xref: /freebsd/sys/netipsec/key.c (revision 8657387683946d0c03e09fe77029edfe309eeb20)
1 /*	$FreeBSD$	*/
2 /*	$KAME: key.c,v 1.191 2001/06/27 10:46:49 sakane Exp $	*/
3 
4 /*-
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * This code is referd to RFC 2367
35  */
36 
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39 #include "opt_ipsec.h"
40 
41 #include <sys/types.h>
42 #include <sys/param.h>
43 #include <sys/systm.h>
44 #include <sys/kernel.h>
45 #include <sys/fnv_hash.h>
46 #include <sys/lock.h>
47 #include <sys/mutex.h>
48 #include <sys/mbuf.h>
49 #include <sys/domain.h>
50 #include <sys/protosw.h>
51 #include <sys/malloc.h>
52 #include <sys/rmlock.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sysctl.h>
56 #include <sys/errno.h>
57 #include <sys/proc.h>
58 #include <sys/queue.h>
59 #include <sys/refcount.h>
60 #include <sys/syslog.h>
61 
62 #include <vm/uma.h>
63 
64 #include <net/if.h>
65 #include <net/if_var.h>
66 #include <net/vnet.h>
67 #include <net/raw_cb.h>
68 
69 #include <netinet/in.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/ip.h>
72 #include <netinet/in_var.h>
73 #include <netinet/udp.h>
74 
75 #ifdef INET6
76 #include <netinet/ip6.h>
77 #include <netinet6/in6_var.h>
78 #include <netinet6/ip6_var.h>
79 #endif /* INET6 */
80 
81 #include <net/pfkeyv2.h>
82 #include <netipsec/keydb.h>
83 #include <netipsec/key.h>
84 #include <netipsec/keysock.h>
85 #include <netipsec/key_debug.h>
86 
87 #include <netipsec/ipsec.h>
88 #ifdef INET6
89 #include <netipsec/ipsec6.h>
90 #endif
91 
92 #include <netipsec/xform.h>
93 #include <machine/in_cksum.h>
94 #include <machine/stdarg.h>
95 
96 /* randomness */
97 #include <sys/random.h>
98 
99 #define FULLMASK	0xff
100 #define	_BITS(bytes)	((bytes) << 3)
101 
102 /*
103  * Note on SA reference counting:
104  * - SAs that are not in DEAD state will have (total external reference + 1)
105  *   following value in reference count field.  they cannot be freed and are
106  *   referenced from SA header.
107  * - SAs that are in DEAD state will have (total external reference)
108  *   in reference count field.  they are ready to be freed.  reference from
109  *   SA header will be removed in key_delsav(), when the reference count
110  *   field hits 0 (= no external reference other than from SA header.
111  */
112 
113 VNET_DEFINE(u_int32_t, key_debug_level) = 0;
114 static VNET_DEFINE(u_int, key_spi_trycnt) = 1000;
115 static VNET_DEFINE(u_int32_t, key_spi_minval) = 0x100;
116 static VNET_DEFINE(u_int32_t, key_spi_maxval) = 0x0fffffff;	/* XXX */
117 static VNET_DEFINE(u_int32_t, policy_id) = 0;
118 /*interval to initialize randseed,1(m)*/
119 static VNET_DEFINE(u_int, key_int_random) = 60;
120 /* interval to expire acquiring, 30(s)*/
121 static VNET_DEFINE(u_int, key_larval_lifetime) = 30;
122 /* counter for blocking SADB_ACQUIRE.*/
123 static VNET_DEFINE(int, key_blockacq_count) = 10;
124 /* lifetime for blocking SADB_ACQUIRE.*/
125 static VNET_DEFINE(int, key_blockacq_lifetime) = 20;
126 /* preferred old sa rather than new sa.*/
127 static VNET_DEFINE(int, key_preferred_oldsa) = 1;
128 #define	V_key_spi_trycnt	VNET(key_spi_trycnt)
129 #define	V_key_spi_minval	VNET(key_spi_minval)
130 #define	V_key_spi_maxval	VNET(key_spi_maxval)
131 #define	V_policy_id		VNET(policy_id)
132 #define	V_key_int_random	VNET(key_int_random)
133 #define	V_key_larval_lifetime	VNET(key_larval_lifetime)
134 #define	V_key_blockacq_count	VNET(key_blockacq_count)
135 #define	V_key_blockacq_lifetime	VNET(key_blockacq_lifetime)
136 #define	V_key_preferred_oldsa	VNET(key_preferred_oldsa)
137 
138 static VNET_DEFINE(u_int32_t, acq_seq) = 0;
139 #define	V_acq_seq		VNET(acq_seq)
140 
141 static VNET_DEFINE(uint32_t, sp_genid) = 0;
142 #define	V_sp_genid		VNET(sp_genid)
143 
144 /* SPD */
145 TAILQ_HEAD(secpolicy_queue, secpolicy);
146 LIST_HEAD(secpolicy_list, secpolicy);
147 static VNET_DEFINE(struct secpolicy_queue, sptree[IPSEC_DIR_MAX]);
148 static VNET_DEFINE(struct secpolicy_queue, sptree_ifnet[IPSEC_DIR_MAX]);
149 static struct rmlock sptree_lock;
150 #define	V_sptree		VNET(sptree)
151 #define	V_sptree_ifnet		VNET(sptree_ifnet)
152 #define	SPTREE_LOCK_INIT()      rm_init(&sptree_lock, "sptree")
153 #define	SPTREE_LOCK_DESTROY()   rm_destroy(&sptree_lock)
154 #define	SPTREE_RLOCK_TRACKER    struct rm_priotracker sptree_tracker
155 #define	SPTREE_RLOCK()          rm_rlock(&sptree_lock, &sptree_tracker)
156 #define	SPTREE_RUNLOCK()        rm_runlock(&sptree_lock, &sptree_tracker)
157 #define	SPTREE_RLOCK_ASSERT()   rm_assert(&sptree_lock, RA_RLOCKED)
158 #define	SPTREE_WLOCK()          rm_wlock(&sptree_lock)
159 #define	SPTREE_WUNLOCK()        rm_wunlock(&sptree_lock)
160 #define	SPTREE_WLOCK_ASSERT()   rm_assert(&sptree_lock, RA_WLOCKED)
161 #define	SPTREE_UNLOCK_ASSERT()  rm_assert(&sptree_lock, RA_UNLOCKED)
162 
163 /* Hash table for lookup SP using unique id */
164 static VNET_DEFINE(struct secpolicy_list *, sphashtbl);
165 static VNET_DEFINE(u_long, sphash_mask);
166 #define	V_sphashtbl		VNET(sphashtbl)
167 #define	V_sphash_mask		VNET(sphash_mask)
168 
169 #define	SPHASH_NHASH_LOG2	7
170 #define	SPHASH_NHASH		(1 << SPHASH_NHASH_LOG2)
171 #define	SPHASH_HASHVAL(id)	(key_u32hash(id) & V_sphash_mask)
172 #define	SPHASH_HASH(id)		&V_sphashtbl[SPHASH_HASHVAL(id)]
173 
174 /* SAD */
175 TAILQ_HEAD(secashead_queue, secashead);
176 LIST_HEAD(secashead_list, secashead);
177 static VNET_DEFINE(struct secashead_queue, sahtree);
178 static struct rmlock sahtree_lock;
179 #define	V_sahtree		VNET(sahtree)
180 #define	SAHTREE_LOCK_INIT()	rm_init(&sahtree_lock, "sahtree")
181 #define	SAHTREE_LOCK_DESTROY()	rm_destroy(&sahtree_lock)
182 #define	SAHTREE_RLOCK_TRACKER	struct rm_priotracker sahtree_tracker
183 #define	SAHTREE_RLOCK()		rm_rlock(&sahtree_lock, &sahtree_tracker)
184 #define	SAHTREE_RUNLOCK()	rm_runlock(&sahtree_lock, &sahtree_tracker)
185 #define	SAHTREE_RLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_RLOCKED)
186 #define	SAHTREE_WLOCK()		rm_wlock(&sahtree_lock)
187 #define	SAHTREE_WUNLOCK()	rm_wunlock(&sahtree_lock)
188 #define	SAHTREE_WLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_WLOCKED)
189 #define	SAHTREE_UNLOCK_ASSERT()	rm_assert(&sahtree_lock, RA_UNLOCKED)
190 
191 /* Hash table for lookup in SAD using SA addresses */
192 static VNET_DEFINE(struct secashead_list *, sahaddrhashtbl);
193 static VNET_DEFINE(u_long, sahaddrhash_mask);
194 #define	V_sahaddrhashtbl	VNET(sahaddrhashtbl)
195 #define	V_sahaddrhash_mask	VNET(sahaddrhash_mask)
196 
197 #define	SAHHASH_NHASH_LOG2	7
198 #define	SAHHASH_NHASH		(1 << SAHHASH_NHASH_LOG2)
199 #define	SAHADDRHASH_HASHVAL(saidx)	\
200     (key_saidxhash(saidx) & V_sahaddrhash_mask)
201 #define	SAHADDRHASH_HASH(saidx)		\
202     &V_sahaddrhashtbl[SAHADDRHASH_HASHVAL(saidx)]
203 
204 /* Hash table for lookup in SAD using SPI */
205 LIST_HEAD(secasvar_list, secasvar);
206 static VNET_DEFINE(struct secasvar_list *, savhashtbl);
207 static VNET_DEFINE(u_long, savhash_mask);
208 #define	V_savhashtbl		VNET(savhashtbl)
209 #define	V_savhash_mask		VNET(savhash_mask)
210 #define	SAVHASH_NHASH_LOG2	7
211 #define	SAVHASH_NHASH		(1 << SAVHASH_NHASH_LOG2)
212 #define	SAVHASH_HASHVAL(spi)	(key_u32hash(spi) & V_savhash_mask)
213 #define	SAVHASH_HASH(spi)	&V_savhashtbl[SAVHASH_HASHVAL(spi)]
214 
215 static uint32_t
216 key_saidxhash(const struct secasindex *saidx)
217 {
218 	uint32_t hval;
219 
220 	hval = fnv_32_buf(&saidx->proto, sizeof(saidx->proto),
221 	    FNV1_32_INIT);
222 	switch (saidx->dst.sa.sa_family) {
223 #ifdef INET
224 	case AF_INET:
225 		hval = fnv_32_buf(&saidx->src.sin.sin_addr,
226 		    sizeof(in_addr_t), hval);
227 		hval = fnv_32_buf(&saidx->dst.sin.sin_addr,
228 		    sizeof(in_addr_t), hval);
229 		break;
230 #endif
231 #ifdef INET6
232 	case AF_INET6:
233 		hval = fnv_32_buf(&saidx->src.sin6.sin6_addr,
234 		    sizeof(struct in6_addr), hval);
235 		hval = fnv_32_buf(&saidx->dst.sin6.sin6_addr,
236 		    sizeof(struct in6_addr), hval);
237 		break;
238 #endif
239 	default:
240 		hval = 0;
241 		ipseclog((LOG_DEBUG, "%s: unknown address family %d",
242 		    __func__, saidx->dst.sa.sa_family));
243 	}
244 	return (hval);
245 }
246 
247 static uint32_t
248 key_u32hash(uint32_t val)
249 {
250 
251 	return (fnv_32_buf(&val, sizeof(val), FNV1_32_INIT));
252 }
253 
254 							/* registed list */
255 static VNET_DEFINE(LIST_HEAD(_regtree, secreg), regtree[SADB_SATYPE_MAX + 1]);
256 #define	V_regtree		VNET(regtree)
257 static struct mtx regtree_lock;
258 #define	REGTREE_LOCK_INIT() \
259 	mtx_init(&regtree_lock, "regtree", "fast ipsec regtree", MTX_DEF)
260 #define	REGTREE_LOCK_DESTROY()	mtx_destroy(&regtree_lock)
261 #define	REGTREE_LOCK()		mtx_lock(&regtree_lock)
262 #define	REGTREE_UNLOCK()	mtx_unlock(&regtree_lock)
263 #define	REGTREE_LOCK_ASSERT()	mtx_assert(&regtree_lock, MA_OWNED)
264 
265 /* Acquiring list */
266 LIST_HEAD(secacq_list, secacq);
267 static VNET_DEFINE(struct secacq_list, acqtree);
268 #define	V_acqtree		VNET(acqtree)
269 static struct mtx acq_lock;
270 #define	ACQ_LOCK_INIT() \
271     mtx_init(&acq_lock, "acqtree", "ipsec SA acquiring list", MTX_DEF)
272 #define	ACQ_LOCK_DESTROY()	mtx_destroy(&acq_lock)
273 #define	ACQ_LOCK()		mtx_lock(&acq_lock)
274 #define	ACQ_UNLOCK()		mtx_unlock(&acq_lock)
275 #define	ACQ_LOCK_ASSERT()	mtx_assert(&acq_lock, MA_OWNED)
276 
277 /* Hash table for lookup in ACQ list using SA addresses */
278 static VNET_DEFINE(struct secacq_list *, acqaddrhashtbl);
279 static VNET_DEFINE(u_long, acqaddrhash_mask);
280 #define	V_acqaddrhashtbl	VNET(acqaddrhashtbl)
281 #define	V_acqaddrhash_mask	VNET(acqaddrhash_mask)
282 
283 /* Hash table for lookup in ACQ list using SEQ number */
284 static VNET_DEFINE(struct secacq_list *, acqseqhashtbl);
285 static VNET_DEFINE(u_long, acqseqhash_mask);
286 #define	V_acqseqhashtbl		VNET(acqseqhashtbl)
287 #define	V_acqseqhash_mask	VNET(acqseqhash_mask)
288 
289 #define	ACQHASH_NHASH_LOG2	7
290 #define	ACQHASH_NHASH		(1 << ACQHASH_NHASH_LOG2)
291 #define	ACQADDRHASH_HASHVAL(saidx)	\
292     (key_saidxhash(saidx) & V_acqaddrhash_mask)
293 #define	ACQSEQHASH_HASHVAL(seq)		\
294     (key_u32hash(seq) & V_acqseqhash_mask)
295 #define	ACQADDRHASH_HASH(saidx)	\
296     &V_acqaddrhashtbl[ACQADDRHASH_HASHVAL(saidx)]
297 #define	ACQSEQHASH_HASH(seq)	\
298     &V_acqseqhashtbl[ACQSEQHASH_HASHVAL(seq)]
299 							/* SP acquiring list */
300 static VNET_DEFINE(LIST_HEAD(_spacqtree, secspacq), spacqtree);
301 #define	V_spacqtree		VNET(spacqtree)
302 static struct mtx spacq_lock;
303 #define	SPACQ_LOCK_INIT() \
304 	mtx_init(&spacq_lock, "spacqtree", \
305 		"fast ipsec security policy acquire list", MTX_DEF)
306 #define	SPACQ_LOCK_DESTROY()	mtx_destroy(&spacq_lock)
307 #define	SPACQ_LOCK()		mtx_lock(&spacq_lock)
308 #define	SPACQ_UNLOCK()		mtx_unlock(&spacq_lock)
309 #define	SPACQ_LOCK_ASSERT()	mtx_assert(&spacq_lock, MA_OWNED)
310 
311 static const int minsize[] = {
312 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
313 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
314 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
315 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
316 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
317 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_SRC */
318 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_DST */
319 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_PROXY */
320 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_AUTH */
321 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_ENCRYPT */
322 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_SRC */
323 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_DST */
324 	sizeof(struct sadb_sens),	/* SADB_EXT_SENSITIVITY */
325 	sizeof(struct sadb_prop),	/* SADB_EXT_PROPOSAL */
326 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_AUTH */
327 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_ENCRYPT */
328 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
329 	0,				/* SADB_X_EXT_KMPRIVATE */
330 	sizeof(struct sadb_x_policy),	/* SADB_X_EXT_POLICY */
331 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
332 	sizeof(struct sadb_x_nat_t_type),/* SADB_X_EXT_NAT_T_TYPE */
333 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_SPORT */
334 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_DPORT */
335 	sizeof(struct sadb_address),	/* SADB_X_EXT_NAT_T_OAI */
336 	sizeof(struct sadb_address),	/* SADB_X_EXT_NAT_T_OAR */
337 	sizeof(struct sadb_x_nat_t_frag),/* SADB_X_EXT_NAT_T_FRAG */
338 	sizeof(struct sadb_x_sa_replay), /* SADB_X_EXT_SA_REPLAY */
339 	sizeof(struct sadb_address),	/* SADB_X_EXT_NEW_ADDRESS_SRC */
340 	sizeof(struct sadb_address),	/* SADB_X_EXT_NEW_ADDRESS_DST */
341 };
342 _Static_assert(sizeof(minsize)/sizeof(int) == SADB_EXT_MAX + 1, "minsize size mismatch");
343 
344 static const int maxsize[] = {
345 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
346 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
347 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
348 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
349 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
350 	0,				/* SADB_EXT_ADDRESS_SRC */
351 	0,				/* SADB_EXT_ADDRESS_DST */
352 	0,				/* SADB_EXT_ADDRESS_PROXY */
353 	0,				/* SADB_EXT_KEY_AUTH */
354 	0,				/* SADB_EXT_KEY_ENCRYPT */
355 	0,				/* SADB_EXT_IDENTITY_SRC */
356 	0,				/* SADB_EXT_IDENTITY_DST */
357 	0,				/* SADB_EXT_SENSITIVITY */
358 	0,				/* SADB_EXT_PROPOSAL */
359 	0,				/* SADB_EXT_SUPPORTED_AUTH */
360 	0,				/* SADB_EXT_SUPPORTED_ENCRYPT */
361 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
362 	0,				/* SADB_X_EXT_KMPRIVATE */
363 	0,				/* SADB_X_EXT_POLICY */
364 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
365 	sizeof(struct sadb_x_nat_t_type),/* SADB_X_EXT_NAT_T_TYPE */
366 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_SPORT */
367 	sizeof(struct sadb_x_nat_t_port),/* SADB_X_EXT_NAT_T_DPORT */
368 	0,				/* SADB_X_EXT_NAT_T_OAI */
369 	0,				/* SADB_X_EXT_NAT_T_OAR */
370 	sizeof(struct sadb_x_nat_t_frag),/* SADB_X_EXT_NAT_T_FRAG */
371 	sizeof(struct sadb_x_sa_replay), /* SADB_X_EXT_SA_REPLAY */
372 	0,				/* SADB_X_EXT_NEW_ADDRESS_SRC */
373 	0,				/* SADB_X_EXT_NEW_ADDRESS_DST */
374 };
375 _Static_assert(sizeof(maxsize)/sizeof(int) == SADB_EXT_MAX + 1, "minsize size mismatch");
376 
377 /*
378  * Internal values for SA flags:
379  * SADB_X_EXT_F_CLONED means that SA was cloned by key_updateaddresses,
380  *	thus we will not free the most of SA content in key_delsav().
381  */
382 #define	SADB_X_EXT_F_CLONED	0x80000000
383 
384 #define	SADB_CHECKLEN(_mhp, _ext)			\
385     ((_mhp)->extlen[(_ext)] < minsize[(_ext)] || (maxsize[(_ext)] != 0 && \
386 	((_mhp)->extlen[(_ext)] > maxsize[(_ext)])))
387 #define	SADB_CHECKHDR(_mhp, _ext)	((_mhp)->ext[(_ext)] == NULL)
388 
389 static VNET_DEFINE(int, ipsec_esp_keymin) = 256;
390 static VNET_DEFINE(int, ipsec_esp_auth) = 0;
391 static VNET_DEFINE(int, ipsec_ah_keymin) = 128;
392 
393 #define	V_ipsec_esp_keymin	VNET(ipsec_esp_keymin)
394 #define	V_ipsec_esp_auth	VNET(ipsec_esp_auth)
395 #define	V_ipsec_ah_keymin	VNET(ipsec_ah_keymin)
396 
397 #ifdef IPSEC_DEBUG
398 VNET_DEFINE(int, ipsec_debug) = 1;
399 #else
400 VNET_DEFINE(int, ipsec_debug) = 0;
401 #endif
402 
403 #ifdef INET
404 SYSCTL_DECL(_net_inet_ipsec);
405 SYSCTL_INT(_net_inet_ipsec, IPSECCTL_DEBUG, debug,
406     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
407     "Enable IPsec debugging output when set.");
408 #endif
409 #ifdef INET6
410 SYSCTL_DECL(_net_inet6_ipsec6);
411 SYSCTL_INT(_net_inet6_ipsec6, IPSECCTL_DEBUG, debug,
412     CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_debug), 0,
413     "Enable IPsec debugging output when set.");
414 #endif
415 
416 SYSCTL_DECL(_net_key);
417 SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,
418 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_debug_level), 0, "");
419 
420 /* max count of trial for the decision of spi value */
421 SYSCTL_INT(_net_key, KEYCTL_SPI_TRY, spi_trycnt,
422 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_trycnt), 0, "");
423 
424 /* minimum spi value to allocate automatically. */
425 SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE, spi_minval,
426 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_minval), 0, "");
427 
428 /* maximun spi value to allocate automatically. */
429 SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE, spi_maxval,
430 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_spi_maxval), 0, "");
431 
432 /* interval to initialize randseed */
433 SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT, int_random,
434 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_int_random), 0, "");
435 
436 /* lifetime for larval SA */
437 SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME, larval_lifetime,
438 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_larval_lifetime), 0, "");
439 
440 /* counter for blocking to send SADB_ACQUIRE to IKEd */
441 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT, blockacq_count,
442 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_count), 0, "");
443 
444 /* lifetime for blocking to send SADB_ACQUIRE to IKEd */
445 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME, blockacq_lifetime,
446 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_blockacq_lifetime), 0, "");
447 
448 /* ESP auth */
449 SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH, esp_auth,
450 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_auth), 0, "");
451 
452 /* minimum ESP key length */
453 SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN, esp_keymin,
454 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_esp_keymin), 0, "");
455 
456 /* minimum AH key length */
457 SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN, ah_keymin,
458 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipsec_ah_keymin), 0, "");
459 
460 /* perfered old SA rather than new SA */
461 SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA, preferred_oldsa,
462 	CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(key_preferred_oldsa), 0, "");
463 
464 #define __LIST_CHAINED(elm) \
465 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
466 
467 MALLOC_DEFINE(M_IPSEC_SA, "secasvar", "ipsec security association");
468 MALLOC_DEFINE(M_IPSEC_SAH, "sahead", "ipsec sa head");
469 MALLOC_DEFINE(M_IPSEC_SP, "ipsecpolicy", "ipsec security policy");
470 MALLOC_DEFINE(M_IPSEC_SR, "ipsecrequest", "ipsec security request");
471 MALLOC_DEFINE(M_IPSEC_MISC, "ipsec-misc", "ipsec miscellaneous");
472 MALLOC_DEFINE(M_IPSEC_SAQ, "ipsec-saq", "ipsec sa acquire");
473 MALLOC_DEFINE(M_IPSEC_SAR, "ipsec-reg", "ipsec sa acquire");
474 
475 static VNET_DEFINE(uma_zone_t, key_lft_zone);
476 #define	V_key_lft_zone		VNET(key_lft_zone)
477 
478 static LIST_HEAD(xforms_list, xformsw) xforms = LIST_HEAD_INITIALIZER();
479 static struct mtx xforms_lock;
480 #define	XFORMS_LOCK_INIT()	\
481     mtx_init(&xforms_lock, "xforms_list", "IPsec transforms list", MTX_DEF)
482 #define	XFORMS_LOCK_DESTROY()	mtx_destroy(&xforms_lock)
483 #define	XFORMS_LOCK()		mtx_lock(&xforms_lock)
484 #define	XFORMS_UNLOCK()		mtx_unlock(&xforms_lock)
485 
486 /*
487  * set parameters into secpolicyindex buffer.
488  * Must allocate secpolicyindex buffer passed to this function.
489  */
490 #define KEY_SETSECSPIDX(_dir, s, d, ps, pd, ulp, idx) \
491 do { \
492 	bzero((idx), sizeof(struct secpolicyindex));                         \
493 	(idx)->dir = (_dir);                                                 \
494 	(idx)->prefs = (ps);                                                 \
495 	(idx)->prefd = (pd);                                                 \
496 	(idx)->ul_proto = (ulp);                                             \
497 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
498 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
499 } while (0)
500 
501 /*
502  * set parameters into secasindex buffer.
503  * Must allocate secasindex buffer before calling this function.
504  */
505 #define KEY_SETSECASIDX(p, m, r, s, d, idx) \
506 do { \
507 	bzero((idx), sizeof(struct secasindex));                             \
508 	(idx)->proto = (p);                                                  \
509 	(idx)->mode = (m);                                                   \
510 	(idx)->reqid = (r);                                                  \
511 	bcopy((s), &(idx)->src, ((const struct sockaddr *)(s))->sa_len);     \
512 	bcopy((d), &(idx)->dst, ((const struct sockaddr *)(d))->sa_len);     \
513 	key_porttosaddr(&(idx)->src.sa, 0);				     \
514 	key_porttosaddr(&(idx)->dst.sa, 0);				     \
515 } while (0)
516 
517 /* key statistics */
518 struct _keystat {
519 	u_long getspi_count; /* the avarage of count to try to get new SPI */
520 } keystat;
521 
522 struct sadb_msghdr {
523 	struct sadb_msg *msg;
524 	struct sadb_ext *ext[SADB_EXT_MAX + 1];
525 	int extoff[SADB_EXT_MAX + 1];
526 	int extlen[SADB_EXT_MAX + 1];
527 };
528 
529 static struct supported_ealgs {
530 	int sadb_alg;
531 	const struct enc_xform *xform;
532 } supported_ealgs[] = {
533 	{ SADB_EALG_DESCBC,		&enc_xform_des },
534 	{ SADB_EALG_3DESCBC,		&enc_xform_3des },
535 	{ SADB_X_EALG_AES,		&enc_xform_rijndael128 },
536 	{ SADB_X_EALG_BLOWFISHCBC,	&enc_xform_blf },
537 	{ SADB_X_EALG_CAST128CBC,	&enc_xform_cast5 },
538 	{ SADB_EALG_NULL,		&enc_xform_null },
539 	{ SADB_X_EALG_CAMELLIACBC,	&enc_xform_camellia },
540 	{ SADB_X_EALG_AESCTR,		&enc_xform_aes_icm },
541 	{ SADB_X_EALG_AESGCM16,		&enc_xform_aes_nist_gcm },
542 	{ SADB_X_EALG_AESGMAC,		&enc_xform_aes_nist_gmac },
543 };
544 
545 static struct supported_aalgs {
546 	int sadb_alg;
547 	const struct auth_hash *xform;
548 } supported_aalgs[] = {
549 	{ SADB_X_AALG_NULL,		&auth_hash_null },
550 	{ SADB_AALG_MD5HMAC,		&auth_hash_hmac_md5 },
551 	{ SADB_AALG_SHA1HMAC,		&auth_hash_hmac_sha1 },
552 	{ SADB_X_AALG_RIPEMD160HMAC,	&auth_hash_hmac_ripemd_160 },
553 	{ SADB_X_AALG_MD5,		&auth_hash_key_md5 },
554 	{ SADB_X_AALG_SHA,		&auth_hash_key_sha1 },
555 	{ SADB_X_AALG_SHA2_256,		&auth_hash_hmac_sha2_256 },
556 	{ SADB_X_AALG_SHA2_384,		&auth_hash_hmac_sha2_384 },
557 	{ SADB_X_AALG_SHA2_512,		&auth_hash_hmac_sha2_512 },
558 	{ SADB_X_AALG_AES128GMAC,	&auth_hash_nist_gmac_aes_128 },
559 	{ SADB_X_AALG_AES192GMAC,	&auth_hash_nist_gmac_aes_192 },
560 	{ SADB_X_AALG_AES256GMAC,	&auth_hash_nist_gmac_aes_256 },
561 };
562 
563 static struct supported_calgs {
564 	int sadb_alg;
565 	const struct comp_algo *xform;
566 } supported_calgs[] = {
567 	{ SADB_X_CALG_DEFLATE,		&comp_algo_deflate },
568 };
569 
570 #ifndef IPSEC_DEBUG2
571 static struct callout key_timer;
572 #endif
573 
574 static void key_unlink(struct secpolicy *);
575 static struct secpolicy *key_getsp(struct secpolicyindex *);
576 static struct secpolicy *key_getspbyid(u_int32_t);
577 static struct mbuf *key_gather_mbuf(struct mbuf *,
578 	const struct sadb_msghdr *, int, int, ...);
579 static int key_spdadd(struct socket *, struct mbuf *,
580 	const struct sadb_msghdr *);
581 static uint32_t key_getnewspid(void);
582 static int key_spddelete(struct socket *, struct mbuf *,
583 	const struct sadb_msghdr *);
584 static int key_spddelete2(struct socket *, struct mbuf *,
585 	const struct sadb_msghdr *);
586 static int key_spdget(struct socket *, struct mbuf *,
587 	const struct sadb_msghdr *);
588 static int key_spdflush(struct socket *, struct mbuf *,
589 	const struct sadb_msghdr *);
590 static int key_spddump(struct socket *, struct mbuf *,
591 	const struct sadb_msghdr *);
592 static struct mbuf *key_setdumpsp(struct secpolicy *,
593 	u_int8_t, u_int32_t, u_int32_t);
594 static struct mbuf *key_sp2mbuf(struct secpolicy *);
595 static size_t key_getspreqmsglen(struct secpolicy *);
596 static int key_spdexpire(struct secpolicy *);
597 static struct secashead *key_newsah(struct secasindex *);
598 static void key_freesah(struct secashead **);
599 static void key_delsah(struct secashead *);
600 static struct secasvar *key_newsav(const struct sadb_msghdr *,
601     struct secasindex *, uint32_t, int *);
602 static void key_delsav(struct secasvar *);
603 static void key_unlinksav(struct secasvar *);
604 static struct secashead *key_getsah(struct secasindex *);
605 static int key_checkspidup(uint32_t);
606 static struct secasvar *key_getsavbyspi(uint32_t);
607 static int key_setnatt(struct secasvar *, const struct sadb_msghdr *);
608 static int key_setsaval(struct secasvar *, const struct sadb_msghdr *);
609 static int key_updatelifetimes(struct secasvar *, const struct sadb_msghdr *);
610 static int key_updateaddresses(struct socket *, struct mbuf *,
611     const struct sadb_msghdr *, struct secasvar *, struct secasindex *);
612 
613 static struct mbuf *key_setdumpsa(struct secasvar *, u_int8_t,
614 	u_int8_t, u_int32_t, u_int32_t);
615 static struct mbuf *key_setsadbmsg(u_int8_t, u_int16_t, u_int8_t,
616 	u_int32_t, pid_t, u_int16_t);
617 static struct mbuf *key_setsadbsa(struct secasvar *);
618 static struct mbuf *key_setsadbaddr(u_int16_t,
619 	const struct sockaddr *, u_int8_t, u_int16_t);
620 static struct mbuf *key_setsadbxport(u_int16_t, u_int16_t);
621 static struct mbuf *key_setsadbxtype(u_int16_t);
622 static struct mbuf *key_setsadbxsa2(u_int8_t, u_int32_t, u_int32_t);
623 static struct mbuf *key_setsadbxsareplay(u_int32_t);
624 static struct mbuf *key_setsadbxpolicy(u_int16_t, u_int8_t,
625 	u_int32_t, u_int32_t);
626 static struct seckey *key_dup_keymsg(const struct sadb_key *, size_t,
627     struct malloc_type *);
628 static struct seclifetime *key_dup_lifemsg(const struct sadb_lifetime *src,
629     struct malloc_type *);
630 
631 /* flags for key_cmpsaidx() */
632 #define CMP_HEAD	1	/* protocol, addresses. */
633 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
634 #define CMP_REQID	3	/* additionally HEAD, reaid. */
635 #define CMP_EXACTLY	4	/* all elements. */
636 static int key_cmpsaidx(const struct secasindex *,
637     const struct secasindex *, int);
638 static int key_cmpspidx_exactly(struct secpolicyindex *,
639     struct secpolicyindex *);
640 static int key_cmpspidx_withmask(struct secpolicyindex *,
641     struct secpolicyindex *);
642 static int key_bbcmp(const void *, const void *, u_int);
643 static uint8_t key_satype2proto(uint8_t);
644 static uint8_t key_proto2satype(uint8_t);
645 
646 static int key_getspi(struct socket *, struct mbuf *,
647 	const struct sadb_msghdr *);
648 static uint32_t key_do_getnewspi(struct sadb_spirange *, struct secasindex *);
649 static int key_update(struct socket *, struct mbuf *,
650 	const struct sadb_msghdr *);
651 static int key_add(struct socket *, struct mbuf *,
652 	const struct sadb_msghdr *);
653 static int key_setident(struct secashead *, const struct sadb_msghdr *);
654 static struct mbuf *key_getmsgbuf_x1(struct mbuf *,
655 	const struct sadb_msghdr *);
656 static int key_delete(struct socket *, struct mbuf *,
657 	const struct sadb_msghdr *);
658 static int key_delete_all(struct socket *, struct mbuf *,
659 	const struct sadb_msghdr *, struct secasindex *);
660 static void key_delete_xform(const struct xformsw *);
661 static int key_get(struct socket *, struct mbuf *,
662 	const struct sadb_msghdr *);
663 
664 static void key_getcomb_setlifetime(struct sadb_comb *);
665 static struct mbuf *key_getcomb_ealg(void);
666 static struct mbuf *key_getcomb_ah(void);
667 static struct mbuf *key_getcomb_ipcomp(void);
668 static struct mbuf *key_getprop(const struct secasindex *);
669 
670 static int key_acquire(const struct secasindex *, struct secpolicy *);
671 static uint32_t key_newacq(const struct secasindex *, int *);
672 static uint32_t key_getacq(const struct secasindex *, int *);
673 static int key_acqdone(const struct secasindex *, uint32_t);
674 static int key_acqreset(uint32_t);
675 static struct secspacq *key_newspacq(struct secpolicyindex *);
676 static struct secspacq *key_getspacq(struct secpolicyindex *);
677 static int key_acquire2(struct socket *, struct mbuf *,
678 	const struct sadb_msghdr *);
679 static int key_register(struct socket *, struct mbuf *,
680 	const struct sadb_msghdr *);
681 static int key_expire(struct secasvar *, int);
682 static int key_flush(struct socket *, struct mbuf *,
683 	const struct sadb_msghdr *);
684 static int key_dump(struct socket *, struct mbuf *,
685 	const struct sadb_msghdr *);
686 static int key_promisc(struct socket *, struct mbuf *,
687 	const struct sadb_msghdr *);
688 static int key_senderror(struct socket *, struct mbuf *, int);
689 static int key_validate_ext(const struct sadb_ext *, int);
690 static int key_align(struct mbuf *, struct sadb_msghdr *);
691 static struct mbuf *key_setlifetime(struct seclifetime *, uint16_t);
692 static struct mbuf *key_setkey(struct seckey *, uint16_t);
693 static int xform_init(struct secasvar *, u_short);
694 
695 #define	DBG_IPSEC_INITREF(t, p)	do {				\
696 	refcount_init(&(p)->refcnt, 1);				\
697 	KEYDBG(KEY_STAMP,					\
698 	    printf("%s: Initialize refcnt %s(%p) = %u\n",	\
699 	    __func__, #t, (p), (p)->refcnt));			\
700 } while (0)
701 #define	DBG_IPSEC_ADDREF(t, p)	do {				\
702 	refcount_acquire(&(p)->refcnt);				\
703 	KEYDBG(KEY_STAMP,					\
704 	    printf("%s: Acquire refcnt %s(%p) -> %u\n",		\
705 	    __func__, #t, (p), (p)->refcnt));			\
706 } while (0)
707 #define	DBG_IPSEC_DELREF(t, p)	do {				\
708 	KEYDBG(KEY_STAMP,					\
709 	    printf("%s: Release refcnt %s(%p) -> %u\n",		\
710 	    __func__, #t, (p), (p)->refcnt - 1));		\
711 	refcount_release(&(p)->refcnt);				\
712 } while (0)
713 
714 #define	IPSEC_INITREF(t, p)	refcount_init(&(p)->refcnt, 1)
715 #define	IPSEC_ADDREF(t, p)	refcount_acquire(&(p)->refcnt)
716 #define	IPSEC_DELREF(t, p)	refcount_release(&(p)->refcnt)
717 
718 #define	SP_INITREF(p)	IPSEC_INITREF(SP, p)
719 #define	SP_ADDREF(p)	IPSEC_ADDREF(SP, p)
720 #define	SP_DELREF(p)	IPSEC_DELREF(SP, p)
721 
722 #define	SAH_INITREF(p)	IPSEC_INITREF(SAH, p)
723 #define	SAH_ADDREF(p)	IPSEC_ADDREF(SAH, p)
724 #define	SAH_DELREF(p)	IPSEC_DELREF(SAH, p)
725 
726 #define	SAV_INITREF(p)	IPSEC_INITREF(SAV, p)
727 #define	SAV_ADDREF(p)	IPSEC_ADDREF(SAV, p)
728 #define	SAV_DELREF(p)	IPSEC_DELREF(SAV, p)
729 
730 /*
731  * Update the refcnt while holding the SPTREE lock.
732  */
733 void
734 key_addref(struct secpolicy *sp)
735 {
736 
737 	SP_ADDREF(sp);
738 }
739 
740 /*
741  * Return 0 when there are known to be no SP's for the specified
742  * direction.  Otherwise return 1.  This is used by IPsec code
743  * to optimize performance.
744  */
745 int
746 key_havesp(u_int dir)
747 {
748 
749 	return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ?
750 		TAILQ_FIRST(&V_sptree[dir]) != NULL : 1);
751 }
752 
753 /* %%% IPsec policy management */
754 /*
755  * Return current SPDB generation.
756  */
757 uint32_t
758 key_getspgen(void)
759 {
760 
761 	return (V_sp_genid);
762 }
763 
764 void
765 key_bumpspgen(void)
766 {
767 
768 	V_sp_genid++;
769 }
770 
771 static int
772 key_checksockaddrs(struct sockaddr *src, struct sockaddr *dst)
773 {
774 
775 	/* family match */
776 	if (src->sa_family != dst->sa_family)
777 		return (EINVAL);
778 	/* sa_len match */
779 	if (src->sa_len != dst->sa_len)
780 		return (EINVAL);
781 	switch (src->sa_family) {
782 #ifdef INET
783 	case AF_INET:
784 		if (src->sa_len != sizeof(struct sockaddr_in))
785 			return (EINVAL);
786 		break;
787 #endif
788 #ifdef INET6
789 	case AF_INET6:
790 		if (src->sa_len != sizeof(struct sockaddr_in6))
791 			return (EINVAL);
792 		break;
793 #endif
794 	default:
795 		return (EAFNOSUPPORT);
796 	}
797 	return (0);
798 }
799 
800 /*
801  * allocating a SP for OUTBOUND or INBOUND packet.
802  * Must call key_freesp() later.
803  * OUT:	NULL:	not found
804  *	others:	found and return the pointer.
805  */
806 struct secpolicy *
807 key_allocsp(struct secpolicyindex *spidx, u_int dir)
808 {
809 	SPTREE_RLOCK_TRACKER;
810 	struct secpolicy *sp;
811 
812 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
813 	IPSEC_ASSERT(dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND,
814 		("invalid direction %u", dir));
815 
816 	SPTREE_RLOCK();
817 	TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
818 		if (key_cmpspidx_withmask(&sp->spidx, spidx)) {
819 			SP_ADDREF(sp);
820 			break;
821 		}
822 	}
823 	SPTREE_RUNLOCK();
824 
825 	if (sp != NULL) {	/* found a SPD entry */
826 		sp->lastused = time_second;
827 		KEYDBG(IPSEC_STAMP,
828 		    printf("%s: return SP(%p)\n", __func__, sp));
829 		KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
830 	} else {
831 		KEYDBG(IPSEC_DATA,
832 		    printf("%s: lookup failed for ", __func__);
833 		    kdebug_secpolicyindex(spidx, NULL));
834 	}
835 	return (sp);
836 }
837 
838 /*
839  * Allocating an SA entry for an *INBOUND* or *OUTBOUND* TCP packet, signed
840  * or should be signed by MD5 signature.
841  * We don't use key_allocsa() for such lookups, because we don't know SPI.
842  * Unlike ESP and AH protocols, SPI isn't transmitted in the TCP header with
843  * signed packet. We use SADB only as storage for password.
844  * OUT:	positive:	corresponding SA for given saidx found.
845  *	NULL:		SA not found
846  */
847 struct secasvar *
848 key_allocsa_tcpmd5(struct secasindex *saidx)
849 {
850 	SAHTREE_RLOCK_TRACKER;
851 	struct secashead *sah;
852 	struct secasvar *sav;
853 
854 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP,
855 	    ("unexpected security protocol %u", saidx->proto));
856 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TCPMD5,
857 	    ("unexpected mode %u", saidx->mode));
858 
859 	SAHTREE_RLOCK();
860 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
861 		KEYDBG(IPSEC_DUMP,
862 		    printf("%s: checking SAH\n", __func__);
863 		    kdebug_secash(sah, "  "));
864 		if (sah->saidx.proto != IPPROTO_TCP)
865 			continue;
866 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
867 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
868 			break;
869 	}
870 	if (sah != NULL) {
871 		if (V_key_preferred_oldsa)
872 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
873 		else
874 			sav = TAILQ_FIRST(&sah->savtree_alive);
875 		if (sav != NULL)
876 			SAV_ADDREF(sav);
877 	} else
878 		sav = NULL;
879 	SAHTREE_RUNLOCK();
880 
881 	if (sav != NULL) {
882 		KEYDBG(IPSEC_STAMP,
883 		    printf("%s: return SA(%p)\n", __func__, sav));
884 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
885 	} else {
886 		KEYDBG(IPSEC_STAMP,
887 		    printf("%s: SA not found\n", __func__));
888 		KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
889 	}
890 	return (sav);
891 }
892 
893 /*
894  * Allocating an SA entry for an *OUTBOUND* packet.
895  * OUT:	positive:	corresponding SA for given saidx found.
896  *	NULL:		SA not found, but will be acquired, check *error
897  *			for acquiring status.
898  */
899 struct secasvar *
900 key_allocsa_policy(struct secpolicy *sp, const struct secasindex *saidx,
901     int *error)
902 {
903 	SAHTREE_RLOCK_TRACKER;
904 	struct secashead *sah;
905 	struct secasvar *sav;
906 
907 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
908 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TRANSPORT ||
909 		saidx->mode == IPSEC_MODE_TUNNEL,
910 		("unexpected policy %u", saidx->mode));
911 
912 	/*
913 	 * We check new SA in the IPsec request because a different
914 	 * SA may be involved each time this request is checked, either
915 	 * because new SAs are being configured, or this request is
916 	 * associated with an unconnected datagram socket, or this request
917 	 * is associated with a system default policy.
918 	 */
919 	SAHTREE_RLOCK();
920 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
921 		KEYDBG(IPSEC_DUMP,
922 		    printf("%s: checking SAH\n", __func__);
923 		    kdebug_secash(sah, "  "));
924 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID))
925 			break;
926 
927 	}
928 	if (sah != NULL) {
929 		/*
930 		 * Allocate the oldest SA available according to
931 		 * draft-jenkins-ipsec-rekeying-03.
932 		 */
933 		if (V_key_preferred_oldsa)
934 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
935 		else
936 			sav = TAILQ_FIRST(&sah->savtree_alive);
937 		if (sav != NULL)
938 			SAV_ADDREF(sav);
939 	} else
940 		sav = NULL;
941 	SAHTREE_RUNLOCK();
942 
943 	if (sav != NULL) {
944 		*error = 0;
945 		KEYDBG(IPSEC_STAMP,
946 		    printf("%s: chosen SA(%p) for SP(%p)\n", __func__,
947 			sav, sp));
948 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
949 		return (sav); /* return referenced SA */
950 	}
951 
952 	/* there is no SA */
953 	*error = key_acquire(saidx, sp);
954 	if ((*error) != 0)
955 		ipseclog((LOG_DEBUG,
956 		    "%s: error %d returned from key_acquire()\n",
957 			__func__, *error));
958 	KEYDBG(IPSEC_STAMP,
959 	    printf("%s: acquire SA for SP(%p), error %d\n",
960 		__func__, sp, *error));
961 	KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
962 	return (NULL);
963 }
964 
965 /*
966  * allocating a usable SA entry for a *INBOUND* packet.
967  * Must call key_freesav() later.
968  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
969  *	NULL:		not found, or error occurred.
970  *
971  * According to RFC 2401 SA is uniquely identified by a triple SPI,
972  * destination address, and security protocol. But according to RFC 4301,
973  * SPI by itself suffices to specify an SA.
974  *
975  * Note that, however, we do need to keep source address in IPsec SA.
976  * IKE specification and PF_KEY specification do assume that we
977  * keep source address in IPsec SA.  We see a tricky situation here.
978  */
979 struct secasvar *
980 key_allocsa(union sockaddr_union *dst, uint8_t proto, uint32_t spi)
981 {
982 	SAHTREE_RLOCK_TRACKER;
983 	struct secasvar *sav;
984 
985 	IPSEC_ASSERT(proto == IPPROTO_ESP || proto == IPPROTO_AH ||
986 	    proto == IPPROTO_IPCOMP, ("unexpected security protocol %u",
987 	    proto));
988 
989 	SAHTREE_RLOCK();
990 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
991 		if (sav->spi == spi)
992 			break;
993 	}
994 	/*
995 	 * We use single SPI namespace for all protocols, so it is
996 	 * impossible to have SPI duplicates in the SAVHASH.
997 	 */
998 	if (sav != NULL) {
999 		if (sav->state != SADB_SASTATE_LARVAL &&
1000 		    sav->sah->saidx.proto == proto &&
1001 		    key_sockaddrcmp(&dst->sa,
1002 			&sav->sah->saidx.dst.sa, 0) == 0)
1003 			SAV_ADDREF(sav);
1004 		else
1005 			sav = NULL;
1006 	}
1007 	SAHTREE_RUNLOCK();
1008 
1009 	if (sav == NULL) {
1010 		KEYDBG(IPSEC_STAMP,
1011 		    char buf[IPSEC_ADDRSTRLEN];
1012 		    printf("%s: SA not found for spi %u proto %u dst %s\n",
1013 			__func__, ntohl(spi), proto, ipsec_address(dst, buf,
1014 			sizeof(buf))));
1015 	} else {
1016 		KEYDBG(IPSEC_STAMP,
1017 		    printf("%s: return SA(%p)\n", __func__, sav));
1018 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1019 	}
1020 	return (sav);
1021 }
1022 
1023 struct secasvar *
1024 key_allocsa_tunnel(union sockaddr_union *src, union sockaddr_union *dst,
1025     uint8_t proto)
1026 {
1027 	SAHTREE_RLOCK_TRACKER;
1028 	struct secasindex saidx;
1029 	struct secashead *sah;
1030 	struct secasvar *sav;
1031 
1032 	IPSEC_ASSERT(src != NULL, ("null src address"));
1033 	IPSEC_ASSERT(dst != NULL, ("null dst address"));
1034 
1035 	KEY_SETSECASIDX(proto, IPSEC_MODE_TUNNEL, 0, &src->sa,
1036 	    &dst->sa, &saidx);
1037 
1038 	sav = NULL;
1039 	SAHTREE_RLOCK();
1040 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
1041 		if (IPSEC_MODE_TUNNEL != sah->saidx.mode)
1042 			continue;
1043 		if (proto != sah->saidx.proto)
1044 			continue;
1045 		if (key_sockaddrcmp(&src->sa, &sah->saidx.src.sa, 0) != 0)
1046 			continue;
1047 		if (key_sockaddrcmp(&dst->sa, &sah->saidx.dst.sa, 0) != 0)
1048 			continue;
1049 		/* XXXAE: is key_preferred_oldsa reasonably?*/
1050 		if (V_key_preferred_oldsa)
1051 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1052 		else
1053 			sav = TAILQ_FIRST(&sah->savtree_alive);
1054 		if (sav != NULL) {
1055 			SAV_ADDREF(sav);
1056 			break;
1057 		}
1058 	}
1059 	SAHTREE_RUNLOCK();
1060 	KEYDBG(IPSEC_STAMP,
1061 	    printf("%s: return SA(%p)\n", __func__, sav));
1062 	if (sav != NULL)
1063 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1064 	return (sav);
1065 }
1066 
1067 /*
1068  * Must be called after calling key_allocsp().
1069  */
1070 void
1071 key_freesp(struct secpolicy **spp)
1072 {
1073 	struct secpolicy *sp = *spp;
1074 
1075 	IPSEC_ASSERT(sp != NULL, ("null sp"));
1076 	if (SP_DELREF(sp) == 0)
1077 		return;
1078 
1079 	KEYDBG(IPSEC_STAMP,
1080 	    printf("%s: last reference to SP(%p)\n", __func__, sp));
1081 	KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1082 
1083 	*spp = NULL;
1084 	while (sp->tcount > 0)
1085 		ipsec_delisr(sp->req[--sp->tcount]);
1086 	free(sp, M_IPSEC_SP);
1087 }
1088 
1089 static void
1090 key_unlink(struct secpolicy *sp)
1091 {
1092 
1093 	IPSEC_ASSERT(sp->spidx.dir == IPSEC_DIR_INBOUND ||
1094 	    sp->spidx.dir == IPSEC_DIR_OUTBOUND,
1095 	    ("invalid direction %u", sp->spidx.dir));
1096 	SPTREE_UNLOCK_ASSERT();
1097 
1098 	KEYDBG(KEY_STAMP,
1099 	    printf("%s: SP(%p)\n", __func__, sp));
1100 	SPTREE_WLOCK();
1101 	if (sp->state != IPSEC_SPSTATE_ALIVE) {
1102 		/* SP is already unlinked */
1103 		SPTREE_WUNLOCK();
1104 		return;
1105 	}
1106 	sp->state = IPSEC_SPSTATE_DEAD;
1107 	TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
1108 	LIST_REMOVE(sp, idhash);
1109 	V_sp_genid++;
1110 	SPTREE_WUNLOCK();
1111 	key_freesp(&sp);
1112 }
1113 
1114 /*
1115  * insert a secpolicy into the SP database. Lower priorities first
1116  */
1117 static void
1118 key_insertsp(struct secpolicy *newsp)
1119 {
1120 	struct secpolicy *sp;
1121 
1122 	SPTREE_WLOCK_ASSERT();
1123 	TAILQ_FOREACH(sp, &V_sptree[newsp->spidx.dir], chain) {
1124 		if (newsp->priority < sp->priority) {
1125 			TAILQ_INSERT_BEFORE(sp, newsp, chain);
1126 			goto done;
1127 		}
1128 	}
1129 	TAILQ_INSERT_TAIL(&V_sptree[newsp->spidx.dir], newsp, chain);
1130 done:
1131 	LIST_INSERT_HEAD(SPHASH_HASH(newsp->id), newsp, idhash);
1132 	newsp->state = IPSEC_SPSTATE_ALIVE;
1133 	V_sp_genid++;
1134 }
1135 
1136 /*
1137  * Insert a bunch of VTI secpolicies into the SPDB.
1138  * We keep VTI policies in the separate list due to following reasons:
1139  * 1) they should be immutable to user's or some deamon's attempts to
1140  *    delete. The only way delete such policies - destroy or unconfigure
1141  *    corresponding virtual inteface.
1142  * 2) such policies have traffic selector that matches all traffic per
1143  *    address family.
1144  * Since all VTI policies have the same priority, we don't care about
1145  * policies order.
1146  */
1147 int
1148 key_register_ifnet(struct secpolicy **spp, u_int count)
1149 {
1150 	struct mbuf *m;
1151 	u_int i;
1152 
1153 	SPTREE_WLOCK();
1154 	/*
1155 	 * First of try to acquire id for each SP.
1156 	 */
1157 	for (i = 0; i < count; i++) {
1158 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1159 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1160 		    ("invalid direction %u", spp[i]->spidx.dir));
1161 
1162 		if ((spp[i]->id = key_getnewspid()) == 0) {
1163 			SPTREE_WUNLOCK();
1164 			return (EAGAIN);
1165 		}
1166 	}
1167 	for (i = 0; i < count; i++) {
1168 		TAILQ_INSERT_TAIL(&V_sptree_ifnet[spp[i]->spidx.dir],
1169 		    spp[i], chain);
1170 		/*
1171 		 * NOTE: despite the fact that we keep VTI SP in the
1172 		 * separate list, SPHASH contains policies from both
1173 		 * sources. Thus SADB_X_SPDGET will correctly return
1174 		 * SP by id, because it uses SPHASH for lookups.
1175 		 */
1176 		LIST_INSERT_HEAD(SPHASH_HASH(spp[i]->id), spp[i], idhash);
1177 		spp[i]->state = IPSEC_SPSTATE_IFNET;
1178 	}
1179 	SPTREE_WUNLOCK();
1180 	/*
1181 	 * Notify user processes about new SP.
1182 	 */
1183 	for (i = 0; i < count; i++) {
1184 		m = key_setdumpsp(spp[i], SADB_X_SPDADD, 0, 0);
1185 		if (m != NULL)
1186 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1187 	}
1188 	return (0);
1189 }
1190 
1191 void
1192 key_unregister_ifnet(struct secpolicy **spp, u_int count)
1193 {
1194 	struct mbuf *m;
1195 	u_int i;
1196 
1197 	SPTREE_WLOCK();
1198 	for (i = 0; i < count; i++) {
1199 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1200 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1201 		    ("invalid direction %u", spp[i]->spidx.dir));
1202 
1203 		if (spp[i]->state != IPSEC_SPSTATE_IFNET)
1204 			continue;
1205 		spp[i]->state = IPSEC_SPSTATE_DEAD;
1206 		TAILQ_REMOVE(&V_sptree_ifnet[spp[i]->spidx.dir],
1207 		    spp[i], chain);
1208 		LIST_REMOVE(spp[i], idhash);
1209 	}
1210 	SPTREE_WUNLOCK();
1211 
1212 	for (i = 0; i < count; i++) {
1213 		m = key_setdumpsp(spp[i], SADB_X_SPDDELETE, 0, 0);
1214 		if (m != NULL)
1215 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1216 	}
1217 }
1218 
1219 /*
1220  * Must be called after calling key_allocsa().
1221  * This function is called by key_freesp() to free some SA allocated
1222  * for a policy.
1223  */
1224 void
1225 key_freesav(struct secasvar **psav)
1226 {
1227 	struct secasvar *sav = *psav;
1228 
1229 	IPSEC_ASSERT(sav != NULL, ("null sav"));
1230 	if (SAV_DELREF(sav) == 0)
1231 		return;
1232 
1233 	KEYDBG(IPSEC_STAMP,
1234 	    printf("%s: last reference to SA(%p)\n", __func__, sav));
1235 
1236 	*psav = NULL;
1237 	key_delsav(sav);
1238 }
1239 
1240 /*
1241  * Unlink SA from SAH and SPI hash under SAHTREE_WLOCK.
1242  * Expect that SA has extra reference due to lookup.
1243  * Release this references, also release SAH reference after unlink.
1244  */
1245 static void
1246 key_unlinksav(struct secasvar *sav)
1247 {
1248 	struct secashead *sah;
1249 
1250 	KEYDBG(KEY_STAMP,
1251 	    printf("%s: SA(%p)\n", __func__, sav));
1252 
1253 	SAHTREE_UNLOCK_ASSERT();
1254 	SAHTREE_WLOCK();
1255 	if (sav->state == SADB_SASTATE_DEAD) {
1256 		/* SA is already unlinked */
1257 		SAHTREE_WUNLOCK();
1258 		return;
1259 	}
1260 	/* Unlink from SAH */
1261 	if (sav->state == SADB_SASTATE_LARVAL)
1262 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
1263 	else
1264 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
1265 	/* Unlink from SPI hash */
1266 	LIST_REMOVE(sav, spihash);
1267 	sav->state = SADB_SASTATE_DEAD;
1268 	sah = sav->sah;
1269 	SAHTREE_WUNLOCK();
1270 	key_freesav(&sav);
1271 	/* Since we are unlinked, release reference to SAH */
1272 	key_freesah(&sah);
1273 }
1274 
1275 /* %%% SPD management */
1276 /*
1277  * search SPD
1278  * OUT:	NULL	: not found
1279  *	others	: found, pointer to a SP.
1280  */
1281 static struct secpolicy *
1282 key_getsp(struct secpolicyindex *spidx)
1283 {
1284 	SPTREE_RLOCK_TRACKER;
1285 	struct secpolicy *sp;
1286 
1287 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
1288 
1289 	SPTREE_RLOCK();
1290 	TAILQ_FOREACH(sp, &V_sptree[spidx->dir], chain) {
1291 		if (key_cmpspidx_exactly(spidx, &sp->spidx)) {
1292 			SP_ADDREF(sp);
1293 			break;
1294 		}
1295 	}
1296 	SPTREE_RUNLOCK();
1297 
1298 	return sp;
1299 }
1300 
1301 /*
1302  * get SP by index.
1303  * OUT:	NULL	: not found
1304  *	others	: found, pointer to referenced SP.
1305  */
1306 static struct secpolicy *
1307 key_getspbyid(uint32_t id)
1308 {
1309 	SPTREE_RLOCK_TRACKER;
1310 	struct secpolicy *sp;
1311 
1312 	SPTREE_RLOCK();
1313 	LIST_FOREACH(sp, SPHASH_HASH(id), idhash) {
1314 		if (sp->id == id) {
1315 			SP_ADDREF(sp);
1316 			break;
1317 		}
1318 	}
1319 	SPTREE_RUNLOCK();
1320 	return (sp);
1321 }
1322 
1323 struct secpolicy *
1324 key_newsp(void)
1325 {
1326 	struct secpolicy *sp;
1327 
1328 	sp = malloc(sizeof(*sp), M_IPSEC_SP, M_NOWAIT | M_ZERO);
1329 	if (sp != NULL)
1330 		SP_INITREF(sp);
1331 	return (sp);
1332 }
1333 
1334 struct ipsecrequest *
1335 ipsec_newisr(void)
1336 {
1337 
1338 	return (malloc(sizeof(struct ipsecrequest), M_IPSEC_SR,
1339 	    M_NOWAIT | M_ZERO));
1340 }
1341 
1342 void
1343 ipsec_delisr(struct ipsecrequest *p)
1344 {
1345 
1346 	free(p, M_IPSEC_SR);
1347 }
1348 
1349 /*
1350  * create secpolicy structure from sadb_x_policy structure.
1351  * NOTE: `state', `secpolicyindex' and 'id' in secpolicy structure
1352  * are not set, so must be set properly later.
1353  */
1354 struct secpolicy *
1355 key_msg2sp(struct sadb_x_policy *xpl0, size_t len, int *error)
1356 {
1357 	struct secpolicy *newsp;
1358 
1359 	IPSEC_ASSERT(xpl0 != NULL, ("null xpl0"));
1360 	IPSEC_ASSERT(len >= sizeof(*xpl0), ("policy too short: %zu", len));
1361 
1362 	if (len != PFKEY_EXTLEN(xpl0)) {
1363 		ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n", __func__));
1364 		*error = EINVAL;
1365 		return NULL;
1366 	}
1367 
1368 	if ((newsp = key_newsp()) == NULL) {
1369 		*error = ENOBUFS;
1370 		return NULL;
1371 	}
1372 
1373 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1374 	newsp->policy = xpl0->sadb_x_policy_type;
1375 	newsp->priority = xpl0->sadb_x_policy_priority;
1376 	newsp->tcount = 0;
1377 
1378 	/* check policy */
1379 	switch (xpl0->sadb_x_policy_type) {
1380 	case IPSEC_POLICY_DISCARD:
1381 	case IPSEC_POLICY_NONE:
1382 	case IPSEC_POLICY_ENTRUST:
1383 	case IPSEC_POLICY_BYPASS:
1384 		break;
1385 
1386 	case IPSEC_POLICY_IPSEC:
1387 	    {
1388 		struct sadb_x_ipsecrequest *xisr;
1389 		struct ipsecrequest *isr;
1390 		int tlen;
1391 
1392 		/* validity check */
1393 		if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1394 			ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n",
1395 				__func__));
1396 			key_freesp(&newsp);
1397 			*error = EINVAL;
1398 			return NULL;
1399 		}
1400 
1401 		tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1402 		xisr = (struct sadb_x_ipsecrequest *)(xpl0 + 1);
1403 
1404 		while (tlen > 0) {
1405 			/* length check */
1406 			if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr) ||
1407 			    xisr->sadb_x_ipsecrequest_len > tlen) {
1408 				ipseclog((LOG_DEBUG, "%s: invalid ipsecrequest "
1409 					"length.\n", __func__));
1410 				key_freesp(&newsp);
1411 				*error = EINVAL;
1412 				return NULL;
1413 			}
1414 
1415 			if (newsp->tcount >= IPSEC_MAXREQ) {
1416 				ipseclog((LOG_DEBUG,
1417 				    "%s: too many ipsecrequests.\n",
1418 				    __func__));
1419 				key_freesp(&newsp);
1420 				*error = EINVAL;
1421 				return (NULL);
1422 			}
1423 
1424 			/* allocate request buffer */
1425 			/* NB: data structure is zero'd */
1426 			isr = ipsec_newisr();
1427 			if (isr == NULL) {
1428 				ipseclog((LOG_DEBUG,
1429 				    "%s: No more memory.\n", __func__));
1430 				key_freesp(&newsp);
1431 				*error = ENOBUFS;
1432 				return NULL;
1433 			}
1434 
1435 			newsp->req[newsp->tcount++] = isr;
1436 
1437 			/* set values */
1438 			switch (xisr->sadb_x_ipsecrequest_proto) {
1439 			case IPPROTO_ESP:
1440 			case IPPROTO_AH:
1441 			case IPPROTO_IPCOMP:
1442 				break;
1443 			default:
1444 				ipseclog((LOG_DEBUG,
1445 				    "%s: invalid proto type=%u\n", __func__,
1446 				    xisr->sadb_x_ipsecrequest_proto));
1447 				key_freesp(&newsp);
1448 				*error = EPROTONOSUPPORT;
1449 				return NULL;
1450 			}
1451 			isr->saidx.proto =
1452 			    (uint8_t)xisr->sadb_x_ipsecrequest_proto;
1453 
1454 			switch (xisr->sadb_x_ipsecrequest_mode) {
1455 			case IPSEC_MODE_TRANSPORT:
1456 			case IPSEC_MODE_TUNNEL:
1457 				break;
1458 			case IPSEC_MODE_ANY:
1459 			default:
1460 				ipseclog((LOG_DEBUG,
1461 				    "%s: invalid mode=%u\n", __func__,
1462 				    xisr->sadb_x_ipsecrequest_mode));
1463 				key_freesp(&newsp);
1464 				*error = EINVAL;
1465 				return NULL;
1466 			}
1467 			isr->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1468 
1469 			switch (xisr->sadb_x_ipsecrequest_level) {
1470 			case IPSEC_LEVEL_DEFAULT:
1471 			case IPSEC_LEVEL_USE:
1472 			case IPSEC_LEVEL_REQUIRE:
1473 				break;
1474 			case IPSEC_LEVEL_UNIQUE:
1475 				/* validity check */
1476 				/*
1477 				 * If range violation of reqid, kernel will
1478 				 * update it, don't refuse it.
1479 				 */
1480 				if (xisr->sadb_x_ipsecrequest_reqid
1481 						> IPSEC_MANUAL_REQID_MAX) {
1482 					ipseclog((LOG_DEBUG,
1483 					    "%s: reqid=%d range "
1484 					    "violation, updated by kernel.\n",
1485 					    __func__,
1486 					    xisr->sadb_x_ipsecrequest_reqid));
1487 					xisr->sadb_x_ipsecrequest_reqid = 0;
1488 				}
1489 
1490 				/* allocate new reqid id if reqid is zero. */
1491 				if (xisr->sadb_x_ipsecrequest_reqid == 0) {
1492 					u_int32_t reqid;
1493 					if ((reqid = key_newreqid()) == 0) {
1494 						key_freesp(&newsp);
1495 						*error = ENOBUFS;
1496 						return NULL;
1497 					}
1498 					isr->saidx.reqid = reqid;
1499 					xisr->sadb_x_ipsecrequest_reqid = reqid;
1500 				} else {
1501 				/* set it for manual keying. */
1502 					isr->saidx.reqid =
1503 					    xisr->sadb_x_ipsecrequest_reqid;
1504 				}
1505 				break;
1506 
1507 			default:
1508 				ipseclog((LOG_DEBUG, "%s: invalid level=%u\n",
1509 					__func__,
1510 					xisr->sadb_x_ipsecrequest_level));
1511 				key_freesp(&newsp);
1512 				*error = EINVAL;
1513 				return NULL;
1514 			}
1515 			isr->level = xisr->sadb_x_ipsecrequest_level;
1516 
1517 			/* set IP addresses if there */
1518 			if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1519 				struct sockaddr *paddr;
1520 
1521 				len = tlen - sizeof(*xisr);
1522 				paddr = (struct sockaddr *)(xisr + 1);
1523 				/* validity check */
1524 				if (len < sizeof(struct sockaddr) ||
1525 				    len < 2 * paddr->sa_len ||
1526 				    paddr->sa_len > sizeof(isr->saidx.src)) {
1527 					ipseclog((LOG_DEBUG, "%s: invalid "
1528 						"request address length.\n",
1529 						__func__));
1530 					key_freesp(&newsp);
1531 					*error = EINVAL;
1532 					return NULL;
1533 				}
1534 				/*
1535 				 * Request length should be enough to keep
1536 				 * source and destination addresses.
1537 				 */
1538 				if (xisr->sadb_x_ipsecrequest_len <
1539 				    sizeof(*xisr) + 2 * paddr->sa_len) {
1540 					ipseclog((LOG_DEBUG, "%s: invalid "
1541 					    "ipsecrequest length.\n",
1542 					    __func__));
1543 					key_freesp(&newsp);
1544 					*error = EINVAL;
1545 					return (NULL);
1546 				}
1547 				bcopy(paddr, &isr->saidx.src, paddr->sa_len);
1548 				paddr = (struct sockaddr *)((caddr_t)paddr +
1549 				    paddr->sa_len);
1550 
1551 				/* validity check */
1552 				if (paddr->sa_len !=
1553 				    isr->saidx.src.sa.sa_len) {
1554 					ipseclog((LOG_DEBUG, "%s: invalid "
1555 						"request address length.\n",
1556 						__func__));
1557 					key_freesp(&newsp);
1558 					*error = EINVAL;
1559 					return NULL;
1560 				}
1561 				/* AF family should match */
1562 				if (paddr->sa_family !=
1563 				    isr->saidx.src.sa.sa_family) {
1564 					ipseclog((LOG_DEBUG, "%s: address "
1565 					    "family doesn't match.\n",
1566 						__func__));
1567 					key_freesp(&newsp);
1568 					*error = EINVAL;
1569 					return (NULL);
1570 				}
1571 				bcopy(paddr, &isr->saidx.dst, paddr->sa_len);
1572 			} else {
1573 				/*
1574 				 * Addresses for TUNNEL mode requests are
1575 				 * mandatory.
1576 				 */
1577 				if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1578 					ipseclog((LOG_DEBUG, "%s: missing "
1579 					    "request addresses.\n", __func__));
1580 					key_freesp(&newsp);
1581 					*error = EINVAL;
1582 					return (NULL);
1583 				}
1584 			}
1585 			tlen -= xisr->sadb_x_ipsecrequest_len;
1586 
1587 			/* validity check */
1588 			if (tlen < 0) {
1589 				ipseclog((LOG_DEBUG, "%s: becoming tlen < 0.\n",
1590 					__func__));
1591 				key_freesp(&newsp);
1592 				*error = EINVAL;
1593 				return NULL;
1594 			}
1595 
1596 			xisr = (struct sadb_x_ipsecrequest *)((caddr_t)xisr
1597 			                 + xisr->sadb_x_ipsecrequest_len);
1598 		}
1599 		/* XXXAE: LARVAL SP */
1600 		if (newsp->tcount < 1) {
1601 			ipseclog((LOG_DEBUG, "%s: valid IPSEC transforms "
1602 			    "not found.\n", __func__));
1603 			key_freesp(&newsp);
1604 			*error = EINVAL;
1605 			return (NULL);
1606 		}
1607 	    }
1608 		break;
1609 	default:
1610 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
1611 		key_freesp(&newsp);
1612 		*error = EINVAL;
1613 		return NULL;
1614 	}
1615 
1616 	*error = 0;
1617 	return (newsp);
1618 }
1619 
1620 uint32_t
1621 key_newreqid(void)
1622 {
1623 	static uint32_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1624 
1625 	if (auto_reqid == ~0)
1626 		auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1627 	else
1628 		auto_reqid++;
1629 
1630 	/* XXX should be unique check */
1631 	return (auto_reqid);
1632 }
1633 
1634 /*
1635  * copy secpolicy struct to sadb_x_policy structure indicated.
1636  */
1637 static struct mbuf *
1638 key_sp2mbuf(struct secpolicy *sp)
1639 {
1640 	struct mbuf *m;
1641 	size_t tlen;
1642 
1643 	tlen = key_getspreqmsglen(sp);
1644 	m = m_get2(tlen, M_NOWAIT, MT_DATA, 0);
1645 	if (m == NULL)
1646 		return (NULL);
1647 	m_align(m, tlen);
1648 	m->m_len = tlen;
1649 	if (key_sp2msg(sp, m->m_data, &tlen) != 0) {
1650 		m_freem(m);
1651 		return (NULL);
1652 	}
1653 	return (m);
1654 }
1655 
1656 int
1657 key_sp2msg(struct secpolicy *sp, void *request, size_t *len)
1658 {
1659 	struct sadb_x_ipsecrequest *xisr;
1660 	struct sadb_x_policy *xpl;
1661 	struct ipsecrequest *isr;
1662 	size_t xlen, ilen;
1663 	caddr_t p;
1664 	int error, i;
1665 
1666 	IPSEC_ASSERT(sp != NULL, ("null policy"));
1667 
1668 	xlen = sizeof(*xpl);
1669 	if (*len < xlen)
1670 		return (EINVAL);
1671 
1672 	error = 0;
1673 	bzero(request, *len);
1674 	xpl = (struct sadb_x_policy *)request;
1675 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
1676 	xpl->sadb_x_policy_type = sp->policy;
1677 	xpl->sadb_x_policy_dir = sp->spidx.dir;
1678 	xpl->sadb_x_policy_id = sp->id;
1679 	xpl->sadb_x_policy_priority = sp->priority;
1680 	switch (sp->state) {
1681 	case IPSEC_SPSTATE_IFNET:
1682 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_IFNET;
1683 		break;
1684 	case IPSEC_SPSTATE_PCB:
1685 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_PCB;
1686 		break;
1687 	default:
1688 		xpl->sadb_x_policy_scope = IPSEC_POLICYSCOPE_GLOBAL;
1689 	}
1690 
1691 	/* if is the policy for ipsec ? */
1692 	if (sp->policy == IPSEC_POLICY_IPSEC) {
1693 		p = (caddr_t)xpl + sizeof(*xpl);
1694 		for (i = 0; i < sp->tcount; i++) {
1695 			isr = sp->req[i];
1696 			ilen = PFKEY_ALIGN8(sizeof(*xisr) +
1697 			    isr->saidx.src.sa.sa_len +
1698 			    isr->saidx.dst.sa.sa_len);
1699 			xlen += ilen;
1700 			if (xlen > *len) {
1701 				error = ENOBUFS;
1702 				/* Calculate needed size */
1703 				continue;
1704 			}
1705 			xisr = (struct sadb_x_ipsecrequest *)p;
1706 			xisr->sadb_x_ipsecrequest_len = ilen;
1707 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
1708 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
1709 			xisr->sadb_x_ipsecrequest_level = isr->level;
1710 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
1711 
1712 			p += sizeof(*xisr);
1713 			bcopy(&isr->saidx.src, p, isr->saidx.src.sa.sa_len);
1714 			p += isr->saidx.src.sa.sa_len;
1715 			bcopy(&isr->saidx.dst, p, isr->saidx.dst.sa.sa_len);
1716 			p += isr->saidx.dst.sa.sa_len;
1717 		}
1718 	}
1719 	xpl->sadb_x_policy_len = PFKEY_UNIT64(xlen);
1720 	if (error == 0)
1721 		*len = xlen;
1722 	else
1723 		*len = sizeof(*xpl);
1724 	return (error);
1725 }
1726 
1727 /* m will not be freed nor modified */
1728 static struct mbuf *
1729 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
1730     int ndeep, int nitem, ...)
1731 {
1732 	va_list ap;
1733 	int idx;
1734 	int i;
1735 	struct mbuf *result = NULL, *n;
1736 	int len;
1737 
1738 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1739 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1740 
1741 	va_start(ap, nitem);
1742 	for (i = 0; i < nitem; i++) {
1743 		idx = va_arg(ap, int);
1744 		if (idx < 0 || idx > SADB_EXT_MAX)
1745 			goto fail;
1746 		/* don't attempt to pull empty extension */
1747 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
1748 			continue;
1749 		if (idx != SADB_EXT_RESERVED  &&
1750 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
1751 			continue;
1752 
1753 		if (idx == SADB_EXT_RESERVED) {
1754 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
1755 
1756 			IPSEC_ASSERT(len <= MHLEN, ("header too big %u", len));
1757 
1758 			MGETHDR(n, M_NOWAIT, MT_DATA);
1759 			if (!n)
1760 				goto fail;
1761 			n->m_len = len;
1762 			n->m_next = NULL;
1763 			m_copydata(m, 0, sizeof(struct sadb_msg),
1764 			    mtod(n, caddr_t));
1765 		} else if (i < ndeep) {
1766 			len = mhp->extlen[idx];
1767 			n = m_get2(len, M_NOWAIT, MT_DATA, 0);
1768 			if (n == NULL)
1769 				goto fail;
1770 			m_align(n, len);
1771 			n->m_len = len;
1772 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
1773 			    mtod(n, caddr_t));
1774 		} else {
1775 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
1776 			    M_NOWAIT);
1777 		}
1778 		if (n == NULL)
1779 			goto fail;
1780 
1781 		if (result)
1782 			m_cat(result, n);
1783 		else
1784 			result = n;
1785 	}
1786 	va_end(ap);
1787 
1788 	if ((result->m_flags & M_PKTHDR) != 0) {
1789 		result->m_pkthdr.len = 0;
1790 		for (n = result; n; n = n->m_next)
1791 			result->m_pkthdr.len += n->m_len;
1792 	}
1793 
1794 	return result;
1795 
1796 fail:
1797 	m_freem(result);
1798 	va_end(ap);
1799 	return NULL;
1800 }
1801 
1802 /*
1803  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
1804  * add an entry to SP database, when received
1805  *   <base, address(SD), (lifetime(H),) policy>
1806  * from the user(?).
1807  * Adding to SP database,
1808  * and send
1809  *   <base, address(SD), (lifetime(H),) policy>
1810  * to the socket which was send.
1811  *
1812  * SPDADD set a unique policy entry.
1813  * SPDSETIDX like SPDADD without a part of policy requests.
1814  * SPDUPDATE replace a unique policy entry.
1815  *
1816  * XXXAE: serialize this in PF_KEY to avoid races.
1817  * m will always be freed.
1818  */
1819 static int
1820 key_spdadd(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
1821 {
1822 	struct secpolicyindex spidx;
1823 	struct sadb_address *src0, *dst0;
1824 	struct sadb_x_policy *xpl0, *xpl;
1825 	struct sadb_lifetime *lft = NULL;
1826 	struct secpolicy *newsp;
1827 	int error;
1828 
1829 	IPSEC_ASSERT(so != NULL, ("null socket"));
1830 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1831 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1832 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
1833 
1834 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
1835 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
1836 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
1837 		ipseclog((LOG_DEBUG,
1838 		    "%s: invalid message: missing required header.\n",
1839 		    __func__));
1840 		return key_senderror(so, m, EINVAL);
1841 	}
1842 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
1843 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
1844 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
1845 		ipseclog((LOG_DEBUG,
1846 		    "%s: invalid message: wrong header size.\n", __func__));
1847 		return key_senderror(so, m, EINVAL);
1848 	}
1849 	if (!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD)) {
1850 		if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD)) {
1851 			ipseclog((LOG_DEBUG,
1852 			    "%s: invalid message: wrong header size.\n",
1853 			    __func__));
1854 			return key_senderror(so, m, EINVAL);
1855 		}
1856 		lft = (struct sadb_lifetime *)mhp->ext[SADB_EXT_LIFETIME_HARD];
1857 	}
1858 
1859 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
1860 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
1861 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
1862 
1863 	/* check the direciton */
1864 	switch (xpl0->sadb_x_policy_dir) {
1865 	case IPSEC_DIR_INBOUND:
1866 	case IPSEC_DIR_OUTBOUND:
1867 		break;
1868 	default:
1869 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
1870 		return key_senderror(so, m, EINVAL);
1871 	}
1872 	/* key_spdadd() accepts DISCARD, NONE and IPSEC. */
1873 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
1874 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
1875 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
1876 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
1877 		return key_senderror(so, m, EINVAL);
1878 	}
1879 
1880 	/* policy requests are mandatory when action is ipsec. */
1881 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
1882 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
1883 		ipseclog((LOG_DEBUG,
1884 		    "%s: policy requests required.\n", __func__));
1885 		return key_senderror(so, m, EINVAL);
1886 	}
1887 
1888 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
1889 	    (struct sockaddr *)(dst0 + 1));
1890 	if (error != 0 ||
1891 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
1892 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
1893 		return key_senderror(so, m, error);
1894 	}
1895 	/* make secindex */
1896 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
1897 	                src0 + 1,
1898 	                dst0 + 1,
1899 	                src0->sadb_address_prefixlen,
1900 	                dst0->sadb_address_prefixlen,
1901 	                src0->sadb_address_proto,
1902 	                &spidx);
1903 	/* Checking there is SP already or not. */
1904 	newsp = key_getsp(&spidx);
1905 	if (newsp != NULL) {
1906 		if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
1907 			KEYDBG(KEY_STAMP,
1908 			    printf("%s: unlink SP(%p) for SPDUPDATE\n",
1909 				__func__, newsp));
1910 			KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
1911 			key_unlink(newsp);
1912 			key_freesp(&newsp);
1913 		} else {
1914 			key_freesp(&newsp);
1915 			ipseclog((LOG_DEBUG, "%s: a SP entry exists already.",
1916 			    __func__));
1917 			return (key_senderror(so, m, EEXIST));
1918 		}
1919 	}
1920 
1921 	/* allocate new SP entry */
1922 	if ((newsp = key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error)) == NULL) {
1923 		return key_senderror(so, m, error);
1924 	}
1925 
1926 	newsp->lastused = newsp->created = time_second;
1927 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
1928 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
1929 	bcopy(&spidx, &newsp->spidx, sizeof(spidx));
1930 
1931 	/* XXXAE: there is race between key_getsp() and key_insertsp() */
1932 	SPTREE_WLOCK();
1933 	if ((newsp->id = key_getnewspid()) == 0) {
1934 		SPTREE_WUNLOCK();
1935 		key_freesp(&newsp);
1936 		return key_senderror(so, m, ENOBUFS);
1937 	}
1938 	key_insertsp(newsp);
1939 	SPTREE_WUNLOCK();
1940 
1941 	KEYDBG(KEY_STAMP,
1942 	    printf("%s: SP(%p)\n", __func__, newsp));
1943 	KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
1944 
1945     {
1946 	struct mbuf *n, *mpolicy;
1947 	struct sadb_msg *newmsg;
1948 	int off;
1949 
1950 	/* create new sadb_msg to reply. */
1951 	if (lft) {
1952 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
1953 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
1954 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
1955 	} else {
1956 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
1957 		    SADB_X_EXT_POLICY,
1958 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
1959 	}
1960 	if (!n)
1961 		return key_senderror(so, m, ENOBUFS);
1962 
1963 	if (n->m_len < sizeof(*newmsg)) {
1964 		n = m_pullup(n, sizeof(*newmsg));
1965 		if (!n)
1966 			return key_senderror(so, m, ENOBUFS);
1967 	}
1968 	newmsg = mtod(n, struct sadb_msg *);
1969 	newmsg->sadb_msg_errno = 0;
1970 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
1971 
1972 	off = 0;
1973 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
1974 	    sizeof(*xpl), &off);
1975 	if (mpolicy == NULL) {
1976 		/* n is already freed */
1977 		return key_senderror(so, m, ENOBUFS);
1978 	}
1979 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, caddr_t) + off);
1980 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
1981 		m_freem(n);
1982 		return key_senderror(so, m, EINVAL);
1983 	}
1984 	xpl->sadb_x_policy_id = newsp->id;
1985 
1986 	m_freem(m);
1987 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
1988     }
1989 }
1990 
1991 /*
1992  * get new policy id.
1993  * OUT:
1994  *	0:	failure.
1995  *	others: success.
1996  */
1997 static uint32_t
1998 key_getnewspid(void)
1999 {
2000 	struct secpolicy *sp;
2001 	uint32_t newid = 0;
2002 	int count = V_key_spi_trycnt;	/* XXX */
2003 
2004 	SPTREE_WLOCK_ASSERT();
2005 	while (count--) {
2006 		if (V_policy_id == ~0) /* overflowed */
2007 			newid = V_policy_id = 1;
2008 		else
2009 			newid = ++V_policy_id;
2010 		LIST_FOREACH(sp, SPHASH_HASH(newid), idhash) {
2011 			if (sp->id == newid)
2012 				break;
2013 		}
2014 		if (sp == NULL)
2015 			break;
2016 	}
2017 	if (count == 0 || newid == 0) {
2018 		ipseclog((LOG_DEBUG, "%s: failed to allocate policy id.\n",
2019 		    __func__));
2020 		return (0);
2021 	}
2022 	return (newid);
2023 }
2024 
2025 /*
2026  * SADB_SPDDELETE processing
2027  * receive
2028  *   <base, address(SD), policy(*)>
2029  * from the user(?), and set SADB_SASTATE_DEAD,
2030  * and send,
2031  *   <base, address(SD), policy(*)>
2032  * to the ikmpd.
2033  * policy(*) including direction of policy.
2034  *
2035  * m will always be freed.
2036  */
2037 static int
2038 key_spddelete(struct socket *so, struct mbuf *m,
2039     const struct sadb_msghdr *mhp)
2040 {
2041 	struct secpolicyindex spidx;
2042 	struct sadb_address *src0, *dst0;
2043 	struct sadb_x_policy *xpl0;
2044 	struct secpolicy *sp;
2045 
2046 	IPSEC_ASSERT(so != NULL, ("null so"));
2047 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2048 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2049 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2050 
2051 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
2052 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
2053 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
2054 		ipseclog((LOG_DEBUG,
2055 		    "%s: invalid message: missing required header.\n",
2056 		    __func__));
2057 		return key_senderror(so, m, EINVAL);
2058 	}
2059 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
2060 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
2061 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2062 		ipseclog((LOG_DEBUG,
2063 		    "%s: invalid message: wrong header size.\n", __func__));
2064 		return key_senderror(so, m, EINVAL);
2065 	}
2066 
2067 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
2068 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
2069 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
2070 
2071 	/* check the direciton */
2072 	switch (xpl0->sadb_x_policy_dir) {
2073 	case IPSEC_DIR_INBOUND:
2074 	case IPSEC_DIR_OUTBOUND:
2075 		break;
2076 	default:
2077 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2078 		return key_senderror(so, m, EINVAL);
2079 	}
2080 	/* Only DISCARD, NONE and IPSEC are allowed */
2081 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2082 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2083 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2084 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2085 		return key_senderror(so, m, EINVAL);
2086 	}
2087 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
2088 	    (struct sockaddr *)(dst0 + 1)) != 0 ||
2089 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2090 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2091 		return key_senderror(so, m, EINVAL);
2092 	}
2093 	/* make secindex */
2094 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2095 	                src0 + 1,
2096 	                dst0 + 1,
2097 	                src0->sadb_address_prefixlen,
2098 	                dst0->sadb_address_prefixlen,
2099 	                src0->sadb_address_proto,
2100 	                &spidx);
2101 
2102 	/* Is there SP in SPD ? */
2103 	if ((sp = key_getsp(&spidx)) == NULL) {
2104 		ipseclog((LOG_DEBUG, "%s: no SP found.\n", __func__));
2105 		return key_senderror(so, m, EINVAL);
2106 	}
2107 
2108 	/* save policy id to buffer to be returned. */
2109 	xpl0->sadb_x_policy_id = sp->id;
2110 
2111 	KEYDBG(KEY_STAMP,
2112 	    printf("%s: SP(%p)\n", __func__, sp));
2113 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2114 	key_unlink(sp);
2115 	key_freesp(&sp);
2116 
2117     {
2118 	struct mbuf *n;
2119 	struct sadb_msg *newmsg;
2120 
2121 	/* create new sadb_msg to reply. */
2122 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2123 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2124 	if (!n)
2125 		return key_senderror(so, m, ENOBUFS);
2126 
2127 	newmsg = mtod(n, struct sadb_msg *);
2128 	newmsg->sadb_msg_errno = 0;
2129 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2130 
2131 	m_freem(m);
2132 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2133     }
2134 }
2135 
2136 /*
2137  * SADB_SPDDELETE2 processing
2138  * receive
2139  *   <base, policy(*)>
2140  * from the user(?), and set SADB_SASTATE_DEAD,
2141  * and send,
2142  *   <base, policy(*)>
2143  * to the ikmpd.
2144  * policy(*) including direction of policy.
2145  *
2146  * m will always be freed.
2147  */
2148 static int
2149 key_spddelete2(struct socket *so, struct mbuf *m,
2150     const struct sadb_msghdr *mhp)
2151 {
2152 	struct secpolicy *sp;
2153 	uint32_t id;
2154 
2155 	IPSEC_ASSERT(so != NULL, ("null socket"));
2156 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2157 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2158 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2159 
2160 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2161 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2162 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2163 		    __func__));
2164 		return key_senderror(so, m, EINVAL);
2165 	}
2166 
2167 	id = ((struct sadb_x_policy *)
2168 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2169 
2170 	/* Is there SP in SPD ? */
2171 	if ((sp = key_getspbyid(id)) == NULL) {
2172 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2173 		    __func__, id));
2174 		return key_senderror(so, m, EINVAL);
2175 	}
2176 
2177 	KEYDBG(KEY_STAMP,
2178 	    printf("%s: SP(%p)\n", __func__, sp));
2179 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2180 	key_unlink(sp);
2181 	if (sp->state != IPSEC_SPSTATE_DEAD) {
2182 		ipseclog((LOG_DEBUG, "%s: failed to delete SP with id %u.\n",
2183 		    __func__, id));
2184 		key_freesp(&sp);
2185 		return (key_senderror(so, m, EACCES));
2186 	}
2187 	key_freesp(&sp);
2188 
2189     {
2190 	struct mbuf *n, *nn;
2191 	struct sadb_msg *newmsg;
2192 	int off, len;
2193 
2194 	/* create new sadb_msg to reply. */
2195 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2196 
2197 	MGETHDR(n, M_NOWAIT, MT_DATA);
2198 	if (n && len > MHLEN) {
2199 		if (!(MCLGET(n, M_NOWAIT))) {
2200 			m_freem(n);
2201 			n = NULL;
2202 		}
2203 	}
2204 	if (!n)
2205 		return key_senderror(so, m, ENOBUFS);
2206 
2207 	n->m_len = len;
2208 	n->m_next = NULL;
2209 	off = 0;
2210 
2211 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
2212 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2213 
2214 	IPSEC_ASSERT(off == len, ("length inconsistency (off %u len %u)",
2215 		off, len));
2216 
2217 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2218 	    mhp->extlen[SADB_X_EXT_POLICY], M_NOWAIT);
2219 	if (!n->m_next) {
2220 		m_freem(n);
2221 		return key_senderror(so, m, ENOBUFS);
2222 	}
2223 
2224 	n->m_pkthdr.len = 0;
2225 	for (nn = n; nn; nn = nn->m_next)
2226 		n->m_pkthdr.len += nn->m_len;
2227 
2228 	newmsg = mtod(n, struct sadb_msg *);
2229 	newmsg->sadb_msg_errno = 0;
2230 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2231 
2232 	m_freem(m);
2233 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2234     }
2235 }
2236 
2237 /*
2238  * SADB_X_SPDGET processing
2239  * receive
2240  *   <base, policy(*)>
2241  * from the user(?),
2242  * and send,
2243  *   <base, address(SD), policy>
2244  * to the ikmpd.
2245  * policy(*) including direction of policy.
2246  *
2247  * m will always be freed.
2248  */
2249 static int
2250 key_spdget(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2251 {
2252 	struct secpolicy *sp;
2253 	struct mbuf *n;
2254 	uint32_t id;
2255 
2256 	IPSEC_ASSERT(so != NULL, ("null socket"));
2257 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2258 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2259 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2260 
2261 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2262 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2263 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2264 		    __func__));
2265 		return key_senderror(so, m, EINVAL);
2266 	}
2267 
2268 	id = ((struct sadb_x_policy *)
2269 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2270 
2271 	/* Is there SP in SPD ? */
2272 	if ((sp = key_getspbyid(id)) == NULL) {
2273 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2274 		    __func__, id));
2275 		return key_senderror(so, m, ENOENT);
2276 	}
2277 
2278 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2279 	    mhp->msg->sadb_msg_pid);
2280 	key_freesp(&sp);
2281 	if (n != NULL) {
2282 		m_freem(m);
2283 		return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2284 	} else
2285 		return key_senderror(so, m, ENOBUFS);
2286 }
2287 
2288 /*
2289  * SADB_X_SPDACQUIRE processing.
2290  * Acquire policy and SA(s) for a *OUTBOUND* packet.
2291  * send
2292  *   <base, policy(*)>
2293  * to KMD, and expect to receive
2294  *   <base> with SADB_X_SPDACQUIRE if error occurred,
2295  * or
2296  *   <base, policy>
2297  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2298  * policy(*) is without policy requests.
2299  *
2300  *    0     : succeed
2301  *    others: error number
2302  */
2303 int
2304 key_spdacquire(struct secpolicy *sp)
2305 {
2306 	struct mbuf *result = NULL, *m;
2307 	struct secspacq *newspacq;
2308 
2309 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2310 	IPSEC_ASSERT(sp->req == NULL, ("policy exists"));
2311 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
2312 		("policy not IPSEC %u", sp->policy));
2313 
2314 	/* Get an entry to check whether sent message or not. */
2315 	newspacq = key_getspacq(&sp->spidx);
2316 	if (newspacq != NULL) {
2317 		if (V_key_blockacq_count < newspacq->count) {
2318 			/* reset counter and do send message. */
2319 			newspacq->count = 0;
2320 		} else {
2321 			/* increment counter and do nothing. */
2322 			newspacq->count++;
2323 			SPACQ_UNLOCK();
2324 			return (0);
2325 		}
2326 		SPACQ_UNLOCK();
2327 	} else {
2328 		/* make new entry for blocking to send SADB_ACQUIRE. */
2329 		newspacq = key_newspacq(&sp->spidx);
2330 		if (newspacq == NULL)
2331 			return ENOBUFS;
2332 	}
2333 
2334 	/* create new sadb_msg to reply. */
2335 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2336 	if (!m)
2337 		return ENOBUFS;
2338 
2339 	result = m;
2340 
2341 	result->m_pkthdr.len = 0;
2342 	for (m = result; m; m = m->m_next)
2343 		result->m_pkthdr.len += m->m_len;
2344 
2345 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2346 	    PFKEY_UNIT64(result->m_pkthdr.len);
2347 
2348 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2349 }
2350 
2351 /*
2352  * SADB_SPDFLUSH processing
2353  * receive
2354  *   <base>
2355  * from the user, and free all entries in secpctree.
2356  * and send,
2357  *   <base>
2358  * to the user.
2359  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2360  *
2361  * m will always be freed.
2362  */
2363 static int
2364 key_spdflush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2365 {
2366 	struct secpolicy_queue drainq;
2367 	struct sadb_msg *newmsg;
2368 	struct secpolicy *sp, *nextsp;
2369 	u_int dir;
2370 
2371 	IPSEC_ASSERT(so != NULL, ("null socket"));
2372 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2373 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2374 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2375 
2376 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2377 		return key_senderror(so, m, EINVAL);
2378 
2379 	TAILQ_INIT(&drainq);
2380 	SPTREE_WLOCK();
2381 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2382 		TAILQ_CONCAT(&drainq, &V_sptree[dir], chain);
2383 	}
2384 	/*
2385 	 * We need to set state to DEAD for each policy to be sure,
2386 	 * that another thread won't try to unlink it.
2387 	 * Also remove SP from sphash.
2388 	 */
2389 	TAILQ_FOREACH(sp, &drainq, chain) {
2390 		sp->state = IPSEC_SPSTATE_DEAD;
2391 		LIST_REMOVE(sp, idhash);
2392 	}
2393 	V_sp_genid++;
2394 	SPTREE_WUNLOCK();
2395 	sp = TAILQ_FIRST(&drainq);
2396 	while (sp != NULL) {
2397 		nextsp = TAILQ_NEXT(sp, chain);
2398 		key_freesp(&sp);
2399 		sp = nextsp;
2400 	}
2401 
2402 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2403 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
2404 		return key_senderror(so, m, ENOBUFS);
2405 	}
2406 
2407 	if (m->m_next)
2408 		m_freem(m->m_next);
2409 	m->m_next = NULL;
2410 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2411 	newmsg = mtod(m, struct sadb_msg *);
2412 	newmsg->sadb_msg_errno = 0;
2413 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2414 
2415 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2416 }
2417 
2418 static uint8_t
2419 key_satype2scopemask(uint8_t satype)
2420 {
2421 
2422 	if (satype == IPSEC_POLICYSCOPE_ANY)
2423 		return (0xff);
2424 	return (satype);
2425 }
2426 /*
2427  * SADB_SPDDUMP processing
2428  * receive
2429  *   <base>
2430  * from the user, and dump all SP leaves and send,
2431  *   <base> .....
2432  * to the ikmpd.
2433  *
2434  * NOTE:
2435  *   sadb_msg_satype is considered as mask of policy scopes.
2436  *   m will always be freed.
2437  */
2438 static int
2439 key_spddump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2440 {
2441 	SPTREE_RLOCK_TRACKER;
2442 	struct secpolicy *sp;
2443 	struct mbuf *n;
2444 	int cnt;
2445 	u_int dir, scope;
2446 
2447 	IPSEC_ASSERT(so != NULL, ("null socket"));
2448 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2449 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2450 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2451 
2452 	/* search SPD entry and get buffer size. */
2453 	cnt = 0;
2454 	scope = key_satype2scopemask(mhp->msg->sadb_msg_satype);
2455 	SPTREE_RLOCK();
2456 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2457 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2458 			TAILQ_FOREACH(sp, &V_sptree[dir], chain)
2459 				cnt++;
2460 		}
2461 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2462 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain)
2463 				cnt++;
2464 		}
2465 	}
2466 
2467 	if (cnt == 0) {
2468 		SPTREE_RUNLOCK();
2469 		return key_senderror(so, m, ENOENT);
2470 	}
2471 
2472 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2473 		if (scope & IPSEC_POLICYSCOPE_GLOBAL) {
2474 			TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
2475 				--cnt;
2476 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2477 				    mhp->msg->sadb_msg_pid);
2478 
2479 				if (n != NULL)
2480 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2481 			}
2482 		}
2483 		if (scope & IPSEC_POLICYSCOPE_IFNET) {
2484 			TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain) {
2485 				--cnt;
2486 				n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2487 				    mhp->msg->sadb_msg_pid);
2488 
2489 				if (n != NULL)
2490 					key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2491 			}
2492 		}
2493 	}
2494 
2495 	SPTREE_RUNLOCK();
2496 	m_freem(m);
2497 	return (0);
2498 }
2499 
2500 static struct mbuf *
2501 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq,
2502     u_int32_t pid)
2503 {
2504 	struct mbuf *result = NULL, *m;
2505 	struct seclifetime lt;
2506 
2507 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid, sp->refcnt);
2508 	if (!m)
2509 		goto fail;
2510 	result = m;
2511 
2512 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2513 	    &sp->spidx.src.sa, sp->spidx.prefs,
2514 	    sp->spidx.ul_proto);
2515 	if (!m)
2516 		goto fail;
2517 	m_cat(result, m);
2518 
2519 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2520 	    &sp->spidx.dst.sa, sp->spidx.prefd,
2521 	    sp->spidx.ul_proto);
2522 	if (!m)
2523 		goto fail;
2524 	m_cat(result, m);
2525 
2526 	m = key_sp2mbuf(sp);
2527 	if (!m)
2528 		goto fail;
2529 	m_cat(result, m);
2530 
2531 	if(sp->lifetime){
2532 		lt.addtime=sp->created;
2533 		lt.usetime= sp->lastused;
2534 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_CURRENT);
2535 		if (!m)
2536 			goto fail;
2537 		m_cat(result, m);
2538 
2539 		lt.addtime=sp->lifetime;
2540 		lt.usetime= sp->validtime;
2541 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_HARD);
2542 		if (!m)
2543 			goto fail;
2544 		m_cat(result, m);
2545 	}
2546 
2547 	if ((result->m_flags & M_PKTHDR) == 0)
2548 		goto fail;
2549 
2550 	if (result->m_len < sizeof(struct sadb_msg)) {
2551 		result = m_pullup(result, sizeof(struct sadb_msg));
2552 		if (result == NULL)
2553 			goto fail;
2554 	}
2555 
2556 	result->m_pkthdr.len = 0;
2557 	for (m = result; m; m = m->m_next)
2558 		result->m_pkthdr.len += m->m_len;
2559 
2560 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2561 	    PFKEY_UNIT64(result->m_pkthdr.len);
2562 
2563 	return result;
2564 
2565 fail:
2566 	m_freem(result);
2567 	return NULL;
2568 }
2569 /*
2570  * get PFKEY message length for security policy and request.
2571  */
2572 static size_t
2573 key_getspreqmsglen(struct secpolicy *sp)
2574 {
2575 	size_t tlen, len;
2576 	int i;
2577 
2578 	tlen = sizeof(struct sadb_x_policy);
2579 	/* if is the policy for ipsec ? */
2580 	if (sp->policy != IPSEC_POLICY_IPSEC)
2581 		return (tlen);
2582 
2583 	/* get length of ipsec requests */
2584 	for (i = 0; i < sp->tcount; i++) {
2585 		len = sizeof(struct sadb_x_ipsecrequest)
2586 			+ sp->req[i]->saidx.src.sa.sa_len
2587 			+ sp->req[i]->saidx.dst.sa.sa_len;
2588 
2589 		tlen += PFKEY_ALIGN8(len);
2590 	}
2591 	return (tlen);
2592 }
2593 
2594 /*
2595  * SADB_SPDEXPIRE processing
2596  * send
2597  *   <base, address(SD), lifetime(CH), policy>
2598  * to KMD by PF_KEY.
2599  *
2600  * OUT:	0	: succeed
2601  *	others	: error number
2602  */
2603 static int
2604 key_spdexpire(struct secpolicy *sp)
2605 {
2606 	struct sadb_lifetime *lt;
2607 	struct mbuf *result = NULL, *m;
2608 	int len, error = -1;
2609 
2610 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2611 
2612 	KEYDBG(KEY_STAMP,
2613 	    printf("%s: SP(%p)\n", __func__, sp));
2614 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2615 
2616 	/* set msg header */
2617 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0);
2618 	if (!m) {
2619 		error = ENOBUFS;
2620 		goto fail;
2621 	}
2622 	result = m;
2623 
2624 	/* create lifetime extension (current and hard) */
2625 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
2626 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
2627 	if (m == NULL) {
2628 		error = ENOBUFS;
2629 		goto fail;
2630 	}
2631 	m_align(m, len);
2632 	m->m_len = len;
2633 	bzero(mtod(m, caddr_t), len);
2634 	lt = mtod(m, struct sadb_lifetime *);
2635 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2636 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
2637 	lt->sadb_lifetime_allocations = 0;
2638 	lt->sadb_lifetime_bytes = 0;
2639 	lt->sadb_lifetime_addtime = sp->created;
2640 	lt->sadb_lifetime_usetime = sp->lastused;
2641 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
2642 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2643 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
2644 	lt->sadb_lifetime_allocations = 0;
2645 	lt->sadb_lifetime_bytes = 0;
2646 	lt->sadb_lifetime_addtime = sp->lifetime;
2647 	lt->sadb_lifetime_usetime = sp->validtime;
2648 	m_cat(result, m);
2649 
2650 	/* set sadb_address for source */
2651 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2652 	    &sp->spidx.src.sa,
2653 	    sp->spidx.prefs, sp->spidx.ul_proto);
2654 	if (!m) {
2655 		error = ENOBUFS;
2656 		goto fail;
2657 	}
2658 	m_cat(result, m);
2659 
2660 	/* set sadb_address for destination */
2661 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2662 	    &sp->spidx.dst.sa,
2663 	    sp->spidx.prefd, sp->spidx.ul_proto);
2664 	if (!m) {
2665 		error = ENOBUFS;
2666 		goto fail;
2667 	}
2668 	m_cat(result, m);
2669 
2670 	/* set secpolicy */
2671 	m = key_sp2mbuf(sp);
2672 	if (!m) {
2673 		error = ENOBUFS;
2674 		goto fail;
2675 	}
2676 	m_cat(result, m);
2677 
2678 	if ((result->m_flags & M_PKTHDR) == 0) {
2679 		error = EINVAL;
2680 		goto fail;
2681 	}
2682 
2683 	if (result->m_len < sizeof(struct sadb_msg)) {
2684 		result = m_pullup(result, sizeof(struct sadb_msg));
2685 		if (result == NULL) {
2686 			error = ENOBUFS;
2687 			goto fail;
2688 		}
2689 	}
2690 
2691 	result->m_pkthdr.len = 0;
2692 	for (m = result; m; m = m->m_next)
2693 		result->m_pkthdr.len += m->m_len;
2694 
2695 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2696 	    PFKEY_UNIT64(result->m_pkthdr.len);
2697 
2698 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
2699 
2700  fail:
2701 	if (result)
2702 		m_freem(result);
2703 	return error;
2704 }
2705 
2706 /* %%% SAD management */
2707 /*
2708  * allocating and initialize new SA head.
2709  * OUT:	NULL	: failure due to the lack of memory.
2710  *	others	: pointer to new SA head.
2711  */
2712 static struct secashead *
2713 key_newsah(struct secasindex *saidx)
2714 {
2715 	struct secashead *sah;
2716 
2717 	sah = malloc(sizeof(struct secashead), M_IPSEC_SAH,
2718 	    M_NOWAIT | M_ZERO);
2719 	if (sah == NULL) {
2720 		PFKEYSTAT_INC(in_nomem);
2721 		return (NULL);
2722 	}
2723 	TAILQ_INIT(&sah->savtree_larval);
2724 	TAILQ_INIT(&sah->savtree_alive);
2725 	sah->saidx = *saidx;
2726 	sah->state = SADB_SASTATE_DEAD;
2727 	SAH_INITREF(sah);
2728 
2729 	KEYDBG(KEY_STAMP,
2730 	    printf("%s: SAH(%p)\n", __func__, sah));
2731 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2732 	return (sah);
2733 }
2734 
2735 static void
2736 key_freesah(struct secashead **psah)
2737 {
2738 	struct secashead *sah = *psah;
2739 
2740 	if (SAH_DELREF(sah) == 0)
2741 		return;
2742 
2743 	KEYDBG(KEY_STAMP,
2744 	    printf("%s: last reference to SAH(%p)\n", __func__, sah));
2745 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2746 
2747 	*psah = NULL;
2748 	key_delsah(sah);
2749 }
2750 
2751 static void
2752 key_delsah(struct secashead *sah)
2753 {
2754 	IPSEC_ASSERT(sah != NULL, ("NULL sah"));
2755 	IPSEC_ASSERT(sah->state == SADB_SASTATE_DEAD,
2756 	    ("Attempt to free non DEAD SAH %p", sah));
2757 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_larval),
2758 	    ("Attempt to free SAH %p with LARVAL SA", sah));
2759 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_alive),
2760 	    ("Attempt to free SAH %p with ALIVE SA", sah));
2761 
2762 	free(sah, M_IPSEC_SAH);
2763 }
2764 
2765 /*
2766  * allocating a new SA for key_add() and key_getspi() call,
2767  * and copy the values of mhp into new buffer.
2768  * When SAD message type is SADB_GETSPI set SA state to LARVAL.
2769  * For SADB_ADD create and initialize SA with MATURE state.
2770  * OUT:	NULL	: fail
2771  *	others	: pointer to new secasvar.
2772  */
2773 static struct secasvar *
2774 key_newsav(const struct sadb_msghdr *mhp, struct secasindex *saidx,
2775     uint32_t spi, int *errp)
2776 {
2777 	struct secashead *sah;
2778 	struct secasvar *sav;
2779 	int isnew;
2780 
2781 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2782 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2783 	IPSEC_ASSERT(mhp->msg->sadb_msg_type == SADB_GETSPI ||
2784 	    mhp->msg->sadb_msg_type == SADB_ADD, ("wrong message type"));
2785 
2786 	sav = NULL;
2787 	sah = NULL;
2788 	/* check SPI value */
2789 	switch (saidx->proto) {
2790 	case IPPROTO_ESP:
2791 	case IPPROTO_AH:
2792 		/*
2793 		 * RFC 4302, 2.4. Security Parameters Index (SPI), SPI values
2794 		 * 1-255 reserved by IANA for future use,
2795 		 * 0 for implementation specific, local use.
2796 		 */
2797 		if (ntohl(spi) <= 255) {
2798 			ipseclog((LOG_DEBUG, "%s: illegal range of SPI %u.\n",
2799 			    __func__, ntohl(spi)));
2800 			*errp = EINVAL;
2801 			goto done;
2802 		}
2803 		break;
2804 	}
2805 
2806 	sav = malloc(sizeof(struct secasvar), M_IPSEC_SA, M_NOWAIT | M_ZERO);
2807 	if (sav == NULL) {
2808 		*errp = ENOBUFS;
2809 		goto done;
2810 	}
2811 	sav->lock = malloc(sizeof(struct mtx), M_IPSEC_MISC,
2812 	    M_NOWAIT | M_ZERO);
2813 	if (sav->lock == NULL) {
2814 		*errp = ENOBUFS;
2815 		goto done;
2816 	}
2817 	mtx_init(sav->lock, "ipsec association", NULL, MTX_DEF);
2818 	sav->lft_c = uma_zalloc(V_key_lft_zone, M_NOWAIT);
2819 	if (sav->lft_c == NULL) {
2820 		*errp = ENOBUFS;
2821 		goto done;
2822 	}
2823 	counter_u64_zero(sav->lft_c_allocations);
2824 	counter_u64_zero(sav->lft_c_bytes);
2825 
2826 	sav->spi = spi;
2827 	sav->seq = mhp->msg->sadb_msg_seq;
2828 	sav->state = SADB_SASTATE_LARVAL;
2829 	sav->pid = (pid_t)mhp->msg->sadb_msg_pid;
2830 	SAV_INITREF(sav);
2831 again:
2832 	sah = key_getsah(saidx);
2833 	if (sah == NULL) {
2834 		/* create a new SA index */
2835 		sah = key_newsah(saidx);
2836 		if (sah == NULL) {
2837 			ipseclog((LOG_DEBUG,
2838 			    "%s: No more memory.\n", __func__));
2839 			*errp = ENOBUFS;
2840 			goto done;
2841 		}
2842 		isnew = 1;
2843 	} else
2844 		isnew = 0;
2845 
2846 	sav->sah = sah;
2847 	if (mhp->msg->sadb_msg_type == SADB_GETSPI) {
2848 		sav->created = time_second;
2849 	} else if (sav->state == SADB_SASTATE_LARVAL) {
2850 		/*
2851 		 * Do not call key_setsaval() second time in case
2852 		 * of `goto again`. We will have MATURE state.
2853 		 */
2854 		*errp = key_setsaval(sav, mhp);
2855 		if (*errp != 0)
2856 			goto done;
2857 		sav->state = SADB_SASTATE_MATURE;
2858 	}
2859 
2860 	SAHTREE_WLOCK();
2861 	/*
2862 	 * Check that existing SAH wasn't unlinked.
2863 	 * Since we didn't hold the SAHTREE lock, it is possible,
2864 	 * that callout handler or key_flush() or key_delete() could
2865 	 * unlink this SAH.
2866 	 */
2867 	if (isnew == 0 && sah->state == SADB_SASTATE_DEAD) {
2868 		SAHTREE_WUNLOCK();
2869 		key_freesah(&sah);	/* reference from key_getsah() */
2870 		goto again;
2871 	}
2872 	if (isnew != 0) {
2873 		/*
2874 		 * Add new SAH into SADB.
2875 		 *
2876 		 * XXXAE: we can serialize key_add and key_getspi calls, so
2877 		 * several threads will not fight in the race.
2878 		 * Otherwise we should check under SAHTREE lock, that this
2879 		 * SAH would not added twice.
2880 		 */
2881 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
2882 		/* Add new SAH into hash by addresses */
2883 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
2884 		/* Now we are linked in the chain */
2885 		sah->state = SADB_SASTATE_MATURE;
2886 		/*
2887 		 * SAV references this new SAH.
2888 		 * In case of existing SAH we reuse reference
2889 		 * from key_getsah().
2890 		 */
2891 		SAH_ADDREF(sah);
2892 	}
2893 	/* Link SAV with SAH */
2894 	if (sav->state == SADB_SASTATE_MATURE)
2895 		TAILQ_INSERT_HEAD(&sah->savtree_alive, sav, chain);
2896 	else
2897 		TAILQ_INSERT_HEAD(&sah->savtree_larval, sav, chain);
2898 	/* Add SAV into SPI hash */
2899 	LIST_INSERT_HEAD(SAVHASH_HASH(sav->spi), sav, spihash);
2900 	SAHTREE_WUNLOCK();
2901 	*errp = 0;	/* success */
2902 done:
2903 	if (*errp != 0) {
2904 		if (sav != NULL) {
2905 			if (sav->lock != NULL) {
2906 				mtx_destroy(sav->lock);
2907 				free(sav->lock, M_IPSEC_MISC);
2908 			}
2909 			if (sav->lft_c != NULL)
2910 				uma_zfree(V_key_lft_zone, sav->lft_c);
2911 			free(sav, M_IPSEC_SA), sav = NULL;
2912 		}
2913 		if (sah != NULL)
2914 			key_freesah(&sah);
2915 		if (*errp == ENOBUFS) {
2916 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
2917 			    __func__));
2918 			PFKEYSTAT_INC(in_nomem);
2919 		}
2920 	}
2921 	return (sav);
2922 }
2923 
2924 /*
2925  * free() SA variable entry.
2926  */
2927 static void
2928 key_cleansav(struct secasvar *sav)
2929 {
2930 
2931 	if (sav->natt != NULL) {
2932 		free(sav->natt, M_IPSEC_MISC);
2933 		sav->natt = NULL;
2934 	}
2935 	if (sav->flags & SADB_X_EXT_F_CLONED)
2936 		return;
2937 	/*
2938 	 * Cleanup xform state.  Note that zeroize'ing causes the
2939 	 * keys to be cleared; otherwise we must do it ourself.
2940 	 */
2941 	if (sav->tdb_xform != NULL) {
2942 		sav->tdb_xform->xf_zeroize(sav);
2943 		sav->tdb_xform = NULL;
2944 	} else {
2945 		if (sav->key_auth != NULL)
2946 			bzero(sav->key_auth->key_data, _KEYLEN(sav->key_auth));
2947 		if (sav->key_enc != NULL)
2948 			bzero(sav->key_enc->key_data, _KEYLEN(sav->key_enc));
2949 	}
2950 	if (sav->key_auth != NULL) {
2951 		if (sav->key_auth->key_data != NULL)
2952 			free(sav->key_auth->key_data, M_IPSEC_MISC);
2953 		free(sav->key_auth, M_IPSEC_MISC);
2954 		sav->key_auth = NULL;
2955 	}
2956 	if (sav->key_enc != NULL) {
2957 		if (sav->key_enc->key_data != NULL)
2958 			free(sav->key_enc->key_data, M_IPSEC_MISC);
2959 		free(sav->key_enc, M_IPSEC_MISC);
2960 		sav->key_enc = NULL;
2961 	}
2962 	if (sav->replay != NULL) {
2963 		if (sav->replay->bitmap != NULL)
2964 			free(sav->replay->bitmap, M_IPSEC_MISC);
2965 		free(sav->replay, M_IPSEC_MISC);
2966 		sav->replay = NULL;
2967 	}
2968 	if (sav->lft_h != NULL) {
2969 		free(sav->lft_h, M_IPSEC_MISC);
2970 		sav->lft_h = NULL;
2971 	}
2972 	if (sav->lft_s != NULL) {
2973 		free(sav->lft_s, M_IPSEC_MISC);
2974 		sav->lft_s = NULL;
2975 	}
2976 }
2977 
2978 /*
2979  * free() SA variable entry.
2980  */
2981 static void
2982 key_delsav(struct secasvar *sav)
2983 {
2984 	IPSEC_ASSERT(sav != NULL, ("null sav"));
2985 	IPSEC_ASSERT(sav->state == SADB_SASTATE_DEAD,
2986 	    ("attempt to free non DEAD SA %p", sav));
2987 	IPSEC_ASSERT(sav->refcnt == 0, ("reference count %u > 0",
2988 	    sav->refcnt));
2989 
2990 	/*
2991 	 * SA must be unlinked from the chain and hashtbl.
2992 	 * If SA was cloned, we leave all fields untouched,
2993 	 * except NAT-T config.
2994 	 */
2995 	key_cleansav(sav);
2996 	if ((sav->flags & SADB_X_EXT_F_CLONED) == 0) {
2997 		mtx_destroy(sav->lock);
2998 		free(sav->lock, M_IPSEC_MISC);
2999 		uma_zfree(V_key_lft_zone, sav->lft_c);
3000 	}
3001 	free(sav, M_IPSEC_SA);
3002 }
3003 
3004 /*
3005  * search SAH.
3006  * OUT:
3007  *	NULL	: not found
3008  *	others	: found, referenced pointer to a SAH.
3009  */
3010 static struct secashead *
3011 key_getsah(struct secasindex *saidx)
3012 {
3013 	SAHTREE_RLOCK_TRACKER;
3014 	struct secashead *sah;
3015 
3016 	SAHTREE_RLOCK();
3017 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
3018 	    if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID) != 0) {
3019 		    SAH_ADDREF(sah);
3020 		    break;
3021 	    }
3022 	}
3023 	SAHTREE_RUNLOCK();
3024 	return (sah);
3025 }
3026 
3027 /*
3028  * Check not to be duplicated SPI.
3029  * OUT:
3030  *	0	: not found
3031  *	1	: found SA with given SPI.
3032  */
3033 static int
3034 key_checkspidup(uint32_t spi)
3035 {
3036 	SAHTREE_RLOCK_TRACKER;
3037 	struct secasvar *sav;
3038 
3039 	/* Assume SPI is in network byte order */
3040 	SAHTREE_RLOCK();
3041 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3042 		if (sav->spi == spi)
3043 			break;
3044 	}
3045 	SAHTREE_RUNLOCK();
3046 	return (sav != NULL);
3047 }
3048 
3049 /*
3050  * Search SA by SPI.
3051  * OUT:
3052  *	NULL	: not found
3053  *	others	: found, referenced pointer to a SA.
3054  */
3055 static struct secasvar *
3056 key_getsavbyspi(uint32_t spi)
3057 {
3058 	SAHTREE_RLOCK_TRACKER;
3059 	struct secasvar *sav;
3060 
3061 	/* Assume SPI is in network byte order */
3062 	SAHTREE_RLOCK();
3063 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3064 		if (sav->spi != spi)
3065 			continue;
3066 		SAV_ADDREF(sav);
3067 		break;
3068 	}
3069 	SAHTREE_RUNLOCK();
3070 	return (sav);
3071 }
3072 
3073 static int
3074 key_updatelifetimes(struct secasvar *sav, const struct sadb_msghdr *mhp)
3075 {
3076 	struct seclifetime *lft_h, *lft_s, *tmp;
3077 
3078 	/* Lifetime extension is optional, check that it is present. */
3079 	if (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3080 	    SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) {
3081 		/*
3082 		 * In case of SADB_UPDATE we may need to change
3083 		 * existing lifetimes.
3084 		 */
3085 		if (sav->state == SADB_SASTATE_MATURE) {
3086 			lft_h = lft_s = NULL;
3087 			goto reset;
3088 		}
3089 		return (0);
3090 	}
3091 	/* Both HARD and SOFT extensions must present */
3092 	if ((SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3093 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
3094 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
3095 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
3096 		ipseclog((LOG_DEBUG,
3097 		    "%s: invalid message: missing required header.\n",
3098 		    __func__));
3099 		return (EINVAL);
3100 	}
3101 	if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD) ||
3102 	    SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_SOFT)) {
3103 		ipseclog((LOG_DEBUG,
3104 		    "%s: invalid message: wrong header size.\n", __func__));
3105 		return (EINVAL);
3106 	}
3107 	lft_h = key_dup_lifemsg((const struct sadb_lifetime *)
3108 	    mhp->ext[SADB_EXT_LIFETIME_HARD], M_IPSEC_MISC);
3109 	if (lft_h == NULL) {
3110 		PFKEYSTAT_INC(in_nomem);
3111 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3112 		return (ENOBUFS);
3113 	}
3114 	lft_s = key_dup_lifemsg((const struct sadb_lifetime *)
3115 	    mhp->ext[SADB_EXT_LIFETIME_SOFT], M_IPSEC_MISC);
3116 	if (lft_s == NULL) {
3117 		PFKEYSTAT_INC(in_nomem);
3118 		free(lft_h, M_IPSEC_MISC);
3119 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3120 		return (ENOBUFS);
3121 	}
3122 reset:
3123 	if (sav->state != SADB_SASTATE_LARVAL) {
3124 		/*
3125 		 * key_update() holds reference to this SA,
3126 		 * so it won't be deleted in meanwhile.
3127 		 */
3128 		SECASVAR_LOCK(sav);
3129 		tmp = sav->lft_h;
3130 		sav->lft_h = lft_h;
3131 		lft_h = tmp;
3132 
3133 		tmp = sav->lft_s;
3134 		sav->lft_s = lft_s;
3135 		lft_s = tmp;
3136 		SECASVAR_UNLOCK(sav);
3137 		if (lft_h != NULL)
3138 			free(lft_h, M_IPSEC_MISC);
3139 		if (lft_s != NULL)
3140 			free(lft_s, M_IPSEC_MISC);
3141 		return (0);
3142 	}
3143 	/* We can update lifetime without holding a lock */
3144 	IPSEC_ASSERT(sav->lft_h == NULL, ("lft_h is already initialized\n"));
3145 	IPSEC_ASSERT(sav->lft_s == NULL, ("lft_s is already initialized\n"));
3146 	sav->lft_h = lft_h;
3147 	sav->lft_s = lft_s;
3148 	return (0);
3149 }
3150 
3151 /*
3152  * copy SA values from PF_KEY message except *SPI, SEQ, PID and TYPE*.
3153  * You must update these if need. Expects only LARVAL SAs.
3154  * OUT:	0:	success.
3155  *	!0:	failure.
3156  */
3157 static int
3158 key_setsaval(struct secasvar *sav, const struct sadb_msghdr *mhp)
3159 {
3160 	const struct sadb_sa *sa0;
3161 	const struct sadb_key *key0;
3162 	uint32_t replay;
3163 	size_t len;
3164 	int error;
3165 
3166 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
3167 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
3168 	IPSEC_ASSERT(sav->state == SADB_SASTATE_LARVAL,
3169 	    ("Attempt to update non LARVAL SA"));
3170 
3171 	/* XXX rewrite */
3172 	error = key_setident(sav->sah, mhp);
3173 	if (error != 0)
3174 		goto fail;
3175 
3176 	/* SA */
3177 	if (!SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
3178 		if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
3179 			error = EINVAL;
3180 			goto fail;
3181 		}
3182 		sa0 = (const struct sadb_sa *)mhp->ext[SADB_EXT_SA];
3183 		sav->alg_auth = sa0->sadb_sa_auth;
3184 		sav->alg_enc = sa0->sadb_sa_encrypt;
3185 		sav->flags = sa0->sadb_sa_flags;
3186 		if ((sav->flags & SADB_KEY_FLAGS_MAX) != sav->flags) {
3187 			ipseclog((LOG_DEBUG,
3188 			    "%s: invalid sa_flags 0x%08x.\n", __func__,
3189 			    sav->flags));
3190 			error = EINVAL;
3191 			goto fail;
3192 		}
3193 
3194 		/* Optional replay window */
3195 		replay = 0;
3196 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0)
3197 			replay = sa0->sadb_sa_replay;
3198 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_SA_REPLAY)) {
3199 			if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA_REPLAY)) {
3200 				error = EINVAL;
3201 				goto fail;
3202 			}
3203 			replay = ((const struct sadb_x_sa_replay *)
3204 			    mhp->ext[SADB_X_EXT_SA_REPLAY])->sadb_x_sa_replay_replay;
3205 
3206 			if (replay > UINT32_MAX - 32) {
3207 				ipseclog((LOG_DEBUG,
3208 				    "%s: replay window too big.\n", __func__));
3209 				error = EINVAL;
3210 				goto fail;
3211 			}
3212 
3213 			replay = (replay + 7) >> 3;
3214 		}
3215 
3216 		sav->replay = malloc(sizeof(struct secreplay), M_IPSEC_MISC,
3217 		    M_NOWAIT | M_ZERO);
3218 		if (sav->replay == NULL) {
3219 			PFKEYSTAT_INC(in_nomem);
3220 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3221 			    __func__));
3222 			error = ENOBUFS;
3223 			goto fail;
3224 		}
3225 
3226 		if (replay != 0) {
3227 			/* number of 32b blocks to be allocated */
3228 			uint32_t bitmap_size;
3229 
3230 			/* RFC 6479:
3231 			 * - the allocated replay window size must be
3232 			 *   a power of two.
3233 			 * - use an extra 32b block as a redundant window.
3234 			 */
3235 			bitmap_size = 1;
3236 			while (replay + 4 > bitmap_size)
3237 				bitmap_size <<= 1;
3238 			bitmap_size = bitmap_size / 4;
3239 
3240 			sav->replay->bitmap = malloc(
3241 			    bitmap_size * sizeof(uint32_t), M_IPSEC_MISC,
3242 			    M_NOWAIT | M_ZERO);
3243 			if (sav->replay->bitmap == NULL) {
3244 				PFKEYSTAT_INC(in_nomem);
3245 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3246 					__func__));
3247 				error = ENOBUFS;
3248 				goto fail;
3249 			}
3250 			sav->replay->bitmap_size = bitmap_size;
3251 			sav->replay->wsize = replay;
3252 		}
3253 	}
3254 
3255 	/* Authentication keys */
3256 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
3257 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH)) {
3258 			error = EINVAL;
3259 			goto fail;
3260 		}
3261 		error = 0;
3262 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_AUTH];
3263 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3264 		switch (mhp->msg->sadb_msg_satype) {
3265 		case SADB_SATYPE_AH:
3266 		case SADB_SATYPE_ESP:
3267 		case SADB_X_SATYPE_TCPSIGNATURE:
3268 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3269 			    sav->alg_auth != SADB_X_AALG_NULL)
3270 				error = EINVAL;
3271 			break;
3272 		case SADB_X_SATYPE_IPCOMP:
3273 		default:
3274 			error = EINVAL;
3275 			break;
3276 		}
3277 		if (error) {
3278 			ipseclog((LOG_DEBUG, "%s: invalid key_auth values.\n",
3279 				__func__));
3280 			goto fail;
3281 		}
3282 
3283 		sav->key_auth = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3284 		if (sav->key_auth == NULL ) {
3285 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3286 				  __func__));
3287 			PFKEYSTAT_INC(in_nomem);
3288 			error = ENOBUFS;
3289 			goto fail;
3290 		}
3291 	}
3292 
3293 	/* Encryption key */
3294 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) {
3295 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT)) {
3296 			error = EINVAL;
3297 			goto fail;
3298 		}
3299 		error = 0;
3300 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_ENCRYPT];
3301 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3302 		switch (mhp->msg->sadb_msg_satype) {
3303 		case SADB_SATYPE_ESP:
3304 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3305 			    sav->alg_enc != SADB_EALG_NULL) {
3306 				error = EINVAL;
3307 				break;
3308 			}
3309 			sav->key_enc = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3310 			if (sav->key_enc == NULL) {
3311 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3312 					__func__));
3313 				PFKEYSTAT_INC(in_nomem);
3314 				error = ENOBUFS;
3315 				goto fail;
3316 			}
3317 			break;
3318 		case SADB_X_SATYPE_IPCOMP:
3319 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3320 				error = EINVAL;
3321 			sav->key_enc = NULL;	/*just in case*/
3322 			break;
3323 		case SADB_SATYPE_AH:
3324 		case SADB_X_SATYPE_TCPSIGNATURE:
3325 		default:
3326 			error = EINVAL;
3327 			break;
3328 		}
3329 		if (error) {
3330 			ipseclog((LOG_DEBUG, "%s: invalid key_enc value.\n",
3331 				__func__));
3332 			goto fail;
3333 		}
3334 	}
3335 
3336 	/* set iv */
3337 	sav->ivlen = 0;
3338 	switch (mhp->msg->sadb_msg_satype) {
3339 	case SADB_SATYPE_AH:
3340 		if (sav->flags & SADB_X_EXT_DERIV) {
3341 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3342 			    "given to AH SA.\n", __func__));
3343 			error = EINVAL;
3344 			goto fail;
3345 		}
3346 		if (sav->alg_enc != SADB_EALG_NONE) {
3347 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3348 			    "mismated.\n", __func__));
3349 			error = EINVAL;
3350 			goto fail;
3351 		}
3352 		error = xform_init(sav, XF_AH);
3353 		break;
3354 	case SADB_SATYPE_ESP:
3355 		if ((sav->flags & (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) ==
3356 		    (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) {
3357 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3358 			    "given to old-esp.\n", __func__));
3359 			error = EINVAL;
3360 			goto fail;
3361 		}
3362 		error = xform_init(sav, XF_ESP);
3363 		break;
3364 	case SADB_X_SATYPE_IPCOMP:
3365 		if (sav->alg_auth != SADB_AALG_NONE) {
3366 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3367 			    "mismated.\n", __func__));
3368 			error = EINVAL;
3369 			goto fail;
3370 		}
3371 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0 &&
3372 		    ntohl(sav->spi) >= 0x10000) {
3373 			ipseclog((LOG_DEBUG, "%s: invalid cpi for IPComp.\n",
3374 			    __func__));
3375 			error = EINVAL;
3376 			goto fail;
3377 		}
3378 		error = xform_init(sav, XF_IPCOMP);
3379 		break;
3380 	case SADB_X_SATYPE_TCPSIGNATURE:
3381 		if (sav->alg_enc != SADB_EALG_NONE) {
3382 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3383 			    "mismated.\n", __func__));
3384 			error = EINVAL;
3385 			goto fail;
3386 		}
3387 		error = xform_init(sav, XF_TCPSIGNATURE);
3388 		break;
3389 	default:
3390 		ipseclog((LOG_DEBUG, "%s: Invalid satype.\n", __func__));
3391 		error = EPROTONOSUPPORT;
3392 		goto fail;
3393 	}
3394 	if (error) {
3395 		ipseclog((LOG_DEBUG, "%s: unable to initialize SA type %u.\n",
3396 		    __func__, mhp->msg->sadb_msg_satype));
3397 		goto fail;
3398 	}
3399 
3400 	/* Handle NAT-T headers */
3401 	error = key_setnatt(sav, mhp);
3402 	if (error != 0)
3403 		goto fail;
3404 
3405 	/* Initialize lifetime for CURRENT */
3406 	sav->firstused = 0;
3407 	sav->created = time_second;
3408 
3409 	/* lifetimes for HARD and SOFT */
3410 	error = key_updatelifetimes(sav, mhp);
3411 	if (error == 0)
3412 		return (0);
3413 fail:
3414 	key_cleansav(sav);
3415 	return (error);
3416 }
3417 
3418 /*
3419  * subroutine for SADB_GET and SADB_DUMP.
3420  */
3421 static struct mbuf *
3422 key_setdumpsa(struct secasvar *sav, uint8_t type, uint8_t satype,
3423     uint32_t seq, uint32_t pid)
3424 {
3425 	struct seclifetime lft_c;
3426 	struct mbuf *result = NULL, *tres = NULL, *m;
3427 	int i, dumporder[] = {
3428 		SADB_EXT_SA, SADB_X_EXT_SA2, SADB_X_EXT_SA_REPLAY,
3429 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3430 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3431 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY,
3432 		SADB_EXT_KEY_AUTH, SADB_EXT_KEY_ENCRYPT,
3433 		SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
3434 		SADB_EXT_SENSITIVITY,
3435 		SADB_X_EXT_NAT_T_TYPE,
3436 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3437 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3438 		SADB_X_EXT_NAT_T_FRAG,
3439 	};
3440 	uint32_t replay_count;
3441 
3442 	m = key_setsadbmsg(type, 0, satype, seq, pid, sav->refcnt);
3443 	if (m == NULL)
3444 		goto fail;
3445 	result = m;
3446 
3447 	for (i = nitems(dumporder) - 1; i >= 0; i--) {
3448 		m = NULL;
3449 		switch (dumporder[i]) {
3450 		case SADB_EXT_SA:
3451 			m = key_setsadbsa(sav);
3452 			if (!m)
3453 				goto fail;
3454 			break;
3455 
3456 		case SADB_X_EXT_SA2:
3457 			SECASVAR_LOCK(sav);
3458 			replay_count = sav->replay ? sav->replay->count : 0;
3459 			SECASVAR_UNLOCK(sav);
3460 			m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
3461 					sav->sah->saidx.reqid);
3462 			if (!m)
3463 				goto fail;
3464 			break;
3465 
3466 		case SADB_X_EXT_SA_REPLAY:
3467 			if (sav->replay == NULL ||
3468 			    sav->replay->wsize <= UINT8_MAX)
3469 				continue;
3470 
3471 			m = key_setsadbxsareplay(sav->replay->wsize);
3472 			if (!m)
3473 				goto fail;
3474 			break;
3475 
3476 		case SADB_EXT_ADDRESS_SRC:
3477 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3478 			    &sav->sah->saidx.src.sa,
3479 			    FULLMASK, IPSEC_ULPROTO_ANY);
3480 			if (!m)
3481 				goto fail;
3482 			break;
3483 
3484 		case SADB_EXT_ADDRESS_DST:
3485 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3486 			    &sav->sah->saidx.dst.sa,
3487 			    FULLMASK, IPSEC_ULPROTO_ANY);
3488 			if (!m)
3489 				goto fail;
3490 			break;
3491 
3492 		case SADB_EXT_KEY_AUTH:
3493 			if (!sav->key_auth)
3494 				continue;
3495 			m = key_setkey(sav->key_auth, SADB_EXT_KEY_AUTH);
3496 			if (!m)
3497 				goto fail;
3498 			break;
3499 
3500 		case SADB_EXT_KEY_ENCRYPT:
3501 			if (!sav->key_enc)
3502 				continue;
3503 			m = key_setkey(sav->key_enc, SADB_EXT_KEY_ENCRYPT);
3504 			if (!m)
3505 				goto fail;
3506 			break;
3507 
3508 		case SADB_EXT_LIFETIME_CURRENT:
3509 			lft_c.addtime = sav->created;
3510 			lft_c.allocations = (uint32_t)counter_u64_fetch(
3511 			    sav->lft_c_allocations);
3512 			lft_c.bytes = counter_u64_fetch(sav->lft_c_bytes);
3513 			lft_c.usetime = sav->firstused;
3514 			m = key_setlifetime(&lft_c, SADB_EXT_LIFETIME_CURRENT);
3515 			if (!m)
3516 				goto fail;
3517 			break;
3518 
3519 		case SADB_EXT_LIFETIME_HARD:
3520 			if (!sav->lft_h)
3521 				continue;
3522 			m = key_setlifetime(sav->lft_h,
3523 					    SADB_EXT_LIFETIME_HARD);
3524 			if (!m)
3525 				goto fail;
3526 			break;
3527 
3528 		case SADB_EXT_LIFETIME_SOFT:
3529 			if (!sav->lft_s)
3530 				continue;
3531 			m = key_setlifetime(sav->lft_s,
3532 					    SADB_EXT_LIFETIME_SOFT);
3533 
3534 			if (!m)
3535 				goto fail;
3536 			break;
3537 
3538 		case SADB_X_EXT_NAT_T_TYPE:
3539 			if (sav->natt == NULL)
3540 				continue;
3541 			m = key_setsadbxtype(UDP_ENCAP_ESPINUDP);
3542 			if (!m)
3543 				goto fail;
3544 			break;
3545 
3546 		case SADB_X_EXT_NAT_T_DPORT:
3547 			if (sav->natt == NULL)
3548 				continue;
3549 			m = key_setsadbxport(sav->natt->dport,
3550 			    SADB_X_EXT_NAT_T_DPORT);
3551 			if (!m)
3552 				goto fail;
3553 			break;
3554 
3555 		case SADB_X_EXT_NAT_T_SPORT:
3556 			if (sav->natt == NULL)
3557 				continue;
3558 			m = key_setsadbxport(sav->natt->sport,
3559 			    SADB_X_EXT_NAT_T_SPORT);
3560 			if (!m)
3561 				goto fail;
3562 			break;
3563 
3564 		case SADB_X_EXT_NAT_T_OAI:
3565 			if (sav->natt == NULL ||
3566 			    (sav->natt->flags & IPSEC_NATT_F_OAI) == 0)
3567 				continue;
3568 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAI,
3569 			    &sav->natt->oai.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3570 			if (!m)
3571 				goto fail;
3572 			break;
3573 		case SADB_X_EXT_NAT_T_OAR:
3574 			if (sav->natt == NULL ||
3575 			    (sav->natt->flags & IPSEC_NATT_F_OAR) == 0)
3576 				continue;
3577 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAR,
3578 			    &sav->natt->oar.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3579 			if (!m)
3580 				goto fail;
3581 			break;
3582 		case SADB_X_EXT_NAT_T_FRAG:
3583 			/* We do not (yet) support those. */
3584 			continue;
3585 
3586 		case SADB_EXT_ADDRESS_PROXY:
3587 		case SADB_EXT_IDENTITY_SRC:
3588 		case SADB_EXT_IDENTITY_DST:
3589 			/* XXX: should we brought from SPD ? */
3590 		case SADB_EXT_SENSITIVITY:
3591 		default:
3592 			continue;
3593 		}
3594 
3595 		if (!m)
3596 			goto fail;
3597 		if (tres)
3598 			m_cat(m, tres);
3599 		tres = m;
3600 	}
3601 
3602 	m_cat(result, tres);
3603 	tres = NULL;
3604 	if (result->m_len < sizeof(struct sadb_msg)) {
3605 		result = m_pullup(result, sizeof(struct sadb_msg));
3606 		if (result == NULL)
3607 			goto fail;
3608 	}
3609 
3610 	result->m_pkthdr.len = 0;
3611 	for (m = result; m; m = m->m_next)
3612 		result->m_pkthdr.len += m->m_len;
3613 
3614 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3615 	    PFKEY_UNIT64(result->m_pkthdr.len);
3616 
3617 	return result;
3618 
3619 fail:
3620 	m_freem(result);
3621 	m_freem(tres);
3622 	return NULL;
3623 }
3624 
3625 /*
3626  * set data into sadb_msg.
3627  */
3628 static struct mbuf *
3629 key_setsadbmsg(u_int8_t type, u_int16_t tlen, u_int8_t satype, u_int32_t seq,
3630     pid_t pid, u_int16_t reserved)
3631 {
3632 	struct mbuf *m;
3633 	struct sadb_msg *p;
3634 	int len;
3635 
3636 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
3637 	if (len > MCLBYTES)
3638 		return NULL;
3639 	MGETHDR(m, M_NOWAIT, MT_DATA);
3640 	if (m && len > MHLEN) {
3641 		if (!(MCLGET(m, M_NOWAIT))) {
3642 			m_freem(m);
3643 			m = NULL;
3644 		}
3645 	}
3646 	if (!m)
3647 		return NULL;
3648 	m->m_pkthdr.len = m->m_len = len;
3649 	m->m_next = NULL;
3650 
3651 	p = mtod(m, struct sadb_msg *);
3652 
3653 	bzero(p, len);
3654 	p->sadb_msg_version = PF_KEY_V2;
3655 	p->sadb_msg_type = type;
3656 	p->sadb_msg_errno = 0;
3657 	p->sadb_msg_satype = satype;
3658 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
3659 	p->sadb_msg_reserved = reserved;
3660 	p->sadb_msg_seq = seq;
3661 	p->sadb_msg_pid = (u_int32_t)pid;
3662 
3663 	return m;
3664 }
3665 
3666 /*
3667  * copy secasvar data into sadb_address.
3668  */
3669 static struct mbuf *
3670 key_setsadbsa(struct secasvar *sav)
3671 {
3672 	struct mbuf *m;
3673 	struct sadb_sa *p;
3674 	int len;
3675 
3676 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
3677 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3678 	if (m == NULL)
3679 		return (NULL);
3680 	m_align(m, len);
3681 	m->m_len = len;
3682 	p = mtod(m, struct sadb_sa *);
3683 	bzero(p, len);
3684 	p->sadb_sa_len = PFKEY_UNIT64(len);
3685 	p->sadb_sa_exttype = SADB_EXT_SA;
3686 	p->sadb_sa_spi = sav->spi;
3687 	p->sadb_sa_replay = sav->replay ?
3688 	    (sav->replay->wsize > UINT8_MAX ? UINT8_MAX :
3689 		sav->replay->wsize): 0;
3690 	p->sadb_sa_state = sav->state;
3691 	p->sadb_sa_auth = sav->alg_auth;
3692 	p->sadb_sa_encrypt = sav->alg_enc;
3693 	p->sadb_sa_flags = sav->flags & SADB_KEY_FLAGS_MAX;
3694 	return (m);
3695 }
3696 
3697 /*
3698  * set data into sadb_address.
3699  */
3700 static struct mbuf *
3701 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
3702     u_int8_t prefixlen, u_int16_t ul_proto)
3703 {
3704 	struct mbuf *m;
3705 	struct sadb_address *p;
3706 	size_t len;
3707 
3708 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
3709 	    PFKEY_ALIGN8(saddr->sa_len);
3710 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3711 	if (m == NULL)
3712 		return (NULL);
3713 	m_align(m, len);
3714 	m->m_len = len;
3715 	p = mtod(m, struct sadb_address *);
3716 
3717 	bzero(p, len);
3718 	p->sadb_address_len = PFKEY_UNIT64(len);
3719 	p->sadb_address_exttype = exttype;
3720 	p->sadb_address_proto = ul_proto;
3721 	if (prefixlen == FULLMASK) {
3722 		switch (saddr->sa_family) {
3723 		case AF_INET:
3724 			prefixlen = sizeof(struct in_addr) << 3;
3725 			break;
3726 		case AF_INET6:
3727 			prefixlen = sizeof(struct in6_addr) << 3;
3728 			break;
3729 		default:
3730 			; /*XXX*/
3731 		}
3732 	}
3733 	p->sadb_address_prefixlen = prefixlen;
3734 	p->sadb_address_reserved = 0;
3735 
3736 	bcopy(saddr,
3737 	    mtod(m, caddr_t) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
3738 	    saddr->sa_len);
3739 
3740 	return m;
3741 }
3742 
3743 /*
3744  * set data into sadb_x_sa2.
3745  */
3746 static struct mbuf *
3747 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int32_t reqid)
3748 {
3749 	struct mbuf *m;
3750 	struct sadb_x_sa2 *p;
3751 	size_t len;
3752 
3753 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
3754 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3755 	if (m == NULL)
3756 		return (NULL);
3757 	m_align(m, len);
3758 	m->m_len = len;
3759 	p = mtod(m, struct sadb_x_sa2 *);
3760 
3761 	bzero(p, len);
3762 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
3763 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
3764 	p->sadb_x_sa2_mode = mode;
3765 	p->sadb_x_sa2_reserved1 = 0;
3766 	p->sadb_x_sa2_reserved2 = 0;
3767 	p->sadb_x_sa2_sequence = seq;
3768 	p->sadb_x_sa2_reqid = reqid;
3769 
3770 	return m;
3771 }
3772 
3773 /*
3774  * Set data into sadb_x_sa_replay.
3775  */
3776 static struct mbuf *
3777 key_setsadbxsareplay(u_int32_t replay)
3778 {
3779 	struct mbuf *m;
3780 	struct sadb_x_sa_replay *p;
3781 	size_t len;
3782 
3783 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa_replay));
3784 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3785 	if (m == NULL)
3786 		return (NULL);
3787 	m_align(m, len);
3788 	m->m_len = len;
3789 	p = mtod(m, struct sadb_x_sa_replay *);
3790 
3791 	bzero(p, len);
3792 	p->sadb_x_sa_replay_len = PFKEY_UNIT64(len);
3793 	p->sadb_x_sa_replay_exttype = SADB_X_EXT_SA_REPLAY;
3794 	p->sadb_x_sa_replay_replay = (replay << 3);
3795 
3796 	return m;
3797 }
3798 
3799 /*
3800  * Set a type in sadb_x_nat_t_type.
3801  */
3802 static struct mbuf *
3803 key_setsadbxtype(u_int16_t type)
3804 {
3805 	struct mbuf *m;
3806 	size_t len;
3807 	struct sadb_x_nat_t_type *p;
3808 
3809 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
3810 
3811 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3812 	if (m == NULL)
3813 		return (NULL);
3814 	m_align(m, len);
3815 	m->m_len = len;
3816 	p = mtod(m, struct sadb_x_nat_t_type *);
3817 
3818 	bzero(p, len);
3819 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
3820 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
3821 	p->sadb_x_nat_t_type_type = type;
3822 
3823 	return (m);
3824 }
3825 /*
3826  * Set a port in sadb_x_nat_t_port.
3827  * In contrast to default RFC 2367 behaviour, port is in network byte order.
3828  */
3829 static struct mbuf *
3830 key_setsadbxport(u_int16_t port, u_int16_t type)
3831 {
3832 	struct mbuf *m;
3833 	size_t len;
3834 	struct sadb_x_nat_t_port *p;
3835 
3836 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
3837 
3838 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3839 	if (m == NULL)
3840 		return (NULL);
3841 	m_align(m, len);
3842 	m->m_len = len;
3843 	p = mtod(m, struct sadb_x_nat_t_port *);
3844 
3845 	bzero(p, len);
3846 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
3847 	p->sadb_x_nat_t_port_exttype = type;
3848 	p->sadb_x_nat_t_port_port = port;
3849 
3850 	return (m);
3851 }
3852 
3853 /*
3854  * Get port from sockaddr. Port is in network byte order.
3855  */
3856 uint16_t
3857 key_portfromsaddr(struct sockaddr *sa)
3858 {
3859 
3860 	switch (sa->sa_family) {
3861 #ifdef INET
3862 	case AF_INET:
3863 		return ((struct sockaddr_in *)sa)->sin_port;
3864 #endif
3865 #ifdef INET6
3866 	case AF_INET6:
3867 		return ((struct sockaddr_in6 *)sa)->sin6_port;
3868 #endif
3869 	}
3870 	return (0);
3871 }
3872 
3873 /*
3874  * Set port in struct sockaddr. Port is in network byte order.
3875  */
3876 void
3877 key_porttosaddr(struct sockaddr *sa, uint16_t port)
3878 {
3879 
3880 	switch (sa->sa_family) {
3881 #ifdef INET
3882 	case AF_INET:
3883 		((struct sockaddr_in *)sa)->sin_port = port;
3884 		break;
3885 #endif
3886 #ifdef INET6
3887 	case AF_INET6:
3888 		((struct sockaddr_in6 *)sa)->sin6_port = port;
3889 		break;
3890 #endif
3891 	default:
3892 		ipseclog((LOG_DEBUG, "%s: unexpected address family %d.\n",
3893 			__func__, sa->sa_family));
3894 		break;
3895 	}
3896 }
3897 
3898 /*
3899  * set data into sadb_x_policy
3900  */
3901 static struct mbuf *
3902 key_setsadbxpolicy(u_int16_t type, u_int8_t dir, u_int32_t id, u_int32_t priority)
3903 {
3904 	struct mbuf *m;
3905 	struct sadb_x_policy *p;
3906 	size_t len;
3907 
3908 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
3909 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3910 	if (m == NULL)
3911 		return (NULL);
3912 	m_align(m, len);
3913 	m->m_len = len;
3914 	p = mtod(m, struct sadb_x_policy *);
3915 
3916 	bzero(p, len);
3917 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
3918 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
3919 	p->sadb_x_policy_type = type;
3920 	p->sadb_x_policy_dir = dir;
3921 	p->sadb_x_policy_id = id;
3922 	p->sadb_x_policy_priority = priority;
3923 
3924 	return m;
3925 }
3926 
3927 /* %%% utilities */
3928 /* Take a key message (sadb_key) from the socket and turn it into one
3929  * of the kernel's key structures (seckey).
3930  *
3931  * IN: pointer to the src
3932  * OUT: NULL no more memory
3933  */
3934 struct seckey *
3935 key_dup_keymsg(const struct sadb_key *src, size_t len,
3936     struct malloc_type *type)
3937 {
3938 	struct seckey *dst;
3939 
3940 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
3941 	if (dst != NULL) {
3942 		dst->bits = src->sadb_key_bits;
3943 		dst->key_data = malloc(len, type, M_NOWAIT);
3944 		if (dst->key_data != NULL) {
3945 			bcopy((const char *)(src + 1), dst->key_data, len);
3946 		} else {
3947 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3948 			    __func__));
3949 			free(dst, type);
3950 			dst = NULL;
3951 		}
3952 	} else {
3953 		ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3954 		    __func__));
3955 
3956 	}
3957 	return (dst);
3958 }
3959 
3960 /* Take a lifetime message (sadb_lifetime) passed in on a socket and
3961  * turn it into one of the kernel's lifetime structures (seclifetime).
3962  *
3963  * IN: pointer to the destination, source and malloc type
3964  * OUT: NULL, no more memory
3965  */
3966 
3967 static struct seclifetime *
3968 key_dup_lifemsg(const struct sadb_lifetime *src, struct malloc_type *type)
3969 {
3970 	struct seclifetime *dst;
3971 
3972 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
3973 	if (dst == NULL) {
3974 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3975 		return (NULL);
3976 	}
3977 	dst->allocations = src->sadb_lifetime_allocations;
3978 	dst->bytes = src->sadb_lifetime_bytes;
3979 	dst->addtime = src->sadb_lifetime_addtime;
3980 	dst->usetime = src->sadb_lifetime_usetime;
3981 	return (dst);
3982 }
3983 
3984 /*
3985  * compare two secasindex structure.
3986  * flag can specify to compare 2 saidxes.
3987  * compare two secasindex structure without both mode and reqid.
3988  * don't compare port.
3989  * IN:
3990  *      saidx0: source, it can be in SAD.
3991  *      saidx1: object.
3992  * OUT:
3993  *      1 : equal
3994  *      0 : not equal
3995  */
3996 static int
3997 key_cmpsaidx(const struct secasindex *saidx0, const struct secasindex *saidx1,
3998     int flag)
3999 {
4000 
4001 	/* sanity */
4002 	if (saidx0 == NULL && saidx1 == NULL)
4003 		return 1;
4004 
4005 	if (saidx0 == NULL || saidx1 == NULL)
4006 		return 0;
4007 
4008 	if (saidx0->proto != saidx1->proto)
4009 		return 0;
4010 
4011 	if (flag == CMP_EXACTLY) {
4012 		if (saidx0->mode != saidx1->mode)
4013 			return 0;
4014 		if (saidx0->reqid != saidx1->reqid)
4015 			return 0;
4016 		if (bcmp(&saidx0->src, &saidx1->src,
4017 		    saidx0->src.sa.sa_len) != 0 ||
4018 		    bcmp(&saidx0->dst, &saidx1->dst,
4019 		    saidx0->dst.sa.sa_len) != 0)
4020 			return 0;
4021 	} else {
4022 
4023 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
4024 		if (flag == CMP_MODE_REQID || flag == CMP_REQID) {
4025 			/*
4026 			 * If reqid of SPD is non-zero, unique SA is required.
4027 			 * The result must be of same reqid in this case.
4028 			 */
4029 			if (saidx1->reqid != 0 &&
4030 			    saidx0->reqid != saidx1->reqid)
4031 				return 0;
4032 		}
4033 
4034 		if (flag == CMP_MODE_REQID) {
4035 			if (saidx0->mode != IPSEC_MODE_ANY
4036 			 && saidx0->mode != saidx1->mode)
4037 				return 0;
4038 		}
4039 
4040 		if (key_sockaddrcmp(&saidx0->src.sa, &saidx1->src.sa, 0) != 0)
4041 			return 0;
4042 		if (key_sockaddrcmp(&saidx0->dst.sa, &saidx1->dst.sa, 0) != 0)
4043 			return 0;
4044 	}
4045 
4046 	return 1;
4047 }
4048 
4049 /*
4050  * compare two secindex structure exactly.
4051  * IN:
4052  *	spidx0: source, it is often in SPD.
4053  *	spidx1: object, it is often from PFKEY message.
4054  * OUT:
4055  *	1 : equal
4056  *	0 : not equal
4057  */
4058 static int
4059 key_cmpspidx_exactly(struct secpolicyindex *spidx0,
4060     struct secpolicyindex *spidx1)
4061 {
4062 	/* sanity */
4063 	if (spidx0 == NULL && spidx1 == NULL)
4064 		return 1;
4065 
4066 	if (spidx0 == NULL || spidx1 == NULL)
4067 		return 0;
4068 
4069 	if (spidx0->prefs != spidx1->prefs
4070 	 || spidx0->prefd != spidx1->prefd
4071 	 || spidx0->ul_proto != spidx1->ul_proto)
4072 		return 0;
4073 
4074 	return key_sockaddrcmp(&spidx0->src.sa, &spidx1->src.sa, 1) == 0 &&
4075 	       key_sockaddrcmp(&spidx0->dst.sa, &spidx1->dst.sa, 1) == 0;
4076 }
4077 
4078 /*
4079  * compare two secindex structure with mask.
4080  * IN:
4081  *	spidx0: source, it is often in SPD.
4082  *	spidx1: object, it is often from IP header.
4083  * OUT:
4084  *	1 : equal
4085  *	0 : not equal
4086  */
4087 static int
4088 key_cmpspidx_withmask(struct secpolicyindex *spidx0,
4089     struct secpolicyindex *spidx1)
4090 {
4091 	/* sanity */
4092 	if (spidx0 == NULL && spidx1 == NULL)
4093 		return 1;
4094 
4095 	if (spidx0 == NULL || spidx1 == NULL)
4096 		return 0;
4097 
4098 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4099 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4100 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4101 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
4102 		return 0;
4103 
4104 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4105 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY
4106 	 && spidx0->ul_proto != spidx1->ul_proto)
4107 		return 0;
4108 
4109 	switch (spidx0->src.sa.sa_family) {
4110 	case AF_INET:
4111 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY
4112 		 && spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
4113 			return 0;
4114 		if (!key_bbcmp(&spidx0->src.sin.sin_addr,
4115 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
4116 			return 0;
4117 		break;
4118 	case AF_INET6:
4119 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY
4120 		 && spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
4121 			return 0;
4122 		/*
4123 		 * scope_id check. if sin6_scope_id is 0, we regard it
4124 		 * as a wildcard scope, which matches any scope zone ID.
4125 		 */
4126 		if (spidx0->src.sin6.sin6_scope_id &&
4127 		    spidx1->src.sin6.sin6_scope_id &&
4128 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
4129 			return 0;
4130 		if (!key_bbcmp(&spidx0->src.sin6.sin6_addr,
4131 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
4132 			return 0;
4133 		break;
4134 	default:
4135 		/* XXX */
4136 		if (bcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
4137 			return 0;
4138 		break;
4139 	}
4140 
4141 	switch (spidx0->dst.sa.sa_family) {
4142 	case AF_INET:
4143 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY
4144 		 && spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
4145 			return 0;
4146 		if (!key_bbcmp(&spidx0->dst.sin.sin_addr,
4147 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
4148 			return 0;
4149 		break;
4150 	case AF_INET6:
4151 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY
4152 		 && spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
4153 			return 0;
4154 		/*
4155 		 * scope_id check. if sin6_scope_id is 0, we regard it
4156 		 * as a wildcard scope, which matches any scope zone ID.
4157 		 */
4158 		if (spidx0->dst.sin6.sin6_scope_id &&
4159 		    spidx1->dst.sin6.sin6_scope_id &&
4160 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
4161 			return 0;
4162 		if (!key_bbcmp(&spidx0->dst.sin6.sin6_addr,
4163 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
4164 			return 0;
4165 		break;
4166 	default:
4167 		/* XXX */
4168 		if (bcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
4169 			return 0;
4170 		break;
4171 	}
4172 
4173 	/* XXX Do we check other field ?  e.g. flowinfo */
4174 
4175 	return 1;
4176 }
4177 
4178 #ifdef satosin
4179 #undef satosin
4180 #endif
4181 #define satosin(s) ((const struct sockaddr_in *)s)
4182 #ifdef satosin6
4183 #undef satosin6
4184 #endif
4185 #define satosin6(s) ((const struct sockaddr_in6 *)s)
4186 /* returns 0 on match */
4187 int
4188 key_sockaddrcmp(const struct sockaddr *sa1, const struct sockaddr *sa2,
4189     int port)
4190 {
4191 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4192 		return 1;
4193 
4194 	switch (sa1->sa_family) {
4195 #ifdef INET
4196 	case AF_INET:
4197 		if (sa1->sa_len != sizeof(struct sockaddr_in))
4198 			return 1;
4199 		if (satosin(sa1)->sin_addr.s_addr !=
4200 		    satosin(sa2)->sin_addr.s_addr) {
4201 			return 1;
4202 		}
4203 		if (port && satosin(sa1)->sin_port != satosin(sa2)->sin_port)
4204 			return 1;
4205 		break;
4206 #endif
4207 #ifdef INET6
4208 	case AF_INET6:
4209 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4210 			return 1;	/*EINVAL*/
4211 		if (satosin6(sa1)->sin6_scope_id !=
4212 		    satosin6(sa2)->sin6_scope_id) {
4213 			return 1;
4214 		}
4215 		if (!IN6_ARE_ADDR_EQUAL(&satosin6(sa1)->sin6_addr,
4216 		    &satosin6(sa2)->sin6_addr)) {
4217 			return 1;
4218 		}
4219 		if (port &&
4220 		    satosin6(sa1)->sin6_port != satosin6(sa2)->sin6_port) {
4221 			return 1;
4222 		}
4223 		break;
4224 #endif
4225 	default:
4226 		if (bcmp(sa1, sa2, sa1->sa_len) != 0)
4227 			return 1;
4228 		break;
4229 	}
4230 
4231 	return 0;
4232 }
4233 
4234 /* returns 0 on match */
4235 int
4236 key_sockaddrcmp_withmask(const struct sockaddr *sa1,
4237     const struct sockaddr *sa2, size_t mask)
4238 {
4239 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4240 		return (1);
4241 
4242 	switch (sa1->sa_family) {
4243 #ifdef INET
4244 	case AF_INET:
4245 		return (!key_bbcmp(&satosin(sa1)->sin_addr,
4246 		    &satosin(sa2)->sin_addr, mask));
4247 #endif
4248 #ifdef INET6
4249 	case AF_INET6:
4250 		if (satosin6(sa1)->sin6_scope_id !=
4251 		    satosin6(sa2)->sin6_scope_id)
4252 			return (1);
4253 		return (!key_bbcmp(&satosin6(sa1)->sin6_addr,
4254 		    &satosin6(sa2)->sin6_addr, mask));
4255 #endif
4256 	}
4257 	return (1);
4258 }
4259 #undef satosin
4260 #undef satosin6
4261 
4262 /*
4263  * compare two buffers with mask.
4264  * IN:
4265  *	addr1: source
4266  *	addr2: object
4267  *	bits:  Number of bits to compare
4268  * OUT:
4269  *	1 : equal
4270  *	0 : not equal
4271  */
4272 static int
4273 key_bbcmp(const void *a1, const void *a2, u_int bits)
4274 {
4275 	const unsigned char *p1 = a1;
4276 	const unsigned char *p2 = a2;
4277 
4278 	/* XXX: This could be considerably faster if we compare a word
4279 	 * at a time, but it is complicated on LSB Endian machines */
4280 
4281 	/* Handle null pointers */
4282 	if (p1 == NULL || p2 == NULL)
4283 		return (p1 == p2);
4284 
4285 	while (bits >= 8) {
4286 		if (*p1++ != *p2++)
4287 			return 0;
4288 		bits -= 8;
4289 	}
4290 
4291 	if (bits > 0) {
4292 		u_int8_t mask = ~((1<<(8-bits))-1);
4293 		if ((*p1 & mask) != (*p2 & mask))
4294 			return 0;
4295 	}
4296 	return 1;	/* Match! */
4297 }
4298 
4299 static void
4300 key_flush_spd(time_t now)
4301 {
4302 	SPTREE_RLOCK_TRACKER;
4303 	struct secpolicy_list drainq;
4304 	struct secpolicy *sp, *nextsp;
4305 	u_int dir;
4306 
4307 	LIST_INIT(&drainq);
4308 	SPTREE_RLOCK();
4309 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4310 		TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
4311 			if (sp->lifetime == 0 && sp->validtime == 0)
4312 				continue;
4313 			if ((sp->lifetime &&
4314 			    now - sp->created > sp->lifetime) ||
4315 			    (sp->validtime &&
4316 			    now - sp->lastused > sp->validtime)) {
4317 				/* Hold extra reference to send SPDEXPIRE */
4318 				SP_ADDREF(sp);
4319 				LIST_INSERT_HEAD(&drainq, sp, drainq);
4320 			}
4321 		}
4322 	}
4323 	SPTREE_RUNLOCK();
4324 	if (LIST_EMPTY(&drainq))
4325 		return;
4326 
4327 	SPTREE_WLOCK();
4328 	sp = LIST_FIRST(&drainq);
4329 	while (sp != NULL) {
4330 		nextsp = LIST_NEXT(sp, drainq);
4331 		/* Check that SP is still linked */
4332 		if (sp->state != IPSEC_SPSTATE_ALIVE) {
4333 			LIST_REMOVE(sp, drainq);
4334 			key_freesp(&sp); /* release extra reference */
4335 			sp = nextsp;
4336 			continue;
4337 		}
4338 		TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
4339 		LIST_REMOVE(sp, idhash);
4340 		sp->state = IPSEC_SPSTATE_DEAD;
4341 		sp = nextsp;
4342 	}
4343 	V_sp_genid++;
4344 	SPTREE_WUNLOCK();
4345 
4346 	sp = LIST_FIRST(&drainq);
4347 	while (sp != NULL) {
4348 		nextsp = LIST_NEXT(sp, drainq);
4349 		key_spdexpire(sp);
4350 		key_freesp(&sp); /* release extra reference */
4351 		key_freesp(&sp); /* release last reference */
4352 		sp = nextsp;
4353 	}
4354 }
4355 
4356 static void
4357 key_flush_sad(time_t now)
4358 {
4359 	SAHTREE_RLOCK_TRACKER;
4360 	struct secashead_list emptyq;
4361 	struct secasvar_list drainq, hexpireq, sexpireq, freeq;
4362 	struct secashead *sah, *nextsah;
4363 	struct secasvar *sav, *nextsav;
4364 
4365 	LIST_INIT(&drainq);
4366 	LIST_INIT(&hexpireq);
4367 	LIST_INIT(&sexpireq);
4368 	LIST_INIT(&emptyq);
4369 
4370 	SAHTREE_RLOCK();
4371 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
4372 		/* Check for empty SAH */
4373 		if (TAILQ_EMPTY(&sah->savtree_larval) &&
4374 		    TAILQ_EMPTY(&sah->savtree_alive)) {
4375 			SAH_ADDREF(sah);
4376 			LIST_INSERT_HEAD(&emptyq, sah, drainq);
4377 			continue;
4378 		}
4379 		/* Add all stale LARVAL SAs into drainq */
4380 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
4381 			if (now - sav->created < V_key_larval_lifetime)
4382 				continue;
4383 			SAV_ADDREF(sav);
4384 			LIST_INSERT_HEAD(&drainq, sav, drainq);
4385 		}
4386 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
4387 			/* lifetimes aren't specified */
4388 			if (sav->lft_h == NULL)
4389 				continue;
4390 			SECASVAR_LOCK(sav);
4391 			/*
4392 			 * Check again with lock held, because it may
4393 			 * be updated by SADB_UPDATE.
4394 			 */
4395 			if (sav->lft_h == NULL) {
4396 				SECASVAR_UNLOCK(sav);
4397 				continue;
4398 			}
4399 			/*
4400 			 * RFC 2367:
4401 			 * HARD lifetimes MUST take precedence over SOFT
4402 			 * lifetimes, meaning if the HARD and SOFT lifetimes
4403 			 * are the same, the HARD lifetime will appear on the
4404 			 * EXPIRE message.
4405 			 */
4406 			/* check HARD lifetime */
4407 			if ((sav->lft_h->addtime != 0 &&
4408 			    now - sav->created > sav->lft_h->addtime) ||
4409 			    (sav->lft_h->usetime != 0 && sav->firstused &&
4410 			    now - sav->firstused > sav->lft_h->usetime) ||
4411 			    (sav->lft_h->bytes != 0 && counter_u64_fetch(
4412 			        sav->lft_c_bytes) > sav->lft_h->bytes)) {
4413 				SECASVAR_UNLOCK(sav);
4414 				SAV_ADDREF(sav);
4415 				LIST_INSERT_HEAD(&hexpireq, sav, drainq);
4416 				continue;
4417 			}
4418 			/* check SOFT lifetime (only for MATURE SAs) */
4419 			if (sav->state == SADB_SASTATE_MATURE && (
4420 			    (sav->lft_s->addtime != 0 &&
4421 			    now - sav->created > sav->lft_s->addtime) ||
4422 			    (sav->lft_s->usetime != 0 && sav->firstused &&
4423 			    now - sav->firstused > sav->lft_s->usetime) ||
4424 			    (sav->lft_s->bytes != 0 && counter_u64_fetch(
4425 				sav->lft_c_bytes) > sav->lft_s->bytes))) {
4426 				SECASVAR_UNLOCK(sav);
4427 				SAV_ADDREF(sav);
4428 				LIST_INSERT_HEAD(&sexpireq, sav, drainq);
4429 				continue;
4430 			}
4431 			SECASVAR_UNLOCK(sav);
4432 		}
4433 	}
4434 	SAHTREE_RUNLOCK();
4435 
4436 	if (LIST_EMPTY(&emptyq) && LIST_EMPTY(&drainq) &&
4437 	    LIST_EMPTY(&hexpireq) && LIST_EMPTY(&sexpireq))
4438 		return;
4439 
4440 	LIST_INIT(&freeq);
4441 	SAHTREE_WLOCK();
4442 	/* Unlink stale LARVAL SAs */
4443 	sav = LIST_FIRST(&drainq);
4444 	while (sav != NULL) {
4445 		nextsav = LIST_NEXT(sav, drainq);
4446 		/* Check that SA is still LARVAL */
4447 		if (sav->state != SADB_SASTATE_LARVAL) {
4448 			LIST_REMOVE(sav, drainq);
4449 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4450 			sav = nextsav;
4451 			continue;
4452 		}
4453 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
4454 		LIST_REMOVE(sav, spihash);
4455 		sav->state = SADB_SASTATE_DEAD;
4456 		sav = nextsav;
4457 	}
4458 	/* Unlink all SAs with expired HARD lifetime */
4459 	sav = LIST_FIRST(&hexpireq);
4460 	while (sav != NULL) {
4461 		nextsav = LIST_NEXT(sav, drainq);
4462 		/* Check that SA is not unlinked */
4463 		if (sav->state == SADB_SASTATE_DEAD) {
4464 			LIST_REMOVE(sav, drainq);
4465 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4466 			sav = nextsav;
4467 			continue;
4468 		}
4469 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
4470 		LIST_REMOVE(sav, spihash);
4471 		sav->state = SADB_SASTATE_DEAD;
4472 		sav = nextsav;
4473 	}
4474 	/* Mark all SAs with expired SOFT lifetime as DYING */
4475 	sav = LIST_FIRST(&sexpireq);
4476 	while (sav != NULL) {
4477 		nextsav = LIST_NEXT(sav, drainq);
4478 		/* Check that SA is not unlinked */
4479 		if (sav->state == SADB_SASTATE_DEAD) {
4480 			LIST_REMOVE(sav, drainq);
4481 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4482 			sav = nextsav;
4483 			continue;
4484 		}
4485 		/*
4486 		 * NOTE: this doesn't change SA order in the chain.
4487 		 */
4488 		sav->state = SADB_SASTATE_DYING;
4489 		sav = nextsav;
4490 	}
4491 	/* Unlink empty SAHs */
4492 	sah = LIST_FIRST(&emptyq);
4493 	while (sah != NULL) {
4494 		nextsah = LIST_NEXT(sah, drainq);
4495 		/* Check that SAH is still empty and not unlinked */
4496 		if (sah->state == SADB_SASTATE_DEAD ||
4497 		    !TAILQ_EMPTY(&sah->savtree_larval) ||
4498 		    !TAILQ_EMPTY(&sah->savtree_alive)) {
4499 			LIST_REMOVE(sah, drainq);
4500 			key_freesah(&sah); /* release extra reference */
4501 			sah = nextsah;
4502 			continue;
4503 		}
4504 		TAILQ_REMOVE(&V_sahtree, sah, chain);
4505 		LIST_REMOVE(sah, addrhash);
4506 		sah->state = SADB_SASTATE_DEAD;
4507 		sah = nextsah;
4508 	}
4509 	SAHTREE_WUNLOCK();
4510 
4511 	/* Send SPDEXPIRE messages */
4512 	sav = LIST_FIRST(&hexpireq);
4513 	while (sav != NULL) {
4514 		nextsav = LIST_NEXT(sav, drainq);
4515 		key_expire(sav, 1);
4516 		key_freesah(&sav->sah); /* release reference from SAV */
4517 		key_freesav(&sav); /* release extra reference */
4518 		key_freesav(&sav); /* release last reference */
4519 		sav = nextsav;
4520 	}
4521 	sav = LIST_FIRST(&sexpireq);
4522 	while (sav != NULL) {
4523 		nextsav = LIST_NEXT(sav, drainq);
4524 		key_expire(sav, 0);
4525 		key_freesav(&sav); /* release extra reference */
4526 		sav = nextsav;
4527 	}
4528 	/* Free stale LARVAL SAs */
4529 	sav = LIST_FIRST(&drainq);
4530 	while (sav != NULL) {
4531 		nextsav = LIST_NEXT(sav, drainq);
4532 		key_freesah(&sav->sah); /* release reference from SAV */
4533 		key_freesav(&sav); /* release extra reference */
4534 		key_freesav(&sav); /* release last reference */
4535 		sav = nextsav;
4536 	}
4537 	/* Free SAs that were unlinked/changed by someone else */
4538 	sav = LIST_FIRST(&freeq);
4539 	while (sav != NULL) {
4540 		nextsav = LIST_NEXT(sav, drainq);
4541 		key_freesav(&sav); /* release extra reference */
4542 		sav = nextsav;
4543 	}
4544 	/* Free empty SAH */
4545 	sah = LIST_FIRST(&emptyq);
4546 	while (sah != NULL) {
4547 		nextsah = LIST_NEXT(sah, drainq);
4548 		key_freesah(&sah); /* release extra reference */
4549 		key_freesah(&sah); /* release last reference */
4550 		sah = nextsah;
4551 	}
4552 }
4553 
4554 static void
4555 key_flush_acq(time_t now)
4556 {
4557 	struct secacq *acq, *nextacq;
4558 
4559 	/* ACQ tree */
4560 	ACQ_LOCK();
4561 	acq = LIST_FIRST(&V_acqtree);
4562 	while (acq != NULL) {
4563 		nextacq = LIST_NEXT(acq, chain);
4564 		if (now - acq->created > V_key_blockacq_lifetime) {
4565 			LIST_REMOVE(acq, chain);
4566 			LIST_REMOVE(acq, addrhash);
4567 			LIST_REMOVE(acq, seqhash);
4568 			free(acq, M_IPSEC_SAQ);
4569 		}
4570 		acq = nextacq;
4571 	}
4572 	ACQ_UNLOCK();
4573 }
4574 
4575 static void
4576 key_flush_spacq(time_t now)
4577 {
4578 	struct secspacq *acq, *nextacq;
4579 
4580 	/* SP ACQ tree */
4581 	SPACQ_LOCK();
4582 	for (acq = LIST_FIRST(&V_spacqtree); acq != NULL; acq = nextacq) {
4583 		nextacq = LIST_NEXT(acq, chain);
4584 		if (now - acq->created > V_key_blockacq_lifetime
4585 		 && __LIST_CHAINED(acq)) {
4586 			LIST_REMOVE(acq, chain);
4587 			free(acq, M_IPSEC_SAQ);
4588 		}
4589 	}
4590 	SPACQ_UNLOCK();
4591 }
4592 
4593 /*
4594  * time handler.
4595  * scanning SPD and SAD to check status for each entries,
4596  * and do to remove or to expire.
4597  * XXX: year 2038 problem may remain.
4598  */
4599 static void
4600 key_timehandler(void *arg)
4601 {
4602 	VNET_ITERATOR_DECL(vnet_iter);
4603 	time_t now = time_second;
4604 
4605 	VNET_LIST_RLOCK_NOSLEEP();
4606 	VNET_FOREACH(vnet_iter) {
4607 		CURVNET_SET(vnet_iter);
4608 		key_flush_spd(now);
4609 		key_flush_sad(now);
4610 		key_flush_acq(now);
4611 		key_flush_spacq(now);
4612 		CURVNET_RESTORE();
4613 	}
4614 	VNET_LIST_RUNLOCK_NOSLEEP();
4615 
4616 #ifndef IPSEC_DEBUG2
4617 	/* do exchange to tick time !! */
4618 	callout_schedule(&key_timer, hz);
4619 #endif /* IPSEC_DEBUG2 */
4620 }
4621 
4622 u_long
4623 key_random()
4624 {
4625 	u_long value;
4626 
4627 	key_randomfill(&value, sizeof(value));
4628 	return value;
4629 }
4630 
4631 void
4632 key_randomfill(void *p, size_t l)
4633 {
4634 	size_t n;
4635 	u_long v;
4636 	static int warn = 1;
4637 
4638 	n = 0;
4639 	n = (size_t)read_random(p, (u_int)l);
4640 	/* last resort */
4641 	while (n < l) {
4642 		v = random();
4643 		bcopy(&v, (u_int8_t *)p + n,
4644 		    l - n < sizeof(v) ? l - n : sizeof(v));
4645 		n += sizeof(v);
4646 
4647 		if (warn) {
4648 			printf("WARNING: pseudo-random number generator "
4649 			    "used for IPsec processing\n");
4650 			warn = 0;
4651 		}
4652 	}
4653 }
4654 
4655 /*
4656  * map SADB_SATYPE_* to IPPROTO_*.
4657  * if satype == SADB_SATYPE then satype is mapped to ~0.
4658  * OUT:
4659  *	0: invalid satype.
4660  */
4661 static uint8_t
4662 key_satype2proto(uint8_t satype)
4663 {
4664 	switch (satype) {
4665 	case SADB_SATYPE_UNSPEC:
4666 		return IPSEC_PROTO_ANY;
4667 	case SADB_SATYPE_AH:
4668 		return IPPROTO_AH;
4669 	case SADB_SATYPE_ESP:
4670 		return IPPROTO_ESP;
4671 	case SADB_X_SATYPE_IPCOMP:
4672 		return IPPROTO_IPCOMP;
4673 	case SADB_X_SATYPE_TCPSIGNATURE:
4674 		return IPPROTO_TCP;
4675 	default:
4676 		return 0;
4677 	}
4678 	/* NOTREACHED */
4679 }
4680 
4681 /*
4682  * map IPPROTO_* to SADB_SATYPE_*
4683  * OUT:
4684  *	0: invalid protocol type.
4685  */
4686 static uint8_t
4687 key_proto2satype(uint8_t proto)
4688 {
4689 	switch (proto) {
4690 	case IPPROTO_AH:
4691 		return SADB_SATYPE_AH;
4692 	case IPPROTO_ESP:
4693 		return SADB_SATYPE_ESP;
4694 	case IPPROTO_IPCOMP:
4695 		return SADB_X_SATYPE_IPCOMP;
4696 	case IPPROTO_TCP:
4697 		return SADB_X_SATYPE_TCPSIGNATURE;
4698 	default:
4699 		return 0;
4700 	}
4701 	/* NOTREACHED */
4702 }
4703 
4704 /* %%% PF_KEY */
4705 /*
4706  * SADB_GETSPI processing is to receive
4707  *	<base, (SA2), src address, dst address, (SPI range)>
4708  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
4709  * tree with the status of LARVAL, and send
4710  *	<base, SA(*), address(SD)>
4711  * to the IKMPd.
4712  *
4713  * IN:	mhp: pointer to the pointer to each header.
4714  * OUT:	NULL if fail.
4715  *	other if success, return pointer to the message to send.
4716  */
4717 static int
4718 key_getspi(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
4719 {
4720 	struct secasindex saidx;
4721 	struct sadb_address *src0, *dst0;
4722 	struct secasvar *sav;
4723 	uint32_t reqid, spi;
4724 	int error;
4725 	uint8_t mode, proto;
4726 
4727 	IPSEC_ASSERT(so != NULL, ("null socket"));
4728 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
4729 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
4730 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
4731 
4732 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
4733 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)
4734 #ifdef PFKEY_STRICT_CHECKS
4735 	    || SADB_CHECKHDR(mhp, SADB_EXT_SPIRANGE)
4736 #endif
4737 	    ) {
4738 		ipseclog((LOG_DEBUG,
4739 		    "%s: invalid message: missing required header.\n",
4740 		    __func__));
4741 		error = EINVAL;
4742 		goto fail;
4743 	}
4744 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
4745 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)
4746 #ifdef PFKEY_STRICT_CHECKS
4747 	    || SADB_CHECKLEN(mhp, SADB_EXT_SPIRANGE)
4748 #endif
4749 	    ) {
4750 		ipseclog((LOG_DEBUG,
4751 		    "%s: invalid message: wrong header size.\n", __func__));
4752 		error = EINVAL;
4753 		goto fail;
4754 	}
4755 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
4756 		mode = IPSEC_MODE_ANY;
4757 		reqid = 0;
4758 	} else {
4759 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
4760 			ipseclog((LOG_DEBUG,
4761 			    "%s: invalid message: wrong header size.\n",
4762 			    __func__));
4763 			error = EINVAL;
4764 			goto fail;
4765 		}
4766 		mode = ((struct sadb_x_sa2 *)
4767 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
4768 		reqid = ((struct sadb_x_sa2 *)
4769 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
4770 	}
4771 
4772 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
4773 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
4774 
4775 	/* map satype to proto */
4776 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
4777 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
4778 			__func__));
4779 		error = EINVAL;
4780 		goto fail;
4781 	}
4782 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
4783 	    (struct sockaddr *)(dst0 + 1));
4784 	if (error != 0) {
4785 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
4786 		error = EINVAL;
4787 		goto fail;
4788 	}
4789 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
4790 
4791 	/* SPI allocation */
4792 	spi = key_do_getnewspi(
4793 	    (struct sadb_spirange *)mhp->ext[SADB_EXT_SPIRANGE], &saidx);
4794 	if (spi == 0) {
4795 		/*
4796 		 * Requested SPI or SPI range is not available or
4797 		 * already used.
4798 		 */
4799 		error = EEXIST;
4800 		goto fail;
4801 	}
4802 	sav = key_newsav(mhp, &saidx, spi, &error);
4803 	if (sav == NULL)
4804 		goto fail;
4805 
4806 	if (sav->seq != 0) {
4807 		/*
4808 		 * RFC2367:
4809 		 * If the SADB_GETSPI message is in response to a
4810 		 * kernel-generated SADB_ACQUIRE, the sadb_msg_seq
4811 		 * MUST be the same as the SADB_ACQUIRE message.
4812 		 *
4813 		 * XXXAE: However it doesn't definethe behaviour how to
4814 		 * check this and what to do if it doesn't match.
4815 		 * Also what we should do if it matches?
4816 		 *
4817 		 * We can compare saidx used in SADB_ACQUIRE with saidx
4818 		 * used in SADB_GETSPI, but this probably can break
4819 		 * existing software. For now just warn if it doesn't match.
4820 		 *
4821 		 * XXXAE: anyway it looks useless.
4822 		 */
4823 		key_acqdone(&saidx, sav->seq);
4824 	}
4825 	KEYDBG(KEY_STAMP,
4826 	    printf("%s: SA(%p)\n", __func__, sav));
4827 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
4828 
4829     {
4830 	struct mbuf *n, *nn;
4831 	struct sadb_sa *m_sa;
4832 	struct sadb_msg *newmsg;
4833 	int off, len;
4834 
4835 	/* create new sadb_msg to reply. */
4836 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
4837 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
4838 
4839 	MGETHDR(n, M_NOWAIT, MT_DATA);
4840 	if (len > MHLEN) {
4841 		if (!(MCLGET(n, M_NOWAIT))) {
4842 			m_freem(n);
4843 			n = NULL;
4844 		}
4845 	}
4846 	if (!n) {
4847 		error = ENOBUFS;
4848 		goto fail;
4849 	}
4850 
4851 	n->m_len = len;
4852 	n->m_next = NULL;
4853 	off = 0;
4854 
4855 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
4856 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
4857 
4858 	m_sa = (struct sadb_sa *)(mtod(n, caddr_t) + off);
4859 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
4860 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
4861 	m_sa->sadb_sa_spi = spi; /* SPI is already in network byte order */
4862 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
4863 
4864 	IPSEC_ASSERT(off == len,
4865 		("length inconsistency (off %u len %u)", off, len));
4866 
4867 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
4868 	    SADB_EXT_ADDRESS_DST);
4869 	if (!n->m_next) {
4870 		m_freem(n);
4871 		error = ENOBUFS;
4872 		goto fail;
4873 	}
4874 
4875 	if (n->m_len < sizeof(struct sadb_msg)) {
4876 		n = m_pullup(n, sizeof(struct sadb_msg));
4877 		if (n == NULL)
4878 			return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
4879 	}
4880 
4881 	n->m_pkthdr.len = 0;
4882 	for (nn = n; nn; nn = nn->m_next)
4883 		n->m_pkthdr.len += nn->m_len;
4884 
4885 	newmsg = mtod(n, struct sadb_msg *);
4886 	newmsg->sadb_msg_seq = sav->seq;
4887 	newmsg->sadb_msg_errno = 0;
4888 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
4889 
4890 	m_freem(m);
4891 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
4892     }
4893 
4894 fail:
4895 	return (key_senderror(so, m, error));
4896 }
4897 
4898 /*
4899  * allocating new SPI
4900  * called by key_getspi().
4901  * OUT:
4902  *	0:	failure.
4903  *	others: success, SPI in network byte order.
4904  */
4905 static uint32_t
4906 key_do_getnewspi(struct sadb_spirange *spirange, struct secasindex *saidx)
4907 {
4908 	uint32_t min, max, newspi, t;
4909 	int count = V_key_spi_trycnt;
4910 
4911 	/* set spi range to allocate */
4912 	if (spirange != NULL) {
4913 		min = spirange->sadb_spirange_min;
4914 		max = spirange->sadb_spirange_max;
4915 	} else {
4916 		min = V_key_spi_minval;
4917 		max = V_key_spi_maxval;
4918 	}
4919 	/* IPCOMP needs 2-byte SPI */
4920 	if (saidx->proto == IPPROTO_IPCOMP) {
4921 		if (min >= 0x10000)
4922 			min = 0xffff;
4923 		if (max >= 0x10000)
4924 			max = 0xffff;
4925 		if (min > max) {
4926 			t = min; min = max; max = t;
4927 		}
4928 	}
4929 
4930 	if (min == max) {
4931 		if (!key_checkspidup(htonl(min))) {
4932 			ipseclog((LOG_DEBUG, "%s: SPI %u exists already.\n",
4933 			    __func__, min));
4934 			return 0;
4935 		}
4936 
4937 		count--; /* taking one cost. */
4938 		newspi = min;
4939 	} else {
4940 
4941 		/* init SPI */
4942 		newspi = 0;
4943 
4944 		/* when requesting to allocate spi ranged */
4945 		while (count--) {
4946 			/* generate pseudo-random SPI value ranged. */
4947 			newspi = min + (key_random() % (max - min + 1));
4948 			if (!key_checkspidup(htonl(newspi)))
4949 				break;
4950 		}
4951 
4952 		if (count == 0 || newspi == 0) {
4953 			ipseclog((LOG_DEBUG,
4954 			    "%s: failed to allocate SPI.\n", __func__));
4955 			return 0;
4956 		}
4957 	}
4958 
4959 	/* statistics */
4960 	keystat.getspi_count =
4961 	    (keystat.getspi_count + V_key_spi_trycnt - count) / 2;
4962 
4963 	return (htonl(newspi));
4964 }
4965 
4966 /*
4967  * Find TCP-MD5 SA with corresponding secasindex.
4968  * If not found, return NULL and fill SPI with usable value if needed.
4969  */
4970 static struct secasvar *
4971 key_getsav_tcpmd5(struct secasindex *saidx, uint32_t *spi)
4972 {
4973 	SAHTREE_RLOCK_TRACKER;
4974 	struct secashead *sah;
4975 	struct secasvar *sav;
4976 
4977 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP, ("wrong proto"));
4978 	SAHTREE_RLOCK();
4979 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
4980 		if (sah->saidx.proto != IPPROTO_TCP)
4981 			continue;
4982 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0) &&
4983 		    !key_sockaddrcmp(&saidx->src.sa, &sah->saidx.src.sa, 0))
4984 			break;
4985 	}
4986 	if (sah != NULL) {
4987 		if (V_key_preferred_oldsa)
4988 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
4989 		else
4990 			sav = TAILQ_FIRST(&sah->savtree_alive);
4991 		if (sav != NULL) {
4992 			SAV_ADDREF(sav);
4993 			SAHTREE_RUNLOCK();
4994 			return (sav);
4995 		}
4996 	}
4997 	if (spi == NULL) {
4998 		/* No SPI required */
4999 		SAHTREE_RUNLOCK();
5000 		return (NULL);
5001 	}
5002 	/* Check that SPI is unique */
5003 	LIST_FOREACH(sav, SAVHASH_HASH(*spi), spihash) {
5004 		if (sav->spi == *spi)
5005 			break;
5006 	}
5007 	if (sav == NULL) {
5008 		SAHTREE_RUNLOCK();
5009 		/* SPI is already unique */
5010 		return (NULL);
5011 	}
5012 	SAHTREE_RUNLOCK();
5013 	/* XXX: not optimal */
5014 	*spi = key_do_getnewspi(NULL, saidx);
5015 	return (NULL);
5016 }
5017 
5018 static int
5019 key_updateaddresses(struct socket *so, struct mbuf *m,
5020     const struct sadb_msghdr *mhp, struct secasvar *sav,
5021     struct secasindex *saidx)
5022 {
5023 	struct sockaddr *newaddr;
5024 	struct secashead *sah;
5025 	struct secasvar *newsav, *tmp;
5026 	struct mbuf *n;
5027 	int error, isnew;
5028 
5029 	/* Check that we need to change SAH */
5030 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC)) {
5031 		newaddr = (struct sockaddr *)(
5032 		    ((struct sadb_address *)
5033 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_SRC]) + 1);
5034 		bcopy(newaddr, &saidx->src, newaddr->sa_len);
5035 		key_porttosaddr(&saidx->src.sa, 0);
5036 	}
5037 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5038 		newaddr = (struct sockaddr *)(
5039 		    ((struct sadb_address *)
5040 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_DST]) + 1);
5041 		bcopy(newaddr, &saidx->dst, newaddr->sa_len);
5042 		key_porttosaddr(&saidx->dst.sa, 0);
5043 	}
5044 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5045 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5046 		error = key_checksockaddrs(&saidx->src.sa, &saidx->dst.sa);
5047 		if (error != 0) {
5048 			ipseclog((LOG_DEBUG, "%s: invalid new sockaddr.\n",
5049 			    __func__));
5050 			return (error);
5051 		}
5052 
5053 		sah = key_getsah(saidx);
5054 		if (sah == NULL) {
5055 			/* create a new SA index */
5056 			sah = key_newsah(saidx);
5057 			if (sah == NULL) {
5058 				ipseclog((LOG_DEBUG,
5059 				    "%s: No more memory.\n", __func__));
5060 				return (ENOBUFS);
5061 			}
5062 			isnew = 2; /* SAH is new */
5063 		} else
5064 			isnew = 1; /* existing SAH is referenced */
5065 	} else {
5066 		/*
5067 		 * src and dst addresses are still the same.
5068 		 * Do we want to change NAT-T config?
5069 		 */
5070 		if (sav->sah->saidx.proto != IPPROTO_ESP ||
5071 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5072 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5073 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5074 			ipseclog((LOG_DEBUG,
5075 			    "%s: invalid message: missing required header.\n",
5076 			    __func__));
5077 			return (EINVAL);
5078 		}
5079 		/* We hold reference to SA, thus SAH will be referenced too. */
5080 		sah = sav->sah;
5081 		isnew = 0;
5082 	}
5083 
5084 	newsav = malloc(sizeof(struct secasvar), M_IPSEC_SA,
5085 	    M_NOWAIT | M_ZERO);
5086 	if (newsav == NULL) {
5087 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5088 		error = ENOBUFS;
5089 		goto fail;
5090 	}
5091 
5092 	/* Clone SA's content into newsav */
5093 	SAV_INITREF(newsav);
5094 	bcopy(sav, newsav, offsetof(struct secasvar, chain));
5095 	/*
5096 	 * We create new NAT-T config if it is needed.
5097 	 * Old NAT-T config will be freed by key_cleansav() when
5098 	 * last reference to SA will be released.
5099 	 */
5100 	newsav->natt = NULL;
5101 	newsav->sah = sah;
5102 	newsav->state = SADB_SASTATE_MATURE;
5103 	error = key_setnatt(sav, mhp);
5104 	if (error != 0)
5105 		goto fail;
5106 
5107 	SAHTREE_WLOCK();
5108 	/* Check that SA is still alive */
5109 	if (sav->state == SADB_SASTATE_DEAD) {
5110 		/* SA was unlinked */
5111 		SAHTREE_WUNLOCK();
5112 		error = ESRCH;
5113 		goto fail;
5114 	}
5115 
5116 	/* Unlink SA from SAH and SPI hash */
5117 	IPSEC_ASSERT((sav->flags & SADB_X_EXT_F_CLONED) == 0,
5118 	    ("SA is already cloned"));
5119 	IPSEC_ASSERT(sav->state == SADB_SASTATE_MATURE ||
5120 	    sav->state == SADB_SASTATE_DYING,
5121 	    ("Wrong SA state %u\n", sav->state));
5122 	TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
5123 	LIST_REMOVE(sav, spihash);
5124 	sav->state = SADB_SASTATE_DEAD;
5125 
5126 	/*
5127 	 * Link new SA with SAH. Keep SAs ordered by
5128 	 * create time (newer are first).
5129 	 */
5130 	TAILQ_FOREACH(tmp, &sah->savtree_alive, chain) {
5131 		if (newsav->created > tmp->created) {
5132 			TAILQ_INSERT_BEFORE(tmp, newsav, chain);
5133 			break;
5134 		}
5135 	}
5136 	if (tmp == NULL)
5137 		TAILQ_INSERT_TAIL(&sah->savtree_alive, newsav, chain);
5138 
5139 	/* Add new SA into SPI hash. */
5140 	LIST_INSERT_HEAD(SAVHASH_HASH(newsav->spi), newsav, spihash);
5141 
5142 	/* Add new SAH into SADB. */
5143 	if (isnew == 2) {
5144 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
5145 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
5146 		sah->state = SADB_SASTATE_MATURE;
5147 		SAH_ADDREF(sah); /* newsav references new SAH */
5148 	}
5149 	/*
5150 	 * isnew == 1 -> @sah was referenced by key_getsah().
5151 	 * isnew == 0 -> we use the same @sah, that was used by @sav,
5152 	 *	and we use its reference for @newsav.
5153 	 */
5154 	SECASVAR_LOCK(sav);
5155 	/* XXX: replace cntr with pointer? */
5156 	newsav->cntr = sav->cntr;
5157 	sav->flags |= SADB_X_EXT_F_CLONED;
5158 	SECASVAR_UNLOCK(sav);
5159 
5160 	SAHTREE_WUNLOCK();
5161 
5162 	KEYDBG(KEY_STAMP,
5163 	    printf("%s: SA(%p) cloned into SA(%p)\n",
5164 	    __func__, sav, newsav));
5165 	KEYDBG(KEY_DATA, kdebug_secasv(newsav));
5166 
5167 	key_freesav(&sav); /* release last reference */
5168 
5169 	/* set msg buf from mhp */
5170 	n = key_getmsgbuf_x1(m, mhp);
5171 	if (n == NULL) {
5172 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5173 		return (ENOBUFS);
5174 	}
5175 	m_freem(m);
5176 	key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5177 	return (0);
5178 fail:
5179 	if (isnew != 0)
5180 		key_freesah(&sah);
5181 	if (newsav != NULL) {
5182 		if (newsav->natt != NULL)
5183 			free(newsav->natt, M_IPSEC_MISC);
5184 		free(newsav, M_IPSEC_SA);
5185 	}
5186 	return (error);
5187 }
5188 
5189 /*
5190  * SADB_UPDATE processing
5191  * receive
5192  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5193  *       key(AE), (identity(SD),) (sensitivity)>
5194  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5195  * and send
5196  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5197  *       (identity(SD),) (sensitivity)>
5198  * to the ikmpd.
5199  *
5200  * m will always be freed.
5201  */
5202 static int
5203 key_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5204 {
5205 	struct secasindex saidx;
5206 	struct sadb_address *src0, *dst0;
5207 	struct sadb_sa *sa0;
5208 	struct secasvar *sav;
5209 	uint32_t reqid;
5210 	int error;
5211 	uint8_t mode, proto;
5212 
5213 	IPSEC_ASSERT(so != NULL, ("null socket"));
5214 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5215 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5216 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5217 
5218 	/* map satype to proto */
5219 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5220 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5221 		    __func__));
5222 		return key_senderror(so, m, EINVAL);
5223 	}
5224 
5225 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5226 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5227 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5228 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5229 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5230 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5231 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5232 		ipseclog((LOG_DEBUG,
5233 		    "%s: invalid message: missing required header.\n",
5234 		    __func__));
5235 		return key_senderror(so, m, EINVAL);
5236 	}
5237 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5238 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5239 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5240 		ipseclog((LOG_DEBUG,
5241 		    "%s: invalid message: wrong header size.\n", __func__));
5242 		return key_senderror(so, m, EINVAL);
5243 	}
5244 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5245 		mode = IPSEC_MODE_ANY;
5246 		reqid = 0;
5247 	} else {
5248 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5249 			ipseclog((LOG_DEBUG,
5250 			    "%s: invalid message: wrong header size.\n",
5251 			    __func__));
5252 			return key_senderror(so, m, EINVAL);
5253 		}
5254 		mode = ((struct sadb_x_sa2 *)
5255 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5256 		reqid = ((struct sadb_x_sa2 *)
5257 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5258 	}
5259 
5260 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5261 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5262 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5263 
5264 	/*
5265 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5266 	 * SADB_UPDATE message.
5267 	 */
5268 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5269 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5270 #ifdef PFKEY_STRICT_CHECKS
5271 		return key_senderror(so, m, EINVAL);
5272 #endif
5273 	}
5274 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5275 	    (struct sockaddr *)(dst0 + 1));
5276 	if (error != 0) {
5277 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5278 		return key_senderror(so, m, error);
5279 	}
5280 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5281 	sav = key_getsavbyspi(sa0->sadb_sa_spi);
5282 	if (sav == NULL) {
5283 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u\n",
5284 		    __func__, ntohl(sa0->sadb_sa_spi)));
5285 		return key_senderror(so, m, EINVAL);
5286 	}
5287 	/*
5288 	 * Check that SADB_UPDATE issued by the same process that did
5289 	 * SADB_GETSPI or SADB_ADD.
5290 	 */
5291 	if (sav->pid != mhp->msg->sadb_msg_pid) {
5292 		ipseclog((LOG_DEBUG,
5293 		    "%s: pid mismatched (SPI %u, pid %u vs. %u)\n", __func__,
5294 		    ntohl(sav->spi), sav->pid, mhp->msg->sadb_msg_pid));
5295 		key_freesav(&sav);
5296 		return key_senderror(so, m, EINVAL);
5297 	}
5298 	/* saidx should match with SA. */
5299 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_MODE_REQID) == 0) {
5300 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u",
5301 		    __func__, ntohl(sav->spi)));
5302 		key_freesav(&sav);
5303 		return key_senderror(so, m, ESRCH);
5304 	}
5305 
5306 	if (sav->state == SADB_SASTATE_LARVAL) {
5307 		if ((mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5308 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) ||
5309 		    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5310 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH))) {
5311 			ipseclog((LOG_DEBUG,
5312 			    "%s: invalid message: missing required header.\n",
5313 			    __func__));
5314 			key_freesav(&sav);
5315 			return key_senderror(so, m, EINVAL);
5316 		}
5317 		/*
5318 		 * We can set any values except src, dst and SPI.
5319 		 */
5320 		error = key_setsaval(sav, mhp);
5321 		if (error != 0) {
5322 			key_freesav(&sav);
5323 			return (key_senderror(so, m, error));
5324 		}
5325 		/* Change SA state to MATURE */
5326 		SAHTREE_WLOCK();
5327 		if (sav->state != SADB_SASTATE_LARVAL) {
5328 			/* SA was deleted or another thread made it MATURE. */
5329 			SAHTREE_WUNLOCK();
5330 			key_freesav(&sav);
5331 			return (key_senderror(so, m, ESRCH));
5332 		}
5333 		/*
5334 		 * NOTE: we keep SAs in savtree_alive ordered by created
5335 		 * time. When SA's state changed from LARVAL to MATURE,
5336 		 * we update its created time in key_setsaval() and move
5337 		 * it into head of savtree_alive.
5338 		 */
5339 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
5340 		TAILQ_INSERT_HEAD(&sav->sah->savtree_alive, sav, chain);
5341 		sav->state = SADB_SASTATE_MATURE;
5342 		SAHTREE_WUNLOCK();
5343 	} else {
5344 		/*
5345 		 * For DYING and MATURE SA we can change only state
5346 		 * and lifetimes. Report EINVAL if something else attempted
5347 		 * to change.
5348 		 */
5349 		if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5350 		    !SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
5351 			key_freesav(&sav);
5352 			return (key_senderror(so, m, EINVAL));
5353 		}
5354 		error = key_updatelifetimes(sav, mhp);
5355 		if (error != 0) {
5356 			key_freesav(&sav);
5357 			return (key_senderror(so, m, error));
5358 		}
5359 		/*
5360 		 * This is FreeBSD extension to RFC2367.
5361 		 * IKEd can specify SADB_X_EXT_NEW_ADDRESS_SRC and/or
5362 		 * SADB_X_EXT_NEW_ADDRESS_DST when it wants to change
5363 		 * SA addresses (for example to implement MOBIKE protocol
5364 		 * as described in RFC4555). Also we allow to change
5365 		 * NAT-T config.
5366 		 */
5367 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5368 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST) ||
5369 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5370 		    sav->natt != NULL) {
5371 			error = key_updateaddresses(so, m, mhp, sav, &saidx);
5372 			key_freesav(&sav);
5373 			if (error != 0)
5374 				return (key_senderror(so, m, error));
5375 			return (0);
5376 		}
5377 		/* Check that SA is still alive */
5378 		SAHTREE_WLOCK();
5379 		if (sav->state == SADB_SASTATE_DEAD) {
5380 			/* SA was unlinked */
5381 			SAHTREE_WUNLOCK();
5382 			key_freesav(&sav);
5383 			return (key_senderror(so, m, ESRCH));
5384 		}
5385 		/*
5386 		 * NOTE: there is possible state moving from DYING to MATURE,
5387 		 * but this doesn't change created time, so we won't reorder
5388 		 * this SA.
5389 		 */
5390 		sav->state = SADB_SASTATE_MATURE;
5391 		SAHTREE_WUNLOCK();
5392 	}
5393 	KEYDBG(KEY_STAMP,
5394 	    printf("%s: SA(%p)\n", __func__, sav));
5395 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5396 	key_freesav(&sav);
5397 
5398     {
5399 	struct mbuf *n;
5400 
5401 	/* set msg buf from mhp */
5402 	n = key_getmsgbuf_x1(m, mhp);
5403 	if (n == NULL) {
5404 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5405 		return key_senderror(so, m, ENOBUFS);
5406 	}
5407 
5408 	m_freem(m);
5409 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5410     }
5411 }
5412 
5413 /*
5414  * SADB_ADD processing
5415  * add an entry to SA database, when received
5416  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5417  *       key(AE), (identity(SD),) (sensitivity)>
5418  * from the ikmpd,
5419  * and send
5420  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5421  *       (identity(SD),) (sensitivity)>
5422  * to the ikmpd.
5423  *
5424  * IGNORE identity and sensitivity messages.
5425  *
5426  * m will always be freed.
5427  */
5428 static int
5429 key_add(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5430 {
5431 	struct secasindex saidx;
5432 	struct sadb_address *src0, *dst0;
5433 	struct sadb_sa *sa0;
5434 	struct secasvar *sav;
5435 	uint32_t reqid, spi;
5436 	uint8_t mode, proto;
5437 	int error;
5438 
5439 	IPSEC_ASSERT(so != NULL, ("null socket"));
5440 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5441 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5442 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5443 
5444 	/* map satype to proto */
5445 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5446 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5447 		    __func__));
5448 		return key_senderror(so, m, EINVAL);
5449 	}
5450 
5451 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5452 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5453 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5454 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP && (
5455 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5456 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT))) ||
5457 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH && (
5458 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH) ||
5459 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH))) ||
5460 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5461 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5462 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5463 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5464 		ipseclog((LOG_DEBUG,
5465 		    "%s: invalid message: missing required header.\n",
5466 		    __func__));
5467 		return key_senderror(so, m, EINVAL);
5468 	}
5469 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5470 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5471 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5472 		ipseclog((LOG_DEBUG,
5473 		    "%s: invalid message: wrong header size.\n", __func__));
5474 		return key_senderror(so, m, EINVAL);
5475 	}
5476 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5477 		mode = IPSEC_MODE_ANY;
5478 		reqid = 0;
5479 	} else {
5480 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5481 			ipseclog((LOG_DEBUG,
5482 			    "%s: invalid message: wrong header size.\n",
5483 			    __func__));
5484 			return key_senderror(so, m, EINVAL);
5485 		}
5486 		mode = ((struct sadb_x_sa2 *)
5487 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5488 		reqid = ((struct sadb_x_sa2 *)
5489 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5490 	}
5491 
5492 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5493 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
5494 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
5495 
5496 	/*
5497 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5498 	 * SADB_ADD message.
5499 	 */
5500 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5501 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5502 #ifdef PFKEY_STRICT_CHECKS
5503 		return key_senderror(so, m, EINVAL);
5504 #endif
5505 	}
5506 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5507 	    (struct sockaddr *)(dst0 + 1));
5508 	if (error != 0) {
5509 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5510 		return key_senderror(so, m, error);
5511 	}
5512 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5513 	spi = sa0->sadb_sa_spi;
5514 	/*
5515 	 * For TCP-MD5 SAs we don't use SPI. Check the uniqueness using
5516 	 * secasindex.
5517 	 * XXXAE: IPComp seems also doesn't use SPI.
5518 	 */
5519 	if (proto == IPPROTO_TCP) {
5520 		sav = key_getsav_tcpmd5(&saidx, &spi);
5521 		if (sav == NULL && spi == 0) {
5522 			/* Failed to allocate SPI */
5523 			ipseclog((LOG_DEBUG, "%s: SA already exists.\n",
5524 			    __func__));
5525 			return key_senderror(so, m, EEXIST);
5526 		}
5527 		/* XXX: SPI that we report back can have another value */
5528 	} else {
5529 		/* We can create new SA only if SPI is different. */
5530 		sav = key_getsavbyspi(spi);
5531 	}
5532 	if (sav != NULL) {
5533 		key_freesav(&sav);
5534 		ipseclog((LOG_DEBUG, "%s: SA already exists.\n", __func__));
5535 		return key_senderror(so, m, EEXIST);
5536 	}
5537 
5538 	sav = key_newsav(mhp, &saidx, spi, &error);
5539 	if (sav == NULL)
5540 		return key_senderror(so, m, error);
5541 	KEYDBG(KEY_STAMP,
5542 	    printf("%s: return SA(%p)\n", __func__, sav));
5543 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5544 	/*
5545 	 * If SADB_ADD was in response to SADB_ACQUIRE, we need to schedule
5546 	 * ACQ for deletion.
5547 	 */
5548 	if (sav->seq != 0)
5549 		key_acqdone(&saidx, sav->seq);
5550 
5551     {
5552 	/*
5553 	 * Don't call key_freesav() on error here, as we would like to
5554 	 * keep the SA in the database.
5555 	 */
5556 	struct mbuf *n;
5557 
5558 	/* set msg buf from mhp */
5559 	n = key_getmsgbuf_x1(m, mhp);
5560 	if (n == NULL) {
5561 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5562 		return key_senderror(so, m, ENOBUFS);
5563 	}
5564 
5565 	m_freem(m);
5566 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5567     }
5568 }
5569 
5570 /*
5571  * NAT-T support.
5572  * IKEd may request the use ESP in UDP encapsulation when it detects the
5573  * presence of NAT. It uses NAT-T extension headers for such SAs to specify
5574  * parameters needed for encapsulation and decapsulation. These PF_KEY
5575  * extension headers are not standardized, so this comment addresses our
5576  * implementation.
5577  * SADB_X_EXT_NAT_T_TYPE specifies type of encapsulation, we support only
5578  * UDP_ENCAP_ESPINUDP as described in RFC3948.
5579  * SADB_X_EXT_NAT_T_SPORT/DPORT specifies source and destination ports for
5580  * UDP header. We use these ports in UDP encapsulation procedure, also we
5581  * can check them in UDP decapsulation procedure.
5582  * SADB_X_EXT_NAT_T_OA[IR] specifies original address of initiator or
5583  * responder. These addresses can be used for transport mode to adjust
5584  * checksum after decapsulation and decryption. Since original IP addresses
5585  * used by peer usually different (we detected presence of NAT), TCP/UDP
5586  * pseudo header checksum and IP header checksum was calculated using original
5587  * addresses. After decapsulation and decryption we need to adjust checksum
5588  * to have correct datagram.
5589  *
5590  * We expect presence of NAT-T extension headers only in SADB_ADD and
5591  * SADB_UPDATE messages. We report NAT-T extension headers in replies
5592  * to SADB_ADD, SADB_UPDATE, SADB_GET, and SADB_DUMP messages.
5593  */
5594 static int
5595 key_setnatt(struct secasvar *sav, const struct sadb_msghdr *mhp)
5596 {
5597 	struct sadb_x_nat_t_port *port;
5598 	struct sadb_x_nat_t_type *type;
5599 	struct sadb_address *oai, *oar;
5600 	struct sockaddr *sa;
5601 	uint32_t addr;
5602 	uint16_t cksum;
5603 
5604 	IPSEC_ASSERT(sav->natt == NULL, ("natt is already initialized"));
5605 	/*
5606 	 * Ignore NAT-T headers if sproto isn't ESP.
5607 	 */
5608 	if (sav->sah->saidx.proto != IPPROTO_ESP)
5609 		return (0);
5610 
5611 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) &&
5612 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) &&
5613 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5614 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5615 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5616 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5617 			ipseclog((LOG_DEBUG,
5618 			    "%s: invalid message: wrong header size.\n",
5619 			    __func__));
5620 			return (EINVAL);
5621 		}
5622 	} else
5623 		return (0);
5624 
5625 	type = (struct sadb_x_nat_t_type *)mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5626 	if (type->sadb_x_nat_t_type_type != UDP_ENCAP_ESPINUDP) {
5627 		ipseclog((LOG_DEBUG, "%s: unsupported NAT-T type %u.\n",
5628 		    __func__, type->sadb_x_nat_t_type_type));
5629 		return (EINVAL);
5630 	}
5631 	/*
5632 	 * Allocate storage for NAT-T config.
5633 	 * On error it will be released by key_cleansav().
5634 	 */
5635 	sav->natt = malloc(sizeof(struct secnatt), M_IPSEC_MISC,
5636 	    M_NOWAIT | M_ZERO);
5637 	if (sav->natt == NULL) {
5638 		PFKEYSTAT_INC(in_nomem);
5639 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5640 		return (ENOBUFS);
5641 	}
5642 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5643 	if (port->sadb_x_nat_t_port_port == 0) {
5644 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T sport specified.\n",
5645 		    __func__));
5646 		return (EINVAL);
5647 	}
5648 	sav->natt->sport = port->sadb_x_nat_t_port_port;
5649 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5650 	if (port->sadb_x_nat_t_port_port == 0) {
5651 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T dport specified.\n",
5652 		    __func__));
5653 		return (EINVAL);
5654 	}
5655 	sav->natt->dport = port->sadb_x_nat_t_port_port;
5656 
5657 	/*
5658 	 * SADB_X_EXT_NAT_T_OAI and SADB_X_EXT_NAT_T_OAR are optional
5659 	 * and needed only for transport mode IPsec.
5660 	 * Usually NAT translates only one address, but it is possible,
5661 	 * that both addresses could be translated.
5662 	 * NOTE: Value of SADB_X_EXT_NAT_T_OAI is equal to SADB_X_EXT_NAT_T_OA.
5663 	 */
5664 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAI)) {
5665 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAI)) {
5666 			ipseclog((LOG_DEBUG,
5667 			    "%s: invalid message: wrong header size.\n",
5668 			    __func__));
5669 			return (EINVAL);
5670 		}
5671 		oai = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAI];
5672 	} else
5673 		oai = NULL;
5674 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAR)) {
5675 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAR)) {
5676 			ipseclog((LOG_DEBUG,
5677 			    "%s: invalid message: wrong header size.\n",
5678 			    __func__));
5679 			return (EINVAL);
5680 		}
5681 		oar = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAR];
5682 	} else
5683 		oar = NULL;
5684 
5685 	/* Initialize addresses only for transport mode */
5686 	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL) {
5687 		cksum = 0;
5688 		if (oai != NULL) {
5689 			/* Currently we support only AF_INET */
5690 			sa = (struct sockaddr *)(oai + 1);
5691 			if (sa->sa_family != AF_INET ||
5692 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5693 				ipseclog((LOG_DEBUG,
5694 				    "%s: wrong NAT-OAi header.\n",
5695 				    __func__));
5696 				return (EINVAL);
5697 			}
5698 			/* Ignore address if it the same */
5699 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5700 			    sav->sah->saidx.src.sin.sin_addr.s_addr) {
5701 				bcopy(sa, &sav->natt->oai.sa, sa->sa_len);
5702 				sav->natt->flags |= IPSEC_NATT_F_OAI;
5703 				/* Calculate checksum delta */
5704 				addr = sav->sah->saidx.src.sin.sin_addr.s_addr;
5705 				cksum = in_addword(cksum, ~addr >> 16);
5706 				cksum = in_addword(cksum, ~addr & 0xffff);
5707 				addr = sav->natt->oai.sin.sin_addr.s_addr;
5708 				cksum = in_addword(cksum, addr >> 16);
5709 				cksum = in_addword(cksum, addr & 0xffff);
5710 			}
5711 		}
5712 		if (oar != NULL) {
5713 			/* Currently we support only AF_INET */
5714 			sa = (struct sockaddr *)(oar + 1);
5715 			if (sa->sa_family != AF_INET ||
5716 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5717 				ipseclog((LOG_DEBUG,
5718 				    "%s: wrong NAT-OAr header.\n",
5719 				    __func__));
5720 				return (EINVAL);
5721 			}
5722 			/* Ignore address if it the same */
5723 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5724 			    sav->sah->saidx.dst.sin.sin_addr.s_addr) {
5725 				bcopy(sa, &sav->natt->oar.sa, sa->sa_len);
5726 				sav->natt->flags |= IPSEC_NATT_F_OAR;
5727 				/* Calculate checksum delta */
5728 				addr = sav->sah->saidx.dst.sin.sin_addr.s_addr;
5729 				cksum = in_addword(cksum, ~addr >> 16);
5730 				cksum = in_addword(cksum, ~addr & 0xffff);
5731 				addr = sav->natt->oar.sin.sin_addr.s_addr;
5732 				cksum = in_addword(cksum, addr >> 16);
5733 				cksum = in_addword(cksum, addr & 0xffff);
5734 			}
5735 		}
5736 		sav->natt->cksum = cksum;
5737 	}
5738 	return (0);
5739 }
5740 
5741 static int
5742 key_setident(struct secashead *sah, const struct sadb_msghdr *mhp)
5743 {
5744 	const struct sadb_ident *idsrc, *iddst;
5745 	int idsrclen, iddstlen;
5746 
5747 	IPSEC_ASSERT(sah != NULL, ("null secashead"));
5748 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5749 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5750 
5751 	/* don't make buffer if not there */
5752 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) &&
5753 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5754 		sah->idents = NULL;
5755 		sah->identd = NULL;
5756 		return (0);
5757 	}
5758 
5759 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) ||
5760 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5761 		ipseclog((LOG_DEBUG, "%s: invalid identity.\n", __func__));
5762 		return (EINVAL);
5763 	}
5764 
5765 	idsrc = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_SRC];
5766 	iddst = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_DST];
5767 	idsrclen = mhp->extlen[SADB_EXT_IDENTITY_SRC];
5768 	iddstlen = mhp->extlen[SADB_EXT_IDENTITY_DST];
5769 
5770 	/* validity check */
5771 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
5772 		ipseclog((LOG_DEBUG, "%s: ident type mismatch.\n", __func__));
5773 		return EINVAL;
5774 	}
5775 
5776 	switch (idsrc->sadb_ident_type) {
5777 	case SADB_IDENTTYPE_PREFIX:
5778 	case SADB_IDENTTYPE_FQDN:
5779 	case SADB_IDENTTYPE_USERFQDN:
5780 	default:
5781 		/* XXX do nothing */
5782 		sah->idents = NULL;
5783 		sah->identd = NULL;
5784 	 	return 0;
5785 	}
5786 
5787 	/* make structure */
5788 	sah->idents = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5789 	if (sah->idents == NULL) {
5790 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5791 		return ENOBUFS;
5792 	}
5793 	sah->identd = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5794 	if (sah->identd == NULL) {
5795 		free(sah->idents, M_IPSEC_MISC);
5796 		sah->idents = NULL;
5797 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5798 		return ENOBUFS;
5799 	}
5800 	sah->idents->type = idsrc->sadb_ident_type;
5801 	sah->idents->id = idsrc->sadb_ident_id;
5802 
5803 	sah->identd->type = iddst->sadb_ident_type;
5804 	sah->identd->id = iddst->sadb_ident_id;
5805 
5806 	return 0;
5807 }
5808 
5809 /*
5810  * m will not be freed on return.
5811  * it is caller's responsibility to free the result.
5812  *
5813  * Called from SADB_ADD and SADB_UPDATE. Reply will contain headers
5814  * from the request in defined order.
5815  */
5816 static struct mbuf *
5817 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
5818 {
5819 	struct mbuf *n;
5820 
5821 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5822 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5823 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5824 
5825 	/* create new sadb_msg to reply. */
5826 	n = key_gather_mbuf(m, mhp, 1, 16, SADB_EXT_RESERVED,
5827 	    SADB_EXT_SA, SADB_X_EXT_SA2,
5828 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
5829 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
5830 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
5831 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
5832 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
5833 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NEW_ADDRESS_SRC,
5834 	    SADB_X_EXT_NEW_ADDRESS_DST);
5835 	if (!n)
5836 		return NULL;
5837 
5838 	if (n->m_len < sizeof(struct sadb_msg)) {
5839 		n = m_pullup(n, sizeof(struct sadb_msg));
5840 		if (n == NULL)
5841 			return NULL;
5842 	}
5843 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
5844 	mtod(n, struct sadb_msg *)->sadb_msg_len =
5845 	    PFKEY_UNIT64(n->m_pkthdr.len);
5846 
5847 	return n;
5848 }
5849 
5850 /*
5851  * SADB_DELETE processing
5852  * receive
5853  *   <base, SA(*), address(SD)>
5854  * from the ikmpd, and set SADB_SASTATE_DEAD,
5855  * and send,
5856  *   <base, SA(*), address(SD)>
5857  * to the ikmpd.
5858  *
5859  * m will always be freed.
5860  */
5861 static int
5862 key_delete(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5863 {
5864 	struct secasindex saidx;
5865 	struct sadb_address *src0, *dst0;
5866 	struct secasvar *sav;
5867 	struct sadb_sa *sa0;
5868 	uint8_t proto;
5869 
5870 	IPSEC_ASSERT(so != NULL, ("null socket"));
5871 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5872 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5873 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5874 
5875 	/* map satype to proto */
5876 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5877 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5878 		    __func__));
5879 		return key_senderror(so, m, EINVAL);
5880 	}
5881 
5882 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5883 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5884 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5885 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5886 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
5887 		    __func__));
5888 		return key_senderror(so, m, EINVAL);
5889 	}
5890 
5891 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5892 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5893 
5894 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
5895 	    (struct sockaddr *)(dst0 + 1)) != 0) {
5896 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5897 		return (key_senderror(so, m, EINVAL));
5898 	}
5899 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
5900 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
5901 		/*
5902 		 * Caller wants us to delete all non-LARVAL SAs
5903 		 * that match the src/dst.  This is used during
5904 		 * IKE INITIAL-CONTACT.
5905 		 * XXXAE: this looks like some extension to RFC2367.
5906 		 */
5907 		ipseclog((LOG_DEBUG, "%s: doing delete all.\n", __func__));
5908 		return (key_delete_all(so, m, mhp, &saidx));
5909 	}
5910 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
5911 		ipseclog((LOG_DEBUG,
5912 		    "%s: invalid message: wrong header size.\n", __func__));
5913 		return (key_senderror(so, m, EINVAL));
5914 	}
5915 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5916 	if (proto == IPPROTO_TCP)
5917 		sav = key_getsav_tcpmd5(&saidx, NULL);
5918 	else
5919 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
5920 	if (sav == NULL) {
5921 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u.\n",
5922 		    __func__, ntohl(sa0->sadb_sa_spi)));
5923 		return (key_senderror(so, m, ESRCH));
5924 	}
5925 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
5926 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
5927 		    __func__, ntohl(sav->spi)));
5928 		key_freesav(&sav);
5929 		return (key_senderror(so, m, ESRCH));
5930 	}
5931 	KEYDBG(KEY_STAMP,
5932 	    printf("%s: SA(%p)\n", __func__, sav));
5933 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5934 	key_unlinksav(sav);
5935 	key_freesav(&sav);
5936 
5937     {
5938 	struct mbuf *n;
5939 	struct sadb_msg *newmsg;
5940 
5941 	/* create new sadb_msg to reply. */
5942 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
5943 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
5944 	if (!n)
5945 		return key_senderror(so, m, ENOBUFS);
5946 
5947 	if (n->m_len < sizeof(struct sadb_msg)) {
5948 		n = m_pullup(n, sizeof(struct sadb_msg));
5949 		if (n == NULL)
5950 			return key_senderror(so, m, ENOBUFS);
5951 	}
5952 	newmsg = mtod(n, struct sadb_msg *);
5953 	newmsg->sadb_msg_errno = 0;
5954 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
5955 
5956 	m_freem(m);
5957 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5958     }
5959 }
5960 
5961 /*
5962  * delete all SAs for src/dst.  Called from key_delete().
5963  */
5964 static int
5965 key_delete_all(struct socket *so, struct mbuf *m,
5966     const struct sadb_msghdr *mhp, struct secasindex *saidx)
5967 {
5968 	struct secasvar_queue drainq;
5969 	struct secashead *sah;
5970 	struct secasvar *sav, *nextsav;
5971 
5972 	TAILQ_INIT(&drainq);
5973 	SAHTREE_WLOCK();
5974 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
5975 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_HEAD) == 0)
5976 			continue;
5977 		/* Move all ALIVE SAs into drainq */
5978 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
5979 	}
5980 	/* Unlink all queued SAs from SPI hash */
5981 	TAILQ_FOREACH(sav, &drainq, chain) {
5982 		sav->state = SADB_SASTATE_DEAD;
5983 		LIST_REMOVE(sav, spihash);
5984 	}
5985 	SAHTREE_WUNLOCK();
5986 	/* Now we can release reference for all SAs in drainq */
5987 	sav = TAILQ_FIRST(&drainq);
5988 	while (sav != NULL) {
5989 		KEYDBG(KEY_STAMP,
5990 		    printf("%s: SA(%p)\n", __func__, sav));
5991 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
5992 		nextsav = TAILQ_NEXT(sav, chain);
5993 		key_freesah(&sav->sah); /* release reference from SAV */
5994 		key_freesav(&sav); /* release last reference */
5995 		sav = nextsav;
5996 	}
5997 
5998     {
5999 	struct mbuf *n;
6000 	struct sadb_msg *newmsg;
6001 
6002 	/* create new sadb_msg to reply. */
6003 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
6004 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6005 	if (!n)
6006 		return key_senderror(so, m, ENOBUFS);
6007 
6008 	if (n->m_len < sizeof(struct sadb_msg)) {
6009 		n = m_pullup(n, sizeof(struct sadb_msg));
6010 		if (n == NULL)
6011 			return key_senderror(so, m, ENOBUFS);
6012 	}
6013 	newmsg = mtod(n, struct sadb_msg *);
6014 	newmsg->sadb_msg_errno = 0;
6015 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
6016 
6017 	m_freem(m);
6018 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6019     }
6020 }
6021 
6022 /*
6023  * Delete all alive SAs for corresponding xform.
6024  * Larval SAs have not initialized tdb_xform, so it is safe to leave them
6025  * here when xform disappears.
6026  */
6027 static void
6028 key_delete_xform(const struct xformsw *xsp)
6029 {
6030 	struct secasvar_queue drainq;
6031 	struct secashead *sah;
6032 	struct secasvar *sav, *nextsav;
6033 
6034 	TAILQ_INIT(&drainq);
6035 	SAHTREE_WLOCK();
6036 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
6037 		sav = TAILQ_FIRST(&sah->savtree_alive);
6038 		if (sav == NULL)
6039 			continue;
6040 		if (sav->tdb_xform != xsp)
6041 			continue;
6042 		/*
6043 		 * It is supposed that all SAs in the chain are related to
6044 		 * one xform.
6045 		 */
6046 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6047 	}
6048 	/* Unlink all queued SAs from SPI hash */
6049 	TAILQ_FOREACH(sav, &drainq, chain) {
6050 		sav->state = SADB_SASTATE_DEAD;
6051 		LIST_REMOVE(sav, spihash);
6052 	}
6053 	SAHTREE_WUNLOCK();
6054 
6055 	/* Now we can release reference for all SAs in drainq */
6056 	sav = TAILQ_FIRST(&drainq);
6057 	while (sav != NULL) {
6058 		KEYDBG(KEY_STAMP,
6059 		    printf("%s: SA(%p)\n", __func__, sav));
6060 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6061 		nextsav = TAILQ_NEXT(sav, chain);
6062 		key_freesah(&sav->sah); /* release reference from SAV */
6063 		key_freesav(&sav); /* release last reference */
6064 		sav = nextsav;
6065 	}
6066 }
6067 
6068 /*
6069  * SADB_GET processing
6070  * receive
6071  *   <base, SA(*), address(SD)>
6072  * from the ikmpd, and get a SP and a SA to respond,
6073  * and send,
6074  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6075  *       (identity(SD),) (sensitivity)>
6076  * to the ikmpd.
6077  *
6078  * m will always be freed.
6079  */
6080 static int
6081 key_get(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6082 {
6083 	struct secasindex saidx;
6084 	struct sadb_address *src0, *dst0;
6085 	struct sadb_sa *sa0;
6086 	struct secasvar *sav;
6087 	uint8_t proto;
6088 
6089 	IPSEC_ASSERT(so != NULL, ("null socket"));
6090 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6091 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6092 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6093 
6094 	/* map satype to proto */
6095 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6096 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6097 			__func__));
6098 		return key_senderror(so, m, EINVAL);
6099 	}
6100 
6101 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
6102 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6103 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)) {
6104 		ipseclog((LOG_DEBUG,
6105 		    "%s: invalid message: missing required header.\n",
6106 		    __func__));
6107 		return key_senderror(so, m, EINVAL);
6108 	}
6109 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
6110 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6111 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6112 		ipseclog((LOG_DEBUG,
6113 		    "%s: invalid message: wrong header size.\n", __func__));
6114 		return key_senderror(so, m, EINVAL);
6115 	}
6116 
6117 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6118 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
6119 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
6120 
6121 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6122 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6123 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6124 		return key_senderror(so, m, EINVAL);
6125 	}
6126 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6127 
6128 	if (proto == IPPROTO_TCP)
6129 		sav = key_getsav_tcpmd5(&saidx, NULL);
6130 	else
6131 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6132 	if (sav == NULL) {
6133 		ipseclog((LOG_DEBUG, "%s: no SA found.\n", __func__));
6134 		return key_senderror(so, m, ESRCH);
6135 	}
6136 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6137 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6138 		    __func__, ntohl(sa0->sadb_sa_spi)));
6139 		key_freesav(&sav);
6140 		return (key_senderror(so, m, ESRCH));
6141 	}
6142 
6143     {
6144 	struct mbuf *n;
6145 	uint8_t satype;
6146 
6147 	/* map proto to satype */
6148 	if ((satype = key_proto2satype(sav->sah->saidx.proto)) == 0) {
6149 		ipseclog((LOG_DEBUG, "%s: there was invalid proto in SAD.\n",
6150 		    __func__));
6151 		key_freesav(&sav);
6152 		return key_senderror(so, m, EINVAL);
6153 	}
6154 
6155 	/* create new sadb_msg to reply. */
6156 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
6157 	    mhp->msg->sadb_msg_pid);
6158 
6159 	key_freesav(&sav);
6160 	if (!n)
6161 		return key_senderror(so, m, ENOBUFS);
6162 
6163 	m_freem(m);
6164 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
6165     }
6166 }
6167 
6168 /* XXX make it sysctl-configurable? */
6169 static void
6170 key_getcomb_setlifetime(struct sadb_comb *comb)
6171 {
6172 
6173 	comb->sadb_comb_soft_allocations = 1;
6174 	comb->sadb_comb_hard_allocations = 1;
6175 	comb->sadb_comb_soft_bytes = 0;
6176 	comb->sadb_comb_hard_bytes = 0;
6177 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
6178 	comb->sadb_comb_soft_addtime = comb->sadb_comb_soft_addtime * 80 / 100;
6179 	comb->sadb_comb_soft_usetime = 28800;	/* 8 hours */
6180 	comb->sadb_comb_hard_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
6181 }
6182 
6183 /*
6184  * XXX reorder combinations by preference
6185  * XXX no idea if the user wants ESP authentication or not
6186  */
6187 static struct mbuf *
6188 key_getcomb_ealg(void)
6189 {
6190 	struct sadb_comb *comb;
6191 	const struct enc_xform *algo;
6192 	struct mbuf *result = NULL, *m, *n;
6193 	int encmin;
6194 	int i, off, o;
6195 	int totlen;
6196 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6197 
6198 	m = NULL;
6199 	for (i = 1; i <= SADB_EALG_MAX; i++) {
6200 		algo = enc_algorithm_lookup(i);
6201 		if (algo == NULL)
6202 			continue;
6203 
6204 		/* discard algorithms with key size smaller than system min */
6205 		if (_BITS(algo->maxkey) < V_ipsec_esp_keymin)
6206 			continue;
6207 		if (_BITS(algo->minkey) < V_ipsec_esp_keymin)
6208 			encmin = V_ipsec_esp_keymin;
6209 		else
6210 			encmin = _BITS(algo->minkey);
6211 
6212 		if (V_ipsec_esp_auth)
6213 			m = key_getcomb_ah();
6214 		else {
6215 			IPSEC_ASSERT(l <= MLEN,
6216 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6217 			MGET(m, M_NOWAIT, MT_DATA);
6218 			if (m) {
6219 				M_ALIGN(m, l);
6220 				m->m_len = l;
6221 				m->m_next = NULL;
6222 				bzero(mtod(m, caddr_t), m->m_len);
6223 			}
6224 		}
6225 		if (!m)
6226 			goto fail;
6227 
6228 		totlen = 0;
6229 		for (n = m; n; n = n->m_next)
6230 			totlen += n->m_len;
6231 		IPSEC_ASSERT((totlen % l) == 0, ("totlen=%u, l=%u", totlen, l));
6232 
6233 		for (off = 0; off < totlen; off += l) {
6234 			n = m_pulldown(m, off, l, &o);
6235 			if (!n) {
6236 				/* m is already freed */
6237 				goto fail;
6238 			}
6239 			comb = (struct sadb_comb *)(mtod(n, caddr_t) + o);
6240 			bzero(comb, sizeof(*comb));
6241 			key_getcomb_setlifetime(comb);
6242 			comb->sadb_comb_encrypt = i;
6243 			comb->sadb_comb_encrypt_minbits = encmin;
6244 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
6245 		}
6246 
6247 		if (!result)
6248 			result = m;
6249 		else
6250 			m_cat(result, m);
6251 	}
6252 
6253 	return result;
6254 
6255  fail:
6256 	if (result)
6257 		m_freem(result);
6258 	return NULL;
6259 }
6260 
6261 static void
6262 key_getsizes_ah(const struct auth_hash *ah, int alg, u_int16_t* min,
6263     u_int16_t* max)
6264 {
6265 
6266 	*min = *max = ah->keysize;
6267 	if (ah->keysize == 0) {
6268 		/*
6269 		 * Transform takes arbitrary key size but algorithm
6270 		 * key size is restricted.  Enforce this here.
6271 		 */
6272 		switch (alg) {
6273 		case SADB_X_AALG_MD5:	*min = *max = 16; break;
6274 		case SADB_X_AALG_SHA:	*min = *max = 20; break;
6275 		case SADB_X_AALG_NULL:	*min = 1; *max = 256; break;
6276 		case SADB_X_AALG_SHA2_256: *min = *max = 32; break;
6277 		case SADB_X_AALG_SHA2_384: *min = *max = 48; break;
6278 		case SADB_X_AALG_SHA2_512: *min = *max = 64; break;
6279 		default:
6280 			DPRINTF(("%s: unknown AH algorithm %u\n",
6281 				__func__, alg));
6282 			break;
6283 		}
6284 	}
6285 }
6286 
6287 /*
6288  * XXX reorder combinations by preference
6289  */
6290 static struct mbuf *
6291 key_getcomb_ah()
6292 {
6293 	const struct auth_hash *algo;
6294 	struct sadb_comb *comb;
6295 	struct mbuf *m;
6296 	u_int16_t minkeysize, maxkeysize;
6297 	int i;
6298 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6299 
6300 	m = NULL;
6301 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6302 #if 1
6303 		/* we prefer HMAC algorithms, not old algorithms */
6304 		if (i != SADB_AALG_SHA1HMAC &&
6305 		    i != SADB_AALG_MD5HMAC  &&
6306 		    i != SADB_X_AALG_SHA2_256 &&
6307 		    i != SADB_X_AALG_SHA2_384 &&
6308 		    i != SADB_X_AALG_SHA2_512)
6309 			continue;
6310 #endif
6311 		algo = auth_algorithm_lookup(i);
6312 		if (!algo)
6313 			continue;
6314 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6315 		/* discard algorithms with key size smaller than system min */
6316 		if (_BITS(minkeysize) < V_ipsec_ah_keymin)
6317 			continue;
6318 
6319 		if (!m) {
6320 			IPSEC_ASSERT(l <= MLEN,
6321 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6322 			MGET(m, M_NOWAIT, MT_DATA);
6323 			if (m) {
6324 				M_ALIGN(m, l);
6325 				m->m_len = l;
6326 				m->m_next = NULL;
6327 			}
6328 		} else
6329 			M_PREPEND(m, l, M_NOWAIT);
6330 		if (!m)
6331 			return NULL;
6332 
6333 		comb = mtod(m, struct sadb_comb *);
6334 		bzero(comb, sizeof(*comb));
6335 		key_getcomb_setlifetime(comb);
6336 		comb->sadb_comb_auth = i;
6337 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6338 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6339 	}
6340 
6341 	return m;
6342 }
6343 
6344 /*
6345  * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6346  * XXX reorder combinations by preference
6347  */
6348 static struct mbuf *
6349 key_getcomb_ipcomp()
6350 {
6351 	const struct comp_algo *algo;
6352 	struct sadb_comb *comb;
6353 	struct mbuf *m;
6354 	int i;
6355 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6356 
6357 	m = NULL;
6358 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6359 		algo = comp_algorithm_lookup(i);
6360 		if (!algo)
6361 			continue;
6362 
6363 		if (!m) {
6364 			IPSEC_ASSERT(l <= MLEN,
6365 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6366 			MGET(m, M_NOWAIT, MT_DATA);
6367 			if (m) {
6368 				M_ALIGN(m, l);
6369 				m->m_len = l;
6370 				m->m_next = NULL;
6371 			}
6372 		} else
6373 			M_PREPEND(m, l, M_NOWAIT);
6374 		if (!m)
6375 			return NULL;
6376 
6377 		comb = mtod(m, struct sadb_comb *);
6378 		bzero(comb, sizeof(*comb));
6379 		key_getcomb_setlifetime(comb);
6380 		comb->sadb_comb_encrypt = i;
6381 		/* what should we set into sadb_comb_*_{min,max}bits? */
6382 	}
6383 
6384 	return m;
6385 }
6386 
6387 /*
6388  * XXX no way to pass mode (transport/tunnel) to userland
6389  * XXX replay checking?
6390  * XXX sysctl interface to ipsec_{ah,esp}_keymin
6391  */
6392 static struct mbuf *
6393 key_getprop(const struct secasindex *saidx)
6394 {
6395 	struct sadb_prop *prop;
6396 	struct mbuf *m, *n;
6397 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6398 	int totlen;
6399 
6400 	switch (saidx->proto)  {
6401 	case IPPROTO_ESP:
6402 		m = key_getcomb_ealg();
6403 		break;
6404 	case IPPROTO_AH:
6405 		m = key_getcomb_ah();
6406 		break;
6407 	case IPPROTO_IPCOMP:
6408 		m = key_getcomb_ipcomp();
6409 		break;
6410 	default:
6411 		return NULL;
6412 	}
6413 
6414 	if (!m)
6415 		return NULL;
6416 	M_PREPEND(m, l, M_NOWAIT);
6417 	if (!m)
6418 		return NULL;
6419 
6420 	totlen = 0;
6421 	for (n = m; n; n = n->m_next)
6422 		totlen += n->m_len;
6423 
6424 	prop = mtod(m, struct sadb_prop *);
6425 	bzero(prop, sizeof(*prop));
6426 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6427 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6428 	prop->sadb_prop_replay = 32;	/* XXX */
6429 
6430 	return m;
6431 }
6432 
6433 /*
6434  * SADB_ACQUIRE processing called by key_checkrequest() and key_acquire2().
6435  * send
6436  *   <base, SA, address(SD), (address(P)), x_policy,
6437  *       (identity(SD),) (sensitivity,) proposal>
6438  * to KMD, and expect to receive
6439  *   <base> with SADB_ACQUIRE if error occurred,
6440  * or
6441  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6442  * from KMD by PF_KEY.
6443  *
6444  * XXX x_policy is outside of RFC2367 (KAME extension).
6445  * XXX sensitivity is not supported.
6446  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6447  * see comment for key_getcomb_ipcomp().
6448  *
6449  * OUT:
6450  *    0     : succeed
6451  *    others: error number
6452  */
6453 static int
6454 key_acquire(const struct secasindex *saidx, struct secpolicy *sp)
6455 {
6456 	union sockaddr_union addr;
6457 	struct mbuf *result, *m;
6458 	uint32_t seq;
6459 	int error;
6460 	uint16_t ul_proto;
6461 	uint8_t mask, satype;
6462 
6463 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
6464 	satype = key_proto2satype(saidx->proto);
6465 	IPSEC_ASSERT(satype != 0, ("null satype, protocol %u", saidx->proto));
6466 
6467 	error = -1;
6468 	result = NULL;
6469 	ul_proto = IPSEC_ULPROTO_ANY;
6470 
6471 	/* Get seq number to check whether sending message or not. */
6472 	seq = key_getacq(saidx, &error);
6473 	if (seq == 0)
6474 		return (error);
6475 
6476 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0);
6477 	if (!m) {
6478 		error = ENOBUFS;
6479 		goto fail;
6480 	}
6481 	result = m;
6482 
6483 	/*
6484 	 * set sadb_address for saidx's.
6485 	 *
6486 	 * Note that if sp is supplied, then we're being called from
6487 	 * key_allocsa_policy() and should supply port and protocol
6488 	 * information.
6489 	 * XXXAE: why only TCP and UDP? ICMP and SCTP looks applicable too.
6490 	 * XXXAE: probably we can handle this in the ipsec[46]_allocsa().
6491 	 * XXXAE: it looks like we should save this info in the ACQ entry.
6492 	 */
6493 	if (sp != NULL && (sp->spidx.ul_proto == IPPROTO_TCP ||
6494 	    sp->spidx.ul_proto == IPPROTO_UDP))
6495 		ul_proto = sp->spidx.ul_proto;
6496 
6497 	addr = saidx->src;
6498 	mask = FULLMASK;
6499 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6500 		switch (sp->spidx.src.sa.sa_family) {
6501 		case AF_INET:
6502 			if (sp->spidx.src.sin.sin_port != IPSEC_PORT_ANY) {
6503 				addr.sin.sin_port = sp->spidx.src.sin.sin_port;
6504 				mask = sp->spidx.prefs;
6505 			}
6506 			break;
6507 		case AF_INET6:
6508 			if (sp->spidx.src.sin6.sin6_port != IPSEC_PORT_ANY) {
6509 				addr.sin6.sin6_port =
6510 				    sp->spidx.src.sin6.sin6_port;
6511 				mask = sp->spidx.prefs;
6512 			}
6513 			break;
6514 		default:
6515 			break;
6516 		}
6517 	}
6518 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &addr.sa, mask, ul_proto);
6519 	if (!m) {
6520 		error = ENOBUFS;
6521 		goto fail;
6522 	}
6523 	m_cat(result, m);
6524 
6525 	addr = saidx->dst;
6526 	mask = FULLMASK;
6527 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6528 		switch (sp->spidx.dst.sa.sa_family) {
6529 		case AF_INET:
6530 			if (sp->spidx.dst.sin.sin_port != IPSEC_PORT_ANY) {
6531 				addr.sin.sin_port = sp->spidx.dst.sin.sin_port;
6532 				mask = sp->spidx.prefd;
6533 			}
6534 			break;
6535 		case AF_INET6:
6536 			if (sp->spidx.dst.sin6.sin6_port != IPSEC_PORT_ANY) {
6537 				addr.sin6.sin6_port =
6538 				    sp->spidx.dst.sin6.sin6_port;
6539 				mask = sp->spidx.prefd;
6540 			}
6541 			break;
6542 		default:
6543 			break;
6544 		}
6545 	}
6546 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &addr.sa, mask, ul_proto);
6547 	if (!m) {
6548 		error = ENOBUFS;
6549 		goto fail;
6550 	}
6551 	m_cat(result, m);
6552 
6553 	/* XXX proxy address (optional) */
6554 
6555 	/* set sadb_x_policy */
6556 	if (sp != NULL) {
6557 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6558 		    sp->priority);
6559 		if (!m) {
6560 			error = ENOBUFS;
6561 			goto fail;
6562 		}
6563 		m_cat(result, m);
6564 	}
6565 
6566 	/* XXX identity (optional) */
6567 #if 0
6568 	if (idexttype && fqdn) {
6569 		/* create identity extension (FQDN) */
6570 		struct sadb_ident *id;
6571 		int fqdnlen;
6572 
6573 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
6574 		id = (struct sadb_ident *)p;
6575 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6576 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6577 		id->sadb_ident_exttype = idexttype;
6578 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
6579 		bcopy(fqdn, id + 1, fqdnlen);
6580 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
6581 	}
6582 
6583 	if (idexttype) {
6584 		/* create identity extension (USERFQDN) */
6585 		struct sadb_ident *id;
6586 		int userfqdnlen;
6587 
6588 		if (userfqdn) {
6589 			/* +1 for terminating-NUL */
6590 			userfqdnlen = strlen(userfqdn) + 1;
6591 		} else
6592 			userfqdnlen = 0;
6593 		id = (struct sadb_ident *)p;
6594 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6595 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6596 		id->sadb_ident_exttype = idexttype;
6597 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
6598 		/* XXX is it correct? */
6599 		if (curproc && curproc->p_cred)
6600 			id->sadb_ident_id = curproc->p_cred->p_ruid;
6601 		if (userfqdn && userfqdnlen)
6602 			bcopy(userfqdn, id + 1, userfqdnlen);
6603 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
6604 	}
6605 #endif
6606 
6607 	/* XXX sensitivity (optional) */
6608 
6609 	/* create proposal/combination extension */
6610 	m = key_getprop(saidx);
6611 #if 0
6612 	/*
6613 	 * spec conformant: always attach proposal/combination extension,
6614 	 * the problem is that we have no way to attach it for ipcomp,
6615 	 * due to the way sadb_comb is declared in RFC2367.
6616 	 */
6617 	if (!m) {
6618 		error = ENOBUFS;
6619 		goto fail;
6620 	}
6621 	m_cat(result, m);
6622 #else
6623 	/*
6624 	 * outside of spec; make proposal/combination extension optional.
6625 	 */
6626 	if (m)
6627 		m_cat(result, m);
6628 #endif
6629 
6630 	if ((result->m_flags & M_PKTHDR) == 0) {
6631 		error = EINVAL;
6632 		goto fail;
6633 	}
6634 
6635 	if (result->m_len < sizeof(struct sadb_msg)) {
6636 		result = m_pullup(result, sizeof(struct sadb_msg));
6637 		if (result == NULL) {
6638 			error = ENOBUFS;
6639 			goto fail;
6640 		}
6641 	}
6642 
6643 	result->m_pkthdr.len = 0;
6644 	for (m = result; m; m = m->m_next)
6645 		result->m_pkthdr.len += m->m_len;
6646 
6647 	mtod(result, struct sadb_msg *)->sadb_msg_len =
6648 	    PFKEY_UNIT64(result->m_pkthdr.len);
6649 
6650 	KEYDBG(KEY_STAMP,
6651 	    printf("%s: SP(%p)\n", __func__, sp));
6652 	KEYDBG(KEY_DATA, kdebug_secasindex(saidx, NULL));
6653 
6654 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
6655 
6656  fail:
6657 	if (result)
6658 		m_freem(result);
6659 	return error;
6660 }
6661 
6662 static uint32_t
6663 key_newacq(const struct secasindex *saidx, int *perror)
6664 {
6665 	struct secacq *acq;
6666 	uint32_t seq;
6667 
6668 	acq = malloc(sizeof(*acq), M_IPSEC_SAQ, M_NOWAIT | M_ZERO);
6669 	if (acq == NULL) {
6670 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6671 		*perror = ENOBUFS;
6672 		return (0);
6673 	}
6674 
6675 	/* copy secindex */
6676 	bcopy(saidx, &acq->saidx, sizeof(acq->saidx));
6677 	acq->created = time_second;
6678 	acq->count = 0;
6679 
6680 	/* add to acqtree */
6681 	ACQ_LOCK();
6682 	seq = acq->seq = (V_acq_seq == ~0 ? 1 : ++V_acq_seq);
6683 	LIST_INSERT_HEAD(&V_acqtree, acq, chain);
6684 	LIST_INSERT_HEAD(ACQADDRHASH_HASH(saidx), acq, addrhash);
6685 	LIST_INSERT_HEAD(ACQSEQHASH_HASH(seq), acq, seqhash);
6686 	ACQ_UNLOCK();
6687 	*perror = 0;
6688 	return (seq);
6689 }
6690 
6691 static uint32_t
6692 key_getacq(const struct secasindex *saidx, int *perror)
6693 {
6694 	struct secacq *acq;
6695 	uint32_t seq;
6696 
6697 	ACQ_LOCK();
6698 	LIST_FOREACH(acq, ACQADDRHASH_HASH(saidx), addrhash) {
6699 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY)) {
6700 			if (acq->count > V_key_blockacq_count) {
6701 				/*
6702 				 * Reset counter and send message.
6703 				 * Also reset created time to keep ACQ for
6704 				 * this saidx.
6705 				 */
6706 				acq->created = time_second;
6707 				acq->count = 0;
6708 				seq = acq->seq;
6709 			} else {
6710 				/*
6711 				 * Increment counter and do nothing.
6712 				 * We send SADB_ACQUIRE message only
6713 				 * for each V_key_blockacq_count packet.
6714 				 */
6715 				acq->count++;
6716 				seq = 0;
6717 			}
6718 			break;
6719 		}
6720 	}
6721 	ACQ_UNLOCK();
6722 	if (acq != NULL) {
6723 		*perror = 0;
6724 		return (seq);
6725 	}
6726 	/* allocate new  entry */
6727 	return (key_newacq(saidx, perror));
6728 }
6729 
6730 static int
6731 key_acqreset(uint32_t seq)
6732 {
6733 	struct secacq *acq;
6734 
6735 	ACQ_LOCK();
6736 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6737 		if (acq->seq == seq) {
6738 			acq->count = 0;
6739 			acq->created = time_second;
6740 			break;
6741 		}
6742 	}
6743 	ACQ_UNLOCK();
6744 	if (acq == NULL)
6745 		return (ESRCH);
6746 	return (0);
6747 }
6748 /*
6749  * Mark ACQ entry as stale to remove it in key_flush_acq().
6750  * Called after successful SADB_GETSPI message.
6751  */
6752 static int
6753 key_acqdone(const struct secasindex *saidx, uint32_t seq)
6754 {
6755 	struct secacq *acq;
6756 
6757 	ACQ_LOCK();
6758 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6759 		if (acq->seq == seq)
6760 			break;
6761 	}
6762 	if (acq != NULL) {
6763 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY) == 0) {
6764 			ipseclog((LOG_DEBUG,
6765 			    "%s: Mismatched saidx for ACQ %u", __func__, seq));
6766 			acq = NULL;
6767 		} else {
6768 			acq->created = 0;
6769 		}
6770 	} else {
6771 		ipseclog((LOG_DEBUG,
6772 		    "%s: ACQ %u is not found.", __func__, seq));
6773 	}
6774 	ACQ_UNLOCK();
6775 	if (acq == NULL)
6776 		return (ESRCH);
6777 	return (0);
6778 }
6779 
6780 static struct secspacq *
6781 key_newspacq(struct secpolicyindex *spidx)
6782 {
6783 	struct secspacq *acq;
6784 
6785 	/* get new entry */
6786 	acq = malloc(sizeof(struct secspacq), M_IPSEC_SAQ, M_NOWAIT|M_ZERO);
6787 	if (acq == NULL) {
6788 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6789 		return NULL;
6790 	}
6791 
6792 	/* copy secindex */
6793 	bcopy(spidx, &acq->spidx, sizeof(acq->spidx));
6794 	acq->created = time_second;
6795 	acq->count = 0;
6796 
6797 	/* add to spacqtree */
6798 	SPACQ_LOCK();
6799 	LIST_INSERT_HEAD(&V_spacqtree, acq, chain);
6800 	SPACQ_UNLOCK();
6801 
6802 	return acq;
6803 }
6804 
6805 static struct secspacq *
6806 key_getspacq(struct secpolicyindex *spidx)
6807 {
6808 	struct secspacq *acq;
6809 
6810 	SPACQ_LOCK();
6811 	LIST_FOREACH(acq, &V_spacqtree, chain) {
6812 		if (key_cmpspidx_exactly(spidx, &acq->spidx)) {
6813 			/* NB: return holding spacq_lock */
6814 			return acq;
6815 		}
6816 	}
6817 	SPACQ_UNLOCK();
6818 
6819 	return NULL;
6820 }
6821 
6822 /*
6823  * SADB_ACQUIRE processing,
6824  * in first situation, is receiving
6825  *   <base>
6826  * from the ikmpd, and clear sequence of its secasvar entry.
6827  *
6828  * In second situation, is receiving
6829  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6830  * from a user land process, and return
6831  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6832  * to the socket.
6833  *
6834  * m will always be freed.
6835  */
6836 static int
6837 key_acquire2(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6838 {
6839 	SAHTREE_RLOCK_TRACKER;
6840 	struct sadb_address *src0, *dst0;
6841 	struct secasindex saidx;
6842 	struct secashead *sah;
6843 	uint32_t reqid;
6844 	int error;
6845 	uint8_t mode, proto;
6846 
6847 	IPSEC_ASSERT(so != NULL, ("null socket"));
6848 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6849 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6850 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6851 
6852 	/*
6853 	 * Error message from KMd.
6854 	 * We assume that if error was occurred in IKEd, the length of PFKEY
6855 	 * message is equal to the size of sadb_msg structure.
6856 	 * We do not raise error even if error occurred in this function.
6857 	 */
6858 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
6859 		/* check sequence number */
6860 		if (mhp->msg->sadb_msg_seq == 0 ||
6861 		    mhp->msg->sadb_msg_errno == 0) {
6862 			ipseclog((LOG_DEBUG, "%s: must specify sequence "
6863 				"number and errno.\n", __func__));
6864 		} else {
6865 			/*
6866 			 * IKEd reported that error occurred.
6867 			 * XXXAE: what it expects from the kernel?
6868 			 * Probably we should send SADB_ACQUIRE again?
6869 			 * If so, reset ACQ's state.
6870 			 * XXXAE: it looks useless.
6871 			 */
6872 			key_acqreset(mhp->msg->sadb_msg_seq);
6873 		}
6874 		m_freem(m);
6875 		return (0);
6876 	}
6877 
6878 	/*
6879 	 * This message is from user land.
6880 	 */
6881 
6882 	/* map satype to proto */
6883 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6884 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6885 		    __func__));
6886 		return key_senderror(so, m, EINVAL);
6887 	}
6888 
6889 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6890 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
6891 	    SADB_CHECKHDR(mhp, SADB_EXT_PROPOSAL)) {
6892 		ipseclog((LOG_DEBUG,
6893 		    "%s: invalid message: missing required header.\n",
6894 		    __func__));
6895 		return key_senderror(so, m, EINVAL);
6896 	}
6897 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6898 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
6899 	    SADB_CHECKLEN(mhp, SADB_EXT_PROPOSAL)) {
6900 		ipseclog((LOG_DEBUG,
6901 		    "%s: invalid message: wrong header size.\n", __func__));
6902 		return key_senderror(so, m, EINVAL);
6903 	}
6904 
6905 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
6906 		mode = IPSEC_MODE_ANY;
6907 		reqid = 0;
6908 	} else {
6909 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
6910 			ipseclog((LOG_DEBUG,
6911 			    "%s: invalid message: wrong header size.\n",
6912 			    __func__));
6913 			return key_senderror(so, m, EINVAL);
6914 		}
6915 		mode = ((struct sadb_x_sa2 *)
6916 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
6917 		reqid = ((struct sadb_x_sa2 *)
6918 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
6919 	}
6920 
6921 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
6922 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
6923 
6924 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
6925 	    (struct sockaddr *)(dst0 + 1));
6926 	if (error != 0) {
6927 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6928 		return key_senderror(so, m, EINVAL);
6929 	}
6930 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
6931 
6932 	/* get a SA index */
6933 	SAHTREE_RLOCK();
6934 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
6935 		if (key_cmpsaidx(&sah->saidx, &saidx, CMP_MODE_REQID))
6936 			break;
6937 	}
6938 	SAHTREE_RUNLOCK();
6939 	if (sah != NULL) {
6940 		ipseclog((LOG_DEBUG, "%s: a SA exists already.\n", __func__));
6941 		return key_senderror(so, m, EEXIST);
6942 	}
6943 
6944 	error = key_acquire(&saidx, NULL);
6945 	if (error != 0) {
6946 		ipseclog((LOG_DEBUG,
6947 		    "%s: error %d returned from key_acquire()\n",
6948 			__func__, error));
6949 		return key_senderror(so, m, error);
6950 	}
6951 	m_freem(m);
6952 	return (0);
6953 }
6954 
6955 /*
6956  * SADB_REGISTER processing.
6957  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
6958  * receive
6959  *   <base>
6960  * from the ikmpd, and register a socket to send PF_KEY messages,
6961  * and send
6962  *   <base, supported>
6963  * to KMD by PF_KEY.
6964  * If socket is detached, must free from regnode.
6965  *
6966  * m will always be freed.
6967  */
6968 static int
6969 key_register(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6970 {
6971 	struct secreg *reg, *newreg = NULL;
6972 
6973 	IPSEC_ASSERT(so != NULL, ("null socket"));
6974 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6975 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6976 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6977 
6978 	/* check for invalid register message */
6979 	if (mhp->msg->sadb_msg_satype >= sizeof(V_regtree)/sizeof(V_regtree[0]))
6980 		return key_senderror(so, m, EINVAL);
6981 
6982 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
6983 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
6984 		goto setmsg;
6985 
6986 	/* check whether existing or not */
6987 	REGTREE_LOCK();
6988 	LIST_FOREACH(reg, &V_regtree[mhp->msg->sadb_msg_satype], chain) {
6989 		if (reg->so == so) {
6990 			REGTREE_UNLOCK();
6991 			ipseclog((LOG_DEBUG, "%s: socket exists already.\n",
6992 				__func__));
6993 			return key_senderror(so, m, EEXIST);
6994 		}
6995 	}
6996 
6997 	/* create regnode */
6998 	newreg =  malloc(sizeof(struct secreg), M_IPSEC_SAR, M_NOWAIT|M_ZERO);
6999 	if (newreg == NULL) {
7000 		REGTREE_UNLOCK();
7001 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7002 		return key_senderror(so, m, ENOBUFS);
7003 	}
7004 
7005 	newreg->so = so;
7006 	((struct keycb *)sotorawcb(so))->kp_registered++;
7007 
7008 	/* add regnode to regtree. */
7009 	LIST_INSERT_HEAD(&V_regtree[mhp->msg->sadb_msg_satype], newreg, chain);
7010 	REGTREE_UNLOCK();
7011 
7012   setmsg:
7013     {
7014 	struct mbuf *n;
7015 	struct sadb_msg *newmsg;
7016 	struct sadb_supported *sup;
7017 	u_int len, alen, elen;
7018 	int off;
7019 	int i;
7020 	struct sadb_alg *alg;
7021 
7022 	/* create new sadb_msg to reply. */
7023 	alen = 0;
7024 	for (i = 1; i <= SADB_AALG_MAX; i++) {
7025 		if (auth_algorithm_lookup(i))
7026 			alen += sizeof(struct sadb_alg);
7027 	}
7028 	if (alen)
7029 		alen += sizeof(struct sadb_supported);
7030 	elen = 0;
7031 	for (i = 1; i <= SADB_EALG_MAX; i++) {
7032 		if (enc_algorithm_lookup(i))
7033 			elen += sizeof(struct sadb_alg);
7034 	}
7035 	if (elen)
7036 		elen += sizeof(struct sadb_supported);
7037 
7038 	len = sizeof(struct sadb_msg) + alen + elen;
7039 
7040 	if (len > MCLBYTES)
7041 		return key_senderror(so, m, ENOBUFS);
7042 
7043 	MGETHDR(n, M_NOWAIT, MT_DATA);
7044 	if (len > MHLEN) {
7045 		if (!(MCLGET(n, M_NOWAIT))) {
7046 			m_freem(n);
7047 			n = NULL;
7048 		}
7049 	}
7050 	if (!n)
7051 		return key_senderror(so, m, ENOBUFS);
7052 
7053 	n->m_pkthdr.len = n->m_len = len;
7054 	n->m_next = NULL;
7055 	off = 0;
7056 
7057 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
7058 	newmsg = mtod(n, struct sadb_msg *);
7059 	newmsg->sadb_msg_errno = 0;
7060 	newmsg->sadb_msg_len = PFKEY_UNIT64(len);
7061 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7062 
7063 	/* for authentication algorithm */
7064 	if (alen) {
7065 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7066 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7067 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7068 		off += PFKEY_ALIGN8(sizeof(*sup));
7069 
7070 		for (i = 1; i <= SADB_AALG_MAX; i++) {
7071 			const struct auth_hash *aalgo;
7072 			u_int16_t minkeysize, maxkeysize;
7073 
7074 			aalgo = auth_algorithm_lookup(i);
7075 			if (!aalgo)
7076 				continue;
7077 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7078 			alg->sadb_alg_id = i;
7079 			alg->sadb_alg_ivlen = 0;
7080 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7081 			alg->sadb_alg_minbits = _BITS(minkeysize);
7082 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7083 			off += PFKEY_ALIGN8(sizeof(*alg));
7084 		}
7085 	}
7086 
7087 	/* for encryption algorithm */
7088 	if (elen) {
7089 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7090 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7091 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7092 		off += PFKEY_ALIGN8(sizeof(*sup));
7093 
7094 		for (i = 1; i <= SADB_EALG_MAX; i++) {
7095 			const struct enc_xform *ealgo;
7096 
7097 			ealgo = enc_algorithm_lookup(i);
7098 			if (!ealgo)
7099 				continue;
7100 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7101 			alg->sadb_alg_id = i;
7102 			alg->sadb_alg_ivlen = ealgo->ivsize;
7103 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7104 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7105 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7106 		}
7107 	}
7108 
7109 	IPSEC_ASSERT(off == len,
7110 		("length assumption failed (off %u len %u)", off, len));
7111 
7112 	m_freem(m);
7113 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7114     }
7115 }
7116 
7117 /*
7118  * free secreg entry registered.
7119  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7120  */
7121 void
7122 key_freereg(struct socket *so)
7123 {
7124 	struct secreg *reg;
7125 	int i;
7126 
7127 	IPSEC_ASSERT(so != NULL, ("NULL so"));
7128 
7129 	/*
7130 	 * check whether existing or not.
7131 	 * check all type of SA, because there is a potential that
7132 	 * one socket is registered to multiple type of SA.
7133 	 */
7134 	REGTREE_LOCK();
7135 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7136 		LIST_FOREACH(reg, &V_regtree[i], chain) {
7137 			if (reg->so == so && __LIST_CHAINED(reg)) {
7138 				LIST_REMOVE(reg, chain);
7139 				free(reg, M_IPSEC_SAR);
7140 				break;
7141 			}
7142 		}
7143 	}
7144 	REGTREE_UNLOCK();
7145 }
7146 
7147 /*
7148  * SADB_EXPIRE processing
7149  * send
7150  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7151  * to KMD by PF_KEY.
7152  * NOTE: We send only soft lifetime extension.
7153  *
7154  * OUT:	0	: succeed
7155  *	others	: error number
7156  */
7157 static int
7158 key_expire(struct secasvar *sav, int hard)
7159 {
7160 	struct mbuf *result = NULL, *m;
7161 	struct sadb_lifetime *lt;
7162 	uint32_t replay_count;
7163 	int error, len;
7164 	uint8_t satype;
7165 
7166 	IPSEC_ASSERT (sav != NULL, ("null sav"));
7167 	IPSEC_ASSERT (sav->sah != NULL, ("null sa header"));
7168 
7169 	KEYDBG(KEY_STAMP,
7170 	    printf("%s: SA(%p) expired %s lifetime\n", __func__,
7171 		sav, hard ? "hard": "soft"));
7172 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
7173 	/* set msg header */
7174 	satype = key_proto2satype(sav->sah->saidx.proto);
7175 	IPSEC_ASSERT(satype != 0, ("invalid proto, satype %u", satype));
7176 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, sav->refcnt);
7177 	if (!m) {
7178 		error = ENOBUFS;
7179 		goto fail;
7180 	}
7181 	result = m;
7182 
7183 	/* create SA extension */
7184 	m = key_setsadbsa(sav);
7185 	if (!m) {
7186 		error = ENOBUFS;
7187 		goto fail;
7188 	}
7189 	m_cat(result, m);
7190 
7191 	/* create SA extension */
7192 	SECASVAR_LOCK(sav);
7193 	replay_count = sav->replay ? sav->replay->count : 0;
7194 	SECASVAR_UNLOCK(sav);
7195 
7196 	m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
7197 			sav->sah->saidx.reqid);
7198 	if (!m) {
7199 		error = ENOBUFS;
7200 		goto fail;
7201 	}
7202 	m_cat(result, m);
7203 
7204 	if (sav->replay && sav->replay->wsize > UINT8_MAX) {
7205 		m = key_setsadbxsareplay(sav->replay->wsize);
7206 		if (!m) {
7207 			error = ENOBUFS;
7208 			goto fail;
7209 		}
7210 		m_cat(result, m);
7211 	}
7212 
7213 	/* create lifetime extension (current and soft) */
7214 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7215 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
7216 	if (m == NULL) {
7217 		error = ENOBUFS;
7218 		goto fail;
7219 	}
7220 	m_align(m, len);
7221 	m->m_len = len;
7222 	bzero(mtod(m, caddr_t), len);
7223 	lt = mtod(m, struct sadb_lifetime *);
7224 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7225 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7226 	lt->sadb_lifetime_allocations =
7227 	    (uint32_t)counter_u64_fetch(sav->lft_c_allocations);
7228 	lt->sadb_lifetime_bytes =
7229 	    counter_u64_fetch(sav->lft_c_bytes);
7230 	lt->sadb_lifetime_addtime = sav->created;
7231 	lt->sadb_lifetime_usetime = sav->firstused;
7232 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
7233 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7234 	if (hard) {
7235 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
7236 		lt->sadb_lifetime_allocations = sav->lft_h->allocations;
7237 		lt->sadb_lifetime_bytes = sav->lft_h->bytes;
7238 		lt->sadb_lifetime_addtime = sav->lft_h->addtime;
7239 		lt->sadb_lifetime_usetime = sav->lft_h->usetime;
7240 	} else {
7241 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_SOFT;
7242 		lt->sadb_lifetime_allocations = sav->lft_s->allocations;
7243 		lt->sadb_lifetime_bytes = sav->lft_s->bytes;
7244 		lt->sadb_lifetime_addtime = sav->lft_s->addtime;
7245 		lt->sadb_lifetime_usetime = sav->lft_s->usetime;
7246 	}
7247 	m_cat(result, m);
7248 
7249 	/* set sadb_address for source */
7250 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
7251 	    &sav->sah->saidx.src.sa,
7252 	    FULLMASK, IPSEC_ULPROTO_ANY);
7253 	if (!m) {
7254 		error = ENOBUFS;
7255 		goto fail;
7256 	}
7257 	m_cat(result, m);
7258 
7259 	/* set sadb_address for destination */
7260 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
7261 	    &sav->sah->saidx.dst.sa,
7262 	    FULLMASK, IPSEC_ULPROTO_ANY);
7263 	if (!m) {
7264 		error = ENOBUFS;
7265 		goto fail;
7266 	}
7267 	m_cat(result, m);
7268 
7269 	/*
7270 	 * XXX-BZ Handle NAT-T extensions here.
7271 	 * XXXAE: it doesn't seem quite useful. IKEs should not depend on
7272 	 * this information, we report only significant SA fields.
7273 	 */
7274 
7275 	if ((result->m_flags & M_PKTHDR) == 0) {
7276 		error = EINVAL;
7277 		goto fail;
7278 	}
7279 
7280 	if (result->m_len < sizeof(struct sadb_msg)) {
7281 		result = m_pullup(result, sizeof(struct sadb_msg));
7282 		if (result == NULL) {
7283 			error = ENOBUFS;
7284 			goto fail;
7285 		}
7286 	}
7287 
7288 	result->m_pkthdr.len = 0;
7289 	for (m = result; m; m = m->m_next)
7290 		result->m_pkthdr.len += m->m_len;
7291 
7292 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7293 	    PFKEY_UNIT64(result->m_pkthdr.len);
7294 
7295 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7296 
7297  fail:
7298 	if (result)
7299 		m_freem(result);
7300 	return error;
7301 }
7302 
7303 static void
7304 key_freesah_flushed(struct secashead_queue *flushq)
7305 {
7306 	struct secashead *sah, *nextsah;
7307 	struct secasvar *sav, *nextsav;
7308 
7309 	sah = TAILQ_FIRST(flushq);
7310 	while (sah != NULL) {
7311 		sav = TAILQ_FIRST(&sah->savtree_larval);
7312 		while (sav != NULL) {
7313 			nextsav = TAILQ_NEXT(sav, chain);
7314 			TAILQ_REMOVE(&sah->savtree_larval, sav, chain);
7315 			key_freesav(&sav); /* release last reference */
7316 			key_freesah(&sah); /* release reference from SAV */
7317 			sav = nextsav;
7318 		}
7319 		sav = TAILQ_FIRST(&sah->savtree_alive);
7320 		while (sav != NULL) {
7321 			nextsav = TAILQ_NEXT(sav, chain);
7322 			TAILQ_REMOVE(&sah->savtree_alive, sav, chain);
7323 			key_freesav(&sav); /* release last reference */
7324 			key_freesah(&sah); /* release reference from SAV */
7325 			sav = nextsav;
7326 		}
7327 		nextsah = TAILQ_NEXT(sah, chain);
7328 		key_freesah(&sah);	/* release last reference */
7329 		sah = nextsah;
7330 	}
7331 }
7332 
7333 /*
7334  * SADB_FLUSH processing
7335  * receive
7336  *   <base>
7337  * from the ikmpd, and free all entries in secastree.
7338  * and send,
7339  *   <base>
7340  * to the ikmpd.
7341  * NOTE: to do is only marking SADB_SASTATE_DEAD.
7342  *
7343  * m will always be freed.
7344  */
7345 static int
7346 key_flush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7347 {
7348 	struct secashead_queue flushq;
7349 	struct sadb_msg *newmsg;
7350 	struct secashead *sah, *nextsah;
7351 	struct secasvar *sav;
7352 	uint8_t proto;
7353 	int i;
7354 
7355 	IPSEC_ASSERT(so != NULL, ("null socket"));
7356 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7357 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7358 
7359 	/* map satype to proto */
7360 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7361 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7362 			__func__));
7363 		return key_senderror(so, m, EINVAL);
7364 	}
7365 	KEYDBG(KEY_STAMP,
7366 	    printf("%s: proto %u\n", __func__, proto));
7367 
7368 	TAILQ_INIT(&flushq);
7369 	if (proto == IPSEC_PROTO_ANY) {
7370 		/* no SATYPE specified, i.e. flushing all SA. */
7371 		SAHTREE_WLOCK();
7372 		/* Move all SAHs into flushq */
7373 		TAILQ_CONCAT(&flushq, &V_sahtree, chain);
7374 		/* Flush all buckets in SPI hash */
7375 		for (i = 0; i < V_savhash_mask + 1; i++)
7376 			LIST_INIT(&V_savhashtbl[i]);
7377 		/* Flush all buckets in SAHADDRHASH */
7378 		for (i = 0; i < V_sahaddrhash_mask + 1; i++)
7379 			LIST_INIT(&V_sahaddrhashtbl[i]);
7380 		/* Mark all SAHs as unlinked */
7381 		TAILQ_FOREACH(sah, &flushq, chain) {
7382 			sah->state = SADB_SASTATE_DEAD;
7383 			/*
7384 			 * Callout handler makes its job using
7385 			 * RLOCK and drain queues. In case, when this
7386 			 * function will be called just before it
7387 			 * acquires WLOCK, we need to mark SAs as
7388 			 * unlinked to prevent second unlink.
7389 			 */
7390 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7391 				sav->state = SADB_SASTATE_DEAD;
7392 			}
7393 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7394 				sav->state = SADB_SASTATE_DEAD;
7395 			}
7396 		}
7397 		SAHTREE_WUNLOCK();
7398 	} else {
7399 		SAHTREE_WLOCK();
7400 		sah = TAILQ_FIRST(&V_sahtree);
7401 		while (sah != NULL) {
7402 			IPSEC_ASSERT(sah->state != SADB_SASTATE_DEAD,
7403 			    ("DEAD SAH %p in SADB_FLUSH", sah));
7404 			nextsah = TAILQ_NEXT(sah, chain);
7405 			if (sah->saidx.proto != proto) {
7406 				sah = nextsah;
7407 				continue;
7408 			}
7409 			sah->state = SADB_SASTATE_DEAD;
7410 			TAILQ_REMOVE(&V_sahtree, sah, chain);
7411 			LIST_REMOVE(sah, addrhash);
7412 			/* Unlink all SAs from SPI hash */
7413 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7414 				LIST_REMOVE(sav, spihash);
7415 				sav->state = SADB_SASTATE_DEAD;
7416 			}
7417 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7418 				LIST_REMOVE(sav, spihash);
7419 				sav->state = SADB_SASTATE_DEAD;
7420 			}
7421 			/* Add SAH into flushq */
7422 			TAILQ_INSERT_HEAD(&flushq, sah, chain);
7423 			sah = nextsah;
7424 		}
7425 		SAHTREE_WUNLOCK();
7426 	}
7427 
7428 	key_freesah_flushed(&flushq);
7429 	/* Free all queued SAs and SAHs */
7430 	if (m->m_len < sizeof(struct sadb_msg) ||
7431 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7432 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7433 		return key_senderror(so, m, ENOBUFS);
7434 	}
7435 
7436 	if (m->m_next)
7437 		m_freem(m->m_next);
7438 	m->m_next = NULL;
7439 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7440 	newmsg = mtod(m, struct sadb_msg *);
7441 	newmsg->sadb_msg_errno = 0;
7442 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7443 
7444 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7445 }
7446 
7447 /*
7448  * SADB_DUMP processing
7449  * dump all entries including status of DEAD in SAD.
7450  * receive
7451  *   <base>
7452  * from the ikmpd, and dump all secasvar leaves
7453  * and send,
7454  *   <base> .....
7455  * to the ikmpd.
7456  *
7457  * m will always be freed.
7458  */
7459 static int
7460 key_dump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7461 {
7462 	SAHTREE_RLOCK_TRACKER;
7463 	struct secashead *sah;
7464 	struct secasvar *sav;
7465 	struct sadb_msg *newmsg;
7466 	struct mbuf *n;
7467 	uint32_t cnt;
7468 	uint8_t proto, satype;
7469 
7470 	IPSEC_ASSERT(so != NULL, ("null socket"));
7471 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7472 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7473 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7474 
7475 	/* map satype to proto */
7476 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7477 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7478 		    __func__));
7479 		return key_senderror(so, m, EINVAL);
7480 	}
7481 
7482 	/* count sav entries to be sent to the userland. */
7483 	cnt = 0;
7484 	SAHTREE_RLOCK();
7485 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7486 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7487 		    proto != sah->saidx.proto)
7488 			continue;
7489 
7490 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain)
7491 			cnt++;
7492 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain)
7493 			cnt++;
7494 	}
7495 
7496 	if (cnt == 0) {
7497 		SAHTREE_RUNLOCK();
7498 		return key_senderror(so, m, ENOENT);
7499 	}
7500 
7501 	/* send this to the userland, one at a time. */
7502 	newmsg = NULL;
7503 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7504 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7505 		    proto != sah->saidx.proto)
7506 			continue;
7507 
7508 		/* map proto to satype */
7509 		if ((satype = key_proto2satype(sah->saidx.proto)) == 0) {
7510 			SAHTREE_RUNLOCK();
7511 			ipseclog((LOG_DEBUG, "%s: there was invalid proto in "
7512 			    "SAD.\n", __func__));
7513 			return key_senderror(so, m, EINVAL);
7514 		}
7515 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7516 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7517 			    --cnt, mhp->msg->sadb_msg_pid);
7518 			if (n == NULL) {
7519 				SAHTREE_RUNLOCK();
7520 				return key_senderror(so, m, ENOBUFS);
7521 			}
7522 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7523 		}
7524 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7525 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7526 			    --cnt, mhp->msg->sadb_msg_pid);
7527 			if (n == NULL) {
7528 				SAHTREE_RUNLOCK();
7529 				return key_senderror(so, m, ENOBUFS);
7530 			}
7531 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7532 		}
7533 	}
7534 	SAHTREE_RUNLOCK();
7535 	m_freem(m);
7536 	return (0);
7537 }
7538 /*
7539  * SADB_X_PROMISC processing
7540  *
7541  * m will always be freed.
7542  */
7543 static int
7544 key_promisc(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7545 {
7546 	int olen;
7547 
7548 	IPSEC_ASSERT(so != NULL, ("null socket"));
7549 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7550 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7551 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7552 
7553 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7554 
7555 	if (olen < sizeof(struct sadb_msg)) {
7556 #if 1
7557 		return key_senderror(so, m, EINVAL);
7558 #else
7559 		m_freem(m);
7560 		return 0;
7561 #endif
7562 	} else if (olen == sizeof(struct sadb_msg)) {
7563 		/* enable/disable promisc mode */
7564 		struct keycb *kp;
7565 
7566 		if ((kp = (struct keycb *)sotorawcb(so)) == NULL)
7567 			return key_senderror(so, m, EINVAL);
7568 		mhp->msg->sadb_msg_errno = 0;
7569 		switch (mhp->msg->sadb_msg_satype) {
7570 		case 0:
7571 		case 1:
7572 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
7573 			break;
7574 		default:
7575 			return key_senderror(so, m, EINVAL);
7576 		}
7577 
7578 		/* send the original message back to everyone */
7579 		mhp->msg->sadb_msg_errno = 0;
7580 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7581 	} else {
7582 		/* send packet as is */
7583 
7584 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
7585 
7586 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
7587 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7588 	}
7589 }
7590 
7591 static int (*key_typesw[])(struct socket *, struct mbuf *,
7592 		const struct sadb_msghdr *) = {
7593 	NULL,		/* SADB_RESERVED */
7594 	key_getspi,	/* SADB_GETSPI */
7595 	key_update,	/* SADB_UPDATE */
7596 	key_add,	/* SADB_ADD */
7597 	key_delete,	/* SADB_DELETE */
7598 	key_get,	/* SADB_GET */
7599 	key_acquire2,	/* SADB_ACQUIRE */
7600 	key_register,	/* SADB_REGISTER */
7601 	NULL,		/* SADB_EXPIRE */
7602 	key_flush,	/* SADB_FLUSH */
7603 	key_dump,	/* SADB_DUMP */
7604 	key_promisc,	/* SADB_X_PROMISC */
7605 	NULL,		/* SADB_X_PCHANGE */
7606 	key_spdadd,	/* SADB_X_SPDUPDATE */
7607 	key_spdadd,	/* SADB_X_SPDADD */
7608 	key_spddelete,	/* SADB_X_SPDDELETE */
7609 	key_spdget,	/* SADB_X_SPDGET */
7610 	NULL,		/* SADB_X_SPDACQUIRE */
7611 	key_spddump,	/* SADB_X_SPDDUMP */
7612 	key_spdflush,	/* SADB_X_SPDFLUSH */
7613 	key_spdadd,	/* SADB_X_SPDSETIDX */
7614 	NULL,		/* SADB_X_SPDEXPIRE */
7615 	key_spddelete2,	/* SADB_X_SPDDELETE2 */
7616 };
7617 
7618 /*
7619  * parse sadb_msg buffer to process PFKEYv2,
7620  * and create a data to response if needed.
7621  * I think to be dealed with mbuf directly.
7622  * IN:
7623  *     msgp  : pointer to pointer to a received buffer pulluped.
7624  *             This is rewrited to response.
7625  *     so    : pointer to socket.
7626  * OUT:
7627  *    length for buffer to send to user process.
7628  */
7629 int
7630 key_parse(struct mbuf *m, struct socket *so)
7631 {
7632 	struct sadb_msg *msg;
7633 	struct sadb_msghdr mh;
7634 	u_int orglen;
7635 	int error;
7636 	int target;
7637 
7638 	IPSEC_ASSERT(so != NULL, ("null socket"));
7639 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7640 
7641 	if (m->m_len < sizeof(struct sadb_msg)) {
7642 		m = m_pullup(m, sizeof(struct sadb_msg));
7643 		if (!m)
7644 			return ENOBUFS;
7645 	}
7646 	msg = mtod(m, struct sadb_msg *);
7647 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
7648 	target = KEY_SENDUP_ONE;
7649 
7650 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len != orglen) {
7651 		ipseclog((LOG_DEBUG, "%s: invalid message length.\n",__func__));
7652 		PFKEYSTAT_INC(out_invlen);
7653 		error = EINVAL;
7654 		goto senderror;
7655 	}
7656 
7657 	if (msg->sadb_msg_version != PF_KEY_V2) {
7658 		ipseclog((LOG_DEBUG, "%s: PF_KEY version %u is mismatched.\n",
7659 		    __func__, msg->sadb_msg_version));
7660 		PFKEYSTAT_INC(out_invver);
7661 		error = EINVAL;
7662 		goto senderror;
7663 	}
7664 
7665 	if (msg->sadb_msg_type > SADB_MAX) {
7666 		ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7667 		    __func__, msg->sadb_msg_type));
7668 		PFKEYSTAT_INC(out_invmsgtype);
7669 		error = EINVAL;
7670 		goto senderror;
7671 	}
7672 
7673 	/* for old-fashioned code - should be nuked */
7674 	if (m->m_pkthdr.len > MCLBYTES) {
7675 		m_freem(m);
7676 		return ENOBUFS;
7677 	}
7678 	if (m->m_next) {
7679 		struct mbuf *n;
7680 
7681 		MGETHDR(n, M_NOWAIT, MT_DATA);
7682 		if (n && m->m_pkthdr.len > MHLEN) {
7683 			if (!(MCLGET(n, M_NOWAIT))) {
7684 				m_free(n);
7685 				n = NULL;
7686 			}
7687 		}
7688 		if (!n) {
7689 			m_freem(m);
7690 			return ENOBUFS;
7691 		}
7692 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
7693 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
7694 		n->m_next = NULL;
7695 		m_freem(m);
7696 		m = n;
7697 	}
7698 
7699 	/* align the mbuf chain so that extensions are in contiguous region. */
7700 	error = key_align(m, &mh);
7701 	if (error)
7702 		return error;
7703 
7704 	msg = mh.msg;
7705 
7706 	/* We use satype as scope mask for spddump */
7707 	if (msg->sadb_msg_type == SADB_X_SPDDUMP) {
7708 		switch (msg->sadb_msg_satype) {
7709 		case IPSEC_POLICYSCOPE_ANY:
7710 		case IPSEC_POLICYSCOPE_GLOBAL:
7711 		case IPSEC_POLICYSCOPE_IFNET:
7712 		case IPSEC_POLICYSCOPE_PCB:
7713 			break;
7714 		default:
7715 			ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7716 			    __func__, msg->sadb_msg_type));
7717 			PFKEYSTAT_INC(out_invsatype);
7718 			error = EINVAL;
7719 			goto senderror;
7720 		}
7721 	} else {
7722 		switch (msg->sadb_msg_satype) { /* check SA type */
7723 		case SADB_SATYPE_UNSPEC:
7724 			switch (msg->sadb_msg_type) {
7725 			case SADB_GETSPI:
7726 			case SADB_UPDATE:
7727 			case SADB_ADD:
7728 			case SADB_DELETE:
7729 			case SADB_GET:
7730 			case SADB_ACQUIRE:
7731 			case SADB_EXPIRE:
7732 				ipseclog((LOG_DEBUG, "%s: must specify satype "
7733 				    "when msg type=%u.\n", __func__,
7734 				    msg->sadb_msg_type));
7735 				PFKEYSTAT_INC(out_invsatype);
7736 				error = EINVAL;
7737 				goto senderror;
7738 			}
7739 			break;
7740 		case SADB_SATYPE_AH:
7741 		case SADB_SATYPE_ESP:
7742 		case SADB_X_SATYPE_IPCOMP:
7743 		case SADB_X_SATYPE_TCPSIGNATURE:
7744 			switch (msg->sadb_msg_type) {
7745 			case SADB_X_SPDADD:
7746 			case SADB_X_SPDDELETE:
7747 			case SADB_X_SPDGET:
7748 			case SADB_X_SPDFLUSH:
7749 			case SADB_X_SPDSETIDX:
7750 			case SADB_X_SPDUPDATE:
7751 			case SADB_X_SPDDELETE2:
7752 				ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7753 				    __func__, msg->sadb_msg_type));
7754 				PFKEYSTAT_INC(out_invsatype);
7755 				error = EINVAL;
7756 				goto senderror;
7757 			}
7758 			break;
7759 		case SADB_SATYPE_RSVP:
7760 		case SADB_SATYPE_OSPFV2:
7761 		case SADB_SATYPE_RIPV2:
7762 		case SADB_SATYPE_MIP:
7763 			ipseclog((LOG_DEBUG, "%s: type %u isn't supported.\n",
7764 			    __func__, msg->sadb_msg_satype));
7765 			PFKEYSTAT_INC(out_invsatype);
7766 			error = EOPNOTSUPP;
7767 			goto senderror;
7768 		case 1:	/* XXX: What does it do? */
7769 			if (msg->sadb_msg_type == SADB_X_PROMISC)
7770 				break;
7771 			/*FALLTHROUGH*/
7772 		default:
7773 			ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7774 			    __func__, msg->sadb_msg_satype));
7775 			PFKEYSTAT_INC(out_invsatype);
7776 			error = EINVAL;
7777 			goto senderror;
7778 		}
7779 	}
7780 
7781 	/* check field of upper layer protocol and address family */
7782 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL
7783 	 && mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
7784 		struct sadb_address *src0, *dst0;
7785 		u_int plen;
7786 
7787 		src0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_SRC]);
7788 		dst0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_DST]);
7789 
7790 		/* check upper layer protocol */
7791 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
7792 			ipseclog((LOG_DEBUG, "%s: upper layer protocol "
7793 				"mismatched.\n", __func__));
7794 			PFKEYSTAT_INC(out_invaddr);
7795 			error = EINVAL;
7796 			goto senderror;
7797 		}
7798 
7799 		/* check family */
7800 		if (PFKEY_ADDR_SADDR(src0)->sa_family !=
7801 		    PFKEY_ADDR_SADDR(dst0)->sa_family) {
7802 			ipseclog((LOG_DEBUG, "%s: address family mismatched.\n",
7803 				__func__));
7804 			PFKEYSTAT_INC(out_invaddr);
7805 			error = EINVAL;
7806 			goto senderror;
7807 		}
7808 		if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7809 		    PFKEY_ADDR_SADDR(dst0)->sa_len) {
7810 			ipseclog((LOG_DEBUG, "%s: address struct size "
7811 				"mismatched.\n", __func__));
7812 			PFKEYSTAT_INC(out_invaddr);
7813 			error = EINVAL;
7814 			goto senderror;
7815 		}
7816 
7817 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7818 		case AF_INET:
7819 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7820 			    sizeof(struct sockaddr_in)) {
7821 				PFKEYSTAT_INC(out_invaddr);
7822 				error = EINVAL;
7823 				goto senderror;
7824 			}
7825 			break;
7826 		case AF_INET6:
7827 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7828 			    sizeof(struct sockaddr_in6)) {
7829 				PFKEYSTAT_INC(out_invaddr);
7830 				error = EINVAL;
7831 				goto senderror;
7832 			}
7833 			break;
7834 		default:
7835 			ipseclog((LOG_DEBUG, "%s: unsupported address family\n",
7836 				__func__));
7837 			PFKEYSTAT_INC(out_invaddr);
7838 			error = EAFNOSUPPORT;
7839 			goto senderror;
7840 		}
7841 
7842 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7843 		case AF_INET:
7844 			plen = sizeof(struct in_addr) << 3;
7845 			break;
7846 		case AF_INET6:
7847 			plen = sizeof(struct in6_addr) << 3;
7848 			break;
7849 		default:
7850 			plen = 0;	/*fool gcc*/
7851 			break;
7852 		}
7853 
7854 		/* check max prefix length */
7855 		if (src0->sadb_address_prefixlen > plen ||
7856 		    dst0->sadb_address_prefixlen > plen) {
7857 			ipseclog((LOG_DEBUG, "%s: illegal prefixlen.\n",
7858 				__func__));
7859 			PFKEYSTAT_INC(out_invaddr);
7860 			error = EINVAL;
7861 			goto senderror;
7862 		}
7863 
7864 		/*
7865 		 * prefixlen == 0 is valid because there can be a case when
7866 		 * all addresses are matched.
7867 		 */
7868 	}
7869 
7870 	if (msg->sadb_msg_type >= nitems(key_typesw) ||
7871 	    key_typesw[msg->sadb_msg_type] == NULL) {
7872 		PFKEYSTAT_INC(out_invmsgtype);
7873 		error = EINVAL;
7874 		goto senderror;
7875 	}
7876 
7877 	return (*key_typesw[msg->sadb_msg_type])(so, m, &mh);
7878 
7879 senderror:
7880 	msg->sadb_msg_errno = error;
7881 	return key_sendup_mbuf(so, m, target);
7882 }
7883 
7884 static int
7885 key_senderror(struct socket *so, struct mbuf *m, int code)
7886 {
7887 	struct sadb_msg *msg;
7888 
7889 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
7890 		("mbuf too small, len %u", m->m_len));
7891 
7892 	msg = mtod(m, struct sadb_msg *);
7893 	msg->sadb_msg_errno = code;
7894 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
7895 }
7896 
7897 /*
7898  * set the pointer to each header into message buffer.
7899  * m will be freed on error.
7900  * XXX larger-than-MCLBYTES extension?
7901  */
7902 static int
7903 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
7904 {
7905 	struct mbuf *n;
7906 	struct sadb_ext *ext;
7907 	size_t off, end;
7908 	int extlen;
7909 	int toff;
7910 
7911 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7912 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7913 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
7914 		("mbuf too small, len %u", m->m_len));
7915 
7916 	/* initialize */
7917 	bzero(mhp, sizeof(*mhp));
7918 
7919 	mhp->msg = mtod(m, struct sadb_msg *);
7920 	mhp->ext[0] = (struct sadb_ext *)mhp->msg;	/*XXX backward compat */
7921 
7922 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7923 	extlen = end;	/*just in case extlen is not updated*/
7924 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
7925 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
7926 		if (!n) {
7927 			/* m is already freed */
7928 			return ENOBUFS;
7929 		}
7930 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
7931 
7932 		/* set pointer */
7933 		switch (ext->sadb_ext_type) {
7934 		case SADB_EXT_SA:
7935 		case SADB_EXT_ADDRESS_SRC:
7936 		case SADB_EXT_ADDRESS_DST:
7937 		case SADB_EXT_ADDRESS_PROXY:
7938 		case SADB_EXT_LIFETIME_CURRENT:
7939 		case SADB_EXT_LIFETIME_HARD:
7940 		case SADB_EXT_LIFETIME_SOFT:
7941 		case SADB_EXT_KEY_AUTH:
7942 		case SADB_EXT_KEY_ENCRYPT:
7943 		case SADB_EXT_IDENTITY_SRC:
7944 		case SADB_EXT_IDENTITY_DST:
7945 		case SADB_EXT_SENSITIVITY:
7946 		case SADB_EXT_PROPOSAL:
7947 		case SADB_EXT_SUPPORTED_AUTH:
7948 		case SADB_EXT_SUPPORTED_ENCRYPT:
7949 		case SADB_EXT_SPIRANGE:
7950 		case SADB_X_EXT_POLICY:
7951 		case SADB_X_EXT_SA2:
7952 		case SADB_X_EXT_NAT_T_TYPE:
7953 		case SADB_X_EXT_NAT_T_SPORT:
7954 		case SADB_X_EXT_NAT_T_DPORT:
7955 		case SADB_X_EXT_NAT_T_OAI:
7956 		case SADB_X_EXT_NAT_T_OAR:
7957 		case SADB_X_EXT_NAT_T_FRAG:
7958 		case SADB_X_EXT_SA_REPLAY:
7959 		case SADB_X_EXT_NEW_ADDRESS_SRC:
7960 		case SADB_X_EXT_NEW_ADDRESS_DST:
7961 			/* duplicate check */
7962 			/*
7963 			 * XXX Are there duplication payloads of either
7964 			 * KEY_AUTH or KEY_ENCRYPT ?
7965 			 */
7966 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
7967 				ipseclog((LOG_DEBUG, "%s: duplicate ext_type "
7968 					"%u\n", __func__, ext->sadb_ext_type));
7969 				m_freem(m);
7970 				PFKEYSTAT_INC(out_dupext);
7971 				return EINVAL;
7972 			}
7973 			break;
7974 		default:
7975 			ipseclog((LOG_DEBUG, "%s: invalid ext_type %u\n",
7976 				__func__, ext->sadb_ext_type));
7977 			m_freem(m);
7978 			PFKEYSTAT_INC(out_invexttype);
7979 			return EINVAL;
7980 		}
7981 
7982 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
7983 
7984 		if (key_validate_ext(ext, extlen)) {
7985 			m_freem(m);
7986 			PFKEYSTAT_INC(out_invlen);
7987 			return EINVAL;
7988 		}
7989 
7990 		n = m_pulldown(m, off, extlen, &toff);
7991 		if (!n) {
7992 			/* m is already freed */
7993 			return ENOBUFS;
7994 		}
7995 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
7996 
7997 		mhp->ext[ext->sadb_ext_type] = ext;
7998 		mhp->extoff[ext->sadb_ext_type] = off;
7999 		mhp->extlen[ext->sadb_ext_type] = extlen;
8000 	}
8001 
8002 	if (off != end) {
8003 		m_freem(m);
8004 		PFKEYSTAT_INC(out_invlen);
8005 		return EINVAL;
8006 	}
8007 
8008 	return 0;
8009 }
8010 
8011 static int
8012 key_validate_ext(const struct sadb_ext *ext, int len)
8013 {
8014 	const struct sockaddr *sa;
8015 	enum { NONE, ADDR } checktype = NONE;
8016 	int baselen = 0;
8017 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
8018 
8019 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
8020 		return EINVAL;
8021 
8022 	/* if it does not match minimum/maximum length, bail */
8023 	if (ext->sadb_ext_type >= nitems(minsize) ||
8024 	    ext->sadb_ext_type >= nitems(maxsize))
8025 		return EINVAL;
8026 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
8027 		return EINVAL;
8028 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
8029 		return EINVAL;
8030 
8031 	/* more checks based on sadb_ext_type XXX need more */
8032 	switch (ext->sadb_ext_type) {
8033 	case SADB_EXT_ADDRESS_SRC:
8034 	case SADB_EXT_ADDRESS_DST:
8035 	case SADB_EXT_ADDRESS_PROXY:
8036 	case SADB_X_EXT_NAT_T_OAI:
8037 	case SADB_X_EXT_NAT_T_OAR:
8038 	case SADB_X_EXT_NEW_ADDRESS_SRC:
8039 	case SADB_X_EXT_NEW_ADDRESS_DST:
8040 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
8041 		checktype = ADDR;
8042 		break;
8043 	case SADB_EXT_IDENTITY_SRC:
8044 	case SADB_EXT_IDENTITY_DST:
8045 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
8046 		    SADB_X_IDENTTYPE_ADDR) {
8047 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
8048 			checktype = ADDR;
8049 		} else
8050 			checktype = NONE;
8051 		break;
8052 	default:
8053 		checktype = NONE;
8054 		break;
8055 	}
8056 
8057 	switch (checktype) {
8058 	case NONE:
8059 		break;
8060 	case ADDR:
8061 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8062 		if (len < baselen + sal)
8063 			return EINVAL;
8064 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8065 			return EINVAL;
8066 		break;
8067 	}
8068 
8069 	return 0;
8070 }
8071 
8072 void
8073 key_init(void)
8074 {
8075 	int i;
8076 
8077 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8078 		TAILQ_INIT(&V_sptree[i]);
8079 		TAILQ_INIT(&V_sptree_ifnet[i]);
8080 	}
8081 
8082 	V_key_lft_zone = uma_zcreate("IPsec SA lft_c",
8083 	    sizeof(uint64_t) * 2, NULL, NULL, NULL, NULL,
8084 	    UMA_ALIGN_PTR, UMA_ZONE_PCPU);
8085 
8086 	TAILQ_INIT(&V_sahtree);
8087 	V_sphashtbl = hashinit(SPHASH_NHASH, M_IPSEC_SP, &V_sphash_mask);
8088 	V_savhashtbl = hashinit(SAVHASH_NHASH, M_IPSEC_SA, &V_savhash_mask);
8089 	V_sahaddrhashtbl = hashinit(SAHHASH_NHASH, M_IPSEC_SAH,
8090 	    &V_sahaddrhash_mask);
8091 	V_acqaddrhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8092 	    &V_acqaddrhash_mask);
8093 	V_acqseqhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8094 	    &V_acqseqhash_mask);
8095 
8096 	for (i = 0; i <= SADB_SATYPE_MAX; i++)
8097 		LIST_INIT(&V_regtree[i]);
8098 
8099 	LIST_INIT(&V_acqtree);
8100 	LIST_INIT(&V_spacqtree);
8101 
8102 	if (!IS_DEFAULT_VNET(curvnet))
8103 		return;
8104 
8105 	XFORMS_LOCK_INIT();
8106 	SPTREE_LOCK_INIT();
8107 	REGTREE_LOCK_INIT();
8108 	SAHTREE_LOCK_INIT();
8109 	ACQ_LOCK_INIT();
8110 	SPACQ_LOCK_INIT();
8111 
8112 #ifndef IPSEC_DEBUG2
8113 	callout_init(&key_timer, 1);
8114 	callout_reset(&key_timer, hz, key_timehandler, NULL);
8115 #endif /*IPSEC_DEBUG2*/
8116 
8117 	/* initialize key statistics */
8118 	keystat.getspi_count = 1;
8119 
8120 	if (bootverbose)
8121 		printf("IPsec: Initialized Security Association Processing.\n");
8122 }
8123 
8124 #ifdef VIMAGE
8125 void
8126 key_destroy(void)
8127 {
8128 	struct secashead_queue sahdrainq;
8129 	struct secpolicy_queue drainq;
8130 	struct secpolicy *sp, *nextsp;
8131 	struct secacq *acq, *nextacq;
8132 	struct secspacq *spacq, *nextspacq;
8133 	struct secashead *sah;
8134 	struct secasvar *sav;
8135 	struct secreg *reg;
8136 	int i;
8137 
8138 	/*
8139 	 * XXX: can we just call free() for each object without
8140 	 * walking through safe way with releasing references?
8141 	 */
8142 	TAILQ_INIT(&drainq);
8143 	SPTREE_WLOCK();
8144 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8145 		TAILQ_CONCAT(&drainq, &V_sptree[i], chain);
8146 		TAILQ_CONCAT(&drainq, &V_sptree_ifnet[i], chain);
8147 	}
8148 	SPTREE_WUNLOCK();
8149 	sp = TAILQ_FIRST(&drainq);
8150 	while (sp != NULL) {
8151 		nextsp = TAILQ_NEXT(sp, chain);
8152 		key_freesp(&sp);
8153 		sp = nextsp;
8154 	}
8155 
8156 	TAILQ_INIT(&sahdrainq);
8157 	SAHTREE_WLOCK();
8158 	TAILQ_CONCAT(&sahdrainq, &V_sahtree, chain);
8159 	for (i = 0; i < V_savhash_mask + 1; i++)
8160 		LIST_INIT(&V_savhashtbl[i]);
8161 	for (i = 0; i < V_sahaddrhash_mask + 1; i++)
8162 		LIST_INIT(&V_sahaddrhashtbl[i]);
8163 	TAILQ_FOREACH(sah, &sahdrainq, chain) {
8164 		sah->state = SADB_SASTATE_DEAD;
8165 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
8166 			sav->state = SADB_SASTATE_DEAD;
8167 		}
8168 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
8169 			sav->state = SADB_SASTATE_DEAD;
8170 		}
8171 	}
8172 	SAHTREE_WUNLOCK();
8173 
8174 	key_freesah_flushed(&sahdrainq);
8175 	hashdestroy(V_sphashtbl, M_IPSEC_SP, V_sphash_mask);
8176 	hashdestroy(V_savhashtbl, M_IPSEC_SA, V_savhash_mask);
8177 	hashdestroy(V_sahaddrhashtbl, M_IPSEC_SAH, V_sahaddrhash_mask);
8178 
8179 	REGTREE_LOCK();
8180 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8181 		LIST_FOREACH(reg, &V_regtree[i], chain) {
8182 			if (__LIST_CHAINED(reg)) {
8183 				LIST_REMOVE(reg, chain);
8184 				free(reg, M_IPSEC_SAR);
8185 				break;
8186 			}
8187 		}
8188 	}
8189 	REGTREE_UNLOCK();
8190 
8191 	ACQ_LOCK();
8192 	acq = LIST_FIRST(&V_acqtree);
8193 	while (acq != NULL) {
8194 		nextacq = LIST_NEXT(acq, chain);
8195 		LIST_REMOVE(acq, chain);
8196 		free(acq, M_IPSEC_SAQ);
8197 		acq = nextacq;
8198 	}
8199 	ACQ_UNLOCK();
8200 
8201 	SPACQ_LOCK();
8202 	for (spacq = LIST_FIRST(&V_spacqtree); spacq != NULL;
8203 	    spacq = nextspacq) {
8204 		nextspacq = LIST_NEXT(spacq, chain);
8205 		if (__LIST_CHAINED(spacq)) {
8206 			LIST_REMOVE(spacq, chain);
8207 			free(spacq, M_IPSEC_SAQ);
8208 		}
8209 	}
8210 	SPACQ_UNLOCK();
8211 	hashdestroy(V_acqaddrhashtbl, M_IPSEC_SAQ, V_acqaddrhash_mask);
8212 	hashdestroy(V_acqseqhashtbl, M_IPSEC_SAQ, V_acqseqhash_mask);
8213 	uma_zdestroy(V_key_lft_zone);
8214 }
8215 #endif
8216 
8217 /* record data transfer on SA, and update timestamps */
8218 void
8219 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8220 {
8221 	IPSEC_ASSERT(sav != NULL, ("Null secasvar"));
8222 	IPSEC_ASSERT(m != NULL, ("Null mbuf"));
8223 
8224 	/*
8225 	 * XXX Currently, there is a difference of bytes size
8226 	 * between inbound and outbound processing.
8227 	 */
8228 	counter_u64_add(sav->lft_c_bytes, m->m_pkthdr.len);
8229 
8230 	/*
8231 	 * We use the number of packets as the unit of
8232 	 * allocations.  We increment the variable
8233 	 * whenever {esp,ah}_{in,out}put is called.
8234 	 */
8235 	counter_u64_add(sav->lft_c_allocations, 1);
8236 
8237 	/*
8238 	 * NOTE: We record CURRENT usetime by using wall clock,
8239 	 * in seconds.  HARD and SOFT lifetime are measured by the time
8240 	 * difference (again in seconds) from usetime.
8241 	 *
8242 	 *	usetime
8243 	 *	v     expire   expire
8244 	 * -----+-----+--------+---> t
8245 	 *	<--------------> HARD
8246 	 *	<-----> SOFT
8247 	 */
8248 	if (sav->firstused == 0)
8249 		sav->firstused = time_second;
8250 }
8251 
8252 /*
8253  * Take one of the kernel's security keys and convert it into a PF_KEY
8254  * structure within an mbuf, suitable for sending up to a waiting
8255  * application in user land.
8256  *
8257  * IN:
8258  *    src: A pointer to a kernel security key.
8259  *    exttype: Which type of key this is. Refer to the PF_KEY data structures.
8260  * OUT:
8261  *    a valid mbuf or NULL indicating an error
8262  *
8263  */
8264 
8265 static struct mbuf *
8266 key_setkey(struct seckey *src, uint16_t exttype)
8267 {
8268 	struct mbuf *m;
8269 	struct sadb_key *p;
8270 	int len;
8271 
8272 	if (src == NULL)
8273 		return NULL;
8274 
8275 	len = PFKEY_ALIGN8(sizeof(struct sadb_key) + _KEYLEN(src));
8276 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8277 	if (m == NULL)
8278 		return NULL;
8279 	m_align(m, len);
8280 	m->m_len = len;
8281 	p = mtod(m, struct sadb_key *);
8282 	bzero(p, len);
8283 	p->sadb_key_len = PFKEY_UNIT64(len);
8284 	p->sadb_key_exttype = exttype;
8285 	p->sadb_key_bits = src->bits;
8286 	bcopy(src->key_data, _KEYBUF(p), _KEYLEN(src));
8287 
8288 	return m;
8289 }
8290 
8291 /*
8292  * Take one of the kernel's lifetime data structures and convert it
8293  * into a PF_KEY structure within an mbuf, suitable for sending up to
8294  * a waiting application in user land.
8295  *
8296  * IN:
8297  *    src: A pointer to a kernel lifetime structure.
8298  *    exttype: Which type of lifetime this is. Refer to the PF_KEY
8299  *             data structures for more information.
8300  * OUT:
8301  *    a valid mbuf or NULL indicating an error
8302  *
8303  */
8304 
8305 static struct mbuf *
8306 key_setlifetime(struct seclifetime *src, uint16_t exttype)
8307 {
8308 	struct mbuf *m = NULL;
8309 	struct sadb_lifetime *p;
8310 	int len = PFKEY_ALIGN8(sizeof(struct sadb_lifetime));
8311 
8312 	if (src == NULL)
8313 		return NULL;
8314 
8315 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8316 	if (m == NULL)
8317 		return m;
8318 	m_align(m, len);
8319 	m->m_len = len;
8320 	p = mtod(m, struct sadb_lifetime *);
8321 
8322 	bzero(p, len);
8323 	p->sadb_lifetime_len = PFKEY_UNIT64(len);
8324 	p->sadb_lifetime_exttype = exttype;
8325 	p->sadb_lifetime_allocations = src->allocations;
8326 	p->sadb_lifetime_bytes = src->bytes;
8327 	p->sadb_lifetime_addtime = src->addtime;
8328 	p->sadb_lifetime_usetime = src->usetime;
8329 
8330 	return m;
8331 
8332 }
8333 
8334 const struct enc_xform *
8335 enc_algorithm_lookup(int alg)
8336 {
8337 	int i;
8338 
8339 	for (i = 0; i < nitems(supported_ealgs); i++)
8340 		if (alg == supported_ealgs[i].sadb_alg)
8341 			return (supported_ealgs[i].xform);
8342 	return (NULL);
8343 }
8344 
8345 const struct auth_hash *
8346 auth_algorithm_lookup(int alg)
8347 {
8348 	int i;
8349 
8350 	for (i = 0; i < nitems(supported_aalgs); i++)
8351 		if (alg == supported_aalgs[i].sadb_alg)
8352 			return (supported_aalgs[i].xform);
8353 	return (NULL);
8354 }
8355 
8356 const struct comp_algo *
8357 comp_algorithm_lookup(int alg)
8358 {
8359 	int i;
8360 
8361 	for (i = 0; i < nitems(supported_calgs); i++)
8362 		if (alg == supported_calgs[i].sadb_alg)
8363 			return (supported_calgs[i].xform);
8364 	return (NULL);
8365 }
8366 
8367 /*
8368  * Register a transform.
8369  */
8370 static int
8371 xform_register(struct xformsw* xsp)
8372 {
8373 	struct xformsw *entry;
8374 
8375 	XFORMS_LOCK();
8376 	LIST_FOREACH(entry, &xforms, chain) {
8377 		if (entry->xf_type == xsp->xf_type) {
8378 			XFORMS_UNLOCK();
8379 			return (EEXIST);
8380 		}
8381 	}
8382 	LIST_INSERT_HEAD(&xforms, xsp, chain);
8383 	XFORMS_UNLOCK();
8384 	return (0);
8385 }
8386 
8387 void
8388 xform_attach(void *data)
8389 {
8390 	struct xformsw *xsp = (struct xformsw *)data;
8391 
8392 	if (xform_register(xsp) != 0)
8393 		printf("%s: failed to register %s xform\n", __func__,
8394 		    xsp->xf_name);
8395 }
8396 
8397 void
8398 xform_detach(void *data)
8399 {
8400 	struct xformsw *xsp = (struct xformsw *)data;
8401 
8402 	XFORMS_LOCK();
8403 	LIST_REMOVE(xsp, chain);
8404 	XFORMS_UNLOCK();
8405 
8406 	/* Delete all SAs related to this xform. */
8407 	key_delete_xform(xsp);
8408 }
8409 
8410 /*
8411  * Initialize transform support in an sav.
8412  */
8413 static int
8414 xform_init(struct secasvar *sav, u_short xftype)
8415 {
8416 	struct xformsw *entry;
8417 	int ret;
8418 
8419 	IPSEC_ASSERT(sav->tdb_xform == NULL,
8420 	    ("tdb_xform is already initialized"));
8421 
8422 	ret = EINVAL;
8423 	XFORMS_LOCK();
8424 	LIST_FOREACH(entry, &xforms, chain) {
8425 	    if (entry->xf_type == xftype) {
8426 		    ret = (*entry->xf_init)(sav, entry);
8427 		    break;
8428 	    }
8429 	}
8430 	XFORMS_UNLOCK();
8431 	return (ret);
8432 }
8433 
8434