Home | History | Annotate | Line # | Download | only in netipsec
key.c revision 1.242
      1 /*	$NetBSD: key.c,v 1.242 2017/11/21 07:33:06 ozaki-r Exp $	*/
      2 /*	$FreeBSD: src/sys/netipsec/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 1.242 2017/11/21 07:33:06 ozaki-r 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 
     76 #include <net/if.h>
     77 #include <net/route.h>
     78 
     79 #include <netinet/in.h>
     80 #include <netinet/in_systm.h>
     81 #include <netinet/ip.h>
     82 #include <netinet/in_var.h>
     83 #ifdef INET
     84 #include <netinet/ip_var.h>
     85 #endif
     86 
     87 #ifdef INET6
     88 #include <netinet/ip6.h>
     89 #include <netinet6/in6_var.h>
     90 #include <netinet6/ip6_var.h>
     91 #endif /* INET6 */
     92 
     93 #ifdef INET
     94 #include <netinet/in_pcb.h>
     95 #endif
     96 #ifdef INET6
     97 #include <netinet6/in6_pcb.h>
     98 #endif /* INET6 */
     99 
    100 #include <net/pfkeyv2.h>
    101 #include <netipsec/keydb.h>
    102 #include <netipsec/key.h>
    103 #include <netipsec/keysock.h>
    104 #include <netipsec/key_debug.h>
    105 
    106 #include <netipsec/ipsec.h>
    107 #ifdef INET6
    108 #include <netipsec/ipsec6.h>
    109 #endif
    110 #include <netipsec/ipsec_private.h>
    111 
    112 #include <netipsec/xform.h>
    113 #include <netipsec/ipcomp.h>
    114 
    115 
    116 #include <net/net_osdep.h>
    117 
    118 #define FULLMASK	0xff
    119 #define	_BITS(bytes)	((bytes) << 3)
    120 
    121 #define PORT_NONE	0
    122 #define PORT_LOOSE	1
    123 #define PORT_STRICT	2
    124 
    125 percpu_t *pfkeystat_percpu;
    126 
    127 /*
    128  * Note on SA reference counting:
    129  * - SAs that are not in DEAD state will have (total external reference + 1)
    130  *   following value in reference count field.  they cannot be freed and are
    131  *   referenced from SA header.
    132  * - SAs that are in DEAD state will have (total external reference)
    133  *   in reference count field.  they are ready to be freed.  reference from
    134  *   SA header will be removed in key_delsav(), when the reference count
    135  *   field hits 0 (= no external reference other than from SA header.
    136  */
    137 
    138 u_int32_t key_debug_level = 0;
    139 static u_int key_spi_trycnt = 1000;
    140 static u_int32_t key_spi_minval = 0x100;
    141 static u_int32_t key_spi_maxval = 0x0fffffff;	/* XXX */
    142 static u_int32_t policy_id = 0;
    143 static u_int key_int_random = 60;	/*interval to initialize randseed,1(m)*/
    144 static u_int key_larval_lifetime = 30;	/* interval to expire acquiring, 30(s)*/
    145 static int key_blockacq_count = 10;	/* counter for blocking SADB_ACQUIRE.*/
    146 static int key_blockacq_lifetime = 20;	/* lifetime for blocking SADB_ACQUIRE.*/
    147 static int key_prefered_oldsa = 0;	/* prefered old sa rather than new sa.*/
    148 
    149 static u_int32_t acq_seq = 0;
    150 
    151 /*
    152  * Locking order: there is no order for now; it means that any locks aren't
    153  * overlapped.
    154  */
    155 /*
    156  * Locking notes on SPD:
    157  * - Modifications to the key_spd.splist must be done with holding key_spd.lock
    158  *   which is a adaptive mutex
    159  * - Read accesses to the key_spd.splist must be in pserialize(9) read sections
    160  * - SP's lifetime is managed by localcount(9)
    161  * - An SP that has been inserted to the key_spd.splist is initially referenced
    162  *   by none, i.e., a reference from the key_spd.splist isn't counted
    163  * - When an SP is being destroyed, we change its state as DEAD, wait for
    164  *   references to the SP to be released, and then deallocate the SP
    165  *   (see key_unlink_sp)
    166  * - Getting an SP
    167  *   - Normally we get an SP from the key_spd.splist (see key_lookup_sp_byspidx)
    168  *     - Must iterate the list and increment the reference count of a found SP
    169  *       (by key_sp_ref) in a pserialize read section
    170  *   - We can gain another reference from a held SP only if we check its state
    171  *     and take its reference in a pserialize read section
    172  *     (see esp_output for example)
    173  *   - We may get an SP from an SP cache. See below
    174  *   - A gotten SP must be released after use by KEY_SP_UNREF (key_sp_unref)
    175  * - Updating member variables of an SP
    176  *   - Most member variables of an SP are immutable
    177  *   - Only sp->state and sp->lastused can be changed
    178  *   - sp->state of an SP is updated only when destroying it under key_spd.lock
    179  * - SP caches
    180  *   - SPs can be cached in PCBs
    181  *   - The lifetime of the caches is controlled by the global generation counter
    182  *     (ipsec_spdgen)
    183  *   - The global counter value is stored when an SP is cached
    184  *   - If the stored value is different from the global counter then the cache
    185  *     is considered invalidated
    186  *   - The counter is incremented when an SP is being destroyed
    187  *   - So checking the generation and taking a reference to an SP should be
    188  *     in a pserialize read section
    189  *   - Note that caching doesn't increment the reference counter of an SP
    190  * - SPs in sockets
    191  *   - Userland programs can set a policy to a socket by
    192  *     setsockopt(IP_IPSEC_POLICY)
    193  *   - Such policies (SPs) are set to a socket (PCB) and also inserted to
    194  *     the key_spd.socksplist list (not the key_spd.splist)
    195  *   - Such a policy is destroyed when a corresponding socket is destroed,
    196  *     however, a socket can be destroyed in softint so we cannot destroy
    197  *     it directly instead we just mark it DEAD and delay the destruction
    198  *     until GC by the timer
    199  */
    200 /*
    201  * Locking notes on SAD:
    202  * - Data structures
    203  *   - SAs are managed by the list called key_sad.sahlist and sav lists of sah
    204  *     entries
    205  *     - An sav is supposed to be an SA from a viewpoint of users
    206  *   - A sah has sav lists for each SA state
    207  *   - Multiple sahs with the same saidx can exist
    208  *     - Only one entry has MATURE state and others should be DEAD
    209  *     - DEAD entries are just ignored from searching
    210  * - Modifications to the key_sad.sahlist and sah.savlist must be done with
    211  *   holding key_sad.lock which is a adaptive mutex
    212  * - Read accesses to the key_sad.sahlist and sah.savlist must be in
    213  *   pserialize(9) read sections
    214  * - sah's lifetime is managed by localcount(9)
    215  * - Getting an sah entry
    216  *   - We get an sah from the key_sad.sahlist
    217  *     - Must iterate the list and increment the reference count of a found sah
    218  *       (by key_sah_ref) in a pserialize read section
    219  *   - A gotten sah must be released after use by key_sah_unref
    220  * - An sah is destroyed when its state become DEAD and no sav is
    221  *   listed to the sah
    222  *   - The destruction is done only in the timer (see key_timehandler_sad)
    223  * - sav's lifetime is managed by localcount(9)
    224  * - Getting an sav entry
    225  *   - First get an sah by saidx and get an sav from either of sah's savlists
    226  *     - Must iterate the list and increment the reference count of a found sav
    227  *       (by key_sa_ref) in a pserialize read section
    228  *   - We can gain another reference from a held SA only if we check its state
    229  *     and take its reference in a pserialize read section
    230  *     (see esp_output for example)
    231  *   - A gotten sav must be released after use by key_sa_unref
    232  * - An sav is destroyed when its state become DEAD
    233  */
    234 /*
    235  * Locking notes on misc data:
    236  * - All lists of key_misc are protected by key_misc.lock
    237  *   - key_misc.lock must be held even for read accesses
    238  */
    239 
    240 /* SPD */
    241 static struct {
    242 	kmutex_t lock;
    243 	kcondvar_t cv_lc;
    244 	struct pslist_head splist[IPSEC_DIR_MAX];
    245 	/*
    246 	 * The list has SPs that are set to a socket via
    247 	 * setsockopt(IP_IPSEC_POLICY) from userland. See ipsec_set_policy.
    248 	 */
    249 	struct pslist_head socksplist;
    250 
    251 	pserialize_t psz;
    252 	kcondvar_t cv_psz;
    253 	bool psz_performing;
    254 } key_spd __cacheline_aligned;
    255 
    256 /* SAD */
    257 static struct {
    258 	kmutex_t lock;
    259 	kcondvar_t cv_lc;
    260 	struct pslist_head sahlist;
    261 
    262 	pserialize_t psz;
    263 	kcondvar_t cv_psz;
    264 	bool psz_performing;
    265 } key_sad __cacheline_aligned;
    266 
    267 /* Misc data */
    268 static struct {
    269 	kmutex_t lock;
    270 	/* registed list */
    271 	LIST_HEAD(_reglist, secreg) reglist[SADB_SATYPE_MAX + 1];
    272 #ifndef IPSEC_NONBLOCK_ACQUIRE
    273 	/* acquiring list */
    274 	LIST_HEAD(_acqlist, secacq) acqlist;
    275 #endif
    276 #ifdef notyet
    277 	/* SP acquiring list */
    278 	LIST_HEAD(_spacqlist, secspacq) spacqlist;
    279 #endif
    280 } key_misc __cacheline_aligned;
    281 
    282 /* Macros for key_spd.splist */
    283 #define SPLIST_ENTRY_INIT(sp)						\
    284 	PSLIST_ENTRY_INIT((sp), pslist_entry)
    285 #define SPLIST_ENTRY_DESTROY(sp)					\
    286 	PSLIST_ENTRY_DESTROY((sp), pslist_entry)
    287 #define SPLIST_WRITER_REMOVE(sp)					\
    288 	PSLIST_WRITER_REMOVE((sp), pslist_entry)
    289 #define SPLIST_READER_EMPTY(dir)					\
    290 	(PSLIST_READER_FIRST(&key_spd.splist[(dir)], struct secpolicy,	\
    291 	                     pslist_entry) == NULL)
    292 #define SPLIST_READER_FOREACH(sp, dir)					\
    293 	PSLIST_READER_FOREACH((sp), &key_spd.splist[(dir)],		\
    294 	                      struct secpolicy, pslist_entry)
    295 #define SPLIST_WRITER_FOREACH(sp, dir)					\
    296 	PSLIST_WRITER_FOREACH((sp), &key_spd.splist[(dir)],		\
    297 	                      struct secpolicy, pslist_entry)
    298 #define SPLIST_WRITER_INSERT_AFTER(sp, new)				\
    299 	PSLIST_WRITER_INSERT_AFTER((sp), (new), pslist_entry)
    300 #define SPLIST_WRITER_EMPTY(dir)					\
    301 	(PSLIST_WRITER_FIRST(&key_spd.splist[(dir)], struct secpolicy,	\
    302 	                     pslist_entry) == NULL)
    303 #define SPLIST_WRITER_INSERT_HEAD(dir, sp)				\
    304 	PSLIST_WRITER_INSERT_HEAD(&key_spd.splist[(dir)], (sp),		\
    305 	                          pslist_entry)
    306 #define SPLIST_WRITER_NEXT(sp)						\
    307 	PSLIST_WRITER_NEXT((sp), struct secpolicy, pslist_entry)
    308 #define SPLIST_WRITER_INSERT_TAIL(dir, new)				\
    309 	do {								\
    310 		if (SPLIST_WRITER_EMPTY((dir))) {			\
    311 			SPLIST_WRITER_INSERT_HEAD((dir), (new));	\
    312 		} else {						\
    313 			struct secpolicy *__sp;				\
    314 			SPLIST_WRITER_FOREACH(__sp, (dir)) {		\
    315 				if (SPLIST_WRITER_NEXT(__sp) == NULL) {	\
    316 					SPLIST_WRITER_INSERT_AFTER(__sp,\
    317 					    (new));			\
    318 					break;				\
    319 				}					\
    320 			}						\
    321 		}							\
    322 	} while (0)
    323 
    324 /* Macros for key_spd.socksplist */
    325 #define SOCKSPLIST_WRITER_FOREACH(sp)					\
    326 	PSLIST_WRITER_FOREACH((sp), &key_spd.socksplist,		\
    327 	                      struct secpolicy,	pslist_entry)
    328 #define SOCKSPLIST_READER_EMPTY()					\
    329 	(PSLIST_READER_FIRST(&key_spd.socksplist, struct secpolicy,	\
    330 	                     pslist_entry) == NULL)
    331 
    332 /* Macros for key_sad.sahlist */
    333 #define SAHLIST_ENTRY_INIT(sah)						\
    334 	PSLIST_ENTRY_INIT((sah), pslist_entry)
    335 #define SAHLIST_ENTRY_DESTROY(sah)					\
    336 	PSLIST_ENTRY_DESTROY((sah), pslist_entry)
    337 #define SAHLIST_WRITER_REMOVE(sah)					\
    338 	PSLIST_WRITER_REMOVE((sah), pslist_entry)
    339 #define SAHLIST_READER_FOREACH(sah)					\
    340 	PSLIST_READER_FOREACH((sah), &key_sad.sahlist, struct secashead,\
    341 	                      pslist_entry)
    342 #define SAHLIST_WRITER_FOREACH(sah)					\
    343 	PSLIST_WRITER_FOREACH((sah), &key_sad.sahlist, struct secashead,\
    344 	                      pslist_entry)
    345 #define SAHLIST_WRITER_INSERT_HEAD(sah)					\
    346 	PSLIST_WRITER_INSERT_HEAD(&key_sad.sahlist, (sah), pslist_entry)
    347 
    348 /* Macros for key_sad.sahlist#savlist */
    349 #define SAVLIST_ENTRY_INIT(sav)						\
    350 	PSLIST_ENTRY_INIT((sav), pslist_entry)
    351 #define SAVLIST_ENTRY_DESTROY(sav)					\
    352 	PSLIST_ENTRY_DESTROY((sav), pslist_entry)
    353 #define SAVLIST_READER_FIRST(sah, state)				\
    354 	PSLIST_READER_FIRST(&(sah)->savlist[(state)], struct secasvar,	\
    355 	                    pslist_entry)
    356 #define SAVLIST_WRITER_REMOVE(sav)					\
    357 	PSLIST_WRITER_REMOVE((sav), pslist_entry)
    358 #define SAVLIST_READER_FOREACH(sav, sah, state)				\
    359 	PSLIST_READER_FOREACH((sav), &(sah)->savlist[(state)],		\
    360 	                      struct secasvar, pslist_entry)
    361 #define SAVLIST_WRITER_FOREACH(sav, sah, state)				\
    362 	PSLIST_WRITER_FOREACH((sav), &(sah)->savlist[(state)],		\
    363 	                      struct secasvar, pslist_entry)
    364 #define SAVLIST_WRITER_INSERT_BEFORE(sav, new)				\
    365 	PSLIST_WRITER_INSERT_BEFORE((sav), (new), pslist_entry)
    366 #define SAVLIST_WRITER_INSERT_AFTER(sav, new)				\
    367 	PSLIST_WRITER_INSERT_AFTER((sav), (new), pslist_entry)
    368 #define SAVLIST_WRITER_EMPTY(sah, state)				\
    369 	(PSLIST_WRITER_FIRST(&(sah)->savlist[(state)], struct secasvar,	\
    370 	                     pslist_entry) == NULL)
    371 #define SAVLIST_WRITER_INSERT_HEAD(sah, state, sav)			\
    372 	PSLIST_WRITER_INSERT_HEAD(&(sah)->savlist[(state)], (sav),	\
    373 	                          pslist_entry)
    374 #define SAVLIST_WRITER_NEXT(sav)					\
    375 	PSLIST_WRITER_NEXT((sav), struct secasvar, pslist_entry)
    376 #define SAVLIST_WRITER_INSERT_TAIL(sah, state, new)			\
    377 	do {								\
    378 		if (SAVLIST_WRITER_EMPTY((sah), (state))) {		\
    379 			SAVLIST_WRITER_INSERT_HEAD((sah), (state), (new));\
    380 		} else {						\
    381 			struct secasvar *__sav;				\
    382 			SAVLIST_WRITER_FOREACH(__sav, (sah), (state)) {	\
    383 				if (SAVLIST_WRITER_NEXT(__sav) == NULL) {\
    384 					SAVLIST_WRITER_INSERT_AFTER(__sav,\
    385 					    (new));			\
    386 					break;				\
    387 				}					\
    388 			}						\
    389 		}							\
    390 	} while (0)
    391 #define SAVLIST_READER_NEXT(sav)					\
    392 	PSLIST_READER_NEXT((sav), struct secasvar, pslist_entry)
    393 
    394 
    395 /* search order for SAs */
    396 	/*
    397 	 * This order is important because we must select the oldest SA
    398 	 * for outbound processing.  For inbound, This is not important.
    399 	 */
    400 static const u_int saorder_state_valid_prefer_old[] = {
    401 	SADB_SASTATE_DYING, SADB_SASTATE_MATURE,
    402 };
    403 static const u_int saorder_state_valid_prefer_new[] = {
    404 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING,
    405 };
    406 
    407 static const u_int saorder_state_alive[] = {
    408 	/* except DEAD */
    409 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING, SADB_SASTATE_LARVAL
    410 };
    411 static const u_int saorder_state_any[] = {
    412 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING,
    413 	SADB_SASTATE_LARVAL, SADB_SASTATE_DEAD
    414 };
    415 
    416 #define SASTATE_ALIVE_FOREACH(s)				\
    417 	for (int _i = 0;					\
    418 	    _i < __arraycount(saorder_state_alive) ?		\
    419 	    (s) = saorder_state_alive[_i], true : false;	\
    420 	    _i++)
    421 #define SASTATE_ANY_FOREACH(s)					\
    422 	for (int _i = 0;					\
    423 	    _i < __arraycount(saorder_state_any) ?		\
    424 	    (s) = saorder_state_any[_i], true : false;		\
    425 	    _i++)
    426 
    427 static const int minsize[] = {
    428 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
    429 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
    430 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
    431 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
    432 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
    433 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_SRC */
    434 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_DST */
    435 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_PROXY */
    436 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_AUTH */
    437 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_ENCRYPT */
    438 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_SRC */
    439 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_DST */
    440 	sizeof(struct sadb_sens),	/* SADB_EXT_SENSITIVITY */
    441 	sizeof(struct sadb_prop),	/* SADB_EXT_PROPOSAL */
    442 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_AUTH */
    443 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_ENCRYPT */
    444 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
    445 	0,				/* SADB_X_EXT_KMPRIVATE */
    446 	sizeof(struct sadb_x_policy),	/* SADB_X_EXT_POLICY */
    447 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
    448 	sizeof(struct sadb_x_nat_t_type),	/* SADB_X_EXT_NAT_T_TYPE */
    449 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_SPORT */
    450 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_DPORT */
    451 	sizeof(struct sadb_address),		/* SADB_X_EXT_NAT_T_OAI */
    452 	sizeof(struct sadb_address),		/* SADB_X_EXT_NAT_T_OAR */
    453 	sizeof(struct sadb_x_nat_t_frag),	/* SADB_X_EXT_NAT_T_FRAG */
    454 };
    455 static const int maxsize[] = {
    456 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
    457 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
    458 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
    459 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
    460 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
    461 	0,				/* SADB_EXT_ADDRESS_SRC */
    462 	0,				/* SADB_EXT_ADDRESS_DST */
    463 	0,				/* SADB_EXT_ADDRESS_PROXY */
    464 	0,				/* SADB_EXT_KEY_AUTH */
    465 	0,				/* SADB_EXT_KEY_ENCRYPT */
    466 	0,				/* SADB_EXT_IDENTITY_SRC */
    467 	0,				/* SADB_EXT_IDENTITY_DST */
    468 	0,				/* SADB_EXT_SENSITIVITY */
    469 	0,				/* SADB_EXT_PROPOSAL */
    470 	0,				/* SADB_EXT_SUPPORTED_AUTH */
    471 	0,				/* SADB_EXT_SUPPORTED_ENCRYPT */
    472 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
    473 	0,				/* SADB_X_EXT_KMPRIVATE */
    474 	0,				/* SADB_X_EXT_POLICY */
    475 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
    476 	sizeof(struct sadb_x_nat_t_type),	/* SADB_X_EXT_NAT_T_TYPE */
    477 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_SPORT */
    478 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_DPORT */
    479 	0,					/* SADB_X_EXT_NAT_T_OAI */
    480 	0,					/* SADB_X_EXT_NAT_T_OAR */
    481 	sizeof(struct sadb_x_nat_t_frag),	/* SADB_X_EXT_NAT_T_FRAG */
    482 };
    483 
    484 static int ipsec_esp_keymin = 256;
    485 static int ipsec_esp_auth = 0;
    486 static int ipsec_ah_keymin = 128;
    487 
    488 #ifdef SYSCTL_DECL
    489 SYSCTL_DECL(_net_key);
    490 #endif
    491 
    492 #ifdef SYSCTL_INT
    493 SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,	CTLFLAG_RW, \
    494 	&key_debug_level,	0,	"");
    495 
    496 /* max count of trial for the decision of spi value */
    497 SYSCTL_INT(_net_key, KEYCTL_SPI_TRY,		spi_trycnt,	CTLFLAG_RW, \
    498 	&key_spi_trycnt,	0,	"");
    499 
    500 /* minimum spi value to allocate automatically. */
    501 SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE,	spi_minval,	CTLFLAG_RW, \
    502 	&key_spi_minval,	0,	"");
    503 
    504 /* maximun spi value to allocate automatically. */
    505 SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE,	spi_maxval,	CTLFLAG_RW, \
    506 	&key_spi_maxval,	0,	"");
    507 
    508 /* interval to initialize randseed */
    509 SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT,	int_random,	CTLFLAG_RW, \
    510 	&key_int_random,	0,	"");
    511 
    512 /* lifetime for larval SA */
    513 SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME,	larval_lifetime, CTLFLAG_RW, \
    514 	&key_larval_lifetime,	0,	"");
    515 
    516 /* counter for blocking to send SADB_ACQUIRE to IKEd */
    517 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT,	blockacq_count,	CTLFLAG_RW, \
    518 	&key_blockacq_count,	0,	"");
    519 
    520 /* lifetime for blocking to send SADB_ACQUIRE to IKEd */
    521 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME,	blockacq_lifetime, CTLFLAG_RW, \
    522 	&key_blockacq_lifetime,	0,	"");
    523 
    524 /* ESP auth */
    525 SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH,	esp_auth, CTLFLAG_RW, \
    526 	&ipsec_esp_auth,	0,	"");
    527 
    528 /* minimum ESP key length */
    529 SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN,	esp_keymin, CTLFLAG_RW, \
    530 	&ipsec_esp_keymin,	0,	"");
    531 
    532 /* minimum AH key length */
    533 SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN,	ah_keymin, CTLFLAG_RW, \
    534 	&ipsec_ah_keymin,	0,	"");
    535 
    536 /* perfered old SA rather than new SA */
    537 SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA,	prefered_oldsa, CTLFLAG_RW,\
    538 	&key_prefered_oldsa,	0,	"");
    539 #endif /* SYSCTL_INT */
    540 
    541 #define __LIST_CHAINED(elm) \
    542 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
    543 #define LIST_INSERT_TAIL(head, elm, type, field) \
    544 do {\
    545 	struct type *curelm = LIST_FIRST(head); \
    546 	if (curelm == NULL) {\
    547 		LIST_INSERT_HEAD(head, elm, field); \
    548 	} else { \
    549 		while (LIST_NEXT(curelm, field)) \
    550 			curelm = LIST_NEXT(curelm, field);\
    551 		LIST_INSERT_AFTER(curelm, elm, field);\
    552 	}\
    553 } while (0)
    554 
    555 #define KEY_CHKSASTATE(head, sav) \
    556 /* do */ { \
    557 	if ((head) != (sav)) {						\
    558 		IPSECLOG(LOG_DEBUG,					\
    559 		    "state mismatched (TREE=%d SA=%d)\n",		\
    560 		    (head), (sav));					\
    561 		continue;						\
    562 	}								\
    563 } /* while (0) */
    564 
    565 #define KEY_CHKSPDIR(head, sp) \
    566 do { \
    567 	if ((head) != (sp)) {						\
    568 		IPSECLOG(LOG_DEBUG,					\
    569 		    "direction mismatched (TREE=%d SP=%d), anyway continue.\n",\
    570 		    (head), (sp));					\
    571 	}								\
    572 } while (0)
    573 
    574 /*
    575  * set parameters into secasindex buffer.
    576  * Must allocate secasindex buffer before calling this function.
    577  */
    578 static int
    579 key_setsecasidx(int, int, int, const struct sockaddr *,
    580     const struct sockaddr *, struct secasindex *);
    581 
    582 /* key statistics */
    583 struct _keystat {
    584 	u_long getspi_count; /* the avarage of count to try to get new SPI */
    585 } keystat;
    586 
    587 struct sadb_msghdr {
    588 	struct sadb_msg *msg;
    589 	void *ext[SADB_EXT_MAX + 1];
    590 	int extoff[SADB_EXT_MAX + 1];
    591 	int extlen[SADB_EXT_MAX + 1];
    592 };
    593 
    594 static void
    595 key_init_spidx_bymsghdr(struct secpolicyindex *, const struct sadb_msghdr *);
    596 
    597 static const struct sockaddr *
    598 key_msghdr_get_sockaddr(const struct sadb_msghdr *mhp, int idx)
    599 {
    600 
    601 	return PFKEY_ADDR_SADDR(mhp->ext[idx]);
    602 }
    603 
    604 static void
    605 key_fill_replymsg(struct mbuf *m, int seq)
    606 {
    607 	struct sadb_msg *msg;
    608 
    609 	KASSERT(m->m_len >= sizeof(*msg));
    610 
    611 	msg = mtod(m, struct sadb_msg *);
    612 	msg->sadb_msg_errno = 0;
    613 	msg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
    614 	if (seq != 0)
    615 		msg->sadb_msg_seq = seq;
    616 }
    617 
    618 #if 0
    619 static void key_freeso(struct socket *);
    620 static void key_freesp_so(struct secpolicy **);
    621 #endif
    622 static struct secpolicy *key_getsp (const struct secpolicyindex *);
    623 static struct secpolicy *key_getspbyid (u_int32_t);
    624 static struct secpolicy *key_lookup_and_remove_sp(const struct secpolicyindex *);
    625 static struct secpolicy *key_lookupbyid_and_remove_sp(u_int32_t);
    626 static void key_destroy_sp(struct secpolicy *);
    627 static u_int16_t key_newreqid (void);
    628 static struct mbuf *key_gather_mbuf (struct mbuf *,
    629 	const struct sadb_msghdr *, int, int, ...);
    630 static int key_api_spdadd(struct socket *, struct mbuf *,
    631 	const struct sadb_msghdr *);
    632 static u_int32_t key_getnewspid (void);
    633 static int key_api_spddelete(struct socket *, struct mbuf *,
    634 	const struct sadb_msghdr *);
    635 static int key_api_spddelete2(struct socket *, struct mbuf *,
    636 	const struct sadb_msghdr *);
    637 static int key_api_spdget(struct socket *, struct mbuf *,
    638 	const struct sadb_msghdr *);
    639 static int key_api_spdflush(struct socket *, struct mbuf *,
    640 	const struct sadb_msghdr *);
    641 static int key_api_spddump(struct socket *, struct mbuf *,
    642 	const struct sadb_msghdr *);
    643 static struct mbuf * key_setspddump (int *errorp, pid_t);
    644 static struct mbuf * key_setspddump_chain (int *errorp, int *lenp, pid_t pid);
    645 static int key_api_nat_map(struct socket *, struct mbuf *,
    646 	const struct sadb_msghdr *);
    647 static struct mbuf *key_setdumpsp (struct secpolicy *,
    648 	u_int8_t, u_int32_t, pid_t);
    649 static u_int key_getspreqmsglen (const struct secpolicy *);
    650 static int key_spdexpire (struct secpolicy *);
    651 static struct secashead *key_newsah (const struct secasindex *);
    652 static void key_unlink_sah(struct secashead *);
    653 static void key_destroy_sah(struct secashead *);
    654 static bool key_sah_has_sav(struct secashead *);
    655 static void key_sah_ref(struct secashead *);
    656 static void key_sah_unref(struct secashead *);
    657 static void key_init_sav(struct secasvar *);
    658 static void key_destroy_sav(struct secasvar *);
    659 static void key_destroy_sav_with_ref(struct secasvar *);
    660 static struct secasvar *key_newsav(struct mbuf *,
    661 	const struct sadb_msghdr *, int *, const char*, int);
    662 #define	KEY_NEWSAV(m, sadb, e)				\
    663 	key_newsav(m, sadb, e, __func__, __LINE__)
    664 static void key_delsav (struct secasvar *);
    665 static struct secashead *key_getsah(const struct secasindex *, int);
    666 static struct secashead *key_getsah_ref(const struct secasindex *, int);
    667 static bool key_checkspidup(const struct secasindex *, u_int32_t);
    668 static struct secasvar *key_getsavbyspi (struct secashead *, u_int32_t);
    669 static int key_setsaval (struct secasvar *, struct mbuf *,
    670 	const struct sadb_msghdr *);
    671 static void key_freesaval(struct secasvar *);
    672 static int key_init_xform(struct secasvar *);
    673 static void key_clear_xform(struct secasvar *);
    674 static struct mbuf *key_setdumpsa (struct secasvar *, u_int8_t,
    675 	u_int8_t, u_int32_t, u_int32_t);
    676 static struct mbuf *key_setsadbxport (u_int16_t, u_int16_t);
    677 static struct mbuf *key_setsadbxtype (u_int16_t);
    678 static struct mbuf *key_setsadbxfrag (u_int16_t);
    679 static void key_porttosaddr (union sockaddr_union *, u_int16_t);
    680 static int key_checksalen (const union sockaddr_union *);
    681 static struct mbuf *key_setsadbmsg (u_int8_t, u_int16_t, u_int8_t,
    682 	u_int32_t, pid_t, u_int16_t, int);
    683 static struct mbuf *key_setsadbsa (struct secasvar *);
    684 static struct mbuf *key_setsadbaddr(u_int16_t,
    685 	const struct sockaddr *, u_int8_t, u_int16_t, int);
    686 #if 0
    687 static struct mbuf *key_setsadbident (u_int16_t, u_int16_t, void *,
    688 	int, u_int64_t);
    689 #endif
    690 static struct mbuf *key_setsadbxsa2 (u_int8_t, u_int32_t, u_int16_t);
    691 static struct mbuf *key_setsadbxpolicy (u_int16_t, u_int8_t,
    692 	u_int32_t, int);
    693 static void *key_newbuf (const void *, u_int);
    694 #ifdef INET6
    695 static int key_ismyaddr6 (const struct sockaddr_in6 *);
    696 #endif
    697 
    698 static void sysctl_net_keyv2_setup(struct sysctllog **);
    699 static void sysctl_net_key_compat_setup(struct sysctllog **);
    700 
    701 /* flags for key_saidx_match() */
    702 #define CMP_HEAD	1	/* protocol, addresses. */
    703 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
    704 #define CMP_REQID	3	/* additionally HEAD, reaid. */
    705 #define CMP_EXACTLY	4	/* all elements. */
    706 static int key_saidx_match(const struct secasindex *,
    707     const struct secasindex *, int);
    708 
    709 static int key_sockaddr_match(const struct sockaddr *,
    710     const struct sockaddr *, int);
    711 static int key_bb_match_withmask(const void *, const void *, u_int);
    712 static u_int16_t key_satype2proto (u_int8_t);
    713 static u_int8_t key_proto2satype (u_int16_t);
    714 
    715 static int key_spidx_match_exactly(const struct secpolicyindex *,
    716     const struct secpolicyindex *);
    717 static int key_spidx_match_withmask(const struct secpolicyindex *,
    718     const struct secpolicyindex *);
    719 
    720 static int key_api_getspi(struct socket *, struct mbuf *,
    721 	const struct sadb_msghdr *);
    722 static u_int32_t key_do_getnewspi (const struct sadb_spirange *,
    723 					const struct secasindex *);
    724 static int key_handle_natt_info (struct secasvar *,
    725 				     const struct sadb_msghdr *);
    726 static int key_set_natt_ports (union sockaddr_union *,
    727 			 	union sockaddr_union *,
    728 				const struct sadb_msghdr *);
    729 static int key_api_update(struct socket *, struct mbuf *,
    730 	const struct sadb_msghdr *);
    731 #ifdef IPSEC_DOSEQCHECK
    732 static struct secasvar *key_getsavbyseq (struct secashead *, u_int32_t);
    733 #endif
    734 static int key_api_add(struct socket *, struct mbuf *,
    735 	const struct sadb_msghdr *);
    736 static int key_setident (struct secashead *, struct mbuf *,
    737 	const struct sadb_msghdr *);
    738 static struct mbuf *key_getmsgbuf_x1 (struct mbuf *,
    739 	const struct sadb_msghdr *);
    740 static int key_api_delete(struct socket *, struct mbuf *,
    741 	const struct sadb_msghdr *);
    742 static int key_api_get(struct socket *, struct mbuf *,
    743 	const struct sadb_msghdr *);
    744 
    745 static void key_getcomb_setlifetime (struct sadb_comb *);
    746 static struct mbuf *key_getcomb_esp(int);
    747 static struct mbuf *key_getcomb_ah(int);
    748 static struct mbuf *key_getcomb_ipcomp(int);
    749 static struct mbuf *key_getprop(const struct secasindex *, int);
    750 
    751 static int key_acquire(const struct secasindex *, const struct secpolicy *,
    752 	    int);
    753 static int key_acquire_sendup_mbuf_later(struct mbuf *);
    754 static void key_acquire_sendup_pending_mbuf(void);
    755 #ifndef IPSEC_NONBLOCK_ACQUIRE
    756 static struct secacq *key_newacq (const struct secasindex *);
    757 static struct secacq *key_getacq (const struct secasindex *);
    758 static struct secacq *key_getacqbyseq (u_int32_t);
    759 #endif
    760 #ifdef notyet
    761 static struct secspacq *key_newspacq (const struct secpolicyindex *);
    762 static struct secspacq *key_getspacq (const struct secpolicyindex *);
    763 #endif
    764 static int key_api_acquire(struct socket *, struct mbuf *,
    765 	const struct sadb_msghdr *);
    766 static int key_api_register(struct socket *, struct mbuf *,
    767 	const struct sadb_msghdr *);
    768 static int key_expire (struct secasvar *);
    769 static int key_api_flush(struct socket *, struct mbuf *,
    770 	const struct sadb_msghdr *);
    771 static struct mbuf *key_setdump_chain (u_int8_t req_satype, int *errorp,
    772 	int *lenp, pid_t pid);
    773 static int key_api_dump(struct socket *, struct mbuf *,
    774 	const struct sadb_msghdr *);
    775 static int key_api_promisc(struct socket *, struct mbuf *,
    776 	const struct sadb_msghdr *);
    777 static int key_senderror (struct socket *, struct mbuf *, int);
    778 static int key_validate_ext (const struct sadb_ext *, int);
    779 static int key_align (struct mbuf *, struct sadb_msghdr *);
    780 #if 0
    781 static const char *key_getfqdn (void);
    782 static const char *key_getuserfqdn (void);
    783 #endif
    784 static void key_sa_chgstate (struct secasvar *, u_int8_t);
    785 
    786 static struct mbuf *key_alloc_mbuf(int, int);
    787 static struct mbuf *key_alloc_mbuf_simple(int, int);
    788 
    789 static void key_timehandler(void *);
    790 static void key_timehandler_work(struct work *, void *);
    791 static struct callout	key_timehandler_ch;
    792 static struct workqueue	*key_timehandler_wq;
    793 static struct work	key_timehandler_wk;
    794 
    795 u_int
    796 key_sp_refcnt(const struct secpolicy *sp)
    797 {
    798 
    799 	/* FIXME */
    800 	return 0;
    801 }
    802 
    803 #ifdef NET_MPSAFE
    804 static void
    805 key_spd_pserialize_perform(void)
    806 {
    807 
    808 	KASSERT(mutex_owned(&key_spd.lock));
    809 
    810 	while (key_spd.psz_performing)
    811 		cv_wait(&key_spd.cv_psz, &key_spd.lock);
    812 	key_spd.psz_performing = true;
    813 	mutex_exit(&key_spd.lock);
    814 
    815 	pserialize_perform(key_spd.psz);
    816 
    817 	mutex_enter(&key_spd.lock);
    818 	key_spd.psz_performing = false;
    819 	cv_broadcast(&key_spd.cv_psz);
    820 }
    821 #endif
    822 
    823 /*
    824  * Remove the sp from the key_spd.splist and wait for references to the sp
    825  * to be released. key_spd.lock must be held.
    826  */
    827 static void
    828 key_unlink_sp(struct secpolicy *sp)
    829 {
    830 
    831 	KASSERT(mutex_owned(&key_spd.lock));
    832 
    833 	sp->state = IPSEC_SPSTATE_DEAD;
    834 	SPLIST_WRITER_REMOVE(sp);
    835 
    836 	/* Invalidate all cached SPD pointers in the PCBs. */
    837 	ipsec_invalpcbcacheall();
    838 
    839 #ifdef NET_MPSAFE
    840 	KASSERT(mutex_ownable(softnet_lock));
    841 	key_spd_pserialize_perform();
    842 #endif
    843 
    844 	localcount_drain(&sp->localcount, &key_spd.cv_lc, &key_spd.lock);
    845 }
    846 
    847 /*
    848  * Return 0 when there are known to be no SP's for the specified
    849  * direction.  Otherwise return 1.  This is used by IPsec code
    850  * to optimize performance.
    851  */
    852 int
    853 key_havesp(u_int dir)
    854 {
    855 	return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ?
    856 		!SPLIST_READER_EMPTY(dir) : 1);
    857 }
    858 
    859 /* %%% IPsec policy management */
    860 /*
    861  * allocating a SP for OUTBOUND or INBOUND packet.
    862  * Must call key_freesp() later.
    863  * OUT:	NULL:	not found
    864  *	others:	found and return the pointer.
    865  */
    866 struct secpolicy *
    867 key_lookup_sp_byspidx(const struct secpolicyindex *spidx,
    868     u_int dir, const char* where, int tag)
    869 {
    870 	struct secpolicy *sp;
    871 	int s;
    872 
    873 	KASSERT(spidx != NULL);
    874 	KASSERTMSG(IPSEC_DIR_IS_INOROUT(dir), "invalid direction %u", dir);
    875 
    876 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
    877 
    878 	/* get a SP entry */
    879 	if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
    880 		kdebug_secpolicyindex("objects", spidx);
    881 	}
    882 
    883 	s = pserialize_read_enter();
    884 	SPLIST_READER_FOREACH(sp, dir) {
    885 		if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
    886 			kdebug_secpolicyindex("in SPD", &sp->spidx);
    887 		}
    888 
    889 		if (sp->state == IPSEC_SPSTATE_DEAD)
    890 			continue;
    891 		if (key_spidx_match_withmask(&sp->spidx, spidx))
    892 			goto found;
    893 	}
    894 	sp = NULL;
    895 found:
    896 	if (sp) {
    897 		/* sanity check */
    898 		KEY_CHKSPDIR(sp->spidx.dir, dir);
    899 
    900 		/* found a SPD entry */
    901 		sp->lastused = time_uptime;
    902 		key_sp_ref(sp, where, tag);
    903 	}
    904 	pserialize_read_exit(s);
    905 
    906 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
    907 	    "DP return SP:%p (ID=%u) refcnt %u\n",
    908 	    sp, sp ? sp->id : 0, key_sp_refcnt(sp));
    909 	return sp;
    910 }
    911 
    912 /*
    913  * return a policy that matches this particular inbound packet.
    914  * XXX slow
    915  */
    916 struct secpolicy *
    917 key_gettunnel(const struct sockaddr *osrc,
    918 	      const struct sockaddr *odst,
    919 	      const struct sockaddr *isrc,
    920 	      const struct sockaddr *idst,
    921 	      const char* where, int tag)
    922 {
    923 	struct secpolicy *sp;
    924 	const int dir = IPSEC_DIR_INBOUND;
    925 	int s;
    926 	struct ipsecrequest *r1, *r2, *p;
    927 	struct secpolicyindex spidx;
    928 
    929 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
    930 
    931 	if (isrc->sa_family != idst->sa_family) {
    932 		IPSECLOG(LOG_ERR, "protocol family mismatched %d != %d\n.",
    933 		    isrc->sa_family, idst->sa_family);
    934 		sp = NULL;
    935 		goto done;
    936 	}
    937 
    938 	s = pserialize_read_enter();
    939 	SPLIST_READER_FOREACH(sp, dir) {
    940 		if (sp->state == IPSEC_SPSTATE_DEAD)
    941 			continue;
    942 
    943 		r1 = r2 = NULL;
    944 		for (p = sp->req; p; p = p->next) {
    945 			if (p->saidx.mode != IPSEC_MODE_TUNNEL)
    946 				continue;
    947 
    948 			r1 = r2;
    949 			r2 = p;
    950 
    951 			if (!r1) {
    952 				/* here we look at address matches only */
    953 				spidx = sp->spidx;
    954 				if (isrc->sa_len > sizeof(spidx.src) ||
    955 				    idst->sa_len > sizeof(spidx.dst))
    956 					continue;
    957 				memcpy(&spidx.src, isrc, isrc->sa_len);
    958 				memcpy(&spidx.dst, idst, idst->sa_len);
    959 				if (!key_spidx_match_withmask(&sp->spidx, &spidx))
    960 					continue;
    961 			} else {
    962 				if (!key_sockaddr_match(&r1->saidx.src.sa, isrc, PORT_NONE) ||
    963 				    !key_sockaddr_match(&r1->saidx.dst.sa, idst, PORT_NONE))
    964 					continue;
    965 			}
    966 
    967 			if (!key_sockaddr_match(&r2->saidx.src.sa, osrc, PORT_NONE) ||
    968 			    !key_sockaddr_match(&r2->saidx.dst.sa, odst, PORT_NONE))
    969 				continue;
    970 
    971 			goto found;
    972 		}
    973 	}
    974 	sp = NULL;
    975 found:
    976 	if (sp) {
    977 		sp->lastused = time_uptime;
    978 		key_sp_ref(sp, where, tag);
    979 	}
    980 	pserialize_read_exit(s);
    981 done:
    982 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
    983 	    "DP return SP:%p (ID=%u) refcnt %u\n",
    984 	    sp, sp ? sp->id : 0, key_sp_refcnt(sp));
    985 	return sp;
    986 }
    987 
    988 /*
    989  * allocating an SA entry for an *OUTBOUND* packet.
    990  * checking each request entries in SP, and acquire an SA if need.
    991  * OUT:	0: there are valid requests.
    992  *	ENOENT: policy may be valid, but SA with REQUIRE is on acquiring.
    993  */
    994 int
    995 key_checkrequest(const struct ipsecrequest *isr, const struct secasindex *saidx,
    996     struct secasvar **ret)
    997 {
    998 	u_int level;
    999 	int error;
   1000 	struct secasvar *sav;
   1001 
   1002 	KASSERT(isr != NULL);
   1003 	KASSERTMSG(saidx->mode == IPSEC_MODE_TRANSPORT ||
   1004 	    saidx->mode == IPSEC_MODE_TUNNEL,
   1005 	    "unexpected policy %u", saidx->mode);
   1006 
   1007 	/* get current level */
   1008 	level = ipsec_get_reqlevel(isr);
   1009 
   1010 	/*
   1011 	 * XXX guard against protocol callbacks from the crypto
   1012 	 * thread as they reference ipsecrequest.sav which we
   1013 	 * temporarily null out below.  Need to rethink how we
   1014 	 * handle bundled SA's in the callback thread.
   1015 	 */
   1016 	IPSEC_SPLASSERT_SOFTNET("key_checkrequest");
   1017 
   1018 	sav = key_lookup_sa_bysaidx(saidx);
   1019 	if (sav != NULL) {
   1020 		*ret = sav;
   1021 		return 0;
   1022 	}
   1023 
   1024 	/* there is no SA */
   1025 	error = key_acquire(saidx, isr->sp, M_NOWAIT);
   1026 	if (error != 0) {
   1027 		/* XXX What should I do ? */
   1028 		IPSECLOG(LOG_DEBUG, "error %d returned from key_acquire.\n",
   1029 		    error);
   1030 		return error;
   1031 	}
   1032 
   1033 	if (level != IPSEC_LEVEL_REQUIRE) {
   1034 		/* XXX sigh, the interface to this routine is botched */
   1035 		*ret = NULL;
   1036 		return 0;
   1037 	} else {
   1038 		return ENOENT;
   1039 	}
   1040 }
   1041 
   1042 /*
   1043  * looking up a SA for policy entry from SAD.
   1044  * NOTE: searching SAD of aliving state.
   1045  * OUT:	NULL:	not found.
   1046  *	others:	found and return the pointer.
   1047  */
   1048 struct secasvar *
   1049 key_lookup_sa_bysaidx(const struct secasindex *saidx)
   1050 {
   1051 	struct secashead *sah;
   1052 	struct secasvar *sav = NULL;
   1053 	u_int stateidx, state;
   1054 	const u_int *saorder_state_valid;
   1055 	int arraysize;
   1056 	int s;
   1057 
   1058 	s = pserialize_read_enter();
   1059 	sah = key_getsah(saidx, CMP_MODE_REQID);
   1060 	if (sah == NULL)
   1061 		goto out;
   1062 
   1063 	/*
   1064 	 * search a valid state list for outbound packet.
   1065 	 * This search order is important.
   1066 	 */
   1067 	if (key_prefered_oldsa) {
   1068 		saorder_state_valid = saorder_state_valid_prefer_old;
   1069 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_old);
   1070 	} else {
   1071 		saorder_state_valid = saorder_state_valid_prefer_new;
   1072 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_new);
   1073 	}
   1074 
   1075 	/* search valid state */
   1076 	for (stateidx = 0;
   1077 	     stateidx < arraysize;
   1078 	     stateidx++) {
   1079 
   1080 		state = saorder_state_valid[stateidx];
   1081 
   1082 		if (key_prefered_oldsa)
   1083 			sav = SAVLIST_READER_FIRST(sah, state);
   1084 		else {
   1085 			/* XXX need O(1) lookup */
   1086 			struct secasvar *last = NULL;
   1087 
   1088 			SAVLIST_READER_FOREACH(sav, sah, state)
   1089 				last = sav;
   1090 			sav = last;
   1091 		}
   1092 		if (sav != NULL) {
   1093 			KEY_SA_REF(sav);
   1094 			KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1095 			    "DP cause refcnt++:%d SA:%p\n",
   1096 			    key_sa_refcnt(sav), sav);
   1097 			break;
   1098 		}
   1099 	}
   1100 out:
   1101 	pserialize_read_exit(s);
   1102 
   1103 	return sav;
   1104 }
   1105 
   1106 #if 0
   1107 static void
   1108 key_sendup_message_delete(struct secasvar *sav)
   1109 {
   1110 	struct mbuf *m, *result = 0;
   1111 	uint8_t satype;
   1112 
   1113 	satype = key_proto2satype(sav->sah->saidx.proto);
   1114 	if (satype == 0)
   1115 		goto msgfail;
   1116 
   1117 	m = key_setsadbmsg(SADB_DELETE, 0, satype, 0, 0, key_sa_refcnt(sav) - 1);
   1118 	if (m == NULL)
   1119 		goto msgfail;
   1120 	result = m;
   1121 
   1122 	/* set sadb_address for saidx's. */
   1123 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa,
   1124 	    sav->sah->saidx.src.sa.sa_len << 3, IPSEC_ULPROTO_ANY);
   1125 	if (m == NULL)
   1126 		goto msgfail;
   1127 	m_cat(result, m);
   1128 
   1129 	/* set sadb_address for saidx's. */
   1130 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.src.sa,
   1131 	    sav->sah->saidx.src.sa.sa_len << 3, IPSEC_ULPROTO_ANY);
   1132 	if (m == NULL)
   1133 		goto msgfail;
   1134 	m_cat(result, m);
   1135 
   1136 	/* create SA extension */
   1137 	m = key_setsadbsa(sav);
   1138 	if (m == NULL)
   1139 		goto msgfail;
   1140 	m_cat(result, m);
   1141 
   1142 	if (result->m_len < sizeof(struct sadb_msg)) {
   1143 		result = m_pullup(result, sizeof(struct sadb_msg));
   1144 		if (result == NULL)
   1145 			goto msgfail;
   1146 	}
   1147 
   1148 	result->m_pkthdr.len = 0;
   1149 	for (m = result; m; m = m->m_next)
   1150 		result->m_pkthdr.len += m->m_len;
   1151 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   1152 	    PFKEY_UNIT64(result->m_pkthdr.len);
   1153 
   1154 	key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
   1155 	result = NULL;
   1156 msgfail:
   1157 	if (result)
   1158 		m_freem(result);
   1159 }
   1160 #endif
   1161 
   1162 /*
   1163  * allocating a usable SA entry for a *INBOUND* packet.
   1164  * Must call key_freesav() later.
   1165  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
   1166  *	NULL:		not found, or error occurred.
   1167  *
   1168  * In the comparison, no source address is used--for RFC2401 conformance.
   1169  * To quote, from section 4.1:
   1170  *	A security association is uniquely identified by a triple consisting
   1171  *	of a Security Parameter Index (SPI), an IP Destination Address, and a
   1172  *	security protocol (AH or ESP) identifier.
   1173  * Note that, however, we do need to keep source address in IPsec SA.
   1174  * IKE specification and PF_KEY specification do assume that we
   1175  * keep source address in IPsec SA.  We see a tricky situation here.
   1176  *
   1177  * sport and dport are used for NAT-T. network order is always used.
   1178  */
   1179 struct secasvar *
   1180 key_lookup_sa(
   1181 	const union sockaddr_union *dst,
   1182 	u_int proto,
   1183 	u_int32_t spi,
   1184 	u_int16_t sport,
   1185 	u_int16_t dport,
   1186 	const char* where, int tag)
   1187 {
   1188 	struct secashead *sah;
   1189 	struct secasvar *sav;
   1190 	u_int stateidx, state;
   1191 	const u_int *saorder_state_valid;
   1192 	int arraysize, chkport;
   1193 	int s;
   1194 
   1195 	int must_check_spi = 1;
   1196 	int must_check_alg = 0;
   1197 	u_int16_t cpi = 0;
   1198 	u_int8_t algo = 0;
   1199 
   1200 	if ((sport != 0) && (dport != 0))
   1201 		chkport = PORT_STRICT;
   1202 	else
   1203 		chkport = PORT_NONE;
   1204 
   1205 	KASSERT(dst != NULL);
   1206 
   1207 	/*
   1208 	 * XXX IPCOMP case
   1209 	 * We use cpi to define spi here. In the case where cpi <=
   1210 	 * IPCOMP_CPI_NEGOTIATE_MIN, cpi just define the algorithm used, not
   1211 	 * the real spi. In this case, don't check the spi but check the
   1212 	 * algorithm
   1213 	 */
   1214 
   1215 	if (proto == IPPROTO_IPCOMP) {
   1216 		u_int32_t tmp;
   1217 		tmp = ntohl(spi);
   1218 		cpi = (u_int16_t) tmp;
   1219 		if (cpi < IPCOMP_CPI_NEGOTIATE_MIN) {
   1220 			algo = (u_int8_t) cpi;
   1221 			must_check_spi = 0;
   1222 			must_check_alg = 1;
   1223 		}
   1224 	}
   1225 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1226 	    "DP from %s:%u check_spi=%d, check_alg=%d\n",
   1227 	    where, tag, must_check_spi, must_check_alg);
   1228 
   1229 
   1230 	/*
   1231 	 * searching SAD.
   1232 	 * XXX: to be checked internal IP header somewhere.  Also when
   1233 	 * IPsec tunnel packet is received.  But ESP tunnel mode is
   1234 	 * encrypted so we can't check internal IP header.
   1235 	 */
   1236 	if (key_prefered_oldsa) {
   1237 		saorder_state_valid = saorder_state_valid_prefer_old;
   1238 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_old);
   1239 	} else {
   1240 		saorder_state_valid = saorder_state_valid_prefer_new;
   1241 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_new);
   1242 	}
   1243 	s = pserialize_read_enter();
   1244 	SAHLIST_READER_FOREACH(sah) {
   1245 		/* search valid state */
   1246 		for (stateidx = 0; stateidx < arraysize; stateidx++) {
   1247 			state = saorder_state_valid[stateidx];
   1248 			SAVLIST_READER_FOREACH(sav, sah, state) {
   1249 				KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   1250 				    "try match spi %#x, %#x\n",
   1251 				    ntohl(spi), ntohl(sav->spi));
   1252 				/* sanity check */
   1253 				KEY_CHKSASTATE(sav->state, state);
   1254 				/* do not return entries w/ unusable state */
   1255 				if (!SADB_SASTATE_USABLE_P(sav)) {
   1256 					KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   1257 					    "bad state %d\n", sav->state);
   1258 					continue;
   1259 				}
   1260 				if (proto != sav->sah->saidx.proto) {
   1261 					KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   1262 					    "proto fail %d != %d\n",
   1263 					    proto, sav->sah->saidx.proto);
   1264 					continue;
   1265 				}
   1266 				if (must_check_spi && spi != sav->spi) {
   1267 					KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   1268 					    "spi fail %#x != %#x\n",
   1269 					    ntohl(spi), ntohl(sav->spi));
   1270 					continue;
   1271 				}
   1272 				/* XXX only on the ipcomp case */
   1273 				if (must_check_alg && algo != sav->alg_comp) {
   1274 					KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   1275 					    "algo fail %d != %d\n",
   1276 					    algo, sav->alg_comp);
   1277 					continue;
   1278 				}
   1279 
   1280 #if 0	/* don't check src */
   1281 	/* Fix port in src->sa */
   1282 
   1283 				/* check src address */
   1284 				if (!key_sockaddr_match(&src->sa, &sav->sah->saidx.src.sa, PORT_NONE))
   1285 					continue;
   1286 #endif
   1287 				/* fix port of dst address XXX*/
   1288 				key_porttosaddr(__UNCONST(dst), dport);
   1289 				/* check dst address */
   1290 				if (!key_sockaddr_match(&dst->sa, &sav->sah->saidx.dst.sa, chkport))
   1291 					continue;
   1292 				key_sa_ref(sav, where, tag);
   1293 				goto done;
   1294 			}
   1295 		}
   1296 	}
   1297 	sav = NULL;
   1298 done:
   1299 	pserialize_read_exit(s);
   1300 
   1301 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1302 	    "DP return SA:%p; refcnt %u\n", sav, key_sa_refcnt(sav));
   1303 	return sav;
   1304 }
   1305 
   1306 static void
   1307 key_validate_savlist(const struct secashead *sah, const u_int state)
   1308 {
   1309 #ifdef DEBUG
   1310 	struct secasvar *sav, *next;
   1311 	int s;
   1312 
   1313 	/*
   1314 	 * The list should be sorted by lft_c->sadb_lifetime_addtime
   1315 	 * in ascending order.
   1316 	 */
   1317 	s = pserialize_read_enter();
   1318 	SAVLIST_READER_FOREACH(sav, sah, state) {
   1319 		next = SAVLIST_READER_NEXT(sav);
   1320 		if (next != NULL &&
   1321 		    sav->lft_c != NULL && next->lft_c != NULL) {
   1322 			KDASSERTMSG(sav->lft_c->sadb_lifetime_addtime <=
   1323 			    next->lft_c->sadb_lifetime_addtime,
   1324 			    "savlist is not sorted: sah=%p, state=%d, "
   1325 			    "sav=%" PRIu64 ", next=%" PRIu64, sah, state,
   1326 			    sav->lft_c->sadb_lifetime_addtime,
   1327 			    next->lft_c->sadb_lifetime_addtime);
   1328 		}
   1329 	}
   1330 	pserialize_read_exit(s);
   1331 #endif
   1332 }
   1333 
   1334 void
   1335 key_init_sp(struct secpolicy *sp)
   1336 {
   1337 
   1338 	ASSERT_SLEEPABLE();
   1339 
   1340 	sp->state = IPSEC_SPSTATE_ALIVE;
   1341 	if (sp->policy == IPSEC_POLICY_IPSEC)
   1342 		KASSERT(sp->req != NULL);
   1343 	localcount_init(&sp->localcount);
   1344 	SPLIST_ENTRY_INIT(sp);
   1345 }
   1346 
   1347 /*
   1348  * Must be called in a pserialize read section. A held SP
   1349  * must be released by key_sp_unref after use.
   1350  */
   1351 void
   1352 key_sp_ref(struct secpolicy *sp, const char* where, int tag)
   1353 {
   1354 
   1355 	localcount_acquire(&sp->localcount);
   1356 
   1357 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1358 	    "DP SP:%p (ID=%u) from %s:%u; refcnt++ now %u\n",
   1359 	    sp, sp->id, where, tag, key_sp_refcnt(sp));
   1360 }
   1361 
   1362 /*
   1363  * Must be called without holding key_spd.lock because the lock
   1364  * would be held in localcount_release.
   1365  */
   1366 void
   1367 key_sp_unref(struct secpolicy *sp, const char* where, int tag)
   1368 {
   1369 
   1370 	KDASSERT(mutex_ownable(&key_spd.lock));
   1371 
   1372 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1373 	    "DP SP:%p (ID=%u) from %s:%u; refcnt-- now %u\n",
   1374 	    sp, sp->id, where, tag, key_sp_refcnt(sp));
   1375 
   1376 	localcount_release(&sp->localcount, &key_spd.cv_lc, &key_spd.lock);
   1377 }
   1378 
   1379 static void
   1380 key_init_sav(struct secasvar *sav)
   1381 {
   1382 
   1383 	ASSERT_SLEEPABLE();
   1384 
   1385 	localcount_init(&sav->localcount);
   1386 	SAVLIST_ENTRY_INIT(sav);
   1387 }
   1388 
   1389 u_int
   1390 key_sa_refcnt(const struct secasvar *sav)
   1391 {
   1392 
   1393 	/* FIXME */
   1394 	return 0;
   1395 }
   1396 
   1397 void
   1398 key_sa_ref(struct secasvar *sav, const char* where, int tag)
   1399 {
   1400 
   1401 	localcount_acquire(&sav->localcount);
   1402 
   1403 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1404 	    "DP cause refcnt++: SA:%p from %s:%u\n",
   1405 	    sav, where, tag);
   1406 }
   1407 
   1408 void
   1409 key_sa_unref(struct secasvar *sav, const char* where, int tag)
   1410 {
   1411 
   1412 	KDASSERT(mutex_ownable(&key_sad.lock));
   1413 
   1414 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1415 	    "DP cause refcnt--: SA:%p from %s:%u\n",
   1416 	    sav, where, tag);
   1417 
   1418 	localcount_release(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
   1419 }
   1420 
   1421 #if 0
   1422 /*
   1423  * Must be called after calling key_lookup_sp*().
   1424  * For the packet with socket.
   1425  */
   1426 static void
   1427 key_freeso(struct socket *so)
   1428 {
   1429 	/* sanity check */
   1430 	KASSERT(so != NULL);
   1431 
   1432 	switch (so->so_proto->pr_domain->dom_family) {
   1433 #ifdef INET
   1434 	case PF_INET:
   1435 	    {
   1436 		struct inpcb *pcb = sotoinpcb(so);
   1437 
   1438 		/* Does it have a PCB ? */
   1439 		if (pcb == NULL)
   1440 			return;
   1441 
   1442 		struct inpcbpolicy *sp = pcb->inp_sp;
   1443 		key_freesp_so(&sp->sp_in);
   1444 		key_freesp_so(&sp->sp_out);
   1445 	    }
   1446 		break;
   1447 #endif
   1448 #ifdef INET6
   1449 	case PF_INET6:
   1450 	    {
   1451 #ifdef HAVE_NRL_INPCB
   1452 		struct inpcb *pcb  = sotoinpcb(so);
   1453 		struct inpcbpolicy *sp = pcb->inp_sp;
   1454 
   1455 		/* Does it have a PCB ? */
   1456 		if (pcb == NULL)
   1457 			return;
   1458 		key_freesp_so(&sp->sp_in);
   1459 		key_freesp_so(&sp->sp_out);
   1460 #else
   1461 		struct in6pcb *pcb  = sotoin6pcb(so);
   1462 
   1463 		/* Does it have a PCB ? */
   1464 		if (pcb == NULL)
   1465 			return;
   1466 		key_freesp_so(&pcb->in6p_sp->sp_in);
   1467 		key_freesp_so(&pcb->in6p_sp->sp_out);
   1468 #endif
   1469 	    }
   1470 		break;
   1471 #endif /* INET6 */
   1472 	default:
   1473 		IPSECLOG(LOG_DEBUG, "unknown address family=%d.\n",
   1474 		    so->so_proto->pr_domain->dom_family);
   1475 		return;
   1476 	}
   1477 }
   1478 
   1479 static void
   1480 key_freesp_so(struct secpolicy **sp)
   1481 {
   1482 
   1483 	KASSERT(sp != NULL);
   1484 	KASSERT(*sp != NULL);
   1485 
   1486 	if ((*sp)->policy == IPSEC_POLICY_ENTRUST ||
   1487 	    (*sp)->policy == IPSEC_POLICY_BYPASS)
   1488 		return;
   1489 
   1490 	KASSERTMSG((*sp)->policy == IPSEC_POLICY_IPSEC,
   1491 	    "invalid policy %u", (*sp)->policy);
   1492 	KEY_SP_UNREF(&sp);
   1493 }
   1494 #endif
   1495 
   1496 #ifdef NET_MPSAFE
   1497 static void
   1498 key_sad_pserialize_perform(void)
   1499 {
   1500 
   1501 	KASSERT(mutex_owned(&key_sad.lock));
   1502 
   1503 	while (key_sad.psz_performing)
   1504 		cv_wait(&key_sad.cv_psz, &key_sad.lock);
   1505 	key_sad.psz_performing = true;
   1506 	mutex_exit(&key_sad.lock);
   1507 
   1508 	pserialize_perform(key_sad.psz);
   1509 
   1510 	mutex_enter(&key_sad.lock);
   1511 	key_sad.psz_performing = false;
   1512 	cv_broadcast(&key_sad.cv_psz);
   1513 }
   1514 #endif
   1515 
   1516 /*
   1517  * Remove the sav from the savlist of its sah and wait for references to the sav
   1518  * to be released. key_sad.lock must be held.
   1519  */
   1520 static void
   1521 key_unlink_sav(struct secasvar *sav)
   1522 {
   1523 
   1524 	KASSERT(mutex_owned(&key_sad.lock));
   1525 
   1526 	SAVLIST_WRITER_REMOVE(sav);
   1527 
   1528 #ifdef NET_MPSAFE
   1529 	KASSERT(mutex_ownable(softnet_lock));
   1530 	key_sad_pserialize_perform();
   1531 #endif
   1532 
   1533 	localcount_drain(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
   1534 }
   1535 
   1536 /*
   1537  * Destroy an sav where the sav must be unlinked from an sah
   1538  * by say key_unlink_sav.
   1539  */
   1540 static void
   1541 key_destroy_sav(struct secasvar *sav)
   1542 {
   1543 
   1544 	ASSERT_SLEEPABLE();
   1545 
   1546 	localcount_fini(&sav->localcount);
   1547 	SAVLIST_ENTRY_DESTROY(sav);
   1548 
   1549 	key_delsav(sav);
   1550 }
   1551 
   1552 /*
   1553  * Destroy sav with holding its reference.
   1554  */
   1555 static void
   1556 key_destroy_sav_with_ref(struct secasvar *sav)
   1557 {
   1558 
   1559 	ASSERT_SLEEPABLE();
   1560 
   1561 	mutex_enter(&key_sad.lock);
   1562 	sav->state = SADB_SASTATE_DEAD;
   1563 	SAVLIST_WRITER_REMOVE(sav);
   1564 	mutex_exit(&key_sad.lock);
   1565 
   1566 	/* We cannot unref with holding key_sad.lock */
   1567 	KEY_SA_UNREF(&sav);
   1568 
   1569 	mutex_enter(&key_sad.lock);
   1570 #ifdef NET_MPSAFE
   1571 	KASSERT(mutex_ownable(softnet_lock));
   1572 	key_sad_pserialize_perform();
   1573 #endif
   1574 	localcount_drain(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
   1575 	mutex_exit(&key_sad.lock);
   1576 
   1577 	key_destroy_sav(sav);
   1578 }
   1579 
   1580 /* %%% SPD management */
   1581 /*
   1582  * free security policy entry.
   1583  */
   1584 static void
   1585 key_destroy_sp(struct secpolicy *sp)
   1586 {
   1587 
   1588 	SPLIST_ENTRY_DESTROY(sp);
   1589 	localcount_fini(&sp->localcount);
   1590 
   1591 	key_free_sp(sp);
   1592 
   1593 	key_update_used();
   1594 }
   1595 
   1596 void
   1597 key_free_sp(struct secpolicy *sp)
   1598 {
   1599 	struct ipsecrequest *isr = sp->req, *nextisr;
   1600 
   1601 	while (isr != NULL) {
   1602 		nextisr = isr->next;
   1603 		kmem_free(isr, sizeof(*isr));
   1604 		isr = nextisr;
   1605 	}
   1606 
   1607 	kmem_free(sp, sizeof(*sp));
   1608 }
   1609 
   1610 void
   1611 key_socksplist_add(struct secpolicy *sp)
   1612 {
   1613 
   1614 	mutex_enter(&key_spd.lock);
   1615 	PSLIST_WRITER_INSERT_HEAD(&key_spd.socksplist, sp, pslist_entry);
   1616 	mutex_exit(&key_spd.lock);
   1617 
   1618 	key_update_used();
   1619 }
   1620 
   1621 /*
   1622  * search SPD
   1623  * OUT:	NULL	: not found
   1624  *	others	: found, pointer to a SP.
   1625  */
   1626 static struct secpolicy *
   1627 key_getsp(const struct secpolicyindex *spidx)
   1628 {
   1629 	struct secpolicy *sp;
   1630 	int s;
   1631 
   1632 	KASSERT(spidx != NULL);
   1633 
   1634 	s = pserialize_read_enter();
   1635 	SPLIST_READER_FOREACH(sp, spidx->dir) {
   1636 		if (sp->state == IPSEC_SPSTATE_DEAD)
   1637 			continue;
   1638 		if (key_spidx_match_exactly(spidx, &sp->spidx)) {
   1639 			KEY_SP_REF(sp);
   1640 			pserialize_read_exit(s);
   1641 			return sp;
   1642 		}
   1643 	}
   1644 	pserialize_read_exit(s);
   1645 
   1646 	return NULL;
   1647 }
   1648 
   1649 /*
   1650  * search SPD and remove found SP
   1651  * OUT:	NULL	: not found
   1652  *	others	: found, pointer to a SP.
   1653  */
   1654 static struct secpolicy *
   1655 key_lookup_and_remove_sp(const struct secpolicyindex *spidx)
   1656 {
   1657 	struct secpolicy *sp = NULL;
   1658 
   1659 	mutex_enter(&key_spd.lock);
   1660 	SPLIST_WRITER_FOREACH(sp, spidx->dir) {
   1661 		KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
   1662 
   1663 		if (key_spidx_match_exactly(spidx, &sp->spidx)) {
   1664 			key_unlink_sp(sp);
   1665 			goto out;
   1666 		}
   1667 	}
   1668 	sp = NULL;
   1669 out:
   1670 	mutex_exit(&key_spd.lock);
   1671 
   1672 	return sp;
   1673 }
   1674 
   1675 /*
   1676  * get SP by index.
   1677  * OUT:	NULL	: not found
   1678  *	others	: found, pointer to a SP.
   1679  */
   1680 static struct secpolicy *
   1681 key_getspbyid(u_int32_t id)
   1682 {
   1683 	struct secpolicy *sp;
   1684 	int s;
   1685 
   1686 	s = pserialize_read_enter();
   1687 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_INBOUND) {
   1688 		if (sp->state == IPSEC_SPSTATE_DEAD)
   1689 			continue;
   1690 		if (sp->id == id) {
   1691 			KEY_SP_REF(sp);
   1692 			goto out;
   1693 		}
   1694 	}
   1695 
   1696 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_OUTBOUND) {
   1697 		if (sp->state == IPSEC_SPSTATE_DEAD)
   1698 			continue;
   1699 		if (sp->id == id) {
   1700 			KEY_SP_REF(sp);
   1701 			goto out;
   1702 		}
   1703 	}
   1704 out:
   1705 	pserialize_read_exit(s);
   1706 	return sp;
   1707 }
   1708 
   1709 /*
   1710  * get SP by index, remove and return it.
   1711  * OUT:	NULL	: not found
   1712  *	others	: found, pointer to a SP.
   1713  */
   1714 static struct secpolicy *
   1715 key_lookupbyid_and_remove_sp(u_int32_t id)
   1716 {
   1717 	struct secpolicy *sp;
   1718 
   1719 	mutex_enter(&key_spd.lock);
   1720 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_INBOUND) {
   1721 		KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
   1722 		if (sp->id == id)
   1723 			goto out;
   1724 	}
   1725 
   1726 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_OUTBOUND) {
   1727 		KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
   1728 		if (sp->id == id)
   1729 			goto out;
   1730 	}
   1731 out:
   1732 	if (sp != NULL)
   1733 		key_unlink_sp(sp);
   1734 	mutex_exit(&key_spd.lock);
   1735 	return sp;
   1736 }
   1737 
   1738 struct secpolicy *
   1739 key_newsp(const char* where, int tag)
   1740 {
   1741 	struct secpolicy *newsp = NULL;
   1742 
   1743 	newsp = kmem_zalloc(sizeof(struct secpolicy), KM_SLEEP);
   1744 
   1745 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1746 	    "DP from %s:%u return SP:%p\n", where, tag, newsp);
   1747 	return newsp;
   1748 }
   1749 
   1750 /*
   1751  * create secpolicy structure from sadb_x_policy structure.
   1752  * NOTE: `state', `secpolicyindex' in secpolicy structure are not set,
   1753  * so must be set properly later.
   1754  */
   1755 struct secpolicy *
   1756 key_msg2sp(const struct sadb_x_policy *xpl0, size_t len, int *error)
   1757 {
   1758 	struct secpolicy *newsp;
   1759 
   1760 	KASSERT(!cpu_softintr_p());
   1761 	KASSERT(xpl0 != NULL);
   1762 	KASSERT(len >= sizeof(*xpl0));
   1763 
   1764 	if (len != PFKEY_EXTLEN(xpl0)) {
   1765 		IPSECLOG(LOG_DEBUG, "Invalid msg length.\n");
   1766 		*error = EINVAL;
   1767 		return NULL;
   1768 	}
   1769 
   1770 	newsp = KEY_NEWSP();
   1771 	if (newsp == NULL) {
   1772 		*error = ENOBUFS;
   1773 		return NULL;
   1774 	}
   1775 
   1776 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
   1777 	newsp->policy = xpl0->sadb_x_policy_type;
   1778 
   1779 	/* check policy */
   1780 	switch (xpl0->sadb_x_policy_type) {
   1781 	case IPSEC_POLICY_DISCARD:
   1782 	case IPSEC_POLICY_NONE:
   1783 	case IPSEC_POLICY_ENTRUST:
   1784 	case IPSEC_POLICY_BYPASS:
   1785 		newsp->req = NULL;
   1786 		*error = 0;
   1787 		return newsp;
   1788 
   1789 	case IPSEC_POLICY_IPSEC:
   1790 		/* Continued */
   1791 		break;
   1792 	default:
   1793 		IPSECLOG(LOG_DEBUG, "invalid policy type.\n");
   1794 		key_free_sp(newsp);
   1795 		*error = EINVAL;
   1796 		return NULL;
   1797 	}
   1798 
   1799 	/* IPSEC_POLICY_IPSEC */
   1800     {
   1801 	int tlen;
   1802 	const struct sadb_x_ipsecrequest *xisr;
   1803 	uint16_t xisr_reqid;
   1804 	struct ipsecrequest **p_isr = &newsp->req;
   1805 
   1806 	/* validity check */
   1807 	if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
   1808 		IPSECLOG(LOG_DEBUG, "Invalid msg length.\n");
   1809 		*error = EINVAL;
   1810 		goto free_exit;
   1811 	}
   1812 
   1813 	tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
   1814 	xisr = (const struct sadb_x_ipsecrequest *)(xpl0 + 1);
   1815 
   1816 	while (tlen > 0) {
   1817 		/* length check */
   1818 		if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr)) {
   1819 			IPSECLOG(LOG_DEBUG, "invalid ipsecrequest length.\n");
   1820 			*error = EINVAL;
   1821 			goto free_exit;
   1822 		}
   1823 
   1824 		/* allocate request buffer */
   1825 		*p_isr = kmem_zalloc(sizeof(**p_isr), KM_SLEEP);
   1826 
   1827 		/* set values */
   1828 		(*p_isr)->next = NULL;
   1829 
   1830 		switch (xisr->sadb_x_ipsecrequest_proto) {
   1831 		case IPPROTO_ESP:
   1832 		case IPPROTO_AH:
   1833 		case IPPROTO_IPCOMP:
   1834 			break;
   1835 		default:
   1836 			IPSECLOG(LOG_DEBUG, "invalid proto type=%u\n",
   1837 			    xisr->sadb_x_ipsecrequest_proto);
   1838 			*error = EPROTONOSUPPORT;
   1839 			goto free_exit;
   1840 		}
   1841 		(*p_isr)->saidx.proto = xisr->sadb_x_ipsecrequest_proto;
   1842 
   1843 		switch (xisr->sadb_x_ipsecrequest_mode) {
   1844 		case IPSEC_MODE_TRANSPORT:
   1845 		case IPSEC_MODE_TUNNEL:
   1846 			break;
   1847 		case IPSEC_MODE_ANY:
   1848 		default:
   1849 			IPSECLOG(LOG_DEBUG, "invalid mode=%u\n",
   1850 			    xisr->sadb_x_ipsecrequest_mode);
   1851 			*error = EINVAL;
   1852 			goto free_exit;
   1853 		}
   1854 		(*p_isr)->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
   1855 
   1856 		switch (xisr->sadb_x_ipsecrequest_level) {
   1857 		case IPSEC_LEVEL_DEFAULT:
   1858 		case IPSEC_LEVEL_USE:
   1859 		case IPSEC_LEVEL_REQUIRE:
   1860 			break;
   1861 		case IPSEC_LEVEL_UNIQUE:
   1862 			xisr_reqid = xisr->sadb_x_ipsecrequest_reqid;
   1863 			/* validity check */
   1864 			/*
   1865 			 * If range violation of reqid, kernel will
   1866 			 * update it, don't refuse it.
   1867 			 */
   1868 			if (xisr_reqid > IPSEC_MANUAL_REQID_MAX) {
   1869 				IPSECLOG(LOG_DEBUG,
   1870 				    "reqid=%d range "
   1871 				    "violation, updated by kernel.\n",
   1872 				    xisr_reqid);
   1873 				xisr_reqid = 0;
   1874 			}
   1875 
   1876 			/* allocate new reqid id if reqid is zero. */
   1877 			if (xisr_reqid == 0) {
   1878 				u_int16_t reqid = key_newreqid();
   1879 				if (reqid == 0) {
   1880 					*error = ENOBUFS;
   1881 					goto free_exit;
   1882 				}
   1883 				(*p_isr)->saidx.reqid = reqid;
   1884 			} else {
   1885 			/* set it for manual keying. */
   1886 				(*p_isr)->saidx.reqid = xisr_reqid;
   1887 			}
   1888 			break;
   1889 
   1890 		default:
   1891 			IPSECLOG(LOG_DEBUG, "invalid level=%u\n",
   1892 			    xisr->sadb_x_ipsecrequest_level);
   1893 			*error = EINVAL;
   1894 			goto free_exit;
   1895 		}
   1896 		(*p_isr)->level = xisr->sadb_x_ipsecrequest_level;
   1897 
   1898 		/* set IP addresses if there */
   1899 		if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
   1900 			const struct sockaddr *paddr;
   1901 
   1902 			paddr = (const struct sockaddr *)(xisr + 1);
   1903 
   1904 			/* validity check */
   1905 			if (paddr->sa_len > sizeof((*p_isr)->saidx.src)) {
   1906 				IPSECLOG(LOG_DEBUG, "invalid request "
   1907 				    "address length.\n");
   1908 				*error = EINVAL;
   1909 				goto free_exit;
   1910 			}
   1911 			memcpy(&(*p_isr)->saidx.src, paddr, paddr->sa_len);
   1912 
   1913 			paddr = (const struct sockaddr *)((const char *)paddr
   1914 			    + paddr->sa_len);
   1915 
   1916 			/* validity check */
   1917 			if (paddr->sa_len > sizeof((*p_isr)->saidx.dst)) {
   1918 				IPSECLOG(LOG_DEBUG, "invalid request "
   1919 				    "address length.\n");
   1920 				*error = EINVAL;
   1921 				goto free_exit;
   1922 			}
   1923 			memcpy(&(*p_isr)->saidx.dst, paddr, paddr->sa_len);
   1924 		}
   1925 
   1926 		(*p_isr)->sp = newsp;
   1927 
   1928 		/* initialization for the next. */
   1929 		p_isr = &(*p_isr)->next;
   1930 		tlen -= xisr->sadb_x_ipsecrequest_len;
   1931 
   1932 		/* validity check */
   1933 		if (tlen < 0) {
   1934 			IPSECLOG(LOG_DEBUG, "becoming tlen < 0.\n");
   1935 			*error = EINVAL;
   1936 			goto free_exit;
   1937 		}
   1938 
   1939 		xisr = (const struct sadb_x_ipsecrequest *)((const char *)xisr +
   1940 		    xisr->sadb_x_ipsecrequest_len);
   1941 	}
   1942     }
   1943 
   1944 	*error = 0;
   1945 	return newsp;
   1946 
   1947 free_exit:
   1948 	key_free_sp(newsp);
   1949 	return NULL;
   1950 }
   1951 
   1952 static u_int16_t
   1953 key_newreqid(void)
   1954 {
   1955 	static u_int16_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
   1956 
   1957 	auto_reqid = (auto_reqid == 0xffff ?
   1958 	    IPSEC_MANUAL_REQID_MAX + 1 : auto_reqid + 1);
   1959 
   1960 	/* XXX should be unique check */
   1961 
   1962 	return auto_reqid;
   1963 }
   1964 
   1965 /*
   1966  * copy secpolicy struct to sadb_x_policy structure indicated.
   1967  */
   1968 struct mbuf *
   1969 key_sp2msg(const struct secpolicy *sp, int mflag)
   1970 {
   1971 	struct sadb_x_policy *xpl;
   1972 	int tlen;
   1973 	char *p;
   1974 	struct mbuf *m;
   1975 
   1976 	KASSERT(sp != NULL);
   1977 
   1978 	tlen = key_getspreqmsglen(sp);
   1979 
   1980 	m = key_alloc_mbuf(tlen, mflag);
   1981 	if (!m || m->m_next) {	/*XXX*/
   1982 		if (m)
   1983 			m_freem(m);
   1984 		return NULL;
   1985 	}
   1986 
   1987 	m->m_len = tlen;
   1988 	m->m_next = NULL;
   1989 	xpl = mtod(m, struct sadb_x_policy *);
   1990 	memset(xpl, 0, tlen);
   1991 
   1992 	xpl->sadb_x_policy_len = PFKEY_UNIT64(tlen);
   1993 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
   1994 	xpl->sadb_x_policy_type = sp->policy;
   1995 	xpl->sadb_x_policy_dir = sp->spidx.dir;
   1996 	xpl->sadb_x_policy_id = sp->id;
   1997 	p = (char *)xpl + sizeof(*xpl);
   1998 
   1999 	/* if is the policy for ipsec ? */
   2000 	if (sp->policy == IPSEC_POLICY_IPSEC) {
   2001 		struct sadb_x_ipsecrequest *xisr;
   2002 		struct ipsecrequest *isr;
   2003 
   2004 		for (isr = sp->req; isr != NULL; isr = isr->next) {
   2005 
   2006 			xisr = (struct sadb_x_ipsecrequest *)p;
   2007 
   2008 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
   2009 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
   2010 			xisr->sadb_x_ipsecrequest_level = isr->level;
   2011 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
   2012 
   2013 			p += sizeof(*xisr);
   2014 			memcpy(p, &isr->saidx.src, isr->saidx.src.sa.sa_len);
   2015 			p += isr->saidx.src.sa.sa_len;
   2016 			memcpy(p, &isr->saidx.dst, isr->saidx.dst.sa.sa_len);
   2017 			p += isr->saidx.src.sa.sa_len;
   2018 
   2019 			xisr->sadb_x_ipsecrequest_len =
   2020 			    PFKEY_ALIGN8(sizeof(*xisr)
   2021 			    + isr->saidx.src.sa.sa_len
   2022 			    + isr->saidx.dst.sa.sa_len);
   2023 		}
   2024 	}
   2025 
   2026 	return m;
   2027 }
   2028 
   2029 /*
   2030  * m will not be freed nor modified. It never return NULL.
   2031  * If it returns a mbuf of M_PKTHDR, the mbuf ensures to have
   2032  * contiguous length at least sizeof(struct sadb_msg).
   2033  */
   2034 static struct mbuf *
   2035 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
   2036 		int ndeep, int nitem, ...)
   2037 {
   2038 	va_list ap;
   2039 	int idx;
   2040 	int i;
   2041 	struct mbuf *result = NULL, *n;
   2042 	int len;
   2043 
   2044 	KASSERT(m != NULL);
   2045 	KASSERT(mhp != NULL);
   2046 	KASSERT(!cpu_softintr_p());
   2047 
   2048 	va_start(ap, nitem);
   2049 	for (i = 0; i < nitem; i++) {
   2050 		idx = va_arg(ap, int);
   2051 		KASSERT(idx >= 0);
   2052 		KASSERT(idx <= SADB_EXT_MAX);
   2053 		/* don't attempt to pull empty extension */
   2054 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
   2055 			continue;
   2056 		if (idx != SADB_EXT_RESERVED &&
   2057 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
   2058 			continue;
   2059 
   2060 		if (idx == SADB_EXT_RESERVED) {
   2061 			CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MHLEN);
   2062 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
   2063 			MGETHDR(n, M_WAITOK, MT_DATA);
   2064 			n->m_len = len;
   2065 			n->m_next = NULL;
   2066 			m_copydata(m, 0, sizeof(struct sadb_msg),
   2067 			    mtod(n, void *));
   2068 		} else if (i < ndeep) {
   2069 			len = mhp->extlen[idx];
   2070 			n = key_alloc_mbuf(len, M_WAITOK);
   2071 			KASSERT(n->m_next == NULL);
   2072 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
   2073 			    mtod(n, void *));
   2074 		} else {
   2075 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
   2076 			    M_WAITOK);
   2077 		}
   2078 		KASSERT(n != NULL);
   2079 
   2080 		if (result)
   2081 			m_cat(result, n);
   2082 		else
   2083 			result = n;
   2084 	}
   2085 	va_end(ap);
   2086 
   2087 	KASSERT(result != NULL);
   2088 	if ((result->m_flags & M_PKTHDR) != 0) {
   2089 		result->m_pkthdr.len = 0;
   2090 		for (n = result; n; n = n->m_next)
   2091 			result->m_pkthdr.len += n->m_len;
   2092 		KASSERT(result->m_len >= sizeof(struct sadb_msg));
   2093 	}
   2094 
   2095 	return result;
   2096 }
   2097 
   2098 /*
   2099  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
   2100  * add an entry to SP database, when received
   2101  *   <base, address(SD), (lifetime(H),) policy>
   2102  * from the user(?).
   2103  * Adding to SP database,
   2104  * and send
   2105  *   <base, address(SD), (lifetime(H),) policy>
   2106  * to the socket which was send.
   2107  *
   2108  * SPDADD set a unique policy entry.
   2109  * SPDSETIDX like SPDADD without a part of policy requests.
   2110  * SPDUPDATE replace a unique policy entry.
   2111  *
   2112  * m will always be freed.
   2113  */
   2114 static int
   2115 key_api_spdadd(struct socket *so, struct mbuf *m,
   2116 	   const struct sadb_msghdr *mhp)
   2117 {
   2118 	const struct sockaddr *src, *dst;
   2119 	const struct sadb_x_policy *xpl0;
   2120 	struct sadb_x_policy *xpl;
   2121 	const struct sadb_lifetime *lft = NULL;
   2122 	struct secpolicyindex spidx;
   2123 	struct secpolicy *newsp;
   2124 	int error;
   2125 
   2126 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   2127 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
   2128 	    mhp->ext[SADB_X_EXT_POLICY] == NULL) {
   2129 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2130 		return key_senderror(so, m, EINVAL);
   2131 	}
   2132 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   2133 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
   2134 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
   2135 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2136 		return key_senderror(so, m, EINVAL);
   2137 	}
   2138 	if (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL) {
   2139 		if (mhp->extlen[SADB_EXT_LIFETIME_HARD] <
   2140 		    sizeof(struct sadb_lifetime)) {
   2141 			IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2142 			return key_senderror(so, m, EINVAL);
   2143 		}
   2144 		lft = mhp->ext[SADB_EXT_LIFETIME_HARD];
   2145 	}
   2146 
   2147 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
   2148 
   2149 	/* checking the direciton. */
   2150 	switch (xpl0->sadb_x_policy_dir) {
   2151 	case IPSEC_DIR_INBOUND:
   2152 	case IPSEC_DIR_OUTBOUND:
   2153 		break;
   2154 	default:
   2155 		IPSECLOG(LOG_DEBUG, "Invalid SP direction.\n");
   2156 		return key_senderror(so, m, EINVAL);
   2157 	}
   2158 
   2159 	/* check policy */
   2160 	/* key_api_spdadd() accepts DISCARD, NONE and IPSEC. */
   2161 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_ENTRUST ||
   2162 	    xpl0->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
   2163 		IPSECLOG(LOG_DEBUG, "Invalid policy type.\n");
   2164 		return key_senderror(so, m, EINVAL);
   2165 	}
   2166 
   2167 	/* policy requests are mandatory when action is ipsec. */
   2168 	if (mhp->msg->sadb_msg_type != SADB_X_SPDSETIDX &&
   2169 	    xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
   2170 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
   2171 		IPSECLOG(LOG_DEBUG, "some policy requests part required.\n");
   2172 		return key_senderror(so, m, EINVAL);
   2173 	}
   2174 
   2175 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   2176 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   2177 
   2178 	/* sanity check on addr pair */
   2179 	if (src->sa_family != dst->sa_family)
   2180 		return key_senderror(so, m, EINVAL);
   2181 	if (src->sa_len != dst->sa_len)
   2182 		return key_senderror(so, m, EINVAL);
   2183 
   2184 	key_init_spidx_bymsghdr(&spidx, mhp);
   2185 
   2186 	/*
   2187 	 * checking there is SP already or not.
   2188 	 * SPDUPDATE doesn't depend on whether there is a SP or not.
   2189 	 * If the type is either SPDADD or SPDSETIDX AND a SP is found,
   2190 	 * then error.
   2191 	 */
   2192     {
   2193 	struct secpolicy *sp;
   2194 
   2195 	if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
   2196 		sp = key_lookup_and_remove_sp(&spidx);
   2197 		if (sp != NULL)
   2198 			key_destroy_sp(sp);
   2199 	} else {
   2200 		sp = key_getsp(&spidx);
   2201 		if (sp != NULL) {
   2202 			KEY_SP_UNREF(&sp);
   2203 			IPSECLOG(LOG_DEBUG, "a SP entry exists already.\n");
   2204 			return key_senderror(so, m, EEXIST);
   2205 		}
   2206 	}
   2207     }
   2208 
   2209 	/* allocation new SP entry */
   2210 	newsp = key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error);
   2211 	if (newsp == NULL) {
   2212 		return key_senderror(so, m, error);
   2213 	}
   2214 
   2215 	newsp->id = key_getnewspid();
   2216 	if (newsp->id == 0) {
   2217 		kmem_free(newsp, sizeof(*newsp));
   2218 		return key_senderror(so, m, ENOBUFS);
   2219 	}
   2220 
   2221 	newsp->spidx = spidx;
   2222 	newsp->created = time_uptime;
   2223 	newsp->lastused = newsp->created;
   2224 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
   2225 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
   2226 
   2227 	key_init_sp(newsp);
   2228 
   2229 	mutex_enter(&key_spd.lock);
   2230 	SPLIST_WRITER_INSERT_TAIL(newsp->spidx.dir, newsp);
   2231 	mutex_exit(&key_spd.lock);
   2232 
   2233 #ifdef notyet
   2234 	/* delete the entry in key_misc.spacqlist */
   2235 	if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
   2236 		struct secspacq *spacq = key_getspacq(&spidx);
   2237 		if (spacq != NULL) {
   2238 			/* reset counter in order to deletion by timehandler. */
   2239 			spacq->created = time_uptime;
   2240 			spacq->count = 0;
   2241 		}
   2242     	}
   2243 #endif
   2244 
   2245 	/* Invalidate all cached SPD pointers in the PCBs. */
   2246 	ipsec_invalpcbcacheall();
   2247 
   2248 #if defined(GATEWAY)
   2249 	/* Invalidate the ipflow cache, as well. */
   2250 	ipflow_invalidate_all(0);
   2251 #ifdef INET6
   2252 	if (in6_present)
   2253 		ip6flow_invalidate_all(0);
   2254 #endif /* INET6 */
   2255 #endif /* GATEWAY */
   2256 
   2257 	key_update_used();
   2258 
   2259     {
   2260 	struct mbuf *n, *mpolicy;
   2261 	int off;
   2262 
   2263 	/* create new sadb_msg to reply. */
   2264 	if (lft) {
   2265 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
   2266 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
   2267 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
   2268 	} else {
   2269 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
   2270 		    SADB_X_EXT_POLICY,
   2271 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
   2272 	}
   2273 
   2274 	key_fill_replymsg(n, 0);
   2275 	off = 0;
   2276 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
   2277 	    sizeof(*xpl), &off);
   2278 	if (mpolicy == NULL) {
   2279 		/* n is already freed */
   2280 		return key_senderror(so, m, ENOBUFS);
   2281 	}
   2282 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, char *) + off);
   2283 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
   2284 		m_freem(n);
   2285 		return key_senderror(so, m, EINVAL);
   2286 	}
   2287 	xpl->sadb_x_policy_id = newsp->id;
   2288 
   2289 	m_freem(m);
   2290 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   2291     }
   2292 }
   2293 
   2294 /*
   2295  * get new policy id.
   2296  * OUT:
   2297  *	0:	failure.
   2298  *	others: success.
   2299  */
   2300 static u_int32_t
   2301 key_getnewspid(void)
   2302 {
   2303 	u_int32_t newid = 0;
   2304 	int count = key_spi_trycnt;	/* XXX */
   2305 	struct secpolicy *sp;
   2306 
   2307 	/* when requesting to allocate spi ranged */
   2308 	while (count--) {
   2309 		newid = (policy_id = (policy_id == ~0 ? 1 : policy_id + 1));
   2310 
   2311 		sp = key_getspbyid(newid);
   2312 		if (sp == NULL)
   2313 			break;
   2314 
   2315 		KEY_SP_UNREF(&sp);
   2316 	}
   2317 
   2318 	if (count == 0 || newid == 0) {
   2319 		IPSECLOG(LOG_DEBUG, "to allocate policy id is failed.\n");
   2320 		return 0;
   2321 	}
   2322 
   2323 	return newid;
   2324 }
   2325 
   2326 /*
   2327  * SADB_SPDDELETE processing
   2328  * receive
   2329  *   <base, address(SD), policy(*)>
   2330  * from the user(?), and set SADB_SASTATE_DEAD,
   2331  * and send,
   2332  *   <base, address(SD), policy(*)>
   2333  * to the ikmpd.
   2334  * policy(*) including direction of policy.
   2335  *
   2336  * m will always be freed.
   2337  */
   2338 static int
   2339 key_api_spddelete(struct socket *so, struct mbuf *m,
   2340               const struct sadb_msghdr *mhp)
   2341 {
   2342 	struct sadb_x_policy *xpl0;
   2343 	struct secpolicyindex spidx;
   2344 	struct secpolicy *sp;
   2345 
   2346 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   2347 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
   2348 	    mhp->ext[SADB_X_EXT_POLICY] == NULL) {
   2349 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2350 		return key_senderror(so, m, EINVAL);
   2351 	}
   2352 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   2353 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
   2354 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
   2355 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2356 		return key_senderror(so, m, EINVAL);
   2357 	}
   2358 
   2359 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
   2360 
   2361 	/* checking the directon. */
   2362 	switch (xpl0->sadb_x_policy_dir) {
   2363 	case IPSEC_DIR_INBOUND:
   2364 	case IPSEC_DIR_OUTBOUND:
   2365 		break;
   2366 	default:
   2367 		IPSECLOG(LOG_DEBUG, "Invalid SP direction.\n");
   2368 		return key_senderror(so, m, EINVAL);
   2369 	}
   2370 
   2371 	/* make secindex */
   2372 	key_init_spidx_bymsghdr(&spidx, mhp);
   2373 
   2374 	/* Is there SP in SPD ? */
   2375 	sp = key_lookup_and_remove_sp(&spidx);
   2376 	if (sp == NULL) {
   2377 		IPSECLOG(LOG_DEBUG, "no SP found.\n");
   2378 		return key_senderror(so, m, EINVAL);
   2379 	}
   2380 
   2381 	/* save policy id to buffer to be returned. */
   2382 	xpl0->sadb_x_policy_id = sp->id;
   2383 
   2384 	key_destroy_sp(sp);
   2385 
   2386 	/* We're deleting policy; no need to invalidate the ipflow cache. */
   2387 
   2388     {
   2389 	struct mbuf *n;
   2390 
   2391 	/* create new sadb_msg to reply. */
   2392 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
   2393 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
   2394 	key_fill_replymsg(n, 0);
   2395 	m_freem(m);
   2396 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   2397     }
   2398 }
   2399 
   2400 static struct mbuf *
   2401 key_alloc_mbuf_simple(int len, int mflag)
   2402 {
   2403 	struct mbuf *n;
   2404 
   2405 	KASSERT(mflag == M_NOWAIT || (mflag == M_WAITOK && !cpu_softintr_p()));
   2406 
   2407 	MGETHDR(n, mflag, MT_DATA);
   2408 	if (n && len > MHLEN) {
   2409 		MCLGET(n, mflag);
   2410 		if ((n->m_flags & M_EXT) == 0) {
   2411 			m_freem(n);
   2412 			n = NULL;
   2413 		}
   2414 	}
   2415 	return n;
   2416 }
   2417 
   2418 /*
   2419  * SADB_SPDDELETE2 processing
   2420  * receive
   2421  *   <base, policy(*)>
   2422  * from the user(?), and set SADB_SASTATE_DEAD,
   2423  * and send,
   2424  *   <base, policy(*)>
   2425  * to the ikmpd.
   2426  * policy(*) including direction of policy.
   2427  *
   2428  * m will always be freed.
   2429  */
   2430 static int
   2431 key_api_spddelete2(struct socket *so, struct mbuf *m,
   2432 	       const struct sadb_msghdr *mhp)
   2433 {
   2434 	u_int32_t id;
   2435 	struct secpolicy *sp;
   2436 	const struct sadb_x_policy *xpl;
   2437 
   2438 	if (mhp->ext[SADB_X_EXT_POLICY] == NULL ||
   2439 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
   2440 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2441 		return key_senderror(so, m, EINVAL);
   2442 	}
   2443 
   2444 	xpl = mhp->ext[SADB_X_EXT_POLICY];
   2445 	id = xpl->sadb_x_policy_id;
   2446 
   2447 	/* Is there SP in SPD ? */
   2448 	sp = key_lookupbyid_and_remove_sp(id);
   2449 	if (sp == NULL) {
   2450 		IPSECLOG(LOG_DEBUG, "no SP found id:%u.\n", id);
   2451 		return key_senderror(so, m, EINVAL);
   2452 	}
   2453 
   2454 	key_destroy_sp(sp);
   2455 
   2456 	/* We're deleting policy; no need to invalidate the ipflow cache. */
   2457 
   2458     {
   2459 	struct mbuf *n, *nn;
   2460 	int off, len;
   2461 
   2462 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MCLBYTES);
   2463 
   2464 	/* create new sadb_msg to reply. */
   2465 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
   2466 
   2467 	n = key_alloc_mbuf_simple(len, M_WAITOK);
   2468 	n->m_len = len;
   2469 	n->m_next = NULL;
   2470 	off = 0;
   2471 
   2472 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
   2473 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
   2474 
   2475 	KASSERTMSG(off == len, "length inconsistency");
   2476 
   2477 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
   2478 	    mhp->extlen[SADB_X_EXT_POLICY], M_WAITOK);
   2479 
   2480 	n->m_pkthdr.len = 0;
   2481 	for (nn = n; nn; nn = nn->m_next)
   2482 		n->m_pkthdr.len += nn->m_len;
   2483 
   2484 	key_fill_replymsg(n, 0);
   2485 	m_freem(m);
   2486 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   2487     }
   2488 }
   2489 
   2490 /*
   2491  * SADB_X_GET processing
   2492  * receive
   2493  *   <base, policy(*)>
   2494  * from the user(?),
   2495  * and send,
   2496  *   <base, address(SD), policy>
   2497  * to the ikmpd.
   2498  * policy(*) including direction of policy.
   2499  *
   2500  * m will always be freed.
   2501  */
   2502 static int
   2503 key_api_spdget(struct socket *so, struct mbuf *m,
   2504 	   const struct sadb_msghdr *mhp)
   2505 {
   2506 	u_int32_t id;
   2507 	struct secpolicy *sp;
   2508 	struct mbuf *n;
   2509 	const struct sadb_x_policy *xpl;
   2510 
   2511 	if (mhp->ext[SADB_X_EXT_POLICY] == NULL ||
   2512 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
   2513 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2514 		return key_senderror(so, m, EINVAL);
   2515 	}
   2516 
   2517 	xpl = mhp->ext[SADB_X_EXT_POLICY];
   2518 	id = xpl->sadb_x_policy_id;
   2519 
   2520 	/* Is there SP in SPD ? */
   2521 	sp = key_getspbyid(id);
   2522 	if (sp == NULL) {
   2523 		IPSECLOG(LOG_DEBUG, "no SP found id:%u.\n", id);
   2524 		return key_senderror(so, m, ENOENT);
   2525 	}
   2526 
   2527 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
   2528 	    mhp->msg->sadb_msg_pid);
   2529 	KEY_SP_UNREF(&sp); /* ref gained by key_getspbyid */
   2530 	m_freem(m);
   2531 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
   2532 }
   2533 
   2534 #ifdef notyet
   2535 /*
   2536  * SADB_X_SPDACQUIRE processing.
   2537  * Acquire policy and SA(s) for a *OUTBOUND* packet.
   2538  * send
   2539  *   <base, policy(*)>
   2540  * to KMD, and expect to receive
   2541  *   <base> with SADB_X_SPDACQUIRE if error occurred,
   2542  * or
   2543  *   <base, policy>
   2544  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
   2545  * policy(*) is without policy requests.
   2546  *
   2547  *    0     : succeed
   2548  *    others: error number
   2549  */
   2550 int
   2551 key_spdacquire(const struct secpolicy *sp)
   2552 {
   2553 	struct mbuf *result = NULL, *m;
   2554 	struct secspacq *newspacq;
   2555 	int error;
   2556 
   2557 	KASSERT(sp != NULL);
   2558 	KASSERTMSG(sp->req == NULL, "called but there is request");
   2559 	KASSERTMSG(sp->policy == IPSEC_POLICY_IPSEC,
   2560 	    "policy mismathed. IPsec is expected");
   2561 
   2562 	/* Get an entry to check whether sent message or not. */
   2563 	newspacq = key_getspacq(&sp->spidx);
   2564 	if (newspacq != NULL) {
   2565 		if (key_blockacq_count < newspacq->count) {
   2566 			/* reset counter and do send message. */
   2567 			newspacq->count = 0;
   2568 		} else {
   2569 			/* increment counter and do nothing. */
   2570 			newspacq->count++;
   2571 			return 0;
   2572 		}
   2573 	} else {
   2574 		/* make new entry for blocking to send SADB_ACQUIRE. */
   2575 		newspacq = key_newspacq(&sp->spidx);
   2576 		if (newspacq == NULL)
   2577 			return ENOBUFS;
   2578 
   2579 		/* add to key_misc.acqlist */
   2580 		LIST_INSERT_HEAD(&key_misc.spacqlist, newspacq, chain);
   2581 	}
   2582 
   2583 	/* create new sadb_msg to reply. */
   2584 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
   2585 	if (!m) {
   2586 		error = ENOBUFS;
   2587 		goto fail;
   2588 	}
   2589 	result = m;
   2590 
   2591 	result->m_pkthdr.len = 0;
   2592 	for (m = result; m; m = m->m_next)
   2593 		result->m_pkthdr.len += m->m_len;
   2594 
   2595 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   2596 	    PFKEY_UNIT64(result->m_pkthdr.len);
   2597 
   2598 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
   2599 
   2600 fail:
   2601 	if (result)
   2602 		m_freem(result);
   2603 	return error;
   2604 }
   2605 #endif /* notyet */
   2606 
   2607 /*
   2608  * SADB_SPDFLUSH processing
   2609  * receive
   2610  *   <base>
   2611  * from the user, and free all entries in secpctree.
   2612  * and send,
   2613  *   <base>
   2614  * to the user.
   2615  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
   2616  *
   2617  * m will always be freed.
   2618  */
   2619 static int
   2620 key_api_spdflush(struct socket *so, struct mbuf *m,
   2621 	     const struct sadb_msghdr *mhp)
   2622 {
   2623 	struct sadb_msg *newmsg;
   2624 	struct secpolicy *sp;
   2625 	u_int dir;
   2626 
   2627 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
   2628 		return key_senderror(so, m, EINVAL);
   2629 
   2630 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
   2631 	    retry:
   2632 		mutex_enter(&key_spd.lock);
   2633 		SPLIST_WRITER_FOREACH(sp, dir) {
   2634 			KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
   2635 			key_unlink_sp(sp);
   2636 			mutex_exit(&key_spd.lock);
   2637 			key_destroy_sp(sp);
   2638 			goto retry;
   2639 		}
   2640 		mutex_exit(&key_spd.lock);
   2641 	}
   2642 
   2643 	/* We're deleting policy; no need to invalidate the ipflow cache. */
   2644 
   2645 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
   2646 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
   2647 		return key_senderror(so, m, ENOBUFS);
   2648 	}
   2649 
   2650 	if (m->m_next)
   2651 		m_freem(m->m_next);
   2652 	m->m_next = NULL;
   2653 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
   2654 	newmsg = mtod(m, struct sadb_msg *);
   2655 	newmsg->sadb_msg_errno = 0;
   2656 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
   2657 
   2658 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
   2659 }
   2660 
   2661 static struct sockaddr key_src = {
   2662 	.sa_len = 2,
   2663 	.sa_family = PF_KEY,
   2664 };
   2665 
   2666 static struct mbuf *
   2667 key_setspddump_chain(int *errorp, int *lenp, pid_t pid)
   2668 {
   2669 	struct secpolicy *sp;
   2670 	int cnt;
   2671 	u_int dir;
   2672 	struct mbuf *m, *n, *prev;
   2673 	int totlen;
   2674 
   2675 	KASSERT(mutex_owned(&key_spd.lock));
   2676 
   2677 	*lenp = 0;
   2678 
   2679 	/* search SPD entry and get buffer size. */
   2680 	cnt = 0;
   2681 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
   2682 		SPLIST_WRITER_FOREACH(sp, dir) {
   2683 			cnt++;
   2684 		}
   2685 	}
   2686 
   2687 	if (cnt == 0) {
   2688 		*errorp = ENOENT;
   2689 		return (NULL);
   2690 	}
   2691 
   2692 	m = NULL;
   2693 	prev = m;
   2694 	totlen = 0;
   2695 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
   2696 		SPLIST_WRITER_FOREACH(sp, dir) {
   2697 			--cnt;
   2698 			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, pid);
   2699 
   2700 			totlen += n->m_pkthdr.len;
   2701 			if (!m) {
   2702 				m = n;
   2703 			} else {
   2704 				prev->m_nextpkt = n;
   2705 			}
   2706 			prev = n;
   2707 		}
   2708 	}
   2709 
   2710 	*lenp = totlen;
   2711 	*errorp = 0;
   2712 	return (m);
   2713 }
   2714 
   2715 /*
   2716  * SADB_SPDDUMP processing
   2717  * receive
   2718  *   <base>
   2719  * from the user, and dump all SP leaves
   2720  * and send,
   2721  *   <base> .....
   2722  * to the ikmpd.
   2723  *
   2724  * m will always be freed.
   2725  */
   2726 static int
   2727 key_api_spddump(struct socket *so, struct mbuf *m0,
   2728  	    const struct sadb_msghdr *mhp)
   2729 {
   2730 	struct mbuf *n;
   2731 	int error, len;
   2732 	int ok;
   2733 	pid_t pid;
   2734 
   2735 	pid = mhp->msg->sadb_msg_pid;
   2736 	/*
   2737 	 * If the requestor has insufficient socket-buffer space
   2738 	 * for the entire chain, nobody gets any response to the DUMP.
   2739 	 * XXX For now, only the requestor ever gets anything.
   2740 	 * Moreover, if the requestor has any space at all, they receive
   2741 	 * the entire chain, otherwise the request is refused with  ENOBUFS.
   2742 	 */
   2743 	if (sbspace(&so->so_rcv) <= 0) {
   2744 		return key_senderror(so, m0, ENOBUFS);
   2745 	}
   2746 
   2747 	mutex_enter(&key_spd.lock);
   2748 	n = key_setspddump_chain(&error, &len, pid);
   2749 	mutex_exit(&key_spd.lock);
   2750 
   2751 	if (n == NULL) {
   2752 		return key_senderror(so, m0, ENOENT);
   2753 	}
   2754 	{
   2755 		uint64_t *ps = PFKEY_STAT_GETREF();
   2756 		ps[PFKEY_STAT_IN_TOTAL]++;
   2757 		ps[PFKEY_STAT_IN_BYTES] += len;
   2758 		PFKEY_STAT_PUTREF();
   2759 	}
   2760 
   2761 	/*
   2762 	 * PF_KEY DUMP responses are no longer broadcast to all PF_KEY sockets.
   2763 	 * The requestor receives either the entire chain, or an
   2764 	 * error message with ENOBUFS.
   2765 	 */
   2766 
   2767 	/*
   2768 	 * sbappendchainwith record takes the chain of entries, one
   2769 	 * packet-record per SPD entry, prepends the key_src sockaddr
   2770 	 * to each packet-record, links the sockaddr mbufs into a new
   2771 	 * list of records, then   appends the entire resulting
   2772 	 * list to the requesting socket.
   2773 	 */
   2774 	ok = sbappendaddrchain(&so->so_rcv, (struct sockaddr *)&key_src, n,
   2775 	    SB_PRIO_ONESHOT_OVERFLOW);
   2776 
   2777 	if (!ok) {
   2778 		PFKEY_STATINC(PFKEY_STAT_IN_NOMEM);
   2779 		m_freem(n);
   2780 		return key_senderror(so, m0, ENOBUFS);
   2781 	}
   2782 
   2783 	m_freem(m0);
   2784 	return error;
   2785 }
   2786 
   2787 /*
   2788  * SADB_X_NAT_T_NEW_MAPPING. Unused by racoon as of 2005/04/23
   2789  */
   2790 static int
   2791 key_api_nat_map(struct socket *so, struct mbuf *m,
   2792 	    const struct sadb_msghdr *mhp)
   2793 {
   2794 	struct sadb_x_nat_t_type *type;
   2795 	struct sadb_x_nat_t_port *sport;
   2796 	struct sadb_x_nat_t_port *dport;
   2797 	struct sadb_address *iaddr, *raddr;
   2798 	struct sadb_x_nat_t_frag *frag;
   2799 
   2800 	if (mhp->ext[SADB_X_EXT_NAT_T_TYPE] == NULL ||
   2801 	    mhp->ext[SADB_X_EXT_NAT_T_SPORT] == NULL ||
   2802 	    mhp->ext[SADB_X_EXT_NAT_T_DPORT] == NULL) {
   2803 		IPSECLOG(LOG_DEBUG, "invalid message.\n");
   2804 		return key_senderror(so, m, EINVAL);
   2805 	}
   2806 	if ((mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) ||
   2807 	    (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) ||
   2808 	    (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport))) {
   2809 		IPSECLOG(LOG_DEBUG, "invalid message.\n");
   2810 		return key_senderror(so, m, EINVAL);
   2811 	}
   2812 
   2813 	if ((mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL) &&
   2814 	    (mhp->extlen[SADB_X_EXT_NAT_T_OAI] < sizeof(*iaddr))) {
   2815 		IPSECLOG(LOG_DEBUG, "invalid message\n");
   2816 		return key_senderror(so, m, EINVAL);
   2817 	}
   2818 
   2819 	if ((mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL) &&
   2820 	    (mhp->extlen[SADB_X_EXT_NAT_T_OAR] < sizeof(*raddr))) {
   2821 		IPSECLOG(LOG_DEBUG, "invalid message\n");
   2822 		return key_senderror(so, m, EINVAL);
   2823 	}
   2824 
   2825 	if ((mhp->ext[SADB_X_EXT_NAT_T_FRAG] != NULL) &&
   2826 	    (mhp->extlen[SADB_X_EXT_NAT_T_FRAG] < sizeof(*frag))) {
   2827 		IPSECLOG(LOG_DEBUG, "invalid message\n");
   2828 		return key_senderror(so, m, EINVAL);
   2829 	}
   2830 
   2831 	type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
   2832 	sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
   2833 	dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
   2834 	iaddr = mhp->ext[SADB_X_EXT_NAT_T_OAI];
   2835 	raddr = mhp->ext[SADB_X_EXT_NAT_T_OAR];
   2836 	frag = mhp->ext[SADB_X_EXT_NAT_T_FRAG];
   2837 
   2838 	/*
   2839 	 * XXX handle that, it should also contain a SA, or anything
   2840 	 * that enable to update the SA information.
   2841 	 */
   2842 
   2843 	return 0;
   2844 }
   2845 
   2846 /*
   2847  * Never return NULL.
   2848  */
   2849 static struct mbuf *
   2850 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq, pid_t pid)
   2851 {
   2852 	struct mbuf *result = NULL, *m;
   2853 
   2854 	KASSERT(!cpu_softintr_p());
   2855 
   2856 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid,
   2857 	    key_sp_refcnt(sp), M_WAITOK);
   2858 	result = m;
   2859 
   2860 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
   2861 	    &sp->spidx.src.sa, sp->spidx.prefs, sp->spidx.ul_proto, M_WAITOK);
   2862 	m_cat(result, m);
   2863 
   2864 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
   2865 	    &sp->spidx.dst.sa, sp->spidx.prefd, sp->spidx.ul_proto, M_WAITOK);
   2866 	m_cat(result, m);
   2867 
   2868 	m = key_sp2msg(sp, M_WAITOK);
   2869 	m_cat(result, m);
   2870 
   2871 	KASSERT(result->m_flags & M_PKTHDR);
   2872 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
   2873 
   2874 	result->m_pkthdr.len = 0;
   2875 	for (m = result; m; m = m->m_next)
   2876 		result->m_pkthdr.len += m->m_len;
   2877 
   2878 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   2879 	    PFKEY_UNIT64(result->m_pkthdr.len);
   2880 
   2881 	return result;
   2882 }
   2883 
   2884 /*
   2885  * get PFKEY message length for security policy and request.
   2886  */
   2887 static u_int
   2888 key_getspreqmsglen(const struct secpolicy *sp)
   2889 {
   2890 	u_int tlen;
   2891 
   2892 	tlen = sizeof(struct sadb_x_policy);
   2893 
   2894 	/* if is the policy for ipsec ? */
   2895 	if (sp->policy != IPSEC_POLICY_IPSEC)
   2896 		return tlen;
   2897 
   2898 	/* get length of ipsec requests */
   2899     {
   2900 	const struct ipsecrequest *isr;
   2901 	int len;
   2902 
   2903 	for (isr = sp->req; isr != NULL; isr = isr->next) {
   2904 		len = sizeof(struct sadb_x_ipsecrequest)
   2905 		    + isr->saidx.src.sa.sa_len + isr->saidx.dst.sa.sa_len;
   2906 
   2907 		tlen += PFKEY_ALIGN8(len);
   2908 	}
   2909     }
   2910 
   2911 	return tlen;
   2912 }
   2913 
   2914 /*
   2915  * SADB_SPDEXPIRE processing
   2916  * send
   2917  *   <base, address(SD), lifetime(CH), policy>
   2918  * to KMD by PF_KEY.
   2919  *
   2920  * OUT:	0	: succeed
   2921  *	others	: error number
   2922  */
   2923 static int
   2924 key_spdexpire(struct secpolicy *sp)
   2925 {
   2926 	int s;
   2927 	struct mbuf *result = NULL, *m;
   2928 	int len;
   2929 	int error = -1;
   2930 	struct sadb_lifetime *lt;
   2931 
   2932 	/* XXX: Why do we lock ? */
   2933 	s = splsoftnet();	/*called from softclock()*/
   2934 
   2935 	KASSERT(sp != NULL);
   2936 
   2937 	/* set msg header */
   2938 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0, M_WAITOK);
   2939 	result = m;
   2940 
   2941 	/* create lifetime extension (current and hard) */
   2942 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
   2943 	m = key_alloc_mbuf(len, M_WAITOK);
   2944 	KASSERT(m->m_next == NULL);
   2945 
   2946 	memset(mtod(m, void *), 0, len);
   2947 	lt = mtod(m, struct sadb_lifetime *);
   2948 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
   2949 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
   2950 	lt->sadb_lifetime_allocations = 0;
   2951 	lt->sadb_lifetime_bytes = 0;
   2952 	lt->sadb_lifetime_addtime = time_mono_to_wall(sp->created);
   2953 	lt->sadb_lifetime_usetime = time_mono_to_wall(sp->lastused);
   2954 	lt = (struct sadb_lifetime *)(mtod(m, char *) + len / 2);
   2955 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
   2956 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
   2957 	lt->sadb_lifetime_allocations = 0;
   2958 	lt->sadb_lifetime_bytes = 0;
   2959 	lt->sadb_lifetime_addtime = sp->lifetime;
   2960 	lt->sadb_lifetime_usetime = sp->validtime;
   2961 	m_cat(result, m);
   2962 
   2963 	/* set sadb_address for source */
   2964 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sp->spidx.src.sa,
   2965 	    sp->spidx.prefs, sp->spidx.ul_proto, M_WAITOK);
   2966 	m_cat(result, m);
   2967 
   2968 	/* set sadb_address for destination */
   2969 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sp->spidx.dst.sa,
   2970 	    sp->spidx.prefd, sp->spidx.ul_proto, M_WAITOK);
   2971 	m_cat(result, m);
   2972 
   2973 	/* set secpolicy */
   2974 	m = key_sp2msg(sp, M_WAITOK);
   2975 	m_cat(result, m);
   2976 
   2977 	KASSERT(result->m_flags & M_PKTHDR);
   2978 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
   2979 
   2980 	result->m_pkthdr.len = 0;
   2981 	for (m = result; m; m = m->m_next)
   2982 		result->m_pkthdr.len += m->m_len;
   2983 
   2984 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   2985 	    PFKEY_UNIT64(result->m_pkthdr.len);
   2986 
   2987 	error = key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
   2988 	splx(s);
   2989 	return error;
   2990 }
   2991 
   2992 /* %%% SAD management */
   2993 /*
   2994  * allocating a memory for new SA head, and copy from the values of mhp.
   2995  * OUT:	NULL	: failure due to the lack of memory.
   2996  *	others	: pointer to new SA head.
   2997  */
   2998 static struct secashead *
   2999 key_newsah(const struct secasindex *saidx)
   3000 {
   3001 	struct secashead *newsah;
   3002 	int i;
   3003 
   3004 	KASSERT(saidx != NULL);
   3005 
   3006 	newsah = kmem_zalloc(sizeof(struct secashead), KM_SLEEP);
   3007 	for (i = 0; i < __arraycount(newsah->savlist); i++)
   3008 		PSLIST_INIT(&newsah->savlist[i]);
   3009 	newsah->saidx = *saidx;
   3010 
   3011 	localcount_init(&newsah->localcount);
   3012 	/* Take a reference for the caller */
   3013 	localcount_acquire(&newsah->localcount);
   3014 
   3015 	/* Add to the sah list */
   3016 	SAHLIST_ENTRY_INIT(newsah);
   3017 	newsah->state = SADB_SASTATE_MATURE;
   3018 	mutex_enter(&key_sad.lock);
   3019 	SAHLIST_WRITER_INSERT_HEAD(newsah);
   3020 	mutex_exit(&key_sad.lock);
   3021 
   3022 	return newsah;
   3023 }
   3024 
   3025 static bool
   3026 key_sah_has_sav(struct secashead *sah)
   3027 {
   3028 	u_int state;
   3029 
   3030 	KASSERT(mutex_owned(&key_sad.lock));
   3031 
   3032 	SASTATE_ANY_FOREACH(state) {
   3033 		if (!SAVLIST_WRITER_EMPTY(sah, state))
   3034 			return true;
   3035 	}
   3036 
   3037 	return false;
   3038 }
   3039 
   3040 static void
   3041 key_unlink_sah(struct secashead *sah)
   3042 {
   3043 
   3044 	KASSERT(!cpu_softintr_p());
   3045 	KASSERT(mutex_owned(&key_sad.lock));
   3046 	KASSERT(sah->state == SADB_SASTATE_DEAD);
   3047 
   3048 	/* Remove from the sah list */
   3049 	SAHLIST_WRITER_REMOVE(sah);
   3050 
   3051 #ifdef NET_MPSAFE
   3052 	KASSERT(mutex_ownable(softnet_lock));
   3053 	key_sad_pserialize_perform();
   3054 #endif
   3055 
   3056 	localcount_drain(&sah->localcount, &key_sad.cv_lc, &key_sad.lock);
   3057 }
   3058 
   3059 static void
   3060 key_destroy_sah(struct secashead *sah)
   3061 {
   3062 
   3063 	rtcache_free(&sah->sa_route);
   3064 
   3065 	SAHLIST_ENTRY_DESTROY(sah);
   3066 	localcount_fini(&sah->localcount);
   3067 
   3068 	if (sah->idents != NULL)
   3069 		kmem_free(sah->idents, sah->idents_len);
   3070 	if (sah->identd != NULL)
   3071 		kmem_free(sah->identd, sah->identd_len);
   3072 
   3073 	kmem_free(sah, sizeof(*sah));
   3074 }
   3075 
   3076 /*
   3077  * allocating a new SA with LARVAL state.
   3078  * key_api_add() and key_api_getspi() call,
   3079  * and copy the values of mhp into new buffer.
   3080  * When SAD message type is GETSPI:
   3081  *	to set sequence number from acq_seq++,
   3082  *	to set zero to SPI.
   3083  *	not to call key_setsava().
   3084  * OUT:	NULL	: fail
   3085  *	others	: pointer to new secasvar.
   3086  *
   3087  * does not modify mbuf.  does not free mbuf on error.
   3088  */
   3089 static struct secasvar *
   3090 key_newsav(struct mbuf *m, const struct sadb_msghdr *mhp,
   3091     int *errp, const char* where, int tag)
   3092 {
   3093 	struct secasvar *newsav;
   3094 	const struct sadb_sa *xsa;
   3095 
   3096 	KASSERT(!cpu_softintr_p());
   3097 	KASSERT(m != NULL);
   3098 	KASSERT(mhp != NULL);
   3099 	KASSERT(mhp->msg != NULL);
   3100 
   3101 	newsav = kmem_zalloc(sizeof(struct secasvar), KM_SLEEP);
   3102 
   3103 	switch (mhp->msg->sadb_msg_type) {
   3104 	case SADB_GETSPI:
   3105 		newsav->spi = 0;
   3106 
   3107 #ifdef IPSEC_DOSEQCHECK
   3108 		/* sync sequence number */
   3109 		if (mhp->msg->sadb_msg_seq == 0)
   3110 			newsav->seq =
   3111 			    (acq_seq = (acq_seq == ~0 ? 1 : ++acq_seq));
   3112 		else
   3113 #endif
   3114 			newsav->seq = mhp->msg->sadb_msg_seq;
   3115 		break;
   3116 
   3117 	case SADB_ADD:
   3118 		/* sanity check */
   3119 		if (mhp->ext[SADB_EXT_SA] == NULL) {
   3120 			IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   3121 			*errp = EINVAL;
   3122 			goto error;
   3123 		}
   3124 		xsa = mhp->ext[SADB_EXT_SA];
   3125 		newsav->spi = xsa->sadb_sa_spi;
   3126 		newsav->seq = mhp->msg->sadb_msg_seq;
   3127 		break;
   3128 	default:
   3129 		*errp = EINVAL;
   3130 		goto error;
   3131 	}
   3132 
   3133 	/* copy sav values */
   3134 	if (mhp->msg->sadb_msg_type != SADB_GETSPI) {
   3135 		*errp = key_setsaval(newsav, m, mhp);
   3136 		if (*errp)
   3137 			goto error;
   3138 	} else {
   3139 		/* We don't allow lft_c to be NULL */
   3140 		newsav->lft_c = kmem_zalloc(sizeof(struct sadb_lifetime),
   3141 		    KM_SLEEP);
   3142 	}
   3143 
   3144 	/* reset created */
   3145 	newsav->created = time_uptime;
   3146 	newsav->pid = mhp->msg->sadb_msg_pid;
   3147 
   3148 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   3149 	    "DP from %s:%u return SA:%p\n", where, tag, newsav);
   3150 	return newsav;
   3151 
   3152 error:
   3153 	KASSERT(*errp != 0);
   3154 	kmem_free(newsav, sizeof(*newsav));
   3155 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   3156 	    "DP from %s:%u return SA:NULL\n", where, tag);
   3157 	return NULL;
   3158 }
   3159 
   3160 
   3161 static void
   3162 key_clear_xform(struct secasvar *sav)
   3163 {
   3164 
   3165 	/*
   3166 	 * Cleanup xform state.  Note that zeroize'ing causes the
   3167 	 * keys to be cleared; otherwise we must do it ourself.
   3168 	 */
   3169 	if (sav->tdb_xform != NULL) {
   3170 		sav->tdb_xform->xf_zeroize(sav);
   3171 		sav->tdb_xform = NULL;
   3172 	} else {
   3173 		if (sav->key_auth != NULL)
   3174 			explicit_memset(_KEYBUF(sav->key_auth), 0,
   3175 			    _KEYLEN(sav->key_auth));
   3176 		if (sav->key_enc != NULL)
   3177 			explicit_memset(_KEYBUF(sav->key_enc), 0,
   3178 			    _KEYLEN(sav->key_enc));
   3179 	}
   3180 }
   3181 
   3182 /*
   3183  * free() SA variable entry.
   3184  */
   3185 static void
   3186 key_delsav(struct secasvar *sav)
   3187 {
   3188 
   3189 	key_clear_xform(sav);
   3190 	key_freesaval(sav);
   3191 	kmem_free(sav, sizeof(*sav));
   3192 }
   3193 
   3194 /*
   3195  * Must be called in a pserialize read section. A held sah
   3196  * must be released by key_sah_unref after use.
   3197  */
   3198 static void
   3199 key_sah_ref(struct secashead *sah)
   3200 {
   3201 
   3202 	localcount_acquire(&sah->localcount);
   3203 }
   3204 
   3205 /*
   3206  * Must be called without holding key_sad.lock because the lock
   3207  * would be held in localcount_release.
   3208  */
   3209 static void
   3210 key_sah_unref(struct secashead *sah)
   3211 {
   3212 
   3213 	KDASSERT(mutex_ownable(&key_sad.lock));
   3214 
   3215 	localcount_release(&sah->localcount, &key_sad.cv_lc, &key_sad.lock);
   3216 }
   3217 
   3218 /*
   3219  * Search SAD and return sah. Must be called in a pserialize
   3220  * read section.
   3221  * OUT:
   3222  *	NULL	: not found
   3223  *	others	: found, pointer to a SA.
   3224  */
   3225 static struct secashead *
   3226 key_getsah(const struct secasindex *saidx, int flag)
   3227 {
   3228 	struct secashead *sah;
   3229 
   3230 	SAHLIST_READER_FOREACH(sah) {
   3231 		if (sah->state == SADB_SASTATE_DEAD)
   3232 			continue;
   3233 		if (key_saidx_match(&sah->saidx, saidx, flag))
   3234 			return sah;
   3235 	}
   3236 
   3237 	return NULL;
   3238 }
   3239 
   3240 /*
   3241  * Search SAD and return sah. If sah is returned, the caller must call
   3242  * key_sah_unref to releaset a reference.
   3243  * OUT:
   3244  *	NULL	: not found
   3245  *	others	: found, pointer to a SA.
   3246  */
   3247 static struct secashead *
   3248 key_getsah_ref(const struct secasindex *saidx, int flag)
   3249 {
   3250 	struct secashead *sah;
   3251 	int s;
   3252 
   3253 	s = pserialize_read_enter();
   3254 	sah = key_getsah(saidx, flag);
   3255 	if (sah != NULL)
   3256 		key_sah_ref(sah);
   3257 	pserialize_read_exit(s);
   3258 
   3259 	return sah;
   3260 }
   3261 
   3262 /*
   3263  * check not to be duplicated SPI.
   3264  * NOTE: this function is too slow due to searching all SAD.
   3265  * OUT:
   3266  *	NULL	: not found
   3267  *	others	: found, pointer to a SA.
   3268  */
   3269 static bool
   3270 key_checkspidup(const struct secasindex *saidx, u_int32_t spi)
   3271 {
   3272 	struct secashead *sah;
   3273 	struct secasvar *sav;
   3274 	int s;
   3275 
   3276 	/* check address family */
   3277 	if (saidx->src.sa.sa_family != saidx->dst.sa.sa_family) {
   3278 		IPSECLOG(LOG_DEBUG, "address family mismatched.\n");
   3279 		return false;
   3280 	}
   3281 
   3282 	/* check all SAD */
   3283 	s = pserialize_read_enter();
   3284 	SAHLIST_READER_FOREACH(sah) {
   3285 		if (!key_ismyaddr((struct sockaddr *)&sah->saidx.dst))
   3286 			continue;
   3287 		sav = key_getsavbyspi(sah, spi);
   3288 		if (sav != NULL) {
   3289 			pserialize_read_exit(s);
   3290 			KEY_SA_UNREF(&sav);
   3291 			return true;
   3292 		}
   3293 	}
   3294 	pserialize_read_exit(s);
   3295 
   3296 	return false;
   3297 }
   3298 
   3299 /*
   3300  * search SAD litmited alive SA, protocol, SPI.
   3301  * OUT:
   3302  *	NULL	: not found
   3303  *	others	: found, pointer to a SA.
   3304  */
   3305 static struct secasvar *
   3306 key_getsavbyspi(struct secashead *sah, u_int32_t spi)
   3307 {
   3308 	struct secasvar *sav = NULL;
   3309 	u_int state;
   3310 	int s;
   3311 
   3312 	/* search all status */
   3313 	s = pserialize_read_enter();
   3314 	SASTATE_ALIVE_FOREACH(state) {
   3315 		SAVLIST_READER_FOREACH(sav, sah, state) {
   3316 			/* sanity check */
   3317 			if (sav->state != state) {
   3318 				IPSECLOG(LOG_DEBUG,
   3319 				    "invalid sav->state (queue: %d SA: %d)\n",
   3320 				    state, sav->state);
   3321 				continue;
   3322 			}
   3323 
   3324 			if (sav->spi == spi) {
   3325 				KEY_SA_REF(sav);
   3326 				goto out;
   3327 			}
   3328 		}
   3329 	}
   3330 out:
   3331 	pserialize_read_exit(s);
   3332 
   3333 	return sav;
   3334 }
   3335 
   3336 /*
   3337  * Free allocated data to member variables of sav:
   3338  * sav->replay, sav->key_* and sav->lft_*.
   3339  */
   3340 static void
   3341 key_freesaval(struct secasvar *sav)
   3342 {
   3343 
   3344 	KASSERT(key_sa_refcnt(sav) == 0);
   3345 
   3346 	if (sav->replay != NULL)
   3347 		kmem_intr_free(sav->replay, sav->replay_len);
   3348 	if (sav->key_auth != NULL)
   3349 		kmem_intr_free(sav->key_auth, sav->key_auth_len);
   3350 	if (sav->key_enc != NULL)
   3351 		kmem_intr_free(sav->key_enc, sav->key_enc_len);
   3352 	if (sav->lft_c != NULL)
   3353 		kmem_intr_free(sav->lft_c, sizeof(*(sav->lft_c)));
   3354 	if (sav->lft_h != NULL)
   3355 		kmem_intr_free(sav->lft_h, sizeof(*(sav->lft_h)));
   3356 	if (sav->lft_s != NULL)
   3357 		kmem_intr_free(sav->lft_s, sizeof(*(sav->lft_s)));
   3358 }
   3359 
   3360 /*
   3361  * copy SA values from PF_KEY message except *SPI, SEQ, PID, STATE and TYPE*.
   3362  * You must update these if need.
   3363  * OUT:	0:	success.
   3364  *	!0:	failure.
   3365  *
   3366  * does not modify mbuf.  does not free mbuf on error.
   3367  */
   3368 static int
   3369 key_setsaval(struct secasvar *sav, struct mbuf *m,
   3370 	     const struct sadb_msghdr *mhp)
   3371 {
   3372 	int error = 0;
   3373 
   3374 	KASSERT(!cpu_softintr_p());
   3375 	KASSERT(m != NULL);
   3376 	KASSERT(mhp != NULL);
   3377 	KASSERT(mhp->msg != NULL);
   3378 
   3379 	/* We shouldn't initialize sav variables while someone uses it. */
   3380 	KASSERT(key_sa_refcnt(sav) == 0);
   3381 
   3382 	/* SA */
   3383 	if (mhp->ext[SADB_EXT_SA] != NULL) {
   3384 		const struct sadb_sa *sa0;
   3385 
   3386 		sa0 = mhp->ext[SADB_EXT_SA];
   3387 		if (mhp->extlen[SADB_EXT_SA] < sizeof(*sa0)) {
   3388 			error = EINVAL;
   3389 			goto fail;
   3390 		}
   3391 
   3392 		sav->alg_auth = sa0->sadb_sa_auth;
   3393 		sav->alg_enc = sa0->sadb_sa_encrypt;
   3394 		sav->flags = sa0->sadb_sa_flags;
   3395 
   3396 		/* replay window */
   3397 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0) {
   3398 			size_t len = sizeof(struct secreplay) +
   3399 			    sa0->sadb_sa_replay;
   3400 			sav->replay = kmem_zalloc(len, KM_SLEEP);
   3401 			sav->replay_len = len;
   3402 			if (sa0->sadb_sa_replay != 0)
   3403 				sav->replay->bitmap = (char*)(sav->replay+1);
   3404 			sav->replay->wsize = sa0->sadb_sa_replay;
   3405 		}
   3406 	}
   3407 
   3408 	/* Authentication keys */
   3409 	if (mhp->ext[SADB_EXT_KEY_AUTH] != NULL) {
   3410 		const struct sadb_key *key0;
   3411 		int len;
   3412 
   3413 		key0 = mhp->ext[SADB_EXT_KEY_AUTH];
   3414 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
   3415 
   3416 		error = 0;
   3417 		if (len < sizeof(*key0)) {
   3418 			error = EINVAL;
   3419 			goto fail;
   3420 		}
   3421 		switch (mhp->msg->sadb_msg_satype) {
   3422 		case SADB_SATYPE_AH:
   3423 		case SADB_SATYPE_ESP:
   3424 		case SADB_X_SATYPE_TCPSIGNATURE:
   3425 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
   3426 			    sav->alg_auth != SADB_X_AALG_NULL)
   3427 				error = EINVAL;
   3428 			break;
   3429 		case SADB_X_SATYPE_IPCOMP:
   3430 		default:
   3431 			error = EINVAL;
   3432 			break;
   3433 		}
   3434 		if (error) {
   3435 			IPSECLOG(LOG_DEBUG, "invalid key_auth values.\n");
   3436 			goto fail;
   3437 		}
   3438 
   3439 		sav->key_auth = key_newbuf(key0, len);
   3440 		sav->key_auth_len = len;
   3441 	}
   3442 
   3443 	/* Encryption key */
   3444 	if (mhp->ext[SADB_EXT_KEY_ENCRYPT] != NULL) {
   3445 		const struct sadb_key *key0;
   3446 		int len;
   3447 
   3448 		key0 = mhp->ext[SADB_EXT_KEY_ENCRYPT];
   3449 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
   3450 
   3451 		error = 0;
   3452 		if (len < sizeof(*key0)) {
   3453 			error = EINVAL;
   3454 			goto fail;
   3455 		}
   3456 		switch (mhp->msg->sadb_msg_satype) {
   3457 		case SADB_SATYPE_ESP:
   3458 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
   3459 			    sav->alg_enc != SADB_EALG_NULL) {
   3460 				error = EINVAL;
   3461 				break;
   3462 			}
   3463 			sav->key_enc = key_newbuf(key0, len);
   3464 			sav->key_enc_len = len;
   3465 			break;
   3466 		case SADB_X_SATYPE_IPCOMP:
   3467 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
   3468 				error = EINVAL;
   3469 			sav->key_enc = NULL;	/*just in case*/
   3470 			break;
   3471 		case SADB_SATYPE_AH:
   3472 		case SADB_X_SATYPE_TCPSIGNATURE:
   3473 		default:
   3474 			error = EINVAL;
   3475 			break;
   3476 		}
   3477 		if (error) {
   3478 			IPSECLOG(LOG_DEBUG, "invalid key_enc value.\n");
   3479 			goto fail;
   3480 		}
   3481 	}
   3482 
   3483 	/* set iv */
   3484 	sav->ivlen = 0;
   3485 
   3486 	switch (mhp->msg->sadb_msg_satype) {
   3487 	case SADB_SATYPE_AH:
   3488 		error = xform_init(sav, XF_AH);
   3489 		break;
   3490 	case SADB_SATYPE_ESP:
   3491 		error = xform_init(sav, XF_ESP);
   3492 		break;
   3493 	case SADB_X_SATYPE_IPCOMP:
   3494 		error = xform_init(sav, XF_IPCOMP);
   3495 		break;
   3496 	case SADB_X_SATYPE_TCPSIGNATURE:
   3497 		error = xform_init(sav, XF_TCPSIGNATURE);
   3498 		break;
   3499 	}
   3500 	if (error) {
   3501 		IPSECLOG(LOG_DEBUG, "unable to initialize SA type %u.\n",
   3502 		    mhp->msg->sadb_msg_satype);
   3503 		goto fail;
   3504 	}
   3505 
   3506 	/* reset created */
   3507 	sav->created = time_uptime;
   3508 
   3509 	/* make lifetime for CURRENT */
   3510 	sav->lft_c = kmem_alloc(sizeof(struct sadb_lifetime), KM_SLEEP);
   3511 
   3512 	sav->lft_c->sadb_lifetime_len =
   3513 	    PFKEY_UNIT64(sizeof(struct sadb_lifetime));
   3514 	sav->lft_c->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
   3515 	sav->lft_c->sadb_lifetime_allocations = 0;
   3516 	sav->lft_c->sadb_lifetime_bytes = 0;
   3517 	sav->lft_c->sadb_lifetime_addtime = time_uptime;
   3518 	sav->lft_c->sadb_lifetime_usetime = 0;
   3519 
   3520 	/* lifetimes for HARD and SOFT */
   3521     {
   3522 	const struct sadb_lifetime *lft0;
   3523 
   3524 	lft0 = mhp->ext[SADB_EXT_LIFETIME_HARD];
   3525 	if (lft0 != NULL) {
   3526 		if (mhp->extlen[SADB_EXT_LIFETIME_HARD] < sizeof(*lft0)) {
   3527 			error = EINVAL;
   3528 			goto fail;
   3529 		}
   3530 		sav->lft_h = key_newbuf(lft0, sizeof(*lft0));
   3531 	}
   3532 
   3533 	lft0 = mhp->ext[SADB_EXT_LIFETIME_SOFT];
   3534 	if (lft0 != NULL) {
   3535 		if (mhp->extlen[SADB_EXT_LIFETIME_SOFT] < sizeof(*lft0)) {
   3536 			error = EINVAL;
   3537 			goto fail;
   3538 		}
   3539 		sav->lft_s = key_newbuf(lft0, sizeof(*lft0));
   3540 		/* to be initialize ? */
   3541 	}
   3542     }
   3543 
   3544 	return 0;
   3545 
   3546  fail:
   3547 	key_clear_xform(sav);
   3548 	key_freesaval(sav);
   3549 
   3550 	return error;
   3551 }
   3552 
   3553 /*
   3554  * validation with a secasvar entry, and set SADB_SATYPE_MATURE.
   3555  * OUT:	0:	valid
   3556  *	other:	errno
   3557  */
   3558 static int
   3559 key_init_xform(struct secasvar *sav)
   3560 {
   3561 	int error;
   3562 
   3563 	/* We shouldn't initialize sav variables while someone uses it. */
   3564 	KASSERT(key_sa_refcnt(sav) == 0);
   3565 
   3566 	/* check SPI value */
   3567 	switch (sav->sah->saidx.proto) {
   3568 	case IPPROTO_ESP:
   3569 	case IPPROTO_AH:
   3570 		if (ntohl(sav->spi) <= 255) {
   3571 			IPSECLOG(LOG_DEBUG, "illegal range of SPI %u.\n",
   3572 			    (u_int32_t)ntohl(sav->spi));
   3573 			return EINVAL;
   3574 		}
   3575 		break;
   3576 	}
   3577 
   3578 	/* check satype */
   3579 	switch (sav->sah->saidx.proto) {
   3580 	case IPPROTO_ESP:
   3581 		/* check flags */
   3582 		if ((sav->flags & (SADB_X_EXT_OLD|SADB_X_EXT_DERIV)) ==
   3583 		    (SADB_X_EXT_OLD|SADB_X_EXT_DERIV)) {
   3584 			IPSECLOG(LOG_DEBUG,
   3585 			    "invalid flag (derived) given to old-esp.\n");
   3586 			return EINVAL;
   3587 		}
   3588 		error = xform_init(sav, XF_ESP);
   3589 		break;
   3590 	case IPPROTO_AH:
   3591 		/* check flags */
   3592 		if (sav->flags & SADB_X_EXT_DERIV) {
   3593 			IPSECLOG(LOG_DEBUG,
   3594 			    "invalid flag (derived) given to AH SA.\n");
   3595 			return EINVAL;
   3596 		}
   3597 		if (sav->alg_enc != SADB_EALG_NONE) {
   3598 			IPSECLOG(LOG_DEBUG,
   3599 			    "protocol and algorithm mismated.\n");
   3600 			return(EINVAL);
   3601 		}
   3602 		error = xform_init(sav, XF_AH);
   3603 		break;
   3604 	case IPPROTO_IPCOMP:
   3605 		if (sav->alg_auth != SADB_AALG_NONE) {
   3606 			IPSECLOG(LOG_DEBUG,
   3607 			    "protocol and algorithm mismated.\n");
   3608 			return(EINVAL);
   3609 		}
   3610 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0
   3611 		 && ntohl(sav->spi) >= 0x10000) {
   3612 			IPSECLOG(LOG_DEBUG, "invalid cpi for IPComp.\n");
   3613 			return(EINVAL);
   3614 		}
   3615 		error = xform_init(sav, XF_IPCOMP);
   3616 		break;
   3617 	case IPPROTO_TCP:
   3618 		if (sav->alg_enc != SADB_EALG_NONE) {
   3619 			IPSECLOG(LOG_DEBUG,
   3620 			    "protocol and algorithm mismated.\n");
   3621 			return(EINVAL);
   3622 		}
   3623 		error = xform_init(sav, XF_TCPSIGNATURE);
   3624 		break;
   3625 	default:
   3626 		IPSECLOG(LOG_DEBUG, "Invalid satype.\n");
   3627 		error = EPROTONOSUPPORT;
   3628 		break;
   3629 	}
   3630 
   3631 	return error;
   3632 }
   3633 
   3634 /*
   3635  * subroutine for SADB_GET and SADB_DUMP. It never return NULL.
   3636  */
   3637 static struct mbuf *
   3638 key_setdumpsa(struct secasvar *sav, u_int8_t type, u_int8_t satype,
   3639 	      u_int32_t seq, u_int32_t pid)
   3640 {
   3641 	struct mbuf *result = NULL, *tres = NULL, *m;
   3642 	int l = 0;
   3643 	int i;
   3644 	void *p;
   3645 	struct sadb_lifetime lt;
   3646 	int dumporder[] = {
   3647 		SADB_EXT_SA, SADB_X_EXT_SA2,
   3648 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
   3649 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
   3650 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY, SADB_EXT_KEY_AUTH,
   3651 		SADB_EXT_KEY_ENCRYPT, SADB_EXT_IDENTITY_SRC,
   3652 		SADB_EXT_IDENTITY_DST, SADB_EXT_SENSITIVITY,
   3653 		SADB_X_EXT_NAT_T_TYPE,
   3654 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
   3655 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
   3656 		SADB_X_EXT_NAT_T_FRAG,
   3657 
   3658 	};
   3659 
   3660 	m = key_setsadbmsg(type, 0, satype, seq, pid, key_sa_refcnt(sav), M_WAITOK);
   3661 	result = m;
   3662 
   3663 	for (i = __arraycount(dumporder) - 1; i >= 0; i--) {
   3664 		m = NULL;
   3665 		p = NULL;
   3666 		switch (dumporder[i]) {
   3667 		case SADB_EXT_SA:
   3668 			m = key_setsadbsa(sav);
   3669 			break;
   3670 
   3671 		case SADB_X_EXT_SA2:
   3672 			m = key_setsadbxsa2(sav->sah->saidx.mode,
   3673 			    sav->replay ? sav->replay->count : 0,
   3674 			    sav->sah->saidx.reqid);
   3675 			break;
   3676 
   3677 		case SADB_EXT_ADDRESS_SRC:
   3678 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
   3679 			    &sav->sah->saidx.src.sa,
   3680 			    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
   3681 			break;
   3682 
   3683 		case SADB_EXT_ADDRESS_DST:
   3684 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
   3685 			    &sav->sah->saidx.dst.sa,
   3686 			    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
   3687 			break;
   3688 
   3689 		case SADB_EXT_KEY_AUTH:
   3690 			if (!sav->key_auth)
   3691 				continue;
   3692 			l = PFKEY_UNUNIT64(sav->key_auth->sadb_key_len);
   3693 			p = sav->key_auth;
   3694 			break;
   3695 
   3696 		case SADB_EXT_KEY_ENCRYPT:
   3697 			if (!sav->key_enc)
   3698 				continue;
   3699 			l = PFKEY_UNUNIT64(sav->key_enc->sadb_key_len);
   3700 			p = sav->key_enc;
   3701 			break;
   3702 
   3703 		case SADB_EXT_LIFETIME_CURRENT:
   3704 			KASSERT(sav->lft_c != NULL);
   3705 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_c)->sadb_ext_len);
   3706 			memcpy(&lt, sav->lft_c, sizeof(struct sadb_lifetime));
   3707 			lt.sadb_lifetime_addtime =
   3708 			    time_mono_to_wall(lt.sadb_lifetime_addtime);
   3709 			lt.sadb_lifetime_usetime =
   3710 			    time_mono_to_wall(lt.sadb_lifetime_usetime);
   3711 			p = &lt;
   3712 			break;
   3713 
   3714 		case SADB_EXT_LIFETIME_HARD:
   3715 			if (!sav->lft_h)
   3716 				continue;
   3717 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_h)->sadb_ext_len);
   3718 			p = sav->lft_h;
   3719 			break;
   3720 
   3721 		case SADB_EXT_LIFETIME_SOFT:
   3722 			if (!sav->lft_s)
   3723 				continue;
   3724 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_s)->sadb_ext_len);
   3725 			p = sav->lft_s;
   3726 			break;
   3727 
   3728 		case SADB_X_EXT_NAT_T_TYPE:
   3729 			m = key_setsadbxtype(sav->natt_type);
   3730 			break;
   3731 
   3732 		case SADB_X_EXT_NAT_T_DPORT:
   3733 			if (sav->natt_type == 0)
   3734 				continue;
   3735 			m = key_setsadbxport(
   3736 			    key_portfromsaddr(&sav->sah->saidx.dst),
   3737 			    SADB_X_EXT_NAT_T_DPORT);
   3738 			break;
   3739 
   3740 		case SADB_X_EXT_NAT_T_SPORT:
   3741 			if (sav->natt_type == 0)
   3742 				continue;
   3743 			m = key_setsadbxport(
   3744 			    key_portfromsaddr(&sav->sah->saidx.src),
   3745 			    SADB_X_EXT_NAT_T_SPORT);
   3746 			break;
   3747 
   3748 		case SADB_X_EXT_NAT_T_FRAG:
   3749 			/* don't send frag info if not set */
   3750 			if (sav->natt_type == 0 || sav->esp_frag == IP_MAXPACKET)
   3751 				continue;
   3752 			m = key_setsadbxfrag(sav->esp_frag);
   3753 			break;
   3754 
   3755 		case SADB_X_EXT_NAT_T_OAI:
   3756 		case SADB_X_EXT_NAT_T_OAR:
   3757 			continue;
   3758 
   3759 		case SADB_EXT_ADDRESS_PROXY:
   3760 		case SADB_EXT_IDENTITY_SRC:
   3761 		case SADB_EXT_IDENTITY_DST:
   3762 			/* XXX: should we brought from SPD ? */
   3763 		case SADB_EXT_SENSITIVITY:
   3764 		default:
   3765 			continue;
   3766 		}
   3767 
   3768 		KASSERT(!(m && p));
   3769 		KASSERT(m != NULL || p != NULL);
   3770 		if (p && tres) {
   3771 			M_PREPEND(tres, l, M_WAITOK);
   3772 			memcpy(mtod(tres, void *), p, l);
   3773 			continue;
   3774 		}
   3775 		if (p) {
   3776 			m = key_alloc_mbuf(l, M_WAITOK);
   3777 			m_copyback(m, 0, l, p);
   3778 		}
   3779 
   3780 		if (tres)
   3781 			m_cat(m, tres);
   3782 		tres = m;
   3783 	}
   3784 
   3785 	m_cat(result, tres);
   3786 	tres = NULL; /* avoid free on error below */
   3787 
   3788 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
   3789 
   3790 	result->m_pkthdr.len = 0;
   3791 	for (m = result; m; m = m->m_next)
   3792 		result->m_pkthdr.len += m->m_len;
   3793 
   3794 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   3795 	    PFKEY_UNIT64(result->m_pkthdr.len);
   3796 
   3797 	return result;
   3798 }
   3799 
   3800 
   3801 /*
   3802  * set a type in sadb_x_nat_t_type
   3803  */
   3804 static struct mbuf *
   3805 key_setsadbxtype(u_int16_t type)
   3806 {
   3807 	struct mbuf *m;
   3808 	size_t len;
   3809 	struct sadb_x_nat_t_type *p;
   3810 
   3811 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
   3812 
   3813 	m = key_alloc_mbuf(len, M_WAITOK);
   3814 	KASSERT(m->m_next == NULL);
   3815 
   3816 	p = mtod(m, struct sadb_x_nat_t_type *);
   3817 
   3818 	memset(p, 0, len);
   3819 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
   3820 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
   3821 	p->sadb_x_nat_t_type_type = type;
   3822 
   3823 	return m;
   3824 }
   3825 /*
   3826  * set a port in sadb_x_nat_t_port. port is in network order
   3827  */
   3828 static struct mbuf *
   3829 key_setsadbxport(u_int16_t port, u_int16_t type)
   3830 {
   3831 	struct mbuf *m;
   3832 	size_t len;
   3833 	struct sadb_x_nat_t_port *p;
   3834 
   3835 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
   3836 
   3837 	m = key_alloc_mbuf(len, M_WAITOK);
   3838 	KASSERT(m->m_next == NULL);
   3839 
   3840 	p = mtod(m, struct sadb_x_nat_t_port *);
   3841 
   3842 	memset(p, 0, len);
   3843 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
   3844 	p->sadb_x_nat_t_port_exttype = type;
   3845 	p->sadb_x_nat_t_port_port = port;
   3846 
   3847 	return m;
   3848 }
   3849 
   3850 /*
   3851  * set fragmentation info in sadb_x_nat_t_frag
   3852  */
   3853 static struct mbuf *
   3854 key_setsadbxfrag(u_int16_t flen)
   3855 {
   3856 	struct mbuf *m;
   3857 	size_t len;
   3858 	struct sadb_x_nat_t_frag *p;
   3859 
   3860 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_frag));
   3861 
   3862 	m = key_alloc_mbuf(len, M_WAITOK);
   3863 	KASSERT(m->m_next == NULL);
   3864 
   3865 	p = mtod(m, struct sadb_x_nat_t_frag *);
   3866 
   3867 	memset(p, 0, len);
   3868 	p->sadb_x_nat_t_frag_len = PFKEY_UNIT64(len);
   3869 	p->sadb_x_nat_t_frag_exttype = SADB_X_EXT_NAT_T_FRAG;
   3870 	p->sadb_x_nat_t_frag_fraglen = flen;
   3871 
   3872 	return m;
   3873 }
   3874 
   3875 /*
   3876  * Get port from sockaddr, port is in network order
   3877  */
   3878 u_int16_t
   3879 key_portfromsaddr(const union sockaddr_union *saddr)
   3880 {
   3881 	u_int16_t port;
   3882 
   3883 	switch (saddr->sa.sa_family) {
   3884 	case AF_INET: {
   3885 		port = saddr->sin.sin_port;
   3886 		break;
   3887 	}
   3888 #ifdef INET6
   3889 	case AF_INET6: {
   3890 		port = saddr->sin6.sin6_port;
   3891 		break;
   3892 	}
   3893 #endif
   3894 	default:
   3895 		printf("%s: unexpected address family\n", __func__);
   3896 		port = 0;
   3897 		break;
   3898 	}
   3899 
   3900 	return port;
   3901 }
   3902 
   3903 
   3904 /*
   3905  * Set port is struct sockaddr. port is in network order
   3906  */
   3907 static void
   3908 key_porttosaddr(union sockaddr_union *saddr, u_int16_t port)
   3909 {
   3910 	switch (saddr->sa.sa_family) {
   3911 	case AF_INET: {
   3912 		saddr->sin.sin_port = port;
   3913 		break;
   3914 	}
   3915 #ifdef INET6
   3916 	case AF_INET6: {
   3917 		saddr->sin6.sin6_port = port;
   3918 		break;
   3919 	}
   3920 #endif
   3921 	default:
   3922 		printf("%s: unexpected address family %d\n", __func__,
   3923 		    saddr->sa.sa_family);
   3924 		break;
   3925 	}
   3926 
   3927 	return;
   3928 }
   3929 
   3930 /*
   3931  * Safety check sa_len
   3932  */
   3933 static int
   3934 key_checksalen(const union sockaddr_union *saddr)
   3935 {
   3936 	switch (saddr->sa.sa_family) {
   3937 	case AF_INET:
   3938 		if (saddr->sa.sa_len != sizeof(struct sockaddr_in))
   3939 			return -1;
   3940 		break;
   3941 #ifdef INET6
   3942 	case AF_INET6:
   3943 		if (saddr->sa.sa_len != sizeof(struct sockaddr_in6))
   3944 			return -1;
   3945 		break;
   3946 #endif
   3947 	default:
   3948 		printf("%s: unexpected sa_family %d\n", __func__,
   3949 		    saddr->sa.sa_family);
   3950 			return -1;
   3951 		break;
   3952 	}
   3953 	return 0;
   3954 }
   3955 
   3956 
   3957 /*
   3958  * set data into sadb_msg.
   3959  */
   3960 static struct mbuf *
   3961 key_setsadbmsg(u_int8_t type,  u_int16_t tlen, u_int8_t satype,
   3962 	       u_int32_t seq, pid_t pid, u_int16_t reserved, int mflag)
   3963 {
   3964 	struct mbuf *m;
   3965 	struct sadb_msg *p;
   3966 	int len;
   3967 
   3968 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MCLBYTES);
   3969 
   3970 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
   3971 
   3972 	m = key_alloc_mbuf_simple(len, mflag);
   3973 	if (!m)
   3974 		return NULL;
   3975 	m->m_pkthdr.len = m->m_len = len;
   3976 	m->m_next = NULL;
   3977 
   3978 	p = mtod(m, struct sadb_msg *);
   3979 
   3980 	memset(p, 0, len);
   3981 	p->sadb_msg_version = PF_KEY_V2;
   3982 	p->sadb_msg_type = type;
   3983 	p->sadb_msg_errno = 0;
   3984 	p->sadb_msg_satype = satype;
   3985 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
   3986 	p->sadb_msg_reserved = reserved;
   3987 	p->sadb_msg_seq = seq;
   3988 	p->sadb_msg_pid = (u_int32_t)pid;
   3989 
   3990 	return m;
   3991 }
   3992 
   3993 /*
   3994  * copy secasvar data into sadb_address.
   3995  */
   3996 static struct mbuf *
   3997 key_setsadbsa(struct secasvar *sav)
   3998 {
   3999 	struct mbuf *m;
   4000 	struct sadb_sa *p;
   4001 	int len;
   4002 
   4003 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
   4004 	m = key_alloc_mbuf(len, M_WAITOK);
   4005 	KASSERT(m->m_next == NULL);
   4006 
   4007 	p = mtod(m, struct sadb_sa *);
   4008 
   4009 	memset(p, 0, len);
   4010 	p->sadb_sa_len = PFKEY_UNIT64(len);
   4011 	p->sadb_sa_exttype = SADB_EXT_SA;
   4012 	p->sadb_sa_spi = sav->spi;
   4013 	p->sadb_sa_replay = (sav->replay != NULL ? sav->replay->wsize : 0);
   4014 	p->sadb_sa_state = sav->state;
   4015 	p->sadb_sa_auth = sav->alg_auth;
   4016 	p->sadb_sa_encrypt = sav->alg_enc;
   4017 	p->sadb_sa_flags = sav->flags;
   4018 
   4019 	return m;
   4020 }
   4021 
   4022 /*
   4023  * set data into sadb_address.
   4024  */
   4025 static struct mbuf *
   4026 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
   4027 		u_int8_t prefixlen, u_int16_t ul_proto, int mflag)
   4028 {
   4029 	struct mbuf *m;
   4030 	struct sadb_address *p;
   4031 	size_t len;
   4032 
   4033 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
   4034 	    PFKEY_ALIGN8(saddr->sa_len);
   4035 	m = key_alloc_mbuf(len, mflag);
   4036 	if (!m || m->m_next) {	/*XXX*/
   4037 		if (m)
   4038 			m_freem(m);
   4039 		return NULL;
   4040 	}
   4041 
   4042 	p = mtod(m, struct sadb_address *);
   4043 
   4044 	memset(p, 0, len);
   4045 	p->sadb_address_len = PFKEY_UNIT64(len);
   4046 	p->sadb_address_exttype = exttype;
   4047 	p->sadb_address_proto = ul_proto;
   4048 	if (prefixlen == FULLMASK) {
   4049 		switch (saddr->sa_family) {
   4050 		case AF_INET:
   4051 			prefixlen = sizeof(struct in_addr) << 3;
   4052 			break;
   4053 		case AF_INET6:
   4054 			prefixlen = sizeof(struct in6_addr) << 3;
   4055 			break;
   4056 		default:
   4057 			; /*XXX*/
   4058 		}
   4059 	}
   4060 	p->sadb_address_prefixlen = prefixlen;
   4061 	p->sadb_address_reserved = 0;
   4062 
   4063 	memcpy(mtod(m, char *) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
   4064 	    saddr, saddr->sa_len);
   4065 
   4066 	return m;
   4067 }
   4068 
   4069 #if 0
   4070 /*
   4071  * set data into sadb_ident.
   4072  */
   4073 static struct mbuf *
   4074 key_setsadbident(u_int16_t exttype, u_int16_t idtype,
   4075 		 void *string, int stringlen, u_int64_t id)
   4076 {
   4077 	struct mbuf *m;
   4078 	struct sadb_ident *p;
   4079 	size_t len;
   4080 
   4081 	len = PFKEY_ALIGN8(sizeof(struct sadb_ident)) + PFKEY_ALIGN8(stringlen);
   4082 	m = key_alloc_mbuf(len);
   4083 	if (!m || m->m_next) {	/*XXX*/
   4084 		if (m)
   4085 			m_freem(m);
   4086 		return NULL;
   4087 	}
   4088 
   4089 	p = mtod(m, struct sadb_ident *);
   4090 
   4091 	memset(p, 0, len);
   4092 	p->sadb_ident_len = PFKEY_UNIT64(len);
   4093 	p->sadb_ident_exttype = exttype;
   4094 	p->sadb_ident_type = idtype;
   4095 	p->sadb_ident_reserved = 0;
   4096 	p->sadb_ident_id = id;
   4097 
   4098 	memcpy(mtod(m, void *) + PFKEY_ALIGN8(sizeof(struct sadb_ident)),
   4099 	   	   string, stringlen);
   4100 
   4101 	return m;
   4102 }
   4103 #endif
   4104 
   4105 /*
   4106  * set data into sadb_x_sa2.
   4107  */
   4108 static struct mbuf *
   4109 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int16_t reqid)
   4110 {
   4111 	struct mbuf *m;
   4112 	struct sadb_x_sa2 *p;
   4113 	size_t len;
   4114 
   4115 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
   4116 	m = key_alloc_mbuf(len, M_WAITOK);
   4117 	KASSERT(m->m_next == NULL);
   4118 
   4119 	p = mtod(m, struct sadb_x_sa2 *);
   4120 
   4121 	memset(p, 0, len);
   4122 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
   4123 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
   4124 	p->sadb_x_sa2_mode = mode;
   4125 	p->sadb_x_sa2_reserved1 = 0;
   4126 	p->sadb_x_sa2_reserved2 = 0;
   4127 	p->sadb_x_sa2_sequence = seq;
   4128 	p->sadb_x_sa2_reqid = reqid;
   4129 
   4130 	return m;
   4131 }
   4132 
   4133 /*
   4134  * set data into sadb_x_policy
   4135  */
   4136 static struct mbuf *
   4137 key_setsadbxpolicy(const u_int16_t type, const u_int8_t dir, const u_int32_t id,
   4138     int mflag)
   4139 {
   4140 	struct mbuf *m;
   4141 	struct sadb_x_policy *p;
   4142 	size_t len;
   4143 
   4144 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
   4145 	m = key_alloc_mbuf(len, mflag);
   4146 	if (!m || m->m_next) {	/*XXX*/
   4147 		if (m)
   4148 			m_freem(m);
   4149 		return NULL;
   4150 	}
   4151 
   4152 	p = mtod(m, struct sadb_x_policy *);
   4153 
   4154 	memset(p, 0, len);
   4155 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
   4156 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
   4157 	p->sadb_x_policy_type = type;
   4158 	p->sadb_x_policy_dir = dir;
   4159 	p->sadb_x_policy_id = id;
   4160 
   4161 	return m;
   4162 }
   4163 
   4164 /* %%% utilities */
   4165 /*
   4166  * copy a buffer into the new buffer allocated.
   4167  */
   4168 static void *
   4169 key_newbuf(const void *src, u_int len)
   4170 {
   4171 	void *new;
   4172 
   4173 	new = kmem_alloc(len, KM_SLEEP);
   4174 	memcpy(new, src, len);
   4175 
   4176 	return new;
   4177 }
   4178 
   4179 /* compare my own address
   4180  * OUT:	1: true, i.e. my address.
   4181  *	0: false
   4182  */
   4183 int
   4184 key_ismyaddr(const struct sockaddr *sa)
   4185 {
   4186 #ifdef INET
   4187 	const struct sockaddr_in *sin;
   4188 	const struct in_ifaddr *ia;
   4189 	int s;
   4190 #endif
   4191 
   4192 	KASSERT(sa != NULL);
   4193 
   4194 	switch (sa->sa_family) {
   4195 #ifdef INET
   4196 	case AF_INET:
   4197 		sin = (const struct sockaddr_in *)sa;
   4198 		s = pserialize_read_enter();
   4199 		IN_ADDRLIST_READER_FOREACH(ia) {
   4200 			if (sin->sin_family == ia->ia_addr.sin_family &&
   4201 			    sin->sin_len == ia->ia_addr.sin_len &&
   4202 			    sin->sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
   4203 			{
   4204 				pserialize_read_exit(s);
   4205 				return 1;
   4206 			}
   4207 		}
   4208 		pserialize_read_exit(s);
   4209 		break;
   4210 #endif
   4211 #ifdef INET6
   4212 	case AF_INET6:
   4213 		return key_ismyaddr6((const struct sockaddr_in6 *)sa);
   4214 #endif
   4215 	}
   4216 
   4217 	return 0;
   4218 }
   4219 
   4220 #ifdef INET6
   4221 /*
   4222  * compare my own address for IPv6.
   4223  * 1: ours
   4224  * 0: other
   4225  * NOTE: derived ip6_input() in KAME. This is necessary to modify more.
   4226  */
   4227 #include <netinet6/in6_var.h>
   4228 
   4229 static int
   4230 key_ismyaddr6(const struct sockaddr_in6 *sin6)
   4231 {
   4232 	struct in6_ifaddr *ia;
   4233 	int s;
   4234 	struct psref psref;
   4235 	int bound;
   4236 	int ours = 1;
   4237 
   4238 	bound = curlwp_bind();
   4239 	s = pserialize_read_enter();
   4240 	IN6_ADDRLIST_READER_FOREACH(ia) {
   4241 		bool ingroup;
   4242 
   4243 		if (key_sockaddr_match((const struct sockaddr *)&sin6,
   4244 		    (const struct sockaddr *)&ia->ia_addr, 0)) {
   4245 			pserialize_read_exit(s);
   4246 			goto ours;
   4247 		}
   4248 		ia6_acquire(ia, &psref);
   4249 		pserialize_read_exit(s);
   4250 
   4251 		/*
   4252 		 * XXX Multicast
   4253 		 * XXX why do we care about multlicast here while we don't care
   4254 		 * about IPv4 multicast??
   4255 		 * XXX scope
   4256 		 */
   4257 		ingroup = in6_multi_group(&sin6->sin6_addr, ia->ia_ifp);
   4258 		if (ingroup) {
   4259 			ia6_release(ia, &psref);
   4260 			goto ours;
   4261 		}
   4262 
   4263 		s = pserialize_read_enter();
   4264 		ia6_release(ia, &psref);
   4265 	}
   4266 	pserialize_read_exit(s);
   4267 
   4268 	/* loopback, just for safety */
   4269 	if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))
   4270 		goto ours;
   4271 
   4272 	ours = 0;
   4273 ours:
   4274 	curlwp_bindx(bound);
   4275 
   4276 	return ours;
   4277 }
   4278 #endif /*INET6*/
   4279 
   4280 /*
   4281  * compare two secasindex structure.
   4282  * flag can specify to compare 2 saidxes.
   4283  * compare two secasindex structure without both mode and reqid.
   4284  * don't compare port.
   4285  * IN:
   4286  *      saidx0: source, it can be in SAD.
   4287  *      saidx1: object.
   4288  * OUT:
   4289  *      1 : equal
   4290  *      0 : not equal
   4291  */
   4292 static int
   4293 key_saidx_match(
   4294 	const struct secasindex *saidx0,
   4295 	const struct secasindex *saidx1,
   4296 	int flag)
   4297 {
   4298 	int chkport;
   4299 	const struct sockaddr *sa0src, *sa0dst, *sa1src, *sa1dst;
   4300 
   4301 	KASSERT(saidx0 != NULL);
   4302 	KASSERT(saidx1 != NULL);
   4303 
   4304 	/* sanity */
   4305 	if (saidx0->proto != saidx1->proto)
   4306 		return 0;
   4307 
   4308 	if (flag == CMP_EXACTLY) {
   4309 		if (saidx0->mode != saidx1->mode)
   4310 			return 0;
   4311 		if (saidx0->reqid != saidx1->reqid)
   4312 			return 0;
   4313 		if (memcmp(&saidx0->src, &saidx1->src, saidx0->src.sa.sa_len) != 0 ||
   4314 		    memcmp(&saidx0->dst, &saidx1->dst, saidx0->dst.sa.sa_len) != 0)
   4315 			return 0;
   4316 	} else {
   4317 
   4318 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
   4319 		if (flag == CMP_MODE_REQID ||flag == CMP_REQID) {
   4320 			/*
   4321 			 * If reqid of SPD is non-zero, unique SA is required.
   4322 			 * The result must be of same reqid in this case.
   4323 			 */
   4324 			if (saidx1->reqid != 0 && saidx0->reqid != saidx1->reqid)
   4325 				return 0;
   4326 		}
   4327 
   4328 		if (flag == CMP_MODE_REQID) {
   4329 			if (saidx0->mode != IPSEC_MODE_ANY &&
   4330 			    saidx0->mode != saidx1->mode)
   4331 				return 0;
   4332 		}
   4333 
   4334 
   4335 		sa0src = &saidx0->src.sa;
   4336 		sa0dst = &saidx0->dst.sa;
   4337 		sa1src = &saidx1->src.sa;
   4338 		sa1dst = &saidx1->dst.sa;
   4339 		/*
   4340 		 * If NAT-T is enabled, check ports for tunnel mode.
   4341 		 * Don't do it for transport mode, as there is no
   4342 		 * port information available in the SP.
   4343 		 * Also don't check ports if they are set to zero
   4344 		 * in the SPD: This means we have a non-generated
   4345 		 * SPD which can't know UDP ports.
   4346 		 */
   4347 		if (saidx1->mode == IPSEC_MODE_TUNNEL)
   4348 			chkport = PORT_LOOSE;
   4349 		else
   4350 			chkport = PORT_NONE;
   4351 
   4352 		if (!key_sockaddr_match(sa0src, sa1src, chkport)) {
   4353 			return 0;
   4354 		}
   4355 		if (!key_sockaddr_match(sa0dst, sa1dst, chkport)) {
   4356 			return 0;
   4357 		}
   4358 	}
   4359 
   4360 	return 1;
   4361 }
   4362 
   4363 /*
   4364  * compare two secindex structure exactly.
   4365  * IN:
   4366  *	spidx0: source, it is often in SPD.
   4367  *	spidx1: object, it is often from PFKEY message.
   4368  * OUT:
   4369  *	1 : equal
   4370  *	0 : not equal
   4371  */
   4372 static int
   4373 key_spidx_match_exactly(
   4374 	const struct secpolicyindex *spidx0,
   4375 	const struct secpolicyindex *spidx1)
   4376 {
   4377 
   4378 	KASSERT(spidx0 != NULL);
   4379 	KASSERT(spidx1 != NULL);
   4380 
   4381 	/* sanity */
   4382 	if (spidx0->prefs != spidx1->prefs ||
   4383 	    spidx0->prefd != spidx1->prefd ||
   4384 	    spidx0->ul_proto != spidx1->ul_proto)
   4385 		return 0;
   4386 
   4387 	return key_sockaddr_match(&spidx0->src.sa, &spidx1->src.sa, PORT_STRICT) &&
   4388 	       key_sockaddr_match(&spidx0->dst.sa, &spidx1->dst.sa, PORT_STRICT);
   4389 }
   4390 
   4391 /*
   4392  * compare two secindex structure with mask.
   4393  * IN:
   4394  *	spidx0: source, it is often in SPD.
   4395  *	spidx1: object, it is often from IP header.
   4396  * OUT:
   4397  *	1 : equal
   4398  *	0 : not equal
   4399  */
   4400 static int
   4401 key_spidx_match_withmask(
   4402 	const struct secpolicyindex *spidx0,
   4403 	const struct secpolicyindex *spidx1)
   4404 {
   4405 
   4406 	KASSERT(spidx0 != NULL);
   4407 	KASSERT(spidx1 != NULL);
   4408 
   4409 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
   4410 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
   4411 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
   4412 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
   4413 		return 0;
   4414 
   4415 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
   4416 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY &&
   4417 	    spidx0->ul_proto != spidx1->ul_proto)
   4418 		return 0;
   4419 
   4420 	switch (spidx0->src.sa.sa_family) {
   4421 	case AF_INET:
   4422 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY &&
   4423 		    spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
   4424 			return 0;
   4425 		if (!key_bb_match_withmask(&spidx0->src.sin.sin_addr,
   4426 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
   4427 			return 0;
   4428 		break;
   4429 	case AF_INET6:
   4430 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY &&
   4431 		    spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
   4432 			return 0;
   4433 		/*
   4434 		 * scope_id check. if sin6_scope_id is 0, we regard it
   4435 		 * as a wildcard scope, which matches any scope zone ID.
   4436 		 */
   4437 		if (spidx0->src.sin6.sin6_scope_id &&
   4438 		    spidx1->src.sin6.sin6_scope_id &&
   4439 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
   4440 			return 0;
   4441 		if (!key_bb_match_withmask(&spidx0->src.sin6.sin6_addr,
   4442 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
   4443 			return 0;
   4444 		break;
   4445 	default:
   4446 		/* XXX */
   4447 		if (memcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
   4448 			return 0;
   4449 		break;
   4450 	}
   4451 
   4452 	switch (spidx0->dst.sa.sa_family) {
   4453 	case AF_INET:
   4454 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY &&
   4455 		    spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
   4456 			return 0;
   4457 		if (!key_bb_match_withmask(&spidx0->dst.sin.sin_addr,
   4458 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
   4459 			return 0;
   4460 		break;
   4461 	case AF_INET6:
   4462 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY &&
   4463 		    spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
   4464 			return 0;
   4465 		/*
   4466 		 * scope_id check. if sin6_scope_id is 0, we regard it
   4467 		 * as a wildcard scope, which matches any scope zone ID.
   4468 		 */
   4469 		if (spidx0->src.sin6.sin6_scope_id &&
   4470 		    spidx1->src.sin6.sin6_scope_id &&
   4471 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
   4472 			return 0;
   4473 		if (!key_bb_match_withmask(&spidx0->dst.sin6.sin6_addr,
   4474 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
   4475 			return 0;
   4476 		break;
   4477 	default:
   4478 		/* XXX */
   4479 		if (memcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
   4480 			return 0;
   4481 		break;
   4482 	}
   4483 
   4484 	/* XXX Do we check other field ?  e.g. flowinfo */
   4485 
   4486 	return 1;
   4487 }
   4488 
   4489 /* returns 0 on match */
   4490 static int
   4491 key_portcomp(in_port_t port1, in_port_t port2, int howport)
   4492 {
   4493 	switch (howport) {
   4494 	case PORT_NONE:
   4495 		return 0;
   4496 	case PORT_LOOSE:
   4497 		if (port1 == 0 || port2 == 0)
   4498 			return 0;
   4499 		/*FALLTHROUGH*/
   4500 	case PORT_STRICT:
   4501 		if (port1 != port2) {
   4502 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   4503 			    "port fail %d != %d\n", port1, port2);
   4504 			return 1;
   4505 		}
   4506 		return 0;
   4507 	default:
   4508 		KASSERT(0);
   4509 		return 1;
   4510 	}
   4511 }
   4512 
   4513 /* returns 1 on match */
   4514 static int
   4515 key_sockaddr_match(
   4516 	const struct sockaddr *sa1,
   4517 	const struct sockaddr *sa2,
   4518 	int howport)
   4519 {
   4520 	const struct sockaddr_in *sin1, *sin2;
   4521 	const struct sockaddr_in6 *sin61, *sin62;
   4522 	char s1[IPSEC_ADDRSTRLEN], s2[IPSEC_ADDRSTRLEN];
   4523 
   4524 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len) {
   4525 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   4526 		    "fam/len fail %d != %d || %d != %d\n",
   4527 			sa1->sa_family, sa2->sa_family, sa1->sa_len,
   4528 			sa2->sa_len);
   4529 		return 0;
   4530 	}
   4531 
   4532 	switch (sa1->sa_family) {
   4533 	case AF_INET:
   4534 		if (sa1->sa_len != sizeof(struct sockaddr_in)) {
   4535 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   4536 			    "len fail %d != %zu\n",
   4537 			    sa1->sa_len, sizeof(struct sockaddr_in));
   4538 			return 0;
   4539 		}
   4540 		sin1 = (const struct sockaddr_in *)sa1;
   4541 		sin2 = (const struct sockaddr_in *)sa2;
   4542 		if (sin1->sin_addr.s_addr != sin2->sin_addr.s_addr) {
   4543 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   4544 			    "addr fail %s != %s\n",
   4545 			    (in_print(s1, sizeof(s1), &sin1->sin_addr), s1),
   4546 			    (in_print(s2, sizeof(s2), &sin2->sin_addr), s2));
   4547 			return 0;
   4548 		}
   4549 		if (key_portcomp(sin1->sin_port, sin2->sin_port, howport)) {
   4550 			return 0;
   4551 		}
   4552 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   4553 		    "addr success %s[%d] == %s[%d]\n",
   4554 		    (in_print(s1, sizeof(s1), &sin1->sin_addr), s1),
   4555 		    sin1->sin_port,
   4556 		    (in_print(s2, sizeof(s2), &sin2->sin_addr), s2),
   4557 		    sin2->sin_port);
   4558 		break;
   4559 	case AF_INET6:
   4560 		sin61 = (const struct sockaddr_in6 *)sa1;
   4561 		sin62 = (const struct sockaddr_in6 *)sa2;
   4562 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
   4563 			return 0;	/*EINVAL*/
   4564 
   4565 		if (sin61->sin6_scope_id != sin62->sin6_scope_id) {
   4566 			return 0;
   4567 		}
   4568 		if (!IN6_ARE_ADDR_EQUAL(&sin61->sin6_addr, &sin62->sin6_addr)) {
   4569 			return 0;
   4570 		}
   4571 		if (key_portcomp(sin61->sin6_port, sin62->sin6_port, howport)) {
   4572 			return 0;
   4573 		}
   4574 		break;
   4575 	default:
   4576 		if (memcmp(sa1, sa2, sa1->sa_len) != 0)
   4577 			return 0;
   4578 		break;
   4579 	}
   4580 
   4581 	return 1;
   4582 }
   4583 
   4584 /*
   4585  * compare two buffers with mask.
   4586  * IN:
   4587  *	addr1: source
   4588  *	addr2: object
   4589  *	bits:  Number of bits to compare
   4590  * OUT:
   4591  *	1 : equal
   4592  *	0 : not equal
   4593  */
   4594 static int
   4595 key_bb_match_withmask(const void *a1, const void *a2, u_int bits)
   4596 {
   4597 	const unsigned char *p1 = a1;
   4598 	const unsigned char *p2 = a2;
   4599 
   4600 	/* XXX: This could be considerably faster if we compare a word
   4601 	 * at a time, but it is complicated on LSB Endian machines */
   4602 
   4603 	/* Handle null pointers */
   4604 	if (p1 == NULL || p2 == NULL)
   4605 		return (p1 == p2);
   4606 
   4607 	while (bits >= 8) {
   4608 		if (*p1++ != *p2++)
   4609 			return 0;
   4610 		bits -= 8;
   4611 	}
   4612 
   4613 	if (bits > 0) {
   4614 		u_int8_t mask = ~((1<<(8-bits))-1);
   4615 		if ((*p1 & mask) != (*p2 & mask))
   4616 			return 0;
   4617 	}
   4618 	return 1;	/* Match! */
   4619 }
   4620 
   4621 static void
   4622 key_timehandler_spd(time_t now)
   4623 {
   4624 	u_int dir;
   4625 	struct secpolicy *sp;
   4626 
   4627 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
   4628 	    retry:
   4629 		mutex_enter(&key_spd.lock);
   4630 		SPLIST_WRITER_FOREACH(sp, dir) {
   4631 			KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
   4632 
   4633 			if (sp->lifetime == 0 && sp->validtime == 0)
   4634 				continue;
   4635 
   4636 			if ((sp->lifetime && now - sp->created > sp->lifetime) ||
   4637 			    (sp->validtime && now - sp->lastused > sp->validtime)) {
   4638 				key_unlink_sp(sp);
   4639 				mutex_exit(&key_spd.lock);
   4640 				key_spdexpire(sp);
   4641 				key_destroy_sp(sp);
   4642 				goto retry;
   4643 			}
   4644 		}
   4645 		mutex_exit(&key_spd.lock);
   4646 	}
   4647 
   4648     retry_socksplist:
   4649 	mutex_enter(&key_spd.lock);
   4650 	SOCKSPLIST_WRITER_FOREACH(sp) {
   4651 		if (sp->state != IPSEC_SPSTATE_DEAD)
   4652 			continue;
   4653 
   4654 		key_unlink_sp(sp);
   4655 		mutex_exit(&key_spd.lock);
   4656 		key_destroy_sp(sp);
   4657 		goto retry_socksplist;
   4658 	}
   4659 	mutex_exit(&key_spd.lock);
   4660 }
   4661 
   4662 static void
   4663 key_timehandler_sad(time_t now)
   4664 {
   4665 	struct secashead *sah;
   4666 	int s;
   4667 
   4668 restart:
   4669 	mutex_enter(&key_sad.lock);
   4670 	SAHLIST_WRITER_FOREACH(sah) {
   4671 		/* If sah has been dead and has no sav, then delete it */
   4672 		if (sah->state == SADB_SASTATE_DEAD &&
   4673 		    !key_sah_has_sav(sah)) {
   4674 			key_unlink_sah(sah);
   4675 			mutex_exit(&key_sad.lock);
   4676 			key_destroy_sah(sah);
   4677 			goto restart;
   4678 		}
   4679 	}
   4680 	mutex_exit(&key_sad.lock);
   4681 
   4682 	s = pserialize_read_enter();
   4683 	SAHLIST_READER_FOREACH(sah) {
   4684 		struct secasvar *sav;
   4685 
   4686 		key_sah_ref(sah);
   4687 		pserialize_read_exit(s);
   4688 
   4689 		/* if LARVAL entry doesn't become MATURE, delete it. */
   4690 		mutex_enter(&key_sad.lock);
   4691 	restart_sav_LARVAL:
   4692 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_LARVAL) {
   4693 			if (now - sav->created > key_larval_lifetime) {
   4694 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
   4695 				goto restart_sav_LARVAL;
   4696 			}
   4697 		}
   4698 		mutex_exit(&key_sad.lock);
   4699 
   4700 		/*
   4701 		 * check MATURE entry to start to send expire message
   4702 		 * whether or not.
   4703 		 */
   4704 	restart_sav_MATURE:
   4705 		mutex_enter(&key_sad.lock);
   4706 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_MATURE) {
   4707 			/* we don't need to check. */
   4708 			if (sav->lft_s == NULL)
   4709 				continue;
   4710 
   4711 			/* sanity check */
   4712 			KASSERT(sav->lft_c != NULL);
   4713 
   4714 			/* check SOFT lifetime */
   4715 			if (sav->lft_s->sadb_lifetime_addtime != 0 &&
   4716 			    now - sav->created > sav->lft_s->sadb_lifetime_addtime) {
   4717 				/*
   4718 				 * check SA to be used whether or not.
   4719 				 * when SA hasn't been used, delete it.
   4720 				 */
   4721 				if (sav->lft_c->sadb_lifetime_usetime == 0) {
   4722 					key_sa_chgstate(sav, SADB_SASTATE_DEAD);
   4723 					mutex_exit(&key_sad.lock);
   4724 				} else {
   4725 					key_sa_chgstate(sav, SADB_SASTATE_DYING);
   4726 					mutex_exit(&key_sad.lock);
   4727 					/*
   4728 					 * XXX If we keep to send expire
   4729 					 * message in the status of
   4730 					 * DYING. Do remove below code.
   4731 					 */
   4732 					key_expire(sav);
   4733 				}
   4734 				goto restart_sav_MATURE;
   4735 			}
   4736 			/* check SOFT lifetime by bytes */
   4737 			/*
   4738 			 * XXX I don't know the way to delete this SA
   4739 			 * when new SA is installed.  Caution when it's
   4740 			 * installed too big lifetime by time.
   4741 			 */
   4742 			else if (sav->lft_s->sadb_lifetime_bytes != 0 &&
   4743 			         sav->lft_s->sadb_lifetime_bytes <
   4744 			         sav->lft_c->sadb_lifetime_bytes) {
   4745 
   4746 				key_sa_chgstate(sav, SADB_SASTATE_DYING);
   4747 				mutex_exit(&key_sad.lock);
   4748 				/*
   4749 				 * XXX If we keep to send expire
   4750 				 * message in the status of
   4751 				 * DYING. Do remove below code.
   4752 				 */
   4753 				key_expire(sav);
   4754 				goto restart_sav_MATURE;
   4755 			}
   4756 		}
   4757 		mutex_exit(&key_sad.lock);
   4758 
   4759 		/* check DYING entry to change status to DEAD. */
   4760 		mutex_enter(&key_sad.lock);
   4761 	restart_sav_DYING:
   4762 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_DYING) {
   4763 			/* we don't need to check. */
   4764 			if (sav->lft_h == NULL)
   4765 				continue;
   4766 
   4767 			/* sanity check */
   4768 			KASSERT(sav->lft_c != NULL);
   4769 
   4770 			if (sav->lft_h->sadb_lifetime_addtime != 0 &&
   4771 			    now - sav->created > sav->lft_h->sadb_lifetime_addtime) {
   4772 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
   4773 				goto restart_sav_DYING;
   4774 			}
   4775 #if 0	/* XXX Should we keep to send expire message until HARD lifetime ? */
   4776 			else if (sav->lft_s != NULL
   4777 			      && sav->lft_s->sadb_lifetime_addtime != 0
   4778 			      && now - sav->created > sav->lft_s->sadb_lifetime_addtime) {
   4779 				/*
   4780 				 * XXX: should be checked to be
   4781 				 * installed the valid SA.
   4782 				 */
   4783 
   4784 				/*
   4785 				 * If there is no SA then sending
   4786 				 * expire message.
   4787 				 */
   4788 				key_expire(sav);
   4789 			}
   4790 #endif
   4791 			/* check HARD lifetime by bytes */
   4792 			else if (sav->lft_h->sadb_lifetime_bytes != 0 &&
   4793 			         sav->lft_h->sadb_lifetime_bytes <
   4794 			         sav->lft_c->sadb_lifetime_bytes) {
   4795 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
   4796 				goto restart_sav_DYING;
   4797 			}
   4798 		}
   4799 		mutex_exit(&key_sad.lock);
   4800 
   4801 		/* delete entry in DEAD */
   4802 	restart_sav_DEAD:
   4803 		mutex_enter(&key_sad.lock);
   4804 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_DEAD) {
   4805 			key_unlink_sav(sav);
   4806 			mutex_exit(&key_sad.lock);
   4807 			key_destroy_sav(sav);
   4808 			goto restart_sav_DEAD;
   4809 		}
   4810 		mutex_exit(&key_sad.lock);
   4811 
   4812 		s = pserialize_read_enter();
   4813 		key_sah_unref(sah);
   4814 	}
   4815 	pserialize_read_exit(s);
   4816 }
   4817 
   4818 static void
   4819 key_timehandler_acq(time_t now)
   4820 {
   4821 #ifndef IPSEC_NONBLOCK_ACQUIRE
   4822 	struct secacq *acq, *nextacq;
   4823 
   4824     restart:
   4825 	mutex_enter(&key_misc.lock);
   4826 	LIST_FOREACH_SAFE(acq, &key_misc.acqlist, chain, nextacq) {
   4827 		if (now - acq->created > key_blockacq_lifetime) {
   4828 			LIST_REMOVE(acq, chain);
   4829 			mutex_exit(&key_misc.lock);
   4830 			kmem_free(acq, sizeof(*acq));
   4831 			goto restart;
   4832 		}
   4833 	}
   4834 	mutex_exit(&key_misc.lock);
   4835 #endif
   4836 }
   4837 
   4838 static void
   4839 key_timehandler_spacq(time_t now)
   4840 {
   4841 #ifdef notyet
   4842 	struct secspacq *acq, *nextacq;
   4843 
   4844 	LIST_FOREACH_SAFE(acq, &key_misc.spacqlist, chain, nextacq) {
   4845 		if (now - acq->created > key_blockacq_lifetime) {
   4846 			KASSERT(__LIST_CHAINED(acq));
   4847 			LIST_REMOVE(acq, chain);
   4848 			kmem_free(acq, sizeof(*acq));
   4849 		}
   4850 	}
   4851 #endif
   4852 }
   4853 
   4854 static unsigned int key_timehandler_work_enqueued = 0;
   4855 
   4856 /*
   4857  * time handler.
   4858  * scanning SPD and SAD to check status for each entries,
   4859  * and do to remove or to expire.
   4860  */
   4861 static void
   4862 key_timehandler_work(struct work *wk, void *arg)
   4863 {
   4864 	time_t now = time_uptime;
   4865 	IPSEC_DECLARE_LOCK_VARIABLE;
   4866 
   4867 	/* We can allow enqueuing another work at this point */
   4868 	atomic_swap_uint(&key_timehandler_work_enqueued, 0);
   4869 
   4870 	IPSEC_ACQUIRE_GLOBAL_LOCKS();
   4871 
   4872 	key_timehandler_spd(now);
   4873 	key_timehandler_sad(now);
   4874 	key_timehandler_acq(now);
   4875 	key_timehandler_spacq(now);
   4876 
   4877 	key_acquire_sendup_pending_mbuf();
   4878 
   4879 	/* do exchange to tick time !! */
   4880 	callout_reset(&key_timehandler_ch, hz, key_timehandler, NULL);
   4881 
   4882 	IPSEC_RELEASE_GLOBAL_LOCKS();
   4883 	return;
   4884 }
   4885 
   4886 static void
   4887 key_timehandler(void *arg)
   4888 {
   4889 
   4890 	/* Avoid enqueuing another work when one is already enqueued */
   4891 	if (atomic_swap_uint(&key_timehandler_work_enqueued, 1) == 1)
   4892 		return;
   4893 
   4894 	workqueue_enqueue(key_timehandler_wq, &key_timehandler_wk, NULL);
   4895 }
   4896 
   4897 u_long
   4898 key_random(void)
   4899 {
   4900 	u_long value;
   4901 
   4902 	key_randomfill(&value, sizeof(value));
   4903 	return value;
   4904 }
   4905 
   4906 void
   4907 key_randomfill(void *p, size_t l)
   4908 {
   4909 
   4910 	cprng_fast(p, l);
   4911 }
   4912 
   4913 /*
   4914  * map SADB_SATYPE_* to IPPROTO_*.
   4915  * if satype == SADB_SATYPE then satype is mapped to ~0.
   4916  * OUT:
   4917  *	0: invalid satype.
   4918  */
   4919 static u_int16_t
   4920 key_satype2proto(u_int8_t satype)
   4921 {
   4922 	switch (satype) {
   4923 	case SADB_SATYPE_UNSPEC:
   4924 		return IPSEC_PROTO_ANY;
   4925 	case SADB_SATYPE_AH:
   4926 		return IPPROTO_AH;
   4927 	case SADB_SATYPE_ESP:
   4928 		return IPPROTO_ESP;
   4929 	case SADB_X_SATYPE_IPCOMP:
   4930 		return IPPROTO_IPCOMP;
   4931 	case SADB_X_SATYPE_TCPSIGNATURE:
   4932 		return IPPROTO_TCP;
   4933 	default:
   4934 		return 0;
   4935 	}
   4936 	/* NOTREACHED */
   4937 }
   4938 
   4939 /*
   4940  * map IPPROTO_* to SADB_SATYPE_*
   4941  * OUT:
   4942  *	0: invalid protocol type.
   4943  */
   4944 static u_int8_t
   4945 key_proto2satype(u_int16_t proto)
   4946 {
   4947 	switch (proto) {
   4948 	case IPPROTO_AH:
   4949 		return SADB_SATYPE_AH;
   4950 	case IPPROTO_ESP:
   4951 		return SADB_SATYPE_ESP;
   4952 	case IPPROTO_IPCOMP:
   4953 		return SADB_X_SATYPE_IPCOMP;
   4954 	case IPPROTO_TCP:
   4955 		return SADB_X_SATYPE_TCPSIGNATURE;
   4956 	default:
   4957 		return 0;
   4958 	}
   4959 	/* NOTREACHED */
   4960 }
   4961 
   4962 static int
   4963 key_setsecasidx(int proto, int mode, int reqid,
   4964     const struct sockaddr *src, const struct sockaddr *dst,
   4965     struct secasindex * saidx)
   4966 {
   4967 	const union sockaddr_union *src_u = (const union sockaddr_union *)src;
   4968 	const union sockaddr_union *dst_u = (const union sockaddr_union *)dst;
   4969 
   4970 	/* sa len safety check */
   4971 	if (key_checksalen(src_u) != 0)
   4972 		return -1;
   4973 	if (key_checksalen(dst_u) != 0)
   4974 		return -1;
   4975 
   4976 	memset(saidx, 0, sizeof(*saidx));
   4977 	saidx->proto = proto;
   4978 	saidx->mode = mode;
   4979 	saidx->reqid = reqid;
   4980 	memcpy(&saidx->src, src_u, src_u->sa.sa_len);
   4981 	memcpy(&saidx->dst, dst_u, dst_u->sa.sa_len);
   4982 
   4983 	key_porttosaddr(&((saidx)->src), 0);
   4984 	key_porttosaddr(&((saidx)->dst), 0);
   4985 	return 0;
   4986 }
   4987 
   4988 static void
   4989 key_init_spidx_bymsghdr(struct secpolicyindex *spidx,
   4990     const struct sadb_msghdr *mhp)
   4991 {
   4992 	const struct sadb_address *src0, *dst0;
   4993 	const struct sockaddr *src, *dst;
   4994 	const struct sadb_x_policy *xpl0;
   4995 
   4996 	src0 = mhp->ext[SADB_EXT_ADDRESS_SRC];
   4997 	dst0 = mhp->ext[SADB_EXT_ADDRESS_DST];
   4998 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   4999 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   5000 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
   5001 
   5002 	memset(spidx, 0, sizeof(*spidx));
   5003 	spidx->dir = xpl0->sadb_x_policy_dir;
   5004 	spidx->prefs = src0->sadb_address_prefixlen;
   5005 	spidx->prefd = dst0->sadb_address_prefixlen;
   5006 	spidx->ul_proto = src0->sadb_address_proto;
   5007 	/* XXX boundary check against sa_len */
   5008 	memcpy(&spidx->src, src, src->sa_len);
   5009 	memcpy(&spidx->dst, dst, dst->sa_len);
   5010 }
   5011 
   5012 /* %%% PF_KEY */
   5013 /*
   5014  * SADB_GETSPI processing is to receive
   5015  *	<base, (SA2), src address, dst address, (SPI range)>
   5016  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
   5017  * tree with the status of LARVAL, and send
   5018  *	<base, SA(*), address(SD)>
   5019  * to the IKMPd.
   5020  *
   5021  * IN:	mhp: pointer to the pointer to each header.
   5022  * OUT:	NULL if fail.
   5023  *	other if success, return pointer to the message to send.
   5024  */
   5025 static int
   5026 key_api_getspi(struct socket *so, struct mbuf *m,
   5027 	   const struct sadb_msghdr *mhp)
   5028 {
   5029 	const struct sockaddr *src, *dst;
   5030 	struct secasindex saidx;
   5031 	struct secashead *sah;
   5032 	struct secasvar *newsav;
   5033 	u_int8_t proto;
   5034 	u_int32_t spi;
   5035 	u_int8_t mode;
   5036 	u_int16_t reqid;
   5037 	int error;
   5038 
   5039 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   5040 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
   5041 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5042 		return key_senderror(so, m, EINVAL);
   5043 	}
   5044 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   5045 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
   5046 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5047 		return key_senderror(so, m, EINVAL);
   5048 	}
   5049 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
   5050 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
   5051 		mode = sa2->sadb_x_sa2_mode;
   5052 		reqid = sa2->sadb_x_sa2_reqid;
   5053 	} else {
   5054 		mode = IPSEC_MODE_ANY;
   5055 		reqid = 0;
   5056 	}
   5057 
   5058 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   5059 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   5060 
   5061 	/* map satype to proto */
   5062 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
   5063 	if (proto == 0) {
   5064 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   5065 		return key_senderror(so, m, EINVAL);
   5066 	}
   5067 
   5068 
   5069 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
   5070 	if (error != 0)
   5071 		return key_senderror(so, m, EINVAL);
   5072 
   5073 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   5074 	if (error != 0)
   5075 		return key_senderror(so, m, EINVAL);
   5076 
   5077 	/* SPI allocation */
   5078 	spi = key_do_getnewspi(mhp->ext[SADB_EXT_SPIRANGE], &saidx);
   5079 	if (spi == 0)
   5080 		return key_senderror(so, m, EINVAL);
   5081 
   5082 	/* get a SA index */
   5083 	sah = key_getsah_ref(&saidx, CMP_REQID);
   5084 	if (sah == NULL) {
   5085 		/* create a new SA index */
   5086 		sah = key_newsah(&saidx);
   5087 		if (sah == NULL) {
   5088 			IPSECLOG(LOG_DEBUG, "No more memory.\n");
   5089 			return key_senderror(so, m, ENOBUFS);
   5090 		}
   5091 	}
   5092 
   5093 	/* get a new SA */
   5094 	/* XXX rewrite */
   5095 	newsav = KEY_NEWSAV(m, mhp, &error);
   5096 	if (newsav == NULL) {
   5097 		key_sah_unref(sah);
   5098 		/* XXX don't free new SA index allocated in above. */
   5099 		return key_senderror(so, m, error);
   5100 	}
   5101 
   5102 	/* set spi */
   5103 	newsav->spi = htonl(spi);
   5104 
   5105 	/* Add to sah#savlist */
   5106 	key_init_sav(newsav);
   5107 	newsav->sah = sah;
   5108 	newsav->state = SADB_SASTATE_LARVAL;
   5109 	mutex_enter(&key_sad.lock);
   5110 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_LARVAL, newsav);
   5111 	mutex_exit(&key_sad.lock);
   5112 	key_validate_savlist(sah, SADB_SASTATE_LARVAL);
   5113 
   5114 	key_sah_unref(sah);
   5115 
   5116 #ifndef IPSEC_NONBLOCK_ACQUIRE
   5117 	/* delete the entry in key_misc.acqlist */
   5118 	if (mhp->msg->sadb_msg_seq != 0) {
   5119 		struct secacq *acq;
   5120 		mutex_enter(&key_misc.lock);
   5121 		acq = key_getacqbyseq(mhp->msg->sadb_msg_seq);
   5122 		if (acq != NULL) {
   5123 			/* reset counter in order to deletion by timehandler. */
   5124 			acq->created = time_uptime;
   5125 			acq->count = 0;
   5126 		}
   5127 		mutex_exit(&key_misc.lock);
   5128 	}
   5129 #endif
   5130 
   5131     {
   5132 	struct mbuf *n, *nn;
   5133 	struct sadb_sa *m_sa;
   5134 	int off, len;
   5135 
   5136 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
   5137 	    PFKEY_ALIGN8(sizeof(struct sadb_sa)) <= MCLBYTES);
   5138 
   5139 	/* create new sadb_msg to reply. */
   5140 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
   5141 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
   5142 
   5143 	n = key_alloc_mbuf_simple(len, M_WAITOK);
   5144 	n->m_len = len;
   5145 	n->m_next = NULL;
   5146 	off = 0;
   5147 
   5148 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
   5149 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
   5150 
   5151 	m_sa = (struct sadb_sa *)(mtod(n, char *) + off);
   5152 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
   5153 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
   5154 	m_sa->sadb_sa_spi = htonl(spi);
   5155 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
   5156 
   5157 	KASSERTMSG(off == len, "length inconsistency");
   5158 
   5159 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
   5160 	    SADB_EXT_ADDRESS_DST);
   5161 
   5162 	KASSERT(n->m_len >= sizeof(struct sadb_msg));
   5163 
   5164 	n->m_pkthdr.len = 0;
   5165 	for (nn = n; nn; nn = nn->m_next)
   5166 		n->m_pkthdr.len += nn->m_len;
   5167 
   5168 	key_fill_replymsg(n, newsav->seq);
   5169 	m_freem(m);
   5170 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
   5171     }
   5172 }
   5173 
   5174 /*
   5175  * allocating new SPI
   5176  * called by key_api_getspi().
   5177  * OUT:
   5178  *	0:	failure.
   5179  *	others: success.
   5180  */
   5181 static u_int32_t
   5182 key_do_getnewspi(const struct sadb_spirange *spirange,
   5183 		 const struct secasindex *saidx)
   5184 {
   5185 	u_int32_t newspi;
   5186 	u_int32_t spmin, spmax;
   5187 	int count = key_spi_trycnt;
   5188 
   5189 	/* set spi range to allocate */
   5190 	if (spirange != NULL) {
   5191 		spmin = spirange->sadb_spirange_min;
   5192 		spmax = spirange->sadb_spirange_max;
   5193 	} else {
   5194 		spmin = key_spi_minval;
   5195 		spmax = key_spi_maxval;
   5196 	}
   5197 	/* IPCOMP needs 2-byte SPI */
   5198 	if (saidx->proto == IPPROTO_IPCOMP) {
   5199 		u_int32_t t;
   5200 		if (spmin >= 0x10000)
   5201 			spmin = 0xffff;
   5202 		if (spmax >= 0x10000)
   5203 			spmax = 0xffff;
   5204 		if (spmin > spmax) {
   5205 			t = spmin; spmin = spmax; spmax = t;
   5206 		}
   5207 	}
   5208 
   5209 	if (spmin == spmax) {
   5210 		if (key_checkspidup(saidx, htonl(spmin))) {
   5211 			IPSECLOG(LOG_DEBUG, "SPI %u exists already.\n", spmin);
   5212 			return 0;
   5213 		}
   5214 
   5215 		count--; /* taking one cost. */
   5216 		newspi = spmin;
   5217 
   5218 	} else {
   5219 
   5220 		/* init SPI */
   5221 		newspi = 0;
   5222 
   5223 		/* when requesting to allocate spi ranged */
   5224 		while (count--) {
   5225 			/* generate pseudo-random SPI value ranged. */
   5226 			newspi = spmin + (key_random() % (spmax - spmin + 1));
   5227 
   5228 			if (!key_checkspidup(saidx, htonl(newspi)))
   5229 				break;
   5230 		}
   5231 
   5232 		if (count == 0 || newspi == 0) {
   5233 			IPSECLOG(LOG_DEBUG, "to allocate spi is failed.\n");
   5234 			return 0;
   5235 		}
   5236 	}
   5237 
   5238 	/* statistics */
   5239 	keystat.getspi_count =
   5240 	    (keystat.getspi_count + key_spi_trycnt - count) / 2;
   5241 
   5242 	return newspi;
   5243 }
   5244 
   5245 static int
   5246 key_handle_natt_info(struct secasvar *sav,
   5247       		     const struct sadb_msghdr *mhp)
   5248 {
   5249 	const char *msg = "?" ;
   5250 	struct sadb_x_nat_t_type *type;
   5251 	struct sadb_x_nat_t_port *sport, *dport;
   5252 	struct sadb_address *iaddr, *raddr;
   5253 	struct sadb_x_nat_t_frag *frag;
   5254 
   5255 	if (mhp->ext[SADB_X_EXT_NAT_T_TYPE] == NULL ||
   5256 	    mhp->ext[SADB_X_EXT_NAT_T_SPORT] == NULL ||
   5257 	    mhp->ext[SADB_X_EXT_NAT_T_DPORT] == NULL)
   5258 		return 0;
   5259 
   5260 	if (mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) {
   5261 		msg = "TYPE";
   5262 		goto bad;
   5263 	}
   5264 
   5265 	if (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) {
   5266 		msg = "SPORT";
   5267 		goto bad;
   5268 	}
   5269 
   5270 	if (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport)) {
   5271 		msg = "DPORT";
   5272 		goto bad;
   5273 	}
   5274 
   5275 	if (mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL) {
   5276 		IPSECLOG(LOG_DEBUG, "NAT-T OAi present\n");
   5277 		if (mhp->extlen[SADB_X_EXT_NAT_T_OAI] < sizeof(*iaddr)) {
   5278 			msg = "OAI";
   5279 			goto bad;
   5280 		}
   5281 	}
   5282 
   5283 	if (mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL) {
   5284 		IPSECLOG(LOG_DEBUG, "NAT-T OAr present\n");
   5285 		if (mhp->extlen[SADB_X_EXT_NAT_T_OAR] < sizeof(*raddr)) {
   5286 			msg = "OAR";
   5287 			goto bad;
   5288 		}
   5289 	}
   5290 
   5291 	if (mhp->ext[SADB_X_EXT_NAT_T_FRAG] != NULL) {
   5292 	    if (mhp->extlen[SADB_X_EXT_NAT_T_FRAG] < sizeof(*frag)) {
   5293 		    msg = "FRAG";
   5294 		    goto bad;
   5295 	    }
   5296 	}
   5297 
   5298 	type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
   5299 	sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
   5300 	dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
   5301 	iaddr = mhp->ext[SADB_X_EXT_NAT_T_OAI];
   5302 	raddr = mhp->ext[SADB_X_EXT_NAT_T_OAR];
   5303 	frag = mhp->ext[SADB_X_EXT_NAT_T_FRAG];
   5304 
   5305 	IPSECLOG(LOG_DEBUG, "type %d, sport = %d, dport = %d\n",
   5306 	    type->sadb_x_nat_t_type_type,
   5307 	    ntohs(sport->sadb_x_nat_t_port_port),
   5308 	    ntohs(dport->sadb_x_nat_t_port_port));
   5309 
   5310 	sav->natt_type = type->sadb_x_nat_t_type_type;
   5311 	key_porttosaddr(&sav->sah->saidx.src, sport->sadb_x_nat_t_port_port);
   5312 	key_porttosaddr(&sav->sah->saidx.dst, dport->sadb_x_nat_t_port_port);
   5313 	if (frag)
   5314 		sav->esp_frag = frag->sadb_x_nat_t_frag_fraglen;
   5315 	else
   5316 		sav->esp_frag = IP_MAXPACKET;
   5317 
   5318 	return 0;
   5319 bad:
   5320 	IPSECLOG(LOG_DEBUG, "invalid message %s\n", msg);
   5321 	__USE(msg);
   5322 	return -1;
   5323 }
   5324 
   5325 /* Just update the IPSEC_NAT_T ports if present */
   5326 static int
   5327 key_set_natt_ports(union sockaddr_union *src, union sockaddr_union *dst,
   5328       		     const struct sadb_msghdr *mhp)
   5329 {
   5330 	if (mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL)
   5331 		IPSECLOG(LOG_DEBUG, "NAT-T OAi present\n");
   5332 	if (mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL)
   5333 		IPSECLOG(LOG_DEBUG, "NAT-T OAr present\n");
   5334 
   5335 	if ((mhp->ext[SADB_X_EXT_NAT_T_TYPE] != NULL) &&
   5336 	    (mhp->ext[SADB_X_EXT_NAT_T_SPORT] != NULL) &&
   5337 	    (mhp->ext[SADB_X_EXT_NAT_T_DPORT] != NULL)) {
   5338 		struct sadb_x_nat_t_type *type;
   5339 		struct sadb_x_nat_t_port *sport;
   5340 		struct sadb_x_nat_t_port *dport;
   5341 
   5342 		if ((mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) ||
   5343 		    (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) ||
   5344 		    (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport))) {
   5345 			IPSECLOG(LOG_DEBUG, "invalid message\n");
   5346 			return -1;
   5347 		}
   5348 
   5349 		type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
   5350 		sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
   5351 		dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
   5352 
   5353 		key_porttosaddr(src, sport->sadb_x_nat_t_port_port);
   5354 		key_porttosaddr(dst, dport->sadb_x_nat_t_port_port);
   5355 
   5356 		IPSECLOG(LOG_DEBUG, "type %d, sport = %d, dport = %d\n",
   5357 		    type->sadb_x_nat_t_type_type,
   5358 		    ntohs(sport->sadb_x_nat_t_port_port),
   5359 		    ntohs(dport->sadb_x_nat_t_port_port));
   5360 	}
   5361 
   5362 	return 0;
   5363 }
   5364 
   5365 
   5366 /*
   5367  * SADB_UPDATE processing
   5368  * receive
   5369  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
   5370  *       key(AE), (identity(SD),) (sensitivity)>
   5371  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
   5372  * and send
   5373  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
   5374  *       (identity(SD),) (sensitivity)>
   5375  * to the ikmpd.
   5376  *
   5377  * m will always be freed.
   5378  */
   5379 static int
   5380 key_api_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
   5381 {
   5382 	struct sadb_sa *sa0;
   5383 	const struct sockaddr *src, *dst;
   5384 	struct secasindex saidx;
   5385 	struct secashead *sah;
   5386 	struct secasvar *sav, *newsav;
   5387 	u_int16_t proto;
   5388 	u_int8_t mode;
   5389 	u_int16_t reqid;
   5390 	int error;
   5391 
   5392 	/* map satype to proto */
   5393 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
   5394 	if (proto == 0) {
   5395 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   5396 		return key_senderror(so, m, EINVAL);
   5397 	}
   5398 
   5399 	if (mhp->ext[SADB_EXT_SA] == NULL ||
   5400 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   5401 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
   5402 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
   5403 	     mhp->ext[SADB_EXT_KEY_ENCRYPT] == NULL) ||
   5404 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
   5405 	     mhp->ext[SADB_EXT_KEY_AUTH] == NULL) ||
   5406 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL &&
   5407 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] == NULL) ||
   5408 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] == NULL &&
   5409 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] != NULL)) {
   5410 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5411 		return key_senderror(so, m, EINVAL);
   5412 	}
   5413 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
   5414 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   5415 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
   5416 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5417 		return key_senderror(so, m, EINVAL);
   5418 	}
   5419 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
   5420 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
   5421 		mode = sa2->sadb_x_sa2_mode;
   5422 		reqid = sa2->sadb_x_sa2_reqid;
   5423 	} else {
   5424 		mode = IPSEC_MODE_ANY;
   5425 		reqid = 0;
   5426 	}
   5427 	/* XXX boundary checking for other extensions */
   5428 
   5429 	sa0 = mhp->ext[SADB_EXT_SA];
   5430 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   5431 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   5432 
   5433 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
   5434 	if (error != 0)
   5435 		return key_senderror(so, m, EINVAL);
   5436 
   5437 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   5438 	if (error != 0)
   5439 		return key_senderror(so, m, EINVAL);
   5440 
   5441 	/* get a SA header */
   5442 	sah = key_getsah_ref(&saidx, CMP_REQID);
   5443 	if (sah == NULL) {
   5444 		IPSECLOG(LOG_DEBUG, "no SA index found.\n");
   5445 		return key_senderror(so, m, ENOENT);
   5446 	}
   5447 
   5448 	/* set spidx if there */
   5449 	/* XXX rewrite */
   5450 	error = key_setident(sah, m, mhp);
   5451 	if (error)
   5452 		goto error_sah;
   5453 
   5454 	/* find a SA with sequence number. */
   5455 #ifdef IPSEC_DOSEQCHECK
   5456 	if (mhp->msg->sadb_msg_seq != 0) {
   5457 		sav = key_getsavbyseq(sah, mhp->msg->sadb_msg_seq);
   5458 		if (sav == NULL) {
   5459 			IPSECLOG(LOG_DEBUG,
   5460 			    "no larval SA with sequence %u exists.\n",
   5461 			    mhp->msg->sadb_msg_seq);
   5462 			error = ENOENT;
   5463 			goto error_sah;
   5464 		}
   5465 	}
   5466 #else
   5467 	sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
   5468 	if (sav == NULL) {
   5469 		IPSECLOG(LOG_DEBUG, "no such a SA found (spi:%u)\n",
   5470 		    (u_int32_t)ntohl(sa0->sadb_sa_spi));
   5471 		error = EINVAL;
   5472 		goto error_sah;
   5473 	}
   5474 #endif
   5475 
   5476 	/* validity check */
   5477 	if (sav->sah->saidx.proto != proto) {
   5478 		IPSECLOG(LOG_DEBUG, "protocol mismatched (DB=%u param=%u)\n",
   5479 		    sav->sah->saidx.proto, proto);
   5480 		error = EINVAL;
   5481 		goto error;
   5482 	}
   5483 #ifdef IPSEC_DOSEQCHECK
   5484 	if (sav->spi != sa0->sadb_sa_spi) {
   5485 		IPSECLOG(LOG_DEBUG, "SPI mismatched (DB:%u param:%u)\n",
   5486 		    (u_int32_t)ntohl(sav->spi),
   5487 		    (u_int32_t)ntohl(sa0->sadb_sa_spi));
   5488 		error = EINVAL;
   5489 		goto error;
   5490 	}
   5491 #endif
   5492 	if (sav->pid != mhp->msg->sadb_msg_pid) {
   5493 		IPSECLOG(LOG_DEBUG, "pid mismatched (DB:%u param:%u)\n",
   5494 		    sav->pid, mhp->msg->sadb_msg_pid);
   5495 		error = EINVAL;
   5496 		goto error;
   5497 	}
   5498 
   5499 	/*
   5500 	 * Allocate a new SA instead of modifying the existing SA directly
   5501 	 * to avoid race conditions.
   5502 	 */
   5503 	newsav = kmem_zalloc(sizeof(struct secasvar), KM_SLEEP);
   5504 
   5505 	/* copy sav values */
   5506 	newsav->spi = sav->spi;
   5507 	newsav->seq = sav->seq;
   5508 	newsav->created = sav->created;
   5509 	newsav->pid = sav->pid;
   5510 	newsav->sah = sav->sah;
   5511 
   5512 	error = key_setsaval(newsav, m, mhp);
   5513 	if (error) {
   5514 		key_delsav(newsav);
   5515 		goto error;
   5516 	}
   5517 
   5518 	error = key_handle_natt_info(newsav, mhp);
   5519 	if (error != 0) {
   5520 		key_delsav(newsav);
   5521 		goto error;
   5522 	}
   5523 
   5524 	error = key_init_xform(newsav);
   5525 	if (error != 0) {
   5526 		key_delsav(newsav);
   5527 		goto error;
   5528 	}
   5529 
   5530 	/* Add to sah#savlist */
   5531 	key_init_sav(newsav);
   5532 	newsav->state = SADB_SASTATE_MATURE;
   5533 	mutex_enter(&key_sad.lock);
   5534 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_MATURE, newsav);
   5535 	mutex_exit(&key_sad.lock);
   5536 	key_validate_savlist(sah, SADB_SASTATE_MATURE);
   5537 
   5538 	key_sah_unref(sah);
   5539 	sah = NULL;
   5540 
   5541 	key_destroy_sav_with_ref(sav);
   5542 	sav = NULL;
   5543 
   5544     {
   5545 	struct mbuf *n;
   5546 
   5547 	/* set msg buf from mhp */
   5548 	n = key_getmsgbuf_x1(m, mhp);
   5549 	if (n == NULL) {
   5550 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
   5551 		return key_senderror(so, m, ENOBUFS);
   5552 	}
   5553 
   5554 	m_freem(m);
   5555 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   5556     }
   5557 error:
   5558 	KEY_SA_UNREF(&sav);
   5559 error_sah:
   5560 	key_sah_unref(sah);
   5561 	return key_senderror(so, m, error);
   5562 }
   5563 
   5564 /*
   5565  * search SAD with sequence for a SA which state is SADB_SASTATE_LARVAL.
   5566  * only called by key_api_update().
   5567  * OUT:
   5568  *	NULL	: not found
   5569  *	others	: found, pointer to a SA.
   5570  */
   5571 #ifdef IPSEC_DOSEQCHECK
   5572 static struct secasvar *
   5573 key_getsavbyseq(struct secashead *sah, u_int32_t seq)
   5574 {
   5575 	struct secasvar *sav;
   5576 	u_int state;
   5577 	int s;
   5578 
   5579 	state = SADB_SASTATE_LARVAL;
   5580 
   5581 	/* search SAD with sequence number ? */
   5582 	s = pserialize_read_enter();
   5583 	SAVLIST_READER_FOREACH(sav, sah, state) {
   5584 		KEY_CHKSASTATE(state, sav->state);
   5585 
   5586 		if (sav->seq == seq) {
   5587 			SA_ADDREF(sav);
   5588 			KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   5589 			    "DP cause refcnt++:%d SA:%p\n",
   5590 			    key_sa_refcnt(sav), sav);
   5591 			break;
   5592 		}
   5593 	}
   5594 	pserialize_read_exit(s);
   5595 
   5596 	return sav;
   5597 }
   5598 #endif
   5599 
   5600 /*
   5601  * SADB_ADD processing
   5602  * add an entry to SA database, when received
   5603  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
   5604  *       key(AE), (identity(SD),) (sensitivity)>
   5605  * from the ikmpd,
   5606  * and send
   5607  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
   5608  *       (identity(SD),) (sensitivity)>
   5609  * to the ikmpd.
   5610  *
   5611  * IGNORE identity and sensitivity messages.
   5612  *
   5613  * m will always be freed.
   5614  */
   5615 static int
   5616 key_api_add(struct socket *so, struct mbuf *m,
   5617 	const struct sadb_msghdr *mhp)
   5618 {
   5619 	struct sadb_sa *sa0;
   5620 	const struct sockaddr *src, *dst;
   5621 	struct secasindex saidx;
   5622 	struct secashead *sah;
   5623 	struct secasvar *newsav;
   5624 	u_int16_t proto;
   5625 	u_int8_t mode;
   5626 	u_int16_t reqid;
   5627 	int error;
   5628 
   5629 	/* map satype to proto */
   5630 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
   5631 	if (proto == 0) {
   5632 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   5633 		return key_senderror(so, m, EINVAL);
   5634 	}
   5635 
   5636 	if (mhp->ext[SADB_EXT_SA] == NULL ||
   5637 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   5638 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
   5639 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
   5640 	     mhp->ext[SADB_EXT_KEY_ENCRYPT] == NULL) ||
   5641 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
   5642 	     mhp->ext[SADB_EXT_KEY_AUTH] == NULL) ||
   5643 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL &&
   5644 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] == NULL) ||
   5645 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] == NULL &&
   5646 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] != NULL)) {
   5647 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5648 		return key_senderror(so, m, EINVAL);
   5649 	}
   5650 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
   5651 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   5652 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
   5653 		/* XXX need more */
   5654 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5655 		return key_senderror(so, m, EINVAL);
   5656 	}
   5657 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
   5658 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
   5659 		mode = sa2->sadb_x_sa2_mode;
   5660 		reqid = sa2->sadb_x_sa2_reqid;
   5661 	} else {
   5662 		mode = IPSEC_MODE_ANY;
   5663 		reqid = 0;
   5664 	}
   5665 
   5666 	sa0 = mhp->ext[SADB_EXT_SA];
   5667 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   5668 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   5669 
   5670 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
   5671 	if (error != 0)
   5672 		return key_senderror(so, m, EINVAL);
   5673 
   5674 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   5675 	if (error != 0)
   5676 		return key_senderror(so, m, EINVAL);
   5677 
   5678 	/* get a SA header */
   5679 	sah = key_getsah_ref(&saidx, CMP_REQID);
   5680 	if (sah == NULL) {
   5681 		/* create a new SA header */
   5682 		sah = key_newsah(&saidx);
   5683 		if (sah == NULL) {
   5684 			IPSECLOG(LOG_DEBUG, "No more memory.\n");
   5685 			return key_senderror(so, m, ENOBUFS);
   5686 		}
   5687 	}
   5688 
   5689 	/* set spidx if there */
   5690 	/* XXX rewrite */
   5691 	error = key_setident(sah, m, mhp);
   5692 	if (error)
   5693 		goto error;
   5694 
   5695     {
   5696 	struct secasvar *sav;
   5697 
   5698 	/* We can create new SA only if SPI is differenct. */
   5699 	sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
   5700 	if (sav != NULL) {
   5701 		KEY_SA_UNREF(&sav);
   5702 		IPSECLOG(LOG_DEBUG, "SA already exists.\n");
   5703 		error = EEXIST;
   5704 		goto error;
   5705 	}
   5706     }
   5707 
   5708 	/* create new SA entry. */
   5709 	newsav = KEY_NEWSAV(m, mhp, &error);
   5710 	if (newsav == NULL)
   5711 		goto error;
   5712 	newsav->sah = sah;
   5713 
   5714 	error = key_handle_natt_info(newsav, mhp);
   5715 	if (error != 0) {
   5716 		key_delsav(newsav);
   5717 		error = EINVAL;
   5718 		goto error;
   5719 	}
   5720 
   5721 	error = key_init_xform(newsav);
   5722 	if (error != 0) {
   5723 		key_delsav(newsav);
   5724 		goto error;
   5725 	}
   5726 
   5727 	/* Add to sah#savlist */
   5728 	key_init_sav(newsav);
   5729 	newsav->state = SADB_SASTATE_MATURE;
   5730 	mutex_enter(&key_sad.lock);
   5731 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_MATURE, newsav);
   5732 	mutex_exit(&key_sad.lock);
   5733 	key_validate_savlist(sah, SADB_SASTATE_MATURE);
   5734 
   5735 	key_sah_unref(sah);
   5736 	sah = NULL;
   5737 
   5738 	/*
   5739 	 * don't call key_freesav() here, as we would like to keep the SA
   5740 	 * in the database on success.
   5741 	 */
   5742 
   5743     {
   5744 	struct mbuf *n;
   5745 
   5746 	/* set msg buf from mhp */
   5747 	n = key_getmsgbuf_x1(m, mhp);
   5748 	if (n == NULL) {
   5749 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
   5750 		return key_senderror(so, m, ENOBUFS);
   5751 	}
   5752 
   5753 	m_freem(m);
   5754 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   5755     }
   5756 error:
   5757 	key_sah_unref(sah);
   5758 	return key_senderror(so, m, error);
   5759 }
   5760 
   5761 /* m is retained */
   5762 static int
   5763 key_setident(struct secashead *sah, struct mbuf *m,
   5764 	     const struct sadb_msghdr *mhp)
   5765 {
   5766 	const struct sadb_ident *idsrc, *iddst;
   5767 	int idsrclen, iddstlen;
   5768 
   5769 	KASSERT(!cpu_softintr_p());
   5770 	KASSERT(sah != NULL);
   5771 	KASSERT(m != NULL);
   5772 	KASSERT(mhp != NULL);
   5773 	KASSERT(mhp->msg != NULL);
   5774 
   5775 	/*
   5776 	 * Can be called with an existing sah from key_api_update().
   5777 	 */
   5778 	if (sah->idents != NULL) {
   5779 		kmem_free(sah->idents, sah->idents_len);
   5780 		sah->idents = NULL;
   5781 		sah->idents_len = 0;
   5782 	}
   5783 	if (sah->identd != NULL) {
   5784 		kmem_free(sah->identd, sah->identd_len);
   5785 		sah->identd = NULL;
   5786 		sah->identd_len = 0;
   5787 	}
   5788 
   5789 	/* don't make buffer if not there */
   5790 	if (mhp->ext[SADB_EXT_IDENTITY_SRC] == NULL &&
   5791 	    mhp->ext[SADB_EXT_IDENTITY_DST] == NULL) {
   5792 		sah->idents = NULL;
   5793 		sah->identd = NULL;
   5794 		return 0;
   5795 	}
   5796 
   5797 	if (mhp->ext[SADB_EXT_IDENTITY_SRC] == NULL ||
   5798 	    mhp->ext[SADB_EXT_IDENTITY_DST] == NULL) {
   5799 		IPSECLOG(LOG_DEBUG, "invalid identity.\n");
   5800 		return EINVAL;
   5801 	}
   5802 
   5803 	idsrc = mhp->ext[SADB_EXT_IDENTITY_SRC];
   5804 	iddst = mhp->ext[SADB_EXT_IDENTITY_DST];
   5805 	idsrclen = mhp->extlen[SADB_EXT_IDENTITY_SRC];
   5806 	iddstlen = mhp->extlen[SADB_EXT_IDENTITY_DST];
   5807 
   5808 	/* validity check */
   5809 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
   5810 		IPSECLOG(LOG_DEBUG, "ident type mismatch.\n");
   5811 		return EINVAL;
   5812 	}
   5813 
   5814 	switch (idsrc->sadb_ident_type) {
   5815 	case SADB_IDENTTYPE_PREFIX:
   5816 	case SADB_IDENTTYPE_FQDN:
   5817 	case SADB_IDENTTYPE_USERFQDN:
   5818 	default:
   5819 		/* XXX do nothing */
   5820 		sah->idents = NULL;
   5821 		sah->identd = NULL;
   5822 	 	return 0;
   5823 	}
   5824 
   5825 	/* make structure */
   5826 	sah->idents = kmem_alloc(idsrclen, KM_SLEEP);
   5827 	sah->idents_len = idsrclen;
   5828 	sah->identd = kmem_alloc(iddstlen, KM_SLEEP);
   5829 	sah->identd_len = iddstlen;
   5830 	memcpy(sah->idents, idsrc, idsrclen);
   5831 	memcpy(sah->identd, iddst, iddstlen);
   5832 
   5833 	return 0;
   5834 }
   5835 
   5836 /*
   5837  * m will not be freed on return. It never return NULL.
   5838  * it is caller's responsibility to free the result.
   5839  */
   5840 static struct mbuf *
   5841 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
   5842 {
   5843 	struct mbuf *n;
   5844 
   5845 	KASSERT(m != NULL);
   5846 	KASSERT(mhp != NULL);
   5847 	KASSERT(mhp->msg != NULL);
   5848 
   5849 	/* create new sadb_msg to reply. */
   5850 	n = key_gather_mbuf(m, mhp, 1, 15, SADB_EXT_RESERVED,
   5851 	    SADB_EXT_SA, SADB_X_EXT_SA2,
   5852 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
   5853 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
   5854 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
   5855 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
   5856 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
   5857 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NAT_T_FRAG);
   5858 
   5859 	KASSERT(n->m_len >= sizeof(struct sadb_msg));
   5860 
   5861 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
   5862 	mtod(n, struct sadb_msg *)->sadb_msg_len =
   5863 	    PFKEY_UNIT64(n->m_pkthdr.len);
   5864 
   5865 	return n;
   5866 }
   5867 
   5868 static int key_delete_all (struct socket *, struct mbuf *,
   5869 			   const struct sadb_msghdr *, u_int16_t);
   5870 
   5871 /*
   5872  * SADB_DELETE processing
   5873  * receive
   5874  *   <base, SA(*), address(SD)>
   5875  * from the ikmpd, and set SADB_SASTATE_DEAD,
   5876  * and send,
   5877  *   <base, SA(*), address(SD)>
   5878  * to the ikmpd.
   5879  *
   5880  * m will always be freed.
   5881  */
   5882 static int
   5883 key_api_delete(struct socket *so, struct mbuf *m,
   5884 	   const struct sadb_msghdr *mhp)
   5885 {
   5886 	struct sadb_sa *sa0;
   5887 	const struct sockaddr *src, *dst;
   5888 	struct secasindex saidx;
   5889 	struct secashead *sah;
   5890 	struct secasvar *sav = NULL;
   5891 	u_int16_t proto;
   5892 	int error;
   5893 
   5894 	/* map satype to proto */
   5895 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
   5896 	if (proto == 0) {
   5897 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   5898 		return key_senderror(so, m, EINVAL);
   5899 	}
   5900 
   5901 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   5902 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
   5903 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5904 		return key_senderror(so, m, EINVAL);
   5905 	}
   5906 
   5907 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   5908 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
   5909 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5910 		return key_senderror(so, m, EINVAL);
   5911 	}
   5912 
   5913 	if (mhp->ext[SADB_EXT_SA] == NULL) {
   5914 		/*
   5915 		 * Caller wants us to delete all non-LARVAL SAs
   5916 		 * that match the src/dst.  This is used during
   5917 		 * IKE INITIAL-CONTACT.
   5918 		 */
   5919 		IPSECLOG(LOG_DEBUG, "doing delete all.\n");
   5920 		return key_delete_all(so, m, mhp, proto);
   5921 	} else if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa)) {
   5922 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5923 		return key_senderror(so, m, EINVAL);
   5924 	}
   5925 
   5926 	sa0 = mhp->ext[SADB_EXT_SA];
   5927 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   5928 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   5929 
   5930 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
   5931 	if (error != 0)
   5932 		return key_senderror(so, m, EINVAL);
   5933 
   5934 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   5935 	if (error != 0)
   5936 		return key_senderror(so, m, EINVAL);
   5937 
   5938 	/* get a SA header */
   5939 	sah = key_getsah_ref(&saidx, CMP_HEAD);
   5940 	if (sah != NULL) {
   5941 		/* get a SA with SPI. */
   5942 		sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
   5943 		key_sah_unref(sah);
   5944 	}
   5945 
   5946 	if (sav == NULL) {
   5947 		IPSECLOG(LOG_DEBUG, "no SA found.\n");
   5948 		return key_senderror(so, m, ENOENT);
   5949 	}
   5950 
   5951 	key_destroy_sav_with_ref(sav);
   5952 	sav = NULL;
   5953 
   5954     {
   5955 	struct mbuf *n;
   5956 
   5957 	/* create new sadb_msg to reply. */
   5958 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
   5959 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
   5960 
   5961 	key_fill_replymsg(n, 0);
   5962 	m_freem(m);
   5963 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   5964     }
   5965 }
   5966 
   5967 /*
   5968  * delete all SAs for src/dst.  Called from key_api_delete().
   5969  */
   5970 static int
   5971 key_delete_all(struct socket *so, struct mbuf *m,
   5972 	       const struct sadb_msghdr *mhp, u_int16_t proto)
   5973 {
   5974 	const struct sockaddr *src, *dst;
   5975 	struct secasindex saidx;
   5976 	struct secashead *sah;
   5977 	struct secasvar *sav;
   5978 	u_int state;
   5979 	int error;
   5980 
   5981 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   5982 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   5983 
   5984 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
   5985 	if (error != 0)
   5986 		return key_senderror(so, m, EINVAL);
   5987 
   5988 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   5989 	if (error != 0)
   5990 		return key_senderror(so, m, EINVAL);
   5991 
   5992 	sah = key_getsah_ref(&saidx, CMP_HEAD);
   5993 	if (sah != NULL) {
   5994 		/* Delete all non-LARVAL SAs. */
   5995 		SASTATE_ALIVE_FOREACH(state) {
   5996 			if (state == SADB_SASTATE_LARVAL)
   5997 				continue;
   5998 		restart:
   5999 			mutex_enter(&key_sad.lock);
   6000 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
   6001 				sav->state = SADB_SASTATE_DEAD;
   6002 				key_unlink_sav(sav);
   6003 				mutex_exit(&key_sad.lock);
   6004 				key_destroy_sav(sav);
   6005 				goto restart;
   6006 			}
   6007 			mutex_exit(&key_sad.lock);
   6008 		}
   6009 		key_sah_unref(sah);
   6010 	}
   6011     {
   6012 	struct mbuf *n;
   6013 
   6014 	/* create new sadb_msg to reply. */
   6015 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
   6016 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
   6017 
   6018 	key_fill_replymsg(n, 0);
   6019 	m_freem(m);
   6020 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   6021     }
   6022 }
   6023 
   6024 /*
   6025  * SADB_GET processing
   6026  * receive
   6027  *   <base, SA(*), address(SD)>
   6028  * from the ikmpd, and get a SP and a SA to respond,
   6029  * and send,
   6030  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
   6031  *       (identity(SD),) (sensitivity)>
   6032  * to the ikmpd.
   6033  *
   6034  * m will always be freed.
   6035  */
   6036 static int
   6037 key_api_get(struct socket *so, struct mbuf *m,
   6038 	const struct sadb_msghdr *mhp)
   6039 {
   6040 	struct sadb_sa *sa0;
   6041 	const struct sockaddr *src, *dst;
   6042 	struct secasindex saidx;
   6043 	struct secasvar *sav = NULL;
   6044 	u_int16_t proto;
   6045 	int error;
   6046 
   6047 	/* map satype to proto */
   6048 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
   6049 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   6050 		return key_senderror(so, m, EINVAL);
   6051 	}
   6052 
   6053 	if (mhp->ext[SADB_EXT_SA] == NULL ||
   6054 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   6055 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
   6056 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   6057 		return key_senderror(so, m, EINVAL);
   6058 	}
   6059 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
   6060 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   6061 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
   6062 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   6063 		return key_senderror(so, m, EINVAL);
   6064 	}
   6065 
   6066 	sa0 = mhp->ext[SADB_EXT_SA];
   6067 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   6068 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   6069 
   6070 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
   6071 	if (error != 0)
   6072 		return key_senderror(so, m, EINVAL);
   6073 
   6074 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   6075 	if (error != 0)
   6076 		return key_senderror(so, m, EINVAL);
   6077 
   6078 	/* get a SA header */
   6079     {
   6080 	struct secashead *sah;
   6081 	int s = pserialize_read_enter();
   6082 
   6083 	sah = key_getsah(&saidx, CMP_HEAD);
   6084 	if (sah != NULL) {
   6085 		/* get a SA with SPI. */
   6086 		sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
   6087 	}
   6088 	pserialize_read_exit(s);
   6089     }
   6090 	if (sav == NULL) {
   6091 		IPSECLOG(LOG_DEBUG, "no SA found.\n");
   6092 		return key_senderror(so, m, ENOENT);
   6093 	}
   6094 
   6095     {
   6096 	struct mbuf *n;
   6097 	u_int8_t satype;
   6098 
   6099 	/* map proto to satype */
   6100 	satype = key_proto2satype(sav->sah->saidx.proto);
   6101 	if (satype == 0) {
   6102 		KEY_SA_UNREF(&sav);
   6103 		IPSECLOG(LOG_DEBUG, "there was invalid proto in SAD.\n");
   6104 		return key_senderror(so, m, EINVAL);
   6105 	}
   6106 
   6107 	/* create new sadb_msg to reply. */
   6108 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
   6109 	    mhp->msg->sadb_msg_pid);
   6110 	KEY_SA_UNREF(&sav);
   6111 	m_freem(m);
   6112 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
   6113     }
   6114 }
   6115 
   6116 /* XXX make it sysctl-configurable? */
   6117 static void
   6118 key_getcomb_setlifetime(struct sadb_comb *comb)
   6119 {
   6120 
   6121 	comb->sadb_comb_soft_allocations = 1;
   6122 	comb->sadb_comb_hard_allocations = 1;
   6123 	comb->sadb_comb_soft_bytes = 0;
   6124 	comb->sadb_comb_hard_bytes = 0;
   6125 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
   6126 	comb->sadb_comb_soft_addtime = comb->sadb_comb_hard_addtime * 80 / 100;
   6127 	comb->sadb_comb_hard_usetime = 28800;	/* 8 hours */
   6128 	comb->sadb_comb_soft_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
   6129 }
   6130 
   6131 /*
   6132  * XXX reorder combinations by preference
   6133  * XXX no idea if the user wants ESP authentication or not
   6134  */
   6135 static struct mbuf *
   6136 key_getcomb_esp(int mflag)
   6137 {
   6138 	struct sadb_comb *comb;
   6139 	const struct enc_xform *algo;
   6140 	struct mbuf *result = NULL, *m, *n;
   6141 	int encmin;
   6142 	int i, off, o;
   6143 	int totlen;
   6144 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
   6145 
   6146 	m = NULL;
   6147 	for (i = 1; i <= SADB_EALG_MAX; i++) {
   6148 		algo = esp_algorithm_lookup(i);
   6149 		if (algo == NULL)
   6150 			continue;
   6151 
   6152 		/* discard algorithms with key size smaller than system min */
   6153 		if (_BITS(algo->maxkey) < ipsec_esp_keymin)
   6154 			continue;
   6155 		if (_BITS(algo->minkey) < ipsec_esp_keymin)
   6156 			encmin = ipsec_esp_keymin;
   6157 		else
   6158 			encmin = _BITS(algo->minkey);
   6159 
   6160 		if (ipsec_esp_auth)
   6161 			m = key_getcomb_ah(mflag);
   6162 		else {
   6163 			KASSERTMSG(l <= MLEN,
   6164 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
   6165 			MGET(m, mflag, MT_DATA);
   6166 			if (m) {
   6167 				M_ALIGN(m, l);
   6168 				m->m_len = l;
   6169 				m->m_next = NULL;
   6170 				memset(mtod(m, void *), 0, m->m_len);
   6171 			}
   6172 		}
   6173 		if (!m)
   6174 			goto fail;
   6175 
   6176 		totlen = 0;
   6177 		for (n = m; n; n = n->m_next)
   6178 			totlen += n->m_len;
   6179 		KASSERTMSG((totlen % l) == 0, "totlen=%u, l=%u", totlen, l);
   6180 
   6181 		for (off = 0; off < totlen; off += l) {
   6182 			n = m_pulldown(m, off, l, &o);
   6183 			if (!n) {
   6184 				/* m is already freed */
   6185 				goto fail;
   6186 			}
   6187 			comb = (struct sadb_comb *)(mtod(n, char *) + o);
   6188 			memset(comb, 0, sizeof(*comb));
   6189 			key_getcomb_setlifetime(comb);
   6190 			comb->sadb_comb_encrypt = i;
   6191 			comb->sadb_comb_encrypt_minbits = encmin;
   6192 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
   6193 		}
   6194 
   6195 		if (!result)
   6196 			result = m;
   6197 		else
   6198 			m_cat(result, m);
   6199 	}
   6200 
   6201 	return result;
   6202 
   6203  fail:
   6204 	if (result)
   6205 		m_freem(result);
   6206 	return NULL;
   6207 }
   6208 
   6209 static void
   6210 key_getsizes_ah(const struct auth_hash *ah, int alg,
   6211 	        u_int16_t* ksmin, u_int16_t* ksmax)
   6212 {
   6213 	*ksmin = *ksmax = ah->keysize;
   6214 	if (ah->keysize == 0) {
   6215 		/*
   6216 		 * Transform takes arbitrary key size but algorithm
   6217 		 * key size is restricted.  Enforce this here.
   6218 		 */
   6219 		switch (alg) {
   6220 		case SADB_X_AALG_MD5:	*ksmin = *ksmax = 16; break;
   6221 		case SADB_X_AALG_SHA:	*ksmin = *ksmax = 20; break;
   6222 		case SADB_X_AALG_NULL:	*ksmin = 0; *ksmax = 256; break;
   6223 		default:
   6224 			IPSECLOG(LOG_DEBUG, "unknown AH algorithm %u\n", alg);
   6225 			break;
   6226 		}
   6227 	}
   6228 }
   6229 
   6230 /*
   6231  * XXX reorder combinations by preference
   6232  */
   6233 static struct mbuf *
   6234 key_getcomb_ah(int mflag)
   6235 {
   6236 	struct sadb_comb *comb;
   6237 	const struct auth_hash *algo;
   6238 	struct mbuf *m;
   6239 	u_int16_t minkeysize, maxkeysize;
   6240 	int i;
   6241 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
   6242 
   6243 	m = NULL;
   6244 	for (i = 1; i <= SADB_AALG_MAX; i++) {
   6245 #if 1
   6246 		/* we prefer HMAC algorithms, not old algorithms */
   6247 		if (i != SADB_AALG_SHA1HMAC &&
   6248 		    i != SADB_AALG_MD5HMAC &&
   6249 		    i != SADB_X_AALG_SHA2_256 &&
   6250 		    i != SADB_X_AALG_SHA2_384 &&
   6251 		    i != SADB_X_AALG_SHA2_512)
   6252 			continue;
   6253 #endif
   6254 		algo = ah_algorithm_lookup(i);
   6255 		if (!algo)
   6256 			continue;
   6257 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
   6258 		/* discard algorithms with key size smaller than system min */
   6259 		if (_BITS(minkeysize) < ipsec_ah_keymin)
   6260 			continue;
   6261 
   6262 		if (!m) {
   6263 			KASSERTMSG(l <= MLEN,
   6264 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
   6265 			MGET(m, mflag, MT_DATA);
   6266 			if (m) {
   6267 				M_ALIGN(m, l);
   6268 				m->m_len = l;
   6269 				m->m_next = NULL;
   6270 			}
   6271 		} else
   6272 			M_PREPEND(m, l, mflag);
   6273 		if (!m)
   6274 			return NULL;
   6275 
   6276 		if (m->m_len < sizeof(struct sadb_comb)) {
   6277 			m = m_pullup(m, sizeof(struct sadb_comb));
   6278 			if (m == NULL)
   6279 				return NULL;
   6280 		}
   6281 
   6282 		comb = mtod(m, struct sadb_comb *);
   6283 		memset(comb, 0, sizeof(*comb));
   6284 		key_getcomb_setlifetime(comb);
   6285 		comb->sadb_comb_auth = i;
   6286 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
   6287 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
   6288 	}
   6289 
   6290 	return m;
   6291 }
   6292 
   6293 /*
   6294  * not really an official behavior.  discussed in pf_key (at) inner.net in Sep2000.
   6295  * XXX reorder combinations by preference
   6296  */
   6297 static struct mbuf *
   6298 key_getcomb_ipcomp(int mflag)
   6299 {
   6300 	struct sadb_comb *comb;
   6301 	const struct comp_algo *algo;
   6302 	struct mbuf *m;
   6303 	int i;
   6304 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
   6305 
   6306 	m = NULL;
   6307 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
   6308 		algo = ipcomp_algorithm_lookup(i);
   6309 		if (!algo)
   6310 			continue;
   6311 
   6312 		if (!m) {
   6313 			KASSERTMSG(l <= MLEN,
   6314 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
   6315 			MGET(m, mflag, MT_DATA);
   6316 			if (m) {
   6317 				M_ALIGN(m, l);
   6318 				m->m_len = l;
   6319 				m->m_next = NULL;
   6320 			}
   6321 		} else
   6322 			M_PREPEND(m, l, mflag);
   6323 		if (!m)
   6324 			return NULL;
   6325 
   6326 		if (m->m_len < sizeof(struct sadb_comb)) {
   6327 			m = m_pullup(m, sizeof(struct sadb_comb));
   6328 			if (m == NULL)
   6329 				return NULL;
   6330 		}
   6331 
   6332 		comb = mtod(m, struct sadb_comb *);
   6333 		memset(comb, 0, sizeof(*comb));
   6334 		key_getcomb_setlifetime(comb);
   6335 		comb->sadb_comb_encrypt = i;
   6336 		/* what should we set into sadb_comb_*_{min,max}bits? */
   6337 	}
   6338 
   6339 	return m;
   6340 }
   6341 
   6342 /*
   6343  * XXX no way to pass mode (transport/tunnel) to userland
   6344  * XXX replay checking?
   6345  * XXX sysctl interface to ipsec_{ah,esp}_keymin
   6346  */
   6347 static struct mbuf *
   6348 key_getprop(const struct secasindex *saidx, int mflag)
   6349 {
   6350 	struct sadb_prop *prop;
   6351 	struct mbuf *m, *n;
   6352 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
   6353 	int totlen;
   6354 
   6355 	switch (saidx->proto)  {
   6356 	case IPPROTO_ESP:
   6357 		m = key_getcomb_esp(mflag);
   6358 		break;
   6359 	case IPPROTO_AH:
   6360 		m = key_getcomb_ah(mflag);
   6361 		break;
   6362 	case IPPROTO_IPCOMP:
   6363 		m = key_getcomb_ipcomp(mflag);
   6364 		break;
   6365 	default:
   6366 		return NULL;
   6367 	}
   6368 
   6369 	if (!m)
   6370 		return NULL;
   6371 	M_PREPEND(m, l, mflag);
   6372 	if (!m)
   6373 		return NULL;
   6374 
   6375 	totlen = 0;
   6376 	for (n = m; n; n = n->m_next)
   6377 		totlen += n->m_len;
   6378 
   6379 	prop = mtod(m, struct sadb_prop *);
   6380 	memset(prop, 0, sizeof(*prop));
   6381 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
   6382 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
   6383 	prop->sadb_prop_replay = 32;	/* XXX */
   6384 
   6385 	return m;
   6386 }
   6387 
   6388 /*
   6389  * SADB_ACQUIRE processing called by key_checkrequest() and key_api_acquire().
   6390  * send
   6391  *   <base, SA, address(SD), (address(P)), x_policy,
   6392  *       (identity(SD),) (sensitivity,) proposal>
   6393  * to KMD, and expect to receive
   6394  *   <base> with SADB_ACQUIRE if error occurred,
   6395  * or
   6396  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
   6397  * from KMD by PF_KEY.
   6398  *
   6399  * XXX x_policy is outside of RFC2367 (KAME extension).
   6400  * XXX sensitivity is not supported.
   6401  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
   6402  * see comment for key_getcomb_ipcomp().
   6403  *
   6404  * OUT:
   6405  *    0     : succeed
   6406  *    others: error number
   6407  */
   6408 static int
   6409 key_acquire(const struct secasindex *saidx, const struct secpolicy *sp, int mflag)
   6410 {
   6411 	struct mbuf *result = NULL, *m;
   6412 #ifndef IPSEC_NONBLOCK_ACQUIRE
   6413 	struct secacq *newacq;
   6414 #endif
   6415 	u_int8_t satype;
   6416 	int error = -1;
   6417 	u_int32_t seq;
   6418 
   6419 	/* sanity check */
   6420 	KASSERT(saidx != NULL);
   6421 	satype = key_proto2satype(saidx->proto);
   6422 	KASSERTMSG(satype != 0, "null satype, protocol %u", saidx->proto);
   6423 
   6424 #ifndef IPSEC_NONBLOCK_ACQUIRE
   6425 	/*
   6426 	 * We never do anything about acquirng SA.  There is anather
   6427 	 * solution that kernel blocks to send SADB_ACQUIRE message until
   6428 	 * getting something message from IKEd.  In later case, to be
   6429 	 * managed with ACQUIRING list.
   6430 	 */
   6431 	/* Get an entry to check whether sending message or not. */
   6432 	mutex_enter(&key_misc.lock);
   6433 	newacq = key_getacq(saidx);
   6434 	if (newacq != NULL) {
   6435 		if (key_blockacq_count < newacq->count) {
   6436 			/* reset counter and do send message. */
   6437 			newacq->count = 0;
   6438 		} else {
   6439 			/* increment counter and do nothing. */
   6440 			newacq->count++;
   6441 			mutex_exit(&key_misc.lock);
   6442 			return 0;
   6443 		}
   6444 	} else {
   6445 		/* make new entry for blocking to send SADB_ACQUIRE. */
   6446 		newacq = key_newacq(saidx);
   6447 		if (newacq == NULL) {
   6448 			mutex_exit(&key_misc.lock);
   6449 			return ENOBUFS;
   6450 		}
   6451 
   6452 		/* add to key_misc.acqlist */
   6453 		LIST_INSERT_HEAD(&key_misc.acqlist, newacq, chain);
   6454 	}
   6455 
   6456 	seq = newacq->seq;
   6457 	mutex_exit(&key_misc.lock);
   6458 #else
   6459 	seq = (acq_seq = (acq_seq == ~0 ? 1 : ++acq_seq));
   6460 #endif
   6461 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0, mflag);
   6462 	if (!m) {
   6463 		error = ENOBUFS;
   6464 		goto fail;
   6465 	}
   6466 	result = m;
   6467 
   6468 	/* set sadb_address for saidx's. */
   6469 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &saidx->src.sa, FULLMASK,
   6470 	    IPSEC_ULPROTO_ANY, mflag);
   6471 	if (!m) {
   6472 		error = ENOBUFS;
   6473 		goto fail;
   6474 	}
   6475 	m_cat(result, m);
   6476 
   6477 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &saidx->dst.sa, FULLMASK,
   6478 	    IPSEC_ULPROTO_ANY, mflag);
   6479 	if (!m) {
   6480 		error = ENOBUFS;
   6481 		goto fail;
   6482 	}
   6483 	m_cat(result, m);
   6484 
   6485 	/* XXX proxy address (optional) */
   6486 
   6487 	/* set sadb_x_policy */
   6488 	if (sp) {
   6489 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
   6490 		    mflag);
   6491 		if (!m) {
   6492 			error = ENOBUFS;
   6493 			goto fail;
   6494 		}
   6495 		m_cat(result, m);
   6496 	}
   6497 
   6498 	/* XXX identity (optional) */
   6499 #if 0
   6500 	if (idexttype && fqdn) {
   6501 		/* create identity extension (FQDN) */
   6502 		struct sadb_ident *id;
   6503 		int fqdnlen;
   6504 
   6505 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
   6506 		id = (struct sadb_ident *)p;
   6507 		memset(id, 0, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
   6508 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
   6509 		id->sadb_ident_exttype = idexttype;
   6510 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
   6511 		memcpy(id + 1, fqdn, fqdnlen);
   6512 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
   6513 	}
   6514 
   6515 	if (idexttype) {
   6516 		/* create identity extension (USERFQDN) */
   6517 		struct sadb_ident *id;
   6518 		int userfqdnlen;
   6519 
   6520 		if (userfqdn) {
   6521 			/* +1 for terminating-NUL */
   6522 			userfqdnlen = strlen(userfqdn) + 1;
   6523 		} else
   6524 			userfqdnlen = 0;
   6525 		id = (struct sadb_ident *)p;
   6526 		memset(id, 0, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
   6527 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
   6528 		id->sadb_ident_exttype = idexttype;
   6529 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
   6530 		/* XXX is it correct? */
   6531 		if (curlwp)
   6532 			id->sadb_ident_id = kauth_cred_getuid(curlwp->l_cred);
   6533 		if (userfqdn && userfqdnlen)
   6534 			memcpy(id + 1, userfqdn, userfqdnlen);
   6535 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
   6536 	}
   6537 #endif
   6538 
   6539 	/* XXX sensitivity (optional) */
   6540 
   6541 	/* create proposal/combination extension */
   6542 	m = key_getprop(saidx, mflag);
   6543 #if 0
   6544 	/*
   6545 	 * spec conformant: always attach proposal/combination extension,
   6546 	 * the problem is that we have no way to attach it for ipcomp,
   6547 	 * due to the way sadb_comb is declared in RFC2367.
   6548 	 */
   6549 	if (!m) {
   6550 		error = ENOBUFS;
   6551 		goto fail;
   6552 	}
   6553 	m_cat(result, m);
   6554 #else
   6555 	/*
   6556 	 * outside of spec; make proposal/combination extension optional.
   6557 	 */
   6558 	if (m)
   6559 		m_cat(result, m);
   6560 #endif
   6561 
   6562 	KASSERT(result->m_flags & M_PKTHDR);
   6563 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
   6564 
   6565 	result->m_pkthdr.len = 0;
   6566 	for (m = result; m; m = m->m_next)
   6567 		result->m_pkthdr.len += m->m_len;
   6568 
   6569 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   6570 	    PFKEY_UNIT64(result->m_pkthdr.len);
   6571 
   6572 	/*
   6573 	 * Called from key_api_acquire that must come from userland, so
   6574 	 * we can call key_sendup_mbuf immediately.
   6575 	 */
   6576 	if (mflag == M_WAITOK)
   6577 		return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
   6578 	/*
   6579 	 * XXX we cannot call key_sendup_mbuf directly here because
   6580 	 * it can cause a deadlock:
   6581 	 * - We have a reference to an SP (and an SA) here
   6582 	 * - key_sendup_mbuf will try to take key_so_mtx
   6583 	 * - Some other thread may try to localcount_drain to the SP with
   6584 	 *   holding key_so_mtx in say key_api_spdflush
   6585 	 * - In this case localcount_drain never return because key_sendup_mbuf
   6586 	 *   that has stuck on key_so_mtx never release a reference to the SP
   6587 	 *
   6588 	 * So defer key_sendup_mbuf to the timer.
   6589 	 */
   6590 	return key_acquire_sendup_mbuf_later(result);
   6591 
   6592  fail:
   6593 	if (result)
   6594 		m_freem(result);
   6595 	return error;
   6596 }
   6597 
   6598 static struct mbuf *key_acquire_mbuf_head = NULL;
   6599 static unsigned key_acquire_mbuf_count = 0;
   6600 #define KEY_ACQUIRE_MBUF_MAX	10
   6601 
   6602 static void
   6603 key_acquire_sendup_pending_mbuf(void)
   6604 {
   6605 	struct mbuf *m, *prev;
   6606 	int error;
   6607 
   6608 again:
   6609 	prev = NULL;
   6610 	mutex_enter(&key_misc.lock);
   6611 	m = key_acquire_mbuf_head;
   6612 	/* Get an earliest mbuf (one at the tail of the list) */
   6613 	while (m != NULL) {
   6614 		if (m->m_nextpkt == NULL) {
   6615 			if (prev != NULL)
   6616 				prev->m_nextpkt = NULL;
   6617 			if (m == key_acquire_mbuf_head)
   6618 				key_acquire_mbuf_head = NULL;
   6619 			key_acquire_mbuf_count--;
   6620 			break;
   6621 		}
   6622 		prev = m;
   6623 		m = m->m_nextpkt;
   6624 	}
   6625 	mutex_exit(&key_misc.lock);
   6626 
   6627 	if (m == NULL)
   6628 		return;
   6629 
   6630 	m->m_nextpkt = NULL;
   6631 	error = key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
   6632 	if (error != 0)
   6633 		IPSECLOG(LOG_WARNING, "key_sendup_mbuf failed (error=%d)\n",
   6634 		    error);
   6635 
   6636 	if (prev != NULL)
   6637 		goto again;
   6638 }
   6639 
   6640 static int
   6641 key_acquire_sendup_mbuf_later(struct mbuf *m)
   6642 {
   6643 
   6644 	mutex_enter(&key_misc.lock);
   6645 	/* Avoid queuing too much mbufs */
   6646 	if (key_acquire_mbuf_count >= KEY_ACQUIRE_MBUF_MAX) {
   6647 		mutex_exit(&key_misc.lock);
   6648 		m_freem(m);
   6649 		return ENOBUFS; /* XXX */
   6650 	}
   6651 	/* Enqueue mbuf at the head of the list */
   6652 	m->m_nextpkt = key_acquire_mbuf_head;
   6653 	key_acquire_mbuf_head = m;
   6654 	key_acquire_mbuf_count++;
   6655 	mutex_exit(&key_misc.lock);
   6656 
   6657 	/* Kick the timer */
   6658 	key_timehandler(NULL);
   6659 
   6660 	return 0;
   6661 }
   6662 
   6663 #ifndef IPSEC_NONBLOCK_ACQUIRE
   6664 static struct secacq *
   6665 key_newacq(const struct secasindex *saidx)
   6666 {
   6667 	struct secacq *newacq;
   6668 
   6669 	/* get new entry */
   6670 	newacq = kmem_intr_zalloc(sizeof(struct secacq), KM_NOSLEEP);
   6671 	if (newacq == NULL) {
   6672 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
   6673 		return NULL;
   6674 	}
   6675 
   6676 	/* copy secindex */
   6677 	memcpy(&newacq->saidx, saidx, sizeof(newacq->saidx));
   6678 	newacq->seq = (acq_seq == ~0 ? 1 : ++acq_seq);
   6679 	newacq->created = time_uptime;
   6680 	newacq->count = 0;
   6681 
   6682 	return newacq;
   6683 }
   6684 
   6685 static struct secacq *
   6686 key_getacq(const struct secasindex *saidx)
   6687 {
   6688 	struct secacq *acq;
   6689 
   6690 	KASSERT(mutex_owned(&key_misc.lock));
   6691 
   6692 	LIST_FOREACH(acq, &key_misc.acqlist, chain) {
   6693 		if (key_saidx_match(saidx, &acq->saidx, CMP_EXACTLY))
   6694 			return acq;
   6695 	}
   6696 
   6697 	return NULL;
   6698 }
   6699 
   6700 static struct secacq *
   6701 key_getacqbyseq(u_int32_t seq)
   6702 {
   6703 	struct secacq *acq;
   6704 
   6705 	KASSERT(mutex_owned(&key_misc.lock));
   6706 
   6707 	LIST_FOREACH(acq, &key_misc.acqlist, chain) {
   6708 		if (acq->seq == seq)
   6709 			return acq;
   6710 	}
   6711 
   6712 	return NULL;
   6713 }
   6714 #endif
   6715 
   6716 #ifdef notyet
   6717 static struct secspacq *
   6718 key_newspacq(const struct secpolicyindex *spidx)
   6719 {
   6720 	struct secspacq *acq;
   6721 
   6722 	/* get new entry */
   6723 	acq = kmem_intr_zalloc(sizeof(struct secspacq), KM_NOSLEEP);
   6724 	if (acq == NULL) {
   6725 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
   6726 		return NULL;
   6727 	}
   6728 
   6729 	/* copy secindex */
   6730 	memcpy(&acq->spidx, spidx, sizeof(acq->spidx));
   6731 	acq->created = time_uptime;
   6732 	acq->count = 0;
   6733 
   6734 	return acq;
   6735 }
   6736 
   6737 static struct secspacq *
   6738 key_getspacq(const struct secpolicyindex *spidx)
   6739 {
   6740 	struct secspacq *acq;
   6741 
   6742 	LIST_FOREACH(acq, &key_misc.spacqlist, chain) {
   6743 		if (key_spidx_match_exactly(spidx, &acq->spidx))
   6744 			return acq;
   6745 	}
   6746 
   6747 	return NULL;
   6748 }
   6749 #endif /* notyet */
   6750 
   6751 /*
   6752  * SADB_ACQUIRE processing,
   6753  * in first situation, is receiving
   6754  *   <base>
   6755  * from the ikmpd, and clear sequence of its secasvar entry.
   6756  *
   6757  * In second situation, is receiving
   6758  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
   6759  * from a user land process, and return
   6760  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
   6761  * to the socket.
   6762  *
   6763  * m will always be freed.
   6764  */
   6765 static int
   6766 key_api_acquire(struct socket *so, struct mbuf *m,
   6767       	     const struct sadb_msghdr *mhp)
   6768 {
   6769 	const struct sockaddr *src, *dst;
   6770 	struct secasindex saidx;
   6771 	u_int16_t proto;
   6772 	int error;
   6773 
   6774 	/*
   6775 	 * Error message from KMd.
   6776 	 * We assume that if error was occurred in IKEd, the length of PFKEY
   6777 	 * message is equal to the size of sadb_msg structure.
   6778 	 * We do not raise error even if error occurred in this function.
   6779 	 */
   6780 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
   6781 #ifndef IPSEC_NONBLOCK_ACQUIRE
   6782 		struct secacq *acq;
   6783 
   6784 		/* check sequence number */
   6785 		if (mhp->msg->sadb_msg_seq == 0) {
   6786 			IPSECLOG(LOG_DEBUG, "must specify sequence number.\n");
   6787 			m_freem(m);
   6788 			return 0;
   6789 		}
   6790 
   6791 		mutex_enter(&key_misc.lock);
   6792 		acq = key_getacqbyseq(mhp->msg->sadb_msg_seq);
   6793 		if (acq == NULL) {
   6794 			mutex_exit(&key_misc.lock);
   6795 			/*
   6796 			 * the specified larval SA is already gone, or we got
   6797 			 * a bogus sequence number.  we can silently ignore it.
   6798 			 */
   6799 			m_freem(m);
   6800 			return 0;
   6801 		}
   6802 
   6803 		/* reset acq counter in order to deletion by timehander. */
   6804 		acq->created = time_uptime;
   6805 		acq->count = 0;
   6806 		mutex_exit(&key_misc.lock);
   6807 #endif
   6808 		m_freem(m);
   6809 		return 0;
   6810 	}
   6811 
   6812 	/*
   6813 	 * This message is from user land.
   6814 	 */
   6815 
   6816 	/* map satype to proto */
   6817 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
   6818 	if (proto == 0) {
   6819 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   6820 		return key_senderror(so, m, EINVAL);
   6821 	}
   6822 
   6823 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   6824 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
   6825 	    mhp->ext[SADB_EXT_PROPOSAL] == NULL) {
   6826 		/* error */
   6827 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   6828 		return key_senderror(so, m, EINVAL);
   6829 	}
   6830 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   6831 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
   6832 	    mhp->extlen[SADB_EXT_PROPOSAL] < sizeof(struct sadb_prop)) {
   6833 		/* error */
   6834 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   6835 		return key_senderror(so, m, EINVAL);
   6836 	}
   6837 
   6838 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   6839 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   6840 
   6841 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
   6842 	if (error != 0)
   6843 		return key_senderror(so, m, EINVAL);
   6844 
   6845 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   6846 	if (error != 0)
   6847 		return key_senderror(so, m, EINVAL);
   6848 
   6849 	/* get a SA index */
   6850     {
   6851 	struct secashead *sah;
   6852 	int s = pserialize_read_enter();
   6853 
   6854 	sah = key_getsah(&saidx, CMP_MODE_REQID);
   6855 	if (sah != NULL) {
   6856 		pserialize_read_exit(s);
   6857 		IPSECLOG(LOG_DEBUG, "a SA exists already.\n");
   6858 		return key_senderror(so, m, EEXIST);
   6859 	}
   6860 	pserialize_read_exit(s);
   6861     }
   6862 
   6863 	error = key_acquire(&saidx, NULL, M_WAITOK);
   6864 	if (error != 0) {
   6865 		IPSECLOG(LOG_DEBUG, "error %d returned from key_acquire.\n",
   6866 		    error);
   6867 		return key_senderror(so, m, error);
   6868 	}
   6869 
   6870 	return key_sendup_mbuf(so, m, KEY_SENDUP_REGISTERED);
   6871 }
   6872 
   6873 /*
   6874  * SADB_REGISTER processing.
   6875  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
   6876  * receive
   6877  *   <base>
   6878  * from the ikmpd, and register a socket to send PF_KEY messages,
   6879  * and send
   6880  *   <base, supported>
   6881  * to KMD by PF_KEY.
   6882  * If socket is detached, must free from regnode.
   6883  *
   6884  * m will always be freed.
   6885  */
   6886 static int
   6887 key_api_register(struct socket *so, struct mbuf *m,
   6888 	     const struct sadb_msghdr *mhp)
   6889 {
   6890 	struct secreg *reg, *newreg = 0;
   6891 
   6892 	/* check for invalid register message */
   6893 	if (mhp->msg->sadb_msg_satype >= __arraycount(key_misc.reglist))
   6894 		return key_senderror(so, m, EINVAL);
   6895 
   6896 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
   6897 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
   6898 		goto setmsg;
   6899 
   6900 	/* Allocate regnode in advance, out of mutex */
   6901 	newreg = kmem_zalloc(sizeof(*newreg), KM_SLEEP);
   6902 
   6903 	/* check whether existing or not */
   6904 	mutex_enter(&key_misc.lock);
   6905 	LIST_FOREACH(reg, &key_misc.reglist[mhp->msg->sadb_msg_satype], chain) {
   6906 		if (reg->so == so) {
   6907 			IPSECLOG(LOG_DEBUG, "socket exists already.\n");
   6908 			mutex_exit(&key_misc.lock);
   6909 			kmem_free(newreg, sizeof(*newreg));
   6910 			return key_senderror(so, m, EEXIST);
   6911 		}
   6912 	}
   6913 
   6914 	newreg->so = so;
   6915 	((struct keycb *)sotorawcb(so))->kp_registered++;
   6916 
   6917 	/* add regnode to key_misc.reglist. */
   6918 	LIST_INSERT_HEAD(&key_misc.reglist[mhp->msg->sadb_msg_satype], newreg, chain);
   6919 	mutex_exit(&key_misc.lock);
   6920 
   6921   setmsg:
   6922     {
   6923 	struct mbuf *n;
   6924 	struct sadb_supported *sup;
   6925 	u_int len, alen, elen;
   6926 	int off;
   6927 	int i;
   6928 	struct sadb_alg *alg;
   6929 
   6930 	/* create new sadb_msg to reply. */
   6931 	alen = 0;
   6932 	for (i = 1; i <= SADB_AALG_MAX; i++) {
   6933 		if (ah_algorithm_lookup(i))
   6934 			alen += sizeof(struct sadb_alg);
   6935 	}
   6936 	if (alen)
   6937 		alen += sizeof(struct sadb_supported);
   6938 	elen = 0;
   6939 	for (i = 1; i <= SADB_EALG_MAX; i++) {
   6940 		if (esp_algorithm_lookup(i))
   6941 			elen += sizeof(struct sadb_alg);
   6942 	}
   6943 	if (elen)
   6944 		elen += sizeof(struct sadb_supported);
   6945 
   6946 	len = sizeof(struct sadb_msg) + alen + elen;
   6947 
   6948 	if (len > MCLBYTES)
   6949 		return key_senderror(so, m, ENOBUFS);
   6950 
   6951 	n = key_alloc_mbuf_simple(len, M_WAITOK);
   6952 	n->m_pkthdr.len = n->m_len = len;
   6953 	n->m_next = NULL;
   6954 	off = 0;
   6955 
   6956 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
   6957 	key_fill_replymsg(n, 0);
   6958 
   6959 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
   6960 
   6961 	/* for authentication algorithm */
   6962 	if (alen) {
   6963 		sup = (struct sadb_supported *)(mtod(n, char *) + off);
   6964 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
   6965 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
   6966 		off += PFKEY_ALIGN8(sizeof(*sup));
   6967 
   6968 		for (i = 1; i <= SADB_AALG_MAX; i++) {
   6969 			const struct auth_hash *aalgo;
   6970 			u_int16_t minkeysize, maxkeysize;
   6971 
   6972 			aalgo = ah_algorithm_lookup(i);
   6973 			if (!aalgo)
   6974 				continue;
   6975 			alg = (struct sadb_alg *)(mtod(n, char *) + off);
   6976 			alg->sadb_alg_id = i;
   6977 			alg->sadb_alg_ivlen = 0;
   6978 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
   6979 			alg->sadb_alg_minbits = _BITS(minkeysize);
   6980 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
   6981 			off += PFKEY_ALIGN8(sizeof(*alg));
   6982 		}
   6983 	}
   6984 
   6985 	/* for encryption algorithm */
   6986 	if (elen) {
   6987 		sup = (struct sadb_supported *)(mtod(n, char *) + off);
   6988 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
   6989 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
   6990 		off += PFKEY_ALIGN8(sizeof(*sup));
   6991 
   6992 		for (i = 1; i <= SADB_EALG_MAX; i++) {
   6993 			const struct enc_xform *ealgo;
   6994 
   6995 			ealgo = esp_algorithm_lookup(i);
   6996 			if (!ealgo)
   6997 				continue;
   6998 			alg = (struct sadb_alg *)(mtod(n, char *) + off);
   6999 			alg->sadb_alg_id = i;
   7000 			alg->sadb_alg_ivlen = ealgo->blocksize;
   7001 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
   7002 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
   7003 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
   7004 		}
   7005 	}
   7006 
   7007 	KASSERTMSG(off == len, "length inconsistency");
   7008 
   7009 	m_freem(m);
   7010 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
   7011     }
   7012 }
   7013 
   7014 /*
   7015  * free secreg entry registered.
   7016  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
   7017  */
   7018 void
   7019 key_freereg(struct socket *so)
   7020 {
   7021 	struct secreg *reg;
   7022 	int i;
   7023 
   7024 	KASSERT(!cpu_softintr_p());
   7025 	KASSERT(so != NULL);
   7026 
   7027 	/*
   7028 	 * check whether existing or not.
   7029 	 * check all type of SA, because there is a potential that
   7030 	 * one socket is registered to multiple type of SA.
   7031 	 */
   7032 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
   7033 		mutex_enter(&key_misc.lock);
   7034 		LIST_FOREACH(reg, &key_misc.reglist[i], chain) {
   7035 			if (reg->so == so) {
   7036 				LIST_REMOVE(reg, chain);
   7037 				break;
   7038 			}
   7039 		}
   7040 		mutex_exit(&key_misc.lock);
   7041 		if (reg != NULL)
   7042 			kmem_free(reg, sizeof(*reg));
   7043 	}
   7044 
   7045 	return;
   7046 }
   7047 
   7048 /*
   7049  * SADB_EXPIRE processing
   7050  * send
   7051  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
   7052  * to KMD by PF_KEY.
   7053  * NOTE: We send only soft lifetime extension.
   7054  *
   7055  * OUT:	0	: succeed
   7056  *	others	: error number
   7057  */
   7058 static int
   7059 key_expire(struct secasvar *sav)
   7060 {
   7061 	int s;
   7062 	int satype;
   7063 	struct mbuf *result = NULL, *m;
   7064 	int len;
   7065 	int error = -1;
   7066 	struct sadb_lifetime *lt;
   7067 
   7068 	/* XXX: Why do we lock ? */
   7069 	s = splsoftnet();	/*called from softclock()*/
   7070 
   7071 	KASSERT(sav != NULL);
   7072 
   7073 	satype = key_proto2satype(sav->sah->saidx.proto);
   7074 	KASSERTMSG(satype != 0, "invalid proto is passed");
   7075 
   7076 	/* set msg header */
   7077 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, key_sa_refcnt(sav),
   7078 	    M_WAITOK);
   7079 	result = m;
   7080 
   7081 	/* create SA extension */
   7082 	m = key_setsadbsa(sav);
   7083 	m_cat(result, m);
   7084 
   7085 	/* create SA extension */
   7086 	m = key_setsadbxsa2(sav->sah->saidx.mode,
   7087 	    sav->replay ? sav->replay->count : 0, sav->sah->saidx.reqid);
   7088 	m_cat(result, m);
   7089 
   7090 	/* create lifetime extension (current and soft) */
   7091 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
   7092 	m = key_alloc_mbuf(len, M_WAITOK);
   7093 	KASSERT(m->m_next == NULL);
   7094 
   7095 	memset(mtod(m, void *), 0, len);
   7096 	lt = mtod(m, struct sadb_lifetime *);
   7097 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
   7098 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
   7099 	lt->sadb_lifetime_allocations = sav->lft_c->sadb_lifetime_allocations;
   7100 	lt->sadb_lifetime_bytes = sav->lft_c->sadb_lifetime_bytes;
   7101 	lt->sadb_lifetime_addtime =
   7102 	    time_mono_to_wall(sav->lft_c->sadb_lifetime_addtime);
   7103 	lt->sadb_lifetime_usetime =
   7104 	    time_mono_to_wall(sav->lft_c->sadb_lifetime_usetime);
   7105 	lt = (struct sadb_lifetime *)(mtod(m, char *) + len / 2);
   7106 	memcpy(lt, sav->lft_s, sizeof(*lt));
   7107 	m_cat(result, m);
   7108 
   7109 	/* set sadb_address for source */
   7110 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa,
   7111 	    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
   7112 	m_cat(result, m);
   7113 
   7114 	/* set sadb_address for destination */
   7115 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.dst.sa,
   7116 	    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
   7117 	m_cat(result, m);
   7118 
   7119 	if ((result->m_flags & M_PKTHDR) == 0) {
   7120 		error = EINVAL;
   7121 		goto fail;
   7122 	}
   7123 
   7124 	if (result->m_len < sizeof(struct sadb_msg)) {
   7125 		result = m_pullup(result, sizeof(struct sadb_msg));
   7126 		if (result == NULL) {
   7127 			error = ENOBUFS;
   7128 			goto fail;
   7129 		}
   7130 	}
   7131 
   7132 	result->m_pkthdr.len = 0;
   7133 	for (m = result; m; m = m->m_next)
   7134 		result->m_pkthdr.len += m->m_len;
   7135 
   7136 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   7137 	    PFKEY_UNIT64(result->m_pkthdr.len);
   7138 
   7139 	splx(s);
   7140 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
   7141 
   7142  fail:
   7143 	if (result)
   7144 		m_freem(result);
   7145 	splx(s);
   7146 	return error;
   7147 }
   7148 
   7149 /*
   7150  * SADB_FLUSH processing
   7151  * receive
   7152  *   <base>
   7153  * from the ikmpd, and free all entries in secastree.
   7154  * and send,
   7155  *   <base>
   7156  * to the ikmpd.
   7157  * NOTE: to do is only marking SADB_SASTATE_DEAD.
   7158  *
   7159  * m will always be freed.
   7160  */
   7161 static int
   7162 key_api_flush(struct socket *so, struct mbuf *m,
   7163           const struct sadb_msghdr *mhp)
   7164 {
   7165 	struct sadb_msg *newmsg;
   7166 	struct secashead *sah;
   7167 	struct secasvar *sav;
   7168 	u_int16_t proto;
   7169 	u_int8_t state;
   7170 	int s;
   7171 
   7172 	/* map satype to proto */
   7173 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
   7174 	if (proto == 0) {
   7175 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   7176 		return key_senderror(so, m, EINVAL);
   7177 	}
   7178 
   7179 	/* no SATYPE specified, i.e. flushing all SA. */
   7180 	s = pserialize_read_enter();
   7181 	SAHLIST_READER_FOREACH(sah) {
   7182 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
   7183 		    proto != sah->saidx.proto)
   7184 			continue;
   7185 
   7186 		key_sah_ref(sah);
   7187 		pserialize_read_exit(s);
   7188 
   7189 		SASTATE_ALIVE_FOREACH(state) {
   7190 		restart:
   7191 			mutex_enter(&key_sad.lock);
   7192 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
   7193 				sav->state = SADB_SASTATE_DEAD;
   7194 				key_unlink_sav(sav);
   7195 				mutex_exit(&key_sad.lock);
   7196 				key_destroy_sav(sav);
   7197 				goto restart;
   7198 			}
   7199 			mutex_exit(&key_sad.lock);
   7200 		}
   7201 
   7202 		s = pserialize_read_enter();
   7203 		sah->state = SADB_SASTATE_DEAD;
   7204 		key_sah_unref(sah);
   7205 	}
   7206 	pserialize_read_exit(s);
   7207 
   7208 	if (m->m_len < sizeof(struct sadb_msg) ||
   7209 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
   7210 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
   7211 		return key_senderror(so, m, ENOBUFS);
   7212 	}
   7213 
   7214 	if (m->m_next)
   7215 		m_freem(m->m_next);
   7216 	m->m_next = NULL;
   7217 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
   7218 	newmsg = mtod(m, struct sadb_msg *);
   7219 	newmsg->sadb_msg_errno = 0;
   7220 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
   7221 
   7222 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
   7223 }
   7224 
   7225 
   7226 static struct mbuf *
   7227 key_setdump_chain(u_int8_t req_satype, int *errorp, int *lenp, pid_t pid)
   7228 {
   7229 	struct secashead *sah;
   7230 	struct secasvar *sav;
   7231 	u_int16_t proto;
   7232 	u_int8_t satype;
   7233 	u_int8_t state;
   7234 	int cnt;
   7235 	struct mbuf *m, *n, *prev;
   7236 
   7237 	KASSERT(mutex_owned(&key_sad.lock));
   7238 
   7239 	*lenp = 0;
   7240 
   7241 	/* map satype to proto */
   7242 	proto = key_satype2proto(req_satype);
   7243 	if (proto == 0) {
   7244 		*errorp = EINVAL;
   7245 		return (NULL);
   7246 	}
   7247 
   7248 	/* count sav entries to be sent to userland. */
   7249 	cnt = 0;
   7250 	SAHLIST_WRITER_FOREACH(sah) {
   7251 		if (req_satype != SADB_SATYPE_UNSPEC &&
   7252 		    proto != sah->saidx.proto)
   7253 			continue;
   7254 
   7255 		SASTATE_ANY_FOREACH(state) {
   7256 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
   7257 				cnt++;
   7258 			}
   7259 		}
   7260 	}
   7261 
   7262 	if (cnt == 0) {
   7263 		*errorp = ENOENT;
   7264 		return (NULL);
   7265 	}
   7266 
   7267 	/* send this to the userland, one at a time. */
   7268 	m = NULL;
   7269 	prev = m;
   7270 	SAHLIST_WRITER_FOREACH(sah) {
   7271 		if (req_satype != SADB_SATYPE_UNSPEC &&
   7272 		    proto != sah->saidx.proto)
   7273 			continue;
   7274 
   7275 		/* map proto to satype */
   7276 		satype = key_proto2satype(sah->saidx.proto);
   7277 		if (satype == 0) {
   7278 			m_freem(m);
   7279 			*errorp = EINVAL;
   7280 			return (NULL);
   7281 		}
   7282 
   7283 		SASTATE_ANY_FOREACH(state) {
   7284 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
   7285 				n = key_setdumpsa(sav, SADB_DUMP, satype,
   7286 				    --cnt, pid);
   7287 				if (!m)
   7288 					m = n;
   7289 				else
   7290 					prev->m_nextpkt = n;
   7291 				prev = n;
   7292 			}
   7293 		}
   7294 	}
   7295 
   7296 	if (!m) {
   7297 		*errorp = EINVAL;
   7298 		return (NULL);
   7299 	}
   7300 
   7301 	if ((m->m_flags & M_PKTHDR) != 0) {
   7302 		m->m_pkthdr.len = 0;
   7303 		for (n = m; n; n = n->m_next)
   7304 			m->m_pkthdr.len += n->m_len;
   7305 	}
   7306 
   7307 	*errorp = 0;
   7308 	return (m);
   7309 }
   7310 
   7311 /*
   7312  * SADB_DUMP processing
   7313  * dump all entries including status of DEAD in SAD.
   7314  * receive
   7315  *   <base>
   7316  * from the ikmpd, and dump all secasvar leaves
   7317  * and send,
   7318  *   <base> .....
   7319  * to the ikmpd.
   7320  *
   7321  * m will always be freed.
   7322  */
   7323 static int
   7324 key_api_dump(struct socket *so, struct mbuf *m0,
   7325 	 const struct sadb_msghdr *mhp)
   7326 {
   7327 	u_int16_t proto;
   7328 	u_int8_t satype;
   7329 	struct mbuf *n;
   7330 	int error, len, ok;
   7331 
   7332 	/* map satype to proto */
   7333 	satype = mhp->msg->sadb_msg_satype;
   7334 	proto = key_satype2proto(satype);
   7335 	if (proto == 0) {
   7336 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   7337 		return key_senderror(so, m0, EINVAL);
   7338 	}
   7339 
   7340 	/*
   7341 	 * If the requestor has insufficient socket-buffer space
   7342 	 * for the entire chain, nobody gets any response to the DUMP.
   7343 	 * XXX For now, only the requestor ever gets anything.
   7344 	 * Moreover, if the requestor has any space at all, they receive
   7345 	 * the entire chain, otherwise the request is refused with ENOBUFS.
   7346 	 */
   7347 	if (sbspace(&so->so_rcv) <= 0) {
   7348 		return key_senderror(so, m0, ENOBUFS);
   7349 	}
   7350 
   7351 	mutex_enter(&key_sad.lock);
   7352 	n = key_setdump_chain(satype, &error, &len, mhp->msg->sadb_msg_pid);
   7353 	mutex_exit(&key_sad.lock);
   7354 
   7355 	if (n == NULL) {
   7356 		return key_senderror(so, m0, ENOENT);
   7357 	}
   7358 	{
   7359 		uint64_t *ps = PFKEY_STAT_GETREF();
   7360 		ps[PFKEY_STAT_IN_TOTAL]++;
   7361 		ps[PFKEY_STAT_IN_BYTES] += len;
   7362 		PFKEY_STAT_PUTREF();
   7363 	}
   7364 
   7365 	/*
   7366 	 * PF_KEY DUMP responses are no longer broadcast to all PF_KEY sockets.
   7367 	 * The requestor receives either the entire chain, or an
   7368 	 * error message with ENOBUFS.
   7369 	 *
   7370 	 * sbappendaddrchain() takes the chain of entries, one
   7371 	 * packet-record per SPD entry, prepends the key_src sockaddr
   7372 	 * to each packet-record, links the sockaddr mbufs into a new
   7373 	 * list of records, then   appends the entire resulting
   7374 	 * list to the requesting socket.
   7375 	 */
   7376 	ok = sbappendaddrchain(&so->so_rcv, (struct sockaddr *)&key_src, n,
   7377 	    SB_PRIO_ONESHOT_OVERFLOW);
   7378 
   7379 	if (!ok) {
   7380 		PFKEY_STATINC(PFKEY_STAT_IN_NOMEM);
   7381 		m_freem(n);
   7382 		return key_senderror(so, m0, ENOBUFS);
   7383 	}
   7384 
   7385 	m_freem(m0);
   7386 	return 0;
   7387 }
   7388 
   7389 /*
   7390  * SADB_X_PROMISC processing
   7391  *
   7392  * m will always be freed.
   7393  */
   7394 static int
   7395 key_api_promisc(struct socket *so, struct mbuf *m,
   7396 	    const struct sadb_msghdr *mhp)
   7397 {
   7398 	int olen;
   7399 
   7400 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
   7401 
   7402 	if (olen < sizeof(struct sadb_msg)) {
   7403 #if 1
   7404 		return key_senderror(so, m, EINVAL);
   7405 #else
   7406 		m_freem(m);
   7407 		return 0;
   7408 #endif
   7409 	} else if (olen == sizeof(struct sadb_msg)) {
   7410 		/* enable/disable promisc mode */
   7411 		struct keycb *kp = (struct keycb *)sotorawcb(so);
   7412 		if (kp == NULL)
   7413 			return key_senderror(so, m, EINVAL);
   7414 		mhp->msg->sadb_msg_errno = 0;
   7415 		switch (mhp->msg->sadb_msg_satype) {
   7416 		case 0:
   7417 		case 1:
   7418 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
   7419 			break;
   7420 		default:
   7421 			return key_senderror(so, m, EINVAL);
   7422 		}
   7423 
   7424 		/* send the original message back to everyone */
   7425 		mhp->msg->sadb_msg_errno = 0;
   7426 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
   7427 	} else {
   7428 		/* send packet as is */
   7429 
   7430 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
   7431 
   7432 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
   7433 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
   7434 	}
   7435 }
   7436 
   7437 static int (*key_api_typesw[]) (struct socket *, struct mbuf *,
   7438 		const struct sadb_msghdr *) = {
   7439 	NULL,			/* SADB_RESERVED */
   7440 	key_api_getspi,		/* SADB_GETSPI */
   7441 	key_api_update,		/* SADB_UPDATE */
   7442 	key_api_add,		/* SADB_ADD */
   7443 	key_api_delete,		/* SADB_DELETE */
   7444 	key_api_get,		/* SADB_GET */
   7445 	key_api_acquire,	/* SADB_ACQUIRE */
   7446 	key_api_register,	/* SADB_REGISTER */
   7447 	NULL,			/* SADB_EXPIRE */
   7448 	key_api_flush,		/* SADB_FLUSH */
   7449 	key_api_dump,		/* SADB_DUMP */
   7450 	key_api_promisc,	/* SADB_X_PROMISC */
   7451 	NULL,			/* SADB_X_PCHANGE */
   7452 	key_api_spdadd,		/* SADB_X_SPDUPDATE */
   7453 	key_api_spdadd,		/* SADB_X_SPDADD */
   7454 	key_api_spddelete,	/* SADB_X_SPDDELETE */
   7455 	key_api_spdget,		/* SADB_X_SPDGET */
   7456 	NULL,			/* SADB_X_SPDACQUIRE */
   7457 	key_api_spddump,	/* SADB_X_SPDDUMP */
   7458 	key_api_spdflush,	/* SADB_X_SPDFLUSH */
   7459 	key_api_spdadd,		/* SADB_X_SPDSETIDX */
   7460 	NULL,			/* SADB_X_SPDEXPIRE */
   7461 	key_api_spddelete2,	/* SADB_X_SPDDELETE2 */
   7462 	key_api_nat_map,	/* SADB_X_NAT_T_NEW_MAPPING */
   7463 };
   7464 
   7465 /*
   7466  * parse sadb_msg buffer to process PFKEYv2,
   7467  * and create a data to response if needed.
   7468  * I think to be dealed with mbuf directly.
   7469  * IN:
   7470  *     msgp  : pointer to pointer to a received buffer pulluped.
   7471  *             This is rewrited to response.
   7472  *     so    : pointer to socket.
   7473  * OUT:
   7474  *    length for buffer to send to user process.
   7475  */
   7476 int
   7477 key_parse(struct mbuf *m, struct socket *so)
   7478 {
   7479 	struct sadb_msg *msg;
   7480 	struct sadb_msghdr mh;
   7481 	u_int orglen;
   7482 	int error;
   7483 
   7484 	KASSERT(m != NULL);
   7485 	KASSERT(so != NULL);
   7486 
   7487 #if 0	/*kdebug_sadb assumes msg in linear buffer*/
   7488 	if (KEYDEBUG_ON(KEYDEBUG_KEY_DUMP)) {
   7489 		kdebug_sadb("passed sadb_msg", msg);
   7490 	}
   7491 #endif
   7492 
   7493 	if (m->m_len < sizeof(struct sadb_msg)) {
   7494 		m = m_pullup(m, sizeof(struct sadb_msg));
   7495 		if (!m)
   7496 			return ENOBUFS;
   7497 	}
   7498 	msg = mtod(m, struct sadb_msg *);
   7499 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
   7500 
   7501 	if ((m->m_flags & M_PKTHDR) == 0 ||
   7502 	    m->m_pkthdr.len != orglen) {
   7503 		IPSECLOG(LOG_DEBUG, "invalid message length.\n");
   7504 		PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
   7505 		error = EINVAL;
   7506 		goto senderror;
   7507 	}
   7508 
   7509 	if (msg->sadb_msg_version != PF_KEY_V2) {
   7510 		IPSECLOG(LOG_DEBUG, "PF_KEY version %u is mismatched.\n",
   7511 		    msg->sadb_msg_version);
   7512 		PFKEY_STATINC(PFKEY_STAT_OUT_INVVER);
   7513 		error = EINVAL;
   7514 		goto senderror;
   7515 	}
   7516 
   7517 	if (msg->sadb_msg_type > SADB_MAX) {
   7518 		IPSECLOG(LOG_DEBUG, "invalid type %u is passed.\n",
   7519 		    msg->sadb_msg_type);
   7520 		PFKEY_STATINC(PFKEY_STAT_OUT_INVMSGTYPE);
   7521 		error = EINVAL;
   7522 		goto senderror;
   7523 	}
   7524 
   7525 	/* for old-fashioned code - should be nuked */
   7526 	if (m->m_pkthdr.len > MCLBYTES) {
   7527 		m_freem(m);
   7528 		return ENOBUFS;
   7529 	}
   7530 	if (m->m_next) {
   7531 		struct mbuf *n;
   7532 
   7533 		n = key_alloc_mbuf_simple(m->m_pkthdr.len, M_WAITOK);
   7534 
   7535 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, void *));
   7536 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
   7537 		n->m_next = NULL;
   7538 		m_freem(m);
   7539 		m = n;
   7540 	}
   7541 
   7542 	/* align the mbuf chain so that extensions are in contiguous region. */
   7543 	error = key_align(m, &mh);
   7544 	if (error)
   7545 		return error;
   7546 
   7547 	if (m->m_next) {	/*XXX*/
   7548 		m_freem(m);
   7549 		return ENOBUFS;
   7550 	}
   7551 
   7552 	msg = mh.msg;
   7553 
   7554 	/* check SA type */
   7555 	switch (msg->sadb_msg_satype) {
   7556 	case SADB_SATYPE_UNSPEC:
   7557 		switch (msg->sadb_msg_type) {
   7558 		case SADB_GETSPI:
   7559 		case SADB_UPDATE:
   7560 		case SADB_ADD:
   7561 		case SADB_DELETE:
   7562 		case SADB_GET:
   7563 		case SADB_ACQUIRE:
   7564 		case SADB_EXPIRE:
   7565 			IPSECLOG(LOG_DEBUG,
   7566 			    "must specify satype when msg type=%u.\n",
   7567 			    msg->sadb_msg_type);
   7568 			PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
   7569 			error = EINVAL;
   7570 			goto senderror;
   7571 		}
   7572 		break;
   7573 	case SADB_SATYPE_AH:
   7574 	case SADB_SATYPE_ESP:
   7575 	case SADB_X_SATYPE_IPCOMP:
   7576 	case SADB_X_SATYPE_TCPSIGNATURE:
   7577 		switch (msg->sadb_msg_type) {
   7578 		case SADB_X_SPDADD:
   7579 		case SADB_X_SPDDELETE:
   7580 		case SADB_X_SPDGET:
   7581 		case SADB_X_SPDDUMP:
   7582 		case SADB_X_SPDFLUSH:
   7583 		case SADB_X_SPDSETIDX:
   7584 		case SADB_X_SPDUPDATE:
   7585 		case SADB_X_SPDDELETE2:
   7586 			IPSECLOG(LOG_DEBUG, "illegal satype=%u\n",
   7587 			    msg->sadb_msg_type);
   7588 			PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
   7589 			error = EINVAL;
   7590 			goto senderror;
   7591 		}
   7592 		break;
   7593 	case SADB_SATYPE_RSVP:
   7594 	case SADB_SATYPE_OSPFV2:
   7595 	case SADB_SATYPE_RIPV2:
   7596 	case SADB_SATYPE_MIP:
   7597 		IPSECLOG(LOG_DEBUG, "type %u isn't supported.\n",
   7598 		    msg->sadb_msg_satype);
   7599 		PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
   7600 		error = EOPNOTSUPP;
   7601 		goto senderror;
   7602 	case 1:	/* XXX: What does it do? */
   7603 		if (msg->sadb_msg_type == SADB_X_PROMISC)
   7604 			break;
   7605 		/*FALLTHROUGH*/
   7606 	default:
   7607 		IPSECLOG(LOG_DEBUG, "invalid type %u is passed.\n",
   7608 		    msg->sadb_msg_satype);
   7609 		PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
   7610 		error = EINVAL;
   7611 		goto senderror;
   7612 	}
   7613 
   7614 	/* check field of upper layer protocol and address family */
   7615 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL &&
   7616 	    mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
   7617 		const struct sadb_address *src0, *dst0;
   7618 		const struct sockaddr *sa0, *da0;
   7619 		u_int plen;
   7620 
   7621 		src0 = mh.ext[SADB_EXT_ADDRESS_SRC];
   7622 		dst0 = mh.ext[SADB_EXT_ADDRESS_DST];
   7623 		sa0 = key_msghdr_get_sockaddr(&mh, SADB_EXT_ADDRESS_SRC);
   7624 		da0 = key_msghdr_get_sockaddr(&mh, SADB_EXT_ADDRESS_DST);
   7625 
   7626 		/* check upper layer protocol */
   7627 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
   7628 			IPSECLOG(LOG_DEBUG,
   7629 			    "upper layer protocol mismatched.\n");
   7630 			goto invaddr;
   7631 		}
   7632 
   7633 		/* check family */
   7634 		if (sa0->sa_family != da0->sa_family) {
   7635 			IPSECLOG(LOG_DEBUG, "address family mismatched.\n");
   7636 			goto invaddr;
   7637 		}
   7638 		if (sa0->sa_len != da0->sa_len) {
   7639 			IPSECLOG(LOG_DEBUG,
   7640 			    "address struct size mismatched.\n");
   7641 			goto invaddr;
   7642 		}
   7643 
   7644 		switch (sa0->sa_family) {
   7645 		case AF_INET:
   7646 			if (sa0->sa_len != sizeof(struct sockaddr_in))
   7647 				goto invaddr;
   7648 			break;
   7649 		case AF_INET6:
   7650 			if (sa0->sa_len != sizeof(struct sockaddr_in6))
   7651 				goto invaddr;
   7652 			break;
   7653 		default:
   7654 			IPSECLOG(LOG_DEBUG, "unsupported address family.\n");
   7655 			error = EAFNOSUPPORT;
   7656 			goto senderror;
   7657 		}
   7658 
   7659 		switch (sa0->sa_family) {
   7660 		case AF_INET:
   7661 			plen = sizeof(struct in_addr) << 3;
   7662 			break;
   7663 		case AF_INET6:
   7664 			plen = sizeof(struct in6_addr) << 3;
   7665 			break;
   7666 		default:
   7667 			plen = 0;	/*fool gcc*/
   7668 			break;
   7669 		}
   7670 
   7671 		/* check max prefix length */
   7672 		if (src0->sadb_address_prefixlen > plen ||
   7673 		    dst0->sadb_address_prefixlen > plen) {
   7674 			IPSECLOG(LOG_DEBUG, "illegal prefixlen.\n");
   7675 			goto invaddr;
   7676 		}
   7677 
   7678 		/*
   7679 		 * prefixlen == 0 is valid because there can be a case when
   7680 		 * all addresses are matched.
   7681 		 */
   7682 	}
   7683 
   7684 	if (msg->sadb_msg_type >= __arraycount(key_api_typesw) ||
   7685 	    key_api_typesw[msg->sadb_msg_type] == NULL) {
   7686 		PFKEY_STATINC(PFKEY_STAT_OUT_INVMSGTYPE);
   7687 		error = EINVAL;
   7688 		goto senderror;
   7689 	}
   7690 
   7691 	return (*key_api_typesw[msg->sadb_msg_type])(so, m, &mh);
   7692 
   7693 invaddr:
   7694 	error = EINVAL;
   7695 senderror:
   7696 	PFKEY_STATINC(PFKEY_STAT_OUT_INVADDR);
   7697 	return key_senderror(so, m, error);
   7698 }
   7699 
   7700 static int
   7701 key_senderror(struct socket *so, struct mbuf *m, int code)
   7702 {
   7703 	struct sadb_msg *msg;
   7704 
   7705 	KASSERT(m->m_len >= sizeof(struct sadb_msg));
   7706 
   7707 	msg = mtod(m, struct sadb_msg *);
   7708 	msg->sadb_msg_errno = code;
   7709 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
   7710 }
   7711 
   7712 /*
   7713  * set the pointer to each header into message buffer.
   7714  * m will be freed on error.
   7715  * XXX larger-than-MCLBYTES extension?
   7716  */
   7717 static int
   7718 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
   7719 {
   7720 	struct mbuf *n;
   7721 	struct sadb_ext *ext;
   7722 	size_t off, end;
   7723 	int extlen;
   7724 	int toff;
   7725 
   7726 	KASSERT(m != NULL);
   7727 	KASSERT(mhp != NULL);
   7728 	KASSERT(m->m_len >= sizeof(struct sadb_msg));
   7729 
   7730 	/* initialize */
   7731 	memset(mhp, 0, sizeof(*mhp));
   7732 
   7733 	mhp->msg = mtod(m, struct sadb_msg *);
   7734 	mhp->ext[0] = mhp->msg;	/*XXX backward compat */
   7735 
   7736 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
   7737 	extlen = end;	/*just in case extlen is not updated*/
   7738 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
   7739 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
   7740 		if (!n) {
   7741 			/* m is already freed */
   7742 			return ENOBUFS;
   7743 		}
   7744 		ext = (struct sadb_ext *)(mtod(n, char *) + toff);
   7745 
   7746 		/* set pointer */
   7747 		switch (ext->sadb_ext_type) {
   7748 		case SADB_EXT_SA:
   7749 		case SADB_EXT_ADDRESS_SRC:
   7750 		case SADB_EXT_ADDRESS_DST:
   7751 		case SADB_EXT_ADDRESS_PROXY:
   7752 		case SADB_EXT_LIFETIME_CURRENT:
   7753 		case SADB_EXT_LIFETIME_HARD:
   7754 		case SADB_EXT_LIFETIME_SOFT:
   7755 		case SADB_EXT_KEY_AUTH:
   7756 		case SADB_EXT_KEY_ENCRYPT:
   7757 		case SADB_EXT_IDENTITY_SRC:
   7758 		case SADB_EXT_IDENTITY_DST:
   7759 		case SADB_EXT_SENSITIVITY:
   7760 		case SADB_EXT_PROPOSAL:
   7761 		case SADB_EXT_SUPPORTED_AUTH:
   7762 		case SADB_EXT_SUPPORTED_ENCRYPT:
   7763 		case SADB_EXT_SPIRANGE:
   7764 		case SADB_X_EXT_POLICY:
   7765 		case SADB_X_EXT_SA2:
   7766 		case SADB_X_EXT_NAT_T_TYPE:
   7767 		case SADB_X_EXT_NAT_T_SPORT:
   7768 		case SADB_X_EXT_NAT_T_DPORT:
   7769 		case SADB_X_EXT_NAT_T_OAI:
   7770 		case SADB_X_EXT_NAT_T_OAR:
   7771 		case SADB_X_EXT_NAT_T_FRAG:
   7772 			/* duplicate check */
   7773 			/*
   7774 			 * XXX Are there duplication payloads of either
   7775 			 * KEY_AUTH or KEY_ENCRYPT ?
   7776 			 */
   7777 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
   7778 				IPSECLOG(LOG_DEBUG,
   7779 				    "duplicate ext_type %u is passed.\n",
   7780 				    ext->sadb_ext_type);
   7781 				m_freem(m);
   7782 				PFKEY_STATINC(PFKEY_STAT_OUT_DUPEXT);
   7783 				return EINVAL;
   7784 			}
   7785 			break;
   7786 		default:
   7787 			IPSECLOG(LOG_DEBUG, "invalid ext_type %u is passed.\n",
   7788 			    ext->sadb_ext_type);
   7789 			m_freem(m);
   7790 			PFKEY_STATINC(PFKEY_STAT_OUT_INVEXTTYPE);
   7791 			return EINVAL;
   7792 		}
   7793 
   7794 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
   7795 
   7796 		if (key_validate_ext(ext, extlen)) {
   7797 			m_freem(m);
   7798 			PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
   7799 			return EINVAL;
   7800 		}
   7801 
   7802 		n = m_pulldown(m, off, extlen, &toff);
   7803 		if (!n) {
   7804 			/* m is already freed */
   7805 			return ENOBUFS;
   7806 		}
   7807 		ext = (struct sadb_ext *)(mtod(n, char *) + toff);
   7808 
   7809 		mhp->ext[ext->sadb_ext_type] = ext;
   7810 		mhp->extoff[ext->sadb_ext_type] = off;
   7811 		mhp->extlen[ext->sadb_ext_type] = extlen;
   7812 	}
   7813 
   7814 	if (off != end) {
   7815 		m_freem(m);
   7816 		PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
   7817 		return EINVAL;
   7818 	}
   7819 
   7820 	return 0;
   7821 }
   7822 
   7823 static int
   7824 key_validate_ext(const struct sadb_ext *ext, int len)
   7825 {
   7826 	const struct sockaddr *sa;
   7827 	enum { NONE, ADDR } checktype = NONE;
   7828 	int baselen = 0;
   7829 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
   7830 
   7831 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
   7832 		return EINVAL;
   7833 
   7834 	/* if it does not match minimum/maximum length, bail */
   7835 	if (ext->sadb_ext_type >= __arraycount(minsize) ||
   7836 	    ext->sadb_ext_type >= __arraycount(maxsize))
   7837 		return EINVAL;
   7838 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
   7839 		return EINVAL;
   7840 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
   7841 		return EINVAL;
   7842 
   7843 	/* more checks based on sadb_ext_type XXX need more */
   7844 	switch (ext->sadb_ext_type) {
   7845 	case SADB_EXT_ADDRESS_SRC:
   7846 	case SADB_EXT_ADDRESS_DST:
   7847 	case SADB_EXT_ADDRESS_PROXY:
   7848 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
   7849 		checktype = ADDR;
   7850 		break;
   7851 	case SADB_EXT_IDENTITY_SRC:
   7852 	case SADB_EXT_IDENTITY_DST:
   7853 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
   7854 		    SADB_X_IDENTTYPE_ADDR) {
   7855 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
   7856 			checktype = ADDR;
   7857 		} else
   7858 			checktype = NONE;
   7859 		break;
   7860 	default:
   7861 		checktype = NONE;
   7862 		break;
   7863 	}
   7864 
   7865 	switch (checktype) {
   7866 	case NONE:
   7867 		break;
   7868 	case ADDR:
   7869 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
   7870 		if (len < baselen + sal)
   7871 			return EINVAL;
   7872 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
   7873 			return EINVAL;
   7874 		break;
   7875 	}
   7876 
   7877 	return 0;
   7878 }
   7879 
   7880 static int
   7881 key_do_init(void)
   7882 {
   7883 	int i, error;
   7884 
   7885 	mutex_init(&key_misc.lock, MUTEX_DEFAULT, IPL_NONE);
   7886 
   7887 	mutex_init(&key_spd.lock, MUTEX_DEFAULT, IPL_NONE);
   7888 	cv_init(&key_spd.cv_lc, "key_sp_lc");
   7889 	key_spd.psz = pserialize_create();
   7890 	cv_init(&key_spd.cv_psz, "key_sp_psz");
   7891 	key_spd.psz_performing = false;
   7892 
   7893 	mutex_init(&key_sad.lock, MUTEX_DEFAULT, IPL_NONE);
   7894 	cv_init(&key_sad.cv_lc, "key_sa_lc");
   7895 	key_sad.psz = pserialize_create();
   7896 	cv_init(&key_sad.cv_psz, "key_sa_psz");
   7897 	key_sad.psz_performing = false;
   7898 
   7899 	pfkeystat_percpu = percpu_alloc(sizeof(uint64_t) * PFKEY_NSTATS);
   7900 
   7901 	callout_init(&key_timehandler_ch, CALLOUT_MPSAFE);
   7902 	error = workqueue_create(&key_timehandler_wq, "key_timehandler",
   7903 	    key_timehandler_work, NULL, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
   7904 	if (error != 0)
   7905 		panic("%s: workqueue_create failed (%d)\n", __func__, error);
   7906 
   7907 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
   7908 		PSLIST_INIT(&key_spd.splist[i]);
   7909 	}
   7910 
   7911 	PSLIST_INIT(&key_spd.socksplist);
   7912 
   7913 	PSLIST_INIT(&key_sad.sahlist);
   7914 
   7915 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
   7916 		LIST_INIT(&key_misc.reglist[i]);
   7917 	}
   7918 
   7919 #ifndef IPSEC_NONBLOCK_ACQUIRE
   7920 	LIST_INIT(&key_misc.acqlist);
   7921 #endif
   7922 #ifdef notyet
   7923 	LIST_INIT(&key_misc.spacqlist);
   7924 #endif
   7925 
   7926 	/* system default */
   7927 	ip4_def_policy.policy = IPSEC_POLICY_NONE;
   7928 	ip4_def_policy.state = IPSEC_SPSTATE_ALIVE;
   7929 	localcount_init(&ip4_def_policy.localcount);
   7930 
   7931 #ifdef INET6
   7932 	ip6_def_policy.policy = IPSEC_POLICY_NONE;
   7933 	ip6_def_policy.state = IPSEC_SPSTATE_ALIVE;
   7934 	localcount_init(&ip6_def_policy.localcount);
   7935 #endif
   7936 
   7937 	callout_reset(&key_timehandler_ch, hz, key_timehandler, NULL);
   7938 
   7939 	/* initialize key statistics */
   7940 	keystat.getspi_count = 1;
   7941 
   7942 	aprint_verbose("IPsec: Initialized Security Association Processing.\n");
   7943 
   7944 	return (0);
   7945 }
   7946 
   7947 void
   7948 key_init(void)
   7949 {
   7950 	static ONCE_DECL(key_init_once);
   7951 
   7952 	sysctl_net_keyv2_setup(NULL);
   7953 	sysctl_net_key_compat_setup(NULL);
   7954 
   7955 	RUN_ONCE(&key_init_once, key_do_init);
   7956 
   7957 	key_init_so();
   7958 }
   7959 
   7960 /*
   7961  * XXX: maybe This function is called after INBOUND IPsec processing.
   7962  *
   7963  * Special check for tunnel-mode packets.
   7964  * We must make some checks for consistency between inner and outer IP header.
   7965  *
   7966  * xxx more checks to be provided
   7967  */
   7968 int
   7969 key_checktunnelsanity(
   7970     struct secasvar *sav,
   7971     u_int family,
   7972     void *src,
   7973     void *dst
   7974 )
   7975 {
   7976 
   7977 	/* XXX: check inner IP header */
   7978 
   7979 	return 1;
   7980 }
   7981 
   7982 #if 0
   7983 #define hostnamelen	strlen(hostname)
   7984 
   7985 /*
   7986  * Get FQDN for the host.
   7987  * If the administrator configured hostname (by hostname(1)) without
   7988  * domain name, returns nothing.
   7989  */
   7990 static const char *
   7991 key_getfqdn(void)
   7992 {
   7993 	int i;
   7994 	int hasdot;
   7995 	static char fqdn[MAXHOSTNAMELEN + 1];
   7996 
   7997 	if (!hostnamelen)
   7998 		return NULL;
   7999 
   8000 	/* check if it comes with domain name. */
   8001 	hasdot = 0;
   8002 	for (i = 0; i < hostnamelen; i++) {
   8003 		if (hostname[i] == '.')
   8004 			hasdot++;
   8005 	}
   8006 	if (!hasdot)
   8007 		return NULL;
   8008 
   8009 	/* NOTE: hostname may not be NUL-terminated. */
   8010 	memset(fqdn, 0, sizeof(fqdn));
   8011 	memcpy(fqdn, hostname, hostnamelen);
   8012 	fqdn[hostnamelen] = '\0';
   8013 	return fqdn;
   8014 }
   8015 
   8016 /*
   8017  * get username@FQDN for the host/user.
   8018  */
   8019 static const char *
   8020 key_getuserfqdn(void)
   8021 {
   8022 	const char *host;
   8023 	static char userfqdn[MAXHOSTNAMELEN + MAXLOGNAME + 2];
   8024 	struct proc *p = curproc;
   8025 	char *q;
   8026 
   8027 	if (!p || !p->p_pgrp || !p->p_pgrp->pg_session)
   8028 		return NULL;
   8029 	if (!(host = key_getfqdn()))
   8030 		return NULL;
   8031 
   8032 	/* NOTE: s_login may not be-NUL terminated. */
   8033 	memset(userfqdn, 0, sizeof(userfqdn));
   8034 	memcpy(userfqdn, Mp->p_pgrp->pg_session->s_login, AXLOGNAME);
   8035 	userfqdn[MAXLOGNAME] = '\0';	/* safeguard */
   8036 	q = userfqdn + strlen(userfqdn);
   8037 	*q++ = '@';
   8038 	memcpy(q, host, strlen(host));
   8039 	q += strlen(host);
   8040 	*q++ = '\0';
   8041 
   8042 	return userfqdn;
   8043 }
   8044 #endif
   8045 
   8046 /* record data transfer on SA, and update timestamps */
   8047 void
   8048 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
   8049 {
   8050 
   8051 	KASSERT(sav != NULL);
   8052 	KASSERT(sav->lft_c != NULL);
   8053 	KASSERT(m != NULL);
   8054 
   8055 	/*
   8056 	 * XXX Currently, there is a difference of bytes size
   8057 	 * between inbound and outbound processing.
   8058 	 */
   8059 	sav->lft_c->sadb_lifetime_bytes += m->m_pkthdr.len;
   8060 	/* to check bytes lifetime is done in key_timehandler(). */
   8061 
   8062 	/*
   8063 	 * We use the number of packets as the unit of
   8064 	 * sadb_lifetime_allocations.  We increment the variable
   8065 	 * whenever {esp,ah}_{in,out}put is called.
   8066 	 */
   8067 	sav->lft_c->sadb_lifetime_allocations++;
   8068 	/* XXX check for expires? */
   8069 
   8070 	/*
   8071 	 * NOTE: We record CURRENT sadb_lifetime_usetime by using wall clock,
   8072 	 * in seconds.  HARD and SOFT lifetime are measured by the time
   8073 	 * difference (again in seconds) from sadb_lifetime_usetime.
   8074 	 *
   8075 	 *	usetime
   8076 	 *	v     expire   expire
   8077 	 * -----+-----+--------+---> t
   8078 	 *	<--------------> HARD
   8079 	 *	<-----> SOFT
   8080 	 */
   8081 	sav->lft_c->sadb_lifetime_usetime = time_uptime;
   8082 	/* XXX check for expires? */
   8083 
   8084 	return;
   8085 }
   8086 
   8087 /* dumb version */
   8088 void
   8089 key_sa_routechange(struct sockaddr *dst)
   8090 {
   8091 	struct secashead *sah;
   8092 	int s;
   8093 
   8094 	s = pserialize_read_enter();
   8095 	SAHLIST_READER_FOREACH(sah) {
   8096 		struct route *ro;
   8097 		const struct sockaddr *sa;
   8098 
   8099 		key_sah_ref(sah);
   8100 		pserialize_read_exit(s);
   8101 
   8102 		ro = &sah->sa_route;
   8103 		sa = rtcache_getdst(ro);
   8104 		if (sa != NULL && dst->sa_len == sa->sa_len &&
   8105 		    memcmp(dst, sa, dst->sa_len) == 0)
   8106 			rtcache_free(ro);
   8107 
   8108 		s = pserialize_read_enter();
   8109 		key_sah_unref(sah);
   8110 	}
   8111 	pserialize_read_exit(s);
   8112 
   8113 	return;
   8114 }
   8115 
   8116 static void
   8117 key_sa_chgstate(struct secasvar *sav, u_int8_t state)
   8118 {
   8119 	struct secasvar *_sav;
   8120 
   8121 	ASSERT_SLEEPABLE();
   8122 	KASSERT(mutex_owned(&key_sad.lock));
   8123 
   8124 	if (sav->state == state)
   8125 		return;
   8126 
   8127 	key_unlink_sav(sav);
   8128 	localcount_fini(&sav->localcount);
   8129 	SAVLIST_ENTRY_DESTROY(sav);
   8130 	key_init_sav(sav);
   8131 
   8132 	sav->state = state;
   8133 	if (!SADB_SASTATE_USABLE_P(sav)) {
   8134 		/* We don't need to care about the order */
   8135 		SAVLIST_WRITER_INSERT_HEAD(sav->sah, state, sav);
   8136 		return;
   8137 	}
   8138 	/*
   8139 	 * Sort the list by lft_c->sadb_lifetime_addtime
   8140 	 * in ascending order.
   8141 	 */
   8142 	SAVLIST_READER_FOREACH(_sav, sav->sah, state) {
   8143 		if (_sav->lft_c->sadb_lifetime_addtime >
   8144 		    sav->lft_c->sadb_lifetime_addtime) {
   8145 			SAVLIST_WRITER_INSERT_BEFORE(_sav, sav);
   8146 			break;
   8147 		}
   8148 	}
   8149 	if (_sav == NULL) {
   8150 		SAVLIST_WRITER_INSERT_TAIL(sav->sah, state, sav);
   8151 	}
   8152 	key_validate_savlist(sav->sah, state);
   8153 }
   8154 
   8155 /* XXX too much? */
   8156 static struct mbuf *
   8157 key_alloc_mbuf(int l, int mflag)
   8158 {
   8159 	struct mbuf *m = NULL, *n;
   8160 	int len, t;
   8161 
   8162 	KASSERT(mflag == M_NOWAIT || (mflag == M_WAITOK && !cpu_softintr_p()));
   8163 
   8164 	len = l;
   8165 	while (len > 0) {
   8166 		MGET(n, mflag, MT_DATA);
   8167 		if (n && len > MLEN) {
   8168 			MCLGET(n, mflag);
   8169 			if ((n->m_flags & M_EXT) == 0) {
   8170 				m_freem(n);
   8171 				n = NULL;
   8172 			}
   8173 		}
   8174 		if (!n) {
   8175 			m_freem(m);
   8176 			return NULL;
   8177 		}
   8178 
   8179 		n->m_next = NULL;
   8180 		n->m_len = 0;
   8181 		n->m_len = M_TRAILINGSPACE(n);
   8182 		/* use the bottom of mbuf, hoping we can prepend afterwards */
   8183 		if (n->m_len > len) {
   8184 			t = (n->m_len - len) & ~(sizeof(long) - 1);
   8185 			n->m_data += t;
   8186 			n->m_len = len;
   8187 		}
   8188 
   8189 		len -= n->m_len;
   8190 
   8191 		if (m)
   8192 			m_cat(m, n);
   8193 		else
   8194 			m = n;
   8195 	}
   8196 
   8197 	return m;
   8198 }
   8199 
   8200 static struct mbuf *
   8201 key_setdump(u_int8_t req_satype, int *errorp, uint32_t pid)
   8202 {
   8203 	struct secashead *sah;
   8204 	struct secasvar *sav;
   8205 	u_int16_t proto;
   8206 	u_int8_t satype;
   8207 	u_int8_t state;
   8208 	int cnt;
   8209 	struct mbuf *m, *n;
   8210 
   8211 	KASSERT(mutex_owned(&key_sad.lock));
   8212 
   8213 	/* map satype to proto */
   8214 	proto = key_satype2proto(req_satype);
   8215 	if (proto == 0) {
   8216 		*errorp = EINVAL;
   8217 		return (NULL);
   8218 	}
   8219 
   8220 	/* count sav entries to be sent to the userland. */
   8221 	cnt = 0;
   8222 	SAHLIST_WRITER_FOREACH(sah) {
   8223 		if (req_satype != SADB_SATYPE_UNSPEC &&
   8224 		    proto != sah->saidx.proto)
   8225 			continue;
   8226 
   8227 		SASTATE_ANY_FOREACH(state) {
   8228 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
   8229 				cnt++;
   8230 			}
   8231 		}
   8232 	}
   8233 
   8234 	if (cnt == 0) {
   8235 		*errorp = ENOENT;
   8236 		return (NULL);
   8237 	}
   8238 
   8239 	/* send this to the userland, one at a time. */
   8240 	m = NULL;
   8241 	SAHLIST_WRITER_FOREACH(sah) {
   8242 		if (req_satype != SADB_SATYPE_UNSPEC &&
   8243 		    proto != sah->saidx.proto)
   8244 			continue;
   8245 
   8246 		/* map proto to satype */
   8247 		satype = key_proto2satype(sah->saidx.proto);
   8248 		if (satype == 0) {
   8249 			m_freem(m);
   8250 			*errorp = EINVAL;
   8251 			return (NULL);
   8252 		}
   8253 
   8254 		SASTATE_ANY_FOREACH(state) {
   8255 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
   8256 				n = key_setdumpsa(sav, SADB_DUMP, satype,
   8257 				    --cnt, pid);
   8258 				if (!m)
   8259 					m = n;
   8260 				else
   8261 					m_cat(m, n);
   8262 			}
   8263 		}
   8264 	}
   8265 
   8266 	if (!m) {
   8267 		*errorp = EINVAL;
   8268 		return (NULL);
   8269 	}
   8270 
   8271 	if ((m->m_flags & M_PKTHDR) != 0) {
   8272 		m->m_pkthdr.len = 0;
   8273 		for (n = m; n; n = n->m_next)
   8274 			m->m_pkthdr.len += n->m_len;
   8275 	}
   8276 
   8277 	*errorp = 0;
   8278 	return (m);
   8279 }
   8280 
   8281 static struct mbuf *
   8282 key_setspddump(int *errorp, pid_t pid)
   8283 {
   8284 	struct secpolicy *sp;
   8285 	int cnt;
   8286 	u_int dir;
   8287 	struct mbuf *m, *n;
   8288 
   8289 	KASSERT(mutex_owned(&key_spd.lock));
   8290 
   8291 	/* search SPD entry and get buffer size. */
   8292 	cnt = 0;
   8293 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
   8294 		SPLIST_WRITER_FOREACH(sp, dir) {
   8295 			cnt++;
   8296 		}
   8297 	}
   8298 
   8299 	if (cnt == 0) {
   8300 		*errorp = ENOENT;
   8301 		return (NULL);
   8302 	}
   8303 
   8304 	m = NULL;
   8305 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
   8306 		SPLIST_WRITER_FOREACH(sp, dir) {
   8307 			--cnt;
   8308 			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, pid);
   8309 
   8310 			if (!m)
   8311 				m = n;
   8312 			else {
   8313 				m->m_pkthdr.len += n->m_pkthdr.len;
   8314 				m_cat(m, n);
   8315 			}
   8316 		}
   8317 	}
   8318 
   8319 	*errorp = 0;
   8320 	return (m);
   8321 }
   8322 
   8323 int
   8324 key_get_used(void) {
   8325 	return !SPLIST_READER_EMPTY(IPSEC_DIR_INBOUND) ||
   8326 	    !SPLIST_READER_EMPTY(IPSEC_DIR_OUTBOUND) ||
   8327 	    !SOCKSPLIST_READER_EMPTY();
   8328 }
   8329 
   8330 void
   8331 key_update_used(void)
   8332 {
   8333 	switch (ipsec_enabled) {
   8334 	default:
   8335 	case 0:
   8336 #ifdef notyet
   8337 		/* XXX: racy */
   8338 		ipsec_used = 0;
   8339 #endif
   8340 		break;
   8341 	case 1:
   8342 #ifndef notyet
   8343 		/* XXX: racy */
   8344 		if (!ipsec_used)
   8345 #endif
   8346 		ipsec_used = key_get_used();
   8347 		break;
   8348 	case 2:
   8349 		ipsec_used = 1;
   8350 		break;
   8351 	}
   8352 }
   8353 
   8354 static int
   8355 sysctl_net_key_dumpsa(SYSCTLFN_ARGS)
   8356 {
   8357 	struct mbuf *m, *n;
   8358 	int err2 = 0;
   8359 	char *p, *ep;
   8360 	size_t len;
   8361 	int error;
   8362 
   8363 	if (newp)
   8364 		return (EPERM);
   8365 	if (namelen != 1)
   8366 		return (EINVAL);
   8367 
   8368 	mutex_enter(&key_sad.lock);
   8369 	m = key_setdump(name[0], &error, l->l_proc->p_pid);
   8370 	mutex_exit(&key_sad.lock);
   8371 	if (!m)
   8372 		return (error);
   8373 	if (!oldp)
   8374 		*oldlenp = m->m_pkthdr.len;
   8375 	else {
   8376 		p = oldp;
   8377 		if (*oldlenp < m->m_pkthdr.len) {
   8378 			err2 = ENOMEM;
   8379 			ep = p + *oldlenp;
   8380 		} else {
   8381 			*oldlenp = m->m_pkthdr.len;
   8382 			ep = p + m->m_pkthdr.len;
   8383 		}
   8384 		for (n = m; n; n = n->m_next) {
   8385 			len =  (ep - p < n->m_len) ?
   8386 				ep - p : n->m_len;
   8387 			error = copyout(mtod(n, const void *), p, len);
   8388 			p += len;
   8389 			if (error)
   8390 				break;
   8391 		}
   8392 		if (error == 0)
   8393 			error = err2;
   8394 	}
   8395 	m_freem(m);
   8396 
   8397 	return (error);
   8398 }
   8399 
   8400 static int
   8401 sysctl_net_key_dumpsp(SYSCTLFN_ARGS)
   8402 {
   8403 	struct mbuf *m, *n;
   8404 	int err2 = 0;
   8405 	char *p, *ep;
   8406 	size_t len;
   8407 	int error;
   8408 
   8409 	if (newp)
   8410 		return (EPERM);
   8411 	if (namelen != 0)
   8412 		return (EINVAL);
   8413 
   8414 	mutex_enter(&key_spd.lock);
   8415 	m = key_setspddump(&error, l->l_proc->p_pid);
   8416 	mutex_exit(&key_spd.lock);
   8417 	if (!m)
   8418 		return (error);
   8419 	if (!oldp)
   8420 		*oldlenp = m->m_pkthdr.len;
   8421 	else {
   8422 		p = oldp;
   8423 		if (*oldlenp < m->m_pkthdr.len) {
   8424 			err2 = ENOMEM;
   8425 			ep = p + *oldlenp;
   8426 		} else {
   8427 			*oldlenp = m->m_pkthdr.len;
   8428 			ep = p + m->m_pkthdr.len;
   8429 		}
   8430 		for (n = m; n; n = n->m_next) {
   8431 			len = (ep - p < n->m_len) ? ep - p : n->m_len;
   8432 			error = copyout(mtod(n, const void *), p, len);
   8433 			p += len;
   8434 			if (error)
   8435 				break;
   8436 		}
   8437 		if (error == 0)
   8438 			error = err2;
   8439 	}
   8440 	m_freem(m);
   8441 
   8442 	return (error);
   8443 }
   8444 
   8445 /*
   8446  * Create sysctl tree for native IPSEC key knobs, originally
   8447  * under name "net.keyv2"  * with MIB number { CTL_NET, PF_KEY_V2. }.
   8448  * However, sysctl(8) never checked for nodes under { CTL_NET, PF_KEY_V2 };
   8449  * and in any case the part of our sysctl namespace used for dumping the
   8450  * SPD and SA database  *HAS* to be compatible with the KAME sysctl
   8451  * namespace, for API reasons.
   8452  *
   8453  * Pending a consensus on the right way  to fix this, add a level of
   8454  * indirection in how we number the `native' IPSEC key nodes;
   8455  * and (as requested by Andrew Brown)  move registration of the
   8456  * KAME-compatible names  to a separate function.
   8457  */
   8458 #if 0
   8459 #  define IPSEC_PFKEY PF_KEY_V2
   8460 # define IPSEC_PFKEY_NAME "keyv2"
   8461 #else
   8462 #  define IPSEC_PFKEY PF_KEY
   8463 # define IPSEC_PFKEY_NAME "key"
   8464 #endif
   8465 
   8466 static int
   8467 sysctl_net_key_stats(SYSCTLFN_ARGS)
   8468 {
   8469 
   8470 	return (NETSTAT_SYSCTL(pfkeystat_percpu, PFKEY_NSTATS));
   8471 }
   8472 
   8473 static void
   8474 sysctl_net_keyv2_setup(struct sysctllog **clog)
   8475 {
   8476 
   8477 	sysctl_createv(clog, 0, NULL, NULL,
   8478 		       CTLFLAG_PERMANENT,
   8479 		       CTLTYPE_NODE, IPSEC_PFKEY_NAME, NULL,
   8480 		       NULL, 0, NULL, 0,
   8481 		       CTL_NET, IPSEC_PFKEY, CTL_EOL);
   8482 
   8483 	sysctl_createv(clog, 0, NULL, NULL,
   8484 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8485 		       CTLTYPE_INT, "debug", NULL,
   8486 		       NULL, 0, &key_debug_level, 0,
   8487 		       CTL_NET, IPSEC_PFKEY, KEYCTL_DEBUG_LEVEL, CTL_EOL);
   8488 	sysctl_createv(clog, 0, NULL, NULL,
   8489 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8490 		       CTLTYPE_INT, "spi_try", NULL,
   8491 		       NULL, 0, &key_spi_trycnt, 0,
   8492 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_TRY, CTL_EOL);
   8493 	sysctl_createv(clog, 0, NULL, NULL,
   8494 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8495 		       CTLTYPE_INT, "spi_min_value", NULL,
   8496 		       NULL, 0, &key_spi_minval, 0,
   8497 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_MIN_VALUE, CTL_EOL);
   8498 	sysctl_createv(clog, 0, NULL, NULL,
   8499 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8500 		       CTLTYPE_INT, "spi_max_value", NULL,
   8501 		       NULL, 0, &key_spi_maxval, 0,
   8502 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_MAX_VALUE, CTL_EOL);
   8503 	sysctl_createv(clog, 0, NULL, NULL,
   8504 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8505 		       CTLTYPE_INT, "random_int", NULL,
   8506 		       NULL, 0, &key_int_random, 0,
   8507 		       CTL_NET, IPSEC_PFKEY, KEYCTL_RANDOM_INT, CTL_EOL);
   8508 	sysctl_createv(clog, 0, NULL, NULL,
   8509 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8510 		       CTLTYPE_INT, "larval_lifetime", NULL,
   8511 		       NULL, 0, &key_larval_lifetime, 0,
   8512 		       CTL_NET, IPSEC_PFKEY, KEYCTL_LARVAL_LIFETIME, CTL_EOL);
   8513 	sysctl_createv(clog, 0, NULL, NULL,
   8514 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8515 		       CTLTYPE_INT, "blockacq_count", NULL,
   8516 		       NULL, 0, &key_blockacq_count, 0,
   8517 		       CTL_NET, IPSEC_PFKEY, KEYCTL_BLOCKACQ_COUNT, CTL_EOL);
   8518 	sysctl_createv(clog, 0, NULL, NULL,
   8519 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8520 		       CTLTYPE_INT, "blockacq_lifetime", NULL,
   8521 		       NULL, 0, &key_blockacq_lifetime, 0,
   8522 		       CTL_NET, IPSEC_PFKEY, KEYCTL_BLOCKACQ_LIFETIME, CTL_EOL);
   8523 	sysctl_createv(clog, 0, NULL, NULL,
   8524 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8525 		       CTLTYPE_INT, "esp_keymin", NULL,
   8526 		       NULL, 0, &ipsec_esp_keymin, 0,
   8527 		       CTL_NET, IPSEC_PFKEY, KEYCTL_ESP_KEYMIN, CTL_EOL);
   8528 	sysctl_createv(clog, 0, NULL, NULL,
   8529 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8530 		       CTLTYPE_INT, "prefered_oldsa", NULL,
   8531 		       NULL, 0, &key_prefered_oldsa, 0,
   8532 		       CTL_NET, PF_KEY, KEYCTL_PREFERED_OLDSA, CTL_EOL);
   8533 	sysctl_createv(clog, 0, NULL, NULL,
   8534 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8535 		       CTLTYPE_INT, "esp_auth", NULL,
   8536 		       NULL, 0, &ipsec_esp_auth, 0,
   8537 		       CTL_NET, IPSEC_PFKEY, KEYCTL_ESP_AUTH, CTL_EOL);
   8538 	sysctl_createv(clog, 0, NULL, NULL,
   8539 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8540 		       CTLTYPE_INT, "ah_keymin", NULL,
   8541 		       NULL, 0, &ipsec_ah_keymin, 0,
   8542 		       CTL_NET, IPSEC_PFKEY, KEYCTL_AH_KEYMIN, CTL_EOL);
   8543 	sysctl_createv(clog, 0, NULL, NULL,
   8544 		       CTLFLAG_PERMANENT,
   8545 		       CTLTYPE_STRUCT, "stats",
   8546 		       SYSCTL_DESCR("PF_KEY statistics"),
   8547 		       sysctl_net_key_stats, 0, NULL, 0,
   8548 		       CTL_NET, IPSEC_PFKEY, CTL_CREATE, CTL_EOL);
   8549 }
   8550 
   8551 /*
   8552  * Register sysctl names used by setkey(8). For historical reasons,
   8553  * and to share a single API, these names appear under { CTL_NET, PF_KEY }
   8554  * for both IPSEC and KAME IPSEC.
   8555  */
   8556 static void
   8557 sysctl_net_key_compat_setup(struct sysctllog **clog)
   8558 {
   8559 
   8560 	sysctl_createv(clog, 0, NULL, NULL,
   8561 		       CTLFLAG_PERMANENT,
   8562 		       CTLTYPE_NODE, "key", NULL,
   8563 		       NULL, 0, NULL, 0,
   8564 		       CTL_NET, PF_KEY, CTL_EOL);
   8565 
   8566 	/* Register the net.key.dump{sa,sp} nodes used by setkey(8). */
   8567 	sysctl_createv(clog, 0, NULL, NULL,
   8568 		       CTLFLAG_PERMANENT,
   8569 		       CTLTYPE_STRUCT, "dumpsa", NULL,
   8570 		       sysctl_net_key_dumpsa, 0, NULL, 0,
   8571 		       CTL_NET, PF_KEY, KEYCTL_DUMPSA, CTL_EOL);
   8572 	sysctl_createv(clog, 0, NULL, NULL,
   8573 		       CTLFLAG_PERMANENT,
   8574 		       CTLTYPE_STRUCT, "dumpsp", NULL,
   8575 		       sysctl_net_key_dumpsp, 0, NULL, 0,
   8576 		       CTL_NET, PF_KEY, KEYCTL_DUMPSP, CTL_EOL);
   8577 }
   8578