Home | History | Annotate | Line # | Download | only in netipsec
key.c revision 1.247
      1 /*	$NetBSD: key.c,v 1.247 2018/01/10 10:56:30 knakahara 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 1.247 2018/01/10 10:56:30 knakahara 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  * - SP origin
    200  *   - SPs can be created by both userland programs and kernel components.
    201  *     The SPs created in kernel must not be removed by userland programs,
    202  *     although the SPs can be read by userland programs.
    203  */
    204 /*
    205  * Locking notes on SAD:
    206  * - Data structures
    207  *   - SAs are managed by the list called key_sad.sahlist and sav lists of sah
    208  *     entries
    209  *     - An sav is supposed to be an SA from a viewpoint of users
    210  *   - A sah has sav lists for each SA state
    211  *   - Multiple sahs with the same saidx can exist
    212  *     - Only one entry has MATURE state and others should be DEAD
    213  *     - DEAD entries are just ignored from searching
    214  * - Modifications to the key_sad.sahlist and sah.savlist must be done with
    215  *   holding key_sad.lock which is a adaptive mutex
    216  * - Read accesses to the key_sad.sahlist and sah.savlist must be in
    217  *   pserialize(9) read sections
    218  * - sah's lifetime is managed by localcount(9)
    219  * - Getting an sah entry
    220  *   - We get an sah from the key_sad.sahlist
    221  *     - Must iterate the list and increment the reference count of a found sah
    222  *       (by key_sah_ref) in a pserialize read section
    223  *   - A gotten sah must be released after use by key_sah_unref
    224  * - An sah is destroyed when its state become DEAD and no sav is
    225  *   listed to the sah
    226  *   - The destruction is done only in the timer (see key_timehandler_sad)
    227  * - sav's lifetime is managed by localcount(9)
    228  * - Getting an sav entry
    229  *   - First get an sah by saidx and get an sav from either of sah's savlists
    230  *     - Must iterate the list and increment the reference count of a found sav
    231  *       (by key_sa_ref) in a pserialize read section
    232  *   - We can gain another reference from a held SA only if we check its state
    233  *     and take its reference in a pserialize read section
    234  *     (see esp_output for example)
    235  *   - A gotten sav must be released after use by key_sa_unref
    236  * - An sav is destroyed when its state become DEAD
    237  */
    238 /*
    239  * Locking notes on misc data:
    240  * - All lists of key_misc are protected by key_misc.lock
    241  *   - key_misc.lock must be held even for read accesses
    242  */
    243 
    244 /* SPD */
    245 static struct {
    246 	kmutex_t lock;
    247 	kcondvar_t cv_lc;
    248 	struct pslist_head splist[IPSEC_DIR_MAX];
    249 	/*
    250 	 * The list has SPs that are set to a socket via
    251 	 * setsockopt(IP_IPSEC_POLICY) from userland. See ipsec_set_policy.
    252 	 */
    253 	struct pslist_head socksplist;
    254 
    255 	pserialize_t psz;
    256 	kcondvar_t cv_psz;
    257 	bool psz_performing;
    258 } key_spd __cacheline_aligned;
    259 
    260 /* SAD */
    261 static struct {
    262 	kmutex_t lock;
    263 	kcondvar_t cv_lc;
    264 	struct pslist_head sahlist;
    265 
    266 	pserialize_t psz;
    267 	kcondvar_t cv_psz;
    268 	bool psz_performing;
    269 } key_sad __cacheline_aligned;
    270 
    271 /* Misc data */
    272 static struct {
    273 	kmutex_t lock;
    274 	/* registed list */
    275 	LIST_HEAD(_reglist, secreg) reglist[SADB_SATYPE_MAX + 1];
    276 #ifndef IPSEC_NONBLOCK_ACQUIRE
    277 	/* acquiring list */
    278 	LIST_HEAD(_acqlist, secacq) acqlist;
    279 #endif
    280 #ifdef notyet
    281 	/* SP acquiring list */
    282 	LIST_HEAD(_spacqlist, secspacq) spacqlist;
    283 #endif
    284 } key_misc __cacheline_aligned;
    285 
    286 /* Macros for key_spd.splist */
    287 #define SPLIST_ENTRY_INIT(sp)						\
    288 	PSLIST_ENTRY_INIT((sp), pslist_entry)
    289 #define SPLIST_ENTRY_DESTROY(sp)					\
    290 	PSLIST_ENTRY_DESTROY((sp), pslist_entry)
    291 #define SPLIST_WRITER_REMOVE(sp)					\
    292 	PSLIST_WRITER_REMOVE((sp), pslist_entry)
    293 #define SPLIST_READER_EMPTY(dir)					\
    294 	(PSLIST_READER_FIRST(&key_spd.splist[(dir)], struct secpolicy,	\
    295 	                     pslist_entry) == NULL)
    296 #define SPLIST_READER_FOREACH(sp, dir)					\
    297 	PSLIST_READER_FOREACH((sp), &key_spd.splist[(dir)],		\
    298 	                      struct secpolicy, pslist_entry)
    299 #define SPLIST_WRITER_FOREACH(sp, dir)					\
    300 	PSLIST_WRITER_FOREACH((sp), &key_spd.splist[(dir)],		\
    301 	                      struct secpolicy, pslist_entry)
    302 #define SPLIST_WRITER_INSERT_AFTER(sp, new)				\
    303 	PSLIST_WRITER_INSERT_AFTER((sp), (new), pslist_entry)
    304 #define SPLIST_WRITER_EMPTY(dir)					\
    305 	(PSLIST_WRITER_FIRST(&key_spd.splist[(dir)], struct secpolicy,	\
    306 	                     pslist_entry) == NULL)
    307 #define SPLIST_WRITER_INSERT_HEAD(dir, sp)				\
    308 	PSLIST_WRITER_INSERT_HEAD(&key_spd.splist[(dir)], (sp),		\
    309 	                          pslist_entry)
    310 #define SPLIST_WRITER_NEXT(sp)						\
    311 	PSLIST_WRITER_NEXT((sp), struct secpolicy, pslist_entry)
    312 #define SPLIST_WRITER_INSERT_TAIL(dir, new)				\
    313 	do {								\
    314 		if (SPLIST_WRITER_EMPTY((dir))) {			\
    315 			SPLIST_WRITER_INSERT_HEAD((dir), (new));	\
    316 		} else {						\
    317 			struct secpolicy *__sp;				\
    318 			SPLIST_WRITER_FOREACH(__sp, (dir)) {		\
    319 				if (SPLIST_WRITER_NEXT(__sp) == NULL) {	\
    320 					SPLIST_WRITER_INSERT_AFTER(__sp,\
    321 					    (new));			\
    322 					break;				\
    323 				}					\
    324 			}						\
    325 		}							\
    326 	} while (0)
    327 
    328 /* Macros for key_spd.socksplist */
    329 #define SOCKSPLIST_WRITER_FOREACH(sp)					\
    330 	PSLIST_WRITER_FOREACH((sp), &key_spd.socksplist,		\
    331 	                      struct secpolicy,	pslist_entry)
    332 #define SOCKSPLIST_READER_EMPTY()					\
    333 	(PSLIST_READER_FIRST(&key_spd.socksplist, struct secpolicy,	\
    334 	                     pslist_entry) == NULL)
    335 
    336 /* Macros for key_sad.sahlist */
    337 #define SAHLIST_ENTRY_INIT(sah)						\
    338 	PSLIST_ENTRY_INIT((sah), pslist_entry)
    339 #define SAHLIST_ENTRY_DESTROY(sah)					\
    340 	PSLIST_ENTRY_DESTROY((sah), pslist_entry)
    341 #define SAHLIST_WRITER_REMOVE(sah)					\
    342 	PSLIST_WRITER_REMOVE((sah), pslist_entry)
    343 #define SAHLIST_READER_FOREACH(sah)					\
    344 	PSLIST_READER_FOREACH((sah), &key_sad.sahlist, struct secashead,\
    345 	                      pslist_entry)
    346 #define SAHLIST_WRITER_FOREACH(sah)					\
    347 	PSLIST_WRITER_FOREACH((sah), &key_sad.sahlist, struct secashead,\
    348 	                      pslist_entry)
    349 #define SAHLIST_WRITER_INSERT_HEAD(sah)					\
    350 	PSLIST_WRITER_INSERT_HEAD(&key_sad.sahlist, (sah), pslist_entry)
    351 
    352 /* Macros for key_sad.sahlist#savlist */
    353 #define SAVLIST_ENTRY_INIT(sav)						\
    354 	PSLIST_ENTRY_INIT((sav), pslist_entry)
    355 #define SAVLIST_ENTRY_DESTROY(sav)					\
    356 	PSLIST_ENTRY_DESTROY((sav), pslist_entry)
    357 #define SAVLIST_READER_FIRST(sah, state)				\
    358 	PSLIST_READER_FIRST(&(sah)->savlist[(state)], struct secasvar,	\
    359 	                    pslist_entry)
    360 #define SAVLIST_WRITER_REMOVE(sav)					\
    361 	PSLIST_WRITER_REMOVE((sav), pslist_entry)
    362 #define SAVLIST_READER_FOREACH(sav, sah, state)				\
    363 	PSLIST_READER_FOREACH((sav), &(sah)->savlist[(state)],		\
    364 	                      struct secasvar, pslist_entry)
    365 #define SAVLIST_WRITER_FOREACH(sav, sah, state)				\
    366 	PSLIST_WRITER_FOREACH((sav), &(sah)->savlist[(state)],		\
    367 	                      struct secasvar, pslist_entry)
    368 #define SAVLIST_WRITER_INSERT_BEFORE(sav, new)				\
    369 	PSLIST_WRITER_INSERT_BEFORE((sav), (new), pslist_entry)
    370 #define SAVLIST_WRITER_INSERT_AFTER(sav, new)				\
    371 	PSLIST_WRITER_INSERT_AFTER((sav), (new), pslist_entry)
    372 #define SAVLIST_WRITER_EMPTY(sah, state)				\
    373 	(PSLIST_WRITER_FIRST(&(sah)->savlist[(state)], struct secasvar,	\
    374 	                     pslist_entry) == NULL)
    375 #define SAVLIST_WRITER_INSERT_HEAD(sah, state, sav)			\
    376 	PSLIST_WRITER_INSERT_HEAD(&(sah)->savlist[(state)], (sav),	\
    377 	                          pslist_entry)
    378 #define SAVLIST_WRITER_NEXT(sav)					\
    379 	PSLIST_WRITER_NEXT((sav), struct secasvar, pslist_entry)
    380 #define SAVLIST_WRITER_INSERT_TAIL(sah, state, new)			\
    381 	do {								\
    382 		if (SAVLIST_WRITER_EMPTY((sah), (state))) {		\
    383 			SAVLIST_WRITER_INSERT_HEAD((sah), (state), (new));\
    384 		} else {						\
    385 			struct secasvar *__sav;				\
    386 			SAVLIST_WRITER_FOREACH(__sav, (sah), (state)) {	\
    387 				if (SAVLIST_WRITER_NEXT(__sav) == NULL) {\
    388 					SAVLIST_WRITER_INSERT_AFTER(__sav,\
    389 					    (new));			\
    390 					break;				\
    391 				}					\
    392 			}						\
    393 		}							\
    394 	} while (0)
    395 #define SAVLIST_READER_NEXT(sav)					\
    396 	PSLIST_READER_NEXT((sav), struct secasvar, pslist_entry)
    397 
    398 
    399 /* search order for SAs */
    400 	/*
    401 	 * This order is important because we must select the oldest SA
    402 	 * for outbound processing.  For inbound, This is not important.
    403 	 */
    404 static const u_int saorder_state_valid_prefer_old[] = {
    405 	SADB_SASTATE_DYING, SADB_SASTATE_MATURE,
    406 };
    407 static const u_int saorder_state_valid_prefer_new[] = {
    408 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING,
    409 };
    410 
    411 static const u_int saorder_state_alive[] = {
    412 	/* except DEAD */
    413 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING, SADB_SASTATE_LARVAL
    414 };
    415 static const u_int saorder_state_any[] = {
    416 	SADB_SASTATE_MATURE, SADB_SASTATE_DYING,
    417 	SADB_SASTATE_LARVAL, SADB_SASTATE_DEAD
    418 };
    419 
    420 #define SASTATE_ALIVE_FOREACH(s)				\
    421 	for (int _i = 0;					\
    422 	    _i < __arraycount(saorder_state_alive) ?		\
    423 	    (s) = saorder_state_alive[_i], true : false;	\
    424 	    _i++)
    425 #define SASTATE_ANY_FOREACH(s)					\
    426 	for (int _i = 0;					\
    427 	    _i < __arraycount(saorder_state_any) ?		\
    428 	    (s) = saorder_state_any[_i], true : false;		\
    429 	    _i++)
    430 
    431 static const int minsize[] = {
    432 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
    433 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
    434 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
    435 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
    436 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
    437 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_SRC */
    438 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_DST */
    439 	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_PROXY */
    440 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_AUTH */
    441 	sizeof(struct sadb_key),	/* SADB_EXT_KEY_ENCRYPT */
    442 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_SRC */
    443 	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_DST */
    444 	sizeof(struct sadb_sens),	/* SADB_EXT_SENSITIVITY */
    445 	sizeof(struct sadb_prop),	/* SADB_EXT_PROPOSAL */
    446 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_AUTH */
    447 	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_ENCRYPT */
    448 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
    449 	0,				/* SADB_X_EXT_KMPRIVATE */
    450 	sizeof(struct sadb_x_policy),	/* SADB_X_EXT_POLICY */
    451 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
    452 	sizeof(struct sadb_x_nat_t_type),	/* SADB_X_EXT_NAT_T_TYPE */
    453 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_SPORT */
    454 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_DPORT */
    455 	sizeof(struct sadb_address),		/* SADB_X_EXT_NAT_T_OAI */
    456 	sizeof(struct sadb_address),		/* SADB_X_EXT_NAT_T_OAR */
    457 	sizeof(struct sadb_x_nat_t_frag),	/* SADB_X_EXT_NAT_T_FRAG */
    458 };
    459 static const int maxsize[] = {
    460 	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
    461 	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
    462 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
    463 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
    464 	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
    465 	0,				/* SADB_EXT_ADDRESS_SRC */
    466 	0,				/* SADB_EXT_ADDRESS_DST */
    467 	0,				/* SADB_EXT_ADDRESS_PROXY */
    468 	0,				/* SADB_EXT_KEY_AUTH */
    469 	0,				/* SADB_EXT_KEY_ENCRYPT */
    470 	0,				/* SADB_EXT_IDENTITY_SRC */
    471 	0,				/* SADB_EXT_IDENTITY_DST */
    472 	0,				/* SADB_EXT_SENSITIVITY */
    473 	0,				/* SADB_EXT_PROPOSAL */
    474 	0,				/* SADB_EXT_SUPPORTED_AUTH */
    475 	0,				/* SADB_EXT_SUPPORTED_ENCRYPT */
    476 	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
    477 	0,				/* SADB_X_EXT_KMPRIVATE */
    478 	0,				/* SADB_X_EXT_POLICY */
    479 	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
    480 	sizeof(struct sadb_x_nat_t_type),	/* SADB_X_EXT_NAT_T_TYPE */
    481 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_SPORT */
    482 	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_DPORT */
    483 	0,					/* SADB_X_EXT_NAT_T_OAI */
    484 	0,					/* SADB_X_EXT_NAT_T_OAR */
    485 	sizeof(struct sadb_x_nat_t_frag),	/* SADB_X_EXT_NAT_T_FRAG */
    486 };
    487 
    488 static int ipsec_esp_keymin = 256;
    489 static int ipsec_esp_auth = 0;
    490 static int ipsec_ah_keymin = 128;
    491 
    492 #ifdef SYSCTL_DECL
    493 SYSCTL_DECL(_net_key);
    494 #endif
    495 
    496 #ifdef SYSCTL_INT
    497 SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,	CTLFLAG_RW, \
    498 	&key_debug_level,	0,	"");
    499 
    500 /* max count of trial for the decision of spi value */
    501 SYSCTL_INT(_net_key, KEYCTL_SPI_TRY,		spi_trycnt,	CTLFLAG_RW, \
    502 	&key_spi_trycnt,	0,	"");
    503 
    504 /* minimum spi value to allocate automatically. */
    505 SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE,	spi_minval,	CTLFLAG_RW, \
    506 	&key_spi_minval,	0,	"");
    507 
    508 /* maximun spi value to allocate automatically. */
    509 SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE,	spi_maxval,	CTLFLAG_RW, \
    510 	&key_spi_maxval,	0,	"");
    511 
    512 /* interval to initialize randseed */
    513 SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT,	int_random,	CTLFLAG_RW, \
    514 	&key_int_random,	0,	"");
    515 
    516 /* lifetime for larval SA */
    517 SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME,	larval_lifetime, CTLFLAG_RW, \
    518 	&key_larval_lifetime,	0,	"");
    519 
    520 /* counter for blocking to send SADB_ACQUIRE to IKEd */
    521 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT,	blockacq_count,	CTLFLAG_RW, \
    522 	&key_blockacq_count,	0,	"");
    523 
    524 /* lifetime for blocking to send SADB_ACQUIRE to IKEd */
    525 SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME,	blockacq_lifetime, CTLFLAG_RW, \
    526 	&key_blockacq_lifetime,	0,	"");
    527 
    528 /* ESP auth */
    529 SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH,	esp_auth, CTLFLAG_RW, \
    530 	&ipsec_esp_auth,	0,	"");
    531 
    532 /* minimum ESP key length */
    533 SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN,	esp_keymin, CTLFLAG_RW, \
    534 	&ipsec_esp_keymin,	0,	"");
    535 
    536 /* minimum AH key length */
    537 SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN,	ah_keymin, CTLFLAG_RW, \
    538 	&ipsec_ah_keymin,	0,	"");
    539 
    540 /* perfered old SA rather than new SA */
    541 SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA,	prefered_oldsa, CTLFLAG_RW,\
    542 	&key_prefered_oldsa,	0,	"");
    543 #endif /* SYSCTL_INT */
    544 
    545 #define __LIST_CHAINED(elm) \
    546 	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
    547 #define LIST_INSERT_TAIL(head, elm, type, field) \
    548 do {\
    549 	struct type *curelm = LIST_FIRST(head); \
    550 	if (curelm == NULL) {\
    551 		LIST_INSERT_HEAD(head, elm, field); \
    552 	} else { \
    553 		while (LIST_NEXT(curelm, field)) \
    554 			curelm = LIST_NEXT(curelm, field);\
    555 		LIST_INSERT_AFTER(curelm, elm, field);\
    556 	}\
    557 } while (0)
    558 
    559 #define KEY_CHKSASTATE(head, sav) \
    560 /* do */ { \
    561 	if ((head) != (sav)) {						\
    562 		IPSECLOG(LOG_DEBUG,					\
    563 		    "state mismatched (TREE=%d SA=%d)\n",		\
    564 		    (head), (sav));					\
    565 		continue;						\
    566 	}								\
    567 } /* while (0) */
    568 
    569 #define KEY_CHKSPDIR(head, sp) \
    570 do { \
    571 	if ((head) != (sp)) {						\
    572 		IPSECLOG(LOG_DEBUG,					\
    573 		    "direction mismatched (TREE=%d SP=%d), anyway continue.\n",\
    574 		    (head), (sp));					\
    575 	}								\
    576 } while (0)
    577 
    578 /*
    579  * set parameters into secasindex buffer.
    580  * Must allocate secasindex buffer before calling this function.
    581  */
    582 static int
    583 key_setsecasidx(int, int, int, const struct sockaddr *,
    584     const struct sockaddr *, struct secasindex *);
    585 
    586 /* key statistics */
    587 struct _keystat {
    588 	u_long getspi_count; /* the avarage of count to try to get new SPI */
    589 } keystat;
    590 
    591 static void
    592 key_init_spidx_bymsghdr(struct secpolicyindex *, const struct sadb_msghdr *);
    593 
    594 static const struct sockaddr *
    595 key_msghdr_get_sockaddr(const struct sadb_msghdr *mhp, int idx)
    596 {
    597 
    598 	return PFKEY_ADDR_SADDR(mhp->ext[idx]);
    599 }
    600 
    601 static void
    602 key_fill_replymsg(struct mbuf *m, int seq)
    603 {
    604 	struct sadb_msg *msg;
    605 
    606 	KASSERT(m->m_len >= sizeof(*msg));
    607 
    608 	msg = mtod(m, struct sadb_msg *);
    609 	msg->sadb_msg_errno = 0;
    610 	msg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
    611 	if (seq != 0)
    612 		msg->sadb_msg_seq = seq;
    613 }
    614 
    615 #if 0
    616 static void key_freeso(struct socket *);
    617 static void key_freesp_so(struct secpolicy **);
    618 #endif
    619 static struct secpolicy *key_getsp (const struct secpolicyindex *);
    620 static struct secpolicy *key_getspbyid (u_int32_t);
    621 static struct secpolicy *key_lookup_and_remove_sp(const struct secpolicyindex *, bool);
    622 static struct secpolicy *key_lookupbyid_and_remove_sp(u_int32_t, bool);
    623 static void key_destroy_sp(struct secpolicy *);
    624 static struct mbuf *key_gather_mbuf (struct mbuf *,
    625 	const struct sadb_msghdr *, int, int, ...);
    626 static int key_api_spdadd(struct socket *, struct mbuf *,
    627 	const struct sadb_msghdr *);
    628 static u_int32_t key_getnewspid (void);
    629 static int key_api_spddelete(struct socket *, struct mbuf *,
    630 	const struct sadb_msghdr *);
    631 static int key_api_spddelete2(struct socket *, struct mbuf *,
    632 	const struct sadb_msghdr *);
    633 static int key_api_spdget(struct socket *, struct mbuf *,
    634 	const struct sadb_msghdr *);
    635 static int key_api_spdflush(struct socket *, struct mbuf *,
    636 	const struct sadb_msghdr *);
    637 static int key_api_spddump(struct socket *, struct mbuf *,
    638 	const struct sadb_msghdr *);
    639 static struct mbuf * key_setspddump (int *errorp, pid_t);
    640 static struct mbuf * key_setspddump_chain (int *errorp, int *lenp, pid_t pid);
    641 static int key_api_nat_map(struct socket *, struct mbuf *,
    642 	const struct sadb_msghdr *);
    643 static struct mbuf *key_setdumpsp (struct secpolicy *,
    644 	u_int8_t, u_int32_t, pid_t);
    645 static u_int key_getspreqmsglen (const struct secpolicy *);
    646 static int key_spdexpire (struct secpolicy *);
    647 static struct secashead *key_newsah (const struct secasindex *);
    648 static void key_unlink_sah(struct secashead *);
    649 static void key_destroy_sah(struct secashead *);
    650 static bool key_sah_has_sav(struct secashead *);
    651 static void key_sah_ref(struct secashead *);
    652 static void key_sah_unref(struct secashead *);
    653 static void key_init_sav(struct secasvar *);
    654 static void key_destroy_sav(struct secasvar *);
    655 static void key_destroy_sav_with_ref(struct secasvar *);
    656 static struct secasvar *key_newsav(struct mbuf *,
    657 	const struct sadb_msghdr *, int *, const char*, int);
    658 #define	KEY_NEWSAV(m, sadb, e)				\
    659 	key_newsav(m, sadb, e, __func__, __LINE__)
    660 static void key_delsav (struct secasvar *);
    661 static struct secashead *key_getsah(const struct secasindex *, int);
    662 static struct secashead *key_getsah_ref(const struct secasindex *, int);
    663 static bool key_checkspidup(const struct secasindex *, u_int32_t);
    664 static struct secasvar *key_getsavbyspi (struct secashead *, u_int32_t);
    665 static int key_setsaval (struct secasvar *, struct mbuf *,
    666 	const struct sadb_msghdr *);
    667 static void key_freesaval(struct secasvar *);
    668 static int key_init_xform(struct secasvar *);
    669 static void key_clear_xform(struct secasvar *);
    670 static struct mbuf *key_setdumpsa (struct secasvar *, u_int8_t,
    671 	u_int8_t, u_int32_t, u_int32_t);
    672 static struct mbuf *key_setsadbxport (u_int16_t, u_int16_t);
    673 static struct mbuf *key_setsadbxtype (u_int16_t);
    674 static struct mbuf *key_setsadbxfrag (u_int16_t);
    675 static void key_porttosaddr (union sockaddr_union *, u_int16_t);
    676 static int key_checksalen (const union sockaddr_union *);
    677 static struct mbuf *key_setsadbmsg (u_int8_t, u_int16_t, u_int8_t,
    678 	u_int32_t, pid_t, u_int16_t, int);
    679 static struct mbuf *key_setsadbsa (struct secasvar *);
    680 static struct mbuf *key_setsadbaddr(u_int16_t,
    681 	const struct sockaddr *, u_int8_t, u_int16_t, int);
    682 #if 0
    683 static struct mbuf *key_setsadbident (u_int16_t, u_int16_t, void *,
    684 	int, u_int64_t);
    685 #endif
    686 static struct mbuf *key_setsadbxsa2 (u_int8_t, u_int32_t, u_int16_t);
    687 static struct mbuf *key_setsadbxpolicy (u_int16_t, u_int8_t,
    688 	u_int32_t, int);
    689 static void *key_newbuf (const void *, u_int);
    690 #ifdef INET6
    691 static int key_ismyaddr6 (const struct sockaddr_in6 *);
    692 #endif
    693 
    694 static void sysctl_net_keyv2_setup(struct sysctllog **);
    695 static void sysctl_net_key_compat_setup(struct sysctllog **);
    696 
    697 /* flags for key_saidx_match() */
    698 #define CMP_HEAD	1	/* protocol, addresses. */
    699 #define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
    700 #define CMP_REQID	3	/* additionally HEAD, reaid. */
    701 #define CMP_EXACTLY	4	/* all elements. */
    702 static int key_saidx_match(const struct secasindex *,
    703     const struct secasindex *, int);
    704 
    705 static int key_sockaddr_match(const struct sockaddr *,
    706     const struct sockaddr *, int);
    707 static int key_bb_match_withmask(const void *, const void *, u_int);
    708 static u_int16_t key_satype2proto (u_int8_t);
    709 static u_int8_t key_proto2satype (u_int16_t);
    710 
    711 static int key_spidx_match_exactly(const struct secpolicyindex *,
    712     const struct secpolicyindex *);
    713 static int key_spidx_match_withmask(const struct secpolicyindex *,
    714     const struct secpolicyindex *);
    715 
    716 static int key_api_getspi(struct socket *, struct mbuf *,
    717 	const struct sadb_msghdr *);
    718 static u_int32_t key_do_getnewspi (const struct sadb_spirange *,
    719 					const struct secasindex *);
    720 static int key_handle_natt_info (struct secasvar *,
    721 				     const struct sadb_msghdr *);
    722 static int key_set_natt_ports (union sockaddr_union *,
    723 			 	union sockaddr_union *,
    724 				const struct sadb_msghdr *);
    725 static int key_api_update(struct socket *, struct mbuf *,
    726 	const struct sadb_msghdr *);
    727 #ifdef IPSEC_DOSEQCHECK
    728 static struct secasvar *key_getsavbyseq (struct secashead *, u_int32_t);
    729 #endif
    730 static int key_api_add(struct socket *, struct mbuf *,
    731 	const struct sadb_msghdr *);
    732 static int key_setident (struct secashead *, struct mbuf *,
    733 	const struct sadb_msghdr *);
    734 static struct mbuf *key_getmsgbuf_x1 (struct mbuf *,
    735 	const struct sadb_msghdr *);
    736 static int key_api_delete(struct socket *, struct mbuf *,
    737 	const struct sadb_msghdr *);
    738 static int key_api_get(struct socket *, struct mbuf *,
    739 	const struct sadb_msghdr *);
    740 
    741 static void key_getcomb_setlifetime (struct sadb_comb *);
    742 static struct mbuf *key_getcomb_esp(int);
    743 static struct mbuf *key_getcomb_ah(int);
    744 static struct mbuf *key_getcomb_ipcomp(int);
    745 static struct mbuf *key_getprop(const struct secasindex *, int);
    746 
    747 static int key_acquire(const struct secasindex *, const struct secpolicy *,
    748 	    int);
    749 static int key_acquire_sendup_mbuf_later(struct mbuf *);
    750 static void key_acquire_sendup_pending_mbuf(void);
    751 #ifndef IPSEC_NONBLOCK_ACQUIRE
    752 static struct secacq *key_newacq (const struct secasindex *);
    753 static struct secacq *key_getacq (const struct secasindex *);
    754 static struct secacq *key_getacqbyseq (u_int32_t);
    755 #endif
    756 #ifdef notyet
    757 static struct secspacq *key_newspacq (const struct secpolicyindex *);
    758 static struct secspacq *key_getspacq (const struct secpolicyindex *);
    759 #endif
    760 static int key_api_acquire(struct socket *, struct mbuf *,
    761 	const struct sadb_msghdr *);
    762 static int key_api_register(struct socket *, struct mbuf *,
    763 	const struct sadb_msghdr *);
    764 static int key_expire (struct secasvar *);
    765 static int key_api_flush(struct socket *, struct mbuf *,
    766 	const struct sadb_msghdr *);
    767 static struct mbuf *key_setdump_chain (u_int8_t req_satype, int *errorp,
    768 	int *lenp, pid_t pid);
    769 static int key_api_dump(struct socket *, struct mbuf *,
    770 	const struct sadb_msghdr *);
    771 static int key_api_promisc(struct socket *, struct mbuf *,
    772 	const struct sadb_msghdr *);
    773 static int key_senderror (struct socket *, struct mbuf *, int);
    774 static int key_validate_ext (const struct sadb_ext *, int);
    775 static int key_align (struct mbuf *, struct sadb_msghdr *);
    776 #if 0
    777 static const char *key_getfqdn (void);
    778 static const char *key_getuserfqdn (void);
    779 #endif
    780 static void key_sa_chgstate (struct secasvar *, u_int8_t);
    781 
    782 static struct mbuf *key_alloc_mbuf(int, int);
    783 static struct mbuf *key_alloc_mbuf_simple(int, int);
    784 
    785 static void key_timehandler(void *);
    786 static void key_timehandler_work(struct work *, void *);
    787 static struct callout	key_timehandler_ch;
    788 static struct workqueue	*key_timehandler_wq;
    789 static struct work	key_timehandler_wk;
    790 
    791 u_int
    792 key_sp_refcnt(const struct secpolicy *sp)
    793 {
    794 
    795 	/* FIXME */
    796 	return 0;
    797 }
    798 
    799 static void
    800 key_spd_pserialize_perform(void)
    801 {
    802 
    803 	KASSERT(mutex_owned(&key_spd.lock));
    804 
    805 	while (key_spd.psz_performing)
    806 		cv_wait(&key_spd.cv_psz, &key_spd.lock);
    807 	key_spd.psz_performing = true;
    808 	mutex_exit(&key_spd.lock);
    809 
    810 	pserialize_perform(key_spd.psz);
    811 
    812 	mutex_enter(&key_spd.lock);
    813 	key_spd.psz_performing = false;
    814 	cv_broadcast(&key_spd.cv_psz);
    815 }
    816 
    817 /*
    818  * Remove the sp from the key_spd.splist and wait for references to the sp
    819  * to be released. key_spd.lock must be held.
    820  */
    821 static void
    822 key_unlink_sp(struct secpolicy *sp)
    823 {
    824 
    825 	KASSERT(mutex_owned(&key_spd.lock));
    826 
    827 	sp->state = IPSEC_SPSTATE_DEAD;
    828 	SPLIST_WRITER_REMOVE(sp);
    829 
    830 	/* Invalidate all cached SPD pointers in the PCBs. */
    831 	ipsec_invalpcbcacheall();
    832 
    833 	KDASSERT(mutex_ownable(softnet_lock));
    834 	key_spd_pserialize_perform();
    835 
    836 	localcount_drain(&sp->localcount, &key_spd.cv_lc, &key_spd.lock);
    837 }
    838 
    839 /*
    840  * Return 0 when there are known to be no SP's for the specified
    841  * direction.  Otherwise return 1.  This is used by IPsec code
    842  * to optimize performance.
    843  */
    844 int
    845 key_havesp(u_int dir)
    846 {
    847 	return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ?
    848 		!SPLIST_READER_EMPTY(dir) : 1);
    849 }
    850 
    851 /* %%% IPsec policy management */
    852 /*
    853  * allocating a SP for OUTBOUND or INBOUND packet.
    854  * Must call key_freesp() later.
    855  * OUT:	NULL:	not found
    856  *	others:	found and return the pointer.
    857  */
    858 struct secpolicy *
    859 key_lookup_sp_byspidx(const struct secpolicyindex *spidx,
    860     u_int dir, const char* where, int tag)
    861 {
    862 	struct secpolicy *sp;
    863 	int s;
    864 
    865 	KASSERT(spidx != NULL);
    866 	KASSERTMSG(IPSEC_DIR_IS_INOROUT(dir), "invalid direction %u", dir);
    867 
    868 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
    869 
    870 	/* get a SP entry */
    871 	if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
    872 		kdebug_secpolicyindex("objects", spidx);
    873 	}
    874 
    875 	s = pserialize_read_enter();
    876 	SPLIST_READER_FOREACH(sp, dir) {
    877 		if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
    878 			kdebug_secpolicyindex("in SPD", &sp->spidx);
    879 		}
    880 
    881 		if (sp->state == IPSEC_SPSTATE_DEAD)
    882 			continue;
    883 		if (key_spidx_match_withmask(&sp->spidx, spidx))
    884 			goto found;
    885 	}
    886 	sp = NULL;
    887 found:
    888 	if (sp) {
    889 		/* sanity check */
    890 		KEY_CHKSPDIR(sp->spidx.dir, dir);
    891 
    892 		/* found a SPD entry */
    893 		sp->lastused = time_uptime;
    894 		key_sp_ref(sp, where, tag);
    895 	}
    896 	pserialize_read_exit(s);
    897 
    898 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
    899 	    "DP return SP:%p (ID=%u) refcnt %u\n",
    900 	    sp, sp ? sp->id : 0, key_sp_refcnt(sp));
    901 	return sp;
    902 }
    903 
    904 /*
    905  * return a policy that matches this particular inbound packet.
    906  * XXX slow
    907  */
    908 struct secpolicy *
    909 key_gettunnel(const struct sockaddr *osrc,
    910 	      const struct sockaddr *odst,
    911 	      const struct sockaddr *isrc,
    912 	      const struct sockaddr *idst,
    913 	      const char* where, int tag)
    914 {
    915 	struct secpolicy *sp;
    916 	const int dir = IPSEC_DIR_INBOUND;
    917 	int s;
    918 	struct ipsecrequest *r1, *r2, *p;
    919 	struct secpolicyindex spidx;
    920 
    921 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
    922 
    923 	if (isrc->sa_family != idst->sa_family) {
    924 		IPSECLOG(LOG_ERR, "protocol family mismatched %d != %d\n.",
    925 		    isrc->sa_family, idst->sa_family);
    926 		sp = NULL;
    927 		goto done;
    928 	}
    929 
    930 	s = pserialize_read_enter();
    931 	SPLIST_READER_FOREACH(sp, dir) {
    932 		if (sp->state == IPSEC_SPSTATE_DEAD)
    933 			continue;
    934 
    935 		r1 = r2 = NULL;
    936 		for (p = sp->req; p; p = p->next) {
    937 			if (p->saidx.mode != IPSEC_MODE_TUNNEL)
    938 				continue;
    939 
    940 			r1 = r2;
    941 			r2 = p;
    942 
    943 			if (!r1) {
    944 				/* here we look at address matches only */
    945 				spidx = sp->spidx;
    946 				if (isrc->sa_len > sizeof(spidx.src) ||
    947 				    idst->sa_len > sizeof(spidx.dst))
    948 					continue;
    949 				memcpy(&spidx.src, isrc, isrc->sa_len);
    950 				memcpy(&spidx.dst, idst, idst->sa_len);
    951 				if (!key_spidx_match_withmask(&sp->spidx, &spidx))
    952 					continue;
    953 			} else {
    954 				if (!key_sockaddr_match(&r1->saidx.src.sa, isrc, PORT_NONE) ||
    955 				    !key_sockaddr_match(&r1->saidx.dst.sa, idst, PORT_NONE))
    956 					continue;
    957 			}
    958 
    959 			if (!key_sockaddr_match(&r2->saidx.src.sa, osrc, PORT_NONE) ||
    960 			    !key_sockaddr_match(&r2->saidx.dst.sa, odst, PORT_NONE))
    961 				continue;
    962 
    963 			goto found;
    964 		}
    965 	}
    966 	sp = NULL;
    967 found:
    968 	if (sp) {
    969 		sp->lastused = time_uptime;
    970 		key_sp_ref(sp, where, tag);
    971 	}
    972 	pserialize_read_exit(s);
    973 done:
    974 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
    975 	    "DP return SP:%p (ID=%u) refcnt %u\n",
    976 	    sp, sp ? sp->id : 0, key_sp_refcnt(sp));
    977 	return sp;
    978 }
    979 
    980 /*
    981  * allocating an SA entry for an *OUTBOUND* packet.
    982  * checking each request entries in SP, and acquire an SA if need.
    983  * OUT:	0: there are valid requests.
    984  *	ENOENT: policy may be valid, but SA with REQUIRE is on acquiring.
    985  */
    986 int
    987 key_checkrequest(const struct ipsecrequest *isr, const struct secasindex *saidx,
    988     struct secasvar **ret)
    989 {
    990 	u_int level;
    991 	int error;
    992 	struct secasvar *sav;
    993 
    994 	KASSERT(isr != NULL);
    995 	KASSERTMSG(saidx->mode == IPSEC_MODE_TRANSPORT ||
    996 	    saidx->mode == IPSEC_MODE_TUNNEL,
    997 	    "unexpected policy %u", saidx->mode);
    998 
    999 	/* get current level */
   1000 	level = ipsec_get_reqlevel(isr);
   1001 
   1002 	/*
   1003 	 * XXX guard against protocol callbacks from the crypto
   1004 	 * thread as they reference ipsecrequest.sav which we
   1005 	 * temporarily null out below.  Need to rethink how we
   1006 	 * handle bundled SA's in the callback thread.
   1007 	 */
   1008 	IPSEC_SPLASSERT_SOFTNET("key_checkrequest");
   1009 
   1010 	sav = key_lookup_sa_bysaidx(saidx);
   1011 	if (sav != NULL) {
   1012 		*ret = sav;
   1013 		return 0;
   1014 	}
   1015 
   1016 	/* there is no SA */
   1017 	error = key_acquire(saidx, isr->sp, M_NOWAIT);
   1018 	if (error != 0) {
   1019 		/* XXX What should I do ? */
   1020 		IPSECLOG(LOG_DEBUG, "error %d returned from key_acquire.\n",
   1021 		    error);
   1022 		return error;
   1023 	}
   1024 
   1025 	if (level != IPSEC_LEVEL_REQUIRE) {
   1026 		/* XXX sigh, the interface to this routine is botched */
   1027 		*ret = NULL;
   1028 		return 0;
   1029 	} else {
   1030 		return ENOENT;
   1031 	}
   1032 }
   1033 
   1034 /*
   1035  * looking up a SA for policy entry from SAD.
   1036  * NOTE: searching SAD of aliving state.
   1037  * OUT:	NULL:	not found.
   1038  *	others:	found and return the pointer.
   1039  */
   1040 struct secasvar *
   1041 key_lookup_sa_bysaidx(const struct secasindex *saidx)
   1042 {
   1043 	struct secashead *sah;
   1044 	struct secasvar *sav = NULL;
   1045 	u_int stateidx, state;
   1046 	const u_int *saorder_state_valid;
   1047 	int arraysize;
   1048 	int s;
   1049 
   1050 	s = pserialize_read_enter();
   1051 	sah = key_getsah(saidx, CMP_MODE_REQID);
   1052 	if (sah == NULL)
   1053 		goto out;
   1054 
   1055 	/*
   1056 	 * search a valid state list for outbound packet.
   1057 	 * This search order is important.
   1058 	 */
   1059 	if (key_prefered_oldsa) {
   1060 		saorder_state_valid = saorder_state_valid_prefer_old;
   1061 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_old);
   1062 	} else {
   1063 		saorder_state_valid = saorder_state_valid_prefer_new;
   1064 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_new);
   1065 	}
   1066 
   1067 	/* search valid state */
   1068 	for (stateidx = 0;
   1069 	     stateidx < arraysize;
   1070 	     stateidx++) {
   1071 
   1072 		state = saorder_state_valid[stateidx];
   1073 
   1074 		if (key_prefered_oldsa)
   1075 			sav = SAVLIST_READER_FIRST(sah, state);
   1076 		else {
   1077 			/* XXX need O(1) lookup */
   1078 			struct secasvar *last = NULL;
   1079 
   1080 			SAVLIST_READER_FOREACH(sav, sah, state)
   1081 				last = sav;
   1082 			sav = last;
   1083 		}
   1084 		if (sav != NULL) {
   1085 			KEY_SA_REF(sav);
   1086 			KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1087 			    "DP cause refcnt++:%d SA:%p\n",
   1088 			    key_sa_refcnt(sav), sav);
   1089 			break;
   1090 		}
   1091 	}
   1092 out:
   1093 	pserialize_read_exit(s);
   1094 
   1095 	return sav;
   1096 }
   1097 
   1098 #if 0
   1099 static void
   1100 key_sendup_message_delete(struct secasvar *sav)
   1101 {
   1102 	struct mbuf *m, *result = 0;
   1103 	uint8_t satype;
   1104 
   1105 	satype = key_proto2satype(sav->sah->saidx.proto);
   1106 	if (satype == 0)
   1107 		goto msgfail;
   1108 
   1109 	m = key_setsadbmsg(SADB_DELETE, 0, satype, 0, 0, key_sa_refcnt(sav) - 1);
   1110 	if (m == NULL)
   1111 		goto msgfail;
   1112 	result = m;
   1113 
   1114 	/* set sadb_address for saidx's. */
   1115 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa,
   1116 	    sav->sah->saidx.src.sa.sa_len << 3, IPSEC_ULPROTO_ANY);
   1117 	if (m == NULL)
   1118 		goto msgfail;
   1119 	m_cat(result, m);
   1120 
   1121 	/* set sadb_address for saidx's. */
   1122 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.src.sa,
   1123 	    sav->sah->saidx.src.sa.sa_len << 3, IPSEC_ULPROTO_ANY);
   1124 	if (m == NULL)
   1125 		goto msgfail;
   1126 	m_cat(result, m);
   1127 
   1128 	/* create SA extension */
   1129 	m = key_setsadbsa(sav);
   1130 	if (m == NULL)
   1131 		goto msgfail;
   1132 	m_cat(result, m);
   1133 
   1134 	if (result->m_len < sizeof(struct sadb_msg)) {
   1135 		result = m_pullup(result, sizeof(struct sadb_msg));
   1136 		if (result == NULL)
   1137 			goto msgfail;
   1138 	}
   1139 
   1140 	result->m_pkthdr.len = 0;
   1141 	for (m = result; m; m = m->m_next)
   1142 		result->m_pkthdr.len += m->m_len;
   1143 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   1144 	    PFKEY_UNIT64(result->m_pkthdr.len);
   1145 
   1146 	key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
   1147 	result = NULL;
   1148 msgfail:
   1149 	if (result)
   1150 		m_freem(result);
   1151 }
   1152 #endif
   1153 
   1154 /*
   1155  * allocating a usable SA entry for a *INBOUND* packet.
   1156  * Must call key_freesav() later.
   1157  * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
   1158  *	NULL:		not found, or error occurred.
   1159  *
   1160  * In the comparison, no source address is used--for RFC2401 conformance.
   1161  * To quote, from section 4.1:
   1162  *	A security association is uniquely identified by a triple consisting
   1163  *	of a Security Parameter Index (SPI), an IP Destination Address, and a
   1164  *	security protocol (AH or ESP) identifier.
   1165  * Note that, however, we do need to keep source address in IPsec SA.
   1166  * IKE specification and PF_KEY specification do assume that we
   1167  * keep source address in IPsec SA.  We see a tricky situation here.
   1168  *
   1169  * sport and dport are used for NAT-T. network order is always used.
   1170  */
   1171 struct secasvar *
   1172 key_lookup_sa(
   1173 	const union sockaddr_union *dst,
   1174 	u_int proto,
   1175 	u_int32_t spi,
   1176 	u_int16_t sport,
   1177 	u_int16_t dport,
   1178 	const char* where, int tag)
   1179 {
   1180 	struct secashead *sah;
   1181 	struct secasvar *sav;
   1182 	u_int stateidx, state;
   1183 	const u_int *saorder_state_valid;
   1184 	int arraysize, chkport;
   1185 	int s;
   1186 
   1187 	int must_check_spi = 1;
   1188 	int must_check_alg = 0;
   1189 	u_int16_t cpi = 0;
   1190 	u_int8_t algo = 0;
   1191 
   1192 	if ((sport != 0) && (dport != 0))
   1193 		chkport = PORT_STRICT;
   1194 	else
   1195 		chkport = PORT_NONE;
   1196 
   1197 	KASSERT(dst != NULL);
   1198 
   1199 	/*
   1200 	 * XXX IPCOMP case
   1201 	 * We use cpi to define spi here. In the case where cpi <=
   1202 	 * IPCOMP_CPI_NEGOTIATE_MIN, cpi just define the algorithm used, not
   1203 	 * the real spi. In this case, don't check the spi but check the
   1204 	 * algorithm
   1205 	 */
   1206 
   1207 	if (proto == IPPROTO_IPCOMP) {
   1208 		u_int32_t tmp;
   1209 		tmp = ntohl(spi);
   1210 		cpi = (u_int16_t) tmp;
   1211 		if (cpi < IPCOMP_CPI_NEGOTIATE_MIN) {
   1212 			algo = (u_int8_t) cpi;
   1213 			must_check_spi = 0;
   1214 			must_check_alg = 1;
   1215 		}
   1216 	}
   1217 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1218 	    "DP from %s:%u check_spi=%d, check_alg=%d\n",
   1219 	    where, tag, must_check_spi, must_check_alg);
   1220 
   1221 
   1222 	/*
   1223 	 * searching SAD.
   1224 	 * XXX: to be checked internal IP header somewhere.  Also when
   1225 	 * IPsec tunnel packet is received.  But ESP tunnel mode is
   1226 	 * encrypted so we can't check internal IP header.
   1227 	 */
   1228 	if (key_prefered_oldsa) {
   1229 		saorder_state_valid = saorder_state_valid_prefer_old;
   1230 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_old);
   1231 	} else {
   1232 		saorder_state_valid = saorder_state_valid_prefer_new;
   1233 		arraysize = _ARRAYLEN(saorder_state_valid_prefer_new);
   1234 	}
   1235 	s = pserialize_read_enter();
   1236 	SAHLIST_READER_FOREACH(sah) {
   1237 		/* search valid state */
   1238 		for (stateidx = 0; stateidx < arraysize; stateidx++) {
   1239 			state = saorder_state_valid[stateidx];
   1240 			SAVLIST_READER_FOREACH(sav, sah, state) {
   1241 				KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   1242 				    "try match spi %#x, %#x\n",
   1243 				    ntohl(spi), ntohl(sav->spi));
   1244 				/* sanity check */
   1245 				KEY_CHKSASTATE(sav->state, state);
   1246 				/* do not return entries w/ unusable state */
   1247 				if (!SADB_SASTATE_USABLE_P(sav)) {
   1248 					KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   1249 					    "bad state %d\n", sav->state);
   1250 					continue;
   1251 				}
   1252 				if (proto != sav->sah->saidx.proto) {
   1253 					KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   1254 					    "proto fail %d != %d\n",
   1255 					    proto, sav->sah->saidx.proto);
   1256 					continue;
   1257 				}
   1258 				if (must_check_spi && spi != sav->spi) {
   1259 					KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   1260 					    "spi fail %#x != %#x\n",
   1261 					    ntohl(spi), ntohl(sav->spi));
   1262 					continue;
   1263 				}
   1264 				/* XXX only on the ipcomp case */
   1265 				if (must_check_alg && algo != sav->alg_comp) {
   1266 					KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   1267 					    "algo fail %d != %d\n",
   1268 					    algo, sav->alg_comp);
   1269 					continue;
   1270 				}
   1271 
   1272 #if 0	/* don't check src */
   1273 	/* Fix port in src->sa */
   1274 
   1275 				/* check src address */
   1276 				if (!key_sockaddr_match(&src->sa, &sav->sah->saidx.src.sa, PORT_NONE))
   1277 					continue;
   1278 #endif
   1279 				/* fix port of dst address XXX*/
   1280 				key_porttosaddr(__UNCONST(dst), dport);
   1281 				/* check dst address */
   1282 				if (!key_sockaddr_match(&dst->sa, &sav->sah->saidx.dst.sa, chkport))
   1283 					continue;
   1284 				key_sa_ref(sav, where, tag);
   1285 				goto done;
   1286 			}
   1287 		}
   1288 	}
   1289 	sav = NULL;
   1290 done:
   1291 	pserialize_read_exit(s);
   1292 
   1293 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1294 	    "DP return SA:%p; refcnt %u\n", sav, key_sa_refcnt(sav));
   1295 	return sav;
   1296 }
   1297 
   1298 static void
   1299 key_validate_savlist(const struct secashead *sah, const u_int state)
   1300 {
   1301 #ifdef DEBUG
   1302 	struct secasvar *sav, *next;
   1303 	int s;
   1304 
   1305 	/*
   1306 	 * The list should be sorted by lft_c->sadb_lifetime_addtime
   1307 	 * in ascending order.
   1308 	 */
   1309 	s = pserialize_read_enter();
   1310 	SAVLIST_READER_FOREACH(sav, sah, state) {
   1311 		next = SAVLIST_READER_NEXT(sav);
   1312 		if (next != NULL &&
   1313 		    sav->lft_c != NULL && next->lft_c != NULL) {
   1314 			KDASSERTMSG(sav->lft_c->sadb_lifetime_addtime <=
   1315 			    next->lft_c->sadb_lifetime_addtime,
   1316 			    "savlist is not sorted: sah=%p, state=%d, "
   1317 			    "sav=%" PRIu64 ", next=%" PRIu64, sah, state,
   1318 			    sav->lft_c->sadb_lifetime_addtime,
   1319 			    next->lft_c->sadb_lifetime_addtime);
   1320 		}
   1321 	}
   1322 	pserialize_read_exit(s);
   1323 #endif
   1324 }
   1325 
   1326 void
   1327 key_init_sp(struct secpolicy *sp)
   1328 {
   1329 
   1330 	ASSERT_SLEEPABLE();
   1331 
   1332 	sp->state = IPSEC_SPSTATE_ALIVE;
   1333 	if (sp->policy == IPSEC_POLICY_IPSEC)
   1334 		KASSERT(sp->req != NULL);
   1335 	localcount_init(&sp->localcount);
   1336 	SPLIST_ENTRY_INIT(sp);
   1337 }
   1338 
   1339 /*
   1340  * Must be called in a pserialize read section. A held SP
   1341  * must be released by key_sp_unref after use.
   1342  */
   1343 void
   1344 key_sp_ref(struct secpolicy *sp, const char* where, int tag)
   1345 {
   1346 
   1347 	localcount_acquire(&sp->localcount);
   1348 
   1349 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1350 	    "DP SP:%p (ID=%u) from %s:%u; refcnt++ now %u\n",
   1351 	    sp, sp->id, where, tag, key_sp_refcnt(sp));
   1352 }
   1353 
   1354 /*
   1355  * Must be called without holding key_spd.lock because the lock
   1356  * would be held in localcount_release.
   1357  */
   1358 void
   1359 key_sp_unref(struct secpolicy *sp, const char* where, int tag)
   1360 {
   1361 
   1362 	KDASSERT(mutex_ownable(&key_spd.lock));
   1363 
   1364 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1365 	    "DP SP:%p (ID=%u) from %s:%u; refcnt-- now %u\n",
   1366 	    sp, sp->id, where, tag, key_sp_refcnt(sp));
   1367 
   1368 	localcount_release(&sp->localcount, &key_spd.cv_lc, &key_spd.lock);
   1369 }
   1370 
   1371 static void
   1372 key_init_sav(struct secasvar *sav)
   1373 {
   1374 
   1375 	ASSERT_SLEEPABLE();
   1376 
   1377 	localcount_init(&sav->localcount);
   1378 	SAVLIST_ENTRY_INIT(sav);
   1379 }
   1380 
   1381 u_int
   1382 key_sa_refcnt(const struct secasvar *sav)
   1383 {
   1384 
   1385 	/* FIXME */
   1386 	return 0;
   1387 }
   1388 
   1389 void
   1390 key_sa_ref(struct secasvar *sav, const char* where, int tag)
   1391 {
   1392 
   1393 	localcount_acquire(&sav->localcount);
   1394 
   1395 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1396 	    "DP cause refcnt++: SA:%p from %s:%u\n",
   1397 	    sav, where, tag);
   1398 }
   1399 
   1400 void
   1401 key_sa_unref(struct secasvar *sav, const char* where, int tag)
   1402 {
   1403 
   1404 	KDASSERT(mutex_ownable(&key_sad.lock));
   1405 
   1406 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1407 	    "DP cause refcnt--: SA:%p from %s:%u\n",
   1408 	    sav, where, tag);
   1409 
   1410 	localcount_release(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
   1411 }
   1412 
   1413 #if 0
   1414 /*
   1415  * Must be called after calling key_lookup_sp*().
   1416  * For the packet with socket.
   1417  */
   1418 static void
   1419 key_freeso(struct socket *so)
   1420 {
   1421 	/* sanity check */
   1422 	KASSERT(so != NULL);
   1423 
   1424 	switch (so->so_proto->pr_domain->dom_family) {
   1425 #ifdef INET
   1426 	case PF_INET:
   1427 	    {
   1428 		struct inpcb *pcb = sotoinpcb(so);
   1429 
   1430 		/* Does it have a PCB ? */
   1431 		if (pcb == NULL)
   1432 			return;
   1433 
   1434 		struct inpcbpolicy *sp = pcb->inp_sp;
   1435 		key_freesp_so(&sp->sp_in);
   1436 		key_freesp_so(&sp->sp_out);
   1437 	    }
   1438 		break;
   1439 #endif
   1440 #ifdef INET6
   1441 	case PF_INET6:
   1442 	    {
   1443 #ifdef HAVE_NRL_INPCB
   1444 		struct inpcb *pcb  = sotoinpcb(so);
   1445 		struct inpcbpolicy *sp = pcb->inp_sp;
   1446 
   1447 		/* Does it have a PCB ? */
   1448 		if (pcb == NULL)
   1449 			return;
   1450 		key_freesp_so(&sp->sp_in);
   1451 		key_freesp_so(&sp->sp_out);
   1452 #else
   1453 		struct in6pcb *pcb  = sotoin6pcb(so);
   1454 
   1455 		/* Does it have a PCB ? */
   1456 		if (pcb == NULL)
   1457 			return;
   1458 		key_freesp_so(&pcb->in6p_sp->sp_in);
   1459 		key_freesp_so(&pcb->in6p_sp->sp_out);
   1460 #endif
   1461 	    }
   1462 		break;
   1463 #endif /* INET6 */
   1464 	default:
   1465 		IPSECLOG(LOG_DEBUG, "unknown address family=%d.\n",
   1466 		    so->so_proto->pr_domain->dom_family);
   1467 		return;
   1468 	}
   1469 }
   1470 
   1471 static void
   1472 key_freesp_so(struct secpolicy **sp)
   1473 {
   1474 
   1475 	KASSERT(sp != NULL);
   1476 	KASSERT(*sp != NULL);
   1477 
   1478 	if ((*sp)->policy == IPSEC_POLICY_ENTRUST ||
   1479 	    (*sp)->policy == IPSEC_POLICY_BYPASS)
   1480 		return;
   1481 
   1482 	KASSERTMSG((*sp)->policy == IPSEC_POLICY_IPSEC,
   1483 	    "invalid policy %u", (*sp)->policy);
   1484 	KEY_SP_UNREF(&sp);
   1485 }
   1486 #endif
   1487 
   1488 static void
   1489 key_sad_pserialize_perform(void)
   1490 {
   1491 
   1492 	KASSERT(mutex_owned(&key_sad.lock));
   1493 
   1494 	while (key_sad.psz_performing)
   1495 		cv_wait(&key_sad.cv_psz, &key_sad.lock);
   1496 	key_sad.psz_performing = true;
   1497 	mutex_exit(&key_sad.lock);
   1498 
   1499 	pserialize_perform(key_sad.psz);
   1500 
   1501 	mutex_enter(&key_sad.lock);
   1502 	key_sad.psz_performing = false;
   1503 	cv_broadcast(&key_sad.cv_psz);
   1504 }
   1505 
   1506 /*
   1507  * Remove the sav from the savlist of its sah and wait for references to the sav
   1508  * to be released. key_sad.lock must be held.
   1509  */
   1510 static void
   1511 key_unlink_sav(struct secasvar *sav)
   1512 {
   1513 
   1514 	KASSERT(mutex_owned(&key_sad.lock));
   1515 
   1516 	SAVLIST_WRITER_REMOVE(sav);
   1517 
   1518 	KDASSERT(mutex_ownable(softnet_lock));
   1519 	key_sad_pserialize_perform();
   1520 
   1521 	localcount_drain(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
   1522 }
   1523 
   1524 /*
   1525  * Destroy an sav where the sav must be unlinked from an sah
   1526  * by say key_unlink_sav.
   1527  */
   1528 static void
   1529 key_destroy_sav(struct secasvar *sav)
   1530 {
   1531 
   1532 	ASSERT_SLEEPABLE();
   1533 
   1534 	localcount_fini(&sav->localcount);
   1535 	SAVLIST_ENTRY_DESTROY(sav);
   1536 
   1537 	key_delsav(sav);
   1538 }
   1539 
   1540 /*
   1541  * Destroy sav with holding its reference.
   1542  */
   1543 static void
   1544 key_destroy_sav_with_ref(struct secasvar *sav)
   1545 {
   1546 
   1547 	ASSERT_SLEEPABLE();
   1548 
   1549 	mutex_enter(&key_sad.lock);
   1550 	sav->state = SADB_SASTATE_DEAD;
   1551 	SAVLIST_WRITER_REMOVE(sav);
   1552 	mutex_exit(&key_sad.lock);
   1553 
   1554 	/* We cannot unref with holding key_sad.lock */
   1555 	KEY_SA_UNREF(&sav);
   1556 
   1557 	mutex_enter(&key_sad.lock);
   1558 	KDASSERT(mutex_ownable(softnet_lock));
   1559 	key_sad_pserialize_perform();
   1560 	localcount_drain(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
   1561 	mutex_exit(&key_sad.lock);
   1562 
   1563 	key_destroy_sav(sav);
   1564 }
   1565 
   1566 /* %%% SPD management */
   1567 /*
   1568  * free security policy entry.
   1569  */
   1570 static void
   1571 key_destroy_sp(struct secpolicy *sp)
   1572 {
   1573 
   1574 	SPLIST_ENTRY_DESTROY(sp);
   1575 	localcount_fini(&sp->localcount);
   1576 
   1577 	key_free_sp(sp);
   1578 
   1579 	key_update_used();
   1580 }
   1581 
   1582 void
   1583 key_free_sp(struct secpolicy *sp)
   1584 {
   1585 	struct ipsecrequest *isr = sp->req, *nextisr;
   1586 
   1587 	while (isr != NULL) {
   1588 		nextisr = isr->next;
   1589 		kmem_free(isr, sizeof(*isr));
   1590 		isr = nextisr;
   1591 	}
   1592 
   1593 	kmem_free(sp, sizeof(*sp));
   1594 }
   1595 
   1596 void
   1597 key_socksplist_add(struct secpolicy *sp)
   1598 {
   1599 
   1600 	mutex_enter(&key_spd.lock);
   1601 	PSLIST_WRITER_INSERT_HEAD(&key_spd.socksplist, sp, pslist_entry);
   1602 	mutex_exit(&key_spd.lock);
   1603 
   1604 	key_update_used();
   1605 }
   1606 
   1607 /*
   1608  * search SPD
   1609  * OUT:	NULL	: not found
   1610  *	others	: found, pointer to a SP.
   1611  */
   1612 static struct secpolicy *
   1613 key_getsp(const struct secpolicyindex *spidx)
   1614 {
   1615 	struct secpolicy *sp;
   1616 	int s;
   1617 
   1618 	KASSERT(spidx != NULL);
   1619 
   1620 	s = pserialize_read_enter();
   1621 	SPLIST_READER_FOREACH(sp, spidx->dir) {
   1622 		if (sp->state == IPSEC_SPSTATE_DEAD)
   1623 			continue;
   1624 		if (key_spidx_match_exactly(spidx, &sp->spidx)) {
   1625 			KEY_SP_REF(sp);
   1626 			pserialize_read_exit(s);
   1627 			return sp;
   1628 		}
   1629 	}
   1630 	pserialize_read_exit(s);
   1631 
   1632 	return NULL;
   1633 }
   1634 
   1635 /*
   1636  * search SPD and remove found SP
   1637  * OUT:	NULL	: not found
   1638  *	others	: found, pointer to a SP.
   1639  */
   1640 static struct secpolicy *
   1641 key_lookup_and_remove_sp(const struct secpolicyindex *spidx, bool from_kernel)
   1642 {
   1643 	struct secpolicy *sp = NULL;
   1644 
   1645 	mutex_enter(&key_spd.lock);
   1646 	SPLIST_WRITER_FOREACH(sp, spidx->dir) {
   1647 		KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
   1648 		/*
   1649 		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
   1650 		 * removed by userland programs.
   1651 		 */
   1652 		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
   1653 			continue;
   1654 		if (key_spidx_match_exactly(spidx, &sp->spidx)) {
   1655 			key_unlink_sp(sp);
   1656 			goto out;
   1657 		}
   1658 	}
   1659 	sp = NULL;
   1660 out:
   1661 	mutex_exit(&key_spd.lock);
   1662 
   1663 	return sp;
   1664 }
   1665 
   1666 /*
   1667  * get SP by index.
   1668  * OUT:	NULL	: not found
   1669  *	others	: found, pointer to a SP.
   1670  */
   1671 static struct secpolicy *
   1672 key_getspbyid(u_int32_t id)
   1673 {
   1674 	struct secpolicy *sp;
   1675 	int s;
   1676 
   1677 	s = pserialize_read_enter();
   1678 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_INBOUND) {
   1679 		if (sp->state == IPSEC_SPSTATE_DEAD)
   1680 			continue;
   1681 		if (sp->id == id) {
   1682 			KEY_SP_REF(sp);
   1683 			goto out;
   1684 		}
   1685 	}
   1686 
   1687 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_OUTBOUND) {
   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 out:
   1696 	pserialize_read_exit(s);
   1697 	return sp;
   1698 }
   1699 
   1700 /*
   1701  * get SP by index, remove and return it.
   1702  * OUT:	NULL	: not found
   1703  *	others	: found, pointer to a SP.
   1704  */
   1705 static struct secpolicy *
   1706 key_lookupbyid_and_remove_sp(u_int32_t id, bool from_kernel)
   1707 {
   1708 	struct secpolicy *sp;
   1709 
   1710 	mutex_enter(&key_spd.lock);
   1711 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_INBOUND) {
   1712 		KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
   1713 		/*
   1714 		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
   1715 		 * removed by userland programs.
   1716 		 */
   1717 		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
   1718 			continue;
   1719 		if (sp->id == id)
   1720 			goto out;
   1721 	}
   1722 
   1723 	SPLIST_READER_FOREACH(sp, IPSEC_DIR_OUTBOUND) {
   1724 		KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
   1725 		/*
   1726 		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
   1727 		 * removed by userland programs.
   1728 		 */
   1729 		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
   1730 			continue;
   1731 		if (sp->id == id)
   1732 			goto out;
   1733 	}
   1734 out:
   1735 	if (sp != NULL)
   1736 		key_unlink_sp(sp);
   1737 	mutex_exit(&key_spd.lock);
   1738 	return sp;
   1739 }
   1740 
   1741 struct secpolicy *
   1742 key_newsp(const char* where, int tag)
   1743 {
   1744 	struct secpolicy *newsp = NULL;
   1745 
   1746 	newsp = kmem_zalloc(sizeof(struct secpolicy), KM_SLEEP);
   1747 
   1748 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   1749 	    "DP from %s:%u return SP:%p\n", where, tag, newsp);
   1750 	return newsp;
   1751 }
   1752 
   1753 /*
   1754  * create secpolicy structure from sadb_x_policy structure.
   1755  * NOTE: `state', `secpolicyindex' in secpolicy structure are not set,
   1756  * so must be set properly later.
   1757  */
   1758 static struct secpolicy *
   1759 _key_msg2sp(const struct sadb_x_policy *xpl0, size_t len, int *error,
   1760     bool from_kernel)
   1761 {
   1762 	struct secpolicy *newsp;
   1763 
   1764 	KASSERT(!cpu_softintr_p());
   1765 	KASSERT(xpl0 != NULL);
   1766 	KASSERT(len >= sizeof(*xpl0));
   1767 
   1768 	if (len != PFKEY_EXTLEN(xpl0)) {
   1769 		IPSECLOG(LOG_DEBUG, "Invalid msg length.\n");
   1770 		*error = EINVAL;
   1771 		return NULL;
   1772 	}
   1773 
   1774 	newsp = KEY_NEWSP();
   1775 	if (newsp == NULL) {
   1776 		*error = ENOBUFS;
   1777 		return NULL;
   1778 	}
   1779 
   1780 	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
   1781 	newsp->policy = xpl0->sadb_x_policy_type;
   1782 
   1783 	/* check policy */
   1784 	switch (xpl0->sadb_x_policy_type) {
   1785 	case IPSEC_POLICY_DISCARD:
   1786 	case IPSEC_POLICY_NONE:
   1787 	case IPSEC_POLICY_ENTRUST:
   1788 	case IPSEC_POLICY_BYPASS:
   1789 		newsp->req = NULL;
   1790 		*error = 0;
   1791 		return newsp;
   1792 
   1793 	case IPSEC_POLICY_IPSEC:
   1794 		/* Continued */
   1795 		break;
   1796 	default:
   1797 		IPSECLOG(LOG_DEBUG, "invalid policy type.\n");
   1798 		key_free_sp(newsp);
   1799 		*error = EINVAL;
   1800 		return NULL;
   1801 	}
   1802 
   1803 	/* IPSEC_POLICY_IPSEC */
   1804     {
   1805 	int tlen;
   1806 	const struct sadb_x_ipsecrequest *xisr;
   1807 	uint16_t xisr_reqid;
   1808 	struct ipsecrequest **p_isr = &newsp->req;
   1809 
   1810 	/* validity check */
   1811 	if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
   1812 		IPSECLOG(LOG_DEBUG, "Invalid msg length.\n");
   1813 		*error = EINVAL;
   1814 		goto free_exit;
   1815 	}
   1816 
   1817 	tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
   1818 	xisr = (const struct sadb_x_ipsecrequest *)(xpl0 + 1);
   1819 
   1820 	while (tlen > 0) {
   1821 		/* length check */
   1822 		if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr)) {
   1823 			IPSECLOG(LOG_DEBUG, "invalid ipsecrequest length.\n");
   1824 			*error = EINVAL;
   1825 			goto free_exit;
   1826 		}
   1827 
   1828 		/* allocate request buffer */
   1829 		*p_isr = kmem_zalloc(sizeof(**p_isr), KM_SLEEP);
   1830 
   1831 		/* set values */
   1832 		(*p_isr)->next = NULL;
   1833 
   1834 		switch (xisr->sadb_x_ipsecrequest_proto) {
   1835 		case IPPROTO_ESP:
   1836 		case IPPROTO_AH:
   1837 		case IPPROTO_IPCOMP:
   1838 			break;
   1839 		default:
   1840 			IPSECLOG(LOG_DEBUG, "invalid proto type=%u\n",
   1841 			    xisr->sadb_x_ipsecrequest_proto);
   1842 			*error = EPROTONOSUPPORT;
   1843 			goto free_exit;
   1844 		}
   1845 		(*p_isr)->saidx.proto = xisr->sadb_x_ipsecrequest_proto;
   1846 
   1847 		switch (xisr->sadb_x_ipsecrequest_mode) {
   1848 		case IPSEC_MODE_TRANSPORT:
   1849 		case IPSEC_MODE_TUNNEL:
   1850 			break;
   1851 		case IPSEC_MODE_ANY:
   1852 		default:
   1853 			IPSECLOG(LOG_DEBUG, "invalid mode=%u\n",
   1854 			    xisr->sadb_x_ipsecrequest_mode);
   1855 			*error = EINVAL;
   1856 			goto free_exit;
   1857 		}
   1858 		(*p_isr)->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
   1859 
   1860 		switch (xisr->sadb_x_ipsecrequest_level) {
   1861 		case IPSEC_LEVEL_DEFAULT:
   1862 		case IPSEC_LEVEL_USE:
   1863 		case IPSEC_LEVEL_REQUIRE:
   1864 			break;
   1865 		case IPSEC_LEVEL_UNIQUE:
   1866 			xisr_reqid = xisr->sadb_x_ipsecrequest_reqid;
   1867 			/* validity check */
   1868 			/*
   1869 			 * case 1) from_kernel == false
   1870 			 * That means the request comes from userland.
   1871 			 * If range violation of reqid, kernel will
   1872 			 * update it, don't refuse it.
   1873 			 *
   1874 			 * case 2) from_kernel == true
   1875 			 * That means the request comes from kernel
   1876 			 * (e.g. ipsec(4) I/F).
   1877 			 * Use thre requested reqid to avoid inconsistency
   1878 			 * between kernel's reqid and the reqid in pf_key
   1879 			 * message sent to userland. The pf_key message is
   1880 			 * built by diverting request mbuf.
   1881 			 */
   1882 			if (!from_kernel &&
   1883 			    xisr_reqid > IPSEC_MANUAL_REQID_MAX) {
   1884 				IPSECLOG(LOG_DEBUG,
   1885 				    "reqid=%d range "
   1886 				    "violation, updated by kernel.\n",
   1887 				    xisr_reqid);
   1888 				xisr_reqid = 0;
   1889 			}
   1890 
   1891 			/* allocate new reqid id if reqid is zero. */
   1892 			if (xisr_reqid == 0) {
   1893 				u_int16_t reqid = key_newreqid();
   1894 				if (reqid == 0) {
   1895 					*error = ENOBUFS;
   1896 					goto free_exit;
   1897 				}
   1898 				(*p_isr)->saidx.reqid = reqid;
   1899 			} else {
   1900 			/* set it for manual keying. */
   1901 				(*p_isr)->saidx.reqid = xisr_reqid;
   1902 			}
   1903 			break;
   1904 
   1905 		default:
   1906 			IPSECLOG(LOG_DEBUG, "invalid level=%u\n",
   1907 			    xisr->sadb_x_ipsecrequest_level);
   1908 			*error = EINVAL;
   1909 			goto free_exit;
   1910 		}
   1911 		(*p_isr)->level = xisr->sadb_x_ipsecrequest_level;
   1912 
   1913 		/* set IP addresses if there */
   1914 		if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
   1915 			const struct sockaddr *paddr;
   1916 
   1917 			paddr = (const struct sockaddr *)(xisr + 1);
   1918 
   1919 			/* validity check */
   1920 			if (paddr->sa_len > sizeof((*p_isr)->saidx.src)) {
   1921 				IPSECLOG(LOG_DEBUG, "invalid request "
   1922 				    "address length.\n");
   1923 				*error = EINVAL;
   1924 				goto free_exit;
   1925 			}
   1926 			memcpy(&(*p_isr)->saidx.src, paddr, paddr->sa_len);
   1927 
   1928 			paddr = (const struct sockaddr *)((const char *)paddr
   1929 			    + paddr->sa_len);
   1930 
   1931 			/* validity check */
   1932 			if (paddr->sa_len > sizeof((*p_isr)->saidx.dst)) {
   1933 				IPSECLOG(LOG_DEBUG, "invalid request "
   1934 				    "address length.\n");
   1935 				*error = EINVAL;
   1936 				goto free_exit;
   1937 			}
   1938 			memcpy(&(*p_isr)->saidx.dst, paddr, paddr->sa_len);
   1939 		}
   1940 
   1941 		(*p_isr)->sp = newsp;
   1942 
   1943 		/* initialization for the next. */
   1944 		p_isr = &(*p_isr)->next;
   1945 		tlen -= xisr->sadb_x_ipsecrequest_len;
   1946 
   1947 		/* validity check */
   1948 		if (tlen < 0) {
   1949 			IPSECLOG(LOG_DEBUG, "becoming tlen < 0.\n");
   1950 			*error = EINVAL;
   1951 			goto free_exit;
   1952 		}
   1953 
   1954 		xisr = (const struct sadb_x_ipsecrequest *)((const char *)xisr +
   1955 		    xisr->sadb_x_ipsecrequest_len);
   1956 	}
   1957     }
   1958 
   1959 	*error = 0;
   1960 	return newsp;
   1961 
   1962 free_exit:
   1963 	key_free_sp(newsp);
   1964 	return NULL;
   1965 }
   1966 
   1967 struct secpolicy *
   1968 key_msg2sp(const struct sadb_x_policy *xpl0, size_t len, int *error)
   1969 {
   1970 
   1971 	return _key_msg2sp(xpl0, len, error, false);
   1972 }
   1973 
   1974 u_int16_t
   1975 key_newreqid(void)
   1976 {
   1977 	static u_int16_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
   1978 
   1979 	auto_reqid = (auto_reqid == 0xffff ?
   1980 	    IPSEC_MANUAL_REQID_MAX + 1 : auto_reqid + 1);
   1981 
   1982 	/* XXX should be unique check */
   1983 
   1984 	return auto_reqid;
   1985 }
   1986 
   1987 /*
   1988  * copy secpolicy struct to sadb_x_policy structure indicated.
   1989  */
   1990 struct mbuf *
   1991 key_sp2msg(const struct secpolicy *sp, int mflag)
   1992 {
   1993 	struct sadb_x_policy *xpl;
   1994 	int tlen;
   1995 	char *p;
   1996 	struct mbuf *m;
   1997 
   1998 	KASSERT(sp != NULL);
   1999 
   2000 	tlen = key_getspreqmsglen(sp);
   2001 
   2002 	m = key_alloc_mbuf(tlen, mflag);
   2003 	if (!m || m->m_next) {	/*XXX*/
   2004 		if (m)
   2005 			m_freem(m);
   2006 		return NULL;
   2007 	}
   2008 
   2009 	m->m_len = tlen;
   2010 	m->m_next = NULL;
   2011 	xpl = mtod(m, struct sadb_x_policy *);
   2012 	memset(xpl, 0, tlen);
   2013 
   2014 	xpl->sadb_x_policy_len = PFKEY_UNIT64(tlen);
   2015 	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
   2016 	xpl->sadb_x_policy_type = sp->policy;
   2017 	xpl->sadb_x_policy_dir = sp->spidx.dir;
   2018 	xpl->sadb_x_policy_id = sp->id;
   2019 	p = (char *)xpl + sizeof(*xpl);
   2020 
   2021 	/* if is the policy for ipsec ? */
   2022 	if (sp->policy == IPSEC_POLICY_IPSEC) {
   2023 		struct sadb_x_ipsecrequest *xisr;
   2024 		struct ipsecrequest *isr;
   2025 
   2026 		for (isr = sp->req; isr != NULL; isr = isr->next) {
   2027 
   2028 			xisr = (struct sadb_x_ipsecrequest *)p;
   2029 
   2030 			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
   2031 			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
   2032 			xisr->sadb_x_ipsecrequest_level = isr->level;
   2033 			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
   2034 
   2035 			p += sizeof(*xisr);
   2036 			memcpy(p, &isr->saidx.src, isr->saidx.src.sa.sa_len);
   2037 			p += isr->saidx.src.sa.sa_len;
   2038 			memcpy(p, &isr->saidx.dst, isr->saidx.dst.sa.sa_len);
   2039 			p += isr->saidx.src.sa.sa_len;
   2040 
   2041 			xisr->sadb_x_ipsecrequest_len =
   2042 			    PFKEY_ALIGN8(sizeof(*xisr)
   2043 			    + isr->saidx.src.sa.sa_len
   2044 			    + isr->saidx.dst.sa.sa_len);
   2045 		}
   2046 	}
   2047 
   2048 	return m;
   2049 }
   2050 
   2051 /*
   2052  * m will not be freed nor modified. It never return NULL.
   2053  * If it returns a mbuf of M_PKTHDR, the mbuf ensures to have
   2054  * contiguous length at least sizeof(struct sadb_msg).
   2055  */
   2056 static struct mbuf *
   2057 key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
   2058 		int ndeep, int nitem, ...)
   2059 {
   2060 	va_list ap;
   2061 	int idx;
   2062 	int i;
   2063 	struct mbuf *result = NULL, *n;
   2064 	int len;
   2065 
   2066 	KASSERT(m != NULL);
   2067 	KASSERT(mhp != NULL);
   2068 	KASSERT(!cpu_softintr_p());
   2069 
   2070 	va_start(ap, nitem);
   2071 	for (i = 0; i < nitem; i++) {
   2072 		idx = va_arg(ap, int);
   2073 		KASSERT(idx >= 0);
   2074 		KASSERT(idx <= SADB_EXT_MAX);
   2075 		/* don't attempt to pull empty extension */
   2076 		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
   2077 			continue;
   2078 		if (idx != SADB_EXT_RESERVED &&
   2079 		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
   2080 			continue;
   2081 
   2082 		if (idx == SADB_EXT_RESERVED) {
   2083 			CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MHLEN);
   2084 			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
   2085 			MGETHDR(n, M_WAITOK, MT_DATA);
   2086 			n->m_len = len;
   2087 			n->m_next = NULL;
   2088 			m_copydata(m, 0, sizeof(struct sadb_msg),
   2089 			    mtod(n, void *));
   2090 		} else if (i < ndeep) {
   2091 			len = mhp->extlen[idx];
   2092 			n = key_alloc_mbuf(len, M_WAITOK);
   2093 			KASSERT(n->m_next == NULL);
   2094 			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
   2095 			    mtod(n, void *));
   2096 		} else {
   2097 			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
   2098 			    M_WAITOK);
   2099 		}
   2100 		KASSERT(n != NULL);
   2101 
   2102 		if (result)
   2103 			m_cat(result, n);
   2104 		else
   2105 			result = n;
   2106 	}
   2107 	va_end(ap);
   2108 
   2109 	KASSERT(result != NULL);
   2110 	if ((result->m_flags & M_PKTHDR) != 0) {
   2111 		result->m_pkthdr.len = 0;
   2112 		for (n = result; n; n = n->m_next)
   2113 			result->m_pkthdr.len += n->m_len;
   2114 		KASSERT(result->m_len >= sizeof(struct sadb_msg));
   2115 	}
   2116 
   2117 	return result;
   2118 }
   2119 
   2120 /*
   2121  * The argument _sp must not overwrite until SP is created and registered
   2122  * successfully.
   2123  */
   2124 static int
   2125 key_spdadd(struct socket *so, struct mbuf *m,
   2126 	   const struct sadb_msghdr *mhp, struct secpolicy **_sp,
   2127 	   bool from_kernel)
   2128 {
   2129 	const struct sockaddr *src, *dst;
   2130 	const struct sadb_x_policy *xpl0;
   2131 	struct sadb_x_policy *xpl;
   2132 	const struct sadb_lifetime *lft = NULL;
   2133 	struct secpolicyindex spidx;
   2134 	struct secpolicy *newsp;
   2135 	int error;
   2136 	uint32_t sadb_x_policy_id;
   2137 
   2138 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   2139 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
   2140 	    mhp->ext[SADB_X_EXT_POLICY] == NULL) {
   2141 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2142 		return key_senderror(so, m, EINVAL);
   2143 	}
   2144 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   2145 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
   2146 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
   2147 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2148 		return key_senderror(so, m, EINVAL);
   2149 	}
   2150 	if (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL) {
   2151 		if (mhp->extlen[SADB_EXT_LIFETIME_HARD] <
   2152 		    sizeof(struct sadb_lifetime)) {
   2153 			IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2154 			return key_senderror(so, m, EINVAL);
   2155 		}
   2156 		lft = mhp->ext[SADB_EXT_LIFETIME_HARD];
   2157 	}
   2158 
   2159 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
   2160 
   2161 	/* checking the direciton. */
   2162 	switch (xpl0->sadb_x_policy_dir) {
   2163 	case IPSEC_DIR_INBOUND:
   2164 	case IPSEC_DIR_OUTBOUND:
   2165 		break;
   2166 	default:
   2167 		IPSECLOG(LOG_DEBUG, "Invalid SP direction.\n");
   2168 		return key_senderror(so, m, EINVAL);
   2169 	}
   2170 
   2171 	/* check policy */
   2172 	/* key_api_spdadd() accepts DISCARD, NONE and IPSEC. */
   2173 	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_ENTRUST ||
   2174 	    xpl0->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
   2175 		IPSECLOG(LOG_DEBUG, "Invalid policy type.\n");
   2176 		return key_senderror(so, m, EINVAL);
   2177 	}
   2178 
   2179 	/* policy requests are mandatory when action is ipsec. */
   2180 	if (mhp->msg->sadb_msg_type != SADB_X_SPDSETIDX &&
   2181 	    xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
   2182 	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
   2183 		IPSECLOG(LOG_DEBUG, "some policy requests part required.\n");
   2184 		return key_senderror(so, m, EINVAL);
   2185 	}
   2186 
   2187 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   2188 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   2189 
   2190 	/* sanity check on addr pair */
   2191 	if (src->sa_family != dst->sa_family)
   2192 		return key_senderror(so, m, EINVAL);
   2193 	if (src->sa_len != dst->sa_len)
   2194 		return key_senderror(so, m, EINVAL);
   2195 
   2196 	key_init_spidx_bymsghdr(&spidx, mhp);
   2197 
   2198 	/*
   2199 	 * checking there is SP already or not.
   2200 	 * SPDUPDATE doesn't depend on whether there is a SP or not.
   2201 	 * If the type is either SPDADD or SPDSETIDX AND a SP is found,
   2202 	 * then error.
   2203 	 */
   2204     {
   2205 	struct secpolicy *sp;
   2206 
   2207 	if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
   2208 		sp = key_lookup_and_remove_sp(&spidx, from_kernel);
   2209 		if (sp != NULL)
   2210 			key_destroy_sp(sp);
   2211 	} else {
   2212 		sp = key_getsp(&spidx);
   2213 		if (sp != NULL) {
   2214 			KEY_SP_UNREF(&sp);
   2215 			IPSECLOG(LOG_DEBUG, "a SP entry exists already.\n");
   2216 			return key_senderror(so, m, EEXIST);
   2217 		}
   2218 	}
   2219     }
   2220 
   2221 	/* allocation new SP entry */
   2222 	newsp = _key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error, from_kernel);
   2223 	if (newsp == NULL) {
   2224 		return key_senderror(so, m, error);
   2225 	}
   2226 
   2227 	newsp->id = key_getnewspid();
   2228 	if (newsp->id == 0) {
   2229 		kmem_free(newsp, sizeof(*newsp));
   2230 		return key_senderror(so, m, ENOBUFS);
   2231 	}
   2232 
   2233 	newsp->spidx = spidx;
   2234 	newsp->created = time_uptime;
   2235 	newsp->lastused = newsp->created;
   2236 	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
   2237 	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
   2238 	if (from_kernel)
   2239 		newsp->origin = IPSEC_SPORIGIN_KERNEL;
   2240 	else
   2241 		newsp->origin = IPSEC_SPORIGIN_USER;
   2242 
   2243 	key_init_sp(newsp);
   2244 	if (from_kernel)
   2245 		KEY_SP_REF(newsp);
   2246 
   2247 	sadb_x_policy_id = newsp->id;
   2248 
   2249 	if (_sp != NULL)
   2250 		*_sp = newsp;
   2251 
   2252 	mutex_enter(&key_spd.lock);
   2253 	SPLIST_WRITER_INSERT_TAIL(newsp->spidx.dir, newsp);
   2254 	mutex_exit(&key_spd.lock);
   2255 	/*
   2256 	 * We don't have a reference to newsp, so we must not touch newsp from
   2257 	 * now on.  If you want to do, you must take a reference beforehand.
   2258 	 */
   2259 	newsp = NULL;
   2260 
   2261 #ifdef notyet
   2262 	/* delete the entry in key_misc.spacqlist */
   2263 	if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
   2264 		struct secspacq *spacq = key_getspacq(&spidx);
   2265 		if (spacq != NULL) {
   2266 			/* reset counter in order to deletion by timehandler. */
   2267 			spacq->created = time_uptime;
   2268 			spacq->count = 0;
   2269 		}
   2270     	}
   2271 #endif
   2272 
   2273 	/* Invalidate all cached SPD pointers in the PCBs. */
   2274 	ipsec_invalpcbcacheall();
   2275 
   2276 #if defined(GATEWAY)
   2277 	/* Invalidate the ipflow cache, as well. */
   2278 	ipflow_invalidate_all(0);
   2279 #ifdef INET6
   2280 	if (in6_present)
   2281 		ip6flow_invalidate_all(0);
   2282 #endif /* INET6 */
   2283 #endif /* GATEWAY */
   2284 
   2285 	key_update_used();
   2286 
   2287     {
   2288 	struct mbuf *n, *mpolicy;
   2289 	int off;
   2290 
   2291 	/* create new sadb_msg to reply. */
   2292 	if (lft) {
   2293 		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
   2294 		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
   2295 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
   2296 	} else {
   2297 		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
   2298 		    SADB_X_EXT_POLICY,
   2299 		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
   2300 	}
   2301 
   2302 	key_fill_replymsg(n, 0);
   2303 	off = 0;
   2304 	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
   2305 	    sizeof(*xpl), &off);
   2306 	if (mpolicy == NULL) {
   2307 		/* n is already freed */
   2308 		/*
   2309 		 * valid sp has been created, so we does not overwrite _sp
   2310 		 * NULL here. let caller decide to use the sp or not.
   2311 		 */
   2312 		return key_senderror(so, m, ENOBUFS);
   2313 	}
   2314 	xpl = (struct sadb_x_policy *)(mtod(mpolicy, char *) + off);
   2315 	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
   2316 		m_freem(n);
   2317 		/* ditto */
   2318 		return key_senderror(so, m, EINVAL);
   2319 	}
   2320 
   2321 	xpl->sadb_x_policy_id = sadb_x_policy_id;
   2322 
   2323 	m_freem(m);
   2324 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   2325     }
   2326 }
   2327 
   2328 /*
   2329  * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
   2330  * add an entry to SP database, when received
   2331  *   <base, address(SD), (lifetime(H),) policy>
   2332  * from the user(?).
   2333  * Adding to SP database,
   2334  * and send
   2335  *   <base, address(SD), (lifetime(H),) policy>
   2336  * to the socket which was send.
   2337  *
   2338  * SPDADD set a unique policy entry.
   2339  * SPDSETIDX like SPDADD without a part of policy requests.
   2340  * SPDUPDATE replace a unique policy entry.
   2341  *
   2342  * m will always be freed.
   2343  */
   2344 static int
   2345 key_api_spdadd(struct socket *so, struct mbuf *m,
   2346 	       const struct sadb_msghdr *mhp)
   2347 {
   2348 
   2349 	return key_spdadd(so, m, mhp, NULL, false);
   2350 }
   2351 
   2352 struct secpolicy *
   2353 key_kpi_spdadd(struct mbuf *m)
   2354 {
   2355 	struct sadb_msghdr mh;
   2356 	int error;
   2357 	struct secpolicy *sp = NULL;
   2358 
   2359 	error = key_align(m, &mh);
   2360 	if (error)
   2361 		return NULL;
   2362 
   2363 	error = key_spdadd(NULL, m, &mh, &sp, true);
   2364 	if (error) {
   2365 		/*
   2366 		 * Currently, when key_spdadd() cannot send a PFKEY message
   2367 		 * which means SP has been created, key_spdadd() returns error
   2368 		 * although SP is created successfully.
   2369 		 * Kernel components would not care PFKEY messages, so return
   2370 		 * the "sp" regardless of error code. key_spdadd() overwrites
   2371 		 * the argument only if SP  is created successfully.
   2372 		 */
   2373 	}
   2374 	return sp;
   2375 }
   2376 
   2377 /*
   2378  * get new policy id.
   2379  * OUT:
   2380  *	0:	failure.
   2381  *	others: success.
   2382  */
   2383 static u_int32_t
   2384 key_getnewspid(void)
   2385 {
   2386 	u_int32_t newid = 0;
   2387 	int count = key_spi_trycnt;	/* XXX */
   2388 	struct secpolicy *sp;
   2389 
   2390 	/* when requesting to allocate spi ranged */
   2391 	while (count--) {
   2392 		newid = (policy_id = (policy_id == ~0 ? 1 : policy_id + 1));
   2393 
   2394 		sp = key_getspbyid(newid);
   2395 		if (sp == NULL)
   2396 			break;
   2397 
   2398 		KEY_SP_UNREF(&sp);
   2399 	}
   2400 
   2401 	if (count == 0 || newid == 0) {
   2402 		IPSECLOG(LOG_DEBUG, "to allocate policy id is failed.\n");
   2403 		return 0;
   2404 	}
   2405 
   2406 	return newid;
   2407 }
   2408 
   2409 /*
   2410  * SADB_SPDDELETE processing
   2411  * receive
   2412  *   <base, address(SD), policy(*)>
   2413  * from the user(?), and set SADB_SASTATE_DEAD,
   2414  * and send,
   2415  *   <base, address(SD), policy(*)>
   2416  * to the ikmpd.
   2417  * policy(*) including direction of policy.
   2418  *
   2419  * m will always be freed.
   2420  */
   2421 static int
   2422 key_api_spddelete(struct socket *so, struct mbuf *m,
   2423               const struct sadb_msghdr *mhp)
   2424 {
   2425 	struct sadb_x_policy *xpl0;
   2426 	struct secpolicyindex spidx;
   2427 	struct secpolicy *sp;
   2428 
   2429 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   2430 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
   2431 	    mhp->ext[SADB_X_EXT_POLICY] == NULL) {
   2432 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2433 		return key_senderror(so, m, EINVAL);
   2434 	}
   2435 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   2436 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
   2437 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
   2438 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2439 		return key_senderror(so, m, EINVAL);
   2440 	}
   2441 
   2442 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
   2443 
   2444 	/* checking the directon. */
   2445 	switch (xpl0->sadb_x_policy_dir) {
   2446 	case IPSEC_DIR_INBOUND:
   2447 	case IPSEC_DIR_OUTBOUND:
   2448 		break;
   2449 	default:
   2450 		IPSECLOG(LOG_DEBUG, "Invalid SP direction.\n");
   2451 		return key_senderror(so, m, EINVAL);
   2452 	}
   2453 
   2454 	/* make secindex */
   2455 	key_init_spidx_bymsghdr(&spidx, mhp);
   2456 
   2457 	/* Is there SP in SPD ? */
   2458 	sp = key_lookup_and_remove_sp(&spidx, false);
   2459 	if (sp == NULL) {
   2460 		IPSECLOG(LOG_DEBUG, "no SP found.\n");
   2461 		return key_senderror(so, m, EINVAL);
   2462 	}
   2463 
   2464 	/* save policy id to buffer to be returned. */
   2465 	xpl0->sadb_x_policy_id = sp->id;
   2466 
   2467 	key_destroy_sp(sp);
   2468 
   2469 	/* We're deleting policy; no need to invalidate the ipflow cache. */
   2470 
   2471     {
   2472 	struct mbuf *n;
   2473 
   2474 	/* create new sadb_msg to reply. */
   2475 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
   2476 	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
   2477 	key_fill_replymsg(n, 0);
   2478 	m_freem(m);
   2479 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   2480     }
   2481 }
   2482 
   2483 static struct mbuf *
   2484 key_alloc_mbuf_simple(int len, int mflag)
   2485 {
   2486 	struct mbuf *n;
   2487 
   2488 	KASSERT(mflag == M_NOWAIT || (mflag == M_WAITOK && !cpu_softintr_p()));
   2489 
   2490 	MGETHDR(n, mflag, MT_DATA);
   2491 	if (n && len > MHLEN) {
   2492 		MCLGET(n, mflag);
   2493 		if ((n->m_flags & M_EXT) == 0) {
   2494 			m_freem(n);
   2495 			n = NULL;
   2496 		}
   2497 	}
   2498 	return n;
   2499 }
   2500 
   2501 /*
   2502  * SADB_SPDDELETE2 processing
   2503  * receive
   2504  *   <base, policy(*)>
   2505  * from the user(?), and set SADB_SASTATE_DEAD,
   2506  * and send,
   2507  *   <base, policy(*)>
   2508  * to the ikmpd.
   2509  * policy(*) including direction of policy.
   2510  *
   2511  * m will always be freed.
   2512  */
   2513 static int
   2514 key_spddelete2(struct socket *so, struct mbuf *m,
   2515 	       const struct sadb_msghdr *mhp, bool from_kernel)
   2516 {
   2517 	u_int32_t id;
   2518 	struct secpolicy *sp;
   2519 	const struct sadb_x_policy *xpl;
   2520 
   2521 	if (mhp->ext[SADB_X_EXT_POLICY] == NULL ||
   2522 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
   2523 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2524 		return key_senderror(so, m, EINVAL);
   2525 	}
   2526 
   2527 	xpl = mhp->ext[SADB_X_EXT_POLICY];
   2528 	id = xpl->sadb_x_policy_id;
   2529 
   2530 	/* Is there SP in SPD ? */
   2531 	sp = key_lookupbyid_and_remove_sp(id, from_kernel);
   2532 	if (sp == NULL) {
   2533 		IPSECLOG(LOG_DEBUG, "no SP found id:%u.\n", id);
   2534 		return key_senderror(so, m, EINVAL);
   2535 	}
   2536 
   2537 	key_destroy_sp(sp);
   2538 
   2539 	/* We're deleting policy; no need to invalidate the ipflow cache. */
   2540 
   2541     {
   2542 	struct mbuf *n, *nn;
   2543 	int off, len;
   2544 
   2545 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MCLBYTES);
   2546 
   2547 	/* create new sadb_msg to reply. */
   2548 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
   2549 
   2550 	n = key_alloc_mbuf_simple(len, M_WAITOK);
   2551 	n->m_len = len;
   2552 	n->m_next = NULL;
   2553 	off = 0;
   2554 
   2555 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
   2556 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
   2557 
   2558 	KASSERTMSG(off == len, "length inconsistency");
   2559 
   2560 	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
   2561 	    mhp->extlen[SADB_X_EXT_POLICY], M_WAITOK);
   2562 
   2563 	n->m_pkthdr.len = 0;
   2564 	for (nn = n; nn; nn = nn->m_next)
   2565 		n->m_pkthdr.len += nn->m_len;
   2566 
   2567 	key_fill_replymsg(n, 0);
   2568 	m_freem(m);
   2569 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   2570     }
   2571 }
   2572 
   2573 /*
   2574  * SADB_SPDDELETE2 processing
   2575  * receive
   2576  *   <base, policy(*)>
   2577  * from the user(?), and set SADB_SASTATE_DEAD,
   2578  * and send,
   2579  *   <base, policy(*)>
   2580  * to the ikmpd.
   2581  * policy(*) including direction of policy.
   2582  *
   2583  * m will always be freed.
   2584  */
   2585 static int
   2586 key_api_spddelete2(struct socket *so, struct mbuf *m,
   2587 	       const struct sadb_msghdr *mhp)
   2588 {
   2589 
   2590 	return key_spddelete2(so, m, mhp, false);
   2591 }
   2592 
   2593 int
   2594 key_kpi_spddelete2(struct mbuf *m)
   2595 {
   2596 	struct sadb_msghdr mh;
   2597 	int error;
   2598 
   2599 	error = key_align(m, &mh);
   2600 	if (error)
   2601 		return EINVAL;
   2602 
   2603 	return key_spddelete2(NULL, m, &mh, true);
   2604 }
   2605 
   2606 /*
   2607  * SADB_X_GET processing
   2608  * receive
   2609  *   <base, policy(*)>
   2610  * from the user(?),
   2611  * and send,
   2612  *   <base, address(SD), policy>
   2613  * to the ikmpd.
   2614  * policy(*) including direction of policy.
   2615  *
   2616  * m will always be freed.
   2617  */
   2618 static int
   2619 key_api_spdget(struct socket *so, struct mbuf *m,
   2620 	   const struct sadb_msghdr *mhp)
   2621 {
   2622 	u_int32_t id;
   2623 	struct secpolicy *sp;
   2624 	struct mbuf *n;
   2625 	const struct sadb_x_policy *xpl;
   2626 
   2627 	if (mhp->ext[SADB_X_EXT_POLICY] == NULL ||
   2628 	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
   2629 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   2630 		return key_senderror(so, m, EINVAL);
   2631 	}
   2632 
   2633 	xpl = mhp->ext[SADB_X_EXT_POLICY];
   2634 	id = xpl->sadb_x_policy_id;
   2635 
   2636 	/* Is there SP in SPD ? */
   2637 	sp = key_getspbyid(id);
   2638 	if (sp == NULL) {
   2639 		IPSECLOG(LOG_DEBUG, "no SP found id:%u.\n", id);
   2640 		return key_senderror(so, m, ENOENT);
   2641 	}
   2642 
   2643 	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
   2644 	    mhp->msg->sadb_msg_pid);
   2645 	KEY_SP_UNREF(&sp); /* ref gained by key_getspbyid */
   2646 	m_freem(m);
   2647 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
   2648 }
   2649 
   2650 #ifdef notyet
   2651 /*
   2652  * SADB_X_SPDACQUIRE processing.
   2653  * Acquire policy and SA(s) for a *OUTBOUND* packet.
   2654  * send
   2655  *   <base, policy(*)>
   2656  * to KMD, and expect to receive
   2657  *   <base> with SADB_X_SPDACQUIRE if error occurred,
   2658  * or
   2659  *   <base, policy>
   2660  * with SADB_X_SPDUPDATE from KMD by PF_KEY.
   2661  * policy(*) is without policy requests.
   2662  *
   2663  *    0     : succeed
   2664  *    others: error number
   2665  */
   2666 int
   2667 key_spdacquire(const struct secpolicy *sp)
   2668 {
   2669 	struct mbuf *result = NULL, *m;
   2670 	struct secspacq *newspacq;
   2671 	int error;
   2672 
   2673 	KASSERT(sp != NULL);
   2674 	KASSERTMSG(sp->req == NULL, "called but there is request");
   2675 	KASSERTMSG(sp->policy == IPSEC_POLICY_IPSEC,
   2676 	    "policy mismathed. IPsec is expected");
   2677 
   2678 	/* Get an entry to check whether sent message or not. */
   2679 	newspacq = key_getspacq(&sp->spidx);
   2680 	if (newspacq != NULL) {
   2681 		if (key_blockacq_count < newspacq->count) {
   2682 			/* reset counter and do send message. */
   2683 			newspacq->count = 0;
   2684 		} else {
   2685 			/* increment counter and do nothing. */
   2686 			newspacq->count++;
   2687 			return 0;
   2688 		}
   2689 	} else {
   2690 		/* make new entry for blocking to send SADB_ACQUIRE. */
   2691 		newspacq = key_newspacq(&sp->spidx);
   2692 		if (newspacq == NULL)
   2693 			return ENOBUFS;
   2694 
   2695 		/* add to key_misc.acqlist */
   2696 		LIST_INSERT_HEAD(&key_misc.spacqlist, newspacq, chain);
   2697 	}
   2698 
   2699 	/* create new sadb_msg to reply. */
   2700 	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
   2701 	if (!m) {
   2702 		error = ENOBUFS;
   2703 		goto fail;
   2704 	}
   2705 	result = m;
   2706 
   2707 	result->m_pkthdr.len = 0;
   2708 	for (m = result; m; m = m->m_next)
   2709 		result->m_pkthdr.len += m->m_len;
   2710 
   2711 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   2712 	    PFKEY_UNIT64(result->m_pkthdr.len);
   2713 
   2714 	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
   2715 
   2716 fail:
   2717 	if (result)
   2718 		m_freem(result);
   2719 	return error;
   2720 }
   2721 #endif /* notyet */
   2722 
   2723 /*
   2724  * SADB_SPDFLUSH processing
   2725  * receive
   2726  *   <base>
   2727  * from the user, and free all entries in secpctree.
   2728  * and send,
   2729  *   <base>
   2730  * to the user.
   2731  * NOTE: what to do is only marking SADB_SASTATE_DEAD.
   2732  *
   2733  * m will always be freed.
   2734  */
   2735 static int
   2736 key_api_spdflush(struct socket *so, struct mbuf *m,
   2737 	     const struct sadb_msghdr *mhp)
   2738 {
   2739 	struct sadb_msg *newmsg;
   2740 	struct secpolicy *sp;
   2741 	u_int dir;
   2742 
   2743 	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
   2744 		return key_senderror(so, m, EINVAL);
   2745 
   2746 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
   2747 	    retry:
   2748 		mutex_enter(&key_spd.lock);
   2749 		SPLIST_WRITER_FOREACH(sp, dir) {
   2750 			KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
   2751 			/*
   2752 			 * Userlang programs can remove SPs created by userland
   2753 			 * probrams only, that is, they cannot remove SPs
   2754 			 * created in kernel(e.g. ipsec(4) I/F).
   2755 			 */
   2756 			if (sp->origin == IPSEC_SPORIGIN_USER) {
   2757 				key_unlink_sp(sp);
   2758 				mutex_exit(&key_spd.lock);
   2759 				key_destroy_sp(sp);
   2760 				goto retry;
   2761 			}
   2762 		}
   2763 		mutex_exit(&key_spd.lock);
   2764 	}
   2765 
   2766 	/* We're deleting policy; no need to invalidate the ipflow cache. */
   2767 
   2768 	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
   2769 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
   2770 		return key_senderror(so, m, ENOBUFS);
   2771 	}
   2772 
   2773 	if (m->m_next)
   2774 		m_freem(m->m_next);
   2775 	m->m_next = NULL;
   2776 	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
   2777 	newmsg = mtod(m, struct sadb_msg *);
   2778 	newmsg->sadb_msg_errno = 0;
   2779 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
   2780 
   2781 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
   2782 }
   2783 
   2784 static struct sockaddr key_src = {
   2785 	.sa_len = 2,
   2786 	.sa_family = PF_KEY,
   2787 };
   2788 
   2789 static struct mbuf *
   2790 key_setspddump_chain(int *errorp, int *lenp, pid_t pid)
   2791 {
   2792 	struct secpolicy *sp;
   2793 	int cnt;
   2794 	u_int dir;
   2795 	struct mbuf *m, *n, *prev;
   2796 	int totlen;
   2797 
   2798 	KASSERT(mutex_owned(&key_spd.lock));
   2799 
   2800 	*lenp = 0;
   2801 
   2802 	/* search SPD entry and get buffer size. */
   2803 	cnt = 0;
   2804 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
   2805 		SPLIST_WRITER_FOREACH(sp, dir) {
   2806 			cnt++;
   2807 		}
   2808 	}
   2809 
   2810 	if (cnt == 0) {
   2811 		*errorp = ENOENT;
   2812 		return (NULL);
   2813 	}
   2814 
   2815 	m = NULL;
   2816 	prev = m;
   2817 	totlen = 0;
   2818 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
   2819 		SPLIST_WRITER_FOREACH(sp, dir) {
   2820 			--cnt;
   2821 			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, pid);
   2822 
   2823 			totlen += n->m_pkthdr.len;
   2824 			if (!m) {
   2825 				m = n;
   2826 			} else {
   2827 				prev->m_nextpkt = n;
   2828 			}
   2829 			prev = n;
   2830 		}
   2831 	}
   2832 
   2833 	*lenp = totlen;
   2834 	*errorp = 0;
   2835 	return (m);
   2836 }
   2837 
   2838 /*
   2839  * SADB_SPDDUMP processing
   2840  * receive
   2841  *   <base>
   2842  * from the user, and dump all SP leaves
   2843  * and send,
   2844  *   <base> .....
   2845  * to the ikmpd.
   2846  *
   2847  * m will always be freed.
   2848  */
   2849 static int
   2850 key_api_spddump(struct socket *so, struct mbuf *m0,
   2851  	    const struct sadb_msghdr *mhp)
   2852 {
   2853 	struct mbuf *n;
   2854 	int error, len;
   2855 	int ok;
   2856 	pid_t pid;
   2857 
   2858 	pid = mhp->msg->sadb_msg_pid;
   2859 	/*
   2860 	 * If the requestor has insufficient socket-buffer space
   2861 	 * for the entire chain, nobody gets any response to the DUMP.
   2862 	 * XXX For now, only the requestor ever gets anything.
   2863 	 * Moreover, if the requestor has any space at all, they receive
   2864 	 * the entire chain, otherwise the request is refused with  ENOBUFS.
   2865 	 */
   2866 	if (sbspace(&so->so_rcv) <= 0) {
   2867 		return key_senderror(so, m0, ENOBUFS);
   2868 	}
   2869 
   2870 	mutex_enter(&key_spd.lock);
   2871 	n = key_setspddump_chain(&error, &len, pid);
   2872 	mutex_exit(&key_spd.lock);
   2873 
   2874 	if (n == NULL) {
   2875 		return key_senderror(so, m0, ENOENT);
   2876 	}
   2877 	{
   2878 		uint64_t *ps = PFKEY_STAT_GETREF();
   2879 		ps[PFKEY_STAT_IN_TOTAL]++;
   2880 		ps[PFKEY_STAT_IN_BYTES] += len;
   2881 		PFKEY_STAT_PUTREF();
   2882 	}
   2883 
   2884 	/*
   2885 	 * PF_KEY DUMP responses are no longer broadcast to all PF_KEY sockets.
   2886 	 * The requestor receives either the entire chain, or an
   2887 	 * error message with ENOBUFS.
   2888 	 */
   2889 
   2890 	/*
   2891 	 * sbappendchainwith record takes the chain of entries, one
   2892 	 * packet-record per SPD entry, prepends the key_src sockaddr
   2893 	 * to each packet-record, links the sockaddr mbufs into a new
   2894 	 * list of records, then   appends the entire resulting
   2895 	 * list to the requesting socket.
   2896 	 */
   2897 	ok = sbappendaddrchain(&so->so_rcv, (struct sockaddr *)&key_src, n,
   2898 	    SB_PRIO_ONESHOT_OVERFLOW);
   2899 
   2900 	if (!ok) {
   2901 		PFKEY_STATINC(PFKEY_STAT_IN_NOMEM);
   2902 		m_freem(n);
   2903 		return key_senderror(so, m0, ENOBUFS);
   2904 	}
   2905 
   2906 	m_freem(m0);
   2907 	return error;
   2908 }
   2909 
   2910 /*
   2911  * SADB_X_NAT_T_NEW_MAPPING. Unused by racoon as of 2005/04/23
   2912  */
   2913 static int
   2914 key_api_nat_map(struct socket *so, struct mbuf *m,
   2915 	    const struct sadb_msghdr *mhp)
   2916 {
   2917 	struct sadb_x_nat_t_type *type;
   2918 	struct sadb_x_nat_t_port *sport;
   2919 	struct sadb_x_nat_t_port *dport;
   2920 	struct sadb_address *iaddr, *raddr;
   2921 	struct sadb_x_nat_t_frag *frag;
   2922 
   2923 	if (mhp->ext[SADB_X_EXT_NAT_T_TYPE] == NULL ||
   2924 	    mhp->ext[SADB_X_EXT_NAT_T_SPORT] == NULL ||
   2925 	    mhp->ext[SADB_X_EXT_NAT_T_DPORT] == NULL) {
   2926 		IPSECLOG(LOG_DEBUG, "invalid message.\n");
   2927 		return key_senderror(so, m, EINVAL);
   2928 	}
   2929 	if ((mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) ||
   2930 	    (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) ||
   2931 	    (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport))) {
   2932 		IPSECLOG(LOG_DEBUG, "invalid message.\n");
   2933 		return key_senderror(so, m, EINVAL);
   2934 	}
   2935 
   2936 	if ((mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL) &&
   2937 	    (mhp->extlen[SADB_X_EXT_NAT_T_OAI] < sizeof(*iaddr))) {
   2938 		IPSECLOG(LOG_DEBUG, "invalid message\n");
   2939 		return key_senderror(so, m, EINVAL);
   2940 	}
   2941 
   2942 	if ((mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL) &&
   2943 	    (mhp->extlen[SADB_X_EXT_NAT_T_OAR] < sizeof(*raddr))) {
   2944 		IPSECLOG(LOG_DEBUG, "invalid message\n");
   2945 		return key_senderror(so, m, EINVAL);
   2946 	}
   2947 
   2948 	if ((mhp->ext[SADB_X_EXT_NAT_T_FRAG] != NULL) &&
   2949 	    (mhp->extlen[SADB_X_EXT_NAT_T_FRAG] < sizeof(*frag))) {
   2950 		IPSECLOG(LOG_DEBUG, "invalid message\n");
   2951 		return key_senderror(so, m, EINVAL);
   2952 	}
   2953 
   2954 	type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
   2955 	sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
   2956 	dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
   2957 	iaddr = mhp->ext[SADB_X_EXT_NAT_T_OAI];
   2958 	raddr = mhp->ext[SADB_X_EXT_NAT_T_OAR];
   2959 	frag = mhp->ext[SADB_X_EXT_NAT_T_FRAG];
   2960 
   2961 	/*
   2962 	 * XXX handle that, it should also contain a SA, or anything
   2963 	 * that enable to update the SA information.
   2964 	 */
   2965 
   2966 	return 0;
   2967 }
   2968 
   2969 /*
   2970  * Never return NULL.
   2971  */
   2972 static struct mbuf *
   2973 key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq, pid_t pid)
   2974 {
   2975 	struct mbuf *result = NULL, *m;
   2976 
   2977 	KASSERT(!cpu_softintr_p());
   2978 
   2979 	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid,
   2980 	    key_sp_refcnt(sp), M_WAITOK);
   2981 	result = m;
   2982 
   2983 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
   2984 	    &sp->spidx.src.sa, sp->spidx.prefs, sp->spidx.ul_proto, M_WAITOK);
   2985 	m_cat(result, m);
   2986 
   2987 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
   2988 	    &sp->spidx.dst.sa, sp->spidx.prefd, sp->spidx.ul_proto, M_WAITOK);
   2989 	m_cat(result, m);
   2990 
   2991 	m = key_sp2msg(sp, M_WAITOK);
   2992 	m_cat(result, m);
   2993 
   2994 	KASSERT(result->m_flags & M_PKTHDR);
   2995 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
   2996 
   2997 	result->m_pkthdr.len = 0;
   2998 	for (m = result; m; m = m->m_next)
   2999 		result->m_pkthdr.len += m->m_len;
   3000 
   3001 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   3002 	    PFKEY_UNIT64(result->m_pkthdr.len);
   3003 
   3004 	return result;
   3005 }
   3006 
   3007 /*
   3008  * get PFKEY message length for security policy and request.
   3009  */
   3010 static u_int
   3011 key_getspreqmsglen(const struct secpolicy *sp)
   3012 {
   3013 	u_int tlen;
   3014 
   3015 	tlen = sizeof(struct sadb_x_policy);
   3016 
   3017 	/* if is the policy for ipsec ? */
   3018 	if (sp->policy != IPSEC_POLICY_IPSEC)
   3019 		return tlen;
   3020 
   3021 	/* get length of ipsec requests */
   3022     {
   3023 	const struct ipsecrequest *isr;
   3024 	int len;
   3025 
   3026 	for (isr = sp->req; isr != NULL; isr = isr->next) {
   3027 		len = sizeof(struct sadb_x_ipsecrequest)
   3028 		    + isr->saidx.src.sa.sa_len + isr->saidx.dst.sa.sa_len;
   3029 
   3030 		tlen += PFKEY_ALIGN8(len);
   3031 	}
   3032     }
   3033 
   3034 	return tlen;
   3035 }
   3036 
   3037 /*
   3038  * SADB_SPDEXPIRE processing
   3039  * send
   3040  *   <base, address(SD), lifetime(CH), policy>
   3041  * to KMD by PF_KEY.
   3042  *
   3043  * OUT:	0	: succeed
   3044  *	others	: error number
   3045  */
   3046 static int
   3047 key_spdexpire(struct secpolicy *sp)
   3048 {
   3049 	int s;
   3050 	struct mbuf *result = NULL, *m;
   3051 	int len;
   3052 	int error = -1;
   3053 	struct sadb_lifetime *lt;
   3054 
   3055 	/* XXX: Why do we lock ? */
   3056 	s = splsoftnet();	/*called from softclock()*/
   3057 
   3058 	KASSERT(sp != NULL);
   3059 
   3060 	/* set msg header */
   3061 	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0, M_WAITOK);
   3062 	result = m;
   3063 
   3064 	/* create lifetime extension (current and hard) */
   3065 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
   3066 	m = key_alloc_mbuf(len, M_WAITOK);
   3067 	KASSERT(m->m_next == NULL);
   3068 
   3069 	memset(mtod(m, void *), 0, len);
   3070 	lt = mtod(m, struct sadb_lifetime *);
   3071 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
   3072 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
   3073 	lt->sadb_lifetime_allocations = 0;
   3074 	lt->sadb_lifetime_bytes = 0;
   3075 	lt->sadb_lifetime_addtime = time_mono_to_wall(sp->created);
   3076 	lt->sadb_lifetime_usetime = time_mono_to_wall(sp->lastused);
   3077 	lt = (struct sadb_lifetime *)(mtod(m, char *) + len / 2);
   3078 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
   3079 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
   3080 	lt->sadb_lifetime_allocations = 0;
   3081 	lt->sadb_lifetime_bytes = 0;
   3082 	lt->sadb_lifetime_addtime = sp->lifetime;
   3083 	lt->sadb_lifetime_usetime = sp->validtime;
   3084 	m_cat(result, m);
   3085 
   3086 	/* set sadb_address for source */
   3087 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sp->spidx.src.sa,
   3088 	    sp->spidx.prefs, sp->spidx.ul_proto, M_WAITOK);
   3089 	m_cat(result, m);
   3090 
   3091 	/* set sadb_address for destination */
   3092 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sp->spidx.dst.sa,
   3093 	    sp->spidx.prefd, sp->spidx.ul_proto, M_WAITOK);
   3094 	m_cat(result, m);
   3095 
   3096 	/* set secpolicy */
   3097 	m = key_sp2msg(sp, M_WAITOK);
   3098 	m_cat(result, m);
   3099 
   3100 	KASSERT(result->m_flags & M_PKTHDR);
   3101 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
   3102 
   3103 	result->m_pkthdr.len = 0;
   3104 	for (m = result; m; m = m->m_next)
   3105 		result->m_pkthdr.len += m->m_len;
   3106 
   3107 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   3108 	    PFKEY_UNIT64(result->m_pkthdr.len);
   3109 
   3110 	error = key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
   3111 	splx(s);
   3112 	return error;
   3113 }
   3114 
   3115 /* %%% SAD management */
   3116 /*
   3117  * allocating a memory for new SA head, and copy from the values of mhp.
   3118  * OUT:	NULL	: failure due to the lack of memory.
   3119  *	others	: pointer to new SA head.
   3120  */
   3121 static struct secashead *
   3122 key_newsah(const struct secasindex *saidx)
   3123 {
   3124 	struct secashead *newsah;
   3125 	int i;
   3126 
   3127 	KASSERT(saidx != NULL);
   3128 
   3129 	newsah = kmem_zalloc(sizeof(struct secashead), KM_SLEEP);
   3130 	for (i = 0; i < __arraycount(newsah->savlist); i++)
   3131 		PSLIST_INIT(&newsah->savlist[i]);
   3132 	newsah->saidx = *saidx;
   3133 
   3134 	localcount_init(&newsah->localcount);
   3135 	/* Take a reference for the caller */
   3136 	localcount_acquire(&newsah->localcount);
   3137 
   3138 	/* Add to the sah list */
   3139 	SAHLIST_ENTRY_INIT(newsah);
   3140 	newsah->state = SADB_SASTATE_MATURE;
   3141 	mutex_enter(&key_sad.lock);
   3142 	SAHLIST_WRITER_INSERT_HEAD(newsah);
   3143 	mutex_exit(&key_sad.lock);
   3144 
   3145 	return newsah;
   3146 }
   3147 
   3148 static bool
   3149 key_sah_has_sav(struct secashead *sah)
   3150 {
   3151 	u_int state;
   3152 
   3153 	KASSERT(mutex_owned(&key_sad.lock));
   3154 
   3155 	SASTATE_ANY_FOREACH(state) {
   3156 		if (!SAVLIST_WRITER_EMPTY(sah, state))
   3157 			return true;
   3158 	}
   3159 
   3160 	return false;
   3161 }
   3162 
   3163 static void
   3164 key_unlink_sah(struct secashead *sah)
   3165 {
   3166 
   3167 	KASSERT(!cpu_softintr_p());
   3168 	KASSERT(mutex_owned(&key_sad.lock));
   3169 	KASSERT(sah->state == SADB_SASTATE_DEAD);
   3170 
   3171 	/* Remove from the sah list */
   3172 	SAHLIST_WRITER_REMOVE(sah);
   3173 
   3174 	KDASSERT(mutex_ownable(softnet_lock));
   3175 	key_sad_pserialize_perform();
   3176 
   3177 	localcount_drain(&sah->localcount, &key_sad.cv_lc, &key_sad.lock);
   3178 }
   3179 
   3180 static void
   3181 key_destroy_sah(struct secashead *sah)
   3182 {
   3183 
   3184 	rtcache_free(&sah->sa_route);
   3185 
   3186 	SAHLIST_ENTRY_DESTROY(sah);
   3187 	localcount_fini(&sah->localcount);
   3188 
   3189 	if (sah->idents != NULL)
   3190 		kmem_free(sah->idents, sah->idents_len);
   3191 	if (sah->identd != NULL)
   3192 		kmem_free(sah->identd, sah->identd_len);
   3193 
   3194 	kmem_free(sah, sizeof(*sah));
   3195 }
   3196 
   3197 /*
   3198  * allocating a new SA with LARVAL state.
   3199  * key_api_add() and key_api_getspi() call,
   3200  * and copy the values of mhp into new buffer.
   3201  * When SAD message type is GETSPI:
   3202  *	to set sequence number from acq_seq++,
   3203  *	to set zero to SPI.
   3204  *	not to call key_setsava().
   3205  * OUT:	NULL	: fail
   3206  *	others	: pointer to new secasvar.
   3207  *
   3208  * does not modify mbuf.  does not free mbuf on error.
   3209  */
   3210 static struct secasvar *
   3211 key_newsav(struct mbuf *m, const struct sadb_msghdr *mhp,
   3212     int *errp, const char* where, int tag)
   3213 {
   3214 	struct secasvar *newsav;
   3215 	const struct sadb_sa *xsa;
   3216 
   3217 	KASSERT(!cpu_softintr_p());
   3218 	KASSERT(m != NULL);
   3219 	KASSERT(mhp != NULL);
   3220 	KASSERT(mhp->msg != NULL);
   3221 
   3222 	newsav = kmem_zalloc(sizeof(struct secasvar), KM_SLEEP);
   3223 
   3224 	switch (mhp->msg->sadb_msg_type) {
   3225 	case SADB_GETSPI:
   3226 		newsav->spi = 0;
   3227 
   3228 #ifdef IPSEC_DOSEQCHECK
   3229 		/* sync sequence number */
   3230 		if (mhp->msg->sadb_msg_seq == 0)
   3231 			newsav->seq =
   3232 			    (acq_seq = (acq_seq == ~0 ? 1 : ++acq_seq));
   3233 		else
   3234 #endif
   3235 			newsav->seq = mhp->msg->sadb_msg_seq;
   3236 		break;
   3237 
   3238 	case SADB_ADD:
   3239 		/* sanity check */
   3240 		if (mhp->ext[SADB_EXT_SA] == NULL) {
   3241 			IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   3242 			*errp = EINVAL;
   3243 			goto error;
   3244 		}
   3245 		xsa = mhp->ext[SADB_EXT_SA];
   3246 		newsav->spi = xsa->sadb_sa_spi;
   3247 		newsav->seq = mhp->msg->sadb_msg_seq;
   3248 		break;
   3249 	default:
   3250 		*errp = EINVAL;
   3251 		goto error;
   3252 	}
   3253 
   3254 	/* copy sav values */
   3255 	if (mhp->msg->sadb_msg_type != SADB_GETSPI) {
   3256 		*errp = key_setsaval(newsav, m, mhp);
   3257 		if (*errp)
   3258 			goto error;
   3259 	} else {
   3260 		/* We don't allow lft_c to be NULL */
   3261 		newsav->lft_c = kmem_zalloc(sizeof(struct sadb_lifetime),
   3262 		    KM_SLEEP);
   3263 	}
   3264 
   3265 	/* reset created */
   3266 	newsav->created = time_uptime;
   3267 	newsav->pid = mhp->msg->sadb_msg_pid;
   3268 
   3269 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   3270 	    "DP from %s:%u return SA:%p\n", where, tag, newsav);
   3271 	return newsav;
   3272 
   3273 error:
   3274 	KASSERT(*errp != 0);
   3275 	kmem_free(newsav, sizeof(*newsav));
   3276 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   3277 	    "DP from %s:%u return SA:NULL\n", where, tag);
   3278 	return NULL;
   3279 }
   3280 
   3281 
   3282 static void
   3283 key_clear_xform(struct secasvar *sav)
   3284 {
   3285 
   3286 	/*
   3287 	 * Cleanup xform state.  Note that zeroize'ing causes the
   3288 	 * keys to be cleared; otherwise we must do it ourself.
   3289 	 */
   3290 	if (sav->tdb_xform != NULL) {
   3291 		sav->tdb_xform->xf_zeroize(sav);
   3292 		sav->tdb_xform = NULL;
   3293 	} else {
   3294 		if (sav->key_auth != NULL)
   3295 			explicit_memset(_KEYBUF(sav->key_auth), 0,
   3296 			    _KEYLEN(sav->key_auth));
   3297 		if (sav->key_enc != NULL)
   3298 			explicit_memset(_KEYBUF(sav->key_enc), 0,
   3299 			    _KEYLEN(sav->key_enc));
   3300 	}
   3301 }
   3302 
   3303 /*
   3304  * free() SA variable entry.
   3305  */
   3306 static void
   3307 key_delsav(struct secasvar *sav)
   3308 {
   3309 
   3310 	key_clear_xform(sav);
   3311 	key_freesaval(sav);
   3312 	kmem_free(sav, sizeof(*sav));
   3313 }
   3314 
   3315 /*
   3316  * Must be called in a pserialize read section. A held sah
   3317  * must be released by key_sah_unref after use.
   3318  */
   3319 static void
   3320 key_sah_ref(struct secashead *sah)
   3321 {
   3322 
   3323 	localcount_acquire(&sah->localcount);
   3324 }
   3325 
   3326 /*
   3327  * Must be called without holding key_sad.lock because the lock
   3328  * would be held in localcount_release.
   3329  */
   3330 static void
   3331 key_sah_unref(struct secashead *sah)
   3332 {
   3333 
   3334 	KDASSERT(mutex_ownable(&key_sad.lock));
   3335 
   3336 	localcount_release(&sah->localcount, &key_sad.cv_lc, &key_sad.lock);
   3337 }
   3338 
   3339 /*
   3340  * Search SAD and return sah. Must be called in a pserialize
   3341  * read section.
   3342  * OUT:
   3343  *	NULL	: not found
   3344  *	others	: found, pointer to a SA.
   3345  */
   3346 static struct secashead *
   3347 key_getsah(const struct secasindex *saidx, int flag)
   3348 {
   3349 	struct secashead *sah;
   3350 
   3351 	SAHLIST_READER_FOREACH(sah) {
   3352 		if (sah->state == SADB_SASTATE_DEAD)
   3353 			continue;
   3354 		if (key_saidx_match(&sah->saidx, saidx, flag))
   3355 			return sah;
   3356 	}
   3357 
   3358 	return NULL;
   3359 }
   3360 
   3361 /*
   3362  * Search SAD and return sah. If sah is returned, the caller must call
   3363  * key_sah_unref to releaset a reference.
   3364  * OUT:
   3365  *	NULL	: not found
   3366  *	others	: found, pointer to a SA.
   3367  */
   3368 static struct secashead *
   3369 key_getsah_ref(const struct secasindex *saidx, int flag)
   3370 {
   3371 	struct secashead *sah;
   3372 	int s;
   3373 
   3374 	s = pserialize_read_enter();
   3375 	sah = key_getsah(saidx, flag);
   3376 	if (sah != NULL)
   3377 		key_sah_ref(sah);
   3378 	pserialize_read_exit(s);
   3379 
   3380 	return sah;
   3381 }
   3382 
   3383 /*
   3384  * check not to be duplicated SPI.
   3385  * NOTE: this function is too slow due to searching all SAD.
   3386  * OUT:
   3387  *	NULL	: not found
   3388  *	others	: found, pointer to a SA.
   3389  */
   3390 static bool
   3391 key_checkspidup(const struct secasindex *saidx, u_int32_t spi)
   3392 {
   3393 	struct secashead *sah;
   3394 	struct secasvar *sav;
   3395 	int s;
   3396 
   3397 	/* check address family */
   3398 	if (saidx->src.sa.sa_family != saidx->dst.sa.sa_family) {
   3399 		IPSECLOG(LOG_DEBUG, "address family mismatched.\n");
   3400 		return false;
   3401 	}
   3402 
   3403 	/* check all SAD */
   3404 	s = pserialize_read_enter();
   3405 	SAHLIST_READER_FOREACH(sah) {
   3406 		if (!key_ismyaddr((struct sockaddr *)&sah->saidx.dst))
   3407 			continue;
   3408 		sav = key_getsavbyspi(sah, spi);
   3409 		if (sav != NULL) {
   3410 			pserialize_read_exit(s);
   3411 			KEY_SA_UNREF(&sav);
   3412 			return true;
   3413 		}
   3414 	}
   3415 	pserialize_read_exit(s);
   3416 
   3417 	return false;
   3418 }
   3419 
   3420 /*
   3421  * search SAD litmited alive SA, protocol, SPI.
   3422  * OUT:
   3423  *	NULL	: not found
   3424  *	others	: found, pointer to a SA.
   3425  */
   3426 static struct secasvar *
   3427 key_getsavbyspi(struct secashead *sah, u_int32_t spi)
   3428 {
   3429 	struct secasvar *sav = NULL;
   3430 	u_int state;
   3431 	int s;
   3432 
   3433 	/* search all status */
   3434 	s = pserialize_read_enter();
   3435 	SASTATE_ALIVE_FOREACH(state) {
   3436 		SAVLIST_READER_FOREACH(sav, sah, state) {
   3437 			/* sanity check */
   3438 			if (sav->state != state) {
   3439 				IPSECLOG(LOG_DEBUG,
   3440 				    "invalid sav->state (queue: %d SA: %d)\n",
   3441 				    state, sav->state);
   3442 				continue;
   3443 			}
   3444 
   3445 			if (sav->spi == spi) {
   3446 				KEY_SA_REF(sav);
   3447 				goto out;
   3448 			}
   3449 		}
   3450 	}
   3451 out:
   3452 	pserialize_read_exit(s);
   3453 
   3454 	return sav;
   3455 }
   3456 
   3457 /*
   3458  * Free allocated data to member variables of sav:
   3459  * sav->replay, sav->key_* and sav->lft_*.
   3460  */
   3461 static void
   3462 key_freesaval(struct secasvar *sav)
   3463 {
   3464 
   3465 	KASSERT(key_sa_refcnt(sav) == 0);
   3466 
   3467 	if (sav->replay != NULL)
   3468 		kmem_intr_free(sav->replay, sav->replay_len);
   3469 	if (sav->key_auth != NULL)
   3470 		kmem_intr_free(sav->key_auth, sav->key_auth_len);
   3471 	if (sav->key_enc != NULL)
   3472 		kmem_intr_free(sav->key_enc, sav->key_enc_len);
   3473 	if (sav->lft_c != NULL)
   3474 		kmem_intr_free(sav->lft_c, sizeof(*(sav->lft_c)));
   3475 	if (sav->lft_h != NULL)
   3476 		kmem_intr_free(sav->lft_h, sizeof(*(sav->lft_h)));
   3477 	if (sav->lft_s != NULL)
   3478 		kmem_intr_free(sav->lft_s, sizeof(*(sav->lft_s)));
   3479 }
   3480 
   3481 /*
   3482  * copy SA values from PF_KEY message except *SPI, SEQ, PID, STATE and TYPE*.
   3483  * You must update these if need.
   3484  * OUT:	0:	success.
   3485  *	!0:	failure.
   3486  *
   3487  * does not modify mbuf.  does not free mbuf on error.
   3488  */
   3489 static int
   3490 key_setsaval(struct secasvar *sav, struct mbuf *m,
   3491 	     const struct sadb_msghdr *mhp)
   3492 {
   3493 	int error = 0;
   3494 
   3495 	KASSERT(!cpu_softintr_p());
   3496 	KASSERT(m != NULL);
   3497 	KASSERT(mhp != NULL);
   3498 	KASSERT(mhp->msg != NULL);
   3499 
   3500 	/* We shouldn't initialize sav variables while someone uses it. */
   3501 	KASSERT(key_sa_refcnt(sav) == 0);
   3502 
   3503 	/* SA */
   3504 	if (mhp->ext[SADB_EXT_SA] != NULL) {
   3505 		const struct sadb_sa *sa0;
   3506 
   3507 		sa0 = mhp->ext[SADB_EXT_SA];
   3508 		if (mhp->extlen[SADB_EXT_SA] < sizeof(*sa0)) {
   3509 			error = EINVAL;
   3510 			goto fail;
   3511 		}
   3512 
   3513 		sav->alg_auth = sa0->sadb_sa_auth;
   3514 		sav->alg_enc = sa0->sadb_sa_encrypt;
   3515 		sav->flags = sa0->sadb_sa_flags;
   3516 
   3517 		/* replay window */
   3518 		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0) {
   3519 			size_t len = sizeof(struct secreplay) +
   3520 			    sa0->sadb_sa_replay;
   3521 			sav->replay = kmem_zalloc(len, KM_SLEEP);
   3522 			sav->replay_len = len;
   3523 			if (sa0->sadb_sa_replay != 0)
   3524 				sav->replay->bitmap = (char*)(sav->replay+1);
   3525 			sav->replay->wsize = sa0->sadb_sa_replay;
   3526 		}
   3527 	}
   3528 
   3529 	/* Authentication keys */
   3530 	if (mhp->ext[SADB_EXT_KEY_AUTH] != NULL) {
   3531 		const struct sadb_key *key0;
   3532 		int len;
   3533 
   3534 		key0 = mhp->ext[SADB_EXT_KEY_AUTH];
   3535 		len = mhp->extlen[SADB_EXT_KEY_AUTH];
   3536 
   3537 		error = 0;
   3538 		if (len < sizeof(*key0)) {
   3539 			error = EINVAL;
   3540 			goto fail;
   3541 		}
   3542 		switch (mhp->msg->sadb_msg_satype) {
   3543 		case SADB_SATYPE_AH:
   3544 		case SADB_SATYPE_ESP:
   3545 		case SADB_X_SATYPE_TCPSIGNATURE:
   3546 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
   3547 			    sav->alg_auth != SADB_X_AALG_NULL)
   3548 				error = EINVAL;
   3549 			break;
   3550 		case SADB_X_SATYPE_IPCOMP:
   3551 		default:
   3552 			error = EINVAL;
   3553 			break;
   3554 		}
   3555 		if (error) {
   3556 			IPSECLOG(LOG_DEBUG, "invalid key_auth values.\n");
   3557 			goto fail;
   3558 		}
   3559 
   3560 		sav->key_auth = key_newbuf(key0, len);
   3561 		sav->key_auth_len = len;
   3562 	}
   3563 
   3564 	/* Encryption key */
   3565 	if (mhp->ext[SADB_EXT_KEY_ENCRYPT] != NULL) {
   3566 		const struct sadb_key *key0;
   3567 		int len;
   3568 
   3569 		key0 = mhp->ext[SADB_EXT_KEY_ENCRYPT];
   3570 		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
   3571 
   3572 		error = 0;
   3573 		if (len < sizeof(*key0)) {
   3574 			error = EINVAL;
   3575 			goto fail;
   3576 		}
   3577 		switch (mhp->msg->sadb_msg_satype) {
   3578 		case SADB_SATYPE_ESP:
   3579 			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
   3580 			    sav->alg_enc != SADB_EALG_NULL) {
   3581 				error = EINVAL;
   3582 				break;
   3583 			}
   3584 			sav->key_enc = key_newbuf(key0, len);
   3585 			sav->key_enc_len = len;
   3586 			break;
   3587 		case SADB_X_SATYPE_IPCOMP:
   3588 			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
   3589 				error = EINVAL;
   3590 			sav->key_enc = NULL;	/*just in case*/
   3591 			break;
   3592 		case SADB_SATYPE_AH:
   3593 		case SADB_X_SATYPE_TCPSIGNATURE:
   3594 		default:
   3595 			error = EINVAL;
   3596 			break;
   3597 		}
   3598 		if (error) {
   3599 			IPSECLOG(LOG_DEBUG, "invalid key_enc value.\n");
   3600 			goto fail;
   3601 		}
   3602 	}
   3603 
   3604 	/* set iv */
   3605 	sav->ivlen = 0;
   3606 
   3607 	switch (mhp->msg->sadb_msg_satype) {
   3608 	case SADB_SATYPE_AH:
   3609 		error = xform_init(sav, XF_AH);
   3610 		break;
   3611 	case SADB_SATYPE_ESP:
   3612 		error = xform_init(sav, XF_ESP);
   3613 		break;
   3614 	case SADB_X_SATYPE_IPCOMP:
   3615 		error = xform_init(sav, XF_IPCOMP);
   3616 		break;
   3617 	case SADB_X_SATYPE_TCPSIGNATURE:
   3618 		error = xform_init(sav, XF_TCPSIGNATURE);
   3619 		break;
   3620 	}
   3621 	if (error) {
   3622 		IPSECLOG(LOG_DEBUG, "unable to initialize SA type %u.\n",
   3623 		    mhp->msg->sadb_msg_satype);
   3624 		goto fail;
   3625 	}
   3626 
   3627 	/* reset created */
   3628 	sav->created = time_uptime;
   3629 
   3630 	/* make lifetime for CURRENT */
   3631 	sav->lft_c = kmem_alloc(sizeof(struct sadb_lifetime), KM_SLEEP);
   3632 
   3633 	sav->lft_c->sadb_lifetime_len =
   3634 	    PFKEY_UNIT64(sizeof(struct sadb_lifetime));
   3635 	sav->lft_c->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
   3636 	sav->lft_c->sadb_lifetime_allocations = 0;
   3637 	sav->lft_c->sadb_lifetime_bytes = 0;
   3638 	sav->lft_c->sadb_lifetime_addtime = time_uptime;
   3639 	sav->lft_c->sadb_lifetime_usetime = 0;
   3640 
   3641 	/* lifetimes for HARD and SOFT */
   3642     {
   3643 	const struct sadb_lifetime *lft0;
   3644 
   3645 	lft0 = mhp->ext[SADB_EXT_LIFETIME_HARD];
   3646 	if (lft0 != NULL) {
   3647 		if (mhp->extlen[SADB_EXT_LIFETIME_HARD] < sizeof(*lft0)) {
   3648 			error = EINVAL;
   3649 			goto fail;
   3650 		}
   3651 		sav->lft_h = key_newbuf(lft0, sizeof(*lft0));
   3652 	}
   3653 
   3654 	lft0 = mhp->ext[SADB_EXT_LIFETIME_SOFT];
   3655 	if (lft0 != NULL) {
   3656 		if (mhp->extlen[SADB_EXT_LIFETIME_SOFT] < sizeof(*lft0)) {
   3657 			error = EINVAL;
   3658 			goto fail;
   3659 		}
   3660 		sav->lft_s = key_newbuf(lft0, sizeof(*lft0));
   3661 		/* to be initialize ? */
   3662 	}
   3663     }
   3664 
   3665 	return 0;
   3666 
   3667  fail:
   3668 	key_clear_xform(sav);
   3669 	key_freesaval(sav);
   3670 
   3671 	return error;
   3672 }
   3673 
   3674 /*
   3675  * validation with a secasvar entry, and set SADB_SATYPE_MATURE.
   3676  * OUT:	0:	valid
   3677  *	other:	errno
   3678  */
   3679 static int
   3680 key_init_xform(struct secasvar *sav)
   3681 {
   3682 	int error;
   3683 
   3684 	/* We shouldn't initialize sav variables while someone uses it. */
   3685 	KASSERT(key_sa_refcnt(sav) == 0);
   3686 
   3687 	/* check SPI value */
   3688 	switch (sav->sah->saidx.proto) {
   3689 	case IPPROTO_ESP:
   3690 	case IPPROTO_AH:
   3691 		if (ntohl(sav->spi) <= 255) {
   3692 			IPSECLOG(LOG_DEBUG, "illegal range of SPI %u.\n",
   3693 			    (u_int32_t)ntohl(sav->spi));
   3694 			return EINVAL;
   3695 		}
   3696 		break;
   3697 	}
   3698 
   3699 	/* check satype */
   3700 	switch (sav->sah->saidx.proto) {
   3701 	case IPPROTO_ESP:
   3702 		/* check flags */
   3703 		if ((sav->flags & (SADB_X_EXT_OLD|SADB_X_EXT_DERIV)) ==
   3704 		    (SADB_X_EXT_OLD|SADB_X_EXT_DERIV)) {
   3705 			IPSECLOG(LOG_DEBUG,
   3706 			    "invalid flag (derived) given to old-esp.\n");
   3707 			return EINVAL;
   3708 		}
   3709 		error = xform_init(sav, XF_ESP);
   3710 		break;
   3711 	case IPPROTO_AH:
   3712 		/* check flags */
   3713 		if (sav->flags & SADB_X_EXT_DERIV) {
   3714 			IPSECLOG(LOG_DEBUG,
   3715 			    "invalid flag (derived) given to AH SA.\n");
   3716 			return EINVAL;
   3717 		}
   3718 		if (sav->alg_enc != SADB_EALG_NONE) {
   3719 			IPSECLOG(LOG_DEBUG,
   3720 			    "protocol and algorithm mismated.\n");
   3721 			return(EINVAL);
   3722 		}
   3723 		error = xform_init(sav, XF_AH);
   3724 		break;
   3725 	case IPPROTO_IPCOMP:
   3726 		if (sav->alg_auth != SADB_AALG_NONE) {
   3727 			IPSECLOG(LOG_DEBUG,
   3728 			    "protocol and algorithm mismated.\n");
   3729 			return(EINVAL);
   3730 		}
   3731 		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0
   3732 		 && ntohl(sav->spi) >= 0x10000) {
   3733 			IPSECLOG(LOG_DEBUG, "invalid cpi for IPComp.\n");
   3734 			return(EINVAL);
   3735 		}
   3736 		error = xform_init(sav, XF_IPCOMP);
   3737 		break;
   3738 	case IPPROTO_TCP:
   3739 		if (sav->alg_enc != SADB_EALG_NONE) {
   3740 			IPSECLOG(LOG_DEBUG,
   3741 			    "protocol and algorithm mismated.\n");
   3742 			return(EINVAL);
   3743 		}
   3744 		error = xform_init(sav, XF_TCPSIGNATURE);
   3745 		break;
   3746 	default:
   3747 		IPSECLOG(LOG_DEBUG, "Invalid satype.\n");
   3748 		error = EPROTONOSUPPORT;
   3749 		break;
   3750 	}
   3751 
   3752 	return error;
   3753 }
   3754 
   3755 /*
   3756  * subroutine for SADB_GET and SADB_DUMP. It never return NULL.
   3757  */
   3758 static struct mbuf *
   3759 key_setdumpsa(struct secasvar *sav, u_int8_t type, u_int8_t satype,
   3760 	      u_int32_t seq, u_int32_t pid)
   3761 {
   3762 	struct mbuf *result = NULL, *tres = NULL, *m;
   3763 	int l = 0;
   3764 	int i;
   3765 	void *p;
   3766 	struct sadb_lifetime lt;
   3767 	int dumporder[] = {
   3768 		SADB_EXT_SA, SADB_X_EXT_SA2,
   3769 		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
   3770 		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
   3771 		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY, SADB_EXT_KEY_AUTH,
   3772 		SADB_EXT_KEY_ENCRYPT, SADB_EXT_IDENTITY_SRC,
   3773 		SADB_EXT_IDENTITY_DST, SADB_EXT_SENSITIVITY,
   3774 		SADB_X_EXT_NAT_T_TYPE,
   3775 		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
   3776 		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
   3777 		SADB_X_EXT_NAT_T_FRAG,
   3778 
   3779 	};
   3780 
   3781 	m = key_setsadbmsg(type, 0, satype, seq, pid, key_sa_refcnt(sav), M_WAITOK);
   3782 	result = m;
   3783 
   3784 	for (i = __arraycount(dumporder) - 1; i >= 0; i--) {
   3785 		m = NULL;
   3786 		p = NULL;
   3787 		switch (dumporder[i]) {
   3788 		case SADB_EXT_SA:
   3789 			m = key_setsadbsa(sav);
   3790 			break;
   3791 
   3792 		case SADB_X_EXT_SA2:
   3793 			m = key_setsadbxsa2(sav->sah->saidx.mode,
   3794 			    sav->replay ? sav->replay->count : 0,
   3795 			    sav->sah->saidx.reqid);
   3796 			break;
   3797 
   3798 		case SADB_EXT_ADDRESS_SRC:
   3799 			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
   3800 			    &sav->sah->saidx.src.sa,
   3801 			    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
   3802 			break;
   3803 
   3804 		case SADB_EXT_ADDRESS_DST:
   3805 			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
   3806 			    &sav->sah->saidx.dst.sa,
   3807 			    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
   3808 			break;
   3809 
   3810 		case SADB_EXT_KEY_AUTH:
   3811 			if (!sav->key_auth)
   3812 				continue;
   3813 			l = PFKEY_UNUNIT64(sav->key_auth->sadb_key_len);
   3814 			p = sav->key_auth;
   3815 			break;
   3816 
   3817 		case SADB_EXT_KEY_ENCRYPT:
   3818 			if (!sav->key_enc)
   3819 				continue;
   3820 			l = PFKEY_UNUNIT64(sav->key_enc->sadb_key_len);
   3821 			p = sav->key_enc;
   3822 			break;
   3823 
   3824 		case SADB_EXT_LIFETIME_CURRENT:
   3825 			KASSERT(sav->lft_c != NULL);
   3826 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_c)->sadb_ext_len);
   3827 			memcpy(&lt, sav->lft_c, sizeof(struct sadb_lifetime));
   3828 			lt.sadb_lifetime_addtime =
   3829 			    time_mono_to_wall(lt.sadb_lifetime_addtime);
   3830 			lt.sadb_lifetime_usetime =
   3831 			    time_mono_to_wall(lt.sadb_lifetime_usetime);
   3832 			p = &lt;
   3833 			break;
   3834 
   3835 		case SADB_EXT_LIFETIME_HARD:
   3836 			if (!sav->lft_h)
   3837 				continue;
   3838 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_h)->sadb_ext_len);
   3839 			p = sav->lft_h;
   3840 			break;
   3841 
   3842 		case SADB_EXT_LIFETIME_SOFT:
   3843 			if (!sav->lft_s)
   3844 				continue;
   3845 			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_s)->sadb_ext_len);
   3846 			p = sav->lft_s;
   3847 			break;
   3848 
   3849 		case SADB_X_EXT_NAT_T_TYPE:
   3850 			m = key_setsadbxtype(sav->natt_type);
   3851 			break;
   3852 
   3853 		case SADB_X_EXT_NAT_T_DPORT:
   3854 			if (sav->natt_type == 0)
   3855 				continue;
   3856 			m = key_setsadbxport(
   3857 			    key_portfromsaddr(&sav->sah->saidx.dst),
   3858 			    SADB_X_EXT_NAT_T_DPORT);
   3859 			break;
   3860 
   3861 		case SADB_X_EXT_NAT_T_SPORT:
   3862 			if (sav->natt_type == 0)
   3863 				continue;
   3864 			m = key_setsadbxport(
   3865 			    key_portfromsaddr(&sav->sah->saidx.src),
   3866 			    SADB_X_EXT_NAT_T_SPORT);
   3867 			break;
   3868 
   3869 		case SADB_X_EXT_NAT_T_FRAG:
   3870 			/* don't send frag info if not set */
   3871 			if (sav->natt_type == 0 || sav->esp_frag == IP_MAXPACKET)
   3872 				continue;
   3873 			m = key_setsadbxfrag(sav->esp_frag);
   3874 			break;
   3875 
   3876 		case SADB_X_EXT_NAT_T_OAI:
   3877 		case SADB_X_EXT_NAT_T_OAR:
   3878 			continue;
   3879 
   3880 		case SADB_EXT_ADDRESS_PROXY:
   3881 		case SADB_EXT_IDENTITY_SRC:
   3882 		case SADB_EXT_IDENTITY_DST:
   3883 			/* XXX: should we brought from SPD ? */
   3884 		case SADB_EXT_SENSITIVITY:
   3885 		default:
   3886 			continue;
   3887 		}
   3888 
   3889 		KASSERT(!(m && p));
   3890 		KASSERT(m != NULL || p != NULL);
   3891 		if (p && tres) {
   3892 			M_PREPEND(tres, l, M_WAITOK);
   3893 			memcpy(mtod(tres, void *), p, l);
   3894 			continue;
   3895 		}
   3896 		if (p) {
   3897 			m = key_alloc_mbuf(l, M_WAITOK);
   3898 			m_copyback(m, 0, l, p);
   3899 		}
   3900 
   3901 		if (tres)
   3902 			m_cat(m, tres);
   3903 		tres = m;
   3904 	}
   3905 
   3906 	m_cat(result, tres);
   3907 	tres = NULL; /* avoid free on error below */
   3908 
   3909 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
   3910 
   3911 	result->m_pkthdr.len = 0;
   3912 	for (m = result; m; m = m->m_next)
   3913 		result->m_pkthdr.len += m->m_len;
   3914 
   3915 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   3916 	    PFKEY_UNIT64(result->m_pkthdr.len);
   3917 
   3918 	return result;
   3919 }
   3920 
   3921 
   3922 /*
   3923  * set a type in sadb_x_nat_t_type
   3924  */
   3925 static struct mbuf *
   3926 key_setsadbxtype(u_int16_t type)
   3927 {
   3928 	struct mbuf *m;
   3929 	size_t len;
   3930 	struct sadb_x_nat_t_type *p;
   3931 
   3932 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
   3933 
   3934 	m = key_alloc_mbuf(len, M_WAITOK);
   3935 	KASSERT(m->m_next == NULL);
   3936 
   3937 	p = mtod(m, struct sadb_x_nat_t_type *);
   3938 
   3939 	memset(p, 0, len);
   3940 	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
   3941 	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
   3942 	p->sadb_x_nat_t_type_type = type;
   3943 
   3944 	return m;
   3945 }
   3946 /*
   3947  * set a port in sadb_x_nat_t_port. port is in network order
   3948  */
   3949 static struct mbuf *
   3950 key_setsadbxport(u_int16_t port, u_int16_t type)
   3951 {
   3952 	struct mbuf *m;
   3953 	size_t len;
   3954 	struct sadb_x_nat_t_port *p;
   3955 
   3956 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
   3957 
   3958 	m = key_alloc_mbuf(len, M_WAITOK);
   3959 	KASSERT(m->m_next == NULL);
   3960 
   3961 	p = mtod(m, struct sadb_x_nat_t_port *);
   3962 
   3963 	memset(p, 0, len);
   3964 	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
   3965 	p->sadb_x_nat_t_port_exttype = type;
   3966 	p->sadb_x_nat_t_port_port = port;
   3967 
   3968 	return m;
   3969 }
   3970 
   3971 /*
   3972  * set fragmentation info in sadb_x_nat_t_frag
   3973  */
   3974 static struct mbuf *
   3975 key_setsadbxfrag(u_int16_t flen)
   3976 {
   3977 	struct mbuf *m;
   3978 	size_t len;
   3979 	struct sadb_x_nat_t_frag *p;
   3980 
   3981 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_frag));
   3982 
   3983 	m = key_alloc_mbuf(len, M_WAITOK);
   3984 	KASSERT(m->m_next == NULL);
   3985 
   3986 	p = mtod(m, struct sadb_x_nat_t_frag *);
   3987 
   3988 	memset(p, 0, len);
   3989 	p->sadb_x_nat_t_frag_len = PFKEY_UNIT64(len);
   3990 	p->sadb_x_nat_t_frag_exttype = SADB_X_EXT_NAT_T_FRAG;
   3991 	p->sadb_x_nat_t_frag_fraglen = flen;
   3992 
   3993 	return m;
   3994 }
   3995 
   3996 /*
   3997  * Get port from sockaddr, port is in network order
   3998  */
   3999 u_int16_t
   4000 key_portfromsaddr(const union sockaddr_union *saddr)
   4001 {
   4002 	u_int16_t port;
   4003 
   4004 	switch (saddr->sa.sa_family) {
   4005 	case AF_INET: {
   4006 		port = saddr->sin.sin_port;
   4007 		break;
   4008 	}
   4009 #ifdef INET6
   4010 	case AF_INET6: {
   4011 		port = saddr->sin6.sin6_port;
   4012 		break;
   4013 	}
   4014 #endif
   4015 	default:
   4016 		printf("%s: unexpected address family\n", __func__);
   4017 		port = 0;
   4018 		break;
   4019 	}
   4020 
   4021 	return port;
   4022 }
   4023 
   4024 
   4025 /*
   4026  * Set port is struct sockaddr. port is in network order
   4027  */
   4028 static void
   4029 key_porttosaddr(union sockaddr_union *saddr, u_int16_t port)
   4030 {
   4031 	switch (saddr->sa.sa_family) {
   4032 	case AF_INET: {
   4033 		saddr->sin.sin_port = port;
   4034 		break;
   4035 	}
   4036 #ifdef INET6
   4037 	case AF_INET6: {
   4038 		saddr->sin6.sin6_port = port;
   4039 		break;
   4040 	}
   4041 #endif
   4042 	default:
   4043 		printf("%s: unexpected address family %d\n", __func__,
   4044 		    saddr->sa.sa_family);
   4045 		break;
   4046 	}
   4047 
   4048 	return;
   4049 }
   4050 
   4051 /*
   4052  * Safety check sa_len
   4053  */
   4054 static int
   4055 key_checksalen(const union sockaddr_union *saddr)
   4056 {
   4057 	switch (saddr->sa.sa_family) {
   4058 	case AF_INET:
   4059 		if (saddr->sa.sa_len != sizeof(struct sockaddr_in))
   4060 			return -1;
   4061 		break;
   4062 #ifdef INET6
   4063 	case AF_INET6:
   4064 		if (saddr->sa.sa_len != sizeof(struct sockaddr_in6))
   4065 			return -1;
   4066 		break;
   4067 #endif
   4068 	default:
   4069 		printf("%s: unexpected sa_family %d\n", __func__,
   4070 		    saddr->sa.sa_family);
   4071 			return -1;
   4072 		break;
   4073 	}
   4074 	return 0;
   4075 }
   4076 
   4077 
   4078 /*
   4079  * set data into sadb_msg.
   4080  */
   4081 static struct mbuf *
   4082 key_setsadbmsg(u_int8_t type,  u_int16_t tlen, u_int8_t satype,
   4083 	       u_int32_t seq, pid_t pid, u_int16_t reserved, int mflag)
   4084 {
   4085 	struct mbuf *m;
   4086 	struct sadb_msg *p;
   4087 	int len;
   4088 
   4089 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MCLBYTES);
   4090 
   4091 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
   4092 
   4093 	m = key_alloc_mbuf_simple(len, mflag);
   4094 	if (!m)
   4095 		return NULL;
   4096 	m->m_pkthdr.len = m->m_len = len;
   4097 	m->m_next = NULL;
   4098 
   4099 	p = mtod(m, struct sadb_msg *);
   4100 
   4101 	memset(p, 0, len);
   4102 	p->sadb_msg_version = PF_KEY_V2;
   4103 	p->sadb_msg_type = type;
   4104 	p->sadb_msg_errno = 0;
   4105 	p->sadb_msg_satype = satype;
   4106 	p->sadb_msg_len = PFKEY_UNIT64(tlen);
   4107 	p->sadb_msg_reserved = reserved;
   4108 	p->sadb_msg_seq = seq;
   4109 	p->sadb_msg_pid = (u_int32_t)pid;
   4110 
   4111 	return m;
   4112 }
   4113 
   4114 /*
   4115  * copy secasvar data into sadb_address.
   4116  */
   4117 static struct mbuf *
   4118 key_setsadbsa(struct secasvar *sav)
   4119 {
   4120 	struct mbuf *m;
   4121 	struct sadb_sa *p;
   4122 	int len;
   4123 
   4124 	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
   4125 	m = key_alloc_mbuf(len, M_WAITOK);
   4126 	KASSERT(m->m_next == NULL);
   4127 
   4128 	p = mtod(m, struct sadb_sa *);
   4129 
   4130 	memset(p, 0, len);
   4131 	p->sadb_sa_len = PFKEY_UNIT64(len);
   4132 	p->sadb_sa_exttype = SADB_EXT_SA;
   4133 	p->sadb_sa_spi = sav->spi;
   4134 	p->sadb_sa_replay = (sav->replay != NULL ? sav->replay->wsize : 0);
   4135 	p->sadb_sa_state = sav->state;
   4136 	p->sadb_sa_auth = sav->alg_auth;
   4137 	p->sadb_sa_encrypt = sav->alg_enc;
   4138 	p->sadb_sa_flags = sav->flags;
   4139 
   4140 	return m;
   4141 }
   4142 
   4143 /*
   4144  * set data into sadb_address.
   4145  */
   4146 static struct mbuf *
   4147 key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
   4148 		u_int8_t prefixlen, u_int16_t ul_proto, int mflag)
   4149 {
   4150 	struct mbuf *m;
   4151 	struct sadb_address *p;
   4152 	size_t len;
   4153 
   4154 	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
   4155 	    PFKEY_ALIGN8(saddr->sa_len);
   4156 	m = key_alloc_mbuf(len, mflag);
   4157 	if (!m || m->m_next) {	/*XXX*/
   4158 		if (m)
   4159 			m_freem(m);
   4160 		return NULL;
   4161 	}
   4162 
   4163 	p = mtod(m, struct sadb_address *);
   4164 
   4165 	memset(p, 0, len);
   4166 	p->sadb_address_len = PFKEY_UNIT64(len);
   4167 	p->sadb_address_exttype = exttype;
   4168 	p->sadb_address_proto = ul_proto;
   4169 	if (prefixlen == FULLMASK) {
   4170 		switch (saddr->sa_family) {
   4171 		case AF_INET:
   4172 			prefixlen = sizeof(struct in_addr) << 3;
   4173 			break;
   4174 		case AF_INET6:
   4175 			prefixlen = sizeof(struct in6_addr) << 3;
   4176 			break;
   4177 		default:
   4178 			; /*XXX*/
   4179 		}
   4180 	}
   4181 	p->sadb_address_prefixlen = prefixlen;
   4182 	p->sadb_address_reserved = 0;
   4183 
   4184 	memcpy(mtod(m, char *) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
   4185 	    saddr, saddr->sa_len);
   4186 
   4187 	return m;
   4188 }
   4189 
   4190 #if 0
   4191 /*
   4192  * set data into sadb_ident.
   4193  */
   4194 static struct mbuf *
   4195 key_setsadbident(u_int16_t exttype, u_int16_t idtype,
   4196 		 void *string, int stringlen, u_int64_t id)
   4197 {
   4198 	struct mbuf *m;
   4199 	struct sadb_ident *p;
   4200 	size_t len;
   4201 
   4202 	len = PFKEY_ALIGN8(sizeof(struct sadb_ident)) + PFKEY_ALIGN8(stringlen);
   4203 	m = key_alloc_mbuf(len);
   4204 	if (!m || m->m_next) {	/*XXX*/
   4205 		if (m)
   4206 			m_freem(m);
   4207 		return NULL;
   4208 	}
   4209 
   4210 	p = mtod(m, struct sadb_ident *);
   4211 
   4212 	memset(p, 0, len);
   4213 	p->sadb_ident_len = PFKEY_UNIT64(len);
   4214 	p->sadb_ident_exttype = exttype;
   4215 	p->sadb_ident_type = idtype;
   4216 	p->sadb_ident_reserved = 0;
   4217 	p->sadb_ident_id = id;
   4218 
   4219 	memcpy(mtod(m, void *) + PFKEY_ALIGN8(sizeof(struct sadb_ident)),
   4220 	   	   string, stringlen);
   4221 
   4222 	return m;
   4223 }
   4224 #endif
   4225 
   4226 /*
   4227  * set data into sadb_x_sa2.
   4228  */
   4229 static struct mbuf *
   4230 key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int16_t reqid)
   4231 {
   4232 	struct mbuf *m;
   4233 	struct sadb_x_sa2 *p;
   4234 	size_t len;
   4235 
   4236 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
   4237 	m = key_alloc_mbuf(len, M_WAITOK);
   4238 	KASSERT(m->m_next == NULL);
   4239 
   4240 	p = mtod(m, struct sadb_x_sa2 *);
   4241 
   4242 	memset(p, 0, len);
   4243 	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
   4244 	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
   4245 	p->sadb_x_sa2_mode = mode;
   4246 	p->sadb_x_sa2_reserved1 = 0;
   4247 	p->sadb_x_sa2_reserved2 = 0;
   4248 	p->sadb_x_sa2_sequence = seq;
   4249 	p->sadb_x_sa2_reqid = reqid;
   4250 
   4251 	return m;
   4252 }
   4253 
   4254 /*
   4255  * set data into sadb_x_policy
   4256  */
   4257 static struct mbuf *
   4258 key_setsadbxpolicy(const u_int16_t type, const u_int8_t dir, const u_int32_t id,
   4259     int mflag)
   4260 {
   4261 	struct mbuf *m;
   4262 	struct sadb_x_policy *p;
   4263 	size_t len;
   4264 
   4265 	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
   4266 	m = key_alloc_mbuf(len, mflag);
   4267 	if (!m || m->m_next) {	/*XXX*/
   4268 		if (m)
   4269 			m_freem(m);
   4270 		return NULL;
   4271 	}
   4272 
   4273 	p = mtod(m, struct sadb_x_policy *);
   4274 
   4275 	memset(p, 0, len);
   4276 	p->sadb_x_policy_len = PFKEY_UNIT64(len);
   4277 	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
   4278 	p->sadb_x_policy_type = type;
   4279 	p->sadb_x_policy_dir = dir;
   4280 	p->sadb_x_policy_id = id;
   4281 
   4282 	return m;
   4283 }
   4284 
   4285 /* %%% utilities */
   4286 /*
   4287  * copy a buffer into the new buffer allocated.
   4288  */
   4289 static void *
   4290 key_newbuf(const void *src, u_int len)
   4291 {
   4292 	void *new;
   4293 
   4294 	new = kmem_alloc(len, KM_SLEEP);
   4295 	memcpy(new, src, len);
   4296 
   4297 	return new;
   4298 }
   4299 
   4300 /* compare my own address
   4301  * OUT:	1: true, i.e. my address.
   4302  *	0: false
   4303  */
   4304 int
   4305 key_ismyaddr(const struct sockaddr *sa)
   4306 {
   4307 #ifdef INET
   4308 	const struct sockaddr_in *sin;
   4309 	const struct in_ifaddr *ia;
   4310 	int s;
   4311 #endif
   4312 
   4313 	KASSERT(sa != NULL);
   4314 
   4315 	switch (sa->sa_family) {
   4316 #ifdef INET
   4317 	case AF_INET:
   4318 		sin = (const struct sockaddr_in *)sa;
   4319 		s = pserialize_read_enter();
   4320 		IN_ADDRLIST_READER_FOREACH(ia) {
   4321 			if (sin->sin_family == ia->ia_addr.sin_family &&
   4322 			    sin->sin_len == ia->ia_addr.sin_len &&
   4323 			    sin->sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
   4324 			{
   4325 				pserialize_read_exit(s);
   4326 				return 1;
   4327 			}
   4328 		}
   4329 		pserialize_read_exit(s);
   4330 		break;
   4331 #endif
   4332 #ifdef INET6
   4333 	case AF_INET6:
   4334 		return key_ismyaddr6((const struct sockaddr_in6 *)sa);
   4335 #endif
   4336 	}
   4337 
   4338 	return 0;
   4339 }
   4340 
   4341 #ifdef INET6
   4342 /*
   4343  * compare my own address for IPv6.
   4344  * 1: ours
   4345  * 0: other
   4346  * NOTE: derived ip6_input() in KAME. This is necessary to modify more.
   4347  */
   4348 #include <netinet6/in6_var.h>
   4349 
   4350 static int
   4351 key_ismyaddr6(const struct sockaddr_in6 *sin6)
   4352 {
   4353 	struct in6_ifaddr *ia;
   4354 	int s;
   4355 	struct psref psref;
   4356 	int bound;
   4357 	int ours = 1;
   4358 
   4359 	bound = curlwp_bind();
   4360 	s = pserialize_read_enter();
   4361 	IN6_ADDRLIST_READER_FOREACH(ia) {
   4362 		bool ingroup;
   4363 
   4364 		if (key_sockaddr_match((const struct sockaddr *)&sin6,
   4365 		    (const struct sockaddr *)&ia->ia_addr, 0)) {
   4366 			pserialize_read_exit(s);
   4367 			goto ours;
   4368 		}
   4369 		ia6_acquire(ia, &psref);
   4370 		pserialize_read_exit(s);
   4371 
   4372 		/*
   4373 		 * XXX Multicast
   4374 		 * XXX why do we care about multlicast here while we don't care
   4375 		 * about IPv4 multicast??
   4376 		 * XXX scope
   4377 		 */
   4378 		ingroup = in6_multi_group(&sin6->sin6_addr, ia->ia_ifp);
   4379 		if (ingroup) {
   4380 			ia6_release(ia, &psref);
   4381 			goto ours;
   4382 		}
   4383 
   4384 		s = pserialize_read_enter();
   4385 		ia6_release(ia, &psref);
   4386 	}
   4387 	pserialize_read_exit(s);
   4388 
   4389 	/* loopback, just for safety */
   4390 	if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))
   4391 		goto ours;
   4392 
   4393 	ours = 0;
   4394 ours:
   4395 	curlwp_bindx(bound);
   4396 
   4397 	return ours;
   4398 }
   4399 #endif /*INET6*/
   4400 
   4401 /*
   4402  * compare two secasindex structure.
   4403  * flag can specify to compare 2 saidxes.
   4404  * compare two secasindex structure without both mode and reqid.
   4405  * don't compare port.
   4406  * IN:
   4407  *      saidx0: source, it can be in SAD.
   4408  *      saidx1: object.
   4409  * OUT:
   4410  *      1 : equal
   4411  *      0 : not equal
   4412  */
   4413 static int
   4414 key_saidx_match(
   4415 	const struct secasindex *saidx0,
   4416 	const struct secasindex *saidx1,
   4417 	int flag)
   4418 {
   4419 	int chkport;
   4420 	const struct sockaddr *sa0src, *sa0dst, *sa1src, *sa1dst;
   4421 
   4422 	KASSERT(saidx0 != NULL);
   4423 	KASSERT(saidx1 != NULL);
   4424 
   4425 	/* sanity */
   4426 	if (saidx0->proto != saidx1->proto)
   4427 		return 0;
   4428 
   4429 	if (flag == CMP_EXACTLY) {
   4430 		if (saidx0->mode != saidx1->mode)
   4431 			return 0;
   4432 		if (saidx0->reqid != saidx1->reqid)
   4433 			return 0;
   4434 		if (memcmp(&saidx0->src, &saidx1->src, saidx0->src.sa.sa_len) != 0 ||
   4435 		    memcmp(&saidx0->dst, &saidx1->dst, saidx0->dst.sa.sa_len) != 0)
   4436 			return 0;
   4437 	} else {
   4438 
   4439 		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
   4440 		if (flag == CMP_MODE_REQID ||flag == CMP_REQID) {
   4441 			/*
   4442 			 * If reqid of SPD is non-zero, unique SA is required.
   4443 			 * The result must be of same reqid in this case.
   4444 			 */
   4445 			if (saidx1->reqid != 0 && saidx0->reqid != saidx1->reqid)
   4446 				return 0;
   4447 		}
   4448 
   4449 		if (flag == CMP_MODE_REQID) {
   4450 			if (saidx0->mode != IPSEC_MODE_ANY &&
   4451 			    saidx0->mode != saidx1->mode)
   4452 				return 0;
   4453 		}
   4454 
   4455 
   4456 		sa0src = &saidx0->src.sa;
   4457 		sa0dst = &saidx0->dst.sa;
   4458 		sa1src = &saidx1->src.sa;
   4459 		sa1dst = &saidx1->dst.sa;
   4460 		/*
   4461 		 * If NAT-T is enabled, check ports for tunnel mode.
   4462 		 * Don't do it for transport mode, as there is no
   4463 		 * port information available in the SP.
   4464 		 * Also don't check ports if they are set to zero
   4465 		 * in the SPD: This means we have a non-generated
   4466 		 * SPD which can't know UDP ports.
   4467 		 */
   4468 		if (saidx1->mode == IPSEC_MODE_TUNNEL)
   4469 			chkport = PORT_LOOSE;
   4470 		else
   4471 			chkport = PORT_NONE;
   4472 
   4473 		if (!key_sockaddr_match(sa0src, sa1src, chkport)) {
   4474 			return 0;
   4475 		}
   4476 		if (!key_sockaddr_match(sa0dst, sa1dst, chkport)) {
   4477 			return 0;
   4478 		}
   4479 	}
   4480 
   4481 	return 1;
   4482 }
   4483 
   4484 /*
   4485  * compare two secindex structure exactly.
   4486  * IN:
   4487  *	spidx0: source, it is often in SPD.
   4488  *	spidx1: object, it is often from PFKEY message.
   4489  * OUT:
   4490  *	1 : equal
   4491  *	0 : not equal
   4492  */
   4493 static int
   4494 key_spidx_match_exactly(
   4495 	const struct secpolicyindex *spidx0,
   4496 	const struct secpolicyindex *spidx1)
   4497 {
   4498 
   4499 	KASSERT(spidx0 != NULL);
   4500 	KASSERT(spidx1 != NULL);
   4501 
   4502 	/* sanity */
   4503 	if (spidx0->prefs != spidx1->prefs ||
   4504 	    spidx0->prefd != spidx1->prefd ||
   4505 	    spidx0->ul_proto != spidx1->ul_proto)
   4506 		return 0;
   4507 
   4508 	return key_sockaddr_match(&spidx0->src.sa, &spidx1->src.sa, PORT_STRICT) &&
   4509 	       key_sockaddr_match(&spidx0->dst.sa, &spidx1->dst.sa, PORT_STRICT);
   4510 }
   4511 
   4512 /*
   4513  * compare two secindex structure with mask.
   4514  * IN:
   4515  *	spidx0: source, it is often in SPD.
   4516  *	spidx1: object, it is often from IP header.
   4517  * OUT:
   4518  *	1 : equal
   4519  *	0 : not equal
   4520  */
   4521 static int
   4522 key_spidx_match_withmask(
   4523 	const struct secpolicyindex *spidx0,
   4524 	const struct secpolicyindex *spidx1)
   4525 {
   4526 
   4527 	KASSERT(spidx0 != NULL);
   4528 	KASSERT(spidx1 != NULL);
   4529 
   4530 	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
   4531 	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
   4532 	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
   4533 	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len)
   4534 		return 0;
   4535 
   4536 	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
   4537 	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY &&
   4538 	    spidx0->ul_proto != spidx1->ul_proto)
   4539 		return 0;
   4540 
   4541 	switch (spidx0->src.sa.sa_family) {
   4542 	case AF_INET:
   4543 		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY &&
   4544 		    spidx0->src.sin.sin_port != spidx1->src.sin.sin_port)
   4545 			return 0;
   4546 		if (!key_bb_match_withmask(&spidx0->src.sin.sin_addr,
   4547 		    &spidx1->src.sin.sin_addr, spidx0->prefs))
   4548 			return 0;
   4549 		break;
   4550 	case AF_INET6:
   4551 		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY &&
   4552 		    spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port)
   4553 			return 0;
   4554 		/*
   4555 		 * scope_id check. if sin6_scope_id is 0, we regard it
   4556 		 * as a wildcard scope, which matches any scope zone ID.
   4557 		 */
   4558 		if (spidx0->src.sin6.sin6_scope_id &&
   4559 		    spidx1->src.sin6.sin6_scope_id &&
   4560 		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id)
   4561 			return 0;
   4562 		if (!key_bb_match_withmask(&spidx0->src.sin6.sin6_addr,
   4563 		    &spidx1->src.sin6.sin6_addr, spidx0->prefs))
   4564 			return 0;
   4565 		break;
   4566 	default:
   4567 		/* XXX */
   4568 		if (memcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0)
   4569 			return 0;
   4570 		break;
   4571 	}
   4572 
   4573 	switch (spidx0->dst.sa.sa_family) {
   4574 	case AF_INET:
   4575 		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY &&
   4576 		    spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port)
   4577 			return 0;
   4578 		if (!key_bb_match_withmask(&spidx0->dst.sin.sin_addr,
   4579 		    &spidx1->dst.sin.sin_addr, spidx0->prefd))
   4580 			return 0;
   4581 		break;
   4582 	case AF_INET6:
   4583 		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY &&
   4584 		    spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port)
   4585 			return 0;
   4586 		/*
   4587 		 * scope_id check. if sin6_scope_id is 0, we regard it
   4588 		 * as a wildcard scope, which matches any scope zone ID.
   4589 		 */
   4590 		if (spidx0->src.sin6.sin6_scope_id &&
   4591 		    spidx1->src.sin6.sin6_scope_id &&
   4592 		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id)
   4593 			return 0;
   4594 		if (!key_bb_match_withmask(&spidx0->dst.sin6.sin6_addr,
   4595 		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd))
   4596 			return 0;
   4597 		break;
   4598 	default:
   4599 		/* XXX */
   4600 		if (memcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0)
   4601 			return 0;
   4602 		break;
   4603 	}
   4604 
   4605 	/* XXX Do we check other field ?  e.g. flowinfo */
   4606 
   4607 	return 1;
   4608 }
   4609 
   4610 /* returns 0 on match */
   4611 static int
   4612 key_portcomp(in_port_t port1, in_port_t port2, int howport)
   4613 {
   4614 	switch (howport) {
   4615 	case PORT_NONE:
   4616 		return 0;
   4617 	case PORT_LOOSE:
   4618 		if (port1 == 0 || port2 == 0)
   4619 			return 0;
   4620 		/*FALLTHROUGH*/
   4621 	case PORT_STRICT:
   4622 		if (port1 != port2) {
   4623 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   4624 			    "port fail %d != %d\n", port1, port2);
   4625 			return 1;
   4626 		}
   4627 		return 0;
   4628 	default:
   4629 		KASSERT(0);
   4630 		return 1;
   4631 	}
   4632 }
   4633 
   4634 /* returns 1 on match */
   4635 static int
   4636 key_sockaddr_match(
   4637 	const struct sockaddr *sa1,
   4638 	const struct sockaddr *sa2,
   4639 	int howport)
   4640 {
   4641 	const struct sockaddr_in *sin1, *sin2;
   4642 	const struct sockaddr_in6 *sin61, *sin62;
   4643 	char s1[IPSEC_ADDRSTRLEN], s2[IPSEC_ADDRSTRLEN];
   4644 
   4645 	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len) {
   4646 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   4647 		    "fam/len fail %d != %d || %d != %d\n",
   4648 			sa1->sa_family, sa2->sa_family, sa1->sa_len,
   4649 			sa2->sa_len);
   4650 		return 0;
   4651 	}
   4652 
   4653 	switch (sa1->sa_family) {
   4654 	case AF_INET:
   4655 		if (sa1->sa_len != sizeof(struct sockaddr_in)) {
   4656 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   4657 			    "len fail %d != %zu\n",
   4658 			    sa1->sa_len, sizeof(struct sockaddr_in));
   4659 			return 0;
   4660 		}
   4661 		sin1 = (const struct sockaddr_in *)sa1;
   4662 		sin2 = (const struct sockaddr_in *)sa2;
   4663 		if (sin1->sin_addr.s_addr != sin2->sin_addr.s_addr) {
   4664 			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   4665 			    "addr fail %s != %s\n",
   4666 			    (in_print(s1, sizeof(s1), &sin1->sin_addr), s1),
   4667 			    (in_print(s2, sizeof(s2), &sin2->sin_addr), s2));
   4668 			return 0;
   4669 		}
   4670 		if (key_portcomp(sin1->sin_port, sin2->sin_port, howport)) {
   4671 			return 0;
   4672 		}
   4673 		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
   4674 		    "addr success %s[%d] == %s[%d]\n",
   4675 		    (in_print(s1, sizeof(s1), &sin1->sin_addr), s1),
   4676 		    sin1->sin_port,
   4677 		    (in_print(s2, sizeof(s2), &sin2->sin_addr), s2),
   4678 		    sin2->sin_port);
   4679 		break;
   4680 	case AF_INET6:
   4681 		sin61 = (const struct sockaddr_in6 *)sa1;
   4682 		sin62 = (const struct sockaddr_in6 *)sa2;
   4683 		if (sa1->sa_len != sizeof(struct sockaddr_in6))
   4684 			return 0;	/*EINVAL*/
   4685 
   4686 		if (sin61->sin6_scope_id != sin62->sin6_scope_id) {
   4687 			return 0;
   4688 		}
   4689 		if (!IN6_ARE_ADDR_EQUAL(&sin61->sin6_addr, &sin62->sin6_addr)) {
   4690 			return 0;
   4691 		}
   4692 		if (key_portcomp(sin61->sin6_port, sin62->sin6_port, howport)) {
   4693 			return 0;
   4694 		}
   4695 		break;
   4696 	default:
   4697 		if (memcmp(sa1, sa2, sa1->sa_len) != 0)
   4698 			return 0;
   4699 		break;
   4700 	}
   4701 
   4702 	return 1;
   4703 }
   4704 
   4705 /*
   4706  * compare two buffers with mask.
   4707  * IN:
   4708  *	addr1: source
   4709  *	addr2: object
   4710  *	bits:  Number of bits to compare
   4711  * OUT:
   4712  *	1 : equal
   4713  *	0 : not equal
   4714  */
   4715 static int
   4716 key_bb_match_withmask(const void *a1, const void *a2, u_int bits)
   4717 {
   4718 	const unsigned char *p1 = a1;
   4719 	const unsigned char *p2 = a2;
   4720 
   4721 	/* XXX: This could be considerably faster if we compare a word
   4722 	 * at a time, but it is complicated on LSB Endian machines */
   4723 
   4724 	/* Handle null pointers */
   4725 	if (p1 == NULL || p2 == NULL)
   4726 		return (p1 == p2);
   4727 
   4728 	while (bits >= 8) {
   4729 		if (*p1++ != *p2++)
   4730 			return 0;
   4731 		bits -= 8;
   4732 	}
   4733 
   4734 	if (bits > 0) {
   4735 		u_int8_t mask = ~((1<<(8-bits))-1);
   4736 		if ((*p1 & mask) != (*p2 & mask))
   4737 			return 0;
   4738 	}
   4739 	return 1;	/* Match! */
   4740 }
   4741 
   4742 static void
   4743 key_timehandler_spd(time_t now)
   4744 {
   4745 	u_int dir;
   4746 	struct secpolicy *sp;
   4747 
   4748 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
   4749 	    retry:
   4750 		mutex_enter(&key_spd.lock);
   4751 		SPLIST_WRITER_FOREACH(sp, dir) {
   4752 			KASSERT(sp->state != IPSEC_SPSTATE_DEAD);
   4753 
   4754 			if (sp->lifetime == 0 && sp->validtime == 0)
   4755 				continue;
   4756 
   4757 			if ((sp->lifetime && now - sp->created > sp->lifetime) ||
   4758 			    (sp->validtime && now - sp->lastused > sp->validtime)) {
   4759 				key_unlink_sp(sp);
   4760 				mutex_exit(&key_spd.lock);
   4761 				key_spdexpire(sp);
   4762 				key_destroy_sp(sp);
   4763 				goto retry;
   4764 			}
   4765 		}
   4766 		mutex_exit(&key_spd.lock);
   4767 	}
   4768 
   4769     retry_socksplist:
   4770 	mutex_enter(&key_spd.lock);
   4771 	SOCKSPLIST_WRITER_FOREACH(sp) {
   4772 		if (sp->state != IPSEC_SPSTATE_DEAD)
   4773 			continue;
   4774 
   4775 		key_unlink_sp(sp);
   4776 		mutex_exit(&key_spd.lock);
   4777 		key_destroy_sp(sp);
   4778 		goto retry_socksplist;
   4779 	}
   4780 	mutex_exit(&key_spd.lock);
   4781 }
   4782 
   4783 static void
   4784 key_timehandler_sad(time_t now)
   4785 {
   4786 	struct secashead *sah;
   4787 	int s;
   4788 
   4789 restart:
   4790 	mutex_enter(&key_sad.lock);
   4791 	SAHLIST_WRITER_FOREACH(sah) {
   4792 		/* If sah has been dead and has no sav, then delete it */
   4793 		if (sah->state == SADB_SASTATE_DEAD &&
   4794 		    !key_sah_has_sav(sah)) {
   4795 			key_unlink_sah(sah);
   4796 			mutex_exit(&key_sad.lock);
   4797 			key_destroy_sah(sah);
   4798 			goto restart;
   4799 		}
   4800 	}
   4801 	mutex_exit(&key_sad.lock);
   4802 
   4803 	s = pserialize_read_enter();
   4804 	SAHLIST_READER_FOREACH(sah) {
   4805 		struct secasvar *sav;
   4806 
   4807 		key_sah_ref(sah);
   4808 		pserialize_read_exit(s);
   4809 
   4810 		/* if LARVAL entry doesn't become MATURE, delete it. */
   4811 		mutex_enter(&key_sad.lock);
   4812 	restart_sav_LARVAL:
   4813 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_LARVAL) {
   4814 			if (now - sav->created > key_larval_lifetime) {
   4815 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
   4816 				goto restart_sav_LARVAL;
   4817 			}
   4818 		}
   4819 		mutex_exit(&key_sad.lock);
   4820 
   4821 		/*
   4822 		 * check MATURE entry to start to send expire message
   4823 		 * whether or not.
   4824 		 */
   4825 	restart_sav_MATURE:
   4826 		mutex_enter(&key_sad.lock);
   4827 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_MATURE) {
   4828 			/* we don't need to check. */
   4829 			if (sav->lft_s == NULL)
   4830 				continue;
   4831 
   4832 			/* sanity check */
   4833 			KASSERT(sav->lft_c != NULL);
   4834 
   4835 			/* check SOFT lifetime */
   4836 			if (sav->lft_s->sadb_lifetime_addtime != 0 &&
   4837 			    now - sav->created > sav->lft_s->sadb_lifetime_addtime) {
   4838 				/*
   4839 				 * check SA to be used whether or not.
   4840 				 * when SA hasn't been used, delete it.
   4841 				 */
   4842 				if (sav->lft_c->sadb_lifetime_usetime == 0) {
   4843 					key_sa_chgstate(sav, SADB_SASTATE_DEAD);
   4844 					mutex_exit(&key_sad.lock);
   4845 				} else {
   4846 					key_sa_chgstate(sav, SADB_SASTATE_DYING);
   4847 					mutex_exit(&key_sad.lock);
   4848 					/*
   4849 					 * XXX If we keep to send expire
   4850 					 * message in the status of
   4851 					 * DYING. Do remove below code.
   4852 					 */
   4853 					key_expire(sav);
   4854 				}
   4855 				goto restart_sav_MATURE;
   4856 			}
   4857 			/* check SOFT lifetime by bytes */
   4858 			/*
   4859 			 * XXX I don't know the way to delete this SA
   4860 			 * when new SA is installed.  Caution when it's
   4861 			 * installed too big lifetime by time.
   4862 			 */
   4863 			else if (sav->lft_s->sadb_lifetime_bytes != 0 &&
   4864 			         sav->lft_s->sadb_lifetime_bytes <
   4865 			         sav->lft_c->sadb_lifetime_bytes) {
   4866 
   4867 				key_sa_chgstate(sav, SADB_SASTATE_DYING);
   4868 				mutex_exit(&key_sad.lock);
   4869 				/*
   4870 				 * XXX If we keep to send expire
   4871 				 * message in the status of
   4872 				 * DYING. Do remove below code.
   4873 				 */
   4874 				key_expire(sav);
   4875 				goto restart_sav_MATURE;
   4876 			}
   4877 		}
   4878 		mutex_exit(&key_sad.lock);
   4879 
   4880 		/* check DYING entry to change status to DEAD. */
   4881 		mutex_enter(&key_sad.lock);
   4882 	restart_sav_DYING:
   4883 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_DYING) {
   4884 			/* we don't need to check. */
   4885 			if (sav->lft_h == NULL)
   4886 				continue;
   4887 
   4888 			/* sanity check */
   4889 			KASSERT(sav->lft_c != NULL);
   4890 
   4891 			if (sav->lft_h->sadb_lifetime_addtime != 0 &&
   4892 			    now - sav->created > sav->lft_h->sadb_lifetime_addtime) {
   4893 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
   4894 				goto restart_sav_DYING;
   4895 			}
   4896 #if 0	/* XXX Should we keep to send expire message until HARD lifetime ? */
   4897 			else if (sav->lft_s != NULL
   4898 			      && sav->lft_s->sadb_lifetime_addtime != 0
   4899 			      && now - sav->created > sav->lft_s->sadb_lifetime_addtime) {
   4900 				/*
   4901 				 * XXX: should be checked to be
   4902 				 * installed the valid SA.
   4903 				 */
   4904 
   4905 				/*
   4906 				 * If there is no SA then sending
   4907 				 * expire message.
   4908 				 */
   4909 				key_expire(sav);
   4910 			}
   4911 #endif
   4912 			/* check HARD lifetime by bytes */
   4913 			else if (sav->lft_h->sadb_lifetime_bytes != 0 &&
   4914 			         sav->lft_h->sadb_lifetime_bytes <
   4915 			         sav->lft_c->sadb_lifetime_bytes) {
   4916 				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
   4917 				goto restart_sav_DYING;
   4918 			}
   4919 		}
   4920 		mutex_exit(&key_sad.lock);
   4921 
   4922 		/* delete entry in DEAD */
   4923 	restart_sav_DEAD:
   4924 		mutex_enter(&key_sad.lock);
   4925 		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_DEAD) {
   4926 			key_unlink_sav(sav);
   4927 			mutex_exit(&key_sad.lock);
   4928 			key_destroy_sav(sav);
   4929 			goto restart_sav_DEAD;
   4930 		}
   4931 		mutex_exit(&key_sad.lock);
   4932 
   4933 		s = pserialize_read_enter();
   4934 		key_sah_unref(sah);
   4935 	}
   4936 	pserialize_read_exit(s);
   4937 }
   4938 
   4939 static void
   4940 key_timehandler_acq(time_t now)
   4941 {
   4942 #ifndef IPSEC_NONBLOCK_ACQUIRE
   4943 	struct secacq *acq, *nextacq;
   4944 
   4945     restart:
   4946 	mutex_enter(&key_misc.lock);
   4947 	LIST_FOREACH_SAFE(acq, &key_misc.acqlist, chain, nextacq) {
   4948 		if (now - acq->created > key_blockacq_lifetime) {
   4949 			LIST_REMOVE(acq, chain);
   4950 			mutex_exit(&key_misc.lock);
   4951 			kmem_free(acq, sizeof(*acq));
   4952 			goto restart;
   4953 		}
   4954 	}
   4955 	mutex_exit(&key_misc.lock);
   4956 #endif
   4957 }
   4958 
   4959 static void
   4960 key_timehandler_spacq(time_t now)
   4961 {
   4962 #ifdef notyet
   4963 	struct secspacq *acq, *nextacq;
   4964 
   4965 	LIST_FOREACH_SAFE(acq, &key_misc.spacqlist, chain, nextacq) {
   4966 		if (now - acq->created > key_blockacq_lifetime) {
   4967 			KASSERT(__LIST_CHAINED(acq));
   4968 			LIST_REMOVE(acq, chain);
   4969 			kmem_free(acq, sizeof(*acq));
   4970 		}
   4971 	}
   4972 #endif
   4973 }
   4974 
   4975 static unsigned int key_timehandler_work_enqueued = 0;
   4976 
   4977 /*
   4978  * time handler.
   4979  * scanning SPD and SAD to check status for each entries,
   4980  * and do to remove or to expire.
   4981  */
   4982 static void
   4983 key_timehandler_work(struct work *wk, void *arg)
   4984 {
   4985 	time_t now = time_uptime;
   4986 
   4987 	/* We can allow enqueuing another work at this point */
   4988 	atomic_swap_uint(&key_timehandler_work_enqueued, 0);
   4989 
   4990 	key_timehandler_spd(now);
   4991 	key_timehandler_sad(now);
   4992 	key_timehandler_acq(now);
   4993 	key_timehandler_spacq(now);
   4994 
   4995 	key_acquire_sendup_pending_mbuf();
   4996 
   4997 	/* do exchange to tick time !! */
   4998 	callout_reset(&key_timehandler_ch, hz, key_timehandler, NULL);
   4999 
   5000 	return;
   5001 }
   5002 
   5003 static void
   5004 key_timehandler(void *arg)
   5005 {
   5006 
   5007 	/* Avoid enqueuing another work when one is already enqueued */
   5008 	if (atomic_swap_uint(&key_timehandler_work_enqueued, 1) == 1)
   5009 		return;
   5010 
   5011 	workqueue_enqueue(key_timehandler_wq, &key_timehandler_wk, NULL);
   5012 }
   5013 
   5014 u_long
   5015 key_random(void)
   5016 {
   5017 	u_long value;
   5018 
   5019 	key_randomfill(&value, sizeof(value));
   5020 	return value;
   5021 }
   5022 
   5023 void
   5024 key_randomfill(void *p, size_t l)
   5025 {
   5026 
   5027 	cprng_fast(p, l);
   5028 }
   5029 
   5030 /*
   5031  * map SADB_SATYPE_* to IPPROTO_*.
   5032  * if satype == SADB_SATYPE then satype is mapped to ~0.
   5033  * OUT:
   5034  *	0: invalid satype.
   5035  */
   5036 static u_int16_t
   5037 key_satype2proto(u_int8_t satype)
   5038 {
   5039 	switch (satype) {
   5040 	case SADB_SATYPE_UNSPEC:
   5041 		return IPSEC_PROTO_ANY;
   5042 	case SADB_SATYPE_AH:
   5043 		return IPPROTO_AH;
   5044 	case SADB_SATYPE_ESP:
   5045 		return IPPROTO_ESP;
   5046 	case SADB_X_SATYPE_IPCOMP:
   5047 		return IPPROTO_IPCOMP;
   5048 	case SADB_X_SATYPE_TCPSIGNATURE:
   5049 		return IPPROTO_TCP;
   5050 	default:
   5051 		return 0;
   5052 	}
   5053 	/* NOTREACHED */
   5054 }
   5055 
   5056 /*
   5057  * map IPPROTO_* to SADB_SATYPE_*
   5058  * OUT:
   5059  *	0: invalid protocol type.
   5060  */
   5061 static u_int8_t
   5062 key_proto2satype(u_int16_t proto)
   5063 {
   5064 	switch (proto) {
   5065 	case IPPROTO_AH:
   5066 		return SADB_SATYPE_AH;
   5067 	case IPPROTO_ESP:
   5068 		return SADB_SATYPE_ESP;
   5069 	case IPPROTO_IPCOMP:
   5070 		return SADB_X_SATYPE_IPCOMP;
   5071 	case IPPROTO_TCP:
   5072 		return SADB_X_SATYPE_TCPSIGNATURE;
   5073 	default:
   5074 		return 0;
   5075 	}
   5076 	/* NOTREACHED */
   5077 }
   5078 
   5079 static int
   5080 key_setsecasidx(int proto, int mode, int reqid,
   5081     const struct sockaddr *src, const struct sockaddr *dst,
   5082     struct secasindex * saidx)
   5083 {
   5084 	const union sockaddr_union *src_u = (const union sockaddr_union *)src;
   5085 	const union sockaddr_union *dst_u = (const union sockaddr_union *)dst;
   5086 
   5087 	/* sa len safety check */
   5088 	if (key_checksalen(src_u) != 0)
   5089 		return -1;
   5090 	if (key_checksalen(dst_u) != 0)
   5091 		return -1;
   5092 
   5093 	memset(saidx, 0, sizeof(*saidx));
   5094 	saidx->proto = proto;
   5095 	saidx->mode = mode;
   5096 	saidx->reqid = reqid;
   5097 	memcpy(&saidx->src, src_u, src_u->sa.sa_len);
   5098 	memcpy(&saidx->dst, dst_u, dst_u->sa.sa_len);
   5099 
   5100 	key_porttosaddr(&((saidx)->src), 0);
   5101 	key_porttosaddr(&((saidx)->dst), 0);
   5102 	return 0;
   5103 }
   5104 
   5105 static void
   5106 key_init_spidx_bymsghdr(struct secpolicyindex *spidx,
   5107     const struct sadb_msghdr *mhp)
   5108 {
   5109 	const struct sadb_address *src0, *dst0;
   5110 	const struct sockaddr *src, *dst;
   5111 	const struct sadb_x_policy *xpl0;
   5112 
   5113 	src0 = mhp->ext[SADB_EXT_ADDRESS_SRC];
   5114 	dst0 = mhp->ext[SADB_EXT_ADDRESS_DST];
   5115 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   5116 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   5117 	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
   5118 
   5119 	memset(spidx, 0, sizeof(*spidx));
   5120 	spidx->dir = xpl0->sadb_x_policy_dir;
   5121 	spidx->prefs = src0->sadb_address_prefixlen;
   5122 	spidx->prefd = dst0->sadb_address_prefixlen;
   5123 	spidx->ul_proto = src0->sadb_address_proto;
   5124 	/* XXX boundary check against sa_len */
   5125 	memcpy(&spidx->src, src, src->sa_len);
   5126 	memcpy(&spidx->dst, dst, dst->sa_len);
   5127 }
   5128 
   5129 /* %%% PF_KEY */
   5130 /*
   5131  * SADB_GETSPI processing is to receive
   5132  *	<base, (SA2), src address, dst address, (SPI range)>
   5133  * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
   5134  * tree with the status of LARVAL, and send
   5135  *	<base, SA(*), address(SD)>
   5136  * to the IKMPd.
   5137  *
   5138  * IN:	mhp: pointer to the pointer to each header.
   5139  * OUT:	NULL if fail.
   5140  *	other if success, return pointer to the message to send.
   5141  */
   5142 static int
   5143 key_api_getspi(struct socket *so, struct mbuf *m,
   5144 	   const struct sadb_msghdr *mhp)
   5145 {
   5146 	const struct sockaddr *src, *dst;
   5147 	struct secasindex saidx;
   5148 	struct secashead *sah;
   5149 	struct secasvar *newsav;
   5150 	u_int8_t proto;
   5151 	u_int32_t spi;
   5152 	u_int8_t mode;
   5153 	u_int16_t reqid;
   5154 	int error;
   5155 
   5156 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   5157 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
   5158 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5159 		return key_senderror(so, m, EINVAL);
   5160 	}
   5161 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   5162 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
   5163 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5164 		return key_senderror(so, m, EINVAL);
   5165 	}
   5166 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
   5167 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
   5168 		mode = sa2->sadb_x_sa2_mode;
   5169 		reqid = sa2->sadb_x_sa2_reqid;
   5170 	} else {
   5171 		mode = IPSEC_MODE_ANY;
   5172 		reqid = 0;
   5173 	}
   5174 
   5175 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   5176 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   5177 
   5178 	/* map satype to proto */
   5179 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
   5180 	if (proto == 0) {
   5181 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   5182 		return key_senderror(so, m, EINVAL);
   5183 	}
   5184 
   5185 
   5186 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
   5187 	if (error != 0)
   5188 		return key_senderror(so, m, EINVAL);
   5189 
   5190 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   5191 	if (error != 0)
   5192 		return key_senderror(so, m, EINVAL);
   5193 
   5194 	/* SPI allocation */
   5195 	spi = key_do_getnewspi(mhp->ext[SADB_EXT_SPIRANGE], &saidx);
   5196 	if (spi == 0)
   5197 		return key_senderror(so, m, EINVAL);
   5198 
   5199 	/* get a SA index */
   5200 	sah = key_getsah_ref(&saidx, CMP_REQID);
   5201 	if (sah == NULL) {
   5202 		/* create a new SA index */
   5203 		sah = key_newsah(&saidx);
   5204 		if (sah == NULL) {
   5205 			IPSECLOG(LOG_DEBUG, "No more memory.\n");
   5206 			return key_senderror(so, m, ENOBUFS);
   5207 		}
   5208 	}
   5209 
   5210 	/* get a new SA */
   5211 	/* XXX rewrite */
   5212 	newsav = KEY_NEWSAV(m, mhp, &error);
   5213 	if (newsav == NULL) {
   5214 		key_sah_unref(sah);
   5215 		/* XXX don't free new SA index allocated in above. */
   5216 		return key_senderror(so, m, error);
   5217 	}
   5218 
   5219 	/* set spi */
   5220 	newsav->spi = htonl(spi);
   5221 
   5222 	/* Add to sah#savlist */
   5223 	key_init_sav(newsav);
   5224 	newsav->sah = sah;
   5225 	newsav->state = SADB_SASTATE_LARVAL;
   5226 	mutex_enter(&key_sad.lock);
   5227 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_LARVAL, newsav);
   5228 	mutex_exit(&key_sad.lock);
   5229 	key_validate_savlist(sah, SADB_SASTATE_LARVAL);
   5230 
   5231 	key_sah_unref(sah);
   5232 
   5233 #ifndef IPSEC_NONBLOCK_ACQUIRE
   5234 	/* delete the entry in key_misc.acqlist */
   5235 	if (mhp->msg->sadb_msg_seq != 0) {
   5236 		struct secacq *acq;
   5237 		mutex_enter(&key_misc.lock);
   5238 		acq = key_getacqbyseq(mhp->msg->sadb_msg_seq);
   5239 		if (acq != NULL) {
   5240 			/* reset counter in order to deletion by timehandler. */
   5241 			acq->created = time_uptime;
   5242 			acq->count = 0;
   5243 		}
   5244 		mutex_exit(&key_misc.lock);
   5245 	}
   5246 #endif
   5247 
   5248     {
   5249 	struct mbuf *n, *nn;
   5250 	struct sadb_sa *m_sa;
   5251 	int off, len;
   5252 
   5253 	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
   5254 	    PFKEY_ALIGN8(sizeof(struct sadb_sa)) <= MCLBYTES);
   5255 
   5256 	/* create new sadb_msg to reply. */
   5257 	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
   5258 	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
   5259 
   5260 	n = key_alloc_mbuf_simple(len, M_WAITOK);
   5261 	n->m_len = len;
   5262 	n->m_next = NULL;
   5263 	off = 0;
   5264 
   5265 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
   5266 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
   5267 
   5268 	m_sa = (struct sadb_sa *)(mtod(n, char *) + off);
   5269 	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
   5270 	m_sa->sadb_sa_exttype = SADB_EXT_SA;
   5271 	m_sa->sadb_sa_spi = htonl(spi);
   5272 	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
   5273 
   5274 	KASSERTMSG(off == len, "length inconsistency");
   5275 
   5276 	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
   5277 	    SADB_EXT_ADDRESS_DST);
   5278 
   5279 	KASSERT(n->m_len >= sizeof(struct sadb_msg));
   5280 
   5281 	n->m_pkthdr.len = 0;
   5282 	for (nn = n; nn; nn = nn->m_next)
   5283 		n->m_pkthdr.len += nn->m_len;
   5284 
   5285 	key_fill_replymsg(n, newsav->seq);
   5286 	m_freem(m);
   5287 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
   5288     }
   5289 }
   5290 
   5291 /*
   5292  * allocating new SPI
   5293  * called by key_api_getspi().
   5294  * OUT:
   5295  *	0:	failure.
   5296  *	others: success.
   5297  */
   5298 static u_int32_t
   5299 key_do_getnewspi(const struct sadb_spirange *spirange,
   5300 		 const struct secasindex *saidx)
   5301 {
   5302 	u_int32_t newspi;
   5303 	u_int32_t spmin, spmax;
   5304 	int count = key_spi_trycnt;
   5305 
   5306 	/* set spi range to allocate */
   5307 	if (spirange != NULL) {
   5308 		spmin = spirange->sadb_spirange_min;
   5309 		spmax = spirange->sadb_spirange_max;
   5310 	} else {
   5311 		spmin = key_spi_minval;
   5312 		spmax = key_spi_maxval;
   5313 	}
   5314 	/* IPCOMP needs 2-byte SPI */
   5315 	if (saidx->proto == IPPROTO_IPCOMP) {
   5316 		u_int32_t t;
   5317 		if (spmin >= 0x10000)
   5318 			spmin = 0xffff;
   5319 		if (spmax >= 0x10000)
   5320 			spmax = 0xffff;
   5321 		if (spmin > spmax) {
   5322 			t = spmin; spmin = spmax; spmax = t;
   5323 		}
   5324 	}
   5325 
   5326 	if (spmin == spmax) {
   5327 		if (key_checkspidup(saidx, htonl(spmin))) {
   5328 			IPSECLOG(LOG_DEBUG, "SPI %u exists already.\n", spmin);
   5329 			return 0;
   5330 		}
   5331 
   5332 		count--; /* taking one cost. */
   5333 		newspi = spmin;
   5334 
   5335 	} else {
   5336 
   5337 		/* init SPI */
   5338 		newspi = 0;
   5339 
   5340 		/* when requesting to allocate spi ranged */
   5341 		while (count--) {
   5342 			/* generate pseudo-random SPI value ranged. */
   5343 			newspi = spmin + (key_random() % (spmax - spmin + 1));
   5344 
   5345 			if (!key_checkspidup(saidx, htonl(newspi)))
   5346 				break;
   5347 		}
   5348 
   5349 		if (count == 0 || newspi == 0) {
   5350 			IPSECLOG(LOG_DEBUG, "to allocate spi is failed.\n");
   5351 			return 0;
   5352 		}
   5353 	}
   5354 
   5355 	/* statistics */
   5356 	keystat.getspi_count =
   5357 	    (keystat.getspi_count + key_spi_trycnt - count) / 2;
   5358 
   5359 	return newspi;
   5360 }
   5361 
   5362 static int
   5363 key_handle_natt_info(struct secasvar *sav,
   5364       		     const struct sadb_msghdr *mhp)
   5365 {
   5366 	const char *msg = "?" ;
   5367 	struct sadb_x_nat_t_type *type;
   5368 	struct sadb_x_nat_t_port *sport, *dport;
   5369 	struct sadb_address *iaddr, *raddr;
   5370 	struct sadb_x_nat_t_frag *frag;
   5371 
   5372 	if (mhp->ext[SADB_X_EXT_NAT_T_TYPE] == NULL ||
   5373 	    mhp->ext[SADB_X_EXT_NAT_T_SPORT] == NULL ||
   5374 	    mhp->ext[SADB_X_EXT_NAT_T_DPORT] == NULL)
   5375 		return 0;
   5376 
   5377 	if (mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) {
   5378 		msg = "TYPE";
   5379 		goto bad;
   5380 	}
   5381 
   5382 	if (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) {
   5383 		msg = "SPORT";
   5384 		goto bad;
   5385 	}
   5386 
   5387 	if (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport)) {
   5388 		msg = "DPORT";
   5389 		goto bad;
   5390 	}
   5391 
   5392 	if (mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL) {
   5393 		IPSECLOG(LOG_DEBUG, "NAT-T OAi present\n");
   5394 		if (mhp->extlen[SADB_X_EXT_NAT_T_OAI] < sizeof(*iaddr)) {
   5395 			msg = "OAI";
   5396 			goto bad;
   5397 		}
   5398 	}
   5399 
   5400 	if (mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL) {
   5401 		IPSECLOG(LOG_DEBUG, "NAT-T OAr present\n");
   5402 		if (mhp->extlen[SADB_X_EXT_NAT_T_OAR] < sizeof(*raddr)) {
   5403 			msg = "OAR";
   5404 			goto bad;
   5405 		}
   5406 	}
   5407 
   5408 	if (mhp->ext[SADB_X_EXT_NAT_T_FRAG] != NULL) {
   5409 	    if (mhp->extlen[SADB_X_EXT_NAT_T_FRAG] < sizeof(*frag)) {
   5410 		    msg = "FRAG";
   5411 		    goto bad;
   5412 	    }
   5413 	}
   5414 
   5415 	type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
   5416 	sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
   5417 	dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
   5418 	iaddr = mhp->ext[SADB_X_EXT_NAT_T_OAI];
   5419 	raddr = mhp->ext[SADB_X_EXT_NAT_T_OAR];
   5420 	frag = mhp->ext[SADB_X_EXT_NAT_T_FRAG];
   5421 
   5422 	IPSECLOG(LOG_DEBUG, "type %d, sport = %d, dport = %d\n",
   5423 	    type->sadb_x_nat_t_type_type,
   5424 	    ntohs(sport->sadb_x_nat_t_port_port),
   5425 	    ntohs(dport->sadb_x_nat_t_port_port));
   5426 
   5427 	sav->natt_type = type->sadb_x_nat_t_type_type;
   5428 	key_porttosaddr(&sav->sah->saidx.src, sport->sadb_x_nat_t_port_port);
   5429 	key_porttosaddr(&sav->sah->saidx.dst, dport->sadb_x_nat_t_port_port);
   5430 	if (frag)
   5431 		sav->esp_frag = frag->sadb_x_nat_t_frag_fraglen;
   5432 	else
   5433 		sav->esp_frag = IP_MAXPACKET;
   5434 
   5435 	return 0;
   5436 bad:
   5437 	IPSECLOG(LOG_DEBUG, "invalid message %s\n", msg);
   5438 	__USE(msg);
   5439 	return -1;
   5440 }
   5441 
   5442 /* Just update the IPSEC_NAT_T ports if present */
   5443 static int
   5444 key_set_natt_ports(union sockaddr_union *src, union sockaddr_union *dst,
   5445       		     const struct sadb_msghdr *mhp)
   5446 {
   5447 	if (mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL)
   5448 		IPSECLOG(LOG_DEBUG, "NAT-T OAi present\n");
   5449 	if (mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL)
   5450 		IPSECLOG(LOG_DEBUG, "NAT-T OAr present\n");
   5451 
   5452 	if ((mhp->ext[SADB_X_EXT_NAT_T_TYPE] != NULL) &&
   5453 	    (mhp->ext[SADB_X_EXT_NAT_T_SPORT] != NULL) &&
   5454 	    (mhp->ext[SADB_X_EXT_NAT_T_DPORT] != NULL)) {
   5455 		struct sadb_x_nat_t_type *type;
   5456 		struct sadb_x_nat_t_port *sport;
   5457 		struct sadb_x_nat_t_port *dport;
   5458 
   5459 		if ((mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) ||
   5460 		    (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) ||
   5461 		    (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport))) {
   5462 			IPSECLOG(LOG_DEBUG, "invalid message\n");
   5463 			return -1;
   5464 		}
   5465 
   5466 		type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
   5467 		sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
   5468 		dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
   5469 
   5470 		key_porttosaddr(src, sport->sadb_x_nat_t_port_port);
   5471 		key_porttosaddr(dst, dport->sadb_x_nat_t_port_port);
   5472 
   5473 		IPSECLOG(LOG_DEBUG, "type %d, sport = %d, dport = %d\n",
   5474 		    type->sadb_x_nat_t_type_type,
   5475 		    ntohs(sport->sadb_x_nat_t_port_port),
   5476 		    ntohs(dport->sadb_x_nat_t_port_port));
   5477 	}
   5478 
   5479 	return 0;
   5480 }
   5481 
   5482 
   5483 /*
   5484  * SADB_UPDATE processing
   5485  * receive
   5486  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
   5487  *       key(AE), (identity(SD),) (sensitivity)>
   5488  * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
   5489  * and send
   5490  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
   5491  *       (identity(SD),) (sensitivity)>
   5492  * to the ikmpd.
   5493  *
   5494  * m will always be freed.
   5495  */
   5496 static int
   5497 key_api_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
   5498 {
   5499 	struct sadb_sa *sa0;
   5500 	const struct sockaddr *src, *dst;
   5501 	struct secasindex saidx;
   5502 	struct secashead *sah;
   5503 	struct secasvar *sav, *newsav;
   5504 	u_int16_t proto;
   5505 	u_int8_t mode;
   5506 	u_int16_t reqid;
   5507 	int error;
   5508 
   5509 	/* map satype to proto */
   5510 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
   5511 	if (proto == 0) {
   5512 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   5513 		return key_senderror(so, m, EINVAL);
   5514 	}
   5515 
   5516 	if (mhp->ext[SADB_EXT_SA] == NULL ||
   5517 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   5518 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
   5519 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
   5520 	     mhp->ext[SADB_EXT_KEY_ENCRYPT] == NULL) ||
   5521 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
   5522 	     mhp->ext[SADB_EXT_KEY_AUTH] == NULL) ||
   5523 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL &&
   5524 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] == NULL) ||
   5525 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] == NULL &&
   5526 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] != NULL)) {
   5527 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5528 		return key_senderror(so, m, EINVAL);
   5529 	}
   5530 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
   5531 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   5532 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
   5533 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5534 		return key_senderror(so, m, EINVAL);
   5535 	}
   5536 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
   5537 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
   5538 		mode = sa2->sadb_x_sa2_mode;
   5539 		reqid = sa2->sadb_x_sa2_reqid;
   5540 	} else {
   5541 		mode = IPSEC_MODE_ANY;
   5542 		reqid = 0;
   5543 	}
   5544 	/* XXX boundary checking for other extensions */
   5545 
   5546 	sa0 = mhp->ext[SADB_EXT_SA];
   5547 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   5548 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   5549 
   5550 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
   5551 	if (error != 0)
   5552 		return key_senderror(so, m, EINVAL);
   5553 
   5554 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   5555 	if (error != 0)
   5556 		return key_senderror(so, m, EINVAL);
   5557 
   5558 	/* get a SA header */
   5559 	sah = key_getsah_ref(&saidx, CMP_REQID);
   5560 	if (sah == NULL) {
   5561 		IPSECLOG(LOG_DEBUG, "no SA index found.\n");
   5562 		return key_senderror(so, m, ENOENT);
   5563 	}
   5564 
   5565 	/* set spidx if there */
   5566 	/* XXX rewrite */
   5567 	error = key_setident(sah, m, mhp);
   5568 	if (error)
   5569 		goto error_sah;
   5570 
   5571 	/* find a SA with sequence number. */
   5572 #ifdef IPSEC_DOSEQCHECK
   5573 	if (mhp->msg->sadb_msg_seq != 0) {
   5574 		sav = key_getsavbyseq(sah, mhp->msg->sadb_msg_seq);
   5575 		if (sav == NULL) {
   5576 			IPSECLOG(LOG_DEBUG,
   5577 			    "no larval SA with sequence %u exists.\n",
   5578 			    mhp->msg->sadb_msg_seq);
   5579 			error = ENOENT;
   5580 			goto error_sah;
   5581 		}
   5582 	}
   5583 #else
   5584 	sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
   5585 	if (sav == NULL) {
   5586 		IPSECLOG(LOG_DEBUG, "no such a SA found (spi:%u)\n",
   5587 		    (u_int32_t)ntohl(sa0->sadb_sa_spi));
   5588 		error = EINVAL;
   5589 		goto error_sah;
   5590 	}
   5591 #endif
   5592 
   5593 	/* validity check */
   5594 	if (sav->sah->saidx.proto != proto) {
   5595 		IPSECLOG(LOG_DEBUG, "protocol mismatched (DB=%u param=%u)\n",
   5596 		    sav->sah->saidx.proto, proto);
   5597 		error = EINVAL;
   5598 		goto error;
   5599 	}
   5600 #ifdef IPSEC_DOSEQCHECK
   5601 	if (sav->spi != sa0->sadb_sa_spi) {
   5602 		IPSECLOG(LOG_DEBUG, "SPI mismatched (DB:%u param:%u)\n",
   5603 		    (u_int32_t)ntohl(sav->spi),
   5604 		    (u_int32_t)ntohl(sa0->sadb_sa_spi));
   5605 		error = EINVAL;
   5606 		goto error;
   5607 	}
   5608 #endif
   5609 	if (sav->pid != mhp->msg->sadb_msg_pid) {
   5610 		IPSECLOG(LOG_DEBUG, "pid mismatched (DB:%u param:%u)\n",
   5611 		    sav->pid, mhp->msg->sadb_msg_pid);
   5612 		error = EINVAL;
   5613 		goto error;
   5614 	}
   5615 
   5616 	/*
   5617 	 * Allocate a new SA instead of modifying the existing SA directly
   5618 	 * to avoid race conditions.
   5619 	 */
   5620 	newsav = kmem_zalloc(sizeof(struct secasvar), KM_SLEEP);
   5621 
   5622 	/* copy sav values */
   5623 	newsav->spi = sav->spi;
   5624 	newsav->seq = sav->seq;
   5625 	newsav->created = sav->created;
   5626 	newsav->pid = sav->pid;
   5627 	newsav->sah = sav->sah;
   5628 
   5629 	error = key_setsaval(newsav, m, mhp);
   5630 	if (error) {
   5631 		key_delsav(newsav);
   5632 		goto error;
   5633 	}
   5634 
   5635 	error = key_handle_natt_info(newsav, mhp);
   5636 	if (error != 0) {
   5637 		key_delsav(newsav);
   5638 		goto error;
   5639 	}
   5640 
   5641 	error = key_init_xform(newsav);
   5642 	if (error != 0) {
   5643 		key_delsav(newsav);
   5644 		goto error;
   5645 	}
   5646 
   5647 	/* Add to sah#savlist */
   5648 	key_init_sav(newsav);
   5649 	newsav->state = SADB_SASTATE_MATURE;
   5650 	mutex_enter(&key_sad.lock);
   5651 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_MATURE, newsav);
   5652 	mutex_exit(&key_sad.lock);
   5653 	key_validate_savlist(sah, SADB_SASTATE_MATURE);
   5654 
   5655 	key_sah_unref(sah);
   5656 	sah = NULL;
   5657 
   5658 	key_destroy_sav_with_ref(sav);
   5659 	sav = NULL;
   5660 
   5661     {
   5662 	struct mbuf *n;
   5663 
   5664 	/* set msg buf from mhp */
   5665 	n = key_getmsgbuf_x1(m, mhp);
   5666 	if (n == NULL) {
   5667 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
   5668 		return key_senderror(so, m, ENOBUFS);
   5669 	}
   5670 
   5671 	m_freem(m);
   5672 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   5673     }
   5674 error:
   5675 	KEY_SA_UNREF(&sav);
   5676 error_sah:
   5677 	key_sah_unref(sah);
   5678 	return key_senderror(so, m, error);
   5679 }
   5680 
   5681 /*
   5682  * search SAD with sequence for a SA which state is SADB_SASTATE_LARVAL.
   5683  * only called by key_api_update().
   5684  * OUT:
   5685  *	NULL	: not found
   5686  *	others	: found, pointer to a SA.
   5687  */
   5688 #ifdef IPSEC_DOSEQCHECK
   5689 static struct secasvar *
   5690 key_getsavbyseq(struct secashead *sah, u_int32_t seq)
   5691 {
   5692 	struct secasvar *sav;
   5693 	u_int state;
   5694 	int s;
   5695 
   5696 	state = SADB_SASTATE_LARVAL;
   5697 
   5698 	/* search SAD with sequence number ? */
   5699 	s = pserialize_read_enter();
   5700 	SAVLIST_READER_FOREACH(sav, sah, state) {
   5701 		KEY_CHKSASTATE(state, sav->state);
   5702 
   5703 		if (sav->seq == seq) {
   5704 			SA_ADDREF(sav);
   5705 			KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
   5706 			    "DP cause refcnt++:%d SA:%p\n",
   5707 			    key_sa_refcnt(sav), sav);
   5708 			break;
   5709 		}
   5710 	}
   5711 	pserialize_read_exit(s);
   5712 
   5713 	return sav;
   5714 }
   5715 #endif
   5716 
   5717 /*
   5718  * SADB_ADD processing
   5719  * add an entry to SA database, when received
   5720  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
   5721  *       key(AE), (identity(SD),) (sensitivity)>
   5722  * from the ikmpd,
   5723  * and send
   5724  *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
   5725  *       (identity(SD),) (sensitivity)>
   5726  * to the ikmpd.
   5727  *
   5728  * IGNORE identity and sensitivity messages.
   5729  *
   5730  * m will always be freed.
   5731  */
   5732 static int
   5733 key_api_add(struct socket *so, struct mbuf *m,
   5734 	const struct sadb_msghdr *mhp)
   5735 {
   5736 	struct sadb_sa *sa0;
   5737 	const struct sockaddr *src, *dst;
   5738 	struct secasindex saidx;
   5739 	struct secashead *sah;
   5740 	struct secasvar *newsav;
   5741 	u_int16_t proto;
   5742 	u_int8_t mode;
   5743 	u_int16_t reqid;
   5744 	int error;
   5745 
   5746 	/* map satype to proto */
   5747 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
   5748 	if (proto == 0) {
   5749 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   5750 		return key_senderror(so, m, EINVAL);
   5751 	}
   5752 
   5753 	if (mhp->ext[SADB_EXT_SA] == NULL ||
   5754 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   5755 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
   5756 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
   5757 	     mhp->ext[SADB_EXT_KEY_ENCRYPT] == NULL) ||
   5758 	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
   5759 	     mhp->ext[SADB_EXT_KEY_AUTH] == NULL) ||
   5760 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL &&
   5761 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] == NULL) ||
   5762 	    (mhp->ext[SADB_EXT_LIFETIME_HARD] == NULL &&
   5763 	     mhp->ext[SADB_EXT_LIFETIME_SOFT] != NULL)) {
   5764 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5765 		return key_senderror(so, m, EINVAL);
   5766 	}
   5767 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
   5768 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   5769 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
   5770 		/* XXX need more */
   5771 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   5772 		return key_senderror(so, m, EINVAL);
   5773 	}
   5774 	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
   5775 		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
   5776 		mode = sa2->sadb_x_sa2_mode;
   5777 		reqid = sa2->sadb_x_sa2_reqid;
   5778 	} else {
   5779 		mode = IPSEC_MODE_ANY;
   5780 		reqid = 0;
   5781 	}
   5782 
   5783 	sa0 = mhp->ext[SADB_EXT_SA];
   5784 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   5785 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   5786 
   5787 	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
   5788 	if (error != 0)
   5789 		return key_senderror(so, m, EINVAL);
   5790 
   5791 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   5792 	if (error != 0)
   5793 		return key_senderror(so, m, EINVAL);
   5794 
   5795 	/* get a SA header */
   5796 	sah = key_getsah_ref(&saidx, CMP_REQID);
   5797 	if (sah == NULL) {
   5798 		/* create a new SA header */
   5799 		sah = key_newsah(&saidx);
   5800 		if (sah == NULL) {
   5801 			IPSECLOG(LOG_DEBUG, "No more memory.\n");
   5802 			return key_senderror(so, m, ENOBUFS);
   5803 		}
   5804 	}
   5805 
   5806 	/* set spidx if there */
   5807 	/* XXX rewrite */
   5808 	error = key_setident(sah, m, mhp);
   5809 	if (error)
   5810 		goto error;
   5811 
   5812     {
   5813 	struct secasvar *sav;
   5814 
   5815 	/* We can create new SA only if SPI is differenct. */
   5816 	sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
   5817 	if (sav != NULL) {
   5818 		KEY_SA_UNREF(&sav);
   5819 		IPSECLOG(LOG_DEBUG, "SA already exists.\n");
   5820 		error = EEXIST;
   5821 		goto error;
   5822 	}
   5823     }
   5824 
   5825 	/* create new SA entry. */
   5826 	newsav = KEY_NEWSAV(m, mhp, &error);
   5827 	if (newsav == NULL)
   5828 		goto error;
   5829 	newsav->sah = sah;
   5830 
   5831 	error = key_handle_natt_info(newsav, mhp);
   5832 	if (error != 0) {
   5833 		key_delsav(newsav);
   5834 		error = EINVAL;
   5835 		goto error;
   5836 	}
   5837 
   5838 	error = key_init_xform(newsav);
   5839 	if (error != 0) {
   5840 		key_delsav(newsav);
   5841 		goto error;
   5842 	}
   5843 
   5844 	/* Add to sah#savlist */
   5845 	key_init_sav(newsav);
   5846 	newsav->state = SADB_SASTATE_MATURE;
   5847 	mutex_enter(&key_sad.lock);
   5848 	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_MATURE, newsav);
   5849 	mutex_exit(&key_sad.lock);
   5850 	key_validate_savlist(sah, SADB_SASTATE_MATURE);
   5851 
   5852 	key_sah_unref(sah);
   5853 	sah = NULL;
   5854 
   5855 	/*
   5856 	 * don't call key_freesav() here, as we would like to keep the SA
   5857 	 * in the database on success.
   5858 	 */
   5859 
   5860     {
   5861 	struct mbuf *n;
   5862 
   5863 	/* set msg buf from mhp */
   5864 	n = key_getmsgbuf_x1(m, mhp);
   5865 	if (n == NULL) {
   5866 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
   5867 		return key_senderror(so, m, ENOBUFS);
   5868 	}
   5869 
   5870 	m_freem(m);
   5871 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   5872     }
   5873 error:
   5874 	key_sah_unref(sah);
   5875 	return key_senderror(so, m, error);
   5876 }
   5877 
   5878 /* m is retained */
   5879 static int
   5880 key_setident(struct secashead *sah, struct mbuf *m,
   5881 	     const struct sadb_msghdr *mhp)
   5882 {
   5883 	const struct sadb_ident *idsrc, *iddst;
   5884 	int idsrclen, iddstlen;
   5885 
   5886 	KASSERT(!cpu_softintr_p());
   5887 	KASSERT(sah != NULL);
   5888 	KASSERT(m != NULL);
   5889 	KASSERT(mhp != NULL);
   5890 	KASSERT(mhp->msg != NULL);
   5891 
   5892 	/*
   5893 	 * Can be called with an existing sah from key_api_update().
   5894 	 */
   5895 	if (sah->idents != NULL) {
   5896 		kmem_free(sah->idents, sah->idents_len);
   5897 		sah->idents = NULL;
   5898 		sah->idents_len = 0;
   5899 	}
   5900 	if (sah->identd != NULL) {
   5901 		kmem_free(sah->identd, sah->identd_len);
   5902 		sah->identd = NULL;
   5903 		sah->identd_len = 0;
   5904 	}
   5905 
   5906 	/* don't make buffer if not there */
   5907 	if (mhp->ext[SADB_EXT_IDENTITY_SRC] == NULL &&
   5908 	    mhp->ext[SADB_EXT_IDENTITY_DST] == NULL) {
   5909 		sah->idents = NULL;
   5910 		sah->identd = NULL;
   5911 		return 0;
   5912 	}
   5913 
   5914 	if (mhp->ext[SADB_EXT_IDENTITY_SRC] == NULL ||
   5915 	    mhp->ext[SADB_EXT_IDENTITY_DST] == NULL) {
   5916 		IPSECLOG(LOG_DEBUG, "invalid identity.\n");
   5917 		return EINVAL;
   5918 	}
   5919 
   5920 	idsrc = mhp->ext[SADB_EXT_IDENTITY_SRC];
   5921 	iddst = mhp->ext[SADB_EXT_IDENTITY_DST];
   5922 	idsrclen = mhp->extlen[SADB_EXT_IDENTITY_SRC];
   5923 	iddstlen = mhp->extlen[SADB_EXT_IDENTITY_DST];
   5924 
   5925 	/* validity check */
   5926 	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
   5927 		IPSECLOG(LOG_DEBUG, "ident type mismatch.\n");
   5928 		return EINVAL;
   5929 	}
   5930 
   5931 	switch (idsrc->sadb_ident_type) {
   5932 	case SADB_IDENTTYPE_PREFIX:
   5933 	case SADB_IDENTTYPE_FQDN:
   5934 	case SADB_IDENTTYPE_USERFQDN:
   5935 	default:
   5936 		/* XXX do nothing */
   5937 		sah->idents = NULL;
   5938 		sah->identd = NULL;
   5939 	 	return 0;
   5940 	}
   5941 
   5942 	/* make structure */
   5943 	sah->idents = kmem_alloc(idsrclen, KM_SLEEP);
   5944 	sah->idents_len = idsrclen;
   5945 	sah->identd = kmem_alloc(iddstlen, KM_SLEEP);
   5946 	sah->identd_len = iddstlen;
   5947 	memcpy(sah->idents, idsrc, idsrclen);
   5948 	memcpy(sah->identd, iddst, iddstlen);
   5949 
   5950 	return 0;
   5951 }
   5952 
   5953 /*
   5954  * m will not be freed on return. It never return NULL.
   5955  * it is caller's responsibility to free the result.
   5956  */
   5957 static struct mbuf *
   5958 key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
   5959 {
   5960 	struct mbuf *n;
   5961 
   5962 	KASSERT(m != NULL);
   5963 	KASSERT(mhp != NULL);
   5964 	KASSERT(mhp->msg != NULL);
   5965 
   5966 	/* create new sadb_msg to reply. */
   5967 	n = key_gather_mbuf(m, mhp, 1, 15, SADB_EXT_RESERVED,
   5968 	    SADB_EXT_SA, SADB_X_EXT_SA2,
   5969 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
   5970 	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
   5971 	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
   5972 	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
   5973 	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
   5974 	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NAT_T_FRAG);
   5975 
   5976 	KASSERT(n->m_len >= sizeof(struct sadb_msg));
   5977 
   5978 	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
   5979 	mtod(n, struct sadb_msg *)->sadb_msg_len =
   5980 	    PFKEY_UNIT64(n->m_pkthdr.len);
   5981 
   5982 	return n;
   5983 }
   5984 
   5985 static int key_delete_all (struct socket *, struct mbuf *,
   5986 			   const struct sadb_msghdr *, u_int16_t);
   5987 
   5988 /*
   5989  * SADB_DELETE processing
   5990  * receive
   5991  *   <base, SA(*), address(SD)>
   5992  * from the ikmpd, and set SADB_SASTATE_DEAD,
   5993  * and send,
   5994  *   <base, SA(*), address(SD)>
   5995  * to the ikmpd.
   5996  *
   5997  * m will always be freed.
   5998  */
   5999 static int
   6000 key_api_delete(struct socket *so, struct mbuf *m,
   6001 	   const struct sadb_msghdr *mhp)
   6002 {
   6003 	struct sadb_sa *sa0;
   6004 	const struct sockaddr *src, *dst;
   6005 	struct secasindex saidx;
   6006 	struct secashead *sah;
   6007 	struct secasvar *sav = NULL;
   6008 	u_int16_t proto;
   6009 	int error;
   6010 
   6011 	/* map satype to proto */
   6012 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
   6013 	if (proto == 0) {
   6014 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   6015 		return key_senderror(so, m, EINVAL);
   6016 	}
   6017 
   6018 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   6019 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
   6020 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   6021 		return key_senderror(so, m, EINVAL);
   6022 	}
   6023 
   6024 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   6025 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
   6026 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   6027 		return key_senderror(so, m, EINVAL);
   6028 	}
   6029 
   6030 	if (mhp->ext[SADB_EXT_SA] == NULL) {
   6031 		/*
   6032 		 * Caller wants us to delete all non-LARVAL SAs
   6033 		 * that match the src/dst.  This is used during
   6034 		 * IKE INITIAL-CONTACT.
   6035 		 */
   6036 		IPSECLOG(LOG_DEBUG, "doing delete all.\n");
   6037 		return key_delete_all(so, m, mhp, proto);
   6038 	} else if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa)) {
   6039 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   6040 		return key_senderror(so, m, EINVAL);
   6041 	}
   6042 
   6043 	sa0 = mhp->ext[SADB_EXT_SA];
   6044 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   6045 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   6046 
   6047 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
   6048 	if (error != 0)
   6049 		return key_senderror(so, m, EINVAL);
   6050 
   6051 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   6052 	if (error != 0)
   6053 		return key_senderror(so, m, EINVAL);
   6054 
   6055 	/* get a SA header */
   6056 	sah = key_getsah_ref(&saidx, CMP_HEAD);
   6057 	if (sah != NULL) {
   6058 		/* get a SA with SPI. */
   6059 		sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
   6060 		key_sah_unref(sah);
   6061 	}
   6062 
   6063 	if (sav == NULL) {
   6064 		IPSECLOG(LOG_DEBUG, "no SA found.\n");
   6065 		return key_senderror(so, m, ENOENT);
   6066 	}
   6067 
   6068 	key_destroy_sav_with_ref(sav);
   6069 	sav = NULL;
   6070 
   6071     {
   6072 	struct mbuf *n;
   6073 
   6074 	/* create new sadb_msg to reply. */
   6075 	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
   6076 	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
   6077 
   6078 	key_fill_replymsg(n, 0);
   6079 	m_freem(m);
   6080 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   6081     }
   6082 }
   6083 
   6084 /*
   6085  * delete all SAs for src/dst.  Called from key_api_delete().
   6086  */
   6087 static int
   6088 key_delete_all(struct socket *so, struct mbuf *m,
   6089 	       const struct sadb_msghdr *mhp, u_int16_t proto)
   6090 {
   6091 	const struct sockaddr *src, *dst;
   6092 	struct secasindex saidx;
   6093 	struct secashead *sah;
   6094 	struct secasvar *sav;
   6095 	u_int state;
   6096 	int error;
   6097 
   6098 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   6099 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   6100 
   6101 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
   6102 	if (error != 0)
   6103 		return key_senderror(so, m, EINVAL);
   6104 
   6105 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   6106 	if (error != 0)
   6107 		return key_senderror(so, m, EINVAL);
   6108 
   6109 	sah = key_getsah_ref(&saidx, CMP_HEAD);
   6110 	if (sah != NULL) {
   6111 		/* Delete all non-LARVAL SAs. */
   6112 		SASTATE_ALIVE_FOREACH(state) {
   6113 			if (state == SADB_SASTATE_LARVAL)
   6114 				continue;
   6115 		restart:
   6116 			mutex_enter(&key_sad.lock);
   6117 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
   6118 				sav->state = SADB_SASTATE_DEAD;
   6119 				key_unlink_sav(sav);
   6120 				mutex_exit(&key_sad.lock);
   6121 				key_destroy_sav(sav);
   6122 				goto restart;
   6123 			}
   6124 			mutex_exit(&key_sad.lock);
   6125 		}
   6126 		key_sah_unref(sah);
   6127 	}
   6128     {
   6129 	struct mbuf *n;
   6130 
   6131 	/* create new sadb_msg to reply. */
   6132 	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
   6133 	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
   6134 
   6135 	key_fill_replymsg(n, 0);
   6136 	m_freem(m);
   6137 	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
   6138     }
   6139 }
   6140 
   6141 /*
   6142  * SADB_GET processing
   6143  * receive
   6144  *   <base, SA(*), address(SD)>
   6145  * from the ikmpd, and get a SP and a SA to respond,
   6146  * and send,
   6147  *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
   6148  *       (identity(SD),) (sensitivity)>
   6149  * to the ikmpd.
   6150  *
   6151  * m will always be freed.
   6152  */
   6153 static int
   6154 key_api_get(struct socket *so, struct mbuf *m,
   6155 	const struct sadb_msghdr *mhp)
   6156 {
   6157 	struct sadb_sa *sa0;
   6158 	const struct sockaddr *src, *dst;
   6159 	struct secasindex saidx;
   6160 	struct secasvar *sav = NULL;
   6161 	u_int16_t proto;
   6162 	int error;
   6163 
   6164 	/* map satype to proto */
   6165 	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
   6166 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   6167 		return key_senderror(so, m, EINVAL);
   6168 	}
   6169 
   6170 	if (mhp->ext[SADB_EXT_SA] == NULL ||
   6171 	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   6172 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
   6173 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   6174 		return key_senderror(so, m, EINVAL);
   6175 	}
   6176 	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
   6177 	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   6178 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
   6179 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   6180 		return key_senderror(so, m, EINVAL);
   6181 	}
   6182 
   6183 	sa0 = mhp->ext[SADB_EXT_SA];
   6184 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   6185 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   6186 
   6187 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
   6188 	if (error != 0)
   6189 		return key_senderror(so, m, EINVAL);
   6190 
   6191 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   6192 	if (error != 0)
   6193 		return key_senderror(so, m, EINVAL);
   6194 
   6195 	/* get a SA header */
   6196     {
   6197 	struct secashead *sah;
   6198 	int s = pserialize_read_enter();
   6199 
   6200 	sah = key_getsah(&saidx, CMP_HEAD);
   6201 	if (sah != NULL) {
   6202 		/* get a SA with SPI. */
   6203 		sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
   6204 	}
   6205 	pserialize_read_exit(s);
   6206     }
   6207 	if (sav == NULL) {
   6208 		IPSECLOG(LOG_DEBUG, "no SA found.\n");
   6209 		return key_senderror(so, m, ENOENT);
   6210 	}
   6211 
   6212     {
   6213 	struct mbuf *n;
   6214 	u_int8_t satype;
   6215 
   6216 	/* map proto to satype */
   6217 	satype = key_proto2satype(sav->sah->saidx.proto);
   6218 	if (satype == 0) {
   6219 		KEY_SA_UNREF(&sav);
   6220 		IPSECLOG(LOG_DEBUG, "there was invalid proto in SAD.\n");
   6221 		return key_senderror(so, m, EINVAL);
   6222 	}
   6223 
   6224 	/* create new sadb_msg to reply. */
   6225 	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
   6226 	    mhp->msg->sadb_msg_pid);
   6227 	KEY_SA_UNREF(&sav);
   6228 	m_freem(m);
   6229 	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
   6230     }
   6231 }
   6232 
   6233 /* XXX make it sysctl-configurable? */
   6234 static void
   6235 key_getcomb_setlifetime(struct sadb_comb *comb)
   6236 {
   6237 
   6238 	comb->sadb_comb_soft_allocations = 1;
   6239 	comb->sadb_comb_hard_allocations = 1;
   6240 	comb->sadb_comb_soft_bytes = 0;
   6241 	comb->sadb_comb_hard_bytes = 0;
   6242 	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
   6243 	comb->sadb_comb_soft_addtime = comb->sadb_comb_hard_addtime * 80 / 100;
   6244 	comb->sadb_comb_hard_usetime = 28800;	/* 8 hours */
   6245 	comb->sadb_comb_soft_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
   6246 }
   6247 
   6248 /*
   6249  * XXX reorder combinations by preference
   6250  * XXX no idea if the user wants ESP authentication or not
   6251  */
   6252 static struct mbuf *
   6253 key_getcomb_esp(int mflag)
   6254 {
   6255 	struct sadb_comb *comb;
   6256 	const struct enc_xform *algo;
   6257 	struct mbuf *result = NULL, *m, *n;
   6258 	int encmin;
   6259 	int i, off, o;
   6260 	int totlen;
   6261 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
   6262 
   6263 	m = NULL;
   6264 	for (i = 1; i <= SADB_EALG_MAX; i++) {
   6265 		algo = esp_algorithm_lookup(i);
   6266 		if (algo == NULL)
   6267 			continue;
   6268 
   6269 		/* discard algorithms with key size smaller than system min */
   6270 		if (_BITS(algo->maxkey) < ipsec_esp_keymin)
   6271 			continue;
   6272 		if (_BITS(algo->minkey) < ipsec_esp_keymin)
   6273 			encmin = ipsec_esp_keymin;
   6274 		else
   6275 			encmin = _BITS(algo->minkey);
   6276 
   6277 		if (ipsec_esp_auth)
   6278 			m = key_getcomb_ah(mflag);
   6279 		else {
   6280 			KASSERTMSG(l <= MLEN,
   6281 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
   6282 			MGET(m, mflag, MT_DATA);
   6283 			if (m) {
   6284 				M_ALIGN(m, l);
   6285 				m->m_len = l;
   6286 				m->m_next = NULL;
   6287 				memset(mtod(m, void *), 0, m->m_len);
   6288 			}
   6289 		}
   6290 		if (!m)
   6291 			goto fail;
   6292 
   6293 		totlen = 0;
   6294 		for (n = m; n; n = n->m_next)
   6295 			totlen += n->m_len;
   6296 		KASSERTMSG((totlen % l) == 0, "totlen=%u, l=%u", totlen, l);
   6297 
   6298 		for (off = 0; off < totlen; off += l) {
   6299 			n = m_pulldown(m, off, l, &o);
   6300 			if (!n) {
   6301 				/* m is already freed */
   6302 				goto fail;
   6303 			}
   6304 			comb = (struct sadb_comb *)(mtod(n, char *) + o);
   6305 			memset(comb, 0, sizeof(*comb));
   6306 			key_getcomb_setlifetime(comb);
   6307 			comb->sadb_comb_encrypt = i;
   6308 			comb->sadb_comb_encrypt_minbits = encmin;
   6309 			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
   6310 		}
   6311 
   6312 		if (!result)
   6313 			result = m;
   6314 		else
   6315 			m_cat(result, m);
   6316 	}
   6317 
   6318 	return result;
   6319 
   6320  fail:
   6321 	if (result)
   6322 		m_freem(result);
   6323 	return NULL;
   6324 }
   6325 
   6326 static void
   6327 key_getsizes_ah(const struct auth_hash *ah, int alg,
   6328 	        u_int16_t* ksmin, u_int16_t* ksmax)
   6329 {
   6330 	*ksmin = *ksmax = ah->keysize;
   6331 	if (ah->keysize == 0) {
   6332 		/*
   6333 		 * Transform takes arbitrary key size but algorithm
   6334 		 * key size is restricted.  Enforce this here.
   6335 		 */
   6336 		switch (alg) {
   6337 		case SADB_X_AALG_MD5:	*ksmin = *ksmax = 16; break;
   6338 		case SADB_X_AALG_SHA:	*ksmin = *ksmax = 20; break;
   6339 		case SADB_X_AALG_NULL:	*ksmin = 0; *ksmax = 256; break;
   6340 		default:
   6341 			IPSECLOG(LOG_DEBUG, "unknown AH algorithm %u\n", alg);
   6342 			break;
   6343 		}
   6344 	}
   6345 }
   6346 
   6347 /*
   6348  * XXX reorder combinations by preference
   6349  */
   6350 static struct mbuf *
   6351 key_getcomb_ah(int mflag)
   6352 {
   6353 	struct sadb_comb *comb;
   6354 	const struct auth_hash *algo;
   6355 	struct mbuf *m;
   6356 	u_int16_t minkeysize, maxkeysize;
   6357 	int i;
   6358 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
   6359 
   6360 	m = NULL;
   6361 	for (i = 1; i <= SADB_AALG_MAX; i++) {
   6362 #if 1
   6363 		/* we prefer HMAC algorithms, not old algorithms */
   6364 		if (i != SADB_AALG_SHA1HMAC &&
   6365 		    i != SADB_AALG_MD5HMAC &&
   6366 		    i != SADB_X_AALG_SHA2_256 &&
   6367 		    i != SADB_X_AALG_SHA2_384 &&
   6368 		    i != SADB_X_AALG_SHA2_512)
   6369 			continue;
   6370 #endif
   6371 		algo = ah_algorithm_lookup(i);
   6372 		if (!algo)
   6373 			continue;
   6374 		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
   6375 		/* discard algorithms with key size smaller than system min */
   6376 		if (_BITS(minkeysize) < ipsec_ah_keymin)
   6377 			continue;
   6378 
   6379 		if (!m) {
   6380 			KASSERTMSG(l <= MLEN,
   6381 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
   6382 			MGET(m, mflag, MT_DATA);
   6383 			if (m) {
   6384 				M_ALIGN(m, l);
   6385 				m->m_len = l;
   6386 				m->m_next = NULL;
   6387 			}
   6388 		} else
   6389 			M_PREPEND(m, l, mflag);
   6390 		if (!m)
   6391 			return NULL;
   6392 
   6393 		if (m->m_len < sizeof(struct sadb_comb)) {
   6394 			m = m_pullup(m, sizeof(struct sadb_comb));
   6395 			if (m == NULL)
   6396 				return NULL;
   6397 		}
   6398 
   6399 		comb = mtod(m, struct sadb_comb *);
   6400 		memset(comb, 0, sizeof(*comb));
   6401 		key_getcomb_setlifetime(comb);
   6402 		comb->sadb_comb_auth = i;
   6403 		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
   6404 		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
   6405 	}
   6406 
   6407 	return m;
   6408 }
   6409 
   6410 /*
   6411  * not really an official behavior.  discussed in pf_key (at) inner.net in Sep2000.
   6412  * XXX reorder combinations by preference
   6413  */
   6414 static struct mbuf *
   6415 key_getcomb_ipcomp(int mflag)
   6416 {
   6417 	struct sadb_comb *comb;
   6418 	const struct comp_algo *algo;
   6419 	struct mbuf *m;
   6420 	int i;
   6421 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
   6422 
   6423 	m = NULL;
   6424 	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
   6425 		algo = ipcomp_algorithm_lookup(i);
   6426 		if (!algo)
   6427 			continue;
   6428 
   6429 		if (!m) {
   6430 			KASSERTMSG(l <= MLEN,
   6431 			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
   6432 			MGET(m, mflag, MT_DATA);
   6433 			if (m) {
   6434 				M_ALIGN(m, l);
   6435 				m->m_len = l;
   6436 				m->m_next = NULL;
   6437 			}
   6438 		} else
   6439 			M_PREPEND(m, l, mflag);
   6440 		if (!m)
   6441 			return NULL;
   6442 
   6443 		if (m->m_len < sizeof(struct sadb_comb)) {
   6444 			m = m_pullup(m, sizeof(struct sadb_comb));
   6445 			if (m == NULL)
   6446 				return NULL;
   6447 		}
   6448 
   6449 		comb = mtod(m, struct sadb_comb *);
   6450 		memset(comb, 0, sizeof(*comb));
   6451 		key_getcomb_setlifetime(comb);
   6452 		comb->sadb_comb_encrypt = i;
   6453 		/* what should we set into sadb_comb_*_{min,max}bits? */
   6454 	}
   6455 
   6456 	return m;
   6457 }
   6458 
   6459 /*
   6460  * XXX no way to pass mode (transport/tunnel) to userland
   6461  * XXX replay checking?
   6462  * XXX sysctl interface to ipsec_{ah,esp}_keymin
   6463  */
   6464 static struct mbuf *
   6465 key_getprop(const struct secasindex *saidx, int mflag)
   6466 {
   6467 	struct sadb_prop *prop;
   6468 	struct mbuf *m, *n;
   6469 	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
   6470 	int totlen;
   6471 
   6472 	switch (saidx->proto)  {
   6473 	case IPPROTO_ESP:
   6474 		m = key_getcomb_esp(mflag);
   6475 		break;
   6476 	case IPPROTO_AH:
   6477 		m = key_getcomb_ah(mflag);
   6478 		break;
   6479 	case IPPROTO_IPCOMP:
   6480 		m = key_getcomb_ipcomp(mflag);
   6481 		break;
   6482 	default:
   6483 		return NULL;
   6484 	}
   6485 
   6486 	if (!m)
   6487 		return NULL;
   6488 	M_PREPEND(m, l, mflag);
   6489 	if (!m)
   6490 		return NULL;
   6491 
   6492 	totlen = 0;
   6493 	for (n = m; n; n = n->m_next)
   6494 		totlen += n->m_len;
   6495 
   6496 	prop = mtod(m, struct sadb_prop *);
   6497 	memset(prop, 0, sizeof(*prop));
   6498 	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
   6499 	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
   6500 	prop->sadb_prop_replay = 32;	/* XXX */
   6501 
   6502 	return m;
   6503 }
   6504 
   6505 /*
   6506  * SADB_ACQUIRE processing called by key_checkrequest() and key_api_acquire().
   6507  * send
   6508  *   <base, SA, address(SD), (address(P)), x_policy,
   6509  *       (identity(SD),) (sensitivity,) proposal>
   6510  * to KMD, and expect to receive
   6511  *   <base> with SADB_ACQUIRE if error occurred,
   6512  * or
   6513  *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
   6514  * from KMD by PF_KEY.
   6515  *
   6516  * XXX x_policy is outside of RFC2367 (KAME extension).
   6517  * XXX sensitivity is not supported.
   6518  * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
   6519  * see comment for key_getcomb_ipcomp().
   6520  *
   6521  * OUT:
   6522  *    0     : succeed
   6523  *    others: error number
   6524  */
   6525 static int
   6526 key_acquire(const struct secasindex *saidx, const struct secpolicy *sp, int mflag)
   6527 {
   6528 	struct mbuf *result = NULL, *m;
   6529 #ifndef IPSEC_NONBLOCK_ACQUIRE
   6530 	struct secacq *newacq;
   6531 #endif
   6532 	u_int8_t satype;
   6533 	int error = -1;
   6534 	u_int32_t seq;
   6535 
   6536 	/* sanity check */
   6537 	KASSERT(saidx != NULL);
   6538 	satype = key_proto2satype(saidx->proto);
   6539 	KASSERTMSG(satype != 0, "null satype, protocol %u", saidx->proto);
   6540 
   6541 #ifndef IPSEC_NONBLOCK_ACQUIRE
   6542 	/*
   6543 	 * We never do anything about acquirng SA.  There is anather
   6544 	 * solution that kernel blocks to send SADB_ACQUIRE message until
   6545 	 * getting something message from IKEd.  In later case, to be
   6546 	 * managed with ACQUIRING list.
   6547 	 */
   6548 	/* Get an entry to check whether sending message or not. */
   6549 	mutex_enter(&key_misc.lock);
   6550 	newacq = key_getacq(saidx);
   6551 	if (newacq != NULL) {
   6552 		if (key_blockacq_count < newacq->count) {
   6553 			/* reset counter and do send message. */
   6554 			newacq->count = 0;
   6555 		} else {
   6556 			/* increment counter and do nothing. */
   6557 			newacq->count++;
   6558 			mutex_exit(&key_misc.lock);
   6559 			return 0;
   6560 		}
   6561 	} else {
   6562 		/* make new entry for blocking to send SADB_ACQUIRE. */
   6563 		newacq = key_newacq(saidx);
   6564 		if (newacq == NULL) {
   6565 			mutex_exit(&key_misc.lock);
   6566 			return ENOBUFS;
   6567 		}
   6568 
   6569 		/* add to key_misc.acqlist */
   6570 		LIST_INSERT_HEAD(&key_misc.acqlist, newacq, chain);
   6571 	}
   6572 
   6573 	seq = newacq->seq;
   6574 	mutex_exit(&key_misc.lock);
   6575 #else
   6576 	seq = (acq_seq = (acq_seq == ~0 ? 1 : ++acq_seq));
   6577 #endif
   6578 	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0, mflag);
   6579 	if (!m) {
   6580 		error = ENOBUFS;
   6581 		goto fail;
   6582 	}
   6583 	result = m;
   6584 
   6585 	/* set sadb_address for saidx's. */
   6586 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &saidx->src.sa, FULLMASK,
   6587 	    IPSEC_ULPROTO_ANY, mflag);
   6588 	if (!m) {
   6589 		error = ENOBUFS;
   6590 		goto fail;
   6591 	}
   6592 	m_cat(result, m);
   6593 
   6594 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &saidx->dst.sa, FULLMASK,
   6595 	    IPSEC_ULPROTO_ANY, mflag);
   6596 	if (!m) {
   6597 		error = ENOBUFS;
   6598 		goto fail;
   6599 	}
   6600 	m_cat(result, m);
   6601 
   6602 	/* XXX proxy address (optional) */
   6603 
   6604 	/* set sadb_x_policy */
   6605 	if (sp) {
   6606 		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
   6607 		    mflag);
   6608 		if (!m) {
   6609 			error = ENOBUFS;
   6610 			goto fail;
   6611 		}
   6612 		m_cat(result, m);
   6613 	}
   6614 
   6615 	/* XXX identity (optional) */
   6616 #if 0
   6617 	if (idexttype && fqdn) {
   6618 		/* create identity extension (FQDN) */
   6619 		struct sadb_ident *id;
   6620 		int fqdnlen;
   6621 
   6622 		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
   6623 		id = (struct sadb_ident *)p;
   6624 		memset(id, 0, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
   6625 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
   6626 		id->sadb_ident_exttype = idexttype;
   6627 		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
   6628 		memcpy(id + 1, fqdn, fqdnlen);
   6629 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
   6630 	}
   6631 
   6632 	if (idexttype) {
   6633 		/* create identity extension (USERFQDN) */
   6634 		struct sadb_ident *id;
   6635 		int userfqdnlen;
   6636 
   6637 		if (userfqdn) {
   6638 			/* +1 for terminating-NUL */
   6639 			userfqdnlen = strlen(userfqdn) + 1;
   6640 		} else
   6641 			userfqdnlen = 0;
   6642 		id = (struct sadb_ident *)p;
   6643 		memset(id, 0, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
   6644 		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
   6645 		id->sadb_ident_exttype = idexttype;
   6646 		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
   6647 		/* XXX is it correct? */
   6648 		if (curlwp)
   6649 			id->sadb_ident_id = kauth_cred_getuid(curlwp->l_cred);
   6650 		if (userfqdn && userfqdnlen)
   6651 			memcpy(id + 1, userfqdn, userfqdnlen);
   6652 		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
   6653 	}
   6654 #endif
   6655 
   6656 	/* XXX sensitivity (optional) */
   6657 
   6658 	/* create proposal/combination extension */
   6659 	m = key_getprop(saidx, mflag);
   6660 #if 0
   6661 	/*
   6662 	 * spec conformant: always attach proposal/combination extension,
   6663 	 * the problem is that we have no way to attach it for ipcomp,
   6664 	 * due to the way sadb_comb is declared in RFC2367.
   6665 	 */
   6666 	if (!m) {
   6667 		error = ENOBUFS;
   6668 		goto fail;
   6669 	}
   6670 	m_cat(result, m);
   6671 #else
   6672 	/*
   6673 	 * outside of spec; make proposal/combination extension optional.
   6674 	 */
   6675 	if (m)
   6676 		m_cat(result, m);
   6677 #endif
   6678 
   6679 	KASSERT(result->m_flags & M_PKTHDR);
   6680 	KASSERT(result->m_len >= sizeof(struct sadb_msg));
   6681 
   6682 	result->m_pkthdr.len = 0;
   6683 	for (m = result; m; m = m->m_next)
   6684 		result->m_pkthdr.len += m->m_len;
   6685 
   6686 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   6687 	    PFKEY_UNIT64(result->m_pkthdr.len);
   6688 
   6689 	/*
   6690 	 * Called from key_api_acquire that must come from userland, so
   6691 	 * we can call key_sendup_mbuf immediately.
   6692 	 */
   6693 	if (mflag == M_WAITOK)
   6694 		return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
   6695 	/*
   6696 	 * XXX we cannot call key_sendup_mbuf directly here because
   6697 	 * it can cause a deadlock:
   6698 	 * - We have a reference to an SP (and an SA) here
   6699 	 * - key_sendup_mbuf will try to take key_so_mtx
   6700 	 * - Some other thread may try to localcount_drain to the SP with
   6701 	 *   holding key_so_mtx in say key_api_spdflush
   6702 	 * - In this case localcount_drain never return because key_sendup_mbuf
   6703 	 *   that has stuck on key_so_mtx never release a reference to the SP
   6704 	 *
   6705 	 * So defer key_sendup_mbuf to the timer.
   6706 	 */
   6707 	return key_acquire_sendup_mbuf_later(result);
   6708 
   6709  fail:
   6710 	if (result)
   6711 		m_freem(result);
   6712 	return error;
   6713 }
   6714 
   6715 static struct mbuf *key_acquire_mbuf_head = NULL;
   6716 static unsigned key_acquire_mbuf_count = 0;
   6717 #define KEY_ACQUIRE_MBUF_MAX	10
   6718 
   6719 static void
   6720 key_acquire_sendup_pending_mbuf(void)
   6721 {
   6722 	struct mbuf *m, *prev;
   6723 	int error;
   6724 
   6725 again:
   6726 	prev = NULL;
   6727 	mutex_enter(&key_misc.lock);
   6728 	m = key_acquire_mbuf_head;
   6729 	/* Get an earliest mbuf (one at the tail of the list) */
   6730 	while (m != NULL) {
   6731 		if (m->m_nextpkt == NULL) {
   6732 			if (prev != NULL)
   6733 				prev->m_nextpkt = NULL;
   6734 			if (m == key_acquire_mbuf_head)
   6735 				key_acquire_mbuf_head = NULL;
   6736 			key_acquire_mbuf_count--;
   6737 			break;
   6738 		}
   6739 		prev = m;
   6740 		m = m->m_nextpkt;
   6741 	}
   6742 	mutex_exit(&key_misc.lock);
   6743 
   6744 	if (m == NULL)
   6745 		return;
   6746 
   6747 	m->m_nextpkt = NULL;
   6748 	error = key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
   6749 	if (error != 0)
   6750 		IPSECLOG(LOG_WARNING, "key_sendup_mbuf failed (error=%d)\n",
   6751 		    error);
   6752 
   6753 	if (prev != NULL)
   6754 		goto again;
   6755 }
   6756 
   6757 static int
   6758 key_acquire_sendup_mbuf_later(struct mbuf *m)
   6759 {
   6760 
   6761 	mutex_enter(&key_misc.lock);
   6762 	/* Avoid queuing too much mbufs */
   6763 	if (key_acquire_mbuf_count >= KEY_ACQUIRE_MBUF_MAX) {
   6764 		mutex_exit(&key_misc.lock);
   6765 		m_freem(m);
   6766 		return ENOBUFS; /* XXX */
   6767 	}
   6768 	/* Enqueue mbuf at the head of the list */
   6769 	m->m_nextpkt = key_acquire_mbuf_head;
   6770 	key_acquire_mbuf_head = m;
   6771 	key_acquire_mbuf_count++;
   6772 	mutex_exit(&key_misc.lock);
   6773 
   6774 	/* Kick the timer */
   6775 	key_timehandler(NULL);
   6776 
   6777 	return 0;
   6778 }
   6779 
   6780 #ifndef IPSEC_NONBLOCK_ACQUIRE
   6781 static struct secacq *
   6782 key_newacq(const struct secasindex *saidx)
   6783 {
   6784 	struct secacq *newacq;
   6785 
   6786 	/* get new entry */
   6787 	newacq = kmem_intr_zalloc(sizeof(struct secacq), KM_NOSLEEP);
   6788 	if (newacq == NULL) {
   6789 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
   6790 		return NULL;
   6791 	}
   6792 
   6793 	/* copy secindex */
   6794 	memcpy(&newacq->saidx, saidx, sizeof(newacq->saidx));
   6795 	newacq->seq = (acq_seq == ~0 ? 1 : ++acq_seq);
   6796 	newacq->created = time_uptime;
   6797 	newacq->count = 0;
   6798 
   6799 	return newacq;
   6800 }
   6801 
   6802 static struct secacq *
   6803 key_getacq(const struct secasindex *saidx)
   6804 {
   6805 	struct secacq *acq;
   6806 
   6807 	KASSERT(mutex_owned(&key_misc.lock));
   6808 
   6809 	LIST_FOREACH(acq, &key_misc.acqlist, chain) {
   6810 		if (key_saidx_match(saidx, &acq->saidx, CMP_EXACTLY))
   6811 			return acq;
   6812 	}
   6813 
   6814 	return NULL;
   6815 }
   6816 
   6817 static struct secacq *
   6818 key_getacqbyseq(u_int32_t seq)
   6819 {
   6820 	struct secacq *acq;
   6821 
   6822 	KASSERT(mutex_owned(&key_misc.lock));
   6823 
   6824 	LIST_FOREACH(acq, &key_misc.acqlist, chain) {
   6825 		if (acq->seq == seq)
   6826 			return acq;
   6827 	}
   6828 
   6829 	return NULL;
   6830 }
   6831 #endif
   6832 
   6833 #ifdef notyet
   6834 static struct secspacq *
   6835 key_newspacq(const struct secpolicyindex *spidx)
   6836 {
   6837 	struct secspacq *acq;
   6838 
   6839 	/* get new entry */
   6840 	acq = kmem_intr_zalloc(sizeof(struct secspacq), KM_NOSLEEP);
   6841 	if (acq == NULL) {
   6842 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
   6843 		return NULL;
   6844 	}
   6845 
   6846 	/* copy secindex */
   6847 	memcpy(&acq->spidx, spidx, sizeof(acq->spidx));
   6848 	acq->created = time_uptime;
   6849 	acq->count = 0;
   6850 
   6851 	return acq;
   6852 }
   6853 
   6854 static struct secspacq *
   6855 key_getspacq(const struct secpolicyindex *spidx)
   6856 {
   6857 	struct secspacq *acq;
   6858 
   6859 	LIST_FOREACH(acq, &key_misc.spacqlist, chain) {
   6860 		if (key_spidx_match_exactly(spidx, &acq->spidx))
   6861 			return acq;
   6862 	}
   6863 
   6864 	return NULL;
   6865 }
   6866 #endif /* notyet */
   6867 
   6868 /*
   6869  * SADB_ACQUIRE processing,
   6870  * in first situation, is receiving
   6871  *   <base>
   6872  * from the ikmpd, and clear sequence of its secasvar entry.
   6873  *
   6874  * In second situation, is receiving
   6875  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
   6876  * from a user land process, and return
   6877  *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
   6878  * to the socket.
   6879  *
   6880  * m will always be freed.
   6881  */
   6882 static int
   6883 key_api_acquire(struct socket *so, struct mbuf *m,
   6884       	     const struct sadb_msghdr *mhp)
   6885 {
   6886 	const struct sockaddr *src, *dst;
   6887 	struct secasindex saidx;
   6888 	u_int16_t proto;
   6889 	int error;
   6890 
   6891 	/*
   6892 	 * Error message from KMd.
   6893 	 * We assume that if error was occurred in IKEd, the length of PFKEY
   6894 	 * message is equal to the size of sadb_msg structure.
   6895 	 * We do not raise error even if error occurred in this function.
   6896 	 */
   6897 	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
   6898 #ifndef IPSEC_NONBLOCK_ACQUIRE
   6899 		struct secacq *acq;
   6900 
   6901 		/* check sequence number */
   6902 		if (mhp->msg->sadb_msg_seq == 0) {
   6903 			IPSECLOG(LOG_DEBUG, "must specify sequence number.\n");
   6904 			m_freem(m);
   6905 			return 0;
   6906 		}
   6907 
   6908 		mutex_enter(&key_misc.lock);
   6909 		acq = key_getacqbyseq(mhp->msg->sadb_msg_seq);
   6910 		if (acq == NULL) {
   6911 			mutex_exit(&key_misc.lock);
   6912 			/*
   6913 			 * the specified larval SA is already gone, or we got
   6914 			 * a bogus sequence number.  we can silently ignore it.
   6915 			 */
   6916 			m_freem(m);
   6917 			return 0;
   6918 		}
   6919 
   6920 		/* reset acq counter in order to deletion by timehander. */
   6921 		acq->created = time_uptime;
   6922 		acq->count = 0;
   6923 		mutex_exit(&key_misc.lock);
   6924 #endif
   6925 		m_freem(m);
   6926 		return 0;
   6927 	}
   6928 
   6929 	/*
   6930 	 * This message is from user land.
   6931 	 */
   6932 
   6933 	/* map satype to proto */
   6934 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
   6935 	if (proto == 0) {
   6936 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   6937 		return key_senderror(so, m, EINVAL);
   6938 	}
   6939 
   6940 	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
   6941 	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
   6942 	    mhp->ext[SADB_EXT_PROPOSAL] == NULL) {
   6943 		/* error */
   6944 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   6945 		return key_senderror(so, m, EINVAL);
   6946 	}
   6947 	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
   6948 	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
   6949 	    mhp->extlen[SADB_EXT_PROPOSAL] < sizeof(struct sadb_prop)) {
   6950 		/* error */
   6951 		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
   6952 		return key_senderror(so, m, EINVAL);
   6953 	}
   6954 
   6955 	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
   6956 	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
   6957 
   6958 	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
   6959 	if (error != 0)
   6960 		return key_senderror(so, m, EINVAL);
   6961 
   6962 	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
   6963 	if (error != 0)
   6964 		return key_senderror(so, m, EINVAL);
   6965 
   6966 	/* get a SA index */
   6967     {
   6968 	struct secashead *sah;
   6969 	int s = pserialize_read_enter();
   6970 
   6971 	sah = key_getsah(&saidx, CMP_MODE_REQID);
   6972 	if (sah != NULL) {
   6973 		pserialize_read_exit(s);
   6974 		IPSECLOG(LOG_DEBUG, "a SA exists already.\n");
   6975 		return key_senderror(so, m, EEXIST);
   6976 	}
   6977 	pserialize_read_exit(s);
   6978     }
   6979 
   6980 	error = key_acquire(&saidx, NULL, M_WAITOK);
   6981 	if (error != 0) {
   6982 		IPSECLOG(LOG_DEBUG, "error %d returned from key_acquire.\n",
   6983 		    error);
   6984 		return key_senderror(so, m, error);
   6985 	}
   6986 
   6987 	return key_sendup_mbuf(so, m, KEY_SENDUP_REGISTERED);
   6988 }
   6989 
   6990 /*
   6991  * SADB_REGISTER processing.
   6992  * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
   6993  * receive
   6994  *   <base>
   6995  * from the ikmpd, and register a socket to send PF_KEY messages,
   6996  * and send
   6997  *   <base, supported>
   6998  * to KMD by PF_KEY.
   6999  * If socket is detached, must free from regnode.
   7000  *
   7001  * m will always be freed.
   7002  */
   7003 static int
   7004 key_api_register(struct socket *so, struct mbuf *m,
   7005 	     const struct sadb_msghdr *mhp)
   7006 {
   7007 	struct secreg *reg, *newreg = 0;
   7008 
   7009 	/* check for invalid register message */
   7010 	if (mhp->msg->sadb_msg_satype >= __arraycount(key_misc.reglist))
   7011 		return key_senderror(so, m, EINVAL);
   7012 
   7013 	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
   7014 	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
   7015 		goto setmsg;
   7016 
   7017 	/* Allocate regnode in advance, out of mutex */
   7018 	newreg = kmem_zalloc(sizeof(*newreg), KM_SLEEP);
   7019 
   7020 	/* check whether existing or not */
   7021 	mutex_enter(&key_misc.lock);
   7022 	LIST_FOREACH(reg, &key_misc.reglist[mhp->msg->sadb_msg_satype], chain) {
   7023 		if (reg->so == so) {
   7024 			IPSECLOG(LOG_DEBUG, "socket exists already.\n");
   7025 			mutex_exit(&key_misc.lock);
   7026 			kmem_free(newreg, sizeof(*newreg));
   7027 			return key_senderror(so, m, EEXIST);
   7028 		}
   7029 	}
   7030 
   7031 	newreg->so = so;
   7032 	((struct keycb *)sotorawcb(so))->kp_registered++;
   7033 
   7034 	/* add regnode to key_misc.reglist. */
   7035 	LIST_INSERT_HEAD(&key_misc.reglist[mhp->msg->sadb_msg_satype], newreg, chain);
   7036 	mutex_exit(&key_misc.lock);
   7037 
   7038   setmsg:
   7039     {
   7040 	struct mbuf *n;
   7041 	struct sadb_supported *sup;
   7042 	u_int len, alen, elen;
   7043 	int off;
   7044 	int i;
   7045 	struct sadb_alg *alg;
   7046 
   7047 	/* create new sadb_msg to reply. */
   7048 	alen = 0;
   7049 	for (i = 1; i <= SADB_AALG_MAX; i++) {
   7050 		if (ah_algorithm_lookup(i))
   7051 			alen += sizeof(struct sadb_alg);
   7052 	}
   7053 	if (alen)
   7054 		alen += sizeof(struct sadb_supported);
   7055 	elen = 0;
   7056 	for (i = 1; i <= SADB_EALG_MAX; i++) {
   7057 		if (esp_algorithm_lookup(i))
   7058 			elen += sizeof(struct sadb_alg);
   7059 	}
   7060 	if (elen)
   7061 		elen += sizeof(struct sadb_supported);
   7062 
   7063 	len = sizeof(struct sadb_msg) + alen + elen;
   7064 
   7065 	if (len > MCLBYTES)
   7066 		return key_senderror(so, m, ENOBUFS);
   7067 
   7068 	n = key_alloc_mbuf_simple(len, M_WAITOK);
   7069 	n->m_pkthdr.len = n->m_len = len;
   7070 	n->m_next = NULL;
   7071 	off = 0;
   7072 
   7073 	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
   7074 	key_fill_replymsg(n, 0);
   7075 
   7076 	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
   7077 
   7078 	/* for authentication algorithm */
   7079 	if (alen) {
   7080 		sup = (struct sadb_supported *)(mtod(n, char *) + off);
   7081 		sup->sadb_supported_len = PFKEY_UNIT64(alen);
   7082 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
   7083 		off += PFKEY_ALIGN8(sizeof(*sup));
   7084 
   7085 		for (i = 1; i <= SADB_AALG_MAX; i++) {
   7086 			const struct auth_hash *aalgo;
   7087 			u_int16_t minkeysize, maxkeysize;
   7088 
   7089 			aalgo = ah_algorithm_lookup(i);
   7090 			if (!aalgo)
   7091 				continue;
   7092 			alg = (struct sadb_alg *)(mtod(n, char *) + off);
   7093 			alg->sadb_alg_id = i;
   7094 			alg->sadb_alg_ivlen = 0;
   7095 			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
   7096 			alg->sadb_alg_minbits = _BITS(minkeysize);
   7097 			alg->sadb_alg_maxbits = _BITS(maxkeysize);
   7098 			off += PFKEY_ALIGN8(sizeof(*alg));
   7099 		}
   7100 	}
   7101 
   7102 	/* for encryption algorithm */
   7103 	if (elen) {
   7104 		sup = (struct sadb_supported *)(mtod(n, char *) + off);
   7105 		sup->sadb_supported_len = PFKEY_UNIT64(elen);
   7106 		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
   7107 		off += PFKEY_ALIGN8(sizeof(*sup));
   7108 
   7109 		for (i = 1; i <= SADB_EALG_MAX; i++) {
   7110 			const struct enc_xform *ealgo;
   7111 
   7112 			ealgo = esp_algorithm_lookup(i);
   7113 			if (!ealgo)
   7114 				continue;
   7115 			alg = (struct sadb_alg *)(mtod(n, char *) + off);
   7116 			alg->sadb_alg_id = i;
   7117 			alg->sadb_alg_ivlen = ealgo->blocksize;
   7118 			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
   7119 			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
   7120 			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
   7121 		}
   7122 	}
   7123 
   7124 	KASSERTMSG(off == len, "length inconsistency");
   7125 
   7126 	m_freem(m);
   7127 	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
   7128     }
   7129 }
   7130 
   7131 /*
   7132  * free secreg entry registered.
   7133  * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
   7134  */
   7135 void
   7136 key_freereg(struct socket *so)
   7137 {
   7138 	struct secreg *reg;
   7139 	int i;
   7140 
   7141 	KASSERT(!cpu_softintr_p());
   7142 	KASSERT(so != NULL);
   7143 
   7144 	/*
   7145 	 * check whether existing or not.
   7146 	 * check all type of SA, because there is a potential that
   7147 	 * one socket is registered to multiple type of SA.
   7148 	 */
   7149 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
   7150 		mutex_enter(&key_misc.lock);
   7151 		LIST_FOREACH(reg, &key_misc.reglist[i], chain) {
   7152 			if (reg->so == so) {
   7153 				LIST_REMOVE(reg, chain);
   7154 				break;
   7155 			}
   7156 		}
   7157 		mutex_exit(&key_misc.lock);
   7158 		if (reg != NULL)
   7159 			kmem_free(reg, sizeof(*reg));
   7160 	}
   7161 
   7162 	return;
   7163 }
   7164 
   7165 /*
   7166  * SADB_EXPIRE processing
   7167  * send
   7168  *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
   7169  * to KMD by PF_KEY.
   7170  * NOTE: We send only soft lifetime extension.
   7171  *
   7172  * OUT:	0	: succeed
   7173  *	others	: error number
   7174  */
   7175 static int
   7176 key_expire(struct secasvar *sav)
   7177 {
   7178 	int s;
   7179 	int satype;
   7180 	struct mbuf *result = NULL, *m;
   7181 	int len;
   7182 	int error = -1;
   7183 	struct sadb_lifetime *lt;
   7184 
   7185 	/* XXX: Why do we lock ? */
   7186 	s = splsoftnet();	/*called from softclock()*/
   7187 
   7188 	KASSERT(sav != NULL);
   7189 
   7190 	satype = key_proto2satype(sav->sah->saidx.proto);
   7191 	KASSERTMSG(satype != 0, "invalid proto is passed");
   7192 
   7193 	/* set msg header */
   7194 	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, key_sa_refcnt(sav),
   7195 	    M_WAITOK);
   7196 	result = m;
   7197 
   7198 	/* create SA extension */
   7199 	m = key_setsadbsa(sav);
   7200 	m_cat(result, m);
   7201 
   7202 	/* create SA extension */
   7203 	m = key_setsadbxsa2(sav->sah->saidx.mode,
   7204 	    sav->replay ? sav->replay->count : 0, sav->sah->saidx.reqid);
   7205 	m_cat(result, m);
   7206 
   7207 	/* create lifetime extension (current and soft) */
   7208 	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
   7209 	m = key_alloc_mbuf(len, M_WAITOK);
   7210 	KASSERT(m->m_next == NULL);
   7211 
   7212 	memset(mtod(m, void *), 0, len);
   7213 	lt = mtod(m, struct sadb_lifetime *);
   7214 	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
   7215 	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
   7216 	lt->sadb_lifetime_allocations = sav->lft_c->sadb_lifetime_allocations;
   7217 	lt->sadb_lifetime_bytes = sav->lft_c->sadb_lifetime_bytes;
   7218 	lt->sadb_lifetime_addtime =
   7219 	    time_mono_to_wall(sav->lft_c->sadb_lifetime_addtime);
   7220 	lt->sadb_lifetime_usetime =
   7221 	    time_mono_to_wall(sav->lft_c->sadb_lifetime_usetime);
   7222 	lt = (struct sadb_lifetime *)(mtod(m, char *) + len / 2);
   7223 	memcpy(lt, sav->lft_s, sizeof(*lt));
   7224 	m_cat(result, m);
   7225 
   7226 	/* set sadb_address for source */
   7227 	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa,
   7228 	    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
   7229 	m_cat(result, m);
   7230 
   7231 	/* set sadb_address for destination */
   7232 	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.dst.sa,
   7233 	    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
   7234 	m_cat(result, m);
   7235 
   7236 	if ((result->m_flags & M_PKTHDR) == 0) {
   7237 		error = EINVAL;
   7238 		goto fail;
   7239 	}
   7240 
   7241 	if (result->m_len < sizeof(struct sadb_msg)) {
   7242 		result = m_pullup(result, sizeof(struct sadb_msg));
   7243 		if (result == NULL) {
   7244 			error = ENOBUFS;
   7245 			goto fail;
   7246 		}
   7247 	}
   7248 
   7249 	result->m_pkthdr.len = 0;
   7250 	for (m = result; m; m = m->m_next)
   7251 		result->m_pkthdr.len += m->m_len;
   7252 
   7253 	mtod(result, struct sadb_msg *)->sadb_msg_len =
   7254 	    PFKEY_UNIT64(result->m_pkthdr.len);
   7255 
   7256 	splx(s);
   7257 	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
   7258 
   7259  fail:
   7260 	if (result)
   7261 		m_freem(result);
   7262 	splx(s);
   7263 	return error;
   7264 }
   7265 
   7266 /*
   7267  * SADB_FLUSH processing
   7268  * receive
   7269  *   <base>
   7270  * from the ikmpd, and free all entries in secastree.
   7271  * and send,
   7272  *   <base>
   7273  * to the ikmpd.
   7274  * NOTE: to do is only marking SADB_SASTATE_DEAD.
   7275  *
   7276  * m will always be freed.
   7277  */
   7278 static int
   7279 key_api_flush(struct socket *so, struct mbuf *m,
   7280           const struct sadb_msghdr *mhp)
   7281 {
   7282 	struct sadb_msg *newmsg;
   7283 	struct secashead *sah;
   7284 	struct secasvar *sav;
   7285 	u_int16_t proto;
   7286 	u_int8_t state;
   7287 	int s;
   7288 
   7289 	/* map satype to proto */
   7290 	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
   7291 	if (proto == 0) {
   7292 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   7293 		return key_senderror(so, m, EINVAL);
   7294 	}
   7295 
   7296 	/* no SATYPE specified, i.e. flushing all SA. */
   7297 	s = pserialize_read_enter();
   7298 	SAHLIST_READER_FOREACH(sah) {
   7299 		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
   7300 		    proto != sah->saidx.proto)
   7301 			continue;
   7302 
   7303 		key_sah_ref(sah);
   7304 		pserialize_read_exit(s);
   7305 
   7306 		SASTATE_ALIVE_FOREACH(state) {
   7307 		restart:
   7308 			mutex_enter(&key_sad.lock);
   7309 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
   7310 				sav->state = SADB_SASTATE_DEAD;
   7311 				key_unlink_sav(sav);
   7312 				mutex_exit(&key_sad.lock);
   7313 				key_destroy_sav(sav);
   7314 				goto restart;
   7315 			}
   7316 			mutex_exit(&key_sad.lock);
   7317 		}
   7318 
   7319 		s = pserialize_read_enter();
   7320 		sah->state = SADB_SASTATE_DEAD;
   7321 		key_sah_unref(sah);
   7322 	}
   7323 	pserialize_read_exit(s);
   7324 
   7325 	if (m->m_len < sizeof(struct sadb_msg) ||
   7326 	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
   7327 		IPSECLOG(LOG_DEBUG, "No more memory.\n");
   7328 		return key_senderror(so, m, ENOBUFS);
   7329 	}
   7330 
   7331 	if (m->m_next)
   7332 		m_freem(m->m_next);
   7333 	m->m_next = NULL;
   7334 	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
   7335 	newmsg = mtod(m, struct sadb_msg *);
   7336 	newmsg->sadb_msg_errno = 0;
   7337 	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
   7338 
   7339 	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
   7340 }
   7341 
   7342 
   7343 static struct mbuf *
   7344 key_setdump_chain(u_int8_t req_satype, int *errorp, int *lenp, pid_t pid)
   7345 {
   7346 	struct secashead *sah;
   7347 	struct secasvar *sav;
   7348 	u_int16_t proto;
   7349 	u_int8_t satype;
   7350 	u_int8_t state;
   7351 	int cnt;
   7352 	struct mbuf *m, *n, *prev;
   7353 
   7354 	KASSERT(mutex_owned(&key_sad.lock));
   7355 
   7356 	*lenp = 0;
   7357 
   7358 	/* map satype to proto */
   7359 	proto = key_satype2proto(req_satype);
   7360 	if (proto == 0) {
   7361 		*errorp = EINVAL;
   7362 		return (NULL);
   7363 	}
   7364 
   7365 	/* count sav entries to be sent to userland. */
   7366 	cnt = 0;
   7367 	SAHLIST_WRITER_FOREACH(sah) {
   7368 		if (req_satype != SADB_SATYPE_UNSPEC &&
   7369 		    proto != sah->saidx.proto)
   7370 			continue;
   7371 
   7372 		SASTATE_ANY_FOREACH(state) {
   7373 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
   7374 				cnt++;
   7375 			}
   7376 		}
   7377 	}
   7378 
   7379 	if (cnt == 0) {
   7380 		*errorp = ENOENT;
   7381 		return (NULL);
   7382 	}
   7383 
   7384 	/* send this to the userland, one at a time. */
   7385 	m = NULL;
   7386 	prev = m;
   7387 	SAHLIST_WRITER_FOREACH(sah) {
   7388 		if (req_satype != SADB_SATYPE_UNSPEC &&
   7389 		    proto != sah->saidx.proto)
   7390 			continue;
   7391 
   7392 		/* map proto to satype */
   7393 		satype = key_proto2satype(sah->saidx.proto);
   7394 		if (satype == 0) {
   7395 			m_freem(m);
   7396 			*errorp = EINVAL;
   7397 			return (NULL);
   7398 		}
   7399 
   7400 		SASTATE_ANY_FOREACH(state) {
   7401 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
   7402 				n = key_setdumpsa(sav, SADB_DUMP, satype,
   7403 				    --cnt, pid);
   7404 				if (!m)
   7405 					m = n;
   7406 				else
   7407 					prev->m_nextpkt = n;
   7408 				prev = n;
   7409 			}
   7410 		}
   7411 	}
   7412 
   7413 	if (!m) {
   7414 		*errorp = EINVAL;
   7415 		return (NULL);
   7416 	}
   7417 
   7418 	if ((m->m_flags & M_PKTHDR) != 0) {
   7419 		m->m_pkthdr.len = 0;
   7420 		for (n = m; n; n = n->m_next)
   7421 			m->m_pkthdr.len += n->m_len;
   7422 	}
   7423 
   7424 	*errorp = 0;
   7425 	return (m);
   7426 }
   7427 
   7428 /*
   7429  * SADB_DUMP processing
   7430  * dump all entries including status of DEAD in SAD.
   7431  * receive
   7432  *   <base>
   7433  * from the ikmpd, and dump all secasvar leaves
   7434  * and send,
   7435  *   <base> .....
   7436  * to the ikmpd.
   7437  *
   7438  * m will always be freed.
   7439  */
   7440 static int
   7441 key_api_dump(struct socket *so, struct mbuf *m0,
   7442 	 const struct sadb_msghdr *mhp)
   7443 {
   7444 	u_int16_t proto;
   7445 	u_int8_t satype;
   7446 	struct mbuf *n;
   7447 	int error, len, ok;
   7448 
   7449 	/* map satype to proto */
   7450 	satype = mhp->msg->sadb_msg_satype;
   7451 	proto = key_satype2proto(satype);
   7452 	if (proto == 0) {
   7453 		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
   7454 		return key_senderror(so, m0, EINVAL);
   7455 	}
   7456 
   7457 	/*
   7458 	 * If the requestor has insufficient socket-buffer space
   7459 	 * for the entire chain, nobody gets any response to the DUMP.
   7460 	 * XXX For now, only the requestor ever gets anything.
   7461 	 * Moreover, if the requestor has any space at all, they receive
   7462 	 * the entire chain, otherwise the request is refused with ENOBUFS.
   7463 	 */
   7464 	if (sbspace(&so->so_rcv) <= 0) {
   7465 		return key_senderror(so, m0, ENOBUFS);
   7466 	}
   7467 
   7468 	mutex_enter(&key_sad.lock);
   7469 	n = key_setdump_chain(satype, &error, &len, mhp->msg->sadb_msg_pid);
   7470 	mutex_exit(&key_sad.lock);
   7471 
   7472 	if (n == NULL) {
   7473 		return key_senderror(so, m0, ENOENT);
   7474 	}
   7475 	{
   7476 		uint64_t *ps = PFKEY_STAT_GETREF();
   7477 		ps[PFKEY_STAT_IN_TOTAL]++;
   7478 		ps[PFKEY_STAT_IN_BYTES] += len;
   7479 		PFKEY_STAT_PUTREF();
   7480 	}
   7481 
   7482 	/*
   7483 	 * PF_KEY DUMP responses are no longer broadcast to all PF_KEY sockets.
   7484 	 * The requestor receives either the entire chain, or an
   7485 	 * error message with ENOBUFS.
   7486 	 *
   7487 	 * sbappendaddrchain() takes the chain of entries, one
   7488 	 * packet-record per SPD entry, prepends the key_src sockaddr
   7489 	 * to each packet-record, links the sockaddr mbufs into a new
   7490 	 * list of records, then   appends the entire resulting
   7491 	 * list to the requesting socket.
   7492 	 */
   7493 	ok = sbappendaddrchain(&so->so_rcv, (struct sockaddr *)&key_src, n,
   7494 	    SB_PRIO_ONESHOT_OVERFLOW);
   7495 
   7496 	if (!ok) {
   7497 		PFKEY_STATINC(PFKEY_STAT_IN_NOMEM);
   7498 		m_freem(n);
   7499 		return key_senderror(so, m0, ENOBUFS);
   7500 	}
   7501 
   7502 	m_freem(m0);
   7503 	return 0;
   7504 }
   7505 
   7506 /*
   7507  * SADB_X_PROMISC processing
   7508  *
   7509  * m will always be freed.
   7510  */
   7511 static int
   7512 key_api_promisc(struct socket *so, struct mbuf *m,
   7513 	    const struct sadb_msghdr *mhp)
   7514 {
   7515 	int olen;
   7516 
   7517 	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
   7518 
   7519 	if (olen < sizeof(struct sadb_msg)) {
   7520 #if 1
   7521 		return key_senderror(so, m, EINVAL);
   7522 #else
   7523 		m_freem(m);
   7524 		return 0;
   7525 #endif
   7526 	} else if (olen == sizeof(struct sadb_msg)) {
   7527 		/* enable/disable promisc mode */
   7528 		struct keycb *kp = (struct keycb *)sotorawcb(so);
   7529 		if (kp == NULL)
   7530 			return key_senderror(so, m, EINVAL);
   7531 		mhp->msg->sadb_msg_errno = 0;
   7532 		switch (mhp->msg->sadb_msg_satype) {
   7533 		case 0:
   7534 		case 1:
   7535 			kp->kp_promisc = mhp->msg->sadb_msg_satype;
   7536 			break;
   7537 		default:
   7538 			return key_senderror(so, m, EINVAL);
   7539 		}
   7540 
   7541 		/* send the original message back to everyone */
   7542 		mhp->msg->sadb_msg_errno = 0;
   7543 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
   7544 	} else {
   7545 		/* send packet as is */
   7546 
   7547 		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
   7548 
   7549 		/* TODO: if sadb_msg_seq is specified, send to specific pid */
   7550 		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
   7551 	}
   7552 }
   7553 
   7554 static int (*key_api_typesw[]) (struct socket *, struct mbuf *,
   7555 		const struct sadb_msghdr *) = {
   7556 	NULL,			/* SADB_RESERVED */
   7557 	key_api_getspi,		/* SADB_GETSPI */
   7558 	key_api_update,		/* SADB_UPDATE */
   7559 	key_api_add,		/* SADB_ADD */
   7560 	key_api_delete,		/* SADB_DELETE */
   7561 	key_api_get,		/* SADB_GET */
   7562 	key_api_acquire,	/* SADB_ACQUIRE */
   7563 	key_api_register,	/* SADB_REGISTER */
   7564 	NULL,			/* SADB_EXPIRE */
   7565 	key_api_flush,		/* SADB_FLUSH */
   7566 	key_api_dump,		/* SADB_DUMP */
   7567 	key_api_promisc,	/* SADB_X_PROMISC */
   7568 	NULL,			/* SADB_X_PCHANGE */
   7569 	key_api_spdadd,		/* SADB_X_SPDUPDATE */
   7570 	key_api_spdadd,		/* SADB_X_SPDADD */
   7571 	key_api_spddelete,	/* SADB_X_SPDDELETE */
   7572 	key_api_spdget,		/* SADB_X_SPDGET */
   7573 	NULL,			/* SADB_X_SPDACQUIRE */
   7574 	key_api_spddump,	/* SADB_X_SPDDUMP */
   7575 	key_api_spdflush,	/* SADB_X_SPDFLUSH */
   7576 	key_api_spdadd,		/* SADB_X_SPDSETIDX */
   7577 	NULL,			/* SADB_X_SPDEXPIRE */
   7578 	key_api_spddelete2,	/* SADB_X_SPDDELETE2 */
   7579 	key_api_nat_map,	/* SADB_X_NAT_T_NEW_MAPPING */
   7580 };
   7581 
   7582 /*
   7583  * parse sadb_msg buffer to process PFKEYv2,
   7584  * and create a data to response if needed.
   7585  * I think to be dealed with mbuf directly.
   7586  * IN:
   7587  *     msgp  : pointer to pointer to a received buffer pulluped.
   7588  *             This is rewrited to response.
   7589  *     so    : pointer to socket.
   7590  * OUT:
   7591  *    length for buffer to send to user process.
   7592  */
   7593 int
   7594 key_parse(struct mbuf *m, struct socket *so)
   7595 {
   7596 	struct sadb_msg *msg;
   7597 	struct sadb_msghdr mh;
   7598 	u_int orglen;
   7599 	int error;
   7600 
   7601 	KASSERT(m != NULL);
   7602 	KASSERT(so != NULL);
   7603 
   7604 #if 0	/*kdebug_sadb assumes msg in linear buffer*/
   7605 	if (KEYDEBUG_ON(KEYDEBUG_KEY_DUMP)) {
   7606 		kdebug_sadb("passed sadb_msg", msg);
   7607 	}
   7608 #endif
   7609 
   7610 	if (m->m_len < sizeof(struct sadb_msg)) {
   7611 		m = m_pullup(m, sizeof(struct sadb_msg));
   7612 		if (!m)
   7613 			return ENOBUFS;
   7614 	}
   7615 	msg = mtod(m, struct sadb_msg *);
   7616 	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
   7617 
   7618 	if ((m->m_flags & M_PKTHDR) == 0 ||
   7619 	    m->m_pkthdr.len != orglen) {
   7620 		IPSECLOG(LOG_DEBUG, "invalid message length.\n");
   7621 		PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
   7622 		error = EINVAL;
   7623 		goto senderror;
   7624 	}
   7625 
   7626 	if (msg->sadb_msg_version != PF_KEY_V2) {
   7627 		IPSECLOG(LOG_DEBUG, "PF_KEY version %u is mismatched.\n",
   7628 		    msg->sadb_msg_version);
   7629 		PFKEY_STATINC(PFKEY_STAT_OUT_INVVER);
   7630 		error = EINVAL;
   7631 		goto senderror;
   7632 	}
   7633 
   7634 	if (msg->sadb_msg_type > SADB_MAX) {
   7635 		IPSECLOG(LOG_DEBUG, "invalid type %u is passed.\n",
   7636 		    msg->sadb_msg_type);
   7637 		PFKEY_STATINC(PFKEY_STAT_OUT_INVMSGTYPE);
   7638 		error = EINVAL;
   7639 		goto senderror;
   7640 	}
   7641 
   7642 	/* for old-fashioned code - should be nuked */
   7643 	if (m->m_pkthdr.len > MCLBYTES) {
   7644 		m_freem(m);
   7645 		return ENOBUFS;
   7646 	}
   7647 	if (m->m_next) {
   7648 		struct mbuf *n;
   7649 
   7650 		n = key_alloc_mbuf_simple(m->m_pkthdr.len, M_WAITOK);
   7651 
   7652 		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, void *));
   7653 		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
   7654 		n->m_next = NULL;
   7655 		m_freem(m);
   7656 		m = n;
   7657 	}
   7658 
   7659 	/* align the mbuf chain so that extensions are in contiguous region. */
   7660 	error = key_align(m, &mh);
   7661 	if (error)
   7662 		return error;
   7663 
   7664 	if (m->m_next) {	/*XXX*/
   7665 		m_freem(m);
   7666 		return ENOBUFS;
   7667 	}
   7668 
   7669 	msg = mh.msg;
   7670 
   7671 	/* check SA type */
   7672 	switch (msg->sadb_msg_satype) {
   7673 	case SADB_SATYPE_UNSPEC:
   7674 		switch (msg->sadb_msg_type) {
   7675 		case SADB_GETSPI:
   7676 		case SADB_UPDATE:
   7677 		case SADB_ADD:
   7678 		case SADB_DELETE:
   7679 		case SADB_GET:
   7680 		case SADB_ACQUIRE:
   7681 		case SADB_EXPIRE:
   7682 			IPSECLOG(LOG_DEBUG,
   7683 			    "must specify satype when msg type=%u.\n",
   7684 			    msg->sadb_msg_type);
   7685 			PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
   7686 			error = EINVAL;
   7687 			goto senderror;
   7688 		}
   7689 		break;
   7690 	case SADB_SATYPE_AH:
   7691 	case SADB_SATYPE_ESP:
   7692 	case SADB_X_SATYPE_IPCOMP:
   7693 	case SADB_X_SATYPE_TCPSIGNATURE:
   7694 		switch (msg->sadb_msg_type) {
   7695 		case SADB_X_SPDADD:
   7696 		case SADB_X_SPDDELETE:
   7697 		case SADB_X_SPDGET:
   7698 		case SADB_X_SPDDUMP:
   7699 		case SADB_X_SPDFLUSH:
   7700 		case SADB_X_SPDSETIDX:
   7701 		case SADB_X_SPDUPDATE:
   7702 		case SADB_X_SPDDELETE2:
   7703 			IPSECLOG(LOG_DEBUG, "illegal satype=%u\n",
   7704 			    msg->sadb_msg_type);
   7705 			PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
   7706 			error = EINVAL;
   7707 			goto senderror;
   7708 		}
   7709 		break;
   7710 	case SADB_SATYPE_RSVP:
   7711 	case SADB_SATYPE_OSPFV2:
   7712 	case SADB_SATYPE_RIPV2:
   7713 	case SADB_SATYPE_MIP:
   7714 		IPSECLOG(LOG_DEBUG, "type %u isn't supported.\n",
   7715 		    msg->sadb_msg_satype);
   7716 		PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
   7717 		error = EOPNOTSUPP;
   7718 		goto senderror;
   7719 	case 1:	/* XXX: What does it do? */
   7720 		if (msg->sadb_msg_type == SADB_X_PROMISC)
   7721 			break;
   7722 		/*FALLTHROUGH*/
   7723 	default:
   7724 		IPSECLOG(LOG_DEBUG, "invalid type %u is passed.\n",
   7725 		    msg->sadb_msg_satype);
   7726 		PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
   7727 		error = EINVAL;
   7728 		goto senderror;
   7729 	}
   7730 
   7731 	/* check field of upper layer protocol and address family */
   7732 	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL &&
   7733 	    mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
   7734 		const struct sadb_address *src0, *dst0;
   7735 		const struct sockaddr *sa0, *da0;
   7736 		u_int plen;
   7737 
   7738 		src0 = mh.ext[SADB_EXT_ADDRESS_SRC];
   7739 		dst0 = mh.ext[SADB_EXT_ADDRESS_DST];
   7740 		sa0 = key_msghdr_get_sockaddr(&mh, SADB_EXT_ADDRESS_SRC);
   7741 		da0 = key_msghdr_get_sockaddr(&mh, SADB_EXT_ADDRESS_DST);
   7742 
   7743 		/* check upper layer protocol */
   7744 		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
   7745 			IPSECLOG(LOG_DEBUG,
   7746 			    "upper layer protocol mismatched.\n");
   7747 			goto invaddr;
   7748 		}
   7749 
   7750 		/* check family */
   7751 		if (sa0->sa_family != da0->sa_family) {
   7752 			IPSECLOG(LOG_DEBUG, "address family mismatched.\n");
   7753 			goto invaddr;
   7754 		}
   7755 		if (sa0->sa_len != da0->sa_len) {
   7756 			IPSECLOG(LOG_DEBUG,
   7757 			    "address struct size mismatched.\n");
   7758 			goto invaddr;
   7759 		}
   7760 
   7761 		switch (sa0->sa_family) {
   7762 		case AF_INET:
   7763 			if (sa0->sa_len != sizeof(struct sockaddr_in))
   7764 				goto invaddr;
   7765 			break;
   7766 		case AF_INET6:
   7767 			if (sa0->sa_len != sizeof(struct sockaddr_in6))
   7768 				goto invaddr;
   7769 			break;
   7770 		default:
   7771 			IPSECLOG(LOG_DEBUG, "unsupported address family.\n");
   7772 			error = EAFNOSUPPORT;
   7773 			goto senderror;
   7774 		}
   7775 
   7776 		switch (sa0->sa_family) {
   7777 		case AF_INET:
   7778 			plen = sizeof(struct in_addr) << 3;
   7779 			break;
   7780 		case AF_INET6:
   7781 			plen = sizeof(struct in6_addr) << 3;
   7782 			break;
   7783 		default:
   7784 			plen = 0;	/*fool gcc*/
   7785 			break;
   7786 		}
   7787 
   7788 		/* check max prefix length */
   7789 		if (src0->sadb_address_prefixlen > plen ||
   7790 		    dst0->sadb_address_prefixlen > plen) {
   7791 			IPSECLOG(LOG_DEBUG, "illegal prefixlen.\n");
   7792 			goto invaddr;
   7793 		}
   7794 
   7795 		/*
   7796 		 * prefixlen == 0 is valid because there can be a case when
   7797 		 * all addresses are matched.
   7798 		 */
   7799 	}
   7800 
   7801 	if (msg->sadb_msg_type >= __arraycount(key_api_typesw) ||
   7802 	    key_api_typesw[msg->sadb_msg_type] == NULL) {
   7803 		PFKEY_STATINC(PFKEY_STAT_OUT_INVMSGTYPE);
   7804 		error = EINVAL;
   7805 		goto senderror;
   7806 	}
   7807 
   7808 	return (*key_api_typesw[msg->sadb_msg_type])(so, m, &mh);
   7809 
   7810 invaddr:
   7811 	error = EINVAL;
   7812 senderror:
   7813 	PFKEY_STATINC(PFKEY_STAT_OUT_INVADDR);
   7814 	return key_senderror(so, m, error);
   7815 }
   7816 
   7817 static int
   7818 key_senderror(struct socket *so, struct mbuf *m, int code)
   7819 {
   7820 	struct sadb_msg *msg;
   7821 
   7822 	KASSERT(m->m_len >= sizeof(struct sadb_msg));
   7823 
   7824 	if (so == NULL) {
   7825 		/*
   7826 		 * This means the request comes from kernel.
   7827 		 * As the request comes from kernel, it is unnecessary to
   7828 		 * send message to userland. Just return errcode directly.
   7829 		 */
   7830 		m_freem(m);
   7831 		return code;
   7832 	}
   7833 
   7834 	msg = mtod(m, struct sadb_msg *);
   7835 	msg->sadb_msg_errno = code;
   7836 	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
   7837 }
   7838 
   7839 /*
   7840  * set the pointer to each header into message buffer.
   7841  * m will be freed on error.
   7842  * XXX larger-than-MCLBYTES extension?
   7843  */
   7844 static int
   7845 key_align(struct mbuf *m, struct sadb_msghdr *mhp)
   7846 {
   7847 	struct mbuf *n;
   7848 	struct sadb_ext *ext;
   7849 	size_t off, end;
   7850 	int extlen;
   7851 	int toff;
   7852 
   7853 	KASSERT(m != NULL);
   7854 	KASSERT(mhp != NULL);
   7855 	KASSERT(m->m_len >= sizeof(struct sadb_msg));
   7856 
   7857 	/* initialize */
   7858 	memset(mhp, 0, sizeof(*mhp));
   7859 
   7860 	mhp->msg = mtod(m, struct sadb_msg *);
   7861 	mhp->ext[0] = mhp->msg;	/*XXX backward compat */
   7862 
   7863 	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
   7864 	extlen = end;	/*just in case extlen is not updated*/
   7865 	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
   7866 		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
   7867 		if (!n) {
   7868 			/* m is already freed */
   7869 			return ENOBUFS;
   7870 		}
   7871 		ext = (struct sadb_ext *)(mtod(n, char *) + toff);
   7872 
   7873 		/* set pointer */
   7874 		switch (ext->sadb_ext_type) {
   7875 		case SADB_EXT_SA:
   7876 		case SADB_EXT_ADDRESS_SRC:
   7877 		case SADB_EXT_ADDRESS_DST:
   7878 		case SADB_EXT_ADDRESS_PROXY:
   7879 		case SADB_EXT_LIFETIME_CURRENT:
   7880 		case SADB_EXT_LIFETIME_HARD:
   7881 		case SADB_EXT_LIFETIME_SOFT:
   7882 		case SADB_EXT_KEY_AUTH:
   7883 		case SADB_EXT_KEY_ENCRYPT:
   7884 		case SADB_EXT_IDENTITY_SRC:
   7885 		case SADB_EXT_IDENTITY_DST:
   7886 		case SADB_EXT_SENSITIVITY:
   7887 		case SADB_EXT_PROPOSAL:
   7888 		case SADB_EXT_SUPPORTED_AUTH:
   7889 		case SADB_EXT_SUPPORTED_ENCRYPT:
   7890 		case SADB_EXT_SPIRANGE:
   7891 		case SADB_X_EXT_POLICY:
   7892 		case SADB_X_EXT_SA2:
   7893 		case SADB_X_EXT_NAT_T_TYPE:
   7894 		case SADB_X_EXT_NAT_T_SPORT:
   7895 		case SADB_X_EXT_NAT_T_DPORT:
   7896 		case SADB_X_EXT_NAT_T_OAI:
   7897 		case SADB_X_EXT_NAT_T_OAR:
   7898 		case SADB_X_EXT_NAT_T_FRAG:
   7899 			/* duplicate check */
   7900 			/*
   7901 			 * XXX Are there duplication payloads of either
   7902 			 * KEY_AUTH or KEY_ENCRYPT ?
   7903 			 */
   7904 			if (mhp->ext[ext->sadb_ext_type] != NULL) {
   7905 				IPSECLOG(LOG_DEBUG,
   7906 				    "duplicate ext_type %u is passed.\n",
   7907 				    ext->sadb_ext_type);
   7908 				m_freem(m);
   7909 				PFKEY_STATINC(PFKEY_STAT_OUT_DUPEXT);
   7910 				return EINVAL;
   7911 			}
   7912 			break;
   7913 		default:
   7914 			IPSECLOG(LOG_DEBUG, "invalid ext_type %u is passed.\n",
   7915 			    ext->sadb_ext_type);
   7916 			m_freem(m);
   7917 			PFKEY_STATINC(PFKEY_STAT_OUT_INVEXTTYPE);
   7918 			return EINVAL;
   7919 		}
   7920 
   7921 		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
   7922 
   7923 		if (key_validate_ext(ext, extlen)) {
   7924 			m_freem(m);
   7925 			PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
   7926 			return EINVAL;
   7927 		}
   7928 
   7929 		n = m_pulldown(m, off, extlen, &toff);
   7930 		if (!n) {
   7931 			/* m is already freed */
   7932 			return ENOBUFS;
   7933 		}
   7934 		ext = (struct sadb_ext *)(mtod(n, char *) + toff);
   7935 
   7936 		mhp->ext[ext->sadb_ext_type] = ext;
   7937 		mhp->extoff[ext->sadb_ext_type] = off;
   7938 		mhp->extlen[ext->sadb_ext_type] = extlen;
   7939 	}
   7940 
   7941 	if (off != end) {
   7942 		m_freem(m);
   7943 		PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
   7944 		return EINVAL;
   7945 	}
   7946 
   7947 	return 0;
   7948 }
   7949 
   7950 static int
   7951 key_validate_ext(const struct sadb_ext *ext, int len)
   7952 {
   7953 	const struct sockaddr *sa;
   7954 	enum { NONE, ADDR } checktype = NONE;
   7955 	int baselen = 0;
   7956 	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
   7957 
   7958 	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
   7959 		return EINVAL;
   7960 
   7961 	/* if it does not match minimum/maximum length, bail */
   7962 	if (ext->sadb_ext_type >= __arraycount(minsize) ||
   7963 	    ext->sadb_ext_type >= __arraycount(maxsize))
   7964 		return EINVAL;
   7965 	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
   7966 		return EINVAL;
   7967 	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
   7968 		return EINVAL;
   7969 
   7970 	/* more checks based on sadb_ext_type XXX need more */
   7971 	switch (ext->sadb_ext_type) {
   7972 	case SADB_EXT_ADDRESS_SRC:
   7973 	case SADB_EXT_ADDRESS_DST:
   7974 	case SADB_EXT_ADDRESS_PROXY:
   7975 		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
   7976 		checktype = ADDR;
   7977 		break;
   7978 	case SADB_EXT_IDENTITY_SRC:
   7979 	case SADB_EXT_IDENTITY_DST:
   7980 		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
   7981 		    SADB_X_IDENTTYPE_ADDR) {
   7982 			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
   7983 			checktype = ADDR;
   7984 		} else
   7985 			checktype = NONE;
   7986 		break;
   7987 	default:
   7988 		checktype = NONE;
   7989 		break;
   7990 	}
   7991 
   7992 	switch (checktype) {
   7993 	case NONE:
   7994 		break;
   7995 	case ADDR:
   7996 		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
   7997 		if (len < baselen + sal)
   7998 			return EINVAL;
   7999 		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
   8000 			return EINVAL;
   8001 		break;
   8002 	}
   8003 
   8004 	return 0;
   8005 }
   8006 
   8007 static int
   8008 key_do_init(void)
   8009 {
   8010 	int i, error;
   8011 
   8012 	mutex_init(&key_misc.lock, MUTEX_DEFAULT, IPL_NONE);
   8013 
   8014 	mutex_init(&key_spd.lock, MUTEX_DEFAULT, IPL_NONE);
   8015 	cv_init(&key_spd.cv_lc, "key_sp_lc");
   8016 	key_spd.psz = pserialize_create();
   8017 	cv_init(&key_spd.cv_psz, "key_sp_psz");
   8018 	key_spd.psz_performing = false;
   8019 
   8020 	mutex_init(&key_sad.lock, MUTEX_DEFAULT, IPL_NONE);
   8021 	cv_init(&key_sad.cv_lc, "key_sa_lc");
   8022 	key_sad.psz = pserialize_create();
   8023 	cv_init(&key_sad.cv_psz, "key_sa_psz");
   8024 	key_sad.psz_performing = false;
   8025 
   8026 	pfkeystat_percpu = percpu_alloc(sizeof(uint64_t) * PFKEY_NSTATS);
   8027 
   8028 	callout_init(&key_timehandler_ch, CALLOUT_MPSAFE);
   8029 	error = workqueue_create(&key_timehandler_wq, "key_timehandler",
   8030 	    key_timehandler_work, NULL, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
   8031 	if (error != 0)
   8032 		panic("%s: workqueue_create failed (%d)\n", __func__, error);
   8033 
   8034 	for (i = 0; i < IPSEC_DIR_MAX; i++) {
   8035 		PSLIST_INIT(&key_spd.splist[i]);
   8036 	}
   8037 
   8038 	PSLIST_INIT(&key_spd.socksplist);
   8039 
   8040 	PSLIST_INIT(&key_sad.sahlist);
   8041 
   8042 	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
   8043 		LIST_INIT(&key_misc.reglist[i]);
   8044 	}
   8045 
   8046 #ifndef IPSEC_NONBLOCK_ACQUIRE
   8047 	LIST_INIT(&key_misc.acqlist);
   8048 #endif
   8049 #ifdef notyet
   8050 	LIST_INIT(&key_misc.spacqlist);
   8051 #endif
   8052 
   8053 	/* system default */
   8054 	ip4_def_policy.policy = IPSEC_POLICY_NONE;
   8055 	ip4_def_policy.state = IPSEC_SPSTATE_ALIVE;
   8056 	localcount_init(&ip4_def_policy.localcount);
   8057 
   8058 #ifdef INET6
   8059 	ip6_def_policy.policy = IPSEC_POLICY_NONE;
   8060 	ip6_def_policy.state = IPSEC_SPSTATE_ALIVE;
   8061 	localcount_init(&ip6_def_policy.localcount);
   8062 #endif
   8063 
   8064 	callout_reset(&key_timehandler_ch, hz, key_timehandler, NULL);
   8065 
   8066 	/* initialize key statistics */
   8067 	keystat.getspi_count = 1;
   8068 
   8069 	aprint_verbose("IPsec: Initialized Security Association Processing.\n");
   8070 
   8071 	return (0);
   8072 }
   8073 
   8074 void
   8075 key_init(void)
   8076 {
   8077 	static ONCE_DECL(key_init_once);
   8078 
   8079 	sysctl_net_keyv2_setup(NULL);
   8080 	sysctl_net_key_compat_setup(NULL);
   8081 
   8082 	RUN_ONCE(&key_init_once, key_do_init);
   8083 
   8084 	key_init_so();
   8085 }
   8086 
   8087 /*
   8088  * XXX: maybe This function is called after INBOUND IPsec processing.
   8089  *
   8090  * Special check for tunnel-mode packets.
   8091  * We must make some checks for consistency between inner and outer IP header.
   8092  *
   8093  * xxx more checks to be provided
   8094  */
   8095 int
   8096 key_checktunnelsanity(
   8097     struct secasvar *sav,
   8098     u_int family,
   8099     void *src,
   8100     void *dst
   8101 )
   8102 {
   8103 
   8104 	/* XXX: check inner IP header */
   8105 
   8106 	return 1;
   8107 }
   8108 
   8109 #if 0
   8110 #define hostnamelen	strlen(hostname)
   8111 
   8112 /*
   8113  * Get FQDN for the host.
   8114  * If the administrator configured hostname (by hostname(1)) without
   8115  * domain name, returns nothing.
   8116  */
   8117 static const char *
   8118 key_getfqdn(void)
   8119 {
   8120 	int i;
   8121 	int hasdot;
   8122 	static char fqdn[MAXHOSTNAMELEN + 1];
   8123 
   8124 	if (!hostnamelen)
   8125 		return NULL;
   8126 
   8127 	/* check if it comes with domain name. */
   8128 	hasdot = 0;
   8129 	for (i = 0; i < hostnamelen; i++) {
   8130 		if (hostname[i] == '.')
   8131 			hasdot++;
   8132 	}
   8133 	if (!hasdot)
   8134 		return NULL;
   8135 
   8136 	/* NOTE: hostname may not be NUL-terminated. */
   8137 	memset(fqdn, 0, sizeof(fqdn));
   8138 	memcpy(fqdn, hostname, hostnamelen);
   8139 	fqdn[hostnamelen] = '\0';
   8140 	return fqdn;
   8141 }
   8142 
   8143 /*
   8144  * get username@FQDN for the host/user.
   8145  */
   8146 static const char *
   8147 key_getuserfqdn(void)
   8148 {
   8149 	const char *host;
   8150 	static char userfqdn[MAXHOSTNAMELEN + MAXLOGNAME + 2];
   8151 	struct proc *p = curproc;
   8152 	char *q;
   8153 
   8154 	if (!p || !p->p_pgrp || !p->p_pgrp->pg_session)
   8155 		return NULL;
   8156 	if (!(host = key_getfqdn()))
   8157 		return NULL;
   8158 
   8159 	/* NOTE: s_login may not be-NUL terminated. */
   8160 	memset(userfqdn, 0, sizeof(userfqdn));
   8161 	memcpy(userfqdn, Mp->p_pgrp->pg_session->s_login, AXLOGNAME);
   8162 	userfqdn[MAXLOGNAME] = '\0';	/* safeguard */
   8163 	q = userfqdn + strlen(userfqdn);
   8164 	*q++ = '@';
   8165 	memcpy(q, host, strlen(host));
   8166 	q += strlen(host);
   8167 	*q++ = '\0';
   8168 
   8169 	return userfqdn;
   8170 }
   8171 #endif
   8172 
   8173 /* record data transfer on SA, and update timestamps */
   8174 void
   8175 key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
   8176 {
   8177 
   8178 	KASSERT(sav != NULL);
   8179 	KASSERT(sav->lft_c != NULL);
   8180 	KASSERT(m != NULL);
   8181 
   8182 	/*
   8183 	 * XXX Currently, there is a difference of bytes size
   8184 	 * between inbound and outbound processing.
   8185 	 */
   8186 	sav->lft_c->sadb_lifetime_bytes += m->m_pkthdr.len;
   8187 	/* to check bytes lifetime is done in key_timehandler(). */
   8188 
   8189 	/*
   8190 	 * We use the number of packets as the unit of
   8191 	 * sadb_lifetime_allocations.  We increment the variable
   8192 	 * whenever {esp,ah}_{in,out}put is called.
   8193 	 */
   8194 	sav->lft_c->sadb_lifetime_allocations++;
   8195 	/* XXX check for expires? */
   8196 
   8197 	/*
   8198 	 * NOTE: We record CURRENT sadb_lifetime_usetime by using wall clock,
   8199 	 * in seconds.  HARD and SOFT lifetime are measured by the time
   8200 	 * difference (again in seconds) from sadb_lifetime_usetime.
   8201 	 *
   8202 	 *	usetime
   8203 	 *	v     expire   expire
   8204 	 * -----+-----+--------+---> t
   8205 	 *	<--------------> HARD
   8206 	 *	<-----> SOFT
   8207 	 */
   8208 	sav->lft_c->sadb_lifetime_usetime = time_uptime;
   8209 	/* XXX check for expires? */
   8210 
   8211 	return;
   8212 }
   8213 
   8214 /* dumb version */
   8215 void
   8216 key_sa_routechange(struct sockaddr *dst)
   8217 {
   8218 	struct secashead *sah;
   8219 	int s;
   8220 
   8221 	s = pserialize_read_enter();
   8222 	SAHLIST_READER_FOREACH(sah) {
   8223 		struct route *ro;
   8224 		const struct sockaddr *sa;
   8225 
   8226 		key_sah_ref(sah);
   8227 		pserialize_read_exit(s);
   8228 
   8229 		ro = &sah->sa_route;
   8230 		sa = rtcache_getdst(ro);
   8231 		if (sa != NULL && dst->sa_len == sa->sa_len &&
   8232 		    memcmp(dst, sa, dst->sa_len) == 0)
   8233 			rtcache_free(ro);
   8234 
   8235 		s = pserialize_read_enter();
   8236 		key_sah_unref(sah);
   8237 	}
   8238 	pserialize_read_exit(s);
   8239 
   8240 	return;
   8241 }
   8242 
   8243 static void
   8244 key_sa_chgstate(struct secasvar *sav, u_int8_t state)
   8245 {
   8246 	struct secasvar *_sav;
   8247 
   8248 	ASSERT_SLEEPABLE();
   8249 	KASSERT(mutex_owned(&key_sad.lock));
   8250 
   8251 	if (sav->state == state)
   8252 		return;
   8253 
   8254 	key_unlink_sav(sav);
   8255 	localcount_fini(&sav->localcount);
   8256 	SAVLIST_ENTRY_DESTROY(sav);
   8257 	key_init_sav(sav);
   8258 
   8259 	sav->state = state;
   8260 	if (!SADB_SASTATE_USABLE_P(sav)) {
   8261 		/* We don't need to care about the order */
   8262 		SAVLIST_WRITER_INSERT_HEAD(sav->sah, state, sav);
   8263 		return;
   8264 	}
   8265 	/*
   8266 	 * Sort the list by lft_c->sadb_lifetime_addtime
   8267 	 * in ascending order.
   8268 	 */
   8269 	SAVLIST_WRITER_FOREACH(_sav, sav->sah, state) {
   8270 		if (_sav->lft_c->sadb_lifetime_addtime >
   8271 		    sav->lft_c->sadb_lifetime_addtime) {
   8272 			SAVLIST_WRITER_INSERT_BEFORE(_sav, sav);
   8273 			break;
   8274 		}
   8275 	}
   8276 	if (_sav == NULL) {
   8277 		SAVLIST_WRITER_INSERT_TAIL(sav->sah, state, sav);
   8278 	}
   8279 	key_validate_savlist(sav->sah, state);
   8280 }
   8281 
   8282 /* XXX too much? */
   8283 static struct mbuf *
   8284 key_alloc_mbuf(int l, int mflag)
   8285 {
   8286 	struct mbuf *m = NULL, *n;
   8287 	int len, t;
   8288 
   8289 	KASSERT(mflag == M_NOWAIT || (mflag == M_WAITOK && !cpu_softintr_p()));
   8290 
   8291 	len = l;
   8292 	while (len > 0) {
   8293 		MGET(n, mflag, MT_DATA);
   8294 		if (n && len > MLEN) {
   8295 			MCLGET(n, mflag);
   8296 			if ((n->m_flags & M_EXT) == 0) {
   8297 				m_freem(n);
   8298 				n = NULL;
   8299 			}
   8300 		}
   8301 		if (!n) {
   8302 			m_freem(m);
   8303 			return NULL;
   8304 		}
   8305 
   8306 		n->m_next = NULL;
   8307 		n->m_len = 0;
   8308 		n->m_len = M_TRAILINGSPACE(n);
   8309 		/* use the bottom of mbuf, hoping we can prepend afterwards */
   8310 		if (n->m_len > len) {
   8311 			t = (n->m_len - len) & ~(sizeof(long) - 1);
   8312 			n->m_data += t;
   8313 			n->m_len = len;
   8314 		}
   8315 
   8316 		len -= n->m_len;
   8317 
   8318 		if (m)
   8319 			m_cat(m, n);
   8320 		else
   8321 			m = n;
   8322 	}
   8323 
   8324 	return m;
   8325 }
   8326 
   8327 static struct mbuf *
   8328 key_setdump(u_int8_t req_satype, int *errorp, uint32_t pid)
   8329 {
   8330 	struct secashead *sah;
   8331 	struct secasvar *sav;
   8332 	u_int16_t proto;
   8333 	u_int8_t satype;
   8334 	u_int8_t state;
   8335 	int cnt;
   8336 	struct mbuf *m, *n;
   8337 
   8338 	KASSERT(mutex_owned(&key_sad.lock));
   8339 
   8340 	/* map satype to proto */
   8341 	proto = key_satype2proto(req_satype);
   8342 	if (proto == 0) {
   8343 		*errorp = EINVAL;
   8344 		return (NULL);
   8345 	}
   8346 
   8347 	/* count sav entries to be sent to the userland. */
   8348 	cnt = 0;
   8349 	SAHLIST_WRITER_FOREACH(sah) {
   8350 		if (req_satype != SADB_SATYPE_UNSPEC &&
   8351 		    proto != sah->saidx.proto)
   8352 			continue;
   8353 
   8354 		SASTATE_ANY_FOREACH(state) {
   8355 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
   8356 				cnt++;
   8357 			}
   8358 		}
   8359 	}
   8360 
   8361 	if (cnt == 0) {
   8362 		*errorp = ENOENT;
   8363 		return (NULL);
   8364 	}
   8365 
   8366 	/* send this to the userland, one at a time. */
   8367 	m = NULL;
   8368 	SAHLIST_WRITER_FOREACH(sah) {
   8369 		if (req_satype != SADB_SATYPE_UNSPEC &&
   8370 		    proto != sah->saidx.proto)
   8371 			continue;
   8372 
   8373 		/* map proto to satype */
   8374 		satype = key_proto2satype(sah->saidx.proto);
   8375 		if (satype == 0) {
   8376 			m_freem(m);
   8377 			*errorp = EINVAL;
   8378 			return (NULL);
   8379 		}
   8380 
   8381 		SASTATE_ANY_FOREACH(state) {
   8382 			SAVLIST_WRITER_FOREACH(sav, sah, state) {
   8383 				n = key_setdumpsa(sav, SADB_DUMP, satype,
   8384 				    --cnt, pid);
   8385 				if (!m)
   8386 					m = n;
   8387 				else
   8388 					m_cat(m, n);
   8389 			}
   8390 		}
   8391 	}
   8392 
   8393 	if (!m) {
   8394 		*errorp = EINVAL;
   8395 		return (NULL);
   8396 	}
   8397 
   8398 	if ((m->m_flags & M_PKTHDR) != 0) {
   8399 		m->m_pkthdr.len = 0;
   8400 		for (n = m; n; n = n->m_next)
   8401 			m->m_pkthdr.len += n->m_len;
   8402 	}
   8403 
   8404 	*errorp = 0;
   8405 	return (m);
   8406 }
   8407 
   8408 static struct mbuf *
   8409 key_setspddump(int *errorp, pid_t pid)
   8410 {
   8411 	struct secpolicy *sp;
   8412 	int cnt;
   8413 	u_int dir;
   8414 	struct mbuf *m, *n;
   8415 
   8416 	KASSERT(mutex_owned(&key_spd.lock));
   8417 
   8418 	/* search SPD entry and get buffer size. */
   8419 	cnt = 0;
   8420 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
   8421 		SPLIST_WRITER_FOREACH(sp, dir) {
   8422 			cnt++;
   8423 		}
   8424 	}
   8425 
   8426 	if (cnt == 0) {
   8427 		*errorp = ENOENT;
   8428 		return (NULL);
   8429 	}
   8430 
   8431 	m = NULL;
   8432 	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
   8433 		SPLIST_WRITER_FOREACH(sp, dir) {
   8434 			--cnt;
   8435 			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, pid);
   8436 
   8437 			if (!m)
   8438 				m = n;
   8439 			else {
   8440 				m->m_pkthdr.len += n->m_pkthdr.len;
   8441 				m_cat(m, n);
   8442 			}
   8443 		}
   8444 	}
   8445 
   8446 	*errorp = 0;
   8447 	return (m);
   8448 }
   8449 
   8450 int
   8451 key_get_used(void) {
   8452 	return !SPLIST_READER_EMPTY(IPSEC_DIR_INBOUND) ||
   8453 	    !SPLIST_READER_EMPTY(IPSEC_DIR_OUTBOUND) ||
   8454 	    !SOCKSPLIST_READER_EMPTY();
   8455 }
   8456 
   8457 void
   8458 key_update_used(void)
   8459 {
   8460 	switch (ipsec_enabled) {
   8461 	default:
   8462 	case 0:
   8463 #ifdef notyet
   8464 		/* XXX: racy */
   8465 		ipsec_used = 0;
   8466 #endif
   8467 		break;
   8468 	case 1:
   8469 #ifndef notyet
   8470 		/* XXX: racy */
   8471 		if (!ipsec_used)
   8472 #endif
   8473 		ipsec_used = key_get_used();
   8474 		break;
   8475 	case 2:
   8476 		ipsec_used = 1;
   8477 		break;
   8478 	}
   8479 }
   8480 
   8481 static int
   8482 sysctl_net_key_dumpsa(SYSCTLFN_ARGS)
   8483 {
   8484 	struct mbuf *m, *n;
   8485 	int err2 = 0;
   8486 	char *p, *ep;
   8487 	size_t len;
   8488 	int error;
   8489 
   8490 	if (newp)
   8491 		return (EPERM);
   8492 	if (namelen != 1)
   8493 		return (EINVAL);
   8494 
   8495 	mutex_enter(&key_sad.lock);
   8496 	m = key_setdump(name[0], &error, l->l_proc->p_pid);
   8497 	mutex_exit(&key_sad.lock);
   8498 	if (!m)
   8499 		return (error);
   8500 	if (!oldp)
   8501 		*oldlenp = m->m_pkthdr.len;
   8502 	else {
   8503 		p = oldp;
   8504 		if (*oldlenp < m->m_pkthdr.len) {
   8505 			err2 = ENOMEM;
   8506 			ep = p + *oldlenp;
   8507 		} else {
   8508 			*oldlenp = m->m_pkthdr.len;
   8509 			ep = p + m->m_pkthdr.len;
   8510 		}
   8511 		for (n = m; n; n = n->m_next) {
   8512 			len =  (ep - p < n->m_len) ?
   8513 				ep - p : n->m_len;
   8514 			error = copyout(mtod(n, const void *), p, len);
   8515 			p += len;
   8516 			if (error)
   8517 				break;
   8518 		}
   8519 		if (error == 0)
   8520 			error = err2;
   8521 	}
   8522 	m_freem(m);
   8523 
   8524 	return (error);
   8525 }
   8526 
   8527 static int
   8528 sysctl_net_key_dumpsp(SYSCTLFN_ARGS)
   8529 {
   8530 	struct mbuf *m, *n;
   8531 	int err2 = 0;
   8532 	char *p, *ep;
   8533 	size_t len;
   8534 	int error;
   8535 
   8536 	if (newp)
   8537 		return (EPERM);
   8538 	if (namelen != 0)
   8539 		return (EINVAL);
   8540 
   8541 	mutex_enter(&key_spd.lock);
   8542 	m = key_setspddump(&error, l->l_proc->p_pid);
   8543 	mutex_exit(&key_spd.lock);
   8544 	if (!m)
   8545 		return (error);
   8546 	if (!oldp)
   8547 		*oldlenp = m->m_pkthdr.len;
   8548 	else {
   8549 		p = oldp;
   8550 		if (*oldlenp < m->m_pkthdr.len) {
   8551 			err2 = ENOMEM;
   8552 			ep = p + *oldlenp;
   8553 		} else {
   8554 			*oldlenp = m->m_pkthdr.len;
   8555 			ep = p + m->m_pkthdr.len;
   8556 		}
   8557 		for (n = m; n; n = n->m_next) {
   8558 			len = (ep - p < n->m_len) ? ep - p : n->m_len;
   8559 			error = copyout(mtod(n, const void *), p, len);
   8560 			p += len;
   8561 			if (error)
   8562 				break;
   8563 		}
   8564 		if (error == 0)
   8565 			error = err2;
   8566 	}
   8567 	m_freem(m);
   8568 
   8569 	return (error);
   8570 }
   8571 
   8572 /*
   8573  * Create sysctl tree for native IPSEC key knobs, originally
   8574  * under name "net.keyv2"  * with MIB number { CTL_NET, PF_KEY_V2. }.
   8575  * However, sysctl(8) never checked for nodes under { CTL_NET, PF_KEY_V2 };
   8576  * and in any case the part of our sysctl namespace used for dumping the
   8577  * SPD and SA database  *HAS* to be compatible with the KAME sysctl
   8578  * namespace, for API reasons.
   8579  *
   8580  * Pending a consensus on the right way  to fix this, add a level of
   8581  * indirection in how we number the `native' IPSEC key nodes;
   8582  * and (as requested by Andrew Brown)  move registration of the
   8583  * KAME-compatible names  to a separate function.
   8584  */
   8585 #if 0
   8586 #  define IPSEC_PFKEY PF_KEY_V2
   8587 # define IPSEC_PFKEY_NAME "keyv2"
   8588 #else
   8589 #  define IPSEC_PFKEY PF_KEY
   8590 # define IPSEC_PFKEY_NAME "key"
   8591 #endif
   8592 
   8593 static int
   8594 sysctl_net_key_stats(SYSCTLFN_ARGS)
   8595 {
   8596 
   8597 	return (NETSTAT_SYSCTL(pfkeystat_percpu, PFKEY_NSTATS));
   8598 }
   8599 
   8600 static void
   8601 sysctl_net_keyv2_setup(struct sysctllog **clog)
   8602 {
   8603 
   8604 	sysctl_createv(clog, 0, NULL, NULL,
   8605 		       CTLFLAG_PERMANENT,
   8606 		       CTLTYPE_NODE, IPSEC_PFKEY_NAME, NULL,
   8607 		       NULL, 0, NULL, 0,
   8608 		       CTL_NET, IPSEC_PFKEY, CTL_EOL);
   8609 
   8610 	sysctl_createv(clog, 0, NULL, NULL,
   8611 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8612 		       CTLTYPE_INT, "debug", NULL,
   8613 		       NULL, 0, &key_debug_level, 0,
   8614 		       CTL_NET, IPSEC_PFKEY, KEYCTL_DEBUG_LEVEL, CTL_EOL);
   8615 	sysctl_createv(clog, 0, NULL, NULL,
   8616 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8617 		       CTLTYPE_INT, "spi_try", NULL,
   8618 		       NULL, 0, &key_spi_trycnt, 0,
   8619 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_TRY, CTL_EOL);
   8620 	sysctl_createv(clog, 0, NULL, NULL,
   8621 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8622 		       CTLTYPE_INT, "spi_min_value", NULL,
   8623 		       NULL, 0, &key_spi_minval, 0,
   8624 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_MIN_VALUE, CTL_EOL);
   8625 	sysctl_createv(clog, 0, NULL, NULL,
   8626 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8627 		       CTLTYPE_INT, "spi_max_value", NULL,
   8628 		       NULL, 0, &key_spi_maxval, 0,
   8629 		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_MAX_VALUE, CTL_EOL);
   8630 	sysctl_createv(clog, 0, NULL, NULL,
   8631 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8632 		       CTLTYPE_INT, "random_int", NULL,
   8633 		       NULL, 0, &key_int_random, 0,
   8634 		       CTL_NET, IPSEC_PFKEY, KEYCTL_RANDOM_INT, CTL_EOL);
   8635 	sysctl_createv(clog, 0, NULL, NULL,
   8636 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8637 		       CTLTYPE_INT, "larval_lifetime", NULL,
   8638 		       NULL, 0, &key_larval_lifetime, 0,
   8639 		       CTL_NET, IPSEC_PFKEY, KEYCTL_LARVAL_LIFETIME, CTL_EOL);
   8640 	sysctl_createv(clog, 0, NULL, NULL,
   8641 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8642 		       CTLTYPE_INT, "blockacq_count", NULL,
   8643 		       NULL, 0, &key_blockacq_count, 0,
   8644 		       CTL_NET, IPSEC_PFKEY, KEYCTL_BLOCKACQ_COUNT, CTL_EOL);
   8645 	sysctl_createv(clog, 0, NULL, NULL,
   8646 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8647 		       CTLTYPE_INT, "blockacq_lifetime", NULL,
   8648 		       NULL, 0, &key_blockacq_lifetime, 0,
   8649 		       CTL_NET, IPSEC_PFKEY, KEYCTL_BLOCKACQ_LIFETIME, CTL_EOL);
   8650 	sysctl_createv(clog, 0, NULL, NULL,
   8651 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8652 		       CTLTYPE_INT, "esp_keymin", NULL,
   8653 		       NULL, 0, &ipsec_esp_keymin, 0,
   8654 		       CTL_NET, IPSEC_PFKEY, KEYCTL_ESP_KEYMIN, CTL_EOL);
   8655 	sysctl_createv(clog, 0, NULL, NULL,
   8656 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8657 		       CTLTYPE_INT, "prefered_oldsa", NULL,
   8658 		       NULL, 0, &key_prefered_oldsa, 0,
   8659 		       CTL_NET, PF_KEY, KEYCTL_PREFERED_OLDSA, CTL_EOL);
   8660 	sysctl_createv(clog, 0, NULL, NULL,
   8661 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8662 		       CTLTYPE_INT, "esp_auth", NULL,
   8663 		       NULL, 0, &ipsec_esp_auth, 0,
   8664 		       CTL_NET, IPSEC_PFKEY, KEYCTL_ESP_AUTH, CTL_EOL);
   8665 	sysctl_createv(clog, 0, NULL, NULL,
   8666 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
   8667 		       CTLTYPE_INT, "ah_keymin", NULL,
   8668 		       NULL, 0, &ipsec_ah_keymin, 0,
   8669 		       CTL_NET, IPSEC_PFKEY, KEYCTL_AH_KEYMIN, CTL_EOL);
   8670 	sysctl_createv(clog, 0, NULL, NULL,
   8671 		       CTLFLAG_PERMANENT,
   8672 		       CTLTYPE_STRUCT, "stats",
   8673 		       SYSCTL_DESCR("PF_KEY statistics"),
   8674 		       sysctl_net_key_stats, 0, NULL, 0,
   8675 		       CTL_NET, IPSEC_PFKEY, CTL_CREATE, CTL_EOL);
   8676 }
   8677 
   8678 /*
   8679  * Register sysctl names used by setkey(8). For historical reasons,
   8680  * and to share a single API, these names appear under { CTL_NET, PF_KEY }
   8681  * for both IPSEC and KAME IPSEC.
   8682  */
   8683 static void
   8684 sysctl_net_key_compat_setup(struct sysctllog **clog)
   8685 {
   8686 
   8687 	sysctl_createv(clog, 0, NULL, NULL,
   8688 		       CTLFLAG_PERMANENT,
   8689 		       CTLTYPE_NODE, "key", NULL,
   8690 		       NULL, 0, NULL, 0,
   8691 		       CTL_NET, PF_KEY, CTL_EOL);
   8692 
   8693 	/* Register the net.key.dump{sa,sp} nodes used by setkey(8). */
   8694 	sysctl_createv(clog, 0, NULL, NULL,
   8695 		       CTLFLAG_PERMANENT,
   8696 		       CTLTYPE_STRUCT, "dumpsa", NULL,
   8697 		       sysctl_net_key_dumpsa, 0, NULL, 0,
   8698 		       CTL_NET, PF_KEY, KEYCTL_DUMPSA, CTL_EOL);
   8699 	sysctl_createv(clog, 0, NULL, NULL,
   8700 		       CTLFLAG_PERMANENT,
   8701 		       CTLTYPE_STRUCT, "dumpsp", NULL,
   8702 		       sysctl_net_key_dumpsp, 0, NULL, 0,
   8703 		       CTL_NET, PF_KEY, KEYCTL_DUMPSP, CTL_EOL);
   8704 }
   8705