xref: /netbsd-src/sys/netipsec/key.c (revision f3cfa6f6ce31685c6c4a758bc430e69eb99f50a4)
1 /*	$NetBSD: key.c,v 1.261 2019/01/27 02:08:48 pgoyette Exp $	*/
2 /*	$FreeBSD: key.c,v 1.3.2.3 2004/02/14 22:23:23 bms Exp $	*/
3 /*	$KAME: key.c,v 1.191 2001/06/27 10:46:49 sakane Exp $	*/
4 
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 #include <sys/cdefs.h>
35 __KERNEL_RCSID(0, "$NetBSD: key.c,v 1.261 2019/01/27 02:08:48 pgoyette Exp $");
36 
37 /*
38  * This code is referred to RFC 2367
39  */
40 
41 #if defined(_KERNEL_OPT)
42 #include "opt_inet.h"
43 #include "opt_ipsec.h"
44 #include "opt_gateway.h"
45 #include "opt_net_mpsafe.h"
46 #endif
47 
48 #include <sys/types.h>
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/callout.h>
52 #include <sys/kernel.h>
53 #include <sys/mbuf.h>
54 #include <sys/domain.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sysctl.h>
58 #include <sys/errno.h>
59 #include <sys/proc.h>
60 #include <sys/queue.h>
61 #include <sys/syslog.h>
62 #include <sys/once.h>
63 #include <sys/cprng.h>
64 #include <sys/psref.h>
65 #include <sys/lwp.h>
66 #include <sys/workqueue.h>
67 #include <sys/kmem.h>
68 #include <sys/cpu.h>
69 #include <sys/atomic.h>
70 #include <sys/pslist.h>
71 #include <sys/mutex.h>
72 #include <sys/condvar.h>
73 #include <sys/localcount.h>
74 #include <sys/pserialize.h>
75 #include <sys/hash.h>
76 
77 #include <net/if.h>
78 #include <net/route.h>
79 
80 #include <netinet/in.h>
81 #include <netinet/in_systm.h>
82 #include <netinet/ip.h>
83 #include <netinet/in_var.h>
84 #ifdef INET
85 #include <netinet/ip_var.h>
86 #endif
87 
88 #ifdef INET6
89 #include <netinet/ip6.h>
90 #include <netinet6/in6_var.h>
91 #include <netinet6/ip6_var.h>
92 #endif /* INET6 */
93 
94 #ifdef INET
95 #include <netinet/in_pcb.h>
96 #endif
97 #ifdef INET6
98 #include <netinet6/in6_pcb.h>
99 #endif /* INET6 */
100 
101 #include <net/pfkeyv2.h>
102 #include <netipsec/keydb.h>
103 #include <netipsec/key.h>
104 #include <netipsec/keysock.h>
105 #include <netipsec/key_debug.h>
106 
107 #include <netipsec/ipsec.h>
108 #ifdef INET6
109 #include <netipsec/ipsec6.h>
110 #endif
111 #include <netipsec/ipsec_private.h>
112 
113 #include <netipsec/xform.h>
114 #include <netipsec/ipcomp.h>
115 
116 #define FULLMASK	0xffu
117 #define	_BITS(bytes)	((bytes) << 3)
118 
119 #define PORT_NONE	0
120 #define PORT_LOOSE	1
121 #define PORT_STRICT	2
122 
123 #ifndef SAHHASH_NHASH
124 #define SAHHASH_NHASH		128
125 #endif
126 
127 #ifndef SAVLUT_NHASH
128 #define SAVLUT_NHASH		128
129 #endif
130 
131 percpu_t *pfkeystat_percpu;
132 
133 /*
134  * Note on SA reference counting:
135  * - SAs that are not in DEAD state will have (total external reference + 1)
136  *   following value in reference count field.  they cannot be freed and are
137  *   referenced from SA header.
138  * - SAs that are in DEAD state will have (total external reference)
139  *   in reference count field.  they are ready to be freed.  reference from
140  *   SA header will be removed in key_delsav(), when the reference count
141  *   field hits 0 (= no external reference other than from SA header.
142  */
143 
144 u_int32_t key_debug_level = 0;
145 static u_int key_spi_trycnt = 1000;
146 static u_int32_t key_spi_minval = 0x100;
147 static u_int32_t key_spi_maxval = 0x0fffffff;	/* XXX */
148 static u_int32_t policy_id = 0;
149 static u_int key_int_random = 60;	/*interval to initialize randseed,1(m)*/
150 static u_int key_larval_lifetime = 30;	/* interval to expire acquiring, 30(s)*/
151 static int key_blockacq_count = 10;	/* counter for blocking SADB_ACQUIRE.*/
152 static int key_blockacq_lifetime = 20;	/* lifetime for blocking SADB_ACQUIRE.*/
153 static int key_prefered_oldsa = 0;	/* prefered old sa rather than new sa.*/
154 
155 static u_int32_t acq_seq = 0;
156 
157 /*
158  * Locking order: there is no order for now; it means that any locks aren't
159  * overlapped.
160  */
161 /*
162  * Locking notes on SPD:
163  * - Modifications to the key_spd.splist must be done with holding key_spd.lock
164  *   which is a adaptive mutex
165  * - Read accesses to the key_spd.splist must be in pserialize(9) read sections
166  * - SP's lifetime is managed by localcount(9)
167  * - An SP that has been inserted to the key_spd.splist is initially referenced
168  *   by none, i.e., a reference from the key_spd.splist isn't counted
169  * - When an SP is being destroyed, we change its state as DEAD, wait for
170  *   references to the SP to be released, and then deallocate the SP
171  *   (see key_unlink_sp)
172  * - Getting an SP
173  *   - Normally we get an SP from the key_spd.splist (see key_lookup_sp_byspidx)
174  *     - Must iterate the list and increment the reference count of a found SP
175  *       (by key_sp_ref) in a pserialize read section
176  *   - We can gain another reference from a held SP only if we check its state
177  *     and take its reference in a pserialize read section
178  *     (see esp_output for example)
179  *   - We may get an SP from an SP cache. See below
180  *   - A gotten SP must be released after use by KEY_SP_UNREF (key_sp_unref)
181  * - Updating member variables of an SP
182  *   - Most member variables of an SP are immutable
183  *   - Only sp->state and sp->lastused can be changed
184  *   - sp->state of an SP is updated only when destroying it under key_spd.lock
185  * - SP caches
186  *   - SPs can be cached in PCBs
187  *   - The lifetime of the caches is controlled by the global generation counter
188  *     (ipsec_spdgen)
189  *   - The global counter value is stored when an SP is cached
190  *   - If the stored value is different from the global counter then the cache
191  *     is considered invalidated
192  *   - The counter is incremented when an SP is being destroyed
193  *   - So checking the generation and taking a reference to an SP should be
194  *     in a pserialize read section
195  *   - Note that caching doesn't increment the reference counter of an SP
196  * - SPs in sockets
197  *   - Userland programs can set a policy to a socket by
198  *     setsockopt(IP_IPSEC_POLICY)
199  *   - Such policies (SPs) are set to a socket (PCB) and also inserted to
200  *     the key_spd.socksplist list (not the key_spd.splist)
201  *   - Such a policy is destroyed when a corresponding socket is destroed,
202  *     however, a socket can be destroyed in softint so we cannot destroy
203  *     it directly instead we just mark it DEAD and delay the destruction
204  *     until GC by the timer
205  * - SP origin
206  *   - SPs can be created by both userland programs and kernel components.
207  *     The SPs created in kernel must not be removed by userland programs,
208  *     although the SPs can be read by userland programs.
209  */
210 /*
211  * Locking notes on SAD:
212  * - Data structures
213  *   - SAs are managed by the list called key_sad.sahlists and sav lists of
214  *     sah entries
215  *     - An sav is supposed to be an SA from a viewpoint of users
216  *   - A sah has sav lists for each SA state
217  *   - Multiple saves with the same saidx can exist
218  *     - Only one entry has MATURE state and others should be DEAD
219  *     - DEAD entries are just ignored from searching
220  *   - All sav whose state is MATURE or DYING are registered to the lookup
221  *     table called key_sad.savlut in addition to the savlists.
222  *     - The table is used to search an sav without use of saidx.
223  * - Modifications to the key_sad.sahlists, sah.savlist and key_sad.savlut
224  *   must be done with holding key_sad.lock which is a adaptive mutex
225  * - Read accesses to the key_sad.sahlists, sah.savlist and key_sad.savlut
226  *   must be in pserialize(9) read sections
227  * - sah's lifetime is managed by localcount(9)
228  * - Getting an sah entry
229  *   - We get an sah from the key_sad.sahlists
230  *     - Must iterate the list and increment the reference count of a found sah
231  *       (by key_sah_ref) in a pserialize read section
232  *   - A gotten sah must be released after use by key_sah_unref
233  * - An sah is destroyed when its state become DEAD and no sav is
234  *   listed to the sah
235  *   - The destruction is done only in the timer (see key_timehandler_sad)
236  * - sav's lifetime is managed by localcount(9)
237  * - Getting an sav entry
238  *   - First get an sah by saidx and get an sav from either of sah's savlists
239  *     - Must iterate the list and increment the reference count of a found sav
240  *       (by key_sa_ref) in a pserialize read section
241  *   - We can gain another reference from a held SA only if we check its state
242  *     and take its reference in a pserialize read section
243  *     (see esp_output for example)
244  *   - A gotten sav must be released after use by key_sa_unref
245  * - An sav is destroyed when its state become DEAD
246  */
247 /*
248  * Locking notes on misc data:
249  * - All lists of key_misc are protected by key_misc.lock
250  *   - key_misc.lock must be held even for read accesses
251  */
252 
253 /* SPD */
254 static struct {
255 	kmutex_t lock;
256 	kcondvar_t cv_lc;
257 	struct pslist_head splist[IPSEC_DIR_MAX];
258 	/*
259 	 * The list has SPs that are set to a socket via
260 	 * setsockopt(IP_IPSEC_POLICY) from userland. See ipsec_set_policy.
261 	 */
262 	struct pslist_head socksplist;
263 
264 	pserialize_t psz;
265 	kcondvar_t cv_psz;
266 	bool psz_performing;
267 } key_spd __cacheline_aligned;
268 
269 /* SAD */
270 static struct {
271 	kmutex_t lock;
272 	kcondvar_t cv_lc;
273 	struct pslist_head *sahlists;
274 	u_long sahlistmask;
275 	struct pslist_head *savlut;
276 	u_long savlutmask;
277 
278 	pserialize_t psz;
279 	kcondvar_t cv_psz;
280 	bool psz_performing;
281 } key_sad __cacheline_aligned;
282 
283 /* Misc data */
284 static struct {
285 	kmutex_t lock;
286 	/* registed list */
287 	LIST_HEAD(_reglist, secreg) reglist[SADB_SATYPE_MAX + 1];
288 #ifndef IPSEC_NONBLOCK_ACQUIRE
289 	/* acquiring list */
290 	LIST_HEAD(_acqlist, secacq) acqlist;
291 #endif
292 #ifdef notyet
293 	/* SP acquiring list */
294 	LIST_HEAD(_spacqlist, secspacq) spacqlist;
295 #endif
296 } key_misc __cacheline_aligned;
297 
298 /* Macros for key_spd.splist */
299 #define SPLIST_ENTRY_INIT(sp)						\
300 	PSLIST_ENTRY_INIT((sp), pslist_entry)
301 #define SPLIST_ENTRY_DESTROY(sp)					\
302 	PSLIST_ENTRY_DESTROY((sp), pslist_entry)
303 #define SPLIST_WRITER_REMOVE(sp)					\
304 	PSLIST_WRITER_REMOVE((sp), pslist_entry)
305 #define SPLIST_READER_EMPTY(dir)					\
306 	(PSLIST_READER_FIRST(&key_spd.splist[(dir)], struct secpolicy,	\
307 	                     pslist_entry) == NULL)
308 #define SPLIST_READER_FOREACH(sp, dir)					\
309 	PSLIST_READER_FOREACH((sp), &key_spd.splist[(dir)],		\
310 	                      struct secpolicy, pslist_entry)
311 #define SPLIST_WRITER_FOREACH(sp, dir)					\
312 	PSLIST_WRITER_FOREACH((sp), &key_spd.splist[(dir)],		\
313 	                      struct secpolicy, pslist_entry)
314 #define SPLIST_WRITER_INSERT_AFTER(sp, new)				\
315 	PSLIST_WRITER_INSERT_AFTER((sp), (new), pslist_entry)
316 #define SPLIST_WRITER_EMPTY(dir)					\
317 	(PSLIST_WRITER_FIRST(&key_spd.splist[(dir)], struct secpolicy,	\
318 	                     pslist_entry) == NULL)
319 #define SPLIST_WRITER_INSERT_HEAD(dir, sp)				\
320 	PSLIST_WRITER_INSERT_HEAD(&key_spd.splist[(dir)], (sp),		\
321 	                          pslist_entry)
322 #define SPLIST_WRITER_NEXT(sp)						\
323 	PSLIST_WRITER_NEXT((sp), struct secpolicy, pslist_entry)
324 #define SPLIST_WRITER_INSERT_TAIL(dir, new)				\
325 	do {								\
326 		if (SPLIST_WRITER_EMPTY((dir))) {			\
327 			SPLIST_WRITER_INSERT_HEAD((dir), (new));	\
328 		} else {						\
329 			struct secpolicy *__sp;				\
330 			SPLIST_WRITER_FOREACH(__sp, (dir)) {		\
331 				if (SPLIST_WRITER_NEXT(__sp) == NULL) {	\
332 					SPLIST_WRITER_INSERT_AFTER(__sp,\
333 					    (new));			\
334 					break;				\
335 				}					\
336 			}						\
337 		}							\
338 	} while (0)
339 
340 /* Macros for key_spd.socksplist */
341 #define SOCKSPLIST_WRITER_FOREACH(sp)					\
342 	PSLIST_WRITER_FOREACH((sp), &key_spd.socksplist,		\
343 	                      struct secpolicy,	pslist_entry)
344 #define SOCKSPLIST_READER_EMPTY()					\
345 	(PSLIST_READER_FIRST(&key_spd.socksplist, struct secpolicy,	\
346 	                     pslist_entry) == NULL)
347 
348 /* Macros for key_sad.sahlist */
349 #define SAHLIST_ENTRY_INIT(sah)						\
350 	PSLIST_ENTRY_INIT((sah), pslist_entry)
351 #define SAHLIST_ENTRY_DESTROY(sah)					\
352 	PSLIST_ENTRY_DESTROY((sah), pslist_entry)
353 #define SAHLIST_WRITER_REMOVE(sah)					\
354 	PSLIST_WRITER_REMOVE((sah), pslist_entry)
355 #define SAHLIST_READER_FOREACH(sah)					\
356 	for(int _i_sah = 0; _i_sah <= key_sad.sahlistmask; _i_sah++)	\
357 		PSLIST_READER_FOREACH((sah), &key_sad.sahlists[_i_sah],	\
358 		                      struct secashead, pslist_entry)
359 #define SAHLIST_READER_FOREACH_SAIDX(sah, saidx)			\
360 	PSLIST_READER_FOREACH((sah),					\
361 	    &key_sad.sahlists[key_saidxhash((saidx),			\
362 	                       key_sad.sahlistmask)],			\
363 	    struct secashead, pslist_entry)
364 #define SAHLIST_WRITER_FOREACH(sah)					\
365 	for(int _i_sah = 0; _i_sah <= key_sad.sahlistmask; _i_sah++)	\
366 		PSLIST_WRITER_FOREACH((sah), &key_sad.sahlists[_i_sah],	\
367 		                     struct secashead, pslist_entry)
368 #define SAHLIST_WRITER_INSERT_HEAD(sah)					\
369 	PSLIST_WRITER_INSERT_HEAD(					\
370 	    &key_sad.sahlists[key_saidxhash(&(sah)->saidx,		\
371 	                      key_sad.sahlistmask)],	\
372 	    (sah), pslist_entry)
373 
374 /* Macros for key_sad.sahlist#savlist */
375 #define SAVLIST_ENTRY_INIT(sav)						\
376 	PSLIST_ENTRY_INIT((sav), pslist_entry)
377 #define SAVLIST_ENTRY_DESTROY(sav)					\
378 	PSLIST_ENTRY_DESTROY((sav), pslist_entry)
379 #define SAVLIST_READER_FIRST(sah, state)				\
380 	PSLIST_READER_FIRST(&(sah)->savlist[(state)], struct secasvar,	\
381 	                    pslist_entry)
382 #define SAVLIST_WRITER_REMOVE(sav)					\
383 	PSLIST_WRITER_REMOVE((sav), pslist_entry)
384 #define SAVLIST_READER_FOREACH(sav, sah, state)				\
385 	PSLIST_READER_FOREACH((sav), &(sah)->savlist[(state)],		\
386 	                      struct secasvar, pslist_entry)
387 #define SAVLIST_WRITER_FOREACH(sav, sah, state)				\
388 	PSLIST_WRITER_FOREACH((sav), &(sah)->savlist[(state)],		\
389 	                      struct secasvar, pslist_entry)
390 #define SAVLIST_WRITER_INSERT_BEFORE(sav, new)				\
391 	PSLIST_WRITER_INSERT_BEFORE((sav), (new), pslist_entry)
392 #define SAVLIST_WRITER_INSERT_AFTER(sav, new)				\
393 	PSLIST_WRITER_INSERT_AFTER((sav), (new), pslist_entry)
394 #define SAVLIST_WRITER_EMPTY(sah, state)				\
395 	(PSLIST_WRITER_FIRST(&(sah)->savlist[(state)], struct secasvar,	\
396 	                     pslist_entry) == NULL)
397 #define SAVLIST_WRITER_INSERT_HEAD(sah, state, sav)			\
398 	PSLIST_WRITER_INSERT_HEAD(&(sah)->savlist[(state)], (sav),	\
399 	                          pslist_entry)
400 #define SAVLIST_WRITER_NEXT(sav)					\
401 	PSLIST_WRITER_NEXT((sav), struct secasvar, pslist_entry)
402 #define SAVLIST_WRITER_INSERT_TAIL(sah, state, new)			\
403 	do {								\
404 		if (SAVLIST_WRITER_EMPTY((sah), (state))) {		\
405 			SAVLIST_WRITER_INSERT_HEAD((sah), (state), (new));\
406 		} else {						\
407 			struct secasvar *__sav;				\
408 			SAVLIST_WRITER_FOREACH(__sav, (sah), (state)) {	\
409 				if (SAVLIST_WRITER_NEXT(__sav) == NULL) {\
410 					SAVLIST_WRITER_INSERT_AFTER(__sav,\
411 					    (new));			\
412 					break;				\
413 				}					\
414 			}						\
415 		}							\
416 	} while (0)
417 #define SAVLIST_READER_NEXT(sav)					\
418 	PSLIST_READER_NEXT((sav), struct secasvar, pslist_entry)
419 
420 /* Macros for key_sad.savlut */
421 #define SAVLUT_ENTRY_INIT(sav)						\
422 	PSLIST_ENTRY_INIT((sav), pslist_entry_savlut)
423 #define SAVLUT_READER_FOREACH(sav, dst, proto, hash_key)		\
424 	PSLIST_READER_FOREACH((sav),					\
425 	&key_sad.savlut[key_savluthash(dst, proto, hash_key,		\
426 	                  key_sad.savlutmask)],				\
427 	struct secasvar, pslist_entry_savlut)
428 #define SAVLUT_WRITER_INSERT_HEAD(sav)					\
429 	key_savlut_writer_insert_head((sav))
430 #define SAVLUT_WRITER_REMOVE(sav)					\
431 	do {								\
432 		if (!(sav)->savlut_added)				\
433 			break;						\
434 		PSLIST_WRITER_REMOVE((sav), pslist_entry_savlut);	\
435 		(sav)->savlut_added = false;				\
436 	} while(0)
437 
438 /* search order for SAs */
439 	/*
440 	 * This order is important because we must select the oldest SA
441 	 * for outbound processing.  For inbound, This is not important.
442 	 */
443 static const u_int saorder_state_valid_prefer_old[] = {
444 	SADB_SASTATE_DYING, SADB_SASTATE_MATURE,
445 };
446 static const u_int saorder_state_valid_prefer_new[] = {
447 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING,
448 };
449 
450 static const u_int saorder_state_alive[] = {
451 	/* except DEAD */
452 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING, SADB_SASTATE_LARVAL
453 };
454 static const u_int saorder_state_any[] = {
455 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING,
456 	SADB_SASTATE_LARVAL, SADB_SASTATE_DEAD
457 };
458 
459 #define SASTATE_ALIVE_FOREACH(s)				\
460 	for (int _i = 0;					\
461 	    _i < __arraycount(saorder_state_alive) ?		\
462 	    (s) = saorder_state_alive[_i], true : false;	\
463 	    _i++)
464 #define SASTATE_ANY_FOREACH(s)					\
465 	for (int _i = 0;					\
466 	    _i < __arraycount(saorder_state_any) ?		\
467 	    (s) = saorder_state_any[_i], true : false;		\
468 	    _i++)
469 #define SASTATE_USABLE_FOREACH(s)				\
470 	for (int _i = 0;					\
471 	    _i < __arraycount(saorder_state_valid_prefer_new) ?	\
472 	    (s) = saorder_state_valid_prefer_new[_i],		\
473 	    true : false;					\
474 	    _i++)
475 
476 static const int minsize[] = {
477 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
478 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
479 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
480 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
481 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
482 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_SRC */
483 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_DST */
484 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_PROXY */
485 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_AUTH */
486 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_ENCRYPT */
487 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_SRC */
488 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_DST */
489 	sizeof(struct sadb_sens),	/* SADB_EXT_SENSITIVITY */
490 	sizeof(struct sadb_prop),	/* SADB_EXT_PROPOSAL */
491 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_AUTH */
492 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_ENCRYPT */
493 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
494 	0,				/* SADB_X_EXT_KMPRIVATE */
495 	sizeof(struct sadb_x_policy),	/* SADB_X_EXT_POLICY */
496 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
497 	sizeof(struct sadb_x_nat_t_type),	/* SADB_X_EXT_NAT_T_TYPE */
498 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_SPORT */
499 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_DPORT */
500 	sizeof(struct sadb_address),		/* SADB_X_EXT_NAT_T_OAI */
501 	sizeof(struct sadb_address),		/* SADB_X_EXT_NAT_T_OAR */
502 	sizeof(struct sadb_x_nat_t_frag),	/* SADB_X_EXT_NAT_T_FRAG */
503 };
504 static const int maxsize[] = {
505 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
506 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
507 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
508 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
509 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
510 	0,				/* SADB_EXT_ADDRESS_SRC */
511 	0,				/* SADB_EXT_ADDRESS_DST */
512 	0,				/* SADB_EXT_ADDRESS_PROXY */
513 	0,				/* SADB_EXT_KEY_AUTH */
514 	0,				/* SADB_EXT_KEY_ENCRYPT */
515 	0,				/* SADB_EXT_IDENTITY_SRC */
516 	0,				/* SADB_EXT_IDENTITY_DST */
517 	0,				/* SADB_EXT_SENSITIVITY */
518 	0,				/* SADB_EXT_PROPOSAL */
519 	0,				/* SADB_EXT_SUPPORTED_AUTH */
520 	0,				/* SADB_EXT_SUPPORTED_ENCRYPT */
521 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
522 	0,				/* SADB_X_EXT_KMPRIVATE */
523 	0,				/* SADB_X_EXT_POLICY */
524 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
525 	sizeof(struct sadb_x_nat_t_type),	/* SADB_X_EXT_NAT_T_TYPE */
526 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_SPORT */
527 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_DPORT */
528 	0,					/* SADB_X_EXT_NAT_T_OAI */
529 	0,					/* SADB_X_EXT_NAT_T_OAR */
530 	sizeof(struct sadb_x_nat_t_frag),	/* SADB_X_EXT_NAT_T_FRAG */
531 };
532 
533 static int ipsec_esp_keymin = 256;
534 static int ipsec_esp_auth = 0;
535 static int ipsec_ah_keymin = 128;
536 
537 #ifdef SYSCTL_DECL
538 SYSCTL_DECL(_net_key);
539 #endif
540 
541 #ifdef SYSCTL_INT
542 SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,	CTLFLAG_RW, \
543 	&key_debug_level,	0,	"");
544 
545 /* max count of trial for the decision of spi value */
546 SYSCTL_INT(_net_key, KEYCTL_SPI_TRY,		spi_trycnt,	CTLFLAG_RW, \
547 	&key_spi_trycnt,	0,	"");
548 
549 /* minimum spi value to allocate automatically. */
550 SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE,	spi_minval,	CTLFLAG_RW, \
551 	&key_spi_minval,	0,	"");
552 
553 /* maximun spi value to allocate automatically. */
554 SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE,	spi_maxval,	CTLFLAG_RW, \
555 	&key_spi_maxval,	0,	"");
556 
557 /* interval to initialize randseed */
558 SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT,	int_random,	CTLFLAG_RW, \
559 	&key_int_random,	0,	"");
560 
561 /* lifetime for larval SA */
562 SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME,	larval_lifetime, CTLFLAG_RW, \
563 	&key_larval_lifetime,	0,	"");
564 
565 /* counter for blocking to send SADB_ACQUIRE to IKEd */
566 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT,	blockacq_count,	CTLFLAG_RW, \
567 	&key_blockacq_count,	0,	"");
568 
569 /* lifetime for blocking to send SADB_ACQUIRE to IKEd */
570 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME,	blockacq_lifetime, CTLFLAG_RW, \
571 	&key_blockacq_lifetime,	0,	"");
572 
573 /* ESP auth */
574 SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH,	esp_auth, CTLFLAG_RW, \
575 	&ipsec_esp_auth,	0,	"");
576 
577 /* minimum ESP key length */
578 SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN,	esp_keymin, CTLFLAG_RW, \
579 	&ipsec_esp_keymin,	0,	"");
580 
581 /* minimum AH key length */
582 SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN,	ah_keymin, CTLFLAG_RW, \
583 	&ipsec_ah_keymin,	0,	"");
584 
585 /* perfered old SA rather than new SA */
586 SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA,	prefered_oldsa, CTLFLAG_RW,\
587 	&key_prefered_oldsa,	0,	"");
588 #endif /* SYSCTL_INT */
589 
590 #define __LIST_CHAINED(elm) \
591 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
592 #define LIST_INSERT_TAIL(head, elm, type, field) \
593 do {\
594 	struct type *curelm = LIST_FIRST(head); \
595 	if (curelm == NULL) {\
596 		LIST_INSERT_HEAD(head, elm, field); \
597 	} else { \
598 		while (LIST_NEXT(curelm, field)) \
599 			curelm = LIST_NEXT(curelm, field);\
600 		LIST_INSERT_AFTER(curelm, elm, field);\
601 	}\
602 } while (0)
603 
604 #define KEY_CHKSASTATE(head, sav) \
605 /* do */ { \
606 	if ((head) != (sav)) {						\
607 		IPSECLOG(LOG_DEBUG,					\
608 		    "state mismatched (TREE=%d SA=%d)\n",		\
609 		    (head), (sav));					\
610 		continue;						\
611 	}								\
612 } /* while (0) */
613 
614 #define KEY_CHKSPDIR(head, sp) \
615 do { \
616 	if ((head) != (sp)) {						\
617 		IPSECLOG(LOG_DEBUG,					\
618 		    "direction mismatched (TREE=%d SP=%d), anyway continue.\n",\
619 		    (head), (sp));					\
620 	}								\
621 } while (0)
622 
623 /*
624  * set parameters into secasindex buffer.
625  * Must allocate secasindex buffer before calling this function.
626  */
627 static int
628 key_setsecasidx(int, int, int, const struct sockaddr *,
629     const struct sockaddr *, struct secasindex *);
630 
631 /* key statistics */
632 struct _keystat {
633 	u_long getspi_count; /* the avarage of count to try to get new SPI */
634 } keystat;
635 
636 static void
637 key_init_spidx_bymsghdr(struct secpolicyindex *, const struct sadb_msghdr *);
638 
639 static const struct sockaddr *
640 key_msghdr_get_sockaddr(const struct sadb_msghdr *mhp, int idx)
641 {
642 
643 	return PFKEY_ADDR_SADDR(mhp->ext[idx]);
644 }
645 
646 static void
647 key_fill_replymsg(struct mbuf *m, int seq)
648 {
649 	struct sadb_msg *msg;
650 
651 	KASSERT(m->m_len >= sizeof(*msg));
652 
653 	msg = mtod(m, struct sadb_msg *);
654 	msg->sadb_msg_errno = 0;
655 	msg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
656 	if (seq != 0)
657 		msg->sadb_msg_seq = seq;
658 }
659 
660 #if 0
661 static void key_freeso(struct socket *);
662 static void key_freesp_so(struct secpolicy **);
663 #endif
664 static struct secpolicy *key_getsp (const struct secpolicyindex *);
665 static struct secpolicy *key_getspbyid (u_int32_t);
666 static struct secpolicy *key_lookup_and_remove_sp(const struct secpolicyindex *, bool);
667 static struct secpolicy *key_lookupbyid_and_remove_sp(u_int32_t, bool);
668 static void key_destroy_sp(struct secpolicy *);
669 static struct mbuf *key_gather_mbuf (struct mbuf *,
670 	const struct sadb_msghdr *, int, int, ...);
671 static int key_api_spdadd(struct socket *, struct mbuf *,
672 	const struct sadb_msghdr *);
673 static u_int32_t key_getnewspid (void);
674 static int key_api_spddelete(struct socket *, struct mbuf *,
675 	const struct sadb_msghdr *);
676 static int key_api_spddelete2(struct socket *, struct mbuf *,
677 	const struct sadb_msghdr *);
678 static int key_api_spdget(struct socket *, struct mbuf *,
679 	const struct sadb_msghdr *);
680 static int key_api_spdflush(struct socket *, struct mbuf *,
681 	const struct sadb_msghdr *);
682 static int key_api_spddump(struct socket *, struct mbuf *,
683 	const struct sadb_msghdr *);
684 static struct mbuf * key_setspddump (int *errorp, pid_t);
685 static struct mbuf * key_setspddump_chain (int *errorp, int *lenp, pid_t pid);
686 static int key_api_nat_map(struct socket *, struct mbuf *,
687 	const struct sadb_msghdr *);
688 static struct mbuf *key_setdumpsp (struct secpolicy *,
689 	u_int8_t, u_int32_t, pid_t);
690 static u_int key_getspreqmsglen (const struct secpolicy *);
691 static int key_spdexpire (struct secpolicy *);
692 static struct secashead *key_newsah (const struct secasindex *);
693 static void key_unlink_sah(struct secashead *);
694 static void key_destroy_sah(struct secashead *);
695 static bool key_sah_has_sav(struct secashead *);
696 static void key_sah_ref(struct secashead *);
697 static void key_sah_unref(struct secashead *);
698 static void key_init_sav(struct secasvar *);
699 static void key_destroy_sav(struct secasvar *);
700 static void key_destroy_sav_with_ref(struct secasvar *);
701 static struct secasvar *key_newsav(struct mbuf *,
702 	const struct sadb_msghdr *, int *, const char*, int);
703 #define	KEY_NEWSAV(m, sadb, e)				\
704 	key_newsav(m, sadb, e, __func__, __LINE__)
705 static void key_delsav (struct secasvar *);
706 static struct secashead *key_getsah(const struct secasindex *, int);
707 static struct secashead *key_getsah_ref(const struct secasindex *, int);
708 static bool key_checkspidup(const struct secasindex *, u_int32_t);
709 static struct secasvar *key_getsavbyspi (struct secashead *, u_int32_t);
710 static int key_setsaval (struct secasvar *, struct mbuf *,
711 	const struct sadb_msghdr *);
712 static void key_freesaval(struct secasvar *);
713 static int key_init_xform(struct secasvar *);
714 static void key_clear_xform(struct secasvar *);
715 static struct mbuf *key_setdumpsa (struct secasvar *, u_int8_t,
716 	u_int8_t, u_int32_t, u_int32_t);
717 static struct mbuf *key_setsadbxport (u_int16_t, u_int16_t);
718 static struct mbuf *key_setsadbxtype (u_int16_t);
719 static struct mbuf *key_setsadbxfrag (u_int16_t);
720 static void key_porttosaddr (union sockaddr_union *, u_int16_t);
721 static int key_checksalen (const union sockaddr_union *);
722 static struct mbuf *key_setsadbmsg (u_int8_t, u_int16_t, u_int8_t,
723 	u_int32_t, pid_t, u_int16_t, int);
724 static struct mbuf *key_setsadbsa (struct secasvar *);
725 static struct mbuf *key_setsadbaddr(u_int16_t,
726 	const struct sockaddr *, u_int8_t, u_int16_t, int);
727 #if 0
728 static struct mbuf *key_setsadbident (u_int16_t, u_int16_t, void *,
729 	int, u_int64_t);
730 #endif
731 static struct mbuf *key_setsadbxsa2 (u_int8_t, u_int32_t, u_int16_t);
732 static struct mbuf *key_setsadbxpolicy (u_int16_t, u_int8_t,
733 	u_int32_t, int);
734 static void *key_newbuf (const void *, u_int);
735 #ifdef INET6
736 static int key_ismyaddr6 (const struct sockaddr_in6 *);
737 #endif
738 
739 static void sysctl_net_keyv2_setup(struct sysctllog **);
740 static void sysctl_net_key_compat_setup(struct sysctllog **);
741 
742 /* flags for key_saidx_match() */
743 #define CMP_HEAD	1	/* protocol, addresses. */
744 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
745 #define CMP_REQID	3	/* additionally HEAD, reaid. */
746 #define CMP_EXACTLY	4	/* all elements. */
747 static int key_saidx_match(const struct secasindex *,
748     const struct secasindex *, int);
749 
750 static int key_sockaddr_match(const struct sockaddr *,
751     const struct sockaddr *, int);
752 static int key_bb_match_withmask(const void *, const void *, u_int);
753 static u_int16_t key_satype2proto (u_int8_t);
754 static u_int8_t key_proto2satype (u_int16_t);
755 
756 static int key_spidx_match_exactly(const struct secpolicyindex *,
757     const struct secpolicyindex *);
758 static int key_spidx_match_withmask(const struct secpolicyindex *,
759     const struct secpolicyindex *);
760 
761 static int key_api_getspi(struct socket *, struct mbuf *,
762 	const struct sadb_msghdr *);
763 static u_int32_t key_do_getnewspi (const struct sadb_spirange *,
764 					const struct secasindex *);
765 static int key_handle_natt_info (struct secasvar *,
766 				     const struct sadb_msghdr *);
767 static int key_set_natt_ports (union sockaddr_union *,
768 			 	union sockaddr_union *,
769 				const struct sadb_msghdr *);
770 static int key_api_update(struct socket *, struct mbuf *,
771 	const struct sadb_msghdr *);
772 #ifdef IPSEC_DOSEQCHECK
773 static struct secasvar *key_getsavbyseq (struct secashead *, u_int32_t);
774 #endif
775 static int key_api_add(struct socket *, struct mbuf *,
776 	const struct sadb_msghdr *);
777 static int key_setident (struct secashead *, struct mbuf *,
778 	const struct sadb_msghdr *);
779 static struct mbuf *key_getmsgbuf_x1 (struct mbuf *,
780 	const struct sadb_msghdr *);
781 static int key_api_delete(struct socket *, struct mbuf *,
782 	const struct sadb_msghdr *);
783 static int key_api_get(struct socket *, struct mbuf *,
784 	const struct sadb_msghdr *);
785 
786 static void key_getcomb_setlifetime (struct sadb_comb *);
787 static struct mbuf *key_getcomb_esp(int);
788 static struct mbuf *key_getcomb_ah(int);
789 static struct mbuf *key_getcomb_ipcomp(int);
790 static struct mbuf *key_getprop(const struct secasindex *, int);
791 
792 static int key_acquire(const struct secasindex *, const struct secpolicy *,
793 	    int);
794 static int key_acquire_sendup_mbuf_later(struct mbuf *);
795 static void key_acquire_sendup_pending_mbuf(void);
796 #ifndef IPSEC_NONBLOCK_ACQUIRE
797 static struct secacq *key_newacq (const struct secasindex *);
798 static struct secacq *key_getacq (const struct secasindex *);
799 static struct secacq *key_getacqbyseq (u_int32_t);
800 #endif
801 #ifdef notyet
802 static struct secspacq *key_newspacq (const struct secpolicyindex *);
803 static struct secspacq *key_getspacq (const struct secpolicyindex *);
804 #endif
805 static int key_api_acquire(struct socket *, struct mbuf *,
806 	const struct sadb_msghdr *);
807 static int key_api_register(struct socket *, struct mbuf *,
808 	const struct sadb_msghdr *);
809 static int key_expire (struct secasvar *);
810 static int key_api_flush(struct socket *, struct mbuf *,
811 	const struct sadb_msghdr *);
812 static struct mbuf *key_setdump_chain (u_int8_t req_satype, int *errorp,
813 	int *lenp, pid_t pid);
814 static int key_api_dump(struct socket *, struct mbuf *,
815 	const struct sadb_msghdr *);
816 static int key_api_promisc(struct socket *, struct mbuf *,
817 	const struct sadb_msghdr *);
818 static int key_senderror (struct socket *, struct mbuf *, int);
819 static int key_validate_ext (const struct sadb_ext *, int);
820 static int key_align (struct mbuf *, struct sadb_msghdr *);
821 #if 0
822 static const char *key_getfqdn (void);
823 static const char *key_getuserfqdn (void);
824 #endif
825 static void key_sa_chgstate (struct secasvar *, u_int8_t);
826 
827 static struct mbuf *key_alloc_mbuf(int, int);
828 static struct mbuf *key_alloc_mbuf_simple(int, int);
829 
830 static void key_timehandler(void *);
831 static void key_timehandler_work(struct work *, void *);
832 static struct callout	key_timehandler_ch;
833 static struct workqueue	*key_timehandler_wq;
834 static struct work	key_timehandler_wk;
835 
836 static inline void
837     key_savlut_writer_insert_head(struct secasvar *sav);
838 static inline uint32_t
839     key_saidxhash(const struct secasindex *, u_long);
840 static inline uint32_t
841     key_savluthash(const struct sockaddr *,
842     uint32_t, uint32_t, u_long);
843 
844 /*
845  * Utilities for percpu counters for sadb_lifetime_allocations and
846  * sadb_lifetime_bytes.
847  */
848 #define LIFETIME_COUNTER_ALLOCATIONS	0
849 #define LIFETIME_COUNTER_BYTES		1
850 #define LIFETIME_COUNTER_SIZE		2
851 
852 typedef uint64_t lifetime_counters_t[LIFETIME_COUNTER_SIZE];
853 
854 static void
855 key_sum_lifetime_counters(void *p, void *arg, struct cpu_info *ci __unused)
856 {
857 	lifetime_counters_t *one = p;
858 	lifetime_counters_t *sum = arg;
859 
860 	(*sum)[LIFETIME_COUNTER_ALLOCATIONS] += (*one)[LIFETIME_COUNTER_ALLOCATIONS];
861 	(*sum)[LIFETIME_COUNTER_BYTES] += (*one)[LIFETIME_COUNTER_BYTES];
862 }
863 
864 u_int
865 key_sp_refcnt(const struct secpolicy *sp)
866 {
867 
868 	/* FIXME */
869 	return 0;
870 }
871 
872 static void
873 key_spd_pserialize_perform(void)
874 {
875 
876 	KASSERT(mutex_owned(&key_spd.lock));
877 
878 	while (key_spd.psz_performing)
879 		cv_wait(&key_spd.cv_psz, &key_spd.lock);
880 	key_spd.psz_performing = true;
881 	mutex_exit(&key_spd.lock);
882 
883 	pserialize_perform(key_spd.psz);
884 
885 	mutex_enter(&key_spd.lock);
886 	key_spd.psz_performing = false;
887 	cv_broadcast(&key_spd.cv_psz);
888 }
889 
890 /*
891  * Remove the sp from the key_spd.splist and wait for references to the sp
892  * to be released. key_spd.lock must be held.
893  */
894 static void
895 key_unlink_sp(struct secpolicy *sp)
896 {
897 
898 	KASSERT(mutex_owned(&key_spd.lock));
899 
900 	sp->state = IPSEC_SPSTATE_DEAD;
901 	SPLIST_WRITER_REMOVE(sp);
902 
903 	/* Invalidate all cached SPD pointers in the PCBs. */
904 	ipsec_invalpcbcacheall();
905 
906 	KDASSERT(mutex_ownable(softnet_lock));
907 	key_spd_pserialize_perform();
908 
909 	localcount_drain(&sp->localcount, &key_spd.cv_lc, &key_spd.lock);
910 }
911 
912 /*
913  * Return 0 when there are known to be no SP's for the specified
914  * direction.  Otherwise return 1.  This is used by IPsec code
915  * to optimize performance.
916  */
917 int
918 key_havesp(u_int dir)
919 {
920 	return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ?
921 		!SPLIST_READER_EMPTY(dir) : 1);
922 }
923 
924 /* %%% IPsec policy management */
925 /*
926  * allocating a SP for OUTBOUND or INBOUND packet.
927  * Must call key_freesp() later.
928  * OUT:	NULL:	not found
929  *	others:	found and return the pointer.
930  */
931 struct secpolicy *
932 key_lookup_sp_byspidx(const struct secpolicyindex *spidx,
933     u_int dir, const char* where, int tag)
934 {
935 	struct secpolicy *sp;
936 	int s;
937 
938 	KASSERT(spidx != NULL);
939 	KASSERTMSG(IPSEC_DIR_IS_INOROUT(dir), "invalid direction %u", dir);
940 
941 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
942 
943 	/* get a SP entry */
944 	if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
945 		kdebug_secpolicyindex("objects", spidx);
946 	}
947 
948 	s = pserialize_read_enter();
949 	SPLIST_READER_FOREACH(sp, dir) {
950 		if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
951 			kdebug_secpolicyindex("in SPD", &sp->spidx);
952 		}
953 
954 		if (sp->state == IPSEC_SPSTATE_DEAD)
955 			continue;
956 		if (key_spidx_match_withmask(&sp->spidx, spidx))
957 			goto found;
958 	}
959 	sp = NULL;
960 found:
961 	if (sp) {
962 		/* sanity check */
963 		KEY_CHKSPDIR(sp->spidx.dir, dir);
964 
965 		/* found a SPD entry */
966 		sp->lastused = time_uptime;
967 		key_sp_ref(sp, where, tag);
968 	}
969 	pserialize_read_exit(s);
970 
971 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
972 	    "DP return SP:%p (ID=%u) refcnt %u\n",
973 	    sp, sp ? sp->id : 0, key_sp_refcnt(sp));
974 	return sp;
975 }
976 
977 /*
978  * return a policy that matches this particular inbound packet.
979  * XXX slow
980  */
981 struct secpolicy *
982 key_gettunnel(const struct sockaddr *osrc,
983 	      const struct sockaddr *odst,
984 	      const struct sockaddr *isrc,
985 	      const struct sockaddr *idst,
986 	      const char* where, int tag)
987 {
988 	struct secpolicy *sp;
989 	const int dir = IPSEC_DIR_INBOUND;
990 	int s;
991 	struct ipsecrequest *r1, *r2, *p;
992 	struct secpolicyindex spidx;
993 
994 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
995 
996 	if (isrc->sa_family != idst->sa_family) {
997 		IPSECLOG(LOG_ERR,
998 		    "address family mismatched src %u, dst %u.\n",
999 		    isrc->sa_family, idst->sa_family);
1000 		sp = NULL;
1001 		goto done;
1002 	}
1003 
1004 	s = pserialize_read_enter();
1005 	SPLIST_READER_FOREACH(sp, dir) {
1006 		if (sp->state == IPSEC_SPSTATE_DEAD)
1007 			continue;
1008 
1009 		r1 = r2 = NULL;
1010 		for (p = sp->req; p; p = p->next) {
1011 			if (p->saidx.mode != IPSEC_MODE_TUNNEL)
1012 				continue;
1013 
1014 			r1 = r2;
1015 			r2 = p;
1016 
1017 			if (!r1) {
1018 				/* here we look at address matches only */
1019 				spidx = sp->spidx;
1020 				if (isrc->sa_len > sizeof(spidx.src) ||
1021 				    idst->sa_len > sizeof(spidx.dst))
1022 					continue;
1023 				memcpy(&spidx.src, isrc, isrc->sa_len);
1024 				memcpy(&spidx.dst, idst, idst->sa_len);
1025 				if (!key_spidx_match_withmask(&sp->spidx, &spidx))
1026 					continue;
1027 			} else {
1028 				if (!key_sockaddr_match(&r1->saidx.src.sa, isrc, PORT_NONE) ||
1029 				    !key_sockaddr_match(&r1->saidx.dst.sa, idst, PORT_NONE))
1030 					continue;
1031 			}
1032 
1033 			if (!key_sockaddr_match(&r2->saidx.src.sa, osrc, PORT_NONE) ||
1034 			    !key_sockaddr_match(&r2->saidx.dst.sa, odst, PORT_NONE))
1035 				continue;
1036 
1037 			goto found;
1038 		}
1039 	}
1040 	sp = NULL;
1041 found:
1042 	if (sp) {
1043 		sp->lastused = time_uptime;
1044 		key_sp_ref(sp, where, tag);
1045 	}
1046 	pserialize_read_exit(s);
1047 done:
1048 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1049 	    "DP return SP:%p (ID=%u) refcnt %u\n",
1050 	    sp, sp ? sp->id : 0, key_sp_refcnt(sp));
1051 	return sp;
1052 }
1053 
1054 /*
1055  * allocating an SA entry for an *OUTBOUND* packet.
1056  * checking each request entries in SP, and acquire an SA if need.
1057  * OUT:	0: there are valid requests.
1058  *	ENOENT: policy may be valid, but SA with REQUIRE is on acquiring.
1059  */
1060 int
1061 key_checkrequest(const struct ipsecrequest *isr, const struct secasindex *saidx,
1062     struct secasvar **ret)
1063 {
1064 	u_int level;
1065 	int error;
1066 	struct secasvar *sav;
1067 
1068 	KASSERT(isr != NULL);
1069 	KASSERTMSG(saidx->mode == IPSEC_MODE_TRANSPORT ||
1070 	    saidx->mode == IPSEC_MODE_TUNNEL,
1071 	    "unexpected policy %u", saidx->mode);
1072 
1073 	/* get current level */
1074 	level = ipsec_get_reqlevel(isr);
1075 
1076 	/*
1077 	 * XXX guard against protocol callbacks from the crypto
1078 	 * thread as they reference ipsecrequest.sav which we
1079 	 * temporarily null out below.  Need to rethink how we
1080 	 * handle bundled SA's in the callback thread.
1081 	 */
1082 
1083 	sav = key_lookup_sa_bysaidx(saidx);
1084 	if (sav != NULL) {
1085 		*ret = sav;
1086 		return 0;
1087 	}
1088 
1089 	/* there is no SA */
1090 	error = key_acquire(saidx, isr->sp, M_NOWAIT);
1091 	if (error != 0) {
1092 		/* XXX What should I do ? */
1093 		IPSECLOG(LOG_DEBUG, "error %d returned from key_acquire.\n",
1094 		    error);
1095 		return error;
1096 	}
1097 
1098 	if (level != IPSEC_LEVEL_REQUIRE) {
1099 		/* XXX sigh, the interface to this routine is botched */
1100 		*ret = NULL;
1101 		return 0;
1102 	} else {
1103 		return ENOENT;
1104 	}
1105 }
1106 
1107 /*
1108  * looking up a SA for policy entry from SAD.
1109  * NOTE: searching SAD of aliving state.
1110  * OUT:	NULL:	not found.
1111  *	others:	found and return the pointer.
1112  */
1113 struct secasvar *
1114 key_lookup_sa_bysaidx(const struct secasindex *saidx)
1115 {
1116 	struct secashead *sah;
1117 	struct secasvar *sav = NULL;
1118 	u_int stateidx, state;
1119 	const u_int *saorder_state_valid;
1120 	int arraysize;
1121 	int s;
1122 
1123 	s = pserialize_read_enter();
1124 	sah = key_getsah(saidx, CMP_MODE_REQID);
1125 	if (sah == NULL)
1126 		goto out;
1127 
1128 	/*
1129 	 * search a valid state list for outbound packet.
1130 	 * This search order is important.
1131 	 */
1132 	if (key_prefered_oldsa) {
1133 		saorder_state_valid = saorder_state_valid_prefer_old;
1134 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_old);
1135 	} else {
1136 		saorder_state_valid = saorder_state_valid_prefer_new;
1137 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_new);
1138 	}
1139 
1140 	/* search valid state */
1141 	for (stateidx = 0;
1142 	     stateidx < arraysize;
1143 	     stateidx++) {
1144 
1145 		state = saorder_state_valid[stateidx];
1146 
1147 		if (key_prefered_oldsa)
1148 			sav = SAVLIST_READER_FIRST(sah, state);
1149 		else {
1150 			/* XXX need O(1) lookup */
1151 			struct secasvar *last = NULL;
1152 
1153 			SAVLIST_READER_FOREACH(sav, sah, state)
1154 				last = sav;
1155 			sav = last;
1156 		}
1157 		if (sav != NULL) {
1158 			KEY_SA_REF(sav);
1159 			KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1160 			    "DP cause refcnt++:%d SA:%p\n",
1161 			    key_sa_refcnt(sav), sav);
1162 			break;
1163 		}
1164 	}
1165 out:
1166 	pserialize_read_exit(s);
1167 
1168 	return sav;
1169 }
1170 
1171 #if 0
1172 static void
1173 key_sendup_message_delete(struct secasvar *sav)
1174 {
1175 	struct mbuf *m, *result = 0;
1176 	uint8_t satype;
1177 
1178 	satype = key_proto2satype(sav->sah->saidx.proto);
1179 	if (satype == 0)
1180 		goto msgfail;
1181 
1182 	m = key_setsadbmsg(SADB_DELETE, 0, satype, 0, 0, key_sa_refcnt(sav) - 1);
1183 	if (m == NULL)
1184 		goto msgfail;
1185 	result = m;
1186 
1187 	/* set sadb_address for saidx's. */
1188 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa,
1189 	    _BITS(sav->sah->saidx.src.sa.sa_len), IPSEC_ULPROTO_ANY);
1190 	if (m == NULL)
1191 		goto msgfail;
1192 	m_cat(result, m);
1193 
1194 	/* set sadb_address for saidx's. */
1195 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.src.sa,
1196 	    _BITS(sav->sah->saidx.src.sa.sa_len), IPSEC_ULPROTO_ANY);
1197 	if (m == NULL)
1198 		goto msgfail;
1199 	m_cat(result, m);
1200 
1201 	/* create SA extension */
1202 	m = key_setsadbsa(sav);
1203 	if (m == NULL)
1204 		goto msgfail;
1205 	m_cat(result, m);
1206 
1207 	if (result->m_len < sizeof(struct sadb_msg)) {
1208 		result = m_pullup(result, sizeof(struct sadb_msg));
1209 		if (result == NULL)
1210 			goto msgfail;
1211 	}
1212 
1213 	result->m_pkthdr.len = 0;
1214 	for (m = result; m; m = m->m_next)
1215 		result->m_pkthdr.len += m->m_len;
1216 	mtod(result, struct sadb_msg *)->sadb_msg_len =
1217 	    PFKEY_UNIT64(result->m_pkthdr.len);
1218 
1219 	key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
1220 	result = NULL;
1221 msgfail:
1222 	if (result)
1223 		m_freem(result);
1224 }
1225 #endif
1226 
1227 /*
1228  * allocating a usable SA entry for a *INBOUND* packet.
1229  * Must call key_freesav() later.
1230  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
1231  *	NULL:		not found, or error occurred.
1232  *
1233  * In the comparison, no source address is used--for RFC2401 conformance.
1234  * To quote, from section 4.1:
1235  *	A security association is uniquely identified by a triple consisting
1236  *	of a Security Parameter Index (SPI), an IP Destination Address, and a
1237  *	security protocol (AH or ESP) identifier.
1238  * Note that, however, we do need to keep source address in IPsec SA.
1239  * IKE specification and PF_KEY specification do assume that we
1240  * keep source address in IPsec SA.  We see a tricky situation here.
1241  *
1242  * sport and dport are used for NAT-T. network order is always used.
1243  */
1244 struct secasvar *
1245 key_lookup_sa(
1246 	const union sockaddr_union *dst,
1247 	u_int proto,
1248 	u_int32_t spi,
1249 	u_int16_t sport,
1250 	u_int16_t dport,
1251 	const char* where, int tag)
1252 {
1253 	struct secasvar *sav;
1254 	int chkport;
1255 	int s;
1256 
1257 	int must_check_spi = 1;
1258 	int must_check_alg = 0;
1259 	u_int16_t cpi = 0;
1260 	u_int8_t algo = 0;
1261 	uint32_t hash_key = spi;
1262 
1263 	if ((sport != 0) && (dport != 0))
1264 		chkport = PORT_STRICT;
1265 	else
1266 		chkport = PORT_NONE;
1267 
1268 	KASSERT(dst != NULL);
1269 
1270 	/*
1271 	 * XXX IPCOMP case
1272 	 * We use cpi to define spi here. In the case where cpi <=
1273 	 * IPCOMP_CPI_NEGOTIATE_MIN, cpi just define the algorithm used, not
1274 	 * the real spi. In this case, don't check the spi but check the
1275 	 * algorithm
1276 	 */
1277 
1278 	if (proto == IPPROTO_IPCOMP) {
1279 		u_int32_t tmp;
1280 		tmp = ntohl(spi);
1281 		cpi = (u_int16_t) tmp;
1282 		if (cpi < IPCOMP_CPI_NEGOTIATE_MIN) {
1283 			algo = (u_int8_t) cpi;
1284 			hash_key = algo;
1285 			must_check_spi = 0;
1286 			must_check_alg = 1;
1287 		}
1288 	}
1289 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1290 	    "DP from %s:%u check_spi=%d, check_alg=%d\n",
1291 	    where, tag, must_check_spi, must_check_alg);
1292 
1293 
1294 	/*
1295 	 * searching SAD.
1296 	 * XXX: to be checked internal IP header somewhere.  Also when
1297 	 * IPsec tunnel packet is received.  But ESP tunnel mode is
1298 	 * encrypted so we can't check internal IP header.
1299 	 */
1300 	s = pserialize_read_enter();
1301 	SAVLUT_READER_FOREACH(sav, &dst->sa, proto, hash_key) {
1302 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1303 		    "try match spi %#x, %#x\n",
1304 		    ntohl(spi), ntohl(sav->spi));
1305 
1306 		/* do not return entries w/ unusable state */
1307 		if (!SADB_SASTATE_USABLE_P(sav)) {
1308 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1309 			    "bad state %d\n", sav->state);
1310 			continue;
1311 		}
1312 		if (proto != sav->sah->saidx.proto) {
1313 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1314 			    "proto fail %d != %d\n",
1315 			    proto, sav->sah->saidx.proto);
1316 			continue;
1317 		}
1318 		if (must_check_spi && spi != sav->spi) {
1319 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1320 			    "spi fail %#x != %#x\n",
1321 			    ntohl(spi), ntohl(sav->spi));
1322 			continue;
1323 		}
1324 		/* XXX only on the ipcomp case */
1325 		if (must_check_alg && algo != sav->alg_comp) {
1326 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1327 			    "algo fail %d != %d\n",
1328 			    algo, sav->alg_comp);
1329 			continue;
1330 		}
1331 
1332 #if 0	/* don't check src */
1333 	/* Fix port in src->sa */
1334 
1335 		/* check src address */
1336 		if (!key_sockaddr_match(&src->sa, &sav->sah->saidx.src.sa, PORT_NONE))
1337 			continue;
1338 #endif
1339 		/* fix port of dst address XXX*/
1340 		key_porttosaddr(__UNCONST(dst), dport);
1341 		/* check dst address */
1342 		if (!key_sockaddr_match(&dst->sa, &sav->sah->saidx.dst.sa, chkport))
1343 			continue;
1344 		key_sa_ref(sav, where, tag);
1345 		goto done;
1346 	}
1347 	sav = NULL;
1348 done:
1349 	pserialize_read_exit(s);
1350 
1351 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1352 	    "DP return SA:%p; refcnt %u\n", sav, key_sa_refcnt(sav));
1353 	return sav;
1354 }
1355 
1356 static void
1357 key_validate_savlist(const struct secashead *sah, const u_int state)
1358 {
1359 #ifdef DEBUG
1360 	struct secasvar *sav, *next;
1361 	int s;
1362 
1363 	/*
1364 	 * The list should be sorted by lft_c->sadb_lifetime_addtime
1365 	 * in ascending order.
1366 	 */
1367 	s = pserialize_read_enter();
1368 	SAVLIST_READER_FOREACH(sav, sah, state) {
1369 		next = SAVLIST_READER_NEXT(sav);
1370 		if (next != NULL &&
1371 		    sav->lft_c != NULL && next->lft_c != NULL) {
1372 			KDASSERTMSG(sav->lft_c->sadb_lifetime_addtime <=
1373 			    next->lft_c->sadb_lifetime_addtime,
1374 			    "savlist is not sorted: sah=%p, state=%d, "
1375 			    "sav=%" PRIu64 ", next=%" PRIu64, sah, state,
1376 			    sav->lft_c->sadb_lifetime_addtime,
1377 			    next->lft_c->sadb_lifetime_addtime);
1378 		}
1379 	}
1380 	pserialize_read_exit(s);
1381 #endif
1382 }
1383 
1384 void
1385 key_init_sp(struct secpolicy *sp)
1386 {
1387 
1388 	ASSERT_SLEEPABLE();
1389 
1390 	sp->state = IPSEC_SPSTATE_ALIVE;
1391 	if (sp->policy == IPSEC_POLICY_IPSEC)
1392 		KASSERT(sp->req != NULL);
1393 	localcount_init(&sp->localcount);
1394 	SPLIST_ENTRY_INIT(sp);
1395 }
1396 
1397 /*
1398  * Must be called in a pserialize read section. A held SP
1399  * must be released by key_sp_unref after use.
1400  */
1401 void
1402 key_sp_ref(struct secpolicy *sp, const char* where, int tag)
1403 {
1404 
1405 	localcount_acquire(&sp->localcount);
1406 
1407 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1408 	    "DP SP:%p (ID=%u) from %s:%u; refcnt++ now %u\n",
1409 	    sp, sp->id, where, tag, key_sp_refcnt(sp));
1410 }
1411 
1412 /*
1413  * Must be called without holding key_spd.lock because the lock
1414  * would be held in localcount_release.
1415  */
1416 void
1417 key_sp_unref(struct secpolicy *sp, const char* where, int tag)
1418 {
1419 
1420 	KDASSERT(mutex_ownable(&key_spd.lock));
1421 
1422 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1423 	    "DP SP:%p (ID=%u) from %s:%u; refcnt-- now %u\n",
1424 	    sp, sp->id, where, tag, key_sp_refcnt(sp));
1425 
1426 	localcount_release(&sp->localcount, &key_spd.cv_lc, &key_spd.lock);
1427 }
1428 
1429 static void
1430 key_init_sav(struct secasvar *sav)
1431 {
1432 
1433 	ASSERT_SLEEPABLE();
1434 
1435 	localcount_init(&sav->localcount);
1436 	SAVLIST_ENTRY_INIT(sav);
1437 	SAVLUT_ENTRY_INIT(sav);
1438 }
1439 
1440 u_int
1441 key_sa_refcnt(const struct secasvar *sav)
1442 {
1443 
1444 	/* FIXME */
1445 	return 0;
1446 }
1447 
1448 void
1449 key_sa_ref(struct secasvar *sav, const char* where, int tag)
1450 {
1451 
1452 	localcount_acquire(&sav->localcount);
1453 
1454 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1455 	    "DP cause refcnt++: SA:%p from %s:%u\n",
1456 	    sav, where, tag);
1457 }
1458 
1459 void
1460 key_sa_unref(struct secasvar *sav, const char* where, int tag)
1461 {
1462 
1463 	KDASSERT(mutex_ownable(&key_sad.lock));
1464 
1465 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1466 	    "DP cause refcnt--: SA:%p from %s:%u\n",
1467 	    sav, where, tag);
1468 
1469 	localcount_release(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
1470 }
1471 
1472 #if 0
1473 /*
1474  * Must be called after calling key_lookup_sp*().
1475  * For the packet with socket.
1476  */
1477 static void
1478 key_freeso(struct socket *so)
1479 {
1480 	/* sanity check */
1481 	KASSERT(so != NULL);
1482 
1483 	switch (so->so_proto->pr_domain->dom_family) {
1484 #ifdef INET
1485 	case PF_INET:
1486 	    {
1487 		struct inpcb *pcb = sotoinpcb(so);
1488 
1489 		/* Does it have a PCB ? */
1490 		if (pcb == NULL)
1491 			return;
1492 
1493 		struct inpcbpolicy *sp = pcb->inp_sp;
1494 		key_freesp_so(&sp->sp_in);
1495 		key_freesp_so(&sp->sp_out);
1496 	    }
1497 		break;
1498 #endif
1499 #ifdef INET6
1500 	case PF_INET6:
1501 	    {
1502 #ifdef HAVE_NRL_INPCB
1503 		struct inpcb *pcb  = sotoinpcb(so);
1504 		struct inpcbpolicy *sp = pcb->inp_sp;
1505 
1506 		/* Does it have a PCB ? */
1507 		if (pcb == NULL)
1508 			return;
1509 		key_freesp_so(&sp->sp_in);
1510 		key_freesp_so(&sp->sp_out);
1511 #else
1512 		struct in6pcb *pcb  = sotoin6pcb(so);
1513 
1514 		/* Does it have a PCB ? */
1515 		if (pcb == NULL)
1516 			return;
1517 		key_freesp_so(&pcb->in6p_sp->sp_in);
1518 		key_freesp_so(&pcb->in6p_sp->sp_out);
1519 #endif
1520 	    }
1521 		break;
1522 #endif /* INET6 */
1523 	default:
1524 		IPSECLOG(LOG_DEBUG, "unknown address family=%d.\n",
1525 		    so->so_proto->pr_domain->dom_family);
1526 		return;
1527 	}
1528 }
1529 
1530 static void
1531 key_freesp_so(struct secpolicy **sp)
1532 {
1533 
1534 	KASSERT(sp != NULL);
1535 	KASSERT(*sp != NULL);
1536 
1537 	if ((*sp)->policy == IPSEC_POLICY_ENTRUST ||
1538 	    (*sp)->policy == IPSEC_POLICY_BYPASS)
1539 		return;
1540 
1541 	KASSERTMSG((*sp)->policy == IPSEC_POLICY_IPSEC,
1542 	    "invalid policy %u", (*sp)->policy);
1543 	KEY_SP_UNREF(&sp);
1544 }
1545 #endif
1546 
1547 static void
1548 key_sad_pserialize_perform(void)
1549 {
1550 
1551 	KASSERT(mutex_owned(&key_sad.lock));
1552 
1553 	while (key_sad.psz_performing)
1554 		cv_wait(&key_sad.cv_psz, &key_sad.lock);
1555 	key_sad.psz_performing = true;
1556 	mutex_exit(&key_sad.lock);
1557 
1558 	pserialize_perform(key_sad.psz);
1559 
1560 	mutex_enter(&key_sad.lock);
1561 	key_sad.psz_performing = false;
1562 	cv_broadcast(&key_sad.cv_psz);
1563 }
1564 
1565 /*
1566  * Remove the sav from the savlist of its sah and wait for references to the sav
1567  * to be released. key_sad.lock must be held.
1568  */
1569 static void
1570 key_unlink_sav(struct secasvar *sav)
1571 {
1572 
1573 	KASSERT(mutex_owned(&key_sad.lock));
1574 
1575 	SAVLIST_WRITER_REMOVE(sav);
1576 	SAVLUT_WRITER_REMOVE(sav);
1577 
1578 	KDASSERT(mutex_ownable(softnet_lock));
1579 	key_sad_pserialize_perform();
1580 
1581 	localcount_drain(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
1582 }
1583 
1584 /*
1585  * Destroy an sav where the sav must be unlinked from an sah
1586  * by say key_unlink_sav.
1587  */
1588 static void
1589 key_destroy_sav(struct secasvar *sav)
1590 {
1591 
1592 	ASSERT_SLEEPABLE();
1593 
1594 	localcount_fini(&sav->localcount);
1595 	SAVLIST_ENTRY_DESTROY(sav);
1596 
1597 	key_delsav(sav);
1598 }
1599 
1600 /*
1601  * Destroy sav with holding its reference.
1602  */
1603 static void
1604 key_destroy_sav_with_ref(struct secasvar *sav)
1605 {
1606 
1607 	ASSERT_SLEEPABLE();
1608 
1609 	mutex_enter(&key_sad.lock);
1610 	sav->state = SADB_SASTATE_DEAD;
1611 	SAVLIST_WRITER_REMOVE(sav);
1612 	SAVLUT_WRITER_REMOVE(sav);
1613 	mutex_exit(&key_sad.lock);
1614 
1615 	/* We cannot unref with holding key_sad.lock */
1616 	KEY_SA_UNREF(&sav);
1617 
1618 	mutex_enter(&key_sad.lock);
1619 	KDASSERT(mutex_ownable(softnet_lock));
1620 	key_sad_pserialize_perform();
1621 	localcount_drain(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
1622 	mutex_exit(&key_sad.lock);
1623 
1624 	key_destroy_sav(sav);
1625 }
1626 
1627 /* %%% SPD management */
1628 /*
1629  * free security policy entry.
1630  */
1631 static void
1632 key_destroy_sp(struct secpolicy *sp)
1633 {
1634 
1635 	SPLIST_ENTRY_DESTROY(sp);
1636 	localcount_fini(&sp->localcount);
1637 
1638 	key_free_sp(sp);
1639 
1640 	key_update_used();
1641 }
1642 
1643 void
1644 key_free_sp(struct secpolicy *sp)
1645 {
1646 	struct ipsecrequest *isr = sp->req, *nextisr;
1647 
1648 	while (isr != NULL) {
1649 		nextisr = isr->next;
1650 		kmem_free(isr, sizeof(*isr));
1651 		isr = nextisr;
1652 	}
1653 
1654 	kmem_free(sp, sizeof(*sp));
1655 }
1656 
1657 void
1658 key_socksplist_add(struct secpolicy *sp)
1659 {
1660 
1661 	mutex_enter(&key_spd.lock);
1662 	PSLIST_WRITER_INSERT_HEAD(&key_spd.socksplist, sp, pslist_entry);
1663 	mutex_exit(&key_spd.lock);
1664 
1665 	key_update_used();
1666 }
1667 
1668 /*
1669  * search SPD
1670  * OUT:	NULL	: not found
1671  *	others	: found, pointer to a SP.
1672  */
1673 static struct secpolicy *
1674 key_getsp(const struct secpolicyindex *spidx)
1675 {
1676 	struct secpolicy *sp;
1677 	int s;
1678 
1679 	KASSERT(spidx != NULL);
1680 
1681 	s = pserialize_read_enter();
1682 	SPLIST_READER_FOREACH(sp, spidx->dir) {
1683 		if (sp->state == IPSEC_SPSTATE_DEAD)
1684 			continue;
1685 		if (key_spidx_match_exactly(spidx, &sp->spidx)) {
1686 			KEY_SP_REF(sp);
1687 			pserialize_read_exit(s);
1688 			return sp;
1689 		}
1690 	}
1691 	pserialize_read_exit(s);
1692 
1693 	return NULL;
1694 }
1695 
1696 /*
1697  * search SPD and remove found SP
1698  * OUT:	NULL	: not found
1699  *	others	: found, pointer to a SP.
1700  */
1701 static struct secpolicy *
1702 key_lookup_and_remove_sp(const struct secpolicyindex *spidx, bool from_kernel)
1703 {
1704 	struct secpolicy *sp = NULL;
1705 
1706 	mutex_enter(&key_spd.lock);
1707 	SPLIST_WRITER_FOREACH(sp, spidx->dir) {
1708 		KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
1709 		/*
1710 		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
1711 		 * removed by userland programs.
1712 		 */
1713 		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
1714 			continue;
1715 		if (key_spidx_match_exactly(spidx, &sp->spidx)) {
1716 			key_unlink_sp(sp);
1717 			goto out;
1718 		}
1719 	}
1720 	sp = NULL;
1721 out:
1722 	mutex_exit(&key_spd.lock);
1723 
1724 	return sp;
1725 }
1726 
1727 /*
1728  * get SP by index.
1729  * OUT:	NULL	: not found
1730  *	others	: found, pointer to a SP.
1731  */
1732 static struct secpolicy *
1733 key_getspbyid(u_int32_t id)
1734 {
1735 	struct secpolicy *sp;
1736 	int s;
1737 
1738 	s = pserialize_read_enter();
1739 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_INBOUND) {
1740 		if (sp->state == IPSEC_SPSTATE_DEAD)
1741 			continue;
1742 		if (sp->id == id) {
1743 			KEY_SP_REF(sp);
1744 			goto out;
1745 		}
1746 	}
1747 
1748 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_OUTBOUND) {
1749 		if (sp->state == IPSEC_SPSTATE_DEAD)
1750 			continue;
1751 		if (sp->id == id) {
1752 			KEY_SP_REF(sp);
1753 			goto out;
1754 		}
1755 	}
1756 out:
1757 	pserialize_read_exit(s);
1758 	return sp;
1759 }
1760 
1761 /*
1762  * get SP by index, remove and return it.
1763  * OUT:	NULL	: not found
1764  *	others	: found, pointer to a SP.
1765  */
1766 static struct secpolicy *
1767 key_lookupbyid_and_remove_sp(u_int32_t id, bool from_kernel)
1768 {
1769 	struct secpolicy *sp;
1770 
1771 	mutex_enter(&key_spd.lock);
1772 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_INBOUND) {
1773 		KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
1774 		/*
1775 		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
1776 		 * removed by userland programs.
1777 		 */
1778 		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
1779 			continue;
1780 		if (sp->id == id)
1781 			goto out;
1782 	}
1783 
1784 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_OUTBOUND) {
1785 		KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
1786 		/*
1787 		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
1788 		 * removed by userland programs.
1789 		 */
1790 		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
1791 			continue;
1792 		if (sp->id == id)
1793 			goto out;
1794 	}
1795 out:
1796 	if (sp != NULL)
1797 		key_unlink_sp(sp);
1798 	mutex_exit(&key_spd.lock);
1799 	return sp;
1800 }
1801 
1802 struct secpolicy *
1803 key_newsp(const char* where, int tag)
1804 {
1805 	struct secpolicy *newsp = NULL;
1806 
1807 	newsp = kmem_zalloc(sizeof(struct secpolicy), KM_SLEEP);
1808 
1809 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1810 	    "DP from %s:%u return SP:%p\n", where, tag, newsp);
1811 	return newsp;
1812 }
1813 
1814 /*
1815  * create secpolicy structure from sadb_x_policy structure.
1816  * NOTE: `state', `secpolicyindex' in secpolicy structure are not set,
1817  * so must be set properly later.
1818  */
1819 static struct secpolicy *
1820 _key_msg2sp(const struct sadb_x_policy *xpl0, size_t len, int *error,
1821     bool from_kernel)
1822 {
1823 	struct secpolicy *newsp;
1824 
1825 	KASSERT(!cpu_softintr_p());
1826 	KASSERT(xpl0 != NULL);
1827 	KASSERT(len >= sizeof(*xpl0));
1828 
1829 	if (len != PFKEY_EXTLEN(xpl0)) {
1830 		IPSECLOG(LOG_DEBUG, "Invalid msg length.\n");
1831 		*error = EINVAL;
1832 		return NULL;
1833 	}
1834 
1835 	newsp = KEY_NEWSP();
1836 	if (newsp == NULL) {
1837 		*error = ENOBUFS;
1838 		return NULL;
1839 	}
1840 
1841 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1842 	newsp->policy = xpl0->sadb_x_policy_type;
1843 
1844 	/* check policy */
1845 	switch (xpl0->sadb_x_policy_type) {
1846 	case IPSEC_POLICY_DISCARD:
1847 	case IPSEC_POLICY_NONE:
1848 	case IPSEC_POLICY_ENTRUST:
1849 	case IPSEC_POLICY_BYPASS:
1850 		newsp->req = NULL;
1851 		*error = 0;
1852 		return newsp;
1853 
1854 	case IPSEC_POLICY_IPSEC:
1855 		/* Continued */
1856 		break;
1857 	default:
1858 		IPSECLOG(LOG_DEBUG, "invalid policy type.\n");
1859 		key_free_sp(newsp);
1860 		*error = EINVAL;
1861 		return NULL;
1862 	}
1863 
1864 	/* IPSEC_POLICY_IPSEC */
1865     {
1866 	int tlen;
1867 	const struct sadb_x_ipsecrequest *xisr;
1868 	uint16_t xisr_reqid;
1869 	struct ipsecrequest **p_isr = &newsp->req;
1870 
1871 	/* validity check */
1872 	if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1873 		IPSECLOG(LOG_DEBUG, "Invalid msg length.\n");
1874 		*error = EINVAL;
1875 		goto free_exit;
1876 	}
1877 
1878 	tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1879 	xisr = (const struct sadb_x_ipsecrequest *)(xpl0 + 1);
1880 
1881 	while (tlen > 0) {
1882 		/* length check */
1883 		if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr)) {
1884 			IPSECLOG(LOG_DEBUG, "invalid ipsecrequest length.\n");
1885 			*error = EINVAL;
1886 			goto free_exit;
1887 		}
1888 
1889 		/* allocate request buffer */
1890 		*p_isr = kmem_zalloc(sizeof(**p_isr), KM_SLEEP);
1891 
1892 		/* set values */
1893 		(*p_isr)->next = NULL;
1894 
1895 		switch (xisr->sadb_x_ipsecrequest_proto) {
1896 		case IPPROTO_ESP:
1897 		case IPPROTO_AH:
1898 		case IPPROTO_IPCOMP:
1899 			break;
1900 		default:
1901 			IPSECLOG(LOG_DEBUG, "invalid proto type=%u\n",
1902 			    xisr->sadb_x_ipsecrequest_proto);
1903 			*error = EPROTONOSUPPORT;
1904 			goto free_exit;
1905 		}
1906 		(*p_isr)->saidx.proto = xisr->sadb_x_ipsecrequest_proto;
1907 
1908 		switch (xisr->sadb_x_ipsecrequest_mode) {
1909 		case IPSEC_MODE_TRANSPORT:
1910 		case IPSEC_MODE_TUNNEL:
1911 			break;
1912 		case IPSEC_MODE_ANY:
1913 		default:
1914 			IPSECLOG(LOG_DEBUG, "invalid mode=%u\n",
1915 			    xisr->sadb_x_ipsecrequest_mode);
1916 			*error = EINVAL;
1917 			goto free_exit;
1918 		}
1919 		(*p_isr)->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1920 
1921 		switch (xisr->sadb_x_ipsecrequest_level) {
1922 		case IPSEC_LEVEL_DEFAULT:
1923 		case IPSEC_LEVEL_USE:
1924 		case IPSEC_LEVEL_REQUIRE:
1925 			break;
1926 		case IPSEC_LEVEL_UNIQUE:
1927 			xisr_reqid = xisr->sadb_x_ipsecrequest_reqid;
1928 			/* validity check */
1929 			/*
1930 			 * case 1) from_kernel == false
1931 			 * That means the request comes from userland.
1932 			 * If range violation of reqid, kernel will
1933 			 * update it, don't refuse it.
1934 			 *
1935 			 * case 2) from_kernel == true
1936 			 * That means the request comes from kernel
1937 			 * (e.g. ipsec(4) I/F).
1938 			 * Use thre requested reqid to avoid inconsistency
1939 			 * between kernel's reqid and the reqid in pf_key
1940 			 * message sent to userland. The pf_key message is
1941 			 * built by diverting request mbuf.
1942 			 */
1943 			if (!from_kernel &&
1944 			    xisr_reqid > IPSEC_MANUAL_REQID_MAX) {
1945 				IPSECLOG(LOG_DEBUG,
1946 				    "reqid=%d range "
1947 				    "violation, updated by kernel.\n",
1948 				    xisr_reqid);
1949 				xisr_reqid = 0;
1950 			}
1951 
1952 			/* allocate new reqid id if reqid is zero. */
1953 			if (xisr_reqid == 0) {
1954 				u_int16_t reqid = key_newreqid();
1955 				if (reqid == 0) {
1956 					*error = ENOBUFS;
1957 					goto free_exit;
1958 				}
1959 				(*p_isr)->saidx.reqid = reqid;
1960 			} else {
1961 			/* set it for manual keying. */
1962 				(*p_isr)->saidx.reqid = xisr_reqid;
1963 			}
1964 			break;
1965 
1966 		default:
1967 			IPSECLOG(LOG_DEBUG, "invalid level=%u\n",
1968 			    xisr->sadb_x_ipsecrequest_level);
1969 			*error = EINVAL;
1970 			goto free_exit;
1971 		}
1972 		(*p_isr)->level = xisr->sadb_x_ipsecrequest_level;
1973 
1974 		/* set IP addresses if there */
1975 		/*
1976 		 * NOTE:
1977 		 * MOBIKE Extensions for PF_KEY draft says:
1978 		 *     If tunnel mode is specified, the sadb_x_ipsecrequest
1979 		 *     structure is followed by two sockaddr structures that
1980 		 *     define the tunnel endpoint addresses.  In the case that
1981 		 *     transport mode is used, no additional addresses are
1982 		 *     specified.
1983 		 * see: https://tools.ietf.org/html/draft-schilcher-mobike-pfkey-extension-01
1984 		 *
1985 		 * And then, the IP addresses will be set by
1986 		 * ipsec_fill_saidx_bymbuf() from packet in transport mode.
1987 		 * This behavior is used by NAT-T enabled ipsecif(4).
1988 		 */
1989 		if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1990 			const struct sockaddr *paddr;
1991 
1992 			paddr = (const struct sockaddr *)(xisr + 1);
1993 
1994 			/* validity check */
1995 			if (paddr->sa_len > sizeof((*p_isr)->saidx.src)) {
1996 				IPSECLOG(LOG_DEBUG, "invalid request "
1997 				    "address length.\n");
1998 				*error = EINVAL;
1999 				goto free_exit;
2000 			}
2001 			memcpy(&(*p_isr)->saidx.src, paddr, paddr->sa_len);
2002 
2003 			paddr = (const struct sockaddr *)((const char *)paddr
2004 			    + paddr->sa_len);
2005 
2006 			/* validity check */
2007 			if (paddr->sa_len > sizeof((*p_isr)->saidx.dst)) {
2008 				IPSECLOG(LOG_DEBUG, "invalid request "
2009 				    "address length.\n");
2010 				*error = EINVAL;
2011 				goto free_exit;
2012 			}
2013 			memcpy(&(*p_isr)->saidx.dst, paddr, paddr->sa_len);
2014 		}
2015 
2016 		(*p_isr)->sp = newsp;
2017 
2018 		/* initialization for the next. */
2019 		p_isr = &(*p_isr)->next;
2020 		tlen -= xisr->sadb_x_ipsecrequest_len;
2021 
2022 		/* validity check */
2023 		if (tlen < 0) {
2024 			IPSECLOG(LOG_DEBUG, "becoming tlen < 0.\n");
2025 			*error = EINVAL;
2026 			goto free_exit;
2027 		}
2028 
2029 		xisr = (const struct sadb_x_ipsecrequest *)((const char *)xisr +
2030 		    xisr->sadb_x_ipsecrequest_len);
2031 	}
2032     }
2033 
2034 	*error = 0;
2035 	return newsp;
2036 
2037 free_exit:
2038 	key_free_sp(newsp);
2039 	return NULL;
2040 }
2041 
2042 struct secpolicy *
2043 key_msg2sp(const struct sadb_x_policy *xpl0, size_t len, int *error)
2044 {
2045 
2046 	return _key_msg2sp(xpl0, len, error, false);
2047 }
2048 
2049 u_int16_t
2050 key_newreqid(void)
2051 {
2052 	static u_int16_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
2053 
2054 	auto_reqid = (auto_reqid == 0xffff ?
2055 	    IPSEC_MANUAL_REQID_MAX + 1 : auto_reqid + 1);
2056 
2057 	/* XXX should be unique check */
2058 
2059 	return auto_reqid;
2060 }
2061 
2062 /*
2063  * copy secpolicy struct to sadb_x_policy structure indicated.
2064  */
2065 struct mbuf *
2066 key_sp2msg(const struct secpolicy *sp, int mflag)
2067 {
2068 	struct sadb_x_policy *xpl;
2069 	int tlen;
2070 	char *p;
2071 	struct mbuf *m;
2072 
2073 	KASSERT(sp != NULL);
2074 
2075 	tlen = key_getspreqmsglen(sp);
2076 
2077 	m = key_alloc_mbuf(tlen, mflag);
2078 	if (!m || m->m_next) {	/*XXX*/
2079 		if (m)
2080 			m_freem(m);
2081 		return NULL;
2082 	}
2083 
2084 	m->m_len = tlen;
2085 	m->m_next = NULL;
2086 	xpl = mtod(m, struct sadb_x_policy *);
2087 	memset(xpl, 0, tlen);
2088 
2089 	xpl->sadb_x_policy_len = PFKEY_UNIT64(tlen);
2090 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
2091 	xpl->sadb_x_policy_type = sp->policy;
2092 	xpl->sadb_x_policy_dir = sp->spidx.dir;
2093 	xpl->sadb_x_policy_id = sp->id;
2094 	p = (char *)xpl + sizeof(*xpl);
2095 
2096 	/* if is the policy for ipsec ? */
2097 	if (sp->policy == IPSEC_POLICY_IPSEC) {
2098 		struct sadb_x_ipsecrequest *xisr;
2099 		struct ipsecrequest *isr;
2100 
2101 		for (isr = sp->req; isr != NULL; isr = isr->next) {
2102 
2103 			xisr = (struct sadb_x_ipsecrequest *)p;
2104 
2105 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
2106 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
2107 			xisr->sadb_x_ipsecrequest_level = isr->level;
2108 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
2109 
2110 			p += sizeof(*xisr);
2111 			memcpy(p, &isr->saidx.src, isr->saidx.src.sa.sa_len);
2112 			p += isr->saidx.src.sa.sa_len;
2113 			memcpy(p, &isr->saidx.dst, isr->saidx.dst.sa.sa_len);
2114 			p += isr->saidx.src.sa.sa_len;
2115 
2116 			xisr->sadb_x_ipsecrequest_len =
2117 			    PFKEY_ALIGN8(sizeof(*xisr)
2118 			    + isr->saidx.src.sa.sa_len
2119 			    + isr->saidx.dst.sa.sa_len);
2120 		}
2121 	}
2122 
2123 	return m;
2124 }
2125 
2126 /*
2127  * m will not be freed nor modified. It never return NULL.
2128  * If it returns a mbuf of M_PKTHDR, the mbuf ensures to have
2129  * contiguous length at least sizeof(struct sadb_msg).
2130  */
2131 static struct mbuf *
2132 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
2133 		int ndeep, int nitem, ...)
2134 {
2135 	va_list ap;
2136 	int idx;
2137 	int i;
2138 	struct mbuf *result = NULL, *n;
2139 	int len;
2140 
2141 	KASSERT(m != NULL);
2142 	KASSERT(mhp != NULL);
2143 	KASSERT(!cpu_softintr_p());
2144 
2145 	va_start(ap, nitem);
2146 	for (i = 0; i < nitem; i++) {
2147 		idx = va_arg(ap, int);
2148 		KASSERT(idx >= 0);
2149 		KASSERT(idx <= SADB_EXT_MAX);
2150 		/* don't attempt to pull empty extension */
2151 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
2152 			continue;
2153 		if (idx != SADB_EXT_RESERVED &&
2154 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
2155 			continue;
2156 
2157 		if (idx == SADB_EXT_RESERVED) {
2158 			CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MHLEN);
2159 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2160 			MGETHDR(n, M_WAITOK, MT_DATA);
2161 			n->m_len = len;
2162 			n->m_next = NULL;
2163 			m_copydata(m, 0, sizeof(struct sadb_msg),
2164 			    mtod(n, void *));
2165 		} else if (i < ndeep) {
2166 			len = mhp->extlen[idx];
2167 			n = key_alloc_mbuf(len, M_WAITOK);
2168 			KASSERT(n->m_next == NULL);
2169 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
2170 			    mtod(n, void *));
2171 		} else {
2172 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
2173 			    M_WAITOK);
2174 		}
2175 		KASSERT(n != NULL);
2176 
2177 		if (result)
2178 			m_cat(result, n);
2179 		else
2180 			result = n;
2181 	}
2182 	va_end(ap);
2183 
2184 	KASSERT(result != NULL);
2185 	if ((result->m_flags & M_PKTHDR) != 0) {
2186 		result->m_pkthdr.len = 0;
2187 		for (n = result; n; n = n->m_next)
2188 			result->m_pkthdr.len += n->m_len;
2189 		KASSERT(result->m_len >= sizeof(struct sadb_msg));
2190 	}
2191 
2192 	return result;
2193 }
2194 
2195 /*
2196  * The argument _sp must not overwrite until SP is created and registered
2197  * successfully.
2198  */
2199 static int
2200 key_spdadd(struct socket *so, struct mbuf *m,
2201 	   const struct sadb_msghdr *mhp, struct secpolicy **_sp,
2202 	   bool from_kernel)
2203 {
2204 	const struct sockaddr *src, *dst;
2205 	const struct sadb_x_policy *xpl0;
2206 	struct sadb_x_policy *xpl;
2207 	const struct sadb_lifetime *lft = NULL;
2208 	struct secpolicyindex spidx;
2209 	struct secpolicy *newsp;
2210 	int error;
2211 	uint32_t sadb_x_policy_id;
2212 
2213 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
2214 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
2215 	    mhp->ext[SADB_X_EXT_POLICY] == NULL) {
2216 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2217 		return key_senderror(so, m, EINVAL);
2218 	}
2219 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
2220 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
2221 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2222 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2223 		return key_senderror(so, m, EINVAL);
2224 	}
2225 	if (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL) {
2226 		if (mhp->extlen[SADB_EXT_LIFETIME_HARD] <
2227 		    sizeof(struct sadb_lifetime)) {
2228 			IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2229 			return key_senderror(so, m, EINVAL);
2230 		}
2231 		lft = mhp->ext[SADB_EXT_LIFETIME_HARD];
2232 	}
2233 
2234 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
2235 
2236 	/* checking the direciton. */
2237 	switch (xpl0->sadb_x_policy_dir) {
2238 	case IPSEC_DIR_INBOUND:
2239 	case IPSEC_DIR_OUTBOUND:
2240 		break;
2241 	default:
2242 		IPSECLOG(LOG_DEBUG, "Invalid SP direction.\n");
2243 		return key_senderror(so, m, EINVAL);
2244 	}
2245 
2246 	/* check policy */
2247 	/* key_api_spdadd() accepts DISCARD, NONE and IPSEC. */
2248 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_ENTRUST ||
2249 	    xpl0->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
2250 		IPSECLOG(LOG_DEBUG, "Invalid policy type.\n");
2251 		return key_senderror(so, m, EINVAL);
2252 	}
2253 
2254 	/* policy requests are mandatory when action is ipsec. */
2255 	if (mhp->msg->sadb_msg_type != SADB_X_SPDSETIDX &&
2256 	    xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
2257 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
2258 		IPSECLOG(LOG_DEBUG, "some policy requests part required.\n");
2259 		return key_senderror(so, m, EINVAL);
2260 	}
2261 
2262 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
2263 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
2264 
2265 	/* sanity check on addr pair */
2266 	if (src->sa_family != dst->sa_family)
2267 		return key_senderror(so, m, EINVAL);
2268 	if (src->sa_len != dst->sa_len)
2269 		return key_senderror(so, m, EINVAL);
2270 
2271 	key_init_spidx_bymsghdr(&spidx, mhp);
2272 
2273 	/*
2274 	 * checking there is SP already or not.
2275 	 * SPDUPDATE doesn't depend on whether there is a SP or not.
2276 	 * If the type is either SPDADD or SPDSETIDX AND a SP is found,
2277 	 * then error.
2278 	 */
2279     {
2280 	struct secpolicy *sp;
2281 
2282 	if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2283 		sp = key_lookup_and_remove_sp(&spidx, from_kernel);
2284 		if (sp != NULL)
2285 			key_destroy_sp(sp);
2286 	} else {
2287 		sp = key_getsp(&spidx);
2288 		if (sp != NULL) {
2289 			KEY_SP_UNREF(&sp);
2290 			IPSECLOG(LOG_DEBUG, "a SP entry exists already.\n");
2291 			return key_senderror(so, m, EEXIST);
2292 		}
2293 	}
2294     }
2295 
2296 	/* allocation new SP entry */
2297 	newsp = _key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error, from_kernel);
2298 	if (newsp == NULL) {
2299 		return key_senderror(so, m, error);
2300 	}
2301 
2302 	newsp->id = key_getnewspid();
2303 	if (newsp->id == 0) {
2304 		kmem_free(newsp, sizeof(*newsp));
2305 		return key_senderror(so, m, ENOBUFS);
2306 	}
2307 
2308 	newsp->spidx = spidx;
2309 	newsp->created = time_uptime;
2310 	newsp->lastused = newsp->created;
2311 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
2312 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
2313 	if (from_kernel)
2314 		newsp->origin = IPSEC_SPORIGIN_KERNEL;
2315 	else
2316 		newsp->origin = IPSEC_SPORIGIN_USER;
2317 
2318 	key_init_sp(newsp);
2319 	if (from_kernel)
2320 		KEY_SP_REF(newsp);
2321 
2322 	sadb_x_policy_id = newsp->id;
2323 
2324 	if (_sp != NULL)
2325 		*_sp = newsp;
2326 
2327 	mutex_enter(&key_spd.lock);
2328 	SPLIST_WRITER_INSERT_TAIL(newsp->spidx.dir, newsp);
2329 	mutex_exit(&key_spd.lock);
2330 	/*
2331 	 * We don't have a reference to newsp, so we must not touch newsp from
2332 	 * now on.  If you want to do, you must take a reference beforehand.
2333 	 */
2334 	newsp = NULL;
2335 
2336 #ifdef notyet
2337 	/* delete the entry in key_misc.spacqlist */
2338 	if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2339 		struct secspacq *spacq = key_getspacq(&spidx);
2340 		if (spacq != NULL) {
2341 			/* reset counter in order to deletion by timehandler. */
2342 			spacq->created = time_uptime;
2343 			spacq->count = 0;
2344 		}
2345     	}
2346 #endif
2347 
2348 	/* Invalidate all cached SPD pointers in the PCBs. */
2349 	ipsec_invalpcbcacheall();
2350 
2351 #if defined(GATEWAY)
2352 	/* Invalidate the ipflow cache, as well. */
2353 	ipflow_invalidate_all(0);
2354 #ifdef INET6
2355 	if (in6_present)
2356 		ip6flow_invalidate_all(0);
2357 #endif /* INET6 */
2358 #endif /* GATEWAY */
2359 
2360 	key_update_used();
2361 
2362     {
2363 	struct mbuf *n, *mpolicy;
2364 	int off;
2365 
2366 	/* create new sadb_msg to reply. */
2367 	if (lft) {
2368 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
2369 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
2370 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2371 	} else {
2372 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
2373 		    SADB_X_EXT_POLICY,
2374 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2375 	}
2376 
2377 	key_fill_replymsg(n, 0);
2378 	off = 0;
2379 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
2380 	    sizeof(*xpl), &off);
2381 	if (mpolicy == NULL) {
2382 		/* n is already freed */
2383 		/*
2384 		 * valid sp has been created, so we does not overwrite _sp
2385 		 * NULL here. let caller decide to use the sp or not.
2386 		 */
2387 		return key_senderror(so, m, ENOBUFS);
2388 	}
2389 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, char *) + off);
2390 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
2391 		m_freem(n);
2392 		/* ditto */
2393 		return key_senderror(so, m, EINVAL);
2394 	}
2395 
2396 	xpl->sadb_x_policy_id = sadb_x_policy_id;
2397 
2398 	m_freem(m);
2399 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2400     }
2401 }
2402 
2403 /*
2404  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
2405  * add an entry to SP database, when received
2406  *   <base, address(SD), (lifetime(H),) policy>
2407  * from the user(?).
2408  * Adding to SP database,
2409  * and send
2410  *   <base, address(SD), (lifetime(H),) policy>
2411  * to the socket which was send.
2412  *
2413  * SPDADD set a unique policy entry.
2414  * SPDSETIDX like SPDADD without a part of policy requests.
2415  * SPDUPDATE replace a unique policy entry.
2416  *
2417  * m will always be freed.
2418  */
2419 static int
2420 key_api_spdadd(struct socket *so, struct mbuf *m,
2421 	       const struct sadb_msghdr *mhp)
2422 {
2423 
2424 	return key_spdadd(so, m, mhp, NULL, false);
2425 }
2426 
2427 struct secpolicy *
2428 key_kpi_spdadd(struct mbuf *m)
2429 {
2430 	struct sadb_msghdr mh;
2431 	int error;
2432 	struct secpolicy *sp = NULL;
2433 
2434 	error = key_align(m, &mh);
2435 	if (error)
2436 		return NULL;
2437 
2438 	error = key_spdadd(NULL, m, &mh, &sp, true);
2439 	if (error) {
2440 		/*
2441 		 * Currently, when key_spdadd() cannot send a PFKEY message
2442 		 * which means SP has been created, key_spdadd() returns error
2443 		 * although SP is created successfully.
2444 		 * Kernel components would not care PFKEY messages, so return
2445 		 * the "sp" regardless of error code. key_spdadd() overwrites
2446 		 * the argument only if SP  is created successfully.
2447 		 */
2448 	}
2449 	return sp;
2450 }
2451 
2452 /*
2453  * get new policy id.
2454  * OUT:
2455  *	0:	failure.
2456  *	others: success.
2457  */
2458 static u_int32_t
2459 key_getnewspid(void)
2460 {
2461 	u_int32_t newid = 0;
2462 	int count = key_spi_trycnt;	/* XXX */
2463 	struct secpolicy *sp;
2464 
2465 	/* when requesting to allocate spi ranged */
2466 	while (count--) {
2467 		newid = (policy_id = (policy_id == ~0 ? 1 : policy_id + 1));
2468 
2469 		sp = key_getspbyid(newid);
2470 		if (sp == NULL)
2471 			break;
2472 
2473 		KEY_SP_UNREF(&sp);
2474 	}
2475 
2476 	if (count == 0 || newid == 0) {
2477 		IPSECLOG(LOG_DEBUG, "to allocate policy id is failed.\n");
2478 		return 0;
2479 	}
2480 
2481 	return newid;
2482 }
2483 
2484 /*
2485  * SADB_SPDDELETE processing
2486  * receive
2487  *   <base, address(SD), policy(*)>
2488  * from the user(?), and set SADB_SASTATE_DEAD,
2489  * and send,
2490  *   <base, address(SD), policy(*)>
2491  * to the ikmpd.
2492  * policy(*) including direction of policy.
2493  *
2494  * m will always be freed.
2495  */
2496 static int
2497 key_api_spddelete(struct socket *so, struct mbuf *m,
2498               const struct sadb_msghdr *mhp)
2499 {
2500 	struct sadb_x_policy *xpl0;
2501 	struct secpolicyindex spidx;
2502 	struct secpolicy *sp;
2503 
2504 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
2505 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
2506 	    mhp->ext[SADB_X_EXT_POLICY] == NULL) {
2507 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2508 		return key_senderror(so, m, EINVAL);
2509 	}
2510 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
2511 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
2512 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2513 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2514 		return key_senderror(so, m, EINVAL);
2515 	}
2516 
2517 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
2518 
2519 	/* checking the directon. */
2520 	switch (xpl0->sadb_x_policy_dir) {
2521 	case IPSEC_DIR_INBOUND:
2522 	case IPSEC_DIR_OUTBOUND:
2523 		break;
2524 	default:
2525 		IPSECLOG(LOG_DEBUG, "Invalid SP direction.\n");
2526 		return key_senderror(so, m, EINVAL);
2527 	}
2528 
2529 	/* make secindex */
2530 	key_init_spidx_bymsghdr(&spidx, mhp);
2531 
2532 	/* Is there SP in SPD ? */
2533 	sp = key_lookup_and_remove_sp(&spidx, false);
2534 	if (sp == NULL) {
2535 		IPSECLOG(LOG_DEBUG, "no SP found.\n");
2536 		return key_senderror(so, m, EINVAL);
2537 	}
2538 
2539 	/* save policy id to buffer to be returned. */
2540 	xpl0->sadb_x_policy_id = sp->id;
2541 
2542 	key_destroy_sp(sp);
2543 
2544 	/* We're deleting policy; no need to invalidate the ipflow cache. */
2545 
2546     {
2547 	struct mbuf *n;
2548 
2549 	/* create new sadb_msg to reply. */
2550 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2551 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2552 	key_fill_replymsg(n, 0);
2553 	m_freem(m);
2554 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2555     }
2556 }
2557 
2558 static struct mbuf *
2559 key_alloc_mbuf_simple(int len, int mflag)
2560 {
2561 	struct mbuf *n;
2562 
2563 	KASSERT(mflag == M_NOWAIT || (mflag == M_WAITOK && !cpu_softintr_p()));
2564 
2565 	MGETHDR(n, mflag, MT_DATA);
2566 	if (n && len > MHLEN) {
2567 		MCLGET(n, mflag);
2568 		if ((n->m_flags & M_EXT) == 0) {
2569 			m_freem(n);
2570 			n = NULL;
2571 		}
2572 	}
2573 	return n;
2574 }
2575 
2576 /*
2577  * SADB_SPDDELETE2 processing
2578  * receive
2579  *   <base, policy(*)>
2580  * from the user(?), and set SADB_SASTATE_DEAD,
2581  * and send,
2582  *   <base, policy(*)>
2583  * to the ikmpd.
2584  * policy(*) including direction of policy.
2585  *
2586  * m will always be freed.
2587  */
2588 static int
2589 key_spddelete2(struct socket *so, struct mbuf *m,
2590 	       const struct sadb_msghdr *mhp, bool from_kernel)
2591 {
2592 	u_int32_t id;
2593 	struct secpolicy *sp;
2594 	const struct sadb_x_policy *xpl;
2595 
2596 	if (mhp->ext[SADB_X_EXT_POLICY] == NULL ||
2597 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2598 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2599 		return key_senderror(so, m, EINVAL);
2600 	}
2601 
2602 	xpl = mhp->ext[SADB_X_EXT_POLICY];
2603 	id = xpl->sadb_x_policy_id;
2604 
2605 	/* Is there SP in SPD ? */
2606 	sp = key_lookupbyid_and_remove_sp(id, from_kernel);
2607 	if (sp == NULL) {
2608 		IPSECLOG(LOG_DEBUG, "no SP found id:%u.\n", id);
2609 		return key_senderror(so, m, EINVAL);
2610 	}
2611 
2612 	key_destroy_sp(sp);
2613 
2614 	/* We're deleting policy; no need to invalidate the ipflow cache. */
2615 
2616     {
2617 	struct mbuf *n, *nn;
2618 	int off, len;
2619 
2620 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MCLBYTES);
2621 
2622 	/* create new sadb_msg to reply. */
2623 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2624 
2625 	n = key_alloc_mbuf_simple(len, M_WAITOK);
2626 	n->m_len = len;
2627 	n->m_next = NULL;
2628 	off = 0;
2629 
2630 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
2631 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2632 
2633 	KASSERTMSG(off == len, "length inconsistency");
2634 
2635 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2636 	    mhp->extlen[SADB_X_EXT_POLICY], M_WAITOK);
2637 
2638 	n->m_pkthdr.len = 0;
2639 	for (nn = n; nn; nn = nn->m_next)
2640 		n->m_pkthdr.len += nn->m_len;
2641 
2642 	key_fill_replymsg(n, 0);
2643 	m_freem(m);
2644 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2645     }
2646 }
2647 
2648 /*
2649  * SADB_SPDDELETE2 processing
2650  * receive
2651  *   <base, policy(*)>
2652  * from the user(?), and set SADB_SASTATE_DEAD,
2653  * and send,
2654  *   <base, policy(*)>
2655  * to the ikmpd.
2656  * policy(*) including direction of policy.
2657  *
2658  * m will always be freed.
2659  */
2660 static int
2661 key_api_spddelete2(struct socket *so, struct mbuf *m,
2662 	       const struct sadb_msghdr *mhp)
2663 {
2664 
2665 	return key_spddelete2(so, m, mhp, false);
2666 }
2667 
2668 int
2669 key_kpi_spddelete2(struct mbuf *m)
2670 {
2671 	struct sadb_msghdr mh;
2672 	int error;
2673 
2674 	error = key_align(m, &mh);
2675 	if (error)
2676 		return EINVAL;
2677 
2678 	return key_spddelete2(NULL, m, &mh, true);
2679 }
2680 
2681 /*
2682  * SADB_X_GET processing
2683  * receive
2684  *   <base, policy(*)>
2685  * from the user(?),
2686  * and send,
2687  *   <base, address(SD), policy>
2688  * to the ikmpd.
2689  * policy(*) including direction of policy.
2690  *
2691  * m will always be freed.
2692  */
2693 static int
2694 key_api_spdget(struct socket *so, struct mbuf *m,
2695 	   const struct sadb_msghdr *mhp)
2696 {
2697 	u_int32_t id;
2698 	struct secpolicy *sp;
2699 	struct mbuf *n;
2700 	const struct sadb_x_policy *xpl;
2701 
2702 	if (mhp->ext[SADB_X_EXT_POLICY] == NULL ||
2703 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2704 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2705 		return key_senderror(so, m, EINVAL);
2706 	}
2707 
2708 	xpl = mhp->ext[SADB_X_EXT_POLICY];
2709 	id = xpl->sadb_x_policy_id;
2710 
2711 	/* Is there SP in SPD ? */
2712 	sp = key_getspbyid(id);
2713 	if (sp == NULL) {
2714 		IPSECLOG(LOG_DEBUG, "no SP found id:%u.\n", id);
2715 		return key_senderror(so, m, ENOENT);
2716 	}
2717 
2718 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2719 	    mhp->msg->sadb_msg_pid);
2720 	KEY_SP_UNREF(&sp); /* ref gained by key_getspbyid */
2721 	m_freem(m);
2722 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2723 }
2724 
2725 #ifdef notyet
2726 /*
2727  * SADB_X_SPDACQUIRE processing.
2728  * Acquire policy and SA(s) for a *OUTBOUND* packet.
2729  * send
2730  *   <base, policy(*)>
2731  * to KMD, and expect to receive
2732  *   <base> with SADB_X_SPDACQUIRE if error occurred,
2733  * or
2734  *   <base, policy>
2735  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2736  * policy(*) is without policy requests.
2737  *
2738  *    0     : succeed
2739  *    others: error number
2740  */
2741 int
2742 key_spdacquire(const struct secpolicy *sp)
2743 {
2744 	struct mbuf *result = NULL, *m;
2745 	struct secspacq *newspacq;
2746 	int error;
2747 
2748 	KASSERT(sp != NULL);
2749 	KASSERTMSG(sp->req == NULL, "called but there is request");
2750 	KASSERTMSG(sp->policy == IPSEC_POLICY_IPSEC,
2751 	    "policy mismathed. IPsec is expected");
2752 
2753 	/* Get an entry to check whether sent message or not. */
2754 	newspacq = key_getspacq(&sp->spidx);
2755 	if (newspacq != NULL) {
2756 		if (key_blockacq_count < newspacq->count) {
2757 			/* reset counter and do send message. */
2758 			newspacq->count = 0;
2759 		} else {
2760 			/* increment counter and do nothing. */
2761 			newspacq->count++;
2762 			return 0;
2763 		}
2764 	} else {
2765 		/* make new entry for blocking to send SADB_ACQUIRE. */
2766 		newspacq = key_newspacq(&sp->spidx);
2767 		if (newspacq == NULL)
2768 			return ENOBUFS;
2769 
2770 		/* add to key_misc.acqlist */
2771 		LIST_INSERT_HEAD(&key_misc.spacqlist, newspacq, chain);
2772 	}
2773 
2774 	/* create new sadb_msg to reply. */
2775 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2776 	if (!m) {
2777 		error = ENOBUFS;
2778 		goto fail;
2779 	}
2780 	result = m;
2781 
2782 	result->m_pkthdr.len = 0;
2783 	for (m = result; m; m = m->m_next)
2784 		result->m_pkthdr.len += m->m_len;
2785 
2786 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2787 	    PFKEY_UNIT64(result->m_pkthdr.len);
2788 
2789 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2790 
2791 fail:
2792 	if (result)
2793 		m_freem(result);
2794 	return error;
2795 }
2796 #endif /* notyet */
2797 
2798 /*
2799  * SADB_SPDFLUSH processing
2800  * receive
2801  *   <base>
2802  * from the user, and free all entries in secpctree.
2803  * and send,
2804  *   <base>
2805  * to the user.
2806  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2807  *
2808  * m will always be freed.
2809  */
2810 static int
2811 key_api_spdflush(struct socket *so, struct mbuf *m,
2812 	     const struct sadb_msghdr *mhp)
2813 {
2814 	struct sadb_msg *newmsg;
2815 	struct secpolicy *sp;
2816 	u_int dir;
2817 
2818 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2819 		return key_senderror(so, m, EINVAL);
2820 
2821 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2822 	    retry:
2823 		mutex_enter(&key_spd.lock);
2824 		SPLIST_WRITER_FOREACH(sp, dir) {
2825 			KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
2826 			/*
2827 			 * Userlang programs can remove SPs created by userland
2828 			 * probrams only, that is, they cannot remove SPs
2829 			 * created in kernel(e.g. ipsec(4) I/F).
2830 			 */
2831 			if (sp->origin == IPSEC_SPORIGIN_USER) {
2832 				key_unlink_sp(sp);
2833 				mutex_exit(&key_spd.lock);
2834 				key_destroy_sp(sp);
2835 				goto retry;
2836 			}
2837 		}
2838 		mutex_exit(&key_spd.lock);
2839 	}
2840 
2841 	/* We're deleting policy; no need to invalidate the ipflow cache. */
2842 
2843 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2844 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
2845 		return key_senderror(so, m, ENOBUFS);
2846 	}
2847 
2848 	if (m->m_next)
2849 		m_freem(m->m_next);
2850 	m->m_next = NULL;
2851 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2852 	newmsg = mtod(m, struct sadb_msg *);
2853 	newmsg->sadb_msg_errno = 0;
2854 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2855 
2856 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2857 }
2858 
2859 static struct sockaddr key_src = {
2860 	.sa_len = 2,
2861 	.sa_family = PF_KEY,
2862 };
2863 
2864 static struct mbuf *
2865 key_setspddump_chain(int *errorp, int *lenp, pid_t pid)
2866 {
2867 	struct secpolicy *sp;
2868 	int cnt;
2869 	u_int dir;
2870 	struct mbuf *m, *n, *prev;
2871 	int totlen;
2872 
2873 	KASSERT(mutex_owned(&key_spd.lock));
2874 
2875 	*lenp = 0;
2876 
2877 	/* search SPD entry and get buffer size. */
2878 	cnt = 0;
2879 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2880 		SPLIST_WRITER_FOREACH(sp, dir) {
2881 			cnt++;
2882 		}
2883 	}
2884 
2885 	if (cnt == 0) {
2886 		*errorp = ENOENT;
2887 		return (NULL);
2888 	}
2889 
2890 	m = NULL;
2891 	prev = m;
2892 	totlen = 0;
2893 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2894 		SPLIST_WRITER_FOREACH(sp, dir) {
2895 			--cnt;
2896 			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, pid);
2897 
2898 			totlen += n->m_pkthdr.len;
2899 			if (!m) {
2900 				m = n;
2901 			} else {
2902 				prev->m_nextpkt = n;
2903 			}
2904 			prev = n;
2905 		}
2906 	}
2907 
2908 	*lenp = totlen;
2909 	*errorp = 0;
2910 	return (m);
2911 }
2912 
2913 /*
2914  * SADB_SPDDUMP processing
2915  * receive
2916  *   <base>
2917  * from the user, and dump all SP leaves
2918  * and send,
2919  *   <base> .....
2920  * to the ikmpd.
2921  *
2922  * m will always be freed.
2923  */
2924 static int
2925 key_api_spddump(struct socket *so, struct mbuf *m0,
2926  	    const struct sadb_msghdr *mhp)
2927 {
2928 	struct mbuf *n;
2929 	int error, len;
2930 	int ok;
2931 	pid_t pid;
2932 
2933 	pid = mhp->msg->sadb_msg_pid;
2934 	/*
2935 	 * If the requestor has insufficient socket-buffer space
2936 	 * for the entire chain, nobody gets any response to the DUMP.
2937 	 * XXX For now, only the requestor ever gets anything.
2938 	 * Moreover, if the requestor has any space at all, they receive
2939 	 * the entire chain, otherwise the request is refused with  ENOBUFS.
2940 	 */
2941 	if (sbspace(&so->so_rcv) <= 0) {
2942 		return key_senderror(so, m0, ENOBUFS);
2943 	}
2944 
2945 	mutex_enter(&key_spd.lock);
2946 	n = key_setspddump_chain(&error, &len, pid);
2947 	mutex_exit(&key_spd.lock);
2948 
2949 	if (n == NULL) {
2950 		return key_senderror(so, m0, ENOENT);
2951 	}
2952 	{
2953 		uint64_t *ps = PFKEY_STAT_GETREF();
2954 		ps[PFKEY_STAT_IN_TOTAL]++;
2955 		ps[PFKEY_STAT_IN_BYTES] += len;
2956 		PFKEY_STAT_PUTREF();
2957 	}
2958 
2959 	/*
2960 	 * PF_KEY DUMP responses are no longer broadcast to all PF_KEY sockets.
2961 	 * The requestor receives either the entire chain, or an
2962 	 * error message with ENOBUFS.
2963 	 */
2964 
2965 	/*
2966 	 * sbappendchainwith record takes the chain of entries, one
2967 	 * packet-record per SPD entry, prepends the key_src sockaddr
2968 	 * to each packet-record, links the sockaddr mbufs into a new
2969 	 * list of records, then   appends the entire resulting
2970 	 * list to the requesting socket.
2971 	 */
2972 	ok = sbappendaddrchain(&so->so_rcv, (struct sockaddr *)&key_src, n,
2973 	    SB_PRIO_ONESHOT_OVERFLOW);
2974 
2975 	if (!ok) {
2976 		PFKEY_STATINC(PFKEY_STAT_IN_NOMEM);
2977 		m_freem(n);
2978 		return key_senderror(so, m0, ENOBUFS);
2979 	}
2980 
2981 	m_freem(m0);
2982 	return error;
2983 }
2984 
2985 /*
2986  * SADB_X_NAT_T_NEW_MAPPING. Unused by racoon as of 2005/04/23
2987  */
2988 static int
2989 key_api_nat_map(struct socket *so, struct mbuf *m,
2990 	    const struct sadb_msghdr *mhp)
2991 {
2992 	struct sadb_x_nat_t_type *type;
2993 	struct sadb_x_nat_t_port *sport;
2994 	struct sadb_x_nat_t_port *dport;
2995 	struct sadb_address *iaddr, *raddr;
2996 	struct sadb_x_nat_t_frag *frag;
2997 
2998 	if (mhp->ext[SADB_X_EXT_NAT_T_TYPE] == NULL ||
2999 	    mhp->ext[SADB_X_EXT_NAT_T_SPORT] == NULL ||
3000 	    mhp->ext[SADB_X_EXT_NAT_T_DPORT] == NULL) {
3001 		IPSECLOG(LOG_DEBUG, "invalid message.\n");
3002 		return key_senderror(so, m, EINVAL);
3003 	}
3004 	if ((mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) ||
3005 	    (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) ||
3006 	    (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport))) {
3007 		IPSECLOG(LOG_DEBUG, "invalid message.\n");
3008 		return key_senderror(so, m, EINVAL);
3009 	}
3010 
3011 	if ((mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL) &&
3012 	    (mhp->extlen[SADB_X_EXT_NAT_T_OAI] < sizeof(*iaddr))) {
3013 		IPSECLOG(LOG_DEBUG, "invalid message\n");
3014 		return key_senderror(so, m, EINVAL);
3015 	}
3016 
3017 	if ((mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL) &&
3018 	    (mhp->extlen[SADB_X_EXT_NAT_T_OAR] < sizeof(*raddr))) {
3019 		IPSECLOG(LOG_DEBUG, "invalid message\n");
3020 		return key_senderror(so, m, EINVAL);
3021 	}
3022 
3023 	if ((mhp->ext[SADB_X_EXT_NAT_T_FRAG] != NULL) &&
3024 	    (mhp->extlen[SADB_X_EXT_NAT_T_FRAG] < sizeof(*frag))) {
3025 		IPSECLOG(LOG_DEBUG, "invalid message\n");
3026 		return key_senderror(so, m, EINVAL);
3027 	}
3028 
3029 	type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
3030 	sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
3031 	dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
3032 	iaddr = mhp->ext[SADB_X_EXT_NAT_T_OAI];
3033 	raddr = mhp->ext[SADB_X_EXT_NAT_T_OAR];
3034 	frag = mhp->ext[SADB_X_EXT_NAT_T_FRAG];
3035 
3036 	/*
3037 	 * XXX handle that, it should also contain a SA, or anything
3038 	 * that enable to update the SA information.
3039 	 */
3040 
3041 	return 0;
3042 }
3043 
3044 /*
3045  * Never return NULL.
3046  */
3047 static struct mbuf *
3048 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq, pid_t pid)
3049 {
3050 	struct mbuf *result = NULL, *m;
3051 
3052 	KASSERT(!cpu_softintr_p());
3053 
3054 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid,
3055 	    key_sp_refcnt(sp), M_WAITOK);
3056 	result = m;
3057 
3058 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3059 	    &sp->spidx.src.sa, sp->spidx.prefs, sp->spidx.ul_proto, M_WAITOK);
3060 	m_cat(result, m);
3061 
3062 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3063 	    &sp->spidx.dst.sa, sp->spidx.prefd, sp->spidx.ul_proto, M_WAITOK);
3064 	m_cat(result, m);
3065 
3066 	m = key_sp2msg(sp, M_WAITOK);
3067 	m_cat(result, m);
3068 
3069 	KASSERT(result->m_flags & M_PKTHDR);
3070 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
3071 
3072 	result->m_pkthdr.len = 0;
3073 	for (m = result; m; m = m->m_next)
3074 		result->m_pkthdr.len += m->m_len;
3075 
3076 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3077 	    PFKEY_UNIT64(result->m_pkthdr.len);
3078 
3079 	return result;
3080 }
3081 
3082 /*
3083  * get PFKEY message length for security policy and request.
3084  */
3085 static u_int
3086 key_getspreqmsglen(const struct secpolicy *sp)
3087 {
3088 	u_int tlen;
3089 
3090 	tlen = sizeof(struct sadb_x_policy);
3091 
3092 	/* if is the policy for ipsec ? */
3093 	if (sp->policy != IPSEC_POLICY_IPSEC)
3094 		return tlen;
3095 
3096 	/* get length of ipsec requests */
3097     {
3098 	const struct ipsecrequest *isr;
3099 	int len;
3100 
3101 	for (isr = sp->req; isr != NULL; isr = isr->next) {
3102 		len = sizeof(struct sadb_x_ipsecrequest)
3103 		    + isr->saidx.src.sa.sa_len + isr->saidx.dst.sa.sa_len;
3104 
3105 		tlen += PFKEY_ALIGN8(len);
3106 	}
3107     }
3108 
3109 	return tlen;
3110 }
3111 
3112 /*
3113  * SADB_SPDEXPIRE processing
3114  * send
3115  *   <base, address(SD), lifetime(CH), policy>
3116  * to KMD by PF_KEY.
3117  *
3118  * OUT:	0	: succeed
3119  *	others	: error number
3120  */
3121 static int
3122 key_spdexpire(struct secpolicy *sp)
3123 {
3124 	int s;
3125 	struct mbuf *result = NULL, *m;
3126 	int len;
3127 	int error = -1;
3128 	struct sadb_lifetime *lt;
3129 
3130 	/* XXX: Why do we lock ? */
3131 	s = splsoftnet();	/*called from softclock()*/
3132 
3133 	KASSERT(sp != NULL);
3134 
3135 	/* set msg header */
3136 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0, M_WAITOK);
3137 	result = m;
3138 
3139 	/* create lifetime extension (current and hard) */
3140 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
3141 	m = key_alloc_mbuf(len, M_WAITOK);
3142 	KASSERT(m->m_next == NULL);
3143 
3144 	memset(mtod(m, void *), 0, len);
3145 	lt = mtod(m, struct sadb_lifetime *);
3146 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
3147 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
3148 	lt->sadb_lifetime_allocations = 0;
3149 	lt->sadb_lifetime_bytes = 0;
3150 	lt->sadb_lifetime_addtime = time_mono_to_wall(sp->created);
3151 	lt->sadb_lifetime_usetime = time_mono_to_wall(sp->lastused);
3152 	lt = (struct sadb_lifetime *)(mtod(m, char *) + len / 2);
3153 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
3154 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
3155 	lt->sadb_lifetime_allocations = 0;
3156 	lt->sadb_lifetime_bytes = 0;
3157 	lt->sadb_lifetime_addtime = sp->lifetime;
3158 	lt->sadb_lifetime_usetime = sp->validtime;
3159 	m_cat(result, m);
3160 
3161 	/* set sadb_address for source */
3162 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sp->spidx.src.sa,
3163 	    sp->spidx.prefs, sp->spidx.ul_proto, M_WAITOK);
3164 	m_cat(result, m);
3165 
3166 	/* set sadb_address for destination */
3167 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sp->spidx.dst.sa,
3168 	    sp->spidx.prefd, sp->spidx.ul_proto, M_WAITOK);
3169 	m_cat(result, m);
3170 
3171 	/* set secpolicy */
3172 	m = key_sp2msg(sp, M_WAITOK);
3173 	m_cat(result, m);
3174 
3175 	KASSERT(result->m_flags & M_PKTHDR);
3176 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
3177 
3178 	result->m_pkthdr.len = 0;
3179 	for (m = result; m; m = m->m_next)
3180 		result->m_pkthdr.len += m->m_len;
3181 
3182 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3183 	    PFKEY_UNIT64(result->m_pkthdr.len);
3184 
3185 	error = key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
3186 	splx(s);
3187 	return error;
3188 }
3189 
3190 /* %%% SAD management */
3191 /*
3192  * allocating a memory for new SA head, and copy from the values of mhp.
3193  * OUT:	NULL	: failure due to the lack of memory.
3194  *	others	: pointer to new SA head.
3195  */
3196 static struct secashead *
3197 key_newsah(const struct secasindex *saidx)
3198 {
3199 	struct secashead *newsah;
3200 	int i;
3201 
3202 	KASSERT(saidx != NULL);
3203 
3204 	newsah = kmem_zalloc(sizeof(struct secashead), KM_SLEEP);
3205 	for (i = 0; i < __arraycount(newsah->savlist); i++)
3206 		PSLIST_INIT(&newsah->savlist[i]);
3207 	newsah->saidx = *saidx;
3208 
3209 	localcount_init(&newsah->localcount);
3210 	/* Take a reference for the caller */
3211 	localcount_acquire(&newsah->localcount);
3212 
3213 	/* Add to the sah list */
3214 	SAHLIST_ENTRY_INIT(newsah);
3215 	newsah->state = SADB_SASTATE_MATURE;
3216 	mutex_enter(&key_sad.lock);
3217 	SAHLIST_WRITER_INSERT_HEAD(newsah);
3218 	mutex_exit(&key_sad.lock);
3219 
3220 	return newsah;
3221 }
3222 
3223 static bool
3224 key_sah_has_sav(struct secashead *sah)
3225 {
3226 	u_int state;
3227 
3228 	KASSERT(mutex_owned(&key_sad.lock));
3229 
3230 	SASTATE_ANY_FOREACH(state) {
3231 		if (!SAVLIST_WRITER_EMPTY(sah, state))
3232 			return true;
3233 	}
3234 
3235 	return false;
3236 }
3237 
3238 static void
3239 key_unlink_sah(struct secashead *sah)
3240 {
3241 
3242 	KASSERT(!cpu_softintr_p());
3243 	KASSERT(mutex_owned(&key_sad.lock));
3244 	KASSERT(sah->state == SADB_SASTATE_DEAD);
3245 
3246 	/* Remove from the sah list */
3247 	SAHLIST_WRITER_REMOVE(sah);
3248 
3249 	KDASSERT(mutex_ownable(softnet_lock));
3250 	key_sad_pserialize_perform();
3251 
3252 	localcount_drain(&sah->localcount, &key_sad.cv_lc, &key_sad.lock);
3253 }
3254 
3255 static void
3256 key_destroy_sah(struct secashead *sah)
3257 {
3258 
3259 	rtcache_free(&sah->sa_route);
3260 
3261 	SAHLIST_ENTRY_DESTROY(sah);
3262 	localcount_fini(&sah->localcount);
3263 
3264 	if (sah->idents != NULL)
3265 		kmem_free(sah->idents, sah->idents_len);
3266 	if (sah->identd != NULL)
3267 		kmem_free(sah->identd, sah->identd_len);
3268 
3269 	kmem_free(sah, sizeof(*sah));
3270 }
3271 
3272 /*
3273  * allocating a new SA with LARVAL state.
3274  * key_api_add() and key_api_getspi() call,
3275  * and copy the values of mhp into new buffer.
3276  * When SAD message type is GETSPI:
3277  *	to set sequence number from acq_seq++,
3278  *	to set zero to SPI.
3279  *	not to call key_setsava().
3280  * OUT:	NULL	: fail
3281  *	others	: pointer to new secasvar.
3282  *
3283  * does not modify mbuf.  does not free mbuf on error.
3284  */
3285 static struct secasvar *
3286 key_newsav(struct mbuf *m, const struct sadb_msghdr *mhp,
3287     int *errp, const char* where, int tag)
3288 {
3289 	struct secasvar *newsav;
3290 	const struct sadb_sa *xsa;
3291 
3292 	KASSERT(!cpu_softintr_p());
3293 	KASSERT(m != NULL);
3294 	KASSERT(mhp != NULL);
3295 	KASSERT(mhp->msg != NULL);
3296 
3297 	newsav = kmem_zalloc(sizeof(struct secasvar), KM_SLEEP);
3298 
3299 	switch (mhp->msg->sadb_msg_type) {
3300 	case SADB_GETSPI:
3301 		newsav->spi = 0;
3302 
3303 #ifdef IPSEC_DOSEQCHECK
3304 		/* sync sequence number */
3305 		if (mhp->msg->sadb_msg_seq == 0)
3306 			newsav->seq =
3307 			    (acq_seq = (acq_seq == ~0 ? 1 : ++acq_seq));
3308 		else
3309 #endif
3310 			newsav->seq = mhp->msg->sadb_msg_seq;
3311 		break;
3312 
3313 	case SADB_ADD:
3314 		/* sanity check */
3315 		if (mhp->ext[SADB_EXT_SA] == NULL) {
3316 			IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
3317 			*errp = EINVAL;
3318 			goto error;
3319 		}
3320 		xsa = mhp->ext[SADB_EXT_SA];
3321 		newsav->spi = xsa->sadb_sa_spi;
3322 		newsav->seq = mhp->msg->sadb_msg_seq;
3323 		break;
3324 	default:
3325 		*errp = EINVAL;
3326 		goto error;
3327 	}
3328 
3329 	/* copy sav values */
3330 	if (mhp->msg->sadb_msg_type != SADB_GETSPI) {
3331 		*errp = key_setsaval(newsav, m, mhp);
3332 		if (*errp)
3333 			goto error;
3334 	} else {
3335 		/* We don't allow lft_c to be NULL */
3336 		newsav->lft_c = kmem_zalloc(sizeof(struct sadb_lifetime),
3337 		    KM_SLEEP);
3338 		newsav->lft_c_counters_percpu =
3339 		    percpu_alloc(sizeof(lifetime_counters_t));
3340 	}
3341 
3342 	/* reset created */
3343 	newsav->created = time_uptime;
3344 	newsav->pid = mhp->msg->sadb_msg_pid;
3345 
3346 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
3347 	    "DP from %s:%u return SA:%p\n", where, tag, newsav);
3348 	return newsav;
3349 
3350 error:
3351 	KASSERT(*errp != 0);
3352 	kmem_free(newsav, sizeof(*newsav));
3353 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
3354 	    "DP from %s:%u return SA:NULL\n", where, tag);
3355 	return NULL;
3356 }
3357 
3358 
3359 static void
3360 key_clear_xform(struct secasvar *sav)
3361 {
3362 
3363 	/*
3364 	 * Cleanup xform state.  Note that zeroize'ing causes the
3365 	 * keys to be cleared; otherwise we must do it ourself.
3366 	 */
3367 	if (sav->tdb_xform != NULL) {
3368 		sav->tdb_xform->xf_zeroize(sav);
3369 		sav->tdb_xform = NULL;
3370 	} else {
3371 		if (sav->key_auth != NULL)
3372 			explicit_memset(_KEYBUF(sav->key_auth), 0,
3373 			    _KEYLEN(sav->key_auth));
3374 		if (sav->key_enc != NULL)
3375 			explicit_memset(_KEYBUF(sav->key_enc), 0,
3376 			    _KEYLEN(sav->key_enc));
3377 	}
3378 }
3379 
3380 /*
3381  * free() SA variable entry.
3382  */
3383 static void
3384 key_delsav(struct secasvar *sav)
3385 {
3386 
3387 	key_clear_xform(sav);
3388 	key_freesaval(sav);
3389 	kmem_free(sav, sizeof(*sav));
3390 }
3391 
3392 /*
3393  * Must be called in a pserialize read section. A held sah
3394  * must be released by key_sah_unref after use.
3395  */
3396 static void
3397 key_sah_ref(struct secashead *sah)
3398 {
3399 
3400 	localcount_acquire(&sah->localcount);
3401 }
3402 
3403 /*
3404  * Must be called without holding key_sad.lock because the lock
3405  * would be held in localcount_release.
3406  */
3407 static void
3408 key_sah_unref(struct secashead *sah)
3409 {
3410 
3411 	KDASSERT(mutex_ownable(&key_sad.lock));
3412 
3413 	localcount_release(&sah->localcount, &key_sad.cv_lc, &key_sad.lock);
3414 }
3415 
3416 /*
3417  * Search SAD and return sah. Must be called in a pserialize
3418  * read section.
3419  * OUT:
3420  *	NULL	: not found
3421  *	others	: found, pointer to a SA.
3422  */
3423 static struct secashead *
3424 key_getsah(const struct secasindex *saidx, int flag)
3425 {
3426 	struct secashead *sah;
3427 
3428 	SAHLIST_READER_FOREACH_SAIDX(sah, saidx) {
3429 		if (sah->state == SADB_SASTATE_DEAD)
3430 			continue;
3431 		if (key_saidx_match(&sah->saidx, saidx, flag))
3432 			return sah;
3433 	}
3434 
3435 	return NULL;
3436 }
3437 
3438 /*
3439  * Search SAD and return sah. If sah is returned, the caller must call
3440  * key_sah_unref to releaset a reference.
3441  * OUT:
3442  *	NULL	: not found
3443  *	others	: found, pointer to a SA.
3444  */
3445 static struct secashead *
3446 key_getsah_ref(const struct secasindex *saidx, int flag)
3447 {
3448 	struct secashead *sah;
3449 	int s;
3450 
3451 	s = pserialize_read_enter();
3452 	sah = key_getsah(saidx, flag);
3453 	if (sah != NULL)
3454 		key_sah_ref(sah);
3455 	pserialize_read_exit(s);
3456 
3457 	return sah;
3458 }
3459 
3460 /*
3461  * check not to be duplicated SPI.
3462  * NOTE: this function is too slow due to searching all SAD.
3463  * OUT:
3464  *	NULL	: not found
3465  *	others	: found, pointer to a SA.
3466  */
3467 static bool
3468 key_checkspidup(const struct secasindex *saidx, u_int32_t spi)
3469 {
3470 	struct secashead *sah;
3471 	struct secasvar *sav;
3472 
3473 	/* check address family */
3474 	if (saidx->src.sa.sa_family != saidx->dst.sa.sa_family) {
3475 		IPSECLOG(LOG_DEBUG,
3476 		    "address family mismatched src %u, dst %u.\n",
3477 		    saidx->src.sa.sa_family, saidx->dst.sa.sa_family);
3478 		return false;
3479 	}
3480 
3481 	/* check all SAD */
3482 	/* key_ismyaddr may sleep, so use mutex, not pserialize, here. */
3483 	mutex_enter(&key_sad.lock);
3484 	SAHLIST_WRITER_FOREACH(sah) {
3485 		if (!key_ismyaddr((struct sockaddr *)&sah->saidx.dst))
3486 			continue;
3487 		sav = key_getsavbyspi(sah, spi);
3488 		if (sav != NULL) {
3489 			KEY_SA_UNREF(&sav);
3490 			mutex_exit(&key_sad.lock);
3491 			return true;
3492 		}
3493 	}
3494 	mutex_exit(&key_sad.lock);
3495 
3496 	return false;
3497 }
3498 
3499 /*
3500  * search SAD litmited alive SA, protocol, SPI.
3501  * OUT:
3502  *	NULL	: not found
3503  *	others	: found, pointer to a SA.
3504  */
3505 static struct secasvar *
3506 key_getsavbyspi(struct secashead *sah, u_int32_t spi)
3507 {
3508 	struct secasvar *sav = NULL;
3509 	u_int state;
3510 	int s;
3511 
3512 	/* search all status */
3513 	s = pserialize_read_enter();
3514 	SASTATE_ALIVE_FOREACH(state) {
3515 		SAVLIST_READER_FOREACH(sav, sah, state) {
3516 			/* sanity check */
3517 			if (sav->state != state) {
3518 				IPSECLOG(LOG_DEBUG,
3519 				    "invalid sav->state (queue: %d SA: %d)\n",
3520 				    state, sav->state);
3521 				continue;
3522 			}
3523 
3524 			if (sav->spi == spi) {
3525 				KEY_SA_REF(sav);
3526 				goto out;
3527 			}
3528 		}
3529 	}
3530 out:
3531 	pserialize_read_exit(s);
3532 
3533 	return sav;
3534 }
3535 
3536 /*
3537  * Free allocated data to member variables of sav:
3538  * sav->replay, sav->key_* and sav->lft_*.
3539  */
3540 static void
3541 key_freesaval(struct secasvar *sav)
3542 {
3543 
3544 	KASSERT(key_sa_refcnt(sav) == 0);
3545 
3546 	if (sav->replay != NULL)
3547 		kmem_intr_free(sav->replay, sav->replay_len);
3548 	if (sav->key_auth != NULL)
3549 		kmem_intr_free(sav->key_auth, sav->key_auth_len);
3550 	if (sav->key_enc != NULL)
3551 		kmem_intr_free(sav->key_enc, sav->key_enc_len);
3552 	if (sav->lft_c_counters_percpu != NULL) {
3553 		percpu_free(sav->lft_c_counters_percpu,
3554 		    sizeof(lifetime_counters_t));
3555 	}
3556 	if (sav->lft_c != NULL)
3557 		kmem_intr_free(sav->lft_c, sizeof(*(sav->lft_c)));
3558 	if (sav->lft_h != NULL)
3559 		kmem_intr_free(sav->lft_h, sizeof(*(sav->lft_h)));
3560 	if (sav->lft_s != NULL)
3561 		kmem_intr_free(sav->lft_s, sizeof(*(sav->lft_s)));
3562 }
3563 
3564 /*
3565  * copy SA values from PF_KEY message except *SPI, SEQ, PID, STATE and TYPE*.
3566  * You must update these if need.
3567  * OUT:	0:	success.
3568  *	!0:	failure.
3569  *
3570  * does not modify mbuf.  does not free mbuf on error.
3571  */
3572 static int
3573 key_setsaval(struct secasvar *sav, struct mbuf *m,
3574 	     const struct sadb_msghdr *mhp)
3575 {
3576 	int error = 0;
3577 
3578 	KASSERT(!cpu_softintr_p());
3579 	KASSERT(m != NULL);
3580 	KASSERT(mhp != NULL);
3581 	KASSERT(mhp->msg != NULL);
3582 
3583 	/* We shouldn't initialize sav variables while someone uses it. */
3584 	KASSERT(key_sa_refcnt(sav) == 0);
3585 
3586 	/* SA */
3587 	if (mhp->ext[SADB_EXT_SA] != NULL) {
3588 		const struct sadb_sa *sa0;
3589 
3590 		sa0 = mhp->ext[SADB_EXT_SA];
3591 		if (mhp->extlen[SADB_EXT_SA] < sizeof(*sa0)) {
3592 			error = EINVAL;
3593 			goto fail;
3594 		}
3595 
3596 		sav->alg_auth = sa0->sadb_sa_auth;
3597 		sav->alg_enc = sa0->sadb_sa_encrypt;
3598 		sav->flags = sa0->sadb_sa_flags;
3599 
3600 		/* replay window */
3601 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0) {
3602 			size_t len = sizeof(struct secreplay) +
3603 			    sa0->sadb_sa_replay;
3604 			sav->replay = kmem_zalloc(len, KM_SLEEP);
3605 			sav->replay_len = len;
3606 			if (sa0->sadb_sa_replay != 0)
3607 				sav->replay->bitmap = (char*)(sav->replay+1);
3608 			sav->replay->wsize = sa0->sadb_sa_replay;
3609 		}
3610 	}
3611 
3612 	/* Authentication keys */
3613 	if (mhp->ext[SADB_EXT_KEY_AUTH] != NULL) {
3614 		const struct sadb_key *key0;
3615 		int len;
3616 
3617 		key0 = mhp->ext[SADB_EXT_KEY_AUTH];
3618 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3619 
3620 		error = 0;
3621 		if (len < sizeof(*key0)) {
3622 			error = EINVAL;
3623 			goto fail;
3624 		}
3625 		switch (mhp->msg->sadb_msg_satype) {
3626 		case SADB_SATYPE_AH:
3627 		case SADB_SATYPE_ESP:
3628 		case SADB_X_SATYPE_TCPSIGNATURE:
3629 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3630 			    sav->alg_auth != SADB_X_AALG_NULL)
3631 				error = EINVAL;
3632 			break;
3633 		case SADB_X_SATYPE_IPCOMP:
3634 		default:
3635 			error = EINVAL;
3636 			break;
3637 		}
3638 		if (error) {
3639 			IPSECLOG(LOG_DEBUG, "invalid key_auth values.\n");
3640 			goto fail;
3641 		}
3642 
3643 		sav->key_auth = key_newbuf(key0, len);
3644 		sav->key_auth_len = len;
3645 	}
3646 
3647 	/* Encryption key */
3648 	if (mhp->ext[SADB_EXT_KEY_ENCRYPT] != NULL) {
3649 		const struct sadb_key *key0;
3650 		int len;
3651 
3652 		key0 = mhp->ext[SADB_EXT_KEY_ENCRYPT];
3653 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3654 
3655 		error = 0;
3656 		if (len < sizeof(*key0)) {
3657 			error = EINVAL;
3658 			goto fail;
3659 		}
3660 		switch (mhp->msg->sadb_msg_satype) {
3661 		case SADB_SATYPE_ESP:
3662 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3663 			    sav->alg_enc != SADB_EALG_NULL) {
3664 				error = EINVAL;
3665 				break;
3666 			}
3667 			sav->key_enc = key_newbuf(key0, len);
3668 			sav->key_enc_len = len;
3669 			break;
3670 		case SADB_X_SATYPE_IPCOMP:
3671 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3672 				error = EINVAL;
3673 			sav->key_enc = NULL;	/*just in case*/
3674 			break;
3675 		case SADB_SATYPE_AH:
3676 		case SADB_X_SATYPE_TCPSIGNATURE:
3677 		default:
3678 			error = EINVAL;
3679 			break;
3680 		}
3681 		if (error) {
3682 			IPSECLOG(LOG_DEBUG, "invalid key_enc value.\n");
3683 			goto fail;
3684 		}
3685 	}
3686 
3687 	/* set iv */
3688 	sav->ivlen = 0;
3689 
3690 	switch (mhp->msg->sadb_msg_satype) {
3691 	case SADB_SATYPE_AH:
3692 		error = xform_init(sav, XF_AH);
3693 		break;
3694 	case SADB_SATYPE_ESP:
3695 		error = xform_init(sav, XF_ESP);
3696 		break;
3697 	case SADB_X_SATYPE_IPCOMP:
3698 		error = xform_init(sav, XF_IPCOMP);
3699 		break;
3700 	case SADB_X_SATYPE_TCPSIGNATURE:
3701 		error = xform_init(sav, XF_TCPSIGNATURE);
3702 		break;
3703 	}
3704 	if (error) {
3705 		IPSECLOG(LOG_DEBUG, "unable to initialize SA type %u.\n",
3706 		    mhp->msg->sadb_msg_satype);
3707 		goto fail;
3708 	}
3709 
3710 	/* reset created */
3711 	sav->created = time_uptime;
3712 
3713 	/* make lifetime for CURRENT */
3714 	sav->lft_c = kmem_alloc(sizeof(struct sadb_lifetime), KM_SLEEP);
3715 
3716 	sav->lft_c->sadb_lifetime_len =
3717 	    PFKEY_UNIT64(sizeof(struct sadb_lifetime));
3718 	sav->lft_c->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
3719 	sav->lft_c->sadb_lifetime_allocations = 0;
3720 	sav->lft_c->sadb_lifetime_bytes = 0;
3721 	sav->lft_c->sadb_lifetime_addtime = time_uptime;
3722 	sav->lft_c->sadb_lifetime_usetime = 0;
3723 
3724 	sav->lft_c_counters_percpu = percpu_alloc(sizeof(lifetime_counters_t));
3725 
3726 	/* lifetimes for HARD and SOFT */
3727     {
3728 	const struct sadb_lifetime *lft0;
3729 
3730 	lft0 = mhp->ext[SADB_EXT_LIFETIME_HARD];
3731 	if (lft0 != NULL) {
3732 		if (mhp->extlen[SADB_EXT_LIFETIME_HARD] < sizeof(*lft0)) {
3733 			error = EINVAL;
3734 			goto fail;
3735 		}
3736 		sav->lft_h = key_newbuf(lft0, sizeof(*lft0));
3737 	}
3738 
3739 	lft0 = mhp->ext[SADB_EXT_LIFETIME_SOFT];
3740 	if (lft0 != NULL) {
3741 		if (mhp->extlen[SADB_EXT_LIFETIME_SOFT] < sizeof(*lft0)) {
3742 			error = EINVAL;
3743 			goto fail;
3744 		}
3745 		sav->lft_s = key_newbuf(lft0, sizeof(*lft0));
3746 		/* to be initialize ? */
3747 	}
3748     }
3749 
3750 	return 0;
3751 
3752  fail:
3753 	key_clear_xform(sav);
3754 	key_freesaval(sav);
3755 
3756 	return error;
3757 }
3758 
3759 /*
3760  * validation with a secasvar entry, and set SADB_SATYPE_MATURE.
3761  * OUT:	0:	valid
3762  *	other:	errno
3763  */
3764 static int
3765 key_init_xform(struct secasvar *sav)
3766 {
3767 	int error;
3768 
3769 	/* We shouldn't initialize sav variables while someone uses it. */
3770 	KASSERT(key_sa_refcnt(sav) == 0);
3771 
3772 	/* check SPI value */
3773 	switch (sav->sah->saidx.proto) {
3774 	case IPPROTO_ESP:
3775 	case IPPROTO_AH:
3776 		if (ntohl(sav->spi) <= 255) {
3777 			IPSECLOG(LOG_DEBUG, "illegal range of SPI %u.\n",
3778 			    (u_int32_t)ntohl(sav->spi));
3779 			return EINVAL;
3780 		}
3781 		break;
3782 	}
3783 
3784 	/* check algo */
3785 	switch (sav->sah->saidx.proto) {
3786 	case IPPROTO_AH:
3787 	case IPPROTO_TCP:
3788 		if (sav->alg_enc != SADB_EALG_NONE) {
3789 			IPSECLOG(LOG_DEBUG,
3790 			    "protocol %u and algorithm mismatched %u != %u.\n",
3791 			    sav->sah->saidx.proto,
3792 			    sav->alg_enc, SADB_EALG_NONE);
3793 			return EINVAL;
3794 		}
3795 		break;
3796 	case IPPROTO_IPCOMP:
3797 		if (sav->alg_auth != SADB_AALG_NONE) {
3798 			IPSECLOG(LOG_DEBUG,
3799 			    "protocol %u and algorithm mismatched %d != %d.\n",
3800 			    sav->sah->saidx.proto,
3801 			    sav->alg_auth, SADB_AALG_NONE);
3802 			return(EINVAL);
3803 		}
3804 		break;
3805 	default:
3806 		break;
3807 	}
3808 
3809 	/* check satype */
3810 	switch (sav->sah->saidx.proto) {
3811 	case IPPROTO_ESP:
3812 		/* check flags */
3813 		if ((sav->flags & (SADB_X_EXT_OLD|SADB_X_EXT_DERIV)) ==
3814 		    (SADB_X_EXT_OLD|SADB_X_EXT_DERIV)) {
3815 			IPSECLOG(LOG_DEBUG,
3816 			    "invalid flag (derived) given to old-esp.\n");
3817 			return EINVAL;
3818 		}
3819 		error = xform_init(sav, XF_ESP);
3820 		break;
3821 	case IPPROTO_AH:
3822 		/* check flags */
3823 		if (sav->flags & SADB_X_EXT_DERIV) {
3824 			IPSECLOG(LOG_DEBUG,
3825 			    "invalid flag (derived) given to AH SA.\n");
3826 			return EINVAL;
3827 		}
3828 		error = xform_init(sav, XF_AH);
3829 		break;
3830 	case IPPROTO_IPCOMP:
3831 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0
3832 		    && ntohl(sav->spi) >= 0x10000) {
3833 			IPSECLOG(LOG_DEBUG, "invalid cpi for IPComp.\n");
3834 			return(EINVAL);
3835 		}
3836 		error = xform_init(sav, XF_IPCOMP);
3837 		break;
3838 	case IPPROTO_TCP:
3839 		error = xform_init(sav, XF_TCPSIGNATURE);
3840 		break;
3841 	default:
3842 		IPSECLOG(LOG_DEBUG, "Invalid satype.\n");
3843 		error = EPROTONOSUPPORT;
3844 		break;
3845 	}
3846 
3847 	return error;
3848 }
3849 
3850 /*
3851  * subroutine for SADB_GET and SADB_DUMP. It never return NULL.
3852  */
3853 static struct mbuf *
3854 key_setdumpsa(struct secasvar *sav, u_int8_t type, u_int8_t satype,
3855 	      u_int32_t seq, u_int32_t pid)
3856 {
3857 	struct mbuf *result = NULL, *tres = NULL, *m;
3858 	int l = 0;
3859 	int i;
3860 	void *p;
3861 	struct sadb_lifetime lt;
3862 	int dumporder[] = {
3863 		SADB_EXT_SA, SADB_X_EXT_SA2,
3864 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3865 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3866 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY, SADB_EXT_KEY_AUTH,
3867 		SADB_EXT_KEY_ENCRYPT, SADB_EXT_IDENTITY_SRC,
3868 		SADB_EXT_IDENTITY_DST, SADB_EXT_SENSITIVITY,
3869 		SADB_X_EXT_NAT_T_TYPE,
3870 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3871 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3872 		SADB_X_EXT_NAT_T_FRAG,
3873 
3874 	};
3875 
3876 	m = key_setsadbmsg(type, 0, satype, seq, pid, key_sa_refcnt(sav), M_WAITOK);
3877 	result = m;
3878 
3879 	for (i = __arraycount(dumporder) - 1; i >= 0; i--) {
3880 		m = NULL;
3881 		p = NULL;
3882 		switch (dumporder[i]) {
3883 		case SADB_EXT_SA:
3884 			m = key_setsadbsa(sav);
3885 			break;
3886 
3887 		case SADB_X_EXT_SA2:
3888 			m = key_setsadbxsa2(sav->sah->saidx.mode,
3889 			    sav->replay ? sav->replay->count : 0,
3890 			    sav->sah->saidx.reqid);
3891 			break;
3892 
3893 		case SADB_EXT_ADDRESS_SRC:
3894 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3895 			    &sav->sah->saidx.src.sa,
3896 			    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
3897 			break;
3898 
3899 		case SADB_EXT_ADDRESS_DST:
3900 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3901 			    &sav->sah->saidx.dst.sa,
3902 			    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
3903 			break;
3904 
3905 		case SADB_EXT_KEY_AUTH:
3906 			if (!sav->key_auth)
3907 				continue;
3908 			l = PFKEY_UNUNIT64(sav->key_auth->sadb_key_len);
3909 			p = sav->key_auth;
3910 			break;
3911 
3912 		case SADB_EXT_KEY_ENCRYPT:
3913 			if (!sav->key_enc)
3914 				continue;
3915 			l = PFKEY_UNUNIT64(sav->key_enc->sadb_key_len);
3916 			p = sav->key_enc;
3917 			break;
3918 
3919 		case SADB_EXT_LIFETIME_CURRENT: {
3920 			lifetime_counters_t sum = {0};
3921 
3922 			KASSERT(sav->lft_c != NULL);
3923 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_c)->sadb_ext_len);
3924 			memcpy(&lt, sav->lft_c, sizeof(struct sadb_lifetime));
3925 			lt.sadb_lifetime_addtime =
3926 			    time_mono_to_wall(lt.sadb_lifetime_addtime);
3927 			lt.sadb_lifetime_usetime =
3928 			    time_mono_to_wall(lt.sadb_lifetime_usetime);
3929 			percpu_foreach(sav->lft_c_counters_percpu,
3930 			    key_sum_lifetime_counters, sum);
3931 			lt.sadb_lifetime_allocations =
3932 			    sum[LIFETIME_COUNTER_ALLOCATIONS];
3933 			lt.sadb_lifetime_bytes =
3934 			    sum[LIFETIME_COUNTER_BYTES];
3935 			p = &lt;
3936 			break;
3937 		    }
3938 
3939 		case SADB_EXT_LIFETIME_HARD:
3940 			if (!sav->lft_h)
3941 				continue;
3942 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_h)->sadb_ext_len);
3943 			p = sav->lft_h;
3944 			break;
3945 
3946 		case SADB_EXT_LIFETIME_SOFT:
3947 			if (!sav->lft_s)
3948 				continue;
3949 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_s)->sadb_ext_len);
3950 			p = sav->lft_s;
3951 			break;
3952 
3953 		case SADB_X_EXT_NAT_T_TYPE:
3954 			m = key_setsadbxtype(sav->natt_type);
3955 			break;
3956 
3957 		case SADB_X_EXT_NAT_T_DPORT:
3958 			if (sav->natt_type == 0)
3959 				continue;
3960 			m = key_setsadbxport(
3961 			    key_portfromsaddr(&sav->sah->saidx.dst),
3962 			    SADB_X_EXT_NAT_T_DPORT);
3963 			break;
3964 
3965 		case SADB_X_EXT_NAT_T_SPORT:
3966 			if (sav->natt_type == 0)
3967 				continue;
3968 			m = key_setsadbxport(
3969 			    key_portfromsaddr(&sav->sah->saidx.src),
3970 			    SADB_X_EXT_NAT_T_SPORT);
3971 			break;
3972 
3973 		case SADB_X_EXT_NAT_T_FRAG:
3974 			/* don't send frag info if not set */
3975 			if (sav->natt_type == 0 || sav->esp_frag == IP_MAXPACKET)
3976 				continue;
3977 			m = key_setsadbxfrag(sav->esp_frag);
3978 			break;
3979 
3980 		case SADB_X_EXT_NAT_T_OAI:
3981 		case SADB_X_EXT_NAT_T_OAR:
3982 			continue;
3983 
3984 		case SADB_EXT_ADDRESS_PROXY:
3985 		case SADB_EXT_IDENTITY_SRC:
3986 		case SADB_EXT_IDENTITY_DST:
3987 			/* XXX: should we brought from SPD ? */
3988 		case SADB_EXT_SENSITIVITY:
3989 		default:
3990 			continue;
3991 		}
3992 
3993 		KASSERT(!(m && p));
3994 		KASSERT(m != NULL || p != NULL);
3995 		if (p && tres) {
3996 			M_PREPEND(tres, l, M_WAITOK);
3997 			memcpy(mtod(tres, void *), p, l);
3998 			continue;
3999 		}
4000 		if (p) {
4001 			m = key_alloc_mbuf(l, M_WAITOK);
4002 			m_copyback(m, 0, l, p);
4003 		}
4004 
4005 		if (tres)
4006 			m_cat(m, tres);
4007 		tres = m;
4008 	}
4009 
4010 	m_cat(result, tres);
4011 	tres = NULL; /* avoid free on error below */
4012 
4013 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
4014 
4015 	result->m_pkthdr.len = 0;
4016 	for (m = result; m; m = m->m_next)
4017 		result->m_pkthdr.len += m->m_len;
4018 
4019 	mtod(result, struct sadb_msg *)->sadb_msg_len =
4020 	    PFKEY_UNIT64(result->m_pkthdr.len);
4021 
4022 	return result;
4023 }
4024 
4025 
4026 /*
4027  * set a type in sadb_x_nat_t_type
4028  */
4029 static struct mbuf *
4030 key_setsadbxtype(u_int16_t type)
4031 {
4032 	struct mbuf *m;
4033 	size_t len;
4034 	struct sadb_x_nat_t_type *p;
4035 
4036 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
4037 
4038 	m = key_alloc_mbuf(len, M_WAITOK);
4039 	KASSERT(m->m_next == NULL);
4040 
4041 	p = mtod(m, struct sadb_x_nat_t_type *);
4042 
4043 	memset(p, 0, len);
4044 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
4045 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
4046 	p->sadb_x_nat_t_type_type = type;
4047 
4048 	return m;
4049 }
4050 /*
4051  * set a port in sadb_x_nat_t_port. port is in network order
4052  */
4053 static struct mbuf *
4054 key_setsadbxport(u_int16_t port, u_int16_t type)
4055 {
4056 	struct mbuf *m;
4057 	size_t len;
4058 	struct sadb_x_nat_t_port *p;
4059 
4060 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
4061 
4062 	m = key_alloc_mbuf(len, M_WAITOK);
4063 	KASSERT(m->m_next == NULL);
4064 
4065 	p = mtod(m, struct sadb_x_nat_t_port *);
4066 
4067 	memset(p, 0, len);
4068 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
4069 	p->sadb_x_nat_t_port_exttype = type;
4070 	p->sadb_x_nat_t_port_port = port;
4071 
4072 	return m;
4073 }
4074 
4075 /*
4076  * set fragmentation info in sadb_x_nat_t_frag
4077  */
4078 static struct mbuf *
4079 key_setsadbxfrag(u_int16_t flen)
4080 {
4081 	struct mbuf *m;
4082 	size_t len;
4083 	struct sadb_x_nat_t_frag *p;
4084 
4085 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_frag));
4086 
4087 	m = key_alloc_mbuf(len, M_WAITOK);
4088 	KASSERT(m->m_next == NULL);
4089 
4090 	p = mtod(m, struct sadb_x_nat_t_frag *);
4091 
4092 	memset(p, 0, len);
4093 	p->sadb_x_nat_t_frag_len = PFKEY_UNIT64(len);
4094 	p->sadb_x_nat_t_frag_exttype = SADB_X_EXT_NAT_T_FRAG;
4095 	p->sadb_x_nat_t_frag_fraglen = flen;
4096 
4097 	return m;
4098 }
4099 
4100 /*
4101  * Get port from sockaddr, port is in network order
4102  */
4103 u_int16_t
4104 key_portfromsaddr(const union sockaddr_union *saddr)
4105 {
4106 	u_int16_t port;
4107 
4108 	switch (saddr->sa.sa_family) {
4109 	case AF_INET: {
4110 		port = saddr->sin.sin_port;
4111 		break;
4112 	}
4113 #ifdef INET6
4114 	case AF_INET6: {
4115 		port = saddr->sin6.sin6_port;
4116 		break;
4117 	}
4118 #endif
4119 	default:
4120 		printf("%s: unexpected address family\n", __func__);
4121 		port = 0;
4122 		break;
4123 	}
4124 
4125 	return port;
4126 }
4127 
4128 
4129 /*
4130  * Set port is struct sockaddr. port is in network order
4131  */
4132 static void
4133 key_porttosaddr(union sockaddr_union *saddr, u_int16_t port)
4134 {
4135 	switch (saddr->sa.sa_family) {
4136 	case AF_INET: {
4137 		saddr->sin.sin_port = port;
4138 		break;
4139 	}
4140 #ifdef INET6
4141 	case AF_INET6: {
4142 		saddr->sin6.sin6_port = port;
4143 		break;
4144 	}
4145 #endif
4146 	default:
4147 		printf("%s: unexpected address family %d\n", __func__,
4148 		    saddr->sa.sa_family);
4149 		break;
4150 	}
4151 
4152 	return;
4153 }
4154 
4155 /*
4156  * Safety check sa_len
4157  */
4158 static int
4159 key_checksalen(const union sockaddr_union *saddr)
4160 {
4161 	switch (saddr->sa.sa_family) {
4162 	case AF_INET:
4163 		if (saddr->sa.sa_len != sizeof(struct sockaddr_in))
4164 			return -1;
4165 		break;
4166 #ifdef INET6
4167 	case AF_INET6:
4168 		if (saddr->sa.sa_len != sizeof(struct sockaddr_in6))
4169 			return -1;
4170 		break;
4171 #endif
4172 	default:
4173 		printf("%s: unexpected sa_family %d\n", __func__,
4174 		    saddr->sa.sa_family);
4175 			return -1;
4176 		break;
4177 	}
4178 	return 0;
4179 }
4180 
4181 
4182 /*
4183  * set data into sadb_msg.
4184  */
4185 static struct mbuf *
4186 key_setsadbmsg(u_int8_t type,  u_int16_t tlen, u_int8_t satype,
4187 	       u_int32_t seq, pid_t pid, u_int16_t reserved, int mflag)
4188 {
4189 	struct mbuf *m;
4190 	struct sadb_msg *p;
4191 	int len;
4192 
4193 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MCLBYTES);
4194 
4195 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
4196 
4197 	m = key_alloc_mbuf_simple(len, mflag);
4198 	if (!m)
4199 		return NULL;
4200 	m->m_pkthdr.len = m->m_len = len;
4201 	m->m_next = NULL;
4202 
4203 	p = mtod(m, struct sadb_msg *);
4204 
4205 	memset(p, 0, len);
4206 	p->sadb_msg_version = PF_KEY_V2;
4207 	p->sadb_msg_type = type;
4208 	p->sadb_msg_errno = 0;
4209 	p->sadb_msg_satype = satype;
4210 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
4211 	p->sadb_msg_reserved = reserved;
4212 	p->sadb_msg_seq = seq;
4213 	p->sadb_msg_pid = (u_int32_t)pid;
4214 
4215 	return m;
4216 }
4217 
4218 /*
4219  * copy secasvar data into sadb_address.
4220  */
4221 static struct mbuf *
4222 key_setsadbsa(struct secasvar *sav)
4223 {
4224 	struct mbuf *m;
4225 	struct sadb_sa *p;
4226 	int len;
4227 
4228 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
4229 	m = key_alloc_mbuf(len, M_WAITOK);
4230 	KASSERT(m->m_next == NULL);
4231 
4232 	p = mtod(m, struct sadb_sa *);
4233 
4234 	memset(p, 0, len);
4235 	p->sadb_sa_len = PFKEY_UNIT64(len);
4236 	p->sadb_sa_exttype = SADB_EXT_SA;
4237 	p->sadb_sa_spi = sav->spi;
4238 	p->sadb_sa_replay = (sav->replay != NULL ? sav->replay->wsize : 0);
4239 	p->sadb_sa_state = sav->state;
4240 	p->sadb_sa_auth = sav->alg_auth;
4241 	p->sadb_sa_encrypt = sav->alg_enc;
4242 	p->sadb_sa_flags = sav->flags;
4243 
4244 	return m;
4245 }
4246 
4247 static uint8_t
4248 key_sabits(const struct sockaddr *saddr)
4249 {
4250 	switch (saddr->sa_family) {
4251 	case AF_INET:
4252 		return _BITS(sizeof(struct in_addr));
4253 	case AF_INET6:
4254 		return _BITS(sizeof(struct in6_addr));
4255 	default:
4256 		return FULLMASK;
4257 	}
4258 }
4259 
4260 /*
4261  * set data into sadb_address.
4262  */
4263 static struct mbuf *
4264 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
4265 		u_int8_t prefixlen, u_int16_t ul_proto, int mflag)
4266 {
4267 	struct mbuf *m;
4268 	struct sadb_address *p;
4269 	size_t len;
4270 
4271 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
4272 	    PFKEY_ALIGN8(saddr->sa_len);
4273 	m = key_alloc_mbuf(len, mflag);
4274 	if (!m || m->m_next) {	/*XXX*/
4275 		if (m)
4276 			m_freem(m);
4277 		return NULL;
4278 	}
4279 
4280 	p = mtod(m, struct sadb_address *);
4281 
4282 	memset(p, 0, len);
4283 	p->sadb_address_len = PFKEY_UNIT64(len);
4284 	p->sadb_address_exttype = exttype;
4285 	p->sadb_address_proto = ul_proto;
4286 	if (prefixlen == FULLMASK) {
4287 		prefixlen = key_sabits(saddr);
4288 	}
4289 	p->sadb_address_prefixlen = prefixlen;
4290 	p->sadb_address_reserved = 0;
4291 
4292 	memcpy(mtod(m, char *) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
4293 	    saddr, saddr->sa_len);
4294 
4295 	return m;
4296 }
4297 
4298 #if 0
4299 /*
4300  * set data into sadb_ident.
4301  */
4302 static struct mbuf *
4303 key_setsadbident(u_int16_t exttype, u_int16_t idtype,
4304 		 void *string, int stringlen, u_int64_t id)
4305 {
4306 	struct mbuf *m;
4307 	struct sadb_ident *p;
4308 	size_t len;
4309 
4310 	len = PFKEY_ALIGN8(sizeof(struct sadb_ident)) + PFKEY_ALIGN8(stringlen);
4311 	m = key_alloc_mbuf(len);
4312 	if (!m || m->m_next) {	/*XXX*/
4313 		if (m)
4314 			m_freem(m);
4315 		return NULL;
4316 	}
4317 
4318 	p = mtod(m, struct sadb_ident *);
4319 
4320 	memset(p, 0, len);
4321 	p->sadb_ident_len = PFKEY_UNIT64(len);
4322 	p->sadb_ident_exttype = exttype;
4323 	p->sadb_ident_type = idtype;
4324 	p->sadb_ident_reserved = 0;
4325 	p->sadb_ident_id = id;
4326 
4327 	memcpy(mtod(m, void *) + PFKEY_ALIGN8(sizeof(struct sadb_ident)),
4328 	   	   string, stringlen);
4329 
4330 	return m;
4331 }
4332 #endif
4333 
4334 /*
4335  * set data into sadb_x_sa2.
4336  */
4337 static struct mbuf *
4338 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int16_t reqid)
4339 {
4340 	struct mbuf *m;
4341 	struct sadb_x_sa2 *p;
4342 	size_t len;
4343 
4344 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
4345 	m = key_alloc_mbuf(len, M_WAITOK);
4346 	KASSERT(m->m_next == NULL);
4347 
4348 	p = mtod(m, struct sadb_x_sa2 *);
4349 
4350 	memset(p, 0, len);
4351 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
4352 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
4353 	p->sadb_x_sa2_mode = mode;
4354 	p->sadb_x_sa2_reserved1 = 0;
4355 	p->sadb_x_sa2_reserved2 = 0;
4356 	p->sadb_x_sa2_sequence = seq;
4357 	p->sadb_x_sa2_reqid = reqid;
4358 
4359 	return m;
4360 }
4361 
4362 /*
4363  * set data into sadb_x_policy
4364  */
4365 static struct mbuf *
4366 key_setsadbxpolicy(const u_int16_t type, const u_int8_t dir, const u_int32_t id,
4367     int mflag)
4368 {
4369 	struct mbuf *m;
4370 	struct sadb_x_policy *p;
4371 	size_t len;
4372 
4373 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
4374 	m = key_alloc_mbuf(len, mflag);
4375 	if (!m || m->m_next) {	/*XXX*/
4376 		if (m)
4377 			m_freem(m);
4378 		return NULL;
4379 	}
4380 
4381 	p = mtod(m, struct sadb_x_policy *);
4382 
4383 	memset(p, 0, len);
4384 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
4385 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
4386 	p->sadb_x_policy_type = type;
4387 	p->sadb_x_policy_dir = dir;
4388 	p->sadb_x_policy_id = id;
4389 
4390 	return m;
4391 }
4392 
4393 /* %%% utilities */
4394 /*
4395  * copy a buffer into the new buffer allocated.
4396  */
4397 static void *
4398 key_newbuf(const void *src, u_int len)
4399 {
4400 	void *new;
4401 
4402 	new = kmem_alloc(len, KM_SLEEP);
4403 	memcpy(new, src, len);
4404 
4405 	return new;
4406 }
4407 
4408 /* compare my own address
4409  * OUT:	1: true, i.e. my address.
4410  *	0: false
4411  */
4412 int
4413 key_ismyaddr(const struct sockaddr *sa)
4414 {
4415 #ifdef INET
4416 	const struct sockaddr_in *sin;
4417 	const struct in_ifaddr *ia;
4418 	int s;
4419 #endif
4420 
4421 	KASSERT(sa != NULL);
4422 
4423 	switch (sa->sa_family) {
4424 #ifdef INET
4425 	case AF_INET:
4426 		sin = (const struct sockaddr_in *)sa;
4427 		s = pserialize_read_enter();
4428 		IN_ADDRLIST_READER_FOREACH(ia) {
4429 			if (sin->sin_family == ia->ia_addr.sin_family &&
4430 			    sin->sin_len == ia->ia_addr.sin_len &&
4431 			    sin->sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
4432 			{
4433 				pserialize_read_exit(s);
4434 				return 1;
4435 			}
4436 		}
4437 		pserialize_read_exit(s);
4438 		break;
4439 #endif
4440 #ifdef INET6
4441 	case AF_INET6:
4442 		return key_ismyaddr6((const struct sockaddr_in6 *)sa);
4443 #endif
4444 	}
4445 
4446 	return 0;
4447 }
4448 
4449 #ifdef INET6
4450 /*
4451  * compare my own address for IPv6.
4452  * 1: ours
4453  * 0: other
4454  * NOTE: derived ip6_input() in KAME. This is necessary to modify more.
4455  */
4456 #include <netinet6/in6_var.h>
4457 
4458 static int
4459 key_ismyaddr6(const struct sockaddr_in6 *sin6)
4460 {
4461 	struct in6_ifaddr *ia;
4462 	int s;
4463 	struct psref psref;
4464 	int bound;
4465 	int ours = 1;
4466 
4467 	bound = curlwp_bind();
4468 	s = pserialize_read_enter();
4469 	IN6_ADDRLIST_READER_FOREACH(ia) {
4470 		bool ingroup;
4471 
4472 		if (key_sockaddr_match((const struct sockaddr *)&sin6,
4473 		    (const struct sockaddr *)&ia->ia_addr, 0)) {
4474 			pserialize_read_exit(s);
4475 			goto ours;
4476 		}
4477 		ia6_acquire(ia, &psref);
4478 		pserialize_read_exit(s);
4479 
4480 		/*
4481 		 * XXX Multicast
4482 		 * XXX why do we care about multlicast here while we don't care
4483 		 * about IPv4 multicast??
4484 		 * XXX scope
4485 		 */
4486 		ingroup = in6_multi_group(&sin6->sin6_addr, ia->ia_ifp);
4487 		if (ingroup) {
4488 			ia6_release(ia, &psref);
4489 			goto ours;
4490 		}
4491 
4492 		s = pserialize_read_enter();
4493 		ia6_release(ia, &psref);
4494 	}
4495 	pserialize_read_exit(s);
4496 
4497 	/* loopback, just for safety */
4498 	if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))
4499 		goto ours;
4500 
4501 	ours = 0;
4502 ours:
4503 	curlwp_bindx(bound);
4504 
4505 	return ours;
4506 }
4507 #endif /*INET6*/
4508 
4509 /*
4510  * compare two secasindex structure.
4511  * flag can specify to compare 2 saidxes.
4512  * compare two secasindex structure without both mode and reqid.
4513  * don't compare port.
4514  * IN:
4515  *      saidx0: source, it can be in SAD.
4516  *      saidx1: object.
4517  * OUT:
4518  *      1 : equal
4519  *      0 : not equal
4520  */
4521 static int
4522 key_saidx_match(
4523 	const struct secasindex *saidx0,
4524 	const struct secasindex *saidx1,
4525 	int flag)
4526 {
4527 	int chkport;
4528 	const struct sockaddr *sa0src, *sa0dst, *sa1src, *sa1dst;
4529 
4530 	KASSERT(saidx0 != NULL);
4531 	KASSERT(saidx1 != NULL);
4532 
4533 	/* sanity */
4534 	if (saidx0->proto != saidx1->proto)
4535 		return 0;
4536 
4537 	if (flag == CMP_EXACTLY) {
4538 		if (saidx0->mode != saidx1->mode)
4539 			return 0;
4540 		if (saidx0->reqid != saidx1->reqid)
4541 			return 0;
4542 		if (memcmp(&saidx0->src, &saidx1->src, saidx0->src.sa.sa_len) != 0 ||
4543 		    memcmp(&saidx0->dst, &saidx1->dst, saidx0->dst.sa.sa_len) != 0)
4544 			return 0;
4545 	} else {
4546 
4547 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
4548 		if (flag == CMP_MODE_REQID ||flag == CMP_REQID) {
4549 			/*
4550 			 * If reqid of SPD is non-zero, unique SA is required.
4551 			 * The result must be of same reqid in this case.
4552 			 */
4553 			if (saidx1->reqid != 0 && saidx0->reqid != saidx1->reqid)
4554 				return 0;
4555 		}
4556 
4557 		if (flag == CMP_MODE_REQID) {
4558 			if (saidx0->mode != IPSEC_MODE_ANY &&
4559 			    saidx0->mode != saidx1->mode)
4560 				return 0;
4561 		}
4562 
4563 
4564 		sa0src = &saidx0->src.sa;
4565 		sa0dst = &saidx0->dst.sa;
4566 		sa1src = &saidx1->src.sa;
4567 		sa1dst = &saidx1->dst.sa;
4568 		/*
4569 		 * If NAT-T is enabled, check ports for tunnel mode.
4570 		 * For ipsecif(4), check ports for transport mode, too.
4571 		 * Don't check ports if they are set to zero
4572 		 * in the SPD: This means we have a non-generated
4573 		 * SPD which can't know UDP ports.
4574 		 */
4575 		if (saidx1->mode == IPSEC_MODE_TUNNEL ||
4576 		    saidx1->mode == IPSEC_MODE_TRANSPORT)
4577 			chkport = PORT_LOOSE;
4578 		else
4579 			chkport = PORT_NONE;
4580 
4581 		if (!key_sockaddr_match(sa0src, sa1src, chkport)) {
4582 			return 0;
4583 		}
4584 		if (!key_sockaddr_match(sa0dst, sa1dst, chkport)) {
4585 			return 0;
4586 		}
4587 	}
4588 
4589 	return 1;
4590 }
4591 
4592 /*
4593  * compare two secindex structure exactly.
4594  * IN:
4595  *	spidx0: source, it is often in SPD.
4596  *	spidx1: object, it is often from PFKEY message.
4597  * OUT:
4598  *	1 : equal
4599  *	0 : not equal
4600  */
4601 static int
4602 key_spidx_match_exactly(
4603 	const struct secpolicyindex *spidx0,
4604 	const struct secpolicyindex *spidx1)
4605 {
4606 
4607 	KASSERT(spidx0 != NULL);
4608 	KASSERT(spidx1 != NULL);
4609 
4610 	/* sanity */
4611 	if (spidx0->prefs != spidx1->prefs ||
4612 	    spidx0->prefd != spidx1->prefd ||
4613 	    spidx0->ul_proto != spidx1->ul_proto)
4614 		return 0;
4615 
4616 	return key_sockaddr_match(&spidx0->src.sa, &spidx1->src.sa, PORT_STRICT) &&
4617 	       key_sockaddr_match(&spidx0->dst.sa, &spidx1->dst.sa, PORT_STRICT);
4618 }
4619 
4620 /*
4621  * compare two secindex structure with mask.
4622  * IN:
4623  *	spidx0: source, it is often in SPD.
4624  *	spidx1: object, it is often from IP header.
4625  * OUT:
4626  *	1 : equal
4627  *	0 : not equal
4628  */
4629 static int
4630 key_spidx_match_withmask(
4631 	const struct secpolicyindex *spidx0,
4632 	const struct secpolicyindex *spidx1)
4633 {
4634 
4635 	KASSERT(spidx0 != NULL);
4636 	KASSERT(spidx1 != NULL);
4637 
4638 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4639 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4640 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4641 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
4642 		return 0;
4643 
4644 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4645 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY &&
4646 	    spidx0->ul_proto != spidx1->ul_proto)
4647 		return 0;
4648 
4649 	switch (spidx0->src.sa.sa_family) {
4650 	case AF_INET:
4651 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY &&
4652 		    spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
4653 			return 0;
4654 		if (!key_bb_match_withmask(&spidx0->src.sin.sin_addr,
4655 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
4656 			return 0;
4657 		break;
4658 	case AF_INET6:
4659 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY &&
4660 		    spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
4661 			return 0;
4662 		/*
4663 		 * scope_id check. if sin6_scope_id is 0, we regard it
4664 		 * as a wildcard scope, which matches any scope zone ID.
4665 		 */
4666 		if (spidx0->src.sin6.sin6_scope_id &&
4667 		    spidx1->src.sin6.sin6_scope_id &&
4668 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
4669 			return 0;
4670 		if (!key_bb_match_withmask(&spidx0->src.sin6.sin6_addr,
4671 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
4672 			return 0;
4673 		break;
4674 	default:
4675 		/* XXX */
4676 		if (memcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
4677 			return 0;
4678 		break;
4679 	}
4680 
4681 	switch (spidx0->dst.sa.sa_family) {
4682 	case AF_INET:
4683 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY &&
4684 		    spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
4685 			return 0;
4686 		if (!key_bb_match_withmask(&spidx0->dst.sin.sin_addr,
4687 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
4688 			return 0;
4689 		break;
4690 	case AF_INET6:
4691 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY &&
4692 		    spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
4693 			return 0;
4694 		/*
4695 		 * scope_id check. if sin6_scope_id is 0, we regard it
4696 		 * as a wildcard scope, which matches any scope zone ID.
4697 		 */
4698 		if (spidx0->src.sin6.sin6_scope_id &&
4699 		    spidx1->src.sin6.sin6_scope_id &&
4700 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
4701 			return 0;
4702 		if (!key_bb_match_withmask(&spidx0->dst.sin6.sin6_addr,
4703 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
4704 			return 0;
4705 		break;
4706 	default:
4707 		/* XXX */
4708 		if (memcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
4709 			return 0;
4710 		break;
4711 	}
4712 
4713 	/* XXX Do we check other field ?  e.g. flowinfo */
4714 
4715 	return 1;
4716 }
4717 
4718 /* returns 0 on match */
4719 static int
4720 key_portcomp(in_port_t port1, in_port_t port2, int howport)
4721 {
4722 	switch (howport) {
4723 	case PORT_NONE:
4724 		return 0;
4725 	case PORT_LOOSE:
4726 		if (port1 == 0 || port2 == 0)
4727 			return 0;
4728 		/*FALLTHROUGH*/
4729 	case PORT_STRICT:
4730 		if (port1 != port2) {
4731 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4732 			    "port fail %d != %d\n", port1, port2);
4733 			return 1;
4734 		}
4735 		return 0;
4736 	default:
4737 		KASSERT(0);
4738 		return 1;
4739 	}
4740 }
4741 
4742 /* returns 1 on match */
4743 static int
4744 key_sockaddr_match(
4745 	const struct sockaddr *sa1,
4746 	const struct sockaddr *sa2,
4747 	int howport)
4748 {
4749 	const struct sockaddr_in *sin1, *sin2;
4750 	const struct sockaddr_in6 *sin61, *sin62;
4751 	char s1[IPSEC_ADDRSTRLEN], s2[IPSEC_ADDRSTRLEN];
4752 
4753 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len) {
4754 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4755 		    "fam/len fail %d != %d || %d != %d\n",
4756 			sa1->sa_family, sa2->sa_family, sa1->sa_len,
4757 			sa2->sa_len);
4758 		return 0;
4759 	}
4760 
4761 	switch (sa1->sa_family) {
4762 	case AF_INET:
4763 		if (sa1->sa_len != sizeof(struct sockaddr_in)) {
4764 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4765 			    "len fail %d != %zu\n",
4766 			    sa1->sa_len, sizeof(struct sockaddr_in));
4767 			return 0;
4768 		}
4769 		sin1 = (const struct sockaddr_in *)sa1;
4770 		sin2 = (const struct sockaddr_in *)sa2;
4771 		if (sin1->sin_addr.s_addr != sin2->sin_addr.s_addr) {
4772 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4773 			    "addr fail %s != %s\n",
4774 			    (in_print(s1, sizeof(s1), &sin1->sin_addr), s1),
4775 			    (in_print(s2, sizeof(s2), &sin2->sin_addr), s2));
4776 			return 0;
4777 		}
4778 		if (key_portcomp(sin1->sin_port, sin2->sin_port, howport)) {
4779 			return 0;
4780 		}
4781 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4782 		    "addr success %s[%d] == %s[%d]\n",
4783 		    (in_print(s1, sizeof(s1), &sin1->sin_addr), s1),
4784 		    sin1->sin_port,
4785 		    (in_print(s2, sizeof(s2), &sin2->sin_addr), s2),
4786 		    sin2->sin_port);
4787 		break;
4788 	case AF_INET6:
4789 		sin61 = (const struct sockaddr_in6 *)sa1;
4790 		sin62 = (const struct sockaddr_in6 *)sa2;
4791 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4792 			return 0;	/*EINVAL*/
4793 
4794 		if (sin61->sin6_scope_id != sin62->sin6_scope_id) {
4795 			return 0;
4796 		}
4797 		if (!IN6_ARE_ADDR_EQUAL(&sin61->sin6_addr, &sin62->sin6_addr)) {
4798 			return 0;
4799 		}
4800 		if (key_portcomp(sin61->sin6_port, sin62->sin6_port, howport)) {
4801 			return 0;
4802 		}
4803 		break;
4804 	default:
4805 		if (memcmp(sa1, sa2, sa1->sa_len) != 0)
4806 			return 0;
4807 		break;
4808 	}
4809 
4810 	return 1;
4811 }
4812 
4813 /*
4814  * compare two buffers with mask.
4815  * IN:
4816  *	addr1: source
4817  *	addr2: object
4818  *	bits:  Number of bits to compare
4819  * OUT:
4820  *	1 : equal
4821  *	0 : not equal
4822  */
4823 static int
4824 key_bb_match_withmask(const void *a1, const void *a2, u_int bits)
4825 {
4826 	const unsigned char *p1 = a1;
4827 	const unsigned char *p2 = a2;
4828 
4829 	/* XXX: This could be considerably faster if we compare a word
4830 	 * at a time, but it is complicated on LSB Endian machines */
4831 
4832 	/* Handle null pointers */
4833 	if (p1 == NULL || p2 == NULL)
4834 		return (p1 == p2);
4835 
4836 	while (bits >= 8) {
4837 		if (*p1++ != *p2++)
4838 			return 0;
4839 		bits -= 8;
4840 	}
4841 
4842 	if (bits > 0) {
4843 		u_int8_t mask = ~((1<<(8-bits))-1);
4844 		if ((*p1 & mask) != (*p2 & mask))
4845 			return 0;
4846 	}
4847 	return 1;	/* Match! */
4848 }
4849 
4850 static void
4851 key_timehandler_spd(time_t now)
4852 {
4853 	u_int dir;
4854 	struct secpolicy *sp;
4855 
4856 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4857 	    retry:
4858 		mutex_enter(&key_spd.lock);
4859 		SPLIST_WRITER_FOREACH(sp, dir) {
4860 			KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
4861 
4862 			if (sp->lifetime == 0 && sp->validtime == 0)
4863 				continue;
4864 
4865 			if ((sp->lifetime && now - sp->created > sp->lifetime) ||
4866 			    (sp->validtime && now - sp->lastused > sp->validtime)) {
4867 				key_unlink_sp(sp);
4868 				mutex_exit(&key_spd.lock);
4869 				key_spdexpire(sp);
4870 				key_destroy_sp(sp);
4871 				goto retry;
4872 			}
4873 		}
4874 		mutex_exit(&key_spd.lock);
4875 	}
4876 
4877     retry_socksplist:
4878 	mutex_enter(&key_spd.lock);
4879 	SOCKSPLIST_WRITER_FOREACH(sp) {
4880 		if (sp->state != IPSEC_SPSTATE_DEAD)
4881 			continue;
4882 
4883 		key_unlink_sp(sp);
4884 		mutex_exit(&key_spd.lock);
4885 		key_destroy_sp(sp);
4886 		goto retry_socksplist;
4887 	}
4888 	mutex_exit(&key_spd.lock);
4889 }
4890 
4891 static void
4892 key_timehandler_sad(time_t now)
4893 {
4894 	struct secashead *sah;
4895 	int s;
4896 
4897 restart:
4898 	mutex_enter(&key_sad.lock);
4899 	SAHLIST_WRITER_FOREACH(sah) {
4900 		/* If sah has been dead and has no sav, then delete it */
4901 		if (sah->state == SADB_SASTATE_DEAD &&
4902 		    !key_sah_has_sav(sah)) {
4903 			key_unlink_sah(sah);
4904 			mutex_exit(&key_sad.lock);
4905 			key_destroy_sah(sah);
4906 			goto restart;
4907 		}
4908 	}
4909 	mutex_exit(&key_sad.lock);
4910 
4911 	s = pserialize_read_enter();
4912 	SAHLIST_READER_FOREACH(sah) {
4913 		struct secasvar *sav;
4914 
4915 		key_sah_ref(sah);
4916 		pserialize_read_exit(s);
4917 
4918 		/* if LARVAL entry doesn't become MATURE, delete it. */
4919 		mutex_enter(&key_sad.lock);
4920 	restart_sav_LARVAL:
4921 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_LARVAL) {
4922 			if (now - sav->created > key_larval_lifetime) {
4923 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
4924 				goto restart_sav_LARVAL;
4925 			}
4926 		}
4927 		mutex_exit(&key_sad.lock);
4928 
4929 		/*
4930 		 * check MATURE entry to start to send expire message
4931 		 * whether or not.
4932 		 */
4933 	restart_sav_MATURE:
4934 		mutex_enter(&key_sad.lock);
4935 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_MATURE) {
4936 			/* we don't need to check. */
4937 			if (sav->lft_s == NULL)
4938 				continue;
4939 
4940 			/* sanity check */
4941 			KASSERT(sav->lft_c != NULL);
4942 
4943 			/* check SOFT lifetime */
4944 			if (sav->lft_s->sadb_lifetime_addtime != 0 &&
4945 			    now - sav->created > sav->lft_s->sadb_lifetime_addtime) {
4946 				/*
4947 				 * check SA to be used whether or not.
4948 				 * when SA hasn't been used, delete it.
4949 				 */
4950 				if (sav->lft_c->sadb_lifetime_usetime == 0) {
4951 					key_sa_chgstate(sav, SADB_SASTATE_DEAD);
4952 					mutex_exit(&key_sad.lock);
4953 				} else {
4954 					key_sa_chgstate(sav, SADB_SASTATE_DYING);
4955 					mutex_exit(&key_sad.lock);
4956 					/*
4957 					 * XXX If we keep to send expire
4958 					 * message in the status of
4959 					 * DYING. Do remove below code.
4960 					 */
4961 					key_expire(sav);
4962 				}
4963 				goto restart_sav_MATURE;
4964 			}
4965 			/* check SOFT lifetime by bytes */
4966 			/*
4967 			 * XXX I don't know the way to delete this SA
4968 			 * when new SA is installed.  Caution when it's
4969 			 * installed too big lifetime by time.
4970 			 */
4971 			else {
4972 				uint64_t lft_c_bytes = 0;
4973 				lifetime_counters_t sum = {0};
4974 
4975 				percpu_foreach(sav->lft_c_counters_percpu,
4976 				    key_sum_lifetime_counters, sum);
4977 				lft_c_bytes = sum[LIFETIME_COUNTER_BYTES];
4978 
4979 				if (sav->lft_s->sadb_lifetime_bytes == 0 ||
4980 				    sav->lft_s->sadb_lifetime_bytes >= lft_c_bytes)
4981 					continue;
4982 
4983 				key_sa_chgstate(sav, SADB_SASTATE_DYING);
4984 				mutex_exit(&key_sad.lock);
4985 				/*
4986 				 * XXX If we keep to send expire
4987 				 * message in the status of
4988 				 * DYING. Do remove below code.
4989 				 */
4990 				key_expire(sav);
4991 				goto restart_sav_MATURE;
4992 			}
4993 		}
4994 		mutex_exit(&key_sad.lock);
4995 
4996 		/* check DYING entry to change status to DEAD. */
4997 		mutex_enter(&key_sad.lock);
4998 	restart_sav_DYING:
4999 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_DYING) {
5000 			/* we don't need to check. */
5001 			if (sav->lft_h == NULL)
5002 				continue;
5003 
5004 			/* sanity check */
5005 			KASSERT(sav->lft_c != NULL);
5006 
5007 			if (sav->lft_h->sadb_lifetime_addtime != 0 &&
5008 			    now - sav->created > sav->lft_h->sadb_lifetime_addtime) {
5009 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
5010 				goto restart_sav_DYING;
5011 			}
5012 #if 0	/* XXX Should we keep to send expire message until HARD lifetime ? */
5013 			else if (sav->lft_s != NULL
5014 			      && sav->lft_s->sadb_lifetime_addtime != 0
5015 			      && now - sav->created > sav->lft_s->sadb_lifetime_addtime) {
5016 				/*
5017 				 * XXX: should be checked to be
5018 				 * installed the valid SA.
5019 				 */
5020 
5021 				/*
5022 				 * If there is no SA then sending
5023 				 * expire message.
5024 				 */
5025 				key_expire(sav);
5026 			}
5027 #endif
5028 			/* check HARD lifetime by bytes */
5029 			else {
5030 				uint64_t lft_c_bytes = 0;
5031 				lifetime_counters_t sum = {0};
5032 
5033 				percpu_foreach(sav->lft_c_counters_percpu,
5034 				    key_sum_lifetime_counters, sum);
5035 				lft_c_bytes = sum[LIFETIME_COUNTER_BYTES];
5036 
5037 				if (sav->lft_h->sadb_lifetime_bytes == 0 ||
5038 				    sav->lft_h->sadb_lifetime_bytes >= lft_c_bytes)
5039 					continue;
5040 
5041 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
5042 				goto restart_sav_DYING;
5043 			}
5044 		}
5045 		mutex_exit(&key_sad.lock);
5046 
5047 		/* delete entry in DEAD */
5048 	restart_sav_DEAD:
5049 		mutex_enter(&key_sad.lock);
5050 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_DEAD) {
5051 			key_unlink_sav(sav);
5052 			mutex_exit(&key_sad.lock);
5053 			key_destroy_sav(sav);
5054 			goto restart_sav_DEAD;
5055 		}
5056 		mutex_exit(&key_sad.lock);
5057 
5058 		s = pserialize_read_enter();
5059 		key_sah_unref(sah);
5060 	}
5061 	pserialize_read_exit(s);
5062 }
5063 
5064 static void
5065 key_timehandler_acq(time_t now)
5066 {
5067 #ifndef IPSEC_NONBLOCK_ACQUIRE
5068 	struct secacq *acq, *nextacq;
5069 
5070     restart:
5071 	mutex_enter(&key_misc.lock);
5072 	LIST_FOREACH_SAFE(acq, &key_misc.acqlist, chain, nextacq) {
5073 		if (now - acq->created > key_blockacq_lifetime) {
5074 			LIST_REMOVE(acq, chain);
5075 			mutex_exit(&key_misc.lock);
5076 			kmem_free(acq, sizeof(*acq));
5077 			goto restart;
5078 		}
5079 	}
5080 	mutex_exit(&key_misc.lock);
5081 #endif
5082 }
5083 
5084 static void
5085 key_timehandler_spacq(time_t now)
5086 {
5087 #ifdef notyet
5088 	struct secspacq *acq, *nextacq;
5089 
5090 	LIST_FOREACH_SAFE(acq, &key_misc.spacqlist, chain, nextacq) {
5091 		if (now - acq->created > key_blockacq_lifetime) {
5092 			KASSERT(__LIST_CHAINED(acq));
5093 			LIST_REMOVE(acq, chain);
5094 			kmem_free(acq, sizeof(*acq));
5095 		}
5096 	}
5097 #endif
5098 }
5099 
5100 static unsigned int key_timehandler_work_enqueued = 0;
5101 
5102 /*
5103  * time handler.
5104  * scanning SPD and SAD to check status for each entries,
5105  * and do to remove or to expire.
5106  */
5107 static void
5108 key_timehandler_work(struct work *wk, void *arg)
5109 {
5110 	time_t now = time_uptime;
5111 
5112 	/* We can allow enqueuing another work at this point */
5113 	atomic_swap_uint(&key_timehandler_work_enqueued, 0);
5114 
5115 	key_timehandler_spd(now);
5116 	key_timehandler_sad(now);
5117 	key_timehandler_acq(now);
5118 	key_timehandler_spacq(now);
5119 
5120 	key_acquire_sendup_pending_mbuf();
5121 
5122 	/* do exchange to tick time !! */
5123 	callout_reset(&key_timehandler_ch, hz, key_timehandler, NULL);
5124 
5125 	return;
5126 }
5127 
5128 static void
5129 key_timehandler(void *arg)
5130 {
5131 
5132 	/* Avoid enqueuing another work when one is already enqueued */
5133 	if (atomic_swap_uint(&key_timehandler_work_enqueued, 1) == 1)
5134 		return;
5135 
5136 	workqueue_enqueue(key_timehandler_wq, &key_timehandler_wk, NULL);
5137 }
5138 
5139 u_long
5140 key_random(void)
5141 {
5142 	u_long value;
5143 
5144 	key_randomfill(&value, sizeof(value));
5145 	return value;
5146 }
5147 
5148 void
5149 key_randomfill(void *p, size_t l)
5150 {
5151 
5152 	cprng_fast(p, l);
5153 }
5154 
5155 /*
5156  * map SADB_SATYPE_* to IPPROTO_*.
5157  * if satype == SADB_SATYPE then satype is mapped to ~0.
5158  * OUT:
5159  *	0: invalid satype.
5160  */
5161 static u_int16_t
5162 key_satype2proto(u_int8_t satype)
5163 {
5164 	switch (satype) {
5165 	case SADB_SATYPE_UNSPEC:
5166 		return IPSEC_PROTO_ANY;
5167 	case SADB_SATYPE_AH:
5168 		return IPPROTO_AH;
5169 	case SADB_SATYPE_ESP:
5170 		return IPPROTO_ESP;
5171 	case SADB_X_SATYPE_IPCOMP:
5172 		return IPPROTO_IPCOMP;
5173 	case SADB_X_SATYPE_TCPSIGNATURE:
5174 		return IPPROTO_TCP;
5175 	default:
5176 		return 0;
5177 	}
5178 	/* NOTREACHED */
5179 }
5180 
5181 /*
5182  * map IPPROTO_* to SADB_SATYPE_*
5183  * OUT:
5184  *	0: invalid protocol type.
5185  */
5186 static u_int8_t
5187 key_proto2satype(u_int16_t proto)
5188 {
5189 	switch (proto) {
5190 	case IPPROTO_AH:
5191 		return SADB_SATYPE_AH;
5192 	case IPPROTO_ESP:
5193 		return SADB_SATYPE_ESP;
5194 	case IPPROTO_IPCOMP:
5195 		return SADB_X_SATYPE_IPCOMP;
5196 	case IPPROTO_TCP:
5197 		return SADB_X_SATYPE_TCPSIGNATURE;
5198 	default:
5199 		return 0;
5200 	}
5201 	/* NOTREACHED */
5202 }
5203 
5204 static int
5205 key_setsecasidx(int proto, int mode, int reqid,
5206     const struct sockaddr *src, const struct sockaddr *dst,
5207     struct secasindex * saidx)
5208 {
5209 	const union sockaddr_union *src_u = (const union sockaddr_union *)src;
5210 	const union sockaddr_union *dst_u = (const union sockaddr_union *)dst;
5211 
5212 	/* sa len safety check */
5213 	if (key_checksalen(src_u) != 0)
5214 		return -1;
5215 	if (key_checksalen(dst_u) != 0)
5216 		return -1;
5217 
5218 	memset(saidx, 0, sizeof(*saidx));
5219 	saidx->proto = proto;
5220 	saidx->mode = mode;
5221 	saidx->reqid = reqid;
5222 	memcpy(&saidx->src, src_u, src_u->sa.sa_len);
5223 	memcpy(&saidx->dst, dst_u, dst_u->sa.sa_len);
5224 
5225 	key_porttosaddr(&((saidx)->src), 0);
5226 	key_porttosaddr(&((saidx)->dst), 0);
5227 	return 0;
5228 }
5229 
5230 static void
5231 key_init_spidx_bymsghdr(struct secpolicyindex *spidx,
5232     const struct sadb_msghdr *mhp)
5233 {
5234 	const struct sadb_address *src0, *dst0;
5235 	const struct sockaddr *src, *dst;
5236 	const struct sadb_x_policy *xpl0;
5237 
5238 	src0 = mhp->ext[SADB_EXT_ADDRESS_SRC];
5239 	dst0 = mhp->ext[SADB_EXT_ADDRESS_DST];
5240 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5241 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5242 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
5243 
5244 	memset(spidx, 0, sizeof(*spidx));
5245 	spidx->dir = xpl0->sadb_x_policy_dir;
5246 	spidx->prefs = src0->sadb_address_prefixlen;
5247 	spidx->prefd = dst0->sadb_address_prefixlen;
5248 	spidx->ul_proto = src0->sadb_address_proto;
5249 	/* XXX boundary check against sa_len */
5250 	memcpy(&spidx->src, src, src->sa_len);
5251 	memcpy(&spidx->dst, dst, dst->sa_len);
5252 }
5253 
5254 /* %%% PF_KEY */
5255 /*
5256  * SADB_GETSPI processing is to receive
5257  *	<base, (SA2), src address, dst address, (SPI range)>
5258  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
5259  * tree with the status of LARVAL, and send
5260  *	<base, SA(*), address(SD)>
5261  * to the IKMPd.
5262  *
5263  * IN:	mhp: pointer to the pointer to each header.
5264  * OUT:	NULL if fail.
5265  *	other if success, return pointer to the message to send.
5266  */
5267 static int
5268 key_api_getspi(struct socket *so, struct mbuf *m,
5269 	   const struct sadb_msghdr *mhp)
5270 {
5271 	const struct sockaddr *src, *dst;
5272 	struct secasindex saidx;
5273 	struct secashead *sah;
5274 	struct secasvar *newsav;
5275 	u_int8_t proto;
5276 	u_int32_t spi;
5277 	u_int8_t mode;
5278 	u_int16_t reqid;
5279 	int error;
5280 
5281 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
5282 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
5283 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5284 		return key_senderror(so, m, EINVAL);
5285 	}
5286 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
5287 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
5288 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5289 		return key_senderror(so, m, EINVAL);
5290 	}
5291 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
5292 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
5293 		mode = sa2->sadb_x_sa2_mode;
5294 		reqid = sa2->sadb_x_sa2_reqid;
5295 	} else {
5296 		mode = IPSEC_MODE_ANY;
5297 		reqid = 0;
5298 	}
5299 
5300 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5301 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5302 
5303 	/* map satype to proto */
5304 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
5305 	if (proto == 0) {
5306 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
5307 		return key_senderror(so, m, EINVAL);
5308 	}
5309 
5310 
5311 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
5312 	if (error != 0)
5313 		return key_senderror(so, m, EINVAL);
5314 
5315 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
5316 	if (error != 0)
5317 		return key_senderror(so, m, EINVAL);
5318 
5319 	/* SPI allocation */
5320 	spi = key_do_getnewspi(mhp->ext[SADB_EXT_SPIRANGE], &saidx);
5321 	if (spi == 0)
5322 		return key_senderror(so, m, EINVAL);
5323 
5324 	/* get a SA index */
5325 	sah = key_getsah_ref(&saidx, CMP_REQID);
5326 	if (sah == NULL) {
5327 		/* create a new SA index */
5328 		sah = key_newsah(&saidx);
5329 		if (sah == NULL) {
5330 			IPSECLOG(LOG_DEBUG, "No more memory.\n");
5331 			return key_senderror(so, m, ENOBUFS);
5332 		}
5333 	}
5334 
5335 	/* get a new SA */
5336 	/* XXX rewrite */
5337 	newsav = KEY_NEWSAV(m, mhp, &error);
5338 	if (newsav == NULL) {
5339 		key_sah_unref(sah);
5340 		/* XXX don't free new SA index allocated in above. */
5341 		return key_senderror(so, m, error);
5342 	}
5343 
5344 	/* set spi */
5345 	newsav->spi = htonl(spi);
5346 
5347 	/* Add to sah#savlist */
5348 	key_init_sav(newsav);
5349 	newsav->sah = sah;
5350 	newsav->state = SADB_SASTATE_LARVAL;
5351 	mutex_enter(&key_sad.lock);
5352 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_LARVAL, newsav);
5353 	mutex_exit(&key_sad.lock);
5354 	key_validate_savlist(sah, SADB_SASTATE_LARVAL);
5355 
5356 	key_sah_unref(sah);
5357 
5358 #ifndef IPSEC_NONBLOCK_ACQUIRE
5359 	/* delete the entry in key_misc.acqlist */
5360 	if (mhp->msg->sadb_msg_seq != 0) {
5361 		struct secacq *acq;
5362 		mutex_enter(&key_misc.lock);
5363 		acq = key_getacqbyseq(mhp->msg->sadb_msg_seq);
5364 		if (acq != NULL) {
5365 			/* reset counter in order to deletion by timehandler. */
5366 			acq->created = time_uptime;
5367 			acq->count = 0;
5368 		}
5369 		mutex_exit(&key_misc.lock);
5370 	}
5371 #endif
5372 
5373     {
5374 	struct mbuf *n, *nn;
5375 	struct sadb_sa *m_sa;
5376 	int off, len;
5377 
5378 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
5379 	    PFKEY_ALIGN8(sizeof(struct sadb_sa)) <= MCLBYTES);
5380 
5381 	/* create new sadb_msg to reply. */
5382 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
5383 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
5384 
5385 	n = key_alloc_mbuf_simple(len, M_WAITOK);
5386 	n->m_len = len;
5387 	n->m_next = NULL;
5388 	off = 0;
5389 
5390 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
5391 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
5392 
5393 	m_sa = (struct sadb_sa *)(mtod(n, char *) + off);
5394 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
5395 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
5396 	m_sa->sadb_sa_spi = htonl(spi);
5397 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
5398 
5399 	KASSERTMSG(off == len, "length inconsistency");
5400 
5401 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
5402 	    SADB_EXT_ADDRESS_DST);
5403 
5404 	KASSERT(n->m_len >= sizeof(struct sadb_msg));
5405 
5406 	n->m_pkthdr.len = 0;
5407 	for (nn = n; nn; nn = nn->m_next)
5408 		n->m_pkthdr.len += nn->m_len;
5409 
5410 	key_fill_replymsg(n, newsav->seq);
5411 	m_freem(m);
5412 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
5413     }
5414 }
5415 
5416 /*
5417  * allocating new SPI
5418  * called by key_api_getspi().
5419  * OUT:
5420  *	0:	failure.
5421  *	others: success.
5422  */
5423 static u_int32_t
5424 key_do_getnewspi(const struct sadb_spirange *spirange,
5425 		 const struct secasindex *saidx)
5426 {
5427 	u_int32_t newspi;
5428 	u_int32_t spmin, spmax;
5429 	int count = key_spi_trycnt;
5430 
5431 	/* set spi range to allocate */
5432 	if (spirange != NULL) {
5433 		spmin = spirange->sadb_spirange_min;
5434 		spmax = spirange->sadb_spirange_max;
5435 	} else {
5436 		spmin = key_spi_minval;
5437 		spmax = key_spi_maxval;
5438 	}
5439 	/* IPCOMP needs 2-byte SPI */
5440 	if (saidx->proto == IPPROTO_IPCOMP) {
5441 		u_int32_t t;
5442 		if (spmin >= 0x10000)
5443 			spmin = 0xffff;
5444 		if (spmax >= 0x10000)
5445 			spmax = 0xffff;
5446 		if (spmin > spmax) {
5447 			t = spmin; spmin = spmax; spmax = t;
5448 		}
5449 	}
5450 
5451 	if (spmin == spmax) {
5452 		if (key_checkspidup(saidx, htonl(spmin))) {
5453 			IPSECLOG(LOG_DEBUG, "SPI %u exists already.\n", spmin);
5454 			return 0;
5455 		}
5456 
5457 		count--; /* taking one cost. */
5458 		newspi = spmin;
5459 
5460 	} else {
5461 
5462 		/* init SPI */
5463 		newspi = 0;
5464 
5465 		/* when requesting to allocate spi ranged */
5466 		while (count--) {
5467 			/* generate pseudo-random SPI value ranged. */
5468 			newspi = spmin + (key_random() % (spmax - spmin + 1));
5469 
5470 			if (!key_checkspidup(saidx, htonl(newspi)))
5471 				break;
5472 		}
5473 
5474 		if (count == 0 || newspi == 0) {
5475 			IPSECLOG(LOG_DEBUG, "to allocate spi is failed.\n");
5476 			return 0;
5477 		}
5478 	}
5479 
5480 	/* statistics */
5481 	keystat.getspi_count =
5482 	    (keystat.getspi_count + key_spi_trycnt - count) / 2;
5483 
5484 	return newspi;
5485 }
5486 
5487 static int
5488 key_handle_natt_info(struct secasvar *sav,
5489       		     const struct sadb_msghdr *mhp)
5490 {
5491 	const char *msg = "?" ;
5492 	struct sadb_x_nat_t_type *type;
5493 	struct sadb_x_nat_t_port *sport, *dport;
5494 	struct sadb_address *iaddr, *raddr;
5495 	struct sadb_x_nat_t_frag *frag;
5496 
5497 	if (mhp->ext[SADB_X_EXT_NAT_T_TYPE] == NULL ||
5498 	    mhp->ext[SADB_X_EXT_NAT_T_SPORT] == NULL ||
5499 	    mhp->ext[SADB_X_EXT_NAT_T_DPORT] == NULL)
5500 		return 0;
5501 
5502 	if (mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) {
5503 		msg = "TYPE";
5504 		goto bad;
5505 	}
5506 
5507 	if (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) {
5508 		msg = "SPORT";
5509 		goto bad;
5510 	}
5511 
5512 	if (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport)) {
5513 		msg = "DPORT";
5514 		goto bad;
5515 	}
5516 
5517 	if (mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL) {
5518 		IPSECLOG(LOG_DEBUG, "NAT-T OAi present\n");
5519 		if (mhp->extlen[SADB_X_EXT_NAT_T_OAI] < sizeof(*iaddr)) {
5520 			msg = "OAI";
5521 			goto bad;
5522 		}
5523 	}
5524 
5525 	if (mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL) {
5526 		IPSECLOG(LOG_DEBUG, "NAT-T OAr present\n");
5527 		if (mhp->extlen[SADB_X_EXT_NAT_T_OAR] < sizeof(*raddr)) {
5528 			msg = "OAR";
5529 			goto bad;
5530 		}
5531 	}
5532 
5533 	if (mhp->ext[SADB_X_EXT_NAT_T_FRAG] != NULL) {
5534 	    if (mhp->extlen[SADB_X_EXT_NAT_T_FRAG] < sizeof(*frag)) {
5535 		    msg = "FRAG";
5536 		    goto bad;
5537 	    }
5538 	}
5539 
5540 	type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5541 	sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5542 	dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5543 	iaddr = mhp->ext[SADB_X_EXT_NAT_T_OAI];
5544 	raddr = mhp->ext[SADB_X_EXT_NAT_T_OAR];
5545 	frag = mhp->ext[SADB_X_EXT_NAT_T_FRAG];
5546 
5547 	IPSECLOG(LOG_DEBUG, "type %d, sport = %d, dport = %d\n",
5548 	    type->sadb_x_nat_t_type_type,
5549 	    ntohs(sport->sadb_x_nat_t_port_port),
5550 	    ntohs(dport->sadb_x_nat_t_port_port));
5551 
5552 	sav->natt_type = type->sadb_x_nat_t_type_type;
5553 	key_porttosaddr(&sav->sah->saidx.src, sport->sadb_x_nat_t_port_port);
5554 	key_porttosaddr(&sav->sah->saidx.dst, dport->sadb_x_nat_t_port_port);
5555 	if (frag)
5556 		sav->esp_frag = frag->sadb_x_nat_t_frag_fraglen;
5557 	else
5558 		sav->esp_frag = IP_MAXPACKET;
5559 
5560 	return 0;
5561 bad:
5562 	IPSECLOG(LOG_DEBUG, "invalid message %s\n", msg);
5563 	__USE(msg);
5564 	return -1;
5565 }
5566 
5567 /* Just update the IPSEC_NAT_T ports if present */
5568 static int
5569 key_set_natt_ports(union sockaddr_union *src, union sockaddr_union *dst,
5570       		     const struct sadb_msghdr *mhp)
5571 {
5572 	if (mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL)
5573 		IPSECLOG(LOG_DEBUG, "NAT-T OAi present\n");
5574 	if (mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL)
5575 		IPSECLOG(LOG_DEBUG, "NAT-T OAr present\n");
5576 
5577 	if ((mhp->ext[SADB_X_EXT_NAT_T_TYPE] != NULL) &&
5578 	    (mhp->ext[SADB_X_EXT_NAT_T_SPORT] != NULL) &&
5579 	    (mhp->ext[SADB_X_EXT_NAT_T_DPORT] != NULL)) {
5580 		struct sadb_x_nat_t_type *type;
5581 		struct sadb_x_nat_t_port *sport;
5582 		struct sadb_x_nat_t_port *dport;
5583 
5584 		if ((mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) ||
5585 		    (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) ||
5586 		    (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport))) {
5587 			IPSECLOG(LOG_DEBUG, "invalid message\n");
5588 			return -1;
5589 		}
5590 
5591 		type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5592 		sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5593 		dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5594 
5595 		key_porttosaddr(src, sport->sadb_x_nat_t_port_port);
5596 		key_porttosaddr(dst, dport->sadb_x_nat_t_port_port);
5597 
5598 		IPSECLOG(LOG_DEBUG, "type %d, sport = %d, dport = %d\n",
5599 		    type->sadb_x_nat_t_type_type,
5600 		    ntohs(sport->sadb_x_nat_t_port_port),
5601 		    ntohs(dport->sadb_x_nat_t_port_port));
5602 	}
5603 
5604 	return 0;
5605 }
5606 
5607 
5608 /*
5609  * SADB_UPDATE processing
5610  * receive
5611  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5612  *       key(AE), (identity(SD),) (sensitivity)>
5613  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5614  * and send
5615  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5616  *       (identity(SD),) (sensitivity)>
5617  * to the ikmpd.
5618  *
5619  * m will always be freed.
5620  */
5621 static int
5622 key_api_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5623 {
5624 	struct sadb_sa *sa0;
5625 	const struct sockaddr *src, *dst;
5626 	struct secasindex saidx;
5627 	struct secashead *sah;
5628 	struct secasvar *sav, *newsav;
5629 	u_int16_t proto;
5630 	u_int8_t mode;
5631 	u_int16_t reqid;
5632 	int error;
5633 
5634 	/* map satype to proto */
5635 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
5636 	if (proto == 0) {
5637 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
5638 		return key_senderror(so, m, EINVAL);
5639 	}
5640 
5641 	if (mhp->ext[SADB_EXT_SA] == NULL ||
5642 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
5643 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
5644 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5645 	     mhp->ext[SADB_EXT_KEY_ENCRYPT] == NULL) ||
5646 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5647 	     mhp->ext[SADB_EXT_KEY_AUTH] == NULL) ||
5648 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL &&
5649 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] == NULL) ||
5650 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] == NULL &&
5651 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] != NULL)) {
5652 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5653 		return key_senderror(so, m, EINVAL);
5654 	}
5655 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
5656 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
5657 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
5658 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5659 		return key_senderror(so, m, EINVAL);
5660 	}
5661 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
5662 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
5663 		mode = sa2->sadb_x_sa2_mode;
5664 		reqid = sa2->sadb_x_sa2_reqid;
5665 	} else {
5666 		mode = IPSEC_MODE_ANY;
5667 		reqid = 0;
5668 	}
5669 	/* XXX boundary checking for other extensions */
5670 
5671 	sa0 = mhp->ext[SADB_EXT_SA];
5672 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5673 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5674 
5675 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
5676 	if (error != 0)
5677 		return key_senderror(so, m, EINVAL);
5678 
5679 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
5680 	if (error != 0)
5681 		return key_senderror(so, m, EINVAL);
5682 
5683 	/* get a SA header */
5684 	sah = key_getsah_ref(&saidx, CMP_REQID);
5685 	if (sah == NULL) {
5686 		IPSECLOG(LOG_DEBUG, "no SA index found.\n");
5687 		return key_senderror(so, m, ENOENT);
5688 	}
5689 
5690 	/* set spidx if there */
5691 	/* XXX rewrite */
5692 	error = key_setident(sah, m, mhp);
5693 	if (error)
5694 		goto error_sah;
5695 
5696 	/* find a SA with sequence number. */
5697 #ifdef IPSEC_DOSEQCHECK
5698 	if (mhp->msg->sadb_msg_seq != 0) {
5699 		sav = key_getsavbyseq(sah, mhp->msg->sadb_msg_seq);
5700 		if (sav == NULL) {
5701 			IPSECLOG(LOG_DEBUG,
5702 			    "no larval SA with sequence %u exists.\n",
5703 			    mhp->msg->sadb_msg_seq);
5704 			error = ENOENT;
5705 			goto error_sah;
5706 		}
5707 	}
5708 #else
5709 	sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
5710 	if (sav == NULL) {
5711 		IPSECLOG(LOG_DEBUG, "no such a SA found (spi:%u)\n",
5712 		    (u_int32_t)ntohl(sa0->sadb_sa_spi));
5713 		error = EINVAL;
5714 		goto error_sah;
5715 	}
5716 #endif
5717 
5718 	/* validity check */
5719 	if (sav->sah->saidx.proto != proto) {
5720 		IPSECLOG(LOG_DEBUG, "protocol mismatched (DB=%u param=%u)\n",
5721 		    sav->sah->saidx.proto, proto);
5722 		error = EINVAL;
5723 		goto error;
5724 	}
5725 #ifdef IPSEC_DOSEQCHECK
5726 	if (sav->spi != sa0->sadb_sa_spi) {
5727 		IPSECLOG(LOG_DEBUG, "SPI mismatched (DB:%u param:%u)\n",
5728 		    (u_int32_t)ntohl(sav->spi),
5729 		    (u_int32_t)ntohl(sa0->sadb_sa_spi));
5730 		error = EINVAL;
5731 		goto error;
5732 	}
5733 #endif
5734 	if (sav->pid != mhp->msg->sadb_msg_pid) {
5735 		IPSECLOG(LOG_DEBUG, "pid mismatched (DB:%u param:%u)\n",
5736 		    sav->pid, mhp->msg->sadb_msg_pid);
5737 		error = EINVAL;
5738 		goto error;
5739 	}
5740 
5741 	/*
5742 	 * Allocate a new SA instead of modifying the existing SA directly
5743 	 * to avoid race conditions.
5744 	 */
5745 	newsav = kmem_zalloc(sizeof(struct secasvar), KM_SLEEP);
5746 
5747 	/* copy sav values */
5748 	newsav->spi = sav->spi;
5749 	newsav->seq = sav->seq;
5750 	newsav->created = sav->created;
5751 	newsav->pid = sav->pid;
5752 	newsav->sah = sav->sah;
5753 
5754 	error = key_setsaval(newsav, m, mhp);
5755 	if (error) {
5756 		key_delsav(newsav);
5757 		goto error;
5758 	}
5759 
5760 	error = key_handle_natt_info(newsav, mhp);
5761 	if (error != 0) {
5762 		key_delsav(newsav);
5763 		goto error;
5764 	}
5765 
5766 	error = key_init_xform(newsav);
5767 	if (error != 0) {
5768 		key_delsav(newsav);
5769 		goto error;
5770 	}
5771 
5772 	/* Add to sah#savlist */
5773 	key_init_sav(newsav);
5774 	newsav->state = SADB_SASTATE_MATURE;
5775 	mutex_enter(&key_sad.lock);
5776 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_MATURE, newsav);
5777 	SAVLUT_WRITER_INSERT_HEAD(newsav);
5778 	mutex_exit(&key_sad.lock);
5779 	key_validate_savlist(sah, SADB_SASTATE_MATURE);
5780 
5781 	key_sah_unref(sah);
5782 	sah = NULL;
5783 
5784 	key_destroy_sav_with_ref(sav);
5785 	sav = NULL;
5786 
5787     {
5788 	struct mbuf *n;
5789 
5790 	/* set msg buf from mhp */
5791 	n = key_getmsgbuf_x1(m, mhp);
5792 	if (n == NULL) {
5793 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
5794 		return key_senderror(so, m, ENOBUFS);
5795 	}
5796 
5797 	m_freem(m);
5798 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5799     }
5800 error:
5801 	KEY_SA_UNREF(&sav);
5802 error_sah:
5803 	key_sah_unref(sah);
5804 	return key_senderror(so, m, error);
5805 }
5806 
5807 /*
5808  * search SAD with sequence for a SA which state is SADB_SASTATE_LARVAL.
5809  * only called by key_api_update().
5810  * OUT:
5811  *	NULL	: not found
5812  *	others	: found, pointer to a SA.
5813  */
5814 #ifdef IPSEC_DOSEQCHECK
5815 static struct secasvar *
5816 key_getsavbyseq(struct secashead *sah, u_int32_t seq)
5817 {
5818 	struct secasvar *sav;
5819 	u_int state;
5820 	int s;
5821 
5822 	state = SADB_SASTATE_LARVAL;
5823 
5824 	/* search SAD with sequence number ? */
5825 	s = pserialize_read_enter();
5826 	SAVLIST_READER_FOREACH(sav, sah, state) {
5827 		KEY_CHKSASTATE(state, sav->state);
5828 
5829 		if (sav->seq == seq) {
5830 			SA_ADDREF(sav);
5831 			KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
5832 			    "DP cause refcnt++:%d SA:%p\n",
5833 			    key_sa_refcnt(sav), sav);
5834 			break;
5835 		}
5836 	}
5837 	pserialize_read_exit(s);
5838 
5839 	return sav;
5840 }
5841 #endif
5842 
5843 /*
5844  * SADB_ADD processing
5845  * add an entry to SA database, when received
5846  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5847  *       key(AE), (identity(SD),) (sensitivity)>
5848  * from the ikmpd,
5849  * and send
5850  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5851  *       (identity(SD),) (sensitivity)>
5852  * to the ikmpd.
5853  *
5854  * IGNORE identity and sensitivity messages.
5855  *
5856  * m will always be freed.
5857  */
5858 static int
5859 key_api_add(struct socket *so, struct mbuf *m,
5860 	const struct sadb_msghdr *mhp)
5861 {
5862 	struct sadb_sa *sa0;
5863 	const struct sockaddr *src, *dst;
5864 	struct secasindex saidx;
5865 	struct secashead *sah;
5866 	struct secasvar *newsav;
5867 	u_int16_t proto;
5868 	u_int8_t mode;
5869 	u_int16_t reqid;
5870 	int error;
5871 
5872 	/* map satype to proto */
5873 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
5874 	if (proto == 0) {
5875 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
5876 		return key_senderror(so, m, EINVAL);
5877 	}
5878 
5879 	if (mhp->ext[SADB_EXT_SA] == NULL ||
5880 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
5881 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
5882 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5883 	     mhp->ext[SADB_EXT_KEY_ENCRYPT] == NULL) ||
5884 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5885 	     mhp->ext[SADB_EXT_KEY_AUTH] == NULL) ||
5886 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL &&
5887 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] == NULL) ||
5888 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] == NULL &&
5889 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] != NULL)) {
5890 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5891 		return key_senderror(so, m, EINVAL);
5892 	}
5893 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
5894 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
5895 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
5896 		/* XXX need more */
5897 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5898 		return key_senderror(so, m, EINVAL);
5899 	}
5900 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
5901 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
5902 		mode = sa2->sadb_x_sa2_mode;
5903 		reqid = sa2->sadb_x_sa2_reqid;
5904 	} else {
5905 		mode = IPSEC_MODE_ANY;
5906 		reqid = 0;
5907 	}
5908 
5909 	sa0 = mhp->ext[SADB_EXT_SA];
5910 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5911 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5912 
5913 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
5914 	if (error != 0)
5915 		return key_senderror(so, m, EINVAL);
5916 
5917 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
5918 	if (error != 0)
5919 		return key_senderror(so, m, EINVAL);
5920 
5921 	/* get a SA header */
5922 	sah = key_getsah_ref(&saidx, CMP_REQID);
5923 	if (sah == NULL) {
5924 		/* create a new SA header */
5925 		sah = key_newsah(&saidx);
5926 		if (sah == NULL) {
5927 			IPSECLOG(LOG_DEBUG, "No more memory.\n");
5928 			return key_senderror(so, m, ENOBUFS);
5929 		}
5930 	}
5931 
5932 	/* set spidx if there */
5933 	/* XXX rewrite */
5934 	error = key_setident(sah, m, mhp);
5935 	if (error)
5936 		goto error;
5937 
5938     {
5939 	struct secasvar *sav;
5940 
5941 	/* We can create new SA only if SPI is differenct. */
5942 	sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
5943 	if (sav != NULL) {
5944 		KEY_SA_UNREF(&sav);
5945 		IPSECLOG(LOG_DEBUG, "SA already exists.\n");
5946 		error = EEXIST;
5947 		goto error;
5948 	}
5949     }
5950 
5951 	/* create new SA entry. */
5952 	newsav = KEY_NEWSAV(m, mhp, &error);
5953 	if (newsav == NULL)
5954 		goto error;
5955 	newsav->sah = sah;
5956 
5957 	error = key_handle_natt_info(newsav, mhp);
5958 	if (error != 0) {
5959 		key_delsav(newsav);
5960 		error = EINVAL;
5961 		goto error;
5962 	}
5963 
5964 	error = key_init_xform(newsav);
5965 	if (error != 0) {
5966 		key_delsav(newsav);
5967 		goto error;
5968 	}
5969 
5970 	/* Add to sah#savlist */
5971 	key_init_sav(newsav);
5972 	newsav->state = SADB_SASTATE_MATURE;
5973 	mutex_enter(&key_sad.lock);
5974 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_MATURE, newsav);
5975 	SAVLUT_WRITER_INSERT_HEAD(newsav);
5976 	mutex_exit(&key_sad.lock);
5977 	key_validate_savlist(sah, SADB_SASTATE_MATURE);
5978 
5979 	key_sah_unref(sah);
5980 	sah = NULL;
5981 
5982 	/*
5983 	 * don't call key_freesav() here, as we would like to keep the SA
5984 	 * in the database on success.
5985 	 */
5986 
5987     {
5988 	struct mbuf *n;
5989 
5990 	/* set msg buf from mhp */
5991 	n = key_getmsgbuf_x1(m, mhp);
5992 	if (n == NULL) {
5993 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
5994 		return key_senderror(so, m, ENOBUFS);
5995 	}
5996 
5997 	m_freem(m);
5998 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5999     }
6000 error:
6001 	key_sah_unref(sah);
6002 	return key_senderror(so, m, error);
6003 }
6004 
6005 /* m is retained */
6006 static int
6007 key_setident(struct secashead *sah, struct mbuf *m,
6008 	     const struct sadb_msghdr *mhp)
6009 {
6010 	const struct sadb_ident *idsrc, *iddst;
6011 	int idsrclen, iddstlen;
6012 
6013 	KASSERT(!cpu_softintr_p());
6014 	KASSERT(sah != NULL);
6015 	KASSERT(m != NULL);
6016 	KASSERT(mhp != NULL);
6017 	KASSERT(mhp->msg != NULL);
6018 
6019 	/*
6020 	 * Can be called with an existing sah from key_api_update().
6021 	 */
6022 	if (sah->idents != NULL) {
6023 		kmem_free(sah->idents, sah->idents_len);
6024 		sah->idents = NULL;
6025 		sah->idents_len = 0;
6026 	}
6027 	if (sah->identd != NULL) {
6028 		kmem_free(sah->identd, sah->identd_len);
6029 		sah->identd = NULL;
6030 		sah->identd_len = 0;
6031 	}
6032 
6033 	/* don't make buffer if not there */
6034 	if (mhp->ext[SADB_EXT_IDENTITY_SRC] == NULL &&
6035 	    mhp->ext[SADB_EXT_IDENTITY_DST] == NULL) {
6036 		sah->idents = NULL;
6037 		sah->identd = NULL;
6038 		return 0;
6039 	}
6040 
6041 	if (mhp->ext[SADB_EXT_IDENTITY_SRC] == NULL ||
6042 	    mhp->ext[SADB_EXT_IDENTITY_DST] == NULL) {
6043 		IPSECLOG(LOG_DEBUG, "invalid identity.\n");
6044 		return EINVAL;
6045 	}
6046 
6047 	idsrc = mhp->ext[SADB_EXT_IDENTITY_SRC];
6048 	iddst = mhp->ext[SADB_EXT_IDENTITY_DST];
6049 	idsrclen = mhp->extlen[SADB_EXT_IDENTITY_SRC];
6050 	iddstlen = mhp->extlen[SADB_EXT_IDENTITY_DST];
6051 
6052 	/* validity check */
6053 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
6054 		IPSECLOG(LOG_DEBUG, "ident type mismatched src %u, dst %u.\n",
6055 		    idsrc->sadb_ident_type, iddst->sadb_ident_type);
6056 		return EINVAL;
6057 	}
6058 
6059 	switch (idsrc->sadb_ident_type) {
6060 	case SADB_IDENTTYPE_PREFIX:
6061 	case SADB_IDENTTYPE_FQDN:
6062 	case SADB_IDENTTYPE_USERFQDN:
6063 	default:
6064 		/* XXX do nothing */
6065 		sah->idents = NULL;
6066 		sah->identd = NULL;
6067 	 	return 0;
6068 	}
6069 
6070 	/* make structure */
6071 	sah->idents = kmem_alloc(idsrclen, KM_SLEEP);
6072 	sah->idents_len = idsrclen;
6073 	sah->identd = kmem_alloc(iddstlen, KM_SLEEP);
6074 	sah->identd_len = iddstlen;
6075 	memcpy(sah->idents, idsrc, idsrclen);
6076 	memcpy(sah->identd, iddst, iddstlen);
6077 
6078 	return 0;
6079 }
6080 
6081 /*
6082  * m will not be freed on return. It never return NULL.
6083  * it is caller's responsibility to free the result.
6084  */
6085 static struct mbuf *
6086 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
6087 {
6088 	struct mbuf *n;
6089 
6090 	KASSERT(m != NULL);
6091 	KASSERT(mhp != NULL);
6092 	KASSERT(mhp->msg != NULL);
6093 
6094 	/* create new sadb_msg to reply. */
6095 	n = key_gather_mbuf(m, mhp, 1, 15, SADB_EXT_RESERVED,
6096 	    SADB_EXT_SA, SADB_X_EXT_SA2,
6097 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
6098 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
6099 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
6100 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
6101 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
6102 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NAT_T_FRAG);
6103 
6104 	KASSERT(n->m_len >= sizeof(struct sadb_msg));
6105 
6106 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
6107 	mtod(n, struct sadb_msg *)->sadb_msg_len =
6108 	    PFKEY_UNIT64(n->m_pkthdr.len);
6109 
6110 	return n;
6111 }
6112 
6113 static int key_delete_all (struct socket *, struct mbuf *,
6114 			   const struct sadb_msghdr *, u_int16_t);
6115 
6116 /*
6117  * SADB_DELETE processing
6118  * receive
6119  *   <base, SA(*), address(SD)>
6120  * from the ikmpd, and set SADB_SASTATE_DEAD,
6121  * and send,
6122  *   <base, SA(*), address(SD)>
6123  * to the ikmpd.
6124  *
6125  * m will always be freed.
6126  */
6127 static int
6128 key_api_delete(struct socket *so, struct mbuf *m,
6129 	   const struct sadb_msghdr *mhp)
6130 {
6131 	struct sadb_sa *sa0;
6132 	const struct sockaddr *src, *dst;
6133 	struct secasindex saidx;
6134 	struct secashead *sah;
6135 	struct secasvar *sav = NULL;
6136 	u_int16_t proto;
6137 	int error;
6138 
6139 	/* map satype to proto */
6140 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
6141 	if (proto == 0) {
6142 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
6143 		return key_senderror(so, m, EINVAL);
6144 	}
6145 
6146 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
6147 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
6148 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6149 		return key_senderror(so, m, EINVAL);
6150 	}
6151 
6152 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
6153 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
6154 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6155 		return key_senderror(so, m, EINVAL);
6156 	}
6157 
6158 	if (mhp->ext[SADB_EXT_SA] == NULL) {
6159 		/*
6160 		 * Caller wants us to delete all non-LARVAL SAs
6161 		 * that match the src/dst.  This is used during
6162 		 * IKE INITIAL-CONTACT.
6163 		 */
6164 		IPSECLOG(LOG_DEBUG, "doing delete all.\n");
6165 		return key_delete_all(so, m, mhp, proto);
6166 	} else if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa)) {
6167 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6168 		return key_senderror(so, m, EINVAL);
6169 	}
6170 
6171 	sa0 = mhp->ext[SADB_EXT_SA];
6172 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6173 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6174 
6175 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
6176 	if (error != 0)
6177 		return key_senderror(so, m, EINVAL);
6178 
6179 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6180 	if (error != 0)
6181 		return key_senderror(so, m, EINVAL);
6182 
6183 	/* get a SA header */
6184 	sah = key_getsah_ref(&saidx, CMP_HEAD);
6185 	if (sah != NULL) {
6186 		/* get a SA with SPI. */
6187 		sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
6188 		key_sah_unref(sah);
6189 	}
6190 
6191 	if (sav == NULL) {
6192 		IPSECLOG(LOG_DEBUG, "no SA found.\n");
6193 		return key_senderror(so, m, ENOENT);
6194 	}
6195 
6196 	key_destroy_sav_with_ref(sav);
6197 	sav = NULL;
6198 
6199     {
6200 	struct mbuf *n;
6201 
6202 	/* create new sadb_msg to reply. */
6203 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
6204 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6205 
6206 	key_fill_replymsg(n, 0);
6207 	m_freem(m);
6208 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6209     }
6210 }
6211 
6212 /*
6213  * delete all SAs for src/dst.  Called from key_api_delete().
6214  */
6215 static int
6216 key_delete_all(struct socket *so, struct mbuf *m,
6217 	       const struct sadb_msghdr *mhp, u_int16_t proto)
6218 {
6219 	const struct sockaddr *src, *dst;
6220 	struct secasindex saidx;
6221 	struct secashead *sah;
6222 	struct secasvar *sav;
6223 	u_int state;
6224 	int error;
6225 
6226 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6227 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6228 
6229 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
6230 	if (error != 0)
6231 		return key_senderror(so, m, EINVAL);
6232 
6233 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6234 	if (error != 0)
6235 		return key_senderror(so, m, EINVAL);
6236 
6237 	sah = key_getsah_ref(&saidx, CMP_HEAD);
6238 	if (sah != NULL) {
6239 		/* Delete all non-LARVAL SAs. */
6240 		SASTATE_ALIVE_FOREACH(state) {
6241 			if (state == SADB_SASTATE_LARVAL)
6242 				continue;
6243 		restart:
6244 			mutex_enter(&key_sad.lock);
6245 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
6246 				sav->state = SADB_SASTATE_DEAD;
6247 				key_unlink_sav(sav);
6248 				mutex_exit(&key_sad.lock);
6249 				key_destroy_sav(sav);
6250 				goto restart;
6251 			}
6252 			mutex_exit(&key_sad.lock);
6253 		}
6254 		key_sah_unref(sah);
6255 	}
6256     {
6257 	struct mbuf *n;
6258 
6259 	/* create new sadb_msg to reply. */
6260 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
6261 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6262 
6263 	key_fill_replymsg(n, 0);
6264 	m_freem(m);
6265 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6266     }
6267 }
6268 
6269 /*
6270  * SADB_GET processing
6271  * receive
6272  *   <base, SA(*), address(SD)>
6273  * from the ikmpd, and get a SP and a SA to respond,
6274  * and send,
6275  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6276  *       (identity(SD),) (sensitivity)>
6277  * to the ikmpd.
6278  *
6279  * m will always be freed.
6280  */
6281 static int
6282 key_api_get(struct socket *so, struct mbuf *m,
6283 	const struct sadb_msghdr *mhp)
6284 {
6285 	struct sadb_sa *sa0;
6286 	const struct sockaddr *src, *dst;
6287 	struct secasindex saidx;
6288 	struct secasvar *sav = NULL;
6289 	u_int16_t proto;
6290 	int error;
6291 
6292 	/* map satype to proto */
6293 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6294 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
6295 		return key_senderror(so, m, EINVAL);
6296 	}
6297 
6298 	if (mhp->ext[SADB_EXT_SA] == NULL ||
6299 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
6300 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
6301 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6302 		return key_senderror(so, m, EINVAL);
6303 	}
6304 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
6305 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
6306 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
6307 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6308 		return key_senderror(so, m, EINVAL);
6309 	}
6310 
6311 	sa0 = mhp->ext[SADB_EXT_SA];
6312 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6313 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6314 
6315 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
6316 	if (error != 0)
6317 		return key_senderror(so, m, EINVAL);
6318 
6319 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6320 	if (error != 0)
6321 		return key_senderror(so, m, EINVAL);
6322 
6323 	/* get a SA header */
6324     {
6325 	struct secashead *sah;
6326 	int s = pserialize_read_enter();
6327 
6328 	sah = key_getsah(&saidx, CMP_HEAD);
6329 	if (sah != NULL) {
6330 		/* get a SA with SPI. */
6331 		sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
6332 	}
6333 	pserialize_read_exit(s);
6334     }
6335 	if (sav == NULL) {
6336 		IPSECLOG(LOG_DEBUG, "no SA found.\n");
6337 		return key_senderror(so, m, ENOENT);
6338 	}
6339 
6340     {
6341 	struct mbuf *n;
6342 	u_int8_t satype;
6343 
6344 	/* map proto to satype */
6345 	satype = key_proto2satype(sav->sah->saidx.proto);
6346 	if (satype == 0) {
6347 		KEY_SA_UNREF(&sav);
6348 		IPSECLOG(LOG_DEBUG, "there was invalid proto in SAD.\n");
6349 		return key_senderror(so, m, EINVAL);
6350 	}
6351 
6352 	/* create new sadb_msg to reply. */
6353 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
6354 	    mhp->msg->sadb_msg_pid);
6355 	KEY_SA_UNREF(&sav);
6356 	m_freem(m);
6357 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
6358     }
6359 }
6360 
6361 /* XXX make it sysctl-configurable? */
6362 static void
6363 key_getcomb_setlifetime(struct sadb_comb *comb)
6364 {
6365 
6366 	comb->sadb_comb_soft_allocations = 1;
6367 	comb->sadb_comb_hard_allocations = 1;
6368 	comb->sadb_comb_soft_bytes = 0;
6369 	comb->sadb_comb_hard_bytes = 0;
6370 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
6371 	comb->sadb_comb_soft_addtime = comb->sadb_comb_hard_addtime * 80 / 100;
6372 	comb->sadb_comb_hard_usetime = 28800;	/* 8 hours */
6373 	comb->sadb_comb_soft_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
6374 }
6375 
6376 /*
6377  * XXX reorder combinations by preference
6378  * XXX no idea if the user wants ESP authentication or not
6379  */
6380 static struct mbuf *
6381 key_getcomb_esp(int mflag)
6382 {
6383 	struct sadb_comb *comb;
6384 	const struct enc_xform *algo;
6385 	struct mbuf *result = NULL, *m, *n;
6386 	int encmin;
6387 	int i, off, o;
6388 	int totlen;
6389 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6390 
6391 	m = NULL;
6392 	for (i = 1; i <= SADB_EALG_MAX; i++) {
6393 		algo = esp_algorithm_lookup(i);
6394 		if (algo == NULL)
6395 			continue;
6396 
6397 		/* discard algorithms with key size smaller than system min */
6398 		if (_BITS(algo->maxkey) < ipsec_esp_keymin)
6399 			continue;
6400 		if (_BITS(algo->minkey) < ipsec_esp_keymin)
6401 			encmin = ipsec_esp_keymin;
6402 		else
6403 			encmin = _BITS(algo->minkey);
6404 
6405 		if (ipsec_esp_auth)
6406 			m = key_getcomb_ah(mflag);
6407 		else {
6408 			KASSERTMSG(l <= MLEN,
6409 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
6410 			MGET(m, mflag, MT_DATA);
6411 			if (m) {
6412 				m_align(m, l);
6413 				m->m_len = l;
6414 				m->m_next = NULL;
6415 				memset(mtod(m, void *), 0, m->m_len);
6416 			}
6417 		}
6418 		if (!m)
6419 			goto fail;
6420 
6421 		totlen = 0;
6422 		for (n = m; n; n = n->m_next)
6423 			totlen += n->m_len;
6424 		KASSERTMSG((totlen % l) == 0, "totlen=%u, l=%u", totlen, l);
6425 
6426 		for (off = 0; off < totlen; off += l) {
6427 			n = m_pulldown(m, off, l, &o);
6428 			if (!n) {
6429 				/* m is already freed */
6430 				goto fail;
6431 			}
6432 			comb = (struct sadb_comb *)(mtod(n, char *) + o);
6433 			memset(comb, 0, sizeof(*comb));
6434 			key_getcomb_setlifetime(comb);
6435 			comb->sadb_comb_encrypt = i;
6436 			comb->sadb_comb_encrypt_minbits = encmin;
6437 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
6438 		}
6439 
6440 		if (!result)
6441 			result = m;
6442 		else
6443 			m_cat(result, m);
6444 	}
6445 
6446 	return result;
6447 
6448  fail:
6449 	if (result)
6450 		m_freem(result);
6451 	return NULL;
6452 }
6453 
6454 static void
6455 key_getsizes_ah(const struct auth_hash *ah, int alg,
6456 	        u_int16_t* ksmin, u_int16_t* ksmax)
6457 {
6458 	*ksmin = *ksmax = ah->keysize;
6459 	if (ah->keysize == 0) {
6460 		/*
6461 		 * Transform takes arbitrary key size but algorithm
6462 		 * key size is restricted.  Enforce this here.
6463 		 */
6464 		switch (alg) {
6465 		case SADB_X_AALG_MD5:	*ksmin = *ksmax = 16; break;
6466 		case SADB_X_AALG_SHA:	*ksmin = *ksmax = 20; break;
6467 		case SADB_X_AALG_NULL:	*ksmin = 0; *ksmax = 256; break;
6468 		default:
6469 			IPSECLOG(LOG_DEBUG, "unknown AH algorithm %u\n", alg);
6470 			break;
6471 		}
6472 	}
6473 }
6474 
6475 /*
6476  * XXX reorder combinations by preference
6477  */
6478 static struct mbuf *
6479 key_getcomb_ah(int mflag)
6480 {
6481 	struct sadb_comb *comb;
6482 	const struct auth_hash *algo;
6483 	struct mbuf *m;
6484 	u_int16_t minkeysize, maxkeysize;
6485 	int i;
6486 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6487 
6488 	m = NULL;
6489 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6490 #if 1
6491 		/* we prefer HMAC algorithms, not old algorithms */
6492 		if (i != SADB_AALG_SHA1HMAC &&
6493 		    i != SADB_AALG_MD5HMAC &&
6494 		    i != SADB_X_AALG_SHA2_256 &&
6495 		    i != SADB_X_AALG_SHA2_384 &&
6496 		    i != SADB_X_AALG_SHA2_512)
6497 			continue;
6498 #endif
6499 		algo = ah_algorithm_lookup(i);
6500 		if (!algo)
6501 			continue;
6502 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6503 		/* discard algorithms with key size smaller than system min */
6504 		if (_BITS(minkeysize) < ipsec_ah_keymin)
6505 			continue;
6506 
6507 		if (!m) {
6508 			KASSERTMSG(l <= MLEN,
6509 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
6510 			MGET(m, mflag, MT_DATA);
6511 			if (m) {
6512 				m_align(m, l);
6513 				m->m_len = l;
6514 				m->m_next = NULL;
6515 			}
6516 		} else
6517 			M_PREPEND(m, l, mflag);
6518 		if (!m)
6519 			return NULL;
6520 
6521 		if (m->m_len < sizeof(struct sadb_comb)) {
6522 			m = m_pullup(m, sizeof(struct sadb_comb));
6523 			if (m == NULL)
6524 				return NULL;
6525 		}
6526 
6527 		comb = mtod(m, struct sadb_comb *);
6528 		memset(comb, 0, sizeof(*comb));
6529 		key_getcomb_setlifetime(comb);
6530 		comb->sadb_comb_auth = i;
6531 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6532 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6533 	}
6534 
6535 	return m;
6536 }
6537 
6538 /*
6539  * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6540  * XXX reorder combinations by preference
6541  */
6542 static struct mbuf *
6543 key_getcomb_ipcomp(int mflag)
6544 {
6545 	struct sadb_comb *comb;
6546 	const struct comp_algo *algo;
6547 	struct mbuf *m;
6548 	int i;
6549 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6550 
6551 	m = NULL;
6552 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6553 		algo = ipcomp_algorithm_lookup(i);
6554 		if (!algo)
6555 			continue;
6556 
6557 		if (!m) {
6558 			KASSERTMSG(l <= MLEN,
6559 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
6560 			MGET(m, mflag, MT_DATA);
6561 			if (m) {
6562 				m_align(m, l);
6563 				m->m_len = l;
6564 				m->m_next = NULL;
6565 			}
6566 		} else
6567 			M_PREPEND(m, l, mflag);
6568 		if (!m)
6569 			return NULL;
6570 
6571 		if (m->m_len < sizeof(struct sadb_comb)) {
6572 			m = m_pullup(m, sizeof(struct sadb_comb));
6573 			if (m == NULL)
6574 				return NULL;
6575 		}
6576 
6577 		comb = mtod(m, struct sadb_comb *);
6578 		memset(comb, 0, sizeof(*comb));
6579 		key_getcomb_setlifetime(comb);
6580 		comb->sadb_comb_encrypt = i;
6581 		/* what should we set into sadb_comb_*_{min,max}bits? */
6582 	}
6583 
6584 	return m;
6585 }
6586 
6587 /*
6588  * XXX no way to pass mode (transport/tunnel) to userland
6589  * XXX replay checking?
6590  * XXX sysctl interface to ipsec_{ah,esp}_keymin
6591  */
6592 static struct mbuf *
6593 key_getprop(const struct secasindex *saidx, int mflag)
6594 {
6595 	struct sadb_prop *prop;
6596 	struct mbuf *m, *n;
6597 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6598 	int totlen;
6599 
6600 	switch (saidx->proto)  {
6601 	case IPPROTO_ESP:
6602 		m = key_getcomb_esp(mflag);
6603 		break;
6604 	case IPPROTO_AH:
6605 		m = key_getcomb_ah(mflag);
6606 		break;
6607 	case IPPROTO_IPCOMP:
6608 		m = key_getcomb_ipcomp(mflag);
6609 		break;
6610 	default:
6611 		return NULL;
6612 	}
6613 
6614 	if (!m)
6615 		return NULL;
6616 	M_PREPEND(m, l, mflag);
6617 	if (!m)
6618 		return NULL;
6619 
6620 	totlen = 0;
6621 	for (n = m; n; n = n->m_next)
6622 		totlen += n->m_len;
6623 
6624 	prop = mtod(m, struct sadb_prop *);
6625 	memset(prop, 0, sizeof(*prop));
6626 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6627 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6628 	prop->sadb_prop_replay = 32;	/* XXX */
6629 
6630 	return m;
6631 }
6632 
6633 /*
6634  * SADB_ACQUIRE processing called by key_checkrequest() and key_api_acquire().
6635  * send
6636  *   <base, SA, address(SD), (address(P)), x_policy,
6637  *       (identity(SD),) (sensitivity,) proposal>
6638  * to KMD, and expect to receive
6639  *   <base> with SADB_ACQUIRE if error occurred,
6640  * or
6641  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6642  * from KMD by PF_KEY.
6643  *
6644  * XXX x_policy is outside of RFC2367 (KAME extension).
6645  * XXX sensitivity is not supported.
6646  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6647  * see comment for key_getcomb_ipcomp().
6648  *
6649  * OUT:
6650  *    0     : succeed
6651  *    others: error number
6652  */
6653 static int
6654 key_acquire(const struct secasindex *saidx, const struct secpolicy *sp, int mflag)
6655 {
6656 	struct mbuf *result = NULL, *m;
6657 #ifndef IPSEC_NONBLOCK_ACQUIRE
6658 	struct secacq *newacq;
6659 #endif
6660 	u_int8_t satype;
6661 	int error = -1;
6662 	u_int32_t seq;
6663 
6664 	/* sanity check */
6665 	KASSERT(saidx != NULL);
6666 	satype = key_proto2satype(saidx->proto);
6667 	KASSERTMSG(satype != 0, "null satype, protocol %u", saidx->proto);
6668 
6669 #ifndef IPSEC_NONBLOCK_ACQUIRE
6670 	/*
6671 	 * We never do anything about acquirng SA.  There is anather
6672 	 * solution that kernel blocks to send SADB_ACQUIRE message until
6673 	 * getting something message from IKEd.  In later case, to be
6674 	 * managed with ACQUIRING list.
6675 	 */
6676 	/* Get an entry to check whether sending message or not. */
6677 	mutex_enter(&key_misc.lock);
6678 	newacq = key_getacq(saidx);
6679 	if (newacq != NULL) {
6680 		if (key_blockacq_count < newacq->count) {
6681 			/* reset counter and do send message. */
6682 			newacq->count = 0;
6683 		} else {
6684 			/* increment counter and do nothing. */
6685 			newacq->count++;
6686 			mutex_exit(&key_misc.lock);
6687 			return 0;
6688 		}
6689 	} else {
6690 		/* make new entry for blocking to send SADB_ACQUIRE. */
6691 		newacq = key_newacq(saidx);
6692 		if (newacq == NULL) {
6693 			mutex_exit(&key_misc.lock);
6694 			return ENOBUFS;
6695 		}
6696 
6697 		/* add to key_misc.acqlist */
6698 		LIST_INSERT_HEAD(&key_misc.acqlist, newacq, chain);
6699 	}
6700 
6701 	seq = newacq->seq;
6702 	mutex_exit(&key_misc.lock);
6703 #else
6704 	seq = (acq_seq = (acq_seq == ~0 ? 1 : ++acq_seq));
6705 #endif
6706 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0, mflag);
6707 	if (!m) {
6708 		error = ENOBUFS;
6709 		goto fail;
6710 	}
6711 	result = m;
6712 
6713 	/* set sadb_address for saidx's. */
6714 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &saidx->src.sa, FULLMASK,
6715 	    IPSEC_ULPROTO_ANY, mflag);
6716 	if (!m) {
6717 		error = ENOBUFS;
6718 		goto fail;
6719 	}
6720 	m_cat(result, m);
6721 
6722 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &saidx->dst.sa, FULLMASK,
6723 	    IPSEC_ULPROTO_ANY, mflag);
6724 	if (!m) {
6725 		error = ENOBUFS;
6726 		goto fail;
6727 	}
6728 	m_cat(result, m);
6729 
6730 	/* XXX proxy address (optional) */
6731 
6732 	/* set sadb_x_policy */
6733 	if (sp) {
6734 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6735 		    mflag);
6736 		if (!m) {
6737 			error = ENOBUFS;
6738 			goto fail;
6739 		}
6740 		m_cat(result, m);
6741 	}
6742 
6743 	/* XXX identity (optional) */
6744 #if 0
6745 	if (idexttype && fqdn) {
6746 		/* create identity extension (FQDN) */
6747 		struct sadb_ident *id;
6748 		int fqdnlen;
6749 
6750 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
6751 		id = (struct sadb_ident *)p;
6752 		memset(id, 0, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6753 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6754 		id->sadb_ident_exttype = idexttype;
6755 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
6756 		memcpy(id + 1, fqdn, fqdnlen);
6757 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
6758 	}
6759 
6760 	if (idexttype) {
6761 		/* create identity extension (USERFQDN) */
6762 		struct sadb_ident *id;
6763 		int userfqdnlen;
6764 
6765 		if (userfqdn) {
6766 			/* +1 for terminating-NUL */
6767 			userfqdnlen = strlen(userfqdn) + 1;
6768 		} else
6769 			userfqdnlen = 0;
6770 		id = (struct sadb_ident *)p;
6771 		memset(id, 0, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6772 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6773 		id->sadb_ident_exttype = idexttype;
6774 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
6775 		/* XXX is it correct? */
6776 		if (curlwp)
6777 			id->sadb_ident_id = kauth_cred_getuid(curlwp->l_cred);
6778 		if (userfqdn && userfqdnlen)
6779 			memcpy(id + 1, userfqdn, userfqdnlen);
6780 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
6781 	}
6782 #endif
6783 
6784 	/* XXX sensitivity (optional) */
6785 
6786 	/* create proposal/combination extension */
6787 	m = key_getprop(saidx, mflag);
6788 #if 0
6789 	/*
6790 	 * spec conformant: always attach proposal/combination extension,
6791 	 * the problem is that we have no way to attach it for ipcomp,
6792 	 * due to the way sadb_comb is declared in RFC2367.
6793 	 */
6794 	if (!m) {
6795 		error = ENOBUFS;
6796 		goto fail;
6797 	}
6798 	m_cat(result, m);
6799 #else
6800 	/*
6801 	 * outside of spec; make proposal/combination extension optional.
6802 	 */
6803 	if (m)
6804 		m_cat(result, m);
6805 #endif
6806 
6807 	KASSERT(result->m_flags & M_PKTHDR);
6808 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
6809 
6810 	result->m_pkthdr.len = 0;
6811 	for (m = result; m; m = m->m_next)
6812 		result->m_pkthdr.len += m->m_len;
6813 
6814 	mtod(result, struct sadb_msg *)->sadb_msg_len =
6815 	    PFKEY_UNIT64(result->m_pkthdr.len);
6816 
6817 	/*
6818 	 * Called from key_api_acquire that must come from userland, so
6819 	 * we can call key_sendup_mbuf immediately.
6820 	 */
6821 	if (mflag == M_WAITOK)
6822 		return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
6823 	/*
6824 	 * XXX we cannot call key_sendup_mbuf directly here because
6825 	 * it can cause a deadlock:
6826 	 * - We have a reference to an SP (and an SA) here
6827 	 * - key_sendup_mbuf will try to take key_so_mtx
6828 	 * - Some other thread may try to localcount_drain to the SP with
6829 	 *   holding key_so_mtx in say key_api_spdflush
6830 	 * - In this case localcount_drain never return because key_sendup_mbuf
6831 	 *   that has stuck on key_so_mtx never release a reference to the SP
6832 	 *
6833 	 * So defer key_sendup_mbuf to the timer.
6834 	 */
6835 	return key_acquire_sendup_mbuf_later(result);
6836 
6837  fail:
6838 	if (result)
6839 		m_freem(result);
6840 	return error;
6841 }
6842 
6843 static struct mbuf *key_acquire_mbuf_head = NULL;
6844 static unsigned key_acquire_mbuf_count = 0;
6845 #define KEY_ACQUIRE_MBUF_MAX	10
6846 
6847 static void
6848 key_acquire_sendup_pending_mbuf(void)
6849 {
6850 	struct mbuf *m, *prev;
6851 	int error;
6852 
6853 again:
6854 	prev = NULL;
6855 	mutex_enter(&key_misc.lock);
6856 	m = key_acquire_mbuf_head;
6857 	/* Get an earliest mbuf (one at the tail of the list) */
6858 	while (m != NULL) {
6859 		if (m->m_nextpkt == NULL) {
6860 			if (prev != NULL)
6861 				prev->m_nextpkt = NULL;
6862 			if (m == key_acquire_mbuf_head)
6863 				key_acquire_mbuf_head = NULL;
6864 			key_acquire_mbuf_count--;
6865 			break;
6866 		}
6867 		prev = m;
6868 		m = m->m_nextpkt;
6869 	}
6870 	mutex_exit(&key_misc.lock);
6871 
6872 	if (m == NULL)
6873 		return;
6874 
6875 	m->m_nextpkt = NULL;
6876 	error = key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
6877 	if (error != 0)
6878 		IPSECLOG(LOG_WARNING, "key_sendup_mbuf failed (error=%d)\n",
6879 		    error);
6880 
6881 	if (prev != NULL)
6882 		goto again;
6883 }
6884 
6885 static int
6886 key_acquire_sendup_mbuf_later(struct mbuf *m)
6887 {
6888 
6889 	mutex_enter(&key_misc.lock);
6890 	/* Avoid queuing too much mbufs */
6891 	if (key_acquire_mbuf_count >= KEY_ACQUIRE_MBUF_MAX) {
6892 		mutex_exit(&key_misc.lock);
6893 		m_freem(m);
6894 		return ENOBUFS; /* XXX */
6895 	}
6896 	/* Enqueue mbuf at the head of the list */
6897 	m->m_nextpkt = key_acquire_mbuf_head;
6898 	key_acquire_mbuf_head = m;
6899 	key_acquire_mbuf_count++;
6900 	mutex_exit(&key_misc.lock);
6901 
6902 	/* Kick the timer */
6903 	key_timehandler(NULL);
6904 
6905 	return 0;
6906 }
6907 
6908 #ifndef IPSEC_NONBLOCK_ACQUIRE
6909 static struct secacq *
6910 key_newacq(const struct secasindex *saidx)
6911 {
6912 	struct secacq *newacq;
6913 
6914 	/* get new entry */
6915 	newacq = kmem_intr_zalloc(sizeof(struct secacq), KM_NOSLEEP);
6916 	if (newacq == NULL) {
6917 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
6918 		return NULL;
6919 	}
6920 
6921 	/* copy secindex */
6922 	memcpy(&newacq->saidx, saidx, sizeof(newacq->saidx));
6923 	newacq->seq = (acq_seq == ~0 ? 1 : ++acq_seq);
6924 	newacq->created = time_uptime;
6925 	newacq->count = 0;
6926 
6927 	return newacq;
6928 }
6929 
6930 static struct secacq *
6931 key_getacq(const struct secasindex *saidx)
6932 {
6933 	struct secacq *acq;
6934 
6935 	KASSERT(mutex_owned(&key_misc.lock));
6936 
6937 	LIST_FOREACH(acq, &key_misc.acqlist, chain) {
6938 		if (key_saidx_match(saidx, &acq->saidx, CMP_EXACTLY))
6939 			return acq;
6940 	}
6941 
6942 	return NULL;
6943 }
6944 
6945 static struct secacq *
6946 key_getacqbyseq(u_int32_t seq)
6947 {
6948 	struct secacq *acq;
6949 
6950 	KASSERT(mutex_owned(&key_misc.lock));
6951 
6952 	LIST_FOREACH(acq, &key_misc.acqlist, chain) {
6953 		if (acq->seq == seq)
6954 			return acq;
6955 	}
6956 
6957 	return NULL;
6958 }
6959 #endif
6960 
6961 #ifdef notyet
6962 static struct secspacq *
6963 key_newspacq(const struct secpolicyindex *spidx)
6964 {
6965 	struct secspacq *acq;
6966 
6967 	/* get new entry */
6968 	acq = kmem_intr_zalloc(sizeof(struct secspacq), KM_NOSLEEP);
6969 	if (acq == NULL) {
6970 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
6971 		return NULL;
6972 	}
6973 
6974 	/* copy secindex */
6975 	memcpy(&acq->spidx, spidx, sizeof(acq->spidx));
6976 	acq->created = time_uptime;
6977 	acq->count = 0;
6978 
6979 	return acq;
6980 }
6981 
6982 static struct secspacq *
6983 key_getspacq(const struct secpolicyindex *spidx)
6984 {
6985 	struct secspacq *acq;
6986 
6987 	LIST_FOREACH(acq, &key_misc.spacqlist, chain) {
6988 		if (key_spidx_match_exactly(spidx, &acq->spidx))
6989 			return acq;
6990 	}
6991 
6992 	return NULL;
6993 }
6994 #endif /* notyet */
6995 
6996 /*
6997  * SADB_ACQUIRE processing,
6998  * in first situation, is receiving
6999  *   <base>
7000  * from the ikmpd, and clear sequence of its secasvar entry.
7001  *
7002  * In second situation, is receiving
7003  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
7004  * from a user land process, and return
7005  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
7006  * to the socket.
7007  *
7008  * m will always be freed.
7009  */
7010 static int
7011 key_api_acquire(struct socket *so, struct mbuf *m,
7012       	     const struct sadb_msghdr *mhp)
7013 {
7014 	const struct sockaddr *src, *dst;
7015 	struct secasindex saidx;
7016 	u_int16_t proto;
7017 	int error;
7018 
7019 	/*
7020 	 * Error message from KMd.
7021 	 * We assume that if error was occurred in IKEd, the length of PFKEY
7022 	 * message is equal to the size of sadb_msg structure.
7023 	 * We do not raise error even if error occurred in this function.
7024 	 */
7025 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
7026 #ifndef IPSEC_NONBLOCK_ACQUIRE
7027 		struct secacq *acq;
7028 
7029 		/* check sequence number */
7030 		if (mhp->msg->sadb_msg_seq == 0) {
7031 			IPSECLOG(LOG_DEBUG, "must specify sequence number.\n");
7032 			m_freem(m);
7033 			return 0;
7034 		}
7035 
7036 		mutex_enter(&key_misc.lock);
7037 		acq = key_getacqbyseq(mhp->msg->sadb_msg_seq);
7038 		if (acq == NULL) {
7039 			mutex_exit(&key_misc.lock);
7040 			/*
7041 			 * the specified larval SA is already gone, or we got
7042 			 * a bogus sequence number.  we can silently ignore it.
7043 			 */
7044 			m_freem(m);
7045 			return 0;
7046 		}
7047 
7048 		/* reset acq counter in order to deletion by timehander. */
7049 		acq->created = time_uptime;
7050 		acq->count = 0;
7051 		mutex_exit(&key_misc.lock);
7052 #endif
7053 		m_freem(m);
7054 		return 0;
7055 	}
7056 
7057 	/*
7058 	 * This message is from user land.
7059 	 */
7060 
7061 	/* map satype to proto */
7062 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
7063 	if (proto == 0) {
7064 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
7065 		return key_senderror(so, m, EINVAL);
7066 	}
7067 
7068 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
7069 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
7070 	    mhp->ext[SADB_EXT_PROPOSAL] == NULL) {
7071 		/* error */
7072 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
7073 		return key_senderror(so, m, EINVAL);
7074 	}
7075 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
7076 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
7077 	    mhp->extlen[SADB_EXT_PROPOSAL] < sizeof(struct sadb_prop)) {
7078 		/* error */
7079 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
7080 		return key_senderror(so, m, EINVAL);
7081 	}
7082 
7083 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
7084 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
7085 
7086 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
7087 	if (error != 0)
7088 		return key_senderror(so, m, EINVAL);
7089 
7090 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
7091 	if (error != 0)
7092 		return key_senderror(so, m, EINVAL);
7093 
7094 	/* get a SA index */
7095     {
7096 	struct secashead *sah;
7097 	int s = pserialize_read_enter();
7098 
7099 	sah = key_getsah(&saidx, CMP_MODE_REQID);
7100 	if (sah != NULL) {
7101 		pserialize_read_exit(s);
7102 		IPSECLOG(LOG_DEBUG, "a SA exists already.\n");
7103 		return key_senderror(so, m, EEXIST);
7104 	}
7105 	pserialize_read_exit(s);
7106     }
7107 
7108 	error = key_acquire(&saidx, NULL, M_WAITOK);
7109 	if (error != 0) {
7110 		IPSECLOG(LOG_DEBUG, "error %d returned from key_acquire.\n",
7111 		    error);
7112 		return key_senderror(so, m, error);
7113 	}
7114 
7115 	return key_sendup_mbuf(so, m, KEY_SENDUP_REGISTERED);
7116 }
7117 
7118 /*
7119  * SADB_REGISTER processing.
7120  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
7121  * receive
7122  *   <base>
7123  * from the ikmpd, and register a socket to send PF_KEY messages,
7124  * and send
7125  *   <base, supported>
7126  * to KMD by PF_KEY.
7127  * If socket is detached, must free from regnode.
7128  *
7129  * m will always be freed.
7130  */
7131 static int
7132 key_api_register(struct socket *so, struct mbuf *m,
7133 	     const struct sadb_msghdr *mhp)
7134 {
7135 	struct secreg *reg, *newreg = 0;
7136 
7137 	/* check for invalid register message */
7138 	if (mhp->msg->sadb_msg_satype >= __arraycount(key_misc.reglist))
7139 		return key_senderror(so, m, EINVAL);
7140 
7141 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
7142 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
7143 		goto setmsg;
7144 
7145 	/* Allocate regnode in advance, out of mutex */
7146 	newreg = kmem_zalloc(sizeof(*newreg), KM_SLEEP);
7147 
7148 	/* check whether existing or not */
7149 	mutex_enter(&key_misc.lock);
7150 	LIST_FOREACH(reg, &key_misc.reglist[mhp->msg->sadb_msg_satype], chain) {
7151 		if (reg->so == so) {
7152 			IPSECLOG(LOG_DEBUG, "socket exists already.\n");
7153 			mutex_exit(&key_misc.lock);
7154 			kmem_free(newreg, sizeof(*newreg));
7155 			return key_senderror(so, m, EEXIST);
7156 		}
7157 	}
7158 
7159 	newreg->so = so;
7160 	((struct keycb *)sotorawcb(so))->kp_registered++;
7161 
7162 	/* add regnode to key_misc.reglist. */
7163 	LIST_INSERT_HEAD(&key_misc.reglist[mhp->msg->sadb_msg_satype], newreg, chain);
7164 	mutex_exit(&key_misc.lock);
7165 
7166   setmsg:
7167     {
7168 	struct mbuf *n;
7169 	struct sadb_supported *sup;
7170 	u_int len, alen, elen;
7171 	int off;
7172 	int i;
7173 	struct sadb_alg *alg;
7174 
7175 	/* create new sadb_msg to reply. */
7176 	alen = 0;
7177 	for (i = 1; i <= SADB_AALG_MAX; i++) {
7178 		if (ah_algorithm_lookup(i))
7179 			alen += sizeof(struct sadb_alg);
7180 	}
7181 	if (alen)
7182 		alen += sizeof(struct sadb_supported);
7183 	elen = 0;
7184 	for (i = 1; i <= SADB_EALG_MAX; i++) {
7185 		if (esp_algorithm_lookup(i))
7186 			elen += sizeof(struct sadb_alg);
7187 	}
7188 	if (elen)
7189 		elen += sizeof(struct sadb_supported);
7190 
7191 	len = sizeof(struct sadb_msg) + alen + elen;
7192 
7193 	if (len > MCLBYTES)
7194 		return key_senderror(so, m, ENOBUFS);
7195 
7196 	n = key_alloc_mbuf_simple(len, M_WAITOK);
7197 	n->m_pkthdr.len = n->m_len = len;
7198 	n->m_next = NULL;
7199 	off = 0;
7200 
7201 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
7202 	key_fill_replymsg(n, 0);
7203 
7204 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7205 
7206 	/* for authentication algorithm */
7207 	if (alen) {
7208 		sup = (struct sadb_supported *)(mtod(n, char *) + off);
7209 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7210 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7211 		off += PFKEY_ALIGN8(sizeof(*sup));
7212 
7213 		for (i = 1; i <= SADB_AALG_MAX; i++) {
7214 			const struct auth_hash *aalgo;
7215 			u_int16_t minkeysize, maxkeysize;
7216 
7217 			aalgo = ah_algorithm_lookup(i);
7218 			if (!aalgo)
7219 				continue;
7220 			alg = (struct sadb_alg *)(mtod(n, char *) + off);
7221 			alg->sadb_alg_id = i;
7222 			alg->sadb_alg_ivlen = 0;
7223 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7224 			alg->sadb_alg_minbits = _BITS(minkeysize);
7225 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7226 			off += PFKEY_ALIGN8(sizeof(*alg));
7227 		}
7228 	}
7229 
7230 	/* for encryption algorithm */
7231 	if (elen) {
7232 		sup = (struct sadb_supported *)(mtod(n, char *) + off);
7233 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7234 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7235 		off += PFKEY_ALIGN8(sizeof(*sup));
7236 
7237 		for (i = 1; i <= SADB_EALG_MAX; i++) {
7238 			const struct enc_xform *ealgo;
7239 
7240 			ealgo = esp_algorithm_lookup(i);
7241 			if (!ealgo)
7242 				continue;
7243 			alg = (struct sadb_alg *)(mtod(n, char *) + off);
7244 			alg->sadb_alg_id = i;
7245 			alg->sadb_alg_ivlen = ealgo->blocksize;
7246 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7247 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7248 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7249 		}
7250 	}
7251 
7252 	KASSERTMSG(off == len, "length inconsistency");
7253 
7254 	m_freem(m);
7255 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7256     }
7257 }
7258 
7259 /*
7260  * free secreg entry registered.
7261  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7262  */
7263 void
7264 key_freereg(struct socket *so)
7265 {
7266 	struct secreg *reg;
7267 	int i;
7268 
7269 	KASSERT(!cpu_softintr_p());
7270 	KASSERT(so != NULL);
7271 
7272 	/*
7273 	 * check whether existing or not.
7274 	 * check all type of SA, because there is a potential that
7275 	 * one socket is registered to multiple type of SA.
7276 	 */
7277 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7278 		mutex_enter(&key_misc.lock);
7279 		LIST_FOREACH(reg, &key_misc.reglist[i], chain) {
7280 			if (reg->so == so) {
7281 				LIST_REMOVE(reg, chain);
7282 				break;
7283 			}
7284 		}
7285 		mutex_exit(&key_misc.lock);
7286 		if (reg != NULL)
7287 			kmem_free(reg, sizeof(*reg));
7288 	}
7289 
7290 	return;
7291 }
7292 
7293 /*
7294  * SADB_EXPIRE processing
7295  * send
7296  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7297  * to KMD by PF_KEY.
7298  * NOTE: We send only soft lifetime extension.
7299  *
7300  * OUT:	0	: succeed
7301  *	others	: error number
7302  */
7303 static int
7304 key_expire(struct secasvar *sav)
7305 {
7306 	int s;
7307 	int satype;
7308 	struct mbuf *result = NULL, *m;
7309 	int len;
7310 	int error = -1;
7311 	struct sadb_lifetime *lt;
7312 	lifetime_counters_t sum = {0};
7313 
7314 	/* XXX: Why do we lock ? */
7315 	s = splsoftnet();	/*called from softclock()*/
7316 
7317 	KASSERT(sav != NULL);
7318 
7319 	satype = key_proto2satype(sav->sah->saidx.proto);
7320 	KASSERTMSG(satype != 0, "invalid proto is passed");
7321 
7322 	/* set msg header */
7323 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, key_sa_refcnt(sav),
7324 	    M_WAITOK);
7325 	result = m;
7326 
7327 	/* create SA extension */
7328 	m = key_setsadbsa(sav);
7329 	m_cat(result, m);
7330 
7331 	/* create SA extension */
7332 	m = key_setsadbxsa2(sav->sah->saidx.mode,
7333 	    sav->replay ? sav->replay->count : 0, sav->sah->saidx.reqid);
7334 	m_cat(result, m);
7335 
7336 	/* create lifetime extension (current and soft) */
7337 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7338 	m = key_alloc_mbuf(len, M_WAITOK);
7339 	KASSERT(m->m_next == NULL);
7340 
7341 	memset(mtod(m, void *), 0, len);
7342 	lt = mtod(m, struct sadb_lifetime *);
7343 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7344 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7345 	percpu_foreach(sav->lft_c_counters_percpu,
7346 	    key_sum_lifetime_counters, sum);
7347 	lt->sadb_lifetime_allocations = sum[LIFETIME_COUNTER_ALLOCATIONS];
7348 	lt->sadb_lifetime_bytes = sum[LIFETIME_COUNTER_BYTES];
7349 	lt->sadb_lifetime_addtime =
7350 	    time_mono_to_wall(sav->lft_c->sadb_lifetime_addtime);
7351 	lt->sadb_lifetime_usetime =
7352 	    time_mono_to_wall(sav->lft_c->sadb_lifetime_usetime);
7353 	lt = (struct sadb_lifetime *)(mtod(m, char *) + len / 2);
7354 	memcpy(lt, sav->lft_s, sizeof(*lt));
7355 	m_cat(result, m);
7356 
7357 	/* set sadb_address for source */
7358 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa,
7359 	    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
7360 	m_cat(result, m);
7361 
7362 	/* set sadb_address for destination */
7363 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.dst.sa,
7364 	    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
7365 	m_cat(result, m);
7366 
7367 	if ((result->m_flags & M_PKTHDR) == 0) {
7368 		error = EINVAL;
7369 		goto fail;
7370 	}
7371 
7372 	if (result->m_len < sizeof(struct sadb_msg)) {
7373 		result = m_pullup(result, sizeof(struct sadb_msg));
7374 		if (result == NULL) {
7375 			error = ENOBUFS;
7376 			goto fail;
7377 		}
7378 	}
7379 
7380 	result->m_pkthdr.len = 0;
7381 	for (m = result; m; m = m->m_next)
7382 		result->m_pkthdr.len += m->m_len;
7383 
7384 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7385 	    PFKEY_UNIT64(result->m_pkthdr.len);
7386 
7387 	splx(s);
7388 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7389 
7390  fail:
7391 	if (result)
7392 		m_freem(result);
7393 	splx(s);
7394 	return error;
7395 }
7396 
7397 /*
7398  * SADB_FLUSH processing
7399  * receive
7400  *   <base>
7401  * from the ikmpd, and free all entries in secastree.
7402  * and send,
7403  *   <base>
7404  * to the ikmpd.
7405  * NOTE: to do is only marking SADB_SASTATE_DEAD.
7406  *
7407  * m will always be freed.
7408  */
7409 static int
7410 key_api_flush(struct socket *so, struct mbuf *m,
7411           const struct sadb_msghdr *mhp)
7412 {
7413 	struct sadb_msg *newmsg;
7414 	struct secashead *sah;
7415 	struct secasvar *sav;
7416 	u_int16_t proto;
7417 	u_int8_t state;
7418 	int s;
7419 
7420 	/* map satype to proto */
7421 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
7422 	if (proto == 0) {
7423 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
7424 		return key_senderror(so, m, EINVAL);
7425 	}
7426 
7427 	/* no SATYPE specified, i.e. flushing all SA. */
7428 	s = pserialize_read_enter();
7429 	SAHLIST_READER_FOREACH(sah) {
7430 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7431 		    proto != sah->saidx.proto)
7432 			continue;
7433 
7434 		key_sah_ref(sah);
7435 		pserialize_read_exit(s);
7436 
7437 		SASTATE_ALIVE_FOREACH(state) {
7438 		restart:
7439 			mutex_enter(&key_sad.lock);
7440 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
7441 				sav->state = SADB_SASTATE_DEAD;
7442 				key_unlink_sav(sav);
7443 				mutex_exit(&key_sad.lock);
7444 				key_destroy_sav(sav);
7445 				goto restart;
7446 			}
7447 			mutex_exit(&key_sad.lock);
7448 		}
7449 
7450 		s = pserialize_read_enter();
7451 		sah->state = SADB_SASTATE_DEAD;
7452 		key_sah_unref(sah);
7453 	}
7454 	pserialize_read_exit(s);
7455 
7456 	if (m->m_len < sizeof(struct sadb_msg) ||
7457 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7458 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
7459 		return key_senderror(so, m, ENOBUFS);
7460 	}
7461 
7462 	if (m->m_next)
7463 		m_freem(m->m_next);
7464 	m->m_next = NULL;
7465 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7466 	newmsg = mtod(m, struct sadb_msg *);
7467 	newmsg->sadb_msg_errno = 0;
7468 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7469 
7470 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7471 }
7472 
7473 
7474 static struct mbuf *
7475 key_setdump_chain(u_int8_t req_satype, int *errorp, int *lenp, pid_t pid)
7476 {
7477 	struct secashead *sah;
7478 	struct secasvar *sav;
7479 	u_int16_t proto;
7480 	u_int8_t satype;
7481 	u_int8_t state;
7482 	int cnt;
7483 	struct mbuf *m, *n, *prev;
7484 
7485 	KASSERT(mutex_owned(&key_sad.lock));
7486 
7487 	*lenp = 0;
7488 
7489 	/* map satype to proto */
7490 	proto = key_satype2proto(req_satype);
7491 	if (proto == 0) {
7492 		*errorp = EINVAL;
7493 		return (NULL);
7494 	}
7495 
7496 	/* count sav entries to be sent to userland. */
7497 	cnt = 0;
7498 	SAHLIST_WRITER_FOREACH(sah) {
7499 		if (req_satype != SADB_SATYPE_UNSPEC &&
7500 		    proto != sah->saidx.proto)
7501 			continue;
7502 
7503 		SASTATE_ANY_FOREACH(state) {
7504 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
7505 				cnt++;
7506 			}
7507 		}
7508 	}
7509 
7510 	if (cnt == 0) {
7511 		*errorp = ENOENT;
7512 		return (NULL);
7513 	}
7514 
7515 	/* send this to the userland, one at a time. */
7516 	m = NULL;
7517 	prev = m;
7518 	SAHLIST_WRITER_FOREACH(sah) {
7519 		if (req_satype != SADB_SATYPE_UNSPEC &&
7520 		    proto != sah->saidx.proto)
7521 			continue;
7522 
7523 		/* map proto to satype */
7524 		satype = key_proto2satype(sah->saidx.proto);
7525 		if (satype == 0) {
7526 			m_freem(m);
7527 			*errorp = EINVAL;
7528 			return (NULL);
7529 		}
7530 
7531 		SASTATE_ANY_FOREACH(state) {
7532 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
7533 				n = key_setdumpsa(sav, SADB_DUMP, satype,
7534 				    --cnt, pid);
7535 				if (!m)
7536 					m = n;
7537 				else
7538 					prev->m_nextpkt = n;
7539 				prev = n;
7540 			}
7541 		}
7542 	}
7543 
7544 	if (!m) {
7545 		*errorp = EINVAL;
7546 		return (NULL);
7547 	}
7548 
7549 	if ((m->m_flags & M_PKTHDR) != 0) {
7550 		m->m_pkthdr.len = 0;
7551 		for (n = m; n; n = n->m_next)
7552 			m->m_pkthdr.len += n->m_len;
7553 	}
7554 
7555 	*errorp = 0;
7556 	return (m);
7557 }
7558 
7559 /*
7560  * SADB_DUMP processing
7561  * dump all entries including status of DEAD in SAD.
7562  * receive
7563  *   <base>
7564  * from the ikmpd, and dump all secasvar leaves
7565  * and send,
7566  *   <base> .....
7567  * to the ikmpd.
7568  *
7569  * m will always be freed.
7570  */
7571 static int
7572 key_api_dump(struct socket *so, struct mbuf *m0,
7573 	 const struct sadb_msghdr *mhp)
7574 {
7575 	u_int16_t proto;
7576 	u_int8_t satype;
7577 	struct mbuf *n;
7578 	int error, len, ok;
7579 
7580 	/* map satype to proto */
7581 	satype = mhp->msg->sadb_msg_satype;
7582 	proto = key_satype2proto(satype);
7583 	if (proto == 0) {
7584 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
7585 		return key_senderror(so, m0, EINVAL);
7586 	}
7587 
7588 	/*
7589 	 * If the requestor has insufficient socket-buffer space
7590 	 * for the entire chain, nobody gets any response to the DUMP.
7591 	 * XXX For now, only the requestor ever gets anything.
7592 	 * Moreover, if the requestor has any space at all, they receive
7593 	 * the entire chain, otherwise the request is refused with ENOBUFS.
7594 	 */
7595 	if (sbspace(&so->so_rcv) <= 0) {
7596 		return key_senderror(so, m0, ENOBUFS);
7597 	}
7598 
7599 	mutex_enter(&key_sad.lock);
7600 	n = key_setdump_chain(satype, &error, &len, mhp->msg->sadb_msg_pid);
7601 	mutex_exit(&key_sad.lock);
7602 
7603 	if (n == NULL) {
7604 		return key_senderror(so, m0, ENOENT);
7605 	}
7606 	{
7607 		uint64_t *ps = PFKEY_STAT_GETREF();
7608 		ps[PFKEY_STAT_IN_TOTAL]++;
7609 		ps[PFKEY_STAT_IN_BYTES] += len;
7610 		PFKEY_STAT_PUTREF();
7611 	}
7612 
7613 	/*
7614 	 * PF_KEY DUMP responses are no longer broadcast to all PF_KEY sockets.
7615 	 * The requestor receives either the entire chain, or an
7616 	 * error message with ENOBUFS.
7617 	 *
7618 	 * sbappendaddrchain() takes the chain of entries, one
7619 	 * packet-record per SPD entry, prepends the key_src sockaddr
7620 	 * to each packet-record, links the sockaddr mbufs into a new
7621 	 * list of records, then   appends the entire resulting
7622 	 * list to the requesting socket.
7623 	 */
7624 	ok = sbappendaddrchain(&so->so_rcv, (struct sockaddr *)&key_src, n,
7625 	    SB_PRIO_ONESHOT_OVERFLOW);
7626 
7627 	if (!ok) {
7628 		PFKEY_STATINC(PFKEY_STAT_IN_NOMEM);
7629 		m_freem(n);
7630 		return key_senderror(so, m0, ENOBUFS);
7631 	}
7632 
7633 	m_freem(m0);
7634 	return 0;
7635 }
7636 
7637 /*
7638  * SADB_X_PROMISC processing
7639  *
7640  * m will always be freed.
7641  */
7642 static int
7643 key_api_promisc(struct socket *so, struct mbuf *m,
7644 	    const struct sadb_msghdr *mhp)
7645 {
7646 	int olen;
7647 
7648 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7649 
7650 	if (olen < sizeof(struct sadb_msg)) {
7651 #if 1
7652 		return key_senderror(so, m, EINVAL);
7653 #else
7654 		m_freem(m);
7655 		return 0;
7656 #endif
7657 	} else if (olen == sizeof(struct sadb_msg)) {
7658 		/* enable/disable promisc mode */
7659 		struct keycb *kp = (struct keycb *)sotorawcb(so);
7660 		if (kp == NULL)
7661 			return key_senderror(so, m, EINVAL);
7662 		mhp->msg->sadb_msg_errno = 0;
7663 		switch (mhp->msg->sadb_msg_satype) {
7664 		case 0:
7665 		case 1:
7666 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
7667 			break;
7668 		default:
7669 			return key_senderror(so, m, EINVAL);
7670 		}
7671 
7672 		/* send the original message back to everyone */
7673 		mhp->msg->sadb_msg_errno = 0;
7674 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7675 	} else {
7676 		/* send packet as is */
7677 
7678 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
7679 
7680 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
7681 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7682 	}
7683 }
7684 
7685 static int (*key_api_typesw[]) (struct socket *, struct mbuf *,
7686 		const struct sadb_msghdr *) = {
7687 	NULL,			/* SADB_RESERVED */
7688 	key_api_getspi,		/* SADB_GETSPI */
7689 	key_api_update,		/* SADB_UPDATE */
7690 	key_api_add,		/* SADB_ADD */
7691 	key_api_delete,		/* SADB_DELETE */
7692 	key_api_get,		/* SADB_GET */
7693 	key_api_acquire,	/* SADB_ACQUIRE */
7694 	key_api_register,	/* SADB_REGISTER */
7695 	NULL,			/* SADB_EXPIRE */
7696 	key_api_flush,		/* SADB_FLUSH */
7697 	key_api_dump,		/* SADB_DUMP */
7698 	key_api_promisc,	/* SADB_X_PROMISC */
7699 	NULL,			/* SADB_X_PCHANGE */
7700 	key_api_spdadd,		/* SADB_X_SPDUPDATE */
7701 	key_api_spdadd,		/* SADB_X_SPDADD */
7702 	key_api_spddelete,	/* SADB_X_SPDDELETE */
7703 	key_api_spdget,		/* SADB_X_SPDGET */
7704 	NULL,			/* SADB_X_SPDACQUIRE */
7705 	key_api_spddump,	/* SADB_X_SPDDUMP */
7706 	key_api_spdflush,	/* SADB_X_SPDFLUSH */
7707 	key_api_spdadd,		/* SADB_X_SPDSETIDX */
7708 	NULL,			/* SADB_X_SPDEXPIRE */
7709 	key_api_spddelete2,	/* SADB_X_SPDDELETE2 */
7710 	key_api_nat_map,	/* SADB_X_NAT_T_NEW_MAPPING */
7711 };
7712 
7713 /*
7714  * parse sadb_msg buffer to process PFKEYv2,
7715  * and create a data to response if needed.
7716  * I think to be dealed with mbuf directly.
7717  * IN:
7718  *     msgp  : pointer to pointer to a received buffer pulluped.
7719  *             This is rewrited to response.
7720  *     so    : pointer to socket.
7721  * OUT:
7722  *    length for buffer to send to user process.
7723  */
7724 int
7725 key_parse(struct mbuf *m, struct socket *so)
7726 {
7727 	struct sadb_msg *msg;
7728 	struct sadb_msghdr mh;
7729 	u_int orglen;
7730 	int error;
7731 
7732 	KASSERT(m != NULL);
7733 	KASSERT(so != NULL);
7734 
7735 #if 0	/*kdebug_sadb assumes msg in linear buffer*/
7736 	if (KEYDEBUG_ON(KEYDEBUG_KEY_DUMP)) {
7737 		kdebug_sadb("passed sadb_msg", msg);
7738 	}
7739 #endif
7740 
7741 	if (m->m_len < sizeof(struct sadb_msg)) {
7742 		m = m_pullup(m, sizeof(struct sadb_msg));
7743 		if (!m)
7744 			return ENOBUFS;
7745 	}
7746 	msg = mtod(m, struct sadb_msg *);
7747 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
7748 
7749 	if ((m->m_flags & M_PKTHDR) == 0 ||
7750 	    m->m_pkthdr.len != orglen) {
7751 		IPSECLOG(LOG_DEBUG, "invalid message length.\n");
7752 		PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
7753 		error = EINVAL;
7754 		goto senderror;
7755 	}
7756 
7757 	if (msg->sadb_msg_version != PF_KEY_V2) {
7758 		IPSECLOG(LOG_DEBUG, "PF_KEY version %u is mismatched.\n",
7759 		    msg->sadb_msg_version);
7760 		PFKEY_STATINC(PFKEY_STAT_OUT_INVVER);
7761 		error = EINVAL;
7762 		goto senderror;
7763 	}
7764 
7765 	if (msg->sadb_msg_type > SADB_MAX) {
7766 		IPSECLOG(LOG_DEBUG, "invalid type %u is passed.\n",
7767 		    msg->sadb_msg_type);
7768 		PFKEY_STATINC(PFKEY_STAT_OUT_INVMSGTYPE);
7769 		error = EINVAL;
7770 		goto senderror;
7771 	}
7772 
7773 	/* for old-fashioned code - should be nuked */
7774 	if (m->m_pkthdr.len > MCLBYTES) {
7775 		m_freem(m);
7776 		return ENOBUFS;
7777 	}
7778 	if (m->m_next) {
7779 		struct mbuf *n;
7780 
7781 		n = key_alloc_mbuf_simple(m->m_pkthdr.len, M_WAITOK);
7782 
7783 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, void *));
7784 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
7785 		n->m_next = NULL;
7786 		m_freem(m);
7787 		m = n;
7788 	}
7789 
7790 	/* align the mbuf chain so that extensions are in contiguous region. */
7791 	error = key_align(m, &mh);
7792 	if (error)
7793 		return error;
7794 
7795 	if (m->m_next) {	/*XXX*/
7796 		m_freem(m);
7797 		return ENOBUFS;
7798 	}
7799 
7800 	msg = mh.msg;
7801 
7802 	/* check SA type */
7803 	switch (msg->sadb_msg_satype) {
7804 	case SADB_SATYPE_UNSPEC:
7805 		switch (msg->sadb_msg_type) {
7806 		case SADB_GETSPI:
7807 		case SADB_UPDATE:
7808 		case SADB_ADD:
7809 		case SADB_DELETE:
7810 		case SADB_GET:
7811 		case SADB_ACQUIRE:
7812 		case SADB_EXPIRE:
7813 			IPSECLOG(LOG_DEBUG,
7814 			    "must specify satype when msg type=%u.\n",
7815 			    msg->sadb_msg_type);
7816 			PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7817 			error = EINVAL;
7818 			goto senderror;
7819 		}
7820 		break;
7821 	case SADB_SATYPE_AH:
7822 	case SADB_SATYPE_ESP:
7823 	case SADB_X_SATYPE_IPCOMP:
7824 	case SADB_X_SATYPE_TCPSIGNATURE:
7825 		switch (msg->sadb_msg_type) {
7826 		case SADB_X_SPDADD:
7827 		case SADB_X_SPDDELETE:
7828 		case SADB_X_SPDGET:
7829 		case SADB_X_SPDDUMP:
7830 		case SADB_X_SPDFLUSH:
7831 		case SADB_X_SPDSETIDX:
7832 		case SADB_X_SPDUPDATE:
7833 		case SADB_X_SPDDELETE2:
7834 			IPSECLOG(LOG_DEBUG, "illegal satype=%u\n",
7835 			    msg->sadb_msg_type);
7836 			PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7837 			error = EINVAL;
7838 			goto senderror;
7839 		}
7840 		break;
7841 	case SADB_SATYPE_RSVP:
7842 	case SADB_SATYPE_OSPFV2:
7843 	case SADB_SATYPE_RIPV2:
7844 	case SADB_SATYPE_MIP:
7845 		IPSECLOG(LOG_DEBUG, "type %u isn't supported.\n",
7846 		    msg->sadb_msg_satype);
7847 		PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7848 		error = EOPNOTSUPP;
7849 		goto senderror;
7850 	case 1:	/* XXX: What does it do? */
7851 		if (msg->sadb_msg_type == SADB_X_PROMISC)
7852 			break;
7853 		/*FALLTHROUGH*/
7854 	default:
7855 		IPSECLOG(LOG_DEBUG, "invalid type %u is passed.\n",
7856 		    msg->sadb_msg_satype);
7857 		PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7858 		error = EINVAL;
7859 		goto senderror;
7860 	}
7861 
7862 	/* check field of upper layer protocol and address family */
7863 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL &&
7864 	    mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
7865 		const struct sadb_address *src0, *dst0;
7866 		const struct sockaddr *sa0, *da0;
7867 		u_int plen;
7868 
7869 		src0 = mh.ext[SADB_EXT_ADDRESS_SRC];
7870 		dst0 = mh.ext[SADB_EXT_ADDRESS_DST];
7871 		sa0 = key_msghdr_get_sockaddr(&mh, SADB_EXT_ADDRESS_SRC);
7872 		da0 = key_msghdr_get_sockaddr(&mh, SADB_EXT_ADDRESS_DST);
7873 
7874 		/* check upper layer protocol */
7875 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
7876 			IPSECLOG(LOG_DEBUG,
7877 			    "upper layer protocol mismatched src %u, dst %u.\n",
7878 			    src0->sadb_address_proto, dst0->sadb_address_proto);
7879 
7880 			goto invaddr;
7881 		}
7882 
7883 		/* check family */
7884 		if (sa0->sa_family != da0->sa_family) {
7885 			IPSECLOG(LOG_DEBUG,
7886 			    "address family mismatched src %u, dst %u.\n",
7887 			    sa0->sa_family, da0->sa_family);
7888 			goto invaddr;
7889 		}
7890 		if (sa0->sa_len != da0->sa_len) {
7891 			IPSECLOG(LOG_DEBUG,
7892 			    "address size mismatched src %u, dst %u.\n",
7893 			    sa0->sa_len, da0->sa_len);
7894 			goto invaddr;
7895 		}
7896 
7897 		switch (sa0->sa_family) {
7898 		case AF_INET:
7899 			if (sa0->sa_len != sizeof(struct sockaddr_in)) {
7900 				IPSECLOG(LOG_DEBUG,
7901 				    "address size mismatched %u != %zu.\n",
7902 				    sa0->sa_len, sizeof(struct sockaddr_in));
7903 				goto invaddr;
7904 			}
7905 			break;
7906 		case AF_INET6:
7907 			if (sa0->sa_len != sizeof(struct sockaddr_in6)) {
7908 				IPSECLOG(LOG_DEBUG,
7909 				    "address size mismatched %u != %zu.\n",
7910 				    sa0->sa_len, sizeof(struct sockaddr_in6));
7911 				goto invaddr;
7912 			}
7913 			break;
7914 		default:
7915 			IPSECLOG(LOG_DEBUG, "unsupported address family %u.\n",
7916 			    sa0->sa_family);
7917 			error = EAFNOSUPPORT;
7918 			goto senderror;
7919 		}
7920 		plen = key_sabits(sa0);
7921 
7922 		/* check max prefix length */
7923 		if (src0->sadb_address_prefixlen > plen ||
7924 		    dst0->sadb_address_prefixlen > plen) {
7925 			IPSECLOG(LOG_DEBUG, "illegal prefixlen.\n");
7926 			goto invaddr;
7927 		}
7928 
7929 		/*
7930 		 * prefixlen == 0 is valid because there can be a case when
7931 		 * all addresses are matched.
7932 		 */
7933 	}
7934 
7935 	if (msg->sadb_msg_type >= __arraycount(key_api_typesw) ||
7936 	    key_api_typesw[msg->sadb_msg_type] == NULL) {
7937 		PFKEY_STATINC(PFKEY_STAT_OUT_INVMSGTYPE);
7938 		error = EINVAL;
7939 		goto senderror;
7940 	}
7941 
7942 	return (*key_api_typesw[msg->sadb_msg_type])(so, m, &mh);
7943 
7944 invaddr:
7945 	error = EINVAL;
7946 senderror:
7947 	PFKEY_STATINC(PFKEY_STAT_OUT_INVADDR);
7948 	return key_senderror(so, m, error);
7949 }
7950 
7951 static int
7952 key_senderror(struct socket *so, struct mbuf *m, int code)
7953 {
7954 	struct sadb_msg *msg;
7955 
7956 	KASSERT(m->m_len >= sizeof(struct sadb_msg));
7957 
7958 	if (so == NULL) {
7959 		/*
7960 		 * This means the request comes from kernel.
7961 		 * As the request comes from kernel, it is unnecessary to
7962 		 * send message to userland. Just return errcode directly.
7963 		 */
7964 		m_freem(m);
7965 		return code;
7966 	}
7967 
7968 	msg = mtod(m, struct sadb_msg *);
7969 	msg->sadb_msg_errno = code;
7970 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
7971 }
7972 
7973 /*
7974  * set the pointer to each header into message buffer.
7975  * m will be freed on error.
7976  * XXX larger-than-MCLBYTES extension?
7977  */
7978 static int
7979 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
7980 {
7981 	struct mbuf *n;
7982 	struct sadb_ext *ext;
7983 	size_t off, end;
7984 	int extlen;
7985 	int toff;
7986 
7987 	KASSERT(m != NULL);
7988 	KASSERT(mhp != NULL);
7989 	KASSERT(m->m_len >= sizeof(struct sadb_msg));
7990 
7991 	/* initialize */
7992 	memset(mhp, 0, sizeof(*mhp));
7993 
7994 	mhp->msg = mtod(m, struct sadb_msg *);
7995 	mhp->ext[0] = mhp->msg;	/*XXX backward compat */
7996 
7997 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7998 	extlen = end;	/*just in case extlen is not updated*/
7999 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
8000 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
8001 		if (!n) {
8002 			/* m is already freed */
8003 			return ENOBUFS;
8004 		}
8005 		ext = (struct sadb_ext *)(mtod(n, char *) + toff);
8006 
8007 		/* set pointer */
8008 		switch (ext->sadb_ext_type) {
8009 		case SADB_EXT_SA:
8010 		case SADB_EXT_ADDRESS_SRC:
8011 		case SADB_EXT_ADDRESS_DST:
8012 		case SADB_EXT_ADDRESS_PROXY:
8013 		case SADB_EXT_LIFETIME_CURRENT:
8014 		case SADB_EXT_LIFETIME_HARD:
8015 		case SADB_EXT_LIFETIME_SOFT:
8016 		case SADB_EXT_KEY_AUTH:
8017 		case SADB_EXT_KEY_ENCRYPT:
8018 		case SADB_EXT_IDENTITY_SRC:
8019 		case SADB_EXT_IDENTITY_DST:
8020 		case SADB_EXT_SENSITIVITY:
8021 		case SADB_EXT_PROPOSAL:
8022 		case SADB_EXT_SUPPORTED_AUTH:
8023 		case SADB_EXT_SUPPORTED_ENCRYPT:
8024 		case SADB_EXT_SPIRANGE:
8025 		case SADB_X_EXT_POLICY:
8026 		case SADB_X_EXT_SA2:
8027 		case SADB_X_EXT_NAT_T_TYPE:
8028 		case SADB_X_EXT_NAT_T_SPORT:
8029 		case SADB_X_EXT_NAT_T_DPORT:
8030 		case SADB_X_EXT_NAT_T_OAI:
8031 		case SADB_X_EXT_NAT_T_OAR:
8032 		case SADB_X_EXT_NAT_T_FRAG:
8033 			/* duplicate check */
8034 			/*
8035 			 * XXX Are there duplication payloads of either
8036 			 * KEY_AUTH or KEY_ENCRYPT ?
8037 			 */
8038 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
8039 				IPSECLOG(LOG_DEBUG,
8040 				    "duplicate ext_type %u is passed.\n",
8041 				    ext->sadb_ext_type);
8042 				m_freem(m);
8043 				PFKEY_STATINC(PFKEY_STAT_OUT_DUPEXT);
8044 				return EINVAL;
8045 			}
8046 			break;
8047 		default:
8048 			IPSECLOG(LOG_DEBUG, "invalid ext_type %u is passed.\n",
8049 			    ext->sadb_ext_type);
8050 			m_freem(m);
8051 			PFKEY_STATINC(PFKEY_STAT_OUT_INVEXTTYPE);
8052 			return EINVAL;
8053 		}
8054 
8055 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
8056 
8057 		if (key_validate_ext(ext, extlen)) {
8058 			m_freem(m);
8059 			PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
8060 			return EINVAL;
8061 		}
8062 
8063 		n = m_pulldown(m, off, extlen, &toff);
8064 		if (!n) {
8065 			/* m is already freed */
8066 			return ENOBUFS;
8067 		}
8068 		ext = (struct sadb_ext *)(mtod(n, char *) + toff);
8069 
8070 		mhp->ext[ext->sadb_ext_type] = ext;
8071 		mhp->extoff[ext->sadb_ext_type] = off;
8072 		mhp->extlen[ext->sadb_ext_type] = extlen;
8073 	}
8074 
8075 	if (off != end) {
8076 		m_freem(m);
8077 		PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
8078 		return EINVAL;
8079 	}
8080 
8081 	return 0;
8082 }
8083 
8084 static int
8085 key_validate_ext(const struct sadb_ext *ext, int len)
8086 {
8087 	const struct sockaddr *sa;
8088 	enum { NONE, ADDR } checktype = NONE;
8089 	int baselen = 0;
8090 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
8091 
8092 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
8093 		return EINVAL;
8094 
8095 	/* if it does not match minimum/maximum length, bail */
8096 	if (ext->sadb_ext_type >= __arraycount(minsize) ||
8097 	    ext->sadb_ext_type >= __arraycount(maxsize))
8098 		return EINVAL;
8099 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
8100 		return EINVAL;
8101 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
8102 		return EINVAL;
8103 
8104 	/* more checks based on sadb_ext_type XXX need more */
8105 	switch (ext->sadb_ext_type) {
8106 	case SADB_EXT_ADDRESS_SRC:
8107 	case SADB_EXT_ADDRESS_DST:
8108 	case SADB_EXT_ADDRESS_PROXY:
8109 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
8110 		checktype = ADDR;
8111 		break;
8112 	case SADB_EXT_IDENTITY_SRC:
8113 	case SADB_EXT_IDENTITY_DST:
8114 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
8115 		    SADB_X_IDENTTYPE_ADDR) {
8116 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
8117 			checktype = ADDR;
8118 		} else
8119 			checktype = NONE;
8120 		break;
8121 	default:
8122 		checktype = NONE;
8123 		break;
8124 	}
8125 
8126 	switch (checktype) {
8127 	case NONE:
8128 		break;
8129 	case ADDR:
8130 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8131 		if (len < baselen + sal)
8132 			return EINVAL;
8133 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8134 			return EINVAL;
8135 		break;
8136 	}
8137 
8138 	return 0;
8139 }
8140 
8141 static int
8142 key_do_init(void)
8143 {
8144 	int i, error;
8145 
8146 	mutex_init(&key_misc.lock, MUTEX_DEFAULT, IPL_NONE);
8147 
8148 	mutex_init(&key_spd.lock, MUTEX_DEFAULT, IPL_NONE);
8149 	cv_init(&key_spd.cv_lc, "key_sp_lc");
8150 	key_spd.psz = pserialize_create();
8151 	cv_init(&key_spd.cv_psz, "key_sp_psz");
8152 	key_spd.psz_performing = false;
8153 
8154 	mutex_init(&key_sad.lock, MUTEX_DEFAULT, IPL_NONE);
8155 	cv_init(&key_sad.cv_lc, "key_sa_lc");
8156 	key_sad.psz = pserialize_create();
8157 	cv_init(&key_sad.cv_psz, "key_sa_psz");
8158 	key_sad.psz_performing = false;
8159 
8160 	pfkeystat_percpu = percpu_alloc(sizeof(uint64_t) * PFKEY_NSTATS);
8161 
8162 	callout_init(&key_timehandler_ch, CALLOUT_MPSAFE);
8163 	error = workqueue_create(&key_timehandler_wq, "key_timehandler",
8164 	    key_timehandler_work, NULL, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
8165 	if (error != 0)
8166 		panic("%s: workqueue_create failed (%d)\n", __func__, error);
8167 
8168 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8169 		PSLIST_INIT(&key_spd.splist[i]);
8170 	}
8171 
8172 	PSLIST_INIT(&key_spd.socksplist);
8173 
8174 	key_sad.sahlists = hashinit(SAHHASH_NHASH, HASH_PSLIST, true,
8175 	    &key_sad.sahlistmask);
8176 	key_sad.savlut = hashinit(SAVLUT_NHASH, HASH_PSLIST, true,
8177 	    &key_sad.savlutmask);
8178 
8179 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8180 		LIST_INIT(&key_misc.reglist[i]);
8181 	}
8182 
8183 #ifndef IPSEC_NONBLOCK_ACQUIRE
8184 	LIST_INIT(&key_misc.acqlist);
8185 #endif
8186 #ifdef notyet
8187 	LIST_INIT(&key_misc.spacqlist);
8188 #endif
8189 
8190 	/* system default */
8191 	ip4_def_policy.policy = IPSEC_POLICY_NONE;
8192 	ip4_def_policy.state = IPSEC_SPSTATE_ALIVE;
8193 	localcount_init(&ip4_def_policy.localcount);
8194 
8195 #ifdef INET6
8196 	ip6_def_policy.policy = IPSEC_POLICY_NONE;
8197 	ip6_def_policy.state = IPSEC_SPSTATE_ALIVE;
8198 	localcount_init(&ip6_def_policy.localcount);
8199 #endif
8200 
8201 	callout_reset(&key_timehandler_ch, hz, key_timehandler, NULL);
8202 
8203 	/* initialize key statistics */
8204 	keystat.getspi_count = 1;
8205 
8206 	aprint_verbose("IPsec: Initialized Security Association Processing.\n");
8207 
8208 	return (0);
8209 }
8210 
8211 void
8212 key_init(void)
8213 {
8214 	static ONCE_DECL(key_init_once);
8215 
8216 	sysctl_net_keyv2_setup(NULL);
8217 	sysctl_net_key_compat_setup(NULL);
8218 
8219 	RUN_ONCE(&key_init_once, key_do_init);
8220 
8221 	key_init_so();
8222 }
8223 
8224 /*
8225  * XXX: maybe This function is called after INBOUND IPsec processing.
8226  *
8227  * Special check for tunnel-mode packets.
8228  * We must make some checks for consistency between inner and outer IP header.
8229  *
8230  * xxx more checks to be provided
8231  */
8232 int
8233 key_checktunnelsanity(
8234     struct secasvar *sav,
8235     u_int family,
8236     void *src,
8237     void *dst
8238 )
8239 {
8240 
8241 	/* XXX: check inner IP header */
8242 
8243 	return 1;
8244 }
8245 
8246 #if 0
8247 #define hostnamelen	strlen(hostname)
8248 
8249 /*
8250  * Get FQDN for the host.
8251  * If the administrator configured hostname (by hostname(1)) without
8252  * domain name, returns nothing.
8253  */
8254 static const char *
8255 key_getfqdn(void)
8256 {
8257 	int i;
8258 	int hasdot;
8259 	static char fqdn[MAXHOSTNAMELEN + 1];
8260 
8261 	if (!hostnamelen)
8262 		return NULL;
8263 
8264 	/* check if it comes with domain name. */
8265 	hasdot = 0;
8266 	for (i = 0; i < hostnamelen; i++) {
8267 		if (hostname[i] == '.')
8268 			hasdot++;
8269 	}
8270 	if (!hasdot)
8271 		return NULL;
8272 
8273 	/* NOTE: hostname may not be NUL-terminated. */
8274 	memset(fqdn, 0, sizeof(fqdn));
8275 	memcpy(fqdn, hostname, hostnamelen);
8276 	fqdn[hostnamelen] = '\0';
8277 	return fqdn;
8278 }
8279 
8280 /*
8281  * get username@FQDN for the host/user.
8282  */
8283 static const char *
8284 key_getuserfqdn(void)
8285 {
8286 	const char *host;
8287 	static char userfqdn[MAXHOSTNAMELEN + MAXLOGNAME + 2];
8288 	struct proc *p = curproc;
8289 	char *q;
8290 
8291 	if (!p || !p->p_pgrp || !p->p_pgrp->pg_session)
8292 		return NULL;
8293 	if (!(host = key_getfqdn()))
8294 		return NULL;
8295 
8296 	/* NOTE: s_login may not be-NUL terminated. */
8297 	memset(userfqdn, 0, sizeof(userfqdn));
8298 	memcpy(userfqdn, Mp->p_pgrp->pg_session->s_login, AXLOGNAME);
8299 	userfqdn[MAXLOGNAME] = '\0';	/* safeguard */
8300 	q = userfqdn + strlen(userfqdn);
8301 	*q++ = '@';
8302 	memcpy(q, host, strlen(host));
8303 	q += strlen(host);
8304 	*q++ = '\0';
8305 
8306 	return userfqdn;
8307 }
8308 #endif
8309 
8310 /* record data transfer on SA, and update timestamps */
8311 void
8312 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8313 {
8314 	lifetime_counters_t *counters;
8315 
8316 	KASSERT(sav != NULL);
8317 	KASSERT(sav->lft_c != NULL);
8318 	KASSERT(m != NULL);
8319 
8320 	counters = percpu_getref(sav->lft_c_counters_percpu);
8321 
8322 	/*
8323 	 * XXX Currently, there is a difference of bytes size
8324 	 * between inbound and outbound processing.
8325 	 */
8326 	(*counters)[LIFETIME_COUNTER_BYTES] += m->m_pkthdr.len;
8327 	/* to check bytes lifetime is done in key_timehandler(). */
8328 
8329 	/*
8330 	 * We use the number of packets as the unit of
8331 	 * sadb_lifetime_allocations.  We increment the variable
8332 	 * whenever {esp,ah}_{in,out}put is called.
8333 	 */
8334 	(*counters)[LIFETIME_COUNTER_ALLOCATIONS]++;
8335 	/* XXX check for expires? */
8336 
8337 	percpu_putref(sav->lft_c_counters_percpu);
8338 
8339 	/*
8340 	 * NOTE: We record CURRENT sadb_lifetime_usetime by using wall clock,
8341 	 * in seconds.  HARD and SOFT lifetime are measured by the time
8342 	 * difference (again in seconds) from sadb_lifetime_usetime.
8343 	 *
8344 	 *	usetime
8345 	 *	v     expire   expire
8346 	 * -----+-----+--------+---> t
8347 	 *	<--------------> HARD
8348 	 *	<-----> SOFT
8349 	 */
8350 	sav->lft_c->sadb_lifetime_usetime = time_uptime;
8351 	/* XXX check for expires? */
8352 
8353 	return;
8354 }
8355 
8356 /* dumb version */
8357 void
8358 key_sa_routechange(struct sockaddr *dst)
8359 {
8360 	struct secashead *sah;
8361 	int s;
8362 
8363 	s = pserialize_read_enter();
8364 	SAHLIST_READER_FOREACH(sah) {
8365 		struct route *ro;
8366 		const struct sockaddr *sa;
8367 
8368 		key_sah_ref(sah);
8369 		pserialize_read_exit(s);
8370 
8371 		ro = &sah->sa_route;
8372 		sa = rtcache_getdst(ro);
8373 		if (sa != NULL && dst->sa_len == sa->sa_len &&
8374 		    memcmp(dst, sa, dst->sa_len) == 0)
8375 			rtcache_free(ro);
8376 
8377 		s = pserialize_read_enter();
8378 		key_sah_unref(sah);
8379 	}
8380 	pserialize_read_exit(s);
8381 
8382 	return;
8383 }
8384 
8385 static void
8386 key_sa_chgstate(struct secasvar *sav, u_int8_t state)
8387 {
8388 	struct secasvar *_sav;
8389 
8390 	ASSERT_SLEEPABLE();
8391 	KASSERT(mutex_owned(&key_sad.lock));
8392 
8393 	if (sav->state == state)
8394 		return;
8395 
8396 	key_unlink_sav(sav);
8397 	localcount_fini(&sav->localcount);
8398 	SAVLIST_ENTRY_DESTROY(sav);
8399 	key_init_sav(sav);
8400 
8401 	sav->state = state;
8402 	if (!SADB_SASTATE_USABLE_P(sav)) {
8403 		/* We don't need to care about the order */
8404 		SAVLIST_WRITER_INSERT_HEAD(sav->sah, state, sav);
8405 		return;
8406 	}
8407 	/*
8408 	 * Sort the list by lft_c->sadb_lifetime_addtime
8409 	 * in ascending order.
8410 	 */
8411 	SAVLIST_WRITER_FOREACH(_sav, sav->sah, state) {
8412 		if (_sav->lft_c->sadb_lifetime_addtime >
8413 		    sav->lft_c->sadb_lifetime_addtime) {
8414 			SAVLIST_WRITER_INSERT_BEFORE(_sav, sav);
8415 			break;
8416 		}
8417 	}
8418 	if (_sav == NULL) {
8419 		SAVLIST_WRITER_INSERT_TAIL(sav->sah, state, sav);
8420 	}
8421 
8422 	SAVLUT_WRITER_INSERT_HEAD(sav);
8423 
8424 	key_validate_savlist(sav->sah, state);
8425 }
8426 
8427 /* XXX too much? */
8428 static struct mbuf *
8429 key_alloc_mbuf(int l, int mflag)
8430 {
8431 	struct mbuf *m = NULL, *n;
8432 	int len, t;
8433 
8434 	KASSERT(mflag == M_NOWAIT || (mflag == M_WAITOK && !cpu_softintr_p()));
8435 
8436 	len = l;
8437 	while (len > 0) {
8438 		MGET(n, mflag, MT_DATA);
8439 		if (n && len > MLEN) {
8440 			MCLGET(n, mflag);
8441 			if ((n->m_flags & M_EXT) == 0) {
8442 				m_freem(n);
8443 				n = NULL;
8444 			}
8445 		}
8446 		if (!n) {
8447 			m_freem(m);
8448 			return NULL;
8449 		}
8450 
8451 		n->m_next = NULL;
8452 		n->m_len = 0;
8453 		n->m_len = M_TRAILINGSPACE(n);
8454 		/* use the bottom of mbuf, hoping we can prepend afterwards */
8455 		if (n->m_len > len) {
8456 			t = (n->m_len - len) & ~(sizeof(long) - 1);
8457 			n->m_data += t;
8458 			n->m_len = len;
8459 		}
8460 
8461 		len -= n->m_len;
8462 
8463 		if (m)
8464 			m_cat(m, n);
8465 		else
8466 			m = n;
8467 	}
8468 
8469 	return m;
8470 }
8471 
8472 static struct mbuf *
8473 key_setdump(u_int8_t req_satype, int *errorp, uint32_t pid)
8474 {
8475 	struct secashead *sah;
8476 	struct secasvar *sav;
8477 	u_int16_t proto;
8478 	u_int8_t satype;
8479 	u_int8_t state;
8480 	int cnt;
8481 	struct mbuf *m, *n;
8482 
8483 	KASSERT(mutex_owned(&key_sad.lock));
8484 
8485 	/* map satype to proto */
8486 	proto = key_satype2proto(req_satype);
8487 	if (proto == 0) {
8488 		*errorp = EINVAL;
8489 		return (NULL);
8490 	}
8491 
8492 	/* count sav entries to be sent to the userland. */
8493 	cnt = 0;
8494 	SAHLIST_WRITER_FOREACH(sah) {
8495 		if (req_satype != SADB_SATYPE_UNSPEC &&
8496 		    proto != sah->saidx.proto)
8497 			continue;
8498 
8499 		SASTATE_ANY_FOREACH(state) {
8500 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
8501 				cnt++;
8502 			}
8503 		}
8504 	}
8505 
8506 	if (cnt == 0) {
8507 		*errorp = ENOENT;
8508 		return (NULL);
8509 	}
8510 
8511 	/* send this to the userland, one at a time. */
8512 	m = NULL;
8513 	SAHLIST_WRITER_FOREACH(sah) {
8514 		if (req_satype != SADB_SATYPE_UNSPEC &&
8515 		    proto != sah->saidx.proto)
8516 			continue;
8517 
8518 		/* map proto to satype */
8519 		satype = key_proto2satype(sah->saidx.proto);
8520 		if (satype == 0) {
8521 			m_freem(m);
8522 			*errorp = EINVAL;
8523 			return (NULL);
8524 		}
8525 
8526 		SASTATE_ANY_FOREACH(state) {
8527 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
8528 				n = key_setdumpsa(sav, SADB_DUMP, satype,
8529 				    --cnt, pid);
8530 				if (!m)
8531 					m = n;
8532 				else
8533 					m_cat(m, n);
8534 			}
8535 		}
8536 	}
8537 
8538 	if (!m) {
8539 		*errorp = EINVAL;
8540 		return (NULL);
8541 	}
8542 
8543 	if ((m->m_flags & M_PKTHDR) != 0) {
8544 		m->m_pkthdr.len = 0;
8545 		for (n = m; n; n = n->m_next)
8546 			m->m_pkthdr.len += n->m_len;
8547 	}
8548 
8549 	*errorp = 0;
8550 	return (m);
8551 }
8552 
8553 static struct mbuf *
8554 key_setspddump(int *errorp, pid_t pid)
8555 {
8556 	struct secpolicy *sp;
8557 	int cnt;
8558 	u_int dir;
8559 	struct mbuf *m, *n;
8560 
8561 	KASSERT(mutex_owned(&key_spd.lock));
8562 
8563 	/* search SPD entry and get buffer size. */
8564 	cnt = 0;
8565 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
8566 		SPLIST_WRITER_FOREACH(sp, dir) {
8567 			cnt++;
8568 		}
8569 	}
8570 
8571 	if (cnt == 0) {
8572 		*errorp = ENOENT;
8573 		return (NULL);
8574 	}
8575 
8576 	m = NULL;
8577 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
8578 		SPLIST_WRITER_FOREACH(sp, dir) {
8579 			--cnt;
8580 			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, pid);
8581 
8582 			if (!m)
8583 				m = n;
8584 			else {
8585 				m->m_pkthdr.len += n->m_pkthdr.len;
8586 				m_cat(m, n);
8587 			}
8588 		}
8589 	}
8590 
8591 	*errorp = 0;
8592 	return (m);
8593 }
8594 
8595 int
8596 key_get_used(void) {
8597 	return !SPLIST_READER_EMPTY(IPSEC_DIR_INBOUND) ||
8598 	    !SPLIST_READER_EMPTY(IPSEC_DIR_OUTBOUND) ||
8599 	    !SOCKSPLIST_READER_EMPTY();
8600 }
8601 
8602 void
8603 key_update_used(void)
8604 {
8605 	switch (ipsec_enabled) {
8606 	default:
8607 	case 0:
8608 #ifdef notyet
8609 		/* XXX: racy */
8610 		ipsec_used = 0;
8611 #endif
8612 		break;
8613 	case 1:
8614 #ifndef notyet
8615 		/* XXX: racy */
8616 		if (!ipsec_used)
8617 #endif
8618 		ipsec_used = key_get_used();
8619 		break;
8620 	case 2:
8621 		ipsec_used = 1;
8622 		break;
8623 	}
8624 }
8625 
8626 static inline void
8627 key_savlut_writer_insert_head(struct secasvar *sav)
8628 {
8629 	uint32_t hash_key;
8630 	uint32_t hash;
8631 
8632 	KASSERT(mutex_owned(&key_sad.lock));
8633 	KASSERT(!sav->savlut_added);
8634 
8635 	if (sav->sah->saidx.proto == IPPROTO_IPCOMP)
8636 		hash_key = sav->alg_comp;
8637 	else
8638 		hash_key = sav->spi;
8639 
8640 	hash = key_savluthash(&sav->sah->saidx.dst.sa,
8641 	    sav->sah->saidx.proto, hash_key, key_sad.savlutmask);
8642 
8643 	PSLIST_WRITER_INSERT_HEAD(&key_sad.savlut[hash], sav,
8644 	    pslist_entry_savlut);
8645 	sav->savlut_added = true;
8646 }
8647 
8648 /*
8649  * Calculate hash using protocol, source address,
8650  * and destination address included in saidx.
8651  */
8652 static inline uint32_t
8653 key_saidxhash(const struct secasindex *saidx, u_long mask)
8654 {
8655 	uint32_t hash32;
8656 	const struct sockaddr_in *sin;
8657 	const struct sockaddr_in6 *sin6;
8658 
8659 	hash32 = saidx->proto;
8660 
8661 	switch (saidx->src.sa.sa_family) {
8662 	case AF_INET:
8663 		sin = &saidx->src.sin;
8664 		hash32 = hash32_buf(&sin->sin_addr,
8665 		    sizeof(sin->sin_addr), hash32);
8666 		sin = &saidx->dst.sin;
8667 		hash32 = hash32_buf(&sin->sin_addr,
8668 		    sizeof(sin->sin_addr), hash32 << 1);
8669 		break;
8670 	case AF_INET6:
8671 		sin6 = &saidx->src.sin6;
8672 		hash32 = hash32_buf(&sin6->sin6_addr,
8673 		    sizeof(sin6->sin6_addr), hash32);
8674 		sin6 = &saidx->dst.sin6;
8675 		hash32 = hash32_buf(&sin6->sin6_addr,
8676 		    sizeof(sin6->sin6_addr), hash32 << 1);
8677 		break;
8678 	default:
8679 		hash32 = 0;
8680 		break;
8681 	}
8682 
8683 	return hash32 & mask;
8684 }
8685 
8686 /*
8687  * Calculate hash using destination address, protocol,
8688  * and spi. Those parameter depend on the search of
8689  * key_lookup_sa().
8690  */
8691 static uint32_t
8692 key_savluthash(const struct sockaddr *dst, uint32_t proto,
8693     uint32_t spi, u_long mask)
8694 {
8695 	uint32_t hash32;
8696 	const struct sockaddr_in *sin;
8697 	const struct sockaddr_in6 *sin6;
8698 
8699 	hash32 = hash32_buf(&proto, sizeof(proto), spi);
8700 
8701 	switch(dst->sa_family) {
8702 	case AF_INET:
8703 		sin = satocsin(dst);
8704 		hash32 = hash32_buf(&sin->sin_addr,
8705 		    sizeof(sin->sin_addr), hash32);
8706 		break;
8707 	case AF_INET6:
8708 		sin6 = satocsin6(dst);
8709 		hash32 = hash32_buf(&sin6->sin6_addr,
8710 		    sizeof(sin6->sin6_addr), hash32);
8711 		break;
8712 	default:
8713 		hash32 = 0;
8714 	}
8715 
8716 	return hash32 & mask;
8717 }
8718 
8719 static int
8720 sysctl_net_key_dumpsa(SYSCTLFN_ARGS)
8721 {
8722 	struct mbuf *m, *n;
8723 	int err2 = 0;
8724 	char *p, *ep;
8725 	size_t len;
8726 	int error;
8727 
8728 	if (newp)
8729 		return (EPERM);
8730 	if (namelen != 1)
8731 		return (EINVAL);
8732 
8733 	mutex_enter(&key_sad.lock);
8734 	m = key_setdump(name[0], &error, l->l_proc->p_pid);
8735 	mutex_exit(&key_sad.lock);
8736 	if (!m)
8737 		return (error);
8738 	if (!oldp)
8739 		*oldlenp = m->m_pkthdr.len;
8740 	else {
8741 		p = oldp;
8742 		if (*oldlenp < m->m_pkthdr.len) {
8743 			err2 = ENOMEM;
8744 			ep = p + *oldlenp;
8745 		} else {
8746 			*oldlenp = m->m_pkthdr.len;
8747 			ep = p + m->m_pkthdr.len;
8748 		}
8749 		for (n = m; n; n = n->m_next) {
8750 			len =  (ep - p < n->m_len) ?
8751 				ep - p : n->m_len;
8752 			error = copyout(mtod(n, const void *), p, len);
8753 			p += len;
8754 			if (error)
8755 				break;
8756 		}
8757 		if (error == 0)
8758 			error = err2;
8759 	}
8760 	m_freem(m);
8761 
8762 	return (error);
8763 }
8764 
8765 static int
8766 sysctl_net_key_dumpsp(SYSCTLFN_ARGS)
8767 {
8768 	struct mbuf *m, *n;
8769 	int err2 = 0;
8770 	char *p, *ep;
8771 	size_t len;
8772 	int error;
8773 
8774 	if (newp)
8775 		return (EPERM);
8776 	if (namelen != 0)
8777 		return (EINVAL);
8778 
8779 	mutex_enter(&key_spd.lock);
8780 	m = key_setspddump(&error, l->l_proc->p_pid);
8781 	mutex_exit(&key_spd.lock);
8782 	if (!m)
8783 		return (error);
8784 	if (!oldp)
8785 		*oldlenp = m->m_pkthdr.len;
8786 	else {
8787 		p = oldp;
8788 		if (*oldlenp < m->m_pkthdr.len) {
8789 			err2 = ENOMEM;
8790 			ep = p + *oldlenp;
8791 		} else {
8792 			*oldlenp = m->m_pkthdr.len;
8793 			ep = p + m->m_pkthdr.len;
8794 		}
8795 		for (n = m; n; n = n->m_next) {
8796 			len = (ep - p < n->m_len) ? ep - p : n->m_len;
8797 			error = copyout(mtod(n, const void *), p, len);
8798 			p += len;
8799 			if (error)
8800 				break;
8801 		}
8802 		if (error == 0)
8803 			error = err2;
8804 	}
8805 	m_freem(m);
8806 
8807 	return (error);
8808 }
8809 
8810 /*
8811  * Create sysctl tree for native IPSEC key knobs, originally
8812  * under name "net.keyv2"  * with MIB number { CTL_NET, PF_KEY_V2. }.
8813  * However, sysctl(8) never checked for nodes under { CTL_NET, PF_KEY_V2 };
8814  * and in any case the part of our sysctl namespace used for dumping the
8815  * SPD and SA database  *HAS* to be compatible with the KAME sysctl
8816  * namespace, for API reasons.
8817  *
8818  * Pending a consensus on the right way  to fix this, add a level of
8819  * indirection in how we number the `native' IPSEC key nodes;
8820  * and (as requested by Andrew Brown)  move registration of the
8821  * KAME-compatible names  to a separate function.
8822  */
8823 #if 0
8824 #  define IPSEC_PFKEY PF_KEY_V2
8825 # define IPSEC_PFKEY_NAME "keyv2"
8826 #else
8827 #  define IPSEC_PFKEY PF_KEY
8828 # define IPSEC_PFKEY_NAME "key"
8829 #endif
8830 
8831 static int
8832 sysctl_net_key_stats(SYSCTLFN_ARGS)
8833 {
8834 
8835 	return (NETSTAT_SYSCTL(pfkeystat_percpu, PFKEY_NSTATS));
8836 }
8837 
8838 static void
8839 sysctl_net_keyv2_setup(struct sysctllog **clog)
8840 {
8841 
8842 	sysctl_createv(clog, 0, NULL, NULL,
8843 		       CTLFLAG_PERMANENT,
8844 		       CTLTYPE_NODE, IPSEC_PFKEY_NAME, NULL,
8845 		       NULL, 0, NULL, 0,
8846 		       CTL_NET, IPSEC_PFKEY, CTL_EOL);
8847 
8848 	sysctl_createv(clog, 0, NULL, NULL,
8849 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8850 		       CTLTYPE_INT, "debug", NULL,
8851 		       NULL, 0, &key_debug_level, 0,
8852 		       CTL_NET, IPSEC_PFKEY, KEYCTL_DEBUG_LEVEL, CTL_EOL);
8853 	sysctl_createv(clog, 0, NULL, NULL,
8854 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8855 		       CTLTYPE_INT, "spi_try", NULL,
8856 		       NULL, 0, &key_spi_trycnt, 0,
8857 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_TRY, CTL_EOL);
8858 	sysctl_createv(clog, 0, NULL, NULL,
8859 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8860 		       CTLTYPE_INT, "spi_min_value", NULL,
8861 		       NULL, 0, &key_spi_minval, 0,
8862 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_MIN_VALUE, CTL_EOL);
8863 	sysctl_createv(clog, 0, NULL, NULL,
8864 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8865 		       CTLTYPE_INT, "spi_max_value", NULL,
8866 		       NULL, 0, &key_spi_maxval, 0,
8867 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_MAX_VALUE, CTL_EOL);
8868 	sysctl_createv(clog, 0, NULL, NULL,
8869 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8870 		       CTLTYPE_INT, "random_int", NULL,
8871 		       NULL, 0, &key_int_random, 0,
8872 		       CTL_NET, IPSEC_PFKEY, KEYCTL_RANDOM_INT, CTL_EOL);
8873 	sysctl_createv(clog, 0, NULL, NULL,
8874 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8875 		       CTLTYPE_INT, "larval_lifetime", NULL,
8876 		       NULL, 0, &key_larval_lifetime, 0,
8877 		       CTL_NET, IPSEC_PFKEY, KEYCTL_LARVAL_LIFETIME, CTL_EOL);
8878 	sysctl_createv(clog, 0, NULL, NULL,
8879 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8880 		       CTLTYPE_INT, "blockacq_count", NULL,
8881 		       NULL, 0, &key_blockacq_count, 0,
8882 		       CTL_NET, IPSEC_PFKEY, KEYCTL_BLOCKACQ_COUNT, CTL_EOL);
8883 	sysctl_createv(clog, 0, NULL, NULL,
8884 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8885 		       CTLTYPE_INT, "blockacq_lifetime", NULL,
8886 		       NULL, 0, &key_blockacq_lifetime, 0,
8887 		       CTL_NET, IPSEC_PFKEY, KEYCTL_BLOCKACQ_LIFETIME, CTL_EOL);
8888 	sysctl_createv(clog, 0, NULL, NULL,
8889 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8890 		       CTLTYPE_INT, "esp_keymin", NULL,
8891 		       NULL, 0, &ipsec_esp_keymin, 0,
8892 		       CTL_NET, IPSEC_PFKEY, KEYCTL_ESP_KEYMIN, CTL_EOL);
8893 	sysctl_createv(clog, 0, NULL, NULL,
8894 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8895 		       CTLTYPE_INT, "prefered_oldsa", NULL,
8896 		       NULL, 0, &key_prefered_oldsa, 0,
8897 		       CTL_NET, PF_KEY, KEYCTL_PREFERED_OLDSA, CTL_EOL);
8898 	sysctl_createv(clog, 0, NULL, NULL,
8899 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8900 		       CTLTYPE_INT, "esp_auth", NULL,
8901 		       NULL, 0, &ipsec_esp_auth, 0,
8902 		       CTL_NET, IPSEC_PFKEY, KEYCTL_ESP_AUTH, CTL_EOL);
8903 	sysctl_createv(clog, 0, NULL, NULL,
8904 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8905 		       CTLTYPE_INT, "ah_keymin", NULL,
8906 		       NULL, 0, &ipsec_ah_keymin, 0,
8907 		       CTL_NET, IPSEC_PFKEY, KEYCTL_AH_KEYMIN, CTL_EOL);
8908 	sysctl_createv(clog, 0, NULL, NULL,
8909 		       CTLFLAG_PERMANENT,
8910 		       CTLTYPE_STRUCT, "stats",
8911 		       SYSCTL_DESCR("PF_KEY statistics"),
8912 		       sysctl_net_key_stats, 0, NULL, 0,
8913 		       CTL_NET, IPSEC_PFKEY, CTL_CREATE, CTL_EOL);
8914 }
8915 
8916 /*
8917  * Register sysctl names used by setkey(8). For historical reasons,
8918  * and to share a single API, these names appear under { CTL_NET, PF_KEY }
8919  * for both IPSEC and KAME IPSEC.
8920  */
8921 static void
8922 sysctl_net_key_compat_setup(struct sysctllog **clog)
8923 {
8924 
8925 	sysctl_createv(clog, 0, NULL, NULL,
8926 		       CTLFLAG_PERMANENT,
8927 		       CTLTYPE_NODE, "key", NULL,
8928 		       NULL, 0, NULL, 0,
8929 		       CTL_NET, PF_KEY, CTL_EOL);
8930 
8931 	/* Register the net.key.dump{sa,sp} nodes used by setkey(8). */
8932 	sysctl_createv(clog, 0, NULL, NULL,
8933 		       CTLFLAG_PERMANENT,
8934 		       CTLTYPE_STRUCT, "dumpsa", NULL,
8935 		       sysctl_net_key_dumpsa, 0, NULL, 0,
8936 		       CTL_NET, PF_KEY, KEYCTL_DUMPSA, CTL_EOL);
8937 	sysctl_createv(clog, 0, NULL, NULL,
8938 		       CTLFLAG_PERMANENT,
8939 		       CTLTYPE_STRUCT, "dumpsp", NULL,
8940 		       sysctl_net_key_dumpsp, 0, NULL, 0,
8941 		       CTL_NET, PF_KEY, KEYCTL_DUMPSP, CTL_EOL);
8942 }
8943