xref: /netbsd-src/sys/netipsec/key.c (revision 2d8e86c2f207da6fbbd50f11b6f33765ebdfa0e9)
1 /*	$NetBSD: key.c,v 1.266 2019/08/04 14:30:36 maxv 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 1.266 2019/08/04 14:30:36 maxv Exp $");
36 
37 /*
38  * This code is referred to RFC 2367
39  */
40 
41 #if defined(_KERNEL_OPT)
42 #include "opt_inet.h"
43 #include "opt_ipsec.h"
44 #include "opt_gateway.h"
45 #include "opt_net_mpsafe.h"
46 #endif
47 
48 #include <sys/types.h>
49 #include <sys/param.h>
50 #include <sys/systm.h>
51 #include <sys/callout.h>
52 #include <sys/kernel.h>
53 #include <sys/mbuf.h>
54 #include <sys/domain.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sysctl.h>
58 #include <sys/errno.h>
59 #include <sys/proc.h>
60 #include <sys/queue.h>
61 #include <sys/syslog.h>
62 #include <sys/once.h>
63 #include <sys/cprng.h>
64 #include <sys/psref.h>
65 #include <sys/lwp.h>
66 #include <sys/workqueue.h>
67 #include <sys/kmem.h>
68 #include <sys/cpu.h>
69 #include <sys/atomic.h>
70 #include <sys/pslist.h>
71 #include <sys/mutex.h>
72 #include <sys/condvar.h>
73 #include <sys/localcount.h>
74 #include <sys/pserialize.h>
75 #include <sys/hash.h>
76 
77 #include <net/if.h>
78 #include <net/route.h>
79 
80 #include <netinet/in.h>
81 #include <netinet/in_systm.h>
82 #include <netinet/ip.h>
83 #include <netinet/in_var.h>
84 #ifdef INET
85 #include <netinet/ip_var.h>
86 #endif
87 
88 #ifdef INET6
89 #include <netinet/ip6.h>
90 #include <netinet6/in6_var.h>
91 #include <netinet6/ip6_var.h>
92 #endif /* INET6 */
93 
94 #ifdef INET
95 #include <netinet/in_pcb.h>
96 #endif
97 #ifdef INET6
98 #include <netinet6/in6_pcb.h>
99 #endif /* INET6 */
100 
101 #include <net/pfkeyv2.h>
102 #include <netipsec/keydb.h>
103 #include <netipsec/key.h>
104 #include <netipsec/keysock.h>
105 #include <netipsec/key_debug.h>
106 
107 #include <netipsec/ipsec.h>
108 #ifdef INET6
109 #include <netipsec/ipsec6.h>
110 #endif
111 #include <netipsec/ipsec_private.h>
112 
113 #include <netipsec/xform.h>
114 #include <netipsec/ipcomp.h>
115 
116 #define FULLMASK	0xffu
117 #define	_BITS(bytes)	((bytes) << 3)
118 
119 #define PORT_NONE	0
120 #define PORT_LOOSE	1
121 #define PORT_STRICT	2
122 
123 #ifndef SAHHASH_NHASH
124 #define SAHHASH_NHASH		128
125 #endif
126 
127 #ifndef SAVLUT_NHASH
128 #define SAVLUT_NHASH		128
129 #endif
130 
131 percpu_t *pfkeystat_percpu;
132 
133 /*
134  * Note on SA reference counting:
135  * - SAs that are not in DEAD state will have (total external reference + 1)
136  *   following value in reference count field.  they cannot be freed and are
137  *   referenced from SA header.
138  * - SAs that are in DEAD state will have (total external reference)
139  *   in reference count field.  they are ready to be freed.  reference from
140  *   SA header will be removed in key_delsav(), when the reference count
141  *   field hits 0 (= no external reference other than from SA header.
142  */
143 
144 u_int32_t key_debug_level = 0;
145 static u_int key_spi_trycnt = 1000;
146 static u_int32_t key_spi_minval = 0x100;
147 static u_int32_t key_spi_maxval = 0x0fffffff;	/* XXX */
148 static u_int32_t policy_id = 0;
149 static u_int key_int_random = 60;	/*interval to initialize randseed,1(m)*/
150 static u_int key_larval_lifetime = 30;	/* interval to expire acquiring, 30(s)*/
151 static int key_blockacq_count = 10;	/* counter for blocking SADB_ACQUIRE.*/
152 static int key_blockacq_lifetime = 20;	/* lifetime for blocking SADB_ACQUIRE.*/
153 static int key_prefered_oldsa = 0;	/* prefered old sa rather than new sa.*/
154 
155 static u_int32_t acq_seq = 0;
156 
157 /*
158  * Locking order: there is no order for now; it means that any locks aren't
159  * overlapped.
160  */
161 /*
162  * Locking notes on SPD:
163  * - Modifications to the key_spd.splist must be done with holding key_spd.lock
164  *   which is a adaptive mutex
165  * - Read accesses to the key_spd.splist must be in pserialize(9) read sections
166  * - SP's lifetime is managed by localcount(9)
167  * - An SP that has been inserted to the key_spd.splist is initially referenced
168  *   by none, i.e., a reference from the key_spd.splist isn't counted
169  * - When an SP is being destroyed, we change its state as DEAD, wait for
170  *   references to the SP to be released, and then deallocate the SP
171  *   (see key_unlink_sp)
172  * - Getting an SP
173  *   - Normally we get an SP from the key_spd.splist (see key_lookup_sp_byspidx)
174  *     - Must iterate the list and increment the reference count of a found SP
175  *       (by key_sp_ref) in a pserialize read section
176  *   - We can gain another reference from a held SP only if we check its state
177  *     and take its reference in a pserialize read section
178  *     (see esp_output for example)
179  *   - We may get an SP from an SP cache. See below
180  *   - A gotten SP must be released after use by KEY_SP_UNREF (key_sp_unref)
181  * - Updating member variables of an SP
182  *   - Most member variables of an SP are immutable
183  *   - Only sp->state and sp->lastused can be changed
184  *   - sp->state of an SP is updated only when destroying it under key_spd.lock
185  * - SP caches
186  *   - SPs can be cached in PCBs
187  *   - The lifetime of the caches is controlled by the global generation counter
188  *     (ipsec_spdgen)
189  *   - The global counter value is stored when an SP is cached
190  *   - If the stored value is different from the global counter then the cache
191  *     is considered invalidated
192  *   - The counter is incremented when an SP is being destroyed
193  *   - So checking the generation and taking a reference to an SP should be
194  *     in a pserialize read section
195  *   - Note that caching doesn't increment the reference counter of an SP
196  * - SPs in sockets
197  *   - Userland programs can set a policy to a socket by
198  *     setsockopt(IP_IPSEC_POLICY)
199  *   - Such policies (SPs) are set to a socket (PCB) and also inserted to
200  *     the key_spd.socksplist list (not the key_spd.splist)
201  *   - Such a policy is destroyed when a corresponding socket is destroed,
202  *     however, a socket can be destroyed in softint so we cannot destroy
203  *     it directly instead we just mark it DEAD and delay the destruction
204  *     until GC by the timer
205  * - SP origin
206  *   - SPs can be created by both userland programs and kernel components.
207  *     The SPs created in kernel must not be removed by userland programs,
208  *     although the SPs can be read by userland programs.
209  */
210 /*
211  * Locking notes on SAD:
212  * - Data structures
213  *   - SAs are managed by the list called key_sad.sahlists and sav lists of
214  *     sah entries
215  *     - An sav is supposed to be an SA from a viewpoint of users
216  *   - A sah has sav lists for each SA state
217  *   - Multiple saves with the same saidx can exist
218  *     - Only one entry has MATURE state and others should be DEAD
219  *     - DEAD entries are just ignored from searching
220  *   - All sav whose state is MATURE or DYING are registered to the lookup
221  *     table called key_sad.savlut in addition to the savlists.
222  *     - The table is used to search an sav without use of saidx.
223  * - Modifications to the key_sad.sahlists, sah.savlist and key_sad.savlut
224  *   must be done with holding key_sad.lock which is a adaptive mutex
225  * - Read accesses to the key_sad.sahlists, sah.savlist and key_sad.savlut
226  *   must be in pserialize(9) read sections
227  * - sah's lifetime is managed by localcount(9)
228  * - Getting an sah entry
229  *   - We get an sah from the key_sad.sahlists
230  *     - Must iterate the list and increment the reference count of a found sah
231  *       (by key_sah_ref) in a pserialize read section
232  *   - A gotten sah must be released after use by key_sah_unref
233  * - An sah is destroyed when its state become DEAD and no sav is
234  *   listed to the sah
235  *   - The destruction is done only in the timer (see key_timehandler_sad)
236  * - sav's lifetime is managed by localcount(9)
237  * - Getting an sav entry
238  *   - First get an sah by saidx and get an sav from either of sah's savlists
239  *     - Must iterate the list and increment the reference count of a found sav
240  *       (by key_sa_ref) in a pserialize read section
241  *   - We can gain another reference from a held SA only if we check its state
242  *     and take its reference in a pserialize read section
243  *     (see esp_output for example)
244  *   - A gotten sav must be released after use by key_sa_unref
245  * - An sav is destroyed when its state become DEAD
246  */
247 /*
248  * Locking notes on misc data:
249  * - All lists of key_misc are protected by key_misc.lock
250  *   - key_misc.lock must be held even for read accesses
251  */
252 
253 /* SPD */
254 static struct {
255 	kmutex_t lock;
256 	kcondvar_t cv_lc;
257 	struct pslist_head splist[IPSEC_DIR_MAX];
258 	/*
259 	 * The list has SPs that are set to a socket via
260 	 * setsockopt(IP_IPSEC_POLICY) from userland. See ipsec_set_policy.
261 	 */
262 	struct pslist_head socksplist;
263 
264 	pserialize_t psz;
265 	kcondvar_t cv_psz;
266 	bool psz_performing;
267 } key_spd __cacheline_aligned;
268 
269 /* SAD */
270 static struct {
271 	kmutex_t lock;
272 	kcondvar_t cv_lc;
273 	struct pslist_head *sahlists;
274 	u_long sahlistmask;
275 	struct pslist_head *savlut;
276 	u_long savlutmask;
277 
278 	pserialize_t psz;
279 	kcondvar_t cv_psz;
280 	bool psz_performing;
281 } key_sad __cacheline_aligned;
282 
283 /* Misc data */
284 static struct {
285 	kmutex_t lock;
286 	/* registed list */
287 	LIST_HEAD(_reglist, secreg) reglist[SADB_SATYPE_MAX + 1];
288 #ifndef IPSEC_NONBLOCK_ACQUIRE
289 	/* acquiring list */
290 	LIST_HEAD(_acqlist, secacq) acqlist;
291 #endif
292 #ifdef notyet
293 	/* SP acquiring list */
294 	LIST_HEAD(_spacqlist, secspacq) spacqlist;
295 #endif
296 } key_misc __cacheline_aligned;
297 
298 /* Macros for key_spd.splist */
299 #define SPLIST_ENTRY_INIT(sp)						\
300 	PSLIST_ENTRY_INIT((sp), pslist_entry)
301 #define SPLIST_ENTRY_DESTROY(sp)					\
302 	PSLIST_ENTRY_DESTROY((sp), pslist_entry)
303 #define SPLIST_WRITER_REMOVE(sp)					\
304 	PSLIST_WRITER_REMOVE((sp), pslist_entry)
305 #define SPLIST_READER_EMPTY(dir)					\
306 	(PSLIST_READER_FIRST(&key_spd.splist[(dir)], struct secpolicy,	\
307 	                     pslist_entry) == NULL)
308 #define SPLIST_READER_FOREACH(sp, dir)					\
309 	PSLIST_READER_FOREACH((sp), &key_spd.splist[(dir)],		\
310 	                      struct secpolicy, pslist_entry)
311 #define SPLIST_WRITER_FOREACH(sp, dir)					\
312 	PSLIST_WRITER_FOREACH((sp), &key_spd.splist[(dir)],		\
313 	                      struct secpolicy, pslist_entry)
314 #define SPLIST_WRITER_INSERT_AFTER(sp, new)				\
315 	PSLIST_WRITER_INSERT_AFTER((sp), (new), pslist_entry)
316 #define SPLIST_WRITER_EMPTY(dir)					\
317 	(PSLIST_WRITER_FIRST(&key_spd.splist[(dir)], struct secpolicy,	\
318 	                     pslist_entry) == NULL)
319 #define SPLIST_WRITER_INSERT_HEAD(dir, sp)				\
320 	PSLIST_WRITER_INSERT_HEAD(&key_spd.splist[(dir)], (sp),		\
321 	                          pslist_entry)
322 #define SPLIST_WRITER_NEXT(sp)						\
323 	PSLIST_WRITER_NEXT((sp), struct secpolicy, pslist_entry)
324 #define SPLIST_WRITER_INSERT_TAIL(dir, new)				\
325 	do {								\
326 		if (SPLIST_WRITER_EMPTY((dir))) {			\
327 			SPLIST_WRITER_INSERT_HEAD((dir), (new));	\
328 		} else {						\
329 			struct secpolicy *__sp;				\
330 			SPLIST_WRITER_FOREACH(__sp, (dir)) {		\
331 				if (SPLIST_WRITER_NEXT(__sp) == NULL) {	\
332 					SPLIST_WRITER_INSERT_AFTER(__sp,\
333 					    (new));			\
334 					break;				\
335 				}					\
336 			}						\
337 		}							\
338 	} while (0)
339 
340 /* Macros for key_spd.socksplist */
341 #define SOCKSPLIST_WRITER_FOREACH(sp)					\
342 	PSLIST_WRITER_FOREACH((sp), &key_spd.socksplist,		\
343 	                      struct secpolicy,	pslist_entry)
344 #define SOCKSPLIST_READER_EMPTY()					\
345 	(PSLIST_READER_FIRST(&key_spd.socksplist, struct secpolicy,	\
346 	                     pslist_entry) == NULL)
347 
348 /* Macros for key_sad.sahlist */
349 #define SAHLIST_ENTRY_INIT(sah)						\
350 	PSLIST_ENTRY_INIT((sah), pslist_entry)
351 #define SAHLIST_ENTRY_DESTROY(sah)					\
352 	PSLIST_ENTRY_DESTROY((sah), pslist_entry)
353 #define SAHLIST_WRITER_REMOVE(sah)					\
354 	PSLIST_WRITER_REMOVE((sah), pslist_entry)
355 #define SAHLIST_READER_FOREACH(sah)					\
356 	for(int _i_sah = 0; _i_sah <= key_sad.sahlistmask; _i_sah++)	\
357 		PSLIST_READER_FOREACH((sah), &key_sad.sahlists[_i_sah],	\
358 		                      struct secashead, pslist_entry)
359 #define SAHLIST_READER_FOREACH_SAIDX(sah, saidx)			\
360 	PSLIST_READER_FOREACH((sah),					\
361 	    &key_sad.sahlists[key_saidxhash((saidx),			\
362 	                       key_sad.sahlistmask)],			\
363 	    struct secashead, pslist_entry)
364 #define SAHLIST_WRITER_FOREACH(sah)					\
365 	for(int _i_sah = 0; _i_sah <= key_sad.sahlistmask; _i_sah++)	\
366 		PSLIST_WRITER_FOREACH((sah), &key_sad.sahlists[_i_sah],	\
367 		                     struct secashead, pslist_entry)
368 #define SAHLIST_WRITER_INSERT_HEAD(sah)					\
369 	PSLIST_WRITER_INSERT_HEAD(					\
370 	    &key_sad.sahlists[key_saidxhash(&(sah)->saidx,		\
371 	                      key_sad.sahlistmask)],	\
372 	    (sah), pslist_entry)
373 
374 /* Macros for key_sad.sahlist#savlist */
375 #define SAVLIST_ENTRY_INIT(sav)						\
376 	PSLIST_ENTRY_INIT((sav), pslist_entry)
377 #define SAVLIST_ENTRY_DESTROY(sav)					\
378 	PSLIST_ENTRY_DESTROY((sav), pslist_entry)
379 #define SAVLIST_READER_FIRST(sah, state)				\
380 	PSLIST_READER_FIRST(&(sah)->savlist[(state)], struct secasvar,	\
381 	                    pslist_entry)
382 #define SAVLIST_WRITER_REMOVE(sav)					\
383 	PSLIST_WRITER_REMOVE((sav), pslist_entry)
384 #define SAVLIST_READER_FOREACH(sav, sah, state)				\
385 	PSLIST_READER_FOREACH((sav), &(sah)->savlist[(state)],		\
386 	                      struct secasvar, pslist_entry)
387 #define SAVLIST_WRITER_FOREACH(sav, sah, state)				\
388 	PSLIST_WRITER_FOREACH((sav), &(sah)->savlist[(state)],		\
389 	                      struct secasvar, pslist_entry)
390 #define SAVLIST_WRITER_INSERT_BEFORE(sav, new)				\
391 	PSLIST_WRITER_INSERT_BEFORE((sav), (new), pslist_entry)
392 #define SAVLIST_WRITER_INSERT_AFTER(sav, new)				\
393 	PSLIST_WRITER_INSERT_AFTER((sav), (new), pslist_entry)
394 #define SAVLIST_WRITER_EMPTY(sah, state)				\
395 	(PSLIST_WRITER_FIRST(&(sah)->savlist[(state)], struct secasvar,	\
396 	                     pslist_entry) == NULL)
397 #define SAVLIST_WRITER_INSERT_HEAD(sah, state, sav)			\
398 	PSLIST_WRITER_INSERT_HEAD(&(sah)->savlist[(state)], (sav),	\
399 	                          pslist_entry)
400 #define SAVLIST_WRITER_NEXT(sav)					\
401 	PSLIST_WRITER_NEXT((sav), struct secasvar, pslist_entry)
402 #define SAVLIST_WRITER_INSERT_TAIL(sah, state, new)			\
403 	do {								\
404 		if (SAVLIST_WRITER_EMPTY((sah), (state))) {		\
405 			SAVLIST_WRITER_INSERT_HEAD((sah), (state), (new));\
406 		} else {						\
407 			struct secasvar *__sav;				\
408 			SAVLIST_WRITER_FOREACH(__sav, (sah), (state)) {	\
409 				if (SAVLIST_WRITER_NEXT(__sav) == NULL) {\
410 					SAVLIST_WRITER_INSERT_AFTER(__sav,\
411 					    (new));			\
412 					break;				\
413 				}					\
414 			}						\
415 		}							\
416 	} while (0)
417 #define SAVLIST_READER_NEXT(sav)					\
418 	PSLIST_READER_NEXT((sav), struct secasvar, pslist_entry)
419 
420 /* Macros for key_sad.savlut */
421 #define SAVLUT_ENTRY_INIT(sav)						\
422 	PSLIST_ENTRY_INIT((sav), pslist_entry_savlut)
423 #define SAVLUT_READER_FOREACH(sav, dst, proto, hash_key)		\
424 	PSLIST_READER_FOREACH((sav),					\
425 	&key_sad.savlut[key_savluthash(dst, proto, hash_key,		\
426 	                  key_sad.savlutmask)],				\
427 	struct secasvar, pslist_entry_savlut)
428 #define SAVLUT_WRITER_INSERT_HEAD(sav)					\
429 	key_savlut_writer_insert_head((sav))
430 #define SAVLUT_WRITER_REMOVE(sav)					\
431 	do {								\
432 		if (!(sav)->savlut_added)				\
433 			break;						\
434 		PSLIST_WRITER_REMOVE((sav), pslist_entry_savlut);	\
435 		(sav)->savlut_added = false;				\
436 	} while(0)
437 
438 /* search order for SAs */
439 	/*
440 	 * This order is important because we must select the oldest SA
441 	 * for outbound processing.  For inbound, This is not important.
442 	 */
443 static const u_int saorder_state_valid_prefer_old[] = {
444 	SADB_SASTATE_DYING, SADB_SASTATE_MATURE,
445 };
446 static const u_int saorder_state_valid_prefer_new[] = {
447 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING,
448 };
449 
450 static const u_int saorder_state_alive[] = {
451 	/* except DEAD */
452 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING, SADB_SASTATE_LARVAL
453 };
454 static const u_int saorder_state_any[] = {
455 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING,
456 	SADB_SASTATE_LARVAL, SADB_SASTATE_DEAD
457 };
458 
459 #define SASTATE_ALIVE_FOREACH(s)				\
460 	for (int _i = 0;					\
461 	    _i < __arraycount(saorder_state_alive) ?		\
462 	    (s) = saorder_state_alive[_i], true : false;	\
463 	    _i++)
464 #define SASTATE_ANY_FOREACH(s)					\
465 	for (int _i = 0;					\
466 	    _i < __arraycount(saorder_state_any) ?		\
467 	    (s) = saorder_state_any[_i], true : false;		\
468 	    _i++)
469 #define SASTATE_USABLE_FOREACH(s)				\
470 	for (int _i = 0;					\
471 	    _i < __arraycount(saorder_state_valid_prefer_new) ?	\
472 	    (s) = saorder_state_valid_prefer_new[_i],		\
473 	    true : false;					\
474 	    _i++)
475 
476 static const int minsize[] = {
477 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
478 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
479 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
480 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
481 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
482 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_SRC */
483 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_DST */
484 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_PROXY */
485 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_AUTH */
486 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_ENCRYPT */
487 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_SRC */
488 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_DST */
489 	sizeof(struct sadb_sens),	/* SADB_EXT_SENSITIVITY */
490 	sizeof(struct sadb_prop),	/* SADB_EXT_PROPOSAL */
491 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_AUTH */
492 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_ENCRYPT */
493 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
494 	0,				/* SADB_X_EXT_KMPRIVATE */
495 	sizeof(struct sadb_x_policy),	/* SADB_X_EXT_POLICY */
496 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
497 	sizeof(struct sadb_x_nat_t_type),	/* SADB_X_EXT_NAT_T_TYPE */
498 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_SPORT */
499 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_DPORT */
500 	sizeof(struct sadb_address),		/* SADB_X_EXT_NAT_T_OAI */
501 	sizeof(struct sadb_address),		/* SADB_X_EXT_NAT_T_OAR */
502 	sizeof(struct sadb_x_nat_t_frag),	/* SADB_X_EXT_NAT_T_FRAG */
503 };
504 static const int maxsize[] = {
505 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
506 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
507 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
508 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
509 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
510 	0,				/* SADB_EXT_ADDRESS_SRC */
511 	0,				/* SADB_EXT_ADDRESS_DST */
512 	0,				/* SADB_EXT_ADDRESS_PROXY */
513 	0,				/* SADB_EXT_KEY_AUTH */
514 	0,				/* SADB_EXT_KEY_ENCRYPT */
515 	0,				/* SADB_EXT_IDENTITY_SRC */
516 	0,				/* SADB_EXT_IDENTITY_DST */
517 	0,				/* SADB_EXT_SENSITIVITY */
518 	0,				/* SADB_EXT_PROPOSAL */
519 	0,				/* SADB_EXT_SUPPORTED_AUTH */
520 	0,				/* SADB_EXT_SUPPORTED_ENCRYPT */
521 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
522 	0,				/* SADB_X_EXT_KMPRIVATE */
523 	0,				/* SADB_X_EXT_POLICY */
524 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
525 	sizeof(struct sadb_x_nat_t_type),	/* SADB_X_EXT_NAT_T_TYPE */
526 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_SPORT */
527 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_DPORT */
528 	0,					/* SADB_X_EXT_NAT_T_OAI */
529 	0,					/* SADB_X_EXT_NAT_T_OAR */
530 	sizeof(struct sadb_x_nat_t_frag),	/* SADB_X_EXT_NAT_T_FRAG */
531 };
532 
533 static int ipsec_esp_keymin = 256;
534 static int ipsec_esp_auth = 0;
535 static int ipsec_ah_keymin = 128;
536 
537 #ifdef SYSCTL_DECL
538 SYSCTL_DECL(_net_key);
539 #endif
540 
541 #ifdef SYSCTL_INT
542 SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,	CTLFLAG_RW, \
543 	&key_debug_level,	0,	"");
544 
545 /* max count of trial for the decision of spi value */
546 SYSCTL_INT(_net_key, KEYCTL_SPI_TRY,		spi_trycnt,	CTLFLAG_RW, \
547 	&key_spi_trycnt,	0,	"");
548 
549 /* minimum spi value to allocate automatically. */
550 SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE,	spi_minval,	CTLFLAG_RW, \
551 	&key_spi_minval,	0,	"");
552 
553 /* maximun spi value to allocate automatically. */
554 SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE,	spi_maxval,	CTLFLAG_RW, \
555 	&key_spi_maxval,	0,	"");
556 
557 /* interval to initialize randseed */
558 SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT,	int_random,	CTLFLAG_RW, \
559 	&key_int_random,	0,	"");
560 
561 /* lifetime for larval SA */
562 SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME,	larval_lifetime, CTLFLAG_RW, \
563 	&key_larval_lifetime,	0,	"");
564 
565 /* counter for blocking to send SADB_ACQUIRE to IKEd */
566 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT,	blockacq_count,	CTLFLAG_RW, \
567 	&key_blockacq_count,	0,	"");
568 
569 /* lifetime for blocking to send SADB_ACQUIRE to IKEd */
570 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME,	blockacq_lifetime, CTLFLAG_RW, \
571 	&key_blockacq_lifetime,	0,	"");
572 
573 /* ESP auth */
574 SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH,	esp_auth, CTLFLAG_RW, \
575 	&ipsec_esp_auth,	0,	"");
576 
577 /* minimum ESP key length */
578 SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN,	esp_keymin, CTLFLAG_RW, \
579 	&ipsec_esp_keymin,	0,	"");
580 
581 /* minimum AH key length */
582 SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN,	ah_keymin, CTLFLAG_RW, \
583 	&ipsec_ah_keymin,	0,	"");
584 
585 /* perfered old SA rather than new SA */
586 SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA,	prefered_oldsa, CTLFLAG_RW,\
587 	&key_prefered_oldsa,	0,	"");
588 #endif /* SYSCTL_INT */
589 
590 #define __LIST_CHAINED(elm) \
591 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
592 #define LIST_INSERT_TAIL(head, elm, type, field) \
593 do {\
594 	struct type *curelm = LIST_FIRST(head); \
595 	if (curelm == NULL) {\
596 		LIST_INSERT_HEAD(head, elm, field); \
597 	} else { \
598 		while (LIST_NEXT(curelm, field)) \
599 			curelm = LIST_NEXT(curelm, field);\
600 		LIST_INSERT_AFTER(curelm, elm, field);\
601 	}\
602 } while (0)
603 
604 #define KEY_CHKSASTATE(head, sav) \
605 /* do */ { \
606 	if ((head) != (sav)) {						\
607 		IPSECLOG(LOG_DEBUG,					\
608 		    "state mismatched (TREE=%d SA=%d)\n",		\
609 		    (head), (sav));					\
610 		continue;						\
611 	}								\
612 } /* while (0) */
613 
614 #define KEY_CHKSPDIR(head, sp) \
615 do { \
616 	if ((head) != (sp)) {						\
617 		IPSECLOG(LOG_DEBUG,					\
618 		    "direction mismatched (TREE=%d SP=%d), anyway continue.\n",\
619 		    (head), (sp));					\
620 	}								\
621 } while (0)
622 
623 /*
624  * set parameters into secasindex buffer.
625  * Must allocate secasindex buffer before calling this function.
626  */
627 static int
628 key_setsecasidx(int, int, int, const struct sockaddr *,
629     const struct sockaddr *, struct secasindex *);
630 
631 /* key statistics */
632 struct _keystat {
633 	u_long getspi_count; /* the avarage of count to try to get new SPI */
634 } keystat;
635 
636 static void
637 key_init_spidx_bymsghdr(struct secpolicyindex *, const struct sadb_msghdr *);
638 
639 static const struct sockaddr *
640 key_msghdr_get_sockaddr(const struct sadb_msghdr *mhp, int idx)
641 {
642 
643 	return PFKEY_ADDR_SADDR(mhp->ext[idx]);
644 }
645 
646 static void
647 key_fill_replymsg(struct mbuf *m, int seq)
648 {
649 	struct sadb_msg *msg;
650 
651 	KASSERT(m->m_len >= sizeof(*msg));
652 
653 	msg = mtod(m, struct sadb_msg *);
654 	msg->sadb_msg_errno = 0;
655 	msg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
656 	if (seq != 0)
657 		msg->sadb_msg_seq = seq;
658 }
659 
660 #if 0
661 static void key_freeso(struct socket *);
662 static void key_freesp_so(struct secpolicy **);
663 #endif
664 static struct secpolicy *key_getsp (const struct secpolicyindex *);
665 static struct secpolicy *key_getspbyid (u_int32_t);
666 static struct secpolicy *key_lookup_and_remove_sp(const struct secpolicyindex *, bool);
667 static struct secpolicy *key_lookupbyid_and_remove_sp(u_int32_t, bool);
668 static void key_destroy_sp(struct secpolicy *);
669 static struct mbuf *key_gather_mbuf (struct mbuf *,
670 	const struct sadb_msghdr *, int, int, ...);
671 static int key_api_spdadd(struct socket *, struct mbuf *,
672 	const struct sadb_msghdr *);
673 static u_int32_t key_getnewspid (void);
674 static int key_api_spddelete(struct socket *, struct mbuf *,
675 	const struct sadb_msghdr *);
676 static int key_api_spddelete2(struct socket *, struct mbuf *,
677 	const struct sadb_msghdr *);
678 static int key_api_spdget(struct socket *, struct mbuf *,
679 	const struct sadb_msghdr *);
680 static int key_api_spdflush(struct socket *, struct mbuf *,
681 	const struct sadb_msghdr *);
682 static int key_api_spddump(struct socket *, struct mbuf *,
683 	const struct sadb_msghdr *);
684 static struct mbuf * key_setspddump (int *errorp, pid_t);
685 static struct mbuf * key_setspddump_chain (int *errorp, int *lenp, pid_t pid);
686 static int key_api_nat_map(struct socket *, struct mbuf *,
687 	const struct sadb_msghdr *);
688 static struct mbuf *key_setdumpsp (struct secpolicy *,
689 	u_int8_t, u_int32_t, pid_t);
690 static u_int key_getspreqmsglen (const struct secpolicy *);
691 static int key_spdexpire (struct secpolicy *);
692 static struct secashead *key_newsah (const struct secasindex *);
693 static void key_unlink_sah(struct secashead *);
694 static void key_destroy_sah(struct secashead *);
695 static bool key_sah_has_sav(struct secashead *);
696 static void key_sah_ref(struct secashead *);
697 static void key_sah_unref(struct secashead *);
698 static void key_init_sav(struct secasvar *);
699 static void key_wait_sav(struct secasvar *);
700 static void key_destroy_sav(struct secasvar *);
701 static struct secasvar *key_newsav(struct mbuf *,
702 	const struct sadb_msghdr *, int *, const char*, int);
703 #define	KEY_NEWSAV(m, sadb, e)				\
704 	key_newsav(m, sadb, e, __func__, __LINE__)
705 static void key_delsav (struct secasvar *);
706 static struct secashead *key_getsah(const struct secasindex *, int);
707 static struct secashead *key_getsah_ref(const struct secasindex *, int);
708 static bool key_checkspidup(const struct secasindex *, u_int32_t);
709 static struct secasvar *key_getsavbyspi (struct secashead *, u_int32_t);
710 static int key_setsaval (struct secasvar *, struct mbuf *,
711 	const struct sadb_msghdr *);
712 static void key_freesaval(struct secasvar *);
713 static int key_init_xform(struct secasvar *);
714 static void key_clear_xform(struct secasvar *);
715 static struct mbuf *key_setdumpsa (struct secasvar *, u_int8_t,
716 	u_int8_t, u_int32_t, u_int32_t);
717 static struct mbuf *key_setsadbxport (u_int16_t, u_int16_t);
718 static struct mbuf *key_setsadbxtype (u_int16_t);
719 static struct mbuf *key_setsadbxfrag (u_int16_t);
720 static void key_porttosaddr (union sockaddr_union *, u_int16_t);
721 static int key_checksalen (const union sockaddr_union *);
722 static struct mbuf *key_setsadbmsg (u_int8_t, u_int16_t, u_int8_t,
723 	u_int32_t, pid_t, u_int16_t, int);
724 static struct mbuf *key_setsadbsa (struct secasvar *);
725 static struct mbuf *key_setsadbaddr(u_int16_t,
726 	const struct sockaddr *, u_int8_t, u_int16_t, int);
727 #if 0
728 static struct mbuf *key_setsadbident (u_int16_t, u_int16_t, void *,
729 	int, u_int64_t);
730 #endif
731 static struct mbuf *key_setsadbxsa2 (u_int8_t, u_int32_t, u_int16_t);
732 static struct mbuf *key_setsadbxpolicy (u_int16_t, u_int8_t,
733 	u_int32_t, int);
734 static void *key_newbuf (const void *, u_int);
735 #ifdef INET6
736 static int key_ismyaddr6 (const struct sockaddr_in6 *);
737 #endif
738 
739 static void sysctl_net_keyv2_setup(struct sysctllog **);
740 static void sysctl_net_key_compat_setup(struct sysctllog **);
741 
742 /* flags for key_saidx_match() */
743 #define CMP_HEAD	1	/* protocol, addresses. */
744 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
745 #define CMP_REQID	3	/* additionally HEAD, reaid. */
746 #define CMP_EXACTLY	4	/* all elements. */
747 static int key_saidx_match(const struct secasindex *,
748     const struct secasindex *, int);
749 
750 static int key_sockaddr_match(const struct sockaddr *,
751     const struct sockaddr *, int);
752 static int key_bb_match_withmask(const void *, const void *, u_int);
753 static u_int16_t key_satype2proto (u_int8_t);
754 static u_int8_t key_proto2satype (u_int16_t);
755 
756 static int key_spidx_match_exactly(const struct secpolicyindex *,
757     const struct secpolicyindex *);
758 static int key_spidx_match_withmask(const struct secpolicyindex *,
759     const struct secpolicyindex *);
760 
761 static int key_api_getspi(struct socket *, struct mbuf *,
762 	const struct sadb_msghdr *);
763 static u_int32_t key_do_getnewspi (const struct sadb_spirange *,
764 					const struct secasindex *);
765 static int key_handle_natt_info (struct secasvar *,
766 				     const struct sadb_msghdr *);
767 static int key_set_natt_ports (union sockaddr_union *,
768 			 	union sockaddr_union *,
769 				const struct sadb_msghdr *);
770 static int key_api_update(struct socket *, struct mbuf *,
771 	const struct sadb_msghdr *);
772 #ifdef IPSEC_DOSEQCHECK
773 static struct secasvar *key_getsavbyseq (struct secashead *, u_int32_t);
774 #endif
775 static int key_api_add(struct socket *, struct mbuf *,
776 	const struct sadb_msghdr *);
777 static int key_setident (struct secashead *, struct mbuf *,
778 	const struct sadb_msghdr *);
779 static struct mbuf *key_getmsgbuf_x1 (struct mbuf *,
780 	const struct sadb_msghdr *);
781 static int key_api_delete(struct socket *, struct mbuf *,
782 	const struct sadb_msghdr *);
783 static int key_api_get(struct socket *, struct mbuf *,
784 	const struct sadb_msghdr *);
785 
786 static void key_getcomb_setlifetime (struct sadb_comb *);
787 static struct mbuf *key_getcomb_esp(int);
788 static struct mbuf *key_getcomb_ah(int);
789 static struct mbuf *key_getcomb_ipcomp(int);
790 static struct mbuf *key_getprop(const struct secasindex *, int);
791 
792 static int key_acquire(const struct secasindex *, const struct secpolicy *,
793 	    int);
794 static int key_acquire_sendup_mbuf_later(struct mbuf *);
795 static void key_acquire_sendup_pending_mbuf(void);
796 #ifndef IPSEC_NONBLOCK_ACQUIRE
797 static struct secacq *key_newacq (const struct secasindex *);
798 static struct secacq *key_getacq (const struct secasindex *);
799 static struct secacq *key_getacqbyseq (u_int32_t);
800 #endif
801 #ifdef notyet
802 static struct secspacq *key_newspacq (const struct secpolicyindex *);
803 static struct secspacq *key_getspacq (const struct secpolicyindex *);
804 #endif
805 static int key_api_acquire(struct socket *, struct mbuf *,
806 	const struct sadb_msghdr *);
807 static int key_api_register(struct socket *, struct mbuf *,
808 	const struct sadb_msghdr *);
809 static int key_expire (struct secasvar *);
810 static int key_api_flush(struct socket *, struct mbuf *,
811 	const struct sadb_msghdr *);
812 static struct mbuf *key_setdump_chain (u_int8_t req_satype, int *errorp,
813 	int *lenp, pid_t pid);
814 static int key_api_dump(struct socket *, struct mbuf *,
815 	const struct sadb_msghdr *);
816 static int key_api_promisc(struct socket *, struct mbuf *,
817 	const struct sadb_msghdr *);
818 static int key_senderror (struct socket *, struct mbuf *, int);
819 static int key_validate_ext (const struct sadb_ext *, int);
820 static int key_align (struct mbuf *, struct sadb_msghdr *);
821 #if 0
822 static const char *key_getfqdn (void);
823 static const char *key_getuserfqdn (void);
824 #endif
825 static void key_sa_chgstate (struct secasvar *, u_int8_t);
826 
827 static struct mbuf *key_alloc_mbuf(int, int);
828 static struct mbuf *key_alloc_mbuf_simple(int, int);
829 
830 static void key_timehandler(void *);
831 static void key_timehandler_work(struct work *, void *);
832 static struct callout	key_timehandler_ch;
833 static struct workqueue	*key_timehandler_wq;
834 static struct work	key_timehandler_wk;
835 
836 static inline void
837     key_savlut_writer_insert_head(struct secasvar *sav);
838 static inline uint32_t
839     key_saidxhash(const struct secasindex *, u_long);
840 static inline uint32_t
841     key_savluthash(const struct sockaddr *,
842     uint32_t, uint32_t, u_long);
843 
844 /*
845  * Utilities for percpu counters for sadb_lifetime_allocations and
846  * sadb_lifetime_bytes.
847  */
848 #define LIFETIME_COUNTER_ALLOCATIONS	0
849 #define LIFETIME_COUNTER_BYTES		1
850 #define LIFETIME_COUNTER_SIZE		2
851 
852 typedef uint64_t lifetime_counters_t[LIFETIME_COUNTER_SIZE];
853 
854 static void
855 key_sum_lifetime_counters(void *p, void *arg, struct cpu_info *ci __unused)
856 {
857 	lifetime_counters_t *one = p;
858 	lifetime_counters_t *sum = arg;
859 
860 	(*sum)[LIFETIME_COUNTER_ALLOCATIONS] += (*one)[LIFETIME_COUNTER_ALLOCATIONS];
861 	(*sum)[LIFETIME_COUNTER_BYTES] += (*one)[LIFETIME_COUNTER_BYTES];
862 }
863 
864 u_int
865 key_sp_refcnt(const struct secpolicy *sp)
866 {
867 
868 	/* FIXME */
869 	return 0;
870 }
871 
872 static void
873 key_spd_pserialize_perform(void)
874 {
875 
876 	KASSERT(mutex_owned(&key_spd.lock));
877 
878 	while (key_spd.psz_performing)
879 		cv_wait(&key_spd.cv_psz, &key_spd.lock);
880 	key_spd.psz_performing = true;
881 	mutex_exit(&key_spd.lock);
882 
883 	pserialize_perform(key_spd.psz);
884 
885 	mutex_enter(&key_spd.lock);
886 	key_spd.psz_performing = false;
887 	cv_broadcast(&key_spd.cv_psz);
888 }
889 
890 /*
891  * Remove the sp from the key_spd.splist and wait for references to the sp
892  * to be released. key_spd.lock must be held.
893  */
894 static void
895 key_unlink_sp(struct secpolicy *sp)
896 {
897 
898 	KASSERT(mutex_owned(&key_spd.lock));
899 
900 	sp->state = IPSEC_SPSTATE_DEAD;
901 	SPLIST_WRITER_REMOVE(sp);
902 
903 	/* Invalidate all cached SPD pointers in the PCBs. */
904 	ipsec_invalpcbcacheall();
905 
906 	KDASSERT(mutex_ownable(softnet_lock));
907 	key_spd_pserialize_perform();
908 
909 	localcount_drain(&sp->localcount, &key_spd.cv_lc, &key_spd.lock);
910 }
911 
912 /*
913  * Return 0 when there are known to be no SP's for the specified
914  * direction.  Otherwise return 1.  This is used by IPsec code
915  * to optimize performance.
916  */
917 int
918 key_havesp(u_int dir)
919 {
920 	return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ?
921 		!SPLIST_READER_EMPTY(dir) : 1);
922 }
923 
924 /* %%% IPsec policy management */
925 /*
926  * allocating a SP for OUTBOUND or INBOUND packet.
927  * Must call key_freesp() later.
928  * OUT:	NULL:	not found
929  *	others:	found and return the pointer.
930  */
931 struct secpolicy *
932 key_lookup_sp_byspidx(const struct secpolicyindex *spidx,
933     u_int dir, const char* where, int tag)
934 {
935 	struct secpolicy *sp;
936 	int s;
937 
938 	KASSERT(spidx != NULL);
939 	KASSERTMSG(IPSEC_DIR_IS_INOROUT(dir), "invalid direction %u", dir);
940 
941 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
942 
943 	/* get a SP entry */
944 	if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
945 		kdebug_secpolicyindex("objects", spidx);
946 	}
947 
948 	s = pserialize_read_enter();
949 	SPLIST_READER_FOREACH(sp, dir) {
950 		if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
951 			kdebug_secpolicyindex("in SPD", &sp->spidx);
952 		}
953 
954 		if (sp->state == IPSEC_SPSTATE_DEAD)
955 			continue;
956 		if (key_spidx_match_withmask(&sp->spidx, spidx))
957 			goto found;
958 	}
959 	sp = NULL;
960 found:
961 	if (sp) {
962 		/* sanity check */
963 		KEY_CHKSPDIR(sp->spidx.dir, dir);
964 
965 		/* found a SPD entry */
966 		sp->lastused = time_uptime;
967 		key_sp_ref(sp, where, tag);
968 	}
969 	pserialize_read_exit(s);
970 
971 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
972 	    "DP return SP:%p (ID=%u) refcnt %u\n",
973 	    sp, sp ? sp->id : 0, key_sp_refcnt(sp));
974 	return sp;
975 }
976 
977 /*
978  * return a policy that matches this particular inbound packet.
979  * XXX slow
980  */
981 struct secpolicy *
982 key_gettunnel(const struct sockaddr *osrc,
983 	      const struct sockaddr *odst,
984 	      const struct sockaddr *isrc,
985 	      const struct sockaddr *idst,
986 	      const char* where, int tag)
987 {
988 	struct secpolicy *sp;
989 	const int dir = IPSEC_DIR_INBOUND;
990 	int s;
991 	struct ipsecrequest *r1, *r2, *p;
992 	struct secpolicyindex spidx;
993 
994 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
995 
996 	if (isrc->sa_family != idst->sa_family) {
997 		IPSECLOG(LOG_ERR,
998 		    "address family mismatched src %u, dst %u.\n",
999 		    isrc->sa_family, idst->sa_family);
1000 		sp = NULL;
1001 		goto done;
1002 	}
1003 
1004 	s = pserialize_read_enter();
1005 	SPLIST_READER_FOREACH(sp, dir) {
1006 		if (sp->state == IPSEC_SPSTATE_DEAD)
1007 			continue;
1008 
1009 		r1 = r2 = NULL;
1010 		for (p = sp->req; p; p = p->next) {
1011 			if (p->saidx.mode != IPSEC_MODE_TUNNEL)
1012 				continue;
1013 
1014 			r1 = r2;
1015 			r2 = p;
1016 
1017 			if (!r1) {
1018 				/* here we look at address matches only */
1019 				spidx = sp->spidx;
1020 				if (isrc->sa_len > sizeof(spidx.src) ||
1021 				    idst->sa_len > sizeof(spidx.dst))
1022 					continue;
1023 				memcpy(&spidx.src, isrc, isrc->sa_len);
1024 				memcpy(&spidx.dst, idst, idst->sa_len);
1025 				if (!key_spidx_match_withmask(&sp->spidx, &spidx))
1026 					continue;
1027 			} else {
1028 				if (!key_sockaddr_match(&r1->saidx.src.sa, isrc, PORT_NONE) ||
1029 				    !key_sockaddr_match(&r1->saidx.dst.sa, idst, PORT_NONE))
1030 					continue;
1031 			}
1032 
1033 			if (!key_sockaddr_match(&r2->saidx.src.sa, osrc, PORT_NONE) ||
1034 			    !key_sockaddr_match(&r2->saidx.dst.sa, odst, PORT_NONE))
1035 				continue;
1036 
1037 			goto found;
1038 		}
1039 	}
1040 	sp = NULL;
1041 found:
1042 	if (sp) {
1043 		sp->lastused = time_uptime;
1044 		key_sp_ref(sp, where, tag);
1045 	}
1046 	pserialize_read_exit(s);
1047 done:
1048 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1049 	    "DP return SP:%p (ID=%u) refcnt %u\n",
1050 	    sp, sp ? sp->id : 0, key_sp_refcnt(sp));
1051 	return sp;
1052 }
1053 
1054 /*
1055  * allocating an SA entry for an *OUTBOUND* packet.
1056  * checking each request entries in SP, and acquire an SA if need.
1057  * OUT:	0: there are valid requests.
1058  *	ENOENT: policy may be valid, but SA with REQUIRE is on acquiring.
1059  */
1060 int
1061 key_checkrequest(const struct ipsecrequest *isr, const struct secasindex *saidx,
1062     struct secasvar **ret)
1063 {
1064 	u_int level;
1065 	int error;
1066 	struct secasvar *sav;
1067 
1068 	KASSERT(isr != NULL);
1069 	KASSERTMSG(saidx->mode == IPSEC_MODE_TRANSPORT ||
1070 	    saidx->mode == IPSEC_MODE_TUNNEL,
1071 	    "unexpected policy %u", saidx->mode);
1072 
1073 	/* get current level */
1074 	level = ipsec_get_reqlevel(isr);
1075 
1076 	/*
1077 	 * XXX guard against protocol callbacks from the crypto
1078 	 * thread as they reference ipsecrequest.sav which we
1079 	 * temporarily null out below.  Need to rethink how we
1080 	 * handle bundled SA's in the callback thread.
1081 	 */
1082 
1083 	sav = key_lookup_sa_bysaidx(saidx);
1084 	if (sav != NULL) {
1085 		*ret = sav;
1086 		return 0;
1087 	}
1088 
1089 	/* there is no SA */
1090 	error = key_acquire(saidx, isr->sp, M_NOWAIT);
1091 	if (error != 0) {
1092 		/* XXX What should I do ? */
1093 		IPSECLOG(LOG_DEBUG, "error %d returned from key_acquire.\n",
1094 		    error);
1095 		return error;
1096 	}
1097 
1098 	if (level != IPSEC_LEVEL_REQUIRE) {
1099 		/* XXX sigh, the interface to this routine is botched */
1100 		*ret = NULL;
1101 		return 0;
1102 	} else {
1103 		return ENOENT;
1104 	}
1105 }
1106 
1107 /*
1108  * looking up a SA for policy entry from SAD.
1109  * NOTE: searching SAD of aliving state.
1110  * OUT:	NULL:	not found.
1111  *	others:	found and return the pointer.
1112  */
1113 struct secasvar *
1114 key_lookup_sa_bysaidx(const struct secasindex *saidx)
1115 {
1116 	struct secashead *sah;
1117 	struct secasvar *sav = NULL;
1118 	u_int stateidx, state;
1119 	const u_int *saorder_state_valid;
1120 	int arraysize;
1121 	int s;
1122 
1123 	s = pserialize_read_enter();
1124 	sah = key_getsah(saidx, CMP_MODE_REQID);
1125 	if (sah == NULL)
1126 		goto out;
1127 
1128 	/*
1129 	 * search a valid state list for outbound packet.
1130 	 * This search order is important.
1131 	 */
1132 	if (key_prefered_oldsa) {
1133 		saorder_state_valid = saorder_state_valid_prefer_old;
1134 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_old);
1135 	} else {
1136 		saorder_state_valid = saorder_state_valid_prefer_new;
1137 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_new);
1138 	}
1139 
1140 	/* search valid state */
1141 	for (stateidx = 0;
1142 	     stateidx < arraysize;
1143 	     stateidx++) {
1144 
1145 		state = saorder_state_valid[stateidx];
1146 
1147 		if (key_prefered_oldsa)
1148 			sav = SAVLIST_READER_FIRST(sah, state);
1149 		else {
1150 			/* XXX need O(1) lookup */
1151 			struct secasvar *last = NULL;
1152 
1153 			SAVLIST_READER_FOREACH(sav, sah, state)
1154 				last = sav;
1155 			sav = last;
1156 		}
1157 		if (sav != NULL) {
1158 			KEY_SA_REF(sav);
1159 			KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1160 			    "DP cause refcnt++:%d SA:%p\n",
1161 			    key_sa_refcnt(sav), sav);
1162 			break;
1163 		}
1164 	}
1165 out:
1166 	pserialize_read_exit(s);
1167 
1168 	return sav;
1169 }
1170 
1171 #if 0
1172 static void
1173 key_sendup_message_delete(struct secasvar *sav)
1174 {
1175 	struct mbuf *m, *result = 0;
1176 	uint8_t satype;
1177 
1178 	satype = key_proto2satype(sav->sah->saidx.proto);
1179 	if (satype == 0)
1180 		goto msgfail;
1181 
1182 	m = key_setsadbmsg(SADB_DELETE, 0, satype, 0, 0, key_sa_refcnt(sav) - 1);
1183 	if (m == NULL)
1184 		goto msgfail;
1185 	result = m;
1186 
1187 	/* set sadb_address for saidx's. */
1188 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa,
1189 	    _BITS(sav->sah->saidx.src.sa.sa_len), IPSEC_ULPROTO_ANY);
1190 	if (m == NULL)
1191 		goto msgfail;
1192 	m_cat(result, m);
1193 
1194 	/* set sadb_address for saidx's. */
1195 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.src.sa,
1196 	    _BITS(sav->sah->saidx.src.sa.sa_len), IPSEC_ULPROTO_ANY);
1197 	if (m == NULL)
1198 		goto msgfail;
1199 	m_cat(result, m);
1200 
1201 	/* create SA extension */
1202 	m = key_setsadbsa(sav);
1203 	if (m == NULL)
1204 		goto msgfail;
1205 	m_cat(result, m);
1206 
1207 	if (result->m_len < sizeof(struct sadb_msg)) {
1208 		result = m_pullup(result, sizeof(struct sadb_msg));
1209 		if (result == NULL)
1210 			goto msgfail;
1211 	}
1212 
1213 	result->m_pkthdr.len = 0;
1214 	for (m = result; m; m = m->m_next)
1215 		result->m_pkthdr.len += m->m_len;
1216 	mtod(result, struct sadb_msg *)->sadb_msg_len =
1217 	    PFKEY_UNIT64(result->m_pkthdr.len);
1218 
1219 	key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
1220 	result = NULL;
1221 msgfail:
1222 	if (result)
1223 		m_freem(result);
1224 }
1225 #endif
1226 
1227 /*
1228  * allocating a usable SA entry for a *INBOUND* packet.
1229  * Must call key_freesav() later.
1230  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
1231  *	NULL:		not found, or error occurred.
1232  *
1233  * In the comparison, no source address is used--for RFC2401 conformance.
1234  * To quote, from section 4.1:
1235  *	A security association is uniquely identified by a triple consisting
1236  *	of a Security Parameter Index (SPI), an IP Destination Address, and a
1237  *	security protocol (AH or ESP) identifier.
1238  * Note that, however, we do need to keep source address in IPsec SA.
1239  * IKE specification and PF_KEY specification do assume that we
1240  * keep source address in IPsec SA.  We see a tricky situation here.
1241  *
1242  * sport and dport are used for NAT-T. network order is always used.
1243  */
1244 struct secasvar *
1245 key_lookup_sa(
1246 	const union sockaddr_union *dst,
1247 	u_int proto,
1248 	u_int32_t spi,
1249 	u_int16_t sport,
1250 	u_int16_t dport,
1251 	const char* where, int tag)
1252 {
1253 	struct secasvar *sav;
1254 	int chkport;
1255 	int s;
1256 
1257 	int must_check_spi = 1;
1258 	int must_check_alg = 0;
1259 	u_int16_t cpi = 0;
1260 	u_int8_t algo = 0;
1261 	uint32_t hash_key = spi;
1262 
1263 	if ((sport != 0) && (dport != 0))
1264 		chkport = PORT_STRICT;
1265 	else
1266 		chkport = PORT_NONE;
1267 
1268 	KASSERT(dst != NULL);
1269 
1270 	/*
1271 	 * XXX IPCOMP case
1272 	 * We use cpi to define spi here. In the case where cpi <=
1273 	 * IPCOMP_CPI_NEGOTIATE_MIN, cpi just define the algorithm used, not
1274 	 * the real spi. In this case, don't check the spi but check the
1275 	 * algorithm
1276 	 */
1277 
1278 	if (proto == IPPROTO_IPCOMP) {
1279 		u_int32_t tmp;
1280 		tmp = ntohl(spi);
1281 		cpi = (u_int16_t) tmp;
1282 		if (cpi < IPCOMP_CPI_NEGOTIATE_MIN) {
1283 			algo = (u_int8_t) cpi;
1284 			hash_key = algo;
1285 			must_check_spi = 0;
1286 			must_check_alg = 1;
1287 		}
1288 	}
1289 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1290 	    "DP from %s:%u check_spi=%d, check_alg=%d\n",
1291 	    where, tag, must_check_spi, must_check_alg);
1292 
1293 
1294 	/*
1295 	 * searching SAD.
1296 	 * XXX: to be checked internal IP header somewhere.  Also when
1297 	 * IPsec tunnel packet is received.  But ESP tunnel mode is
1298 	 * encrypted so we can't check internal IP header.
1299 	 */
1300 	s = pserialize_read_enter();
1301 	SAVLUT_READER_FOREACH(sav, &dst->sa, proto, hash_key) {
1302 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1303 		    "try match spi %#x, %#x\n",
1304 		    ntohl(spi), ntohl(sav->spi));
1305 
1306 		/* do not return entries w/ unusable state */
1307 		if (!SADB_SASTATE_USABLE_P(sav)) {
1308 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1309 			    "bad state %d\n", sav->state);
1310 			continue;
1311 		}
1312 		if (proto != sav->sah->saidx.proto) {
1313 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1314 			    "proto fail %d != %d\n",
1315 			    proto, sav->sah->saidx.proto);
1316 			continue;
1317 		}
1318 		if (must_check_spi && spi != sav->spi) {
1319 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1320 			    "spi fail %#x != %#x\n",
1321 			    ntohl(spi), ntohl(sav->spi));
1322 			continue;
1323 		}
1324 		/* XXX only on the ipcomp case */
1325 		if (must_check_alg && algo != sav->alg_comp) {
1326 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1327 			    "algo fail %d != %d\n",
1328 			    algo, sav->alg_comp);
1329 			continue;
1330 		}
1331 
1332 #if 0	/* don't check src */
1333 	/* Fix port in src->sa */
1334 
1335 		/* check src address */
1336 		if (!key_sockaddr_match(&src->sa, &sav->sah->saidx.src.sa, PORT_NONE))
1337 			continue;
1338 #endif
1339 		/* fix port of dst address XXX*/
1340 		key_porttosaddr(__UNCONST(dst), dport);
1341 		/* check dst address */
1342 		if (!key_sockaddr_match(&dst->sa, &sav->sah->saidx.dst.sa, chkport))
1343 			continue;
1344 		key_sa_ref(sav, where, tag);
1345 		goto done;
1346 	}
1347 	sav = NULL;
1348 done:
1349 	pserialize_read_exit(s);
1350 
1351 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1352 	    "DP return SA:%p; refcnt %u\n", sav, key_sa_refcnt(sav));
1353 	return sav;
1354 }
1355 
1356 static void
1357 key_validate_savlist(const struct secashead *sah, const u_int state)
1358 {
1359 #ifdef DEBUG
1360 	struct secasvar *sav, *next;
1361 	int s;
1362 
1363 	/*
1364 	 * The list should be sorted by lft_c->sadb_lifetime_addtime
1365 	 * in ascending order.
1366 	 */
1367 	s = pserialize_read_enter();
1368 	SAVLIST_READER_FOREACH(sav, sah, state) {
1369 		next = SAVLIST_READER_NEXT(sav);
1370 		if (next != NULL &&
1371 		    sav->lft_c != NULL && next->lft_c != NULL) {
1372 			KDASSERTMSG(sav->lft_c->sadb_lifetime_addtime <=
1373 			    next->lft_c->sadb_lifetime_addtime,
1374 			    "savlist is not sorted: sah=%p, state=%d, "
1375 			    "sav=%" PRIu64 ", next=%" PRIu64, sah, state,
1376 			    sav->lft_c->sadb_lifetime_addtime,
1377 			    next->lft_c->sadb_lifetime_addtime);
1378 		}
1379 	}
1380 	pserialize_read_exit(s);
1381 #endif
1382 }
1383 
1384 void
1385 key_init_sp(struct secpolicy *sp)
1386 {
1387 
1388 	ASSERT_SLEEPABLE();
1389 
1390 	sp->state = IPSEC_SPSTATE_ALIVE;
1391 	if (sp->policy == IPSEC_POLICY_IPSEC)
1392 		KASSERT(sp->req != NULL);
1393 	localcount_init(&sp->localcount);
1394 	SPLIST_ENTRY_INIT(sp);
1395 }
1396 
1397 /*
1398  * Must be called in a pserialize read section. A held SP
1399  * must be released by key_sp_unref after use.
1400  */
1401 void
1402 key_sp_ref(struct secpolicy *sp, const char* where, int tag)
1403 {
1404 
1405 	localcount_acquire(&sp->localcount);
1406 
1407 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1408 	    "DP SP:%p (ID=%u) from %s:%u; refcnt++ now %u\n",
1409 	    sp, sp->id, where, tag, key_sp_refcnt(sp));
1410 }
1411 
1412 /*
1413  * Must be called without holding key_spd.lock because the lock
1414  * would be held in localcount_release.
1415  */
1416 void
1417 key_sp_unref(struct secpolicy *sp, const char* where, int tag)
1418 {
1419 
1420 	KDASSERT(mutex_ownable(&key_spd.lock));
1421 
1422 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1423 	    "DP SP:%p (ID=%u) from %s:%u; refcnt-- now %u\n",
1424 	    sp, sp->id, where, tag, key_sp_refcnt(sp));
1425 
1426 	localcount_release(&sp->localcount, &key_spd.cv_lc, &key_spd.lock);
1427 }
1428 
1429 static void
1430 key_init_sav(struct secasvar *sav)
1431 {
1432 
1433 	ASSERT_SLEEPABLE();
1434 
1435 	localcount_init(&sav->localcount);
1436 	SAVLIST_ENTRY_INIT(sav);
1437 	SAVLUT_ENTRY_INIT(sav);
1438 }
1439 
1440 u_int
1441 key_sa_refcnt(const struct secasvar *sav)
1442 {
1443 
1444 	/* FIXME */
1445 	return 0;
1446 }
1447 
1448 void
1449 key_sa_ref(struct secasvar *sav, const char* where, int tag)
1450 {
1451 
1452 	localcount_acquire(&sav->localcount);
1453 
1454 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1455 	    "DP cause refcnt++: SA:%p from %s:%u\n",
1456 	    sav, where, tag);
1457 }
1458 
1459 void
1460 key_sa_unref(struct secasvar *sav, const char* where, int tag)
1461 {
1462 
1463 	KDASSERT(mutex_ownable(&key_sad.lock));
1464 
1465 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1466 	    "DP cause refcnt--: SA:%p from %s:%u\n",
1467 	    sav, where, tag);
1468 
1469 	localcount_release(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
1470 }
1471 
1472 #if 0
1473 /*
1474  * Must be called after calling key_lookup_sp*().
1475  * For the packet with socket.
1476  */
1477 static void
1478 key_freeso(struct socket *so)
1479 {
1480 	/* sanity check */
1481 	KASSERT(so != NULL);
1482 
1483 	switch (so->so_proto->pr_domain->dom_family) {
1484 #ifdef INET
1485 	case PF_INET:
1486 	    {
1487 		struct inpcb *pcb = sotoinpcb(so);
1488 
1489 		/* Does it have a PCB ? */
1490 		if (pcb == NULL)
1491 			return;
1492 
1493 		struct inpcbpolicy *sp = pcb->inp_sp;
1494 		key_freesp_so(&sp->sp_in);
1495 		key_freesp_so(&sp->sp_out);
1496 	    }
1497 		break;
1498 #endif
1499 #ifdef INET6
1500 	case PF_INET6:
1501 	    {
1502 #ifdef HAVE_NRL_INPCB
1503 		struct inpcb *pcb  = sotoinpcb(so);
1504 		struct inpcbpolicy *sp = pcb->inp_sp;
1505 
1506 		/* Does it have a PCB ? */
1507 		if (pcb == NULL)
1508 			return;
1509 		key_freesp_so(&sp->sp_in);
1510 		key_freesp_so(&sp->sp_out);
1511 #else
1512 		struct in6pcb *pcb  = sotoin6pcb(so);
1513 
1514 		/* Does it have a PCB ? */
1515 		if (pcb == NULL)
1516 			return;
1517 		key_freesp_so(&pcb->in6p_sp->sp_in);
1518 		key_freesp_so(&pcb->in6p_sp->sp_out);
1519 #endif
1520 	    }
1521 		break;
1522 #endif /* INET6 */
1523 	default:
1524 		IPSECLOG(LOG_DEBUG, "unknown address family=%d.\n",
1525 		    so->so_proto->pr_domain->dom_family);
1526 		return;
1527 	}
1528 }
1529 
1530 static void
1531 key_freesp_so(struct secpolicy **sp)
1532 {
1533 
1534 	KASSERT(sp != NULL);
1535 	KASSERT(*sp != NULL);
1536 
1537 	if ((*sp)->policy == IPSEC_POLICY_ENTRUST ||
1538 	    (*sp)->policy == IPSEC_POLICY_BYPASS)
1539 		return;
1540 
1541 	KASSERTMSG((*sp)->policy == IPSEC_POLICY_IPSEC,
1542 	    "invalid policy %u", (*sp)->policy);
1543 	KEY_SP_UNREF(&sp);
1544 }
1545 #endif
1546 
1547 static void
1548 key_sad_pserialize_perform(void)
1549 {
1550 
1551 	KASSERT(mutex_owned(&key_sad.lock));
1552 
1553 	while (key_sad.psz_performing)
1554 		cv_wait(&key_sad.cv_psz, &key_sad.lock);
1555 	key_sad.psz_performing = true;
1556 	mutex_exit(&key_sad.lock);
1557 
1558 	pserialize_perform(key_sad.psz);
1559 
1560 	mutex_enter(&key_sad.lock);
1561 	key_sad.psz_performing = false;
1562 	cv_broadcast(&key_sad.cv_psz);
1563 }
1564 
1565 /*
1566  * Remove the sav from the savlist of its sah and wait for references to the sav
1567  * to be released. key_sad.lock must be held.
1568  */
1569 static void
1570 key_unlink_sav(struct secasvar *sav)
1571 {
1572 
1573 	KASSERT(mutex_owned(&key_sad.lock));
1574 
1575 	SAVLIST_WRITER_REMOVE(sav);
1576 	SAVLUT_WRITER_REMOVE(sav);
1577 
1578 	KDASSERT(mutex_ownable(softnet_lock));
1579 	key_sad_pserialize_perform();
1580 
1581 	localcount_drain(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
1582 }
1583 
1584 /*
1585  * Destroy an sav where the sav must be unlinked from an sah
1586  * by say key_unlink_sav.
1587  */
1588 static void
1589 key_destroy_sav(struct secasvar *sav)
1590 {
1591 
1592 	ASSERT_SLEEPABLE();
1593 
1594 	localcount_fini(&sav->localcount);
1595 	SAVLIST_ENTRY_DESTROY(sav);
1596 
1597 	key_delsav(sav);
1598 }
1599 
1600 /*
1601  * Wait for references of a passed sav to go away.
1602  */
1603 static void
1604 key_wait_sav(struct secasvar *sav)
1605 {
1606 
1607 	ASSERT_SLEEPABLE();
1608 
1609 	mutex_enter(&key_sad.lock);
1610 	KASSERT(sav->state == SADB_SASTATE_DEAD);
1611 	KDASSERT(mutex_ownable(softnet_lock));
1612 	key_sad_pserialize_perform();
1613 	localcount_drain(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
1614 	mutex_exit(&key_sad.lock);
1615 }
1616 
1617 /* %%% SPD management */
1618 /*
1619  * free security policy entry.
1620  */
1621 static void
1622 key_destroy_sp(struct secpolicy *sp)
1623 {
1624 
1625 	SPLIST_ENTRY_DESTROY(sp);
1626 	localcount_fini(&sp->localcount);
1627 
1628 	key_free_sp(sp);
1629 
1630 	key_update_used();
1631 }
1632 
1633 void
1634 key_free_sp(struct secpolicy *sp)
1635 {
1636 	struct ipsecrequest *isr = sp->req, *nextisr;
1637 
1638 	while (isr != NULL) {
1639 		nextisr = isr->next;
1640 		kmem_free(isr, sizeof(*isr));
1641 		isr = nextisr;
1642 	}
1643 
1644 	kmem_free(sp, sizeof(*sp));
1645 }
1646 
1647 void
1648 key_socksplist_add(struct secpolicy *sp)
1649 {
1650 
1651 	mutex_enter(&key_spd.lock);
1652 	PSLIST_WRITER_INSERT_HEAD(&key_spd.socksplist, sp, pslist_entry);
1653 	mutex_exit(&key_spd.lock);
1654 
1655 	key_update_used();
1656 }
1657 
1658 /*
1659  * search SPD
1660  * OUT:	NULL	: not found
1661  *	others	: found, pointer to a SP.
1662  */
1663 static struct secpolicy *
1664 key_getsp(const struct secpolicyindex *spidx)
1665 {
1666 	struct secpolicy *sp;
1667 	int s;
1668 
1669 	KASSERT(spidx != NULL);
1670 
1671 	s = pserialize_read_enter();
1672 	SPLIST_READER_FOREACH(sp, spidx->dir) {
1673 		if (sp->state == IPSEC_SPSTATE_DEAD)
1674 			continue;
1675 		if (key_spidx_match_exactly(spidx, &sp->spidx)) {
1676 			KEY_SP_REF(sp);
1677 			pserialize_read_exit(s);
1678 			return sp;
1679 		}
1680 	}
1681 	pserialize_read_exit(s);
1682 
1683 	return NULL;
1684 }
1685 
1686 /*
1687  * search SPD and remove found SP
1688  * OUT:	NULL	: not found
1689  *	others	: found, pointer to a SP.
1690  */
1691 static struct secpolicy *
1692 key_lookup_and_remove_sp(const struct secpolicyindex *spidx, bool from_kernel)
1693 {
1694 	struct secpolicy *sp = NULL;
1695 
1696 	mutex_enter(&key_spd.lock);
1697 	SPLIST_WRITER_FOREACH(sp, spidx->dir) {
1698 		KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
1699 		/*
1700 		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
1701 		 * removed by userland programs.
1702 		 */
1703 		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
1704 			continue;
1705 		if (key_spidx_match_exactly(spidx, &sp->spidx)) {
1706 			key_unlink_sp(sp);
1707 			goto out;
1708 		}
1709 	}
1710 	sp = NULL;
1711 out:
1712 	mutex_exit(&key_spd.lock);
1713 
1714 	return sp;
1715 }
1716 
1717 /*
1718  * get SP by index.
1719  * OUT:	NULL	: not found
1720  *	others	: found, pointer to a SP.
1721  */
1722 static struct secpolicy *
1723 key_getspbyid(u_int32_t id)
1724 {
1725 	struct secpolicy *sp;
1726 	int s;
1727 
1728 	s = pserialize_read_enter();
1729 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_INBOUND) {
1730 		if (sp->state == IPSEC_SPSTATE_DEAD)
1731 			continue;
1732 		if (sp->id == id) {
1733 			KEY_SP_REF(sp);
1734 			goto out;
1735 		}
1736 	}
1737 
1738 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_OUTBOUND) {
1739 		if (sp->state == IPSEC_SPSTATE_DEAD)
1740 			continue;
1741 		if (sp->id == id) {
1742 			KEY_SP_REF(sp);
1743 			goto out;
1744 		}
1745 	}
1746 out:
1747 	pserialize_read_exit(s);
1748 	return sp;
1749 }
1750 
1751 /*
1752  * get SP by index, remove and return it.
1753  * OUT:	NULL	: not found
1754  *	others	: found, pointer to a SP.
1755  */
1756 static struct secpolicy *
1757 key_lookupbyid_and_remove_sp(u_int32_t id, bool from_kernel)
1758 {
1759 	struct secpolicy *sp;
1760 
1761 	mutex_enter(&key_spd.lock);
1762 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_INBOUND) {
1763 		KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
1764 		/*
1765 		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
1766 		 * removed by userland programs.
1767 		 */
1768 		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
1769 			continue;
1770 		if (sp->id == id)
1771 			goto out;
1772 	}
1773 
1774 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_OUTBOUND) {
1775 		KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
1776 		/*
1777 		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
1778 		 * removed by userland programs.
1779 		 */
1780 		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
1781 			continue;
1782 		if (sp->id == id)
1783 			goto out;
1784 	}
1785 out:
1786 	if (sp != NULL)
1787 		key_unlink_sp(sp);
1788 	mutex_exit(&key_spd.lock);
1789 	return sp;
1790 }
1791 
1792 struct secpolicy *
1793 key_newsp(const char* where, int tag)
1794 {
1795 	struct secpolicy *newsp = NULL;
1796 
1797 	newsp = kmem_zalloc(sizeof(struct secpolicy), KM_SLEEP);
1798 
1799 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1800 	    "DP from %s:%u return SP:%p\n", where, tag, newsp);
1801 	return newsp;
1802 }
1803 
1804 /*
1805  * create secpolicy structure from sadb_x_policy structure.
1806  * NOTE: `state', `secpolicyindex' in secpolicy structure are not set,
1807  * so must be set properly later.
1808  */
1809 static struct secpolicy *
1810 _key_msg2sp(const struct sadb_x_policy *xpl0, size_t len, int *error,
1811     bool from_kernel)
1812 {
1813 	struct secpolicy *newsp;
1814 
1815 	KASSERT(!cpu_softintr_p());
1816 	KASSERT(xpl0 != NULL);
1817 	KASSERT(len >= sizeof(*xpl0));
1818 
1819 	if (len != PFKEY_EXTLEN(xpl0)) {
1820 		IPSECLOG(LOG_DEBUG, "Invalid msg length.\n");
1821 		*error = EINVAL;
1822 		return NULL;
1823 	}
1824 
1825 	newsp = KEY_NEWSP();
1826 	if (newsp == NULL) {
1827 		*error = ENOBUFS;
1828 		return NULL;
1829 	}
1830 
1831 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1832 	newsp->policy = xpl0->sadb_x_policy_type;
1833 
1834 	/* check policy */
1835 	switch (xpl0->sadb_x_policy_type) {
1836 	case IPSEC_POLICY_DISCARD:
1837 	case IPSEC_POLICY_NONE:
1838 	case IPSEC_POLICY_ENTRUST:
1839 	case IPSEC_POLICY_BYPASS:
1840 		newsp->req = NULL;
1841 		*error = 0;
1842 		return newsp;
1843 
1844 	case IPSEC_POLICY_IPSEC:
1845 		/* Continued */
1846 		break;
1847 	default:
1848 		IPSECLOG(LOG_DEBUG, "invalid policy type.\n");
1849 		key_free_sp(newsp);
1850 		*error = EINVAL;
1851 		return NULL;
1852 	}
1853 
1854 	/* IPSEC_POLICY_IPSEC */
1855     {
1856 	int tlen;
1857 	const struct sadb_x_ipsecrequest *xisr;
1858 	uint16_t xisr_reqid;
1859 	struct ipsecrequest **p_isr = &newsp->req;
1860 
1861 	/* validity check */
1862 	if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1863 		IPSECLOG(LOG_DEBUG, "Invalid msg length.\n");
1864 		*error = EINVAL;
1865 		goto free_exit;
1866 	}
1867 
1868 	tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1869 	xisr = (const struct sadb_x_ipsecrequest *)(xpl0 + 1);
1870 
1871 	while (tlen > 0) {
1872 		/* length check */
1873 		if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr)) {
1874 			IPSECLOG(LOG_DEBUG, "invalid ipsecrequest length.\n");
1875 			*error = EINVAL;
1876 			goto free_exit;
1877 		}
1878 
1879 		/* allocate request buffer */
1880 		*p_isr = kmem_zalloc(sizeof(**p_isr), KM_SLEEP);
1881 
1882 		/* set values */
1883 		(*p_isr)->next = NULL;
1884 
1885 		switch (xisr->sadb_x_ipsecrequest_proto) {
1886 		case IPPROTO_ESP:
1887 		case IPPROTO_AH:
1888 		case IPPROTO_IPCOMP:
1889 			break;
1890 		default:
1891 			IPSECLOG(LOG_DEBUG, "invalid proto type=%u\n",
1892 			    xisr->sadb_x_ipsecrequest_proto);
1893 			*error = EPROTONOSUPPORT;
1894 			goto free_exit;
1895 		}
1896 		(*p_isr)->saidx.proto = xisr->sadb_x_ipsecrequest_proto;
1897 
1898 		switch (xisr->sadb_x_ipsecrequest_mode) {
1899 		case IPSEC_MODE_TRANSPORT:
1900 		case IPSEC_MODE_TUNNEL:
1901 			break;
1902 		case IPSEC_MODE_ANY:
1903 		default:
1904 			IPSECLOG(LOG_DEBUG, "invalid mode=%u\n",
1905 			    xisr->sadb_x_ipsecrequest_mode);
1906 			*error = EINVAL;
1907 			goto free_exit;
1908 		}
1909 		(*p_isr)->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1910 
1911 		switch (xisr->sadb_x_ipsecrequest_level) {
1912 		case IPSEC_LEVEL_DEFAULT:
1913 		case IPSEC_LEVEL_USE:
1914 		case IPSEC_LEVEL_REQUIRE:
1915 			break;
1916 		case IPSEC_LEVEL_UNIQUE:
1917 			xisr_reqid = xisr->sadb_x_ipsecrequest_reqid;
1918 			/* validity check */
1919 			/*
1920 			 * case 1) from_kernel == false
1921 			 * That means the request comes from userland.
1922 			 * If range violation of reqid, kernel will
1923 			 * update it, don't refuse it.
1924 			 *
1925 			 * case 2) from_kernel == true
1926 			 * That means the request comes from kernel
1927 			 * (e.g. ipsec(4) I/F).
1928 			 * Use thre requested reqid to avoid inconsistency
1929 			 * between kernel's reqid and the reqid in pf_key
1930 			 * message sent to userland. The pf_key message is
1931 			 * built by diverting request mbuf.
1932 			 */
1933 			if (!from_kernel &&
1934 			    xisr_reqid > IPSEC_MANUAL_REQID_MAX) {
1935 				IPSECLOG(LOG_DEBUG,
1936 				    "reqid=%d range "
1937 				    "violation, updated by kernel.\n",
1938 				    xisr_reqid);
1939 				xisr_reqid = 0;
1940 			}
1941 
1942 			/* allocate new reqid id if reqid is zero. */
1943 			if (xisr_reqid == 0) {
1944 				u_int16_t reqid = key_newreqid();
1945 				if (reqid == 0) {
1946 					*error = ENOBUFS;
1947 					goto free_exit;
1948 				}
1949 				(*p_isr)->saidx.reqid = reqid;
1950 			} else {
1951 			/* set it for manual keying. */
1952 				(*p_isr)->saidx.reqid = xisr_reqid;
1953 			}
1954 			break;
1955 
1956 		default:
1957 			IPSECLOG(LOG_DEBUG, "invalid level=%u\n",
1958 			    xisr->sadb_x_ipsecrequest_level);
1959 			*error = EINVAL;
1960 			goto free_exit;
1961 		}
1962 		(*p_isr)->level = xisr->sadb_x_ipsecrequest_level;
1963 
1964 		/* set IP addresses if there */
1965 		/*
1966 		 * NOTE:
1967 		 * MOBIKE Extensions for PF_KEY draft says:
1968 		 *     If tunnel mode is specified, the sadb_x_ipsecrequest
1969 		 *     structure is followed by two sockaddr structures that
1970 		 *     define the tunnel endpoint addresses.  In the case that
1971 		 *     transport mode is used, no additional addresses are
1972 		 *     specified.
1973 		 * see: https://tools.ietf.org/html/draft-schilcher-mobike-pfkey-extension-01
1974 		 *
1975 		 * And then, the IP addresses will be set by
1976 		 * ipsec_fill_saidx_bymbuf() from packet in transport mode.
1977 		 * This behavior is used by NAT-T enabled ipsecif(4).
1978 		 */
1979 		if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1980 			const struct sockaddr *paddr;
1981 
1982 			paddr = (const struct sockaddr *)(xisr + 1);
1983 
1984 			/* validity check */
1985 			if (paddr->sa_len > sizeof((*p_isr)->saidx.src)) {
1986 				IPSECLOG(LOG_DEBUG, "invalid request "
1987 				    "address length.\n");
1988 				*error = EINVAL;
1989 				goto free_exit;
1990 			}
1991 			memcpy(&(*p_isr)->saidx.src, paddr, paddr->sa_len);
1992 
1993 			paddr = (const struct sockaddr *)((const char *)paddr
1994 			    + paddr->sa_len);
1995 
1996 			/* validity check */
1997 			if (paddr->sa_len > sizeof((*p_isr)->saidx.dst)) {
1998 				IPSECLOG(LOG_DEBUG, "invalid request "
1999 				    "address length.\n");
2000 				*error = EINVAL;
2001 				goto free_exit;
2002 			}
2003 			memcpy(&(*p_isr)->saidx.dst, paddr, paddr->sa_len);
2004 		}
2005 
2006 		(*p_isr)->sp = newsp;
2007 
2008 		/* initialization for the next. */
2009 		p_isr = &(*p_isr)->next;
2010 		tlen -= xisr->sadb_x_ipsecrequest_len;
2011 
2012 		/* validity check */
2013 		if (tlen < 0) {
2014 			IPSECLOG(LOG_DEBUG, "becoming tlen < 0.\n");
2015 			*error = EINVAL;
2016 			goto free_exit;
2017 		}
2018 
2019 		xisr = (const struct sadb_x_ipsecrequest *)((const char *)xisr +
2020 		    xisr->sadb_x_ipsecrequest_len);
2021 	}
2022     }
2023 
2024 	*error = 0;
2025 	return newsp;
2026 
2027 free_exit:
2028 	key_free_sp(newsp);
2029 	return NULL;
2030 }
2031 
2032 struct secpolicy *
2033 key_msg2sp(const struct sadb_x_policy *xpl0, size_t len, int *error)
2034 {
2035 
2036 	return _key_msg2sp(xpl0, len, error, false);
2037 }
2038 
2039 u_int16_t
2040 key_newreqid(void)
2041 {
2042 	static u_int16_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
2043 
2044 	auto_reqid = (auto_reqid == 0xffff ?
2045 	    IPSEC_MANUAL_REQID_MAX + 1 : auto_reqid + 1);
2046 
2047 	/* XXX should be unique check */
2048 
2049 	return auto_reqid;
2050 }
2051 
2052 /*
2053  * copy secpolicy struct to sadb_x_policy structure indicated.
2054  */
2055 struct mbuf *
2056 key_sp2msg(const struct secpolicy *sp, int mflag)
2057 {
2058 	struct sadb_x_policy *xpl;
2059 	int tlen;
2060 	char *p;
2061 	struct mbuf *m;
2062 
2063 	KASSERT(sp != NULL);
2064 
2065 	tlen = key_getspreqmsglen(sp);
2066 
2067 	m = key_alloc_mbuf(tlen, mflag);
2068 	if (!m || m->m_next) {	/*XXX*/
2069 		if (m)
2070 			m_freem(m);
2071 		return NULL;
2072 	}
2073 
2074 	m->m_len = tlen;
2075 	m->m_next = NULL;
2076 	xpl = mtod(m, struct sadb_x_policy *);
2077 	memset(xpl, 0, tlen);
2078 
2079 	xpl->sadb_x_policy_len = PFKEY_UNIT64(tlen);
2080 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
2081 	xpl->sadb_x_policy_type = sp->policy;
2082 	xpl->sadb_x_policy_dir = sp->spidx.dir;
2083 	xpl->sadb_x_policy_id = sp->id;
2084 	p = (char *)xpl + sizeof(*xpl);
2085 
2086 	/* if is the policy for ipsec ? */
2087 	if (sp->policy == IPSEC_POLICY_IPSEC) {
2088 		struct sadb_x_ipsecrequest *xisr;
2089 		struct ipsecrequest *isr;
2090 
2091 		for (isr = sp->req; isr != NULL; isr = isr->next) {
2092 
2093 			xisr = (struct sadb_x_ipsecrequest *)p;
2094 
2095 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
2096 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
2097 			xisr->sadb_x_ipsecrequest_level = isr->level;
2098 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
2099 
2100 			p += sizeof(*xisr);
2101 			memcpy(p, &isr->saidx.src, isr->saidx.src.sa.sa_len);
2102 			p += isr->saidx.src.sa.sa_len;
2103 			memcpy(p, &isr->saidx.dst, isr->saidx.dst.sa.sa_len);
2104 			p += isr->saidx.src.sa.sa_len;
2105 
2106 			xisr->sadb_x_ipsecrequest_len =
2107 			    PFKEY_ALIGN8(sizeof(*xisr)
2108 			    + isr->saidx.src.sa.sa_len
2109 			    + isr->saidx.dst.sa.sa_len);
2110 		}
2111 	}
2112 
2113 	return m;
2114 }
2115 
2116 /*
2117  * m will not be freed nor modified. It never return NULL.
2118  * If it returns a mbuf of M_PKTHDR, the mbuf ensures to have
2119  * contiguous length at least sizeof(struct sadb_msg).
2120  */
2121 static struct mbuf *
2122 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
2123 		int ndeep, int nitem, ...)
2124 {
2125 	va_list ap;
2126 	int idx;
2127 	int i;
2128 	struct mbuf *result = NULL, *n;
2129 	int len;
2130 
2131 	KASSERT(m != NULL);
2132 	KASSERT(mhp != NULL);
2133 	KASSERT(!cpu_softintr_p());
2134 
2135 	va_start(ap, nitem);
2136 	for (i = 0; i < nitem; i++) {
2137 		idx = va_arg(ap, int);
2138 		KASSERT(idx >= 0);
2139 		KASSERT(idx <= SADB_EXT_MAX);
2140 		/* don't attempt to pull empty extension */
2141 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
2142 			continue;
2143 		if (idx != SADB_EXT_RESERVED &&
2144 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
2145 			continue;
2146 
2147 		if (idx == SADB_EXT_RESERVED) {
2148 			CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MHLEN);
2149 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2150 			MGETHDR(n, M_WAITOK, MT_DATA);
2151 			n->m_len = len;
2152 			n->m_next = NULL;
2153 			m_copydata(m, 0, sizeof(struct sadb_msg),
2154 			    mtod(n, void *));
2155 		} else if (i < ndeep) {
2156 			len = mhp->extlen[idx];
2157 			n = key_alloc_mbuf(len, M_WAITOK);
2158 			KASSERT(n->m_next == NULL);
2159 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
2160 			    mtod(n, void *));
2161 		} else {
2162 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
2163 			    M_WAITOK);
2164 		}
2165 		KASSERT(n != NULL);
2166 
2167 		if (result)
2168 			m_cat(result, n);
2169 		else
2170 			result = n;
2171 	}
2172 	va_end(ap);
2173 
2174 	KASSERT(result != NULL);
2175 	if ((result->m_flags & M_PKTHDR) != 0) {
2176 		result->m_pkthdr.len = 0;
2177 		for (n = result; n; n = n->m_next)
2178 			result->m_pkthdr.len += n->m_len;
2179 		KASSERT(result->m_len >= sizeof(struct sadb_msg));
2180 	}
2181 
2182 	return result;
2183 }
2184 
2185 /*
2186  * The argument _sp must not overwrite until SP is created and registered
2187  * successfully.
2188  */
2189 static int
2190 key_spdadd(struct socket *so, struct mbuf *m,
2191 	   const struct sadb_msghdr *mhp, struct secpolicy **_sp,
2192 	   bool from_kernel)
2193 {
2194 	const struct sockaddr *src, *dst;
2195 	const struct sadb_x_policy *xpl0;
2196 	struct sadb_x_policy *xpl;
2197 	const struct sadb_lifetime *lft = NULL;
2198 	struct secpolicyindex spidx;
2199 	struct secpolicy *newsp;
2200 	int error;
2201 	uint32_t sadb_x_policy_id;
2202 
2203 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
2204 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
2205 	    mhp->ext[SADB_X_EXT_POLICY] == NULL) {
2206 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2207 		return key_senderror(so, m, EINVAL);
2208 	}
2209 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
2210 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
2211 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2212 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2213 		return key_senderror(so, m, EINVAL);
2214 	}
2215 	if (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL) {
2216 		if (mhp->extlen[SADB_EXT_LIFETIME_HARD] <
2217 		    sizeof(struct sadb_lifetime)) {
2218 			IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2219 			return key_senderror(so, m, EINVAL);
2220 		}
2221 		lft = mhp->ext[SADB_EXT_LIFETIME_HARD];
2222 	}
2223 
2224 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
2225 
2226 	/* checking the direciton. */
2227 	switch (xpl0->sadb_x_policy_dir) {
2228 	case IPSEC_DIR_INBOUND:
2229 	case IPSEC_DIR_OUTBOUND:
2230 		break;
2231 	default:
2232 		IPSECLOG(LOG_DEBUG, "Invalid SP direction.\n");
2233 		return key_senderror(so, m, EINVAL);
2234 	}
2235 
2236 	/* check policy */
2237 	/* key_api_spdadd() accepts DISCARD, NONE and IPSEC. */
2238 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_ENTRUST ||
2239 	    xpl0->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
2240 		IPSECLOG(LOG_DEBUG, "Invalid policy type.\n");
2241 		return key_senderror(so, m, EINVAL);
2242 	}
2243 
2244 	/* policy requests are mandatory when action is ipsec. */
2245 	if (mhp->msg->sadb_msg_type != SADB_X_SPDSETIDX &&
2246 	    xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
2247 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
2248 		IPSECLOG(LOG_DEBUG, "some policy requests part required.\n");
2249 		return key_senderror(so, m, EINVAL);
2250 	}
2251 
2252 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
2253 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
2254 
2255 	/* sanity check on addr pair */
2256 	if (src->sa_family != dst->sa_family)
2257 		return key_senderror(so, m, EINVAL);
2258 	if (src->sa_len != dst->sa_len)
2259 		return key_senderror(so, m, EINVAL);
2260 
2261 	key_init_spidx_bymsghdr(&spidx, mhp);
2262 
2263 	/*
2264 	 * checking there is SP already or not.
2265 	 * SPDUPDATE doesn't depend on whether there is a SP or not.
2266 	 * If the type is either SPDADD or SPDSETIDX AND a SP is found,
2267 	 * then error.
2268 	 */
2269     {
2270 	struct secpolicy *sp;
2271 
2272 	if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2273 		sp = key_lookup_and_remove_sp(&spidx, from_kernel);
2274 		if (sp != NULL)
2275 			key_destroy_sp(sp);
2276 	} else {
2277 		sp = key_getsp(&spidx);
2278 		if (sp != NULL) {
2279 			KEY_SP_UNREF(&sp);
2280 			IPSECLOG(LOG_DEBUG, "a SP entry exists already.\n");
2281 			return key_senderror(so, m, EEXIST);
2282 		}
2283 	}
2284     }
2285 
2286 	/* allocation new SP entry */
2287 	newsp = _key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error, from_kernel);
2288 	if (newsp == NULL) {
2289 		return key_senderror(so, m, error);
2290 	}
2291 
2292 	newsp->id = key_getnewspid();
2293 	if (newsp->id == 0) {
2294 		kmem_free(newsp, sizeof(*newsp));
2295 		return key_senderror(so, m, ENOBUFS);
2296 	}
2297 
2298 	newsp->spidx = spidx;
2299 	newsp->created = time_uptime;
2300 	newsp->lastused = newsp->created;
2301 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
2302 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
2303 	if (from_kernel)
2304 		newsp->origin = IPSEC_SPORIGIN_KERNEL;
2305 	else
2306 		newsp->origin = IPSEC_SPORIGIN_USER;
2307 
2308 	key_init_sp(newsp);
2309 	if (from_kernel)
2310 		KEY_SP_REF(newsp);
2311 
2312 	sadb_x_policy_id = newsp->id;
2313 
2314 	if (_sp != NULL)
2315 		*_sp = newsp;
2316 
2317 	mutex_enter(&key_spd.lock);
2318 	SPLIST_WRITER_INSERT_TAIL(newsp->spidx.dir, newsp);
2319 	mutex_exit(&key_spd.lock);
2320 	/*
2321 	 * We don't have a reference to newsp, so we must not touch newsp from
2322 	 * now on.  If you want to do, you must take a reference beforehand.
2323 	 */
2324 	newsp = NULL;
2325 
2326 #ifdef notyet
2327 	/* delete the entry in key_misc.spacqlist */
2328 	if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2329 		struct secspacq *spacq = key_getspacq(&spidx);
2330 		if (spacq != NULL) {
2331 			/* reset counter in order to deletion by timehandler. */
2332 			spacq->created = time_uptime;
2333 			spacq->count = 0;
2334 		}
2335     	}
2336 #endif
2337 
2338 	/* Invalidate all cached SPD pointers in the PCBs. */
2339 	ipsec_invalpcbcacheall();
2340 
2341 #if defined(GATEWAY)
2342 	/* Invalidate the ipflow cache, as well. */
2343 	ipflow_invalidate_all(0);
2344 #ifdef INET6
2345 	if (in6_present)
2346 		ip6flow_invalidate_all(0);
2347 #endif /* INET6 */
2348 #endif /* GATEWAY */
2349 
2350 	key_update_used();
2351 
2352     {
2353 	struct mbuf *n, *mpolicy;
2354 	int off;
2355 
2356 	/* create new sadb_msg to reply. */
2357 	if (lft) {
2358 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
2359 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
2360 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2361 	} else {
2362 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
2363 		    SADB_X_EXT_POLICY,
2364 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2365 	}
2366 
2367 	key_fill_replymsg(n, 0);
2368 	off = 0;
2369 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
2370 	    sizeof(*xpl), &off);
2371 	if (mpolicy == NULL) {
2372 		/* n is already freed */
2373 		/*
2374 		 * valid sp has been created, so we does not overwrite _sp
2375 		 * NULL here. let caller decide to use the sp or not.
2376 		 */
2377 		return key_senderror(so, m, ENOBUFS);
2378 	}
2379 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, char *) + off);
2380 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
2381 		m_freem(n);
2382 		/* ditto */
2383 		return key_senderror(so, m, EINVAL);
2384 	}
2385 
2386 	xpl->sadb_x_policy_id = sadb_x_policy_id;
2387 
2388 	m_freem(m);
2389 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2390     }
2391 }
2392 
2393 /*
2394  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
2395  * add an entry to SP database, when received
2396  *   <base, address(SD), (lifetime(H),) policy>
2397  * from the user(?).
2398  * Adding to SP database,
2399  * and send
2400  *   <base, address(SD), (lifetime(H),) policy>
2401  * to the socket which was send.
2402  *
2403  * SPDADD set a unique policy entry.
2404  * SPDSETIDX like SPDADD without a part of policy requests.
2405  * SPDUPDATE replace a unique policy entry.
2406  *
2407  * m will always be freed.
2408  */
2409 static int
2410 key_api_spdadd(struct socket *so, struct mbuf *m,
2411 	       const struct sadb_msghdr *mhp)
2412 {
2413 
2414 	return key_spdadd(so, m, mhp, NULL, false);
2415 }
2416 
2417 struct secpolicy *
2418 key_kpi_spdadd(struct mbuf *m)
2419 {
2420 	struct sadb_msghdr mh;
2421 	int error;
2422 	struct secpolicy *sp = NULL;
2423 
2424 	error = key_align(m, &mh);
2425 	if (error)
2426 		return NULL;
2427 
2428 	error = key_spdadd(NULL, m, &mh, &sp, true);
2429 	if (error) {
2430 		/*
2431 		 * Currently, when key_spdadd() cannot send a PFKEY message
2432 		 * which means SP has been created, key_spdadd() returns error
2433 		 * although SP is created successfully.
2434 		 * Kernel components would not care PFKEY messages, so return
2435 		 * the "sp" regardless of error code. key_spdadd() overwrites
2436 		 * the argument only if SP  is created successfully.
2437 		 */
2438 	}
2439 	return sp;
2440 }
2441 
2442 /*
2443  * get new policy id.
2444  * OUT:
2445  *	0:	failure.
2446  *	others: success.
2447  */
2448 static u_int32_t
2449 key_getnewspid(void)
2450 {
2451 	u_int32_t newid = 0;
2452 	int count = key_spi_trycnt;	/* XXX */
2453 	struct secpolicy *sp;
2454 
2455 	/* when requesting to allocate spi ranged */
2456 	while (count--) {
2457 		newid = (policy_id = (policy_id == ~0 ? 1 : policy_id + 1));
2458 
2459 		sp = key_getspbyid(newid);
2460 		if (sp == NULL)
2461 			break;
2462 
2463 		KEY_SP_UNREF(&sp);
2464 	}
2465 
2466 	if (count == 0 || newid == 0) {
2467 		IPSECLOG(LOG_DEBUG, "to allocate policy id is failed.\n");
2468 		return 0;
2469 	}
2470 
2471 	return newid;
2472 }
2473 
2474 /*
2475  * SADB_SPDDELETE processing
2476  * receive
2477  *   <base, address(SD), policy(*)>
2478  * from the user(?), and set SADB_SASTATE_DEAD,
2479  * and send,
2480  *   <base, address(SD), policy(*)>
2481  * to the ikmpd.
2482  * policy(*) including direction of policy.
2483  *
2484  * m will always be freed.
2485  */
2486 static int
2487 key_api_spddelete(struct socket *so, struct mbuf *m,
2488               const struct sadb_msghdr *mhp)
2489 {
2490 	struct sadb_x_policy *xpl0;
2491 	struct secpolicyindex spidx;
2492 	struct secpolicy *sp;
2493 
2494 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
2495 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
2496 	    mhp->ext[SADB_X_EXT_POLICY] == NULL) {
2497 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2498 		return key_senderror(so, m, EINVAL);
2499 	}
2500 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
2501 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
2502 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2503 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2504 		return key_senderror(so, m, EINVAL);
2505 	}
2506 
2507 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
2508 
2509 	/* checking the directon. */
2510 	switch (xpl0->sadb_x_policy_dir) {
2511 	case IPSEC_DIR_INBOUND:
2512 	case IPSEC_DIR_OUTBOUND:
2513 		break;
2514 	default:
2515 		IPSECLOG(LOG_DEBUG, "Invalid SP direction.\n");
2516 		return key_senderror(so, m, EINVAL);
2517 	}
2518 
2519 	/* make secindex */
2520 	key_init_spidx_bymsghdr(&spidx, mhp);
2521 
2522 	/* Is there SP in SPD ? */
2523 	sp = key_lookup_and_remove_sp(&spidx, false);
2524 	if (sp == NULL) {
2525 		IPSECLOG(LOG_DEBUG, "no SP found.\n");
2526 		return key_senderror(so, m, EINVAL);
2527 	}
2528 
2529 	/* save policy id to buffer to be returned. */
2530 	xpl0->sadb_x_policy_id = sp->id;
2531 
2532 	key_destroy_sp(sp);
2533 
2534 	/* We're deleting policy; no need to invalidate the ipflow cache. */
2535 
2536     {
2537 	struct mbuf *n;
2538 
2539 	/* create new sadb_msg to reply. */
2540 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2541 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2542 	key_fill_replymsg(n, 0);
2543 	m_freem(m);
2544 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2545     }
2546 }
2547 
2548 static struct mbuf *
2549 key_alloc_mbuf_simple(int len, int mflag)
2550 {
2551 	struct mbuf *n;
2552 
2553 	KASSERT(mflag == M_NOWAIT || (mflag == M_WAITOK && !cpu_softintr_p()));
2554 
2555 	MGETHDR(n, mflag, MT_DATA);
2556 	if (n && len > MHLEN) {
2557 		MCLGET(n, mflag);
2558 		if ((n->m_flags & M_EXT) == 0) {
2559 			m_freem(n);
2560 			n = NULL;
2561 		}
2562 	}
2563 	return n;
2564 }
2565 
2566 /*
2567  * SADB_SPDDELETE2 processing
2568  * receive
2569  *   <base, policy(*)>
2570  * from the user(?), and set SADB_SASTATE_DEAD,
2571  * and send,
2572  *   <base, policy(*)>
2573  * to the ikmpd.
2574  * policy(*) including direction of policy.
2575  *
2576  * m will always be freed.
2577  */
2578 static int
2579 key_spddelete2(struct socket *so, struct mbuf *m,
2580 	       const struct sadb_msghdr *mhp, bool from_kernel)
2581 {
2582 	u_int32_t id;
2583 	struct secpolicy *sp;
2584 	const struct sadb_x_policy *xpl;
2585 
2586 	if (mhp->ext[SADB_X_EXT_POLICY] == NULL ||
2587 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2588 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2589 		return key_senderror(so, m, EINVAL);
2590 	}
2591 
2592 	xpl = mhp->ext[SADB_X_EXT_POLICY];
2593 	id = xpl->sadb_x_policy_id;
2594 
2595 	/* Is there SP in SPD ? */
2596 	sp = key_lookupbyid_and_remove_sp(id, from_kernel);
2597 	if (sp == NULL) {
2598 		IPSECLOG(LOG_DEBUG, "no SP found id:%u.\n", id);
2599 		return key_senderror(so, m, EINVAL);
2600 	}
2601 
2602 	key_destroy_sp(sp);
2603 
2604 	/* We're deleting policy; no need to invalidate the ipflow cache. */
2605 
2606     {
2607 	struct mbuf *n, *nn;
2608 	int off, len;
2609 
2610 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MCLBYTES);
2611 
2612 	/* create new sadb_msg to reply. */
2613 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2614 
2615 	n = key_alloc_mbuf_simple(len, M_WAITOK);
2616 	n->m_len = len;
2617 	n->m_next = NULL;
2618 	off = 0;
2619 
2620 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
2621 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2622 
2623 	KASSERTMSG(off == len, "length inconsistency");
2624 
2625 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2626 	    mhp->extlen[SADB_X_EXT_POLICY], M_WAITOK);
2627 
2628 	n->m_pkthdr.len = 0;
2629 	for (nn = n; nn; nn = nn->m_next)
2630 		n->m_pkthdr.len += nn->m_len;
2631 
2632 	key_fill_replymsg(n, 0);
2633 	m_freem(m);
2634 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2635     }
2636 }
2637 
2638 /*
2639  * SADB_SPDDELETE2 processing
2640  * receive
2641  *   <base, policy(*)>
2642  * from the user(?), and set SADB_SASTATE_DEAD,
2643  * and send,
2644  *   <base, policy(*)>
2645  * to the ikmpd.
2646  * policy(*) including direction of policy.
2647  *
2648  * m will always be freed.
2649  */
2650 static int
2651 key_api_spddelete2(struct socket *so, struct mbuf *m,
2652 	       const struct sadb_msghdr *mhp)
2653 {
2654 
2655 	return key_spddelete2(so, m, mhp, false);
2656 }
2657 
2658 int
2659 key_kpi_spddelete2(struct mbuf *m)
2660 {
2661 	struct sadb_msghdr mh;
2662 	int error;
2663 
2664 	error = key_align(m, &mh);
2665 	if (error)
2666 		return EINVAL;
2667 
2668 	return key_spddelete2(NULL, m, &mh, true);
2669 }
2670 
2671 /*
2672  * SADB_X_GET processing
2673  * receive
2674  *   <base, policy(*)>
2675  * from the user(?),
2676  * and send,
2677  *   <base, address(SD), policy>
2678  * to the ikmpd.
2679  * policy(*) including direction of policy.
2680  *
2681  * m will always be freed.
2682  */
2683 static int
2684 key_api_spdget(struct socket *so, struct mbuf *m,
2685 	   const struct sadb_msghdr *mhp)
2686 {
2687 	u_int32_t id;
2688 	struct secpolicy *sp;
2689 	struct mbuf *n;
2690 	const struct sadb_x_policy *xpl;
2691 
2692 	if (mhp->ext[SADB_X_EXT_POLICY] == NULL ||
2693 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2694 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2695 		return key_senderror(so, m, EINVAL);
2696 	}
2697 
2698 	xpl = mhp->ext[SADB_X_EXT_POLICY];
2699 	id = xpl->sadb_x_policy_id;
2700 
2701 	/* Is there SP in SPD ? */
2702 	sp = key_getspbyid(id);
2703 	if (sp == NULL) {
2704 		IPSECLOG(LOG_DEBUG, "no SP found id:%u.\n", id);
2705 		return key_senderror(so, m, ENOENT);
2706 	}
2707 
2708 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2709 	    mhp->msg->sadb_msg_pid);
2710 	KEY_SP_UNREF(&sp); /* ref gained by key_getspbyid */
2711 	m_freem(m);
2712 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2713 }
2714 
2715 #ifdef notyet
2716 /*
2717  * SADB_X_SPDACQUIRE processing.
2718  * Acquire policy and SA(s) for a *OUTBOUND* packet.
2719  * send
2720  *   <base, policy(*)>
2721  * to KMD, and expect to receive
2722  *   <base> with SADB_X_SPDACQUIRE if error occurred,
2723  * or
2724  *   <base, policy>
2725  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2726  * policy(*) is without policy requests.
2727  *
2728  *    0     : succeed
2729  *    others: error number
2730  */
2731 int
2732 key_spdacquire(const struct secpolicy *sp)
2733 {
2734 	struct mbuf *result = NULL, *m;
2735 	struct secspacq *newspacq;
2736 	int error;
2737 
2738 	KASSERT(sp != NULL);
2739 	KASSERTMSG(sp->req == NULL, "called but there is request");
2740 	KASSERTMSG(sp->policy == IPSEC_POLICY_IPSEC,
2741 	    "policy mismathed. IPsec is expected");
2742 
2743 	/* Get an entry to check whether sent message or not. */
2744 	newspacq = key_getspacq(&sp->spidx);
2745 	if (newspacq != NULL) {
2746 		if (key_blockacq_count < newspacq->count) {
2747 			/* reset counter and do send message. */
2748 			newspacq->count = 0;
2749 		} else {
2750 			/* increment counter and do nothing. */
2751 			newspacq->count++;
2752 			return 0;
2753 		}
2754 	} else {
2755 		/* make new entry for blocking to send SADB_ACQUIRE. */
2756 		newspacq = key_newspacq(&sp->spidx);
2757 		if (newspacq == NULL)
2758 			return ENOBUFS;
2759 
2760 		/* add to key_misc.acqlist */
2761 		LIST_INSERT_HEAD(&key_misc.spacqlist, newspacq, chain);
2762 	}
2763 
2764 	/* create new sadb_msg to reply. */
2765 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2766 	if (!m) {
2767 		error = ENOBUFS;
2768 		goto fail;
2769 	}
2770 	result = m;
2771 
2772 	result->m_pkthdr.len = 0;
2773 	for (m = result; m; m = m->m_next)
2774 		result->m_pkthdr.len += m->m_len;
2775 
2776 	mtod(result, struct sadb_msg *)->sadb_msg_len =
2777 	    PFKEY_UNIT64(result->m_pkthdr.len);
2778 
2779 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2780 
2781 fail:
2782 	if (result)
2783 		m_freem(result);
2784 	return error;
2785 }
2786 #endif /* notyet */
2787 
2788 /*
2789  * SADB_SPDFLUSH processing
2790  * receive
2791  *   <base>
2792  * from the user, and free all entries in secpctree.
2793  * and send,
2794  *   <base>
2795  * to the user.
2796  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2797  *
2798  * m will always be freed.
2799  */
2800 static int
2801 key_api_spdflush(struct socket *so, struct mbuf *m,
2802 	     const struct sadb_msghdr *mhp)
2803 {
2804 	struct sadb_msg *newmsg;
2805 	struct secpolicy *sp;
2806 	u_int dir;
2807 
2808 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2809 		return key_senderror(so, m, EINVAL);
2810 
2811 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2812 	    retry:
2813 		mutex_enter(&key_spd.lock);
2814 		SPLIST_WRITER_FOREACH(sp, dir) {
2815 			KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
2816 			/*
2817 			 * Userlang programs can remove SPs created by userland
2818 			 * probrams only, that is, they cannot remove SPs
2819 			 * created in kernel(e.g. ipsec(4) I/F).
2820 			 */
2821 			if (sp->origin == IPSEC_SPORIGIN_USER) {
2822 				key_unlink_sp(sp);
2823 				mutex_exit(&key_spd.lock);
2824 				key_destroy_sp(sp);
2825 				goto retry;
2826 			}
2827 		}
2828 		mutex_exit(&key_spd.lock);
2829 	}
2830 
2831 	/* We're deleting policy; no need to invalidate the ipflow cache. */
2832 
2833 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2834 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
2835 		return key_senderror(so, m, ENOBUFS);
2836 	}
2837 
2838 	if (m->m_next)
2839 		m_freem(m->m_next);
2840 	m->m_next = NULL;
2841 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2842 	newmsg = mtod(m, struct sadb_msg *);
2843 	newmsg->sadb_msg_errno = 0;
2844 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2845 
2846 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2847 }
2848 
2849 static struct sockaddr key_src = {
2850 	.sa_len = 2,
2851 	.sa_family = PF_KEY,
2852 };
2853 
2854 static struct mbuf *
2855 key_setspddump_chain(int *errorp, int *lenp, pid_t pid)
2856 {
2857 	struct secpolicy *sp;
2858 	int cnt;
2859 	u_int dir;
2860 	struct mbuf *m, *n, *prev;
2861 	int totlen;
2862 
2863 	KASSERT(mutex_owned(&key_spd.lock));
2864 
2865 	*lenp = 0;
2866 
2867 	/* search SPD entry and get buffer size. */
2868 	cnt = 0;
2869 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2870 		SPLIST_WRITER_FOREACH(sp, dir) {
2871 			cnt++;
2872 		}
2873 	}
2874 
2875 	if (cnt == 0) {
2876 		*errorp = ENOENT;
2877 		return (NULL);
2878 	}
2879 
2880 	m = NULL;
2881 	prev = m;
2882 	totlen = 0;
2883 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2884 		SPLIST_WRITER_FOREACH(sp, dir) {
2885 			--cnt;
2886 			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, pid);
2887 
2888 			totlen += n->m_pkthdr.len;
2889 			if (!m) {
2890 				m = n;
2891 			} else {
2892 				prev->m_nextpkt = n;
2893 			}
2894 			prev = n;
2895 		}
2896 	}
2897 
2898 	*lenp = totlen;
2899 	*errorp = 0;
2900 	return (m);
2901 }
2902 
2903 /*
2904  * SADB_SPDDUMP processing
2905  * receive
2906  *   <base>
2907  * from the user, and dump all SP leaves
2908  * and send,
2909  *   <base> .....
2910  * to the ikmpd.
2911  *
2912  * m will always be freed.
2913  */
2914 static int
2915 key_api_spddump(struct socket *so, struct mbuf *m0,
2916  	    const struct sadb_msghdr *mhp)
2917 {
2918 	struct mbuf *n;
2919 	int error, len;
2920 	int ok;
2921 	pid_t pid;
2922 
2923 	pid = mhp->msg->sadb_msg_pid;
2924 	/*
2925 	 * If the requestor has insufficient socket-buffer space
2926 	 * for the entire chain, nobody gets any response to the DUMP.
2927 	 * XXX For now, only the requestor ever gets anything.
2928 	 * Moreover, if the requestor has any space at all, they receive
2929 	 * the entire chain, otherwise the request is refused with  ENOBUFS.
2930 	 */
2931 	if (sbspace(&so->so_rcv) <= 0) {
2932 		return key_senderror(so, m0, ENOBUFS);
2933 	}
2934 
2935 	mutex_enter(&key_spd.lock);
2936 	n = key_setspddump_chain(&error, &len, pid);
2937 	mutex_exit(&key_spd.lock);
2938 
2939 	if (n == NULL) {
2940 		return key_senderror(so, m0, ENOENT);
2941 	}
2942 	{
2943 		uint64_t *ps = PFKEY_STAT_GETREF();
2944 		ps[PFKEY_STAT_IN_TOTAL]++;
2945 		ps[PFKEY_STAT_IN_BYTES] += len;
2946 		PFKEY_STAT_PUTREF();
2947 	}
2948 
2949 	/*
2950 	 * PF_KEY DUMP responses are no longer broadcast to all PF_KEY sockets.
2951 	 * The requestor receives either the entire chain, or an
2952 	 * error message with ENOBUFS.
2953 	 */
2954 
2955 	/*
2956 	 * sbappendchainwith record takes the chain of entries, one
2957 	 * packet-record per SPD entry, prepends the key_src sockaddr
2958 	 * to each packet-record, links the sockaddr mbufs into a new
2959 	 * list of records, then   appends the entire resulting
2960 	 * list to the requesting socket.
2961 	 */
2962 	ok = sbappendaddrchain(&so->so_rcv, (struct sockaddr *)&key_src, n,
2963 	    SB_PRIO_ONESHOT_OVERFLOW);
2964 
2965 	if (!ok) {
2966 		PFKEY_STATINC(PFKEY_STAT_IN_NOMEM);
2967 		m_freem(n);
2968 		return key_senderror(so, m0, ENOBUFS);
2969 	}
2970 
2971 	m_freem(m0);
2972 	return error;
2973 }
2974 
2975 /*
2976  * SADB_X_NAT_T_NEW_MAPPING. Unused by racoon as of 2005/04/23
2977  */
2978 static int
2979 key_api_nat_map(struct socket *so, struct mbuf *m,
2980 	    const struct sadb_msghdr *mhp)
2981 {
2982 	struct sadb_x_nat_t_type *type;
2983 	struct sadb_x_nat_t_port *sport;
2984 	struct sadb_x_nat_t_port *dport;
2985 	struct sadb_address *iaddr, *raddr;
2986 	struct sadb_x_nat_t_frag *frag;
2987 
2988 	if (mhp->ext[SADB_X_EXT_NAT_T_TYPE] == NULL ||
2989 	    mhp->ext[SADB_X_EXT_NAT_T_SPORT] == NULL ||
2990 	    mhp->ext[SADB_X_EXT_NAT_T_DPORT] == NULL) {
2991 		IPSECLOG(LOG_DEBUG, "invalid message.\n");
2992 		return key_senderror(so, m, EINVAL);
2993 	}
2994 	if ((mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) ||
2995 	    (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) ||
2996 	    (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport))) {
2997 		IPSECLOG(LOG_DEBUG, "invalid message.\n");
2998 		return key_senderror(so, m, EINVAL);
2999 	}
3000 
3001 	if ((mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL) &&
3002 	    (mhp->extlen[SADB_X_EXT_NAT_T_OAI] < sizeof(*iaddr))) {
3003 		IPSECLOG(LOG_DEBUG, "invalid message\n");
3004 		return key_senderror(so, m, EINVAL);
3005 	}
3006 
3007 	if ((mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL) &&
3008 	    (mhp->extlen[SADB_X_EXT_NAT_T_OAR] < sizeof(*raddr))) {
3009 		IPSECLOG(LOG_DEBUG, "invalid message\n");
3010 		return key_senderror(so, m, EINVAL);
3011 	}
3012 
3013 	if ((mhp->ext[SADB_X_EXT_NAT_T_FRAG] != NULL) &&
3014 	    (mhp->extlen[SADB_X_EXT_NAT_T_FRAG] < sizeof(*frag))) {
3015 		IPSECLOG(LOG_DEBUG, "invalid message\n");
3016 		return key_senderror(so, m, EINVAL);
3017 	}
3018 
3019 	type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
3020 	sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
3021 	dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
3022 	iaddr = mhp->ext[SADB_X_EXT_NAT_T_OAI];
3023 	raddr = mhp->ext[SADB_X_EXT_NAT_T_OAR];
3024 	frag = mhp->ext[SADB_X_EXT_NAT_T_FRAG];
3025 
3026 	/*
3027 	 * XXX handle that, it should also contain a SA, or anything
3028 	 * that enable to update the SA information.
3029 	 */
3030 
3031 	return 0;
3032 }
3033 
3034 /*
3035  * Never return NULL.
3036  */
3037 static struct mbuf *
3038 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq, pid_t pid)
3039 {
3040 	struct mbuf *result = NULL, *m;
3041 
3042 	KASSERT(!cpu_softintr_p());
3043 
3044 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid,
3045 	    key_sp_refcnt(sp), M_WAITOK);
3046 	result = m;
3047 
3048 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3049 	    &sp->spidx.src.sa, sp->spidx.prefs, sp->spidx.ul_proto, M_WAITOK);
3050 	m_cat(result, m);
3051 
3052 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3053 	    &sp->spidx.dst.sa, sp->spidx.prefd, sp->spidx.ul_proto, M_WAITOK);
3054 	m_cat(result, m);
3055 
3056 	m = key_sp2msg(sp, M_WAITOK);
3057 	m_cat(result, m);
3058 
3059 	KASSERT(result->m_flags & M_PKTHDR);
3060 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
3061 
3062 	result->m_pkthdr.len = 0;
3063 	for (m = result; m; m = m->m_next)
3064 		result->m_pkthdr.len += m->m_len;
3065 
3066 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3067 	    PFKEY_UNIT64(result->m_pkthdr.len);
3068 
3069 	return result;
3070 }
3071 
3072 /*
3073  * get PFKEY message length for security policy and request.
3074  */
3075 static u_int
3076 key_getspreqmsglen(const struct secpolicy *sp)
3077 {
3078 	u_int tlen;
3079 
3080 	tlen = sizeof(struct sadb_x_policy);
3081 
3082 	/* if is the policy for ipsec ? */
3083 	if (sp->policy != IPSEC_POLICY_IPSEC)
3084 		return tlen;
3085 
3086 	/* get length of ipsec requests */
3087     {
3088 	const struct ipsecrequest *isr;
3089 	int len;
3090 
3091 	for (isr = sp->req; isr != NULL; isr = isr->next) {
3092 		len = sizeof(struct sadb_x_ipsecrequest)
3093 		    + isr->saidx.src.sa.sa_len + isr->saidx.dst.sa.sa_len;
3094 
3095 		tlen += PFKEY_ALIGN8(len);
3096 	}
3097     }
3098 
3099 	return tlen;
3100 }
3101 
3102 /*
3103  * SADB_SPDEXPIRE processing
3104  * send
3105  *   <base, address(SD), lifetime(CH), policy>
3106  * to KMD by PF_KEY.
3107  *
3108  * OUT:	0	: succeed
3109  *	others	: error number
3110  */
3111 static int
3112 key_spdexpire(struct secpolicy *sp)
3113 {
3114 	int s;
3115 	struct mbuf *result = NULL, *m;
3116 	int len;
3117 	int error = -1;
3118 	struct sadb_lifetime *lt;
3119 
3120 	/* XXX: Why do we lock ? */
3121 	s = splsoftnet();	/*called from softclock()*/
3122 
3123 	KASSERT(sp != NULL);
3124 
3125 	/* set msg header */
3126 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0, M_WAITOK);
3127 	result = m;
3128 
3129 	/* create lifetime extension (current and hard) */
3130 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
3131 	m = key_alloc_mbuf(len, M_WAITOK);
3132 	KASSERT(m->m_next == NULL);
3133 
3134 	memset(mtod(m, void *), 0, len);
3135 	lt = mtod(m, struct sadb_lifetime *);
3136 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
3137 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
3138 	lt->sadb_lifetime_allocations = 0;
3139 	lt->sadb_lifetime_bytes = 0;
3140 	lt->sadb_lifetime_addtime = time_mono_to_wall(sp->created);
3141 	lt->sadb_lifetime_usetime = time_mono_to_wall(sp->lastused);
3142 	lt = (struct sadb_lifetime *)(mtod(m, char *) + len / 2);
3143 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
3144 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
3145 	lt->sadb_lifetime_allocations = 0;
3146 	lt->sadb_lifetime_bytes = 0;
3147 	lt->sadb_lifetime_addtime = sp->lifetime;
3148 	lt->sadb_lifetime_usetime = sp->validtime;
3149 	m_cat(result, m);
3150 
3151 	/* set sadb_address for source */
3152 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sp->spidx.src.sa,
3153 	    sp->spidx.prefs, sp->spidx.ul_proto, M_WAITOK);
3154 	m_cat(result, m);
3155 
3156 	/* set sadb_address for destination */
3157 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sp->spidx.dst.sa,
3158 	    sp->spidx.prefd, sp->spidx.ul_proto, M_WAITOK);
3159 	m_cat(result, m);
3160 
3161 	/* set secpolicy */
3162 	m = key_sp2msg(sp, M_WAITOK);
3163 	m_cat(result, m);
3164 
3165 	KASSERT(result->m_flags & M_PKTHDR);
3166 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
3167 
3168 	result->m_pkthdr.len = 0;
3169 	for (m = result; m; m = m->m_next)
3170 		result->m_pkthdr.len += m->m_len;
3171 
3172 	mtod(result, struct sadb_msg *)->sadb_msg_len =
3173 	    PFKEY_UNIT64(result->m_pkthdr.len);
3174 
3175 	error = key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
3176 	splx(s);
3177 	return error;
3178 }
3179 
3180 /* %%% SAD management */
3181 /*
3182  * allocating a memory for new SA head, and copy from the values of mhp.
3183  * OUT:	NULL	: failure due to the lack of memory.
3184  *	others	: pointer to new SA head.
3185  */
3186 static struct secashead *
3187 key_newsah(const struct secasindex *saidx)
3188 {
3189 	struct secashead *newsah;
3190 	int i;
3191 
3192 	KASSERT(saidx != NULL);
3193 
3194 	newsah = kmem_zalloc(sizeof(struct secashead), KM_SLEEP);
3195 	for (i = 0; i < __arraycount(newsah->savlist); i++)
3196 		PSLIST_INIT(&newsah->savlist[i]);
3197 	newsah->saidx = *saidx;
3198 
3199 	localcount_init(&newsah->localcount);
3200 	/* Take a reference for the caller */
3201 	localcount_acquire(&newsah->localcount);
3202 
3203 	/* Add to the sah list */
3204 	SAHLIST_ENTRY_INIT(newsah);
3205 	newsah->state = SADB_SASTATE_MATURE;
3206 	mutex_enter(&key_sad.lock);
3207 	SAHLIST_WRITER_INSERT_HEAD(newsah);
3208 	mutex_exit(&key_sad.lock);
3209 
3210 	return newsah;
3211 }
3212 
3213 static bool
3214 key_sah_has_sav(struct secashead *sah)
3215 {
3216 	u_int state;
3217 
3218 	KASSERT(mutex_owned(&key_sad.lock));
3219 
3220 	SASTATE_ANY_FOREACH(state) {
3221 		if (!SAVLIST_WRITER_EMPTY(sah, state))
3222 			return true;
3223 	}
3224 
3225 	return false;
3226 }
3227 
3228 static void
3229 key_unlink_sah(struct secashead *sah)
3230 {
3231 
3232 	KASSERT(!cpu_softintr_p());
3233 	KASSERT(mutex_owned(&key_sad.lock));
3234 	KASSERT(sah->state == SADB_SASTATE_DEAD);
3235 
3236 	/* Remove from the sah list */
3237 	SAHLIST_WRITER_REMOVE(sah);
3238 
3239 	KDASSERT(mutex_ownable(softnet_lock));
3240 	key_sad_pserialize_perform();
3241 
3242 	localcount_drain(&sah->localcount, &key_sad.cv_lc, &key_sad.lock);
3243 }
3244 
3245 static void
3246 key_destroy_sah(struct secashead *sah)
3247 {
3248 
3249 	rtcache_free(&sah->sa_route);
3250 
3251 	SAHLIST_ENTRY_DESTROY(sah);
3252 	localcount_fini(&sah->localcount);
3253 
3254 	if (sah->idents != NULL)
3255 		kmem_free(sah->idents, sah->idents_len);
3256 	if (sah->identd != NULL)
3257 		kmem_free(sah->identd, sah->identd_len);
3258 
3259 	kmem_free(sah, sizeof(*sah));
3260 }
3261 
3262 /*
3263  * allocating a new SA with LARVAL state.
3264  * key_api_add() and key_api_getspi() call,
3265  * and copy the values of mhp into new buffer.
3266  * When SAD message type is GETSPI:
3267  *	to set sequence number from acq_seq++,
3268  *	to set zero to SPI.
3269  *	not to call key_setsaval().
3270  * OUT:	NULL	: fail
3271  *	others	: pointer to new secasvar.
3272  *
3273  * does not modify mbuf.  does not free mbuf on error.
3274  */
3275 static struct secasvar *
3276 key_newsav(struct mbuf *m, const struct sadb_msghdr *mhp,
3277     int *errp, const char* where, int tag)
3278 {
3279 	struct secasvar *newsav;
3280 	const struct sadb_sa *xsa;
3281 
3282 	KASSERT(!cpu_softintr_p());
3283 	KASSERT(m != NULL);
3284 	KASSERT(mhp != NULL);
3285 	KASSERT(mhp->msg != NULL);
3286 
3287 	newsav = kmem_zalloc(sizeof(struct secasvar), KM_SLEEP);
3288 
3289 	switch (mhp->msg->sadb_msg_type) {
3290 	case SADB_GETSPI:
3291 		newsav->spi = 0;
3292 
3293 #ifdef IPSEC_DOSEQCHECK
3294 		/* sync sequence number */
3295 		if (mhp->msg->sadb_msg_seq == 0)
3296 			newsav->seq =
3297 			    (acq_seq = (acq_seq == ~0 ? 1 : ++acq_seq));
3298 		else
3299 #endif
3300 			newsav->seq = mhp->msg->sadb_msg_seq;
3301 		break;
3302 
3303 	case SADB_ADD:
3304 		/* sanity check */
3305 		if (mhp->ext[SADB_EXT_SA] == NULL) {
3306 			IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
3307 			*errp = EINVAL;
3308 			goto error;
3309 		}
3310 		xsa = mhp->ext[SADB_EXT_SA];
3311 		newsav->spi = xsa->sadb_sa_spi;
3312 		newsav->seq = mhp->msg->sadb_msg_seq;
3313 		break;
3314 	default:
3315 		*errp = EINVAL;
3316 		goto error;
3317 	}
3318 
3319 	/* copy sav values */
3320 	if (mhp->msg->sadb_msg_type != SADB_GETSPI) {
3321 		*errp = key_setsaval(newsav, m, mhp);
3322 		if (*errp)
3323 			goto error;
3324 	} else {
3325 		/* We don't allow lft_c to be NULL */
3326 		newsav->lft_c = kmem_zalloc(sizeof(struct sadb_lifetime),
3327 		    KM_SLEEP);
3328 		newsav->lft_c_counters_percpu =
3329 		    percpu_alloc(sizeof(lifetime_counters_t));
3330 	}
3331 
3332 	/* reset created */
3333 	newsav->created = time_uptime;
3334 	newsav->pid = mhp->msg->sadb_msg_pid;
3335 
3336 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
3337 	    "DP from %s:%u return SA:%p\n", where, tag, newsav);
3338 	return newsav;
3339 
3340 error:
3341 	KASSERT(*errp != 0);
3342 	kmem_free(newsav, sizeof(*newsav));
3343 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
3344 	    "DP from %s:%u return SA:NULL\n", where, tag);
3345 	return NULL;
3346 }
3347 
3348 
3349 static void
3350 key_clear_xform(struct secasvar *sav)
3351 {
3352 
3353 	/*
3354 	 * Cleanup xform state.  Note that zeroize'ing causes the
3355 	 * keys to be cleared; otherwise we must do it ourself.
3356 	 */
3357 	if (sav->tdb_xform != NULL) {
3358 		sav->tdb_xform->xf_zeroize(sav);
3359 		sav->tdb_xform = NULL;
3360 	} else {
3361 		if (sav->key_auth != NULL)
3362 			explicit_memset(_KEYBUF(sav->key_auth), 0,
3363 			    _KEYLEN(sav->key_auth));
3364 		if (sav->key_enc != NULL)
3365 			explicit_memset(_KEYBUF(sav->key_enc), 0,
3366 			    _KEYLEN(sav->key_enc));
3367 	}
3368 }
3369 
3370 /*
3371  * free() SA variable entry.
3372  */
3373 static void
3374 key_delsav(struct secasvar *sav)
3375 {
3376 
3377 	key_clear_xform(sav);
3378 	key_freesaval(sav);
3379 	kmem_free(sav, sizeof(*sav));
3380 }
3381 
3382 /*
3383  * Must be called in a pserialize read section. A held sah
3384  * must be released by key_sah_unref after use.
3385  */
3386 static void
3387 key_sah_ref(struct secashead *sah)
3388 {
3389 
3390 	localcount_acquire(&sah->localcount);
3391 }
3392 
3393 /*
3394  * Must be called without holding key_sad.lock because the lock
3395  * would be held in localcount_release.
3396  */
3397 static void
3398 key_sah_unref(struct secashead *sah)
3399 {
3400 
3401 	KDASSERT(mutex_ownable(&key_sad.lock));
3402 
3403 	localcount_release(&sah->localcount, &key_sad.cv_lc, &key_sad.lock);
3404 }
3405 
3406 /*
3407  * Search SAD and return sah. Must be called in a pserialize
3408  * read section.
3409  * OUT:
3410  *	NULL	: not found
3411  *	others	: found, pointer to a SA.
3412  */
3413 static struct secashead *
3414 key_getsah(const struct secasindex *saidx, int flag)
3415 {
3416 	struct secashead *sah;
3417 
3418 	SAHLIST_READER_FOREACH_SAIDX(sah, saidx) {
3419 		if (sah->state == SADB_SASTATE_DEAD)
3420 			continue;
3421 		if (key_saidx_match(&sah->saidx, saidx, flag))
3422 			return sah;
3423 	}
3424 
3425 	return NULL;
3426 }
3427 
3428 /*
3429  * Search SAD and return sah. If sah is returned, the caller must call
3430  * key_sah_unref to releaset a reference.
3431  * OUT:
3432  *	NULL	: not found
3433  *	others	: found, pointer to a SA.
3434  */
3435 static struct secashead *
3436 key_getsah_ref(const struct secasindex *saidx, int flag)
3437 {
3438 	struct secashead *sah;
3439 	int s;
3440 
3441 	s = pserialize_read_enter();
3442 	sah = key_getsah(saidx, flag);
3443 	if (sah != NULL)
3444 		key_sah_ref(sah);
3445 	pserialize_read_exit(s);
3446 
3447 	return sah;
3448 }
3449 
3450 /*
3451  * check not to be duplicated SPI.
3452  * NOTE: this function is too slow due to searching all SAD.
3453  * OUT:
3454  *	NULL	: not found
3455  *	others	: found, pointer to a SA.
3456  */
3457 static bool
3458 key_checkspidup(const struct secasindex *saidx, u_int32_t spi)
3459 {
3460 	struct secashead *sah;
3461 	struct secasvar *sav;
3462 
3463 	/* check address family */
3464 	if (saidx->src.sa.sa_family != saidx->dst.sa.sa_family) {
3465 		IPSECLOG(LOG_DEBUG,
3466 		    "address family mismatched src %u, dst %u.\n",
3467 		    saidx->src.sa.sa_family, saidx->dst.sa.sa_family);
3468 		return false;
3469 	}
3470 
3471 	/* check all SAD */
3472 	/* key_ismyaddr may sleep, so use mutex, not pserialize, here. */
3473 	mutex_enter(&key_sad.lock);
3474 	SAHLIST_WRITER_FOREACH(sah) {
3475 		if (!key_ismyaddr((struct sockaddr *)&sah->saidx.dst))
3476 			continue;
3477 		sav = key_getsavbyspi(sah, spi);
3478 		if (sav != NULL) {
3479 			KEY_SA_UNREF(&sav);
3480 			mutex_exit(&key_sad.lock);
3481 			return true;
3482 		}
3483 	}
3484 	mutex_exit(&key_sad.lock);
3485 
3486 	return false;
3487 }
3488 
3489 /*
3490  * search SAD litmited alive SA, protocol, SPI.
3491  * OUT:
3492  *	NULL	: not found
3493  *	others	: found, pointer to a SA.
3494  */
3495 static struct secasvar *
3496 key_getsavbyspi(struct secashead *sah, u_int32_t spi)
3497 {
3498 	struct secasvar *sav = NULL;
3499 	u_int state;
3500 	int s;
3501 
3502 	/* search all status */
3503 	s = pserialize_read_enter();
3504 	SASTATE_ALIVE_FOREACH(state) {
3505 		SAVLIST_READER_FOREACH(sav, sah, state) {
3506 			/* sanity check */
3507 			if (sav->state != state) {
3508 				IPSECLOG(LOG_DEBUG,
3509 				    "invalid sav->state (queue: %d SA: %d)\n",
3510 				    state, sav->state);
3511 				continue;
3512 			}
3513 
3514 			if (sav->spi == spi) {
3515 				KEY_SA_REF(sav);
3516 				goto out;
3517 			}
3518 		}
3519 	}
3520 out:
3521 	pserialize_read_exit(s);
3522 
3523 	return sav;
3524 }
3525 
3526 /*
3527  * Search SAD litmited alive SA by an SPI and remove it from a list.
3528  * OUT:
3529  *	NULL	: not found
3530  *	others	: found, pointer to a SA.
3531  */
3532 static struct secasvar *
3533 key_lookup_and_remove_sav(struct secashead *sah, u_int32_t spi,
3534     const struct secasvar *hint)
3535 {
3536 	struct secasvar *sav = NULL;
3537 	u_int state;
3538 
3539 	/* search all status */
3540 	mutex_enter(&key_sad.lock);
3541 	SASTATE_ALIVE_FOREACH(state) {
3542 		SAVLIST_WRITER_FOREACH(sav, sah, state) {
3543 			KASSERT(sav->state == state);
3544 
3545 			if (sav->spi == spi) {
3546 				if (hint != NULL && hint != sav)
3547 					continue;
3548 				sav->state = SADB_SASTATE_DEAD;
3549 				SAVLIST_WRITER_REMOVE(sav);
3550 				SAVLUT_WRITER_REMOVE(sav);
3551 				goto out;
3552 			}
3553 		}
3554 	}
3555 out:
3556 	mutex_exit(&key_sad.lock);
3557 
3558 	return sav;
3559 }
3560 
3561 /*
3562  * Free allocated data to member variables of sav:
3563  * sav->replay, sav->key_* and sav->lft_*.
3564  */
3565 static void
3566 key_freesaval(struct secasvar *sav)
3567 {
3568 
3569 	KASSERT(key_sa_refcnt(sav) == 0);
3570 
3571 	if (sav->replay != NULL)
3572 		kmem_intr_free(sav->replay, sav->replay_len);
3573 	if (sav->key_auth != NULL)
3574 		kmem_intr_free(sav->key_auth, sav->key_auth_len);
3575 	if (sav->key_enc != NULL)
3576 		kmem_intr_free(sav->key_enc, sav->key_enc_len);
3577 	if (sav->lft_c_counters_percpu != NULL) {
3578 		percpu_free(sav->lft_c_counters_percpu,
3579 		    sizeof(lifetime_counters_t));
3580 	}
3581 	if (sav->lft_c != NULL)
3582 		kmem_intr_free(sav->lft_c, sizeof(*(sav->lft_c)));
3583 	if (sav->lft_h != NULL)
3584 		kmem_intr_free(sav->lft_h, sizeof(*(sav->lft_h)));
3585 	if (sav->lft_s != NULL)
3586 		kmem_intr_free(sav->lft_s, sizeof(*(sav->lft_s)));
3587 }
3588 
3589 /*
3590  * copy SA values from PF_KEY message except *SPI, SEQ, PID, STATE and TYPE*.
3591  * You must update these if need.
3592  * OUT:	0:	success.
3593  *	!0:	failure.
3594  *
3595  * does not modify mbuf.  does not free mbuf on error.
3596  */
3597 static int
3598 key_setsaval(struct secasvar *sav, struct mbuf *m,
3599 	     const struct sadb_msghdr *mhp)
3600 {
3601 	int error = 0;
3602 
3603 	KASSERT(!cpu_softintr_p());
3604 	KASSERT(m != NULL);
3605 	KASSERT(mhp != NULL);
3606 	KASSERT(mhp->msg != NULL);
3607 
3608 	/* We shouldn't initialize sav variables while someone uses it. */
3609 	KASSERT(key_sa_refcnt(sav) == 0);
3610 
3611 	/* SA */
3612 	if (mhp->ext[SADB_EXT_SA] != NULL) {
3613 		const struct sadb_sa *sa0;
3614 
3615 		sa0 = mhp->ext[SADB_EXT_SA];
3616 		if (mhp->extlen[SADB_EXT_SA] < sizeof(*sa0)) {
3617 			error = EINVAL;
3618 			goto fail;
3619 		}
3620 
3621 		sav->alg_auth = sa0->sadb_sa_auth;
3622 		sav->alg_enc = sa0->sadb_sa_encrypt;
3623 		sav->flags = sa0->sadb_sa_flags;
3624 
3625 		/* replay window */
3626 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0) {
3627 			size_t len = sizeof(struct secreplay) +
3628 			    sa0->sadb_sa_replay;
3629 			sav->replay = kmem_zalloc(len, KM_SLEEP);
3630 			sav->replay_len = len;
3631 			if (sa0->sadb_sa_replay != 0)
3632 				sav->replay->bitmap = (char*)(sav->replay+1);
3633 			sav->replay->wsize = sa0->sadb_sa_replay;
3634 		}
3635 	}
3636 
3637 	/* Authentication keys */
3638 	if (mhp->ext[SADB_EXT_KEY_AUTH] != NULL) {
3639 		const struct sadb_key *key0;
3640 		int len;
3641 
3642 		key0 = mhp->ext[SADB_EXT_KEY_AUTH];
3643 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3644 
3645 		error = 0;
3646 		if (len < sizeof(*key0)) {
3647 			error = EINVAL;
3648 			goto fail;
3649 		}
3650 		switch (mhp->msg->sadb_msg_satype) {
3651 		case SADB_SATYPE_AH:
3652 		case SADB_SATYPE_ESP:
3653 		case SADB_X_SATYPE_TCPSIGNATURE:
3654 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3655 			    sav->alg_auth != SADB_X_AALG_NULL)
3656 				error = EINVAL;
3657 			break;
3658 		case SADB_X_SATYPE_IPCOMP:
3659 		default:
3660 			error = EINVAL;
3661 			break;
3662 		}
3663 		if (error) {
3664 			IPSECLOG(LOG_DEBUG, "invalid key_auth values.\n");
3665 			goto fail;
3666 		}
3667 
3668 		sav->key_auth = key_newbuf(key0, len);
3669 		sav->key_auth_len = len;
3670 	}
3671 
3672 	/* Encryption key */
3673 	if (mhp->ext[SADB_EXT_KEY_ENCRYPT] != NULL) {
3674 		const struct sadb_key *key0;
3675 		int len;
3676 
3677 		key0 = mhp->ext[SADB_EXT_KEY_ENCRYPT];
3678 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3679 
3680 		error = 0;
3681 		if (len < sizeof(*key0)) {
3682 			error = EINVAL;
3683 			goto fail;
3684 		}
3685 		switch (mhp->msg->sadb_msg_satype) {
3686 		case SADB_SATYPE_ESP:
3687 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3688 			    sav->alg_enc != SADB_EALG_NULL) {
3689 				error = EINVAL;
3690 				break;
3691 			}
3692 			sav->key_enc = key_newbuf(key0, len);
3693 			sav->key_enc_len = len;
3694 			break;
3695 		case SADB_X_SATYPE_IPCOMP:
3696 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3697 				error = EINVAL;
3698 			sav->key_enc = NULL;	/*just in case*/
3699 			break;
3700 		case SADB_SATYPE_AH:
3701 		case SADB_X_SATYPE_TCPSIGNATURE:
3702 		default:
3703 			error = EINVAL;
3704 			break;
3705 		}
3706 		if (error) {
3707 			IPSECLOG(LOG_DEBUG, "invalid key_enc value.\n");
3708 			goto fail;
3709 		}
3710 	}
3711 
3712 	/* set iv */
3713 	sav->ivlen = 0;
3714 
3715 	switch (mhp->msg->sadb_msg_satype) {
3716 	case SADB_SATYPE_AH:
3717 		error = xform_init(sav, XF_AH);
3718 		break;
3719 	case SADB_SATYPE_ESP:
3720 		error = xform_init(sav, XF_ESP);
3721 		break;
3722 	case SADB_X_SATYPE_IPCOMP:
3723 		error = xform_init(sav, XF_IPCOMP);
3724 		break;
3725 	case SADB_X_SATYPE_TCPSIGNATURE:
3726 		error = xform_init(sav, XF_TCPSIGNATURE);
3727 		break;
3728 	default:
3729 		error = EOPNOTSUPP;
3730 		break;
3731 	}
3732 	if (error) {
3733 		IPSECLOG(LOG_DEBUG, "unable to initialize SA type %u (%d)\n",
3734 		    mhp->msg->sadb_msg_satype, error);
3735 		goto fail;
3736 	}
3737 
3738 	/* reset created */
3739 	sav->created = time_uptime;
3740 
3741 	/* make lifetime for CURRENT */
3742 	sav->lft_c = kmem_alloc(sizeof(struct sadb_lifetime), KM_SLEEP);
3743 
3744 	sav->lft_c->sadb_lifetime_len =
3745 	    PFKEY_UNIT64(sizeof(struct sadb_lifetime));
3746 	sav->lft_c->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
3747 	sav->lft_c->sadb_lifetime_allocations = 0;
3748 	sav->lft_c->sadb_lifetime_bytes = 0;
3749 	sav->lft_c->sadb_lifetime_addtime = time_uptime;
3750 	sav->lft_c->sadb_lifetime_usetime = 0;
3751 
3752 	sav->lft_c_counters_percpu = percpu_alloc(sizeof(lifetime_counters_t));
3753 
3754 	/* lifetimes for HARD and SOFT */
3755     {
3756 	const struct sadb_lifetime *lft0;
3757 
3758 	lft0 = mhp->ext[SADB_EXT_LIFETIME_HARD];
3759 	if (lft0 != NULL) {
3760 		if (mhp->extlen[SADB_EXT_LIFETIME_HARD] < sizeof(*lft0)) {
3761 			error = EINVAL;
3762 			goto fail;
3763 		}
3764 		sav->lft_h = key_newbuf(lft0, sizeof(*lft0));
3765 	}
3766 
3767 	lft0 = mhp->ext[SADB_EXT_LIFETIME_SOFT];
3768 	if (lft0 != NULL) {
3769 		if (mhp->extlen[SADB_EXT_LIFETIME_SOFT] < sizeof(*lft0)) {
3770 			error = EINVAL;
3771 			goto fail;
3772 		}
3773 		sav->lft_s = key_newbuf(lft0, sizeof(*lft0));
3774 		/* to be initialize ? */
3775 	}
3776     }
3777 
3778 	return 0;
3779 
3780  fail:
3781 	key_clear_xform(sav);
3782 	key_freesaval(sav);
3783 
3784 	return error;
3785 }
3786 
3787 /*
3788  * validation with a secasvar entry, and set SADB_SATYPE_MATURE.
3789  * OUT:	0:	valid
3790  *	other:	errno
3791  */
3792 static int
3793 key_init_xform(struct secasvar *sav)
3794 {
3795 	int error;
3796 
3797 	/* We shouldn't initialize sav variables while someone uses it. */
3798 	KASSERT(key_sa_refcnt(sav) == 0);
3799 
3800 	/* check SPI value */
3801 	switch (sav->sah->saidx.proto) {
3802 	case IPPROTO_ESP:
3803 	case IPPROTO_AH:
3804 		if (ntohl(sav->spi) <= 255) {
3805 			IPSECLOG(LOG_DEBUG, "illegal range of SPI %u.\n",
3806 			    (u_int32_t)ntohl(sav->spi));
3807 			return EINVAL;
3808 		}
3809 		break;
3810 	}
3811 
3812 	/* check algo */
3813 	switch (sav->sah->saidx.proto) {
3814 	case IPPROTO_AH:
3815 	case IPPROTO_TCP:
3816 		if (sav->alg_enc != SADB_EALG_NONE) {
3817 			IPSECLOG(LOG_DEBUG,
3818 			    "protocol %u and algorithm mismatched %u != %u.\n",
3819 			    sav->sah->saidx.proto,
3820 			    sav->alg_enc, SADB_EALG_NONE);
3821 			return EINVAL;
3822 		}
3823 		break;
3824 	case IPPROTO_IPCOMP:
3825 		if (sav->alg_auth != SADB_AALG_NONE) {
3826 			IPSECLOG(LOG_DEBUG,
3827 			    "protocol %u and algorithm mismatched %d != %d.\n",
3828 			    sav->sah->saidx.proto,
3829 			    sav->alg_auth, SADB_AALG_NONE);
3830 			return(EINVAL);
3831 		}
3832 		break;
3833 	default:
3834 		break;
3835 	}
3836 
3837 	/* check satype */
3838 	switch (sav->sah->saidx.proto) {
3839 	case IPPROTO_ESP:
3840 		/* check flags */
3841 		if ((sav->flags & (SADB_X_EXT_OLD|SADB_X_EXT_DERIV)) ==
3842 		    (SADB_X_EXT_OLD|SADB_X_EXT_DERIV)) {
3843 			IPSECLOG(LOG_DEBUG,
3844 			    "invalid flag (derived) given to old-esp.\n");
3845 			return EINVAL;
3846 		}
3847 		error = xform_init(sav, XF_ESP);
3848 		break;
3849 	case IPPROTO_AH:
3850 		/* check flags */
3851 		if (sav->flags & SADB_X_EXT_DERIV) {
3852 			IPSECLOG(LOG_DEBUG,
3853 			    "invalid flag (derived) given to AH SA.\n");
3854 			return EINVAL;
3855 		}
3856 		error = xform_init(sav, XF_AH);
3857 		break;
3858 	case IPPROTO_IPCOMP:
3859 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0
3860 		    && ntohl(sav->spi) >= 0x10000) {
3861 			IPSECLOG(LOG_DEBUG, "invalid cpi for IPComp.\n");
3862 			return(EINVAL);
3863 		}
3864 		error = xform_init(sav, XF_IPCOMP);
3865 		break;
3866 	case IPPROTO_TCP:
3867 		error = xform_init(sav, XF_TCPSIGNATURE);
3868 		break;
3869 	default:
3870 		IPSECLOG(LOG_DEBUG, "Invalid satype.\n");
3871 		error = EPROTONOSUPPORT;
3872 		break;
3873 	}
3874 
3875 	return error;
3876 }
3877 
3878 /*
3879  * subroutine for SADB_GET and SADB_DUMP. It never return NULL.
3880  */
3881 static struct mbuf *
3882 key_setdumpsa(struct secasvar *sav, u_int8_t type, u_int8_t satype,
3883 	      u_int32_t seq, u_int32_t pid)
3884 {
3885 	struct mbuf *result = NULL, *tres = NULL, *m;
3886 	int l = 0;
3887 	int i;
3888 	void *p;
3889 	struct sadb_lifetime lt;
3890 	int dumporder[] = {
3891 		SADB_EXT_SA, SADB_X_EXT_SA2,
3892 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3893 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3894 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY, SADB_EXT_KEY_AUTH,
3895 		SADB_EXT_KEY_ENCRYPT, SADB_EXT_IDENTITY_SRC,
3896 		SADB_EXT_IDENTITY_DST, SADB_EXT_SENSITIVITY,
3897 		SADB_X_EXT_NAT_T_TYPE,
3898 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3899 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3900 		SADB_X_EXT_NAT_T_FRAG,
3901 
3902 	};
3903 
3904 	m = key_setsadbmsg(type, 0, satype, seq, pid, key_sa_refcnt(sav), M_WAITOK);
3905 	result = m;
3906 
3907 	for (i = __arraycount(dumporder) - 1; i >= 0; i--) {
3908 		m = NULL;
3909 		p = NULL;
3910 		switch (dumporder[i]) {
3911 		case SADB_EXT_SA:
3912 			m = key_setsadbsa(sav);
3913 			break;
3914 
3915 		case SADB_X_EXT_SA2:
3916 			m = key_setsadbxsa2(sav->sah->saidx.mode,
3917 			    sav->replay ? sav->replay->count : 0,
3918 			    sav->sah->saidx.reqid);
3919 			break;
3920 
3921 		case SADB_EXT_ADDRESS_SRC:
3922 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3923 			    &sav->sah->saidx.src.sa,
3924 			    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
3925 			break;
3926 
3927 		case SADB_EXT_ADDRESS_DST:
3928 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3929 			    &sav->sah->saidx.dst.sa,
3930 			    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
3931 			break;
3932 
3933 		case SADB_EXT_KEY_AUTH:
3934 			if (!sav->key_auth)
3935 				continue;
3936 			l = PFKEY_UNUNIT64(sav->key_auth->sadb_key_len);
3937 			p = sav->key_auth;
3938 			break;
3939 
3940 		case SADB_EXT_KEY_ENCRYPT:
3941 			if (!sav->key_enc)
3942 				continue;
3943 			l = PFKEY_UNUNIT64(sav->key_enc->sadb_key_len);
3944 			p = sav->key_enc;
3945 			break;
3946 
3947 		case SADB_EXT_LIFETIME_CURRENT: {
3948 			lifetime_counters_t sum = {0};
3949 
3950 			KASSERT(sav->lft_c != NULL);
3951 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_c)->sadb_ext_len);
3952 			memcpy(&lt, sav->lft_c, sizeof(struct sadb_lifetime));
3953 			lt.sadb_lifetime_addtime =
3954 			    time_mono_to_wall(lt.sadb_lifetime_addtime);
3955 			lt.sadb_lifetime_usetime =
3956 			    time_mono_to_wall(lt.sadb_lifetime_usetime);
3957 			percpu_foreach(sav->lft_c_counters_percpu,
3958 			    key_sum_lifetime_counters, sum);
3959 			lt.sadb_lifetime_allocations =
3960 			    sum[LIFETIME_COUNTER_ALLOCATIONS];
3961 			lt.sadb_lifetime_bytes =
3962 			    sum[LIFETIME_COUNTER_BYTES];
3963 			p = &lt;
3964 			break;
3965 		    }
3966 
3967 		case SADB_EXT_LIFETIME_HARD:
3968 			if (!sav->lft_h)
3969 				continue;
3970 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_h)->sadb_ext_len);
3971 			p = sav->lft_h;
3972 			break;
3973 
3974 		case SADB_EXT_LIFETIME_SOFT:
3975 			if (!sav->lft_s)
3976 				continue;
3977 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_s)->sadb_ext_len);
3978 			p = sav->lft_s;
3979 			break;
3980 
3981 		case SADB_X_EXT_NAT_T_TYPE:
3982 			m = key_setsadbxtype(sav->natt_type);
3983 			break;
3984 
3985 		case SADB_X_EXT_NAT_T_DPORT:
3986 			if (sav->natt_type == 0)
3987 				continue;
3988 			m = key_setsadbxport(
3989 			    key_portfromsaddr(&sav->sah->saidx.dst),
3990 			    SADB_X_EXT_NAT_T_DPORT);
3991 			break;
3992 
3993 		case SADB_X_EXT_NAT_T_SPORT:
3994 			if (sav->natt_type == 0)
3995 				continue;
3996 			m = key_setsadbxport(
3997 			    key_portfromsaddr(&sav->sah->saidx.src),
3998 			    SADB_X_EXT_NAT_T_SPORT);
3999 			break;
4000 
4001 		case SADB_X_EXT_NAT_T_FRAG:
4002 			/* don't send frag info if not set */
4003 			if (sav->natt_type == 0 || sav->esp_frag == IP_MAXPACKET)
4004 				continue;
4005 			m = key_setsadbxfrag(sav->esp_frag);
4006 			break;
4007 
4008 		case SADB_X_EXT_NAT_T_OAI:
4009 		case SADB_X_EXT_NAT_T_OAR:
4010 			continue;
4011 
4012 		case SADB_EXT_ADDRESS_PROXY:
4013 		case SADB_EXT_IDENTITY_SRC:
4014 		case SADB_EXT_IDENTITY_DST:
4015 			/* XXX: should we brought from SPD ? */
4016 		case SADB_EXT_SENSITIVITY:
4017 		default:
4018 			continue;
4019 		}
4020 
4021 		KASSERT(!(m && p));
4022 		KASSERT(m != NULL || p != NULL);
4023 		if (p && tres) {
4024 			M_PREPEND(tres, l, M_WAITOK);
4025 			memcpy(mtod(tres, void *), p, l);
4026 			continue;
4027 		}
4028 		if (p) {
4029 			m = key_alloc_mbuf(l, M_WAITOK);
4030 			m_copyback(m, 0, l, p);
4031 		}
4032 
4033 		if (tres)
4034 			m_cat(m, tres);
4035 		tres = m;
4036 	}
4037 
4038 	m_cat(result, tres);
4039 	tres = NULL; /* avoid free on error below */
4040 
4041 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
4042 
4043 	result->m_pkthdr.len = 0;
4044 	for (m = result; m; m = m->m_next)
4045 		result->m_pkthdr.len += m->m_len;
4046 
4047 	mtod(result, struct sadb_msg *)->sadb_msg_len =
4048 	    PFKEY_UNIT64(result->m_pkthdr.len);
4049 
4050 	return result;
4051 }
4052 
4053 
4054 /*
4055  * set a type in sadb_x_nat_t_type
4056  */
4057 static struct mbuf *
4058 key_setsadbxtype(u_int16_t type)
4059 {
4060 	struct mbuf *m;
4061 	size_t len;
4062 	struct sadb_x_nat_t_type *p;
4063 
4064 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
4065 
4066 	m = key_alloc_mbuf(len, M_WAITOK);
4067 	KASSERT(m->m_next == NULL);
4068 
4069 	p = mtod(m, struct sadb_x_nat_t_type *);
4070 
4071 	memset(p, 0, len);
4072 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
4073 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
4074 	p->sadb_x_nat_t_type_type = type;
4075 
4076 	return m;
4077 }
4078 /*
4079  * set a port in sadb_x_nat_t_port. port is in network order
4080  */
4081 static struct mbuf *
4082 key_setsadbxport(u_int16_t port, u_int16_t type)
4083 {
4084 	struct mbuf *m;
4085 	size_t len;
4086 	struct sadb_x_nat_t_port *p;
4087 
4088 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
4089 
4090 	m = key_alloc_mbuf(len, M_WAITOK);
4091 	KASSERT(m->m_next == NULL);
4092 
4093 	p = mtod(m, struct sadb_x_nat_t_port *);
4094 
4095 	memset(p, 0, len);
4096 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
4097 	p->sadb_x_nat_t_port_exttype = type;
4098 	p->sadb_x_nat_t_port_port = port;
4099 
4100 	return m;
4101 }
4102 
4103 /*
4104  * set fragmentation info in sadb_x_nat_t_frag
4105  */
4106 static struct mbuf *
4107 key_setsadbxfrag(u_int16_t flen)
4108 {
4109 	struct mbuf *m;
4110 	size_t len;
4111 	struct sadb_x_nat_t_frag *p;
4112 
4113 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_frag));
4114 
4115 	m = key_alloc_mbuf(len, M_WAITOK);
4116 	KASSERT(m->m_next == NULL);
4117 
4118 	p = mtod(m, struct sadb_x_nat_t_frag *);
4119 
4120 	memset(p, 0, len);
4121 	p->sadb_x_nat_t_frag_len = PFKEY_UNIT64(len);
4122 	p->sadb_x_nat_t_frag_exttype = SADB_X_EXT_NAT_T_FRAG;
4123 	p->sadb_x_nat_t_frag_fraglen = flen;
4124 
4125 	return m;
4126 }
4127 
4128 /*
4129  * Get port from sockaddr, port is in network order
4130  */
4131 u_int16_t
4132 key_portfromsaddr(const union sockaddr_union *saddr)
4133 {
4134 	u_int16_t port;
4135 
4136 	switch (saddr->sa.sa_family) {
4137 	case AF_INET: {
4138 		port = saddr->sin.sin_port;
4139 		break;
4140 	}
4141 #ifdef INET6
4142 	case AF_INET6: {
4143 		port = saddr->sin6.sin6_port;
4144 		break;
4145 	}
4146 #endif
4147 	default:
4148 		printf("%s: unexpected address family\n", __func__);
4149 		port = 0;
4150 		break;
4151 	}
4152 
4153 	return port;
4154 }
4155 
4156 
4157 /*
4158  * Set port is struct sockaddr. port is in network order
4159  */
4160 static void
4161 key_porttosaddr(union sockaddr_union *saddr, u_int16_t port)
4162 {
4163 	switch (saddr->sa.sa_family) {
4164 	case AF_INET: {
4165 		saddr->sin.sin_port = port;
4166 		break;
4167 	}
4168 #ifdef INET6
4169 	case AF_INET6: {
4170 		saddr->sin6.sin6_port = port;
4171 		break;
4172 	}
4173 #endif
4174 	default:
4175 		printf("%s: unexpected address family %d\n", __func__,
4176 		    saddr->sa.sa_family);
4177 		break;
4178 	}
4179 
4180 	return;
4181 }
4182 
4183 /*
4184  * Safety check sa_len
4185  */
4186 static int
4187 key_checksalen(const union sockaddr_union *saddr)
4188 {
4189 	switch (saddr->sa.sa_family) {
4190 	case AF_INET:
4191 		if (saddr->sa.sa_len != sizeof(struct sockaddr_in))
4192 			return -1;
4193 		break;
4194 #ifdef INET6
4195 	case AF_INET6:
4196 		if (saddr->sa.sa_len != sizeof(struct sockaddr_in6))
4197 			return -1;
4198 		break;
4199 #endif
4200 	default:
4201 		printf("%s: unexpected sa_family %d\n", __func__,
4202 		    saddr->sa.sa_family);
4203 			return -1;
4204 		break;
4205 	}
4206 	return 0;
4207 }
4208 
4209 
4210 /*
4211  * set data into sadb_msg.
4212  */
4213 static struct mbuf *
4214 key_setsadbmsg(u_int8_t type,  u_int16_t tlen, u_int8_t satype,
4215 	       u_int32_t seq, pid_t pid, u_int16_t reserved, int mflag)
4216 {
4217 	struct mbuf *m;
4218 	struct sadb_msg *p;
4219 	int len;
4220 
4221 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MCLBYTES);
4222 
4223 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
4224 
4225 	m = key_alloc_mbuf_simple(len, mflag);
4226 	if (!m)
4227 		return NULL;
4228 	m->m_pkthdr.len = m->m_len = len;
4229 	m->m_next = NULL;
4230 
4231 	p = mtod(m, struct sadb_msg *);
4232 
4233 	memset(p, 0, len);
4234 	p->sadb_msg_version = PF_KEY_V2;
4235 	p->sadb_msg_type = type;
4236 	p->sadb_msg_errno = 0;
4237 	p->sadb_msg_satype = satype;
4238 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
4239 	p->sadb_msg_reserved = reserved;
4240 	p->sadb_msg_seq = seq;
4241 	p->sadb_msg_pid = (u_int32_t)pid;
4242 
4243 	return m;
4244 }
4245 
4246 /*
4247  * copy secasvar data into sadb_address.
4248  */
4249 static struct mbuf *
4250 key_setsadbsa(struct secasvar *sav)
4251 {
4252 	struct mbuf *m;
4253 	struct sadb_sa *p;
4254 	int len;
4255 
4256 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
4257 	m = key_alloc_mbuf(len, M_WAITOK);
4258 	KASSERT(m->m_next == NULL);
4259 
4260 	p = mtod(m, struct sadb_sa *);
4261 
4262 	memset(p, 0, len);
4263 	p->sadb_sa_len = PFKEY_UNIT64(len);
4264 	p->sadb_sa_exttype = SADB_EXT_SA;
4265 	p->sadb_sa_spi = sav->spi;
4266 	p->sadb_sa_replay = (sav->replay != NULL ? sav->replay->wsize : 0);
4267 	p->sadb_sa_state = sav->state;
4268 	p->sadb_sa_auth = sav->alg_auth;
4269 	p->sadb_sa_encrypt = sav->alg_enc;
4270 	p->sadb_sa_flags = sav->flags;
4271 
4272 	return m;
4273 }
4274 
4275 static uint8_t
4276 key_sabits(const struct sockaddr *saddr)
4277 {
4278 	switch (saddr->sa_family) {
4279 	case AF_INET:
4280 		return _BITS(sizeof(struct in_addr));
4281 	case AF_INET6:
4282 		return _BITS(sizeof(struct in6_addr));
4283 	default:
4284 		return FULLMASK;
4285 	}
4286 }
4287 
4288 /*
4289  * set data into sadb_address.
4290  */
4291 static struct mbuf *
4292 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
4293 		u_int8_t prefixlen, u_int16_t ul_proto, int mflag)
4294 {
4295 	struct mbuf *m;
4296 	struct sadb_address *p;
4297 	size_t len;
4298 
4299 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
4300 	    PFKEY_ALIGN8(saddr->sa_len);
4301 	m = key_alloc_mbuf(len, mflag);
4302 	if (!m || m->m_next) {	/*XXX*/
4303 		if (m)
4304 			m_freem(m);
4305 		return NULL;
4306 	}
4307 
4308 	p = mtod(m, struct sadb_address *);
4309 
4310 	memset(p, 0, len);
4311 	p->sadb_address_len = PFKEY_UNIT64(len);
4312 	p->sadb_address_exttype = exttype;
4313 	p->sadb_address_proto = ul_proto;
4314 	if (prefixlen == FULLMASK) {
4315 		prefixlen = key_sabits(saddr);
4316 	}
4317 	p->sadb_address_prefixlen = prefixlen;
4318 	p->sadb_address_reserved = 0;
4319 
4320 	memcpy(mtod(m, char *) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
4321 	    saddr, saddr->sa_len);
4322 
4323 	return m;
4324 }
4325 
4326 #if 0
4327 /*
4328  * set data into sadb_ident.
4329  */
4330 static struct mbuf *
4331 key_setsadbident(u_int16_t exttype, u_int16_t idtype,
4332 		 void *string, int stringlen, u_int64_t id)
4333 {
4334 	struct mbuf *m;
4335 	struct sadb_ident *p;
4336 	size_t len;
4337 
4338 	len = PFKEY_ALIGN8(sizeof(struct sadb_ident)) + PFKEY_ALIGN8(stringlen);
4339 	m = key_alloc_mbuf(len);
4340 	if (!m || m->m_next) {	/*XXX*/
4341 		if (m)
4342 			m_freem(m);
4343 		return NULL;
4344 	}
4345 
4346 	p = mtod(m, struct sadb_ident *);
4347 
4348 	memset(p, 0, len);
4349 	p->sadb_ident_len = PFKEY_UNIT64(len);
4350 	p->sadb_ident_exttype = exttype;
4351 	p->sadb_ident_type = idtype;
4352 	p->sadb_ident_reserved = 0;
4353 	p->sadb_ident_id = id;
4354 
4355 	memcpy(mtod(m, void *) + PFKEY_ALIGN8(sizeof(struct sadb_ident)),
4356 	   	   string, stringlen);
4357 
4358 	return m;
4359 }
4360 #endif
4361 
4362 /*
4363  * set data into sadb_x_sa2.
4364  */
4365 static struct mbuf *
4366 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int16_t reqid)
4367 {
4368 	struct mbuf *m;
4369 	struct sadb_x_sa2 *p;
4370 	size_t len;
4371 
4372 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
4373 	m = key_alloc_mbuf(len, M_WAITOK);
4374 	KASSERT(m->m_next == NULL);
4375 
4376 	p = mtod(m, struct sadb_x_sa2 *);
4377 
4378 	memset(p, 0, len);
4379 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
4380 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
4381 	p->sadb_x_sa2_mode = mode;
4382 	p->sadb_x_sa2_reserved1 = 0;
4383 	p->sadb_x_sa2_reserved2 = 0;
4384 	p->sadb_x_sa2_sequence = seq;
4385 	p->sadb_x_sa2_reqid = reqid;
4386 
4387 	return m;
4388 }
4389 
4390 /*
4391  * set data into sadb_x_policy
4392  */
4393 static struct mbuf *
4394 key_setsadbxpolicy(const u_int16_t type, const u_int8_t dir, const u_int32_t id,
4395     int mflag)
4396 {
4397 	struct mbuf *m;
4398 	struct sadb_x_policy *p;
4399 	size_t len;
4400 
4401 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
4402 	m = key_alloc_mbuf(len, mflag);
4403 	if (!m || m->m_next) {	/*XXX*/
4404 		if (m)
4405 			m_freem(m);
4406 		return NULL;
4407 	}
4408 
4409 	p = mtod(m, struct sadb_x_policy *);
4410 
4411 	memset(p, 0, len);
4412 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
4413 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
4414 	p->sadb_x_policy_type = type;
4415 	p->sadb_x_policy_dir = dir;
4416 	p->sadb_x_policy_id = id;
4417 
4418 	return m;
4419 }
4420 
4421 /* %%% utilities */
4422 /*
4423  * copy a buffer into the new buffer allocated.
4424  */
4425 static void *
4426 key_newbuf(const void *src, u_int len)
4427 {
4428 	void *new;
4429 
4430 	new = kmem_alloc(len, KM_SLEEP);
4431 	memcpy(new, src, len);
4432 
4433 	return new;
4434 }
4435 
4436 /* compare my own address
4437  * OUT:	1: true, i.e. my address.
4438  *	0: false
4439  */
4440 int
4441 key_ismyaddr(const struct sockaddr *sa)
4442 {
4443 #ifdef INET
4444 	const struct sockaddr_in *sin;
4445 	const struct in_ifaddr *ia;
4446 	int s;
4447 #endif
4448 
4449 	KASSERT(sa != NULL);
4450 
4451 	switch (sa->sa_family) {
4452 #ifdef INET
4453 	case AF_INET:
4454 		sin = (const struct sockaddr_in *)sa;
4455 		s = pserialize_read_enter();
4456 		IN_ADDRLIST_READER_FOREACH(ia) {
4457 			if (sin->sin_family == ia->ia_addr.sin_family &&
4458 			    sin->sin_len == ia->ia_addr.sin_len &&
4459 			    sin->sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
4460 			{
4461 				pserialize_read_exit(s);
4462 				return 1;
4463 			}
4464 		}
4465 		pserialize_read_exit(s);
4466 		break;
4467 #endif
4468 #ifdef INET6
4469 	case AF_INET6:
4470 		return key_ismyaddr6((const struct sockaddr_in6 *)sa);
4471 #endif
4472 	}
4473 
4474 	return 0;
4475 }
4476 
4477 #ifdef INET6
4478 /*
4479  * compare my own address for IPv6.
4480  * 1: ours
4481  * 0: other
4482  * NOTE: derived ip6_input() in KAME. This is necessary to modify more.
4483  */
4484 #include <netinet6/in6_var.h>
4485 
4486 static int
4487 key_ismyaddr6(const struct sockaddr_in6 *sin6)
4488 {
4489 	struct in6_ifaddr *ia;
4490 	int s;
4491 	struct psref psref;
4492 	int bound;
4493 	int ours = 1;
4494 
4495 	bound = curlwp_bind();
4496 	s = pserialize_read_enter();
4497 	IN6_ADDRLIST_READER_FOREACH(ia) {
4498 		bool ingroup;
4499 
4500 		if (key_sockaddr_match((const struct sockaddr *)&sin6,
4501 		    (const struct sockaddr *)&ia->ia_addr, 0)) {
4502 			pserialize_read_exit(s);
4503 			goto ours;
4504 		}
4505 		ia6_acquire(ia, &psref);
4506 		pserialize_read_exit(s);
4507 
4508 		/*
4509 		 * XXX Multicast
4510 		 * XXX why do we care about multlicast here while we don't care
4511 		 * about IPv4 multicast??
4512 		 * XXX scope
4513 		 */
4514 		ingroup = in6_multi_group(&sin6->sin6_addr, ia->ia_ifp);
4515 		if (ingroup) {
4516 			ia6_release(ia, &psref);
4517 			goto ours;
4518 		}
4519 
4520 		s = pserialize_read_enter();
4521 		ia6_release(ia, &psref);
4522 	}
4523 	pserialize_read_exit(s);
4524 
4525 	/* loopback, just for safety */
4526 	if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))
4527 		goto ours;
4528 
4529 	ours = 0;
4530 ours:
4531 	curlwp_bindx(bound);
4532 
4533 	return ours;
4534 }
4535 #endif /*INET6*/
4536 
4537 /*
4538  * compare two secasindex structure.
4539  * flag can specify to compare 2 saidxes.
4540  * compare two secasindex structure without both mode and reqid.
4541  * don't compare port.
4542  * IN:
4543  *      saidx0: source, it can be in SAD.
4544  *      saidx1: object.
4545  * OUT:
4546  *      1 : equal
4547  *      0 : not equal
4548  */
4549 static int
4550 key_saidx_match(
4551 	const struct secasindex *saidx0,
4552 	const struct secasindex *saidx1,
4553 	int flag)
4554 {
4555 	int chkport;
4556 	const struct sockaddr *sa0src, *sa0dst, *sa1src, *sa1dst;
4557 
4558 	KASSERT(saidx0 != NULL);
4559 	KASSERT(saidx1 != NULL);
4560 
4561 	/* sanity */
4562 	if (saidx0->proto != saidx1->proto)
4563 		return 0;
4564 
4565 	if (flag == CMP_EXACTLY) {
4566 		if (saidx0->mode != saidx1->mode)
4567 			return 0;
4568 		if (saidx0->reqid != saidx1->reqid)
4569 			return 0;
4570 		if (memcmp(&saidx0->src, &saidx1->src, saidx0->src.sa.sa_len) != 0 ||
4571 		    memcmp(&saidx0->dst, &saidx1->dst, saidx0->dst.sa.sa_len) != 0)
4572 			return 0;
4573 	} else {
4574 
4575 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
4576 		if (flag == CMP_MODE_REQID ||flag == CMP_REQID) {
4577 			/*
4578 			 * If reqid of SPD is non-zero, unique SA is required.
4579 			 * The result must be of same reqid in this case.
4580 			 */
4581 			if (saidx1->reqid != 0 && saidx0->reqid != saidx1->reqid)
4582 				return 0;
4583 		}
4584 
4585 		if (flag == CMP_MODE_REQID) {
4586 			if (saidx0->mode != IPSEC_MODE_ANY &&
4587 			    saidx0->mode != saidx1->mode)
4588 				return 0;
4589 		}
4590 
4591 
4592 		sa0src = &saidx0->src.sa;
4593 		sa0dst = &saidx0->dst.sa;
4594 		sa1src = &saidx1->src.sa;
4595 		sa1dst = &saidx1->dst.sa;
4596 		/*
4597 		 * If NAT-T is enabled, check ports for tunnel mode.
4598 		 * For ipsecif(4), check ports for transport mode, too.
4599 		 * Don't check ports if they are set to zero
4600 		 * in the SPD: This means we have a non-generated
4601 		 * SPD which can't know UDP ports.
4602 		 */
4603 		if (saidx1->mode == IPSEC_MODE_TUNNEL ||
4604 		    saidx1->mode == IPSEC_MODE_TRANSPORT)
4605 			chkport = PORT_LOOSE;
4606 		else
4607 			chkport = PORT_NONE;
4608 
4609 		if (!key_sockaddr_match(sa0src, sa1src, chkport)) {
4610 			return 0;
4611 		}
4612 		if (!key_sockaddr_match(sa0dst, sa1dst, chkport)) {
4613 			return 0;
4614 		}
4615 	}
4616 
4617 	return 1;
4618 }
4619 
4620 /*
4621  * compare two secindex structure exactly.
4622  * IN:
4623  *	spidx0: source, it is often in SPD.
4624  *	spidx1: object, it is often from PFKEY message.
4625  * OUT:
4626  *	1 : equal
4627  *	0 : not equal
4628  */
4629 static int
4630 key_spidx_match_exactly(
4631 	const struct secpolicyindex *spidx0,
4632 	const struct secpolicyindex *spidx1)
4633 {
4634 
4635 	KASSERT(spidx0 != NULL);
4636 	KASSERT(spidx1 != NULL);
4637 
4638 	/* sanity */
4639 	if (spidx0->prefs != spidx1->prefs ||
4640 	    spidx0->prefd != spidx1->prefd ||
4641 	    spidx0->ul_proto != spidx1->ul_proto)
4642 		return 0;
4643 
4644 	return key_sockaddr_match(&spidx0->src.sa, &spidx1->src.sa, PORT_STRICT) &&
4645 	       key_sockaddr_match(&spidx0->dst.sa, &spidx1->dst.sa, PORT_STRICT);
4646 }
4647 
4648 /*
4649  * compare two secindex structure with mask.
4650  * IN:
4651  *	spidx0: source, it is often in SPD.
4652  *	spidx1: object, it is often from IP header.
4653  * OUT:
4654  *	1 : equal
4655  *	0 : not equal
4656  */
4657 static int
4658 key_spidx_match_withmask(
4659 	const struct secpolicyindex *spidx0,
4660 	const struct secpolicyindex *spidx1)
4661 {
4662 
4663 	KASSERT(spidx0 != NULL);
4664 	KASSERT(spidx1 != NULL);
4665 
4666 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4667 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4668 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4669 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
4670 		return 0;
4671 
4672 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4673 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY &&
4674 	    spidx0->ul_proto != spidx1->ul_proto)
4675 		return 0;
4676 
4677 	switch (spidx0->src.sa.sa_family) {
4678 	case AF_INET:
4679 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY &&
4680 		    spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
4681 			return 0;
4682 		if (!key_bb_match_withmask(&spidx0->src.sin.sin_addr,
4683 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
4684 			return 0;
4685 		break;
4686 	case AF_INET6:
4687 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY &&
4688 		    spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
4689 			return 0;
4690 		/*
4691 		 * scope_id check. if sin6_scope_id is 0, we regard it
4692 		 * as a wildcard scope, which matches any scope zone ID.
4693 		 */
4694 		if (spidx0->src.sin6.sin6_scope_id &&
4695 		    spidx1->src.sin6.sin6_scope_id &&
4696 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
4697 			return 0;
4698 		if (!key_bb_match_withmask(&spidx0->src.sin6.sin6_addr,
4699 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
4700 			return 0;
4701 		break;
4702 	default:
4703 		/* XXX */
4704 		if (memcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
4705 			return 0;
4706 		break;
4707 	}
4708 
4709 	switch (spidx0->dst.sa.sa_family) {
4710 	case AF_INET:
4711 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY &&
4712 		    spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
4713 			return 0;
4714 		if (!key_bb_match_withmask(&spidx0->dst.sin.sin_addr,
4715 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
4716 			return 0;
4717 		break;
4718 	case AF_INET6:
4719 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY &&
4720 		    spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
4721 			return 0;
4722 		/*
4723 		 * scope_id check. if sin6_scope_id is 0, we regard it
4724 		 * as a wildcard scope, which matches any scope zone ID.
4725 		 */
4726 		if (spidx0->src.sin6.sin6_scope_id &&
4727 		    spidx1->src.sin6.sin6_scope_id &&
4728 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
4729 			return 0;
4730 		if (!key_bb_match_withmask(&spidx0->dst.sin6.sin6_addr,
4731 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
4732 			return 0;
4733 		break;
4734 	default:
4735 		/* XXX */
4736 		if (memcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
4737 			return 0;
4738 		break;
4739 	}
4740 
4741 	/* XXX Do we check other field ?  e.g. flowinfo */
4742 
4743 	return 1;
4744 }
4745 
4746 /* returns 0 on match */
4747 static int
4748 key_portcomp(in_port_t port1, in_port_t port2, int howport)
4749 {
4750 	switch (howport) {
4751 	case PORT_NONE:
4752 		return 0;
4753 	case PORT_LOOSE:
4754 		if (port1 == 0 || port2 == 0)
4755 			return 0;
4756 		/*FALLTHROUGH*/
4757 	case PORT_STRICT:
4758 		if (port1 != port2) {
4759 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4760 			    "port fail %d != %d\n", port1, port2);
4761 			return 1;
4762 		}
4763 		return 0;
4764 	default:
4765 		KASSERT(0);
4766 		return 1;
4767 	}
4768 }
4769 
4770 /* returns 1 on match */
4771 static int
4772 key_sockaddr_match(
4773 	const struct sockaddr *sa1,
4774 	const struct sockaddr *sa2,
4775 	int howport)
4776 {
4777 	const struct sockaddr_in *sin1, *sin2;
4778 	const struct sockaddr_in6 *sin61, *sin62;
4779 	char s1[IPSEC_ADDRSTRLEN], s2[IPSEC_ADDRSTRLEN];
4780 
4781 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len) {
4782 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4783 		    "fam/len fail %d != %d || %d != %d\n",
4784 			sa1->sa_family, sa2->sa_family, sa1->sa_len,
4785 			sa2->sa_len);
4786 		return 0;
4787 	}
4788 
4789 	switch (sa1->sa_family) {
4790 	case AF_INET:
4791 		if (sa1->sa_len != sizeof(struct sockaddr_in)) {
4792 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4793 			    "len fail %d != %zu\n",
4794 			    sa1->sa_len, sizeof(struct sockaddr_in));
4795 			return 0;
4796 		}
4797 		sin1 = (const struct sockaddr_in *)sa1;
4798 		sin2 = (const struct sockaddr_in *)sa2;
4799 		if (sin1->sin_addr.s_addr != sin2->sin_addr.s_addr) {
4800 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4801 			    "addr fail %s != %s\n",
4802 			    (in_print(s1, sizeof(s1), &sin1->sin_addr), s1),
4803 			    (in_print(s2, sizeof(s2), &sin2->sin_addr), s2));
4804 			return 0;
4805 		}
4806 		if (key_portcomp(sin1->sin_port, sin2->sin_port, howport)) {
4807 			return 0;
4808 		}
4809 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4810 		    "addr success %s[%d] == %s[%d]\n",
4811 		    (in_print(s1, sizeof(s1), &sin1->sin_addr), s1),
4812 		    sin1->sin_port,
4813 		    (in_print(s2, sizeof(s2), &sin2->sin_addr), s2),
4814 		    sin2->sin_port);
4815 		break;
4816 	case AF_INET6:
4817 		sin61 = (const struct sockaddr_in6 *)sa1;
4818 		sin62 = (const struct sockaddr_in6 *)sa2;
4819 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4820 			return 0;	/*EINVAL*/
4821 
4822 		if (sin61->sin6_scope_id != sin62->sin6_scope_id) {
4823 			return 0;
4824 		}
4825 		if (!IN6_ARE_ADDR_EQUAL(&sin61->sin6_addr, &sin62->sin6_addr)) {
4826 			return 0;
4827 		}
4828 		if (key_portcomp(sin61->sin6_port, sin62->sin6_port, howport)) {
4829 			return 0;
4830 		}
4831 		break;
4832 	default:
4833 		if (memcmp(sa1, sa2, sa1->sa_len) != 0)
4834 			return 0;
4835 		break;
4836 	}
4837 
4838 	return 1;
4839 }
4840 
4841 /*
4842  * compare two buffers with mask.
4843  * IN:
4844  *	addr1: source
4845  *	addr2: object
4846  *	bits:  Number of bits to compare
4847  * OUT:
4848  *	1 : equal
4849  *	0 : not equal
4850  */
4851 static int
4852 key_bb_match_withmask(const void *a1, const void *a2, u_int bits)
4853 {
4854 	const unsigned char *p1 = a1;
4855 	const unsigned char *p2 = a2;
4856 
4857 	/* XXX: This could be considerably faster if we compare a word
4858 	 * at a time, but it is complicated on LSB Endian machines */
4859 
4860 	/* Handle null pointers */
4861 	if (p1 == NULL || p2 == NULL)
4862 		return (p1 == p2);
4863 
4864 	while (bits >= 8) {
4865 		if (*p1++ != *p2++)
4866 			return 0;
4867 		bits -= 8;
4868 	}
4869 
4870 	if (bits > 0) {
4871 		u_int8_t mask = ~((1<<(8-bits))-1);
4872 		if ((*p1 & mask) != (*p2 & mask))
4873 			return 0;
4874 	}
4875 	return 1;	/* Match! */
4876 }
4877 
4878 static void
4879 key_timehandler_spd(time_t now)
4880 {
4881 	u_int dir;
4882 	struct secpolicy *sp;
4883 
4884 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4885 	    retry:
4886 		mutex_enter(&key_spd.lock);
4887 		SPLIST_WRITER_FOREACH(sp, dir) {
4888 			KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
4889 
4890 			if (sp->lifetime == 0 && sp->validtime == 0)
4891 				continue;
4892 
4893 			if ((sp->lifetime && now - sp->created > sp->lifetime) ||
4894 			    (sp->validtime && now - sp->lastused > sp->validtime)) {
4895 				key_unlink_sp(sp);
4896 				mutex_exit(&key_spd.lock);
4897 				key_spdexpire(sp);
4898 				key_destroy_sp(sp);
4899 				goto retry;
4900 			}
4901 		}
4902 		mutex_exit(&key_spd.lock);
4903 	}
4904 
4905     retry_socksplist:
4906 	mutex_enter(&key_spd.lock);
4907 	SOCKSPLIST_WRITER_FOREACH(sp) {
4908 		if (sp->state != IPSEC_SPSTATE_DEAD)
4909 			continue;
4910 
4911 		key_unlink_sp(sp);
4912 		mutex_exit(&key_spd.lock);
4913 		key_destroy_sp(sp);
4914 		goto retry_socksplist;
4915 	}
4916 	mutex_exit(&key_spd.lock);
4917 }
4918 
4919 static void
4920 key_timehandler_sad(time_t now)
4921 {
4922 	struct secashead *sah;
4923 	int s;
4924 
4925 restart:
4926 	mutex_enter(&key_sad.lock);
4927 	SAHLIST_WRITER_FOREACH(sah) {
4928 		/* If sah has been dead and has no sav, then delete it */
4929 		if (sah->state == SADB_SASTATE_DEAD &&
4930 		    !key_sah_has_sav(sah)) {
4931 			key_unlink_sah(sah);
4932 			mutex_exit(&key_sad.lock);
4933 			key_destroy_sah(sah);
4934 			goto restart;
4935 		}
4936 	}
4937 	mutex_exit(&key_sad.lock);
4938 
4939 	s = pserialize_read_enter();
4940 	SAHLIST_READER_FOREACH(sah) {
4941 		struct secasvar *sav;
4942 
4943 		key_sah_ref(sah);
4944 		pserialize_read_exit(s);
4945 
4946 		/* if LARVAL entry doesn't become MATURE, delete it. */
4947 		mutex_enter(&key_sad.lock);
4948 	restart_sav_LARVAL:
4949 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_LARVAL) {
4950 			if (now - sav->created > key_larval_lifetime) {
4951 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
4952 				goto restart_sav_LARVAL;
4953 			}
4954 		}
4955 		mutex_exit(&key_sad.lock);
4956 
4957 		/*
4958 		 * check MATURE entry to start to send expire message
4959 		 * whether or not.
4960 		 */
4961 	restart_sav_MATURE:
4962 		mutex_enter(&key_sad.lock);
4963 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_MATURE) {
4964 			/* we don't need to check. */
4965 			if (sav->lft_s == NULL)
4966 				continue;
4967 
4968 			/* sanity check */
4969 			KASSERT(sav->lft_c != NULL);
4970 
4971 			/* check SOFT lifetime */
4972 			if (sav->lft_s->sadb_lifetime_addtime != 0 &&
4973 			    now - sav->created > sav->lft_s->sadb_lifetime_addtime) {
4974 				/*
4975 				 * check SA to be used whether or not.
4976 				 * when SA hasn't been used, delete it.
4977 				 */
4978 				if (sav->lft_c->sadb_lifetime_usetime == 0) {
4979 					key_sa_chgstate(sav, SADB_SASTATE_DEAD);
4980 					mutex_exit(&key_sad.lock);
4981 				} else {
4982 					key_sa_chgstate(sav, SADB_SASTATE_DYING);
4983 					mutex_exit(&key_sad.lock);
4984 					/*
4985 					 * XXX If we keep to send expire
4986 					 * message in the status of
4987 					 * DYING. Do remove below code.
4988 					 */
4989 					key_expire(sav);
4990 				}
4991 				goto restart_sav_MATURE;
4992 			}
4993 			/* check SOFT lifetime by bytes */
4994 			/*
4995 			 * XXX I don't know the way to delete this SA
4996 			 * when new SA is installed.  Caution when it's
4997 			 * installed too big lifetime by time.
4998 			 */
4999 			else {
5000 				uint64_t lft_c_bytes = 0;
5001 				lifetime_counters_t sum = {0};
5002 
5003 				percpu_foreach(sav->lft_c_counters_percpu,
5004 				    key_sum_lifetime_counters, sum);
5005 				lft_c_bytes = sum[LIFETIME_COUNTER_BYTES];
5006 
5007 				if (sav->lft_s->sadb_lifetime_bytes == 0 ||
5008 				    sav->lft_s->sadb_lifetime_bytes >= lft_c_bytes)
5009 					continue;
5010 
5011 				key_sa_chgstate(sav, SADB_SASTATE_DYING);
5012 				mutex_exit(&key_sad.lock);
5013 				/*
5014 				 * XXX If we keep to send expire
5015 				 * message in the status of
5016 				 * DYING. Do remove below code.
5017 				 */
5018 				key_expire(sav);
5019 				goto restart_sav_MATURE;
5020 			}
5021 		}
5022 		mutex_exit(&key_sad.lock);
5023 
5024 		/* check DYING entry to change status to DEAD. */
5025 		mutex_enter(&key_sad.lock);
5026 	restart_sav_DYING:
5027 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_DYING) {
5028 			/* we don't need to check. */
5029 			if (sav->lft_h == NULL)
5030 				continue;
5031 
5032 			/* sanity check */
5033 			KASSERT(sav->lft_c != NULL);
5034 
5035 			if (sav->lft_h->sadb_lifetime_addtime != 0 &&
5036 			    now - sav->created > sav->lft_h->sadb_lifetime_addtime) {
5037 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
5038 				goto restart_sav_DYING;
5039 			}
5040 #if 0	/* XXX Should we keep to send expire message until HARD lifetime ? */
5041 			else if (sav->lft_s != NULL
5042 			      && sav->lft_s->sadb_lifetime_addtime != 0
5043 			      && now - sav->created > sav->lft_s->sadb_lifetime_addtime) {
5044 				/*
5045 				 * XXX: should be checked to be
5046 				 * installed the valid SA.
5047 				 */
5048 
5049 				/*
5050 				 * If there is no SA then sending
5051 				 * expire message.
5052 				 */
5053 				key_expire(sav);
5054 			}
5055 #endif
5056 			/* check HARD lifetime by bytes */
5057 			else {
5058 				uint64_t lft_c_bytes = 0;
5059 				lifetime_counters_t sum = {0};
5060 
5061 				percpu_foreach(sav->lft_c_counters_percpu,
5062 				    key_sum_lifetime_counters, sum);
5063 				lft_c_bytes = sum[LIFETIME_COUNTER_BYTES];
5064 
5065 				if (sav->lft_h->sadb_lifetime_bytes == 0 ||
5066 				    sav->lft_h->sadb_lifetime_bytes >= lft_c_bytes)
5067 					continue;
5068 
5069 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
5070 				goto restart_sav_DYING;
5071 			}
5072 		}
5073 		mutex_exit(&key_sad.lock);
5074 
5075 		/* delete entry in DEAD */
5076 	restart_sav_DEAD:
5077 		mutex_enter(&key_sad.lock);
5078 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_DEAD) {
5079 			key_unlink_sav(sav);
5080 			mutex_exit(&key_sad.lock);
5081 			key_destroy_sav(sav);
5082 			goto restart_sav_DEAD;
5083 		}
5084 		mutex_exit(&key_sad.lock);
5085 
5086 		s = pserialize_read_enter();
5087 		key_sah_unref(sah);
5088 	}
5089 	pserialize_read_exit(s);
5090 }
5091 
5092 static void
5093 key_timehandler_acq(time_t now)
5094 {
5095 #ifndef IPSEC_NONBLOCK_ACQUIRE
5096 	struct secacq *acq, *nextacq;
5097 
5098     restart:
5099 	mutex_enter(&key_misc.lock);
5100 	LIST_FOREACH_SAFE(acq, &key_misc.acqlist, chain, nextacq) {
5101 		if (now - acq->created > key_blockacq_lifetime) {
5102 			LIST_REMOVE(acq, chain);
5103 			mutex_exit(&key_misc.lock);
5104 			kmem_free(acq, sizeof(*acq));
5105 			goto restart;
5106 		}
5107 	}
5108 	mutex_exit(&key_misc.lock);
5109 #endif
5110 }
5111 
5112 static void
5113 key_timehandler_spacq(time_t now)
5114 {
5115 #ifdef notyet
5116 	struct secspacq *acq, *nextacq;
5117 
5118 	LIST_FOREACH_SAFE(acq, &key_misc.spacqlist, chain, nextacq) {
5119 		if (now - acq->created > key_blockacq_lifetime) {
5120 			KASSERT(__LIST_CHAINED(acq));
5121 			LIST_REMOVE(acq, chain);
5122 			kmem_free(acq, sizeof(*acq));
5123 		}
5124 	}
5125 #endif
5126 }
5127 
5128 static unsigned int key_timehandler_work_enqueued = 0;
5129 
5130 /*
5131  * time handler.
5132  * scanning SPD and SAD to check status for each entries,
5133  * and do to remove or to expire.
5134  */
5135 static void
5136 key_timehandler_work(struct work *wk, void *arg)
5137 {
5138 	time_t now = time_uptime;
5139 
5140 	/* We can allow enqueuing another work at this point */
5141 	atomic_swap_uint(&key_timehandler_work_enqueued, 0);
5142 
5143 	key_timehandler_spd(now);
5144 	key_timehandler_sad(now);
5145 	key_timehandler_acq(now);
5146 	key_timehandler_spacq(now);
5147 
5148 	key_acquire_sendup_pending_mbuf();
5149 
5150 	/* do exchange to tick time !! */
5151 	callout_reset(&key_timehandler_ch, hz, key_timehandler, NULL);
5152 
5153 	return;
5154 }
5155 
5156 static void
5157 key_timehandler(void *arg)
5158 {
5159 
5160 	/* Avoid enqueuing another work when one is already enqueued */
5161 	if (atomic_swap_uint(&key_timehandler_work_enqueued, 1) == 1)
5162 		return;
5163 
5164 	workqueue_enqueue(key_timehandler_wq, &key_timehandler_wk, NULL);
5165 }
5166 
5167 u_long
5168 key_random(void)
5169 {
5170 	u_long value;
5171 
5172 	key_randomfill(&value, sizeof(value));
5173 	return value;
5174 }
5175 
5176 void
5177 key_randomfill(void *p, size_t l)
5178 {
5179 
5180 	cprng_fast(p, l);
5181 }
5182 
5183 /*
5184  * map SADB_SATYPE_* to IPPROTO_*.
5185  * if satype == SADB_SATYPE then satype is mapped to ~0.
5186  * OUT:
5187  *	0: invalid satype.
5188  */
5189 static u_int16_t
5190 key_satype2proto(u_int8_t satype)
5191 {
5192 	switch (satype) {
5193 	case SADB_SATYPE_UNSPEC:
5194 		return IPSEC_PROTO_ANY;
5195 	case SADB_SATYPE_AH:
5196 		return IPPROTO_AH;
5197 	case SADB_SATYPE_ESP:
5198 		return IPPROTO_ESP;
5199 	case SADB_X_SATYPE_IPCOMP:
5200 		return IPPROTO_IPCOMP;
5201 	case SADB_X_SATYPE_TCPSIGNATURE:
5202 		return IPPROTO_TCP;
5203 	default:
5204 		return 0;
5205 	}
5206 	/* NOTREACHED */
5207 }
5208 
5209 /*
5210  * map IPPROTO_* to SADB_SATYPE_*
5211  * OUT:
5212  *	0: invalid protocol type.
5213  */
5214 static u_int8_t
5215 key_proto2satype(u_int16_t proto)
5216 {
5217 	switch (proto) {
5218 	case IPPROTO_AH:
5219 		return SADB_SATYPE_AH;
5220 	case IPPROTO_ESP:
5221 		return SADB_SATYPE_ESP;
5222 	case IPPROTO_IPCOMP:
5223 		return SADB_X_SATYPE_IPCOMP;
5224 	case IPPROTO_TCP:
5225 		return SADB_X_SATYPE_TCPSIGNATURE;
5226 	default:
5227 		return 0;
5228 	}
5229 	/* NOTREACHED */
5230 }
5231 
5232 static int
5233 key_setsecasidx(int proto, int mode, int reqid,
5234     const struct sockaddr *src, const struct sockaddr *dst,
5235     struct secasindex * saidx)
5236 {
5237 	const union sockaddr_union *src_u = (const union sockaddr_union *)src;
5238 	const union sockaddr_union *dst_u = (const union sockaddr_union *)dst;
5239 
5240 	/* sa len safety check */
5241 	if (key_checksalen(src_u) != 0)
5242 		return -1;
5243 	if (key_checksalen(dst_u) != 0)
5244 		return -1;
5245 
5246 	memset(saidx, 0, sizeof(*saidx));
5247 	saidx->proto = proto;
5248 	saidx->mode = mode;
5249 	saidx->reqid = reqid;
5250 	memcpy(&saidx->src, src_u, src_u->sa.sa_len);
5251 	memcpy(&saidx->dst, dst_u, dst_u->sa.sa_len);
5252 
5253 	key_porttosaddr(&((saidx)->src), 0);
5254 	key_porttosaddr(&((saidx)->dst), 0);
5255 	return 0;
5256 }
5257 
5258 static void
5259 key_init_spidx_bymsghdr(struct secpolicyindex *spidx,
5260     const struct sadb_msghdr *mhp)
5261 {
5262 	const struct sadb_address *src0, *dst0;
5263 	const struct sockaddr *src, *dst;
5264 	const struct sadb_x_policy *xpl0;
5265 
5266 	src0 = mhp->ext[SADB_EXT_ADDRESS_SRC];
5267 	dst0 = mhp->ext[SADB_EXT_ADDRESS_DST];
5268 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5269 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5270 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
5271 
5272 	memset(spidx, 0, sizeof(*spidx));
5273 	spidx->dir = xpl0->sadb_x_policy_dir;
5274 	spidx->prefs = src0->sadb_address_prefixlen;
5275 	spidx->prefd = dst0->sadb_address_prefixlen;
5276 	spidx->ul_proto = src0->sadb_address_proto;
5277 	/* XXX boundary check against sa_len */
5278 	memcpy(&spidx->src, src, src->sa_len);
5279 	memcpy(&spidx->dst, dst, dst->sa_len);
5280 }
5281 
5282 /* %%% PF_KEY */
5283 /*
5284  * SADB_GETSPI processing is to receive
5285  *	<base, (SA2), src address, dst address, (SPI range)>
5286  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
5287  * tree with the status of LARVAL, and send
5288  *	<base, SA(*), address(SD)>
5289  * to the IKMPd.
5290  *
5291  * IN:	mhp: pointer to the pointer to each header.
5292  * OUT:	NULL if fail.
5293  *	other if success, return pointer to the message to send.
5294  */
5295 static int
5296 key_api_getspi(struct socket *so, struct mbuf *m,
5297 	   const struct sadb_msghdr *mhp)
5298 {
5299 	const struct sockaddr *src, *dst;
5300 	struct secasindex saidx;
5301 	struct secashead *sah;
5302 	struct secasvar *newsav;
5303 	u_int8_t proto;
5304 	u_int32_t spi;
5305 	u_int8_t mode;
5306 	u_int16_t reqid;
5307 	int error;
5308 
5309 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
5310 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
5311 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5312 		return key_senderror(so, m, EINVAL);
5313 	}
5314 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
5315 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
5316 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5317 		return key_senderror(so, m, EINVAL);
5318 	}
5319 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
5320 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
5321 		mode = sa2->sadb_x_sa2_mode;
5322 		reqid = sa2->sadb_x_sa2_reqid;
5323 	} else {
5324 		mode = IPSEC_MODE_ANY;
5325 		reqid = 0;
5326 	}
5327 
5328 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5329 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5330 
5331 	/* map satype to proto */
5332 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
5333 	if (proto == 0) {
5334 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
5335 		return key_senderror(so, m, EINVAL);
5336 	}
5337 
5338 
5339 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
5340 	if (error != 0)
5341 		return key_senderror(so, m, EINVAL);
5342 
5343 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
5344 	if (error != 0)
5345 		return key_senderror(so, m, EINVAL);
5346 
5347 	/* SPI allocation */
5348 	spi = key_do_getnewspi(mhp->ext[SADB_EXT_SPIRANGE], &saidx);
5349 	if (spi == 0)
5350 		return key_senderror(so, m, EINVAL);
5351 
5352 	/* get a SA index */
5353 	sah = key_getsah_ref(&saidx, CMP_REQID);
5354 	if (sah == NULL) {
5355 		/* create a new SA index */
5356 		sah = key_newsah(&saidx);
5357 		if (sah == NULL) {
5358 			IPSECLOG(LOG_DEBUG, "No more memory.\n");
5359 			return key_senderror(so, m, ENOBUFS);
5360 		}
5361 	}
5362 
5363 	/* get a new SA */
5364 	/* XXX rewrite */
5365 	newsav = KEY_NEWSAV(m, mhp, &error);
5366 	if (newsav == NULL) {
5367 		key_sah_unref(sah);
5368 		/* XXX don't free new SA index allocated in above. */
5369 		return key_senderror(so, m, error);
5370 	}
5371 
5372 	/* set spi */
5373 	newsav->spi = htonl(spi);
5374 
5375 	/* Add to sah#savlist */
5376 	key_init_sav(newsav);
5377 	newsav->sah = sah;
5378 	newsav->state = SADB_SASTATE_LARVAL;
5379 	mutex_enter(&key_sad.lock);
5380 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_LARVAL, newsav);
5381 	mutex_exit(&key_sad.lock);
5382 	key_validate_savlist(sah, SADB_SASTATE_LARVAL);
5383 
5384 	key_sah_unref(sah);
5385 
5386 #ifndef IPSEC_NONBLOCK_ACQUIRE
5387 	/* delete the entry in key_misc.acqlist */
5388 	if (mhp->msg->sadb_msg_seq != 0) {
5389 		struct secacq *acq;
5390 		mutex_enter(&key_misc.lock);
5391 		acq = key_getacqbyseq(mhp->msg->sadb_msg_seq);
5392 		if (acq != NULL) {
5393 			/* reset counter in order to deletion by timehandler. */
5394 			acq->created = time_uptime;
5395 			acq->count = 0;
5396 		}
5397 		mutex_exit(&key_misc.lock);
5398 	}
5399 #endif
5400 
5401     {
5402 	struct mbuf *n, *nn;
5403 	struct sadb_sa *m_sa;
5404 	int off, len;
5405 
5406 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
5407 	    PFKEY_ALIGN8(sizeof(struct sadb_sa)) <= MCLBYTES);
5408 
5409 	/* create new sadb_msg to reply. */
5410 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
5411 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
5412 
5413 	n = key_alloc_mbuf_simple(len, M_WAITOK);
5414 	n->m_len = len;
5415 	n->m_next = NULL;
5416 	off = 0;
5417 
5418 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
5419 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
5420 
5421 	m_sa = (struct sadb_sa *)(mtod(n, char *) + off);
5422 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
5423 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
5424 	m_sa->sadb_sa_spi = htonl(spi);
5425 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
5426 
5427 	KASSERTMSG(off == len, "length inconsistency");
5428 
5429 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
5430 	    SADB_EXT_ADDRESS_DST);
5431 
5432 	KASSERT(n->m_len >= sizeof(struct sadb_msg));
5433 
5434 	n->m_pkthdr.len = 0;
5435 	for (nn = n; nn; nn = nn->m_next)
5436 		n->m_pkthdr.len += nn->m_len;
5437 
5438 	key_fill_replymsg(n, newsav->seq);
5439 	m_freem(m);
5440 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
5441     }
5442 }
5443 
5444 /*
5445  * allocating new SPI
5446  * called by key_api_getspi().
5447  * OUT:
5448  *	0:	failure.
5449  *	others: success.
5450  */
5451 static u_int32_t
5452 key_do_getnewspi(const struct sadb_spirange *spirange,
5453 		 const struct secasindex *saidx)
5454 {
5455 	u_int32_t newspi;
5456 	u_int32_t spmin, spmax;
5457 	int count = key_spi_trycnt;
5458 
5459 	/* set spi range to allocate */
5460 	if (spirange != NULL) {
5461 		spmin = spirange->sadb_spirange_min;
5462 		spmax = spirange->sadb_spirange_max;
5463 	} else {
5464 		spmin = key_spi_minval;
5465 		spmax = key_spi_maxval;
5466 	}
5467 	/* IPCOMP needs 2-byte SPI */
5468 	if (saidx->proto == IPPROTO_IPCOMP) {
5469 		u_int32_t t;
5470 		if (spmin >= 0x10000)
5471 			spmin = 0xffff;
5472 		if (spmax >= 0x10000)
5473 			spmax = 0xffff;
5474 		if (spmin > spmax) {
5475 			t = spmin; spmin = spmax; spmax = t;
5476 		}
5477 	}
5478 
5479 	if (spmin == spmax) {
5480 		if (key_checkspidup(saidx, htonl(spmin))) {
5481 			IPSECLOG(LOG_DEBUG, "SPI %u exists already.\n", spmin);
5482 			return 0;
5483 		}
5484 
5485 		count--; /* taking one cost. */
5486 		newspi = spmin;
5487 
5488 	} else {
5489 
5490 		/* init SPI */
5491 		newspi = 0;
5492 
5493 		/* when requesting to allocate spi ranged */
5494 		while (count--) {
5495 			/* generate pseudo-random SPI value ranged. */
5496 			newspi = spmin + (key_random() % (spmax - spmin + 1));
5497 
5498 			if (!key_checkspidup(saidx, htonl(newspi)))
5499 				break;
5500 		}
5501 
5502 		if (count == 0 || newspi == 0) {
5503 			IPSECLOG(LOG_DEBUG, "to allocate spi is failed.\n");
5504 			return 0;
5505 		}
5506 	}
5507 
5508 	/* statistics */
5509 	keystat.getspi_count =
5510 	    (keystat.getspi_count + key_spi_trycnt - count) / 2;
5511 
5512 	return newspi;
5513 }
5514 
5515 static int
5516 key_handle_natt_info(struct secasvar *sav,
5517       		     const struct sadb_msghdr *mhp)
5518 {
5519 	const char *msg = "?" ;
5520 	struct sadb_x_nat_t_type *type;
5521 	struct sadb_x_nat_t_port *sport, *dport;
5522 	struct sadb_address *iaddr, *raddr;
5523 	struct sadb_x_nat_t_frag *frag;
5524 
5525 	if (mhp->ext[SADB_X_EXT_NAT_T_TYPE] == NULL ||
5526 	    mhp->ext[SADB_X_EXT_NAT_T_SPORT] == NULL ||
5527 	    mhp->ext[SADB_X_EXT_NAT_T_DPORT] == NULL)
5528 		return 0;
5529 
5530 	if (mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) {
5531 		msg = "TYPE";
5532 		goto bad;
5533 	}
5534 
5535 	if (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) {
5536 		msg = "SPORT";
5537 		goto bad;
5538 	}
5539 
5540 	if (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport)) {
5541 		msg = "DPORT";
5542 		goto bad;
5543 	}
5544 
5545 	if (mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL) {
5546 		IPSECLOG(LOG_DEBUG, "NAT-T OAi present\n");
5547 		if (mhp->extlen[SADB_X_EXT_NAT_T_OAI] < sizeof(*iaddr)) {
5548 			msg = "OAI";
5549 			goto bad;
5550 		}
5551 	}
5552 
5553 	if (mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL) {
5554 		IPSECLOG(LOG_DEBUG, "NAT-T OAr present\n");
5555 		if (mhp->extlen[SADB_X_EXT_NAT_T_OAR] < sizeof(*raddr)) {
5556 			msg = "OAR";
5557 			goto bad;
5558 		}
5559 	}
5560 
5561 	if (mhp->ext[SADB_X_EXT_NAT_T_FRAG] != NULL) {
5562 	    if (mhp->extlen[SADB_X_EXT_NAT_T_FRAG] < sizeof(*frag)) {
5563 		    msg = "FRAG";
5564 		    goto bad;
5565 	    }
5566 	}
5567 
5568 	type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5569 	sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5570 	dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5571 	iaddr = mhp->ext[SADB_X_EXT_NAT_T_OAI];
5572 	raddr = mhp->ext[SADB_X_EXT_NAT_T_OAR];
5573 	frag = mhp->ext[SADB_X_EXT_NAT_T_FRAG];
5574 
5575 	IPSECLOG(LOG_DEBUG, "type %d, sport = %d, dport = %d\n",
5576 	    type->sadb_x_nat_t_type_type,
5577 	    ntohs(sport->sadb_x_nat_t_port_port),
5578 	    ntohs(dport->sadb_x_nat_t_port_port));
5579 
5580 	sav->natt_type = type->sadb_x_nat_t_type_type;
5581 	key_porttosaddr(&sav->sah->saidx.src, sport->sadb_x_nat_t_port_port);
5582 	key_porttosaddr(&sav->sah->saidx.dst, dport->sadb_x_nat_t_port_port);
5583 	if (frag)
5584 		sav->esp_frag = frag->sadb_x_nat_t_frag_fraglen;
5585 	else
5586 		sav->esp_frag = IP_MAXPACKET;
5587 
5588 	return 0;
5589 bad:
5590 	IPSECLOG(LOG_DEBUG, "invalid message %s\n", msg);
5591 	__USE(msg);
5592 	return -1;
5593 }
5594 
5595 /* Just update the IPSEC_NAT_T ports if present */
5596 static int
5597 key_set_natt_ports(union sockaddr_union *src, union sockaddr_union *dst,
5598       		     const struct sadb_msghdr *mhp)
5599 {
5600 	if (mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL)
5601 		IPSECLOG(LOG_DEBUG, "NAT-T OAi present\n");
5602 	if (mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL)
5603 		IPSECLOG(LOG_DEBUG, "NAT-T OAr present\n");
5604 
5605 	if ((mhp->ext[SADB_X_EXT_NAT_T_TYPE] != NULL) &&
5606 	    (mhp->ext[SADB_X_EXT_NAT_T_SPORT] != NULL) &&
5607 	    (mhp->ext[SADB_X_EXT_NAT_T_DPORT] != NULL)) {
5608 		struct sadb_x_nat_t_type *type;
5609 		struct sadb_x_nat_t_port *sport;
5610 		struct sadb_x_nat_t_port *dport;
5611 
5612 		if ((mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) ||
5613 		    (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) ||
5614 		    (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport))) {
5615 			IPSECLOG(LOG_DEBUG, "invalid message\n");
5616 			return -1;
5617 		}
5618 
5619 		type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5620 		sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5621 		dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5622 
5623 		key_porttosaddr(src, sport->sadb_x_nat_t_port_port);
5624 		key_porttosaddr(dst, dport->sadb_x_nat_t_port_port);
5625 
5626 		IPSECLOG(LOG_DEBUG, "type %d, sport = %d, dport = %d\n",
5627 		    type->sadb_x_nat_t_type_type,
5628 		    ntohs(sport->sadb_x_nat_t_port_port),
5629 		    ntohs(dport->sadb_x_nat_t_port_port));
5630 	}
5631 
5632 	return 0;
5633 }
5634 
5635 
5636 /*
5637  * SADB_UPDATE processing
5638  * receive
5639  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5640  *       key(AE), (identity(SD),) (sensitivity)>
5641  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5642  * and send
5643  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5644  *       (identity(SD),) (sensitivity)>
5645  * to the ikmpd.
5646  *
5647  * m will always be freed.
5648  */
5649 static int
5650 key_api_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5651 {
5652 	struct sadb_sa *sa0;
5653 	const struct sockaddr *src, *dst;
5654 	struct secasindex saidx;
5655 	struct secashead *sah;
5656 	struct secasvar *sav, *newsav, *oldsav;
5657 	u_int16_t proto;
5658 	u_int8_t mode;
5659 	u_int16_t reqid;
5660 	int error;
5661 
5662 	/* map satype to proto */
5663 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
5664 	if (proto == 0) {
5665 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
5666 		return key_senderror(so, m, EINVAL);
5667 	}
5668 
5669 	if (mhp->ext[SADB_EXT_SA] == NULL ||
5670 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
5671 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
5672 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5673 	     mhp->ext[SADB_EXT_KEY_ENCRYPT] == NULL) ||
5674 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5675 	     mhp->ext[SADB_EXT_KEY_AUTH] == NULL) ||
5676 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL &&
5677 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] == NULL) ||
5678 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] == NULL &&
5679 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] != NULL)) {
5680 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5681 		return key_senderror(so, m, EINVAL);
5682 	}
5683 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
5684 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
5685 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
5686 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5687 		return key_senderror(so, m, EINVAL);
5688 	}
5689 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
5690 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
5691 		mode = sa2->sadb_x_sa2_mode;
5692 		reqid = sa2->sadb_x_sa2_reqid;
5693 	} else {
5694 		mode = IPSEC_MODE_ANY;
5695 		reqid = 0;
5696 	}
5697 	/* XXX boundary checking for other extensions */
5698 
5699 	sa0 = mhp->ext[SADB_EXT_SA];
5700 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5701 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5702 
5703 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
5704 	if (error != 0)
5705 		return key_senderror(so, m, EINVAL);
5706 
5707 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
5708 	if (error != 0)
5709 		return key_senderror(so, m, EINVAL);
5710 
5711 	/* get a SA header */
5712 	sah = key_getsah_ref(&saidx, CMP_REQID);
5713 	if (sah == NULL) {
5714 		IPSECLOG(LOG_DEBUG, "no SA index found.\n");
5715 		return key_senderror(so, m, ENOENT);
5716 	}
5717 
5718 	/* set spidx if there */
5719 	/* XXX rewrite */
5720 	error = key_setident(sah, m, mhp);
5721 	if (error)
5722 		goto error_sah;
5723 
5724 	/* find a SA with sequence number. */
5725 #ifdef IPSEC_DOSEQCHECK
5726 	if (mhp->msg->sadb_msg_seq != 0) {
5727 		sav = key_getsavbyseq(sah, mhp->msg->sadb_msg_seq);
5728 		if (sav == NULL) {
5729 			IPSECLOG(LOG_DEBUG,
5730 			    "no larval SA with sequence %u exists.\n",
5731 			    mhp->msg->sadb_msg_seq);
5732 			error = ENOENT;
5733 			goto error_sah;
5734 		}
5735 	}
5736 #else
5737 	sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
5738 	if (sav == NULL) {
5739 		IPSECLOG(LOG_DEBUG, "no such a SA found (spi:%u)\n",
5740 		    (u_int32_t)ntohl(sa0->sadb_sa_spi));
5741 		error = EINVAL;
5742 		goto error_sah;
5743 	}
5744 #endif
5745 
5746 	/* validity check */
5747 	if (sav->sah->saidx.proto != proto) {
5748 		IPSECLOG(LOG_DEBUG, "protocol mismatched (DB=%u param=%u)\n",
5749 		    sav->sah->saidx.proto, proto);
5750 		error = EINVAL;
5751 		goto error;
5752 	}
5753 #ifdef IPSEC_DOSEQCHECK
5754 	if (sav->spi != sa0->sadb_sa_spi) {
5755 		IPSECLOG(LOG_DEBUG, "SPI mismatched (DB:%u param:%u)\n",
5756 		    (u_int32_t)ntohl(sav->spi),
5757 		    (u_int32_t)ntohl(sa0->sadb_sa_spi));
5758 		error = EINVAL;
5759 		goto error;
5760 	}
5761 #endif
5762 	if (sav->pid != mhp->msg->sadb_msg_pid) {
5763 		IPSECLOG(LOG_DEBUG, "pid mismatched (DB:%u param:%u)\n",
5764 		    sav->pid, mhp->msg->sadb_msg_pid);
5765 		error = EINVAL;
5766 		goto error;
5767 	}
5768 
5769 	/*
5770 	 * Allocate a new SA instead of modifying the existing SA directly
5771 	 * to avoid race conditions.
5772 	 */
5773 	newsav = kmem_zalloc(sizeof(struct secasvar), KM_SLEEP);
5774 
5775 	/* copy sav values */
5776 	newsav->spi = sav->spi;
5777 	newsav->seq = sav->seq;
5778 	newsav->created = sav->created;
5779 	newsav->pid = sav->pid;
5780 	newsav->sah = sav->sah;
5781 
5782 	error = key_setsaval(newsav, m, mhp);
5783 	if (error) {
5784 		kmem_free(newsav, sizeof(*newsav));
5785 		goto error;
5786 	}
5787 
5788 	error = key_handle_natt_info(newsav, mhp);
5789 	if (error != 0) {
5790 		key_delsav(newsav);
5791 		goto error;
5792 	}
5793 
5794 	error = key_init_xform(newsav);
5795 	if (error != 0) {
5796 		key_delsav(newsav);
5797 		goto error;
5798 	}
5799 
5800 	/* Add to sah#savlist */
5801 	key_init_sav(newsav);
5802 	newsav->state = SADB_SASTATE_MATURE;
5803 	mutex_enter(&key_sad.lock);
5804 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_MATURE, newsav);
5805 	SAVLUT_WRITER_INSERT_HEAD(newsav);
5806 	mutex_exit(&key_sad.lock);
5807 	key_validate_savlist(sah, SADB_SASTATE_MATURE);
5808 
5809 	/*
5810 	 * We need to lookup and remove the sav atomically, so get it again
5811 	 * here by a special API while we have a reference to it.
5812 	 */
5813 	oldsav = key_lookup_and_remove_sav(sah, sa0->sadb_sa_spi, sav);
5814 	KASSERT(oldsav == NULL || oldsav == sav);
5815 	/* We can release the reference because of oldsav */
5816 	KEY_SA_UNREF(&sav);
5817 	if (oldsav == NULL) {
5818 		/* Someone has already removed the sav.  Nothing to do. */
5819 	} else {
5820 		key_wait_sav(oldsav);
5821 		key_destroy_sav(oldsav);
5822 		oldsav = NULL;
5823 	}
5824 	sav = NULL;
5825 
5826 	key_sah_unref(sah);
5827 	sah = NULL;
5828 
5829     {
5830 	struct mbuf *n;
5831 
5832 	/* set msg buf from mhp */
5833 	n = key_getmsgbuf_x1(m, mhp);
5834 	if (n == NULL) {
5835 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
5836 		return key_senderror(so, m, ENOBUFS);
5837 	}
5838 
5839 	m_freem(m);
5840 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5841     }
5842 error:
5843 	KEY_SA_UNREF(&sav);
5844 error_sah:
5845 	key_sah_unref(sah);
5846 	return key_senderror(so, m, error);
5847 }
5848 
5849 /*
5850  * search SAD with sequence for a SA which state is SADB_SASTATE_LARVAL.
5851  * only called by key_api_update().
5852  * OUT:
5853  *	NULL	: not found
5854  *	others	: found, pointer to a SA.
5855  */
5856 #ifdef IPSEC_DOSEQCHECK
5857 static struct secasvar *
5858 key_getsavbyseq(struct secashead *sah, u_int32_t seq)
5859 {
5860 	struct secasvar *sav;
5861 	u_int state;
5862 	int s;
5863 
5864 	state = SADB_SASTATE_LARVAL;
5865 
5866 	/* search SAD with sequence number ? */
5867 	s = pserialize_read_enter();
5868 	SAVLIST_READER_FOREACH(sav, sah, state) {
5869 		KEY_CHKSASTATE(state, sav->state);
5870 
5871 		if (sav->seq == seq) {
5872 			SA_ADDREF(sav);
5873 			KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
5874 			    "DP cause refcnt++:%d SA:%p\n",
5875 			    key_sa_refcnt(sav), sav);
5876 			break;
5877 		}
5878 	}
5879 	pserialize_read_exit(s);
5880 
5881 	return sav;
5882 }
5883 #endif
5884 
5885 /*
5886  * SADB_ADD processing
5887  * add an entry to SA database, when received
5888  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5889  *       key(AE), (identity(SD),) (sensitivity)>
5890  * from the ikmpd,
5891  * and send
5892  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5893  *       (identity(SD),) (sensitivity)>
5894  * to the ikmpd.
5895  *
5896  * IGNORE identity and sensitivity messages.
5897  *
5898  * m will always be freed.
5899  */
5900 static int
5901 key_api_add(struct socket *so, struct mbuf *m,
5902 	const struct sadb_msghdr *mhp)
5903 {
5904 	struct sadb_sa *sa0;
5905 	const struct sockaddr *src, *dst;
5906 	struct secasindex saidx;
5907 	struct secashead *sah;
5908 	struct secasvar *newsav;
5909 	u_int16_t proto;
5910 	u_int8_t mode;
5911 	u_int16_t reqid;
5912 	int error;
5913 
5914 	/* map satype to proto */
5915 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
5916 	if (proto == 0) {
5917 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
5918 		return key_senderror(so, m, EINVAL);
5919 	}
5920 
5921 	if (mhp->ext[SADB_EXT_SA] == NULL ||
5922 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
5923 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
5924 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5925 	     mhp->ext[SADB_EXT_KEY_ENCRYPT] == NULL) ||
5926 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5927 	     mhp->ext[SADB_EXT_KEY_AUTH] == NULL) ||
5928 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL &&
5929 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] == NULL) ||
5930 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] == NULL &&
5931 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] != NULL)) {
5932 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5933 		return key_senderror(so, m, EINVAL);
5934 	}
5935 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
5936 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
5937 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
5938 		/* XXX need more */
5939 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5940 		return key_senderror(so, m, EINVAL);
5941 	}
5942 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
5943 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
5944 		mode = sa2->sadb_x_sa2_mode;
5945 		reqid = sa2->sadb_x_sa2_reqid;
5946 	} else {
5947 		mode = IPSEC_MODE_ANY;
5948 		reqid = 0;
5949 	}
5950 
5951 	sa0 = mhp->ext[SADB_EXT_SA];
5952 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5953 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5954 
5955 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
5956 	if (error != 0)
5957 		return key_senderror(so, m, EINVAL);
5958 
5959 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
5960 	if (error != 0)
5961 		return key_senderror(so, m, EINVAL);
5962 
5963 	/* get a SA header */
5964 	sah = key_getsah_ref(&saidx, CMP_REQID);
5965 	if (sah == NULL) {
5966 		/* create a new SA header */
5967 		sah = key_newsah(&saidx);
5968 		if (sah == NULL) {
5969 			IPSECLOG(LOG_DEBUG, "No more memory.\n");
5970 			return key_senderror(so, m, ENOBUFS);
5971 		}
5972 	}
5973 
5974 	/* set spidx if there */
5975 	/* XXX rewrite */
5976 	error = key_setident(sah, m, mhp);
5977 	if (error)
5978 		goto error;
5979 
5980     {
5981 	struct secasvar *sav;
5982 
5983 	/* We can create new SA only if SPI is differenct. */
5984 	sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
5985 	if (sav != NULL) {
5986 		KEY_SA_UNREF(&sav);
5987 		IPSECLOG(LOG_DEBUG, "SA already exists.\n");
5988 		error = EEXIST;
5989 		goto error;
5990 	}
5991     }
5992 
5993 	/* create new SA entry. */
5994 	newsav = KEY_NEWSAV(m, mhp, &error);
5995 	if (newsav == NULL)
5996 		goto error;
5997 	newsav->sah = sah;
5998 
5999 	error = key_handle_natt_info(newsav, mhp);
6000 	if (error != 0) {
6001 		key_delsav(newsav);
6002 		error = EINVAL;
6003 		goto error;
6004 	}
6005 
6006 	error = key_init_xform(newsav);
6007 	if (error != 0) {
6008 		key_delsav(newsav);
6009 		goto error;
6010 	}
6011 
6012 	/* Add to sah#savlist */
6013 	key_init_sav(newsav);
6014 	newsav->state = SADB_SASTATE_MATURE;
6015 	mutex_enter(&key_sad.lock);
6016 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_MATURE, newsav);
6017 	SAVLUT_WRITER_INSERT_HEAD(newsav);
6018 	mutex_exit(&key_sad.lock);
6019 	key_validate_savlist(sah, SADB_SASTATE_MATURE);
6020 
6021 	key_sah_unref(sah);
6022 	sah = NULL;
6023 
6024 	/*
6025 	 * don't call key_freesav() here, as we would like to keep the SA
6026 	 * in the database on success.
6027 	 */
6028 
6029     {
6030 	struct mbuf *n;
6031 
6032 	/* set msg buf from mhp */
6033 	n = key_getmsgbuf_x1(m, mhp);
6034 	if (n == NULL) {
6035 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
6036 		return key_senderror(so, m, ENOBUFS);
6037 	}
6038 
6039 	m_freem(m);
6040 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6041     }
6042 error:
6043 	key_sah_unref(sah);
6044 	return key_senderror(so, m, error);
6045 }
6046 
6047 /* m is retained */
6048 static int
6049 key_setident(struct secashead *sah, struct mbuf *m,
6050 	     const struct sadb_msghdr *mhp)
6051 {
6052 	const struct sadb_ident *idsrc, *iddst;
6053 	int idsrclen, iddstlen;
6054 
6055 	KASSERT(!cpu_softintr_p());
6056 	KASSERT(sah != NULL);
6057 	KASSERT(m != NULL);
6058 	KASSERT(mhp != NULL);
6059 	KASSERT(mhp->msg != NULL);
6060 
6061 	/*
6062 	 * Can be called with an existing sah from key_api_update().
6063 	 */
6064 	if (sah->idents != NULL) {
6065 		kmem_free(sah->idents, sah->idents_len);
6066 		sah->idents = NULL;
6067 		sah->idents_len = 0;
6068 	}
6069 	if (sah->identd != NULL) {
6070 		kmem_free(sah->identd, sah->identd_len);
6071 		sah->identd = NULL;
6072 		sah->identd_len = 0;
6073 	}
6074 
6075 	/* don't make buffer if not there */
6076 	if (mhp->ext[SADB_EXT_IDENTITY_SRC] == NULL &&
6077 	    mhp->ext[SADB_EXT_IDENTITY_DST] == NULL) {
6078 		sah->idents = NULL;
6079 		sah->identd = NULL;
6080 		return 0;
6081 	}
6082 
6083 	if (mhp->ext[SADB_EXT_IDENTITY_SRC] == NULL ||
6084 	    mhp->ext[SADB_EXT_IDENTITY_DST] == NULL) {
6085 		IPSECLOG(LOG_DEBUG, "invalid identity.\n");
6086 		return EINVAL;
6087 	}
6088 
6089 	idsrc = mhp->ext[SADB_EXT_IDENTITY_SRC];
6090 	iddst = mhp->ext[SADB_EXT_IDENTITY_DST];
6091 	idsrclen = mhp->extlen[SADB_EXT_IDENTITY_SRC];
6092 	iddstlen = mhp->extlen[SADB_EXT_IDENTITY_DST];
6093 
6094 	/* validity check */
6095 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
6096 		IPSECLOG(LOG_DEBUG, "ident type mismatched src %u, dst %u.\n",
6097 		    idsrc->sadb_ident_type, iddst->sadb_ident_type);
6098 		return EINVAL;
6099 	}
6100 
6101 	switch (idsrc->sadb_ident_type) {
6102 	case SADB_IDENTTYPE_PREFIX:
6103 	case SADB_IDENTTYPE_FQDN:
6104 	case SADB_IDENTTYPE_USERFQDN:
6105 	default:
6106 		/* XXX do nothing */
6107 		sah->idents = NULL;
6108 		sah->identd = NULL;
6109 	 	return 0;
6110 	}
6111 
6112 	/* make structure */
6113 	sah->idents = kmem_alloc(idsrclen, KM_SLEEP);
6114 	sah->idents_len = idsrclen;
6115 	sah->identd = kmem_alloc(iddstlen, KM_SLEEP);
6116 	sah->identd_len = iddstlen;
6117 	memcpy(sah->idents, idsrc, idsrclen);
6118 	memcpy(sah->identd, iddst, iddstlen);
6119 
6120 	return 0;
6121 }
6122 
6123 /*
6124  * m will not be freed on return. It never return NULL.
6125  * it is caller's responsibility to free the result.
6126  */
6127 static struct mbuf *
6128 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
6129 {
6130 	struct mbuf *n;
6131 
6132 	KASSERT(m != NULL);
6133 	KASSERT(mhp != NULL);
6134 	KASSERT(mhp->msg != NULL);
6135 
6136 	/* create new sadb_msg to reply. */
6137 	n = key_gather_mbuf(m, mhp, 1, 15, SADB_EXT_RESERVED,
6138 	    SADB_EXT_SA, SADB_X_EXT_SA2,
6139 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
6140 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
6141 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
6142 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
6143 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
6144 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NAT_T_FRAG);
6145 
6146 	KASSERT(n->m_len >= sizeof(struct sadb_msg));
6147 
6148 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
6149 	mtod(n, struct sadb_msg *)->sadb_msg_len =
6150 	    PFKEY_UNIT64(n->m_pkthdr.len);
6151 
6152 	return n;
6153 }
6154 
6155 static int key_delete_all (struct socket *, struct mbuf *,
6156 			   const struct sadb_msghdr *, u_int16_t);
6157 
6158 /*
6159  * SADB_DELETE processing
6160  * receive
6161  *   <base, SA(*), address(SD)>
6162  * from the ikmpd, and set SADB_SASTATE_DEAD,
6163  * and send,
6164  *   <base, SA(*), address(SD)>
6165  * to the ikmpd.
6166  *
6167  * m will always be freed.
6168  */
6169 static int
6170 key_api_delete(struct socket *so, struct mbuf *m,
6171 	   const struct sadb_msghdr *mhp)
6172 {
6173 	struct sadb_sa *sa0;
6174 	const struct sockaddr *src, *dst;
6175 	struct secasindex saidx;
6176 	struct secashead *sah;
6177 	struct secasvar *sav = NULL;
6178 	u_int16_t proto;
6179 	int error;
6180 
6181 	/* map satype to proto */
6182 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
6183 	if (proto == 0) {
6184 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
6185 		return key_senderror(so, m, EINVAL);
6186 	}
6187 
6188 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
6189 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
6190 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6191 		return key_senderror(so, m, EINVAL);
6192 	}
6193 
6194 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
6195 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
6196 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6197 		return key_senderror(so, m, EINVAL);
6198 	}
6199 
6200 	if (mhp->ext[SADB_EXT_SA] == NULL) {
6201 		/*
6202 		 * Caller wants us to delete all non-LARVAL SAs
6203 		 * that match the src/dst.  This is used during
6204 		 * IKE INITIAL-CONTACT.
6205 		 */
6206 		IPSECLOG(LOG_DEBUG, "doing delete all.\n");
6207 		return key_delete_all(so, m, mhp, proto);
6208 	} else if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa)) {
6209 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6210 		return key_senderror(so, m, EINVAL);
6211 	}
6212 
6213 	sa0 = mhp->ext[SADB_EXT_SA];
6214 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6215 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6216 
6217 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
6218 	if (error != 0)
6219 		return key_senderror(so, m, EINVAL);
6220 
6221 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6222 	if (error != 0)
6223 		return key_senderror(so, m, EINVAL);
6224 
6225 	/* get a SA header */
6226 	sah = key_getsah_ref(&saidx, CMP_HEAD);
6227 	if (sah != NULL) {
6228 		/* get a SA with SPI. */
6229 		sav = key_lookup_and_remove_sav(sah, sa0->sadb_sa_spi, NULL);
6230 		key_sah_unref(sah);
6231 	}
6232 
6233 	if (sav == NULL) {
6234 		IPSECLOG(LOG_DEBUG, "no SA found.\n");
6235 		return key_senderror(so, m, ENOENT);
6236 	}
6237 
6238 	key_wait_sav(sav);
6239 	key_destroy_sav(sav);
6240 	sav = NULL;
6241 
6242     {
6243 	struct mbuf *n;
6244 
6245 	/* create new sadb_msg to reply. */
6246 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
6247 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6248 
6249 	key_fill_replymsg(n, 0);
6250 	m_freem(m);
6251 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6252     }
6253 }
6254 
6255 /*
6256  * delete all SAs for src/dst.  Called from key_api_delete().
6257  */
6258 static int
6259 key_delete_all(struct socket *so, struct mbuf *m,
6260 	       const struct sadb_msghdr *mhp, u_int16_t proto)
6261 {
6262 	const struct sockaddr *src, *dst;
6263 	struct secasindex saidx;
6264 	struct secashead *sah;
6265 	struct secasvar *sav;
6266 	u_int state;
6267 	int error;
6268 
6269 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6270 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6271 
6272 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
6273 	if (error != 0)
6274 		return key_senderror(so, m, EINVAL);
6275 
6276 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6277 	if (error != 0)
6278 		return key_senderror(so, m, EINVAL);
6279 
6280 	sah = key_getsah_ref(&saidx, CMP_HEAD);
6281 	if (sah != NULL) {
6282 		/* Delete all non-LARVAL SAs. */
6283 		SASTATE_ALIVE_FOREACH(state) {
6284 			if (state == SADB_SASTATE_LARVAL)
6285 				continue;
6286 		restart:
6287 			mutex_enter(&key_sad.lock);
6288 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
6289 				sav->state = SADB_SASTATE_DEAD;
6290 				key_unlink_sav(sav);
6291 				mutex_exit(&key_sad.lock);
6292 				key_destroy_sav(sav);
6293 				goto restart;
6294 			}
6295 			mutex_exit(&key_sad.lock);
6296 		}
6297 		key_sah_unref(sah);
6298 	}
6299     {
6300 	struct mbuf *n;
6301 
6302 	/* create new sadb_msg to reply. */
6303 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
6304 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6305 
6306 	key_fill_replymsg(n, 0);
6307 	m_freem(m);
6308 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6309     }
6310 }
6311 
6312 /*
6313  * SADB_GET processing
6314  * receive
6315  *   <base, SA(*), address(SD)>
6316  * from the ikmpd, and get a SP and a SA to respond,
6317  * and send,
6318  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6319  *       (identity(SD),) (sensitivity)>
6320  * to the ikmpd.
6321  *
6322  * m will always be freed.
6323  */
6324 static int
6325 key_api_get(struct socket *so, struct mbuf *m,
6326 	const struct sadb_msghdr *mhp)
6327 {
6328 	struct sadb_sa *sa0;
6329 	const struct sockaddr *src, *dst;
6330 	struct secasindex saidx;
6331 	struct secasvar *sav = NULL;
6332 	u_int16_t proto;
6333 	int error;
6334 
6335 	/* map satype to proto */
6336 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6337 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
6338 		return key_senderror(so, m, EINVAL);
6339 	}
6340 
6341 	if (mhp->ext[SADB_EXT_SA] == NULL ||
6342 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
6343 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
6344 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6345 		return key_senderror(so, m, EINVAL);
6346 	}
6347 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
6348 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
6349 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
6350 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6351 		return key_senderror(so, m, EINVAL);
6352 	}
6353 
6354 	sa0 = mhp->ext[SADB_EXT_SA];
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 	/* get a SA header */
6367     {
6368 	struct secashead *sah;
6369 	int s = pserialize_read_enter();
6370 
6371 	sah = key_getsah(&saidx, CMP_HEAD);
6372 	if (sah != NULL) {
6373 		/* get a SA with SPI. */
6374 		sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
6375 	}
6376 	pserialize_read_exit(s);
6377     }
6378 	if (sav == NULL) {
6379 		IPSECLOG(LOG_DEBUG, "no SA found.\n");
6380 		return key_senderror(so, m, ENOENT);
6381 	}
6382 
6383     {
6384 	struct mbuf *n;
6385 	u_int8_t satype;
6386 
6387 	/* map proto to satype */
6388 	satype = key_proto2satype(sav->sah->saidx.proto);
6389 	if (satype == 0) {
6390 		KEY_SA_UNREF(&sav);
6391 		IPSECLOG(LOG_DEBUG, "there was invalid proto in SAD.\n");
6392 		return key_senderror(so, m, EINVAL);
6393 	}
6394 
6395 	/* create new sadb_msg to reply. */
6396 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
6397 	    mhp->msg->sadb_msg_pid);
6398 	KEY_SA_UNREF(&sav);
6399 	m_freem(m);
6400 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
6401     }
6402 }
6403 
6404 /* XXX make it sysctl-configurable? */
6405 static void
6406 key_getcomb_setlifetime(struct sadb_comb *comb)
6407 {
6408 
6409 	comb->sadb_comb_soft_allocations = 1;
6410 	comb->sadb_comb_hard_allocations = 1;
6411 	comb->sadb_comb_soft_bytes = 0;
6412 	comb->sadb_comb_hard_bytes = 0;
6413 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
6414 	comb->sadb_comb_soft_addtime = comb->sadb_comb_hard_addtime * 80 / 100;
6415 	comb->sadb_comb_hard_usetime = 28800;	/* 8 hours */
6416 	comb->sadb_comb_soft_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
6417 }
6418 
6419 /*
6420  * XXX reorder combinations by preference
6421  * XXX no idea if the user wants ESP authentication or not
6422  */
6423 static struct mbuf *
6424 key_getcomb_esp(int mflag)
6425 {
6426 	struct sadb_comb *comb;
6427 	const struct enc_xform *algo;
6428 	struct mbuf *result = NULL, *m, *n;
6429 	int encmin;
6430 	int i, off, o;
6431 	int totlen;
6432 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6433 
6434 	m = NULL;
6435 	for (i = 1; i <= SADB_EALG_MAX; i++) {
6436 		algo = esp_algorithm_lookup(i);
6437 		if (algo == NULL)
6438 			continue;
6439 
6440 		/* discard algorithms with key size smaller than system min */
6441 		if (_BITS(algo->maxkey) < ipsec_esp_keymin)
6442 			continue;
6443 		if (_BITS(algo->minkey) < ipsec_esp_keymin)
6444 			encmin = ipsec_esp_keymin;
6445 		else
6446 			encmin = _BITS(algo->minkey);
6447 
6448 		if (ipsec_esp_auth)
6449 			m = key_getcomb_ah(mflag);
6450 		else {
6451 			KASSERTMSG(l <= MLEN,
6452 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
6453 			MGET(m, mflag, MT_DATA);
6454 			if (m) {
6455 				m_align(m, l);
6456 				m->m_len = l;
6457 				m->m_next = NULL;
6458 				memset(mtod(m, void *), 0, m->m_len);
6459 			}
6460 		}
6461 		if (!m)
6462 			goto fail;
6463 
6464 		totlen = 0;
6465 		for (n = m; n; n = n->m_next)
6466 			totlen += n->m_len;
6467 		KASSERTMSG((totlen % l) == 0, "totlen=%u, l=%u", totlen, l);
6468 
6469 		for (off = 0; off < totlen; off += l) {
6470 			n = m_pulldown(m, off, l, &o);
6471 			if (!n) {
6472 				/* m is already freed */
6473 				goto fail;
6474 			}
6475 			comb = (struct sadb_comb *)(mtod(n, char *) + o);
6476 			memset(comb, 0, sizeof(*comb));
6477 			key_getcomb_setlifetime(comb);
6478 			comb->sadb_comb_encrypt = i;
6479 			comb->sadb_comb_encrypt_minbits = encmin;
6480 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
6481 		}
6482 
6483 		if (!result)
6484 			result = m;
6485 		else
6486 			m_cat(result, m);
6487 	}
6488 
6489 	return result;
6490 
6491  fail:
6492 	if (result)
6493 		m_freem(result);
6494 	return NULL;
6495 }
6496 
6497 static void
6498 key_getsizes_ah(const struct auth_hash *ah, int alg,
6499 	        u_int16_t* ksmin, u_int16_t* ksmax)
6500 {
6501 	*ksmin = *ksmax = ah->keysize;
6502 	if (ah->keysize == 0) {
6503 		/*
6504 		 * Transform takes arbitrary key size but algorithm
6505 		 * key size is restricted.  Enforce this here.
6506 		 */
6507 		switch (alg) {
6508 		case SADB_X_AALG_MD5:	*ksmin = *ksmax = 16; break;
6509 		case SADB_X_AALG_SHA:	*ksmin = *ksmax = 20; break;
6510 		case SADB_X_AALG_NULL:	*ksmin = 0; *ksmax = 256; break;
6511 		default:
6512 			IPSECLOG(LOG_DEBUG, "unknown AH algorithm %u\n", alg);
6513 			break;
6514 		}
6515 	}
6516 }
6517 
6518 /*
6519  * XXX reorder combinations by preference
6520  */
6521 static struct mbuf *
6522 key_getcomb_ah(int mflag)
6523 {
6524 	struct sadb_comb *comb;
6525 	const struct auth_hash *algo;
6526 	struct mbuf *m;
6527 	u_int16_t minkeysize, maxkeysize;
6528 	int i;
6529 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6530 
6531 	m = NULL;
6532 	for (i = 1; i <= SADB_AALG_MAX; i++) {
6533 #if 1
6534 		/* we prefer HMAC algorithms, not old algorithms */
6535 		if (i != SADB_AALG_SHA1HMAC &&
6536 		    i != SADB_AALG_MD5HMAC &&
6537 		    i != SADB_X_AALG_SHA2_256 &&
6538 		    i != SADB_X_AALG_SHA2_384 &&
6539 		    i != SADB_X_AALG_SHA2_512)
6540 			continue;
6541 #endif
6542 		algo = ah_algorithm_lookup(i);
6543 		if (!algo)
6544 			continue;
6545 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6546 		/* discard algorithms with key size smaller than system min */
6547 		if (_BITS(minkeysize) < ipsec_ah_keymin)
6548 			continue;
6549 
6550 		if (!m) {
6551 			KASSERTMSG(l <= MLEN,
6552 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
6553 			MGET(m, mflag, MT_DATA);
6554 			if (m) {
6555 				m_align(m, l);
6556 				m->m_len = l;
6557 				m->m_next = NULL;
6558 			}
6559 		} else
6560 			M_PREPEND(m, l, mflag);
6561 		if (!m)
6562 			return NULL;
6563 
6564 		if (m->m_len < sizeof(struct sadb_comb)) {
6565 			m = m_pullup(m, sizeof(struct sadb_comb));
6566 			if (m == NULL)
6567 				return NULL;
6568 		}
6569 
6570 		comb = mtod(m, struct sadb_comb *);
6571 		memset(comb, 0, sizeof(*comb));
6572 		key_getcomb_setlifetime(comb);
6573 		comb->sadb_comb_auth = i;
6574 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6575 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6576 	}
6577 
6578 	return m;
6579 }
6580 
6581 /*
6582  * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6583  * XXX reorder combinations by preference
6584  */
6585 static struct mbuf *
6586 key_getcomb_ipcomp(int mflag)
6587 {
6588 	struct sadb_comb *comb;
6589 	const struct comp_algo *algo;
6590 	struct mbuf *m;
6591 	int i;
6592 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6593 
6594 	m = NULL;
6595 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6596 		algo = ipcomp_algorithm_lookup(i);
6597 		if (!algo)
6598 			continue;
6599 
6600 		if (!m) {
6601 			KASSERTMSG(l <= MLEN,
6602 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
6603 			MGET(m, mflag, MT_DATA);
6604 			if (m) {
6605 				m_align(m, l);
6606 				m->m_len = l;
6607 				m->m_next = NULL;
6608 			}
6609 		} else
6610 			M_PREPEND(m, l, mflag);
6611 		if (!m)
6612 			return NULL;
6613 
6614 		if (m->m_len < sizeof(struct sadb_comb)) {
6615 			m = m_pullup(m, sizeof(struct sadb_comb));
6616 			if (m == NULL)
6617 				return NULL;
6618 		}
6619 
6620 		comb = mtod(m, struct sadb_comb *);
6621 		memset(comb, 0, sizeof(*comb));
6622 		key_getcomb_setlifetime(comb);
6623 		comb->sadb_comb_encrypt = i;
6624 		/* what should we set into sadb_comb_*_{min,max}bits? */
6625 	}
6626 
6627 	return m;
6628 }
6629 
6630 /*
6631  * XXX no way to pass mode (transport/tunnel) to userland
6632  * XXX replay checking?
6633  * XXX sysctl interface to ipsec_{ah,esp}_keymin
6634  */
6635 static struct mbuf *
6636 key_getprop(const struct secasindex *saidx, int mflag)
6637 {
6638 	struct sadb_prop *prop;
6639 	struct mbuf *m, *n;
6640 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6641 	int totlen;
6642 
6643 	switch (saidx->proto)  {
6644 	case IPPROTO_ESP:
6645 		m = key_getcomb_esp(mflag);
6646 		break;
6647 	case IPPROTO_AH:
6648 		m = key_getcomb_ah(mflag);
6649 		break;
6650 	case IPPROTO_IPCOMP:
6651 		m = key_getcomb_ipcomp(mflag);
6652 		break;
6653 	default:
6654 		return NULL;
6655 	}
6656 
6657 	if (!m)
6658 		return NULL;
6659 	M_PREPEND(m, l, mflag);
6660 	if (!m)
6661 		return NULL;
6662 
6663 	totlen = 0;
6664 	for (n = m; n; n = n->m_next)
6665 		totlen += n->m_len;
6666 
6667 	prop = mtod(m, struct sadb_prop *);
6668 	memset(prop, 0, sizeof(*prop));
6669 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6670 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6671 	prop->sadb_prop_replay = 32;	/* XXX */
6672 
6673 	return m;
6674 }
6675 
6676 /*
6677  * SADB_ACQUIRE processing called by key_checkrequest() and key_api_acquire().
6678  * send
6679  *   <base, SA, address(SD), (address(P)), x_policy,
6680  *       (identity(SD),) (sensitivity,) proposal>
6681  * to KMD, and expect to receive
6682  *   <base> with SADB_ACQUIRE if error occurred,
6683  * or
6684  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6685  * from KMD by PF_KEY.
6686  *
6687  * XXX x_policy is outside of RFC2367 (KAME extension).
6688  * XXX sensitivity is not supported.
6689  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6690  * see comment for key_getcomb_ipcomp().
6691  *
6692  * OUT:
6693  *    0     : succeed
6694  *    others: error number
6695  */
6696 static int
6697 key_acquire(const struct secasindex *saidx, const struct secpolicy *sp, int mflag)
6698 {
6699 	struct mbuf *result = NULL, *m;
6700 #ifndef IPSEC_NONBLOCK_ACQUIRE
6701 	struct secacq *newacq;
6702 #endif
6703 	u_int8_t satype;
6704 	int error = -1;
6705 	u_int32_t seq;
6706 
6707 	/* sanity check */
6708 	KASSERT(saidx != NULL);
6709 	satype = key_proto2satype(saidx->proto);
6710 	KASSERTMSG(satype != 0, "null satype, protocol %u", saidx->proto);
6711 
6712 #ifndef IPSEC_NONBLOCK_ACQUIRE
6713 	/*
6714 	 * We never do anything about acquirng SA.  There is anather
6715 	 * solution that kernel blocks to send SADB_ACQUIRE message until
6716 	 * getting something message from IKEd.  In later case, to be
6717 	 * managed with ACQUIRING list.
6718 	 */
6719 	/* Get an entry to check whether sending message or not. */
6720 	mutex_enter(&key_misc.lock);
6721 	newacq = key_getacq(saidx);
6722 	if (newacq != NULL) {
6723 		if (key_blockacq_count < newacq->count) {
6724 			/* reset counter and do send message. */
6725 			newacq->count = 0;
6726 		} else {
6727 			/* increment counter and do nothing. */
6728 			newacq->count++;
6729 			mutex_exit(&key_misc.lock);
6730 			return 0;
6731 		}
6732 	} else {
6733 		/* make new entry for blocking to send SADB_ACQUIRE. */
6734 		newacq = key_newacq(saidx);
6735 		if (newacq == NULL) {
6736 			mutex_exit(&key_misc.lock);
6737 			return ENOBUFS;
6738 		}
6739 
6740 		/* add to key_misc.acqlist */
6741 		LIST_INSERT_HEAD(&key_misc.acqlist, newacq, chain);
6742 	}
6743 
6744 	seq = newacq->seq;
6745 	mutex_exit(&key_misc.lock);
6746 #else
6747 	seq = (acq_seq = (acq_seq == ~0 ? 1 : ++acq_seq));
6748 #endif
6749 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0, mflag);
6750 	if (!m) {
6751 		error = ENOBUFS;
6752 		goto fail;
6753 	}
6754 	result = m;
6755 
6756 	/* set sadb_address for saidx's. */
6757 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &saidx->src.sa, FULLMASK,
6758 	    IPSEC_ULPROTO_ANY, mflag);
6759 	if (!m) {
6760 		error = ENOBUFS;
6761 		goto fail;
6762 	}
6763 	m_cat(result, m);
6764 
6765 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &saidx->dst.sa, FULLMASK,
6766 	    IPSEC_ULPROTO_ANY, mflag);
6767 	if (!m) {
6768 		error = ENOBUFS;
6769 		goto fail;
6770 	}
6771 	m_cat(result, m);
6772 
6773 	/* XXX proxy address (optional) */
6774 
6775 	/* set sadb_x_policy */
6776 	if (sp) {
6777 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6778 		    mflag);
6779 		if (!m) {
6780 			error = ENOBUFS;
6781 			goto fail;
6782 		}
6783 		m_cat(result, m);
6784 	}
6785 
6786 	/* XXX identity (optional) */
6787 #if 0
6788 	if (idexttype && fqdn) {
6789 		/* create identity extension (FQDN) */
6790 		struct sadb_ident *id;
6791 		int fqdnlen;
6792 
6793 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
6794 		id = (struct sadb_ident *)p;
6795 		memset(id, 0, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6796 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6797 		id->sadb_ident_exttype = idexttype;
6798 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
6799 		memcpy(id + 1, fqdn, fqdnlen);
6800 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
6801 	}
6802 
6803 	if (idexttype) {
6804 		/* create identity extension (USERFQDN) */
6805 		struct sadb_ident *id;
6806 		int userfqdnlen;
6807 
6808 		if (userfqdn) {
6809 			/* +1 for terminating-NUL */
6810 			userfqdnlen = strlen(userfqdn) + 1;
6811 		} else
6812 			userfqdnlen = 0;
6813 		id = (struct sadb_ident *)p;
6814 		memset(id, 0, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6815 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6816 		id->sadb_ident_exttype = idexttype;
6817 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
6818 		/* XXX is it correct? */
6819 		if (curlwp)
6820 			id->sadb_ident_id = kauth_cred_getuid(curlwp->l_cred);
6821 		if (userfqdn && userfqdnlen)
6822 			memcpy(id + 1, userfqdn, userfqdnlen);
6823 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
6824 	}
6825 #endif
6826 
6827 	/* XXX sensitivity (optional) */
6828 
6829 	/* create proposal/combination extension */
6830 	m = key_getprop(saidx, mflag);
6831 #if 0
6832 	/*
6833 	 * spec conformant: always attach proposal/combination extension,
6834 	 * the problem is that we have no way to attach it for ipcomp,
6835 	 * due to the way sadb_comb is declared in RFC2367.
6836 	 */
6837 	if (!m) {
6838 		error = ENOBUFS;
6839 		goto fail;
6840 	}
6841 	m_cat(result, m);
6842 #else
6843 	/*
6844 	 * outside of spec; make proposal/combination extension optional.
6845 	 */
6846 	if (m)
6847 		m_cat(result, m);
6848 #endif
6849 
6850 	KASSERT(result->m_flags & M_PKTHDR);
6851 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
6852 
6853 	result->m_pkthdr.len = 0;
6854 	for (m = result; m; m = m->m_next)
6855 		result->m_pkthdr.len += m->m_len;
6856 
6857 	mtod(result, struct sadb_msg *)->sadb_msg_len =
6858 	    PFKEY_UNIT64(result->m_pkthdr.len);
6859 
6860 	/*
6861 	 * Called from key_api_acquire that must come from userland, so
6862 	 * we can call key_sendup_mbuf immediately.
6863 	 */
6864 	if (mflag == M_WAITOK)
6865 		return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
6866 	/*
6867 	 * XXX we cannot call key_sendup_mbuf directly here because
6868 	 * it can cause a deadlock:
6869 	 * - We have a reference to an SP (and an SA) here
6870 	 * - key_sendup_mbuf will try to take key_so_mtx
6871 	 * - Some other thread may try to localcount_drain to the SP with
6872 	 *   holding key_so_mtx in say key_api_spdflush
6873 	 * - In this case localcount_drain never return because key_sendup_mbuf
6874 	 *   that has stuck on key_so_mtx never release a reference to the SP
6875 	 *
6876 	 * So defer key_sendup_mbuf to the timer.
6877 	 */
6878 	return key_acquire_sendup_mbuf_later(result);
6879 
6880  fail:
6881 	if (result)
6882 		m_freem(result);
6883 	return error;
6884 }
6885 
6886 static struct mbuf *key_acquire_mbuf_head = NULL;
6887 static unsigned key_acquire_mbuf_count = 0;
6888 #define KEY_ACQUIRE_MBUF_MAX	10
6889 
6890 static void
6891 key_acquire_sendup_pending_mbuf(void)
6892 {
6893 	struct mbuf *m, *prev;
6894 	int error;
6895 
6896 again:
6897 	prev = NULL;
6898 	mutex_enter(&key_misc.lock);
6899 	m = key_acquire_mbuf_head;
6900 	/* Get an earliest mbuf (one at the tail of the list) */
6901 	while (m != NULL) {
6902 		if (m->m_nextpkt == NULL) {
6903 			if (prev != NULL)
6904 				prev->m_nextpkt = NULL;
6905 			if (m == key_acquire_mbuf_head)
6906 				key_acquire_mbuf_head = NULL;
6907 			key_acquire_mbuf_count--;
6908 			break;
6909 		}
6910 		prev = m;
6911 		m = m->m_nextpkt;
6912 	}
6913 	mutex_exit(&key_misc.lock);
6914 
6915 	if (m == NULL)
6916 		return;
6917 
6918 	m->m_nextpkt = NULL;
6919 	error = key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
6920 	if (error != 0)
6921 		IPSECLOG(LOG_WARNING, "key_sendup_mbuf failed (error=%d)\n",
6922 		    error);
6923 
6924 	if (prev != NULL)
6925 		goto again;
6926 }
6927 
6928 static int
6929 key_acquire_sendup_mbuf_later(struct mbuf *m)
6930 {
6931 
6932 	mutex_enter(&key_misc.lock);
6933 	/* Avoid queuing too much mbufs */
6934 	if (key_acquire_mbuf_count >= KEY_ACQUIRE_MBUF_MAX) {
6935 		mutex_exit(&key_misc.lock);
6936 		m_freem(m);
6937 		return ENOBUFS; /* XXX */
6938 	}
6939 	/* Enqueue mbuf at the head of the list */
6940 	m->m_nextpkt = key_acquire_mbuf_head;
6941 	key_acquire_mbuf_head = m;
6942 	key_acquire_mbuf_count++;
6943 	mutex_exit(&key_misc.lock);
6944 
6945 	/* Kick the timer */
6946 	key_timehandler(NULL);
6947 
6948 	return 0;
6949 }
6950 
6951 #ifndef IPSEC_NONBLOCK_ACQUIRE
6952 static struct secacq *
6953 key_newacq(const struct secasindex *saidx)
6954 {
6955 	struct secacq *newacq;
6956 
6957 	/* get new entry */
6958 	newacq = kmem_intr_zalloc(sizeof(struct secacq), KM_NOSLEEP);
6959 	if (newacq == NULL) {
6960 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
6961 		return NULL;
6962 	}
6963 
6964 	/* copy secindex */
6965 	memcpy(&newacq->saidx, saidx, sizeof(newacq->saidx));
6966 	newacq->seq = (acq_seq == ~0 ? 1 : ++acq_seq);
6967 	newacq->created = time_uptime;
6968 	newacq->count = 0;
6969 
6970 	return newacq;
6971 }
6972 
6973 static struct secacq *
6974 key_getacq(const struct secasindex *saidx)
6975 {
6976 	struct secacq *acq;
6977 
6978 	KASSERT(mutex_owned(&key_misc.lock));
6979 
6980 	LIST_FOREACH(acq, &key_misc.acqlist, chain) {
6981 		if (key_saidx_match(saidx, &acq->saidx, CMP_EXACTLY))
6982 			return acq;
6983 	}
6984 
6985 	return NULL;
6986 }
6987 
6988 static struct secacq *
6989 key_getacqbyseq(u_int32_t seq)
6990 {
6991 	struct secacq *acq;
6992 
6993 	KASSERT(mutex_owned(&key_misc.lock));
6994 
6995 	LIST_FOREACH(acq, &key_misc.acqlist, chain) {
6996 		if (acq->seq == seq)
6997 			return acq;
6998 	}
6999 
7000 	return NULL;
7001 }
7002 #endif
7003 
7004 #ifdef notyet
7005 static struct secspacq *
7006 key_newspacq(const struct secpolicyindex *spidx)
7007 {
7008 	struct secspacq *acq;
7009 
7010 	/* get new entry */
7011 	acq = kmem_intr_zalloc(sizeof(struct secspacq), KM_NOSLEEP);
7012 	if (acq == NULL) {
7013 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
7014 		return NULL;
7015 	}
7016 
7017 	/* copy secindex */
7018 	memcpy(&acq->spidx, spidx, sizeof(acq->spidx));
7019 	acq->created = time_uptime;
7020 	acq->count = 0;
7021 
7022 	return acq;
7023 }
7024 
7025 static struct secspacq *
7026 key_getspacq(const struct secpolicyindex *spidx)
7027 {
7028 	struct secspacq *acq;
7029 
7030 	LIST_FOREACH(acq, &key_misc.spacqlist, chain) {
7031 		if (key_spidx_match_exactly(spidx, &acq->spidx))
7032 			return acq;
7033 	}
7034 
7035 	return NULL;
7036 }
7037 #endif /* notyet */
7038 
7039 /*
7040  * SADB_ACQUIRE processing,
7041  * in first situation, is receiving
7042  *   <base>
7043  * from the ikmpd, and clear sequence of its secasvar entry.
7044  *
7045  * In second situation, is receiving
7046  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
7047  * from a user land process, and return
7048  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
7049  * to the socket.
7050  *
7051  * m will always be freed.
7052  */
7053 static int
7054 key_api_acquire(struct socket *so, struct mbuf *m,
7055       	     const struct sadb_msghdr *mhp)
7056 {
7057 	const struct sockaddr *src, *dst;
7058 	struct secasindex saidx;
7059 	u_int16_t proto;
7060 	int error;
7061 
7062 	/*
7063 	 * Error message from KMd.
7064 	 * We assume that if error was occurred in IKEd, the length of PFKEY
7065 	 * message is equal to the size of sadb_msg structure.
7066 	 * We do not raise error even if error occurred in this function.
7067 	 */
7068 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
7069 #ifndef IPSEC_NONBLOCK_ACQUIRE
7070 		struct secacq *acq;
7071 
7072 		/* check sequence number */
7073 		if (mhp->msg->sadb_msg_seq == 0) {
7074 			IPSECLOG(LOG_DEBUG, "must specify sequence number.\n");
7075 			m_freem(m);
7076 			return 0;
7077 		}
7078 
7079 		mutex_enter(&key_misc.lock);
7080 		acq = key_getacqbyseq(mhp->msg->sadb_msg_seq);
7081 		if (acq == NULL) {
7082 			mutex_exit(&key_misc.lock);
7083 			/*
7084 			 * the specified larval SA is already gone, or we got
7085 			 * a bogus sequence number.  we can silently ignore it.
7086 			 */
7087 			m_freem(m);
7088 			return 0;
7089 		}
7090 
7091 		/* reset acq counter in order to deletion by timehander. */
7092 		acq->created = time_uptime;
7093 		acq->count = 0;
7094 		mutex_exit(&key_misc.lock);
7095 #endif
7096 		m_freem(m);
7097 		return 0;
7098 	}
7099 
7100 	/*
7101 	 * This message is from user land.
7102 	 */
7103 
7104 	/* map satype to proto */
7105 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
7106 	if (proto == 0) {
7107 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
7108 		return key_senderror(so, m, EINVAL);
7109 	}
7110 
7111 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
7112 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
7113 	    mhp->ext[SADB_EXT_PROPOSAL] == NULL) {
7114 		/* error */
7115 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
7116 		return key_senderror(so, m, EINVAL);
7117 	}
7118 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
7119 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
7120 	    mhp->extlen[SADB_EXT_PROPOSAL] < sizeof(struct sadb_prop)) {
7121 		/* error */
7122 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
7123 		return key_senderror(so, m, EINVAL);
7124 	}
7125 
7126 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
7127 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
7128 
7129 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
7130 	if (error != 0)
7131 		return key_senderror(so, m, EINVAL);
7132 
7133 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
7134 	if (error != 0)
7135 		return key_senderror(so, m, EINVAL);
7136 
7137 	/* get a SA index */
7138     {
7139 	struct secashead *sah;
7140 	int s = pserialize_read_enter();
7141 
7142 	sah = key_getsah(&saidx, CMP_MODE_REQID);
7143 	if (sah != NULL) {
7144 		pserialize_read_exit(s);
7145 		IPSECLOG(LOG_DEBUG, "a SA exists already.\n");
7146 		return key_senderror(so, m, EEXIST);
7147 	}
7148 	pserialize_read_exit(s);
7149     }
7150 
7151 	error = key_acquire(&saidx, NULL, M_WAITOK);
7152 	if (error != 0) {
7153 		IPSECLOG(LOG_DEBUG, "error %d returned from key_acquire.\n",
7154 		    error);
7155 		return key_senderror(so, m, error);
7156 	}
7157 
7158 	return key_sendup_mbuf(so, m, KEY_SENDUP_REGISTERED);
7159 }
7160 
7161 /*
7162  * SADB_REGISTER processing.
7163  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
7164  * receive
7165  *   <base>
7166  * from the ikmpd, and register a socket to send PF_KEY messages,
7167  * and send
7168  *   <base, supported>
7169  * to KMD by PF_KEY.
7170  * If socket is detached, must free from regnode.
7171  *
7172  * m will always be freed.
7173  */
7174 static int
7175 key_api_register(struct socket *so, struct mbuf *m,
7176 	     const struct sadb_msghdr *mhp)
7177 {
7178 	struct secreg *reg, *newreg = 0;
7179 
7180 	/* check for invalid register message */
7181 	if (mhp->msg->sadb_msg_satype >= __arraycount(key_misc.reglist))
7182 		return key_senderror(so, m, EINVAL);
7183 
7184 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
7185 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
7186 		goto setmsg;
7187 
7188 	/* Allocate regnode in advance, out of mutex */
7189 	newreg = kmem_zalloc(sizeof(*newreg), KM_SLEEP);
7190 
7191 	/* check whether existing or not */
7192 	mutex_enter(&key_misc.lock);
7193 	LIST_FOREACH(reg, &key_misc.reglist[mhp->msg->sadb_msg_satype], chain) {
7194 		if (reg->so == so) {
7195 			IPSECLOG(LOG_DEBUG, "socket exists already.\n");
7196 			mutex_exit(&key_misc.lock);
7197 			kmem_free(newreg, sizeof(*newreg));
7198 			return key_senderror(so, m, EEXIST);
7199 		}
7200 	}
7201 
7202 	newreg->so = so;
7203 	((struct keycb *)sotorawcb(so))->kp_registered++;
7204 
7205 	/* add regnode to key_misc.reglist. */
7206 	LIST_INSERT_HEAD(&key_misc.reglist[mhp->msg->sadb_msg_satype], newreg, chain);
7207 	mutex_exit(&key_misc.lock);
7208 
7209   setmsg:
7210     {
7211 	struct mbuf *n;
7212 	struct sadb_supported *sup;
7213 	u_int len, alen, elen;
7214 	int off;
7215 	int i;
7216 	struct sadb_alg *alg;
7217 
7218 	/* create new sadb_msg to reply. */
7219 	alen = 0;
7220 	for (i = 1; i <= SADB_AALG_MAX; i++) {
7221 		if (ah_algorithm_lookup(i))
7222 			alen += sizeof(struct sadb_alg);
7223 	}
7224 	if (alen)
7225 		alen += sizeof(struct sadb_supported);
7226 	elen = 0;
7227 	for (i = 1; i <= SADB_EALG_MAX; i++) {
7228 		if (esp_algorithm_lookup(i))
7229 			elen += sizeof(struct sadb_alg);
7230 	}
7231 	if (elen)
7232 		elen += sizeof(struct sadb_supported);
7233 
7234 	len = sizeof(struct sadb_msg) + alen + elen;
7235 
7236 	if (len > MCLBYTES)
7237 		return key_senderror(so, m, ENOBUFS);
7238 
7239 	n = key_alloc_mbuf_simple(len, M_WAITOK);
7240 	n->m_pkthdr.len = n->m_len = len;
7241 	n->m_next = NULL;
7242 	off = 0;
7243 
7244 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
7245 	key_fill_replymsg(n, 0);
7246 
7247 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7248 
7249 	/* for authentication algorithm */
7250 	if (alen) {
7251 		sup = (struct sadb_supported *)(mtod(n, char *) + off);
7252 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7253 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7254 		sup->sadb_supported_reserved = 0;
7255 		off += PFKEY_ALIGN8(sizeof(*sup));
7256 
7257 		for (i = 1; i <= SADB_AALG_MAX; i++) {
7258 			const struct auth_hash *aalgo;
7259 			u_int16_t minkeysize, maxkeysize;
7260 
7261 			aalgo = ah_algorithm_lookup(i);
7262 			if (!aalgo)
7263 				continue;
7264 			alg = (struct sadb_alg *)(mtod(n, char *) + off);
7265 			alg->sadb_alg_id = i;
7266 			alg->sadb_alg_ivlen = 0;
7267 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7268 			alg->sadb_alg_minbits = _BITS(minkeysize);
7269 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7270 			alg->sadb_alg_reserved = 0;
7271 			off += PFKEY_ALIGN8(sizeof(*alg));
7272 		}
7273 	}
7274 
7275 	/* for encryption algorithm */
7276 	if (elen) {
7277 		sup = (struct sadb_supported *)(mtod(n, char *) + off);
7278 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7279 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7280 		sup->sadb_supported_reserved = 0;
7281 		off += PFKEY_ALIGN8(sizeof(*sup));
7282 
7283 		for (i = 1; i <= SADB_EALG_MAX; i++) {
7284 			const struct enc_xform *ealgo;
7285 
7286 			ealgo = esp_algorithm_lookup(i);
7287 			if (!ealgo)
7288 				continue;
7289 			alg = (struct sadb_alg *)(mtod(n, char *) + off);
7290 			alg->sadb_alg_id = i;
7291 			alg->sadb_alg_ivlen = ealgo->blocksize;
7292 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7293 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7294 			alg->sadb_alg_reserved = 0;
7295 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7296 		}
7297 	}
7298 
7299 	KASSERTMSG(off == len, "length inconsistency");
7300 
7301 	m_freem(m);
7302 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7303     }
7304 }
7305 
7306 /*
7307  * free secreg entry registered.
7308  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7309  */
7310 void
7311 key_freereg(struct socket *so)
7312 {
7313 	struct secreg *reg;
7314 	int i;
7315 
7316 	KASSERT(!cpu_softintr_p());
7317 	KASSERT(so != NULL);
7318 
7319 	/*
7320 	 * check whether existing or not.
7321 	 * check all type of SA, because there is a potential that
7322 	 * one socket is registered to multiple type of SA.
7323 	 */
7324 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7325 		mutex_enter(&key_misc.lock);
7326 		LIST_FOREACH(reg, &key_misc.reglist[i], chain) {
7327 			if (reg->so == so) {
7328 				LIST_REMOVE(reg, chain);
7329 				break;
7330 			}
7331 		}
7332 		mutex_exit(&key_misc.lock);
7333 		if (reg != NULL)
7334 			kmem_free(reg, sizeof(*reg));
7335 	}
7336 
7337 	return;
7338 }
7339 
7340 /*
7341  * SADB_EXPIRE processing
7342  * send
7343  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7344  * to KMD by PF_KEY.
7345  * NOTE: We send only soft lifetime extension.
7346  *
7347  * OUT:	0	: succeed
7348  *	others	: error number
7349  */
7350 static int
7351 key_expire(struct secasvar *sav)
7352 {
7353 	int s;
7354 	int satype;
7355 	struct mbuf *result = NULL, *m;
7356 	int len;
7357 	int error = -1;
7358 	struct sadb_lifetime *lt;
7359 	lifetime_counters_t sum = {0};
7360 
7361 	/* XXX: Why do we lock ? */
7362 	s = splsoftnet();	/*called from softclock()*/
7363 
7364 	KASSERT(sav != NULL);
7365 
7366 	satype = key_proto2satype(sav->sah->saidx.proto);
7367 	KASSERTMSG(satype != 0, "invalid proto is passed");
7368 
7369 	/* set msg header */
7370 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, key_sa_refcnt(sav),
7371 	    M_WAITOK);
7372 	result = m;
7373 
7374 	/* create SA extension */
7375 	m = key_setsadbsa(sav);
7376 	m_cat(result, m);
7377 
7378 	/* create SA extension */
7379 	m = key_setsadbxsa2(sav->sah->saidx.mode,
7380 	    sav->replay ? sav->replay->count : 0, sav->sah->saidx.reqid);
7381 	m_cat(result, m);
7382 
7383 	/* create lifetime extension (current and soft) */
7384 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7385 	m = key_alloc_mbuf(len, M_WAITOK);
7386 	KASSERT(m->m_next == NULL);
7387 
7388 	memset(mtod(m, void *), 0, len);
7389 	lt = mtod(m, struct sadb_lifetime *);
7390 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7391 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7392 	percpu_foreach(sav->lft_c_counters_percpu,
7393 	    key_sum_lifetime_counters, sum);
7394 	lt->sadb_lifetime_allocations = sum[LIFETIME_COUNTER_ALLOCATIONS];
7395 	lt->sadb_lifetime_bytes = sum[LIFETIME_COUNTER_BYTES];
7396 	lt->sadb_lifetime_addtime =
7397 	    time_mono_to_wall(sav->lft_c->sadb_lifetime_addtime);
7398 	lt->sadb_lifetime_usetime =
7399 	    time_mono_to_wall(sav->lft_c->sadb_lifetime_usetime);
7400 	lt = (struct sadb_lifetime *)(mtod(m, char *) + len / 2);
7401 	memcpy(lt, sav->lft_s, sizeof(*lt));
7402 	m_cat(result, m);
7403 
7404 	/* set sadb_address for source */
7405 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa,
7406 	    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
7407 	m_cat(result, m);
7408 
7409 	/* set sadb_address for destination */
7410 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.dst.sa,
7411 	    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
7412 	m_cat(result, m);
7413 
7414 	if ((result->m_flags & M_PKTHDR) == 0) {
7415 		error = EINVAL;
7416 		goto fail;
7417 	}
7418 
7419 	if (result->m_len < sizeof(struct sadb_msg)) {
7420 		result = m_pullup(result, sizeof(struct sadb_msg));
7421 		if (result == NULL) {
7422 			error = ENOBUFS;
7423 			goto fail;
7424 		}
7425 	}
7426 
7427 	result->m_pkthdr.len = 0;
7428 	for (m = result; m; m = m->m_next)
7429 		result->m_pkthdr.len += m->m_len;
7430 
7431 	mtod(result, struct sadb_msg *)->sadb_msg_len =
7432 	    PFKEY_UNIT64(result->m_pkthdr.len);
7433 
7434 	splx(s);
7435 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7436 
7437  fail:
7438 	if (result)
7439 		m_freem(result);
7440 	splx(s);
7441 	return error;
7442 }
7443 
7444 /*
7445  * SADB_FLUSH processing
7446  * receive
7447  *   <base>
7448  * from the ikmpd, and free all entries in secastree.
7449  * and send,
7450  *   <base>
7451  * to the ikmpd.
7452  * NOTE: to do is only marking SADB_SASTATE_DEAD.
7453  *
7454  * m will always be freed.
7455  */
7456 static int
7457 key_api_flush(struct socket *so, struct mbuf *m,
7458           const struct sadb_msghdr *mhp)
7459 {
7460 	struct sadb_msg *newmsg;
7461 	struct secashead *sah;
7462 	struct secasvar *sav;
7463 	u_int16_t proto;
7464 	u_int8_t state;
7465 	int s;
7466 
7467 	/* map satype to proto */
7468 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
7469 	if (proto == 0) {
7470 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
7471 		return key_senderror(so, m, EINVAL);
7472 	}
7473 
7474 	/* no SATYPE specified, i.e. flushing all SA. */
7475 	s = pserialize_read_enter();
7476 	SAHLIST_READER_FOREACH(sah) {
7477 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7478 		    proto != sah->saidx.proto)
7479 			continue;
7480 
7481 		key_sah_ref(sah);
7482 		pserialize_read_exit(s);
7483 
7484 		SASTATE_ALIVE_FOREACH(state) {
7485 		restart:
7486 			mutex_enter(&key_sad.lock);
7487 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
7488 				sav->state = SADB_SASTATE_DEAD;
7489 				key_unlink_sav(sav);
7490 				mutex_exit(&key_sad.lock);
7491 				key_destroy_sav(sav);
7492 				goto restart;
7493 			}
7494 			mutex_exit(&key_sad.lock);
7495 		}
7496 
7497 		s = pserialize_read_enter();
7498 		sah->state = SADB_SASTATE_DEAD;
7499 		key_sah_unref(sah);
7500 	}
7501 	pserialize_read_exit(s);
7502 
7503 	if (m->m_len < sizeof(struct sadb_msg) ||
7504 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7505 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
7506 		return key_senderror(so, m, ENOBUFS);
7507 	}
7508 
7509 	if (m->m_next)
7510 		m_freem(m->m_next);
7511 	m->m_next = NULL;
7512 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7513 	newmsg = mtod(m, struct sadb_msg *);
7514 	newmsg->sadb_msg_errno = 0;
7515 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7516 
7517 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7518 }
7519 
7520 
7521 static struct mbuf *
7522 key_setdump_chain(u_int8_t req_satype, int *errorp, int *lenp, pid_t pid)
7523 {
7524 	struct secashead *sah;
7525 	struct secasvar *sav;
7526 	u_int16_t proto;
7527 	u_int8_t satype;
7528 	u_int8_t state;
7529 	int cnt;
7530 	struct mbuf *m, *n, *prev;
7531 
7532 	KASSERT(mutex_owned(&key_sad.lock));
7533 
7534 	*lenp = 0;
7535 
7536 	/* map satype to proto */
7537 	proto = key_satype2proto(req_satype);
7538 	if (proto == 0) {
7539 		*errorp = EINVAL;
7540 		return (NULL);
7541 	}
7542 
7543 	/* count sav entries to be sent to userland. */
7544 	cnt = 0;
7545 	SAHLIST_WRITER_FOREACH(sah) {
7546 		if (req_satype != SADB_SATYPE_UNSPEC &&
7547 		    proto != sah->saidx.proto)
7548 			continue;
7549 
7550 		SASTATE_ANY_FOREACH(state) {
7551 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
7552 				cnt++;
7553 			}
7554 		}
7555 	}
7556 
7557 	if (cnt == 0) {
7558 		*errorp = ENOENT;
7559 		return (NULL);
7560 	}
7561 
7562 	/* send this to the userland, one at a time. */
7563 	m = NULL;
7564 	prev = m;
7565 	SAHLIST_WRITER_FOREACH(sah) {
7566 		if (req_satype != SADB_SATYPE_UNSPEC &&
7567 		    proto != sah->saidx.proto)
7568 			continue;
7569 
7570 		/* map proto to satype */
7571 		satype = key_proto2satype(sah->saidx.proto);
7572 		if (satype == 0) {
7573 			m_freem(m);
7574 			*errorp = EINVAL;
7575 			return (NULL);
7576 		}
7577 
7578 		SASTATE_ANY_FOREACH(state) {
7579 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
7580 				n = key_setdumpsa(sav, SADB_DUMP, satype,
7581 				    --cnt, pid);
7582 				if (!m)
7583 					m = n;
7584 				else
7585 					prev->m_nextpkt = n;
7586 				prev = n;
7587 			}
7588 		}
7589 	}
7590 
7591 	if (!m) {
7592 		*errorp = EINVAL;
7593 		return (NULL);
7594 	}
7595 
7596 	if ((m->m_flags & M_PKTHDR) != 0) {
7597 		m->m_pkthdr.len = 0;
7598 		for (n = m; n; n = n->m_next)
7599 			m->m_pkthdr.len += n->m_len;
7600 	}
7601 
7602 	*errorp = 0;
7603 	return (m);
7604 }
7605 
7606 /*
7607  * SADB_DUMP processing
7608  * dump all entries including status of DEAD in SAD.
7609  * receive
7610  *   <base>
7611  * from the ikmpd, and dump all secasvar leaves
7612  * and send,
7613  *   <base> .....
7614  * to the ikmpd.
7615  *
7616  * m will always be freed.
7617  */
7618 static int
7619 key_api_dump(struct socket *so, struct mbuf *m0,
7620 	 const struct sadb_msghdr *mhp)
7621 {
7622 	u_int16_t proto;
7623 	u_int8_t satype;
7624 	struct mbuf *n;
7625 	int error, len, ok;
7626 
7627 	/* map satype to proto */
7628 	satype = mhp->msg->sadb_msg_satype;
7629 	proto = key_satype2proto(satype);
7630 	if (proto == 0) {
7631 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
7632 		return key_senderror(so, m0, EINVAL);
7633 	}
7634 
7635 	/*
7636 	 * If the requestor has insufficient socket-buffer space
7637 	 * for the entire chain, nobody gets any response to the DUMP.
7638 	 * XXX For now, only the requestor ever gets anything.
7639 	 * Moreover, if the requestor has any space at all, they receive
7640 	 * the entire chain, otherwise the request is refused with ENOBUFS.
7641 	 */
7642 	if (sbspace(&so->so_rcv) <= 0) {
7643 		return key_senderror(so, m0, ENOBUFS);
7644 	}
7645 
7646 	mutex_enter(&key_sad.lock);
7647 	n = key_setdump_chain(satype, &error, &len, mhp->msg->sadb_msg_pid);
7648 	mutex_exit(&key_sad.lock);
7649 
7650 	if (n == NULL) {
7651 		return key_senderror(so, m0, ENOENT);
7652 	}
7653 	{
7654 		uint64_t *ps = PFKEY_STAT_GETREF();
7655 		ps[PFKEY_STAT_IN_TOTAL]++;
7656 		ps[PFKEY_STAT_IN_BYTES] += len;
7657 		PFKEY_STAT_PUTREF();
7658 	}
7659 
7660 	/*
7661 	 * PF_KEY DUMP responses are no longer broadcast to all PF_KEY sockets.
7662 	 * The requestor receives either the entire chain, or an
7663 	 * error message with ENOBUFS.
7664 	 *
7665 	 * sbappendaddrchain() takes the chain of entries, one
7666 	 * packet-record per SPD entry, prepends the key_src sockaddr
7667 	 * to each packet-record, links the sockaddr mbufs into a new
7668 	 * list of records, then   appends the entire resulting
7669 	 * list to the requesting socket.
7670 	 */
7671 	ok = sbappendaddrchain(&so->so_rcv, (struct sockaddr *)&key_src, n,
7672 	    SB_PRIO_ONESHOT_OVERFLOW);
7673 
7674 	if (!ok) {
7675 		PFKEY_STATINC(PFKEY_STAT_IN_NOMEM);
7676 		m_freem(n);
7677 		return key_senderror(so, m0, ENOBUFS);
7678 	}
7679 
7680 	m_freem(m0);
7681 	return 0;
7682 }
7683 
7684 /*
7685  * SADB_X_PROMISC processing
7686  *
7687  * m will always be freed.
7688  */
7689 static int
7690 key_api_promisc(struct socket *so, struct mbuf *m,
7691 	    const struct sadb_msghdr *mhp)
7692 {
7693 	int olen;
7694 
7695 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7696 
7697 	if (olen < sizeof(struct sadb_msg)) {
7698 #if 1
7699 		return key_senderror(so, m, EINVAL);
7700 #else
7701 		m_freem(m);
7702 		return 0;
7703 #endif
7704 	} else if (olen == sizeof(struct sadb_msg)) {
7705 		/* enable/disable promisc mode */
7706 		struct keycb *kp = (struct keycb *)sotorawcb(so);
7707 		if (kp == NULL)
7708 			return key_senderror(so, m, EINVAL);
7709 		mhp->msg->sadb_msg_errno = 0;
7710 		switch (mhp->msg->sadb_msg_satype) {
7711 		case 0:
7712 		case 1:
7713 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
7714 			break;
7715 		default:
7716 			return key_senderror(so, m, EINVAL);
7717 		}
7718 
7719 		/* send the original message back to everyone */
7720 		mhp->msg->sadb_msg_errno = 0;
7721 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7722 	} else {
7723 		/* send packet as is */
7724 
7725 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
7726 
7727 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
7728 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7729 	}
7730 }
7731 
7732 static int (*key_api_typesw[]) (struct socket *, struct mbuf *,
7733 		const struct sadb_msghdr *) = {
7734 	NULL,			/* SADB_RESERVED */
7735 	key_api_getspi,		/* SADB_GETSPI */
7736 	key_api_update,		/* SADB_UPDATE */
7737 	key_api_add,		/* SADB_ADD */
7738 	key_api_delete,		/* SADB_DELETE */
7739 	key_api_get,		/* SADB_GET */
7740 	key_api_acquire,	/* SADB_ACQUIRE */
7741 	key_api_register,	/* SADB_REGISTER */
7742 	NULL,			/* SADB_EXPIRE */
7743 	key_api_flush,		/* SADB_FLUSH */
7744 	key_api_dump,		/* SADB_DUMP */
7745 	key_api_promisc,	/* SADB_X_PROMISC */
7746 	NULL,			/* SADB_X_PCHANGE */
7747 	key_api_spdadd,		/* SADB_X_SPDUPDATE */
7748 	key_api_spdadd,		/* SADB_X_SPDADD */
7749 	key_api_spddelete,	/* SADB_X_SPDDELETE */
7750 	key_api_spdget,		/* SADB_X_SPDGET */
7751 	NULL,			/* SADB_X_SPDACQUIRE */
7752 	key_api_spddump,	/* SADB_X_SPDDUMP */
7753 	key_api_spdflush,	/* SADB_X_SPDFLUSH */
7754 	key_api_spdadd,		/* SADB_X_SPDSETIDX */
7755 	NULL,			/* SADB_X_SPDEXPIRE */
7756 	key_api_spddelete2,	/* SADB_X_SPDDELETE2 */
7757 	key_api_nat_map,	/* SADB_X_NAT_T_NEW_MAPPING */
7758 };
7759 
7760 /*
7761  * parse sadb_msg buffer to process PFKEYv2,
7762  * and create a data to response if needed.
7763  * I think to be dealed with mbuf directly.
7764  * IN:
7765  *     msgp  : pointer to pointer to a received buffer pulluped.
7766  *             This is rewrited to response.
7767  *     so    : pointer to socket.
7768  * OUT:
7769  *    length for buffer to send to user process.
7770  */
7771 int
7772 key_parse(struct mbuf *m, struct socket *so)
7773 {
7774 	struct sadb_msg *msg;
7775 	struct sadb_msghdr mh;
7776 	u_int orglen;
7777 	int error;
7778 
7779 	KASSERT(m != NULL);
7780 	KASSERT(so != NULL);
7781 
7782 #if 0	/*kdebug_sadb assumes msg in linear buffer*/
7783 	if (KEYDEBUG_ON(KEYDEBUG_KEY_DUMP)) {
7784 		kdebug_sadb("passed sadb_msg", msg);
7785 	}
7786 #endif
7787 
7788 	if (m->m_len < sizeof(struct sadb_msg)) {
7789 		m = m_pullup(m, sizeof(struct sadb_msg));
7790 		if (!m)
7791 			return ENOBUFS;
7792 	}
7793 	msg = mtod(m, struct sadb_msg *);
7794 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
7795 
7796 	if ((m->m_flags & M_PKTHDR) == 0 ||
7797 	    m->m_pkthdr.len != orglen) {
7798 		IPSECLOG(LOG_DEBUG, "invalid message length.\n");
7799 		PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
7800 		error = EINVAL;
7801 		goto senderror;
7802 	}
7803 
7804 	if (msg->sadb_msg_version != PF_KEY_V2) {
7805 		IPSECLOG(LOG_DEBUG, "PF_KEY version %u is mismatched.\n",
7806 		    msg->sadb_msg_version);
7807 		PFKEY_STATINC(PFKEY_STAT_OUT_INVVER);
7808 		error = EINVAL;
7809 		goto senderror;
7810 	}
7811 
7812 	if (msg->sadb_msg_type > SADB_MAX) {
7813 		IPSECLOG(LOG_DEBUG, "invalid type %u is passed.\n",
7814 		    msg->sadb_msg_type);
7815 		PFKEY_STATINC(PFKEY_STAT_OUT_INVMSGTYPE);
7816 		error = EINVAL;
7817 		goto senderror;
7818 	}
7819 
7820 	/* for old-fashioned code - should be nuked */
7821 	if (m->m_pkthdr.len > MCLBYTES) {
7822 		m_freem(m);
7823 		return ENOBUFS;
7824 	}
7825 	if (m->m_next) {
7826 		struct mbuf *n;
7827 
7828 		n = key_alloc_mbuf_simple(m->m_pkthdr.len, M_WAITOK);
7829 
7830 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, void *));
7831 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
7832 		n->m_next = NULL;
7833 		m_freem(m);
7834 		m = n;
7835 	}
7836 
7837 	/* align the mbuf chain so that extensions are in contiguous region. */
7838 	error = key_align(m, &mh);
7839 	if (error)
7840 		return error;
7841 
7842 	if (m->m_next) {	/*XXX*/
7843 		m_freem(m);
7844 		return ENOBUFS;
7845 	}
7846 
7847 	msg = mh.msg;
7848 
7849 	/* check SA type */
7850 	switch (msg->sadb_msg_satype) {
7851 	case SADB_SATYPE_UNSPEC:
7852 		switch (msg->sadb_msg_type) {
7853 		case SADB_GETSPI:
7854 		case SADB_UPDATE:
7855 		case SADB_ADD:
7856 		case SADB_DELETE:
7857 		case SADB_GET:
7858 		case SADB_ACQUIRE:
7859 		case SADB_EXPIRE:
7860 			IPSECLOG(LOG_DEBUG,
7861 			    "must specify satype when msg type=%u.\n",
7862 			    msg->sadb_msg_type);
7863 			PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7864 			error = EINVAL;
7865 			goto senderror;
7866 		}
7867 		break;
7868 	case SADB_SATYPE_AH:
7869 	case SADB_SATYPE_ESP:
7870 	case SADB_X_SATYPE_IPCOMP:
7871 	case SADB_X_SATYPE_TCPSIGNATURE:
7872 		switch (msg->sadb_msg_type) {
7873 		case SADB_X_SPDADD:
7874 		case SADB_X_SPDDELETE:
7875 		case SADB_X_SPDGET:
7876 		case SADB_X_SPDDUMP:
7877 		case SADB_X_SPDFLUSH:
7878 		case SADB_X_SPDSETIDX:
7879 		case SADB_X_SPDUPDATE:
7880 		case SADB_X_SPDDELETE2:
7881 			IPSECLOG(LOG_DEBUG, "illegal satype=%u\n",
7882 			    msg->sadb_msg_type);
7883 			PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7884 			error = EINVAL;
7885 			goto senderror;
7886 		}
7887 		break;
7888 	case SADB_SATYPE_RSVP:
7889 	case SADB_SATYPE_OSPFV2:
7890 	case SADB_SATYPE_RIPV2:
7891 	case SADB_SATYPE_MIP:
7892 		IPSECLOG(LOG_DEBUG, "type %u isn't supported.\n",
7893 		    msg->sadb_msg_satype);
7894 		PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7895 		error = EOPNOTSUPP;
7896 		goto senderror;
7897 	case 1:	/* XXX: What does it do? */
7898 		if (msg->sadb_msg_type == SADB_X_PROMISC)
7899 			break;
7900 		/*FALLTHROUGH*/
7901 	default:
7902 		IPSECLOG(LOG_DEBUG, "invalid type %u is passed.\n",
7903 		    msg->sadb_msg_satype);
7904 		PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7905 		error = EINVAL;
7906 		goto senderror;
7907 	}
7908 
7909 	/* check field of upper layer protocol and address family */
7910 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL &&
7911 	    mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
7912 		const struct sadb_address *src0, *dst0;
7913 		const struct sockaddr *sa0, *da0;
7914 		u_int plen;
7915 
7916 		src0 = mh.ext[SADB_EXT_ADDRESS_SRC];
7917 		dst0 = mh.ext[SADB_EXT_ADDRESS_DST];
7918 		sa0 = key_msghdr_get_sockaddr(&mh, SADB_EXT_ADDRESS_SRC);
7919 		da0 = key_msghdr_get_sockaddr(&mh, SADB_EXT_ADDRESS_DST);
7920 
7921 		/* check upper layer protocol */
7922 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
7923 			IPSECLOG(LOG_DEBUG,
7924 			    "upper layer protocol mismatched src %u, dst %u.\n",
7925 			    src0->sadb_address_proto, dst0->sadb_address_proto);
7926 
7927 			goto invaddr;
7928 		}
7929 
7930 		/* check family */
7931 		if (sa0->sa_family != da0->sa_family) {
7932 			IPSECLOG(LOG_DEBUG,
7933 			    "address family mismatched src %u, dst %u.\n",
7934 			    sa0->sa_family, da0->sa_family);
7935 			goto invaddr;
7936 		}
7937 		if (sa0->sa_len != da0->sa_len) {
7938 			IPSECLOG(LOG_DEBUG,
7939 			    "address size mismatched src %u, dst %u.\n",
7940 			    sa0->sa_len, da0->sa_len);
7941 			goto invaddr;
7942 		}
7943 
7944 		switch (sa0->sa_family) {
7945 		case AF_INET:
7946 			if (sa0->sa_len != sizeof(struct sockaddr_in)) {
7947 				IPSECLOG(LOG_DEBUG,
7948 				    "address size mismatched %u != %zu.\n",
7949 				    sa0->sa_len, sizeof(struct sockaddr_in));
7950 				goto invaddr;
7951 			}
7952 			break;
7953 		case AF_INET6:
7954 			if (sa0->sa_len != sizeof(struct sockaddr_in6)) {
7955 				IPSECLOG(LOG_DEBUG,
7956 				    "address size mismatched %u != %zu.\n",
7957 				    sa0->sa_len, sizeof(struct sockaddr_in6));
7958 				goto invaddr;
7959 			}
7960 			break;
7961 		default:
7962 			IPSECLOG(LOG_DEBUG, "unsupported address family %u.\n",
7963 			    sa0->sa_family);
7964 			error = EAFNOSUPPORT;
7965 			goto senderror;
7966 		}
7967 		plen = key_sabits(sa0);
7968 
7969 		/* check max prefix length */
7970 		if (src0->sadb_address_prefixlen > plen ||
7971 		    dst0->sadb_address_prefixlen > plen) {
7972 			IPSECLOG(LOG_DEBUG, "illegal prefixlen.\n");
7973 			goto invaddr;
7974 		}
7975 
7976 		/*
7977 		 * prefixlen == 0 is valid because there can be a case when
7978 		 * all addresses are matched.
7979 		 */
7980 	}
7981 
7982 	if (msg->sadb_msg_type >= __arraycount(key_api_typesw) ||
7983 	    key_api_typesw[msg->sadb_msg_type] == NULL) {
7984 		PFKEY_STATINC(PFKEY_STAT_OUT_INVMSGTYPE);
7985 		error = EINVAL;
7986 		goto senderror;
7987 	}
7988 
7989 	return (*key_api_typesw[msg->sadb_msg_type])(so, m, &mh);
7990 
7991 invaddr:
7992 	error = EINVAL;
7993 senderror:
7994 	PFKEY_STATINC(PFKEY_STAT_OUT_INVADDR);
7995 	return key_senderror(so, m, error);
7996 }
7997 
7998 static int
7999 key_senderror(struct socket *so, struct mbuf *m, int code)
8000 {
8001 	struct sadb_msg *msg;
8002 
8003 	KASSERT(m->m_len >= sizeof(struct sadb_msg));
8004 
8005 	if (so == NULL) {
8006 		/*
8007 		 * This means the request comes from kernel.
8008 		 * As the request comes from kernel, it is unnecessary to
8009 		 * send message to userland. Just return errcode directly.
8010 		 */
8011 		m_freem(m);
8012 		return code;
8013 	}
8014 
8015 	msg = mtod(m, struct sadb_msg *);
8016 	msg->sadb_msg_errno = code;
8017 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
8018 }
8019 
8020 /*
8021  * set the pointer to each header into message buffer.
8022  * m will be freed on error.
8023  * XXX larger-than-MCLBYTES extension?
8024  */
8025 static int
8026 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
8027 {
8028 	struct mbuf *n;
8029 	struct sadb_ext *ext;
8030 	size_t off, end;
8031 	int extlen;
8032 	int toff;
8033 
8034 	KASSERT(m != NULL);
8035 	KASSERT(mhp != NULL);
8036 	KASSERT(m->m_len >= sizeof(struct sadb_msg));
8037 
8038 	/* initialize */
8039 	memset(mhp, 0, sizeof(*mhp));
8040 
8041 	mhp->msg = mtod(m, struct sadb_msg *);
8042 	mhp->ext[0] = mhp->msg;	/*XXX backward compat */
8043 
8044 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
8045 	extlen = end;	/*just in case extlen is not updated*/
8046 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
8047 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
8048 		if (!n) {
8049 			/* m is already freed */
8050 			return ENOBUFS;
8051 		}
8052 		ext = (struct sadb_ext *)(mtod(n, char *) + toff);
8053 
8054 		/* set pointer */
8055 		switch (ext->sadb_ext_type) {
8056 		case SADB_EXT_SA:
8057 		case SADB_EXT_ADDRESS_SRC:
8058 		case SADB_EXT_ADDRESS_DST:
8059 		case SADB_EXT_ADDRESS_PROXY:
8060 		case SADB_EXT_LIFETIME_CURRENT:
8061 		case SADB_EXT_LIFETIME_HARD:
8062 		case SADB_EXT_LIFETIME_SOFT:
8063 		case SADB_EXT_KEY_AUTH:
8064 		case SADB_EXT_KEY_ENCRYPT:
8065 		case SADB_EXT_IDENTITY_SRC:
8066 		case SADB_EXT_IDENTITY_DST:
8067 		case SADB_EXT_SENSITIVITY:
8068 		case SADB_EXT_PROPOSAL:
8069 		case SADB_EXT_SUPPORTED_AUTH:
8070 		case SADB_EXT_SUPPORTED_ENCRYPT:
8071 		case SADB_EXT_SPIRANGE:
8072 		case SADB_X_EXT_POLICY:
8073 		case SADB_X_EXT_SA2:
8074 		case SADB_X_EXT_NAT_T_TYPE:
8075 		case SADB_X_EXT_NAT_T_SPORT:
8076 		case SADB_X_EXT_NAT_T_DPORT:
8077 		case SADB_X_EXT_NAT_T_OAI:
8078 		case SADB_X_EXT_NAT_T_OAR:
8079 		case SADB_X_EXT_NAT_T_FRAG:
8080 			/* duplicate check */
8081 			/*
8082 			 * XXX Are there duplication payloads of either
8083 			 * KEY_AUTH or KEY_ENCRYPT ?
8084 			 */
8085 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
8086 				IPSECLOG(LOG_DEBUG,
8087 				    "duplicate ext_type %u is passed.\n",
8088 				    ext->sadb_ext_type);
8089 				m_freem(m);
8090 				PFKEY_STATINC(PFKEY_STAT_OUT_DUPEXT);
8091 				return EINVAL;
8092 			}
8093 			break;
8094 		default:
8095 			IPSECLOG(LOG_DEBUG, "invalid ext_type %u is passed.\n",
8096 			    ext->sadb_ext_type);
8097 			m_freem(m);
8098 			PFKEY_STATINC(PFKEY_STAT_OUT_INVEXTTYPE);
8099 			return EINVAL;
8100 		}
8101 
8102 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
8103 
8104 		if (key_validate_ext(ext, extlen)) {
8105 			m_freem(m);
8106 			PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
8107 			return EINVAL;
8108 		}
8109 
8110 		n = m_pulldown(m, off, extlen, &toff);
8111 		if (!n) {
8112 			/* m is already freed */
8113 			return ENOBUFS;
8114 		}
8115 		ext = (struct sadb_ext *)(mtod(n, char *) + toff);
8116 
8117 		mhp->ext[ext->sadb_ext_type] = ext;
8118 		mhp->extoff[ext->sadb_ext_type] = off;
8119 		mhp->extlen[ext->sadb_ext_type] = extlen;
8120 	}
8121 
8122 	if (off != end) {
8123 		m_freem(m);
8124 		PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
8125 		return EINVAL;
8126 	}
8127 
8128 	return 0;
8129 }
8130 
8131 static int
8132 key_validate_ext(const struct sadb_ext *ext, int len)
8133 {
8134 	const struct sockaddr *sa;
8135 	enum { NONE, ADDR } checktype = NONE;
8136 	int baselen = 0;
8137 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
8138 
8139 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
8140 		return EINVAL;
8141 
8142 	/* if it does not match minimum/maximum length, bail */
8143 	if (ext->sadb_ext_type >= __arraycount(minsize) ||
8144 	    ext->sadb_ext_type >= __arraycount(maxsize))
8145 		return EINVAL;
8146 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
8147 		return EINVAL;
8148 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
8149 		return EINVAL;
8150 
8151 	/* more checks based on sadb_ext_type XXX need more */
8152 	switch (ext->sadb_ext_type) {
8153 	case SADB_EXT_ADDRESS_SRC:
8154 	case SADB_EXT_ADDRESS_DST:
8155 	case SADB_EXT_ADDRESS_PROXY:
8156 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
8157 		checktype = ADDR;
8158 		break;
8159 	case SADB_EXT_IDENTITY_SRC:
8160 	case SADB_EXT_IDENTITY_DST:
8161 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
8162 		    SADB_X_IDENTTYPE_ADDR) {
8163 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
8164 			checktype = ADDR;
8165 		} else
8166 			checktype = NONE;
8167 		break;
8168 	default:
8169 		checktype = NONE;
8170 		break;
8171 	}
8172 
8173 	switch (checktype) {
8174 	case NONE:
8175 		break;
8176 	case ADDR:
8177 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8178 		if (len < baselen + sal)
8179 			return EINVAL;
8180 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8181 			return EINVAL;
8182 		break;
8183 	}
8184 
8185 	return 0;
8186 }
8187 
8188 static int
8189 key_do_init(void)
8190 {
8191 	int i, error;
8192 
8193 	mutex_init(&key_misc.lock, MUTEX_DEFAULT, IPL_NONE);
8194 
8195 	mutex_init(&key_spd.lock, MUTEX_DEFAULT, IPL_NONE);
8196 	cv_init(&key_spd.cv_lc, "key_sp_lc");
8197 	key_spd.psz = pserialize_create();
8198 	cv_init(&key_spd.cv_psz, "key_sp_psz");
8199 	key_spd.psz_performing = false;
8200 
8201 	mutex_init(&key_sad.lock, MUTEX_DEFAULT, IPL_NONE);
8202 	cv_init(&key_sad.cv_lc, "key_sa_lc");
8203 	key_sad.psz = pserialize_create();
8204 	cv_init(&key_sad.cv_psz, "key_sa_psz");
8205 	key_sad.psz_performing = false;
8206 
8207 	pfkeystat_percpu = percpu_alloc(sizeof(uint64_t) * PFKEY_NSTATS);
8208 
8209 	callout_init(&key_timehandler_ch, CALLOUT_MPSAFE);
8210 	error = workqueue_create(&key_timehandler_wq, "key_timehandler",
8211 	    key_timehandler_work, NULL, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
8212 	if (error != 0)
8213 		panic("%s: workqueue_create failed (%d)\n", __func__, error);
8214 
8215 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8216 		PSLIST_INIT(&key_spd.splist[i]);
8217 	}
8218 
8219 	PSLIST_INIT(&key_spd.socksplist);
8220 
8221 	key_sad.sahlists = hashinit(SAHHASH_NHASH, HASH_PSLIST, true,
8222 	    &key_sad.sahlistmask);
8223 	key_sad.savlut = hashinit(SAVLUT_NHASH, HASH_PSLIST, true,
8224 	    &key_sad.savlutmask);
8225 
8226 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8227 		LIST_INIT(&key_misc.reglist[i]);
8228 	}
8229 
8230 #ifndef IPSEC_NONBLOCK_ACQUIRE
8231 	LIST_INIT(&key_misc.acqlist);
8232 #endif
8233 #ifdef notyet
8234 	LIST_INIT(&key_misc.spacqlist);
8235 #endif
8236 
8237 	/* system default */
8238 	ip4_def_policy.policy = IPSEC_POLICY_NONE;
8239 	ip4_def_policy.state = IPSEC_SPSTATE_ALIVE;
8240 	localcount_init(&ip4_def_policy.localcount);
8241 
8242 #ifdef INET6
8243 	ip6_def_policy.policy = IPSEC_POLICY_NONE;
8244 	ip6_def_policy.state = IPSEC_SPSTATE_ALIVE;
8245 	localcount_init(&ip6_def_policy.localcount);
8246 #endif
8247 
8248 	callout_reset(&key_timehandler_ch, hz, key_timehandler, NULL);
8249 
8250 	/* initialize key statistics */
8251 	keystat.getspi_count = 1;
8252 
8253 	aprint_verbose("IPsec: Initialized Security Association Processing.\n");
8254 
8255 	return (0);
8256 }
8257 
8258 void
8259 key_init(void)
8260 {
8261 	static ONCE_DECL(key_init_once);
8262 
8263 	sysctl_net_keyv2_setup(NULL);
8264 	sysctl_net_key_compat_setup(NULL);
8265 
8266 	RUN_ONCE(&key_init_once, key_do_init);
8267 
8268 	key_init_so();
8269 }
8270 
8271 /*
8272  * XXX: maybe This function is called after INBOUND IPsec processing.
8273  *
8274  * Special check for tunnel-mode packets.
8275  * We must make some checks for consistency between inner and outer IP header.
8276  *
8277  * xxx more checks to be provided
8278  */
8279 int
8280 key_checktunnelsanity(
8281     struct secasvar *sav,
8282     u_int family,
8283     void *src,
8284     void *dst
8285 )
8286 {
8287 
8288 	/* XXX: check inner IP header */
8289 
8290 	return 1;
8291 }
8292 
8293 #if 0
8294 #define hostnamelen	strlen(hostname)
8295 
8296 /*
8297  * Get FQDN for the host.
8298  * If the administrator configured hostname (by hostname(1)) without
8299  * domain name, returns nothing.
8300  */
8301 static const char *
8302 key_getfqdn(void)
8303 {
8304 	int i;
8305 	int hasdot;
8306 	static char fqdn[MAXHOSTNAMELEN + 1];
8307 
8308 	if (!hostnamelen)
8309 		return NULL;
8310 
8311 	/* check if it comes with domain name. */
8312 	hasdot = 0;
8313 	for (i = 0; i < hostnamelen; i++) {
8314 		if (hostname[i] == '.')
8315 			hasdot++;
8316 	}
8317 	if (!hasdot)
8318 		return NULL;
8319 
8320 	/* NOTE: hostname may not be NUL-terminated. */
8321 	memset(fqdn, 0, sizeof(fqdn));
8322 	memcpy(fqdn, hostname, hostnamelen);
8323 	fqdn[hostnamelen] = '\0';
8324 	return fqdn;
8325 }
8326 
8327 /*
8328  * get username@FQDN for the host/user.
8329  */
8330 static const char *
8331 key_getuserfqdn(void)
8332 {
8333 	const char *host;
8334 	static char userfqdn[MAXHOSTNAMELEN + MAXLOGNAME + 2];
8335 	struct proc *p = curproc;
8336 	char *q;
8337 
8338 	if (!p || !p->p_pgrp || !p->p_pgrp->pg_session)
8339 		return NULL;
8340 	if (!(host = key_getfqdn()))
8341 		return NULL;
8342 
8343 	/* NOTE: s_login may not be-NUL terminated. */
8344 	memset(userfqdn, 0, sizeof(userfqdn));
8345 	memcpy(userfqdn, Mp->p_pgrp->pg_session->s_login, AXLOGNAME);
8346 	userfqdn[MAXLOGNAME] = '\0';	/* safeguard */
8347 	q = userfqdn + strlen(userfqdn);
8348 	*q++ = '@';
8349 	memcpy(q, host, strlen(host));
8350 	q += strlen(host);
8351 	*q++ = '\0';
8352 
8353 	return userfqdn;
8354 }
8355 #endif
8356 
8357 /* record data transfer on SA, and update timestamps */
8358 void
8359 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8360 {
8361 	lifetime_counters_t *counters;
8362 
8363 	KASSERT(sav != NULL);
8364 	KASSERT(sav->lft_c != NULL);
8365 	KASSERT(m != NULL);
8366 
8367 	counters = percpu_getref(sav->lft_c_counters_percpu);
8368 
8369 	/*
8370 	 * XXX Currently, there is a difference of bytes size
8371 	 * between inbound and outbound processing.
8372 	 */
8373 	(*counters)[LIFETIME_COUNTER_BYTES] += m->m_pkthdr.len;
8374 	/* to check bytes lifetime is done in key_timehandler(). */
8375 
8376 	/*
8377 	 * We use the number of packets as the unit of
8378 	 * sadb_lifetime_allocations.  We increment the variable
8379 	 * whenever {esp,ah}_{in,out}put is called.
8380 	 */
8381 	(*counters)[LIFETIME_COUNTER_ALLOCATIONS]++;
8382 	/* XXX check for expires? */
8383 
8384 	percpu_putref(sav->lft_c_counters_percpu);
8385 
8386 	/*
8387 	 * NOTE: We record CURRENT sadb_lifetime_usetime by using wall clock,
8388 	 * in seconds.  HARD and SOFT lifetime are measured by the time
8389 	 * difference (again in seconds) from sadb_lifetime_usetime.
8390 	 *
8391 	 *	usetime
8392 	 *	v     expire   expire
8393 	 * -----+-----+--------+---> t
8394 	 *	<--------------> HARD
8395 	 *	<-----> SOFT
8396 	 */
8397 	sav->lft_c->sadb_lifetime_usetime = time_uptime;
8398 	/* XXX check for expires? */
8399 
8400 	return;
8401 }
8402 
8403 /* dumb version */
8404 void
8405 key_sa_routechange(struct sockaddr *dst)
8406 {
8407 	struct secashead *sah;
8408 	int s;
8409 
8410 	s = pserialize_read_enter();
8411 	SAHLIST_READER_FOREACH(sah) {
8412 		struct route *ro;
8413 		const struct sockaddr *sa;
8414 
8415 		key_sah_ref(sah);
8416 		pserialize_read_exit(s);
8417 
8418 		ro = &sah->sa_route;
8419 		sa = rtcache_getdst(ro);
8420 		if (sa != NULL && dst->sa_len == sa->sa_len &&
8421 		    memcmp(dst, sa, dst->sa_len) == 0)
8422 			rtcache_free(ro);
8423 
8424 		s = pserialize_read_enter();
8425 		key_sah_unref(sah);
8426 	}
8427 	pserialize_read_exit(s);
8428 
8429 	return;
8430 }
8431 
8432 static void
8433 key_sa_chgstate(struct secasvar *sav, u_int8_t state)
8434 {
8435 	struct secasvar *_sav;
8436 
8437 	ASSERT_SLEEPABLE();
8438 	KASSERT(mutex_owned(&key_sad.lock));
8439 
8440 	if (sav->state == state)
8441 		return;
8442 
8443 	key_unlink_sav(sav);
8444 	localcount_fini(&sav->localcount);
8445 	SAVLIST_ENTRY_DESTROY(sav);
8446 	key_init_sav(sav);
8447 
8448 	sav->state = state;
8449 	if (!SADB_SASTATE_USABLE_P(sav)) {
8450 		/* We don't need to care about the order */
8451 		SAVLIST_WRITER_INSERT_HEAD(sav->sah, state, sav);
8452 		return;
8453 	}
8454 	/*
8455 	 * Sort the list by lft_c->sadb_lifetime_addtime
8456 	 * in ascending order.
8457 	 */
8458 	SAVLIST_WRITER_FOREACH(_sav, sav->sah, state) {
8459 		if (_sav->lft_c->sadb_lifetime_addtime >
8460 		    sav->lft_c->sadb_lifetime_addtime) {
8461 			SAVLIST_WRITER_INSERT_BEFORE(_sav, sav);
8462 			break;
8463 		}
8464 	}
8465 	if (_sav == NULL) {
8466 		SAVLIST_WRITER_INSERT_TAIL(sav->sah, state, sav);
8467 	}
8468 
8469 	SAVLUT_WRITER_INSERT_HEAD(sav);
8470 
8471 	key_validate_savlist(sav->sah, state);
8472 }
8473 
8474 /* XXX too much? */
8475 static struct mbuf *
8476 key_alloc_mbuf(int l, int mflag)
8477 {
8478 	struct mbuf *m = NULL, *n;
8479 	int len, t;
8480 
8481 	KASSERT(mflag == M_NOWAIT || (mflag == M_WAITOK && !cpu_softintr_p()));
8482 
8483 	len = l;
8484 	while (len > 0) {
8485 		MGET(n, mflag, MT_DATA);
8486 		if (n && len > MLEN) {
8487 			MCLGET(n, mflag);
8488 			if ((n->m_flags & M_EXT) == 0) {
8489 				m_freem(n);
8490 				n = NULL;
8491 			}
8492 		}
8493 		if (!n) {
8494 			m_freem(m);
8495 			return NULL;
8496 		}
8497 
8498 		n->m_next = NULL;
8499 		n->m_len = 0;
8500 		n->m_len = M_TRAILINGSPACE(n);
8501 		/* use the bottom of mbuf, hoping we can prepend afterwards */
8502 		if (n->m_len > len) {
8503 			t = (n->m_len - len) & ~(sizeof(long) - 1);
8504 			n->m_data += t;
8505 			n->m_len = len;
8506 		}
8507 
8508 		len -= n->m_len;
8509 
8510 		if (m)
8511 			m_cat(m, n);
8512 		else
8513 			m = n;
8514 	}
8515 
8516 	return m;
8517 }
8518 
8519 static struct mbuf *
8520 key_setdump(u_int8_t req_satype, int *errorp, uint32_t pid)
8521 {
8522 	struct secashead *sah;
8523 	struct secasvar *sav;
8524 	u_int16_t proto;
8525 	u_int8_t satype;
8526 	u_int8_t state;
8527 	int cnt;
8528 	struct mbuf *m, *n;
8529 
8530 	KASSERT(mutex_owned(&key_sad.lock));
8531 
8532 	/* map satype to proto */
8533 	proto = key_satype2proto(req_satype);
8534 	if (proto == 0) {
8535 		*errorp = EINVAL;
8536 		return (NULL);
8537 	}
8538 
8539 	/* count sav entries to be sent to the userland. */
8540 	cnt = 0;
8541 	SAHLIST_WRITER_FOREACH(sah) {
8542 		if (req_satype != SADB_SATYPE_UNSPEC &&
8543 		    proto != sah->saidx.proto)
8544 			continue;
8545 
8546 		SASTATE_ANY_FOREACH(state) {
8547 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
8548 				cnt++;
8549 			}
8550 		}
8551 	}
8552 
8553 	if (cnt == 0) {
8554 		*errorp = ENOENT;
8555 		return (NULL);
8556 	}
8557 
8558 	/* send this to the userland, one at a time. */
8559 	m = NULL;
8560 	SAHLIST_WRITER_FOREACH(sah) {
8561 		if (req_satype != SADB_SATYPE_UNSPEC &&
8562 		    proto != sah->saidx.proto)
8563 			continue;
8564 
8565 		/* map proto to satype */
8566 		satype = key_proto2satype(sah->saidx.proto);
8567 		if (satype == 0) {
8568 			m_freem(m);
8569 			*errorp = EINVAL;
8570 			return (NULL);
8571 		}
8572 
8573 		SASTATE_ANY_FOREACH(state) {
8574 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
8575 				n = key_setdumpsa(sav, SADB_DUMP, satype,
8576 				    --cnt, pid);
8577 				if (!m)
8578 					m = n;
8579 				else
8580 					m_cat(m, n);
8581 			}
8582 		}
8583 	}
8584 
8585 	if (!m) {
8586 		*errorp = EINVAL;
8587 		return (NULL);
8588 	}
8589 
8590 	if ((m->m_flags & M_PKTHDR) != 0) {
8591 		m->m_pkthdr.len = 0;
8592 		for (n = m; n; n = n->m_next)
8593 			m->m_pkthdr.len += n->m_len;
8594 	}
8595 
8596 	*errorp = 0;
8597 	return (m);
8598 }
8599 
8600 static struct mbuf *
8601 key_setspddump(int *errorp, pid_t pid)
8602 {
8603 	struct secpolicy *sp;
8604 	int cnt;
8605 	u_int dir;
8606 	struct mbuf *m, *n;
8607 
8608 	KASSERT(mutex_owned(&key_spd.lock));
8609 
8610 	/* search SPD entry and get buffer size. */
8611 	cnt = 0;
8612 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
8613 		SPLIST_WRITER_FOREACH(sp, dir) {
8614 			cnt++;
8615 		}
8616 	}
8617 
8618 	if (cnt == 0) {
8619 		*errorp = ENOENT;
8620 		return (NULL);
8621 	}
8622 
8623 	m = NULL;
8624 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
8625 		SPLIST_WRITER_FOREACH(sp, dir) {
8626 			--cnt;
8627 			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, pid);
8628 
8629 			if (!m)
8630 				m = n;
8631 			else {
8632 				m->m_pkthdr.len += n->m_pkthdr.len;
8633 				m_cat(m, n);
8634 			}
8635 		}
8636 	}
8637 
8638 	*errorp = 0;
8639 	return (m);
8640 }
8641 
8642 int
8643 key_get_used(void) {
8644 	return !SPLIST_READER_EMPTY(IPSEC_DIR_INBOUND) ||
8645 	    !SPLIST_READER_EMPTY(IPSEC_DIR_OUTBOUND) ||
8646 	    !SOCKSPLIST_READER_EMPTY();
8647 }
8648 
8649 void
8650 key_update_used(void)
8651 {
8652 	switch (ipsec_enabled) {
8653 	default:
8654 	case 0:
8655 #ifdef notyet
8656 		/* XXX: racy */
8657 		ipsec_used = 0;
8658 #endif
8659 		break;
8660 	case 1:
8661 #ifndef notyet
8662 		/* XXX: racy */
8663 		if (!ipsec_used)
8664 #endif
8665 		ipsec_used = key_get_used();
8666 		break;
8667 	case 2:
8668 		ipsec_used = 1;
8669 		break;
8670 	}
8671 }
8672 
8673 static inline void
8674 key_savlut_writer_insert_head(struct secasvar *sav)
8675 {
8676 	uint32_t hash_key;
8677 	uint32_t hash;
8678 
8679 	KASSERT(mutex_owned(&key_sad.lock));
8680 	KASSERT(!sav->savlut_added);
8681 
8682 	if (sav->sah->saidx.proto == IPPROTO_IPCOMP)
8683 		hash_key = sav->alg_comp;
8684 	else
8685 		hash_key = sav->spi;
8686 
8687 	hash = key_savluthash(&sav->sah->saidx.dst.sa,
8688 	    sav->sah->saidx.proto, hash_key, key_sad.savlutmask);
8689 
8690 	PSLIST_WRITER_INSERT_HEAD(&key_sad.savlut[hash], sav,
8691 	    pslist_entry_savlut);
8692 	sav->savlut_added = true;
8693 }
8694 
8695 /*
8696  * Calculate hash using protocol, source address,
8697  * and destination address included in saidx.
8698  */
8699 static inline uint32_t
8700 key_saidxhash(const struct secasindex *saidx, u_long mask)
8701 {
8702 	uint32_t hash32;
8703 	const struct sockaddr_in *sin;
8704 	const struct sockaddr_in6 *sin6;
8705 
8706 	hash32 = saidx->proto;
8707 
8708 	switch (saidx->src.sa.sa_family) {
8709 	case AF_INET:
8710 		sin = &saidx->src.sin;
8711 		hash32 = hash32_buf(&sin->sin_addr,
8712 		    sizeof(sin->sin_addr), hash32);
8713 		sin = &saidx->dst.sin;
8714 		hash32 = hash32_buf(&sin->sin_addr,
8715 		    sizeof(sin->sin_addr), hash32 << 1);
8716 		break;
8717 	case AF_INET6:
8718 		sin6 = &saidx->src.sin6;
8719 		hash32 = hash32_buf(&sin6->sin6_addr,
8720 		    sizeof(sin6->sin6_addr), hash32);
8721 		sin6 = &saidx->dst.sin6;
8722 		hash32 = hash32_buf(&sin6->sin6_addr,
8723 		    sizeof(sin6->sin6_addr), hash32 << 1);
8724 		break;
8725 	default:
8726 		hash32 = 0;
8727 		break;
8728 	}
8729 
8730 	return hash32 & mask;
8731 }
8732 
8733 /*
8734  * Calculate hash using destination address, protocol,
8735  * and spi. Those parameter depend on the search of
8736  * key_lookup_sa().
8737  */
8738 static uint32_t
8739 key_savluthash(const struct sockaddr *dst, uint32_t proto,
8740     uint32_t spi, u_long mask)
8741 {
8742 	uint32_t hash32;
8743 	const struct sockaddr_in *sin;
8744 	const struct sockaddr_in6 *sin6;
8745 
8746 	hash32 = hash32_buf(&proto, sizeof(proto), spi);
8747 
8748 	switch(dst->sa_family) {
8749 	case AF_INET:
8750 		sin = satocsin(dst);
8751 		hash32 = hash32_buf(&sin->sin_addr,
8752 		    sizeof(sin->sin_addr), hash32);
8753 		break;
8754 	case AF_INET6:
8755 		sin6 = satocsin6(dst);
8756 		hash32 = hash32_buf(&sin6->sin6_addr,
8757 		    sizeof(sin6->sin6_addr), hash32);
8758 		break;
8759 	default:
8760 		hash32 = 0;
8761 	}
8762 
8763 	return hash32 & mask;
8764 }
8765 
8766 static int
8767 sysctl_net_key_dumpsa(SYSCTLFN_ARGS)
8768 {
8769 	struct mbuf *m, *n;
8770 	int err2 = 0;
8771 	char *p, *ep;
8772 	size_t len;
8773 	int error;
8774 
8775 	if (newp)
8776 		return (EPERM);
8777 	if (namelen != 1)
8778 		return (EINVAL);
8779 
8780 	mutex_enter(&key_sad.lock);
8781 	m = key_setdump(name[0], &error, l->l_proc->p_pid);
8782 	mutex_exit(&key_sad.lock);
8783 	if (!m)
8784 		return (error);
8785 	if (!oldp)
8786 		*oldlenp = m->m_pkthdr.len;
8787 	else {
8788 		p = oldp;
8789 		if (*oldlenp < m->m_pkthdr.len) {
8790 			err2 = ENOMEM;
8791 			ep = p + *oldlenp;
8792 		} else {
8793 			*oldlenp = m->m_pkthdr.len;
8794 			ep = p + m->m_pkthdr.len;
8795 		}
8796 		for (n = m; n; n = n->m_next) {
8797 			len =  (ep - p < n->m_len) ?
8798 				ep - p : n->m_len;
8799 			error = copyout(mtod(n, const void *), p, len);
8800 			p += len;
8801 			if (error)
8802 				break;
8803 		}
8804 		if (error == 0)
8805 			error = err2;
8806 	}
8807 	m_freem(m);
8808 
8809 	return (error);
8810 }
8811 
8812 static int
8813 sysctl_net_key_dumpsp(SYSCTLFN_ARGS)
8814 {
8815 	struct mbuf *m, *n;
8816 	int err2 = 0;
8817 	char *p, *ep;
8818 	size_t len;
8819 	int error;
8820 
8821 	if (newp)
8822 		return (EPERM);
8823 	if (namelen != 0)
8824 		return (EINVAL);
8825 
8826 	mutex_enter(&key_spd.lock);
8827 	m = key_setspddump(&error, l->l_proc->p_pid);
8828 	mutex_exit(&key_spd.lock);
8829 	if (!m)
8830 		return (error);
8831 	if (!oldp)
8832 		*oldlenp = m->m_pkthdr.len;
8833 	else {
8834 		p = oldp;
8835 		if (*oldlenp < m->m_pkthdr.len) {
8836 			err2 = ENOMEM;
8837 			ep = p + *oldlenp;
8838 		} else {
8839 			*oldlenp = m->m_pkthdr.len;
8840 			ep = p + m->m_pkthdr.len;
8841 		}
8842 		for (n = m; n; n = n->m_next) {
8843 			len = (ep - p < n->m_len) ? ep - p : n->m_len;
8844 			error = copyout(mtod(n, const void *), p, len);
8845 			p += len;
8846 			if (error)
8847 				break;
8848 		}
8849 		if (error == 0)
8850 			error = err2;
8851 	}
8852 	m_freem(m);
8853 
8854 	return (error);
8855 }
8856 
8857 /*
8858  * Create sysctl tree for native IPSEC key knobs, originally
8859  * under name "net.keyv2"  * with MIB number { CTL_NET, PF_KEY_V2. }.
8860  * However, sysctl(8) never checked for nodes under { CTL_NET, PF_KEY_V2 };
8861  * and in any case the part of our sysctl namespace used for dumping the
8862  * SPD and SA database  *HAS* to be compatible with the KAME sysctl
8863  * namespace, for API reasons.
8864  *
8865  * Pending a consensus on the right way  to fix this, add a level of
8866  * indirection in how we number the `native' IPSEC key nodes;
8867  * and (as requested by Andrew Brown)  move registration of the
8868  * KAME-compatible names  to a separate function.
8869  */
8870 #if 0
8871 #  define IPSEC_PFKEY PF_KEY_V2
8872 # define IPSEC_PFKEY_NAME "keyv2"
8873 #else
8874 #  define IPSEC_PFKEY PF_KEY
8875 # define IPSEC_PFKEY_NAME "key"
8876 #endif
8877 
8878 static int
8879 sysctl_net_key_stats(SYSCTLFN_ARGS)
8880 {
8881 
8882 	return (NETSTAT_SYSCTL(pfkeystat_percpu, PFKEY_NSTATS));
8883 }
8884 
8885 static void
8886 sysctl_net_keyv2_setup(struct sysctllog **clog)
8887 {
8888 
8889 	sysctl_createv(clog, 0, NULL, NULL,
8890 		       CTLFLAG_PERMANENT,
8891 		       CTLTYPE_NODE, IPSEC_PFKEY_NAME, NULL,
8892 		       NULL, 0, NULL, 0,
8893 		       CTL_NET, IPSEC_PFKEY, CTL_EOL);
8894 
8895 	sysctl_createv(clog, 0, NULL, NULL,
8896 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8897 		       CTLTYPE_INT, "debug", NULL,
8898 		       NULL, 0, &key_debug_level, 0,
8899 		       CTL_NET, IPSEC_PFKEY, KEYCTL_DEBUG_LEVEL, CTL_EOL);
8900 	sysctl_createv(clog, 0, NULL, NULL,
8901 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8902 		       CTLTYPE_INT, "spi_try", NULL,
8903 		       NULL, 0, &key_spi_trycnt, 0,
8904 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_TRY, CTL_EOL);
8905 	sysctl_createv(clog, 0, NULL, NULL,
8906 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8907 		       CTLTYPE_INT, "spi_min_value", NULL,
8908 		       NULL, 0, &key_spi_minval, 0,
8909 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_MIN_VALUE, CTL_EOL);
8910 	sysctl_createv(clog, 0, NULL, NULL,
8911 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8912 		       CTLTYPE_INT, "spi_max_value", NULL,
8913 		       NULL, 0, &key_spi_maxval, 0,
8914 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_MAX_VALUE, CTL_EOL);
8915 	sysctl_createv(clog, 0, NULL, NULL,
8916 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8917 		       CTLTYPE_INT, "random_int", NULL,
8918 		       NULL, 0, &key_int_random, 0,
8919 		       CTL_NET, IPSEC_PFKEY, KEYCTL_RANDOM_INT, CTL_EOL);
8920 	sysctl_createv(clog, 0, NULL, NULL,
8921 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8922 		       CTLTYPE_INT, "larval_lifetime", NULL,
8923 		       NULL, 0, &key_larval_lifetime, 0,
8924 		       CTL_NET, IPSEC_PFKEY, KEYCTL_LARVAL_LIFETIME, CTL_EOL);
8925 	sysctl_createv(clog, 0, NULL, NULL,
8926 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8927 		       CTLTYPE_INT, "blockacq_count", NULL,
8928 		       NULL, 0, &key_blockacq_count, 0,
8929 		       CTL_NET, IPSEC_PFKEY, KEYCTL_BLOCKACQ_COUNT, CTL_EOL);
8930 	sysctl_createv(clog, 0, NULL, NULL,
8931 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8932 		       CTLTYPE_INT, "blockacq_lifetime", NULL,
8933 		       NULL, 0, &key_blockacq_lifetime, 0,
8934 		       CTL_NET, IPSEC_PFKEY, KEYCTL_BLOCKACQ_LIFETIME, CTL_EOL);
8935 	sysctl_createv(clog, 0, NULL, NULL,
8936 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8937 		       CTLTYPE_INT, "esp_keymin", NULL,
8938 		       NULL, 0, &ipsec_esp_keymin, 0,
8939 		       CTL_NET, IPSEC_PFKEY, KEYCTL_ESP_KEYMIN, CTL_EOL);
8940 	sysctl_createv(clog, 0, NULL, NULL,
8941 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8942 		       CTLTYPE_INT, "prefered_oldsa", NULL,
8943 		       NULL, 0, &key_prefered_oldsa, 0,
8944 		       CTL_NET, PF_KEY, KEYCTL_PREFERED_OLDSA, CTL_EOL);
8945 	sysctl_createv(clog, 0, NULL, NULL,
8946 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8947 		       CTLTYPE_INT, "esp_auth", NULL,
8948 		       NULL, 0, &ipsec_esp_auth, 0,
8949 		       CTL_NET, IPSEC_PFKEY, KEYCTL_ESP_AUTH, CTL_EOL);
8950 	sysctl_createv(clog, 0, NULL, NULL,
8951 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8952 		       CTLTYPE_INT, "ah_keymin", NULL,
8953 		       NULL, 0, &ipsec_ah_keymin, 0,
8954 		       CTL_NET, IPSEC_PFKEY, KEYCTL_AH_KEYMIN, CTL_EOL);
8955 	sysctl_createv(clog, 0, NULL, NULL,
8956 		       CTLFLAG_PERMANENT,
8957 		       CTLTYPE_STRUCT, "stats",
8958 		       SYSCTL_DESCR("PF_KEY statistics"),
8959 		       sysctl_net_key_stats, 0, NULL, 0,
8960 		       CTL_NET, IPSEC_PFKEY, CTL_CREATE, CTL_EOL);
8961 }
8962 
8963 /*
8964  * Register sysctl names used by setkey(8). For historical reasons,
8965  * and to share a single API, these names appear under { CTL_NET, PF_KEY }
8966  * for both IPSEC and KAME IPSEC.
8967  */
8968 static void
8969 sysctl_net_key_compat_setup(struct sysctllog **clog)
8970 {
8971 
8972 	sysctl_createv(clog, 0, NULL, NULL,
8973 		       CTLFLAG_PERMANENT,
8974 		       CTLTYPE_NODE, "key", NULL,
8975 		       NULL, 0, NULL, 0,
8976 		       CTL_NET, PF_KEY, CTL_EOL);
8977 
8978 	/* Register the net.key.dump{sa,sp} nodes used by setkey(8). */
8979 	sysctl_createv(clog, 0, NULL, NULL,
8980 		       CTLFLAG_PERMANENT,
8981 		       CTLTYPE_STRUCT, "dumpsa", NULL,
8982 		       sysctl_net_key_dumpsa, 0, NULL, 0,
8983 		       CTL_NET, PF_KEY, KEYCTL_DUMPSA, CTL_EOL);
8984 	sysctl_createv(clog, 0, NULL, NULL,
8985 		       CTLFLAG_PERMANENT,
8986 		       CTLTYPE_STRUCT, "dumpsp", NULL,
8987 		       sysctl_net_key_dumpsp, 0, NULL, 0,
8988 		       CTL_NET, PF_KEY, KEYCTL_DUMPSP, CTL_EOL);
8989 }
8990