xref: /netbsd-src/sys/netipsec/key.c (revision 5299f67511feea2eccdeab8fd06f5b28df504b37)
1 /*	$NetBSD: key.c,v 1.285 2024/09/02 18:56:20 andvar 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 1.285 2024/09/02 18:56:20 andvar 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 #include <sys/xcall.h>
77 
78 #include <net/if.h>
79 #include <net/route.h>
80 
81 #include <netinet/in.h>
82 #include <netinet/in_systm.h>
83 #include <netinet/ip.h>
84 #include <netinet/in_var.h>
85 #ifdef INET
86 #include <netinet/ip_var.h>
87 #endif
88 
89 #ifdef INET6
90 #include <netinet/ip6.h>
91 #include <netinet6/in6_var.h>
92 #include <netinet6/ip6_var.h>
93 #endif /* INET6 */
94 
95 #ifdef INET
96 #include <netinet/in_pcb.h>
97 #endif
98 #ifdef INET6
99 #include <netinet6/in6_pcb.h>
100 #endif /* INET6 */
101 
102 #include <net/pfkeyv2.h>
103 #include <netipsec/keydb.h>
104 #include <netipsec/key.h>
105 #include <netipsec/keysock.h>
106 #include <netipsec/key_debug.h>
107 
108 #include <netipsec/ipsec.h>
109 #ifdef INET6
110 #include <netipsec/ipsec6.h>
111 #endif
112 #include <netipsec/ipsec_private.h>
113 
114 #include <netipsec/xform.h>
115 #include <netipsec/ipcomp.h>
116 
117 #define FULLMASK	0xffu
118 #define	_BITS(bytes)	((bytes) << 3)
119 
120 #define PORT_NONE	0
121 #define PORT_LOOSE	1
122 #define PORT_STRICT	2
123 
124 #ifndef SAHHASH_NHASH
125 #define SAHHASH_NHASH		128
126 #endif
127 
128 #ifndef SAVLUT_NHASH
129 #define SAVLUT_NHASH		128
130 #endif
131 
132 percpu_t *pfkeystat_percpu;
133 
134 /*
135  * Note on SA reference counting:
136  * - SAs that are not in DEAD state will have (total external reference + 1)
137  *   following value in reference count field.  they cannot be freed and are
138  *   referenced from SA header.
139  * - SAs that are in DEAD state will have (total external reference)
140  *   in reference count field.  they are ready to be freed.  reference from
141  *   SA header will be removed in key_delsav(), when the reference count
142  *   field hits 0 (= no external reference other than from SA header.
143  */
144 
145 u_int32_t key_debug_level = 0;
146 static u_int key_spi_trycnt = 1000;
147 static u_int32_t key_spi_minval = 0x100;
148 static u_int32_t key_spi_maxval = 0x0fffffff;	/* XXX */
149 static u_int32_t policy_id = 0;
150 static u_int key_int_random = 60;	/*interval to initialize randseed,1(m)*/
151 static u_int key_larval_lifetime = 30;	/* interval to expire acquiring, 30(s)*/
152 static int key_blockacq_count = 10;	/* counter for blocking SADB_ACQUIRE.*/
153 static int key_blockacq_lifetime = 20;	/* lifetime for blocking SADB_ACQUIRE.*/
154 static int key_prefered_oldsa = 0;	/* prefered old sa rather than new sa.*/
155 
156 static u_int32_t acq_seq = 0;
157 
158 /*
159  * Locking order: there is no order for now; it means that any locks aren't
160  * overlapped.
161  */
162 /*
163  * Locking notes on SPD:
164  * - Modifications to the key_spd.splist must be done with holding key_spd.lock
165  *   which is a adaptive mutex
166  * - Read accesses to the key_spd.splist must be in pserialize(9) read sections
167  * - SP's lifetime is managed by localcount(9)
168  * - An SP that has been inserted to the key_spd.splist is initially referenced
169  *   by none, i.e., a reference from the key_spd.splist isn't counted
170  * - When an SP is being destroyed, we change its state as DEAD, wait for
171  *   references to the SP to be released, and then deallocate the SP
172  *   (see key_unlink_sp)
173  * - Getting an SP
174  *   - Normally we get an SP from the key_spd.splist (see key_lookup_sp_byspidx)
175  *     - Must iterate the list and increment the reference count of a found SP
176  *       (by key_sp_ref) in a pserialize read section
177  *   - We can gain another reference from a held SP only if we check its state
178  *     and take its reference in a pserialize read section
179  *     (see esp_output for example)
180  *   - We may get an SP from an SP cache. See below
181  *   - A gotten SP must be released after use by KEY_SP_UNREF (key_sp_unref)
182  * - Updating member variables of an SP
183  *   - Most member variables of an SP are immutable
184  *   - Only sp->state and sp->lastused can be changed
185  *   - sp->state of an SP is updated only when destroying it under key_spd.lock
186  * - SP caches
187  *   - SPs can be cached in PCBs
188  *   - The lifetime of the caches is controlled by the global generation counter
189  *     (ipsec_spdgen)
190  *   - The global counter value is stored when an SP is cached
191  *   - If the stored value is different from the global counter then the cache
192  *     is considered invalidated
193  *   - The counter is incremented when an SP is being destroyed
194  *   - So checking the generation and taking a reference to an SP should be
195  *     in a pserialize read section
196  *   - Note that caching doesn't increment the reference counter of an SP
197  * - SPs in sockets
198  *   - Userland programs can set a policy to a socket by
199  *     setsockopt(IP_IPSEC_POLICY)
200  *   - Such policies (SPs) are set to a socket (PCB) and also inserted to
201  *     the key_spd.socksplist list (not the key_spd.splist)
202  *   - Such a policy is destroyed when a corresponding socket is destroed,
203  *     however, a socket can be destroyed in softint so we cannot destroy
204  *     it directly instead we just mark it DEAD and delay the destruction
205  *     until GC by the timer
206  * - SP origin
207  *   - SPs can be created by both userland programs and kernel components.
208  *     The SPs created in kernel must not be removed by userland programs,
209  *     although the SPs can be read by userland programs.
210  */
211 /*
212  * Locking notes on SAD:
213  * - Data structures
214  *   - SAs are managed by the list called key_sad.sahlists and sav lists of
215  *     sah entries
216  *     - An sav is supposed to be an SA from a viewpoint of users
217  *   - A sah has sav lists for each SA state
218  *   - Multiple saves with the same saidx can exist
219  *     - Only one entry has MATURE state and others should be DEAD
220  *     - DEAD entries are just ignored from searching
221  *   - All sav whose state is MATURE or DYING are registered to the lookup
222  *     table called key_sad.savlut in addition to the savlists.
223  *     - The table is used to search an sav without use of saidx.
224  * - Modifications to the key_sad.sahlists, sah.savlist and key_sad.savlut
225  *   must be done with holding key_sad.lock which is a adaptive mutex
226  * - Read accesses to the key_sad.sahlists, sah.savlist and key_sad.savlut
227  *   must be in pserialize(9) read sections
228  * - sah's lifetime is managed by localcount(9)
229  * - Getting an sah entry
230  *   - We get an sah from the key_sad.sahlists
231  *     - Must iterate the list and increment the reference count of a found sah
232  *       (by key_sah_ref) in a pserialize read section
233  *   - A gotten sah must be released after use by key_sah_unref
234  * - An sah is destroyed when its state become DEAD and no sav is
235  *   listed to the sah
236  *   - The destruction is done only in the timer (see key_timehandler_sad)
237  * - sav's lifetime is managed by localcount(9)
238  * - Getting an sav entry
239  *   - First get an sah by saidx and get an sav from either of sah's savlists
240  *     - Must iterate the list and increment the reference count of a found sav
241  *       (by key_sa_ref) in a pserialize read section
242  *   - We can gain another reference from a held SA only if we check its state
243  *     and take its reference in a pserialize read section
244  *     (see esp_output for example)
245  *   - A gotten sav must be released after use by key_sa_unref
246  * - An sav is destroyed when its state become DEAD
247  */
248 /*
249  * Locking notes on misc data:
250  * - All lists of key_misc are protected by key_misc.lock
251  *   - key_misc.lock must be held even for read accesses
252  */
253 
254 /* SPD */
255 static struct {
256 	kmutex_t lock;
257 	kcondvar_t cv_lc;
258 	struct pslist_head splist[IPSEC_DIR_MAX];
259 	/*
260 	 * The list has SPs that are set to a socket via
261 	 * setsockopt(IP_IPSEC_POLICY) from userland. See ipsec_set_policy.
262 	 */
263 	struct pslist_head socksplist;
264 
265 	pserialize_t psz;
266 	kcondvar_t cv_psz;
267 	bool psz_performing;
268 } key_spd __cacheline_aligned;
269 
270 /* SAD */
271 static struct {
272 	kmutex_t lock;
273 	kcondvar_t cv_lc;
274 	struct pslist_head *sahlists;
275 	u_long sahlistmask;
276 	struct pslist_head *savlut;
277 	u_long savlutmask;
278 
279 	pserialize_t psz;
280 	kcondvar_t cv_psz;
281 	bool psz_performing;
282 } key_sad __cacheline_aligned;
283 
284 /* Misc data */
285 static struct {
286 	kmutex_t lock;
287 	/* registed list */
288 	LIST_HEAD(_reglist, secreg) reglist[SADB_SATYPE_MAX + 1];
289 #ifndef IPSEC_NONBLOCK_ACQUIRE
290 	/* acquiring list */
291 	LIST_HEAD(_acqlist, secacq) acqlist;
292 #endif
293 #ifdef notyet
294 	/* SP acquiring list */
295 	LIST_HEAD(_spacqlist, secspacq) spacqlist;
296 #endif
297 } key_misc __cacheline_aligned;
298 
299 /* Macros for key_spd.splist */
300 #define SPLIST_ENTRY_INIT(sp)						\
301 	PSLIST_ENTRY_INIT((sp), pslist_entry)
302 #define SPLIST_ENTRY_DESTROY(sp)					\
303 	PSLIST_ENTRY_DESTROY((sp), pslist_entry)
304 #define SPLIST_WRITER_REMOVE(sp)					\
305 	PSLIST_WRITER_REMOVE((sp), pslist_entry)
306 #define SPLIST_READER_EMPTY(dir)					\
307 	(PSLIST_READER_FIRST(&key_spd.splist[(dir)], struct secpolicy,	\
308 	                     pslist_entry) == NULL)
309 #define SPLIST_READER_FOREACH(sp, dir)					\
310 	PSLIST_READER_FOREACH((sp), &key_spd.splist[(dir)],		\
311 	                      struct secpolicy, pslist_entry)
312 #define SPLIST_WRITER_FOREACH(sp, dir)					\
313 	PSLIST_WRITER_FOREACH((sp), &key_spd.splist[(dir)],		\
314 	                      struct secpolicy, pslist_entry)
315 #define SPLIST_WRITER_INSERT_AFTER(sp, new)				\
316 	PSLIST_WRITER_INSERT_AFTER((sp), (new), pslist_entry)
317 #define SPLIST_WRITER_EMPTY(dir)					\
318 	(PSLIST_WRITER_FIRST(&key_spd.splist[(dir)], struct secpolicy,	\
319 	                     pslist_entry) == NULL)
320 #define SPLIST_WRITER_INSERT_HEAD(dir, sp)				\
321 	PSLIST_WRITER_INSERT_HEAD(&key_spd.splist[(dir)], (sp),		\
322 	                          pslist_entry)
323 #define SPLIST_WRITER_NEXT(sp)						\
324 	PSLIST_WRITER_NEXT((sp), struct secpolicy, pslist_entry)
325 #define SPLIST_WRITER_INSERT_TAIL(dir, new)				\
326 	do {								\
327 		if (SPLIST_WRITER_EMPTY((dir))) {			\
328 			SPLIST_WRITER_INSERT_HEAD((dir), (new));	\
329 		} else {						\
330 			struct secpolicy *__sp;				\
331 			SPLIST_WRITER_FOREACH(__sp, (dir)) {		\
332 				if (SPLIST_WRITER_NEXT(__sp) == NULL) {	\
333 					SPLIST_WRITER_INSERT_AFTER(__sp,\
334 					    (new));			\
335 					break;				\
336 				}					\
337 			}						\
338 		}							\
339 	} while (0)
340 
341 /* Macros for key_spd.socksplist */
342 #define SOCKSPLIST_WRITER_FOREACH(sp)					\
343 	PSLIST_WRITER_FOREACH((sp), &key_spd.socksplist,		\
344 	                      struct secpolicy,	pslist_entry)
345 #define SOCKSPLIST_READER_EMPTY()					\
346 	(PSLIST_READER_FIRST(&key_spd.socksplist, struct secpolicy,	\
347 	                     pslist_entry) == NULL)
348 
349 /* Macros for key_sad.sahlist */
350 #define SAHLIST_ENTRY_INIT(sah)						\
351 	PSLIST_ENTRY_INIT((sah), pslist_entry)
352 #define SAHLIST_ENTRY_DESTROY(sah)					\
353 	PSLIST_ENTRY_DESTROY((sah), pslist_entry)
354 #define SAHLIST_WRITER_REMOVE(sah)					\
355 	PSLIST_WRITER_REMOVE((sah), pslist_entry)
356 #define SAHLIST_READER_FOREACH(sah)					\
357 	for(int _i_sah = 0; _i_sah <= key_sad.sahlistmask; _i_sah++)	\
358 		PSLIST_READER_FOREACH((sah), &key_sad.sahlists[_i_sah],	\
359 		                      struct secashead, pslist_entry)
360 #define SAHLIST_READER_FOREACH_SAIDX(sah, saidx)			\
361 	PSLIST_READER_FOREACH((sah),					\
362 	    &key_sad.sahlists[key_saidxhash((saidx),			\
363 	                       key_sad.sahlistmask)],			\
364 	    struct secashead, pslist_entry)
365 #define SAHLIST_WRITER_FOREACH(sah)					\
366 	for(int _i_sah = 0; _i_sah <= key_sad.sahlistmask; _i_sah++)	\
367 		PSLIST_WRITER_FOREACH((sah), &key_sad.sahlists[_i_sah],	\
368 		                     struct secashead, pslist_entry)
369 #define SAHLIST_WRITER_INSERT_HEAD(sah)					\
370 	PSLIST_WRITER_INSERT_HEAD(					\
371 	    &key_sad.sahlists[key_saidxhash(&(sah)->saidx,		\
372 	                      key_sad.sahlistmask)],	\
373 	    (sah), pslist_entry)
374 
375 /* Macros for key_sad.sahlist#savlist */
376 #define SAVLIST_ENTRY_INIT(sav)						\
377 	PSLIST_ENTRY_INIT((sav), pslist_entry)
378 #define SAVLIST_ENTRY_DESTROY(sav)					\
379 	PSLIST_ENTRY_DESTROY((sav), pslist_entry)
380 #define SAVLIST_READER_FIRST(sah, state)				\
381 	PSLIST_READER_FIRST(&(sah)->savlist[(state)], struct secasvar,	\
382 	                    pslist_entry)
383 #define SAVLIST_WRITER_REMOVE(sav)					\
384 	PSLIST_WRITER_REMOVE((sav), pslist_entry)
385 #define SAVLIST_READER_FOREACH(sav, sah, state)				\
386 	PSLIST_READER_FOREACH((sav), &(sah)->savlist[(state)],		\
387 	                      struct secasvar, pslist_entry)
388 #define SAVLIST_WRITER_FOREACH(sav, sah, state)				\
389 	PSLIST_WRITER_FOREACH((sav), &(sah)->savlist[(state)],		\
390 	                      struct secasvar, pslist_entry)
391 #define SAVLIST_WRITER_INSERT_BEFORE(sav, new)				\
392 	PSLIST_WRITER_INSERT_BEFORE((sav), (new), pslist_entry)
393 #define SAVLIST_WRITER_INSERT_AFTER(sav, new)				\
394 	PSLIST_WRITER_INSERT_AFTER((sav), (new), pslist_entry)
395 #define SAVLIST_WRITER_EMPTY(sah, state)				\
396 	(PSLIST_WRITER_FIRST(&(sah)->savlist[(state)], struct secasvar,	\
397 	                     pslist_entry) == NULL)
398 #define SAVLIST_WRITER_INSERT_HEAD(sah, state, sav)			\
399 	PSLIST_WRITER_INSERT_HEAD(&(sah)->savlist[(state)], (sav),	\
400 	                          pslist_entry)
401 #define SAVLIST_WRITER_NEXT(sav)					\
402 	PSLIST_WRITER_NEXT((sav), struct secasvar, pslist_entry)
403 #define SAVLIST_WRITER_INSERT_TAIL(sah, state, new)			\
404 	do {								\
405 		if (SAVLIST_WRITER_EMPTY((sah), (state))) {		\
406 			SAVLIST_WRITER_INSERT_HEAD((sah), (state), (new));\
407 		} else {						\
408 			struct secasvar *__sav;				\
409 			SAVLIST_WRITER_FOREACH(__sav, (sah), (state)) {	\
410 				if (SAVLIST_WRITER_NEXT(__sav) == NULL) {\
411 					SAVLIST_WRITER_INSERT_AFTER(__sav,\
412 					    (new));			\
413 					break;				\
414 				}					\
415 			}						\
416 		}							\
417 	} while (0)
418 #define SAVLIST_READER_NEXT(sav)					\
419 	PSLIST_READER_NEXT((sav), struct secasvar, pslist_entry)
420 
421 /* Macros for key_sad.savlut */
422 #define SAVLUT_ENTRY_INIT(sav)						\
423 	PSLIST_ENTRY_INIT((sav), pslist_entry_savlut)
424 #define SAVLUT_READER_FOREACH(sav, dst, proto, hash_key)		\
425 	PSLIST_READER_FOREACH((sav),					\
426 	&key_sad.savlut[key_savluthash(dst, proto, hash_key,		\
427 	                  key_sad.savlutmask)],				\
428 	struct secasvar, pslist_entry_savlut)
429 #define SAVLUT_WRITER_INSERT_HEAD(sav)					\
430 	key_savlut_writer_insert_head((sav))
431 #define SAVLUT_WRITER_REMOVE(sav)					\
432 	do {								\
433 		if (!(sav)->savlut_added)				\
434 			break;						\
435 		PSLIST_WRITER_REMOVE((sav), pslist_entry_savlut);	\
436 		(sav)->savlut_added = false;				\
437 	} while(0)
438 
439 /* search order for SAs */
440 	/*
441 	 * This order is important because we must select the oldest SA
442 	 * for outbound processing.  For inbound, This is not important.
443 	 */
444 static const u_int saorder_state_valid_prefer_old[] = {
445 	SADB_SASTATE_DYING, SADB_SASTATE_MATURE,
446 };
447 static const u_int saorder_state_valid_prefer_new[] = {
448 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING,
449 };
450 
451 static const u_int saorder_state_alive[] = {
452 	/* except DEAD */
453 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING, SADB_SASTATE_LARVAL
454 };
455 static const u_int saorder_state_any[] = {
456 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING,
457 	SADB_SASTATE_LARVAL, SADB_SASTATE_DEAD
458 };
459 
460 #define SASTATE_ALIVE_FOREACH(s)				\
461 	for (int _i = 0;					\
462 	    _i < __arraycount(saorder_state_alive) ?		\
463 	    (s) = saorder_state_alive[_i], true : false;	\
464 	    _i++)
465 #define SASTATE_ANY_FOREACH(s)					\
466 	for (int _i = 0;					\
467 	    _i < __arraycount(saorder_state_any) ?		\
468 	    (s) = saorder_state_any[_i], true : false;		\
469 	    _i++)
470 #define SASTATE_USABLE_FOREACH(s)				\
471 	for (int _i = 0;					\
472 	    _i < __arraycount(saorder_state_valid_prefer_new) ?	\
473 	    (s) = saorder_state_valid_prefer_new[_i],		\
474 	    true : false;					\
475 	    _i++)
476 
477 static const int minsize[] = {
478 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
479 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
480 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
481 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
482 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
483 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_SRC */
484 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_DST */
485 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_PROXY */
486 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_AUTH */
487 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_ENCRYPT */
488 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_SRC */
489 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_DST */
490 	sizeof(struct sadb_sens),	/* SADB_EXT_SENSITIVITY */
491 	sizeof(struct sadb_prop),	/* SADB_EXT_PROPOSAL */
492 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_AUTH */
493 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_ENCRYPT */
494 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
495 	0,				/* SADB_X_EXT_KMPRIVATE */
496 	sizeof(struct sadb_x_policy),	/* SADB_X_EXT_POLICY */
497 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
498 	sizeof(struct sadb_x_nat_t_type),	/* SADB_X_EXT_NAT_T_TYPE */
499 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_SPORT */
500 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_DPORT */
501 	sizeof(struct sadb_address),		/* SADB_X_EXT_NAT_T_OAI */
502 	sizeof(struct sadb_address),		/* SADB_X_EXT_NAT_T_OAR */
503 	sizeof(struct sadb_x_nat_t_frag),	/* SADB_X_EXT_NAT_T_FRAG */
504 };
505 static const int maxsize[] = {
506 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
507 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
508 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
509 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
510 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
511 	0,				/* SADB_EXT_ADDRESS_SRC */
512 	0,				/* SADB_EXT_ADDRESS_DST */
513 	0,				/* SADB_EXT_ADDRESS_PROXY */
514 	0,				/* SADB_EXT_KEY_AUTH */
515 	0,				/* SADB_EXT_KEY_ENCRYPT */
516 	0,				/* SADB_EXT_IDENTITY_SRC */
517 	0,				/* SADB_EXT_IDENTITY_DST */
518 	0,				/* SADB_EXT_SENSITIVITY */
519 	0,				/* SADB_EXT_PROPOSAL */
520 	0,				/* SADB_EXT_SUPPORTED_AUTH */
521 	0,				/* SADB_EXT_SUPPORTED_ENCRYPT */
522 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
523 	0,				/* SADB_X_EXT_KMPRIVATE */
524 	0,				/* SADB_X_EXT_POLICY */
525 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
526 	sizeof(struct sadb_x_nat_t_type),	/* SADB_X_EXT_NAT_T_TYPE */
527 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_SPORT */
528 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_DPORT */
529 	0,					/* SADB_X_EXT_NAT_T_OAI */
530 	0,					/* SADB_X_EXT_NAT_T_OAR */
531 	sizeof(struct sadb_x_nat_t_frag),	/* SADB_X_EXT_NAT_T_FRAG */
532 };
533 
534 static int ipsec_esp_keymin = 256;
535 static int ipsec_esp_auth = 0;
536 static int ipsec_ah_keymin = 128;
537 static bool ipsec_allow_different_idtype = false;
538 
539 #ifdef SYSCTL_DECL
540 SYSCTL_DECL(_net_key);
541 #endif
542 
543 #ifdef SYSCTL_INT
544 SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,	CTLFLAG_RW, \
545 	&key_debug_level,	0,	"");
546 
547 /* max count of trial for the decision of spi value */
548 SYSCTL_INT(_net_key, KEYCTL_SPI_TRY,		spi_trycnt,	CTLFLAG_RW, \
549 	&key_spi_trycnt,	0,	"");
550 
551 /* minimum spi value to allocate automatically. */
552 SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE,	spi_minval,	CTLFLAG_RW, \
553 	&key_spi_minval,	0,	"");
554 
555 /* maximun spi value to allocate automatically. */
556 SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE,	spi_maxval,	CTLFLAG_RW, \
557 	&key_spi_maxval,	0,	"");
558 
559 /* interval to initialize randseed */
560 SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT,	int_random,	CTLFLAG_RW, \
561 	&key_int_random,	0,	"");
562 
563 /* lifetime for larval SA */
564 SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME,	larval_lifetime, CTLFLAG_RW, \
565 	&key_larval_lifetime,	0,	"");
566 
567 /* counter for blocking to send SADB_ACQUIRE to IKEd */
568 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT,	blockacq_count,	CTLFLAG_RW, \
569 	&key_blockacq_count,	0,	"");
570 
571 /* lifetime for blocking to send SADB_ACQUIRE to IKEd */
572 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME,	blockacq_lifetime, CTLFLAG_RW, \
573 	&key_blockacq_lifetime,	0,	"");
574 
575 /* ESP auth */
576 SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH,	esp_auth, CTLFLAG_RW, \
577 	&ipsec_esp_auth,	0,	"");
578 
579 /* minimum ESP key length */
580 SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN,	esp_keymin, CTLFLAG_RW, \
581 	&ipsec_esp_keymin,	0,	"");
582 
583 /* minimum AH key length */
584 SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN,	ah_keymin, CTLFLAG_RW, \
585 	&ipsec_ah_keymin,	0,	"");
586 
587 /* perfered old SA rather than new SA */
588 SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA,	prefered_oldsa, CTLFLAG_RW,\
589 	&key_prefered_oldsa,	0,	"");
590 #endif /* SYSCTL_INT */
591 
592 #define __LIST_CHAINED(elm) \
593 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
594 #define LIST_INSERT_TAIL(head, elm, type, field) \
595 do {\
596 	struct type *curelm = LIST_FIRST(head); \
597 	if (curelm == NULL) {\
598 		LIST_INSERT_HEAD(head, elm, field); \
599 	} else { \
600 		while (LIST_NEXT(curelm, field)) \
601 			curelm = LIST_NEXT(curelm, field);\
602 		LIST_INSERT_AFTER(curelm, elm, field);\
603 	}\
604 } while (0)
605 
606 #define KEY_CHKSASTATE(head, sav) \
607 /* do */ { \
608 	if ((head) != (sav)) {						\
609 		IPSECLOG(LOG_DEBUG,					\
610 		    "state mismatched (TREE=%d SA=%d)\n",		\
611 		    (head), (sav));					\
612 		continue;						\
613 	}								\
614 } /* while (0) */
615 
616 #define KEY_CHKSPDIR(head, sp) \
617 do { \
618 	if ((head) != (sp)) {						\
619 		IPSECLOG(LOG_DEBUG,					\
620 		    "direction mismatched (TREE=%d SP=%d), anyway continue.\n",\
621 		    (head), (sp));					\
622 	}								\
623 } while (0)
624 
625 /*
626  * set parameters into secasindex buffer.
627  * Must allocate secasindex buffer before calling this function.
628  */
629 static int
630 key_setsecasidx(int, int, int, const struct sockaddr *,
631     const struct sockaddr *, struct secasindex *);
632 
633 /* key statistics */
634 struct _keystat {
635 	u_long getspi_count; /* the avarage of count to try to get new SPI */
636 } keystat;
637 
638 static void
639 key_init_spidx_bymsghdr(struct secpolicyindex *, const struct sadb_msghdr *);
640 
641 static const struct sockaddr *
642 key_msghdr_get_sockaddr(const struct sadb_msghdr *mhp, int idx)
643 {
644 
645 	return PFKEY_ADDR_SADDR(mhp->ext[idx]);
646 }
647 
648 static void
649 key_fill_replymsg(struct mbuf *m, int seq)
650 {
651 	struct sadb_msg *msg;
652 
653 	KASSERT(m->m_len >= sizeof(*msg));
654 
655 	msg = mtod(m, struct sadb_msg *);
656 	msg->sadb_msg_errno = 0;
657 	msg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
658 	if (seq != 0)
659 		msg->sadb_msg_seq = seq;
660 }
661 
662 #if 0
663 static void key_freeso(struct socket *);
664 static void key_freesp_so(struct secpolicy **);
665 #endif
666 static struct secpolicy *key_getsp (const struct secpolicyindex *);
667 static struct secpolicy *key_getspbyid (u_int32_t);
668 static struct secpolicy *key_lookup_and_remove_sp(const struct secpolicyindex *, bool);
669 static struct secpolicy *key_lookupbyid_and_remove_sp(u_int32_t, bool);
670 static void key_destroy_sp(struct secpolicy *);
671 static struct mbuf *key_gather_mbuf (struct mbuf *,
672 	const struct sadb_msghdr *, int, int, ...);
673 static int key_api_spdadd(struct socket *, struct mbuf *,
674 	const struct sadb_msghdr *);
675 static u_int32_t key_getnewspid (void);
676 static int key_api_spddelete(struct socket *, struct mbuf *,
677 	const struct sadb_msghdr *);
678 static int key_api_spddelete2(struct socket *, struct mbuf *,
679 	const struct sadb_msghdr *);
680 static int key_api_spdget(struct socket *, struct mbuf *,
681 	const struct sadb_msghdr *);
682 static int key_api_spdflush(struct socket *, struct mbuf *,
683 	const struct sadb_msghdr *);
684 static int key_api_spddump(struct socket *, struct mbuf *,
685 	const struct sadb_msghdr *);
686 static struct mbuf * key_setspddump (int *errorp, pid_t);
687 static struct mbuf * key_setspddump_chain (int *errorp, int *lenp, pid_t pid);
688 static int key_api_nat_map(struct socket *, struct mbuf *,
689 	const struct sadb_msghdr *);
690 static struct mbuf *key_setdumpsp (struct secpolicy *,
691 	u_int8_t, u_int32_t, pid_t);
692 static u_int key_getspreqmsglen (const struct secpolicy *);
693 static int key_spdexpire (struct secpolicy *);
694 static struct secashead *key_newsah (const struct secasindex *);
695 static void key_unlink_sah(struct secashead *);
696 static void key_destroy_sah(struct secashead *);
697 static bool key_sah_has_sav(struct secashead *);
698 static void key_sah_ref(struct secashead *);
699 static void key_sah_unref(struct secashead *);
700 static void key_init_sav(struct secasvar *);
701 static void key_wait_sav(struct secasvar *);
702 static void key_destroy_sav(struct secasvar *);
703 static struct secasvar *key_newsav(struct mbuf *,
704 	const struct sadb_msghdr *, int *, int, const char*, int);
705 #define	KEY_NEWSAV(m, sadb, e, proto)				\
706 	key_newsav(m, sadb, e, proto, __func__, __LINE__)
707 static void key_delsav (struct secasvar *);
708 static struct secashead *key_getsah(const struct secasindex *, int);
709 static struct secashead *key_getsah_ref(const struct secasindex *, int);
710 static bool key_checkspidup(const struct secasindex *, u_int32_t);
711 static struct secasvar *key_getsavbyspi (struct secashead *, u_int32_t);
712 static int key_setsaval (struct secasvar *, struct mbuf *,
713 	const struct sadb_msghdr *);
714 static void key_freesaval(struct secasvar *);
715 static int key_init_xform(struct secasvar *);
716 static void key_clear_xform(struct secasvar *);
717 static struct mbuf *key_setdumpsa (struct secasvar *, u_int8_t,
718 	u_int8_t, u_int32_t, u_int32_t);
719 static struct mbuf *key_setsadbxport (u_int16_t, u_int16_t);
720 static struct mbuf *key_setsadbxtype (u_int16_t);
721 static struct mbuf *key_setsadbxfrag (u_int16_t);
722 static void key_porttosaddr (union sockaddr_union *, u_int16_t);
723 static int key_checksalen (const union sockaddr_union *);
724 static struct mbuf *key_setsadbmsg (u_int8_t, u_int16_t, u_int8_t,
725 	u_int32_t, pid_t, u_int16_t, int);
726 static struct mbuf *key_setsadbsa (struct secasvar *);
727 static struct mbuf *key_setsadbaddr(u_int16_t,
728 	const struct sockaddr *, u_int8_t, u_int16_t, int);
729 #if 0
730 static struct mbuf *key_setsadbident (u_int16_t, u_int16_t, void *,
731 	int, u_int64_t);
732 #endif
733 static struct mbuf *key_setsadbxsa2 (u_int8_t, u_int32_t, u_int16_t);
734 static struct mbuf *key_setsadbxpolicy (u_int16_t, u_int8_t,
735 	u_int32_t, int);
736 static void *key_newbuf (const void *, u_int);
737 #ifdef INET6
738 static int key_ismyaddr6 (const struct sockaddr_in6 *);
739 #endif
740 
741 static void sysctl_net_keyv2_setup(struct sysctllog **);
742 static void sysctl_net_key_compat_setup(struct sysctllog **);
743 
744 /* flags for key_saidx_match() */
745 #define CMP_HEAD	1	/* protocol, addresses. */
746 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
747 #define CMP_REQID	3	/* additionally HEAD, reaid. */
748 #define CMP_EXACTLY	4	/* all elements. */
749 static int key_saidx_match(const struct secasindex *,
750     const struct secasindex *, int);
751 
752 static int key_sockaddr_match(const struct sockaddr *,
753     const struct sockaddr *, int);
754 static int key_bb_match_withmask(const void *, const void *, u_int);
755 static u_int16_t key_satype2proto (u_int8_t);
756 static u_int8_t key_proto2satype (u_int16_t);
757 
758 static int key_spidx_match_exactly(const struct secpolicyindex *,
759     const struct secpolicyindex *);
760 static int key_spidx_match_withmask(const struct secpolicyindex *,
761     const struct secpolicyindex *);
762 
763 static int key_api_getspi(struct socket *, struct mbuf *,
764 	const struct sadb_msghdr *);
765 static u_int32_t key_do_getnewspi (const struct sadb_spirange *,
766 					const struct secasindex *);
767 static int key_handle_natt_info (struct secasvar *,
768 				     const struct sadb_msghdr *);
769 static int key_set_natt_ports (union sockaddr_union *,
770 			 	union sockaddr_union *,
771 				const struct sadb_msghdr *);
772 static int key_api_update(struct socket *, struct mbuf *,
773 	const struct sadb_msghdr *);
774 #ifdef IPSEC_DOSEQCHECK
775 static struct secasvar *key_getsavbyseq (struct secashead *, u_int32_t);
776 #endif
777 static int key_api_add(struct socket *, struct mbuf *,
778 	const struct sadb_msghdr *);
779 static int key_setident (struct secashead *, struct mbuf *,
780 	const struct sadb_msghdr *);
781 static struct mbuf *key_getmsgbuf_x1 (struct mbuf *,
782 	const struct sadb_msghdr *);
783 static int key_api_delete(struct socket *, struct mbuf *,
784 	const struct sadb_msghdr *);
785 static int key_api_get(struct socket *, struct mbuf *,
786 	const struct sadb_msghdr *);
787 
788 static void key_getcomb_setlifetime (struct sadb_comb *);
789 static struct mbuf *key_getcomb_esp(int);
790 static struct mbuf *key_getcomb_ah(int);
791 static struct mbuf *key_getcomb_ipcomp(int);
792 static struct mbuf *key_getprop(const struct secasindex *, int);
793 
794 static int key_acquire(const struct secasindex *, const struct secpolicy *,
795 	    int);
796 static int key_acquire_sendup_mbuf_later(struct mbuf *);
797 static void key_acquire_sendup_pending_mbuf(void);
798 #ifndef IPSEC_NONBLOCK_ACQUIRE
799 static struct secacq *key_newacq (const struct secasindex *);
800 static struct secacq *key_getacq (const struct secasindex *);
801 static struct secacq *key_getacqbyseq (u_int32_t);
802 #endif
803 #ifdef notyet
804 static struct secspacq *key_newspacq (const struct secpolicyindex *);
805 static struct secspacq *key_getspacq (const struct secpolicyindex *);
806 #endif
807 static int key_api_acquire(struct socket *, struct mbuf *,
808 	const struct sadb_msghdr *);
809 static int key_api_register(struct socket *, struct mbuf *,
810 	const struct sadb_msghdr *);
811 static int key_expire (struct secasvar *);
812 static int key_api_flush(struct socket *, struct mbuf *,
813 	const struct sadb_msghdr *);
814 static struct mbuf *key_setdump_chain (u_int8_t req_satype, int *errorp,
815 	int *lenp, pid_t pid);
816 static int key_api_dump(struct socket *, struct mbuf *,
817 	const struct sadb_msghdr *);
818 static int key_api_promisc(struct socket *, struct mbuf *,
819 	const struct sadb_msghdr *);
820 static int key_senderror (struct socket *, struct mbuf *, int);
821 static int key_validate_ext (const struct sadb_ext *, int);
822 static int key_align (struct mbuf *, struct sadb_msghdr *);
823 #if 0
824 static const char *key_getfqdn (void);
825 static const char *key_getuserfqdn (void);
826 #endif
827 static void key_sa_chgstate (struct secasvar *, u_int8_t);
828 
829 static struct mbuf *key_alloc_mbuf(int, int);
830 static struct mbuf *key_alloc_mbuf_simple(int, int);
831 
832 static void key_timehandler(void *);
833 static void key_timehandler_work(struct work *, void *);
834 static struct callout	key_timehandler_ch;
835 static struct workqueue	*key_timehandler_wq;
836 static struct work	key_timehandler_wk;
837 
838 static inline void
839     key_savlut_writer_insert_head(struct secasvar *sav);
840 static inline uint32_t
841     key_saidxhash(const struct secasindex *, u_long);
842 static inline uint32_t
843     key_savluthash(const struct sockaddr *,
844     uint32_t, uint32_t, u_long);
845 
846 /*
847  * Utilities for percpu counters for sadb_lifetime_allocations and
848  * sadb_lifetime_bytes.
849  */
850 #define LIFETIME_COUNTER_ALLOCATIONS	0
851 #define LIFETIME_COUNTER_BYTES		1
852 #define LIFETIME_COUNTER_SIZE		2
853 
854 typedef uint64_t lifetime_counters_t[LIFETIME_COUNTER_SIZE];
855 
856 static void
857 key_sum_lifetime_counters(void *p, void *arg, struct cpu_info *ci __unused)
858 {
859 	lifetime_counters_t *one = p;
860 	lifetime_counters_t *sum = arg;
861 
862 	(*sum)[LIFETIME_COUNTER_ALLOCATIONS] += (*one)[LIFETIME_COUNTER_ALLOCATIONS];
863 	(*sum)[LIFETIME_COUNTER_BYTES] += (*one)[LIFETIME_COUNTER_BYTES];
864 }
865 
866 u_int
867 key_sp_refcnt(const struct secpolicy *sp)
868 {
869 
870 	/* FIXME */
871 	return 0;
872 }
873 
874 void
875 key_sp_touch(struct secpolicy *sp)
876 {
877 
878 	sp->lastused = time_uptime;
879 }
880 
881 static void
882 key_spd_pserialize_perform(void)
883 {
884 
885 	KASSERT(mutex_owned(&key_spd.lock));
886 
887 	while (key_spd.psz_performing)
888 		cv_wait(&key_spd.cv_psz, &key_spd.lock);
889 	key_spd.psz_performing = true;
890 	mutex_exit(&key_spd.lock);
891 
892 	pserialize_perform(key_spd.psz);
893 
894 	mutex_enter(&key_spd.lock);
895 	key_spd.psz_performing = false;
896 	cv_broadcast(&key_spd.cv_psz);
897 }
898 
899 /*
900  * Remove the sp from the key_spd.splist and wait for references to the sp
901  * to be released. key_spd.lock must be held.
902  */
903 static void
904 key_unlink_sp(struct secpolicy *sp)
905 {
906 
907 	KASSERT(mutex_owned(&key_spd.lock));
908 
909 	sp->state = IPSEC_SPSTATE_DEAD;
910 	SPLIST_WRITER_REMOVE(sp);
911 
912 	/* Invalidate all cached SPD pointers in the PCBs. */
913 	ipsec_invalpcbcacheall();
914 
915 	KDASSERT(mutex_ownable(softnet_lock));
916 	key_spd_pserialize_perform();
917 
918 	localcount_drain(&sp->localcount, &key_spd.cv_lc, &key_spd.lock);
919 }
920 
921 /*
922  * Return 0 when there are known to be no SP's for the specified
923  * direction.  Otherwise return 1.  This is used by IPsec code
924  * to optimize performance.
925  */
926 int
927 key_havesp(u_int dir)
928 {
929 	return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ?
930 		!SPLIST_READER_EMPTY(dir) : 1);
931 }
932 
933 /* %%% IPsec policy management */
934 /*
935  * allocating a SP for OUTBOUND or INBOUND packet.
936  * Must call key_freesp() later.
937  * OUT:	NULL:	not found
938  *	others:	found and return the pointer.
939  */
940 struct secpolicy *
941 key_lookup_sp_byspidx(const struct secpolicyindex *spidx,
942     u_int dir, const char* where, int tag)
943 {
944 	struct secpolicy *sp;
945 	int s;
946 
947 	KASSERT(spidx != NULL);
948 	KASSERTMSG(IPSEC_DIR_IS_INOROUT(dir), "invalid direction %u", dir);
949 
950 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
951 
952 	/* get a SP entry */
953 	if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
954 		kdebug_secpolicyindex("objects", spidx);
955 	}
956 
957 	s = pserialize_read_enter();
958 	SPLIST_READER_FOREACH(sp, dir) {
959 		if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
960 			kdebug_secpolicyindex("in SPD", &sp->spidx);
961 		}
962 
963 		if (sp->state == IPSEC_SPSTATE_DEAD)
964 			continue;
965 		if (key_spidx_match_withmask(&sp->spidx, spidx))
966 			goto found;
967 	}
968 	sp = NULL;
969 found:
970 	if (sp) {
971 		/* sanity check */
972 		KEY_CHKSPDIR(sp->spidx.dir, dir);
973 
974 		/* found a SPD entry */
975 		key_sp_touch(sp);
976 		key_sp_ref(sp, where, tag);
977 	}
978 	pserialize_read_exit(s);
979 
980 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
981 	    "DP return SP:%p (ID=%u) refcnt %u\n",
982 	    sp, sp ? sp->id : 0, key_sp_refcnt(sp));
983 	return sp;
984 }
985 
986 /*
987  * return a policy that matches this particular inbound packet.
988  * XXX slow
989  */
990 struct secpolicy *
991 key_gettunnel(const struct sockaddr *osrc,
992 	      const struct sockaddr *odst,
993 	      const struct sockaddr *isrc,
994 	      const struct sockaddr *idst,
995 	      const char* where, int tag)
996 {
997 	struct secpolicy *sp;
998 	const int dir = IPSEC_DIR_INBOUND;
999 	int s;
1000 	struct ipsecrequest *r1, *r2, *p;
1001 	struct secpolicyindex spidx;
1002 
1003 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
1004 
1005 	if (isrc->sa_family != idst->sa_family) {
1006 		IPSECLOG(LOG_ERR,
1007 		    "address family mismatched src %u, dst %u.\n",
1008 		    isrc->sa_family, idst->sa_family);
1009 		sp = NULL;
1010 		goto done;
1011 	}
1012 
1013 	s = pserialize_read_enter();
1014 	SPLIST_READER_FOREACH(sp, dir) {
1015 		if (sp->state == IPSEC_SPSTATE_DEAD)
1016 			continue;
1017 
1018 		r1 = r2 = NULL;
1019 		for (p = sp->req; p; p = p->next) {
1020 			if (p->saidx.mode != IPSEC_MODE_TUNNEL)
1021 				continue;
1022 
1023 			r1 = r2;
1024 			r2 = p;
1025 
1026 			if (!r1) {
1027 				/* here we look at address matches only */
1028 				spidx = sp->spidx;
1029 				if (isrc->sa_len > sizeof(spidx.src) ||
1030 				    idst->sa_len > sizeof(spidx.dst))
1031 					continue;
1032 				memcpy(&spidx.src, isrc, isrc->sa_len);
1033 				memcpy(&spidx.dst, idst, idst->sa_len);
1034 				if (!key_spidx_match_withmask(&sp->spidx, &spidx))
1035 					continue;
1036 			} else {
1037 				if (!key_sockaddr_match(&r1->saidx.src.sa, isrc, PORT_NONE) ||
1038 				    !key_sockaddr_match(&r1->saidx.dst.sa, idst, PORT_NONE))
1039 					continue;
1040 			}
1041 
1042 			if (!key_sockaddr_match(&r2->saidx.src.sa, osrc, PORT_NONE) ||
1043 			    !key_sockaddr_match(&r2->saidx.dst.sa, odst, PORT_NONE))
1044 				continue;
1045 
1046 			goto found;
1047 		}
1048 	}
1049 	sp = NULL;
1050 found:
1051 	if (sp) {
1052 		key_sp_touch(sp);
1053 		key_sp_ref(sp, where, tag);
1054 	}
1055 	pserialize_read_exit(s);
1056 done:
1057 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1058 	    "DP return SP:%p (ID=%u) refcnt %u\n",
1059 	    sp, sp ? sp->id : 0, key_sp_refcnt(sp));
1060 	return sp;
1061 }
1062 
1063 /*
1064  * allocating an SA entry for an *OUTBOUND* packet.
1065  * checking each request entries in SP, and acquire an SA if need.
1066  * OUT:	0: there are valid requests.
1067  *	ENOENT: policy may be valid, but SA with REQUIRE is on acquiring.
1068  */
1069 int
1070 key_checkrequest(const struct ipsecrequest *isr, const struct secasindex *saidx,
1071     struct secasvar **ret)
1072 {
1073 	u_int level;
1074 	int error;
1075 	struct secasvar *sav;
1076 
1077 	KASSERT(isr != NULL);
1078 	KASSERTMSG(saidx->mode == IPSEC_MODE_TRANSPORT ||
1079 	    saidx->mode == IPSEC_MODE_TUNNEL,
1080 	    "unexpected policy %u", saidx->mode);
1081 
1082 	/* get current level */
1083 	level = ipsec_get_reqlevel(isr);
1084 
1085 	/*
1086 	 * XXX guard against protocol callbacks from the crypto
1087 	 * thread as they reference ipsecrequest.sav which we
1088 	 * temporarily null out below.  Need to rethink how we
1089 	 * handle bundled SA's in the callback thread.
1090 	 */
1091 
1092 	sav = key_lookup_sa_bysaidx(saidx);
1093 	if (sav != NULL) {
1094 		*ret = sav;
1095 		return 0;
1096 	}
1097 
1098 	/* there is no SA */
1099 	error = key_acquire(saidx, isr->sp, M_NOWAIT);
1100 	if (error != 0) {
1101 		/* XXX What should I do ? */
1102 		IPSECLOG(LOG_DEBUG, "error %d returned from key_acquire.\n",
1103 		    error);
1104 		return error;
1105 	}
1106 
1107 	if (level != IPSEC_LEVEL_REQUIRE) {
1108 		/* XXX sigh, the interface to this routine is botched */
1109 		*ret = NULL;
1110 		return 0;
1111 	} else {
1112 		return ENOENT;
1113 	}
1114 }
1115 
1116 /*
1117  * looking up a SA for policy entry from SAD.
1118  * NOTE: searching SAD of aliving state.
1119  * OUT:	NULL:	not found.
1120  *	others:	found and return the pointer.
1121  */
1122 struct secasvar *
1123 key_lookup_sa_bysaidx(const struct secasindex *saidx)
1124 {
1125 	struct secashead *sah;
1126 	struct secasvar *sav = NULL;
1127 	u_int stateidx, state;
1128 	const u_int *saorder_state_valid;
1129 	int arraysize;
1130 	int s;
1131 
1132 	s = pserialize_read_enter();
1133 	sah = key_getsah(saidx, CMP_MODE_REQID);
1134 	if (sah == NULL)
1135 		goto out;
1136 
1137 	/*
1138 	 * search a valid state list for outbound packet.
1139 	 * This search order is important.
1140 	 */
1141 	if (key_prefered_oldsa) {
1142 		saorder_state_valid = saorder_state_valid_prefer_old;
1143 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_old);
1144 	} else {
1145 		saorder_state_valid = saorder_state_valid_prefer_new;
1146 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_new);
1147 	}
1148 
1149 	/* search valid state */
1150 	for (stateidx = 0;
1151 	     stateidx < arraysize;
1152 	     stateidx++) {
1153 
1154 		state = saorder_state_valid[stateidx];
1155 
1156 		if (key_prefered_oldsa)
1157 			sav = SAVLIST_READER_FIRST(sah, state);
1158 		else {
1159 			/* XXX need O(1) lookup */
1160 			struct secasvar *last = NULL;
1161 
1162 			SAVLIST_READER_FOREACH(sav, sah, state)
1163 				last = sav;
1164 			sav = last;
1165 		}
1166 		if (sav != NULL) {
1167 			KEY_SA_REF(sav);
1168 			KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1169 			    "DP cause refcnt++:%d SA:%p\n",
1170 			    key_sa_refcnt(sav), sav);
1171 			break;
1172 		}
1173 	}
1174 out:
1175 	pserialize_read_exit(s);
1176 
1177 	return sav;
1178 }
1179 
1180 #if 0
1181 static void
1182 key_sendup_message_delete(struct secasvar *sav)
1183 {
1184 	struct mbuf *m, *result = 0;
1185 	uint8_t satype;
1186 
1187 	satype = key_proto2satype(sav->sah->saidx.proto);
1188 	if (satype == 0)
1189 		goto msgfail;
1190 
1191 	m = key_setsadbmsg(SADB_DELETE, 0, satype, 0, 0, key_sa_refcnt(sav) - 1);
1192 	if (m == NULL)
1193 		goto msgfail;
1194 	result = m;
1195 
1196 	/* set sadb_address for saidx's. */
1197 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa,
1198 	    _BITS(sav->sah->saidx.src.sa.sa_len), IPSEC_ULPROTO_ANY);
1199 	if (m == NULL)
1200 		goto msgfail;
1201 	m_cat(result, m);
1202 
1203 	/* set sadb_address for saidx's. */
1204 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.src.sa,
1205 	    _BITS(sav->sah->saidx.src.sa.sa_len), IPSEC_ULPROTO_ANY);
1206 	if (m == NULL)
1207 		goto msgfail;
1208 	m_cat(result, m);
1209 
1210 	/* create SA extension */
1211 	m = key_setsadbsa(sav);
1212 	if (m == NULL)
1213 		goto msgfail;
1214 	m_cat(result, m);
1215 
1216 	if (result->m_len < sizeof(struct sadb_msg)) {
1217 		result = m_pullup(result, sizeof(struct sadb_msg));
1218 		if (result == NULL)
1219 			goto msgfail;
1220 	}
1221 
1222 	result->m_pkthdr.len = 0;
1223 	for (m = result; m; m = m->m_next)
1224 		result->m_pkthdr.len += m->m_len;
1225 	mtod(result, struct sadb_msg *)->sadb_msg_len =
1226 	    PFKEY_UNIT64(result->m_pkthdr.len);
1227 
1228 	key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
1229 	result = NULL;
1230 msgfail:
1231 	m_freem(result);
1232 }
1233 #endif
1234 
1235 /*
1236  * allocating a usable SA entry for a *INBOUND* packet.
1237  * Must call key_freesav() later.
1238  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
1239  *	NULL:		not found, or error occurred.
1240  *
1241  * In the comparison, no source address is used--for RFC2401 conformance.
1242  * To quote, from section 4.1:
1243  *	A security association is uniquely identified by a triple consisting
1244  *	of a Security Parameter Index (SPI), an IP Destination Address, and a
1245  *	security protocol (AH or ESP) identifier.
1246  * Note that, however, we do need to keep source address in IPsec SA.
1247  * IKE specification and PF_KEY specification do assume that we
1248  * keep source address in IPsec SA.  We see a tricky situation here.
1249  *
1250  * sport and dport are used for NAT-T. network order is always used.
1251  */
1252 struct secasvar *
1253 key_lookup_sa(
1254 	const union sockaddr_union *dst,
1255 	u_int proto,
1256 	u_int32_t spi,
1257 	u_int16_t sport,
1258 	u_int16_t dport,
1259 	const char* where, int tag)
1260 {
1261 	struct secasvar *sav;
1262 	int chkport;
1263 	int s;
1264 
1265 	int must_check_spi = 1;
1266 	int must_check_alg = 0;
1267 	u_int16_t cpi = 0;
1268 	u_int8_t algo = 0;
1269 	uint32_t hash_key = spi;
1270 
1271 	if ((sport != 0) && (dport != 0))
1272 		chkport = PORT_STRICT;
1273 	else
1274 		chkport = PORT_NONE;
1275 
1276 	KASSERT(dst != NULL);
1277 
1278 	/*
1279 	 * XXX IPCOMP case
1280 	 * We use cpi to define spi here. In the case where cpi <=
1281 	 * IPCOMP_CPI_NEGOTIATE_MIN, cpi just define the algorithm used, not
1282 	 * the real spi. In this case, don't check the spi but check the
1283 	 * algorithm
1284 	 */
1285 
1286 	if (proto == IPPROTO_IPCOMP) {
1287 		u_int32_t tmp;
1288 		tmp = ntohl(spi);
1289 		cpi = (u_int16_t) tmp;
1290 		if (cpi < IPCOMP_CPI_NEGOTIATE_MIN) {
1291 			algo = (u_int8_t) cpi;
1292 			hash_key = algo;
1293 			must_check_spi = 0;
1294 			must_check_alg = 1;
1295 		}
1296 	}
1297 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1298 	    "DP from %s:%u check_spi=%d(%#x), check_alg=%d(%d), proto=%d\n",
1299 	    where, tag,
1300 	    must_check_spi, ntohl(spi),
1301 	    must_check_alg, algo,
1302 	    proto);
1303 
1304 
1305 	/*
1306 	 * searching SAD.
1307 	 * XXX: to be checked internal IP header somewhere.  Also when
1308 	 * IPsec tunnel packet is received.  But ESP tunnel mode is
1309 	 * encrypted so we can't check internal IP header.
1310 	 */
1311 	s = pserialize_read_enter();
1312 	SAVLUT_READER_FOREACH(sav, &dst->sa, proto, hash_key) {
1313 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1314 		    "try match spi %#x, %#x\n",
1315 		    ntohl(spi), ntohl(sav->spi));
1316 
1317 		/* do not return entries w/ unusable state */
1318 		if (!SADB_SASTATE_USABLE_P(sav)) {
1319 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1320 			    "bad state %d\n", sav->state);
1321 			continue;
1322 		}
1323 		if (proto != sav->sah->saidx.proto) {
1324 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1325 			    "proto fail %d != %d\n",
1326 			    proto, sav->sah->saidx.proto);
1327 			continue;
1328 		}
1329 		if (must_check_spi && spi != sav->spi) {
1330 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1331 			    "spi fail %#x != %#x\n",
1332 			    ntohl(spi), ntohl(sav->spi));
1333 			continue;
1334 		}
1335 		/* XXX only on the ipcomp case */
1336 		if (must_check_alg && algo != sav->alg_comp) {
1337 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1338 			    "algo fail %d != %d\n",
1339 			    algo, sav->alg_comp);
1340 			continue;
1341 		}
1342 
1343 #if 0	/* don't check src */
1344 	/* Fix port in src->sa */
1345 
1346 		/* check src address */
1347 		if (!key_sockaddr_match(&src->sa, &sav->sah->saidx.src.sa, PORT_NONE))
1348 			continue;
1349 #endif
1350 		/* fix port of dst address XXX*/
1351 		key_porttosaddr(__UNCONST(dst), dport);
1352 		/* check dst address */
1353 		if (!key_sockaddr_match(&dst->sa, &sav->sah->saidx.dst.sa, chkport))
1354 			continue;
1355 		key_sa_ref(sav, where, tag);
1356 		goto done;
1357 	}
1358 	sav = NULL;
1359 done:
1360 	pserialize_read_exit(s);
1361 
1362 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1363 	    "DP return SA:%p; refcnt %u\n", sav, key_sa_refcnt(sav));
1364 	return sav;
1365 }
1366 
1367 static void
1368 key_validate_savlist(const struct secashead *sah, const u_int state)
1369 {
1370 #ifdef DEBUG
1371 	struct secasvar *sav, *next;
1372 	int s;
1373 
1374 	/*
1375 	 * The list should be sorted by lft_c->sadb_lifetime_addtime
1376 	 * in ascending order.
1377 	 */
1378 	s = pserialize_read_enter();
1379 	SAVLIST_READER_FOREACH(sav, sah, state) {
1380 		next = SAVLIST_READER_NEXT(sav);
1381 		if (next != NULL &&
1382 		    sav->lft_c != NULL && next->lft_c != NULL) {
1383 			KDASSERTMSG(sav->lft_c->sadb_lifetime_addtime <=
1384 			    next->lft_c->sadb_lifetime_addtime,
1385 			    "savlist is not sorted: sah=%p, state=%d, "
1386 			    "sav=%" PRIu64 ", next=%" PRIu64, sah, state,
1387 			    sav->lft_c->sadb_lifetime_addtime,
1388 			    next->lft_c->sadb_lifetime_addtime);
1389 		}
1390 	}
1391 	pserialize_read_exit(s);
1392 #endif
1393 }
1394 
1395 void
1396 key_init_sp(struct secpolicy *sp)
1397 {
1398 
1399 	ASSERT_SLEEPABLE();
1400 
1401 	sp->state = IPSEC_SPSTATE_ALIVE;
1402 	if (sp->policy == IPSEC_POLICY_IPSEC)
1403 		KASSERT(sp->req != NULL);
1404 	localcount_init(&sp->localcount);
1405 	SPLIST_ENTRY_INIT(sp);
1406 }
1407 
1408 /*
1409  * Must be called in a pserialize read section. A held SP
1410  * must be released by key_sp_unref after use.
1411  */
1412 void
1413 key_sp_ref(struct secpolicy *sp, const char* where, int tag)
1414 {
1415 
1416 	localcount_acquire(&sp->localcount);
1417 
1418 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1419 	    "DP SP:%p (ID=%u) from %s:%u; refcnt++ now %u\n",
1420 	    sp, sp->id, where, tag, key_sp_refcnt(sp));
1421 }
1422 
1423 /*
1424  * Must be called without holding key_spd.lock because the lock
1425  * would be held in localcount_release.
1426  */
1427 void
1428 key_sp_unref(struct secpolicy *sp, const char* where, int tag)
1429 {
1430 
1431 	KDASSERT(mutex_ownable(&key_spd.lock));
1432 
1433 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1434 	    "DP SP:%p (ID=%u) from %s:%u; refcnt-- now %u\n",
1435 	    sp, sp->id, where, tag, key_sp_refcnt(sp));
1436 
1437 	localcount_release(&sp->localcount, &key_spd.cv_lc, &key_spd.lock);
1438 }
1439 
1440 static void
1441 key_init_sav(struct secasvar *sav)
1442 {
1443 
1444 	ASSERT_SLEEPABLE();
1445 
1446 	localcount_init(&sav->localcount);
1447 	SAVLIST_ENTRY_INIT(sav);
1448 	SAVLUT_ENTRY_INIT(sav);
1449 }
1450 
1451 u_int
1452 key_sa_refcnt(const struct secasvar *sav)
1453 {
1454 
1455 	/* FIXME */
1456 	return 0;
1457 }
1458 
1459 void
1460 key_sa_ref(struct secasvar *sav, const char* where, int tag)
1461 {
1462 
1463 	localcount_acquire(&sav->localcount);
1464 
1465 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1466 	    "DP cause refcnt++: SA:%p from %s:%u\n",
1467 	    sav, where, tag);
1468 }
1469 
1470 void
1471 key_sa_unref(struct secasvar *sav, const char* where, int tag)
1472 {
1473 
1474 	KDASSERT(mutex_ownable(&key_sad.lock));
1475 
1476 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1477 	    "DP cause refcnt--: SA:%p from %s:%u\n",
1478 	    sav, where, tag);
1479 
1480 	localcount_release(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
1481 }
1482 
1483 #if 0
1484 /*
1485  * Must be called after calling key_lookup_sp*().
1486  * For the packet with socket.
1487  */
1488 static void
1489 key_freeso(struct socket *so)
1490 {
1491 	/* sanity check */
1492 	KASSERT(so != NULL);
1493 
1494 	switch (so->so_proto->pr_domain->dom_family) {
1495 #ifdef INET
1496 	case PF_INET:
1497 	    {
1498 		struct inpcb *pcb = sotoinpcb(so);
1499 
1500 		/* Does it have a PCB ? */
1501 		if (pcb == NULL)
1502 			return;
1503 
1504 		struct inpcbpolicy *sp = pcb->inp_sp;
1505 		key_freesp_so(&sp->sp_in);
1506 		key_freesp_so(&sp->sp_out);
1507 	    }
1508 		break;
1509 #endif
1510 #ifdef INET6
1511 	case PF_INET6:
1512 	    {
1513 #ifdef HAVE_NRL_INPCB
1514 		struct inpcb *pcb  = sotoinpcb(so);
1515 		struct inpcbpolicy *sp = pcb->inp_sp;
1516 
1517 		/* Does it have a PCB ? */
1518 		if (pcb == NULL)
1519 			return;
1520 		key_freesp_so(&sp->sp_in);
1521 		key_freesp_so(&sp->sp_out);
1522 #else
1523 		struct in6pcb *pcb  = sotoin6pcb(so);
1524 
1525 		/* Does it have a PCB ? */
1526 		if (pcb == NULL)
1527 			return;
1528 		key_freesp_so(&pcb->in6p_sp->sp_in);
1529 		key_freesp_so(&pcb->in6p_sp->sp_out);
1530 #endif
1531 	    }
1532 		break;
1533 #endif /* INET6 */
1534 	default:
1535 		IPSECLOG(LOG_DEBUG, "unknown address family=%d.\n",
1536 		    so->so_proto->pr_domain->dom_family);
1537 		return;
1538 	}
1539 }
1540 
1541 static void
1542 key_freesp_so(struct secpolicy **sp)
1543 {
1544 
1545 	KASSERT(sp != NULL);
1546 	KASSERT(*sp != NULL);
1547 
1548 	if ((*sp)->policy == IPSEC_POLICY_ENTRUST ||
1549 	    (*sp)->policy == IPSEC_POLICY_BYPASS)
1550 		return;
1551 
1552 	KASSERTMSG((*sp)->policy == IPSEC_POLICY_IPSEC,
1553 	    "invalid policy %u", (*sp)->policy);
1554 	KEY_SP_UNREF(&sp);
1555 }
1556 #endif
1557 
1558 static void
1559 key_sad_pserialize_perform(void)
1560 {
1561 
1562 	KASSERT(mutex_owned(&key_sad.lock));
1563 
1564 	while (key_sad.psz_performing)
1565 		cv_wait(&key_sad.cv_psz, &key_sad.lock);
1566 	key_sad.psz_performing = true;
1567 	mutex_exit(&key_sad.lock);
1568 
1569 	pserialize_perform(key_sad.psz);
1570 
1571 	mutex_enter(&key_sad.lock);
1572 	key_sad.psz_performing = false;
1573 	cv_broadcast(&key_sad.cv_psz);
1574 }
1575 
1576 /*
1577  * Remove the sav from the savlist of its sah and wait for references to the sav
1578  * to be released. key_sad.lock must be held.
1579  */
1580 static void
1581 key_unlink_sav(struct secasvar *sav)
1582 {
1583 
1584 	KASSERT(mutex_owned(&key_sad.lock));
1585 
1586 	SAVLIST_WRITER_REMOVE(sav);
1587 	SAVLUT_WRITER_REMOVE(sav);
1588 
1589 	KDASSERT(mutex_ownable(softnet_lock));
1590 	key_sad_pserialize_perform();
1591 
1592 	localcount_drain(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
1593 }
1594 
1595 /*
1596  * Destroy an sav where the sav must be unlinked from an sah
1597  * by say key_unlink_sav.
1598  */
1599 static void
1600 key_destroy_sav(struct secasvar *sav)
1601 {
1602 
1603 	ASSERT_SLEEPABLE();
1604 
1605 	localcount_fini(&sav->localcount);
1606 	SAVLIST_ENTRY_DESTROY(sav);
1607 
1608 	key_delsav(sav);
1609 }
1610 
1611 /*
1612  * Wait for references of a passed sav to go away.
1613  */
1614 static void
1615 key_wait_sav(struct secasvar *sav)
1616 {
1617 
1618 	ASSERT_SLEEPABLE();
1619 
1620 	mutex_enter(&key_sad.lock);
1621 	KASSERT(sav->state == SADB_SASTATE_DEAD);
1622 	KDASSERT(mutex_ownable(softnet_lock));
1623 	key_sad_pserialize_perform();
1624 	localcount_drain(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
1625 	mutex_exit(&key_sad.lock);
1626 }
1627 
1628 /* %%% SPD management */
1629 /*
1630  * free security policy entry.
1631  */
1632 static void
1633 key_destroy_sp(struct secpolicy *sp)
1634 {
1635 
1636 	SPLIST_ENTRY_DESTROY(sp);
1637 	localcount_fini(&sp->localcount);
1638 
1639 	key_free_sp(sp);
1640 
1641 	key_update_used();
1642 }
1643 
1644 void
1645 key_free_sp(struct secpolicy *sp)
1646 {
1647 	struct ipsecrequest *isr = sp->req, *nextisr;
1648 
1649 	while (isr != NULL) {
1650 		nextisr = isr->next;
1651 		kmem_free(isr, sizeof(*isr));
1652 		isr = nextisr;
1653 	}
1654 
1655 	kmem_free(sp, sizeof(*sp));
1656 }
1657 
1658 void
1659 key_socksplist_add(struct secpolicy *sp)
1660 {
1661 
1662 	mutex_enter(&key_spd.lock);
1663 	PSLIST_WRITER_INSERT_HEAD(&key_spd.socksplist, sp, pslist_entry);
1664 	mutex_exit(&key_spd.lock);
1665 
1666 	key_update_used();
1667 }
1668 
1669 /*
1670  * search SPD
1671  * OUT:	NULL	: not found
1672  *	others	: found, pointer to a SP.
1673  */
1674 static struct secpolicy *
1675 key_getsp(const struct secpolicyindex *spidx)
1676 {
1677 	struct secpolicy *sp;
1678 	int s;
1679 
1680 	KASSERT(spidx != NULL);
1681 
1682 	s = pserialize_read_enter();
1683 	SPLIST_READER_FOREACH(sp, spidx->dir) {
1684 		if (sp->state == IPSEC_SPSTATE_DEAD)
1685 			continue;
1686 		if (key_spidx_match_exactly(spidx, &sp->spidx)) {
1687 			KEY_SP_REF(sp);
1688 			pserialize_read_exit(s);
1689 			return sp;
1690 		}
1691 	}
1692 	pserialize_read_exit(s);
1693 
1694 	return NULL;
1695 }
1696 
1697 /*
1698  * search SPD and remove found SP
1699  * OUT:	NULL	: not found
1700  *	others	: found, pointer to a SP.
1701  */
1702 static struct secpolicy *
1703 key_lookup_and_remove_sp(const struct secpolicyindex *spidx, bool from_kernel)
1704 {
1705 	struct secpolicy *sp = NULL;
1706 
1707 	mutex_enter(&key_spd.lock);
1708 	SPLIST_WRITER_FOREACH(sp, spidx->dir) {
1709 		KASSERTMSG(sp->state != IPSEC_SPSTATE_DEAD, "sp->state=%u",
1710 		    sp->state);
1711 		/*
1712 		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
1713 		 * removed by userland programs.
1714 		 */
1715 		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
1716 			continue;
1717 		if (key_spidx_match_exactly(spidx, &sp->spidx)) {
1718 			key_unlink_sp(sp);
1719 			goto out;
1720 		}
1721 	}
1722 	sp = NULL;
1723 out:
1724 	mutex_exit(&key_spd.lock);
1725 
1726 	return sp;
1727 }
1728 
1729 /*
1730  * get SP by index.
1731  * OUT:	NULL	: not found
1732  *	others	: found, pointer to a SP.
1733  */
1734 static struct secpolicy *
1735 key_getspbyid(u_int32_t id)
1736 {
1737 	struct secpolicy *sp;
1738 	int s;
1739 
1740 	s = pserialize_read_enter();
1741 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_INBOUND) {
1742 		if (sp->state == IPSEC_SPSTATE_DEAD)
1743 			continue;
1744 		if (sp->id == id) {
1745 			KEY_SP_REF(sp);
1746 			goto out;
1747 		}
1748 	}
1749 
1750 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_OUTBOUND) {
1751 		if (sp->state == IPSEC_SPSTATE_DEAD)
1752 			continue;
1753 		if (sp->id == id) {
1754 			KEY_SP_REF(sp);
1755 			goto out;
1756 		}
1757 	}
1758 out:
1759 	pserialize_read_exit(s);
1760 	return sp;
1761 }
1762 
1763 /*
1764  * get SP by index, remove and return it.
1765  * OUT:	NULL	: not found
1766  *	others	: found, pointer to a SP.
1767  */
1768 static struct secpolicy *
1769 key_lookupbyid_and_remove_sp(u_int32_t id, bool from_kernel)
1770 {
1771 	struct secpolicy *sp;
1772 
1773 	mutex_enter(&key_spd.lock);
1774 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_INBOUND) {
1775 		KASSERTMSG(sp->state != IPSEC_SPSTATE_DEAD, "sp->state=%u",
1776 		    sp->state);
1777 		/*
1778 		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
1779 		 * removed by userland programs.
1780 		 */
1781 		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
1782 			continue;
1783 		if (sp->id == id)
1784 			goto out;
1785 	}
1786 
1787 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_OUTBOUND) {
1788 		KASSERTMSG(sp->state != IPSEC_SPSTATE_DEAD, "sp->state=%u",
1789 		    sp->state);
1790 		/*
1791 		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
1792 		 * removed by userland programs.
1793 		 */
1794 		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
1795 			continue;
1796 		if (sp->id == id)
1797 			goto out;
1798 	}
1799 out:
1800 	if (sp != NULL)
1801 		key_unlink_sp(sp);
1802 	mutex_exit(&key_spd.lock);
1803 	return sp;
1804 }
1805 
1806 struct secpolicy *
1807 key_newsp(const char* where, int tag)
1808 {
1809 	struct secpolicy *newsp = NULL;
1810 
1811 	newsp = kmem_zalloc(sizeof(struct secpolicy), KM_SLEEP);
1812 
1813 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1814 	    "DP from %s:%u return SP:%p\n", where, tag, newsp);
1815 	return newsp;
1816 }
1817 
1818 /*
1819  * create secpolicy structure from sadb_x_policy structure.
1820  * NOTE: `state', `secpolicyindex' in secpolicy structure are not set,
1821  * so must be set properly later.
1822  */
1823 static struct secpolicy *
1824 _key_msg2sp(const struct sadb_x_policy *xpl0, size_t len, int *error,
1825     bool from_kernel)
1826 {
1827 	struct secpolicy *newsp;
1828 
1829 	KASSERT(!cpu_softintr_p());
1830 	KASSERT(xpl0 != NULL);
1831 	KASSERT(len >= sizeof(*xpl0));
1832 
1833 	if (len != PFKEY_EXTLEN(xpl0)) {
1834 		IPSECLOG(LOG_DEBUG, "Invalid msg length.\n");
1835 		*error = EINVAL;
1836 		return NULL;
1837 	}
1838 
1839 	newsp = KEY_NEWSP();
1840 	if (newsp == NULL) {
1841 		*error = ENOBUFS;
1842 		return NULL;
1843 	}
1844 
1845 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1846 	newsp->policy = xpl0->sadb_x_policy_type;
1847 
1848 	/* check policy */
1849 	switch (xpl0->sadb_x_policy_type) {
1850 	case IPSEC_POLICY_DISCARD:
1851 	case IPSEC_POLICY_NONE:
1852 	case IPSEC_POLICY_ENTRUST:
1853 	case IPSEC_POLICY_BYPASS:
1854 		newsp->req = NULL;
1855 		*error = 0;
1856 		return newsp;
1857 
1858 	case IPSEC_POLICY_IPSEC:
1859 		/* Continued */
1860 		break;
1861 	default:
1862 		IPSECLOG(LOG_DEBUG, "invalid policy type.\n");
1863 		key_free_sp(newsp);
1864 		*error = EINVAL;
1865 		return NULL;
1866 	}
1867 
1868 	/* IPSEC_POLICY_IPSEC */
1869     {
1870 	int tlen;
1871 	const struct sadb_x_ipsecrequest *xisr;
1872 	uint16_t xisr_reqid;
1873 	struct ipsecrequest **p_isr = &newsp->req;
1874 
1875 	/* validity check */
1876 	if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1877 		IPSECLOG(LOG_DEBUG, "Invalid msg length.\n");
1878 		*error = EINVAL;
1879 		goto free_exit;
1880 	}
1881 
1882 	tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1883 	xisr = (const struct sadb_x_ipsecrequest *)(xpl0 + 1);
1884 
1885 	while (tlen > 0) {
1886 		/* length check */
1887 		if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr)) {
1888 			IPSECLOG(LOG_DEBUG, "invalid ipsecrequest length.\n");
1889 			*error = EINVAL;
1890 			goto free_exit;
1891 		}
1892 
1893 		/* allocate request buffer */
1894 		*p_isr = kmem_zalloc(sizeof(**p_isr), KM_SLEEP);
1895 
1896 		/* set values */
1897 		(*p_isr)->next = NULL;
1898 
1899 		switch (xisr->sadb_x_ipsecrequest_proto) {
1900 		case IPPROTO_ESP:
1901 		case IPPROTO_AH:
1902 		case IPPROTO_IPCOMP:
1903 			break;
1904 		default:
1905 			IPSECLOG(LOG_DEBUG, "invalid proto type=%u\n",
1906 			    xisr->sadb_x_ipsecrequest_proto);
1907 			*error = EPROTONOSUPPORT;
1908 			goto free_exit;
1909 		}
1910 		(*p_isr)->saidx.proto = xisr->sadb_x_ipsecrequest_proto;
1911 
1912 		switch (xisr->sadb_x_ipsecrequest_mode) {
1913 		case IPSEC_MODE_TRANSPORT:
1914 		case IPSEC_MODE_TUNNEL:
1915 			break;
1916 		case IPSEC_MODE_ANY:
1917 		default:
1918 			IPSECLOG(LOG_DEBUG, "invalid mode=%u\n",
1919 			    xisr->sadb_x_ipsecrequest_mode);
1920 			*error = EINVAL;
1921 			goto free_exit;
1922 		}
1923 		(*p_isr)->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1924 
1925 		switch (xisr->sadb_x_ipsecrequest_level) {
1926 		case IPSEC_LEVEL_DEFAULT:
1927 		case IPSEC_LEVEL_USE:
1928 		case IPSEC_LEVEL_REQUIRE:
1929 			break;
1930 		case IPSEC_LEVEL_UNIQUE:
1931 			xisr_reqid = xisr->sadb_x_ipsecrequest_reqid;
1932 			/* validity check */
1933 			/*
1934 			 * case 1) from_kernel == false
1935 			 * That means the request comes from userland.
1936 			 * If range violation of reqid, kernel will
1937 			 * update it, don't refuse it.
1938 			 *
1939 			 * case 2) from_kernel == true
1940 			 * That means the request comes from kernel
1941 			 * (e.g. ipsec(4) I/F).
1942 			 * Use thre requested reqid to avoid inconsistency
1943 			 * between kernel's reqid and the reqid in pf_key
1944 			 * message sent to userland. The pf_key message is
1945 			 * built by diverting request mbuf.
1946 			 */
1947 			if (!from_kernel &&
1948 			    xisr_reqid > IPSEC_MANUAL_REQID_MAX) {
1949 				IPSECLOG(LOG_DEBUG,
1950 				    "reqid=%d range "
1951 				    "violation, updated by kernel.\n",
1952 				    xisr_reqid);
1953 				xisr_reqid = 0;
1954 			}
1955 
1956 			/* allocate new reqid id if reqid is zero. */
1957 			if (xisr_reqid == 0) {
1958 				u_int16_t reqid = key_newreqid();
1959 				if (reqid == 0) {
1960 					*error = ENOBUFS;
1961 					goto free_exit;
1962 				}
1963 				(*p_isr)->saidx.reqid = reqid;
1964 			} else {
1965 			/* set it for manual keying. */
1966 				(*p_isr)->saidx.reqid = xisr_reqid;
1967 			}
1968 			break;
1969 
1970 		default:
1971 			IPSECLOG(LOG_DEBUG, "invalid level=%u\n",
1972 			    xisr->sadb_x_ipsecrequest_level);
1973 			*error = EINVAL;
1974 			goto free_exit;
1975 		}
1976 		(*p_isr)->level = xisr->sadb_x_ipsecrequest_level;
1977 
1978 		/* set IP addresses if there */
1979 		/*
1980 		 * NOTE:
1981 		 * MOBIKE Extensions for PF_KEY draft says:
1982 		 *     If tunnel mode is specified, the sadb_x_ipsecrequest
1983 		 *     structure is followed by two sockaddr structures that
1984 		 *     define the tunnel endpoint addresses.  In the case that
1985 		 *     transport mode is used, no additional addresses are
1986 		 *     specified.
1987 		 * see: https://tools.ietf.org/html/draft-schilcher-mobike-pfkey-extension-01
1988 		 *
1989 		 * And then, the IP addresses will be set by
1990 		 * ipsec_fill_saidx_bymbuf() from packet in transport mode.
1991 		 * This behavior is used by NAT-T enabled ipsecif(4).
1992 		 */
1993 		if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1994 			const struct sockaddr *paddr;
1995 
1996 			paddr = (const struct sockaddr *)(xisr + 1);
1997 
1998 			/* validity check */
1999 			if (paddr->sa_len > sizeof((*p_isr)->saidx.src)) {
2000 				IPSECLOG(LOG_DEBUG, "invalid request "
2001 				    "address length.\n");
2002 				*error = EINVAL;
2003 				goto free_exit;
2004 			}
2005 			memcpy(&(*p_isr)->saidx.src, paddr, paddr->sa_len);
2006 
2007 			paddr = (const struct sockaddr *)((const char *)paddr
2008 			    + paddr->sa_len);
2009 
2010 			/* validity check */
2011 			if (paddr->sa_len > sizeof((*p_isr)->saidx.dst)) {
2012 				IPSECLOG(LOG_DEBUG, "invalid request "
2013 				    "address length.\n");
2014 				*error = EINVAL;
2015 				goto free_exit;
2016 			}
2017 			memcpy(&(*p_isr)->saidx.dst, paddr, paddr->sa_len);
2018 		}
2019 
2020 		(*p_isr)->sp = newsp;
2021 
2022 		/* initialization for the next. */
2023 		p_isr = &(*p_isr)->next;
2024 		tlen -= xisr->sadb_x_ipsecrequest_len;
2025 
2026 		/* validity check */
2027 		if (tlen < 0) {
2028 			IPSECLOG(LOG_DEBUG, "becoming tlen < 0.\n");
2029 			*error = EINVAL;
2030 			goto free_exit;
2031 		}
2032 
2033 		xisr = (const struct sadb_x_ipsecrequest *)((const char *)xisr +
2034 		    xisr->sadb_x_ipsecrequest_len);
2035 	}
2036     }
2037 
2038 	*error = 0;
2039 	return newsp;
2040 
2041 free_exit:
2042 	key_free_sp(newsp);
2043 	return NULL;
2044 }
2045 
2046 struct secpolicy *
2047 key_msg2sp(const struct sadb_x_policy *xpl0, size_t len, int *error)
2048 {
2049 
2050 	return _key_msg2sp(xpl0, len, error, false);
2051 }
2052 
2053 u_int16_t
2054 key_newreqid(void)
2055 {
2056 	static u_int16_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
2057 
2058 	auto_reqid = (auto_reqid == 0xffff ?
2059 	    IPSEC_MANUAL_REQID_MAX + 1 : auto_reqid + 1);
2060 
2061 	/* XXX should be unique check */
2062 
2063 	return auto_reqid;
2064 }
2065 
2066 /*
2067  * copy secpolicy struct to sadb_x_policy structure indicated.
2068  */
2069 struct mbuf *
2070 key_sp2msg(const struct secpolicy *sp, int mflag)
2071 {
2072 	struct sadb_x_policy *xpl;
2073 	int tlen;
2074 	char *p;
2075 	struct mbuf *m;
2076 
2077 	KASSERT(sp != NULL);
2078 
2079 	tlen = key_getspreqmsglen(sp);
2080 
2081 	m = key_alloc_mbuf(tlen, mflag);
2082 	if (!m || m->m_next) {	/*XXX*/
2083 		m_freem(m);
2084 		return NULL;
2085 	}
2086 
2087 	m->m_len = tlen;
2088 	m->m_next = NULL;
2089 	xpl = mtod(m, struct sadb_x_policy *);
2090 	memset(xpl, 0, tlen);
2091 
2092 	xpl->sadb_x_policy_len = PFKEY_UNIT64(tlen);
2093 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
2094 	xpl->sadb_x_policy_type = sp->policy;
2095 	xpl->sadb_x_policy_dir = sp->spidx.dir;
2096 	xpl->sadb_x_policy_id = sp->id;
2097 	if (sp->origin == IPSEC_SPORIGIN_KERNEL)
2098 		xpl->sadb_x_policy_flags |= IPSEC_POLICY_FLAG_ORIGIN_KERNEL;
2099 	p = (char *)xpl + sizeof(*xpl);
2100 
2101 	/* if is the policy for ipsec ? */
2102 	if (sp->policy == IPSEC_POLICY_IPSEC) {
2103 		struct sadb_x_ipsecrequest *xisr;
2104 		struct ipsecrequest *isr;
2105 
2106 		for (isr = sp->req; isr != NULL; isr = isr->next) {
2107 
2108 			xisr = (struct sadb_x_ipsecrequest *)p;
2109 
2110 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
2111 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
2112 			xisr->sadb_x_ipsecrequest_level = isr->level;
2113 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
2114 
2115 			p += sizeof(*xisr);
2116 			memcpy(p, &isr->saidx.src, isr->saidx.src.sa.sa_len);
2117 			p += isr->saidx.src.sa.sa_len;
2118 			memcpy(p, &isr->saidx.dst, isr->saidx.dst.sa.sa_len);
2119 			p += isr->saidx.src.sa.sa_len;
2120 
2121 			xisr->sadb_x_ipsecrequest_len =
2122 			    PFKEY_ALIGN8(sizeof(*xisr)
2123 			    + isr->saidx.src.sa.sa_len
2124 			    + isr->saidx.dst.sa.sa_len);
2125 		}
2126 	}
2127 
2128 	return m;
2129 }
2130 
2131 /*
2132  * m will not be freed nor modified. It never return NULL.
2133  * If it returns a mbuf of M_PKTHDR, the mbuf ensures to have
2134  * contiguous length at least sizeof(struct sadb_msg).
2135  */
2136 static struct mbuf *
2137 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
2138 		int ndeep, int nitem, ...)
2139 {
2140 	va_list ap;
2141 	int idx;
2142 	int i;
2143 	struct mbuf *result = NULL, *n;
2144 	int len;
2145 
2146 	KASSERT(m != NULL);
2147 	KASSERT(mhp != NULL);
2148 	KASSERT(!cpu_softintr_p());
2149 
2150 	va_start(ap, nitem);
2151 	for (i = 0; i < nitem; i++) {
2152 		idx = va_arg(ap, int);
2153 		KASSERT(idx >= 0);
2154 		KASSERT(idx <= SADB_EXT_MAX);
2155 		/* don't attempt to pull empty extension */
2156 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
2157 			continue;
2158 		if (idx != SADB_EXT_RESERVED &&
2159 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
2160 			continue;
2161 
2162 		if (idx == SADB_EXT_RESERVED) {
2163 			CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MHLEN);
2164 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2165 			MGETHDR(n, M_WAITOK, MT_DATA);
2166 			n->m_len = len;
2167 			n->m_next = NULL;
2168 			m_copydata(m, 0, sizeof(struct sadb_msg),
2169 			    mtod(n, void *));
2170 		} else if (i < ndeep) {
2171 			len = mhp->extlen[idx];
2172 			n = key_alloc_mbuf(len, M_WAITOK);
2173 			KASSERT(n->m_next == NULL);
2174 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
2175 			    mtod(n, void *));
2176 		} else {
2177 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
2178 			    M_WAITOK);
2179 		}
2180 		KASSERT(n != NULL);
2181 
2182 		if (result)
2183 			m_cat(result, n);
2184 		else
2185 			result = n;
2186 	}
2187 	va_end(ap);
2188 
2189 	KASSERT(result != NULL);
2190 	if ((result->m_flags & M_PKTHDR) != 0) {
2191 		result->m_pkthdr.len = 0;
2192 		for (n = result; n; n = n->m_next)
2193 			result->m_pkthdr.len += n->m_len;
2194 		KASSERT(result->m_len >= sizeof(struct sadb_msg));
2195 	}
2196 
2197 	return result;
2198 }
2199 
2200 /*
2201  * The argument _sp must not overwrite until SP is created and registered
2202  * successfully.
2203  */
2204 static int
2205 key_spdadd(struct socket *so, struct mbuf *m,
2206 	   const struct sadb_msghdr *mhp, struct secpolicy **_sp,
2207 	   bool from_kernel)
2208 {
2209 	const struct sockaddr *src, *dst;
2210 	const struct sadb_x_policy *xpl0;
2211 	struct sadb_x_policy *xpl;
2212 	const struct sadb_lifetime *lft = NULL;
2213 	struct secpolicyindex spidx;
2214 	struct secpolicy *newsp;
2215 	int error;
2216 	uint32_t sadb_x_policy_id;
2217 
2218 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
2219 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
2220 	    mhp->ext[SADB_X_EXT_POLICY] == NULL) {
2221 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2222 		return key_senderror(so, m, EINVAL);
2223 	}
2224 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
2225 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
2226 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2227 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2228 		return key_senderror(so, m, EINVAL);
2229 	}
2230 	if (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL) {
2231 		if (mhp->extlen[SADB_EXT_LIFETIME_HARD] <
2232 		    sizeof(struct sadb_lifetime)) {
2233 			IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2234 			return key_senderror(so, m, EINVAL);
2235 		}
2236 		lft = mhp->ext[SADB_EXT_LIFETIME_HARD];
2237 	}
2238 
2239 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
2240 
2241 	/* checking the direction. */
2242 	switch (xpl0->sadb_x_policy_dir) {
2243 	case IPSEC_DIR_INBOUND:
2244 	case IPSEC_DIR_OUTBOUND:
2245 		break;
2246 	default:
2247 		IPSECLOG(LOG_DEBUG, "Invalid SP direction.\n");
2248 		return key_senderror(so, m, EINVAL);
2249 	}
2250 
2251 	/* check policy */
2252 	/* key_api_spdadd() accepts DISCARD, NONE and IPSEC. */
2253 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_ENTRUST ||
2254 	    xpl0->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
2255 		IPSECLOG(LOG_DEBUG, "Invalid policy type.\n");
2256 		return key_senderror(so, m, EINVAL);
2257 	}
2258 
2259 	/* policy requests are mandatory when action is ipsec. */
2260 	if (mhp->msg->sadb_msg_type != SADB_X_SPDSETIDX &&
2261 	    xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
2262 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
2263 		IPSECLOG(LOG_DEBUG, "some policy requests part required.\n");
2264 		return key_senderror(so, m, EINVAL);
2265 	}
2266 
2267 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
2268 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
2269 
2270 	/* sanity check on addr pair */
2271 	if (src->sa_family != dst->sa_family)
2272 		return key_senderror(so, m, EINVAL);
2273 	if (src->sa_len != dst->sa_len)
2274 		return key_senderror(so, m, EINVAL);
2275 
2276 	key_init_spidx_bymsghdr(&spidx, mhp);
2277 
2278 	/*
2279 	 * checking there is SP already or not.
2280 	 * SPDUPDATE doesn't depend on whether there is a SP or not.
2281 	 * If the type is either SPDADD or SPDSETIDX AND a SP is found,
2282 	 * then error.
2283 	 */
2284     {
2285 	struct secpolicy *sp;
2286 
2287 	if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2288 		sp = key_lookup_and_remove_sp(&spidx, from_kernel);
2289 		if (sp != NULL)
2290 			key_destroy_sp(sp);
2291 	} else {
2292 		sp = key_getsp(&spidx);
2293 		if (sp != NULL) {
2294 			KEY_SP_UNREF(&sp);
2295 			IPSECLOG(LOG_DEBUG, "a SP entry exists already.\n");
2296 			return key_senderror(so, m, EEXIST);
2297 		}
2298 	}
2299     }
2300 
2301 	/* allocation new SP entry */
2302 	newsp = _key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error, from_kernel);
2303 	if (newsp == NULL) {
2304 		return key_senderror(so, m, error);
2305 	}
2306 
2307 	newsp->id = key_getnewspid();
2308 	if (newsp->id == 0) {
2309 		kmem_free(newsp, sizeof(*newsp));
2310 		return key_senderror(so, m, ENOBUFS);
2311 	}
2312 
2313 	newsp->spidx = spidx;
2314 	newsp->created = time_uptime;
2315 	newsp->lastused = newsp->created;
2316 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
2317 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
2318 	if (from_kernel)
2319 		newsp->origin = IPSEC_SPORIGIN_KERNEL;
2320 	else
2321 		newsp->origin = IPSEC_SPORIGIN_USER;
2322 
2323 	key_init_sp(newsp);
2324 	if (from_kernel)
2325 		KEY_SP_REF(newsp);
2326 
2327 	sadb_x_policy_id = newsp->id;
2328 
2329 	if (_sp != NULL)
2330 		*_sp = newsp;
2331 
2332 	mutex_enter(&key_spd.lock);
2333 	SPLIST_WRITER_INSERT_TAIL(newsp->spidx.dir, newsp);
2334 	mutex_exit(&key_spd.lock);
2335 	/*
2336 	 * We don't have a reference to newsp, so we must not touch newsp from
2337 	 * now on.  If you want to do, you must take a reference beforehand.
2338 	 */
2339 	newsp = NULL;
2340 
2341 #ifdef notyet
2342 	/* delete the entry in key_misc.spacqlist */
2343 	if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2344 		struct secspacq *spacq = key_getspacq(&spidx);
2345 		if (spacq != NULL) {
2346 			/* reset counter in order to deletion by timehandler. */
2347 			spacq->created = time_uptime;
2348 			spacq->count = 0;
2349 		}
2350     	}
2351 #endif
2352 
2353 	/* Invalidate all cached SPD pointers in the PCBs. */
2354 	ipsec_invalpcbcacheall();
2355 
2356 #if defined(GATEWAY)
2357 	/* Invalidate the ipflow cache, as well. */
2358 	ipflow_invalidate_all(0);
2359 #ifdef INET6
2360 	if (in6_present)
2361 		ip6flow_invalidate_all(0);
2362 #endif /* INET6 */
2363 #endif /* GATEWAY */
2364 
2365 	key_update_used();
2366 
2367     {
2368 	struct mbuf *n, *mpolicy;
2369 	int off;
2370 
2371 	/* create new sadb_msg to reply. */
2372 	if (lft) {
2373 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
2374 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
2375 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2376 	} else {
2377 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
2378 		    SADB_X_EXT_POLICY,
2379 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2380 	}
2381 
2382 	key_fill_replymsg(n, 0);
2383 	off = 0;
2384 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
2385 	    sizeof(*xpl), &off);
2386 	if (mpolicy == NULL) {
2387 		/* n is already freed */
2388 		/*
2389 		 * valid sp has been created, so we does not overwrite _sp
2390 		 * NULL here. let caller decide to use the sp or not.
2391 		 */
2392 		return key_senderror(so, m, ENOBUFS);
2393 	}
2394 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, char *) + off);
2395 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
2396 		m_freem(n);
2397 		/* ditto */
2398 		return key_senderror(so, m, EINVAL);
2399 	}
2400 
2401 	xpl->sadb_x_policy_id = sadb_x_policy_id;
2402 
2403 	m_freem(m);
2404 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2405     }
2406 }
2407 
2408 /*
2409  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
2410  * add an entry to SP database, when received
2411  *   <base, address(SD), (lifetime(H),) policy>
2412  * from the user(?).
2413  * Adding to SP database,
2414  * and send
2415  *   <base, address(SD), (lifetime(H),) policy>
2416  * to the socket which was send.
2417  *
2418  * SPDADD set a unique policy entry.
2419  * SPDSETIDX like SPDADD without a part of policy requests.
2420  * SPDUPDATE replace a unique policy entry.
2421  *
2422  * m will always be freed.
2423  */
2424 static int
2425 key_api_spdadd(struct socket *so, struct mbuf *m,
2426 	       const struct sadb_msghdr *mhp)
2427 {
2428 
2429 	return key_spdadd(so, m, mhp, NULL, false);
2430 }
2431 
2432 struct secpolicy *
2433 key_kpi_spdadd(struct mbuf *m)
2434 {
2435 	struct sadb_msghdr mh;
2436 	int error;
2437 	struct secpolicy *sp = NULL;
2438 
2439 	error = key_align(m, &mh);
2440 	if (error)
2441 		return NULL;
2442 
2443 	error = key_spdadd(NULL, m, &mh, &sp, true);
2444 	if (error) {
2445 		/*
2446 		 * Currently, when key_spdadd() cannot send a PFKEY message
2447 		 * which means SP has been created, key_spdadd() returns error
2448 		 * although SP is created successfully.
2449 		 * Kernel components would not care PFKEY messages, so return
2450 		 * the "sp" regardless of error code. key_spdadd() overwrites
2451 		 * the argument only if SP  is created successfully.
2452 		 */
2453 	}
2454 	return sp;
2455 }
2456 
2457 /*
2458  * get new policy id.
2459  * OUT:
2460  *	0:	failure.
2461  *	others: success.
2462  */
2463 static u_int32_t
2464 key_getnewspid(void)
2465 {
2466 	u_int32_t newid = 0;
2467 	int count = key_spi_trycnt;	/* XXX */
2468 	struct secpolicy *sp;
2469 
2470 	/* when requesting to allocate spi ranged */
2471 	while (count--) {
2472 		newid = (policy_id = (policy_id == ~0 ? 1 : policy_id + 1));
2473 
2474 		sp = key_getspbyid(newid);
2475 		if (sp == NULL)
2476 			break;
2477 
2478 		KEY_SP_UNREF(&sp);
2479 	}
2480 
2481 	if (count == 0 || newid == 0) {
2482 		IPSECLOG(LOG_DEBUG, "to allocate policy id is failed.\n");
2483 		return 0;
2484 	}
2485 
2486 	return newid;
2487 }
2488 
2489 /*
2490  * SADB_SPDDELETE processing
2491  * receive
2492  *   <base, address(SD), policy(*)>
2493  * from the user(?), and set SADB_SASTATE_DEAD,
2494  * and send,
2495  *   <base, address(SD), policy(*)>
2496  * to the ikmpd.
2497  * policy(*) including direction of policy.
2498  *
2499  * m will always be freed.
2500  */
2501 static int
2502 key_api_spddelete(struct socket *so, struct mbuf *m,
2503               const struct sadb_msghdr *mhp)
2504 {
2505 	struct sadb_x_policy *xpl0;
2506 	struct secpolicyindex spidx;
2507 	struct secpolicy *sp;
2508 
2509 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
2510 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
2511 	    mhp->ext[SADB_X_EXT_POLICY] == NULL) {
2512 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2513 		return key_senderror(so, m, EINVAL);
2514 	}
2515 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
2516 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
2517 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2518 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2519 		return key_senderror(so, m, EINVAL);
2520 	}
2521 
2522 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
2523 
2524 	/* checking the direction. */
2525 	switch (xpl0->sadb_x_policy_dir) {
2526 	case IPSEC_DIR_INBOUND:
2527 	case IPSEC_DIR_OUTBOUND:
2528 		break;
2529 	default:
2530 		IPSECLOG(LOG_DEBUG, "Invalid SP direction.\n");
2531 		return key_senderror(so, m, EINVAL);
2532 	}
2533 
2534 	/* make secindex */
2535 	key_init_spidx_bymsghdr(&spidx, mhp);
2536 
2537 	/* Is there SP in SPD ? */
2538 	sp = key_lookup_and_remove_sp(&spidx, false);
2539 	if (sp == NULL) {
2540 		IPSECLOG(LOG_DEBUG, "no SP found.\n");
2541 		return key_senderror(so, m, EINVAL);
2542 	}
2543 
2544 	/* save policy id to buffer to be returned. */
2545 	xpl0->sadb_x_policy_id = sp->id;
2546 
2547 	key_destroy_sp(sp);
2548 
2549 	/* We're deleting policy; no need to invalidate the ipflow cache. */
2550 
2551     {
2552 	struct mbuf *n;
2553 
2554 	/* create new sadb_msg to reply. */
2555 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2556 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2557 	key_fill_replymsg(n, 0);
2558 	m_freem(m);
2559 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2560     }
2561 }
2562 
2563 static struct mbuf *
2564 key_alloc_mbuf_simple(int len, int mflag)
2565 {
2566 	struct mbuf *n;
2567 
2568 	KASSERT(mflag == M_NOWAIT || (mflag == M_WAITOK && !cpu_softintr_p()));
2569 
2570 	MGETHDR(n, mflag, MT_DATA);
2571 	if (n && len > MHLEN) {
2572 		MCLGET(n, mflag);
2573 		if ((n->m_flags & M_EXT) == 0) {
2574 			m_freem(n);
2575 			n = NULL;
2576 		}
2577 	}
2578 	return n;
2579 }
2580 
2581 /*
2582  * SADB_SPDDELETE2 processing
2583  * receive
2584  *   <base, policy(*)>
2585  * from the user(?), and set SADB_SASTATE_DEAD,
2586  * and send,
2587  *   <base, policy(*)>
2588  * to the ikmpd.
2589  * policy(*) including direction of policy.
2590  *
2591  * m will always be freed.
2592  */
2593 static int
2594 key_spddelete2(struct socket *so, struct mbuf *m,
2595 	       const struct sadb_msghdr *mhp, bool from_kernel)
2596 {
2597 	u_int32_t id;
2598 	struct secpolicy *sp;
2599 	const struct sadb_x_policy *xpl;
2600 
2601 	if (mhp->ext[SADB_X_EXT_POLICY] == NULL ||
2602 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2603 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2604 		return key_senderror(so, m, EINVAL);
2605 	}
2606 
2607 	xpl = mhp->ext[SADB_X_EXT_POLICY];
2608 	id = xpl->sadb_x_policy_id;
2609 
2610 	/* Is there SP in SPD ? */
2611 	sp = key_lookupbyid_and_remove_sp(id, from_kernel);
2612 	if (sp == NULL) {
2613 		IPSECLOG(LOG_DEBUG, "no SP found id:%u.\n", id);
2614 		return key_senderror(so, m, EINVAL);
2615 	}
2616 
2617 	key_destroy_sp(sp);
2618 
2619 	/* We're deleting policy; no need to invalidate the ipflow cache. */
2620 
2621     {
2622 	struct mbuf *n, *nn;
2623 	int off, len;
2624 
2625 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MCLBYTES);
2626 
2627 	/* create new sadb_msg to reply. */
2628 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2629 
2630 	n = key_alloc_mbuf_simple(len, M_WAITOK);
2631 	n->m_len = len;
2632 	n->m_next = NULL;
2633 	off = 0;
2634 
2635 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
2636 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2637 
2638 	KASSERTMSG(off == len, "length inconsistency");
2639 
2640 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2641 	    mhp->extlen[SADB_X_EXT_POLICY], M_WAITOK);
2642 
2643 	n->m_pkthdr.len = 0;
2644 	for (nn = n; nn; nn = nn->m_next)
2645 		n->m_pkthdr.len += nn->m_len;
2646 
2647 	key_fill_replymsg(n, 0);
2648 	m_freem(m);
2649 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2650     }
2651 }
2652 
2653 /*
2654  * SADB_SPDDELETE2 processing
2655  * receive
2656  *   <base, policy(*)>
2657  * from the user(?), and set SADB_SASTATE_DEAD,
2658  * and send,
2659  *   <base, policy(*)>
2660  * to the ikmpd.
2661  * policy(*) including direction of policy.
2662  *
2663  * m will always be freed.
2664  */
2665 static int
2666 key_api_spddelete2(struct socket *so, struct mbuf *m,
2667 	       const struct sadb_msghdr *mhp)
2668 {
2669 
2670 	return key_spddelete2(so, m, mhp, false);
2671 }
2672 
2673 int
2674 key_kpi_spddelete2(struct mbuf *m)
2675 {
2676 	struct sadb_msghdr mh;
2677 	int error;
2678 
2679 	error = key_align(m, &mh);
2680 	if (error)
2681 		return EINVAL;
2682 
2683 	return key_spddelete2(NULL, m, &mh, true);
2684 }
2685 
2686 /*
2687  * SADB_X_GET processing
2688  * receive
2689  *   <base, policy(*)>
2690  * from the user(?),
2691  * and send,
2692  *   <base, address(SD), policy>
2693  * to the ikmpd.
2694  * policy(*) including direction of policy.
2695  *
2696  * m will always be freed.
2697  */
2698 static int
2699 key_api_spdget(struct socket *so, struct mbuf *m,
2700 	   const struct sadb_msghdr *mhp)
2701 {
2702 	u_int32_t id;
2703 	struct secpolicy *sp;
2704 	struct mbuf *n;
2705 	const struct sadb_x_policy *xpl;
2706 
2707 	if (mhp->ext[SADB_X_EXT_POLICY] == NULL ||
2708 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2709 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2710 		return key_senderror(so, m, EINVAL);
2711 	}
2712 
2713 	xpl = mhp->ext[SADB_X_EXT_POLICY];
2714 	id = xpl->sadb_x_policy_id;
2715 
2716 	/* Is there SP in SPD ? */
2717 	sp = key_getspbyid(id);
2718 	if (sp == NULL) {
2719 		IPSECLOG(LOG_DEBUG, "no SP found id:%u.\n", id);
2720 		return key_senderror(so, m, ENOENT);
2721 	}
2722 
2723 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2724 	    mhp->msg->sadb_msg_pid);
2725 	KEY_SP_UNREF(&sp); /* ref gained by key_getspbyid */
2726 	m_freem(m);
2727 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2728 }
2729 
2730 #ifdef notyet
2731 /*
2732  * SADB_X_SPDACQUIRE processing.
2733  * Acquire policy and SA(s) for a *OUTBOUND* packet.
2734  * send
2735  *   <base, policy(*)>
2736  * to KMD, and expect to receive
2737  *   <base> with SADB_X_SPDACQUIRE if error occurred,
2738  * or
2739  *   <base, policy>
2740  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2741  * policy(*) is without policy requests.
2742  *
2743  *    0     : succeed
2744  *    others: error number
2745  */
2746 int
2747 key_spdacquire(const struct secpolicy *sp)
2748 {
2749 	struct mbuf *result = NULL, *m;
2750 	struct secspacq *newspacq;
2751 	int error;
2752 
2753 	KASSERT(sp != NULL);
2754 	KASSERTMSG(sp->req == NULL, "called but there is request");
2755 	KASSERTMSG(sp->policy == IPSEC_POLICY_IPSEC,
2756 	    "policy mismathed. IPsec is expected");
2757 
2758 	/* Get an entry to check whether sent message or not. */
2759 	newspacq = key_getspacq(&sp->spidx);
2760 	if (newspacq != NULL) {
2761 		if (key_blockacq_count < newspacq->count) {
2762 			/* reset counter and do send message. */
2763 			newspacq->count = 0;
2764 		} else {
2765 			/* increment counter and do nothing. */
2766 			newspacq->count++;
2767 			return 0;
2768 		}
2769 	} else {
2770 		/* make new entry for blocking to send SADB_ACQUIRE. */
2771 		newspacq = key_newspacq(&sp->spidx);
2772 		if (newspacq == NULL)
2773 			return ENOBUFS;
2774 
2775 		/* add to key_misc.acqlist */
2776 		LIST_INSERT_HEAD(&key_misc.spacqlist, newspacq, chain);
2777 	}
2778 
2779 	/* create new sadb_msg to reply. */
2780 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2781 	if (!m) {
2782 		error = ENOBUFS;
2783 		goto fail;
2784 	}
2785 	result = m;
2786 
2787 	result->m_pkthdr.len = 0;
2788 	for (m = result; m; m = m->m_next)
2789 		result->m_pkthdr.len += m->m_len;
2790 
2791 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2792 	    PFKEY_UNIT64(result->m_pkthdr.len);
2793 
2794 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2795 
2796 fail:
2797 	m_freem(result);
2798 	return error;
2799 }
2800 #endif /* notyet */
2801 
2802 /*
2803  * SADB_SPDFLUSH processing
2804  * receive
2805  *   <base>
2806  * from the user, and free all entries in secpctree.
2807  * and send,
2808  *   <base>
2809  * to the user.
2810  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2811  *
2812  * m will always be freed.
2813  */
2814 static int
2815 key_api_spdflush(struct socket *so, struct mbuf *m,
2816 	     const struct sadb_msghdr *mhp)
2817 {
2818 	struct sadb_msg *newmsg;
2819 	struct secpolicy *sp;
2820 	u_int dir;
2821 
2822 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2823 		return key_senderror(so, m, EINVAL);
2824 
2825 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2826 	    retry:
2827 		mutex_enter(&key_spd.lock);
2828 		SPLIST_WRITER_FOREACH(sp, dir) {
2829 			KASSERTMSG(sp->state != IPSEC_SPSTATE_DEAD,
2830 			    "sp->state=%u", sp->state);
2831 			/*
2832 			 * Userlang programs can remove SPs created by userland
2833 			 * probrams only, that is, they cannot remove SPs
2834 			 * created in kernel(e.g. ipsec(4) I/F).
2835 			 */
2836 			if (sp->origin == IPSEC_SPORIGIN_USER) {
2837 				key_unlink_sp(sp);
2838 				mutex_exit(&key_spd.lock);
2839 				key_destroy_sp(sp);
2840 				goto retry;
2841 			}
2842 		}
2843 		mutex_exit(&key_spd.lock);
2844 	}
2845 
2846 	/* We're deleting policy; no need to invalidate the ipflow cache. */
2847 
2848 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2849 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
2850 		return key_senderror(so, m, ENOBUFS);
2851 	}
2852 
2853 	m_freem(m->m_next);
2854 	m->m_next = NULL;
2855 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2856 	newmsg = mtod(m, struct sadb_msg *);
2857 	newmsg->sadb_msg_errno = 0;
2858 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2859 
2860 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2861 }
2862 
2863 static struct sockaddr key_src = {
2864 	.sa_len = 2,
2865 	.sa_family = PF_KEY,
2866 };
2867 
2868 static struct mbuf *
2869 key_setspddump_chain(int *errorp, int *lenp, pid_t pid)
2870 {
2871 	struct secpolicy *sp;
2872 	int cnt;
2873 	u_int dir;
2874 	struct mbuf *m, *n, *prev;
2875 	int totlen;
2876 
2877 	KASSERT(mutex_owned(&key_spd.lock));
2878 
2879 	*lenp = 0;
2880 
2881 	/* search SPD entry and get buffer size. */
2882 	cnt = 0;
2883 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2884 		SPLIST_WRITER_FOREACH(sp, dir) {
2885 			cnt++;
2886 		}
2887 	}
2888 
2889 	if (cnt == 0) {
2890 		*errorp = ENOENT;
2891 		return (NULL);
2892 	}
2893 
2894 	m = NULL;
2895 	prev = m;
2896 	totlen = 0;
2897 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2898 		SPLIST_WRITER_FOREACH(sp, dir) {
2899 			--cnt;
2900 			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, pid);
2901 
2902 			totlen += n->m_pkthdr.len;
2903 			if (!m) {
2904 				m = n;
2905 			} else {
2906 				prev->m_nextpkt = n;
2907 			}
2908 			prev = n;
2909 		}
2910 	}
2911 
2912 	*lenp = totlen;
2913 	*errorp = 0;
2914 	return (m);
2915 }
2916 
2917 /*
2918  * SADB_SPDDUMP processing
2919  * receive
2920  *   <base>
2921  * from the user, and dump all SP leaves
2922  * and send,
2923  *   <base> .....
2924  * to the ikmpd.
2925  *
2926  * m will always be freed.
2927  */
2928 static int
2929 key_api_spddump(struct socket *so, struct mbuf *m0,
2930  	    const struct sadb_msghdr *mhp)
2931 {
2932 	struct mbuf *n;
2933 	int error, len;
2934 	int ok;
2935 	pid_t pid;
2936 
2937 	pid = mhp->msg->sadb_msg_pid;
2938 	/*
2939 	 * If the requestor has insufficient socket-buffer space
2940 	 * for the entire chain, nobody gets any response to the DUMP.
2941 	 * XXX For now, only the requestor ever gets anything.
2942 	 * Moreover, if the requestor has any space at all, they receive
2943 	 * the entire chain, otherwise the request is refused with  ENOBUFS.
2944 	 */
2945 	if (sbspace(&so->so_rcv) <= 0) {
2946 		return key_senderror(so, m0, ENOBUFS);
2947 	}
2948 
2949 	mutex_enter(&key_spd.lock);
2950 	n = key_setspddump_chain(&error, &len, pid);
2951 	mutex_exit(&key_spd.lock);
2952 
2953 	if (n == NULL) {
2954 		return key_senderror(so, m0, ENOENT);
2955 	}
2956 	{
2957 		net_stat_ref_t ps = PFKEY_STAT_GETREF();
2958 		_NET_STATINC_REF(ps, PFKEY_STAT_IN_TOTAL);
2959 		_NET_STATADD_REF(ps, PFKEY_STAT_IN_BYTES, len);
2960 		PFKEY_STAT_PUTREF();
2961 	}
2962 
2963 	/*
2964 	 * PF_KEY DUMP responses are no longer broadcast to all PF_KEY sockets.
2965 	 * The requestor receives either the entire chain, or an
2966 	 * error message with ENOBUFS.
2967 	 */
2968 
2969 	/*
2970 	 * sbappendchainwith record takes the chain of entries, one
2971 	 * packet-record per SPD entry, prepends the key_src sockaddr
2972 	 * to each packet-record, links the sockaddr mbufs into a new
2973 	 * list of records, then   appends the entire resulting
2974 	 * list to the requesting socket.
2975 	 */
2976 	ok = sbappendaddrchain(&so->so_rcv, (struct sockaddr *)&key_src, n,
2977 	    SB_PRIO_ONESHOT_OVERFLOW);
2978 
2979 	if (!ok) {
2980 		PFKEY_STATINC(PFKEY_STAT_IN_NOMEM);
2981 		m_freem(n);
2982 		return key_senderror(so, m0, ENOBUFS);
2983 	}
2984 
2985 	m_freem(m0);
2986 	return error;
2987 }
2988 
2989 /*
2990  * SADB_X_NAT_T_NEW_MAPPING. Unused by racoon as of 2005/04/23
2991  */
2992 static int
2993 key_api_nat_map(struct socket *so, struct mbuf *m,
2994 	    const struct sadb_msghdr *mhp)
2995 {
2996 	struct sadb_x_nat_t_type *type;
2997 	struct sadb_x_nat_t_port *sport;
2998 	struct sadb_x_nat_t_port *dport;
2999 	struct sadb_address *iaddr, *raddr;
3000 	struct sadb_x_nat_t_frag *frag;
3001 
3002 	if (mhp->ext[SADB_X_EXT_NAT_T_TYPE] == NULL ||
3003 	    mhp->ext[SADB_X_EXT_NAT_T_SPORT] == NULL ||
3004 	    mhp->ext[SADB_X_EXT_NAT_T_DPORT] == NULL) {
3005 		IPSECLOG(LOG_DEBUG, "invalid message.\n");
3006 		return key_senderror(so, m, EINVAL);
3007 	}
3008 	if ((mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) ||
3009 	    (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) ||
3010 	    (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport))) {
3011 		IPSECLOG(LOG_DEBUG, "invalid message.\n");
3012 		return key_senderror(so, m, EINVAL);
3013 	}
3014 
3015 	if ((mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL) &&
3016 	    (mhp->extlen[SADB_X_EXT_NAT_T_OAI] < sizeof(*iaddr))) {
3017 		IPSECLOG(LOG_DEBUG, "invalid message\n");
3018 		return key_senderror(so, m, EINVAL);
3019 	}
3020 
3021 	if ((mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL) &&
3022 	    (mhp->extlen[SADB_X_EXT_NAT_T_OAR] < sizeof(*raddr))) {
3023 		IPSECLOG(LOG_DEBUG, "invalid message\n");
3024 		return key_senderror(so, m, EINVAL);
3025 	}
3026 
3027 	if ((mhp->ext[SADB_X_EXT_NAT_T_FRAG] != NULL) &&
3028 	    (mhp->extlen[SADB_X_EXT_NAT_T_FRAG] < sizeof(*frag))) {
3029 		IPSECLOG(LOG_DEBUG, "invalid message\n");
3030 		return key_senderror(so, m, EINVAL);
3031 	}
3032 
3033 	type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
3034 	sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
3035 	dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
3036 	iaddr = mhp->ext[SADB_X_EXT_NAT_T_OAI];
3037 	raddr = mhp->ext[SADB_X_EXT_NAT_T_OAR];
3038 	frag = mhp->ext[SADB_X_EXT_NAT_T_FRAG];
3039 
3040 	/*
3041 	 * XXX handle that, it should also contain a SA, or anything
3042 	 * that enable to update the SA information.
3043 	 */
3044 
3045 	return 0;
3046 }
3047 
3048 /*
3049  * Never return NULL.
3050  */
3051 static struct mbuf *
3052 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq, pid_t pid)
3053 {
3054 	struct mbuf *result = NULL, *m;
3055 
3056 	KASSERT(!cpu_softintr_p());
3057 
3058 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid,
3059 	    key_sp_refcnt(sp), M_WAITOK);
3060 	result = m;
3061 
3062 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3063 	    &sp->spidx.src.sa, sp->spidx.prefs, sp->spidx.ul_proto, M_WAITOK);
3064 	m_cat(result, m);
3065 
3066 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3067 	    &sp->spidx.dst.sa, sp->spidx.prefd, sp->spidx.ul_proto, M_WAITOK);
3068 	m_cat(result, m);
3069 
3070 	m = key_sp2msg(sp, M_WAITOK);
3071 	m_cat(result, m);
3072 
3073 	KASSERT(result->m_flags & M_PKTHDR);
3074 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
3075 
3076 	result->m_pkthdr.len = 0;
3077 	for (m = result; m; m = m->m_next)
3078 		result->m_pkthdr.len += m->m_len;
3079 
3080 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3081 	    PFKEY_UNIT64(result->m_pkthdr.len);
3082 
3083 	return result;
3084 }
3085 
3086 /*
3087  * get PFKEY message length for security policy and request.
3088  */
3089 static u_int
3090 key_getspreqmsglen(const struct secpolicy *sp)
3091 {
3092 	u_int tlen;
3093 
3094 	tlen = sizeof(struct sadb_x_policy);
3095 
3096 	/* if is the policy for ipsec ? */
3097 	if (sp->policy != IPSEC_POLICY_IPSEC)
3098 		return tlen;
3099 
3100 	/* get length of ipsec requests */
3101     {
3102 	const struct ipsecrequest *isr;
3103 	int len;
3104 
3105 	for (isr = sp->req; isr != NULL; isr = isr->next) {
3106 		len = sizeof(struct sadb_x_ipsecrequest)
3107 		    + isr->saidx.src.sa.sa_len + isr->saidx.dst.sa.sa_len;
3108 
3109 		tlen += PFKEY_ALIGN8(len);
3110 	}
3111     }
3112 
3113 	return tlen;
3114 }
3115 
3116 /*
3117  * SADB_SPDEXPIRE processing
3118  * send
3119  *   <base, address(SD), lifetime(CH), policy>
3120  * to KMD by PF_KEY.
3121  *
3122  * OUT:	0	: succeed
3123  *	others	: error number
3124  */
3125 static int
3126 key_spdexpire(struct secpolicy *sp)
3127 {
3128 	int s;
3129 	struct mbuf *result = NULL, *m;
3130 	int len;
3131 	int error = -1;
3132 	struct sadb_lifetime *lt;
3133 
3134 	/* XXX: Why do we lock ? */
3135 	s = splsoftnet();	/*called from softclock()*/
3136 
3137 	KASSERT(sp != NULL);
3138 
3139 	/* set msg header */
3140 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0, M_WAITOK);
3141 	result = m;
3142 
3143 	/* create lifetime extension (current and hard) */
3144 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
3145 	m = key_alloc_mbuf(len, M_WAITOK);
3146 	KASSERT(m->m_next == NULL);
3147 
3148 	memset(mtod(m, void *), 0, len);
3149 	lt = mtod(m, struct sadb_lifetime *);
3150 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
3151 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
3152 	lt->sadb_lifetime_allocations = 0;
3153 	lt->sadb_lifetime_bytes = 0;
3154 	lt->sadb_lifetime_addtime = time_mono_to_wall(sp->created);
3155 	lt->sadb_lifetime_usetime = time_mono_to_wall(sp->lastused);
3156 	lt = (struct sadb_lifetime *)(mtod(m, char *) + len / 2);
3157 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
3158 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
3159 	lt->sadb_lifetime_allocations = 0;
3160 	lt->sadb_lifetime_bytes = 0;
3161 	lt->sadb_lifetime_addtime = sp->lifetime;
3162 	lt->sadb_lifetime_usetime = sp->validtime;
3163 	m_cat(result, m);
3164 
3165 	/* set sadb_address for source */
3166 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sp->spidx.src.sa,
3167 	    sp->spidx.prefs, sp->spidx.ul_proto, M_WAITOK);
3168 	m_cat(result, m);
3169 
3170 	/* set sadb_address for destination */
3171 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sp->spidx.dst.sa,
3172 	    sp->spidx.prefd, sp->spidx.ul_proto, M_WAITOK);
3173 	m_cat(result, m);
3174 
3175 	/* set secpolicy */
3176 	m = key_sp2msg(sp, M_WAITOK);
3177 	m_cat(result, m);
3178 
3179 	KASSERT(result->m_flags & M_PKTHDR);
3180 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
3181 
3182 	result->m_pkthdr.len = 0;
3183 	for (m = result; m; m = m->m_next)
3184 		result->m_pkthdr.len += m->m_len;
3185 
3186 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3187 	    PFKEY_UNIT64(result->m_pkthdr.len);
3188 
3189 	error = key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
3190 	splx(s);
3191 	return error;
3192 }
3193 
3194 /* %%% SAD management */
3195 /*
3196  * allocating a memory for new SA head, and copy from the values of mhp.
3197  * OUT:	NULL	: failure due to the lack of memory.
3198  *	others	: pointer to new SA head.
3199  */
3200 static struct secashead *
3201 key_newsah(const struct secasindex *saidx)
3202 {
3203 	struct secashead *newsah;
3204 	int i;
3205 
3206 	KASSERT(saidx != NULL);
3207 
3208 	newsah = kmem_zalloc(sizeof(struct secashead), KM_SLEEP);
3209 	for (i = 0; i < __arraycount(newsah->savlist); i++)
3210 		PSLIST_INIT(&newsah->savlist[i]);
3211 	newsah->saidx = *saidx;
3212 
3213 	localcount_init(&newsah->localcount);
3214 	/* Take a reference for the caller */
3215 	localcount_acquire(&newsah->localcount);
3216 
3217 	/* Add to the sah list */
3218 	SAHLIST_ENTRY_INIT(newsah);
3219 	newsah->state = SADB_SASTATE_MATURE;
3220 	mutex_enter(&key_sad.lock);
3221 	SAHLIST_WRITER_INSERT_HEAD(newsah);
3222 	mutex_exit(&key_sad.lock);
3223 
3224 	return newsah;
3225 }
3226 
3227 static bool
3228 key_sah_has_sav(struct secashead *sah)
3229 {
3230 	u_int state;
3231 
3232 	KASSERT(mutex_owned(&key_sad.lock));
3233 
3234 	SASTATE_ANY_FOREACH(state) {
3235 		if (!SAVLIST_WRITER_EMPTY(sah, state))
3236 			return true;
3237 	}
3238 
3239 	return false;
3240 }
3241 
3242 static void
3243 key_unlink_sah(struct secashead *sah)
3244 {
3245 
3246 	KASSERT(!cpu_softintr_p());
3247 	KASSERT(mutex_owned(&key_sad.lock));
3248 	KASSERTMSG(sah->state == SADB_SASTATE_DEAD, "sah->state=%u", sah->state);
3249 
3250 	/* Remove from the sah list */
3251 	SAHLIST_WRITER_REMOVE(sah);
3252 
3253 	KDASSERT(mutex_ownable(softnet_lock));
3254 	key_sad_pserialize_perform();
3255 
3256 	localcount_drain(&sah->localcount, &key_sad.cv_lc, &key_sad.lock);
3257 }
3258 
3259 static void
3260 key_destroy_sah(struct secashead *sah)
3261 {
3262 
3263 	rtcache_free(&sah->sa_route);
3264 
3265 	SAHLIST_ENTRY_DESTROY(sah);
3266 	localcount_fini(&sah->localcount);
3267 
3268 	if (sah->idents != NULL)
3269 		kmem_free(sah->idents, sah->idents_len);
3270 	if (sah->identd != NULL)
3271 		kmem_free(sah->identd, sah->identd_len);
3272 
3273 	kmem_free(sah, sizeof(*sah));
3274 }
3275 
3276 /*
3277  * allocating a new SA with LARVAL state.
3278  * key_api_add() and key_api_getspi() call,
3279  * and copy the values of mhp into new buffer.
3280  * When SAD message type is GETSPI:
3281  *	to set sequence number from acq_seq++,
3282  *	to set zero to SPI.
3283  *	not to call key_setsaval().
3284  * OUT:	NULL	: fail
3285  *	others	: pointer to new secasvar.
3286  *
3287  * does not modify mbuf.  does not free mbuf on error.
3288  */
3289 static struct secasvar *
3290 key_newsav(struct mbuf *m, const struct sadb_msghdr *mhp,
3291     int *errp, int proto, const char* where, int tag)
3292 {
3293 	struct secasvar *newsav;
3294 	const struct sadb_sa *xsa;
3295 
3296 	KASSERT(!cpu_softintr_p());
3297 	KASSERT(m != NULL);
3298 	KASSERT(mhp != NULL);
3299 	KASSERT(mhp->msg != NULL);
3300 
3301 	newsav = kmem_zalloc(sizeof(struct secasvar), KM_SLEEP);
3302 
3303 	switch (mhp->msg->sadb_msg_type) {
3304 	case SADB_GETSPI:
3305 		newsav->spi = 0;
3306 
3307 #ifdef IPSEC_DOSEQCHECK
3308 		/* sync sequence number */
3309 		if (mhp->msg->sadb_msg_seq == 0)
3310 			newsav->seq =
3311 			    (acq_seq = (acq_seq == ~0 ? 1 : ++acq_seq));
3312 		else
3313 #endif
3314 			newsav->seq = mhp->msg->sadb_msg_seq;
3315 		break;
3316 
3317 	case SADB_ADD:
3318 		/* sanity check */
3319 		if (mhp->ext[SADB_EXT_SA] == NULL) {
3320 			IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
3321 			*errp = EINVAL;
3322 			goto error;
3323 		}
3324 		xsa = mhp->ext[SADB_EXT_SA];
3325 		newsav->spi = xsa->sadb_sa_spi;
3326 		newsav->seq = mhp->msg->sadb_msg_seq;
3327 		break;
3328 	default:
3329 		*errp = EINVAL;
3330 		goto error;
3331 	}
3332 
3333 	/* copy sav values */
3334 	if (mhp->msg->sadb_msg_type != SADB_GETSPI) {
3335 		*errp = key_setsaval(newsav, m, mhp);
3336 		if (*errp)
3337 			goto error;
3338 	} else {
3339 		/* We don't allow lft_c to be NULL */
3340 		newsav->lft_c = kmem_zalloc(sizeof(struct sadb_lifetime),
3341 		    KM_SLEEP);
3342 		newsav->lft_c_counters_percpu =
3343 		    percpu_alloc(sizeof(lifetime_counters_t));
3344 	}
3345 
3346 	/* reset created */
3347 	newsav->created = time_uptime;
3348 	newsav->pid = mhp->msg->sadb_msg_pid;
3349 
3350 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
3351 	    "DP from %s:%u return SA:%p spi=%#x proto=%d\n",
3352 	    where, tag, newsav, ntohl(newsav->spi), proto);
3353 	return newsav;
3354 
3355 error:
3356 	KASSERT(*errp != 0);
3357 	kmem_free(newsav, sizeof(*newsav));
3358 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
3359 	    "DP from %s:%u return SA:NULL\n", where, tag);
3360 	return NULL;
3361 }
3362 
3363 
3364 static void
3365 key_clear_xform(struct secasvar *sav)
3366 {
3367 
3368 	/*
3369 	 * Cleanup xform state.  Note that zeroize'ing causes the
3370 	 * keys to be cleared; otherwise we must do it ourself.
3371 	 */
3372 	if (sav->tdb_xform != NULL) {
3373 		sav->tdb_xform->xf_zeroize(sav);
3374 		sav->tdb_xform = NULL;
3375 	} else {
3376 		if (sav->key_auth != NULL)
3377 			explicit_memset(_KEYBUF(sav->key_auth), 0,
3378 			    _KEYLEN(sav->key_auth));
3379 		if (sav->key_enc != NULL)
3380 			explicit_memset(_KEYBUF(sav->key_enc), 0,
3381 			    _KEYLEN(sav->key_enc));
3382 	}
3383 }
3384 
3385 /*
3386  * free() SA variable entry.
3387  */
3388 static void
3389 key_delsav(struct secasvar *sav)
3390 {
3391 
3392 	key_clear_xform(sav);
3393 	key_freesaval(sav);
3394 	kmem_free(sav, sizeof(*sav));
3395 }
3396 
3397 /*
3398  * Must be called in a pserialize read section. A held sah
3399  * must be released by key_sah_unref after use.
3400  */
3401 static void
3402 key_sah_ref(struct secashead *sah)
3403 {
3404 
3405 	localcount_acquire(&sah->localcount);
3406 }
3407 
3408 /*
3409  * Must be called without holding key_sad.lock because the lock
3410  * would be held in localcount_release.
3411  */
3412 static void
3413 key_sah_unref(struct secashead *sah)
3414 {
3415 
3416 	KDASSERT(mutex_ownable(&key_sad.lock));
3417 
3418 	localcount_release(&sah->localcount, &key_sad.cv_lc, &key_sad.lock);
3419 }
3420 
3421 /*
3422  * Search SAD and return sah. Must be called in a pserialize
3423  * read section.
3424  * OUT:
3425  *	NULL	: not found
3426  *	others	: found, pointer to a SA.
3427  */
3428 static struct secashead *
3429 key_getsah(const struct secasindex *saidx, int flag)
3430 {
3431 	struct secashead *sah;
3432 
3433 	SAHLIST_READER_FOREACH_SAIDX(sah, saidx) {
3434 		if (sah->state == SADB_SASTATE_DEAD)
3435 			continue;
3436 		if (key_saidx_match(&sah->saidx, saidx, flag))
3437 			return sah;
3438 	}
3439 
3440 	return NULL;
3441 }
3442 
3443 /*
3444  * Search SAD and return sah. If sah is returned, the caller must call
3445  * key_sah_unref to releaset a reference.
3446  * OUT:
3447  *	NULL	: not found
3448  *	others	: found, pointer to a SA.
3449  */
3450 static struct secashead *
3451 key_getsah_ref(const struct secasindex *saidx, int flag)
3452 {
3453 	struct secashead *sah;
3454 	int s;
3455 
3456 	s = pserialize_read_enter();
3457 	sah = key_getsah(saidx, flag);
3458 	if (sah != NULL)
3459 		key_sah_ref(sah);
3460 	pserialize_read_exit(s);
3461 
3462 	return sah;
3463 }
3464 
3465 /*
3466  * check not to be duplicated SPI.
3467  * NOTE: this function is too slow due to searching all SAD.
3468  * OUT:
3469  *	NULL	: not found
3470  *	others	: found, pointer to a SA.
3471  */
3472 static bool
3473 key_checkspidup(const struct secasindex *saidx, u_int32_t spi)
3474 {
3475 	struct secashead *sah;
3476 	struct secasvar *sav;
3477 
3478 	/* check address family */
3479 	if (saidx->src.sa.sa_family != saidx->dst.sa.sa_family) {
3480 		IPSECLOG(LOG_DEBUG,
3481 		    "address family mismatched src %u, dst %u.\n",
3482 		    saidx->src.sa.sa_family, saidx->dst.sa.sa_family);
3483 		return false;
3484 	}
3485 
3486 	/* check all SAD */
3487 	/* key_ismyaddr may sleep, so use mutex, not pserialize, here. */
3488 	mutex_enter(&key_sad.lock);
3489 	SAHLIST_WRITER_FOREACH(sah) {
3490 		if (!key_ismyaddr((struct sockaddr *)&sah->saidx.dst))
3491 			continue;
3492 		sav = key_getsavbyspi(sah, spi);
3493 		if (sav != NULL) {
3494 			KEY_SA_UNREF(&sav);
3495 			mutex_exit(&key_sad.lock);
3496 			return true;
3497 		}
3498 	}
3499 	mutex_exit(&key_sad.lock);
3500 
3501 	return false;
3502 }
3503 
3504 /*
3505  * search SAD litmited alive SA, protocol, SPI.
3506  * OUT:
3507  *	NULL	: not found
3508  *	others	: found, pointer to a SA.
3509  */
3510 static struct secasvar *
3511 key_getsavbyspi(struct secashead *sah, u_int32_t spi)
3512 {
3513 	struct secasvar *sav = NULL;
3514 	u_int state;
3515 	int s;
3516 
3517 	/* search all status */
3518 	s = pserialize_read_enter();
3519 	SASTATE_ALIVE_FOREACH(state) {
3520 		SAVLIST_READER_FOREACH(sav, sah, state) {
3521 			/* sanity check */
3522 			if (sav->state != state) {
3523 				IPSECLOG(LOG_DEBUG,
3524 				    "invalid sav->state (queue: %d SA: %d)\n",
3525 				    state, sav->state);
3526 				continue;
3527 			}
3528 
3529 			if (sav->spi == spi) {
3530 				KEY_SA_REF(sav);
3531 				goto out;
3532 			}
3533 		}
3534 	}
3535 out:
3536 	pserialize_read_exit(s);
3537 
3538 	return sav;
3539 }
3540 
3541 /*
3542  * Search SAD litmited alive SA by an SPI and remove it from a list.
3543  * OUT:
3544  *	NULL	: not found
3545  *	others	: found, pointer to a SA.
3546  */
3547 static struct secasvar *
3548 key_lookup_and_remove_sav(struct secashead *sah, u_int32_t spi,
3549     const struct secasvar *hint)
3550 {
3551 	struct secasvar *sav = NULL;
3552 	u_int state;
3553 
3554 	/* search all status */
3555 	mutex_enter(&key_sad.lock);
3556 	SASTATE_ALIVE_FOREACH(state) {
3557 		SAVLIST_WRITER_FOREACH(sav, sah, state) {
3558 			KASSERT(sav->state == state);
3559 
3560 			if (sav->spi == spi) {
3561 				if (hint != NULL && hint != sav)
3562 					continue;
3563 				sav->state = SADB_SASTATE_DEAD;
3564 				SAVLIST_WRITER_REMOVE(sav);
3565 				SAVLUT_WRITER_REMOVE(sav);
3566 				goto out;
3567 			}
3568 		}
3569 	}
3570 out:
3571 	mutex_exit(&key_sad.lock);
3572 
3573 	return sav;
3574 }
3575 
3576 /*
3577  * Free allocated data to member variables of sav:
3578  * sav->replay, sav->key_* and sav->lft_*.
3579  */
3580 static void
3581 key_freesaval(struct secasvar *sav)
3582 {
3583 
3584 	KASSERTMSG(key_sa_refcnt(sav) == 0, "key_sa_refcnt(sav)=%u",
3585 	    key_sa_refcnt(sav));
3586 
3587 	if (sav->replay != NULL)
3588 		kmem_free(sav->replay, sav->replay_len);
3589 	if (sav->key_auth != NULL)
3590 		kmem_free(sav->key_auth, sav->key_auth_len);
3591 	if (sav->key_enc != NULL)
3592 		kmem_free(sav->key_enc, sav->key_enc_len);
3593 	if (sav->lft_c_counters_percpu != NULL) {
3594 		percpu_free(sav->lft_c_counters_percpu,
3595 		    sizeof(lifetime_counters_t));
3596 	}
3597 	if (sav->lft_c != NULL)
3598 		kmem_free(sav->lft_c, sizeof(*(sav->lft_c)));
3599 	if (sav->lft_h != NULL)
3600 		kmem_free(sav->lft_h, sizeof(*(sav->lft_h)));
3601 	if (sav->lft_s != NULL)
3602 		kmem_free(sav->lft_s, sizeof(*(sav->lft_s)));
3603 }
3604 
3605 /*
3606  * copy SA values from PF_KEY message except *SPI, SEQ, PID, STATE and TYPE*.
3607  * You must update these if need.
3608  * OUT:	0:	success.
3609  *	!0:	failure.
3610  *
3611  * does not modify mbuf.  does not free mbuf on error.
3612  */
3613 static int
3614 key_setsaval(struct secasvar *sav, struct mbuf *m,
3615 	     const struct sadb_msghdr *mhp)
3616 {
3617 	int error = 0;
3618 
3619 	KASSERT(!cpu_softintr_p());
3620 	KASSERT(m != NULL);
3621 	KASSERT(mhp != NULL);
3622 	KASSERT(mhp->msg != NULL);
3623 
3624 	/* We shouldn't initialize sav variables while someone uses it. */
3625 	KASSERTMSG(key_sa_refcnt(sav) == 0, "key_sa_refcnt(sav)=%u",
3626 	    key_sa_refcnt(sav));
3627 
3628 	/* SA */
3629 	if (mhp->ext[SADB_EXT_SA] != NULL) {
3630 		const struct sadb_sa *sa0;
3631 
3632 		sa0 = mhp->ext[SADB_EXT_SA];
3633 		if (mhp->extlen[SADB_EXT_SA] < sizeof(*sa0)) {
3634 			error = EINVAL;
3635 			goto fail;
3636 		}
3637 
3638 		sav->alg_auth = sa0->sadb_sa_auth;
3639 		sav->alg_enc = sa0->sadb_sa_encrypt;
3640 		sav->flags = sa0->sadb_sa_flags;
3641 
3642 		/* replay window */
3643 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0) {
3644 			size_t len = sizeof(struct secreplay) +
3645 			    sa0->sadb_sa_replay;
3646 			sav->replay = kmem_zalloc(len, KM_SLEEP);
3647 			sav->replay_len = len;
3648 			if (sa0->sadb_sa_replay != 0)
3649 				sav->replay->bitmap = (char*)(sav->replay+1);
3650 			sav->replay->wsize = sa0->sadb_sa_replay;
3651 		}
3652 	}
3653 
3654 	/* Authentication keys */
3655 	if (mhp->ext[SADB_EXT_KEY_AUTH] != NULL) {
3656 		const struct sadb_key *key0;
3657 		int len;
3658 
3659 		key0 = mhp->ext[SADB_EXT_KEY_AUTH];
3660 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3661 
3662 		error = 0;
3663 		if (len < sizeof(*key0)) {
3664 			error = EINVAL;
3665 			goto fail;
3666 		}
3667 		switch (mhp->msg->sadb_msg_satype) {
3668 		case SADB_SATYPE_AH:
3669 		case SADB_SATYPE_ESP:
3670 		case SADB_X_SATYPE_TCPSIGNATURE:
3671 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3672 			    sav->alg_auth != SADB_X_AALG_NULL)
3673 				error = EINVAL;
3674 			break;
3675 		case SADB_X_SATYPE_IPCOMP:
3676 		default:
3677 			error = EINVAL;
3678 			break;
3679 		}
3680 		if (error) {
3681 			IPSECLOG(LOG_DEBUG, "invalid key_auth values.\n");
3682 			goto fail;
3683 		}
3684 
3685 		sav->key_auth = key_newbuf(key0, len);
3686 		sav->key_auth_len = len;
3687 	}
3688 
3689 	/* Encryption key */
3690 	if (mhp->ext[SADB_EXT_KEY_ENCRYPT] != NULL) {
3691 		const struct sadb_key *key0;
3692 		int len;
3693 
3694 		key0 = mhp->ext[SADB_EXT_KEY_ENCRYPT];
3695 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3696 
3697 		error = 0;
3698 		if (len < sizeof(*key0)) {
3699 			error = EINVAL;
3700 			goto fail;
3701 		}
3702 		switch (mhp->msg->sadb_msg_satype) {
3703 		case SADB_SATYPE_ESP:
3704 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3705 			    sav->alg_enc != SADB_EALG_NULL) {
3706 				error = EINVAL;
3707 				break;
3708 			}
3709 			sav->key_enc = key_newbuf(key0, len);
3710 			sav->key_enc_len = len;
3711 			break;
3712 		case SADB_X_SATYPE_IPCOMP:
3713 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3714 				error = EINVAL;
3715 			sav->key_enc = NULL;	/*just in case*/
3716 			break;
3717 		case SADB_SATYPE_AH:
3718 		case SADB_X_SATYPE_TCPSIGNATURE:
3719 		default:
3720 			error = EINVAL;
3721 			break;
3722 		}
3723 		if (error) {
3724 			IPSECLOG(LOG_DEBUG, "invalid key_enc value.\n");
3725 			goto fail;
3726 		}
3727 	}
3728 
3729 	/* set iv */
3730 	sav->ivlen = 0;
3731 
3732 	switch (mhp->msg->sadb_msg_satype) {
3733 	case SADB_SATYPE_AH:
3734 		error = xform_init(sav, XF_AH);
3735 		break;
3736 	case SADB_SATYPE_ESP:
3737 		error = xform_init(sav, XF_ESP);
3738 		break;
3739 	case SADB_X_SATYPE_IPCOMP:
3740 		error = xform_init(sav, XF_IPCOMP);
3741 		break;
3742 	case SADB_X_SATYPE_TCPSIGNATURE:
3743 		error = xform_init(sav, XF_TCPSIGNATURE);
3744 		break;
3745 	default:
3746 		error = EOPNOTSUPP;
3747 		break;
3748 	}
3749 	if (error) {
3750 		IPSECLOG(LOG_DEBUG, "unable to initialize SA type %u (%d)\n",
3751 		    mhp->msg->sadb_msg_satype, error);
3752 		goto fail;
3753 	}
3754 
3755 	/* reset created */
3756 	sav->created = time_uptime;
3757 
3758 	/* make lifetime for CURRENT */
3759 	sav->lft_c = kmem_alloc(sizeof(struct sadb_lifetime), KM_SLEEP);
3760 
3761 	sav->lft_c->sadb_lifetime_len =
3762 	    PFKEY_UNIT64(sizeof(struct sadb_lifetime));
3763 	sav->lft_c->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
3764 	sav->lft_c->sadb_lifetime_allocations = 0;
3765 	sav->lft_c->sadb_lifetime_bytes = 0;
3766 	sav->lft_c->sadb_lifetime_addtime = time_uptime;
3767 	sav->lft_c->sadb_lifetime_usetime = 0;
3768 
3769 	sav->lft_c_counters_percpu = percpu_alloc(sizeof(lifetime_counters_t));
3770 
3771 	/* lifetimes for HARD and SOFT */
3772     {
3773 	const struct sadb_lifetime *lft0;
3774 
3775 	lft0 = mhp->ext[SADB_EXT_LIFETIME_HARD];
3776 	if (lft0 != NULL) {
3777 		if (mhp->extlen[SADB_EXT_LIFETIME_HARD] < sizeof(*lft0)) {
3778 			error = EINVAL;
3779 			goto fail;
3780 		}
3781 		sav->lft_h = key_newbuf(lft0, sizeof(*lft0));
3782 	}
3783 
3784 	lft0 = mhp->ext[SADB_EXT_LIFETIME_SOFT];
3785 	if (lft0 != NULL) {
3786 		if (mhp->extlen[SADB_EXT_LIFETIME_SOFT] < sizeof(*lft0)) {
3787 			error = EINVAL;
3788 			goto fail;
3789 		}
3790 		sav->lft_s = key_newbuf(lft0, sizeof(*lft0));
3791 		/* to be initialize ? */
3792 	}
3793     }
3794 
3795 	return 0;
3796 
3797  fail:
3798 	key_clear_xform(sav);
3799 	key_freesaval(sav);
3800 
3801 	return error;
3802 }
3803 
3804 /*
3805  * validation with a secasvar entry, and set SADB_SATYPE_MATURE.
3806  * OUT:	0:	valid
3807  *	other:	errno
3808  */
3809 static int
3810 key_init_xform(struct secasvar *sav)
3811 {
3812 	int error;
3813 
3814 	/* We shouldn't initialize sav variables while someone uses it. */
3815 	KASSERTMSG(key_sa_refcnt(sav) == 0, "key_sa_refcnt(sav)=%u",
3816 	    key_sa_refcnt(sav));
3817 
3818 	/* check SPI value */
3819 	switch (sav->sah->saidx.proto) {
3820 	case IPPROTO_ESP:
3821 	case IPPROTO_AH:
3822 		if (ntohl(sav->spi) <= 255) {
3823 			IPSECLOG(LOG_DEBUG, "illegal range of SPI %u.\n",
3824 			    (u_int32_t)ntohl(sav->spi));
3825 			return EINVAL;
3826 		}
3827 		break;
3828 	}
3829 
3830 	/* check algo */
3831 	switch (sav->sah->saidx.proto) {
3832 	case IPPROTO_AH:
3833 	case IPPROTO_TCP:
3834 		if (sav->alg_enc != SADB_EALG_NONE) {
3835 			IPSECLOG(LOG_DEBUG,
3836 			    "protocol %u and algorithm mismatched %u != %u.\n",
3837 			    sav->sah->saidx.proto,
3838 			    sav->alg_enc, SADB_EALG_NONE);
3839 			return EINVAL;
3840 		}
3841 		break;
3842 	case IPPROTO_IPCOMP:
3843 		if (sav->alg_auth != SADB_AALG_NONE) {
3844 			IPSECLOG(LOG_DEBUG,
3845 			    "protocol %u and algorithm mismatched %d != %d.\n",
3846 			    sav->sah->saidx.proto,
3847 			    sav->alg_auth, SADB_AALG_NONE);
3848 			return(EINVAL);
3849 		}
3850 		break;
3851 	default:
3852 		break;
3853 	}
3854 
3855 	/* check satype */
3856 	switch (sav->sah->saidx.proto) {
3857 	case IPPROTO_ESP:
3858 		/* check flags */
3859 		if ((sav->flags & (SADB_X_EXT_OLD|SADB_X_EXT_DERIV)) ==
3860 		    (SADB_X_EXT_OLD|SADB_X_EXT_DERIV)) {
3861 			IPSECLOG(LOG_DEBUG,
3862 			    "invalid flag (derived) given to old-esp.\n");
3863 			return EINVAL;
3864 		}
3865 		error = xform_init(sav, XF_ESP);
3866 		break;
3867 	case IPPROTO_AH:
3868 		/* check flags */
3869 		if (sav->flags & SADB_X_EXT_DERIV) {
3870 			IPSECLOG(LOG_DEBUG,
3871 			    "invalid flag (derived) given to AH SA.\n");
3872 			return EINVAL;
3873 		}
3874 		error = xform_init(sav, XF_AH);
3875 		break;
3876 	case IPPROTO_IPCOMP:
3877 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0
3878 		    && ntohl(sav->spi) >= 0x10000) {
3879 			IPSECLOG(LOG_DEBUG, "invalid cpi for IPComp.\n");
3880 			return(EINVAL);
3881 		}
3882 		error = xform_init(sav, XF_IPCOMP);
3883 		break;
3884 	case IPPROTO_TCP:
3885 		error = xform_init(sav, XF_TCPSIGNATURE);
3886 		break;
3887 	default:
3888 		IPSECLOG(LOG_DEBUG, "Invalid satype.\n");
3889 		error = EPROTONOSUPPORT;
3890 		break;
3891 	}
3892 
3893 	return error;
3894 }
3895 
3896 /*
3897  * subroutine for SADB_GET and SADB_DUMP. It never return NULL.
3898  */
3899 static struct mbuf *
3900 key_setdumpsa(struct secasvar *sav, u_int8_t type, u_int8_t satype,
3901 	      u_int32_t seq, u_int32_t pid)
3902 {
3903 	struct mbuf *result = NULL, *tres = NULL, *m;
3904 	int l = 0;
3905 	int i;
3906 	void *p;
3907 	struct sadb_lifetime lt;
3908 	int dumporder[] = {
3909 		SADB_EXT_SA, SADB_X_EXT_SA2,
3910 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3911 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3912 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY, SADB_EXT_KEY_AUTH,
3913 		SADB_EXT_KEY_ENCRYPT, SADB_EXT_IDENTITY_SRC,
3914 		SADB_EXT_IDENTITY_DST, SADB_EXT_SENSITIVITY,
3915 		SADB_X_EXT_NAT_T_TYPE,
3916 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3917 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3918 		SADB_X_EXT_NAT_T_FRAG,
3919 
3920 	};
3921 
3922 	m = key_setsadbmsg(type, 0, satype, seq, pid, key_sa_refcnt(sav), M_WAITOK);
3923 	result = m;
3924 
3925 	for (i = __arraycount(dumporder) - 1; i >= 0; i--) {
3926 		m = NULL;
3927 		p = NULL;
3928 		switch (dumporder[i]) {
3929 		case SADB_EXT_SA:
3930 			m = key_setsadbsa(sav);
3931 			break;
3932 
3933 		case SADB_X_EXT_SA2:
3934 			m = key_setsadbxsa2(sav->sah->saidx.mode,
3935 			    sav->replay ? sav->replay->count : 0,
3936 			    sav->sah->saidx.reqid);
3937 			break;
3938 
3939 		case SADB_EXT_ADDRESS_SRC:
3940 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3941 			    &sav->sah->saidx.src.sa,
3942 			    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
3943 			break;
3944 
3945 		case SADB_EXT_ADDRESS_DST:
3946 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3947 			    &sav->sah->saidx.dst.sa,
3948 			    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
3949 			break;
3950 
3951 		case SADB_EXT_KEY_AUTH:
3952 			if (!sav->key_auth)
3953 				continue;
3954 			l = PFKEY_UNUNIT64(sav->key_auth->sadb_key_len);
3955 			p = sav->key_auth;
3956 			break;
3957 
3958 		case SADB_EXT_KEY_ENCRYPT:
3959 			if (!sav->key_enc)
3960 				continue;
3961 			l = PFKEY_UNUNIT64(sav->key_enc->sadb_key_len);
3962 			p = sav->key_enc;
3963 			break;
3964 
3965 		case SADB_EXT_LIFETIME_CURRENT: {
3966 			lifetime_counters_t sum = {0};
3967 
3968 			KASSERT(sav->lft_c != NULL);
3969 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_c)->sadb_ext_len);
3970 			memcpy(&lt, sav->lft_c, sizeof(struct sadb_lifetime));
3971 			lt.sadb_lifetime_addtime =
3972 			    time_mono_to_wall(lt.sadb_lifetime_addtime);
3973 			lt.sadb_lifetime_usetime =
3974 			    time_mono_to_wall(lt.sadb_lifetime_usetime);
3975 			percpu_foreach_xcall(sav->lft_c_counters_percpu,
3976 			    XC_HIGHPRI_IPL(IPL_SOFTNET),
3977 			    key_sum_lifetime_counters, sum);
3978 			lt.sadb_lifetime_allocations =
3979 			    sum[LIFETIME_COUNTER_ALLOCATIONS];
3980 			lt.sadb_lifetime_bytes =
3981 			    sum[LIFETIME_COUNTER_BYTES];
3982 			p = &lt;
3983 			break;
3984 		    }
3985 
3986 		case SADB_EXT_LIFETIME_HARD:
3987 			if (!sav->lft_h)
3988 				continue;
3989 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_h)->sadb_ext_len);
3990 			p = sav->lft_h;
3991 			break;
3992 
3993 		case SADB_EXT_LIFETIME_SOFT:
3994 			if (!sav->lft_s)
3995 				continue;
3996 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_s)->sadb_ext_len);
3997 			p = sav->lft_s;
3998 			break;
3999 
4000 		case SADB_X_EXT_NAT_T_TYPE:
4001 			m = key_setsadbxtype(sav->natt_type);
4002 			break;
4003 
4004 		case SADB_X_EXT_NAT_T_DPORT:
4005 			if (sav->natt_type == 0)
4006 				continue;
4007 			m = key_setsadbxport(
4008 			    key_portfromsaddr(&sav->sah->saidx.dst),
4009 			    SADB_X_EXT_NAT_T_DPORT);
4010 			break;
4011 
4012 		case SADB_X_EXT_NAT_T_SPORT:
4013 			if (sav->natt_type == 0)
4014 				continue;
4015 			m = key_setsadbxport(
4016 			    key_portfromsaddr(&sav->sah->saidx.src),
4017 			    SADB_X_EXT_NAT_T_SPORT);
4018 			break;
4019 
4020 		case SADB_X_EXT_NAT_T_FRAG:
4021 			/* don't send frag info if not set */
4022 			if (sav->natt_type == 0 || sav->esp_frag == IP_MAXPACKET)
4023 				continue;
4024 			m = key_setsadbxfrag(sav->esp_frag);
4025 			break;
4026 
4027 		case SADB_X_EXT_NAT_T_OAI:
4028 		case SADB_X_EXT_NAT_T_OAR:
4029 			continue;
4030 
4031 		case SADB_EXT_ADDRESS_PROXY:
4032 		case SADB_EXT_IDENTITY_SRC:
4033 		case SADB_EXT_IDENTITY_DST:
4034 			/* XXX: should we brought from SPD ? */
4035 		case SADB_EXT_SENSITIVITY:
4036 		default:
4037 			continue;
4038 		}
4039 
4040 		KASSERT(!(m && p));
4041 		KASSERT(m != NULL || p != NULL);
4042 		if (p && tres) {
4043 			M_PREPEND(tres, l, M_WAITOK);
4044 			memcpy(mtod(tres, void *), p, l);
4045 			continue;
4046 		}
4047 		if (p) {
4048 			m = key_alloc_mbuf(l, M_WAITOK);
4049 			m_copyback(m, 0, l, p);
4050 		}
4051 
4052 		if (tres)
4053 			m_cat(m, tres);
4054 		tres = m;
4055 	}
4056 
4057 	m_cat(result, tres);
4058 	tres = NULL; /* avoid free on error below */
4059 
4060 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
4061 
4062 	result->m_pkthdr.len = 0;
4063 	for (m = result; m; m = m->m_next)
4064 		result->m_pkthdr.len += m->m_len;
4065 
4066 	mtod(result, struct sadb_msg *)->sadb_msg_len =
4067 	    PFKEY_UNIT64(result->m_pkthdr.len);
4068 
4069 	return result;
4070 }
4071 
4072 
4073 /*
4074  * set a type in sadb_x_nat_t_type
4075  */
4076 static struct mbuf *
4077 key_setsadbxtype(u_int16_t type)
4078 {
4079 	struct mbuf *m;
4080 	size_t len;
4081 	struct sadb_x_nat_t_type *p;
4082 
4083 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
4084 
4085 	m = key_alloc_mbuf(len, M_WAITOK);
4086 	KASSERT(m->m_next == NULL);
4087 
4088 	p = mtod(m, struct sadb_x_nat_t_type *);
4089 
4090 	memset(p, 0, len);
4091 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
4092 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
4093 	p->sadb_x_nat_t_type_type = type;
4094 
4095 	return m;
4096 }
4097 /*
4098  * set a port in sadb_x_nat_t_port. port is in network order
4099  */
4100 static struct mbuf *
4101 key_setsadbxport(u_int16_t port, u_int16_t type)
4102 {
4103 	struct mbuf *m;
4104 	size_t len;
4105 	struct sadb_x_nat_t_port *p;
4106 
4107 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
4108 
4109 	m = key_alloc_mbuf(len, M_WAITOK);
4110 	KASSERT(m->m_next == NULL);
4111 
4112 	p = mtod(m, struct sadb_x_nat_t_port *);
4113 
4114 	memset(p, 0, len);
4115 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
4116 	p->sadb_x_nat_t_port_exttype = type;
4117 	p->sadb_x_nat_t_port_port = port;
4118 
4119 	return m;
4120 }
4121 
4122 /*
4123  * set fragmentation info in sadb_x_nat_t_frag
4124  */
4125 static struct mbuf *
4126 key_setsadbxfrag(u_int16_t flen)
4127 {
4128 	struct mbuf *m;
4129 	size_t len;
4130 	struct sadb_x_nat_t_frag *p;
4131 
4132 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_frag));
4133 
4134 	m = key_alloc_mbuf(len, M_WAITOK);
4135 	KASSERT(m->m_next == NULL);
4136 
4137 	p = mtod(m, struct sadb_x_nat_t_frag *);
4138 
4139 	memset(p, 0, len);
4140 	p->sadb_x_nat_t_frag_len = PFKEY_UNIT64(len);
4141 	p->sadb_x_nat_t_frag_exttype = SADB_X_EXT_NAT_T_FRAG;
4142 	p->sadb_x_nat_t_frag_fraglen = flen;
4143 
4144 	return m;
4145 }
4146 
4147 /*
4148  * Get port from sockaddr, port is in network order
4149  */
4150 u_int16_t
4151 key_portfromsaddr(const union sockaddr_union *saddr)
4152 {
4153 	u_int16_t port;
4154 
4155 	switch (saddr->sa.sa_family) {
4156 	case AF_INET: {
4157 		port = saddr->sin.sin_port;
4158 		break;
4159 	}
4160 #ifdef INET6
4161 	case AF_INET6: {
4162 		port = saddr->sin6.sin6_port;
4163 		break;
4164 	}
4165 #endif
4166 	default:
4167 		printf("%s: unexpected address family\n", __func__);
4168 		port = 0;
4169 		break;
4170 	}
4171 
4172 	return port;
4173 }
4174 
4175 
4176 /*
4177  * Set port is struct sockaddr. port is in network order
4178  */
4179 static void
4180 key_porttosaddr(union sockaddr_union *saddr, u_int16_t port)
4181 {
4182 	switch (saddr->sa.sa_family) {
4183 	case AF_INET: {
4184 		saddr->sin.sin_port = port;
4185 		break;
4186 	}
4187 #ifdef INET6
4188 	case AF_INET6: {
4189 		saddr->sin6.sin6_port = port;
4190 		break;
4191 	}
4192 #endif
4193 	default:
4194 		printf("%s: unexpected address family %d\n", __func__,
4195 		    saddr->sa.sa_family);
4196 		break;
4197 	}
4198 
4199 	return;
4200 }
4201 
4202 /*
4203  * Safety check sa_len
4204  */
4205 static int
4206 key_checksalen(const union sockaddr_union *saddr)
4207 {
4208 	switch (saddr->sa.sa_family) {
4209 	case AF_INET:
4210 		if (saddr->sa.sa_len != sizeof(struct sockaddr_in))
4211 			return -1;
4212 		break;
4213 #ifdef INET6
4214 	case AF_INET6:
4215 		if (saddr->sa.sa_len != sizeof(struct sockaddr_in6))
4216 			return -1;
4217 		break;
4218 #endif
4219 	default:
4220 		printf("%s: unexpected sa_family %d\n", __func__,
4221 		    saddr->sa.sa_family);
4222 			return -1;
4223 		break;
4224 	}
4225 	return 0;
4226 }
4227 
4228 
4229 /*
4230  * set data into sadb_msg.
4231  */
4232 static struct mbuf *
4233 key_setsadbmsg(u_int8_t type,  u_int16_t tlen, u_int8_t satype,
4234 	       u_int32_t seq, pid_t pid, u_int16_t reserved, int mflag)
4235 {
4236 	struct mbuf *m;
4237 	struct sadb_msg *p;
4238 	int len;
4239 
4240 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MCLBYTES);
4241 
4242 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
4243 
4244 	m = key_alloc_mbuf_simple(len, mflag);
4245 	if (!m)
4246 		return NULL;
4247 	m->m_pkthdr.len = m->m_len = len;
4248 	m->m_next = NULL;
4249 
4250 	p = mtod(m, struct sadb_msg *);
4251 
4252 	memset(p, 0, len);
4253 	p->sadb_msg_version = PF_KEY_V2;
4254 	p->sadb_msg_type = type;
4255 	p->sadb_msg_errno = 0;
4256 	p->sadb_msg_satype = satype;
4257 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
4258 	p->sadb_msg_reserved = reserved;
4259 	p->sadb_msg_seq = seq;
4260 	p->sadb_msg_pid = (u_int32_t)pid;
4261 
4262 	return m;
4263 }
4264 
4265 /*
4266  * copy secasvar data into sadb_address.
4267  */
4268 static struct mbuf *
4269 key_setsadbsa(struct secasvar *sav)
4270 {
4271 	struct mbuf *m;
4272 	struct sadb_sa *p;
4273 	int len;
4274 
4275 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
4276 	m = key_alloc_mbuf(len, M_WAITOK);
4277 	KASSERT(m->m_next == NULL);
4278 
4279 	p = mtod(m, struct sadb_sa *);
4280 
4281 	memset(p, 0, len);
4282 	p->sadb_sa_len = PFKEY_UNIT64(len);
4283 	p->sadb_sa_exttype = SADB_EXT_SA;
4284 	p->sadb_sa_spi = sav->spi;
4285 	p->sadb_sa_replay = (sav->replay != NULL ? sav->replay->wsize : 0);
4286 	p->sadb_sa_state = sav->state;
4287 	p->sadb_sa_auth = sav->alg_auth;
4288 	p->sadb_sa_encrypt = sav->alg_enc;
4289 	p->sadb_sa_flags = sav->flags;
4290 
4291 	return m;
4292 }
4293 
4294 static uint8_t
4295 key_sabits(const struct sockaddr *saddr)
4296 {
4297 	switch (saddr->sa_family) {
4298 	case AF_INET:
4299 		return _BITS(sizeof(struct in_addr));
4300 	case AF_INET6:
4301 		return _BITS(sizeof(struct in6_addr));
4302 	default:
4303 		return FULLMASK;
4304 	}
4305 }
4306 
4307 /*
4308  * set data into sadb_address.
4309  */
4310 static struct mbuf *
4311 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
4312 		u_int8_t prefixlen, u_int16_t ul_proto, int mflag)
4313 {
4314 	struct mbuf *m;
4315 	struct sadb_address *p;
4316 	size_t len;
4317 
4318 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
4319 	    PFKEY_ALIGN8(saddr->sa_len);
4320 	m = key_alloc_mbuf(len, mflag);
4321 	if (!m || m->m_next) {	/*XXX*/
4322 		m_freem(m);
4323 		return NULL;
4324 	}
4325 
4326 	p = mtod(m, struct sadb_address *);
4327 
4328 	memset(p, 0, len);
4329 	p->sadb_address_len = PFKEY_UNIT64(len);
4330 	p->sadb_address_exttype = exttype;
4331 	p->sadb_address_proto = ul_proto;
4332 	if (prefixlen == FULLMASK) {
4333 		prefixlen = key_sabits(saddr);
4334 	}
4335 	p->sadb_address_prefixlen = prefixlen;
4336 	p->sadb_address_reserved = 0;
4337 
4338 	memcpy(mtod(m, char *) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
4339 	    saddr, saddr->sa_len);
4340 
4341 	return m;
4342 }
4343 
4344 #if 0
4345 /*
4346  * set data into sadb_ident.
4347  */
4348 static struct mbuf *
4349 key_setsadbident(u_int16_t exttype, u_int16_t idtype,
4350 		 void *string, int stringlen, u_int64_t id)
4351 {
4352 	struct mbuf *m;
4353 	struct sadb_ident *p;
4354 	size_t len;
4355 
4356 	len = PFKEY_ALIGN8(sizeof(struct sadb_ident)) + PFKEY_ALIGN8(stringlen);
4357 	m = key_alloc_mbuf(len);
4358 	if (!m || m->m_next) {	/*XXX*/
4359 		m_freem(m);
4360 		return NULL;
4361 	}
4362 
4363 	p = mtod(m, struct sadb_ident *);
4364 
4365 	memset(p, 0, len);
4366 	p->sadb_ident_len = PFKEY_UNIT64(len);
4367 	p->sadb_ident_exttype = exttype;
4368 	p->sadb_ident_type = idtype;
4369 	p->sadb_ident_reserved = 0;
4370 	p->sadb_ident_id = id;
4371 
4372 	memcpy(mtod(m, void *) + PFKEY_ALIGN8(sizeof(struct sadb_ident)),
4373 	   	   string, stringlen);
4374 
4375 	return m;
4376 }
4377 #endif
4378 
4379 /*
4380  * set data into sadb_x_sa2.
4381  */
4382 static struct mbuf *
4383 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int16_t reqid)
4384 {
4385 	struct mbuf *m;
4386 	struct sadb_x_sa2 *p;
4387 	size_t len;
4388 
4389 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
4390 	m = key_alloc_mbuf(len, M_WAITOK);
4391 	KASSERT(m->m_next == NULL);
4392 
4393 	p = mtod(m, struct sadb_x_sa2 *);
4394 
4395 	memset(p, 0, len);
4396 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
4397 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
4398 	p->sadb_x_sa2_mode = mode;
4399 	p->sadb_x_sa2_reserved1 = 0;
4400 	p->sadb_x_sa2_reserved2 = 0;
4401 	p->sadb_x_sa2_sequence = seq;
4402 	p->sadb_x_sa2_reqid = reqid;
4403 
4404 	return m;
4405 }
4406 
4407 /*
4408  * set data into sadb_x_policy
4409  */
4410 static struct mbuf *
4411 key_setsadbxpolicy(const u_int16_t type, const u_int8_t dir, const u_int32_t id,
4412     int mflag)
4413 {
4414 	struct mbuf *m;
4415 	struct sadb_x_policy *p;
4416 	size_t len;
4417 
4418 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
4419 	m = key_alloc_mbuf(len, mflag);
4420 	if (!m || m->m_next) {	/*XXX*/
4421 		m_freem(m);
4422 		return NULL;
4423 	}
4424 
4425 	p = mtod(m, struct sadb_x_policy *);
4426 
4427 	memset(p, 0, len);
4428 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
4429 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
4430 	p->sadb_x_policy_type = type;
4431 	p->sadb_x_policy_dir = dir;
4432 	p->sadb_x_policy_id = id;
4433 
4434 	return m;
4435 }
4436 
4437 /* %%% utilities */
4438 /*
4439  * copy a buffer into the new buffer allocated.
4440  */
4441 static void *
4442 key_newbuf(const void *src, u_int len)
4443 {
4444 	void *new;
4445 
4446 	new = kmem_alloc(len, KM_SLEEP);
4447 	memcpy(new, src, len);
4448 
4449 	return new;
4450 }
4451 
4452 /* compare my own address
4453  * OUT:	1: true, i.e. my address.
4454  *	0: false
4455  */
4456 int
4457 key_ismyaddr(const struct sockaddr *sa)
4458 {
4459 #ifdef INET
4460 	const struct sockaddr_in *sin;
4461 	const struct in_ifaddr *ia;
4462 	int s;
4463 #endif
4464 
4465 	KASSERT(sa != NULL);
4466 
4467 	switch (sa->sa_family) {
4468 #ifdef INET
4469 	case AF_INET:
4470 		sin = (const struct sockaddr_in *)sa;
4471 		s = pserialize_read_enter();
4472 		IN_ADDRLIST_READER_FOREACH(ia) {
4473 			if (sin->sin_family == ia->ia_addr.sin_family &&
4474 			    sin->sin_len == ia->ia_addr.sin_len &&
4475 			    sin->sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
4476 			{
4477 				pserialize_read_exit(s);
4478 				return 1;
4479 			}
4480 		}
4481 		pserialize_read_exit(s);
4482 		break;
4483 #endif
4484 #ifdef INET6
4485 	case AF_INET6:
4486 		return key_ismyaddr6((const struct sockaddr_in6 *)sa);
4487 #endif
4488 	}
4489 
4490 	return 0;
4491 }
4492 
4493 #ifdef INET6
4494 /*
4495  * compare my own address for IPv6.
4496  * 1: ours
4497  * 0: other
4498  * NOTE: derived ip6_input() in KAME. This is necessary to modify more.
4499  */
4500 #include <netinet6/in6_var.h>
4501 
4502 static int
4503 key_ismyaddr6(const struct sockaddr_in6 *sin6)
4504 {
4505 	struct in6_ifaddr *ia;
4506 	int s;
4507 	struct psref psref;
4508 	int bound;
4509 	int ours = 1;
4510 
4511 	bound = curlwp_bind();
4512 	s = pserialize_read_enter();
4513 	IN6_ADDRLIST_READER_FOREACH(ia) {
4514 		if (key_sockaddr_match((const struct sockaddr *)&sin6,
4515 		    (const struct sockaddr *)&ia->ia_addr, 0)) {
4516 			pserialize_read_exit(s);
4517 			goto ours;
4518 		}
4519 
4520 		if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4521 			bool ingroup;
4522 
4523 			ia6_acquire(ia, &psref);
4524 			pserialize_read_exit(s);
4525 
4526 			/*
4527 			 * XXX Multicast
4528 			 * XXX why do we care about multlicast here while we don't care
4529 			 * about IPv4 multicast??
4530 			 * XXX scope
4531 			 */
4532 			ingroup = in6_multi_group(&sin6->sin6_addr, ia->ia_ifp);
4533 			if (ingroup) {
4534 				ia6_release(ia, &psref);
4535 				goto ours;
4536 			}
4537 
4538 			s = pserialize_read_enter();
4539 			ia6_release(ia, &psref);
4540 		}
4541 
4542 	}
4543 	pserialize_read_exit(s);
4544 
4545 	/* loopback, just for safety */
4546 	if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))
4547 		goto ours;
4548 
4549 	ours = 0;
4550 ours:
4551 	curlwp_bindx(bound);
4552 
4553 	return ours;
4554 }
4555 #endif /*INET6*/
4556 
4557 /*
4558  * compare two secasindex structure.
4559  * flag can specify to compare 2 saidxes.
4560  * compare two secasindex structure without both mode and reqid.
4561  * don't compare port.
4562  * IN:
4563  *      saidx0: source, it can be in SAD.
4564  *      saidx1: object.
4565  * OUT:
4566  *      1 : equal
4567  *      0 : not equal
4568  */
4569 static int
4570 key_saidx_match(
4571 	const struct secasindex *saidx0,
4572 	const struct secasindex *saidx1,
4573 	int flag)
4574 {
4575 	int chkport;
4576 	const struct sockaddr *sa0src, *sa0dst, *sa1src, *sa1dst;
4577 
4578 	KASSERT(saidx0 != NULL);
4579 	KASSERT(saidx1 != NULL);
4580 
4581 	/* sanity */
4582 	if (saidx0->proto != saidx1->proto)
4583 		return 0;
4584 
4585 	if (flag == CMP_EXACTLY) {
4586 		if (saidx0->mode != saidx1->mode)
4587 			return 0;
4588 		if (saidx0->reqid != saidx1->reqid)
4589 			return 0;
4590 		if (memcmp(&saidx0->src, &saidx1->src, saidx0->src.sa.sa_len) != 0 ||
4591 		    memcmp(&saidx0->dst, &saidx1->dst, saidx0->dst.sa.sa_len) != 0)
4592 			return 0;
4593 	} else {
4594 
4595 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
4596 		if (flag == CMP_MODE_REQID ||flag == CMP_REQID) {
4597 			/*
4598 			 * If reqid of SPD is non-zero, unique SA is required.
4599 			 * The result must be of same reqid in this case.
4600 			 */
4601 			if (saidx1->reqid != 0 && saidx0->reqid != saidx1->reqid)
4602 				return 0;
4603 		}
4604 
4605 		if (flag == CMP_MODE_REQID) {
4606 			if (saidx0->mode != IPSEC_MODE_ANY &&
4607 			    saidx0->mode != saidx1->mode)
4608 				return 0;
4609 		}
4610 
4611 
4612 		sa0src = &saidx0->src.sa;
4613 		sa0dst = &saidx0->dst.sa;
4614 		sa1src = &saidx1->src.sa;
4615 		sa1dst = &saidx1->dst.sa;
4616 		/*
4617 		 * If NAT-T is enabled, check ports for tunnel mode.
4618 		 * For ipsecif(4), check ports for transport mode, too.
4619 		 * Don't check ports if they are set to zero
4620 		 * in the SPD: This means we have a non-generated
4621 		 * SPD which can't know UDP ports.
4622 		 */
4623 		if (saidx1->mode == IPSEC_MODE_TUNNEL ||
4624 		    saidx1->mode == IPSEC_MODE_TRANSPORT)
4625 			chkport = PORT_LOOSE;
4626 		else
4627 			chkport = PORT_NONE;
4628 
4629 		if (!key_sockaddr_match(sa0src, sa1src, chkport)) {
4630 			return 0;
4631 		}
4632 		if (!key_sockaddr_match(sa0dst, sa1dst, chkport)) {
4633 			return 0;
4634 		}
4635 	}
4636 
4637 	return 1;
4638 }
4639 
4640 /*
4641  * compare two secindex structure exactly.
4642  * IN:
4643  *	spidx0: source, it is often in SPD.
4644  *	spidx1: object, it is often from PFKEY message.
4645  * OUT:
4646  *	1 : equal
4647  *	0 : not equal
4648  */
4649 static int
4650 key_spidx_match_exactly(
4651 	const struct secpolicyindex *spidx0,
4652 	const struct secpolicyindex *spidx1)
4653 {
4654 
4655 	KASSERT(spidx0 != NULL);
4656 	KASSERT(spidx1 != NULL);
4657 
4658 	/* sanity */
4659 	if (spidx0->prefs != spidx1->prefs ||
4660 	    spidx0->prefd != spidx1->prefd ||
4661 	    spidx0->ul_proto != spidx1->ul_proto)
4662 		return 0;
4663 
4664 	return key_sockaddr_match(&spidx0->src.sa, &spidx1->src.sa, PORT_STRICT) &&
4665 	       key_sockaddr_match(&spidx0->dst.sa, &spidx1->dst.sa, PORT_STRICT);
4666 }
4667 
4668 /*
4669  * compare two secindex structure with mask.
4670  * IN:
4671  *	spidx0: source, it is often in SPD.
4672  *	spidx1: object, it is often from IP header.
4673  * OUT:
4674  *	1 : equal
4675  *	0 : not equal
4676  */
4677 static int
4678 key_spidx_match_withmask(
4679 	const struct secpolicyindex *spidx0,
4680 	const struct secpolicyindex *spidx1)
4681 {
4682 
4683 	KASSERT(spidx0 != NULL);
4684 	KASSERT(spidx1 != NULL);
4685 
4686 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4687 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4688 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4689 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len) {
4690 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH, ".sa wrong\n");
4691 		return 0;
4692 	}
4693 
4694 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4695 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY &&
4696 	    spidx0->ul_proto != spidx1->ul_proto) {
4697 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "proto wrong\n");
4698 		return 0;
4699 	}
4700 
4701 	switch (spidx0->src.sa.sa_family) {
4702 	case AF_INET:
4703 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY &&
4704 		    spidx0->src.sin.sin_port != spidx1->src.sin.sin_port) {
4705 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v4 src port wrong\n");
4706 			return 0;
4707 		}
4708 		if (!key_bb_match_withmask(&spidx0->src.sin.sin_addr,
4709 					   &spidx1->src.sin.sin_addr, spidx0->prefs)) {
4710 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v4 src addr wrong\n");
4711 			return 0;
4712 		}
4713 		break;
4714 	case AF_INET6:
4715 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY &&
4716 		    spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port) {
4717 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v6 src port wrong\n");
4718 			return 0;
4719 		}
4720 		/*
4721 		 * scope_id check. if sin6_scope_id is 0, we regard it
4722 		 * as a wildcard scope, which matches any scope zone ID.
4723 		 */
4724 		if (spidx0->src.sin6.sin6_scope_id &&
4725 		    spidx1->src.sin6.sin6_scope_id &&
4726 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id) {
4727 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v6 src scope wrong\n");
4728 			return 0;
4729 		}
4730 		if (!key_bb_match_withmask(&spidx0->src.sin6.sin6_addr,
4731 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs)) {
4732 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v6 src addr wrong\n");
4733 			return 0;
4734 		}
4735 		break;
4736 	default:
4737 		/* XXX */
4738 		if (memcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0) {
4739 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "src memcmp wrong\n");
4740 			return 0;
4741 		}
4742 		break;
4743 	}
4744 
4745 	switch (spidx0->dst.sa.sa_family) {
4746 	case AF_INET:
4747 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY &&
4748 		    spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port) {
4749 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v4 dst port wrong\n");
4750 			return 0;
4751 		}
4752 		if (!key_bb_match_withmask(&spidx0->dst.sin.sin_addr,
4753 		    &spidx1->dst.sin.sin_addr, spidx0->prefd)) {
4754 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v4 dst addr wrong\n");
4755 			return 0;
4756 		}
4757 		break;
4758 	case AF_INET6:
4759 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY &&
4760 		    spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port) {
4761 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v6 dst port wrong\n");
4762 			return 0;
4763 		}
4764 		/*
4765 		 * scope_id check. if sin6_scope_id is 0, we regard it
4766 		 * as a wildcard scope, which matches any scope zone ID.
4767 		 */
4768 		if (spidx0->src.sin6.sin6_scope_id &&
4769 		    spidx1->src.sin6.sin6_scope_id &&
4770 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id) {
4771 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "DP v6 dst scope wrong\n");
4772 			return 0;
4773 		}
4774 		if (!key_bb_match_withmask(&spidx0->dst.sin6.sin6_addr,
4775 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd)) {
4776 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v6 dst addr wrong\n");
4777 			return 0;
4778 		}
4779 		break;
4780 	default:
4781 		/* XXX */
4782 		if (memcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0) {
4783 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "dst memcmp wrong\n");
4784 			return 0;
4785 		}
4786 		break;
4787 	}
4788 
4789 	/* XXX Do we check other field ?  e.g. flowinfo */
4790 
4791 	return 1;
4792 }
4793 
4794 /* returns 0 on match */
4795 static int
4796 key_portcomp(in_port_t port1, in_port_t port2, int howport)
4797 {
4798 	switch (howport) {
4799 	case PORT_NONE:
4800 		return 0;
4801 	case PORT_LOOSE:
4802 		if (port1 == 0 || port2 == 0)
4803 			return 0;
4804 		/*FALLTHROUGH*/
4805 	case PORT_STRICT:
4806 		if (port1 != port2) {
4807 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4808 			    "port fail %d != %d\n", ntohs(port1), ntohs(port2));
4809 			return 1;
4810 		}
4811 		return 0;
4812 	default:
4813 		KASSERT(0);
4814 		return 1;
4815 	}
4816 }
4817 
4818 /* returns 1 on match */
4819 static int
4820 key_sockaddr_match(
4821 	const struct sockaddr *sa1,
4822 	const struct sockaddr *sa2,
4823 	int howport)
4824 {
4825 	const struct sockaddr_in *sin1, *sin2;
4826 	const struct sockaddr_in6 *sin61, *sin62;
4827 	char s1[IPSEC_ADDRSTRLEN], s2[IPSEC_ADDRSTRLEN];
4828 
4829 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len) {
4830 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4831 		    "fam/len fail %d != %d || %d != %d\n",
4832 			sa1->sa_family, sa2->sa_family, sa1->sa_len,
4833 			sa2->sa_len);
4834 		return 0;
4835 	}
4836 
4837 	switch (sa1->sa_family) {
4838 	case AF_INET:
4839 		if (sa1->sa_len != sizeof(struct sockaddr_in)) {
4840 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4841 			    "len fail %d != %zu\n",
4842 			    sa1->sa_len, sizeof(struct sockaddr_in));
4843 			return 0;
4844 		}
4845 		sin1 = (const struct sockaddr_in *)sa1;
4846 		sin2 = (const struct sockaddr_in *)sa2;
4847 		if (sin1->sin_addr.s_addr != sin2->sin_addr.s_addr) {
4848 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4849 			    "addr fail %s != %s\n",
4850 			    (in_print(s1, sizeof(s1), &sin1->sin_addr), s1),
4851 			    (in_print(s2, sizeof(s2), &sin2->sin_addr), s2));
4852 			return 0;
4853 		}
4854 		if (key_portcomp(sin1->sin_port, sin2->sin_port, howport)) {
4855 			return 0;
4856 		}
4857 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4858 		    "addr success %s[%d] == %s[%d]\n",
4859 		    (in_print(s1, sizeof(s1), &sin1->sin_addr), s1),
4860 		    ntohs(sin1->sin_port),
4861 		    (in_print(s2, sizeof(s2), &sin2->sin_addr), s2),
4862 		    ntohs(sin2->sin_port));
4863 		break;
4864 	case AF_INET6:
4865 		sin61 = (const struct sockaddr_in6 *)sa1;
4866 		sin62 = (const struct sockaddr_in6 *)sa2;
4867 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4868 			return 0;	/*EINVAL*/
4869 
4870 		if (sin61->sin6_scope_id != sin62->sin6_scope_id) {
4871 			return 0;
4872 		}
4873 		if (!IN6_ARE_ADDR_EQUAL(&sin61->sin6_addr, &sin62->sin6_addr)) {
4874 			return 0;
4875 		}
4876 		if (key_portcomp(sin61->sin6_port, sin62->sin6_port, howport)) {
4877 			return 0;
4878 		}
4879 		break;
4880 	default:
4881 		if (memcmp(sa1, sa2, sa1->sa_len) != 0)
4882 			return 0;
4883 		break;
4884 	}
4885 
4886 	return 1;
4887 }
4888 
4889 /*
4890  * compare two buffers with mask.
4891  * IN:
4892  *	addr1: source
4893  *	addr2: object
4894  *	bits:  Number of bits to compare
4895  * OUT:
4896  *	1 : equal
4897  *	0 : not equal
4898  */
4899 static int
4900 key_bb_match_withmask(const void *a1, const void *a2, u_int bits)
4901 {
4902 	const unsigned char *p1 = a1;
4903 	const unsigned char *p2 = a2;
4904 
4905 	/* XXX: This could be considerably faster if we compare a word
4906 	 * at a time, but it is complicated on LSB Endian machines */
4907 
4908 	/* Handle null pointers */
4909 	if (p1 == NULL || p2 == NULL)
4910 		return (p1 == p2);
4911 
4912 	while (bits >= 8) {
4913 		if (*p1++ != *p2++)
4914 			return 0;
4915 		bits -= 8;
4916 	}
4917 
4918 	if (bits > 0) {
4919 		u_int8_t mask = ~((1<<(8-bits))-1);
4920 		if ((*p1 & mask) != (*p2 & mask))
4921 			return 0;
4922 	}
4923 	return 1;	/* Match! */
4924 }
4925 
4926 static void
4927 key_timehandler_spd(void)
4928 {
4929 	u_int dir;
4930 	struct secpolicy *sp;
4931 	volatile time_t now;
4932 
4933 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4934 	    retry:
4935 		mutex_enter(&key_spd.lock);
4936 		/*
4937 		 * To avoid for sp->created to overtake "now" because of
4938 		 * waiting mutex, set time_uptime here.
4939 		 */
4940 		now = time_uptime;
4941 		SPLIST_WRITER_FOREACH(sp, dir) {
4942 			KASSERTMSG(sp->state != IPSEC_SPSTATE_DEAD,
4943 			    "sp->state=%u", sp->state);
4944 
4945 			if (sp->lifetime == 0 && sp->validtime == 0)
4946 				continue;
4947 
4948 			if ((sp->lifetime && now - sp->created > sp->lifetime) ||
4949 			    (sp->validtime && now - sp->lastused > sp->validtime)) {
4950 				key_unlink_sp(sp);
4951 				mutex_exit(&key_spd.lock);
4952 				key_spdexpire(sp);
4953 				key_destroy_sp(sp);
4954 				goto retry;
4955 			}
4956 		}
4957 		mutex_exit(&key_spd.lock);
4958 	}
4959 
4960     retry_socksplist:
4961 	mutex_enter(&key_spd.lock);
4962 	SOCKSPLIST_WRITER_FOREACH(sp) {
4963 		if (sp->state != IPSEC_SPSTATE_DEAD)
4964 			continue;
4965 
4966 		key_unlink_sp(sp);
4967 		mutex_exit(&key_spd.lock);
4968 		key_destroy_sp(sp);
4969 		goto retry_socksplist;
4970 	}
4971 	mutex_exit(&key_spd.lock);
4972 }
4973 
4974 static void
4975 key_timehandler_sad(void)
4976 {
4977 	struct secashead *sah;
4978 	int s;
4979 	volatile time_t now;
4980 
4981 restart:
4982 	mutex_enter(&key_sad.lock);
4983 	SAHLIST_WRITER_FOREACH(sah) {
4984 		/* If sah has been dead and has no sav, then delete it */
4985 		if (sah->state == SADB_SASTATE_DEAD &&
4986 		    !key_sah_has_sav(sah)) {
4987 			key_unlink_sah(sah);
4988 			mutex_exit(&key_sad.lock);
4989 			key_destroy_sah(sah);
4990 			goto restart;
4991 		}
4992 	}
4993 	mutex_exit(&key_sad.lock);
4994 
4995 	s = pserialize_read_enter();
4996 	SAHLIST_READER_FOREACH(sah) {
4997 		struct secasvar *sav;
4998 
4999 		key_sah_ref(sah);
5000 		pserialize_read_exit(s);
5001 
5002 		/* if LARVAL entry doesn't become MATURE, delete it. */
5003 		mutex_enter(&key_sad.lock);
5004 	restart_sav_LARVAL:
5005 		/*
5006 		 * Same as key_timehandler_spd(), set time_uptime here.
5007 		 */
5008 		now = time_uptime;
5009 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_LARVAL) {
5010 			if (now - sav->created > key_larval_lifetime) {
5011 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
5012 				goto restart_sav_LARVAL;
5013 			}
5014 		}
5015 		mutex_exit(&key_sad.lock);
5016 
5017 		/*
5018 		 * check MATURE entry to start to send expire message
5019 		 * whether or not.
5020 		 */
5021 	restart_sav_MATURE:
5022 		mutex_enter(&key_sad.lock);
5023 		/*
5024 		 * ditto
5025 		 */
5026 		now = time_uptime;
5027 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_MATURE) {
5028 			/* we don't need to check. */
5029 			if (sav->lft_s == NULL)
5030 				continue;
5031 
5032 			/* sanity check */
5033 			KASSERT(sav->lft_c != NULL);
5034 
5035 			/* check SOFT lifetime */
5036 			if (sav->lft_s->sadb_lifetime_addtime != 0 &&
5037 			    now - sav->created > sav->lft_s->sadb_lifetime_addtime) {
5038 				/*
5039 				 * check SA to be used whether or not.
5040 				 * when SA hasn't been used, delete it.
5041 				 */
5042 				if (sav->lft_c->sadb_lifetime_usetime == 0) {
5043 					key_sa_chgstate(sav, SADB_SASTATE_DEAD);
5044 					mutex_exit(&key_sad.lock);
5045 				} else {
5046 					key_sa_chgstate(sav, SADB_SASTATE_DYING);
5047 					mutex_exit(&key_sad.lock);
5048 					/*
5049 					 * XXX If we keep to send expire
5050 					 * message in the status of
5051 					 * DYING. Do remove below code.
5052 					 */
5053 					key_expire(sav);
5054 				}
5055 				goto restart_sav_MATURE;
5056 			}
5057 			/* check SOFT lifetime by bytes */
5058 			/*
5059 			 * XXX I don't know the way to delete this SA
5060 			 * when new SA is installed.  Caution when it's
5061 			 * installed too big lifetime by time.
5062 			 */
5063 			else {
5064 				uint64_t lft_c_bytes = 0;
5065 				lifetime_counters_t sum = {0};
5066 
5067 				percpu_foreach_xcall(sav->lft_c_counters_percpu,
5068 				    XC_HIGHPRI_IPL(IPL_SOFTNET),
5069 				    key_sum_lifetime_counters, sum);
5070 				lft_c_bytes = sum[LIFETIME_COUNTER_BYTES];
5071 
5072 				if (sav->lft_s->sadb_lifetime_bytes == 0 ||
5073 				    sav->lft_s->sadb_lifetime_bytes >= lft_c_bytes)
5074 					continue;
5075 
5076 				key_sa_chgstate(sav, SADB_SASTATE_DYING);
5077 				mutex_exit(&key_sad.lock);
5078 				/*
5079 				 * XXX If we keep to send expire
5080 				 * message in the status of
5081 				 * DYING. Do remove below code.
5082 				 */
5083 				key_expire(sav);
5084 				goto restart_sav_MATURE;
5085 			}
5086 		}
5087 		mutex_exit(&key_sad.lock);
5088 
5089 		/* check DYING entry to change status to DEAD. */
5090 		mutex_enter(&key_sad.lock);
5091 	restart_sav_DYING:
5092 		/*
5093 		 * ditto
5094 		 */
5095 		now = time_uptime;
5096 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_DYING) {
5097 			/* we don't need to check. */
5098 			if (sav->lft_h == NULL)
5099 				continue;
5100 
5101 			/* sanity check */
5102 			KASSERT(sav->lft_c != NULL);
5103 
5104 			if (sav->lft_h->sadb_lifetime_addtime != 0 &&
5105 			    now - sav->created > sav->lft_h->sadb_lifetime_addtime) {
5106 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
5107 				goto restart_sav_DYING;
5108 			}
5109 #if 0	/* XXX Should we keep to send expire message until HARD lifetime ? */
5110 			else if (sav->lft_s != NULL
5111 			      && sav->lft_s->sadb_lifetime_addtime != 0
5112 			      && now - sav->created > sav->lft_s->sadb_lifetime_addtime) {
5113 				/*
5114 				 * XXX: should be checked to be
5115 				 * installed the valid SA.
5116 				 */
5117 
5118 				/*
5119 				 * If there is no SA then sending
5120 				 * expire message.
5121 				 */
5122 				key_expire(sav);
5123 			}
5124 #endif
5125 			/* check HARD lifetime by bytes */
5126 			else {
5127 				uint64_t lft_c_bytes = 0;
5128 				lifetime_counters_t sum = {0};
5129 
5130 				percpu_foreach_xcall(sav->lft_c_counters_percpu,
5131 				    XC_HIGHPRI_IPL(IPL_SOFTNET),
5132 				    key_sum_lifetime_counters, sum);
5133 				lft_c_bytes = sum[LIFETIME_COUNTER_BYTES];
5134 
5135 				if (sav->lft_h->sadb_lifetime_bytes == 0 ||
5136 				    sav->lft_h->sadb_lifetime_bytes >= lft_c_bytes)
5137 					continue;
5138 
5139 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
5140 				goto restart_sav_DYING;
5141 			}
5142 		}
5143 		mutex_exit(&key_sad.lock);
5144 
5145 		/* delete entry in DEAD */
5146 	restart_sav_DEAD:
5147 		mutex_enter(&key_sad.lock);
5148 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_DEAD) {
5149 			key_unlink_sav(sav);
5150 			mutex_exit(&key_sad.lock);
5151 			key_destroy_sav(sav);
5152 			goto restart_sav_DEAD;
5153 		}
5154 		mutex_exit(&key_sad.lock);
5155 
5156 		s = pserialize_read_enter();
5157 		key_sah_unref(sah);
5158 	}
5159 	pserialize_read_exit(s);
5160 }
5161 
5162 static void
5163 key_timehandler_acq(void)
5164 {
5165 #ifndef IPSEC_NONBLOCK_ACQUIRE
5166 	struct secacq *acq, *nextacq;
5167 	volatile time_t now;
5168 
5169     restart:
5170 	mutex_enter(&key_misc.lock);
5171 	/*
5172 	 * Same as key_timehandler_spd(), set time_uptime here.
5173 	 */
5174 	now = time_uptime;
5175 	LIST_FOREACH_SAFE(acq, &key_misc.acqlist, chain, nextacq) {
5176 		if (now - acq->created > key_blockacq_lifetime) {
5177 			LIST_REMOVE(acq, chain);
5178 			mutex_exit(&key_misc.lock);
5179 			kmem_free(acq, sizeof(*acq));
5180 			goto restart;
5181 		}
5182 	}
5183 	mutex_exit(&key_misc.lock);
5184 #endif
5185 }
5186 
5187 static void
5188 key_timehandler_spacq(void)
5189 {
5190 #ifdef notyet
5191 	struct secspacq *acq, *nextacq;
5192 	time_t now = time_uptime;
5193 
5194 	LIST_FOREACH_SAFE(acq, &key_misc.spacqlist, chain, nextacq) {
5195 		if (now - acq->created > key_blockacq_lifetime) {
5196 			KASSERT(__LIST_CHAINED(acq));
5197 			LIST_REMOVE(acq, chain);
5198 			kmem_free(acq, sizeof(*acq));
5199 		}
5200 	}
5201 #endif
5202 }
5203 
5204 static unsigned int key_timehandler_work_enqueued = 0;
5205 
5206 /*
5207  * time handler.
5208  * scanning SPD and SAD to check status for each entries,
5209  * and do to remove or to expire.
5210  */
5211 static void
5212 key_timehandler_work(struct work *wk, void *arg)
5213 {
5214 
5215 	/* We can allow enqueuing another work at this point */
5216 	atomic_swap_uint(&key_timehandler_work_enqueued, 0);
5217 
5218 	key_timehandler_spd();
5219 	key_timehandler_sad();
5220 	key_timehandler_acq();
5221 	key_timehandler_spacq();
5222 
5223 	key_acquire_sendup_pending_mbuf();
5224 
5225 	/* do exchange to tick time !! */
5226 	callout_reset(&key_timehandler_ch, hz, key_timehandler, NULL);
5227 
5228 	return;
5229 }
5230 
5231 static void
5232 key_timehandler(void *arg)
5233 {
5234 
5235 	/* Avoid enqueuing another work when one is already enqueued */
5236 	if (atomic_swap_uint(&key_timehandler_work_enqueued, 1) == 1)
5237 		return;
5238 
5239 	workqueue_enqueue(key_timehandler_wq, &key_timehandler_wk, NULL);
5240 }
5241 
5242 u_long
5243 key_random(void)
5244 {
5245 	u_long value;
5246 
5247 	key_randomfill(&value, sizeof(value));
5248 	return value;
5249 }
5250 
5251 void
5252 key_randomfill(void *p, size_t l)
5253 {
5254 
5255 	cprng_fast(p, l);
5256 }
5257 
5258 /*
5259  * map SADB_SATYPE_* to IPPROTO_*.
5260  * if satype == SADB_SATYPE then satype is mapped to ~0.
5261  * OUT:
5262  *	0: invalid satype.
5263  */
5264 static u_int16_t
5265 key_satype2proto(u_int8_t satype)
5266 {
5267 	switch (satype) {
5268 	case SADB_SATYPE_UNSPEC:
5269 		return IPSEC_PROTO_ANY;
5270 	case SADB_SATYPE_AH:
5271 		return IPPROTO_AH;
5272 	case SADB_SATYPE_ESP:
5273 		return IPPROTO_ESP;
5274 	case SADB_X_SATYPE_IPCOMP:
5275 		return IPPROTO_IPCOMP;
5276 	case SADB_X_SATYPE_TCPSIGNATURE:
5277 		return IPPROTO_TCP;
5278 	default:
5279 		return 0;
5280 	}
5281 	/* NOTREACHED */
5282 }
5283 
5284 /*
5285  * map IPPROTO_* to SADB_SATYPE_*
5286  * OUT:
5287  *	0: invalid protocol type.
5288  */
5289 static u_int8_t
5290 key_proto2satype(u_int16_t proto)
5291 {
5292 	switch (proto) {
5293 	case IPPROTO_AH:
5294 		return SADB_SATYPE_AH;
5295 	case IPPROTO_ESP:
5296 		return SADB_SATYPE_ESP;
5297 	case IPPROTO_IPCOMP:
5298 		return SADB_X_SATYPE_IPCOMP;
5299 	case IPPROTO_TCP:
5300 		return SADB_X_SATYPE_TCPSIGNATURE;
5301 	default:
5302 		return 0;
5303 	}
5304 	/* NOTREACHED */
5305 }
5306 
5307 static int
5308 key_setsecasidx(int proto, int mode, int reqid,
5309     const struct sockaddr *src, const struct sockaddr *dst,
5310     struct secasindex * saidx)
5311 {
5312 	const union sockaddr_union *src_u = (const union sockaddr_union *)src;
5313 	const union sockaddr_union *dst_u = (const union sockaddr_union *)dst;
5314 
5315 	/* sa len safety check */
5316 	if (key_checksalen(src_u) != 0)
5317 		return -1;
5318 	if (key_checksalen(dst_u) != 0)
5319 		return -1;
5320 
5321 	memset(saidx, 0, sizeof(*saidx));
5322 	saidx->proto = proto;
5323 	saidx->mode = mode;
5324 	saidx->reqid = reqid;
5325 	memcpy(&saidx->src, src_u, src_u->sa.sa_len);
5326 	memcpy(&saidx->dst, dst_u, dst_u->sa.sa_len);
5327 
5328 	key_porttosaddr(&((saidx)->src), 0);
5329 	key_porttosaddr(&((saidx)->dst), 0);
5330 	return 0;
5331 }
5332 
5333 static void
5334 key_init_spidx_bymsghdr(struct secpolicyindex *spidx,
5335     const struct sadb_msghdr *mhp)
5336 {
5337 	const struct sadb_address *src0, *dst0;
5338 	const struct sockaddr *src, *dst;
5339 	const struct sadb_x_policy *xpl0;
5340 
5341 	src0 = mhp->ext[SADB_EXT_ADDRESS_SRC];
5342 	dst0 = mhp->ext[SADB_EXT_ADDRESS_DST];
5343 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5344 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5345 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
5346 
5347 	memset(spidx, 0, sizeof(*spidx));
5348 	spidx->dir = xpl0->sadb_x_policy_dir;
5349 	spidx->prefs = src0->sadb_address_prefixlen;
5350 	spidx->prefd = dst0->sadb_address_prefixlen;
5351 	spidx->ul_proto = src0->sadb_address_proto;
5352 	/* XXX boundary check against sa_len */
5353 	memcpy(&spidx->src, src, src->sa_len);
5354 	memcpy(&spidx->dst, dst, dst->sa_len);
5355 }
5356 
5357 /* %%% PF_KEY */
5358 /*
5359  * SADB_GETSPI processing is to receive
5360  *	<base, (SA2), src address, dst address, (SPI range)>
5361  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
5362  * tree with the status of LARVAL, and send
5363  *	<base, SA(*), address(SD)>
5364  * to the IKMPd.
5365  *
5366  * IN:	mhp: pointer to the pointer to each header.
5367  * OUT:	NULL if fail.
5368  *	other if success, return pointer to the message to send.
5369  */
5370 static int
5371 key_api_getspi(struct socket *so, struct mbuf *m,
5372 	   const struct sadb_msghdr *mhp)
5373 {
5374 	const struct sockaddr *src, *dst;
5375 	struct secasindex saidx;
5376 	struct secashead *sah;
5377 	struct secasvar *newsav;
5378 	u_int8_t proto;
5379 	u_int32_t spi;
5380 	u_int8_t mode;
5381 	u_int16_t reqid;
5382 	int error;
5383 
5384 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
5385 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
5386 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5387 		return key_senderror(so, m, EINVAL);
5388 	}
5389 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
5390 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
5391 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5392 		return key_senderror(so, m, EINVAL);
5393 	}
5394 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
5395 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
5396 		mode = sa2->sadb_x_sa2_mode;
5397 		reqid = sa2->sadb_x_sa2_reqid;
5398 	} else {
5399 		mode = IPSEC_MODE_ANY;
5400 		reqid = 0;
5401 	}
5402 
5403 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5404 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5405 
5406 	/* map satype to proto */
5407 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
5408 	if (proto == 0) {
5409 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
5410 		return key_senderror(so, m, EINVAL);
5411 	}
5412 
5413 
5414 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
5415 	if (error != 0)
5416 		return key_senderror(so, m, EINVAL);
5417 
5418 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
5419 	if (error != 0)
5420 		return key_senderror(so, m, EINVAL);
5421 
5422 	/* SPI allocation */
5423 	spi = key_do_getnewspi(mhp->ext[SADB_EXT_SPIRANGE], &saidx);
5424 	if (spi == 0)
5425 		return key_senderror(so, m, EINVAL);
5426 
5427 	/* get a SA index */
5428 	sah = key_getsah_ref(&saidx, CMP_REQID);
5429 	if (sah == NULL) {
5430 		/* create a new SA index */
5431 		sah = key_newsah(&saidx);
5432 		if (sah == NULL) {
5433 			IPSECLOG(LOG_DEBUG, "No more memory.\n");
5434 			return key_senderror(so, m, ENOBUFS);
5435 		}
5436 	}
5437 
5438 	/* get a new SA */
5439 	/* XXX rewrite */
5440 	newsav = KEY_NEWSAV(m, mhp, &error, proto);
5441 	if (newsav == NULL) {
5442 		key_sah_unref(sah);
5443 		/* XXX don't free new SA index allocated in above. */
5444 		return key_senderror(so, m, error);
5445 	}
5446 
5447 	/* set spi */
5448 	newsav->spi = htonl(spi);
5449 
5450 	/* Add to sah#savlist */
5451 	key_init_sav(newsav);
5452 	newsav->sah = sah;
5453 	newsav->state = SADB_SASTATE_LARVAL;
5454 	mutex_enter(&key_sad.lock);
5455 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_LARVAL, newsav);
5456 	mutex_exit(&key_sad.lock);
5457 	key_validate_savlist(sah, SADB_SASTATE_LARVAL);
5458 
5459 	key_sah_unref(sah);
5460 
5461 #ifndef IPSEC_NONBLOCK_ACQUIRE
5462 	/* delete the entry in key_misc.acqlist */
5463 	if (mhp->msg->sadb_msg_seq != 0) {
5464 		struct secacq *acq;
5465 		mutex_enter(&key_misc.lock);
5466 		acq = key_getacqbyseq(mhp->msg->sadb_msg_seq);
5467 		if (acq != NULL) {
5468 			/* reset counter in order to deletion by timehandler. */
5469 			acq->created = time_uptime;
5470 			acq->count = 0;
5471 		}
5472 		mutex_exit(&key_misc.lock);
5473 	}
5474 #endif
5475 
5476     {
5477 	struct mbuf *n, *nn;
5478 	struct sadb_sa *m_sa;
5479 	int off, len;
5480 
5481 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
5482 	    PFKEY_ALIGN8(sizeof(struct sadb_sa)) <= MCLBYTES);
5483 
5484 	/* create new sadb_msg to reply. */
5485 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
5486 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
5487 
5488 	n = key_alloc_mbuf_simple(len, M_WAITOK);
5489 	n->m_len = len;
5490 	n->m_next = NULL;
5491 	off = 0;
5492 
5493 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
5494 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
5495 
5496 	m_sa = (struct sadb_sa *)(mtod(n, char *) + off);
5497 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
5498 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
5499 	m_sa->sadb_sa_spi = htonl(spi);
5500 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
5501 
5502 	KASSERTMSG(off == len, "length inconsistency");
5503 
5504 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
5505 	    SADB_EXT_ADDRESS_DST);
5506 
5507 	KASSERT(n->m_len >= sizeof(struct sadb_msg));
5508 
5509 	n->m_pkthdr.len = 0;
5510 	for (nn = n; nn; nn = nn->m_next)
5511 		n->m_pkthdr.len += nn->m_len;
5512 
5513 	key_fill_replymsg(n, newsav->seq);
5514 	m_freem(m);
5515 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
5516     }
5517 }
5518 
5519 /*
5520  * allocating new SPI
5521  * called by key_api_getspi().
5522  * OUT:
5523  *	0:	failure.
5524  *	others: success.
5525  */
5526 static u_int32_t
5527 key_do_getnewspi(const struct sadb_spirange *spirange,
5528 		 const struct secasindex *saidx)
5529 {
5530 	u_int32_t newspi;
5531 	u_int32_t spmin, spmax;
5532 	int count = key_spi_trycnt;
5533 
5534 	/* set spi range to allocate */
5535 	if (spirange != NULL) {
5536 		spmin = spirange->sadb_spirange_min;
5537 		spmax = spirange->sadb_spirange_max;
5538 	} else {
5539 		spmin = key_spi_minval;
5540 		spmax = key_spi_maxval;
5541 	}
5542 	/* IPCOMP needs 2-byte SPI */
5543 	if (saidx->proto == IPPROTO_IPCOMP) {
5544 		u_int32_t t;
5545 		if (spmin >= 0x10000)
5546 			spmin = 0xffff;
5547 		if (spmax >= 0x10000)
5548 			spmax = 0xffff;
5549 		if (spmin > spmax) {
5550 			t = spmin; spmin = spmax; spmax = t;
5551 		}
5552 	}
5553 
5554 	if (spmin == spmax) {
5555 		if (key_checkspidup(saidx, htonl(spmin))) {
5556 			IPSECLOG(LOG_DEBUG, "SPI %u exists already.\n", spmin);
5557 			return 0;
5558 		}
5559 
5560 		count--; /* taking one cost. */
5561 		newspi = spmin;
5562 
5563 	} else {
5564 
5565 		/* init SPI */
5566 		newspi = 0;
5567 
5568 		/* when requesting to allocate spi ranged */
5569 		while (count--) {
5570 			/* generate pseudo-random SPI value ranged. */
5571 			newspi = spmin + (key_random() % (spmax - spmin + 1));
5572 
5573 			if (!key_checkspidup(saidx, htonl(newspi)))
5574 				break;
5575 		}
5576 
5577 		if (count == 0 || newspi == 0) {
5578 			IPSECLOG(LOG_DEBUG, "to allocate spi is failed.\n");
5579 			return 0;
5580 		}
5581 	}
5582 
5583 	/* statistics */
5584 	keystat.getspi_count =
5585 	    (keystat.getspi_count + key_spi_trycnt - count) / 2;
5586 
5587 	return newspi;
5588 }
5589 
5590 static int
5591 key_handle_natt_info(struct secasvar *sav,
5592       		     const struct sadb_msghdr *mhp)
5593 {
5594 	const char *msg = "?" ;
5595 	struct sadb_x_nat_t_type *type;
5596 	struct sadb_x_nat_t_port *sport, *dport;
5597 	struct sadb_address *iaddr, *raddr;
5598 	struct sadb_x_nat_t_frag *frag;
5599 
5600 	if (mhp->ext[SADB_X_EXT_NAT_T_TYPE] == NULL ||
5601 	    mhp->ext[SADB_X_EXT_NAT_T_SPORT] == NULL ||
5602 	    mhp->ext[SADB_X_EXT_NAT_T_DPORT] == NULL)
5603 		return 0;
5604 
5605 	if (mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) {
5606 		msg = "TYPE";
5607 		goto bad;
5608 	}
5609 
5610 	if (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) {
5611 		msg = "SPORT";
5612 		goto bad;
5613 	}
5614 
5615 	if (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport)) {
5616 		msg = "DPORT";
5617 		goto bad;
5618 	}
5619 
5620 	if (mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL) {
5621 		IPSECLOG(LOG_DEBUG, "NAT-T OAi present\n");
5622 		if (mhp->extlen[SADB_X_EXT_NAT_T_OAI] < sizeof(*iaddr)) {
5623 			msg = "OAI";
5624 			goto bad;
5625 		}
5626 	}
5627 
5628 	if (mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL) {
5629 		IPSECLOG(LOG_DEBUG, "NAT-T OAr present\n");
5630 		if (mhp->extlen[SADB_X_EXT_NAT_T_OAR] < sizeof(*raddr)) {
5631 			msg = "OAR";
5632 			goto bad;
5633 		}
5634 	}
5635 
5636 	if (mhp->ext[SADB_X_EXT_NAT_T_FRAG] != NULL) {
5637 	    if (mhp->extlen[SADB_X_EXT_NAT_T_FRAG] < sizeof(*frag)) {
5638 		    msg = "FRAG";
5639 		    goto bad;
5640 	    }
5641 	}
5642 
5643 	type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5644 	sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5645 	dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5646 	iaddr = mhp->ext[SADB_X_EXT_NAT_T_OAI];
5647 	raddr = mhp->ext[SADB_X_EXT_NAT_T_OAR];
5648 	frag = mhp->ext[SADB_X_EXT_NAT_T_FRAG];
5649 
5650 	IPSECLOG(LOG_DEBUG, "type %d, sport = %d, dport = %d\n",
5651 	    type->sadb_x_nat_t_type_type,
5652 	    ntohs(sport->sadb_x_nat_t_port_port),
5653 	    ntohs(dport->sadb_x_nat_t_port_port));
5654 
5655 	sav->natt_type = type->sadb_x_nat_t_type_type;
5656 	key_porttosaddr(&sav->sah->saidx.src, sport->sadb_x_nat_t_port_port);
5657 	key_porttosaddr(&sav->sah->saidx.dst, dport->sadb_x_nat_t_port_port);
5658 	if (frag)
5659 		sav->esp_frag = frag->sadb_x_nat_t_frag_fraglen;
5660 	else
5661 		sav->esp_frag = IP_MAXPACKET;
5662 
5663 	return 0;
5664 bad:
5665 	IPSECLOG(LOG_DEBUG, "invalid message %s\n", msg);
5666 	__USE(msg);
5667 	return -1;
5668 }
5669 
5670 /* Just update the IPSEC_NAT_T ports if present */
5671 static int
5672 key_set_natt_ports(union sockaddr_union *src, union sockaddr_union *dst,
5673       		     const struct sadb_msghdr *mhp)
5674 {
5675 	if (mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL)
5676 		IPSECLOG(LOG_DEBUG, "NAT-T OAi present\n");
5677 	if (mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL)
5678 		IPSECLOG(LOG_DEBUG, "NAT-T OAr present\n");
5679 
5680 	if ((mhp->ext[SADB_X_EXT_NAT_T_TYPE] != NULL) &&
5681 	    (mhp->ext[SADB_X_EXT_NAT_T_SPORT] != NULL) &&
5682 	    (mhp->ext[SADB_X_EXT_NAT_T_DPORT] != NULL)) {
5683 		struct sadb_x_nat_t_type *type;
5684 		struct sadb_x_nat_t_port *sport;
5685 		struct sadb_x_nat_t_port *dport;
5686 
5687 		if ((mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) ||
5688 		    (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) ||
5689 		    (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport))) {
5690 			IPSECLOG(LOG_DEBUG, "invalid message\n");
5691 			return -1;
5692 		}
5693 
5694 		type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5695 		sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5696 		dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5697 
5698 		key_porttosaddr(src, sport->sadb_x_nat_t_port_port);
5699 		key_porttosaddr(dst, dport->sadb_x_nat_t_port_port);
5700 
5701 		IPSECLOG(LOG_DEBUG, "type %d, sport = %d, dport = %d\n",
5702 		    type->sadb_x_nat_t_type_type,
5703 		    ntohs(sport->sadb_x_nat_t_port_port),
5704 		    ntohs(dport->sadb_x_nat_t_port_port));
5705 	}
5706 
5707 	return 0;
5708 }
5709 
5710 
5711 /*
5712  * SADB_UPDATE processing
5713  * receive
5714  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5715  *       key(AE), (identity(SD),) (sensitivity)>
5716  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5717  * and send
5718  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5719  *       (identity(SD),) (sensitivity)>
5720  * to the ikmpd.
5721  *
5722  * m will always be freed.
5723  */
5724 static int
5725 key_api_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5726 {
5727 	struct sadb_sa *sa0;
5728 	const struct sockaddr *src, *dst;
5729 	struct secasindex saidx;
5730 	struct secashead *sah;
5731 	struct secasvar *sav, *newsav, *oldsav;
5732 	u_int16_t proto;
5733 	u_int8_t mode;
5734 	u_int16_t reqid;
5735 	int error;
5736 
5737 	/* map satype to proto */
5738 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
5739 	if (proto == 0) {
5740 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
5741 		return key_senderror(so, m, EINVAL);
5742 	}
5743 
5744 	if (mhp->ext[SADB_EXT_SA] == NULL ||
5745 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
5746 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
5747 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5748 	     mhp->ext[SADB_EXT_KEY_ENCRYPT] == NULL) ||
5749 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5750 	     mhp->ext[SADB_EXT_KEY_AUTH] == NULL) ||
5751 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL &&
5752 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] == NULL) ||
5753 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] == NULL &&
5754 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] != NULL)) {
5755 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5756 		return key_senderror(so, m, EINVAL);
5757 	}
5758 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
5759 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
5760 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
5761 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5762 		return key_senderror(so, m, EINVAL);
5763 	}
5764 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
5765 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
5766 		mode = sa2->sadb_x_sa2_mode;
5767 		reqid = sa2->sadb_x_sa2_reqid;
5768 	} else {
5769 		mode = IPSEC_MODE_ANY;
5770 		reqid = 0;
5771 	}
5772 	/* XXX boundary checking for other extensions */
5773 
5774 	sa0 = mhp->ext[SADB_EXT_SA];
5775 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5776 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5777 
5778 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
5779 	if (error != 0)
5780 		return key_senderror(so, m, EINVAL);
5781 
5782 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
5783 	if (error != 0)
5784 		return key_senderror(so, m, EINVAL);
5785 
5786 	/* get a SA header */
5787 	sah = key_getsah_ref(&saidx, CMP_REQID);
5788 	if (sah == NULL) {
5789 		IPSECLOG(LOG_DEBUG, "no SA index found.\n");
5790 		return key_senderror(so, m, ENOENT);
5791 	}
5792 
5793 	/* set spidx if there */
5794 	/* XXX rewrite */
5795 	error = key_setident(sah, m, mhp);
5796 	if (error)
5797 		goto error_sah;
5798 
5799 	/* find a SA with sequence number. */
5800 #ifdef IPSEC_DOSEQCHECK
5801 	if (mhp->msg->sadb_msg_seq != 0) {
5802 		sav = key_getsavbyseq(sah, mhp->msg->sadb_msg_seq);
5803 		if (sav == NULL) {
5804 			IPSECLOG(LOG_DEBUG,
5805 			    "no larval SA with sequence %u exists.\n",
5806 			    mhp->msg->sadb_msg_seq);
5807 			error = ENOENT;
5808 			goto error_sah;
5809 		}
5810 	}
5811 #else
5812 	sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
5813 	if (sav == NULL) {
5814 		IPSECLOG(LOG_DEBUG, "no such a SA found (spi:%u)\n",
5815 		    (u_int32_t)ntohl(sa0->sadb_sa_spi));
5816 		error = EINVAL;
5817 		goto error_sah;
5818 	}
5819 #endif
5820 
5821 	/* validity check */
5822 	if (sav->sah->saidx.proto != proto) {
5823 		IPSECLOG(LOG_DEBUG, "protocol mismatched (DB=%u param=%u)\n",
5824 		    sav->sah->saidx.proto, proto);
5825 		error = EINVAL;
5826 		goto error;
5827 	}
5828 #ifdef IPSEC_DOSEQCHECK
5829 	if (sav->spi != sa0->sadb_sa_spi) {
5830 		IPSECLOG(LOG_DEBUG, "SPI mismatched (DB:%u param:%u)\n",
5831 		    (u_int32_t)ntohl(sav->spi),
5832 		    (u_int32_t)ntohl(sa0->sadb_sa_spi));
5833 		error = EINVAL;
5834 		goto error;
5835 	}
5836 #endif
5837 	if (sav->pid != mhp->msg->sadb_msg_pid) {
5838 		IPSECLOG(LOG_DEBUG, "pid mismatched (DB:%u param:%u)\n",
5839 		    sav->pid, mhp->msg->sadb_msg_pid);
5840 		error = EINVAL;
5841 		goto error;
5842 	}
5843 
5844 	/*
5845 	 * Allocate a new SA instead of modifying the existing SA directly
5846 	 * to avoid race conditions.
5847 	 */
5848 	newsav = kmem_zalloc(sizeof(struct secasvar), KM_SLEEP);
5849 
5850 	/* copy sav values */
5851 	newsav->spi = sav->spi;
5852 	newsav->seq = sav->seq;
5853 	newsav->created = sav->created;
5854 	newsav->pid = sav->pid;
5855 	newsav->sah = sav->sah;
5856  	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
5857 	    "DP from %s:%u update SA:%p to SA:%p spi=%#x proto=%d\n",
5858 	    __func__, __LINE__, sav, newsav,
5859 	    ntohl(newsav->spi), proto);
5860 
5861 	error = key_setsaval(newsav, m, mhp);
5862 	if (error) {
5863 		kmem_free(newsav, sizeof(*newsav));
5864 		goto error;
5865 	}
5866 
5867 	error = key_handle_natt_info(newsav, mhp);
5868 	if (error != 0) {
5869 		key_delsav(newsav);
5870 		goto error;
5871 	}
5872 
5873 	error = key_init_xform(newsav);
5874 	if (error != 0) {
5875 		key_delsav(newsav);
5876 		goto error;
5877 	}
5878 
5879 	/* Add to sah#savlist */
5880 	key_init_sav(newsav);
5881 	newsav->state = SADB_SASTATE_MATURE;
5882 	mutex_enter(&key_sad.lock);
5883 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_MATURE, newsav);
5884 	SAVLUT_WRITER_INSERT_HEAD(newsav);
5885 	mutex_exit(&key_sad.lock);
5886 	key_validate_savlist(sah, SADB_SASTATE_MATURE);
5887 
5888 	/*
5889 	 * We need to lookup and remove the sav atomically, so get it again
5890 	 * here by a special API while we have a reference to it.
5891 	 */
5892 	oldsav = key_lookup_and_remove_sav(sah, sa0->sadb_sa_spi, sav);
5893 	KASSERT(oldsav == NULL || oldsav == sav);
5894 	/* We can release the reference because of oldsav */
5895 	KEY_SA_UNREF(&sav);
5896 	if (oldsav == NULL) {
5897 		/* Someone has already removed the sav.  Nothing to do. */
5898 	} else {
5899 		key_wait_sav(oldsav);
5900 		key_destroy_sav(oldsav);
5901 		oldsav = NULL;
5902 	}
5903 	sav = NULL;
5904 
5905 	key_sah_unref(sah);
5906 	sah = NULL;
5907 
5908     {
5909 	struct mbuf *n;
5910 
5911 	/* set msg buf from mhp */
5912 	n = key_getmsgbuf_x1(m, mhp);
5913 	if (n == NULL) {
5914 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
5915 		return key_senderror(so, m, ENOBUFS);
5916 	}
5917 
5918 	m_freem(m);
5919 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5920     }
5921 error:
5922 	KEY_SA_UNREF(&sav);
5923 error_sah:
5924 	key_sah_unref(sah);
5925 	return key_senderror(so, m, error);
5926 }
5927 
5928 /*
5929  * search SAD with sequence for a SA which state is SADB_SASTATE_LARVAL.
5930  * only called by key_api_update().
5931  * OUT:
5932  *	NULL	: not found
5933  *	others	: found, pointer to a SA.
5934  */
5935 #ifdef IPSEC_DOSEQCHECK
5936 static struct secasvar *
5937 key_getsavbyseq(struct secashead *sah, u_int32_t seq)
5938 {
5939 	struct secasvar *sav;
5940 	u_int state;
5941 	int s;
5942 
5943 	state = SADB_SASTATE_LARVAL;
5944 
5945 	/* search SAD with sequence number ? */
5946 	s = pserialize_read_enter();
5947 	SAVLIST_READER_FOREACH(sav, sah, state) {
5948 		KEY_CHKSASTATE(state, sav->state);
5949 
5950 		if (sav->seq == seq) {
5951 			SA_ADDREF(sav);
5952 			KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
5953 			    "DP cause refcnt++:%d SA:%p\n",
5954 			    key_sa_refcnt(sav), sav);
5955 			break;
5956 		}
5957 	}
5958 	pserialize_read_exit(s);
5959 
5960 	return sav;
5961 }
5962 #endif
5963 
5964 /*
5965  * SADB_ADD processing
5966  * add an entry to SA database, when received
5967  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5968  *       key(AE), (identity(SD),) (sensitivity)>
5969  * from the ikmpd,
5970  * and send
5971  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5972  *       (identity(SD),) (sensitivity)>
5973  * to the ikmpd.
5974  *
5975  * IGNORE identity and sensitivity messages.
5976  *
5977  * m will always be freed.
5978  */
5979 static int
5980 key_api_add(struct socket *so, struct mbuf *m,
5981 	const struct sadb_msghdr *mhp)
5982 {
5983 	struct sadb_sa *sa0;
5984 	const struct sockaddr *src, *dst;
5985 	struct secasindex saidx;
5986 	struct secashead *sah;
5987 	struct secasvar *newsav;
5988 	u_int16_t proto;
5989 	u_int8_t mode;
5990 	u_int16_t reqid;
5991 	int error;
5992 
5993 	/* map satype to proto */
5994 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
5995 	if (proto == 0) {
5996 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
5997 		return key_senderror(so, m, EINVAL);
5998 	}
5999 
6000 	if (mhp->ext[SADB_EXT_SA] == NULL ||
6001 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
6002 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
6003 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
6004 	     mhp->ext[SADB_EXT_KEY_ENCRYPT] == NULL) ||
6005 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
6006 	     mhp->ext[SADB_EXT_KEY_AUTH] == NULL) ||
6007 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL &&
6008 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] == NULL) ||
6009 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] == NULL &&
6010 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] != NULL)) {
6011 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6012 		return key_senderror(so, m, EINVAL);
6013 	}
6014 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
6015 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
6016 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
6017 		/* XXX need more */
6018 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6019 		return key_senderror(so, m, EINVAL);
6020 	}
6021 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
6022 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
6023 		mode = sa2->sadb_x_sa2_mode;
6024 		reqid = sa2->sadb_x_sa2_reqid;
6025 	} else {
6026 		mode = IPSEC_MODE_ANY;
6027 		reqid = 0;
6028 	}
6029 
6030 	sa0 = mhp->ext[SADB_EXT_SA];
6031 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6032 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6033 
6034 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
6035 	if (error != 0)
6036 		return key_senderror(so, m, EINVAL);
6037 
6038 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6039 	if (error != 0)
6040 		return key_senderror(so, m, EINVAL);
6041 
6042 	/* get a SA header */
6043 	sah = key_getsah_ref(&saidx, CMP_REQID);
6044 	if (sah == NULL) {
6045 		/* create a new SA header */
6046 		sah = key_newsah(&saidx);
6047 		if (sah == NULL) {
6048 			IPSECLOG(LOG_DEBUG, "No more memory.\n");
6049 			return key_senderror(so, m, ENOBUFS);
6050 		}
6051 	}
6052 
6053 	/* set spidx if there */
6054 	/* XXX rewrite */
6055 	error = key_setident(sah, m, mhp);
6056 	if (error)
6057 		goto error;
6058 
6059     {
6060 	struct secasvar *sav;
6061 
6062 	/* We can create new SA only if SPI is differenct. */
6063 	sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
6064 	if (sav != NULL) {
6065 		KEY_SA_UNREF(&sav);
6066 		IPSECLOG(LOG_DEBUG, "SA already exists.\n");
6067 		error = EEXIST;
6068 		goto error;
6069 	}
6070     }
6071 
6072 	/* create new SA entry. */
6073 	newsav = KEY_NEWSAV(m, mhp, &error, proto);
6074 	if (newsav == NULL)
6075 		goto error;
6076 	newsav->sah = sah;
6077 
6078 	error = key_handle_natt_info(newsav, mhp);
6079 	if (error != 0) {
6080 		key_delsav(newsav);
6081 		error = EINVAL;
6082 		goto error;
6083 	}
6084 
6085 	error = key_init_xform(newsav);
6086 	if (error != 0) {
6087 		key_delsav(newsav);
6088 		goto error;
6089 	}
6090 
6091 	/* Add to sah#savlist */
6092 	key_init_sav(newsav);
6093 	newsav->state = SADB_SASTATE_MATURE;
6094 	mutex_enter(&key_sad.lock);
6095 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_MATURE, newsav);
6096 	SAVLUT_WRITER_INSERT_HEAD(newsav);
6097 	mutex_exit(&key_sad.lock);
6098 	key_validate_savlist(sah, SADB_SASTATE_MATURE);
6099 
6100 	key_sah_unref(sah);
6101 	sah = NULL;
6102 
6103 	/*
6104 	 * don't call key_freesav() here, as we would like to keep the SA
6105 	 * in the database on success.
6106 	 */
6107 
6108     {
6109 	struct mbuf *n;
6110 
6111 	/* set msg buf from mhp */
6112 	n = key_getmsgbuf_x1(m, mhp);
6113 	if (n == NULL) {
6114 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
6115 		return key_senderror(so, m, ENOBUFS);
6116 	}
6117 
6118 	m_freem(m);
6119 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6120     }
6121 error:
6122 	key_sah_unref(sah);
6123 	return key_senderror(so, m, error);
6124 }
6125 
6126 /* m is retained */
6127 static int
6128 key_setident(struct secashead *sah, struct mbuf *m,
6129 	     const struct sadb_msghdr *mhp)
6130 {
6131 	const struct sadb_ident *idsrc, *iddst;
6132 	int idsrclen, iddstlen;
6133 
6134 	KASSERT(!cpu_softintr_p());
6135 	KASSERT(sah != NULL);
6136 	KASSERT(m != NULL);
6137 	KASSERT(mhp != NULL);
6138 	KASSERT(mhp->msg != NULL);
6139 
6140 	/*
6141 	 * Can be called with an existing sah from key_api_update().
6142 	 */
6143 	if (sah->idents != NULL) {
6144 		kmem_free(sah->idents, sah->idents_len);
6145 		sah->idents = NULL;
6146 		sah->idents_len = 0;
6147 	}
6148 	if (sah->identd != NULL) {
6149 		kmem_free(sah->identd, sah->identd_len);
6150 		sah->identd = NULL;
6151 		sah->identd_len = 0;
6152 	}
6153 
6154 	/* don't make buffer if not there */
6155 	if (mhp->ext[SADB_EXT_IDENTITY_SRC] == NULL &&
6156 	    mhp->ext[SADB_EXT_IDENTITY_DST] == NULL) {
6157 		sah->idents = NULL;
6158 		sah->identd = NULL;
6159 		return 0;
6160 	}
6161 
6162 	if (mhp->ext[SADB_EXT_IDENTITY_SRC] == NULL ||
6163 	    mhp->ext[SADB_EXT_IDENTITY_DST] == NULL) {
6164 		IPSECLOG(LOG_DEBUG, "invalid identity.\n");
6165 		return EINVAL;
6166 	}
6167 
6168 	idsrc = mhp->ext[SADB_EXT_IDENTITY_SRC];
6169 	iddst = mhp->ext[SADB_EXT_IDENTITY_DST];
6170 	idsrclen = mhp->extlen[SADB_EXT_IDENTITY_SRC];
6171 	iddstlen = mhp->extlen[SADB_EXT_IDENTITY_DST];
6172 
6173 	/* validity check */
6174 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
6175 		IPSECLOG(LOG_DEBUG, "ident type mismatched src %u, dst %u.\n",
6176 		    idsrc->sadb_ident_type, iddst->sadb_ident_type);
6177 		/*
6178 		 * Some VPN appliances(e.g. NetScreen) can send different
6179 		 * identifier types on IDii and IDir, so be able to allow
6180 		 * such message.
6181 		 */
6182 		if (!ipsec_allow_different_idtype) {
6183 			return EINVAL;
6184 		}
6185 	}
6186 
6187 	switch (idsrc->sadb_ident_type) {
6188 	case SADB_IDENTTYPE_PREFIX:
6189 	case SADB_IDENTTYPE_FQDN:
6190 	case SADB_IDENTTYPE_USERFQDN:
6191 	default:
6192 		/* XXX do nothing */
6193 		sah->idents = NULL;
6194 		sah->identd = NULL;
6195 	 	return 0;
6196 	}
6197 
6198 	/* make structure */
6199 	sah->idents = kmem_alloc(idsrclen, KM_SLEEP);
6200 	sah->idents_len = idsrclen;
6201 	sah->identd = kmem_alloc(iddstlen, KM_SLEEP);
6202 	sah->identd_len = iddstlen;
6203 	memcpy(sah->idents, idsrc, idsrclen);
6204 	memcpy(sah->identd, iddst, iddstlen);
6205 
6206 	return 0;
6207 }
6208 
6209 /*
6210  * m will not be freed on return. It never return NULL.
6211  * it is caller's responsibility to free the result.
6212  */
6213 static struct mbuf *
6214 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
6215 {
6216 	struct mbuf *n;
6217 
6218 	KASSERT(m != NULL);
6219 	KASSERT(mhp != NULL);
6220 	KASSERT(mhp->msg != NULL);
6221 
6222 	/* create new sadb_msg to reply. */
6223 	n = key_gather_mbuf(m, mhp, 1, 15, SADB_EXT_RESERVED,
6224 	    SADB_EXT_SA, SADB_X_EXT_SA2,
6225 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
6226 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
6227 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
6228 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
6229 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
6230 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NAT_T_FRAG);
6231 
6232 	KASSERT(n->m_len >= sizeof(struct sadb_msg));
6233 
6234 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
6235 	mtod(n, struct sadb_msg *)->sadb_msg_len =
6236 	    PFKEY_UNIT64(n->m_pkthdr.len);
6237 
6238 	return n;
6239 }
6240 
6241 static int key_delete_all (struct socket *, struct mbuf *,
6242 			   const struct sadb_msghdr *, u_int16_t);
6243 
6244 /*
6245  * SADB_DELETE processing
6246  * receive
6247  *   <base, SA(*), address(SD)>
6248  * from the ikmpd, and set SADB_SASTATE_DEAD,
6249  * and send,
6250  *   <base, SA(*), address(SD)>
6251  * to the ikmpd.
6252  *
6253  * m will always be freed.
6254  */
6255 static int
6256 key_api_delete(struct socket *so, struct mbuf *m,
6257 	   const struct sadb_msghdr *mhp)
6258 {
6259 	struct sadb_sa *sa0;
6260 	const struct sockaddr *src, *dst;
6261 	struct secasindex saidx;
6262 	struct secashead *sah;
6263 	struct secasvar *sav = NULL;
6264 	u_int16_t proto;
6265 	int error;
6266 
6267 	/* map satype to proto */
6268 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
6269 	if (proto == 0) {
6270 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
6271 		return key_senderror(so, m, EINVAL);
6272 	}
6273 
6274 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
6275 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
6276 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6277 		return key_senderror(so, m, EINVAL);
6278 	}
6279 
6280 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
6281 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
6282 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6283 		return key_senderror(so, m, EINVAL);
6284 	}
6285 
6286 	if (mhp->ext[SADB_EXT_SA] == NULL) {
6287 		/*
6288 		 * Caller wants us to delete all non-LARVAL SAs
6289 		 * that match the src/dst.  This is used during
6290 		 * IKE INITIAL-CONTACT.
6291 		 */
6292 		IPSECLOG(LOG_DEBUG, "doing delete all.\n");
6293 		return key_delete_all(so, m, mhp, proto);
6294 	} else if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa)) {
6295 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6296 		return key_senderror(so, m, EINVAL);
6297 	}
6298 
6299 	sa0 = mhp->ext[SADB_EXT_SA];
6300 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6301 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6302 
6303 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
6304 	if (error != 0)
6305 		return key_senderror(so, m, EINVAL);
6306 
6307 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6308 	if (error != 0)
6309 		return key_senderror(so, m, EINVAL);
6310 
6311 	/* get a SA header */
6312 	sah = key_getsah_ref(&saidx, CMP_HEAD);
6313 	if (sah != NULL) {
6314 		/* get a SA with SPI. */
6315 		sav = key_lookup_and_remove_sav(sah, sa0->sadb_sa_spi, NULL);
6316 		key_sah_unref(sah);
6317 	}
6318 
6319 	if (sav == NULL) {
6320 		IPSECLOG(LOG_DEBUG, "no SA found.\n");
6321 		return key_senderror(so, m, ENOENT);
6322 	}
6323 
6324 	key_wait_sav(sav);
6325 	key_destroy_sav(sav);
6326 	sav = NULL;
6327 
6328     {
6329 	struct mbuf *n;
6330 
6331 	/* create new sadb_msg to reply. */
6332 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
6333 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6334 
6335 	key_fill_replymsg(n, 0);
6336 	m_freem(m);
6337 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6338     }
6339 }
6340 
6341 /*
6342  * delete all SAs for src/dst.  Called from key_api_delete().
6343  */
6344 static int
6345 key_delete_all(struct socket *so, struct mbuf *m,
6346 	       const struct sadb_msghdr *mhp, u_int16_t proto)
6347 {
6348 	const struct sockaddr *src, *dst;
6349 	struct secasindex saidx;
6350 	struct secashead *sah;
6351 	struct secasvar *sav;
6352 	u_int state;
6353 	int error;
6354 
6355 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6356 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6357 
6358 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
6359 	if (error != 0)
6360 		return key_senderror(so, m, EINVAL);
6361 
6362 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6363 	if (error != 0)
6364 		return key_senderror(so, m, EINVAL);
6365 
6366 	sah = key_getsah_ref(&saidx, CMP_HEAD);
6367 	if (sah != NULL) {
6368 		/* Delete all non-LARVAL SAs. */
6369 		SASTATE_ALIVE_FOREACH(state) {
6370 			if (state == SADB_SASTATE_LARVAL)
6371 				continue;
6372 		restart:
6373 			mutex_enter(&key_sad.lock);
6374 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
6375 				sav->state = SADB_SASTATE_DEAD;
6376 				key_unlink_sav(sav);
6377 				mutex_exit(&key_sad.lock);
6378 				key_destroy_sav(sav);
6379 				goto restart;
6380 			}
6381 			mutex_exit(&key_sad.lock);
6382 		}
6383 		key_sah_unref(sah);
6384 	}
6385     {
6386 	struct mbuf *n;
6387 
6388 	/* create new sadb_msg to reply. */
6389 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
6390 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6391 
6392 	key_fill_replymsg(n, 0);
6393 	m_freem(m);
6394 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6395     }
6396 }
6397 
6398 /*
6399  * SADB_GET processing
6400  * receive
6401  *   <base, SA(*), address(SD)>
6402  * from the ikmpd, and get a SP and a SA to respond,
6403  * and send,
6404  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6405  *       (identity(SD),) (sensitivity)>
6406  * to the ikmpd.
6407  *
6408  * m will always be freed.
6409  */
6410 static int
6411 key_api_get(struct socket *so, struct mbuf *m,
6412 	const struct sadb_msghdr *mhp)
6413 {
6414 	struct sadb_sa *sa0;
6415 	const struct sockaddr *src, *dst;
6416 	struct secasindex saidx;
6417 	struct secasvar *sav = NULL;
6418 	u_int16_t proto;
6419 	int error;
6420 
6421 	/* map satype to proto */
6422 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6423 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
6424 		return key_senderror(so, m, EINVAL);
6425 	}
6426 
6427 	if (mhp->ext[SADB_EXT_SA] == NULL ||
6428 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
6429 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
6430 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6431 		return key_senderror(so, m, EINVAL);
6432 	}
6433 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
6434 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
6435 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
6436 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6437 		return key_senderror(so, m, EINVAL);
6438 	}
6439 
6440 	sa0 = mhp->ext[SADB_EXT_SA];
6441 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6442 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6443 
6444 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
6445 	if (error != 0)
6446 		return key_senderror(so, m, EINVAL);
6447 
6448 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6449 	if (error != 0)
6450 		return key_senderror(so, m, EINVAL);
6451 
6452 	/* get a SA header */
6453     {
6454 	struct secashead *sah;
6455 	int s = pserialize_read_enter();
6456 
6457 	sah = key_getsah(&saidx, CMP_HEAD);
6458 	if (sah != NULL) {
6459 		/* get a SA with SPI. */
6460 		sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
6461 	}
6462 	pserialize_read_exit(s);
6463     }
6464 	if (sav == NULL) {
6465 		IPSECLOG(LOG_DEBUG, "no SA found.\n");
6466 		return key_senderror(so, m, ENOENT);
6467 	}
6468 
6469     {
6470 	struct mbuf *n;
6471 	u_int8_t satype;
6472 
6473 	/* map proto to satype */
6474 	satype = key_proto2satype(sav->sah->saidx.proto);
6475 	if (satype == 0) {
6476 		KEY_SA_UNREF(&sav);
6477 		IPSECLOG(LOG_DEBUG, "there was invalid proto in SAD.\n");
6478 		return key_senderror(so, m, EINVAL);
6479 	}
6480 
6481 	/* create new sadb_msg to reply. */
6482 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
6483 	    mhp->msg->sadb_msg_pid);
6484 	KEY_SA_UNREF(&sav);
6485 	m_freem(m);
6486 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
6487     }
6488 }
6489 
6490 /* XXX make it sysctl-configurable? */
6491 static void
6492 key_getcomb_setlifetime(struct sadb_comb *comb)
6493 {
6494 
6495 	comb->sadb_comb_soft_allocations = 1;
6496 	comb->sadb_comb_hard_allocations = 1;
6497 	comb->sadb_comb_soft_bytes = 0;
6498 	comb->sadb_comb_hard_bytes = 0;
6499 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
6500 	comb->sadb_comb_soft_addtime = comb->sadb_comb_hard_addtime * 80 / 100;
6501 	comb->sadb_comb_hard_usetime = 28800;	/* 8 hours */
6502 	comb->sadb_comb_soft_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
6503 }
6504 
6505 /*
6506  * XXX reorder combinations by preference
6507  * XXX no idea if the user wants ESP authentication or not
6508  */
6509 static struct mbuf *
6510 key_getcomb_esp(int mflag)
6511 {
6512 	struct sadb_comb *comb;
6513 	const struct enc_xform *algo;
6514 	struct mbuf *result = NULL, *m, *n;
6515 	int encmin;
6516 	int i, off, o;
6517 	int totlen;
6518 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6519 
6520 	m = NULL;
6521 	for (i = 1; i <= SADB_EALG_MAX; i++) {
6522 		algo = esp_algorithm_lookup(i);
6523 		if (algo == NULL)
6524 			continue;
6525 
6526 		/* discard algorithms with key size smaller than system min */
6527 		if (_BITS(algo->maxkey) < ipsec_esp_keymin)
6528 			continue;
6529 		if (_BITS(algo->minkey) < ipsec_esp_keymin)
6530 			encmin = ipsec_esp_keymin;
6531 		else
6532 			encmin = _BITS(algo->minkey);
6533 
6534 		if (ipsec_esp_auth)
6535 			m = key_getcomb_ah(mflag);
6536 		else {
6537 			KASSERTMSG(l <= MLEN,
6538 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
6539 			MGET(m, mflag, MT_DATA);
6540 			if (m) {
6541 				m_align(m, l);
6542 				m->m_len = l;
6543 				m->m_next = NULL;
6544 				memset(mtod(m, void *), 0, m->m_len);
6545 			}
6546 		}
6547 		if (!m)
6548 			goto fail;
6549 
6550 		totlen = 0;
6551 		for (n = m; n; n = n->m_next)
6552 			totlen += n->m_len;
6553 		KASSERTMSG((totlen % l) == 0, "totlen=%u, l=%u", totlen, l);
6554 
6555 		for (off = 0; off < totlen; off += l) {
6556 			n = m_pulldown(m, off, l, &o);
6557 			if (!n) {
6558 				/* m is already freed */
6559 				goto fail;
6560 			}
6561 			comb = (struct sadb_comb *)(mtod(n, char *) + o);
6562 			memset(comb, 0, sizeof(*comb));
6563 			key_getcomb_setlifetime(comb);
6564 			comb->sadb_comb_encrypt = i;
6565 			comb->sadb_comb_encrypt_minbits = encmin;
6566 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
6567 		}
6568 
6569 		if (!result)
6570 			result = m;
6571 		else
6572 			m_cat(result, m);
6573 	}
6574 
6575 	return result;
6576 
6577  fail:
6578 	m_freem(result);
6579 	return NULL;
6580 }
6581 
6582 static void
6583 key_getsizes_ah(const struct auth_hash *ah, int alg,
6584 	        u_int16_t* ksmin, u_int16_t* ksmax)
6585 {
6586 	*ksmin = *ksmax = ah->keysize;
6587 	if (ah->keysize == 0) {
6588 		/*
6589 		 * Transform takes arbitrary key size but algorithm
6590 		 * key size is restricted.  Enforce this here.
6591 		 */
6592 		switch (alg) {
6593 		case SADB_X_AALG_MD5:	*ksmin = *ksmax = 16; break;
6594 		case SADB_X_AALG_SHA:	*ksmin = *ksmax = 20; break;
6595 		case SADB_X_AALG_NULL:	*ksmin = 0; *ksmax = 256; break;
6596 		default:
6597 			IPSECLOG(LOG_DEBUG, "unknown AH algorithm %u\n", alg);
6598 			break;
6599 		}
6600 	}
6601 }
6602 
6603 /*
6604  * XXX reorder combinations by preference
6605  */
6606 static struct mbuf *
6607 key_getcomb_ah(int mflag)
6608 {
6609 	struct sadb_comb *comb;
6610 	const struct auth_hash *algo;
6611 	struct mbuf *m;
6612 	u_int16_t minkeysize, maxkeysize;
6613 	int i;
6614 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6615 
6616 	m = NULL;
6617 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6618 #if 1
6619 		/* we prefer HMAC algorithms, not old algorithms */
6620 		if (i != SADB_AALG_SHA1HMAC &&
6621 		    i != SADB_AALG_MD5HMAC &&
6622 		    i != SADB_X_AALG_SHA2_256 &&
6623 		    i != SADB_X_AALG_SHA2_384 &&
6624 		    i != SADB_X_AALG_SHA2_512)
6625 			continue;
6626 #endif
6627 		algo = ah_algorithm_lookup(i);
6628 		if (!algo)
6629 			continue;
6630 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6631 		/* discard algorithms with key size smaller than system min */
6632 		if (_BITS(minkeysize) < ipsec_ah_keymin)
6633 			continue;
6634 
6635 		if (!m) {
6636 			KASSERTMSG(l <= MLEN,
6637 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
6638 			MGET(m, mflag, MT_DATA);
6639 			if (m) {
6640 				m_align(m, l);
6641 				m->m_len = l;
6642 				m->m_next = NULL;
6643 			}
6644 		} else
6645 			M_PREPEND(m, l, mflag);
6646 		if (!m)
6647 			return NULL;
6648 
6649 		if (m->m_len < sizeof(struct sadb_comb)) {
6650 			m = m_pullup(m, sizeof(struct sadb_comb));
6651 			if (m == NULL)
6652 				return NULL;
6653 		}
6654 
6655 		comb = mtod(m, struct sadb_comb *);
6656 		memset(comb, 0, sizeof(*comb));
6657 		key_getcomb_setlifetime(comb);
6658 		comb->sadb_comb_auth = i;
6659 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6660 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6661 	}
6662 
6663 	return m;
6664 }
6665 
6666 /*
6667  * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6668  * XXX reorder combinations by preference
6669  */
6670 static struct mbuf *
6671 key_getcomb_ipcomp(int mflag)
6672 {
6673 	struct sadb_comb *comb;
6674 	const struct comp_algo *algo;
6675 	struct mbuf *m;
6676 	int i;
6677 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6678 
6679 	m = NULL;
6680 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6681 		algo = ipcomp_algorithm_lookup(i);
6682 		if (!algo)
6683 			continue;
6684 
6685 		if (!m) {
6686 			KASSERTMSG(l <= MLEN,
6687 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
6688 			MGET(m, mflag, MT_DATA);
6689 			if (m) {
6690 				m_align(m, l);
6691 				m->m_len = l;
6692 				m->m_next = NULL;
6693 			}
6694 		} else
6695 			M_PREPEND(m, l, mflag);
6696 		if (!m)
6697 			return NULL;
6698 
6699 		if (m->m_len < sizeof(struct sadb_comb)) {
6700 			m = m_pullup(m, sizeof(struct sadb_comb));
6701 			if (m == NULL)
6702 				return NULL;
6703 		}
6704 
6705 		comb = mtod(m, struct sadb_comb *);
6706 		memset(comb, 0, sizeof(*comb));
6707 		key_getcomb_setlifetime(comb);
6708 		comb->sadb_comb_encrypt = i;
6709 		/* what should we set into sadb_comb_*_{min,max}bits? */
6710 	}
6711 
6712 	return m;
6713 }
6714 
6715 /*
6716  * XXX no way to pass mode (transport/tunnel) to userland
6717  * XXX replay checking?
6718  * XXX sysctl interface to ipsec_{ah,esp}_keymin
6719  */
6720 static struct mbuf *
6721 key_getprop(const struct secasindex *saidx, int mflag)
6722 {
6723 	struct sadb_prop *prop;
6724 	struct mbuf *m, *n;
6725 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6726 	int totlen;
6727 
6728 	switch (saidx->proto)  {
6729 	case IPPROTO_ESP:
6730 		m = key_getcomb_esp(mflag);
6731 		break;
6732 	case IPPROTO_AH:
6733 		m = key_getcomb_ah(mflag);
6734 		break;
6735 	case IPPROTO_IPCOMP:
6736 		m = key_getcomb_ipcomp(mflag);
6737 		break;
6738 	default:
6739 		return NULL;
6740 	}
6741 
6742 	if (!m)
6743 		return NULL;
6744 	M_PREPEND(m, l, mflag);
6745 	if (!m)
6746 		return NULL;
6747 
6748 	totlen = 0;
6749 	for (n = m; n; n = n->m_next)
6750 		totlen += n->m_len;
6751 
6752 	prop = mtod(m, struct sadb_prop *);
6753 	memset(prop, 0, sizeof(*prop));
6754 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6755 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6756 	prop->sadb_prop_replay = 32;	/* XXX */
6757 
6758 	return m;
6759 }
6760 
6761 /*
6762  * SADB_ACQUIRE processing called by key_checkrequest() and key_api_acquire().
6763  * send
6764  *   <base, SA, address(SD), (address(P)), x_policy,
6765  *       (identity(SD),) (sensitivity,) proposal>
6766  * to KMD, and expect to receive
6767  *   <base> with SADB_ACQUIRE if error occurred,
6768  * or
6769  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6770  * from KMD by PF_KEY.
6771  *
6772  * XXX x_policy is outside of RFC2367 (KAME extension).
6773  * XXX sensitivity is not supported.
6774  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6775  * see comment for key_getcomb_ipcomp().
6776  *
6777  * OUT:
6778  *    0     : succeed
6779  *    others: error number
6780  */
6781 static int
6782 key_acquire(const struct secasindex *saidx, const struct secpolicy *sp, int mflag)
6783 {
6784 	struct mbuf *result = NULL, *m;
6785 #ifndef IPSEC_NONBLOCK_ACQUIRE
6786 	struct secacq *newacq;
6787 #endif
6788 	u_int8_t satype;
6789 	int error = -1;
6790 	u_int32_t seq;
6791 
6792 	/* sanity check */
6793 	KASSERT(saidx != NULL);
6794 	satype = key_proto2satype(saidx->proto);
6795 	KASSERTMSG(satype != 0, "null satype, protocol %u", saidx->proto);
6796 
6797 #ifndef IPSEC_NONBLOCK_ACQUIRE
6798 	/*
6799 	 * We never do anything about acquiring SA.  There is another
6800 	 * solution that kernel blocks to send SADB_ACQUIRE message until
6801 	 * getting something message from IKEd.  In later case, to be
6802 	 * managed with ACQUIRING list.
6803 	 */
6804 	/* Get an entry to check whether sending message or not. */
6805 	mutex_enter(&key_misc.lock);
6806 	newacq = key_getacq(saidx);
6807 	if (newacq != NULL) {
6808 		if (key_blockacq_count < newacq->count) {
6809 			/* reset counter and do send message. */
6810 			newacq->count = 0;
6811 		} else {
6812 			/* increment counter and do nothing. */
6813 			newacq->count++;
6814 			mutex_exit(&key_misc.lock);
6815 			return 0;
6816 		}
6817 	} else {
6818 		/* make new entry for blocking to send SADB_ACQUIRE. */
6819 		newacq = key_newacq(saidx);
6820 		if (newacq == NULL) {
6821 			mutex_exit(&key_misc.lock);
6822 			return ENOBUFS;
6823 		}
6824 
6825 		/* add to key_misc.acqlist */
6826 		LIST_INSERT_HEAD(&key_misc.acqlist, newacq, chain);
6827 	}
6828 
6829 	seq = newacq->seq;
6830 	mutex_exit(&key_misc.lock);
6831 #else
6832 	seq = (acq_seq = (acq_seq == ~0 ? 1 : ++acq_seq));
6833 #endif
6834 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0, mflag);
6835 	if (!m) {
6836 		error = ENOBUFS;
6837 		goto fail;
6838 	}
6839 	result = m;
6840 
6841 	/* set sadb_address for saidx's. */
6842 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &saidx->src.sa, FULLMASK,
6843 	    IPSEC_ULPROTO_ANY, mflag);
6844 	if (!m) {
6845 		error = ENOBUFS;
6846 		goto fail;
6847 	}
6848 	m_cat(result, m);
6849 
6850 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &saidx->dst.sa, FULLMASK,
6851 	    IPSEC_ULPROTO_ANY, mflag);
6852 	if (!m) {
6853 		error = ENOBUFS;
6854 		goto fail;
6855 	}
6856 	m_cat(result, m);
6857 
6858 	/* XXX proxy address (optional) */
6859 
6860 	/* set sadb_x_policy */
6861 	if (sp) {
6862 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6863 		    mflag);
6864 		if (!m) {
6865 			error = ENOBUFS;
6866 			goto fail;
6867 		}
6868 		m_cat(result, m);
6869 	}
6870 
6871 	/* XXX identity (optional) */
6872 #if 0
6873 	if (idexttype && fqdn) {
6874 		/* create identity extension (FQDN) */
6875 		struct sadb_ident *id;
6876 		int fqdnlen;
6877 
6878 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
6879 		id = (struct sadb_ident *)p;
6880 		memset(id, 0, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6881 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6882 		id->sadb_ident_exttype = idexttype;
6883 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
6884 		memcpy(id + 1, fqdn, fqdnlen);
6885 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
6886 	}
6887 
6888 	if (idexttype) {
6889 		/* create identity extension (USERFQDN) */
6890 		struct sadb_ident *id;
6891 		int userfqdnlen;
6892 
6893 		if (userfqdn) {
6894 			/* +1 for terminating-NUL */
6895 			userfqdnlen = strlen(userfqdn) + 1;
6896 		} else
6897 			userfqdnlen = 0;
6898 		id = (struct sadb_ident *)p;
6899 		memset(id, 0, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6900 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6901 		id->sadb_ident_exttype = idexttype;
6902 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
6903 		/* XXX is it correct? */
6904 		if (curlwp)
6905 			id->sadb_ident_id = kauth_cred_getuid(curlwp->l_cred);
6906 		if (userfqdn && userfqdnlen)
6907 			memcpy(id + 1, userfqdn, userfqdnlen);
6908 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
6909 	}
6910 #endif
6911 
6912 	/* XXX sensitivity (optional) */
6913 
6914 	/* create proposal/combination extension */
6915 	m = key_getprop(saidx, mflag);
6916 #if 0
6917 	/*
6918 	 * spec conformant: always attach proposal/combination extension,
6919 	 * the problem is that we have no way to attach it for ipcomp,
6920 	 * due to the way sadb_comb is declared in RFC2367.
6921 	 */
6922 	if (!m) {
6923 		error = ENOBUFS;
6924 		goto fail;
6925 	}
6926 	m_cat(result, m);
6927 #else
6928 	/*
6929 	 * outside of spec; make proposal/combination extension optional.
6930 	 */
6931 	if (m)
6932 		m_cat(result, m);
6933 #endif
6934 
6935 	KASSERT(result->m_flags & M_PKTHDR);
6936 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
6937 
6938 	result->m_pkthdr.len = 0;
6939 	for (m = result; m; m = m->m_next)
6940 		result->m_pkthdr.len += m->m_len;
6941 
6942 	mtod(result, struct sadb_msg *)->sadb_msg_len =
6943 	    PFKEY_UNIT64(result->m_pkthdr.len);
6944 
6945 	/*
6946 	 * Called from key_api_acquire that must come from userland, so
6947 	 * we can call key_sendup_mbuf immediately.
6948 	 */
6949 	if (mflag == M_WAITOK)
6950 		return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
6951 	/*
6952 	 * XXX we cannot call key_sendup_mbuf directly here because
6953 	 * it can cause a deadlock:
6954 	 * - We have a reference to an SP (and an SA) here
6955 	 * - key_sendup_mbuf will try to take key_so_mtx
6956 	 * - Some other thread may try to localcount_drain to the SP with
6957 	 *   holding key_so_mtx in say key_api_spdflush
6958 	 * - In this case localcount_drain never return because key_sendup_mbuf
6959 	 *   that has stuck on key_so_mtx never release a reference to the SP
6960 	 *
6961 	 * So defer key_sendup_mbuf to the timer.
6962 	 */
6963 	return key_acquire_sendup_mbuf_later(result);
6964 
6965  fail:
6966 	m_freem(result);
6967 	return error;
6968 }
6969 
6970 static struct mbuf *key_acquire_mbuf_head = NULL;
6971 static unsigned key_acquire_mbuf_count = 0;
6972 #define KEY_ACQUIRE_MBUF_MAX	10
6973 
6974 static void
6975 key_acquire_sendup_pending_mbuf(void)
6976 {
6977 	struct mbuf *m, *prev;
6978 	int error;
6979 
6980 again:
6981 	prev = NULL;
6982 	mutex_enter(&key_misc.lock);
6983 	m = key_acquire_mbuf_head;
6984 	/* Get an earliest mbuf (one at the tail of the list) */
6985 	while (m != NULL) {
6986 		if (m->m_nextpkt == NULL) {
6987 			if (prev != NULL)
6988 				prev->m_nextpkt = NULL;
6989 			if (m == key_acquire_mbuf_head)
6990 				key_acquire_mbuf_head = NULL;
6991 			key_acquire_mbuf_count--;
6992 			break;
6993 		}
6994 		prev = m;
6995 		m = m->m_nextpkt;
6996 	}
6997 	mutex_exit(&key_misc.lock);
6998 
6999 	if (m == NULL)
7000 		return;
7001 
7002 	m->m_nextpkt = NULL;
7003 	error = key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
7004 	if (error != 0)
7005 		IPSECLOG(LOG_WARNING, "key_sendup_mbuf failed (error=%d)\n",
7006 		    error);
7007 
7008 	if (prev != NULL)
7009 		goto again;
7010 }
7011 
7012 static int
7013 key_acquire_sendup_mbuf_later(struct mbuf *m)
7014 {
7015 
7016 	mutex_enter(&key_misc.lock);
7017 	/* Avoid queuing too much mbufs */
7018 	if (key_acquire_mbuf_count >= KEY_ACQUIRE_MBUF_MAX) {
7019 		mutex_exit(&key_misc.lock);
7020 		m_freem(m);
7021 		return ENOBUFS; /* XXX */
7022 	}
7023 	/* Enqueue mbuf at the head of the list */
7024 	m->m_nextpkt = key_acquire_mbuf_head;
7025 	key_acquire_mbuf_head = m;
7026 	key_acquire_mbuf_count++;
7027 	mutex_exit(&key_misc.lock);
7028 
7029 	/* Kick the timer */
7030 	key_timehandler(NULL);
7031 
7032 	return 0;
7033 }
7034 
7035 #ifndef IPSEC_NONBLOCK_ACQUIRE
7036 static struct secacq *
7037 key_newacq(const struct secasindex *saidx)
7038 {
7039 	struct secacq *newacq;
7040 
7041 	/* get new entry */
7042 	newacq = kmem_intr_zalloc(sizeof(struct secacq), KM_NOSLEEP);
7043 	if (newacq == NULL) {
7044 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
7045 		return NULL;
7046 	}
7047 
7048 	/* copy secindex */
7049 	memcpy(&newacq->saidx, saidx, sizeof(newacq->saidx));
7050 	newacq->seq = (acq_seq == ~0 ? 1 : ++acq_seq);
7051 	newacq->created = time_uptime;
7052 	newacq->count = 0;
7053 
7054 	return newacq;
7055 }
7056 
7057 static struct secacq *
7058 key_getacq(const struct secasindex *saidx)
7059 {
7060 	struct secacq *acq;
7061 
7062 	KASSERT(mutex_owned(&key_misc.lock));
7063 
7064 	LIST_FOREACH(acq, &key_misc.acqlist, chain) {
7065 		if (key_saidx_match(saidx, &acq->saidx, CMP_EXACTLY))
7066 			return acq;
7067 	}
7068 
7069 	return NULL;
7070 }
7071 
7072 static struct secacq *
7073 key_getacqbyseq(u_int32_t seq)
7074 {
7075 	struct secacq *acq;
7076 
7077 	KASSERT(mutex_owned(&key_misc.lock));
7078 
7079 	LIST_FOREACH(acq, &key_misc.acqlist, chain) {
7080 		if (acq->seq == seq)
7081 			return acq;
7082 	}
7083 
7084 	return NULL;
7085 }
7086 #endif
7087 
7088 #ifdef notyet
7089 static struct secspacq *
7090 key_newspacq(const struct secpolicyindex *spidx)
7091 {
7092 	struct secspacq *acq;
7093 
7094 	/* get new entry */
7095 	acq = kmem_intr_zalloc(sizeof(struct secspacq), KM_NOSLEEP);
7096 	if (acq == NULL) {
7097 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
7098 		return NULL;
7099 	}
7100 
7101 	/* copy secindex */
7102 	memcpy(&acq->spidx, spidx, sizeof(acq->spidx));
7103 	acq->created = time_uptime;
7104 	acq->count = 0;
7105 
7106 	return acq;
7107 }
7108 
7109 static struct secspacq *
7110 key_getspacq(const struct secpolicyindex *spidx)
7111 {
7112 	struct secspacq *acq;
7113 
7114 	LIST_FOREACH(acq, &key_misc.spacqlist, chain) {
7115 		if (key_spidx_match_exactly(spidx, &acq->spidx))
7116 			return acq;
7117 	}
7118 
7119 	return NULL;
7120 }
7121 #endif /* notyet */
7122 
7123 /*
7124  * SADB_ACQUIRE processing,
7125  * in first situation, is receiving
7126  *   <base>
7127  * from the ikmpd, and clear sequence of its secasvar entry.
7128  *
7129  * In second situation, is receiving
7130  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
7131  * from a user land process, and return
7132  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
7133  * to the socket.
7134  *
7135  * m will always be freed.
7136  */
7137 static int
7138 key_api_acquire(struct socket *so, struct mbuf *m,
7139       	     const struct sadb_msghdr *mhp)
7140 {
7141 	const struct sockaddr *src, *dst;
7142 	struct secasindex saidx;
7143 	u_int16_t proto;
7144 	int error;
7145 
7146 	/*
7147 	 * Error message from KMd.
7148 	 * We assume that if error was occurred in IKEd, the length of PFKEY
7149 	 * message is equal to the size of sadb_msg structure.
7150 	 * We do not raise error even if error occurred in this function.
7151 	 */
7152 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
7153 #ifndef IPSEC_NONBLOCK_ACQUIRE
7154 		struct secacq *acq;
7155 
7156 		/* check sequence number */
7157 		if (mhp->msg->sadb_msg_seq == 0) {
7158 			IPSECLOG(LOG_DEBUG, "must specify sequence number.\n");
7159 			m_freem(m);
7160 			return 0;
7161 		}
7162 
7163 		mutex_enter(&key_misc.lock);
7164 		acq = key_getacqbyseq(mhp->msg->sadb_msg_seq);
7165 		if (acq == NULL) {
7166 			mutex_exit(&key_misc.lock);
7167 			/*
7168 			 * the specified larval SA is already gone, or we got
7169 			 * a bogus sequence number.  we can silently ignore it.
7170 			 */
7171 			m_freem(m);
7172 			return 0;
7173 		}
7174 
7175 		/* reset acq counter in order to deletion by timehandler. */
7176 		acq->created = time_uptime;
7177 		acq->count = 0;
7178 		mutex_exit(&key_misc.lock);
7179 #endif
7180 		m_freem(m);
7181 		return 0;
7182 	}
7183 
7184 	/*
7185 	 * This message is from user land.
7186 	 */
7187 
7188 	/* map satype to proto */
7189 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
7190 	if (proto == 0) {
7191 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
7192 		return key_senderror(so, m, EINVAL);
7193 	}
7194 
7195 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
7196 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
7197 	    mhp->ext[SADB_EXT_PROPOSAL] == NULL) {
7198 		/* error */
7199 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
7200 		return key_senderror(so, m, EINVAL);
7201 	}
7202 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
7203 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
7204 	    mhp->extlen[SADB_EXT_PROPOSAL] < sizeof(struct sadb_prop)) {
7205 		/* error */
7206 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
7207 		return key_senderror(so, m, EINVAL);
7208 	}
7209 
7210 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
7211 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
7212 
7213 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
7214 	if (error != 0)
7215 		return key_senderror(so, m, EINVAL);
7216 
7217 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
7218 	if (error != 0)
7219 		return key_senderror(so, m, EINVAL);
7220 
7221 	/* get a SA index */
7222     {
7223 	struct secashead *sah;
7224 	int s = pserialize_read_enter();
7225 
7226 	sah = key_getsah(&saidx, CMP_MODE_REQID);
7227 	if (sah != NULL) {
7228 		pserialize_read_exit(s);
7229 		IPSECLOG(LOG_DEBUG, "a SA exists already.\n");
7230 		return key_senderror(so, m, EEXIST);
7231 	}
7232 	pserialize_read_exit(s);
7233     }
7234 
7235 	error = key_acquire(&saidx, NULL, M_WAITOK);
7236 	if (error != 0) {
7237 		IPSECLOG(LOG_DEBUG, "error %d returned from key_acquire.\n",
7238 		    error);
7239 		return key_senderror(so, m, error);
7240 	}
7241 
7242 	return key_sendup_mbuf(so, m, KEY_SENDUP_REGISTERED);
7243 }
7244 
7245 /*
7246  * SADB_REGISTER processing.
7247  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
7248  * receive
7249  *   <base>
7250  * from the ikmpd, and register a socket to send PF_KEY messages,
7251  * and send
7252  *   <base, supported>
7253  * to KMD by PF_KEY.
7254  * If socket is detached, must free from regnode.
7255  *
7256  * m will always be freed.
7257  */
7258 static int
7259 key_api_register(struct socket *so, struct mbuf *m,
7260 	     const struct sadb_msghdr *mhp)
7261 {
7262 	struct secreg *reg, *newreg = 0;
7263 
7264 	/* check for invalid register message */
7265 	if (mhp->msg->sadb_msg_satype >= __arraycount(key_misc.reglist))
7266 		return key_senderror(so, m, EINVAL);
7267 
7268 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
7269 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
7270 		goto setmsg;
7271 
7272 	/* Allocate regnode in advance, out of mutex */
7273 	newreg = kmem_zalloc(sizeof(*newreg), KM_SLEEP);
7274 
7275 	/* check whether existing or not */
7276 	mutex_enter(&key_misc.lock);
7277 	LIST_FOREACH(reg, &key_misc.reglist[mhp->msg->sadb_msg_satype], chain) {
7278 		if (reg->so == so) {
7279 			IPSECLOG(LOG_DEBUG, "socket exists already.\n");
7280 			mutex_exit(&key_misc.lock);
7281 			kmem_free(newreg, sizeof(*newreg));
7282 			return key_senderror(so, m, EEXIST);
7283 		}
7284 	}
7285 
7286 	newreg->so = so;
7287 	((struct keycb *)sotorawcb(so))->kp_registered++;
7288 
7289 	/* add regnode to key_misc.reglist. */
7290 	LIST_INSERT_HEAD(&key_misc.reglist[mhp->msg->sadb_msg_satype], newreg, chain);
7291 	mutex_exit(&key_misc.lock);
7292 
7293   setmsg:
7294     {
7295 	struct mbuf *n;
7296 	struct sadb_supported *sup;
7297 	u_int len, alen, elen;
7298 	int off;
7299 	int i;
7300 	struct sadb_alg *alg;
7301 
7302 	/* create new sadb_msg to reply. */
7303 	alen = 0;
7304 	for (i = 1; i <= SADB_AALG_MAX; i++) {
7305 		if (ah_algorithm_lookup(i))
7306 			alen += sizeof(struct sadb_alg);
7307 	}
7308 	if (alen)
7309 		alen += sizeof(struct sadb_supported);
7310 	elen = 0;
7311 	for (i = 1; i <= SADB_EALG_MAX; i++) {
7312 		if (esp_algorithm_lookup(i))
7313 			elen += sizeof(struct sadb_alg);
7314 	}
7315 	if (elen)
7316 		elen += sizeof(struct sadb_supported);
7317 
7318 	len = sizeof(struct sadb_msg) + alen + elen;
7319 
7320 	if (len > MCLBYTES)
7321 		return key_senderror(so, m, ENOBUFS);
7322 
7323 	n = key_alloc_mbuf_simple(len, M_WAITOK);
7324 	n->m_pkthdr.len = n->m_len = len;
7325 	n->m_next = NULL;
7326 	off = 0;
7327 
7328 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
7329 	key_fill_replymsg(n, 0);
7330 
7331 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7332 
7333 	/* for authentication algorithm */
7334 	if (alen) {
7335 		sup = (struct sadb_supported *)(mtod(n, char *) + off);
7336 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7337 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7338 		sup->sadb_supported_reserved = 0;
7339 		off += PFKEY_ALIGN8(sizeof(*sup));
7340 
7341 		for (i = 1; i <= SADB_AALG_MAX; i++) {
7342 			const struct auth_hash *aalgo;
7343 			u_int16_t minkeysize, maxkeysize;
7344 
7345 			aalgo = ah_algorithm_lookup(i);
7346 			if (!aalgo)
7347 				continue;
7348 			alg = (struct sadb_alg *)(mtod(n, char *) + off);
7349 			alg->sadb_alg_id = i;
7350 			alg->sadb_alg_ivlen = 0;
7351 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7352 			alg->sadb_alg_minbits = _BITS(minkeysize);
7353 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7354 			alg->sadb_alg_reserved = 0;
7355 			off += PFKEY_ALIGN8(sizeof(*alg));
7356 		}
7357 	}
7358 
7359 	/* for encryption algorithm */
7360 	if (elen) {
7361 		sup = (struct sadb_supported *)(mtod(n, char *) + off);
7362 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7363 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7364 		sup->sadb_supported_reserved = 0;
7365 		off += PFKEY_ALIGN8(sizeof(*sup));
7366 
7367 		for (i = 1; i <= SADB_EALG_MAX; i++) {
7368 			const struct enc_xform *ealgo;
7369 
7370 			ealgo = esp_algorithm_lookup(i);
7371 			if (!ealgo)
7372 				continue;
7373 			alg = (struct sadb_alg *)(mtod(n, char *) + off);
7374 			alg->sadb_alg_id = i;
7375 			alg->sadb_alg_ivlen = ealgo->blocksize;
7376 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7377 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7378 			alg->sadb_alg_reserved = 0;
7379 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7380 		}
7381 	}
7382 
7383 	KASSERTMSG(off == len, "length inconsistency");
7384 
7385 	m_freem(m);
7386 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7387     }
7388 }
7389 
7390 /*
7391  * free secreg entry registered.
7392  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7393  */
7394 void
7395 key_freereg(struct socket *so)
7396 {
7397 	struct secreg *reg;
7398 	int i;
7399 
7400 	KASSERT(!cpu_softintr_p());
7401 	KASSERT(so != NULL);
7402 
7403 	/*
7404 	 * check whether existing or not.
7405 	 * check all type of SA, because there is a potential that
7406 	 * one socket is registered to multiple type of SA.
7407 	 */
7408 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7409 		mutex_enter(&key_misc.lock);
7410 		LIST_FOREACH(reg, &key_misc.reglist[i], chain) {
7411 			if (reg->so == so) {
7412 				LIST_REMOVE(reg, chain);
7413 				break;
7414 			}
7415 		}
7416 		mutex_exit(&key_misc.lock);
7417 		if (reg != NULL)
7418 			kmem_free(reg, sizeof(*reg));
7419 	}
7420 
7421 	return;
7422 }
7423 
7424 /*
7425  * SADB_EXPIRE processing
7426  * send
7427  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7428  * to KMD by PF_KEY.
7429  * NOTE: We send only soft lifetime extension.
7430  *
7431  * OUT:	0	: succeed
7432  *	others	: error number
7433  */
7434 static int
7435 key_expire(struct secasvar *sav)
7436 {
7437 	int s;
7438 	int satype;
7439 	struct mbuf *result = NULL, *m;
7440 	int len;
7441 	int error = -1;
7442 	struct sadb_lifetime *lt;
7443 	lifetime_counters_t sum = {0};
7444 
7445 	/* XXX: Why do we lock ? */
7446 	s = splsoftnet();	/*called from softclock()*/
7447 
7448 	KASSERT(sav != NULL);
7449 
7450 	satype = key_proto2satype(sav->sah->saidx.proto);
7451 	KASSERTMSG(satype != 0, "invalid proto is passed");
7452 
7453 	/* set msg header */
7454 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, key_sa_refcnt(sav),
7455 	    M_WAITOK);
7456 	result = m;
7457 
7458 	/* create SA extension */
7459 	m = key_setsadbsa(sav);
7460 	m_cat(result, m);
7461 
7462 	/* create SA extension */
7463 	m = key_setsadbxsa2(sav->sah->saidx.mode,
7464 	    sav->replay ? sav->replay->count : 0, sav->sah->saidx.reqid);
7465 	m_cat(result, m);
7466 
7467 	/* create lifetime extension (current and soft) */
7468 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7469 	m = key_alloc_mbuf(len, M_WAITOK);
7470 	KASSERT(m->m_next == NULL);
7471 
7472 	memset(mtod(m, void *), 0, len);
7473 	lt = mtod(m, struct sadb_lifetime *);
7474 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7475 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7476 	percpu_foreach_xcall(sav->lft_c_counters_percpu,
7477 	    XC_HIGHPRI_IPL(IPL_SOFTNET), key_sum_lifetime_counters, sum);
7478 	lt->sadb_lifetime_allocations = sum[LIFETIME_COUNTER_ALLOCATIONS];
7479 	lt->sadb_lifetime_bytes = sum[LIFETIME_COUNTER_BYTES];
7480 	lt->sadb_lifetime_addtime =
7481 	    time_mono_to_wall(sav->lft_c->sadb_lifetime_addtime);
7482 	lt->sadb_lifetime_usetime =
7483 	    time_mono_to_wall(sav->lft_c->sadb_lifetime_usetime);
7484 	lt = (struct sadb_lifetime *)(mtod(m, char *) + len / 2);
7485 	memcpy(lt, sav->lft_s, sizeof(*lt));
7486 	m_cat(result, m);
7487 
7488 	/* set sadb_address for source */
7489 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa,
7490 	    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
7491 	m_cat(result, m);
7492 
7493 	/* set sadb_address for destination */
7494 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.dst.sa,
7495 	    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
7496 	m_cat(result, m);
7497 
7498 	if ((result->m_flags & M_PKTHDR) == 0) {
7499 		error = EINVAL;
7500 		goto fail;
7501 	}
7502 
7503 	if (result->m_len < sizeof(struct sadb_msg)) {
7504 		result = m_pullup(result, sizeof(struct sadb_msg));
7505 		if (result == NULL) {
7506 			error = ENOBUFS;
7507 			goto fail;
7508 		}
7509 	}
7510 
7511 	result->m_pkthdr.len = 0;
7512 	for (m = result; m; m = m->m_next)
7513 		result->m_pkthdr.len += m->m_len;
7514 
7515 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7516 	    PFKEY_UNIT64(result->m_pkthdr.len);
7517 
7518 	splx(s);
7519 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7520 
7521  fail:
7522 	m_freem(result);
7523 	splx(s);
7524 	return error;
7525 }
7526 
7527 /*
7528  * SADB_FLUSH processing
7529  * receive
7530  *   <base>
7531  * from the ikmpd, and free all entries in secastree.
7532  * and send,
7533  *   <base>
7534  * to the ikmpd.
7535  * NOTE: to do is only marking SADB_SASTATE_DEAD.
7536  *
7537  * m will always be freed.
7538  */
7539 static int
7540 key_api_flush(struct socket *so, struct mbuf *m,
7541           const struct sadb_msghdr *mhp)
7542 {
7543 	struct sadb_msg *newmsg;
7544 	struct secashead *sah;
7545 	struct secasvar *sav;
7546 	u_int16_t proto;
7547 	u_int8_t state;
7548 	int s;
7549 
7550 	/* map satype to proto */
7551 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
7552 	if (proto == 0) {
7553 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
7554 		return key_senderror(so, m, EINVAL);
7555 	}
7556 
7557 	/* no SATYPE specified, i.e. flushing all SA. */
7558 	s = pserialize_read_enter();
7559 	SAHLIST_READER_FOREACH(sah) {
7560 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7561 		    proto != sah->saidx.proto)
7562 			continue;
7563 
7564 		key_sah_ref(sah);
7565 		pserialize_read_exit(s);
7566 
7567 		SASTATE_ALIVE_FOREACH(state) {
7568 		restart:
7569 			mutex_enter(&key_sad.lock);
7570 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
7571 				sav->state = SADB_SASTATE_DEAD;
7572 				key_unlink_sav(sav);
7573 				mutex_exit(&key_sad.lock);
7574 				key_destroy_sav(sav);
7575 				goto restart;
7576 			}
7577 			mutex_exit(&key_sad.lock);
7578 		}
7579 
7580 		s = pserialize_read_enter();
7581 		sah->state = SADB_SASTATE_DEAD;
7582 		key_sah_unref(sah);
7583 	}
7584 	pserialize_read_exit(s);
7585 
7586 	if (m->m_len < sizeof(struct sadb_msg) ||
7587 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7588 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
7589 		return key_senderror(so, m, ENOBUFS);
7590 	}
7591 
7592 	m_freem(m->m_next);
7593 	m->m_next = NULL;
7594 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7595 	newmsg = mtod(m, struct sadb_msg *);
7596 	newmsg->sadb_msg_errno = 0;
7597 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7598 
7599 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7600 }
7601 
7602 
7603 static struct mbuf *
7604 key_setdump_chain(u_int8_t req_satype, int *errorp, int *lenp, pid_t pid)
7605 {
7606 	struct secashead *sah;
7607 	struct secasvar *sav;
7608 	u_int16_t proto;
7609 	u_int8_t satype;
7610 	u_int8_t state;
7611 	int cnt;
7612 	struct mbuf *m, *n, *prev;
7613 
7614 	KASSERT(mutex_owned(&key_sad.lock));
7615 
7616 	*lenp = 0;
7617 
7618 	/* map satype to proto */
7619 	proto = key_satype2proto(req_satype);
7620 	if (proto == 0) {
7621 		*errorp = EINVAL;
7622 		return (NULL);
7623 	}
7624 
7625 	/* count sav entries to be sent to userland. */
7626 	cnt = 0;
7627 	SAHLIST_WRITER_FOREACH(sah) {
7628 		if (req_satype != SADB_SATYPE_UNSPEC &&
7629 		    proto != sah->saidx.proto)
7630 			continue;
7631 
7632 		SASTATE_ANY_FOREACH(state) {
7633 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
7634 				cnt++;
7635 			}
7636 		}
7637 	}
7638 
7639 	if (cnt == 0) {
7640 		*errorp = ENOENT;
7641 		return (NULL);
7642 	}
7643 
7644 	/* send this to the userland, one at a time. */
7645 	m = NULL;
7646 	prev = m;
7647 	SAHLIST_WRITER_FOREACH(sah) {
7648 		if (req_satype != SADB_SATYPE_UNSPEC &&
7649 		    proto != sah->saidx.proto)
7650 			continue;
7651 
7652 		/* map proto to satype */
7653 		satype = key_proto2satype(sah->saidx.proto);
7654 		if (satype == 0) {
7655 			m_freem(m);
7656 			*errorp = EINVAL;
7657 			return (NULL);
7658 		}
7659 
7660 		SASTATE_ANY_FOREACH(state) {
7661 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
7662 				n = key_setdumpsa(sav, SADB_DUMP, satype,
7663 				    --cnt, pid);
7664 				if (!m)
7665 					m = n;
7666 				else
7667 					prev->m_nextpkt = n;
7668 				prev = n;
7669 			}
7670 		}
7671 	}
7672 
7673 	if (!m) {
7674 		*errorp = EINVAL;
7675 		return (NULL);
7676 	}
7677 
7678 	if ((m->m_flags & M_PKTHDR) != 0) {
7679 		m->m_pkthdr.len = 0;
7680 		for (n = m; n; n = n->m_next)
7681 			m->m_pkthdr.len += n->m_len;
7682 	}
7683 
7684 	*errorp = 0;
7685 	return (m);
7686 }
7687 
7688 /*
7689  * SADB_DUMP processing
7690  * dump all entries including status of DEAD in SAD.
7691  * receive
7692  *   <base>
7693  * from the ikmpd, and dump all secasvar leaves
7694  * and send,
7695  *   <base> .....
7696  * to the ikmpd.
7697  *
7698  * m will always be freed.
7699  */
7700 static int
7701 key_api_dump(struct socket *so, struct mbuf *m0,
7702 	 const struct sadb_msghdr *mhp)
7703 {
7704 	u_int16_t proto;
7705 	u_int8_t satype;
7706 	struct mbuf *n;
7707 	int error, len, ok;
7708 
7709 	/* map satype to proto */
7710 	satype = mhp->msg->sadb_msg_satype;
7711 	proto = key_satype2proto(satype);
7712 	if (proto == 0) {
7713 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
7714 		return key_senderror(so, m0, EINVAL);
7715 	}
7716 
7717 	/*
7718 	 * If the requestor has insufficient socket-buffer space
7719 	 * for the entire chain, nobody gets any response to the DUMP.
7720 	 * XXX For now, only the requestor ever gets anything.
7721 	 * Moreover, if the requestor has any space at all, they receive
7722 	 * the entire chain, otherwise the request is refused with ENOBUFS.
7723 	 */
7724 	if (sbspace(&so->so_rcv) <= 0) {
7725 		return key_senderror(so, m0, ENOBUFS);
7726 	}
7727 
7728 	mutex_enter(&key_sad.lock);
7729 	n = key_setdump_chain(satype, &error, &len, mhp->msg->sadb_msg_pid);
7730 	mutex_exit(&key_sad.lock);
7731 
7732 	if (n == NULL) {
7733 		return key_senderror(so, m0, ENOENT);
7734 	}
7735 	{
7736 		net_stat_ref_t ps = PFKEY_STAT_GETREF();
7737 		_NET_STATINC_REF(ps, PFKEY_STAT_IN_TOTAL);
7738 		_NET_STATADD_REF(ps, PFKEY_STAT_IN_BYTES, len);
7739 		PFKEY_STAT_PUTREF();
7740 	}
7741 
7742 	/*
7743 	 * PF_KEY DUMP responses are no longer broadcast to all PF_KEY sockets.
7744 	 * The requestor receives either the entire chain, or an
7745 	 * error message with ENOBUFS.
7746 	 *
7747 	 * sbappendaddrchain() takes the chain of entries, one
7748 	 * packet-record per SPD entry, prepends the key_src sockaddr
7749 	 * to each packet-record, links the sockaddr mbufs into a new
7750 	 * list of records, then   appends the entire resulting
7751 	 * list to the requesting socket.
7752 	 */
7753 	ok = sbappendaddrchain(&so->so_rcv, (struct sockaddr *)&key_src, n,
7754 	    SB_PRIO_ONESHOT_OVERFLOW);
7755 
7756 	if (!ok) {
7757 		PFKEY_STATINC(PFKEY_STAT_IN_NOMEM);
7758 		m_freem(n);
7759 		return key_senderror(so, m0, ENOBUFS);
7760 	}
7761 
7762 	m_freem(m0);
7763 	return 0;
7764 }
7765 
7766 /*
7767  * SADB_X_PROMISC processing
7768  *
7769  * m will always be freed.
7770  */
7771 static int
7772 key_api_promisc(struct socket *so, struct mbuf *m,
7773 	    const struct sadb_msghdr *mhp)
7774 {
7775 	int olen;
7776 
7777 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7778 
7779 	if (olen < sizeof(struct sadb_msg)) {
7780 #if 1
7781 		return key_senderror(so, m, EINVAL);
7782 #else
7783 		m_freem(m);
7784 		return 0;
7785 #endif
7786 	} else if (olen == sizeof(struct sadb_msg)) {
7787 		/* enable/disable promisc mode */
7788 		struct keycb *kp = (struct keycb *)sotorawcb(so);
7789 		if (kp == NULL)
7790 			return key_senderror(so, m, EINVAL);
7791 		mhp->msg->sadb_msg_errno = 0;
7792 		switch (mhp->msg->sadb_msg_satype) {
7793 		case 0:
7794 		case 1:
7795 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
7796 			break;
7797 		default:
7798 			return key_senderror(so, m, EINVAL);
7799 		}
7800 
7801 		/* send the original message back to everyone */
7802 		mhp->msg->sadb_msg_errno = 0;
7803 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7804 	} else {
7805 		/* send packet as is */
7806 
7807 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
7808 
7809 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
7810 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7811 	}
7812 }
7813 
7814 static int (*key_api_typesw[]) (struct socket *, struct mbuf *,
7815 		const struct sadb_msghdr *) = {
7816 	NULL,			/* SADB_RESERVED */
7817 	key_api_getspi,		/* SADB_GETSPI */
7818 	key_api_update,		/* SADB_UPDATE */
7819 	key_api_add,		/* SADB_ADD */
7820 	key_api_delete,		/* SADB_DELETE */
7821 	key_api_get,		/* SADB_GET */
7822 	key_api_acquire,	/* SADB_ACQUIRE */
7823 	key_api_register,	/* SADB_REGISTER */
7824 	NULL,			/* SADB_EXPIRE */
7825 	key_api_flush,		/* SADB_FLUSH */
7826 	key_api_dump,		/* SADB_DUMP */
7827 	key_api_promisc,	/* SADB_X_PROMISC */
7828 	NULL,			/* SADB_X_PCHANGE */
7829 	key_api_spdadd,		/* SADB_X_SPDUPDATE */
7830 	key_api_spdadd,		/* SADB_X_SPDADD */
7831 	key_api_spddelete,	/* SADB_X_SPDDELETE */
7832 	key_api_spdget,		/* SADB_X_SPDGET */
7833 	NULL,			/* SADB_X_SPDACQUIRE */
7834 	key_api_spddump,	/* SADB_X_SPDDUMP */
7835 	key_api_spdflush,	/* SADB_X_SPDFLUSH */
7836 	key_api_spdadd,		/* SADB_X_SPDSETIDX */
7837 	NULL,			/* SADB_X_SPDEXPIRE */
7838 	key_api_spddelete2,	/* SADB_X_SPDDELETE2 */
7839 	key_api_nat_map,	/* SADB_X_NAT_T_NEW_MAPPING */
7840 };
7841 
7842 /*
7843  * parse sadb_msg buffer to process PFKEYv2,
7844  * and create a data to response if needed.
7845  * I think to be dealed with mbuf directly.
7846  * IN:
7847  *     msgp  : pointer to pointer to a received buffer pulluped.
7848  *             This is rewrited to response.
7849  *     so    : pointer to socket.
7850  * OUT:
7851  *    length for buffer to send to user process.
7852  */
7853 int
7854 key_parse(struct mbuf *m, struct socket *so)
7855 {
7856 	struct sadb_msg *msg;
7857 	struct sadb_msghdr mh;
7858 	u_int orglen;
7859 	int error;
7860 
7861 	KASSERT(m != NULL);
7862 	KASSERT(so != NULL);
7863 
7864 #if 0	/*kdebug_sadb assumes msg in linear buffer*/
7865 	if (KEYDEBUG_ON(KEYDEBUG_KEY_DUMP)) {
7866 		kdebug_sadb("passed sadb_msg", msg);
7867 	}
7868 #endif
7869 
7870 	if (m->m_len < sizeof(struct sadb_msg)) {
7871 		m = m_pullup(m, sizeof(struct sadb_msg));
7872 		if (!m)
7873 			return ENOBUFS;
7874 	}
7875 	msg = mtod(m, struct sadb_msg *);
7876 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
7877 
7878 	if ((m->m_flags & M_PKTHDR) == 0 ||
7879 	    m->m_pkthdr.len != orglen) {
7880 		IPSECLOG(LOG_DEBUG, "invalid message length.\n");
7881 		PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
7882 		error = EINVAL;
7883 		goto senderror;
7884 	}
7885 
7886 	if (msg->sadb_msg_version != PF_KEY_V2) {
7887 		IPSECLOG(LOG_DEBUG, "PF_KEY version %u is mismatched.\n",
7888 		    msg->sadb_msg_version);
7889 		PFKEY_STATINC(PFKEY_STAT_OUT_INVVER);
7890 		error = EINVAL;
7891 		goto senderror;
7892 	}
7893 
7894 	if (msg->sadb_msg_type > SADB_MAX) {
7895 		IPSECLOG(LOG_DEBUG, "invalid type %u is passed.\n",
7896 		    msg->sadb_msg_type);
7897 		PFKEY_STATINC(PFKEY_STAT_OUT_INVMSGTYPE);
7898 		error = EINVAL;
7899 		goto senderror;
7900 	}
7901 
7902 	/* for old-fashioned code - should be nuked */
7903 	if (m->m_pkthdr.len > MCLBYTES) {
7904 		m_freem(m);
7905 		return ENOBUFS;
7906 	}
7907 	if (m->m_next) {
7908 		struct mbuf *n;
7909 
7910 		n = key_alloc_mbuf_simple(m->m_pkthdr.len, M_WAITOK);
7911 
7912 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, void *));
7913 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
7914 		n->m_next = NULL;
7915 		m_freem(m);
7916 		m = n;
7917 	}
7918 
7919 	/* align the mbuf chain so that extensions are in contiguous region. */
7920 	error = key_align(m, &mh);
7921 	if (error)
7922 		return error;
7923 
7924 	if (m->m_next) {	/*XXX*/
7925 		m_freem(m);
7926 		return ENOBUFS;
7927 	}
7928 
7929 	msg = mh.msg;
7930 
7931 	/* check SA type */
7932 	switch (msg->sadb_msg_satype) {
7933 	case SADB_SATYPE_UNSPEC:
7934 		switch (msg->sadb_msg_type) {
7935 		case SADB_GETSPI:
7936 		case SADB_UPDATE:
7937 		case SADB_ADD:
7938 		case SADB_DELETE:
7939 		case SADB_GET:
7940 		case SADB_ACQUIRE:
7941 		case SADB_EXPIRE:
7942 			IPSECLOG(LOG_DEBUG,
7943 			    "must specify satype when msg type=%u.\n",
7944 			    msg->sadb_msg_type);
7945 			PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7946 			error = EINVAL;
7947 			goto senderror;
7948 		}
7949 		break;
7950 	case SADB_SATYPE_AH:
7951 	case SADB_SATYPE_ESP:
7952 	case SADB_X_SATYPE_IPCOMP:
7953 	case SADB_X_SATYPE_TCPSIGNATURE:
7954 		switch (msg->sadb_msg_type) {
7955 		case SADB_X_SPDADD:
7956 		case SADB_X_SPDDELETE:
7957 		case SADB_X_SPDGET:
7958 		case SADB_X_SPDDUMP:
7959 		case SADB_X_SPDFLUSH:
7960 		case SADB_X_SPDSETIDX:
7961 		case SADB_X_SPDUPDATE:
7962 		case SADB_X_SPDDELETE2:
7963 			IPSECLOG(LOG_DEBUG, "illegal satype=%u\n",
7964 			    msg->sadb_msg_type);
7965 			PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7966 			error = EINVAL;
7967 			goto senderror;
7968 		}
7969 		break;
7970 	case SADB_SATYPE_RSVP:
7971 	case SADB_SATYPE_OSPFV2:
7972 	case SADB_SATYPE_RIPV2:
7973 	case SADB_SATYPE_MIP:
7974 		IPSECLOG(LOG_DEBUG, "type %u isn't supported.\n",
7975 		    msg->sadb_msg_satype);
7976 		PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7977 		error = EOPNOTSUPP;
7978 		goto senderror;
7979 	case 1:	/* XXX: What does it do? */
7980 		if (msg->sadb_msg_type == SADB_X_PROMISC)
7981 			break;
7982 		/*FALLTHROUGH*/
7983 	default:
7984 		IPSECLOG(LOG_DEBUG, "invalid type %u is passed.\n",
7985 		    msg->sadb_msg_satype);
7986 		PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7987 		error = EINVAL;
7988 		goto senderror;
7989 	}
7990 
7991 	/* check field of upper layer protocol and address family */
7992 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL &&
7993 	    mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
7994 		const struct sadb_address *src0, *dst0;
7995 		const struct sockaddr *sa0, *da0;
7996 		u_int plen;
7997 
7998 		src0 = mh.ext[SADB_EXT_ADDRESS_SRC];
7999 		dst0 = mh.ext[SADB_EXT_ADDRESS_DST];
8000 		sa0 = key_msghdr_get_sockaddr(&mh, SADB_EXT_ADDRESS_SRC);
8001 		da0 = key_msghdr_get_sockaddr(&mh, SADB_EXT_ADDRESS_DST);
8002 
8003 		/* check upper layer protocol */
8004 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
8005 			IPSECLOG(LOG_DEBUG,
8006 			    "upper layer protocol mismatched src %u, dst %u.\n",
8007 			    src0->sadb_address_proto, dst0->sadb_address_proto);
8008 
8009 			goto invaddr;
8010 		}
8011 
8012 		/* check family */
8013 		if (sa0->sa_family != da0->sa_family) {
8014 			IPSECLOG(LOG_DEBUG,
8015 			    "address family mismatched src %u, dst %u.\n",
8016 			    sa0->sa_family, da0->sa_family);
8017 			goto invaddr;
8018 		}
8019 		if (sa0->sa_len != da0->sa_len) {
8020 			IPSECLOG(LOG_DEBUG,
8021 			    "address size mismatched src %u, dst %u.\n",
8022 			    sa0->sa_len, da0->sa_len);
8023 			goto invaddr;
8024 		}
8025 
8026 		switch (sa0->sa_family) {
8027 		case AF_INET:
8028 			if (sa0->sa_len != sizeof(struct sockaddr_in)) {
8029 				IPSECLOG(LOG_DEBUG,
8030 				    "address size mismatched %u != %zu.\n",
8031 				    sa0->sa_len, sizeof(struct sockaddr_in));
8032 				goto invaddr;
8033 			}
8034 			break;
8035 		case AF_INET6:
8036 			if (sa0->sa_len != sizeof(struct sockaddr_in6)) {
8037 				IPSECLOG(LOG_DEBUG,
8038 				    "address size mismatched %u != %zu.\n",
8039 				    sa0->sa_len, sizeof(struct sockaddr_in6));
8040 				goto invaddr;
8041 			}
8042 			break;
8043 		default:
8044 			IPSECLOG(LOG_DEBUG, "unsupported address family %u.\n",
8045 			    sa0->sa_family);
8046 			error = EAFNOSUPPORT;
8047 			goto senderror;
8048 		}
8049 		plen = key_sabits(sa0);
8050 
8051 		/* check max prefix length */
8052 		if (src0->sadb_address_prefixlen > plen ||
8053 		    dst0->sadb_address_prefixlen > plen) {
8054 			IPSECLOG(LOG_DEBUG, "illegal prefixlen.\n");
8055 			goto invaddr;
8056 		}
8057 
8058 		/*
8059 		 * prefixlen == 0 is valid because there can be a case when
8060 		 * all addresses are matched.
8061 		 */
8062 	}
8063 
8064 	if (msg->sadb_msg_type >= __arraycount(key_api_typesw) ||
8065 	    key_api_typesw[msg->sadb_msg_type] == NULL) {
8066 		PFKEY_STATINC(PFKEY_STAT_OUT_INVMSGTYPE);
8067 		error = EINVAL;
8068 		goto senderror;
8069 	}
8070 
8071 	return (*key_api_typesw[msg->sadb_msg_type])(so, m, &mh);
8072 
8073 invaddr:
8074 	error = EINVAL;
8075 senderror:
8076 	PFKEY_STATINC(PFKEY_STAT_OUT_INVADDR);
8077 	return key_senderror(so, m, error);
8078 }
8079 
8080 static int
8081 key_senderror(struct socket *so, struct mbuf *m, int code)
8082 {
8083 	struct sadb_msg *msg;
8084 
8085 	KASSERT(m->m_len >= sizeof(struct sadb_msg));
8086 
8087 	if (so == NULL) {
8088 		/*
8089 		 * This means the request comes from kernel.
8090 		 * As the request comes from kernel, it is unnecessary to
8091 		 * send message to userland. Just return errcode directly.
8092 		 */
8093 		m_freem(m);
8094 		return code;
8095 	}
8096 
8097 	msg = mtod(m, struct sadb_msg *);
8098 	msg->sadb_msg_errno = code;
8099 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
8100 }
8101 
8102 /*
8103  * set the pointer to each header into message buffer.
8104  * m will be freed on error.
8105  * XXX larger-than-MCLBYTES extension?
8106  */
8107 static int
8108 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
8109 {
8110 	struct mbuf *n;
8111 	struct sadb_ext *ext;
8112 	size_t off, end;
8113 	int extlen;
8114 	int toff;
8115 
8116 	KASSERT(m != NULL);
8117 	KASSERT(mhp != NULL);
8118 	KASSERT(m->m_len >= sizeof(struct sadb_msg));
8119 
8120 	/* initialize */
8121 	memset(mhp, 0, sizeof(*mhp));
8122 
8123 	mhp->msg = mtod(m, struct sadb_msg *);
8124 	mhp->ext[0] = mhp->msg;	/*XXX backward compat */
8125 
8126 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
8127 	extlen = end;	/*just in case extlen is not updated*/
8128 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
8129 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
8130 		if (!n) {
8131 			/* m is already freed */
8132 			return ENOBUFS;
8133 		}
8134 		ext = (struct sadb_ext *)(mtod(n, char *) + toff);
8135 
8136 		/* set pointer */
8137 		switch (ext->sadb_ext_type) {
8138 		case SADB_EXT_SA:
8139 		case SADB_EXT_ADDRESS_SRC:
8140 		case SADB_EXT_ADDRESS_DST:
8141 		case SADB_EXT_ADDRESS_PROXY:
8142 		case SADB_EXT_LIFETIME_CURRENT:
8143 		case SADB_EXT_LIFETIME_HARD:
8144 		case SADB_EXT_LIFETIME_SOFT:
8145 		case SADB_EXT_KEY_AUTH:
8146 		case SADB_EXT_KEY_ENCRYPT:
8147 		case SADB_EXT_IDENTITY_SRC:
8148 		case SADB_EXT_IDENTITY_DST:
8149 		case SADB_EXT_SENSITIVITY:
8150 		case SADB_EXT_PROPOSAL:
8151 		case SADB_EXT_SUPPORTED_AUTH:
8152 		case SADB_EXT_SUPPORTED_ENCRYPT:
8153 		case SADB_EXT_SPIRANGE:
8154 		case SADB_X_EXT_POLICY:
8155 		case SADB_X_EXT_SA2:
8156 		case SADB_X_EXT_NAT_T_TYPE:
8157 		case SADB_X_EXT_NAT_T_SPORT:
8158 		case SADB_X_EXT_NAT_T_DPORT:
8159 		case SADB_X_EXT_NAT_T_OAI:
8160 		case SADB_X_EXT_NAT_T_OAR:
8161 		case SADB_X_EXT_NAT_T_FRAG:
8162 			/* duplicate check */
8163 			/*
8164 			 * XXX Are there duplication payloads of either
8165 			 * KEY_AUTH or KEY_ENCRYPT ?
8166 			 */
8167 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
8168 				IPSECLOG(LOG_DEBUG,
8169 				    "duplicate ext_type %u is passed.\n",
8170 				    ext->sadb_ext_type);
8171 				m_freem(m);
8172 				PFKEY_STATINC(PFKEY_STAT_OUT_DUPEXT);
8173 				return EINVAL;
8174 			}
8175 			break;
8176 		default:
8177 			IPSECLOG(LOG_DEBUG, "invalid ext_type %u is passed.\n",
8178 			    ext->sadb_ext_type);
8179 			m_freem(m);
8180 			PFKEY_STATINC(PFKEY_STAT_OUT_INVEXTTYPE);
8181 			return EINVAL;
8182 		}
8183 
8184 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
8185 
8186 		if (key_validate_ext(ext, extlen)) {
8187 			m_freem(m);
8188 			PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
8189 			return EINVAL;
8190 		}
8191 
8192 		n = m_pulldown(m, off, extlen, &toff);
8193 		if (!n) {
8194 			/* m is already freed */
8195 			return ENOBUFS;
8196 		}
8197 		ext = (struct sadb_ext *)(mtod(n, char *) + toff);
8198 
8199 		mhp->ext[ext->sadb_ext_type] = ext;
8200 		mhp->extoff[ext->sadb_ext_type] = off;
8201 		mhp->extlen[ext->sadb_ext_type] = extlen;
8202 	}
8203 
8204 	if (off != end) {
8205 		m_freem(m);
8206 		PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
8207 		return EINVAL;
8208 	}
8209 
8210 	return 0;
8211 }
8212 
8213 static int
8214 key_validate_ext(const struct sadb_ext *ext, int len)
8215 {
8216 	const struct sockaddr *sa;
8217 	enum { NONE, ADDR } checktype = NONE;
8218 	int baselen = 0;
8219 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
8220 
8221 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
8222 		return EINVAL;
8223 
8224 	/* if it does not match minimum/maximum length, bail */
8225 	if (ext->sadb_ext_type >= __arraycount(minsize) ||
8226 	    ext->sadb_ext_type >= __arraycount(maxsize))
8227 		return EINVAL;
8228 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
8229 		return EINVAL;
8230 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
8231 		return EINVAL;
8232 
8233 	/* more checks based on sadb_ext_type XXX need more */
8234 	switch (ext->sadb_ext_type) {
8235 	case SADB_EXT_ADDRESS_SRC:
8236 	case SADB_EXT_ADDRESS_DST:
8237 	case SADB_EXT_ADDRESS_PROXY:
8238 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
8239 		checktype = ADDR;
8240 		break;
8241 	case SADB_EXT_IDENTITY_SRC:
8242 	case SADB_EXT_IDENTITY_DST:
8243 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
8244 		    SADB_X_IDENTTYPE_ADDR) {
8245 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
8246 			checktype = ADDR;
8247 		} else
8248 			checktype = NONE;
8249 		break;
8250 	default:
8251 		checktype = NONE;
8252 		break;
8253 	}
8254 
8255 	switch (checktype) {
8256 	case NONE:
8257 		break;
8258 	case ADDR:
8259 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8260 		if (len < baselen + sal)
8261 			return EINVAL;
8262 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8263 			return EINVAL;
8264 		break;
8265 	}
8266 
8267 	return 0;
8268 }
8269 
8270 static int
8271 key_do_init(void)
8272 {
8273 	int i, error;
8274 
8275 	mutex_init(&key_misc.lock, MUTEX_DEFAULT, IPL_NONE);
8276 
8277 	mutex_init(&key_spd.lock, MUTEX_DEFAULT, IPL_NONE);
8278 	cv_init(&key_spd.cv_lc, "key_sp_lc");
8279 	key_spd.psz = pserialize_create();
8280 	cv_init(&key_spd.cv_psz, "key_sp_psz");
8281 	key_spd.psz_performing = false;
8282 
8283 	mutex_init(&key_sad.lock, MUTEX_DEFAULT, IPL_NONE);
8284 	cv_init(&key_sad.cv_lc, "key_sa_lc");
8285 	key_sad.psz = pserialize_create();
8286 	cv_init(&key_sad.cv_psz, "key_sa_psz");
8287 	key_sad.psz_performing = false;
8288 
8289 	pfkeystat_percpu = percpu_alloc(sizeof(uint64_t) * PFKEY_NSTATS);
8290 
8291 	callout_init(&key_timehandler_ch, CALLOUT_MPSAFE);
8292 	error = workqueue_create(&key_timehandler_wq, "key_timehandler",
8293 	    key_timehandler_work, NULL, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
8294 	if (error != 0)
8295 		panic("%s: workqueue_create failed (%d)\n", __func__, error);
8296 
8297 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8298 		PSLIST_INIT(&key_spd.splist[i]);
8299 	}
8300 
8301 	PSLIST_INIT(&key_spd.socksplist);
8302 
8303 	key_sad.sahlists = hashinit(SAHHASH_NHASH, HASH_PSLIST, true,
8304 	    &key_sad.sahlistmask);
8305 	key_sad.savlut = hashinit(SAVLUT_NHASH, HASH_PSLIST, true,
8306 	    &key_sad.savlutmask);
8307 
8308 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8309 		LIST_INIT(&key_misc.reglist[i]);
8310 	}
8311 
8312 #ifndef IPSEC_NONBLOCK_ACQUIRE
8313 	LIST_INIT(&key_misc.acqlist);
8314 #endif
8315 #ifdef notyet
8316 	LIST_INIT(&key_misc.spacqlist);
8317 #endif
8318 
8319 	/* system default */
8320 	ip4_def_policy.policy = IPSEC_POLICY_NONE;
8321 	ip4_def_policy.state = IPSEC_SPSTATE_ALIVE;
8322 	localcount_init(&ip4_def_policy.localcount);
8323 
8324 #ifdef INET6
8325 	ip6_def_policy.policy = IPSEC_POLICY_NONE;
8326 	ip6_def_policy.state = IPSEC_SPSTATE_ALIVE;
8327 	localcount_init(&ip6_def_policy.localcount);
8328 #endif
8329 
8330 	callout_reset(&key_timehandler_ch, hz, key_timehandler, NULL);
8331 
8332 	/* initialize key statistics */
8333 	keystat.getspi_count = 1;
8334 
8335 	aprint_verbose("IPsec: Initialized Security Association Processing.\n");
8336 
8337 	return (0);
8338 }
8339 
8340 void
8341 key_init(void)
8342 {
8343 	static ONCE_DECL(key_init_once);
8344 
8345 	sysctl_net_keyv2_setup(NULL);
8346 	sysctl_net_key_compat_setup(NULL);
8347 
8348 	RUN_ONCE(&key_init_once, key_do_init);
8349 
8350 	key_init_so();
8351 }
8352 
8353 /*
8354  * XXX: maybe This function is called after INBOUND IPsec processing.
8355  *
8356  * Special check for tunnel-mode packets.
8357  * We must make some checks for consistency between inner and outer IP header.
8358  *
8359  * xxx more checks to be provided
8360  */
8361 int
8362 key_checktunnelsanity(
8363     struct secasvar *sav,
8364     u_int family,
8365     void *src,
8366     void *dst
8367 )
8368 {
8369 
8370 	/* XXX: check inner IP header */
8371 
8372 	return 1;
8373 }
8374 
8375 #if 0
8376 #define hostnamelen	strlen(hostname)
8377 
8378 /*
8379  * Get FQDN for the host.
8380  * If the administrator configured hostname (by hostname(1)) without
8381  * domain name, returns nothing.
8382  */
8383 static const char *
8384 key_getfqdn(void)
8385 {
8386 	int i;
8387 	int hasdot;
8388 	static char fqdn[MAXHOSTNAMELEN + 1];
8389 
8390 	if (!hostnamelen)
8391 		return NULL;
8392 
8393 	/* check if it comes with domain name. */
8394 	hasdot = 0;
8395 	for (i = 0; i < hostnamelen; i++) {
8396 		if (hostname[i] == '.')
8397 			hasdot++;
8398 	}
8399 	if (!hasdot)
8400 		return NULL;
8401 
8402 	/* NOTE: hostname may not be NUL-terminated. */
8403 	memset(fqdn, 0, sizeof(fqdn));
8404 	memcpy(fqdn, hostname, hostnamelen);
8405 	fqdn[hostnamelen] = '\0';
8406 	return fqdn;
8407 }
8408 
8409 /*
8410  * get username@FQDN for the host/user.
8411  */
8412 static const char *
8413 key_getuserfqdn(void)
8414 {
8415 	const char *host;
8416 	static char userfqdn[MAXHOSTNAMELEN + MAXLOGNAME + 2];
8417 	struct proc *p = curproc;
8418 	char *q;
8419 
8420 	if (!p || !p->p_pgrp || !p->p_pgrp->pg_session)
8421 		return NULL;
8422 	if (!(host = key_getfqdn()))
8423 		return NULL;
8424 
8425 	/* NOTE: s_login may not be-NUL terminated. */
8426 	memset(userfqdn, 0, sizeof(userfqdn));
8427 	memcpy(userfqdn, Mp->p_pgrp->pg_session->s_login, AXLOGNAME);
8428 	userfqdn[MAXLOGNAME] = '\0';	/* safeguard */
8429 	q = userfqdn + strlen(userfqdn);
8430 	*q++ = '@';
8431 	memcpy(q, host, strlen(host));
8432 	q += strlen(host);
8433 	*q++ = '\0';
8434 
8435 	return userfqdn;
8436 }
8437 #endif
8438 
8439 /* record data transfer on SA, and update timestamps */
8440 void
8441 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8442 {
8443 	lifetime_counters_t *counters;
8444 
8445 	KASSERT(sav != NULL);
8446 	KASSERT(sav->lft_c != NULL);
8447 	KASSERT(m != NULL);
8448 
8449 	counters = percpu_getref(sav->lft_c_counters_percpu);
8450 
8451 	/*
8452 	 * XXX Currently, there is a difference of bytes size
8453 	 * between inbound and outbound processing.
8454 	 */
8455 	(*counters)[LIFETIME_COUNTER_BYTES] += m->m_pkthdr.len;
8456 	/* to check bytes lifetime is done in key_timehandler(). */
8457 
8458 	/*
8459 	 * We use the number of packets as the unit of
8460 	 * sadb_lifetime_allocations.  We increment the variable
8461 	 * whenever {esp,ah}_{in,out}put is called.
8462 	 */
8463 	(*counters)[LIFETIME_COUNTER_ALLOCATIONS]++;
8464 	/* XXX check for expires? */
8465 
8466 	percpu_putref(sav->lft_c_counters_percpu);
8467 
8468 	/*
8469 	 * NOTE: We record CURRENT sadb_lifetime_usetime by using wall clock,
8470 	 * in seconds.  HARD and SOFT lifetime are measured by the time
8471 	 * difference (again in seconds) from sadb_lifetime_usetime.
8472 	 *
8473 	 *	usetime
8474 	 *	v     expire   expire
8475 	 * -----+-----+--------+---> t
8476 	 *	<--------------> HARD
8477 	 *	<-----> SOFT
8478 	 */
8479 	sav->lft_c->sadb_lifetime_usetime = time_uptime;
8480 	/* XXX check for expires? */
8481 
8482 	return;
8483 }
8484 
8485 /* dumb version */
8486 void
8487 key_sa_routechange(struct sockaddr *dst)
8488 {
8489 	struct secashead *sah;
8490 	int s;
8491 
8492 	s = pserialize_read_enter();
8493 	SAHLIST_READER_FOREACH(sah) {
8494 		struct route *ro;
8495 		const struct sockaddr *sa;
8496 
8497 		key_sah_ref(sah);
8498 		pserialize_read_exit(s);
8499 
8500 		ro = &sah->sa_route;
8501 		sa = rtcache_getdst(ro);
8502 		if (sa != NULL && dst->sa_len == sa->sa_len &&
8503 		    memcmp(dst, sa, dst->sa_len) == 0)
8504 			rtcache_free(ro);
8505 
8506 		s = pserialize_read_enter();
8507 		key_sah_unref(sah);
8508 	}
8509 	pserialize_read_exit(s);
8510 
8511 	return;
8512 }
8513 
8514 static void
8515 key_sa_chgstate(struct secasvar *sav, u_int8_t state)
8516 {
8517 	struct secasvar *_sav;
8518 
8519 	ASSERT_SLEEPABLE();
8520 	KASSERT(mutex_owned(&key_sad.lock));
8521 
8522 	if (sav->state == state)
8523 		return;
8524 
8525 	key_unlink_sav(sav);
8526 	localcount_fini(&sav->localcount);
8527 	SAVLIST_ENTRY_DESTROY(sav);
8528 	key_init_sav(sav);
8529 
8530 	sav->state = state;
8531 	if (!SADB_SASTATE_USABLE_P(sav)) {
8532 		/* We don't need to care about the order */
8533 		SAVLIST_WRITER_INSERT_HEAD(sav->sah, state, sav);
8534 		return;
8535 	}
8536 	/*
8537 	 * Sort the list by lft_c->sadb_lifetime_addtime
8538 	 * in ascending order.
8539 	 */
8540 	SAVLIST_WRITER_FOREACH(_sav, sav->sah, state) {
8541 		if (_sav->lft_c->sadb_lifetime_addtime >
8542 		    sav->lft_c->sadb_lifetime_addtime) {
8543 			SAVLIST_WRITER_INSERT_BEFORE(_sav, sav);
8544 			break;
8545 		}
8546 	}
8547 	if (_sav == NULL) {
8548 		SAVLIST_WRITER_INSERT_TAIL(sav->sah, state, sav);
8549 	}
8550 
8551 	SAVLUT_WRITER_INSERT_HEAD(sav);
8552 
8553 	key_validate_savlist(sav->sah, state);
8554 }
8555 
8556 /* XXX too much? */
8557 static struct mbuf *
8558 key_alloc_mbuf(int l, int mflag)
8559 {
8560 	struct mbuf *m = NULL, *n;
8561 	int len, t;
8562 
8563 	KASSERT(mflag == M_NOWAIT || (mflag == M_WAITOK && !cpu_softintr_p()));
8564 
8565 	len = l;
8566 	while (len > 0) {
8567 		MGET(n, mflag, MT_DATA);
8568 		if (n && len > MLEN) {
8569 			MCLGET(n, mflag);
8570 			if ((n->m_flags & M_EXT) == 0) {
8571 				m_freem(n);
8572 				n = NULL;
8573 			}
8574 		}
8575 		if (!n) {
8576 			m_freem(m);
8577 			return NULL;
8578 		}
8579 
8580 		n->m_next = NULL;
8581 		n->m_len = 0;
8582 		n->m_len = M_TRAILINGSPACE(n);
8583 		/* use the bottom of mbuf, hoping we can prepend afterwards */
8584 		if (n->m_len > len) {
8585 			t = (n->m_len - len) & ~(sizeof(long) - 1);
8586 			n->m_data += t;
8587 			n->m_len = len;
8588 		}
8589 
8590 		len -= n->m_len;
8591 
8592 		if (m)
8593 			m_cat(m, n);
8594 		else
8595 			m = n;
8596 	}
8597 
8598 	return m;
8599 }
8600 
8601 static struct mbuf *
8602 key_setdump(u_int8_t req_satype, int *errorp, uint32_t pid)
8603 {
8604 	struct secashead *sah;
8605 	struct secasvar *sav;
8606 	u_int16_t proto;
8607 	u_int8_t satype;
8608 	u_int8_t state;
8609 	int cnt;
8610 	struct mbuf *m, *n;
8611 
8612 	KASSERT(mutex_owned(&key_sad.lock));
8613 
8614 	/* map satype to proto */
8615 	proto = key_satype2proto(req_satype);
8616 	if (proto == 0) {
8617 		*errorp = EINVAL;
8618 		return (NULL);
8619 	}
8620 
8621 	/* count sav entries to be sent to the userland. */
8622 	cnt = 0;
8623 	SAHLIST_WRITER_FOREACH(sah) {
8624 		if (req_satype != SADB_SATYPE_UNSPEC &&
8625 		    proto != sah->saidx.proto)
8626 			continue;
8627 
8628 		SASTATE_ANY_FOREACH(state) {
8629 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
8630 				cnt++;
8631 			}
8632 		}
8633 	}
8634 
8635 	if (cnt == 0) {
8636 		*errorp = ENOENT;
8637 		return (NULL);
8638 	}
8639 
8640 	/* send this to the userland, one at a time. */
8641 	m = NULL;
8642 	SAHLIST_WRITER_FOREACH(sah) {
8643 		if (req_satype != SADB_SATYPE_UNSPEC &&
8644 		    proto != sah->saidx.proto)
8645 			continue;
8646 
8647 		/* map proto to satype */
8648 		satype = key_proto2satype(sah->saidx.proto);
8649 		if (satype == 0) {
8650 			m_freem(m);
8651 			*errorp = EINVAL;
8652 			return (NULL);
8653 		}
8654 
8655 		SASTATE_ANY_FOREACH(state) {
8656 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
8657 				n = key_setdumpsa(sav, SADB_DUMP, satype,
8658 				    --cnt, pid);
8659 				if (!m)
8660 					m = n;
8661 				else
8662 					m_cat(m, n);
8663 			}
8664 		}
8665 	}
8666 
8667 	if (!m) {
8668 		*errorp = EINVAL;
8669 		return (NULL);
8670 	}
8671 
8672 	if ((m->m_flags & M_PKTHDR) != 0) {
8673 		m->m_pkthdr.len = 0;
8674 		for (n = m; n; n = n->m_next)
8675 			m->m_pkthdr.len += n->m_len;
8676 	}
8677 
8678 	*errorp = 0;
8679 	return (m);
8680 }
8681 
8682 static struct mbuf *
8683 key_setspddump(int *errorp, pid_t pid)
8684 {
8685 	struct secpolicy *sp;
8686 	int cnt;
8687 	u_int dir;
8688 	struct mbuf *m, *n;
8689 
8690 	KASSERT(mutex_owned(&key_spd.lock));
8691 
8692 	/* search SPD entry and get buffer size. */
8693 	cnt = 0;
8694 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
8695 		SPLIST_WRITER_FOREACH(sp, dir) {
8696 			cnt++;
8697 		}
8698 	}
8699 
8700 	if (cnt == 0) {
8701 		*errorp = ENOENT;
8702 		return (NULL);
8703 	}
8704 
8705 	m = NULL;
8706 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
8707 		SPLIST_WRITER_FOREACH(sp, dir) {
8708 			--cnt;
8709 			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, pid);
8710 
8711 			if (!m)
8712 				m = n;
8713 			else {
8714 				m->m_pkthdr.len += n->m_pkthdr.len;
8715 				m_cat(m, n);
8716 			}
8717 		}
8718 	}
8719 
8720 	*errorp = 0;
8721 	return (m);
8722 }
8723 
8724 int
8725 key_get_used(void) {
8726 	return !SPLIST_READER_EMPTY(IPSEC_DIR_INBOUND) ||
8727 	    !SPLIST_READER_EMPTY(IPSEC_DIR_OUTBOUND) ||
8728 	    !SOCKSPLIST_READER_EMPTY();
8729 }
8730 
8731 void
8732 key_update_used(void)
8733 {
8734 	switch (ipsec_enabled) {
8735 	default:
8736 	case 0:
8737 #ifdef notyet
8738 		/* XXX: racy */
8739 		ipsec_used = 0;
8740 #endif
8741 		break;
8742 	case 1:
8743 #ifndef notyet
8744 		/* XXX: racy */
8745 		if (!ipsec_used)
8746 #endif
8747 		ipsec_used = key_get_used();
8748 		break;
8749 	case 2:
8750 		ipsec_used = 1;
8751 		break;
8752 	}
8753 }
8754 
8755 static inline void
8756 key_savlut_writer_insert_head(struct secasvar *sav)
8757 {
8758 	uint32_t hash_key;
8759 	uint32_t hash;
8760 
8761 	KASSERT(mutex_owned(&key_sad.lock));
8762 	KASSERT(!sav->savlut_added);
8763 
8764 	hash_key = sav->spi;
8765 
8766 	hash = key_savluthash(&sav->sah->saidx.dst.sa,
8767 	    sav->sah->saidx.proto, hash_key, key_sad.savlutmask);
8768 
8769 	PSLIST_WRITER_INSERT_HEAD(&key_sad.savlut[hash], sav,
8770 	    pslist_entry_savlut);
8771 	sav->savlut_added = true;
8772 }
8773 
8774 /*
8775  * Calculate hash using protocol, source address,
8776  * and destination address included in saidx.
8777  */
8778 static inline uint32_t
8779 key_saidxhash(const struct secasindex *saidx, u_long mask)
8780 {
8781 	uint32_t hash32;
8782 	const struct sockaddr_in *sin;
8783 	const struct sockaddr_in6 *sin6;
8784 
8785 	hash32 = saidx->proto;
8786 
8787 	switch (saidx->src.sa.sa_family) {
8788 	case AF_INET:
8789 		sin = &saidx->src.sin;
8790 		hash32 = hash32_buf(&sin->sin_addr,
8791 		    sizeof(sin->sin_addr), hash32);
8792 		sin = &saidx->dst.sin;
8793 		hash32 = hash32_buf(&sin->sin_addr,
8794 		    sizeof(sin->sin_addr), hash32 << 1);
8795 		break;
8796 	case AF_INET6:
8797 		sin6 = &saidx->src.sin6;
8798 		hash32 = hash32_buf(&sin6->sin6_addr,
8799 		    sizeof(sin6->sin6_addr), hash32);
8800 		sin6 = &saidx->dst.sin6;
8801 		hash32 = hash32_buf(&sin6->sin6_addr,
8802 		    sizeof(sin6->sin6_addr), hash32 << 1);
8803 		break;
8804 	default:
8805 		hash32 = 0;
8806 		break;
8807 	}
8808 
8809 	return hash32 & mask;
8810 }
8811 
8812 /*
8813  * Calculate hash using destination address, protocol,
8814  * and spi. Those parameter depend on the search of
8815  * key_lookup_sa().
8816  */
8817 static uint32_t
8818 key_savluthash(const struct sockaddr *dst, uint32_t proto,
8819     uint32_t spi, u_long mask)
8820 {
8821 	uint32_t hash32;
8822 	const struct sockaddr_in *sin;
8823 	const struct sockaddr_in6 *sin6;
8824 
8825 	hash32 = hash32_buf(&proto, sizeof(proto), spi);
8826 
8827 	switch(dst->sa_family) {
8828 	case AF_INET:
8829 		sin = satocsin(dst);
8830 		hash32 = hash32_buf(&sin->sin_addr,
8831 		    sizeof(sin->sin_addr), hash32);
8832 		break;
8833 	case AF_INET6:
8834 		sin6 = satocsin6(dst);
8835 		hash32 = hash32_buf(&sin6->sin6_addr,
8836 		    sizeof(sin6->sin6_addr), hash32);
8837 		break;
8838 	default:
8839 		hash32 = 0;
8840 	}
8841 
8842 	return hash32 & mask;
8843 }
8844 
8845 static int
8846 sysctl_net_key_dumpsa(SYSCTLFN_ARGS)
8847 {
8848 	struct mbuf *m, *n;
8849 	int err2 = 0;
8850 	char *p, *ep;
8851 	size_t len;
8852 	int error;
8853 
8854 	if (newp)
8855 		return (EPERM);
8856 	if (namelen != 1)
8857 		return (EINVAL);
8858 
8859 	mutex_enter(&key_sad.lock);
8860 	m = key_setdump(name[0], &error, l->l_proc->p_pid);
8861 	mutex_exit(&key_sad.lock);
8862 	if (!m)
8863 		return (error);
8864 	if (!oldp)
8865 		*oldlenp = m->m_pkthdr.len;
8866 	else {
8867 		p = oldp;
8868 		if (*oldlenp < m->m_pkthdr.len) {
8869 			err2 = ENOMEM;
8870 			ep = p + *oldlenp;
8871 		} else {
8872 			*oldlenp = m->m_pkthdr.len;
8873 			ep = p + m->m_pkthdr.len;
8874 		}
8875 		for (n = m; n; n = n->m_next) {
8876 			len =  (ep - p < n->m_len) ?
8877 				ep - p : n->m_len;
8878 			error = copyout(mtod(n, const void *), p, len);
8879 			p += len;
8880 			if (error)
8881 				break;
8882 		}
8883 		if (error == 0)
8884 			error = err2;
8885 	}
8886 	m_freem(m);
8887 
8888 	return (error);
8889 }
8890 
8891 static int
8892 sysctl_net_key_dumpsp(SYSCTLFN_ARGS)
8893 {
8894 	struct mbuf *m, *n;
8895 	int err2 = 0;
8896 	char *p, *ep;
8897 	size_t len;
8898 	int error;
8899 
8900 	if (newp)
8901 		return (EPERM);
8902 	if (namelen != 0)
8903 		return (EINVAL);
8904 
8905 	mutex_enter(&key_spd.lock);
8906 	m = key_setspddump(&error, l->l_proc->p_pid);
8907 	mutex_exit(&key_spd.lock);
8908 	if (!m)
8909 		return (error);
8910 	if (!oldp)
8911 		*oldlenp = m->m_pkthdr.len;
8912 	else {
8913 		p = oldp;
8914 		if (*oldlenp < m->m_pkthdr.len) {
8915 			err2 = ENOMEM;
8916 			ep = p + *oldlenp;
8917 		} else {
8918 			*oldlenp = m->m_pkthdr.len;
8919 			ep = p + m->m_pkthdr.len;
8920 		}
8921 		for (n = m; n; n = n->m_next) {
8922 			len = (ep - p < n->m_len) ? ep - p : n->m_len;
8923 			error = copyout(mtod(n, const void *), p, len);
8924 			p += len;
8925 			if (error)
8926 				break;
8927 		}
8928 		if (error == 0)
8929 			error = err2;
8930 	}
8931 	m_freem(m);
8932 
8933 	return (error);
8934 }
8935 
8936 /*
8937  * Create sysctl tree for native IPSEC key knobs, originally
8938  * under name "net.keyv2"  * with MIB number { CTL_NET, PF_KEY_V2. }.
8939  * However, sysctl(8) never checked for nodes under { CTL_NET, PF_KEY_V2 };
8940  * and in any case the part of our sysctl namespace used for dumping the
8941  * SPD and SA database  *HAS* to be compatible with the KAME sysctl
8942  * namespace, for API reasons.
8943  *
8944  * Pending a consensus on the right way  to fix this, add a level of
8945  * indirection in how we number the `native' IPSEC key nodes;
8946  * and (as requested by Andrew Brown)  move registration of the
8947  * KAME-compatible names  to a separate function.
8948  */
8949 #if 0
8950 #  define IPSEC_PFKEY PF_KEY_V2
8951 # define IPSEC_PFKEY_NAME "keyv2"
8952 #else
8953 #  define IPSEC_PFKEY PF_KEY
8954 # define IPSEC_PFKEY_NAME "key"
8955 #endif
8956 
8957 static int
8958 sysctl_net_key_stats(SYSCTLFN_ARGS)
8959 {
8960 
8961 	return (NETSTAT_SYSCTL(pfkeystat_percpu, PFKEY_NSTATS));
8962 }
8963 
8964 static void
8965 sysctl_net_keyv2_setup(struct sysctllog **clog)
8966 {
8967 
8968 	sysctl_createv(clog, 0, NULL, NULL,
8969 		       CTLFLAG_PERMANENT,
8970 		       CTLTYPE_NODE, IPSEC_PFKEY_NAME, NULL,
8971 		       NULL, 0, NULL, 0,
8972 		       CTL_NET, IPSEC_PFKEY, CTL_EOL);
8973 
8974 	sysctl_createv(clog, 0, NULL, NULL,
8975 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8976 		       CTLTYPE_INT, "debug", NULL,
8977 		       NULL, 0, &key_debug_level, 0,
8978 		       CTL_NET, IPSEC_PFKEY, KEYCTL_DEBUG_LEVEL, CTL_EOL);
8979 	sysctl_createv(clog, 0, NULL, NULL,
8980 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8981 		       CTLTYPE_INT, "spi_try", NULL,
8982 		       NULL, 0, &key_spi_trycnt, 0,
8983 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_TRY, CTL_EOL);
8984 	sysctl_createv(clog, 0, NULL, NULL,
8985 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8986 		       CTLTYPE_INT, "spi_min_value", NULL,
8987 		       NULL, 0, &key_spi_minval, 0,
8988 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_MIN_VALUE, CTL_EOL);
8989 	sysctl_createv(clog, 0, NULL, NULL,
8990 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8991 		       CTLTYPE_INT, "spi_max_value", NULL,
8992 		       NULL, 0, &key_spi_maxval, 0,
8993 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_MAX_VALUE, CTL_EOL);
8994 	sysctl_createv(clog, 0, NULL, NULL,
8995 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8996 		       CTLTYPE_INT, "random_int", NULL,
8997 		       NULL, 0, &key_int_random, 0,
8998 		       CTL_NET, IPSEC_PFKEY, KEYCTL_RANDOM_INT, CTL_EOL);
8999 	sysctl_createv(clog, 0, NULL, NULL,
9000 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9001 		       CTLTYPE_INT, "larval_lifetime", NULL,
9002 		       NULL, 0, &key_larval_lifetime, 0,
9003 		       CTL_NET, IPSEC_PFKEY, KEYCTL_LARVAL_LIFETIME, CTL_EOL);
9004 	sysctl_createv(clog, 0, NULL, NULL,
9005 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9006 		       CTLTYPE_INT, "blockacq_count", NULL,
9007 		       NULL, 0, &key_blockacq_count, 0,
9008 		       CTL_NET, IPSEC_PFKEY, KEYCTL_BLOCKACQ_COUNT, CTL_EOL);
9009 	sysctl_createv(clog, 0, NULL, NULL,
9010 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9011 		       CTLTYPE_INT, "blockacq_lifetime", NULL,
9012 		       NULL, 0, &key_blockacq_lifetime, 0,
9013 		       CTL_NET, IPSEC_PFKEY, KEYCTL_BLOCKACQ_LIFETIME, CTL_EOL);
9014 	sysctl_createv(clog, 0, NULL, NULL,
9015 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9016 		       CTLTYPE_INT, "esp_keymin", NULL,
9017 		       NULL, 0, &ipsec_esp_keymin, 0,
9018 		       CTL_NET, IPSEC_PFKEY, KEYCTL_ESP_KEYMIN, CTL_EOL);
9019 	sysctl_createv(clog, 0, NULL, NULL,
9020 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9021 		       CTLTYPE_INT, "prefered_oldsa", NULL,
9022 		       NULL, 0, &key_prefered_oldsa, 0,
9023 		       CTL_NET, PF_KEY, KEYCTL_PREFERED_OLDSA, CTL_EOL);
9024 	sysctl_createv(clog, 0, NULL, NULL,
9025 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9026 		       CTLTYPE_INT, "esp_auth", NULL,
9027 		       NULL, 0, &ipsec_esp_auth, 0,
9028 		       CTL_NET, IPSEC_PFKEY, KEYCTL_ESP_AUTH, CTL_EOL);
9029 	sysctl_createv(clog, 0, NULL, NULL,
9030 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9031 		       CTLTYPE_INT, "ah_keymin", NULL,
9032 		       NULL, 0, &ipsec_ah_keymin, 0,
9033 		       CTL_NET, IPSEC_PFKEY, KEYCTL_AH_KEYMIN, CTL_EOL);
9034 	sysctl_createv(clog, 0, NULL, NULL,
9035 		       CTLFLAG_PERMANENT,
9036 		       CTLTYPE_STRUCT, "stats",
9037 		       SYSCTL_DESCR("PF_KEY statistics"),
9038 		       sysctl_net_key_stats, 0, NULL, 0,
9039 		       CTL_NET, IPSEC_PFKEY, CTL_CREATE, CTL_EOL);
9040 	sysctl_createv(clog, 0, NULL, NULL,
9041 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9042 		       CTLTYPE_BOOL, "allow_different_idtype", NULL,
9043 		       NULL, 0, &ipsec_allow_different_idtype, 0,
9044 		       CTL_NET, IPSEC_PFKEY, KEYCTL_ALLOW_DIFFERENT_IDTYPE, CTL_EOL);
9045 }
9046 
9047 /*
9048  * Register sysctl names used by setkey(8). For historical reasons,
9049  * and to share a single API, these names appear under { CTL_NET, PF_KEY }
9050  * for both IPSEC and KAME IPSEC.
9051  */
9052 static void
9053 sysctl_net_key_compat_setup(struct sysctllog **clog)
9054 {
9055 
9056 	sysctl_createv(clog, 0, NULL, NULL,
9057 		       CTLFLAG_PERMANENT,
9058 		       CTLTYPE_NODE, "key", NULL,
9059 		       NULL, 0, NULL, 0,
9060 		       CTL_NET, PF_KEY, CTL_EOL);
9061 
9062 	/* Register the net.key.dump{sa,sp} nodes used by setkey(8). */
9063 	sysctl_createv(clog, 0, NULL, NULL,
9064 		       CTLFLAG_PERMANENT,
9065 		       CTLTYPE_STRUCT, "dumpsa", NULL,
9066 		       sysctl_net_key_dumpsa, 0, NULL, 0,
9067 		       CTL_NET, PF_KEY, KEYCTL_DUMPSA, CTL_EOL);
9068 	sysctl_createv(clog, 0, NULL, NULL,
9069 		       CTLFLAG_PERMANENT,
9070 		       CTLTYPE_STRUCT, "dumpsp", NULL,
9071 		       sysctl_net_key_dumpsp, 0, NULL, 0,
9072 		       CTL_NET, PF_KEY, KEYCTL_DUMPSP, CTL_EOL);
9073 }
9074