xref: /freebsd/sys/netipsec/key.c (revision 5e3190f700637fcfc1a52daeaa4a031fdd2557c7)
1 /*	$KAME: key.c,v 1.191 2001/06/27 10:46:49 sakane Exp $	*/
2 
3 /*-
4  * SPDX-License-Identifier: BSD-3-Clause
5  *
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * This code is referd to RFC 2367
36  */
37 
38 #include "opt_inet.h"
39 #include "opt_inet6.h"
40 #include "opt_ipsec.h"
41 
42 #include <sys/types.h>
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
46 #include <sys/fnv_hash.h>
47 #include <sys/lock.h>
48 #include <sys/mutex.h>
49 #include <sys/mbuf.h>
50 #include <sys/domain.h>
51 #include <sys/protosw.h>
52 #include <sys/malloc.h>
53 #include <sys/rmlock.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/sysctl.h>
57 #include <sys/errno.h>
58 #include <sys/proc.h>
59 #include <sys/queue.h>
60 #include <sys/refcount.h>
61 #include <sys/syslog.h>
62 
63 #include <vm/uma.h>
64 
65 #include <net/if.h>
66 #include <net/if_var.h>
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/ip.h>
72 #include <netinet/in_var.h>
73 #include <netinet/udp.h>
74 
75 #ifdef INET6
76 #include <netinet/ip6.h>
77 #include <netinet6/in6_var.h>
78 #include <netinet6/ip6_var.h>
79 #endif /* INET6 */
80 
81 #include <net/pfkeyv2.h>
82 #include <netipsec/keydb.h>
83 #include <netipsec/key.h>
84 #include <netipsec/keysock.h>
85 #include <netipsec/key_debug.h>
86 
87 #include <netipsec/ipsec.h>
88 #ifdef INET6
89 #include <netipsec/ipsec6.h>
90 #endif
91 
92 #include <netipsec/xform.h>
93 #include <machine/in_cksum.h>
94 #include <machine/stdarg.h>
95 
96 /* randomness */
97 #include <sys/random.h>
98 
99 #define FULLMASK	0xff
100 #define	_BITS(bytes)	((bytes) << 3)
101 
102 #define	UINT32_80PCT	0xcccccccc
103 /*
104  * Note on SA reference counting:
105  * - SAs that are not in DEAD state will have (total external reference + 1)
106  *   following value in reference count field.  they cannot be freed and are
107  *   referenced from SA header.
108  * - SAs that are in DEAD state will have (total external reference)
109  *   in reference count field.  they are ready to be freed.  reference from
110  *   SA header will be removed in key_delsav(), when the reference count
111  *   field hits 0 (= no external reference other than from SA header.
112  */
113 
114 VNET_DEFINE(u_int32_t, key_debug_level) = 0;
115 VNET_DEFINE_STATIC(u_int, key_spi_trycnt) = 1000;
116 VNET_DEFINE_STATIC(u_int32_t, key_spi_minval) = 0x100;
117 VNET_DEFINE_STATIC(u_int32_t, key_spi_maxval) = 0x0fffffff;	/* XXX */
118 VNET_DEFINE_STATIC(u_int32_t, policy_id) = 0;
119 /*interval to initialize randseed,1(m)*/
120 VNET_DEFINE_STATIC(u_int, key_int_random) = 60;
121 /* interval to expire acquiring, 30(s)*/
122 VNET_DEFINE_STATIC(u_int, key_larval_lifetime) = 30;
123 /* counter for blocking SADB_ACQUIRE.*/
124 VNET_DEFINE_STATIC(int, key_blockacq_count) = 10;
125 /* lifetime for blocking SADB_ACQUIRE.*/
126 VNET_DEFINE_STATIC(int, key_blockacq_lifetime) = 20;
127 /* preferred old sa rather than new sa.*/
128 VNET_DEFINE_STATIC(int, key_preferred_oldsa) = 1;
129 #define	V_key_spi_trycnt	VNET(key_spi_trycnt)
130 #define	V_key_spi_minval	VNET(key_spi_minval)
131 #define	V_key_spi_maxval	VNET(key_spi_maxval)
132 #define	V_policy_id		VNET(policy_id)
133 #define	V_key_int_random	VNET(key_int_random)
134 #define	V_key_larval_lifetime	VNET(key_larval_lifetime)
135 #define	V_key_blockacq_count	VNET(key_blockacq_count)
136 #define	V_key_blockacq_lifetime	VNET(key_blockacq_lifetime)
137 #define	V_key_preferred_oldsa	VNET(key_preferred_oldsa)
138 
139 VNET_DEFINE_STATIC(u_int32_t, acq_seq) = 0;
140 #define	V_acq_seq		VNET(acq_seq)
141 
142 VNET_DEFINE_STATIC(uint32_t, sp_genid) = 0;
143 #define	V_sp_genid		VNET(sp_genid)
144 
145 /* SPD */
146 TAILQ_HEAD(secpolicy_queue, secpolicy);
147 LIST_HEAD(secpolicy_list, secpolicy);
148 VNET_DEFINE_STATIC(struct secpolicy_queue, sptree[IPSEC_DIR_MAX]);
149 VNET_DEFINE_STATIC(struct secpolicy_queue, sptree_ifnet[IPSEC_DIR_MAX]);
150 static struct rmlock sptree_lock;
151 #define	V_sptree		VNET(sptree)
152 #define	V_sptree_ifnet		VNET(sptree_ifnet)
153 #define	SPTREE_LOCK_INIT()      rm_init(&sptree_lock, "sptree")
154 #define	SPTREE_LOCK_DESTROY()   rm_destroy(&sptree_lock)
155 #define	SPTREE_RLOCK_TRACKER    struct rm_priotracker sptree_tracker
156 #define	SPTREE_RLOCK()          rm_rlock(&sptree_lock, &sptree_tracker)
157 #define	SPTREE_RUNLOCK()        rm_runlock(&sptree_lock, &sptree_tracker)
158 #define	SPTREE_RLOCK_ASSERT()   rm_assert(&sptree_lock, RA_RLOCKED)
159 #define	SPTREE_WLOCK()          rm_wlock(&sptree_lock)
160 #define	SPTREE_WUNLOCK()        rm_wunlock(&sptree_lock)
161 #define	SPTREE_WLOCK_ASSERT()   rm_assert(&sptree_lock, RA_WLOCKED)
162 #define	SPTREE_UNLOCK_ASSERT()  rm_assert(&sptree_lock, RA_UNLOCKED)
163 
164 /* Hash table for lookup SP using unique id */
165 VNET_DEFINE_STATIC(struct secpolicy_list *, sphashtbl);
166 VNET_DEFINE_STATIC(u_long, sphash_mask);
167 #define	V_sphashtbl		VNET(sphashtbl)
168 #define	V_sphash_mask		VNET(sphash_mask)
169 
170 #define	SPHASH_NHASH_LOG2	7
171 #define	SPHASH_NHASH		(1 << SPHASH_NHASH_LOG2)
172 #define	SPHASH_HASHVAL(id)	(key_u32hash(id) & V_sphash_mask)
173 #define	SPHASH_HASH(id)		&V_sphashtbl[SPHASH_HASHVAL(id)]
174 
175 /* SPD cache */
176 struct spdcache_entry {
177    struct secpolicyindex spidx;	/* secpolicyindex */
178    struct secpolicy *sp;	/* cached policy to be used */
179 
180    LIST_ENTRY(spdcache_entry) chain;
181 };
182 LIST_HEAD(spdcache_entry_list, spdcache_entry);
183 
184 #define	SPDCACHE_MAX_ENTRIES_PER_HASH	8
185 
186 VNET_DEFINE_STATIC(u_int, key_spdcache_maxentries) = 0;
187 #define	V_key_spdcache_maxentries	VNET(key_spdcache_maxentries)
188 VNET_DEFINE_STATIC(u_int, key_spdcache_threshold) = 32;
189 #define	V_key_spdcache_threshold	VNET(key_spdcache_threshold)
190 VNET_DEFINE_STATIC(unsigned long, spd_size) = 0;
191 #define	V_spd_size		VNET(spd_size)
192 
193 #define SPDCACHE_ENABLED()	(V_key_spdcache_maxentries != 0)
194 #define SPDCACHE_ACTIVE() \
195 	(SPDCACHE_ENABLED() && V_spd_size >= V_key_spdcache_threshold)
196 
197 VNET_DEFINE_STATIC(struct spdcache_entry_list *, spdcachehashtbl);
198 VNET_DEFINE_STATIC(u_long, spdcachehash_mask);
199 #define	V_spdcachehashtbl	VNET(spdcachehashtbl)
200 #define	V_spdcachehash_mask	VNET(spdcachehash_mask)
201 
202 #define	SPDCACHE_HASHVAL(idx) \
203 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->ul_proto) &  \
204 	    V_spdcachehash_mask)
205 
206 /* Each cache line is protected by a mutex */
207 VNET_DEFINE_STATIC(struct mtx *, spdcache_lock);
208 #define	V_spdcache_lock		VNET(spdcache_lock)
209 
210 #define	SPDCACHE_LOCK_INIT(a) \
211 	mtx_init(&V_spdcache_lock[a], "spdcache", \
212 	    "fast ipsec SPD cache", MTX_DEF|MTX_DUPOK)
213 #define	SPDCACHE_LOCK_DESTROY(a)	mtx_destroy(&V_spdcache_lock[a])
214 #define	SPDCACHE_LOCK(a)		mtx_lock(&V_spdcache_lock[a]);
215 #define	SPDCACHE_UNLOCK(a)		mtx_unlock(&V_spdcache_lock[a]);
216 
217 static struct sx spi_alloc_lock;
218 #define	SPI_ALLOC_LOCK_INIT()		sx_init(&spi_alloc_lock, "spialloc")
219 #define	SPI_ALLOC_LOCK_DESTROY()	sx_destroy(&spi_alloc_lock)
220 #define	SPI_ALLOC_LOCK()          	sx_xlock(&spi_alloc_lock)
221 #define	SPI_ALLOC_UNLOCK()        	sx_unlock(&spi_alloc_lock)
222 #define	SPI_ALLOC_LOCK_ASSERT()   	sx_assert(&spi_alloc_lock, SA_XLOCKED)
223 
224 /* SAD */
225 TAILQ_HEAD(secashead_queue, secashead);
226 LIST_HEAD(secashead_list, secashead);
227 VNET_DEFINE_STATIC(struct secashead_queue, sahtree);
228 static struct rmlock sahtree_lock;
229 #define	V_sahtree		VNET(sahtree)
230 #define	SAHTREE_LOCK_INIT()	rm_init(&sahtree_lock, "sahtree")
231 #define	SAHTREE_LOCK_DESTROY()	rm_destroy(&sahtree_lock)
232 #define	SAHTREE_RLOCK_TRACKER	struct rm_priotracker sahtree_tracker
233 #define	SAHTREE_RLOCK()		rm_rlock(&sahtree_lock, &sahtree_tracker)
234 #define	SAHTREE_RUNLOCK()	rm_runlock(&sahtree_lock, &sahtree_tracker)
235 #define	SAHTREE_RLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_RLOCKED)
236 #define	SAHTREE_WLOCK()		rm_wlock(&sahtree_lock)
237 #define	SAHTREE_WUNLOCK()	rm_wunlock(&sahtree_lock)
238 #define	SAHTREE_WLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_WLOCKED)
239 #define	SAHTREE_UNLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_UNLOCKED)
240 
241 /* Hash table for lookup in SAD using SA addresses */
242 VNET_DEFINE_STATIC(struct secashead_list *, sahaddrhashtbl);
243 VNET_DEFINE_STATIC(u_long, sahaddrhash_mask);
244 #define	V_sahaddrhashtbl	VNET(sahaddrhashtbl)
245 #define	V_sahaddrhash_mask	VNET(sahaddrhash_mask)
246 
247 #define	SAHHASH_NHASH_LOG2	7
248 #define	SAHHASH_NHASH		(1 << SAHHASH_NHASH_LOG2)
249 #define	SAHADDRHASH_HASHVAL(idx)	\
250 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->proto) & \
251 	    V_sahaddrhash_mask)
252 #define	SAHADDRHASH_HASH(saidx)		\
253     &V_sahaddrhashtbl[SAHADDRHASH_HASHVAL(saidx)]
254 
255 /* Hash table for lookup in SAD using SPI */
256 LIST_HEAD(secasvar_list, secasvar);
257 VNET_DEFINE_STATIC(struct secasvar_list *, savhashtbl);
258 VNET_DEFINE_STATIC(u_long, savhash_mask);
259 #define	V_savhashtbl		VNET(savhashtbl)
260 #define	V_savhash_mask		VNET(savhash_mask)
261 #define	SAVHASH_NHASH_LOG2	7
262 #define	SAVHASH_NHASH		(1 << SAVHASH_NHASH_LOG2)
263 #define	SAVHASH_HASHVAL(spi)	(key_u32hash(spi) & V_savhash_mask)
264 #define	SAVHASH_HASH(spi)	&V_savhashtbl[SAVHASH_HASHVAL(spi)]
265 
266 static uint32_t
267 key_addrprotohash(const union sockaddr_union *src,
268     const union sockaddr_union *dst, const uint8_t *proto)
269 {
270 	uint32_t hval;
271 
272 	hval = fnv_32_buf(proto, sizeof(*proto),
273 	    FNV1_32_INIT);
274 	switch (dst->sa.sa_family) {
275 #ifdef INET
276 	case AF_INET:
277 		hval = fnv_32_buf(&src->sin.sin_addr,
278 		    sizeof(in_addr_t), hval);
279 		hval = fnv_32_buf(&dst->sin.sin_addr,
280 		    sizeof(in_addr_t), hval);
281 		break;
282 #endif
283 #ifdef INET6
284 	case AF_INET6:
285 		hval = fnv_32_buf(&src->sin6.sin6_addr,
286 		    sizeof(struct in6_addr), hval);
287 		hval = fnv_32_buf(&dst->sin6.sin6_addr,
288 		    sizeof(struct in6_addr), hval);
289 		break;
290 #endif
291 	default:
292 		hval = 0;
293 		ipseclog((LOG_DEBUG, "%s: unknown address family %d\n",
294 		    __func__, dst->sa.sa_family));
295 	}
296 	return (hval);
297 }
298 
299 static uint32_t
300 key_u32hash(uint32_t val)
301 {
302 
303 	return (fnv_32_buf(&val, sizeof(val), FNV1_32_INIT));
304 }
305 
306 							/* registed list */
307 VNET_DEFINE_STATIC(LIST_HEAD(_regtree, secreg), regtree[SADB_SATYPE_MAX + 1]);
308 #define	V_regtree		VNET(regtree)
309 static struct mtx regtree_lock;
310 #define	REGTREE_LOCK_INIT() \
311 	mtx_init(&regtree_lock, "regtree", "fast ipsec regtree", MTX_DEF)
312 #define	REGTREE_LOCK_DESTROY()	mtx_destroy(&regtree_lock)
313 #define	REGTREE_LOCK()		mtx_lock(&regtree_lock)
314 #define	REGTREE_UNLOCK()	mtx_unlock(&regtree_lock)
315 #define	REGTREE_LOCK_ASSERT()	mtx_assert(&regtree_lock, MA_OWNED)
316 
317 /* Acquiring list */
318 LIST_HEAD(secacq_list, secacq);
319 VNET_DEFINE_STATIC(struct secacq_list, acqtree);
320 #define	V_acqtree		VNET(acqtree)
321 static struct mtx acq_lock;
322 #define	ACQ_LOCK_INIT() \
323     mtx_init(&acq_lock, "acqtree", "ipsec SA acquiring list", MTX_DEF)
324 #define	ACQ_LOCK_DESTROY()	mtx_destroy(&acq_lock)
325 #define	ACQ_LOCK()		mtx_lock(&acq_lock)
326 #define	ACQ_UNLOCK()		mtx_unlock(&acq_lock)
327 #define	ACQ_LOCK_ASSERT()	mtx_assert(&acq_lock, MA_OWNED)
328 
329 /* Hash table for lookup in ACQ list using SA addresses */
330 VNET_DEFINE_STATIC(struct secacq_list *, acqaddrhashtbl);
331 VNET_DEFINE_STATIC(u_long, acqaddrhash_mask);
332 #define	V_acqaddrhashtbl	VNET(acqaddrhashtbl)
333 #define	V_acqaddrhash_mask	VNET(acqaddrhash_mask)
334 
335 /* Hash table for lookup in ACQ list using SEQ number */
336 VNET_DEFINE_STATIC(struct secacq_list *, acqseqhashtbl);
337 VNET_DEFINE_STATIC(u_long, acqseqhash_mask);
338 #define	V_acqseqhashtbl		VNET(acqseqhashtbl)
339 #define	V_acqseqhash_mask	VNET(acqseqhash_mask)
340 
341 #define	ACQHASH_NHASH_LOG2	7
342 #define	ACQHASH_NHASH		(1 << ACQHASH_NHASH_LOG2)
343 #define	ACQADDRHASH_HASHVAL(idx)	\
344 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->proto) & \
345 	    V_acqaddrhash_mask)
346 #define	ACQSEQHASH_HASHVAL(seq)		\
347     (key_u32hash(seq) & V_acqseqhash_mask)
348 #define	ACQADDRHASH_HASH(saidx)	\
349     &V_acqaddrhashtbl[ACQADDRHASH_HASHVAL(saidx)]
350 #define	ACQSEQHASH_HASH(seq)	\
351     &V_acqseqhashtbl[ACQSEQHASH_HASHVAL(seq)]
352 							/* SP acquiring list */
353 VNET_DEFINE_STATIC(LIST_HEAD(_spacqtree, secspacq), spacqtree);
354 #define	V_spacqtree		VNET(spacqtree)
355 static struct mtx spacq_lock;
356 #define	SPACQ_LOCK_INIT() \
357 	mtx_init(&spacq_lock, "spacqtree", \
358 		"fast ipsec security policy acquire list", MTX_DEF)
359 #define	SPACQ_LOCK_DESTROY()	mtx_destroy(&spacq_lock)
360 #define	SPACQ_LOCK()		mtx_lock(&spacq_lock)
361 #define	SPACQ_UNLOCK()		mtx_unlock(&spacq_lock)
362 #define	SPACQ_LOCK_ASSERT()	mtx_assert(&spacq_lock, MA_OWNED)
363 
364 static const int minsize[] = {
365 	[SADB_EXT_RESERVED] = sizeof(struct sadb_msg),
366 	[SADB_EXT_SA] = sizeof(struct sadb_sa),
367 	[SADB_EXT_LIFETIME_CURRENT] = sizeof(struct sadb_lifetime),
368 	[SADB_EXT_LIFETIME_HARD] = sizeof(struct sadb_lifetime),
369 	[SADB_EXT_LIFETIME_SOFT] = sizeof(struct sadb_lifetime),
370 	[SADB_EXT_ADDRESS_SRC] = sizeof(struct sadb_address),
371 	[SADB_EXT_ADDRESS_DST] = sizeof(struct sadb_address),
372 	[SADB_EXT_ADDRESS_PROXY] = sizeof(struct sadb_address),
373 	[SADB_EXT_KEY_AUTH] = sizeof(struct sadb_key),
374 	[SADB_EXT_KEY_ENCRYPT] = sizeof(struct sadb_key),
375 	[SADB_EXT_IDENTITY_SRC] = sizeof(struct sadb_ident),
376 	[SADB_EXT_IDENTITY_DST] = sizeof(struct sadb_ident),
377 	[SADB_EXT_SENSITIVITY] = sizeof(struct sadb_sens),
378 	[SADB_EXT_PROPOSAL] = sizeof(struct sadb_prop),
379 	[SADB_EXT_SUPPORTED_AUTH] = sizeof(struct sadb_supported),
380 	[SADB_EXT_SUPPORTED_ENCRYPT] = sizeof(struct sadb_supported),
381 	[SADB_EXT_SPIRANGE] = sizeof(struct sadb_spirange),
382 	[SADB_X_EXT_KMPRIVATE] = 0,
383 	[SADB_X_EXT_POLICY] = sizeof(struct sadb_x_policy),
384 	[SADB_X_EXT_SA2] = sizeof(struct sadb_x_sa2),
385 	[SADB_X_EXT_NAT_T_TYPE] = sizeof(struct sadb_x_nat_t_type),
386 	[SADB_X_EXT_NAT_T_SPORT] = sizeof(struct sadb_x_nat_t_port),
387 	[SADB_X_EXT_NAT_T_DPORT] = sizeof(struct sadb_x_nat_t_port),
388 	[SADB_X_EXT_NAT_T_OAI] = sizeof(struct sadb_address),
389 	[SADB_X_EXT_NAT_T_OAR] = sizeof(struct sadb_address),
390 	[SADB_X_EXT_NAT_T_FRAG] = sizeof(struct sadb_x_nat_t_frag),
391 	[SADB_X_EXT_SA_REPLAY] = sizeof(struct sadb_x_sa_replay),
392 	[SADB_X_EXT_NEW_ADDRESS_SRC] = sizeof(struct sadb_address),
393 	[SADB_X_EXT_NEW_ADDRESS_DST] = sizeof(struct sadb_address),
394 };
395 _Static_assert(nitems(minsize) == SADB_EXT_MAX + 1, "minsize size mismatch");
396 
397 static const int maxsize[] = {
398 	[SADB_EXT_RESERVED] = sizeof(struct sadb_msg),
399 	[SADB_EXT_SA] = sizeof(struct sadb_sa),
400 	[SADB_EXT_LIFETIME_CURRENT] = sizeof(struct sadb_lifetime),
401 	[SADB_EXT_LIFETIME_HARD] = sizeof(struct sadb_lifetime),
402 	[SADB_EXT_LIFETIME_SOFT] = sizeof(struct sadb_lifetime),
403 	[SADB_EXT_ADDRESS_SRC] = 0,
404 	[SADB_EXT_ADDRESS_DST] = 0,
405 	[SADB_EXT_ADDRESS_PROXY] = 0,
406 	[SADB_EXT_KEY_AUTH] = 0,
407 	[SADB_EXT_KEY_ENCRYPT] = 0,
408 	[SADB_EXT_IDENTITY_SRC] = 0,
409 	[SADB_EXT_IDENTITY_DST] = 0,
410 	[SADB_EXT_SENSITIVITY] = 0,
411 	[SADB_EXT_PROPOSAL] = 0,
412 	[SADB_EXT_SUPPORTED_AUTH] = 0,
413 	[SADB_EXT_SUPPORTED_ENCRYPT] = 0,
414 	[SADB_EXT_SPIRANGE] = sizeof(struct sadb_spirange),
415 	[SADB_X_EXT_KMPRIVATE] = 0,
416 	[SADB_X_EXT_POLICY] = 0,
417 	[SADB_X_EXT_SA2] = sizeof(struct sadb_x_sa2),
418 	[SADB_X_EXT_NAT_T_TYPE] = sizeof(struct sadb_x_nat_t_type),
419 	[SADB_X_EXT_NAT_T_SPORT] = sizeof(struct sadb_x_nat_t_port),
420 	[SADB_X_EXT_NAT_T_DPORT] = sizeof(struct sadb_x_nat_t_port),
421 	[SADB_X_EXT_NAT_T_OAI] = 0,
422 	[SADB_X_EXT_NAT_T_OAR] = 0,
423 	[SADB_X_EXT_NAT_T_FRAG] = sizeof(struct sadb_x_nat_t_frag),
424 	[SADB_X_EXT_SA_REPLAY] = sizeof(struct sadb_x_sa_replay),
425 	[SADB_X_EXT_NEW_ADDRESS_SRC] = 0,
426 	[SADB_X_EXT_NEW_ADDRESS_DST] = 0,
427 };
428 _Static_assert(nitems(maxsize) == SADB_EXT_MAX + 1, "maxsize size mismatch");
429 
430 /*
431  * Internal values for SA flags:
432  * SADB_X_EXT_F_CLONED means that SA was cloned by key_updateaddresses,
433  *	thus we will not free the most of SA content in key_delsav().
434  */
435 #define	SADB_X_EXT_F_CLONED	0x80000000
436 
437 #define	SADB_CHECKLEN(_mhp, _ext)			\
438     ((_mhp)->extlen[(_ext)] < minsize[(_ext)] || (maxsize[(_ext)] != 0 && \
439 	((_mhp)->extlen[(_ext)] > maxsize[(_ext)])))
440 #define	SADB_CHECKHDR(_mhp, _ext)	((_mhp)->ext[(_ext)] == NULL)
441 
442 VNET_DEFINE_STATIC(int, ipsec_esp_keymin) = 256;
443 VNET_DEFINE_STATIC(int, ipsec_esp_auth) = 0;
444 VNET_DEFINE_STATIC(int, ipsec_ah_keymin) = 128;
445 
446 #define	V_ipsec_esp_keymin	VNET(ipsec_esp_keymin)
447 #define	V_ipsec_esp_auth	VNET(ipsec_esp_auth)
448 #define	V_ipsec_ah_keymin	VNET(ipsec_ah_keymin)
449 
450 #ifdef IPSEC_DEBUG
451 VNET_DEFINE(int, ipsec_debug) = 1;
452 #else
453 VNET_DEFINE(int, ipsec_debug) = 0;
454 #endif
455 
456 #ifdef INET
457 SYSCTL_DECL(_net_inet_ipsec);
458 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
459     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
460     "Enable IPsec debugging output when set.");
461 #endif
462 #ifdef INET6
463 SYSCTL_DECL(_net_inet6_ipsec6);
464 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug,
465     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
466     "Enable IPsec debugging output when set.");
467 #endif
468 
469 SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,
470 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_debug_level), 0, "");
471 
472 /* max count of trial for the decision of spi value */
473 SYSCTL_INT(_net_key, KEYCTL_SPI_TRY, spi_trycnt,
474 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_trycnt), 0, "");
475 
476 /* minimum spi value to allocate automatically. */
477 SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE, spi_minval,
478 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_minval), 0, "");
479 
480 /* maximun spi value to allocate automatically. */
481 SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE, spi_maxval,
482 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_maxval), 0, "");
483 
484 /* interval to initialize randseed */
485 SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT, int_random,
486 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_int_random), 0, "");
487 
488 /* lifetime for larval SA */
489 SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME, larval_lifetime,
490 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_larval_lifetime), 0, "");
491 
492 /* counter for blocking to send SADB_ACQUIRE to IKEd */
493 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT, blockacq_count,
494 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_count), 0, "");
495 
496 /* lifetime for blocking to send SADB_ACQUIRE to IKEd */
497 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME, blockacq_lifetime,
498 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_lifetime), 0, "");
499 
500 /* ESP auth */
501 SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH, esp_auth,
502 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_auth), 0, "");
503 
504 /* minimum ESP key length */
505 SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN, esp_keymin,
506 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_keymin), 0, "");
507 
508 /* minimum AH key length */
509 SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN, ah_keymin,
510 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_ah_keymin), 0, "");
511 
512 /* perfered old SA rather than new SA */
513 SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA, preferred_oldsa,
514 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_preferred_oldsa), 0, "");
515 
516 SYSCTL_NODE(_net_key, OID_AUTO, spdcache, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
517     "SPD cache");
518 
519 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, maxentries,
520 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_maxentries), 0,
521 	"Maximum number of entries in the SPD cache"
522 	" (power of 2, 0 to disable)");
523 
524 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, threshold,
525 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_threshold), 0,
526 	"Number of SPs that make the SPD cache active");
527 
528 #define __LIST_CHAINED(elm) \
529 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
530 
531 MALLOC_DEFINE(M_IPSEC_SA, "secasvar", "ipsec security association");
532 MALLOC_DEFINE(M_IPSEC_SAH, "sahead", "ipsec sa head");
533 MALLOC_DEFINE(M_IPSEC_SP, "ipsecpolicy", "ipsec security policy");
534 MALLOC_DEFINE(M_IPSEC_SR, "ipsecrequest", "ipsec security request");
535 MALLOC_DEFINE(M_IPSEC_MISC, "ipsec-misc", "ipsec miscellaneous");
536 MALLOC_DEFINE(M_IPSEC_SAQ, "ipsec-saq", "ipsec sa acquire");
537 MALLOC_DEFINE(M_IPSEC_SAR, "ipsec-reg", "ipsec sa acquire");
538 MALLOC_DEFINE(M_IPSEC_SPDCACHE, "ipsec-spdcache", "ipsec SPD cache");
539 
540 static uma_zone_t __read_mostly ipsec_key_lft_zone;
541 
542 /*
543  * set parameters into secpolicyindex buffer.
544  * Must allocate secpolicyindex buffer passed to this function.
545  */
546 #define KEY_SETSECSPIDX(_dir, s, d, ps, pd, ulp, idx) \
547 do { \
548 	bzero((idx), sizeof(struct secpolicyindex));                         \
549 	(idx)->dir = (_dir);                                                 \
550 	(idx)->prefs = (ps);                                                 \
551 	(idx)->prefd = (pd);                                                 \
552 	(idx)->ul_proto = (ulp);                                             \
553 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
554 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
555 } while (0)
556 
557 /*
558  * set parameters into secasindex buffer.
559  * Must allocate secasindex buffer before calling this function.
560  */
561 #define KEY_SETSECASIDX(p, m, r, s, d, idx) \
562 do { \
563 	bzero((idx), sizeof(struct secasindex));                             \
564 	(idx)->proto = (p);                                                  \
565 	(idx)->mode = (m);                                                   \
566 	(idx)->reqid = (r);                                                  \
567 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
568 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
569 	key_porttosaddr(&(idx)->src.sa, 0);				     \
570 	key_porttosaddr(&(idx)->dst.sa, 0);				     \
571 } while (0)
572 
573 /* key statistics */
574 struct _keystat {
575 	u_long getspi_count; /* the avarage of count to try to get new SPI */
576 } keystat;
577 
578 struct sadb_msghdr {
579 	struct sadb_msg *msg;
580 	struct sadb_ext *ext[SADB_EXT_MAX + 1];
581 	int extoff[SADB_EXT_MAX + 1];
582 	int extlen[SADB_EXT_MAX + 1];
583 };
584 
585 static const struct supported_ealgs {
586 	int sadb_alg;
587 	const struct enc_xform *xform;
588 } supported_ealgs[] = {
589 	{ SADB_X_EALG_AES,		&enc_xform_aes_cbc },
590 	{ SADB_EALG_NULL,		&enc_xform_null },
591 	{ SADB_X_EALG_AESCTR,		&enc_xform_aes_icm },
592 	{ SADB_X_EALG_AESGCM16,		&enc_xform_aes_nist_gcm },
593 	{ SADB_X_EALG_AESGMAC,		&enc_xform_aes_nist_gmac },
594 	{ SADB_X_EALG_CHACHA20POLY1305,	&enc_xform_chacha20_poly1305 },
595 };
596 
597 static const struct supported_aalgs {
598 	int sadb_alg;
599 	const struct auth_hash *xform;
600 } supported_aalgs[] = {
601 	{ SADB_X_AALG_NULL,		&auth_hash_null },
602 	{ SADB_AALG_SHA1HMAC,		&auth_hash_hmac_sha1 },
603 	{ SADB_X_AALG_SHA2_256,		&auth_hash_hmac_sha2_256 },
604 	{ SADB_X_AALG_SHA2_384,		&auth_hash_hmac_sha2_384 },
605 	{ SADB_X_AALG_SHA2_512,		&auth_hash_hmac_sha2_512 },
606 	{ SADB_X_AALG_AES128GMAC,	&auth_hash_nist_gmac_aes_128 },
607 	{ SADB_X_AALG_AES192GMAC,	&auth_hash_nist_gmac_aes_192 },
608 	{ SADB_X_AALG_AES256GMAC,	&auth_hash_nist_gmac_aes_256 },
609 	{ SADB_X_AALG_CHACHA20POLY1305,	&auth_hash_poly1305 },
610 };
611 
612 static const struct supported_calgs {
613 	int sadb_alg;
614 	const struct comp_algo *xform;
615 } supported_calgs[] = {
616 	{ SADB_X_CALG_DEFLATE,		&comp_algo_deflate },
617 };
618 
619 #ifndef IPSEC_DEBUG2
620 static struct callout key_timer;
621 #endif
622 
623 static void key_unlink(struct secpolicy *);
624 static void key_detach(struct secpolicy *);
625 static struct secpolicy *key_do_allocsp(struct secpolicyindex *spidx, u_int dir);
626 static struct secpolicy *key_getsp(struct secpolicyindex *);
627 static struct secpolicy *key_getspbyid(u_int32_t);
628 static struct mbuf *key_gather_mbuf(struct mbuf *,
629 	const struct sadb_msghdr *, int, int, ...);
630 static int key_spdadd(struct socket *, struct mbuf *,
631 	const struct sadb_msghdr *);
632 static uint32_t key_getnewspid(void);
633 static int key_spddelete(struct socket *, struct mbuf *,
634 	const struct sadb_msghdr *);
635 static int key_spddelete2(struct socket *, struct mbuf *,
636 	const struct sadb_msghdr *);
637 static int key_spdget(struct socket *, struct mbuf *,
638 	const struct sadb_msghdr *);
639 static int key_spdflush(struct socket *, struct mbuf *,
640 	const struct sadb_msghdr *);
641 static int key_spddump(struct socket *, struct mbuf *,
642 	const struct sadb_msghdr *);
643 static struct mbuf *key_setdumpsp(struct secpolicy *,
644 	u_int8_t, u_int32_t, u_int32_t);
645 static struct mbuf *key_sp2mbuf(struct secpolicy *);
646 static size_t key_getspreqmsglen(struct secpolicy *);
647 static int key_spdexpire(struct secpolicy *);
648 static struct secashead *key_newsah(struct secasindex *);
649 static void key_freesah(struct secashead **);
650 static void key_delsah(struct secashead *);
651 static struct secasvar *key_newsav(const struct sadb_msghdr *,
652     struct secasindex *, uint32_t, int *);
653 static void key_delsav(struct secasvar *);
654 static void key_unlinksav(struct secasvar *);
655 static struct secashead *key_getsah(struct secasindex *);
656 static int key_checkspidup(uint32_t);
657 static struct secasvar *key_getsavbyspi(uint32_t);
658 static int key_setnatt(struct secasvar *, const struct sadb_msghdr *);
659 static int key_setsaval(struct secasvar *, const struct sadb_msghdr *);
660 static int key_updatelifetimes(struct secasvar *, const struct sadb_msghdr *);
661 static int key_updateaddresses(struct socket *, struct mbuf *,
662     const struct sadb_msghdr *, struct secasvar *, struct secasindex *);
663 
664 static struct mbuf *key_setdumpsa(struct secasvar *, u_int8_t,
665 	u_int8_t, u_int32_t, u_int32_t);
666 static struct mbuf *key_setsadbmsg(u_int8_t, u_int16_t, u_int8_t,
667 	u_int32_t, pid_t, u_int16_t);
668 static struct mbuf *key_setsadbsa(struct secasvar *);
669 static struct mbuf *key_setsadbaddr(u_int16_t,
670 	const struct sockaddr *, u_int8_t, u_int16_t);
671 static struct mbuf *key_setsadbxport(u_int16_t, u_int16_t);
672 static struct mbuf *key_setsadbxtype(u_int16_t);
673 static struct mbuf *key_setsadbxsa2(u_int8_t, u_int32_t, u_int32_t);
674 static struct mbuf *key_setsadbxsareplay(u_int32_t);
675 static struct mbuf *key_setsadbxpolicy(u_int16_t, u_int8_t,
676 	u_int32_t, u_int32_t);
677 static struct seckey *key_dup_keymsg(const struct sadb_key *, size_t,
678     struct malloc_type *);
679 static struct seclifetime *key_dup_lifemsg(const struct sadb_lifetime *src,
680     struct malloc_type *);
681 
682 /* flags for key_cmpsaidx() */
683 #define CMP_HEAD	1	/* protocol, addresses. */
684 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
685 #define CMP_REQID	3	/* additionally HEAD, reaid. */
686 #define CMP_EXACTLY	4	/* all elements. */
687 static int key_cmpsaidx(const struct secasindex *,
688     const struct secasindex *, int);
689 static int key_cmpspidx_exactly(struct secpolicyindex *,
690     struct secpolicyindex *);
691 static int key_cmpspidx_withmask(struct secpolicyindex *,
692     struct secpolicyindex *);
693 static int key_bbcmp(const void *, const void *, u_int);
694 static uint8_t key_satype2proto(uint8_t);
695 static uint8_t key_proto2satype(uint8_t);
696 
697 static int key_getspi(struct socket *, struct mbuf *,
698 	const struct sadb_msghdr *);
699 static uint32_t key_do_getnewspi(struct sadb_spirange *, struct secasindex *);
700 static int key_update(struct socket *, struct mbuf *,
701 	const struct sadb_msghdr *);
702 static int key_add(struct socket *, struct mbuf *,
703 	const struct sadb_msghdr *);
704 static int key_setident(struct secashead *, const struct sadb_msghdr *);
705 static struct mbuf *key_getmsgbuf_x1(struct mbuf *,
706 	const struct sadb_msghdr *);
707 static int key_delete(struct socket *, struct mbuf *,
708 	const struct sadb_msghdr *);
709 static int key_delete_all(struct socket *, struct mbuf *,
710 	const struct sadb_msghdr *, struct secasindex *);
711 static int key_get(struct socket *, struct mbuf *,
712 	const struct sadb_msghdr *);
713 
714 static void key_getcomb_setlifetime(struct sadb_comb *);
715 static struct mbuf *key_getcomb_ealg(void);
716 static struct mbuf *key_getcomb_ah(void);
717 static struct mbuf *key_getcomb_ipcomp(void);
718 static struct mbuf *key_getprop(const struct secasindex *);
719 
720 static int key_acquire(const struct secasindex *, struct secpolicy *);
721 static uint32_t key_newacq(const struct secasindex *, int *);
722 static uint32_t key_getacq(const struct secasindex *, int *);
723 static int key_acqdone(const struct secasindex *, uint32_t);
724 static int key_acqreset(uint32_t);
725 static struct secspacq *key_newspacq(struct secpolicyindex *);
726 static struct secspacq *key_getspacq(struct secpolicyindex *);
727 static int key_acquire2(struct socket *, struct mbuf *,
728 	const struct sadb_msghdr *);
729 static int key_register(struct socket *, struct mbuf *,
730 	const struct sadb_msghdr *);
731 static int key_expire(struct secasvar *, int);
732 static int key_flush(struct socket *, struct mbuf *,
733 	const struct sadb_msghdr *);
734 static int key_dump(struct socket *, struct mbuf *,
735 	const struct sadb_msghdr *);
736 static int key_promisc(struct socket *, struct mbuf *,
737 	const struct sadb_msghdr *);
738 static int key_senderror(struct socket *, struct mbuf *, int);
739 static int key_validate_ext(const struct sadb_ext *, int);
740 static int key_align(struct mbuf *, struct sadb_msghdr *);
741 static struct mbuf *key_setlifetime(struct seclifetime *, uint16_t);
742 static struct mbuf *key_setkey(struct seckey *, uint16_t);
743 
744 static void spdcache_init(void);
745 static void spdcache_clear(void);
746 static struct spdcache_entry *spdcache_entry_alloc(
747 	const struct secpolicyindex *spidx,
748 	struct secpolicy *policy);
749 static void spdcache_entry_free(struct spdcache_entry *entry);
750 #ifdef VIMAGE
751 static void spdcache_destroy(void);
752 #endif
753 
754 #define	DBG_IPSEC_INITREF(t, p)	do {				\
755 	refcount_init(&(p)->refcnt, 1);				\
756 	KEYDBG(KEY_STAMP,					\
757 	    printf("%s: Initialize refcnt %s(%p) = %u\n",	\
758 	    __func__, #t, (p), (p)->refcnt));			\
759 } while (0)
760 #define	DBG_IPSEC_ADDREF(t, p)	do {				\
761 	refcount_acquire(&(p)->refcnt);				\
762 	KEYDBG(KEY_STAMP,					\
763 	    printf("%s: Acquire refcnt %s(%p) -> %u\n",		\
764 	    __func__, #t, (p), (p)->refcnt));			\
765 } while (0)
766 #define	DBG_IPSEC_DELREF(t, p)	do {				\
767 	KEYDBG(KEY_STAMP,					\
768 	    printf("%s: Release refcnt %s(%p) -> %u\n",		\
769 	    __func__, #t, (p), (p)->refcnt - 1));		\
770 	refcount_release(&(p)->refcnt);				\
771 } while (0)
772 
773 #define	IPSEC_INITREF(t, p)	refcount_init(&(p)->refcnt, 1)
774 #define	IPSEC_ADDREF(t, p)	refcount_acquire(&(p)->refcnt)
775 #define	IPSEC_DELREF(t, p)	refcount_release(&(p)->refcnt)
776 
777 #define	SP_INITREF(p)	IPSEC_INITREF(SP, p)
778 #define	SP_ADDREF(p)	IPSEC_ADDREF(SP, p)
779 #define	SP_DELREF(p)	IPSEC_DELREF(SP, p)
780 
781 #define	SAH_INITREF(p)	IPSEC_INITREF(SAH, p)
782 #define	SAH_ADDREF(p)	IPSEC_ADDREF(SAH, p)
783 #define	SAH_DELREF(p)	IPSEC_DELREF(SAH, p)
784 
785 #define	SAV_INITREF(p)	IPSEC_INITREF(SAV, p)
786 #define	SAV_ADDREF(p)	IPSEC_ADDREF(SAV, p)
787 #define	SAV_DELREF(p)	IPSEC_DELREF(SAV, p)
788 
789 /*
790  * Update the refcnt while holding the SPTREE lock.
791  */
792 void
793 key_addref(struct secpolicy *sp)
794 {
795 
796 	SP_ADDREF(sp);
797 }
798 
799 /*
800  * Return 0 when there are known to be no SP's for the specified
801  * direction.  Otherwise return 1.  This is used by IPsec code
802  * to optimize performance.
803  */
804 int
805 key_havesp(u_int dir)
806 {
807 
808 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
809 		("invalid direction %u", dir));
810 	return (TAILQ_FIRST(&V_sptree[dir]) != NULL);
811 }
812 
813 int
814 key_havesp_any(void)
815 {
816 
817 	return (V_spd_size != 0);
818 }
819 
820 /*
821  * Allocate a single mbuf with a buffer of the desired length.  The buffer is
822  * pre-zeroed to help ensure that uninitialized pad bytes are not leaked.
823  */
824 static struct mbuf *
825 key_mget(u_int len)
826 {
827 	struct mbuf *m;
828 
829 	KASSERT(len <= MCLBYTES,
830 	    ("%s: invalid buffer length %u", __func__, len));
831 
832 	m = m_get2(len, M_NOWAIT, MT_DATA, M_PKTHDR);
833 	if (m == NULL)
834 		return (NULL);
835 	memset(mtod(m, void *), 0, len);
836 	return (m);
837 }
838 
839 /* %%% IPsec policy management */
840 /*
841  * Return current SPDB generation.
842  */
843 uint32_t
844 key_getspgen(void)
845 {
846 
847 	return (V_sp_genid);
848 }
849 
850 void
851 key_bumpspgen(void)
852 {
853 
854 	V_sp_genid++;
855 }
856 
857 static int
858 key_checksockaddrs(struct sockaddr *src, struct sockaddr *dst)
859 {
860 
861 	/* family match */
862 	if (src->sa_family != dst->sa_family)
863 		return (EINVAL);
864 	/* sa_len match */
865 	if (src->sa_len != dst->sa_len)
866 		return (EINVAL);
867 	switch (src->sa_family) {
868 #ifdef INET
869 	case AF_INET:
870 		if (src->sa_len != sizeof(struct sockaddr_in))
871 			return (EINVAL);
872 		break;
873 #endif
874 #ifdef INET6
875 	case AF_INET6:
876 		if (src->sa_len != sizeof(struct sockaddr_in6))
877 			return (EINVAL);
878 		break;
879 #endif
880 	default:
881 		return (EAFNOSUPPORT);
882 	}
883 	return (0);
884 }
885 
886 struct secpolicy *
887 key_do_allocsp(struct secpolicyindex *spidx, u_int dir)
888 {
889 	SPTREE_RLOCK_TRACKER;
890 	struct secpolicy *sp;
891 
892 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
893 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
894 		("invalid direction %u", dir));
895 
896 	SPTREE_RLOCK();
897 	TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
898 		if (key_cmpspidx_withmask(&sp->spidx, spidx)) {
899 			SP_ADDREF(sp);
900 			break;
901 		}
902 	}
903 	SPTREE_RUNLOCK();
904 	return (sp);
905 }
906 
907 /*
908  * allocating a SP for OUTBOUND or INBOUND packet.
909  * Must call key_freesp() later.
910  * OUT:	NULL:	not found
911  *	others:	found and return the pointer.
912  */
913 struct secpolicy *
914 key_allocsp(struct secpolicyindex *spidx, u_int dir)
915 {
916 	struct spdcache_entry *entry, *lastentry, *tmpentry;
917 	struct secpolicy *sp;
918 	uint32_t hashv;
919 	time_t ts;
920 	int nb_entries;
921 
922 	if (!SPDCACHE_ACTIVE()) {
923 		sp = key_do_allocsp(spidx, dir);
924 		goto out;
925 	}
926 
927 	hashv = SPDCACHE_HASHVAL(spidx);
928 	SPDCACHE_LOCK(hashv);
929 	nb_entries = 0;
930 	LIST_FOREACH_SAFE(entry, &V_spdcachehashtbl[hashv], chain, tmpentry) {
931 		/* Removed outdated entries */
932 		if (entry->sp != NULL &&
933 		    entry->sp->state == IPSEC_SPSTATE_DEAD) {
934 			LIST_REMOVE(entry, chain);
935 			spdcache_entry_free(entry);
936 			continue;
937 		}
938 
939 		nb_entries++;
940 		if (!key_cmpspidx_exactly(&entry->spidx, spidx)) {
941 			lastentry = entry;
942 			continue;
943 		}
944 
945 		sp = entry->sp;
946 		if (entry->sp != NULL)
947 			SP_ADDREF(sp);
948 
949 		/* IPSECSTAT_INC(ips_spdcache_hits); */
950 
951 		SPDCACHE_UNLOCK(hashv);
952 		goto out;
953 	}
954 
955 	/* IPSECSTAT_INC(ips_spdcache_misses); */
956 
957 	sp = key_do_allocsp(spidx, dir);
958 	entry = spdcache_entry_alloc(spidx, sp);
959 	if (entry != NULL) {
960 		if (nb_entries >= SPDCACHE_MAX_ENTRIES_PER_HASH) {
961 			LIST_REMOVE(lastentry, chain);
962 			spdcache_entry_free(lastentry);
963 		}
964 
965 		LIST_INSERT_HEAD(&V_spdcachehashtbl[hashv], entry, chain);
966 	}
967 
968 	SPDCACHE_UNLOCK(hashv);
969 
970 out:
971 	if (sp != NULL) {	/* found a SPD entry */
972 		ts = time_second;
973 		if (__predict_false(sp->lastused != ts))
974 			sp->lastused = ts;
975 		KEYDBG(IPSEC_STAMP,
976 		    printf("%s: return SP(%p)\n", __func__, sp));
977 		KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
978 	} else {
979 		KEYDBG(IPSEC_DATA,
980 		    printf("%s: lookup failed for ", __func__);
981 		    kdebug_secpolicyindex(spidx, NULL));
982 	}
983 	return (sp);
984 }
985 
986 /*
987  * Allocating an SA entry for an *INBOUND* or *OUTBOUND* TCP packet, signed
988  * or should be signed by MD5 signature.
989  * We don't use key_allocsa() for such lookups, because we don't know SPI.
990  * Unlike ESP and AH protocols, SPI isn't transmitted in the TCP header with
991  * signed packet. We use SADB only as storage for password.
992  * OUT:	positive:	corresponding SA for given saidx found.
993  *	NULL:		SA not found
994  */
995 struct secasvar *
996 key_allocsa_tcpmd5(struct secasindex *saidx)
997 {
998 	SAHTREE_RLOCK_TRACKER;
999 	struct secashead *sah;
1000 	struct secasvar *sav;
1001 
1002 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP,
1003 	    ("unexpected security protocol %u", saidx->proto));
1004 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TCPMD5,
1005 	    ("unexpected mode %u", saidx->mode));
1006 
1007 	SAHTREE_RLOCK();
1008 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
1009 		KEYDBG(IPSEC_DUMP,
1010 		    printf("%s: checking SAH\n", __func__);
1011 		    kdebug_secash(sah, "  "));
1012 		if (sah->saidx.proto != IPPROTO_TCP)
1013 			continue;
1014 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
1015 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
1016 			break;
1017 	}
1018 	if (sah != NULL) {
1019 		if (V_key_preferred_oldsa)
1020 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1021 		else
1022 			sav = TAILQ_FIRST(&sah->savtree_alive);
1023 		if (sav != NULL)
1024 			SAV_ADDREF(sav);
1025 	} else
1026 		sav = NULL;
1027 	SAHTREE_RUNLOCK();
1028 
1029 	if (sav != NULL) {
1030 		KEYDBG(IPSEC_STAMP,
1031 		    printf("%s: return SA(%p)\n", __func__, sav));
1032 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1033 	} else {
1034 		KEYDBG(IPSEC_STAMP,
1035 		    printf("%s: SA not found\n", __func__));
1036 		KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1037 	}
1038 	return (sav);
1039 }
1040 
1041 /*
1042  * Allocating an SA entry for an *OUTBOUND* packet.
1043  * OUT:	positive:	corresponding SA for given saidx found.
1044  *	NULL:		SA not found, but will be acquired, check *error
1045  *			for acquiring status.
1046  */
1047 struct secasvar *
1048 key_allocsa_policy(struct secpolicy *sp, const struct secasindex *saidx,
1049     int *error)
1050 {
1051 	SAHTREE_RLOCK_TRACKER;
1052 	struct secashead *sah;
1053 	struct secasvar *sav;
1054 
1055 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
1056 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TRANSPORT ||
1057 		saidx->mode == IPSEC_MODE_TUNNEL,
1058 		("unexpected policy %u", saidx->mode));
1059 
1060 	/*
1061 	 * We check new SA in the IPsec request because a different
1062 	 * SA may be involved each time this request is checked, either
1063 	 * because new SAs are being configured, or this request is
1064 	 * associated with an unconnected datagram socket, or this request
1065 	 * is associated with a system default policy.
1066 	 */
1067 	SAHTREE_RLOCK();
1068 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
1069 		KEYDBG(IPSEC_DUMP,
1070 		    printf("%s: checking SAH\n", __func__);
1071 		    kdebug_secash(sah, "  "));
1072 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID))
1073 			break;
1074 	}
1075 	if (sah != NULL) {
1076 		/*
1077 		 * Allocate the oldest SA available according to
1078 		 * draft-jenkins-ipsec-rekeying-03.
1079 		 */
1080 		if (V_key_preferred_oldsa)
1081 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1082 		else
1083 			sav = TAILQ_FIRST(&sah->savtree_alive);
1084 		if (sav != NULL)
1085 			SAV_ADDREF(sav);
1086 	} else
1087 		sav = NULL;
1088 	SAHTREE_RUNLOCK();
1089 
1090 	if (sav != NULL) {
1091 		*error = 0;
1092 		KEYDBG(IPSEC_STAMP,
1093 		    printf("%s: chosen SA(%p) for SP(%p)\n", __func__,
1094 			sav, sp));
1095 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1096 		return (sav); /* return referenced SA */
1097 	}
1098 
1099 	/* there is no SA */
1100 	*error = key_acquire(saidx, sp);
1101 	if ((*error) != 0)
1102 		ipseclog((LOG_DEBUG,
1103 		    "%s: error %d returned from key_acquire()\n",
1104 			__func__, *error));
1105 	KEYDBG(IPSEC_STAMP,
1106 	    printf("%s: acquire SA for SP(%p), error %d\n",
1107 		__func__, sp, *error));
1108 	KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1109 	return (NULL);
1110 }
1111 
1112 /*
1113  * allocating a usable SA entry for a *INBOUND* packet.
1114  * Must call key_freesav() later.
1115  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
1116  *	NULL:		not found, or error occurred.
1117  *
1118  * According to RFC 2401 SA is uniquely identified by a triple SPI,
1119  * destination address, and security protocol. But according to RFC 4301,
1120  * SPI by itself suffices to specify an SA.
1121  *
1122  * Note that, however, we do need to keep source address in IPsec SA.
1123  * IKE specification and PF_KEY specification do assume that we
1124  * keep source address in IPsec SA.  We see a tricky situation here.
1125  */
1126 struct secasvar *
1127 key_allocsa(union sockaddr_union *dst, uint8_t proto, uint32_t spi)
1128 {
1129 	SAHTREE_RLOCK_TRACKER;
1130 	struct secasvar *sav;
1131 
1132 	IPSEC_ASSERT(proto == IPPROTO_ESP || proto == IPPROTO_AH ||
1133 	    proto == IPPROTO_IPCOMP, ("unexpected security protocol %u",
1134 	    proto));
1135 
1136 	SAHTREE_RLOCK();
1137 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
1138 		if (sav->spi == spi)
1139 			break;
1140 	}
1141 	/*
1142 	 * We use single SPI namespace for all protocols, so it is
1143 	 * impossible to have SPI duplicates in the SAVHASH.
1144 	 */
1145 	if (sav != NULL) {
1146 		if (sav->state != SADB_SASTATE_LARVAL &&
1147 		    sav->sah->saidx.proto == proto &&
1148 		    key_sockaddrcmp(&dst->sa,
1149 			&sav->sah->saidx.dst.sa, 0) == 0)
1150 			SAV_ADDREF(sav);
1151 		else
1152 			sav = NULL;
1153 	}
1154 	SAHTREE_RUNLOCK();
1155 
1156 	if (sav == NULL) {
1157 		KEYDBG(IPSEC_STAMP,
1158 		    char buf[IPSEC_ADDRSTRLEN];
1159 		    printf("%s: SA not found for spi %u proto %u dst %s\n",
1160 			__func__, ntohl(spi), proto, ipsec_address(dst, buf,
1161 			sizeof(buf))));
1162 	} else {
1163 		KEYDBG(IPSEC_STAMP,
1164 		    printf("%s: return SA(%p)\n", __func__, sav));
1165 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1166 	}
1167 	return (sav);
1168 }
1169 
1170 struct secasvar *
1171 key_allocsa_tunnel(union sockaddr_union *src, union sockaddr_union *dst,
1172     uint8_t proto)
1173 {
1174 	SAHTREE_RLOCK_TRACKER;
1175 	struct secasindex saidx;
1176 	struct secashead *sah;
1177 	struct secasvar *sav;
1178 
1179 	IPSEC_ASSERT(src != NULL, ("null src address"));
1180 	IPSEC_ASSERT(dst != NULL, ("null dst address"));
1181 
1182 	KEY_SETSECASIDX(proto, IPSEC_MODE_TUNNEL, 0, &src->sa,
1183 	    &dst->sa, &saidx);
1184 
1185 	sav = NULL;
1186 	SAHTREE_RLOCK();
1187 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
1188 		if (IPSEC_MODE_TUNNEL != sah->saidx.mode)
1189 			continue;
1190 		if (proto != sah->saidx.proto)
1191 			continue;
1192 		if (key_sockaddrcmp(&src->sa, &sah->saidx.src.sa, 0) != 0)
1193 			continue;
1194 		if (key_sockaddrcmp(&dst->sa, &sah->saidx.dst.sa, 0) != 0)
1195 			continue;
1196 		/* XXXAE: is key_preferred_oldsa reasonably?*/
1197 		if (V_key_preferred_oldsa)
1198 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1199 		else
1200 			sav = TAILQ_FIRST(&sah->savtree_alive);
1201 		if (sav != NULL) {
1202 			SAV_ADDREF(sav);
1203 			break;
1204 		}
1205 	}
1206 	SAHTREE_RUNLOCK();
1207 	KEYDBG(IPSEC_STAMP,
1208 	    printf("%s: return SA(%p)\n", __func__, sav));
1209 	if (sav != NULL)
1210 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1211 	return (sav);
1212 }
1213 
1214 /*
1215  * Must be called after calling key_allocsp().
1216  */
1217 void
1218 key_freesp(struct secpolicy **spp)
1219 {
1220 	struct secpolicy *sp = *spp;
1221 
1222 	IPSEC_ASSERT(sp != NULL, ("null sp"));
1223 	if (SP_DELREF(sp) == 0)
1224 		return;
1225 
1226 	KEYDBG(IPSEC_STAMP,
1227 	    printf("%s: last reference to SP(%p)\n", __func__, sp));
1228 	KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1229 
1230 	*spp = NULL;
1231 	while (sp->tcount > 0)
1232 		ipsec_delisr(sp->req[--sp->tcount]);
1233 	free(sp, M_IPSEC_SP);
1234 }
1235 
1236 static void
1237 key_unlink(struct secpolicy *sp)
1238 {
1239 	SPTREE_WLOCK();
1240 	key_detach(sp);
1241 	SPTREE_WUNLOCK();
1242 	if (SPDCACHE_ENABLED())
1243 		spdcache_clear();
1244 	key_freesp(&sp);
1245 }
1246 
1247 static void
1248 key_detach(struct secpolicy *sp)
1249 {
1250 	IPSEC_ASSERT(sp->spidx.dir == IPSEC_DIR_INBOUND ||
1251 	    sp->spidx.dir == IPSEC_DIR_OUTBOUND,
1252 	    ("invalid direction %u", sp->spidx.dir));
1253 	SPTREE_WLOCK_ASSERT();
1254 
1255 	KEYDBG(KEY_STAMP,
1256 	    printf("%s: SP(%p)\n", __func__, sp));
1257 	if (sp->state != IPSEC_SPSTATE_ALIVE) {
1258 		/* SP is already unlinked */
1259 		return;
1260 	}
1261 	sp->state = IPSEC_SPSTATE_DEAD;
1262 	TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
1263 	V_spd_size--;
1264 	LIST_REMOVE(sp, idhash);
1265 	V_sp_genid++;
1266 }
1267 
1268 /*
1269  * insert a secpolicy into the SP database. Lower priorities first
1270  */
1271 static void
1272 key_insertsp(struct secpolicy *newsp)
1273 {
1274 	struct secpolicy *sp;
1275 
1276 	SPTREE_WLOCK_ASSERT();
1277 	TAILQ_FOREACH(sp, &V_sptree[newsp->spidx.dir], chain) {
1278 		if (newsp->priority < sp->priority) {
1279 			TAILQ_INSERT_BEFORE(sp, newsp, chain);
1280 			goto done;
1281 		}
1282 	}
1283 	TAILQ_INSERT_TAIL(&V_sptree[newsp->spidx.dir], newsp, chain);
1284 done:
1285 	LIST_INSERT_HEAD(SPHASH_HASH(newsp->id), newsp, idhash);
1286 	newsp->state = IPSEC_SPSTATE_ALIVE;
1287 	V_spd_size++;
1288 	V_sp_genid++;
1289 }
1290 
1291 /*
1292  * Insert a bunch of VTI secpolicies into the SPDB.
1293  * We keep VTI policies in the separate list due to following reasons:
1294  * 1) they should be immutable to user's or some deamon's attempts to
1295  *    delete. The only way delete such policies - destroy or unconfigure
1296  *    corresponding virtual inteface.
1297  * 2) such policies have traffic selector that matches all traffic per
1298  *    address family.
1299  * Since all VTI policies have the same priority, we don't care about
1300  * policies order.
1301  */
1302 int
1303 key_register_ifnet(struct secpolicy **spp, u_int count)
1304 {
1305 	struct mbuf *m;
1306 	u_int i;
1307 
1308 	SPTREE_WLOCK();
1309 	/*
1310 	 * First of try to acquire id for each SP.
1311 	 */
1312 	for (i = 0; i < count; i++) {
1313 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1314 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1315 		    ("invalid direction %u", spp[i]->spidx.dir));
1316 
1317 		if ((spp[i]->id = key_getnewspid()) == 0) {
1318 			SPTREE_WUNLOCK();
1319 			return (EAGAIN);
1320 		}
1321 	}
1322 	for (i = 0; i < count; i++) {
1323 		TAILQ_INSERT_TAIL(&V_sptree_ifnet[spp[i]->spidx.dir],
1324 		    spp[i], chain);
1325 		/*
1326 		 * NOTE: despite the fact that we keep VTI SP in the
1327 		 * separate list, SPHASH contains policies from both
1328 		 * sources. Thus SADB_X_SPDGET will correctly return
1329 		 * SP by id, because it uses SPHASH for lookups.
1330 		 */
1331 		LIST_INSERT_HEAD(SPHASH_HASH(spp[i]->id), spp[i], idhash);
1332 		spp[i]->state = IPSEC_SPSTATE_IFNET;
1333 	}
1334 	SPTREE_WUNLOCK();
1335 	/*
1336 	 * Notify user processes about new SP.
1337 	 */
1338 	for (i = 0; i < count; i++) {
1339 		m = key_setdumpsp(spp[i], SADB_X_SPDADD, 0, 0);
1340 		if (m != NULL)
1341 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1342 	}
1343 	return (0);
1344 }
1345 
1346 void
1347 key_unregister_ifnet(struct secpolicy **spp, u_int count)
1348 {
1349 	struct mbuf *m;
1350 	u_int i;
1351 
1352 	SPTREE_WLOCK();
1353 	for (i = 0; i < count; i++) {
1354 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1355 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1356 		    ("invalid direction %u", spp[i]->spidx.dir));
1357 
1358 		if (spp[i]->state != IPSEC_SPSTATE_IFNET)
1359 			continue;
1360 		spp[i]->state = IPSEC_SPSTATE_DEAD;
1361 		TAILQ_REMOVE(&V_sptree_ifnet[spp[i]->spidx.dir],
1362 		    spp[i], chain);
1363 		V_spd_size--;
1364 		LIST_REMOVE(spp[i], idhash);
1365 	}
1366 	SPTREE_WUNLOCK();
1367 	if (SPDCACHE_ENABLED())
1368 		spdcache_clear();
1369 
1370 	for (i = 0; i < count; i++) {
1371 		m = key_setdumpsp(spp[i], SADB_X_SPDDELETE, 0, 0);
1372 		if (m != NULL)
1373 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1374 	}
1375 }
1376 
1377 /*
1378  * Must be called after calling key_allocsa().
1379  * This function is called by key_freesp() to free some SA allocated
1380  * for a policy.
1381  */
1382 void
1383 key_freesav(struct secasvar **psav)
1384 {
1385 	struct secasvar *sav = *psav;
1386 
1387 	IPSEC_ASSERT(sav != NULL, ("null sav"));
1388 	CURVNET_ASSERT_SET();
1389 	if (SAV_DELREF(sav) == 0)
1390 		return;
1391 
1392 	KEYDBG(IPSEC_STAMP,
1393 	    printf("%s: last reference to SA(%p)\n", __func__, sav));
1394 
1395 	*psav = NULL;
1396 	key_delsav(sav);
1397 }
1398 
1399 /*
1400  * Unlink SA from SAH and SPI hash under SAHTREE_WLOCK.
1401  * Expect that SA has extra reference due to lookup.
1402  * Release this references, also release SAH reference after unlink.
1403  */
1404 static void
1405 key_unlinksav(struct secasvar *sav)
1406 {
1407 	struct secashead *sah;
1408 
1409 	KEYDBG(KEY_STAMP,
1410 	    printf("%s: SA(%p)\n", __func__, sav));
1411 
1412 	CURVNET_ASSERT_SET();
1413 	SAHTREE_UNLOCK_ASSERT();
1414 	SAHTREE_WLOCK();
1415 	if (sav->state == SADB_SASTATE_DEAD) {
1416 		/* SA is already unlinked */
1417 		SAHTREE_WUNLOCK();
1418 		return;
1419 	}
1420 	/* Unlink from SAH */
1421 	if (sav->state == SADB_SASTATE_LARVAL)
1422 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
1423 	else
1424 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
1425 	/* Unlink from SPI hash */
1426 	LIST_REMOVE(sav, spihash);
1427 	sav->state = SADB_SASTATE_DEAD;
1428 	sah = sav->sah;
1429 	SAHTREE_WUNLOCK();
1430 	key_freesav(&sav);
1431 	/* Since we are unlinked, release reference to SAH */
1432 	key_freesah(&sah);
1433 }
1434 
1435 /* %%% SPD management */
1436 /*
1437  * search SPD
1438  * OUT:	NULL	: not found
1439  *	others	: found, pointer to a SP.
1440  */
1441 static struct secpolicy *
1442 key_getsp(struct secpolicyindex *spidx)
1443 {
1444 	SPTREE_RLOCK_TRACKER;
1445 	struct secpolicy *sp;
1446 
1447 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
1448 
1449 	SPTREE_RLOCK();
1450 	TAILQ_FOREACH(sp, &V_sptree[spidx->dir], chain) {
1451 		if (key_cmpspidx_exactly(spidx, &sp->spidx)) {
1452 			SP_ADDREF(sp);
1453 			break;
1454 		}
1455 	}
1456 	SPTREE_RUNLOCK();
1457 
1458 	return sp;
1459 }
1460 
1461 /*
1462  * get SP by index.
1463  * OUT:	NULL	: not found
1464  *	others	: found, pointer to referenced SP.
1465  */
1466 static struct secpolicy *
1467 key_getspbyid(uint32_t id)
1468 {
1469 	SPTREE_RLOCK_TRACKER;
1470 	struct secpolicy *sp;
1471 
1472 	SPTREE_RLOCK();
1473 	LIST_FOREACH(sp, SPHASH_HASH(id), idhash) {
1474 		if (sp->id == id) {
1475 			SP_ADDREF(sp);
1476 			break;
1477 		}
1478 	}
1479 	SPTREE_RUNLOCK();
1480 	return (sp);
1481 }
1482 
1483 struct secpolicy *
1484 key_newsp(void)
1485 {
1486 	struct secpolicy *sp;
1487 
1488 	sp = malloc(sizeof(*sp), M_IPSEC_SP, M_NOWAIT | M_ZERO);
1489 	if (sp != NULL)
1490 		SP_INITREF(sp);
1491 	return (sp);
1492 }
1493 
1494 struct ipsecrequest *
1495 ipsec_newisr(void)
1496 {
1497 
1498 	return (malloc(sizeof(struct ipsecrequest), M_IPSEC_SR,
1499 	    M_NOWAIT | M_ZERO));
1500 }
1501 
1502 void
1503 ipsec_delisr(struct ipsecrequest *p)
1504 {
1505 
1506 	free(p, M_IPSEC_SR);
1507 }
1508 
1509 /*
1510  * create secpolicy structure from sadb_x_policy structure.
1511  * NOTE: `state', `secpolicyindex' and 'id' in secpolicy structure
1512  * are not set, so must be set properly later.
1513  */
1514 struct secpolicy *
1515 key_msg2sp(struct sadb_x_policy *xpl0, size_t len, int *error)
1516 {
1517 	struct secpolicy *newsp;
1518 
1519 	IPSEC_ASSERT(xpl0 != NULL, ("null xpl0"));
1520 	IPSEC_ASSERT(len >= sizeof(*xpl0), ("policy too short: %zu", len));
1521 
1522 	if (len != PFKEY_EXTLEN(xpl0)) {
1523 		ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n", __func__));
1524 		*error = EINVAL;
1525 		return NULL;
1526 	}
1527 
1528 	if ((newsp = key_newsp()) == NULL) {
1529 		*error = ENOBUFS;
1530 		return NULL;
1531 	}
1532 
1533 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1534 	newsp->policy = xpl0->sadb_x_policy_type;
1535 	newsp->priority = xpl0->sadb_x_policy_priority;
1536 	newsp->tcount = 0;
1537 
1538 	/* check policy */
1539 	switch (xpl0->sadb_x_policy_type) {
1540 	case IPSEC_POLICY_DISCARD:
1541 	case IPSEC_POLICY_NONE:
1542 	case IPSEC_POLICY_ENTRUST:
1543 	case IPSEC_POLICY_BYPASS:
1544 		break;
1545 
1546 	case IPSEC_POLICY_IPSEC:
1547 	    {
1548 		struct sadb_x_ipsecrequest *xisr;
1549 		struct ipsecrequest *isr;
1550 		int tlen;
1551 
1552 		/* validity check */
1553 		if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1554 			ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n",
1555 				__func__));
1556 			key_freesp(&newsp);
1557 			*error = EINVAL;
1558 			return NULL;
1559 		}
1560 
1561 		tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1562 		xisr = (struct sadb_x_ipsecrequest *)(xpl0 + 1);
1563 
1564 		while (tlen > 0) {
1565 			/* length check */
1566 			if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr) ||
1567 			    xisr->sadb_x_ipsecrequest_len > tlen) {
1568 				ipseclog((LOG_DEBUG, "%s: invalid ipsecrequest "
1569 					"length.\n", __func__));
1570 				key_freesp(&newsp);
1571 				*error = EINVAL;
1572 				return NULL;
1573 			}
1574 
1575 			if (newsp->tcount >= IPSEC_MAXREQ) {
1576 				ipseclog((LOG_DEBUG,
1577 				    "%s: too many ipsecrequests.\n",
1578 				    __func__));
1579 				key_freesp(&newsp);
1580 				*error = EINVAL;
1581 				return (NULL);
1582 			}
1583 
1584 			/* allocate request buffer */
1585 			/* NB: data structure is zero'd */
1586 			isr = ipsec_newisr();
1587 			if (isr == NULL) {
1588 				ipseclog((LOG_DEBUG,
1589 				    "%s: No more memory.\n", __func__));
1590 				key_freesp(&newsp);
1591 				*error = ENOBUFS;
1592 				return NULL;
1593 			}
1594 
1595 			newsp->req[newsp->tcount++] = isr;
1596 
1597 			/* set values */
1598 			switch (xisr->sadb_x_ipsecrequest_proto) {
1599 			case IPPROTO_ESP:
1600 			case IPPROTO_AH:
1601 			case IPPROTO_IPCOMP:
1602 				break;
1603 			default:
1604 				ipseclog((LOG_DEBUG,
1605 				    "%s: invalid proto type=%u\n", __func__,
1606 				    xisr->sadb_x_ipsecrequest_proto));
1607 				key_freesp(&newsp);
1608 				*error = EPROTONOSUPPORT;
1609 				return NULL;
1610 			}
1611 			isr->saidx.proto =
1612 			    (uint8_t)xisr->sadb_x_ipsecrequest_proto;
1613 
1614 			switch (xisr->sadb_x_ipsecrequest_mode) {
1615 			case IPSEC_MODE_TRANSPORT:
1616 			case IPSEC_MODE_TUNNEL:
1617 				break;
1618 			case IPSEC_MODE_ANY:
1619 			default:
1620 				ipseclog((LOG_DEBUG,
1621 				    "%s: invalid mode=%u\n", __func__,
1622 				    xisr->sadb_x_ipsecrequest_mode));
1623 				key_freesp(&newsp);
1624 				*error = EINVAL;
1625 				return NULL;
1626 			}
1627 			isr->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1628 
1629 			switch (xisr->sadb_x_ipsecrequest_level) {
1630 			case IPSEC_LEVEL_DEFAULT:
1631 			case IPSEC_LEVEL_USE:
1632 			case IPSEC_LEVEL_REQUIRE:
1633 				break;
1634 			case IPSEC_LEVEL_UNIQUE:
1635 				/* validity check */
1636 				/*
1637 				 * If range violation of reqid, kernel will
1638 				 * update it, don't refuse it.
1639 				 */
1640 				if (xisr->sadb_x_ipsecrequest_reqid
1641 						> IPSEC_MANUAL_REQID_MAX) {
1642 					ipseclog((LOG_DEBUG,
1643 					    "%s: reqid=%d range "
1644 					    "violation, updated by kernel.\n",
1645 					    __func__,
1646 					    xisr->sadb_x_ipsecrequest_reqid));
1647 					xisr->sadb_x_ipsecrequest_reqid = 0;
1648 				}
1649 
1650 				/* allocate new reqid id if reqid is zero. */
1651 				if (xisr->sadb_x_ipsecrequest_reqid == 0) {
1652 					u_int32_t reqid;
1653 					if ((reqid = key_newreqid()) == 0) {
1654 						key_freesp(&newsp);
1655 						*error = ENOBUFS;
1656 						return NULL;
1657 					}
1658 					isr->saidx.reqid = reqid;
1659 					xisr->sadb_x_ipsecrequest_reqid = reqid;
1660 				} else {
1661 				/* set it for manual keying. */
1662 					isr->saidx.reqid =
1663 					    xisr->sadb_x_ipsecrequest_reqid;
1664 				}
1665 				break;
1666 
1667 			default:
1668 				ipseclog((LOG_DEBUG, "%s: invalid level=%u\n",
1669 					__func__,
1670 					xisr->sadb_x_ipsecrequest_level));
1671 				key_freesp(&newsp);
1672 				*error = EINVAL;
1673 				return NULL;
1674 			}
1675 			isr->level = xisr->sadb_x_ipsecrequest_level;
1676 
1677 			/* set IP addresses if there */
1678 			if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1679 				struct sockaddr *paddr;
1680 
1681 				len = tlen - sizeof(*xisr);
1682 				paddr = (struct sockaddr *)(xisr + 1);
1683 				/* validity check */
1684 				if (len < sizeof(struct sockaddr) ||
1685 				    len < 2 * paddr->sa_len ||
1686 				    paddr->sa_len > sizeof(isr->saidx.src)) {
1687 					ipseclog((LOG_DEBUG, "%s: invalid "
1688 						"request address length.\n",
1689 						__func__));
1690 					key_freesp(&newsp);
1691 					*error = EINVAL;
1692 					return NULL;
1693 				}
1694 				/*
1695 				 * Request length should be enough to keep
1696 				 * source and destination addresses.
1697 				 */
1698 				if (xisr->sadb_x_ipsecrequest_len <
1699 				    sizeof(*xisr) + 2 * paddr->sa_len) {
1700 					ipseclog((LOG_DEBUG, "%s: invalid "
1701 					    "ipsecrequest length.\n",
1702 					    __func__));
1703 					key_freesp(&newsp);
1704 					*error = EINVAL;
1705 					return (NULL);
1706 				}
1707 				bcopy(paddr, &isr->saidx.src, paddr->sa_len);
1708 				paddr = (struct sockaddr *)((caddr_t)paddr +
1709 				    paddr->sa_len);
1710 
1711 				/* validity check */
1712 				if (paddr->sa_len !=
1713 				    isr->saidx.src.sa.sa_len) {
1714 					ipseclog((LOG_DEBUG, "%s: invalid "
1715 						"request address length.\n",
1716 						__func__));
1717 					key_freesp(&newsp);
1718 					*error = EINVAL;
1719 					return NULL;
1720 				}
1721 				/* AF family should match */
1722 				if (paddr->sa_family !=
1723 				    isr->saidx.src.sa.sa_family) {
1724 					ipseclog((LOG_DEBUG, "%s: address "
1725 					    "family doesn't match.\n",
1726 						__func__));
1727 					key_freesp(&newsp);
1728 					*error = EINVAL;
1729 					return (NULL);
1730 				}
1731 				bcopy(paddr, &isr->saidx.dst, paddr->sa_len);
1732 			} else {
1733 				/*
1734 				 * Addresses for TUNNEL mode requests are
1735 				 * mandatory.
1736 				 */
1737 				if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1738 					ipseclog((LOG_DEBUG, "%s: missing "
1739 					    "request addresses.\n", __func__));
1740 					key_freesp(&newsp);
1741 					*error = EINVAL;
1742 					return (NULL);
1743 				}
1744 			}
1745 			tlen -= xisr->sadb_x_ipsecrequest_len;
1746 
1747 			/* validity check */
1748 			if (tlen < 0) {
1749 				ipseclog((LOG_DEBUG, "%s: becoming tlen < 0.\n",
1750 					__func__));
1751 				key_freesp(&newsp);
1752 				*error = EINVAL;
1753 				return NULL;
1754 			}
1755 
1756 			xisr = (struct sadb_x_ipsecrequest *)((caddr_t)xisr
1757 			                 + xisr->sadb_x_ipsecrequest_len);
1758 		}
1759 		/* XXXAE: LARVAL SP */
1760 		if (newsp->tcount < 1) {
1761 			ipseclog((LOG_DEBUG, "%s: valid IPSEC transforms "
1762 			    "not found.\n", __func__));
1763 			key_freesp(&newsp);
1764 			*error = EINVAL;
1765 			return (NULL);
1766 		}
1767 	    }
1768 		break;
1769 	default:
1770 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
1771 		key_freesp(&newsp);
1772 		*error = EINVAL;
1773 		return NULL;
1774 	}
1775 
1776 	*error = 0;
1777 	return (newsp);
1778 }
1779 
1780 uint32_t
1781 key_newreqid(void)
1782 {
1783 	static uint32_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1784 
1785 	if (auto_reqid == ~0)
1786 		auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1787 	else
1788 		auto_reqid++;
1789 
1790 	/* XXX should be unique check */
1791 	return (auto_reqid);
1792 }
1793 
1794 /*
1795  * copy secpolicy struct to sadb_x_policy structure indicated.
1796  */
1797 static struct mbuf *
1798 key_sp2mbuf(struct secpolicy *sp)
1799 {
1800 	struct mbuf *m;
1801 	size_t tlen;
1802 
1803 	tlen = key_getspreqmsglen(sp);
1804 	m = m_get2(tlen, M_NOWAIT, MT_DATA, 0);
1805 	if (m == NULL)
1806 		return (NULL);
1807 	m_align(m, tlen);
1808 	m->m_len = tlen;
1809 	if (key_sp2msg(sp, m->m_data, &tlen) != 0) {
1810 		m_freem(m);
1811 		return (NULL);
1812 	}
1813 	return (m);
1814 }
1815 
1816 int
1817 key_sp2msg(struct secpolicy *sp, void *request, size_t *len)
1818 {
1819 	struct sadb_x_ipsecrequest *xisr;
1820 	struct sadb_x_policy *xpl;
1821 	struct ipsecrequest *isr;
1822 	size_t xlen, ilen;
1823 	caddr_t p;
1824 	int error, i;
1825 
1826 	IPSEC_ASSERT(sp != NULL, ("null policy"));
1827 
1828 	xlen = sizeof(*xpl);
1829 	if (*len < xlen)
1830 		return (EINVAL);
1831 
1832 	error = 0;
1833 	bzero(request, *len);
1834 	xpl = (struct sadb_x_policy *)request;
1835 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
1836 	xpl->sadb_x_policy_type = sp->policy;
1837 	xpl->sadb_x_policy_dir = sp->spidx.dir;
1838 	xpl->sadb_x_policy_id = sp->id;
1839 	xpl->sadb_x_policy_priority = sp->priority;
1840 	switch (sp->state) {
1841 	case IPSEC_SPSTATE_IFNET:
1842 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_IFNET;
1843 		break;
1844 	case IPSEC_SPSTATE_PCB:
1845 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_PCB;
1846 		break;
1847 	default:
1848 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_GLOBAL;
1849 	}
1850 
1851 	/* if is the policy for ipsec ? */
1852 	if (sp->policy == IPSEC_POLICY_IPSEC) {
1853 		p = (caddr_t)xpl + sizeof(*xpl);
1854 		for (i = 0; i < sp->tcount; i++) {
1855 			isr = sp->req[i];
1856 			ilen = PFKEY_ALIGN8(sizeof(*xisr) +
1857 			    isr->saidx.src.sa.sa_len +
1858 			    isr->saidx.dst.sa.sa_len);
1859 			xlen += ilen;
1860 			if (xlen > *len) {
1861 				error = ENOBUFS;
1862 				/* Calculate needed size */
1863 				continue;
1864 			}
1865 			xisr = (struct sadb_x_ipsecrequest *)p;
1866 			xisr->sadb_x_ipsecrequest_len = ilen;
1867 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
1868 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
1869 			xisr->sadb_x_ipsecrequest_level = isr->level;
1870 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
1871 
1872 			p += sizeof(*xisr);
1873 			bcopy(&isr->saidx.src, p, isr->saidx.src.sa.sa_len);
1874 			p += isr->saidx.src.sa.sa_len;
1875 			bcopy(&isr->saidx.dst, p, isr->saidx.dst.sa.sa_len);
1876 			p += isr->saidx.dst.sa.sa_len;
1877 		}
1878 	}
1879 	xpl->sadb_x_policy_len = PFKEY_UNIT64(xlen);
1880 	if (error == 0)
1881 		*len = xlen;
1882 	else
1883 		*len = sizeof(*xpl);
1884 	return (error);
1885 }
1886 
1887 /* m will not be freed nor modified */
1888 static struct mbuf *
1889 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
1890     int ndeep, int nitem, ...)
1891 {
1892 	va_list ap;
1893 	int idx;
1894 	int i;
1895 	struct mbuf *result = NULL, *n;
1896 	int len;
1897 
1898 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1899 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1900 
1901 	va_start(ap, nitem);
1902 	for (i = 0; i < nitem; i++) {
1903 		idx = va_arg(ap, int);
1904 		if (idx < 0 || idx > SADB_EXT_MAX)
1905 			goto fail;
1906 		/* don't attempt to pull empty extension */
1907 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
1908 			continue;
1909 		if (idx != SADB_EXT_RESERVED  &&
1910 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
1911 			continue;
1912 
1913 		if (idx == SADB_EXT_RESERVED) {
1914 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
1915 
1916 			IPSEC_ASSERT(len <= MHLEN, ("header too big %u", len));
1917 
1918 			MGETHDR(n, M_NOWAIT, MT_DATA);
1919 			if (!n)
1920 				goto fail;
1921 			n->m_len = len;
1922 			n->m_next = NULL;
1923 			m_copydata(m, 0, sizeof(struct sadb_msg),
1924 			    mtod(n, caddr_t));
1925 		} else if (i < ndeep) {
1926 			len = mhp->extlen[idx];
1927 			n = m_get2(len, M_NOWAIT, MT_DATA, 0);
1928 			if (n == NULL)
1929 				goto fail;
1930 			m_align(n, len);
1931 			n->m_len = len;
1932 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
1933 			    mtod(n, caddr_t));
1934 		} else {
1935 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
1936 			    M_NOWAIT);
1937 		}
1938 		if (n == NULL)
1939 			goto fail;
1940 
1941 		if (result)
1942 			m_cat(result, n);
1943 		else
1944 			result = n;
1945 	}
1946 	va_end(ap);
1947 
1948 	if ((result->m_flags & M_PKTHDR) != 0) {
1949 		result->m_pkthdr.len = 0;
1950 		for (n = result; n; n = n->m_next)
1951 			result->m_pkthdr.len += n->m_len;
1952 	}
1953 
1954 	return result;
1955 
1956 fail:
1957 	m_freem(result);
1958 	va_end(ap);
1959 	return NULL;
1960 }
1961 
1962 /*
1963  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
1964  * add an entry to SP database, when received
1965  *   <base, address(SD), (lifetime(H),) policy>
1966  * from the user(?).
1967  * Adding to SP database,
1968  * and send
1969  *   <base, address(SD), (lifetime(H),) policy>
1970  * to the socket which was send.
1971  *
1972  * SPDADD set a unique policy entry.
1973  * SPDSETIDX like SPDADD without a part of policy requests.
1974  * SPDUPDATE replace a unique policy entry.
1975  *
1976  * XXXAE: serialize this in PF_KEY to avoid races.
1977  * m will always be freed.
1978  */
1979 static int
1980 key_spdadd(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
1981 {
1982 	struct secpolicyindex spidx;
1983 	struct sadb_address *src0, *dst0;
1984 	struct sadb_x_policy *xpl0, *xpl;
1985 	struct sadb_lifetime *lft = NULL;
1986 	struct secpolicy *newsp, *oldsp;
1987 	int error;
1988 
1989 	IPSEC_ASSERT(so != NULL, ("null socket"));
1990 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1991 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1992 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
1993 
1994 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
1995 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
1996 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
1997 		ipseclog((LOG_DEBUG,
1998 		    "%s: invalid message: missing required header.\n",
1999 		    __func__));
2000 		return key_senderror(so, m, EINVAL);
2001 	}
2002 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
2003 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
2004 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2005 		ipseclog((LOG_DEBUG,
2006 		    "%s: invalid message: wrong header size.\n", __func__));
2007 		return key_senderror(so, m, EINVAL);
2008 	}
2009 	if (!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD)) {
2010 		if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD)) {
2011 			ipseclog((LOG_DEBUG,
2012 			    "%s: invalid message: wrong header size.\n",
2013 			    __func__));
2014 			return key_senderror(so, m, EINVAL);
2015 		}
2016 		lft = (struct sadb_lifetime *)mhp->ext[SADB_EXT_LIFETIME_HARD];
2017 	}
2018 
2019 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
2020 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
2021 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
2022 
2023 	/* check the direciton */
2024 	switch (xpl0->sadb_x_policy_dir) {
2025 	case IPSEC_DIR_INBOUND:
2026 	case IPSEC_DIR_OUTBOUND:
2027 		break;
2028 	default:
2029 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2030 		return key_senderror(so, m, EINVAL);
2031 	}
2032 	/* key_spdadd() accepts DISCARD, NONE and IPSEC. */
2033 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2034 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2035 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2036 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2037 		return key_senderror(so, m, EINVAL);
2038 	}
2039 
2040 	/* policy requests are mandatory when action is ipsec. */
2041 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
2042 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
2043 		ipseclog((LOG_DEBUG,
2044 		    "%s: policy requests required.\n", __func__));
2045 		return key_senderror(so, m, EINVAL);
2046 	}
2047 
2048 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
2049 	    (struct sockaddr *)(dst0 + 1));
2050 	if (error != 0 ||
2051 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2052 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2053 		return key_senderror(so, m, error);
2054 	}
2055 	/* make secindex */
2056 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2057 	                src0 + 1,
2058 	                dst0 + 1,
2059 	                src0->sadb_address_prefixlen,
2060 	                dst0->sadb_address_prefixlen,
2061 	                src0->sadb_address_proto,
2062 	                &spidx);
2063 	/* Checking there is SP already or not. */
2064 	oldsp = key_getsp(&spidx);
2065 	if (oldsp != NULL) {
2066 		if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2067 			KEYDBG(KEY_STAMP,
2068 			    printf("%s: unlink SP(%p) for SPDUPDATE\n",
2069 				__func__, oldsp));
2070 			KEYDBG(KEY_DATA, kdebug_secpolicy(oldsp));
2071 		} else {
2072 			key_freesp(&oldsp);
2073 			ipseclog((LOG_DEBUG,
2074 			    "%s: a SP entry exists already.\n", __func__));
2075 			return (key_senderror(so, m, EEXIST));
2076 		}
2077 	}
2078 
2079 	/* allocate new SP entry */
2080 	if ((newsp = key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error)) == NULL) {
2081 		if (oldsp != NULL) {
2082 			key_unlink(oldsp);
2083 			key_freesp(&oldsp); /* second for our reference */
2084 		}
2085 		return key_senderror(so, m, error);
2086 	}
2087 
2088 	newsp->lastused = newsp->created = time_second;
2089 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
2090 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
2091 	bcopy(&spidx, &newsp->spidx, sizeof(spidx));
2092 
2093 	SPTREE_WLOCK();
2094 	if ((newsp->id = key_getnewspid()) == 0) {
2095 		if (oldsp != NULL)
2096 			key_detach(oldsp);
2097 		SPTREE_WUNLOCK();
2098 		if (oldsp != NULL) {
2099 			key_freesp(&oldsp); /* first for key_detach */
2100 			IPSEC_ASSERT(oldsp != NULL, ("null oldsp: refcount bug"));
2101 			key_freesp(&oldsp); /* second for our reference */
2102 			if (SPDCACHE_ENABLED()) /* refresh cache because of key_detach */
2103 				spdcache_clear();
2104 		}
2105 		key_freesp(&newsp);
2106 		return key_senderror(so, m, ENOBUFS);
2107 	}
2108 	if (oldsp != NULL)
2109 		key_detach(oldsp);
2110 	key_insertsp(newsp);
2111 	SPTREE_WUNLOCK();
2112 	if (oldsp != NULL) {
2113 		key_freesp(&oldsp); /* first for key_detach */
2114 		IPSEC_ASSERT(oldsp != NULL, ("null oldsp: refcount bug"));
2115 		key_freesp(&oldsp); /* second for our reference */
2116 	}
2117 	if (SPDCACHE_ENABLED())
2118 		spdcache_clear();
2119 	KEYDBG(KEY_STAMP,
2120 	    printf("%s: SP(%p)\n", __func__, newsp));
2121 	KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
2122 
2123     {
2124 	struct mbuf *n, *mpolicy;
2125 	struct sadb_msg *newmsg;
2126 	int off;
2127 
2128 	/* create new sadb_msg to reply. */
2129 	if (lft) {
2130 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
2131 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
2132 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2133 	} else {
2134 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
2135 		    SADB_X_EXT_POLICY,
2136 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2137 	}
2138 	if (!n)
2139 		return key_senderror(so, m, ENOBUFS);
2140 
2141 	if (n->m_len < sizeof(*newmsg)) {
2142 		n = m_pullup(n, sizeof(*newmsg));
2143 		if (!n)
2144 			return key_senderror(so, m, ENOBUFS);
2145 	}
2146 	newmsg = mtod(n, struct sadb_msg *);
2147 	newmsg->sadb_msg_errno = 0;
2148 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2149 
2150 	off = 0;
2151 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
2152 	    sizeof(*xpl), &off);
2153 	if (mpolicy == NULL) {
2154 		/* n is already freed */
2155 		return key_senderror(so, m, ENOBUFS);
2156 	}
2157 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, caddr_t) + off);
2158 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
2159 		m_freem(n);
2160 		return key_senderror(so, m, EINVAL);
2161 	}
2162 	xpl->sadb_x_policy_id = newsp->id;
2163 
2164 	m_freem(m);
2165 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2166     }
2167 }
2168 
2169 /*
2170  * get new policy id.
2171  * OUT:
2172  *	0:	failure.
2173  *	others: success.
2174  */
2175 static uint32_t
2176 key_getnewspid(void)
2177 {
2178 	struct secpolicy *sp;
2179 	uint32_t newid = 0;
2180 	int tries, limit;
2181 
2182 	SPTREE_WLOCK_ASSERT();
2183 
2184 	limit = atomic_load_int(&V_key_spi_trycnt);
2185 	for (tries = 0; tries < limit; tries++) {
2186 		if (V_policy_id == ~0) /* overflowed */
2187 			newid = V_policy_id = 1;
2188 		else
2189 			newid = ++V_policy_id;
2190 		LIST_FOREACH(sp, SPHASH_HASH(newid), idhash) {
2191 			if (sp->id == newid)
2192 				break;
2193 		}
2194 		if (sp == NULL)
2195 			break;
2196 	}
2197 	if (tries == limit || newid == 0) {
2198 		ipseclog((LOG_DEBUG, "%s: failed to allocate policy id.\n",
2199 		    __func__));
2200 		return (0);
2201 	}
2202 	return (newid);
2203 }
2204 
2205 /*
2206  * SADB_SPDDELETE processing
2207  * receive
2208  *   <base, address(SD), policy(*)>
2209  * from the user(?), and set SADB_SASTATE_DEAD,
2210  * and send,
2211  *   <base, address(SD), policy(*)>
2212  * to the ikmpd.
2213  * policy(*) including direction of policy.
2214  *
2215  * m will always be freed.
2216  */
2217 static int
2218 key_spddelete(struct socket *so, struct mbuf *m,
2219     const struct sadb_msghdr *mhp)
2220 {
2221 	struct secpolicyindex spidx;
2222 	struct sadb_address *src0, *dst0;
2223 	struct sadb_x_policy *xpl0;
2224 	struct secpolicy *sp;
2225 
2226 	IPSEC_ASSERT(so != NULL, ("null so"));
2227 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2228 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2229 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2230 
2231 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
2232 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
2233 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
2234 		ipseclog((LOG_DEBUG,
2235 		    "%s: invalid message: missing required header.\n",
2236 		    __func__));
2237 		return key_senderror(so, m, EINVAL);
2238 	}
2239 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
2240 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
2241 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2242 		ipseclog((LOG_DEBUG,
2243 		    "%s: invalid message: wrong header size.\n", __func__));
2244 		return key_senderror(so, m, EINVAL);
2245 	}
2246 
2247 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
2248 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
2249 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
2250 
2251 	/* check the direciton */
2252 	switch (xpl0->sadb_x_policy_dir) {
2253 	case IPSEC_DIR_INBOUND:
2254 	case IPSEC_DIR_OUTBOUND:
2255 		break;
2256 	default:
2257 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2258 		return key_senderror(so, m, EINVAL);
2259 	}
2260 	/* Only DISCARD, NONE and IPSEC are allowed */
2261 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2262 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2263 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2264 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2265 		return key_senderror(so, m, EINVAL);
2266 	}
2267 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
2268 	    (struct sockaddr *)(dst0 + 1)) != 0 ||
2269 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2270 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2271 		return key_senderror(so, m, EINVAL);
2272 	}
2273 	/* make secindex */
2274 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2275 	                src0 + 1,
2276 	                dst0 + 1,
2277 	                src0->sadb_address_prefixlen,
2278 	                dst0->sadb_address_prefixlen,
2279 	                src0->sadb_address_proto,
2280 	                &spidx);
2281 
2282 	/* Is there SP in SPD ? */
2283 	if ((sp = key_getsp(&spidx)) == NULL) {
2284 		ipseclog((LOG_DEBUG, "%s: no SP found.\n", __func__));
2285 		return key_senderror(so, m, EINVAL);
2286 	}
2287 
2288 	/* save policy id to buffer to be returned. */
2289 	xpl0->sadb_x_policy_id = sp->id;
2290 
2291 	KEYDBG(KEY_STAMP,
2292 	    printf("%s: SP(%p)\n", __func__, sp));
2293 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2294 	key_unlink(sp);
2295 	key_freesp(&sp);
2296 
2297     {
2298 	struct mbuf *n;
2299 	struct sadb_msg *newmsg;
2300 
2301 	/* create new sadb_msg to reply. */
2302 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2303 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2304 	if (!n)
2305 		return key_senderror(so, m, ENOBUFS);
2306 
2307 	newmsg = mtod(n, struct sadb_msg *);
2308 	newmsg->sadb_msg_errno = 0;
2309 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2310 
2311 	m_freem(m);
2312 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2313     }
2314 }
2315 
2316 /*
2317  * SADB_SPDDELETE2 processing
2318  * receive
2319  *   <base, policy(*)>
2320  * from the user(?), and set SADB_SASTATE_DEAD,
2321  * and send,
2322  *   <base, policy(*)>
2323  * to the ikmpd.
2324  * policy(*) including direction of policy.
2325  *
2326  * m will always be freed.
2327  */
2328 static int
2329 key_spddelete2(struct socket *so, struct mbuf *m,
2330     const struct sadb_msghdr *mhp)
2331 {
2332 	struct secpolicy *sp;
2333 	uint32_t id;
2334 
2335 	IPSEC_ASSERT(so != NULL, ("null socket"));
2336 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2337 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2338 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2339 
2340 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2341 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2342 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2343 		    __func__));
2344 		return key_senderror(so, m, EINVAL);
2345 	}
2346 
2347 	id = ((struct sadb_x_policy *)
2348 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2349 
2350 	/* Is there SP in SPD ? */
2351 	if ((sp = key_getspbyid(id)) == NULL) {
2352 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2353 		    __func__, id));
2354 		return key_senderror(so, m, EINVAL);
2355 	}
2356 
2357 	KEYDBG(KEY_STAMP,
2358 	    printf("%s: SP(%p)\n", __func__, sp));
2359 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2360 	key_unlink(sp);
2361 	if (sp->state != IPSEC_SPSTATE_DEAD) {
2362 		ipseclog((LOG_DEBUG, "%s: failed to delete SP with id %u.\n",
2363 		    __func__, id));
2364 		key_freesp(&sp);
2365 		return (key_senderror(so, m, EACCES));
2366 	}
2367 	key_freesp(&sp);
2368 
2369     {
2370 	struct mbuf *n, *nn;
2371 	struct sadb_msg *newmsg;
2372 	int off, len;
2373 
2374 	/* create new sadb_msg to reply. */
2375 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2376 
2377 	n = key_mget(len);
2378 	if (n == NULL)
2379 		return key_senderror(so, m, ENOBUFS);
2380 
2381 	n->m_len = len;
2382 	n->m_next = NULL;
2383 	off = 0;
2384 
2385 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
2386 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2387 
2388 	IPSEC_ASSERT(off == len, ("length inconsistency (off %u len %u)",
2389 		off, len));
2390 
2391 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2392 	    mhp->extlen[SADB_X_EXT_POLICY], M_NOWAIT);
2393 	if (!n->m_next) {
2394 		m_freem(n);
2395 		return key_senderror(so, m, ENOBUFS);
2396 	}
2397 
2398 	n->m_pkthdr.len = 0;
2399 	for (nn = n; nn; nn = nn->m_next)
2400 		n->m_pkthdr.len += nn->m_len;
2401 
2402 	newmsg = mtod(n, struct sadb_msg *);
2403 	newmsg->sadb_msg_errno = 0;
2404 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2405 
2406 	m_freem(m);
2407 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2408     }
2409 }
2410 
2411 /*
2412  * SADB_X_SPDGET processing
2413  * receive
2414  *   <base, policy(*)>
2415  * from the user(?),
2416  * and send,
2417  *   <base, address(SD), policy>
2418  * to the ikmpd.
2419  * policy(*) including direction of policy.
2420  *
2421  * m will always be freed.
2422  */
2423 static int
2424 key_spdget(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2425 {
2426 	struct secpolicy *sp;
2427 	struct mbuf *n;
2428 	uint32_t id;
2429 
2430 	IPSEC_ASSERT(so != NULL, ("null socket"));
2431 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2432 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2433 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2434 
2435 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2436 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2437 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2438 		    __func__));
2439 		return key_senderror(so, m, EINVAL);
2440 	}
2441 
2442 	id = ((struct sadb_x_policy *)
2443 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2444 
2445 	/* Is there SP in SPD ? */
2446 	if ((sp = key_getspbyid(id)) == NULL) {
2447 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2448 		    __func__, id));
2449 		return key_senderror(so, m, ENOENT);
2450 	}
2451 
2452 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2453 	    mhp->msg->sadb_msg_pid);
2454 	key_freesp(&sp);
2455 	if (n != NULL) {
2456 		m_freem(m);
2457 		return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2458 	} else
2459 		return key_senderror(so, m, ENOBUFS);
2460 }
2461 
2462 /*
2463  * SADB_X_SPDACQUIRE processing.
2464  * Acquire policy and SA(s) for a *OUTBOUND* packet.
2465  * send
2466  *   <base, policy(*)>
2467  * to KMD, and expect to receive
2468  *   <base> with SADB_X_SPDACQUIRE if error occurred,
2469  * or
2470  *   <base, policy>
2471  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2472  * policy(*) is without policy requests.
2473  *
2474  *    0     : succeed
2475  *    others: error number
2476  */
2477 int
2478 key_spdacquire(struct secpolicy *sp)
2479 {
2480 	struct mbuf *result = NULL, *m;
2481 	struct secspacq *newspacq;
2482 
2483 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2484 	IPSEC_ASSERT(sp->req == NULL, ("policy exists"));
2485 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
2486 		("policy not IPSEC %u", sp->policy));
2487 
2488 	/* Get an entry to check whether sent message or not. */
2489 	newspacq = key_getspacq(&sp->spidx);
2490 	if (newspacq != NULL) {
2491 		if (V_key_blockacq_count < newspacq->count) {
2492 			/* reset counter and do send message. */
2493 			newspacq->count = 0;
2494 		} else {
2495 			/* increment counter and do nothing. */
2496 			newspacq->count++;
2497 			SPACQ_UNLOCK();
2498 			return (0);
2499 		}
2500 		SPACQ_UNLOCK();
2501 	} else {
2502 		/* make new entry for blocking to send SADB_ACQUIRE. */
2503 		newspacq = key_newspacq(&sp->spidx);
2504 		if (newspacq == NULL)
2505 			return ENOBUFS;
2506 	}
2507 
2508 	/* create new sadb_msg to reply. */
2509 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2510 	if (!m)
2511 		return ENOBUFS;
2512 
2513 	result = m;
2514 
2515 	result->m_pkthdr.len = 0;
2516 	for (m = result; m; m = m->m_next)
2517 		result->m_pkthdr.len += m->m_len;
2518 
2519 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2520 	    PFKEY_UNIT64(result->m_pkthdr.len);
2521 
2522 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2523 }
2524 
2525 /*
2526  * SADB_SPDFLUSH processing
2527  * receive
2528  *   <base>
2529  * from the user, and free all entries in secpctree.
2530  * and send,
2531  *   <base>
2532  * to the user.
2533  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2534  *
2535  * m will always be freed.
2536  */
2537 static int
2538 key_spdflush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2539 {
2540 	struct secpolicy_queue drainq;
2541 	struct sadb_msg *newmsg;
2542 	struct secpolicy *sp, *nextsp;
2543 	u_int dir;
2544 
2545 	IPSEC_ASSERT(so != NULL, ("null socket"));
2546 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2547 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2548 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2549 
2550 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2551 		return key_senderror(so, m, EINVAL);
2552 
2553 	TAILQ_INIT(&drainq);
2554 	SPTREE_WLOCK();
2555 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2556 		TAILQ_CONCAT(&drainq, &V_sptree[dir], chain);
2557 	}
2558 	/*
2559 	 * We need to set state to DEAD for each policy to be sure,
2560 	 * that another thread won't try to unlink it.
2561 	 * Also remove SP from sphash.
2562 	 */
2563 	TAILQ_FOREACH(sp, &drainq, chain) {
2564 		sp->state = IPSEC_SPSTATE_DEAD;
2565 		LIST_REMOVE(sp, idhash);
2566 	}
2567 	V_sp_genid++;
2568 	V_spd_size = 0;
2569 	SPTREE_WUNLOCK();
2570 	if (SPDCACHE_ENABLED())
2571 		spdcache_clear();
2572 	sp = TAILQ_FIRST(&drainq);
2573 	while (sp != NULL) {
2574 		nextsp = TAILQ_NEXT(sp, chain);
2575 		key_freesp(&sp);
2576 		sp = nextsp;
2577 	}
2578 
2579 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2580 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
2581 		return key_senderror(so, m, ENOBUFS);
2582 	}
2583 
2584 	if (m->m_next)
2585 		m_freem(m->m_next);
2586 	m->m_next = NULL;
2587 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2588 	newmsg = mtod(m, struct sadb_msg *);
2589 	newmsg->sadb_msg_errno = 0;
2590 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2591 
2592 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2593 }
2594 
2595 static uint8_t
2596 key_satype2scopemask(uint8_t satype)
2597 {
2598 
2599 	if (satype == IPSEC_POLICYSCOPE_ANY)
2600 		return (0xff);
2601 	return (satype);
2602 }
2603 /*
2604  * SADB_SPDDUMP processing
2605  * receive
2606  *   <base>
2607  * from the user, and dump all SP leaves and send,
2608  *   <base> .....
2609  * to the ikmpd.
2610  *
2611  * NOTE:
2612  *   sadb_msg_satype is considered as mask of policy scopes.
2613  *   m will always be freed.
2614  */
2615 static int
2616 key_spddump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2617 {
2618 	SPTREE_RLOCK_TRACKER;
2619 	struct secpolicy *sp;
2620 	struct mbuf *n;
2621 	int cnt;
2622 	u_int dir, scope;
2623 
2624 	IPSEC_ASSERT(so != NULL, ("null socket"));
2625 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2626 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2627 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2628 
2629 	/* search SPD entry and get buffer size. */
2630 	cnt = 0;
2631 	scope = key_satype2scopemask(mhp->msg->sadb_msg_satype);
2632 	SPTREE_RLOCK();
2633 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2634 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2635 			TAILQ_FOREACH(sp, &V_sptree[dir], chain)
2636 				cnt++;
2637 		}
2638 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2639 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain)
2640 				cnt++;
2641 		}
2642 	}
2643 
2644 	if (cnt == 0) {
2645 		SPTREE_RUNLOCK();
2646 		return key_senderror(so, m, ENOENT);
2647 	}
2648 
2649 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2650 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2651 			TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
2652 				--cnt;
2653 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2654 				    mhp->msg->sadb_msg_pid);
2655 
2656 				if (n != NULL)
2657 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2658 			}
2659 		}
2660 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2661 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain) {
2662 				--cnt;
2663 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2664 				    mhp->msg->sadb_msg_pid);
2665 
2666 				if (n != NULL)
2667 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2668 			}
2669 		}
2670 	}
2671 
2672 	SPTREE_RUNLOCK();
2673 	m_freem(m);
2674 	return (0);
2675 }
2676 
2677 static struct mbuf *
2678 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq,
2679     u_int32_t pid)
2680 {
2681 	struct mbuf *result = NULL, *m;
2682 	struct seclifetime lt;
2683 
2684 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid, sp->refcnt);
2685 	if (!m)
2686 		goto fail;
2687 	result = m;
2688 
2689 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2690 	    &sp->spidx.src.sa, sp->spidx.prefs,
2691 	    sp->spidx.ul_proto);
2692 	if (!m)
2693 		goto fail;
2694 	m_cat(result, m);
2695 
2696 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2697 	    &sp->spidx.dst.sa, sp->spidx.prefd,
2698 	    sp->spidx.ul_proto);
2699 	if (!m)
2700 		goto fail;
2701 	m_cat(result, m);
2702 
2703 	m = key_sp2mbuf(sp);
2704 	if (!m)
2705 		goto fail;
2706 	m_cat(result, m);
2707 
2708 	if(sp->lifetime){
2709 		lt.addtime=sp->created;
2710 		lt.usetime= sp->lastused;
2711 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_CURRENT);
2712 		if (!m)
2713 			goto fail;
2714 		m_cat(result, m);
2715 
2716 		lt.addtime=sp->lifetime;
2717 		lt.usetime= sp->validtime;
2718 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_HARD);
2719 		if (!m)
2720 			goto fail;
2721 		m_cat(result, m);
2722 	}
2723 
2724 	if ((result->m_flags & M_PKTHDR) == 0)
2725 		goto fail;
2726 
2727 	if (result->m_len < sizeof(struct sadb_msg)) {
2728 		result = m_pullup(result, sizeof(struct sadb_msg));
2729 		if (result == NULL)
2730 			goto fail;
2731 	}
2732 
2733 	result->m_pkthdr.len = 0;
2734 	for (m = result; m; m = m->m_next)
2735 		result->m_pkthdr.len += m->m_len;
2736 
2737 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2738 	    PFKEY_UNIT64(result->m_pkthdr.len);
2739 
2740 	return result;
2741 
2742 fail:
2743 	m_freem(result);
2744 	return NULL;
2745 }
2746 /*
2747  * get PFKEY message length for security policy and request.
2748  */
2749 static size_t
2750 key_getspreqmsglen(struct secpolicy *sp)
2751 {
2752 	size_t tlen, len;
2753 	int i;
2754 
2755 	tlen = sizeof(struct sadb_x_policy);
2756 	/* if is the policy for ipsec ? */
2757 	if (sp->policy != IPSEC_POLICY_IPSEC)
2758 		return (tlen);
2759 
2760 	/* get length of ipsec requests */
2761 	for (i = 0; i < sp->tcount; i++) {
2762 		len = sizeof(struct sadb_x_ipsecrequest)
2763 			+ sp->req[i]->saidx.src.sa.sa_len
2764 			+ sp->req[i]->saidx.dst.sa.sa_len;
2765 
2766 		tlen += PFKEY_ALIGN8(len);
2767 	}
2768 	return (tlen);
2769 }
2770 
2771 /*
2772  * SADB_SPDEXPIRE processing
2773  * send
2774  *   <base, address(SD), lifetime(CH), policy>
2775  * to KMD by PF_KEY.
2776  *
2777  * OUT:	0	: succeed
2778  *	others	: error number
2779  */
2780 static int
2781 key_spdexpire(struct secpolicy *sp)
2782 {
2783 	struct sadb_lifetime *lt;
2784 	struct mbuf *result = NULL, *m;
2785 	int len, error = -1;
2786 
2787 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2788 
2789 	KEYDBG(KEY_STAMP,
2790 	    printf("%s: SP(%p)\n", __func__, sp));
2791 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2792 
2793 	/* set msg header */
2794 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0);
2795 	if (!m) {
2796 		error = ENOBUFS;
2797 		goto fail;
2798 	}
2799 	result = m;
2800 
2801 	/* create lifetime extension (current and hard) */
2802 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
2803 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
2804 	if (m == NULL) {
2805 		error = ENOBUFS;
2806 		goto fail;
2807 	}
2808 	m_align(m, len);
2809 	m->m_len = len;
2810 	bzero(mtod(m, caddr_t), len);
2811 	lt = mtod(m, struct sadb_lifetime *);
2812 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2813 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
2814 	lt->sadb_lifetime_allocations = 0;
2815 	lt->sadb_lifetime_bytes = 0;
2816 	lt->sadb_lifetime_addtime = sp->created;
2817 	lt->sadb_lifetime_usetime = sp->lastused;
2818 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
2819 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2820 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
2821 	lt->sadb_lifetime_allocations = 0;
2822 	lt->sadb_lifetime_bytes = 0;
2823 	lt->sadb_lifetime_addtime = sp->lifetime;
2824 	lt->sadb_lifetime_usetime = sp->validtime;
2825 	m_cat(result, m);
2826 
2827 	/* set sadb_address for source */
2828 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2829 	    &sp->spidx.src.sa,
2830 	    sp->spidx.prefs, sp->spidx.ul_proto);
2831 	if (!m) {
2832 		error = ENOBUFS;
2833 		goto fail;
2834 	}
2835 	m_cat(result, m);
2836 
2837 	/* set sadb_address for destination */
2838 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2839 	    &sp->spidx.dst.sa,
2840 	    sp->spidx.prefd, sp->spidx.ul_proto);
2841 	if (!m) {
2842 		error = ENOBUFS;
2843 		goto fail;
2844 	}
2845 	m_cat(result, m);
2846 
2847 	/* set secpolicy */
2848 	m = key_sp2mbuf(sp);
2849 	if (!m) {
2850 		error = ENOBUFS;
2851 		goto fail;
2852 	}
2853 	m_cat(result, m);
2854 
2855 	if ((result->m_flags & M_PKTHDR) == 0) {
2856 		error = EINVAL;
2857 		goto fail;
2858 	}
2859 
2860 	if (result->m_len < sizeof(struct sadb_msg)) {
2861 		result = m_pullup(result, sizeof(struct sadb_msg));
2862 		if (result == NULL) {
2863 			error = ENOBUFS;
2864 			goto fail;
2865 		}
2866 	}
2867 
2868 	result->m_pkthdr.len = 0;
2869 	for (m = result; m; m = m->m_next)
2870 		result->m_pkthdr.len += m->m_len;
2871 
2872 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2873 	    PFKEY_UNIT64(result->m_pkthdr.len);
2874 
2875 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
2876 
2877  fail:
2878 	if (result)
2879 		m_freem(result);
2880 	return error;
2881 }
2882 
2883 /* %%% SAD management */
2884 /*
2885  * allocating and initialize new SA head.
2886  * OUT:	NULL	: failure due to the lack of memory.
2887  *	others	: pointer to new SA head.
2888  */
2889 static struct secashead *
2890 key_newsah(struct secasindex *saidx)
2891 {
2892 	struct secashead *sah;
2893 
2894 	sah = malloc(sizeof(struct secashead), M_IPSEC_SAH,
2895 	    M_NOWAIT | M_ZERO);
2896 	if (sah == NULL) {
2897 		PFKEYSTAT_INC(in_nomem);
2898 		return (NULL);
2899 	}
2900 	TAILQ_INIT(&sah->savtree_larval);
2901 	TAILQ_INIT(&sah->savtree_alive);
2902 	sah->saidx = *saidx;
2903 	sah->state = SADB_SASTATE_DEAD;
2904 	SAH_INITREF(sah);
2905 
2906 	KEYDBG(KEY_STAMP,
2907 	    printf("%s: SAH(%p)\n", __func__, sah));
2908 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2909 	return (sah);
2910 }
2911 
2912 static void
2913 key_freesah(struct secashead **psah)
2914 {
2915 	struct secashead *sah = *psah;
2916 
2917 	CURVNET_ASSERT_SET();
2918 
2919 	if (SAH_DELREF(sah) == 0)
2920 		return;
2921 
2922 	KEYDBG(KEY_STAMP,
2923 	    printf("%s: last reference to SAH(%p)\n", __func__, sah));
2924 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2925 
2926 	*psah = NULL;
2927 	key_delsah(sah);
2928 }
2929 
2930 static void
2931 key_delsah(struct secashead *sah)
2932 {
2933 	IPSEC_ASSERT(sah != NULL, ("NULL sah"));
2934 	IPSEC_ASSERT(sah->state == SADB_SASTATE_DEAD,
2935 	    ("Attempt to free non DEAD SAH %p", sah));
2936 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_larval),
2937 	    ("Attempt to free SAH %p with LARVAL SA", sah));
2938 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_alive),
2939 	    ("Attempt to free SAH %p with ALIVE SA", sah));
2940 
2941 	free(sah, M_IPSEC_SAH);
2942 }
2943 
2944 /*
2945  * allocating a new SA for key_add() and key_getspi() call,
2946  * and copy the values of mhp into new buffer.
2947  * When SAD message type is SADB_GETSPI set SA state to LARVAL.
2948  * For SADB_ADD create and initialize SA with MATURE state.
2949  * OUT:	NULL	: fail
2950  *	others	: pointer to new secasvar.
2951  */
2952 static struct secasvar *
2953 key_newsav(const struct sadb_msghdr *mhp, struct secasindex *saidx,
2954     uint32_t spi, int *errp)
2955 {
2956 	struct secashead *sah;
2957 	struct secasvar *sav;
2958 	int isnew;
2959 
2960 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2961 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2962 	IPSEC_ASSERT(mhp->msg->sadb_msg_type == SADB_GETSPI ||
2963 	    mhp->msg->sadb_msg_type == SADB_ADD, ("wrong message type"));
2964 
2965 	sav = NULL;
2966 	sah = NULL;
2967 	/* check SPI value */
2968 	switch (saidx->proto) {
2969 	case IPPROTO_ESP:
2970 	case IPPROTO_AH:
2971 		/*
2972 		 * RFC 4302, 2.4. Security Parameters Index (SPI), SPI values
2973 		 * 1-255 reserved by IANA for future use,
2974 		 * 0 for implementation specific, local use.
2975 		 */
2976 		if (ntohl(spi) <= 255) {
2977 			ipseclog((LOG_DEBUG, "%s: illegal range of SPI %u.\n",
2978 			    __func__, ntohl(spi)));
2979 			*errp = EINVAL;
2980 			goto done;
2981 		}
2982 		break;
2983 	}
2984 
2985 	sav = malloc(sizeof(struct secasvar), M_IPSEC_SA, M_NOWAIT | M_ZERO);
2986 	if (sav == NULL) {
2987 		*errp = ENOBUFS;
2988 		goto done;
2989 	}
2990 	sav->lock = malloc_aligned(max(sizeof(struct rmlock),
2991 	    CACHE_LINE_SIZE), CACHE_LINE_SIZE, M_IPSEC_MISC,
2992 	    M_NOWAIT | M_ZERO);
2993 	if (sav->lock == NULL) {
2994 		*errp = ENOBUFS;
2995 		goto done;
2996 	}
2997 	rm_init(sav->lock, "ipsec association");
2998 	sav->lft_c = uma_zalloc_pcpu(ipsec_key_lft_zone, M_NOWAIT | M_ZERO);
2999 	if (sav->lft_c == NULL) {
3000 		*errp = ENOBUFS;
3001 		goto done;
3002 	}
3003 
3004 	sav->spi = spi;
3005 	sav->seq = mhp->msg->sadb_msg_seq;
3006 	sav->state = SADB_SASTATE_LARVAL;
3007 	sav->pid = (pid_t)mhp->msg->sadb_msg_pid;
3008 	SAV_INITREF(sav);
3009 again:
3010 	sah = key_getsah(saidx);
3011 	if (sah == NULL) {
3012 		/* create a new SA index */
3013 		sah = key_newsah(saidx);
3014 		if (sah == NULL) {
3015 			ipseclog((LOG_DEBUG,
3016 			    "%s: No more memory.\n", __func__));
3017 			*errp = ENOBUFS;
3018 			goto done;
3019 		}
3020 		isnew = 1;
3021 	} else
3022 		isnew = 0;
3023 
3024 	sav->sah = sah;
3025 	if (mhp->msg->sadb_msg_type == SADB_GETSPI) {
3026 		sav->created = time_second;
3027 	} else if (sav->state == SADB_SASTATE_LARVAL) {
3028 		/*
3029 		 * Do not call key_setsaval() second time in case
3030 		 * of `goto again`. We will have MATURE state.
3031 		 */
3032 		*errp = key_setsaval(sav, mhp);
3033 		if (*errp != 0)
3034 			goto done;
3035 		sav->state = SADB_SASTATE_MATURE;
3036 	}
3037 
3038 	SAHTREE_WLOCK();
3039 	/*
3040 	 * Check that existing SAH wasn't unlinked.
3041 	 * Since we didn't hold the SAHTREE lock, it is possible,
3042 	 * that callout handler or key_flush() or key_delete() could
3043 	 * unlink this SAH.
3044 	 */
3045 	if (isnew == 0 && sah->state == SADB_SASTATE_DEAD) {
3046 		SAHTREE_WUNLOCK();
3047 		key_freesah(&sah);	/* reference from key_getsah() */
3048 		goto again;
3049 	}
3050 	if (isnew != 0) {
3051 		/*
3052 		 * Add new SAH into SADB.
3053 		 *
3054 		 * XXXAE: we can serialize key_add and key_getspi calls, so
3055 		 * several threads will not fight in the race.
3056 		 * Otherwise we should check under SAHTREE lock, that this
3057 		 * SAH would not added twice.
3058 		 */
3059 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
3060 		/* Add new SAH into hash by addresses */
3061 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
3062 		/* Now we are linked in the chain */
3063 		sah->state = SADB_SASTATE_MATURE;
3064 		/*
3065 		 * SAV references this new SAH.
3066 		 * In case of existing SAH we reuse reference
3067 		 * from key_getsah().
3068 		 */
3069 		SAH_ADDREF(sah);
3070 	}
3071 	/* Link SAV with SAH */
3072 	if (sav->state == SADB_SASTATE_MATURE)
3073 		TAILQ_INSERT_HEAD(&sah->savtree_alive, sav, chain);
3074 	else
3075 		TAILQ_INSERT_HEAD(&sah->savtree_larval, sav, chain);
3076 	/* Add SAV into SPI hash */
3077 	LIST_INSERT_HEAD(SAVHASH_HASH(sav->spi), sav, spihash);
3078 	SAHTREE_WUNLOCK();
3079 	*errp = 0;	/* success */
3080 done:
3081 	if (*errp != 0) {
3082 		if (sav != NULL) {
3083 			if (sav->lock != NULL) {
3084 				rm_destroy(sav->lock);
3085 				free(sav->lock, M_IPSEC_MISC);
3086 			}
3087 			if (sav->lft_c != NULL)
3088 				uma_zfree_pcpu(ipsec_key_lft_zone, sav->lft_c);
3089 			free(sav, M_IPSEC_SA), sav = NULL;
3090 		}
3091 		if (sah != NULL)
3092 			key_freesah(&sah);
3093 		if (*errp == ENOBUFS) {
3094 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3095 			    __func__));
3096 			PFKEYSTAT_INC(in_nomem);
3097 		}
3098 	}
3099 	return (sav);
3100 }
3101 
3102 /*
3103  * free() SA variable entry.
3104  */
3105 static void
3106 key_cleansav(struct secasvar *sav)
3107 {
3108 
3109 	if (sav->natt != NULL) {
3110 		free(sav->natt, M_IPSEC_MISC);
3111 		sav->natt = NULL;
3112 	}
3113 	if (sav->flags & SADB_X_EXT_F_CLONED)
3114 		return;
3115 	if (sav->tdb_xform != NULL) {
3116 		sav->tdb_xform->xf_cleanup(sav);
3117 		sav->tdb_xform = NULL;
3118 	}
3119 	if (sav->key_auth != NULL) {
3120 		zfree(sav->key_auth->key_data, M_IPSEC_MISC);
3121 		free(sav->key_auth, M_IPSEC_MISC);
3122 		sav->key_auth = NULL;
3123 	}
3124 	if (sav->key_enc != NULL) {
3125 		zfree(sav->key_enc->key_data, M_IPSEC_MISC);
3126 		free(sav->key_enc, M_IPSEC_MISC);
3127 		sav->key_enc = NULL;
3128 	}
3129 	if (sav->replay != NULL) {
3130 		mtx_destroy(&sav->replay->lock);
3131 		if (sav->replay->bitmap != NULL)
3132 			free(sav->replay->bitmap, M_IPSEC_MISC);
3133 		free(sav->replay, M_IPSEC_MISC);
3134 		sav->replay = NULL;
3135 	}
3136 	if (sav->lft_h != NULL) {
3137 		free(sav->lft_h, M_IPSEC_MISC);
3138 		sav->lft_h = NULL;
3139 	}
3140 	if (sav->lft_s != NULL) {
3141 		free(sav->lft_s, M_IPSEC_MISC);
3142 		sav->lft_s = NULL;
3143 	}
3144 }
3145 
3146 /*
3147  * free() SA variable entry.
3148  */
3149 static void
3150 key_delsav(struct secasvar *sav)
3151 {
3152 	IPSEC_ASSERT(sav != NULL, ("null sav"));
3153 	IPSEC_ASSERT(sav->state == SADB_SASTATE_DEAD,
3154 	    ("attempt to free non DEAD SA %p", sav));
3155 	IPSEC_ASSERT(sav->refcnt == 0, ("reference count %u > 0",
3156 	    sav->refcnt));
3157 
3158 	/*
3159 	 * SA must be unlinked from the chain and hashtbl.
3160 	 * If SA was cloned, we leave all fields untouched,
3161 	 * except NAT-T config.
3162 	 */
3163 	key_cleansav(sav);
3164 	if ((sav->flags & SADB_X_EXT_F_CLONED) == 0) {
3165 		rm_destroy(sav->lock);
3166 		free(sav->lock, M_IPSEC_MISC);
3167 		uma_zfree_pcpu(ipsec_key_lft_zone, sav->lft_c);
3168 	}
3169 	free(sav, M_IPSEC_SA);
3170 }
3171 
3172 /*
3173  * search SAH.
3174  * OUT:
3175  *	NULL	: not found
3176  *	others	: found, referenced pointer to a SAH.
3177  */
3178 static struct secashead *
3179 key_getsah(struct secasindex *saidx)
3180 {
3181 	SAHTREE_RLOCK_TRACKER;
3182 	struct secashead *sah;
3183 
3184 	SAHTREE_RLOCK();
3185 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
3186 	    if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID) != 0) {
3187 		    SAH_ADDREF(sah);
3188 		    break;
3189 	    }
3190 	}
3191 	SAHTREE_RUNLOCK();
3192 	return (sah);
3193 }
3194 
3195 /*
3196  * Check not to be duplicated SPI.
3197  * OUT:
3198  *	0	: not found
3199  *	1	: found SA with given SPI.
3200  */
3201 static int
3202 key_checkspidup(uint32_t spi)
3203 {
3204 	SAHTREE_RLOCK_TRACKER;
3205 	struct secasvar *sav;
3206 
3207 	/* Assume SPI is in network byte order */
3208 	SAHTREE_RLOCK();
3209 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3210 		if (sav->spi == spi)
3211 			break;
3212 	}
3213 	SAHTREE_RUNLOCK();
3214 	return (sav != NULL);
3215 }
3216 
3217 /*
3218  * Search SA by SPI.
3219  * OUT:
3220  *	NULL	: not found
3221  *	others	: found, referenced pointer to a SA.
3222  */
3223 static struct secasvar *
3224 key_getsavbyspi(uint32_t spi)
3225 {
3226 	SAHTREE_RLOCK_TRACKER;
3227 	struct secasvar *sav;
3228 
3229 	/* Assume SPI is in network byte order */
3230 	SAHTREE_RLOCK();
3231 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3232 		if (sav->spi != spi)
3233 			continue;
3234 		SAV_ADDREF(sav);
3235 		break;
3236 	}
3237 	SAHTREE_RUNLOCK();
3238 	return (sav);
3239 }
3240 
3241 static int
3242 key_updatelifetimes(struct secasvar *sav, const struct sadb_msghdr *mhp)
3243 {
3244 	struct seclifetime *lft_h, *lft_s, *tmp;
3245 
3246 	/* Lifetime extension is optional, check that it is present. */
3247 	if (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3248 	    SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) {
3249 		/*
3250 		 * In case of SADB_UPDATE we may need to change
3251 		 * existing lifetimes.
3252 		 */
3253 		if (sav->state == SADB_SASTATE_MATURE) {
3254 			lft_h = lft_s = NULL;
3255 			goto reset;
3256 		}
3257 		return (0);
3258 	}
3259 	/* Both HARD and SOFT extensions must present */
3260 	if ((SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3261 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
3262 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
3263 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
3264 		ipseclog((LOG_DEBUG,
3265 		    "%s: invalid message: missing required header.\n",
3266 		    __func__));
3267 		return (EINVAL);
3268 	}
3269 	if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD) ||
3270 	    SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_SOFT)) {
3271 		ipseclog((LOG_DEBUG,
3272 		    "%s: invalid message: wrong header size.\n", __func__));
3273 		return (EINVAL);
3274 	}
3275 	lft_h = key_dup_lifemsg((const struct sadb_lifetime *)
3276 	    mhp->ext[SADB_EXT_LIFETIME_HARD], M_IPSEC_MISC);
3277 	if (lft_h == NULL) {
3278 		PFKEYSTAT_INC(in_nomem);
3279 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3280 		return (ENOBUFS);
3281 	}
3282 	lft_s = key_dup_lifemsg((const struct sadb_lifetime *)
3283 	    mhp->ext[SADB_EXT_LIFETIME_SOFT], M_IPSEC_MISC);
3284 	if (lft_s == NULL) {
3285 		PFKEYSTAT_INC(in_nomem);
3286 		free(lft_h, M_IPSEC_MISC);
3287 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3288 		return (ENOBUFS);
3289 	}
3290 reset:
3291 	if (sav->state != SADB_SASTATE_LARVAL) {
3292 		/*
3293 		 * key_update() holds reference to this SA,
3294 		 * so it won't be deleted in meanwhile.
3295 		 */
3296 		SECASVAR_WLOCK(sav);
3297 		tmp = sav->lft_h;
3298 		sav->lft_h = lft_h;
3299 		lft_h = tmp;
3300 
3301 		tmp = sav->lft_s;
3302 		sav->lft_s = lft_s;
3303 		lft_s = tmp;
3304 		SECASVAR_WUNLOCK(sav);
3305 		if (lft_h != NULL)
3306 			free(lft_h, M_IPSEC_MISC);
3307 		if (lft_s != NULL)
3308 			free(lft_s, M_IPSEC_MISC);
3309 		return (0);
3310 	}
3311 	/* We can update lifetime without holding a lock */
3312 	IPSEC_ASSERT(sav->lft_h == NULL, ("lft_h is already initialized\n"));
3313 	IPSEC_ASSERT(sav->lft_s == NULL, ("lft_s is already initialized\n"));
3314 	sav->lft_h = lft_h;
3315 	sav->lft_s = lft_s;
3316 	return (0);
3317 }
3318 
3319 /*
3320  * copy SA values from PF_KEY message except *SPI, SEQ, PID and TYPE*.
3321  * You must update these if need. Expects only LARVAL SAs.
3322  * OUT:	0:	success.
3323  *	!0:	failure.
3324  */
3325 static int
3326 key_setsaval(struct secasvar *sav, const struct sadb_msghdr *mhp)
3327 {
3328 	const struct sadb_sa *sa0;
3329 	const struct sadb_key *key0;
3330 	uint32_t replay;
3331 	size_t len;
3332 	int error;
3333 
3334 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
3335 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
3336 	IPSEC_ASSERT(sav->state == SADB_SASTATE_LARVAL,
3337 	    ("Attempt to update non LARVAL SA"));
3338 
3339 	/* XXX rewrite */
3340 	error = key_setident(sav->sah, mhp);
3341 	if (error != 0)
3342 		goto fail;
3343 
3344 	/* SA */
3345 	if (!SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
3346 		if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
3347 			error = EINVAL;
3348 			goto fail;
3349 		}
3350 		sa0 = (const struct sadb_sa *)mhp->ext[SADB_EXT_SA];
3351 		sav->alg_auth = sa0->sadb_sa_auth;
3352 		sav->alg_enc = sa0->sadb_sa_encrypt;
3353 		sav->flags = sa0->sadb_sa_flags;
3354 		if ((sav->flags & SADB_KEY_FLAGS_MAX) != sav->flags) {
3355 			ipseclog((LOG_DEBUG,
3356 			    "%s: invalid sa_flags 0x%08x.\n", __func__,
3357 			    sav->flags));
3358 			error = EINVAL;
3359 			goto fail;
3360 		}
3361 
3362 		/* Optional replay window */
3363 		replay = 0;
3364 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0)
3365 			replay = sa0->sadb_sa_replay;
3366 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_SA_REPLAY)) {
3367 			if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA_REPLAY)) {
3368 				error = EINVAL;
3369 				goto fail;
3370 			}
3371 			replay = ((const struct sadb_x_sa_replay *)
3372 			    mhp->ext[SADB_X_EXT_SA_REPLAY])->sadb_x_sa_replay_replay;
3373 
3374 			if (replay > UINT32_MAX - 32) {
3375 				ipseclog((LOG_DEBUG,
3376 				    "%s: replay window too big.\n", __func__));
3377 				error = EINVAL;
3378 				goto fail;
3379 			}
3380 
3381 			replay = (replay + 7) >> 3;
3382 		}
3383 
3384 		sav->replay = malloc(sizeof(struct secreplay), M_IPSEC_MISC,
3385 		    M_NOWAIT | M_ZERO);
3386 		if (sav->replay == NULL) {
3387 			PFKEYSTAT_INC(in_nomem);
3388 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3389 			    __func__));
3390 			error = ENOBUFS;
3391 			goto fail;
3392 		}
3393 		mtx_init(&sav->replay->lock, "ipsec replay", NULL, MTX_DEF);
3394 
3395 		if (replay != 0) {
3396 			/* number of 32b blocks to be allocated */
3397 			uint32_t bitmap_size;
3398 
3399 			/* RFC 6479:
3400 			 * - the allocated replay window size must be
3401 			 *   a power of two.
3402 			 * - use an extra 32b block as a redundant window.
3403 			 */
3404 			bitmap_size = 1;
3405 			while (replay + 4 > bitmap_size)
3406 				bitmap_size <<= 1;
3407 			bitmap_size = bitmap_size / 4;
3408 
3409 			sav->replay->bitmap = malloc(
3410 			    bitmap_size * sizeof(uint32_t), M_IPSEC_MISC,
3411 			    M_NOWAIT | M_ZERO);
3412 			if (sav->replay->bitmap == NULL) {
3413 				PFKEYSTAT_INC(in_nomem);
3414 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3415 					__func__));
3416 				error = ENOBUFS;
3417 				goto fail;
3418 			}
3419 			sav->replay->bitmap_size = bitmap_size;
3420 			sav->replay->wsize = replay;
3421 		}
3422 	}
3423 
3424 	/* Authentication keys */
3425 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
3426 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH)) {
3427 			error = EINVAL;
3428 			goto fail;
3429 		}
3430 		error = 0;
3431 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_AUTH];
3432 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3433 		switch (mhp->msg->sadb_msg_satype) {
3434 		case SADB_SATYPE_AH:
3435 		case SADB_SATYPE_ESP:
3436 		case SADB_X_SATYPE_TCPSIGNATURE:
3437 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3438 			    sav->alg_auth != SADB_X_AALG_NULL)
3439 				error = EINVAL;
3440 			break;
3441 		case SADB_X_SATYPE_IPCOMP:
3442 		default:
3443 			error = EINVAL;
3444 			break;
3445 		}
3446 		if (error) {
3447 			ipseclog((LOG_DEBUG, "%s: invalid key_auth values.\n",
3448 				__func__));
3449 			goto fail;
3450 		}
3451 
3452 		sav->key_auth = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3453 		if (sav->key_auth == NULL ) {
3454 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3455 				  __func__));
3456 			PFKEYSTAT_INC(in_nomem);
3457 			error = ENOBUFS;
3458 			goto fail;
3459 		}
3460 	}
3461 
3462 	/* Encryption key */
3463 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) {
3464 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT)) {
3465 			error = EINVAL;
3466 			goto fail;
3467 		}
3468 		error = 0;
3469 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_ENCRYPT];
3470 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3471 		switch (mhp->msg->sadb_msg_satype) {
3472 		case SADB_SATYPE_ESP:
3473 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3474 			    sav->alg_enc != SADB_EALG_NULL) {
3475 				error = EINVAL;
3476 				break;
3477 			}
3478 			sav->key_enc = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3479 			if (sav->key_enc == NULL) {
3480 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3481 					__func__));
3482 				PFKEYSTAT_INC(in_nomem);
3483 				error = ENOBUFS;
3484 				goto fail;
3485 			}
3486 			break;
3487 		case SADB_X_SATYPE_IPCOMP:
3488 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3489 				error = EINVAL;
3490 			sav->key_enc = NULL;	/*just in case*/
3491 			break;
3492 		case SADB_SATYPE_AH:
3493 		case SADB_X_SATYPE_TCPSIGNATURE:
3494 		default:
3495 			error = EINVAL;
3496 			break;
3497 		}
3498 		if (error) {
3499 			ipseclog((LOG_DEBUG, "%s: invalid key_enc value.\n",
3500 				__func__));
3501 			goto fail;
3502 		}
3503 	}
3504 
3505 	/* set iv */
3506 	sav->ivlen = 0;
3507 	switch (mhp->msg->sadb_msg_satype) {
3508 	case SADB_SATYPE_AH:
3509 		if (sav->flags & SADB_X_EXT_DERIV) {
3510 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3511 			    "given to AH SA.\n", __func__));
3512 			error = EINVAL;
3513 			goto fail;
3514 		}
3515 		if (sav->alg_enc != SADB_EALG_NONE) {
3516 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3517 			    "mismated.\n", __func__));
3518 			error = EINVAL;
3519 			goto fail;
3520 		}
3521 		error = xform_init(sav, XF_AH);
3522 		break;
3523 	case SADB_SATYPE_ESP:
3524 		if ((sav->flags & (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) ==
3525 		    (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) {
3526 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3527 			    "given to old-esp.\n", __func__));
3528 			error = EINVAL;
3529 			goto fail;
3530 		}
3531 		error = xform_init(sav, XF_ESP);
3532 		break;
3533 	case SADB_X_SATYPE_IPCOMP:
3534 		if (sav->alg_auth != SADB_AALG_NONE) {
3535 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3536 			    "mismated.\n", __func__));
3537 			error = EINVAL;
3538 			goto fail;
3539 		}
3540 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0 &&
3541 		    ntohl(sav->spi) >= 0x10000) {
3542 			ipseclog((LOG_DEBUG, "%s: invalid cpi for IPComp.\n",
3543 			    __func__));
3544 			error = EINVAL;
3545 			goto fail;
3546 		}
3547 		error = xform_init(sav, XF_IPCOMP);
3548 		break;
3549 	case SADB_X_SATYPE_TCPSIGNATURE:
3550 		if (sav->alg_enc != SADB_EALG_NONE) {
3551 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3552 			    "mismated.\n", __func__));
3553 			error = EINVAL;
3554 			goto fail;
3555 		}
3556 		error = xform_init(sav, XF_TCPSIGNATURE);
3557 		break;
3558 	default:
3559 		ipseclog((LOG_DEBUG, "%s: Invalid satype.\n", __func__));
3560 		error = EPROTONOSUPPORT;
3561 		goto fail;
3562 	}
3563 	if (error) {
3564 		ipseclog((LOG_DEBUG, "%s: unable to initialize SA type %u.\n",
3565 		    __func__, mhp->msg->sadb_msg_satype));
3566 		goto fail;
3567 	}
3568 
3569 	/* Handle NAT-T headers */
3570 	error = key_setnatt(sav, mhp);
3571 	if (error != 0)
3572 		goto fail;
3573 
3574 	/* Initialize lifetime for CURRENT */
3575 	sav->firstused = 0;
3576 	sav->created = time_second;
3577 
3578 	/* lifetimes for HARD and SOFT */
3579 	error = key_updatelifetimes(sav, mhp);
3580 	if (error == 0)
3581 		return (0);
3582 fail:
3583 	key_cleansav(sav);
3584 	return (error);
3585 }
3586 
3587 /*
3588  * subroutine for SADB_GET and SADB_DUMP.
3589  */
3590 static struct mbuf *
3591 key_setdumpsa(struct secasvar *sav, uint8_t type, uint8_t satype,
3592     uint32_t seq, uint32_t pid)
3593 {
3594 	struct seclifetime lft_c;
3595 	struct mbuf *result = NULL, *tres = NULL, *m;
3596 	int i, dumporder[] = {
3597 		SADB_EXT_SA, SADB_X_EXT_SA2, SADB_X_EXT_SA_REPLAY,
3598 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3599 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3600 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY,
3601 		SADB_EXT_KEY_AUTH, SADB_EXT_KEY_ENCRYPT,
3602 		SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
3603 		SADB_EXT_SENSITIVITY,
3604 		SADB_X_EXT_NAT_T_TYPE,
3605 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3606 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3607 		SADB_X_EXT_NAT_T_FRAG,
3608 	};
3609 	uint32_t replay_count;
3610 
3611 	SECASVAR_RLOCK_TRACKER;
3612 
3613 	m = key_setsadbmsg(type, 0, satype, seq, pid, sav->refcnt);
3614 	if (m == NULL)
3615 		goto fail;
3616 	result = m;
3617 
3618 	for (i = nitems(dumporder) - 1; i >= 0; i--) {
3619 		m = NULL;
3620 		switch (dumporder[i]) {
3621 		case SADB_EXT_SA:
3622 			m = key_setsadbsa(sav);
3623 			if (!m)
3624 				goto fail;
3625 			break;
3626 
3627 		case SADB_X_EXT_SA2: {
3628 			SECASVAR_RLOCK(sav);
3629 			replay_count = sav->replay ? sav->replay->count : 0;
3630 			SECASVAR_RUNLOCK(sav);
3631 			m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
3632 					sav->sah->saidx.reqid);
3633 			if (!m)
3634 				goto fail;
3635 			break;
3636 		}
3637 		case SADB_X_EXT_SA_REPLAY:
3638 			if (sav->replay == NULL ||
3639 			    sav->replay->wsize <= UINT8_MAX)
3640 				continue;
3641 
3642 			m = key_setsadbxsareplay(sav->replay->wsize);
3643 			if (!m)
3644 				goto fail;
3645 			break;
3646 
3647 		case SADB_EXT_ADDRESS_SRC:
3648 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3649 			    &sav->sah->saidx.src.sa,
3650 			    FULLMASK, IPSEC_ULPROTO_ANY);
3651 			if (!m)
3652 				goto fail;
3653 			break;
3654 
3655 		case SADB_EXT_ADDRESS_DST:
3656 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3657 			    &sav->sah->saidx.dst.sa,
3658 			    FULLMASK, IPSEC_ULPROTO_ANY);
3659 			if (!m)
3660 				goto fail;
3661 			break;
3662 
3663 		case SADB_EXT_KEY_AUTH:
3664 			if (!sav->key_auth)
3665 				continue;
3666 			m = key_setkey(sav->key_auth, SADB_EXT_KEY_AUTH);
3667 			if (!m)
3668 				goto fail;
3669 			break;
3670 
3671 		case SADB_EXT_KEY_ENCRYPT:
3672 			if (!sav->key_enc)
3673 				continue;
3674 			m = key_setkey(sav->key_enc, SADB_EXT_KEY_ENCRYPT);
3675 			if (!m)
3676 				goto fail;
3677 			break;
3678 
3679 		case SADB_EXT_LIFETIME_CURRENT:
3680 			lft_c.addtime = sav->created;
3681 			lft_c.allocations = (uint32_t)counter_u64_fetch(
3682 			    sav->lft_c_allocations);
3683 			lft_c.bytes = counter_u64_fetch(sav->lft_c_bytes);
3684 			lft_c.usetime = sav->firstused;
3685 			m = key_setlifetime(&lft_c, SADB_EXT_LIFETIME_CURRENT);
3686 			if (!m)
3687 				goto fail;
3688 			break;
3689 
3690 		case SADB_EXT_LIFETIME_HARD:
3691 			if (!sav->lft_h)
3692 				continue;
3693 			m = key_setlifetime(sav->lft_h,
3694 					    SADB_EXT_LIFETIME_HARD);
3695 			if (!m)
3696 				goto fail;
3697 			break;
3698 
3699 		case SADB_EXT_LIFETIME_SOFT:
3700 			if (!sav->lft_s)
3701 				continue;
3702 			m = key_setlifetime(sav->lft_s,
3703 					    SADB_EXT_LIFETIME_SOFT);
3704 
3705 			if (!m)
3706 				goto fail;
3707 			break;
3708 
3709 		case SADB_X_EXT_NAT_T_TYPE:
3710 			if (sav->natt == NULL)
3711 				continue;
3712 			m = key_setsadbxtype(UDP_ENCAP_ESPINUDP);
3713 			if (!m)
3714 				goto fail;
3715 			break;
3716 
3717 		case SADB_X_EXT_NAT_T_DPORT:
3718 			if (sav->natt == NULL)
3719 				continue;
3720 			m = key_setsadbxport(sav->natt->dport,
3721 			    SADB_X_EXT_NAT_T_DPORT);
3722 			if (!m)
3723 				goto fail;
3724 			break;
3725 
3726 		case SADB_X_EXT_NAT_T_SPORT:
3727 			if (sav->natt == NULL)
3728 				continue;
3729 			m = key_setsadbxport(sav->natt->sport,
3730 			    SADB_X_EXT_NAT_T_SPORT);
3731 			if (!m)
3732 				goto fail;
3733 			break;
3734 
3735 		case SADB_X_EXT_NAT_T_OAI:
3736 			if (sav->natt == NULL ||
3737 			    (sav->natt->flags & IPSEC_NATT_F_OAI) == 0)
3738 				continue;
3739 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAI,
3740 			    &sav->natt->oai.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3741 			if (!m)
3742 				goto fail;
3743 			break;
3744 		case SADB_X_EXT_NAT_T_OAR:
3745 			if (sav->natt == NULL ||
3746 			    (sav->natt->flags & IPSEC_NATT_F_OAR) == 0)
3747 				continue;
3748 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAR,
3749 			    &sav->natt->oar.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3750 			if (!m)
3751 				goto fail;
3752 			break;
3753 		case SADB_X_EXT_NAT_T_FRAG:
3754 			/* We do not (yet) support those. */
3755 			continue;
3756 
3757 		case SADB_EXT_ADDRESS_PROXY:
3758 		case SADB_EXT_IDENTITY_SRC:
3759 		case SADB_EXT_IDENTITY_DST:
3760 			/* XXX: should we brought from SPD ? */
3761 		case SADB_EXT_SENSITIVITY:
3762 		default:
3763 			continue;
3764 		}
3765 
3766 		if (!m)
3767 			goto fail;
3768 		if (tres)
3769 			m_cat(m, tres);
3770 		tres = m;
3771 	}
3772 
3773 	m_cat(result, tres);
3774 	tres = NULL;
3775 	if (result->m_len < sizeof(struct sadb_msg)) {
3776 		result = m_pullup(result, sizeof(struct sadb_msg));
3777 		if (result == NULL)
3778 			goto fail;
3779 	}
3780 
3781 	result->m_pkthdr.len = 0;
3782 	for (m = result; m; m = m->m_next)
3783 		result->m_pkthdr.len += m->m_len;
3784 
3785 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3786 	    PFKEY_UNIT64(result->m_pkthdr.len);
3787 
3788 	return result;
3789 
3790 fail:
3791 	m_freem(result);
3792 	m_freem(tres);
3793 	return NULL;
3794 }
3795 
3796 /*
3797  * set data into sadb_msg.
3798  */
3799 static struct mbuf *
3800 key_setsadbmsg(u_int8_t type, u_int16_t tlen, u_int8_t satype, u_int32_t seq,
3801     pid_t pid, u_int16_t reserved)
3802 {
3803 	struct mbuf *m;
3804 	struct sadb_msg *p;
3805 	int len;
3806 
3807 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
3808 	if (len > MCLBYTES)
3809 		return NULL;
3810 	m = key_mget(len);
3811 	if (m == NULL)
3812 		return NULL;
3813 	m->m_pkthdr.len = m->m_len = len;
3814 	m->m_next = NULL;
3815 
3816 	p = mtod(m, struct sadb_msg *);
3817 
3818 	bzero(p, len);
3819 	p->sadb_msg_version = PF_KEY_V2;
3820 	p->sadb_msg_type = type;
3821 	p->sadb_msg_errno = 0;
3822 	p->sadb_msg_satype = satype;
3823 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
3824 	p->sadb_msg_reserved = reserved;
3825 	p->sadb_msg_seq = seq;
3826 	p->sadb_msg_pid = (u_int32_t)pid;
3827 
3828 	return m;
3829 }
3830 
3831 /*
3832  * copy secasvar data into sadb_address.
3833  */
3834 static struct mbuf *
3835 key_setsadbsa(struct secasvar *sav)
3836 {
3837 	struct mbuf *m;
3838 	struct sadb_sa *p;
3839 	int len;
3840 
3841 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
3842 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3843 	if (m == NULL)
3844 		return (NULL);
3845 	m_align(m, len);
3846 	m->m_len = len;
3847 	p = mtod(m, struct sadb_sa *);
3848 	bzero(p, len);
3849 	p->sadb_sa_len = PFKEY_UNIT64(len);
3850 	p->sadb_sa_exttype = SADB_EXT_SA;
3851 	p->sadb_sa_spi = sav->spi;
3852 	p->sadb_sa_replay = sav->replay ?
3853 	    (sav->replay->wsize > UINT8_MAX ? UINT8_MAX :
3854 		sav->replay->wsize): 0;
3855 	p->sadb_sa_state = sav->state;
3856 	p->sadb_sa_auth = sav->alg_auth;
3857 	p->sadb_sa_encrypt = sav->alg_enc;
3858 	p->sadb_sa_flags = sav->flags & SADB_KEY_FLAGS_MAX;
3859 	return (m);
3860 }
3861 
3862 /*
3863  * set data into sadb_address.
3864  */
3865 static struct mbuf *
3866 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
3867     u_int8_t prefixlen, u_int16_t ul_proto)
3868 {
3869 	struct mbuf *m;
3870 	struct sadb_address *p;
3871 	size_t len;
3872 
3873 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
3874 	    PFKEY_ALIGN8(saddr->sa_len);
3875 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3876 	if (m == NULL)
3877 		return (NULL);
3878 	m_align(m, len);
3879 	m->m_len = len;
3880 	p = mtod(m, struct sadb_address *);
3881 
3882 	bzero(p, len);
3883 	p->sadb_address_len = PFKEY_UNIT64(len);
3884 	p->sadb_address_exttype = exttype;
3885 	p->sadb_address_proto = ul_proto;
3886 	if (prefixlen == FULLMASK) {
3887 		switch (saddr->sa_family) {
3888 		case AF_INET:
3889 			prefixlen = sizeof(struct in_addr) << 3;
3890 			break;
3891 		case AF_INET6:
3892 			prefixlen = sizeof(struct in6_addr) << 3;
3893 			break;
3894 		default:
3895 			; /*XXX*/
3896 		}
3897 	}
3898 	p->sadb_address_prefixlen = prefixlen;
3899 	p->sadb_address_reserved = 0;
3900 
3901 	bcopy(saddr,
3902 	    mtod(m, caddr_t) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
3903 	    saddr->sa_len);
3904 
3905 	return m;
3906 }
3907 
3908 /*
3909  * set data into sadb_x_sa2.
3910  */
3911 static struct mbuf *
3912 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int32_t reqid)
3913 {
3914 	struct mbuf *m;
3915 	struct sadb_x_sa2 *p;
3916 	size_t len;
3917 
3918 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
3919 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3920 	if (m == NULL)
3921 		return (NULL);
3922 	m_align(m, len);
3923 	m->m_len = len;
3924 	p = mtod(m, struct sadb_x_sa2 *);
3925 
3926 	bzero(p, len);
3927 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
3928 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
3929 	p->sadb_x_sa2_mode = mode;
3930 	p->sadb_x_sa2_reserved1 = 0;
3931 	p->sadb_x_sa2_reserved2 = 0;
3932 	p->sadb_x_sa2_sequence = seq;
3933 	p->sadb_x_sa2_reqid = reqid;
3934 
3935 	return m;
3936 }
3937 
3938 /*
3939  * Set data into sadb_x_sa_replay.
3940  */
3941 static struct mbuf *
3942 key_setsadbxsareplay(u_int32_t replay)
3943 {
3944 	struct mbuf *m;
3945 	struct sadb_x_sa_replay *p;
3946 	size_t len;
3947 
3948 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa_replay));
3949 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3950 	if (m == NULL)
3951 		return (NULL);
3952 	m_align(m, len);
3953 	m->m_len = len;
3954 	p = mtod(m, struct sadb_x_sa_replay *);
3955 
3956 	bzero(p, len);
3957 	p->sadb_x_sa_replay_len = PFKEY_UNIT64(len);
3958 	p->sadb_x_sa_replay_exttype = SADB_X_EXT_SA_REPLAY;
3959 	p->sadb_x_sa_replay_replay = (replay << 3);
3960 
3961 	return m;
3962 }
3963 
3964 /*
3965  * Set a type in sadb_x_nat_t_type.
3966  */
3967 static struct mbuf *
3968 key_setsadbxtype(u_int16_t type)
3969 {
3970 	struct mbuf *m;
3971 	size_t len;
3972 	struct sadb_x_nat_t_type *p;
3973 
3974 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
3975 
3976 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3977 	if (m == NULL)
3978 		return (NULL);
3979 	m_align(m, len);
3980 	m->m_len = len;
3981 	p = mtod(m, struct sadb_x_nat_t_type *);
3982 
3983 	bzero(p, len);
3984 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
3985 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
3986 	p->sadb_x_nat_t_type_type = type;
3987 
3988 	return (m);
3989 }
3990 /*
3991  * Set a port in sadb_x_nat_t_port.
3992  * In contrast to default RFC 2367 behaviour, port is in network byte order.
3993  */
3994 static struct mbuf *
3995 key_setsadbxport(u_int16_t port, u_int16_t type)
3996 {
3997 	struct mbuf *m;
3998 	size_t len;
3999 	struct sadb_x_nat_t_port *p;
4000 
4001 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
4002 
4003 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4004 	if (m == NULL)
4005 		return (NULL);
4006 	m_align(m, len);
4007 	m->m_len = len;
4008 	p = mtod(m, struct sadb_x_nat_t_port *);
4009 
4010 	bzero(p, len);
4011 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
4012 	p->sadb_x_nat_t_port_exttype = type;
4013 	p->sadb_x_nat_t_port_port = port;
4014 
4015 	return (m);
4016 }
4017 
4018 /*
4019  * Get port from sockaddr. Port is in network byte order.
4020  */
4021 uint16_t
4022 key_portfromsaddr(struct sockaddr *sa)
4023 {
4024 
4025 	switch (sa->sa_family) {
4026 #ifdef INET
4027 	case AF_INET:
4028 		return ((struct sockaddr_in *)sa)->sin_port;
4029 #endif
4030 #ifdef INET6
4031 	case AF_INET6:
4032 		return ((struct sockaddr_in6 *)sa)->sin6_port;
4033 #endif
4034 	}
4035 	return (0);
4036 }
4037 
4038 /*
4039  * Set port in struct sockaddr. Port is in network byte order.
4040  */
4041 void
4042 key_porttosaddr(struct sockaddr *sa, uint16_t port)
4043 {
4044 
4045 	switch (sa->sa_family) {
4046 #ifdef INET
4047 	case AF_INET:
4048 		((struct sockaddr_in *)sa)->sin_port = port;
4049 		break;
4050 #endif
4051 #ifdef INET6
4052 	case AF_INET6:
4053 		((struct sockaddr_in6 *)sa)->sin6_port = port;
4054 		break;
4055 #endif
4056 	default:
4057 		ipseclog((LOG_DEBUG, "%s: unexpected address family %d.\n",
4058 			__func__, sa->sa_family));
4059 		break;
4060 	}
4061 }
4062 
4063 /*
4064  * set data into sadb_x_policy
4065  */
4066 static struct mbuf *
4067 key_setsadbxpolicy(u_int16_t type, u_int8_t dir, u_int32_t id, u_int32_t priority)
4068 {
4069 	struct mbuf *m;
4070 	struct sadb_x_policy *p;
4071 	size_t len;
4072 
4073 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
4074 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4075 	if (m == NULL)
4076 		return (NULL);
4077 	m_align(m, len);
4078 	m->m_len = len;
4079 	p = mtod(m, struct sadb_x_policy *);
4080 
4081 	bzero(p, len);
4082 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
4083 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
4084 	p->sadb_x_policy_type = type;
4085 	p->sadb_x_policy_dir = dir;
4086 	p->sadb_x_policy_id = id;
4087 	p->sadb_x_policy_priority = priority;
4088 
4089 	return m;
4090 }
4091 
4092 /* %%% utilities */
4093 /* Take a key message (sadb_key) from the socket and turn it into one
4094  * of the kernel's key structures (seckey).
4095  *
4096  * IN: pointer to the src
4097  * OUT: NULL no more memory
4098  */
4099 struct seckey *
4100 key_dup_keymsg(const struct sadb_key *src, size_t len,
4101     struct malloc_type *type)
4102 {
4103 	struct seckey *dst;
4104 
4105 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4106 	if (dst != NULL) {
4107 		dst->bits = src->sadb_key_bits;
4108 		dst->key_data = malloc(len, type, M_NOWAIT);
4109 		if (dst->key_data != NULL) {
4110 			bcopy((const char *)(src + 1), dst->key_data, len);
4111 		} else {
4112 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4113 			    __func__));
4114 			free(dst, type);
4115 			dst = NULL;
4116 		}
4117 	} else {
4118 		ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4119 		    __func__));
4120 	}
4121 	return (dst);
4122 }
4123 
4124 /* Take a lifetime message (sadb_lifetime) passed in on a socket and
4125  * turn it into one of the kernel's lifetime structures (seclifetime).
4126  *
4127  * IN: pointer to the destination, source and malloc type
4128  * OUT: NULL, no more memory
4129  */
4130 
4131 static struct seclifetime *
4132 key_dup_lifemsg(const struct sadb_lifetime *src, struct malloc_type *type)
4133 {
4134 	struct seclifetime *dst;
4135 
4136 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4137 	if (dst == NULL) {
4138 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
4139 		return (NULL);
4140 	}
4141 	dst->allocations = src->sadb_lifetime_allocations;
4142 	dst->bytes = src->sadb_lifetime_bytes;
4143 	dst->addtime = src->sadb_lifetime_addtime;
4144 	dst->usetime = src->sadb_lifetime_usetime;
4145 	return (dst);
4146 }
4147 
4148 /*
4149  * compare two secasindex structure.
4150  * flag can specify to compare 2 saidxes.
4151  * compare two secasindex structure without both mode and reqid.
4152  * don't compare port.
4153  * IN:
4154  *      saidx0: source, it can be in SAD.
4155  *      saidx1: object.
4156  * OUT:
4157  *      1 : equal
4158  *      0 : not equal
4159  */
4160 static int
4161 key_cmpsaidx(const struct secasindex *saidx0, const struct secasindex *saidx1,
4162     int flag)
4163 {
4164 
4165 	/* sanity */
4166 	if (saidx0 == NULL && saidx1 == NULL)
4167 		return 1;
4168 
4169 	if (saidx0 == NULL || saidx1 == NULL)
4170 		return 0;
4171 
4172 	if (saidx0->proto != saidx1->proto)
4173 		return 0;
4174 
4175 	if (flag == CMP_EXACTLY) {
4176 		if (saidx0->mode != saidx1->mode)
4177 			return 0;
4178 		if (saidx0->reqid != saidx1->reqid)
4179 			return 0;
4180 		if (bcmp(&saidx0->src, &saidx1->src,
4181 		    saidx0->src.sa.sa_len) != 0 ||
4182 		    bcmp(&saidx0->dst, &saidx1->dst,
4183 		    saidx0->dst.sa.sa_len) != 0)
4184 			return 0;
4185 	} else {
4186 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
4187 		if (flag == CMP_MODE_REQID || flag == CMP_REQID) {
4188 			/*
4189 			 * If reqid of SPD is non-zero, unique SA is required.
4190 			 * The result must be of same reqid in this case.
4191 			 */
4192 			if (saidx1->reqid != 0 &&
4193 			    saidx0->reqid != saidx1->reqid)
4194 				return 0;
4195 		}
4196 
4197 		if (flag == CMP_MODE_REQID) {
4198 			if (saidx0->mode != IPSEC_MODE_ANY
4199 			 && saidx0->mode != saidx1->mode)
4200 				return 0;
4201 		}
4202 
4203 		if (key_sockaddrcmp(&saidx0->src.sa, &saidx1->src.sa, 0) != 0)
4204 			return 0;
4205 		if (key_sockaddrcmp(&saidx0->dst.sa, &saidx1->dst.sa, 0) != 0)
4206 			return 0;
4207 	}
4208 
4209 	return 1;
4210 }
4211 
4212 /*
4213  * compare two secindex structure exactly.
4214  * IN:
4215  *	spidx0: source, it is often in SPD.
4216  *	spidx1: object, it is often from PFKEY message.
4217  * OUT:
4218  *	1 : equal
4219  *	0 : not equal
4220  */
4221 static int
4222 key_cmpspidx_exactly(struct secpolicyindex *spidx0,
4223     struct secpolicyindex *spidx1)
4224 {
4225 	/* sanity */
4226 	if (spidx0 == NULL && spidx1 == NULL)
4227 		return 1;
4228 
4229 	if (spidx0 == NULL || spidx1 == NULL)
4230 		return 0;
4231 
4232 	if (spidx0->prefs != spidx1->prefs
4233 	 || spidx0->prefd != spidx1->prefd
4234 	 || spidx0->ul_proto != spidx1->ul_proto
4235 	 || spidx0->dir != spidx1->dir)
4236 		return 0;
4237 
4238 	return key_sockaddrcmp(&spidx0->src.sa, &spidx1->src.sa, 1) == 0 &&
4239 	       key_sockaddrcmp(&spidx0->dst.sa, &spidx1->dst.sa, 1) == 0;
4240 }
4241 
4242 /*
4243  * compare two secindex structure with mask.
4244  * IN:
4245  *	spidx0: source, it is often in SPD.
4246  *	spidx1: object, it is often from IP header.
4247  * OUT:
4248  *	1 : equal
4249  *	0 : not equal
4250  */
4251 static int
4252 key_cmpspidx_withmask(struct secpolicyindex *spidx0,
4253     struct secpolicyindex *spidx1)
4254 {
4255 	/* sanity */
4256 	if (spidx0 == NULL && spidx1 == NULL)
4257 		return 1;
4258 
4259 	if (spidx0 == NULL || spidx1 == NULL)
4260 		return 0;
4261 
4262 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4263 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4264 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4265 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
4266 		return 0;
4267 
4268 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4269 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY
4270 	 && spidx0->ul_proto != spidx1->ul_proto)
4271 		return 0;
4272 
4273 	switch (spidx0->src.sa.sa_family) {
4274 	case AF_INET:
4275 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY
4276 		 && spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
4277 			return 0;
4278 		if (!key_bbcmp(&spidx0->src.sin.sin_addr,
4279 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
4280 			return 0;
4281 		break;
4282 	case AF_INET6:
4283 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY
4284 		 && spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
4285 			return 0;
4286 		/*
4287 		 * scope_id check. if sin6_scope_id is 0, we regard it
4288 		 * as a wildcard scope, which matches any scope zone ID.
4289 		 */
4290 		if (spidx0->src.sin6.sin6_scope_id &&
4291 		    spidx1->src.sin6.sin6_scope_id &&
4292 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
4293 			return 0;
4294 		if (!key_bbcmp(&spidx0->src.sin6.sin6_addr,
4295 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
4296 			return 0;
4297 		break;
4298 	default:
4299 		/* XXX */
4300 		if (bcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
4301 			return 0;
4302 		break;
4303 	}
4304 
4305 	switch (spidx0->dst.sa.sa_family) {
4306 	case AF_INET:
4307 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY
4308 		 && spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
4309 			return 0;
4310 		if (!key_bbcmp(&spidx0->dst.sin.sin_addr,
4311 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
4312 			return 0;
4313 		break;
4314 	case AF_INET6:
4315 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY
4316 		 && spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
4317 			return 0;
4318 		/*
4319 		 * scope_id check. if sin6_scope_id is 0, we regard it
4320 		 * as a wildcard scope, which matches any scope zone ID.
4321 		 */
4322 		if (spidx0->dst.sin6.sin6_scope_id &&
4323 		    spidx1->dst.sin6.sin6_scope_id &&
4324 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
4325 			return 0;
4326 		if (!key_bbcmp(&spidx0->dst.sin6.sin6_addr,
4327 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
4328 			return 0;
4329 		break;
4330 	default:
4331 		/* XXX */
4332 		if (bcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
4333 			return 0;
4334 		break;
4335 	}
4336 
4337 	/* XXX Do we check other field ?  e.g. flowinfo */
4338 
4339 	return 1;
4340 }
4341 
4342 #ifdef satosin
4343 #undef satosin
4344 #endif
4345 #define satosin(s) ((const struct sockaddr_in *)s)
4346 #ifdef satosin6
4347 #undef satosin6
4348 #endif
4349 #define satosin6(s) ((const struct sockaddr_in6 *)s)
4350 /* returns 0 on match */
4351 int
4352 key_sockaddrcmp(const struct sockaddr *sa1, const struct sockaddr *sa2,
4353     int port)
4354 {
4355 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4356 		return 1;
4357 
4358 	switch (sa1->sa_family) {
4359 #ifdef INET
4360 	case AF_INET:
4361 		if (sa1->sa_len != sizeof(struct sockaddr_in))
4362 			return 1;
4363 		if (satosin(sa1)->sin_addr.s_addr !=
4364 		    satosin(sa2)->sin_addr.s_addr) {
4365 			return 1;
4366 		}
4367 		if (port && satosin(sa1)->sin_port != satosin(sa2)->sin_port)
4368 			return 1;
4369 		break;
4370 #endif
4371 #ifdef INET6
4372 	case AF_INET6:
4373 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4374 			return 1;	/*EINVAL*/
4375 		if (satosin6(sa1)->sin6_scope_id !=
4376 		    satosin6(sa2)->sin6_scope_id) {
4377 			return 1;
4378 		}
4379 		if (!IN6_ARE_ADDR_EQUAL(&satosin6(sa1)->sin6_addr,
4380 		    &satosin6(sa2)->sin6_addr)) {
4381 			return 1;
4382 		}
4383 		if (port &&
4384 		    satosin6(sa1)->sin6_port != satosin6(sa2)->sin6_port) {
4385 			return 1;
4386 		}
4387 		break;
4388 #endif
4389 	default:
4390 		if (bcmp(sa1, sa2, sa1->sa_len) != 0)
4391 			return 1;
4392 		break;
4393 	}
4394 
4395 	return 0;
4396 }
4397 
4398 /* returns 0 on match */
4399 int
4400 key_sockaddrcmp_withmask(const struct sockaddr *sa1,
4401     const struct sockaddr *sa2, size_t mask)
4402 {
4403 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4404 		return (1);
4405 
4406 	switch (sa1->sa_family) {
4407 #ifdef INET
4408 	case AF_INET:
4409 		return (!key_bbcmp(&satosin(sa1)->sin_addr,
4410 		    &satosin(sa2)->sin_addr, mask));
4411 #endif
4412 #ifdef INET6
4413 	case AF_INET6:
4414 		if (satosin6(sa1)->sin6_scope_id !=
4415 		    satosin6(sa2)->sin6_scope_id)
4416 			return (1);
4417 		return (!key_bbcmp(&satosin6(sa1)->sin6_addr,
4418 		    &satosin6(sa2)->sin6_addr, mask));
4419 #endif
4420 	}
4421 	return (1);
4422 }
4423 #undef satosin
4424 #undef satosin6
4425 
4426 /*
4427  * compare two buffers with mask.
4428  * IN:
4429  *	addr1: source
4430  *	addr2: object
4431  *	bits:  Number of bits to compare
4432  * OUT:
4433  *	1 : equal
4434  *	0 : not equal
4435  */
4436 static int
4437 key_bbcmp(const void *a1, const void *a2, u_int bits)
4438 {
4439 	const unsigned char *p1 = a1;
4440 	const unsigned char *p2 = a2;
4441 
4442 	/* XXX: This could be considerably faster if we compare a word
4443 	 * at a time, but it is complicated on LSB Endian machines */
4444 
4445 	/* Handle null pointers */
4446 	if (p1 == NULL || p2 == NULL)
4447 		return (p1 == p2);
4448 
4449 	while (bits >= 8) {
4450 		if (*p1++ != *p2++)
4451 			return 0;
4452 		bits -= 8;
4453 	}
4454 
4455 	if (bits > 0) {
4456 		u_int8_t mask = ~((1<<(8-bits))-1);
4457 		if ((*p1 & mask) != (*p2 & mask))
4458 			return 0;
4459 	}
4460 	return 1;	/* Match! */
4461 }
4462 
4463 static void
4464 key_flush_spd(time_t now)
4465 {
4466 	SPTREE_RLOCK_TRACKER;
4467 	struct secpolicy_list drainq;
4468 	struct secpolicy *sp, *nextsp;
4469 	u_int dir;
4470 
4471 	LIST_INIT(&drainq);
4472 	SPTREE_RLOCK();
4473 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4474 		TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
4475 			if (sp->lifetime == 0 && sp->validtime == 0)
4476 				continue;
4477 			if ((sp->lifetime &&
4478 			    now - sp->created > sp->lifetime) ||
4479 			    (sp->validtime &&
4480 			    now - sp->lastused > sp->validtime)) {
4481 				/* Hold extra reference to send SPDEXPIRE */
4482 				SP_ADDREF(sp);
4483 				LIST_INSERT_HEAD(&drainq, sp, drainq);
4484 			}
4485 		}
4486 	}
4487 	SPTREE_RUNLOCK();
4488 	if (LIST_EMPTY(&drainq))
4489 		return;
4490 
4491 	SPTREE_WLOCK();
4492 	sp = LIST_FIRST(&drainq);
4493 	while (sp != NULL) {
4494 		nextsp = LIST_NEXT(sp, drainq);
4495 		/* Check that SP is still linked */
4496 		if (sp->state != IPSEC_SPSTATE_ALIVE) {
4497 			LIST_REMOVE(sp, drainq);
4498 			key_freesp(&sp); /* release extra reference */
4499 			sp = nextsp;
4500 			continue;
4501 		}
4502 		TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
4503 		V_spd_size--;
4504 		LIST_REMOVE(sp, idhash);
4505 		sp->state = IPSEC_SPSTATE_DEAD;
4506 		sp = nextsp;
4507 	}
4508 	V_sp_genid++;
4509 	SPTREE_WUNLOCK();
4510 	if (SPDCACHE_ENABLED())
4511 		spdcache_clear();
4512 
4513 	sp = LIST_FIRST(&drainq);
4514 	while (sp != NULL) {
4515 		nextsp = LIST_NEXT(sp, drainq);
4516 		key_spdexpire(sp);
4517 		key_freesp(&sp); /* release extra reference */
4518 		key_freesp(&sp); /* release last reference */
4519 		sp = nextsp;
4520 	}
4521 }
4522 
4523 static void
4524 key_flush_sad(time_t now)
4525 {
4526 	SAHTREE_RLOCK_TRACKER;
4527 	struct secashead_list emptyq;
4528 	struct secasvar_list drainq, hexpireq, sexpireq, freeq;
4529 	struct secashead *sah, *nextsah;
4530 	struct secasvar *sav, *nextsav;
4531 
4532 	SECASVAR_RLOCK_TRACKER;
4533 
4534 	LIST_INIT(&drainq);
4535 	LIST_INIT(&hexpireq);
4536 	LIST_INIT(&sexpireq);
4537 	LIST_INIT(&emptyq);
4538 
4539 	SAHTREE_RLOCK();
4540 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
4541 		/* Check for empty SAH */
4542 		if (TAILQ_EMPTY(&sah->savtree_larval) &&
4543 		    TAILQ_EMPTY(&sah->savtree_alive)) {
4544 			SAH_ADDREF(sah);
4545 			LIST_INSERT_HEAD(&emptyq, sah, drainq);
4546 			continue;
4547 		}
4548 		/* Add all stale LARVAL SAs into drainq */
4549 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
4550 			if (now - sav->created < V_key_larval_lifetime)
4551 				continue;
4552 			SAV_ADDREF(sav);
4553 			LIST_INSERT_HEAD(&drainq, sav, drainq);
4554 		}
4555 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
4556 			/* lifetimes aren't specified */
4557 			if (sav->lft_h == NULL)
4558 				continue;
4559 			SECASVAR_RLOCK(sav);
4560 			/*
4561 			 * Check again with lock held, because it may
4562 			 * be updated by SADB_UPDATE.
4563 			 */
4564 			if (sav->lft_h == NULL) {
4565 				SECASVAR_RUNLOCK(sav);
4566 				continue;
4567 			}
4568 			/*
4569 			 * RFC 2367:
4570 			 * HARD lifetimes MUST take precedence over SOFT
4571 			 * lifetimes, meaning if the HARD and SOFT lifetimes
4572 			 * are the same, the HARD lifetime will appear on the
4573 			 * EXPIRE message.
4574 			 */
4575 			/* check HARD lifetime */
4576 			if ((sav->lft_h->addtime != 0 &&
4577 			    now - sav->created > sav->lft_h->addtime) ||
4578 			    (sav->lft_h->usetime != 0 && sav->firstused &&
4579 			    now - sav->firstused > sav->lft_h->usetime) ||
4580 			    (sav->lft_h->bytes != 0 && counter_u64_fetch(
4581 			        sav->lft_c_bytes) > sav->lft_h->bytes)) {
4582 				SECASVAR_RUNLOCK(sav);
4583 				SAV_ADDREF(sav);
4584 				LIST_INSERT_HEAD(&hexpireq, sav, drainq);
4585 				continue;
4586 			}
4587 			/* check SOFT lifetime (only for MATURE SAs) */
4588 			if (sav->state == SADB_SASTATE_MATURE && (
4589 			    (sav->lft_s->addtime != 0 &&
4590 			    now - sav->created > sav->lft_s->addtime) ||
4591 			    (sav->lft_s->usetime != 0 && sav->firstused &&
4592 			    now - sav->firstused > sav->lft_s->usetime) ||
4593 			    (sav->lft_s->bytes != 0 && counter_u64_fetch(
4594 				sav->lft_c_bytes) > sav->lft_s->bytes) ||
4595 			    (!(sav->flags & SADB_X_SAFLAGS_ESN) &&
4596 			    (sav->replay != NULL) && (
4597 			    (sav->replay->count > UINT32_80PCT) ||
4598 			    (sav->replay->last > UINT32_80PCT))))) {
4599 				SECASVAR_RUNLOCK(sav);
4600 				SAV_ADDREF(sav);
4601 				LIST_INSERT_HEAD(&sexpireq, sav, drainq);
4602 				continue;
4603 			}
4604 			SECASVAR_RUNLOCK(sav);
4605 		}
4606 	}
4607 	SAHTREE_RUNLOCK();
4608 
4609 	if (LIST_EMPTY(&emptyq) && LIST_EMPTY(&drainq) &&
4610 	    LIST_EMPTY(&hexpireq) && LIST_EMPTY(&sexpireq))
4611 		return;
4612 
4613 	LIST_INIT(&freeq);
4614 	SAHTREE_WLOCK();
4615 	/* Unlink stale LARVAL SAs */
4616 	sav = LIST_FIRST(&drainq);
4617 	while (sav != NULL) {
4618 		nextsav = LIST_NEXT(sav, drainq);
4619 		/* Check that SA is still LARVAL */
4620 		if (sav->state != SADB_SASTATE_LARVAL) {
4621 			LIST_REMOVE(sav, drainq);
4622 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4623 			sav = nextsav;
4624 			continue;
4625 		}
4626 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
4627 		LIST_REMOVE(sav, spihash);
4628 		sav->state = SADB_SASTATE_DEAD;
4629 		sav = nextsav;
4630 	}
4631 	/* Unlink all SAs with expired HARD lifetime */
4632 	sav = LIST_FIRST(&hexpireq);
4633 	while (sav != NULL) {
4634 		nextsav = LIST_NEXT(sav, drainq);
4635 		/* Check that SA is not unlinked */
4636 		if (sav->state == SADB_SASTATE_DEAD) {
4637 			LIST_REMOVE(sav, drainq);
4638 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4639 			sav = nextsav;
4640 			continue;
4641 		}
4642 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
4643 		LIST_REMOVE(sav, spihash);
4644 		sav->state = SADB_SASTATE_DEAD;
4645 		sav = nextsav;
4646 	}
4647 	/* Mark all SAs with expired SOFT lifetime as DYING */
4648 	sav = LIST_FIRST(&sexpireq);
4649 	while (sav != NULL) {
4650 		nextsav = LIST_NEXT(sav, drainq);
4651 		/* Check that SA is not unlinked */
4652 		if (sav->state == SADB_SASTATE_DEAD) {
4653 			LIST_REMOVE(sav, drainq);
4654 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4655 			sav = nextsav;
4656 			continue;
4657 		}
4658 		/*
4659 		 * NOTE: this doesn't change SA order in the chain.
4660 		 */
4661 		sav->state = SADB_SASTATE_DYING;
4662 		sav = nextsav;
4663 	}
4664 	/* Unlink empty SAHs */
4665 	sah = LIST_FIRST(&emptyq);
4666 	while (sah != NULL) {
4667 		nextsah = LIST_NEXT(sah, drainq);
4668 		/* Check that SAH is still empty and not unlinked */
4669 		if (sah->state == SADB_SASTATE_DEAD ||
4670 		    !TAILQ_EMPTY(&sah->savtree_larval) ||
4671 		    !TAILQ_EMPTY(&sah->savtree_alive)) {
4672 			LIST_REMOVE(sah, drainq);
4673 			key_freesah(&sah); /* release extra reference */
4674 			sah = nextsah;
4675 			continue;
4676 		}
4677 		TAILQ_REMOVE(&V_sahtree, sah, chain);
4678 		LIST_REMOVE(sah, addrhash);
4679 		sah->state = SADB_SASTATE_DEAD;
4680 		sah = nextsah;
4681 	}
4682 	SAHTREE_WUNLOCK();
4683 
4684 	/* Send SPDEXPIRE messages */
4685 	sav = LIST_FIRST(&hexpireq);
4686 	while (sav != NULL) {
4687 		nextsav = LIST_NEXT(sav, drainq);
4688 		key_expire(sav, 1);
4689 		key_freesah(&sav->sah); /* release reference from SAV */
4690 		key_freesav(&sav); /* release extra reference */
4691 		key_freesav(&sav); /* release last reference */
4692 		sav = nextsav;
4693 	}
4694 	sav = LIST_FIRST(&sexpireq);
4695 	while (sav != NULL) {
4696 		nextsav = LIST_NEXT(sav, drainq);
4697 		key_expire(sav, 0);
4698 		key_freesav(&sav); /* release extra reference */
4699 		sav = nextsav;
4700 	}
4701 	/* Free stale LARVAL SAs */
4702 	sav = LIST_FIRST(&drainq);
4703 	while (sav != NULL) {
4704 		nextsav = LIST_NEXT(sav, drainq);
4705 		key_freesah(&sav->sah); /* release reference from SAV */
4706 		key_freesav(&sav); /* release extra reference */
4707 		key_freesav(&sav); /* release last reference */
4708 		sav = nextsav;
4709 	}
4710 	/* Free SAs that were unlinked/changed by someone else */
4711 	sav = LIST_FIRST(&freeq);
4712 	while (sav != NULL) {
4713 		nextsav = LIST_NEXT(sav, drainq);
4714 		key_freesav(&sav); /* release extra reference */
4715 		sav = nextsav;
4716 	}
4717 	/* Free empty SAH */
4718 	sah = LIST_FIRST(&emptyq);
4719 	while (sah != NULL) {
4720 		nextsah = LIST_NEXT(sah, drainq);
4721 		key_freesah(&sah); /* release extra reference */
4722 		key_freesah(&sah); /* release last reference */
4723 		sah = nextsah;
4724 	}
4725 }
4726 
4727 static void
4728 key_flush_acq(time_t now)
4729 {
4730 	struct secacq *acq, *nextacq;
4731 
4732 	/* ACQ tree */
4733 	ACQ_LOCK();
4734 	acq = LIST_FIRST(&V_acqtree);
4735 	while (acq != NULL) {
4736 		nextacq = LIST_NEXT(acq, chain);
4737 		if (now - acq->created > V_key_blockacq_lifetime) {
4738 			LIST_REMOVE(acq, chain);
4739 			LIST_REMOVE(acq, addrhash);
4740 			LIST_REMOVE(acq, seqhash);
4741 			free(acq, M_IPSEC_SAQ);
4742 		}
4743 		acq = nextacq;
4744 	}
4745 	ACQ_UNLOCK();
4746 }
4747 
4748 static void
4749 key_flush_spacq(time_t now)
4750 {
4751 	struct secspacq *acq, *nextacq;
4752 
4753 	/* SP ACQ tree */
4754 	SPACQ_LOCK();
4755 	for (acq = LIST_FIRST(&V_spacqtree); acq != NULL; acq = nextacq) {
4756 		nextacq = LIST_NEXT(acq, chain);
4757 		if (now - acq->created > V_key_blockacq_lifetime
4758 		 && __LIST_CHAINED(acq)) {
4759 			LIST_REMOVE(acq, chain);
4760 			free(acq, M_IPSEC_SAQ);
4761 		}
4762 	}
4763 	SPACQ_UNLOCK();
4764 }
4765 
4766 /*
4767  * time handler.
4768  * scanning SPD and SAD to check status for each entries,
4769  * and do to remove or to expire.
4770  * XXX: year 2038 problem may remain.
4771  */
4772 static void
4773 key_timehandler(void *arg)
4774 {
4775 	VNET_ITERATOR_DECL(vnet_iter);
4776 	time_t now = time_second;
4777 
4778 	VNET_LIST_RLOCK_NOSLEEP();
4779 	VNET_FOREACH(vnet_iter) {
4780 		CURVNET_SET(vnet_iter);
4781 		key_flush_spd(now);
4782 		key_flush_sad(now);
4783 		key_flush_acq(now);
4784 		key_flush_spacq(now);
4785 		CURVNET_RESTORE();
4786 	}
4787 	VNET_LIST_RUNLOCK_NOSLEEP();
4788 
4789 #ifndef IPSEC_DEBUG2
4790 	/* do exchange to tick time !! */
4791 	callout_schedule(&key_timer, hz);
4792 #endif /* IPSEC_DEBUG2 */
4793 }
4794 
4795 u_long
4796 key_random(void)
4797 {
4798 	u_long value;
4799 
4800 	arc4random_buf(&value, sizeof(value));
4801 	return value;
4802 }
4803 
4804 /*
4805  * map SADB_SATYPE_* to IPPROTO_*.
4806  * if satype == SADB_SATYPE then satype is mapped to ~0.
4807  * OUT:
4808  *	0: invalid satype.
4809  */
4810 static uint8_t
4811 key_satype2proto(uint8_t satype)
4812 {
4813 	switch (satype) {
4814 	case SADB_SATYPE_UNSPEC:
4815 		return IPSEC_PROTO_ANY;
4816 	case SADB_SATYPE_AH:
4817 		return IPPROTO_AH;
4818 	case SADB_SATYPE_ESP:
4819 		return IPPROTO_ESP;
4820 	case SADB_X_SATYPE_IPCOMP:
4821 		return IPPROTO_IPCOMP;
4822 	case SADB_X_SATYPE_TCPSIGNATURE:
4823 		return IPPROTO_TCP;
4824 	default:
4825 		return 0;
4826 	}
4827 	/* NOTREACHED */
4828 }
4829 
4830 /*
4831  * map IPPROTO_* to SADB_SATYPE_*
4832  * OUT:
4833  *	0: invalid protocol type.
4834  */
4835 static uint8_t
4836 key_proto2satype(uint8_t proto)
4837 {
4838 	switch (proto) {
4839 	case IPPROTO_AH:
4840 		return SADB_SATYPE_AH;
4841 	case IPPROTO_ESP:
4842 		return SADB_SATYPE_ESP;
4843 	case IPPROTO_IPCOMP:
4844 		return SADB_X_SATYPE_IPCOMP;
4845 	case IPPROTO_TCP:
4846 		return SADB_X_SATYPE_TCPSIGNATURE;
4847 	default:
4848 		return 0;
4849 	}
4850 	/* NOTREACHED */
4851 }
4852 
4853 /* %%% PF_KEY */
4854 /*
4855  * SADB_GETSPI processing is to receive
4856  *	<base, (SA2), src address, dst address, (SPI range)>
4857  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
4858  * tree with the status of LARVAL, and send
4859  *	<base, SA(*), address(SD)>
4860  * to the IKMPd.
4861  *
4862  * IN:	mhp: pointer to the pointer to each header.
4863  * OUT:	NULL if fail.
4864  *	other if success, return pointer to the message to send.
4865  */
4866 static int
4867 key_getspi(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
4868 {
4869 	struct secasindex saidx;
4870 	struct sadb_address *src0, *dst0;
4871 	struct secasvar *sav;
4872 	uint32_t reqid, spi;
4873 	int error;
4874 	uint8_t mode, proto;
4875 
4876 	IPSEC_ASSERT(so != NULL, ("null socket"));
4877 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
4878 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
4879 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
4880 
4881 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
4882 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)
4883 #ifdef PFKEY_STRICT_CHECKS
4884 	    || SADB_CHECKHDR(mhp, SADB_EXT_SPIRANGE)
4885 #endif
4886 	    ) {
4887 		ipseclog((LOG_DEBUG,
4888 		    "%s: invalid message: missing required header.\n",
4889 		    __func__));
4890 		error = EINVAL;
4891 		goto fail;
4892 	}
4893 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
4894 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)
4895 #ifdef PFKEY_STRICT_CHECKS
4896 	    || SADB_CHECKLEN(mhp, SADB_EXT_SPIRANGE)
4897 #endif
4898 	    ) {
4899 		ipseclog((LOG_DEBUG,
4900 		    "%s: invalid message: wrong header size.\n", __func__));
4901 		error = EINVAL;
4902 		goto fail;
4903 	}
4904 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
4905 		mode = IPSEC_MODE_ANY;
4906 		reqid = 0;
4907 	} else {
4908 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
4909 			ipseclog((LOG_DEBUG,
4910 			    "%s: invalid message: wrong header size.\n",
4911 			    __func__));
4912 			error = EINVAL;
4913 			goto fail;
4914 		}
4915 		mode = ((struct sadb_x_sa2 *)
4916 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
4917 		reqid = ((struct sadb_x_sa2 *)
4918 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
4919 	}
4920 
4921 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
4922 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
4923 
4924 	/* map satype to proto */
4925 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
4926 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
4927 			__func__));
4928 		error = EINVAL;
4929 		goto fail;
4930 	}
4931 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
4932 	    (struct sockaddr *)(dst0 + 1));
4933 	if (error != 0) {
4934 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
4935 		error = EINVAL;
4936 		goto fail;
4937 	}
4938 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
4939 
4940 	/* SPI allocation */
4941 	SPI_ALLOC_LOCK();
4942 	spi = key_do_getnewspi(
4943 	    (struct sadb_spirange *)mhp->ext[SADB_EXT_SPIRANGE], &saidx);
4944 	if (spi == 0) {
4945 		/*
4946 		 * Requested SPI or SPI range is not available or
4947 		 * already used.
4948 		 */
4949 		SPI_ALLOC_UNLOCK();
4950 		error = EEXIST;
4951 		goto fail;
4952 	}
4953 	sav = key_newsav(mhp, &saidx, spi, &error);
4954 	SPI_ALLOC_UNLOCK();
4955 	if (sav == NULL)
4956 		goto fail;
4957 
4958 	if (sav->seq != 0) {
4959 		/*
4960 		 * RFC2367:
4961 		 * If the SADB_GETSPI message is in response to a
4962 		 * kernel-generated SADB_ACQUIRE, the sadb_msg_seq
4963 		 * MUST be the same as the SADB_ACQUIRE message.
4964 		 *
4965 		 * XXXAE: However it doesn't definethe behaviour how to
4966 		 * check this and what to do if it doesn't match.
4967 		 * Also what we should do if it matches?
4968 		 *
4969 		 * We can compare saidx used in SADB_ACQUIRE with saidx
4970 		 * used in SADB_GETSPI, but this probably can break
4971 		 * existing software. For now just warn if it doesn't match.
4972 		 *
4973 		 * XXXAE: anyway it looks useless.
4974 		 */
4975 		key_acqdone(&saidx, sav->seq);
4976 	}
4977 	KEYDBG(KEY_STAMP,
4978 	    printf("%s: SA(%p)\n", __func__, sav));
4979 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
4980 
4981     {
4982 	struct mbuf *n, *nn;
4983 	struct sadb_sa *m_sa;
4984 	struct sadb_msg *newmsg;
4985 	int off, len;
4986 
4987 	/* create new sadb_msg to reply. */
4988 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
4989 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
4990 
4991 	n = key_mget(len);
4992 	if (n == NULL) {
4993 		error = ENOBUFS;
4994 		goto fail;
4995 	}
4996 
4997 	n->m_len = len;
4998 	n->m_next = NULL;
4999 	off = 0;
5000 
5001 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
5002 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
5003 
5004 	m_sa = (struct sadb_sa *)(mtod(n, caddr_t) + off);
5005 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
5006 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
5007 	m_sa->sadb_sa_spi = spi; /* SPI is already in network byte order */
5008 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
5009 
5010 	IPSEC_ASSERT(off == len,
5011 		("length inconsistency (off %u len %u)", off, len));
5012 
5013 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
5014 	    SADB_EXT_ADDRESS_DST);
5015 	if (!n->m_next) {
5016 		m_freem(n);
5017 		error = ENOBUFS;
5018 		goto fail;
5019 	}
5020 
5021 	if (n->m_len < sizeof(struct sadb_msg)) {
5022 		n = m_pullup(n, sizeof(struct sadb_msg));
5023 		if (n == NULL)
5024 			return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
5025 	}
5026 
5027 	n->m_pkthdr.len = 0;
5028 	for (nn = n; nn; nn = nn->m_next)
5029 		n->m_pkthdr.len += nn->m_len;
5030 
5031 	newmsg = mtod(n, struct sadb_msg *);
5032 	newmsg->sadb_msg_seq = sav->seq;
5033 	newmsg->sadb_msg_errno = 0;
5034 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
5035 
5036 	m_freem(m);
5037 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
5038     }
5039 
5040 fail:
5041 	return (key_senderror(so, m, error));
5042 }
5043 
5044 /*
5045  * allocating new SPI
5046  * called by key_getspi().
5047  * OUT:
5048  *	0:	failure.
5049  *	others: success, SPI in network byte order.
5050  */
5051 static uint32_t
5052 key_do_getnewspi(struct sadb_spirange *spirange, struct secasindex *saidx)
5053 {
5054 	uint32_t min, max, newspi, t;
5055 	int tries, limit;
5056 
5057 	SPI_ALLOC_LOCK_ASSERT();
5058 
5059 	/* set spi range to allocate */
5060 	if (spirange != NULL) {
5061 		min = spirange->sadb_spirange_min;
5062 		max = spirange->sadb_spirange_max;
5063 	} else {
5064 		min = V_key_spi_minval;
5065 		max = V_key_spi_maxval;
5066 	}
5067 	/* IPCOMP needs 2-byte SPI */
5068 	if (saidx->proto == IPPROTO_IPCOMP) {
5069 		if (min >= 0x10000)
5070 			min = 0xffff;
5071 		if (max >= 0x10000)
5072 			max = 0xffff;
5073 		if (min > max) {
5074 			t = min; min = max; max = t;
5075 		}
5076 	}
5077 
5078 	if (min == max) {
5079 		if (key_checkspidup(htonl(min))) {
5080 			ipseclog((LOG_DEBUG, "%s: SPI %u exists already.\n",
5081 			    __func__, min));
5082 			return 0;
5083 		}
5084 
5085 		tries = 1;
5086 		newspi = min;
5087 	} else {
5088 		/* init SPI */
5089 		newspi = 0;
5090 
5091 		limit = atomic_load_int(&V_key_spi_trycnt);
5092 		/* when requesting to allocate spi ranged */
5093 		for (tries = 0; tries < limit; tries++) {
5094 			/* generate pseudo-random SPI value ranged. */
5095 			newspi = min + (key_random() % (max - min + 1));
5096 			if (!key_checkspidup(htonl(newspi)))
5097 				break;
5098 		}
5099 
5100 		if (tries == limit || newspi == 0) {
5101 			ipseclog((LOG_DEBUG,
5102 			    "%s: failed to allocate SPI.\n", __func__));
5103 			return 0;
5104 		}
5105 	}
5106 
5107 	/* statistics */
5108 	keystat.getspi_count =
5109 	    (keystat.getspi_count + tries) / 2;
5110 
5111 	return (htonl(newspi));
5112 }
5113 
5114 /*
5115  * Find TCP-MD5 SA with corresponding secasindex.
5116  * If not found, return NULL and fill SPI with usable value if needed.
5117  */
5118 static struct secasvar *
5119 key_getsav_tcpmd5(struct secasindex *saidx, uint32_t *spi)
5120 {
5121 	SAHTREE_RLOCK_TRACKER;
5122 	struct secashead *sah;
5123 	struct secasvar *sav;
5124 
5125 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP, ("wrong proto"));
5126 	SAHTREE_RLOCK();
5127 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
5128 		if (sah->saidx.proto != IPPROTO_TCP)
5129 			continue;
5130 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
5131 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
5132 			break;
5133 	}
5134 	if (sah != NULL) {
5135 		if (V_key_preferred_oldsa)
5136 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
5137 		else
5138 			sav = TAILQ_FIRST(&sah->savtree_alive);
5139 		if (sav != NULL) {
5140 			SAV_ADDREF(sav);
5141 			SAHTREE_RUNLOCK();
5142 			return (sav);
5143 		}
5144 	}
5145 	if (spi == NULL) {
5146 		/* No SPI required */
5147 		SAHTREE_RUNLOCK();
5148 		return (NULL);
5149 	}
5150 	/* Check that SPI is unique */
5151 	LIST_FOREACH(sav, SAVHASH_HASH(*spi), spihash) {
5152 		if (sav->spi == *spi)
5153 			break;
5154 	}
5155 	if (sav == NULL) {
5156 		SAHTREE_RUNLOCK();
5157 		/* SPI is already unique */
5158 		return (NULL);
5159 	}
5160 	SAHTREE_RUNLOCK();
5161 	/* XXX: not optimal */
5162 	*spi = key_do_getnewspi(NULL, saidx);
5163 	return (NULL);
5164 }
5165 
5166 static int
5167 key_updateaddresses(struct socket *so, struct mbuf *m,
5168     const struct sadb_msghdr *mhp, struct secasvar *sav,
5169     struct secasindex *saidx)
5170 {
5171 	struct sockaddr *newaddr;
5172 	struct secashead *sah;
5173 	struct secasvar *newsav, *tmp;
5174 	struct mbuf *n;
5175 	int error, isnew;
5176 
5177 	/* Check that we need to change SAH */
5178 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC)) {
5179 		newaddr = (struct sockaddr *)(
5180 		    ((struct sadb_address *)
5181 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_SRC]) + 1);
5182 		bcopy(newaddr, &saidx->src, newaddr->sa_len);
5183 		key_porttosaddr(&saidx->src.sa, 0);
5184 	}
5185 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5186 		newaddr = (struct sockaddr *)(
5187 		    ((struct sadb_address *)
5188 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_DST]) + 1);
5189 		bcopy(newaddr, &saidx->dst, newaddr->sa_len);
5190 		key_porttosaddr(&saidx->dst.sa, 0);
5191 	}
5192 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5193 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5194 		error = key_checksockaddrs(&saidx->src.sa, &saidx->dst.sa);
5195 		if (error != 0) {
5196 			ipseclog((LOG_DEBUG, "%s: invalid new sockaddr.\n",
5197 			    __func__));
5198 			return (error);
5199 		}
5200 
5201 		sah = key_getsah(saidx);
5202 		if (sah == NULL) {
5203 			/* create a new SA index */
5204 			sah = key_newsah(saidx);
5205 			if (sah == NULL) {
5206 				ipseclog((LOG_DEBUG,
5207 				    "%s: No more memory.\n", __func__));
5208 				return (ENOBUFS);
5209 			}
5210 			isnew = 2; /* SAH is new */
5211 		} else
5212 			isnew = 1; /* existing SAH is referenced */
5213 	} else {
5214 		/*
5215 		 * src and dst addresses are still the same.
5216 		 * Do we want to change NAT-T config?
5217 		 */
5218 		if (sav->sah->saidx.proto != IPPROTO_ESP ||
5219 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5220 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5221 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5222 			ipseclog((LOG_DEBUG,
5223 			    "%s: invalid message: missing required header.\n",
5224 			    __func__));
5225 			return (EINVAL);
5226 		}
5227 		/* We hold reference to SA, thus SAH will be referenced too. */
5228 		sah = sav->sah;
5229 		isnew = 0;
5230 	}
5231 
5232 	newsav = malloc(sizeof(struct secasvar), M_IPSEC_SA,
5233 	    M_NOWAIT | M_ZERO);
5234 	if (newsav == NULL) {
5235 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5236 		error = ENOBUFS;
5237 		goto fail;
5238 	}
5239 
5240 	/* Clone SA's content into newsav */
5241 	SAV_INITREF(newsav);
5242 	bcopy(sav, newsav, offsetof(struct secasvar, chain));
5243 	/*
5244 	 * We create new NAT-T config if it is needed.
5245 	 * Old NAT-T config will be freed by key_cleansav() when
5246 	 * last reference to SA will be released.
5247 	 */
5248 	newsav->natt = NULL;
5249 	newsav->sah = sah;
5250 	newsav->state = SADB_SASTATE_MATURE;
5251 	error = key_setnatt(newsav, mhp);
5252 	if (error != 0)
5253 		goto fail;
5254 
5255 	SAHTREE_WLOCK();
5256 	/* Check that SA is still alive */
5257 	if (sav->state == SADB_SASTATE_DEAD) {
5258 		/* SA was unlinked */
5259 		SAHTREE_WUNLOCK();
5260 		error = ESRCH;
5261 		goto fail;
5262 	}
5263 
5264 	/* Unlink SA from SAH and SPI hash */
5265 	IPSEC_ASSERT((sav->flags & SADB_X_EXT_F_CLONED) == 0,
5266 	    ("SA is already cloned"));
5267 	IPSEC_ASSERT(sav->state == SADB_SASTATE_MATURE ||
5268 	    sav->state == SADB_SASTATE_DYING,
5269 	    ("Wrong SA state %u\n", sav->state));
5270 	TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
5271 	LIST_REMOVE(sav, spihash);
5272 	sav->state = SADB_SASTATE_DEAD;
5273 
5274 	/*
5275 	 * Link new SA with SAH. Keep SAs ordered by
5276 	 * create time (newer are first).
5277 	 */
5278 	TAILQ_FOREACH(tmp, &sah->savtree_alive, chain) {
5279 		if (newsav->created > tmp->created) {
5280 			TAILQ_INSERT_BEFORE(tmp, newsav, chain);
5281 			break;
5282 		}
5283 	}
5284 	if (tmp == NULL)
5285 		TAILQ_INSERT_TAIL(&sah->savtree_alive, newsav, chain);
5286 
5287 	/* Add new SA into SPI hash. */
5288 	LIST_INSERT_HEAD(SAVHASH_HASH(newsav->spi), newsav, spihash);
5289 
5290 	/* Add new SAH into SADB. */
5291 	if (isnew == 2) {
5292 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
5293 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
5294 		sah->state = SADB_SASTATE_MATURE;
5295 		SAH_ADDREF(sah); /* newsav references new SAH */
5296 	}
5297 	/*
5298 	 * isnew == 1 -> @sah was referenced by key_getsah().
5299 	 * isnew == 0 -> we use the same @sah, that was used by @sav,
5300 	 *	and we use its reference for @newsav.
5301 	 */
5302 	SECASVAR_WLOCK(sav);
5303 	/* XXX: replace cntr with pointer? */
5304 	newsav->cntr = sav->cntr;
5305 	sav->flags |= SADB_X_EXT_F_CLONED;
5306 	SECASVAR_WUNLOCK(sav);
5307 
5308 	SAHTREE_WUNLOCK();
5309 
5310 	KEYDBG(KEY_STAMP,
5311 	    printf("%s: SA(%p) cloned into SA(%p)\n",
5312 	    __func__, sav, newsav));
5313 	KEYDBG(KEY_DATA, kdebug_secasv(newsav));
5314 
5315 	key_freesav(&sav); /* release last reference */
5316 
5317 	/* set msg buf from mhp */
5318 	n = key_getmsgbuf_x1(m, mhp);
5319 	if (n == NULL) {
5320 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5321 		return (ENOBUFS);
5322 	}
5323 	m_freem(m);
5324 	key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5325 	return (0);
5326 fail:
5327 	if (isnew != 0)
5328 		key_freesah(&sah);
5329 	if (newsav != NULL) {
5330 		if (newsav->natt != NULL)
5331 			free(newsav->natt, M_IPSEC_MISC);
5332 		free(newsav, M_IPSEC_SA);
5333 	}
5334 	return (error);
5335 }
5336 
5337 /*
5338  * SADB_UPDATE processing
5339  * receive
5340  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5341  *       key(AE), (identity(SD),) (sensitivity)>
5342  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5343  * and send
5344  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5345  *       (identity(SD),) (sensitivity)>
5346  * to the ikmpd.
5347  *
5348  * m will always be freed.
5349  */
5350 static int
5351 key_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5352 {
5353 	struct secasindex saidx;
5354 	struct sadb_address *src0, *dst0;
5355 	struct sadb_sa *sa0;
5356 	struct secasvar *sav;
5357 	uint32_t reqid;
5358 	int error;
5359 	uint8_t mode, proto;
5360 
5361 	IPSEC_ASSERT(so != NULL, ("null socket"));
5362 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5363 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5364 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5365 
5366 	/* map satype to proto */
5367 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5368 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5369 		    __func__));
5370 		return key_senderror(so, m, EINVAL);
5371 	}
5372 
5373 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5374 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5375 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5376 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5377 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5378 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5379 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5380 		ipseclog((LOG_DEBUG,
5381 		    "%s: invalid message: missing required header.\n",
5382 		    __func__));
5383 		return key_senderror(so, m, EINVAL);
5384 	}
5385 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5386 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5387 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5388 		ipseclog((LOG_DEBUG,
5389 		    "%s: invalid message: wrong header size.\n", __func__));
5390 		return key_senderror(so, m, EINVAL);
5391 	}
5392 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5393 		mode = IPSEC_MODE_ANY;
5394 		reqid = 0;
5395 	} else {
5396 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5397 			ipseclog((LOG_DEBUG,
5398 			    "%s: invalid message: wrong header size.\n",
5399 			    __func__));
5400 			return key_senderror(so, m, EINVAL);
5401 		}
5402 		mode = ((struct sadb_x_sa2 *)
5403 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5404 		reqid = ((struct sadb_x_sa2 *)
5405 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5406 	}
5407 
5408 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5409 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5410 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5411 
5412 	/*
5413 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5414 	 * SADB_UPDATE message.
5415 	 */
5416 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5417 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5418 #ifdef PFKEY_STRICT_CHECKS
5419 		return key_senderror(so, m, EINVAL);
5420 #endif
5421 	}
5422 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5423 	    (struct sockaddr *)(dst0 + 1));
5424 	if (error != 0) {
5425 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5426 		return key_senderror(so, m, error);
5427 	}
5428 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5429 	sav = key_getsavbyspi(sa0->sadb_sa_spi);
5430 	if (sav == NULL) {
5431 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u\n",
5432 		    __func__, ntohl(sa0->sadb_sa_spi)));
5433 		return key_senderror(so, m, EINVAL);
5434 	}
5435 	/*
5436 	 * Check that SADB_UPDATE issued by the same process that did
5437 	 * SADB_GETSPI or SADB_ADD.
5438 	 */
5439 	if (sav->pid != mhp->msg->sadb_msg_pid) {
5440 		ipseclog((LOG_DEBUG,
5441 		    "%s: pid mismatched (SPI %u, pid %u vs. %u)\n", __func__,
5442 		    ntohl(sav->spi), sav->pid, mhp->msg->sadb_msg_pid));
5443 		key_freesav(&sav);
5444 		return key_senderror(so, m, EINVAL);
5445 	}
5446 	/* saidx should match with SA. */
5447 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_MODE_REQID) == 0) {
5448 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u\n",
5449 		    __func__, ntohl(sav->spi)));
5450 		key_freesav(&sav);
5451 		return key_senderror(so, m, ESRCH);
5452 	}
5453 
5454 	if (sav->state == SADB_SASTATE_LARVAL) {
5455 		if ((mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5456 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) ||
5457 		    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5458 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH))) {
5459 			ipseclog((LOG_DEBUG,
5460 			    "%s: invalid message: missing required header.\n",
5461 			    __func__));
5462 			key_freesav(&sav);
5463 			return key_senderror(so, m, EINVAL);
5464 		}
5465 		/*
5466 		 * We can set any values except src, dst and SPI.
5467 		 */
5468 		error = key_setsaval(sav, mhp);
5469 		if (error != 0) {
5470 			key_freesav(&sav);
5471 			return (key_senderror(so, m, error));
5472 		}
5473 		/* Change SA state to MATURE */
5474 		SAHTREE_WLOCK();
5475 		if (sav->state != SADB_SASTATE_LARVAL) {
5476 			/* SA was deleted or another thread made it MATURE. */
5477 			SAHTREE_WUNLOCK();
5478 			key_freesav(&sav);
5479 			return (key_senderror(so, m, ESRCH));
5480 		}
5481 		/*
5482 		 * NOTE: we keep SAs in savtree_alive ordered by created
5483 		 * time. When SA's state changed from LARVAL to MATURE,
5484 		 * we update its created time in key_setsaval() and move
5485 		 * it into head of savtree_alive.
5486 		 */
5487 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
5488 		TAILQ_INSERT_HEAD(&sav->sah->savtree_alive, sav, chain);
5489 		sav->state = SADB_SASTATE_MATURE;
5490 		SAHTREE_WUNLOCK();
5491 	} else {
5492 		/*
5493 		 * For DYING and MATURE SA we can change only state
5494 		 * and lifetimes. Report EINVAL if something else attempted
5495 		 * to change.
5496 		 */
5497 		if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5498 		    !SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
5499 			key_freesav(&sav);
5500 			return (key_senderror(so, m, EINVAL));
5501 		}
5502 		error = key_updatelifetimes(sav, mhp);
5503 		if (error != 0) {
5504 			key_freesav(&sav);
5505 			return (key_senderror(so, m, error));
5506 		}
5507 		/*
5508 		 * This is FreeBSD extension to RFC2367.
5509 		 * IKEd can specify SADB_X_EXT_NEW_ADDRESS_SRC and/or
5510 		 * SADB_X_EXT_NEW_ADDRESS_DST when it wants to change
5511 		 * SA addresses (for example to implement MOBIKE protocol
5512 		 * as described in RFC4555). Also we allow to change
5513 		 * NAT-T config.
5514 		 */
5515 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5516 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST) ||
5517 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5518 		    sav->natt != NULL) {
5519 			error = key_updateaddresses(so, m, mhp, sav, &saidx);
5520 			key_freesav(&sav);
5521 			if (error != 0)
5522 				return (key_senderror(so, m, error));
5523 			return (0);
5524 		}
5525 		/* Check that SA is still alive */
5526 		SAHTREE_WLOCK();
5527 		if (sav->state == SADB_SASTATE_DEAD) {
5528 			/* SA was unlinked */
5529 			SAHTREE_WUNLOCK();
5530 			key_freesav(&sav);
5531 			return (key_senderror(so, m, ESRCH));
5532 		}
5533 		/*
5534 		 * NOTE: there is possible state moving from DYING to MATURE,
5535 		 * but this doesn't change created time, so we won't reorder
5536 		 * this SA.
5537 		 */
5538 		sav->state = SADB_SASTATE_MATURE;
5539 		SAHTREE_WUNLOCK();
5540 	}
5541 	KEYDBG(KEY_STAMP,
5542 	    printf("%s: SA(%p)\n", __func__, sav));
5543 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5544 	key_freesav(&sav);
5545 
5546     {
5547 	struct mbuf *n;
5548 
5549 	/* set msg buf from mhp */
5550 	n = key_getmsgbuf_x1(m, mhp);
5551 	if (n == NULL) {
5552 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5553 		return key_senderror(so, m, ENOBUFS);
5554 	}
5555 
5556 	m_freem(m);
5557 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5558     }
5559 }
5560 
5561 /*
5562  * SADB_ADD processing
5563  * add an entry to SA database, when received
5564  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5565  *       key(AE), (identity(SD),) (sensitivity)>
5566  * from the ikmpd,
5567  * and send
5568  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5569  *       (identity(SD),) (sensitivity)>
5570  * to the ikmpd.
5571  *
5572  * IGNORE identity and sensitivity messages.
5573  *
5574  * m will always be freed.
5575  */
5576 static int
5577 key_add(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5578 {
5579 	struct secasindex saidx;
5580 	struct sadb_address *src0, *dst0;
5581 	struct sadb_sa *sa0;
5582 	struct secasvar *sav;
5583 	uint32_t reqid, spi;
5584 	uint8_t mode, proto;
5585 	int error;
5586 
5587 	IPSEC_ASSERT(so != NULL, ("null socket"));
5588 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5589 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5590 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5591 
5592 	/* map satype to proto */
5593 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5594 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5595 		    __func__));
5596 		return key_senderror(so, m, EINVAL);
5597 	}
5598 
5599 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5600 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5601 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5602 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP && (
5603 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5604 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT))) ||
5605 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH && (
5606 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH) ||
5607 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH))) ||
5608 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5609 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5610 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5611 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5612 		ipseclog((LOG_DEBUG,
5613 		    "%s: invalid message: missing required header.\n",
5614 		    __func__));
5615 		return key_senderror(so, m, EINVAL);
5616 	}
5617 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5618 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5619 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5620 		ipseclog((LOG_DEBUG,
5621 		    "%s: invalid message: wrong header size.\n", __func__));
5622 		return key_senderror(so, m, EINVAL);
5623 	}
5624 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5625 		mode = IPSEC_MODE_ANY;
5626 		reqid = 0;
5627 	} else {
5628 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5629 			ipseclog((LOG_DEBUG,
5630 			    "%s: invalid message: wrong header size.\n",
5631 			    __func__));
5632 			return key_senderror(so, m, EINVAL);
5633 		}
5634 		mode = ((struct sadb_x_sa2 *)
5635 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5636 		reqid = ((struct sadb_x_sa2 *)
5637 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5638 	}
5639 
5640 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5641 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
5642 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
5643 
5644 	/*
5645 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5646 	 * SADB_ADD message.
5647 	 */
5648 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5649 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5650 #ifdef PFKEY_STRICT_CHECKS
5651 		return key_senderror(so, m, EINVAL);
5652 #endif
5653 	}
5654 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5655 	    (struct sockaddr *)(dst0 + 1));
5656 	if (error != 0) {
5657 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5658 		return key_senderror(so, m, error);
5659 	}
5660 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5661 	spi = sa0->sadb_sa_spi;
5662 	/*
5663 	 * For TCP-MD5 SAs we don't use SPI. Check the uniqueness using
5664 	 * secasindex.
5665 	 * XXXAE: IPComp seems also doesn't use SPI.
5666 	 */
5667 	SPI_ALLOC_LOCK();
5668 	if (proto == IPPROTO_TCP) {
5669 		sav = key_getsav_tcpmd5(&saidx, &spi);
5670 		if (sav == NULL && spi == 0) {
5671 			SPI_ALLOC_UNLOCK();
5672 			/* Failed to allocate SPI */
5673 			ipseclog((LOG_DEBUG, "%s: SA already exists.\n",
5674 			    __func__));
5675 			return key_senderror(so, m, EEXIST);
5676 		}
5677 		/* XXX: SPI that we report back can have another value */
5678 	} else {
5679 		/* We can create new SA only if SPI is different. */
5680 		sav = key_getsavbyspi(spi);
5681 	}
5682 	if (sav != NULL) {
5683 		SPI_ALLOC_UNLOCK();
5684 		key_freesav(&sav);
5685 		ipseclog((LOG_DEBUG, "%s: SA already exists.\n", __func__));
5686 		return key_senderror(so, m, EEXIST);
5687 	}
5688 
5689 	sav = key_newsav(mhp, &saidx, spi, &error);
5690 	SPI_ALLOC_UNLOCK();
5691 	if (sav == NULL)
5692 		return key_senderror(so, m, error);
5693 	KEYDBG(KEY_STAMP,
5694 	    printf("%s: return SA(%p)\n", __func__, sav));
5695 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5696 	/*
5697 	 * If SADB_ADD was in response to SADB_ACQUIRE, we need to schedule
5698 	 * ACQ for deletion.
5699 	 */
5700 	if (sav->seq != 0)
5701 		key_acqdone(&saidx, sav->seq);
5702 
5703     {
5704 	/*
5705 	 * Don't call key_freesav() on error here, as we would like to
5706 	 * keep the SA in the database.
5707 	 */
5708 	struct mbuf *n;
5709 
5710 	/* set msg buf from mhp */
5711 	n = key_getmsgbuf_x1(m, mhp);
5712 	if (n == NULL) {
5713 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5714 		return key_senderror(so, m, ENOBUFS);
5715 	}
5716 
5717 	m_freem(m);
5718 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5719     }
5720 }
5721 
5722 /*
5723  * NAT-T support.
5724  * IKEd may request the use ESP in UDP encapsulation when it detects the
5725  * presence of NAT. It uses NAT-T extension headers for such SAs to specify
5726  * parameters needed for encapsulation and decapsulation. These PF_KEY
5727  * extension headers are not standardized, so this comment addresses our
5728  * implementation.
5729  * SADB_X_EXT_NAT_T_TYPE specifies type of encapsulation, we support only
5730  * UDP_ENCAP_ESPINUDP as described in RFC3948.
5731  * SADB_X_EXT_NAT_T_SPORT/DPORT specifies source and destination ports for
5732  * UDP header. We use these ports in UDP encapsulation procedure, also we
5733  * can check them in UDP decapsulation procedure.
5734  * SADB_X_EXT_NAT_T_OA[IR] specifies original address of initiator or
5735  * responder. These addresses can be used for transport mode to adjust
5736  * checksum after decapsulation and decryption. Since original IP addresses
5737  * used by peer usually different (we detected presence of NAT), TCP/UDP
5738  * pseudo header checksum and IP header checksum was calculated using original
5739  * addresses. After decapsulation and decryption we need to adjust checksum
5740  * to have correct datagram.
5741  *
5742  * We expect presence of NAT-T extension headers only in SADB_ADD and
5743  * SADB_UPDATE messages. We report NAT-T extension headers in replies
5744  * to SADB_ADD, SADB_UPDATE, SADB_GET, and SADB_DUMP messages.
5745  */
5746 static int
5747 key_setnatt(struct secasvar *sav, const struct sadb_msghdr *mhp)
5748 {
5749 	struct sadb_x_nat_t_port *port;
5750 	struct sadb_x_nat_t_type *type;
5751 	struct sadb_address *oai, *oar;
5752 	struct sockaddr *sa;
5753 	uint32_t addr;
5754 	uint16_t cksum;
5755 
5756 	IPSEC_ASSERT(sav->natt == NULL, ("natt is already initialized"));
5757 	/*
5758 	 * Ignore NAT-T headers if sproto isn't ESP.
5759 	 */
5760 	if (sav->sah->saidx.proto != IPPROTO_ESP)
5761 		return (0);
5762 
5763 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) &&
5764 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) &&
5765 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5766 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5767 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5768 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5769 			ipseclog((LOG_DEBUG,
5770 			    "%s: invalid message: wrong header size.\n",
5771 			    __func__));
5772 			return (EINVAL);
5773 		}
5774 	} else
5775 		return (0);
5776 
5777 	type = (struct sadb_x_nat_t_type *)mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5778 	if (type->sadb_x_nat_t_type_type != UDP_ENCAP_ESPINUDP) {
5779 		ipseclog((LOG_DEBUG, "%s: unsupported NAT-T type %u.\n",
5780 		    __func__, type->sadb_x_nat_t_type_type));
5781 		return (EINVAL);
5782 	}
5783 	/*
5784 	 * Allocate storage for NAT-T config.
5785 	 * On error it will be released by key_cleansav().
5786 	 */
5787 	sav->natt = malloc(sizeof(struct secnatt), M_IPSEC_MISC,
5788 	    M_NOWAIT | M_ZERO);
5789 	if (sav->natt == NULL) {
5790 		PFKEYSTAT_INC(in_nomem);
5791 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5792 		return (ENOBUFS);
5793 	}
5794 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5795 	if (port->sadb_x_nat_t_port_port == 0) {
5796 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T sport specified.\n",
5797 		    __func__));
5798 		return (EINVAL);
5799 	}
5800 	sav->natt->sport = port->sadb_x_nat_t_port_port;
5801 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5802 	if (port->sadb_x_nat_t_port_port == 0) {
5803 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T dport specified.\n",
5804 		    __func__));
5805 		return (EINVAL);
5806 	}
5807 	sav->natt->dport = port->sadb_x_nat_t_port_port;
5808 
5809 	/*
5810 	 * SADB_X_EXT_NAT_T_OAI and SADB_X_EXT_NAT_T_OAR are optional
5811 	 * and needed only for transport mode IPsec.
5812 	 * Usually NAT translates only one address, but it is possible,
5813 	 * that both addresses could be translated.
5814 	 * NOTE: Value of SADB_X_EXT_NAT_T_OAI is equal to SADB_X_EXT_NAT_T_OA.
5815 	 */
5816 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAI)) {
5817 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAI)) {
5818 			ipseclog((LOG_DEBUG,
5819 			    "%s: invalid message: wrong header size.\n",
5820 			    __func__));
5821 			return (EINVAL);
5822 		}
5823 		oai = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAI];
5824 	} else
5825 		oai = NULL;
5826 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAR)) {
5827 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAR)) {
5828 			ipseclog((LOG_DEBUG,
5829 			    "%s: invalid message: wrong header size.\n",
5830 			    __func__));
5831 			return (EINVAL);
5832 		}
5833 		oar = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAR];
5834 	} else
5835 		oar = NULL;
5836 
5837 	/* Initialize addresses only for transport mode */
5838 	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL) {
5839 		cksum = 0;
5840 		if (oai != NULL) {
5841 			/* Currently we support only AF_INET */
5842 			sa = (struct sockaddr *)(oai + 1);
5843 			if (sa->sa_family != AF_INET ||
5844 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5845 				ipseclog((LOG_DEBUG,
5846 				    "%s: wrong NAT-OAi header.\n",
5847 				    __func__));
5848 				return (EINVAL);
5849 			}
5850 			/* Ignore address if it the same */
5851 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5852 			    sav->sah->saidx.src.sin.sin_addr.s_addr) {
5853 				bcopy(sa, &sav->natt->oai.sa, sa->sa_len);
5854 				sav->natt->flags |= IPSEC_NATT_F_OAI;
5855 				/* Calculate checksum delta */
5856 				addr = sav->sah->saidx.src.sin.sin_addr.s_addr;
5857 				cksum = in_addword(cksum, ~addr >> 16);
5858 				cksum = in_addword(cksum, ~addr & 0xffff);
5859 				addr = sav->natt->oai.sin.sin_addr.s_addr;
5860 				cksum = in_addword(cksum, addr >> 16);
5861 				cksum = in_addword(cksum, addr & 0xffff);
5862 			}
5863 		}
5864 		if (oar != NULL) {
5865 			/* Currently we support only AF_INET */
5866 			sa = (struct sockaddr *)(oar + 1);
5867 			if (sa->sa_family != AF_INET ||
5868 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5869 				ipseclog((LOG_DEBUG,
5870 				    "%s: wrong NAT-OAr header.\n",
5871 				    __func__));
5872 				return (EINVAL);
5873 			}
5874 			/* Ignore address if it the same */
5875 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5876 			    sav->sah->saidx.dst.sin.sin_addr.s_addr) {
5877 				bcopy(sa, &sav->natt->oar.sa, sa->sa_len);
5878 				sav->natt->flags |= IPSEC_NATT_F_OAR;
5879 				/* Calculate checksum delta */
5880 				addr = sav->sah->saidx.dst.sin.sin_addr.s_addr;
5881 				cksum = in_addword(cksum, ~addr >> 16);
5882 				cksum = in_addword(cksum, ~addr & 0xffff);
5883 				addr = sav->natt->oar.sin.sin_addr.s_addr;
5884 				cksum = in_addword(cksum, addr >> 16);
5885 				cksum = in_addword(cksum, addr & 0xffff);
5886 			}
5887 		}
5888 		sav->natt->cksum = cksum;
5889 	}
5890 	return (0);
5891 }
5892 
5893 static int
5894 key_setident(struct secashead *sah, const struct sadb_msghdr *mhp)
5895 {
5896 	const struct sadb_ident *idsrc, *iddst;
5897 
5898 	IPSEC_ASSERT(sah != NULL, ("null secashead"));
5899 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5900 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5901 
5902 	/* don't make buffer if not there */
5903 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) &&
5904 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5905 		sah->idents = NULL;
5906 		sah->identd = NULL;
5907 		return (0);
5908 	}
5909 
5910 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) ||
5911 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5912 		ipseclog((LOG_DEBUG, "%s: invalid identity.\n", __func__));
5913 		return (EINVAL);
5914 	}
5915 
5916 	idsrc = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_SRC];
5917 	iddst = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_DST];
5918 
5919 	/* validity check */
5920 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
5921 		ipseclog((LOG_DEBUG, "%s: ident type mismatch.\n", __func__));
5922 		return EINVAL;
5923 	}
5924 
5925 	switch (idsrc->sadb_ident_type) {
5926 	case SADB_IDENTTYPE_PREFIX:
5927 	case SADB_IDENTTYPE_FQDN:
5928 	case SADB_IDENTTYPE_USERFQDN:
5929 	default:
5930 		/* XXX do nothing */
5931 		sah->idents = NULL;
5932 		sah->identd = NULL;
5933 	 	return 0;
5934 	}
5935 
5936 	/* make structure */
5937 	sah->idents = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5938 	if (sah->idents == NULL) {
5939 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5940 		return ENOBUFS;
5941 	}
5942 	sah->identd = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5943 	if (sah->identd == NULL) {
5944 		free(sah->idents, M_IPSEC_MISC);
5945 		sah->idents = NULL;
5946 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5947 		return ENOBUFS;
5948 	}
5949 	sah->idents->type = idsrc->sadb_ident_type;
5950 	sah->idents->id = idsrc->sadb_ident_id;
5951 
5952 	sah->identd->type = iddst->sadb_ident_type;
5953 	sah->identd->id = iddst->sadb_ident_id;
5954 
5955 	return 0;
5956 }
5957 
5958 /*
5959  * m will not be freed on return.
5960  * it is caller's responsibility to free the result.
5961  *
5962  * Called from SADB_ADD and SADB_UPDATE. Reply will contain headers
5963  * from the request in defined order.
5964  */
5965 static struct mbuf *
5966 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
5967 {
5968 	struct mbuf *n;
5969 
5970 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5971 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5972 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5973 
5974 	/* create new sadb_msg to reply. */
5975 	n = key_gather_mbuf(m, mhp, 1, 16, SADB_EXT_RESERVED,
5976 	    SADB_EXT_SA, SADB_X_EXT_SA2,
5977 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
5978 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
5979 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
5980 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
5981 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
5982 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NEW_ADDRESS_SRC,
5983 	    SADB_X_EXT_NEW_ADDRESS_DST);
5984 	if (!n)
5985 		return NULL;
5986 
5987 	if (n->m_len < sizeof(struct sadb_msg)) {
5988 		n = m_pullup(n, sizeof(struct sadb_msg));
5989 		if (n == NULL)
5990 			return NULL;
5991 	}
5992 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
5993 	mtod(n, struct sadb_msg *)->sadb_msg_len =
5994 	    PFKEY_UNIT64(n->m_pkthdr.len);
5995 
5996 	return n;
5997 }
5998 
5999 /*
6000  * SADB_DELETE processing
6001  * receive
6002  *   <base, SA(*), address(SD)>
6003  * from the ikmpd, and set SADB_SASTATE_DEAD,
6004  * and send,
6005  *   <base, SA(*), address(SD)>
6006  * to the ikmpd.
6007  *
6008  * m will always be freed.
6009  */
6010 static int
6011 key_delete(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6012 {
6013 	struct secasindex saidx;
6014 	struct sadb_address *src0, *dst0;
6015 	struct secasvar *sav;
6016 	struct sadb_sa *sa0;
6017 	uint8_t proto;
6018 
6019 	IPSEC_ASSERT(so != NULL, ("null socket"));
6020 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6021 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6022 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6023 
6024 	/* map satype to proto */
6025 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6026 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6027 		    __func__));
6028 		return key_senderror(so, m, EINVAL);
6029 	}
6030 
6031 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6032 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
6033 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6034 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6035 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
6036 		    __func__));
6037 		return key_senderror(so, m, EINVAL);
6038 	}
6039 
6040 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
6041 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
6042 
6043 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6044 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6045 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6046 		return (key_senderror(so, m, EINVAL));
6047 	}
6048 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6049 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
6050 		/*
6051 		 * Caller wants us to delete all non-LARVAL SAs
6052 		 * that match the src/dst.  This is used during
6053 		 * IKE INITIAL-CONTACT.
6054 		 * XXXAE: this looks like some extension to RFC2367.
6055 		 */
6056 		ipseclog((LOG_DEBUG, "%s: doing delete all.\n", __func__));
6057 		return (key_delete_all(so, m, mhp, &saidx));
6058 	}
6059 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
6060 		ipseclog((LOG_DEBUG,
6061 		    "%s: invalid message: wrong header size.\n", __func__));
6062 		return (key_senderror(so, m, EINVAL));
6063 	}
6064 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6065 	SPI_ALLOC_LOCK();
6066 	if (proto == IPPROTO_TCP)
6067 		sav = key_getsav_tcpmd5(&saidx, NULL);
6068 	else
6069 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6070 	SPI_ALLOC_UNLOCK();
6071 	if (sav == NULL) {
6072 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u.\n",
6073 		    __func__, ntohl(sa0->sadb_sa_spi)));
6074 		return (key_senderror(so, m, ESRCH));
6075 	}
6076 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6077 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6078 		    __func__, ntohl(sav->spi)));
6079 		key_freesav(&sav);
6080 		return (key_senderror(so, m, ESRCH));
6081 	}
6082 	KEYDBG(KEY_STAMP,
6083 	    printf("%s: SA(%p)\n", __func__, sav));
6084 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
6085 	key_unlinksav(sav);
6086 	key_freesav(&sav);
6087 
6088     {
6089 	struct mbuf *n;
6090 	struct sadb_msg *newmsg;
6091 
6092 	/* create new sadb_msg to reply. */
6093 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
6094 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6095 	if (!n)
6096 		return key_senderror(so, m, ENOBUFS);
6097 
6098 	if (n->m_len < sizeof(struct sadb_msg)) {
6099 		n = m_pullup(n, sizeof(struct sadb_msg));
6100 		if (n == NULL)
6101 			return key_senderror(so, m, ENOBUFS);
6102 	}
6103 	newmsg = mtod(n, struct sadb_msg *);
6104 	newmsg->sadb_msg_errno = 0;
6105 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6106 
6107 	m_freem(m);
6108 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6109     }
6110 }
6111 
6112 /*
6113  * delete all SAs for src/dst.  Called from key_delete().
6114  */
6115 static int
6116 key_delete_all(struct socket *so, struct mbuf *m,
6117     const struct sadb_msghdr *mhp, struct secasindex *saidx)
6118 {
6119 	struct secasvar_queue drainq;
6120 	struct secashead *sah;
6121 	struct secasvar *sav, *nextsav;
6122 
6123 	TAILQ_INIT(&drainq);
6124 	SAHTREE_WLOCK();
6125 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
6126 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_HEAD) == 0)
6127 			continue;
6128 		/* Move all ALIVE SAs into drainq */
6129 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6130 	}
6131 	/* Unlink all queued SAs from SPI hash */
6132 	TAILQ_FOREACH(sav, &drainq, chain) {
6133 		sav->state = SADB_SASTATE_DEAD;
6134 		LIST_REMOVE(sav, spihash);
6135 	}
6136 	SAHTREE_WUNLOCK();
6137 	/* Now we can release reference for all SAs in drainq */
6138 	sav = TAILQ_FIRST(&drainq);
6139 	while (sav != NULL) {
6140 		KEYDBG(KEY_STAMP,
6141 		    printf("%s: SA(%p)\n", __func__, sav));
6142 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6143 		nextsav = TAILQ_NEXT(sav, chain);
6144 		key_freesah(&sav->sah); /* release reference from SAV */
6145 		key_freesav(&sav); /* release last reference */
6146 		sav = nextsav;
6147 	}
6148 
6149     {
6150 	struct mbuf *n;
6151 	struct sadb_msg *newmsg;
6152 
6153 	/* create new sadb_msg to reply. */
6154 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
6155 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6156 	if (!n)
6157 		return key_senderror(so, m, ENOBUFS);
6158 
6159 	if (n->m_len < sizeof(struct sadb_msg)) {
6160 		n = m_pullup(n, sizeof(struct sadb_msg));
6161 		if (n == NULL)
6162 			return key_senderror(so, m, ENOBUFS);
6163 	}
6164 	newmsg = mtod(n, struct sadb_msg *);
6165 	newmsg->sadb_msg_errno = 0;
6166 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6167 
6168 	m_freem(m);
6169 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6170     }
6171 }
6172 
6173 /*
6174  * Delete all alive SAs for corresponding xform.
6175  * Larval SAs have not initialized tdb_xform, so it is safe to leave them
6176  * here when xform disappears.
6177  */
6178 void
6179 key_delete_xform(const struct xformsw *xsp)
6180 {
6181 	struct secasvar_queue drainq;
6182 	struct secashead *sah;
6183 	struct secasvar *sav, *nextsav;
6184 
6185 	TAILQ_INIT(&drainq);
6186 	SAHTREE_WLOCK();
6187 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
6188 		sav = TAILQ_FIRST(&sah->savtree_alive);
6189 		if (sav == NULL)
6190 			continue;
6191 		if (sav->tdb_xform != xsp)
6192 			continue;
6193 		/*
6194 		 * It is supposed that all SAs in the chain are related to
6195 		 * one xform.
6196 		 */
6197 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6198 	}
6199 	/* Unlink all queued SAs from SPI hash */
6200 	TAILQ_FOREACH(sav, &drainq, chain) {
6201 		sav->state = SADB_SASTATE_DEAD;
6202 		LIST_REMOVE(sav, spihash);
6203 	}
6204 	SAHTREE_WUNLOCK();
6205 
6206 	/* Now we can release reference for all SAs in drainq */
6207 	sav = TAILQ_FIRST(&drainq);
6208 	while (sav != NULL) {
6209 		KEYDBG(KEY_STAMP,
6210 		    printf("%s: SA(%p)\n", __func__, sav));
6211 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6212 		nextsav = TAILQ_NEXT(sav, chain);
6213 		key_freesah(&sav->sah); /* release reference from SAV */
6214 		key_freesav(&sav); /* release last reference */
6215 		sav = nextsav;
6216 	}
6217 }
6218 
6219 /*
6220  * SADB_GET processing
6221  * receive
6222  *   <base, SA(*), address(SD)>
6223  * from the ikmpd, and get a SP and a SA to respond,
6224  * and send,
6225  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6226  *       (identity(SD),) (sensitivity)>
6227  * to the ikmpd.
6228  *
6229  * m will always be freed.
6230  */
6231 static int
6232 key_get(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6233 {
6234 	struct secasindex saidx;
6235 	struct sadb_address *src0, *dst0;
6236 	struct sadb_sa *sa0;
6237 	struct secasvar *sav;
6238 	uint8_t proto;
6239 
6240 	IPSEC_ASSERT(so != NULL, ("null socket"));
6241 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6242 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6243 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6244 
6245 	/* map satype to proto */
6246 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6247 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6248 			__func__));
6249 		return key_senderror(so, m, EINVAL);
6250 	}
6251 
6252 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
6253 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6254 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)) {
6255 		ipseclog((LOG_DEBUG,
6256 		    "%s: invalid message: missing required header.\n",
6257 		    __func__));
6258 		return key_senderror(so, m, EINVAL);
6259 	}
6260 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
6261 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6262 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6263 		ipseclog((LOG_DEBUG,
6264 		    "%s: invalid message: wrong header size.\n", __func__));
6265 		return key_senderror(so, m, EINVAL);
6266 	}
6267 
6268 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6269 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
6270 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
6271 
6272 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6273 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6274 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6275 		return key_senderror(so, m, EINVAL);
6276 	}
6277 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6278 
6279 	SPI_ALLOC_LOCK();
6280 	if (proto == IPPROTO_TCP)
6281 		sav = key_getsav_tcpmd5(&saidx, NULL);
6282 	else
6283 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6284 	SPI_ALLOC_UNLOCK();
6285 	if (sav == NULL) {
6286 		ipseclog((LOG_DEBUG, "%s: no SA found.\n", __func__));
6287 		return key_senderror(so, m, ESRCH);
6288 	}
6289 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6290 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6291 		    __func__, ntohl(sa0->sadb_sa_spi)));
6292 		key_freesav(&sav);
6293 		return (key_senderror(so, m, ESRCH));
6294 	}
6295 
6296     {
6297 	struct mbuf *n;
6298 	uint8_t satype;
6299 
6300 	/* map proto to satype */
6301 	if ((satype = key_proto2satype(sav->sah->saidx.proto)) == 0) {
6302 		ipseclog((LOG_DEBUG, "%s: there was invalid proto in SAD.\n",
6303 		    __func__));
6304 		key_freesav(&sav);
6305 		return key_senderror(so, m, EINVAL);
6306 	}
6307 
6308 	/* create new sadb_msg to reply. */
6309 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
6310 	    mhp->msg->sadb_msg_pid);
6311 
6312 	key_freesav(&sav);
6313 	if (!n)
6314 		return key_senderror(so, m, ENOBUFS);
6315 
6316 	m_freem(m);
6317 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
6318     }
6319 }
6320 
6321 /* XXX make it sysctl-configurable? */
6322 static void
6323 key_getcomb_setlifetime(struct sadb_comb *comb)
6324 {
6325 
6326 	comb->sadb_comb_soft_allocations = 1;
6327 	comb->sadb_comb_hard_allocations = 1;
6328 	comb->sadb_comb_soft_bytes = 0;
6329 	comb->sadb_comb_hard_bytes = 0;
6330 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
6331 	comb->sadb_comb_soft_addtime = comb->sadb_comb_soft_addtime * 80 / 100;
6332 	comb->sadb_comb_soft_usetime = 28800;	/* 8 hours */
6333 	comb->sadb_comb_hard_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
6334 }
6335 
6336 /*
6337  * XXX reorder combinations by preference
6338  * XXX no idea if the user wants ESP authentication or not
6339  */
6340 static struct mbuf *
6341 key_getcomb_ealg(void)
6342 {
6343 	struct sadb_comb *comb;
6344 	const struct enc_xform *algo;
6345 	struct mbuf *result = NULL, *m, *n;
6346 	int encmin;
6347 	int i, off, o;
6348 	int totlen;
6349 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6350 
6351 	m = NULL;
6352 	for (i = 1; i <= SADB_EALG_MAX; i++) {
6353 		algo = enc_algorithm_lookup(i);
6354 		if (algo == NULL)
6355 			continue;
6356 
6357 		/* discard algorithms with key size smaller than system min */
6358 		if (_BITS(algo->maxkey) < V_ipsec_esp_keymin)
6359 			continue;
6360 		if (_BITS(algo->minkey) < V_ipsec_esp_keymin)
6361 			encmin = V_ipsec_esp_keymin;
6362 		else
6363 			encmin = _BITS(algo->minkey);
6364 
6365 		if (V_ipsec_esp_auth)
6366 			m = key_getcomb_ah();
6367 		else {
6368 			IPSEC_ASSERT(l <= MLEN,
6369 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6370 			MGET(m, M_NOWAIT, MT_DATA);
6371 			if (m) {
6372 				M_ALIGN(m, l);
6373 				m->m_len = l;
6374 				m->m_next = NULL;
6375 				bzero(mtod(m, caddr_t), m->m_len);
6376 			}
6377 		}
6378 		if (!m)
6379 			goto fail;
6380 
6381 		totlen = 0;
6382 		for (n = m; n; n = n->m_next)
6383 			totlen += n->m_len;
6384 		IPSEC_ASSERT((totlen % l) == 0, ("totlen=%u, l=%u", totlen, l));
6385 
6386 		for (off = 0; off < totlen; off += l) {
6387 			n = m_pulldown(m, off, l, &o);
6388 			if (!n) {
6389 				/* m is already freed */
6390 				goto fail;
6391 			}
6392 			comb = (struct sadb_comb *)(mtod(n, caddr_t) + o);
6393 			bzero(comb, sizeof(*comb));
6394 			key_getcomb_setlifetime(comb);
6395 			comb->sadb_comb_encrypt = i;
6396 			comb->sadb_comb_encrypt_minbits = encmin;
6397 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
6398 		}
6399 
6400 		if (!result)
6401 			result = m;
6402 		else
6403 			m_cat(result, m);
6404 	}
6405 
6406 	return result;
6407 
6408  fail:
6409 	if (result)
6410 		m_freem(result);
6411 	return NULL;
6412 }
6413 
6414 static void
6415 key_getsizes_ah(const struct auth_hash *ah, int alg, u_int16_t* min,
6416     u_int16_t* max)
6417 {
6418 
6419 	*min = *max = ah->hashsize;
6420 	if (ah->keysize == 0) {
6421 		/*
6422 		 * Transform takes arbitrary key size but algorithm
6423 		 * key size is restricted.  Enforce this here.
6424 		 */
6425 		switch (alg) {
6426 		case SADB_X_AALG_NULL:	*min = 1; *max = 256; break;
6427 		case SADB_X_AALG_SHA2_256: *min = *max = 32; break;
6428 		case SADB_X_AALG_SHA2_384: *min = *max = 48; break;
6429 		case SADB_X_AALG_SHA2_512: *min = *max = 64; break;
6430 		default:
6431 			DPRINTF(("%s: unknown AH algorithm %u\n",
6432 				__func__, alg));
6433 			break;
6434 		}
6435 	}
6436 }
6437 
6438 /*
6439  * XXX reorder combinations by preference
6440  */
6441 static struct mbuf *
6442 key_getcomb_ah(void)
6443 {
6444 	const struct auth_hash *algo;
6445 	struct sadb_comb *comb;
6446 	struct mbuf *m;
6447 	u_int16_t minkeysize, maxkeysize;
6448 	int i;
6449 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6450 
6451 	m = NULL;
6452 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6453 #if 1
6454 		/* we prefer HMAC algorithms, not old algorithms */
6455 		if (i != SADB_AALG_SHA1HMAC &&
6456 		    i != SADB_X_AALG_SHA2_256 &&
6457 		    i != SADB_X_AALG_SHA2_384 &&
6458 		    i != SADB_X_AALG_SHA2_512)
6459 			continue;
6460 #endif
6461 		algo = auth_algorithm_lookup(i);
6462 		if (!algo)
6463 			continue;
6464 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6465 		/* discard algorithms with key size smaller than system min */
6466 		if (_BITS(minkeysize) < V_ipsec_ah_keymin)
6467 			continue;
6468 
6469 		if (!m) {
6470 			IPSEC_ASSERT(l <= MLEN,
6471 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6472 			MGET(m, M_NOWAIT, MT_DATA);
6473 			if (m) {
6474 				M_ALIGN(m, l);
6475 				m->m_len = l;
6476 				m->m_next = NULL;
6477 			}
6478 		} else
6479 			M_PREPEND(m, l, M_NOWAIT);
6480 		if (!m)
6481 			return NULL;
6482 
6483 		comb = mtod(m, struct sadb_comb *);
6484 		bzero(comb, sizeof(*comb));
6485 		key_getcomb_setlifetime(comb);
6486 		comb->sadb_comb_auth = i;
6487 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6488 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6489 	}
6490 
6491 	return m;
6492 }
6493 
6494 /*
6495  * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6496  * XXX reorder combinations by preference
6497  */
6498 static struct mbuf *
6499 key_getcomb_ipcomp(void)
6500 {
6501 	const struct comp_algo *algo;
6502 	struct sadb_comb *comb;
6503 	struct mbuf *m;
6504 	int i;
6505 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6506 
6507 	m = NULL;
6508 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6509 		algo = comp_algorithm_lookup(i);
6510 		if (!algo)
6511 			continue;
6512 
6513 		if (!m) {
6514 			IPSEC_ASSERT(l <= MLEN,
6515 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6516 			MGET(m, M_NOWAIT, MT_DATA);
6517 			if (m) {
6518 				M_ALIGN(m, l);
6519 				m->m_len = l;
6520 				m->m_next = NULL;
6521 			}
6522 		} else
6523 			M_PREPEND(m, l, M_NOWAIT);
6524 		if (!m)
6525 			return NULL;
6526 
6527 		comb = mtod(m, struct sadb_comb *);
6528 		bzero(comb, sizeof(*comb));
6529 		key_getcomb_setlifetime(comb);
6530 		comb->sadb_comb_encrypt = i;
6531 		/* what should we set into sadb_comb_*_{min,max}bits? */
6532 	}
6533 
6534 	return m;
6535 }
6536 
6537 /*
6538  * XXX no way to pass mode (transport/tunnel) to userland
6539  * XXX replay checking?
6540  * XXX sysctl interface to ipsec_{ah,esp}_keymin
6541  */
6542 static struct mbuf *
6543 key_getprop(const struct secasindex *saidx)
6544 {
6545 	struct sadb_prop *prop;
6546 	struct mbuf *m, *n;
6547 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6548 	int totlen;
6549 
6550 	switch (saidx->proto)  {
6551 	case IPPROTO_ESP:
6552 		m = key_getcomb_ealg();
6553 		break;
6554 	case IPPROTO_AH:
6555 		m = key_getcomb_ah();
6556 		break;
6557 	case IPPROTO_IPCOMP:
6558 		m = key_getcomb_ipcomp();
6559 		break;
6560 	default:
6561 		return NULL;
6562 	}
6563 
6564 	if (!m)
6565 		return NULL;
6566 	M_PREPEND(m, l, M_NOWAIT);
6567 	if (!m)
6568 		return NULL;
6569 
6570 	totlen = 0;
6571 	for (n = m; n; n = n->m_next)
6572 		totlen += n->m_len;
6573 
6574 	prop = mtod(m, struct sadb_prop *);
6575 	bzero(prop, sizeof(*prop));
6576 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6577 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6578 	prop->sadb_prop_replay = 32;	/* XXX */
6579 
6580 	return m;
6581 }
6582 
6583 /*
6584  * SADB_ACQUIRE processing called by key_checkrequest() and key_acquire2().
6585  * send
6586  *   <base, SA, address(SD), (address(P)), x_policy,
6587  *       (identity(SD),) (sensitivity,) proposal>
6588  * to KMD, and expect to receive
6589  *   <base> with SADB_ACQUIRE if error occurred,
6590  * or
6591  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6592  * from KMD by PF_KEY.
6593  *
6594  * XXX x_policy is outside of RFC2367 (KAME extension).
6595  * XXX sensitivity is not supported.
6596  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6597  * see comment for key_getcomb_ipcomp().
6598  *
6599  * OUT:
6600  *    0     : succeed
6601  *    others: error number
6602  */
6603 static int
6604 key_acquire(const struct secasindex *saidx, struct secpolicy *sp)
6605 {
6606 	union sockaddr_union addr;
6607 	struct mbuf *result, *m;
6608 	uint32_t seq;
6609 	int error;
6610 	uint16_t ul_proto;
6611 	uint8_t mask, satype;
6612 
6613 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
6614 	satype = key_proto2satype(saidx->proto);
6615 	IPSEC_ASSERT(satype != 0, ("null satype, protocol %u", saidx->proto));
6616 
6617 	error = -1;
6618 	result = NULL;
6619 	ul_proto = IPSEC_ULPROTO_ANY;
6620 
6621 	/* Get seq number to check whether sending message or not. */
6622 	seq = key_getacq(saidx, &error);
6623 	if (seq == 0)
6624 		return (error);
6625 
6626 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0);
6627 	if (!m) {
6628 		error = ENOBUFS;
6629 		goto fail;
6630 	}
6631 	result = m;
6632 
6633 	/*
6634 	 * set sadb_address for saidx's.
6635 	 *
6636 	 * Note that if sp is supplied, then we're being called from
6637 	 * key_allocsa_policy() and should supply port and protocol
6638 	 * information.
6639 	 * XXXAE: why only TCP and UDP? ICMP and SCTP looks applicable too.
6640 	 * XXXAE: probably we can handle this in the ipsec[46]_allocsa().
6641 	 * XXXAE: it looks like we should save this info in the ACQ entry.
6642 	 */
6643 	if (sp != NULL && (sp->spidx.ul_proto == IPPROTO_TCP ||
6644 	    sp->spidx.ul_proto == IPPROTO_UDP))
6645 		ul_proto = sp->spidx.ul_proto;
6646 
6647 	addr = saidx->src;
6648 	mask = FULLMASK;
6649 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6650 		switch (sp->spidx.src.sa.sa_family) {
6651 		case AF_INET:
6652 			if (sp->spidx.src.sin.sin_port != IPSEC_PORT_ANY) {
6653 				addr.sin.sin_port = sp->spidx.src.sin.sin_port;
6654 				mask = sp->spidx.prefs;
6655 			}
6656 			break;
6657 		case AF_INET6:
6658 			if (sp->spidx.src.sin6.sin6_port != IPSEC_PORT_ANY) {
6659 				addr.sin6.sin6_port =
6660 				    sp->spidx.src.sin6.sin6_port;
6661 				mask = sp->spidx.prefs;
6662 			}
6663 			break;
6664 		default:
6665 			break;
6666 		}
6667 	}
6668 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &addr.sa, mask, ul_proto);
6669 	if (!m) {
6670 		error = ENOBUFS;
6671 		goto fail;
6672 	}
6673 	m_cat(result, m);
6674 
6675 	addr = saidx->dst;
6676 	mask = FULLMASK;
6677 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6678 		switch (sp->spidx.dst.sa.sa_family) {
6679 		case AF_INET:
6680 			if (sp->spidx.dst.sin.sin_port != IPSEC_PORT_ANY) {
6681 				addr.sin.sin_port = sp->spidx.dst.sin.sin_port;
6682 				mask = sp->spidx.prefd;
6683 			}
6684 			break;
6685 		case AF_INET6:
6686 			if (sp->spidx.dst.sin6.sin6_port != IPSEC_PORT_ANY) {
6687 				addr.sin6.sin6_port =
6688 				    sp->spidx.dst.sin6.sin6_port;
6689 				mask = sp->spidx.prefd;
6690 			}
6691 			break;
6692 		default:
6693 			break;
6694 		}
6695 	}
6696 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &addr.sa, mask, ul_proto);
6697 	if (!m) {
6698 		error = ENOBUFS;
6699 		goto fail;
6700 	}
6701 	m_cat(result, m);
6702 
6703 	/* XXX proxy address (optional) */
6704 
6705 	/*
6706 	 * Set sadb_x_policy. This is KAME extension to RFC2367.
6707 	 */
6708 	if (sp != NULL) {
6709 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6710 		    sp->priority);
6711 		if (!m) {
6712 			error = ENOBUFS;
6713 			goto fail;
6714 		}
6715 		m_cat(result, m);
6716 	}
6717 
6718 	/*
6719 	 * Set sadb_x_sa2 extension if saidx->reqid is not zero.
6720 	 * This is FreeBSD extension to RFC2367.
6721 	 */
6722 	if (saidx->reqid != 0) {
6723 		m = key_setsadbxsa2(saidx->mode, 0, saidx->reqid);
6724 		if (m == NULL) {
6725 			error = ENOBUFS;
6726 			goto fail;
6727 		}
6728 		m_cat(result, m);
6729 	}
6730 	/* XXX identity (optional) */
6731 #if 0
6732 	if (idexttype && fqdn) {
6733 		/* create identity extension (FQDN) */
6734 		struct sadb_ident *id;
6735 		int fqdnlen;
6736 
6737 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
6738 		id = (struct sadb_ident *)p;
6739 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6740 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6741 		id->sadb_ident_exttype = idexttype;
6742 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
6743 		bcopy(fqdn, id + 1, fqdnlen);
6744 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
6745 	}
6746 
6747 	if (idexttype) {
6748 		/* create identity extension (USERFQDN) */
6749 		struct sadb_ident *id;
6750 		int userfqdnlen;
6751 
6752 		if (userfqdn) {
6753 			/* +1 for terminating-NUL */
6754 			userfqdnlen = strlen(userfqdn) + 1;
6755 		} else
6756 			userfqdnlen = 0;
6757 		id = (struct sadb_ident *)p;
6758 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6759 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6760 		id->sadb_ident_exttype = idexttype;
6761 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
6762 		/* XXX is it correct? */
6763 		if (curproc && curproc->p_cred)
6764 			id->sadb_ident_id = curproc->p_cred->p_ruid;
6765 		if (userfqdn && userfqdnlen)
6766 			bcopy(userfqdn, id + 1, userfqdnlen);
6767 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
6768 	}
6769 #endif
6770 
6771 	/* XXX sensitivity (optional) */
6772 
6773 	/* create proposal/combination extension */
6774 	m = key_getprop(saidx);
6775 #if 0
6776 	/*
6777 	 * spec conformant: always attach proposal/combination extension,
6778 	 * the problem is that we have no way to attach it for ipcomp,
6779 	 * due to the way sadb_comb is declared in RFC2367.
6780 	 */
6781 	if (!m) {
6782 		error = ENOBUFS;
6783 		goto fail;
6784 	}
6785 	m_cat(result, m);
6786 #else
6787 	/*
6788 	 * outside of spec; make proposal/combination extension optional.
6789 	 */
6790 	if (m)
6791 		m_cat(result, m);
6792 #endif
6793 
6794 	if ((result->m_flags & M_PKTHDR) == 0) {
6795 		error = EINVAL;
6796 		goto fail;
6797 	}
6798 
6799 	if (result->m_len < sizeof(struct sadb_msg)) {
6800 		result = m_pullup(result, sizeof(struct sadb_msg));
6801 		if (result == NULL) {
6802 			error = ENOBUFS;
6803 			goto fail;
6804 		}
6805 	}
6806 
6807 	result->m_pkthdr.len = 0;
6808 	for (m = result; m; m = m->m_next)
6809 		result->m_pkthdr.len += m->m_len;
6810 
6811 	mtod(result, struct sadb_msg *)->sadb_msg_len =
6812 	    PFKEY_UNIT64(result->m_pkthdr.len);
6813 
6814 	KEYDBG(KEY_STAMP,
6815 	    printf("%s: SP(%p)\n", __func__, sp));
6816 	KEYDBG(KEY_DATA, kdebug_secasindex(saidx, NULL));
6817 
6818 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
6819 
6820  fail:
6821 	if (result)
6822 		m_freem(result);
6823 	return error;
6824 }
6825 
6826 static uint32_t
6827 key_newacq(const struct secasindex *saidx, int *perror)
6828 {
6829 	struct secacq *acq;
6830 	uint32_t seq;
6831 
6832 	acq = malloc(sizeof(*acq), M_IPSEC_SAQ, M_NOWAIT | M_ZERO);
6833 	if (acq == NULL) {
6834 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6835 		*perror = ENOBUFS;
6836 		return (0);
6837 	}
6838 
6839 	/* copy secindex */
6840 	bcopy(saidx, &acq->saidx, sizeof(acq->saidx));
6841 	acq->created = time_second;
6842 	acq->count = 0;
6843 
6844 	/* add to acqtree */
6845 	ACQ_LOCK();
6846 	seq = acq->seq = (V_acq_seq == ~0 ? 1 : ++V_acq_seq);
6847 	LIST_INSERT_HEAD(&V_acqtree, acq, chain);
6848 	LIST_INSERT_HEAD(ACQADDRHASH_HASH(saidx), acq, addrhash);
6849 	LIST_INSERT_HEAD(ACQSEQHASH_HASH(seq), acq, seqhash);
6850 	ACQ_UNLOCK();
6851 	*perror = 0;
6852 	return (seq);
6853 }
6854 
6855 static uint32_t
6856 key_getacq(const struct secasindex *saidx, int *perror)
6857 {
6858 	struct secacq *acq;
6859 	uint32_t seq;
6860 
6861 	ACQ_LOCK();
6862 	LIST_FOREACH(acq, ACQADDRHASH_HASH(saidx), addrhash) {
6863 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY)) {
6864 			if (acq->count > V_key_blockacq_count) {
6865 				/*
6866 				 * Reset counter and send message.
6867 				 * Also reset created time to keep ACQ for
6868 				 * this saidx.
6869 				 */
6870 				acq->created = time_second;
6871 				acq->count = 0;
6872 				seq = acq->seq;
6873 			} else {
6874 				/*
6875 				 * Increment counter and do nothing.
6876 				 * We send SADB_ACQUIRE message only
6877 				 * for each V_key_blockacq_count packet.
6878 				 */
6879 				acq->count++;
6880 				seq = 0;
6881 			}
6882 			break;
6883 		}
6884 	}
6885 	ACQ_UNLOCK();
6886 	if (acq != NULL) {
6887 		*perror = 0;
6888 		return (seq);
6889 	}
6890 	/* allocate new  entry */
6891 	return (key_newacq(saidx, perror));
6892 }
6893 
6894 static int
6895 key_acqreset(uint32_t seq)
6896 {
6897 	struct secacq *acq;
6898 
6899 	ACQ_LOCK();
6900 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6901 		if (acq->seq == seq) {
6902 			acq->count = 0;
6903 			acq->created = time_second;
6904 			break;
6905 		}
6906 	}
6907 	ACQ_UNLOCK();
6908 	if (acq == NULL)
6909 		return (ESRCH);
6910 	return (0);
6911 }
6912 /*
6913  * Mark ACQ entry as stale to remove it in key_flush_acq().
6914  * Called after successful SADB_GETSPI message.
6915  */
6916 static int
6917 key_acqdone(const struct secasindex *saidx, uint32_t seq)
6918 {
6919 	struct secacq *acq;
6920 
6921 	ACQ_LOCK();
6922 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6923 		if (acq->seq == seq)
6924 			break;
6925 	}
6926 	if (acq != NULL) {
6927 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY) == 0) {
6928 			ipseclog((LOG_DEBUG,
6929 			    "%s: Mismatched saidx for ACQ %u\n", __func__, seq));
6930 			acq = NULL;
6931 		} else {
6932 			acq->created = 0;
6933 		}
6934 	} else {
6935 		ipseclog((LOG_DEBUG,
6936 		    "%s: ACQ %u is not found.\n", __func__, seq));
6937 	}
6938 	ACQ_UNLOCK();
6939 	if (acq == NULL)
6940 		return (ESRCH);
6941 	return (0);
6942 }
6943 
6944 static struct secspacq *
6945 key_newspacq(struct secpolicyindex *spidx)
6946 {
6947 	struct secspacq *acq;
6948 
6949 	/* get new entry */
6950 	acq = malloc(sizeof(struct secspacq), M_IPSEC_SAQ, M_NOWAIT|M_ZERO);
6951 	if (acq == NULL) {
6952 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6953 		return NULL;
6954 	}
6955 
6956 	/* copy secindex */
6957 	bcopy(spidx, &acq->spidx, sizeof(acq->spidx));
6958 	acq->created = time_second;
6959 	acq->count = 0;
6960 
6961 	/* add to spacqtree */
6962 	SPACQ_LOCK();
6963 	LIST_INSERT_HEAD(&V_spacqtree, acq, chain);
6964 	SPACQ_UNLOCK();
6965 
6966 	return acq;
6967 }
6968 
6969 static struct secspacq *
6970 key_getspacq(struct secpolicyindex *spidx)
6971 {
6972 	struct secspacq *acq;
6973 
6974 	SPACQ_LOCK();
6975 	LIST_FOREACH(acq, &V_spacqtree, chain) {
6976 		if (key_cmpspidx_exactly(spidx, &acq->spidx)) {
6977 			/* NB: return holding spacq_lock */
6978 			return acq;
6979 		}
6980 	}
6981 	SPACQ_UNLOCK();
6982 
6983 	return NULL;
6984 }
6985 
6986 /*
6987  * SADB_ACQUIRE processing,
6988  * in first situation, is receiving
6989  *   <base>
6990  * from the ikmpd, and clear sequence of its secasvar entry.
6991  *
6992  * In second situation, is receiving
6993  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6994  * from a user land process, and return
6995  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6996  * to the socket.
6997  *
6998  * m will always be freed.
6999  */
7000 static int
7001 key_acquire2(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7002 {
7003 	SAHTREE_RLOCK_TRACKER;
7004 	struct sadb_address *src0, *dst0;
7005 	struct secasindex saidx;
7006 	struct secashead *sah;
7007 	uint32_t reqid;
7008 	int error;
7009 	uint8_t mode, proto;
7010 
7011 	IPSEC_ASSERT(so != NULL, ("null socket"));
7012 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7013 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7014 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7015 
7016 	/*
7017 	 * Error message from KMd.
7018 	 * We assume that if error was occurred in IKEd, the length of PFKEY
7019 	 * message is equal to the size of sadb_msg structure.
7020 	 * We do not raise error even if error occurred in this function.
7021 	 */
7022 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
7023 		/* check sequence number */
7024 		if (mhp->msg->sadb_msg_seq == 0 ||
7025 		    mhp->msg->sadb_msg_errno == 0) {
7026 			ipseclog((LOG_DEBUG, "%s: must specify sequence "
7027 				"number and errno.\n", __func__));
7028 		} else {
7029 			/*
7030 			 * IKEd reported that error occurred.
7031 			 * XXXAE: what it expects from the kernel?
7032 			 * Probably we should send SADB_ACQUIRE again?
7033 			 * If so, reset ACQ's state.
7034 			 * XXXAE: it looks useless.
7035 			 */
7036 			key_acqreset(mhp->msg->sadb_msg_seq);
7037 		}
7038 		m_freem(m);
7039 		return (0);
7040 	}
7041 
7042 	/*
7043 	 * This message is from user land.
7044 	 */
7045 
7046 	/* map satype to proto */
7047 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7048 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7049 		    __func__));
7050 		return key_senderror(so, m, EINVAL);
7051 	}
7052 
7053 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
7054 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
7055 	    SADB_CHECKHDR(mhp, SADB_EXT_PROPOSAL)) {
7056 		ipseclog((LOG_DEBUG,
7057 		    "%s: invalid message: missing required header.\n",
7058 		    __func__));
7059 		return key_senderror(so, m, EINVAL);
7060 	}
7061 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
7062 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
7063 	    SADB_CHECKLEN(mhp, SADB_EXT_PROPOSAL)) {
7064 		ipseclog((LOG_DEBUG,
7065 		    "%s: invalid message: wrong header size.\n", __func__));
7066 		return key_senderror(so, m, EINVAL);
7067 	}
7068 
7069 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
7070 		mode = IPSEC_MODE_ANY;
7071 		reqid = 0;
7072 	} else {
7073 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
7074 			ipseclog((LOG_DEBUG,
7075 			    "%s: invalid message: wrong header size.\n",
7076 			    __func__));
7077 			return key_senderror(so, m, EINVAL);
7078 		}
7079 		mode = ((struct sadb_x_sa2 *)
7080 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
7081 		reqid = ((struct sadb_x_sa2 *)
7082 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
7083 	}
7084 
7085 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
7086 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
7087 
7088 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
7089 	    (struct sockaddr *)(dst0 + 1));
7090 	if (error != 0) {
7091 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
7092 		return key_senderror(so, m, EINVAL);
7093 	}
7094 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
7095 
7096 	/* get a SA index */
7097 	SAHTREE_RLOCK();
7098 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
7099 		if (key_cmpsaidx(&sah->saidx, &saidx, CMP_MODE_REQID))
7100 			break;
7101 	}
7102 	SAHTREE_RUNLOCK();
7103 	if (sah != NULL) {
7104 		ipseclog((LOG_DEBUG, "%s: a SA exists already.\n", __func__));
7105 		return key_senderror(so, m, EEXIST);
7106 	}
7107 
7108 	error = key_acquire(&saidx, NULL);
7109 	if (error != 0) {
7110 		ipseclog((LOG_DEBUG,
7111 		    "%s: error %d returned from key_acquire()\n",
7112 			__func__, error));
7113 		return key_senderror(so, m, error);
7114 	}
7115 	m_freem(m);
7116 	return (0);
7117 }
7118 
7119 /*
7120  * SADB_REGISTER processing.
7121  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
7122  * receive
7123  *   <base>
7124  * from the ikmpd, and register a socket to send PF_KEY messages,
7125  * and send
7126  *   <base, supported>
7127  * to KMD by PF_KEY.
7128  * If socket is detached, must free from regnode.
7129  *
7130  * m will always be freed.
7131  */
7132 static int
7133 key_register(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7134 {
7135 	struct secreg *reg, *newreg = NULL;
7136 
7137 	IPSEC_ASSERT(so != NULL, ("null socket"));
7138 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7139 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7140 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7141 
7142 	/* check for invalid register message */
7143 	if (mhp->msg->sadb_msg_satype >= sizeof(V_regtree)/sizeof(V_regtree[0]))
7144 		return key_senderror(so, m, EINVAL);
7145 
7146 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
7147 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
7148 		goto setmsg;
7149 
7150 	/* check whether existing or not */
7151 	REGTREE_LOCK();
7152 	LIST_FOREACH(reg, &V_regtree[mhp->msg->sadb_msg_satype], chain) {
7153 		if (reg->so == so) {
7154 			REGTREE_UNLOCK();
7155 			ipseclog((LOG_DEBUG, "%s: socket exists already.\n",
7156 				__func__));
7157 			return key_senderror(so, m, EEXIST);
7158 		}
7159 	}
7160 
7161 	/* create regnode */
7162 	newreg =  malloc(sizeof(struct secreg), M_IPSEC_SAR, M_NOWAIT|M_ZERO);
7163 	if (newreg == NULL) {
7164 		REGTREE_UNLOCK();
7165 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7166 		return key_senderror(so, m, ENOBUFS);
7167 	}
7168 
7169 	newreg->so = so;
7170 	((struct keycb *)(so->so_pcb))->kp_registered++;
7171 
7172 	/* add regnode to regtree. */
7173 	LIST_INSERT_HEAD(&V_regtree[mhp->msg->sadb_msg_satype], newreg, chain);
7174 	REGTREE_UNLOCK();
7175 
7176   setmsg:
7177     {
7178 	struct mbuf *n;
7179 	struct sadb_msg *newmsg;
7180 	struct sadb_supported *sup;
7181 	u_int len, alen, elen;
7182 	int off;
7183 	int i;
7184 	struct sadb_alg *alg;
7185 
7186 	/* create new sadb_msg to reply. */
7187 	alen = 0;
7188 	for (i = 1; i <= SADB_AALG_MAX; i++) {
7189 		if (auth_algorithm_lookup(i))
7190 			alen += sizeof(struct sadb_alg);
7191 	}
7192 	if (alen)
7193 		alen += sizeof(struct sadb_supported);
7194 	elen = 0;
7195 	for (i = 1; i <= SADB_EALG_MAX; i++) {
7196 		if (enc_algorithm_lookup(i))
7197 			elen += sizeof(struct sadb_alg);
7198 	}
7199 	if (elen)
7200 		elen += sizeof(struct sadb_supported);
7201 
7202 	len = sizeof(struct sadb_msg) + alen + elen;
7203 
7204 	if (len > MCLBYTES)
7205 		return key_senderror(so, m, ENOBUFS);
7206 
7207 	n = key_mget(len);
7208 	if (n == NULL)
7209 		return key_senderror(so, m, ENOBUFS);
7210 
7211 	n->m_pkthdr.len = n->m_len = len;
7212 	n->m_next = NULL;
7213 	off = 0;
7214 
7215 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
7216 	newmsg = mtod(n, struct sadb_msg *);
7217 	newmsg->sadb_msg_errno = 0;
7218 	newmsg->sadb_msg_len = PFKEY_UNIT64(len);
7219 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7220 
7221 	/* for authentication algorithm */
7222 	if (alen) {
7223 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7224 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7225 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7226 		off += PFKEY_ALIGN8(sizeof(*sup));
7227 
7228 		for (i = 1; i <= SADB_AALG_MAX; i++) {
7229 			const struct auth_hash *aalgo;
7230 			u_int16_t minkeysize, maxkeysize;
7231 
7232 			aalgo = auth_algorithm_lookup(i);
7233 			if (!aalgo)
7234 				continue;
7235 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7236 			alg->sadb_alg_id = i;
7237 			alg->sadb_alg_ivlen = 0;
7238 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7239 			alg->sadb_alg_minbits = _BITS(minkeysize);
7240 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7241 			off += PFKEY_ALIGN8(sizeof(*alg));
7242 		}
7243 	}
7244 
7245 	/* for encryption algorithm */
7246 	if (elen) {
7247 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7248 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7249 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7250 		off += PFKEY_ALIGN8(sizeof(*sup));
7251 
7252 		for (i = 1; i <= SADB_EALG_MAX; i++) {
7253 			const struct enc_xform *ealgo;
7254 
7255 			ealgo = enc_algorithm_lookup(i);
7256 			if (!ealgo)
7257 				continue;
7258 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7259 			alg->sadb_alg_id = i;
7260 			alg->sadb_alg_ivlen = ealgo->ivsize;
7261 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7262 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7263 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7264 		}
7265 	}
7266 
7267 	IPSEC_ASSERT(off == len,
7268 		("length assumption failed (off %u len %u)", off, len));
7269 
7270 	m_freem(m);
7271 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7272     }
7273 }
7274 
7275 /*
7276  * free secreg entry registered.
7277  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7278  */
7279 void
7280 key_freereg(struct socket *so)
7281 {
7282 	struct secreg *reg;
7283 	int i;
7284 
7285 	IPSEC_ASSERT(so != NULL, ("NULL so"));
7286 
7287 	/*
7288 	 * check whether existing or not.
7289 	 * check all type of SA, because there is a potential that
7290 	 * one socket is registered to multiple type of SA.
7291 	 */
7292 	REGTREE_LOCK();
7293 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7294 		LIST_FOREACH(reg, &V_regtree[i], chain) {
7295 			if (reg->so == so && __LIST_CHAINED(reg)) {
7296 				LIST_REMOVE(reg, chain);
7297 				free(reg, M_IPSEC_SAR);
7298 				break;
7299 			}
7300 		}
7301 	}
7302 	REGTREE_UNLOCK();
7303 }
7304 
7305 /*
7306  * SADB_EXPIRE processing
7307  * send
7308  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7309  * to KMD by PF_KEY.
7310  * NOTE: We send only soft lifetime extension.
7311  *
7312  * OUT:	0	: succeed
7313  *	others	: error number
7314  */
7315 static int
7316 key_expire(struct secasvar *sav, int hard)
7317 {
7318 	struct mbuf *result = NULL, *m;
7319 	struct sadb_lifetime *lt;
7320 	uint32_t replay_count;
7321 	int error, len;
7322 	uint8_t satype;
7323 
7324 	SECASVAR_RLOCK_TRACKER;
7325 
7326 	IPSEC_ASSERT (sav != NULL, ("null sav"));
7327 	IPSEC_ASSERT (sav->sah != NULL, ("null sa header"));
7328 
7329 	KEYDBG(KEY_STAMP,
7330 	    printf("%s: SA(%p) expired %s lifetime\n", __func__,
7331 		sav, hard ? "hard": "soft"));
7332 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
7333 	/* set msg header */
7334 	satype = key_proto2satype(sav->sah->saidx.proto);
7335 	IPSEC_ASSERT(satype != 0, ("invalid proto, satype %u", satype));
7336 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, sav->refcnt);
7337 	if (!m) {
7338 		error = ENOBUFS;
7339 		goto fail;
7340 	}
7341 	result = m;
7342 
7343 	/* create SA extension */
7344 	m = key_setsadbsa(sav);
7345 	if (!m) {
7346 		error = ENOBUFS;
7347 		goto fail;
7348 	}
7349 	m_cat(result, m);
7350 
7351 	/* create SA extension */
7352 	SECASVAR_RLOCK(sav);
7353 	replay_count = sav->replay ? sav->replay->count : 0;
7354 	SECASVAR_RUNLOCK(sav);
7355 
7356 	m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
7357 			sav->sah->saidx.reqid);
7358 	if (!m) {
7359 		error = ENOBUFS;
7360 		goto fail;
7361 	}
7362 	m_cat(result, m);
7363 
7364 	if (sav->replay && sav->replay->wsize > UINT8_MAX) {
7365 		m = key_setsadbxsareplay(sav->replay->wsize);
7366 		if (!m) {
7367 			error = ENOBUFS;
7368 			goto fail;
7369 		}
7370 		m_cat(result, m);
7371 	}
7372 
7373 	/* create lifetime extension (current and soft) */
7374 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7375 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
7376 	if (m == NULL) {
7377 		error = ENOBUFS;
7378 		goto fail;
7379 	}
7380 	m_align(m, len);
7381 	m->m_len = len;
7382 	bzero(mtod(m, caddr_t), len);
7383 	lt = mtod(m, struct sadb_lifetime *);
7384 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7385 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7386 	lt->sadb_lifetime_allocations =
7387 	    (uint32_t)counter_u64_fetch(sav->lft_c_allocations);
7388 	lt->sadb_lifetime_bytes =
7389 	    counter_u64_fetch(sav->lft_c_bytes);
7390 	lt->sadb_lifetime_addtime = sav->created;
7391 	lt->sadb_lifetime_usetime = sav->firstused;
7392 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
7393 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7394 	if (hard) {
7395 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
7396 		lt->sadb_lifetime_allocations = sav->lft_h->allocations;
7397 		lt->sadb_lifetime_bytes = sav->lft_h->bytes;
7398 		lt->sadb_lifetime_addtime = sav->lft_h->addtime;
7399 		lt->sadb_lifetime_usetime = sav->lft_h->usetime;
7400 	} else {
7401 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_SOFT;
7402 		lt->sadb_lifetime_allocations = sav->lft_s->allocations;
7403 		lt->sadb_lifetime_bytes = sav->lft_s->bytes;
7404 		lt->sadb_lifetime_addtime = sav->lft_s->addtime;
7405 		lt->sadb_lifetime_usetime = sav->lft_s->usetime;
7406 	}
7407 	m_cat(result, m);
7408 
7409 	/* set sadb_address for source */
7410 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
7411 	    &sav->sah->saidx.src.sa,
7412 	    FULLMASK, IPSEC_ULPROTO_ANY);
7413 	if (!m) {
7414 		error = ENOBUFS;
7415 		goto fail;
7416 	}
7417 	m_cat(result, m);
7418 
7419 	/* set sadb_address for destination */
7420 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
7421 	    &sav->sah->saidx.dst.sa,
7422 	    FULLMASK, IPSEC_ULPROTO_ANY);
7423 	if (!m) {
7424 		error = ENOBUFS;
7425 		goto fail;
7426 	}
7427 	m_cat(result, m);
7428 
7429 	/*
7430 	 * XXX-BZ Handle NAT-T extensions here.
7431 	 * XXXAE: it doesn't seem quite useful. IKEs should not depend on
7432 	 * this information, we report only significant SA fields.
7433 	 */
7434 
7435 	if ((result->m_flags & M_PKTHDR) == 0) {
7436 		error = EINVAL;
7437 		goto fail;
7438 	}
7439 
7440 	if (result->m_len < sizeof(struct sadb_msg)) {
7441 		result = m_pullup(result, sizeof(struct sadb_msg));
7442 		if (result == NULL) {
7443 			error = ENOBUFS;
7444 			goto fail;
7445 		}
7446 	}
7447 
7448 	result->m_pkthdr.len = 0;
7449 	for (m = result; m; m = m->m_next)
7450 		result->m_pkthdr.len += m->m_len;
7451 
7452 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7453 	    PFKEY_UNIT64(result->m_pkthdr.len);
7454 
7455 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7456 
7457  fail:
7458 	if (result)
7459 		m_freem(result);
7460 	return error;
7461 }
7462 
7463 static void
7464 key_freesah_flushed(struct secashead_queue *flushq)
7465 {
7466 	struct secashead *sah, *nextsah;
7467 	struct secasvar *sav, *nextsav;
7468 
7469 	sah = TAILQ_FIRST(flushq);
7470 	while (sah != NULL) {
7471 		sav = TAILQ_FIRST(&sah->savtree_larval);
7472 		while (sav != NULL) {
7473 			nextsav = TAILQ_NEXT(sav, chain);
7474 			TAILQ_REMOVE(&sah->savtree_larval, sav, chain);
7475 			key_freesav(&sav); /* release last reference */
7476 			key_freesah(&sah); /* release reference from SAV */
7477 			sav = nextsav;
7478 		}
7479 		sav = TAILQ_FIRST(&sah->savtree_alive);
7480 		while (sav != NULL) {
7481 			nextsav = TAILQ_NEXT(sav, chain);
7482 			TAILQ_REMOVE(&sah->savtree_alive, sav, chain);
7483 			key_freesav(&sav); /* release last reference */
7484 			key_freesah(&sah); /* release reference from SAV */
7485 			sav = nextsav;
7486 		}
7487 		nextsah = TAILQ_NEXT(sah, chain);
7488 		key_freesah(&sah);	/* release last reference */
7489 		sah = nextsah;
7490 	}
7491 }
7492 
7493 /*
7494  * SADB_FLUSH processing
7495  * receive
7496  *   <base>
7497  * from the ikmpd, and free all entries in secastree.
7498  * and send,
7499  *   <base>
7500  * to the ikmpd.
7501  * NOTE: to do is only marking SADB_SASTATE_DEAD.
7502  *
7503  * m will always be freed.
7504  */
7505 static int
7506 key_flush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7507 {
7508 	struct secashead_queue flushq;
7509 	struct sadb_msg *newmsg;
7510 	struct secashead *sah, *nextsah;
7511 	struct secasvar *sav;
7512 	uint8_t proto;
7513 	int i;
7514 
7515 	IPSEC_ASSERT(so != NULL, ("null socket"));
7516 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7517 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7518 
7519 	/* map satype to proto */
7520 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7521 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7522 			__func__));
7523 		return key_senderror(so, m, EINVAL);
7524 	}
7525 	KEYDBG(KEY_STAMP,
7526 	    printf("%s: proto %u\n", __func__, proto));
7527 
7528 	TAILQ_INIT(&flushq);
7529 	if (proto == IPSEC_PROTO_ANY) {
7530 		/* no SATYPE specified, i.e. flushing all SA. */
7531 		SAHTREE_WLOCK();
7532 		/* Move all SAHs into flushq */
7533 		TAILQ_CONCAT(&flushq, &V_sahtree, chain);
7534 		/* Flush all buckets in SPI hash */
7535 		for (i = 0; i < V_savhash_mask + 1; i++)
7536 			LIST_INIT(&V_savhashtbl[i]);
7537 		/* Flush all buckets in SAHADDRHASH */
7538 		for (i = 0; i < V_sahaddrhash_mask + 1; i++)
7539 			LIST_INIT(&V_sahaddrhashtbl[i]);
7540 		/* Mark all SAHs as unlinked */
7541 		TAILQ_FOREACH(sah, &flushq, chain) {
7542 			sah->state = SADB_SASTATE_DEAD;
7543 			/*
7544 			 * Callout handler makes its job using
7545 			 * RLOCK and drain queues. In case, when this
7546 			 * function will be called just before it
7547 			 * acquires WLOCK, we need to mark SAs as
7548 			 * unlinked to prevent second unlink.
7549 			 */
7550 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7551 				sav->state = SADB_SASTATE_DEAD;
7552 			}
7553 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7554 				sav->state = SADB_SASTATE_DEAD;
7555 			}
7556 		}
7557 		SAHTREE_WUNLOCK();
7558 	} else {
7559 		SAHTREE_WLOCK();
7560 		sah = TAILQ_FIRST(&V_sahtree);
7561 		while (sah != NULL) {
7562 			IPSEC_ASSERT(sah->state != SADB_SASTATE_DEAD,
7563 			    ("DEAD SAH %p in SADB_FLUSH", sah));
7564 			nextsah = TAILQ_NEXT(sah, chain);
7565 			if (sah->saidx.proto != proto) {
7566 				sah = nextsah;
7567 				continue;
7568 			}
7569 			sah->state = SADB_SASTATE_DEAD;
7570 			TAILQ_REMOVE(&V_sahtree, sah, chain);
7571 			LIST_REMOVE(sah, addrhash);
7572 			/* Unlink all SAs from SPI hash */
7573 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7574 				LIST_REMOVE(sav, spihash);
7575 				sav->state = SADB_SASTATE_DEAD;
7576 			}
7577 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7578 				LIST_REMOVE(sav, spihash);
7579 				sav->state = SADB_SASTATE_DEAD;
7580 			}
7581 			/* Add SAH into flushq */
7582 			TAILQ_INSERT_HEAD(&flushq, sah, chain);
7583 			sah = nextsah;
7584 		}
7585 		SAHTREE_WUNLOCK();
7586 	}
7587 
7588 	key_freesah_flushed(&flushq);
7589 	/* Free all queued SAs and SAHs */
7590 	if (m->m_len < sizeof(struct sadb_msg) ||
7591 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7592 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7593 		return key_senderror(so, m, ENOBUFS);
7594 	}
7595 
7596 	if (m->m_next)
7597 		m_freem(m->m_next);
7598 	m->m_next = NULL;
7599 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7600 	newmsg = mtod(m, struct sadb_msg *);
7601 	newmsg->sadb_msg_errno = 0;
7602 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7603 
7604 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7605 }
7606 
7607 /*
7608  * SADB_DUMP processing
7609  * dump all entries including status of DEAD in SAD.
7610  * receive
7611  *   <base>
7612  * from the ikmpd, and dump all secasvar leaves
7613  * and send,
7614  *   <base> .....
7615  * to the ikmpd.
7616  *
7617  * m will always be freed.
7618  */
7619 static int
7620 key_dump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7621 {
7622 	SAHTREE_RLOCK_TRACKER;
7623 	struct secashead *sah;
7624 	struct secasvar *sav;
7625 	struct mbuf *n;
7626 	uint32_t cnt;
7627 	uint8_t proto, satype;
7628 
7629 	IPSEC_ASSERT(so != NULL, ("null socket"));
7630 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7631 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7632 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7633 
7634 	/* map satype to proto */
7635 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7636 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7637 		    __func__));
7638 		return key_senderror(so, m, EINVAL);
7639 	}
7640 
7641 	/* count sav entries to be sent to the userland. */
7642 	cnt = 0;
7643 	SAHTREE_RLOCK();
7644 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7645 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7646 		    proto != sah->saidx.proto)
7647 			continue;
7648 
7649 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain)
7650 			cnt++;
7651 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain)
7652 			cnt++;
7653 	}
7654 
7655 	if (cnt == 0) {
7656 		SAHTREE_RUNLOCK();
7657 		return key_senderror(so, m, ENOENT);
7658 	}
7659 
7660 	/* send this to the userland, one at a time. */
7661 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7662 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7663 		    proto != sah->saidx.proto)
7664 			continue;
7665 
7666 		/* map proto to satype */
7667 		if ((satype = key_proto2satype(sah->saidx.proto)) == 0) {
7668 			SAHTREE_RUNLOCK();
7669 			ipseclog((LOG_DEBUG, "%s: there was invalid proto in "
7670 			    "SAD.\n", __func__));
7671 			return key_senderror(so, m, EINVAL);
7672 		}
7673 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7674 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7675 			    --cnt, mhp->msg->sadb_msg_pid);
7676 			if (n == NULL) {
7677 				SAHTREE_RUNLOCK();
7678 				return key_senderror(so, m, ENOBUFS);
7679 			}
7680 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7681 		}
7682 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7683 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7684 			    --cnt, mhp->msg->sadb_msg_pid);
7685 			if (n == NULL) {
7686 				SAHTREE_RUNLOCK();
7687 				return key_senderror(so, m, ENOBUFS);
7688 			}
7689 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7690 		}
7691 	}
7692 	SAHTREE_RUNLOCK();
7693 	m_freem(m);
7694 	return (0);
7695 }
7696 /*
7697  * SADB_X_PROMISC processing
7698  *
7699  * m will always be freed.
7700  */
7701 static int
7702 key_promisc(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7703 {
7704 	int olen;
7705 
7706 	IPSEC_ASSERT(so != NULL, ("null socket"));
7707 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7708 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7709 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7710 
7711 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7712 
7713 	if (olen < sizeof(struct sadb_msg)) {
7714 #if 1
7715 		return key_senderror(so, m, EINVAL);
7716 #else
7717 		m_freem(m);
7718 		return 0;
7719 #endif
7720 	} else if (olen == sizeof(struct sadb_msg)) {
7721 		/* enable/disable promisc mode */
7722 		struct keycb *kp;
7723 
7724 		if ((kp = so->so_pcb) == NULL)
7725 			return key_senderror(so, m, EINVAL);
7726 		mhp->msg->sadb_msg_errno = 0;
7727 		switch (mhp->msg->sadb_msg_satype) {
7728 		case 0:
7729 		case 1:
7730 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
7731 			break;
7732 		default:
7733 			return key_senderror(so, m, EINVAL);
7734 		}
7735 
7736 		/* send the original message back to everyone */
7737 		mhp->msg->sadb_msg_errno = 0;
7738 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7739 	} else {
7740 		/* send packet as is */
7741 
7742 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
7743 
7744 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
7745 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7746 	}
7747 }
7748 
7749 static int (*key_typesw[])(struct socket *, struct mbuf *,
7750     const struct sadb_msghdr *) = {
7751 	[SADB_RESERVED] =	NULL,
7752 	[SADB_GETSPI] =		key_getspi,
7753 	[SADB_UPDATE] =		key_update,
7754 	[SADB_ADD] =		key_add,
7755 	[SADB_DELETE] =		key_delete,
7756 	[SADB_GET] =		key_get,
7757 	[SADB_ACQUIRE] =	key_acquire2,
7758 	[SADB_REGISTER] =	key_register,
7759 	[SADB_EXPIRE] =		NULL,
7760 	[SADB_FLUSH] =		key_flush,
7761 	[SADB_DUMP] =		key_dump,
7762 	[SADB_X_PROMISC] =	key_promisc,
7763 	[SADB_X_PCHANGE] =	NULL,
7764 	[SADB_X_SPDUPDATE] =	key_spdadd,
7765 	[SADB_X_SPDADD] =	key_spdadd,
7766 	[SADB_X_SPDDELETE] =	key_spddelete,
7767 	[SADB_X_SPDGET] =	key_spdget,
7768 	[SADB_X_SPDACQUIRE] =	NULL,
7769 	[SADB_X_SPDDUMP] =	key_spddump,
7770 	[SADB_X_SPDFLUSH] =	key_spdflush,
7771 	[SADB_X_SPDSETIDX] =	key_spdadd,
7772 	[SADB_X_SPDEXPIRE] =	NULL,
7773 	[SADB_X_SPDDELETE2] =	key_spddelete2,
7774 };
7775 
7776 /*
7777  * parse sadb_msg buffer to process PFKEYv2,
7778  * and create a data to response if needed.
7779  * I think to be dealed with mbuf directly.
7780  * IN:
7781  *     msgp  : pointer to pointer to a received buffer pulluped.
7782  *             This is rewrited to response.
7783  *     so    : pointer to socket.
7784  * OUT:
7785  *    length for buffer to send to user process.
7786  */
7787 int
7788 key_parse(struct mbuf *m, struct socket *so)
7789 {
7790 	struct sadb_msg *msg;
7791 	struct sadb_msghdr mh;
7792 	u_int orglen;
7793 	int error;
7794 	int target;
7795 
7796 	IPSEC_ASSERT(so != NULL, ("null socket"));
7797 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7798 
7799 	if (m->m_len < sizeof(struct sadb_msg)) {
7800 		m = m_pullup(m, sizeof(struct sadb_msg));
7801 		if (!m)
7802 			return ENOBUFS;
7803 	}
7804 	msg = mtod(m, struct sadb_msg *);
7805 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
7806 	target = KEY_SENDUP_ONE;
7807 
7808 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len != orglen) {
7809 		ipseclog((LOG_DEBUG, "%s: invalid message length.\n",__func__));
7810 		PFKEYSTAT_INC(out_invlen);
7811 		error = EINVAL;
7812 		goto senderror;
7813 	}
7814 
7815 	if (msg->sadb_msg_version != PF_KEY_V2) {
7816 		ipseclog((LOG_DEBUG, "%s: PF_KEY version %u is mismatched.\n",
7817 		    __func__, msg->sadb_msg_version));
7818 		PFKEYSTAT_INC(out_invver);
7819 		error = EINVAL;
7820 		goto senderror;
7821 	}
7822 
7823 	if (msg->sadb_msg_type > SADB_MAX) {
7824 		ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7825 		    __func__, msg->sadb_msg_type));
7826 		PFKEYSTAT_INC(out_invmsgtype);
7827 		error = EINVAL;
7828 		goto senderror;
7829 	}
7830 
7831 	/* for old-fashioned code - should be nuked */
7832 	if (m->m_pkthdr.len > MCLBYTES) {
7833 		m_freem(m);
7834 		return ENOBUFS;
7835 	}
7836 	if (m->m_next) {
7837 		struct mbuf *n;
7838 
7839 		n = key_mget(m->m_pkthdr.len);
7840 		if (n == NULL) {
7841 			m_freem(m);
7842 			return ENOBUFS;
7843 		}
7844 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
7845 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
7846 		n->m_next = NULL;
7847 		m_freem(m);
7848 		m = n;
7849 	}
7850 
7851 	/* align the mbuf chain so that extensions are in contiguous region. */
7852 	error = key_align(m, &mh);
7853 	if (error)
7854 		return error;
7855 
7856 	msg = mh.msg;
7857 
7858 	/* We use satype as scope mask for spddump */
7859 	if (msg->sadb_msg_type == SADB_X_SPDDUMP) {
7860 		switch (msg->sadb_msg_satype) {
7861 		case IPSEC_POLICYSCOPE_ANY:
7862 		case IPSEC_POLICYSCOPE_GLOBAL:
7863 		case IPSEC_POLICYSCOPE_IFNET:
7864 		case IPSEC_POLICYSCOPE_PCB:
7865 			break;
7866 		default:
7867 			ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7868 			    __func__, msg->sadb_msg_type));
7869 			PFKEYSTAT_INC(out_invsatype);
7870 			error = EINVAL;
7871 			goto senderror;
7872 		}
7873 	} else {
7874 		switch (msg->sadb_msg_satype) { /* check SA type */
7875 		case SADB_SATYPE_UNSPEC:
7876 			switch (msg->sadb_msg_type) {
7877 			case SADB_GETSPI:
7878 			case SADB_UPDATE:
7879 			case SADB_ADD:
7880 			case SADB_DELETE:
7881 			case SADB_GET:
7882 			case SADB_ACQUIRE:
7883 			case SADB_EXPIRE:
7884 				ipseclog((LOG_DEBUG, "%s: must specify satype "
7885 				    "when msg type=%u.\n", __func__,
7886 				    msg->sadb_msg_type));
7887 				PFKEYSTAT_INC(out_invsatype);
7888 				error = EINVAL;
7889 				goto senderror;
7890 			}
7891 			break;
7892 		case SADB_SATYPE_AH:
7893 		case SADB_SATYPE_ESP:
7894 		case SADB_X_SATYPE_IPCOMP:
7895 		case SADB_X_SATYPE_TCPSIGNATURE:
7896 			switch (msg->sadb_msg_type) {
7897 			case SADB_X_SPDADD:
7898 			case SADB_X_SPDDELETE:
7899 			case SADB_X_SPDGET:
7900 			case SADB_X_SPDFLUSH:
7901 			case SADB_X_SPDSETIDX:
7902 			case SADB_X_SPDUPDATE:
7903 			case SADB_X_SPDDELETE2:
7904 				ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7905 				    __func__, msg->sadb_msg_type));
7906 				PFKEYSTAT_INC(out_invsatype);
7907 				error = EINVAL;
7908 				goto senderror;
7909 			}
7910 			break;
7911 		case SADB_SATYPE_RSVP:
7912 		case SADB_SATYPE_OSPFV2:
7913 		case SADB_SATYPE_RIPV2:
7914 		case SADB_SATYPE_MIP:
7915 			ipseclog((LOG_DEBUG, "%s: type %u isn't supported.\n",
7916 			    __func__, msg->sadb_msg_satype));
7917 			PFKEYSTAT_INC(out_invsatype);
7918 			error = EOPNOTSUPP;
7919 			goto senderror;
7920 		case 1:	/* XXX: What does it do? */
7921 			if (msg->sadb_msg_type == SADB_X_PROMISC)
7922 				break;
7923 			/*FALLTHROUGH*/
7924 		default:
7925 			ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7926 			    __func__, msg->sadb_msg_satype));
7927 			PFKEYSTAT_INC(out_invsatype);
7928 			error = EINVAL;
7929 			goto senderror;
7930 		}
7931 	}
7932 
7933 	/* check field of upper layer protocol and address family */
7934 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL
7935 	 && mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
7936 		struct sadb_address *src0, *dst0;
7937 		u_int plen;
7938 
7939 		src0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_SRC]);
7940 		dst0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_DST]);
7941 
7942 		/* check upper layer protocol */
7943 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
7944 			ipseclog((LOG_DEBUG, "%s: upper layer protocol "
7945 				"mismatched.\n", __func__));
7946 			PFKEYSTAT_INC(out_invaddr);
7947 			error = EINVAL;
7948 			goto senderror;
7949 		}
7950 
7951 		/* check family */
7952 		if (PFKEY_ADDR_SADDR(src0)->sa_family !=
7953 		    PFKEY_ADDR_SADDR(dst0)->sa_family) {
7954 			ipseclog((LOG_DEBUG, "%s: address family mismatched.\n",
7955 				__func__));
7956 			PFKEYSTAT_INC(out_invaddr);
7957 			error = EINVAL;
7958 			goto senderror;
7959 		}
7960 		if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7961 		    PFKEY_ADDR_SADDR(dst0)->sa_len) {
7962 			ipseclog((LOG_DEBUG, "%s: address struct size "
7963 				"mismatched.\n", __func__));
7964 			PFKEYSTAT_INC(out_invaddr);
7965 			error = EINVAL;
7966 			goto senderror;
7967 		}
7968 
7969 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7970 		case AF_INET:
7971 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7972 			    sizeof(struct sockaddr_in)) {
7973 				PFKEYSTAT_INC(out_invaddr);
7974 				error = EINVAL;
7975 				goto senderror;
7976 			}
7977 			break;
7978 		case AF_INET6:
7979 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7980 			    sizeof(struct sockaddr_in6)) {
7981 				PFKEYSTAT_INC(out_invaddr);
7982 				error = EINVAL;
7983 				goto senderror;
7984 			}
7985 			break;
7986 		default:
7987 			ipseclog((LOG_DEBUG, "%s: unsupported address family\n",
7988 				__func__));
7989 			PFKEYSTAT_INC(out_invaddr);
7990 			error = EAFNOSUPPORT;
7991 			goto senderror;
7992 		}
7993 
7994 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7995 		case AF_INET:
7996 			plen = sizeof(struct in_addr) << 3;
7997 			break;
7998 		case AF_INET6:
7999 			plen = sizeof(struct in6_addr) << 3;
8000 			break;
8001 		default:
8002 			plen = 0;	/*fool gcc*/
8003 			break;
8004 		}
8005 
8006 		/* check max prefix length */
8007 		if (src0->sadb_address_prefixlen > plen ||
8008 		    dst0->sadb_address_prefixlen > plen) {
8009 			ipseclog((LOG_DEBUG, "%s: illegal prefixlen.\n",
8010 				__func__));
8011 			PFKEYSTAT_INC(out_invaddr);
8012 			error = EINVAL;
8013 			goto senderror;
8014 		}
8015 
8016 		/*
8017 		 * prefixlen == 0 is valid because there can be a case when
8018 		 * all addresses are matched.
8019 		 */
8020 	}
8021 
8022 	if (msg->sadb_msg_type >= nitems(key_typesw) ||
8023 	    key_typesw[msg->sadb_msg_type] == NULL) {
8024 		PFKEYSTAT_INC(out_invmsgtype);
8025 		error = EINVAL;
8026 		goto senderror;
8027 	}
8028 
8029 	return (*key_typesw[msg->sadb_msg_type])(so, m, &mh);
8030 
8031 senderror:
8032 	msg->sadb_msg_errno = error;
8033 	return key_sendup_mbuf(so, m, target);
8034 }
8035 
8036 static int
8037 key_senderror(struct socket *so, struct mbuf *m, int code)
8038 {
8039 	struct sadb_msg *msg;
8040 
8041 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8042 		("mbuf too small, len %u", m->m_len));
8043 
8044 	msg = mtod(m, struct sadb_msg *);
8045 	msg->sadb_msg_errno = code;
8046 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
8047 }
8048 
8049 /*
8050  * set the pointer to each header into message buffer.
8051  * m will be freed on error.
8052  * XXX larger-than-MCLBYTES extension?
8053  */
8054 static int
8055 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
8056 {
8057 	struct mbuf *n;
8058 	struct sadb_ext *ext;
8059 	size_t off, end;
8060 	int extlen;
8061 	int toff;
8062 
8063 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
8064 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
8065 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8066 		("mbuf too small, len %u", m->m_len));
8067 
8068 	/* initialize */
8069 	bzero(mhp, sizeof(*mhp));
8070 
8071 	mhp->msg = mtod(m, struct sadb_msg *);
8072 	mhp->ext[0] = (struct sadb_ext *)mhp->msg;	/*XXX backward compat */
8073 
8074 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
8075 	extlen = end;	/*just in case extlen is not updated*/
8076 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
8077 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
8078 		if (!n) {
8079 			/* m is already freed */
8080 			return ENOBUFS;
8081 		}
8082 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8083 
8084 		/* set pointer */
8085 		switch (ext->sadb_ext_type) {
8086 		case SADB_EXT_SA:
8087 		case SADB_EXT_ADDRESS_SRC:
8088 		case SADB_EXT_ADDRESS_DST:
8089 		case SADB_EXT_ADDRESS_PROXY:
8090 		case SADB_EXT_LIFETIME_CURRENT:
8091 		case SADB_EXT_LIFETIME_HARD:
8092 		case SADB_EXT_LIFETIME_SOFT:
8093 		case SADB_EXT_KEY_AUTH:
8094 		case SADB_EXT_KEY_ENCRYPT:
8095 		case SADB_EXT_IDENTITY_SRC:
8096 		case SADB_EXT_IDENTITY_DST:
8097 		case SADB_EXT_SENSITIVITY:
8098 		case SADB_EXT_PROPOSAL:
8099 		case SADB_EXT_SUPPORTED_AUTH:
8100 		case SADB_EXT_SUPPORTED_ENCRYPT:
8101 		case SADB_EXT_SPIRANGE:
8102 		case SADB_X_EXT_POLICY:
8103 		case SADB_X_EXT_SA2:
8104 		case SADB_X_EXT_NAT_T_TYPE:
8105 		case SADB_X_EXT_NAT_T_SPORT:
8106 		case SADB_X_EXT_NAT_T_DPORT:
8107 		case SADB_X_EXT_NAT_T_OAI:
8108 		case SADB_X_EXT_NAT_T_OAR:
8109 		case SADB_X_EXT_NAT_T_FRAG:
8110 		case SADB_X_EXT_SA_REPLAY:
8111 		case SADB_X_EXT_NEW_ADDRESS_SRC:
8112 		case SADB_X_EXT_NEW_ADDRESS_DST:
8113 			/* duplicate check */
8114 			/*
8115 			 * XXX Are there duplication payloads of either
8116 			 * KEY_AUTH or KEY_ENCRYPT ?
8117 			 */
8118 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
8119 				ipseclog((LOG_DEBUG, "%s: duplicate ext_type "
8120 					"%u\n", __func__, ext->sadb_ext_type));
8121 				m_freem(m);
8122 				PFKEYSTAT_INC(out_dupext);
8123 				return EINVAL;
8124 			}
8125 			break;
8126 		default:
8127 			ipseclog((LOG_DEBUG, "%s: invalid ext_type %u\n",
8128 				__func__, ext->sadb_ext_type));
8129 			m_freem(m);
8130 			PFKEYSTAT_INC(out_invexttype);
8131 			return EINVAL;
8132 		}
8133 
8134 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
8135 
8136 		if (key_validate_ext(ext, extlen)) {
8137 			m_freem(m);
8138 			PFKEYSTAT_INC(out_invlen);
8139 			return EINVAL;
8140 		}
8141 
8142 		n = m_pulldown(m, off, extlen, &toff);
8143 		if (!n) {
8144 			/* m is already freed */
8145 			return ENOBUFS;
8146 		}
8147 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8148 
8149 		mhp->ext[ext->sadb_ext_type] = ext;
8150 		mhp->extoff[ext->sadb_ext_type] = off;
8151 		mhp->extlen[ext->sadb_ext_type] = extlen;
8152 	}
8153 
8154 	if (off != end) {
8155 		m_freem(m);
8156 		PFKEYSTAT_INC(out_invlen);
8157 		return EINVAL;
8158 	}
8159 
8160 	return 0;
8161 }
8162 
8163 static int
8164 key_validate_ext(const struct sadb_ext *ext, int len)
8165 {
8166 	const struct sockaddr *sa;
8167 	enum { NONE, ADDR } checktype = NONE;
8168 	int baselen = 0;
8169 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
8170 
8171 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
8172 		return EINVAL;
8173 
8174 	/* if it does not match minimum/maximum length, bail */
8175 	if (ext->sadb_ext_type >= nitems(minsize) ||
8176 	    ext->sadb_ext_type >= nitems(maxsize))
8177 		return EINVAL;
8178 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
8179 		return EINVAL;
8180 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
8181 		return EINVAL;
8182 
8183 	/* more checks based on sadb_ext_type XXX need more */
8184 	switch (ext->sadb_ext_type) {
8185 	case SADB_EXT_ADDRESS_SRC:
8186 	case SADB_EXT_ADDRESS_DST:
8187 	case SADB_EXT_ADDRESS_PROXY:
8188 	case SADB_X_EXT_NAT_T_OAI:
8189 	case SADB_X_EXT_NAT_T_OAR:
8190 	case SADB_X_EXT_NEW_ADDRESS_SRC:
8191 	case SADB_X_EXT_NEW_ADDRESS_DST:
8192 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
8193 		checktype = ADDR;
8194 		break;
8195 	case SADB_EXT_IDENTITY_SRC:
8196 	case SADB_EXT_IDENTITY_DST:
8197 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
8198 		    SADB_X_IDENTTYPE_ADDR) {
8199 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
8200 			checktype = ADDR;
8201 		} else
8202 			checktype = NONE;
8203 		break;
8204 	default:
8205 		checktype = NONE;
8206 		break;
8207 	}
8208 
8209 	switch (checktype) {
8210 	case NONE:
8211 		break;
8212 	case ADDR:
8213 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8214 		if (len < baselen + sal)
8215 			return EINVAL;
8216 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8217 			return EINVAL;
8218 		break;
8219 	}
8220 
8221 	return 0;
8222 }
8223 
8224 void
8225 spdcache_init(void)
8226 {
8227 	int i;
8228 
8229 	TUNABLE_INT_FETCH("net.key.spdcache.maxentries",
8230 	    &V_key_spdcache_maxentries);
8231 	TUNABLE_INT_FETCH("net.key.spdcache.threshold",
8232 	    &V_key_spdcache_threshold);
8233 
8234 	if (V_key_spdcache_maxentries) {
8235 		V_key_spdcache_maxentries = MAX(V_key_spdcache_maxentries,
8236 		    SPDCACHE_MAX_ENTRIES_PER_HASH);
8237 		V_spdcachehashtbl = hashinit(V_key_spdcache_maxentries /
8238 		    SPDCACHE_MAX_ENTRIES_PER_HASH,
8239 		    M_IPSEC_SPDCACHE, &V_spdcachehash_mask);
8240 		V_key_spdcache_maxentries = (V_spdcachehash_mask + 1)
8241 		    * SPDCACHE_MAX_ENTRIES_PER_HASH;
8242 
8243 		V_spdcache_lock = malloc(sizeof(struct mtx) *
8244 		    (V_spdcachehash_mask + 1),
8245 		    M_IPSEC_SPDCACHE, M_WAITOK | M_ZERO);
8246 
8247 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8248 			SPDCACHE_LOCK_INIT(i);
8249 	}
8250 }
8251 
8252 struct spdcache_entry *
8253 spdcache_entry_alloc(const struct secpolicyindex *spidx, struct secpolicy *sp)
8254 {
8255 	struct spdcache_entry *entry;
8256 
8257 	entry = malloc(sizeof(struct spdcache_entry), M_IPSEC_SPDCACHE,
8258 	    M_NOWAIT | M_ZERO);
8259 	if (entry == NULL)
8260 		return (NULL);
8261 
8262 	if (sp != NULL)
8263 		SP_ADDREF(sp);
8264 
8265 	entry->spidx = *spidx;
8266 	entry->sp = sp;
8267 
8268 	return (entry);
8269 }
8270 
8271 void
8272 spdcache_entry_free(struct spdcache_entry *entry)
8273 {
8274 
8275 	if (entry->sp != NULL)
8276 		key_freesp(&entry->sp);
8277 	free(entry, M_IPSEC_SPDCACHE);
8278 }
8279 
8280 void
8281 spdcache_clear(void)
8282 {
8283 	struct spdcache_entry *entry;
8284 	int i;
8285 
8286 	for (i = 0; i < V_spdcachehash_mask + 1; ++i) {
8287 		SPDCACHE_LOCK(i);
8288 		while (!LIST_EMPTY(&V_spdcachehashtbl[i])) {
8289 			entry = LIST_FIRST(&V_spdcachehashtbl[i]);
8290 			LIST_REMOVE(entry, chain);
8291 			spdcache_entry_free(entry);
8292 		}
8293 		SPDCACHE_UNLOCK(i);
8294 	}
8295 }
8296 
8297 #ifdef VIMAGE
8298 void
8299 spdcache_destroy(void)
8300 {
8301 	int i;
8302 
8303 	if (SPDCACHE_ENABLED()) {
8304 		spdcache_clear();
8305 		hashdestroy(V_spdcachehashtbl, M_IPSEC_SPDCACHE, V_spdcachehash_mask);
8306 
8307 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8308 			SPDCACHE_LOCK_DESTROY(i);
8309 
8310 		free(V_spdcache_lock, M_IPSEC_SPDCACHE);
8311 	}
8312 }
8313 #endif
8314 
8315 static void
8316 key_vnet_init(void *arg __unused)
8317 {
8318 	int i;
8319 
8320 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8321 		TAILQ_INIT(&V_sptree[i]);
8322 		TAILQ_INIT(&V_sptree_ifnet[i]);
8323 	}
8324 
8325 	TAILQ_INIT(&V_sahtree);
8326 	V_sphashtbl = hashinit(SPHASH_NHASH, M_IPSEC_SP, &V_sphash_mask);
8327 	V_savhashtbl = hashinit(SAVHASH_NHASH, M_IPSEC_SA, &V_savhash_mask);
8328 	V_sahaddrhashtbl = hashinit(SAHHASH_NHASH, M_IPSEC_SAH,
8329 	    &V_sahaddrhash_mask);
8330 	V_acqaddrhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8331 	    &V_acqaddrhash_mask);
8332 	V_acqseqhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8333 	    &V_acqseqhash_mask);
8334 
8335 	spdcache_init();
8336 
8337 	for (i = 0; i <= SADB_SATYPE_MAX; i++)
8338 		LIST_INIT(&V_regtree[i]);
8339 
8340 	LIST_INIT(&V_acqtree);
8341 	LIST_INIT(&V_spacqtree);
8342 }
8343 VNET_SYSINIT(key_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_SECOND,
8344     key_vnet_init, NULL);
8345 
8346 static void
8347 key_init(void *arg __unused)
8348 {
8349 
8350 	ipsec_key_lft_zone = uma_zcreate("IPsec SA lft_c",
8351 	    sizeof(uint64_t) * 2, NULL, NULL, NULL, NULL,
8352 	    UMA_ALIGN_PTR, UMA_ZONE_PCPU);
8353 
8354 	SPTREE_LOCK_INIT();
8355 	REGTREE_LOCK_INIT();
8356 	SAHTREE_LOCK_INIT();
8357 	ACQ_LOCK_INIT();
8358 	SPACQ_LOCK_INIT();
8359 	SPI_ALLOC_LOCK_INIT();
8360 
8361 #ifndef IPSEC_DEBUG2
8362 	callout_init(&key_timer, 1);
8363 	callout_reset(&key_timer, hz, key_timehandler, NULL);
8364 #endif /*IPSEC_DEBUG2*/
8365 
8366 	/* initialize key statistics */
8367 	keystat.getspi_count = 1;
8368 
8369 	if (bootverbose)
8370 		printf("IPsec: Initialized Security Association Processing.\n");
8371 }
8372 SYSINIT(key_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, key_init, NULL);
8373 
8374 #ifdef VIMAGE
8375 static void
8376 key_vnet_destroy(void *arg __unused)
8377 {
8378 	struct secashead_queue sahdrainq;
8379 	struct secpolicy_queue drainq;
8380 	struct secpolicy *sp, *nextsp;
8381 	struct secacq *acq, *nextacq;
8382 	struct secspacq *spacq, *nextspacq;
8383 	struct secashead *sah;
8384 	struct secasvar *sav;
8385 	struct secreg *reg;
8386 	int i;
8387 
8388 	/*
8389 	 * XXX: can we just call free() for each object without
8390 	 * walking through safe way with releasing references?
8391 	 */
8392 	TAILQ_INIT(&drainq);
8393 	SPTREE_WLOCK();
8394 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8395 		TAILQ_CONCAT(&drainq, &V_sptree[i], chain);
8396 		TAILQ_CONCAT(&drainq, &V_sptree_ifnet[i], chain);
8397 	}
8398 	for (i = 0; i < V_sphash_mask + 1; i++)
8399 		LIST_INIT(&V_sphashtbl[i]);
8400 	SPTREE_WUNLOCK();
8401 	spdcache_destroy();
8402 
8403 	sp = TAILQ_FIRST(&drainq);
8404 	while (sp != NULL) {
8405 		nextsp = TAILQ_NEXT(sp, chain);
8406 		key_freesp(&sp);
8407 		sp = nextsp;
8408 	}
8409 
8410 	TAILQ_INIT(&sahdrainq);
8411 	SAHTREE_WLOCK();
8412 	TAILQ_CONCAT(&sahdrainq, &V_sahtree, chain);
8413 	for (i = 0; i < V_savhash_mask + 1; i++)
8414 		LIST_INIT(&V_savhashtbl[i]);
8415 	for (i = 0; i < V_sahaddrhash_mask + 1; i++)
8416 		LIST_INIT(&V_sahaddrhashtbl[i]);
8417 	TAILQ_FOREACH(sah, &sahdrainq, chain) {
8418 		sah->state = SADB_SASTATE_DEAD;
8419 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
8420 			sav->state = SADB_SASTATE_DEAD;
8421 		}
8422 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
8423 			sav->state = SADB_SASTATE_DEAD;
8424 		}
8425 	}
8426 	SAHTREE_WUNLOCK();
8427 
8428 	key_freesah_flushed(&sahdrainq);
8429 	hashdestroy(V_sphashtbl, M_IPSEC_SP, V_sphash_mask);
8430 	hashdestroy(V_savhashtbl, M_IPSEC_SA, V_savhash_mask);
8431 	hashdestroy(V_sahaddrhashtbl, M_IPSEC_SAH, V_sahaddrhash_mask);
8432 
8433 	REGTREE_LOCK();
8434 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8435 		LIST_FOREACH(reg, &V_regtree[i], chain) {
8436 			if (__LIST_CHAINED(reg)) {
8437 				LIST_REMOVE(reg, chain);
8438 				free(reg, M_IPSEC_SAR);
8439 				break;
8440 			}
8441 		}
8442 	}
8443 	REGTREE_UNLOCK();
8444 
8445 	ACQ_LOCK();
8446 	acq = LIST_FIRST(&V_acqtree);
8447 	while (acq != NULL) {
8448 		nextacq = LIST_NEXT(acq, chain);
8449 		LIST_REMOVE(acq, chain);
8450 		free(acq, M_IPSEC_SAQ);
8451 		acq = nextacq;
8452 	}
8453 	for (i = 0; i < V_acqaddrhash_mask + 1; i++)
8454 		LIST_INIT(&V_acqaddrhashtbl[i]);
8455 	for (i = 0; i < V_acqseqhash_mask + 1; i++)
8456 		LIST_INIT(&V_acqseqhashtbl[i]);
8457 	ACQ_UNLOCK();
8458 
8459 	SPACQ_LOCK();
8460 	for (spacq = LIST_FIRST(&V_spacqtree); spacq != NULL;
8461 	    spacq = nextspacq) {
8462 		nextspacq = LIST_NEXT(spacq, chain);
8463 		if (__LIST_CHAINED(spacq)) {
8464 			LIST_REMOVE(spacq, chain);
8465 			free(spacq, M_IPSEC_SAQ);
8466 		}
8467 	}
8468 	SPACQ_UNLOCK();
8469 	hashdestroy(V_acqaddrhashtbl, M_IPSEC_SAQ, V_acqaddrhash_mask);
8470 	hashdestroy(V_acqseqhashtbl, M_IPSEC_SAQ, V_acqseqhash_mask);
8471 }
8472 VNET_SYSUNINIT(key_vnet_destroy, SI_SUB_PROTO_DOMAIN, SI_ORDER_SECOND,
8473     key_vnet_destroy, NULL);
8474 #endif
8475 
8476 /*
8477  * XXX: as long as domains are not unloadable, this function is never called,
8478  * provided for consistensy and future unload support.
8479  */
8480 static void
8481 key_destroy(void *arg __unused)
8482 {
8483 	uma_zdestroy(ipsec_key_lft_zone);
8484 
8485 #ifndef IPSEC_DEBUG2
8486 	callout_drain(&key_timer);
8487 #endif
8488 	SPTREE_LOCK_DESTROY();
8489 	REGTREE_LOCK_DESTROY();
8490 	SAHTREE_LOCK_DESTROY();
8491 	ACQ_LOCK_DESTROY();
8492 	SPACQ_LOCK_DESTROY();
8493 	SPI_ALLOC_LOCK_DESTROY();
8494 }
8495 SYSUNINIT(key_destroy, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, key_destroy, NULL);
8496 
8497 /* record data transfer on SA, and update timestamps */
8498 void
8499 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8500 {
8501 	IPSEC_ASSERT(sav != NULL, ("Null secasvar"));
8502 	IPSEC_ASSERT(m != NULL, ("Null mbuf"));
8503 
8504 	/*
8505 	 * XXX Currently, there is a difference of bytes size
8506 	 * between inbound and outbound processing.
8507 	 */
8508 	counter_u64_add(sav->lft_c_bytes, m->m_pkthdr.len);
8509 
8510 	/*
8511 	 * We use the number of packets as the unit of
8512 	 * allocations.  We increment the variable
8513 	 * whenever {esp,ah}_{in,out}put is called.
8514 	 */
8515 	counter_u64_add(sav->lft_c_allocations, 1);
8516 
8517 	/*
8518 	 * NOTE: We record CURRENT usetime by using wall clock,
8519 	 * in seconds.  HARD and SOFT lifetime are measured by the time
8520 	 * difference (again in seconds) from usetime.
8521 	 *
8522 	 *	usetime
8523 	 *	v     expire   expire
8524 	 * -----+-----+--------+---> t
8525 	 *	<--------------> HARD
8526 	 *	<-----> SOFT
8527 	 */
8528 	if (sav->firstused == 0)
8529 		sav->firstused = time_second;
8530 }
8531 
8532 /*
8533  * Take one of the kernel's security keys and convert it into a PF_KEY
8534  * structure within an mbuf, suitable for sending up to a waiting
8535  * application in user land.
8536  *
8537  * IN:
8538  *    src: A pointer to a kernel security key.
8539  *    exttype: Which type of key this is. Refer to the PF_KEY data structures.
8540  * OUT:
8541  *    a valid mbuf or NULL indicating an error
8542  *
8543  */
8544 
8545 static struct mbuf *
8546 key_setkey(struct seckey *src, uint16_t exttype)
8547 {
8548 	struct mbuf *m;
8549 	struct sadb_key *p;
8550 	int len;
8551 
8552 	if (src == NULL)
8553 		return NULL;
8554 
8555 	len = PFKEY_ALIGN8(sizeof(struct sadb_key) + _KEYLEN(src));
8556 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8557 	if (m == NULL)
8558 		return NULL;
8559 	m_align(m, len);
8560 	m->m_len = len;
8561 	p = mtod(m, struct sadb_key *);
8562 	bzero(p, len);
8563 	p->sadb_key_len = PFKEY_UNIT64(len);
8564 	p->sadb_key_exttype = exttype;
8565 	p->sadb_key_bits = src->bits;
8566 	bcopy(src->key_data, _KEYBUF(p), _KEYLEN(src));
8567 
8568 	return m;
8569 }
8570 
8571 /*
8572  * Take one of the kernel's lifetime data structures and convert it
8573  * into a PF_KEY structure within an mbuf, suitable for sending up to
8574  * a waiting application in user land.
8575  *
8576  * IN:
8577  *    src: A pointer to a kernel lifetime structure.
8578  *    exttype: Which type of lifetime this is. Refer to the PF_KEY
8579  *             data structures for more information.
8580  * OUT:
8581  *    a valid mbuf or NULL indicating an error
8582  *
8583  */
8584 
8585 static struct mbuf *
8586 key_setlifetime(struct seclifetime *src, uint16_t exttype)
8587 {
8588 	struct mbuf *m = NULL;
8589 	struct sadb_lifetime *p;
8590 	int len = PFKEY_ALIGN8(sizeof(struct sadb_lifetime));
8591 
8592 	if (src == NULL)
8593 		return NULL;
8594 
8595 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8596 	if (m == NULL)
8597 		return m;
8598 	m_align(m, len);
8599 	m->m_len = len;
8600 	p = mtod(m, struct sadb_lifetime *);
8601 
8602 	bzero(p, len);
8603 	p->sadb_lifetime_len = PFKEY_UNIT64(len);
8604 	p->sadb_lifetime_exttype = exttype;
8605 	p->sadb_lifetime_allocations = src->allocations;
8606 	p->sadb_lifetime_bytes = src->bytes;
8607 	p->sadb_lifetime_addtime = src->addtime;
8608 	p->sadb_lifetime_usetime = src->usetime;
8609 
8610 	return m;
8611 
8612 }
8613 
8614 const struct enc_xform *
8615 enc_algorithm_lookup(int alg)
8616 {
8617 	int i;
8618 
8619 	for (i = 0; i < nitems(supported_ealgs); i++)
8620 		if (alg == supported_ealgs[i].sadb_alg)
8621 			return (supported_ealgs[i].xform);
8622 	return (NULL);
8623 }
8624 
8625 const struct auth_hash *
8626 auth_algorithm_lookup(int alg)
8627 {
8628 	int i;
8629 
8630 	for (i = 0; i < nitems(supported_aalgs); i++)
8631 		if (alg == supported_aalgs[i].sadb_alg)
8632 			return (supported_aalgs[i].xform);
8633 	return (NULL);
8634 }
8635 
8636 const struct comp_algo *
8637 comp_algorithm_lookup(int alg)
8638 {
8639 	int i;
8640 
8641 	for (i = 0; i < nitems(supported_calgs); i++)
8642 		if (alg == supported_calgs[i].sadb_alg)
8643 			return (supported_calgs[i].xform);
8644 	return (NULL);
8645 }
8646