EINVAL);
5821 		}
5822 		oar = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAR];
5823 	} else
5824 		oar = NULL;
5825 
5826 	/* Initialize addresses only for transport mode */
5827 	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL) {
5828 		cksum = 0;
5829 		if (oai != NULL) {
5830 			/* Currently we support only AF_INET */
5831 			sa = (struct sockaddr *)(oai + 1);
5832 			if (sa->sa_family != AF_INET ||
5833 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5834 				ipseclog((LOG_DEBUG,
5835 				    "%s: wrong NAT-OAi header.\n",
5836 				    __func__));
5837 				return (EINVAL);
5838 			}
5839 			/* Ignore address if it the same */
5840 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5841 			    sav->sah->saidx.src.sin.sin_addr.s_addr) {
5842 				bcopy(sa, &sav->natt->oai.sa, sa->sa_len);
5843 				sav->natt->flags |= IPSEC_NATT_F_OAI;
5844 				/* Calculate checksum delta */
5845 				addr = sav->sah->saidx.src.sin.sin_addr.s_addr;
5846 				cksum = in_addword(cksum, ~addr >> 16);
5847 				cksum = in_addword(cksum, ~addr & 0xffff);
5848 				addr = sav->natt->oai.sin.sin_addr.s_addr;
5849 				cksum = in_addword(cksum, addr >> 16);
5850 				cksum = in_addword(cksum, addr & 0xffff);
5851 			}
5852 		}
5853 		if (oar != NULL) {
5854 			/* Currently we support only AF_INET */
5855 			sa = (struct sockaddr *)(oar + 1);
5856 			if (sa->sa_family != AF_INET ||
5857 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5858 				ipseclog((LOG_DEBUG,
5859 				    "%s: wrong NAT-OAr header.\n",
5860 				    __func__));
5861 				return (EINVAL);
5862 			}
5863 			/* Ignore address if it the same */
5864 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5865 			    sav->sah->saidx.dst.sin.sin_addr.s_addr) {
5866 				bcopy(sa, &sav->natt->oar.sa, sa->sa_len);
5867 				sav->natt->flags |= IPSEC_NATT_F_OAR;
5868 				/* Calculate checksum delta */
5869 				addr = sav->sah->saidx.dst.sin.sin_addr.s_addr;
5870 				cksum = in_addword(cksum, ~addr >> 16);
5871 				cksum = in_addword(cksum, ~addr & 0xffff);
5872 				addr = sav->natt->oar.sin.sin_addr.s_addr;
5873 				cksum = in_addword(cksum, addr >> 16);
5874 				cksum = in_addword(cksum, addr & 0xffff);
5875 			}
5876 		}
5877 		sav->natt->cksum = cksum;
5878 	}
5879 	return (0);
5880 }
5881 
5882 static int
5883 key_setident(struct secashead *sah, const struct sadb_msghdr *mhp)
5884 {
5885 	const struct sadb_ident *idsrc, *iddst;
5886 
5887 	IPSEC_ASSERT(sah != NULL, ("null secashead"));
5888 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5889 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5890 
5891 	/* don't make buffer if not there */
5892 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) &&
5893 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5894 		sah->idents = NULL;
5895 		sah->identd = NULL;
5896 		return (0);
5897 	}
5898 
5899 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) ||
5900 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5901 		ipseclog((LOG_DEBUG, "%s: invalid identity.\n", __func__));
5902 		return (EINVAL);
5903 	}
5904 
5905 	idsrc = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_SRC];
5906 	iddst = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_DST];
5907 
5908 	/* validity check */
5909 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
5910 		ipseclog((LOG_DEBUG, "%s: ident type mismatch.\n", __func__));
5911 		return EINVAL;
5912 	}
5913 
5914 	switch (idsrc->sadb_ident_type) {
5915 	case SADB_IDENTTYPE_PREFIX:
5916 	case SADB_IDENTTYPE_FQDN:
5917 	case SADB_IDENTTYPE_USERFQDN:
5918 	default:
5919 		/* XXX do nothing */
5920 		sah->idents = NULL;
5921 		sah->identd = NULL;
5922 	 	return 0;
5923 	}
5924 
5925 	/* make structure */
5926 	sah->idents = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5927 	if (sah->idents == NULL) {
5928 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5929 		return ENOBUFS;
5930 	}
5931 	sah->identd = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5932 	if (sah->identd == NULL) {
5933 		free(sah->idents, M_IPSEC_MISC);
5934 		sah->idents = NULL;
5935 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5936 		return ENOBUFS;
5937 	}
5938 	sah->idents->type = idsrc->sadb_ident_type;
5939 	sah->idents->id = idsrc->sadb_ident_id;
5940 
5941 	sah->identd->type = iddst->sadb_ident_type;
5942 	sah->identd->id = iddst->sadb_ident_id;
5943 
5944 	return 0;
5945 }
5946 
5947 /*
5948  * m will not be freed on return.
5949  * it is caller's responsibility to free the result.
5950  *
5951  * Called from SADB_ADD and SADB_UPDATE. Reply will contain headers
5952  * from the request in defined order.
5953  */
5954 static struct mbuf *
5955 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
5956 {
5957 	struct mbuf *n;
5958 
5959 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5960 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5961 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5962 
5963 	/* create new sadb_msg to reply. */
5964 	n = key_gather_mbuf(m, mhp, 1, 16, SADB_EXT_RESERVED,
5965 	    SADB_EXT_SA, SADB_X_EXT_SA2,
5966 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
5967 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
5968 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
5969 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
5970 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
5971 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NEW_ADDRESS_SRC,
5972 	    SADB_X_EXT_NEW_ADDRESS_DST);
5973 	if (!n)
5974 		return NULL;
5975 
5976 	if (n->m_len < sizeof(struct sadb_msg)) {
5977 		n = m_pullup(n, sizeof(struct sadb_msg));
5978 		if (n == NULL)
5979 			return NULL;
5980 	}
5981 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
5982 	mtod(n, struct sadb_msg *)->sadb_msg_len =
5983 	    PFKEY_UNIT64(n->m_pkthdr.len);
5984 
5985 	return n;
5986 }
5987 
5988 /*
5989  * SADB_DELETE processing
5990  * receive
5991  *   <base, SA(*), address(SD)>
5992  * from the ikmpd, and set SADB_SASTATE_DEAD,
5993  * and send,
5994  *   <base, SA(*), address(SD)>
5995  * to the ikmpd.
5996  *
5997  * m will always be freed.
5998  */
5999 static int
6000 key_delete(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6001 {
6002 	struct secasindex saidx;
6003 	struct sadb_address *src0, *dst0;
6004 	struct secasvar *sav;
6005 	struct sadb_sa *sa0;
6006 	uint8_t proto;
6007 
6008 	IPSEC_ASSERT(so != NULL, ("null socket"));
6009 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6010 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6011 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6012 
6013 	/* map satype to proto */
6014 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6015 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6016 		    __func__));
6017 		return key_senderror(so, m, EINVAL);
6018 	}
6019 
6020 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6021 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
6022 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6023 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6024 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
6025 		    __func__));
6026 		return key_senderror(so, m, EINVAL);
6027 	}
6028 
6029 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
6030 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
6031 
6032 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6033 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6034 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6035 		return (key_senderror(so, m, EINVAL));
6036 	}
6037 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6038 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
6039 		/*
6040 		 * Caller wants us to delete all non-LARVAL SAs
6041 		 * that match the src/dst.  This is used during
6042 		 * IKE INITIAL-CONTACT.
6043 		 * XXXAE: this looks like some extension to RFC2367.
6044 		 */
6045 		ipseclog((LOG_DEBUG, "%s: doing delete all.\n", __func__));
6046 		return (key_delete_all(so, m, mhp, &saidx));
6047 	}
6048 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
6049 		ipseclog((LOG_DEBUG,
6050 		    "%s: invalid message: wrong header size.\n", __func__));
6051 		return (key_senderror(so, m, EINVAL));
6052 	}
6053 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6054 	SPI_ALLOC_LOCK();
6055 	if (proto == IPPROTO_TCP)
6056 		sav = key_getsav_tcpmd5(&saidx, NULL);
6057 	else
6058 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6059 	SPI_ALLOC_UNLOCK();
6060 	if (sav == NULL) {
6061 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u.\n",
6062 		    __func__, ntohl(sa0->sadb_sa_spi)));
6063 		return (key_senderror(so, m, ESRCH));
6064 	}
6065 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6066 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6067 		    __func__, ntohl(sav->spi)));
6068 		key_freesav(&sav);
6069 		return (key_senderror(so, m, ESRCH));
6070 	}
6071 	KEYDBG(KEY_STAMP,
6072 	    printf("%s: SA(%p)\n", __func__, sav));
6073 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
6074 	key_unlinksav(sav);
6075 	key_freesav(&sav);
6076 
6077     {
6078 	struct mbuf *n;
6079 	struct sadb_msg *newmsg;
6080 
6081 	/* create new sadb_msg to reply. */
6082 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
6083 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6084 	if (!n)
6085 		return key_senderror(so, m, ENOBUFS);
6086 
6087 	if (n->m_len < sizeof(struct sadb_msg)) {
6088 		n = m_pullup(n, sizeof(struct sadb_msg));
6089 		if (n == NULL)
6090 			return key_senderror(so, m, ENOBUFS);
6091 	}
6092 	newmsg = mtod(n, struct sadb_msg *);
6093 	newmsg->sadb_msg_errno = 0;
6094 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6095 
6096 	m_freem(m);
6097 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6098     }
6099 }
6100 
6101 /*
6102  * delete all SAs for src/dst.  Called from key_delete().
6103  */
6104 static int
6105 key_delete_all(struct socket *so, struct mbuf *m,
6106     const struct sadb_msghdr *mhp, struct secasindex *saidx)
6107 {
6108 	struct secasvar_queue drainq;
6109 	struct secashead *sah;
6110 	struct secasvar *sav, *nextsav;
6111 
6112 	TAILQ_INIT(&drainq);
6113 	SAHTREE_WLOCK();
6114 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
6115 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_HEAD) == 0)
6116 			continue;
6117 		/* Move all ALIVE SAs into drainq */
6118 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6119 	}
6120 	/* Unlink all queued SAs from SPI hash */
6121 	TAILQ_FOREACH(sav, &drainq, chain) {
6122 		sav->state = SADB_SASTATE_DEAD;
6123 		LIST_REMOVE(sav, spihash);
6124 	}
6125 	SAHTREE_WUNLOCK();
6126 	/* Now we can release reference for all SAs in drainq */
6127 	sav = TAILQ_FIRST(&drainq);
6128 	while (sav != NULL) {
6129 		KEYDBG(KEY_STAMP,
6130 		    printf("%s: SA(%p)\n", __func__, sav));
6131 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6132 		nextsav = TAILQ_NEXT(sav, chain);
6133 		key_freesah(&sav->sah); /* release reference from SAV */
6134 		key_freesav(&sav); /* release last reference */
6135 		sav = nextsav;
6136 	}
6137 
6138     {
6139 	struct mbuf *n;
6140 	struct sadb_msg *newmsg;
6141 
6142 	/* create new sadb_msg to reply. */
6143 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
6144 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6145 	if (!n)
6146 		return key_senderror(so, m, ENOBUFS);
6147 
6148 	if (n->m_len < sizeof(struct sadb_msg)) {
6149 		n = m_pullup(n, sizeof(struct sadb_msg));
6150 		if (n == NULL)
6151 			return key_senderror(so, m, ENOBUFS);
6152 	}
6153 	newmsg = mtod(n, struct sadb_msg *);
6154 	newmsg->sadb_msg_errno = 0;
6155 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6156 
6157 	m_freem(m);
6158 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6159     }
6160 }
6161 
6162 /*
6163  * Delete all alive SAs for corresponding xform.
6164  * Larval SAs have not initialized tdb_xform, so it is safe to leave them
6165  * here when xform disappears.
6166  */
6167 void
6168 key_delete_xform(const struct xformsw *xsp)
6169 {
6170 	struct secasvar_queue drainq;
6171 	struct secashead *sah;
6172 	struct secasvar *sav, *nextsav;
6173 
6174 	TAILQ_INIT(&drainq);
6175 	SAHTREE_WLOCK();
6176 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
6177 		sav = TAILQ_FIRST(&sah->savtree_alive);
6178 		if (sav == NULL)
6179 			continue;
6180 		if (sav->tdb_xform != xsp)
6181 			continue;
6182 		/*
6183 		 * It is supposed that all SAs in the chain are related to
6184 		 * one xform.
6185 		 */
6186 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6187 	}
6188 	/* Unlink all queued SAs from SPI hash */
6189 	TAILQ_FOREACH(sav, &drainq, chain) {
6190 		sav->state = SADB_SASTATE_DEAD;
6191 		LIST_REMOVE(sav, spihash);
6192 	}
6193 	SAHTREE_WUNLOCK();
6194 
6195 	/* Now we can release reference for all SAs in drainq */
6196 	sav = TAILQ_FIRST(&drainq);
6197 	while (sav != NULL) {
6198 		KEYDBG(KEY_STAMP,
6199 		    printf("%s: SA(%p)\n", __func__, sav));
6200 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6201 		nextsav = TAILQ_NEXT(sav, chain);
6202 		key_freesah(&sav->sah); /* release reference from SAV */
6203 		key_freesav(&sav); /* release last reference */
6204 		sav = nextsav;
6205 	}
6206 }
6207 
6208 /*
6209  * SADB_GET processing
6210  * receive
6211  *   <base, SA(*), address(SD)>
6212  * from the ikmpd, and get a SP and a SA to respond,
6213  * and send,
6214  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6215  *       (identity(SD),) (sensitivity)>
6216  * to the ikmpd.
6217  *
6218  * m will always be freed.
6219  */
6220 static int
6221 key_get(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6222 {
6223 	struct secasindex saidx;
6224 	struct sadb_address *src0, *dst0;
6225 	struct sadb_sa *sa0;
6226 	struct secasvar *sav;
6227 	uint8_t proto;
6228 
6229 	IPSEC_ASSERT(so != NULL, ("null socket"));
6230 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6231 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6232 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6233 
6234 	/* map satype to proto */
6235 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6236 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6237 			__func__));
6238 		return key_senderror(so, m, EINVAL);
6239 	}
6240 
6241 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
6242 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6243 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)) {
6244 		ipseclog((LOG_DEBUG,
6245 		    "%s: invalid message: missing required header.\n",
6246 		    __func__));
6247 		return key_senderror(so, m, EINVAL);
6248 	}
6249 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
6250 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6251 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6252 		ipseclog((LOG_DEBUG,
6253 		    "%s: invalid message: wrong header size.\n", __func__));
6254 		return key_senderror(so, m, EINVAL);
6255 	}
6256 
6257 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6258 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
6259 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
6260 
6261 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6262 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6263 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6264 		return key_senderror(so, m, EINVAL);
6265 	}
6266 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6267 
6268 	SPI_ALLOC_LOCK();
6269 	if (proto == IPPROTO_TCP)
6270 		sav = key_getsav_tcpmd5(&saidx, NULL);
6271 	else
6272 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6273 	SPI_ALLOC_UNLOCK();
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_NULL:	*min = 1; *max = 256; break;
6416 		case SADB_X_AALG_SHA2_256: *min = *max = 32; break;
6417 		case SADB_X_AALG_SHA2_384: *min = *max = 48; break;
6418 		case SADB_X_AALG_SHA2_512: *min = *max = 64; break;
6419 		default:
6420 			DPRINTF(("%s: unknown AH algorithm %u\n",
6421 				__func__, alg));
6422 			break;
6423 		}
6424 	}
6425 }
6426 
6427 /*
6428  * XXX reorder combinations by preference
6429  */
6430 static struct mbuf *
6431 key_getcomb_ah(void)
6432 {
6433 	const struct auth_hash *algo;
6434 	struct sadb_comb *comb;
6435 	struct mbuf *m;
6436 	u_int16_t minkeysize, maxkeysize;
6437 	int i;
6438 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6439 
6440 	m = NULL;
6441 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6442 #if 1
6443 		/* we prefer HMAC algorithms, not old algorithms */
6444 		if (i != SADB_AALG_SHA1HMAC &&
6445 		    i != SADB_X_AALG_SHA2_256 &&
6446 		    i != SADB_X_AALG_SHA2_384 &&
6447 		    i != SADB_X_AALG_SHA2_512)
6448 			continue;
6449 #endif
6450 		algo = auth_algorithm_lookup(i);
6451 		if (!algo)
6452 			continue;
6453 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6454 		/* discard algorithms with key size smaller than system min */
6455 		if (_BITS(minkeysize) < V_ipsec_ah_keymin)
6456 			continue;
6457 
6458 		if (!m) {
6459 			IPSEC_ASSERT(l <= MLEN,
6460 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6461 			MGET(m, M_NOWAIT, MT_DATA);
6462 			if (m) {
6463 				M_ALIGN(m, l);
6464 				m->m_len = l;
6465 				m->m_next = NULL;
6466 			}
6467 		} else
6468 			M_PREPEND(m, l, M_NOWAIT);
6469 		if (!m)
6470 			return NULL;
6471 
6472 		comb = mtod(m, struct sadb_comb *);
6473 		bzero(comb, sizeof(*comb));
6474 		key_getcomb_setlifetime(comb);
6475 		comb->sadb_comb_auth = i;
6476 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6477 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6478 	}
6479 
6480 	return m;
6481 }
6482 
6483 /*
6484  * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6485  * XXX reorder combinations by preference
6486  */
6487 static struct mbuf *
6488 key_getcomb_ipcomp(void)
6489 {
6490 	const struct comp_algo *algo;
6491 	struct sadb_comb *comb;
6492 	struct mbuf *m;
6493 	int i;
6494 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6495 
6496 	m = NULL;
6497 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6498 		algo = comp_algorithm_lookup(i);
6499 		if (!algo)
6500 			continue;
6501 
6502 		if (!m) {
6503 			IPSEC_ASSERT(l <= MLEN,
6504 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6505 			MGET(m, M_NOWAIT, MT_DATA);
6506 			if (m) {
6507 				M_ALIGN(m, l);
6508 				m->m_len = l;
6509 				m->m_next = NULL;
6510 			}
6511 		} else
6512 			M_PREPEND(m, l, M_NOWAIT);
6513 		if (!m)
6514 			return NULL;
6515 
6516 		comb = mtod(m, struct sadb_comb *);
6517 		bzero(comb, sizeof(*comb));
6518 		key_getcomb_setlifetime(comb);
6519 		comb->sadb_comb_encrypt = i;
6520 		/* what should we set into sadb_comb_*_{min,max}bits? */
6521 	}
6522 
6523 	return m;
6524 }
6525 
6526 /*
6527  * XXX no way to pass mode (transport/tunnel) to userland
6528  * XXX replay checking?
6529  * XXX sysctl interface to ipsec_{ah,esp}_keymin
6530  */
6531 static struct mbuf *
6532 key_getprop(const struct secasindex *saidx)
6533 {
6534 	struct sadb_prop *prop;
6535 	struct mbuf *m, *n;
6536 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6537 	int totlen;
6538 
6539 	switch (saidx->proto)  {
6540 	case IPPROTO_ESP:
6541 		m = key_getcomb_ealg();
6542 		break;
6543 	case IPPROTO_AH:
6544 		m = key_getcomb_ah();
6545 		break;
6546 	case IPPROTO_IPCOMP:
6547 		m = key_getcomb_ipcomp();
6548 		break;
6549 	default:
6550 		return NULL;
6551 	}
6552 
6553 	if (!m)
6554 		return NULL;
6555 	M_PREPEND(m, l, M_NOWAIT);
6556 	if (!m)
6557 		return NULL;
6558 
6559 	totlen = 0;
6560 	for (n = m; n; n = n->m_next)
6561 		totlen += n->m_len;
6562 
6563 	prop = mtod(m, struct sadb_prop *);
6564 	bzero(prop, sizeof(*prop));
6565 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6566 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6567 	prop->sadb_prop_replay = 32;	/* XXX */
6568 
6569 	return m;
6570 }
6571 
6572 /*
6573  * SADB_ACQUIRE processing called by key_checkrequest() and key_acquire2().
6574  * send
6575  *   <base, SA, address(SD), (address(P)), x_policy,
6576  *       (identity(SD),) (sensitivity,) proposal>
6577  * to KMD, and expect to receive
6578  *   <base> with SADB_ACQUIRE if error occurred,
6579  * or
6580  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6581  * from KMD by PF_KEY.
6582  *
6583  * XXX x_policy is outside of RFC2367 (KAME extension).
6584  * XXX sensitivity is not supported.
6585  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6586  * see comment for key_getcomb_ipcomp().
6587  *
6588  * OUT:
6589  *    0     : succeed
6590  *    others: error number
6591  */
6592 static int
6593 key_acquire(const struct secasindex *saidx, struct secpolicy *sp)
6594 {
6595 	union sockaddr_union addr;
6596 	struct mbuf *result, *m;
6597 	uint32_t seq;
6598 	int error;
6599 	uint16_t ul_proto;
6600 	uint8_t mask, satype;
6601 
6602 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
6603 	satype = key_proto2satype(saidx->proto);
6604 	IPSEC_ASSERT(satype != 0, ("null satype, protocol %u", saidx->proto));
6605 
6606 	error = -1;
6607 	result = NULL;
6608 	ul_proto = IPSEC_ULPROTO_ANY;
6609 
6610 	/* Get seq number to check whether sending message or not. */
6611 	seq = key_getacq(saidx, &error);
6612 	if (seq == 0)
6613 		return (error);
6614 
6615 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0);
6616 	if (!m) {
6617 		error = ENOBUFS;
6618 		goto fail;
6619 	}
6620 	result = m;
6621 
6622 	/*
6623 	 * set sadb_address for saidx's.
6624 	 *
6625 	 * Note that if sp is supplied, then we're being called from
6626 	 * key_allocsa_policy() and should supply port and protocol
6627 	 * information.
6628 	 * XXXAE: why only TCP and UDP? ICMP and SCTP looks applicable too.
6629 	 * XXXAE: probably we can handle this in the ipsec[46]_allocsa().
6630 	 * XXXAE: it looks like we should save this info in the ACQ entry.
6631 	 */
6632 	if (sp != NULL && (sp->spidx.ul_proto == IPPROTO_TCP ||
6633 	    sp->spidx.ul_proto == IPPROTO_UDP))
6634 		ul_proto = sp->spidx.ul_proto;
6635 
6636 	addr = saidx->src;
6637 	mask = FULLMASK;
6638 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6639 		switch (sp->spidx.src.sa.sa_family) {
6640 		case AF_INET:
6641 			if (sp->spidx.src.sin.sin_port != IPSEC_PORT_ANY) {
6642 				addr.sin.sin_port = sp->spidx.src.sin.sin_port;
6643 				mask = sp->spidx.prefs;
6644 			}
6645 			break;
6646 		case AF_INET6:
6647 			if (sp->spidx.src.sin6.sin6_port != IPSEC_PORT_ANY) {
6648 				addr.sin6.sin6_port =
6649 				    sp->spidx.src.sin6.sin6_port;
6650 				mask = sp->spidx.prefs;
6651 			}
6652 			break;
6653 		default:
6654 			break;
6655 		}
6656 	}
6657 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &addr.sa, mask, ul_proto);
6658 	if (!m) {
6659 		error = ENOBUFS;
6660 		goto fail;
6661 	}
6662 	m_cat(result, m);
6663 
6664 	addr = saidx->dst;
6665 	mask = FULLMASK;
6666 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6667 		switch (sp->spidx.dst.sa.sa_family) {
6668 		case AF_INET:
6669 			if (sp->spidx.dst.sin.sin_port != IPSEC_PORT_ANY) {
6670 				addr.sin.sin_port = sp->spidx.dst.sin.sin_port;
6671 				mask = sp->spidx.prefd;
6672 			}
6673 			break;
6674 		case AF_INET6:
6675 			if (sp->spidx.dst.sin6.sin6_port != IPSEC_PORT_ANY) {
6676 				addr.sin6.sin6_port =
6677 				    sp->spidx.dst.sin6.sin6_port;
6678 				mask = sp->spidx.prefd;
6679 			}
6680 			break;
6681 		default:
6682 			break;
6683 		}
6684 	}
6685 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &addr.sa, mask, ul_proto);
6686 	if (!m) {
6687 		error = ENOBUFS;
6688 		goto fail;
6689 	}
6690 	m_cat(result, m);
6691 
6692 	/* XXX proxy address (optional) */
6693 
6694 	/*
6695 	 * Set sadb_x_policy. This is KAME extension to RFC2367.
6696 	 */
6697 	if (sp != NULL) {
6698 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6699 		    sp->priority);
6700 		if (!m) {
6701 			error = ENOBUFS;
6702 			goto fail;
6703 		}
6704 		m_cat(result, m);
6705 	}
6706 
6707 	/*
6708 	 * Set sadb_x_sa2 extension if saidx->reqid is not zero.
6709 	 * This is FreeBSD extension to RFC2367.
6710 	 */
6711 	if (saidx->reqid != 0) {
6712 		m = key_setsadbxsa2(saidx->mode, 0, saidx->reqid);
6713 		if (m == NULL) {
6714 			error = ENOBUFS;
6715 			goto fail;
6716 		}
6717 		m_cat(result, m);
6718 	}
6719 	/* XXX identity (optional) */
6720 #if 0
6721 	if (idexttype && fqdn) {
6722 		/* create identity extension (FQDN) */
6723 		struct sadb_ident *id;
6724 		int fqdnlen;
6725 
6726 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
6727 		id = (struct sadb_ident *)p;
6728 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6729 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6730 		id->sadb_ident_exttype = idexttype;
6731 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
6732 		bcopy(fqdn, id + 1, fqdnlen);
6733 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
6734 	}
6735 
6736 	if (idexttype) {
6737 		/* create identity extension (USERFQDN) */
6738 		struct sadb_ident *id;
6739 		int userfqdnlen;
6740 
6741 		if (userfqdn) {
6742 			/* +1 for terminating-NUL */
6743 			userfqdnlen = strlen(userfqdn) + 1;
6744 		} else
6745 			userfqdnlen = 0;
6746 		id = (struct sadb_ident *)p;
6747 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6748 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6749 		id->sadb_ident_exttype = idexttype;
6750 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
6751 		/* XXX is it correct? */
6752 		if (curproc && curproc->p_cred)
6753 			id->sadb_ident_id = curproc->p_cred->p_ruid;
6754 		if (userfqdn && userfqdnlen)
6755 			bcopy(userfqdn, id + 1, userfqdnlen);
6756 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
6757 	}
6758 #endif
6759 
6760 	/* XXX sensitivity (optional) */
6761 
6762 	/* create proposal/combination extension */
6763 	m = key_getprop(saidx);
6764 #if 0
6765 	/*
6766 	 * spec conformant: always attach proposal/combination extension,
6767 	 * the problem is that we have no way to attach it for ipcomp,
6768 	 * due to the way sadb_comb is declared in RFC2367.
6769 	 */
6770 	if (!m) {
6771 		error = ENOBUFS;
6772 		goto fail;
6773 	}
6774 	m_cat(result, m);
6775 #else
6776 	/*
6777 	 * outside of spec; make proposal/combination extension optional.
6778 	 */
6779 	if (m)
6780 		m_cat(result, m);
6781 #endif
6782 
6783 	if ((result->m_flags & M_PKTHDR) == 0) {
6784 		error = EINVAL;
6785 		goto fail;
6786 	}
6787 
6788 	if (result->m_len < sizeof(struct sadb_msg)) {
6789 		result = m_pullup(result, sizeof(struct sadb_msg));
6790 		if (result == NULL) {
6791 			error = ENOBUFS;
6792 			goto fail;
6793 		}
6794 	}
6795 
6796 	result->m_pkthdr.len = 0;
6797 	for (m = result; m; m = m->m_next)
6798 		result->m_pkthdr.len += m->m_len;
6799 
6800 	mtod(result, struct sadb_msg *)->sadb_msg_len =
6801 	    PFKEY_UNIT64(result->m_pkthdr.len);
6802 
6803 	KEYDBG(KEY_STAMP,
6804 	    printf("%s: SP(%p)\n", __func__, sp));
6805 	KEYDBG(KEY_DATA, kdebug_secasindex(saidx, NULL));
6806 
6807 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
6808 
6809  fail:
6810 	if (result)
6811 		m_freem(result);
6812 	return error;
6813 }
6814 
6815 static uint32_t
6816 key_newacq(const struct secasindex *saidx, int *perror)
6817 {
6818 	struct secacq *acq;
6819 	uint32_t seq;
6820 
6821 	acq = malloc(sizeof(*acq), M_IPSEC_SAQ, M_NOWAIT | M_ZERO);
6822 	if (acq == NULL) {
6823 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6824 		*perror = ENOBUFS;
6825 		return (0);
6826 	}
6827 
6828 	/* copy secindex */
6829 	bcopy(saidx, &acq->saidx, sizeof(acq->saidx));
6830 	acq->created = time_second;
6831 	acq->count = 0;
6832 
6833 	/* add to acqtree */
6834 	ACQ_LOCK();
6835 	seq = acq->seq = (V_acq_seq == ~0 ? 1 : ++V_acq_seq);
6836 	LIST_INSERT_HEAD(&V_acqtree, acq, chain);
6837 	LIST_INSERT_HEAD(ACQADDRHASH_HASH(saidx), acq, addrhash);
6838 	LIST_INSERT_HEAD(ACQSEQHASH_HASH(seq), acq, seqhash);
6839 	ACQ_UNLOCK();
6840 	*perror = 0;
6841 	return (seq);
6842 }
6843 
6844 static uint32_t
6845 key_getacq(const struct secasindex *saidx, int *perror)
6846 {
6847 	struct secacq *acq;
6848 	uint32_t seq;
6849 
6850 	ACQ_LOCK();
6851 	LIST_FOREACH(acq, ACQADDRHASH_HASH(saidx), addrhash) {
6852 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY)) {
6853 			if (acq->count > V_key_blockacq_count) {
6854 				/*
6855 				 * Reset counter and send message.
6856 				 * Also reset created time to keep ACQ for
6857 				 * this saidx.
6858 				 */
6859 				acq->created = time_second;
6860 				acq->count = 0;
6861 				seq = acq->seq;
6862 			} else {
6863 				/*
6864 				 * Increment counter and do nothing.
6865 				 * We send SADB_ACQUIRE message only
6866 				 * for each V_key_blockacq_count packet.
6867 				 */
6868 				acq->count++;
6869 				seq = 0;
6870 			}
6871 			break;
6872 		}
6873 	}
6874 	ACQ_UNLOCK();
6875 	if (acq != NULL) {
6876 		*perror = 0;
6877 		return (seq);
6878 	}
6879 	/* allocate new  entry */
6880 	return (key_newacq(saidx, perror));
6881 }
6882 
6883 static int
6884 key_acqreset(uint32_t seq)
6885 {
6886 	struct secacq *acq;
6887 
6888 	ACQ_LOCK();
6889 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6890 		if (acq->seq == seq) {
6891 			acq->count = 0;
6892 			acq->created = time_second;
6893 			break;
6894 		}
6895 	}
6896 	ACQ_UNLOCK();
6897 	if (acq == NULL)
6898 		return (ESRCH);
6899 	return (0);
6900 }
6901 /*
6902  * Mark ACQ entry as stale to remove it in key_flush_acq().
6903  * Called after successful SADB_GETSPI message.
6904  */
6905 static int
6906 key_acqdone(const struct secasindex *saidx, uint32_t seq)
6907 {
6908 	struct secacq *acq;
6909 
6910 	ACQ_LOCK();
6911 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6912 		if (acq->seq == seq)
6913 			break;
6914 	}
6915 	if (acq != NULL) {
6916 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY) == 0) {
6917 			ipseclog((LOG_DEBUG,
6918 			    "%s: Mismatched saidx for ACQ %u\n", __func__, seq));
6919 			acq = NULL;
6920 		} else {
6921 			acq->created = 0;
6922 		}
6923 	} else {
6924 		ipseclog((LOG_DEBUG,
6925 		    "%s: ACQ %u is not found.\n", __func__, seq));
6926 	}
6927 	ACQ_UNLOCK();
6928 	if (acq == NULL)
6929 		return (ESRCH);
6930 	return (0);
6931 }
6932 
6933 static struct secspacq *
6934 key_newspacq(struct secpolicyindex *spidx)
6935 {
6936 	struct secspacq *acq;
6937 
6938 	/* get new entry */
6939 	acq = malloc(sizeof(struct secspacq), M_IPSEC_SAQ, M_NOWAIT|M_ZERO);
6940 	if (acq == NULL) {
6941 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6942 		return NULL;
6943 	}
6944 
6945 	/* copy secindex */
6946 	bcopy(spidx, &acq->spidx, sizeof(acq->spidx));
6947 	acq->created = time_second;
6948 	acq->count = 0;
6949 
6950 	/* add to spacqtree */
6951 	SPACQ_LOCK();
6952 	LIST_INSERT_HEAD(&V_spacqtree, acq, chain);
6953 	SPACQ_UNLOCK();
6954 
6955 	return acq;
6956 }
6957 
6958 static struct secspacq *
6959 key_getspacq(struct secpolicyindex *spidx)
6960 {
6961 	struct secspacq *acq;
6962 
6963 	SPACQ_LOCK();
6964 	LIST_FOREACH(acq, &V_spacqtree, chain) {
6965 		if (key_cmpspidx_exactly(spidx, &acq->spidx)) {
6966 			/* NB: return holding spacq_lock */
6967 			return acq;
6968 		}
6969 	}
6970 	SPACQ_UNLOCK();
6971 
6972 	return NULL;
6973 }
6974 
6975 /*
6976  * SADB_ACQUIRE processing,
6977  * in first situation, is receiving
6978  *   <base>
6979  * from the ikmpd, and clear sequence of its secasvar entry.
6980  *
6981  * In second situation, is receiving
6982  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6983  * from a user land process, and return
6984  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6985  * to the socket.
6986  *
6987  * m will always be freed.
6988  */
6989 static int
6990 key_acquire2(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6991 {
6992 	SAHTREE_RLOCK_TRACKER;
6993 	struct sadb_address *src0, *dst0;
6994 	struct secasindex saidx;
6995 	struct secashead *sah;
6996 	uint32_t reqid;
6997 	int error;
6998 	uint8_t mode, proto;
6999 
7000 	IPSEC_ASSERT(so != NULL, ("null socket"));
7001 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7002 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7003 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7004 
7005 	/*
7006 	 * Error message from KMd.
7007 	 * We assume that if error was occurred in IKEd, the length of PFKEY
7008 	 * message is equal to the size of sadb_msg structure.
7009 	 * We do not raise error even if error occurred in this function.
7010 	 */
7011 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
7012 		/* check sequence number */
7013 		if (mhp->msg->sadb_msg_seq == 0 ||
7014 		    mhp->msg->sadb_msg_errno == 0) {
7015 			ipseclog((LOG_DEBUG, "%s: must specify sequence "
7016 				"number and errno.\n", __func__));
7017 		} else {
7018 			/*
7019 			 * IKEd reported that error occurred.
7020 			 * XXXAE: what it expects from the kernel?
7021 			 * Probably we should send SADB_ACQUIRE again?
7022 			 * If so, reset ACQ's state.
7023 			 * XXXAE: it looks useless.
7024 			 */
7025 			key_acqreset(mhp->msg->sadb_msg_seq);
7026 		}
7027 		m_freem(m);
7028 		return (0);
7029 	}
7030 
7031 	/*
7032 	 * This message is from user land.
7033 	 */
7034 
7035 	/* map satype to proto */
7036 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7037 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7038 		    __func__));
7039 		return key_senderror(so, m, EINVAL);
7040 	}
7041 
7042 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
7043 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
7044 	    SADB_CHECKHDR(mhp, SADB_EXT_PROPOSAL)) {
7045 		ipseclog((LOG_DEBUG,
7046 		    "%s: invalid message: missing required header.\n",
7047 		    __func__));
7048 		return key_senderror(so, m, EINVAL);
7049 	}
7050 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
7051 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
7052 	    SADB_CHECKLEN(mhp, SADB_EXT_PROPOSAL)) {
7053 		ipseclog((LOG_DEBUG,
7054 		    "%s: invalid message: wrong header size.\n", __func__));
7055 		return key_senderror(so, m, EINVAL);
7056 	}
7057 
7058 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
7059 		mode = IPSEC_MODE_ANY;
7060 		reqid = 0;
7061 	} else {
7062 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
7063 			ipseclog((LOG_DEBUG,
7064 			    "%s: invalid message: wrong header size.\n",
7065 			    __func__));
7066 			return key_senderror(so, m, EINVAL);
7067 		}
7068 		mode = ((struct sadb_x_sa2 *)
7069 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
7070 		reqid = ((struct sadb_x_sa2 *)
7071 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
7072 	}
7073 
7074 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
7075 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
7076 
7077 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
7078 	    (struct sockaddr *)(dst0 + 1));
7079 	if (error != 0) {
7080 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
7081 		return key_senderror(so, m, EINVAL);
7082 	}
7083 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
7084 
7085 	/* get a SA index */
7086 	SAHTREE_RLOCK();
7087 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
7088 		if (key_cmpsaidx(&sah->saidx, &saidx, CMP_MODE_REQID))
7089 			break;
7090 	}
7091 	SAHTREE_RUNLOCK();
7092 	if (sah != NULL) {
7093 		ipseclog((LOG_DEBUG, "%s: a SA exists already.\n", __func__));
7094 		return key_senderror(so, m, EEXIST);
7095 	}
7096 
7097 	error = key_acquire(&saidx, NULL);
7098 	if (error != 0) {
7099 		ipseclog((LOG_DEBUG,
7100 		    "%s: error %d returned from key_acquire()\n",
7101 			__func__, error));
7102 		return key_senderror(so, m, error);
7103 	}
7104 	m_freem(m);
7105 	return (0);
7106 }
7107 
7108 /*
7109  * SADB_REGISTER processing.
7110  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
7111  * receive
7112  *   <base>
7113  * from the ikmpd, and register a socket to send PF_KEY messages,
7114  * and send
7115  *   <base, supported>
7116  * to KMD by PF_KEY.
7117  * If socket is detached, must free from regnode.
7118  *
7119  * m will always be freed.
7120  */
7121 static int
7122 key_register(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7123 {
7124 	struct secreg *reg, *newreg = NULL;
7125 
7126 	IPSEC_ASSERT(so != NULL, ("null socket"));
7127 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7128 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7129 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7130 
7131 	/* check for invalid register message */
7132 	if (mhp->msg->sadb_msg_satype >= sizeof(V_regtree)/sizeof(V_regtree[0]))
7133 		return key_senderror(so, m, EINVAL);
7134 
7135 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
7136 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
7137 		goto setmsg;
7138 
7139 	/* check whether existing or not */
7140 	REGTREE_LOCK();
7141 	LIST_FOREACH(reg, &V_regtree[mhp->msg->sadb_msg_satype], chain) {
7142 		if (reg->so == so) {
7143 			REGTREE_UNLOCK();
7144 			ipseclog((LOG_DEBUG, "%s: socket exists already.\n",
7145 				__func__));
7146 			return key_senderror(so, m, EEXIST);
7147 		}
7148 	}
7149 
7150 	/* create regnode */
7151 	newreg =  malloc(sizeof(struct secreg), M_IPSEC_SAR, M_NOWAIT|M_ZERO);
7152 	if (newreg == NULL) {
7153 		REGTREE_UNLOCK();
7154 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7155 		return key_senderror(so, m, ENOBUFS);
7156 	}
7157 
7158 	newreg->so = so;
7159 	((struct keycb *)(so->so_pcb))->kp_registered++;
7160 
7161 	/* add regnode to regtree. */
7162 	LIST_INSERT_HEAD(&V_regtree[mhp->msg->sadb_msg_satype], newreg, chain);
7163 	REGTREE_UNLOCK();
7164 
7165   setmsg:
7166     {
7167 	struct mbuf *n;
7168 	struct sadb_msg *newmsg;
7169 	struct sadb_supported *sup;
7170 	u_int len, alen, elen;
7171 	int off;
7172 	int i;
7173 	struct sadb_alg *alg;
7174 
7175 	/* create new sadb_msg to reply. */
7176 	alen = 0;
7177 	for (i = 1; i <= SADB_AALG_MAX; i++) {
7178 		if (auth_algorithm_lookup(i))
7179 			alen += sizeof(struct sadb_alg);
7180 	}
7181 	if (alen)
7182 		alen += sizeof(struct sadb_supported);
7183 	elen = 0;
7184 	for (i = 1; i <= SADB_EALG_MAX; i++) {
7185 		if (enc_algorithm_lookup(i))
7186 			elen += sizeof(struct sadb_alg);
7187 	}
7188 	if (elen)
7189 		elen += sizeof(struct sadb_supported);
7190 
7191 	len = sizeof(struct sadb_msg) + alen + elen;
7192 
7193 	if (len > MCLBYTES)
7194 		return key_senderror(so, m, ENOBUFS);
7195 
7196 	MGETHDR(n, M_NOWAIT, MT_DATA);
7197 	if (n != NULL && len > MHLEN) {
7198 		if (!(MCLGET(n, M_NOWAIT))) {
7199 			m_freem(n);
7200 			n = NULL;
7201 		}
7202 	}
7203 	if (!n)
7204 		return key_senderror(so, m, ENOBUFS);
7205 
7206 	n->m_pkthdr.len = n->m_len = len;
7207 	n->m_next = NULL;
7208 	off = 0;
7209 
7210 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
7211 	newmsg = mtod(n, struct sadb_msg *);
7212 	newmsg->sadb_msg_errno = 0;
7213 	newmsg->sadb_msg_len = PFKEY_UNIT64(len);
7214 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7215 
7216 	/* for authentication algorithm */
7217 	if (alen) {
7218 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7219 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7220 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7221 		off += PFKEY_ALIGN8(sizeof(*sup));
7222 
7223 		for (i = 1; i <= SADB_AALG_MAX; i++) {
7224 			const struct auth_hash *aalgo;
7225 			u_int16_t minkeysize, maxkeysize;
7226 
7227 			aalgo = auth_algorithm_lookup(i);
7228 			if (!aalgo)
7229 				continue;
7230 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7231 			alg->sadb_alg_id = i;
7232 			alg->sadb_alg_ivlen = 0;
7233 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7234 			alg->sadb_alg_minbits = _BITS(minkeysize);
7235 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7236 			off += PFKEY_ALIGN8(sizeof(*alg));
7237 		}
7238 	}
7239 
7240 	/* for encryption algorithm */
7241 	if (elen) {
7242 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7243 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7244 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7245 		off += PFKEY_ALIGN8(sizeof(*sup));
7246 
7247 		for (i = 1; i <= SADB_EALG_MAX; i++) {
7248 			const struct enc_xform *ealgo;
7249 
7250 			ealgo = enc_algorithm_lookup(i);
7251 			if (!ealgo)
7252 				continue;
7253 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7254 			alg->sadb_alg_id = i;
7255 			alg->sadb_alg_ivlen = ealgo->ivsize;
7256 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7257 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7258 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7259 		}
7260 	}
7261 
7262 	IPSEC_ASSERT(off == len,
7263 		("length assumption failed (off %u len %u)", off, len));
7264 
7265 	m_freem(m);
7266 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7267     }
7268 }
7269 
7270 /*
7271  * free secreg entry registered.
7272  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7273  */
7274 void
7275 key_freereg(struct socket *so)
7276 {
7277 	struct secreg *reg;
7278 	int i;
7279 
7280 	IPSEC_ASSERT(so != NULL, ("NULL so"));
7281 
7282 	/*
7283 	 * check whether existing or not.
7284 	 * check all type of SA, because there is a potential that
7285 	 * one socket is registered to multiple type of SA.
7286 	 */
7287 	REGTREE_LOCK();
7288 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7289 		LIST_FOREACH(reg, &V_regtree[i], chain) {
7290 			if (reg->so == so && __LIST_CHAINED(reg)) {
7291 				LIST_REMOVE(reg, chain);
7292 				free(reg, M_IPSEC_SAR);
7293 				break;
7294 			}
7295 		}
7296 	}
7297 	REGTREE_UNLOCK();
7298 }
7299 
7300 /*
7301  * SADB_EXPIRE processing
7302  * send
7303  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7304  * to KMD by PF_KEY.
7305  * NOTE: We send only soft lifetime extension.
7306  *
7307  * OUT:	0	: succeed
7308  *	others	: error number
7309  */
7310 static int
7311 key_expire(struct secasvar *sav, int hard)
7312 {
7313 	struct mbuf *result = NULL, *m;
7314 	struct sadb_lifetime *lt;
7315 	uint32_t replay_count;
7316 	int error, len;
7317 	uint8_t satype;
7318 
7319 	SECASVAR_RLOCK_TRACKER;
7320 
7321 	IPSEC_ASSERT (sav != NULL, ("null sav"));
7322 	IPSEC_ASSERT (sav->sah != NULL, ("null sa header"));
7323 
7324 	KEYDBG(KEY_STAMP,
7325 	    printf("%s: SA(%p) expired %s lifetime\n", __func__,
7326 		sav, hard ? "hard": "soft"));
7327 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
7328 	/* set msg header */
7329 	satype = key_proto2satype(sav->sah->saidx.proto);
7330 	IPSEC_ASSERT(satype != 0, ("invalid proto, satype %u", satype));
7331 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, sav->refcnt);
7332 	if (!m) {
7333 		error = ENOBUFS;
7334 		goto fail;
7335 	}
7336 	result = m;
7337 
7338 	/* create SA extension */
7339 	m = key_setsadbsa(sav);
7340 	if (!m) {
7341 		error = ENOBUFS;
7342 		goto fail;
7343 	}
7344 	m_cat(result, m);
7345 
7346 	/* create SA extension */
7347 	SECASVAR_RLOCK(sav);
7348 	replay_count = sav->replay ? sav->replay->count : 0;
7349 	SECASVAR_RUNLOCK(sav);
7350 
7351 	m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
7352 			sav->sah->saidx.reqid);
7353 	if (!m) {
7354 		error = ENOBUFS;
7355 		goto fail;
7356 	}
7357 	m_cat(result, m);
7358 
7359 	if (sav->replay && sav->replay->wsize > UINT8_MAX) {
7360 		m = key_setsadbxsareplay(sav->replay->wsize);
7361 		if (!m) {
7362 			error = ENOBUFS;
7363 			goto fail;
7364 		}
7365 		m_cat(result, m);
7366 	}
7367 
7368 	/* create lifetime extension (current and soft) */
7369 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7370 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
7371 	if (m == NULL) {
7372 		error = ENOBUFS;
7373 		goto fail;
7374 	}
7375 	m_align(m, len);
7376 	m->m_len = len;
7377 	bzero(mtod(m, caddr_t), len);
7378 	lt = mtod(m, struct sadb_lifetime *);
7379 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7380 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7381 	lt->sadb_lifetime_allocations =
7382 	    (uint32_t)counter_u64_fetch(sav->lft_c_allocations);
7383 	lt->sadb_lifetime_bytes =
7384 	    counter_u64_fetch(sav->lft_c_bytes);
7385 	lt->sadb_lifetime_addtime = sav->created;
7386 	lt->sadb_lifetime_usetime = sav->firstused;
7387 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
7388 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7389 	if (hard) {
7390 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
7391 		lt->sadb_lifetime_allocations = sav->lft_h->allocations;
7392 		lt->sadb_lifetime_bytes = sav->lft_h->bytes;
7393 		lt->sadb_lifetime_addtime = sav->lft_h->addtime;
7394 		lt->sadb_lifetime_usetime = sav->lft_h->usetime;
7395 	} else {
7396 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_SOFT;
7397 		lt->sadb_lifetime_allocations = sav->lft_s->allocations;
7398 		lt->sadb_lifetime_bytes = sav->lft_s->bytes;
7399 		lt->sadb_lifetime_addtime = sav->lft_s->addtime;
7400 		lt->sadb_lifetime_usetime = sav->lft_s->usetime;
7401 	}
7402 	m_cat(result, m);
7403 
7404 	/* set sadb_address for source */
7405 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
7406 	    &sav->sah->saidx.src.sa,
7407 	    FULLMASK, IPSEC_ULPROTO_ANY);
7408 	if (!m) {
7409 		error = ENOBUFS;
7410 		goto fail;
7411 	}
7412 	m_cat(result, m);
7413 
7414 	/* set sadb_address for destination */
7415 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
7416 	    &sav->sah->saidx.dst.sa,
7417 	    FULLMASK, IPSEC_ULPROTO_ANY);
7418 	if (!m) {
7419 		error = ENOBUFS;
7420 		goto fail;
7421 	}
7422 	m_cat(result, m);
7423 
7424 	/*
7425 	 * XXX-BZ Handle NAT-T extensions here.
7426 	 * XXXAE: it doesn't seem quite useful. IKEs should not depend on
7427 	 * this information, we report only significant SA fields.
7428 	 */
7429 
7430 	if ((result->m_flags & M_PKTHDR) == 0) {
7431 		error = EINVAL;
7432 		goto fail;
7433 	}
7434 
7435 	if (result->m_len < sizeof(struct sadb_msg)) {
7436 		result = m_pullup(result, sizeof(struct sadb_msg));
7437 		if (result == NULL) {
7438 			error = ENOBUFS;
7439 			goto fail;
7440 		}
7441 	}
7442 
7443 	result->m_pkthdr.len = 0;
7444 	for (m = result; m; m = m->m_next)
7445 		result->m_pkthdr.len += m->m_len;
7446 
7447 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7448 	    PFKEY_UNIT64(result->m_pkthdr.len);
7449 
7450 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7451 
7452  fail:
7453 	if (result)
7454 		m_freem(result);
7455 	return error;
7456 }
7457 
7458 static void
7459 key_freesah_flushed(struct secashead_queue *flushq)
7460 {
7461 	struct secashead *sah, *nextsah;
7462 	struct secasvar *sav, *nextsav;
7463 
7464 	sah = TAILQ_FIRST(flushq);
7465 	while (sah != NULL) {
7466 		sav = TAILQ_FIRST(&sah->savtree_larval);
7467 		while (sav != NULL) {
7468 			nextsav = TAILQ_NEXT(sav, chain);
7469 			TAILQ_REMOVE(&sah->savtree_larval, sav, chain);
7470 			key_freesav(&sav); /* release last reference */
7471 			key_freesah(&sah); /* release reference from SAV */
7472 			sav = nextsav;
7473 		}
7474 		sav = TAILQ_FIRST(&sah->savtree_alive);
7475 		while (sav != NULL) {
7476 			nextsav = TAILQ_NEXT(sav, chain);
7477 			TAILQ_REMOVE(&sah->savtree_alive, sav, chain);
7478 			key_freesav(&sav); /* release last reference */
7479 			key_freesah(&sah); /* release reference from SAV */
7480 			sav = nextsav;
7481 		}
7482 		nextsah = TAILQ_NEXT(sah, chain);
7483 		key_freesah(&sah);	/* release last reference */
7484 		sah = nextsah;
7485 	}
7486 }
7487 
7488 /*
7489  * SADB_FLUSH processing
7490  * receive
7491  *   <base>
7492  * from the ikmpd, and free all entries in secastree.
7493  * and send,
7494  *   <base>
7495  * to the ikmpd.
7496  * NOTE: to do is only marking SADB_SASTATE_DEAD.
7497  *
7498  * m will always be freed.
7499  */
7500 static int
7501 key_flush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7502 {
7503 	struct secashead_queue flushq;
7504 	struct sadb_msg *newmsg;
7505 	struct secashead *sah, *nextsah;
7506 	struct secasvar *sav;
7507 	uint8_t proto;
7508 	int i;
7509 
7510 	IPSEC_ASSERT(so != NULL, ("null socket"));
7511 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7512 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7513 
7514 	/* map satype to proto */
7515 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7516 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7517 			__func__));
7518 		return key_senderror(so, m, EINVAL);
7519 	}
7520 	KEYDBG(KEY_STAMP,
7521 	    printf("%s: proto %u\n", __func__, proto));
7522 
7523 	TAILQ_INIT(&flushq);
7524 	if (proto == IPSEC_PROTO_ANY) {
7525 		/* no SATYPE specified, i.e. flushing all SA. */
7526 		SAHTREE_WLOCK();
7527 		/* Move all SAHs into flushq */
7528 		TAILQ_CONCAT(&flushq, &V_sahtree, chain);
7529 		/* Flush all buckets in SPI hash */
7530 		for (i = 0; i < V_savhash_mask + 1; i++)
7531 			LIST_INIT(&V_savhashtbl[i]);
7532 		/* Flush all buckets in SAHADDRHASH */
7533 		for (i = 0; i < V_sahaddrhash_mask + 1; i++)
7534 			LIST_INIT(&V_sahaddrhashtbl[i]);
7535 		/* Mark all SAHs as unlinked */
7536 		TAILQ_FOREACH(sah, &flushq, chain) {
7537 			sah->state = SADB_SASTATE_DEAD;
7538 			/*
7539 			 * Callout handler makes its job using
7540 			 * RLOCK and drain queues. In case, when this
7541 			 * function will be called just before it
7542 			 * acquires WLOCK, we need to mark SAs as
7543 			 * unlinked to prevent second unlink.
7544 			 */
7545 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7546 				sav->state = SADB_SASTATE_DEAD;
7547 			}
7548 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7549 				sav->state = SADB_SASTATE_DEAD;
7550 			}
7551 		}
7552 		SAHTREE_WUNLOCK();
7553 	} else {
7554 		SAHTREE_WLOCK();
7555 		sah = TAILQ_FIRST(&V_sahtree);
7556 		while (sah != NULL) {
7557 			IPSEC_ASSERT(sah->state != SADB_SASTATE_DEAD,
7558 			    ("DEAD SAH %p in SADB_FLUSH", sah));
7559 			nextsah = TAILQ_NEXT(sah, chain);
7560 			if (sah->saidx.proto != proto) {
7561 				sah = nextsah;
7562 				continue;
7563 			}
7564 			sah->state = SADB_SASTATE_DEAD;
7565 			TAILQ_REMOVE(&V_sahtree, sah, chain);
7566 			LIST_REMOVE(sah, addrhash);
7567 			/* Unlink all SAs from SPI hash */
7568 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7569 				LIST_REMOVE(sav, spihash);
7570 				sav->state = SADB_SASTATE_DEAD;
7571 			}
7572 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7573 				LIST_REMOVE(sav, spihash);
7574 				sav->state = SADB_SASTATE_DEAD;
7575 			}
7576 			/* Add SAH into flushq */
7577 			TAILQ_INSERT_HEAD(&flushq, sah, chain);
7578 			sah = nextsah;
7579 		}
7580 		SAHTREE_WUNLOCK();
7581 	}
7582 
7583 	key_freesah_flushed(&flushq);
7584 	/* Free all queued SAs and SAHs */
7585 	if (m->m_len < sizeof(struct sadb_msg) ||
7586 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7587 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7588 		return key_senderror(so, m, ENOBUFS);
7589 	}
7590 
7591 	if (m->m_next)
7592 		m_freem(m->m_next);
7593 	m->m_next = NULL;
7594 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7595 	newmsg = mtod(m, struct sadb_msg *);
7596 	newmsg->sadb_msg_errno = 0;
7597 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7598 
7599 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7600 }
7601 
7602 /*
7603  * SADB_DUMP processing
7604  * dump all entries including status of DEAD in SAD.
7605  * receive
7606  *   <base>
7607  * from the ikmpd, and dump all secasvar leaves
7608  * and send,
7609  *   <base> .....
7610  * to the ikmpd.
7611  *
7612  * m will always be freed.
7613  */
7614 static int
7615 key_dump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7616 {
7617 	SAHTREE_RLOCK_TRACKER;
7618 	struct secashead *sah;
7619 	struct secasvar *sav;
7620 	struct mbuf *n;
7621 	uint32_t cnt;
7622 	uint8_t proto, satype;
7623 
7624 	IPSEC_ASSERT(so != NULL, ("null socket"));
7625 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7626 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7627 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7628 
7629 	/* map satype to proto */
7630 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7631 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7632 		    __func__));
7633 		return key_senderror(so, m, EINVAL);
7634 	}
7635 
7636 	/* count sav entries to be sent to the userland. */
7637 	cnt = 0;
7638 	SAHTREE_RLOCK();
7639 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7640 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7641 		    proto != sah->saidx.proto)
7642 			continue;
7643 
7644 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain)
7645 			cnt++;
7646 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain)
7647 			cnt++;
7648 	}
7649 
7650 	if (cnt == 0) {
7651 		SAHTREE_RUNLOCK();
7652 		return key_senderror(so, m, ENOENT);
7653 	}
7654 
7655 	/* send this to the userland, one at a time. */
7656 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7657 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7658 		    proto != sah->saidx.proto)
7659 			continue;
7660 
7661 		/* map proto to satype */
7662 		if ((satype = key_proto2satype(sah->saidx.proto)) == 0) {
7663 			SAHTREE_RUNLOCK();
7664 			ipseclog((LOG_DEBUG, "%s: there was invalid proto in "
7665 			    "SAD.\n", __func__));
7666 			return key_senderror(so, m, EINVAL);
7667 		}
7668 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7669 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7670 			    --cnt, mhp->msg->sadb_msg_pid);
7671 			if (n == NULL) {
7672 				SAHTREE_RUNLOCK();
7673 				return key_senderror(so, m, ENOBUFS);
7674 			}
7675 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7676 		}
7677 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7678 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7679 			    --cnt, mhp->msg->sadb_msg_pid);
7680 			if (n == NULL) {
7681 				SAHTREE_RUNLOCK();
7682 				return key_senderror(so, m, ENOBUFS);
7683 			}
7684 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7685 		}
7686 	}
7687 	SAHTREE_RUNLOCK();
7688 	m_freem(m);
7689 	return (0);
7690 }
7691 /*
7692  * SADB_X_PROMISC processing
7693  *
7694  * m will always be freed.
7695  */
7696 static int
7697 key_promisc(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7698 {
7699 	int olen;
7700 
7701 	IPSEC_ASSERT(so != NULL, ("null socket"));
7702 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7703 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7704 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7705 
7706 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7707 
7708 	if (olen < sizeof(struct sadb_msg)) {
7709 #if 1
7710 		return key_senderror(so, m, EINVAL);
7711 #else
7712 		m_freem(m);
7713 		return 0;
7714 #endif
7715 	} else if (olen == sizeof(struct sadb_msg)) {
7716 		/* enable/disable promisc mode */
7717 		struct keycb *kp;
7718 
7719 		if ((kp = so->so_pcb) == NULL)
7720 			return key_senderror(so, m, EINVAL);
7721 		mhp->msg->sadb_msg_errno = 0;
7722 		switch (mhp->msg->sadb_msg_satype) {
7723 		case 0:
7724 		case 1:
7725 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
7726 			break;
7727 		default:
7728 			return key_senderror(so, m, EINVAL);
7729 		}
7730 
7731 		/* send the original message back to everyone */
7732 		mhp->msg->sadb_msg_errno = 0;
7733 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7734 	} else {
7735 		/* send packet as is */
7736 
7737 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
7738 
7739 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
7740 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7741 	}
7742 }
7743 
7744 static int (*key_typesw[])(struct socket *, struct mbuf *,
7745 		const struct sadb_msghdr *) = {
7746 	NULL,		/* SADB_RESERVED */
7747 	key_getspi,	/* SADB_GETSPI */
7748 	key_update,	/* SADB_UPDATE */
7749 	key_add,	/* SADB_ADD */
7750 	key_delete,	/* SADB_DELETE */
7751 	key_get,	/* SADB_GET */
7752 	key_acquire2,	/* SADB_ACQUIRE */
7753 	key_register,	/* SADB_REGISTER */
7754 	NULL,		/* SADB_EXPIRE */
7755 	key_flush,	/* SADB_FLUSH */
7756 	key_dump,	/* SADB_DUMP */
7757 	key_promisc,	/* SADB_X_PROMISC */
7758 	NULL,		/* SADB_X_PCHANGE */
7759 	key_spdadd,	/* SADB_X_SPDUPDATE */
7760 	key_spdadd,	/* SADB_X_SPDADD */
7761 	key_spddelete,	/* SADB_X_SPDDELETE */
7762 	key_spdget,	/* SADB_X_SPDGET */
7763 	NULL,		/* SADB_X_SPDACQUIRE */
7764 	key_spddump,	/* SADB_X_SPDDUMP */
7765 	key_spdflush,	/* SADB_X_SPDFLUSH */
7766 	key_spdadd,	/* SADB_X_SPDSETIDX */
7767 	NULL,		/* SADB_X_SPDEXPIRE */
7768 	key_spddelete2,	/* SADB_X_SPDDELETE2 */
7769 };
7770 
7771 /*
7772  * parse sadb_msg buffer to process PFKEYv2,
7773  * and create a data to response if needed.
7774  * I think to be dealed with mbuf directly.
7775  * IN:
7776  *     msgp  : pointer to pointer to a received buffer pulluped.
7777  *             This is rewrited to response.
7778  *     so    : pointer to socket.
7779  * OUT:
7780  *    length for buffer to send to user process.
7781  */
7782 int
7783 key_parse(struct mbuf *m, struct socket *so)
7784 {
7785 	struct sadb_msg *msg;
7786 	struct sadb_msghdr mh;
7787 	u_int orglen;
7788 	int error;
7789 	int target;
7790 
7791 	IPSEC_ASSERT(so != NULL, ("null socket"));
7792 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7793 
7794 	if (m->m_len < sizeof(struct sadb_msg)) {
7795 		m = m_pullup(m, sizeof(struct sadb_msg));
7796 		if (!m)
7797 			return ENOBUFS;
7798 	}
7799 	msg = mtod(m, struct sadb_msg *);
7800 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
7801 	target = KEY_SENDUP_ONE;
7802 
7803 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len != orglen) {
7804 		ipseclog((LOG_DEBUG, "%s: invalid message length.\n",__func__));
7805 		PFKEYSTAT_INC(out_invlen);
7806 		error = EINVAL;
7807 		goto senderror;
7808 	}
7809 
7810 	if (msg->sadb_msg_version != PF_KEY_V2) {
7811 		ipseclog((LOG_DEBUG, "%s: PF_KEY version %u is mismatched.\n",
7812 		    __func__, msg->sadb_msg_version));
7813 		PFKEYSTAT_INC(out_invver);
7814 		error = EINVAL;
7815 		goto senderror;
7816 	}
7817 
7818 	if (msg->sadb_msg_type > SADB_MAX) {
7819 		ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7820 		    __func__, msg->sadb_msg_type));
7821 		PFKEYSTAT_INC(out_invmsgtype);
7822 		error = EINVAL;
7823 		goto senderror;
7824 	}
7825 
7826 	/* for old-fashioned code - should be nuked */
7827 	if (m->m_pkthdr.len > MCLBYTES) {
7828 		m_freem(m);
7829 		return ENOBUFS;
7830 	}
7831 	if (m->m_next) {
7832 		struct mbuf *n;
7833 
7834 		MGETHDR(n, M_NOWAIT, MT_DATA);
7835 		if (n && m->m_pkthdr.len > MHLEN) {
7836 			if (!(MCLGET(n, M_NOWAIT))) {
7837 				m_free(n);
7838 				n = NULL;
7839 			}
7840 		}
7841 		if (!n) {
7842 			m_freem(m);
7843 			return ENOBUFS;
7844 		}
7845 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
7846 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
7847 		n->m_next = NULL;
7848 		m_freem(m);
7849 		m = n;
7850 	}
7851 
7852 	/* align the mbuf chain so that extensions are in contiguous region. */
7853 	error = key_align(m, &mh);
7854 	if (error)
7855 		return error;
7856 
7857 	msg = mh.msg;
7858 
7859 	/* We use satype as scope mask for spddump */
7860 	if (msg->sadb_msg_type == SADB_X_SPDDUMP) {
7861 		switch (msg->sadb_msg_satype) {
7862 		case IPSEC_POLICYSCOPE_ANY:
7863 		case IPSEC_POLICYSCOPE_GLOBAL:
7864 		case IPSEC_POLICYSCOPE_IFNET:
7865 		case IPSEC_POLICYSCOPE_PCB:
7866 			break;
7867 		default:
7868 			ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7869 			    __func__, msg->sadb_msg_type));
7870 			PFKEYSTAT_INC(out_invsatype);
7871 			error = EINVAL;
7872 			goto senderror;
7873 		}
7874 	} else {
7875 		switch (msg->sadb_msg_satype) { /* check SA type */
7876 		case SADB_SATYPE_UNSPEC:
7877 			switch (msg->sadb_msg_type) {
7878 			case SADB_GETSPI:
7879 			case SADB_UPDATE:
7880 			case SADB_ADD:
7881 			case SADB_DELETE:
7882 			case SADB_GET:
7883 			case SADB_ACQUIRE:
7884 			case SADB_EXPIRE:
7885 				ipseclog((LOG_DEBUG, "%s: must specify satype "
7886 				    "when msg type=%u.\n", __func__,
7887 				    msg->sadb_msg_type));
7888 				PFKEYSTAT_INC(out_invsatype);
7889 				error = EINVAL;
7890 				goto senderror;
7891 			}
7892 			break;
7893 		case SADB_SATYPE_AH:
7894 		case SADB_SATYPE_ESP:
7895 		case SADB_X_SATYPE_IPCOMP:
7896 		case SADB_X_SATYPE_TCPSIGNATURE:
7897 			switch (msg->sadb_msg_type) {
7898 			case SADB_X_SPDADD:
7899 			case SADB_X_SPDDELETE:
7900 			case SADB_X_SPDGET:
7901 			case SADB_X_SPDFLUSH:
7902 			case SADB_X_SPDSETIDX:
7903 			case SADB_X_SPDUPDATE:
7904 			case SADB_X_SPDDELETE2:
7905 				ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7906 				    __func__, msg->sadb_msg_type));
7907 				PFKEYSTAT_INC(out_invsatype);
7908 				error = EINVAL;
7909 				goto senderror;
7910 			}
7911 			break;
7912 		case SADB_SATYPE_RSVP:
7913 		case SADB_SATYPE_OSPFV2:
7914 		case SADB_SATYPE_RIPV2:
7915 		case SADB_SATYPE_MIP:
7916 			ipseclog((LOG_DEBUG, "%s: type %u isn't supported.\n",
7917 			    __func__, msg->sadb_msg_satype));
7918 			PFKEYSTAT_INC(out_invsatype);
7919 			error = EOPNOTSUPP;
7920 			goto senderror;
7921 		case 1:	/* XXX: What does it do? */
7922 			if (msg->sadb_msg_type == SADB_X_PROMISC)
7923 				break;
7924 			/*FALLTHROUGH*/
7925 		default:
7926 			ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7927 			    __func__, msg->sadb_msg_satype));
7928 			PFKEYSTAT_INC(out_invsatype);
7929 			error = EINVAL;
7930 			goto senderror;
7931 		}
7932 	}
7933 
7934 	/* check field of upper layer protocol and address family */
7935 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL
7936 	 && mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
7937 		struct sadb_address *src0, *dst0;
7938 		u_int plen;
7939 
7940 		src0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_SRC]);
7941 		dst0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_DST]);
7942 
7943 		/* check upper layer protocol */
7944 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
7945 			ipseclog((LOG_DEBUG, "%s: upper layer protocol "
7946 				"mismatched.\n", __func__));
7947 			PFKEYSTAT_INC(out_invaddr);
7948 			error = EINVAL;
7949 			goto senderror;
7950 		}
7951 
7952 		/* check family */
7953 		if (PFKEY_ADDR_SADDR(src0)->sa_family !=
7954 		    PFKEY_ADDR_SADDR(dst0)->sa_family) {
7955 			ipseclog((LOG_DEBUG, "%s: address family mismatched.\n",
7956 				__func__));
7957 			PFKEYSTAT_INC(out_invaddr);
7958 			error = EINVAL;
7959 			goto senderror;
7960 		}
7961 		if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7962 		    PFKEY_ADDR_SADDR(dst0)->sa_len) {
7963 			ipseclog((LOG_DEBUG, "%s: address struct size "
7964 				"mismatched.\n", __func__));
7965 			PFKEYSTAT_INC(out_invaddr);
7966 			error = EINVAL;
7967 			goto senderror;
7968 		}
7969 
7970 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7971 		case AF_INET:
7972 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7973 			    sizeof(struct sockaddr_in)) {
7974 				PFKEYSTAT_INC(out_invaddr);
7975 				error = EINVAL;
7976 				goto senderror;
7977 			}
7978 			break;
7979 		case AF_INET6:
7980 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7981 			    sizeof(struct sockaddr_in6)) {
7982 				PFKEYSTAT_INC(out_invaddr);
7983 				error = EINVAL;
7984 				goto senderror;
7985 			}
7986 			break;
7987 		default:
7988 			ipseclog((LOG_DEBUG, "%s: unsupported address family\n",
7989 				__func__));
7990 			PFKEYSTAT_INC(out_invaddr);
7991 			error = EAFNOSUPPORT;
7992 			goto senderror;
7993 		}
7994 
7995 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7996 		case AF_INET:
7997 			plen = sizeof(struct in_addr) << 3;
7998 			break;
7999 		case AF_INET6:
8000 			plen = sizeof(struct in6_addr) << 3;
8001 			break;
8002 		default:
8003 			plen = 0;	/*fool gcc*/
8004 			break;
8005 		}
8006 
8007 		/* check max prefix length */
8008 		if (src0->sadb_address_prefixlen > plen ||
8009 		    dst0->sadb_address_prefixlen > plen) {
8010 			ipseclog((LOG_DEBUG, "%s: illegal prefixlen.\n",
8011 				__func__));
8012 			PFKEYSTAT_INC(out_invaddr);
8013 			error = EINVAL;
8014 			goto senderror;
8015 		}
8016 
8017 		/*
8018 		 * prefixlen == 0 is valid because there can be a case when
8019 		 * all addresses are matched.
8020 		 */
8021 	}
8022 
8023 	if (msg->sadb_msg_type >= nitems(key_typesw) ||
8024 	    key_typesw[msg->sadb_msg_type] == NULL) {
8025 		PFKEYSTAT_INC(out_invmsgtype);
8026 		error = EINVAL;
8027 		goto senderror;
8028 	}
8029 
8030 	return (*key_typesw[msg->sadb_msg_type])(so, m, &mh);
8031 
8032 senderror:
8033 	msg->sadb_msg_errno = error;
8034 	return key_sendup_mbuf(so, m, target);
8035 }
8036 
8037 static int
8038 key_senderror(struct socket *so, struct mbuf *m, int code)
8039 {
8040 	struct sadb_msg *msg;
8041 
8042 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8043 		("mbuf too small, len %u", m->m_len));
8044 
8045 	msg = mtod(m, struct sadb_msg *);
8046 	msg->sadb_msg_errno = code;
8047 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
8048 }
8049 
8050 /*
8051  * set the pointer to each header into message buffer.
8052  * m will be freed on error.
8053  * XXX larger-than-MCLBYTES extension?
8054  */
8055 static int
8056 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
8057 {
8058 	struct mbuf *n;
8059 	struct sadb_ext *ext;
8060 	size_t off, end;
8061 	int extlen;
8062 	int toff;
8063 
8064 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
8065 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
8066 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8067 		("mbuf too small, len %u", m->m_len));
8068 
8069 	/* initialize */
8070 	bzero(mhp, sizeof(*mhp));
8071 
8072 	mhp->msg = mtod(m, struct sadb_msg *);
8073 	mhp->ext[0] = (struct sadb_ext *)mhp->msg;	/*XXX backward compat */
8074 
8075 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
8076 	extlen = end;	/*just in case extlen is not updated*/
8077 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
8078 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
8079 		if (!n) {
8080 			/* m is already freed */
8081 			return ENOBUFS;
8082 		}
8083 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8084 
8085 		/* set pointer */
8086 		switch (ext->sadb_ext_type) {
8087 		case SADB_EXT_SA:
8088 		case SADB_EXT_ADDRESS_SRC:
8089 		case SADB_EXT_ADDRESS_DST:
8090 		case SADB_EXT_ADDRESS_PROXY:
8091 		case SADB_EXT_LIFETIME_CURRENT:
8092 		case SADB_EXT_LIFETIME_HARD:
8093 		case SADB_EXT_LIFETIME_SOFT:
8094 		case SADB_EXT_KEY_AUTH:
8095 		case SADB_EXT_KEY_ENCRYPT:
8096 		case SADB_EXT_IDENTITY_SRC:
8097 		case SADB_EXT_IDENTITY_DST:
8098 		case SADB_EXT_SENSITIVITY:
8099 		case SADB_EXT_PROPOSAL:
8100 		case SADB_EXT_SUPPORTED_AUTH:
8101 		case SADB_EXT_SUPPORTED_ENCRYPT:
8102 		case SADB_EXT_SPIRANGE:
8103 		case SADB_X_EXT_POLICY:
8104 		case SADB_X_EXT_SA2:
8105 		case SADB_X_EXT_NAT_T_TYPE:
8106 		case SADB_X_EXT_NAT_T_SPORT:
8107 		case SADB_X_EXT_NAT_T_DPORT:
8108 		case SADB_X_EXT_NAT_T_OAI:
8109 		case SADB_X_EXT_NAT_T_OAR:
8110 		case SADB_X_EXT_NAT_T_FRAG:
8111 		case SADB_X_EXT_SA_REPLAY:
8112 		case SADB_X_EXT_NEW_ADDRESS_SRC:
8113 		case SADB_X_EXT_NEW_ADDRESS_DST:
8114 			/* duplicate check */
8115 			/*
8116 			 * XXX Are there duplication payloads of either
8117 			 * KEY_AUTH or KEY_ENCRYPT ?
8118 			 */
8119 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
8120 				ipseclog((LOG_DEBUG, "%s: duplicate ext_type "
8121 					"%u\n", __func__, ext->sadb_ext_type));
8122 				m_freem(m);
8123 				PFKEYSTAT_INC(out_dupext);
8124 				return EINVAL;
8125 			}
8126 			break;
8127 		default:
8128 			ipseclog((LOG_DEBUG, "%s: invalid ext_type %u\n",
8129 				__func__, ext->sadb_ext_type));
8130 			m_freem(m);
8131 			PFKEYSTAT_INC(out_invexttype);
8132 			return EINVAL;
8133 		}
8134 
8135 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
8136 
8137 		if (key_validate_ext(ext, extlen)) {
8138 			m_freem(m);
8139 			PFKEYSTAT_INC(out_invlen);
8140 			return EINVAL;
8141 		}
8142 
8143 		n = m_pulldown(m, off, extlen, &toff);
8144 		if (!n) {
8145 			/* m is already freed */
8146 			return ENOBUFS;
8147 		}
8148 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8149 
8150 		mhp->ext[ext->sadb_ext_type] = ext;
8151 		mhp->extoff[ext->sadb_ext_type] = off;
8152 		mhp->extlen[ext->sadb_ext_type] = extlen;
8153 	}
8154 
8155 	if (off != end) {
8156 		m_freem(m);
8157 		PFKEYSTAT_INC(out_invlen);
8158 		return EINVAL;
8159 	}
8160 
8161 	return 0;
8162 }
8163 
8164 static int
8165 key_validate_ext(const struct sadb_ext *ext, int len)
8166 {
8167 	const struct sockaddr *sa;
8168 	enum { NONE, ADDR } checktype = NONE;
8169 	int baselen = 0;
8170 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
8171 
8172 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
8173 		return EINVAL;
8174 
8175 	/* if it does not match minimum/maximum length, bail */
8176 	if (ext->sadb_ext_type >= nitems(minsize) ||
8177 	    ext->sadb_ext_type >= nitems(maxsize))
8178 		return EINVAL;
8179 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
8180 		return EINVAL;
8181 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
8182 		return EINVAL;
8183 
8184 	/* more checks based on sadb_ext_type XXX need more */
8185 	switch (ext->sadb_ext_type) {
8186 	case SADB_EXT_ADDRESS_SRC:
8187 	case SADB_EXT_ADDRESS_DST:
8188 	case SADB_EXT_ADDRESS_PROXY:
8189 	case SADB_X_EXT_NAT_T_OAI:
8190 	case SADB_X_EXT_NAT_T_OAR:
8191 	case SADB_X_EXT_NEW_ADDRESS_SRC:
8192 	case SADB_X_EXT_NEW_ADDRESS_DST:
8193 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
8194 		checktype = ADDR;
8195 		break;
8196 	case SADB_EXT_IDENTITY_SRC:
8197 	case SADB_EXT_IDENTITY_DST:
8198 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
8199 		    SADB_X_IDENTTYPE_ADDR) {
8200 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
8201 			checktype = ADDR;
8202 		} else
8203 			checktype = NONE;
8204 		break;
8205 	default:
8206 		checktype = NONE;
8207 		break;
8208 	}
8209 
8210 	switch (checktype) {
8211 	case NONE:
8212 		break;
8213 	case ADDR:
8214 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8215 		if (len < baselen + sal)
8216 			return EINVAL;
8217 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8218 			return EINVAL;
8219 		break;
8220 	}
8221 
8222 	return 0;
8223 }
8224 
8225 void
8226 spdcache_init(void)
8227 {
8228 	int i;
8229 
8230 	TUNABLE_INT_FETCH("net.key.spdcache.maxentries",
8231 	    &V_key_spdcache_maxentries);
8232 	TUNABLE_INT_FETCH("net.key.spdcache.threshold",
8233 	    &V_key_spdcache_threshold);
8234 
8235 	if (V_key_spdcache_maxentries) {
8236 		V_key_spdcache_maxentries = MAX(V_key_spdcache_maxentries,
8237 		    SPDCACHE_MAX_ENTRIES_PER_HASH);
8238 		V_spdcachehashtbl = hashinit(V_key_spdcache_maxentries /
8239 		    SPDCACHE_MAX_ENTRIES_PER_HASH,
8240 		    M_IPSEC_SPDCACHE, &V_spdcachehash_mask);
8241 		V_key_spdcache_maxentries = (V_spdcachehash_mask + 1)
8242 		    * SPDCACHE_MAX_ENTRIES_PER_HASH;
8243 
8244 		V_spdcache_lock = malloc(sizeof(struct mtx) *
8245 		    (V_spdcachehash_mask + 1),
8246 		    M_IPSEC_SPDCACHE, M_WAITOK|M_ZERO);
8247 
8248 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8249 			SPDCACHE_LOCK_INIT(i);
8250 	}
8251 }
8252 
8253 struct spdcache_entry *
8254 spdcache_entry_alloc(const struct secpolicyindex *spidx, struct secpolicy *sp)
8255 {
8256 	struct spdcache_entry *entry;
8257 
8258 	entry = malloc(sizeof(struct spdcache_entry),
8259 		    M_IPSEC_SPDCACHE, M_NOWAIT|M_ZERO);
8260 	if (entry == NULL)
8261 		return NULL;
8262 
8263 	if (sp != NULL)
8264 		SP_ADDREF(sp);
8265 
8266 	entry->spidx = *spidx;
8267 	entry->sp = sp;
8268 
8269 	return (entry);
8270 }
8271 
8272 void
8273 spdcache_entry_free(struct spdcache_entry *entry)
8274 {
8275 
8276 	if (entry->sp != NULL)
8277 		key_freesp(&entry->sp);
8278 	free(entry, M_IPSEC_SPDCACHE);
8279 }
8280 
8281 void
8282 spdcache_clear(void)
8283 {
8284 	struct spdcache_entry *entry;
8285 	int i;
8286 
8287 	for (i = 0; i < V_spdcachehash_mask + 1; ++i) {
8288 		SPDCACHE_LOCK(i);
8289 		while (!LIST_EMPTY(&V_spdcachehashtbl[i])) {
8290 			entry = LIST_FIRST(&V_spdcachehashtbl[i]);
8291 			LIST_REMOVE(entry, chain);
8292 			spdcache_entry_free(entry);
8293 		}
8294 		SPDCACHE_UNLOCK(i);
8295 	}
8296 }
8297 
8298 #ifdef VIMAGE
8299 void
8300 spdcache_destroy(void)
8301 {
8302 	int i;
8303 
8304 	if (SPDCACHE_ENABLED()) {
8305 		spdcache_clear();
8306 		hashdestroy(V_spdcachehashtbl, M_IPSEC_SPDCACHE, V_spdcachehash_mask);
8307 
8308 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8309 			SPDCACHE_LOCK_DESTROY(i);
8310 
8311 		free(V_spdcache_lock, M_IPSEC_SPDCACHE);
8312 	}
8313 }
8314 #endif
8315 
8316 static void
8317 key_vnet_init(void *arg __unused)
8318 {
8319 	int i;
8320 
8321 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8322 		TAILQ_INIT(&V_sptree[i]);
8323 		TAILQ_INIT(&V_sptree_ifnet[i]);
8324 	}
8325 
8326 	TAILQ_INIT(&V_sahtree);
8327 	V_sphashtbl = hashinit(SPHASH_NHASH, M_IPSEC_SP, &V_sphash_mask);
8328 	V_savhashtbl = hashinit(SAVHASH_NHASH, M_IPSEC_SA, &V_savhash_mask);
8329 	V_sahaddrhashtbl = hashinit(SAHHASH_NHASH, M_IPSEC_SAH,
8330 	    &V_sahaddrhash_mask);
8331 	V_acqaddrhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8332 	    &V_acqaddrhash_mask);
8333 	V_acqseqhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8334 	    &V_acqseqhash_mask);
8335 
8336 	spdcache_init();
8337 
8338 	for (i = 0; i <= SADB_SATYPE_MAX; i++)
8339 		LIST_INIT(&V_regtree[i]);
8340 
8341 	LIST_INIT(&V_acqtree);
8342 	LIST_INIT(&V_spacqtree);
8343 }
8344 VNET_SYSINIT(key_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_SECOND,
8345     key_vnet_init, NULL);
8346 
8347 static void
8348 key_init(void *arg __unused)
8349 {
8350 
8351 	ipsec_key_lft_zone = uma_zcreate("IPsec SA lft_c",
8352 	    sizeof(uint64_t) * 2, NULL, NULL, NULL, NULL,
8353 	    UMA_ALIGN_PTR, UMA_ZONE_PCPU);
8354 
8355 	SPTREE_LOCK_INIT();
8356 	REGTREE_LOCK_INIT();
8357 	SAHTREE_LOCK_INIT();
8358 	ACQ_LOCK_INIT();
8359 	SPACQ_LOCK_INIT();
8360 	SPI_ALLOC_LOCK_INIT();
8361 
8362 #ifndef IPSEC_DEBUG2
8363 	callout_init(&key_timer, 1);
8364 	callout_reset(&key_timer, hz, key_timehandler, NULL);
8365 #endif /*IPSEC_DEBUG2*/
8366 
8367 	/* initialize key statistics */
8368 	keystat.getspi_count = 1;
8369 
8370 	if (bootverbose)
8371 		printf("IPsec: Initialized Security Association Processing.\n");
8372 }
8373 SYSINIT(key_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, key_init, NULL);
8374 
8375 #ifdef VIMAGE
8376 static void
8377 key_vnet_destroy(void *arg __unused)
8378 {
8379 	struct secashead_queue sahdrainq;
8380 	struct secpolicy_queue drainq;
8381 	struct secpolicy *sp, *nextsp;
8382 	struct secacq *acq, *nextacq;
8383 	struct secspacq *spacq, *nextspacq;
8384 	struct secashead *sah;
8385 	struct secasvar *sav;
8386 	struct secreg *reg;
8387 	int i;
8388 
8389 	/*
8390 	 * XXX: can we just call free() for each object without
8391 	 * walking through safe way with releasing references?
8392 	 */
8393 	TAILQ_INIT(&drainq);
8394 	SPTREE_WLOCK();
8395 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8396 		TAILQ_CONCAT(&drainq, &V_sptree[i], chain);
8397 		TAILQ_CONCAT(&drainq, &V_sptree_ifnet[i], chain);
8398 	}
8399 	for (i = 0; i < V_sphash_mask + 1; i++)
8400 		LIST_INIT(&V_sphashtbl[i]);
8401 	SPTREE_WUNLOCK();
8402 	spdcache_destroy();
8403 
8404 	sp = TAILQ_FIRST(&drainq);
8405 	while (sp != NULL) {
8406 		nextsp = TAILQ_NEXT(sp, chain);
8407 		key_freesp(&sp);
8408 		sp = nextsp;
8409 	}
8410 
8411 	TAILQ_INIT(&sahdrainq);
8412 	SAHTREE_WLOCK();
8413 	TAILQ_CONCAT(&sahdrainq, &V_sahtree, chain);
8414 	for (i = 0; i < V_savhash_mask + 1; i++)
8415 		LIST_INIT(&V_savhashtbl[i]);
8416 	for (i = 0; i < V_sahaddrhash_mask + 1; i++)
8417 		LIST_INIT(&V_sahaddrhashtbl[i]);
8418 	TAILQ_FOREACH(sah, &sahdrainq, chain) {
8419 		sah->state = SADB_SASTATE_DEAD;
8420 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
8421 			sav->state = SADB_SASTATE_DEAD;
8422 		}
8423 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
8424 			sav->state = SADB_SASTATE_DEAD;
8425 		}
8426 	}
8427 	SAHTREE_WUNLOCK();
8428 
8429 	key_freesah_flushed(&sahdrainq);
8430 	hashdestroy(V_sphashtbl, M_IPSEC_SP, V_sphash_mask);
8431 	hashdestroy(V_savhashtbl, M_IPSEC_SA, V_savhash_mask);
8432 	hashdestroy(V_sahaddrhashtbl, M_IPSEC_SAH, V_sahaddrhash_mask);
8433 
8434 	REGTREE_LOCK();
8435 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8436 		LIST_FOREACH(reg, &V_regtree[i], chain) {
8437 			if (__LIST_CHAINED(reg)) {
8438 				LIST_REMOVE(reg, chain);
8439 				free(reg, M_IPSEC_SAR);
8440 				break;
8441 			}
8442 		}
8443 	}
8444 	REGTREE_UNLOCK();
8445 
8446 	ACQ_LOCK();
8447 	acq = LIST_FIRST(&V_acqtree);
8448 	while (acq != NULL) {
8449 		nextacq = LIST_NEXT(acq, chain);
8450 		LIST_REMOVE(acq, chain);
8451 		free(acq, M_IPSEC_SAQ);
8452 		acq = nextacq;
8453 	}
8454 	for (i = 0; i < V_acqaddrhash_mask + 1; i++)
8455 		LIST_INIT(&V_acqaddrhashtbl[i]);
8456 	for (i = 0; i < V_acqseqhash_mask + 1; i++)
8457 		LIST_INIT(&V_acqseqhashtbl[i]);
8458 	ACQ_UNLOCK();
8459 
8460 	SPACQ_LOCK();
8461 	for (spacq = LIST_FIRST(&V_spacqtree); spacq != NULL;
8462 	    spacq = nextspacq) {
8463 		nextspacq = LIST_NEXT(spacq, chain);
8464 		if (__LIST_CHAINED(spacq)) {
8465 			LIST_REMOVE(spacq, chain);
8466 			free(spacq, M_IPSEC_SAQ);
8467 		}
8468 	}
8469 	SPACQ_UNLOCK();
8470 	hashdestroy(V_acqaddrhashtbl, M_IPSEC_SAQ, V_acqaddrhash_mask);
8471 	hashdestroy(V_acqseqhashtbl, M_IPSEC_SAQ, V_acqseqhash_mask);
8472 }
8473 VNET_SYSUNINIT(key_vnet_destroy, SI_SUB_PROTO_DOMAIN, SI_ORDER_SECOND,
8474     key_vnet_destroy, NULL);
8475 #endif
8476 
8477 /*
8478  * XXX: as long as domains are not unloadable, this function is never called,
8479  * provided for consistensy and future unload support.
8480  */
8481 static void
8482 key_destroy(void *arg __unused)
8483 {
8484 	uma_zdestroy(ipsec_key_lft_zone);
8485 
8486 #ifndef IPSEC_DEBUG2
8487 	callout_drain(&key_timer);
8488 #endif
8489 	SPTREE_LOCK_DESTROY();
8490 	REGTREE_LOCK_DESTROY();
8491 	SAHTREE_LOCK_DESTROY();
8492 	ACQ_LOCK_DESTROY();
8493 	SPACQ_LOCK_DESTROY();
8494 	SPI_ALLOC_LOCK_DESTROY();
8495 }
8496 SYSUNINIT(key_destroy, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, key_destroy, NULL);
8497 
8498 /* record data transfer on SA, and update timestamps */
8499 void
8500 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8501 {
8502 	IPSEC_ASSERT(sav != NULL, ("Null secasvar"));
8503 	IPSEC_ASSERT(m != NULL, ("Null mbuf"));
8504 
8505 	/*
8506 	 * XXX Currently, there is a difference of bytes size
8507 	 * between inbound and outbound processing.
8508 	 */
8509 	counter_u64_add(sav->lft_c_bytes, m->m_pkthdr.len);
8510 
8511 	/*
8512 	 * We use the number of packets as the unit of
8513 	 * allocations.  We increment the variable
8514 	 * whenever {esp,ah}_{in,out}put is called.
8515 	 */
8516 	counter_u64_add(sav->lft_c_allocations, 1);
8517 
8518 	/*
8519 	 * NOTE: We record CURRENT usetime by using wall clock,
8520 	 * in seconds.  HARD and SOFT lifetime are measured by the time
8521 	 * difference (again in seconds) from usetime.
8522 	 *
8523 	 *	usetime
8524 	 *	v     expire   expire
8525 	 * -----+-----+--------+---> t
8526 	 *	<--------------> HARD
8527 	 *	<-----> SOFT
8528 	 */
8529 	if (sav->firstused == 0)
8530 		sav->firstused = time_second;
8531 }
8532 
8533 /*
8534  * Take one of the kernel's security keys and convert it into a PF_KEY
8535  * structure within an mbuf, suitable for sending up to a waiting
8536  * application in user land.
8537  *
8538  * IN:
8539  *    src: A pointer to a kernel security key.
8540  *    exttype: Which type of key this is. Refer to the PF_KEY data structures.
8541  * OUT:
8542  *    a valid mbuf or NULL indicating an error
8543  *
8544  */
8545 
8546 static struct mbuf *
8547 key_setkey(struct seckey *src, uint16_t exttype)
8548 {
8549 	struct mbuf *m;
8550 	struct sadb_key *p;
8551 	int len;
8552 
8553 	if (src == NULL)
8554 		return NULL;
8555 
8556 	len = PFKEY_ALIGN8(sizeof(struct sadb_key) + _KEYLEN(src));
8557 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8558 	if (m == NULL)
8559 		return NULL;
8560 	m_align(m, len);
8561 	m->m_len = len;
8562 	p = mtod(m, struct sadb_key *);
8563 	bzero(p, len);
8564 	p->sadb_key_len = PFKEY_UNIT64(len);
8565 	p->sadb_key_exttype = exttype;
8566 	p->sadb_key_bits = src->bits;
8567 	bcopy(src->key_data, _KEYBUF(p), _KEYLEN(src));
8568 
8569 	return m;
8570 }
8571 
8572 /*
8573  * Take one of the kernel's lifetime data structures and convert it
8574  * into a PF_KEY structure within an mbuf, suitable for sending up to
8575  * a waiting application in user land.
8576  *
8577  * IN:
8578  *    src: A pointer to a kernel lifetime structure.
8579  *    exttype: Which type of lifetime this is. Refer to the PF_KEY
8580  *             data structures for more information.
8581  * OUT:
8582  *    a valid mbuf or NULL indicating an error
8583  *
8584  */
8585 
8586 static struct mbuf *
8587 key_setlifetime(struct seclifetime *src, uint16_t exttype)
8588 {
8589 	struct mbuf *m = NULL;
8590 	struct sadb_lifetime *p;
8591 	int len = PFKEY_ALIGN8(sizeof(struct sadb_lifetime));
8592 
8593 	if (src == NULL)
8594 		return NULL;
8595 
8596 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8597 	if (m == NULL)
8598 		return m;
8599 	m_align(m, len);
8600 	m->m_len = len;
8601 	p = mtod(m, struct sadb_lifetime *);
8602 
8603 	bzero(p, len);
8604 	p->sadb_lifetime_len = PFKEY_UNIT64(len);
8605 	p->sadb_lifetime_exttype = exttype;
8606 	p->sadb_lifetime_allocations = src->allocations;
8607 	p->sadb_lifetime_bytes = src->bytes;
8608 	p->sadb_lifetime_addtime = src->addtime;
8609 	p->sadb_lifetime_usetime = src->usetime;
8610 
8611 	return m;
8612 
8613 }
8614 
8615 const struct enc_xform *
8616 enc_algorithm_lookup(int alg)
8617 {
8618 	int i;
8619 
8620 	for (i = 0; i < nitems(supported_ealgs); i++)
8621 		if (alg == supported_ealgs[i].sadb_alg)
8622 			return (supported_ealgs[i].xform);
8623 	return (NULL);
8624 }
8625 
8626 const struct auth_hash *
8627 auth_algorithm_lookup(int alg)
8628 {
8629 	int i;
8630 
8631 	for (i = 0; i < nitems(supported_aalgs); i++)
8632 		if (alg == supported_aalgs[i].sadb_alg)
8633 			return (supported_aalgs[i].xform);
8634 	return (NULL);
8635 }
8636 
8637 const struct comp_algo *
8638 comp_algorithm_lookup(int alg)
8639 {
8640 	int i;
8641 
8642 	for (i = 0; i < nitems(supported_calgs); i++)
8643 		if (alg == supported_calgs[i].sadb_alg)
8644 			return (supported_calgs[i].xform);
8645 	return (NULL);
8646 }
8647