xref: /freebsd/sys/netipsec/key.c (revision 7937bfbc0ca53fe7cdd0d54414f9296e273a518e)
1 /*	$KAME: key.c,v 1.191 2001/06/27 10:46:49 sakane Exp $	*/
2 
3 /*-
4  * SPDX-License-Identifier: BSD-3-Clause
5  *
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * This code is referd to RFC 2367
36  */
37 
38 #include "opt_inet.h"
39 #include "opt_inet6.h"
40 #include "opt_ipsec.h"
41 
42 #include <sys/types.h>
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/kernel.h>
46 #include <sys/fnv_hash.h>
47 #include <sys/lock.h>
48 #include <sys/mutex.h>
49 #include <sys/mbuf.h>
50 #include <sys/domain.h>
51 #include <sys/protosw.h>
52 #include <sys/malloc.h>
53 #include <sys/rmlock.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/sysctl.h>
57 #include <sys/errno.h>
58 #include <sys/proc.h>
59 #include <sys/queue.h>
60 #include <sys/refcount.h>
61 #include <sys/syslog.h>
62 
63 #include <vm/uma.h>
64 
65 #include <net/if.h>
66 #include <net/if_var.h>
67 #include <net/vnet.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/ip.h>
72 #include <netinet/in_var.h>
73 #include <netinet/udp.h>
74 
75 #ifdef INET6
76 #include <netinet/ip6.h>
77 #include <netinet6/in6_var.h>
78 #include <netinet6/ip6_var.h>
79 #endif /* INET6 */
80 
81 #include <net/pfkeyv2.h>
82 #include <netipsec/keydb.h>
83 #include <netipsec/key.h>
84 #include <netipsec/keysock.h>
85 #include <netipsec/key_debug.h>
86 #include <netipsec/ipsec_offload.h>
87 
88 #include <netipsec/ipsec.h>
89 #ifdef INET6
90 #include <netipsec/ipsec6.h>
91 #endif
92 
93 #include <netipsec/xform.h>
94 #include <netipsec/ipsec_offload.h>
95 #include <machine/in_cksum.h>
96 #include <machine/stdarg.h>
97 
98 /* randomness */
99 #include <sys/random.h>
100 
101 #ifdef IPSEC_OFFLOAD
102 void (*ipsec_accel_sa_newkey_p)(struct secasvar *sav);
103 void (*ipsec_accel_forget_sav_p)(struct secasvar *sav);
104 void (*ipsec_accel_spdadd_p)(struct secpolicy *sp, struct inpcb *inp);
105 void (*ipsec_accel_spddel_p)(struct secpolicy *sp);
106 int (*ipsec_accel_sa_lifetime_op_p)(struct secasvar *sav,
107     struct seclifetime *lft_c, if_t ifp, enum IF_SA_CNT_WHICH op,
108     struct rm_priotracker *sahtree_trackerp);
109 void (*ipsec_accel_sync_p)(void);
110 bool (*ipsec_accel_is_accel_sav_p)(struct secasvar *sav);
111 struct mbuf *(*ipsec_accel_key_setaccelif_p)(struct secasvar *sav);
112 void (*ipsec_accel_on_ifdown_p)(struct ifnet *ifp);
113 void (*ipsec_accel_drv_sa_lifetime_update_p)(struct secasvar *sav, if_t ifp,
114     u_int drv_spi, uint64_t octets, uint64_t allocs);
115 int (*ipsec_accel_drv_sa_lifetime_fetch_p)(struct secasvar *sav, if_t ifp,
116     u_int drv_spi, uint64_t *octets, uint64_t *allocs);
117 #endif
118 
119 #define FULLMASK	0xff
120 #define	_BITS(bytes)	((bytes) << 3)
121 
122 #define	UINT32_80PCT	0xcccccccc
123 /*
124  * Note on SA reference counting:
125  * - SAs that are not in DEAD state will have (total external reference + 1)
126  *   following value in reference count field.  they cannot be freed and are
127  *   referenced from SA header.
128  * - SAs that are in DEAD state will have (total external reference)
129  *   in reference count field.  they are ready to be freed.  reference from
130  *   SA header will be removed in key_delsav(), when the reference count
131  *   field hits 0 (= no external reference other than from SA header.
132  */
133 
134 VNET_DEFINE(u_int32_t, key_debug_level) = 0;
135 VNET_DEFINE_STATIC(u_int, key_spi_trycnt) = 1000;
136 VNET_DEFINE_STATIC(u_int32_t, key_spi_minval) = 0x100;
137 VNET_DEFINE_STATIC(u_int32_t, key_spi_maxval) = 0x0fffffff;	/* XXX */
138 VNET_DEFINE_STATIC(u_int32_t, policy_id) = 0;
139 /*interval to initialize randseed,1(m)*/
140 VNET_DEFINE_STATIC(u_int, key_int_random) = 60;
141 /* interval to expire acquiring, 30(s)*/
142 VNET_DEFINE_STATIC(u_int, key_larval_lifetime) = 30;
143 /* counter for blocking SADB_ACQUIRE.*/
144 VNET_DEFINE_STATIC(int, key_blockacq_count) = 10;
145 /* lifetime for blocking SADB_ACQUIRE.*/
146 VNET_DEFINE_STATIC(int, key_blockacq_lifetime) = 20;
147 /* preferred old sa rather than new sa.*/
148 VNET_DEFINE_STATIC(int, key_preferred_oldsa) = 1;
149 #define	V_key_spi_trycnt	VNET(key_spi_trycnt)
150 #define	V_key_spi_minval	VNET(key_spi_minval)
151 #define	V_key_spi_maxval	VNET(key_spi_maxval)
152 #define	V_policy_id		VNET(policy_id)
153 #define	V_key_int_random	VNET(key_int_random)
154 #define	V_key_larval_lifetime	VNET(key_larval_lifetime)
155 #define	V_key_blockacq_count	VNET(key_blockacq_count)
156 #define	V_key_blockacq_lifetime	VNET(key_blockacq_lifetime)
157 #define	V_key_preferred_oldsa	VNET(key_preferred_oldsa)
158 
159 VNET_DEFINE_STATIC(u_int32_t, acq_seq) = 0;
160 #define	V_acq_seq		VNET(acq_seq)
161 
162 VNET_DEFINE_STATIC(uint32_t, sp_genid) = 0;
163 #define	V_sp_genid		VNET(sp_genid)
164 
165 /* SPD */
166 TAILQ_HEAD(secpolicy_queue, secpolicy);
167 LIST_HEAD(secpolicy_list, secpolicy);
168 VNET_DEFINE_STATIC(struct secpolicy_queue, sptree[IPSEC_DIR_MAX]);
169 VNET_DEFINE_STATIC(struct secpolicy_queue, sptree_ifnet[IPSEC_DIR_MAX]);
170 static struct rmlock sptree_lock;
171 #define	V_sptree		VNET(sptree)
172 #define	V_sptree_ifnet		VNET(sptree_ifnet)
173 #define	SPTREE_LOCK_INIT()      rm_init(&sptree_lock, "sptree")
174 #define	SPTREE_LOCK_DESTROY()   rm_destroy(&sptree_lock)
175 #define	SPTREE_RLOCK_TRACKER    struct rm_priotracker sptree_tracker
176 #define	SPTREE_RLOCK()          rm_rlock(&sptree_lock, &sptree_tracker)
177 #define	SPTREE_RUNLOCK()        rm_runlock(&sptree_lock, &sptree_tracker)
178 #define	SPTREE_RLOCK_ASSERT()   rm_assert(&sptree_lock, RA_RLOCKED)
179 #define	SPTREE_WLOCK()          rm_wlock(&sptree_lock)
180 #define	SPTREE_WUNLOCK()        rm_wunlock(&sptree_lock)
181 #define	SPTREE_WLOCK_ASSERT()   rm_assert(&sptree_lock, RA_WLOCKED)
182 #define	SPTREE_UNLOCK_ASSERT()  rm_assert(&sptree_lock, RA_UNLOCKED)
183 
184 /* Hash table for lookup SP using unique id */
185 VNET_DEFINE_STATIC(struct secpolicy_list *, sphashtbl);
186 VNET_DEFINE_STATIC(u_long, sphash_mask);
187 #define	V_sphashtbl		VNET(sphashtbl)
188 #define	V_sphash_mask		VNET(sphash_mask)
189 
190 #define	SPHASH_NHASH_LOG2	7
191 #define	SPHASH_NHASH		(1 << SPHASH_NHASH_LOG2)
192 #define	SPHASH_HASHVAL(id)	(key_u32hash(id) & V_sphash_mask)
193 #define	SPHASH_HASH(id)		&V_sphashtbl[SPHASH_HASHVAL(id)]
194 
195 /* SPD cache */
196 struct spdcache_entry {
197    struct secpolicyindex spidx;	/* secpolicyindex */
198    struct secpolicy *sp;	/* cached policy to be used */
199 
200    LIST_ENTRY(spdcache_entry) chain;
201 };
202 LIST_HEAD(spdcache_entry_list, spdcache_entry);
203 
204 #define	SPDCACHE_MAX_ENTRIES_PER_HASH	8
205 
206 VNET_DEFINE_STATIC(u_int, key_spdcache_maxentries) = 0;
207 #define	V_key_spdcache_maxentries	VNET(key_spdcache_maxentries)
208 VNET_DEFINE_STATIC(u_int, key_spdcache_threshold) = 32;
209 #define	V_key_spdcache_threshold	VNET(key_spdcache_threshold)
210 VNET_DEFINE_STATIC(unsigned long, spd_size) = 0;
211 #define	V_spd_size		VNET(spd_size)
212 
213 #define SPDCACHE_ENABLED()	(V_key_spdcache_maxentries != 0)
214 #define SPDCACHE_ACTIVE() \
215 	(SPDCACHE_ENABLED() && V_spd_size >= V_key_spdcache_threshold)
216 
217 VNET_DEFINE_STATIC(struct spdcache_entry_list *, spdcachehashtbl);
218 VNET_DEFINE_STATIC(u_long, spdcachehash_mask);
219 #define	V_spdcachehashtbl	VNET(spdcachehashtbl)
220 #define	V_spdcachehash_mask	VNET(spdcachehash_mask)
221 
222 #define	SPDCACHE_HASHVAL(idx) \
223 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->ul_proto) &  \
224 	    V_spdcachehash_mask)
225 
226 /* Each cache line is protected by a mutex */
227 VNET_DEFINE_STATIC(struct mtx *, spdcache_lock);
228 #define	V_spdcache_lock		VNET(spdcache_lock)
229 
230 #define	SPDCACHE_LOCK_INIT(a) \
231 	mtx_init(&V_spdcache_lock[a], "spdcache", \
232 	    "fast ipsec SPD cache", MTX_DEF|MTX_DUPOK)
233 #define	SPDCACHE_LOCK_DESTROY(a)	mtx_destroy(&V_spdcache_lock[a])
234 #define	SPDCACHE_LOCK(a)		mtx_lock(&V_spdcache_lock[a]);
235 #define	SPDCACHE_UNLOCK(a)		mtx_unlock(&V_spdcache_lock[a]);
236 
237 static struct sx spi_alloc_lock;
238 #define	SPI_ALLOC_LOCK_INIT()		sx_init(&spi_alloc_lock, "spialloc")
239 #define	SPI_ALLOC_LOCK_DESTROY()	sx_destroy(&spi_alloc_lock)
240 #define	SPI_ALLOC_LOCK()          	sx_xlock(&spi_alloc_lock)
241 #define	SPI_ALLOC_UNLOCK()        	sx_unlock(&spi_alloc_lock)
242 #define	SPI_ALLOC_LOCK_ASSERT()   	sx_assert(&spi_alloc_lock, SA_XLOCKED)
243 
244 /* SAD */
245 TAILQ_HEAD(secashead_queue, secashead);
246 LIST_HEAD(secashead_list, secashead);
247 VNET_DEFINE_STATIC(struct secashead_queue, sahtree);
248 static struct rmlock sahtree_lock;
249 #define	V_sahtree		VNET(sahtree)
250 #define	SAHTREE_LOCK_INIT()	rm_init(&sahtree_lock, "sahtree")
251 #define	SAHTREE_LOCK_DESTROY()	rm_destroy(&sahtree_lock)
252 #define	SAHTREE_RLOCK_TRACKER	struct rm_priotracker sahtree_tracker
253 #define	SAHTREE_RLOCK()		rm_rlock(&sahtree_lock, &sahtree_tracker)
254 #define	SAHTREE_RUNLOCK()	rm_runlock(&sahtree_lock, &sahtree_tracker)
255 #define	SAHTREE_RLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_RLOCKED)
256 #define	SAHTREE_WLOCK()		rm_wlock(&sahtree_lock)
257 #define	SAHTREE_WUNLOCK()	rm_wunlock(&sahtree_lock)
258 #define	SAHTREE_WLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_WLOCKED)
259 #define	SAHTREE_UNLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_UNLOCKED)
260 
261 /* Hash table for lookup in SAD using SA addresses */
262 VNET_DEFINE_STATIC(struct secashead_list *, sahaddrhashtbl);
263 VNET_DEFINE_STATIC(u_long, sahaddrhash_mask);
264 #define	V_sahaddrhashtbl	VNET(sahaddrhashtbl)
265 #define	V_sahaddrhash_mask	VNET(sahaddrhash_mask)
266 
267 #define	SAHHASH_NHASH_LOG2	7
268 #define	SAHHASH_NHASH		(1 << SAHHASH_NHASH_LOG2)
269 #define	SAHADDRHASH_HASHVAL(idx)	\
270 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->proto) & \
271 	    V_sahaddrhash_mask)
272 #define	SAHADDRHASH_HASH(saidx)		\
273     &V_sahaddrhashtbl[SAHADDRHASH_HASHVAL(saidx)]
274 
275 /* Hash table for lookup in SAD using SPI */
276 LIST_HEAD(secasvar_list, secasvar);
277 VNET_DEFINE_STATIC(struct secasvar_list *, savhashtbl);
278 VNET_DEFINE_STATIC(u_long, savhash_mask);
279 #define	V_savhashtbl		VNET(savhashtbl)
280 #define	V_savhash_mask		VNET(savhash_mask)
281 #define	SAVHASH_NHASH_LOG2	7
282 #define	SAVHASH_NHASH		(1 << SAVHASH_NHASH_LOG2)
283 #define	SAVHASH_HASHVAL(spi)	(key_u32hash(spi) & V_savhash_mask)
284 #define	SAVHASH_HASH(spi)	&V_savhashtbl[SAVHASH_HASHVAL(spi)]
285 
286 static uint32_t
287 key_addrprotohash(const union sockaddr_union *src,
288     const union sockaddr_union *dst, const uint8_t *proto)
289 {
290 	uint32_t hval;
291 
292 	hval = fnv_32_buf(proto, sizeof(*proto),
293 	    FNV1_32_INIT);
294 	switch (dst->sa.sa_family) {
295 #ifdef INET
296 	case AF_INET:
297 		hval = fnv_32_buf(&src->sin.sin_addr,
298 		    sizeof(in_addr_t), hval);
299 		hval = fnv_32_buf(&dst->sin.sin_addr,
300 		    sizeof(in_addr_t), hval);
301 		break;
302 #endif
303 #ifdef INET6
304 	case AF_INET6:
305 		hval = fnv_32_buf(&src->sin6.sin6_addr,
306 		    sizeof(struct in6_addr), hval);
307 		hval = fnv_32_buf(&dst->sin6.sin6_addr,
308 		    sizeof(struct in6_addr), hval);
309 		break;
310 #endif
311 	default:
312 		hval = 0;
313 		ipseclog((LOG_DEBUG, "%s: unknown address family %d\n",
314 		    __func__, dst->sa.sa_family));
315 	}
316 	return (hval);
317 }
318 
319 static uint32_t
320 key_u32hash(uint32_t val)
321 {
322 
323 	return (fnv_32_buf(&val, sizeof(val), FNV1_32_INIT));
324 }
325 
326 							/* registed list */
327 VNET_DEFINE_STATIC(LIST_HEAD(_regtree, secreg), regtree[SADB_SATYPE_MAX + 1]);
328 #define	V_regtree		VNET(regtree)
329 static struct mtx regtree_lock;
330 #define	REGTREE_LOCK_INIT() \
331 	mtx_init(&regtree_lock, "regtree", "fast ipsec regtree", MTX_DEF)
332 #define	REGTREE_LOCK_DESTROY()	mtx_destroy(&regtree_lock)
333 #define	REGTREE_LOCK()		mtx_lock(&regtree_lock)
334 #define	REGTREE_UNLOCK()	mtx_unlock(&regtree_lock)
335 #define	REGTREE_LOCK_ASSERT()	mtx_assert(&regtree_lock, MA_OWNED)
336 
337 /* Acquiring list */
338 LIST_HEAD(secacq_list, secacq);
339 VNET_DEFINE_STATIC(struct secacq_list, acqtree);
340 #define	V_acqtree		VNET(acqtree)
341 static struct mtx acq_lock;
342 #define	ACQ_LOCK_INIT() \
343     mtx_init(&acq_lock, "acqtree", "ipsec SA acquiring list", MTX_DEF)
344 #define	ACQ_LOCK_DESTROY()	mtx_destroy(&acq_lock)
345 #define	ACQ_LOCK()		mtx_lock(&acq_lock)
346 #define	ACQ_UNLOCK()		mtx_unlock(&acq_lock)
347 #define	ACQ_LOCK_ASSERT()	mtx_assert(&acq_lock, MA_OWNED)
348 
349 /* Hash table for lookup in ACQ list using SA addresses */
350 VNET_DEFINE_STATIC(struct secacq_list *, acqaddrhashtbl);
351 VNET_DEFINE_STATIC(u_long, acqaddrhash_mask);
352 #define	V_acqaddrhashtbl	VNET(acqaddrhashtbl)
353 #define	V_acqaddrhash_mask	VNET(acqaddrhash_mask)
354 
355 /* Hash table for lookup in ACQ list using SEQ number */
356 VNET_DEFINE_STATIC(struct secacq_list *, acqseqhashtbl);
357 VNET_DEFINE_STATIC(u_long, acqseqhash_mask);
358 #define	V_acqseqhashtbl		VNET(acqseqhashtbl)
359 #define	V_acqseqhash_mask	VNET(acqseqhash_mask)
360 
361 #define	ACQHASH_NHASH_LOG2	7
362 #define	ACQHASH_NHASH		(1 << ACQHASH_NHASH_LOG2)
363 #define	ACQADDRHASH_HASHVAL(idx)	\
364 	(key_addrprotohash(&(idx)->src, &(idx)->dst, &(idx)->proto) & \
365 	    V_acqaddrhash_mask)
366 #define	ACQSEQHASH_HASHVAL(seq)		\
367     (key_u32hash(seq) & V_acqseqhash_mask)
368 #define	ACQADDRHASH_HASH(saidx)	\
369     &V_acqaddrhashtbl[ACQADDRHASH_HASHVAL(saidx)]
370 #define	ACQSEQHASH_HASH(seq)	\
371     &V_acqseqhashtbl[ACQSEQHASH_HASHVAL(seq)]
372 							/* SP acquiring list */
373 VNET_DEFINE_STATIC(LIST_HEAD(_spacqtree, secspacq), spacqtree);
374 #define	V_spacqtree		VNET(spacqtree)
375 static struct mtx spacq_lock;
376 #define	SPACQ_LOCK_INIT() \
377 	mtx_init(&spacq_lock, "spacqtree", \
378 		"fast ipsec security policy acquire list", MTX_DEF)
379 #define	SPACQ_LOCK_DESTROY()	mtx_destroy(&spacq_lock)
380 #define	SPACQ_LOCK()		mtx_lock(&spacq_lock)
381 #define	SPACQ_UNLOCK()		mtx_unlock(&spacq_lock)
382 #define	SPACQ_LOCK_ASSERT()	mtx_assert(&spacq_lock, MA_OWNED)
383 
384 static const int minsize[] = {
385 	[SADB_EXT_RESERVED] = sizeof(struct sadb_msg),
386 	[SADB_EXT_SA] = sizeof(struct sadb_sa),
387 	[SADB_EXT_LIFETIME_CURRENT] = sizeof(struct sadb_lifetime),
388 	[SADB_EXT_LIFETIME_HARD] = sizeof(struct sadb_lifetime),
389 	[SADB_EXT_LIFETIME_SOFT] = sizeof(struct sadb_lifetime),
390 	[SADB_EXT_ADDRESS_SRC] = sizeof(struct sadb_address),
391 	[SADB_EXT_ADDRESS_DST] = sizeof(struct sadb_address),
392 	[SADB_EXT_ADDRESS_PROXY] = sizeof(struct sadb_address),
393 	[SADB_EXT_KEY_AUTH] = sizeof(struct sadb_key),
394 	[SADB_EXT_KEY_ENCRYPT] = sizeof(struct sadb_key),
395 	[SADB_EXT_IDENTITY_SRC] = sizeof(struct sadb_ident),
396 	[SADB_EXT_IDENTITY_DST] = sizeof(struct sadb_ident),
397 	[SADB_EXT_SENSITIVITY] = sizeof(struct sadb_sens),
398 	[SADB_EXT_PROPOSAL] = sizeof(struct sadb_prop),
399 	[SADB_EXT_SUPPORTED_AUTH] = sizeof(struct sadb_supported),
400 	[SADB_EXT_SUPPORTED_ENCRYPT] = sizeof(struct sadb_supported),
401 	[SADB_EXT_SPIRANGE] = sizeof(struct sadb_spirange),
402 	[SADB_X_EXT_KMPRIVATE] = 0,
403 	[SADB_X_EXT_POLICY] = sizeof(struct sadb_x_policy),
404 	[SADB_X_EXT_SA2] = sizeof(struct sadb_x_sa2),
405 	[SADB_X_EXT_NAT_T_TYPE] = sizeof(struct sadb_x_nat_t_type),
406 	[SADB_X_EXT_NAT_T_SPORT] = sizeof(struct sadb_x_nat_t_port),
407 	[SADB_X_EXT_NAT_T_DPORT] = sizeof(struct sadb_x_nat_t_port),
408 	[SADB_X_EXT_NAT_T_OAI] = sizeof(struct sadb_address),
409 	[SADB_X_EXT_NAT_T_OAR] = sizeof(struct sadb_address),
410 	[SADB_X_EXT_NAT_T_FRAG] = sizeof(struct sadb_x_nat_t_frag),
411 	[SADB_X_EXT_SA_REPLAY] = sizeof(struct sadb_x_sa_replay),
412 	[SADB_X_EXT_NEW_ADDRESS_SRC] = sizeof(struct sadb_address),
413 	[SADB_X_EXT_NEW_ADDRESS_DST] = sizeof(struct sadb_address),
414 	[SADB_X_EXT_LFT_CUR_SW_OFFL] = sizeof(struct sadb_lifetime),
415 	[SADB_X_EXT_LFT_CUR_HW_OFFL] = sizeof(struct sadb_lifetime),
416 	[SADB_X_EXT_IF_HW_OFFL] = sizeof(struct sadb_x_if_hw_offl),
417 };
418 _Static_assert(nitems(minsize) == SADB_EXT_MAX + 1, "minsize size mismatch");
419 
420 static const int maxsize[] = {
421 	[SADB_EXT_RESERVED] = sizeof(struct sadb_msg),
422 	[SADB_EXT_SA] = sizeof(struct sadb_sa),
423 	[SADB_EXT_LIFETIME_CURRENT] = sizeof(struct sadb_lifetime),
424 	[SADB_EXT_LIFETIME_HARD] = sizeof(struct sadb_lifetime),
425 	[SADB_EXT_LIFETIME_SOFT] = sizeof(struct sadb_lifetime),
426 	[SADB_EXT_ADDRESS_SRC] = 0,
427 	[SADB_EXT_ADDRESS_DST] = 0,
428 	[SADB_EXT_ADDRESS_PROXY] = 0,
429 	[SADB_EXT_KEY_AUTH] = 0,
430 	[SADB_EXT_KEY_ENCRYPT] = 0,
431 	[SADB_EXT_IDENTITY_SRC] = 0,
432 	[SADB_EXT_IDENTITY_DST] = 0,
433 	[SADB_EXT_SENSITIVITY] = 0,
434 	[SADB_EXT_PROPOSAL] = 0,
435 	[SADB_EXT_SUPPORTED_AUTH] = 0,
436 	[SADB_EXT_SUPPORTED_ENCRYPT] = 0,
437 	[SADB_EXT_SPIRANGE] = sizeof(struct sadb_spirange),
438 	[SADB_X_EXT_KMPRIVATE] = 0,
439 	[SADB_X_EXT_POLICY] = 0,
440 	[SADB_X_EXT_SA2] = sizeof(struct sadb_x_sa2),
441 	[SADB_X_EXT_NAT_T_TYPE] = sizeof(struct sadb_x_nat_t_type),
442 	[SADB_X_EXT_NAT_T_SPORT] = sizeof(struct sadb_x_nat_t_port),
443 	[SADB_X_EXT_NAT_T_DPORT] = sizeof(struct sadb_x_nat_t_port),
444 	[SADB_X_EXT_NAT_T_OAI] = 0,
445 	[SADB_X_EXT_NAT_T_OAR] = 0,
446 	[SADB_X_EXT_NAT_T_FRAG] = sizeof(struct sadb_x_nat_t_frag),
447 	[SADB_X_EXT_SA_REPLAY] = sizeof(struct sadb_x_sa_replay),
448 	[SADB_X_EXT_NEW_ADDRESS_SRC] = 0,
449 	[SADB_X_EXT_NEW_ADDRESS_DST] = 0,
450 	[SADB_X_EXT_LFT_CUR_SW_OFFL] = sizeof(struct sadb_lifetime),
451 	[SADB_X_EXT_LFT_CUR_HW_OFFL] = sizeof(struct sadb_lifetime),
452 	[SADB_X_EXT_IF_HW_OFFL] = sizeof(struct sadb_x_if_hw_offl),
453 };
454 _Static_assert(nitems(maxsize) == SADB_EXT_MAX + 1, "maxsize size mismatch");
455 
456 /*
457  * Internal values for SA flags:
458  * SADB_X_EXT_F_CLONED means that SA was cloned by key_updateaddresses,
459  *	thus we will not free the most of SA content in key_delsav().
460  */
461 #define	SADB_X_EXT_F_CLONED	0x80000000
462 
463 #define	SADB_CHECKLEN(_mhp, _ext)			\
464     ((_mhp)->extlen[(_ext)] < minsize[(_ext)] || (maxsize[(_ext)] != 0 && \
465 	((_mhp)->extlen[(_ext)] > maxsize[(_ext)])))
466 #define	SADB_CHECKHDR(_mhp, _ext)	((_mhp)->ext[(_ext)] == NULL)
467 
468 VNET_DEFINE_STATIC(int, ipsec_esp_keymin) = 256;
469 VNET_DEFINE_STATIC(int, ipsec_esp_auth) = 0;
470 VNET_DEFINE_STATIC(int, ipsec_ah_keymin) = 128;
471 
472 #define	V_ipsec_esp_keymin	VNET(ipsec_esp_keymin)
473 #define	V_ipsec_esp_auth	VNET(ipsec_esp_auth)
474 #define	V_ipsec_ah_keymin	VNET(ipsec_ah_keymin)
475 
476 #ifdef IPSEC_DEBUG
477 VNET_DEFINE(int, ipsec_debug) = 1;
478 #else
479 VNET_DEFINE(int, ipsec_debug) = 0;
480 #endif
481 
482 #ifdef INET
483 SYSCTL_DECL(_net_inet_ipsec);
484 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
485     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
486     "Enable IPsec debugging output when set.");
487 #endif
488 #ifdef INET6
489 SYSCTL_DECL(_net_inet6_ipsec6);
490 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug,
491     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
492     "Enable IPsec debugging output when set.");
493 #endif
494 
495 SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,
496 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_debug_level), 0, "");
497 
498 /* max count of trial for the decision of spi value */
499 SYSCTL_INT(_net_key, KEYCTL_SPI_TRY, spi_trycnt,
500 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_trycnt), 0, "");
501 
502 /* minimum spi value to allocate automatically. */
503 SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE, spi_minval,
504 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_minval), 0, "");
505 
506 /* maximun spi value to allocate automatically. */
507 SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE, spi_maxval,
508 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_maxval), 0, "");
509 
510 /* interval to initialize randseed */
511 SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT, int_random,
512 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_int_random), 0, "");
513 
514 /* lifetime for larval SA */
515 SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME, larval_lifetime,
516 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_larval_lifetime), 0, "");
517 
518 /* counter for blocking to send SADB_ACQUIRE to IKEd */
519 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT, blockacq_count,
520 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_count), 0, "");
521 
522 /* lifetime for blocking to send SADB_ACQUIRE to IKEd */
523 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME, blockacq_lifetime,
524 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_lifetime), 0, "");
525 
526 /* ESP auth */
527 SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH, esp_auth,
528 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_auth), 0, "");
529 
530 /* minimum ESP key length */
531 SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN, esp_keymin,
532 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_keymin), 0, "");
533 
534 /* minimum AH key length */
535 SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN, ah_keymin,
536 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_ah_keymin), 0, "");
537 
538 /* perfered old SA rather than new SA */
539 SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA, preferred_oldsa,
540 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_preferred_oldsa), 0, "");
541 
542 SYSCTL_NODE(_net_key, OID_AUTO, spdcache, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
543     "SPD cache");
544 
545 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, maxentries,
546 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_maxentries), 0,
547 	"Maximum number of entries in the SPD cache"
548 	" (power of 2, 0 to disable)");
549 
550 SYSCTL_UINT(_net_key_spdcache, OID_AUTO, threshold,
551 	CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(key_spdcache_threshold), 0,
552 	"Number of SPs that make the SPD cache active");
553 
554 #define __LIST_CHAINED(elm) \
555 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
556 
557 MALLOC_DEFINE(M_IPSEC_SA, "secasvar", "ipsec security association");
558 MALLOC_DEFINE(M_IPSEC_SAH, "sahead", "ipsec sa head");
559 MALLOC_DEFINE(M_IPSEC_SP, "ipsecpolicy", "ipsec security policy");
560 MALLOC_DEFINE(M_IPSEC_SR, "ipsecrequest", "ipsec security request");
561 MALLOC_DEFINE(M_IPSEC_MISC, "ipsec-misc", "ipsec miscellaneous");
562 MALLOC_DEFINE(M_IPSEC_SAQ, "ipsec-saq", "ipsec sa acquire");
563 MALLOC_DEFINE(M_IPSEC_SAR, "ipsec-reg", "ipsec sa acquire");
564 MALLOC_DEFINE(M_IPSEC_SPDCACHE, "ipsec-spdcache", "ipsec SPD cache");
565 
566 static uma_zone_t __read_mostly ipsec_key_lft_zone;
567 
568 /*
569  * set parameters into secpolicyindex buffer.
570  * Must allocate secpolicyindex buffer passed to this function.
571  */
572 #define KEY_SETSECSPIDX(_dir, s, d, ps, pd, ulp, idx) \
573 do { \
574 	bzero((idx), sizeof(struct secpolicyindex));                         \
575 	(idx)->dir = (_dir);                                                 \
576 	(idx)->prefs = (ps);                                                 \
577 	(idx)->prefd = (pd);                                                 \
578 	(idx)->ul_proto = (ulp);                                             \
579 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
580 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
581 } while (0)
582 
583 /*
584  * set parameters into secasindex buffer.
585  * Must allocate secasindex buffer before calling this function.
586  */
587 #define KEY_SETSECASIDX(p, m, r, s, d, idx) \
588 do { \
589 	bzero((idx), sizeof(struct secasindex));                             \
590 	(idx)->proto = (p);                                                  \
591 	(idx)->mode = (m);                                                   \
592 	(idx)->reqid = (r);                                                  \
593 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
594 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
595 	key_porttosaddr(&(idx)->src.sa, 0);				     \
596 	key_porttosaddr(&(idx)->dst.sa, 0);				     \
597 } while (0)
598 
599 /* key statistics */
600 struct _keystat {
601 	u_long getspi_count; /* the avarage of count to try to get new SPI */
602 } keystat;
603 
604 struct sadb_msghdr {
605 	struct sadb_msg *msg;
606 	struct sadb_ext *ext[SADB_EXT_MAX + 1];
607 	int extoff[SADB_EXT_MAX + 1];
608 	int extlen[SADB_EXT_MAX + 1];
609 };
610 
611 static const struct supported_ealgs {
612 	int sadb_alg;
613 	const struct enc_xform *xform;
614 } supported_ealgs[] = {
615 	{ SADB_X_EALG_AES,		&enc_xform_aes_cbc },
616 	{ SADB_EALG_NULL,		&enc_xform_null },
617 	{ SADB_X_EALG_AESCTR,		&enc_xform_aes_icm },
618 	{ SADB_X_EALG_AESGCM16,		&enc_xform_aes_nist_gcm },
619 	{ SADB_X_EALG_AESGMAC,		&enc_xform_aes_nist_gmac },
620 	{ SADB_X_EALG_CHACHA20POLY1305,	&enc_xform_chacha20_poly1305 },
621 };
622 
623 static const struct supported_aalgs {
624 	int sadb_alg;
625 	const struct auth_hash *xform;
626 } supported_aalgs[] = {
627 	{ SADB_X_AALG_NULL,		&auth_hash_null },
628 	{ SADB_AALG_SHA1HMAC,		&auth_hash_hmac_sha1 },
629 	{ SADB_X_AALG_SHA2_256,		&auth_hash_hmac_sha2_256 },
630 	{ SADB_X_AALG_SHA2_384,		&auth_hash_hmac_sha2_384 },
631 	{ SADB_X_AALG_SHA2_512,		&auth_hash_hmac_sha2_512 },
632 	{ SADB_X_AALG_AES128GMAC,	&auth_hash_nist_gmac_aes_128 },
633 	{ SADB_X_AALG_AES192GMAC,	&auth_hash_nist_gmac_aes_192 },
634 	{ SADB_X_AALG_AES256GMAC,	&auth_hash_nist_gmac_aes_256 },
635 	{ SADB_X_AALG_CHACHA20POLY1305,	&auth_hash_poly1305 },
636 };
637 
638 static const struct supported_calgs {
639 	int sadb_alg;
640 	const struct comp_algo *xform;
641 } supported_calgs[] = {
642 	{ SADB_X_CALG_DEFLATE,		&comp_algo_deflate },
643 };
644 
645 #ifndef IPSEC_DEBUG2
646 static struct callout key_timer;
647 #endif
648 
649 static void key_unlink(struct secpolicy *);
650 static void key_detach(struct secpolicy *);
651 static struct secpolicy *key_getsp(struct secpolicyindex *);
652 static struct secpolicy *key_getspbyid(u_int32_t);
653 static struct mbuf *key_gather_mbuf(struct mbuf *,
654 	const struct sadb_msghdr *, int, int, ...);
655 static int key_spdadd(struct socket *, struct mbuf *,
656 	const struct sadb_msghdr *);
657 static uint32_t key_getnewspid(void);
658 static int key_spddelete(struct socket *, struct mbuf *,
659 	const struct sadb_msghdr *);
660 static int key_spddelete2(struct socket *, struct mbuf *,
661 	const struct sadb_msghdr *);
662 static int key_spdget(struct socket *, struct mbuf *,
663 	const struct sadb_msghdr *);
664 static int key_spdflush(struct socket *, struct mbuf *,
665 	const struct sadb_msghdr *);
666 static int key_spddump(struct socket *, struct mbuf *,
667 	const struct sadb_msghdr *);
668 static struct mbuf *key_setdumpsp(struct secpolicy *,
669 	u_int8_t, u_int32_t, u_int32_t);
670 static struct mbuf *key_sp2mbuf(struct secpolicy *);
671 static size_t key_getspreqmsglen(struct secpolicy *);
672 static int key_spdexpire(struct secpolicy *);
673 static struct secashead *key_newsah(struct secasindex *);
674 static void key_freesah(struct secashead **);
675 static void key_delsah(struct secashead *);
676 static struct secasvar *key_newsav(const struct sadb_msghdr *,
677     struct secasindex *, uint32_t, int *);
678 static void key_delsav(struct secasvar *);
679 static void key_unlinksav(struct secasvar *);
680 static struct secashead *key_getsah(struct secasindex *);
681 static int key_checkspidup(uint32_t);
682 static struct secasvar *key_getsavbyspi(uint32_t);
683 static int key_setnatt(struct secasvar *, const struct sadb_msghdr *);
684 static int key_setsaval(struct secasvar *, const struct sadb_msghdr *);
685 static int key_updatelifetimes(struct secasvar *, const struct sadb_msghdr *);
686 static int key_updateaddresses(struct socket *, struct mbuf *,
687     const struct sadb_msghdr *, struct secasvar *, struct secasindex *);
688 
689 static struct mbuf *key_setdumpsa(struct secasvar *, u_int8_t,
690 	u_int8_t, u_int32_t, u_int32_t, struct rm_priotracker *);
691 static struct mbuf *key_setsadbmsg(u_int8_t, u_int16_t, u_int8_t,
692 	u_int32_t, pid_t, u_int16_t);
693 static struct mbuf *key_setsadbsa(struct secasvar *);
694 static struct mbuf *key_setsadbaddr(u_int16_t,
695 	const struct sockaddr *, u_int8_t, u_int16_t);
696 static struct mbuf *key_setsadbxport(u_int16_t, u_int16_t);
697 static struct mbuf *key_setsadbxtype(u_int16_t);
698 static struct mbuf *key_setsadbxsa2(u_int8_t, u_int32_t, u_int32_t);
699 static struct mbuf *key_setsadbxsareplay(u_int32_t);
700 static struct mbuf *key_setsadbxpolicy(u_int16_t, u_int8_t,
701 	u_int32_t, u_int32_t);
702 static struct seckey *key_dup_keymsg(const struct sadb_key *, size_t,
703     struct malloc_type *);
704 static struct seclifetime *key_dup_lifemsg(const struct sadb_lifetime *src,
705     struct malloc_type *);
706 
707 /* flags for key_cmpsaidx() */
708 #define CMP_HEAD	1	/* protocol, addresses. */
709 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
710 #define CMP_REQID	3	/* additionally HEAD, reaid. */
711 #define CMP_EXACTLY	4	/* all elements. */
712 static int key_cmpsaidx(const struct secasindex *,
713     const struct secasindex *, int);
714 static int key_cmpspidx_exactly(struct secpolicyindex *,
715     struct secpolicyindex *);
716 static int key_cmpspidx_withmask(struct secpolicyindex *,
717     struct secpolicyindex *);
718 static int key_bbcmp(const void *, const void *, u_int);
719 static uint8_t key_satype2proto(uint8_t);
720 static uint8_t key_proto2satype(uint8_t);
721 
722 static int key_getspi(struct socket *, struct mbuf *,
723 	const struct sadb_msghdr *);
724 static uint32_t key_do_getnewspi(struct sadb_spirange *, struct secasindex *);
725 static int key_update(struct socket *, struct mbuf *,
726 	const struct sadb_msghdr *);
727 static int key_add(struct socket *, struct mbuf *,
728 	const struct sadb_msghdr *);
729 static int key_setident(struct secashead *, const struct sadb_msghdr *);
730 static struct mbuf *key_getmsgbuf_x1(struct mbuf *,
731 	const struct sadb_msghdr *);
732 static int key_delete(struct socket *, struct mbuf *,
733 	const struct sadb_msghdr *);
734 static int key_delete_all(struct socket *, struct mbuf *,
735 	const struct sadb_msghdr *, struct secasindex *);
736 static int key_get(struct socket *, struct mbuf *,
737 	const struct sadb_msghdr *);
738 
739 static void key_getcomb_setlifetime(struct sadb_comb *);
740 static struct mbuf *key_getcomb_ealg(void);
741 static struct mbuf *key_getcomb_ah(void);
742 static struct mbuf *key_getcomb_ipcomp(void);
743 static struct mbuf *key_getprop(const struct secasindex *);
744 
745 static int key_acquire(const struct secasindex *, struct secpolicy *);
746 static uint32_t key_newacq(const struct secasindex *, int *);
747 static uint32_t key_getacq(const struct secasindex *, int *);
748 static int key_acqdone(const struct secasindex *, uint32_t);
749 static int key_acqreset(uint32_t);
750 static struct secspacq *key_newspacq(struct secpolicyindex *);
751 static struct secspacq *key_getspacq(struct secpolicyindex *);
752 static int key_acquire2(struct socket *, struct mbuf *,
753 	const struct sadb_msghdr *);
754 static int key_register(struct socket *, struct mbuf *,
755 	const struct sadb_msghdr *);
756 static int key_expire(struct secasvar *, int);
757 static int key_flush(struct socket *, struct mbuf *,
758 	const struct sadb_msghdr *);
759 static int key_dump(struct socket *, struct mbuf *,
760 	const struct sadb_msghdr *);
761 static int key_promisc(struct socket *, struct mbuf *,
762 	const struct sadb_msghdr *);
763 static int key_senderror(struct socket *, struct mbuf *, int);
764 static int key_validate_ext(const struct sadb_ext *, int);
765 static int key_align(struct mbuf *, struct sadb_msghdr *);
766 static struct mbuf *key_setlifetime(struct seclifetime *, uint16_t);
767 static struct mbuf *key_setkey(struct seckey *, uint16_t);
768 
769 static void spdcache_init(void);
770 static void spdcache_clear(void);
771 static struct spdcache_entry *spdcache_entry_alloc(
772 	const struct secpolicyindex *spidx,
773 	struct secpolicy *policy);
774 static void spdcache_entry_free(struct spdcache_entry *entry);
775 #ifdef VIMAGE
776 static void spdcache_destroy(void);
777 #endif
778 
779 #define	DBG_IPSEC_INITREF(t, p)	do {				\
780 	refcount_init(&(p)->refcnt, 1);				\
781 	KEYDBG(KEY_STAMP,					\
782 	    printf("%s: Initialize refcnt %s(%p) = %u\n",	\
783 	    __func__, #t, (p), (p)->refcnt));			\
784 } while (0)
785 #define	DBG_IPSEC_ADDREF(t, p)	do {				\
786 	refcount_acquire(&(p)->refcnt);				\
787 	KEYDBG(KEY_STAMP,					\
788 	    printf("%s: Acquire refcnt %s(%p) -> %u\n",		\
789 	    __func__, #t, (p), (p)->refcnt));			\
790 } while (0)
791 #define	DBG_IPSEC_DELREF(t, p)	do {				\
792 	KEYDBG(KEY_STAMP,					\
793 	    printf("%s: Release refcnt %s(%p) -> %u\n",		\
794 	    __func__, #t, (p), (p)->refcnt - 1));		\
795 	refcount_release(&(p)->refcnt);				\
796 } while (0)
797 
798 #define	IPSEC_INITREF(t, p)	refcount_init(&(p)->refcnt, 1)
799 #define	IPSEC_ADDREF(t, p)	refcount_acquire(&(p)->refcnt)
800 #define	IPSEC_DELREF(t, p)	refcount_release(&(p)->refcnt)
801 
802 #define	SP_INITREF(p)	IPSEC_INITREF(SP, p)
803 #define	SP_ADDREF(p)	IPSEC_ADDREF(SP, p)
804 #define	SP_DELREF(p)	IPSEC_DELREF(SP, p)
805 
806 #define	SAH_INITREF(p)	IPSEC_INITREF(SAH, p)
807 #define	SAH_ADDREF(p)	IPSEC_ADDREF(SAH, p)
808 #define	SAH_DELREF(p)	IPSEC_DELREF(SAH, p)
809 
810 #define	SAV_INITREF(p)	IPSEC_INITREF(SAV, p)
811 #define	SAV_ADDREF(p)	IPSEC_ADDREF(SAV, p)
812 #define	SAV_DELREF(p)	IPSEC_DELREF(SAV, p)
813 
814 /*
815  * Update the refcnt while holding the SPTREE lock.
816  */
817 void
818 key_addref(struct secpolicy *sp)
819 {
820 
821 	SP_ADDREF(sp);
822 }
823 
824 /*
825  * Return 0 when there are known to be no SP's for the specified
826  * direction.  Otherwise return 1.  This is used by IPsec code
827  * to optimize performance.
828  */
829 int
830 key_havesp(u_int dir)
831 {
832 
833 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
834 		("invalid direction %u", dir));
835 	return (TAILQ_FIRST(&V_sptree[dir]) != NULL);
836 }
837 
838 int
839 key_havesp_any(void)
840 {
841 
842 	return (V_spd_size != 0);
843 }
844 
845 /*
846  * Allocate a single mbuf with a buffer of the desired length.  The buffer is
847  * pre-zeroed to help ensure that uninitialized pad bytes are not leaked.
848  */
849 static struct mbuf *
850 key_mget(u_int len)
851 {
852 	struct mbuf *m;
853 
854 	KASSERT(len <= MCLBYTES,
855 	    ("%s: invalid buffer length %u", __func__, len));
856 
857 	m = m_get2(len, M_NOWAIT, MT_DATA, M_PKTHDR);
858 	if (m == NULL)
859 		return (NULL);
860 	memset(mtod(m, void *), 0, len);
861 	return (m);
862 }
863 
864 /* %%% IPsec policy management */
865 /*
866  * Return current SPDB generation.
867  */
868 uint32_t
869 key_getspgen(void)
870 {
871 
872 	return (V_sp_genid);
873 }
874 
875 void
876 key_bumpspgen(void)
877 {
878 
879 	V_sp_genid++;
880 }
881 
882 static int
883 key_checksockaddrs(struct sockaddr *src, struct sockaddr *dst)
884 {
885 
886 	/* family match */
887 	if (src->sa_family != dst->sa_family)
888 		return (EINVAL);
889 	/* sa_len match */
890 	if (src->sa_len != dst->sa_len)
891 		return (EINVAL);
892 	switch (src->sa_family) {
893 #ifdef INET
894 	case AF_INET:
895 		if (src->sa_len != sizeof(struct sockaddr_in))
896 			return (EINVAL);
897 		break;
898 #endif
899 #ifdef INET6
900 	case AF_INET6:
901 		if (src->sa_len != sizeof(struct sockaddr_in6))
902 			return (EINVAL);
903 		break;
904 #endif
905 	default:
906 		return (EAFNOSUPPORT);
907 	}
908 	return (0);
909 }
910 
911 struct secpolicy *
912 key_do_allocsp(struct secpolicyindex *spidx, u_int dir)
913 {
914 	SPTREE_RLOCK_TRACKER;
915 	struct secpolicy *sp;
916 
917 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
918 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
919 		("invalid direction %u", dir));
920 
921 	SPTREE_RLOCK();
922 	TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
923 		if (key_cmpspidx_withmask(&sp->spidx, spidx)) {
924 			SP_ADDREF(sp);
925 			break;
926 		}
927 	}
928 	SPTREE_RUNLOCK();
929 	return (sp);
930 }
931 
932 /*
933  * allocating a SP for OUTBOUND or INBOUND packet.
934  * Must call key_freesp() later.
935  * OUT:	NULL:	not found
936  *	others:	found and return the pointer.
937  */
938 struct secpolicy *
939 key_allocsp(struct secpolicyindex *spidx, u_int dir)
940 {
941 	struct spdcache_entry *entry, *lastentry, *tmpentry;
942 	struct secpolicy *sp;
943 	uint32_t hashv;
944 	time_t ts;
945 	int nb_entries;
946 
947 	if (!SPDCACHE_ACTIVE()) {
948 		sp = key_do_allocsp(spidx, dir);
949 		goto out;
950 	}
951 
952 	hashv = SPDCACHE_HASHVAL(spidx);
953 	SPDCACHE_LOCK(hashv);
954 	nb_entries = 0;
955 	LIST_FOREACH_SAFE(entry, &V_spdcachehashtbl[hashv], chain, tmpentry) {
956 		/* Removed outdated entries */
957 		if (entry->sp != NULL &&
958 		    entry->sp->state == IPSEC_SPSTATE_DEAD) {
959 			LIST_REMOVE(entry, chain);
960 			spdcache_entry_free(entry);
961 			continue;
962 		}
963 
964 		nb_entries++;
965 		if (!key_cmpspidx_exactly(&entry->spidx, spidx)) {
966 			lastentry = entry;
967 			continue;
968 		}
969 
970 		sp = entry->sp;
971 		if (entry->sp != NULL)
972 			SP_ADDREF(sp);
973 
974 		/* IPSECSTAT_INC(ips_spdcache_hits); */
975 
976 		SPDCACHE_UNLOCK(hashv);
977 		goto out;
978 	}
979 
980 	/* IPSECSTAT_INC(ips_spdcache_misses); */
981 
982 	sp = key_do_allocsp(spidx, dir);
983 	entry = spdcache_entry_alloc(spidx, sp);
984 	if (entry != NULL) {
985 		if (nb_entries >= SPDCACHE_MAX_ENTRIES_PER_HASH) {
986 			LIST_REMOVE(lastentry, chain);
987 			spdcache_entry_free(lastentry);
988 		}
989 
990 		LIST_INSERT_HEAD(&V_spdcachehashtbl[hashv], entry, chain);
991 	}
992 
993 	SPDCACHE_UNLOCK(hashv);
994 
995 out:
996 	if (sp != NULL) {	/* found a SPD entry */
997 		ts = time_second;
998 		if (__predict_false(sp->lastused != ts))
999 			sp->lastused = ts;
1000 		KEYDBG(IPSEC_STAMP,
1001 		    printf("%s: return SP(%p)\n", __func__, sp));
1002 		KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1003 	} else {
1004 		KEYDBG(IPSEC_DATA,
1005 		    printf("%s: lookup failed for ", __func__);
1006 		    kdebug_secpolicyindex(spidx, NULL));
1007 	}
1008 	return (sp);
1009 }
1010 
1011 /*
1012  * Allocating an SA entry for an *INBOUND* or *OUTBOUND* TCP packet, signed
1013  * or should be signed by MD5 signature.
1014  * We don't use key_allocsa() for such lookups, because we don't know SPI.
1015  * Unlike ESP and AH protocols, SPI isn't transmitted in the TCP header with
1016  * signed packet. We use SADB only as storage for password.
1017  * OUT:	positive:	corresponding SA for given saidx found.
1018  *	NULL:		SA not found
1019  */
1020 struct secasvar *
1021 key_allocsa_tcpmd5(struct secasindex *saidx)
1022 {
1023 	SAHTREE_RLOCK_TRACKER;
1024 	struct secashead *sah;
1025 	struct secasvar *sav;
1026 
1027 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP,
1028 	    ("unexpected security protocol %u", saidx->proto));
1029 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TCPMD5,
1030 	    ("unexpected mode %u", saidx->mode));
1031 
1032 	SAHTREE_RLOCK();
1033 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
1034 		KEYDBG(IPSEC_DUMP,
1035 		    printf("%s: checking SAH\n", __func__);
1036 		    kdebug_secash(sah, "  "));
1037 		if (sah->saidx.proto != IPPROTO_TCP)
1038 			continue;
1039 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
1040 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
1041 			break;
1042 	}
1043 	if (sah != NULL) {
1044 		if (V_key_preferred_oldsa)
1045 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1046 		else
1047 			sav = TAILQ_FIRST(&sah->savtree_alive);
1048 		if (sav != NULL)
1049 			SAV_ADDREF(sav);
1050 	} else
1051 		sav = NULL;
1052 	SAHTREE_RUNLOCK();
1053 
1054 	if (sav != NULL) {
1055 		KEYDBG(IPSEC_STAMP,
1056 		    printf("%s: return SA(%p)\n", __func__, sav));
1057 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1058 	} else {
1059 		KEYDBG(IPSEC_STAMP,
1060 		    printf("%s: SA not found\n", __func__));
1061 		KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1062 	}
1063 	return (sav);
1064 }
1065 
1066 /*
1067  * Allocating an SA entry for an *OUTBOUND* packet.
1068  * OUT:	positive:	corresponding SA for given saidx found.
1069  *	NULL:		SA not found, but will be acquired, check *error
1070  *			for acquiring status.
1071  */
1072 struct secasvar *
1073 key_allocsa_policy(struct secpolicy *sp, const struct secasindex *saidx,
1074     int *error)
1075 {
1076 	SAHTREE_RLOCK_TRACKER;
1077 	struct secashead *sah;
1078 	struct secasvar *sav;
1079 
1080 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
1081 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TRANSPORT ||
1082 		saidx->mode == IPSEC_MODE_TUNNEL,
1083 		("unexpected policy %u", saidx->mode));
1084 
1085 	/*
1086 	 * We check new SA in the IPsec request because a different
1087 	 * SA may be involved each time this request is checked, either
1088 	 * because new SAs are being configured, or this request is
1089 	 * associated with an unconnected datagram socket, or this request
1090 	 * is associated with a system default policy.
1091 	 */
1092 	SAHTREE_RLOCK();
1093 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
1094 		KEYDBG(IPSEC_DUMP,
1095 		    printf("%s: checking SAH\n", __func__);
1096 		    kdebug_secash(sah, "  "));
1097 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID))
1098 			break;
1099 	}
1100 	if (sah != NULL) {
1101 		/*
1102 		 * Allocate the oldest SA available according to
1103 		 * draft-jenkins-ipsec-rekeying-03.
1104 		 */
1105 		if (V_key_preferred_oldsa)
1106 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1107 		else
1108 			sav = TAILQ_FIRST(&sah->savtree_alive);
1109 		if (sav != NULL)
1110 			SAV_ADDREF(sav);
1111 	} else
1112 		sav = NULL;
1113 	SAHTREE_RUNLOCK();
1114 
1115 	if (sav != NULL) {
1116 		*error = 0;
1117 		KEYDBG(IPSEC_STAMP,
1118 		    printf("%s: chosen SA(%p) for SP(%p)\n", __func__,
1119 			sav, sp));
1120 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1121 		return (sav); /* return referenced SA */
1122 	}
1123 
1124 	/* there is no SA */
1125 	*error = key_acquire(saidx, sp);
1126 	if ((*error) != 0)
1127 		ipseclog((LOG_DEBUG,
1128 		    "%s: error %d returned from key_acquire()\n",
1129 			__func__, *error));
1130 	KEYDBG(IPSEC_STAMP,
1131 	    printf("%s: acquire SA for SP(%p), error %d\n",
1132 		__func__, sp, *error));
1133 	KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
1134 	return (NULL);
1135 }
1136 
1137 /*
1138  * allocating a usable SA entry for a *INBOUND* packet.
1139  * Must call key_freesav() later.
1140  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
1141  *	NULL:		not found, or error occurred.
1142  *
1143  * According to RFC 2401 SA is uniquely identified by a triple SPI,
1144  * destination address, and security protocol. But according to RFC 4301,
1145  * SPI by itself suffices to specify an SA.
1146  *
1147  * Note that, however, we do need to keep source address in IPsec SA.
1148  * IKE specification and PF_KEY specification do assume that we
1149  * keep source address in IPsec SA.  We see a tricky situation here.
1150  */
1151 struct secasvar *
1152 key_allocsa(union sockaddr_union *dst, uint8_t proto, uint32_t spi)
1153 {
1154 	SAHTREE_RLOCK_TRACKER;
1155 	struct secasvar *sav;
1156 
1157 	IPSEC_ASSERT(proto == IPPROTO_ESP || proto == IPPROTO_AH ||
1158 	    proto == IPPROTO_IPCOMP, ("unexpected security protocol %u",
1159 	    proto));
1160 
1161 	SAHTREE_RLOCK();
1162 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
1163 		if (sav->spi == spi)
1164 			break;
1165 	}
1166 	/*
1167 	 * We use single SPI namespace for all protocols, so it is
1168 	 * impossible to have SPI duplicates in the SAVHASH.
1169 	 */
1170 	if (sav != NULL) {
1171 		if (sav->state != SADB_SASTATE_LARVAL &&
1172 		    sav->sah->saidx.proto == proto &&
1173 		    key_sockaddrcmp(&dst->sa,
1174 			&sav->sah->saidx.dst.sa, 0) == 0)
1175 			SAV_ADDREF(sav);
1176 		else
1177 			sav = NULL;
1178 	}
1179 	SAHTREE_RUNLOCK();
1180 
1181 	if (sav == NULL) {
1182 		KEYDBG(IPSEC_STAMP,
1183 		    char buf[IPSEC_ADDRSTRLEN];
1184 		    printf("%s: SA not found for spi %u proto %u dst %s\n",
1185 			__func__, ntohl(spi), proto, ipsec_address(dst, buf,
1186 			sizeof(buf))));
1187 	} else {
1188 		KEYDBG(IPSEC_STAMP,
1189 		    printf("%s: return SA(%p)\n", __func__, sav));
1190 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1191 	}
1192 	return (sav);
1193 }
1194 
1195 struct secasvar *
1196 key_allocsa_tunnel(union sockaddr_union *src, union sockaddr_union *dst,
1197     uint8_t proto)
1198 {
1199 	SAHTREE_RLOCK_TRACKER;
1200 	struct secasindex saidx;
1201 	struct secashead *sah;
1202 	struct secasvar *sav;
1203 
1204 	IPSEC_ASSERT(src != NULL, ("null src address"));
1205 	IPSEC_ASSERT(dst != NULL, ("null dst address"));
1206 
1207 	KEY_SETSECASIDX(proto, IPSEC_MODE_TUNNEL, 0, &src->sa,
1208 	    &dst->sa, &saidx);
1209 
1210 	sav = NULL;
1211 	SAHTREE_RLOCK();
1212 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
1213 		if (IPSEC_MODE_TUNNEL != sah->saidx.mode)
1214 			continue;
1215 		if (proto != sah->saidx.proto)
1216 			continue;
1217 		if (key_sockaddrcmp(&src->sa, &sah->saidx.src.sa, 0) != 0)
1218 			continue;
1219 		if (key_sockaddrcmp(&dst->sa, &sah->saidx.dst.sa, 0) != 0)
1220 			continue;
1221 		/* XXXAE: is key_preferred_oldsa reasonably?*/
1222 		if (V_key_preferred_oldsa)
1223 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1224 		else
1225 			sav = TAILQ_FIRST(&sah->savtree_alive);
1226 		if (sav != NULL) {
1227 			SAV_ADDREF(sav);
1228 			break;
1229 		}
1230 	}
1231 	SAHTREE_RUNLOCK();
1232 	KEYDBG(IPSEC_STAMP,
1233 	    printf("%s: return SA(%p)\n", __func__, sav));
1234 	if (sav != NULL)
1235 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1236 	return (sav);
1237 }
1238 
1239 /*
1240  * Must be called after calling key_allocsp().
1241  */
1242 void
1243 key_freesp(struct secpolicy **spp)
1244 {
1245 	struct secpolicy *sp = *spp;
1246 
1247 	IPSEC_ASSERT(sp != NULL, ("null sp"));
1248 	if (SP_DELREF(sp) == 0)
1249 		return;
1250 
1251 	KEYDBG(IPSEC_STAMP,
1252 	    printf("%s: last reference to SP(%p)\n", __func__, sp));
1253 	KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1254 
1255 	*spp = NULL;
1256 #ifdef IPSEC_OFFLOAD
1257 	KASSERT(CK_LIST_EMPTY(&sp->accel_ifps),
1258 	    ("key_freesp: sp %p still offloaded", sp));
1259 	free(__DECONST(char *, sp->accel_ifname), M_IPSEC_MISC);
1260 #endif
1261 	while (sp->tcount > 0)
1262 		ipsec_delisr(sp->req[--sp->tcount]);
1263 	free(sp, M_IPSEC_SP);
1264 }
1265 
1266 static void
1267 key_unlink(struct secpolicy *sp)
1268 {
1269 	SPTREE_WLOCK();
1270 	key_detach(sp);
1271 	SPTREE_WUNLOCK();
1272 	if (SPDCACHE_ENABLED())
1273 		spdcache_clear();
1274 	ipsec_accel_sync();
1275 	key_freesp(&sp);
1276 }
1277 
1278 static void
1279 key_detach(struct secpolicy *sp)
1280 {
1281 	IPSEC_ASSERT(sp->spidx.dir == IPSEC_DIR_INBOUND ||
1282 	    sp->spidx.dir == IPSEC_DIR_OUTBOUND,
1283 	    ("invalid direction %u", sp->spidx.dir));
1284 	SPTREE_WLOCK_ASSERT();
1285 
1286 	KEYDBG(KEY_STAMP,
1287 	    printf("%s: SP(%p)\n", __func__, sp));
1288 	if (sp->state != IPSEC_SPSTATE_ALIVE) {
1289 		/* SP is already unlinked */
1290 		return;
1291 	}
1292 	sp->state = IPSEC_SPSTATE_DEAD;
1293 	ipsec_accel_spddel(sp);
1294 	TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
1295 	V_spd_size--;
1296 	LIST_REMOVE(sp, idhash);
1297 	V_sp_genid++;
1298 }
1299 
1300 /*
1301  * insert a secpolicy into the SP database. Lower priorities first
1302  */
1303 static void
1304 key_insertsp(struct secpolicy *newsp)
1305 {
1306 	struct secpolicy *sp;
1307 
1308 	SPTREE_WLOCK_ASSERT();
1309 	TAILQ_FOREACH(sp, &V_sptree[newsp->spidx.dir], chain) {
1310 		if (newsp->priority < sp->priority) {
1311 			TAILQ_INSERT_BEFORE(sp, newsp, chain);
1312 			goto done;
1313 		}
1314 	}
1315 	TAILQ_INSERT_TAIL(&V_sptree[newsp->spidx.dir], newsp, chain);
1316 done:
1317 	LIST_INSERT_HEAD(SPHASH_HASH(newsp->id), newsp, idhash);
1318 	newsp->state = IPSEC_SPSTATE_ALIVE;
1319 	V_spd_size++;
1320 	V_sp_genid++;
1321 	ipsec_accel_spdadd(newsp, NULL);
1322 }
1323 
1324 /*
1325  * Insert a bunch of VTI secpolicies into the SPDB.
1326  * We keep VTI policies in the separate list due to following reasons:
1327  * 1) they should be immutable to user's or some deamon's attempts to
1328  *    delete. The only way delete such policies - destroy or unconfigure
1329  *    corresponding virtual inteface.
1330  * 2) such policies have traffic selector that matches all traffic per
1331  *    address family.
1332  * Since all VTI policies have the same priority, we don't care about
1333  * policies order.
1334  */
1335 int
1336 key_register_ifnet(struct secpolicy **spp, u_int count)
1337 {
1338 	struct mbuf *m;
1339 	u_int i;
1340 
1341 	SPTREE_WLOCK();
1342 	/*
1343 	 * First of try to acquire id for each SP.
1344 	 */
1345 	for (i = 0; i < count; i++) {
1346 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1347 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1348 		    ("invalid direction %u", spp[i]->spidx.dir));
1349 
1350 		if ((spp[i]->id = key_getnewspid()) == 0) {
1351 			SPTREE_WUNLOCK();
1352 			return (EAGAIN);
1353 		}
1354 	}
1355 	for (i = 0; i < count; i++) {
1356 		TAILQ_INSERT_TAIL(&V_sptree_ifnet[spp[i]->spidx.dir],
1357 		    spp[i], chain);
1358 		/*
1359 		 * NOTE: despite the fact that we keep VTI SP in the
1360 		 * separate list, SPHASH contains policies from both
1361 		 * sources. Thus SADB_X_SPDGET will correctly return
1362 		 * SP by id, because it uses SPHASH for lookups.
1363 		 */
1364 		LIST_INSERT_HEAD(SPHASH_HASH(spp[i]->id), spp[i], idhash);
1365 		spp[i]->state = IPSEC_SPSTATE_IFNET;
1366 		ipsec_accel_spdadd(spp[i], NULL);
1367 	}
1368 	SPTREE_WUNLOCK();
1369 	/*
1370 	 * Notify user processes about new SP.
1371 	 */
1372 	for (i = 0; i < count; i++) {
1373 		m = key_setdumpsp(spp[i], SADB_X_SPDADD, 0, 0);
1374 		if (m != NULL)
1375 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1376 	}
1377 	return (0);
1378 }
1379 
1380 void
1381 key_unregister_ifnet(struct secpolicy **spp, u_int count)
1382 {
1383 	struct mbuf *m;
1384 	u_int i;
1385 
1386 	SPTREE_WLOCK();
1387 	for (i = 0; i < count; i++) {
1388 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1389 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1390 		    ("invalid direction %u", spp[i]->spidx.dir));
1391 
1392 		if (spp[i]->state != IPSEC_SPSTATE_IFNET)
1393 			continue;
1394 		spp[i]->state = IPSEC_SPSTATE_DEAD;
1395 		ipsec_accel_spddel(spp[i]);
1396 		TAILQ_REMOVE(&V_sptree_ifnet[spp[i]->spidx.dir],
1397 		    spp[i], chain);
1398 		V_spd_size--;
1399 		LIST_REMOVE(spp[i], idhash);
1400 	}
1401 	SPTREE_WUNLOCK();
1402 	if (SPDCACHE_ENABLED())
1403 		spdcache_clear();
1404 	ipsec_accel_sync();
1405 
1406 	for (i = 0; i < count; i++) {
1407 		m = key_setdumpsp(spp[i], SADB_X_SPDDELETE, 0, 0);
1408 		if (m != NULL)
1409 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1410 	}
1411 }
1412 
1413 /*
1414  * Must be called after calling key_allocsa().
1415  * This function is called by key_freesp() to free some SA allocated
1416  * for a policy.
1417  */
1418 void
1419 key_freesav(struct secasvar **psav)
1420 {
1421 	struct secasvar *sav = *psav;
1422 
1423 	IPSEC_ASSERT(sav != NULL, ("null sav"));
1424 	CURVNET_ASSERT_SET();
1425 	if (SAV_DELREF(sav) == 0)
1426 		return;
1427 
1428 	KEYDBG(IPSEC_STAMP,
1429 	    printf("%s: last reference to SA(%p)\n", __func__, sav));
1430 
1431 	*psav = NULL;
1432 	key_delsav(sav);
1433 }
1434 
1435 /*
1436  * Unlink SA from SAH and SPI hash under SAHTREE_WLOCK.
1437  * Expect that SA has extra reference due to lookup.
1438  * Release this references, also release SAH reference after unlink.
1439  */
1440 static void
1441 key_unlinksav(struct secasvar *sav)
1442 {
1443 	struct secashead *sah;
1444 
1445 	KEYDBG(KEY_STAMP,
1446 	    printf("%s: SA(%p)\n", __func__, sav));
1447 
1448 	CURVNET_ASSERT_SET();
1449 	SAHTREE_UNLOCK_ASSERT();
1450 	SAHTREE_WLOCK();
1451 	if (sav->state == SADB_SASTATE_DEAD) {
1452 		/* SA is already unlinked */
1453 		SAHTREE_WUNLOCK();
1454 		return;
1455 	}
1456 	/* Unlink from SAH */
1457 	if (sav->state == SADB_SASTATE_LARVAL)
1458 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
1459 	else
1460 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
1461 	/* Unlink from SPI hash */
1462 	LIST_REMOVE(sav, spihash);
1463 	sav->state = SADB_SASTATE_DEAD;
1464 	ipsec_accel_forget_sav(sav);
1465 	sah = sav->sah;
1466 	SAHTREE_WUNLOCK();
1467 	key_freesav(&sav);
1468 	/* Since we are unlinked, release reference to SAH */
1469 	key_freesah(&sah);
1470 }
1471 
1472 /* %%% SPD management */
1473 /*
1474  * search SPD
1475  * OUT:	NULL	: not found
1476  *	others	: found, pointer to a SP.
1477  */
1478 static struct secpolicy *
1479 key_getsp(struct secpolicyindex *spidx)
1480 {
1481 	SPTREE_RLOCK_TRACKER;
1482 	struct secpolicy *sp;
1483 
1484 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
1485 
1486 	SPTREE_RLOCK();
1487 	TAILQ_FOREACH(sp, &V_sptree[spidx->dir], chain) {
1488 		if (key_cmpspidx_exactly(spidx, &sp->spidx)) {
1489 			SP_ADDREF(sp);
1490 			break;
1491 		}
1492 	}
1493 	SPTREE_RUNLOCK();
1494 
1495 	return sp;
1496 }
1497 
1498 /*
1499  * get SP by index.
1500  * OUT:	NULL	: not found
1501  *	others	: found, pointer to referenced SP.
1502  */
1503 static struct secpolicy *
1504 key_getspbyid(uint32_t id)
1505 {
1506 	SPTREE_RLOCK_TRACKER;
1507 	struct secpolicy *sp;
1508 
1509 	SPTREE_RLOCK();
1510 	LIST_FOREACH(sp, SPHASH_HASH(id), idhash) {
1511 		if (sp->id == id) {
1512 			SP_ADDREF(sp);
1513 			break;
1514 		}
1515 	}
1516 	SPTREE_RUNLOCK();
1517 	return (sp);
1518 }
1519 
1520 struct secpolicy *
1521 key_newsp(void)
1522 {
1523 	struct secpolicy *sp;
1524 
1525 	sp = malloc(sizeof(*sp), M_IPSEC_SP, M_NOWAIT | M_ZERO);
1526 	if (sp != NULL)
1527 		SP_INITREF(sp);
1528 	return (sp);
1529 }
1530 
1531 struct ipsecrequest *
1532 ipsec_newisr(void)
1533 {
1534 
1535 	return (malloc(sizeof(struct ipsecrequest), M_IPSEC_SR,
1536 	    M_NOWAIT | M_ZERO));
1537 }
1538 
1539 void
1540 ipsec_delisr(struct ipsecrequest *p)
1541 {
1542 
1543 	free(p, M_IPSEC_SR);
1544 }
1545 
1546 /*
1547  * create secpolicy structure from sadb_x_policy structure.
1548  * NOTE: `state', `secpolicyindex' and 'id' in secpolicy structure
1549  * are not set, so must be set properly later.
1550  */
1551 struct secpolicy *
1552 key_msg2sp(struct sadb_x_policy *xpl0, size_t len, int *error)
1553 {
1554 	struct secpolicy *newsp;
1555 
1556 	IPSEC_ASSERT(xpl0 != NULL, ("null xpl0"));
1557 	IPSEC_ASSERT(len >= sizeof(*xpl0), ("policy too short: %zu", len));
1558 
1559 	if (len != PFKEY_EXTLEN(xpl0)) {
1560 		ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n", __func__));
1561 		*error = EINVAL;
1562 		return NULL;
1563 	}
1564 
1565 	if ((newsp = key_newsp()) == NULL) {
1566 		*error = ENOBUFS;
1567 		return NULL;
1568 	}
1569 
1570 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1571 	newsp->policy = xpl0->sadb_x_policy_type;
1572 	newsp->priority = xpl0->sadb_x_policy_priority;
1573 	newsp->tcount = 0;
1574 
1575 	/* check policy */
1576 	switch (xpl0->sadb_x_policy_type) {
1577 	case IPSEC_POLICY_DISCARD:
1578 	case IPSEC_POLICY_NONE:
1579 	case IPSEC_POLICY_ENTRUST:
1580 	case IPSEC_POLICY_BYPASS:
1581 		break;
1582 
1583 	case IPSEC_POLICY_IPSEC:
1584 	    {
1585 		struct sadb_x_ipsecrequest *xisr;
1586 		struct ipsecrequest *isr;
1587 		int tlen;
1588 
1589 		/* validity check */
1590 		if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1591 			ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n",
1592 				__func__));
1593 			key_freesp(&newsp);
1594 			*error = EINVAL;
1595 			return NULL;
1596 		}
1597 
1598 		tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1599 		xisr = (struct sadb_x_ipsecrequest *)(xpl0 + 1);
1600 
1601 		while (tlen > 0) {
1602 			/* length check */
1603 			if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr) ||
1604 			    xisr->sadb_x_ipsecrequest_len > tlen) {
1605 				ipseclog((LOG_DEBUG, "%s: invalid ipsecrequest "
1606 					"length.\n", __func__));
1607 				key_freesp(&newsp);
1608 				*error = EINVAL;
1609 				return NULL;
1610 			}
1611 
1612 			if (newsp->tcount >= IPSEC_MAXREQ) {
1613 				ipseclog((LOG_DEBUG,
1614 				    "%s: too many ipsecrequests.\n",
1615 				    __func__));
1616 				key_freesp(&newsp);
1617 				*error = EINVAL;
1618 				return (NULL);
1619 			}
1620 
1621 			/* allocate request buffer */
1622 			/* NB: data structure is zero'd */
1623 			isr = ipsec_newisr();
1624 			if (isr == NULL) {
1625 				ipseclog((LOG_DEBUG,
1626 				    "%s: No more memory.\n", __func__));
1627 				key_freesp(&newsp);
1628 				*error = ENOBUFS;
1629 				return NULL;
1630 			}
1631 
1632 			newsp->req[newsp->tcount++] = isr;
1633 
1634 			/* set values */
1635 			switch (xisr->sadb_x_ipsecrequest_proto) {
1636 			case IPPROTO_ESP:
1637 			case IPPROTO_AH:
1638 			case IPPROTO_IPCOMP:
1639 				break;
1640 			default:
1641 				ipseclog((LOG_DEBUG,
1642 				    "%s: invalid proto type=%u\n", __func__,
1643 				    xisr->sadb_x_ipsecrequest_proto));
1644 				key_freesp(&newsp);
1645 				*error = EPROTONOSUPPORT;
1646 				return NULL;
1647 			}
1648 			isr->saidx.proto =
1649 			    (uint8_t)xisr->sadb_x_ipsecrequest_proto;
1650 
1651 			switch (xisr->sadb_x_ipsecrequest_mode) {
1652 			case IPSEC_MODE_TRANSPORT:
1653 			case IPSEC_MODE_TUNNEL:
1654 				break;
1655 			case IPSEC_MODE_ANY:
1656 			default:
1657 				ipseclog((LOG_DEBUG,
1658 				    "%s: invalid mode=%u\n", __func__,
1659 				    xisr->sadb_x_ipsecrequest_mode));
1660 				key_freesp(&newsp);
1661 				*error = EINVAL;
1662 				return NULL;
1663 			}
1664 			isr->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1665 
1666 			switch (xisr->sadb_x_ipsecrequest_level) {
1667 			case IPSEC_LEVEL_DEFAULT:
1668 			case IPSEC_LEVEL_USE:
1669 			case IPSEC_LEVEL_REQUIRE:
1670 				break;
1671 			case IPSEC_LEVEL_UNIQUE:
1672 				/* validity check */
1673 				/*
1674 				 * If range violation of reqid, kernel will
1675 				 * update it, don't refuse it.
1676 				 */
1677 				if (xisr->sadb_x_ipsecrequest_reqid
1678 						> IPSEC_MANUAL_REQID_MAX) {
1679 					ipseclog((LOG_DEBUG,
1680 					    "%s: reqid=%d range "
1681 					    "violation, updated by kernel.\n",
1682 					    __func__,
1683 					    xisr->sadb_x_ipsecrequest_reqid));
1684 					xisr->sadb_x_ipsecrequest_reqid = 0;
1685 				}
1686 
1687 				/* allocate new reqid id if reqid is zero. */
1688 				if (xisr->sadb_x_ipsecrequest_reqid == 0) {
1689 					u_int32_t reqid;
1690 					if ((reqid = key_newreqid()) == 0) {
1691 						key_freesp(&newsp);
1692 						*error = ENOBUFS;
1693 						return NULL;
1694 					}
1695 					isr->saidx.reqid = reqid;
1696 					xisr->sadb_x_ipsecrequest_reqid = reqid;
1697 				} else {
1698 				/* set it for manual keying. */
1699 					isr->saidx.reqid =
1700 					    xisr->sadb_x_ipsecrequest_reqid;
1701 				}
1702 				break;
1703 
1704 			default:
1705 				ipseclog((LOG_DEBUG, "%s: invalid level=%u\n",
1706 					__func__,
1707 					xisr->sadb_x_ipsecrequest_level));
1708 				key_freesp(&newsp);
1709 				*error = EINVAL;
1710 				return NULL;
1711 			}
1712 			isr->level = xisr->sadb_x_ipsecrequest_level;
1713 
1714 			/* set IP addresses if there */
1715 			if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1716 				struct sockaddr *paddr;
1717 
1718 				len = tlen - sizeof(*xisr);
1719 				paddr = (struct sockaddr *)(xisr + 1);
1720 				/* validity check */
1721 				if (len < sizeof(struct sockaddr) ||
1722 				    len < 2 * paddr->sa_len ||
1723 				    paddr->sa_len > sizeof(isr->saidx.src)) {
1724 					ipseclog((LOG_DEBUG, "%s: invalid "
1725 						"request address length.\n",
1726 						__func__));
1727 					key_freesp(&newsp);
1728 					*error = EINVAL;
1729 					return NULL;
1730 				}
1731 				/*
1732 				 * Request length should be enough to keep
1733 				 * source and destination addresses.
1734 				 */
1735 				if (xisr->sadb_x_ipsecrequest_len <
1736 				    sizeof(*xisr) + 2 * paddr->sa_len) {
1737 					ipseclog((LOG_DEBUG, "%s: invalid "
1738 					    "ipsecrequest length.\n",
1739 					    __func__));
1740 					key_freesp(&newsp);
1741 					*error = EINVAL;
1742 					return (NULL);
1743 				}
1744 				bcopy(paddr, &isr->saidx.src, paddr->sa_len);
1745 				paddr = (struct sockaddr *)((caddr_t)paddr +
1746 				    paddr->sa_len);
1747 
1748 				/* validity check */
1749 				if (paddr->sa_len !=
1750 				    isr->saidx.src.sa.sa_len) {
1751 					ipseclog((LOG_DEBUG, "%s: invalid "
1752 						"request address length.\n",
1753 						__func__));
1754 					key_freesp(&newsp);
1755 					*error = EINVAL;
1756 					return NULL;
1757 				}
1758 				/* AF family should match */
1759 				if (paddr->sa_family !=
1760 				    isr->saidx.src.sa.sa_family) {
1761 					ipseclog((LOG_DEBUG, "%s: address "
1762 					    "family doesn't match.\n",
1763 						__func__));
1764 					key_freesp(&newsp);
1765 					*error = EINVAL;
1766 					return (NULL);
1767 				}
1768 				bcopy(paddr, &isr->saidx.dst, paddr->sa_len);
1769 			} else {
1770 				/*
1771 				 * Addresses for TUNNEL mode requests are
1772 				 * mandatory.
1773 				 */
1774 				if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1775 					ipseclog((LOG_DEBUG, "%s: missing "
1776 					    "request addresses.\n", __func__));
1777 					key_freesp(&newsp);
1778 					*error = EINVAL;
1779 					return (NULL);
1780 				}
1781 			}
1782 			tlen -= xisr->sadb_x_ipsecrequest_len;
1783 
1784 			/* validity check */
1785 			if (tlen < 0) {
1786 				ipseclog((LOG_DEBUG, "%s: becoming tlen < 0.\n",
1787 					__func__));
1788 				key_freesp(&newsp);
1789 				*error = EINVAL;
1790 				return NULL;
1791 			}
1792 
1793 			xisr = (struct sadb_x_ipsecrequest *)((caddr_t)xisr
1794 			                 + xisr->sadb_x_ipsecrequest_len);
1795 		}
1796 		/* XXXAE: LARVAL SP */
1797 		if (newsp->tcount < 1) {
1798 			ipseclog((LOG_DEBUG, "%s: valid IPSEC transforms "
1799 			    "not found.\n", __func__));
1800 			key_freesp(&newsp);
1801 			*error = EINVAL;
1802 			return (NULL);
1803 		}
1804 	    }
1805 		break;
1806 	default:
1807 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
1808 		key_freesp(&newsp);
1809 		*error = EINVAL;
1810 		return NULL;
1811 	}
1812 
1813 	*error = 0;
1814 	return (newsp);
1815 }
1816 
1817 uint32_t
1818 key_newreqid(void)
1819 {
1820 	static uint32_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1821 
1822 	if (auto_reqid == ~0)
1823 		auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1824 	else
1825 		auto_reqid++;
1826 
1827 	/* XXX should be unique check */
1828 	return (auto_reqid);
1829 }
1830 
1831 /*
1832  * copy secpolicy struct to sadb_x_policy structure indicated.
1833  */
1834 static struct mbuf *
1835 key_sp2mbuf(struct secpolicy *sp)
1836 {
1837 	struct mbuf *m;
1838 	size_t tlen;
1839 
1840 	tlen = key_getspreqmsglen(sp);
1841 	m = m_get2(tlen, M_NOWAIT, MT_DATA, 0);
1842 	if (m == NULL)
1843 		return (NULL);
1844 	m_align(m, tlen);
1845 	m->m_len = tlen;
1846 	if (key_sp2msg(sp, m->m_data, &tlen) != 0) {
1847 		m_freem(m);
1848 		return (NULL);
1849 	}
1850 	return (m);
1851 }
1852 
1853 int
1854 key_sp2msg(struct secpolicy *sp, void *request, size_t *len)
1855 {
1856 	struct sadb_x_ipsecrequest *xisr;
1857 	struct sadb_x_policy *xpl;
1858 	struct ipsecrequest *isr;
1859 	size_t xlen, ilen;
1860 	caddr_t p;
1861 	int error, i;
1862 #ifdef IPSEC_OFFLOAD
1863 	struct sadb_x_if_hw_offl *xif;
1864 #endif
1865 
1866 	IPSEC_ASSERT(sp != NULL, ("null policy"));
1867 
1868 	xlen = sizeof(*xpl);
1869 	if (*len < xlen)
1870 		return (EINVAL);
1871 
1872 	error = 0;
1873 	bzero(request, *len);
1874 	xpl = (struct sadb_x_policy *)request;
1875 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
1876 	xpl->sadb_x_policy_type = sp->policy;
1877 	xpl->sadb_x_policy_dir = sp->spidx.dir;
1878 	xpl->sadb_x_policy_id = sp->id;
1879 	xpl->sadb_x_policy_priority = sp->priority;
1880 	switch (sp->state) {
1881 	case IPSEC_SPSTATE_IFNET:
1882 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_IFNET;
1883 		break;
1884 	case IPSEC_SPSTATE_PCB:
1885 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_PCB;
1886 		break;
1887 	default:
1888 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_GLOBAL;
1889 	}
1890 
1891 	/* if is the policy for ipsec ? */
1892 	if (sp->policy == IPSEC_POLICY_IPSEC) {
1893 		p = (caddr_t)xpl + sizeof(*xpl);
1894 		for (i = 0; i < sp->tcount; i++) {
1895 			isr = sp->req[i];
1896 			ilen = PFKEY_ALIGN8(sizeof(*xisr) +
1897 			    isr->saidx.src.sa.sa_len +
1898 			    isr->saidx.dst.sa.sa_len);
1899 			xlen += ilen;
1900 			if (xlen > *len) {
1901 				error = ENOBUFS;
1902 				/* Calculate needed size */
1903 				continue;
1904 			}
1905 			xisr = (struct sadb_x_ipsecrequest *)p;
1906 			xisr->sadb_x_ipsecrequest_len = ilen;
1907 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
1908 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
1909 			xisr->sadb_x_ipsecrequest_level = isr->level;
1910 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
1911 
1912 			p += sizeof(*xisr);
1913 			bcopy(&isr->saidx.src, p, isr->saidx.src.sa.sa_len);
1914 			p += isr->saidx.src.sa.sa_len;
1915 			bcopy(&isr->saidx.dst, p, isr->saidx.dst.sa.sa_len);
1916 			p += isr->saidx.dst.sa.sa_len;
1917 		}
1918 	}
1919 	xpl->sadb_x_policy_len = PFKEY_UNIT64(xlen);
1920 #ifdef IPSEC_OFFLOAD
1921 	if (error == 0 && sp->accel_ifname != NULL) {
1922 		xif = (struct sadb_x_if_hw_offl *)(xpl + 1);
1923 		bzero(xif, sizeof(*xif));
1924 		xif->sadb_x_if_hw_offl_len = PFKEY_UNIT64(sizeof(*xif));
1925 		xif->sadb_x_if_hw_offl_exttype = SADB_X_EXT_IF_HW_OFFL;
1926 		xif->sadb_x_if_hw_offl_flags = 0;
1927 		strncpy(xif->sadb_x_if_hw_offl_if, sp->accel_ifname,
1928 		    sizeof(xif->sadb_x_if_hw_offl_if));
1929 		xlen += sizeof(*xif);
1930 	}
1931 #endif
1932 	if (error == 0)
1933 		*len = xlen;
1934 	else
1935 		*len = sizeof(*xpl);
1936 	return (error);
1937 }
1938 
1939 /* m will not be freed nor modified */
1940 static struct mbuf *
1941 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
1942     int ndeep, int nitem, ...)
1943 {
1944 	va_list ap;
1945 	int idx;
1946 	int i;
1947 	struct mbuf *result = NULL, *n;
1948 	int len;
1949 
1950 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1951 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1952 
1953 	va_start(ap, nitem);
1954 	for (i = 0; i < nitem; i++) {
1955 		idx = va_arg(ap, int);
1956 		if (idx < 0 || idx > SADB_EXT_MAX)
1957 			goto fail;
1958 		/* don't attempt to pull empty extension */
1959 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
1960 			continue;
1961 		if (idx != SADB_EXT_RESERVED  &&
1962 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
1963 			continue;
1964 
1965 		if (idx == SADB_EXT_RESERVED) {
1966 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
1967 
1968 			IPSEC_ASSERT(len <= MHLEN, ("header too big %u", len));
1969 
1970 			MGETHDR(n, M_NOWAIT, MT_DATA);
1971 			if (!n)
1972 				goto fail;
1973 			n->m_len = len;
1974 			n->m_next = NULL;
1975 			m_copydata(m, 0, sizeof(struct sadb_msg),
1976 			    mtod(n, caddr_t));
1977 		} else if (i < ndeep) {
1978 			len = mhp->extlen[idx];
1979 			n = m_get2(len, M_NOWAIT, MT_DATA, 0);
1980 			if (n == NULL)
1981 				goto fail;
1982 			m_align(n, len);
1983 			n->m_len = len;
1984 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
1985 			    mtod(n, caddr_t));
1986 		} else {
1987 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
1988 			    M_NOWAIT);
1989 		}
1990 		if (n == NULL)
1991 			goto fail;
1992 
1993 		if (result)
1994 			m_cat(result, n);
1995 		else
1996 			result = n;
1997 	}
1998 	va_end(ap);
1999 
2000 	if ((result->m_flags & M_PKTHDR) != 0) {
2001 		result->m_pkthdr.len = 0;
2002 		for (n = result; n; n = n->m_next)
2003 			result->m_pkthdr.len += n->m_len;
2004 	}
2005 
2006 	return result;
2007 
2008 fail:
2009 	m_freem(result);
2010 	va_end(ap);
2011 	return NULL;
2012 }
2013 
2014 /*
2015  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
2016  * add an entry to SP database, when received
2017  *   <base, address(SD), (lifetime(H),) policy>
2018  * from the user(?).
2019  * Adding to SP database,
2020  * and send
2021  *   <base, address(SD), (lifetime(H),) policy>
2022  * to the socket which was send.
2023  *
2024  * SPDADD set a unique policy entry.
2025  * SPDSETIDX like SPDADD without a part of policy requests.
2026  * SPDUPDATE replace a unique policy entry.
2027  *
2028  * XXXAE: serialize this in PF_KEY to avoid races.
2029  * m will always be freed.
2030  */
2031 static int
2032 key_spdadd(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2033 {
2034 	struct secpolicyindex spidx;
2035 	struct sadb_address *src0, *dst0;
2036 	struct sadb_x_policy *xpl0, *xpl;
2037 	struct sadb_lifetime *lft = NULL;
2038 	struct secpolicy *newsp, *oldsp;
2039 	int error;
2040 
2041 	IPSEC_ASSERT(so != NULL, ("null socket"));
2042 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2043 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2044 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2045 
2046 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
2047 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
2048 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
2049 		ipseclog((LOG_DEBUG,
2050 		    "%s: invalid message: missing required header.\n",
2051 		    __func__));
2052 		return key_senderror(so, m, EINVAL);
2053 	}
2054 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
2055 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
2056 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2057 		ipseclog((LOG_DEBUG,
2058 		    "%s: invalid message: wrong header size.\n", __func__));
2059 		return key_senderror(so, m, EINVAL);
2060 	}
2061 	if (!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD)) {
2062 		if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD)) {
2063 			ipseclog((LOG_DEBUG,
2064 			    "%s: invalid message: wrong header size.\n",
2065 			    __func__));
2066 			return key_senderror(so, m, EINVAL);
2067 		}
2068 		lft = (struct sadb_lifetime *)mhp->ext[SADB_EXT_LIFETIME_HARD];
2069 	}
2070 
2071 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
2072 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
2073 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
2074 
2075 	/* check the direciton */
2076 	switch (xpl0->sadb_x_policy_dir) {
2077 	case IPSEC_DIR_INBOUND:
2078 	case IPSEC_DIR_OUTBOUND:
2079 		break;
2080 	default:
2081 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2082 		return key_senderror(so, m, EINVAL);
2083 	}
2084 	/* key_spdadd() accepts DISCARD, NONE and IPSEC. */
2085 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2086 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2087 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2088 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2089 		return key_senderror(so, m, EINVAL);
2090 	}
2091 
2092 	/* policy requests are mandatory when action is ipsec. */
2093 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
2094 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
2095 		ipseclog((LOG_DEBUG,
2096 		    "%s: policy requests required.\n", __func__));
2097 		return key_senderror(so, m, EINVAL);
2098 	}
2099 
2100 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
2101 	    (struct sockaddr *)(dst0 + 1));
2102 	if (error != 0 ||
2103 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2104 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2105 		return key_senderror(so, m, error);
2106 	}
2107 	/* make secindex */
2108 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2109 	                src0 + 1,
2110 	                dst0 + 1,
2111 	                src0->sadb_address_prefixlen,
2112 	                dst0->sadb_address_prefixlen,
2113 	                src0->sadb_address_proto,
2114 	                &spidx);
2115 	/* Checking there is SP already or not. */
2116 	oldsp = key_getsp(&spidx);
2117 	if (oldsp != NULL) {
2118 		if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2119 			KEYDBG(KEY_STAMP,
2120 			    printf("%s: unlink SP(%p) for SPDUPDATE\n",
2121 				__func__, oldsp));
2122 			KEYDBG(KEY_DATA, kdebug_secpolicy(oldsp));
2123 		} else {
2124 			key_freesp(&oldsp);
2125 			ipseclog((LOG_DEBUG,
2126 			    "%s: a SP entry exists already.\n", __func__));
2127 			return (key_senderror(so, m, EEXIST));
2128 		}
2129 	}
2130 
2131 	/* allocate new SP entry */
2132 	if ((newsp = key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error)) == NULL) {
2133 		if (oldsp != NULL) {
2134 			key_unlink(oldsp);
2135 			key_freesp(&oldsp); /* second for our reference */
2136 		}
2137 		return key_senderror(so, m, error);
2138 	}
2139 
2140 	newsp->lastused = newsp->created = time_second;
2141 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
2142 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
2143 	bcopy(&spidx, &newsp->spidx, sizeof(spidx));
2144 #ifdef IPSEC_OFFLOAD
2145 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_IF_HW_OFFL) &&
2146 	    !SADB_CHECKLEN(mhp, SADB_X_EXT_IF_HW_OFFL)) {
2147 		struct sadb_x_if_hw_offl *xof;
2148 
2149 		xof = (struct sadb_x_if_hw_offl *)mhp->ext[
2150 		    SADB_X_EXT_IF_HW_OFFL];
2151 		newsp->accel_ifname = malloc(sizeof(xof->sadb_x_if_hw_offl_if),
2152 		    M_IPSEC_MISC, M_NOWAIT);
2153 		if (newsp->accel_ifname == NULL) {
2154 			ipseclog((LOG_DEBUG, "%s: cannot alloc accel_ifname.\n",
2155 			    __func__));
2156 			key_freesp(&newsp);
2157 			return (key_senderror(so, m, error));
2158 		}
2159 		strncpy(__DECONST(char *, newsp->accel_ifname),
2160 		    xof->sadb_x_if_hw_offl_if,
2161 		    sizeof(xof->sadb_x_if_hw_offl_if));
2162 	}
2163 
2164 #endif
2165 
2166 	SPTREE_WLOCK();
2167 	if ((newsp->id = key_getnewspid()) == 0) {
2168 		if (oldsp != NULL)
2169 			key_detach(oldsp);
2170 		SPTREE_WUNLOCK();
2171 		if (oldsp != NULL) {
2172 			ipsec_accel_sync();
2173 			key_freesp(&oldsp); /* first for key_detach */
2174 			IPSEC_ASSERT(oldsp != NULL, ("null oldsp: refcount bug"));
2175 			key_freesp(&oldsp); /* second for our reference */
2176 			if (SPDCACHE_ENABLED()) /* refresh cache because of key_detach */
2177 				spdcache_clear();
2178 		}
2179 		key_freesp(&newsp);
2180 		return key_senderror(so, m, ENOBUFS);
2181 	}
2182 	if (oldsp != NULL)
2183 		key_detach(oldsp);
2184 	key_insertsp(newsp);
2185 	SPTREE_WUNLOCK();
2186 	if (oldsp != NULL) {
2187 		ipsec_accel_sync();
2188 		key_freesp(&oldsp); /* first for key_detach */
2189 		IPSEC_ASSERT(oldsp != NULL, ("null oldsp: refcount bug"));
2190 		key_freesp(&oldsp); /* second for our reference */
2191 	}
2192 	if (SPDCACHE_ENABLED())
2193 		spdcache_clear();
2194 	KEYDBG(KEY_STAMP,
2195 	    printf("%s: SP(%p)\n", __func__, newsp));
2196 	KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
2197 
2198     {
2199 	struct mbuf *n, *mpolicy;
2200 	struct sadb_msg *newmsg;
2201 	int off;
2202 
2203 	/* create new sadb_msg to reply. */
2204 	if (lft) {
2205 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
2206 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
2207 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2208 	} else {
2209 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
2210 		    SADB_X_EXT_POLICY,
2211 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2212 	}
2213 	if (!n)
2214 		return key_senderror(so, m, ENOBUFS);
2215 
2216 	if (n->m_len < sizeof(*newmsg)) {
2217 		n = m_pullup(n, sizeof(*newmsg));
2218 		if (!n)
2219 			return key_senderror(so, m, ENOBUFS);
2220 	}
2221 	newmsg = mtod(n, struct sadb_msg *);
2222 	newmsg->sadb_msg_errno = 0;
2223 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2224 
2225 	off = 0;
2226 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
2227 	    sizeof(*xpl), &off);
2228 	if (mpolicy == NULL) {
2229 		/* n is already freed */
2230 		return key_senderror(so, m, ENOBUFS);
2231 	}
2232 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, caddr_t) + off);
2233 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
2234 		m_freem(n);
2235 		return key_senderror(so, m, EINVAL);
2236 	}
2237 	xpl->sadb_x_policy_id = newsp->id;
2238 
2239 	m_freem(m);
2240 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2241     }
2242 }
2243 
2244 /*
2245  * get new policy id.
2246  * OUT:
2247  *	0:	failure.
2248  *	others: success.
2249  */
2250 static uint32_t
2251 key_getnewspid(void)
2252 {
2253 	struct secpolicy *sp;
2254 	uint32_t newid = 0;
2255 	int tries, limit;
2256 
2257 	SPTREE_WLOCK_ASSERT();
2258 
2259 	limit = atomic_load_int(&V_key_spi_trycnt);
2260 	for (tries = 0; tries < limit; tries++) {
2261 		if (V_policy_id == ~0) /* overflowed */
2262 			newid = V_policy_id = 1;
2263 		else
2264 			newid = ++V_policy_id;
2265 		LIST_FOREACH(sp, SPHASH_HASH(newid), idhash) {
2266 			if (sp->id == newid)
2267 				break;
2268 		}
2269 		if (sp == NULL)
2270 			break;
2271 	}
2272 	if (tries == limit || newid == 0) {
2273 		ipseclog((LOG_DEBUG, "%s: failed to allocate policy id.\n",
2274 		    __func__));
2275 		return (0);
2276 	}
2277 	return (newid);
2278 }
2279 
2280 /*
2281  * SADB_SPDDELETE processing
2282  * receive
2283  *   <base, address(SD), policy(*)>
2284  * from the user(?), and set SADB_SASTATE_DEAD,
2285  * and send,
2286  *   <base, address(SD), policy(*)>
2287  * to the ikmpd.
2288  * policy(*) including direction of policy.
2289  *
2290  * m will always be freed.
2291  */
2292 static int
2293 key_spddelete(struct socket *so, struct mbuf *m,
2294     const struct sadb_msghdr *mhp)
2295 {
2296 	struct secpolicyindex spidx;
2297 	struct sadb_address *src0, *dst0;
2298 	struct sadb_x_policy *xpl0;
2299 	struct secpolicy *sp;
2300 
2301 	IPSEC_ASSERT(so != NULL, ("null so"));
2302 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2303 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2304 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2305 
2306 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
2307 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
2308 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
2309 		ipseclog((LOG_DEBUG,
2310 		    "%s: invalid message: missing required header.\n",
2311 		    __func__));
2312 		return key_senderror(so, m, EINVAL);
2313 	}
2314 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
2315 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
2316 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2317 		ipseclog((LOG_DEBUG,
2318 		    "%s: invalid message: wrong header size.\n", __func__));
2319 		return key_senderror(so, m, EINVAL);
2320 	}
2321 
2322 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
2323 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
2324 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
2325 
2326 	/* check the direciton */
2327 	switch (xpl0->sadb_x_policy_dir) {
2328 	case IPSEC_DIR_INBOUND:
2329 	case IPSEC_DIR_OUTBOUND:
2330 		break;
2331 	default:
2332 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2333 		return key_senderror(so, m, EINVAL);
2334 	}
2335 	/* Only DISCARD, NONE and IPSEC are allowed */
2336 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2337 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2338 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2339 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2340 		return key_senderror(so, m, EINVAL);
2341 	}
2342 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
2343 	    (struct sockaddr *)(dst0 + 1)) != 0 ||
2344 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2345 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2346 		return key_senderror(so, m, EINVAL);
2347 	}
2348 	/* make secindex */
2349 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2350 	                src0 + 1,
2351 	                dst0 + 1,
2352 	                src0->sadb_address_prefixlen,
2353 	                dst0->sadb_address_prefixlen,
2354 	                src0->sadb_address_proto,
2355 	                &spidx);
2356 
2357 	/* Is there SP in SPD ? */
2358 	if ((sp = key_getsp(&spidx)) == NULL) {
2359 		ipseclog((LOG_DEBUG, "%s: no SP found.\n", __func__));
2360 		return key_senderror(so, m, EINVAL);
2361 	}
2362 
2363 	/* save policy id to buffer to be returned. */
2364 	xpl0->sadb_x_policy_id = sp->id;
2365 
2366 	KEYDBG(KEY_STAMP,
2367 	    printf("%s: SP(%p)\n", __func__, sp));
2368 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2369 	ipsec_accel_spddel(sp);
2370 	key_unlink(sp);
2371 	key_freesp(&sp);
2372 
2373     {
2374 	struct mbuf *n;
2375 	struct sadb_msg *newmsg;
2376 
2377 	/* create new sadb_msg to reply. */
2378 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2379 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2380 	if (!n)
2381 		return key_senderror(so, m, ENOBUFS);
2382 
2383 	newmsg = mtod(n, struct sadb_msg *);
2384 	newmsg->sadb_msg_errno = 0;
2385 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2386 
2387 	m_freem(m);
2388 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2389     }
2390 }
2391 
2392 /*
2393  * SADB_SPDDELETE2 processing
2394  * receive
2395  *   <base, policy(*)>
2396  * from the user(?), and set SADB_SASTATE_DEAD,
2397  * and send,
2398  *   <base, policy(*)>
2399  * to the ikmpd.
2400  * policy(*) including direction of policy.
2401  *
2402  * m will always be freed.
2403  */
2404 static int
2405 key_spddelete2(struct socket *so, struct mbuf *m,
2406     const struct sadb_msghdr *mhp)
2407 {
2408 	struct secpolicy *sp;
2409 	uint32_t id;
2410 
2411 	IPSEC_ASSERT(so != NULL, ("null socket"));
2412 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2413 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2414 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2415 
2416 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2417 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2418 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2419 		    __func__));
2420 		return key_senderror(so, m, EINVAL);
2421 	}
2422 
2423 	id = ((struct sadb_x_policy *)
2424 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2425 
2426 	/* Is there SP in SPD ? */
2427 	if ((sp = key_getspbyid(id)) == NULL) {
2428 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2429 		    __func__, id));
2430 		return key_senderror(so, m, EINVAL);
2431 	}
2432 
2433 	KEYDBG(KEY_STAMP,
2434 	    printf("%s: SP(%p)\n", __func__, sp));
2435 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2436 	key_unlink(sp);
2437 	if (sp->state != IPSEC_SPSTATE_DEAD) {
2438 		ipseclog((LOG_DEBUG, "%s: failed to delete SP with id %u.\n",
2439 		    __func__, id));
2440 		key_freesp(&sp);
2441 		return (key_senderror(so, m, EACCES));
2442 	}
2443 	key_freesp(&sp);
2444 
2445     {
2446 	struct mbuf *n, *nn;
2447 	struct sadb_msg *newmsg;
2448 	int off, len;
2449 
2450 	/* create new sadb_msg to reply. */
2451 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2452 
2453 	n = key_mget(len);
2454 	if (n == NULL)
2455 		return key_senderror(so, m, ENOBUFS);
2456 
2457 	n->m_len = len;
2458 	n->m_next = NULL;
2459 	off = 0;
2460 
2461 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
2462 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2463 
2464 	IPSEC_ASSERT(off == len, ("length inconsistency (off %u len %u)",
2465 		off, len));
2466 
2467 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2468 	    mhp->extlen[SADB_X_EXT_POLICY], M_NOWAIT);
2469 	if (!n->m_next) {
2470 		m_freem(n);
2471 		return key_senderror(so, m, ENOBUFS);
2472 	}
2473 
2474 	n->m_pkthdr.len = 0;
2475 	for (nn = n; nn; nn = nn->m_next)
2476 		n->m_pkthdr.len += nn->m_len;
2477 
2478 	newmsg = mtod(n, struct sadb_msg *);
2479 	newmsg->sadb_msg_errno = 0;
2480 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2481 
2482 	m_freem(m);
2483 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2484     }
2485 }
2486 
2487 /*
2488  * SADB_X_SPDGET processing
2489  * receive
2490  *   <base, policy(*)>
2491  * from the user(?),
2492  * and send,
2493  *   <base, address(SD), policy>
2494  * to the ikmpd.
2495  * policy(*) including direction of policy.
2496  *
2497  * m will always be freed.
2498  */
2499 static int
2500 key_spdget(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2501 {
2502 	struct secpolicy *sp;
2503 	struct mbuf *n;
2504 	uint32_t id;
2505 
2506 	IPSEC_ASSERT(so != NULL, ("null socket"));
2507 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2508 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2509 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2510 
2511 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2512 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2513 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2514 		    __func__));
2515 		return key_senderror(so, m, EINVAL);
2516 	}
2517 
2518 	id = ((struct sadb_x_policy *)
2519 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2520 
2521 	/* Is there SP in SPD ? */
2522 	if ((sp = key_getspbyid(id)) == NULL) {
2523 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2524 		    __func__, id));
2525 		return key_senderror(so, m, ENOENT);
2526 	}
2527 
2528 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2529 	    mhp->msg->sadb_msg_pid);
2530 	key_freesp(&sp);
2531 	if (n != NULL) {
2532 		m_freem(m);
2533 		return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2534 	} else
2535 		return key_senderror(so, m, ENOBUFS);
2536 }
2537 
2538 /*
2539  * SADB_X_SPDACQUIRE processing.
2540  * Acquire policy and SA(s) for a *OUTBOUND* packet.
2541  * send
2542  *   <base, policy(*)>
2543  * to KMD, and expect to receive
2544  *   <base> with SADB_X_SPDACQUIRE if error occurred,
2545  * or
2546  *   <base, policy>
2547  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2548  * policy(*) is without policy requests.
2549  *
2550  *    0     : succeed
2551  *    others: error number
2552  */
2553 int
2554 key_spdacquire(struct secpolicy *sp)
2555 {
2556 	struct mbuf *result = NULL, *m;
2557 	struct secspacq *newspacq;
2558 
2559 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2560 	IPSEC_ASSERT(sp->req == NULL, ("policy exists"));
2561 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
2562 		("policy not IPSEC %u", sp->policy));
2563 
2564 	/* Get an entry to check whether sent message or not. */
2565 	newspacq = key_getspacq(&sp->spidx);
2566 	if (newspacq != NULL) {
2567 		if (V_key_blockacq_count < newspacq->count) {
2568 			/* reset counter and do send message. */
2569 			newspacq->count = 0;
2570 		} else {
2571 			/* increment counter and do nothing. */
2572 			newspacq->count++;
2573 			SPACQ_UNLOCK();
2574 			return (0);
2575 		}
2576 		SPACQ_UNLOCK();
2577 	} else {
2578 		/* make new entry for blocking to send SADB_ACQUIRE. */
2579 		newspacq = key_newspacq(&sp->spidx);
2580 		if (newspacq == NULL)
2581 			return ENOBUFS;
2582 	}
2583 
2584 	/* create new sadb_msg to reply. */
2585 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2586 	if (!m)
2587 		return ENOBUFS;
2588 
2589 	result = m;
2590 
2591 	result->m_pkthdr.len = 0;
2592 	for (m = result; m; m = m->m_next)
2593 		result->m_pkthdr.len += m->m_len;
2594 
2595 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2596 	    PFKEY_UNIT64(result->m_pkthdr.len);
2597 
2598 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2599 }
2600 
2601 /*
2602  * SADB_SPDFLUSH processing
2603  * receive
2604  *   <base>
2605  * from the user, and free all entries in secpctree.
2606  * and send,
2607  *   <base>
2608  * to the user.
2609  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2610  *
2611  * m will always be freed.
2612  */
2613 static int
2614 key_spdflush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2615 {
2616 	struct secpolicy_queue drainq;
2617 	struct sadb_msg *newmsg;
2618 	struct secpolicy *sp, *nextsp;
2619 	u_int dir;
2620 
2621 	IPSEC_ASSERT(so != NULL, ("null socket"));
2622 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2623 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2624 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2625 
2626 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2627 		return key_senderror(so, m, EINVAL);
2628 
2629 	TAILQ_INIT(&drainq);
2630 	SPTREE_WLOCK();
2631 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2632 		TAILQ_CONCAT(&drainq, &V_sptree[dir], chain);
2633 	}
2634 	/*
2635 	 * We need to set state to DEAD for each policy to be sure,
2636 	 * that another thread won't try to unlink it.
2637 	 * Also remove SP from sphash.
2638 	 */
2639 	TAILQ_FOREACH(sp, &drainq, chain) {
2640 		sp->state = IPSEC_SPSTATE_DEAD;
2641 		ipsec_accel_spddel(sp);
2642 		LIST_REMOVE(sp, idhash);
2643 	}
2644 	V_sp_genid++;
2645 	V_spd_size = 0;
2646 	SPTREE_WUNLOCK();
2647 	if (SPDCACHE_ENABLED())
2648 		spdcache_clear();
2649 	sp = TAILQ_FIRST(&drainq);
2650 	while (sp != NULL) {
2651 		nextsp = TAILQ_NEXT(sp, chain);
2652 		key_freesp(&sp);
2653 		sp = nextsp;
2654 	}
2655 
2656 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2657 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
2658 		return key_senderror(so, m, ENOBUFS);
2659 	}
2660 
2661 	if (m->m_next)
2662 		m_freem(m->m_next);
2663 	m->m_next = NULL;
2664 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2665 	newmsg = mtod(m, struct sadb_msg *);
2666 	newmsg->sadb_msg_errno = 0;
2667 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2668 
2669 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2670 }
2671 
2672 static uint8_t
2673 key_satype2scopemask(uint8_t satype)
2674 {
2675 
2676 	if (satype == IPSEC_POLICYSCOPE_ANY)
2677 		return (0xff);
2678 	return (satype);
2679 }
2680 /*
2681  * SADB_SPDDUMP processing
2682  * receive
2683  *   <base>
2684  * from the user, and dump all SP leaves and send,
2685  *   <base> .....
2686  * to the ikmpd.
2687  *
2688  * NOTE:
2689  *   sadb_msg_satype is considered as mask of policy scopes.
2690  *   m will always be freed.
2691  */
2692 static int
2693 key_spddump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2694 {
2695 	SPTREE_RLOCK_TRACKER;
2696 	struct secpolicy *sp;
2697 	struct mbuf *n;
2698 	int cnt;
2699 	u_int dir, scope;
2700 
2701 	IPSEC_ASSERT(so != NULL, ("null socket"));
2702 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2703 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2704 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2705 
2706 	/* search SPD entry and get buffer size. */
2707 	cnt = 0;
2708 	scope = key_satype2scopemask(mhp->msg->sadb_msg_satype);
2709 	SPTREE_RLOCK();
2710 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2711 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2712 			TAILQ_FOREACH(sp, &V_sptree[dir], chain)
2713 				cnt++;
2714 		}
2715 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2716 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain)
2717 				cnt++;
2718 		}
2719 	}
2720 
2721 	if (cnt == 0) {
2722 		SPTREE_RUNLOCK();
2723 		return key_senderror(so, m, ENOENT);
2724 	}
2725 
2726 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2727 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2728 			TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
2729 				--cnt;
2730 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2731 				    mhp->msg->sadb_msg_pid);
2732 
2733 				if (n != NULL)
2734 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2735 			}
2736 		}
2737 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2738 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain) {
2739 				--cnt;
2740 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2741 				    mhp->msg->sadb_msg_pid);
2742 
2743 				if (n != NULL)
2744 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2745 			}
2746 		}
2747 	}
2748 
2749 	SPTREE_RUNLOCK();
2750 	m_freem(m);
2751 	return (0);
2752 }
2753 
2754 static struct mbuf *
2755 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq,
2756     u_int32_t pid)
2757 {
2758 	struct mbuf *result = NULL, *m;
2759 	struct seclifetime lt;
2760 
2761 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid, sp->refcnt);
2762 	if (!m)
2763 		goto fail;
2764 	result = m;
2765 
2766 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2767 	    &sp->spidx.src.sa, sp->spidx.prefs,
2768 	    sp->spidx.ul_proto);
2769 	if (!m)
2770 		goto fail;
2771 	m_cat(result, m);
2772 
2773 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2774 	    &sp->spidx.dst.sa, sp->spidx.prefd,
2775 	    sp->spidx.ul_proto);
2776 	if (!m)
2777 		goto fail;
2778 	m_cat(result, m);
2779 
2780 	m = key_sp2mbuf(sp);
2781 	if (!m)
2782 		goto fail;
2783 	m_cat(result, m);
2784 
2785 	if(sp->lifetime){
2786 		lt.addtime=sp->created;
2787 		lt.usetime= sp->lastused;
2788 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_CURRENT);
2789 		if (!m)
2790 			goto fail;
2791 		m_cat(result, m);
2792 
2793 		lt.addtime=sp->lifetime;
2794 		lt.usetime= sp->validtime;
2795 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_HARD);
2796 		if (!m)
2797 			goto fail;
2798 		m_cat(result, m);
2799 	}
2800 
2801 	if ((result->m_flags & M_PKTHDR) == 0)
2802 		goto fail;
2803 
2804 	if (result->m_len < sizeof(struct sadb_msg)) {
2805 		result = m_pullup(result, sizeof(struct sadb_msg));
2806 		if (result == NULL)
2807 			goto fail;
2808 	}
2809 
2810 	result->m_pkthdr.len = 0;
2811 	for (m = result; m; m = m->m_next)
2812 		result->m_pkthdr.len += m->m_len;
2813 
2814 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2815 	    PFKEY_UNIT64(result->m_pkthdr.len);
2816 
2817 	return result;
2818 
2819 fail:
2820 	m_freem(result);
2821 	return NULL;
2822 }
2823 /*
2824  * get PFKEY message length for security policy and request.
2825  */
2826 static size_t
2827 key_getspreqmsglen(struct secpolicy *sp)
2828 {
2829 	size_t tlen, len;
2830 	int i;
2831 
2832 	tlen = sizeof(struct sadb_x_policy);
2833 	/* if is the policy for ipsec ? */
2834 	if (sp->policy != IPSEC_POLICY_IPSEC)
2835 		return (tlen);
2836 
2837 	/* get length of ipsec requests */
2838 	for (i = 0; i < sp->tcount; i++) {
2839 		len = sizeof(struct sadb_x_ipsecrequest)
2840 			+ sp->req[i]->saidx.src.sa.sa_len
2841 			+ sp->req[i]->saidx.dst.sa.sa_len;
2842 
2843 		tlen += PFKEY_ALIGN8(len);
2844 	}
2845 #ifdef IPSEC_OFFLOAD
2846 	if (sp->accel_ifname != NULL)
2847 		tlen += sizeof(struct sadb_x_if_hw_offl);
2848 #endif
2849 	return (tlen);
2850 }
2851 
2852 /*
2853  * SADB_SPDEXPIRE processing
2854  * send
2855  *   <base, address(SD), lifetime(CH), policy>
2856  * to KMD by PF_KEY.
2857  *
2858  * OUT:	0	: succeed
2859  *	others	: error number
2860  */
2861 static int
2862 key_spdexpire(struct secpolicy *sp)
2863 {
2864 	struct sadb_lifetime *lt;
2865 	struct mbuf *result = NULL, *m;
2866 	int len, error = -1;
2867 
2868 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2869 
2870 	KEYDBG(KEY_STAMP,
2871 	    printf("%s: SP(%p)\n", __func__, sp));
2872 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2873 
2874 	/* set msg header */
2875 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0);
2876 	if (!m) {
2877 		error = ENOBUFS;
2878 		goto fail;
2879 	}
2880 	result = m;
2881 
2882 	/* create lifetime extension (current and hard) */
2883 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
2884 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
2885 	if (m == NULL) {
2886 		error = ENOBUFS;
2887 		goto fail;
2888 	}
2889 	m_align(m, len);
2890 	m->m_len = len;
2891 	bzero(mtod(m, caddr_t), len);
2892 	lt = mtod(m, struct sadb_lifetime *);
2893 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2894 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
2895 	lt->sadb_lifetime_allocations = 0;
2896 	lt->sadb_lifetime_bytes = 0;
2897 	lt->sadb_lifetime_addtime = sp->created;
2898 	lt->sadb_lifetime_usetime = sp->lastused;
2899 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
2900 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2901 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
2902 	lt->sadb_lifetime_allocations = 0;
2903 	lt->sadb_lifetime_bytes = 0;
2904 	lt->sadb_lifetime_addtime = sp->lifetime;
2905 	lt->sadb_lifetime_usetime = sp->validtime;
2906 	m_cat(result, m);
2907 
2908 	/* set sadb_address for source */
2909 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2910 	    &sp->spidx.src.sa,
2911 	    sp->spidx.prefs, sp->spidx.ul_proto);
2912 	if (!m) {
2913 		error = ENOBUFS;
2914 		goto fail;
2915 	}
2916 	m_cat(result, m);
2917 
2918 	/* set sadb_address for destination */
2919 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2920 	    &sp->spidx.dst.sa,
2921 	    sp->spidx.prefd, sp->spidx.ul_proto);
2922 	if (!m) {
2923 		error = ENOBUFS;
2924 		goto fail;
2925 	}
2926 	m_cat(result, m);
2927 
2928 	/* set secpolicy */
2929 	m = key_sp2mbuf(sp);
2930 	if (!m) {
2931 		error = ENOBUFS;
2932 		goto fail;
2933 	}
2934 	m_cat(result, m);
2935 
2936 	if ((result->m_flags & M_PKTHDR) == 0) {
2937 		error = EINVAL;
2938 		goto fail;
2939 	}
2940 
2941 	if (result->m_len < sizeof(struct sadb_msg)) {
2942 		result = m_pullup(result, sizeof(struct sadb_msg));
2943 		if (result == NULL) {
2944 			error = ENOBUFS;
2945 			goto fail;
2946 		}
2947 	}
2948 
2949 	result->m_pkthdr.len = 0;
2950 	for (m = result; m; m = m->m_next)
2951 		result->m_pkthdr.len += m->m_len;
2952 
2953 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2954 	    PFKEY_UNIT64(result->m_pkthdr.len);
2955 
2956 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
2957 
2958  fail:
2959 	if (result)
2960 		m_freem(result);
2961 	return error;
2962 }
2963 
2964 /* %%% SAD management */
2965 /*
2966  * allocating and initialize new SA head.
2967  * OUT:	NULL	: failure due to the lack of memory.
2968  *	others	: pointer to new SA head.
2969  */
2970 static struct secashead *
2971 key_newsah(struct secasindex *saidx)
2972 {
2973 	struct secashead *sah;
2974 
2975 	sah = malloc(sizeof(struct secashead), M_IPSEC_SAH,
2976 	    M_NOWAIT | M_ZERO);
2977 	if (sah == NULL) {
2978 		PFKEYSTAT_INC(in_nomem);
2979 		return (NULL);
2980 	}
2981 	TAILQ_INIT(&sah->savtree_larval);
2982 	TAILQ_INIT(&sah->savtree_alive);
2983 	sah->saidx = *saidx;
2984 	sah->state = SADB_SASTATE_DEAD;
2985 	SAH_INITREF(sah);
2986 
2987 	KEYDBG(KEY_STAMP,
2988 	    printf("%s: SAH(%p)\n", __func__, sah));
2989 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2990 	return (sah);
2991 }
2992 
2993 static void
2994 key_freesah(struct secashead **psah)
2995 {
2996 	struct secashead *sah = *psah;
2997 
2998 	CURVNET_ASSERT_SET();
2999 
3000 	if (SAH_DELREF(sah) == 0)
3001 		return;
3002 
3003 	KEYDBG(KEY_STAMP,
3004 	    printf("%s: last reference to SAH(%p)\n", __func__, sah));
3005 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
3006 
3007 	*psah = NULL;
3008 	key_delsah(sah);
3009 }
3010 
3011 static void
3012 key_delsah(struct secashead *sah)
3013 {
3014 	IPSEC_ASSERT(sah != NULL, ("NULL sah"));
3015 	IPSEC_ASSERT(sah->state == SADB_SASTATE_DEAD,
3016 	    ("Attempt to free non DEAD SAH %p", sah));
3017 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_larval),
3018 	    ("Attempt to free SAH %p with LARVAL SA", sah));
3019 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_alive),
3020 	    ("Attempt to free SAH %p with ALIVE SA", sah));
3021 
3022 	free(sah, M_IPSEC_SAH);
3023 }
3024 
3025 /*
3026  * allocating a new SA for key_add() and key_getspi() call,
3027  * and copy the values of mhp into new buffer.
3028  * When SAD message type is SADB_GETSPI set SA state to LARVAL.
3029  * For SADB_ADD create and initialize SA with MATURE state.
3030  * OUT:	NULL	: fail
3031  *	others	: pointer to new secasvar.
3032  */
3033 static struct secasvar *
3034 key_newsav(const struct sadb_msghdr *mhp, struct secasindex *saidx,
3035     uint32_t spi, int *errp)
3036 {
3037 	struct secashead *sah;
3038 	struct secasvar *sav;
3039 	int isnew;
3040 
3041 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
3042 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
3043 	IPSEC_ASSERT(mhp->msg->sadb_msg_type == SADB_GETSPI ||
3044 	    mhp->msg->sadb_msg_type == SADB_ADD, ("wrong message type"));
3045 
3046 	sav = NULL;
3047 	sah = NULL;
3048 	/* check SPI value */
3049 	switch (saidx->proto) {
3050 	case IPPROTO_ESP:
3051 	case IPPROTO_AH:
3052 		/*
3053 		 * RFC 4302, 2.4. Security Parameters Index (SPI), SPI values
3054 		 * 1-255 reserved by IANA for future use,
3055 		 * 0 for implementation specific, local use.
3056 		 */
3057 		if (ntohl(spi) <= 255) {
3058 			ipseclog((LOG_DEBUG, "%s: illegal range of SPI %u.\n",
3059 			    __func__, ntohl(spi)));
3060 			*errp = EINVAL;
3061 			goto done;
3062 		}
3063 		break;
3064 	}
3065 
3066 	sav = malloc(sizeof(struct secasvar), M_IPSEC_SA, M_NOWAIT | M_ZERO);
3067 	if (sav == NULL) {
3068 		*errp = ENOBUFS;
3069 		goto done;
3070 	}
3071 	sav->lock = malloc_aligned(max(sizeof(struct rmlock),
3072 	    CACHE_LINE_SIZE), CACHE_LINE_SIZE, M_IPSEC_MISC,
3073 	    M_NOWAIT | M_ZERO);
3074 	if (sav->lock == NULL) {
3075 		*errp = ENOBUFS;
3076 		goto done;
3077 	}
3078 	rm_init(sav->lock, "ipsec association");
3079 	sav->lft_c = uma_zalloc_pcpu(ipsec_key_lft_zone, M_NOWAIT | M_ZERO);
3080 	if (sav->lft_c == NULL) {
3081 		*errp = ENOBUFS;
3082 		goto done;
3083 	}
3084 
3085 	sav->spi = spi;
3086 	sav->seq = mhp->msg->sadb_msg_seq;
3087 	sav->state = SADB_SASTATE_LARVAL;
3088 	sav->pid = (pid_t)mhp->msg->sadb_msg_pid;
3089 	SAV_INITREF(sav);
3090 #ifdef IPSEC_OFFLOAD
3091 	CK_LIST_INIT(&sav->accel_ifps);
3092 	sav->accel_forget_tq = 0;
3093 	sav->accel_lft_sw = uma_zalloc_pcpu(ipsec_key_lft_zone,
3094 	    M_NOWAIT | M_ZERO);
3095 	if (sav->accel_lft_sw == NULL) {
3096 		*errp = ENOBUFS;
3097 		goto done;
3098 	}
3099 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_IF_HW_OFFL) &&
3100 	    !SADB_CHECKLEN(mhp, SADB_X_EXT_IF_HW_OFFL)) {
3101 		struct sadb_x_if_hw_offl *xof;
3102 
3103 		xof = (struct sadb_x_if_hw_offl *)mhp->ext[
3104 		    SADB_X_EXT_IF_HW_OFFL];
3105 		sav->accel_ifname = malloc(sizeof(xof->sadb_x_if_hw_offl_if),
3106 		    M_IPSEC_MISC, M_NOWAIT);
3107 		if (sav->accel_ifname == NULL) {
3108 			*errp = ENOBUFS;
3109 			goto done;
3110 		}
3111 		strncpy(__DECONST(char *, sav->accel_ifname),
3112 		    xof->sadb_x_if_hw_offl_if,
3113 		    sizeof(xof->sadb_x_if_hw_offl_if));
3114 	}
3115 #endif
3116 again:
3117 	sah = key_getsah(saidx);
3118 	if (sah == NULL) {
3119 		/* create a new SA index */
3120 		sah = key_newsah(saidx);
3121 		if (sah == NULL) {
3122 			ipseclog((LOG_DEBUG,
3123 			    "%s: No more memory.\n", __func__));
3124 			*errp = ENOBUFS;
3125 			goto done;
3126 		}
3127 		isnew = 1;
3128 	} else
3129 		isnew = 0;
3130 
3131 	sav->sah = sah;
3132 	if (mhp->msg->sadb_msg_type == SADB_GETSPI) {
3133 		sav->created = time_second;
3134 	} else if (sav->state == SADB_SASTATE_LARVAL) {
3135 		/*
3136 		 * Do not call key_setsaval() second time in case
3137 		 * of `goto again`. We will have MATURE state.
3138 		 */
3139 		*errp = key_setsaval(sav, mhp);
3140 		if (*errp != 0)
3141 			goto done;
3142 		sav->state = SADB_SASTATE_MATURE;
3143 	}
3144 
3145 	SAHTREE_WLOCK();
3146 	/*
3147 	 * Check that existing SAH wasn't unlinked.
3148 	 * Since we didn't hold the SAHTREE lock, it is possible,
3149 	 * that callout handler or key_flush() or key_delete() could
3150 	 * unlink this SAH.
3151 	 */
3152 	if (isnew == 0 && sah->state == SADB_SASTATE_DEAD) {
3153 		SAHTREE_WUNLOCK();
3154 		key_freesah(&sah);	/* reference from key_getsah() */
3155 		goto again;
3156 	}
3157 	if (isnew != 0) {
3158 		/*
3159 		 * Add new SAH into SADB.
3160 		 *
3161 		 * XXXAE: we can serialize key_add and key_getspi calls, so
3162 		 * several threads will not fight in the race.
3163 		 * Otherwise we should check under SAHTREE lock, that this
3164 		 * SAH would not added twice.
3165 		 */
3166 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
3167 		/* Add new SAH into hash by addresses */
3168 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
3169 		/* Now we are linked in the chain */
3170 		sah->state = SADB_SASTATE_MATURE;
3171 		/*
3172 		 * SAV references this new SAH.
3173 		 * In case of existing SAH we reuse reference
3174 		 * from key_getsah().
3175 		 */
3176 		SAH_ADDREF(sah);
3177 	}
3178 	/* Link SAV with SAH */
3179 	if (sav->state == SADB_SASTATE_MATURE) {
3180 		TAILQ_INSERT_HEAD(&sah->savtree_alive, sav, chain);
3181 		ipsec_accel_sa_newkey(sav);
3182 	} else
3183 		TAILQ_INSERT_HEAD(&sah->savtree_larval, sav, chain);
3184 	/* Add SAV into SPI hash */
3185 	LIST_INSERT_HEAD(SAVHASH_HASH(sav->spi), sav, spihash);
3186 	SAHTREE_WUNLOCK();
3187 	*errp = 0;	/* success */
3188 done:
3189 	if (*errp != 0) {
3190 		if (sav != NULL) {
3191 			if (sav->lock != NULL) {
3192 				rm_destroy(sav->lock);
3193 				free(sav->lock, M_IPSEC_MISC);
3194 			}
3195 			if (sav->lft_c != NULL)
3196 				uma_zfree_pcpu(ipsec_key_lft_zone, sav->lft_c);
3197 #ifdef IPSEC_OFFLOAD
3198 			if (sav->accel_lft_sw != NULL)
3199 				uma_zfree_pcpu(ipsec_key_lft_zone,
3200 				    sav->accel_lft_sw);
3201 			free(__DECONST(char *, sav->accel_ifname),
3202 			    M_IPSEC_MISC);
3203 #endif
3204 			free(sav, M_IPSEC_SA), sav = NULL;
3205 		}
3206 		if (sah != NULL)
3207 			key_freesah(&sah);
3208 		if (*errp == ENOBUFS) {
3209 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3210 			    __func__));
3211 			PFKEYSTAT_INC(in_nomem);
3212 		}
3213 	}
3214 	return (sav);
3215 }
3216 
3217 /*
3218  * free() SA variable entry.
3219  */
3220 static void
3221 key_cleansav(struct secasvar *sav)
3222 {
3223 
3224 	if (sav->natt != NULL) {
3225 		free(sav->natt, M_IPSEC_MISC);
3226 		sav->natt = NULL;
3227 	}
3228 	if (sav->flags & SADB_X_EXT_F_CLONED)
3229 		return;
3230 	if (sav->tdb_xform != NULL) {
3231 		sav->tdb_xform->xf_cleanup(sav);
3232 		sav->tdb_xform = NULL;
3233 	}
3234 	if (sav->key_auth != NULL) {
3235 		zfree(sav->key_auth->key_data, M_IPSEC_MISC);
3236 		free(sav->key_auth, M_IPSEC_MISC);
3237 		sav->key_auth = NULL;
3238 	}
3239 	if (sav->key_enc != NULL) {
3240 		zfree(sav->key_enc->key_data, M_IPSEC_MISC);
3241 		free(sav->key_enc, M_IPSEC_MISC);
3242 		sav->key_enc = NULL;
3243 	}
3244 	if (sav->replay != NULL) {
3245 		mtx_destroy(&sav->replay->lock);
3246 		if (sav->replay->bitmap != NULL)
3247 			free(sav->replay->bitmap, M_IPSEC_MISC);
3248 		free(sav->replay, M_IPSEC_MISC);
3249 		sav->replay = NULL;
3250 	}
3251 	if (sav->lft_h != NULL) {
3252 		free(sav->lft_h, M_IPSEC_MISC);
3253 		sav->lft_h = NULL;
3254 	}
3255 	if (sav->lft_s != NULL) {
3256 		free(sav->lft_s, M_IPSEC_MISC);
3257 		sav->lft_s = NULL;
3258 	}
3259 }
3260 
3261 /*
3262  * free() SA variable entry.
3263  */
3264 static void
3265 key_delsav(struct secasvar *sav)
3266 {
3267 	IPSEC_ASSERT(sav != NULL, ("null sav"));
3268 	IPSEC_ASSERT(sav->state == SADB_SASTATE_DEAD,
3269 	    ("attempt to free non DEAD SA %p", sav));
3270 	IPSEC_ASSERT(sav->refcnt == 0, ("reference count %u > 0",
3271 	    sav->refcnt));
3272 #ifdef IPSEC_OFFLOAD
3273 	KASSERT(CK_LIST_EMPTY(&sav->accel_ifps),
3274 	    ("key_unlinksav: sav %p still offloaded", sav));
3275 #endif
3276 
3277 	/*
3278 	 * SA must be unlinked from the chain and hashtbl.
3279 	 * If SA was cloned, we leave all fields untouched,
3280 	 * except NAT-T config.
3281 	 */
3282 	key_cleansav(sav);
3283 	if ((sav->flags & SADB_X_EXT_F_CLONED) == 0) {
3284 		rm_destroy(sav->lock);
3285 		free(sav->lock, M_IPSEC_MISC);
3286 		uma_zfree_pcpu(ipsec_key_lft_zone, sav->lft_c);
3287 	}
3288 #ifdef IPSEC_OFFLOAD
3289 	/* XXXKIB should this be moved to key_cleansav()? */
3290 	uma_zfree_pcpu(ipsec_key_lft_zone, sav->accel_lft_sw);
3291 	free(__DECONST(char *, sav->accel_ifname), M_IPSEC_MISC);
3292 #endif
3293 	free(sav, M_IPSEC_SA);
3294 }
3295 
3296 /*
3297  * search SAH.
3298  * OUT:
3299  *	NULL	: not found
3300  *	others	: found, referenced pointer to a SAH.
3301  */
3302 static struct secashead *
3303 key_getsah(struct secasindex *saidx)
3304 {
3305 	SAHTREE_RLOCK_TRACKER;
3306 	struct secashead *sah;
3307 
3308 	SAHTREE_RLOCK();
3309 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
3310 	    if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID) != 0) {
3311 		    SAH_ADDREF(sah);
3312 		    break;
3313 	    }
3314 	}
3315 	SAHTREE_RUNLOCK();
3316 	return (sah);
3317 }
3318 
3319 /*
3320  * Check not to be duplicated SPI.
3321  * OUT:
3322  *	0	: not found
3323  *	1	: found SA with given SPI.
3324  */
3325 static int
3326 key_checkspidup(uint32_t spi)
3327 {
3328 	SAHTREE_RLOCK_TRACKER;
3329 	struct secasvar *sav;
3330 
3331 	/* Assume SPI is in network byte order */
3332 	SAHTREE_RLOCK();
3333 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3334 		if (sav->spi == spi)
3335 			break;
3336 	}
3337 	SAHTREE_RUNLOCK();
3338 	return (sav != NULL);
3339 }
3340 
3341 /*
3342  * Search SA by SPI.
3343  * OUT:
3344  *	NULL	: not found
3345  *	others	: found, referenced pointer to a SA.
3346  */
3347 static struct secasvar *
3348 key_getsavbyspi(uint32_t spi)
3349 {
3350 	SAHTREE_RLOCK_TRACKER;
3351 	struct secasvar *sav;
3352 
3353 	/* Assume SPI is in network byte order */
3354 	SAHTREE_RLOCK();
3355 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3356 		if (sav->spi != spi)
3357 			continue;
3358 		SAV_ADDREF(sav);
3359 		break;
3360 	}
3361 	SAHTREE_RUNLOCK();
3362 	return (sav);
3363 }
3364 
3365 static int
3366 key_updatelifetimes(struct secasvar *sav, const struct sadb_msghdr *mhp)
3367 {
3368 	struct seclifetime *lft_h, *lft_s, *tmp;
3369 
3370 	/* Lifetime extension is optional, check that it is present. */
3371 	if (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3372 	    SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) {
3373 		/*
3374 		 * In case of SADB_UPDATE we may need to change
3375 		 * existing lifetimes.
3376 		 */
3377 		if (sav->state == SADB_SASTATE_MATURE) {
3378 			lft_h = lft_s = NULL;
3379 			goto reset;
3380 		}
3381 		return (0);
3382 	}
3383 	/* Both HARD and SOFT extensions must present */
3384 	if ((SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3385 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
3386 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
3387 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
3388 		ipseclog((LOG_DEBUG,
3389 		    "%s: invalid message: missing required header.\n",
3390 		    __func__));
3391 		return (EINVAL);
3392 	}
3393 	if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD) ||
3394 	    SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_SOFT)) {
3395 		ipseclog((LOG_DEBUG,
3396 		    "%s: invalid message: wrong header size.\n", __func__));
3397 		return (EINVAL);
3398 	}
3399 	lft_h = key_dup_lifemsg((const struct sadb_lifetime *)
3400 	    mhp->ext[SADB_EXT_LIFETIME_HARD], M_IPSEC_MISC);
3401 	if (lft_h == NULL) {
3402 		PFKEYSTAT_INC(in_nomem);
3403 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3404 		return (ENOBUFS);
3405 	}
3406 	lft_s = key_dup_lifemsg((const struct sadb_lifetime *)
3407 	    mhp->ext[SADB_EXT_LIFETIME_SOFT], M_IPSEC_MISC);
3408 	if (lft_s == NULL) {
3409 		PFKEYSTAT_INC(in_nomem);
3410 		free(lft_h, M_IPSEC_MISC);
3411 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3412 		return (ENOBUFS);
3413 	}
3414 reset:
3415 	if (sav->state != SADB_SASTATE_LARVAL) {
3416 		/*
3417 		 * key_update() holds reference to this SA,
3418 		 * so it won't be deleted in meanwhile.
3419 		 */
3420 		SECASVAR_WLOCK(sav);
3421 		tmp = sav->lft_h;
3422 		sav->lft_h = lft_h;
3423 		lft_h = tmp;
3424 
3425 		tmp = sav->lft_s;
3426 		sav->lft_s = lft_s;
3427 		lft_s = tmp;
3428 		SECASVAR_WUNLOCK(sav);
3429 		if (lft_h != NULL)
3430 			free(lft_h, M_IPSEC_MISC);
3431 		if (lft_s != NULL)
3432 			free(lft_s, M_IPSEC_MISC);
3433 		return (0);
3434 	}
3435 	/* We can update lifetime without holding a lock */
3436 	IPSEC_ASSERT(sav->lft_h == NULL, ("lft_h is already initialized\n"));
3437 	IPSEC_ASSERT(sav->lft_s == NULL, ("lft_s is already initialized\n"));
3438 	sav->lft_h = lft_h;
3439 	sav->lft_s = lft_s;
3440 	return (0);
3441 }
3442 
3443 /*
3444  * copy SA values from PF_KEY message except *SPI, SEQ, PID and TYPE*.
3445  * You must update these if need. Expects only LARVAL SAs.
3446  * OUT:	0:	success.
3447  *	!0:	failure.
3448  */
3449 static int
3450 key_setsaval(struct secasvar *sav, const struct sadb_msghdr *mhp)
3451 {
3452 	const struct sadb_sa *sa0;
3453 	const struct sadb_key *key0;
3454 	uint32_t replay;
3455 	size_t len;
3456 	int error;
3457 
3458 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
3459 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
3460 	IPSEC_ASSERT(sav->state == SADB_SASTATE_LARVAL,
3461 	    ("Attempt to update non LARVAL SA"));
3462 
3463 	/* XXX rewrite */
3464 	error = key_setident(sav->sah, mhp);
3465 	if (error != 0)
3466 		goto fail;
3467 
3468 	/* SA */
3469 	if (!SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
3470 		if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
3471 			error = EINVAL;
3472 			goto fail;
3473 		}
3474 		sa0 = (const struct sadb_sa *)mhp->ext[SADB_EXT_SA];
3475 		sav->alg_auth = sa0->sadb_sa_auth;
3476 		sav->alg_enc = sa0->sadb_sa_encrypt;
3477 		sav->flags = sa0->sadb_sa_flags;
3478 		if ((sav->flags & SADB_KEY_FLAGS_MAX) != sav->flags) {
3479 			ipseclog((LOG_DEBUG,
3480 			    "%s: invalid sa_flags 0x%08x.\n", __func__,
3481 			    sav->flags));
3482 			error = EINVAL;
3483 			goto fail;
3484 		}
3485 
3486 		/* Optional replay window */
3487 		replay = 0;
3488 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0)
3489 			replay = sa0->sadb_sa_replay;
3490 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_SA_REPLAY)) {
3491 			if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA_REPLAY)) {
3492 				error = EINVAL;
3493 				goto fail;
3494 			}
3495 			replay = ((const struct sadb_x_sa_replay *)
3496 			    mhp->ext[SADB_X_EXT_SA_REPLAY])->sadb_x_sa_replay_replay;
3497 
3498 			if (replay > UINT32_MAX - 32) {
3499 				ipseclog((LOG_DEBUG,
3500 				    "%s: replay window too big.\n", __func__));
3501 				error = EINVAL;
3502 				goto fail;
3503 			}
3504 
3505 			replay = (replay + 7) >> 3;
3506 		}
3507 
3508 		sav->replay = malloc(sizeof(struct secreplay), M_IPSEC_MISC,
3509 		    M_NOWAIT | M_ZERO);
3510 		if (sav->replay == NULL) {
3511 			PFKEYSTAT_INC(in_nomem);
3512 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3513 			    __func__));
3514 			error = ENOBUFS;
3515 			goto fail;
3516 		}
3517 		mtx_init(&sav->replay->lock, "ipsec replay", NULL, MTX_DEF);
3518 
3519 		if (replay != 0) {
3520 			/* number of 32b blocks to be allocated */
3521 			uint32_t bitmap_size;
3522 
3523 			/* RFC 6479:
3524 			 * - the allocated replay window size must be
3525 			 *   a power of two.
3526 			 * - use an extra 32b block as a redundant window.
3527 			 */
3528 			bitmap_size = 1;
3529 			while (replay + 4 > bitmap_size)
3530 				bitmap_size <<= 1;
3531 			bitmap_size = bitmap_size / 4;
3532 
3533 			sav->replay->bitmap = malloc(
3534 			    bitmap_size * sizeof(uint32_t), M_IPSEC_MISC,
3535 			    M_NOWAIT | M_ZERO);
3536 			if (sav->replay->bitmap == NULL) {
3537 				PFKEYSTAT_INC(in_nomem);
3538 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3539 					__func__));
3540 				error = ENOBUFS;
3541 				goto fail;
3542 			}
3543 			sav->replay->bitmap_size = bitmap_size;
3544 			sav->replay->wsize = replay;
3545 		}
3546 	}
3547 
3548 	/* Authentication keys */
3549 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
3550 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH)) {
3551 			error = EINVAL;
3552 			goto fail;
3553 		}
3554 		error = 0;
3555 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_AUTH];
3556 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3557 		switch (mhp->msg->sadb_msg_satype) {
3558 		case SADB_SATYPE_AH:
3559 		case SADB_SATYPE_ESP:
3560 		case SADB_X_SATYPE_TCPSIGNATURE:
3561 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3562 			    sav->alg_auth != SADB_X_AALG_NULL)
3563 				error = EINVAL;
3564 			break;
3565 		case SADB_X_SATYPE_IPCOMP:
3566 		default:
3567 			error = EINVAL;
3568 			break;
3569 		}
3570 		if (error) {
3571 			ipseclog((LOG_DEBUG, "%s: invalid key_auth values.\n",
3572 				__func__));
3573 			goto fail;
3574 		}
3575 
3576 		sav->key_auth = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3577 		if (sav->key_auth == NULL ) {
3578 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3579 				  __func__));
3580 			PFKEYSTAT_INC(in_nomem);
3581 			error = ENOBUFS;
3582 			goto fail;
3583 		}
3584 	}
3585 
3586 	/* Encryption key */
3587 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) {
3588 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT)) {
3589 			error = EINVAL;
3590 			goto fail;
3591 		}
3592 		error = 0;
3593 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_ENCRYPT];
3594 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3595 		switch (mhp->msg->sadb_msg_satype) {
3596 		case SADB_SATYPE_ESP:
3597 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3598 			    sav->alg_enc != SADB_EALG_NULL) {
3599 				error = EINVAL;
3600 				break;
3601 			}
3602 			sav->key_enc = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3603 			if (sav->key_enc == NULL) {
3604 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3605 					__func__));
3606 				PFKEYSTAT_INC(in_nomem);
3607 				error = ENOBUFS;
3608 				goto fail;
3609 			}
3610 			break;
3611 		case SADB_X_SATYPE_IPCOMP:
3612 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3613 				error = EINVAL;
3614 			sav->key_enc = NULL;	/*just in case*/
3615 			break;
3616 		case SADB_SATYPE_AH:
3617 		case SADB_X_SATYPE_TCPSIGNATURE:
3618 		default:
3619 			error = EINVAL;
3620 			break;
3621 		}
3622 		if (error) {
3623 			ipseclog((LOG_DEBUG, "%s: invalid key_enc value.\n",
3624 				__func__));
3625 			goto fail;
3626 		}
3627 	}
3628 
3629 	/* set iv */
3630 	sav->ivlen = 0;
3631 	switch (mhp->msg->sadb_msg_satype) {
3632 	case SADB_SATYPE_AH:
3633 		if (sav->flags & SADB_X_EXT_DERIV) {
3634 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3635 			    "given to AH SA.\n", __func__));
3636 			error = EINVAL;
3637 			goto fail;
3638 		}
3639 		if (sav->alg_enc != SADB_EALG_NONE) {
3640 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3641 			    "mismated.\n", __func__));
3642 			error = EINVAL;
3643 			goto fail;
3644 		}
3645 		error = xform_init(sav, XF_AH);
3646 		break;
3647 	case SADB_SATYPE_ESP:
3648 		if ((sav->flags & (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) ==
3649 		    (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) {
3650 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3651 			    "given to old-esp.\n", __func__));
3652 			error = EINVAL;
3653 			goto fail;
3654 		}
3655 		error = xform_init(sav, XF_ESP);
3656 		break;
3657 	case SADB_X_SATYPE_IPCOMP:
3658 		if (sav->alg_auth != SADB_AALG_NONE) {
3659 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3660 			    "mismated.\n", __func__));
3661 			error = EINVAL;
3662 			goto fail;
3663 		}
3664 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0 &&
3665 		    ntohl(sav->spi) >= 0x10000) {
3666 			ipseclog((LOG_DEBUG, "%s: invalid cpi for IPComp.\n",
3667 			    __func__));
3668 			error = EINVAL;
3669 			goto fail;
3670 		}
3671 		error = xform_init(sav, XF_IPCOMP);
3672 		break;
3673 	case SADB_X_SATYPE_TCPSIGNATURE:
3674 		if (sav->alg_enc != SADB_EALG_NONE) {
3675 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3676 			    "mismated.\n", __func__));
3677 			error = EINVAL;
3678 			goto fail;
3679 		}
3680 		error = xform_init(sav, XF_TCPSIGNATURE);
3681 		break;
3682 	default:
3683 		ipseclog((LOG_DEBUG, "%s: Invalid satype.\n", __func__));
3684 		error = EPROTONOSUPPORT;
3685 		goto fail;
3686 	}
3687 	if (error) {
3688 		ipseclog((LOG_DEBUG, "%s: unable to initialize SA type %u.\n",
3689 		    __func__, mhp->msg->sadb_msg_satype));
3690 		goto fail;
3691 	}
3692 
3693 	/* Handle NAT-T headers */
3694 	error = key_setnatt(sav, mhp);
3695 	if (error != 0)
3696 		goto fail;
3697 
3698 	/* Initialize lifetime for CURRENT */
3699 	sav->firstused = 0;
3700 	sav->created = time_second;
3701 
3702 	/* lifetimes for HARD and SOFT */
3703 	error = key_updatelifetimes(sav, mhp);
3704 	if (error == 0)
3705 		return (0);
3706 fail:
3707 	key_cleansav(sav);
3708 	return (error);
3709 }
3710 
3711 /*
3712  * subroutine for SADB_GET and SADB_DUMP.
3713  */
3714 static struct mbuf *
3715 key_setdumpsa(struct secasvar *sav, uint8_t type, uint8_t satype,
3716     uint32_t seq, uint32_t pid, struct rm_priotracker *sahtree_trackerp)
3717 {
3718 	struct seclifetime lft_c;
3719 	struct mbuf *result = NULL, *tres = NULL, *m;
3720 	int i, dumporder[] = {
3721 		SADB_EXT_SA, SADB_X_EXT_SA2, SADB_X_EXT_SA_REPLAY,
3722 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3723 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3724 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY,
3725 		SADB_EXT_KEY_AUTH, SADB_EXT_KEY_ENCRYPT,
3726 		SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
3727 		SADB_EXT_SENSITIVITY,
3728 		SADB_X_EXT_NAT_T_TYPE,
3729 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3730 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3731 		SADB_X_EXT_NAT_T_FRAG,
3732 #ifdef IPSEC_OFFLOAD
3733 		SADB_X_EXT_LFT_CUR_SW_OFFL, SADB_X_EXT_LFT_CUR_HW_OFFL,
3734 		SADB_X_EXT_IF_HW_OFFL,
3735 #endif
3736 	};
3737 	uint32_t replay_count;
3738 #ifdef IPSEC_OFFLOAD
3739 	int error;
3740 #endif
3741 
3742 	SECASVAR_RLOCK_TRACKER;
3743 
3744 	m = key_setsadbmsg(type, 0, satype, seq, pid, sav->refcnt);
3745 	if (m == NULL)
3746 		goto fail;
3747 	result = m;
3748 
3749 	for (i = nitems(dumporder) - 1; i >= 0; i--) {
3750 		m = NULL;
3751 		switch (dumporder[i]) {
3752 		case SADB_EXT_SA:
3753 			m = key_setsadbsa(sav);
3754 			if (!m)
3755 				goto fail;
3756 			break;
3757 
3758 		case SADB_X_EXT_SA2: {
3759 			SECASVAR_RLOCK(sav);
3760 			replay_count = sav->replay ? sav->replay->count : 0;
3761 			SECASVAR_RUNLOCK(sav);
3762 			m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
3763 					sav->sah->saidx.reqid);
3764 			if (!m)
3765 				goto fail;
3766 			break;
3767 		}
3768 		case SADB_X_EXT_SA_REPLAY:
3769 			if (sav->replay == NULL ||
3770 			    sav->replay->wsize <= UINT8_MAX)
3771 				continue;
3772 
3773 			m = key_setsadbxsareplay(sav->replay->wsize);
3774 			if (!m)
3775 				goto fail;
3776 			break;
3777 
3778 		case SADB_EXT_ADDRESS_SRC:
3779 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3780 			    &sav->sah->saidx.src.sa,
3781 			    FULLMASK, IPSEC_ULPROTO_ANY);
3782 			if (!m)
3783 				goto fail;
3784 			break;
3785 
3786 		case SADB_EXT_ADDRESS_DST:
3787 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3788 			    &sav->sah->saidx.dst.sa,
3789 			    FULLMASK, IPSEC_ULPROTO_ANY);
3790 			if (!m)
3791 				goto fail;
3792 			break;
3793 
3794 		case SADB_EXT_KEY_AUTH:
3795 			if (!sav->key_auth)
3796 				continue;
3797 			m = key_setkey(sav->key_auth, SADB_EXT_KEY_AUTH);
3798 			if (!m)
3799 				goto fail;
3800 			break;
3801 
3802 		case SADB_EXT_KEY_ENCRYPT:
3803 			if (!sav->key_enc)
3804 				continue;
3805 			m = key_setkey(sav->key_enc, SADB_EXT_KEY_ENCRYPT);
3806 			if (!m)
3807 				goto fail;
3808 			break;
3809 
3810 		case SADB_EXT_LIFETIME_CURRENT:
3811 			lft_c.addtime = sav->created;
3812 			lft_c.allocations = (uint32_t)counter_u64_fetch(
3813 			    sav->lft_c_allocations);
3814 			lft_c.bytes = counter_u64_fetch(sav->lft_c_bytes);
3815 			lft_c.usetime = sav->firstused;
3816 			m = key_setlifetime(&lft_c, SADB_EXT_LIFETIME_CURRENT);
3817 			if (!m)
3818 				goto fail;
3819 			break;
3820 
3821 		case SADB_EXT_LIFETIME_HARD:
3822 			if (!sav->lft_h)
3823 				continue;
3824 			m = key_setlifetime(sav->lft_h,
3825 					    SADB_EXT_LIFETIME_HARD);
3826 			if (!m)
3827 				goto fail;
3828 			break;
3829 
3830 		case SADB_EXT_LIFETIME_SOFT:
3831 			if (!sav->lft_s)
3832 				continue;
3833 			m = key_setlifetime(sav->lft_s,
3834 					    SADB_EXT_LIFETIME_SOFT);
3835 
3836 			if (!m)
3837 				goto fail;
3838 			break;
3839 
3840 		case SADB_X_EXT_NAT_T_TYPE:
3841 			if (sav->natt == NULL)
3842 				continue;
3843 			m = key_setsadbxtype(UDP_ENCAP_ESPINUDP);
3844 			if (!m)
3845 				goto fail;
3846 			break;
3847 
3848 		case SADB_X_EXT_NAT_T_DPORT:
3849 			if (sav->natt == NULL)
3850 				continue;
3851 			m = key_setsadbxport(sav->natt->dport,
3852 			    SADB_X_EXT_NAT_T_DPORT);
3853 			if (!m)
3854 				goto fail;
3855 			break;
3856 
3857 		case SADB_X_EXT_NAT_T_SPORT:
3858 			if (sav->natt == NULL)
3859 				continue;
3860 			m = key_setsadbxport(sav->natt->sport,
3861 			    SADB_X_EXT_NAT_T_SPORT);
3862 			if (!m)
3863 				goto fail;
3864 			break;
3865 
3866 		case SADB_X_EXT_NAT_T_OAI:
3867 			if (sav->natt == NULL ||
3868 			    (sav->natt->flags & IPSEC_NATT_F_OAI) == 0)
3869 				continue;
3870 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAI,
3871 			    &sav->natt->oai.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3872 			if (!m)
3873 				goto fail;
3874 			break;
3875 		case SADB_X_EXT_NAT_T_OAR:
3876 			if (sav->natt == NULL ||
3877 			    (sav->natt->flags & IPSEC_NATT_F_OAR) == 0)
3878 				continue;
3879 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAR,
3880 			    &sav->natt->oar.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3881 			if (!m)
3882 				goto fail;
3883 			break;
3884 		case SADB_X_EXT_NAT_T_FRAG:
3885 			/* We do not (yet) support those. */
3886 			continue;
3887 #ifdef IPSEC_OFFLOAD
3888 		case SADB_X_EXT_LFT_CUR_SW_OFFL:
3889 			if (!ipsec_accel_is_accel_sav(sav))
3890 				continue;
3891 			SAV_ADDREF(sav);
3892 			error = ipsec_accel_sa_lifetime_op(sav, &lft_c,
3893 			    NULL, IF_SA_CNT_TOTAL_SW_VAL, sahtree_trackerp);
3894 			if (error != 0) {
3895 				m = NULL;
3896 				goto fail;
3897 			}
3898 			m = key_setlifetime(&lft_c, dumporder[i]);
3899 			if (m == NULL)
3900 				goto fail;
3901 			key_freesav(&sav);
3902 			if (sav == NULL) {
3903 				m_freem(m);
3904 				goto fail;
3905 			}
3906 			break;
3907 		case SADB_X_EXT_LFT_CUR_HW_OFFL:
3908 			if (!ipsec_accel_is_accel_sav(sav))
3909 				continue;
3910 			memset(&lft_c, 0, sizeof(lft_c));
3911 			lft_c.bytes = sav->accel_hw_octets;
3912 			lft_c.allocations = sav->accel_hw_allocs;
3913 			m = key_setlifetime(&lft_c, dumporder[i]);
3914 			if (m == NULL)
3915 				goto fail;
3916 			break;
3917 		case SADB_X_EXT_IF_HW_OFFL:
3918 			if (!ipsec_accel_is_accel_sav(sav))
3919 				continue;
3920 			m = ipsec_accel_key_setaccelif(sav);
3921 			if (m == NULL)
3922 				continue;	/* benigh */
3923 			break;
3924 #endif
3925 
3926 		case SADB_EXT_ADDRESS_PROXY:
3927 		case SADB_EXT_IDENTITY_SRC:
3928 		case SADB_EXT_IDENTITY_DST:
3929 			/* XXX: should we brought from SPD ? */
3930 		case SADB_EXT_SENSITIVITY:
3931 		default:
3932 			continue;
3933 		}
3934 
3935 		if (!m)
3936 			goto fail;
3937 		if (tres)
3938 			m_cat(m, tres);
3939 		tres = m;
3940 	}
3941 
3942 	m_cat(result, tres);
3943 	tres = NULL;
3944 	if (result->m_len < sizeof(struct sadb_msg)) {
3945 		result = m_pullup(result, sizeof(struct sadb_msg));
3946 		if (result == NULL)
3947 			goto fail;
3948 	}
3949 
3950 	result->m_pkthdr.len = 0;
3951 	for (m = result; m; m = m->m_next)
3952 		result->m_pkthdr.len += m->m_len;
3953 
3954 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3955 	    PFKEY_UNIT64(result->m_pkthdr.len);
3956 
3957 	return result;
3958 
3959 fail:
3960 	m_freem(result);
3961 	m_freem(tres);
3962 	return NULL;
3963 }
3964 
3965 /*
3966  * set data into sadb_msg.
3967  */
3968 static struct mbuf *
3969 key_setsadbmsg(u_int8_t type, u_int16_t tlen, u_int8_t satype, u_int32_t seq,
3970     pid_t pid, u_int16_t reserved)
3971 {
3972 	struct mbuf *m;
3973 	struct sadb_msg *p;
3974 	int len;
3975 
3976 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
3977 	if (len > MCLBYTES)
3978 		return NULL;
3979 	m = key_mget(len);
3980 	if (m == NULL)
3981 		return NULL;
3982 	m->m_pkthdr.len = m->m_len = len;
3983 	m->m_next = NULL;
3984 
3985 	p = mtod(m, struct sadb_msg *);
3986 
3987 	bzero(p, len);
3988 	p->sadb_msg_version = PF_KEY_V2;
3989 	p->sadb_msg_type = type;
3990 	p->sadb_msg_errno = 0;
3991 	p->sadb_msg_satype = satype;
3992 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
3993 	p->sadb_msg_reserved = reserved;
3994 	p->sadb_msg_seq = seq;
3995 	p->sadb_msg_pid = (u_int32_t)pid;
3996 
3997 	return m;
3998 }
3999 
4000 /*
4001  * copy secasvar data into sadb_address.
4002  */
4003 static struct mbuf *
4004 key_setsadbsa(struct secasvar *sav)
4005 {
4006 	struct mbuf *m;
4007 	struct sadb_sa *p;
4008 	int len;
4009 
4010 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
4011 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4012 	if (m == NULL)
4013 		return (NULL);
4014 	m_align(m, len);
4015 	m->m_len = len;
4016 	p = mtod(m, struct sadb_sa *);
4017 	bzero(p, len);
4018 	p->sadb_sa_len = PFKEY_UNIT64(len);
4019 	p->sadb_sa_exttype = SADB_EXT_SA;
4020 	p->sadb_sa_spi = sav->spi;
4021 	p->sadb_sa_replay = sav->replay ?
4022 	    (sav->replay->wsize > UINT8_MAX ? UINT8_MAX :
4023 		sav->replay->wsize): 0;
4024 	p->sadb_sa_state = sav->state;
4025 	p->sadb_sa_auth = sav->alg_auth;
4026 	p->sadb_sa_encrypt = sav->alg_enc;
4027 	p->sadb_sa_flags = sav->flags & SADB_KEY_FLAGS_MAX;
4028 	return (m);
4029 }
4030 
4031 /*
4032  * set data into sadb_address.
4033  */
4034 static struct mbuf *
4035 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
4036     u_int8_t prefixlen, u_int16_t ul_proto)
4037 {
4038 	struct mbuf *m;
4039 	struct sadb_address *p;
4040 	size_t len;
4041 
4042 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
4043 	    PFKEY_ALIGN8(saddr->sa_len);
4044 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4045 	if (m == NULL)
4046 		return (NULL);
4047 	m_align(m, len);
4048 	m->m_len = len;
4049 	p = mtod(m, struct sadb_address *);
4050 
4051 	bzero(p, len);
4052 	p->sadb_address_len = PFKEY_UNIT64(len);
4053 	p->sadb_address_exttype = exttype;
4054 	p->sadb_address_proto = ul_proto;
4055 	if (prefixlen == FULLMASK) {
4056 		switch (saddr->sa_family) {
4057 		case AF_INET:
4058 			prefixlen = sizeof(struct in_addr) << 3;
4059 			break;
4060 		case AF_INET6:
4061 			prefixlen = sizeof(struct in6_addr) << 3;
4062 			break;
4063 		default:
4064 			; /*XXX*/
4065 		}
4066 	}
4067 	p->sadb_address_prefixlen = prefixlen;
4068 	p->sadb_address_reserved = 0;
4069 
4070 	bcopy(saddr,
4071 	    mtod(m, caddr_t) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
4072 	    saddr->sa_len);
4073 
4074 	return m;
4075 }
4076 
4077 /*
4078  * set data into sadb_x_sa2.
4079  */
4080 static struct mbuf *
4081 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int32_t reqid)
4082 {
4083 	struct mbuf *m;
4084 	struct sadb_x_sa2 *p;
4085 	size_t len;
4086 
4087 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
4088 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4089 	if (m == NULL)
4090 		return (NULL);
4091 	m_align(m, len);
4092 	m->m_len = len;
4093 	p = mtod(m, struct sadb_x_sa2 *);
4094 
4095 	bzero(p, len);
4096 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
4097 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
4098 	p->sadb_x_sa2_mode = mode;
4099 	p->sadb_x_sa2_reserved1 = 0;
4100 	p->sadb_x_sa2_reserved2 = 0;
4101 	p->sadb_x_sa2_sequence = seq;
4102 	p->sadb_x_sa2_reqid = reqid;
4103 
4104 	return m;
4105 }
4106 
4107 /*
4108  * Set data into sadb_x_sa_replay.
4109  */
4110 static struct mbuf *
4111 key_setsadbxsareplay(u_int32_t replay)
4112 {
4113 	struct mbuf *m;
4114 	struct sadb_x_sa_replay *p;
4115 	size_t len;
4116 
4117 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa_replay));
4118 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4119 	if (m == NULL)
4120 		return (NULL);
4121 	m_align(m, len);
4122 	m->m_len = len;
4123 	p = mtod(m, struct sadb_x_sa_replay *);
4124 
4125 	bzero(p, len);
4126 	p->sadb_x_sa_replay_len = PFKEY_UNIT64(len);
4127 	p->sadb_x_sa_replay_exttype = SADB_X_EXT_SA_REPLAY;
4128 	p->sadb_x_sa_replay_replay = (replay << 3);
4129 
4130 	return m;
4131 }
4132 
4133 /*
4134  * Set a type in sadb_x_nat_t_type.
4135  */
4136 static struct mbuf *
4137 key_setsadbxtype(u_int16_t type)
4138 {
4139 	struct mbuf *m;
4140 	size_t len;
4141 	struct sadb_x_nat_t_type *p;
4142 
4143 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
4144 
4145 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4146 	if (m == NULL)
4147 		return (NULL);
4148 	m_align(m, len);
4149 	m->m_len = len;
4150 	p = mtod(m, struct sadb_x_nat_t_type *);
4151 
4152 	bzero(p, len);
4153 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
4154 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
4155 	p->sadb_x_nat_t_type_type = type;
4156 
4157 	return (m);
4158 }
4159 /*
4160  * Set a port in sadb_x_nat_t_port.
4161  * In contrast to default RFC 2367 behaviour, port is in network byte order.
4162  */
4163 static struct mbuf *
4164 key_setsadbxport(u_int16_t port, u_int16_t type)
4165 {
4166 	struct mbuf *m;
4167 	size_t len;
4168 	struct sadb_x_nat_t_port *p;
4169 
4170 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
4171 
4172 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4173 	if (m == NULL)
4174 		return (NULL);
4175 	m_align(m, len);
4176 	m->m_len = len;
4177 	p = mtod(m, struct sadb_x_nat_t_port *);
4178 
4179 	bzero(p, len);
4180 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
4181 	p->sadb_x_nat_t_port_exttype = type;
4182 	p->sadb_x_nat_t_port_port = port;
4183 
4184 	return (m);
4185 }
4186 
4187 /*
4188  * Get port from sockaddr. Port is in network byte order.
4189  */
4190 uint16_t
4191 key_portfromsaddr(struct sockaddr *sa)
4192 {
4193 
4194 	switch (sa->sa_family) {
4195 #ifdef INET
4196 	case AF_INET:
4197 		return ((struct sockaddr_in *)sa)->sin_port;
4198 #endif
4199 #ifdef INET6
4200 	case AF_INET6:
4201 		return ((struct sockaddr_in6 *)sa)->sin6_port;
4202 #endif
4203 	}
4204 	return (0);
4205 }
4206 
4207 /*
4208  * Set port in struct sockaddr. Port is in network byte order.
4209  */
4210 void
4211 key_porttosaddr(struct sockaddr *sa, uint16_t port)
4212 {
4213 
4214 	switch (sa->sa_family) {
4215 #ifdef INET
4216 	case AF_INET:
4217 		((struct sockaddr_in *)sa)->sin_port = port;
4218 		break;
4219 #endif
4220 #ifdef INET6
4221 	case AF_INET6:
4222 		((struct sockaddr_in6 *)sa)->sin6_port = port;
4223 		break;
4224 #endif
4225 	default:
4226 		ipseclog((LOG_DEBUG, "%s: unexpected address family %d.\n",
4227 			__func__, sa->sa_family));
4228 		break;
4229 	}
4230 }
4231 
4232 /*
4233  * set data into sadb_x_policy
4234  */
4235 static struct mbuf *
4236 key_setsadbxpolicy(u_int16_t type, u_int8_t dir, u_int32_t id, u_int32_t priority)
4237 {
4238 	struct mbuf *m;
4239 	struct sadb_x_policy *p;
4240 	size_t len;
4241 
4242 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
4243 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
4244 	if (m == NULL)
4245 		return (NULL);
4246 	m_align(m, len);
4247 	m->m_len = len;
4248 	p = mtod(m, struct sadb_x_policy *);
4249 
4250 	bzero(p, len);
4251 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
4252 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
4253 	p->sadb_x_policy_type = type;
4254 	p->sadb_x_policy_dir = dir;
4255 	p->sadb_x_policy_id = id;
4256 	p->sadb_x_policy_priority = priority;
4257 
4258 	return m;
4259 }
4260 
4261 /* %%% utilities */
4262 /* Take a key message (sadb_key) from the socket and turn it into one
4263  * of the kernel's key structures (seckey).
4264  *
4265  * IN: pointer to the src
4266  * OUT: NULL no more memory
4267  */
4268 struct seckey *
4269 key_dup_keymsg(const struct sadb_key *src, size_t len,
4270     struct malloc_type *type)
4271 {
4272 	struct seckey *dst;
4273 
4274 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4275 	if (dst != NULL) {
4276 		dst->bits = src->sadb_key_bits;
4277 		dst->key_data = malloc(len, type, M_NOWAIT);
4278 		if (dst->key_data != NULL) {
4279 			bcopy((const char *)(src + 1), dst->key_data, len);
4280 		} else {
4281 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4282 			    __func__));
4283 			free(dst, type);
4284 			dst = NULL;
4285 		}
4286 	} else {
4287 		ipseclog((LOG_DEBUG, "%s: No more memory.\n",
4288 		    __func__));
4289 	}
4290 	return (dst);
4291 }
4292 
4293 /* Take a lifetime message (sadb_lifetime) passed in on a socket and
4294  * turn it into one of the kernel's lifetime structures (seclifetime).
4295  *
4296  * IN: pointer to the destination, source and malloc type
4297  * OUT: NULL, no more memory
4298  */
4299 
4300 static struct seclifetime *
4301 key_dup_lifemsg(const struct sadb_lifetime *src, struct malloc_type *type)
4302 {
4303 	struct seclifetime *dst;
4304 
4305 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
4306 	if (dst == NULL) {
4307 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
4308 		return (NULL);
4309 	}
4310 	dst->allocations = src->sadb_lifetime_allocations;
4311 	dst->bytes = src->sadb_lifetime_bytes;
4312 	dst->addtime = src->sadb_lifetime_addtime;
4313 	dst->usetime = src->sadb_lifetime_usetime;
4314 	return (dst);
4315 }
4316 
4317 /*
4318  * compare two secasindex structure.
4319  * flag can specify to compare 2 saidxes.
4320  * compare two secasindex structure without both mode and reqid.
4321  * don't compare port.
4322  * IN:
4323  *      saidx0: source, it can be in SAD.
4324  *      saidx1: object.
4325  * OUT:
4326  *      1 : equal
4327  *      0 : not equal
4328  */
4329 static int
4330 key_cmpsaidx(const struct secasindex *saidx0, const struct secasindex *saidx1,
4331     int flag)
4332 {
4333 
4334 	/* sanity */
4335 	if (saidx0 == NULL && saidx1 == NULL)
4336 		return 1;
4337 
4338 	if (saidx0 == NULL || saidx1 == NULL)
4339 		return 0;
4340 
4341 	if (saidx0->proto != saidx1->proto)
4342 		return 0;
4343 
4344 	if (flag == CMP_EXACTLY) {
4345 		if (saidx0->mode != saidx1->mode)
4346 			return 0;
4347 		if (saidx0->reqid != saidx1->reqid)
4348 			return 0;
4349 		if (bcmp(&saidx0->src, &saidx1->src,
4350 		    saidx0->src.sa.sa_len) != 0 ||
4351 		    bcmp(&saidx0->dst, &saidx1->dst,
4352 		    saidx0->dst.sa.sa_len) != 0)
4353 			return 0;
4354 	} else {
4355 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
4356 		if (flag == CMP_MODE_REQID || flag == CMP_REQID) {
4357 			/*
4358 			 * If reqid of SPD is non-zero, unique SA is required.
4359 			 * The result must be of same reqid in this case.
4360 			 */
4361 			if (saidx1->reqid != 0 &&
4362 			    saidx0->reqid != saidx1->reqid)
4363 				return 0;
4364 		}
4365 
4366 		if (flag == CMP_MODE_REQID) {
4367 			if (saidx0->mode != IPSEC_MODE_ANY
4368 			 && saidx0->mode != saidx1->mode)
4369 				return 0;
4370 		}
4371 
4372 		if (key_sockaddrcmp(&saidx0->src.sa, &saidx1->src.sa, 0) != 0)
4373 			return 0;
4374 		if (key_sockaddrcmp(&saidx0->dst.sa, &saidx1->dst.sa, 0) != 0)
4375 			return 0;
4376 	}
4377 
4378 	return 1;
4379 }
4380 
4381 /*
4382  * compare two secindex structure exactly.
4383  * IN:
4384  *	spidx0: source, it is often in SPD.
4385  *	spidx1: object, it is often from PFKEY message.
4386  * OUT:
4387  *	1 : equal
4388  *	0 : not equal
4389  */
4390 static int
4391 key_cmpspidx_exactly(struct secpolicyindex *spidx0,
4392     struct secpolicyindex *spidx1)
4393 {
4394 	/* sanity */
4395 	if (spidx0 == NULL && spidx1 == NULL)
4396 		return 1;
4397 
4398 	if (spidx0 == NULL || spidx1 == NULL)
4399 		return 0;
4400 
4401 	if (spidx0->prefs != spidx1->prefs
4402 	 || spidx0->prefd != spidx1->prefd
4403 	 || spidx0->ul_proto != spidx1->ul_proto
4404 	 || spidx0->dir != spidx1->dir)
4405 		return 0;
4406 
4407 	return key_sockaddrcmp(&spidx0->src.sa, &spidx1->src.sa, 1) == 0 &&
4408 	       key_sockaddrcmp(&spidx0->dst.sa, &spidx1->dst.sa, 1) == 0;
4409 }
4410 
4411 /*
4412  * compare two secindex structure with mask.
4413  * IN:
4414  *	spidx0: source, it is often in SPD.
4415  *	spidx1: object, it is often from IP header.
4416  * OUT:
4417  *	1 : equal
4418  *	0 : not equal
4419  */
4420 static int
4421 key_cmpspidx_withmask(struct secpolicyindex *spidx0,
4422     struct secpolicyindex *spidx1)
4423 {
4424 	/* sanity */
4425 	if (spidx0 == NULL && spidx1 == NULL)
4426 		return 1;
4427 
4428 	if (spidx0 == NULL || spidx1 == NULL)
4429 		return 0;
4430 
4431 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4432 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4433 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4434 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
4435 		return 0;
4436 
4437 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4438 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY
4439 	 && spidx0->ul_proto != spidx1->ul_proto)
4440 		return 0;
4441 
4442 	switch (spidx0->src.sa.sa_family) {
4443 	case AF_INET:
4444 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY
4445 		 && spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
4446 			return 0;
4447 		if (!key_bbcmp(&spidx0->src.sin.sin_addr,
4448 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
4449 			return 0;
4450 		break;
4451 	case AF_INET6:
4452 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY
4453 		 && spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
4454 			return 0;
4455 		/*
4456 		 * scope_id check. if sin6_scope_id is 0, we regard it
4457 		 * as a wildcard scope, which matches any scope zone ID.
4458 		 */
4459 		if (spidx0->src.sin6.sin6_scope_id &&
4460 		    spidx1->src.sin6.sin6_scope_id &&
4461 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
4462 			return 0;
4463 		if (!key_bbcmp(&spidx0->src.sin6.sin6_addr,
4464 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
4465 			return 0;
4466 		break;
4467 	default:
4468 		/* XXX */
4469 		if (bcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
4470 			return 0;
4471 		break;
4472 	}
4473 
4474 	switch (spidx0->dst.sa.sa_family) {
4475 	case AF_INET:
4476 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY
4477 		 && spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
4478 			return 0;
4479 		if (!key_bbcmp(&spidx0->dst.sin.sin_addr,
4480 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
4481 			return 0;
4482 		break;
4483 	case AF_INET6:
4484 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY
4485 		 && spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
4486 			return 0;
4487 		/*
4488 		 * scope_id check. if sin6_scope_id is 0, we regard it
4489 		 * as a wildcard scope, which matches any scope zone ID.
4490 		 */
4491 		if (spidx0->dst.sin6.sin6_scope_id &&
4492 		    spidx1->dst.sin6.sin6_scope_id &&
4493 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
4494 			return 0;
4495 		if (!key_bbcmp(&spidx0->dst.sin6.sin6_addr,
4496 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
4497 			return 0;
4498 		break;
4499 	default:
4500 		/* XXX */
4501 		if (bcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
4502 			return 0;
4503 		break;
4504 	}
4505 
4506 	/* XXX Do we check other field ?  e.g. flowinfo */
4507 
4508 	return 1;
4509 }
4510 
4511 #ifdef satosin
4512 #undef satosin
4513 #endif
4514 #define satosin(s) ((const struct sockaddr_in *)s)
4515 #ifdef satosin6
4516 #undef satosin6
4517 #endif
4518 #define satosin6(s) ((const struct sockaddr_in6 *)s)
4519 /* returns 0 on match */
4520 int
4521 key_sockaddrcmp(const struct sockaddr *sa1, const struct sockaddr *sa2,
4522     int port)
4523 {
4524 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4525 		return 1;
4526 
4527 	switch (sa1->sa_family) {
4528 #ifdef INET
4529 	case AF_INET:
4530 		if (sa1->sa_len != sizeof(struct sockaddr_in))
4531 			return 1;
4532 		if (satosin(sa1)->sin_addr.s_addr !=
4533 		    satosin(sa2)->sin_addr.s_addr) {
4534 			return 1;
4535 		}
4536 		if (port && satosin(sa1)->sin_port != satosin(sa2)->sin_port)
4537 			return 1;
4538 		break;
4539 #endif
4540 #ifdef INET6
4541 	case AF_INET6:
4542 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4543 			return 1;	/*EINVAL*/
4544 		if (satosin6(sa1)->sin6_scope_id !=
4545 		    satosin6(sa2)->sin6_scope_id) {
4546 			return 1;
4547 		}
4548 		if (!IN6_ARE_ADDR_EQUAL(&satosin6(sa1)->sin6_addr,
4549 		    &satosin6(sa2)->sin6_addr)) {
4550 			return 1;
4551 		}
4552 		if (port &&
4553 		    satosin6(sa1)->sin6_port != satosin6(sa2)->sin6_port) {
4554 			return 1;
4555 		}
4556 		break;
4557 #endif
4558 	default:
4559 		if (bcmp(sa1, sa2, sa1->sa_len) != 0)
4560 			return 1;
4561 		break;
4562 	}
4563 
4564 	return 0;
4565 }
4566 
4567 /* returns 0 on match */
4568 int
4569 key_sockaddrcmp_withmask(const struct sockaddr *sa1,
4570     const struct sockaddr *sa2, size_t mask)
4571 {
4572 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4573 		return (1);
4574 
4575 	switch (sa1->sa_family) {
4576 #ifdef INET
4577 	case AF_INET:
4578 		return (!key_bbcmp(&satosin(sa1)->sin_addr,
4579 		    &satosin(sa2)->sin_addr, mask));
4580 #endif
4581 #ifdef INET6
4582 	case AF_INET6:
4583 		if (satosin6(sa1)->sin6_scope_id !=
4584 		    satosin6(sa2)->sin6_scope_id)
4585 			return (1);
4586 		return (!key_bbcmp(&satosin6(sa1)->sin6_addr,
4587 		    &satosin6(sa2)->sin6_addr, mask));
4588 #endif
4589 	}
4590 	return (1);
4591 }
4592 #undef satosin
4593 #undef satosin6
4594 
4595 /*
4596  * compare two buffers with mask.
4597  * IN:
4598  *	addr1: source
4599  *	addr2: object
4600  *	bits:  Number of bits to compare
4601  * OUT:
4602  *	1 : equal
4603  *	0 : not equal
4604  */
4605 static int
4606 key_bbcmp(const void *a1, const void *a2, u_int bits)
4607 {
4608 	const unsigned char *p1 = a1;
4609 	const unsigned char *p2 = a2;
4610 
4611 	/* XXX: This could be considerably faster if we compare a word
4612 	 * at a time, but it is complicated on LSB Endian machines */
4613 
4614 	/* Handle null pointers */
4615 	if (p1 == NULL || p2 == NULL)
4616 		return (p1 == p2);
4617 
4618 	while (bits >= 8) {
4619 		if (*p1++ != *p2++)
4620 			return 0;
4621 		bits -= 8;
4622 	}
4623 
4624 	if (bits > 0) {
4625 		u_int8_t mask = ~((1<<(8-bits))-1);
4626 		if ((*p1 & mask) != (*p2 & mask))
4627 			return 0;
4628 	}
4629 	return 1;	/* Match! */
4630 }
4631 
4632 static void
4633 key_flush_spd(time_t now)
4634 {
4635 	SPTREE_RLOCK_TRACKER;
4636 	struct secpolicy_list drainq;
4637 	struct secpolicy *sp, *nextsp;
4638 	u_int dir;
4639 
4640 	LIST_INIT(&drainq);
4641 	SPTREE_RLOCK();
4642 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4643 		TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
4644 			if (sp->lifetime == 0 && sp->validtime == 0)
4645 				continue;
4646 			if ((sp->lifetime &&
4647 			    now - sp->created > sp->lifetime) ||
4648 			    (sp->validtime &&
4649 			    now - sp->lastused > sp->validtime)) {
4650 				/* Hold extra reference to send SPDEXPIRE */
4651 				SP_ADDREF(sp);
4652 				LIST_INSERT_HEAD(&drainq, sp, drainq);
4653 			}
4654 		}
4655 	}
4656 	SPTREE_RUNLOCK();
4657 	if (LIST_EMPTY(&drainq))
4658 		return;
4659 
4660 	SPTREE_WLOCK();
4661 	sp = LIST_FIRST(&drainq);
4662 	while (sp != NULL) {
4663 		nextsp = LIST_NEXT(sp, drainq);
4664 		/* Check that SP is still linked */
4665 		if (sp->state != IPSEC_SPSTATE_ALIVE) {
4666 			LIST_REMOVE(sp, drainq);
4667 			key_freesp(&sp); /* release extra reference */
4668 			sp = nextsp;
4669 			continue;
4670 		}
4671 		TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
4672 		V_spd_size--;
4673 		LIST_REMOVE(sp, idhash);
4674 		sp->state = IPSEC_SPSTATE_DEAD;
4675 		ipsec_accel_spddel(sp);
4676 		sp = nextsp;
4677 	}
4678 	V_sp_genid++;
4679 	SPTREE_WUNLOCK();
4680 	if (SPDCACHE_ENABLED())
4681 		spdcache_clear();
4682 
4683 	sp = LIST_FIRST(&drainq);
4684 	while (sp != NULL) {
4685 		nextsp = LIST_NEXT(sp, drainq);
4686 		key_spdexpire(sp);
4687 		key_freesp(&sp); /* release extra reference */
4688 		key_freesp(&sp); /* release last reference */
4689 		sp = nextsp;
4690 	}
4691 }
4692 
4693 static void
4694 key_flush_sad(time_t now)
4695 {
4696 	SAHTREE_RLOCK_TRACKER;
4697 	struct secashead_list emptyq;
4698 	struct secasvar_list drainq, hexpireq, sexpireq, freeq;
4699 	struct secashead *sah, *nextsah;
4700 	struct secasvar *sav, *nextsav;
4701 
4702 	SECASVAR_RLOCK_TRACKER;
4703 
4704 	LIST_INIT(&drainq);
4705 	LIST_INIT(&hexpireq);
4706 	LIST_INIT(&sexpireq);
4707 	LIST_INIT(&emptyq);
4708 
4709 	SAHTREE_RLOCK();
4710 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
4711 		/* Check for empty SAH */
4712 		if (TAILQ_EMPTY(&sah->savtree_larval) &&
4713 		    TAILQ_EMPTY(&sah->savtree_alive)) {
4714 			SAH_ADDREF(sah);
4715 			LIST_INSERT_HEAD(&emptyq, sah, drainq);
4716 			continue;
4717 		}
4718 		/* Add all stale LARVAL SAs into drainq */
4719 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
4720 			if (now - sav->created < V_key_larval_lifetime)
4721 				continue;
4722 			SAV_ADDREF(sav);
4723 			LIST_INSERT_HEAD(&drainq, sav, drainq);
4724 		}
4725 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
4726 			/* lifetimes aren't specified */
4727 			if (sav->lft_h == NULL)
4728 				continue;
4729 			SECASVAR_RLOCK(sav);
4730 			/*
4731 			 * Check again with lock held, because it may
4732 			 * be updated by SADB_UPDATE.
4733 			 */
4734 			if (sav->lft_h == NULL) {
4735 				SECASVAR_RUNLOCK(sav);
4736 				continue;
4737 			}
4738 			/*
4739 			 * RFC 2367:
4740 			 * HARD lifetimes MUST take precedence over SOFT
4741 			 * lifetimes, meaning if the HARD and SOFT lifetimes
4742 			 * are the same, the HARD lifetime will appear on the
4743 			 * EXPIRE message.
4744 			 */
4745 			/* check HARD lifetime */
4746 			if ((sav->lft_h->addtime != 0 &&
4747 			    now - sav->created > sav->lft_h->addtime) ||
4748 			    (sav->lft_h->usetime != 0 && sav->firstused &&
4749 			    now - sav->firstused > sav->lft_h->usetime) ||
4750 			    (sav->lft_h->bytes != 0 && counter_u64_fetch(
4751 			        sav->lft_c_bytes) > sav->lft_h->bytes)) {
4752 				SECASVAR_RUNLOCK(sav);
4753 				SAV_ADDREF(sav);
4754 				LIST_INSERT_HEAD(&hexpireq, sav, drainq);
4755 				continue;
4756 			}
4757 			/* check SOFT lifetime (only for MATURE SAs) */
4758 			if (sav->state == SADB_SASTATE_MATURE && (
4759 			    (sav->lft_s->addtime != 0 &&
4760 			    now - sav->created > sav->lft_s->addtime) ||
4761 			    (sav->lft_s->usetime != 0 && sav->firstused &&
4762 			    now - sav->firstused > sav->lft_s->usetime) ||
4763 			    (sav->lft_s->bytes != 0 && counter_u64_fetch(
4764 				sav->lft_c_bytes) > sav->lft_s->bytes) ||
4765 			    (!(sav->flags & SADB_X_SAFLAGS_ESN) &&
4766 			    (sav->replay != NULL) && (
4767 			    (sav->replay->count > UINT32_80PCT) ||
4768 			    (sav->replay->last > UINT32_80PCT))))) {
4769 				SECASVAR_RUNLOCK(sav);
4770 				SAV_ADDREF(sav);
4771 				LIST_INSERT_HEAD(&sexpireq, sav, drainq);
4772 				continue;
4773 			}
4774 			SECASVAR_RUNLOCK(sav);
4775 		}
4776 	}
4777 	SAHTREE_RUNLOCK();
4778 
4779 	if (LIST_EMPTY(&emptyq) && LIST_EMPTY(&drainq) &&
4780 	    LIST_EMPTY(&hexpireq) && LIST_EMPTY(&sexpireq))
4781 		return;
4782 
4783 	LIST_INIT(&freeq);
4784 	SAHTREE_WLOCK();
4785 	/* Unlink stale LARVAL SAs */
4786 	sav = LIST_FIRST(&drainq);
4787 	while (sav != NULL) {
4788 		nextsav = LIST_NEXT(sav, drainq);
4789 		/* Check that SA is still LARVAL */
4790 		if (sav->state != SADB_SASTATE_LARVAL) {
4791 			LIST_REMOVE(sav, drainq);
4792 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4793 			sav = nextsav;
4794 			continue;
4795 		}
4796 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
4797 		LIST_REMOVE(sav, spihash);
4798 		sav->state = SADB_SASTATE_DEAD;
4799 		ipsec_accel_forget_sav(sav);
4800 		sav = nextsav;
4801 	}
4802 	/* Unlink all SAs with expired HARD lifetime */
4803 	sav = LIST_FIRST(&hexpireq);
4804 	while (sav != NULL) {
4805 		nextsav = LIST_NEXT(sav, drainq);
4806 		/* Check that SA is not unlinked */
4807 		if (sav->state == SADB_SASTATE_DEAD) {
4808 			LIST_REMOVE(sav, drainq);
4809 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4810 			sav = nextsav;
4811 			continue;
4812 		}
4813 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
4814 		LIST_REMOVE(sav, spihash);
4815 		sav->state = SADB_SASTATE_DEAD;
4816 		ipsec_accel_forget_sav(sav);
4817 		sav = nextsav;
4818 	}
4819 	/* Mark all SAs with expired SOFT lifetime as DYING */
4820 	sav = LIST_FIRST(&sexpireq);
4821 	while (sav != NULL) {
4822 		nextsav = LIST_NEXT(sav, drainq);
4823 		/* Check that SA is not unlinked */
4824 		if (sav->state == SADB_SASTATE_DEAD) {
4825 			LIST_REMOVE(sav, drainq);
4826 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4827 			sav = nextsav;
4828 			continue;
4829 		}
4830 		/*
4831 		 * NOTE: this doesn't change SA order in the chain.
4832 		 */
4833 		sav->state = SADB_SASTATE_DYING;
4834 		sav = nextsav;
4835 	}
4836 	/* Unlink empty SAHs */
4837 	sah = LIST_FIRST(&emptyq);
4838 	while (sah != NULL) {
4839 		nextsah = LIST_NEXT(sah, drainq);
4840 		/* Check that SAH is still empty and not unlinked */
4841 		if (sah->state == SADB_SASTATE_DEAD ||
4842 		    !TAILQ_EMPTY(&sah->savtree_larval) ||
4843 		    !TAILQ_EMPTY(&sah->savtree_alive)) {
4844 			LIST_REMOVE(sah, drainq);
4845 			key_freesah(&sah); /* release extra reference */
4846 			sah = nextsah;
4847 			continue;
4848 		}
4849 		TAILQ_REMOVE(&V_sahtree, sah, chain);
4850 		LIST_REMOVE(sah, addrhash);
4851 		sah->state = SADB_SASTATE_DEAD;
4852 		sah = nextsah;
4853 	}
4854 	SAHTREE_WUNLOCK();
4855 
4856 	/* Send SPDEXPIRE messages */
4857 	sav = LIST_FIRST(&hexpireq);
4858 	while (sav != NULL) {
4859 		nextsav = LIST_NEXT(sav, drainq);
4860 		key_expire(sav, 1);
4861 		key_freesah(&sav->sah); /* release reference from SAV */
4862 		key_freesav(&sav); /* release extra reference */
4863 		key_freesav(&sav); /* release last reference */
4864 		sav = nextsav;
4865 	}
4866 	sav = LIST_FIRST(&sexpireq);
4867 	while (sav != NULL) {
4868 		nextsav = LIST_NEXT(sav, drainq);
4869 		key_expire(sav, 0);
4870 		key_freesav(&sav); /* release extra reference */
4871 		sav = nextsav;
4872 	}
4873 	/* Free stale LARVAL SAs */
4874 	sav = LIST_FIRST(&drainq);
4875 	while (sav != NULL) {
4876 		nextsav = LIST_NEXT(sav, drainq);
4877 		key_freesah(&sav->sah); /* release reference from SAV */
4878 		key_freesav(&sav); /* release extra reference */
4879 		key_freesav(&sav); /* release last reference */
4880 		sav = nextsav;
4881 	}
4882 	/* Free SAs that were unlinked/changed by someone else */
4883 	sav = LIST_FIRST(&freeq);
4884 	while (sav != NULL) {
4885 		nextsav = LIST_NEXT(sav, drainq);
4886 		key_freesav(&sav); /* release extra reference */
4887 		sav = nextsav;
4888 	}
4889 	/* Free empty SAH */
4890 	sah = LIST_FIRST(&emptyq);
4891 	while (sah != NULL) {
4892 		nextsah = LIST_NEXT(sah, drainq);
4893 		key_freesah(&sah); /* release extra reference */
4894 		key_freesah(&sah); /* release last reference */
4895 		sah = nextsah;
4896 	}
4897 }
4898 
4899 static void
4900 key_flush_acq(time_t now)
4901 {
4902 	struct secacq *acq, *nextacq;
4903 
4904 	/* ACQ tree */
4905 	ACQ_LOCK();
4906 	acq = LIST_FIRST(&V_acqtree);
4907 	while (acq != NULL) {
4908 		nextacq = LIST_NEXT(acq, chain);
4909 		if (now - acq->created > V_key_blockacq_lifetime) {
4910 			LIST_REMOVE(acq, chain);
4911 			LIST_REMOVE(acq, addrhash);
4912 			LIST_REMOVE(acq, seqhash);
4913 			free(acq, M_IPSEC_SAQ);
4914 		}
4915 		acq = nextacq;
4916 	}
4917 	ACQ_UNLOCK();
4918 }
4919 
4920 static void
4921 key_flush_spacq(time_t now)
4922 {
4923 	struct secspacq *acq, *nextacq;
4924 
4925 	/* SP ACQ tree */
4926 	SPACQ_LOCK();
4927 	for (acq = LIST_FIRST(&V_spacqtree); acq != NULL; acq = nextacq) {
4928 		nextacq = LIST_NEXT(acq, chain);
4929 		if (now - acq->created > V_key_blockacq_lifetime
4930 		 && __LIST_CHAINED(acq)) {
4931 			LIST_REMOVE(acq, chain);
4932 			free(acq, M_IPSEC_SAQ);
4933 		}
4934 	}
4935 	SPACQ_UNLOCK();
4936 }
4937 
4938 /*
4939  * time handler.
4940  * scanning SPD and SAD to check status for each entries,
4941  * and do to remove or to expire.
4942  * XXX: year 2038 problem may remain.
4943  */
4944 static void
4945 key_timehandler(void *arg)
4946 {
4947 	VNET_ITERATOR_DECL(vnet_iter);
4948 	time_t now = time_second;
4949 
4950 	VNET_LIST_RLOCK_NOSLEEP();
4951 	VNET_FOREACH(vnet_iter) {
4952 		CURVNET_SET(vnet_iter);
4953 		key_flush_spd(now);
4954 		key_flush_sad(now);
4955 		key_flush_acq(now);
4956 		key_flush_spacq(now);
4957 		CURVNET_RESTORE();
4958 	}
4959 	VNET_LIST_RUNLOCK_NOSLEEP();
4960 
4961 #ifndef IPSEC_DEBUG2
4962 	/* do exchange to tick time !! */
4963 	callout_schedule(&key_timer, hz);
4964 #endif /* IPSEC_DEBUG2 */
4965 }
4966 
4967 u_long
4968 key_random(void)
4969 {
4970 	u_long value;
4971 
4972 	arc4random_buf(&value, sizeof(value));
4973 	return value;
4974 }
4975 
4976 /*
4977  * map SADB_SATYPE_* to IPPROTO_*.
4978  * if satype == SADB_SATYPE then satype is mapped to ~0.
4979  * OUT:
4980  *	0: invalid satype.
4981  */
4982 static uint8_t
4983 key_satype2proto(uint8_t satype)
4984 {
4985 	switch (satype) {
4986 	case SADB_SATYPE_UNSPEC:
4987 		return IPSEC_PROTO_ANY;
4988 	case SADB_SATYPE_AH:
4989 		return IPPROTO_AH;
4990 	case SADB_SATYPE_ESP:
4991 		return IPPROTO_ESP;
4992 	case SADB_X_SATYPE_IPCOMP:
4993 		return IPPROTO_IPCOMP;
4994 	case SADB_X_SATYPE_TCPSIGNATURE:
4995 		return IPPROTO_TCP;
4996 	default:
4997 		return 0;
4998 	}
4999 	/* NOTREACHED */
5000 }
5001 
5002 /*
5003  * map IPPROTO_* to SADB_SATYPE_*
5004  * OUT:
5005  *	0: invalid protocol type.
5006  */
5007 static uint8_t
5008 key_proto2satype(uint8_t proto)
5009 {
5010 	switch (proto) {
5011 	case IPPROTO_AH:
5012 		return SADB_SATYPE_AH;
5013 	case IPPROTO_ESP:
5014 		return SADB_SATYPE_ESP;
5015 	case IPPROTO_IPCOMP:
5016 		return SADB_X_SATYPE_IPCOMP;
5017 	case IPPROTO_TCP:
5018 		return SADB_X_SATYPE_TCPSIGNATURE;
5019 	default:
5020 		return 0;
5021 	}
5022 	/* NOTREACHED */
5023 }
5024 
5025 /* %%% PF_KEY */
5026 /*
5027  * SADB_GETSPI processing is to receive
5028  *	<base, (SA2), src address, dst address, (SPI range)>
5029  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
5030  * tree with the status of LARVAL, and send
5031  *	<base, SA(*), address(SD)>
5032  * to the IKMPd.
5033  *
5034  * IN:	mhp: pointer to the pointer to each header.
5035  * OUT:	NULL if fail.
5036  *	other if success, return pointer to the message to send.
5037  */
5038 static int
5039 key_getspi(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5040 {
5041 	struct secasindex saidx;
5042 	struct sadb_address *src0, *dst0;
5043 	struct secasvar *sav;
5044 	uint32_t reqid, spi;
5045 	int error;
5046 	uint8_t mode, proto;
5047 
5048 	IPSEC_ASSERT(so != NULL, ("null socket"));
5049 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5050 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5051 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5052 
5053 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5054 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)
5055 #ifdef PFKEY_STRICT_CHECKS
5056 	    || SADB_CHECKHDR(mhp, SADB_EXT_SPIRANGE)
5057 #endif
5058 	    ) {
5059 		ipseclog((LOG_DEBUG,
5060 		    "%s: invalid message: missing required header.\n",
5061 		    __func__));
5062 		error = EINVAL;
5063 		goto fail;
5064 	}
5065 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5066 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)
5067 #ifdef PFKEY_STRICT_CHECKS
5068 	    || SADB_CHECKLEN(mhp, SADB_EXT_SPIRANGE)
5069 #endif
5070 	    ) {
5071 		ipseclog((LOG_DEBUG,
5072 		    "%s: invalid message: wrong header size.\n", __func__));
5073 		error = EINVAL;
5074 		goto fail;
5075 	}
5076 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5077 		mode = IPSEC_MODE_ANY;
5078 		reqid = 0;
5079 	} else {
5080 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5081 			ipseclog((LOG_DEBUG,
5082 			    "%s: invalid message: wrong header size.\n",
5083 			    __func__));
5084 			error = EINVAL;
5085 			goto fail;
5086 		}
5087 		mode = ((struct sadb_x_sa2 *)
5088 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5089 		reqid = ((struct sadb_x_sa2 *)
5090 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5091 	}
5092 
5093 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5094 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5095 
5096 	/* map satype to proto */
5097 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5098 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5099 			__func__));
5100 		error = EINVAL;
5101 		goto fail;
5102 	}
5103 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5104 	    (struct sockaddr *)(dst0 + 1));
5105 	if (error != 0) {
5106 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5107 		error = EINVAL;
5108 		goto fail;
5109 	}
5110 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5111 
5112 	/* SPI allocation */
5113 	SPI_ALLOC_LOCK();
5114 	spi = key_do_getnewspi(
5115 	    (struct sadb_spirange *)mhp->ext[SADB_EXT_SPIRANGE], &saidx);
5116 	if (spi == 0) {
5117 		/*
5118 		 * Requested SPI or SPI range is not available or
5119 		 * already used.
5120 		 */
5121 		SPI_ALLOC_UNLOCK();
5122 		error = EEXIST;
5123 		goto fail;
5124 	}
5125 	sav = key_newsav(mhp, &saidx, spi, &error);
5126 	SPI_ALLOC_UNLOCK();
5127 	if (sav == NULL)
5128 		goto fail;
5129 
5130 	if (sav->seq != 0) {
5131 		/*
5132 		 * RFC2367:
5133 		 * If the SADB_GETSPI message is in response to a
5134 		 * kernel-generated SADB_ACQUIRE, the sadb_msg_seq
5135 		 * MUST be the same as the SADB_ACQUIRE message.
5136 		 *
5137 		 * XXXAE: However it doesn't definethe behaviour how to
5138 		 * check this and what to do if it doesn't match.
5139 		 * Also what we should do if it matches?
5140 		 *
5141 		 * We can compare saidx used in SADB_ACQUIRE with saidx
5142 		 * used in SADB_GETSPI, but this probably can break
5143 		 * existing software. For now just warn if it doesn't match.
5144 		 *
5145 		 * XXXAE: anyway it looks useless.
5146 		 */
5147 		key_acqdone(&saidx, sav->seq);
5148 	}
5149 	KEYDBG(KEY_STAMP,
5150 	    printf("%s: SA(%p)\n", __func__, sav));
5151 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5152 
5153     {
5154 	struct mbuf *n, *nn;
5155 	struct sadb_sa *m_sa;
5156 	struct sadb_msg *newmsg;
5157 	int off, len;
5158 
5159 	/* create new sadb_msg to reply. */
5160 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
5161 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
5162 
5163 	n = key_mget(len);
5164 	if (n == NULL) {
5165 		error = ENOBUFS;
5166 		goto fail;
5167 	}
5168 
5169 	n->m_len = len;
5170 	n->m_next = NULL;
5171 	off = 0;
5172 
5173 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
5174 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
5175 
5176 	m_sa = (struct sadb_sa *)(mtod(n, caddr_t) + off);
5177 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
5178 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
5179 	m_sa->sadb_sa_spi = spi; /* SPI is already in network byte order */
5180 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
5181 
5182 	IPSEC_ASSERT(off == len,
5183 		("length inconsistency (off %u len %u)", off, len));
5184 
5185 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
5186 	    SADB_EXT_ADDRESS_DST);
5187 	if (!n->m_next) {
5188 		m_freem(n);
5189 		error = ENOBUFS;
5190 		goto fail;
5191 	}
5192 
5193 	if (n->m_len < sizeof(struct sadb_msg)) {
5194 		n = m_pullup(n, sizeof(struct sadb_msg));
5195 		if (n == NULL)
5196 			return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
5197 	}
5198 
5199 	n->m_pkthdr.len = 0;
5200 	for (nn = n; nn; nn = nn->m_next)
5201 		n->m_pkthdr.len += nn->m_len;
5202 
5203 	newmsg = mtod(n, struct sadb_msg *);
5204 	newmsg->sadb_msg_seq = sav->seq;
5205 	newmsg->sadb_msg_errno = 0;
5206 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
5207 
5208 	m_freem(m);
5209 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
5210     }
5211 
5212 fail:
5213 	return (key_senderror(so, m, error));
5214 }
5215 
5216 /*
5217  * allocating new SPI
5218  * called by key_getspi().
5219  * OUT:
5220  *	0:	failure.
5221  *	others: success, SPI in network byte order.
5222  */
5223 static uint32_t
5224 key_do_getnewspi(struct sadb_spirange *spirange, struct secasindex *saidx)
5225 {
5226 	uint32_t min, max, newspi, t;
5227 	int tries, limit;
5228 
5229 	SPI_ALLOC_LOCK_ASSERT();
5230 
5231 	/* set spi range to allocate */
5232 	if (spirange != NULL) {
5233 		min = spirange->sadb_spirange_min;
5234 		max = spirange->sadb_spirange_max;
5235 	} else {
5236 		min = V_key_spi_minval;
5237 		max = V_key_spi_maxval;
5238 	}
5239 	/* IPCOMP needs 2-byte SPI */
5240 	if (saidx->proto == IPPROTO_IPCOMP) {
5241 		if (min >= 0x10000)
5242 			min = 0xffff;
5243 		if (max >= 0x10000)
5244 			max = 0xffff;
5245 		if (min > max) {
5246 			t = min; min = max; max = t;
5247 		}
5248 	}
5249 
5250 	if (min == max) {
5251 		if (key_checkspidup(htonl(min))) {
5252 			ipseclog((LOG_DEBUG, "%s: SPI %u exists already.\n",
5253 			    __func__, min));
5254 			return 0;
5255 		}
5256 
5257 		tries = 1;
5258 		newspi = min;
5259 	} else {
5260 		/* init SPI */
5261 		newspi = 0;
5262 
5263 		limit = atomic_load_int(&V_key_spi_trycnt);
5264 		/* when requesting to allocate spi ranged */
5265 		for (tries = 0; tries < limit; tries++) {
5266 			/* generate pseudo-random SPI value ranged. */
5267 			newspi = min + (key_random() % (max - min + 1));
5268 			if (!key_checkspidup(htonl(newspi)))
5269 				break;
5270 		}
5271 
5272 		if (tries == limit || newspi == 0) {
5273 			ipseclog((LOG_DEBUG,
5274 			    "%s: failed to allocate SPI.\n", __func__));
5275 			return 0;
5276 		}
5277 	}
5278 
5279 	/* statistics */
5280 	keystat.getspi_count =
5281 	    (keystat.getspi_count + tries) / 2;
5282 
5283 	return (htonl(newspi));
5284 }
5285 
5286 /*
5287  * Find TCP-MD5 SA with corresponding secasindex.
5288  * If not found, return NULL and fill SPI with usable value if needed.
5289  */
5290 static struct secasvar *
5291 key_getsav_tcpmd5(struct secasindex *saidx, uint32_t *spi)
5292 {
5293 	SAHTREE_RLOCK_TRACKER;
5294 	struct secashead *sah;
5295 	struct secasvar *sav;
5296 
5297 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP, ("wrong proto"));
5298 	SAHTREE_RLOCK();
5299 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
5300 		if (sah->saidx.proto != IPPROTO_TCP)
5301 			continue;
5302 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
5303 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
5304 			break;
5305 	}
5306 	if (sah != NULL) {
5307 		if (V_key_preferred_oldsa)
5308 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
5309 		else
5310 			sav = TAILQ_FIRST(&sah->savtree_alive);
5311 		if (sav != NULL) {
5312 			SAV_ADDREF(sav);
5313 			SAHTREE_RUNLOCK();
5314 			return (sav);
5315 		}
5316 	}
5317 	if (spi == NULL) {
5318 		/* No SPI required */
5319 		SAHTREE_RUNLOCK();
5320 		return (NULL);
5321 	}
5322 	/* Check that SPI is unique */
5323 	LIST_FOREACH(sav, SAVHASH_HASH(*spi), spihash) {
5324 		if (sav->spi == *spi)
5325 			break;
5326 	}
5327 	if (sav == NULL) {
5328 		SAHTREE_RUNLOCK();
5329 		/* SPI is already unique */
5330 		return (NULL);
5331 	}
5332 	SAHTREE_RUNLOCK();
5333 	/* XXX: not optimal */
5334 	*spi = key_do_getnewspi(NULL, saidx);
5335 	return (NULL);
5336 }
5337 
5338 static int
5339 key_updateaddresses(struct socket *so, struct mbuf *m,
5340     const struct sadb_msghdr *mhp, struct secasvar *sav,
5341     struct secasindex *saidx)
5342 {
5343 	struct sockaddr *newaddr;
5344 	struct secashead *sah;
5345 	struct secasvar *newsav, *tmp;
5346 	struct mbuf *n;
5347 	int error, isnew;
5348 
5349 	/* Check that we need to change SAH */
5350 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC)) {
5351 		newaddr = (struct sockaddr *)(
5352 		    ((struct sadb_address *)
5353 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_SRC]) + 1);
5354 		bcopy(newaddr, &saidx->src, newaddr->sa_len);
5355 		key_porttosaddr(&saidx->src.sa, 0);
5356 	}
5357 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5358 		newaddr = (struct sockaddr *)(
5359 		    ((struct sadb_address *)
5360 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_DST]) + 1);
5361 		bcopy(newaddr, &saidx->dst, newaddr->sa_len);
5362 		key_porttosaddr(&saidx->dst.sa, 0);
5363 	}
5364 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5365 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5366 		error = key_checksockaddrs(&saidx->src.sa, &saidx->dst.sa);
5367 		if (error != 0) {
5368 			ipseclog((LOG_DEBUG, "%s: invalid new sockaddr.\n",
5369 			    __func__));
5370 			return (error);
5371 		}
5372 
5373 		sah = key_getsah(saidx);
5374 		if (sah == NULL) {
5375 			/* create a new SA index */
5376 			sah = key_newsah(saidx);
5377 			if (sah == NULL) {
5378 				ipseclog((LOG_DEBUG,
5379 				    "%s: No more memory.\n", __func__));
5380 				return (ENOBUFS);
5381 			}
5382 			isnew = 2; /* SAH is new */
5383 		} else
5384 			isnew = 1; /* existing SAH is referenced */
5385 	} else {
5386 		/*
5387 		 * src and dst addresses are still the same.
5388 		 * Do we want to change NAT-T config?
5389 		 */
5390 		if (sav->sah->saidx.proto != IPPROTO_ESP ||
5391 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5392 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5393 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5394 			ipseclog((LOG_DEBUG,
5395 			    "%s: invalid message: missing required header.\n",
5396 			    __func__));
5397 			return (EINVAL);
5398 		}
5399 		/* We hold reference to SA, thus SAH will be referenced too. */
5400 		sah = sav->sah;
5401 		isnew = 0;
5402 	}
5403 
5404 	newsav = malloc(sizeof(struct secasvar), M_IPSEC_SA,
5405 	    M_NOWAIT | M_ZERO);
5406 	if (newsav == NULL) {
5407 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5408 		error = ENOBUFS;
5409 		goto fail;
5410 	}
5411 
5412 	/* Clone SA's content into newsav */
5413 	SAV_INITREF(newsav);
5414 	bcopy(sav, newsav, offsetof(struct secasvar, chain));
5415 #ifdef IPSEC_OFFLOAD
5416 	CK_LIST_INIT(&newsav->accel_ifps);
5417 	newsav->accel_forget_tq = 0;
5418 	newsav->accel_lft_sw = uma_zalloc_pcpu(ipsec_key_lft_zone,
5419 	    M_NOWAIT | M_ZERO);
5420 	if (newsav->accel_lft_sw == NULL) {
5421 		error = ENOBUFS;
5422 		goto fail;
5423 	}
5424 	if (sav->accel_ifname != NULL) {
5425 		struct sadb_x_if_hw_offl xof;
5426 
5427 		newsav->accel_ifname = malloc(sizeof(xof.sadb_x_if_hw_offl_if),
5428 		    M_IPSEC_MISC, M_NOWAIT);
5429 		if (newsav->accel_ifname == NULL) {
5430 			error = ENOBUFS;
5431 			goto fail;
5432 		}
5433 		strncpy(__DECONST(char *, sav->accel_ifname),
5434 		    newsav->accel_ifname,
5435 		    sizeof(xof.sadb_x_if_hw_offl_if));
5436 	}
5437 #endif
5438 
5439 	/*
5440 	 * We create new NAT-T config if it is needed.
5441 	 * Old NAT-T config will be freed by key_cleansav() when
5442 	 * last reference to SA will be released.
5443 	 */
5444 	newsav->natt = NULL;
5445 	newsav->sah = sah;
5446 	newsav->state = SADB_SASTATE_MATURE;
5447 	error = key_setnatt(newsav, mhp);
5448 	if (error != 0)
5449 		goto fail;
5450 
5451 	SAHTREE_WLOCK();
5452 	/* Check that SA is still alive */
5453 	if (sav->state == SADB_SASTATE_DEAD) {
5454 		/* SA was unlinked */
5455 		SAHTREE_WUNLOCK();
5456 		error = ESRCH;
5457 		goto fail;
5458 	}
5459 
5460 	/* Unlink SA from SAH and SPI hash */
5461 	IPSEC_ASSERT((sav->flags & SADB_X_EXT_F_CLONED) == 0,
5462 	    ("SA is already cloned"));
5463 	IPSEC_ASSERT(sav->state == SADB_SASTATE_MATURE ||
5464 	    sav->state == SADB_SASTATE_DYING,
5465 	    ("Wrong SA state %u\n", sav->state));
5466 	TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
5467 	LIST_REMOVE(sav, spihash);
5468 	sav->state = SADB_SASTATE_DEAD;
5469 	ipsec_accel_forget_sav(sav);
5470 
5471 	/*
5472 	 * Link new SA with SAH. Keep SAs ordered by
5473 	 * create time (newer are first).
5474 	 */
5475 	TAILQ_FOREACH(tmp, &sah->savtree_alive, chain) {
5476 		if (newsav->created > tmp->created) {
5477 			TAILQ_INSERT_BEFORE(tmp, newsav, chain);
5478 			break;
5479 		}
5480 	}
5481 	if (tmp == NULL)
5482 		TAILQ_INSERT_TAIL(&sah->savtree_alive, newsav, chain);
5483 
5484 	/* Add new SA into SPI hash. */
5485 	LIST_INSERT_HEAD(SAVHASH_HASH(newsav->spi), newsav, spihash);
5486 
5487 	/* Add new SAH into SADB. */
5488 	if (isnew == 2) {
5489 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
5490 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
5491 		sah->state = SADB_SASTATE_MATURE;
5492 		SAH_ADDREF(sah); /* newsav references new SAH */
5493 	}
5494 	/*
5495 	 * isnew == 1 -> @sah was referenced by key_getsah().
5496 	 * isnew == 0 -> we use the same @sah, that was used by @sav,
5497 	 *	and we use its reference for @newsav.
5498 	 */
5499 	SECASVAR_WLOCK(sav);
5500 	/* XXX: replace cntr with pointer? */
5501 	newsav->cntr = sav->cntr;
5502 	sav->flags |= SADB_X_EXT_F_CLONED;
5503 	SECASVAR_WUNLOCK(sav);
5504 
5505 	SAHTREE_WUNLOCK();
5506 
5507 	KEYDBG(KEY_STAMP,
5508 	    printf("%s: SA(%p) cloned into SA(%p)\n",
5509 	    __func__, sav, newsav));
5510 	KEYDBG(KEY_DATA, kdebug_secasv(newsav));
5511 
5512 	key_freesav(&sav); /* release last reference */
5513 
5514 	/* set msg buf from mhp */
5515 	n = key_getmsgbuf_x1(m, mhp);
5516 	if (n == NULL) {
5517 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5518 		return (ENOBUFS);
5519 	}
5520 	m_freem(m);
5521 	key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5522 	return (0);
5523 fail:
5524 	if (isnew != 0)
5525 		key_freesah(&sah);
5526 	if (newsav != NULL) {
5527 #ifdef IPSEC_OFFLOAD
5528 		uma_zfree_pcpu(ipsec_key_lft_zone, newsav->accel_lft_sw);
5529 		free(__DECONST(char *, newsav->accel_ifname), M_IPSEC_MISC);
5530 #endif
5531 		if (newsav->natt != NULL)
5532 			free(newsav->natt, M_IPSEC_MISC);
5533 		free(newsav, M_IPSEC_SA);
5534 	}
5535 	return (error);
5536 }
5537 
5538 /*
5539  * SADB_UPDATE processing
5540  * receive
5541  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5542  *       key(AE), (identity(SD),) (sensitivity)>
5543  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5544  * and send
5545  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5546  *       (identity(SD),) (sensitivity)>
5547  * to the ikmpd.
5548  *
5549  * m will always be freed.
5550  */
5551 static int
5552 key_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5553 {
5554 	struct secasindex saidx;
5555 	struct sadb_address *src0, *dst0;
5556 	struct sadb_sa *sa0;
5557 	struct secasvar *sav;
5558 	uint32_t reqid;
5559 	int error;
5560 	uint8_t mode, proto;
5561 
5562 	IPSEC_ASSERT(so != NULL, ("null socket"));
5563 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5564 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5565 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5566 
5567 	/* map satype to proto */
5568 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5569 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5570 		    __func__));
5571 		return key_senderror(so, m, EINVAL);
5572 	}
5573 
5574 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5575 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5576 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5577 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5578 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5579 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5580 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5581 		ipseclog((LOG_DEBUG,
5582 		    "%s: invalid message: missing required header.\n",
5583 		    __func__));
5584 		return key_senderror(so, m, EINVAL);
5585 	}
5586 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5587 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5588 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5589 		ipseclog((LOG_DEBUG,
5590 		    "%s: invalid message: wrong header size.\n", __func__));
5591 		return key_senderror(so, m, EINVAL);
5592 	}
5593 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5594 		mode = IPSEC_MODE_ANY;
5595 		reqid = 0;
5596 	} else {
5597 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5598 			ipseclog((LOG_DEBUG,
5599 			    "%s: invalid message: wrong header size.\n",
5600 			    __func__));
5601 			return key_senderror(so, m, EINVAL);
5602 		}
5603 		mode = ((struct sadb_x_sa2 *)
5604 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5605 		reqid = ((struct sadb_x_sa2 *)
5606 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5607 	}
5608 
5609 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5610 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5611 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5612 
5613 	/*
5614 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5615 	 * SADB_UPDATE message.
5616 	 */
5617 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5618 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5619 #ifdef PFKEY_STRICT_CHECKS
5620 		return key_senderror(so, m, EINVAL);
5621 #endif
5622 	}
5623 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5624 	    (struct sockaddr *)(dst0 + 1));
5625 	if (error != 0) {
5626 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5627 		return key_senderror(so, m, error);
5628 	}
5629 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5630 	sav = key_getsavbyspi(sa0->sadb_sa_spi);
5631 	if (sav == NULL) {
5632 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u\n",
5633 		    __func__, ntohl(sa0->sadb_sa_spi)));
5634 		return key_senderror(so, m, EINVAL);
5635 	}
5636 	/*
5637 	 * Check that SADB_UPDATE issued by the same process that did
5638 	 * SADB_GETSPI or SADB_ADD.
5639 	 */
5640 	if (sav->pid != mhp->msg->sadb_msg_pid) {
5641 		ipseclog((LOG_DEBUG,
5642 		    "%s: pid mismatched (SPI %u, pid %u vs. %u)\n", __func__,
5643 		    ntohl(sav->spi), sav->pid, mhp->msg->sadb_msg_pid));
5644 		key_freesav(&sav);
5645 		return key_senderror(so, m, EINVAL);
5646 	}
5647 	/* saidx should match with SA. */
5648 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_MODE_REQID) == 0) {
5649 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u\n",
5650 		    __func__, ntohl(sav->spi)));
5651 		key_freesav(&sav);
5652 		return key_senderror(so, m, ESRCH);
5653 	}
5654 
5655 	if (sav->state == SADB_SASTATE_LARVAL) {
5656 		if ((mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5657 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) ||
5658 		    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5659 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH))) {
5660 			ipseclog((LOG_DEBUG,
5661 			    "%s: invalid message: missing required header.\n",
5662 			    __func__));
5663 			key_freesav(&sav);
5664 			return key_senderror(so, m, EINVAL);
5665 		}
5666 		/*
5667 		 * We can set any values except src, dst and SPI.
5668 		 */
5669 		error = key_setsaval(sav, mhp);
5670 		if (error != 0) {
5671 			key_freesav(&sav);
5672 			return (key_senderror(so, m, error));
5673 		}
5674 		/* Change SA state to MATURE */
5675 		SAHTREE_WLOCK();
5676 		if (sav->state != SADB_SASTATE_LARVAL) {
5677 			/* SA was deleted or another thread made it MATURE. */
5678 			SAHTREE_WUNLOCK();
5679 			key_freesav(&sav);
5680 			return (key_senderror(so, m, ESRCH));
5681 		}
5682 		/*
5683 		 * NOTE: we keep SAs in savtree_alive ordered by created
5684 		 * time. When SA's state changed from LARVAL to MATURE,
5685 		 * we update its created time in key_setsaval() and move
5686 		 * it into head of savtree_alive.
5687 		 */
5688 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
5689 		TAILQ_INSERT_HEAD(&sav->sah->savtree_alive, sav, chain);
5690 		sav->state = SADB_SASTATE_MATURE;
5691 		SAHTREE_WUNLOCK();
5692 	} else {
5693 		/*
5694 		 * For DYING and MATURE SA we can change only state
5695 		 * and lifetimes. Report EINVAL if something else attempted
5696 		 * to change.
5697 		 */
5698 		if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5699 		    !SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
5700 			key_freesav(&sav);
5701 			return (key_senderror(so, m, EINVAL));
5702 		}
5703 		error = key_updatelifetimes(sav, mhp);
5704 		if (error != 0) {
5705 			key_freesav(&sav);
5706 			return (key_senderror(so, m, error));
5707 		}
5708 		/*
5709 		 * This is FreeBSD extension to RFC2367.
5710 		 * IKEd can specify SADB_X_EXT_NEW_ADDRESS_SRC and/or
5711 		 * SADB_X_EXT_NEW_ADDRESS_DST when it wants to change
5712 		 * SA addresses (for example to implement MOBIKE protocol
5713 		 * as described in RFC4555). Also we allow to change
5714 		 * NAT-T config.
5715 		 */
5716 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5717 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST) ||
5718 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5719 		    sav->natt != NULL) {
5720 			error = key_updateaddresses(so, m, mhp, sav, &saidx);
5721 			key_freesav(&sav);
5722 			if (error != 0)
5723 				return (key_senderror(so, m, error));
5724 			return (0);
5725 		}
5726 		/* Check that SA is still alive */
5727 		SAHTREE_WLOCK();
5728 		if (sav->state == SADB_SASTATE_DEAD) {
5729 			/* SA was unlinked */
5730 			SAHTREE_WUNLOCK();
5731 			key_freesav(&sav);
5732 			return (key_senderror(so, m, ESRCH));
5733 		}
5734 		/*
5735 		 * NOTE: there is possible state moving from DYING to MATURE,
5736 		 * but this doesn't change created time, so we won't reorder
5737 		 * this SA.
5738 		 */
5739 		sav->state = SADB_SASTATE_MATURE;
5740 		SAHTREE_WUNLOCK();
5741 	}
5742 	KEYDBG(KEY_STAMP,
5743 	    printf("%s: SA(%p)\n", __func__, sav));
5744 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5745 	ipsec_accel_sa_newkey(sav);
5746 	key_freesav(&sav);
5747 
5748     {
5749 	struct mbuf *n;
5750 
5751 	/* set msg buf from mhp */
5752 	n = key_getmsgbuf_x1(m, mhp);
5753 	if (n == NULL) {
5754 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5755 		return key_senderror(so, m, ENOBUFS);
5756 	}
5757 
5758 	m_freem(m);
5759 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5760     }
5761 }
5762 
5763 /*
5764  * SADB_ADD processing
5765  * add an entry to SA database, when received
5766  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5767  *       key(AE), (identity(SD),) (sensitivity)>
5768  * from the ikmpd,
5769  * and send
5770  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5771  *       (identity(SD),) (sensitivity)>
5772  * to the ikmpd.
5773  *
5774  * IGNORE identity and sensitivity messages.
5775  *
5776  * m will always be freed.
5777  */
5778 static int
5779 key_add(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5780 {
5781 	struct secasindex saidx;
5782 	struct sadb_address *src0, *dst0;
5783 	struct sadb_sa *sa0;
5784 	struct secasvar *sav;
5785 	uint32_t reqid, spi;
5786 	uint8_t mode, proto;
5787 	int error;
5788 
5789 	IPSEC_ASSERT(so != NULL, ("null socket"));
5790 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5791 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5792 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5793 
5794 	/* map satype to proto */
5795 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5796 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5797 		    __func__));
5798 		return key_senderror(so, m, EINVAL);
5799 	}
5800 
5801 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5802 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5803 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5804 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP && (
5805 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5806 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT))) ||
5807 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH && (
5808 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH) ||
5809 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH))) ||
5810 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5811 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5812 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5813 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5814 		ipseclog((LOG_DEBUG,
5815 		    "%s: invalid message: missing required header.\n",
5816 		    __func__));
5817 		return key_senderror(so, m, EINVAL);
5818 	}
5819 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5820 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5821 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5822 		ipseclog((LOG_DEBUG,
5823 		    "%s: invalid message: wrong header size.\n", __func__));
5824 		return key_senderror(so, m, EINVAL);
5825 	}
5826 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5827 		mode = IPSEC_MODE_ANY;
5828 		reqid = 0;
5829 	} else {
5830 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5831 			ipseclog((LOG_DEBUG,
5832 			    "%s: invalid message: wrong header size.\n",
5833 			    __func__));
5834 			return key_senderror(so, m, EINVAL);
5835 		}
5836 		mode = ((struct sadb_x_sa2 *)
5837 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5838 		reqid = ((struct sadb_x_sa2 *)
5839 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5840 	}
5841 
5842 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5843 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
5844 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
5845 
5846 	/*
5847 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5848 	 * SADB_ADD message.
5849 	 */
5850 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5851 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5852 #ifdef PFKEY_STRICT_CHECKS
5853 		return key_senderror(so, m, EINVAL);
5854 #endif
5855 	}
5856 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5857 	    (struct sockaddr *)(dst0 + 1));
5858 	if (error != 0) {
5859 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5860 		return key_senderror(so, m, error);
5861 	}
5862 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5863 	spi = sa0->sadb_sa_spi;
5864 	/*
5865 	 * For TCP-MD5 SAs we don't use SPI. Check the uniqueness using
5866 	 * secasindex.
5867 	 * XXXAE: IPComp seems also doesn't use SPI.
5868 	 */
5869 	SPI_ALLOC_LOCK();
5870 	if (proto == IPPROTO_TCP) {
5871 		sav = key_getsav_tcpmd5(&saidx, &spi);
5872 		if (sav == NULL && spi == 0) {
5873 			SPI_ALLOC_UNLOCK();
5874 			/* Failed to allocate SPI */
5875 			ipseclog((LOG_DEBUG, "%s: SA already exists.\n",
5876 			    __func__));
5877 			return key_senderror(so, m, EEXIST);
5878 		}
5879 		/* XXX: SPI that we report back can have another value */
5880 	} else {
5881 		/* We can create new SA only if SPI is different. */
5882 		sav = key_getsavbyspi(spi);
5883 	}
5884 	if (sav != NULL) {
5885 		SPI_ALLOC_UNLOCK();
5886 		key_freesav(&sav);
5887 		ipseclog((LOG_DEBUG, "%s: SA already exists.\n", __func__));
5888 		return key_senderror(so, m, EEXIST);
5889 	}
5890 
5891 	sav = key_newsav(mhp, &saidx, spi, &error);
5892 	SPI_ALLOC_UNLOCK();
5893 	if (sav == NULL)
5894 		return key_senderror(so, m, error);
5895 	KEYDBG(KEY_STAMP,
5896 	    printf("%s: return SA(%p)\n", __func__, sav));
5897 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5898 	ipsec_accel_sa_newkey(sav);
5899 	/*
5900 	 * If SADB_ADD was in response to SADB_ACQUIRE, we need to schedule
5901 	 * ACQ for deletion.
5902 	 */
5903 	if (sav->seq != 0)
5904 		key_acqdone(&saidx, sav->seq);
5905 
5906     {
5907 	/*
5908 	 * Don't call key_freesav() on error here, as we would like to
5909 	 * keep the SA in the database.
5910 	 */
5911 	struct mbuf *n;
5912 
5913 	/* set msg buf from mhp */
5914 	n = key_getmsgbuf_x1(m, mhp);
5915 	if (n == NULL) {
5916 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5917 		return key_senderror(so, m, ENOBUFS);
5918 	}
5919 
5920 	m_freem(m);
5921 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5922     }
5923 }
5924 
5925 /*
5926  * NAT-T support.
5927  * IKEd may request the use ESP in UDP encapsulation when it detects the
5928  * presence of NAT. It uses NAT-T extension headers for such SAs to specify
5929  * parameters needed for encapsulation and decapsulation. These PF_KEY
5930  * extension headers are not standardized, so this comment addresses our
5931  * implementation.
5932  * SADB_X_EXT_NAT_T_TYPE specifies type of encapsulation, we support only
5933  * UDP_ENCAP_ESPINUDP as described in RFC3948.
5934  * SADB_X_EXT_NAT_T_SPORT/DPORT specifies source and destination ports for
5935  * UDP header. We use these ports in UDP encapsulation procedure, also we
5936  * can check them in UDP decapsulation procedure.
5937  * SADB_X_EXT_NAT_T_OA[IR] specifies original address of initiator or
5938  * responder. These addresses can be used for transport mode to adjust
5939  * checksum after decapsulation and decryption. Since original IP addresses
5940  * used by peer usually different (we detected presence of NAT), TCP/UDP
5941  * pseudo header checksum and IP header checksum was calculated using original
5942  * addresses. After decapsulation and decryption we need to adjust checksum
5943  * to have correct datagram.
5944  *
5945  * We expect presence of NAT-T extension headers only in SADB_ADD and
5946  * SADB_UPDATE messages. We report NAT-T extension headers in replies
5947  * to SADB_ADD, SADB_UPDATE, SADB_GET, and SADB_DUMP messages.
5948  */
5949 static int
5950 key_setnatt(struct secasvar *sav, const struct sadb_msghdr *mhp)
5951 {
5952 	struct sadb_x_nat_t_port *port;
5953 	struct sadb_x_nat_t_type *type;
5954 	struct sadb_address *oai, *oar;
5955 	struct sockaddr *sa;
5956 	uint32_t addr;
5957 	uint16_t cksum;
5958 	int i;
5959 
5960 	IPSEC_ASSERT(sav->natt == NULL, ("natt is already initialized"));
5961 	/*
5962 	 * Ignore NAT-T headers if sproto isn't ESP.
5963 	 */
5964 	if (sav->sah->saidx.proto != IPPROTO_ESP)
5965 		return (0);
5966 
5967 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) &&
5968 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) &&
5969 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5970 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5971 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5972 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5973 			ipseclog((LOG_DEBUG,
5974 			    "%s: invalid message: wrong header size.\n",
5975 			    __func__));
5976 			return (EINVAL);
5977 		}
5978 	} else
5979 		return (0);
5980 
5981 	type = (struct sadb_x_nat_t_type *)mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5982 	if (type->sadb_x_nat_t_type_type != UDP_ENCAP_ESPINUDP) {
5983 		ipseclog((LOG_DEBUG, "%s: unsupported NAT-T type %u.\n",
5984 		    __func__, type->sadb_x_nat_t_type_type));
5985 		return (EINVAL);
5986 	}
5987 	/*
5988 	 * Allocate storage for NAT-T config.
5989 	 * On error it will be released by key_cleansav().
5990 	 */
5991 	sav->natt = malloc(sizeof(struct secnatt), M_IPSEC_MISC,
5992 	    M_NOWAIT | M_ZERO);
5993 	if (sav->natt == NULL) {
5994 		PFKEYSTAT_INC(in_nomem);
5995 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5996 		return (ENOBUFS);
5997 	}
5998 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5999 	if (port->sadb_x_nat_t_port_port == 0) {
6000 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T sport specified.\n",
6001 		    __func__));
6002 		return (EINVAL);
6003 	}
6004 	sav->natt->sport = port->sadb_x_nat_t_port_port;
6005 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_DPORT];
6006 	if (port->sadb_x_nat_t_port_port == 0) {
6007 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T dport specified.\n",
6008 		    __func__));
6009 		return (EINVAL);
6010 	}
6011 	sav->natt->dport = port->sadb_x_nat_t_port_port;
6012 
6013 	/*
6014 	 * SADB_X_EXT_NAT_T_OAI and SADB_X_EXT_NAT_T_OAR are optional
6015 	 * and needed only for transport mode IPsec.
6016 	 * Usually NAT translates only one address, but it is possible,
6017 	 * that both addresses could be translated.
6018 	 * NOTE: Value of SADB_X_EXT_NAT_T_OAI is equal to SADB_X_EXT_NAT_T_OA.
6019 	 */
6020 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAI)) {
6021 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAI)) {
6022 			ipseclog((LOG_DEBUG,
6023 			    "%s: invalid message: wrong header size.\n",
6024 			    __func__));
6025 			return (EINVAL);
6026 		}
6027 		oai = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAI];
6028 	} else
6029 		oai = NULL;
6030 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAR)) {
6031 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAR)) {
6032 			ipseclog((LOG_DEBUG,
6033 			    "%s: invalid message: wrong header size.\n",
6034 			    __func__));
6035 			return (EINVAL);
6036 		}
6037 		oar = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAR];
6038 	} else
6039 		oar = NULL;
6040 
6041 	/* Initialize addresses only for transport mode */
6042 	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL) {
6043 		cksum = 0;
6044 		if (oai != NULL) {
6045 			sa = (struct sockaddr *)(oai + 1);
6046 			switch (sa->sa_family) {
6047 #ifdef AF_INET
6048 			case AF_INET:
6049 				if (sa->sa_len != sizeof(struct sockaddr_in)) {
6050 					ipseclog((LOG_DEBUG,
6051 					    "%s: wrong NAT-OAi header.\n",
6052 					    __func__));
6053 					return (EINVAL);
6054 				}
6055 				/* Ignore address if it the same */
6056 				if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
6057 				    sav->sah->saidx.src.sin.sin_addr.s_addr) {
6058 					bcopy(sa, &sav->natt->oai.sa, sa->sa_len);
6059 					sav->natt->flags |= IPSEC_NATT_F_OAI;
6060 					/* Calculate checksum delta */
6061 					addr = sav->sah->saidx.src.sin.sin_addr.s_addr;
6062 					cksum = in_addword(cksum, ~addr >> 16);
6063 					cksum = in_addword(cksum, ~addr & 0xffff);
6064 					addr = sav->natt->oai.sin.sin_addr.s_addr;
6065 					cksum = in_addword(cksum, addr >> 16);
6066 					cksum = in_addword(cksum, addr & 0xffff);
6067 				}
6068 				break;
6069 #endif
6070 #ifdef AF_INET6
6071 			case AF_INET6:
6072 				if (sa->sa_len != sizeof(struct sockaddr_in6)) {
6073 					ipseclog((LOG_DEBUG,
6074 					    "%s: wrong NAT-OAi header.\n",
6075 					    __func__));
6076 					return (EINVAL);
6077 				}
6078 				/* Ignore address if it the same */
6079 				if (memcmp(&((struct sockaddr_in6 *)sa)->sin6_addr.s6_addr,
6080 				    &sav->sah->saidx.src.sin6.sin6_addr.s6_addr,
6081 				    sizeof(struct in6_addr)) != 0) {
6082 					bcopy(sa, &sav->natt->oai.sa, sa->sa_len);
6083 					sav->natt->flags |= IPSEC_NATT_F_OAI;
6084 					/* Calculate checksum delta */
6085 					for (i = 0; i < 8; i++) {
6086 						cksum = in_addword(cksum,
6087 						  ~sav->sah->saidx.src.sin6.sin6_addr.s6_addr16[i]);
6088 						cksum = in_addword(cksum,
6089 						   sav->natt->oai.sin6.sin6_addr.s6_addr16[i]);
6090 					}
6091 				}
6092 				break;
6093 #endif
6094 			default:
6095 				ipseclog((LOG_DEBUG,
6096 				    "%s: wrong NAT-OAi header.\n",
6097 				    __func__));
6098 				return (EINVAL);
6099 			}
6100 		}
6101 		if (oar != NULL) {
6102 			sa = (struct sockaddr *)(oar + 1);
6103 			switch (sa->sa_family) {
6104 #ifdef AF_INET
6105 			case AF_INET:
6106 				if (sa->sa_len != sizeof(struct sockaddr_in)) {
6107 					ipseclog((LOG_DEBUG,
6108 					    "%s: wrong NAT-OAr header.\n",
6109 					    __func__));
6110 					return (EINVAL);
6111 				}
6112 				/* Ignore address if it the same */
6113 				if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
6114 				    sav->sah->saidx.dst.sin.sin_addr.s_addr) {
6115 					bcopy(sa, &sav->natt->oar.sa, sa->sa_len);
6116 					sav->natt->flags |= IPSEC_NATT_F_OAR;
6117 					/* Calculate checksum delta */
6118 					addr = sav->sah->saidx.dst.sin.sin_addr.s_addr;
6119 					cksum = in_addword(cksum, ~addr >> 16);
6120 					cksum = in_addword(cksum, ~addr & 0xffff);
6121 					addr = sav->natt->oar.sin.sin_addr.s_addr;
6122 					cksum = in_addword(cksum, addr >> 16);
6123 					cksum = in_addword(cksum, addr & 0xffff);
6124 				}
6125 				break;
6126 #endif
6127 #ifdef AF_INET6
6128 			case AF_INET6:
6129 				if (sa->sa_len != sizeof(struct sockaddr_in6)) {
6130 					ipseclog((LOG_DEBUG,
6131 					    "%s: wrong NAT-OAr header.\n",
6132 					    __func__));
6133 					return (EINVAL);
6134 				}
6135 				/* Ignore address if it the same */
6136 				if (memcmp(&((struct sockaddr_in6 *)sa)->sin6_addr.s6_addr,
6137 				           &sav->sah->saidx.dst.sin6.sin6_addr.s6_addr, 16) != 0) {
6138 					bcopy(sa, &sav->natt->oar.sa, sa->sa_len);
6139 					sav->natt->flags |= IPSEC_NATT_F_OAR;
6140 					/* Calculate checksum delta */
6141 					for (i = 0; i < 8; i++) {
6142 						cksum = in_addword(cksum,
6143 						   ~sav->sah->saidx.dst.sin6.sin6_addr.s6_addr16[i]);
6144 						cksum = in_addword(cksum,
6145 						   sav->natt->oar.sin6.sin6_addr.s6_addr16[i]);
6146 					}
6147 				}
6148 				break;
6149 #endif
6150 			default:
6151 				ipseclog((LOG_DEBUG,
6152 				    "%s: wrong NAT-OAr header.\n",
6153 				    __func__));
6154 				return (EINVAL);
6155 			}
6156 		}
6157 		sav->natt->cksum = cksum;
6158 	}
6159 	return (0);
6160 }
6161 
6162 static int
6163 key_setident(struct secashead *sah, const struct sadb_msghdr *mhp)
6164 {
6165 	const struct sadb_ident *idsrc, *iddst;
6166 
6167 	IPSEC_ASSERT(sah != NULL, ("null secashead"));
6168 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6169 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6170 
6171 	/* don't make buffer if not there */
6172 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) &&
6173 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
6174 		sah->idents = NULL;
6175 		sah->identd = NULL;
6176 		return (0);
6177 	}
6178 
6179 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) ||
6180 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
6181 		ipseclog((LOG_DEBUG, "%s: invalid identity.\n", __func__));
6182 		return (EINVAL);
6183 	}
6184 
6185 	idsrc = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_SRC];
6186 	iddst = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_DST];
6187 
6188 	/* validity check */
6189 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
6190 		ipseclog((LOG_DEBUG, "%s: ident type mismatch.\n", __func__));
6191 		return EINVAL;
6192 	}
6193 
6194 	switch (idsrc->sadb_ident_type) {
6195 	case SADB_IDENTTYPE_PREFIX:
6196 	case SADB_IDENTTYPE_FQDN:
6197 	case SADB_IDENTTYPE_USERFQDN:
6198 	default:
6199 		/* XXX do nothing */
6200 		sah->idents = NULL;
6201 		sah->identd = NULL;
6202 	 	return 0;
6203 	}
6204 
6205 	/* make structure */
6206 	sah->idents = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
6207 	if (sah->idents == NULL) {
6208 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6209 		return ENOBUFS;
6210 	}
6211 	sah->identd = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
6212 	if (sah->identd == NULL) {
6213 		free(sah->idents, M_IPSEC_MISC);
6214 		sah->idents = NULL;
6215 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6216 		return ENOBUFS;
6217 	}
6218 	sah->idents->type = idsrc->sadb_ident_type;
6219 	sah->idents->id = idsrc->sadb_ident_id;
6220 
6221 	sah->identd->type = iddst->sadb_ident_type;
6222 	sah->identd->id = iddst->sadb_ident_id;
6223 
6224 	return 0;
6225 }
6226 
6227 /*
6228  * m will not be freed on return.
6229  * it is caller's responsibility to free the result.
6230  *
6231  * Called from SADB_ADD and SADB_UPDATE. Reply will contain headers
6232  * from the request in defined order.
6233  */
6234 static struct mbuf *
6235 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
6236 {
6237 	struct mbuf *n;
6238 
6239 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6240 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6241 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6242 
6243 	/* create new sadb_msg to reply. */
6244 	n = key_gather_mbuf(m, mhp, 1, 16, SADB_EXT_RESERVED,
6245 	    SADB_EXT_SA, SADB_X_EXT_SA2,
6246 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
6247 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
6248 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
6249 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
6250 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
6251 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NEW_ADDRESS_SRC,
6252 	    SADB_X_EXT_NEW_ADDRESS_DST);
6253 	if (!n)
6254 		return NULL;
6255 
6256 	if (n->m_len < sizeof(struct sadb_msg)) {
6257 		n = m_pullup(n, sizeof(struct sadb_msg));
6258 		if (n == NULL)
6259 			return NULL;
6260 	}
6261 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
6262 	mtod(n, struct sadb_msg *)->sadb_msg_len =
6263 	    PFKEY_UNIT64(n->m_pkthdr.len);
6264 
6265 	return n;
6266 }
6267 
6268 /*
6269  * SADB_DELETE processing
6270  * receive
6271  *   <base, SA(*), address(SD)>
6272  * from the ikmpd, and set SADB_SASTATE_DEAD,
6273  * and send,
6274  *   <base, SA(*), address(SD)>
6275  * to the ikmpd.
6276  *
6277  * m will always be freed.
6278  */
6279 static int
6280 key_delete(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6281 {
6282 	struct secasindex saidx;
6283 	struct sadb_address *src0, *dst0;
6284 	struct secasvar *sav;
6285 	struct sadb_sa *sa0;
6286 	uint8_t proto;
6287 
6288 	IPSEC_ASSERT(so != NULL, ("null socket"));
6289 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6290 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6291 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6292 
6293 	/* map satype to proto */
6294 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6295 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6296 		    __func__));
6297 		return key_senderror(so, m, EINVAL);
6298 	}
6299 
6300 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6301 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
6302 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6303 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6304 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
6305 		    __func__));
6306 		return key_senderror(so, m, EINVAL);
6307 	}
6308 
6309 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
6310 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
6311 
6312 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6313 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6314 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6315 		return (key_senderror(so, m, EINVAL));
6316 	}
6317 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6318 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
6319 		/*
6320 		 * Caller wants us to delete all non-LARVAL SAs
6321 		 * that match the src/dst.  This is used during
6322 		 * IKE INITIAL-CONTACT.
6323 		 * XXXAE: this looks like some extension to RFC2367.
6324 		 */
6325 		ipseclog((LOG_DEBUG, "%s: doing delete all.\n", __func__));
6326 		return (key_delete_all(so, m, mhp, &saidx));
6327 	}
6328 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
6329 		ipseclog((LOG_DEBUG,
6330 		    "%s: invalid message: wrong header size.\n", __func__));
6331 		return (key_senderror(so, m, EINVAL));
6332 	}
6333 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6334 	SPI_ALLOC_LOCK();
6335 	if (proto == IPPROTO_TCP)
6336 		sav = key_getsav_tcpmd5(&saidx, NULL);
6337 	else
6338 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6339 	SPI_ALLOC_UNLOCK();
6340 	if (sav == NULL) {
6341 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u.\n",
6342 		    __func__, ntohl(sa0->sadb_sa_spi)));
6343 		return (key_senderror(so, m, ESRCH));
6344 	}
6345 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6346 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6347 		    __func__, ntohl(sav->spi)));
6348 		key_freesav(&sav);
6349 		return (key_senderror(so, m, ESRCH));
6350 	}
6351 	KEYDBG(KEY_STAMP,
6352 	    printf("%s: SA(%p)\n", __func__, sav));
6353 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
6354 	key_unlinksav(sav);
6355 	key_freesav(&sav);
6356 
6357     {
6358 	struct mbuf *n;
6359 	struct sadb_msg *newmsg;
6360 
6361 	/* create new sadb_msg to reply. */
6362 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
6363 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6364 	if (!n)
6365 		return key_senderror(so, m, ENOBUFS);
6366 
6367 	if (n->m_len < sizeof(struct sadb_msg)) {
6368 		n = m_pullup(n, sizeof(struct sadb_msg));
6369 		if (n == NULL)
6370 			return key_senderror(so, m, ENOBUFS);
6371 	}
6372 	newmsg = mtod(n, struct sadb_msg *);
6373 	newmsg->sadb_msg_errno = 0;
6374 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6375 
6376 	m_freem(m);
6377 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6378     }
6379 }
6380 
6381 /*
6382  * delete all SAs for src/dst.  Called from key_delete().
6383  */
6384 static int
6385 key_delete_all(struct socket *so, struct mbuf *m,
6386     const struct sadb_msghdr *mhp, struct secasindex *saidx)
6387 {
6388 	struct secasvar_queue drainq;
6389 	struct secashead *sah;
6390 	struct secasvar *sav, *nextsav;
6391 
6392 	TAILQ_INIT(&drainq);
6393 	SAHTREE_WLOCK();
6394 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
6395 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_HEAD) == 0)
6396 			continue;
6397 		/* Move all ALIVE SAs into drainq */
6398 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6399 	}
6400 	/* Unlink all queued SAs from SPI hash */
6401 	TAILQ_FOREACH(sav, &drainq, chain) {
6402 		sav->state = SADB_SASTATE_DEAD;
6403 		ipsec_accel_forget_sav(sav);
6404 		LIST_REMOVE(sav, spihash);
6405 	}
6406 	SAHTREE_WUNLOCK();
6407 	/* Now we can release reference for all SAs in drainq */
6408 	sav = TAILQ_FIRST(&drainq);
6409 	while (sav != NULL) {
6410 		KEYDBG(KEY_STAMP,
6411 		    printf("%s: SA(%p)\n", __func__, sav));
6412 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6413 		nextsav = TAILQ_NEXT(sav, chain);
6414 		key_freesah(&sav->sah); /* release reference from SAV */
6415 		key_freesav(&sav); /* release last reference */
6416 		sav = nextsav;
6417 	}
6418 
6419     {
6420 	struct mbuf *n;
6421 	struct sadb_msg *newmsg;
6422 
6423 	/* create new sadb_msg to reply. */
6424 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
6425 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6426 	if (!n)
6427 		return key_senderror(so, m, ENOBUFS);
6428 
6429 	if (n->m_len < sizeof(struct sadb_msg)) {
6430 		n = m_pullup(n, sizeof(struct sadb_msg));
6431 		if (n == NULL)
6432 			return key_senderror(so, m, ENOBUFS);
6433 	}
6434 	newmsg = mtod(n, struct sadb_msg *);
6435 	newmsg->sadb_msg_errno = 0;
6436 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6437 
6438 	m_freem(m);
6439 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6440     }
6441 }
6442 
6443 /*
6444  * Delete all alive SAs for corresponding xform.
6445  * Larval SAs have not initialized tdb_xform, so it is safe to leave them
6446  * here when xform disappears.
6447  */
6448 void
6449 key_delete_xform(const struct xformsw *xsp)
6450 {
6451 	struct secasvar_queue drainq;
6452 	struct secashead *sah;
6453 	struct secasvar *sav, *nextsav;
6454 
6455 	TAILQ_INIT(&drainq);
6456 	SAHTREE_WLOCK();
6457 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
6458 		sav = TAILQ_FIRST(&sah->savtree_alive);
6459 		if (sav == NULL)
6460 			continue;
6461 		if (sav->tdb_xform != xsp)
6462 			continue;
6463 		/*
6464 		 * It is supposed that all SAs in the chain are related to
6465 		 * one xform.
6466 		 */
6467 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6468 	}
6469 	/* Unlink all queued SAs from SPI hash */
6470 	TAILQ_FOREACH(sav, &drainq, chain) {
6471 		sav->state = SADB_SASTATE_DEAD;
6472 		ipsec_accel_forget_sav(sav);
6473 		LIST_REMOVE(sav, spihash);
6474 	}
6475 	SAHTREE_WUNLOCK();
6476 
6477 	/* Now we can release reference for all SAs in drainq */
6478 	sav = TAILQ_FIRST(&drainq);
6479 	while (sav != NULL) {
6480 		KEYDBG(KEY_STAMP,
6481 		    printf("%s: SA(%p)\n", __func__, sav));
6482 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6483 		nextsav = TAILQ_NEXT(sav, chain);
6484 		key_freesah(&sav->sah); /* release reference from SAV */
6485 		key_freesav(&sav); /* release last reference */
6486 		sav = nextsav;
6487 	}
6488 }
6489 
6490 /*
6491  * SADB_GET processing
6492  * receive
6493  *   <base, SA(*), address(SD)>
6494  * from the ikmpd, and get a SP and a SA to respond,
6495  * and send,
6496  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6497  *       (identity(SD),) (sensitivity)>
6498  * to the ikmpd.
6499  *
6500  * m will always be freed.
6501  */
6502 static int
6503 key_get(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6504 {
6505 	struct secasindex saidx;
6506 	struct sadb_address *src0, *dst0;
6507 	struct sadb_sa *sa0;
6508 	struct secasvar *sav;
6509 	uint8_t proto;
6510 
6511 	IPSEC_ASSERT(so != NULL, ("null socket"));
6512 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6513 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6514 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6515 
6516 	/* map satype to proto */
6517 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6518 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6519 			__func__));
6520 		return key_senderror(so, m, EINVAL);
6521 	}
6522 
6523 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
6524 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6525 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)) {
6526 		ipseclog((LOG_DEBUG,
6527 		    "%s: invalid message: missing required header.\n",
6528 		    __func__));
6529 		return key_senderror(so, m, EINVAL);
6530 	}
6531 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
6532 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6533 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6534 		ipseclog((LOG_DEBUG,
6535 		    "%s: invalid message: wrong header size.\n", __func__));
6536 		return key_senderror(so, m, EINVAL);
6537 	}
6538 
6539 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6540 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
6541 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
6542 
6543 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6544 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6545 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6546 		return key_senderror(so, m, EINVAL);
6547 	}
6548 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6549 
6550 	SPI_ALLOC_LOCK();
6551 	if (proto == IPPROTO_TCP)
6552 		sav = key_getsav_tcpmd5(&saidx, NULL);
6553 	else
6554 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6555 	SPI_ALLOC_UNLOCK();
6556 	if (sav == NULL) {
6557 		ipseclog((LOG_DEBUG, "%s: no SA found.\n", __func__));
6558 		return key_senderror(so, m, ESRCH);
6559 	}
6560 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6561 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6562 		    __func__, ntohl(sa0->sadb_sa_spi)));
6563 		key_freesav(&sav);
6564 		return (key_senderror(so, m, ESRCH));
6565 	}
6566 
6567     {
6568 	struct mbuf *n;
6569 	uint8_t satype;
6570 
6571 	/* map proto to satype */
6572 	if ((satype = key_proto2satype(sav->sah->saidx.proto)) == 0) {
6573 		ipseclog((LOG_DEBUG, "%s: there was invalid proto in SAD.\n",
6574 		    __func__));
6575 		key_freesav(&sav);
6576 		return key_senderror(so, m, EINVAL);
6577 	}
6578 
6579 	/* create new sadb_msg to reply. */
6580 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
6581 	    mhp->msg->sadb_msg_pid, NULL);
6582 
6583 	key_freesav(&sav);
6584 	if (!n)
6585 		return key_senderror(so, m, ENOBUFS);
6586 
6587 	m_freem(m);
6588 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
6589     }
6590 }
6591 
6592 /* XXX make it sysctl-configurable? */
6593 static void
6594 key_getcomb_setlifetime(struct sadb_comb *comb)
6595 {
6596 
6597 	comb->sadb_comb_soft_allocations = 1;
6598 	comb->sadb_comb_hard_allocations = 1;
6599 	comb->sadb_comb_soft_bytes = 0;
6600 	comb->sadb_comb_hard_bytes = 0;
6601 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
6602 	comb->sadb_comb_soft_addtime = comb->sadb_comb_soft_addtime * 80 / 100;
6603 	comb->sadb_comb_soft_usetime = 28800;	/* 8 hours */
6604 	comb->sadb_comb_hard_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
6605 }
6606 
6607 /*
6608  * XXX reorder combinations by preference
6609  * XXX no idea if the user wants ESP authentication or not
6610  */
6611 static struct mbuf *
6612 key_getcomb_ealg(void)
6613 {
6614 	struct sadb_comb *comb;
6615 	const struct enc_xform *algo;
6616 	struct mbuf *result = NULL, *m, *n;
6617 	int encmin;
6618 	int i, off, o;
6619 	int totlen;
6620 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6621 
6622 	m = NULL;
6623 	for (i = 1; i <= SADB_EALG_MAX; i++) {
6624 		algo = enc_algorithm_lookup(i);
6625 		if (algo == NULL)
6626 			continue;
6627 
6628 		/* discard algorithms with key size smaller than system min */
6629 		if (_BITS(algo->maxkey) < V_ipsec_esp_keymin)
6630 			continue;
6631 		if (_BITS(algo->minkey) < V_ipsec_esp_keymin)
6632 			encmin = V_ipsec_esp_keymin;
6633 		else
6634 			encmin = _BITS(algo->minkey);
6635 
6636 		if (V_ipsec_esp_auth)
6637 			m = key_getcomb_ah();
6638 		else {
6639 			IPSEC_ASSERT(l <= MLEN,
6640 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6641 			MGET(m, M_NOWAIT, MT_DATA);
6642 			if (m) {
6643 				M_ALIGN(m, l);
6644 				m->m_len = l;
6645 				m->m_next = NULL;
6646 				bzero(mtod(m, caddr_t), m->m_len);
6647 			}
6648 		}
6649 		if (!m)
6650 			goto fail;
6651 
6652 		totlen = 0;
6653 		for (n = m; n; n = n->m_next)
6654 			totlen += n->m_len;
6655 		IPSEC_ASSERT((totlen % l) == 0, ("totlen=%u, l=%u", totlen, l));
6656 
6657 		for (off = 0; off < totlen; off += l) {
6658 			n = m_pulldown(m, off, l, &o);
6659 			if (!n) {
6660 				/* m is already freed */
6661 				goto fail;
6662 			}
6663 			comb = (struct sadb_comb *)(mtod(n, caddr_t) + o);
6664 			bzero(comb, sizeof(*comb));
6665 			key_getcomb_setlifetime(comb);
6666 			comb->sadb_comb_encrypt = i;
6667 			comb->sadb_comb_encrypt_minbits = encmin;
6668 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
6669 		}
6670 
6671 		if (!result)
6672 			result = m;
6673 		else
6674 			m_cat(result, m);
6675 	}
6676 
6677 	return result;
6678 
6679  fail:
6680 	if (result)
6681 		m_freem(result);
6682 	return NULL;
6683 }
6684 
6685 static void
6686 key_getsizes_ah(const struct auth_hash *ah, int alg, u_int16_t* min,
6687     u_int16_t* max)
6688 {
6689 
6690 	*min = *max = ah->hashsize;
6691 	if (ah->keysize == 0) {
6692 		/*
6693 		 * Transform takes arbitrary key size but algorithm
6694 		 * key size is restricted.  Enforce this here.
6695 		 */
6696 		switch (alg) {
6697 		case SADB_X_AALG_NULL:	*min = 1; *max = 256; break;
6698 		case SADB_X_AALG_SHA2_256: *min = *max = 32; break;
6699 		case SADB_X_AALG_SHA2_384: *min = *max = 48; break;
6700 		case SADB_X_AALG_SHA2_512: *min = *max = 64; break;
6701 		default:
6702 			DPRINTF(("%s: unknown AH algorithm %u\n",
6703 				__func__, alg));
6704 			break;
6705 		}
6706 	}
6707 }
6708 
6709 /*
6710  * XXX reorder combinations by preference
6711  */
6712 static struct mbuf *
6713 key_getcomb_ah(void)
6714 {
6715 	const struct auth_hash *algo;
6716 	struct sadb_comb *comb;
6717 	struct mbuf *m;
6718 	u_int16_t minkeysize, maxkeysize;
6719 	int i;
6720 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6721 
6722 	m = NULL;
6723 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6724 #if 1
6725 		/* we prefer HMAC algorithms, not old algorithms */
6726 		if (i != SADB_AALG_SHA1HMAC &&
6727 		    i != SADB_X_AALG_SHA2_256 &&
6728 		    i != SADB_X_AALG_SHA2_384 &&
6729 		    i != SADB_X_AALG_SHA2_512)
6730 			continue;
6731 #endif
6732 		algo = auth_algorithm_lookup(i);
6733 		if (!algo)
6734 			continue;
6735 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6736 		/* discard algorithms with key size smaller than system min */
6737 		if (_BITS(minkeysize) < V_ipsec_ah_keymin)
6738 			continue;
6739 
6740 		if (!m) {
6741 			IPSEC_ASSERT(l <= MLEN,
6742 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6743 			MGET(m, M_NOWAIT, MT_DATA);
6744 			if (m) {
6745 				M_ALIGN(m, l);
6746 				m->m_len = l;
6747 				m->m_next = NULL;
6748 			}
6749 		} else
6750 			M_PREPEND(m, l, M_NOWAIT);
6751 		if (!m)
6752 			return NULL;
6753 
6754 		comb = mtod(m, struct sadb_comb *);
6755 		bzero(comb, sizeof(*comb));
6756 		key_getcomb_setlifetime(comb);
6757 		comb->sadb_comb_auth = i;
6758 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6759 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6760 	}
6761 
6762 	return m;
6763 }
6764 
6765 /*
6766  * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6767  * XXX reorder combinations by preference
6768  */
6769 static struct mbuf *
6770 key_getcomb_ipcomp(void)
6771 {
6772 	const struct comp_algo *algo;
6773 	struct sadb_comb *comb;
6774 	struct mbuf *m;
6775 	int i;
6776 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6777 
6778 	m = NULL;
6779 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6780 		algo = comp_algorithm_lookup(i);
6781 		if (!algo)
6782 			continue;
6783 
6784 		if (!m) {
6785 			IPSEC_ASSERT(l <= MLEN,
6786 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6787 			MGET(m, M_NOWAIT, MT_DATA);
6788 			if (m) {
6789 				M_ALIGN(m, l);
6790 				m->m_len = l;
6791 				m->m_next = NULL;
6792 			}
6793 		} else
6794 			M_PREPEND(m, l, M_NOWAIT);
6795 		if (!m)
6796 			return NULL;
6797 
6798 		comb = mtod(m, struct sadb_comb *);
6799 		bzero(comb, sizeof(*comb));
6800 		key_getcomb_setlifetime(comb);
6801 		comb->sadb_comb_encrypt = i;
6802 		/* what should we set into sadb_comb_*_{min,max}bits? */
6803 	}
6804 
6805 	return m;
6806 }
6807 
6808 /*
6809  * XXX no way to pass mode (transport/tunnel) to userland
6810  * XXX replay checking?
6811  * XXX sysctl interface to ipsec_{ah,esp}_keymin
6812  */
6813 static struct mbuf *
6814 key_getprop(const struct secasindex *saidx)
6815 {
6816 	struct sadb_prop *prop;
6817 	struct mbuf *m, *n;
6818 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6819 	int totlen;
6820 
6821 	switch (saidx->proto)  {
6822 	case IPPROTO_ESP:
6823 		m = key_getcomb_ealg();
6824 		break;
6825 	case IPPROTO_AH:
6826 		m = key_getcomb_ah();
6827 		break;
6828 	case IPPROTO_IPCOMP:
6829 		m = key_getcomb_ipcomp();
6830 		break;
6831 	default:
6832 		return NULL;
6833 	}
6834 
6835 	if (!m)
6836 		return NULL;
6837 	M_PREPEND(m, l, M_NOWAIT);
6838 	if (!m)
6839 		return NULL;
6840 
6841 	totlen = 0;
6842 	for (n = m; n; n = n->m_next)
6843 		totlen += n->m_len;
6844 
6845 	prop = mtod(m, struct sadb_prop *);
6846 	bzero(prop, sizeof(*prop));
6847 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6848 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6849 	prop->sadb_prop_replay = 32;	/* XXX */
6850 
6851 	return m;
6852 }
6853 
6854 /*
6855  * SADB_ACQUIRE processing called by key_checkrequest() and key_acquire2().
6856  * send
6857  *   <base, SA, address(SD), (address(P)), x_policy,
6858  *       (identity(SD),) (sensitivity,) proposal>
6859  * to KMD, and expect to receive
6860  *   <base> with SADB_ACQUIRE if error occurred,
6861  * or
6862  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6863  * from KMD by PF_KEY.
6864  *
6865  * XXX x_policy is outside of RFC2367 (KAME extension).
6866  * XXX sensitivity is not supported.
6867  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6868  * see comment for key_getcomb_ipcomp().
6869  *
6870  * OUT:
6871  *    0     : succeed
6872  *    others: error number
6873  */
6874 static int
6875 key_acquire(const struct secasindex *saidx, struct secpolicy *sp)
6876 {
6877 	union sockaddr_union addr;
6878 	struct mbuf *result, *m;
6879 	uint32_t seq;
6880 	int error;
6881 	uint16_t ul_proto;
6882 	uint8_t mask, satype;
6883 
6884 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
6885 	satype = key_proto2satype(saidx->proto);
6886 	IPSEC_ASSERT(satype != 0, ("null satype, protocol %u", saidx->proto));
6887 
6888 	error = -1;
6889 	result = NULL;
6890 	ul_proto = IPSEC_ULPROTO_ANY;
6891 
6892 	/* Get seq number to check whether sending message or not. */
6893 	seq = key_getacq(saidx, &error);
6894 	if (seq == 0)
6895 		return (error);
6896 
6897 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0);
6898 	if (!m) {
6899 		error = ENOBUFS;
6900 		goto fail;
6901 	}
6902 	result = m;
6903 
6904 	/*
6905 	 * set sadb_address for saidx's.
6906 	 *
6907 	 * Note that if sp is supplied, then we're being called from
6908 	 * key_allocsa_policy() and should supply port and protocol
6909 	 * information.
6910 	 * XXXAE: why only TCP and UDP? ICMP and SCTP looks applicable too.
6911 	 * XXXAE: probably we can handle this in the ipsec[46]_allocsa().
6912 	 * XXXAE: it looks like we should save this info in the ACQ entry.
6913 	 */
6914 	if (sp != NULL && (sp->spidx.ul_proto == IPPROTO_TCP ||
6915 	    sp->spidx.ul_proto == IPPROTO_UDP))
6916 		ul_proto = sp->spidx.ul_proto;
6917 
6918 	addr = saidx->src;
6919 	mask = FULLMASK;
6920 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6921 		switch (sp->spidx.src.sa.sa_family) {
6922 		case AF_INET:
6923 			if (sp->spidx.src.sin.sin_port != IPSEC_PORT_ANY) {
6924 				addr.sin.sin_port = sp->spidx.src.sin.sin_port;
6925 				mask = sp->spidx.prefs;
6926 			}
6927 			break;
6928 		case AF_INET6:
6929 			if (sp->spidx.src.sin6.sin6_port != IPSEC_PORT_ANY) {
6930 				addr.sin6.sin6_port =
6931 				    sp->spidx.src.sin6.sin6_port;
6932 				mask = sp->spidx.prefs;
6933 			}
6934 			break;
6935 		default:
6936 			break;
6937 		}
6938 	}
6939 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &addr.sa, mask, ul_proto);
6940 	if (!m) {
6941 		error = ENOBUFS;
6942 		goto fail;
6943 	}
6944 	m_cat(result, m);
6945 
6946 	addr = saidx->dst;
6947 	mask = FULLMASK;
6948 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6949 		switch (sp->spidx.dst.sa.sa_family) {
6950 		case AF_INET:
6951 			if (sp->spidx.dst.sin.sin_port != IPSEC_PORT_ANY) {
6952 				addr.sin.sin_port = sp->spidx.dst.sin.sin_port;
6953 				mask = sp->spidx.prefd;
6954 			}
6955 			break;
6956 		case AF_INET6:
6957 			if (sp->spidx.dst.sin6.sin6_port != IPSEC_PORT_ANY) {
6958 				addr.sin6.sin6_port =
6959 				    sp->spidx.dst.sin6.sin6_port;
6960 				mask = sp->spidx.prefd;
6961 			}
6962 			break;
6963 		default:
6964 			break;
6965 		}
6966 	}
6967 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &addr.sa, mask, ul_proto);
6968 	if (!m) {
6969 		error = ENOBUFS;
6970 		goto fail;
6971 	}
6972 	m_cat(result, m);
6973 
6974 	/* XXX proxy address (optional) */
6975 
6976 	/*
6977 	 * Set sadb_x_policy. This is KAME extension to RFC2367.
6978 	 */
6979 	if (sp != NULL) {
6980 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6981 		    sp->priority);
6982 		if (!m) {
6983 			error = ENOBUFS;
6984 			goto fail;
6985 		}
6986 		m_cat(result, m);
6987 	}
6988 
6989 	/*
6990 	 * Set sadb_x_sa2 extension if saidx->reqid is not zero.
6991 	 * This is FreeBSD extension to RFC2367.
6992 	 */
6993 	if (saidx->reqid != 0) {
6994 		m = key_setsadbxsa2(saidx->mode, 0, saidx->reqid);
6995 		if (m == NULL) {
6996 			error = ENOBUFS;
6997 			goto fail;
6998 		}
6999 		m_cat(result, m);
7000 	}
7001 	/* XXX identity (optional) */
7002 #if 0
7003 	if (idexttype && fqdn) {
7004 		/* create identity extension (FQDN) */
7005 		struct sadb_ident *id;
7006 		int fqdnlen;
7007 
7008 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
7009 		id = (struct sadb_ident *)p;
7010 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
7011 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
7012 		id->sadb_ident_exttype = idexttype;
7013 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
7014 		bcopy(fqdn, id + 1, fqdnlen);
7015 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
7016 	}
7017 
7018 	if (idexttype) {
7019 		/* create identity extension (USERFQDN) */
7020 		struct sadb_ident *id;
7021 		int userfqdnlen;
7022 
7023 		if (userfqdn) {
7024 			/* +1 for terminating-NUL */
7025 			userfqdnlen = strlen(userfqdn) + 1;
7026 		} else
7027 			userfqdnlen = 0;
7028 		id = (struct sadb_ident *)p;
7029 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
7030 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
7031 		id->sadb_ident_exttype = idexttype;
7032 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
7033 		/* XXX is it correct? */
7034 		if (curproc && curproc->p_cred)
7035 			id->sadb_ident_id = curproc->p_cred->p_ruid;
7036 		if (userfqdn && userfqdnlen)
7037 			bcopy(userfqdn, id + 1, userfqdnlen);
7038 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
7039 	}
7040 #endif
7041 
7042 	/* XXX sensitivity (optional) */
7043 
7044 	/* create proposal/combination extension */
7045 	m = key_getprop(saidx);
7046 #if 0
7047 	/*
7048 	 * spec conformant: always attach proposal/combination extension,
7049 	 * the problem is that we have no way to attach it for ipcomp,
7050 	 * due to the way sadb_comb is declared in RFC2367.
7051 	 */
7052 	if (!m) {
7053 		error = ENOBUFS;
7054 		goto fail;
7055 	}
7056 	m_cat(result, m);
7057 #else
7058 	/*
7059 	 * outside of spec; make proposal/combination extension optional.
7060 	 */
7061 	if (m)
7062 		m_cat(result, m);
7063 #endif
7064 
7065 	if ((result->m_flags & M_PKTHDR) == 0) {
7066 		error = EINVAL;
7067 		goto fail;
7068 	}
7069 
7070 	if (result->m_len < sizeof(struct sadb_msg)) {
7071 		result = m_pullup(result, sizeof(struct sadb_msg));
7072 		if (result == NULL) {
7073 			error = ENOBUFS;
7074 			goto fail;
7075 		}
7076 	}
7077 
7078 	result->m_pkthdr.len = 0;
7079 	for (m = result; m; m = m->m_next)
7080 		result->m_pkthdr.len += m->m_len;
7081 
7082 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7083 	    PFKEY_UNIT64(result->m_pkthdr.len);
7084 
7085 	KEYDBG(KEY_STAMP,
7086 	    printf("%s: SP(%p)\n", __func__, sp));
7087 	KEYDBG(KEY_DATA, kdebug_secasindex(saidx, NULL));
7088 
7089 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7090 
7091  fail:
7092 	if (result)
7093 		m_freem(result);
7094 	return error;
7095 }
7096 
7097 static uint32_t
7098 key_newacq(const struct secasindex *saidx, int *perror)
7099 {
7100 	struct secacq *acq;
7101 	uint32_t seq;
7102 
7103 	acq = malloc(sizeof(*acq), M_IPSEC_SAQ, M_NOWAIT | M_ZERO);
7104 	if (acq == NULL) {
7105 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7106 		*perror = ENOBUFS;
7107 		return (0);
7108 	}
7109 
7110 	/* copy secindex */
7111 	bcopy(saidx, &acq->saidx, sizeof(acq->saidx));
7112 	acq->created = time_second;
7113 	acq->count = 0;
7114 
7115 	/* add to acqtree */
7116 	ACQ_LOCK();
7117 	seq = acq->seq = (V_acq_seq == ~0 ? 1 : ++V_acq_seq);
7118 	LIST_INSERT_HEAD(&V_acqtree, acq, chain);
7119 	LIST_INSERT_HEAD(ACQADDRHASH_HASH(saidx), acq, addrhash);
7120 	LIST_INSERT_HEAD(ACQSEQHASH_HASH(seq), acq, seqhash);
7121 	ACQ_UNLOCK();
7122 	*perror = 0;
7123 	return (seq);
7124 }
7125 
7126 static uint32_t
7127 key_getacq(const struct secasindex *saidx, int *perror)
7128 {
7129 	struct secacq *acq;
7130 	uint32_t seq;
7131 
7132 	ACQ_LOCK();
7133 	LIST_FOREACH(acq, ACQADDRHASH_HASH(saidx), addrhash) {
7134 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY)) {
7135 			if (acq->count > V_key_blockacq_count) {
7136 				/*
7137 				 * Reset counter and send message.
7138 				 * Also reset created time to keep ACQ for
7139 				 * this saidx.
7140 				 */
7141 				acq->created = time_second;
7142 				acq->count = 0;
7143 				seq = acq->seq;
7144 			} else {
7145 				/*
7146 				 * Increment counter and do nothing.
7147 				 * We send SADB_ACQUIRE message only
7148 				 * for each V_key_blockacq_count packet.
7149 				 */
7150 				acq->count++;
7151 				seq = 0;
7152 			}
7153 			break;
7154 		}
7155 	}
7156 	ACQ_UNLOCK();
7157 	if (acq != NULL) {
7158 		*perror = 0;
7159 		return (seq);
7160 	}
7161 	/* allocate new  entry */
7162 	return (key_newacq(saidx, perror));
7163 }
7164 
7165 static int
7166 key_acqreset(uint32_t seq)
7167 {
7168 	struct secacq *acq;
7169 
7170 	ACQ_LOCK();
7171 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
7172 		if (acq->seq == seq) {
7173 			acq->count = 0;
7174 			acq->created = time_second;
7175 			break;
7176 		}
7177 	}
7178 	ACQ_UNLOCK();
7179 	if (acq == NULL)
7180 		return (ESRCH);
7181 	return (0);
7182 }
7183 /*
7184  * Mark ACQ entry as stale to remove it in key_flush_acq().
7185  * Called after successful SADB_GETSPI message.
7186  */
7187 static int
7188 key_acqdone(const struct secasindex *saidx, uint32_t seq)
7189 {
7190 	struct secacq *acq;
7191 
7192 	ACQ_LOCK();
7193 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
7194 		if (acq->seq == seq)
7195 			break;
7196 	}
7197 	if (acq != NULL) {
7198 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY) == 0) {
7199 			ipseclog((LOG_DEBUG,
7200 			    "%s: Mismatched saidx for ACQ %u\n", __func__, seq));
7201 			acq = NULL;
7202 		} else {
7203 			acq->created = 0;
7204 		}
7205 	} else {
7206 		ipseclog((LOG_DEBUG,
7207 		    "%s: ACQ %u is not found.\n", __func__, seq));
7208 	}
7209 	ACQ_UNLOCK();
7210 	if (acq == NULL)
7211 		return (ESRCH);
7212 	return (0);
7213 }
7214 
7215 static struct secspacq *
7216 key_newspacq(struct secpolicyindex *spidx)
7217 {
7218 	struct secspacq *acq;
7219 
7220 	/* get new entry */
7221 	acq = malloc(sizeof(struct secspacq), M_IPSEC_SAQ, M_NOWAIT|M_ZERO);
7222 	if (acq == NULL) {
7223 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7224 		return NULL;
7225 	}
7226 
7227 	/* copy secindex */
7228 	bcopy(spidx, &acq->spidx, sizeof(acq->spidx));
7229 	acq->created = time_second;
7230 	acq->count = 0;
7231 
7232 	/* add to spacqtree */
7233 	SPACQ_LOCK();
7234 	LIST_INSERT_HEAD(&V_spacqtree, acq, chain);
7235 	SPACQ_UNLOCK();
7236 
7237 	return acq;
7238 }
7239 
7240 static struct secspacq *
7241 key_getspacq(struct secpolicyindex *spidx)
7242 {
7243 	struct secspacq *acq;
7244 
7245 	SPACQ_LOCK();
7246 	LIST_FOREACH(acq, &V_spacqtree, chain) {
7247 		if (key_cmpspidx_exactly(spidx, &acq->spidx)) {
7248 			/* NB: return holding spacq_lock */
7249 			return acq;
7250 		}
7251 	}
7252 	SPACQ_UNLOCK();
7253 
7254 	return NULL;
7255 }
7256 
7257 /*
7258  * SADB_ACQUIRE processing,
7259  * in first situation, is receiving
7260  *   <base>
7261  * from the ikmpd, and clear sequence of its secasvar entry.
7262  *
7263  * In second situation, is receiving
7264  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
7265  * from a user land process, and return
7266  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
7267  * to the socket.
7268  *
7269  * m will always be freed.
7270  */
7271 static int
7272 key_acquire2(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7273 {
7274 	SAHTREE_RLOCK_TRACKER;
7275 	struct sadb_address *src0, *dst0;
7276 	struct secasindex saidx;
7277 	struct secashead *sah;
7278 	uint32_t reqid;
7279 	int error;
7280 	uint8_t mode, proto;
7281 
7282 	IPSEC_ASSERT(so != NULL, ("null socket"));
7283 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7284 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7285 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7286 
7287 	/*
7288 	 * Error message from KMd.
7289 	 * We assume that if error was occurred in IKEd, the length of PFKEY
7290 	 * message is equal to the size of sadb_msg structure.
7291 	 * We do not raise error even if error occurred in this function.
7292 	 */
7293 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
7294 		/* check sequence number */
7295 		if (mhp->msg->sadb_msg_seq == 0 ||
7296 		    mhp->msg->sadb_msg_errno == 0) {
7297 			ipseclog((LOG_DEBUG, "%s: must specify sequence "
7298 				"number and errno.\n", __func__));
7299 		} else {
7300 			/*
7301 			 * IKEd reported that error occurred.
7302 			 * XXXAE: what it expects from the kernel?
7303 			 * Probably we should send SADB_ACQUIRE again?
7304 			 * If so, reset ACQ's state.
7305 			 * XXXAE: it looks useless.
7306 			 */
7307 			key_acqreset(mhp->msg->sadb_msg_seq);
7308 		}
7309 		m_freem(m);
7310 		return (0);
7311 	}
7312 
7313 	/*
7314 	 * This message is from user land.
7315 	 */
7316 
7317 	/* map satype to proto */
7318 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7319 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7320 		    __func__));
7321 		return key_senderror(so, m, EINVAL);
7322 	}
7323 
7324 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
7325 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
7326 	    SADB_CHECKHDR(mhp, SADB_EXT_PROPOSAL)) {
7327 		ipseclog((LOG_DEBUG,
7328 		    "%s: invalid message: missing required header.\n",
7329 		    __func__));
7330 		return key_senderror(so, m, EINVAL);
7331 	}
7332 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
7333 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
7334 	    SADB_CHECKLEN(mhp, SADB_EXT_PROPOSAL)) {
7335 		ipseclog((LOG_DEBUG,
7336 		    "%s: invalid message: wrong header size.\n", __func__));
7337 		return key_senderror(so, m, EINVAL);
7338 	}
7339 
7340 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
7341 		mode = IPSEC_MODE_ANY;
7342 		reqid = 0;
7343 	} else {
7344 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
7345 			ipseclog((LOG_DEBUG,
7346 			    "%s: invalid message: wrong header size.\n",
7347 			    __func__));
7348 			return key_senderror(so, m, EINVAL);
7349 		}
7350 		mode = ((struct sadb_x_sa2 *)
7351 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
7352 		reqid = ((struct sadb_x_sa2 *)
7353 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
7354 	}
7355 
7356 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
7357 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
7358 
7359 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
7360 	    (struct sockaddr *)(dst0 + 1));
7361 	if (error != 0) {
7362 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
7363 		return key_senderror(so, m, EINVAL);
7364 	}
7365 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
7366 
7367 	/* get a SA index */
7368 	SAHTREE_RLOCK();
7369 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
7370 		if (key_cmpsaidx(&sah->saidx, &saidx, CMP_MODE_REQID))
7371 			break;
7372 	}
7373 	SAHTREE_RUNLOCK();
7374 	if (sah != NULL) {
7375 		ipseclog((LOG_DEBUG, "%s: a SA exists already.\n", __func__));
7376 		return key_senderror(so, m, EEXIST);
7377 	}
7378 
7379 	error = key_acquire(&saidx, NULL);
7380 	if (error != 0) {
7381 		ipseclog((LOG_DEBUG,
7382 		    "%s: error %d returned from key_acquire()\n",
7383 			__func__, error));
7384 		return key_senderror(so, m, error);
7385 	}
7386 	m_freem(m);
7387 	return (0);
7388 }
7389 
7390 /*
7391  * SADB_REGISTER processing.
7392  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
7393  * receive
7394  *   <base>
7395  * from the ikmpd, and register a socket to send PF_KEY messages,
7396  * and send
7397  *   <base, supported>
7398  * to KMD by PF_KEY.
7399  * If socket is detached, must free from regnode.
7400  *
7401  * m will always be freed.
7402  */
7403 static int
7404 key_register(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7405 {
7406 	struct secreg *reg, *newreg = NULL;
7407 
7408 	IPSEC_ASSERT(so != NULL, ("null socket"));
7409 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7410 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7411 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7412 
7413 	/* check for invalid register message */
7414 	if (mhp->msg->sadb_msg_satype >= sizeof(V_regtree)/sizeof(V_regtree[0]))
7415 		return key_senderror(so, m, EINVAL);
7416 
7417 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
7418 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
7419 		goto setmsg;
7420 
7421 	/* check whether existing or not */
7422 	REGTREE_LOCK();
7423 	LIST_FOREACH(reg, &V_regtree[mhp->msg->sadb_msg_satype], chain) {
7424 		if (reg->so == so) {
7425 			REGTREE_UNLOCK();
7426 			ipseclog((LOG_DEBUG, "%s: socket exists already.\n",
7427 				__func__));
7428 			return key_senderror(so, m, EEXIST);
7429 		}
7430 	}
7431 
7432 	/* create regnode */
7433 	newreg =  malloc(sizeof(struct secreg), M_IPSEC_SAR, M_NOWAIT|M_ZERO);
7434 	if (newreg == NULL) {
7435 		REGTREE_UNLOCK();
7436 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7437 		return key_senderror(so, m, ENOBUFS);
7438 	}
7439 
7440 	newreg->so = so;
7441 	((struct keycb *)(so->so_pcb))->kp_registered++;
7442 
7443 	/* add regnode to regtree. */
7444 	LIST_INSERT_HEAD(&V_regtree[mhp->msg->sadb_msg_satype], newreg, chain);
7445 	REGTREE_UNLOCK();
7446 
7447   setmsg:
7448     {
7449 	struct mbuf *n;
7450 	struct sadb_msg *newmsg;
7451 	struct sadb_supported *sup;
7452 	u_int len, alen, elen;
7453 	int off;
7454 	int i;
7455 	struct sadb_alg *alg;
7456 
7457 	/* create new sadb_msg to reply. */
7458 	alen = 0;
7459 	for (i = 1; i <= SADB_AALG_MAX; i++) {
7460 		if (auth_algorithm_lookup(i))
7461 			alen += sizeof(struct sadb_alg);
7462 	}
7463 	if (alen)
7464 		alen += sizeof(struct sadb_supported);
7465 	elen = 0;
7466 	for (i = 1; i <= SADB_EALG_MAX; i++) {
7467 		if (enc_algorithm_lookup(i))
7468 			elen += sizeof(struct sadb_alg);
7469 	}
7470 	if (elen)
7471 		elen += sizeof(struct sadb_supported);
7472 
7473 	len = sizeof(struct sadb_msg) + alen + elen;
7474 
7475 	if (len > MCLBYTES)
7476 		return key_senderror(so, m, ENOBUFS);
7477 
7478 	n = key_mget(len);
7479 	if (n == NULL)
7480 		return key_senderror(so, m, ENOBUFS);
7481 
7482 	n->m_pkthdr.len = n->m_len = len;
7483 	n->m_next = NULL;
7484 	off = 0;
7485 
7486 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
7487 	newmsg = mtod(n, struct sadb_msg *);
7488 	newmsg->sadb_msg_errno = 0;
7489 	newmsg->sadb_msg_len = PFKEY_UNIT64(len);
7490 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7491 
7492 	/* for authentication algorithm */
7493 	if (alen) {
7494 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7495 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7496 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7497 		off += PFKEY_ALIGN8(sizeof(*sup));
7498 
7499 		for (i = 1; i <= SADB_AALG_MAX; i++) {
7500 			const struct auth_hash *aalgo;
7501 			u_int16_t minkeysize, maxkeysize;
7502 
7503 			aalgo = auth_algorithm_lookup(i);
7504 			if (!aalgo)
7505 				continue;
7506 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7507 			alg->sadb_alg_id = i;
7508 			alg->sadb_alg_ivlen = 0;
7509 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7510 			alg->sadb_alg_minbits = _BITS(minkeysize);
7511 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7512 			off += PFKEY_ALIGN8(sizeof(*alg));
7513 		}
7514 	}
7515 
7516 	/* for encryption algorithm */
7517 	if (elen) {
7518 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7519 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7520 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7521 		off += PFKEY_ALIGN8(sizeof(*sup));
7522 
7523 		for (i = 1; i <= SADB_EALG_MAX; i++) {
7524 			const struct enc_xform *ealgo;
7525 
7526 			ealgo = enc_algorithm_lookup(i);
7527 			if (!ealgo)
7528 				continue;
7529 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7530 			alg->sadb_alg_id = i;
7531 			alg->sadb_alg_ivlen = ealgo->ivsize;
7532 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7533 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7534 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7535 		}
7536 	}
7537 
7538 	IPSEC_ASSERT(off == len,
7539 		("length assumption failed (off %u len %u)", off, len));
7540 
7541 	m_freem(m);
7542 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7543     }
7544 }
7545 
7546 /*
7547  * free secreg entry registered.
7548  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7549  */
7550 void
7551 key_freereg(struct socket *so)
7552 {
7553 	struct secreg *reg;
7554 	int i;
7555 
7556 	IPSEC_ASSERT(so != NULL, ("NULL so"));
7557 
7558 	/*
7559 	 * check whether existing or not.
7560 	 * check all type of SA, because there is a potential that
7561 	 * one socket is registered to multiple type of SA.
7562 	 */
7563 	REGTREE_LOCK();
7564 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7565 		LIST_FOREACH(reg, &V_regtree[i], chain) {
7566 			if (reg->so == so && __LIST_CHAINED(reg)) {
7567 				LIST_REMOVE(reg, chain);
7568 				free(reg, M_IPSEC_SAR);
7569 				break;
7570 			}
7571 		}
7572 	}
7573 	REGTREE_UNLOCK();
7574 }
7575 
7576 /*
7577  * SADB_EXPIRE processing
7578  * send
7579  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7580  * to KMD by PF_KEY.
7581  * NOTE: We send only soft lifetime extension.
7582  *
7583  * OUT:	0	: succeed
7584  *	others	: error number
7585  */
7586 static int
7587 key_expire(struct secasvar *sav, int hard)
7588 {
7589 	struct mbuf *result = NULL, *m;
7590 	struct sadb_lifetime *lt;
7591 	uint32_t replay_count;
7592 	int error, len;
7593 	uint8_t satype;
7594 
7595 	SECASVAR_RLOCK_TRACKER;
7596 
7597 	IPSEC_ASSERT (sav != NULL, ("null sav"));
7598 	IPSEC_ASSERT (sav->sah != NULL, ("null sa header"));
7599 
7600 	KEYDBG(KEY_STAMP,
7601 	    printf("%s: SA(%p) expired %s lifetime\n", __func__,
7602 		sav, hard ? "hard": "soft"));
7603 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
7604 	/* set msg header */
7605 	satype = key_proto2satype(sav->sah->saidx.proto);
7606 	IPSEC_ASSERT(satype != 0, ("invalid proto, satype %u", satype));
7607 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, sav->refcnt);
7608 	if (!m) {
7609 		error = ENOBUFS;
7610 		goto fail;
7611 	}
7612 	result = m;
7613 
7614 	/* create SA extension */
7615 	m = key_setsadbsa(sav);
7616 	if (!m) {
7617 		error = ENOBUFS;
7618 		goto fail;
7619 	}
7620 	m_cat(result, m);
7621 
7622 	/* create SA extension */
7623 	SECASVAR_RLOCK(sav);
7624 	replay_count = sav->replay ? sav->replay->count : 0;
7625 	SECASVAR_RUNLOCK(sav);
7626 
7627 	m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
7628 			sav->sah->saidx.reqid);
7629 	if (!m) {
7630 		error = ENOBUFS;
7631 		goto fail;
7632 	}
7633 	m_cat(result, m);
7634 
7635 	if (sav->replay && sav->replay->wsize > UINT8_MAX) {
7636 		m = key_setsadbxsareplay(sav->replay->wsize);
7637 		if (!m) {
7638 			error = ENOBUFS;
7639 			goto fail;
7640 		}
7641 		m_cat(result, m);
7642 	}
7643 
7644 	/* create lifetime extension (current and soft) */
7645 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7646 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
7647 	if (m == NULL) {
7648 		error = ENOBUFS;
7649 		goto fail;
7650 	}
7651 	m_align(m, len);
7652 	m->m_len = len;
7653 	bzero(mtod(m, caddr_t), len);
7654 	lt = mtod(m, struct sadb_lifetime *);
7655 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7656 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7657 	lt->sadb_lifetime_allocations =
7658 	    (uint32_t)counter_u64_fetch(sav->lft_c_allocations);
7659 	lt->sadb_lifetime_bytes =
7660 	    counter_u64_fetch(sav->lft_c_bytes);
7661 	lt->sadb_lifetime_addtime = sav->created;
7662 	lt->sadb_lifetime_usetime = sav->firstused;
7663 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
7664 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7665 	if (hard) {
7666 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
7667 		lt->sadb_lifetime_allocations = sav->lft_h->allocations;
7668 		lt->sadb_lifetime_bytes = sav->lft_h->bytes;
7669 		lt->sadb_lifetime_addtime = sav->lft_h->addtime;
7670 		lt->sadb_lifetime_usetime = sav->lft_h->usetime;
7671 	} else {
7672 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_SOFT;
7673 		lt->sadb_lifetime_allocations = sav->lft_s->allocations;
7674 		lt->sadb_lifetime_bytes = sav->lft_s->bytes;
7675 		lt->sadb_lifetime_addtime = sav->lft_s->addtime;
7676 		lt->sadb_lifetime_usetime = sav->lft_s->usetime;
7677 	}
7678 	m_cat(result, m);
7679 
7680 	/* set sadb_address for source */
7681 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
7682 	    &sav->sah->saidx.src.sa,
7683 	    FULLMASK, IPSEC_ULPROTO_ANY);
7684 	if (!m) {
7685 		error = ENOBUFS;
7686 		goto fail;
7687 	}
7688 	m_cat(result, m);
7689 
7690 	/* set sadb_address for destination */
7691 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
7692 	    &sav->sah->saidx.dst.sa,
7693 	    FULLMASK, IPSEC_ULPROTO_ANY);
7694 	if (!m) {
7695 		error = ENOBUFS;
7696 		goto fail;
7697 	}
7698 	m_cat(result, m);
7699 
7700 	/*
7701 	 * XXX-BZ Handle NAT-T extensions here.
7702 	 * XXXAE: it doesn't seem quite useful. IKEs should not depend on
7703 	 * this information, we report only significant SA fields.
7704 	 */
7705 
7706 	if ((result->m_flags & M_PKTHDR) == 0) {
7707 		error = EINVAL;
7708 		goto fail;
7709 	}
7710 
7711 	if (result->m_len < sizeof(struct sadb_msg)) {
7712 		result = m_pullup(result, sizeof(struct sadb_msg));
7713 		if (result == NULL) {
7714 			error = ENOBUFS;
7715 			goto fail;
7716 		}
7717 	}
7718 
7719 	result->m_pkthdr.len = 0;
7720 	for (m = result; m; m = m->m_next)
7721 		result->m_pkthdr.len += m->m_len;
7722 
7723 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7724 	    PFKEY_UNIT64(result->m_pkthdr.len);
7725 
7726 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7727 
7728  fail:
7729 	if (result)
7730 		m_freem(result);
7731 	return error;
7732 }
7733 
7734 static void
7735 key_freesah_flushed(struct secashead_queue *flushq)
7736 {
7737 	struct secashead *sah, *nextsah;
7738 	struct secasvar *sav, *nextsav;
7739 
7740 	sah = TAILQ_FIRST(flushq);
7741 	while (sah != NULL) {
7742 		sav = TAILQ_FIRST(&sah->savtree_larval);
7743 		while (sav != NULL) {
7744 			nextsav = TAILQ_NEXT(sav, chain);
7745 			TAILQ_REMOVE(&sah->savtree_larval, sav, chain);
7746 			key_freesav(&sav); /* release last reference */
7747 			key_freesah(&sah); /* release reference from SAV */
7748 			sav = nextsav;
7749 		}
7750 		sav = TAILQ_FIRST(&sah->savtree_alive);
7751 		while (sav != NULL) {
7752 			nextsav = TAILQ_NEXT(sav, chain);
7753 			TAILQ_REMOVE(&sah->savtree_alive, sav, chain);
7754 			key_freesav(&sav); /* release last reference */
7755 			key_freesah(&sah); /* release reference from SAV */
7756 			sav = nextsav;
7757 		}
7758 		nextsah = TAILQ_NEXT(sah, chain);
7759 		key_freesah(&sah);	/* release last reference */
7760 		sah = nextsah;
7761 	}
7762 }
7763 
7764 /*
7765  * SADB_FLUSH processing
7766  * receive
7767  *   <base>
7768  * from the ikmpd, and free all entries in secastree.
7769  * and send,
7770  *   <base>
7771  * to the ikmpd.
7772  * NOTE: to do is only marking SADB_SASTATE_DEAD.
7773  *
7774  * m will always be freed.
7775  */
7776 static int
7777 key_flush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7778 {
7779 	struct secashead_queue flushq;
7780 	struct sadb_msg *newmsg;
7781 	struct secashead *sah, *nextsah;
7782 	struct secasvar *sav;
7783 	uint8_t proto;
7784 	int i;
7785 
7786 	IPSEC_ASSERT(so != NULL, ("null socket"));
7787 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7788 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7789 
7790 	/* map satype to proto */
7791 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7792 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7793 			__func__));
7794 		return key_senderror(so, m, EINVAL);
7795 	}
7796 	KEYDBG(KEY_STAMP,
7797 	    printf("%s: proto %u\n", __func__, proto));
7798 
7799 	TAILQ_INIT(&flushq);
7800 	if (proto == IPSEC_PROTO_ANY) {
7801 		/* no SATYPE specified, i.e. flushing all SA. */
7802 		SAHTREE_WLOCK();
7803 		/* Move all SAHs into flushq */
7804 		TAILQ_CONCAT(&flushq, &V_sahtree, chain);
7805 		/* Flush all buckets in SPI hash */
7806 		for (i = 0; i < V_savhash_mask + 1; i++)
7807 			LIST_INIT(&V_savhashtbl[i]);
7808 		/* Flush all buckets in SAHADDRHASH */
7809 		for (i = 0; i < V_sahaddrhash_mask + 1; i++)
7810 			LIST_INIT(&V_sahaddrhashtbl[i]);
7811 		/* Mark all SAHs as unlinked */
7812 		TAILQ_FOREACH(sah, &flushq, chain) {
7813 			sah->state = SADB_SASTATE_DEAD;
7814 			/*
7815 			 * Callout handler makes its job using
7816 			 * RLOCK and drain queues. In case, when this
7817 			 * function will be called just before it
7818 			 * acquires WLOCK, we need to mark SAs as
7819 			 * unlinked to prevent second unlink.
7820 			 */
7821 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7822 				sav->state = SADB_SASTATE_DEAD;
7823 				ipsec_accel_forget_sav(sav);
7824 			}
7825 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7826 				sav->state = SADB_SASTATE_DEAD;
7827 				ipsec_accel_forget_sav(sav);
7828 			}
7829 		}
7830 		SAHTREE_WUNLOCK();
7831 	} else {
7832 		SAHTREE_WLOCK();
7833 		sah = TAILQ_FIRST(&V_sahtree);
7834 		while (sah != NULL) {
7835 			IPSEC_ASSERT(sah->state != SADB_SASTATE_DEAD,
7836 			    ("DEAD SAH %p in SADB_FLUSH", sah));
7837 			nextsah = TAILQ_NEXT(sah, chain);
7838 			if (sah->saidx.proto != proto) {
7839 				sah = nextsah;
7840 				continue;
7841 			}
7842 			sah->state = SADB_SASTATE_DEAD;
7843 			TAILQ_REMOVE(&V_sahtree, sah, chain);
7844 			LIST_REMOVE(sah, addrhash);
7845 			/* Unlink all SAs from SPI hash */
7846 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7847 				LIST_REMOVE(sav, spihash);
7848 				sav->state = SADB_SASTATE_DEAD;
7849 				ipsec_accel_forget_sav(sav);
7850 			}
7851 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7852 				LIST_REMOVE(sav, spihash);
7853 				sav->state = SADB_SASTATE_DEAD;
7854 				ipsec_accel_forget_sav(sav);
7855 			}
7856 			/* Add SAH into flushq */
7857 			TAILQ_INSERT_HEAD(&flushq, sah, chain);
7858 			sah = nextsah;
7859 		}
7860 		SAHTREE_WUNLOCK();
7861 	}
7862 
7863 	key_freesah_flushed(&flushq);
7864 	/* Free all queued SAs and SAHs */
7865 	if (m->m_len < sizeof(struct sadb_msg) ||
7866 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7867 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7868 		return key_senderror(so, m, ENOBUFS);
7869 	}
7870 
7871 	if (m->m_next)
7872 		m_freem(m->m_next);
7873 	m->m_next = NULL;
7874 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7875 	newmsg = mtod(m, struct sadb_msg *);
7876 	newmsg->sadb_msg_errno = 0;
7877 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7878 
7879 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7880 }
7881 
7882 /*
7883  * SADB_DUMP processing
7884  * dump all entries including status of DEAD in SAD.
7885  * receive
7886  *   <base>
7887  * from the ikmpd, and dump all secasvar leaves
7888  * and send,
7889  *   <base> .....
7890  * to the ikmpd.
7891  *
7892  * m will always be freed.
7893  */
7894 static int
7895 key_dump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7896 {
7897 	SAHTREE_RLOCK_TRACKER;
7898 	struct secashead *sah;
7899 	struct secasvar *sav;
7900 	struct mbuf *n;
7901 	uint32_t cnt;
7902 	uint8_t proto, satype;
7903 
7904 	IPSEC_ASSERT(so != NULL, ("null socket"));
7905 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7906 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7907 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7908 
7909 	/* map satype to proto */
7910 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7911 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7912 		    __func__));
7913 		return key_senderror(so, m, EINVAL);
7914 	}
7915 
7916 	/* count sav entries to be sent to the userland. */
7917 	cnt = 0;
7918 	IFNET_RLOCK();
7919 	SAHTREE_RLOCK();
7920 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7921 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7922 		    proto != sah->saidx.proto)
7923 			continue;
7924 
7925 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain)
7926 			cnt++;
7927 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain)
7928 			cnt++;
7929 	}
7930 
7931 	if (cnt == 0) {
7932 		SAHTREE_RUNLOCK();
7933 		IFNET_RUNLOCK();
7934 		return key_senderror(so, m, ENOENT);
7935 	}
7936 
7937 	/* send this to the userland, one at a time. */
7938 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7939 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7940 		    proto != sah->saidx.proto)
7941 			continue;
7942 
7943 		/* map proto to satype */
7944 		if ((satype = key_proto2satype(sah->saidx.proto)) == 0) {
7945 			SAHTREE_RUNLOCK();
7946 			IFNET_RUNLOCK();
7947 			ipseclog((LOG_DEBUG, "%s: there was invalid proto in "
7948 			    "SAD.\n", __func__));
7949 			return key_senderror(so, m, EINVAL);
7950 		}
7951 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7952 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7953 			    --cnt, mhp->msg->sadb_msg_pid, &sahtree_tracker);
7954 			if (n == NULL) {
7955 				SAHTREE_RUNLOCK();
7956 				IFNET_RUNLOCK();
7957 				return key_senderror(so, m, ENOBUFS);
7958 			}
7959 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7960 		}
7961 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7962 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7963 			    --cnt, mhp->msg->sadb_msg_pid, &sahtree_tracker);
7964 			if (n == NULL) {
7965 				SAHTREE_RUNLOCK();
7966 				IFNET_RUNLOCK();
7967 				return key_senderror(so, m, ENOBUFS);
7968 			}
7969 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7970 		}
7971 	}
7972 	SAHTREE_RUNLOCK();
7973 	IFNET_RUNLOCK();
7974 	m_freem(m);
7975 	return (0);
7976 }
7977 /*
7978  * SADB_X_PROMISC processing
7979  *
7980  * m will always be freed.
7981  */
7982 static int
7983 key_promisc(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7984 {
7985 	int olen;
7986 
7987 	IPSEC_ASSERT(so != NULL, ("null socket"));
7988 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7989 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7990 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7991 
7992 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7993 
7994 	if (olen < sizeof(struct sadb_msg)) {
7995 #if 1
7996 		return key_senderror(so, m, EINVAL);
7997 #else
7998 		m_freem(m);
7999 		return 0;
8000 #endif
8001 	} else if (olen == sizeof(struct sadb_msg)) {
8002 		/* enable/disable promisc mode */
8003 		struct keycb *kp;
8004 
8005 		if ((kp = so->so_pcb) == NULL)
8006 			return key_senderror(so, m, EINVAL);
8007 		mhp->msg->sadb_msg_errno = 0;
8008 		switch (mhp->msg->sadb_msg_satype) {
8009 		case 0:
8010 		case 1:
8011 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
8012 			break;
8013 		default:
8014 			return key_senderror(so, m, EINVAL);
8015 		}
8016 
8017 		/* send the original message back to everyone */
8018 		mhp->msg->sadb_msg_errno = 0;
8019 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
8020 	} else {
8021 		/* send packet as is */
8022 
8023 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
8024 
8025 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
8026 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
8027 	}
8028 }
8029 
8030 static int (*key_typesw[])(struct socket *, struct mbuf *,
8031     const struct sadb_msghdr *) = {
8032 	[SADB_RESERVED] =	NULL,
8033 	[SADB_GETSPI] =		key_getspi,
8034 	[SADB_UPDATE] =		key_update,
8035 	[SADB_ADD] =		key_add,
8036 	[SADB_DELETE] =		key_delete,
8037 	[SADB_GET] =		key_get,
8038 	[SADB_ACQUIRE] =	key_acquire2,
8039 	[SADB_REGISTER] =	key_register,
8040 	[SADB_EXPIRE] =		NULL,
8041 	[SADB_FLUSH] =		key_flush,
8042 	[SADB_DUMP] =		key_dump,
8043 	[SADB_X_PROMISC] =	key_promisc,
8044 	[SADB_X_PCHANGE] =	NULL,
8045 	[SADB_X_SPDUPDATE] =	key_spdadd,
8046 	[SADB_X_SPDADD] =	key_spdadd,
8047 	[SADB_X_SPDDELETE] =	key_spddelete,
8048 	[SADB_X_SPDGET] =	key_spdget,
8049 	[SADB_X_SPDACQUIRE] =	NULL,
8050 	[SADB_X_SPDDUMP] =	key_spddump,
8051 	[SADB_X_SPDFLUSH] =	key_spdflush,
8052 	[SADB_X_SPDSETIDX] =	key_spdadd,
8053 	[SADB_X_SPDEXPIRE] =	NULL,
8054 	[SADB_X_SPDDELETE2] =	key_spddelete2,
8055 };
8056 
8057 /*
8058  * parse sadb_msg buffer to process PFKEYv2,
8059  * and create a data to response if needed.
8060  * I think to be dealed with mbuf directly.
8061  * IN:
8062  *     msgp  : pointer to pointer to a received buffer pulluped.
8063  *             This is rewrited to response.
8064  *     so    : pointer to socket.
8065  * OUT:
8066  *    length for buffer to send to user process.
8067  */
8068 int
8069 key_parse(struct mbuf *m, struct socket *so)
8070 {
8071 	struct sadb_msg *msg;
8072 	struct sadb_msghdr mh;
8073 	u_int orglen;
8074 	int error;
8075 	int target;
8076 
8077 	IPSEC_ASSERT(so != NULL, ("null socket"));
8078 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
8079 
8080 	if (m->m_len < sizeof(struct sadb_msg)) {
8081 		m = m_pullup(m, sizeof(struct sadb_msg));
8082 		if (!m)
8083 			return ENOBUFS;
8084 	}
8085 	msg = mtod(m, struct sadb_msg *);
8086 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
8087 	target = KEY_SENDUP_ONE;
8088 
8089 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len != orglen) {
8090 		ipseclog((LOG_DEBUG, "%s: invalid message length.\n",__func__));
8091 		PFKEYSTAT_INC(out_invlen);
8092 		error = EINVAL;
8093 		goto senderror;
8094 	}
8095 
8096 	if (msg->sadb_msg_version != PF_KEY_V2) {
8097 		ipseclog((LOG_DEBUG, "%s: PF_KEY version %u is mismatched.\n",
8098 		    __func__, msg->sadb_msg_version));
8099 		PFKEYSTAT_INC(out_invver);
8100 		error = EINVAL;
8101 		goto senderror;
8102 	}
8103 
8104 	if (msg->sadb_msg_type > SADB_MAX) {
8105 		ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
8106 		    __func__, msg->sadb_msg_type));
8107 		PFKEYSTAT_INC(out_invmsgtype);
8108 		error = EINVAL;
8109 		goto senderror;
8110 	}
8111 
8112 	/* for old-fashioned code - should be nuked */
8113 	if (m->m_pkthdr.len > MCLBYTES) {
8114 		m_freem(m);
8115 		return ENOBUFS;
8116 	}
8117 	if (m->m_next) {
8118 		struct mbuf *n;
8119 
8120 		n = key_mget(m->m_pkthdr.len);
8121 		if (n == NULL) {
8122 			m_freem(m);
8123 			return ENOBUFS;
8124 		}
8125 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
8126 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
8127 		n->m_next = NULL;
8128 		m_freem(m);
8129 		m = n;
8130 	}
8131 
8132 	/* align the mbuf chain so that extensions are in contiguous region. */
8133 	error = key_align(m, &mh);
8134 	if (error)
8135 		return error;
8136 
8137 	msg = mh.msg;
8138 
8139 	/* We use satype as scope mask for spddump */
8140 	if (msg->sadb_msg_type == SADB_X_SPDDUMP) {
8141 		switch (msg->sadb_msg_satype) {
8142 		case IPSEC_POLICYSCOPE_ANY:
8143 		case IPSEC_POLICYSCOPE_GLOBAL:
8144 		case IPSEC_POLICYSCOPE_IFNET:
8145 		case IPSEC_POLICYSCOPE_PCB:
8146 			break;
8147 		default:
8148 			ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
8149 			    __func__, msg->sadb_msg_type));
8150 			PFKEYSTAT_INC(out_invsatype);
8151 			error = EINVAL;
8152 			goto senderror;
8153 		}
8154 	} else {
8155 		switch (msg->sadb_msg_satype) { /* check SA type */
8156 		case SADB_SATYPE_UNSPEC:
8157 			switch (msg->sadb_msg_type) {
8158 			case SADB_GETSPI:
8159 			case SADB_UPDATE:
8160 			case SADB_ADD:
8161 			case SADB_DELETE:
8162 			case SADB_GET:
8163 			case SADB_ACQUIRE:
8164 			case SADB_EXPIRE:
8165 				ipseclog((LOG_DEBUG, "%s: must specify satype "
8166 				    "when msg type=%u.\n", __func__,
8167 				    msg->sadb_msg_type));
8168 				PFKEYSTAT_INC(out_invsatype);
8169 				error = EINVAL;
8170 				goto senderror;
8171 			}
8172 			break;
8173 		case SADB_SATYPE_AH:
8174 		case SADB_SATYPE_ESP:
8175 		case SADB_X_SATYPE_IPCOMP:
8176 		case SADB_X_SATYPE_TCPSIGNATURE:
8177 			switch (msg->sadb_msg_type) {
8178 			case SADB_X_SPDADD:
8179 			case SADB_X_SPDDELETE:
8180 			case SADB_X_SPDGET:
8181 			case SADB_X_SPDFLUSH:
8182 			case SADB_X_SPDSETIDX:
8183 			case SADB_X_SPDUPDATE:
8184 			case SADB_X_SPDDELETE2:
8185 				ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
8186 				    __func__, msg->sadb_msg_type));
8187 				PFKEYSTAT_INC(out_invsatype);
8188 				error = EINVAL;
8189 				goto senderror;
8190 			}
8191 			break;
8192 		case SADB_SATYPE_RSVP:
8193 		case SADB_SATYPE_OSPFV2:
8194 		case SADB_SATYPE_RIPV2:
8195 		case SADB_SATYPE_MIP:
8196 			ipseclog((LOG_DEBUG, "%s: type %u isn't supported.\n",
8197 			    __func__, msg->sadb_msg_satype));
8198 			PFKEYSTAT_INC(out_invsatype);
8199 			error = EOPNOTSUPP;
8200 			goto senderror;
8201 		case 1:	/* XXX: What does it do? */
8202 			if (msg->sadb_msg_type == SADB_X_PROMISC)
8203 				break;
8204 			/*FALLTHROUGH*/
8205 		default:
8206 			ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
8207 			    __func__, msg->sadb_msg_satype));
8208 			PFKEYSTAT_INC(out_invsatype);
8209 			error = EINVAL;
8210 			goto senderror;
8211 		}
8212 	}
8213 
8214 	/* check field of upper layer protocol and address family */
8215 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL
8216 	 && mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
8217 		struct sadb_address *src0, *dst0;
8218 		u_int plen;
8219 
8220 		src0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_SRC]);
8221 		dst0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_DST]);
8222 
8223 		/* check upper layer protocol */
8224 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
8225 			ipseclog((LOG_DEBUG, "%s: upper layer protocol "
8226 				"mismatched.\n", __func__));
8227 			PFKEYSTAT_INC(out_invaddr);
8228 			error = EINVAL;
8229 			goto senderror;
8230 		}
8231 
8232 		/* check family */
8233 		if (PFKEY_ADDR_SADDR(src0)->sa_family !=
8234 		    PFKEY_ADDR_SADDR(dst0)->sa_family) {
8235 			ipseclog((LOG_DEBUG, "%s: address family mismatched.\n",
8236 				__func__));
8237 			PFKEYSTAT_INC(out_invaddr);
8238 			error = EINVAL;
8239 			goto senderror;
8240 		}
8241 		if (PFKEY_ADDR_SADDR(src0)->sa_len !=
8242 		    PFKEY_ADDR_SADDR(dst0)->sa_len) {
8243 			ipseclog((LOG_DEBUG, "%s: address struct size "
8244 				"mismatched.\n", __func__));
8245 			PFKEYSTAT_INC(out_invaddr);
8246 			error = EINVAL;
8247 			goto senderror;
8248 		}
8249 
8250 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
8251 		case AF_INET:
8252 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
8253 			    sizeof(struct sockaddr_in)) {
8254 				PFKEYSTAT_INC(out_invaddr);
8255 				error = EINVAL;
8256 				goto senderror;
8257 			}
8258 			break;
8259 		case AF_INET6:
8260 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
8261 			    sizeof(struct sockaddr_in6)) {
8262 				PFKEYSTAT_INC(out_invaddr);
8263 				error = EINVAL;
8264 				goto senderror;
8265 			}
8266 			break;
8267 		default:
8268 			ipseclog((LOG_DEBUG, "%s: unsupported address family\n",
8269 				__func__));
8270 			PFKEYSTAT_INC(out_invaddr);
8271 			error = EAFNOSUPPORT;
8272 			goto senderror;
8273 		}
8274 
8275 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
8276 		case AF_INET:
8277 			plen = sizeof(struct in_addr) << 3;
8278 			break;
8279 		case AF_INET6:
8280 			plen = sizeof(struct in6_addr) << 3;
8281 			break;
8282 		default:
8283 			plen = 0;	/*fool gcc*/
8284 			break;
8285 		}
8286 
8287 		/* check max prefix length */
8288 		if (src0->sadb_address_prefixlen > plen ||
8289 		    dst0->sadb_address_prefixlen > plen) {
8290 			ipseclog((LOG_DEBUG, "%s: illegal prefixlen.\n",
8291 				__func__));
8292 			PFKEYSTAT_INC(out_invaddr);
8293 			error = EINVAL;
8294 			goto senderror;
8295 		}
8296 
8297 		/*
8298 		 * prefixlen == 0 is valid because there can be a case when
8299 		 * all addresses are matched.
8300 		 */
8301 	}
8302 
8303 	if (msg->sadb_msg_type >= nitems(key_typesw) ||
8304 	    key_typesw[msg->sadb_msg_type] == NULL) {
8305 		PFKEYSTAT_INC(out_invmsgtype);
8306 		error = EINVAL;
8307 		goto senderror;
8308 	}
8309 
8310 	return (*key_typesw[msg->sadb_msg_type])(so, m, &mh);
8311 
8312 senderror:
8313 	msg->sadb_msg_errno = error;
8314 	return key_sendup_mbuf(so, m, target);
8315 }
8316 
8317 static int
8318 key_senderror(struct socket *so, struct mbuf *m, int code)
8319 {
8320 	struct sadb_msg *msg;
8321 
8322 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8323 		("mbuf too small, len %u", m->m_len));
8324 
8325 	msg = mtod(m, struct sadb_msg *);
8326 	msg->sadb_msg_errno = code;
8327 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
8328 }
8329 
8330 /*
8331  * set the pointer to each header into message buffer.
8332  * m will be freed on error.
8333  * XXX larger-than-MCLBYTES extension?
8334  */
8335 static int
8336 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
8337 {
8338 	struct mbuf *n;
8339 	struct sadb_ext *ext;
8340 	size_t off, end;
8341 	int extlen;
8342 	int toff;
8343 
8344 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
8345 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
8346 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
8347 		("mbuf too small, len %u", m->m_len));
8348 
8349 	/* initialize */
8350 	bzero(mhp, sizeof(*mhp));
8351 
8352 	mhp->msg = mtod(m, struct sadb_msg *);
8353 	mhp->ext[0] = (struct sadb_ext *)mhp->msg;	/*XXX backward compat */
8354 
8355 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
8356 	extlen = end;	/*just in case extlen is not updated*/
8357 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
8358 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
8359 		if (!n) {
8360 			/* m is already freed */
8361 			return ENOBUFS;
8362 		}
8363 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8364 
8365 		/* set pointer */
8366 		switch (ext->sadb_ext_type) {
8367 		case SADB_EXT_SA:
8368 		case SADB_EXT_ADDRESS_SRC:
8369 		case SADB_EXT_ADDRESS_DST:
8370 		case SADB_EXT_ADDRESS_PROXY:
8371 		case SADB_EXT_LIFETIME_CURRENT:
8372 		case SADB_EXT_LIFETIME_HARD:
8373 		case SADB_EXT_LIFETIME_SOFT:
8374 		case SADB_EXT_KEY_AUTH:
8375 		case SADB_EXT_KEY_ENCRYPT:
8376 		case SADB_EXT_IDENTITY_SRC:
8377 		case SADB_EXT_IDENTITY_DST:
8378 		case SADB_EXT_SENSITIVITY:
8379 		case SADB_EXT_PROPOSAL:
8380 		case SADB_EXT_SUPPORTED_AUTH:
8381 		case SADB_EXT_SUPPORTED_ENCRYPT:
8382 		case SADB_EXT_SPIRANGE:
8383 		case SADB_X_EXT_POLICY:
8384 		case SADB_X_EXT_SA2:
8385 		case SADB_X_EXT_NAT_T_TYPE:
8386 		case SADB_X_EXT_NAT_T_SPORT:
8387 		case SADB_X_EXT_NAT_T_DPORT:
8388 		case SADB_X_EXT_NAT_T_OAI:
8389 		case SADB_X_EXT_NAT_T_OAR:
8390 		case SADB_X_EXT_NAT_T_FRAG:
8391 		case SADB_X_EXT_SA_REPLAY:
8392 		case SADB_X_EXT_NEW_ADDRESS_SRC:
8393 		case SADB_X_EXT_NEW_ADDRESS_DST:
8394 #ifdef IPSEC_OFFLOAD
8395 		case SADB_X_EXT_LFT_CUR_SW_OFFL:
8396 		case SADB_X_EXT_LFT_CUR_HW_OFFL:
8397 		case SADB_X_EXT_IF_HW_OFFL:
8398 #endif
8399 			/* duplicate check */
8400 			/*
8401 			 * XXX Are there duplication payloads of either
8402 			 * KEY_AUTH or KEY_ENCRYPT ?
8403 			 */
8404 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
8405 				ipseclog((LOG_DEBUG, "%s: duplicate ext_type "
8406 					"%u\n", __func__, ext->sadb_ext_type));
8407 				m_freem(m);
8408 				PFKEYSTAT_INC(out_dupext);
8409 				return EINVAL;
8410 			}
8411 			break;
8412 		default:
8413 			ipseclog((LOG_DEBUG, "%s: invalid ext_type %u\n",
8414 				__func__, ext->sadb_ext_type));
8415 			m_freem(m);
8416 			PFKEYSTAT_INC(out_invexttype);
8417 			return EINVAL;
8418 		}
8419 
8420 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
8421 
8422 		if (key_validate_ext(ext, extlen)) {
8423 			m_freem(m);
8424 			PFKEYSTAT_INC(out_invlen);
8425 			return EINVAL;
8426 		}
8427 
8428 		n = m_pulldown(m, off, extlen, &toff);
8429 		if (!n) {
8430 			/* m is already freed */
8431 			return ENOBUFS;
8432 		}
8433 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
8434 
8435 		mhp->ext[ext->sadb_ext_type] = ext;
8436 		mhp->extoff[ext->sadb_ext_type] = off;
8437 		mhp->extlen[ext->sadb_ext_type] = extlen;
8438 	}
8439 
8440 	if (off != end) {
8441 		m_freem(m);
8442 		PFKEYSTAT_INC(out_invlen);
8443 		return EINVAL;
8444 	}
8445 
8446 	return 0;
8447 }
8448 
8449 static int
8450 key_validate_ext(const struct sadb_ext *ext, int len)
8451 {
8452 	const struct sockaddr *sa;
8453 	enum { NONE, ADDR } checktype = NONE;
8454 	int baselen = 0;
8455 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
8456 
8457 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
8458 		return EINVAL;
8459 
8460 	/* if it does not match minimum/maximum length, bail */
8461 	if (ext->sadb_ext_type >= nitems(minsize) ||
8462 	    ext->sadb_ext_type >= nitems(maxsize))
8463 		return EINVAL;
8464 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
8465 		return EINVAL;
8466 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
8467 		return EINVAL;
8468 
8469 	/* more checks based on sadb_ext_type XXX need more */
8470 	switch (ext->sadb_ext_type) {
8471 	case SADB_EXT_ADDRESS_SRC:
8472 	case SADB_EXT_ADDRESS_DST:
8473 	case SADB_EXT_ADDRESS_PROXY:
8474 	case SADB_X_EXT_NAT_T_OAI:
8475 	case SADB_X_EXT_NAT_T_OAR:
8476 	case SADB_X_EXT_NEW_ADDRESS_SRC:
8477 	case SADB_X_EXT_NEW_ADDRESS_DST:
8478 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
8479 		checktype = ADDR;
8480 		break;
8481 	case SADB_EXT_IDENTITY_SRC:
8482 	case SADB_EXT_IDENTITY_DST:
8483 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
8484 		    SADB_X_IDENTTYPE_ADDR) {
8485 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
8486 			checktype = ADDR;
8487 		} else
8488 			checktype = NONE;
8489 		break;
8490 	default:
8491 		checktype = NONE;
8492 		break;
8493 	}
8494 
8495 	switch (checktype) {
8496 	case NONE:
8497 		break;
8498 	case ADDR:
8499 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8500 		if (len < baselen + sal)
8501 			return EINVAL;
8502 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8503 			return EINVAL;
8504 		break;
8505 	}
8506 
8507 	return 0;
8508 }
8509 
8510 void
8511 spdcache_init(void)
8512 {
8513 	int i;
8514 
8515 	TUNABLE_INT_FETCH("net.key.spdcache.maxentries",
8516 	    &V_key_spdcache_maxentries);
8517 	TUNABLE_INT_FETCH("net.key.spdcache.threshold",
8518 	    &V_key_spdcache_threshold);
8519 
8520 	if (V_key_spdcache_maxentries) {
8521 		V_key_spdcache_maxentries = MAX(V_key_spdcache_maxentries,
8522 		    SPDCACHE_MAX_ENTRIES_PER_HASH);
8523 		V_spdcachehashtbl = hashinit(V_key_spdcache_maxentries /
8524 		    SPDCACHE_MAX_ENTRIES_PER_HASH,
8525 		    M_IPSEC_SPDCACHE, &V_spdcachehash_mask);
8526 		V_key_spdcache_maxentries = (V_spdcachehash_mask + 1)
8527 		    * SPDCACHE_MAX_ENTRIES_PER_HASH;
8528 
8529 		V_spdcache_lock = malloc(sizeof(struct mtx) *
8530 		    (V_spdcachehash_mask + 1),
8531 		    M_IPSEC_SPDCACHE, M_WAITOK | M_ZERO);
8532 
8533 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8534 			SPDCACHE_LOCK_INIT(i);
8535 	}
8536 }
8537 
8538 struct spdcache_entry *
8539 spdcache_entry_alloc(const struct secpolicyindex *spidx, struct secpolicy *sp)
8540 {
8541 	struct spdcache_entry *entry;
8542 
8543 	entry = malloc(sizeof(struct spdcache_entry), M_IPSEC_SPDCACHE,
8544 	    M_NOWAIT | M_ZERO);
8545 	if (entry == NULL)
8546 		return (NULL);
8547 
8548 	if (sp != NULL)
8549 		SP_ADDREF(sp);
8550 
8551 	entry->spidx = *spidx;
8552 	entry->sp = sp;
8553 
8554 	return (entry);
8555 }
8556 
8557 void
8558 spdcache_entry_free(struct spdcache_entry *entry)
8559 {
8560 
8561 	if (entry->sp != NULL)
8562 		key_freesp(&entry->sp);
8563 	free(entry, M_IPSEC_SPDCACHE);
8564 }
8565 
8566 void
8567 spdcache_clear(void)
8568 {
8569 	struct spdcache_entry *entry;
8570 	int i;
8571 
8572 	for (i = 0; i < V_spdcachehash_mask + 1; ++i) {
8573 		SPDCACHE_LOCK(i);
8574 		while (!LIST_EMPTY(&V_spdcachehashtbl[i])) {
8575 			entry = LIST_FIRST(&V_spdcachehashtbl[i]);
8576 			LIST_REMOVE(entry, chain);
8577 			spdcache_entry_free(entry);
8578 		}
8579 		SPDCACHE_UNLOCK(i);
8580 	}
8581 }
8582 
8583 #ifdef VIMAGE
8584 void
8585 spdcache_destroy(void)
8586 {
8587 	int i;
8588 
8589 	if (SPDCACHE_ENABLED()) {
8590 		spdcache_clear();
8591 		hashdestroy(V_spdcachehashtbl, M_IPSEC_SPDCACHE, V_spdcachehash_mask);
8592 
8593 		for (i = 0; i < V_spdcachehash_mask + 1; ++i)
8594 			SPDCACHE_LOCK_DESTROY(i);
8595 
8596 		free(V_spdcache_lock, M_IPSEC_SPDCACHE);
8597 	}
8598 }
8599 #endif
8600 
8601 static void
8602 key_vnet_init(void *arg __unused)
8603 {
8604 	int i;
8605 
8606 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8607 		TAILQ_INIT(&V_sptree[i]);
8608 		TAILQ_INIT(&V_sptree_ifnet[i]);
8609 	}
8610 
8611 	TAILQ_INIT(&V_sahtree);
8612 	V_sphashtbl = hashinit(SPHASH_NHASH, M_IPSEC_SP, &V_sphash_mask);
8613 	V_savhashtbl = hashinit(SAVHASH_NHASH, M_IPSEC_SA, &V_savhash_mask);
8614 	V_sahaddrhashtbl = hashinit(SAHHASH_NHASH, M_IPSEC_SAH,
8615 	    &V_sahaddrhash_mask);
8616 	V_acqaddrhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8617 	    &V_acqaddrhash_mask);
8618 	V_acqseqhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8619 	    &V_acqseqhash_mask);
8620 
8621 	spdcache_init();
8622 
8623 	for (i = 0; i <= SADB_SATYPE_MAX; i++)
8624 		LIST_INIT(&V_regtree[i]);
8625 
8626 	LIST_INIT(&V_acqtree);
8627 	LIST_INIT(&V_spacqtree);
8628 }
8629 VNET_SYSINIT(key_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_SECOND,
8630     key_vnet_init, NULL);
8631 
8632 static void
8633 key_init(void *arg __unused)
8634 {
8635 
8636 	ipsec_key_lft_zone = uma_zcreate("IPsec SA lft_c",
8637 	    sizeof(uint64_t) * 2, NULL, NULL, NULL, NULL,
8638 	    UMA_ALIGN_PTR, UMA_ZONE_PCPU);
8639 
8640 	SPTREE_LOCK_INIT();
8641 	REGTREE_LOCK_INIT();
8642 	SAHTREE_LOCK_INIT();
8643 	ACQ_LOCK_INIT();
8644 	SPACQ_LOCK_INIT();
8645 	SPI_ALLOC_LOCK_INIT();
8646 
8647 #ifndef IPSEC_DEBUG2
8648 	callout_init(&key_timer, 1);
8649 	callout_reset(&key_timer, hz, key_timehandler, NULL);
8650 #endif /*IPSEC_DEBUG2*/
8651 
8652 	/* initialize key statistics */
8653 	keystat.getspi_count = 1;
8654 
8655 	if (bootverbose)
8656 		printf("IPsec: Initialized Security Association Processing.\n");
8657 }
8658 SYSINIT(key_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, key_init, NULL);
8659 
8660 #ifdef VIMAGE
8661 static void
8662 key_vnet_destroy(void *arg __unused)
8663 {
8664 	struct secashead_queue sahdrainq;
8665 	struct secpolicy_queue drainq;
8666 	struct secpolicy *sp, *nextsp;
8667 	struct secacq *acq, *nextacq;
8668 	struct secspacq *spacq, *nextspacq;
8669 	struct secashead *sah;
8670 	struct secasvar *sav;
8671 	struct secreg *reg;
8672 	int i;
8673 
8674 	/*
8675 	 * XXX: can we just call free() for each object without
8676 	 * walking through safe way with releasing references?
8677 	 */
8678 	TAILQ_INIT(&drainq);
8679 	SPTREE_WLOCK();
8680 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8681 		TAILQ_CONCAT(&drainq, &V_sptree[i], chain);
8682 		TAILQ_CONCAT(&drainq, &V_sptree_ifnet[i], chain);
8683 	}
8684 	for (i = 0; i < V_sphash_mask + 1; i++)
8685 		LIST_INIT(&V_sphashtbl[i]);
8686 	SPTREE_WUNLOCK();
8687 	spdcache_destroy();
8688 
8689 	sp = TAILQ_FIRST(&drainq);
8690 	while (sp != NULL) {
8691 		nextsp = TAILQ_NEXT(sp, chain);
8692 		key_freesp(&sp);
8693 		sp = nextsp;
8694 	}
8695 
8696 	TAILQ_INIT(&sahdrainq);
8697 	SAHTREE_WLOCK();
8698 	TAILQ_CONCAT(&sahdrainq, &V_sahtree, chain);
8699 	for (i = 0; i < V_savhash_mask + 1; i++)
8700 		LIST_INIT(&V_savhashtbl[i]);
8701 	for (i = 0; i < V_sahaddrhash_mask + 1; i++)
8702 		LIST_INIT(&V_sahaddrhashtbl[i]);
8703 	TAILQ_FOREACH(sah, &sahdrainq, chain) {
8704 		sah->state = SADB_SASTATE_DEAD;
8705 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
8706 			sav->state = SADB_SASTATE_DEAD;
8707 			ipsec_accel_forget_sav(sav);
8708 		}
8709 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
8710 			sav->state = SADB_SASTATE_DEAD;
8711 			ipsec_accel_forget_sav(sav);
8712 		}
8713 	}
8714 	SAHTREE_WUNLOCK();
8715 
8716 	/* Wait for async work referencing this VNET to finish. */
8717 	ipsec_accel_sync();
8718 
8719 	key_freesah_flushed(&sahdrainq);
8720 	hashdestroy(V_sphashtbl, M_IPSEC_SP, V_sphash_mask);
8721 	hashdestroy(V_savhashtbl, M_IPSEC_SA, V_savhash_mask);
8722 	hashdestroy(V_sahaddrhashtbl, M_IPSEC_SAH, V_sahaddrhash_mask);
8723 
8724 	REGTREE_LOCK();
8725 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8726 		LIST_FOREACH(reg, &V_regtree[i], chain) {
8727 			if (__LIST_CHAINED(reg)) {
8728 				LIST_REMOVE(reg, chain);
8729 				free(reg, M_IPSEC_SAR);
8730 				break;
8731 			}
8732 		}
8733 	}
8734 	REGTREE_UNLOCK();
8735 
8736 	ACQ_LOCK();
8737 	acq = LIST_FIRST(&V_acqtree);
8738 	while (acq != NULL) {
8739 		nextacq = LIST_NEXT(acq, chain);
8740 		LIST_REMOVE(acq, chain);
8741 		free(acq, M_IPSEC_SAQ);
8742 		acq = nextacq;
8743 	}
8744 	for (i = 0; i < V_acqaddrhash_mask + 1; i++)
8745 		LIST_INIT(&V_acqaddrhashtbl[i]);
8746 	for (i = 0; i < V_acqseqhash_mask + 1; i++)
8747 		LIST_INIT(&V_acqseqhashtbl[i]);
8748 	ACQ_UNLOCK();
8749 
8750 	SPACQ_LOCK();
8751 	for (spacq = LIST_FIRST(&V_spacqtree); spacq != NULL;
8752 	    spacq = nextspacq) {
8753 		nextspacq = LIST_NEXT(spacq, chain);
8754 		if (__LIST_CHAINED(spacq)) {
8755 			LIST_REMOVE(spacq, chain);
8756 			free(spacq, M_IPSEC_SAQ);
8757 		}
8758 	}
8759 	SPACQ_UNLOCK();
8760 	hashdestroy(V_acqaddrhashtbl, M_IPSEC_SAQ, V_acqaddrhash_mask);
8761 	hashdestroy(V_acqseqhashtbl, M_IPSEC_SAQ, V_acqseqhash_mask);
8762 }
8763 VNET_SYSUNINIT(key_vnet_destroy, SI_SUB_PROTO_DOMAIN, SI_ORDER_SECOND,
8764     key_vnet_destroy, NULL);
8765 #endif
8766 
8767 /*
8768  * XXX: as long as domains are not unloadable, this function is never called,
8769  * provided for consistensy and future unload support.
8770  */
8771 static void
8772 key_destroy(void *arg __unused)
8773 {
8774 	uma_zdestroy(ipsec_key_lft_zone);
8775 
8776 #ifndef IPSEC_DEBUG2
8777 	callout_drain(&key_timer);
8778 #endif
8779 	SPTREE_LOCK_DESTROY();
8780 	REGTREE_LOCK_DESTROY();
8781 	SAHTREE_LOCK_DESTROY();
8782 	ACQ_LOCK_DESTROY();
8783 	SPACQ_LOCK_DESTROY();
8784 	SPI_ALLOC_LOCK_DESTROY();
8785 }
8786 SYSUNINIT(key_destroy, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, key_destroy, NULL);
8787 
8788 /* record data transfer on SA, and update timestamps */
8789 void
8790 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8791 {
8792 	IPSEC_ASSERT(sav != NULL, ("Null secasvar"));
8793 	IPSEC_ASSERT(m != NULL, ("Null mbuf"));
8794 
8795 	/*
8796 	 * XXX Currently, there is a difference of bytes size
8797 	 * between inbound and outbound processing.
8798 	 */
8799 	counter_u64_add(sav->lft_c_bytes, m->m_pkthdr.len);
8800 
8801 	/*
8802 	 * We use the number of packets as the unit of
8803 	 * allocations.  We increment the variable
8804 	 * whenever {esp,ah}_{in,out}put is called.
8805 	 */
8806 	counter_u64_add(sav->lft_c_allocations, 1);
8807 
8808 	/*
8809 	 * NOTE: We record CURRENT usetime by using wall clock,
8810 	 * in seconds.  HARD and SOFT lifetime are measured by the time
8811 	 * difference (again in seconds) from usetime.
8812 	 *
8813 	 *	usetime
8814 	 *	v     expire   expire
8815 	 * -----+-----+--------+---> t
8816 	 *	<--------------> HARD
8817 	 *	<-----> SOFT
8818 	 */
8819 	if (sav->firstused == 0)
8820 		sav->firstused = time_second;
8821 }
8822 
8823 /*
8824  * Take one of the kernel's security keys and convert it into a PF_KEY
8825  * structure within an mbuf, suitable for sending up to a waiting
8826  * application in user land.
8827  *
8828  * IN:
8829  *    src: A pointer to a kernel security key.
8830  *    exttype: Which type of key this is. Refer to the PF_KEY data structures.
8831  * OUT:
8832  *    a valid mbuf or NULL indicating an error
8833  *
8834  */
8835 
8836 static struct mbuf *
8837 key_setkey(struct seckey *src, uint16_t exttype)
8838 {
8839 	struct mbuf *m;
8840 	struct sadb_key *p;
8841 	int len;
8842 
8843 	if (src == NULL)
8844 		return NULL;
8845 
8846 	len = PFKEY_ALIGN8(sizeof(struct sadb_key) + _KEYLEN(src));
8847 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8848 	if (m == NULL)
8849 		return NULL;
8850 	m_align(m, len);
8851 	m->m_len = len;
8852 	p = mtod(m, struct sadb_key *);
8853 	bzero(p, len);
8854 	p->sadb_key_len = PFKEY_UNIT64(len);
8855 	p->sadb_key_exttype = exttype;
8856 	p->sadb_key_bits = src->bits;
8857 	bcopy(src->key_data, _KEYBUF(p), _KEYLEN(src));
8858 
8859 	return m;
8860 }
8861 
8862 #ifdef IPSEC_OFFLOAD
8863 struct mbuf *
8864 key_setaccelif(const char *ifname)
8865 {
8866 	struct mbuf *m = NULL;
8867 	struct sadb_x_if_hw_offl *p;
8868 	int len = PFKEY_ALIGN8(sizeof(*p));
8869 
8870 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8871 	if (m == NULL)
8872 		return (m);
8873 	m_align(m, len);
8874 	m->m_len = len;
8875 	p = mtod(m, struct sadb_x_if_hw_offl *);
8876 
8877 	bzero(p, len);
8878 	p->sadb_x_if_hw_offl_len = PFKEY_UNIT64(len);
8879 	p->sadb_x_if_hw_offl_exttype = SADB_X_EXT_IF_HW_OFFL;
8880 	p->sadb_x_if_hw_offl_flags = 0;
8881 	strncpy(p->sadb_x_if_hw_offl_if, ifname,
8882 	    sizeof(p->sadb_x_if_hw_offl_if));
8883 
8884 	return (m);
8885 }
8886 #endif
8887 
8888 /*
8889  * Take one of the kernel's lifetime data structures and convert it
8890  * into a PF_KEY structure within an mbuf, suitable for sending up to
8891  * a waiting application in user land.
8892  *
8893  * IN:
8894  *    src: A pointer to a kernel lifetime structure.
8895  *    exttype: Which type of lifetime this is. Refer to the PF_KEY
8896  *             data structures for more information.
8897  * OUT:
8898  *    a valid mbuf or NULL indicating an error
8899  *
8900  */
8901 
8902 static struct mbuf *
8903 key_setlifetime(struct seclifetime *src, uint16_t exttype)
8904 {
8905 	struct mbuf *m = NULL;
8906 	struct sadb_lifetime *p;
8907 	int len = PFKEY_ALIGN8(sizeof(struct sadb_lifetime));
8908 
8909 	if (src == NULL)
8910 		return NULL;
8911 
8912 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8913 	if (m == NULL)
8914 		return m;
8915 	m_align(m, len);
8916 	m->m_len = len;
8917 	p = mtod(m, struct sadb_lifetime *);
8918 
8919 	bzero(p, len);
8920 	p->sadb_lifetime_len = PFKEY_UNIT64(len);
8921 	p->sadb_lifetime_exttype = exttype;
8922 	p->sadb_lifetime_allocations = src->allocations;
8923 	p->sadb_lifetime_bytes = src->bytes;
8924 	p->sadb_lifetime_addtime = src->addtime;
8925 	p->sadb_lifetime_usetime = src->usetime;
8926 
8927 	return m;
8928 
8929 }
8930 
8931 const struct enc_xform *
8932 enc_algorithm_lookup(int alg)
8933 {
8934 	int i;
8935 
8936 	for (i = 0; i < nitems(supported_ealgs); i++)
8937 		if (alg == supported_ealgs[i].sadb_alg)
8938 			return (supported_ealgs[i].xform);
8939 	return (NULL);
8940 }
8941 
8942 const struct auth_hash *
8943 auth_algorithm_lookup(int alg)
8944 {
8945 	int i;
8946 
8947 	for (i = 0; i < nitems(supported_aalgs); i++)
8948 		if (alg == supported_aalgs[i].sadb_alg)
8949 			return (supported_aalgs[i].xform);
8950 	return (NULL);
8951 }
8952 
8953 const struct comp_algo *
8954 comp_algorithm_lookup(int alg)
8955 {
8956 	int i;
8957 
8958 	for (i = 0; i < nitems(supported_calgs); i++)
8959 		if (alg == supported_calgs[i].sadb_alg)
8960 			return (supported_calgs[i].xform);
8961 	return (NULL);
8962 }
8963 
8964 void
8965 ipsec_sahtree_runlock(struct rm_priotracker *sahtree_trackerp)
8966 {
8967 	rm_runlock(&sahtree_lock, sahtree_trackerp);
8968 }
8969 
8970 void
8971 ipsec_sahtree_rlock(struct rm_priotracker *sahtree_trackerp)
8972 {
8973 	rm_rlock(&sahtree_lock, sahtree_trackerp);
8974 }
8975 
8976 #ifdef IPSEC_OFFLOAD
8977 void
8978 ipsec_accel_on_ifdown(struct ifnet *ifp)
8979 {
8980 	void (*p)(struct ifnet *ifp);
8981 
8982 	p = atomic_load_ptr(&ipsec_accel_on_ifdown_p);
8983 	if (p != NULL)
8984 		p(ifp);
8985 }
8986 
8987 void
8988 ipsec_accel_drv_sa_lifetime_update(struct secasvar *sav, if_t ifp,
8989     u_int drv_spi, uint64_t octets, uint64_t allocs)
8990 {
8991 	void (*p)(struct secasvar *sav, if_t ifp, u_int drv_spi,
8992 	    uint64_t octets, uint64_t allocs);
8993 
8994 	p = atomic_load_ptr(&ipsec_accel_drv_sa_lifetime_update_p);
8995 	if (p != NULL)
8996 		p(sav, ifp, drv_spi, octets, allocs);
8997 }
8998 
8999 int
9000 ipsec_accel_drv_sa_lifetime_fetch(struct secasvar *sav,
9001     if_t ifp, u_int drv_spi, uint64_t *octets, uint64_t *allocs)
9002 {
9003 	int (*p)(struct secasvar *sav, if_t ifp, u_int drv_spi,
9004 	    uint64_t *octets, uint64_t *allocs);
9005 
9006 	p = atomic_load_ptr(&ipsec_accel_drv_sa_lifetime_fetch_p);
9007 	if (p == NULL)
9008 		return (EOPNOTSUPP);
9009 	return (p(sav, ifp, drv_spi, octets, allocs));
9010 }
9011 #endif
9012