xref: /freebsd/sys/netipsec/key.c (revision 076ad2f8)
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 			break;
868 	}
869 	if (sah != NULL) {
870 		if (V_key_preferred_oldsa)
871 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
872 		else
873 			sav = TAILQ_FIRST(&sah->savtree_alive);
874 		if (sav != NULL)
875 			SAV_ADDREF(sav);
876 	} else
877 		sav = NULL;
878 	SAHTREE_RUNLOCK();
879 
880 	if (sav != NULL) {
881 		KEYDBG(IPSEC_STAMP,
882 		    printf("%s: return SA(%p)\n", __func__, sav));
883 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
884 	} else {
885 		KEYDBG(IPSEC_STAMP,
886 		    printf("%s: SA not found\n", __func__));
887 		KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
888 	}
889 	return (sav);
890 }
891 
892 /*
893  * Allocating an SA entry for an *OUTBOUND* packet.
894  * OUT:	positive:	corresponding SA for given saidx found.
895  *	NULL:		SA not found, but will be acquired, check *error
896  *			for acquiring status.
897  */
898 struct secasvar *
899 key_allocsa_policy(struct secpolicy *sp, const struct secasindex *saidx,
900     int *error)
901 {
902 	SAHTREE_RLOCK_TRACKER;
903 	struct secashead *sah;
904 	struct secasvar *sav;
905 
906 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
907 	IPSEC_ASSERT(saidx->mode == IPSEC_MODE_TRANSPORT ||
908 		saidx->mode == IPSEC_MODE_TUNNEL,
909 		("unexpected policy %u", saidx->mode));
910 
911 	/*
912 	 * We check new SA in the IPsec request because a different
913 	 * SA may be involved each time this request is checked, either
914 	 * because new SAs are being configured, or this request is
915 	 * associated with an unconnected datagram socket, or this request
916 	 * is associated with a system default policy.
917 	 */
918 	SAHTREE_RLOCK();
919 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
920 		KEYDBG(IPSEC_DUMP,
921 		    printf("%s: checking SAH\n", __func__);
922 		    kdebug_secash(sah, "  "));
923 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID))
924 			break;
925 
926 	}
927 	if (sah != NULL) {
928 		/*
929 		 * Allocate the oldest SA available according to
930 		 * draft-jenkins-ipsec-rekeying-03.
931 		 */
932 		if (V_key_preferred_oldsa)
933 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
934 		else
935 			sav = TAILQ_FIRST(&sah->savtree_alive);
936 		if (sav != NULL)
937 			SAV_ADDREF(sav);
938 	} else
939 		sav = NULL;
940 	SAHTREE_RUNLOCK();
941 
942 	if (sav != NULL) {
943 		*error = 0;
944 		KEYDBG(IPSEC_STAMP,
945 		    printf("%s: chosen SA(%p) for SP(%p)\n", __func__,
946 			sav, sp));
947 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
948 		return (sav); /* return referenced SA */
949 	}
950 
951 	/* there is no SA */
952 	*error = key_acquire(saidx, sp);
953 	if ((*error) != 0)
954 		ipseclog((LOG_DEBUG,
955 		    "%s: error %d returned from key_acquire()\n",
956 			__func__, *error));
957 	KEYDBG(IPSEC_STAMP,
958 	    printf("%s: acquire SA for SP(%p), error %d\n",
959 		__func__, sp, *error));
960 	KEYDBG(IPSEC_DATA, kdebug_secasindex(saidx, NULL));
961 	return (NULL);
962 }
963 
964 /*
965  * allocating a usable SA entry for a *INBOUND* packet.
966  * Must call key_freesav() later.
967  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
968  *	NULL:		not found, or error occurred.
969  *
970  * According to RFC 2401 SA is uniquely identified by a triple SPI,
971  * destination address, and security protocol. But according to RFC 4301,
972  * SPI by itself suffices to specify an SA.
973  *
974  * Note that, however, we do need to keep source address in IPsec SA.
975  * IKE specification and PF_KEY specification do assume that we
976  * keep source address in IPsec SA.  We see a tricky situation here.
977  */
978 struct secasvar *
979 key_allocsa(union sockaddr_union *dst, uint8_t proto, uint32_t spi)
980 {
981 	SAHTREE_RLOCK_TRACKER;
982 	struct secasvar *sav;
983 
984 	IPSEC_ASSERT(proto == IPPROTO_ESP || proto == IPPROTO_AH ||
985 	    proto == IPPROTO_IPCOMP, ("unexpected security protocol %u",
986 	    proto));
987 
988 	SAHTREE_RLOCK();
989 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
990 		if (sav->spi == spi)
991 			break;
992 	}
993 	/*
994 	 * We use single SPI namespace for all protocols, so it is
995 	 * impossible to have SPI duplicates in the SAVHASH.
996 	 */
997 	if (sav != NULL) {
998 		if (sav->state != SADB_SASTATE_LARVAL &&
999 		    sav->sah->saidx.proto == proto &&
1000 		    key_sockaddrcmp(&dst->sa,
1001 			&sav->sah->saidx.dst.sa, 0) == 0)
1002 			SAV_ADDREF(sav);
1003 		else
1004 			sav = NULL;
1005 	}
1006 	SAHTREE_RUNLOCK();
1007 
1008 	if (sav == NULL) {
1009 		KEYDBG(IPSEC_STAMP,
1010 		    char buf[IPSEC_ADDRSTRLEN];
1011 		    printf("%s: SA not found for spi %u proto %u dst %s\n",
1012 			__func__, ntohl(spi), proto, ipsec_address(dst, buf,
1013 			sizeof(buf))));
1014 	} else {
1015 		KEYDBG(IPSEC_STAMP,
1016 		    printf("%s: return SA(%p)\n", __func__, sav));
1017 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1018 	}
1019 	return (sav);
1020 }
1021 
1022 struct secasvar *
1023 key_allocsa_tunnel(union sockaddr_union *src, union sockaddr_union *dst,
1024     uint8_t proto)
1025 {
1026 	SAHTREE_RLOCK_TRACKER;
1027 	struct secasindex saidx;
1028 	struct secashead *sah;
1029 	struct secasvar *sav;
1030 
1031 	IPSEC_ASSERT(src != NULL, ("null src address"));
1032 	IPSEC_ASSERT(dst != NULL, ("null dst address"));
1033 
1034 	KEY_SETSECASIDX(proto, IPSEC_MODE_TUNNEL, 0, &src->sa,
1035 	    &dst->sa, &saidx);
1036 
1037 	sav = NULL;
1038 	SAHTREE_RLOCK();
1039 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
1040 		if (IPSEC_MODE_TUNNEL != sah->saidx.mode)
1041 			continue;
1042 		if (proto != sah->saidx.proto)
1043 			continue;
1044 		if (key_sockaddrcmp(&src->sa, &sav->sah->saidx.src.sa, 0) != 0)
1045 			continue;
1046 		if (key_sockaddrcmp(&dst->sa, &sav->sah->saidx.dst.sa, 0) != 0)
1047 			continue;
1048 		/* XXXAE: is key_preferred_oldsa reasonably?*/
1049 		if (V_key_preferred_oldsa)
1050 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
1051 		else
1052 			sav = TAILQ_FIRST(&sah->savtree_alive);
1053 		if (sav != NULL) {
1054 			SAV_ADDREF(sav);
1055 			break;
1056 		}
1057 	}
1058 	SAHTREE_RUNLOCK();
1059 	KEYDBG(IPSEC_STAMP,
1060 	    printf("%s: return SA(%p)\n", __func__, sav));
1061 	if (sav != NULL)
1062 		KEYDBG(IPSEC_DATA, kdebug_secasv(sav));
1063 	return (sav);
1064 }
1065 
1066 /*
1067  * Must be called after calling key_allocsp().
1068  */
1069 void
1070 key_freesp(struct secpolicy **spp)
1071 {
1072 	struct secpolicy *sp = *spp;
1073 
1074 	IPSEC_ASSERT(sp != NULL, ("null sp"));
1075 	if (SP_DELREF(sp) == 0)
1076 		return;
1077 
1078 	KEYDBG(IPSEC_STAMP,
1079 	    printf("%s: last reference to SP(%p)\n", __func__, sp));
1080 	KEYDBG(IPSEC_DATA, kdebug_secpolicy(sp));
1081 
1082 	*spp = NULL;
1083 	while (sp->tcount > 0)
1084 		ipsec_delisr(sp->req[--sp->tcount]);
1085 	free(sp, M_IPSEC_SP);
1086 }
1087 
1088 static void
1089 key_unlink(struct secpolicy *sp)
1090 {
1091 
1092 	IPSEC_ASSERT(sp->spidx.dir == IPSEC_DIR_INBOUND ||
1093 	    sp->spidx.dir == IPSEC_DIR_OUTBOUND,
1094 	    ("invalid direction %u", sp->spidx.dir));
1095 	SPTREE_UNLOCK_ASSERT();
1096 
1097 	KEYDBG(KEY_STAMP,
1098 	    printf("%s: SP(%p)\n", __func__, sp));
1099 	SPTREE_WLOCK();
1100 	if (sp->state != IPSEC_SPSTATE_ALIVE) {
1101 		/* SP is already unlinked */
1102 		SPTREE_WUNLOCK();
1103 		return;
1104 	}
1105 	sp->state = IPSEC_SPSTATE_DEAD;
1106 	TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
1107 	LIST_REMOVE(sp, idhash);
1108 	V_sp_genid++;
1109 	SPTREE_WUNLOCK();
1110 	key_freesp(&sp);
1111 }
1112 
1113 /*
1114  * insert a secpolicy into the SP database. Lower priorities first
1115  */
1116 static void
1117 key_insertsp(struct secpolicy *newsp)
1118 {
1119 	struct secpolicy *sp;
1120 
1121 	SPTREE_WLOCK_ASSERT();
1122 	TAILQ_FOREACH(sp, &V_sptree[newsp->spidx.dir], chain) {
1123 		if (newsp->priority < sp->priority) {
1124 			TAILQ_INSERT_BEFORE(sp, newsp, chain);
1125 			goto done;
1126 		}
1127 	}
1128 	TAILQ_INSERT_TAIL(&V_sptree[newsp->spidx.dir], newsp, chain);
1129 done:
1130 	LIST_INSERT_HEAD(SPHASH_HASH(newsp->id), newsp, idhash);
1131 	newsp->state = IPSEC_SPSTATE_ALIVE;
1132 	V_sp_genid++;
1133 }
1134 
1135 /*
1136  * Insert a bunch of VTI secpolicies into the SPDB.
1137  * We keep VTI policies in the separate list due to following reasons:
1138  * 1) they should be immutable to user's or some deamon's attempts to
1139  *    delete. The only way delete such policies - destroy or unconfigure
1140  *    corresponding virtual inteface.
1141  * 2) such policies have traffic selector that matches all traffic per
1142  *    address family.
1143  * Since all VTI policies have the same priority, we don't care about
1144  * policies order.
1145  */
1146 int
1147 key_register_ifnet(struct secpolicy **spp, u_int count)
1148 {
1149 	struct mbuf *m;
1150 	u_int i;
1151 
1152 	SPTREE_WLOCK();
1153 	/*
1154 	 * First of try to acquire id for each SP.
1155 	 */
1156 	for (i = 0; i < count; i++) {
1157 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1158 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1159 		    ("invalid direction %u", spp[i]->spidx.dir));
1160 
1161 		if ((spp[i]->id = key_getnewspid()) == 0) {
1162 			SPTREE_WUNLOCK();
1163 			return (EAGAIN);
1164 		}
1165 	}
1166 	for (i = 0; i < count; i++) {
1167 		TAILQ_INSERT_TAIL(&V_sptree_ifnet[spp[i]->spidx.dir],
1168 		    spp[i], chain);
1169 		/*
1170 		 * NOTE: despite the fact that we keep VTI SP in the
1171 		 * separate list, SPHASH contains policies from both
1172 		 * sources. Thus SADB_X_SPDGET will correctly return
1173 		 * SP by id, because it uses SPHASH for lookups.
1174 		 */
1175 		LIST_INSERT_HEAD(SPHASH_HASH(spp[i]->id), spp[i], idhash);
1176 		spp[i]->state = IPSEC_SPSTATE_IFNET;
1177 	}
1178 	SPTREE_WUNLOCK();
1179 	/*
1180 	 * Notify user processes about new SP.
1181 	 */
1182 	for (i = 0; i < count; i++) {
1183 		m = key_setdumpsp(spp[i], SADB_X_SPDADD, 0, 0);
1184 		if (m != NULL)
1185 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1186 	}
1187 	return (0);
1188 }
1189 
1190 void
1191 key_unregister_ifnet(struct secpolicy **spp, u_int count)
1192 {
1193 	struct mbuf *m;
1194 	u_int i;
1195 
1196 	SPTREE_WLOCK();
1197 	for (i = 0; i < count; i++) {
1198 		IPSEC_ASSERT(spp[i]->spidx.dir == IPSEC_DIR_INBOUND ||
1199 		    spp[i]->spidx.dir == IPSEC_DIR_OUTBOUND,
1200 		    ("invalid direction %u", spp[i]->spidx.dir));
1201 
1202 		if (spp[i]->state != IPSEC_SPSTATE_IFNET)
1203 			continue;
1204 		spp[i]->state = IPSEC_SPSTATE_DEAD;
1205 		TAILQ_REMOVE(&V_sptree_ifnet[spp[i]->spidx.dir],
1206 		    spp[i], chain);
1207 		LIST_REMOVE(spp[i], idhash);
1208 	}
1209 	SPTREE_WUNLOCK();
1210 
1211 	for (i = 0; i < count; i++) {
1212 		m = key_setdumpsp(spp[i], SADB_X_SPDDELETE, 0, 0);
1213 		if (m != NULL)
1214 			key_sendup_mbuf(NULL, m, KEY_SENDUP_ALL);
1215 	}
1216 }
1217 
1218 /*
1219  * Must be called after calling key_allocsa().
1220  * This function is called by key_freesp() to free some SA allocated
1221  * for a policy.
1222  */
1223 void
1224 key_freesav(struct secasvar **psav)
1225 {
1226 	struct secasvar *sav = *psav;
1227 
1228 	IPSEC_ASSERT(sav != NULL, ("null sav"));
1229 	if (SAV_DELREF(sav) == 0)
1230 		return;
1231 
1232 	KEYDBG(IPSEC_STAMP,
1233 	    printf("%s: last reference to SA(%p)\n", __func__, sav));
1234 
1235 	*psav = NULL;
1236 	key_delsav(sav);
1237 }
1238 
1239 /*
1240  * Unlink SA from SAH and SPI hash under SAHTREE_WLOCK.
1241  * Expect that SA has extra reference due to lookup.
1242  * Release this references, also release SAH reference after unlink.
1243  */
1244 static void
1245 key_unlinksav(struct secasvar *sav)
1246 {
1247 	struct secashead *sah;
1248 
1249 	KEYDBG(KEY_STAMP,
1250 	    printf("%s: SA(%p)\n", __func__, sav));
1251 
1252 	SAHTREE_UNLOCK_ASSERT();
1253 	SAHTREE_WLOCK();
1254 	if (sav->state == SADB_SASTATE_DEAD) {
1255 		/* SA is already unlinked */
1256 		SAHTREE_WUNLOCK();
1257 		return;
1258 	}
1259 	/* Unlink from SAH */
1260 	if (sav->state == SADB_SASTATE_LARVAL)
1261 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
1262 	else
1263 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
1264 	/* Unlink from SPI hash */
1265 	LIST_REMOVE(sav, spihash);
1266 	sav->state = SADB_SASTATE_DEAD;
1267 	sah = sav->sah;
1268 	SAHTREE_WUNLOCK();
1269 	key_freesav(&sav);
1270 	/* Since we are unlinked, release reference to SAH */
1271 	key_freesah(&sah);
1272 }
1273 
1274 /* %%% SPD management */
1275 /*
1276  * search SPD
1277  * OUT:	NULL	: not found
1278  *	others	: found, pointer to a SP.
1279  */
1280 static struct secpolicy *
1281 key_getsp(struct secpolicyindex *spidx)
1282 {
1283 	SPTREE_RLOCK_TRACKER;
1284 	struct secpolicy *sp;
1285 
1286 	IPSEC_ASSERT(spidx != NULL, ("null spidx"));
1287 
1288 	SPTREE_RLOCK();
1289 	TAILQ_FOREACH(sp, &V_sptree[spidx->dir], chain) {
1290 		if (key_cmpspidx_exactly(spidx, &sp->spidx)) {
1291 			SP_ADDREF(sp);
1292 			break;
1293 		}
1294 	}
1295 	SPTREE_RUNLOCK();
1296 
1297 	return sp;
1298 }
1299 
1300 /*
1301  * get SP by index.
1302  * OUT:	NULL	: not found
1303  *	others	: found, pointer to referenced SP.
1304  */
1305 static struct secpolicy *
1306 key_getspbyid(uint32_t id)
1307 {
1308 	SPTREE_RLOCK_TRACKER;
1309 	struct secpolicy *sp;
1310 
1311 	SPTREE_RLOCK();
1312 	LIST_FOREACH(sp, SPHASH_HASH(id), idhash) {
1313 		if (sp->id == id) {
1314 			SP_ADDREF(sp);
1315 			break;
1316 		}
1317 	}
1318 	SPTREE_RUNLOCK();
1319 	return (sp);
1320 }
1321 
1322 struct secpolicy *
1323 key_newsp(void)
1324 {
1325 	struct secpolicy *sp;
1326 
1327 	sp = malloc(sizeof(*sp), M_IPSEC_SP, M_NOWAIT | M_ZERO);
1328 	if (sp != NULL)
1329 		SP_INITREF(sp);
1330 	return (sp);
1331 }
1332 
1333 struct ipsecrequest *
1334 ipsec_newisr(void)
1335 {
1336 
1337 	return (malloc(sizeof(struct ipsecrequest), M_IPSEC_SR,
1338 	    M_NOWAIT | M_ZERO));
1339 }
1340 
1341 void
1342 ipsec_delisr(struct ipsecrequest *p)
1343 {
1344 
1345 	free(p, M_IPSEC_SR);
1346 }
1347 
1348 /*
1349  * create secpolicy structure from sadb_x_policy structure.
1350  * NOTE: `state', `secpolicyindex' and 'id' in secpolicy structure
1351  * are not set, so must be set properly later.
1352  */
1353 struct secpolicy *
1354 key_msg2sp(struct sadb_x_policy *xpl0, size_t len, int *error)
1355 {
1356 	struct secpolicy *newsp;
1357 
1358 	IPSEC_ASSERT(xpl0 != NULL, ("null xpl0"));
1359 	IPSEC_ASSERT(len >= sizeof(*xpl0), ("policy too short: %zu", len));
1360 
1361 	if (len != PFKEY_EXTLEN(xpl0)) {
1362 		ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n", __func__));
1363 		*error = EINVAL;
1364 		return NULL;
1365 	}
1366 
1367 	if ((newsp = key_newsp()) == NULL) {
1368 		*error = ENOBUFS;
1369 		return NULL;
1370 	}
1371 
1372 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1373 	newsp->policy = xpl0->sadb_x_policy_type;
1374 	newsp->priority = xpl0->sadb_x_policy_priority;
1375 	newsp->tcount = 0;
1376 
1377 	/* check policy */
1378 	switch (xpl0->sadb_x_policy_type) {
1379 	case IPSEC_POLICY_DISCARD:
1380 	case IPSEC_POLICY_NONE:
1381 	case IPSEC_POLICY_ENTRUST:
1382 	case IPSEC_POLICY_BYPASS:
1383 		break;
1384 
1385 	case IPSEC_POLICY_IPSEC:
1386 	    {
1387 		struct sadb_x_ipsecrequest *xisr;
1388 		struct ipsecrequest *isr;
1389 		int tlen;
1390 
1391 		/* validity check */
1392 		if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1393 			ipseclog((LOG_DEBUG, "%s: Invalid msg length.\n",
1394 				__func__));
1395 			key_freesp(&newsp);
1396 			*error = EINVAL;
1397 			return NULL;
1398 		}
1399 
1400 		tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1401 		xisr = (struct sadb_x_ipsecrequest *)(xpl0 + 1);
1402 
1403 		while (tlen > 0) {
1404 			/* length check */
1405 			if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr)) {
1406 				ipseclog((LOG_DEBUG, "%s: invalid ipsecrequest "
1407 					"length.\n", __func__));
1408 				key_freesp(&newsp);
1409 				*error = EINVAL;
1410 				return NULL;
1411 			}
1412 
1413 			if (newsp->tcount >= IPSEC_MAXREQ) {
1414 				ipseclog((LOG_DEBUG,
1415 				    "%s: too many ipsecrequests.\n",
1416 				    __func__));
1417 				key_freesp(&newsp);
1418 				*error = EINVAL;
1419 				return (NULL);
1420 			}
1421 
1422 			/* allocate request buffer */
1423 			/* NB: data structure is zero'd */
1424 			isr = ipsec_newisr();
1425 			if (isr == NULL) {
1426 				ipseclog((LOG_DEBUG,
1427 				    "%s: No more memory.\n", __func__));
1428 				key_freesp(&newsp);
1429 				*error = ENOBUFS;
1430 				return NULL;
1431 			}
1432 
1433 			newsp->req[newsp->tcount++] = isr;
1434 
1435 			/* set values */
1436 			switch (xisr->sadb_x_ipsecrequest_proto) {
1437 			case IPPROTO_ESP:
1438 			case IPPROTO_AH:
1439 			case IPPROTO_IPCOMP:
1440 				break;
1441 			default:
1442 				ipseclog((LOG_DEBUG,
1443 				    "%s: invalid proto type=%u\n", __func__,
1444 				    xisr->sadb_x_ipsecrequest_proto));
1445 				key_freesp(&newsp);
1446 				*error = EPROTONOSUPPORT;
1447 				return NULL;
1448 			}
1449 			isr->saidx.proto =
1450 			    (uint8_t)xisr->sadb_x_ipsecrequest_proto;
1451 
1452 			switch (xisr->sadb_x_ipsecrequest_mode) {
1453 			case IPSEC_MODE_TRANSPORT:
1454 			case IPSEC_MODE_TUNNEL:
1455 				break;
1456 			case IPSEC_MODE_ANY:
1457 			default:
1458 				ipseclog((LOG_DEBUG,
1459 				    "%s: invalid mode=%u\n", __func__,
1460 				    xisr->sadb_x_ipsecrequest_mode));
1461 				key_freesp(&newsp);
1462 				*error = EINVAL;
1463 				return NULL;
1464 			}
1465 			isr->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1466 
1467 			switch (xisr->sadb_x_ipsecrequest_level) {
1468 			case IPSEC_LEVEL_DEFAULT:
1469 			case IPSEC_LEVEL_USE:
1470 			case IPSEC_LEVEL_REQUIRE:
1471 				break;
1472 			case IPSEC_LEVEL_UNIQUE:
1473 				/* validity check */
1474 				/*
1475 				 * If range violation of reqid, kernel will
1476 				 * update it, don't refuse it.
1477 				 */
1478 				if (xisr->sadb_x_ipsecrequest_reqid
1479 						> IPSEC_MANUAL_REQID_MAX) {
1480 					ipseclog((LOG_DEBUG,
1481 					    "%s: reqid=%d range "
1482 					    "violation, updated by kernel.\n",
1483 					    __func__,
1484 					    xisr->sadb_x_ipsecrequest_reqid));
1485 					xisr->sadb_x_ipsecrequest_reqid = 0;
1486 				}
1487 
1488 				/* allocate new reqid id if reqid is zero. */
1489 				if (xisr->sadb_x_ipsecrequest_reqid == 0) {
1490 					u_int32_t reqid;
1491 					if ((reqid = key_newreqid()) == 0) {
1492 						key_freesp(&newsp);
1493 						*error = ENOBUFS;
1494 						return NULL;
1495 					}
1496 					isr->saidx.reqid = reqid;
1497 					xisr->sadb_x_ipsecrequest_reqid = reqid;
1498 				} else {
1499 				/* set it for manual keying. */
1500 					isr->saidx.reqid =
1501 					    xisr->sadb_x_ipsecrequest_reqid;
1502 				}
1503 				break;
1504 
1505 			default:
1506 				ipseclog((LOG_DEBUG, "%s: invalid level=%u\n",
1507 					__func__,
1508 					xisr->sadb_x_ipsecrequest_level));
1509 				key_freesp(&newsp);
1510 				*error = EINVAL;
1511 				return NULL;
1512 			}
1513 			isr->level = xisr->sadb_x_ipsecrequest_level;
1514 
1515 			/* set IP addresses if there */
1516 			if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1517 				struct sockaddr *paddr;
1518 
1519 				paddr = (struct sockaddr *)(xisr + 1);
1520 				/* validity check */
1521 				if (paddr->sa_len
1522 				    > sizeof(isr->saidx.src)) {
1523 					ipseclog((LOG_DEBUG, "%s: invalid "
1524 						"request address length.\n",
1525 						__func__));
1526 					key_freesp(&newsp);
1527 					*error = EINVAL;
1528 					return NULL;
1529 				}
1530 				bcopy(paddr, &isr->saidx.src, paddr->sa_len);
1531 				paddr = (struct sockaddr *)((caddr_t)paddr +
1532 				    paddr->sa_len);
1533 
1534 				/* validity check */
1535 				if (paddr->sa_len
1536 				    > sizeof(isr->saidx.dst)) {
1537 					ipseclog((LOG_DEBUG, "%s: invalid "
1538 						"request address length.\n",
1539 						__func__));
1540 					key_freesp(&newsp);
1541 					*error = EINVAL;
1542 					return NULL;
1543 				}
1544 				/* AF family should match */
1545 				if (paddr->sa_family !=
1546 				    isr->saidx.src.sa.sa_family) {
1547 					ipseclog((LOG_DEBUG, "%s: address "
1548 					    "family doesn't match.\n",
1549 						__func__));
1550 					key_freesp(&newsp);
1551 					*error = EINVAL;
1552 					return (NULL);
1553 				}
1554 				bcopy(paddr, &isr->saidx.dst, paddr->sa_len);
1555 			} else {
1556 				/*
1557 				 * Addresses for TUNNEL mode requests are
1558 				 * mandatory.
1559 				 */
1560 				if (isr->saidx.mode == IPSEC_MODE_TUNNEL) {
1561 					ipseclog((LOG_DEBUG, "%s: missing "
1562 					    "request addresses.\n", __func__));
1563 					key_freesp(&newsp);
1564 					*error = EINVAL;
1565 					return (NULL);
1566 				}
1567 			}
1568 			tlen -= xisr->sadb_x_ipsecrequest_len;
1569 
1570 			/* validity check */
1571 			if (tlen < 0) {
1572 				ipseclog((LOG_DEBUG, "%s: becoming tlen < 0.\n",
1573 					__func__));
1574 				key_freesp(&newsp);
1575 				*error = EINVAL;
1576 				return NULL;
1577 			}
1578 
1579 			xisr = (struct sadb_x_ipsecrequest *)((caddr_t)xisr
1580 			                 + xisr->sadb_x_ipsecrequest_len);
1581 		}
1582 		/* XXXAE: LARVAL SP */
1583 		if (newsp->tcount < 1) {
1584 			ipseclog((LOG_DEBUG, "%s: valid IPSEC transforms "
1585 			    "not found.\n", __func__));
1586 			key_freesp(&newsp);
1587 			*error = EINVAL;
1588 			return (NULL);
1589 		}
1590 	    }
1591 		break;
1592 	default:
1593 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
1594 		key_freesp(&newsp);
1595 		*error = EINVAL;
1596 		return NULL;
1597 	}
1598 
1599 	*error = 0;
1600 	return (newsp);
1601 }
1602 
1603 uint32_t
1604 key_newreqid(void)
1605 {
1606 	static uint32_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1607 
1608 	if (auto_reqid == ~0)
1609 		auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
1610 	else
1611 		auto_reqid++;
1612 
1613 	/* XXX should be unique check */
1614 	return (auto_reqid);
1615 }
1616 
1617 /*
1618  * copy secpolicy struct to sadb_x_policy structure indicated.
1619  */
1620 static struct mbuf *
1621 key_sp2mbuf(struct secpolicy *sp)
1622 {
1623 	struct mbuf *m;
1624 	size_t tlen;
1625 
1626 	tlen = key_getspreqmsglen(sp);
1627 	m = m_get2(tlen, M_NOWAIT, MT_DATA, 0);
1628 	if (m == NULL)
1629 		return (NULL);
1630 	m_align(m, tlen);
1631 	m->m_len = tlen;
1632 	if (key_sp2msg(sp, m->m_data, &tlen) != 0) {
1633 		m_freem(m);
1634 		return (NULL);
1635 	}
1636 	return (m);
1637 }
1638 
1639 int
1640 key_sp2msg(struct secpolicy *sp, void *request, size_t *len)
1641 {
1642 	struct sadb_x_ipsecrequest *xisr;
1643 	struct sadb_x_policy *xpl;
1644 	struct ipsecrequest *isr;
1645 	size_t xlen, ilen;
1646 	caddr_t p;
1647 	int error, i;
1648 
1649 	IPSEC_ASSERT(sp != NULL, ("null policy"));
1650 
1651 	xlen = sizeof(*xpl);
1652 	if (*len < xlen)
1653 		return (EINVAL);
1654 
1655 	error = 0;
1656 	bzero(request, *len);
1657 	xpl = (struct sadb_x_policy *)request;
1658 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
1659 	xpl->sadb_x_policy_type = sp->policy;
1660 	xpl->sadb_x_policy_dir = sp->spidx.dir;
1661 	xpl->sadb_x_policy_id = sp->id;
1662 	xpl->sadb_x_policy_priority = sp->priority;
1663 
1664 	/* if is the policy for ipsec ? */
1665 	if (sp->policy == IPSEC_POLICY_IPSEC) {
1666 		p = (caddr_t)xpl + sizeof(*xpl);
1667 		for (i = 0; i < sp->tcount; i++) {
1668 			isr = sp->req[i];
1669 			ilen = PFKEY_ALIGN8(sizeof(*xisr) +
1670 			    isr->saidx.src.sa.sa_len +
1671 			    isr->saidx.dst.sa.sa_len);
1672 			xlen += ilen;
1673 			if (xlen > *len) {
1674 				error = ENOBUFS;
1675 				/* Calculate needed size */
1676 				continue;
1677 			}
1678 			xisr = (struct sadb_x_ipsecrequest *)p;
1679 			xisr->sadb_x_ipsecrequest_len = ilen;
1680 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
1681 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
1682 			xisr->sadb_x_ipsecrequest_level = isr->level;
1683 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
1684 
1685 			p += sizeof(*xisr);
1686 			bcopy(&isr->saidx.src, p, isr->saidx.src.sa.sa_len);
1687 			p += isr->saidx.src.sa.sa_len;
1688 			bcopy(&isr->saidx.dst, p, isr->saidx.dst.sa.sa_len);
1689 			p += isr->saidx.dst.sa.sa_len;
1690 		}
1691 	}
1692 	xpl->sadb_x_policy_len = PFKEY_UNIT64(xlen);
1693 	if (error == 0)
1694 		*len = xlen;
1695 	else
1696 		*len = sizeof(*xpl);
1697 	return (error);
1698 }
1699 
1700 /* m will not be freed nor modified */
1701 static struct mbuf *
1702 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
1703     int ndeep, int nitem, ...)
1704 {
1705 	va_list ap;
1706 	int idx;
1707 	int i;
1708 	struct mbuf *result = NULL, *n;
1709 	int len;
1710 
1711 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1712 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1713 
1714 	va_start(ap, nitem);
1715 	for (i = 0; i < nitem; i++) {
1716 		idx = va_arg(ap, int);
1717 		if (idx < 0 || idx > SADB_EXT_MAX)
1718 			goto fail;
1719 		/* don't attempt to pull empty extension */
1720 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
1721 			continue;
1722 		if (idx != SADB_EXT_RESERVED  &&
1723 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
1724 			continue;
1725 
1726 		if (idx == SADB_EXT_RESERVED) {
1727 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
1728 
1729 			IPSEC_ASSERT(len <= MHLEN, ("header too big %u", len));
1730 
1731 			MGETHDR(n, M_NOWAIT, MT_DATA);
1732 			if (!n)
1733 				goto fail;
1734 			n->m_len = len;
1735 			n->m_next = NULL;
1736 			m_copydata(m, 0, sizeof(struct sadb_msg),
1737 			    mtod(n, caddr_t));
1738 		} else if (i < ndeep) {
1739 			len = mhp->extlen[idx];
1740 			n = m_get2(len, M_NOWAIT, MT_DATA, 0);
1741 			if (n == NULL)
1742 				goto fail;
1743 			m_align(n, len);
1744 			n->m_len = len;
1745 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
1746 			    mtod(n, caddr_t));
1747 		} else {
1748 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
1749 			    M_NOWAIT);
1750 		}
1751 		if (n == NULL)
1752 			goto fail;
1753 
1754 		if (result)
1755 			m_cat(result, n);
1756 		else
1757 			result = n;
1758 	}
1759 	va_end(ap);
1760 
1761 	if ((result->m_flags & M_PKTHDR) != 0) {
1762 		result->m_pkthdr.len = 0;
1763 		for (n = result; n; n = n->m_next)
1764 			result->m_pkthdr.len += n->m_len;
1765 	}
1766 
1767 	return result;
1768 
1769 fail:
1770 	m_freem(result);
1771 	va_end(ap);
1772 	return NULL;
1773 }
1774 
1775 /*
1776  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
1777  * add an entry to SP database, when received
1778  *   <base, address(SD), (lifetime(H),) policy>
1779  * from the user(?).
1780  * Adding to SP database,
1781  * and send
1782  *   <base, address(SD), (lifetime(H),) policy>
1783  * to the socket which was send.
1784  *
1785  * SPDADD set a unique policy entry.
1786  * SPDSETIDX like SPDADD without a part of policy requests.
1787  * SPDUPDATE replace a unique policy entry.
1788  *
1789  * XXXAE: serialize this in PF_KEY to avoid races.
1790  * m will always be freed.
1791  */
1792 static int
1793 key_spdadd(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
1794 {
1795 	struct secpolicyindex spidx;
1796 	struct sadb_address *src0, *dst0;
1797 	struct sadb_x_policy *xpl0, *xpl;
1798 	struct sadb_lifetime *lft = NULL;
1799 	struct secpolicy *newsp;
1800 	int error;
1801 
1802 	IPSEC_ASSERT(so != NULL, ("null socket"));
1803 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
1804 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
1805 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
1806 
1807 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
1808 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
1809 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
1810 		ipseclog((LOG_DEBUG,
1811 		    "%s: invalid message: missing required header.\n",
1812 		    __func__));
1813 		return key_senderror(so, m, EINVAL);
1814 	}
1815 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
1816 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
1817 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
1818 		ipseclog((LOG_DEBUG,
1819 		    "%s: invalid message: wrong header size.\n", __func__));
1820 		return key_senderror(so, m, EINVAL);
1821 	}
1822 	if (!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD)) {
1823 		if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD)) {
1824 			ipseclog((LOG_DEBUG,
1825 			    "%s: invalid message: wrong header size.\n",
1826 			    __func__));
1827 			return key_senderror(so, m, EINVAL);
1828 		}
1829 		lft = (struct sadb_lifetime *)mhp->ext[SADB_EXT_LIFETIME_HARD];
1830 	}
1831 
1832 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
1833 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
1834 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
1835 
1836 	/* check the direciton */
1837 	switch (xpl0->sadb_x_policy_dir) {
1838 	case IPSEC_DIR_INBOUND:
1839 	case IPSEC_DIR_OUTBOUND:
1840 		break;
1841 	default:
1842 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
1843 		return key_senderror(so, m, EINVAL);
1844 	}
1845 	/* key_spdadd() accepts DISCARD, NONE and IPSEC. */
1846 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
1847 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
1848 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
1849 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
1850 		return key_senderror(so, m, EINVAL);
1851 	}
1852 
1853 	/* policy requests are mandatory when action is ipsec. */
1854 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
1855 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
1856 		ipseclog((LOG_DEBUG,
1857 		    "%s: policy requests required.\n", __func__));
1858 		return key_senderror(so, m, EINVAL);
1859 	}
1860 
1861 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
1862 	    (struct sockaddr *)(dst0 + 1));
1863 	if (error != 0 ||
1864 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
1865 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
1866 		return key_senderror(so, m, error);
1867 	}
1868 	/* make secindex */
1869 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
1870 	                src0 + 1,
1871 	                dst0 + 1,
1872 	                src0->sadb_address_prefixlen,
1873 	                dst0->sadb_address_prefixlen,
1874 	                src0->sadb_address_proto,
1875 	                &spidx);
1876 	/* Checking there is SP already or not. */
1877 	newsp = key_getsp(&spidx);
1878 	if (newsp != NULL) {
1879 		if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
1880 			KEYDBG(KEY_STAMP,
1881 			    printf("%s: unlink SP(%p) for SPDUPDATE\n",
1882 				__func__, newsp));
1883 			KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
1884 			key_unlink(newsp);
1885 			key_freesp(&newsp);
1886 		} else {
1887 			key_freesp(&newsp);
1888 			ipseclog((LOG_DEBUG, "%s: a SP entry exists already.",
1889 			    __func__));
1890 			return (key_senderror(so, m, EEXIST));
1891 		}
1892 	}
1893 
1894 	/* allocate new SP entry */
1895 	if ((newsp = key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error)) == NULL) {
1896 		return key_senderror(so, m, error);
1897 	}
1898 
1899 	newsp->lastused = newsp->created = time_second;
1900 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
1901 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
1902 	bcopy(&spidx, &newsp->spidx, sizeof(spidx));
1903 
1904 	/* XXXAE: there is race between key_getsp() and key_insertsp() */
1905 	SPTREE_WLOCK();
1906 	if ((newsp->id = key_getnewspid()) == 0) {
1907 		SPTREE_WUNLOCK();
1908 		key_freesp(&newsp);
1909 		return key_senderror(so, m, ENOBUFS);
1910 	}
1911 	key_insertsp(newsp);
1912 	SPTREE_WUNLOCK();
1913 
1914 	KEYDBG(KEY_STAMP,
1915 	    printf("%s: SP(%p)\n", __func__, newsp));
1916 	KEYDBG(KEY_DATA, kdebug_secpolicy(newsp));
1917 
1918     {
1919 	struct mbuf *n, *mpolicy;
1920 	struct sadb_msg *newmsg;
1921 	int off;
1922 
1923 	/* create new sadb_msg to reply. */
1924 	if (lft) {
1925 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
1926 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
1927 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
1928 	} else {
1929 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
1930 		    SADB_X_EXT_POLICY,
1931 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
1932 	}
1933 	if (!n)
1934 		return key_senderror(so, m, ENOBUFS);
1935 
1936 	if (n->m_len < sizeof(*newmsg)) {
1937 		n = m_pullup(n, sizeof(*newmsg));
1938 		if (!n)
1939 			return key_senderror(so, m, ENOBUFS);
1940 	}
1941 	newmsg = mtod(n, struct sadb_msg *);
1942 	newmsg->sadb_msg_errno = 0;
1943 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
1944 
1945 	off = 0;
1946 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
1947 	    sizeof(*xpl), &off);
1948 	if (mpolicy == NULL) {
1949 		/* n is already freed */
1950 		return key_senderror(so, m, ENOBUFS);
1951 	}
1952 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, caddr_t) + off);
1953 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
1954 		m_freem(n);
1955 		return key_senderror(so, m, EINVAL);
1956 	}
1957 	xpl->sadb_x_policy_id = newsp->id;
1958 
1959 	m_freem(m);
1960 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
1961     }
1962 }
1963 
1964 /*
1965  * get new policy id.
1966  * OUT:
1967  *	0:	failure.
1968  *	others: success.
1969  */
1970 static uint32_t
1971 key_getnewspid(void)
1972 {
1973 	struct secpolicy *sp;
1974 	uint32_t newid = 0;
1975 	int count = V_key_spi_trycnt;	/* XXX */
1976 
1977 	SPTREE_WLOCK_ASSERT();
1978 	while (count--) {
1979 		if (V_policy_id == ~0) /* overflowed */
1980 			newid = V_policy_id = 1;
1981 		else
1982 			newid = ++V_policy_id;
1983 		LIST_FOREACH(sp, SPHASH_HASH(newid), idhash) {
1984 			if (sp->id == newid)
1985 				break;
1986 		}
1987 		if (sp == NULL)
1988 			break;
1989 	}
1990 	if (count == 0 || newid == 0) {
1991 		ipseclog((LOG_DEBUG, "%s: failed to allocate policy id.\n",
1992 		    __func__));
1993 		return (0);
1994 	}
1995 	return (newid);
1996 }
1997 
1998 /*
1999  * SADB_SPDDELETE processing
2000  * receive
2001  *   <base, address(SD), policy(*)>
2002  * from the user(?), and set SADB_SASTATE_DEAD,
2003  * and send,
2004  *   <base, address(SD), policy(*)>
2005  * to the ikmpd.
2006  * policy(*) including direction of policy.
2007  *
2008  * m will always be freed.
2009  */
2010 static int
2011 key_spddelete(struct socket *so, struct mbuf *m,
2012     const struct sadb_msghdr *mhp)
2013 {
2014 	struct secpolicyindex spidx;
2015 	struct sadb_address *src0, *dst0;
2016 	struct sadb_x_policy *xpl0;
2017 	struct secpolicy *sp;
2018 
2019 	IPSEC_ASSERT(so != NULL, ("null so"));
2020 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2021 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2022 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2023 
2024 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
2025 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
2026 	    SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY)) {
2027 		ipseclog((LOG_DEBUG,
2028 		    "%s: invalid message: missing required header.\n",
2029 		    __func__));
2030 		return key_senderror(so, m, EINVAL);
2031 	}
2032 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
2033 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
2034 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2035 		ipseclog((LOG_DEBUG,
2036 		    "%s: invalid message: wrong header size.\n", __func__));
2037 		return key_senderror(so, m, EINVAL);
2038 	}
2039 
2040 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
2041 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
2042 	xpl0 = (struct sadb_x_policy *)mhp->ext[SADB_X_EXT_POLICY];
2043 
2044 	/* check the direciton */
2045 	switch (xpl0->sadb_x_policy_dir) {
2046 	case IPSEC_DIR_INBOUND:
2047 	case IPSEC_DIR_OUTBOUND:
2048 		break;
2049 	default:
2050 		ipseclog((LOG_DEBUG, "%s: invalid SP direction.\n", __func__));
2051 		return key_senderror(so, m, EINVAL);
2052 	}
2053 	/* Only DISCARD, NONE and IPSEC are allowed */
2054 	if (xpl0->sadb_x_policy_type != IPSEC_POLICY_DISCARD &&
2055 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_NONE &&
2056 	    xpl0->sadb_x_policy_type != IPSEC_POLICY_IPSEC) {
2057 		ipseclog((LOG_DEBUG, "%s: invalid policy type.\n", __func__));
2058 		return key_senderror(so, m, EINVAL);
2059 	}
2060 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
2061 	    (struct sockaddr *)(dst0 + 1)) != 0 ||
2062 	    src0->sadb_address_proto != dst0->sadb_address_proto) {
2063 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
2064 		return key_senderror(so, m, EINVAL);
2065 	}
2066 	/* make secindex */
2067 	KEY_SETSECSPIDX(xpl0->sadb_x_policy_dir,
2068 	                src0 + 1,
2069 	                dst0 + 1,
2070 	                src0->sadb_address_prefixlen,
2071 	                dst0->sadb_address_prefixlen,
2072 	                src0->sadb_address_proto,
2073 	                &spidx);
2074 
2075 	/* Is there SP in SPD ? */
2076 	if ((sp = key_getsp(&spidx)) == NULL) {
2077 		ipseclog((LOG_DEBUG, "%s: no SP found.\n", __func__));
2078 		return key_senderror(so, m, EINVAL);
2079 	}
2080 
2081 	/* save policy id to buffer to be returned. */
2082 	xpl0->sadb_x_policy_id = sp->id;
2083 
2084 	KEYDBG(KEY_STAMP,
2085 	    printf("%s: SP(%p)\n", __func__, sp));
2086 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2087 	key_unlink(sp);
2088 	key_freesp(&sp);
2089 
2090     {
2091 	struct mbuf *n;
2092 	struct sadb_msg *newmsg;
2093 
2094 	/* create new sadb_msg to reply. */
2095 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2096 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2097 	if (!n)
2098 		return key_senderror(so, m, ENOBUFS);
2099 
2100 	newmsg = mtod(n, struct sadb_msg *);
2101 	newmsg->sadb_msg_errno = 0;
2102 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2103 
2104 	m_freem(m);
2105 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2106     }
2107 }
2108 
2109 /*
2110  * SADB_SPDDELETE2 processing
2111  * receive
2112  *   <base, policy(*)>
2113  * from the user(?), and set SADB_SASTATE_DEAD,
2114  * and send,
2115  *   <base, policy(*)>
2116  * to the ikmpd.
2117  * policy(*) including direction of policy.
2118  *
2119  * m will always be freed.
2120  */
2121 static int
2122 key_spddelete2(struct socket *so, struct mbuf *m,
2123     const struct sadb_msghdr *mhp)
2124 {
2125 	struct secpolicy *sp;
2126 	uint32_t id;
2127 
2128 	IPSEC_ASSERT(so != NULL, ("null socket"));
2129 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2130 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2131 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2132 
2133 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2134 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2135 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2136 		    __func__));
2137 		return key_senderror(so, m, EINVAL);
2138 	}
2139 
2140 	id = ((struct sadb_x_policy *)
2141 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2142 
2143 	/* Is there SP in SPD ? */
2144 	if ((sp = key_getspbyid(id)) == NULL) {
2145 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2146 		    __func__, id));
2147 		return key_senderror(so, m, EINVAL);
2148 	}
2149 
2150 	KEYDBG(KEY_STAMP,
2151 	    printf("%s: SP(%p)\n", __func__, sp));
2152 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2153 	key_unlink(sp);
2154 	if (sp->state != IPSEC_SPSTATE_DEAD) {
2155 		ipseclog((LOG_DEBUG, "%s: failed to delete SP with id %u.\n",
2156 		    __func__, id));
2157 		key_freesp(&sp);
2158 		return (key_senderror(so, m, EACCES));
2159 	}
2160 	key_freesp(&sp);
2161 
2162     {
2163 	struct mbuf *n, *nn;
2164 	struct sadb_msg *newmsg;
2165 	int off, len;
2166 
2167 	/* create new sadb_msg to reply. */
2168 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2169 
2170 	MGETHDR(n, M_NOWAIT, MT_DATA);
2171 	if (n && len > MHLEN) {
2172 		if (!(MCLGET(n, M_NOWAIT))) {
2173 			m_freem(n);
2174 			n = NULL;
2175 		}
2176 	}
2177 	if (!n)
2178 		return key_senderror(so, m, ENOBUFS);
2179 
2180 	n->m_len = len;
2181 	n->m_next = NULL;
2182 	off = 0;
2183 
2184 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
2185 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2186 
2187 	IPSEC_ASSERT(off == len, ("length inconsistency (off %u len %u)",
2188 		off, len));
2189 
2190 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2191 	    mhp->extlen[SADB_X_EXT_POLICY], M_NOWAIT);
2192 	if (!n->m_next) {
2193 		m_freem(n);
2194 		return key_senderror(so, m, ENOBUFS);
2195 	}
2196 
2197 	n->m_pkthdr.len = 0;
2198 	for (nn = n; nn; nn = nn->m_next)
2199 		n->m_pkthdr.len += nn->m_len;
2200 
2201 	newmsg = mtod(n, struct sadb_msg *);
2202 	newmsg->sadb_msg_errno = 0;
2203 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
2204 
2205 	m_freem(m);
2206 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2207     }
2208 }
2209 
2210 /*
2211  * SADB_X_SPDGET processing
2212  * receive
2213  *   <base, policy(*)>
2214  * from the user(?),
2215  * and send,
2216  *   <base, address(SD), policy>
2217  * to the ikmpd.
2218  * policy(*) including direction of policy.
2219  *
2220  * m will always be freed.
2221  */
2222 static int
2223 key_spdget(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2224 {
2225 	struct secpolicy *sp;
2226 	struct mbuf *n;
2227 	uint32_t id;
2228 
2229 	IPSEC_ASSERT(so != NULL, ("null socket"));
2230 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2231 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2232 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2233 
2234 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_POLICY) ||
2235 	    SADB_CHECKLEN(mhp, SADB_X_EXT_POLICY)) {
2236 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
2237 		    __func__));
2238 		return key_senderror(so, m, EINVAL);
2239 	}
2240 
2241 	id = ((struct sadb_x_policy *)
2242 	    mhp->ext[SADB_X_EXT_POLICY])->sadb_x_policy_id;
2243 
2244 	/* Is there SP in SPD ? */
2245 	if ((sp = key_getspbyid(id)) == NULL) {
2246 		ipseclog((LOG_DEBUG, "%s: no SP found for id %u.\n",
2247 		    __func__, id));
2248 		return key_senderror(so, m, ENOENT);
2249 	}
2250 
2251 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2252 	    mhp->msg->sadb_msg_pid);
2253 	key_freesp(&sp);
2254 	if (n != NULL) {
2255 		m_freem(m);
2256 		return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2257 	} else
2258 		return key_senderror(so, m, ENOBUFS);
2259 }
2260 
2261 /*
2262  * SADB_X_SPDACQUIRE processing.
2263  * Acquire policy and SA(s) for a *OUTBOUND* packet.
2264  * send
2265  *   <base, policy(*)>
2266  * to KMD, and expect to receive
2267  *   <base> with SADB_X_SPDACQUIRE if error occurred,
2268  * or
2269  *   <base, policy>
2270  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2271  * policy(*) is without policy requests.
2272  *
2273  *    0     : succeed
2274  *    others: error number
2275  */
2276 int
2277 key_spdacquire(struct secpolicy *sp)
2278 {
2279 	struct mbuf *result = NULL, *m;
2280 	struct secspacq *newspacq;
2281 
2282 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2283 	IPSEC_ASSERT(sp->req == NULL, ("policy exists"));
2284 	IPSEC_ASSERT(sp->policy == IPSEC_POLICY_IPSEC,
2285 		("policy not IPSEC %u", sp->policy));
2286 
2287 	/* Get an entry to check whether sent message or not. */
2288 	newspacq = key_getspacq(&sp->spidx);
2289 	if (newspacq != NULL) {
2290 		if (V_key_blockacq_count < newspacq->count) {
2291 			/* reset counter and do send message. */
2292 			newspacq->count = 0;
2293 		} else {
2294 			/* increment counter and do nothing. */
2295 			newspacq->count++;
2296 			SPACQ_UNLOCK();
2297 			return (0);
2298 		}
2299 		SPACQ_UNLOCK();
2300 	} else {
2301 		/* make new entry for blocking to send SADB_ACQUIRE. */
2302 		newspacq = key_newspacq(&sp->spidx);
2303 		if (newspacq == NULL)
2304 			return ENOBUFS;
2305 	}
2306 
2307 	/* create new sadb_msg to reply. */
2308 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2309 	if (!m)
2310 		return ENOBUFS;
2311 
2312 	result = m;
2313 
2314 	result->m_pkthdr.len = 0;
2315 	for (m = result; m; m = m->m_next)
2316 		result->m_pkthdr.len += m->m_len;
2317 
2318 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2319 	    PFKEY_UNIT64(result->m_pkthdr.len);
2320 
2321 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2322 }
2323 
2324 /*
2325  * SADB_SPDFLUSH processing
2326  * receive
2327  *   <base>
2328  * from the user, and free all entries in secpctree.
2329  * and send,
2330  *   <base>
2331  * to the user.
2332  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2333  *
2334  * m will always be freed.
2335  */
2336 static int
2337 key_spdflush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2338 {
2339 	struct secpolicy_queue drainq;
2340 	struct sadb_msg *newmsg;
2341 	struct secpolicy *sp, *nextsp;
2342 	u_int dir;
2343 
2344 	IPSEC_ASSERT(so != NULL, ("null socket"));
2345 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2346 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2347 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2348 
2349 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2350 		return key_senderror(so, m, EINVAL);
2351 
2352 	TAILQ_INIT(&drainq);
2353 	SPTREE_WLOCK();
2354 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2355 		TAILQ_CONCAT(&drainq, &V_sptree[dir], chain);
2356 	}
2357 	/*
2358 	 * We need to set state to DEAD for each policy to be sure,
2359 	 * that another thread won't try to unlink it.
2360 	 * Also remove SP from sphash.
2361 	 */
2362 	TAILQ_FOREACH(sp, &drainq, chain) {
2363 		sp->state = IPSEC_SPSTATE_DEAD;
2364 		LIST_REMOVE(sp, idhash);
2365 	}
2366 	V_sp_genid++;
2367 	SPTREE_WUNLOCK();
2368 	sp = TAILQ_FIRST(&drainq);
2369 	while (sp != NULL) {
2370 		nextsp = TAILQ_NEXT(sp, chain);
2371 		key_freesp(&sp);
2372 		sp = nextsp;
2373 	}
2374 
2375 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2376 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
2377 		return key_senderror(so, m, ENOBUFS);
2378 	}
2379 
2380 	if (m->m_next)
2381 		m_freem(m->m_next);
2382 	m->m_next = NULL;
2383 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2384 	newmsg = mtod(m, struct sadb_msg *);
2385 	newmsg->sadb_msg_errno = 0;
2386 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2387 
2388 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2389 }
2390 
2391 /*
2392  * SADB_SPDDUMP processing
2393  * receive
2394  *   <base>
2395  * from the user, and dump all SP leaves
2396  * and send,
2397  *   <base> .....
2398  * to the ikmpd.
2399  *
2400  * m will always be freed.
2401  */
2402 static int
2403 key_spddump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
2404 {
2405 	SPTREE_RLOCK_TRACKER;
2406 	struct secpolicy *sp;
2407 	struct mbuf *n;
2408 	int cnt;
2409 	u_int dir;
2410 
2411 	IPSEC_ASSERT(so != NULL, ("null socket"));
2412 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
2413 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2414 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2415 
2416 	/* search SPD entry and get buffer size. */
2417 	cnt = 0;
2418 	SPTREE_RLOCK();
2419 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2420 		TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
2421 			cnt++;
2422 		}
2423 		TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain) {
2424 			cnt++;
2425 		}
2426 	}
2427 
2428 	if (cnt == 0) {
2429 		SPTREE_RUNLOCK();
2430 		return key_senderror(so, m, ENOENT);
2431 	}
2432 
2433 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2434 		TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
2435 			--cnt;
2436 			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2437 			    mhp->msg->sadb_msg_pid);
2438 
2439 			if (n)
2440 				key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2441 		}
2442 		TAILQ_FOREACH(sp, &V_sptree_ifnet[dir], chain) {
2443 			--cnt;
2444 			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt,
2445 			    mhp->msg->sadb_msg_pid);
2446 
2447 			if (n)
2448 				key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2449 		}
2450 	}
2451 
2452 	SPTREE_RUNLOCK();
2453 	m_freem(m);
2454 	return (0);
2455 }
2456 
2457 static struct mbuf *
2458 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq,
2459     u_int32_t pid)
2460 {
2461 	struct mbuf *result = NULL, *m;
2462 	struct seclifetime lt;
2463 
2464 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid, sp->refcnt);
2465 	if (!m)
2466 		goto fail;
2467 	result = m;
2468 
2469 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2470 	    &sp->spidx.src.sa, sp->spidx.prefs,
2471 	    sp->spidx.ul_proto);
2472 	if (!m)
2473 		goto fail;
2474 	m_cat(result, m);
2475 
2476 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2477 	    &sp->spidx.dst.sa, sp->spidx.prefd,
2478 	    sp->spidx.ul_proto);
2479 	if (!m)
2480 		goto fail;
2481 	m_cat(result, m);
2482 
2483 	m = key_sp2mbuf(sp);
2484 	if (!m)
2485 		goto fail;
2486 	m_cat(result, m);
2487 
2488 	if(sp->lifetime){
2489 		lt.addtime=sp->created;
2490 		lt.usetime= sp->lastused;
2491 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_CURRENT);
2492 		if (!m)
2493 			goto fail;
2494 		m_cat(result, m);
2495 
2496 		lt.addtime=sp->lifetime;
2497 		lt.usetime= sp->validtime;
2498 		m = key_setlifetime(&lt, SADB_EXT_LIFETIME_HARD);
2499 		if (!m)
2500 			goto fail;
2501 		m_cat(result, m);
2502 	}
2503 
2504 	if ((result->m_flags & M_PKTHDR) == 0)
2505 		goto fail;
2506 
2507 	if (result->m_len < sizeof(struct sadb_msg)) {
2508 		result = m_pullup(result, sizeof(struct sadb_msg));
2509 		if (result == NULL)
2510 			goto fail;
2511 	}
2512 
2513 	result->m_pkthdr.len = 0;
2514 	for (m = result; m; m = m->m_next)
2515 		result->m_pkthdr.len += m->m_len;
2516 
2517 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2518 	    PFKEY_UNIT64(result->m_pkthdr.len);
2519 
2520 	return result;
2521 
2522 fail:
2523 	m_freem(result);
2524 	return NULL;
2525 }
2526 /*
2527  * get PFKEY message length for security policy and request.
2528  */
2529 static size_t
2530 key_getspreqmsglen(struct secpolicy *sp)
2531 {
2532 	size_t tlen, len;
2533 	int i;
2534 
2535 	tlen = sizeof(struct sadb_x_policy);
2536 	/* if is the policy for ipsec ? */
2537 	if (sp->policy != IPSEC_POLICY_IPSEC)
2538 		return (tlen);
2539 
2540 	/* get length of ipsec requests */
2541 	for (i = 0; i < sp->tcount; i++) {
2542 		len = sizeof(struct sadb_x_ipsecrequest)
2543 			+ sp->req[i]->saidx.src.sa.sa_len
2544 			+ sp->req[i]->saidx.dst.sa.sa_len;
2545 
2546 		tlen += PFKEY_ALIGN8(len);
2547 	}
2548 	return (tlen);
2549 }
2550 
2551 /*
2552  * SADB_SPDEXPIRE processing
2553  * send
2554  *   <base, address(SD), lifetime(CH), policy>
2555  * to KMD by PF_KEY.
2556  *
2557  * OUT:	0	: succeed
2558  *	others	: error number
2559  */
2560 static int
2561 key_spdexpire(struct secpolicy *sp)
2562 {
2563 	struct sadb_lifetime *lt;
2564 	struct mbuf *result = NULL, *m;
2565 	int len, error = -1;
2566 
2567 	IPSEC_ASSERT(sp != NULL, ("null secpolicy"));
2568 
2569 	KEYDBG(KEY_STAMP,
2570 	    printf("%s: SP(%p)\n", __func__, sp));
2571 	KEYDBG(KEY_DATA, kdebug_secpolicy(sp));
2572 
2573 	/* set msg header */
2574 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0);
2575 	if (!m) {
2576 		error = ENOBUFS;
2577 		goto fail;
2578 	}
2579 	result = m;
2580 
2581 	/* create lifetime extension (current and hard) */
2582 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
2583 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
2584 	if (m == NULL) {
2585 		error = ENOBUFS;
2586 		goto fail;
2587 	}
2588 	m_align(m, len);
2589 	m->m_len = len;
2590 	bzero(mtod(m, caddr_t), len);
2591 	lt = mtod(m, struct sadb_lifetime *);
2592 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2593 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
2594 	lt->sadb_lifetime_allocations = 0;
2595 	lt->sadb_lifetime_bytes = 0;
2596 	lt->sadb_lifetime_addtime = sp->created;
2597 	lt->sadb_lifetime_usetime = sp->lastused;
2598 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
2599 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
2600 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
2601 	lt->sadb_lifetime_allocations = 0;
2602 	lt->sadb_lifetime_bytes = 0;
2603 	lt->sadb_lifetime_addtime = sp->lifetime;
2604 	lt->sadb_lifetime_usetime = sp->validtime;
2605 	m_cat(result, m);
2606 
2607 	/* set sadb_address for source */
2608 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
2609 	    &sp->spidx.src.sa,
2610 	    sp->spidx.prefs, sp->spidx.ul_proto);
2611 	if (!m) {
2612 		error = ENOBUFS;
2613 		goto fail;
2614 	}
2615 	m_cat(result, m);
2616 
2617 	/* set sadb_address for destination */
2618 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
2619 	    &sp->spidx.dst.sa,
2620 	    sp->spidx.prefd, sp->spidx.ul_proto);
2621 	if (!m) {
2622 		error = ENOBUFS;
2623 		goto fail;
2624 	}
2625 	m_cat(result, m);
2626 
2627 	/* set secpolicy */
2628 	m = key_sp2mbuf(sp);
2629 	if (!m) {
2630 		error = ENOBUFS;
2631 		goto fail;
2632 	}
2633 	m_cat(result, m);
2634 
2635 	if ((result->m_flags & M_PKTHDR) == 0) {
2636 		error = EINVAL;
2637 		goto fail;
2638 	}
2639 
2640 	if (result->m_len < sizeof(struct sadb_msg)) {
2641 		result = m_pullup(result, sizeof(struct sadb_msg));
2642 		if (result == NULL) {
2643 			error = ENOBUFS;
2644 			goto fail;
2645 		}
2646 	}
2647 
2648 	result->m_pkthdr.len = 0;
2649 	for (m = result; m; m = m->m_next)
2650 		result->m_pkthdr.len += m->m_len;
2651 
2652 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2653 	    PFKEY_UNIT64(result->m_pkthdr.len);
2654 
2655 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
2656 
2657  fail:
2658 	if (result)
2659 		m_freem(result);
2660 	return error;
2661 }
2662 
2663 /* %%% SAD management */
2664 /*
2665  * allocating and initialize new SA head.
2666  * OUT:	NULL	: failure due to the lack of memory.
2667  *	others	: pointer to new SA head.
2668  */
2669 static struct secashead *
2670 key_newsah(struct secasindex *saidx)
2671 {
2672 	struct secashead *sah;
2673 
2674 	sah = malloc(sizeof(struct secashead), M_IPSEC_SAH,
2675 	    M_NOWAIT | M_ZERO);
2676 	if (sah == NULL) {
2677 		PFKEYSTAT_INC(in_nomem);
2678 		return (NULL);
2679 	}
2680 	TAILQ_INIT(&sah->savtree_larval);
2681 	TAILQ_INIT(&sah->savtree_alive);
2682 	sah->saidx = *saidx;
2683 	sah->state = SADB_SASTATE_DEAD;
2684 	SAH_INITREF(sah);
2685 
2686 	KEYDBG(KEY_STAMP,
2687 	    printf("%s: SAH(%p)\n", __func__, sah));
2688 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2689 	return (sah);
2690 }
2691 
2692 static void
2693 key_freesah(struct secashead **psah)
2694 {
2695 	struct secashead *sah = *psah;
2696 
2697 	if (SAH_DELREF(sah) == 0)
2698 		return;
2699 
2700 	KEYDBG(KEY_STAMP,
2701 	    printf("%s: last reference to SAH(%p)\n", __func__, sah));
2702 	KEYDBG(KEY_DATA, kdebug_secash(sah, NULL));
2703 
2704 	*psah = NULL;
2705 	key_delsah(sah);
2706 }
2707 
2708 static void
2709 key_delsah(struct secashead *sah)
2710 {
2711 	IPSEC_ASSERT(sah != NULL, ("NULL sah"));
2712 	IPSEC_ASSERT(sah->state == SADB_SASTATE_DEAD,
2713 	    ("Attempt to free non DEAD SAH %p", sah));
2714 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_larval),
2715 	    ("Attempt to free SAH %p with LARVAL SA", sah));
2716 	IPSEC_ASSERT(TAILQ_EMPTY(&sah->savtree_alive),
2717 	    ("Attempt to free SAH %p with ALIVE SA", sah));
2718 
2719 	free(sah, M_IPSEC_SAH);
2720 }
2721 
2722 /*
2723  * allocating a new SA for key_add() and key_getspi() call,
2724  * and copy the values of mhp into new buffer.
2725  * When SAD message type is SADB_GETSPI set SA state to LARVAL.
2726  * For SADB_ADD create and initialize SA with MATURE state.
2727  * OUT:	NULL	: fail
2728  *	others	: pointer to new secasvar.
2729  */
2730 static struct secasvar *
2731 key_newsav(const struct sadb_msghdr *mhp, struct secasindex *saidx,
2732     uint32_t spi, int *errp)
2733 {
2734 	struct secashead *sah;
2735 	struct secasvar *sav;
2736 	int isnew;
2737 
2738 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
2739 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
2740 	IPSEC_ASSERT(mhp->msg->sadb_msg_type == SADB_GETSPI ||
2741 	    mhp->msg->sadb_msg_type == SADB_ADD, ("wrong message type"));
2742 
2743 	sav = NULL;
2744 	sah = NULL;
2745 	/* check SPI value */
2746 	switch (saidx->proto) {
2747 	case IPPROTO_ESP:
2748 	case IPPROTO_AH:
2749 		/*
2750 		 * RFC 4302, 2.4. Security Parameters Index (SPI), SPI values
2751 		 * 1-255 reserved by IANA for future use,
2752 		 * 0 for implementation specific, local use.
2753 		 */
2754 		if (ntohl(spi) <= 255) {
2755 			ipseclog((LOG_DEBUG, "%s: illegal range of SPI %u.\n",
2756 			    __func__, ntohl(spi)));
2757 			*errp = EINVAL;
2758 			goto done;
2759 		}
2760 		break;
2761 	}
2762 
2763 	sav = malloc(sizeof(struct secasvar), M_IPSEC_SA, M_NOWAIT | M_ZERO);
2764 	if (sav == NULL) {
2765 		*errp = ENOBUFS;
2766 		goto done;
2767 	}
2768 	sav->lock = malloc(sizeof(struct mtx), M_IPSEC_MISC,
2769 	    M_NOWAIT | M_ZERO);
2770 	if (sav->lock == NULL) {
2771 		*errp = ENOBUFS;
2772 		goto done;
2773 	}
2774 	mtx_init(sav->lock, "ipsec association", NULL, MTX_DEF);
2775 	sav->lft_c = uma_zalloc(V_key_lft_zone, M_NOWAIT);
2776 	if (sav->lft_c == NULL) {
2777 		*errp = ENOBUFS;
2778 		goto done;
2779 	}
2780 	counter_u64_zero(sav->lft_c_allocations);
2781 	counter_u64_zero(sav->lft_c_bytes);
2782 
2783 	sav->spi = spi;
2784 	sav->seq = mhp->msg->sadb_msg_seq;
2785 	sav->state = SADB_SASTATE_LARVAL;
2786 	sav->pid = (pid_t)mhp->msg->sadb_msg_pid;
2787 	SAV_INITREF(sav);
2788 again:
2789 	sah = key_getsah(saidx);
2790 	if (sah == NULL) {
2791 		/* create a new SA index */
2792 		sah = key_newsah(saidx);
2793 		if (sah == NULL) {
2794 			ipseclog((LOG_DEBUG,
2795 			    "%s: No more memory.\n", __func__));
2796 			*errp = ENOBUFS;
2797 			goto done;
2798 		}
2799 		isnew = 1;
2800 	} else
2801 		isnew = 0;
2802 
2803 	sav->sah = sah;
2804 	if (mhp->msg->sadb_msg_type == SADB_GETSPI) {
2805 		sav->created = time_second;
2806 	} else if (sav->state == SADB_SASTATE_LARVAL) {
2807 		/*
2808 		 * Do not call key_setsaval() second time in case
2809 		 * of `goto again`. We will have MATURE state.
2810 		 */
2811 		*errp = key_setsaval(sav, mhp);
2812 		if (*errp != 0)
2813 			goto done;
2814 		sav->state = SADB_SASTATE_MATURE;
2815 	}
2816 
2817 	SAHTREE_WLOCK();
2818 	/*
2819 	 * Check that existing SAH wasn't unlinked.
2820 	 * Since we didn't hold the SAHTREE lock, it is possible,
2821 	 * that callout handler or key_flush() or key_delete() could
2822 	 * unlink this SAH.
2823 	 */
2824 	if (isnew == 0 && sah->state == SADB_SASTATE_DEAD) {
2825 		SAHTREE_WUNLOCK();
2826 		key_freesah(&sah);	/* reference from key_getsah() */
2827 		goto again;
2828 	}
2829 	if (isnew != 0) {
2830 		/*
2831 		 * Add new SAH into SADB.
2832 		 *
2833 		 * XXXAE: we can serialize key_add and key_getspi calls, so
2834 		 * several threads will not fight in the race.
2835 		 * Otherwise we should check under SAHTREE lock, that this
2836 		 * SAH would not added twice.
2837 		 */
2838 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
2839 		/* Add new SAH into hash by addresses */
2840 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
2841 		/* Now we are linked in the chain */
2842 		sah->state = SADB_SASTATE_MATURE;
2843 		/*
2844 		 * SAV references this new SAH.
2845 		 * In case of existing SAH we reuse reference
2846 		 * from key_getsah().
2847 		 */
2848 		SAH_ADDREF(sah);
2849 	}
2850 	/* Link SAV with SAH */
2851 	if (sav->state == SADB_SASTATE_MATURE)
2852 		TAILQ_INSERT_HEAD(&sah->savtree_alive, sav, chain);
2853 	else
2854 		TAILQ_INSERT_HEAD(&sah->savtree_larval, sav, chain);
2855 	/* Add SAV into SPI hash */
2856 	LIST_INSERT_HEAD(SAVHASH_HASH(sav->spi), sav, spihash);
2857 	SAHTREE_WUNLOCK();
2858 	*errp = 0;	/* success */
2859 done:
2860 	if (*errp != 0) {
2861 		if (sav != NULL) {
2862 			if (sav->lock != NULL) {
2863 				mtx_destroy(sav->lock);
2864 				free(sav->lock, M_IPSEC_MISC);
2865 			}
2866 			if (sav->lft_c != NULL)
2867 				uma_zfree(V_key_lft_zone, sav->lft_c);
2868 			free(sav, M_IPSEC_SA), sav = NULL;
2869 		}
2870 		if (sah != NULL)
2871 			key_freesah(&sah);
2872 		if (*errp == ENOBUFS) {
2873 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
2874 			    __func__));
2875 			PFKEYSTAT_INC(in_nomem);
2876 		}
2877 	}
2878 	return (sav);
2879 }
2880 
2881 /*
2882  * free() SA variable entry.
2883  */
2884 static void
2885 key_cleansav(struct secasvar *sav)
2886 {
2887 
2888 	if (sav->natt != NULL) {
2889 		free(sav->natt, M_IPSEC_MISC);
2890 		sav->natt = NULL;
2891 	}
2892 	if (sav->flags & SADB_X_EXT_F_CLONED)
2893 		return;
2894 	/*
2895 	 * Cleanup xform state.  Note that zeroize'ing causes the
2896 	 * keys to be cleared; otherwise we must do it ourself.
2897 	 */
2898 	if (sav->tdb_xform != NULL) {
2899 		sav->tdb_xform->xf_zeroize(sav);
2900 		sav->tdb_xform = NULL;
2901 	} else {
2902 		if (sav->key_auth != NULL)
2903 			bzero(sav->key_auth->key_data, _KEYLEN(sav->key_auth));
2904 		if (sav->key_enc != NULL)
2905 			bzero(sav->key_enc->key_data, _KEYLEN(sav->key_enc));
2906 	}
2907 	if (sav->key_auth != NULL) {
2908 		if (sav->key_auth->key_data != NULL)
2909 			free(sav->key_auth->key_data, M_IPSEC_MISC);
2910 		free(sav->key_auth, M_IPSEC_MISC);
2911 		sav->key_auth = NULL;
2912 	}
2913 	if (sav->key_enc != NULL) {
2914 		if (sav->key_enc->key_data != NULL)
2915 			free(sav->key_enc->key_data, M_IPSEC_MISC);
2916 		free(sav->key_enc, M_IPSEC_MISC);
2917 		sav->key_enc = NULL;
2918 	}
2919 	if (sav->replay != NULL) {
2920 		if (sav->replay->bitmap != NULL)
2921 			free(sav->replay->bitmap, M_IPSEC_MISC);
2922 		free(sav->replay, M_IPSEC_MISC);
2923 		sav->replay = NULL;
2924 	}
2925 	if (sav->lft_h != NULL) {
2926 		free(sav->lft_h, M_IPSEC_MISC);
2927 		sav->lft_h = NULL;
2928 	}
2929 	if (sav->lft_s != NULL) {
2930 		free(sav->lft_s, M_IPSEC_MISC);
2931 		sav->lft_s = NULL;
2932 	}
2933 }
2934 
2935 /*
2936  * free() SA variable entry.
2937  */
2938 static void
2939 key_delsav(struct secasvar *sav)
2940 {
2941 	IPSEC_ASSERT(sav != NULL, ("null sav"));
2942 	IPSEC_ASSERT(sav->state == SADB_SASTATE_DEAD,
2943 	    ("attempt to free non DEAD SA %p", sav));
2944 	IPSEC_ASSERT(sav->refcnt == 0, ("reference count %u > 0",
2945 	    sav->refcnt));
2946 
2947 	/*
2948 	 * SA must be unlinked from the chain and hashtbl.
2949 	 * If SA was cloned, we leave all fields untouched,
2950 	 * except NAT-T config.
2951 	 */
2952 	key_cleansav(sav);
2953 	if ((sav->flags & SADB_X_EXT_F_CLONED) == 0) {
2954 		mtx_destroy(sav->lock);
2955 		free(sav->lock, M_IPSEC_MISC);
2956 		uma_zfree(V_key_lft_zone, sav->lft_c);
2957 	}
2958 	free(sav, M_IPSEC_SA);
2959 }
2960 
2961 /*
2962  * search SAH.
2963  * OUT:
2964  *	NULL	: not found
2965  *	others	: found, referenced pointer to a SAH.
2966  */
2967 static struct secashead *
2968 key_getsah(struct secasindex *saidx)
2969 {
2970 	SAHTREE_RLOCK_TRACKER;
2971 	struct secashead *sah;
2972 
2973 	SAHTREE_RLOCK();
2974 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
2975 	    if (key_cmpsaidx(&sah->saidx, saidx, CMP_MODE_REQID) != 0) {
2976 		    SAH_ADDREF(sah);
2977 		    break;
2978 	    }
2979 	}
2980 	SAHTREE_RUNLOCK();
2981 	return (sah);
2982 }
2983 
2984 /*
2985  * Check not to be duplicated SPI.
2986  * OUT:
2987  *	0	: not found
2988  *	1	: found SA with given SPI.
2989  */
2990 static int
2991 key_checkspidup(uint32_t spi)
2992 {
2993 	SAHTREE_RLOCK_TRACKER;
2994 	struct secasvar *sav;
2995 
2996 	/* Assume SPI is in network byte order */
2997 	SAHTREE_RLOCK();
2998 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
2999 		if (sav->spi == spi)
3000 			break;
3001 	}
3002 	SAHTREE_RUNLOCK();
3003 	return (sav != NULL);
3004 }
3005 
3006 /*
3007  * Search SA by SPI.
3008  * OUT:
3009  *	NULL	: not found
3010  *	others	: found, referenced pointer to a SA.
3011  */
3012 static struct secasvar *
3013 key_getsavbyspi(uint32_t spi)
3014 {
3015 	SAHTREE_RLOCK_TRACKER;
3016 	struct secasvar *sav;
3017 
3018 	/* Assume SPI is in network byte order */
3019 	SAHTREE_RLOCK();
3020 	LIST_FOREACH(sav, SAVHASH_HASH(spi), spihash) {
3021 		if (sav->spi != spi)
3022 			continue;
3023 		SAV_ADDREF(sav);
3024 		break;
3025 	}
3026 	SAHTREE_RUNLOCK();
3027 	return (sav);
3028 }
3029 
3030 static int
3031 key_updatelifetimes(struct secasvar *sav, const struct sadb_msghdr *mhp)
3032 {
3033 	struct seclifetime *lft_h, *lft_s, *tmp;
3034 
3035 	/* Lifetime extension is optional, check that it is present. */
3036 	if (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3037 	    SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) {
3038 		/*
3039 		 * In case of SADB_UPDATE we may need to change
3040 		 * existing lifetimes.
3041 		 */
3042 		if (sav->state == SADB_SASTATE_MATURE) {
3043 			lft_h = lft_s = NULL;
3044 			goto reset;
3045 		}
3046 		return (0);
3047 	}
3048 	/* Both HARD and SOFT extensions must present */
3049 	if ((SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
3050 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
3051 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
3052 	    !SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
3053 		ipseclog((LOG_DEBUG,
3054 		    "%s: invalid message: missing required header.\n",
3055 		    __func__));
3056 		return (EINVAL);
3057 	}
3058 	if (SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_HARD) ||
3059 	    SADB_CHECKLEN(mhp, SADB_EXT_LIFETIME_SOFT)) {
3060 		ipseclog((LOG_DEBUG,
3061 		    "%s: invalid message: wrong header size.\n", __func__));
3062 		return (EINVAL);
3063 	}
3064 	lft_h = key_dup_lifemsg((const struct sadb_lifetime *)
3065 	    mhp->ext[SADB_EXT_LIFETIME_HARD], M_IPSEC_MISC);
3066 	if (lft_h == NULL) {
3067 		PFKEYSTAT_INC(in_nomem);
3068 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3069 		return (ENOBUFS);
3070 	}
3071 	lft_s = key_dup_lifemsg((const struct sadb_lifetime *)
3072 	    mhp->ext[SADB_EXT_LIFETIME_SOFT], M_IPSEC_MISC);
3073 	if (lft_s == NULL) {
3074 		PFKEYSTAT_INC(in_nomem);
3075 		free(lft_h, M_IPSEC_MISC);
3076 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3077 		return (ENOBUFS);
3078 	}
3079 reset:
3080 	if (sav->state != SADB_SASTATE_LARVAL) {
3081 		/*
3082 		 * key_update() holds reference to this SA,
3083 		 * so it won't be deleted in meanwhile.
3084 		 */
3085 		SECASVAR_LOCK(sav);
3086 		tmp = sav->lft_h;
3087 		sav->lft_h = lft_h;
3088 		lft_h = tmp;
3089 
3090 		tmp = sav->lft_s;
3091 		sav->lft_s = lft_s;
3092 		lft_s = tmp;
3093 		SECASVAR_UNLOCK(sav);
3094 		if (lft_h != NULL)
3095 			free(lft_h, M_IPSEC_MISC);
3096 		if (lft_s != NULL)
3097 			free(lft_s, M_IPSEC_MISC);
3098 		return (0);
3099 	}
3100 	/* We can update lifetime without holding a lock */
3101 	IPSEC_ASSERT(sav->lft_h == NULL, ("lft_h is already initialized\n"));
3102 	IPSEC_ASSERT(sav->lft_s == NULL, ("lft_s is already initialized\n"));
3103 	sav->lft_h = lft_h;
3104 	sav->lft_s = lft_s;
3105 	return (0);
3106 }
3107 
3108 /*
3109  * copy SA values from PF_KEY message except *SPI, SEQ, PID and TYPE*.
3110  * You must update these if need. Expects only LARVAL SAs.
3111  * OUT:	0:	success.
3112  *	!0:	failure.
3113  */
3114 static int
3115 key_setsaval(struct secasvar *sav, const struct sadb_msghdr *mhp)
3116 {
3117 	const struct sadb_sa *sa0;
3118 	const struct sadb_key *key0;
3119 	uint32_t replay;
3120 	size_t len;
3121 	int error;
3122 
3123 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
3124 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
3125 	IPSEC_ASSERT(sav->state == SADB_SASTATE_LARVAL,
3126 	    ("Attempt to update non LARVAL SA"));
3127 
3128 	/* XXX rewrite */
3129 	error = key_setident(sav->sah, mhp);
3130 	if (error != 0)
3131 		goto fail;
3132 
3133 	/* SA */
3134 	if (!SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
3135 		if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
3136 			error = EINVAL;
3137 			goto fail;
3138 		}
3139 		sa0 = (const struct sadb_sa *)mhp->ext[SADB_EXT_SA];
3140 		sav->alg_auth = sa0->sadb_sa_auth;
3141 		sav->alg_enc = sa0->sadb_sa_encrypt;
3142 		sav->flags = sa0->sadb_sa_flags;
3143 		if ((sav->flags & SADB_KEY_FLAGS_MAX) != sav->flags) {
3144 			ipseclog((LOG_DEBUG,
3145 			    "%s: invalid sa_flags 0x%08x.\n", __func__,
3146 			    sav->flags));
3147 			error = EINVAL;
3148 			goto fail;
3149 		}
3150 
3151 		/* Optional replay window */
3152 		replay = 0;
3153 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0)
3154 			replay = sa0->sadb_sa_replay;
3155 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_SA_REPLAY)) {
3156 			if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA_REPLAY)) {
3157 				error = EINVAL;
3158 				goto fail;
3159 			}
3160 			replay = ((const struct sadb_x_sa_replay *)
3161 			    mhp->ext[SADB_X_EXT_SA_REPLAY])->sadb_x_sa_replay_replay;
3162 
3163 			if (replay > UINT32_MAX - 32) {
3164 				ipseclog((LOG_DEBUG,
3165 				    "%s: replay window too big.\n", __func__));
3166 				error = EINVAL;
3167 				goto fail;
3168 			}
3169 
3170 			replay = (replay + 7) >> 3;
3171 		}
3172 
3173 		sav->replay = malloc(sizeof(struct secreplay), M_IPSEC_MISC,
3174 		    M_NOWAIT | M_ZERO);
3175 		if (sav->replay == NULL) {
3176 			PFKEYSTAT_INC(in_nomem);
3177 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3178 			    __func__));
3179 			error = ENOBUFS;
3180 			goto fail;
3181 		}
3182 
3183 		if (replay != 0) {
3184 			/* number of 32b blocks to be allocated */
3185 			uint32_t bitmap_size;
3186 
3187 			/* RFC 6479:
3188 			 * - the allocated replay window size must be
3189 			 *   a power of two.
3190 			 * - use an extra 32b block as a redundant window.
3191 			 */
3192 			bitmap_size = 1;
3193 			while (replay + 4 > bitmap_size)
3194 				bitmap_size <<= 1;
3195 			bitmap_size = bitmap_size / 4;
3196 
3197 			sav->replay->bitmap = malloc(
3198 			    bitmap_size * sizeof(uint32_t), M_IPSEC_MISC,
3199 			    M_NOWAIT | M_ZERO);
3200 			if (sav->replay->bitmap == NULL) {
3201 				PFKEYSTAT_INC(in_nomem);
3202 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3203 					__func__));
3204 				error = ENOBUFS;
3205 				goto fail;
3206 			}
3207 			sav->replay->bitmap_size = bitmap_size;
3208 			sav->replay->wsize = replay;
3209 		}
3210 	}
3211 
3212 	/* Authentication keys */
3213 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
3214 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH)) {
3215 			error = EINVAL;
3216 			goto fail;
3217 		}
3218 		error = 0;
3219 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_AUTH];
3220 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3221 		switch (mhp->msg->sadb_msg_satype) {
3222 		case SADB_SATYPE_AH:
3223 		case SADB_SATYPE_ESP:
3224 		case SADB_X_SATYPE_TCPSIGNATURE:
3225 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3226 			    sav->alg_auth != SADB_X_AALG_NULL)
3227 				error = EINVAL;
3228 			break;
3229 		case SADB_X_SATYPE_IPCOMP:
3230 		default:
3231 			error = EINVAL;
3232 			break;
3233 		}
3234 		if (error) {
3235 			ipseclog((LOG_DEBUG, "%s: invalid key_auth values.\n",
3236 				__func__));
3237 			goto fail;
3238 		}
3239 
3240 		sav->key_auth = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3241 		if (sav->key_auth == NULL ) {
3242 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3243 				  __func__));
3244 			PFKEYSTAT_INC(in_nomem);
3245 			error = ENOBUFS;
3246 			goto fail;
3247 		}
3248 	}
3249 
3250 	/* Encryption key */
3251 	if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) {
3252 		if (SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT)) {
3253 			error = EINVAL;
3254 			goto fail;
3255 		}
3256 		error = 0;
3257 		key0 = (const struct sadb_key *)mhp->ext[SADB_EXT_KEY_ENCRYPT];
3258 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3259 		switch (mhp->msg->sadb_msg_satype) {
3260 		case SADB_SATYPE_ESP:
3261 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3262 			    sav->alg_enc != SADB_EALG_NULL) {
3263 				error = EINVAL;
3264 				break;
3265 			}
3266 			sav->key_enc = key_dup_keymsg(key0, len, M_IPSEC_MISC);
3267 			if (sav->key_enc == NULL) {
3268 				ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3269 					__func__));
3270 				PFKEYSTAT_INC(in_nomem);
3271 				error = ENOBUFS;
3272 				goto fail;
3273 			}
3274 			break;
3275 		case SADB_X_SATYPE_IPCOMP:
3276 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3277 				error = EINVAL;
3278 			sav->key_enc = NULL;	/*just in case*/
3279 			break;
3280 		case SADB_SATYPE_AH:
3281 		case SADB_X_SATYPE_TCPSIGNATURE:
3282 		default:
3283 			error = EINVAL;
3284 			break;
3285 		}
3286 		if (error) {
3287 			ipseclog((LOG_DEBUG, "%s: invalid key_enc value.\n",
3288 				__func__));
3289 			goto fail;
3290 		}
3291 	}
3292 
3293 	/* set iv */
3294 	sav->ivlen = 0;
3295 	switch (mhp->msg->sadb_msg_satype) {
3296 	case SADB_SATYPE_AH:
3297 		if (sav->flags & SADB_X_EXT_DERIV) {
3298 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3299 			    "given to AH SA.\n", __func__));
3300 			error = EINVAL;
3301 			goto fail;
3302 		}
3303 		if (sav->alg_enc != SADB_EALG_NONE) {
3304 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3305 			    "mismated.\n", __func__));
3306 			error = EINVAL;
3307 			goto fail;
3308 		}
3309 		error = xform_init(sav, XF_AH);
3310 		break;
3311 	case SADB_SATYPE_ESP:
3312 		if ((sav->flags & (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) ==
3313 		    (SADB_X_EXT_OLD | SADB_X_EXT_DERIV)) {
3314 			ipseclog((LOG_DEBUG, "%s: invalid flag (derived) "
3315 			    "given to old-esp.\n", __func__));
3316 			error = EINVAL;
3317 			goto fail;
3318 		}
3319 		error = xform_init(sav, XF_ESP);
3320 		break;
3321 	case SADB_X_SATYPE_IPCOMP:
3322 		if (sav->alg_auth != SADB_AALG_NONE) {
3323 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3324 			    "mismated.\n", __func__));
3325 			error = EINVAL;
3326 			goto fail;
3327 		}
3328 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0 &&
3329 		    ntohl(sav->spi) >= 0x10000) {
3330 			ipseclog((LOG_DEBUG, "%s: invalid cpi for IPComp.\n",
3331 			    __func__));
3332 			error = EINVAL;
3333 			goto fail;
3334 		}
3335 		error = xform_init(sav, XF_IPCOMP);
3336 		break;
3337 	case SADB_X_SATYPE_TCPSIGNATURE:
3338 		if (sav->alg_enc != SADB_EALG_NONE) {
3339 			ipseclog((LOG_DEBUG, "%s: protocol and algorithm "
3340 			    "mismated.\n", __func__));
3341 			error = EINVAL;
3342 			goto fail;
3343 		}
3344 		error = xform_init(sav, XF_TCPSIGNATURE);
3345 		break;
3346 	default:
3347 		ipseclog((LOG_DEBUG, "%s: Invalid satype.\n", __func__));
3348 		error = EPROTONOSUPPORT;
3349 		goto fail;
3350 	}
3351 	if (error) {
3352 		ipseclog((LOG_DEBUG, "%s: unable to initialize SA type %u.\n",
3353 		    __func__, mhp->msg->sadb_msg_satype));
3354 		goto fail;
3355 	}
3356 
3357 	/* Handle NAT-T headers */
3358 	error = key_setnatt(sav, mhp);
3359 	if (error != 0)
3360 		goto fail;
3361 
3362 	/* Initialize lifetime for CURRENT */
3363 	sav->firstused = 0;
3364 	sav->created = time_second;
3365 
3366 	/* lifetimes for HARD and SOFT */
3367 	error = key_updatelifetimes(sav, mhp);
3368 	if (error == 0)
3369 		return (0);
3370 fail:
3371 	key_cleansav(sav);
3372 	return (error);
3373 }
3374 
3375 /*
3376  * subroutine for SADB_GET and SADB_DUMP.
3377  */
3378 static struct mbuf *
3379 key_setdumpsa(struct secasvar *sav, uint8_t type, uint8_t satype,
3380     uint32_t seq, uint32_t pid)
3381 {
3382 	struct seclifetime lft_c;
3383 	struct mbuf *result = NULL, *tres = NULL, *m;
3384 	int i, dumporder[] = {
3385 		SADB_EXT_SA, SADB_X_EXT_SA2, SADB_X_EXT_SA_REPLAY,
3386 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3387 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3388 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY,
3389 		SADB_EXT_KEY_AUTH, SADB_EXT_KEY_ENCRYPT,
3390 		SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
3391 		SADB_EXT_SENSITIVITY,
3392 		SADB_X_EXT_NAT_T_TYPE,
3393 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3394 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3395 		SADB_X_EXT_NAT_T_FRAG,
3396 	};
3397 	uint32_t replay_count;
3398 
3399 	m = key_setsadbmsg(type, 0, satype, seq, pid, sav->refcnt);
3400 	if (m == NULL)
3401 		goto fail;
3402 	result = m;
3403 
3404 	for (i = nitems(dumporder) - 1; i >= 0; i--) {
3405 		m = NULL;
3406 		switch (dumporder[i]) {
3407 		case SADB_EXT_SA:
3408 			m = key_setsadbsa(sav);
3409 			if (!m)
3410 				goto fail;
3411 			break;
3412 
3413 		case SADB_X_EXT_SA2:
3414 			SECASVAR_LOCK(sav);
3415 			replay_count = sav->replay ? sav->replay->count : 0;
3416 			SECASVAR_UNLOCK(sav);
3417 			m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
3418 					sav->sah->saidx.reqid);
3419 			if (!m)
3420 				goto fail;
3421 			break;
3422 
3423 		case SADB_X_EXT_SA_REPLAY:
3424 			if (sav->replay == NULL ||
3425 			    sav->replay->wsize <= UINT8_MAX)
3426 				continue;
3427 
3428 			m = key_setsadbxsareplay(sav->replay->wsize);
3429 			if (!m)
3430 				goto fail;
3431 			break;
3432 
3433 		case SADB_EXT_ADDRESS_SRC:
3434 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3435 			    &sav->sah->saidx.src.sa,
3436 			    FULLMASK, IPSEC_ULPROTO_ANY);
3437 			if (!m)
3438 				goto fail;
3439 			break;
3440 
3441 		case SADB_EXT_ADDRESS_DST:
3442 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3443 			    &sav->sah->saidx.dst.sa,
3444 			    FULLMASK, IPSEC_ULPROTO_ANY);
3445 			if (!m)
3446 				goto fail;
3447 			break;
3448 
3449 		case SADB_EXT_KEY_AUTH:
3450 			if (!sav->key_auth)
3451 				continue;
3452 			m = key_setkey(sav->key_auth, SADB_EXT_KEY_AUTH);
3453 			if (!m)
3454 				goto fail;
3455 			break;
3456 
3457 		case SADB_EXT_KEY_ENCRYPT:
3458 			if (!sav->key_enc)
3459 				continue;
3460 			m = key_setkey(sav->key_enc, SADB_EXT_KEY_ENCRYPT);
3461 			if (!m)
3462 				goto fail;
3463 			break;
3464 
3465 		case SADB_EXT_LIFETIME_CURRENT:
3466 			lft_c.addtime = sav->created;
3467 			lft_c.allocations = (uint32_t)counter_u64_fetch(
3468 			    sav->lft_c_allocations);
3469 			lft_c.bytes = counter_u64_fetch(sav->lft_c_bytes);
3470 			lft_c.usetime = sav->firstused;
3471 			m = key_setlifetime(&lft_c, SADB_EXT_LIFETIME_CURRENT);
3472 			if (!m)
3473 				goto fail;
3474 			break;
3475 
3476 		case SADB_EXT_LIFETIME_HARD:
3477 			if (!sav->lft_h)
3478 				continue;
3479 			m = key_setlifetime(sav->lft_h,
3480 					    SADB_EXT_LIFETIME_HARD);
3481 			if (!m)
3482 				goto fail;
3483 			break;
3484 
3485 		case SADB_EXT_LIFETIME_SOFT:
3486 			if (!sav->lft_s)
3487 				continue;
3488 			m = key_setlifetime(sav->lft_s,
3489 					    SADB_EXT_LIFETIME_SOFT);
3490 
3491 			if (!m)
3492 				goto fail;
3493 			break;
3494 
3495 		case SADB_X_EXT_NAT_T_TYPE:
3496 			if (sav->natt == NULL)
3497 				continue;
3498 			m = key_setsadbxtype(UDP_ENCAP_ESPINUDP);
3499 			if (!m)
3500 				goto fail;
3501 			break;
3502 
3503 		case SADB_X_EXT_NAT_T_DPORT:
3504 			if (sav->natt == NULL)
3505 				continue;
3506 			m = key_setsadbxport(sav->natt->dport,
3507 			    SADB_X_EXT_NAT_T_DPORT);
3508 			if (!m)
3509 				goto fail;
3510 			break;
3511 
3512 		case SADB_X_EXT_NAT_T_SPORT:
3513 			if (sav->natt == NULL)
3514 				continue;
3515 			m = key_setsadbxport(sav->natt->sport,
3516 			    SADB_X_EXT_NAT_T_SPORT);
3517 			if (!m)
3518 				goto fail;
3519 			break;
3520 
3521 		case SADB_X_EXT_NAT_T_OAI:
3522 			if (sav->natt == NULL ||
3523 			    (sav->natt->flags & IPSEC_NATT_F_OAI) == 0)
3524 				continue;
3525 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAI,
3526 			    &sav->natt->oai.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3527 			if (!m)
3528 				goto fail;
3529 			break;
3530 		case SADB_X_EXT_NAT_T_OAR:
3531 			if (sav->natt == NULL ||
3532 			    (sav->natt->flags & IPSEC_NATT_F_OAR) == 0)
3533 				continue;
3534 			m = key_setsadbaddr(SADB_X_EXT_NAT_T_OAR,
3535 			    &sav->natt->oar.sa, FULLMASK, IPSEC_ULPROTO_ANY);
3536 			if (!m)
3537 				goto fail;
3538 			break;
3539 		case SADB_X_EXT_NAT_T_FRAG:
3540 			/* We do not (yet) support those. */
3541 			continue;
3542 
3543 		case SADB_EXT_ADDRESS_PROXY:
3544 		case SADB_EXT_IDENTITY_SRC:
3545 		case SADB_EXT_IDENTITY_DST:
3546 			/* XXX: should we brought from SPD ? */
3547 		case SADB_EXT_SENSITIVITY:
3548 		default:
3549 			continue;
3550 		}
3551 
3552 		if (!m)
3553 			goto fail;
3554 		if (tres)
3555 			m_cat(m, tres);
3556 		tres = m;
3557 	}
3558 
3559 	m_cat(result, tres);
3560 	tres = NULL;
3561 	if (result->m_len < sizeof(struct sadb_msg)) {
3562 		result = m_pullup(result, sizeof(struct sadb_msg));
3563 		if (result == NULL)
3564 			goto fail;
3565 	}
3566 
3567 	result->m_pkthdr.len = 0;
3568 	for (m = result; m; m = m->m_next)
3569 		result->m_pkthdr.len += m->m_len;
3570 
3571 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3572 	    PFKEY_UNIT64(result->m_pkthdr.len);
3573 
3574 	return result;
3575 
3576 fail:
3577 	m_freem(result);
3578 	m_freem(tres);
3579 	return NULL;
3580 }
3581 
3582 /*
3583  * set data into sadb_msg.
3584  */
3585 static struct mbuf *
3586 key_setsadbmsg(u_int8_t type, u_int16_t tlen, u_int8_t satype, u_int32_t seq,
3587     pid_t pid, u_int16_t reserved)
3588 {
3589 	struct mbuf *m;
3590 	struct sadb_msg *p;
3591 	int len;
3592 
3593 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
3594 	if (len > MCLBYTES)
3595 		return NULL;
3596 	MGETHDR(m, M_NOWAIT, MT_DATA);
3597 	if (m && len > MHLEN) {
3598 		if (!(MCLGET(m, M_NOWAIT))) {
3599 			m_freem(m);
3600 			m = NULL;
3601 		}
3602 	}
3603 	if (!m)
3604 		return NULL;
3605 	m->m_pkthdr.len = m->m_len = len;
3606 	m->m_next = NULL;
3607 
3608 	p = mtod(m, struct sadb_msg *);
3609 
3610 	bzero(p, len);
3611 	p->sadb_msg_version = PF_KEY_V2;
3612 	p->sadb_msg_type = type;
3613 	p->sadb_msg_errno = 0;
3614 	p->sadb_msg_satype = satype;
3615 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
3616 	p->sadb_msg_reserved = reserved;
3617 	p->sadb_msg_seq = seq;
3618 	p->sadb_msg_pid = (u_int32_t)pid;
3619 
3620 	return m;
3621 }
3622 
3623 /*
3624  * copy secasvar data into sadb_address.
3625  */
3626 static struct mbuf *
3627 key_setsadbsa(struct secasvar *sav)
3628 {
3629 	struct mbuf *m;
3630 	struct sadb_sa *p;
3631 	int len;
3632 
3633 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
3634 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3635 	if (m == NULL)
3636 		return (NULL);
3637 	m_align(m, len);
3638 	m->m_len = len;
3639 	p = mtod(m, struct sadb_sa *);
3640 	bzero(p, len);
3641 	p->sadb_sa_len = PFKEY_UNIT64(len);
3642 	p->sadb_sa_exttype = SADB_EXT_SA;
3643 	p->sadb_sa_spi = sav->spi;
3644 	p->sadb_sa_replay = sav->replay ?
3645 	    (sav->replay->wsize > UINT8_MAX ? UINT8_MAX :
3646 		sav->replay->wsize): 0;
3647 	p->sadb_sa_state = sav->state;
3648 	p->sadb_sa_auth = sav->alg_auth;
3649 	p->sadb_sa_encrypt = sav->alg_enc;
3650 	p->sadb_sa_flags = sav->flags & SADB_KEY_FLAGS_MAX;
3651 	return (m);
3652 }
3653 
3654 /*
3655  * set data into sadb_address.
3656  */
3657 static struct mbuf *
3658 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
3659     u_int8_t prefixlen, u_int16_t ul_proto)
3660 {
3661 	struct mbuf *m;
3662 	struct sadb_address *p;
3663 	size_t len;
3664 
3665 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
3666 	    PFKEY_ALIGN8(saddr->sa_len);
3667 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3668 	if (m == NULL)
3669 		return (NULL);
3670 	m_align(m, len);
3671 	m->m_len = len;
3672 	p = mtod(m, struct sadb_address *);
3673 
3674 	bzero(p, len);
3675 	p->sadb_address_len = PFKEY_UNIT64(len);
3676 	p->sadb_address_exttype = exttype;
3677 	p->sadb_address_proto = ul_proto;
3678 	if (prefixlen == FULLMASK) {
3679 		switch (saddr->sa_family) {
3680 		case AF_INET:
3681 			prefixlen = sizeof(struct in_addr) << 3;
3682 			break;
3683 		case AF_INET6:
3684 			prefixlen = sizeof(struct in6_addr) << 3;
3685 			break;
3686 		default:
3687 			; /*XXX*/
3688 		}
3689 	}
3690 	p->sadb_address_prefixlen = prefixlen;
3691 	p->sadb_address_reserved = 0;
3692 
3693 	bcopy(saddr,
3694 	    mtod(m, caddr_t) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
3695 	    saddr->sa_len);
3696 
3697 	return m;
3698 }
3699 
3700 /*
3701  * set data into sadb_x_sa2.
3702  */
3703 static struct mbuf *
3704 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int32_t reqid)
3705 {
3706 	struct mbuf *m;
3707 	struct sadb_x_sa2 *p;
3708 	size_t len;
3709 
3710 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
3711 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3712 	if (m == NULL)
3713 		return (NULL);
3714 	m_align(m, len);
3715 	m->m_len = len;
3716 	p = mtod(m, struct sadb_x_sa2 *);
3717 
3718 	bzero(p, len);
3719 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
3720 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
3721 	p->sadb_x_sa2_mode = mode;
3722 	p->sadb_x_sa2_reserved1 = 0;
3723 	p->sadb_x_sa2_reserved2 = 0;
3724 	p->sadb_x_sa2_sequence = seq;
3725 	p->sadb_x_sa2_reqid = reqid;
3726 
3727 	return m;
3728 }
3729 
3730 /*
3731  * Set data into sadb_x_sa_replay.
3732  */
3733 static struct mbuf *
3734 key_setsadbxsareplay(u_int32_t replay)
3735 {
3736 	struct mbuf *m;
3737 	struct sadb_x_sa_replay *p;
3738 	size_t len;
3739 
3740 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa_replay));
3741 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3742 	if (m == NULL)
3743 		return (NULL);
3744 	m_align(m, len);
3745 	m->m_len = len;
3746 	p = mtod(m, struct sadb_x_sa_replay *);
3747 
3748 	bzero(p, len);
3749 	p->sadb_x_sa_replay_len = PFKEY_UNIT64(len);
3750 	p->sadb_x_sa_replay_exttype = SADB_X_EXT_SA_REPLAY;
3751 	p->sadb_x_sa_replay_replay = (replay << 3);
3752 
3753 	return m;
3754 }
3755 
3756 /*
3757  * Set a type in sadb_x_nat_t_type.
3758  */
3759 static struct mbuf *
3760 key_setsadbxtype(u_int16_t type)
3761 {
3762 	struct mbuf *m;
3763 	size_t len;
3764 	struct sadb_x_nat_t_type *p;
3765 
3766 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
3767 
3768 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3769 	if (m == NULL)
3770 		return (NULL);
3771 	m_align(m, len);
3772 	m->m_len = len;
3773 	p = mtod(m, struct sadb_x_nat_t_type *);
3774 
3775 	bzero(p, len);
3776 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
3777 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
3778 	p->sadb_x_nat_t_type_type = type;
3779 
3780 	return (m);
3781 }
3782 /*
3783  * Set a port in sadb_x_nat_t_port.
3784  * In contrast to default RFC 2367 behaviour, port is in network byte order.
3785  */
3786 static struct mbuf *
3787 key_setsadbxport(u_int16_t port, u_int16_t type)
3788 {
3789 	struct mbuf *m;
3790 	size_t len;
3791 	struct sadb_x_nat_t_port *p;
3792 
3793 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
3794 
3795 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3796 	if (m == NULL)
3797 		return (NULL);
3798 	m_align(m, len);
3799 	m->m_len = len;
3800 	p = mtod(m, struct sadb_x_nat_t_port *);
3801 
3802 	bzero(p, len);
3803 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
3804 	p->sadb_x_nat_t_port_exttype = type;
3805 	p->sadb_x_nat_t_port_port = port;
3806 
3807 	return (m);
3808 }
3809 
3810 /*
3811  * Get port from sockaddr. Port is in network byte order.
3812  */
3813 uint16_t
3814 key_portfromsaddr(struct sockaddr *sa)
3815 {
3816 
3817 	switch (sa->sa_family) {
3818 #ifdef INET
3819 	case AF_INET:
3820 		return ((struct sockaddr_in *)sa)->sin_port;
3821 #endif
3822 #ifdef INET6
3823 	case AF_INET6:
3824 		return ((struct sockaddr_in6 *)sa)->sin6_port;
3825 #endif
3826 	}
3827 	return (0);
3828 }
3829 
3830 /*
3831  * Set port in struct sockaddr. Port is in network byte order.
3832  */
3833 void
3834 key_porttosaddr(struct sockaddr *sa, uint16_t port)
3835 {
3836 
3837 	switch (sa->sa_family) {
3838 #ifdef INET
3839 	case AF_INET:
3840 		((struct sockaddr_in *)sa)->sin_port = port;
3841 		break;
3842 #endif
3843 #ifdef INET6
3844 	case AF_INET6:
3845 		((struct sockaddr_in6 *)sa)->sin6_port = port;
3846 		break;
3847 #endif
3848 	default:
3849 		ipseclog((LOG_DEBUG, "%s: unexpected address family %d.\n",
3850 			__func__, sa->sa_family));
3851 		break;
3852 	}
3853 }
3854 
3855 /*
3856  * set data into sadb_x_policy
3857  */
3858 static struct mbuf *
3859 key_setsadbxpolicy(u_int16_t type, u_int8_t dir, u_int32_t id, u_int32_t priority)
3860 {
3861 	struct mbuf *m;
3862 	struct sadb_x_policy *p;
3863 	size_t len;
3864 
3865 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
3866 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
3867 	if (m == NULL)
3868 		return (NULL);
3869 	m_align(m, len);
3870 	m->m_len = len;
3871 	p = mtod(m, struct sadb_x_policy *);
3872 
3873 	bzero(p, len);
3874 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
3875 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
3876 	p->sadb_x_policy_type = type;
3877 	p->sadb_x_policy_dir = dir;
3878 	p->sadb_x_policy_id = id;
3879 	p->sadb_x_policy_priority = priority;
3880 
3881 	return m;
3882 }
3883 
3884 /* %%% utilities */
3885 /* Take a key message (sadb_key) from the socket and turn it into one
3886  * of the kernel's key structures (seckey).
3887  *
3888  * IN: pointer to the src
3889  * OUT: NULL no more memory
3890  */
3891 struct seckey *
3892 key_dup_keymsg(const struct sadb_key *src, size_t len,
3893     struct malloc_type *type)
3894 {
3895 	struct seckey *dst;
3896 
3897 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
3898 	if (dst != NULL) {
3899 		dst->bits = src->sadb_key_bits;
3900 		dst->key_data = malloc(len, type, M_NOWAIT);
3901 		if (dst->key_data != NULL) {
3902 			bcopy((const char *)(src + 1), dst->key_data, len);
3903 		} else {
3904 			ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3905 			    __func__));
3906 			free(dst, type);
3907 			dst = NULL;
3908 		}
3909 	} else {
3910 		ipseclog((LOG_DEBUG, "%s: No more memory.\n",
3911 		    __func__));
3912 
3913 	}
3914 	return (dst);
3915 }
3916 
3917 /* Take a lifetime message (sadb_lifetime) passed in on a socket and
3918  * turn it into one of the kernel's lifetime structures (seclifetime).
3919  *
3920  * IN: pointer to the destination, source and malloc type
3921  * OUT: NULL, no more memory
3922  */
3923 
3924 static struct seclifetime *
3925 key_dup_lifemsg(const struct sadb_lifetime *src, struct malloc_type *type)
3926 {
3927 	struct seclifetime *dst;
3928 
3929 	dst = malloc(sizeof(*dst), type, M_NOWAIT);
3930 	if (dst == NULL) {
3931 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
3932 		return (NULL);
3933 	}
3934 	dst->allocations = src->sadb_lifetime_allocations;
3935 	dst->bytes = src->sadb_lifetime_bytes;
3936 	dst->addtime = src->sadb_lifetime_addtime;
3937 	dst->usetime = src->sadb_lifetime_usetime;
3938 	return (dst);
3939 }
3940 
3941 /*
3942  * compare two secasindex structure.
3943  * flag can specify to compare 2 saidxes.
3944  * compare two secasindex structure without both mode and reqid.
3945  * don't compare port.
3946  * IN:
3947  *      saidx0: source, it can be in SAD.
3948  *      saidx1: object.
3949  * OUT:
3950  *      1 : equal
3951  *      0 : not equal
3952  */
3953 static int
3954 key_cmpsaidx(const struct secasindex *saidx0, const struct secasindex *saidx1,
3955     int flag)
3956 {
3957 
3958 	/* sanity */
3959 	if (saidx0 == NULL && saidx1 == NULL)
3960 		return 1;
3961 
3962 	if (saidx0 == NULL || saidx1 == NULL)
3963 		return 0;
3964 
3965 	if (saidx0->proto != saidx1->proto)
3966 		return 0;
3967 
3968 	if (flag == CMP_EXACTLY) {
3969 		if (saidx0->mode != saidx1->mode)
3970 			return 0;
3971 		if (saidx0->reqid != saidx1->reqid)
3972 			return 0;
3973 		if (bcmp(&saidx0->src, &saidx1->src,
3974 		    saidx0->src.sa.sa_len) != 0 ||
3975 		    bcmp(&saidx0->dst, &saidx1->dst,
3976 		    saidx0->dst.sa.sa_len) != 0)
3977 			return 0;
3978 	} else {
3979 
3980 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
3981 		if (flag == CMP_MODE_REQID || flag == CMP_REQID) {
3982 			/*
3983 			 * If reqid of SPD is non-zero, unique SA is required.
3984 			 * The result must be of same reqid in this case.
3985 			 */
3986 			if (saidx1->reqid != 0 &&
3987 			    saidx0->reqid != saidx1->reqid)
3988 				return 0;
3989 		}
3990 
3991 		if (flag == CMP_MODE_REQID) {
3992 			if (saidx0->mode != IPSEC_MODE_ANY
3993 			 && saidx0->mode != saidx1->mode)
3994 				return 0;
3995 		}
3996 
3997 		if (key_sockaddrcmp(&saidx0->src.sa, &saidx1->src.sa, 0) != 0)
3998 			return 0;
3999 		if (key_sockaddrcmp(&saidx0->dst.sa, &saidx1->dst.sa, 0) != 0)
4000 			return 0;
4001 	}
4002 
4003 	return 1;
4004 }
4005 
4006 /*
4007  * compare two secindex structure exactly.
4008  * IN:
4009  *	spidx0: source, it is often in SPD.
4010  *	spidx1: object, it is often from PFKEY message.
4011  * OUT:
4012  *	1 : equal
4013  *	0 : not equal
4014  */
4015 static int
4016 key_cmpspidx_exactly(struct secpolicyindex *spidx0,
4017     struct secpolicyindex *spidx1)
4018 {
4019 	/* sanity */
4020 	if (spidx0 == NULL && spidx1 == NULL)
4021 		return 1;
4022 
4023 	if (spidx0 == NULL || spidx1 == NULL)
4024 		return 0;
4025 
4026 	if (spidx0->prefs != spidx1->prefs
4027 	 || spidx0->prefd != spidx1->prefd
4028 	 || spidx0->ul_proto != spidx1->ul_proto)
4029 		return 0;
4030 
4031 	return key_sockaddrcmp(&spidx0->src.sa, &spidx1->src.sa, 1) == 0 &&
4032 	       key_sockaddrcmp(&spidx0->dst.sa, &spidx1->dst.sa, 1) == 0;
4033 }
4034 
4035 /*
4036  * compare two secindex structure with mask.
4037  * IN:
4038  *	spidx0: source, it is often in SPD.
4039  *	spidx1: object, it is often from IP header.
4040  * OUT:
4041  *	1 : equal
4042  *	0 : not equal
4043  */
4044 static int
4045 key_cmpspidx_withmask(struct secpolicyindex *spidx0,
4046     struct secpolicyindex *spidx1)
4047 {
4048 	/* sanity */
4049 	if (spidx0 == NULL && spidx1 == NULL)
4050 		return 1;
4051 
4052 	if (spidx0 == NULL || spidx1 == NULL)
4053 		return 0;
4054 
4055 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4056 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4057 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4058 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
4059 		return 0;
4060 
4061 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4062 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY
4063 	 && spidx0->ul_proto != spidx1->ul_proto)
4064 		return 0;
4065 
4066 	switch (spidx0->src.sa.sa_family) {
4067 	case AF_INET:
4068 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY
4069 		 && spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
4070 			return 0;
4071 		if (!key_bbcmp(&spidx0->src.sin.sin_addr,
4072 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
4073 			return 0;
4074 		break;
4075 	case AF_INET6:
4076 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY
4077 		 && spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
4078 			return 0;
4079 		/*
4080 		 * scope_id check. if sin6_scope_id is 0, we regard it
4081 		 * as a wildcard scope, which matches any scope zone ID.
4082 		 */
4083 		if (spidx0->src.sin6.sin6_scope_id &&
4084 		    spidx1->src.sin6.sin6_scope_id &&
4085 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
4086 			return 0;
4087 		if (!key_bbcmp(&spidx0->src.sin6.sin6_addr,
4088 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
4089 			return 0;
4090 		break;
4091 	default:
4092 		/* XXX */
4093 		if (bcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
4094 			return 0;
4095 		break;
4096 	}
4097 
4098 	switch (spidx0->dst.sa.sa_family) {
4099 	case AF_INET:
4100 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY
4101 		 && spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
4102 			return 0;
4103 		if (!key_bbcmp(&spidx0->dst.sin.sin_addr,
4104 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
4105 			return 0;
4106 		break;
4107 	case AF_INET6:
4108 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY
4109 		 && spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
4110 			return 0;
4111 		/*
4112 		 * scope_id check. if sin6_scope_id is 0, we regard it
4113 		 * as a wildcard scope, which matches any scope zone ID.
4114 		 */
4115 		if (spidx0->dst.sin6.sin6_scope_id &&
4116 		    spidx1->dst.sin6.sin6_scope_id &&
4117 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
4118 			return 0;
4119 		if (!key_bbcmp(&spidx0->dst.sin6.sin6_addr,
4120 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
4121 			return 0;
4122 		break;
4123 	default:
4124 		/* XXX */
4125 		if (bcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
4126 			return 0;
4127 		break;
4128 	}
4129 
4130 	/* XXX Do we check other field ?  e.g. flowinfo */
4131 
4132 	return 1;
4133 }
4134 
4135 #ifdef satosin
4136 #undef satosin
4137 #endif
4138 #define satosin(s) ((const struct sockaddr_in *)s)
4139 #ifdef satosin6
4140 #undef satosin6
4141 #endif
4142 #define satosin6(s) ((const struct sockaddr_in6 *)s)
4143 /* returns 0 on match */
4144 int
4145 key_sockaddrcmp(const struct sockaddr *sa1, const struct sockaddr *sa2,
4146     int port)
4147 {
4148 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4149 		return 1;
4150 
4151 	switch (sa1->sa_family) {
4152 #ifdef INET
4153 	case AF_INET:
4154 		if (sa1->sa_len != sizeof(struct sockaddr_in))
4155 			return 1;
4156 		if (satosin(sa1)->sin_addr.s_addr !=
4157 		    satosin(sa2)->sin_addr.s_addr) {
4158 			return 1;
4159 		}
4160 		if (port && satosin(sa1)->sin_port != satosin(sa2)->sin_port)
4161 			return 1;
4162 		break;
4163 #endif
4164 #ifdef INET6
4165 	case AF_INET6:
4166 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4167 			return 1;	/*EINVAL*/
4168 		if (satosin6(sa1)->sin6_scope_id !=
4169 		    satosin6(sa2)->sin6_scope_id) {
4170 			return 1;
4171 		}
4172 		if (!IN6_ARE_ADDR_EQUAL(&satosin6(sa1)->sin6_addr,
4173 		    &satosin6(sa2)->sin6_addr)) {
4174 			return 1;
4175 		}
4176 		if (port &&
4177 		    satosin6(sa1)->sin6_port != satosin6(sa2)->sin6_port) {
4178 			return 1;
4179 		}
4180 		break;
4181 #endif
4182 	default:
4183 		if (bcmp(sa1, sa2, sa1->sa_len) != 0)
4184 			return 1;
4185 		break;
4186 	}
4187 
4188 	return 0;
4189 }
4190 
4191 /* returns 0 on match */
4192 int
4193 key_sockaddrcmp_withmask(const struct sockaddr *sa1,
4194     const struct sockaddr *sa2, size_t mask)
4195 {
4196 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len)
4197 		return (1);
4198 
4199 	switch (sa1->sa_family) {
4200 #ifdef INET
4201 	case AF_INET:
4202 		return (!key_bbcmp(&satosin(sa1)->sin_addr,
4203 		    &satosin(sa2)->sin_addr, mask));
4204 #endif
4205 #ifdef INET6
4206 	case AF_INET6:
4207 		if (satosin6(sa1)->sin6_scope_id !=
4208 		    satosin6(sa2)->sin6_scope_id)
4209 			return (1);
4210 		return (!key_bbcmp(&satosin6(sa1)->sin6_addr,
4211 		    &satosin6(sa2)->sin6_addr, mask));
4212 #endif
4213 	}
4214 	return (1);
4215 }
4216 #undef satosin
4217 #undef satosin6
4218 
4219 /*
4220  * compare two buffers with mask.
4221  * IN:
4222  *	addr1: source
4223  *	addr2: object
4224  *	bits:  Number of bits to compare
4225  * OUT:
4226  *	1 : equal
4227  *	0 : not equal
4228  */
4229 static int
4230 key_bbcmp(const void *a1, const void *a2, u_int bits)
4231 {
4232 	const unsigned char *p1 = a1;
4233 	const unsigned char *p2 = a2;
4234 
4235 	/* XXX: This could be considerably faster if we compare a word
4236 	 * at a time, but it is complicated on LSB Endian machines */
4237 
4238 	/* Handle null pointers */
4239 	if (p1 == NULL || p2 == NULL)
4240 		return (p1 == p2);
4241 
4242 	while (bits >= 8) {
4243 		if (*p1++ != *p2++)
4244 			return 0;
4245 		bits -= 8;
4246 	}
4247 
4248 	if (bits > 0) {
4249 		u_int8_t mask = ~((1<<(8-bits))-1);
4250 		if ((*p1 & mask) != (*p2 & mask))
4251 			return 0;
4252 	}
4253 	return 1;	/* Match! */
4254 }
4255 
4256 static void
4257 key_flush_spd(time_t now)
4258 {
4259 	SPTREE_RLOCK_TRACKER;
4260 	struct secpolicy_list drainq;
4261 	struct secpolicy *sp, *nextsp;
4262 	u_int dir;
4263 
4264 	LIST_INIT(&drainq);
4265 	SPTREE_RLOCK();
4266 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4267 		TAILQ_FOREACH(sp, &V_sptree[dir], chain) {
4268 			if (sp->lifetime == 0 && sp->validtime == 0)
4269 				continue;
4270 			if ((sp->lifetime &&
4271 			    now - sp->created > sp->lifetime) ||
4272 			    (sp->validtime &&
4273 			    now - sp->lastused > sp->validtime)) {
4274 				/* Hold extra reference to send SPDEXPIRE */
4275 				SP_ADDREF(sp);
4276 				LIST_INSERT_HEAD(&drainq, sp, drainq);
4277 			}
4278 		}
4279 	}
4280 	SPTREE_RUNLOCK();
4281 	if (LIST_EMPTY(&drainq))
4282 		return;
4283 
4284 	SPTREE_WLOCK();
4285 	sp = LIST_FIRST(&drainq);
4286 	while (sp != NULL) {
4287 		nextsp = LIST_NEXT(sp, drainq);
4288 		/* Check that SP is still linked */
4289 		if (sp->state != IPSEC_SPSTATE_ALIVE) {
4290 			LIST_REMOVE(sp, drainq);
4291 			key_freesp(&sp); /* release extra reference */
4292 			sp = nextsp;
4293 			continue;
4294 		}
4295 		TAILQ_REMOVE(&V_sptree[sp->spidx.dir], sp, chain);
4296 		LIST_REMOVE(sp, idhash);
4297 		sp->state = IPSEC_SPSTATE_DEAD;
4298 		sp = nextsp;
4299 	}
4300 	V_sp_genid++;
4301 	SPTREE_WUNLOCK();
4302 
4303 	sp = LIST_FIRST(&drainq);
4304 	while (sp != NULL) {
4305 		nextsp = LIST_NEXT(sp, drainq);
4306 		key_spdexpire(sp);
4307 		key_freesp(&sp); /* release extra reference */
4308 		key_freesp(&sp); /* release last reference */
4309 		sp = nextsp;
4310 	}
4311 }
4312 
4313 static void
4314 key_flush_sad(time_t now)
4315 {
4316 	SAHTREE_RLOCK_TRACKER;
4317 	struct secashead_list emptyq;
4318 	struct secasvar_list drainq, hexpireq, sexpireq, freeq;
4319 	struct secashead *sah, *nextsah;
4320 	struct secasvar *sav, *nextsav;
4321 
4322 	LIST_INIT(&drainq);
4323 	LIST_INIT(&hexpireq);
4324 	LIST_INIT(&sexpireq);
4325 	LIST_INIT(&emptyq);
4326 
4327 	SAHTREE_RLOCK();
4328 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
4329 		/* Check for empty SAH */
4330 		if (TAILQ_EMPTY(&sah->savtree_larval) &&
4331 		    TAILQ_EMPTY(&sah->savtree_alive)) {
4332 			SAH_ADDREF(sah);
4333 			LIST_INSERT_HEAD(&emptyq, sah, drainq);
4334 			continue;
4335 		}
4336 		/* Add all stale LARVAL SAs into drainq */
4337 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
4338 			if (now - sav->created < V_key_larval_lifetime)
4339 				continue;
4340 			SAV_ADDREF(sav);
4341 			LIST_INSERT_HEAD(&drainq, sav, drainq);
4342 		}
4343 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
4344 			/* lifetimes aren't specified */
4345 			if (sav->lft_h == NULL)
4346 				continue;
4347 			SECASVAR_LOCK(sav);
4348 			/*
4349 			 * Check again with lock held, because it may
4350 			 * be updated by SADB_UPDATE.
4351 			 */
4352 			if (sav->lft_h == NULL) {
4353 				SECASVAR_UNLOCK(sav);
4354 				continue;
4355 			}
4356 			/*
4357 			 * RFC 2367:
4358 			 * HARD lifetimes MUST take precedence over SOFT
4359 			 * lifetimes, meaning if the HARD and SOFT lifetimes
4360 			 * are the same, the HARD lifetime will appear on the
4361 			 * EXPIRE message.
4362 			 */
4363 			/* check HARD lifetime */
4364 			if ((sav->lft_h->addtime != 0 &&
4365 			    now - sav->created > sav->lft_h->addtime) ||
4366 			    (sav->lft_h->usetime != 0 && sav->firstused &&
4367 			    now - sav->firstused > sav->lft_h->usetime) ||
4368 			    (sav->lft_h->bytes != 0 && counter_u64_fetch(
4369 			        sav->lft_c_bytes) > sav->lft_h->bytes)) {
4370 				SECASVAR_UNLOCK(sav);
4371 				SAV_ADDREF(sav);
4372 				LIST_INSERT_HEAD(&hexpireq, sav, drainq);
4373 				continue;
4374 			}
4375 			/* check SOFT lifetime (only for MATURE SAs) */
4376 			if (sav->state == SADB_SASTATE_MATURE && (
4377 			    (sav->lft_s->addtime != 0 &&
4378 			    now - sav->created > sav->lft_s->addtime) ||
4379 			    (sav->lft_s->usetime != 0 && sav->firstused &&
4380 			    now - sav->firstused > sav->lft_s->usetime) ||
4381 			    (sav->lft_s->bytes != 0 && counter_u64_fetch(
4382 				sav->lft_c_bytes) > sav->lft_s->bytes))) {
4383 				SECASVAR_UNLOCK(sav);
4384 				SAV_ADDREF(sav);
4385 				LIST_INSERT_HEAD(&sexpireq, sav, drainq);
4386 				continue;
4387 			}
4388 			SECASVAR_UNLOCK(sav);
4389 		}
4390 	}
4391 	SAHTREE_RUNLOCK();
4392 
4393 	if (LIST_EMPTY(&emptyq) && LIST_EMPTY(&drainq) &&
4394 	    LIST_EMPTY(&hexpireq) && LIST_EMPTY(&sexpireq))
4395 		return;
4396 
4397 	LIST_INIT(&freeq);
4398 	SAHTREE_WLOCK();
4399 	/* Unlink stale LARVAL SAs */
4400 	sav = LIST_FIRST(&drainq);
4401 	while (sav != NULL) {
4402 		nextsav = LIST_NEXT(sav, drainq);
4403 		/* Check that SA is still LARVAL */
4404 		if (sav->state != SADB_SASTATE_LARVAL) {
4405 			LIST_REMOVE(sav, drainq);
4406 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4407 			sav = nextsav;
4408 			continue;
4409 		}
4410 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
4411 		LIST_REMOVE(sav, spihash);
4412 		sav->state = SADB_SASTATE_DEAD;
4413 		sav = nextsav;
4414 	}
4415 	/* Unlink all SAs with expired HARD lifetime */
4416 	sav = LIST_FIRST(&hexpireq);
4417 	while (sav != NULL) {
4418 		nextsav = LIST_NEXT(sav, drainq);
4419 		/* Check that SA is not unlinked */
4420 		if (sav->state == SADB_SASTATE_DEAD) {
4421 			LIST_REMOVE(sav, drainq);
4422 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4423 			sav = nextsav;
4424 			continue;
4425 		}
4426 		TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
4427 		LIST_REMOVE(sav, spihash);
4428 		sav->state = SADB_SASTATE_DEAD;
4429 		sav = nextsav;
4430 	}
4431 	/* Mark all SAs with expired SOFT lifetime as DYING */
4432 	sav = LIST_FIRST(&sexpireq);
4433 	while (sav != NULL) {
4434 		nextsav = LIST_NEXT(sav, drainq);
4435 		/* Check that SA is not unlinked */
4436 		if (sav->state == SADB_SASTATE_DEAD) {
4437 			LIST_REMOVE(sav, drainq);
4438 			LIST_INSERT_HEAD(&freeq, sav, drainq);
4439 			sav = nextsav;
4440 			continue;
4441 		}
4442 		/*
4443 		 * NOTE: this doesn't change SA order in the chain.
4444 		 */
4445 		sav->state = SADB_SASTATE_DYING;
4446 		sav = nextsav;
4447 	}
4448 	/* Unlink empty SAHs */
4449 	sah = LIST_FIRST(&emptyq);
4450 	while (sah != NULL) {
4451 		nextsah = LIST_NEXT(sah, drainq);
4452 		/* Check that SAH is still empty and not unlinked */
4453 		if (sah->state == SADB_SASTATE_DEAD ||
4454 		    !TAILQ_EMPTY(&sah->savtree_larval) ||
4455 		    !TAILQ_EMPTY(&sah->savtree_alive)) {
4456 			LIST_REMOVE(sah, drainq);
4457 			key_freesah(&sah); /* release extra reference */
4458 			sah = nextsah;
4459 			continue;
4460 		}
4461 		TAILQ_REMOVE(&V_sahtree, sah, chain);
4462 		LIST_REMOVE(sah, addrhash);
4463 		sah->state = SADB_SASTATE_DEAD;
4464 		sah = nextsah;
4465 	}
4466 	SAHTREE_WUNLOCK();
4467 
4468 	/* Send SPDEXPIRE messages */
4469 	sav = LIST_FIRST(&hexpireq);
4470 	while (sav != NULL) {
4471 		nextsav = LIST_NEXT(sav, drainq);
4472 		key_expire(sav, 1);
4473 		key_freesah(&sav->sah); /* release reference from SAV */
4474 		key_freesav(&sav); /* release extra reference */
4475 		key_freesav(&sav); /* release last reference */
4476 		sav = nextsav;
4477 	}
4478 	sav = LIST_FIRST(&sexpireq);
4479 	while (sav != NULL) {
4480 		nextsav = LIST_NEXT(sav, drainq);
4481 		key_expire(sav, 0);
4482 		key_freesav(&sav); /* release extra reference */
4483 		sav = nextsav;
4484 	}
4485 	/* Free stale LARVAL SAs */
4486 	sav = LIST_FIRST(&drainq);
4487 	while (sav != NULL) {
4488 		nextsav = LIST_NEXT(sav, drainq);
4489 		key_freesah(&sav->sah); /* release reference from SAV */
4490 		key_freesav(&sav); /* release extra reference */
4491 		key_freesav(&sav); /* release last reference */
4492 		sav = nextsav;
4493 	}
4494 	/* Free SAs that were unlinked/changed by someone else */
4495 	sav = LIST_FIRST(&freeq);
4496 	while (sav != NULL) {
4497 		nextsav = LIST_NEXT(sav, drainq);
4498 		key_freesav(&sav); /* release extra reference */
4499 		sav = nextsav;
4500 	}
4501 	/* Free empty SAH */
4502 	sah = LIST_FIRST(&emptyq);
4503 	while (sah != NULL) {
4504 		nextsah = LIST_NEXT(sah, drainq);
4505 		key_freesah(&sah); /* release extra reference */
4506 		key_freesah(&sah); /* release last reference */
4507 		sah = nextsah;
4508 	}
4509 }
4510 
4511 static void
4512 key_flush_acq(time_t now)
4513 {
4514 	struct secacq *acq, *nextacq;
4515 
4516 	/* ACQ tree */
4517 	ACQ_LOCK();
4518 	acq = LIST_FIRST(&V_acqtree);
4519 	while (acq != NULL) {
4520 		nextacq = LIST_NEXT(acq, chain);
4521 		if (now - acq->created > V_key_blockacq_lifetime) {
4522 			LIST_REMOVE(acq, chain);
4523 			LIST_REMOVE(acq, addrhash);
4524 			LIST_REMOVE(acq, seqhash);
4525 			free(acq, M_IPSEC_SAQ);
4526 		}
4527 		acq = nextacq;
4528 	}
4529 	ACQ_UNLOCK();
4530 }
4531 
4532 static void
4533 key_flush_spacq(time_t now)
4534 {
4535 	struct secspacq *acq, *nextacq;
4536 
4537 	/* SP ACQ tree */
4538 	SPACQ_LOCK();
4539 	for (acq = LIST_FIRST(&V_spacqtree); acq != NULL; acq = nextacq) {
4540 		nextacq = LIST_NEXT(acq, chain);
4541 		if (now - acq->created > V_key_blockacq_lifetime
4542 		 && __LIST_CHAINED(acq)) {
4543 			LIST_REMOVE(acq, chain);
4544 			free(acq, M_IPSEC_SAQ);
4545 		}
4546 	}
4547 	SPACQ_UNLOCK();
4548 }
4549 
4550 /*
4551  * time handler.
4552  * scanning SPD and SAD to check status for each entries,
4553  * and do to remove or to expire.
4554  * XXX: year 2038 problem may remain.
4555  */
4556 static void
4557 key_timehandler(void *arg)
4558 {
4559 	VNET_ITERATOR_DECL(vnet_iter);
4560 	time_t now = time_second;
4561 
4562 	VNET_LIST_RLOCK_NOSLEEP();
4563 	VNET_FOREACH(vnet_iter) {
4564 		CURVNET_SET(vnet_iter);
4565 		key_flush_spd(now);
4566 		key_flush_sad(now);
4567 		key_flush_acq(now);
4568 		key_flush_spacq(now);
4569 		CURVNET_RESTORE();
4570 	}
4571 	VNET_LIST_RUNLOCK_NOSLEEP();
4572 
4573 #ifndef IPSEC_DEBUG2
4574 	/* do exchange to tick time !! */
4575 	callout_schedule(&key_timer, hz);
4576 #endif /* IPSEC_DEBUG2 */
4577 }
4578 
4579 u_long
4580 key_random()
4581 {
4582 	u_long value;
4583 
4584 	key_randomfill(&value, sizeof(value));
4585 	return value;
4586 }
4587 
4588 void
4589 key_randomfill(void *p, size_t l)
4590 {
4591 	size_t n;
4592 	u_long v;
4593 	static int warn = 1;
4594 
4595 	n = 0;
4596 	n = (size_t)read_random(p, (u_int)l);
4597 	/* last resort */
4598 	while (n < l) {
4599 		v = random();
4600 		bcopy(&v, (u_int8_t *)p + n,
4601 		    l - n < sizeof(v) ? l - n : sizeof(v));
4602 		n += sizeof(v);
4603 
4604 		if (warn) {
4605 			printf("WARNING: pseudo-random number generator "
4606 			    "used for IPsec processing\n");
4607 			warn = 0;
4608 		}
4609 	}
4610 }
4611 
4612 /*
4613  * map SADB_SATYPE_* to IPPROTO_*.
4614  * if satype == SADB_SATYPE then satype is mapped to ~0.
4615  * OUT:
4616  *	0: invalid satype.
4617  */
4618 static uint8_t
4619 key_satype2proto(uint8_t satype)
4620 {
4621 	switch (satype) {
4622 	case SADB_SATYPE_UNSPEC:
4623 		return IPSEC_PROTO_ANY;
4624 	case SADB_SATYPE_AH:
4625 		return IPPROTO_AH;
4626 	case SADB_SATYPE_ESP:
4627 		return IPPROTO_ESP;
4628 	case SADB_X_SATYPE_IPCOMP:
4629 		return IPPROTO_IPCOMP;
4630 	case SADB_X_SATYPE_TCPSIGNATURE:
4631 		return IPPROTO_TCP;
4632 	default:
4633 		return 0;
4634 	}
4635 	/* NOTREACHED */
4636 }
4637 
4638 /*
4639  * map IPPROTO_* to SADB_SATYPE_*
4640  * OUT:
4641  *	0: invalid protocol type.
4642  */
4643 static uint8_t
4644 key_proto2satype(uint8_t proto)
4645 {
4646 	switch (proto) {
4647 	case IPPROTO_AH:
4648 		return SADB_SATYPE_AH;
4649 	case IPPROTO_ESP:
4650 		return SADB_SATYPE_ESP;
4651 	case IPPROTO_IPCOMP:
4652 		return SADB_X_SATYPE_IPCOMP;
4653 	case IPPROTO_TCP:
4654 		return SADB_X_SATYPE_TCPSIGNATURE;
4655 	default:
4656 		return 0;
4657 	}
4658 	/* NOTREACHED */
4659 }
4660 
4661 /* %%% PF_KEY */
4662 /*
4663  * SADB_GETSPI processing is to receive
4664  *	<base, (SA2), src address, dst address, (SPI range)>
4665  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
4666  * tree with the status of LARVAL, and send
4667  *	<base, SA(*), address(SD)>
4668  * to the IKMPd.
4669  *
4670  * IN:	mhp: pointer to the pointer to each header.
4671  * OUT:	NULL if fail.
4672  *	other if success, return pointer to the message to send.
4673  */
4674 static int
4675 key_getspi(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
4676 {
4677 	struct secasindex saidx;
4678 	struct sadb_address *src0, *dst0;
4679 	struct secasvar *sav;
4680 	uint32_t reqid, spi;
4681 	int error;
4682 	uint8_t mode, proto;
4683 
4684 	IPSEC_ASSERT(so != NULL, ("null socket"));
4685 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
4686 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
4687 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
4688 
4689 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
4690 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)
4691 #ifdef PFKEY_STRICT_CHECKS
4692 	    || SADB_CHECKHDR(mhp, SADB_EXT_SPIRANGE)
4693 #endif
4694 	    ) {
4695 		ipseclog((LOG_DEBUG,
4696 		    "%s: invalid message: missing required header.\n",
4697 		    __func__));
4698 		error = EINVAL;
4699 		goto fail;
4700 	}
4701 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
4702 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)
4703 #ifdef PFKEY_STRICT_CHECKS
4704 	    || SADB_CHECKLEN(mhp, SADB_EXT_SPIRANGE)
4705 #endif
4706 	    ) {
4707 		ipseclog((LOG_DEBUG,
4708 		    "%s: invalid message: wrong header size.\n", __func__));
4709 		error = EINVAL;
4710 		goto fail;
4711 	}
4712 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
4713 		mode = IPSEC_MODE_ANY;
4714 		reqid = 0;
4715 	} else {
4716 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
4717 			ipseclog((LOG_DEBUG,
4718 			    "%s: invalid message: wrong header size.\n",
4719 			    __func__));
4720 			error = EINVAL;
4721 			goto fail;
4722 		}
4723 		mode = ((struct sadb_x_sa2 *)
4724 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
4725 		reqid = ((struct sadb_x_sa2 *)
4726 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
4727 	}
4728 
4729 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
4730 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
4731 
4732 	/* map satype to proto */
4733 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
4734 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
4735 			__func__));
4736 		error = EINVAL;
4737 		goto fail;
4738 	}
4739 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
4740 	    (struct sockaddr *)(dst0 + 1));
4741 	if (error != 0) {
4742 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
4743 		error = EINVAL;
4744 		goto fail;
4745 	}
4746 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
4747 
4748 	/* SPI allocation */
4749 	spi = key_do_getnewspi(
4750 	    (struct sadb_spirange *)mhp->ext[SADB_EXT_SPIRANGE], &saidx);
4751 	if (spi == 0) {
4752 		/*
4753 		 * Requested SPI or SPI range is not available or
4754 		 * already used.
4755 		 */
4756 		error = EEXIST;
4757 		goto fail;
4758 	}
4759 	sav = key_newsav(mhp, &saidx, spi, &error);
4760 	if (sav == NULL)
4761 		goto fail;
4762 
4763 	if (sav->seq != 0) {
4764 		/*
4765 		 * RFC2367:
4766 		 * If the SADB_GETSPI message is in response to a
4767 		 * kernel-generated SADB_ACQUIRE, the sadb_msg_seq
4768 		 * MUST be the same as the SADB_ACQUIRE message.
4769 		 *
4770 		 * XXXAE: However it doesn't definethe behaviour how to
4771 		 * check this and what to do if it doesn't match.
4772 		 * Also what we should do if it matches?
4773 		 *
4774 		 * We can compare saidx used in SADB_ACQUIRE with saidx
4775 		 * used in SADB_GETSPI, but this probably can break
4776 		 * existing software. For now just warn if it doesn't match.
4777 		 *
4778 		 * XXXAE: anyway it looks useless.
4779 		 */
4780 		key_acqdone(&saidx, sav->seq);
4781 	}
4782 	KEYDBG(KEY_STAMP,
4783 	    printf("%s: SA(%p)\n", __func__, sav));
4784 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
4785 
4786     {
4787 	struct mbuf *n, *nn;
4788 	struct sadb_sa *m_sa;
4789 	struct sadb_msg *newmsg;
4790 	int off, len;
4791 
4792 	/* create new sadb_msg to reply. */
4793 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
4794 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
4795 
4796 	MGETHDR(n, M_NOWAIT, MT_DATA);
4797 	if (len > MHLEN) {
4798 		if (!(MCLGET(n, M_NOWAIT))) {
4799 			m_freem(n);
4800 			n = NULL;
4801 		}
4802 	}
4803 	if (!n) {
4804 		error = ENOBUFS;
4805 		goto fail;
4806 	}
4807 
4808 	n->m_len = len;
4809 	n->m_next = NULL;
4810 	off = 0;
4811 
4812 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
4813 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
4814 
4815 	m_sa = (struct sadb_sa *)(mtod(n, caddr_t) + off);
4816 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
4817 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
4818 	m_sa->sadb_sa_spi = spi; /* SPI is already in network byte order */
4819 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
4820 
4821 	IPSEC_ASSERT(off == len,
4822 		("length inconsistency (off %u len %u)", off, len));
4823 
4824 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
4825 	    SADB_EXT_ADDRESS_DST);
4826 	if (!n->m_next) {
4827 		m_freem(n);
4828 		error = ENOBUFS;
4829 		goto fail;
4830 	}
4831 
4832 	if (n->m_len < sizeof(struct sadb_msg)) {
4833 		n = m_pullup(n, sizeof(struct sadb_msg));
4834 		if (n == NULL)
4835 			return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
4836 	}
4837 
4838 	n->m_pkthdr.len = 0;
4839 	for (nn = n; nn; nn = nn->m_next)
4840 		n->m_pkthdr.len += nn->m_len;
4841 
4842 	newmsg = mtod(n, struct sadb_msg *);
4843 	newmsg->sadb_msg_seq = sav->seq;
4844 	newmsg->sadb_msg_errno = 0;
4845 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
4846 
4847 	m_freem(m);
4848 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
4849     }
4850 
4851 fail:
4852 	return (key_senderror(so, m, error));
4853 }
4854 
4855 /*
4856  * allocating new SPI
4857  * called by key_getspi().
4858  * OUT:
4859  *	0:	failure.
4860  *	others: success, SPI in network byte order.
4861  */
4862 static uint32_t
4863 key_do_getnewspi(struct sadb_spirange *spirange, struct secasindex *saidx)
4864 {
4865 	uint32_t min, max, newspi, t;
4866 	int count = V_key_spi_trycnt;
4867 
4868 	/* set spi range to allocate */
4869 	if (spirange != NULL) {
4870 		min = spirange->sadb_spirange_min;
4871 		max = spirange->sadb_spirange_max;
4872 	} else {
4873 		min = V_key_spi_minval;
4874 		max = V_key_spi_maxval;
4875 	}
4876 	/* IPCOMP needs 2-byte SPI */
4877 	if (saidx->proto == IPPROTO_IPCOMP) {
4878 		if (min >= 0x10000)
4879 			min = 0xffff;
4880 		if (max >= 0x10000)
4881 			max = 0xffff;
4882 		if (min > max) {
4883 			t = min; min = max; max = t;
4884 		}
4885 	}
4886 
4887 	if (min == max) {
4888 		if (!key_checkspidup(htonl(min))) {
4889 			ipseclog((LOG_DEBUG, "%s: SPI %u exists already.\n",
4890 			    __func__, min));
4891 			return 0;
4892 		}
4893 
4894 		count--; /* taking one cost. */
4895 		newspi = min;
4896 	} else {
4897 
4898 		/* init SPI */
4899 		newspi = 0;
4900 
4901 		/* when requesting to allocate spi ranged */
4902 		while (count--) {
4903 			/* generate pseudo-random SPI value ranged. */
4904 			newspi = min + (key_random() % (max - min + 1));
4905 			if (!key_checkspidup(htonl(newspi)))
4906 				break;
4907 		}
4908 
4909 		if (count == 0 || newspi == 0) {
4910 			ipseclog((LOG_DEBUG,
4911 			    "%s: failed to allocate SPI.\n", __func__));
4912 			return 0;
4913 		}
4914 	}
4915 
4916 	/* statistics */
4917 	keystat.getspi_count =
4918 	    (keystat.getspi_count + V_key_spi_trycnt - count) / 2;
4919 
4920 	return (htonl(newspi));
4921 }
4922 
4923 /*
4924  * Find TCP-MD5 SA with corresponding secasindex.
4925  * If not found, return NULL and fill SPI with usable value if needed.
4926  */
4927 static struct secasvar *
4928 key_getsav_tcpmd5(struct secasindex *saidx, uint32_t *spi)
4929 {
4930 	SAHTREE_RLOCK_TRACKER;
4931 	struct secashead *sah;
4932 	struct secasvar *sav;
4933 
4934 	IPSEC_ASSERT(saidx->proto == IPPROTO_TCP, ("wrong proto"));
4935 	SAHTREE_RLOCK();
4936 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
4937 		if (sah->saidx.proto != IPPROTO_TCP)
4938 			continue;
4939 		if (!key_sockaddrcmp(&saidx->dst.sa, &sah->saidx.dst.sa, 0))
4940 			break;
4941 	}
4942 	if (sah != NULL) {
4943 		if (V_key_preferred_oldsa)
4944 			sav = TAILQ_LAST(&sah->savtree_alive, secasvar_queue);
4945 		else
4946 			sav = TAILQ_FIRST(&sah->savtree_alive);
4947 		if (sav != NULL) {
4948 			SAV_ADDREF(sav);
4949 			SAHTREE_RUNLOCK();
4950 			return (sav);
4951 		}
4952 	}
4953 	if (spi == NULL) {
4954 		/* No SPI required */
4955 		SAHTREE_RUNLOCK();
4956 		return (NULL);
4957 	}
4958 	/* Check that SPI is unique */
4959 	LIST_FOREACH(sav, SAVHASH_HASH(*spi), spihash) {
4960 		if (sav->spi == *spi)
4961 			break;
4962 	}
4963 	if (sav == NULL) {
4964 		SAHTREE_RUNLOCK();
4965 		/* SPI is already unique */
4966 		return (NULL);
4967 	}
4968 	SAHTREE_RUNLOCK();
4969 	/* XXX: not optimal */
4970 	*spi = key_do_getnewspi(NULL, saidx);
4971 	return (NULL);
4972 }
4973 
4974 static int
4975 key_updateaddresses(struct socket *so, struct mbuf *m,
4976     const struct sadb_msghdr *mhp, struct secasvar *sav,
4977     struct secasindex *saidx)
4978 {
4979 	struct sockaddr *newaddr;
4980 	struct secashead *sah;
4981 	struct secasvar *newsav, *tmp;
4982 	struct mbuf *n;
4983 	int error, isnew;
4984 
4985 	/* Check that we need to change SAH */
4986 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC)) {
4987 		newaddr = (struct sockaddr *)(
4988 		    ((struct sadb_address *)
4989 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_SRC]) + 1);
4990 		bcopy(newaddr, &saidx->src, newaddr->sa_len);
4991 		key_porttosaddr(&saidx->src.sa, 0);
4992 	}
4993 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
4994 		newaddr = (struct sockaddr *)(
4995 		    ((struct sadb_address *)
4996 		    mhp->ext[SADB_X_EXT_NEW_ADDRESS_DST]) + 1);
4997 		bcopy(newaddr, &saidx->dst, newaddr->sa_len);
4998 		key_porttosaddr(&saidx->dst.sa, 0);
4999 	}
5000 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5001 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST)) {
5002 		error = key_checksockaddrs(&saidx->src.sa, &saidx->dst.sa);
5003 		if (error != 0) {
5004 			ipseclog((LOG_DEBUG, "%s: invalid new sockaddr.\n",
5005 			    __func__));
5006 			return (error);
5007 		}
5008 
5009 		sah = key_getsah(saidx);
5010 		if (sah == NULL) {
5011 			/* create a new SA index */
5012 			sah = key_newsah(saidx);
5013 			if (sah == NULL) {
5014 				ipseclog((LOG_DEBUG,
5015 				    "%s: No more memory.\n", __func__));
5016 				return (ENOBUFS);
5017 			}
5018 			isnew = 2; /* SAH is new */
5019 		} else
5020 			isnew = 1; /* existing SAH is referenced */
5021 	} else {
5022 		/*
5023 		 * src and dst addresses are still the same.
5024 		 * Do we want to change NAT-T config?
5025 		 */
5026 		if (sav->sah->saidx.proto != IPPROTO_ESP ||
5027 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5028 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5029 		    SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5030 			ipseclog((LOG_DEBUG,
5031 			    "%s: invalid message: missing required header.\n",
5032 			    __func__));
5033 			return (EINVAL);
5034 		}
5035 		/* We hold reference to SA, thus SAH will be referenced too. */
5036 		sah = sav->sah;
5037 		isnew = 0;
5038 	}
5039 
5040 	newsav = malloc(sizeof(struct secasvar), M_IPSEC_SA,
5041 	    M_NOWAIT | M_ZERO);
5042 	if (newsav == NULL) {
5043 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5044 		error = ENOBUFS;
5045 		goto fail;
5046 	}
5047 
5048 	/* Clone SA's content into newsav */
5049 	SAV_INITREF(newsav);
5050 	bcopy(sav, newsav, offsetof(struct secasvar, chain));
5051 	/*
5052 	 * We create new NAT-T config if it is needed.
5053 	 * Old NAT-T config will be freed by key_cleansav() when
5054 	 * last reference to SA will be released.
5055 	 */
5056 	newsav->natt = NULL;
5057 	newsav->sah = sah;
5058 	newsav->state = SADB_SASTATE_MATURE;
5059 	error = key_setnatt(sav, mhp);
5060 	if (error != 0)
5061 		goto fail;
5062 
5063 	SAHTREE_WLOCK();
5064 	/* Check that SA is still alive */
5065 	if (sav->state == SADB_SASTATE_DEAD) {
5066 		/* SA was unlinked */
5067 		SAHTREE_WUNLOCK();
5068 		error = ESRCH;
5069 		goto fail;
5070 	}
5071 
5072 	/* Unlink SA from SAH and SPI hash */
5073 	IPSEC_ASSERT((sav->flags & SADB_X_EXT_F_CLONED) == 0,
5074 	    ("SA is already cloned"));
5075 	IPSEC_ASSERT(sav->state == SADB_SASTATE_MATURE ||
5076 	    sav->state == SADB_SASTATE_DYING,
5077 	    ("Wrong SA state %u\n", sav->state));
5078 	TAILQ_REMOVE(&sav->sah->savtree_alive, sav, chain);
5079 	LIST_REMOVE(sav, spihash);
5080 	sav->state = SADB_SASTATE_DEAD;
5081 
5082 	/*
5083 	 * Link new SA with SAH. Keep SAs ordered by
5084 	 * create time (newer are first).
5085 	 */
5086 	TAILQ_FOREACH(tmp, &sah->savtree_alive, chain) {
5087 		if (newsav->created > tmp->created) {
5088 			TAILQ_INSERT_BEFORE(tmp, newsav, chain);
5089 			break;
5090 		}
5091 	}
5092 	if (tmp == NULL)
5093 		TAILQ_INSERT_TAIL(&sah->savtree_alive, newsav, chain);
5094 
5095 	/* Add new SA into SPI hash. */
5096 	LIST_INSERT_HEAD(SAVHASH_HASH(newsav->spi), newsav, spihash);
5097 
5098 	/* Add new SAH into SADB. */
5099 	if (isnew == 2) {
5100 		TAILQ_INSERT_HEAD(&V_sahtree, sah, chain);
5101 		LIST_INSERT_HEAD(SAHADDRHASH_HASH(saidx), sah, addrhash);
5102 		sah->state = SADB_SASTATE_MATURE;
5103 		SAH_ADDREF(sah); /* newsav references new SAH */
5104 	}
5105 	/*
5106 	 * isnew == 1 -> @sah was referenced by key_getsah().
5107 	 * isnew == 0 -> we use the same @sah, that was used by @sav,
5108 	 *	and we use its reference for @newsav.
5109 	 */
5110 	SECASVAR_LOCK(sav);
5111 	/* XXX: replace cntr with pointer? */
5112 	newsav->cntr = sav->cntr;
5113 	sav->flags |= SADB_X_EXT_F_CLONED;
5114 	SECASVAR_UNLOCK(sav);
5115 
5116 	SAHTREE_WUNLOCK();
5117 
5118 	KEYDBG(KEY_STAMP,
5119 	    printf("%s: SA(%p) cloned into SA(%p)\n",
5120 	    __func__, sav, newsav));
5121 	KEYDBG(KEY_DATA, kdebug_secasv(newsav));
5122 
5123 	key_freesav(&sav); /* release last reference */
5124 
5125 	/* set msg buf from mhp */
5126 	n = key_getmsgbuf_x1(m, mhp);
5127 	if (n == NULL) {
5128 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5129 		return (ENOBUFS);
5130 	}
5131 	m_freem(m);
5132 	key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5133 	return (0);
5134 fail:
5135 	if (isnew != 0)
5136 		key_freesah(&sah);
5137 	if (newsav != NULL) {
5138 		if (newsav->natt != NULL)
5139 			free(newsav->natt, M_IPSEC_MISC);
5140 		free(newsav, M_IPSEC_SA);
5141 	}
5142 	return (error);
5143 }
5144 
5145 /*
5146  * SADB_UPDATE processing
5147  * receive
5148  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5149  *       key(AE), (identity(SD),) (sensitivity)>
5150  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5151  * and send
5152  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5153  *       (identity(SD),) (sensitivity)>
5154  * to the ikmpd.
5155  *
5156  * m will always be freed.
5157  */
5158 static int
5159 key_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5160 {
5161 	struct secasindex saidx;
5162 	struct sadb_address *src0, *dst0;
5163 	struct sadb_sa *sa0;
5164 	struct secasvar *sav;
5165 	uint32_t reqid;
5166 	int error;
5167 	uint8_t mode, proto;
5168 
5169 	IPSEC_ASSERT(so != NULL, ("null socket"));
5170 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5171 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5172 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5173 
5174 	/* map satype to proto */
5175 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5176 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5177 		    __func__));
5178 		return key_senderror(so, m, EINVAL);
5179 	}
5180 
5181 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5182 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5183 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5184 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5185 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5186 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5187 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5188 		ipseclog((LOG_DEBUG,
5189 		    "%s: invalid message: missing required header.\n",
5190 		    __func__));
5191 		return key_senderror(so, m, EINVAL);
5192 	}
5193 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5194 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5195 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5196 		ipseclog((LOG_DEBUG,
5197 		    "%s: invalid message: wrong header size.\n", __func__));
5198 		return key_senderror(so, m, EINVAL);
5199 	}
5200 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5201 		mode = IPSEC_MODE_ANY;
5202 		reqid = 0;
5203 	} else {
5204 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5205 			ipseclog((LOG_DEBUG,
5206 			    "%s: invalid message: wrong header size.\n",
5207 			    __func__));
5208 			return key_senderror(so, m, EINVAL);
5209 		}
5210 		mode = ((struct sadb_x_sa2 *)
5211 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5212 		reqid = ((struct sadb_x_sa2 *)
5213 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5214 	}
5215 
5216 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5217 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5218 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5219 
5220 	/*
5221 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5222 	 * SADB_UPDATE message.
5223 	 */
5224 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5225 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5226 #ifdef PFKEY_STRICT_CHECKS
5227 		return key_senderror(so, m, EINVAL);
5228 #endif
5229 	}
5230 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5231 	    (struct sockaddr *)(dst0 + 1));
5232 	if (error != 0) {
5233 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5234 		return key_senderror(so, m, error);
5235 	}
5236 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5237 	sav = key_getsavbyspi(sa0->sadb_sa_spi);
5238 	if (sav == NULL) {
5239 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u\n",
5240 		    __func__, ntohl(sa0->sadb_sa_spi)));
5241 		return key_senderror(so, m, EINVAL);
5242 	}
5243 	/*
5244 	 * Check that SADB_UPDATE issued by the same process that did
5245 	 * SADB_GETSPI or SADB_ADD.
5246 	 */
5247 	if (sav->pid != mhp->msg->sadb_msg_pid) {
5248 		ipseclog((LOG_DEBUG,
5249 		    "%s: pid mismatched (SPI %u, pid %u vs. %u)\n", __func__,
5250 		    ntohl(sav->spi), sav->pid, mhp->msg->sadb_msg_pid));
5251 		key_freesav(&sav);
5252 		return key_senderror(so, m, EINVAL);
5253 	}
5254 	/* saidx should match with SA. */
5255 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_MODE_REQID) == 0) {
5256 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u",
5257 		    __func__, ntohl(sav->spi)));
5258 		key_freesav(&sav);
5259 		return key_senderror(so, m, ESRCH);
5260 	}
5261 
5262 	if (sav->state == SADB_SASTATE_LARVAL) {
5263 		if ((mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5264 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT)) ||
5265 		    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5266 		    SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH))) {
5267 			ipseclog((LOG_DEBUG,
5268 			    "%s: invalid message: missing required header.\n",
5269 			    __func__));
5270 			key_freesav(&sav);
5271 			return key_senderror(so, m, EINVAL);
5272 		}
5273 		/*
5274 		 * We can set any values except src, dst and SPI.
5275 		 */
5276 		error = key_setsaval(sav, mhp);
5277 		if (error != 0) {
5278 			key_freesav(&sav);
5279 			return (key_senderror(so, m, error));
5280 		}
5281 		/* Change SA state to MATURE */
5282 		SAHTREE_WLOCK();
5283 		if (sav->state != SADB_SASTATE_LARVAL) {
5284 			/* SA was deleted or another thread made it MATURE. */
5285 			SAHTREE_WUNLOCK();
5286 			key_freesav(&sav);
5287 			return (key_senderror(so, m, ESRCH));
5288 		}
5289 		/*
5290 		 * NOTE: we keep SAs in savtree_alive ordered by created
5291 		 * time. When SA's state changed from LARVAL to MATURE,
5292 		 * we update its created time in key_setsaval() and move
5293 		 * it into head of savtree_alive.
5294 		 */
5295 		TAILQ_REMOVE(&sav->sah->savtree_larval, sav, chain);
5296 		TAILQ_INSERT_HEAD(&sav->sah->savtree_alive, sav, chain);
5297 		sav->state = SADB_SASTATE_MATURE;
5298 		SAHTREE_WUNLOCK();
5299 	} else {
5300 		/*
5301 		 * For DYING and MATURE SA we can change only state
5302 		 * and lifetimes. Report EINVAL if something else attempted
5303 		 * to change.
5304 		 */
5305 		if (!SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5306 		    !SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH)) {
5307 			key_freesav(&sav);
5308 			return (key_senderror(so, m, EINVAL));
5309 		}
5310 		error = key_updatelifetimes(sav, mhp);
5311 		if (error != 0) {
5312 			key_freesav(&sav);
5313 			return (key_senderror(so, m, error));
5314 		}
5315 		/*
5316 		 * This is FreeBSD extension to RFC2367.
5317 		 * IKEd can specify SADB_X_EXT_NEW_ADDRESS_SRC and/or
5318 		 * SADB_X_EXT_NEW_ADDRESS_DST when it wants to change
5319 		 * SA addresses (for example to implement MOBIKE protocol
5320 		 * as described in RFC4555). Also we allow to change
5321 		 * NAT-T config.
5322 		 */
5323 		if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_SRC) ||
5324 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NEW_ADDRESS_DST) ||
5325 		    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5326 		    sav->natt != NULL) {
5327 			error = key_updateaddresses(so, m, mhp, sav, &saidx);
5328 			key_freesav(&sav);
5329 			if (error != 0)
5330 				return (key_senderror(so, m, error));
5331 			return (0);
5332 		}
5333 		/* Check that SA is still alive */
5334 		SAHTREE_WLOCK();
5335 		if (sav->state == SADB_SASTATE_DEAD) {
5336 			/* SA was unlinked */
5337 			SAHTREE_WUNLOCK();
5338 			key_freesav(&sav);
5339 			return (key_senderror(so, m, ESRCH));
5340 		}
5341 		/*
5342 		 * NOTE: there is possible state moving from DYING to MATURE,
5343 		 * but this doesn't change created time, so we won't reorder
5344 		 * this SA.
5345 		 */
5346 		sav->state = SADB_SASTATE_MATURE;
5347 		SAHTREE_WUNLOCK();
5348 	}
5349 	KEYDBG(KEY_STAMP,
5350 	    printf("%s: SA(%p)\n", __func__, sav));
5351 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5352 	key_freesav(&sav);
5353 
5354     {
5355 	struct mbuf *n;
5356 
5357 	/* set msg buf from mhp */
5358 	n = key_getmsgbuf_x1(m, mhp);
5359 	if (n == NULL) {
5360 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5361 		return key_senderror(so, m, ENOBUFS);
5362 	}
5363 
5364 	m_freem(m);
5365 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5366     }
5367 }
5368 
5369 /*
5370  * SADB_ADD processing
5371  * add an entry to SA database, when received
5372  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5373  *       key(AE), (identity(SD),) (sensitivity)>
5374  * from the ikmpd,
5375  * and send
5376  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5377  *       (identity(SD),) (sensitivity)>
5378  * to the ikmpd.
5379  *
5380  * IGNORE identity and sensitivity messages.
5381  *
5382  * m will always be freed.
5383  */
5384 static int
5385 key_add(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5386 {
5387 	struct secasindex saidx;
5388 	struct sadb_address *src0, *dst0;
5389 	struct sadb_sa *sa0;
5390 	struct secasvar *sav;
5391 	uint32_t reqid, spi;
5392 	uint8_t mode, proto;
5393 	int error;
5394 
5395 	IPSEC_ASSERT(so != NULL, ("null socket"));
5396 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5397 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5398 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5399 
5400 	/* map satype to proto */
5401 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5402 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5403 		    __func__));
5404 		return key_senderror(so, m, EINVAL);
5405 	}
5406 
5407 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
5408 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5409 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5410 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP && (
5411 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_ENCRYPT) ||
5412 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_ENCRYPT))) ||
5413 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH && (
5414 		SADB_CHECKHDR(mhp, SADB_EXT_KEY_AUTH) ||
5415 		SADB_CHECKLEN(mhp, SADB_EXT_KEY_AUTH))) ||
5416 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD) &&
5417 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT)) ||
5418 	    (SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_SOFT) &&
5419 		!SADB_CHECKHDR(mhp, SADB_EXT_LIFETIME_HARD))) {
5420 		ipseclog((LOG_DEBUG,
5421 		    "%s: invalid message: missing required header.\n",
5422 		    __func__));
5423 		return key_senderror(so, m, EINVAL);
5424 	}
5425 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
5426 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5427 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5428 		ipseclog((LOG_DEBUG,
5429 		    "%s: invalid message: wrong header size.\n", __func__));
5430 		return key_senderror(so, m, EINVAL);
5431 	}
5432 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
5433 		mode = IPSEC_MODE_ANY;
5434 		reqid = 0;
5435 	} else {
5436 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
5437 			ipseclog((LOG_DEBUG,
5438 			    "%s: invalid message: wrong header size.\n",
5439 			    __func__));
5440 			return key_senderror(so, m, EINVAL);
5441 		}
5442 		mode = ((struct sadb_x_sa2 *)
5443 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
5444 		reqid = ((struct sadb_x_sa2 *)
5445 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
5446 	}
5447 
5448 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5449 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
5450 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
5451 
5452 	/*
5453 	 * Only SADB_SASTATE_MATURE SAs may be submitted in an
5454 	 * SADB_ADD message.
5455 	 */
5456 	if (sa0->sadb_sa_state != SADB_SASTATE_MATURE) {
5457 		ipseclog((LOG_DEBUG, "%s: invalid state.\n", __func__));
5458 #ifdef PFKEY_STRICT_CHECKS
5459 		return key_senderror(so, m, EINVAL);
5460 #endif
5461 	}
5462 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
5463 	    (struct sockaddr *)(dst0 + 1));
5464 	if (error != 0) {
5465 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5466 		return key_senderror(so, m, error);
5467 	}
5468 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
5469 	spi = sa0->sadb_sa_spi;
5470 	/*
5471 	 * For TCP-MD5 SAs we don't use SPI. Check the uniqueness using
5472 	 * secasindex.
5473 	 * XXXAE: IPComp seems also doesn't use SPI.
5474 	 */
5475 	if (proto == IPPROTO_TCP) {
5476 		sav = key_getsav_tcpmd5(&saidx, &spi);
5477 		if (sav == NULL && spi == 0) {
5478 			/* Failed to allocate SPI */
5479 			ipseclog((LOG_DEBUG, "%s: SA already exists.\n",
5480 			    __func__));
5481 			return key_senderror(so, m, EEXIST);
5482 		}
5483 		/* XXX: SPI that we report back can have another value */
5484 	} else {
5485 		/* We can create new SA only if SPI is different. */
5486 		sav = key_getsavbyspi(spi);
5487 	}
5488 	if (sav != NULL) {
5489 		key_freesav(&sav);
5490 		ipseclog((LOG_DEBUG, "%s: SA already exists.\n", __func__));
5491 		return key_senderror(so, m, EEXIST);
5492 	}
5493 
5494 	sav = key_newsav(mhp, &saidx, spi, &error);
5495 	if (sav == NULL)
5496 		return key_senderror(so, m, error);
5497 	KEYDBG(KEY_STAMP,
5498 	    printf("%s: return SA(%p)\n", __func__, sav));
5499 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5500 	/*
5501 	 * If SADB_ADD was in response to SADB_ACQUIRE, we need to schedule
5502 	 * ACQ for deletion.
5503 	 */
5504 	if (sav->seq != 0)
5505 		key_acqdone(&saidx, sav->seq);
5506 
5507     {
5508 	/*
5509 	 * Don't call key_freesav() on error here, as we would like to
5510 	 * keep the SA in the database.
5511 	 */
5512 	struct mbuf *n;
5513 
5514 	/* set msg buf from mhp */
5515 	n = key_getmsgbuf_x1(m, mhp);
5516 	if (n == NULL) {
5517 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5518 		return key_senderror(so, m, ENOBUFS);
5519 	}
5520 
5521 	m_freem(m);
5522 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5523     }
5524 }
5525 
5526 /*
5527  * NAT-T support.
5528  * IKEd may request the use ESP in UDP encapsulation when it detects the
5529  * presence of NAT. It uses NAT-T extension headers for such SAs to specify
5530  * parameters needed for encapsulation and decapsulation. These PF_KEY
5531  * extension headers are not standardized, so this comment addresses our
5532  * implementation.
5533  * SADB_X_EXT_NAT_T_TYPE specifies type of encapsulation, we support only
5534  * UDP_ENCAP_ESPINUDP as described in RFC3948.
5535  * SADB_X_EXT_NAT_T_SPORT/DPORT specifies source and destination ports for
5536  * UDP header. We use these ports in UDP encapsulation procedure, also we
5537  * can check them in UDP decapsulation procedure.
5538  * SADB_X_EXT_NAT_T_OA[IR] specifies original address of initiator or
5539  * responder. These addresses can be used for transport mode to adjust
5540  * checksum after decapsulation and decryption. Since original IP addresses
5541  * used by peer usually different (we detected presence of NAT), TCP/UDP
5542  * pseudo header checksum and IP header checksum was calculated using original
5543  * addresses. After decapsulation and decryption we need to adjust checksum
5544  * to have correct datagram.
5545  *
5546  * We expect presence of NAT-T extension headers only in SADB_ADD and
5547  * SADB_UPDATE messages. We report NAT-T extension headers in replies
5548  * to SADB_ADD, SADB_UPDATE, SADB_GET, and SADB_DUMP messages.
5549  */
5550 static int
5551 key_setnatt(struct secasvar *sav, const struct sadb_msghdr *mhp)
5552 {
5553 	struct sadb_x_nat_t_port *port;
5554 	struct sadb_x_nat_t_type *type;
5555 	struct sadb_address *oai, *oar;
5556 	struct sockaddr *sa;
5557 	uint32_t addr;
5558 	uint16_t cksum;
5559 
5560 	IPSEC_ASSERT(sav->natt == NULL, ("natt is already initialized"));
5561 	/*
5562 	 * Ignore NAT-T headers if sproto isn't ESP.
5563 	 */
5564 	if (sav->sah->saidx.proto != IPPROTO_ESP)
5565 		return (0);
5566 
5567 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_TYPE) &&
5568 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_SPORT) &&
5569 	    !SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5570 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_TYPE) ||
5571 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_SPORT) ||
5572 		    SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_DPORT)) {
5573 			ipseclog((LOG_DEBUG,
5574 			    "%s: invalid message: wrong header size.\n",
5575 			    __func__));
5576 			return (EINVAL);
5577 		}
5578 	} else
5579 		return (0);
5580 
5581 	type = (struct sadb_x_nat_t_type *)mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5582 	if (type->sadb_x_nat_t_type_type != UDP_ENCAP_ESPINUDP) {
5583 		ipseclog((LOG_DEBUG, "%s: unsupported NAT-T type %u.\n",
5584 		    __func__, type->sadb_x_nat_t_type_type));
5585 		return (EINVAL);
5586 	}
5587 	/*
5588 	 * Allocate storage for NAT-T config.
5589 	 * On error it will be released by key_cleansav().
5590 	 */
5591 	sav->natt = malloc(sizeof(struct secnatt), M_IPSEC_MISC,
5592 	    M_NOWAIT | M_ZERO);
5593 	if (sav->natt == NULL) {
5594 		PFKEYSTAT_INC(in_nomem);
5595 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5596 		return (ENOBUFS);
5597 	}
5598 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5599 	if (port->sadb_x_nat_t_port_port == 0) {
5600 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T sport specified.\n",
5601 		    __func__));
5602 		return (EINVAL);
5603 	}
5604 	sav->natt->sport = port->sadb_x_nat_t_port_port;
5605 	port = (struct sadb_x_nat_t_port *)mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5606 	if (port->sadb_x_nat_t_port_port == 0) {
5607 		ipseclog((LOG_DEBUG, "%s: invalid NAT-T dport specified.\n",
5608 		    __func__));
5609 		return (EINVAL);
5610 	}
5611 	sav->natt->dport = port->sadb_x_nat_t_port_port;
5612 
5613 	/*
5614 	 * SADB_X_EXT_NAT_T_OAI and SADB_X_EXT_NAT_T_OAR are optional
5615 	 * and needed only for transport mode IPsec.
5616 	 * Usually NAT translates only one address, but it is possible,
5617 	 * that both addresses could be translated.
5618 	 * NOTE: Value of SADB_X_EXT_NAT_T_OAI is equal to SADB_X_EXT_NAT_T_OA.
5619 	 */
5620 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAI)) {
5621 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAI)) {
5622 			ipseclog((LOG_DEBUG,
5623 			    "%s: invalid message: wrong header size.\n",
5624 			    __func__));
5625 			return (EINVAL);
5626 		}
5627 		oai = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAI];
5628 	} else
5629 		oai = NULL;
5630 	if (!SADB_CHECKHDR(mhp, SADB_X_EXT_NAT_T_OAR)) {
5631 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_NAT_T_OAR)) {
5632 			ipseclog((LOG_DEBUG,
5633 			    "%s: invalid message: wrong header size.\n",
5634 			    __func__));
5635 			return (EINVAL);
5636 		}
5637 		oar = (struct sadb_address *)mhp->ext[SADB_X_EXT_NAT_T_OAR];
5638 	} else
5639 		oar = NULL;
5640 
5641 	/* Initialize addresses only for transport mode */
5642 	if (sav->sah->saidx.mode != IPSEC_MODE_TUNNEL) {
5643 		cksum = 0;
5644 		if (oai != NULL) {
5645 			/* Currently we support only AF_INET */
5646 			sa = (struct sockaddr *)(oai + 1);
5647 			if (sa->sa_family != AF_INET ||
5648 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5649 				ipseclog((LOG_DEBUG,
5650 				    "%s: wrong NAT-OAi header.\n",
5651 				    __func__));
5652 				return (EINVAL);
5653 			}
5654 			/* Ignore address if it the same */
5655 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5656 			    sav->sah->saidx.src.sin.sin_addr.s_addr) {
5657 				bcopy(sa, &sav->natt->oai.sa, sa->sa_len);
5658 				sav->natt->flags |= IPSEC_NATT_F_OAI;
5659 				/* Calculate checksum delta */
5660 				addr = sav->sah->saidx.src.sin.sin_addr.s_addr;
5661 				cksum = in_addword(cksum, ~addr >> 16);
5662 				cksum = in_addword(cksum, ~addr & 0xffff);
5663 				addr = sav->natt->oai.sin.sin_addr.s_addr;
5664 				cksum = in_addword(cksum, addr >> 16);
5665 				cksum = in_addword(cksum, addr & 0xffff);
5666 			}
5667 		}
5668 		if (oar != NULL) {
5669 			/* Currently we support only AF_INET */
5670 			sa = (struct sockaddr *)(oar + 1);
5671 			if (sa->sa_family != AF_INET ||
5672 			    sa->sa_len != sizeof(struct sockaddr_in)) {
5673 				ipseclog((LOG_DEBUG,
5674 				    "%s: wrong NAT-OAr header.\n",
5675 				    __func__));
5676 				return (EINVAL);
5677 			}
5678 			/* Ignore address if it the same */
5679 			if (((struct sockaddr_in *)sa)->sin_addr.s_addr !=
5680 			    sav->sah->saidx.dst.sin.sin_addr.s_addr) {
5681 				bcopy(sa, &sav->natt->oar.sa, sa->sa_len);
5682 				sav->natt->flags |= IPSEC_NATT_F_OAR;
5683 				/* Calculate checksum delta */
5684 				addr = sav->sah->saidx.dst.sin.sin_addr.s_addr;
5685 				cksum = in_addword(cksum, ~addr >> 16);
5686 				cksum = in_addword(cksum, ~addr & 0xffff);
5687 				addr = sav->natt->oar.sin.sin_addr.s_addr;
5688 				cksum = in_addword(cksum, addr >> 16);
5689 				cksum = in_addword(cksum, addr & 0xffff);
5690 			}
5691 		}
5692 		sav->natt->cksum = cksum;
5693 	}
5694 	return (0);
5695 }
5696 
5697 static int
5698 key_setident(struct secashead *sah, const struct sadb_msghdr *mhp)
5699 {
5700 	const struct sadb_ident *idsrc, *iddst;
5701 	int idsrclen, iddstlen;
5702 
5703 	IPSEC_ASSERT(sah != NULL, ("null secashead"));
5704 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5705 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5706 
5707 	/* don't make buffer if not there */
5708 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) &&
5709 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5710 		sah->idents = NULL;
5711 		sah->identd = NULL;
5712 		return (0);
5713 	}
5714 
5715 	if (SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_SRC) ||
5716 	    SADB_CHECKHDR(mhp, SADB_EXT_IDENTITY_DST)) {
5717 		ipseclog((LOG_DEBUG, "%s: invalid identity.\n", __func__));
5718 		return (EINVAL);
5719 	}
5720 
5721 	idsrc = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_SRC];
5722 	iddst = (const struct sadb_ident *)mhp->ext[SADB_EXT_IDENTITY_DST];
5723 	idsrclen = mhp->extlen[SADB_EXT_IDENTITY_SRC];
5724 	iddstlen = mhp->extlen[SADB_EXT_IDENTITY_DST];
5725 
5726 	/* validity check */
5727 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
5728 		ipseclog((LOG_DEBUG, "%s: ident type mismatch.\n", __func__));
5729 		return EINVAL;
5730 	}
5731 
5732 	switch (idsrc->sadb_ident_type) {
5733 	case SADB_IDENTTYPE_PREFIX:
5734 	case SADB_IDENTTYPE_FQDN:
5735 	case SADB_IDENTTYPE_USERFQDN:
5736 	default:
5737 		/* XXX do nothing */
5738 		sah->idents = NULL;
5739 		sah->identd = NULL;
5740 	 	return 0;
5741 	}
5742 
5743 	/* make structure */
5744 	sah->idents = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5745 	if (sah->idents == NULL) {
5746 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5747 		return ENOBUFS;
5748 	}
5749 	sah->identd = malloc(sizeof(struct secident), M_IPSEC_MISC, M_NOWAIT);
5750 	if (sah->identd == NULL) {
5751 		free(sah->idents, M_IPSEC_MISC);
5752 		sah->idents = NULL;
5753 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
5754 		return ENOBUFS;
5755 	}
5756 	sah->idents->type = idsrc->sadb_ident_type;
5757 	sah->idents->id = idsrc->sadb_ident_id;
5758 
5759 	sah->identd->type = iddst->sadb_ident_type;
5760 	sah->identd->id = iddst->sadb_ident_id;
5761 
5762 	return 0;
5763 }
5764 
5765 /*
5766  * m will not be freed on return.
5767  * it is caller's responsibility to free the result.
5768  *
5769  * Called from SADB_ADD and SADB_UPDATE. Reply will contain headers
5770  * from the request in defined order.
5771  */
5772 static struct mbuf *
5773 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
5774 {
5775 	struct mbuf *n;
5776 
5777 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5778 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5779 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5780 
5781 	/* create new sadb_msg to reply. */
5782 	n = key_gather_mbuf(m, mhp, 1, 16, SADB_EXT_RESERVED,
5783 	    SADB_EXT_SA, SADB_X_EXT_SA2,
5784 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
5785 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
5786 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
5787 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
5788 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
5789 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NEW_ADDRESS_SRC,
5790 	    SADB_X_EXT_NEW_ADDRESS_DST);
5791 	if (!n)
5792 		return NULL;
5793 
5794 	if (n->m_len < sizeof(struct sadb_msg)) {
5795 		n = m_pullup(n, sizeof(struct sadb_msg));
5796 		if (n == NULL)
5797 			return NULL;
5798 	}
5799 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
5800 	mtod(n, struct sadb_msg *)->sadb_msg_len =
5801 	    PFKEY_UNIT64(n->m_pkthdr.len);
5802 
5803 	return n;
5804 }
5805 
5806 /*
5807  * SADB_DELETE processing
5808  * receive
5809  *   <base, SA(*), address(SD)>
5810  * from the ikmpd, and set SADB_SASTATE_DEAD,
5811  * and send,
5812  *   <base, SA(*), address(SD)>
5813  * to the ikmpd.
5814  *
5815  * m will always be freed.
5816  */
5817 static int
5818 key_delete(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5819 {
5820 	struct secasindex saidx;
5821 	struct sadb_address *src0, *dst0;
5822 	struct secasvar *sav;
5823 	struct sadb_sa *sa0;
5824 	uint8_t proto;
5825 
5826 	IPSEC_ASSERT(so != NULL, ("null socket"));
5827 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
5828 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
5829 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
5830 
5831 	/* map satype to proto */
5832 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
5833 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
5834 		    __func__));
5835 		return key_senderror(so, m, EINVAL);
5836 	}
5837 
5838 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
5839 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
5840 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
5841 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
5842 		ipseclog((LOG_DEBUG, "%s: invalid message is passed.\n",
5843 		    __func__));
5844 		return key_senderror(so, m, EINVAL);
5845 	}
5846 
5847 	src0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_SRC]);
5848 	dst0 = (struct sadb_address *)(mhp->ext[SADB_EXT_ADDRESS_DST]);
5849 
5850 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
5851 	    (struct sockaddr *)(dst0 + 1)) != 0) {
5852 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
5853 		return (key_senderror(so, m, EINVAL));
5854 	}
5855 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
5856 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA)) {
5857 		/*
5858 		 * Caller wants us to delete all non-LARVAL SAs
5859 		 * that match the src/dst.  This is used during
5860 		 * IKE INITIAL-CONTACT.
5861 		 * XXXAE: this looks like some extension to RFC2367.
5862 		 */
5863 		ipseclog((LOG_DEBUG, "%s: doing delete all.\n", __func__));
5864 		return (key_delete_all(so, m, mhp, &saidx));
5865 	}
5866 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA)) {
5867 		ipseclog((LOG_DEBUG,
5868 		    "%s: invalid message: wrong header size.\n", __func__));
5869 		return (key_senderror(so, m, EINVAL));
5870 	}
5871 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
5872 	if (proto == IPPROTO_TCP)
5873 		sav = key_getsav_tcpmd5(&saidx, NULL);
5874 	else
5875 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
5876 	if (sav == NULL) {
5877 		ipseclog((LOG_DEBUG, "%s: no SA found for SPI %u.\n",
5878 		    __func__, ntohl(sa0->sadb_sa_spi)));
5879 		return (key_senderror(so, m, ESRCH));
5880 	}
5881 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
5882 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
5883 		    __func__, ntohl(sav->spi)));
5884 		key_freesav(&sav);
5885 		return (key_senderror(so, m, ESRCH));
5886 	}
5887 	KEYDBG(KEY_STAMP,
5888 	    printf("%s: SA(%p)\n", __func__, sav));
5889 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
5890 	key_unlinksav(sav);
5891 	key_freesav(&sav);
5892 
5893     {
5894 	struct mbuf *n;
5895 	struct sadb_msg *newmsg;
5896 
5897 	/* create new sadb_msg to reply. */
5898 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
5899 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
5900 	if (!n)
5901 		return key_senderror(so, m, ENOBUFS);
5902 
5903 	if (n->m_len < sizeof(struct sadb_msg)) {
5904 		n = m_pullup(n, sizeof(struct sadb_msg));
5905 		if (n == NULL)
5906 			return key_senderror(so, m, ENOBUFS);
5907 	}
5908 	newmsg = mtod(n, struct sadb_msg *);
5909 	newmsg->sadb_msg_errno = 0;
5910 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
5911 
5912 	m_freem(m);
5913 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5914     }
5915 }
5916 
5917 /*
5918  * delete all SAs for src/dst.  Called from key_delete().
5919  */
5920 static int
5921 key_delete_all(struct socket *so, struct mbuf *m,
5922     const struct sadb_msghdr *mhp, struct secasindex *saidx)
5923 {
5924 	struct secasvar_queue drainq;
5925 	struct secashead *sah;
5926 	struct secasvar *sav, *nextsav;
5927 
5928 	TAILQ_INIT(&drainq);
5929 	SAHTREE_WLOCK();
5930 	LIST_FOREACH(sah, SAHADDRHASH_HASH(saidx), addrhash) {
5931 		if (key_cmpsaidx(&sah->saidx, saidx, CMP_HEAD) == 0)
5932 			continue;
5933 		/* Move all ALIVE SAs into drainq */
5934 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
5935 	}
5936 	/* Unlink all queued SAs from SPI hash */
5937 	TAILQ_FOREACH(sav, &drainq, chain) {
5938 		sav->state = SADB_SASTATE_DEAD;
5939 		LIST_REMOVE(sav, spihash);
5940 	}
5941 	SAHTREE_WUNLOCK();
5942 	/* Now we can release reference for all SAs in drainq */
5943 	sav = TAILQ_FIRST(&drainq);
5944 	while (sav != NULL) {
5945 		KEYDBG(KEY_STAMP,
5946 		    printf("%s: SA(%p)\n", __func__, sav));
5947 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
5948 		nextsav = TAILQ_NEXT(sav, chain);
5949 		key_freesah(&sav->sah); /* release reference from SAV */
5950 		key_freesav(&sav); /* release last reference */
5951 		sav = nextsav;
5952 	}
5953 
5954     {
5955 	struct mbuf *n;
5956 	struct sadb_msg *newmsg;
5957 
5958 	/* create new sadb_msg to reply. */
5959 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
5960 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
5961 	if (!n)
5962 		return key_senderror(so, m, ENOBUFS);
5963 
5964 	if (n->m_len < sizeof(struct sadb_msg)) {
5965 		n = m_pullup(n, sizeof(struct sadb_msg));
5966 		if (n == NULL)
5967 			return key_senderror(so, m, ENOBUFS);
5968 	}
5969 	newmsg = mtod(n, struct sadb_msg *);
5970 	newmsg->sadb_msg_errno = 0;
5971 	newmsg->sadb_msg_len = PFKEY_UNIT64(n->m_pkthdr.len);
5972 
5973 	m_freem(m);
5974 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5975     }
5976 }
5977 
5978 /*
5979  * Delete all alive SAs for corresponding xform.
5980  * Larval SAs have not initialized tdb_xform, so it is safe to leave them
5981  * here when xform disappears.
5982  */
5983 static void
5984 key_delete_xform(const struct xformsw *xsp)
5985 {
5986 	struct secasvar_queue drainq;
5987 	struct secashead *sah;
5988 	struct secasvar *sav, *nextsav;
5989 
5990 	TAILQ_INIT(&drainq);
5991 	SAHTREE_WLOCK();
5992 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
5993 		sav = TAILQ_FIRST(&sah->savtree_alive);
5994 		if (sav == NULL)
5995 			continue;
5996 		if (sav->tdb_xform != xsp)
5997 			continue;
5998 		/*
5999 		 * It is supposed that all SAs in the chain are related to
6000 		 * one xform.
6001 		 */
6002 		TAILQ_CONCAT(&drainq, &sah->savtree_alive, chain);
6003 	}
6004 	/* Unlink all queued SAs from SPI hash */
6005 	TAILQ_FOREACH(sav, &drainq, chain) {
6006 		sav->state = SADB_SASTATE_DEAD;
6007 		LIST_REMOVE(sav, spihash);
6008 	}
6009 	SAHTREE_WUNLOCK();
6010 
6011 	/* Now we can release reference for all SAs in drainq */
6012 	sav = TAILQ_FIRST(&drainq);
6013 	while (sav != NULL) {
6014 		KEYDBG(KEY_STAMP,
6015 		    printf("%s: SA(%p)\n", __func__, sav));
6016 		KEYDBG(KEY_DATA, kdebug_secasv(sav));
6017 		nextsav = TAILQ_NEXT(sav, chain);
6018 		key_freesah(&sav->sah); /* release reference from SAV */
6019 		key_freesav(&sav); /* release last reference */
6020 		sav = nextsav;
6021 	}
6022 }
6023 
6024 /*
6025  * SADB_GET processing
6026  * receive
6027  *   <base, SA(*), address(SD)>
6028  * from the ikmpd, and get a SP and a SA to respond,
6029  * and send,
6030  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6031  *       (identity(SD),) (sensitivity)>
6032  * to the ikmpd.
6033  *
6034  * m will always be freed.
6035  */
6036 static int
6037 key_get(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6038 {
6039 	struct secasindex saidx;
6040 	struct sadb_address *src0, *dst0;
6041 	struct sadb_sa *sa0;
6042 	struct secasvar *sav;
6043 	uint8_t proto;
6044 
6045 	IPSEC_ASSERT(so != NULL, ("null socket"));
6046 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6047 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6048 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6049 
6050 	/* map satype to proto */
6051 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6052 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6053 			__func__));
6054 		return key_senderror(so, m, EINVAL);
6055 	}
6056 
6057 	if (SADB_CHECKHDR(mhp, SADB_EXT_SA) ||
6058 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6059 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST)) {
6060 		ipseclog((LOG_DEBUG,
6061 		    "%s: invalid message: missing required header.\n",
6062 		    __func__));
6063 		return key_senderror(so, m, EINVAL);
6064 	}
6065 	if (SADB_CHECKLEN(mhp, SADB_EXT_SA) ||
6066 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6067 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST)) {
6068 		ipseclog((LOG_DEBUG,
6069 		    "%s: invalid message: wrong header size.\n", __func__));
6070 		return key_senderror(so, m, EINVAL);
6071 	}
6072 
6073 	sa0 = (struct sadb_sa *)mhp->ext[SADB_EXT_SA];
6074 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
6075 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
6076 
6077 	if (key_checksockaddrs((struct sockaddr *)(src0 + 1),
6078 	    (struct sockaddr *)(dst0 + 1)) != 0) {
6079 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6080 		return key_senderror(so, m, EINVAL);
6081 	}
6082 	KEY_SETSECASIDX(proto, IPSEC_MODE_ANY, 0, src0 + 1, dst0 + 1, &saidx);
6083 
6084 	if (proto == IPPROTO_TCP)
6085 		sav = key_getsav_tcpmd5(&saidx, NULL);
6086 	else
6087 		sav = key_getsavbyspi(sa0->sadb_sa_spi);
6088 	if (sav == NULL) {
6089 		ipseclog((LOG_DEBUG, "%s: no SA found.\n", __func__));
6090 		return key_senderror(so, m, ESRCH);
6091 	}
6092 	if (key_cmpsaidx(&sav->sah->saidx, &saidx, CMP_HEAD) == 0) {
6093 		ipseclog((LOG_DEBUG, "%s: saidx mismatched for SPI %u.\n",
6094 		    __func__, ntohl(sa0->sadb_sa_spi)));
6095 		key_freesav(&sav);
6096 		return (key_senderror(so, m, ESRCH));
6097 	}
6098 
6099     {
6100 	struct mbuf *n;
6101 	uint8_t satype;
6102 
6103 	/* map proto to satype */
6104 	if ((satype = key_proto2satype(sav->sah->saidx.proto)) == 0) {
6105 		ipseclog((LOG_DEBUG, "%s: there was invalid proto in SAD.\n",
6106 		    __func__));
6107 		key_freesav(&sav);
6108 		return key_senderror(so, m, EINVAL);
6109 	}
6110 
6111 	/* create new sadb_msg to reply. */
6112 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
6113 	    mhp->msg->sadb_msg_pid);
6114 
6115 	key_freesav(&sav);
6116 	if (!n)
6117 		return key_senderror(so, m, ENOBUFS);
6118 
6119 	m_freem(m);
6120 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
6121     }
6122 }
6123 
6124 /* XXX make it sysctl-configurable? */
6125 static void
6126 key_getcomb_setlifetime(struct sadb_comb *comb)
6127 {
6128 
6129 	comb->sadb_comb_soft_allocations = 1;
6130 	comb->sadb_comb_hard_allocations = 1;
6131 	comb->sadb_comb_soft_bytes = 0;
6132 	comb->sadb_comb_hard_bytes = 0;
6133 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
6134 	comb->sadb_comb_soft_addtime = comb->sadb_comb_soft_addtime * 80 / 100;
6135 	comb->sadb_comb_soft_usetime = 28800;	/* 8 hours */
6136 	comb->sadb_comb_hard_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
6137 }
6138 
6139 /*
6140  * XXX reorder combinations by preference
6141  * XXX no idea if the user wants ESP authentication or not
6142  */
6143 static struct mbuf *
6144 key_getcomb_ealg(void)
6145 {
6146 	struct sadb_comb *comb;
6147 	const struct enc_xform *algo;
6148 	struct mbuf *result = NULL, *m, *n;
6149 	int encmin;
6150 	int i, off, o;
6151 	int totlen;
6152 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6153 
6154 	m = NULL;
6155 	for (i = 1; i <= SADB_EALG_MAX; i++) {
6156 		algo = enc_algorithm_lookup(i);
6157 		if (algo == NULL)
6158 			continue;
6159 
6160 		/* discard algorithms with key size smaller than system min */
6161 		if (_BITS(algo->maxkey) < V_ipsec_esp_keymin)
6162 			continue;
6163 		if (_BITS(algo->minkey) < V_ipsec_esp_keymin)
6164 			encmin = V_ipsec_esp_keymin;
6165 		else
6166 			encmin = _BITS(algo->minkey);
6167 
6168 		if (V_ipsec_esp_auth)
6169 			m = key_getcomb_ah();
6170 		else {
6171 			IPSEC_ASSERT(l <= MLEN,
6172 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6173 			MGET(m, M_NOWAIT, MT_DATA);
6174 			if (m) {
6175 				M_ALIGN(m, l);
6176 				m->m_len = l;
6177 				m->m_next = NULL;
6178 				bzero(mtod(m, caddr_t), m->m_len);
6179 			}
6180 		}
6181 		if (!m)
6182 			goto fail;
6183 
6184 		totlen = 0;
6185 		for (n = m; n; n = n->m_next)
6186 			totlen += n->m_len;
6187 		IPSEC_ASSERT((totlen % l) == 0, ("totlen=%u, l=%u", totlen, l));
6188 
6189 		for (off = 0; off < totlen; off += l) {
6190 			n = m_pulldown(m, off, l, &o);
6191 			if (!n) {
6192 				/* m is already freed */
6193 				goto fail;
6194 			}
6195 			comb = (struct sadb_comb *)(mtod(n, caddr_t) + o);
6196 			bzero(comb, sizeof(*comb));
6197 			key_getcomb_setlifetime(comb);
6198 			comb->sadb_comb_encrypt = i;
6199 			comb->sadb_comb_encrypt_minbits = encmin;
6200 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
6201 		}
6202 
6203 		if (!result)
6204 			result = m;
6205 		else
6206 			m_cat(result, m);
6207 	}
6208 
6209 	return result;
6210 
6211  fail:
6212 	if (result)
6213 		m_freem(result);
6214 	return NULL;
6215 }
6216 
6217 static void
6218 key_getsizes_ah(const struct auth_hash *ah, int alg, u_int16_t* min,
6219     u_int16_t* max)
6220 {
6221 
6222 	*min = *max = ah->keysize;
6223 	if (ah->keysize == 0) {
6224 		/*
6225 		 * Transform takes arbitrary key size but algorithm
6226 		 * key size is restricted.  Enforce this here.
6227 		 */
6228 		switch (alg) {
6229 		case SADB_X_AALG_MD5:	*min = *max = 16; break;
6230 		case SADB_X_AALG_SHA:	*min = *max = 20; break;
6231 		case SADB_X_AALG_NULL:	*min = 1; *max = 256; break;
6232 		case SADB_X_AALG_SHA2_256: *min = *max = 32; break;
6233 		case SADB_X_AALG_SHA2_384: *min = *max = 48; break;
6234 		case SADB_X_AALG_SHA2_512: *min = *max = 64; break;
6235 		default:
6236 			DPRINTF(("%s: unknown AH algorithm %u\n",
6237 				__func__, alg));
6238 			break;
6239 		}
6240 	}
6241 }
6242 
6243 /*
6244  * XXX reorder combinations by preference
6245  */
6246 static struct mbuf *
6247 key_getcomb_ah()
6248 {
6249 	const struct auth_hash *algo;
6250 	struct sadb_comb *comb;
6251 	struct mbuf *m;
6252 	u_int16_t minkeysize, maxkeysize;
6253 	int i;
6254 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6255 
6256 	m = NULL;
6257 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6258 #if 1
6259 		/* we prefer HMAC algorithms, not old algorithms */
6260 		if (i != SADB_AALG_SHA1HMAC &&
6261 		    i != SADB_AALG_MD5HMAC  &&
6262 		    i != SADB_X_AALG_SHA2_256 &&
6263 		    i != SADB_X_AALG_SHA2_384 &&
6264 		    i != SADB_X_AALG_SHA2_512)
6265 			continue;
6266 #endif
6267 		algo = auth_algorithm_lookup(i);
6268 		if (!algo)
6269 			continue;
6270 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6271 		/* discard algorithms with key size smaller than system min */
6272 		if (_BITS(minkeysize) < V_ipsec_ah_keymin)
6273 			continue;
6274 
6275 		if (!m) {
6276 			IPSEC_ASSERT(l <= MLEN,
6277 				("l=%u > MLEN=%lu", l, (u_long) MLEN));
6278 			MGET(m, M_NOWAIT, MT_DATA);
6279 			if (m) {
6280 				M_ALIGN(m, l);
6281 				m->m_len = l;
6282 				m->m_next = NULL;
6283 			}
6284 		} else
6285 			M_PREPEND(m, l, M_NOWAIT);
6286 		if (!m)
6287 			return NULL;
6288 
6289 		comb = mtod(m, struct sadb_comb *);
6290 		bzero(comb, sizeof(*comb));
6291 		key_getcomb_setlifetime(comb);
6292 		comb->sadb_comb_auth = i;
6293 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6294 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6295 	}
6296 
6297 	return m;
6298 }
6299 
6300 /*
6301  * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6302  * XXX reorder combinations by preference
6303  */
6304 static struct mbuf *
6305 key_getcomb_ipcomp()
6306 {
6307 	const struct comp_algo *algo;
6308 	struct sadb_comb *comb;
6309 	struct mbuf *m;
6310 	int i;
6311 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6312 
6313 	m = NULL;
6314 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6315 		algo = comp_algorithm_lookup(i);
6316 		if (!algo)
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_encrypt = i;
6337 		/* what should we set into sadb_comb_*_{min,max}bits? */
6338 	}
6339 
6340 	return m;
6341 }
6342 
6343 /*
6344  * XXX no way to pass mode (transport/tunnel) to userland
6345  * XXX replay checking?
6346  * XXX sysctl interface to ipsec_{ah,esp}_keymin
6347  */
6348 static struct mbuf *
6349 key_getprop(const struct secasindex *saidx)
6350 {
6351 	struct sadb_prop *prop;
6352 	struct mbuf *m, *n;
6353 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6354 	int totlen;
6355 
6356 	switch (saidx->proto)  {
6357 	case IPPROTO_ESP:
6358 		m = key_getcomb_ealg();
6359 		break;
6360 	case IPPROTO_AH:
6361 		m = key_getcomb_ah();
6362 		break;
6363 	case IPPROTO_IPCOMP:
6364 		m = key_getcomb_ipcomp();
6365 		break;
6366 	default:
6367 		return NULL;
6368 	}
6369 
6370 	if (!m)
6371 		return NULL;
6372 	M_PREPEND(m, l, M_NOWAIT);
6373 	if (!m)
6374 		return NULL;
6375 
6376 	totlen = 0;
6377 	for (n = m; n; n = n->m_next)
6378 		totlen += n->m_len;
6379 
6380 	prop = mtod(m, struct sadb_prop *);
6381 	bzero(prop, sizeof(*prop));
6382 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6383 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6384 	prop->sadb_prop_replay = 32;	/* XXX */
6385 
6386 	return m;
6387 }
6388 
6389 /*
6390  * SADB_ACQUIRE processing called by key_checkrequest() and key_acquire2().
6391  * send
6392  *   <base, SA, address(SD), (address(P)), x_policy,
6393  *       (identity(SD),) (sensitivity,) proposal>
6394  * to KMD, and expect to receive
6395  *   <base> with SADB_ACQUIRE if error occurred,
6396  * or
6397  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6398  * from KMD by PF_KEY.
6399  *
6400  * XXX x_policy is outside of RFC2367 (KAME extension).
6401  * XXX sensitivity is not supported.
6402  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6403  * see comment for key_getcomb_ipcomp().
6404  *
6405  * OUT:
6406  *    0     : succeed
6407  *    others: error number
6408  */
6409 static int
6410 key_acquire(const struct secasindex *saidx, struct secpolicy *sp)
6411 {
6412 	union sockaddr_union addr;
6413 	struct mbuf *result, *m;
6414 	uint32_t seq;
6415 	int error;
6416 	uint16_t ul_proto;
6417 	uint8_t mask, satype;
6418 
6419 	IPSEC_ASSERT(saidx != NULL, ("null saidx"));
6420 	satype = key_proto2satype(saidx->proto);
6421 	IPSEC_ASSERT(satype != 0, ("null satype, protocol %u", saidx->proto));
6422 
6423 	error = -1;
6424 	result = NULL;
6425 	ul_proto = IPSEC_ULPROTO_ANY;
6426 
6427 	/* Get seq number to check whether sending message or not. */
6428 	seq = key_getacq(saidx, &error);
6429 	if (seq == 0)
6430 		return (error);
6431 
6432 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0);
6433 	if (!m) {
6434 		error = ENOBUFS;
6435 		goto fail;
6436 	}
6437 	result = m;
6438 
6439 	/*
6440 	 * set sadb_address for saidx's.
6441 	 *
6442 	 * Note that if sp is supplied, then we're being called from
6443 	 * key_allocsa_policy() and should supply port and protocol
6444 	 * information.
6445 	 * XXXAE: why only TCP and UDP? ICMP and SCTP looks applicable too.
6446 	 * XXXAE: probably we can handle this in the ipsec[46]_allocsa().
6447 	 * XXXAE: it looks like we should save this info in the ACQ entry.
6448 	 */
6449 	if (sp != NULL && (sp->spidx.ul_proto == IPPROTO_TCP ||
6450 	    sp->spidx.ul_proto == IPPROTO_UDP))
6451 		ul_proto = sp->spidx.ul_proto;
6452 
6453 	addr = saidx->src;
6454 	mask = FULLMASK;
6455 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6456 		switch (sp->spidx.src.sa.sa_family) {
6457 		case AF_INET:
6458 			if (sp->spidx.src.sin.sin_port != IPSEC_PORT_ANY) {
6459 				addr.sin.sin_port = sp->spidx.src.sin.sin_port;
6460 				mask = sp->spidx.prefs;
6461 			}
6462 			break;
6463 		case AF_INET6:
6464 			if (sp->spidx.src.sin6.sin6_port != IPSEC_PORT_ANY) {
6465 				addr.sin6.sin6_port =
6466 				    sp->spidx.src.sin6.sin6_port;
6467 				mask = sp->spidx.prefs;
6468 			}
6469 			break;
6470 		default:
6471 			break;
6472 		}
6473 	}
6474 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &addr.sa, mask, ul_proto);
6475 	if (!m) {
6476 		error = ENOBUFS;
6477 		goto fail;
6478 	}
6479 	m_cat(result, m);
6480 
6481 	addr = saidx->dst;
6482 	mask = FULLMASK;
6483 	if (ul_proto != IPSEC_ULPROTO_ANY) {
6484 		switch (sp->spidx.dst.sa.sa_family) {
6485 		case AF_INET:
6486 			if (sp->spidx.dst.sin.sin_port != IPSEC_PORT_ANY) {
6487 				addr.sin.sin_port = sp->spidx.dst.sin.sin_port;
6488 				mask = sp->spidx.prefd;
6489 			}
6490 			break;
6491 		case AF_INET6:
6492 			if (sp->spidx.dst.sin6.sin6_port != IPSEC_PORT_ANY) {
6493 				addr.sin6.sin6_port =
6494 				    sp->spidx.dst.sin6.sin6_port;
6495 				mask = sp->spidx.prefd;
6496 			}
6497 			break;
6498 		default:
6499 			break;
6500 		}
6501 	}
6502 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &addr.sa, mask, ul_proto);
6503 	if (!m) {
6504 		error = ENOBUFS;
6505 		goto fail;
6506 	}
6507 	m_cat(result, m);
6508 
6509 	/* XXX proxy address (optional) */
6510 
6511 	/* set sadb_x_policy */
6512 	if (sp != NULL) {
6513 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6514 		    sp->priority);
6515 		if (!m) {
6516 			error = ENOBUFS;
6517 			goto fail;
6518 		}
6519 		m_cat(result, m);
6520 	}
6521 
6522 	/* XXX identity (optional) */
6523 #if 0
6524 	if (idexttype && fqdn) {
6525 		/* create identity extension (FQDN) */
6526 		struct sadb_ident *id;
6527 		int fqdnlen;
6528 
6529 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
6530 		id = (struct sadb_ident *)p;
6531 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6532 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6533 		id->sadb_ident_exttype = idexttype;
6534 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
6535 		bcopy(fqdn, id + 1, fqdnlen);
6536 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
6537 	}
6538 
6539 	if (idexttype) {
6540 		/* create identity extension (USERFQDN) */
6541 		struct sadb_ident *id;
6542 		int userfqdnlen;
6543 
6544 		if (userfqdn) {
6545 			/* +1 for terminating-NUL */
6546 			userfqdnlen = strlen(userfqdn) + 1;
6547 		} else
6548 			userfqdnlen = 0;
6549 		id = (struct sadb_ident *)p;
6550 		bzero(id, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6551 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6552 		id->sadb_ident_exttype = idexttype;
6553 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
6554 		/* XXX is it correct? */
6555 		if (curproc && curproc->p_cred)
6556 			id->sadb_ident_id = curproc->p_cred->p_ruid;
6557 		if (userfqdn && userfqdnlen)
6558 			bcopy(userfqdn, id + 1, userfqdnlen);
6559 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
6560 	}
6561 #endif
6562 
6563 	/* XXX sensitivity (optional) */
6564 
6565 	/* create proposal/combination extension */
6566 	m = key_getprop(saidx);
6567 #if 0
6568 	/*
6569 	 * spec conformant: always attach proposal/combination extension,
6570 	 * the problem is that we have no way to attach it for ipcomp,
6571 	 * due to the way sadb_comb is declared in RFC2367.
6572 	 */
6573 	if (!m) {
6574 		error = ENOBUFS;
6575 		goto fail;
6576 	}
6577 	m_cat(result, m);
6578 #else
6579 	/*
6580 	 * outside of spec; make proposal/combination extension optional.
6581 	 */
6582 	if (m)
6583 		m_cat(result, m);
6584 #endif
6585 
6586 	if ((result->m_flags & M_PKTHDR) == 0) {
6587 		error = EINVAL;
6588 		goto fail;
6589 	}
6590 
6591 	if (result->m_len < sizeof(struct sadb_msg)) {
6592 		result = m_pullup(result, sizeof(struct sadb_msg));
6593 		if (result == NULL) {
6594 			error = ENOBUFS;
6595 			goto fail;
6596 		}
6597 	}
6598 
6599 	result->m_pkthdr.len = 0;
6600 	for (m = result; m; m = m->m_next)
6601 		result->m_pkthdr.len += m->m_len;
6602 
6603 	mtod(result, struct sadb_msg *)->sadb_msg_len =
6604 	    PFKEY_UNIT64(result->m_pkthdr.len);
6605 
6606 	KEYDBG(KEY_STAMP,
6607 	    printf("%s: SP(%p)\n", __func__, sp));
6608 	KEYDBG(KEY_DATA, kdebug_secasindex(saidx, NULL));
6609 
6610 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
6611 
6612  fail:
6613 	if (result)
6614 		m_freem(result);
6615 	return error;
6616 }
6617 
6618 static uint32_t
6619 key_newacq(const struct secasindex *saidx, int *perror)
6620 {
6621 	struct secacq *acq;
6622 	uint32_t seq;
6623 
6624 	acq = malloc(sizeof(*acq), M_IPSEC_SAQ, M_NOWAIT | M_ZERO);
6625 	if (acq == NULL) {
6626 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6627 		*perror = ENOBUFS;
6628 		return (0);
6629 	}
6630 
6631 	/* copy secindex */
6632 	bcopy(saidx, &acq->saidx, sizeof(acq->saidx));
6633 	acq->created = time_second;
6634 	acq->count = 0;
6635 
6636 	/* add to acqtree */
6637 	ACQ_LOCK();
6638 	seq = acq->seq = (V_acq_seq == ~0 ? 1 : ++V_acq_seq);
6639 	LIST_INSERT_HEAD(&V_acqtree, acq, chain);
6640 	LIST_INSERT_HEAD(ACQADDRHASH_HASH(saidx), acq, addrhash);
6641 	LIST_INSERT_HEAD(ACQSEQHASH_HASH(seq), acq, seqhash);
6642 	ACQ_UNLOCK();
6643 	*perror = 0;
6644 	return (seq);
6645 }
6646 
6647 static uint32_t
6648 key_getacq(const struct secasindex *saidx, int *perror)
6649 {
6650 	struct secacq *acq;
6651 	uint32_t seq;
6652 
6653 	ACQ_LOCK();
6654 	LIST_FOREACH(acq, ACQADDRHASH_HASH(saidx), addrhash) {
6655 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY)) {
6656 			if (acq->count > V_key_blockacq_count) {
6657 				/*
6658 				 * Reset counter and send message.
6659 				 * Also reset created time to keep ACQ for
6660 				 * this saidx.
6661 				 */
6662 				acq->created = time_second;
6663 				acq->count = 0;
6664 				seq = acq->seq;
6665 			} else {
6666 				/*
6667 				 * Increment counter and do nothing.
6668 				 * We send SADB_ACQUIRE message only
6669 				 * for each V_key_blockacq_count packet.
6670 				 */
6671 				acq->count++;
6672 				seq = 0;
6673 			}
6674 			break;
6675 		}
6676 	}
6677 	ACQ_UNLOCK();
6678 	if (acq != NULL) {
6679 		*perror = 0;
6680 		return (seq);
6681 	}
6682 	/* allocate new  entry */
6683 	return (key_newacq(saidx, perror));
6684 }
6685 
6686 static int
6687 key_acqreset(uint32_t seq)
6688 {
6689 	struct secacq *acq;
6690 
6691 	ACQ_LOCK();
6692 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6693 		if (acq->seq == seq) {
6694 			acq->count = 0;
6695 			acq->created = time_second;
6696 			break;
6697 		}
6698 	}
6699 	ACQ_UNLOCK();
6700 	if (acq == NULL)
6701 		return (ESRCH);
6702 	return (0);
6703 }
6704 /*
6705  * Mark ACQ entry as stale to remove it in key_flush_acq().
6706  * Called after successful SADB_GETSPI message.
6707  */
6708 static int
6709 key_acqdone(const struct secasindex *saidx, uint32_t seq)
6710 {
6711 	struct secacq *acq;
6712 
6713 	ACQ_LOCK();
6714 	LIST_FOREACH(acq, ACQSEQHASH_HASH(seq), seqhash) {
6715 		if (acq->seq == seq)
6716 			break;
6717 	}
6718 	if (acq != NULL) {
6719 		if (key_cmpsaidx(&acq->saidx, saidx, CMP_EXACTLY) == 0) {
6720 			ipseclog((LOG_DEBUG,
6721 			    "%s: Mismatched saidx for ACQ %u", __func__, seq));
6722 			acq = NULL;
6723 		} else {
6724 			acq->created = 0;
6725 		}
6726 	} else {
6727 		ipseclog((LOG_DEBUG,
6728 		    "%s: ACQ %u is not found.", __func__, seq));
6729 	}
6730 	ACQ_UNLOCK();
6731 	if (acq == NULL)
6732 		return (ESRCH);
6733 	return (0);
6734 }
6735 
6736 static struct secspacq *
6737 key_newspacq(struct secpolicyindex *spidx)
6738 {
6739 	struct secspacq *acq;
6740 
6741 	/* get new entry */
6742 	acq = malloc(sizeof(struct secspacq), M_IPSEC_SAQ, M_NOWAIT|M_ZERO);
6743 	if (acq == NULL) {
6744 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6745 		return NULL;
6746 	}
6747 
6748 	/* copy secindex */
6749 	bcopy(spidx, &acq->spidx, sizeof(acq->spidx));
6750 	acq->created = time_second;
6751 	acq->count = 0;
6752 
6753 	/* add to spacqtree */
6754 	SPACQ_LOCK();
6755 	LIST_INSERT_HEAD(&V_spacqtree, acq, chain);
6756 	SPACQ_UNLOCK();
6757 
6758 	return acq;
6759 }
6760 
6761 static struct secspacq *
6762 key_getspacq(struct secpolicyindex *spidx)
6763 {
6764 	struct secspacq *acq;
6765 
6766 	SPACQ_LOCK();
6767 	LIST_FOREACH(acq, &V_spacqtree, chain) {
6768 		if (key_cmpspidx_exactly(spidx, &acq->spidx)) {
6769 			/* NB: return holding spacq_lock */
6770 			return acq;
6771 		}
6772 	}
6773 	SPACQ_UNLOCK();
6774 
6775 	return NULL;
6776 }
6777 
6778 /*
6779  * SADB_ACQUIRE processing,
6780  * in first situation, is receiving
6781  *   <base>
6782  * from the ikmpd, and clear sequence of its secasvar entry.
6783  *
6784  * In second situation, is receiving
6785  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6786  * from a user land process, and return
6787  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
6788  * to the socket.
6789  *
6790  * m will always be freed.
6791  */
6792 static int
6793 key_acquire2(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6794 {
6795 	SAHTREE_RLOCK_TRACKER;
6796 	struct sadb_address *src0, *dst0;
6797 	struct secasindex saidx;
6798 	struct secashead *sah;
6799 	uint32_t reqid;
6800 	int error;
6801 	uint8_t mode, proto;
6802 
6803 	IPSEC_ASSERT(so != NULL, ("null socket"));
6804 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6805 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6806 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6807 
6808 	/*
6809 	 * Error message from KMd.
6810 	 * We assume that if error was occurred in IKEd, the length of PFKEY
6811 	 * message is equal to the size of sadb_msg structure.
6812 	 * We do not raise error even if error occurred in this function.
6813 	 */
6814 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
6815 		/* check sequence number */
6816 		if (mhp->msg->sadb_msg_seq == 0 ||
6817 		    mhp->msg->sadb_msg_errno == 0) {
6818 			ipseclog((LOG_DEBUG, "%s: must specify sequence "
6819 				"number and errno.\n", __func__));
6820 		} else {
6821 			/*
6822 			 * IKEd reported that error occurred.
6823 			 * XXXAE: what it expects from the kernel?
6824 			 * Probably we should send SADB_ACQUIRE again?
6825 			 * If so, reset ACQ's state.
6826 			 * XXXAE: it looks useless.
6827 			 */
6828 			key_acqreset(mhp->msg->sadb_msg_seq);
6829 		}
6830 		m_freem(m);
6831 		return (0);
6832 	}
6833 
6834 	/*
6835 	 * This message is from user land.
6836 	 */
6837 
6838 	/* map satype to proto */
6839 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6840 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
6841 		    __func__));
6842 		return key_senderror(so, m, EINVAL);
6843 	}
6844 
6845 	if (SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_SRC) ||
6846 	    SADB_CHECKHDR(mhp, SADB_EXT_ADDRESS_DST) ||
6847 	    SADB_CHECKHDR(mhp, SADB_EXT_PROPOSAL)) {
6848 		ipseclog((LOG_DEBUG,
6849 		    "%s: invalid message: missing required header.\n",
6850 		    __func__));
6851 		return key_senderror(so, m, EINVAL);
6852 	}
6853 	if (SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_SRC) ||
6854 	    SADB_CHECKLEN(mhp, SADB_EXT_ADDRESS_DST) ||
6855 	    SADB_CHECKLEN(mhp, SADB_EXT_PROPOSAL)) {
6856 		ipseclog((LOG_DEBUG,
6857 		    "%s: invalid message: wrong header size.\n", __func__));
6858 		return key_senderror(so, m, EINVAL);
6859 	}
6860 
6861 	if (SADB_CHECKHDR(mhp, SADB_X_EXT_SA2)) {
6862 		mode = IPSEC_MODE_ANY;
6863 		reqid = 0;
6864 	} else {
6865 		if (SADB_CHECKLEN(mhp, SADB_X_EXT_SA2)) {
6866 			ipseclog((LOG_DEBUG,
6867 			    "%s: invalid message: wrong header size.\n",
6868 			    __func__));
6869 			return key_senderror(so, m, EINVAL);
6870 		}
6871 		mode = ((struct sadb_x_sa2 *)
6872 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_mode;
6873 		reqid = ((struct sadb_x_sa2 *)
6874 		    mhp->ext[SADB_X_EXT_SA2])->sadb_x_sa2_reqid;
6875 	}
6876 
6877 	src0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_SRC];
6878 	dst0 = (struct sadb_address *)mhp->ext[SADB_EXT_ADDRESS_DST];
6879 
6880 	error = key_checksockaddrs((struct sockaddr *)(src0 + 1),
6881 	    (struct sockaddr *)(dst0 + 1));
6882 	if (error != 0) {
6883 		ipseclog((LOG_DEBUG, "%s: invalid sockaddr.\n", __func__));
6884 		return key_senderror(so, m, EINVAL);
6885 	}
6886 	KEY_SETSECASIDX(proto, mode, reqid, src0 + 1, dst0 + 1, &saidx);
6887 
6888 	/* get a SA index */
6889 	SAHTREE_RLOCK();
6890 	LIST_FOREACH(sah, SAHADDRHASH_HASH(&saidx), addrhash) {
6891 		if (key_cmpsaidx(&sah->saidx, &saidx, CMP_MODE_REQID))
6892 			break;
6893 	}
6894 	SAHTREE_RUNLOCK();
6895 	if (sah != NULL) {
6896 		ipseclog((LOG_DEBUG, "%s: a SA exists already.\n", __func__));
6897 		return key_senderror(so, m, EEXIST);
6898 	}
6899 
6900 	error = key_acquire(&saidx, NULL);
6901 	if (error != 0) {
6902 		ipseclog((LOG_DEBUG,
6903 		    "%s: error %d returned from key_acquire()\n",
6904 			__func__, error));
6905 		return key_senderror(so, m, error);
6906 	}
6907 	m_freem(m);
6908 	return (0);
6909 }
6910 
6911 /*
6912  * SADB_REGISTER processing.
6913  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
6914  * receive
6915  *   <base>
6916  * from the ikmpd, and register a socket to send PF_KEY messages,
6917  * and send
6918  *   <base, supported>
6919  * to KMD by PF_KEY.
6920  * If socket is detached, must free from regnode.
6921  *
6922  * m will always be freed.
6923  */
6924 static int
6925 key_register(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
6926 {
6927 	struct secreg *reg, *newreg = NULL;
6928 
6929 	IPSEC_ASSERT(so != NULL, ("null socket"));
6930 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
6931 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
6932 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
6933 
6934 	/* check for invalid register message */
6935 	if (mhp->msg->sadb_msg_satype >= sizeof(V_regtree)/sizeof(V_regtree[0]))
6936 		return key_senderror(so, m, EINVAL);
6937 
6938 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
6939 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
6940 		goto setmsg;
6941 
6942 	/* check whether existing or not */
6943 	REGTREE_LOCK();
6944 	LIST_FOREACH(reg, &V_regtree[mhp->msg->sadb_msg_satype], chain) {
6945 		if (reg->so == so) {
6946 			REGTREE_UNLOCK();
6947 			ipseclog((LOG_DEBUG, "%s: socket exists already.\n",
6948 				__func__));
6949 			return key_senderror(so, m, EEXIST);
6950 		}
6951 	}
6952 
6953 	/* create regnode */
6954 	newreg =  malloc(sizeof(struct secreg), M_IPSEC_SAR, M_NOWAIT|M_ZERO);
6955 	if (newreg == NULL) {
6956 		REGTREE_UNLOCK();
6957 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
6958 		return key_senderror(so, m, ENOBUFS);
6959 	}
6960 
6961 	newreg->so = so;
6962 	((struct keycb *)sotorawcb(so))->kp_registered++;
6963 
6964 	/* add regnode to regtree. */
6965 	LIST_INSERT_HEAD(&V_regtree[mhp->msg->sadb_msg_satype], newreg, chain);
6966 	REGTREE_UNLOCK();
6967 
6968   setmsg:
6969     {
6970 	struct mbuf *n;
6971 	struct sadb_msg *newmsg;
6972 	struct sadb_supported *sup;
6973 	u_int len, alen, elen;
6974 	int off;
6975 	int i;
6976 	struct sadb_alg *alg;
6977 
6978 	/* create new sadb_msg to reply. */
6979 	alen = 0;
6980 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6981 		if (auth_algorithm_lookup(i))
6982 			alen += sizeof(struct sadb_alg);
6983 	}
6984 	if (alen)
6985 		alen += sizeof(struct sadb_supported);
6986 	elen = 0;
6987 	for (i = 1; i <= SADB_EALG_MAX; i++) {
6988 		if (enc_algorithm_lookup(i))
6989 			elen += sizeof(struct sadb_alg);
6990 	}
6991 	if (elen)
6992 		elen += sizeof(struct sadb_supported);
6993 
6994 	len = sizeof(struct sadb_msg) + alen + elen;
6995 
6996 	if (len > MCLBYTES)
6997 		return key_senderror(so, m, ENOBUFS);
6998 
6999 	MGETHDR(n, M_NOWAIT, MT_DATA);
7000 	if (len > MHLEN) {
7001 		if (!(MCLGET(n, M_NOWAIT))) {
7002 			m_freem(n);
7003 			n = NULL;
7004 		}
7005 	}
7006 	if (!n)
7007 		return key_senderror(so, m, ENOBUFS);
7008 
7009 	n->m_pkthdr.len = n->m_len = len;
7010 	n->m_next = NULL;
7011 	off = 0;
7012 
7013 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, caddr_t) + off);
7014 	newmsg = mtod(n, struct sadb_msg *);
7015 	newmsg->sadb_msg_errno = 0;
7016 	newmsg->sadb_msg_len = PFKEY_UNIT64(len);
7017 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7018 
7019 	/* for authentication algorithm */
7020 	if (alen) {
7021 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7022 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7023 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7024 		off += PFKEY_ALIGN8(sizeof(*sup));
7025 
7026 		for (i = 1; i <= SADB_AALG_MAX; i++) {
7027 			const struct auth_hash *aalgo;
7028 			u_int16_t minkeysize, maxkeysize;
7029 
7030 			aalgo = auth_algorithm_lookup(i);
7031 			if (!aalgo)
7032 				continue;
7033 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7034 			alg->sadb_alg_id = i;
7035 			alg->sadb_alg_ivlen = 0;
7036 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7037 			alg->sadb_alg_minbits = _BITS(minkeysize);
7038 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7039 			off += PFKEY_ALIGN8(sizeof(*alg));
7040 		}
7041 	}
7042 
7043 	/* for encryption algorithm */
7044 	if (elen) {
7045 		sup = (struct sadb_supported *)(mtod(n, caddr_t) + off);
7046 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7047 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7048 		off += PFKEY_ALIGN8(sizeof(*sup));
7049 
7050 		for (i = 1; i <= SADB_EALG_MAX; i++) {
7051 			const struct enc_xform *ealgo;
7052 
7053 			ealgo = enc_algorithm_lookup(i);
7054 			if (!ealgo)
7055 				continue;
7056 			alg = (struct sadb_alg *)(mtod(n, caddr_t) + off);
7057 			alg->sadb_alg_id = i;
7058 			alg->sadb_alg_ivlen = ealgo->ivsize;
7059 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7060 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7061 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7062 		}
7063 	}
7064 
7065 	IPSEC_ASSERT(off == len,
7066 		("length assumption failed (off %u len %u)", off, len));
7067 
7068 	m_freem(m);
7069 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7070     }
7071 }
7072 
7073 /*
7074  * free secreg entry registered.
7075  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7076  */
7077 void
7078 key_freereg(struct socket *so)
7079 {
7080 	struct secreg *reg;
7081 	int i;
7082 
7083 	IPSEC_ASSERT(so != NULL, ("NULL so"));
7084 
7085 	/*
7086 	 * check whether existing or not.
7087 	 * check all type of SA, because there is a potential that
7088 	 * one socket is registered to multiple type of SA.
7089 	 */
7090 	REGTREE_LOCK();
7091 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7092 		LIST_FOREACH(reg, &V_regtree[i], chain) {
7093 			if (reg->so == so && __LIST_CHAINED(reg)) {
7094 				LIST_REMOVE(reg, chain);
7095 				free(reg, M_IPSEC_SAR);
7096 				break;
7097 			}
7098 		}
7099 	}
7100 	REGTREE_UNLOCK();
7101 }
7102 
7103 /*
7104  * SADB_EXPIRE processing
7105  * send
7106  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7107  * to KMD by PF_KEY.
7108  * NOTE: We send only soft lifetime extension.
7109  *
7110  * OUT:	0	: succeed
7111  *	others	: error number
7112  */
7113 static int
7114 key_expire(struct secasvar *sav, int hard)
7115 {
7116 	struct mbuf *result = NULL, *m;
7117 	struct sadb_lifetime *lt;
7118 	uint32_t replay_count;
7119 	int error, len;
7120 	uint8_t satype;
7121 
7122 	IPSEC_ASSERT (sav != NULL, ("null sav"));
7123 	IPSEC_ASSERT (sav->sah != NULL, ("null sa header"));
7124 
7125 	KEYDBG(KEY_STAMP,
7126 	    printf("%s: SA(%p) expired %s lifetime\n", __func__,
7127 		sav, hard ? "hard": "soft"));
7128 	KEYDBG(KEY_DATA, kdebug_secasv(sav));
7129 	/* set msg header */
7130 	satype = key_proto2satype(sav->sah->saidx.proto);
7131 	IPSEC_ASSERT(satype != 0, ("invalid proto, satype %u", satype));
7132 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, sav->refcnt);
7133 	if (!m) {
7134 		error = ENOBUFS;
7135 		goto fail;
7136 	}
7137 	result = m;
7138 
7139 	/* create SA extension */
7140 	m = key_setsadbsa(sav);
7141 	if (!m) {
7142 		error = ENOBUFS;
7143 		goto fail;
7144 	}
7145 	m_cat(result, m);
7146 
7147 	/* create SA extension */
7148 	SECASVAR_LOCK(sav);
7149 	replay_count = sav->replay ? sav->replay->count : 0;
7150 	SECASVAR_UNLOCK(sav);
7151 
7152 	m = key_setsadbxsa2(sav->sah->saidx.mode, replay_count,
7153 			sav->sah->saidx.reqid);
7154 	if (!m) {
7155 		error = ENOBUFS;
7156 		goto fail;
7157 	}
7158 	m_cat(result, m);
7159 
7160 	if (sav->replay && sav->replay->wsize > UINT8_MAX) {
7161 		m = key_setsadbxsareplay(sav->replay->wsize);
7162 		if (!m) {
7163 			error = ENOBUFS;
7164 			goto fail;
7165 		}
7166 		m_cat(result, m);
7167 	}
7168 
7169 	/* create lifetime extension (current and soft) */
7170 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7171 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
7172 	if (m == NULL) {
7173 		error = ENOBUFS;
7174 		goto fail;
7175 	}
7176 	m_align(m, len);
7177 	m->m_len = len;
7178 	bzero(mtod(m, caddr_t), len);
7179 	lt = mtod(m, struct sadb_lifetime *);
7180 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7181 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7182 	lt->sadb_lifetime_allocations =
7183 	    (uint32_t)counter_u64_fetch(sav->lft_c_allocations);
7184 	lt->sadb_lifetime_bytes =
7185 	    counter_u64_fetch(sav->lft_c_bytes);
7186 	lt->sadb_lifetime_addtime = sav->created;
7187 	lt->sadb_lifetime_usetime = sav->firstused;
7188 	lt = (struct sadb_lifetime *)(mtod(m, caddr_t) + len / 2);
7189 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7190 	if (hard) {
7191 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
7192 		lt->sadb_lifetime_allocations = sav->lft_h->allocations;
7193 		lt->sadb_lifetime_bytes = sav->lft_h->bytes;
7194 		lt->sadb_lifetime_addtime = sav->lft_h->addtime;
7195 		lt->sadb_lifetime_usetime = sav->lft_h->usetime;
7196 	} else {
7197 		lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_SOFT;
7198 		lt->sadb_lifetime_allocations = sav->lft_s->allocations;
7199 		lt->sadb_lifetime_bytes = sav->lft_s->bytes;
7200 		lt->sadb_lifetime_addtime = sav->lft_s->addtime;
7201 		lt->sadb_lifetime_usetime = sav->lft_s->usetime;
7202 	}
7203 	m_cat(result, m);
7204 
7205 	/* set sadb_address for source */
7206 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
7207 	    &sav->sah->saidx.src.sa,
7208 	    FULLMASK, IPSEC_ULPROTO_ANY);
7209 	if (!m) {
7210 		error = ENOBUFS;
7211 		goto fail;
7212 	}
7213 	m_cat(result, m);
7214 
7215 	/* set sadb_address for destination */
7216 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
7217 	    &sav->sah->saidx.dst.sa,
7218 	    FULLMASK, IPSEC_ULPROTO_ANY);
7219 	if (!m) {
7220 		error = ENOBUFS;
7221 		goto fail;
7222 	}
7223 	m_cat(result, m);
7224 
7225 	/*
7226 	 * XXX-BZ Handle NAT-T extensions here.
7227 	 * XXXAE: it doesn't seem quite useful. IKEs should not depend on
7228 	 * this information, we report only significant SA fields.
7229 	 */
7230 
7231 	if ((result->m_flags & M_PKTHDR) == 0) {
7232 		error = EINVAL;
7233 		goto fail;
7234 	}
7235 
7236 	if (result->m_len < sizeof(struct sadb_msg)) {
7237 		result = m_pullup(result, sizeof(struct sadb_msg));
7238 		if (result == NULL) {
7239 			error = ENOBUFS;
7240 			goto fail;
7241 		}
7242 	}
7243 
7244 	result->m_pkthdr.len = 0;
7245 	for (m = result; m; m = m->m_next)
7246 		result->m_pkthdr.len += m->m_len;
7247 
7248 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7249 	    PFKEY_UNIT64(result->m_pkthdr.len);
7250 
7251 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7252 
7253  fail:
7254 	if (result)
7255 		m_freem(result);
7256 	return error;
7257 }
7258 
7259 static void
7260 key_freesah_flushed(struct secashead_queue *flushq)
7261 {
7262 	struct secashead *sah, *nextsah;
7263 	struct secasvar *sav, *nextsav;
7264 
7265 	sah = TAILQ_FIRST(flushq);
7266 	while (sah != NULL) {
7267 		sav = TAILQ_FIRST(&sah->savtree_larval);
7268 		while (sav != NULL) {
7269 			nextsav = TAILQ_NEXT(sav, chain);
7270 			TAILQ_REMOVE(&sah->savtree_larval, sav, chain);
7271 			key_freesav(&sav); /* release last reference */
7272 			key_freesah(&sah); /* release reference from SAV */
7273 			sav = nextsav;
7274 		}
7275 		sav = TAILQ_FIRST(&sah->savtree_alive);
7276 		while (sav != NULL) {
7277 			nextsav = TAILQ_NEXT(sav, chain);
7278 			TAILQ_REMOVE(&sah->savtree_alive, sav, chain);
7279 			key_freesav(&sav); /* release last reference */
7280 			key_freesah(&sah); /* release reference from SAV */
7281 			sav = nextsav;
7282 		}
7283 		nextsah = TAILQ_NEXT(sah, chain);
7284 		key_freesah(&sah);	/* release last reference */
7285 		sah = nextsah;
7286 	}
7287 }
7288 
7289 /*
7290  * SADB_FLUSH processing
7291  * receive
7292  *   <base>
7293  * from the ikmpd, and free all entries in secastree.
7294  * and send,
7295  *   <base>
7296  * to the ikmpd.
7297  * NOTE: to do is only marking SADB_SASTATE_DEAD.
7298  *
7299  * m will always be freed.
7300  */
7301 static int
7302 key_flush(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7303 {
7304 	struct secashead_queue flushq;
7305 	struct sadb_msg *newmsg;
7306 	struct secashead *sah, *nextsah;
7307 	struct secasvar *sav;
7308 	uint8_t proto;
7309 	int i;
7310 
7311 	IPSEC_ASSERT(so != NULL, ("null socket"));
7312 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7313 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7314 
7315 	/* map satype to proto */
7316 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7317 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7318 			__func__));
7319 		return key_senderror(so, m, EINVAL);
7320 	}
7321 	KEYDBG(KEY_STAMP,
7322 	    printf("%s: proto %u\n", __func__, proto));
7323 
7324 	TAILQ_INIT(&flushq);
7325 	if (proto == IPSEC_PROTO_ANY) {
7326 		/* no SATYPE specified, i.e. flushing all SA. */
7327 		SAHTREE_WLOCK();
7328 		/* Move all SAHs into flushq */
7329 		TAILQ_CONCAT(&flushq, &V_sahtree, chain);
7330 		/* Flush all buckets in SPI hash */
7331 		for (i = 0; i < V_savhash_mask + 1; i++)
7332 			LIST_INIT(&V_savhashtbl[i]);
7333 		/* Flush all buckets in SAHADDRHASH */
7334 		for (i = 0; i < V_sahaddrhash_mask + 1; i++)
7335 			LIST_INIT(&V_sahaddrhashtbl[i]);
7336 		/* Mark all SAHs as unlinked */
7337 		TAILQ_FOREACH(sah, &flushq, chain) {
7338 			sah->state = SADB_SASTATE_DEAD;
7339 			/*
7340 			 * Callout handler makes its job using
7341 			 * RLOCK and drain queues. In case, when this
7342 			 * function will be called just before it
7343 			 * acquires WLOCK, we need to mark SAs as
7344 			 * unlinked to prevent second unlink.
7345 			 */
7346 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7347 				sav->state = SADB_SASTATE_DEAD;
7348 			}
7349 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7350 				sav->state = SADB_SASTATE_DEAD;
7351 			}
7352 		}
7353 		SAHTREE_WUNLOCK();
7354 	} else {
7355 		SAHTREE_WLOCK();
7356 		sah = TAILQ_FIRST(&V_sahtree);
7357 		while (sah != NULL) {
7358 			IPSEC_ASSERT(sah->state != SADB_SASTATE_DEAD,
7359 			    ("DEAD SAH %p in SADB_FLUSH", sah));
7360 			nextsah = TAILQ_NEXT(sah, chain);
7361 			if (sah->saidx.proto != proto) {
7362 				sah = nextsah;
7363 				continue;
7364 			}
7365 			sah->state = SADB_SASTATE_DEAD;
7366 			TAILQ_REMOVE(&V_sahtree, sah, chain);
7367 			LIST_REMOVE(sah, addrhash);
7368 			/* Unlink all SAs from SPI hash */
7369 			TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7370 				LIST_REMOVE(sav, spihash);
7371 				sav->state = SADB_SASTATE_DEAD;
7372 			}
7373 			TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7374 				LIST_REMOVE(sav, spihash);
7375 				sav->state = SADB_SASTATE_DEAD;
7376 			}
7377 			/* Add SAH into flushq */
7378 			TAILQ_INSERT_HEAD(&flushq, sah, chain);
7379 			sah = nextsah;
7380 		}
7381 		SAHTREE_WUNLOCK();
7382 	}
7383 
7384 	key_freesah_flushed(&flushq);
7385 	/* Free all queued SAs and SAHs */
7386 	if (m->m_len < sizeof(struct sadb_msg) ||
7387 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7388 		ipseclog((LOG_DEBUG, "%s: No more memory.\n", __func__));
7389 		return key_senderror(so, m, ENOBUFS);
7390 	}
7391 
7392 	if (m->m_next)
7393 		m_freem(m->m_next);
7394 	m->m_next = NULL;
7395 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7396 	newmsg = mtod(m, struct sadb_msg *);
7397 	newmsg->sadb_msg_errno = 0;
7398 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7399 
7400 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7401 }
7402 
7403 /*
7404  * SADB_DUMP processing
7405  * dump all entries including status of DEAD in SAD.
7406  * receive
7407  *   <base>
7408  * from the ikmpd, and dump all secasvar leaves
7409  * and send,
7410  *   <base> .....
7411  * to the ikmpd.
7412  *
7413  * m will always be freed.
7414  */
7415 static int
7416 key_dump(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7417 {
7418 	SAHTREE_RLOCK_TRACKER;
7419 	struct secashead *sah;
7420 	struct secasvar *sav;
7421 	struct sadb_msg *newmsg;
7422 	struct mbuf *n;
7423 	uint32_t cnt;
7424 	uint8_t proto, satype;
7425 
7426 	IPSEC_ASSERT(so != NULL, ("null socket"));
7427 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7428 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7429 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7430 
7431 	/* map satype to proto */
7432 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
7433 		ipseclog((LOG_DEBUG, "%s: invalid satype is passed.\n",
7434 		    __func__));
7435 		return key_senderror(so, m, EINVAL);
7436 	}
7437 
7438 	/* count sav entries to be sent to the userland. */
7439 	cnt = 0;
7440 	SAHTREE_RLOCK();
7441 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7442 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7443 		    proto != sah->saidx.proto)
7444 			continue;
7445 
7446 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain)
7447 			cnt++;
7448 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain)
7449 			cnt++;
7450 	}
7451 
7452 	if (cnt == 0) {
7453 		SAHTREE_RUNLOCK();
7454 		return key_senderror(so, m, ENOENT);
7455 	}
7456 
7457 	/* send this to the userland, one at a time. */
7458 	newmsg = NULL;
7459 	TAILQ_FOREACH(sah, &V_sahtree, chain) {
7460 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7461 		    proto != sah->saidx.proto)
7462 			continue;
7463 
7464 		/* map proto to satype */
7465 		if ((satype = key_proto2satype(sah->saidx.proto)) == 0) {
7466 			SAHTREE_RUNLOCK();
7467 			ipseclog((LOG_DEBUG, "%s: there was invalid proto in "
7468 			    "SAD.\n", __func__));
7469 			return key_senderror(so, m, EINVAL);
7470 		}
7471 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
7472 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7473 			    --cnt, mhp->msg->sadb_msg_pid);
7474 			if (n == NULL) {
7475 				SAHTREE_RUNLOCK();
7476 				return key_senderror(so, m, ENOBUFS);
7477 			}
7478 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7479 		}
7480 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
7481 			n = key_setdumpsa(sav, SADB_DUMP, satype,
7482 			    --cnt, mhp->msg->sadb_msg_pid);
7483 			if (n == NULL) {
7484 				SAHTREE_RUNLOCK();
7485 				return key_senderror(so, m, ENOBUFS);
7486 			}
7487 			key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
7488 		}
7489 	}
7490 	SAHTREE_RUNLOCK();
7491 	m_freem(m);
7492 	return (0);
7493 }
7494 /*
7495  * SADB_X_PROMISC processing
7496  *
7497  * m will always be freed.
7498  */
7499 static int
7500 key_promisc(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
7501 {
7502 	int olen;
7503 
7504 	IPSEC_ASSERT(so != NULL, ("null socket"));
7505 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7506 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7507 	IPSEC_ASSERT(mhp->msg != NULL, ("null msg"));
7508 
7509 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7510 
7511 	if (olen < sizeof(struct sadb_msg)) {
7512 #if 1
7513 		return key_senderror(so, m, EINVAL);
7514 #else
7515 		m_freem(m);
7516 		return 0;
7517 #endif
7518 	} else if (olen == sizeof(struct sadb_msg)) {
7519 		/* enable/disable promisc mode */
7520 		struct keycb *kp;
7521 
7522 		if ((kp = (struct keycb *)sotorawcb(so)) == NULL)
7523 			return key_senderror(so, m, EINVAL);
7524 		mhp->msg->sadb_msg_errno = 0;
7525 		switch (mhp->msg->sadb_msg_satype) {
7526 		case 0:
7527 		case 1:
7528 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
7529 			break;
7530 		default:
7531 			return key_senderror(so, m, EINVAL);
7532 		}
7533 
7534 		/* send the original message back to everyone */
7535 		mhp->msg->sadb_msg_errno = 0;
7536 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7537 	} else {
7538 		/* send packet as is */
7539 
7540 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
7541 
7542 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
7543 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7544 	}
7545 }
7546 
7547 static int (*key_typesw[])(struct socket *, struct mbuf *,
7548 		const struct sadb_msghdr *) = {
7549 	NULL,		/* SADB_RESERVED */
7550 	key_getspi,	/* SADB_GETSPI */
7551 	key_update,	/* SADB_UPDATE */
7552 	key_add,	/* SADB_ADD */
7553 	key_delete,	/* SADB_DELETE */
7554 	key_get,	/* SADB_GET */
7555 	key_acquire2,	/* SADB_ACQUIRE */
7556 	key_register,	/* SADB_REGISTER */
7557 	NULL,		/* SADB_EXPIRE */
7558 	key_flush,	/* SADB_FLUSH */
7559 	key_dump,	/* SADB_DUMP */
7560 	key_promisc,	/* SADB_X_PROMISC */
7561 	NULL,		/* SADB_X_PCHANGE */
7562 	key_spdadd,	/* SADB_X_SPDUPDATE */
7563 	key_spdadd,	/* SADB_X_SPDADD */
7564 	key_spddelete,	/* SADB_X_SPDDELETE */
7565 	key_spdget,	/* SADB_X_SPDGET */
7566 	NULL,		/* SADB_X_SPDACQUIRE */
7567 	key_spddump,	/* SADB_X_SPDDUMP */
7568 	key_spdflush,	/* SADB_X_SPDFLUSH */
7569 	key_spdadd,	/* SADB_X_SPDSETIDX */
7570 	NULL,		/* SADB_X_SPDEXPIRE */
7571 	key_spddelete2,	/* SADB_X_SPDDELETE2 */
7572 };
7573 
7574 /*
7575  * parse sadb_msg buffer to process PFKEYv2,
7576  * and create a data to response if needed.
7577  * I think to be dealed with mbuf directly.
7578  * IN:
7579  *     msgp  : pointer to pointer to a received buffer pulluped.
7580  *             This is rewrited to response.
7581  *     so    : pointer to socket.
7582  * OUT:
7583  *    length for buffer to send to user process.
7584  */
7585 int
7586 key_parse(struct mbuf *m, struct socket *so)
7587 {
7588 	struct sadb_msg *msg;
7589 	struct sadb_msghdr mh;
7590 	u_int orglen;
7591 	int error;
7592 	int target;
7593 
7594 	IPSEC_ASSERT(so != NULL, ("null socket"));
7595 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7596 
7597 	if (m->m_len < sizeof(struct sadb_msg)) {
7598 		m = m_pullup(m, sizeof(struct sadb_msg));
7599 		if (!m)
7600 			return ENOBUFS;
7601 	}
7602 	msg = mtod(m, struct sadb_msg *);
7603 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
7604 	target = KEY_SENDUP_ONE;
7605 
7606 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len != orglen) {
7607 		ipseclog((LOG_DEBUG, "%s: invalid message length.\n",__func__));
7608 		PFKEYSTAT_INC(out_invlen);
7609 		error = EINVAL;
7610 		goto senderror;
7611 	}
7612 
7613 	if (msg->sadb_msg_version != PF_KEY_V2) {
7614 		ipseclog((LOG_DEBUG, "%s: PF_KEY version %u is mismatched.\n",
7615 		    __func__, msg->sadb_msg_version));
7616 		PFKEYSTAT_INC(out_invver);
7617 		error = EINVAL;
7618 		goto senderror;
7619 	}
7620 
7621 	if (msg->sadb_msg_type > SADB_MAX) {
7622 		ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7623 		    __func__, msg->sadb_msg_type));
7624 		PFKEYSTAT_INC(out_invmsgtype);
7625 		error = EINVAL;
7626 		goto senderror;
7627 	}
7628 
7629 	/* for old-fashioned code - should be nuked */
7630 	if (m->m_pkthdr.len > MCLBYTES) {
7631 		m_freem(m);
7632 		return ENOBUFS;
7633 	}
7634 	if (m->m_next) {
7635 		struct mbuf *n;
7636 
7637 		MGETHDR(n, M_NOWAIT, MT_DATA);
7638 		if (n && m->m_pkthdr.len > MHLEN) {
7639 			if (!(MCLGET(n, M_NOWAIT))) {
7640 				m_free(n);
7641 				n = NULL;
7642 			}
7643 		}
7644 		if (!n) {
7645 			m_freem(m);
7646 			return ENOBUFS;
7647 		}
7648 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
7649 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
7650 		n->m_next = NULL;
7651 		m_freem(m);
7652 		m = n;
7653 	}
7654 
7655 	/* align the mbuf chain so that extensions are in contiguous region. */
7656 	error = key_align(m, &mh);
7657 	if (error)
7658 		return error;
7659 
7660 	msg = mh.msg;
7661 
7662 	/* check SA type */
7663 	switch (msg->sadb_msg_satype) {
7664 	case SADB_SATYPE_UNSPEC:
7665 		switch (msg->sadb_msg_type) {
7666 		case SADB_GETSPI:
7667 		case SADB_UPDATE:
7668 		case SADB_ADD:
7669 		case SADB_DELETE:
7670 		case SADB_GET:
7671 		case SADB_ACQUIRE:
7672 		case SADB_EXPIRE:
7673 			ipseclog((LOG_DEBUG, "%s: must specify satype "
7674 			    "when msg type=%u.\n", __func__,
7675 			    msg->sadb_msg_type));
7676 			PFKEYSTAT_INC(out_invsatype);
7677 			error = EINVAL;
7678 			goto senderror;
7679 		}
7680 		break;
7681 	case SADB_SATYPE_AH:
7682 	case SADB_SATYPE_ESP:
7683 	case SADB_X_SATYPE_IPCOMP:
7684 	case SADB_X_SATYPE_TCPSIGNATURE:
7685 		switch (msg->sadb_msg_type) {
7686 		case SADB_X_SPDADD:
7687 		case SADB_X_SPDDELETE:
7688 		case SADB_X_SPDGET:
7689 		case SADB_X_SPDDUMP:
7690 		case SADB_X_SPDFLUSH:
7691 		case SADB_X_SPDSETIDX:
7692 		case SADB_X_SPDUPDATE:
7693 		case SADB_X_SPDDELETE2:
7694 			ipseclog((LOG_DEBUG, "%s: illegal satype=%u\n",
7695 				__func__, msg->sadb_msg_type));
7696 			PFKEYSTAT_INC(out_invsatype);
7697 			error = EINVAL;
7698 			goto senderror;
7699 		}
7700 		break;
7701 	case SADB_SATYPE_RSVP:
7702 	case SADB_SATYPE_OSPFV2:
7703 	case SADB_SATYPE_RIPV2:
7704 	case SADB_SATYPE_MIP:
7705 		ipseclog((LOG_DEBUG, "%s: type %u isn't supported.\n",
7706 			__func__, msg->sadb_msg_satype));
7707 		PFKEYSTAT_INC(out_invsatype);
7708 		error = EOPNOTSUPP;
7709 		goto senderror;
7710 	case 1:	/* XXX: What does it do? */
7711 		if (msg->sadb_msg_type == SADB_X_PROMISC)
7712 			break;
7713 		/*FALLTHROUGH*/
7714 	default:
7715 		ipseclog((LOG_DEBUG, "%s: invalid type %u is passed.\n",
7716 			__func__, msg->sadb_msg_satype));
7717 		PFKEYSTAT_INC(out_invsatype);
7718 		error = EINVAL;
7719 		goto senderror;
7720 	}
7721 
7722 	/* check field of upper layer protocol and address family */
7723 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL
7724 	 && mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
7725 		struct sadb_address *src0, *dst0;
7726 		u_int plen;
7727 
7728 		src0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_SRC]);
7729 		dst0 = (struct sadb_address *)(mh.ext[SADB_EXT_ADDRESS_DST]);
7730 
7731 		/* check upper layer protocol */
7732 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
7733 			ipseclog((LOG_DEBUG, "%s: upper layer protocol "
7734 				"mismatched.\n", __func__));
7735 			PFKEYSTAT_INC(out_invaddr);
7736 			error = EINVAL;
7737 			goto senderror;
7738 		}
7739 
7740 		/* check family */
7741 		if (PFKEY_ADDR_SADDR(src0)->sa_family !=
7742 		    PFKEY_ADDR_SADDR(dst0)->sa_family) {
7743 			ipseclog((LOG_DEBUG, "%s: address family mismatched.\n",
7744 				__func__));
7745 			PFKEYSTAT_INC(out_invaddr);
7746 			error = EINVAL;
7747 			goto senderror;
7748 		}
7749 		if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7750 		    PFKEY_ADDR_SADDR(dst0)->sa_len) {
7751 			ipseclog((LOG_DEBUG, "%s: address struct size "
7752 				"mismatched.\n", __func__));
7753 			PFKEYSTAT_INC(out_invaddr);
7754 			error = EINVAL;
7755 			goto senderror;
7756 		}
7757 
7758 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7759 		case AF_INET:
7760 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7761 			    sizeof(struct sockaddr_in)) {
7762 				PFKEYSTAT_INC(out_invaddr);
7763 				error = EINVAL;
7764 				goto senderror;
7765 			}
7766 			break;
7767 		case AF_INET6:
7768 			if (PFKEY_ADDR_SADDR(src0)->sa_len !=
7769 			    sizeof(struct sockaddr_in6)) {
7770 				PFKEYSTAT_INC(out_invaddr);
7771 				error = EINVAL;
7772 				goto senderror;
7773 			}
7774 			break;
7775 		default:
7776 			ipseclog((LOG_DEBUG, "%s: unsupported address family\n",
7777 				__func__));
7778 			PFKEYSTAT_INC(out_invaddr);
7779 			error = EAFNOSUPPORT;
7780 			goto senderror;
7781 		}
7782 
7783 		switch (PFKEY_ADDR_SADDR(src0)->sa_family) {
7784 		case AF_INET:
7785 			plen = sizeof(struct in_addr) << 3;
7786 			break;
7787 		case AF_INET6:
7788 			plen = sizeof(struct in6_addr) << 3;
7789 			break;
7790 		default:
7791 			plen = 0;	/*fool gcc*/
7792 			break;
7793 		}
7794 
7795 		/* check max prefix length */
7796 		if (src0->sadb_address_prefixlen > plen ||
7797 		    dst0->sadb_address_prefixlen > plen) {
7798 			ipseclog((LOG_DEBUG, "%s: illegal prefixlen.\n",
7799 				__func__));
7800 			PFKEYSTAT_INC(out_invaddr);
7801 			error = EINVAL;
7802 			goto senderror;
7803 		}
7804 
7805 		/*
7806 		 * prefixlen == 0 is valid because there can be a case when
7807 		 * all addresses are matched.
7808 		 */
7809 	}
7810 
7811 	if (msg->sadb_msg_type >= nitems(key_typesw) ||
7812 	    key_typesw[msg->sadb_msg_type] == NULL) {
7813 		PFKEYSTAT_INC(out_invmsgtype);
7814 		error = EINVAL;
7815 		goto senderror;
7816 	}
7817 
7818 	return (*key_typesw[msg->sadb_msg_type])(so, m, &mh);
7819 
7820 senderror:
7821 	msg->sadb_msg_errno = error;
7822 	return key_sendup_mbuf(so, m, target);
7823 }
7824 
7825 static int
7826 key_senderror(struct socket *so, struct mbuf *m, int code)
7827 {
7828 	struct sadb_msg *msg;
7829 
7830 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
7831 		("mbuf too small, len %u", m->m_len));
7832 
7833 	msg = mtod(m, struct sadb_msg *);
7834 	msg->sadb_msg_errno = code;
7835 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
7836 }
7837 
7838 /*
7839  * set the pointer to each header into message buffer.
7840  * m will be freed on error.
7841  * XXX larger-than-MCLBYTES extension?
7842  */
7843 static int
7844 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
7845 {
7846 	struct mbuf *n;
7847 	struct sadb_ext *ext;
7848 	size_t off, end;
7849 	int extlen;
7850 	int toff;
7851 
7852 	IPSEC_ASSERT(m != NULL, ("null mbuf"));
7853 	IPSEC_ASSERT(mhp != NULL, ("null msghdr"));
7854 	IPSEC_ASSERT(m->m_len >= sizeof(struct sadb_msg),
7855 		("mbuf too small, len %u", m->m_len));
7856 
7857 	/* initialize */
7858 	bzero(mhp, sizeof(*mhp));
7859 
7860 	mhp->msg = mtod(m, struct sadb_msg *);
7861 	mhp->ext[0] = (struct sadb_ext *)mhp->msg;	/*XXX backward compat */
7862 
7863 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7864 	extlen = end;	/*just in case extlen is not updated*/
7865 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
7866 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
7867 		if (!n) {
7868 			/* m is already freed */
7869 			return ENOBUFS;
7870 		}
7871 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
7872 
7873 		/* set pointer */
7874 		switch (ext->sadb_ext_type) {
7875 		case SADB_EXT_SA:
7876 		case SADB_EXT_ADDRESS_SRC:
7877 		case SADB_EXT_ADDRESS_DST:
7878 		case SADB_EXT_ADDRESS_PROXY:
7879 		case SADB_EXT_LIFETIME_CURRENT:
7880 		case SADB_EXT_LIFETIME_HARD:
7881 		case SADB_EXT_LIFETIME_SOFT:
7882 		case SADB_EXT_KEY_AUTH:
7883 		case SADB_EXT_KEY_ENCRYPT:
7884 		case SADB_EXT_IDENTITY_SRC:
7885 		case SADB_EXT_IDENTITY_DST:
7886 		case SADB_EXT_SENSITIVITY:
7887 		case SADB_EXT_PROPOSAL:
7888 		case SADB_EXT_SUPPORTED_AUTH:
7889 		case SADB_EXT_SUPPORTED_ENCRYPT:
7890 		case SADB_EXT_SPIRANGE:
7891 		case SADB_X_EXT_POLICY:
7892 		case SADB_X_EXT_SA2:
7893 		case SADB_X_EXT_NAT_T_TYPE:
7894 		case SADB_X_EXT_NAT_T_SPORT:
7895 		case SADB_X_EXT_NAT_T_DPORT:
7896 		case SADB_X_EXT_NAT_T_OAI:
7897 		case SADB_X_EXT_NAT_T_OAR:
7898 		case SADB_X_EXT_NAT_T_FRAG:
7899 		case SADB_X_EXT_SA_REPLAY:
7900 		case SADB_X_EXT_NEW_ADDRESS_SRC:
7901 		case SADB_X_EXT_NEW_ADDRESS_DST:
7902 			/* duplicate check */
7903 			/*
7904 			 * XXX Are there duplication payloads of either
7905 			 * KEY_AUTH or KEY_ENCRYPT ?
7906 			 */
7907 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
7908 				ipseclog((LOG_DEBUG, "%s: duplicate ext_type "
7909 					"%u\n", __func__, ext->sadb_ext_type));
7910 				m_freem(m);
7911 				PFKEYSTAT_INC(out_dupext);
7912 				return EINVAL;
7913 			}
7914 			break;
7915 		default:
7916 			ipseclog((LOG_DEBUG, "%s: invalid ext_type %u\n",
7917 				__func__, ext->sadb_ext_type));
7918 			m_freem(m);
7919 			PFKEYSTAT_INC(out_invexttype);
7920 			return EINVAL;
7921 		}
7922 
7923 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
7924 
7925 		if (key_validate_ext(ext, extlen)) {
7926 			m_freem(m);
7927 			PFKEYSTAT_INC(out_invlen);
7928 			return EINVAL;
7929 		}
7930 
7931 		n = m_pulldown(m, off, extlen, &toff);
7932 		if (!n) {
7933 			/* m is already freed */
7934 			return ENOBUFS;
7935 		}
7936 		ext = (struct sadb_ext *)(mtod(n, caddr_t) + toff);
7937 
7938 		mhp->ext[ext->sadb_ext_type] = ext;
7939 		mhp->extoff[ext->sadb_ext_type] = off;
7940 		mhp->extlen[ext->sadb_ext_type] = extlen;
7941 	}
7942 
7943 	if (off != end) {
7944 		m_freem(m);
7945 		PFKEYSTAT_INC(out_invlen);
7946 		return EINVAL;
7947 	}
7948 
7949 	return 0;
7950 }
7951 
7952 static int
7953 key_validate_ext(const struct sadb_ext *ext, int len)
7954 {
7955 	const struct sockaddr *sa;
7956 	enum { NONE, ADDR } checktype = NONE;
7957 	int baselen = 0;
7958 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
7959 
7960 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
7961 		return EINVAL;
7962 
7963 	/* if it does not match minimum/maximum length, bail */
7964 	if (ext->sadb_ext_type >= nitems(minsize) ||
7965 	    ext->sadb_ext_type >= nitems(maxsize))
7966 		return EINVAL;
7967 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
7968 		return EINVAL;
7969 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
7970 		return EINVAL;
7971 
7972 	/* more checks based on sadb_ext_type XXX need more */
7973 	switch (ext->sadb_ext_type) {
7974 	case SADB_EXT_ADDRESS_SRC:
7975 	case SADB_EXT_ADDRESS_DST:
7976 	case SADB_EXT_ADDRESS_PROXY:
7977 	case SADB_X_EXT_NAT_T_OAI:
7978 	case SADB_X_EXT_NAT_T_OAR:
7979 	case SADB_X_EXT_NEW_ADDRESS_SRC:
7980 	case SADB_X_EXT_NEW_ADDRESS_DST:
7981 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
7982 		checktype = ADDR;
7983 		break;
7984 	case SADB_EXT_IDENTITY_SRC:
7985 	case SADB_EXT_IDENTITY_DST:
7986 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
7987 		    SADB_X_IDENTTYPE_ADDR) {
7988 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
7989 			checktype = ADDR;
7990 		} else
7991 			checktype = NONE;
7992 		break;
7993 	default:
7994 		checktype = NONE;
7995 		break;
7996 	}
7997 
7998 	switch (checktype) {
7999 	case NONE:
8000 		break;
8001 	case ADDR:
8002 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8003 		if (len < baselen + sal)
8004 			return EINVAL;
8005 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8006 			return EINVAL;
8007 		break;
8008 	}
8009 
8010 	return 0;
8011 }
8012 
8013 void
8014 key_init(void)
8015 {
8016 	int i;
8017 
8018 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8019 		TAILQ_INIT(&V_sptree[i]);
8020 		TAILQ_INIT(&V_sptree_ifnet[i]);
8021 	}
8022 
8023 	V_key_lft_zone = uma_zcreate("IPsec SA lft_c",
8024 	    sizeof(uint64_t) * 2, NULL, NULL, NULL, NULL,
8025 	    UMA_ALIGN_PTR, UMA_ZONE_PCPU);
8026 
8027 	TAILQ_INIT(&V_sahtree);
8028 	V_sphashtbl = hashinit(SPHASH_NHASH, M_IPSEC_SP, &V_sphash_mask);
8029 	V_savhashtbl = hashinit(SAVHASH_NHASH, M_IPSEC_SA, &V_savhash_mask);
8030 	V_sahaddrhashtbl = hashinit(SAHHASH_NHASH, M_IPSEC_SAH,
8031 	    &V_sahaddrhash_mask);
8032 	V_acqaddrhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8033 	    &V_acqaddrhash_mask);
8034 	V_acqseqhashtbl = hashinit(ACQHASH_NHASH, M_IPSEC_SAQ,
8035 	    &V_acqseqhash_mask);
8036 
8037 	for (i = 0; i <= SADB_SATYPE_MAX; i++)
8038 		LIST_INIT(&V_regtree[i]);
8039 
8040 	LIST_INIT(&V_acqtree);
8041 	LIST_INIT(&V_spacqtree);
8042 
8043 	if (!IS_DEFAULT_VNET(curvnet))
8044 		return;
8045 
8046 	XFORMS_LOCK_INIT();
8047 	SPTREE_LOCK_INIT();
8048 	REGTREE_LOCK_INIT();
8049 	SAHTREE_LOCK_INIT();
8050 	ACQ_LOCK_INIT();
8051 	SPACQ_LOCK_INIT();
8052 
8053 #ifndef IPSEC_DEBUG2
8054 	callout_init(&key_timer, 1);
8055 	callout_reset(&key_timer, hz, key_timehandler, NULL);
8056 #endif /*IPSEC_DEBUG2*/
8057 
8058 	/* initialize key statistics */
8059 	keystat.getspi_count = 1;
8060 
8061 	if (bootverbose)
8062 		printf("IPsec: Initialized Security Association Processing.\n");
8063 }
8064 
8065 #ifdef VIMAGE
8066 void
8067 key_destroy(void)
8068 {
8069 	struct secashead_queue sahdrainq;
8070 	struct secpolicy_queue drainq;
8071 	struct secpolicy *sp, *nextsp;
8072 	struct secacq *acq, *nextacq;
8073 	struct secspacq *spacq, *nextspacq;
8074 	struct secashead *sah;
8075 	struct secasvar *sav;
8076 	struct secreg *reg;
8077 	int i;
8078 
8079 	/*
8080 	 * XXX: can we just call free() for each object without
8081 	 * walking through safe way with releasing references?
8082 	 */
8083 	TAILQ_INIT(&drainq);
8084 	SPTREE_WLOCK();
8085 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8086 		TAILQ_CONCAT(&drainq, &V_sptree[i], chain);
8087 		TAILQ_CONCAT(&drainq, &V_sptree_ifnet[i], chain);
8088 	}
8089 	SPTREE_WUNLOCK();
8090 	sp = TAILQ_FIRST(&drainq);
8091 	while (sp != NULL) {
8092 		nextsp = TAILQ_NEXT(sp, chain);
8093 		key_freesp(&sp);
8094 		sp = nextsp;
8095 	}
8096 
8097 	TAILQ_INIT(&sahdrainq);
8098 	SAHTREE_WLOCK();
8099 	TAILQ_CONCAT(&sahdrainq, &V_sahtree, chain);
8100 	for (i = 0; i < V_savhash_mask + 1; i++)
8101 		LIST_INIT(&V_savhashtbl[i]);
8102 	for (i = 0; i < V_sahaddrhash_mask + 1; i++)
8103 		LIST_INIT(&V_sahaddrhashtbl[i]);
8104 	TAILQ_FOREACH(sah, &sahdrainq, chain) {
8105 		sah->state = SADB_SASTATE_DEAD;
8106 		TAILQ_FOREACH(sav, &sah->savtree_larval, chain) {
8107 			sav->state = SADB_SASTATE_DEAD;
8108 		}
8109 		TAILQ_FOREACH(sav, &sah->savtree_alive, chain) {
8110 			sav->state = SADB_SASTATE_DEAD;
8111 		}
8112 	}
8113 	SAHTREE_WUNLOCK();
8114 
8115 	key_freesah_flushed(&sahdrainq);
8116 	hashdestroy(V_sphashtbl, M_IPSEC_SP, V_sphash_mask);
8117 	hashdestroy(V_savhashtbl, M_IPSEC_SA, V_savhash_mask);
8118 	hashdestroy(V_sahaddrhashtbl, M_IPSEC_SAH, V_sahaddrhash_mask);
8119 
8120 	REGTREE_LOCK();
8121 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8122 		LIST_FOREACH(reg, &V_regtree[i], chain) {
8123 			if (__LIST_CHAINED(reg)) {
8124 				LIST_REMOVE(reg, chain);
8125 				free(reg, M_IPSEC_SAR);
8126 				break;
8127 			}
8128 		}
8129 	}
8130 	REGTREE_UNLOCK();
8131 
8132 	ACQ_LOCK();
8133 	acq = LIST_FIRST(&V_acqtree);
8134 	while (acq != NULL) {
8135 		nextacq = LIST_NEXT(acq, chain);
8136 		LIST_REMOVE(acq, chain);
8137 		free(acq, M_IPSEC_SAQ);
8138 		acq = nextacq;
8139 	}
8140 	ACQ_UNLOCK();
8141 
8142 	SPACQ_LOCK();
8143 	for (spacq = LIST_FIRST(&V_spacqtree); spacq != NULL;
8144 	    spacq = nextspacq) {
8145 		nextspacq = LIST_NEXT(spacq, chain);
8146 		if (__LIST_CHAINED(spacq)) {
8147 			LIST_REMOVE(spacq, chain);
8148 			free(spacq, M_IPSEC_SAQ);
8149 		}
8150 	}
8151 	SPACQ_UNLOCK();
8152 	hashdestroy(V_acqaddrhashtbl, M_IPSEC_SAQ, V_acqaddrhash_mask);
8153 	hashdestroy(V_acqseqhashtbl, M_IPSEC_SAQ, V_acqseqhash_mask);
8154 	uma_zdestroy(V_key_lft_zone);
8155 }
8156 #endif
8157 
8158 /* record data transfer on SA, and update timestamps */
8159 void
8160 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8161 {
8162 	IPSEC_ASSERT(sav != NULL, ("Null secasvar"));
8163 	IPSEC_ASSERT(m != NULL, ("Null mbuf"));
8164 
8165 	/*
8166 	 * XXX Currently, there is a difference of bytes size
8167 	 * between inbound and outbound processing.
8168 	 */
8169 	counter_u64_add(sav->lft_c_bytes, m->m_pkthdr.len);
8170 
8171 	/*
8172 	 * We use the number of packets as the unit of
8173 	 * allocations.  We increment the variable
8174 	 * whenever {esp,ah}_{in,out}put is called.
8175 	 */
8176 	counter_u64_add(sav->lft_c_allocations, 1);
8177 
8178 	/*
8179 	 * NOTE: We record CURRENT usetime by using wall clock,
8180 	 * in seconds.  HARD and SOFT lifetime are measured by the time
8181 	 * difference (again in seconds) from usetime.
8182 	 *
8183 	 *	usetime
8184 	 *	v     expire   expire
8185 	 * -----+-----+--------+---> t
8186 	 *	<--------------> HARD
8187 	 *	<-----> SOFT
8188 	 */
8189 	if (sav->firstused == 0)
8190 		sav->firstused = time_second;
8191 }
8192 
8193 /*
8194  * Take one of the kernel's security keys and convert it into a PF_KEY
8195  * structure within an mbuf, suitable for sending up to a waiting
8196  * application in user land.
8197  *
8198  * IN:
8199  *    src: A pointer to a kernel security key.
8200  *    exttype: Which type of key this is. Refer to the PF_KEY data structures.
8201  * OUT:
8202  *    a valid mbuf or NULL indicating an error
8203  *
8204  */
8205 
8206 static struct mbuf *
8207 key_setkey(struct seckey *src, uint16_t exttype)
8208 {
8209 	struct mbuf *m;
8210 	struct sadb_key *p;
8211 	int len;
8212 
8213 	if (src == NULL)
8214 		return NULL;
8215 
8216 	len = PFKEY_ALIGN8(sizeof(struct sadb_key) + _KEYLEN(src));
8217 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8218 	if (m == NULL)
8219 		return NULL;
8220 	m_align(m, len);
8221 	m->m_len = len;
8222 	p = mtod(m, struct sadb_key *);
8223 	bzero(p, len);
8224 	p->sadb_key_len = PFKEY_UNIT64(len);
8225 	p->sadb_key_exttype = exttype;
8226 	p->sadb_key_bits = src->bits;
8227 	bcopy(src->key_data, _KEYBUF(p), _KEYLEN(src));
8228 
8229 	return m;
8230 }
8231 
8232 /*
8233  * Take one of the kernel's lifetime data structures and convert it
8234  * into a PF_KEY structure within an mbuf, suitable for sending up to
8235  * a waiting application in user land.
8236  *
8237  * IN:
8238  *    src: A pointer to a kernel lifetime structure.
8239  *    exttype: Which type of lifetime this is. Refer to the PF_KEY
8240  *             data structures for more information.
8241  * OUT:
8242  *    a valid mbuf or NULL indicating an error
8243  *
8244  */
8245 
8246 static struct mbuf *
8247 key_setlifetime(struct seclifetime *src, uint16_t exttype)
8248 {
8249 	struct mbuf *m = NULL;
8250 	struct sadb_lifetime *p;
8251 	int len = PFKEY_ALIGN8(sizeof(struct sadb_lifetime));
8252 
8253 	if (src == NULL)
8254 		return NULL;
8255 
8256 	m = m_get2(len, M_NOWAIT, MT_DATA, 0);
8257 	if (m == NULL)
8258 		return m;
8259 	m_align(m, len);
8260 	m->m_len = len;
8261 	p = mtod(m, struct sadb_lifetime *);
8262 
8263 	bzero(p, len);
8264 	p->sadb_lifetime_len = PFKEY_UNIT64(len);
8265 	p->sadb_lifetime_exttype = exttype;
8266 	p->sadb_lifetime_allocations = src->allocations;
8267 	p->sadb_lifetime_bytes = src->bytes;
8268 	p->sadb_lifetime_addtime = src->addtime;
8269 	p->sadb_lifetime_usetime = src->usetime;
8270 
8271 	return m;
8272 
8273 }
8274 
8275 const struct enc_xform *
8276 enc_algorithm_lookup(int alg)
8277 {
8278 	int i;
8279 
8280 	for (i = 0; i < nitems(supported_ealgs); i++)
8281 		if (alg == supported_ealgs[i].sadb_alg)
8282 			return (supported_ealgs[i].xform);
8283 	return (NULL);
8284 }
8285 
8286 const struct auth_hash *
8287 auth_algorithm_lookup(int alg)
8288 {
8289 	int i;
8290 
8291 	for (i = 0; i < nitems(supported_aalgs); i++)
8292 		if (alg == supported_aalgs[i].sadb_alg)
8293 			return (supported_aalgs[i].xform);
8294 	return (NULL);
8295 }
8296 
8297 const struct comp_algo *
8298 comp_algorithm_lookup(int alg)
8299 {
8300 	int i;
8301 
8302 	for (i = 0; i < nitems(supported_calgs); i++)
8303 		if (alg == supported_calgs[i].sadb_alg)
8304 			return (supported_calgs[i].xform);
8305 	return (NULL);
8306 }
8307 
8308 /*
8309  * Register a transform.
8310  */
8311 static int
8312 xform_register(struct xformsw* xsp)
8313 {
8314 	struct xformsw *entry;
8315 
8316 	XFORMS_LOCK();
8317 	LIST_FOREACH(entry, &xforms, chain) {
8318 		if (entry->xf_type == xsp->xf_type) {
8319 			XFORMS_UNLOCK();
8320 			return (EEXIST);
8321 		}
8322 	}
8323 	LIST_INSERT_HEAD(&xforms, xsp, chain);
8324 	XFORMS_UNLOCK();
8325 	return (0);
8326 }
8327 
8328 void
8329 xform_attach(void *data)
8330 {
8331 	struct xformsw *xsp = (struct xformsw *)data;
8332 
8333 	if (xform_register(xsp) != 0)
8334 		printf("%s: failed to register %s xform\n", __func__,
8335 		    xsp->xf_name);
8336 }
8337 
8338 void
8339 xform_detach(void *data)
8340 {
8341 	struct xformsw *xsp = (struct xformsw *)data;
8342 
8343 	XFORMS_LOCK();
8344 	LIST_REMOVE(xsp, chain);
8345 	XFORMS_UNLOCK();
8346 
8347 	/* Delete all SAs related to this xform. */
8348 	key_delete_xform(xsp);
8349 }
8350 
8351 /*
8352  * Initialize transform support in an sav.
8353  */
8354 static int
8355 xform_init(struct secasvar *sav, u_short xftype)
8356 {
8357 	struct xformsw *entry;
8358 	int ret;
8359 
8360 	IPSEC_ASSERT(sav->tdb_xform == NULL,
8361 	    ("tdb_xform is already initialized"));
8362 
8363 	ret = EINVAL;
8364 	XFORMS_LOCK();
8365 	LIST_FOREACH(entry, &xforms, chain) {
8366 	    if (entry->xf_type == xftype) {
8367 		    ret = (*entry->xf_init)(sav, entry);
8368 		    break;
8369 	    }
8370 	}
8371 	XFORMS_UNLOCK();
8372 	return (ret);
8373 }
8374 
8375