xref: /freebsd/sys/netipsec/key.c (revision 732a02b4e77866604a120a275c082bb6221bd2ff)
1 /*	$FreeBSD$	*/
2 /*	$KAME: key.c,v 1.191 2001/06/27 10:46:49 sakane Exp $	*/
3 
4 /*-
5  * SPDX-License-Identifier: BSD-3-Clause
6  *
7  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
8  * All rights reserved.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 3. Neither the name of the project nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 /*
36  * This code is referd to RFC 2367
37  */
38 
39 #include "opt_inet.h"
40 #include "opt_inet6.h"
41 #include "opt_ipsec.h"
42 
43 #include <sys/types.h>
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/kernel.h>
47 #include <sys/fnv_hash.h>
48 #include <sys/lock.h>
49 #include <sys/mutex.h>
50 #include <sys/mbuf.h>
51 #include <sys/domain.h>
52 #include <sys/protosw.h>
53 #include <sys/malloc.h>
54 #include <sys/rmlock.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sysctl.h>
58 #include <sys/errno.h>
59 #include <sys/proc.h>
60 #include <sys/queue.h>
61 #include <sys/refcount.h>
62 #include <sys/syslog.h>
63 
64 #include <vm/uma.h>
65 
66 #include <net/if.h>
67 #include <net/if_var.h>
68 #include <net/vnet.h>
69 #include <net/raw_cb.h>
70 
71 #include <netinet/in.h>
72 #include <netinet/in_systm.h>
73 #include <netinet/ip.h>
74 #include <netinet/in_var.h>
75 #include <netinet/udp.h>
76 
77 #ifdef INET6
78 #include <netinet/ip6.h>
79 #include <netinet6/in6_var.h>
80 #include <netinet6/ip6_var.h>
81 #endif /* INET6 */
82 
83 #include <net/pfkeyv2.h>
84 #include <netipsec/keydb.h>
85 #include <netipsec/key.h>
86 #include <netipsec/keysock.h>
87 #include <netipsec/key_debug.h>
88 
89 #include <netipsec/ipsec.h>
90 #ifdef INET6
91 #include <netipsec/ipsec6.h>
92 #endif
93 
94 #include <netipsec/xform.h>
95 #include <machine/in_cksum.h>
96 #include <machine/stdarg.h>
97 
98 /* randomness */
99 #include <sys/random.h>
100 
101 #define FULLMASK	0xff
102 #define	_BITS(bytes)	((bytes) << 3)
103 
104 /*
105  * Note on SA reference counting:
106  * - SAs that are not in DEAD state will have (total external reference + 1)
107  *   following value in reference count field.  they cannot be freed and are
108  *   referenced from SA header.
109  * - SAs that are in DEAD state will have (total external reference)
110  *   in reference count field.  they are ready to be freed.  reference from
111  *   SA header will be removed in key_delsav(), when the reference count
112  *   field hits 0 (= no external reference other than from SA header.
113  */
114 
115 VNET_DEFINE(u_int32_t, key_debug_level) = 0;
116 VNET_DEFINE_STATIC(u_int, key_spi_trycnt) = 1000;
117 VNET_DEFINE_STATIC(u_int32_t, key_spi_minval) = 0x100;
118 VNET_DEFINE_STATIC(u_int32_t, key_spi_maxval) = 0x0fffffff;	/* XXX */
119 VNET_DEFINE_STATIC(u_int32_t, policy_id) = 0;
120 /*interval to initialize randseed,1(m)*/
121 VNET_DEFINE_STATIC(u_int, key_int_random) = 60;
122 /* interval to expire acquiring, 30(s)*/
123 VNET_DEFINE_STATIC(u_int, key_larval_lifetime) = 30;
124 /* counter for blocking SADB_ACQUIRE.*/
125 VNET_DEFINE_STATIC(int, key_blockacq_count) = 10;
126 /* lifetime for blocking SADB_ACQUIRE.*/
127 VNET_DEFINE_STATIC(int, key_blockacq_lifetime) = 20;
128 /* preferred old sa rather than new sa.*/
129 VNET_DEFINE_STATIC(int, key_preferred_oldsa) = 1;
130 #define	V_key_spi_trycnt	VNET(key_spi_trycnt)
131 #define	V_key_spi_minval	VNET(key_spi_minval)
132 #define	V_key_spi_maxval	VNET(key_spi_maxval)
133 #define	V_policy_id		VNET(policy_id)
134 #define	V_key_int_random	VNET(key_int_random)
135 #define	V_key_larval_lifetime	VNET(key_larval_lifetime)
136 #define	V_key_blockacq_count	VNET(key_blockacq_count)
137 #define	V_key_blockacq_lifetime	VNET(key_blockacq_lifetime)
138 #define	V_key_preferred_oldsa	VNET(key_preferred_oldsa)
139 
140 VNET_DEFINE_STATIC(u_int32_t, acq_seq) = 0;
141 #define	V_acq_seq		VNET(acq_seq)
142 
143 VNET_DEFINE_STATIC(uint32_t, sp_genid) = 0;
144 #define	V_sp_genid		VNET(sp_genid)
145 
146 /* SPD */
147 TAILQ_HEAD(secpolicy_queue, secpolicy);
148 LIST_HEAD(secpolicy_list, secpolicy);
149 VNET_DEFINE_STATIC(struct secpolicy_queue, sptree[IPSEC_DIR_MAX]);
150 VNET_DEFINE_STATIC(struct secpolicy_queue, sptree_ifnet[IPSEC_DIR_MAX]);
151 static struct rmlock sptree_lock;
152 #define	V_sptree		VNET(sptree)
153 #define	V_sptree_ifnet		VNET(sptree_ifnet)
154 #define	SPTREE_LOCK_INIT()      rm_init(&sptree_lock, "sptree")
155 #define	SPTREE_LOCK_DESTROY()   rm_destroy(&sptree_lock)
156 #define	SPTREE_RLOCK_TRACKER    struct rm_priotracker sptree_tracker
157 #define	SPTREE_RLOCK()          rm_rlock(&sptree_lock, &sptree_tracker)
158 #define	SPTREE_RUNLOCK()        rm_runlock(&sptree_lock, &sptree_tracker)
159 #define	SPTREE_RLOCK_ASSERT()   rm_assert(&sptree_lock, RA_RLOCKED)
160 #define	SPTREE_WLOCK()          rm_wlock(&sptree_lock)
161 #define	SPTREE_WUNLOCK()        rm_wunlock(&sptree_lock)
162 #define	SPTREE_WLOCK_ASSERT()   rm_assert(&sptree_lock, RA_WLOCKED)
163 #define	SPTREE_UNLOCK_ASSERT()  rm_assert(&sptree_lock, RA_UNLOCKED)
164 
165 /* Hash table for lookup SP using unique id */
166 VNET_DEFINE_STATIC(struct secpolicy_list *, sphashtbl);
167 VNET_DEFINE_STATIC(u_long, sphash_mask);
168 #define	V_sphashtbl		VNET(sphashtbl)
169 #define	V_sphash_mask		VNET(sphash_mask)
170 
171 #define	SPHASH_NHASH_LOG2	7
172 #define	SPHASH_NHASH		(1 << SPHASH_NHASH_LOG2)
173 #define	SPHASH_HASHVAL(id)	(key_u32hash(id) & V_sphash_mask)
174 #define	SPHASH_HASH(id)		&V_sphashtbl[SPHASH_HASHVAL(id)]
175 
176 /* SPD cache */
177 struct spdcache_entry {
178    struct secpolicyindex spidx;	/* secpolicyindex */
179    struct secpolicy *sp;	/* cached policy to be used */
180 
181    LIST_ENTRY(spdcache_entry) chain;
182 };
183 LIST_HEAD(spdcache_entry_list, spdcache_entry);
184 
185 #define	SPDCACHE_MAX_ENTRIES_PER_HASH	8
186 
187 VNET_DEFINE_STATIC(u_int, key_spdcache_maxentries) = 0;
188 #define	V_key_spdcache_maxentries	VNET(key_spdcache_maxentries)
189 VNET_DEFINE_STATIC(u_int, key_spdcache_threshold) = 32;
190 #define	V_key_spdcache_threshold	VNET(key_spdcache_threshold)
191 VNET_DEFINE_STATIC(unsigned long, spd_size) = 0;
192 #define	V_spd_size		VNET(spd_size)
193 
194 #define SPDCACHE_ENABLED()	(V_key_spdcache_maxentries != 0)
195 #define SPDCACHE_ACTIVE() \
196 	(SPDCACHE_ENABLED() && V_spd_size >= V_key_spdcache_threshold)
197 
198 VNET_DEFINE_STATIC(struct spdcache_entry_list *, spdcachehashtbl);
199 VNET_DEFINE_STATIC(u_long, spdcachehash_mask);
200 #define	V_spdcachehashtbl	VNET(spdcachehashtbl)
201 #define	V_spdcachehash_mask	VNET(spdcachehash_mask)
202 
203 #define	SPDCACHE_HASHVAL(idx) \
204 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->ul_proto) &  \
205 	    V_spdcachehash_mask)
206 
207 /* Each cache line is protected by a mutex */
208 VNET_DEFINE_STATIC(struct mtx *, spdcache_lock);
209 #define	V_spdcache_lock		VNET(spdcache_lock)
210 
211 #define	SPDCACHE_LOCK_INIT(a) \
212 	mtx_init(&V_spdcache_lock[a], "spdcache", \
213 	    "fast ipsec SPD cache", MTX_DEF|MTX_DUPOK)
214 #define	SPDCACHE_LOCK_DESTROY(a)	mtx_destroy(&V_spdcache_lock[a])
215 #define	SPDCACHE_LOCK(a)		mtx_lock(&V_spdcache_lock[a]);
216 #define	SPDCACHE_UNLOCK(a)		mtx_unlock(&V_spdcache_lock[a]);
217 
218 /* SAD */
219 TAILQ_HEAD(secashead_queue, secashead);
220 LIST_HEAD(secashead_list, secashead);
221 VNET_DEFINE_STATIC(struct secashead_queue, sahtree);
222 static struct rmlock sahtree_lock;
223 #define	V_sahtree		VNET(sahtree)
224 #define	SAHTREE_LOCK_INIT()	rm_init(&sahtree_lock, "sahtree")
225 #define	SAHTREE_LOCK_DESTROY()	rm_destroy(&sahtree_lock)
226 #define	SAHTREE_RLOCK_TRACKER	struct rm_priotracker sahtree_tracker
227 #define	SAHTREE_RLOCK()		rm_rlock(&sahtree_lock, &sahtree_tracker)
228 #define	SAHTREE_RUNLOCK()	rm_runlock(&sahtree_lock, &sahtree_tracker)
229 #define	SAHTREE_RLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_RLOCKED)
230 #define	SAHTREE_WLOCK()		rm_wlock(&sahtree_lock)
231 #define	SAHTREE_WUNLOCK()	rm_wunlock(&sahtree_lock)
232 #define	SAHTREE_WLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_WLOCKED)
233 #define	SAHTREE_UNLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_UNLOCKED)
234 
235 /* Hash table for lookup in SAD using SA addresses */
236 VNET_DEFINE_STATIC(struct secashead_list *, sahaddrhashtbl);
237 VNET_DEFINE_STATIC(u_long, sahaddrhash_mask);
238 #define	V_sahaddrhashtbl	VNET(sahaddrhashtbl)
239 #define	V_sahaddrhash_mask	VNET(sahaddrhash_mask)
240 
241 #define	SAHHASH_NHASH_LOG2	7
242 #define	SAHHASH_NHASH		(1 << SAHHASH_NHASH_LOG2)
243 #define	SAHADDRHASH_HASHVAL(idx)	\
244 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->proto) & \
245 	    V_sahaddrhash_mask)
246 #define	SAHADDRHASH_HASH(saidx)		\
247     &V_sahaddrhashtbl[SAHADDRHASH_HASHVAL(saidx)]
248 
249 /* Hash table for lookup in SAD using SPI */
250 LIST_HEAD(secasvar_list, secasvar);
251 VNET_DEFINE_STATIC(struct secasvar_list *, savhashtbl);
252 VNET_DEFINE_STATIC(u_long, savhash_mask);
253 #define	V_savhashtbl		VNET(savhashtbl)
254 #define	V_savhash_mask		VNET(savhash_mask)
255 #define	SAVHASH_NHASH_LOG2	7
256 #define	SAVHASH_NHASH		(1 << SAVHASH_NHASH_LOG2)
257 #define	SAVHASH_HASHVAL(spi)	(key_u32hash(spi) & V_savhash_mask)
258 #define	SAVHASH_HASH(spi)	&V_savhashtbl[SAVHASH_HASHVAL(spi)]
259 
260 static uint32_t
261 key_addrprotohash(const union sockaddr_union *src,
262     const union sockaddr_union *dst, const uint8_t *proto)
263 {
264 	uint32_t hval;
265 
266 	hval = fnv_32_buf(proto, sizeof(*proto),
267 	    FNV1_32_INIT);
268 	switch (dst->sa.sa_family) {
269 #ifdef INET
270 	case AF_INET:
271 		hval = fnv_32_buf(&src->sin.sin_addr,
272 		    sizeof(in_addr_t), hval);
273 		hval = fnv_32_buf(&dst->sin.sin_addr,
274 		    sizeof(in_addr_t), hval);
275 		break;
276 #endif
277 #ifdef INET6
278 	case AF_INET6:
279 		hval = fnv_32_buf(&src->sin6.sin6_addr,
280 		    sizeof(struct in6_addr), hval);
281 		hval = fnv_32_buf(&dst->sin6.sin6_addr,
282 		    sizeof(struct in6_addr), hval);
283 		break;
284 #endif
285 	default:
286 		hval = 0;
287 		ipseclog((LOG_DEBUG, "%s: unknown address family %d\n",
288 		    __func__, dst->sa.sa_family));
289 	}
290 	return (hval);
291 }
292 
293 static uint32_t
294 key_u32hash(uint32_t val)
295 {
296 
297 	return (fnv_32_buf(&val, sizeof(val), FNV1_32_INIT));
298 }
299 
300 							/* registed list */
301 VNET_DEFINE_STATIC(LIST_HEAD(_regtree, secreg), regtree[SADB_SATYPE_MAX + 1]);
302 #define	V_regtree		VNET(regtree)
303 static struct mtx regtree_lock;
304 #define	REGTREE_LOCK_INIT() \
305 	mtx_init(&regtree_lock, "regtree", "fast ipsec regtree", MTX_DEF)
306 #define	REGTREE_LOCK_DESTROY()	mtx_destroy(&regtree_lock)
307 #define	REGTREE_LOCK()		mtx_lock(&regtree_lock)
308 #define	REGTREE_UNLOCK()	mtx_unlock(&regtree_lock)
309 #define	REGTREE_LOCK_ASSERT()	mtx_assert(&regtree_lock, MA_OWNED)
310 
311 /* Acquiring list */
312 LIST_HEAD(secacq_list, secacq);
313 VNET_DEFINE_STATIC(struct secacq_list, acqtree);
314 #define	V_acqtree		VNET(acqtree)
315 static struct mtx acq_lock;
316 #define	ACQ_LOCK_INIT() \
317     mtx_init(&acq_lock, "acqtree", "ipsec SA acquiring list", MTX_DEF)
318 #define	ACQ_LOCK_DESTROY()	mtx_destroy(&acq_lock)
319 #define	ACQ_LOCK()		mtx_lock(&acq_lock)
320 #define	ACQ_UNLOCK()		mtx_unlock(&acq_lock)
321 #define	ACQ_LOCK_ASSERT()	mtx_assert(&acq_lock, MA_OWNED)
322 
323 /* Hash table for lookup in ACQ list using SA addresses */
324 VNET_DEFINE_STATIC(struct secacq_list *, acqaddrhashtbl);
325 VNET_DEFINE_STATIC(u_long, acqaddrhash_mask);
326 #define	V_acqaddrhashtbl	VNET(acqaddrhashtbl)
327 #define	V_acqaddrhash_mask	VNET(acqaddrhash_mask)
328 
329 /* Hash table for lookup in ACQ list using SEQ number */
330 VNET_DEFINE_STATIC(struct secacq_list *, acqseqhashtbl);
331 VNET_DEFINE_STATIC(u_long, acqseqhash_mask);
332 #define	V_acqseqhashtbl		VNET(acqseqhashtbl)
333 #define	V_acqseqhash_mask	VNET(acqseqhash_mask)
334 
335 #define	ACQHASH_NHASH_LOG2	7
336 #define	ACQHASH_NHASH		(1 << ACQHASH_NHASH_LOG2)
337 #define	ACQADDRHASH_HASHVAL(idx)	\
338 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->proto) & \
339 	    V_acqaddrhash_mask)
340 #define	ACQSEQHASH_HASHVAL(seq)		\
341     (key_u32hash(seq) & V_acqseqhash_mask)
342 #define	ACQADDRHASH_HASH(saidx)	\
343     &V_acqaddrhashtbl[ACQADDRHASH_HASHVAL(saidx)]
344 #define	ACQSEQHASH_HASH(seq)	\
345     &V_acqseqhashtbl[ACQSEQHASH_HASHVAL(seq)]
346 							/* SP acquiring list */
347 VNET_DEFINE_STATIC(LIST_HEAD(_spacqtree, secspacq), spacqtree);
348 #define	V_spacqtree		VNET(spacqtree)
349 static struct mtx spacq_lock;
350 #define	SPACQ_LOCK_INIT() \
351 	mtx_init(&spacq_lock, "spacqtree", \
352 		"fast ipsec security policy acquire list", MTX_DEF)
353 #define	SPACQ_LOCK_DESTROY()	mtx_destroy(&spacq_lock)
354 #define	SPACQ_LOCK()		mtx_lock(&spacq_lock)
355 #define	SPACQ_UNLOCK()		mtx_unlock(&spacq_lock)
356 #define	SPACQ_LOCK_ASSERT()	mtx_assert(&spacq_lock, MA_OWNED)
357 
358 static const int minsize[] = {
359 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
360 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
361 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
362 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
363 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
364 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_SRC */
365 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_DST */
366 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_PROXY */
367 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_AUTH */
368 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_ENCRYPT */
369 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_SRC */
370 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_DST */
371 	sizeof(struct sadb_sens),	/* SADB_EXT_SENSITIVITY */
372 	sizeof(struct sadb_prop),	/* SADB_EXT_PROPOSAL */
373 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_AUTH */
374 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_ENCRYPT */
375 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
376 	0,				/* SADB_X_EXT_KMPRIVATE */
377 	sizeof(struct sadb_x_policy),	/* SADB_X_EXT_POLICY */
378 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
379 	sizeof(struct sadb_x_nat_t_type),/* SADB_X_EXT_NAT_T_TYPE */
380 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_SPORT */
381 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_DPORT */
382 	sizeof(struct sadb_address),	/* SADB_X_EXT_NAT_T_OAI */
383 	sizeof(struct sadb_address),	/* SADB_X_EXT_NAT_T_OAR */
384 	sizeof(struct sadb_x_nat_t_frag),/* SADB_X_EXT_NAT_T_FRAG */
385 	sizeof(struct sadb_x_sa_replay), /* SADB_X_EXT_SA_REPLAY */
386 	sizeof(struct sadb_address),	/* SADB_X_EXT_NEW_ADDRESS_SRC */
387 	sizeof(struct sadb_address),	/* SADB_X_EXT_NEW_ADDRESS_DST */
388 };
389 _Static_assert(sizeof(minsize)/sizeof(int) == SADB_EXT_MAX + 1, "minsize size mismatch");
390 
391 static const int maxsize[] = {
392 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
393 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
394 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
395 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
396 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
397 	0,				/* SADB_EXT_ADDRESS_SRC */
398 	0,				/* SADB_EXT_ADDRESS_DST */
399 	0,				/* SADB_EXT_ADDRESS_PROXY */
400 	0,				/* SADB_EXT_KEY_AUTH */
401 	0,				/* SADB_EXT_KEY_ENCRYPT */
402 	0,				/* SADB_EXT_IDENTITY_SRC */
403 	0,				/* SADB_EXT_IDENTITY_DST */
404 	0,				/* SADB_EXT_SENSITIVITY */
405 	0,				/* SADB_EXT_PROPOSAL */
406 	0,				/* SADB_EXT_SUPPORTED_AUTH */
407 	0,				/* SADB_EXT_SUPPORTED_ENCRYPT */
408 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
409 	0,				/* SADB_X_EXT_KMPRIVATE */
410 	0,				/* SADB_X_EXT_POLICY */
411 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
412 	sizeof(struct sadb_x_nat_t_type),/* SADB_X_EXT_NAT_T_TYPE */
413 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_SPORT */
414 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_DPORT */
415 	0,				/* SADB_X_EXT_NAT_T_OAI */
416 	0,				/* SADB_X_EXT_NAT_T_OAR */
417 	sizeof(struct sadb_x_nat_t_frag),/* SADB_X_EXT_NAT_T_FRAG */
418 	sizeof(struct sadb_x_sa_replay), /* SADB_X_EXT_SA_REPLAY */
419 	0,				/* SADB_X_EXT_NEW_ADDRESS_SRC */
420 	0,				/* SADB_X_EXT_NEW_ADDRESS_DST */
421 };
422 _Static_assert(sizeof(maxsize)/sizeof(int) == SADB_EXT_MAX + 1, "minsize size mismatch");
423 
424 /*
425  * Internal values for SA flags:
426  * SADB_X_EXT_F_CLONED means that SA was cloned by key_updateaddresses,
427  *	thus we will not free the most of SA content in key_delsav().
428  */
429 #define	SADB_X_EXT_F_CLONED	0x80000000
430 
431 #define	SADB_CHECKLEN(_mhp, _ext)			\
432     ((_mhp)->extlen[(_ext)] < minsize[(_ext)] || (maxsize[(_ext)] != 0 && \
433 	((_mhp)->extlen[(_ext)] > maxsize[(_ext)])))
434 #define	SADB_CHECKHDR(_mhp, _ext)	((_mhp)->ext[(_ext)] == NULL)
435 
436 VNET_DEFINE_STATIC(int, ipsec_esp_keymin) = 256;
437 VNET_DEFINE_STATIC(int, ipsec_esp_auth) = 0;
438 VNET_DEFINE_STATIC(int, ipsec_ah_keymin) = 128;
439 
440 #define	V_ipsec_esp_keymin	VNET(ipsec_esp_keymin)
441 #define	V_ipsec_esp_auth	VNET(ipsec_esp_auth)
442 #define	V_ipsec_ah_keymin	VNET(ipsec_ah_keymin)
443 
444 #ifdef IPSEC_DEBUG
445 VNET_DEFINE(int, ipsec_debug) = 1;
446 #else
447 VNET_DEFINE(int, ipsec_debug) = 0;
448 #endif
449 
450 #ifdef INET
451 SYSCTL_DECL(_net_inet_ipsec);
452 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
453     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
454     "Enable IPsec debugging output when set.");
455 #endif
456 #ifdef INET6
457 SYSCTL_DECL(_net_inet6_ipsec6);
458 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug,
459     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
460     "Enable IPsec debugging output when set.");
461 #endif
462 
463 SYSCTL_DECL(_net_key);
464 SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,
465 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_debug_level), 0, "");
466 
467 /* max count of trial for the decision of spi value */
468 SYSCTL_INT(_net_key, KEYCTL_SPI_TRY, spi_trycnt,
469 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_trycnt), 0, "");
470 
471 /* minimum spi value to allocate automatically. */
472 SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE, spi_minval,
473 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_minval), 0, "");
474 
475 /* maximun spi value to allocate automatically. */
476 SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE, spi_maxval,
477 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_maxval), 0, "");
478 
479 /* interval to initialize randseed */
480 SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT, int_random,
481 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_int_random), 0, "");
482 
483 /* lifetime for larval SA */
484 SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME, larval_lifetime,
485 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_larval_lifetime), 0, "");
486 
487 /* counter for blocking to send SADB_ACQUIRE to IKEd */
488 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT, blockacq_count,
489 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_count), 0, "");
490 
491 /* lifetime for blocking to send SADB_ACQUIRE to IKEd */
492 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME, blockacq_lifetime,
493 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_lifetime), 0, "");
494 
495 /* ESP auth */
496 SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH, esp_auth,
497 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_auth), 0, "");
498 
499 /* minimum ESP key length */
500 SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN, esp_keymin,
501 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_keymin), 0, "");
502 
503 /* minimum AH key length */
504 SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN, ah_keymin,
505 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_ah_keymin), 0, "");
506 
507 /* perfered old SA rather than new SA */
508 SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA, preferred_oldsa,
509 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_preferred_oldsa), 0, "");
510 
511 static SYSCTL_NODE(_net_key, OID_AUTO, spdcache,
512     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
513     "SPD cache");
514 
515 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, maxentries,
516 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_maxentries), 0,
517 	"Maximum number of entries in the SPD cache"
518 	" (power of 2, 0 to disable)");
519 
520 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, threshold,
521 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_threshold), 0,
522 	"Number of SPs that make the SPD cache active");
523 
524 #define __LIST_CHAINED(elm) \
525 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
526 
527 MALLOC_DEFINE(M_IPSEC_SA, "secasvar", "ipsec security association");
528 MALLOC_DEFINE(M_IPSEC_SAH, "sahead", "ipsec sa head");
529 MALLOC_DEFINE(M_IPSEC_SP, "ipsecpolicy", "ipsec security policy");
530 MALLOC_DEFINE(M_IPSEC_SR, "ipsecrequest", "ipsec security request");
531 MALLOC_DEFINE(M_IPSEC_MISC, "ipsec-misc", "ipsec miscellaneous");
532 MALLOC_DEFINE(M_IPSEC_SAQ, "ipsec-saq", "ipsec sa acquire");
533 MALLOC_DEFINE(M_IPSEC_SAR, "ipsec-reg", "ipsec sa acquire");
534 MALLOC_DEFINE(M_IPSEC_SPDCACHE, "ipsec-spdcache", "ipsec SPD cache");
535 
536 VNET_DEFINE_STATIC(uma_zone_t, key_lft_zone);
537 #define	V_key_lft_zone		VNET(key_lft_zone)
538 
539 /*
540  * set parameters into secpolicyindex buffer.
541  * Must allocate secpolicyindex buffer passed to this function.
542  */
543 #define KEY_SETSECSPIDX(_dir, s, d, ps, pd, ulp, idx) \
544 do { \
545 	bzero((idx), sizeof(struct secpolicyindex));                         \
546 	(idx)->dir = (_dir);                                                 \
547 	(idx)->prefs = (ps);                                                 \
548 	(idx)->prefd = (pd);                                                 \
549 	(idx)->ul_proto = (ulp);                                             \
550 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
551 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
552 } while (0)
553 
554 /*
555  * set parameters into secasindex buffer.
556  * Must allocate secasindex buffer before calling this function.
557  */
558 #define KEY_SETSECASIDX(p, m, r, s, d, idx) \
559 do { \
560 	bzero((idx), sizeof(struct secasindex));                             \
561 	(idx)->proto = (p);                                                  \
562 	(idx)->mode = (m);                                                   \
563 	(idx)->reqid = (r);                                                  \
564 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
565 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
566 	key_porttosaddr(&(idx)->src.sa, 0);				     \
567 	key_porttosaddr(&(idx)->dst.sa, 0);				     \
568 } while (0)
569 
570 /* key statistics */
571 struct _keystat {
572 	u_long getspi_count; /* the avarage of count to try to get new SPI */
573 } keystat;
574 
575 struct sadb_msghdr {
576 	struct sadb_msg *msg;
577 	struct sadb_ext *ext[SADB_EXT_MAX + 1];
578 	int extoff[SADB_EXT_MAX + 1];
579 	int extlen[SADB_EXT_MAX + 1];
580 };
581 
582 static struct supported_ealgs {
583 	int sadb_alg;
584 	const struct enc_xform *xform;
585 } supported_ealgs[] = {
586 	{ SADB_EALG_DESCBC,		&enc_xform_des },
587 	{ SADB_EALG_3DESCBC,		&enc_xform_3des },
588 	{ SADB_X_EALG_AES,		&enc_xform_rijndael128 },
589 	{ SADB_X_EALG_BLOWFISHCBC,	&enc_xform_blf },
590 	{ SADB_X_EALG_CAST128CBC,	&enc_xform_cast5 },
591 	{ SADB_EALG_NULL,		&enc_xform_null },
592 	{ SADB_X_EALG_CAMELLIACBC,	&enc_xform_camellia },
593 	{ SADB_X_EALG_AESCTR,		&enc_xform_aes_icm },
594 	{ SADB_X_EALG_AESGCM16,		&enc_xform_aes_nist_gcm },
595 	{ SADB_X_EALG_AESGMAC,		&enc_xform_aes_nist_gmac },
596 };
597 
598 static struct supported_aalgs {
599 	int sadb_alg;
600 	const struct auth_hash *xform;
601 } supported_aalgs[] = {
602 	{ SADB_X_AALG_NULL,		&auth_hash_null },
603 	{ SADB_AALG_MD5HMAC,		&auth_hash_hmac_md5 },
604 	{ SADB_AALG_SHA1HMAC,		&auth_hash_hmac_sha1 },
605 	{ SADB_X_AALG_RIPEMD160HMAC,	&auth_hash_hmac_ripemd_160 },
606 	{ SADB_X_AALG_MD5,		&auth_hash_key_md5 },
607 	{ SADB_X_AALG_SHA,		&auth_hash_key_sha1 },
608 	{ SADB_X_AALG_SHA2_256,		&auth_hash_hmac_sha2_256 },
609 	{ SADB_X_AALG_SHA2_384,		&auth_hash_hmac_sha2_384 },
610 	{ SADB_X_AALG_SHA2_512,		&auth_hash_hmac_sha2_512 },
611 	{ SADB_X_AALG_AES128GMAC,	&auth_hash_nist_gmac_aes_128 },
612 	{ SADB_X_AALG_AES192GMAC,	&auth_hash_nist_gmac_aes_192 },
613 	{ SADB_X_AALG_AES256GMAC,	&auth_hash_nist_gmac_aes_256 },
614 };
615 
616 static struct supported_calgs {
617 	int sadb_alg;
618 	const struct comp_algo *xform;
619 } supported_calgs[] = {
620 	{ SADB_X_CALG_DEFLATE,		&comp_algo_deflate },
621 };
622 
623 #ifndef IPSEC_DEBUG2
624 static struct callout key_timer;
625 #endif
626 
627 static void key_unlink(struct secpolicy *);
628 static struct secpolicy *key_do_allocsp(struct secpolicyindex *spidx, u_int dir);
629 static struct secpolicy *key_getsp(struct secpolicyindex *);
630 static struct secpolicy *key_getspbyid(u_int32_t);
631 static struct mbuf *key_gather_mbuf(struct mbuf *,
632 	const struct sadb_msghdr *, int, int, ...);
633 static int key_spdadd(struct socket *, struct mbuf *,
634 	const struct sadb_msghdr *);
635 static uint32_t key_getnewspid(void);
636 static int key_spddelete(struct socket *, struct mbuf *,
637 	const struct sadb_msghdr *);
638 static int key_spddelete2(struct socket *, struct mbuf *,
639 	const struct sadb_msghdr *);
640 static int key_spdget(struct socket *, struct mbuf *,
641 	const struct sadb_msghdr *);
642 static int key_spdflush(struct socket *, struct mbuf *,
643 	const struct sadb_msghdr *);
644 static int key_spddump(struct socket *, struct mbuf *,
645 	const struct sadb_msghdr *);
646 static struct mbuf *key_setdumpsp(struct secpolicy *,
647 	u_int8_t, u_int32_t, u_int32_t);
648 static struct mbuf *key_sp2mbuf(struct secpolicy *);
649 static size_t key_getspreqmsglen(struct secpolicy *);
650 static int key_spdexpire(struct secpolicy *);
651 static struct secashead *key_newsah(struct secasindex *);
652 static void key_freesah(struct secashead **);
653 static void key_delsah(struct secashead *);
654 static struct secasvar *key_newsav(const struct sadb_msghdr *,
655     struct secasindex *, uint32_t, int *);
656 static void key_delsav(struct secasvar *);
657 static void key_unlinksav(struct secasvar *);
658 static struct secashead *key_getsah(struct secasindex *);
659 static int key_checkspidup(uint32_t);
660 static struct secasvar *key_getsavbyspi(uint32_t);
661 static int key_setnatt(struct secasvar *, const struct sadb_msghdr *);
662 static int key_setsaval(struct secasvar *, const struct sadb_msghdr *);
663 static int key_updatelifetimes(struct secasvar *, const struct sadb_msghdr *);
664 static int key_updateaddresses(struct socket *, struct mbuf *,
665     const struct sadb_msghdr *, struct secasvar *, struct secasindex *);
666 
667 static struct mbuf *key_setdumpsa(struct secasvar *, u_int8_t,
668 	u_int8_t, u_int32_t, u_int32_t);
669 static struct mbuf *key_setsadbmsg(u_int8_t, u_int16_t, u_int8_t,
670 	u_int32_t, pid_t, u_int16_t);
671 static struct mbuf *key_setsadbsa(struct secasvar *);
672 static struct mbuf *key_setsadbaddr(u_int16_t,
673 	const struct sockaddr *, u_int8_t, u_int16_t);
674 static struct mbuf *key_setsadbxport(u_int16_t, u_int16_t);
675 static struct mbuf *key_setsadbxtype(u_int16_t);
676 static struct mbuf *key_setsadbxsa2(u_int8_t, u_int32_t, u_int32_t);
677 static struct mbuf *key_setsadbxsareplay(u_int32_t);
678 static struct mbuf *key_setsadbxpolicy(u_int16_t, u_int8_t,
679 	u_int32_t, u_int32_t);
680 static struct seckey *key_dup_keymsg(const struct sadb_key *, size_t,
681     struct malloc_type *);
682 static struct seclifetime *key_dup_lifemsg(const struct sadb_lifetime *src,
683     struct malloc_type *);
684 
685 /* flags for key_cmpsaidx() */
686 #define CMP_HEAD	1	/* protocol, addresses. */
687 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
688 #define CMP_REQID	3	/* additionally HEAD, reaid. */
689 #define CMP_EXACTLY	4	/* all elements. */
690 static int key_cmpsaidx(const struct secasindex *,
691     const struct secasindex *, int);
692 static int key_cmpspidx_exactly(struct secpolicyindex *,
693     struct secpolicyindex *);
694 static int key_cmpspidx_withmask(struct secpolicyindex *,
695     struct secpolicyindex *);
696 static int key_bbcmp(const void *, const void *, u_int);
697 static uint8_t key_satype2proto(uint8_t);
698 static uint8_t key_proto2satype(uint8_t);
699 
700 static int key_getspi(struct socket *, struct mbuf *,
701 	const struct sadb_msghdr *);
702 static uint32_t key_do_getnewspi(struct sadb_spirange *, struct secasindex *);
703 static int key_update(struct socket *, struct mbuf *,
704 	const struct sadb_msghdr *);
705 static int key_add(struct socket *, struct mbuf *,
706 	const struct sadb_msghdr *);
707 static int key_setident(struct secashead *, const struct sadb_msghdr *);
708 static struct mbuf *key_getmsgbuf_x1(struct mbuf *,
709 	const struct sadb_msghdr *);
710 static int key_delete(struct socket *, struct mbuf *,
711 	const struct sadb_msghdr *);
712 static int key_delete_all(struct socket *, struct mbuf *,
713 	const struct sadb_msghdr *, struct secasindex *);
714 static int key_get(struct socket *, struct mbuf *,
715 	const struct sadb_msghdr *);
716 
717 static void key_getcomb_setlifetime(struct sadb_comb *);
718 static struct mbuf *key_getcomb_ealg(void);
719 static struct mbuf *key_getcomb_ah(void);
720 static struct mbuf *key_getcomb_ipcomp(void);
721 static struct mbuf *key_getprop(const struct secasindex *);
722 
723 static int key_acquire(const struct secasindex *, struct secpolicy *);
724 static uint32_t key_newacq(const struct secasindex *, int *);
725 static uint32_t key_getacq(const struct secasindex *, int *);
726 static int key_acqdone(const struct secasindex *, uint32_t);
727 static int key_acqreset(uint32_t);
728 static struct secspacq *key_newspacq(struct secpolicyindex *);
729 static struct secspacq *key_getspacq(struct secpolicyindex *);
730 static int key_acquire2(struct socket *, struct mbuf *,
731 	const struct sadb_msghdr *);
732 static int key_register(struct socket *, struct mbuf *,
733 	const struct sadb_msghdr *);
734 static int key_expire(struct secasvar *, int);
735 static int key_flush(struct socket *, struct mbuf *,
736 	const struct sadb_msghdr *);
737 static int key_dump(struct socket *, struct mbuf *,
738 	const struct sadb_msghdr *);
739 static int key_promisc(struct socket *, struct mbuf *,
740 	const struct sadb_msghdr *);
741 static int key_senderror(struct socket *, struct mbuf *, int);
742 static int key_validate_ext(const struct sadb_ext *, int);
743 static int key_align(struct mbuf *, struct sadb_msghdr *);
744 static struct mbuf *key_setlifetime(struct seclifetime *, uint16_t);
745 static struct mbuf *key_setkey(struct seckey *, uint16_t);
746 
747 static void spdcache_init(void);
748 static void spdcache_clear(void);
749 static struct spdcache_entry *spdcache_entry_alloc(
750 	const struct secpolicyindex *spidx,
751 	struct secpolicy *policy);
752 static void spdcache_entry_free(struct spdcache_entry *entry);
753 #ifdef VIMAGE
754 static void spdcache_destroy(void);
755 #endif
756 
757 #define	DBG_IPSEC_INITREF(t, p)	do {				\
758 	refcount_init(&(p)->refcnt, 1);				\
759 	KEYDBG(KEY_STAMP,					\
760 	    printf("%s: Initialize refcnt %s(%p) = %u\n",	\
761 	    __func__, #t, (p), (p)->refcnt));			\
762 } while (0)
763 #define	DBG_IPSEC_ADDREF(t, p)	do {				\
764 	refcount_acquire(&(p)->refcnt);				\
765 	KEYDBG(KEY_STAMP,					\
766 	    printf("%s: Acquire refcnt %s(%p) -> %u\n",		\
767 	    __func__, #t, (p), (p)->refcnt));			\
768 } while (0)
769 #define	DBG_IPSEC_DELREF(t, p)	do {				\
770 	KEYDBG(KEY_STAMP,					\
771 	    printf("%s: Release refcnt %s(%p) -> %u\n",		\
772 	    __func__, #t, (p), (p)->refcnt - 1));		\
773 	refcount_release(&(p)->refcnt);				\
774 } while (0)
775 
776 #define	IPSEC_INITREF(t, p)	refcount_init(&(p)->refcnt, 1)
777 #define	IPSEC_ADDREF(t, p)	refcount_acquire(&(p)->refcnt)
778 #define	IPSEC_DELREF(t, p)	refcount_release(&(p)->refcnt)
779 
780 #define	SP_INITREF(p)	IPSEC_INITREF(SP, p)
781 #define	SP_ADDREF(p)	IPSEC_ADDREF(SP, p)
782 #define	SP_DELREF(p)	IPSEC_DELREF(SP, p)
783 
784 #define	SAH_INITREF(p)	IPSEC_INITREF(SAH, p)
785 #define	SAH_ADDREF(p)	IPSEC_ADDREF(SAH, p)
786 #define	SAH_DELREF(p)	IPSEC_DELREF(SAH, p)
787 
788 #define	SAV_INITREF(p)	IPSEC_INITREF(SAV, p)
789 #define	SAV_ADDREF(p)	IPSEC_ADDREF(SAV, p)
790 #define	SAV_DELREF(p)	IPSEC_DELREF(SAV, p)
791 
792 /*
793  * Update the refcnt while holding the SPTREE lock.
794  */
795 void
796 key_addref(struct secpolicy *sp)
797 {
798 
799 	SP_ADDREF(sp);
800 }
801 
802 /*
803  * Return 0 when there are known to be no SP's for the specified
804  * direction.  Otherwise return 1.  This is used by IPsec code
805  * to optimize performance.
806  */
807 int
808 key_havesp(u_int dir)
809 {
810 
811 	return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ?
812 		TAILQ_FIRST(&V_sptree[dir]) != NULL : 1);
813 }
814 
815 /* %%% IPsec policy management */
816 /*
817  * Return current SPDB generation.
818  */
819 uint32_t
820 key_getspgen(void)
821 {
822 
823 	return (V_sp_genid);
824 }
825 
826 void
827 key_bumpspgen(void)
828 {
829 
830 	V_sp_genid++;
831 }
832 
833 static int
834 key_checksockaddrs(struct sockaddr *src, struct sockaddr *dst)
835 {
836 
837 	/* family match */
838 	if (src->sa_family != dst->sa_family)
839 		return (EINVAL);
840 	/* sa_len match */
841 	if (src->sa_len != dst->sa_len)
842 		return (EINVAL);
843 	switch (src->sa_family) {
844 #ifdef INET
845 	case AF_INET:
846 		if (src->sa_len != sizeof(struct sockaddr_in))
847 			return (EINVAL);
848 		break;
849 #endif
850 #ifdef INET6
851 	case AF_INET6:
852 		if (src->sa_len != sizeof(struct sockaddr_in6))
853 			return (EINVAL);
854 		break;
855 #endif
856 	default:
857 		return (EAFNOSUPPORT);
858 	}
859 	return (0);
860 }
861 
862 struct secpolicy *
863 key_do_allocsp(struct secpolicyindex *spidx, u_int dir)
864 {
865 	SPTREE_RLOCK_TRACKER;
866 	struct secpolicy *sp;
867 
868 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
869 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
870 		("invalid direction %u", dir));
871 
872 	SPTREE_RLOCK();
873 	TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
874 		if (key_cmpspidx_withmask(&sp->spidx, spidx)) {
875 			SP_ADDREF(sp);
876 			break;
877 		}
878 	}
879 	SPTREE_RUNLOCK();
880 	return (sp);
881 }
882 
883 
884 /*
885  * allocating a SP for OUTBOUND or INBOUND packet.
886  * Must call key_freesp() later.
887  * OUT:	NULL:	not found
888  *	others:	found and return the pointer.
889  */
890 struct secpolicy *
891 key_allocsp(struct secpolicyindex *spidx, u_int dir)
892 {
893 	struct spdcache_entry *entry, *lastentry, *tmpentry;
894 	struct secpolicy *sp;
895 	uint32_t hashv;
896 	int nb_entries;
897 
898 	if (!SPDCACHE_ACTIVE()) {
899 		sp = key_do_allocsp(spidx, dir);
900 		goto out;
901 	}
902 
903 	hashv = SPDCACHE_HASHVAL(spidx);
904 	SPDCACHE_LOCK(hashv);
905 	nb_entries = 0;
906 	LIST_FOREACH_SAFE(entry, &V_spdcachehashtbl[hashv], chain, tmpentry) {
907 		/* Removed outdated entries */
908 		if (entry->sp != NULL &&
909 		    entry->sp->state == IPSEC_SPSTATE_DEAD) {
910 			LIST_REMOVE(entry, chain);
911 			spdcache_entry_free(entry);
912 			continue;
913 		}
914 
915 		nb_entries++;
916 		if (!key_cmpspidx_exactly(&entry->spidx, spidx)) {
917 			lastentry = entry;
918 			continue;
919 		}
920 
921 		sp = entry->sp;
922 		if (entry->sp != NULL)
923 			SP_ADDREF(sp);
924 
925 		/* IPSECSTAT_INC(ips_spdcache_hits); */
926 
927 		SPDCACHE_UNLOCK(hashv);
928 		goto out;
929 	}
930 
931 	/* IPSECSTAT_INC(ips_spdcache_misses); */
932 
933 	sp = key_do_allocsp(spidx, dir);
934 	entry = spdcache_entry_alloc(spidx, sp);
935 	if (entry != NULL) {
936 		if (nb_entries >= SPDCACHE_MAX_ENTRIES_PER_HASH) {
937 			LIST_REMOVE(lastentry, chain);
938 			spdcache_entry_free(lastentry);
939 		}
940 
941 		LIST_INSERT_HEAD(&V_spdcachehashtbl[hashv], entry, chain);
942 	}
943 
944 	SPDCACHE_UNLOCK(hashv);
945 
946 out:
947 	if (sp != NULL) {	/* found a SPD entry */
948 		sp->lastused = time_second;
949 		KEYDBG(IPSEC_STAMP,
950 		    printf("%s: return SP(%p)\n", __func__, sp));
951 		KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
952 	} else {
953 		KEYDBG(IPSEC_DATA,
954 		    printf("%s: lookup failed for ", __func__);
955 		    kdebug_secpolicyindex(spidx, NULL));
956 	}
957 	return (sp);
958 }
959 
960 /*
961  * Allocating an SA entry for an *INBOUND* or *OUTBOUND* TCP packet, signed
962  * or should be signed by MD5 signature.
963  * We don't use key_allocsa() for such lookups, because we don't know SPI.
964  * Unlike ESP and AH protocols, SPI isn't transmitted in the TCP header with
965  * signed packet. We use SADB only as storage for password.
966  * OUT:	positive:	corresponding SA for given saidx found.
967  *	NULL:		SA not found
968  */
969 struct secasvar *
970 key_allocsa_tcpmd5(struct secasindex *saidx)
971 {
972 	SAHTREE_RLOCK_TRACKER;
973 	struct secashead *sah;
974 	struct secasvar *sav;
975 
976 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP,
977 	    ("unexpected security protocol %u", saidx->proto));
978 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TCPMD5,
979 	    ("unexpected mode %u", saidx->mode));
980 
981 	SAHTREE_RLOCK();
982 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
983 		KEYDBG(IPSEC_DUMP,
984 		    printf("%s: checking SAH\n", __func__);
985 		    kdebug_secash(sah, "  "));
986 		if (sah->saidx.proto != IPPROTO_TCP)
987 			continue;
988 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
989 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
990 			break;
991 	}
992 	if (sah != NULL) {
993 		if (V_key_preferred_oldsa)
994 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
995 		else
996 			sav = TAILQ_FIRST(&sah->savtree_alive);
997 		if (sav != NULL)
998 			SAV_ADDREF(sav);
999 	} else
1000 		sav = NULL;
1001 	SAHTREE_RUNLOCK();
1002 
1003 	if (sav != NULL) {
1004 		KEYDBG(IPSEC_STAMP,
1005 		    printf("%s: return SA(%p)\n", __func__, sav));
1006 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1007 	} else {
1008 		KEYDBG(IPSEC_STAMP,
1009 		    printf("%s: SA not found\n", __func__));
1010 		KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1011 	}
1012 	return (sav);
1013 }
1014 
1015 /*
1016  * Allocating an SA entry for an *OUTBOUND* packet.
1017  * OUT:	positive:	corresponding SA for given saidx found.
1018  *	NULL:		SA not found, but will be acquired, check *error
1019  *			for acquiring status.
1020  */
1021 struct secasvar *
1022 key_allocsa_policy(struct secpolicy *sp, const struct secasindex *saidx,
1023     int *error)
1024 {
1025 	SAHTREE_RLOCK_TRACKER;
1026 	struct secashead *sah;
1027 	struct secasvar *sav;
1028 
1029 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
1030 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TRANSPORT ||
1031 		saidx->mode == IPSEC_MODE_TUNNEL,
1032 		("unexpected policy %u", saidx->mode));
1033 
1034 	/*
1035 	 * We check new SA in the IPsec request because a different
1036 	 * SA may be involved each time this request is checked, either
1037 	 * because new SAs are being configured, or this request is
1038 	 * associated with an unconnected datagram socket, or this request
1039 	 * is associated with a system default policy.
1040 	 */
1041 	SAHTREE_RLOCK();
1042 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
1043 		KEYDBG(IPSEC_DUMP,
1044 		    printf("%s: checking SAH\n", __func__);
1045 		    kdebug_secash(sah, "  "));
1046 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID))
1047 			break;
1048 
1049 	}
1050 	if (sah != NULL) {
1051 		/*
1052 		 * Allocate the oldest SA available according to
1053 		 * draft-jenkins-ipsec-rekeying-03.
1054 		 */
1055 		if (V_key_preferred_oldsa)
1056 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1057 		else
1058 			sav = TAILQ_FIRST(&sah->savtree_alive);
1059 		if (sav != NULL)
1060 			SAV_ADDREF(sav);
1061 	} else
1062 		sav = NULL;
1063 	SAHTREE_RUNLOCK();
1064 
1065 	if (sav != NULL) {
1066 		*error = 0;
1067 		KEYDBG(IPSEC_STAMP,
1068 		    printf("%s: chosen SA(%p) for SP(%p)\n", __func__,
1069 			sav, sp));
1070 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1071 		return (sav); /* return referenced SA */
1072 	}
1073 
1074 	/* there is no SA */
1075 	*error = key_acquire(saidx, sp);
1076 	if ((*error) != 0)
1077 		ipseclog((LOG_DEBUG,
1078 		    "%s: error %d returned from key_acquire()\n",
1079 			__func__, *error));
1080 	KEYDBG(IPSEC_STAMP,
1081 	    printf("%s: acquire SA for SP(%p), error %d\n",
1082 		__func__, sp, *error));
1083 	KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1084 	return (NULL);
1085 }
1086 
1087 /*
1088  * allocating a usable SA entry for a *INBOUND* packet.
1089  * Must call key_freesav() later.
1090  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
1091  *	NULL:		not found, or error occurred.
1092  *
1093  * According to RFC 2401 SA is uniquely identified by a triple SPI,
1094  * destination address, and security protocol. But according to RFC 4301,
1095  * SPI by itself suffices to specify an SA.
1096  *
1097  * Note that, however, we do need to keep source address in IPsec SA.
1098  * IKE specification and PF_KEY specification do assume that we
1099  * keep source address in IPsec SA.  We see a tricky situation here.
1100  */
1101 struct secasvar *
1102 key_allocsa(union sockaddr_union *dst, uint8_t proto, uint32_t spi)
1103 {
1104 	SAHTREE_RLOCK_TRACKER;
1105 	struct secasvar *sav;
1106 
1107 	IPSEC_ASSERT(proto == IPPROTO_ESP || proto == IPPROTO_AH ||
1108 	    proto == IPPROTO_IPCOMP, ("unexpected security protocol %u",
1109 	    proto));
1110 
1111 	SAHTREE_RLOCK();
1112 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
1113 		if (sav->spi == spi)
1114 			break;
1115 	}
1116 	/*
1117 	 * We use single SPI namespace for all protocols, so it is
1118 	 * impossible to have SPI duplicates in the SAVHASH.
1119 	 */
1120 	if (sav != NULL) {
1121 		if (sav->state != SADB_SASTATE_LARVAL &&
1122 		    sav->sah->saidx.proto == proto &&
1123 		    key_sockaddrcmp(&dst->sa,
1124 			&sav->sah->saidx.dst.sa, 0) == 0)
1125 			SAV_ADDREF(sav);
1126 		else
1127 			sav = NULL;
1128 	}
1129 	SAHTREE_RUNLOCK();
1130 
1131 	if (sav == NULL) {
1132 		KEYDBG(IPSEC_STAMP,
1133 		    char buf[IPSEC_ADDRSTRLEN];
1134 		    printf("%s: SA not found for spi %u proto %u dst %s\n",
1135 			__func__, ntohl(spi), proto, ipsec_address(dst, buf,
1136 			sizeof(buf))));
1137 	} else {
1138 		KEYDBG(IPSEC_STAMP,
1139 		    printf("%s: return SA(%p)\n", __func__, sav));
1140 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1141 	}
1142 	return (sav);
1143 }
1144 
1145 struct secasvar *
1146 key_allocsa_tunnel(union sockaddr_union *src, union sockaddr_union *dst,
1147     uint8_t proto)
1148 {
1149 	SAHTREE_RLOCK_TRACKER;
1150 	struct secasindex saidx;
1151 	struct secashead *sah;
1152 	struct secasvar *sav;
1153 
1154 	IPSEC_ASSERT(src != NULL, ("null src address"));
1155 	IPSEC_ASSERT(dst != NULL, ("null dst address"));
1156 
1157 	KEY_SETSECASIDX(proto, IPSEC_MODE_TUNNEL, 0, &src->sa,
1158 	    &dst->sa, &saidx);
1159 
1160 	sav = NULL;
1161 	SAHTREE_RLOCK();
1162 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
1163 		if (IPSEC_MODE_TUNNEL != sah->saidx.mode)
1164 			continue;
1165 		if (proto != sah->saidx.proto)
1166 			continue;
1167 		if (key_sockaddrcmp(&src->sa, &sah->saidx.src.sa, 0) != 0)
1168 			continue;
1169 		if (key_sockaddrcmp(&dst->sa, &sah->saidx.dst.sa, 0) != 0)
1170 			continue;
1171 		/* XXXAE: is key_preferred_oldsa reasonably?*/
1172 		if (V_key_preferred_oldsa)
1173 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1174 		else
1175 			sav = TAILQ_FIRST(&sah->savtree_alive);
1176 		if (sav != NULL) {
1177 			SAV_ADDREF(sav);
1178 			break;
1179 		}
1180 	}
1181 	SAHTREE_RUNLOCK();
1182 	KEYDBG(IPSEC_STAMP,
1183 	    printf("%s: return SA(%p)\n", __func__, sav));
1184 	if (sav != NULL)
1185 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1186 	return (sav);
1187 }
1188 
1189 /*
1190  * Must be called after calling key_allocsp().
1191  */
1192 void
1193 key_freesp(struct secpolicy **spp)
1194 {
1195 	struct secpolicy *sp = *spp;
1196 
1197 	IPSEC_ASSERT(sp != NULL, ("null sp"));
1198 	if (SP_DELREF(sp) == 0)
1199 		return;
1200 
1201 	KEYDBG(IPSEC_STAMP,
1202 	    printf("%s: last reference to SP(%p)\n", __func__, sp));
1203 	KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1204 
1205 	*spp = NULL;
1206 	while (sp->tcount > 0)
1207 		ipsec_delisr(sp->req[--sp->tcount]);
1208 	free(sp, M_IPSEC_SP);
1209 }
1210 
1211 static void
1212 key_unlink(struct secpolicy *sp)
1213 {
1214 
1215 	IPSEC_ASSERT(sp->spidx.dir == IPSEC_DIR_INBOUND ||
1216 	    sp->spidx.dir == IPSEC_DIR_OUTBOUND,
1217 	    ("invalid direction %u", sp->spidx.dir));
1218 	SPTREE_UNLOCK_ASSERT();
1219 
1220 	KEYDBG(KEY_STAMP,
1221 	    printf("%s: SP(%p)\n", __func__, sp));
1222 	SPTREE_WLOCK();
1223 	if (sp->state != IPSEC_SPSTATE_ALIVE) {
1224 		/* SP is already unlinked */
1225 		SPTREE_WUNLOCK();
1226 		return;
1227 	}
1228 	sp->state = IPSEC_SPSTATE_DEAD;
1229 	TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
1230 	V_spd_size--;
1231 	LIST_REMOVE(sp, idhash);
1232 	V_sp_genid++;
1233 	SPTREE_WUNLOCK();
1234 	if (SPDCACHE_ENABLED())
1235 		spdcache_clear();
1236 	key_freesp(&sp);
1237 }
1238 
1239 /*
1240  * insert a secpolicy into the SP database. Lower priorities first
1241  */
1242 static void
1243 key_insertsp(struct secpolicy *newsp)
1244 {
1245 	struct secpolicy *sp;
1246 
1247 	SPTREE_WLOCK_ASSERT();
1248 	TAILQ_FOREACH(sp, &V_sptree[newsp->spidx.dir], chain) {
1249 		if (newsp->priority < sp->priority) {
1250 			TAILQ_INSERT_BEFORE(sp, newsp, chain);
1251 			goto done;
1252 		}
1253 	}
1254 	TAILQ_INSERT_TAIL(&V_sptree[newsp->spidx.dir], newsp, chain);
1255 done:
1256 	LIST_INSERT_HEAD(SPHASH_HASH(newsp->id), newsp, idhash);
1257 	newsp->state = IPSEC_SPSTATE_ALIVE;
1258 	V_spd_size++;
1259 	V_sp_genid++;
1260 }
1261 
1262 /*
1263  * Insert a bunch of VTI secpolicies into the SPDB.
1264  * We keep VTI policies in the separate list due to following reasons:
1265  * 1) they should be immutable to user's or some deamon's attempts to
1266  *    delete. The only way delete such policies - destroy or unconfigure
1267  *    corresponding virtual inteface.
1268  * 2) such policies have traffic selector that matches all traffic per
1269  *    address family.
1270  * Since all VTI policies have the same priority, we don't care about
1271  * policies order.
1272  */
1273 int
1274 key_register_ifnet(struct secpolicy **spp, u_int count)
1275 {
1276 	struct mbuf *m;
1277 	u_int i;
1278 
1279 	SPTREE_WLOCK();
1280 	/*
1281 	 * First of try to acquire id for each SP.
1282 	 */
1283 	for (i = 0; i < count; i++) {
1284 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1285 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1286 		    ("invalid direction %u", spp[i]->spidx.dir));
1287 
1288 		if ((spp[i]->id = key_getnewspid()) == 0) {
1289 			SPTREE_WUNLOCK();
1290 			return (EAGAIN);
1291 		}
1292 	}
1293 	for (i = 0; i < count; i++) {
1294 		TAILQ_INSERT_TAIL(&V_sptree_ifnet[spp[i]->spidx.dir],
1295 		    spp[i], chain);
1296 		/*
1297 		 * NOTE: despite the fact that we keep VTI SP in the
1298 		 * separate list, SPHASH contains policies from both
1299 		 * sources. Thus SADB_X_SPDGET will correctly return
1300 		 * SP by id, because it uses SPHASH for lookups.
1301 		 */
1302 		LIST_INSERT_HEAD(SPHASH_HASH(spp[i]->id), spp[i], idhash);
1303 		spp[i]->state = IPSEC_SPSTATE_IFNET;
1304 	}
1305 	SPTREE_WUNLOCK();
1306 	/*
1307 	 * Notify user processes about new SP.
1308 	 */
1309 	for (i = 0; i < count; i++) {
1310 		m = key_setdumpsp(spp[i], SADB_X_SPDADD, 0, 0);
1311 		if (m != NULL)
1312 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1313 	}
1314 	return (0);
1315 }
1316 
1317 void
1318 key_unregister_ifnet(struct secpolicy **spp, u_int count)
1319 {
1320 	struct mbuf *m;
1321 	u_int i;
1322 
1323 	SPTREE_WLOCK();
1324 	for (i = 0; i < count; i++) {
1325 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1326 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1327 		    ("invalid direction %u", spp[i]->spidx.dir));
1328 
1329 		if (spp[i]->state != IPSEC_SPSTATE_IFNET)
1330 			continue;
1331 		spp[i]->state = IPSEC_SPSTATE_DEAD;
1332 		TAILQ_REMOVE(&V_sptree_ifnet[spp[i]->spidx.dir],
1333 		    spp[i], chain);
1334 		V_spd_size--;
1335 		LIST_REMOVE(spp[i], idhash);
1336 	}
1337 	SPTREE_WUNLOCK();
1338 	if (SPDCACHE_ENABLED())
1339 		spdcache_clear();
1340 
1341 	for (i = 0; i < count; i++) {
1342 		m = key_setdumpsp(spp[i], SADB_X_SPDDELETE, 0, 0);
1343 		if (m != NULL)
1344 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1345 	}
1346 }
1347 
1348 /*
1349  * Must be called after calling key_allocsa().
1350  * This function is called by key_freesp() to free some SA allocated
1351  * for a policy.
1352  */
1353 void
1354 key_freesav(struct secasvar **psav)
1355 {
1356 	struct secasvar *sav = *psav;
1357 
1358 	IPSEC_ASSERT(sav != NULL, ("null sav"));
1359 	if (SAV_DELREF(sav) == 0)
1360 		return;
1361 
1362 	KEYDBG(IPSEC_STAMP,
1363 	    printf("%s: last reference to SA(%p)\n", __func__, sav));
1364 
1365 	*psav = NULL;
1366 	key_delsav(sav);
1367 }
1368 
1369 /*
1370  * Unlink SA from SAH and SPI hash under SAHTREE_WLOCK.
1371  * Expect that SA has extra reference due to lookup.
1372  * Release this references, also release SAH reference after unlink.
1373  */
1374 static void
1375 key_unlinksav(struct secasvar *sav)
1376 {
1377 	struct secashead *sah;
1378 
1379 	KEYDBG(KEY_STAMP,
1380 	    printf("%s: SA(%p)\n", __func__, sav));
1381 
1382 	SAHTREE_UNLOCK_ASSERT();
1383 	SAHTREE_WLOCK();
1384 	if (sav->state == SADB_SASTATE_DEAD) {
1385 		/* SA is already unlinked */
1386 		SAHTREE_WUNLOCK();
1387 		return;
1388 	}
1389 	/* Unlink from SAH */
1390 	if (sav->state == SADB_SASTATE_LARVAL)
1391 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
1392 	else
1393 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
1394 	/* Unlink from SPI hash */
1395 	LIST_REMOVE(sav, spihash);
1396 	sav->state = SADB_SASTATE_DEAD;
1397 	sah = sav->sah;
1398 	SAHTREE_WUNLOCK();
1399 	key_freesav(&sav);
1400 	/* Since we are unlinked, release reference to SAH */
1401 	key_freesah(&sah);
1402 }
1403 
1404 /* %%% SPD management */
1405 /*
1406  * search SPD
1407  * OUT:	NULL	: not found
1408  *	others	: found, pointer to a SP.
1409  */
1410 static struct secpolicy *
1411 key_getsp(struct secpolicyindex *spidx)
1412 {
1413 	SPTREE_RLOCK_TRACKER;
1414 	struct secpolicy *sp;
1415 
1416 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
1417 
1418 	SPTREE_RLOCK();
1419 	TAILQ_FOREACH(sp, &V_sptree[spidx->dir], chain) {
1420 		if (key_cmpspidx_exactly(spidx, &sp->spidx)) {
1421 			SP_ADDREF(sp);
1422 			break;
1423 		}
1424 	}
1425 	SPTREE_RUNLOCK();
1426 
1427 	return sp;
1428 }
1429 
1430 /*
1431  * get SP by index.
1432  * OUT:	NULL	: not found
1433  *	others	: found, pointer to referenced SP.
1434  */
1435 static struct secpolicy *
1436 key_getspbyid(uint32_t id)
1437 {
1438 	SPTREE_RLOCK_TRACKER;
1439 	struct secpolicy *sp;
1440 
1441 	SPTREE_RLOCK();
1442 	LIST_FOREACH(sp, SPHASH_HASH(id), idhash) {
1443 		if (sp->id == id) {
1444 			SP_ADDREF(sp);
1445 			break;
1446 		}
1447 	}
1448 	SPTREE_RUNLOCK();
1449 	return (sp);
1450 }
1451 
1452 struct secpolicy *
1453 key_newsp(void)
1454 {
1455 	struct secpolicy *sp;
1456 
1457 	sp = malloc(sizeof(*sp), M_IPSEC_SP, M_NOWAIT | M_ZERO);
1458 	if (sp != NULL)
1459 		SP_INITREF(sp);
1460 	return (sp);
1461 }
1462 
1463 struct ipsecrequest *
1464 ipsec_newisr(void)
1465 {
1466 
1467 	return (malloc(sizeof(struct ipsecrequest), M_IPSEC_SR,
1468 	    M_NOWAIT | M_ZERO));
1469 }
1470 
1471 void
1472 ipsec_delisr(struct ipsecrequest *p)
1473 {
1474 
1475 	free(p, M_IPSEC_SR);
1476 }
1477 
1478 /*
1479  * create secpolicy structure from sadb_x_policy structure.
1480  * NOTE: `state', `secpolicyindex' and 'id' in secpolicy structure
1481  * are not set, so must be set properly later.
1482  */
1483 struct secpolicy *
1484 key_msg2sp(struct sadb_x_policy *xpl0, size_t len, int *error)
1485 {
1486 	struct secpolicy *newsp;
1487 
1488 	IPSEC_ASSERT(xpl0 != NULL, ("null xpl0"));
1489 	IPSEC_ASSERT(len >= sizeof(*xpl0), ("policy too short: %zu", len));
1490 
1491 	if (len != PFKEY_EXTLEN(xpl0)) {
1492 		ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n", __func__));
1493 		*error = EINVAL;
1494 		return NULL;
1495 	}
1496 
1497 	if ((newsp = key_newsp()) == NULL) {
1498 		*error = ENOBUFS;
1499 		return NULL;
1500 	}
1501 
1502 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1503 	newsp->policy = xpl0->sadb_x_policy_type;
1504 	newsp->priority = xpl0->sadb_x_policy_priority;
1505 	newsp->tcount = 0;
1506 
1507 	/* check policy */
1508 	switch (xpl0->sadb_x_policy_type) {
1509 	case IPSEC_POLICY_DISCARD:
1510 	case IPSEC_POLICY_NONE:
1511 	case IPSEC_POLICY_ENTRUST:
1512 	case IPSEC_POLICY_BYPASS:
1513 		break;
1514 
1515 	case IPSEC_POLICY_IPSEC:
1516 	    {
1517 		struct sadb_x_ipsecrequest *xisr;
1518 		struct ipsecrequest *isr;
1519 		int tlen;
1520 
1521 		/* validity check */
1522 		if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1523 			ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n",
1524 				__func__));
1525 			key_freesp(&newsp);
1526 			*error = EINVAL;
1527 			return NULL;
1528 		}
1529 
1530 		tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1531 		xisr = (struct sadb_x_ipsecrequest *)(xpl0 + 1);
1532 
1533 		while (tlen > 0) {
1534 			/* length check */
1535 			if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr) ||
1536 			    xisr->sadb_x_ipsecrequest_len > tlen) {
1537 				ipseclog((LOG_DEBUG, "%s: invalid ipsecrequest "
1538 					"length.\n", __func__));
1539 				key_freesp(&newsp);
1540 				*error = EINVAL;
1541 				return NULL;
1542 			}
1543 
1544 			if (newsp->tcount >= IPSEC_MAXREQ) {
1545 				ipseclog((LOG_DEBUG,
1546 				    "%s: too many ipsecrequests.\n",
1547 				    __func__));
1548 				key_freesp(&newsp);
1549 				*error = EINVAL;
1550 				return (NULL);
1551 			}
1552 
1553 			/* allocate request buffer */
1554 			/* NB: data structure is zero'd */
1555 			isr = ipsec_newisr();
1556 			if (isr == NULL) {
1557 				ipseclog((LOG_DEBUG,
1558 				    "%s: No more memory.\n", __func__));
1559 				key_freesp(&newsp);
1560 				*error = ENOBUFS;
1561 				return NULL;
1562 			}
1563 
1564 			newsp->req[newsp->tcount++] = isr;
1565 
1566 			/* set values */
1567 			switch (xisr->sadb_x_ipsecrequest_proto) {
1568 			case IPPROTO_ESP:
1569 			case IPPROTO_AH:
1570 			case IPPROTO_IPCOMP:
1571 				break;
1572 			default:
1573 				ipseclog((LOG_DEBUG,
1574 				    "%s: invalid proto type=%u\n", __func__,
1575 				    xisr->sadb_x_ipsecrequest_proto));
1576 				key_freesp(&newsp);
1577 				*error = EPROTONOSUPPORT;
1578 				return NULL;
1579 			}
1580 			isr->saidx.proto =
1581 			    (uint8_t)xisr->sadb_x_ipsecrequest_proto;
1582 
1583 			switch (xisr->sadb_x_ipsecrequest_mode) {
1584 			case IPSEC_MODE_TRANSPORT:
1585 			case IPSEC_MODE_TUNNEL:
1586 				break;
1587 			case IPSEC_MODE_ANY:
1588 			default:
1589 				ipseclog((LOG_DEBUG,
1590 				    "%s: invalid mode=%u\n", __func__,
1591 				    xisr->sadb_x_ipsecrequest_mode));
1592 				key_freesp(&newsp);
1593 				*error = EINVAL;
1594 				return NULL;
1595 			}
1596 			isr->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1597 
1598 			switch (xisr->sadb_x_ipsecrequest_level) {
1599 			case IPSEC_LEVEL_DEFAULT:
1600 			case IPSEC_LEVEL_USE:
1601 			case IPSEC_LEVEL_REQUIRE:
1602 				break;
1603 			case IPSEC_LEVEL_UNIQUE:
1604 				/* validity check */
1605 				/*
1606 				 * If range violation of reqid, kernel will
1607 				 * update it, don't refuse it.
1608 				 */
1609 				if (xisr->sadb_x_ipsecrequest_reqid
1610 						> IPSEC_MANUAL_REQID_MAX) {
1611 					ipseclog((LOG_DEBUG,
1612 					    "%s: reqid=%d range "
1613 					    "violation, updated by kernel.\n",
1614 					    __func__,
1615 					    xisr->sadb_x_ipsecrequest_reqid));
1616 					xisr->sadb_x_ipsecrequest_reqid = 0;
1617 				}
1618 
1619 				/* allocate new reqid id if reqid is zero. */
1620 				if (xisr->sadb_x_ipsecrequest_reqid == 0) {
1621 					u_int32_t reqid;
1622 					if ((reqid = key_newreqid()) == 0) {
1623 						key_freesp(&newsp);
1624 						*error = ENOBUFS;
1625 						return NULL;
1626 					}
1627 					isr->saidx.reqid = reqid;
1628 					xisr->sadb_x_ipsecrequest_reqid = reqid;
1629 				} else {
1630 				/* set it for manual keying. */
1631 					isr->saidx.reqid =
1632 					    xisr->sadb_x_ipsecrequest_reqid;
1633 				}
1634 				break;
1635 
1636 			default:
1637 				ipseclog((LOG_DEBUG, "%s: invalid level=%u\n",
1638 					__func__,
1639 					xisr->sadb_x_ipsecrequest_level));
1640 				key_freesp(&newsp);
1641 				*error = EINVAL;
1642 				return NULL;
1643 			}
1644 			isr->level = xisr->sadb_x_ipsecrequest_level;
1645 
1646 			/* set IP addresses if there */
1647 			if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1648 				struct sockaddr *paddr;
1649 
1650 				len = tlen - sizeof(*xisr);
1651 				paddr = (struct sockaddr *)(xisr + 1);
1652 				/* validity check */
1653 				if (len < sizeof(struct sockaddr) ||
1654 				    len < 2 * paddr->sa_len ||
1655 				    paddr->sa_len > sizeof(isr->saidx.src)) {
1656 					ipseclog((LOG_DEBUG, "%s: invalid "
1657 						"request address length.\n",
1658 						__func__));
1659 					key_freesp(&newsp);
1660 					*error = EINVAL;
1661 					return NULL;
1662 				}
1663 				/*
1664 				 * Request length should be enough to keep
1665 				 * source and destination addresses.
1666 				 */
1667 				if (xisr->sadb_x_ipsecrequest_len <
1668 				    sizeof(*xisr) + 2 * paddr->sa_len) {
1669 					ipseclog((LOG_DEBUG, "%s: invalid "
1670 					    "ipsecrequest length.\n",
1671 					    __func__));
1672 					key_freesp(&newsp);
1673 					*error = EINVAL;
1674 					return (NULL);
1675 				}
1676 				bcopy(paddr, &isr->saidx.src, paddr->sa_len);
1677 				paddr = (struct sockaddr *)((caddr_t)paddr +
1678 				    paddr->sa_len);
1679 
1680 				/* validity check */
1681 				if (paddr->sa_len !=
1682 				    isr->saidx.src.sa.sa_len) {
1683 					ipseclog((LOG_DEBUG, "%s: invalid "
1684 						"request address length.\n",
1685 						__func__));
1686 					key_freesp(&newsp);
1687 					*error = EINVAL;
1688 					return NULL;
1689 				}
1690 				/* AF family should match */
1691 				if (paddr->sa_family !=
1692 				    isr->saidx.src.sa.sa_family) {
1693 					ipseclog((LOG_DEBUG, "%s: address "
1694 					    "family doesn't match.\n",
1695 						__func__));
1696 					key_freesp(&newsp);
1697 					*error = EINVAL;
1698 					return (NULL);
1699 				}
1700 				bcopy(paddr, &isr->saidx.dst, paddr->sa_len);
1701 			} else {
1702 				/*
1703 				 * Addresses for TUNNEL mode requests are
1704 				 * mandatory.
1705 				 */
1706 				if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1707 					ipseclog((LOG_DEBUG, "%s: missing "
1708 					    "request addresses.\n", __func__));
1709 					key_freesp(&newsp);
1710 					*error = EINVAL;
1711 					return (NULL);
1712 				}
1713 			}
1714 			tlen -= xisr->sadb_x_ipsecrequest_len;
1715 
1716 			/* validity check */
1717 			if (tlen < 0) {
1718 				ipseclog((LOG_DEBUG, "%s: becoming tlen < 0.\n",
1719 					__func__));
1720 				key_freesp(&newsp);
1721 				*error = EINVAL;
1722 				return NULL;
1723 			}
1724 
1725 			xisr = (struct sadb_x_ipsecrequest *)((caddr_t)xisr
1726 			                 + xisr->sadb_x_ipsecrequest_len);
1727 		}
1728 		/* XXXAE: LARVAL SP */
1729 		if (newsp->tcount < 1) {
1730 			ipseclog((LOG_DEBUG, "%s: valid IPSEC transforms "
1731 			    "not found.\n", __func__));
1732 			key_freesp(&newsp);
1733 			*error = EINVAL;
1734 			return (NULL);
1735 		}
1736 	    }
1737 		break;
1738 	default:
1739 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
1740 		key_freesp(&newsp);
1741 		*error = EINVAL;
1742 		return NULL;
1743 	}
1744 
1745 	*error = 0;
1746 	return (newsp);
1747 }
1748 
1749 uint32_t
1750 key_newreqid(void)
1751 {
1752 	static uint32_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1753 
1754 	if (auto_reqid == ~0)
1755 		auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1756 	else
1757 		auto_reqid++;
1758 
1759 	/* XXX should be unique check */
1760 	return (auto_reqid);
1761 }
1762 
1763 /*
1764  * copy secpolicy struct to sadb_x_policy structure indicated.
1765  */
1766 static struct mbuf *
1767 key_sp2mbuf(struct secpolicy *sp)
1768 {
1769 	struct mbuf *m;
1770 	size_t tlen;
1771 
1772 	tlen = key_getspreqmsglen(sp);
1773 	m = m_get2(tlen, M_NOWAIT, MT_DATA, 0);
1774 	if (m == NULL)
1775 		return (NULL);
1776 	m_align(m, tlen);
1777 	m->m_len = tlen;
1778 	if (key_sp2msg(sp, m->m_data, &tlen) != 0) {
1779 		m_freem(m);
1780 		return (NULL);
1781 	}
1782 	return (m);
1783 }
1784 
1785 int
1786 key_sp2msg(struct secpolicy *sp, void *request, size_t *len)
1787 {
1788 	struct sadb_x_ipsecrequest *xisr;
1789 	struct sadb_x_policy *xpl;
1790 	struct ipsecrequest *isr;
1791 	size_t xlen, ilen;
1792 	caddr_t p;
1793 	int error, i;
1794 
1795 	IPSEC_ASSERT(sp != NULL, ("null policy"));
1796 
1797 	xlen = sizeof(*xpl);
1798 	if (*len < xlen)
1799 		return (EINVAL);
1800 
1801 	error = 0;
1802 	bzero(request, *len);
1803 	xpl = (struct sadb_x_policy *)request;
1804 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
1805 	xpl->sadb_x_policy_type = sp->policy;
1806 	xpl->sadb_x_policy_dir = sp->spidx.dir;
1807 	xpl->sadb_x_policy_id = sp->id;
1808 	xpl->sadb_x_policy_priority = sp->priority;
1809 	switch (sp->state) {
1810 	case IPSEC_SPSTATE_IFNET:
1811 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_IFNET;
1812 		break;
1813 	case IPSEC_SPSTATE_PCB:
1814 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_PCB;
1815 		break;
1816 	default:
1817 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_GLOBAL;
1818 	}
1819 
1820 	/* if is the policy for ipsec ? */
1821 	if (sp->policy == IPSEC_POLICY_IPSEC) {
1822 		p = (caddr_t)xpl + sizeof(*xpl);
1823 		for (i = 0; i < sp->tcount; i++) {
1824 			isr = sp->req[i];
1825 			ilen = PFKEY_ALIGN8(sizeof(*xisr) +
1826 			    isr->saidx.src.sa.sa_len +
1827 			    isr->saidx.dst.sa.sa_len);
1828 			xlen += ilen;
1829 			if (xlen > *len) {
1830 				error = ENOBUFS;
1831 				/* Calculate needed size */
1832 				continue;
1833 			}
1834 			xisr = (struct sadb_x_ipsecrequest *)p;
1835 			xisr->sadb_x_ipsecrequest_len = ilen;
1836 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
1837 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
1838 			xisr->sadb_x_ipsecrequest_level = isr->level;
1839 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
1840 
1841 			p += sizeof(*xisr);
1842 			bcopy(&isr->saidx.src, p, isr->saidx.src.sa.sa_len);
1843 			p += isr->saidx.src.sa.sa_len;
1844 			bcopy(&isr->saidx.dst, p, isr->saidx.dst.sa.sa_len);
1845 			p += isr->saidx.dst.sa.sa_len;
1846 		}
1847 	}
1848 	xpl->sadb_x_policy_len = PFKEY_UNIT64(xlen);
1849 	if (error == 0)
1850 		*len = xlen;
1851 	else
1852 		*len = sizeof(*xpl);
1853 	return (error);
1854 }
1855 
1856 /* m will not be freed nor modified */
1857 static struct mbuf *
1858 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
1859     int ndeep, int nitem, ...)
1860 {
1861 	va_list ap;
1862 	int idx;
1863 	int i;
1864 	struct mbuf *result = NULL, *n;
1865 	int len;
1866 
1867 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1868 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1869 
1870 	va_start(ap, nitem);
1871 	for (i = 0; i < nitem; i++) {
1872 		idx = va_arg(ap, int);
1873 		if (idx < 0 || idx > SADB_EXT_MAX)
1874 			goto fail;
1875 		/* don't attempt to pull empty extension */
1876 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
1877 			continue;
1878 		if (idx != SADB_EXT_RESERVED  &&
1879 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
1880 			continue;
1881 
1882 		if (idx == SADB_EXT_RESERVED) {
1883 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
1884 
1885 			IPSEC_ASSERT(len <= MHLEN, ("header too big %u", len));
1886 
1887 			MGETHDR(n, M_NOWAIT, MT_DATA);
1888 			if (!n)
1889 				goto fail;
1890 			n->m_len = len;
1891 			n->m_next = NULL;
1892 			m_copydata(m, 0, sizeof(struct sadb_msg),
1893 			    mtod(n, caddr_t));
1894 		} else if (i < ndeep) {
1895 			len = mhp->extlen[idx];
1896 			n = m_get2(len, M_NOWAIT, MT_DATA, 0);
1897 			if (n == NULL)
1898 				goto fail;
1899 			m_align(n, len);
1900 			n->m_len = len;
1901 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
1902 			    mtod(n, caddr_t));
1903 		} else {
1904 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
1905 			    M_NOWAIT);
1906 		}
1907 		if (n == NULL)
1908 			goto fail;
1909 
1910 		if (result)
1911 			m_cat(result, n);
1912 		else
1913 			result = n;
1914 	}
1915 	va_end(ap);
1916 
1917 	if ((result->m_flags & M_PKTHDR) != 0) {
1918 		result->m_pkthdr.len = 0;
1919 		for (n = result; n; n = n->m_next)
1920 			result->m_pkthdr.len += n->m_len;
1921 	}
1922 
1923 	return result;
1924 
1925 fail:
1926 	m_freem(result);
1927 	va_end(ap);
1928 	return NULL;
1929 }
1930 
1931 /*
1932  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
1933  * add an entry to SP database, when received
1934  *   <base, address(SD), (lifetime(H),) policy>
1935  * from the user(?).
1936  * Adding to SP database,
1937  * and send
1938  *   <base, address(SD), (lifetime(H),) policy>
1939  * to the socket which was send.
1940  *
1941  * SPDADD set a unique policy entry.
1942  * SPDSETIDX like SPDADD without a part of policy requests.
1943  * SPDUPDATE replace a unique policy entry.
1944  *
1945  * XXXAE: serialize this in PF_KEY to avoid races.
1946  * m will always be freed.
1947  */
1948 static int
1949 key_spdadd(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
1950 {
1951 	struct secpolicyindex spidx;
1952 	struct sadb_address *src0, *dst0;
1953 	struct sadb_x_policy *xpl0, *xpl;
1954 	struct sadb_lifetime *lft = NULL;
1955 	struct secpolicy *newsp;
1956 	int error;
1957 
1958 	IPSEC_ASSERT(so != NULL, ("null socket"));
1959 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1960 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1961 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
1962 
1963 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
1964 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
1965 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
1966 		ipseclog((LOG_DEBUG,
1967 		    "%s: invalid message: missing required header.\n",
1968 		    __func__));
1969 		return key_senderror(so, m, EINVAL);
1970 	}
1971 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
1972 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
1973 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
1974 		ipseclog((LOG_DEBUG,
1975 		    "%s: invalid message: wrong header size.\n", __func__));
1976 		return key_senderror(so, m, EINVAL);
1977 	}
1978 	if (!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD)) {
1979 		if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD)) {
1980 			ipseclog((LOG_DEBUG,
1981 			    "%s: invalid message: wrong header size.\n",
1982 			    __func__));
1983 			return key_senderror(so, m, EINVAL);
1984 		}
1985 		lft = (struct sadb_lifetime *)mhp->ext[SADB_EXT_LIFETIME_HARD];
1986 	}
1987 
1988 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
1989 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
1990 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
1991 
1992 	/* check the direciton */
1993 	switch (xpl0->sadb_x_policy_dir) {
1994 	case IPSEC_DIR_INBOUND:
1995 	case IPSEC_DIR_OUTBOUND:
1996 		break;
1997 	default:
1998 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
1999 		return key_senderror(so, m, EINVAL);
2000 	}
2001 	/* key_spdadd() accepts DISCARD, NONE and IPSEC. */
2002 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2003 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2004 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2005 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2006 		return key_senderror(so, m, EINVAL);
2007 	}
2008 
2009 	/* policy requests are mandatory when action is ipsec. */
2010 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
2011 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
2012 		ipseclog((LOG_DEBUG,
2013 		    "%s: policy requests required.\n", __func__));
2014 		return key_senderror(so, m, EINVAL);
2015 	}
2016 
2017 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
2018 	    (struct sockaddr *)(dst0 + 1));
2019 	if (error != 0 ||
2020 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2021 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2022 		return key_senderror(so, m, error);
2023 	}
2024 	/* make secindex */
2025 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2026 	                src0 + 1,
2027 	                dst0 + 1,
2028 	                src0->sadb_address_prefixlen,
2029 	                dst0->sadb_address_prefixlen,
2030 	                src0->sadb_address_proto,
2031 	                &spidx);
2032 	/* Checking there is SP already or not. */
2033 	newsp = key_getsp(&spidx);
2034 	if (newsp != NULL) {
2035 		if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2036 			KEYDBG(KEY_STAMP,
2037 			    printf("%s: unlink SP(%p) for SPDUPDATE\n",
2038 				__func__, newsp));
2039 			KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
2040 			key_unlink(newsp);
2041 			key_freesp(&newsp);
2042 		} else {
2043 			key_freesp(&newsp);
2044 			ipseclog((LOG_DEBUG,
2045 			    "%s: a SP entry exists already.\n", __func__));
2046 			return (key_senderror(so, m, EEXIST));
2047 		}
2048 	}
2049 
2050 	/* allocate new SP entry */
2051 	if ((newsp = key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error)) == NULL) {
2052 		return key_senderror(so, m, error);
2053 	}
2054 
2055 	newsp->lastused = newsp->created = time_second;
2056 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
2057 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
2058 	bcopy(&spidx, &newsp->spidx, sizeof(spidx));
2059 
2060 	/* XXXAE: there is race between key_getsp() and key_insertsp() */
2061 	SPTREE_WLOCK();
2062 	if ((newsp->id = key_getnewspid()) == 0) {
2063 		SPTREE_WUNLOCK();
2064 		key_freesp(&newsp);
2065 		return key_senderror(so, m, ENOBUFS);
2066 	}
2067 	key_insertsp(newsp);
2068 	SPTREE_WUNLOCK();
2069 	if (SPDCACHE_ENABLED())
2070 		spdcache_clear();
2071 
2072 	KEYDBG(KEY_STAMP,
2073 	    printf("%s: SP(%p)\n", __func__, newsp));
2074 	KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
2075 
2076     {
2077 	struct mbuf *n, *mpolicy;
2078 	struct sadb_msg *newmsg;
2079 	int off;
2080 
2081 	/* create new sadb_msg to reply. */
2082 	if (lft) {
2083 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
2084 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
2085 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2086 	} else {
2087 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
2088 		    SADB_X_EXT_POLICY,
2089 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2090 	}
2091 	if (!n)
2092 		return key_senderror(so, m, ENOBUFS);
2093 
2094 	if (n->m_len < sizeof(*newmsg)) {
2095 		n = m_pullup(n, sizeof(*newmsg));
2096 		if (!n)
2097 			return key_senderror(so, m, ENOBUFS);
2098 	}
2099 	newmsg = mtod(n, struct sadb_msg *);
2100 	newmsg->sadb_msg_errno = 0;
2101 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2102 
2103 	off = 0;
2104 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
2105 	    sizeof(*xpl), &off);
2106 	if (mpolicy == NULL) {
2107 		/* n is already freed */
2108 		return key_senderror(so, m, ENOBUFS);
2109 	}
2110 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, caddr_t) + off);
2111 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
2112 		m_freem(n);
2113 		return key_senderror(so, m, EINVAL);
2114 	}
2115 	xpl->sadb_x_policy_id = newsp->id;
2116 
2117 	m_freem(m);
2118 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2119     }
2120 }
2121 
2122 /*
2123  * get new policy id.
2124  * OUT:
2125  *	0:	failure.
2126  *	others: success.
2127  */
2128 static uint32_t
2129 key_getnewspid(void)
2130 {
2131 	struct secpolicy *sp;
2132 	uint32_t newid = 0;
2133 	int count = V_key_spi_trycnt;	/* XXX */
2134 
2135 	SPTREE_WLOCK_ASSERT();
2136 	while (count--) {
2137 		if (V_policy_id == ~0) /* overflowed */
2138 			newid = V_policy_id = 1;
2139 		else
2140 			newid = ++V_policy_id;
2141 		LIST_FOREACH(sp, SPHASH_HASH(newid), idhash) {
2142 			if (sp->id == newid)
2143 				break;
2144 		}
2145 		if (sp == NULL)
2146 			break;
2147 	}
2148 	if (count == 0 || newid == 0) {
2149 		ipseclog((LOG_DEBUG, "%s: failed to allocate policy id.\n",
2150 		    __func__));
2151 		return (0);
2152 	}
2153 	return (newid);
2154 }
2155 
2156 /*
2157  * SADB_SPDDELETE processing
2158  * receive
2159  *   <base, address(SD), policy(*)>
2160  * from the user(?), and set SADB_SASTATE_DEAD,
2161  * and send,
2162  *   <base, address(SD), policy(*)>
2163  * to the ikmpd.
2164  * policy(*) including direction of policy.
2165  *
2166  * m will always be freed.
2167  */
2168 static int
2169 key_spddelete(struct socket *so, struct mbuf *m,
2170     const struct sadb_msghdr *mhp)
2171 {
2172 	struct secpolicyindex spidx;
2173 	struct sadb_address *src0, *dst0;
2174 	struct sadb_x_policy *xpl0;
2175 	struct secpolicy *sp;
2176 
2177 	IPSEC_ASSERT(so != NULL, ("null so"));
2178 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2179 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2180 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2181 
2182 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
2183 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
2184 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
2185 		ipseclog((LOG_DEBUG,
2186 		    "%s: invalid message: missing required header.\n",
2187 		    __func__));
2188 		return key_senderror(so, m, EINVAL);
2189 	}
2190 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
2191 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
2192 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2193 		ipseclog((LOG_DEBUG,
2194 		    "%s: invalid message: wrong header size.\n", __func__));
2195 		return key_senderror(so, m, EINVAL);
2196 	}
2197 
2198 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
2199 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
2200 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
2201 
2202 	/* check the direciton */
2203 	switch (xpl0->sadb_x_policy_dir) {
2204 	case IPSEC_DIR_INBOUND:
2205 	case IPSEC_DIR_OUTBOUND:
2206 		break;
2207 	default:
2208 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2209 		return key_senderror(so, m, EINVAL);
2210 	}
2211 	/* Only DISCARD, NONE and IPSEC are allowed */
2212 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2213 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2214 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2215 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2216 		return key_senderror(so, m, EINVAL);
2217 	}
2218 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
2219 	    (struct sockaddr *)(dst0 + 1)) != 0 ||
2220 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2221 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2222 		return key_senderror(so, m, EINVAL);
2223 	}
2224 	/* make secindex */
2225 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2226 	                src0 + 1,
2227 	                dst0 + 1,
2228 	                src0->sadb_address_prefixlen,
2229 	                dst0->sadb_address_prefixlen,
2230 	                src0->sadb_address_proto,
2231 	                &spidx);
2232 
2233 	/* Is there SP in SPD ? */
2234 	if ((sp = key_getsp(&spidx)) == NULL) {
2235 		ipseclog((LOG_DEBUG, "%s: no SP found.\n", __func__));
2236 		return key_senderror(so, m, EINVAL);
2237 	}
2238 
2239 	/* save policy id to buffer to be returned. */
2240 	xpl0->sadb_x_policy_id = sp->id;
2241 
2242 	KEYDBG(KEY_STAMP,
2243 	    printf("%s: SP(%p)\n", __func__, sp));
2244 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2245 	key_unlink(sp);
2246 	key_freesp(&sp);
2247 
2248     {
2249 	struct mbuf *n;
2250 	struct sadb_msg *newmsg;
2251 
2252 	/* create new sadb_msg to reply. */
2253 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2254 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2255 	if (!n)
2256 		return key_senderror(so, m, ENOBUFS);
2257 
2258 	newmsg = mtod(n, struct sadb_msg *);
2259 	newmsg->sadb_msg_errno = 0;
2260 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2261 
2262 	m_freem(m);
2263 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2264     }
2265 }
2266 
2267 /*
2268  * SADB_SPDDELETE2 processing
2269  * receive
2270  *   <base, policy(*)>
2271  * from the user(?), and set SADB_SASTATE_DEAD,
2272  * and send,
2273  *   <base, policy(*)>
2274  * to the ikmpd.
2275  * policy(*) including direction of policy.
2276  *
2277  * m will always be freed.
2278  */
2279 static int
2280 key_spddelete2(struct socket *so, struct mbuf *m,
2281     const struct sadb_msghdr *mhp)
2282 {
2283 	struct secpolicy *sp;
2284 	uint32_t id;
2285 
2286 	IPSEC_ASSERT(so != NULL, ("null socket"));
2287 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2288 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2289 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2290 
2291 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2292 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2293 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2294 		    __func__));
2295 		return key_senderror(so, m, EINVAL);
2296 	}
2297 
2298 	id = ((struct sadb_x_policy *)
2299 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2300 
2301 	/* Is there SP in SPD ? */
2302 	if ((sp = key_getspbyid(id)) == NULL) {
2303 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2304 		    __func__, id));
2305 		return key_senderror(so, m, EINVAL);
2306 	}
2307 
2308 	KEYDBG(KEY_STAMP,
2309 	    printf("%s: SP(%p)\n", __func__, sp));
2310 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2311 	key_unlink(sp);
2312 	if (sp->state != IPSEC_SPSTATE_DEAD) {
2313 		ipseclog((LOG_DEBUG, "%s: failed to delete SP with id %u.\n",
2314 		    __func__, id));
2315 		key_freesp(&sp);
2316 		return (key_senderror(so, m, EACCES));
2317 	}
2318 	key_freesp(&sp);
2319 
2320     {
2321 	struct mbuf *n, *nn;
2322 	struct sadb_msg *newmsg;
2323 	int off, len;
2324 
2325 	/* create new sadb_msg to reply. */
2326 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2327 
2328 	MGETHDR(n, M_NOWAIT, MT_DATA);
2329 	if (n && len > MHLEN) {
2330 		if (!(MCLGET(n, M_NOWAIT))) {
2331 			m_freem(n);
2332 			n = NULL;
2333 		}
2334 	}
2335 	if (!n)
2336 		return key_senderror(so, m, ENOBUFS);
2337 
2338 	n->m_len = len;
2339 	n->m_next = NULL;
2340 	off = 0;
2341 
2342 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
2343 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2344 
2345 	IPSEC_ASSERT(off == len, ("length inconsistency (off %u len %u)",
2346 		off, len));
2347 
2348 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2349 	    mhp->extlen[SADB_X_EXT_POLICY], M_NOWAIT);
2350 	if (!n->m_next) {
2351 		m_freem(n);
2352 		return key_senderror(so, m, ENOBUFS);
2353 	}
2354 
2355 	n->m_pkthdr.len = 0;
2356 	for (nn = n; nn; nn = nn->m_next)
2357 		n->m_pkthdr.len += nn->m_len;
2358 
2359 	newmsg = mtod(n, struct sadb_msg *);
2360 	newmsg->sadb_msg_errno = 0;
2361 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2362 
2363 	m_freem(m);
2364 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2365     }
2366 }
2367 
2368 /*
2369  * SADB_X_SPDGET processing
2370  * receive
2371  *   <base, policy(*)>
2372  * from the user(?),
2373  * and send,
2374  *   <base, address(SD), policy>
2375  * to the ikmpd.
2376  * policy(*) including direction of policy.
2377  *
2378  * m will always be freed.
2379  */
2380 static int
2381 key_spdget(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2382 {
2383 	struct secpolicy *sp;
2384 	struct mbuf *n;
2385 	uint32_t id;
2386 
2387 	IPSEC_ASSERT(so != NULL, ("null socket"));
2388 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2389 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2390 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2391 
2392 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2393 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2394 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2395 		    __func__));
2396 		return key_senderror(so, m, EINVAL);
2397 	}
2398 
2399 	id = ((struct sadb_x_policy *)
2400 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2401 
2402 	/* Is there SP in SPD ? */
2403 	if ((sp = key_getspbyid(id)) == NULL) {
2404 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2405 		    __func__, id));
2406 		return key_senderror(so, m, ENOENT);
2407 	}
2408 
2409 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2410 	    mhp->msg->sadb_msg_pid);
2411 	key_freesp(&sp);
2412 	if (n != NULL) {
2413 		m_freem(m);
2414 		return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2415 	} else
2416 		return key_senderror(so, m, ENOBUFS);
2417 }
2418 
2419 /*
2420  * SADB_X_SPDACQUIRE processing.
2421  * Acquire policy and SA(s) for a *OUTBOUND* packet.
2422  * send
2423  *   <base, policy(*)>
2424  * to KMD, and expect to receive
2425  *   <base> with SADB_X_SPDACQUIRE if error occurred,
2426  * or
2427  *   <base, policy>
2428  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2429  * policy(*) is without policy requests.
2430  *
2431  *    0     : succeed
2432  *    others: error number
2433  */
2434 int
2435 key_spdacquire(struct secpolicy *sp)
2436 {
2437 	struct mbuf *result = NULL, *m;
2438 	struct secspacq *newspacq;
2439 
2440 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2441 	IPSEC_ASSERT(sp->req == NULL, ("policy exists"));
2442 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
2443 		("policy not IPSEC %u", sp->policy));
2444 
2445 	/* Get an entry to check whether sent message or not. */
2446 	newspacq = key_getspacq(&sp->spidx);
2447 	if (newspacq != NULL) {
2448 		if (V_key_blockacq_count < newspacq->count) {
2449 			/* reset counter and do send message. */
2450 			newspacq->count = 0;
2451 		} else {
2452 			/* increment counter and do nothing. */
2453 			newspacq->count++;
2454 			SPACQ_UNLOCK();
2455 			return (0);
2456 		}
2457 		SPACQ_UNLOCK();
2458 	} else {
2459 		/* make new entry for blocking to send SADB_ACQUIRE. */
2460 		newspacq = key_newspacq(&sp->spidx);
2461 		if (newspacq == NULL)
2462 			return ENOBUFS;
2463 	}
2464 
2465 	/* create new sadb_msg to reply. */
2466 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2467 	if (!m)
2468 		return ENOBUFS;
2469 
2470 	result = m;
2471 
2472 	result->m_pkthdr.len = 0;
2473 	for (m = result; m; m = m->m_next)
2474 		result->m_pkthdr.len += m->m_len;
2475 
2476 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2477 	    PFKEY_UNIT64(result->m_pkthdr.len);
2478 
2479 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2480 }
2481 
2482 /*
2483  * SADB_SPDFLUSH processing
2484  * receive
2485  *   <base>
2486  * from the user, and free all entries in secpctree.
2487  * and send,
2488  *   <base>
2489  * to the user.
2490  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2491  *
2492  * m will always be freed.
2493  */
2494 static int
2495 key_spdflush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2496 {
2497 	struct secpolicy_queue drainq;
2498 	struct sadb_msg *newmsg;
2499 	struct secpolicy *sp, *nextsp;
2500 	u_int dir;
2501 
2502 	IPSEC_ASSERT(so != NULL, ("null socket"));
2503 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2504 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2505 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2506 
2507 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2508 		return key_senderror(so, m, EINVAL);
2509 
2510 	TAILQ_INIT(&drainq);
2511 	SPTREE_WLOCK();
2512 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2513 		TAILQ_CONCAT(&drainq, &V_sptree[dir], chain);
2514 	}
2515 	/*
2516 	 * We need to set state to DEAD for each policy to be sure,
2517 	 * that another thread won't try to unlink it.
2518 	 * Also remove SP from sphash.
2519 	 */
2520 	TAILQ_FOREACH(sp, &drainq, chain) {
2521 		sp->state = IPSEC_SPSTATE_DEAD;
2522 		LIST_REMOVE(sp, idhash);
2523 	}
2524 	V_sp_genid++;
2525 	V_spd_size = 0;
2526 	SPTREE_WUNLOCK();
2527 	if (SPDCACHE_ENABLED())
2528 		spdcache_clear();
2529 	sp = TAILQ_FIRST(&drainq);
2530 	while (sp != NULL) {
2531 		nextsp = TAILQ_NEXT(sp, chain);
2532 		key_freesp(&sp);
2533 		sp = nextsp;
2534 	}
2535 
2536 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2537 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
2538 		return key_senderror(so, m, ENOBUFS);
2539 	}
2540 
2541 	if (m->m_next)
2542 		m_freem(m->m_next);
2543 	m->m_next = NULL;
2544 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2545 	newmsg = mtod(m, struct sadb_msg *);
2546 	newmsg->sadb_msg_errno = 0;
2547 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2548 
2549 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2550 }
2551 
2552 static uint8_t
2553 key_satype2scopemask(uint8_t satype)
2554 {
2555 
2556 	if (satype == IPSEC_POLICYSCOPE_ANY)
2557 		return (0xff);
2558 	return (satype);
2559 }
2560 /*
2561  * SADB_SPDDUMP processing
2562  * receive
2563  *   <base>
2564  * from the user, and dump all SP leaves and send,
2565  *   <base> .....
2566  * to the ikmpd.
2567  *
2568  * NOTE:
2569  *   sadb_msg_satype is considered as mask of policy scopes.
2570  *   m will always be freed.
2571  */
2572 static int
2573 key_spddump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2574 {
2575 	SPTREE_RLOCK_TRACKER;
2576 	struct secpolicy *sp;
2577 	struct mbuf *n;
2578 	int cnt;
2579 	u_int dir, scope;
2580 
2581 	IPSEC_ASSERT(so != NULL, ("null socket"));
2582 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2583 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2584 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2585 
2586 	/* search SPD entry and get buffer size. */
2587 	cnt = 0;
2588 	scope = key_satype2scopemask(mhp->msg->sadb_msg_satype);
2589 	SPTREE_RLOCK();
2590 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2591 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2592 			TAILQ_FOREACH(sp, &V_sptree[dir], chain)
2593 				cnt++;
2594 		}
2595 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2596 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain)
2597 				cnt++;
2598 		}
2599 	}
2600 
2601 	if (cnt == 0) {
2602 		SPTREE_RUNLOCK();
2603 		return key_senderror(so, m, ENOENT);
2604 	}
2605 
2606 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2607 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2608 			TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
2609 				--cnt;
2610 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2611 				    mhp->msg->sadb_msg_pid);
2612 
2613 				if (n != NULL)
2614 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2615 			}
2616 		}
2617 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2618 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain) {
2619 				--cnt;
2620 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2621 				    mhp->msg->sadb_msg_pid);
2622 
2623 				if (n != NULL)
2624 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2625 			}
2626 		}
2627 	}
2628 
2629 	SPTREE_RUNLOCK();
2630 	m_freem(m);
2631 	return (0);
2632 }
2633 
2634 static struct mbuf *
2635 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq,
2636     u_int32_t pid)
2637 {
2638 	struct mbuf *result = NULL, *m;
2639 	struct seclifetime lt;
2640 
2641 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid, sp->refcnt);
2642 	if (!m)
2643 		goto fail;
2644 	result = m;
2645 
2646 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2647 	    &sp->spidx.src.sa, sp->spidx.prefs,
2648 	    sp->spidx.ul_proto);
2649 	if (!m)
2650 		goto fail;
2651 	m_cat(result, m);
2652 
2653 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2654 	    &sp->spidx.dst.sa, sp->spidx.prefd,
2655 	    sp->spidx.ul_proto);
2656 	if (!m)
2657 		goto fail;
2658 	m_cat(result, m);
2659 
2660 	m = key_sp2mbuf(sp);
2661 	if (!m)
2662 		goto fail;
2663 	m_cat(result, m);
2664 
2665 	if(sp->lifetime){
2666 		lt.addtime=sp->created;
2667 		lt.usetime= sp->lastused;
2668 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_CURRENT);
2669 		if (!m)
2670 			goto fail;
2671 		m_cat(result, m);
2672 
2673 		lt.addtime=sp->lifetime;
2674 		lt.usetime= sp->validtime;
2675 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_HARD);
2676 		if (!m)
2677 			goto fail;
2678 		m_cat(result, m);
2679 	}
2680 
2681 	if ((result->m_flags & M_PKTHDR) == 0)
2682 		goto fail;
2683 
2684 	if (result->m_len < sizeof(struct sadb_msg)) {
2685 		result = m_pullup(result, sizeof(struct sadb_msg));
2686 		if (result == NULL)
2687 			goto fail;
2688 	}
2689 
2690 	result->m_pkthdr.len = 0;
2691 	for (m = result; m; m = m->m_next)
2692 		result->m_pkthdr.len += m->m_len;
2693 
2694 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2695 	    PFKEY_UNIT64(result->m_pkthdr.len);
2696 
2697 	return result;
2698 
2699 fail:
2700 	m_freem(result);
2701 	return NULL;
2702 }
2703 /*
2704  * get PFKEY message length for security policy and request.
2705  */
2706 static size_t
2707 key_getspreqmsglen(struct secpolicy *sp)
2708 {
2709 	size_t tlen, len;
2710 	int i;
2711 
2712 	tlen = sizeof(struct sadb_x_policy);
2713 	/* if is the policy for ipsec ? */
2714 	if (sp->policy != IPSEC_POLICY_IPSEC)
2715 		return (tlen);
2716 
2717 	/* get length of ipsec requests */
2718 	for (i = 0; i < sp->tcount; i++) {
2719 		len = sizeof(struct sadb_x_ipsecrequest)
2720 			+ sp->req[i]->saidx.src.sa.sa_len
2721 			+ sp->req[i]->saidx.dst.sa.sa_len;
2722 
2723 		tlen += PFKEY_ALIGN8(len);
2724 	}
2725 	return (tlen);
2726 }
2727 
2728 /*
2729  * SADB_SPDEXPIRE processing
2730  * send
2731  *   <base, address(SD), lifetime(CH), policy>
2732  * to KMD by PF_KEY.
2733  *
2734  * OUT:	0	: succeed
2735  *	others	: error number
2736  */
2737 static int
2738 key_spdexpire(struct secpolicy *sp)
2739 {
2740 	struct sadb_lifetime *lt;
2741 	struct mbuf *result = NULL, *m;
2742 	int len, error = -1;
2743 
2744 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2745 
2746 	KEYDBG(KEY_STAMP,
2747 	    printf("%s: SP(%p)\n", __func__, sp));
2748 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2749 
2750 	/* set msg header */
2751 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0);
2752 	if (!m) {
2753 		error = ENOBUFS;
2754 		goto fail;
2755 	}
2756 	result = m;
2757 
2758 	/* create lifetime extension (current and hard) */
2759 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
2760 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
2761 	if (m == NULL) {
2762 		error = ENOBUFS;
2763 		goto fail;
2764 	}
2765 	m_align(m, len);
2766 	m->m_len = len;
2767 	bzero(mtod(m, caddr_t), len);
2768 	lt = mtod(m, struct sadb_lifetime *);
2769 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2770 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
2771 	lt->sadb_lifetime_allocations = 0;
2772 	lt->sadb_lifetime_bytes = 0;
2773 	lt->sadb_lifetime_addtime = sp->created;
2774 	lt->sadb_lifetime_usetime = sp->lastused;
2775 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
2776 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2777 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
2778 	lt->sadb_lifetime_allocations = 0;
2779 	lt->sadb_lifetime_bytes = 0;
2780 	lt->sadb_lifetime_addtime = sp->lifetime;
2781 	lt->sadb_lifetime_usetime = sp->validtime;
2782 	m_cat(result, m);
2783 
2784 	/* set sadb_address for source */
2785 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2786 	    &sp->spidx.src.sa,
2787 	    sp->spidx.prefs, sp->spidx.ul_proto);
2788 	if (!m) {
2789 		error = ENOBUFS;
2790 		goto fail;
2791 	}
2792 	m_cat(result, m);
2793 
2794 	/* set sadb_address for destination */
2795 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2796 	    &sp->spidx.dst.sa,
2797 	    sp->spidx.prefd, sp->spidx.ul_proto);
2798 	if (!m) {
2799 		error = ENOBUFS;
2800 		goto fail;
2801 	}
2802 	m_cat(result, m);
2803 
2804 	/* set secpolicy */
2805 	m = key_sp2mbuf(sp);
2806 	if (!m) {
2807 		error = ENOBUFS;
2808 		goto fail;
2809 	}
2810 	m_cat(result, m);
2811 
2812 	if ((result->m_flags & M_PKTHDR) == 0) {
2813 		error = EINVAL;
2814 		goto fail;
2815 	}
2816 
2817 	if (result->m_len < sizeof(struct sadb_msg)) {
2818 		result = m_pullup(result, sizeof(struct sadb_msg));
2819 		if (result == NULL) {
2820 			error = ENOBUFS;
2821 			goto fail;
2822 		}
2823 	}
2824 
2825 	result->m_pkthdr.len = 0;
2826 	for (m = result; m; m = m->m_next)
2827 		result->m_pkthdr.len += m->m_len;
2828 
2829 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2830 	    PFKEY_UNIT64(result->m_pkthdr.len);
2831 
2832 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
2833 
2834  fail:
2835 	if (result)
2836 		m_freem(result);
2837 	return error;
2838 }
2839 
2840 /* %%% SAD management */
2841 /*
2842  * allocating and initialize new SA head.
2843  * OUT:	NULL	: failure due to the lack of memory.
2844  *	others	: pointer to new SA head.
2845  */
2846 static struct secashead *
2847 key_newsah(struct secasindex *saidx)
2848 {
2849 	struct secashead *sah;
2850 
2851 	sah = malloc(sizeof(struct secashead), M_IPSEC_SAH,
2852 	    M_NOWAIT | M_ZERO);
2853 	if (sah == NULL) {
2854 		PFKEYSTAT_INC(in_nomem);
2855 		return (NULL);
2856 	}
2857 	TAILQ_INIT(&sah->savtree_larval);
2858 	TAILQ_INIT(&sah->savtree_alive);
2859 	sah->saidx = *saidx;
2860 	sah->state = SADB_SASTATE_DEAD;
2861 	SAH_INITREF(sah);
2862 
2863 	KEYDBG(KEY_STAMP,
2864 	    printf("%s: SAH(%p)\n", __func__, sah));
2865 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2866 	return (sah);
2867 }
2868 
2869 static void
2870 key_freesah(struct secashead **psah)
2871 {
2872 	struct secashead *sah = *psah;
2873 
2874 	if (SAH_DELREF(sah) == 0)
2875 		return;
2876 
2877 	KEYDBG(KEY_STAMP,
2878 	    printf("%s: last reference to SAH(%p)\n", __func__, sah));
2879 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2880 
2881 	*psah = NULL;
2882 	key_delsah(sah);
2883 }
2884 
2885 static void
2886 key_delsah(struct secashead *sah)
2887 {
2888 	IPSEC_ASSERT(sah != NULL, ("NULL sah"));
2889 	IPSEC_ASSERT(sah->state == SADB_SASTATE_DEAD,
2890 	    ("Attempt to free non DEAD SAH %p", sah));
2891 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_larval),
2892 	    ("Attempt to free SAH %p with LARVAL SA", sah));
2893 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_alive),
2894 	    ("Attempt to free SAH %p with ALIVE SA", sah));
2895 
2896 	free(sah, M_IPSEC_SAH);
2897 }
2898 
2899 /*
2900  * allocating a new SA for key_add() and key_getspi() call,
2901  * and copy the values of mhp into new buffer.
2902  * When SAD message type is SADB_GETSPI set SA state to LARVAL.
2903  * For SADB_ADD create and initialize SA with MATURE state.
2904  * OUT:	NULL	: fail
2905  *	others	: pointer to new secasvar.
2906  */
2907 static struct secasvar *
2908 key_newsav(const struct sadb_msghdr *mhp, struct secasindex *saidx,
2909     uint32_t spi, int *errp)
2910 {
2911 	struct secashead *sah;
2912 	struct secasvar *sav;
2913 	int isnew;
2914 
2915 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2916 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2917 	IPSEC_ASSERT(mhp->msg->sadb_msg_type == SADB_GETSPI ||
2918 	    mhp->msg->sadb_msg_type == SADB_ADD, ("wrong message type"));
2919 
2920 	sav = NULL;
2921 	sah = NULL;
2922 	/* check SPI value */
2923 	switch (saidx->proto) {
2924 	case IPPROTO_ESP:
2925 	case IPPROTO_AH:
2926 		/*
2927 		 * RFC 4302, 2.4. Security Parameters Index (SPI), SPI values
2928 		 * 1-255 reserved by IANA for future use,
2929 		 * 0 for implementation specific, local use.
2930 		 */
2931 		if (ntohl(spi) <= 255) {
2932 			ipseclog((LOG_DEBUG, "%s: illegal range of SPI %u.\n",
2933 			    __func__, ntohl(spi)));
2934 			*errp = EINVAL;
2935 			goto done;
2936 		}
2937 		break;
2938 	}
2939 
2940 	sav = malloc(sizeof(struct secasvar), M_IPSEC_SA, M_NOWAIT | M_ZERO);
2941 	if (sav == NULL) {
2942 		*errp = ENOBUFS;
2943 		goto done;
2944 	}
2945 	sav->lock = malloc(sizeof(struct mtx), M_IPSEC_MISC,
2946 	    M_NOWAIT | M_ZERO);
2947 	if (sav->lock == NULL) {
2948 		*errp = ENOBUFS;
2949 		goto done;
2950 	}
2951 	mtx_init(sav->lock, "ipsec association", NULL, MTX_DEF);
2952 	sav->lft_c = uma_zalloc_pcpu(V_key_lft_zone, M_NOWAIT);
2953 	if (sav->lft_c == NULL) {
2954 		*errp = ENOBUFS;
2955 		goto done;
2956 	}
2957 	counter_u64_zero(sav->lft_c_allocations);
2958 	counter_u64_zero(sav->lft_c_bytes);
2959 
2960 	sav->spi = spi;
2961 	sav->seq = mhp->msg->sadb_msg_seq;
2962 	sav->state = SADB_SASTATE_LARVAL;
2963 	sav->pid = (pid_t)mhp->msg->sadb_msg_pid;
2964 	SAV_INITREF(sav);
2965 again:
2966 	sah = key_getsah(saidx);
2967 	if (sah == NULL) {
2968 		/* create a new SA index */
2969 		sah = key_newsah(saidx);
2970 		if (sah == NULL) {
2971 			ipseclog((LOG_DEBUG,
2972 			    "%s: No more memory.\n", __func__));
2973 			*errp = ENOBUFS;
2974 			goto done;
2975 		}
2976 		isnew = 1;
2977 	} else
2978 		isnew = 0;
2979 
2980 	sav->sah = sah;
2981 	if (mhp->msg->sadb_msg_type == SADB_GETSPI) {
2982 		sav->created = time_second;
2983 	} else if (sav->state == SADB_SASTATE_LARVAL) {
2984 		/*
2985 		 * Do not call key_setsaval() second time in case
2986 		 * of `goto again`. We will have MATURE state.
2987 		 */
2988 		*errp = key_setsaval(sav, mhp);
2989 		if (*errp != 0)
2990 			goto done;
2991 		sav->state = SADB_SASTATE_MATURE;
2992 	}
2993 
2994 	SAHTREE_WLOCK();
2995 	/*
2996 	 * Check that existing SAH wasn't unlinked.
2997 	 * Since we didn't hold the SAHTREE lock, it is possible,
2998 	 * that callout handler or key_flush() or key_delete() could
2999 	 * unlink this SAH.
3000 	 */
3001 	if (isnew == 0 && sah->state == SADB_SASTATE_DEAD) {
3002 		SAHTREE_WUNLOCK();
3003 		key_freesah(&sah);	/* reference from key_getsah() */
3004 		goto again;
3005 	}
3006 	if (isnew != 0) {
3007 		/*
3008 		 * Add new SAH into SADB.
3009 		 *
3010 		 * XXXAE: we can serialize key_add and key_getspi calls, so
3011 		 * several threads will not fight in the race.
3012 		 * Otherwise we should check under SAHTREE lock, that this
3013 		 * SAH would not added twice.
3014 		 */
3015 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
3016 		/* Add new SAH into hash by addresses */
3017 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
3018 		/* Now we are linked in the chain */
3019 		sah->state = SADB_SASTATE_MATURE;
3020 		/*
3021 		 * SAV references this new SAH.
3022 		 * In case of existing SAH we reuse reference
3023 		 * from key_getsah().
3024 		 */
3025 		SAH_ADDREF(sah);
3026 	}
3027 	/* Link SAV with SAH */
3028 	if (sav->state == SADB_SASTATE_MATURE)
3029 		TAILQ_INSERT_HEAD(&sah->savtree_alive, sav, chain);
3030 	else
3031 		TAILQ_INSERT_HEAD(&sah->savtree_larval, sav, chain);
3032 	/* Add SAV into SPI hash */
3033 	LIST_INSERT_HEAD(SAVHASH_HASH(sav->spi), sav, spihash);
3034 	SAHTREE_WUNLOCK();
3035 	*errp = 0;	/* success */
3036 done:
3037 	if (*errp != 0) {
3038 		if (sav != NULL) {
3039 			if (sav->lock != NULL) {
3040 				mtx_destroy(sav->lock);
3041 				free(sav->lock, M_IPSEC_MISC);
3042 			}
3043 			if (sav->lft_c != NULL)
3044 				uma_zfree_pcpu(V_key_lft_zone, sav->lft_c);
3045 			free(sav, M_IPSEC_SA), sav = NULL;
3046 		}
3047 		if (sah != NULL)
3048 			key_freesah(&sah);
3049 		if (*errp == ENOBUFS) {
3050 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3051 			    __func__));
3052 			PFKEYSTAT_INC(in_nomem);
3053 		}
3054 	}
3055 	return (sav);
3056 }
3057 
3058 /*
3059  * free() SA variable entry.
3060  */
3061 static void
3062 key_cleansav(struct secasvar *sav)
3063 {
3064 
3065 	if (sav->natt != NULL) {
3066 		free(sav->natt, M_IPSEC_MISC);
3067 		sav->natt = NULL;
3068 	}
3069 	if (sav->flags & SADB_X_EXT_F_CLONED)
3070 		return;
3071 	/*
3072 	 * Cleanup xform state.  Note that zeroize'ing causes the
3073 	 * keys to be cleared; otherwise we must do it ourself.
3074 	 */
3075 	if (sav->tdb_xform != NULL) {
3076 		sav->tdb_xform->xf_zeroize(sav);
3077 		sav->tdb_xform = NULL;
3078 	} else {
3079 		if (sav->key_auth != NULL)
3080 			bzero(sav->key_auth->key_data, _KEYLEN(sav->key_auth));
3081 		if (sav->key_enc != NULL)
3082 			bzero(sav->key_enc->key_data, _KEYLEN(sav->key_enc));
3083 	}
3084 	if (sav->key_auth != NULL) {
3085 		if (sav->key_auth->key_data != NULL)
3086 			free(sav->key_auth->key_data, M_IPSEC_MISC);
3087 		free(sav->key_auth, M_IPSEC_MISC);
3088 		sav->key_auth = NULL;
3089 	}
3090 	if (sav->key_enc != NULL) {
3091 		if (sav->key_enc->key_data != NULL)
3092 			free(sav->key_enc->key_data, M_IPSEC_MISC);
3093 		free(sav->key_enc, M_IPSEC_MISC);
3094 		sav->key_enc = NULL;
3095 	}
3096 	if (sav->replay != NULL) {
3097 		if (sav->replay->bitmap != NULL)
3098 			free(sav->replay->bitmap, M_IPSEC_MISC);
3099 		free(sav->replay, M_IPSEC_MISC);
3100 		sav->replay = NULL;
3101 	}
3102 	if (sav->lft_h != NULL) {
3103 		free(sav->lft_h, M_IPSEC_MISC);
3104 		sav->lft_h = NULL;
3105 	}
3106 	if (sav->lft_s != NULL) {
3107 		free(sav->lft_s, M_IPSEC_MISC);
3108 		sav->lft_s = NULL;
3109 	}
3110 }
3111 
3112 /*
3113  * free() SA variable entry.
3114  */
3115 static void
3116 key_delsav(struct secasvar *sav)
3117 {
3118 	IPSEC_ASSERT(sav != NULL, ("null sav"));
3119 	IPSEC_ASSERT(sav->state == SADB_SASTATE_DEAD,
3120 	    ("attempt to free non DEAD SA %p", sav));
3121 	IPSEC_ASSERT(sav->refcnt == 0, ("reference count %u > 0",
3122 	    sav->refcnt));
3123 
3124 	/*
3125 	 * SA must be unlinked from the chain and hashtbl.
3126 	 * If SA was cloned, we leave all fields untouched,
3127 	 * except NAT-T config.
3128 	 */
3129 	key_cleansav(sav);
3130 	if ((sav->flags & SADB_X_EXT_F_CLONED) == 0) {
3131 		mtx_destroy(sav->lock);
3132 		free(sav->lock, M_IPSEC_MISC);
3133 		uma_zfree_pcpu(V_key_lft_zone, sav->lft_c);
3134 	}
3135 	free(sav, M_IPSEC_SA);
3136 }
3137 
3138 /*
3139  * search SAH.
3140  * OUT:
3141  *	NULL	: not found
3142  *	others	: found, referenced pointer to a SAH.
3143  */
3144 static struct secashead *
3145 key_getsah(struct secasindex *saidx)
3146 {
3147 	SAHTREE_RLOCK_TRACKER;
3148 	struct secashead *sah;
3149 
3150 	SAHTREE_RLOCK();
3151 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
3152 	    if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID) != 0) {
3153 		    SAH_ADDREF(sah);
3154 		    break;
3155 	    }
3156 	}
3157 	SAHTREE_RUNLOCK();
3158 	return (sah);
3159 }
3160 
3161 /*
3162  * Check not to be duplicated SPI.
3163  * OUT:
3164  *	0	: not found
3165  *	1	: found SA with given SPI.
3166  */
3167 static int
3168 key_checkspidup(uint32_t spi)
3169 {
3170 	SAHTREE_RLOCK_TRACKER;
3171 	struct secasvar *sav;
3172 
3173 	/* Assume SPI is in network byte order */
3174 	SAHTREE_RLOCK();
3175 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3176 		if (sav->spi == spi)
3177 			break;
3178 	}
3179 	SAHTREE_RUNLOCK();
3180 	return (sav != NULL);
3181 }
3182 
3183 /*
3184  * Search SA by SPI.
3185  * OUT:
3186  *	NULL	: not found
3187  *	others	: found, referenced pointer to a SA.
3188  */
3189 static struct secasvar *
3190 key_getsavbyspi(uint32_t spi)
3191 {
3192 	SAHTREE_RLOCK_TRACKER;
3193 	struct secasvar *sav;
3194 
3195 	/* Assume SPI is in network byte order */
3196 	SAHTREE_RLOCK();
3197 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3198 		if (sav->spi != spi)
3199 			continue;
3200 		SAV_ADDREF(sav);
3201 		break;
3202 	}
3203 	SAHTREE_RUNLOCK();
3204 	return (sav);
3205 }
3206 
3207 static int
3208 key_updatelifetimes(struct secasvar *sav, const struct sadb_msghdr *mhp)
3209 {
3210 	struct seclifetime *lft_h, *lft_s, *tmp;
3211 
3212 	/* Lifetime extension is optional, check that it is present. */
3213 	if (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3214 	    SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) {
3215 		/*
3216 		 * In case of SADB_UPDATE we may need to change
3217 		 * existing lifetimes.
3218 		 */
3219 		if (sav->state == SADB_SASTATE_MATURE) {
3220 			lft_h = lft_s = NULL;
3221 			goto reset;
3222 		}
3223 		return (0);
3224 	}
3225 	/* Both HARD and SOFT extensions must present */
3226 	if ((SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3227 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
3228 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
3229 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
3230 		ipseclog((LOG_DEBUG,
3231 		    "%s: invalid message: missing required header.\n",
3232 		    __func__));
3233 		return (EINVAL);
3234 	}
3235 	if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD) ||
3236 	    SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_SOFT)) {
3237 		ipseclog((LOG_DEBUG,
3238 		    "%s: invalid message: wrong header size.\n", __func__));
3239 		return (EINVAL);
3240 	}
3241 	lft_h = key_dup_lifemsg((const struct sadb_lifetime *)
3242 	    mhp->ext[SADB_EXT_LIFETIME_HARD], M_IPSEC_MISC);
3243 	if (lft_h == NULL) {
3244 		PFKEYSTAT_INC(in_nomem);
3245 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3246 		return (ENOBUFS);
3247 	}
3248 	lft_s = key_dup_lifemsg((const struct sadb_lifetime *)
3249 	    mhp->ext[SADB_EXT_LIFETIME_SOFT], M_IPSEC_MISC);
3250 	if (lft_s == NULL) {
3251 		PFKEYSTAT_INC(in_nomem);
3252 		free(lft_h, M_IPSEC_MISC);
3253 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3254 		return (ENOBUFS);
3255 	}
3256 reset:
3257 	if (sav->state != SADB_SASTATE_LARVAL) {
3258 		/*
3259 		 * key_update() holds reference to this SA,
3260 		 * so it won't be deleted in meanwhile.
3261 		 */
3262 		SECASVAR_LOCK(sav);
3263 		tmp = sav->lft_h;
3264 		sav->lft_h = lft_h;
3265 		lft_h = tmp;
3266 
3267 		tmp = sav->lft_s;
3268 		sav->lft_s = lft_s;
3269 		lft_s = tmp;
3270 		SECASVAR_UNLOCK(sav);
3271 		if (lft_h != NULL)
3272 			free(lft_h, M_IPSEC_MISC);
3273 		if (lft_s != NULL)
3274 			free(lft_s, M_IPSEC_MISC);
3275 		return (0);
3276 	}
3277 	/* We can update lifetime without holding a lock */
3278 	IPSEC_ASSERT(sav->lft_h == NULL, ("lft_h is already initialized\n"));
3279 	IPSEC_ASSERT(sav->lft_s == NULL, ("lft_s is already initialized\n"));
3280 	sav->lft_h = lft_h;
3281 	sav->lft_s = lft_s;
3282 	return (0);
3283 }
3284 
3285 /*
3286  * copy SA values from PF_KEY message except *SPI, SEQ, PID and TYPE*.
3287  * You must update these if need. Expects only LARVAL SAs.
3288  * OUT:	0:	success.
3289  *	!0:	failure.
3290  */
3291 static int
3292 key_setsaval(struct secasvar *sav, const struct sadb_msghdr *mhp)
3293 {
3294 	const struct sadb_sa *sa0;
3295 	const struct sadb_key *key0;
3296 	uint32_t replay;
3297 	size_t len;
3298 	int error;
3299 
3300 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
3301 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
3302 	IPSEC_ASSERT(sav->state == SADB_SASTATE_LARVAL,
3303 	    ("Attempt to update non LARVAL SA"));
3304 
3305 	/* XXX rewrite */
3306 	error = key_setident(sav->sah, mhp);
3307 	if (error != 0)
3308 		goto fail;
3309 
3310 	/* SA */
3311 	if (!SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
3312 		if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
3313 			error = EINVAL;
3314 			goto fail;
3315 		}
3316 		sa0 = (const struct sadb_sa *)mhp->ext[SADB_EXT_SA];
3317 		sav->alg_auth = sa0->sadb_sa_auth;
3318 		sav->alg_enc = sa0->sadb_sa_encrypt;
3319 		sav->flags = sa0->sadb_sa_flags;
3320 		if ((sav->flags & SADB_KEY_FLAGS_MAX) != sav->flags) {
3321 			ipseclog((LOG_DEBUG,
3322 			    "%s: invalid sa_flags 0x%08x.\n", __func__,
3323 			    sav->flags));
3324 			error = EINVAL;
3325 			goto fail;
3326 		}
3327 
3328 		/* Optional replay window */
3329 		replay = 0;
3330 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0)
3331 			replay = sa0->sadb_sa_replay;
3332 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_SA_REPLAY)) {
3333 			if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA_REPLAY)) {
3334 				error = EINVAL;
3335 				goto fail;
3336 			}
3337 			replay = ((const struct sadb_x_sa_replay *)
3338 			    mhp->ext[SADB_X_EXT_SA_REPLAY])->sadb_x_sa_replay_replay;
3339 
3340 			if (replay > UINT32_MAX - 32) {
3341 				ipseclog((LOG_DEBUG,
3342 				    "%s: replay window too big.\n", __func__));
3343 				error = EINVAL;
3344 				goto fail;
3345 			}
3346 
3347 			replay = (replay + 7) >> 3;
3348 		}
3349 
3350 		sav->replay = malloc(sizeof(struct secreplay), M_IPSEC_MISC,
3351 		    M_NOWAIT | M_ZERO);
3352 		if (sav->replay == NULL) {
3353 			PFKEYSTAT_INC(in_nomem);
3354 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3355 			    __func__));
3356 			error = ENOBUFS;
3357 			goto fail;
3358 		}
3359 
3360 		if (replay != 0) {
3361 			/* number of 32b blocks to be allocated */
3362 			uint32_t bitmap_size;
3363 
3364 			/* RFC 6479:
3365 			 * - the allocated replay window size must be
3366 			 *   a power of two.
3367 			 * - use an extra 32b block as a redundant window.
3368 			 */
3369 			bitmap_size = 1;
3370 			while (replay + 4 > bitmap_size)
3371 				bitmap_size <<= 1;
3372 			bitmap_size = bitmap_size / 4;
3373 
3374 			sav->replay->bitmap = malloc(
3375 			    bitmap_size * sizeof(uint32_t), M_IPSEC_MISC,
3376 			    M_NOWAIT | M_ZERO);
3377 			if (sav->replay->bitmap == NULL) {
3378 				PFKEYSTAT_INC(in_nomem);
3379 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3380 					__func__));
3381 				error = ENOBUFS;
3382 				goto fail;
3383 			}
3384 			sav->replay->bitmap_size = bitmap_size;
3385 			sav->replay->wsize = replay;
3386 		}
3387 	}
3388 
3389 	/* Authentication keys */
3390 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
3391 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH)) {
3392 			error = EINVAL;
3393 			goto fail;
3394 		}
3395 		error = 0;
3396 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_AUTH];
3397 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3398 		switch (mhp->msg->sadb_msg_satype) {
3399 		case SADB_SATYPE_AH:
3400 		case SADB_SATYPE_ESP:
3401 		case SADB_X_SATYPE_TCPSIGNATURE:
3402 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3403 			    sav->alg_auth != SADB_X_AALG_NULL)
3404 				error = EINVAL;
3405 			break;
3406 		case SADB_X_SATYPE_IPCOMP:
3407 		default:
3408 			error = EINVAL;
3409 			break;
3410 		}
3411 		if (error) {
3412 			ipseclog((LOG_DEBUG, "%s: invalid key_auth values.\n",
3413 				__func__));
3414 			goto fail;
3415 		}
3416 
3417 		sav->key_auth = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3418 		if (sav->key_auth == NULL ) {
3419 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3420 				  __func__));
3421 			PFKEYSTAT_INC(in_nomem);
3422 			error = ENOBUFS;
3423 			goto fail;
3424 		}
3425 	}
3426 
3427 	/* Encryption key */
3428 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) {
3429 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT)) {
3430 			error = EINVAL;
3431 			goto fail;
3432 		}
3433 		error = 0;
3434 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_ENCRYPT];
3435 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3436 		switch (mhp->msg->sadb_msg_satype) {
3437 		case SADB_SATYPE_ESP:
3438 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3439 			    sav->alg_enc != SADB_EALG_NULL) {
3440 				error = EINVAL;
3441 				break;
3442 			}
3443 			sav->key_enc = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3444 			if (sav->key_enc == NULL) {
3445 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3446 					__func__));
3447 				PFKEYSTAT_INC(in_nomem);
3448 				error = ENOBUFS;
3449 				goto fail;
3450 			}
3451 			break;
3452 		case SADB_X_SATYPE_IPCOMP:
3453 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3454 				error = EINVAL;
3455 			sav->key_enc = NULL;	/*just in case*/
3456 			break;
3457 		case SADB_SATYPE_AH:
3458 		case SADB_X_SATYPE_TCPSIGNATURE:
3459 		default:
3460 			error = EINVAL;
3461 			break;
3462 		}
3463 		if (error) {
3464 			ipseclog((LOG_DEBUG, "%s: invalid key_enc value.\n",
3465 				__func__));
3466 			goto fail;
3467 		}
3468 	}
3469 
3470 	/* set iv */
3471 	sav->ivlen = 0;
3472 	switch (mhp->msg->sadb_msg_satype) {
3473 	case SADB_SATYPE_AH:
3474 		if (sav->flags & SADB_X_EXT_DERIV) {
3475 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3476 			    "given to AH SA.\n", __func__));
3477 			error = EINVAL;
3478 			goto fail;
3479 		}
3480 		if (sav->alg_enc != SADB_EALG_NONE) {
3481 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3482 			    "mismated.\n", __func__));
3483 			error = EINVAL;
3484 			goto fail;
3485 		}
3486 		error = xform_init(sav, XF_AH);
3487 		break;
3488 	case SADB_SATYPE_ESP:
3489 		if ((sav->flags & (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) ==
3490 		    (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) {
3491 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3492 			    "given to old-esp.\n", __func__));
3493 			error = EINVAL;
3494 			goto fail;
3495 		}
3496 		error = xform_init(sav, XF_ESP);
3497 		break;
3498 	case SADB_X_SATYPE_IPCOMP:
3499 		if (sav->alg_auth != SADB_AALG_NONE) {
3500 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3501 			    "mismated.\n", __func__));
3502 			error = EINVAL;
3503 			goto fail;
3504 		}
3505 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0 &&
3506 		    ntohl(sav->spi) >= 0x10000) {
3507 			ipseclog((LOG_DEBUG, "%s: invalid cpi for IPComp.\n",
3508 			    __func__));
3509 			error = EINVAL;
3510 			goto fail;
3511 		}
3512 		error = xform_init(sav, XF_IPCOMP);
3513 		break;
3514 	case SADB_X_SATYPE_TCPSIGNATURE:
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_TCPSIGNATURE);
3522 		break;
3523 	default:
3524 		ipseclog((LOG_DEBUG, "%s: Invalid satype.\n", __func__));
3525 		error = EPROTONOSUPPORT;
3526 		goto fail;
3527 	}
3528 	if (error) {
3529 		ipseclog((LOG_DEBUG, "%s: unable to initialize SA type %u.\n",
3530 		    __func__, mhp->msg->sadb_msg_satype));
3531 		goto fail;
3532 	}
3533 
3534 	/* Handle NAT-T headers */
3535 	error = key_setnatt(sav, mhp);
3536 	if (error != 0)
3537 		goto fail;
3538 
3539 	/* Initialize lifetime for CURRENT */
3540 	sav->firstused = 0;
3541 	sav->created = time_second;
3542 
3543 	/* lifetimes for HARD and SOFT */
3544 	error = key_updatelifetimes(sav, mhp);
3545 	if (error == 0)
3546 		return (0);
3547 fail:
3548 	key_cleansav(sav);
3549 	return (error);
3550 }
3551 
3552 /*
3553  * subroutine for SADB_GET and SADB_DUMP.
3554  */
3555 static struct mbuf *
3556 key_setdumpsa(struct secasvar *sav, uint8_t type, uint8_t satype,
3557     uint32_t seq, uint32_t pid)
3558 {
3559 	struct seclifetime lft_c;
3560 	struct mbuf *result = NULL, *tres = NULL, *m;
3561 	int i, dumporder[] = {
3562 		SADB_EXT_SA, SADB_X_EXT_SA2, SADB_X_EXT_SA_REPLAY,
3563 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3564 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3565 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY,
3566 		SADB_EXT_KEY_AUTH, SADB_EXT_KEY_ENCRYPT,
3567 		SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
3568 		SADB_EXT_SENSITIVITY,
3569 		SADB_X_EXT_NAT_T_TYPE,
3570 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3571 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3572 		SADB_X_EXT_NAT_T_FRAG,
3573 	};
3574 	uint32_t replay_count;
3575 
3576 	m = key_setsadbmsg(type, 0, satype, seq, pid, sav->refcnt);
3577 	if (m == NULL)
3578 		goto fail;
3579 	result = m;
3580 
3581 	for (i = nitems(dumporder) - 1; i >= 0; i--) {
3582 		m = NULL;
3583 		switch (dumporder[i]) {
3584 		case SADB_EXT_SA:
3585 			m = key_setsadbsa(sav);
3586 			if (!m)
3587 				goto fail;
3588 			break;
3589 
3590 		case SADB_X_EXT_SA2:
3591 			SECASVAR_LOCK(sav);
3592 			replay_count = sav->replay ? sav->replay->count : 0;
3593 			SECASVAR_UNLOCK(sav);
3594 			m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
3595 					sav->sah->saidx.reqid);
3596 			if (!m)
3597 				goto fail;
3598 			break;
3599 
3600 		case SADB_X_EXT_SA_REPLAY:
3601 			if (sav->replay == NULL ||
3602 			    sav->replay->wsize <= UINT8_MAX)
3603 				continue;
3604 
3605 			m = key_setsadbxsareplay(sav->replay->wsize);
3606 			if (!m)
3607 				goto fail;
3608 			break;
3609 
3610 		case SADB_EXT_ADDRESS_SRC:
3611 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3612 			    &sav->sah->saidx.src.sa,
3613 			    FULLMASK, IPSEC_ULPROTO_ANY);
3614 			if (!m)
3615 				goto fail;
3616 			break;
3617 
3618 		case SADB_EXT_ADDRESS_DST:
3619 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3620 			    &sav->sah->saidx.dst.sa,
3621 			    FULLMASK, IPSEC_ULPROTO_ANY);
3622 			if (!m)
3623 				goto fail;
3624 			break;
3625 
3626 		case SADB_EXT_KEY_AUTH:
3627 			if (!sav->key_auth)
3628 				continue;
3629 			m = key_setkey(sav->key_auth, SADB_EXT_KEY_AUTH);
3630 			if (!m)
3631 				goto fail;
3632 			break;
3633 
3634 		case SADB_EXT_KEY_ENCRYPT:
3635 			if (!sav->key_enc)
3636 				continue;
3637 			m = key_setkey(sav->key_enc, SADB_EXT_KEY_ENCRYPT);
3638 			if (!m)
3639 				goto fail;
3640 			break;
3641 
3642 		case SADB_EXT_LIFETIME_CURRENT:
3643 			lft_c.addtime = sav->created;
3644 			lft_c.allocations = (uint32_t)counter_u64_fetch(
3645 			    sav->lft_c_allocations);
3646 			lft_c.bytes = counter_u64_fetch(sav->lft_c_bytes);
3647 			lft_c.usetime = sav->firstused;
3648 			m = key_setlifetime(&lft_c, SADB_EXT_LIFETIME_CURRENT);
3649 			if (!m)
3650 				goto fail;
3651 			break;
3652 
3653 		case SADB_EXT_LIFETIME_HARD:
3654 			if (!sav->lft_h)
3655 				continue;
3656 			m = key_setlifetime(sav->lft_h,
3657 					    SADB_EXT_LIFETIME_HARD);
3658 			if (!m)
3659 				goto fail;
3660 			break;
3661 
3662 		case SADB_EXT_LIFETIME_SOFT:
3663 			if (!sav->lft_s)
3664 				continue;
3665 			m = key_setlifetime(sav->lft_s,
3666 					    SADB_EXT_LIFETIME_SOFT);
3667 
3668 			if (!m)
3669 				goto fail;
3670 			break;
3671 
3672 		case SADB_X_EXT_NAT_T_TYPE:
3673 			if (sav->natt == NULL)
3674 				continue;
3675 			m = key_setsadbxtype(UDP_ENCAP_ESPINUDP);
3676 			if (!m)
3677 				goto fail;
3678 			break;
3679 
3680 		case SADB_X_EXT_NAT_T_DPORT:
3681 			if (sav->natt == NULL)
3682 				continue;
3683 			m = key_setsadbxport(sav->natt->dport,
3684 			    SADB_X_EXT_NAT_T_DPORT);
3685 			if (!m)
3686 				goto fail;
3687 			break;
3688 
3689 		case SADB_X_EXT_NAT_T_SPORT:
3690 			if (sav->natt == NULL)
3691 				continue;
3692 			m = key_setsadbxport(sav->natt->sport,
3693 			    SADB_X_EXT_NAT_T_SPORT);
3694 			if (!m)
3695 				goto fail;
3696 			break;
3697 
3698 		case SADB_X_EXT_NAT_T_OAI:
3699 			if (sav->natt == NULL ||
3700 			    (sav->natt->flags & IPSEC_NATT_F_OAI) == 0)
3701 				continue;
3702 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAI,
3703 			    &sav->natt->oai.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3704 			if (!m)
3705 				goto fail;
3706 			break;
3707 		case SADB_X_EXT_NAT_T_OAR:
3708 			if (sav->natt == NULL ||
3709 			    (sav->natt->flags & IPSEC_NATT_F_OAR) == 0)
3710 				continue;
3711 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAR,
3712 			    &sav->natt->oar.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3713 			if (!m)
3714 				goto fail;
3715 			break;
3716 		case SADB_X_EXT_NAT_T_FRAG:
3717 			/* We do not (yet) support those. */
3718 			continue;
3719 
3720 		case SADB_EXT_ADDRESS_PROXY:
3721 		case SADB_EXT_IDENTITY_SRC:
3722 		case SADB_EXT_IDENTITY_DST:
3723 			/* XXX: should we brought from SPD ? */
3724 		case SADB_EXT_SENSITIVITY:
3725 		default:
3726 			continue;
3727 		}
3728 
3729 		if (!m)
3730 			goto fail;
3731 		if (tres)
3732 			m_cat(m, tres);
3733 		tres = m;
3734 	}
3735 
3736 	m_cat(result, tres);
3737 	tres = NULL;
3738 	if (result->m_len < sizeof(struct sadb_msg)) {
3739 		result = m_pullup(result, sizeof(struct sadb_msg));
3740 		if (result == NULL)
3741 			goto fail;
3742 	}
3743 
3744 	result->m_pkthdr.len = 0;
3745 	for (m = result; m; m = m->m_next)
3746 		result->m_pkthdr.len += m->m_len;
3747 
3748 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3749 	    PFKEY_UNIT64(result->m_pkthdr.len);
3750 
3751 	return result;
3752 
3753 fail:
3754 	m_freem(result);
3755 	m_freem(tres);
3756 	return NULL;
3757 }
3758 
3759 /*
3760  * set data into sadb_msg.
3761  */
3762 static struct mbuf *
3763 key_setsadbmsg(u_int8_t type, u_int16_t tlen, u_int8_t satype, u_int32_t seq,
3764     pid_t pid, u_int16_t reserved)
3765 {
3766 	struct mbuf *m;
3767 	struct sadb_msg *p;
3768 	int len;
3769 
3770 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
3771 	if (len > MCLBYTES)
3772 		return NULL;
3773 	MGETHDR(m, M_NOWAIT, MT_DATA);
3774 	if (m && len > MHLEN) {
3775 		if (!(MCLGET(m, M_NOWAIT))) {
3776 			m_freem(m);
3777 			m = NULL;
3778 		}
3779 	}
3780 	if (!m)
3781 		return NULL;
3782 	m->m_pkthdr.len = m->m_len = len;
3783 	m->m_next = NULL;
3784 
3785 	p = mtod(m, struct sadb_msg *);
3786 
3787 	bzero(p, len);
3788 	p->sadb_msg_version = PF_KEY_V2;
3789 	p->sadb_msg_type = type;
3790 	p->sadb_msg_errno = 0;
3791 	p->sadb_msg_satype = satype;
3792 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
3793 	p->sadb_msg_reserved = reserved;
3794 	p->sadb_msg_seq = seq;
3795 	p->sadb_msg_pid = (u_int32_t)pid;
3796 
3797 	return m;
3798 }
3799 
3800 /*
3801  * copy secasvar data into sadb_address.
3802  */
3803 static struct mbuf *
3804 key_setsadbsa(struct secasvar *sav)
3805 {
3806 	struct mbuf *m;
3807 	struct sadb_sa *p;
3808 	int len;
3809 
3810 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
3811 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3812 	if (m == NULL)
3813 		return (NULL);
3814 	m_align(m, len);
3815 	m->m_len = len;
3816 	p = mtod(m, struct sadb_sa *);
3817 	bzero(p, len);
3818 	p->sadb_sa_len = PFKEY_UNIT64(len);
3819 	p->sadb_sa_exttype = SADB_EXT_SA;
3820 	p->sadb_sa_spi = sav->spi;
3821 	p->sadb_sa_replay = sav->replay ?
3822 	    (sav->replay->wsize > UINT8_MAX ? UINT8_MAX :
3823 		sav->replay->wsize): 0;
3824 	p->sadb_sa_state = sav->state;
3825 	p->sadb_sa_auth = sav->alg_auth;
3826 	p->sadb_sa_encrypt = sav->alg_enc;
3827 	p->sadb_sa_flags = sav->flags & SADB_KEY_FLAGS_MAX;
3828 	return (m);
3829 }
3830 
3831 /*
3832  * set data into sadb_address.
3833  */
3834 static struct mbuf *
3835 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
3836     u_int8_t prefixlen, u_int16_t ul_proto)
3837 {
3838 	struct mbuf *m;
3839 	struct sadb_address *p;
3840 	size_t len;
3841 
3842 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
3843 	    PFKEY_ALIGN8(saddr->sa_len);
3844 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3845 	if (m == NULL)
3846 		return (NULL);
3847 	m_align(m, len);
3848 	m->m_len = len;
3849 	p = mtod(m, struct sadb_address *);
3850 
3851 	bzero(p, len);
3852 	p->sadb_address_len = PFKEY_UNIT64(len);
3853 	p->sadb_address_exttype = exttype;
3854 	p->sadb_address_proto = ul_proto;
3855 	if (prefixlen == FULLMASK) {
3856 		switch (saddr->sa_family) {
3857 		case AF_INET:
3858 			prefixlen = sizeof(struct in_addr) << 3;
3859 			break;
3860 		case AF_INET6:
3861 			prefixlen = sizeof(struct in6_addr) << 3;
3862 			break;
3863 		default:
3864 			; /*XXX*/
3865 		}
3866 	}
3867 	p->sadb_address_prefixlen = prefixlen;
3868 	p->sadb_address_reserved = 0;
3869 
3870 	bcopy(saddr,
3871 	    mtod(m, caddr_t) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
3872 	    saddr->sa_len);
3873 
3874 	return m;
3875 }
3876 
3877 /*
3878  * set data into sadb_x_sa2.
3879  */
3880 static struct mbuf *
3881 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int32_t reqid)
3882 {
3883 	struct mbuf *m;
3884 	struct sadb_x_sa2 *p;
3885 	size_t len;
3886 
3887 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
3888 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3889 	if (m == NULL)
3890 		return (NULL);
3891 	m_align(m, len);
3892 	m->m_len = len;
3893 	p = mtod(m, struct sadb_x_sa2 *);
3894 
3895 	bzero(p, len);
3896 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
3897 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
3898 	p->sadb_x_sa2_mode = mode;
3899 	p->sadb_x_sa2_reserved1 = 0;
3900 	p->sadb_x_sa2_reserved2 = 0;
3901 	p->sadb_x_sa2_sequence = seq;
3902 	p->sadb_x_sa2_reqid = reqid;
3903 
3904 	return m;
3905 }
3906 
3907 /*
3908  * Set data into sadb_x_sa_replay.
3909  */
3910 static struct mbuf *
3911 key_setsadbxsareplay(u_int32_t replay)
3912 {
3913 	struct mbuf *m;
3914 	struct sadb_x_sa_replay *p;
3915 	size_t len;
3916 
3917 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa_replay));
3918 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3919 	if (m == NULL)
3920 		return (NULL);
3921 	m_align(m, len);
3922 	m->m_len = len;
3923 	p = mtod(m, struct sadb_x_sa_replay *);
3924 
3925 	bzero(p, len);
3926 	p->sadb_x_sa_replay_len = PFKEY_UNIT64(len);
3927 	p->sadb_x_sa_replay_exttype = SADB_X_EXT_SA_REPLAY;
3928 	p->sadb_x_sa_replay_replay = (replay << 3);
3929 
3930 	return m;
3931 }
3932 
3933 /*
3934  * Set a type in sadb_x_nat_t_type.
3935  */
3936 static struct mbuf *
3937 key_setsadbxtype(u_int16_t type)
3938 {
3939 	struct mbuf *m;
3940 	size_t len;
3941 	struct sadb_x_nat_t_type *p;
3942 
3943 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
3944 
3945 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3946 	if (m == NULL)
3947 		return (NULL);
3948 	m_align(m, len);
3949 	m->m_len = len;
3950 	p = mtod(m, struct sadb_x_nat_t_type *);
3951 
3952 	bzero(p, len);
3953 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
3954 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
3955 	p->sadb_x_nat_t_type_type = type;
3956 
3957 	return (m);
3958 }
3959 /*
3960  * Set a port in sadb_x_nat_t_port.
3961  * In contrast to default RFC 2367 behaviour, port is in network byte order.
3962  */
3963 static struct mbuf *
3964 key_setsadbxport(u_int16_t port, u_int16_t type)
3965 {
3966 	struct mbuf *m;
3967 	size_t len;
3968 	struct sadb_x_nat_t_port *p;
3969 
3970 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
3971 
3972 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3973 	if (m == NULL)
3974 		return (NULL);
3975 	m_align(m, len);
3976 	m->m_len = len;
3977 	p = mtod(m, struct sadb_x_nat_t_port *);
3978 
3979 	bzero(p, len);
3980 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
3981 	p->sadb_x_nat_t_port_exttype = type;
3982 	p->sadb_x_nat_t_port_port = port;
3983 
3984 	return (m);
3985 }
3986 
3987 /*
3988  * Get port from sockaddr. Port is in network byte order.
3989  */
3990 uint16_t
3991 key_portfromsaddr(struct sockaddr *sa)
3992 {
3993 
3994 	switch (sa->sa_family) {
3995 #ifdef INET
3996 	case AF_INET:
3997 		return ((struct sockaddr_in *)sa)->sin_port;
3998 #endif
3999 #ifdef INET6
4000 	case AF_INET6:
4001 		return ((struct sockaddr_in6 *)sa)->sin6_port;
4002 #endif
4003 	}
4004 	return (0);
4005 }
4006 
4007 /*
4008  * Set port in struct sockaddr. Port is in network byte order.
4009  */
4010 void
4011 key_porttosaddr(struct sockaddr *sa, uint16_t port)
4012 {
4013 
4014 	switch (sa->sa_family) {
4015 #ifdef INET
4016 	case AF_INET:
4017 		((struct sockaddr_in *)sa)->sin_port = port;
4018 		break;
4019 #endif
4020 #ifdef INET6
4021 	case AF_INET6:
4022 		((struct sockaddr_in6 *)sa)->sin6_port = port;
4023 		break;
4024 #endif
4025 	default:
4026 		ipseclog((LOG_DEBUG, "%s: unexpected address family %d.\n",
4027 			__func__, sa->sa_family));
4028 		break;
4029 	}
4030 }
4031 
4032 /*
4033  * set data into sadb_x_policy
4034  */
4035 static struct mbuf *
4036 key_setsadbxpolicy(u_int16_t type, u_int8_t dir, u_int32_t id, u_int32_t priority)
4037 {
4038 	struct mbuf *m;
4039 	struct sadb_x_policy *p;
4040 	size_t len;
4041 
4042 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
4043 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4044 	if (m == NULL)
4045 		return (NULL);
4046 	m_align(m, len);
4047 	m->m_len = len;
4048 	p = mtod(m, struct sadb_x_policy *);
4049 
4050 	bzero(p, len);
4051 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
4052 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
4053 	p->sadb_x_policy_type = type;
4054 	p->sadb_x_policy_dir = dir;
4055 	p->sadb_x_policy_id = id;
4056 	p->sadb_x_policy_priority = priority;
4057 
4058 	return m;
4059 }
4060 
4061 /* %%% utilities */
4062 /* Take a key message (sadb_key) from the socket and turn it into one
4063  * of the kernel's key structures (seckey).
4064  *
4065  * IN: pointer to the src
4066  * OUT: NULL no more memory
4067  */
4068 struct seckey *
4069 key_dup_keymsg(const struct sadb_key *src, size_t len,
4070     struct malloc_type *type)
4071 {
4072 	struct seckey *dst;
4073 
4074 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4075 	if (dst != NULL) {
4076 		dst->bits = src->sadb_key_bits;
4077 		dst->key_data = malloc(len, type, M_NOWAIT);
4078 		if (dst->key_data != NULL) {
4079 			bcopy((const char *)(src + 1), dst->key_data, len);
4080 		} else {
4081 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4082 			    __func__));
4083 			free(dst, type);
4084 			dst = NULL;
4085 		}
4086 	} else {
4087 		ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4088 		    __func__));
4089 
4090 	}
4091 	return (dst);
4092 }
4093 
4094 /* Take a lifetime message (sadb_lifetime) passed in on a socket and
4095  * turn it into one of the kernel's lifetime structures (seclifetime).
4096  *
4097  * IN: pointer to the destination, source and malloc type
4098  * OUT: NULL, no more memory
4099  */
4100 
4101 static struct seclifetime *
4102 key_dup_lifemsg(const struct sadb_lifetime *src, struct malloc_type *type)
4103 {
4104 	struct seclifetime *dst;
4105 
4106 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4107 	if (dst == NULL) {
4108 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
4109 		return (NULL);
4110 	}
4111 	dst->allocations = src->sadb_lifetime_allocations;
4112 	dst->bytes = src->sadb_lifetime_bytes;
4113 	dst->addtime = src->sadb_lifetime_addtime;
4114 	dst->usetime = src->sadb_lifetime_usetime;
4115 	return (dst);
4116 }
4117 
4118 /*
4119  * compare two secasindex structure.
4120  * flag can specify to compare 2 saidxes.
4121  * compare two secasindex structure without both mode and reqid.
4122  * don't compare port.
4123  * IN:
4124  *      saidx0: source, it can be in SAD.
4125  *      saidx1: object.
4126  * OUT:
4127  *      1 : equal
4128  *      0 : not equal
4129  */
4130 static int
4131 key_cmpsaidx(const struct secasindex *saidx0, const struct secasindex *saidx1,
4132     int flag)
4133 {
4134 
4135 	/* sanity */
4136 	if (saidx0 == NULL && saidx1 == NULL)
4137 		return 1;
4138 
4139 	if (saidx0 == NULL || saidx1 == NULL)
4140 		return 0;
4141 
4142 	if (saidx0->proto != saidx1->proto)
4143 		return 0;
4144 
4145 	if (flag == CMP_EXACTLY) {
4146 		if (saidx0->mode != saidx1->mode)
4147 			return 0;
4148 		if (saidx0->reqid != saidx1->reqid)
4149 			return 0;
4150 		if (bcmp(&saidx0->src, &saidx1->src,
4151 		    saidx0->src.sa.sa_len) != 0 ||
4152 		    bcmp(&saidx0->dst, &saidx1->dst,
4153 		    saidx0->dst.sa.sa_len) != 0)
4154 			return 0;
4155 	} else {
4156 
4157 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
4158 		if (flag == CMP_MODE_REQID || flag == CMP_REQID) {
4159 			/*
4160 			 * If reqid of SPD is non-zero, unique SA is required.
4161 			 * The result must be of same reqid in this case.
4162 			 */
4163 			if (saidx1->reqid != 0 &&
4164 			    saidx0->reqid != saidx1->reqid)
4165 				return 0;
4166 		}
4167 
4168 		if (flag == CMP_MODE_REQID) {
4169 			if (saidx0->mode != IPSEC_MODE_ANY
4170 			 && saidx0->mode != saidx1->mode)
4171 				return 0;
4172 		}
4173 
4174 		if (key_sockaddrcmp(&saidx0->src.sa, &saidx1->src.sa, 0) != 0)
4175 			return 0;
4176 		if (key_sockaddrcmp(&saidx0->dst.sa, &saidx1->dst.sa, 0) != 0)
4177 			return 0;
4178 	}
4179 
4180 	return 1;
4181 }
4182 
4183 /*
4184  * compare two secindex structure exactly.
4185  * IN:
4186  *	spidx0: source, it is often in SPD.
4187  *	spidx1: object, it is often from PFKEY message.
4188  * OUT:
4189  *	1 : equal
4190  *	0 : not equal
4191  */
4192 static int
4193 key_cmpspidx_exactly(struct secpolicyindex *spidx0,
4194     struct secpolicyindex *spidx1)
4195 {
4196 	/* sanity */
4197 	if (spidx0 == NULL && spidx1 == NULL)
4198 		return 1;
4199 
4200 	if (spidx0 == NULL || spidx1 == NULL)
4201 		return 0;
4202 
4203 	if (spidx0->prefs != spidx1->prefs
4204 	 || spidx0->prefd != spidx1->prefd
4205 	 || spidx0->ul_proto != spidx1->ul_proto
4206 	 || spidx0->dir != spidx1->dir)
4207 		return 0;
4208 
4209 	return key_sockaddrcmp(&spidx0->src.sa, &spidx1->src.sa, 1) == 0 &&
4210 	       key_sockaddrcmp(&spidx0->dst.sa, &spidx1->dst.sa, 1) == 0;
4211 }
4212 
4213 /*
4214  * compare two secindex structure with mask.
4215  * IN:
4216  *	spidx0: source, it is often in SPD.
4217  *	spidx1: object, it is often from IP header.
4218  * OUT:
4219  *	1 : equal
4220  *	0 : not equal
4221  */
4222 static int
4223 key_cmpspidx_withmask(struct secpolicyindex *spidx0,
4224     struct secpolicyindex *spidx1)
4225 {
4226 	/* sanity */
4227 	if (spidx0 == NULL && spidx1 == NULL)
4228 		return 1;
4229 
4230 	if (spidx0 == NULL || spidx1 == NULL)
4231 		return 0;
4232 
4233 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4234 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4235 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4236 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
4237 		return 0;
4238 
4239 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4240 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY
4241 	 && spidx0->ul_proto != spidx1->ul_proto)
4242 		return 0;
4243 
4244 	switch (spidx0->src.sa.sa_family) {
4245 	case AF_INET:
4246 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY
4247 		 && spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
4248 			return 0;
4249 		if (!key_bbcmp(&spidx0->src.sin.sin_addr,
4250 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
4251 			return 0;
4252 		break;
4253 	case AF_INET6:
4254 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY
4255 		 && spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
4256 			return 0;
4257 		/*
4258 		 * scope_id check. if sin6_scope_id is 0, we regard it
4259 		 * as a wildcard scope, which matches any scope zone ID.
4260 		 */
4261 		if (spidx0->src.sin6.sin6_scope_id &&
4262 		    spidx1->src.sin6.sin6_scope_id &&
4263 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
4264 			return 0;
4265 		if (!key_bbcmp(&spidx0->src.sin6.sin6_addr,
4266 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
4267 			return 0;
4268 		break;
4269 	default:
4270 		/* XXX */
4271 		if (bcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
4272 			return 0;
4273 		break;
4274 	}
4275 
4276 	switch (spidx0->dst.sa.sa_family) {
4277 	case AF_INET:
4278 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY
4279 		 && spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
4280 			return 0;
4281 		if (!key_bbcmp(&spidx0->dst.sin.sin_addr,
4282 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
4283 			return 0;
4284 		break;
4285 	case AF_INET6:
4286 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY
4287 		 && spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
4288 			return 0;
4289 		/*
4290 		 * scope_id check. if sin6_scope_id is 0, we regard it
4291 		 * as a wildcard scope, which matches any scope zone ID.
4292 		 */
4293 		if (spidx0->dst.sin6.sin6_scope_id &&
4294 		    spidx1->dst.sin6.sin6_scope_id &&
4295 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
4296 			return 0;
4297 		if (!key_bbcmp(&spidx0->dst.sin6.sin6_addr,
4298 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
4299 			return 0;
4300 		break;
4301 	default:
4302 		/* XXX */
4303 		if (bcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
4304 			return 0;
4305 		break;
4306 	}
4307 
4308 	/* XXX Do we check other field ?  e.g. flowinfo */
4309 
4310 	return 1;
4311 }
4312 
4313 #ifdef satosin
4314 #undef satosin
4315 #endif
4316 #define satosin(s) ((const struct sockaddr_in *)s)
4317 #ifdef satosin6
4318 #undef satosin6
4319 #endif
4320 #define satosin6(s) ((const struct sockaddr_in6 *)s)
4321 /* returns 0 on match */
4322 int
4323 key_sockaddrcmp(const struct sockaddr *sa1, const struct sockaddr *sa2,
4324     int port)
4325 {
4326 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4327 		return 1;
4328 
4329 	switch (sa1->sa_family) {
4330 #ifdef INET
4331 	case AF_INET:
4332 		if (sa1->sa_len != sizeof(struct sockaddr_in))
4333 			return 1;
4334 		if (satosin(sa1)->sin_addr.s_addr !=
4335 		    satosin(sa2)->sin_addr.s_addr) {
4336 			return 1;
4337 		}
4338 		if (port && satosin(sa1)->sin_port != satosin(sa2)->sin_port)
4339 			return 1;
4340 		break;
4341 #endif
4342 #ifdef INET6
4343 	case AF_INET6:
4344 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4345 			return 1;	/*EINVAL*/
4346 		if (satosin6(sa1)->sin6_scope_id !=
4347 		    satosin6(sa2)->sin6_scope_id) {
4348 			return 1;
4349 		}
4350 		if (!IN6_ARE_ADDR_EQUAL(&satosin6(sa1)->sin6_addr,
4351 		    &satosin6(sa2)->sin6_addr)) {
4352 			return 1;
4353 		}
4354 		if (port &&
4355 		    satosin6(sa1)->sin6_port != satosin6(sa2)->sin6_port) {
4356 			return 1;
4357 		}
4358 		break;
4359 #endif
4360 	default:
4361 		if (bcmp(sa1, sa2, sa1->sa_len) != 0)
4362 			return 1;
4363 		break;
4364 	}
4365 
4366 	return 0;
4367 }
4368 
4369 /* returns 0 on match */
4370 int
4371 key_sockaddrcmp_withmask(const struct sockaddr *sa1,
4372     const struct sockaddr *sa2, size_t mask)
4373 {
4374 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4375 		return (1);
4376 
4377 	switch (sa1->sa_family) {
4378 #ifdef INET
4379 	case AF_INET:
4380 		return (!key_bbcmp(&satosin(sa1)->sin_addr,
4381 		    &satosin(sa2)->sin_addr, mask));
4382 #endif
4383 #ifdef INET6
4384 	case AF_INET6:
4385 		if (satosin6(sa1)->sin6_scope_id !=
4386 		    satosin6(sa2)->sin6_scope_id)
4387 			return (1);
4388 		return (!key_bbcmp(&satosin6(sa1)->sin6_addr,
4389 		    &satosin6(sa2)->sin6_addr, mask));
4390 #endif
4391 	}
4392 	return (1);
4393 }
4394 #undef satosin
4395 #undef satosin6
4396 
4397 /*
4398  * compare two buffers with mask.
4399  * IN:
4400  *	addr1: source
4401  *	addr2: object
4402  *	bits:  Number of bits to compare
4403  * OUT:
4404  *	1 : equal
4405  *	0 : not equal
4406  */
4407 static int
4408 key_bbcmp(const void *a1, const void *a2, u_int bits)
4409 {
4410 	const unsigned char *p1 = a1;
4411 	const unsigned char *p2 = a2;
4412 
4413 	/* XXX: This could be considerably faster if we compare a word
4414 	 * at a time, but it is complicated on LSB Endian machines */
4415 
4416 	/* Handle null pointers */
4417 	if (p1 == NULL || p2 == NULL)
4418 		return (p1 == p2);
4419 
4420 	while (bits >= 8) {
4421 		if (*p1++ != *p2++)
4422 			return 0;
4423 		bits -= 8;
4424 	}
4425 
4426 	if (bits > 0) {
4427 		u_int8_t mask = ~((1<<(8-bits))-1);
4428 		if ((*p1 & mask) != (*p2 & mask))
4429 			return 0;
4430 	}
4431 	return 1;	/* Match! */
4432 }
4433 
4434 static void
4435 key_flush_spd(time_t now)
4436 {
4437 	SPTREE_RLOCK_TRACKER;
4438 	struct secpolicy_list drainq;
4439 	struct secpolicy *sp, *nextsp;
4440 	u_int dir;
4441 
4442 	LIST_INIT(&drainq);
4443 	SPTREE_RLOCK();
4444 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4445 		TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
4446 			if (sp->lifetime == 0 && sp->validtime == 0)
4447 				continue;
4448 			if ((sp->lifetime &&
4449 			    now - sp->created > sp->lifetime) ||
4450 			    (sp->validtime &&
4451 			    now - sp->lastused > sp->validtime)) {
4452 				/* Hold extra reference to send SPDEXPIRE */
4453 				SP_ADDREF(sp);
4454 				LIST_INSERT_HEAD(&drainq, sp, drainq);
4455 			}
4456 		}
4457 	}
4458 	SPTREE_RUNLOCK();
4459 	if (LIST_EMPTY(&drainq))
4460 		return;
4461 
4462 	SPTREE_WLOCK();
4463 	sp = LIST_FIRST(&drainq);
4464 	while (sp != NULL) {
4465 		nextsp = LIST_NEXT(sp, drainq);
4466 		/* Check that SP is still linked */
4467 		if (sp->state != IPSEC_SPSTATE_ALIVE) {
4468 			LIST_REMOVE(sp, drainq);
4469 			key_freesp(&sp); /* release extra reference */
4470 			sp = nextsp;
4471 			continue;
4472 		}
4473 		TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
4474 		V_spd_size--;
4475 		LIST_REMOVE(sp, idhash);
4476 		sp->state = IPSEC_SPSTATE_DEAD;
4477 		sp = nextsp;
4478 	}
4479 	V_sp_genid++;
4480 	SPTREE_WUNLOCK();
4481 	if (SPDCACHE_ENABLED())
4482 		spdcache_clear();
4483 
4484 	sp = LIST_FIRST(&drainq);
4485 	while (sp != NULL) {
4486 		nextsp = LIST_NEXT(sp, drainq);
4487 		key_spdexpire(sp);
4488 		key_freesp(&sp); /* release extra reference */
4489 		key_freesp(&sp); /* release last reference */
4490 		sp = nextsp;
4491 	}
4492 }
4493 
4494 static void
4495 key_flush_sad(time_t now)
4496 {
4497 	SAHTREE_RLOCK_TRACKER;
4498 	struct secashead_list emptyq;
4499 	struct secasvar_list drainq, hexpireq, sexpireq, freeq;
4500 	struct secashead *sah, *nextsah;
4501 	struct secasvar *sav, *nextsav;
4502 
4503 	LIST_INIT(&drainq);
4504 	LIST_INIT(&hexpireq);
4505 	LIST_INIT(&sexpireq);
4506 	LIST_INIT(&emptyq);
4507 
4508 	SAHTREE_RLOCK();
4509 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
4510 		/* Check for empty SAH */
4511 		if (TAILQ_EMPTY(&sah->savtree_larval) &&
4512 		    TAILQ_EMPTY(&sah->savtree_alive)) {
4513 			SAH_ADDREF(sah);
4514 			LIST_INSERT_HEAD(&emptyq, sah, drainq);
4515 			continue;
4516 		}
4517 		/* Add all stale LARVAL SAs into drainq */
4518 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
4519 			if (now - sav->created < V_key_larval_lifetime)
4520 				continue;
4521 			SAV_ADDREF(sav);
4522 			LIST_INSERT_HEAD(&drainq, sav, drainq);
4523 		}
4524 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
4525 			/* lifetimes aren't specified */
4526 			if (sav->lft_h == NULL)
4527 				continue;
4528 			SECASVAR_LOCK(sav);
4529 			/*
4530 			 * Check again with lock held, because it may
4531 			 * be updated by SADB_UPDATE.
4532 			 */
4533 			if (sav->lft_h == NULL) {
4534 				SECASVAR_UNLOCK(sav);
4535 				continue;
4536 			}
4537 			/*
4538 			 * RFC 2367:
4539 			 * HARD lifetimes MUST take precedence over SOFT
4540 			 * lifetimes, meaning if the HARD and SOFT lifetimes
4541 			 * are the same, the HARD lifetime will appear on the
4542 			 * EXPIRE message.
4543 			 */
4544 			/* check HARD lifetime */
4545 			if ((sav->lft_h->addtime != 0 &&
4546 			    now - sav->created > sav->lft_h->addtime) ||
4547 			    (sav->lft_h->usetime != 0 && sav->firstused &&
4548 			    now - sav->firstused > sav->lft_h->usetime) ||
4549 			    (sav->lft_h->bytes != 0 && counter_u64_fetch(
4550 			        sav->lft_c_bytes) > sav->lft_h->bytes)) {
4551 				SECASVAR_UNLOCK(sav);
4552 				SAV_ADDREF(sav);
4553 				LIST_INSERT_HEAD(&hexpireq, sav, drainq);
4554 				continue;
4555 			}
4556 			/* check SOFT lifetime (only for MATURE SAs) */
4557 			if (sav->state == SADB_SASTATE_MATURE && (
4558 			    (sav->lft_s->addtime != 0 &&
4559 			    now - sav->created > sav->lft_s->addtime) ||
4560 			    (sav->lft_s->usetime != 0 && sav->firstused &&
4561 			    now - sav->firstused > sav->lft_s->usetime) ||
4562 			    (sav->lft_s->bytes != 0 && counter_u64_fetch(
4563 				sav->lft_c_bytes) > sav->lft_s->bytes))) {
4564 				SECASVAR_UNLOCK(sav);
4565 				SAV_ADDREF(sav);
4566 				LIST_INSERT_HEAD(&sexpireq, sav, drainq);
4567 				continue;
4568 			}
4569 			SECASVAR_UNLOCK(sav);
4570 		}
4571 	}
4572 	SAHTREE_RUNLOCK();
4573 
4574 	if (LIST_EMPTY(&emptyq) && LIST_EMPTY(&drainq) &&
4575 	    LIST_EMPTY(&hexpireq) && LIST_EMPTY(&sexpireq))
4576 		return;
4577 
4578 	LIST_INIT(&freeq);
4579 	SAHTREE_WLOCK();
4580 	/* Unlink stale LARVAL SAs */
4581 	sav = LIST_FIRST(&drainq);
4582 	while (sav != NULL) {
4583 		nextsav = LIST_NEXT(sav, drainq);
4584 		/* Check that SA is still LARVAL */
4585 		if (sav->state != SADB_SASTATE_LARVAL) {
4586 			LIST_REMOVE(sav, drainq);
4587 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4588 			sav = nextsav;
4589 			continue;
4590 		}
4591 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
4592 		LIST_REMOVE(sav, spihash);
4593 		sav->state = SADB_SASTATE_DEAD;
4594 		sav = nextsav;
4595 	}
4596 	/* Unlink all SAs with expired HARD lifetime */
4597 	sav = LIST_FIRST(&hexpireq);
4598 	while (sav != NULL) {
4599 		nextsav = LIST_NEXT(sav, drainq);
4600 		/* Check that SA is not unlinked */
4601 		if (sav->state == SADB_SASTATE_DEAD) {
4602 			LIST_REMOVE(sav, drainq);
4603 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4604 			sav = nextsav;
4605 			continue;
4606 		}
4607 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
4608 		LIST_REMOVE(sav, spihash);
4609 		sav->state = SADB_SASTATE_DEAD;
4610 		sav = nextsav;
4611 	}
4612 	/* Mark all SAs with expired SOFT lifetime as DYING */
4613 	sav = LIST_FIRST(&sexpireq);
4614 	while (sav != NULL) {
4615 		nextsav = LIST_NEXT(sav, drainq);
4616 		/* Check that SA is not unlinked */
4617 		if (sav->state == SADB_SASTATE_DEAD) {
4618 			LIST_REMOVE(sav, drainq);
4619 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4620 			sav = nextsav;
4621 			continue;
4622 		}
4623 		/*
4624 		 * NOTE: this doesn't change SA order in the chain.
4625 		 */
4626 		sav->state = SADB_SASTATE_DYING;
4627 		sav = nextsav;
4628 	}
4629 	/* Unlink empty SAHs */
4630 	sah = LIST_FIRST(&emptyq);
4631 	while (sah != NULL) {
4632 		nextsah = LIST_NEXT(sah, drainq);
4633 		/* Check that SAH is still empty and not unlinked */
4634 		if (sah->state == SADB_SASTATE_DEAD ||
4635 		    !TAILQ_EMPTY(&sah->savtree_larval) ||
4636 		    !TAILQ_EMPTY(&sah->savtree_alive)) {
4637 			LIST_REMOVE(sah, drainq);
4638 			key_freesah(&sah); /* release extra reference */
4639 			sah = nextsah;
4640 			continue;
4641 		}
4642 		TAILQ_REMOVE(&V_sahtree, sah, chain);
4643 		LIST_REMOVE(sah, addrhash);
4644 		sah->state = SADB_SASTATE_DEAD;
4645 		sah = nextsah;
4646 	}
4647 	SAHTREE_WUNLOCK();
4648 
4649 	/* Send SPDEXPIRE messages */
4650 	sav = LIST_FIRST(&hexpireq);
4651 	while (sav != NULL) {
4652 		nextsav = LIST_NEXT(sav, drainq);
4653 		key_expire(sav, 1);
4654 		key_freesah(&sav->sah); /* release reference from SAV */
4655 		key_freesav(&sav); /* release extra reference */
4656 		key_freesav(&sav); /* release last reference */
4657 		sav = nextsav;
4658 	}
4659 	sav = LIST_FIRST(&sexpireq);
4660 	while (sav != NULL) {
4661 		nextsav = LIST_NEXT(sav, drainq);
4662 		key_expire(sav, 0);
4663 		key_freesav(&sav); /* release extra reference */
4664 		sav = nextsav;
4665 	}
4666 	/* Free stale LARVAL SAs */
4667 	sav = LIST_FIRST(&drainq);
4668 	while (sav != NULL) {
4669 		nextsav = LIST_NEXT(sav, drainq);
4670 		key_freesah(&sav->sah); /* release reference from SAV */
4671 		key_freesav(&sav); /* release extra reference */
4672 		key_freesav(&sav); /* release last reference */
4673 		sav = nextsav;
4674 	}
4675 	/* Free SAs that were unlinked/changed by someone else */
4676 	sav = LIST_FIRST(&freeq);
4677 	while (sav != NULL) {
4678 		nextsav = LIST_NEXT(sav, drainq);
4679 		key_freesav(&sav); /* release extra reference */
4680 		sav = nextsav;
4681 	}
4682 	/* Free empty SAH */
4683 	sah = LIST_FIRST(&emptyq);
4684 	while (sah != NULL) {
4685 		nextsah = LIST_NEXT(sah, drainq);
4686 		key_freesah(&sah); /* release extra reference */
4687 		key_freesah(&sah); /* release last reference */
4688 		sah = nextsah;
4689 	}
4690 }
4691 
4692 static void
4693 key_flush_acq(time_t now)
4694 {
4695 	struct secacq *acq, *nextacq;
4696 
4697 	/* ACQ tree */
4698 	ACQ_LOCK();
4699 	acq = LIST_FIRST(&V_acqtree);
4700 	while (acq != NULL) {
4701 		nextacq = LIST_NEXT(acq, chain);
4702 		if (now - acq->created > V_key_blockacq_lifetime) {
4703 			LIST_REMOVE(acq, chain);
4704 			LIST_REMOVE(acq, addrhash);
4705 			LIST_REMOVE(acq, seqhash);
4706 			free(acq, M_IPSEC_SAQ);
4707 		}
4708 		acq = nextacq;
4709 	}
4710 	ACQ_UNLOCK();
4711 }
4712 
4713 static void
4714 key_flush_spacq(time_t now)
4715 {
4716 	struct secspacq *acq, *nextacq;
4717 
4718 	/* SP ACQ tree */
4719 	SPACQ_LOCK();
4720 	for (acq = LIST_FIRST(&V_spacqtree); acq != NULL; acq = nextacq) {
4721 		nextacq = LIST_NEXT(acq, chain);
4722 		if (now - acq->created > V_key_blockacq_lifetime
4723 		 && __LIST_CHAINED(acq)) {
4724 			LIST_REMOVE(acq, chain);
4725 			free(acq, M_IPSEC_SAQ);
4726 		}
4727 	}
4728 	SPACQ_UNLOCK();
4729 }
4730 
4731 /*
4732  * time handler.
4733  * scanning SPD and SAD to check status for each entries,
4734  * and do to remove or to expire.
4735  * XXX: year 2038 problem may remain.
4736  */
4737 static void
4738 key_timehandler(void *arg)
4739 {
4740 	VNET_ITERATOR_DECL(vnet_iter);
4741 	time_t now = time_second;
4742 
4743 	VNET_LIST_RLOCK_NOSLEEP();
4744 	VNET_FOREACH(vnet_iter) {
4745 		CURVNET_SET(vnet_iter);
4746 		key_flush_spd(now);
4747 		key_flush_sad(now);
4748 		key_flush_acq(now);
4749 		key_flush_spacq(now);
4750 		CURVNET_RESTORE();
4751 	}
4752 	VNET_LIST_RUNLOCK_NOSLEEP();
4753 
4754 #ifndef IPSEC_DEBUG2
4755 	/* do exchange to tick time !! */
4756 	callout_schedule(&key_timer, hz);
4757 #endif /* IPSEC_DEBUG2 */
4758 }
4759 
4760 u_long
4761 key_random()
4762 {
4763 	u_long value;
4764 
4765 	arc4random_buf(&value, sizeof(value));
4766 	return value;
4767 }
4768 
4769 /*
4770  * map SADB_SATYPE_* to IPPROTO_*.
4771  * if satype == SADB_SATYPE then satype is mapped to ~0.
4772  * OUT:
4773  *	0: invalid satype.
4774  */
4775 static uint8_t
4776 key_satype2proto(uint8_t satype)
4777 {
4778 	switch (satype) {
4779 	case SADB_SATYPE_UNSPEC:
4780 		return IPSEC_PROTO_ANY;
4781 	case SADB_SATYPE_AH:
4782 		return IPPROTO_AH;
4783 	case SADB_SATYPE_ESP:
4784 		return IPPROTO_ESP;
4785 	case SADB_X_SATYPE_IPCOMP:
4786 		return IPPROTO_IPCOMP;
4787 	case SADB_X_SATYPE_TCPSIGNATURE:
4788 		return IPPROTO_TCP;
4789 	default:
4790 		return 0;
4791 	}
4792 	/* NOTREACHED */
4793 }
4794 
4795 /*
4796  * map IPPROTO_* to SADB_SATYPE_*
4797  * OUT:
4798  *	0: invalid protocol type.
4799  */
4800 static uint8_t
4801 key_proto2satype(uint8_t proto)
4802 {
4803 	switch (proto) {
4804 	case IPPROTO_AH:
4805 		return SADB_SATYPE_AH;
4806 	case IPPROTO_ESP:
4807 		return SADB_SATYPE_ESP;
4808 	case IPPROTO_IPCOMP:
4809 		return SADB_X_SATYPE_IPCOMP;
4810 	case IPPROTO_TCP:
4811 		return SADB_X_SATYPE_TCPSIGNATURE;
4812 	default:
4813 		return 0;
4814 	}
4815 	/* NOTREACHED */
4816 }
4817 
4818 /* %%% PF_KEY */
4819 /*
4820  * SADB_GETSPI processing is to receive
4821  *	<base, (SA2), src address, dst address, (SPI range)>
4822  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
4823  * tree with the status of LARVAL, and send
4824  *	<base, SA(*), address(SD)>
4825  * to the IKMPd.
4826  *
4827  * IN:	mhp: pointer to the pointer to each header.
4828  * OUT:	NULL if fail.
4829  *	other if success, return pointer to the message to send.
4830  */
4831 static int
4832 key_getspi(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
4833 {
4834 	struct secasindex saidx;
4835 	struct sadb_address *src0, *dst0;
4836 	struct secasvar *sav;
4837 	uint32_t reqid, spi;
4838 	int error;
4839 	uint8_t mode, proto;
4840 
4841 	IPSEC_ASSERT(so != NULL, ("null socket"));
4842 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
4843 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
4844 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
4845 
4846 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
4847 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)
4848 #ifdef PFKEY_STRICT_CHECKS
4849 	    || SADB_CHECKHDR(mhp, SADB_EXT_SPIRANGE)
4850 #endif
4851 	    ) {
4852 		ipseclog((LOG_DEBUG,
4853 		    "%s: invalid message: missing required header.\n",
4854 		    __func__));
4855 		error = EINVAL;
4856 		goto fail;
4857 	}
4858 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
4859 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)
4860 #ifdef PFKEY_STRICT_CHECKS
4861 	    || SADB_CHECKLEN(mhp, SADB_EXT_SPIRANGE)
4862 #endif
4863 	    ) {
4864 		ipseclog((LOG_DEBUG,
4865 		    "%s: invalid message: wrong header size.\n", __func__));
4866 		error = EINVAL;
4867 		goto fail;
4868 	}
4869 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
4870 		mode = IPSEC_MODE_ANY;
4871 		reqid = 0;
4872 	} else {
4873 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
4874 			ipseclog((LOG_DEBUG,
4875 			    "%s: invalid message: wrong header size.\n",
4876 			    __func__));
4877 			error = EINVAL;
4878 			goto fail;
4879 		}
4880 		mode = ((struct sadb_x_sa2 *)
4881 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
4882 		reqid = ((struct sadb_x_sa2 *)
4883 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
4884 	}
4885 
4886 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
4887 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
4888 
4889 	/* map satype to proto */
4890 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
4891 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
4892 			__func__));
4893 		error = EINVAL;
4894 		goto fail;
4895 	}
4896 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
4897 	    (struct sockaddr *)(dst0 + 1));
4898 	if (error != 0) {
4899 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
4900 		error = EINVAL;
4901 		goto fail;
4902 	}
4903 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
4904 
4905 	/* SPI allocation */
4906 	spi = key_do_getnewspi(
4907 	    (struct sadb_spirange *)mhp->ext[SADB_EXT_SPIRANGE], &saidx);
4908 	if (spi == 0) {
4909 		/*
4910 		 * Requested SPI or SPI range is not available or
4911 		 * already used.
4912 		 */
4913 		error = EEXIST;
4914 		goto fail;
4915 	}
4916 	sav = key_newsav(mhp, &saidx, spi, &error);
4917 	if (sav == NULL)
4918 		goto fail;
4919 
4920 	if (sav->seq != 0) {
4921 		/*
4922 		 * RFC2367:
4923 		 * If the SADB_GETSPI message is in response to a
4924 		 * kernel-generated SADB_ACQUIRE, the sadb_msg_seq
4925 		 * MUST be the same as the SADB_ACQUIRE message.
4926 		 *
4927 		 * XXXAE: However it doesn't definethe behaviour how to
4928 		 * check this and what to do if it doesn't match.
4929 		 * Also what we should do if it matches?
4930 		 *
4931 		 * We can compare saidx used in SADB_ACQUIRE with saidx
4932 		 * used in SADB_GETSPI, but this probably can break
4933 		 * existing software. For now just warn if it doesn't match.
4934 		 *
4935 		 * XXXAE: anyway it looks useless.
4936 		 */
4937 		key_acqdone(&saidx, sav->seq);
4938 	}
4939 	KEYDBG(KEY_STAMP,
4940 	    printf("%s: SA(%p)\n", __func__, sav));
4941 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
4942 
4943     {
4944 	struct mbuf *n, *nn;
4945 	struct sadb_sa *m_sa;
4946 	struct sadb_msg *newmsg;
4947 	int off, len;
4948 
4949 	/* create new sadb_msg to reply. */
4950 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
4951 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
4952 
4953 	MGETHDR(n, M_NOWAIT, MT_DATA);
4954 	if (len > MHLEN) {
4955 		if (!(MCLGET(n, M_NOWAIT))) {
4956 			m_freem(n);
4957 			n = NULL;
4958 		}
4959 	}
4960 	if (!n) {
4961 		error = ENOBUFS;
4962 		goto fail;
4963 	}
4964 
4965 	n->m_len = len;
4966 	n->m_next = NULL;
4967 	off = 0;
4968 
4969 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
4970 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
4971 
4972 	m_sa = (struct sadb_sa *)(mtod(n, caddr_t) + off);
4973 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
4974 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
4975 	m_sa->sadb_sa_spi = spi; /* SPI is already in network byte order */
4976 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
4977 
4978 	IPSEC_ASSERT(off == len,
4979 		("length inconsistency (off %u len %u)", off, len));
4980 
4981 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
4982 	    SADB_EXT_ADDRESS_DST);
4983 	if (!n->m_next) {
4984 		m_freem(n);
4985 		error = ENOBUFS;
4986 		goto fail;
4987 	}
4988 
4989 	if (n->m_len < sizeof(struct sadb_msg)) {
4990 		n = m_pullup(n, sizeof(struct sadb_msg));
4991 		if (n == NULL)
4992 			return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
4993 	}
4994 
4995 	n->m_pkthdr.len = 0;
4996 	for (nn = n; nn; nn = nn->m_next)
4997 		n->m_pkthdr.len += nn->m_len;
4998 
4999 	newmsg = mtod(n, struct sadb_msg *);
5000 	newmsg->sadb_msg_seq = sav->seq;
5001 	newmsg->sadb_msg_errno = 0;
5002 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
5003 
5004 	m_freem(m);
5005 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
5006     }
5007 
5008 fail:
5009 	return (key_senderror(so, m, error));
5010 }
5011 
5012 /*
5013  * allocating new SPI
5014  * called by key_getspi().
5015  * OUT:
5016  *	0:	failure.
5017  *	others: success, SPI in network byte order.
5018  */
5019 static uint32_t
5020 key_do_getnewspi(struct sadb_spirange *spirange, struct secasindex *saidx)
5021 {
5022 	uint32_t min, max, newspi, t;
5023 	int count = V_key_spi_trycnt;
5024 
5025 	/* set spi range to allocate */
5026 	if (spirange != NULL) {
5027 		min = spirange->sadb_spirange_min;
5028 		max = spirange->sadb_spirange_max;
5029 	} else {
5030 		min = V_key_spi_minval;
5031 		max = V_key_spi_maxval;
5032 	}
5033 	/* IPCOMP needs 2-byte SPI */
5034 	if (saidx->proto == IPPROTO_IPCOMP) {
5035 		if (min >= 0x10000)
5036 			min = 0xffff;
5037 		if (max >= 0x10000)
5038 			max = 0xffff;
5039 		if (min > max) {
5040 			t = min; min = max; max = t;
5041 		}
5042 	}
5043 
5044 	if (min == max) {
5045 		if (!key_checkspidup(htonl(min))) {
5046 			ipseclog((LOG_DEBUG, "%s: SPI %u exists already.\n",
5047 			    __func__, min));
5048 			return 0;
5049 		}
5050 
5051 		count--; /* taking one cost. */
5052 		newspi = min;
5053 	} else {
5054 
5055 		/* init SPI */
5056 		newspi = 0;
5057 
5058 		/* when requesting to allocate spi ranged */
5059 		while (count--) {
5060 			/* generate pseudo-random SPI value ranged. */
5061 			newspi = min + (key_random() % (max - min + 1));
5062 			if (!key_checkspidup(htonl(newspi)))
5063 				break;
5064 		}
5065 
5066 		if (count == 0 || newspi == 0) {
5067 			ipseclog((LOG_DEBUG,
5068 			    "%s: failed to allocate SPI.\n", __func__));
5069 			return 0;
5070 		}
5071 	}
5072 
5073 	/* statistics */
5074 	keystat.getspi_count =
5075 	    (keystat.getspi_count + V_key_spi_trycnt - count) / 2;
5076 
5077 	return (htonl(newspi));
5078 }
5079 
5080 /*
5081  * Find TCP-MD5 SA with corresponding secasindex.
5082  * If not found, return NULL and fill SPI with usable value if needed.
5083  */
5084 static struct secasvar *
5085 key_getsav_tcpmd5(struct secasindex *saidx, uint32_t *spi)
5086 {
5087 	SAHTREE_RLOCK_TRACKER;
5088 	struct secashead *sah;
5089 	struct secasvar *sav;
5090 
5091 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP, ("wrong proto"));
5092 	SAHTREE_RLOCK();
5093 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
5094 		if (sah->saidx.proto != IPPROTO_TCP)
5095 			continue;
5096 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
5097 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
5098 			break;
5099 	}
5100 	if (sah != NULL) {
5101 		if (V_key_preferred_oldsa)
5102 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
5103 		else
5104 			sav = TAILQ_FIRST(&sah->savtree_alive);
5105 		if (sav != NULL) {
5106 			SAV_ADDREF(sav);
5107 			SAHTREE_RUNLOCK();
5108 			return (sav);
5109 		}
5110 	}
5111 	if (spi == NULL) {
5112 		/* No SPI required */
5113 		SAHTREE_RUNLOCK();
5114 		return (NULL);
5115 	}
5116 	/* Check that SPI is unique */
5117 	LIST_FOREACH(sav, SAVHASH_HASH(*spi), spihash) {
5118 		if (sav->spi == *spi)
5119 			break;
5120 	}
5121 	if (sav == NULL) {
5122 		SAHTREE_RUNLOCK();
5123 		/* SPI is already unique */
5124 		return (NULL);
5125 	}
5126 	SAHTREE_RUNLOCK();
5127 	/* XXX: not optimal */
5128 	*spi = key_do_getnewspi(NULL, saidx);
5129 	return (NULL);
5130 }
5131 
5132 static int
5133 key_updateaddresses(struct socket *so, struct mbuf *m,
5134     const struct sadb_msghdr *mhp, struct secasvar *sav,
5135     struct secasindex *saidx)
5136 {
5137 	struct sockaddr *newaddr;
5138 	struct secashead *sah;
5139 	struct secasvar *newsav, *tmp;
5140 	struct mbuf *n;
5141 	int error, isnew;
5142 
5143 	/* Check that we need to change SAH */
5144 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC)) {
5145 		newaddr = (struct sockaddr *)(
5146 		    ((struct sadb_address *)
5147 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_SRC]) + 1);
5148 		bcopy(newaddr, &saidx->src, newaddr->sa_len);
5149 		key_porttosaddr(&saidx->src.sa, 0);
5150 	}
5151 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5152 		newaddr = (struct sockaddr *)(
5153 		    ((struct sadb_address *)
5154 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_DST]) + 1);
5155 		bcopy(newaddr, &saidx->dst, newaddr->sa_len);
5156 		key_porttosaddr(&saidx->dst.sa, 0);
5157 	}
5158 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5159 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5160 		error = key_checksockaddrs(&saidx->src.sa, &saidx->dst.sa);
5161 		if (error != 0) {
5162 			ipseclog((LOG_DEBUG, "%s: invalid new sockaddr.\n",
5163 			    __func__));
5164 			return (error);
5165 		}
5166 
5167 		sah = key_getsah(saidx);
5168 		if (sah == NULL) {
5169 			/* create a new SA index */
5170 			sah = key_newsah(saidx);
5171 			if (sah == NULL) {
5172 				ipseclog((LOG_DEBUG,
5173 				    "%s: No more memory.\n", __func__));
5174 				return (ENOBUFS);
5175 			}
5176 			isnew = 2; /* SAH is new */
5177 		} else
5178 			isnew = 1; /* existing SAH is referenced */
5179 	} else {
5180 		/*
5181 		 * src and dst addresses are still the same.
5182 		 * Do we want to change NAT-T config?
5183 		 */
5184 		if (sav->sah->saidx.proto != IPPROTO_ESP ||
5185 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5186 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5187 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5188 			ipseclog((LOG_DEBUG,
5189 			    "%s: invalid message: missing required header.\n",
5190 			    __func__));
5191 			return (EINVAL);
5192 		}
5193 		/* We hold reference to SA, thus SAH will be referenced too. */
5194 		sah = sav->sah;
5195 		isnew = 0;
5196 	}
5197 
5198 	newsav = malloc(sizeof(struct secasvar), M_IPSEC_SA,
5199 	    M_NOWAIT | M_ZERO);
5200 	if (newsav == NULL) {
5201 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5202 		error = ENOBUFS;
5203 		goto fail;
5204 	}
5205 
5206 	/* Clone SA's content into newsav */
5207 	SAV_INITREF(newsav);
5208 	bcopy(sav, newsav, offsetof(struct secasvar, chain));
5209 	/*
5210 	 * We create new NAT-T config if it is needed.
5211 	 * Old NAT-T config will be freed by key_cleansav() when
5212 	 * last reference to SA will be released.
5213 	 */
5214 	newsav->natt = NULL;
5215 	newsav->sah = sah;
5216 	newsav->state = SADB_SASTATE_MATURE;
5217 	error = key_setnatt(newsav, mhp);
5218 	if (error != 0)
5219 		goto fail;
5220 
5221 	SAHTREE_WLOCK();
5222 	/* Check that SA is still alive */
5223 	if (sav->state == SADB_SASTATE_DEAD) {
5224 		/* SA was unlinked */
5225 		SAHTREE_WUNLOCK();
5226 		error = ESRCH;
5227 		goto fail;
5228 	}
5229 
5230 	/* Unlink SA from SAH and SPI hash */
5231 	IPSEC_ASSERT((sav->flags & SADB_X_EXT_F_CLONED) == 0,
5232 	    ("SA is already cloned"));
5233 	IPSEC_ASSERT(sav->state == SADB_SASTATE_MATURE ||
5234 	    sav->state == SADB_SASTATE_DYING,
5235 	    ("Wrong SA state %u\n", sav->state));
5236 	TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
5237 	LIST_REMOVE(sav, spihash);
5238 	sav->state = SADB_SASTATE_DEAD;
5239 
5240 	/*
5241 	 * Link new SA with SAH. Keep SAs ordered by
5242 	 * create time (newer are first).
5243 	 */
5244 	TAILQ_FOREACH(tmp, &sah->savtree_alive, chain) {
5245 		if (newsav->created > tmp->created) {
5246 			TAILQ_INSERT_BEFORE(tmp, newsav, chain);
5247 			break;
5248 		}
5249 	}
5250 	if (tmp == NULL)
5251 		TAILQ_INSERT_TAIL(&sah->savtree_alive, newsav, chain);
5252 
5253 	/* Add new SA into SPI hash. */
5254 	LIST_INSERT_HEAD(SAVHASH_HASH(newsav->spi), newsav, spihash);
5255 
5256 	/* Add new SAH into SADB. */
5257 	if (isnew == 2) {
5258 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
5259 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
5260 		sah->state = SADB_SASTATE_MATURE;
5261 		SAH_ADDREF(sah); /* newsav references new SAH */
5262 	}
5263 	/*
5264 	 * isnew == 1 -> @sah was referenced by key_getsah().
5265 	 * isnew == 0 -> we use the same @sah, that was used by @sav,
5266 	 *	and we use its reference for @newsav.
5267 	 */
5268 	SECASVAR_LOCK(sav);
5269 	/* XXX: replace cntr with pointer? */
5270 	newsav->cntr = sav->cntr;
5271 	sav->flags |= SADB_X_EXT_F_CLONED;
5272 	SECASVAR_UNLOCK(sav);
5273 
5274 	SAHTREE_WUNLOCK();
5275 
5276 	KEYDBG(KEY_STAMP,
5277 	    printf("%s: SA(%p) cloned into SA(%p)\n",
5278 	    __func__, sav, newsav));
5279 	KEYDBG(KEY_DATA, kdebug_secasv(newsav));
5280 
5281 	key_freesav(&sav); /* release last reference */
5282 
5283 	/* set msg buf from mhp */
5284 	n = key_getmsgbuf_x1(m, mhp);
5285 	if (n == NULL) {
5286 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5287 		return (ENOBUFS);
5288 	}
5289 	m_freem(m);
5290 	key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5291 	return (0);
5292 fail:
5293 	if (isnew != 0)
5294 		key_freesah(&sah);
5295 	if (newsav != NULL) {
5296 		if (newsav->natt != NULL)
5297 			free(newsav->natt, M_IPSEC_MISC);
5298 		free(newsav, M_IPSEC_SA);
5299 	}
5300 	return (error);
5301 }
5302 
5303 /*
5304  * SADB_UPDATE processing
5305  * receive
5306  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5307  *       key(AE), (identity(SD),) (sensitivity)>
5308  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5309  * and send
5310  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5311  *       (identity(SD),) (sensitivity)>
5312  * to the ikmpd.
5313  *
5314  * m will always be freed.
5315  */
5316 static int
5317 key_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5318 {
5319 	struct secasindex saidx;
5320 	struct sadb_address *src0, *dst0;
5321 	struct sadb_sa *sa0;
5322 	struct secasvar *sav;
5323 	uint32_t reqid;
5324 	int error;
5325 	uint8_t mode, proto;
5326 
5327 	IPSEC_ASSERT(so != NULL, ("null socket"));
5328 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5329 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5330 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5331 
5332 	/* map satype to proto */
5333 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5334 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5335 		    __func__));
5336 		return key_senderror(so, m, EINVAL);
5337 	}
5338 
5339 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5340 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5341 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5342 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5343 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5344 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5345 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5346 		ipseclog((LOG_DEBUG,
5347 		    "%s: invalid message: missing required header.\n",
5348 		    __func__));
5349 		return key_senderror(so, m, EINVAL);
5350 	}
5351 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5352 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5353 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5354 		ipseclog((LOG_DEBUG,
5355 		    "%s: invalid message: wrong header size.\n", __func__));
5356 		return key_senderror(so, m, EINVAL);
5357 	}
5358 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5359 		mode = IPSEC_MODE_ANY;
5360 		reqid = 0;
5361 	} else {
5362 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5363 			ipseclog((LOG_DEBUG,
5364 			    "%s: invalid message: wrong header size.\n",
5365 			    __func__));
5366 			return key_senderror(so, m, EINVAL);
5367 		}
5368 		mode = ((struct sadb_x_sa2 *)
5369 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5370 		reqid = ((struct sadb_x_sa2 *)
5371 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5372 	}
5373 
5374 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5375 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5376 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5377 
5378 	/*
5379 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5380 	 * SADB_UPDATE message.
5381 	 */
5382 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5383 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5384 #ifdef PFKEY_STRICT_CHECKS
5385 		return key_senderror(so, m, EINVAL);
5386 #endif
5387 	}
5388 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5389 	    (struct sockaddr *)(dst0 + 1));
5390 	if (error != 0) {
5391 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5392 		return key_senderror(so, m, error);
5393 	}
5394 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5395 	sav = key_getsavbyspi(sa0->sadb_sa_spi);
5396 	if (sav == NULL) {
5397 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u\n",
5398 		    __func__, ntohl(sa0->sadb_sa_spi)));
5399 		return key_senderror(so, m, EINVAL);
5400 	}
5401 	/*
5402 	 * Check that SADB_UPDATE issued by the same process that did
5403 	 * SADB_GETSPI or SADB_ADD.
5404 	 */
5405 	if (sav->pid != mhp->msg->sadb_msg_pid) {
5406 		ipseclog((LOG_DEBUG,
5407 		    "%s: pid mismatched (SPI %u, pid %u vs. %u)\n", __func__,
5408 		    ntohl(sav->spi), sav->pid, mhp->msg->sadb_msg_pid));
5409 		key_freesav(&sav);
5410 		return key_senderror(so, m, EINVAL);
5411 	}
5412 	/* saidx should match with SA. */
5413 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_MODE_REQID) == 0) {
5414 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u\n",
5415 		    __func__, ntohl(sav->spi)));
5416 		key_freesav(&sav);
5417 		return key_senderror(so, m, ESRCH);
5418 	}
5419 
5420 	if (sav->state == SADB_SASTATE_LARVAL) {
5421 		if ((mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5422 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) ||
5423 		    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5424 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH))) {
5425 			ipseclog((LOG_DEBUG,
5426 			    "%s: invalid message: missing required header.\n",
5427 			    __func__));
5428 			key_freesav(&sav);
5429 			return key_senderror(so, m, EINVAL);
5430 		}
5431 		/*
5432 		 * We can set any values except src, dst and SPI.
5433 		 */
5434 		error = key_setsaval(sav, mhp);
5435 		if (error != 0) {
5436 			key_freesav(&sav);
5437 			return (key_senderror(so, m, error));
5438 		}
5439 		/* Change SA state to MATURE */
5440 		SAHTREE_WLOCK();
5441 		if (sav->state != SADB_SASTATE_LARVAL) {
5442 			/* SA was deleted or another thread made it MATURE. */
5443 			SAHTREE_WUNLOCK();
5444 			key_freesav(&sav);
5445 			return (key_senderror(so, m, ESRCH));
5446 		}
5447 		/*
5448 		 * NOTE: we keep SAs in savtree_alive ordered by created
5449 		 * time. When SA's state changed from LARVAL to MATURE,
5450 		 * we update its created time in key_setsaval() and move
5451 		 * it into head of savtree_alive.
5452 		 */
5453 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
5454 		TAILQ_INSERT_HEAD(&sav->sah->savtree_alive, sav, chain);
5455 		sav->state = SADB_SASTATE_MATURE;
5456 		SAHTREE_WUNLOCK();
5457 	} else {
5458 		/*
5459 		 * For DYING and MATURE SA we can change only state
5460 		 * and lifetimes. Report EINVAL if something else attempted
5461 		 * to change.
5462 		 */
5463 		if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5464 		    !SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
5465 			key_freesav(&sav);
5466 			return (key_senderror(so, m, EINVAL));
5467 		}
5468 		error = key_updatelifetimes(sav, mhp);
5469 		if (error != 0) {
5470 			key_freesav(&sav);
5471 			return (key_senderror(so, m, error));
5472 		}
5473 		/*
5474 		 * This is FreeBSD extension to RFC2367.
5475 		 * IKEd can specify SADB_X_EXT_NEW_ADDRESS_SRC and/or
5476 		 * SADB_X_EXT_NEW_ADDRESS_DST when it wants to change
5477 		 * SA addresses (for example to implement MOBIKE protocol
5478 		 * as described in RFC4555). Also we allow to change
5479 		 * NAT-T config.
5480 		 */
5481 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5482 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST) ||
5483 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5484 		    sav->natt != NULL) {
5485 			error = key_updateaddresses(so, m, mhp, sav, &saidx);
5486 			key_freesav(&sav);
5487 			if (error != 0)
5488 				return (key_senderror(so, m, error));
5489 			return (0);
5490 		}
5491 		/* Check that SA is still alive */
5492 		SAHTREE_WLOCK();
5493 		if (sav->state == SADB_SASTATE_DEAD) {
5494 			/* SA was unlinked */
5495 			SAHTREE_WUNLOCK();
5496 			key_freesav(&sav);
5497 			return (key_senderror(so, m, ESRCH));
5498 		}
5499 		/*
5500 		 * NOTE: there is possible state moving from DYING to MATURE,
5501 		 * but this doesn't change created time, so we won't reorder
5502 		 * this SA.
5503 		 */
5504 		sav->state = SADB_SASTATE_MATURE;
5505 		SAHTREE_WUNLOCK();
5506 	}
5507 	KEYDBG(KEY_STAMP,
5508 	    printf("%s: SA(%p)\n", __func__, sav));
5509 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5510 	key_freesav(&sav);
5511 
5512     {
5513 	struct mbuf *n;
5514 
5515 	/* set msg buf from mhp */
5516 	n = key_getmsgbuf_x1(m, mhp);
5517 	if (n == NULL) {
5518 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5519 		return key_senderror(so, m, ENOBUFS);
5520 	}
5521 
5522 	m_freem(m);
5523 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5524     }
5525 }
5526 
5527 /*
5528  * SADB_ADD processing
5529  * add an entry to SA database, when received
5530  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5531  *       key(AE), (identity(SD),) (sensitivity)>
5532  * from the ikmpd,
5533  * and send
5534  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5535  *       (identity(SD),) (sensitivity)>
5536  * to the ikmpd.
5537  *
5538  * IGNORE identity and sensitivity messages.
5539  *
5540  * m will always be freed.
5541  */
5542 static int
5543 key_add(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5544 {
5545 	struct secasindex saidx;
5546 	struct sadb_address *src0, *dst0;
5547 	struct sadb_sa *sa0;
5548 	struct secasvar *sav;
5549 	uint32_t reqid, spi;
5550 	uint8_t mode, proto;
5551 	int error;
5552 
5553 	IPSEC_ASSERT(so != NULL, ("null socket"));
5554 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5555 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5556 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5557 
5558 	/* map satype to proto */
5559 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5560 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5561 		    __func__));
5562 		return key_senderror(so, m, EINVAL);
5563 	}
5564 
5565 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5566 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5567 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5568 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP && (
5569 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5570 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT))) ||
5571 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH && (
5572 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH) ||
5573 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH))) ||
5574 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5575 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5576 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5577 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5578 		ipseclog((LOG_DEBUG,
5579 		    "%s: invalid message: missing required header.\n",
5580 		    __func__));
5581 		return key_senderror(so, m, EINVAL);
5582 	}
5583 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5584 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5585 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5586 		ipseclog((LOG_DEBUG,
5587 		    "%s: invalid message: wrong header size.\n", __func__));
5588 		return key_senderror(so, m, EINVAL);
5589 	}
5590 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5591 		mode = IPSEC_MODE_ANY;
5592 		reqid = 0;
5593 	} else {
5594 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5595 			ipseclog((LOG_DEBUG,
5596 			    "%s: invalid message: wrong header size.\n",
5597 			    __func__));
5598 			return key_senderror(so, m, EINVAL);
5599 		}
5600 		mode = ((struct sadb_x_sa2 *)
5601 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5602 		reqid = ((struct sadb_x_sa2 *)
5603 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5604 	}
5605 
5606 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5607 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
5608 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
5609 
5610 	/*
5611 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5612 	 * SADB_ADD message.
5613 	 */
5614 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5615 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5616 #ifdef PFKEY_STRICT_CHECKS
5617 		return key_senderror(so, m, EINVAL);
5618 #endif
5619 	}
5620 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5621 	    (struct sockaddr *)(dst0 + 1));
5622 	if (error != 0) {
5623 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5624 		return key_senderror(so, m, error);
5625 	}
5626 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5627 	spi = sa0->sadb_sa_spi;
5628 	/*
5629 	 * For TCP-MD5 SAs we don't use SPI. Check the uniqueness using
5630 	 * secasindex.
5631 	 * XXXAE: IPComp seems also doesn't use SPI.
5632 	 */
5633 	if (proto == IPPROTO_TCP) {
5634 		sav = key_getsav_tcpmd5(&saidx, &spi);
5635 		if (sav == NULL && spi == 0) {
5636 			/* Failed to allocate SPI */
5637 			ipseclog((LOG_DEBUG, "%s: SA already exists.\n",
5638 			    __func__));
5639 			return key_senderror(so, m, EEXIST);
5640 		}
5641 		/* XXX: SPI that we report back can have another value */
5642 	} else {
5643 		/* We can create new SA only if SPI is different. */
5644 		sav = key_getsavbyspi(spi);
5645 	}
5646 	if (sav != NULL) {
5647 		key_freesav(&sav);
5648 		ipseclog((LOG_DEBUG, "%s: SA already exists.\n", __func__));
5649 		return key_senderror(so, m, EEXIST);
5650 	}
5651 
5652 	sav = key_newsav(mhp, &saidx, spi, &error);
5653 	if (sav == NULL)
5654 		return key_senderror(so, m, error);
5655 	KEYDBG(KEY_STAMP,
5656 	    printf("%s: return SA(%p)\n", __func__, sav));
5657 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5658 	/*
5659 	 * If SADB_ADD was in response to SADB_ACQUIRE, we need to schedule
5660 	 * ACQ for deletion.
5661 	 */
5662 	if (sav->seq != 0)
5663 		key_acqdone(&saidx, sav->seq);
5664 
5665     {
5666 	/*
5667 	 * Don't call key_freesav() on error here, as we would like to
5668 	 * keep the SA in the database.
5669 	 */
5670 	struct mbuf *n;
5671 
5672 	/* set msg buf from mhp */
5673 	n = key_getmsgbuf_x1(m, mhp);
5674 	if (n == NULL) {
5675 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5676 		return key_senderror(so, m, ENOBUFS);
5677 	}
5678 
5679 	m_freem(m);
5680 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5681     }
5682 }
5683 
5684 /*
5685  * NAT-T support.
5686  * IKEd may request the use ESP in UDP encapsulation when it detects the
5687  * presence of NAT. It uses NAT-T extension headers for such SAs to specify
5688  * parameters needed for encapsulation and decapsulation. These PF_KEY
5689  * extension headers are not standardized, so this comment addresses our
5690  * implementation.
5691  * SADB_X_EXT_NAT_T_TYPE specifies type of encapsulation, we support only
5692  * UDP_ENCAP_ESPINUDP as described in RFC3948.
5693  * SADB_X_EXT_NAT_T_SPORT/DPORT specifies source and destination ports for
5694  * UDP header. We use these ports in UDP encapsulation procedure, also we
5695  * can check them in UDP decapsulation procedure.
5696  * SADB_X_EXT_NAT_T_OA[IR] specifies original address of initiator or
5697  * responder. These addresses can be used for transport mode to adjust
5698  * checksum after decapsulation and decryption. Since original IP addresses
5699  * used by peer usually different (we detected presence of NAT), TCP/UDP
5700  * pseudo header checksum and IP header checksum was calculated using original
5701  * addresses. After decapsulation and decryption we need to adjust checksum
5702  * to have correct datagram.
5703  *
5704  * We expect presence of NAT-T extension headers only in SADB_ADD and
5705  * SADB_UPDATE messages. We report NAT-T extension headers in replies
5706  * to SADB_ADD, SADB_UPDATE, SADB_GET, and SADB_DUMP messages.
5707  */
5708 static int
5709 key_setnatt(struct secasvar *sav, const struct sadb_msghdr *mhp)
5710 {
5711 	struct sadb_x_nat_t_port *port;
5712 	struct sadb_x_nat_t_type *type;
5713 	struct sadb_address *oai, *oar;
5714 	struct sockaddr *sa;
5715 	uint32_t addr;
5716 	uint16_t cksum;
5717 
5718 	IPSEC_ASSERT(sav->natt == NULL, ("natt is already initialized"));
5719 	/*
5720 	 * Ignore NAT-T headers if sproto isn't ESP.
5721 	 */
5722 	if (sav->sah->saidx.proto != IPPROTO_ESP)
5723 		return (0);
5724 
5725 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) &&
5726 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) &&
5727 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5728 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5729 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5730 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5731 			ipseclog((LOG_DEBUG,
5732 			    "%s: invalid message: wrong header size.\n",
5733 			    __func__));
5734 			return (EINVAL);
5735 		}
5736 	} else
5737 		return (0);
5738 
5739 	type = (struct sadb_x_nat_t_type *)mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5740 	if (type->sadb_x_nat_t_type_type != UDP_ENCAP_ESPINUDP) {
5741 		ipseclog((LOG_DEBUG, "%s: unsupported NAT-T type %u.\n",
5742 		    __func__, type->sadb_x_nat_t_type_type));
5743 		return (EINVAL);
5744 	}
5745 	/*
5746 	 * Allocate storage for NAT-T config.
5747 	 * On error it will be released by key_cleansav().
5748 	 */
5749 	sav->natt = malloc(sizeof(struct secnatt), M_IPSEC_MISC,
5750 	    M_NOWAIT | M_ZERO);
5751 	if (sav->natt == NULL) {
5752 		PFKEYSTAT_INC(in_nomem);
5753 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5754 		return (ENOBUFS);
5755 	}
5756 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5757 	if (port->sadb_x_nat_t_port_port == 0) {
5758 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T sport specified.\n",
5759 		    __func__));
5760 		return (EINVAL);
5761 	}
5762 	sav->natt->sport = port->sadb_x_nat_t_port_port;
5763 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5764 	if (port->sadb_x_nat_t_port_port == 0) {
5765 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T dport specified.\n",
5766 		    __func__));
5767 		return (EINVAL);
5768 	}
5769 	sav->natt->dport = port->sadb_x_nat_t_port_port;
5770 
5771 	/*
5772 	 * SADB_X_EXT_NAT_T_OAI and SADB_X_EXT_NAT_T_OAR are optional
5773 	 * and needed only for transport mode IPsec.
5774 	 * Usually NAT translates only one address, but it is possible,
5775 	 * that both addresses could be translated.
5776 	 * NOTE: Value of SADB_X_EXT_NAT_T_OAI is equal to SADB_X_EXT_NAT_T_OA.
5777 	 */
5778 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAI)) {
5779 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAI)) {
5780 			ipseclog((LOG_DEBUG,
5781 			    "%s: invalid message: wrong header size.\n",
5782 			    __func__));
5783 			return (EINVAL);
5784 		}
5785 		oai = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAI];
5786 	} else
5787 		oai = NULL;
5788 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAR)) {
5789 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAR)) {
5790 			ipseclog((LOG_DEBUG,
5791 			    "%s: invalid message: wrong header size.\n",
5792 			    __func__));
5793 			return (EINVAL);
5794 		}
5795 		oar = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAR];
5796 	} else
5797 		oar = NULL;
5798 
5799 	/* Initialize addresses only for transport mode */
5800 	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL) {
5801 		cksum = 0;
5802 		if (oai != NULL) {
5803 			/* Currently we support only AF_INET */
5804 			sa = (struct sockaddr *)(oai + 1);
5805 			if (sa->sa_family != AF_INET ||
5806 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5807 				ipseclog((LOG_DEBUG,
5808 				    "%s: wrong NAT-OAi header.\n",
5809 				    __func__));
5810 				return (EINVAL);
5811 			}
5812 			/* Ignore address if it the same */
5813 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5814 			    sav->sah->saidx.src.sin.sin_addr.s_addr) {
5815 				bcopy(sa, &sav->natt->oai.sa, sa->sa_len);
5816 				sav->natt->flags |= IPSEC_NATT_F_OAI;
5817 				/* Calculate checksum delta */
5818 				addr = sav->sah->saidx.src.sin.sin_addr.s_addr;
5819 				cksum = in_addword(cksum, ~addr >> 16);
5820 				cksum = in_addword(cksum, ~addr & 0xffff);
5821 				addr = sav->natt->oai.sin.sin_addr.s_addr;
5822 				cksum = in_addword(cksum, addr >> 16);
5823 				cksum = in_addword(cksum, addr & 0xffff);
5824 			}
5825 		}
5826 		if (oar != NULL) {
5827 			/* Currently we support only AF_INET */
5828 			sa = (struct sockaddr *)(oar + 1);
5829 			if (sa->sa_family != AF_INET ||
5830 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5831 				ipseclog((LOG_DEBUG,
5832 				    "%s: wrong NAT-OAr header.\n",
5833 				    __func__));
5834 				return (EINVAL);
5835 			}
5836 			/* Ignore address if it the same */
5837 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5838 			    sav->sah->saidx.dst.sin.sin_addr.s_addr) {
5839 				bcopy(sa, &sav->natt->oar.sa, sa->sa_len);
5840 				sav->natt->flags |= IPSEC_NATT_F_OAR;
5841 				/* Calculate checksum delta */
5842 				addr = sav->sah->saidx.dst.sin.sin_addr.s_addr;
5843 				cksum = in_addword(cksum, ~addr >> 16);
5844 				cksum = in_addword(cksum, ~addr & 0xffff);
5845 				addr = sav->natt->oar.sin.sin_addr.s_addr;
5846 				cksum = in_addword(cksum, addr >> 16);
5847 				cksum = in_addword(cksum, addr & 0xffff);
5848 			}
5849 		}
5850 		sav->natt->cksum = cksum;
5851 	}
5852 	return (0);
5853 }
5854 
5855 static int
5856 key_setident(struct secashead *sah, const struct sadb_msghdr *mhp)
5857 {
5858 	const struct sadb_ident *idsrc, *iddst;
5859 
5860 	IPSEC_ASSERT(sah != NULL, ("null secashead"));
5861 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5862 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5863 
5864 	/* don't make buffer if not there */
5865 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) &&
5866 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5867 		sah->idents = NULL;
5868 		sah->identd = NULL;
5869 		return (0);
5870 	}
5871 
5872 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) ||
5873 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5874 		ipseclog((LOG_DEBUG, "%s: invalid identity.\n", __func__));
5875 		return (EINVAL);
5876 	}
5877 
5878 	idsrc = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_SRC];
5879 	iddst = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_DST];
5880 
5881 	/* validity check */
5882 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
5883 		ipseclog((LOG_DEBUG, "%s: ident type mismatch.\n", __func__));
5884 		return EINVAL;
5885 	}
5886 
5887 	switch (idsrc->sadb_ident_type) {
5888 	case SADB_IDENTTYPE_PREFIX:
5889 	case SADB_IDENTTYPE_FQDN:
5890 	case SADB_IDENTTYPE_USERFQDN:
5891 	default:
5892 		/* XXX do nothing */
5893 		sah->idents = NULL;
5894 		sah->identd = NULL;
5895 	 	return 0;
5896 	}
5897 
5898 	/* make structure */
5899 	sah->idents = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5900 	if (sah->idents == NULL) {
5901 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5902 		return ENOBUFS;
5903 	}
5904 	sah->identd = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5905 	if (sah->identd == NULL) {
5906 		free(sah->idents, M_IPSEC_MISC);
5907 		sah->idents = NULL;
5908 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5909 		return ENOBUFS;
5910 	}
5911 	sah->idents->type = idsrc->sadb_ident_type;
5912 	sah->idents->id = idsrc->sadb_ident_id;
5913 
5914 	sah->identd->type = iddst->sadb_ident_type;
5915 	sah->identd->id = iddst->sadb_ident_id;
5916 
5917 	return 0;
5918 }
5919 
5920 /*
5921  * m will not be freed on return.
5922  * it is caller's responsibility to free the result.
5923  *
5924  * Called from SADB_ADD and SADB_UPDATE. Reply will contain headers
5925  * from the request in defined order.
5926  */
5927 static struct mbuf *
5928 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
5929 {
5930 	struct mbuf *n;
5931 
5932 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5933 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5934 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5935 
5936 	/* create new sadb_msg to reply. */
5937 	n = key_gather_mbuf(m, mhp, 1, 16, SADB_EXT_RESERVED,
5938 	    SADB_EXT_SA, SADB_X_EXT_SA2,
5939 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
5940 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
5941 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
5942 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
5943 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
5944 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NEW_ADDRESS_SRC,
5945 	    SADB_X_EXT_NEW_ADDRESS_DST);
5946 	if (!n)
5947 		return NULL;
5948 
5949 	if (n->m_len < sizeof(struct sadb_msg)) {
5950 		n = m_pullup(n, sizeof(struct sadb_msg));
5951 		if (n == NULL)
5952 			return NULL;
5953 	}
5954 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
5955 	mtod(n, struct sadb_msg *)->sadb_msg_len =
5956 	    PFKEY_UNIT64(n->m_pkthdr.len);
5957 
5958 	return n;
5959 }
5960 
5961 /*
5962  * SADB_DELETE processing
5963  * receive
5964  *   <base, SA(*), address(SD)>
5965  * from the ikmpd, and set SADB_SASTATE_DEAD,
5966  * and send,
5967  *   <base, SA(*), address(SD)>
5968  * to the ikmpd.
5969  *
5970  * m will always be freed.
5971  */
5972 static int
5973 key_delete(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5974 {
5975 	struct secasindex saidx;
5976 	struct sadb_address *src0, *dst0;
5977 	struct secasvar *sav;
5978 	struct sadb_sa *sa0;
5979 	uint8_t proto;
5980 
5981 	IPSEC_ASSERT(so != NULL, ("null socket"));
5982 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5983 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5984 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5985 
5986 	/* map satype to proto */
5987 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5988 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5989 		    __func__));
5990 		return key_senderror(so, m, EINVAL);
5991 	}
5992 
5993 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5994 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5995 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5996 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5997 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
5998 		    __func__));
5999 		return key_senderror(so, m, EINVAL);
6000 	}
6001 
6002 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
6003 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
6004 
6005 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6006 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6007 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6008 		return (key_senderror(so, m, EINVAL));
6009 	}
6010 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6011 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
6012 		/*
6013 		 * Caller wants us to delete all non-LARVAL SAs
6014 		 * that match the src/dst.  This is used during
6015 		 * IKE INITIAL-CONTACT.
6016 		 * XXXAE: this looks like some extension to RFC2367.
6017 		 */
6018 		ipseclog((LOG_DEBUG, "%s: doing delete all.\n", __func__));
6019 		return (key_delete_all(so, m, mhp, &saidx));
6020 	}
6021 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
6022 		ipseclog((LOG_DEBUG,
6023 		    "%s: invalid message: wrong header size.\n", __func__));
6024 		return (key_senderror(so, m, EINVAL));
6025 	}
6026 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6027 	if (proto == IPPROTO_TCP)
6028 		sav = key_getsav_tcpmd5(&saidx, NULL);
6029 	else
6030 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6031 	if (sav == NULL) {
6032 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u.\n",
6033 		    __func__, ntohl(sa0->sadb_sa_spi)));
6034 		return (key_senderror(so, m, ESRCH));
6035 	}
6036 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6037 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6038 		    __func__, ntohl(sav->spi)));
6039 		key_freesav(&sav);
6040 		return (key_senderror(so, m, ESRCH));
6041 	}
6042 	KEYDBG(KEY_STAMP,
6043 	    printf("%s: SA(%p)\n", __func__, sav));
6044 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
6045 	key_unlinksav(sav);
6046 	key_freesav(&sav);
6047 
6048     {
6049 	struct mbuf *n;
6050 	struct sadb_msg *newmsg;
6051 
6052 	/* create new sadb_msg to reply. */
6053 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
6054 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6055 	if (!n)
6056 		return key_senderror(so, m, ENOBUFS);
6057 
6058 	if (n->m_len < sizeof(struct sadb_msg)) {
6059 		n = m_pullup(n, sizeof(struct sadb_msg));
6060 		if (n == NULL)
6061 			return key_senderror(so, m, ENOBUFS);
6062 	}
6063 	newmsg = mtod(n, struct sadb_msg *);
6064 	newmsg->sadb_msg_errno = 0;
6065 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6066 
6067 	m_freem(m);
6068 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6069     }
6070 }
6071 
6072 /*
6073  * delete all SAs for src/dst.  Called from key_delete().
6074  */
6075 static int
6076 key_delete_all(struct socket *so, struct mbuf *m,
6077     const struct sadb_msghdr *mhp, struct secasindex *saidx)
6078 {
6079 	struct secasvar_queue drainq;
6080 	struct secashead *sah;
6081 	struct secasvar *sav, *nextsav;
6082 
6083 	TAILQ_INIT(&drainq);
6084 	SAHTREE_WLOCK();
6085 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
6086 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_HEAD) == 0)
6087 			continue;
6088 		/* Move all ALIVE SAs into drainq */
6089 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6090 	}
6091 	/* Unlink all queued SAs from SPI hash */
6092 	TAILQ_FOREACH(sav, &drainq, chain) {
6093 		sav->state = SADB_SASTATE_DEAD;
6094 		LIST_REMOVE(sav, spihash);
6095 	}
6096 	SAHTREE_WUNLOCK();
6097 	/* Now we can release reference for all SAs in drainq */
6098 	sav = TAILQ_FIRST(&drainq);
6099 	while (sav != NULL) {
6100 		KEYDBG(KEY_STAMP,
6101 		    printf("%s: SA(%p)\n", __func__, sav));
6102 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6103 		nextsav = TAILQ_NEXT(sav, chain);
6104 		key_freesah(&sav->sah); /* release reference from SAV */
6105 		key_freesav(&sav); /* release last reference */
6106 		sav = nextsav;
6107 	}
6108 
6109     {
6110 	struct mbuf *n;
6111 	struct sadb_msg *newmsg;
6112 
6113 	/* create new sadb_msg to reply. */
6114 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
6115 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6116 	if (!n)
6117 		return key_senderror(so, m, ENOBUFS);
6118 
6119 	if (n->m_len < sizeof(struct sadb_msg)) {
6120 		n = m_pullup(n, sizeof(struct sadb_msg));
6121 		if (n == NULL)
6122 			return key_senderror(so, m, ENOBUFS);
6123 	}
6124 	newmsg = mtod(n, struct sadb_msg *);
6125 	newmsg->sadb_msg_errno = 0;
6126 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6127 
6128 	m_freem(m);
6129 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6130     }
6131 }
6132 
6133 /*
6134  * Delete all alive SAs for corresponding xform.
6135  * Larval SAs have not initialized tdb_xform, so it is safe to leave them
6136  * here when xform disappears.
6137  */
6138 void
6139 key_delete_xform(const struct xformsw *xsp)
6140 {
6141 	struct secasvar_queue drainq;
6142 	struct secashead *sah;
6143 	struct secasvar *sav, *nextsav;
6144 
6145 	TAILQ_INIT(&drainq);
6146 	SAHTREE_WLOCK();
6147 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
6148 		sav = TAILQ_FIRST(&sah->savtree_alive);
6149 		if (sav == NULL)
6150 			continue;
6151 		if (sav->tdb_xform != xsp)
6152 			continue;
6153 		/*
6154 		 * It is supposed that all SAs in the chain are related to
6155 		 * one xform.
6156 		 */
6157 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6158 	}
6159 	/* Unlink all queued SAs from SPI hash */
6160 	TAILQ_FOREACH(sav, &drainq, chain) {
6161 		sav->state = SADB_SASTATE_DEAD;
6162 		LIST_REMOVE(sav, spihash);
6163 	}
6164 	SAHTREE_WUNLOCK();
6165 
6166 	/* Now we can release reference for all SAs in drainq */
6167 	sav = TAILQ_FIRST(&drainq);
6168 	while (sav != NULL) {
6169 		KEYDBG(KEY_STAMP,
6170 		    printf("%s: SA(%p)\n", __func__, sav));
6171 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6172 		nextsav = TAILQ_NEXT(sav, chain);
6173 		key_freesah(&sav->sah); /* release reference from SAV */
6174 		key_freesav(&sav); /* release last reference */
6175 		sav = nextsav;
6176 	}
6177 }
6178 
6179 /*
6180  * SADB_GET processing
6181  * receive
6182  *   <base, SA(*), address(SD)>
6183  * from the ikmpd, and get a SP and a SA to respond,
6184  * and send,
6185  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6186  *       (identity(SD),) (sensitivity)>
6187  * to the ikmpd.
6188  *
6189  * m will always be freed.
6190  */
6191 static int
6192 key_get(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6193 {
6194 	struct secasindex saidx;
6195 	struct sadb_address *src0, *dst0;
6196 	struct sadb_sa *sa0;
6197 	struct secasvar *sav;
6198 	uint8_t proto;
6199 
6200 	IPSEC_ASSERT(so != NULL, ("null socket"));
6201 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6202 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6203 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6204 
6205 	/* map satype to proto */
6206 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6207 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6208 			__func__));
6209 		return key_senderror(so, m, EINVAL);
6210 	}
6211 
6212 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
6213 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6214 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)) {
6215 		ipseclog((LOG_DEBUG,
6216 		    "%s: invalid message: missing required header.\n",
6217 		    __func__));
6218 		return key_senderror(so, m, EINVAL);
6219 	}
6220 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
6221 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6222 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6223 		ipseclog((LOG_DEBUG,
6224 		    "%s: invalid message: wrong header size.\n", __func__));
6225 		return key_senderror(so, m, EINVAL);
6226 	}
6227 
6228 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6229 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
6230 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
6231 
6232 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6233 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6234 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6235 		return key_senderror(so, m, EINVAL);
6236 	}
6237 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6238 
6239 	if (proto == IPPROTO_TCP)
6240 		sav = key_getsav_tcpmd5(&saidx, NULL);
6241 	else
6242 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6243 	if (sav == NULL) {
6244 		ipseclog((LOG_DEBUG, "%s: no SA found.\n", __func__));
6245 		return key_senderror(so, m, ESRCH);
6246 	}
6247 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6248 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6249 		    __func__, ntohl(sa0->sadb_sa_spi)));
6250 		key_freesav(&sav);
6251 		return (key_senderror(so, m, ESRCH));
6252 	}
6253 
6254     {
6255 	struct mbuf *n;
6256 	uint8_t satype;
6257 
6258 	/* map proto to satype */
6259 	if ((satype = key_proto2satype(sav->sah->saidx.proto)) == 0) {
6260 		ipseclog((LOG_DEBUG, "%s: there was invalid proto in SAD.\n",
6261 		    __func__));
6262 		key_freesav(&sav);
6263 		return key_senderror(so, m, EINVAL);
6264 	}
6265 
6266 	/* create new sadb_msg to reply. */
6267 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
6268 	    mhp->msg->sadb_msg_pid);
6269 
6270 	key_freesav(&sav);
6271 	if (!n)
6272 		return key_senderror(so, m, ENOBUFS);
6273 
6274 	m_freem(m);
6275 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
6276     }
6277 }
6278 
6279 /* XXX make it sysctl-configurable? */
6280 static void
6281 key_getcomb_setlifetime(struct sadb_comb *comb)
6282 {
6283 
6284 	comb->sadb_comb_soft_allocations = 1;
6285 	comb->sadb_comb_hard_allocations = 1;
6286 	comb->sadb_comb_soft_bytes = 0;
6287 	comb->sadb_comb_hard_bytes = 0;
6288 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
6289 	comb->sadb_comb_soft_addtime = comb->sadb_comb_soft_addtime * 80 / 100;
6290 	comb->sadb_comb_soft_usetime = 28800;	/* 8 hours */
6291 	comb->sadb_comb_hard_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
6292 }
6293 
6294 /*
6295  * XXX reorder combinations by preference
6296  * XXX no idea if the user wants ESP authentication or not
6297  */
6298 static struct mbuf *
6299 key_getcomb_ealg(void)
6300 {
6301 	struct sadb_comb *comb;
6302 	const struct enc_xform *algo;
6303 	struct mbuf *result = NULL, *m, *n;
6304 	int encmin;
6305 	int i, off, o;
6306 	int totlen;
6307 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6308 
6309 	m = NULL;
6310 	for (i = 1; i <= SADB_EALG_MAX; i++) {
6311 		algo = enc_algorithm_lookup(i);
6312 		if (algo == NULL)
6313 			continue;
6314 
6315 		/* discard algorithms with key size smaller than system min */
6316 		if (_BITS(algo->maxkey) < V_ipsec_esp_keymin)
6317 			continue;
6318 		if (_BITS(algo->minkey) < V_ipsec_esp_keymin)
6319 			encmin = V_ipsec_esp_keymin;
6320 		else
6321 			encmin = _BITS(algo->minkey);
6322 
6323 		if (V_ipsec_esp_auth)
6324 			m = key_getcomb_ah();
6325 		else {
6326 			IPSEC_ASSERT(l <= MLEN,
6327 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6328 			MGET(m, M_NOWAIT, MT_DATA);
6329 			if (m) {
6330 				M_ALIGN(m, l);
6331 				m->m_len = l;
6332 				m->m_next = NULL;
6333 				bzero(mtod(m, caddr_t), m->m_len);
6334 			}
6335 		}
6336 		if (!m)
6337 			goto fail;
6338 
6339 		totlen = 0;
6340 		for (n = m; n; n = n->m_next)
6341 			totlen += n->m_len;
6342 		IPSEC_ASSERT((totlen % l) == 0, ("totlen=%u, l=%u", totlen, l));
6343 
6344 		for (off = 0; off < totlen; off += l) {
6345 			n = m_pulldown(m, off, l, &o);
6346 			if (!n) {
6347 				/* m is already freed */
6348 				goto fail;
6349 			}
6350 			comb = (struct sadb_comb *)(mtod(n, caddr_t) + o);
6351 			bzero(comb, sizeof(*comb));
6352 			key_getcomb_setlifetime(comb);
6353 			comb->sadb_comb_encrypt = i;
6354 			comb->sadb_comb_encrypt_minbits = encmin;
6355 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
6356 		}
6357 
6358 		if (!result)
6359 			result = m;
6360 		else
6361 			m_cat(result, m);
6362 	}
6363 
6364 	return result;
6365 
6366  fail:
6367 	if (result)
6368 		m_freem(result);
6369 	return NULL;
6370 }
6371 
6372 static void
6373 key_getsizes_ah(const struct auth_hash *ah, int alg, u_int16_t* min,
6374     u_int16_t* max)
6375 {
6376 
6377 	*min = *max = ah->hashsize;
6378 	if (ah->keysize == 0) {
6379 		/*
6380 		 * Transform takes arbitrary key size but algorithm
6381 		 * key size is restricted.  Enforce this here.
6382 		 */
6383 		switch (alg) {
6384 		case SADB_X_AALG_MD5:	*min = *max = 16; break;
6385 		case SADB_X_AALG_SHA:	*min = *max = 20; break;
6386 		case SADB_X_AALG_NULL:	*min = 1; *max = 256; break;
6387 		case SADB_X_AALG_SHA2_256: *min = *max = 32; break;
6388 		case SADB_X_AALG_SHA2_384: *min = *max = 48; break;
6389 		case SADB_X_AALG_SHA2_512: *min = *max = 64; break;
6390 		default:
6391 			DPRINTF(("%s: unknown AH algorithm %u\n",
6392 				__func__, alg));
6393 			break;
6394 		}
6395 	}
6396 }
6397 
6398 /*
6399  * XXX reorder combinations by preference
6400  */
6401 static struct mbuf *
6402 key_getcomb_ah()
6403 {
6404 	const struct auth_hash *algo;
6405 	struct sadb_comb *comb;
6406 	struct mbuf *m;
6407 	u_int16_t minkeysize, maxkeysize;
6408 	int i;
6409 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6410 
6411 	m = NULL;
6412 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6413 #if 1
6414 		/* we prefer HMAC algorithms, not old algorithms */
6415 		if (i != SADB_AALG_SHA1HMAC &&
6416 		    i != SADB_AALG_MD5HMAC  &&
6417 		    i != SADB_X_AALG_SHA2_256 &&
6418 		    i != SADB_X_AALG_SHA2_384 &&
6419 		    i != SADB_X_AALG_SHA2_512)
6420 			continue;
6421 #endif
6422 		algo = auth_algorithm_lookup(i);
6423 		if (!algo)
6424 			continue;
6425 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6426 		/* discard algorithms with key size smaller than system min */
6427 		if (_BITS(minkeysize) < V_ipsec_ah_keymin)
6428 			continue;
6429 
6430 		if (!m) {
6431 			IPSEC_ASSERT(l <= MLEN,
6432 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6433 			MGET(m, M_NOWAIT, MT_DATA);
6434 			if (m) {
6435 				M_ALIGN(m, l);
6436 				m->m_len = l;
6437 				m->m_next = NULL;
6438 			}
6439 		} else
6440 			M_PREPEND(m, l, M_NOWAIT);
6441 		if (!m)
6442 			return NULL;
6443 
6444 		comb = mtod(m, struct sadb_comb *);
6445 		bzero(comb, sizeof(*comb));
6446 		key_getcomb_setlifetime(comb);
6447 		comb->sadb_comb_auth = i;
6448 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6449 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6450 	}
6451 
6452 	return m;
6453 }
6454 
6455 /*
6456  * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6457  * XXX reorder combinations by preference
6458  */
6459 static struct mbuf *
6460 key_getcomb_ipcomp()
6461 {
6462 	const struct comp_algo *algo;
6463 	struct sadb_comb *comb;
6464 	struct mbuf *m;
6465 	int i;
6466 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6467 
6468 	m = NULL;
6469 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6470 		algo = comp_algorithm_lookup(i);
6471 		if (!algo)
6472 			continue;
6473 
6474 		if (!m) {
6475 			IPSEC_ASSERT(l <= MLEN,
6476 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6477 			MGET(m, M_NOWAIT, MT_DATA);
6478 			if (m) {
6479 				M_ALIGN(m, l);
6480 				m->m_len = l;
6481 				m->m_next = NULL;
6482 			}
6483 		} else
6484 			M_PREPEND(m, l, M_NOWAIT);
6485 		if (!m)
6486 			return NULL;
6487 
6488 		comb = mtod(m, struct sadb_comb *);
6489 		bzero(comb, sizeof(*comb));
6490 		key_getcomb_setlifetime(comb);
6491 		comb->sadb_comb_encrypt = i;
6492 		/* what should we set into sadb_comb_*_{min,max}bits? */
6493 	}
6494 
6495 	return m;
6496 }
6497 
6498 /*
6499  * XXX no way to pass mode (transport/tunnel) to userland
6500  * XXX replay checking?
6501  * XXX sysctl interface to ipsec_{ah,esp}_keymin
6502  */
6503 static struct mbuf *
6504 key_getprop(const struct secasindex *saidx)
6505 {
6506 	struct sadb_prop *prop;
6507 	struct mbuf *m, *n;
6508 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6509 	int totlen;
6510 
6511 	switch (saidx->proto)  {
6512 	case IPPROTO_ESP:
6513 		m = key_getcomb_ealg();
6514 		break;
6515 	case IPPROTO_AH:
6516 		m = key_getcomb_ah();
6517 		break;
6518 	case IPPROTO_IPCOMP:
6519 		m = key_getcomb_ipcomp();
6520 		break;
6521 	default:
6522 		return NULL;
6523 	}
6524 
6525 	if (!m)
6526 		return NULL;
6527 	M_PREPEND(m, l, M_NOWAIT);
6528 	if (!m)
6529 		return NULL;
6530 
6531 	totlen = 0;
6532 	for (n = m; n; n = n->m_next)
6533 		totlen += n->m_len;
6534 
6535 	prop = mtod(m, struct sadb_prop *);
6536 	bzero(prop, sizeof(*prop));
6537 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6538 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6539 	prop->sadb_prop_replay = 32;	/* XXX */
6540 
6541 	return m;
6542 }
6543 
6544 /*
6545  * SADB_ACQUIRE processing called by key_checkrequest() and key_acquire2().
6546  * send
6547  *   <base, SA, address(SD), (address(P)), x_policy,
6548  *       (identity(SD),) (sensitivity,) proposal>
6549  * to KMD, and expect to receive
6550  *   <base> with SADB_ACQUIRE if error occurred,
6551  * or
6552  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6553  * from KMD by PF_KEY.
6554  *
6555  * XXX x_policy is outside of RFC2367 (KAME extension).
6556  * XXX sensitivity is not supported.
6557  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6558  * see comment for key_getcomb_ipcomp().
6559  *
6560  * OUT:
6561  *    0     : succeed
6562  *    others: error number
6563  */
6564 static int
6565 key_acquire(const struct secasindex *saidx, struct secpolicy *sp)
6566 {
6567 	union sockaddr_union addr;
6568 	struct mbuf *result, *m;
6569 	uint32_t seq;
6570 	int error;
6571 	uint16_t ul_proto;
6572 	uint8_t mask, satype;
6573 
6574 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
6575 	satype = key_proto2satype(saidx->proto);
6576 	IPSEC_ASSERT(satype != 0, ("null satype, protocol %u", saidx->proto));
6577 
6578 	error = -1;
6579 	result = NULL;
6580 	ul_proto = IPSEC_ULPROTO_ANY;
6581 
6582 	/* Get seq number to check whether sending message or not. */
6583 	seq = key_getacq(saidx, &error);
6584 	if (seq == 0)
6585 		return (error);
6586 
6587 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0);
6588 	if (!m) {
6589 		error = ENOBUFS;
6590 		goto fail;
6591 	}
6592 	result = m;
6593 
6594 	/*
6595 	 * set sadb_address for saidx's.
6596 	 *
6597 	 * Note that if sp is supplied, then we're being called from
6598 	 * key_allocsa_policy() and should supply port and protocol
6599 	 * information.
6600 	 * XXXAE: why only TCP and UDP? ICMP and SCTP looks applicable too.
6601 	 * XXXAE: probably we can handle this in the ipsec[46]_allocsa().
6602 	 * XXXAE: it looks like we should save this info in the ACQ entry.
6603 	 */
6604 	if (sp != NULL && (sp->spidx.ul_proto == IPPROTO_TCP ||
6605 	    sp->spidx.ul_proto == IPPROTO_UDP))
6606 		ul_proto = sp->spidx.ul_proto;
6607 
6608 	addr = saidx->src;
6609 	mask = FULLMASK;
6610 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6611 		switch (sp->spidx.src.sa.sa_family) {
6612 		case AF_INET:
6613 			if (sp->spidx.src.sin.sin_port != IPSEC_PORT_ANY) {
6614 				addr.sin.sin_port = sp->spidx.src.sin.sin_port;
6615 				mask = sp->spidx.prefs;
6616 			}
6617 			break;
6618 		case AF_INET6:
6619 			if (sp->spidx.src.sin6.sin6_port != IPSEC_PORT_ANY) {
6620 				addr.sin6.sin6_port =
6621 				    sp->spidx.src.sin6.sin6_port;
6622 				mask = sp->spidx.prefs;
6623 			}
6624 			break;
6625 		default:
6626 			break;
6627 		}
6628 	}
6629 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &addr.sa, mask, ul_proto);
6630 	if (!m) {
6631 		error = ENOBUFS;
6632 		goto fail;
6633 	}
6634 	m_cat(result, m);
6635 
6636 	addr = saidx->dst;
6637 	mask = FULLMASK;
6638 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6639 		switch (sp->spidx.dst.sa.sa_family) {
6640 		case AF_INET:
6641 			if (sp->spidx.dst.sin.sin_port != IPSEC_PORT_ANY) {
6642 				addr.sin.sin_port = sp->spidx.dst.sin.sin_port;
6643 				mask = sp->spidx.prefd;
6644 			}
6645 			break;
6646 		case AF_INET6:
6647 			if (sp->spidx.dst.sin6.sin6_port != IPSEC_PORT_ANY) {
6648 				addr.sin6.sin6_port =
6649 				    sp->spidx.dst.sin6.sin6_port;
6650 				mask = sp->spidx.prefd;
6651 			}
6652 			break;
6653 		default:
6654 			break;
6655 		}
6656 	}
6657 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &addr.sa, mask, ul_proto);
6658 	if (!m) {
6659 		error = ENOBUFS;
6660 		goto fail;
6661 	}
6662 	m_cat(result, m);
6663 
6664 	/* XXX proxy address (optional) */
6665 
6666 	/*
6667 	 * Set sadb_x_policy. This is KAME extension to RFC2367.
6668 	 */
6669 	if (sp != NULL) {
6670 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6671 		    sp->priority);
6672 		if (!m) {
6673 			error = ENOBUFS;
6674 			goto fail;
6675 		}
6676 		m_cat(result, m);
6677 	}
6678 
6679 	/*
6680 	 * Set sadb_x_sa2 extension if saidx->reqid is not zero.
6681 	 * This is FreeBSD extension to RFC2367.
6682 	 */
6683 	if (saidx->reqid != 0) {
6684 		m = key_setsadbxsa2(saidx->mode, 0, saidx->reqid);
6685 		if (m == NULL) {
6686 			error = ENOBUFS;
6687 			goto fail;
6688 		}
6689 		m_cat(result, m);
6690 	}
6691 	/* XXX identity (optional) */
6692 #if 0
6693 	if (idexttype && fqdn) {
6694 		/* create identity extension (FQDN) */
6695 		struct sadb_ident *id;
6696 		int fqdnlen;
6697 
6698 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
6699 		id = (struct sadb_ident *)p;
6700 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6701 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6702 		id->sadb_ident_exttype = idexttype;
6703 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
6704 		bcopy(fqdn, id + 1, fqdnlen);
6705 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
6706 	}
6707 
6708 	if (idexttype) {
6709 		/* create identity extension (USERFQDN) */
6710 		struct sadb_ident *id;
6711 		int userfqdnlen;
6712 
6713 		if (userfqdn) {
6714 			/* +1 for terminating-NUL */
6715 			userfqdnlen = strlen(userfqdn) + 1;
6716 		} else
6717 			userfqdnlen = 0;
6718 		id = (struct sadb_ident *)p;
6719 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6720 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6721 		id->sadb_ident_exttype = idexttype;
6722 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
6723 		/* XXX is it correct? */
6724 		if (curproc && curproc->p_cred)
6725 			id->sadb_ident_id = curproc->p_cred->p_ruid;
6726 		if (userfqdn && userfqdnlen)
6727 			bcopy(userfqdn, id + 1, userfqdnlen);
6728 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
6729 	}
6730 #endif
6731 
6732 	/* XXX sensitivity (optional) */
6733 
6734 	/* create proposal/combination extension */
6735 	m = key_getprop(saidx);
6736 #if 0
6737 	/*
6738 	 * spec conformant: always attach proposal/combination extension,
6739 	 * the problem is that we have no way to attach it for ipcomp,
6740 	 * due to the way sadb_comb is declared in RFC2367.
6741 	 */
6742 	if (!m) {
6743 		error = ENOBUFS;
6744 		goto fail;
6745 	}
6746 	m_cat(result, m);
6747 #else
6748 	/*
6749 	 * outside of spec; make proposal/combination extension optional.
6750 	 */
6751 	if (m)
6752 		m_cat(result, m);
6753 #endif
6754 
6755 	if ((result->m_flags & M_PKTHDR) == 0) {
6756 		error = EINVAL;
6757 		goto fail;
6758 	}
6759 
6760 	if (result->m_len < sizeof(struct sadb_msg)) {
6761 		result = m_pullup(result, sizeof(struct sadb_msg));
6762 		if (result == NULL) {
6763 			error = ENOBUFS;
6764 			goto fail;
6765 		}
6766 	}
6767 
6768 	result->m_pkthdr.len = 0;
6769 	for (m = result; m; m = m->m_next)
6770 		result->m_pkthdr.len += m->m_len;
6771 
6772 	mtod(result, struct sadb_msg *)->sadb_msg_len =
6773 	    PFKEY_UNIT64(result->m_pkthdr.len);
6774 
6775 	KEYDBG(KEY_STAMP,
6776 	    printf("%s: SP(%p)\n", __func__, sp));
6777 	KEYDBG(KEY_DATA, kdebug_secasindex(saidx, NULL));
6778 
6779 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
6780 
6781  fail:
6782 	if (result)
6783 		m_freem(result);
6784 	return error;
6785 }
6786 
6787 static uint32_t
6788 key_newacq(const struct secasindex *saidx, int *perror)
6789 {
6790 	struct secacq *acq;
6791 	uint32_t seq;
6792 
6793 	acq = malloc(sizeof(*acq), M_IPSEC_SAQ, M_NOWAIT | M_ZERO);
6794 	if (acq == NULL) {
6795 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6796 		*perror = ENOBUFS;
6797 		return (0);
6798 	}
6799 
6800 	/* copy secindex */
6801 	bcopy(saidx, &acq->saidx, sizeof(acq->saidx));
6802 	acq->created = time_second;
6803 	acq->count = 0;
6804 
6805 	/* add to acqtree */
6806 	ACQ_LOCK();
6807 	seq = acq->seq = (V_acq_seq == ~0 ? 1 : ++V_acq_seq);
6808 	LIST_INSERT_HEAD(&V_acqtree, acq, chain);
6809 	LIST_INSERT_HEAD(ACQADDRHASH_HASH(saidx), acq, addrhash);
6810 	LIST_INSERT_HEAD(ACQSEQHASH_HASH(seq), acq, seqhash);
6811 	ACQ_UNLOCK();
6812 	*perror = 0;
6813 	return (seq);
6814 }
6815 
6816 static uint32_t
6817 key_getacq(const struct secasindex *saidx, int *perror)
6818 {
6819 	struct secacq *acq;
6820 	uint32_t seq;
6821 
6822 	ACQ_LOCK();
6823 	LIST_FOREACH(acq, ACQADDRHASH_HASH(saidx), addrhash) {
6824 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY)) {
6825 			if (acq->count > V_key_blockacq_count) {
6826 				/*
6827 				 * Reset counter and send message.
6828 				 * Also reset created time to keep ACQ for
6829 				 * this saidx.
6830 				 */
6831 				acq->created = time_second;
6832 				acq->count = 0;
6833 				seq = acq->seq;
6834 			} else {
6835 				/*
6836 				 * Increment counter and do nothing.
6837 				 * We send SADB_ACQUIRE message only
6838 				 * for each V_key_blockacq_count packet.
6839 				 */
6840 				acq->count++;
6841 				seq = 0;
6842 			}
6843 			break;
6844 		}
6845 	}
6846 	ACQ_UNLOCK();
6847 	if (acq != NULL) {
6848 		*perror = 0;
6849 		return (seq);
6850 	}
6851 	/* allocate new  entry */
6852 	return (key_newacq(saidx, perror));
6853 }
6854 
6855 static int
6856 key_acqreset(uint32_t seq)
6857 {
6858 	struct secacq *acq;
6859 
6860 	ACQ_LOCK();
6861 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6862 		if (acq->seq == seq) {
6863 			acq->count = 0;
6864 			acq->created = time_second;
6865 			break;
6866 		}
6867 	}
6868 	ACQ_UNLOCK();
6869 	if (acq == NULL)
6870 		return (ESRCH);
6871 	return (0);
6872 }
6873 /*
6874  * Mark ACQ entry as stale to remove it in key_flush_acq().
6875  * Called after successful SADB_GETSPI message.
6876  */
6877 static int
6878 key_acqdone(const struct secasindex *saidx, uint32_t seq)
6879 {
6880 	struct secacq *acq;
6881 
6882 	ACQ_LOCK();
6883 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6884 		if (acq->seq == seq)
6885 			break;
6886 	}
6887 	if (acq != NULL) {
6888 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY) == 0) {
6889 			ipseclog((LOG_DEBUG,
6890 			    "%s: Mismatched saidx for ACQ %u\n", __func__, seq));
6891 			acq = NULL;
6892 		} else {
6893 			acq->created = 0;
6894 		}
6895 	} else {
6896 		ipseclog((LOG_DEBUG,
6897 		    "%s: ACQ %u is not found.\n", __func__, seq));
6898 	}
6899 	ACQ_UNLOCK();
6900 	if (acq == NULL)
6901 		return (ESRCH);
6902 	return (0);
6903 }
6904 
6905 static struct secspacq *
6906 key_newspacq(struct secpolicyindex *spidx)
6907 {
6908 	struct secspacq *acq;
6909 
6910 	/* get new entry */
6911 	acq = malloc(sizeof(struct secspacq), M_IPSEC_SAQ, M_NOWAIT|M_ZERO);
6912 	if (acq == NULL) {
6913 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6914 		return NULL;
6915 	}
6916 
6917 	/* copy secindex */
6918 	bcopy(spidx, &acq->spidx, sizeof(acq->spidx));
6919 	acq->created = time_second;
6920 	acq->count = 0;
6921 
6922 	/* add to spacqtree */
6923 	SPACQ_LOCK();
6924 	LIST_INSERT_HEAD(&V_spacqtree, acq, chain);
6925 	SPACQ_UNLOCK();
6926 
6927 	return acq;
6928 }
6929 
6930 static struct secspacq *
6931 key_getspacq(struct secpolicyindex *spidx)
6932 {
6933 	struct secspacq *acq;
6934 
6935 	SPACQ_LOCK();
6936 	LIST_FOREACH(acq, &V_spacqtree, chain) {
6937 		if (key_cmpspidx_exactly(spidx, &acq->spidx)) {
6938 			/* NB: return holding spacq_lock */
6939 			return acq;
6940 		}
6941 	}
6942 	SPACQ_UNLOCK();
6943 
6944 	return NULL;
6945 }
6946 
6947 /*
6948  * SADB_ACQUIRE processing,
6949  * in first situation, is receiving
6950  *   <base>
6951  * from the ikmpd, and clear sequence of its secasvar entry.
6952  *
6953  * In second situation, is receiving
6954  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6955  * from a user land process, and return
6956  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6957  * to the socket.
6958  *
6959  * m will always be freed.
6960  */
6961 static int
6962 key_acquire2(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6963 {
6964 	SAHTREE_RLOCK_TRACKER;
6965 	struct sadb_address *src0, *dst0;
6966 	struct secasindex saidx;
6967 	struct secashead *sah;
6968 	uint32_t reqid;
6969 	int error;
6970 	uint8_t mode, proto;
6971 
6972 	IPSEC_ASSERT(so != NULL, ("null socket"));
6973 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6974 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6975 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6976 
6977 	/*
6978 	 * Error message from KMd.
6979 	 * We assume that if error was occurred in IKEd, the length of PFKEY
6980 	 * message is equal to the size of sadb_msg structure.
6981 	 * We do not raise error even if error occurred in this function.
6982 	 */
6983 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
6984 		/* check sequence number */
6985 		if (mhp->msg->sadb_msg_seq == 0 ||
6986 		    mhp->msg->sadb_msg_errno == 0) {
6987 			ipseclog((LOG_DEBUG, "%s: must specify sequence "
6988 				"number and errno.\n", __func__));
6989 		} else {
6990 			/*
6991 			 * IKEd reported that error occurred.
6992 			 * XXXAE: what it expects from the kernel?
6993 			 * Probably we should send SADB_ACQUIRE again?
6994 			 * If so, reset ACQ's state.
6995 			 * XXXAE: it looks useless.
6996 			 */
6997 			key_acqreset(mhp->msg->sadb_msg_seq);
6998 		}
6999 		m_freem(m);
7000 		return (0);
7001 	}
7002 
7003 	/*
7004 	 * This message is from user land.
7005 	 */
7006 
7007 	/* map satype to proto */
7008 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7009 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7010 		    __func__));
7011 		return key_senderror(so, m, EINVAL);
7012 	}
7013 
7014 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
7015 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
7016 	    SADB_CHECKHDR(mhp, SADB_EXT_PROPOSAL)) {
7017 		ipseclog((LOG_DEBUG,
7018 		    "%s: invalid message: missing required header.\n",
7019 		    __func__));
7020 		return key_senderror(so, m, EINVAL);
7021 	}
7022 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
7023 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
7024 	    SADB_CHECKLEN(mhp, SADB_EXT_PROPOSAL)) {
7025 		ipseclog((LOG_DEBUG,
7026 		    "%s: invalid message: wrong header size.\n", __func__));
7027 		return key_senderror(so, m, EINVAL);
7028 	}
7029 
7030 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
7031 		mode = IPSEC_MODE_ANY;
7032 		reqid = 0;
7033 	} else {
7034 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
7035 			ipseclog((LOG_DEBUG,
7036 			    "%s: invalid message: wrong header size.\n",
7037 			    __func__));
7038 			return key_senderror(so, m, EINVAL);
7039 		}
7040 		mode = ((struct sadb_x_sa2 *)
7041 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
7042 		reqid = ((struct sadb_x_sa2 *)
7043 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
7044 	}
7045 
7046 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
7047 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
7048 
7049 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
7050 	    (struct sockaddr *)(dst0 + 1));
7051 	if (error != 0) {
7052 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
7053 		return key_senderror(so, m, EINVAL);
7054 	}
7055 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
7056 
7057 	/* get a SA index */
7058 	SAHTREE_RLOCK();
7059 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
7060 		if (key_cmpsaidx(&sah->saidx, &saidx, CMP_MODE_REQID))
7061 			break;
7062 	}
7063 	SAHTREE_RUNLOCK();
7064 	if (sah != NULL) {
7065 		ipseclog((LOG_DEBUG, "%s: a SA exists already.\n", __func__));
7066 		return key_senderror(so, m, EEXIST);
7067 	}
7068 
7069 	error = key_acquire(&saidx, NULL);
7070 	if (error != 0) {
7071 		ipseclog((LOG_DEBUG,
7072 		    "%s: error %d returned from key_acquire()\n",
7073 			__func__, error));
7074 		return key_senderror(so, m, error);
7075 	}
7076 	m_freem(m);
7077 	return (0);
7078 }
7079 
7080 /*
7081  * SADB_REGISTER processing.
7082  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
7083  * receive
7084  *   <base>
7085  * from the ikmpd, and register a socket to send PF_KEY messages,
7086  * and send
7087  *   <base, supported>
7088  * to KMD by PF_KEY.
7089  * If socket is detached, must free from regnode.
7090  *
7091  * m will always be freed.
7092  */
7093 static int
7094 key_register(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7095 {
7096 	struct secreg *reg, *newreg = NULL;
7097 
7098 	IPSEC_ASSERT(so != NULL, ("null socket"));
7099 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7100 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7101 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7102 
7103 	/* check for invalid register message */
7104 	if (mhp->msg->sadb_msg_satype >= sizeof(V_regtree)/sizeof(V_regtree[0]))
7105 		return key_senderror(so, m, EINVAL);
7106 
7107 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
7108 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
7109 		goto setmsg;
7110 
7111 	/* check whether existing or not */
7112 	REGTREE_LOCK();
7113 	LIST_FOREACH(reg, &V_regtree[mhp->msg->sadb_msg_satype], chain) {
7114 		if (reg->so == so) {
7115 			REGTREE_UNLOCK();
7116 			ipseclog((LOG_DEBUG, "%s: socket exists already.\n",
7117 				__func__));
7118 			return key_senderror(so, m, EEXIST);
7119 		}
7120 	}
7121 
7122 	/* create regnode */
7123 	newreg =  malloc(sizeof(struct secreg), M_IPSEC_SAR, M_NOWAIT|M_ZERO);
7124 	if (newreg == NULL) {
7125 		REGTREE_UNLOCK();
7126 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7127 		return key_senderror(so, m, ENOBUFS);
7128 	}
7129 
7130 	newreg->so = so;
7131 	((struct keycb *)sotorawcb(so))->kp_registered++;
7132 
7133 	/* add regnode to regtree. */
7134 	LIST_INSERT_HEAD(&V_regtree[mhp->msg->sadb_msg_satype], newreg, chain);
7135 	REGTREE_UNLOCK();
7136 
7137   setmsg:
7138     {
7139 	struct mbuf *n;
7140 	struct sadb_msg *newmsg;
7141 	struct sadb_supported *sup;
7142 	u_int len, alen, elen;
7143 	int off;
7144 	int i;
7145 	struct sadb_alg *alg;
7146 
7147 	/* create new sadb_msg to reply. */
7148 	alen = 0;
7149 	for (i = 1; i <= SADB_AALG_MAX; i++) {
7150 		if (auth_algorithm_lookup(i))
7151 			alen += sizeof(struct sadb_alg);
7152 	}
7153 	if (alen)
7154 		alen += sizeof(struct sadb_supported);
7155 	elen = 0;
7156 	for (i = 1; i <= SADB_EALG_MAX; i++) {
7157 		if (enc_algorithm_lookup(i))
7158 			elen += sizeof(struct sadb_alg);
7159 	}
7160 	if (elen)
7161 		elen += sizeof(struct sadb_supported);
7162 
7163 	len = sizeof(struct sadb_msg) + alen + elen;
7164 
7165 	if (len > MCLBYTES)
7166 		return key_senderror(so, m, ENOBUFS);
7167 
7168 	MGETHDR(n, M_NOWAIT, MT_DATA);
7169 	if (n != NULL && len > MHLEN) {
7170 		if (!(MCLGET(n, M_NOWAIT))) {
7171 			m_freem(n);
7172 			n = NULL;
7173 		}
7174 	}
7175 	if (!n)
7176 		return key_senderror(so, m, ENOBUFS);
7177 
7178 	n->m_pkthdr.len = n->m_len = len;
7179 	n->m_next = NULL;
7180 	off = 0;
7181 
7182 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
7183 	newmsg = mtod(n, struct sadb_msg *);
7184 	newmsg->sadb_msg_errno = 0;
7185 	newmsg->sadb_msg_len = PFKEY_UNIT64(len);
7186 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7187 
7188 	/* for authentication algorithm */
7189 	if (alen) {
7190 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7191 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7192 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7193 		off += PFKEY_ALIGN8(sizeof(*sup));
7194 
7195 		for (i = 1; i <= SADB_AALG_MAX; i++) {
7196 			const struct auth_hash *aalgo;
7197 			u_int16_t minkeysize, maxkeysize;
7198 
7199 			aalgo = auth_algorithm_lookup(i);
7200 			if (!aalgo)
7201 				continue;
7202 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7203 			alg->sadb_alg_id = i;
7204 			alg->sadb_alg_ivlen = 0;
7205 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7206 			alg->sadb_alg_minbits = _BITS(minkeysize);
7207 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7208 			off += PFKEY_ALIGN8(sizeof(*alg));
7209 		}
7210 	}
7211 
7212 	/* for encryption algorithm */
7213 	if (elen) {
7214 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7215 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7216 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7217 		off += PFKEY_ALIGN8(sizeof(*sup));
7218 
7219 		for (i = 1; i <= SADB_EALG_MAX; i++) {
7220 			const struct enc_xform *ealgo;
7221 
7222 			ealgo = enc_algorithm_lookup(i);
7223 			if (!ealgo)
7224 				continue;
7225 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7226 			alg->sadb_alg_id = i;
7227 			alg->sadb_alg_ivlen = ealgo->ivsize;
7228 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7229 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7230 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7231 		}
7232 	}
7233 
7234 	IPSEC_ASSERT(off == len,
7235 		("length assumption failed (off %u len %u)", off, len));
7236 
7237 	m_freem(m);
7238 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7239     }
7240 }
7241 
7242 /*
7243  * free secreg entry registered.
7244  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7245  */
7246 void
7247 key_freereg(struct socket *so)
7248 {
7249 	struct secreg *reg;
7250 	int i;
7251 
7252 	IPSEC_ASSERT(so != NULL, ("NULL so"));
7253 
7254 	/*
7255 	 * check whether existing or not.
7256 	 * check all type of SA, because there is a potential that
7257 	 * one socket is registered to multiple type of SA.
7258 	 */
7259 	REGTREE_LOCK();
7260 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7261 		LIST_FOREACH(reg, &V_regtree[i], chain) {
7262 			if (reg->so == so && __LIST_CHAINED(reg)) {
7263 				LIST_REMOVE(reg, chain);
7264 				free(reg, M_IPSEC_SAR);
7265 				break;
7266 			}
7267 		}
7268 	}
7269 	REGTREE_UNLOCK();
7270 }
7271 
7272 /*
7273  * SADB_EXPIRE processing
7274  * send
7275  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7276  * to KMD by PF_KEY.
7277  * NOTE: We send only soft lifetime extension.
7278  *
7279  * OUT:	0	: succeed
7280  *	others	: error number
7281  */
7282 static int
7283 key_expire(struct secasvar *sav, int hard)
7284 {
7285 	struct mbuf *result = NULL, *m;
7286 	struct sadb_lifetime *lt;
7287 	uint32_t replay_count;
7288 	int error, len;
7289 	uint8_t satype;
7290 
7291 	IPSEC_ASSERT (sav != NULL, ("null sav"));
7292 	IPSEC_ASSERT (sav->sah != NULL, ("null sa header"));
7293 
7294 	KEYDBG(KEY_STAMP,
7295 	    printf("%s: SA(%p) expired %s lifetime\n", __func__,
7296 		sav, hard ? "hard": "soft"));
7297 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
7298 	/* set msg header */
7299 	satype = key_proto2satype(sav->sah->saidx.proto);
7300 	IPSEC_ASSERT(satype != 0, ("invalid proto, satype %u", satype));
7301 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, sav->refcnt);
7302 	if (!m) {
7303 		error = ENOBUFS;
7304 		goto fail;
7305 	}
7306 	result = m;
7307 
7308 	/* create SA extension */
7309 	m = key_setsadbsa(sav);
7310 	if (!m) {
7311 		error = ENOBUFS;
7312 		goto fail;
7313 	}
7314 	m_cat(result, m);
7315 
7316 	/* create SA extension */
7317 	SECASVAR_LOCK(sav);
7318 	replay_count = sav->replay ? sav->replay->count : 0;
7319 	SECASVAR_UNLOCK(sav);
7320 
7321 	m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
7322 			sav->sah->saidx.reqid);
7323 	if (!m) {
7324 		error = ENOBUFS;
7325 		goto fail;
7326 	}
7327 	m_cat(result, m);
7328 
7329 	if (sav->replay && sav->replay->wsize > UINT8_MAX) {
7330 		m = key_setsadbxsareplay(sav->replay->wsize);
7331 		if (!m) {
7332 			error = ENOBUFS;
7333 			goto fail;
7334 		}
7335 		m_cat(result, m);
7336 	}
7337 
7338 	/* create lifetime extension (current and soft) */
7339 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7340 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
7341 	if (m == NULL) {
7342 		error = ENOBUFS;
7343 		goto fail;
7344 	}
7345 	m_align(m, len);
7346 	m->m_len = len;
7347 	bzero(mtod(m, caddr_t), len);
7348 	lt = mtod(m, struct sadb_lifetime *);
7349 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7350 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7351 	lt->sadb_lifetime_allocations =
7352 	    (uint32_t)counter_u64_fetch(sav->lft_c_allocations);
7353 	lt->sadb_lifetime_bytes =
7354 	    counter_u64_fetch(sav->lft_c_bytes);
7355 	lt->sadb_lifetime_addtime = sav->created;
7356 	lt->sadb_lifetime_usetime = sav->firstused;
7357 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
7358 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7359 	if (hard) {
7360 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
7361 		lt->sadb_lifetime_allocations = sav->lft_h->allocations;
7362 		lt->sadb_lifetime_bytes = sav->lft_h->bytes;
7363 		lt->sadb_lifetime_addtime = sav->lft_h->addtime;
7364 		lt->sadb_lifetime_usetime = sav->lft_h->usetime;
7365 	} else {
7366 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_SOFT;
7367 		lt->sadb_lifetime_allocations = sav->lft_s->allocations;
7368 		lt->sadb_lifetime_bytes = sav->lft_s->bytes;
7369 		lt->sadb_lifetime_addtime = sav->lft_s->addtime;
7370 		lt->sadb_lifetime_usetime = sav->lft_s->usetime;
7371 	}
7372 	m_cat(result, m);
7373 
7374 	/* set sadb_address for source */
7375 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
7376 	    &sav->sah->saidx.src.sa,
7377 	    FULLMASK, IPSEC_ULPROTO_ANY);
7378 	if (!m) {
7379 		error = ENOBUFS;
7380 		goto fail;
7381 	}
7382 	m_cat(result, m);
7383 
7384 	/* set sadb_address for destination */
7385 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
7386 	    &sav->sah->saidx.dst.sa,
7387 	    FULLMASK, IPSEC_ULPROTO_ANY);
7388 	if (!m) {
7389 		error = ENOBUFS;
7390 		goto fail;
7391 	}
7392 	m_cat(result, m);
7393 
7394 	/*
7395 	 * XXX-BZ Handle NAT-T extensions here.
7396 	 * XXXAE: it doesn't seem quite useful. IKEs should not depend on
7397 	 * this information, we report only significant SA fields.
7398 	 */
7399 
7400 	if ((result->m_flags & M_PKTHDR) == 0) {
7401 		error = EINVAL;
7402 		goto fail;
7403 	}
7404 
7405 	if (result->m_len < sizeof(struct sadb_msg)) {
7406 		result = m_pullup(result, sizeof(struct sadb_msg));
7407 		if (result == NULL) {
7408 			error = ENOBUFS;
7409 			goto fail;
7410 		}
7411 	}
7412 
7413 	result->m_pkthdr.len = 0;
7414 	for (m = result; m; m = m->m_next)
7415 		result->m_pkthdr.len += m->m_len;
7416 
7417 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7418 	    PFKEY_UNIT64(result->m_pkthdr.len);
7419 
7420 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7421 
7422  fail:
7423 	if (result)
7424 		m_freem(result);
7425 	return error;
7426 }
7427 
7428 static void
7429 key_freesah_flushed(struct secashead_queue *flushq)
7430 {
7431 	struct secashead *sah, *nextsah;
7432 	struct secasvar *sav, *nextsav;
7433 
7434 	sah = TAILQ_FIRST(flushq);
7435 	while (sah != NULL) {
7436 		sav = TAILQ_FIRST(&sah->savtree_larval);
7437 		while (sav != NULL) {
7438 			nextsav = TAILQ_NEXT(sav, chain);
7439 			TAILQ_REMOVE(&sah->savtree_larval, sav, chain);
7440 			key_freesav(&sav); /* release last reference */
7441 			key_freesah(&sah); /* release reference from SAV */
7442 			sav = nextsav;
7443 		}
7444 		sav = TAILQ_FIRST(&sah->savtree_alive);
7445 		while (sav != NULL) {
7446 			nextsav = TAILQ_NEXT(sav, chain);
7447 			TAILQ_REMOVE(&sah->savtree_alive, sav, chain);
7448 			key_freesav(&sav); /* release last reference */
7449 			key_freesah(&sah); /* release reference from SAV */
7450 			sav = nextsav;
7451 		}
7452 		nextsah = TAILQ_NEXT(sah, chain);
7453 		key_freesah(&sah);	/* release last reference */
7454 		sah = nextsah;
7455 	}
7456 }
7457 
7458 /*
7459  * SADB_FLUSH processing
7460  * receive
7461  *   <base>
7462  * from the ikmpd, and free all entries in secastree.
7463  * and send,
7464  *   <base>
7465  * to the ikmpd.
7466  * NOTE: to do is only marking SADB_SASTATE_DEAD.
7467  *
7468  * m will always be freed.
7469  */
7470 static int
7471 key_flush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7472 {
7473 	struct secashead_queue flushq;
7474 	struct sadb_msg *newmsg;
7475 	struct secashead *sah, *nextsah;
7476 	struct secasvar *sav;
7477 	uint8_t proto;
7478 	int i;
7479 
7480 	IPSEC_ASSERT(so != NULL, ("null socket"));
7481 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7482 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7483 
7484 	/* map satype to proto */
7485 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7486 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7487 			__func__));
7488 		return key_senderror(so, m, EINVAL);
7489 	}
7490 	KEYDBG(KEY_STAMP,
7491 	    printf("%s: proto %u\n", __func__, proto));
7492 
7493 	TAILQ_INIT(&flushq);
7494 	if (proto == IPSEC_PROTO_ANY) {
7495 		/* no SATYPE specified, i.e. flushing all SA. */
7496 		SAHTREE_WLOCK();
7497 		/* Move all SAHs into flushq */
7498 		TAILQ_CONCAT(&flushq, &V_sahtree, chain);
7499 		/* Flush all buckets in SPI hash */
7500 		for (i = 0; i < V_savhash_mask + 1; i++)
7501 			LIST_INIT(&V_savhashtbl[i]);
7502 		/* Flush all buckets in SAHADDRHASH */
7503 		for (i = 0; i < V_sahaddrhash_mask + 1; i++)
7504 			LIST_INIT(&V_sahaddrhashtbl[i]);
7505 		/* Mark all SAHs as unlinked */
7506 		TAILQ_FOREACH(sah, &flushq, chain) {
7507 			sah->state = SADB_SASTATE_DEAD;
7508 			/*
7509 			 * Callout handler makes its job using
7510 			 * RLOCK and drain queues. In case, when this
7511 			 * function will be called just before it
7512 			 * acquires WLOCK, we need to mark SAs as
7513 			 * unlinked to prevent second unlink.
7514 			 */
7515 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7516 				sav->state = SADB_SASTATE_DEAD;
7517 			}
7518 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7519 				sav->state = SADB_SASTATE_DEAD;
7520 			}
7521 		}
7522 		SAHTREE_WUNLOCK();
7523 	} else {
7524 		SAHTREE_WLOCK();
7525 		sah = TAILQ_FIRST(&V_sahtree);
7526 		while (sah != NULL) {
7527 			IPSEC_ASSERT(sah->state != SADB_SASTATE_DEAD,
7528 			    ("DEAD SAH %p in SADB_FLUSH", sah));
7529 			nextsah = TAILQ_NEXT(sah, chain);
7530 			if (sah->saidx.proto != proto) {
7531 				sah = nextsah;
7532 				continue;
7533 			}
7534 			sah->state = SADB_SASTATE_DEAD;
7535 			TAILQ_REMOVE(&V_sahtree, sah, chain);
7536 			LIST_REMOVE(sah, addrhash);
7537 			/* Unlink all SAs from SPI hash */
7538 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7539 				LIST_REMOVE(sav, spihash);
7540 				sav->state = SADB_SASTATE_DEAD;
7541 			}
7542 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7543 				LIST_REMOVE(sav, spihash);
7544 				sav->state = SADB_SASTATE_DEAD;
7545 			}
7546 			/* Add SAH into flushq */
7547 			TAILQ_INSERT_HEAD(&flushq, sah, chain);
7548 			sah = nextsah;
7549 		}
7550 		SAHTREE_WUNLOCK();
7551 	}
7552 
7553 	key_freesah_flushed(&flushq);
7554 	/* Free all queued SAs and SAHs */
7555 	if (m->m_len < sizeof(struct sadb_msg) ||
7556 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7557 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7558 		return key_senderror(so, m, ENOBUFS);
7559 	}
7560 
7561 	if (m->m_next)
7562 		m_freem(m->m_next);
7563 	m->m_next = NULL;
7564 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7565 	newmsg = mtod(m, struct sadb_msg *);
7566 	newmsg->sadb_msg_errno = 0;
7567 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7568 
7569 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7570 }
7571 
7572 /*
7573  * SADB_DUMP processing
7574  * dump all entries including status of DEAD in SAD.
7575  * receive
7576  *   <base>
7577  * from the ikmpd, and dump all secasvar leaves
7578  * and send,
7579  *   <base> .....
7580  * to the ikmpd.
7581  *
7582  * m will always be freed.
7583  */
7584 static int
7585 key_dump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7586 {
7587 	SAHTREE_RLOCK_TRACKER;
7588 	struct secashead *sah;
7589 	struct secasvar *sav;
7590 	struct mbuf *n;
7591 	uint32_t cnt;
7592 	uint8_t proto, satype;
7593 
7594 	IPSEC_ASSERT(so != NULL, ("null socket"));
7595 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7596 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7597 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7598 
7599 	/* map satype to proto */
7600 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7601 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7602 		    __func__));
7603 		return key_senderror(so, m, EINVAL);
7604 	}
7605 
7606 	/* count sav entries to be sent to the userland. */
7607 	cnt = 0;
7608 	SAHTREE_RLOCK();
7609 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7610 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7611 		    proto != sah->saidx.proto)
7612 			continue;
7613 
7614 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain)
7615 			cnt++;
7616 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain)
7617 			cnt++;
7618 	}
7619 
7620 	if (cnt == 0) {
7621 		SAHTREE_RUNLOCK();
7622 		return key_senderror(so, m, ENOENT);
7623 	}
7624 
7625 	/* send this to the userland, one at a time. */
7626 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7627 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7628 		    proto != sah->saidx.proto)
7629 			continue;
7630 
7631 		/* map proto to satype */
7632 		if ((satype = key_proto2satype(sah->saidx.proto)) == 0) {
7633 			SAHTREE_RUNLOCK();
7634 			ipseclog((LOG_DEBUG, "%s: there was invalid proto in "
7635 			    "SAD.\n", __func__));
7636 			return key_senderror(so, m, EINVAL);
7637 		}
7638 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7639 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7640 			    --cnt, mhp->msg->sadb_msg_pid);
7641 			if (n == NULL) {
7642 				SAHTREE_RUNLOCK();
7643 				return key_senderror(so, m, ENOBUFS);
7644 			}
7645 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7646 		}
7647 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7648 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7649 			    --cnt, mhp->msg->sadb_msg_pid);
7650 			if (n == NULL) {
7651 				SAHTREE_RUNLOCK();
7652 				return key_senderror(so, m, ENOBUFS);
7653 			}
7654 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7655 		}
7656 	}
7657 	SAHTREE_RUNLOCK();
7658 	m_freem(m);
7659 	return (0);
7660 }
7661 /*
7662  * SADB_X_PROMISC processing
7663  *
7664  * m will always be freed.
7665  */
7666 static int
7667 key_promisc(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7668 {
7669 	int olen;
7670 
7671 	IPSEC_ASSERT(so != NULL, ("null socket"));
7672 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7673 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7674 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7675 
7676 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7677 
7678 	if (olen < sizeof(struct sadb_msg)) {
7679 #if 1
7680 		return key_senderror(so, m, EINVAL);
7681 #else
7682 		m_freem(m);
7683 		return 0;
7684 #endif
7685 	} else if (olen == sizeof(struct sadb_msg)) {
7686 		/* enable/disable promisc mode */
7687 		struct keycb *kp;
7688 
7689 		if ((kp = (struct keycb *)sotorawcb(so)) == NULL)
7690 			return key_senderror(so, m, EINVAL);
7691 		mhp->msg->sadb_msg_errno = 0;
7692 		switch (mhp->msg->sadb_msg_satype) {
7693 		case 0:
7694 		case 1:
7695 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
7696 			break;
7697 		default:
7698 			return key_senderror(so, m, EINVAL);
7699 		}
7700 
7701 		/* send the original message back to everyone */
7702 		mhp->msg->sadb_msg_errno = 0;
7703 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7704 	} else {
7705 		/* send packet as is */
7706 
7707 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
7708 
7709 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
7710 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7711 	}
7712 }
7713 
7714 static int (*key_typesw[])(struct socket *, struct mbuf *,
7715 		const struct sadb_msghdr *) = {
7716 	NULL,		/* SADB_RESERVED */
7717 	key_getspi,	/* SADB_GETSPI */
7718 	key_update,	/* SADB_UPDATE */
7719 	key_add,	/* SADB_ADD */
7720 	key_delete,	/* SADB_DELETE */
7721 	key_get,	/* SADB_GET */
7722 	key_acquire2,	/* SADB_ACQUIRE */
7723 	key_register,	/* SADB_REGISTER */
7724 	NULL,		/* SADB_EXPIRE */
7725 	key_flush,	/* SADB_FLUSH */
7726 	key_dump,	/* SADB_DUMP */
7727 	key_promisc,	/* SADB_X_PROMISC */
7728 	NULL,		/* SADB_X_PCHANGE */
7729 	key_spdadd,	/* SADB_X_SPDUPDATE */
7730 	key_spdadd,	/* SADB_X_SPDADD */
7731 	key_spddelete,	/* SADB_X_SPDDELETE */
7732 	key_spdget,	/* SADB_X_SPDGET */
7733 	NULL,		/* SADB_X_SPDACQUIRE */
7734 	key_spddump,	/* SADB_X_SPDDUMP */
7735 	key_spdflush,	/* SADB_X_SPDFLUSH */
7736 	key_spdadd,	/* SADB_X_SPDSETIDX */
7737 	NULL,		/* SADB_X_SPDEXPIRE */
7738 	key_spddelete2,	/* SADB_X_SPDDELETE2 */
7739 };
7740 
7741 /*
7742  * parse sadb_msg buffer to process PFKEYv2,
7743  * and create a data to response if needed.
7744  * I think to be dealed with mbuf directly.
7745  * IN:
7746  *     msgp  : pointer to pointer to a received buffer pulluped.
7747  *             This is rewrited to response.
7748  *     so    : pointer to socket.
7749  * OUT:
7750  *    length for buffer to send to user process.
7751  */
7752 int
7753 key_parse(struct mbuf *m, struct socket *so)
7754 {
7755 	struct sadb_msg *msg;
7756 	struct sadb_msghdr mh;
7757 	u_int orglen;
7758 	int error;
7759 	int target;
7760 
7761 	IPSEC_ASSERT(so != NULL, ("null socket"));
7762 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7763 
7764 	if (m->m_len < sizeof(struct sadb_msg)) {
7765 		m = m_pullup(m, sizeof(struct sadb_msg));
7766 		if (!m)
7767 			return ENOBUFS;
7768 	}
7769 	msg = mtod(m, struct sadb_msg *);
7770 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
7771 	target = KEY_SENDUP_ONE;
7772 
7773 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len != orglen) {
7774 		ipseclog((LOG_DEBUG, "%s: invalid message length.\n",__func__));
7775 		PFKEYSTAT_INC(out_invlen);
7776 		error = EINVAL;
7777 		goto senderror;
7778 	}
7779 
7780 	if (msg->sadb_msg_version != PF_KEY_V2) {
7781 		ipseclog((LOG_DEBUG, "%s: PF_KEY version %u is mismatched.\n",
7782 		    __func__, msg->sadb_msg_version));
7783 		PFKEYSTAT_INC(out_invver);
7784 		error = EINVAL;
7785 		goto senderror;
7786 	}
7787 
7788 	if (msg->sadb_msg_type > SADB_MAX) {
7789 		ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7790 		    __func__, msg->sadb_msg_type));
7791 		PFKEYSTAT_INC(out_invmsgtype);
7792 		error = EINVAL;
7793 		goto senderror;
7794 	}
7795 
7796 	/* for old-fashioned code - should be nuked */
7797 	if (m->m_pkthdr.len > MCLBYTES) {
7798 		m_freem(m);
7799 		return ENOBUFS;
7800 	}
7801 	if (m->m_next) {
7802 		struct mbuf *n;
7803 
7804 		MGETHDR(n, M_NOWAIT, MT_DATA);
7805 		if (n && m->m_pkthdr.len > MHLEN) {
7806 			if (!(MCLGET(n, M_NOWAIT))) {
7807 				m_free(n);
7808 				n = NULL;
7809 			}
7810 		}
7811 		if (!n) {
7812 			m_freem(m);
7813 			return ENOBUFS;
7814 		}
7815 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
7816 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
7817 		n->m_next = NULL;
7818 		m_freem(m);
7819 		m = n;
7820 	}
7821 
7822 	/* align the mbuf chain so that extensions are in contiguous region. */
7823 	error = key_align(m, &mh);
7824 	if (error)
7825 		return error;
7826 
7827 	msg = mh.msg;
7828 
7829 	/* We use satype as scope mask for spddump */
7830 	if (msg->sadb_msg_type == SADB_X_SPDDUMP) {
7831 		switch (msg->sadb_msg_satype) {
7832 		case IPSEC_POLICYSCOPE_ANY:
7833 		case IPSEC_POLICYSCOPE_GLOBAL:
7834 		case IPSEC_POLICYSCOPE_IFNET:
7835 		case IPSEC_POLICYSCOPE_PCB:
7836 			break;
7837 		default:
7838 			ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7839 			    __func__, msg->sadb_msg_type));
7840 			PFKEYSTAT_INC(out_invsatype);
7841 			error = EINVAL;
7842 			goto senderror;
7843 		}
7844 	} else {
7845 		switch (msg->sadb_msg_satype) { /* check SA type */
7846 		case SADB_SATYPE_UNSPEC:
7847 			switch (msg->sadb_msg_type) {
7848 			case SADB_GETSPI:
7849 			case SADB_UPDATE:
7850 			case SADB_ADD:
7851 			case SADB_DELETE:
7852 			case SADB_GET:
7853 			case SADB_ACQUIRE:
7854 			case SADB_EXPIRE:
7855 				ipseclog((LOG_DEBUG, "%s: must specify satype "
7856 				    "when msg type=%u.\n", __func__,
7857 				    msg->sadb_msg_type));
7858 				PFKEYSTAT_INC(out_invsatype);
7859 				error = EINVAL;
7860 				goto senderror;
7861 			}
7862 			break;
7863 		case SADB_SATYPE_AH:
7864 		case SADB_SATYPE_ESP:
7865 		case SADB_X_SATYPE_IPCOMP:
7866 		case SADB_X_SATYPE_TCPSIGNATURE:
7867 			switch (msg->sadb_msg_type) {
7868 			case SADB_X_SPDADD:
7869 			case SADB_X_SPDDELETE:
7870 			case SADB_X_SPDGET:
7871 			case SADB_X_SPDFLUSH:
7872 			case SADB_X_SPDSETIDX:
7873 			case SADB_X_SPDUPDATE:
7874 			case SADB_X_SPDDELETE2:
7875 				ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7876 				    __func__, msg->sadb_msg_type));
7877 				PFKEYSTAT_INC(out_invsatype);
7878 				error = EINVAL;
7879 				goto senderror;
7880 			}
7881 			break;
7882 		case SADB_SATYPE_RSVP:
7883 		case SADB_SATYPE_OSPFV2:
7884 		case SADB_SATYPE_RIPV2:
7885 		case SADB_SATYPE_MIP:
7886 			ipseclog((LOG_DEBUG, "%s: type %u isn't supported.\n",
7887 			    __func__, msg->sadb_msg_satype));
7888 			PFKEYSTAT_INC(out_invsatype);
7889 			error = EOPNOTSUPP;
7890 			goto senderror;
7891 		case 1:	/* XXX: What does it do? */
7892 			if (msg->sadb_msg_type == SADB_X_PROMISC)
7893 				break;
7894 			/*FALLTHROUGH*/
7895 		default:
7896 			ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7897 			    __func__, msg->sadb_msg_satype));
7898 			PFKEYSTAT_INC(out_invsatype);
7899 			error = EINVAL;
7900 			goto senderror;
7901 		}
7902 	}
7903 
7904 	/* check field of upper layer protocol and address family */
7905 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL
7906 	 && mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
7907 		struct sadb_address *src0, *dst0;
7908 		u_int plen;
7909 
7910 		src0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_SRC]);
7911 		dst0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_DST]);
7912 
7913 		/* check upper layer protocol */
7914 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
7915 			ipseclog((LOG_DEBUG, "%s: upper layer protocol "
7916 				"mismatched.\n", __func__));
7917 			PFKEYSTAT_INC(out_invaddr);
7918 			error = EINVAL;
7919 			goto senderror;
7920 		}
7921 
7922 		/* check family */
7923 		if (PFKEY_ADDR_SADDR(src0)->sa_family !=
7924 		    PFKEY_ADDR_SADDR(dst0)->sa_family) {
7925 			ipseclog((LOG_DEBUG, "%s: address family mismatched.\n",
7926 				__func__));
7927 			PFKEYSTAT_INC(out_invaddr);
7928 			error = EINVAL;
7929 			goto senderror;
7930 		}
7931 		if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7932 		    PFKEY_ADDR_SADDR(dst0)->sa_len) {
7933 			ipseclog((LOG_DEBUG, "%s: address struct size "
7934 				"mismatched.\n", __func__));
7935 			PFKEYSTAT_INC(out_invaddr);
7936 			error = EINVAL;
7937 			goto senderror;
7938 		}
7939 
7940 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7941 		case AF_INET:
7942 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7943 			    sizeof(struct sockaddr_in)) {
7944 				PFKEYSTAT_INC(out_invaddr);
7945 				error = EINVAL;
7946 				goto senderror;
7947 			}
7948 			break;
7949 		case AF_INET6:
7950 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7951 			    sizeof(struct sockaddr_in6)) {
7952 				PFKEYSTAT_INC(out_invaddr);
7953 				error = EINVAL;
7954 				goto senderror;
7955 			}
7956 			break;
7957 		default:
7958 			ipseclog((LOG_DEBUG, "%s: unsupported address family\n",
7959 				__func__));
7960 			PFKEYSTAT_INC(out_invaddr);
7961 			error = EAFNOSUPPORT;
7962 			goto senderror;
7963 		}
7964 
7965 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7966 		case AF_INET:
7967 			plen = sizeof(struct in_addr) << 3;
7968 			break;
7969 		case AF_INET6:
7970 			plen = sizeof(struct in6_addr) << 3;
7971 			break;
7972 		default:
7973 			plen = 0;	/*fool gcc*/
7974 			break;
7975 		}
7976 
7977 		/* check max prefix length */
7978 		if (src0->sadb_address_prefixlen > plen ||
7979 		    dst0->sadb_address_prefixlen > plen) {
7980 			ipseclog((LOG_DEBUG, "%s: illegal prefixlen.\n",
7981 				__func__));
7982 			PFKEYSTAT_INC(out_invaddr);
7983 			error = EINVAL;
7984 			goto senderror;
7985 		}
7986 
7987 		/*
7988 		 * prefixlen == 0 is valid because there can be a case when
7989 		 * all addresses are matched.
7990 		 */
7991 	}
7992 
7993 	if (msg->sadb_msg_type >= nitems(key_typesw) ||
7994 	    key_typesw[msg->sadb_msg_type] == NULL) {
7995 		PFKEYSTAT_INC(out_invmsgtype);
7996 		error = EINVAL;
7997 		goto senderror;
7998 	}
7999 
8000 	return (*key_typesw[msg->sadb_msg_type])(so, m, &mh);
8001 
8002 senderror:
8003 	msg->sadb_msg_errno = error;
8004 	return key_sendup_mbuf(so, m, target);
8005 }
8006 
8007 static int
8008 key_senderror(struct socket *so, struct mbuf *m, int code)
8009 {
8010 	struct sadb_msg *msg;
8011 
8012 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8013 		("mbuf too small, len %u", m->m_len));
8014 
8015 	msg = mtod(m, struct sadb_msg *);
8016 	msg->sadb_msg_errno = code;
8017 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
8018 }
8019 
8020 /*
8021  * set the pointer to each header into message buffer.
8022  * m will be freed on error.
8023  * XXX larger-than-MCLBYTES extension?
8024  */
8025 static int
8026 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
8027 {
8028 	struct mbuf *n;
8029 	struct sadb_ext *ext;
8030 	size_t off, end;
8031 	int extlen;
8032 	int toff;
8033 
8034 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
8035 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
8036 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8037 		("mbuf too small, len %u", m->m_len));
8038 
8039 	/* initialize */
8040 	bzero(mhp, sizeof(*mhp));
8041 
8042 	mhp->msg = mtod(m, struct sadb_msg *);
8043 	mhp->ext[0] = (struct sadb_ext *)mhp->msg;	/*XXX backward compat */
8044 
8045 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
8046 	extlen = end;	/*just in case extlen is not updated*/
8047 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
8048 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
8049 		if (!n) {
8050 			/* m is already freed */
8051 			return ENOBUFS;
8052 		}
8053 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8054 
8055 		/* set pointer */
8056 		switch (ext->sadb_ext_type) {
8057 		case SADB_EXT_SA:
8058 		case SADB_EXT_ADDRESS_SRC:
8059 		case SADB_EXT_ADDRESS_DST:
8060 		case SADB_EXT_ADDRESS_PROXY:
8061 		case SADB_EXT_LIFETIME_CURRENT:
8062 		case SADB_EXT_LIFETIME_HARD:
8063 		case SADB_EXT_LIFETIME_SOFT:
8064 		case SADB_EXT_KEY_AUTH:
8065 		case SADB_EXT_KEY_ENCRYPT:
8066 		case SADB_EXT_IDENTITY_SRC:
8067 		case SADB_EXT_IDENTITY_DST:
8068 		case SADB_EXT_SENSITIVITY:
8069 		case SADB_EXT_PROPOSAL:
8070 		case SADB_EXT_SUPPORTED_AUTH:
8071 		case SADB_EXT_SUPPORTED_ENCRYPT:
8072 		case SADB_EXT_SPIRANGE:
8073 		case SADB_X_EXT_POLICY:
8074 		case SADB_X_EXT_SA2:
8075 		case SADB_X_EXT_NAT_T_TYPE:
8076 		case SADB_X_EXT_NAT_T_SPORT:
8077 		case SADB_X_EXT_NAT_T_DPORT:
8078 		case SADB_X_EXT_NAT_T_OAI:
8079 		case SADB_X_EXT_NAT_T_OAR:
8080 		case SADB_X_EXT_NAT_T_FRAG:
8081 		case SADB_X_EXT_SA_REPLAY:
8082 		case SADB_X_EXT_NEW_ADDRESS_SRC:
8083 		case SADB_X_EXT_NEW_ADDRESS_DST:
8084 			/* duplicate check */
8085 			/*
8086 			 * XXX Are there duplication payloads of either
8087 			 * KEY_AUTH or KEY_ENCRYPT ?
8088 			 */
8089 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
8090 				ipseclog((LOG_DEBUG, "%s: duplicate ext_type "
8091 					"%u\n", __func__, ext->sadb_ext_type));
8092 				m_freem(m);
8093 				PFKEYSTAT_INC(out_dupext);
8094 				return EINVAL;
8095 			}
8096 			break;
8097 		default:
8098 			ipseclog((LOG_DEBUG, "%s: invalid ext_type %u\n",
8099 				__func__, ext->sadb_ext_type));
8100 			m_freem(m);
8101 			PFKEYSTAT_INC(out_invexttype);
8102 			return EINVAL;
8103 		}
8104 
8105 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
8106 
8107 		if (key_validate_ext(ext, extlen)) {
8108 			m_freem(m);
8109 			PFKEYSTAT_INC(out_invlen);
8110 			return EINVAL;
8111 		}
8112 
8113 		n = m_pulldown(m, off, extlen, &toff);
8114 		if (!n) {
8115 			/* m is already freed */
8116 			return ENOBUFS;
8117 		}
8118 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8119 
8120 		mhp->ext[ext->sadb_ext_type] = ext;
8121 		mhp->extoff[ext->sadb_ext_type] = off;
8122 		mhp->extlen[ext->sadb_ext_type] = extlen;
8123 	}
8124 
8125 	if (off != end) {
8126 		m_freem(m);
8127 		PFKEYSTAT_INC(out_invlen);
8128 		return EINVAL;
8129 	}
8130 
8131 	return 0;
8132 }
8133 
8134 static int
8135 key_validate_ext(const struct sadb_ext *ext, int len)
8136 {
8137 	const struct sockaddr *sa;
8138 	enum { NONE, ADDR } checktype = NONE;
8139 	int baselen = 0;
8140 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
8141 
8142 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
8143 		return EINVAL;
8144 
8145 	/* if it does not match minimum/maximum length, bail */
8146 	if (ext->sadb_ext_type >= nitems(minsize) ||
8147 	    ext->sadb_ext_type >= nitems(maxsize))
8148 		return EINVAL;
8149 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
8150 		return EINVAL;
8151 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
8152 		return EINVAL;
8153 
8154 	/* more checks based on sadb_ext_type XXX need more */
8155 	switch (ext->sadb_ext_type) {
8156 	case SADB_EXT_ADDRESS_SRC:
8157 	case SADB_EXT_ADDRESS_DST:
8158 	case SADB_EXT_ADDRESS_PROXY:
8159 	case SADB_X_EXT_NAT_T_OAI:
8160 	case SADB_X_EXT_NAT_T_OAR:
8161 	case SADB_X_EXT_NEW_ADDRESS_SRC:
8162 	case SADB_X_EXT_NEW_ADDRESS_DST:
8163 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
8164 		checktype = ADDR;
8165 		break;
8166 	case SADB_EXT_IDENTITY_SRC:
8167 	case SADB_EXT_IDENTITY_DST:
8168 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
8169 		    SADB_X_IDENTTYPE_ADDR) {
8170 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
8171 			checktype = ADDR;
8172 		} else
8173 			checktype = NONE;
8174 		break;
8175 	default:
8176 		checktype = NONE;
8177 		break;
8178 	}
8179 
8180 	switch (checktype) {
8181 	case NONE:
8182 		break;
8183 	case ADDR:
8184 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8185 		if (len < baselen + sal)
8186 			return EINVAL;
8187 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8188 			return EINVAL;
8189 		break;
8190 	}
8191 
8192 	return 0;
8193 }
8194 
8195 void
8196 spdcache_init(void)
8197 {
8198 	int i;
8199 
8200 	TUNABLE_INT_FETCH("net.key.spdcache.maxentries",
8201 	    &V_key_spdcache_maxentries);
8202 	TUNABLE_INT_FETCH("net.key.spdcache.threshold",
8203 	    &V_key_spdcache_threshold);
8204 
8205 	if (V_key_spdcache_maxentries) {
8206 		V_key_spdcache_maxentries = MAX(V_key_spdcache_maxentries,
8207 		    SPDCACHE_MAX_ENTRIES_PER_HASH);
8208 		V_spdcachehashtbl = hashinit(V_key_spdcache_maxentries /
8209 		    SPDCACHE_MAX_ENTRIES_PER_HASH,
8210 		    M_IPSEC_SPDCACHE, &V_spdcachehash_mask);
8211 		V_key_spdcache_maxentries = (V_spdcachehash_mask + 1)
8212 		    * SPDCACHE_MAX_ENTRIES_PER_HASH;
8213 
8214 		V_spdcache_lock = malloc(sizeof(struct mtx) *
8215 		    (V_spdcachehash_mask + 1),
8216 		    M_IPSEC_SPDCACHE, M_WAITOK|M_ZERO);
8217 
8218 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8219 			SPDCACHE_LOCK_INIT(i);
8220 	}
8221 }
8222 
8223 struct spdcache_entry *
8224 spdcache_entry_alloc(const struct secpolicyindex *spidx, struct secpolicy *sp)
8225 {
8226 	struct spdcache_entry *entry;
8227 
8228 	entry = malloc(sizeof(struct spdcache_entry),
8229 		    M_IPSEC_SPDCACHE, M_NOWAIT|M_ZERO);
8230 	if (entry == NULL)
8231 		return NULL;
8232 
8233 	if (sp != NULL)
8234 		SP_ADDREF(sp);
8235 
8236 	entry->spidx = *spidx;
8237 	entry->sp = sp;
8238 
8239 	return (entry);
8240 }
8241 
8242 void
8243 spdcache_entry_free(struct spdcache_entry *entry)
8244 {
8245 
8246 	if (entry->sp != NULL)
8247 		key_freesp(&entry->sp);
8248 	free(entry, M_IPSEC_SPDCACHE);
8249 }
8250 
8251 void
8252 spdcache_clear(void)
8253 {
8254 	struct spdcache_entry *entry;
8255 	int i;
8256 
8257 	for (i = 0; i < V_spdcachehash_mask + 1; ++i) {
8258 		SPDCACHE_LOCK(i);
8259 		while (!LIST_EMPTY(&V_spdcachehashtbl[i])) {
8260 			entry = LIST_FIRST(&V_spdcachehashtbl[i]);
8261 			LIST_REMOVE(entry, chain);
8262 			spdcache_entry_free(entry);
8263 		}
8264 		SPDCACHE_UNLOCK(i);
8265 	}
8266 }
8267 
8268 #ifdef VIMAGE
8269 void
8270 spdcache_destroy(void)
8271 {
8272 	int i;
8273 
8274 	if (SPDCACHE_ENABLED()) {
8275 		spdcache_clear();
8276 		hashdestroy(V_spdcachehashtbl, M_IPSEC_SPDCACHE, V_spdcachehash_mask);
8277 
8278 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8279 			SPDCACHE_LOCK_DESTROY(i);
8280 
8281 		free(V_spdcache_lock, M_IPSEC_SPDCACHE);
8282 	}
8283 }
8284 #endif
8285 void
8286 key_init(void)
8287 {
8288 	int i;
8289 
8290 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8291 		TAILQ_INIT(&V_sptree[i]);
8292 		TAILQ_INIT(&V_sptree_ifnet[i]);
8293 	}
8294 
8295 	V_key_lft_zone = uma_zcreate("IPsec SA lft_c",
8296 	    sizeof(uint64_t) * 2, NULL, NULL, NULL, NULL,
8297 	    UMA_ALIGN_PTR, UMA_ZONE_PCPU);
8298 
8299 	TAILQ_INIT(&V_sahtree);
8300 	V_sphashtbl = hashinit(SPHASH_NHASH, M_IPSEC_SP, &V_sphash_mask);
8301 	V_savhashtbl = hashinit(SAVHASH_NHASH, M_IPSEC_SA, &V_savhash_mask);
8302 	V_sahaddrhashtbl = hashinit(SAHHASH_NHASH, M_IPSEC_SAH,
8303 	    &V_sahaddrhash_mask);
8304 	V_acqaddrhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8305 	    &V_acqaddrhash_mask);
8306 	V_acqseqhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8307 	    &V_acqseqhash_mask);
8308 
8309 	spdcache_init();
8310 
8311 	for (i = 0; i <= SADB_SATYPE_MAX; i++)
8312 		LIST_INIT(&V_regtree[i]);
8313 
8314 	LIST_INIT(&V_acqtree);
8315 	LIST_INIT(&V_spacqtree);
8316 
8317 	if (!IS_DEFAULT_VNET(curvnet))
8318 		return;
8319 
8320 	SPTREE_LOCK_INIT();
8321 	REGTREE_LOCK_INIT();
8322 	SAHTREE_LOCK_INIT();
8323 	ACQ_LOCK_INIT();
8324 	SPACQ_LOCK_INIT();
8325 
8326 #ifndef IPSEC_DEBUG2
8327 	callout_init(&key_timer, 1);
8328 	callout_reset(&key_timer, hz, key_timehandler, NULL);
8329 #endif /*IPSEC_DEBUG2*/
8330 
8331 	/* initialize key statistics */
8332 	keystat.getspi_count = 1;
8333 
8334 	if (bootverbose)
8335 		printf("IPsec: Initialized Security Association Processing.\n");
8336 }
8337 
8338 #ifdef VIMAGE
8339 void
8340 key_destroy(void)
8341 {
8342 	struct secashead_queue sahdrainq;
8343 	struct secpolicy_queue drainq;
8344 	struct secpolicy *sp, *nextsp;
8345 	struct secacq *acq, *nextacq;
8346 	struct secspacq *spacq, *nextspacq;
8347 	struct secashead *sah;
8348 	struct secasvar *sav;
8349 	struct secreg *reg;
8350 	int i;
8351 
8352 	/*
8353 	 * XXX: can we just call free() for each object without
8354 	 * walking through safe way with releasing references?
8355 	 */
8356 	TAILQ_INIT(&drainq);
8357 	SPTREE_WLOCK();
8358 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8359 		TAILQ_CONCAT(&drainq, &V_sptree[i], chain);
8360 		TAILQ_CONCAT(&drainq, &V_sptree_ifnet[i], chain);
8361 	}
8362 	for (i = 0; i < V_sphash_mask + 1; i++)
8363 		LIST_INIT(&V_sphashtbl[i]);
8364 	SPTREE_WUNLOCK();
8365 	spdcache_destroy();
8366 
8367 	sp = TAILQ_FIRST(&drainq);
8368 	while (sp != NULL) {
8369 		nextsp = TAILQ_NEXT(sp, chain);
8370 		key_freesp(&sp);
8371 		sp = nextsp;
8372 	}
8373 
8374 	TAILQ_INIT(&sahdrainq);
8375 	SAHTREE_WLOCK();
8376 	TAILQ_CONCAT(&sahdrainq, &V_sahtree, chain);
8377 	for (i = 0; i < V_savhash_mask + 1; i++)
8378 		LIST_INIT(&V_savhashtbl[i]);
8379 	for (i = 0; i < V_sahaddrhash_mask + 1; i++)
8380 		LIST_INIT(&V_sahaddrhashtbl[i]);
8381 	TAILQ_FOREACH(sah, &sahdrainq, chain) {
8382 		sah->state = SADB_SASTATE_DEAD;
8383 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
8384 			sav->state = SADB_SASTATE_DEAD;
8385 		}
8386 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
8387 			sav->state = SADB_SASTATE_DEAD;
8388 		}
8389 	}
8390 	SAHTREE_WUNLOCK();
8391 
8392 	key_freesah_flushed(&sahdrainq);
8393 	hashdestroy(V_sphashtbl, M_IPSEC_SP, V_sphash_mask);
8394 	hashdestroy(V_savhashtbl, M_IPSEC_SA, V_savhash_mask);
8395 	hashdestroy(V_sahaddrhashtbl, M_IPSEC_SAH, V_sahaddrhash_mask);
8396 
8397 	REGTREE_LOCK();
8398 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8399 		LIST_FOREACH(reg, &V_regtree[i], chain) {
8400 			if (__LIST_CHAINED(reg)) {
8401 				LIST_REMOVE(reg, chain);
8402 				free(reg, M_IPSEC_SAR);
8403 				break;
8404 			}
8405 		}
8406 	}
8407 	REGTREE_UNLOCK();
8408 
8409 	ACQ_LOCK();
8410 	acq = LIST_FIRST(&V_acqtree);
8411 	while (acq != NULL) {
8412 		nextacq = LIST_NEXT(acq, chain);
8413 		LIST_REMOVE(acq, chain);
8414 		free(acq, M_IPSEC_SAQ);
8415 		acq = nextacq;
8416 	}
8417 	for (i = 0; i < V_acqaddrhash_mask + 1; i++)
8418 		LIST_INIT(&V_acqaddrhashtbl[i]);
8419 	for (i = 0; i < V_acqseqhash_mask + 1; i++)
8420 		LIST_INIT(&V_acqseqhashtbl[i]);
8421 	ACQ_UNLOCK();
8422 
8423 	SPACQ_LOCK();
8424 	for (spacq = LIST_FIRST(&V_spacqtree); spacq != NULL;
8425 	    spacq = nextspacq) {
8426 		nextspacq = LIST_NEXT(spacq, chain);
8427 		if (__LIST_CHAINED(spacq)) {
8428 			LIST_REMOVE(spacq, chain);
8429 			free(spacq, M_IPSEC_SAQ);
8430 		}
8431 	}
8432 	SPACQ_UNLOCK();
8433 	hashdestroy(V_acqaddrhashtbl, M_IPSEC_SAQ, V_acqaddrhash_mask);
8434 	hashdestroy(V_acqseqhashtbl, M_IPSEC_SAQ, V_acqseqhash_mask);
8435 	uma_zdestroy(V_key_lft_zone);
8436 
8437 	if (!IS_DEFAULT_VNET(curvnet))
8438 		return;
8439 #ifndef IPSEC_DEBUG2
8440 	callout_drain(&key_timer);
8441 #endif
8442 	SPTREE_LOCK_DESTROY();
8443 	REGTREE_LOCK_DESTROY();
8444 	SAHTREE_LOCK_DESTROY();
8445 	ACQ_LOCK_DESTROY();
8446 	SPACQ_LOCK_DESTROY();
8447 }
8448 #endif
8449 
8450 /* record data transfer on SA, and update timestamps */
8451 void
8452 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8453 {
8454 	IPSEC_ASSERT(sav != NULL, ("Null secasvar"));
8455 	IPSEC_ASSERT(m != NULL, ("Null mbuf"));
8456 
8457 	/*
8458 	 * XXX Currently, there is a difference of bytes size
8459 	 * between inbound and outbound processing.
8460 	 */
8461 	counter_u64_add(sav->lft_c_bytes, m->m_pkthdr.len);
8462 
8463 	/*
8464 	 * We use the number of packets as the unit of
8465 	 * allocations.  We increment the variable
8466 	 * whenever {esp,ah}_{in,out}put is called.
8467 	 */
8468 	counter_u64_add(sav->lft_c_allocations, 1);
8469 
8470 	/*
8471 	 * NOTE: We record CURRENT usetime by using wall clock,
8472 	 * in seconds.  HARD and SOFT lifetime are measured by the time
8473 	 * difference (again in seconds) from usetime.
8474 	 *
8475 	 *	usetime
8476 	 *	v     expire   expire
8477 	 * -----+-----+--------+---> t
8478 	 *	<--------------> HARD
8479 	 *	<-----> SOFT
8480 	 */
8481 	if (sav->firstused == 0)
8482 		sav->firstused = time_second;
8483 }
8484 
8485 /*
8486  * Take one of the kernel's security keys and convert it into a PF_KEY
8487  * structure within an mbuf, suitable for sending up to a waiting
8488  * application in user land.
8489  *
8490  * IN:
8491  *    src: A pointer to a kernel security key.
8492  *    exttype: Which type of key this is. Refer to the PF_KEY data structures.
8493  * OUT:
8494  *    a valid mbuf or NULL indicating an error
8495  *
8496  */
8497 
8498 static struct mbuf *
8499 key_setkey(struct seckey *src, uint16_t exttype)
8500 {
8501 	struct mbuf *m;
8502 	struct sadb_key *p;
8503 	int len;
8504 
8505 	if (src == NULL)
8506 		return NULL;
8507 
8508 	len = PFKEY_ALIGN8(sizeof(struct sadb_key) + _KEYLEN(src));
8509 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8510 	if (m == NULL)
8511 		return NULL;
8512 	m_align(m, len);
8513 	m->m_len = len;
8514 	p = mtod(m, struct sadb_key *);
8515 	bzero(p, len);
8516 	p->sadb_key_len = PFKEY_UNIT64(len);
8517 	p->sadb_key_exttype = exttype;
8518 	p->sadb_key_bits = src->bits;
8519 	bcopy(src->key_data, _KEYBUF(p), _KEYLEN(src));
8520 
8521 	return m;
8522 }
8523 
8524 /*
8525  * Take one of the kernel's lifetime data structures and convert it
8526  * into a PF_KEY structure within an mbuf, suitable for sending up to
8527  * a waiting application in user land.
8528  *
8529  * IN:
8530  *    src: A pointer to a kernel lifetime structure.
8531  *    exttype: Which type of lifetime this is. Refer to the PF_KEY
8532  *             data structures for more information.
8533  * OUT:
8534  *    a valid mbuf or NULL indicating an error
8535  *
8536  */
8537 
8538 static struct mbuf *
8539 key_setlifetime(struct seclifetime *src, uint16_t exttype)
8540 {
8541 	struct mbuf *m = NULL;
8542 	struct sadb_lifetime *p;
8543 	int len = PFKEY_ALIGN8(sizeof(struct sadb_lifetime));
8544 
8545 	if (src == NULL)
8546 		return NULL;
8547 
8548 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8549 	if (m == NULL)
8550 		return m;
8551 	m_align(m, len);
8552 	m->m_len = len;
8553 	p = mtod(m, struct sadb_lifetime *);
8554 
8555 	bzero(p, len);
8556 	p->sadb_lifetime_len = PFKEY_UNIT64(len);
8557 	p->sadb_lifetime_exttype = exttype;
8558 	p->sadb_lifetime_allocations = src->allocations;
8559 	p->sadb_lifetime_bytes = src->bytes;
8560 	p->sadb_lifetime_addtime = src->addtime;
8561 	p->sadb_lifetime_usetime = src->usetime;
8562 
8563 	return m;
8564 
8565 }
8566 
8567 const struct enc_xform *
8568 enc_algorithm_lookup(int alg)
8569 {
8570 	int i;
8571 
8572 	for (i = 0; i < nitems(supported_ealgs); i++)
8573 		if (alg == supported_ealgs[i].sadb_alg)
8574 			return (supported_ealgs[i].xform);
8575 	return (NULL);
8576 }
8577 
8578 const struct auth_hash *
8579 auth_algorithm_lookup(int alg)
8580 {
8581 	int i;
8582 
8583 	for (i = 0; i < nitems(supported_aalgs); i++)
8584 		if (alg == supported_aalgs[i].sadb_alg)
8585 			return (supported_aalgs[i].xform);
8586 	return (NULL);
8587 }
8588 
8589 const struct comp_algo *
8590 comp_algorithm_lookup(int alg)
8591 {
8592 	int i;
8593 
8594 	for (i = 0; i < nitems(supported_calgs); i++)
8595 		if (alg == supported_calgs[i].sadb_alg)
8596 			return (supported_calgs[i].xform);
8597 	return (NULL);
8598 }
8599 
8600