1/*	$NetBSD: key.c,v 1.282 2023/08/10 06:44:12 andvar Exp $	*/
2/*	$FreeBSD: key.c,v 1.3.2.3 2004/02/14 22:23:23 bms Exp $	*/
3/*	$KAME: key.c,v 1.191 2001/06/27 10:46:49 sakane Exp $	*/
4
5/*
6 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 *    notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 *    notice, this list of conditions and the following disclaimer in the
16 *    documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 */
33
34#include <sys/cdefs.h>
35__KERNEL_RCSID(0, "$NetBSD: key.c,v 1.282 2023/08/10 06:44:12 andvar Exp $");
36
37/*
38 * This code is referred to RFC 2367
39 */
40
41#if defined(_KERNEL_OPT)
42#include "opt_inet.h"
43#include "opt_ipsec.h"
44#include "opt_gateway.h"
45#include "opt_net_mpsafe.h"
46#endif
47
48#include <sys/types.h>
49#include <sys/param.h>
50#include <sys/systm.h>
51#include <sys/callout.h>
52#include <sys/kernel.h>
53#include <sys/mbuf.h>
54#include <sys/domain.h>
55#include <sys/socket.h>
56#include <sys/socketvar.h>
57#include <sys/sysctl.h>
58#include <sys/errno.h>
59#include <sys/proc.h>
60#include <sys/queue.h>
61#include <sys/syslog.h>
62#include <sys/once.h>
63#include <sys/cprng.h>
64#include <sys/psref.h>
65#include <sys/lwp.h>
66#include <sys/workqueue.h>
67#include <sys/kmem.h>
68#include <sys/cpu.h>
69#include <sys/atomic.h>
70#include <sys/pslist.h>
71#include <sys/mutex.h>
72#include <sys/condvar.h>
73#include <sys/localcount.h>
74#include <sys/pserialize.h>
75#include <sys/hash.h>
76#include <sys/xcall.h>
77
78#include <net/if.h>
79#include <net/route.h>
80
81#include <netinet/in.h>
82#include <netinet/in_systm.h>
83#include <netinet/ip.h>
84#include <netinet/in_var.h>
85#ifdef INET
86#include <netinet/ip_var.h>
87#endif
88
89#ifdef INET6
90#include <netinet/ip6.h>
91#include <netinet6/in6_var.h>
92#include <netinet6/ip6_var.h>
93#endif /* INET6 */
94
95#ifdef INET
96#include <netinet/in_pcb.h>
97#endif
98#ifdef INET6
99#include <netinet6/in6_pcb.h>
100#endif /* INET6 */
101
102#include <net/pfkeyv2.h>
103#include <netipsec/keydb.h>
104#include <netipsec/key.h>
105#include <netipsec/keysock.h>
106#include <netipsec/key_debug.h>
107
108#include <netipsec/ipsec.h>
109#ifdef INET6
110#include <netipsec/ipsec6.h>
111#endif
112#include <netipsec/ipsec_private.h>
113
114#include <netipsec/xform.h>
115#include <netipsec/ipcomp.h>
116
117#define FULLMASK	0xffu
118#define	_BITS(bytes)	((bytes) << 3)
119
120#define PORT_NONE	0
121#define PORT_LOOSE	1
122#define PORT_STRICT	2
123
124#ifndef SAHHASH_NHASH
125#define SAHHASH_NHASH		128
126#endif
127
128#ifndef SAVLUT_NHASH
129#define SAVLUT_NHASH		128
130#endif
131
132percpu_t *pfkeystat_percpu;
133
134/*
135 * Note on SA reference counting:
136 * - SAs that are not in DEAD state will have (total external reference + 1)
137 *   following value in reference count field.  they cannot be freed and are
138 *   referenced from SA header.
139 * - SAs that are in DEAD state will have (total external reference)
140 *   in reference count field.  they are ready to be freed.  reference from
141 *   SA header will be removed in key_delsav(), when the reference count
142 *   field hits 0 (= no external reference other than from SA header.
143 */
144
145u_int32_t key_debug_level = 0;
146static u_int key_spi_trycnt = 1000;
147static u_int32_t key_spi_minval = 0x100;
148static u_int32_t key_spi_maxval = 0x0fffffff;	/* XXX */
149static u_int32_t policy_id = 0;
150static u_int key_int_random = 60;	/*interval to initialize randseed,1(m)*/
151static u_int key_larval_lifetime = 30;	/* interval to expire acquiring, 30(s)*/
152static int key_blockacq_count = 10;	/* counter for blocking SADB_ACQUIRE.*/
153static int key_blockacq_lifetime = 20;	/* lifetime for blocking SADB_ACQUIRE.*/
154static int key_prefered_oldsa = 0;	/* prefered old sa rather than new sa.*/
155
156static u_int32_t acq_seq = 0;
157
158/*
159 * Locking order: there is no order for now; it means that any locks aren't
160 * overlapped.
161 */
162/*
163 * Locking notes on SPD:
164 * - Modifications to the key_spd.splist must be done with holding key_spd.lock
165 *   which is a adaptive mutex
166 * - Read accesses to the key_spd.splist must be in pserialize(9) read sections
167 * - SP's lifetime is managed by localcount(9)
168 * - An SP that has been inserted to the key_spd.splist is initially referenced
169 *   by none, i.e., a reference from the key_spd.splist isn't counted
170 * - When an SP is being destroyed, we change its state as DEAD, wait for
171 *   references to the SP to be released, and then deallocate the SP
172 *   (see key_unlink_sp)
173 * - Getting an SP
174 *   - Normally we get an SP from the key_spd.splist (see key_lookup_sp_byspidx)
175 *     - Must iterate the list and increment the reference count of a found SP
176 *       (by key_sp_ref) in a pserialize read section
177 *   - We can gain another reference from a held SP only if we check its state
178 *     and take its reference in a pserialize read section
179 *     (see esp_output for example)
180 *   - We may get an SP from an SP cache. See below
181 *   - A gotten SP must be released after use by KEY_SP_UNREF (key_sp_unref)
182 * - Updating member variables of an SP
183 *   - Most member variables of an SP are immutable
184 *   - Only sp->state and sp->lastused can be changed
185 *   - sp->state of an SP is updated only when destroying it under key_spd.lock
186 * - SP caches
187 *   - SPs can be cached in PCBs
188 *   - The lifetime of the caches is controlled by the global generation counter
189 *     (ipsec_spdgen)
190 *   - The global counter value is stored when an SP is cached
191 *   - If the stored value is different from the global counter then the cache
192 *     is considered invalidated
193 *   - The counter is incremented when an SP is being destroyed
194 *   - So checking the generation and taking a reference to an SP should be
195 *     in a pserialize read section
196 *   - Note that caching doesn't increment the reference counter of an SP
197 * - SPs in sockets
198 *   - Userland programs can set a policy to a socket by
199 *     setsockopt(IP_IPSEC_POLICY)
200 *   - Such policies (SPs) are set to a socket (PCB) and also inserted to
201 *     the key_spd.socksplist list (not the key_spd.splist)
202 *   - Such a policy is destroyed when a corresponding socket is destroed,
203 *     however, a socket can be destroyed in softint so we cannot destroy
204 *     it directly instead we just mark it DEAD and delay the destruction
205 *     until GC by the timer
206 * - SP origin
207 *   - SPs can be created by both userland programs and kernel components.
208 *     The SPs created in kernel must not be removed by userland programs,
209 *     although the SPs can be read by userland programs.
210 */
211/*
212 * Locking notes on SAD:
213 * - Data structures
214 *   - SAs are managed by the list called key_sad.sahlists and sav lists of
215 *     sah entries
216 *     - An sav is supposed to be an SA from a viewpoint of users
217 *   - A sah has sav lists for each SA state
218 *   - Multiple saves with the same saidx can exist
219 *     - Only one entry has MATURE state and others should be DEAD
220 *     - DEAD entries are just ignored from searching
221 *   - All sav whose state is MATURE or DYING are registered to the lookup
222 *     table called key_sad.savlut in addition to the savlists.
223 *     - The table is used to search an sav without use of saidx.
224 * - Modifications to the key_sad.sahlists, sah.savlist and key_sad.savlut
225 *   must be done with holding key_sad.lock which is a adaptive mutex
226 * - Read accesses to the key_sad.sahlists, sah.savlist and key_sad.savlut
227 *   must be in pserialize(9) read sections
228 * - sah's lifetime is managed by localcount(9)
229 * - Getting an sah entry
230 *   - We get an sah from the key_sad.sahlists
231 *     - Must iterate the list and increment the reference count of a found sah
232 *       (by key_sah_ref) in a pserialize read section
233 *   - A gotten sah must be released after use by key_sah_unref
234 * - An sah is destroyed when its state become DEAD and no sav is
235 *   listed to the sah
236 *   - The destruction is done only in the timer (see key_timehandler_sad)
237 * - sav's lifetime is managed by localcount(9)
238 * - Getting an sav entry
239 *   - First get an sah by saidx and get an sav from either of sah's savlists
240 *     - Must iterate the list and increment the reference count of a found sav
241 *       (by key_sa_ref) in a pserialize read section
242 *   - We can gain another reference from a held SA only if we check its state
243 *     and take its reference in a pserialize read section
244 *     (see esp_output for example)
245 *   - A gotten sav must be released after use by key_sa_unref
246 * - An sav is destroyed when its state become DEAD
247 */
248/*
249 * Locking notes on misc data:
250 * - All lists of key_misc are protected by key_misc.lock
251 *   - key_misc.lock must be held even for read accesses
252 */
253
254/* SPD */
255static struct {
256	kmutex_t lock;
257	kcondvar_t cv_lc;
258	struct pslist_head splist[IPSEC_DIR_MAX];
259	/*
260	 * The list has SPs that are set to a socket via
261	 * setsockopt(IP_IPSEC_POLICY) from userland. See ipsec_set_policy.
262	 */
263	struct pslist_head socksplist;
264
265	pserialize_t psz;
266	kcondvar_t cv_psz;
267	bool psz_performing;
268} key_spd __cacheline_aligned;
269
270/* SAD */
271static struct {
272	kmutex_t lock;
273	kcondvar_t cv_lc;
274	struct pslist_head *sahlists;
275	u_long sahlistmask;
276	struct pslist_head *savlut;
277	u_long savlutmask;
278
279	pserialize_t psz;
280	kcondvar_t cv_psz;
281	bool psz_performing;
282} key_sad __cacheline_aligned;
283
284/* Misc data */
285static struct {
286	kmutex_t lock;
287	/* registed list */
288	LIST_HEAD(_reglist, secreg) reglist[SADB_SATYPE_MAX + 1];
289#ifndef IPSEC_NONBLOCK_ACQUIRE
290	/* acquiring list */
291	LIST_HEAD(_acqlist, secacq) acqlist;
292#endif
293#ifdef notyet
294	/* SP acquiring list */
295	LIST_HEAD(_spacqlist, secspacq) spacqlist;
296#endif
297} key_misc __cacheline_aligned;
298
299/* Macros for key_spd.splist */
300#define SPLIST_ENTRY_INIT(sp)						\
301	PSLIST_ENTRY_INIT((sp), pslist_entry)
302#define SPLIST_ENTRY_DESTROY(sp)					\
303	PSLIST_ENTRY_DESTROY((sp), pslist_entry)
304#define SPLIST_WRITER_REMOVE(sp)					\
305	PSLIST_WRITER_REMOVE((sp), pslist_entry)
306#define SPLIST_READER_EMPTY(dir)					\
307	(PSLIST_READER_FIRST(&key_spd.splist[(dir)], struct secpolicy,	\
308	                     pslist_entry) == NULL)
309#define SPLIST_READER_FOREACH(sp, dir)					\
310	PSLIST_READER_FOREACH((sp), &key_spd.splist[(dir)],		\
311	                      struct secpolicy, pslist_entry)
312#define SPLIST_WRITER_FOREACH(sp, dir)					\
313	PSLIST_WRITER_FOREACH((sp), &key_spd.splist[(dir)],		\
314	                      struct secpolicy, pslist_entry)
315#define SPLIST_WRITER_INSERT_AFTER(sp, new)				\
316	PSLIST_WRITER_INSERT_AFTER((sp), (new), pslist_entry)
317#define SPLIST_WRITER_EMPTY(dir)					\
318	(PSLIST_WRITER_FIRST(&key_spd.splist[(dir)], struct secpolicy,	\
319	                     pslist_entry) == NULL)
320#define SPLIST_WRITER_INSERT_HEAD(dir, sp)				\
321	PSLIST_WRITER_INSERT_HEAD(&key_spd.splist[(dir)], (sp),		\
322	                          pslist_entry)
323#define SPLIST_WRITER_NEXT(sp)						\
324	PSLIST_WRITER_NEXT((sp), struct secpolicy, pslist_entry)
325#define SPLIST_WRITER_INSERT_TAIL(dir, new)				\
326	do {								\
327		if (SPLIST_WRITER_EMPTY((dir))) {			\
328			SPLIST_WRITER_INSERT_HEAD((dir), (new));	\
329		} else {						\
330			struct secpolicy *__sp;				\
331			SPLIST_WRITER_FOREACH(__sp, (dir)) {		\
332				if (SPLIST_WRITER_NEXT(__sp) == NULL) {	\
333					SPLIST_WRITER_INSERT_AFTER(__sp,\
334					    (new));			\
335					break;				\
336				}					\
337			}						\
338		}							\
339	} while (0)
340
341/* Macros for key_spd.socksplist */
342#define SOCKSPLIST_WRITER_FOREACH(sp)					\
343	PSLIST_WRITER_FOREACH((sp), &key_spd.socksplist,		\
344	                      struct secpolicy,	pslist_entry)
345#define SOCKSPLIST_READER_EMPTY()					\
346	(PSLIST_READER_FIRST(&key_spd.socksplist, struct secpolicy,	\
347	                     pslist_entry) == NULL)
348
349/* Macros for key_sad.sahlist */
350#define SAHLIST_ENTRY_INIT(sah)						\
351	PSLIST_ENTRY_INIT((sah), pslist_entry)
352#define SAHLIST_ENTRY_DESTROY(sah)					\
353	PSLIST_ENTRY_DESTROY((sah), pslist_entry)
354#define SAHLIST_WRITER_REMOVE(sah)					\
355	PSLIST_WRITER_REMOVE((sah), pslist_entry)
356#define SAHLIST_READER_FOREACH(sah)					\
357	for(int _i_sah = 0; _i_sah <= key_sad.sahlistmask; _i_sah++)	\
358		PSLIST_READER_FOREACH((sah), &key_sad.sahlists[_i_sah],	\
359		                      struct secashead, pslist_entry)
360#define SAHLIST_READER_FOREACH_SAIDX(sah, saidx)			\
361	PSLIST_READER_FOREACH((sah),					\
362	    &key_sad.sahlists[key_saidxhash((saidx),			\
363	                       key_sad.sahlistmask)],			\
364	    struct secashead, pslist_entry)
365#define SAHLIST_WRITER_FOREACH(sah)					\
366	for(int _i_sah = 0; _i_sah <= key_sad.sahlistmask; _i_sah++)	\
367		PSLIST_WRITER_FOREACH((sah), &key_sad.sahlists[_i_sah],	\
368		                     struct secashead, pslist_entry)
369#define SAHLIST_WRITER_INSERT_HEAD(sah)					\
370	PSLIST_WRITER_INSERT_HEAD(					\
371	    &key_sad.sahlists[key_saidxhash(&(sah)->saidx,		\
372	                      key_sad.sahlistmask)],	\
373	    (sah), pslist_entry)
374
375/* Macros for key_sad.sahlist#savlist */
376#define SAVLIST_ENTRY_INIT(sav)						\
377	PSLIST_ENTRY_INIT((sav), pslist_entry)
378#define SAVLIST_ENTRY_DESTROY(sav)					\
379	PSLIST_ENTRY_DESTROY((sav), pslist_entry)
380#define SAVLIST_READER_FIRST(sah, state)				\
381	PSLIST_READER_FIRST(&(sah)->savlist[(state)], struct secasvar,	\
382	                    pslist_entry)
383#define SAVLIST_WRITER_REMOVE(sav)					\
384	PSLIST_WRITER_REMOVE((sav), pslist_entry)
385#define SAVLIST_READER_FOREACH(sav, sah, state)				\
386	PSLIST_READER_FOREACH((sav), &(sah)->savlist[(state)],		\
387	                      struct secasvar, pslist_entry)
388#define SAVLIST_WRITER_FOREACH(sav, sah, state)				\
389	PSLIST_WRITER_FOREACH((sav), &(sah)->savlist[(state)],		\
390	                      struct secasvar, pslist_entry)
391#define SAVLIST_WRITER_INSERT_BEFORE(sav, new)				\
392	PSLIST_WRITER_INSERT_BEFORE((sav), (new), pslist_entry)
393#define SAVLIST_WRITER_INSERT_AFTER(sav, new)				\
394	PSLIST_WRITER_INSERT_AFTER((sav), (new), pslist_entry)
395#define SAVLIST_WRITER_EMPTY(sah, state)				\
396	(PSLIST_WRITER_FIRST(&(sah)->savlist[(state)], struct secasvar,	\
397	                     pslist_entry) == NULL)
398#define SAVLIST_WRITER_INSERT_HEAD(sah, state, sav)			\
399	PSLIST_WRITER_INSERT_HEAD(&(sah)->savlist[(state)], (sav),	\
400	                          pslist_entry)
401#define SAVLIST_WRITER_NEXT(sav)					\
402	PSLIST_WRITER_NEXT((sav), struct secasvar, pslist_entry)
403#define SAVLIST_WRITER_INSERT_TAIL(sah, state, new)			\
404	do {								\
405		if (SAVLIST_WRITER_EMPTY((sah), (state))) {		\
406			SAVLIST_WRITER_INSERT_HEAD((sah), (state), (new));\
407		} else {						\
408			struct secasvar *__sav;				\
409			SAVLIST_WRITER_FOREACH(__sav, (sah), (state)) {	\
410				if (SAVLIST_WRITER_NEXT(__sav) == NULL) {\
411					SAVLIST_WRITER_INSERT_AFTER(__sav,\
412					    (new));			\
413					break;				\
414				}					\
415			}						\
416		}							\
417	} while (0)
418#define SAVLIST_READER_NEXT(sav)					\
419	PSLIST_READER_NEXT((sav), struct secasvar, pslist_entry)
420
421/* Macros for key_sad.savlut */
422#define SAVLUT_ENTRY_INIT(sav)						\
423	PSLIST_ENTRY_INIT((sav), pslist_entry_savlut)
424#define SAVLUT_READER_FOREACH(sav, dst, proto, hash_key)		\
425	PSLIST_READER_FOREACH((sav),					\
426	&key_sad.savlut[key_savluthash(dst, proto, hash_key,		\
427	                  key_sad.savlutmask)],				\
428	struct secasvar, pslist_entry_savlut)
429#define SAVLUT_WRITER_INSERT_HEAD(sav)					\
430	key_savlut_writer_insert_head((sav))
431#define SAVLUT_WRITER_REMOVE(sav)					\
432	do {								\
433		if (!(sav)->savlut_added)				\
434			break;						\
435		PSLIST_WRITER_REMOVE((sav), pslist_entry_savlut);	\
436		(sav)->savlut_added = false;				\
437	} while(0)
438
439/* search order for SAs */
440	/*
441	 * This order is important because we must select the oldest SA
442	 * for outbound processing.  For inbound, This is not important.
443	 */
444static const u_int saorder_state_valid_prefer_old[] = {
445	SADB_SASTATE_DYING, SADB_SASTATE_MATURE,
446};
447static const u_int saorder_state_valid_prefer_new[] = {
448	SADB_SASTATE_MATURE, SADB_SASTATE_DYING,
449};
450
451static const u_int saorder_state_alive[] = {
452	/* except DEAD */
453	SADB_SASTATE_MATURE, SADB_SASTATE_DYING, SADB_SASTATE_LARVAL
454};
455static const u_int saorder_state_any[] = {
456	SADB_SASTATE_MATURE, SADB_SASTATE_DYING,
457	SADB_SASTATE_LARVAL, SADB_SASTATE_DEAD
458};
459
460#define SASTATE_ALIVE_FOREACH(s)				\
461	for (int _i = 0;					\
462	    _i < __arraycount(saorder_state_alive) ?		\
463	    (s) = saorder_state_alive[_i], true : false;	\
464	    _i++)
465#define SASTATE_ANY_FOREACH(s)					\
466	for (int _i = 0;					\
467	    _i < __arraycount(saorder_state_any) ?		\
468	    (s) = saorder_state_any[_i], true : false;		\
469	    _i++)
470#define SASTATE_USABLE_FOREACH(s)				\
471	for (int _i = 0;					\
472	    _i < __arraycount(saorder_state_valid_prefer_new) ?	\
473	    (s) = saorder_state_valid_prefer_new[_i],		\
474	    true : false;					\
475	    _i++)
476
477static const int minsize[] = {
478	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
479	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
480	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
481	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
482	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
483	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_SRC */
484	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_DST */
485	sizeof(struct sadb_address),	/* SADB_EXT_ADDRESS_PROXY */
486	sizeof(struct sadb_key),	/* SADB_EXT_KEY_AUTH */
487	sizeof(struct sadb_key),	/* SADB_EXT_KEY_ENCRYPT */
488	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_SRC */
489	sizeof(struct sadb_ident),	/* SADB_EXT_IDENTITY_DST */
490	sizeof(struct sadb_sens),	/* SADB_EXT_SENSITIVITY */
491	sizeof(struct sadb_prop),	/* SADB_EXT_PROPOSAL */
492	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_AUTH */
493	sizeof(struct sadb_supported),	/* SADB_EXT_SUPPORTED_ENCRYPT */
494	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
495	0,				/* SADB_X_EXT_KMPRIVATE */
496	sizeof(struct sadb_x_policy),	/* SADB_X_EXT_POLICY */
497	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
498	sizeof(struct sadb_x_nat_t_type),	/* SADB_X_EXT_NAT_T_TYPE */
499	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_SPORT */
500	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_DPORT */
501	sizeof(struct sadb_address),		/* SADB_X_EXT_NAT_T_OAI */
502	sizeof(struct sadb_address),		/* SADB_X_EXT_NAT_T_OAR */
503	sizeof(struct sadb_x_nat_t_frag),	/* SADB_X_EXT_NAT_T_FRAG */
504};
505static const int maxsize[] = {
506	sizeof(struct sadb_msg),	/* SADB_EXT_RESERVED */
507	sizeof(struct sadb_sa),		/* SADB_EXT_SA */
508	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_CURRENT */
509	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_HARD */
510	sizeof(struct sadb_lifetime),	/* SADB_EXT_LIFETIME_SOFT */
511	0,				/* SADB_EXT_ADDRESS_SRC */
512	0,				/* SADB_EXT_ADDRESS_DST */
513	0,				/* SADB_EXT_ADDRESS_PROXY */
514	0,				/* SADB_EXT_KEY_AUTH */
515	0,				/* SADB_EXT_KEY_ENCRYPT */
516	0,				/* SADB_EXT_IDENTITY_SRC */
517	0,				/* SADB_EXT_IDENTITY_DST */
518	0,				/* SADB_EXT_SENSITIVITY */
519	0,				/* SADB_EXT_PROPOSAL */
520	0,				/* SADB_EXT_SUPPORTED_AUTH */
521	0,				/* SADB_EXT_SUPPORTED_ENCRYPT */
522	sizeof(struct sadb_spirange),	/* SADB_EXT_SPIRANGE */
523	0,				/* SADB_X_EXT_KMPRIVATE */
524	0,				/* SADB_X_EXT_POLICY */
525	sizeof(struct sadb_x_sa2),	/* SADB_X_SA2 */
526	sizeof(struct sadb_x_nat_t_type),	/* SADB_X_EXT_NAT_T_TYPE */
527	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_SPORT */
528	sizeof(struct sadb_x_nat_t_port),	/* SADB_X_EXT_NAT_T_DPORT */
529	0,					/* SADB_X_EXT_NAT_T_OAI */
530	0,					/* SADB_X_EXT_NAT_T_OAR */
531	sizeof(struct sadb_x_nat_t_frag),	/* SADB_X_EXT_NAT_T_FRAG */
532};
533
534static int ipsec_esp_keymin = 256;
535static int ipsec_esp_auth = 0;
536static int ipsec_ah_keymin = 128;
537static bool ipsec_allow_different_idtype = false;
538
539#ifdef SYSCTL_DECL
540SYSCTL_DECL(_net_key);
541#endif
542
543#ifdef SYSCTL_INT
544SYSCTL_INT(_net_key, KEYCTL_DEBUG_LEVEL,	debug,	CTLFLAG_RW, \
545	&key_debug_level,	0,	"");
546
547/* max count of trial for the decision of spi value */
548SYSCTL_INT(_net_key, KEYCTL_SPI_TRY,		spi_trycnt,	CTLFLAG_RW, \
549	&key_spi_trycnt,	0,	"");
550
551/* minimum spi value to allocate automatically. */
552SYSCTL_INT(_net_key, KEYCTL_SPI_MIN_VALUE,	spi_minval,	CTLFLAG_RW, \
553	&key_spi_minval,	0,	"");
554
555/* maximun spi value to allocate automatically. */
556SYSCTL_INT(_net_key, KEYCTL_SPI_MAX_VALUE,	spi_maxval,	CTLFLAG_RW, \
557	&key_spi_maxval,	0,	"");
558
559/* interval to initialize randseed */
560SYSCTL_INT(_net_key, KEYCTL_RANDOM_INT,	int_random,	CTLFLAG_RW, \
561	&key_int_random,	0,	"");
562
563/* lifetime for larval SA */
564SYSCTL_INT(_net_key, KEYCTL_LARVAL_LIFETIME,	larval_lifetime, CTLFLAG_RW, \
565	&key_larval_lifetime,	0,	"");
566
567/* counter for blocking to send SADB_ACQUIRE to IKEd */
568SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_COUNT,	blockacq_count,	CTLFLAG_RW, \
569	&key_blockacq_count,	0,	"");
570
571/* lifetime for blocking to send SADB_ACQUIRE to IKEd */
572SYSCTL_INT(_net_key, KEYCTL_BLOCKACQ_LIFETIME,	blockacq_lifetime, CTLFLAG_RW, \
573	&key_blockacq_lifetime,	0,	"");
574
575/* ESP auth */
576SYSCTL_INT(_net_key, KEYCTL_ESP_AUTH,	esp_auth, CTLFLAG_RW, \
577	&ipsec_esp_auth,	0,	"");
578
579/* minimum ESP key length */
580SYSCTL_INT(_net_key, KEYCTL_ESP_KEYMIN,	esp_keymin, CTLFLAG_RW, \
581	&ipsec_esp_keymin,	0,	"");
582
583/* minimum AH key length */
584SYSCTL_INT(_net_key, KEYCTL_AH_KEYMIN,	ah_keymin, CTLFLAG_RW, \
585	&ipsec_ah_keymin,	0,	"");
586
587/* perfered old SA rather than new SA */
588SYSCTL_INT(_net_key, KEYCTL_PREFERED_OLDSA,	prefered_oldsa, CTLFLAG_RW,\
589	&key_prefered_oldsa,	0,	"");
590#endif /* SYSCTL_INT */
591
592#define __LIST_CHAINED(elm) \
593	(!((elm)->chain.le_next == NULL && (elm)->chain.le_prev == NULL))
594#define LIST_INSERT_TAIL(head, elm, type, field) \
595do {\
596	struct type *curelm = LIST_FIRST(head); \
597	if (curelm == NULL) {\
598		LIST_INSERT_HEAD(head, elm, field); \
599	} else { \
600		while (LIST_NEXT(curelm, field)) \
601			curelm = LIST_NEXT(curelm, field);\
602		LIST_INSERT_AFTER(curelm, elm, field);\
603	}\
604} while (0)
605
606#define KEY_CHKSASTATE(head, sav) \
607/* do */ { \
608	if ((head) != (sav)) {						\
609		IPSECLOG(LOG_DEBUG,					\
610		    "state mismatched (TREE=%d SA=%d)\n",		\
611		    (head), (sav));					\
612		continue;						\
613	}								\
614} /* while (0) */
615
616#define KEY_CHKSPDIR(head, sp) \
617do { \
618	if ((head) != (sp)) {						\
619		IPSECLOG(LOG_DEBUG,					\
620		    "direction mismatched (TREE=%d SP=%d), anyway continue.\n",\
621		    (head), (sp));					\
622	}								\
623} while (0)
624
625/*
626 * set parameters into secasindex buffer.
627 * Must allocate secasindex buffer before calling this function.
628 */
629static int
630key_setsecasidx(int, int, int, const struct sockaddr *,
631    const struct sockaddr *, struct secasindex *);
632
633/* key statistics */
634struct _keystat {
635	u_long getspi_count; /* the avarage of count to try to get new SPI */
636} keystat;
637
638static void
639key_init_spidx_bymsghdr(struct secpolicyindex *, const struct sadb_msghdr *);
640
641static const struct sockaddr *
642key_msghdr_get_sockaddr(const struct sadb_msghdr *mhp, int idx)
643{
644
645	return PFKEY_ADDR_SADDR(mhp->ext[idx]);
646}
647
648static void
649key_fill_replymsg(struct mbuf *m, int seq)
650{
651	struct sadb_msg *msg;
652
653	KASSERT(m->m_len >= sizeof(*msg));
654
655	msg = mtod(m, struct sadb_msg *);
656	msg->sadb_msg_errno = 0;
657	msg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
658	if (seq != 0)
659		msg->sadb_msg_seq = seq;
660}
661
662#if 0
663static void key_freeso(struct socket *);
664static void key_freesp_so(struct secpolicy **);
665#endif
666static struct secpolicy *key_getsp (const struct secpolicyindex *);
667static struct secpolicy *key_getspbyid (u_int32_t);
668static struct secpolicy *key_lookup_and_remove_sp(const struct secpolicyindex *, bool);
669static struct secpolicy *key_lookupbyid_and_remove_sp(u_int32_t, bool);
670static void key_destroy_sp(struct secpolicy *);
671static struct mbuf *key_gather_mbuf (struct mbuf *,
672	const struct sadb_msghdr *, int, int, ...);
673static int key_api_spdadd(struct socket *, struct mbuf *,
674	const struct sadb_msghdr *);
675static u_int32_t key_getnewspid (void);
676static int key_api_spddelete(struct socket *, struct mbuf *,
677	const struct sadb_msghdr *);
678static int key_api_spddelete2(struct socket *, struct mbuf *,
679	const struct sadb_msghdr *);
680static int key_api_spdget(struct socket *, struct mbuf *,
681	const struct sadb_msghdr *);
682static int key_api_spdflush(struct socket *, struct mbuf *,
683	const struct sadb_msghdr *);
684static int key_api_spddump(struct socket *, struct mbuf *,
685	const struct sadb_msghdr *);
686static struct mbuf * key_setspddump (int *errorp, pid_t);
687static struct mbuf * key_setspddump_chain (int *errorp, int *lenp, pid_t pid);
688static int key_api_nat_map(struct socket *, struct mbuf *,
689	const struct sadb_msghdr *);
690static struct mbuf *key_setdumpsp (struct secpolicy *,
691	u_int8_t, u_int32_t, pid_t);
692static u_int key_getspreqmsglen (const struct secpolicy *);
693static int key_spdexpire (struct secpolicy *);
694static struct secashead *key_newsah (const struct secasindex *);
695static void key_unlink_sah(struct secashead *);
696static void key_destroy_sah(struct secashead *);
697static bool key_sah_has_sav(struct secashead *);
698static void key_sah_ref(struct secashead *);
699static void key_sah_unref(struct secashead *);
700static void key_init_sav(struct secasvar *);
701static void key_wait_sav(struct secasvar *);
702static void key_destroy_sav(struct secasvar *);
703static struct secasvar *key_newsav(struct mbuf *,
704	const struct sadb_msghdr *, int *, int, const char*, int);
705#define	KEY_NEWSAV(m, sadb, e, proto)				\
706	key_newsav(m, sadb, e, proto, __func__, __LINE__)
707static void key_delsav (struct secasvar *);
708static struct secashead *key_getsah(const struct secasindex *, int);
709static struct secashead *key_getsah_ref(const struct secasindex *, int);
710static bool key_checkspidup(const struct secasindex *, u_int32_t);
711static struct secasvar *key_getsavbyspi (struct secashead *, u_int32_t);
712static int key_setsaval (struct secasvar *, struct mbuf *,
713	const struct sadb_msghdr *);
714static void key_freesaval(struct secasvar *);
715static int key_init_xform(struct secasvar *);
716static void key_clear_xform(struct secasvar *);
717static struct mbuf *key_setdumpsa (struct secasvar *, u_int8_t,
718	u_int8_t, u_int32_t, u_int32_t);
719static struct mbuf *key_setsadbxport (u_int16_t, u_int16_t);
720static struct mbuf *key_setsadbxtype (u_int16_t);
721static struct mbuf *key_setsadbxfrag (u_int16_t);
722static void key_porttosaddr (union sockaddr_union *, u_int16_t);
723static int key_checksalen (const union sockaddr_union *);
724static struct mbuf *key_setsadbmsg (u_int8_t, u_int16_t, u_int8_t,
725	u_int32_t, pid_t, u_int16_t, int);
726static struct mbuf *key_setsadbsa (struct secasvar *);
727static struct mbuf *key_setsadbaddr(u_int16_t,
728	const struct sockaddr *, u_int8_t, u_int16_t, int);
729#if 0
730static struct mbuf *key_setsadbident (u_int16_t, u_int16_t, void *,
731	int, u_int64_t);
732#endif
733static struct mbuf *key_setsadbxsa2 (u_int8_t, u_int32_t, u_int16_t);
734static struct mbuf *key_setsadbxpolicy (u_int16_t, u_int8_t,
735	u_int32_t, int);
736static void *key_newbuf (const void *, u_int);
737#ifdef INET6
738static int key_ismyaddr6 (const struct sockaddr_in6 *);
739#endif
740
741static void sysctl_net_keyv2_setup(struct sysctllog **);
742static void sysctl_net_key_compat_setup(struct sysctllog **);
743
744/* flags for key_saidx_match() */
745#define CMP_HEAD	1	/* protocol, addresses. */
746#define CMP_MODE_REQID	2	/* additionally HEAD, reqid, mode. */
747#define CMP_REQID	3	/* additionally HEAD, reaid. */
748#define CMP_EXACTLY	4	/* all elements. */
749static int key_saidx_match(const struct secasindex *,
750    const struct secasindex *, int);
751
752static int key_sockaddr_match(const struct sockaddr *,
753    const struct sockaddr *, int);
754static int key_bb_match_withmask(const void *, const void *, u_int);
755static u_int16_t key_satype2proto (u_int8_t);
756static u_int8_t key_proto2satype (u_int16_t);
757
758static int key_spidx_match_exactly(const struct secpolicyindex *,
759    const struct secpolicyindex *);
760static int key_spidx_match_withmask(const struct secpolicyindex *,
761    const struct secpolicyindex *);
762
763static int key_api_getspi(struct socket *, struct mbuf *,
764	const struct sadb_msghdr *);
765static u_int32_t key_do_getnewspi (const struct sadb_spirange *,
766					const struct secasindex *);
767static int key_handle_natt_info (struct secasvar *,
768				     const struct sadb_msghdr *);
769static int key_set_natt_ports (union sockaddr_union *,
770			 	union sockaddr_union *,
771				const struct sadb_msghdr *);
772static int key_api_update(struct socket *, struct mbuf *,
773	const struct sadb_msghdr *);
774#ifdef IPSEC_DOSEQCHECK
775static struct secasvar *key_getsavbyseq (struct secashead *, u_int32_t);
776#endif
777static int key_api_add(struct socket *, struct mbuf *,
778	const struct sadb_msghdr *);
779static int key_setident (struct secashead *, struct mbuf *,
780	const struct sadb_msghdr *);
781static struct mbuf *key_getmsgbuf_x1 (struct mbuf *,
782	const struct sadb_msghdr *);
783static int key_api_delete(struct socket *, struct mbuf *,
784	const struct sadb_msghdr *);
785static int key_api_get(struct socket *, struct mbuf *,
786	const struct sadb_msghdr *);
787
788static void key_getcomb_setlifetime (struct sadb_comb *);
789static struct mbuf *key_getcomb_esp(int);
790static struct mbuf *key_getcomb_ah(int);
791static struct mbuf *key_getcomb_ipcomp(int);
792static struct mbuf *key_getprop(const struct secasindex *, int);
793
794static int key_acquire(const struct secasindex *, const struct secpolicy *,
795	    int);
796static int key_acquire_sendup_mbuf_later(struct mbuf *);
797static void key_acquire_sendup_pending_mbuf(void);
798#ifndef IPSEC_NONBLOCK_ACQUIRE
799static struct secacq *key_newacq (const struct secasindex *);
800static struct secacq *key_getacq (const struct secasindex *);
801static struct secacq *key_getacqbyseq (u_int32_t);
802#endif
803#ifdef notyet
804static struct secspacq *key_newspacq (const struct secpolicyindex *);
805static struct secspacq *key_getspacq (const struct secpolicyindex *);
806#endif
807static int key_api_acquire(struct socket *, struct mbuf *,
808	const struct sadb_msghdr *);
809static int key_api_register(struct socket *, struct mbuf *,
810	const struct sadb_msghdr *);
811static int key_expire (struct secasvar *);
812static int key_api_flush(struct socket *, struct mbuf *,
813	const struct sadb_msghdr *);
814static struct mbuf *key_setdump_chain (u_int8_t req_satype, int *errorp,
815	int *lenp, pid_t pid);
816static int key_api_dump(struct socket *, struct mbuf *,
817	const struct sadb_msghdr *);
818static int key_api_promisc(struct socket *, struct mbuf *,
819	const struct sadb_msghdr *);
820static int key_senderror (struct socket *, struct mbuf *, int);
821static int key_validate_ext (const struct sadb_ext *, int);
822static int key_align (struct mbuf *, struct sadb_msghdr *);
823#if 0
824static const char *key_getfqdn (void);
825static const char *key_getuserfqdn (void);
826#endif
827static void key_sa_chgstate (struct secasvar *, u_int8_t);
828
829static struct mbuf *key_alloc_mbuf(int, int);
830static struct mbuf *key_alloc_mbuf_simple(int, int);
831
832static void key_timehandler(void *);
833static void key_timehandler_work(struct work *, void *);
834static struct callout	key_timehandler_ch;
835static struct workqueue	*key_timehandler_wq;
836static struct work	key_timehandler_wk;
837
838static inline void
839    key_savlut_writer_insert_head(struct secasvar *sav);
840static inline uint32_t
841    key_saidxhash(const struct secasindex *, u_long);
842static inline uint32_t
843    key_savluthash(const struct sockaddr *,
844    uint32_t, uint32_t, u_long);
845
846/*
847 * Utilities for percpu counters for sadb_lifetime_allocations and
848 * sadb_lifetime_bytes.
849 */
850#define LIFETIME_COUNTER_ALLOCATIONS	0
851#define LIFETIME_COUNTER_BYTES		1
852#define LIFETIME_COUNTER_SIZE		2
853
854typedef uint64_t lifetime_counters_t[LIFETIME_COUNTER_SIZE];
855
856static void
857key_sum_lifetime_counters(void *p, void *arg, struct cpu_info *ci __unused)
858{
859	lifetime_counters_t *one = p;
860	lifetime_counters_t *sum = arg;
861
862	(*sum)[LIFETIME_COUNTER_ALLOCATIONS] += (*one)[LIFETIME_COUNTER_ALLOCATIONS];
863	(*sum)[LIFETIME_COUNTER_BYTES] += (*one)[LIFETIME_COUNTER_BYTES];
864}
865
866u_int
867key_sp_refcnt(const struct secpolicy *sp)
868{
869
870	/* FIXME */
871	return 0;
872}
873
874void
875key_sp_touch(struct secpolicy *sp)
876{
877
878	sp->lastused = time_uptime;
879}
880
881static void
882key_spd_pserialize_perform(void)
883{
884
885	KASSERT(mutex_owned(&key_spd.lock));
886
887	while (key_spd.psz_performing)
888		cv_wait(&key_spd.cv_psz, &key_spd.lock);
889	key_spd.psz_performing = true;
890	mutex_exit(&key_spd.lock);
891
892	pserialize_perform(key_spd.psz);
893
894	mutex_enter(&key_spd.lock);
895	key_spd.psz_performing = false;
896	cv_broadcast(&key_spd.cv_psz);
897}
898
899/*
900 * Remove the sp from the key_spd.splist and wait for references to the sp
901 * to be released. key_spd.lock must be held.
902 */
903static void
904key_unlink_sp(struct secpolicy *sp)
905{
906
907	KASSERT(mutex_owned(&key_spd.lock));
908
909	sp->state = IPSEC_SPSTATE_DEAD;
910	SPLIST_WRITER_REMOVE(sp);
911
912	/* Invalidate all cached SPD pointers in the PCBs. */
913	ipsec_invalpcbcacheall();
914
915	KDASSERT(mutex_ownable(softnet_lock));
916	key_spd_pserialize_perform();
917
918	localcount_drain(&sp->localcount, &key_spd.cv_lc, &key_spd.lock);
919}
920
921/*
922 * Return 0 when there are known to be no SP's for the specified
923 * direction.  Otherwise return 1.  This is used by IPsec code
924 * to optimize performance.
925 */
926int
927key_havesp(u_int dir)
928{
929	return (dir == IPSEC_DIR_INBOUND || dir == IPSEC_DIR_OUTBOUND ?
930		!SPLIST_READER_EMPTY(dir) : 1);
931}
932
933/* %%% IPsec policy management */
934/*
935 * allocating a SP for OUTBOUND or INBOUND packet.
936 * Must call key_freesp() later.
937 * OUT:	NULL:	not found
938 *	others:	found and return the pointer.
939 */
940struct secpolicy *
941key_lookup_sp_byspidx(const struct secpolicyindex *spidx,
942    u_int dir, const char* where, int tag)
943{
944	struct secpolicy *sp;
945	int s;
946
947	KASSERT(spidx != NULL);
948	KASSERTMSG(IPSEC_DIR_IS_INOROUT(dir), "invalid direction %u", dir);
949
950	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
951
952	/* get a SP entry */
953	if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
954		kdebug_secpolicyindex("objects", spidx);
955	}
956
957	s = pserialize_read_enter();
958	SPLIST_READER_FOREACH(sp, dir) {
959		if (KEYDEBUG_ON(KEYDEBUG_IPSEC_DATA)) {
960			kdebug_secpolicyindex("in SPD", &sp->spidx);
961		}
962
963		if (sp->state == IPSEC_SPSTATE_DEAD)
964			continue;
965		if (key_spidx_match_withmask(&sp->spidx, spidx))
966			goto found;
967	}
968	sp = NULL;
969found:
970	if (sp) {
971		/* sanity check */
972		KEY_CHKSPDIR(sp->spidx.dir, dir);
973
974		/* found a SPD entry */
975		key_sp_touch(sp);
976		key_sp_ref(sp, where, tag);
977	}
978	pserialize_read_exit(s);
979
980	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
981	    "DP return SP:%p (ID=%u) refcnt %u\n",
982	    sp, sp ? sp->id : 0, key_sp_refcnt(sp));
983	return sp;
984}
985
986/*
987 * return a policy that matches this particular inbound packet.
988 * XXX slow
989 */
990struct secpolicy *
991key_gettunnel(const struct sockaddr *osrc,
992	      const struct sockaddr *odst,
993	      const struct sockaddr *isrc,
994	      const struct sockaddr *idst,
995	      const char* where, int tag)
996{
997	struct secpolicy *sp;
998	const int dir = IPSEC_DIR_INBOUND;
999	int s;
1000	struct ipsecrequest *r1, *r2, *p;
1001	struct secpolicyindex spidx;
1002
1003	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP, "DP from %s:%u\n", where, tag);
1004
1005	if (isrc->sa_family != idst->sa_family) {
1006		IPSECLOG(LOG_ERR,
1007		    "address family mismatched src %u, dst %u.\n",
1008		    isrc->sa_family, idst->sa_family);
1009		sp = NULL;
1010		goto done;
1011	}
1012
1013	s = pserialize_read_enter();
1014	SPLIST_READER_FOREACH(sp, dir) {
1015		if (sp->state == IPSEC_SPSTATE_DEAD)
1016			continue;
1017
1018		r1 = r2 = NULL;
1019		for (p = sp->req; p; p = p->next) {
1020			if (p->saidx.mode != IPSEC_MODE_TUNNEL)
1021				continue;
1022
1023			r1 = r2;
1024			r2 = p;
1025
1026			if (!r1) {
1027				/* here we look at address matches only */
1028				spidx = sp->spidx;
1029				if (isrc->sa_len > sizeof(spidx.src) ||
1030				    idst->sa_len > sizeof(spidx.dst))
1031					continue;
1032				memcpy(&spidx.src, isrc, isrc->sa_len);
1033				memcpy(&spidx.dst, idst, idst->sa_len);
1034				if (!key_spidx_match_withmask(&sp->spidx, &spidx))
1035					continue;
1036			} else {
1037				if (!key_sockaddr_match(&r1->saidx.src.sa, isrc, PORT_NONE) ||
1038				    !key_sockaddr_match(&r1->saidx.dst.sa, idst, PORT_NONE))
1039					continue;
1040			}
1041
1042			if (!key_sockaddr_match(&r2->saidx.src.sa, osrc, PORT_NONE) ||
1043			    !key_sockaddr_match(&r2->saidx.dst.sa, odst, PORT_NONE))
1044				continue;
1045
1046			goto found;
1047		}
1048	}
1049	sp = NULL;
1050found:
1051	if (sp) {
1052		key_sp_touch(sp);
1053		key_sp_ref(sp, where, tag);
1054	}
1055	pserialize_read_exit(s);
1056done:
1057	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1058	    "DP return SP:%p (ID=%u) refcnt %u\n",
1059	    sp, sp ? sp->id : 0, key_sp_refcnt(sp));
1060	return sp;
1061}
1062
1063/*
1064 * allocating an SA entry for an *OUTBOUND* packet.
1065 * checking each request entries in SP, and acquire an SA if need.
1066 * OUT:	0: there are valid requests.
1067 *	ENOENT: policy may be valid, but SA with REQUIRE is on acquiring.
1068 */
1069int
1070key_checkrequest(const struct ipsecrequest *isr, const struct secasindex *saidx,
1071    struct secasvar **ret)
1072{
1073	u_int level;
1074	int error;
1075	struct secasvar *sav;
1076
1077	KASSERT(isr != NULL);
1078	KASSERTMSG(saidx->mode == IPSEC_MODE_TRANSPORT ||
1079	    saidx->mode == IPSEC_MODE_TUNNEL,
1080	    "unexpected policy %u", saidx->mode);
1081
1082	/* get current level */
1083	level = ipsec_get_reqlevel(isr);
1084
1085	/*
1086	 * XXX guard against protocol callbacks from the crypto
1087	 * thread as they reference ipsecrequest.sav which we
1088	 * temporarily null out below.  Need to rethink how we
1089	 * handle bundled SA's in the callback thread.
1090	 */
1091
1092	sav = key_lookup_sa_bysaidx(saidx);
1093	if (sav != NULL) {
1094		*ret = sav;
1095		return 0;
1096	}
1097
1098	/* there is no SA */
1099	error = key_acquire(saidx, isr->sp, M_NOWAIT);
1100	if (error != 0) {
1101		/* XXX What should I do ? */
1102		IPSECLOG(LOG_DEBUG, "error %d returned from key_acquire.\n",
1103		    error);
1104		return error;
1105	}
1106
1107	if (level != IPSEC_LEVEL_REQUIRE) {
1108		/* XXX sigh, the interface to this routine is botched */
1109		*ret = NULL;
1110		return 0;
1111	} else {
1112		return ENOENT;
1113	}
1114}
1115
1116/*
1117 * looking up a SA for policy entry from SAD.
1118 * NOTE: searching SAD of aliving state.
1119 * OUT:	NULL:	not found.
1120 *	others:	found and return the pointer.
1121 */
1122struct secasvar *
1123key_lookup_sa_bysaidx(const struct secasindex *saidx)
1124{
1125	struct secashead *sah;
1126	struct secasvar *sav = NULL;
1127	u_int stateidx, state;
1128	const u_int *saorder_state_valid;
1129	int arraysize;
1130	int s;
1131
1132	s = pserialize_read_enter();
1133	sah = key_getsah(saidx, CMP_MODE_REQID);
1134	if (sah == NULL)
1135		goto out;
1136
1137	/*
1138	 * search a valid state list for outbound packet.
1139	 * This search order is important.
1140	 */
1141	if (key_prefered_oldsa) {
1142		saorder_state_valid = saorder_state_valid_prefer_old;
1143		arraysize = _ARRAYLEN(saorder_state_valid_prefer_old);
1144	} else {
1145		saorder_state_valid = saorder_state_valid_prefer_new;
1146		arraysize = _ARRAYLEN(saorder_state_valid_prefer_new);
1147	}
1148
1149	/* search valid state */
1150	for (stateidx = 0;
1151	     stateidx < arraysize;
1152	     stateidx++) {
1153
1154		state = saorder_state_valid[stateidx];
1155
1156		if (key_prefered_oldsa)
1157			sav = SAVLIST_READER_FIRST(sah, state);
1158		else {
1159			/* XXX need O(1) lookup */
1160			struct secasvar *last = NULL;
1161
1162			SAVLIST_READER_FOREACH(sav, sah, state)
1163				last = sav;
1164			sav = last;
1165		}
1166		if (sav != NULL) {
1167			KEY_SA_REF(sav);
1168			KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1169			    "DP cause refcnt++:%d SA:%p\n",
1170			    key_sa_refcnt(sav), sav);
1171			break;
1172		}
1173	}
1174out:
1175	pserialize_read_exit(s);
1176
1177	return sav;
1178}
1179
1180#if 0
1181static void
1182key_sendup_message_delete(struct secasvar *sav)
1183{
1184	struct mbuf *m, *result = 0;
1185	uint8_t satype;
1186
1187	satype = key_proto2satype(sav->sah->saidx.proto);
1188	if (satype == 0)
1189		goto msgfail;
1190
1191	m = key_setsadbmsg(SADB_DELETE, 0, satype, 0, 0, key_sa_refcnt(sav) - 1);
1192	if (m == NULL)
1193		goto msgfail;
1194	result = m;
1195
1196	/* set sadb_address for saidx's. */
1197	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa,
1198	    _BITS(sav->sah->saidx.src.sa.sa_len), IPSEC_ULPROTO_ANY);
1199	if (m == NULL)
1200		goto msgfail;
1201	m_cat(result, m);
1202
1203	/* set sadb_address for saidx's. */
1204	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.src.sa,
1205	    _BITS(sav->sah->saidx.src.sa.sa_len), IPSEC_ULPROTO_ANY);
1206	if (m == NULL)
1207		goto msgfail;
1208	m_cat(result, m);
1209
1210	/* create SA extension */
1211	m = key_setsadbsa(sav);
1212	if (m == NULL)
1213		goto msgfail;
1214	m_cat(result, m);
1215
1216	if (result->m_len < sizeof(struct sadb_msg)) {
1217		result = m_pullup(result, sizeof(struct sadb_msg));
1218		if (result == NULL)
1219			goto msgfail;
1220	}
1221
1222	result->m_pkthdr.len = 0;
1223	for (m = result; m; m = m->m_next)
1224		result->m_pkthdr.len += m->m_len;
1225	mtod(result, struct sadb_msg *)->sadb_msg_len =
1226	    PFKEY_UNIT64(result->m_pkthdr.len);
1227
1228	key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
1229	result = NULL;
1230msgfail:
1231	if (result)
1232		m_freem(result);
1233}
1234#endif
1235
1236/*
1237 * allocating a usable SA entry for a *INBOUND* packet.
1238 * Must call key_freesav() later.
1239 * OUT: positive:	pointer to a usable sav (i.e. MATURE or DYING state).
1240 *	NULL:		not found, or error occurred.
1241 *
1242 * In the comparison, no source address is used--for RFC2401 conformance.
1243 * To quote, from section 4.1:
1244 *	A security association is uniquely identified by a triple consisting
1245 *	of a Security Parameter Index (SPI), an IP Destination Address, and a
1246 *	security protocol (AH or ESP) identifier.
1247 * Note that, however, we do need to keep source address in IPsec SA.
1248 * IKE specification and PF_KEY specification do assume that we
1249 * keep source address in IPsec SA.  We see a tricky situation here.
1250 *
1251 * sport and dport are used for NAT-T. network order is always used.
1252 */
1253struct secasvar *
1254key_lookup_sa(
1255	const union sockaddr_union *dst,
1256	u_int proto,
1257	u_int32_t spi,
1258	u_int16_t sport,
1259	u_int16_t dport,
1260	const char* where, int tag)
1261{
1262	struct secasvar *sav;
1263	int chkport;
1264	int s;
1265
1266	int must_check_spi = 1;
1267	int must_check_alg = 0;
1268	u_int16_t cpi = 0;
1269	u_int8_t algo = 0;
1270	uint32_t hash_key = spi;
1271
1272	if ((sport != 0) && (dport != 0))
1273		chkport = PORT_STRICT;
1274	else
1275		chkport = PORT_NONE;
1276
1277	KASSERT(dst != NULL);
1278
1279	/*
1280	 * XXX IPCOMP case
1281	 * We use cpi to define spi here. In the case where cpi <=
1282	 * IPCOMP_CPI_NEGOTIATE_MIN, cpi just define the algorithm used, not
1283	 * the real spi. In this case, don't check the spi but check the
1284	 * algorithm
1285	 */
1286
1287	if (proto == IPPROTO_IPCOMP) {
1288		u_int32_t tmp;
1289		tmp = ntohl(spi);
1290		cpi = (u_int16_t) tmp;
1291		if (cpi < IPCOMP_CPI_NEGOTIATE_MIN) {
1292			algo = (u_int8_t) cpi;
1293			hash_key = algo;
1294			must_check_spi = 0;
1295			must_check_alg = 1;
1296		}
1297	}
1298	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1299	    "DP from %s:%u check_spi=%d(%#x), check_alg=%d(%d), proto=%d\n",
1300	    where, tag,
1301	    must_check_spi, ntohl(spi),
1302	    must_check_alg, algo,
1303	    proto);
1304
1305
1306	/*
1307	 * searching SAD.
1308	 * XXX: to be checked internal IP header somewhere.  Also when
1309	 * IPsec tunnel packet is received.  But ESP tunnel mode is
1310	 * encrypted so we can't check internal IP header.
1311	 */
1312	s = pserialize_read_enter();
1313	SAVLUT_READER_FOREACH(sav, &dst->sa, proto, hash_key) {
1314		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1315		    "try match spi %#x, %#x\n",
1316		    ntohl(spi), ntohl(sav->spi));
1317
1318		/* do not return entries w/ unusable state */
1319		if (!SADB_SASTATE_USABLE_P(sav)) {
1320			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1321			    "bad state %d\n", sav->state);
1322			continue;
1323		}
1324		if (proto != sav->sah->saidx.proto) {
1325			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1326			    "proto fail %d != %d\n",
1327			    proto, sav->sah->saidx.proto);
1328			continue;
1329		}
1330		if (must_check_spi && spi != sav->spi) {
1331			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1332			    "spi fail %#x != %#x\n",
1333			    ntohl(spi), ntohl(sav->spi));
1334			continue;
1335		}
1336		/* XXX only on the ipcomp case */
1337		if (must_check_alg && algo != sav->alg_comp) {
1338			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
1339			    "algo fail %d != %d\n",
1340			    algo, sav->alg_comp);
1341			continue;
1342		}
1343
1344#if 0	/* don't check src */
1345	/* Fix port in src->sa */
1346
1347		/* check src address */
1348		if (!key_sockaddr_match(&src->sa, &sav->sah->saidx.src.sa, PORT_NONE))
1349			continue;
1350#endif
1351		/* fix port of dst address XXX*/
1352		key_porttosaddr(__UNCONST(dst), dport);
1353		/* check dst address */
1354		if (!key_sockaddr_match(&dst->sa, &sav->sah->saidx.dst.sa, chkport))
1355			continue;
1356		key_sa_ref(sav, where, tag);
1357		goto done;
1358	}
1359	sav = NULL;
1360done:
1361	pserialize_read_exit(s);
1362
1363	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1364	    "DP return SA:%p; refcnt %u\n", sav, key_sa_refcnt(sav));
1365	return sav;
1366}
1367
1368static void
1369key_validate_savlist(const struct secashead *sah, const u_int state)
1370{
1371#ifdef DEBUG
1372	struct secasvar *sav, *next;
1373	int s;
1374
1375	/*
1376	 * The list should be sorted by lft_c->sadb_lifetime_addtime
1377	 * in ascending order.
1378	 */
1379	s = pserialize_read_enter();
1380	SAVLIST_READER_FOREACH(sav, sah, state) {
1381		next = SAVLIST_READER_NEXT(sav);
1382		if (next != NULL &&
1383		    sav->lft_c != NULL && next->lft_c != NULL) {
1384			KDASSERTMSG(sav->lft_c->sadb_lifetime_addtime <=
1385			    next->lft_c->sadb_lifetime_addtime,
1386			    "savlist is not sorted: sah=%p, state=%d, "
1387			    "sav=%" PRIu64 ", next=%" PRIu64, sah, state,
1388			    sav->lft_c->sadb_lifetime_addtime,
1389			    next->lft_c->sadb_lifetime_addtime);
1390		}
1391	}
1392	pserialize_read_exit(s);
1393#endif
1394}
1395
1396void
1397key_init_sp(struct secpolicy *sp)
1398{
1399
1400	ASSERT_SLEEPABLE();
1401
1402	sp->state = IPSEC_SPSTATE_ALIVE;
1403	if (sp->policy == IPSEC_POLICY_IPSEC)
1404		KASSERT(sp->req != NULL);
1405	localcount_init(&sp->localcount);
1406	SPLIST_ENTRY_INIT(sp);
1407}
1408
1409/*
1410 * Must be called in a pserialize read section. A held SP
1411 * must be released by key_sp_unref after use.
1412 */
1413void
1414key_sp_ref(struct secpolicy *sp, const char* where, int tag)
1415{
1416
1417	localcount_acquire(&sp->localcount);
1418
1419	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1420	    "DP SP:%p (ID=%u) from %s:%u; refcnt++ now %u\n",
1421	    sp, sp->id, where, tag, key_sp_refcnt(sp));
1422}
1423
1424/*
1425 * Must be called without holding key_spd.lock because the lock
1426 * would be held in localcount_release.
1427 */
1428void
1429key_sp_unref(struct secpolicy *sp, const char* where, int tag)
1430{
1431
1432	KDASSERT(mutex_ownable(&key_spd.lock));
1433
1434	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1435	    "DP SP:%p (ID=%u) from %s:%u; refcnt-- now %u\n",
1436	    sp, sp->id, where, tag, key_sp_refcnt(sp));
1437
1438	localcount_release(&sp->localcount, &key_spd.cv_lc, &key_spd.lock);
1439}
1440
1441static void
1442key_init_sav(struct secasvar *sav)
1443{
1444
1445	ASSERT_SLEEPABLE();
1446
1447	localcount_init(&sav->localcount);
1448	SAVLIST_ENTRY_INIT(sav);
1449	SAVLUT_ENTRY_INIT(sav);
1450}
1451
1452u_int
1453key_sa_refcnt(const struct secasvar *sav)
1454{
1455
1456	/* FIXME */
1457	return 0;
1458}
1459
1460void
1461key_sa_ref(struct secasvar *sav, const char* where, int tag)
1462{
1463
1464	localcount_acquire(&sav->localcount);
1465
1466	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1467	    "DP cause refcnt++: SA:%p from %s:%u\n",
1468	    sav, where, tag);
1469}
1470
1471void
1472key_sa_unref(struct secasvar *sav, const char* where, int tag)
1473{
1474
1475	KDASSERT(mutex_ownable(&key_sad.lock));
1476
1477	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1478	    "DP cause refcnt--: SA:%p from %s:%u\n",
1479	    sav, where, tag);
1480
1481	localcount_release(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
1482}
1483
1484#if 0
1485/*
1486 * Must be called after calling key_lookup_sp*().
1487 * For the packet with socket.
1488 */
1489static void
1490key_freeso(struct socket *so)
1491{
1492	/* sanity check */
1493	KASSERT(so != NULL);
1494
1495	switch (so->so_proto->pr_domain->dom_family) {
1496#ifdef INET
1497	case PF_INET:
1498	    {
1499		struct inpcb *pcb = sotoinpcb(so);
1500
1501		/* Does it have a PCB ? */
1502		if (pcb == NULL)
1503			return;
1504
1505		struct inpcbpolicy *sp = pcb->inp_sp;
1506		key_freesp_so(&sp->sp_in);
1507		key_freesp_so(&sp->sp_out);
1508	    }
1509		break;
1510#endif
1511#ifdef INET6
1512	case PF_INET6:
1513	    {
1514#ifdef HAVE_NRL_INPCB
1515		struct inpcb *pcb  = sotoinpcb(so);
1516		struct inpcbpolicy *sp = pcb->inp_sp;
1517
1518		/* Does it have a PCB ? */
1519		if (pcb == NULL)
1520			return;
1521		key_freesp_so(&sp->sp_in);
1522		key_freesp_so(&sp->sp_out);
1523#else
1524		struct in6pcb *pcb  = sotoin6pcb(so);
1525
1526		/* Does it have a PCB ? */
1527		if (pcb == NULL)
1528			return;
1529		key_freesp_so(&pcb->in6p_sp->sp_in);
1530		key_freesp_so(&pcb->in6p_sp->sp_out);
1531#endif
1532	    }
1533		break;
1534#endif /* INET6 */
1535	default:
1536		IPSECLOG(LOG_DEBUG, "unknown address family=%d.\n",
1537		    so->so_proto->pr_domain->dom_family);
1538		return;
1539	}
1540}
1541
1542static void
1543key_freesp_so(struct secpolicy **sp)
1544{
1545
1546	KASSERT(sp != NULL);
1547	KASSERT(*sp != NULL);
1548
1549	if ((*sp)->policy == IPSEC_POLICY_ENTRUST ||
1550	    (*sp)->policy == IPSEC_POLICY_BYPASS)
1551		return;
1552
1553	KASSERTMSG((*sp)->policy == IPSEC_POLICY_IPSEC,
1554	    "invalid policy %u", (*sp)->policy);
1555	KEY_SP_UNREF(&sp);
1556}
1557#endif
1558
1559static void
1560key_sad_pserialize_perform(void)
1561{
1562
1563	KASSERT(mutex_owned(&key_sad.lock));
1564
1565	while (key_sad.psz_performing)
1566		cv_wait(&key_sad.cv_psz, &key_sad.lock);
1567	key_sad.psz_performing = true;
1568	mutex_exit(&key_sad.lock);
1569
1570	pserialize_perform(key_sad.psz);
1571
1572	mutex_enter(&key_sad.lock);
1573	key_sad.psz_performing = false;
1574	cv_broadcast(&key_sad.cv_psz);
1575}
1576
1577/*
1578 * Remove the sav from the savlist of its sah and wait for references to the sav
1579 * to be released. key_sad.lock must be held.
1580 */
1581static void
1582key_unlink_sav(struct secasvar *sav)
1583{
1584
1585	KASSERT(mutex_owned(&key_sad.lock));
1586
1587	SAVLIST_WRITER_REMOVE(sav);
1588	SAVLUT_WRITER_REMOVE(sav);
1589
1590	KDASSERT(mutex_ownable(softnet_lock));
1591	key_sad_pserialize_perform();
1592
1593	localcount_drain(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
1594}
1595
1596/*
1597 * Destroy an sav where the sav must be unlinked from an sah
1598 * by say key_unlink_sav.
1599 */
1600static void
1601key_destroy_sav(struct secasvar *sav)
1602{
1603
1604	ASSERT_SLEEPABLE();
1605
1606	localcount_fini(&sav->localcount);
1607	SAVLIST_ENTRY_DESTROY(sav);
1608
1609	key_delsav(sav);
1610}
1611
1612/*
1613 * Wait for references of a passed sav to go away.
1614 */
1615static void
1616key_wait_sav(struct secasvar *sav)
1617{
1618
1619	ASSERT_SLEEPABLE();
1620
1621	mutex_enter(&key_sad.lock);
1622	KASSERT(sav->state == SADB_SASTATE_DEAD);
1623	KDASSERT(mutex_ownable(softnet_lock));
1624	key_sad_pserialize_perform();
1625	localcount_drain(&sav->localcount, &key_sad.cv_lc, &key_sad.lock);
1626	mutex_exit(&key_sad.lock);
1627}
1628
1629/* %%% SPD management */
1630/*
1631 * free security policy entry.
1632 */
1633static void
1634key_destroy_sp(struct secpolicy *sp)
1635{
1636
1637	SPLIST_ENTRY_DESTROY(sp);
1638	localcount_fini(&sp->localcount);
1639
1640	key_free_sp(sp);
1641
1642	key_update_used();
1643}
1644
1645void
1646key_free_sp(struct secpolicy *sp)
1647{
1648	struct ipsecrequest *isr = sp->req, *nextisr;
1649
1650	while (isr != NULL) {
1651		nextisr = isr->next;
1652		kmem_free(isr, sizeof(*isr));
1653		isr = nextisr;
1654	}
1655
1656	kmem_free(sp, sizeof(*sp));
1657}
1658
1659void
1660key_socksplist_add(struct secpolicy *sp)
1661{
1662
1663	mutex_enter(&key_spd.lock);
1664	PSLIST_WRITER_INSERT_HEAD(&key_spd.socksplist, sp, pslist_entry);
1665	mutex_exit(&key_spd.lock);
1666
1667	key_update_used();
1668}
1669
1670/*
1671 * search SPD
1672 * OUT:	NULL	: not found
1673 *	others	: found, pointer to a SP.
1674 */
1675static struct secpolicy *
1676key_getsp(const struct secpolicyindex *spidx)
1677{
1678	struct secpolicy *sp;
1679	int s;
1680
1681	KASSERT(spidx != NULL);
1682
1683	s = pserialize_read_enter();
1684	SPLIST_READER_FOREACH(sp, spidx->dir) {
1685		if (sp->state == IPSEC_SPSTATE_DEAD)
1686			continue;
1687		if (key_spidx_match_exactly(spidx, &sp->spidx)) {
1688			KEY_SP_REF(sp);
1689			pserialize_read_exit(s);
1690			return sp;
1691		}
1692	}
1693	pserialize_read_exit(s);
1694
1695	return NULL;
1696}
1697
1698/*
1699 * search SPD and remove found SP
1700 * OUT:	NULL	: not found
1701 *	others	: found, pointer to a SP.
1702 */
1703static struct secpolicy *
1704key_lookup_and_remove_sp(const struct secpolicyindex *spidx, bool from_kernel)
1705{
1706	struct secpolicy *sp = NULL;
1707
1708	mutex_enter(&key_spd.lock);
1709	SPLIST_WRITER_FOREACH(sp, spidx->dir) {
1710		KASSERTMSG(sp->state != IPSEC_SPSTATE_DEAD, "sp->state=%u",
1711		    sp->state);
1712		/*
1713		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
1714		 * removed by userland programs.
1715		 */
1716		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
1717			continue;
1718		if (key_spidx_match_exactly(spidx, &sp->spidx)) {
1719			key_unlink_sp(sp);
1720			goto out;
1721		}
1722	}
1723	sp = NULL;
1724out:
1725	mutex_exit(&key_spd.lock);
1726
1727	return sp;
1728}
1729
1730/*
1731 * get SP by index.
1732 * OUT:	NULL	: not found
1733 *	others	: found, pointer to a SP.
1734 */
1735static struct secpolicy *
1736key_getspbyid(u_int32_t id)
1737{
1738	struct secpolicy *sp;
1739	int s;
1740
1741	s = pserialize_read_enter();
1742	SPLIST_READER_FOREACH(sp, IPSEC_DIR_INBOUND) {
1743		if (sp->state == IPSEC_SPSTATE_DEAD)
1744			continue;
1745		if (sp->id == id) {
1746			KEY_SP_REF(sp);
1747			goto out;
1748		}
1749	}
1750
1751	SPLIST_READER_FOREACH(sp, IPSEC_DIR_OUTBOUND) {
1752		if (sp->state == IPSEC_SPSTATE_DEAD)
1753			continue;
1754		if (sp->id == id) {
1755			KEY_SP_REF(sp);
1756			goto out;
1757		}
1758	}
1759out:
1760	pserialize_read_exit(s);
1761	return sp;
1762}
1763
1764/*
1765 * get SP by index, remove and return it.
1766 * OUT:	NULL	: not found
1767 *	others	: found, pointer to a SP.
1768 */
1769static struct secpolicy *
1770key_lookupbyid_and_remove_sp(u_int32_t id, bool from_kernel)
1771{
1772	struct secpolicy *sp;
1773
1774	mutex_enter(&key_spd.lock);
1775	SPLIST_READER_FOREACH(sp, IPSEC_DIR_INBOUND) {
1776		KASSERTMSG(sp->state != IPSEC_SPSTATE_DEAD, "sp->state=%u",
1777		    sp->state);
1778		/*
1779		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
1780		 * removed by userland programs.
1781		 */
1782		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
1783			continue;
1784		if (sp->id == id)
1785			goto out;
1786	}
1787
1788	SPLIST_READER_FOREACH(sp, IPSEC_DIR_OUTBOUND) {
1789		KASSERTMSG(sp->state != IPSEC_SPSTATE_DEAD, "sp->state=%u",
1790		    sp->state);
1791		/*
1792		 * SPs created in kernel(e.g. ipsec(4) I/F) must not be
1793		 * removed by userland programs.
1794		 */
1795		if (!from_kernel && sp->origin == IPSEC_SPORIGIN_KERNEL)
1796			continue;
1797		if (sp->id == id)
1798			goto out;
1799	}
1800out:
1801	if (sp != NULL)
1802		key_unlink_sp(sp);
1803	mutex_exit(&key_spd.lock);
1804	return sp;
1805}
1806
1807struct secpolicy *
1808key_newsp(const char* where, int tag)
1809{
1810	struct secpolicy *newsp = NULL;
1811
1812	newsp = kmem_zalloc(sizeof(struct secpolicy), KM_SLEEP);
1813
1814	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
1815	    "DP from %s:%u return SP:%p\n", where, tag, newsp);
1816	return newsp;
1817}
1818
1819/*
1820 * create secpolicy structure from sadb_x_policy structure.
1821 * NOTE: `state', `secpolicyindex' in secpolicy structure are not set,
1822 * so must be set properly later.
1823 */
1824static struct secpolicy *
1825_key_msg2sp(const struct sadb_x_policy *xpl0, size_t len, int *error,
1826    bool from_kernel)
1827{
1828	struct secpolicy *newsp;
1829
1830	KASSERT(!cpu_softintr_p());
1831	KASSERT(xpl0 != NULL);
1832	KASSERT(len >= sizeof(*xpl0));
1833
1834	if (len != PFKEY_EXTLEN(xpl0)) {
1835		IPSECLOG(LOG_DEBUG, "Invalid msg length.\n");
1836		*error = EINVAL;
1837		return NULL;
1838	}
1839
1840	newsp = KEY_NEWSP();
1841	if (newsp == NULL) {
1842		*error = ENOBUFS;
1843		return NULL;
1844	}
1845
1846	newsp->spidx.dir = xpl0->sadb_x_policy_dir;
1847	newsp->policy = xpl0->sadb_x_policy_type;
1848
1849	/* check policy */
1850	switch (xpl0->sadb_x_policy_type) {
1851	case IPSEC_POLICY_DISCARD:
1852	case IPSEC_POLICY_NONE:
1853	case IPSEC_POLICY_ENTRUST:
1854	case IPSEC_POLICY_BYPASS:
1855		newsp->req = NULL;
1856		*error = 0;
1857		return newsp;
1858
1859	case IPSEC_POLICY_IPSEC:
1860		/* Continued */
1861		break;
1862	default:
1863		IPSECLOG(LOG_DEBUG, "invalid policy type.\n");
1864		key_free_sp(newsp);
1865		*error = EINVAL;
1866		return NULL;
1867	}
1868
1869	/* IPSEC_POLICY_IPSEC */
1870    {
1871	int tlen;
1872	const struct sadb_x_ipsecrequest *xisr;
1873	uint16_t xisr_reqid;
1874	struct ipsecrequest **p_isr = &newsp->req;
1875
1876	/* validity check */
1877	if (PFKEY_EXTLEN(xpl0) < sizeof(*xpl0)) {
1878		IPSECLOG(LOG_DEBUG, "Invalid msg length.\n");
1879		*error = EINVAL;
1880		goto free_exit;
1881	}
1882
1883	tlen = PFKEY_EXTLEN(xpl0) - sizeof(*xpl0);
1884	xisr = (const struct sadb_x_ipsecrequest *)(xpl0 + 1);
1885
1886	while (tlen > 0) {
1887		/* length check */
1888		if (xisr->sadb_x_ipsecrequest_len < sizeof(*xisr)) {
1889			IPSECLOG(LOG_DEBUG, "invalid ipsecrequest length.\n");
1890			*error = EINVAL;
1891			goto free_exit;
1892		}
1893
1894		/* allocate request buffer */
1895		*p_isr = kmem_zalloc(sizeof(**p_isr), KM_SLEEP);
1896
1897		/* set values */
1898		(*p_isr)->next = NULL;
1899
1900		switch (xisr->sadb_x_ipsecrequest_proto) {
1901		case IPPROTO_ESP:
1902		case IPPROTO_AH:
1903		case IPPROTO_IPCOMP:
1904			break;
1905		default:
1906			IPSECLOG(LOG_DEBUG, "invalid proto type=%u\n",
1907			    xisr->sadb_x_ipsecrequest_proto);
1908			*error = EPROTONOSUPPORT;
1909			goto free_exit;
1910		}
1911		(*p_isr)->saidx.proto = xisr->sadb_x_ipsecrequest_proto;
1912
1913		switch (xisr->sadb_x_ipsecrequest_mode) {
1914		case IPSEC_MODE_TRANSPORT:
1915		case IPSEC_MODE_TUNNEL:
1916			break;
1917		case IPSEC_MODE_ANY:
1918		default:
1919			IPSECLOG(LOG_DEBUG, "invalid mode=%u\n",
1920			    xisr->sadb_x_ipsecrequest_mode);
1921			*error = EINVAL;
1922			goto free_exit;
1923		}
1924		(*p_isr)->saidx.mode = xisr->sadb_x_ipsecrequest_mode;
1925
1926		switch (xisr->sadb_x_ipsecrequest_level) {
1927		case IPSEC_LEVEL_DEFAULT:
1928		case IPSEC_LEVEL_USE:
1929		case IPSEC_LEVEL_REQUIRE:
1930			break;
1931		case IPSEC_LEVEL_UNIQUE:
1932			xisr_reqid = xisr->sadb_x_ipsecrequest_reqid;
1933			/* validity check */
1934			/*
1935			 * case 1) from_kernel == false
1936			 * That means the request comes from userland.
1937			 * If range violation of reqid, kernel will
1938			 * update it, don't refuse it.
1939			 *
1940			 * case 2) from_kernel == true
1941			 * That means the request comes from kernel
1942			 * (e.g. ipsec(4) I/F).
1943			 * Use thre requested reqid to avoid inconsistency
1944			 * between kernel's reqid and the reqid in pf_key
1945			 * message sent to userland. The pf_key message is
1946			 * built by diverting request mbuf.
1947			 */
1948			if (!from_kernel &&
1949			    xisr_reqid > IPSEC_MANUAL_REQID_MAX) {
1950				IPSECLOG(LOG_DEBUG,
1951				    "reqid=%d range "
1952				    "violation, updated by kernel.\n",
1953				    xisr_reqid);
1954				xisr_reqid = 0;
1955			}
1956
1957			/* allocate new reqid id if reqid is zero. */
1958			if (xisr_reqid == 0) {
1959				u_int16_t reqid = key_newreqid();
1960				if (reqid == 0) {
1961					*error = ENOBUFS;
1962					goto free_exit;
1963				}
1964				(*p_isr)->saidx.reqid = reqid;
1965			} else {
1966			/* set it for manual keying. */
1967				(*p_isr)->saidx.reqid = xisr_reqid;
1968			}
1969			break;
1970
1971		default:
1972			IPSECLOG(LOG_DEBUG, "invalid level=%u\n",
1973			    xisr->sadb_x_ipsecrequest_level);
1974			*error = EINVAL;
1975			goto free_exit;
1976		}
1977		(*p_isr)->level = xisr->sadb_x_ipsecrequest_level;
1978
1979		/* set IP addresses if there */
1980		/*
1981		 * NOTE:
1982		 * MOBIKE Extensions for PF_KEY draft says:
1983		 *     If tunnel mode is specified, the sadb_x_ipsecrequest
1984		 *     structure is followed by two sockaddr structures that
1985		 *     define the tunnel endpoint addresses.  In the case that
1986		 *     transport mode is used, no additional addresses are
1987		 *     specified.
1988		 * see: https://tools.ietf.org/html/draft-schilcher-mobike-pfkey-extension-01
1989		 *
1990		 * And then, the IP addresses will be set by
1991		 * ipsec_fill_saidx_bymbuf() from packet in transport mode.
1992		 * This behavior is used by NAT-T enabled ipsecif(4).
1993		 */
1994		if (xisr->sadb_x_ipsecrequest_len > sizeof(*xisr)) {
1995			const struct sockaddr *paddr;
1996
1997			paddr = (const struct sockaddr *)(xisr + 1);
1998
1999			/* validity check */
2000			if (paddr->sa_len > sizeof((*p_isr)->saidx.src)) {
2001				IPSECLOG(LOG_DEBUG, "invalid request "
2002				    "address length.\n");
2003				*error = EINVAL;
2004				goto free_exit;
2005			}
2006			memcpy(&(*p_isr)->saidx.src, paddr, paddr->sa_len);
2007
2008			paddr = (const struct sockaddr *)((const char *)paddr
2009			    + paddr->sa_len);
2010
2011			/* validity check */
2012			if (paddr->sa_len > sizeof((*p_isr)->saidx.dst)) {
2013				IPSECLOG(LOG_DEBUG, "invalid request "
2014				    "address length.\n");
2015				*error = EINVAL;
2016				goto free_exit;
2017			}
2018			memcpy(&(*p_isr)->saidx.dst, paddr, paddr->sa_len);
2019		}
2020
2021		(*p_isr)->sp = newsp;
2022
2023		/* initialization for the next. */
2024		p_isr = &(*p_isr)->next;
2025		tlen -= xisr->sadb_x_ipsecrequest_len;
2026
2027		/* validity check */
2028		if (tlen < 0) {
2029			IPSECLOG(LOG_DEBUG, "becoming tlen < 0.\n");
2030			*error = EINVAL;
2031			goto free_exit;
2032		}
2033
2034		xisr = (const struct sadb_x_ipsecrequest *)((const char *)xisr +
2035		    xisr->sadb_x_ipsecrequest_len);
2036	}
2037    }
2038
2039	*error = 0;
2040	return newsp;
2041
2042free_exit:
2043	key_free_sp(newsp);
2044	return NULL;
2045}
2046
2047struct secpolicy *
2048key_msg2sp(const struct sadb_x_policy *xpl0, size_t len, int *error)
2049{
2050
2051	return _key_msg2sp(xpl0, len, error, false);
2052}
2053
2054u_int16_t
2055key_newreqid(void)
2056{
2057	static u_int16_t auto_reqid = IPSEC_MANUAL_REQID_MAX + 1;
2058
2059	auto_reqid = (auto_reqid == 0xffff ?
2060	    IPSEC_MANUAL_REQID_MAX + 1 : auto_reqid + 1);
2061
2062	/* XXX should be unique check */
2063
2064	return auto_reqid;
2065}
2066
2067/*
2068 * copy secpolicy struct to sadb_x_policy structure indicated.
2069 */
2070struct mbuf *
2071key_sp2msg(const struct secpolicy *sp, int mflag)
2072{
2073	struct sadb_x_policy *xpl;
2074	int tlen;
2075	char *p;
2076	struct mbuf *m;
2077
2078	KASSERT(sp != NULL);
2079
2080	tlen = key_getspreqmsglen(sp);
2081
2082	m = key_alloc_mbuf(tlen, mflag);
2083	if (!m || m->m_next) {	/*XXX*/
2084		if (m)
2085			m_freem(m);
2086		return NULL;
2087	}
2088
2089	m->m_len = tlen;
2090	m->m_next = NULL;
2091	xpl = mtod(m, struct sadb_x_policy *);
2092	memset(xpl, 0, tlen);
2093
2094	xpl->sadb_x_policy_len = PFKEY_UNIT64(tlen);
2095	xpl->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
2096	xpl->sadb_x_policy_type = sp->policy;
2097	xpl->sadb_x_policy_dir = sp->spidx.dir;
2098	xpl->sadb_x_policy_id = sp->id;
2099	if (sp->origin == IPSEC_SPORIGIN_KERNEL)
2100		xpl->sadb_x_policy_flags |= IPSEC_POLICY_FLAG_ORIGIN_KERNEL;
2101	p = (char *)xpl + sizeof(*xpl);
2102
2103	/* if is the policy for ipsec ? */
2104	if (sp->policy == IPSEC_POLICY_IPSEC) {
2105		struct sadb_x_ipsecrequest *xisr;
2106		struct ipsecrequest *isr;
2107
2108		for (isr = sp->req; isr != NULL; isr = isr->next) {
2109
2110			xisr = (struct sadb_x_ipsecrequest *)p;
2111
2112			xisr->sadb_x_ipsecrequest_proto = isr->saidx.proto;
2113			xisr->sadb_x_ipsecrequest_mode = isr->saidx.mode;
2114			xisr->sadb_x_ipsecrequest_level = isr->level;
2115			xisr->sadb_x_ipsecrequest_reqid = isr->saidx.reqid;
2116
2117			p += sizeof(*xisr);
2118			memcpy(p, &isr->saidx.src, isr->saidx.src.sa.sa_len);
2119			p += isr->saidx.src.sa.sa_len;
2120			memcpy(p, &isr->saidx.dst, isr->saidx.dst.sa.sa_len);
2121			p += isr->saidx.src.sa.sa_len;
2122
2123			xisr->sadb_x_ipsecrequest_len =
2124			    PFKEY_ALIGN8(sizeof(*xisr)
2125			    + isr->saidx.src.sa.sa_len
2126			    + isr->saidx.dst.sa.sa_len);
2127		}
2128	}
2129
2130	return m;
2131}
2132
2133/*
2134 * m will not be freed nor modified. It never return NULL.
2135 * If it returns a mbuf of M_PKTHDR, the mbuf ensures to have
2136 * contiguous length at least sizeof(struct sadb_msg).
2137 */
2138static struct mbuf *
2139key_gather_mbuf(struct mbuf *m, const struct sadb_msghdr *mhp,
2140		int ndeep, int nitem, ...)
2141{
2142	va_list ap;
2143	int idx;
2144	int i;
2145	struct mbuf *result = NULL, *n;
2146	int len;
2147
2148	KASSERT(m != NULL);
2149	KASSERT(mhp != NULL);
2150	KASSERT(!cpu_softintr_p());
2151
2152	va_start(ap, nitem);
2153	for (i = 0; i < nitem; i++) {
2154		idx = va_arg(ap, int);
2155		KASSERT(idx >= 0);
2156		KASSERT(idx <= SADB_EXT_MAX);
2157		/* don't attempt to pull empty extension */
2158		if (idx == SADB_EXT_RESERVED && mhp->msg == NULL)
2159			continue;
2160		if (idx != SADB_EXT_RESERVED &&
2161		    (mhp->ext[idx] == NULL || mhp->extlen[idx] == 0))
2162			continue;
2163
2164		if (idx == SADB_EXT_RESERVED) {
2165			CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MHLEN);
2166			len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2167			MGETHDR(n, M_WAITOK, MT_DATA);
2168			n->m_len = len;
2169			n->m_next = NULL;
2170			m_copydata(m, 0, sizeof(struct sadb_msg),
2171			    mtod(n, void *));
2172		} else if (i < ndeep) {
2173			len = mhp->extlen[idx];
2174			n = key_alloc_mbuf(len, M_WAITOK);
2175			KASSERT(n->m_next == NULL);
2176			m_copydata(m, mhp->extoff[idx], mhp->extlen[idx],
2177			    mtod(n, void *));
2178		} else {
2179			n = m_copym(m, mhp->extoff[idx], mhp->extlen[idx],
2180			    M_WAITOK);
2181		}
2182		KASSERT(n != NULL);
2183
2184		if (result)
2185			m_cat(result, n);
2186		else
2187			result = n;
2188	}
2189	va_end(ap);
2190
2191	KASSERT(result != NULL);
2192	if ((result->m_flags & M_PKTHDR) != 0) {
2193		result->m_pkthdr.len = 0;
2194		for (n = result; n; n = n->m_next)
2195			result->m_pkthdr.len += n->m_len;
2196		KASSERT(result->m_len >= sizeof(struct sadb_msg));
2197	}
2198
2199	return result;
2200}
2201
2202/*
2203 * The argument _sp must not overwrite until SP is created and registered
2204 * successfully.
2205 */
2206static int
2207key_spdadd(struct socket *so, struct mbuf *m,
2208	   const struct sadb_msghdr *mhp, struct secpolicy **_sp,
2209	   bool from_kernel)
2210{
2211	const struct sockaddr *src, *dst;
2212	const struct sadb_x_policy *xpl0;
2213	struct sadb_x_policy *xpl;
2214	const struct sadb_lifetime *lft = NULL;
2215	struct secpolicyindex spidx;
2216	struct secpolicy *newsp;
2217	int error;
2218	uint32_t sadb_x_policy_id;
2219
2220	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
2221	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
2222	    mhp->ext[SADB_X_EXT_POLICY] == NULL) {
2223		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2224		return key_senderror(so, m, EINVAL);
2225	}
2226	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
2227	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
2228	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2229		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2230		return key_senderror(so, m, EINVAL);
2231	}
2232	if (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL) {
2233		if (mhp->extlen[SADB_EXT_LIFETIME_HARD] <
2234		    sizeof(struct sadb_lifetime)) {
2235			IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2236			return key_senderror(so, m, EINVAL);
2237		}
2238		lft = mhp->ext[SADB_EXT_LIFETIME_HARD];
2239	}
2240
2241	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
2242
2243	/* checking the direction. */
2244	switch (xpl0->sadb_x_policy_dir) {
2245	case IPSEC_DIR_INBOUND:
2246	case IPSEC_DIR_OUTBOUND:
2247		break;
2248	default:
2249		IPSECLOG(LOG_DEBUG, "Invalid SP direction.\n");
2250		return key_senderror(so, m, EINVAL);
2251	}
2252
2253	/* check policy */
2254	/* key_api_spdadd() accepts DISCARD, NONE and IPSEC. */
2255	if (xpl0->sadb_x_policy_type == IPSEC_POLICY_ENTRUST ||
2256	    xpl0->sadb_x_policy_type == IPSEC_POLICY_BYPASS) {
2257		IPSECLOG(LOG_DEBUG, "Invalid policy type.\n");
2258		return key_senderror(so, m, EINVAL);
2259	}
2260
2261	/* policy requests are mandatory when action is ipsec. */
2262	if (mhp->msg->sadb_msg_type != SADB_X_SPDSETIDX &&
2263	    xpl0->sadb_x_policy_type == IPSEC_POLICY_IPSEC &&
2264	    mhp->extlen[SADB_X_EXT_POLICY] <= sizeof(*xpl0)) {
2265		IPSECLOG(LOG_DEBUG, "some policy requests part required.\n");
2266		return key_senderror(so, m, EINVAL);
2267	}
2268
2269	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
2270	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
2271
2272	/* sanity check on addr pair */
2273	if (src->sa_family != dst->sa_family)
2274		return key_senderror(so, m, EINVAL);
2275	if (src->sa_len != dst->sa_len)
2276		return key_senderror(so, m, EINVAL);
2277
2278	key_init_spidx_bymsghdr(&spidx, mhp);
2279
2280	/*
2281	 * checking there is SP already or not.
2282	 * SPDUPDATE doesn't depend on whether there is a SP or not.
2283	 * If the type is either SPDADD or SPDSETIDX AND a SP is found,
2284	 * then error.
2285	 */
2286    {
2287	struct secpolicy *sp;
2288
2289	if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2290		sp = key_lookup_and_remove_sp(&spidx, from_kernel);
2291		if (sp != NULL)
2292			key_destroy_sp(sp);
2293	} else {
2294		sp = key_getsp(&spidx);
2295		if (sp != NULL) {
2296			KEY_SP_UNREF(&sp);
2297			IPSECLOG(LOG_DEBUG, "a SP entry exists already.\n");
2298			return key_senderror(so, m, EEXIST);
2299		}
2300	}
2301    }
2302
2303	/* allocation new SP entry */
2304	newsp = _key_msg2sp(xpl0, PFKEY_EXTLEN(xpl0), &error, from_kernel);
2305	if (newsp == NULL) {
2306		return key_senderror(so, m, error);
2307	}
2308
2309	newsp->id = key_getnewspid();
2310	if (newsp->id == 0) {
2311		kmem_free(newsp, sizeof(*newsp));
2312		return key_senderror(so, m, ENOBUFS);
2313	}
2314
2315	newsp->spidx = spidx;
2316	newsp->created = time_uptime;
2317	newsp->lastused = newsp->created;
2318	newsp->lifetime = lft ? lft->sadb_lifetime_addtime : 0;
2319	newsp->validtime = lft ? lft->sadb_lifetime_usetime : 0;
2320	if (from_kernel)
2321		newsp->origin = IPSEC_SPORIGIN_KERNEL;
2322	else
2323		newsp->origin = IPSEC_SPORIGIN_USER;
2324
2325	key_init_sp(newsp);
2326	if (from_kernel)
2327		KEY_SP_REF(newsp);
2328
2329	sadb_x_policy_id = newsp->id;
2330
2331	if (_sp != NULL)
2332		*_sp = newsp;
2333
2334	mutex_enter(&key_spd.lock);
2335	SPLIST_WRITER_INSERT_TAIL(newsp->spidx.dir, newsp);
2336	mutex_exit(&key_spd.lock);
2337	/*
2338	 * We don't have a reference to newsp, so we must not touch newsp from
2339	 * now on.  If you want to do, you must take a reference beforehand.
2340	 */
2341	newsp = NULL;
2342
2343#ifdef notyet
2344	/* delete the entry in key_misc.spacqlist */
2345	if (mhp->msg->sadb_msg_type == SADB_X_SPDUPDATE) {
2346		struct secspacq *spacq = key_getspacq(&spidx);
2347		if (spacq != NULL) {
2348			/* reset counter in order to deletion by timehandler. */
2349			spacq->created = time_uptime;
2350			spacq->count = 0;
2351		}
2352    	}
2353#endif
2354
2355	/* Invalidate all cached SPD pointers in the PCBs. */
2356	ipsec_invalpcbcacheall();
2357
2358#if defined(GATEWAY)
2359	/* Invalidate the ipflow cache, as well. */
2360	ipflow_invalidate_all(0);
2361#ifdef INET6
2362	if (in6_present)
2363		ip6flow_invalidate_all(0);
2364#endif /* INET6 */
2365#endif /* GATEWAY */
2366
2367	key_update_used();
2368
2369    {
2370	struct mbuf *n, *mpolicy;
2371	int off;
2372
2373	/* create new sadb_msg to reply. */
2374	if (lft) {
2375		n = key_gather_mbuf(m, mhp, 2, 5, SADB_EXT_RESERVED,
2376		    SADB_X_EXT_POLICY, SADB_EXT_LIFETIME_HARD,
2377		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2378	} else {
2379		n = key_gather_mbuf(m, mhp, 2, 4, SADB_EXT_RESERVED,
2380		    SADB_X_EXT_POLICY,
2381		    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2382	}
2383
2384	key_fill_replymsg(n, 0);
2385	off = 0;
2386	mpolicy = m_pulldown(n, PFKEY_ALIGN8(sizeof(struct sadb_msg)),
2387	    sizeof(*xpl), &off);
2388	if (mpolicy == NULL) {
2389		/* n is already freed */
2390		/*
2391		 * valid sp has been created, so we does not overwrite _sp
2392		 * NULL here. let caller decide to use the sp or not.
2393		 */
2394		return key_senderror(so, m, ENOBUFS);
2395	}
2396	xpl = (struct sadb_x_policy *)(mtod(mpolicy, char *) + off);
2397	if (xpl->sadb_x_policy_exttype != SADB_X_EXT_POLICY) {
2398		m_freem(n);
2399		/* ditto */
2400		return key_senderror(so, m, EINVAL);
2401	}
2402
2403	xpl->sadb_x_policy_id = sadb_x_policy_id;
2404
2405	m_freem(m);
2406	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2407    }
2408}
2409
2410/*
2411 * SADB_X_SPDADD, SADB_X_SPDSETIDX or SADB_X_SPDUPDATE processing
2412 * add an entry to SP database, when received
2413 *   <base, address(SD), (lifetime(H),) policy>
2414 * from the user(?).
2415 * Adding to SP database,
2416 * and send
2417 *   <base, address(SD), (lifetime(H),) policy>
2418 * to the socket which was send.
2419 *
2420 * SPDADD set a unique policy entry.
2421 * SPDSETIDX like SPDADD without a part of policy requests.
2422 * SPDUPDATE replace a unique policy entry.
2423 *
2424 * m will always be freed.
2425 */
2426static int
2427key_api_spdadd(struct socket *so, struct mbuf *m,
2428	       const struct sadb_msghdr *mhp)
2429{
2430
2431	return key_spdadd(so, m, mhp, NULL, false);
2432}
2433
2434struct secpolicy *
2435key_kpi_spdadd(struct mbuf *m)
2436{
2437	struct sadb_msghdr mh;
2438	int error;
2439	struct secpolicy *sp = NULL;
2440
2441	error = key_align(m, &mh);
2442	if (error)
2443		return NULL;
2444
2445	error = key_spdadd(NULL, m, &mh, &sp, true);
2446	if (error) {
2447		/*
2448		 * Currently, when key_spdadd() cannot send a PFKEY message
2449		 * which means SP has been created, key_spdadd() returns error
2450		 * although SP is created successfully.
2451		 * Kernel components would not care PFKEY messages, so return
2452		 * the "sp" regardless of error code. key_spdadd() overwrites
2453		 * the argument only if SP  is created successfully.
2454		 */
2455	}
2456	return sp;
2457}
2458
2459/*
2460 * get new policy id.
2461 * OUT:
2462 *	0:	failure.
2463 *	others: success.
2464 */
2465static u_int32_t
2466key_getnewspid(void)
2467{
2468	u_int32_t newid = 0;
2469	int count = key_spi_trycnt;	/* XXX */
2470	struct secpolicy *sp;
2471
2472	/* when requesting to allocate spi ranged */
2473	while (count--) {
2474		newid = (policy_id = (policy_id == ~0 ? 1 : policy_id + 1));
2475
2476		sp = key_getspbyid(newid);
2477		if (sp == NULL)
2478			break;
2479
2480		KEY_SP_UNREF(&sp);
2481	}
2482
2483	if (count == 0 || newid == 0) {
2484		IPSECLOG(LOG_DEBUG, "to allocate policy id is failed.\n");
2485		return 0;
2486	}
2487
2488	return newid;
2489}
2490
2491/*
2492 * SADB_SPDDELETE processing
2493 * receive
2494 *   <base, address(SD), policy(*)>
2495 * from the user(?), and set SADB_SASTATE_DEAD,
2496 * and send,
2497 *   <base, address(SD), policy(*)>
2498 * to the ikmpd.
2499 * policy(*) including direction of policy.
2500 *
2501 * m will always be freed.
2502 */
2503static int
2504key_api_spddelete(struct socket *so, struct mbuf *m,
2505              const struct sadb_msghdr *mhp)
2506{
2507	struct sadb_x_policy *xpl0;
2508	struct secpolicyindex spidx;
2509	struct secpolicy *sp;
2510
2511	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
2512	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
2513	    mhp->ext[SADB_X_EXT_POLICY] == NULL) {
2514		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2515		return key_senderror(so, m, EINVAL);
2516	}
2517	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
2518	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
2519	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2520		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2521		return key_senderror(so, m, EINVAL);
2522	}
2523
2524	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
2525
2526	/* checking the direction. */
2527	switch (xpl0->sadb_x_policy_dir) {
2528	case IPSEC_DIR_INBOUND:
2529	case IPSEC_DIR_OUTBOUND:
2530		break;
2531	default:
2532		IPSECLOG(LOG_DEBUG, "Invalid SP direction.\n");
2533		return key_senderror(so, m, EINVAL);
2534	}
2535
2536	/* make secindex */
2537	key_init_spidx_bymsghdr(&spidx, mhp);
2538
2539	/* Is there SP in SPD ? */
2540	sp = key_lookup_and_remove_sp(&spidx, false);
2541	if (sp == NULL) {
2542		IPSECLOG(LOG_DEBUG, "no SP found.\n");
2543		return key_senderror(so, m, EINVAL);
2544	}
2545
2546	/* save policy id to buffer to be returned. */
2547	xpl0->sadb_x_policy_id = sp->id;
2548
2549	key_destroy_sp(sp);
2550
2551	/* We're deleting policy; no need to invalidate the ipflow cache. */
2552
2553    {
2554	struct mbuf *n;
2555
2556	/* create new sadb_msg to reply. */
2557	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
2558	    SADB_X_EXT_POLICY, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
2559	key_fill_replymsg(n, 0);
2560	m_freem(m);
2561	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2562    }
2563}
2564
2565static struct mbuf *
2566key_alloc_mbuf_simple(int len, int mflag)
2567{
2568	struct mbuf *n;
2569
2570	KASSERT(mflag == M_NOWAIT || (mflag == M_WAITOK && !cpu_softintr_p()));
2571
2572	MGETHDR(n, mflag, MT_DATA);
2573	if (n && len > MHLEN) {
2574		MCLGET(n, mflag);
2575		if ((n->m_flags & M_EXT) == 0) {
2576			m_freem(n);
2577			n = NULL;
2578		}
2579	}
2580	return n;
2581}
2582
2583/*
2584 * SADB_SPDDELETE2 processing
2585 * receive
2586 *   <base, policy(*)>
2587 * from the user(?), and set SADB_SASTATE_DEAD,
2588 * and send,
2589 *   <base, policy(*)>
2590 * to the ikmpd.
2591 * policy(*) including direction of policy.
2592 *
2593 * m will always be freed.
2594 */
2595static int
2596key_spddelete2(struct socket *so, struct mbuf *m,
2597	       const struct sadb_msghdr *mhp, bool from_kernel)
2598{
2599	u_int32_t id;
2600	struct secpolicy *sp;
2601	const struct sadb_x_policy *xpl;
2602
2603	if (mhp->ext[SADB_X_EXT_POLICY] == NULL ||
2604	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2605		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2606		return key_senderror(so, m, EINVAL);
2607	}
2608
2609	xpl = mhp->ext[SADB_X_EXT_POLICY];
2610	id = xpl->sadb_x_policy_id;
2611
2612	/* Is there SP in SPD ? */
2613	sp = key_lookupbyid_and_remove_sp(id, from_kernel);
2614	if (sp == NULL) {
2615		IPSECLOG(LOG_DEBUG, "no SP found id:%u.\n", id);
2616		return key_senderror(so, m, EINVAL);
2617	}
2618
2619	key_destroy_sp(sp);
2620
2621	/* We're deleting policy; no need to invalidate the ipflow cache. */
2622
2623    {
2624	struct mbuf *n, *nn;
2625	int off, len;
2626
2627	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MCLBYTES);
2628
2629	/* create new sadb_msg to reply. */
2630	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2631
2632	n = key_alloc_mbuf_simple(len, M_WAITOK);
2633	n->m_len = len;
2634	n->m_next = NULL;
2635	off = 0;
2636
2637	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
2638	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
2639
2640	KASSERTMSG(off == len, "length inconsistency");
2641
2642	n->m_next = m_copym(m, mhp->extoff[SADB_X_EXT_POLICY],
2643	    mhp->extlen[SADB_X_EXT_POLICY], M_WAITOK);
2644
2645	n->m_pkthdr.len = 0;
2646	for (nn = n; nn; nn = nn->m_next)
2647		n->m_pkthdr.len += nn->m_len;
2648
2649	key_fill_replymsg(n, 0);
2650	m_freem(m);
2651	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
2652    }
2653}
2654
2655/*
2656 * SADB_SPDDELETE2 processing
2657 * receive
2658 *   <base, policy(*)>
2659 * from the user(?), and set SADB_SASTATE_DEAD,
2660 * and send,
2661 *   <base, policy(*)>
2662 * to the ikmpd.
2663 * policy(*) including direction of policy.
2664 *
2665 * m will always be freed.
2666 */
2667static int
2668key_api_spddelete2(struct socket *so, struct mbuf *m,
2669	       const struct sadb_msghdr *mhp)
2670{
2671
2672	return key_spddelete2(so, m, mhp, false);
2673}
2674
2675int
2676key_kpi_spddelete2(struct mbuf *m)
2677{
2678	struct sadb_msghdr mh;
2679	int error;
2680
2681	error = key_align(m, &mh);
2682	if (error)
2683		return EINVAL;
2684
2685	return key_spddelete2(NULL, m, &mh, true);
2686}
2687
2688/*
2689 * SADB_X_GET processing
2690 * receive
2691 *   <base, policy(*)>
2692 * from the user(?),
2693 * and send,
2694 *   <base, address(SD), policy>
2695 * to the ikmpd.
2696 * policy(*) including direction of policy.
2697 *
2698 * m will always be freed.
2699 */
2700static int
2701key_api_spdget(struct socket *so, struct mbuf *m,
2702	   const struct sadb_msghdr *mhp)
2703{
2704	u_int32_t id;
2705	struct secpolicy *sp;
2706	struct mbuf *n;
2707	const struct sadb_x_policy *xpl;
2708
2709	if (mhp->ext[SADB_X_EXT_POLICY] == NULL ||
2710	    mhp->extlen[SADB_X_EXT_POLICY] < sizeof(struct sadb_x_policy)) {
2711		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
2712		return key_senderror(so, m, EINVAL);
2713	}
2714
2715	xpl = mhp->ext[SADB_X_EXT_POLICY];
2716	id = xpl->sadb_x_policy_id;
2717
2718	/* Is there SP in SPD ? */
2719	sp = key_getspbyid(id);
2720	if (sp == NULL) {
2721		IPSECLOG(LOG_DEBUG, "no SP found id:%u.\n", id);
2722		return key_senderror(so, m, ENOENT);
2723	}
2724
2725	n = key_setdumpsp(sp, SADB_X_SPDGET, mhp->msg->sadb_msg_seq,
2726	    mhp->msg->sadb_msg_pid);
2727	KEY_SP_UNREF(&sp); /* ref gained by key_getspbyid */
2728	m_freem(m);
2729	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
2730}
2731
2732#ifdef notyet
2733/*
2734 * SADB_X_SPDACQUIRE processing.
2735 * Acquire policy and SA(s) for a *OUTBOUND* packet.
2736 * send
2737 *   <base, policy(*)>
2738 * to KMD, and expect to receive
2739 *   <base> with SADB_X_SPDACQUIRE if error occurred,
2740 * or
2741 *   <base, policy>
2742 * with SADB_X_SPDUPDATE from KMD by PF_KEY.
2743 * policy(*) is without policy requests.
2744 *
2745 *    0     : succeed
2746 *    others: error number
2747 */
2748int
2749key_spdacquire(const struct secpolicy *sp)
2750{
2751	struct mbuf *result = NULL, *m;
2752	struct secspacq *newspacq;
2753	int error;
2754
2755	KASSERT(sp != NULL);
2756	KASSERTMSG(sp->req == NULL, "called but there is request");
2757	KASSERTMSG(sp->policy == IPSEC_POLICY_IPSEC,
2758	    "policy mismathed. IPsec is expected");
2759
2760	/* Get an entry to check whether sent message or not. */
2761	newspacq = key_getspacq(&sp->spidx);
2762	if (newspacq != NULL) {
2763		if (key_blockacq_count < newspacq->count) {
2764			/* reset counter and do send message. */
2765			newspacq->count = 0;
2766		} else {
2767			/* increment counter and do nothing. */
2768			newspacq->count++;
2769			return 0;
2770		}
2771	} else {
2772		/* make new entry for blocking to send SADB_ACQUIRE. */
2773		newspacq = key_newspacq(&sp->spidx);
2774		if (newspacq == NULL)
2775			return ENOBUFS;
2776
2777		/* add to key_misc.acqlist */
2778		LIST_INSERT_HEAD(&key_misc.spacqlist, newspacq, chain);
2779	}
2780
2781	/* create new sadb_msg to reply. */
2782	m = key_setsadbmsg(SADB_X_SPDACQUIRE, 0, 0, 0, 0, 0);
2783	if (!m) {
2784		error = ENOBUFS;
2785		goto fail;
2786	}
2787	result = m;
2788
2789	result->m_pkthdr.len = 0;
2790	for (m = result; m; m = m->m_next)
2791		result->m_pkthdr.len += m->m_len;
2792
2793	mtod(result, struct sadb_msg *)->sadb_msg_len =
2794	    PFKEY_UNIT64(result->m_pkthdr.len);
2795
2796	return key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
2797
2798fail:
2799	if (result)
2800		m_freem(result);
2801	return error;
2802}
2803#endif /* notyet */
2804
2805/*
2806 * SADB_SPDFLUSH processing
2807 * receive
2808 *   <base>
2809 * from the user, and free all entries in secpctree.
2810 * and send,
2811 *   <base>
2812 * to the user.
2813 * NOTE: what to do is only marking SADB_SASTATE_DEAD.
2814 *
2815 * m will always be freed.
2816 */
2817static int
2818key_api_spdflush(struct socket *so, struct mbuf *m,
2819	     const struct sadb_msghdr *mhp)
2820{
2821	struct sadb_msg *newmsg;
2822	struct secpolicy *sp;
2823	u_int dir;
2824
2825	if (m->m_len != PFKEY_ALIGN8(sizeof(struct sadb_msg)))
2826		return key_senderror(so, m, EINVAL);
2827
2828	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2829	    retry:
2830		mutex_enter(&key_spd.lock);
2831		SPLIST_WRITER_FOREACH(sp, dir) {
2832			KASSERTMSG(sp->state != IPSEC_SPSTATE_DEAD,
2833			    "sp->state=%u", sp->state);
2834			/*
2835			 * Userlang programs can remove SPs created by userland
2836			 * probrams only, that is, they cannot remove SPs
2837			 * created in kernel(e.g. ipsec(4) I/F).
2838			 */
2839			if (sp->origin == IPSEC_SPORIGIN_USER) {
2840				key_unlink_sp(sp);
2841				mutex_exit(&key_spd.lock);
2842				key_destroy_sp(sp);
2843				goto retry;
2844			}
2845		}
2846		mutex_exit(&key_spd.lock);
2847	}
2848
2849	/* We're deleting policy; no need to invalidate the ipflow cache. */
2850
2851	if (sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
2852		IPSECLOG(LOG_DEBUG, "No more memory.\n");
2853		return key_senderror(so, m, ENOBUFS);
2854	}
2855
2856	if (m->m_next)
2857		m_freem(m->m_next);
2858	m->m_next = NULL;
2859	m->m_pkthdr.len = m->m_len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
2860	newmsg = mtod(m, struct sadb_msg *);
2861	newmsg->sadb_msg_errno = 0;
2862	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
2863
2864	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
2865}
2866
2867static struct sockaddr key_src = {
2868	.sa_len = 2,
2869	.sa_family = PF_KEY,
2870};
2871
2872static struct mbuf *
2873key_setspddump_chain(int *errorp, int *lenp, pid_t pid)
2874{
2875	struct secpolicy *sp;
2876	int cnt;
2877	u_int dir;
2878	struct mbuf *m, *n, *prev;
2879	int totlen;
2880
2881	KASSERT(mutex_owned(&key_spd.lock));
2882
2883	*lenp = 0;
2884
2885	/* search SPD entry and get buffer size. */
2886	cnt = 0;
2887	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2888		SPLIST_WRITER_FOREACH(sp, dir) {
2889			cnt++;
2890		}
2891	}
2892
2893	if (cnt == 0) {
2894		*errorp = ENOENT;
2895		return (NULL);
2896	}
2897
2898	m = NULL;
2899	prev = m;
2900	totlen = 0;
2901	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
2902		SPLIST_WRITER_FOREACH(sp, dir) {
2903			--cnt;
2904			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, pid);
2905
2906			totlen += n->m_pkthdr.len;
2907			if (!m) {
2908				m = n;
2909			} else {
2910				prev->m_nextpkt = n;
2911			}
2912			prev = n;
2913		}
2914	}
2915
2916	*lenp = totlen;
2917	*errorp = 0;
2918	return (m);
2919}
2920
2921/*
2922 * SADB_SPDDUMP processing
2923 * receive
2924 *   <base>
2925 * from the user, and dump all SP leaves
2926 * and send,
2927 *   <base> .....
2928 * to the ikmpd.
2929 *
2930 * m will always be freed.
2931 */
2932static int
2933key_api_spddump(struct socket *so, struct mbuf *m0,
2934 	    const struct sadb_msghdr *mhp)
2935{
2936	struct mbuf *n;
2937	int error, len;
2938	int ok;
2939	pid_t pid;
2940
2941	pid = mhp->msg->sadb_msg_pid;
2942	/*
2943	 * If the requestor has insufficient socket-buffer space
2944	 * for the entire chain, nobody gets any response to the DUMP.
2945	 * XXX For now, only the requestor ever gets anything.
2946	 * Moreover, if the requestor has any space at all, they receive
2947	 * the entire chain, otherwise the request is refused with  ENOBUFS.
2948	 */
2949	if (sbspace(&so->so_rcv) <= 0) {
2950		return key_senderror(so, m0, ENOBUFS);
2951	}
2952
2953	mutex_enter(&key_spd.lock);
2954	n = key_setspddump_chain(&error, &len, pid);
2955	mutex_exit(&key_spd.lock);
2956
2957	if (n == NULL) {
2958		return key_senderror(so, m0, ENOENT);
2959	}
2960	{
2961		uint64_t *ps = PFKEY_STAT_GETREF();
2962		ps[PFKEY_STAT_IN_TOTAL]++;
2963		ps[PFKEY_STAT_IN_BYTES] += len;
2964		PFKEY_STAT_PUTREF();
2965	}
2966
2967	/*
2968	 * PF_KEY DUMP responses are no longer broadcast to all PF_KEY sockets.
2969	 * The requestor receives either the entire chain, or an
2970	 * error message with ENOBUFS.
2971	 */
2972
2973	/*
2974	 * sbappendchainwith record takes the chain of entries, one
2975	 * packet-record per SPD entry, prepends the key_src sockaddr
2976	 * to each packet-record, links the sockaddr mbufs into a new
2977	 * list of records, then   appends the entire resulting
2978	 * list to the requesting socket.
2979	 */
2980	ok = sbappendaddrchain(&so->so_rcv, (struct sockaddr *)&key_src, n,
2981	    SB_PRIO_ONESHOT_OVERFLOW);
2982
2983	if (!ok) {
2984		PFKEY_STATINC(PFKEY_STAT_IN_NOMEM);
2985		m_freem(n);
2986		return key_senderror(so, m0, ENOBUFS);
2987	}
2988
2989	m_freem(m0);
2990	return error;
2991}
2992
2993/*
2994 * SADB_X_NAT_T_NEW_MAPPING. Unused by racoon as of 2005/04/23
2995 */
2996static int
2997key_api_nat_map(struct socket *so, struct mbuf *m,
2998	    const struct sadb_msghdr *mhp)
2999{
3000	struct sadb_x_nat_t_type *type;
3001	struct sadb_x_nat_t_port *sport;
3002	struct sadb_x_nat_t_port *dport;
3003	struct sadb_address *iaddr, *raddr;
3004	struct sadb_x_nat_t_frag *frag;
3005
3006	if (mhp->ext[SADB_X_EXT_NAT_T_TYPE] == NULL ||
3007	    mhp->ext[SADB_X_EXT_NAT_T_SPORT] == NULL ||
3008	    mhp->ext[SADB_X_EXT_NAT_T_DPORT] == NULL) {
3009		IPSECLOG(LOG_DEBUG, "invalid message.\n");
3010		return key_senderror(so, m, EINVAL);
3011	}
3012	if ((mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) ||
3013	    (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) ||
3014	    (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport))) {
3015		IPSECLOG(LOG_DEBUG, "invalid message.\n");
3016		return key_senderror(so, m, EINVAL);
3017	}
3018
3019	if ((mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL) &&
3020	    (mhp->extlen[SADB_X_EXT_NAT_T_OAI] < sizeof(*iaddr))) {
3021		IPSECLOG(LOG_DEBUG, "invalid message\n");
3022		return key_senderror(so, m, EINVAL);
3023	}
3024
3025	if ((mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL) &&
3026	    (mhp->extlen[SADB_X_EXT_NAT_T_OAR] < sizeof(*raddr))) {
3027		IPSECLOG(LOG_DEBUG, "invalid message\n");
3028		return key_senderror(so, m, EINVAL);
3029	}
3030
3031	if ((mhp->ext[SADB_X_EXT_NAT_T_FRAG] != NULL) &&
3032	    (mhp->extlen[SADB_X_EXT_NAT_T_FRAG] < sizeof(*frag))) {
3033		IPSECLOG(LOG_DEBUG, "invalid message\n");
3034		return key_senderror(so, m, EINVAL);
3035	}
3036
3037	type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
3038	sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
3039	dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
3040	iaddr = mhp->ext[SADB_X_EXT_NAT_T_OAI];
3041	raddr = mhp->ext[SADB_X_EXT_NAT_T_OAR];
3042	frag = mhp->ext[SADB_X_EXT_NAT_T_FRAG];
3043
3044	/*
3045	 * XXX handle that, it should also contain a SA, or anything
3046	 * that enable to update the SA information.
3047	 */
3048
3049	return 0;
3050}
3051
3052/*
3053 * Never return NULL.
3054 */
3055static struct mbuf *
3056key_setdumpsp(struct secpolicy *sp, u_int8_t type, u_int32_t seq, pid_t pid)
3057{
3058	struct mbuf *result = NULL, *m;
3059
3060	KASSERT(!cpu_softintr_p());
3061
3062	m = key_setsadbmsg(type, 0, SADB_SATYPE_UNSPEC, seq, pid,
3063	    key_sp_refcnt(sp), M_WAITOK);
3064	result = m;
3065
3066	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3067	    &sp->spidx.src.sa, sp->spidx.prefs, sp->spidx.ul_proto, M_WAITOK);
3068	m_cat(result, m);
3069
3070	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3071	    &sp->spidx.dst.sa, sp->spidx.prefd, sp->spidx.ul_proto, M_WAITOK);
3072	m_cat(result, m);
3073
3074	m = key_sp2msg(sp, M_WAITOK);
3075	m_cat(result, m);
3076
3077	KASSERT(result->m_flags & M_PKTHDR);
3078	KASSERT(result->m_len >= sizeof(struct sadb_msg));
3079
3080	result->m_pkthdr.len = 0;
3081	for (m = result; m; m = m->m_next)
3082		result->m_pkthdr.len += m->m_len;
3083
3084	mtod(result, struct sadb_msg *)->sadb_msg_len =
3085	    PFKEY_UNIT64(result->m_pkthdr.len);
3086
3087	return result;
3088}
3089
3090/*
3091 * get PFKEY message length for security policy and request.
3092 */
3093static u_int
3094key_getspreqmsglen(const struct secpolicy *sp)
3095{
3096	u_int tlen;
3097
3098	tlen = sizeof(struct sadb_x_policy);
3099
3100	/* if is the policy for ipsec ? */
3101	if (sp->policy != IPSEC_POLICY_IPSEC)
3102		return tlen;
3103
3104	/* get length of ipsec requests */
3105    {
3106	const struct ipsecrequest *isr;
3107	int len;
3108
3109	for (isr = sp->req; isr != NULL; isr = isr->next) {
3110		len = sizeof(struct sadb_x_ipsecrequest)
3111		    + isr->saidx.src.sa.sa_len + isr->saidx.dst.sa.sa_len;
3112
3113		tlen += PFKEY_ALIGN8(len);
3114	}
3115    }
3116
3117	return tlen;
3118}
3119
3120/*
3121 * SADB_SPDEXPIRE processing
3122 * send
3123 *   <base, address(SD), lifetime(CH), policy>
3124 * to KMD by PF_KEY.
3125 *
3126 * OUT:	0	: succeed
3127 *	others	: error number
3128 */
3129static int
3130key_spdexpire(struct secpolicy *sp)
3131{
3132	int s;
3133	struct mbuf *result = NULL, *m;
3134	int len;
3135	int error = -1;
3136	struct sadb_lifetime *lt;
3137
3138	/* XXX: Why do we lock ? */
3139	s = splsoftnet();	/*called from softclock()*/
3140
3141	KASSERT(sp != NULL);
3142
3143	/* set msg header */
3144	m = key_setsadbmsg(SADB_X_SPDEXPIRE, 0, 0, 0, 0, 0, M_WAITOK);
3145	result = m;
3146
3147	/* create lifetime extension (current and hard) */
3148	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
3149	m = key_alloc_mbuf(len, M_WAITOK);
3150	KASSERT(m->m_next == NULL);
3151
3152	memset(mtod(m, void *), 0, len);
3153	lt = mtod(m, struct sadb_lifetime *);
3154	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
3155	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
3156	lt->sadb_lifetime_allocations = 0;
3157	lt->sadb_lifetime_bytes = 0;
3158	lt->sadb_lifetime_addtime = time_mono_to_wall(sp->created);
3159	lt->sadb_lifetime_usetime = time_mono_to_wall(sp->lastused);
3160	lt = (struct sadb_lifetime *)(mtod(m, char *) + len / 2);
3161	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
3162	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_HARD;
3163	lt->sadb_lifetime_allocations = 0;
3164	lt->sadb_lifetime_bytes = 0;
3165	lt->sadb_lifetime_addtime = sp->lifetime;
3166	lt->sadb_lifetime_usetime = sp->validtime;
3167	m_cat(result, m);
3168
3169	/* set sadb_address for source */
3170	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sp->spidx.src.sa,
3171	    sp->spidx.prefs, sp->spidx.ul_proto, M_WAITOK);
3172	m_cat(result, m);
3173
3174	/* set sadb_address for destination */
3175	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sp->spidx.dst.sa,
3176	    sp->spidx.prefd, sp->spidx.ul_proto, M_WAITOK);
3177	m_cat(result, m);
3178
3179	/* set secpolicy */
3180	m = key_sp2msg(sp, M_WAITOK);
3181	m_cat(result, m);
3182
3183	KASSERT(result->m_flags & M_PKTHDR);
3184	KASSERT(result->m_len >= sizeof(struct sadb_msg));
3185
3186	result->m_pkthdr.len = 0;
3187	for (m = result; m; m = m->m_next)
3188		result->m_pkthdr.len += m->m_len;
3189
3190	mtod(result, struct sadb_msg *)->sadb_msg_len =
3191	    PFKEY_UNIT64(result->m_pkthdr.len);
3192
3193	error = key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
3194	splx(s);
3195	return error;
3196}
3197
3198/* %%% SAD management */
3199/*
3200 * allocating a memory for new SA head, and copy from the values of mhp.
3201 * OUT:	NULL	: failure due to the lack of memory.
3202 *	others	: pointer to new SA head.
3203 */
3204static struct secashead *
3205key_newsah(const struct secasindex *saidx)
3206{
3207	struct secashead *newsah;
3208	int i;
3209
3210	KASSERT(saidx != NULL);
3211
3212	newsah = kmem_zalloc(sizeof(struct secashead), KM_SLEEP);
3213	for (i = 0; i < __arraycount(newsah->savlist); i++)
3214		PSLIST_INIT(&newsah->savlist[i]);
3215	newsah->saidx = *saidx;
3216
3217	localcount_init(&newsah->localcount);
3218	/* Take a reference for the caller */
3219	localcount_acquire(&newsah->localcount);
3220
3221	/* Add to the sah list */
3222	SAHLIST_ENTRY_INIT(newsah);
3223	newsah->state = SADB_SASTATE_MATURE;
3224	mutex_enter(&key_sad.lock);
3225	SAHLIST_WRITER_INSERT_HEAD(newsah);
3226	mutex_exit(&key_sad.lock);
3227
3228	return newsah;
3229}
3230
3231static bool
3232key_sah_has_sav(struct secashead *sah)
3233{
3234	u_int state;
3235
3236	KASSERT(mutex_owned(&key_sad.lock));
3237
3238	SASTATE_ANY_FOREACH(state) {
3239		if (!SAVLIST_WRITER_EMPTY(sah, state))
3240			return true;
3241	}
3242
3243	return false;
3244}
3245
3246static void
3247key_unlink_sah(struct secashead *sah)
3248{
3249
3250	KASSERT(!cpu_softintr_p());
3251	KASSERT(mutex_owned(&key_sad.lock));
3252	KASSERTMSG(sah->state == SADB_SASTATE_DEAD, "sah->state=%u", sah->state);
3253
3254	/* Remove from the sah list */
3255	SAHLIST_WRITER_REMOVE(sah);
3256
3257	KDASSERT(mutex_ownable(softnet_lock));
3258	key_sad_pserialize_perform();
3259
3260	localcount_drain(&sah->localcount, &key_sad.cv_lc, &key_sad.lock);
3261}
3262
3263static void
3264key_destroy_sah(struct secashead *sah)
3265{
3266
3267	rtcache_free(&sah->sa_route);
3268
3269	SAHLIST_ENTRY_DESTROY(sah);
3270	localcount_fini(&sah->localcount);
3271
3272	if (sah->idents != NULL)
3273		kmem_free(sah->idents, sah->idents_len);
3274	if (sah->identd != NULL)
3275		kmem_free(sah->identd, sah->identd_len);
3276
3277	kmem_free(sah, sizeof(*sah));
3278}
3279
3280/*
3281 * allocating a new SA with LARVAL state.
3282 * key_api_add() and key_api_getspi() call,
3283 * and copy the values of mhp into new buffer.
3284 * When SAD message type is GETSPI:
3285 *	to set sequence number from acq_seq++,
3286 *	to set zero to SPI.
3287 *	not to call key_setsaval().
3288 * OUT:	NULL	: fail
3289 *	others	: pointer to new secasvar.
3290 *
3291 * does not modify mbuf.  does not free mbuf on error.
3292 */
3293static struct secasvar *
3294key_newsav(struct mbuf *m, const struct sadb_msghdr *mhp,
3295    int *errp, int proto, const char* where, int tag)
3296{
3297	struct secasvar *newsav;
3298	const struct sadb_sa *xsa;
3299
3300	KASSERT(!cpu_softintr_p());
3301	KASSERT(m != NULL);
3302	KASSERT(mhp != NULL);
3303	KASSERT(mhp->msg != NULL);
3304
3305	newsav = kmem_zalloc(sizeof(struct secasvar), KM_SLEEP);
3306
3307	switch (mhp->msg->sadb_msg_type) {
3308	case SADB_GETSPI:
3309		newsav->spi = 0;
3310
3311#ifdef IPSEC_DOSEQCHECK
3312		/* sync sequence number */
3313		if (mhp->msg->sadb_msg_seq == 0)
3314			newsav->seq =
3315			    (acq_seq = (acq_seq == ~0 ? 1 : ++acq_seq));
3316		else
3317#endif
3318			newsav->seq = mhp->msg->sadb_msg_seq;
3319		break;
3320
3321	case SADB_ADD:
3322		/* sanity check */
3323		if (mhp->ext[SADB_EXT_SA] == NULL) {
3324			IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
3325			*errp = EINVAL;
3326			goto error;
3327		}
3328		xsa = mhp->ext[SADB_EXT_SA];
3329		newsav->spi = xsa->sadb_sa_spi;
3330		newsav->seq = mhp->msg->sadb_msg_seq;
3331		break;
3332	default:
3333		*errp = EINVAL;
3334		goto error;
3335	}
3336
3337	/* copy sav values */
3338	if (mhp->msg->sadb_msg_type != SADB_GETSPI) {
3339		*errp = key_setsaval(newsav, m, mhp);
3340		if (*errp)
3341			goto error;
3342	} else {
3343		/* We don't allow lft_c to be NULL */
3344		newsav->lft_c = kmem_zalloc(sizeof(struct sadb_lifetime),
3345		    KM_SLEEP);
3346		newsav->lft_c_counters_percpu =
3347		    percpu_alloc(sizeof(lifetime_counters_t));
3348	}
3349
3350	/* reset created */
3351	newsav->created = time_uptime;
3352	newsav->pid = mhp->msg->sadb_msg_pid;
3353
3354	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
3355	    "DP from %s:%u return SA:%p spi=%#x proto=%d\n",
3356	    where, tag, newsav, ntohl(newsav->spi), proto);
3357	return newsav;
3358
3359error:
3360	KASSERT(*errp != 0);
3361	kmem_free(newsav, sizeof(*newsav));
3362	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
3363	    "DP from %s:%u return SA:NULL\n", where, tag);
3364	return NULL;
3365}
3366
3367
3368static void
3369key_clear_xform(struct secasvar *sav)
3370{
3371
3372	/*
3373	 * Cleanup xform state.  Note that zeroize'ing causes the
3374	 * keys to be cleared; otherwise we must do it ourself.
3375	 */
3376	if (sav->tdb_xform != NULL) {
3377		sav->tdb_xform->xf_zeroize(sav);
3378		sav->tdb_xform = NULL;
3379	} else {
3380		if (sav->key_auth != NULL)
3381			explicit_memset(_KEYBUF(sav->key_auth), 0,
3382			    _KEYLEN(sav->key_auth));
3383		if (sav->key_enc != NULL)
3384			explicit_memset(_KEYBUF(sav->key_enc), 0,
3385			    _KEYLEN(sav->key_enc));
3386	}
3387}
3388
3389/*
3390 * free() SA variable entry.
3391 */
3392static void
3393key_delsav(struct secasvar *sav)
3394{
3395
3396	key_clear_xform(sav);
3397	key_freesaval(sav);
3398	kmem_free(sav, sizeof(*sav));
3399}
3400
3401/*
3402 * Must be called in a pserialize read section. A held sah
3403 * must be released by key_sah_unref after use.
3404 */
3405static void
3406key_sah_ref(struct secashead *sah)
3407{
3408
3409	localcount_acquire(&sah->localcount);
3410}
3411
3412/*
3413 * Must be called without holding key_sad.lock because the lock
3414 * would be held in localcount_release.
3415 */
3416static void
3417key_sah_unref(struct secashead *sah)
3418{
3419
3420	KDASSERT(mutex_ownable(&key_sad.lock));
3421
3422	localcount_release(&sah->localcount, &key_sad.cv_lc, &key_sad.lock);
3423}
3424
3425/*
3426 * Search SAD and return sah. Must be called in a pserialize
3427 * read section.
3428 * OUT:
3429 *	NULL	: not found
3430 *	others	: found, pointer to a SA.
3431 */
3432static struct secashead *
3433key_getsah(const struct secasindex *saidx, int flag)
3434{
3435	struct secashead *sah;
3436
3437	SAHLIST_READER_FOREACH_SAIDX(sah, saidx) {
3438		if (sah->state == SADB_SASTATE_DEAD)
3439			continue;
3440		if (key_saidx_match(&sah->saidx, saidx, flag))
3441			return sah;
3442	}
3443
3444	return NULL;
3445}
3446
3447/*
3448 * Search SAD and return sah. If sah is returned, the caller must call
3449 * key_sah_unref to releaset a reference.
3450 * OUT:
3451 *	NULL	: not found
3452 *	others	: found, pointer to a SA.
3453 */
3454static struct secashead *
3455key_getsah_ref(const struct secasindex *saidx, int flag)
3456{
3457	struct secashead *sah;
3458	int s;
3459
3460	s = pserialize_read_enter();
3461	sah = key_getsah(saidx, flag);
3462	if (sah != NULL)
3463		key_sah_ref(sah);
3464	pserialize_read_exit(s);
3465
3466	return sah;
3467}
3468
3469/*
3470 * check not to be duplicated SPI.
3471 * NOTE: this function is too slow due to searching all SAD.
3472 * OUT:
3473 *	NULL	: not found
3474 *	others	: found, pointer to a SA.
3475 */
3476static bool
3477key_checkspidup(const struct secasindex *saidx, u_int32_t spi)
3478{
3479	struct secashead *sah;
3480	struct secasvar *sav;
3481
3482	/* check address family */
3483	if (saidx->src.sa.sa_family != saidx->dst.sa.sa_family) {
3484		IPSECLOG(LOG_DEBUG,
3485		    "address family mismatched src %u, dst %u.\n",
3486		    saidx->src.sa.sa_family, saidx->dst.sa.sa_family);
3487		return false;
3488	}
3489
3490	/* check all SAD */
3491	/* key_ismyaddr may sleep, so use mutex, not pserialize, here. */
3492	mutex_enter(&key_sad.lock);
3493	SAHLIST_WRITER_FOREACH(sah) {
3494		if (!key_ismyaddr((struct sockaddr *)&sah->saidx.dst))
3495			continue;
3496		sav = key_getsavbyspi(sah, spi);
3497		if (sav != NULL) {
3498			KEY_SA_UNREF(&sav);
3499			mutex_exit(&key_sad.lock);
3500			return true;
3501		}
3502	}
3503	mutex_exit(&key_sad.lock);
3504
3505	return false;
3506}
3507
3508/*
3509 * search SAD litmited alive SA, protocol, SPI.
3510 * OUT:
3511 *	NULL	: not found
3512 *	others	: found, pointer to a SA.
3513 */
3514static struct secasvar *
3515key_getsavbyspi(struct secashead *sah, u_int32_t spi)
3516{
3517	struct secasvar *sav = NULL;
3518	u_int state;
3519	int s;
3520
3521	/* search all status */
3522	s = pserialize_read_enter();
3523	SASTATE_ALIVE_FOREACH(state) {
3524		SAVLIST_READER_FOREACH(sav, sah, state) {
3525			/* sanity check */
3526			if (sav->state != state) {
3527				IPSECLOG(LOG_DEBUG,
3528				    "invalid sav->state (queue: %d SA: %d)\n",
3529				    state, sav->state);
3530				continue;
3531			}
3532
3533			if (sav->spi == spi) {
3534				KEY_SA_REF(sav);
3535				goto out;
3536			}
3537		}
3538	}
3539out:
3540	pserialize_read_exit(s);
3541
3542	return sav;
3543}
3544
3545/*
3546 * Search SAD litmited alive SA by an SPI and remove it from a list.
3547 * OUT:
3548 *	NULL	: not found
3549 *	others	: found, pointer to a SA.
3550 */
3551static struct secasvar *
3552key_lookup_and_remove_sav(struct secashead *sah, u_int32_t spi,
3553    const struct secasvar *hint)
3554{
3555	struct secasvar *sav = NULL;
3556	u_int state;
3557
3558	/* search all status */
3559	mutex_enter(&key_sad.lock);
3560	SASTATE_ALIVE_FOREACH(state) {
3561		SAVLIST_WRITER_FOREACH(sav, sah, state) {
3562			KASSERT(sav->state == state);
3563
3564			if (sav->spi == spi) {
3565				if (hint != NULL && hint != sav)
3566					continue;
3567				sav->state = SADB_SASTATE_DEAD;
3568				SAVLIST_WRITER_REMOVE(sav);
3569				SAVLUT_WRITER_REMOVE(sav);
3570				goto out;
3571			}
3572		}
3573	}
3574out:
3575	mutex_exit(&key_sad.lock);
3576
3577	return sav;
3578}
3579
3580/*
3581 * Free allocated data to member variables of sav:
3582 * sav->replay, sav->key_* and sav->lft_*.
3583 */
3584static void
3585key_freesaval(struct secasvar *sav)
3586{
3587
3588	KASSERTMSG(key_sa_refcnt(sav) == 0, "key_sa_refcnt(sav)=%u",
3589	    key_sa_refcnt(sav));
3590
3591	if (sav->replay != NULL)
3592		kmem_free(sav->replay, sav->replay_len);
3593	if (sav->key_auth != NULL)
3594		kmem_free(sav->key_auth, sav->key_auth_len);
3595	if (sav->key_enc != NULL)
3596		kmem_free(sav->key_enc, sav->key_enc_len);
3597	if (sav->lft_c_counters_percpu != NULL) {
3598		percpu_free(sav->lft_c_counters_percpu,
3599		    sizeof(lifetime_counters_t));
3600	}
3601	if (sav->lft_c != NULL)
3602		kmem_free(sav->lft_c, sizeof(*(sav->lft_c)));
3603	if (sav->lft_h != NULL)
3604		kmem_free(sav->lft_h, sizeof(*(sav->lft_h)));
3605	if (sav->lft_s != NULL)
3606		kmem_free(sav->lft_s, sizeof(*(sav->lft_s)));
3607}
3608
3609/*
3610 * copy SA values from PF_KEY message except *SPI, SEQ, PID, STATE and TYPE*.
3611 * You must update these if need.
3612 * OUT:	0:	success.
3613 *	!0:	failure.
3614 *
3615 * does not modify mbuf.  does not free mbuf on error.
3616 */
3617static int
3618key_setsaval(struct secasvar *sav, struct mbuf *m,
3619	     const struct sadb_msghdr *mhp)
3620{
3621	int error = 0;
3622
3623	KASSERT(!cpu_softintr_p());
3624	KASSERT(m != NULL);
3625	KASSERT(mhp != NULL);
3626	KASSERT(mhp->msg != NULL);
3627
3628	/* We shouldn't initialize sav variables while someone uses it. */
3629	KASSERTMSG(key_sa_refcnt(sav) == 0, "key_sa_refcnt(sav)=%u",
3630	    key_sa_refcnt(sav));
3631
3632	/* SA */
3633	if (mhp->ext[SADB_EXT_SA] != NULL) {
3634		const struct sadb_sa *sa0;
3635
3636		sa0 = mhp->ext[SADB_EXT_SA];
3637		if (mhp->extlen[SADB_EXT_SA] < sizeof(*sa0)) {
3638			error = EINVAL;
3639			goto fail;
3640		}
3641
3642		sav->alg_auth = sa0->sadb_sa_auth;
3643		sav->alg_enc = sa0->sadb_sa_encrypt;
3644		sav->flags = sa0->sadb_sa_flags;
3645
3646		/* replay window */
3647		if ((sa0->sadb_sa_flags & SADB_X_EXT_OLD) == 0) {
3648			size_t len = sizeof(struct secreplay) +
3649			    sa0->sadb_sa_replay;
3650			sav->replay = kmem_zalloc(len, KM_SLEEP);
3651			sav->replay_len = len;
3652			if (sa0->sadb_sa_replay != 0)
3653				sav->replay->bitmap = (char*)(sav->replay+1);
3654			sav->replay->wsize = sa0->sadb_sa_replay;
3655		}
3656	}
3657
3658	/* Authentication keys */
3659	if (mhp->ext[SADB_EXT_KEY_AUTH] != NULL) {
3660		const struct sadb_key *key0;
3661		int len;
3662
3663		key0 = mhp->ext[SADB_EXT_KEY_AUTH];
3664		len = mhp->extlen[SADB_EXT_KEY_AUTH];
3665
3666		error = 0;
3667		if (len < sizeof(*key0)) {
3668			error = EINVAL;
3669			goto fail;
3670		}
3671		switch (mhp->msg->sadb_msg_satype) {
3672		case SADB_SATYPE_AH:
3673		case SADB_SATYPE_ESP:
3674		case SADB_X_SATYPE_TCPSIGNATURE:
3675			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3676			    sav->alg_auth != SADB_X_AALG_NULL)
3677				error = EINVAL;
3678			break;
3679		case SADB_X_SATYPE_IPCOMP:
3680		default:
3681			error = EINVAL;
3682			break;
3683		}
3684		if (error) {
3685			IPSECLOG(LOG_DEBUG, "invalid key_auth values.\n");
3686			goto fail;
3687		}
3688
3689		sav->key_auth = key_newbuf(key0, len);
3690		sav->key_auth_len = len;
3691	}
3692
3693	/* Encryption key */
3694	if (mhp->ext[SADB_EXT_KEY_ENCRYPT] != NULL) {
3695		const struct sadb_key *key0;
3696		int len;
3697
3698		key0 = mhp->ext[SADB_EXT_KEY_ENCRYPT];
3699		len = mhp->extlen[SADB_EXT_KEY_ENCRYPT];
3700
3701		error = 0;
3702		if (len < sizeof(*key0)) {
3703			error = EINVAL;
3704			goto fail;
3705		}
3706		switch (mhp->msg->sadb_msg_satype) {
3707		case SADB_SATYPE_ESP:
3708			if (len == PFKEY_ALIGN8(sizeof(struct sadb_key)) &&
3709			    sav->alg_enc != SADB_EALG_NULL) {
3710				error = EINVAL;
3711				break;
3712			}
3713			sav->key_enc = key_newbuf(key0, len);
3714			sav->key_enc_len = len;
3715			break;
3716		case SADB_X_SATYPE_IPCOMP:
3717			if (len != PFKEY_ALIGN8(sizeof(struct sadb_key)))
3718				error = EINVAL;
3719			sav->key_enc = NULL;	/*just in case*/
3720			break;
3721		case SADB_SATYPE_AH:
3722		case SADB_X_SATYPE_TCPSIGNATURE:
3723		default:
3724			error = EINVAL;
3725			break;
3726		}
3727		if (error) {
3728			IPSECLOG(LOG_DEBUG, "invalid key_enc value.\n");
3729			goto fail;
3730		}
3731	}
3732
3733	/* set iv */
3734	sav->ivlen = 0;
3735
3736	switch (mhp->msg->sadb_msg_satype) {
3737	case SADB_SATYPE_AH:
3738		error = xform_init(sav, XF_AH);
3739		break;
3740	case SADB_SATYPE_ESP:
3741		error = xform_init(sav, XF_ESP);
3742		break;
3743	case SADB_X_SATYPE_IPCOMP:
3744		error = xform_init(sav, XF_IPCOMP);
3745		break;
3746	case SADB_X_SATYPE_TCPSIGNATURE:
3747		error = xform_init(sav, XF_TCPSIGNATURE);
3748		break;
3749	default:
3750		error = EOPNOTSUPP;
3751		break;
3752	}
3753	if (error) {
3754		IPSECLOG(LOG_DEBUG, "unable to initialize SA type %u (%d)\n",
3755		    mhp->msg->sadb_msg_satype, error);
3756		goto fail;
3757	}
3758
3759	/* reset created */
3760	sav->created = time_uptime;
3761
3762	/* make lifetime for CURRENT */
3763	sav->lft_c = kmem_alloc(sizeof(struct sadb_lifetime), KM_SLEEP);
3764
3765	sav->lft_c->sadb_lifetime_len =
3766	    PFKEY_UNIT64(sizeof(struct sadb_lifetime));
3767	sav->lft_c->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
3768	sav->lft_c->sadb_lifetime_allocations = 0;
3769	sav->lft_c->sadb_lifetime_bytes = 0;
3770	sav->lft_c->sadb_lifetime_addtime = time_uptime;
3771	sav->lft_c->sadb_lifetime_usetime = 0;
3772
3773	sav->lft_c_counters_percpu = percpu_alloc(sizeof(lifetime_counters_t));
3774
3775	/* lifetimes for HARD and SOFT */
3776    {
3777	const struct sadb_lifetime *lft0;
3778
3779	lft0 = mhp->ext[SADB_EXT_LIFETIME_HARD];
3780	if (lft0 != NULL) {
3781		if (mhp->extlen[SADB_EXT_LIFETIME_HARD] < sizeof(*lft0)) {
3782			error = EINVAL;
3783			goto fail;
3784		}
3785		sav->lft_h = key_newbuf(lft0, sizeof(*lft0));
3786	}
3787
3788	lft0 = mhp->ext[SADB_EXT_LIFETIME_SOFT];
3789	if (lft0 != NULL) {
3790		if (mhp->extlen[SADB_EXT_LIFETIME_SOFT] < sizeof(*lft0)) {
3791			error = EINVAL;
3792			goto fail;
3793		}
3794		sav->lft_s = key_newbuf(lft0, sizeof(*lft0));
3795		/* to be initialize ? */
3796	}
3797    }
3798
3799	return 0;
3800
3801 fail:
3802	key_clear_xform(sav);
3803	key_freesaval(sav);
3804
3805	return error;
3806}
3807
3808/*
3809 * validation with a secasvar entry, and set SADB_SATYPE_MATURE.
3810 * OUT:	0:	valid
3811 *	other:	errno
3812 */
3813static int
3814key_init_xform(struct secasvar *sav)
3815{
3816	int error;
3817
3818	/* We shouldn't initialize sav variables while someone uses it. */
3819	KASSERTMSG(key_sa_refcnt(sav) == 0, "key_sa_refcnt(sav)=%u",
3820	    key_sa_refcnt(sav));
3821
3822	/* check SPI value */
3823	switch (sav->sah->saidx.proto) {
3824	case IPPROTO_ESP:
3825	case IPPROTO_AH:
3826		if (ntohl(sav->spi) <= 255) {
3827			IPSECLOG(LOG_DEBUG, "illegal range of SPI %u.\n",
3828			    (u_int32_t)ntohl(sav->spi));
3829			return EINVAL;
3830		}
3831		break;
3832	}
3833
3834	/* check algo */
3835	switch (sav->sah->saidx.proto) {
3836	case IPPROTO_AH:
3837	case IPPROTO_TCP:
3838		if (sav->alg_enc != SADB_EALG_NONE) {
3839			IPSECLOG(LOG_DEBUG,
3840			    "protocol %u and algorithm mismatched %u != %u.\n",
3841			    sav->sah->saidx.proto,
3842			    sav->alg_enc, SADB_EALG_NONE);
3843			return EINVAL;
3844		}
3845		break;
3846	case IPPROTO_IPCOMP:
3847		if (sav->alg_auth != SADB_AALG_NONE) {
3848			IPSECLOG(LOG_DEBUG,
3849			    "protocol %u and algorithm mismatched %d != %d.\n",
3850			    sav->sah->saidx.proto,
3851			    sav->alg_auth, SADB_AALG_NONE);
3852			return(EINVAL);
3853		}
3854		break;
3855	default:
3856		break;
3857	}
3858
3859	/* check satype */
3860	switch (sav->sah->saidx.proto) {
3861	case IPPROTO_ESP:
3862		/* check flags */
3863		if ((sav->flags & (SADB_X_EXT_OLD|SADB_X_EXT_DERIV)) ==
3864		    (SADB_X_EXT_OLD|SADB_X_EXT_DERIV)) {
3865			IPSECLOG(LOG_DEBUG,
3866			    "invalid flag (derived) given to old-esp.\n");
3867			return EINVAL;
3868		}
3869		error = xform_init(sav, XF_ESP);
3870		break;
3871	case IPPROTO_AH:
3872		/* check flags */
3873		if (sav->flags & SADB_X_EXT_DERIV) {
3874			IPSECLOG(LOG_DEBUG,
3875			    "invalid flag (derived) given to AH SA.\n");
3876			return EINVAL;
3877		}
3878		error = xform_init(sav, XF_AH);
3879		break;
3880	case IPPROTO_IPCOMP:
3881		if ((sav->flags & SADB_X_EXT_RAWCPI) == 0
3882		    && ntohl(sav->spi) >= 0x10000) {
3883			IPSECLOG(LOG_DEBUG, "invalid cpi for IPComp.\n");
3884			return(EINVAL);
3885		}
3886		error = xform_init(sav, XF_IPCOMP);
3887		break;
3888	case IPPROTO_TCP:
3889		error = xform_init(sav, XF_TCPSIGNATURE);
3890		break;
3891	default:
3892		IPSECLOG(LOG_DEBUG, "Invalid satype.\n");
3893		error = EPROTONOSUPPORT;
3894		break;
3895	}
3896
3897	return error;
3898}
3899
3900/*
3901 * subroutine for SADB_GET and SADB_DUMP. It never return NULL.
3902 */
3903static struct mbuf *
3904key_setdumpsa(struct secasvar *sav, u_int8_t type, u_int8_t satype,
3905	      u_int32_t seq, u_int32_t pid)
3906{
3907	struct mbuf *result = NULL, *tres = NULL, *m;
3908	int l = 0;
3909	int i;
3910	void *p;
3911	struct sadb_lifetime lt;
3912	int dumporder[] = {
3913		SADB_EXT_SA, SADB_X_EXT_SA2,
3914		SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
3915		SADB_EXT_LIFETIME_CURRENT, SADB_EXT_ADDRESS_SRC,
3916		SADB_EXT_ADDRESS_DST, SADB_EXT_ADDRESS_PROXY, SADB_EXT_KEY_AUTH,
3917		SADB_EXT_KEY_ENCRYPT, SADB_EXT_IDENTITY_SRC,
3918		SADB_EXT_IDENTITY_DST, SADB_EXT_SENSITIVITY,
3919		SADB_X_EXT_NAT_T_TYPE,
3920		SADB_X_EXT_NAT_T_SPORT, SADB_X_EXT_NAT_T_DPORT,
3921		SADB_X_EXT_NAT_T_OAI, SADB_X_EXT_NAT_T_OAR,
3922		SADB_X_EXT_NAT_T_FRAG,
3923
3924	};
3925
3926	m = key_setsadbmsg(type, 0, satype, seq, pid, key_sa_refcnt(sav), M_WAITOK);
3927	result = m;
3928
3929	for (i = __arraycount(dumporder) - 1; i >= 0; i--) {
3930		m = NULL;
3931		p = NULL;
3932		switch (dumporder[i]) {
3933		case SADB_EXT_SA:
3934			m = key_setsadbsa(sav);
3935			break;
3936
3937		case SADB_X_EXT_SA2:
3938			m = key_setsadbxsa2(sav->sah->saidx.mode,
3939			    sav->replay ? sav->replay->count : 0,
3940			    sav->sah->saidx.reqid);
3941			break;
3942
3943		case SADB_EXT_ADDRESS_SRC:
3944			m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC,
3945			    &sav->sah->saidx.src.sa,
3946			    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
3947			break;
3948
3949		case SADB_EXT_ADDRESS_DST:
3950			m = key_setsadbaddr(SADB_EXT_ADDRESS_DST,
3951			    &sav->sah->saidx.dst.sa,
3952			    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
3953			break;
3954
3955		case SADB_EXT_KEY_AUTH:
3956			if (!sav->key_auth)
3957				continue;
3958			l = PFKEY_UNUNIT64(sav->key_auth->sadb_key_len);
3959			p = sav->key_auth;
3960			break;
3961
3962		case SADB_EXT_KEY_ENCRYPT:
3963			if (!sav->key_enc)
3964				continue;
3965			l = PFKEY_UNUNIT64(sav->key_enc->sadb_key_len);
3966			p = sav->key_enc;
3967			break;
3968
3969		case SADB_EXT_LIFETIME_CURRENT: {
3970			lifetime_counters_t sum = {0};
3971
3972			KASSERT(sav->lft_c != NULL);
3973			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_c)->sadb_ext_len);
3974			memcpy(&lt, sav->lft_c, sizeof(struct sadb_lifetime));
3975			lt.sadb_lifetime_addtime =
3976			    time_mono_to_wall(lt.sadb_lifetime_addtime);
3977			lt.sadb_lifetime_usetime =
3978			    time_mono_to_wall(lt.sadb_lifetime_usetime);
3979			percpu_foreach_xcall(sav->lft_c_counters_percpu,
3980			    XC_HIGHPRI_IPL(IPL_SOFTNET),
3981			    key_sum_lifetime_counters, sum);
3982			lt.sadb_lifetime_allocations =
3983			    sum[LIFETIME_COUNTER_ALLOCATIONS];
3984			lt.sadb_lifetime_bytes =
3985			    sum[LIFETIME_COUNTER_BYTES];
3986			p = &lt;
3987			break;
3988		    }
3989
3990		case SADB_EXT_LIFETIME_HARD:
3991			if (!sav->lft_h)
3992				continue;
3993			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_h)->sadb_ext_len);
3994			p = sav->lft_h;
3995			break;
3996
3997		case SADB_EXT_LIFETIME_SOFT:
3998			if (!sav->lft_s)
3999				continue;
4000			l = PFKEY_UNUNIT64(((struct sadb_ext *)sav->lft_s)->sadb_ext_len);
4001			p = sav->lft_s;
4002			break;
4003
4004		case SADB_X_EXT_NAT_T_TYPE:
4005			m = key_setsadbxtype(sav->natt_type);
4006			break;
4007
4008		case SADB_X_EXT_NAT_T_DPORT:
4009			if (sav->natt_type == 0)
4010				continue;
4011			m = key_setsadbxport(
4012			    key_portfromsaddr(&sav->sah->saidx.dst),
4013			    SADB_X_EXT_NAT_T_DPORT);
4014			break;
4015
4016		case SADB_X_EXT_NAT_T_SPORT:
4017			if (sav->natt_type == 0)
4018				continue;
4019			m = key_setsadbxport(
4020			    key_portfromsaddr(&sav->sah->saidx.src),
4021			    SADB_X_EXT_NAT_T_SPORT);
4022			break;
4023
4024		case SADB_X_EXT_NAT_T_FRAG:
4025			/* don't send frag info if not set */
4026			if (sav->natt_type == 0 || sav->esp_frag == IP_MAXPACKET)
4027				continue;
4028			m = key_setsadbxfrag(sav->esp_frag);
4029			break;
4030
4031		case SADB_X_EXT_NAT_T_OAI:
4032		case SADB_X_EXT_NAT_T_OAR:
4033			continue;
4034
4035		case SADB_EXT_ADDRESS_PROXY:
4036		case SADB_EXT_IDENTITY_SRC:
4037		case SADB_EXT_IDENTITY_DST:
4038			/* XXX: should we brought from SPD ? */
4039		case SADB_EXT_SENSITIVITY:
4040		default:
4041			continue;
4042		}
4043
4044		KASSERT(!(m && p));
4045		KASSERT(m != NULL || p != NULL);
4046		if (p && tres) {
4047			M_PREPEND(tres, l, M_WAITOK);
4048			memcpy(mtod(tres, void *), p, l);
4049			continue;
4050		}
4051		if (p) {
4052			m = key_alloc_mbuf(l, M_WAITOK);
4053			m_copyback(m, 0, l, p);
4054		}
4055
4056		if (tres)
4057			m_cat(m, tres);
4058		tres = m;
4059	}
4060
4061	m_cat(result, tres);
4062	tres = NULL; /* avoid free on error below */
4063
4064	KASSERT(result->m_len >= sizeof(struct sadb_msg));
4065
4066	result->m_pkthdr.len = 0;
4067	for (m = result; m; m = m->m_next)
4068		result->m_pkthdr.len += m->m_len;
4069
4070	mtod(result, struct sadb_msg *)->sadb_msg_len =
4071	    PFKEY_UNIT64(result->m_pkthdr.len);
4072
4073	return result;
4074}
4075
4076
4077/*
4078 * set a type in sadb_x_nat_t_type
4079 */
4080static struct mbuf *
4081key_setsadbxtype(u_int16_t type)
4082{
4083	struct mbuf *m;
4084	size_t len;
4085	struct sadb_x_nat_t_type *p;
4086
4087	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_type));
4088
4089	m = key_alloc_mbuf(len, M_WAITOK);
4090	KASSERT(m->m_next == NULL);
4091
4092	p = mtod(m, struct sadb_x_nat_t_type *);
4093
4094	memset(p, 0, len);
4095	p->sadb_x_nat_t_type_len = PFKEY_UNIT64(len);
4096	p->sadb_x_nat_t_type_exttype = SADB_X_EXT_NAT_T_TYPE;
4097	p->sadb_x_nat_t_type_type = type;
4098
4099	return m;
4100}
4101/*
4102 * set a port in sadb_x_nat_t_port. port is in network order
4103 */
4104static struct mbuf *
4105key_setsadbxport(u_int16_t port, u_int16_t type)
4106{
4107	struct mbuf *m;
4108	size_t len;
4109	struct sadb_x_nat_t_port *p;
4110
4111	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_port));
4112
4113	m = key_alloc_mbuf(len, M_WAITOK);
4114	KASSERT(m->m_next == NULL);
4115
4116	p = mtod(m, struct sadb_x_nat_t_port *);
4117
4118	memset(p, 0, len);
4119	p->sadb_x_nat_t_port_len = PFKEY_UNIT64(len);
4120	p->sadb_x_nat_t_port_exttype = type;
4121	p->sadb_x_nat_t_port_port = port;
4122
4123	return m;
4124}
4125
4126/*
4127 * set fragmentation info in sadb_x_nat_t_frag
4128 */
4129static struct mbuf *
4130key_setsadbxfrag(u_int16_t flen)
4131{
4132	struct mbuf *m;
4133	size_t len;
4134	struct sadb_x_nat_t_frag *p;
4135
4136	len = PFKEY_ALIGN8(sizeof(struct sadb_x_nat_t_frag));
4137
4138	m = key_alloc_mbuf(len, M_WAITOK);
4139	KASSERT(m->m_next == NULL);
4140
4141	p = mtod(m, struct sadb_x_nat_t_frag *);
4142
4143	memset(p, 0, len);
4144	p->sadb_x_nat_t_frag_len = PFKEY_UNIT64(len);
4145	p->sadb_x_nat_t_frag_exttype = SADB_X_EXT_NAT_T_FRAG;
4146	p->sadb_x_nat_t_frag_fraglen = flen;
4147
4148	return m;
4149}
4150
4151/*
4152 * Get port from sockaddr, port is in network order
4153 */
4154u_int16_t
4155key_portfromsaddr(const union sockaddr_union *saddr)
4156{
4157	u_int16_t port;
4158
4159	switch (saddr->sa.sa_family) {
4160	case AF_INET: {
4161		port = saddr->sin.sin_port;
4162		break;
4163	}
4164#ifdef INET6
4165	case AF_INET6: {
4166		port = saddr->sin6.sin6_port;
4167		break;
4168	}
4169#endif
4170	default:
4171		printf("%s: unexpected address family\n", __func__);
4172		port = 0;
4173		break;
4174	}
4175
4176	return port;
4177}
4178
4179
4180/*
4181 * Set port is struct sockaddr. port is in network order
4182 */
4183static void
4184key_porttosaddr(union sockaddr_union *saddr, u_int16_t port)
4185{
4186	switch (saddr->sa.sa_family) {
4187	case AF_INET: {
4188		saddr->sin.sin_port = port;
4189		break;
4190	}
4191#ifdef INET6
4192	case AF_INET6: {
4193		saddr->sin6.sin6_port = port;
4194		break;
4195	}
4196#endif
4197	default:
4198		printf("%s: unexpected address family %d\n", __func__,
4199		    saddr->sa.sa_family);
4200		break;
4201	}
4202
4203	return;
4204}
4205
4206/*
4207 * Safety check sa_len
4208 */
4209static int
4210key_checksalen(const union sockaddr_union *saddr)
4211{
4212	switch (saddr->sa.sa_family) {
4213	case AF_INET:
4214		if (saddr->sa.sa_len != sizeof(struct sockaddr_in))
4215			return -1;
4216		break;
4217#ifdef INET6
4218	case AF_INET6:
4219		if (saddr->sa.sa_len != sizeof(struct sockaddr_in6))
4220			return -1;
4221		break;
4222#endif
4223	default:
4224		printf("%s: unexpected sa_family %d\n", __func__,
4225		    saddr->sa.sa_family);
4226			return -1;
4227		break;
4228	}
4229	return 0;
4230}
4231
4232
4233/*
4234 * set data into sadb_msg.
4235 */
4236static struct mbuf *
4237key_setsadbmsg(u_int8_t type,  u_int16_t tlen, u_int8_t satype,
4238	       u_int32_t seq, pid_t pid, u_int16_t reserved, int mflag)
4239{
4240	struct mbuf *m;
4241	struct sadb_msg *p;
4242	int len;
4243
4244	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) <= MCLBYTES);
4245
4246	len = PFKEY_ALIGN8(sizeof(struct sadb_msg));
4247
4248	m = key_alloc_mbuf_simple(len, mflag);
4249	if (!m)
4250		return NULL;
4251	m->m_pkthdr.len = m->m_len = len;
4252	m->m_next = NULL;
4253
4254	p = mtod(m, struct sadb_msg *);
4255
4256	memset(p, 0, len);
4257	p->sadb_msg_version = PF_KEY_V2;
4258	p->sadb_msg_type = type;
4259	p->sadb_msg_errno = 0;
4260	p->sadb_msg_satype = satype;
4261	p->sadb_msg_len = PFKEY_UNIT64(tlen);
4262	p->sadb_msg_reserved = reserved;
4263	p->sadb_msg_seq = seq;
4264	p->sadb_msg_pid = (u_int32_t)pid;
4265
4266	return m;
4267}
4268
4269/*
4270 * copy secasvar data into sadb_address.
4271 */
4272static struct mbuf *
4273key_setsadbsa(struct secasvar *sav)
4274{
4275	struct mbuf *m;
4276	struct sadb_sa *p;
4277	int len;
4278
4279	len = PFKEY_ALIGN8(sizeof(struct sadb_sa));
4280	m = key_alloc_mbuf(len, M_WAITOK);
4281	KASSERT(m->m_next == NULL);
4282
4283	p = mtod(m, struct sadb_sa *);
4284
4285	memset(p, 0, len);
4286	p->sadb_sa_len = PFKEY_UNIT64(len);
4287	p->sadb_sa_exttype = SADB_EXT_SA;
4288	p->sadb_sa_spi = sav->spi;
4289	p->sadb_sa_replay = (sav->replay != NULL ? sav->replay->wsize : 0);
4290	p->sadb_sa_state = sav->state;
4291	p->sadb_sa_auth = sav->alg_auth;
4292	p->sadb_sa_encrypt = sav->alg_enc;
4293	p->sadb_sa_flags = sav->flags;
4294
4295	return m;
4296}
4297
4298static uint8_t
4299key_sabits(const struct sockaddr *saddr)
4300{
4301	switch (saddr->sa_family) {
4302	case AF_INET:
4303		return _BITS(sizeof(struct in_addr));
4304	case AF_INET6:
4305		return _BITS(sizeof(struct in6_addr));
4306	default:
4307		return FULLMASK;
4308	}
4309}
4310
4311/*
4312 * set data into sadb_address.
4313 */
4314static struct mbuf *
4315key_setsadbaddr(u_int16_t exttype, const struct sockaddr *saddr,
4316		u_int8_t prefixlen, u_int16_t ul_proto, int mflag)
4317{
4318	struct mbuf *m;
4319	struct sadb_address *p;
4320	size_t len;
4321
4322	len = PFKEY_ALIGN8(sizeof(struct sadb_address)) +
4323	    PFKEY_ALIGN8(saddr->sa_len);
4324	m = key_alloc_mbuf(len, mflag);
4325	if (!m || m->m_next) {	/*XXX*/
4326		if (m)
4327			m_freem(m);
4328		return NULL;
4329	}
4330
4331	p = mtod(m, struct sadb_address *);
4332
4333	memset(p, 0, len);
4334	p->sadb_address_len = PFKEY_UNIT64(len);
4335	p->sadb_address_exttype = exttype;
4336	p->sadb_address_proto = ul_proto;
4337	if (prefixlen == FULLMASK) {
4338		prefixlen = key_sabits(saddr);
4339	}
4340	p->sadb_address_prefixlen = prefixlen;
4341	p->sadb_address_reserved = 0;
4342
4343	memcpy(mtod(m, char *) + PFKEY_ALIGN8(sizeof(struct sadb_address)),
4344	    saddr, saddr->sa_len);
4345
4346	return m;
4347}
4348
4349#if 0
4350/*
4351 * set data into sadb_ident.
4352 */
4353static struct mbuf *
4354key_setsadbident(u_int16_t exttype, u_int16_t idtype,
4355		 void *string, int stringlen, u_int64_t id)
4356{
4357	struct mbuf *m;
4358	struct sadb_ident *p;
4359	size_t len;
4360
4361	len = PFKEY_ALIGN8(sizeof(struct sadb_ident)) + PFKEY_ALIGN8(stringlen);
4362	m = key_alloc_mbuf(len);
4363	if (!m || m->m_next) {	/*XXX*/
4364		if (m)
4365			m_freem(m);
4366		return NULL;
4367	}
4368
4369	p = mtod(m, struct sadb_ident *);
4370
4371	memset(p, 0, len);
4372	p->sadb_ident_len = PFKEY_UNIT64(len);
4373	p->sadb_ident_exttype = exttype;
4374	p->sadb_ident_type = idtype;
4375	p->sadb_ident_reserved = 0;
4376	p->sadb_ident_id = id;
4377
4378	memcpy(mtod(m, void *) + PFKEY_ALIGN8(sizeof(struct sadb_ident)),
4379	   	   string, stringlen);
4380
4381	return m;
4382}
4383#endif
4384
4385/*
4386 * set data into sadb_x_sa2.
4387 */
4388static struct mbuf *
4389key_setsadbxsa2(u_int8_t mode, u_int32_t seq, u_int16_t reqid)
4390{
4391	struct mbuf *m;
4392	struct sadb_x_sa2 *p;
4393	size_t len;
4394
4395	len = PFKEY_ALIGN8(sizeof(struct sadb_x_sa2));
4396	m = key_alloc_mbuf(len, M_WAITOK);
4397	KASSERT(m->m_next == NULL);
4398
4399	p = mtod(m, struct sadb_x_sa2 *);
4400
4401	memset(p, 0, len);
4402	p->sadb_x_sa2_len = PFKEY_UNIT64(len);
4403	p->sadb_x_sa2_exttype = SADB_X_EXT_SA2;
4404	p->sadb_x_sa2_mode = mode;
4405	p->sadb_x_sa2_reserved1 = 0;
4406	p->sadb_x_sa2_reserved2 = 0;
4407	p->sadb_x_sa2_sequence = seq;
4408	p->sadb_x_sa2_reqid = reqid;
4409
4410	return m;
4411}
4412
4413/*
4414 * set data into sadb_x_policy
4415 */
4416static struct mbuf *
4417key_setsadbxpolicy(const u_int16_t type, const u_int8_t dir, const u_int32_t id,
4418    int mflag)
4419{
4420	struct mbuf *m;
4421	struct sadb_x_policy *p;
4422	size_t len;
4423
4424	len = PFKEY_ALIGN8(sizeof(struct sadb_x_policy));
4425	m = key_alloc_mbuf(len, mflag);
4426	if (!m || m->m_next) {	/*XXX*/
4427		if (m)
4428			m_freem(m);
4429		return NULL;
4430	}
4431
4432	p = mtod(m, struct sadb_x_policy *);
4433
4434	memset(p, 0, len);
4435	p->sadb_x_policy_len = PFKEY_UNIT64(len);
4436	p->sadb_x_policy_exttype = SADB_X_EXT_POLICY;
4437	p->sadb_x_policy_type = type;
4438	p->sadb_x_policy_dir = dir;
4439	p->sadb_x_policy_id = id;
4440
4441	return m;
4442}
4443
4444/* %%% utilities */
4445/*
4446 * copy a buffer into the new buffer allocated.
4447 */
4448static void *
4449key_newbuf(const void *src, u_int len)
4450{
4451	void *new;
4452
4453	new = kmem_alloc(len, KM_SLEEP);
4454	memcpy(new, src, len);
4455
4456	return new;
4457}
4458
4459/* compare my own address
4460 * OUT:	1: true, i.e. my address.
4461 *	0: false
4462 */
4463int
4464key_ismyaddr(const struct sockaddr *sa)
4465{
4466#ifdef INET
4467	const struct sockaddr_in *sin;
4468	const struct in_ifaddr *ia;
4469	int s;
4470#endif
4471
4472	KASSERT(sa != NULL);
4473
4474	switch (sa->sa_family) {
4475#ifdef INET
4476	case AF_INET:
4477		sin = (const struct sockaddr_in *)sa;
4478		s = pserialize_read_enter();
4479		IN_ADDRLIST_READER_FOREACH(ia) {
4480			if (sin->sin_family == ia->ia_addr.sin_family &&
4481			    sin->sin_len == ia->ia_addr.sin_len &&
4482			    sin->sin_addr.s_addr == ia->ia_addr.sin_addr.s_addr)
4483			{
4484				pserialize_read_exit(s);
4485				return 1;
4486			}
4487		}
4488		pserialize_read_exit(s);
4489		break;
4490#endif
4491#ifdef INET6
4492	case AF_INET6:
4493		return key_ismyaddr6((const struct sockaddr_in6 *)sa);
4494#endif
4495	}
4496
4497	return 0;
4498}
4499
4500#ifdef INET6
4501/*
4502 * compare my own address for IPv6.
4503 * 1: ours
4504 * 0: other
4505 * NOTE: derived ip6_input() in KAME. This is necessary to modify more.
4506 */
4507#include <netinet6/in6_var.h>
4508
4509static int
4510key_ismyaddr6(const struct sockaddr_in6 *sin6)
4511{
4512	struct in6_ifaddr *ia;
4513	int s;
4514	struct psref psref;
4515	int bound;
4516	int ours = 1;
4517
4518	bound = curlwp_bind();
4519	s = pserialize_read_enter();
4520	IN6_ADDRLIST_READER_FOREACH(ia) {
4521		if (key_sockaddr_match((const struct sockaddr *)&sin6,
4522		    (const struct sockaddr *)&ia->ia_addr, 0)) {
4523			pserialize_read_exit(s);
4524			goto ours;
4525		}
4526
4527		if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4528			bool ingroup;
4529
4530			ia6_acquire(ia, &psref);
4531			pserialize_read_exit(s);
4532
4533			/*
4534			 * XXX Multicast
4535			 * XXX why do we care about multlicast here while we don't care
4536			 * about IPv4 multicast??
4537			 * XXX scope
4538			 */
4539			ingroup = in6_multi_group(&sin6->sin6_addr, ia->ia_ifp);
4540			if (ingroup) {
4541				ia6_release(ia, &psref);
4542				goto ours;
4543			}
4544
4545			s = pserialize_read_enter();
4546			ia6_release(ia, &psref);
4547		}
4548
4549	}
4550	pserialize_read_exit(s);
4551
4552	/* loopback, just for safety */
4553	if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))
4554		goto ours;
4555
4556	ours = 0;
4557ours:
4558	curlwp_bindx(bound);
4559
4560	return ours;
4561}
4562#endif /*INET6*/
4563
4564/*
4565 * compare two secasindex structure.
4566 * flag can specify to compare 2 saidxes.
4567 * compare two secasindex structure without both mode and reqid.
4568 * don't compare port.
4569 * IN:
4570 *      saidx0: source, it can be in SAD.
4571 *      saidx1: object.
4572 * OUT:
4573 *      1 : equal
4574 *      0 : not equal
4575 */
4576static int
4577key_saidx_match(
4578	const struct secasindex *saidx0,
4579	const struct secasindex *saidx1,
4580	int flag)
4581{
4582	int chkport;
4583	const struct sockaddr *sa0src, *sa0dst, *sa1src, *sa1dst;
4584
4585	KASSERT(saidx0 != NULL);
4586	KASSERT(saidx1 != NULL);
4587
4588	/* sanity */
4589	if (saidx0->proto != saidx1->proto)
4590		return 0;
4591
4592	if (flag == CMP_EXACTLY) {
4593		if (saidx0->mode != saidx1->mode)
4594			return 0;
4595		if (saidx0->reqid != saidx1->reqid)
4596			return 0;
4597		if (memcmp(&saidx0->src, &saidx1->src, saidx0->src.sa.sa_len) != 0 ||
4598		    memcmp(&saidx0->dst, &saidx1->dst, saidx0->dst.sa.sa_len) != 0)
4599			return 0;
4600	} else {
4601
4602		/* CMP_MODE_REQID, CMP_REQID, CMP_HEAD */
4603		if (flag == CMP_MODE_REQID ||flag == CMP_REQID) {
4604			/*
4605			 * If reqid of SPD is non-zero, unique SA is required.
4606			 * The result must be of same reqid in this case.
4607			 */
4608			if (saidx1->reqid != 0 && saidx0->reqid != saidx1->reqid)
4609				return 0;
4610		}
4611
4612		if (flag == CMP_MODE_REQID) {
4613			if (saidx0->mode != IPSEC_MODE_ANY &&
4614			    saidx0->mode != saidx1->mode)
4615				return 0;
4616		}
4617
4618
4619		sa0src = &saidx0->src.sa;
4620		sa0dst = &saidx0->dst.sa;
4621		sa1src = &saidx1->src.sa;
4622		sa1dst = &saidx1->dst.sa;
4623		/*
4624		 * If NAT-T is enabled, check ports for tunnel mode.
4625		 * For ipsecif(4), check ports for transport mode, too.
4626		 * Don't check ports if they are set to zero
4627		 * in the SPD: This means we have a non-generated
4628		 * SPD which can't know UDP ports.
4629		 */
4630		if (saidx1->mode == IPSEC_MODE_TUNNEL ||
4631		    saidx1->mode == IPSEC_MODE_TRANSPORT)
4632			chkport = PORT_LOOSE;
4633		else
4634			chkport = PORT_NONE;
4635
4636		if (!key_sockaddr_match(sa0src, sa1src, chkport)) {
4637			return 0;
4638		}
4639		if (!key_sockaddr_match(sa0dst, sa1dst, chkport)) {
4640			return 0;
4641		}
4642	}
4643
4644	return 1;
4645}
4646
4647/*
4648 * compare two secindex structure exactly.
4649 * IN:
4650 *	spidx0: source, it is often in SPD.
4651 *	spidx1: object, it is often from PFKEY message.
4652 * OUT:
4653 *	1 : equal
4654 *	0 : not equal
4655 */
4656static int
4657key_spidx_match_exactly(
4658	const struct secpolicyindex *spidx0,
4659	const struct secpolicyindex *spidx1)
4660{
4661
4662	KASSERT(spidx0 != NULL);
4663	KASSERT(spidx1 != NULL);
4664
4665	/* sanity */
4666	if (spidx0->prefs != spidx1->prefs ||
4667	    spidx0->prefd != spidx1->prefd ||
4668	    spidx0->ul_proto != spidx1->ul_proto)
4669		return 0;
4670
4671	return key_sockaddr_match(&spidx0->src.sa, &spidx1->src.sa, PORT_STRICT) &&
4672	       key_sockaddr_match(&spidx0->dst.sa, &spidx1->dst.sa, PORT_STRICT);
4673}
4674
4675/*
4676 * compare two secindex structure with mask.
4677 * IN:
4678 *	spidx0: source, it is often in SPD.
4679 *	spidx1: object, it is often from IP header.
4680 * OUT:
4681 *	1 : equal
4682 *	0 : not equal
4683 */
4684static int
4685key_spidx_match_withmask(
4686	const struct secpolicyindex *spidx0,
4687	const struct secpolicyindex *spidx1)
4688{
4689
4690	KASSERT(spidx0 != NULL);
4691	KASSERT(spidx1 != NULL);
4692
4693	if (spidx0->src.sa.sa_family != spidx1->src.sa.sa_family ||
4694	    spidx0->dst.sa.sa_family != spidx1->dst.sa.sa_family ||
4695	    spidx0->src.sa.sa_len != spidx1->src.sa.sa_len ||
4696	    spidx0->dst.sa.sa_len != spidx1->dst.sa.sa_len) {
4697		KEYDEBUG_PRINTF(KEYDEBUG_MATCH, ".sa wrong\n");
4698		return 0;
4699	}
4700
4701	/* if spidx.ul_proto == IPSEC_ULPROTO_ANY, ignore. */
4702	if (spidx0->ul_proto != (u_int16_t)IPSEC_ULPROTO_ANY &&
4703	    spidx0->ul_proto != spidx1->ul_proto) {
4704		KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "proto wrong\n");
4705		return 0;
4706	}
4707
4708	switch (spidx0->src.sa.sa_family) {
4709	case AF_INET:
4710		if (spidx0->src.sin.sin_port != IPSEC_PORT_ANY &&
4711		    spidx0->src.sin.sin_port != spidx1->src.sin.sin_port) {
4712			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v4 src port wrong\n");
4713			return 0;
4714		}
4715		if (!key_bb_match_withmask(&spidx0->src.sin.sin_addr,
4716					   &spidx1->src.sin.sin_addr, spidx0->prefs)) {
4717			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v4 src addr wrong\n");
4718			return 0;
4719		}
4720		break;
4721	case AF_INET6:
4722		if (spidx0->src.sin6.sin6_port != IPSEC_PORT_ANY &&
4723		    spidx0->src.sin6.sin6_port != spidx1->src.sin6.sin6_port) {
4724			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v6 src port wrong\n");
4725			return 0;
4726		}
4727		/*
4728		 * scope_id check. if sin6_scope_id is 0, we regard it
4729		 * as a wildcard scope, which matches any scope zone ID.
4730		 */
4731		if (spidx0->src.sin6.sin6_scope_id &&
4732		    spidx1->src.sin6.sin6_scope_id &&
4733		    spidx0->src.sin6.sin6_scope_id != spidx1->src.sin6.sin6_scope_id) {
4734			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v6 src scope wrong\n");
4735			return 0;
4736		}
4737		if (!key_bb_match_withmask(&spidx0->src.sin6.sin6_addr,
4738		    &spidx1->src.sin6.sin6_addr, spidx0->prefs)) {
4739			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v6 src addr wrong\n");
4740			return 0;
4741		}
4742		break;
4743	default:
4744		/* XXX */
4745		if (memcmp(&spidx0->src, &spidx1->src, spidx0->src.sa.sa_len) != 0) {
4746			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "src memcmp wrong\n");
4747			return 0;
4748		}
4749		break;
4750	}
4751
4752	switch (spidx0->dst.sa.sa_family) {
4753	case AF_INET:
4754		if (spidx0->dst.sin.sin_port != IPSEC_PORT_ANY &&
4755		    spidx0->dst.sin.sin_port != spidx1->dst.sin.sin_port) {
4756			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v4 dst port wrong\n");
4757			return 0;
4758		}
4759		if (!key_bb_match_withmask(&spidx0->dst.sin.sin_addr,
4760		    &spidx1->dst.sin.sin_addr, spidx0->prefd)) {
4761			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v4 dst addr wrong\n");
4762			return 0;
4763		}
4764		break;
4765	case AF_INET6:
4766		if (spidx0->dst.sin6.sin6_port != IPSEC_PORT_ANY &&
4767		    spidx0->dst.sin6.sin6_port != spidx1->dst.sin6.sin6_port) {
4768			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v6 dst port wrong\n");
4769			return 0;
4770		}
4771		/*
4772		 * scope_id check. if sin6_scope_id is 0, we regard it
4773		 * as a wildcard scope, which matches any scope zone ID.
4774		 */
4775		if (spidx0->src.sin6.sin6_scope_id &&
4776		    spidx1->src.sin6.sin6_scope_id &&
4777		    spidx0->dst.sin6.sin6_scope_id != spidx1->dst.sin6.sin6_scope_id) {
4778			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "DP v6 dst scope wrong\n");
4779			return 0;
4780		}
4781		if (!key_bb_match_withmask(&spidx0->dst.sin6.sin6_addr,
4782		    &spidx1->dst.sin6.sin6_addr, spidx0->prefd)) {
4783			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "v6 dst addr wrong\n");
4784			return 0;
4785		}
4786		break;
4787	default:
4788		/* XXX */
4789		if (memcmp(&spidx0->dst, &spidx1->dst, spidx0->dst.sa.sa_len) != 0) {
4790			KEYDEBUG_PRINTF(KEYDEBUG_MATCH, "dst memcmp wrong\n");
4791			return 0;
4792		}
4793		break;
4794	}
4795
4796	/* XXX Do we check other field ?  e.g. flowinfo */
4797
4798	return 1;
4799}
4800
4801/* returns 0 on match */
4802static int
4803key_portcomp(in_port_t port1, in_port_t port2, int howport)
4804{
4805	switch (howport) {
4806	case PORT_NONE:
4807		return 0;
4808	case PORT_LOOSE:
4809		if (port1 == 0 || port2 == 0)
4810			return 0;
4811		/*FALLTHROUGH*/
4812	case PORT_STRICT:
4813		if (port1 != port2) {
4814			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4815			    "port fail %d != %d\n", ntohs(port1), ntohs(port2));
4816			return 1;
4817		}
4818		return 0;
4819	default:
4820		KASSERT(0);
4821		return 1;
4822	}
4823}
4824
4825/* returns 1 on match */
4826static int
4827key_sockaddr_match(
4828	const struct sockaddr *sa1,
4829	const struct sockaddr *sa2,
4830	int howport)
4831{
4832	const struct sockaddr_in *sin1, *sin2;
4833	const struct sockaddr_in6 *sin61, *sin62;
4834	char s1[IPSEC_ADDRSTRLEN], s2[IPSEC_ADDRSTRLEN];
4835
4836	if (sa1->sa_family != sa2->sa_family || sa1->sa_len != sa2->sa_len) {
4837		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4838		    "fam/len fail %d != %d || %d != %d\n",
4839			sa1->sa_family, sa2->sa_family, sa1->sa_len,
4840			sa2->sa_len);
4841		return 0;
4842	}
4843
4844	switch (sa1->sa_family) {
4845	case AF_INET:
4846		if (sa1->sa_len != sizeof(struct sockaddr_in)) {
4847			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4848			    "len fail %d != %zu\n",
4849			    sa1->sa_len, sizeof(struct sockaddr_in));
4850			return 0;
4851		}
4852		sin1 = (const struct sockaddr_in *)sa1;
4853		sin2 = (const struct sockaddr_in *)sa2;
4854		if (sin1->sin_addr.s_addr != sin2->sin_addr.s_addr) {
4855			KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4856			    "addr fail %s != %s\n",
4857			    (in_print(s1, sizeof(s1), &sin1->sin_addr), s1),
4858			    (in_print(s2, sizeof(s2), &sin2->sin_addr), s2));
4859			return 0;
4860		}
4861		if (key_portcomp(sin1->sin_port, sin2->sin_port, howport)) {
4862			return 0;
4863		}
4864		KEYDEBUG_PRINTF(KEYDEBUG_MATCH,
4865		    "addr success %s[%d] == %s[%d]\n",
4866		    (in_print(s1, sizeof(s1), &sin1->sin_addr), s1),
4867		    ntohs(sin1->sin_port),
4868		    (in_print(s2, sizeof(s2), &sin2->sin_addr), s2),
4869		    ntohs(sin2->sin_port));
4870		break;
4871	case AF_INET6:
4872		sin61 = (const struct sockaddr_in6 *)sa1;
4873		sin62 = (const struct sockaddr_in6 *)sa2;
4874		if (sa1->sa_len != sizeof(struct sockaddr_in6))
4875			return 0;	/*EINVAL*/
4876
4877		if (sin61->sin6_scope_id != sin62->sin6_scope_id) {
4878			return 0;
4879		}
4880		if (!IN6_ARE_ADDR_EQUAL(&sin61->sin6_addr, &sin62->sin6_addr)) {
4881			return 0;
4882		}
4883		if (key_portcomp(sin61->sin6_port, sin62->sin6_port, howport)) {
4884			return 0;
4885		}
4886		break;
4887	default:
4888		if (memcmp(sa1, sa2, sa1->sa_len) != 0)
4889			return 0;
4890		break;
4891	}
4892
4893	return 1;
4894}
4895
4896/*
4897 * compare two buffers with mask.
4898 * IN:
4899 *	addr1: source
4900 *	addr2: object
4901 *	bits:  Number of bits to compare
4902 * OUT:
4903 *	1 : equal
4904 *	0 : not equal
4905 */
4906static int
4907key_bb_match_withmask(const void *a1, const void *a2, u_int bits)
4908{
4909	const unsigned char *p1 = a1;
4910	const unsigned char *p2 = a2;
4911
4912	/* XXX: This could be considerably faster if we compare a word
4913	 * at a time, but it is complicated on LSB Endian machines */
4914
4915	/* Handle null pointers */
4916	if (p1 == NULL || p2 == NULL)
4917		return (p1 == p2);
4918
4919	while (bits >= 8) {
4920		if (*p1++ != *p2++)
4921			return 0;
4922		bits -= 8;
4923	}
4924
4925	if (bits > 0) {
4926		u_int8_t mask = ~((1<<(8-bits))-1);
4927		if ((*p1 & mask) != (*p2 & mask))
4928			return 0;
4929	}
4930	return 1;	/* Match! */
4931}
4932
4933static void
4934key_timehandler_spd(void)
4935{
4936	u_int dir;
4937	struct secpolicy *sp;
4938	volatile time_t now;
4939
4940	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
4941	    retry:
4942		mutex_enter(&key_spd.lock);
4943		/*
4944		 * To avoid for sp->created to overtake "now" because of
4945		 * waiting mutex, set time_uptime here.
4946		 */
4947		now = time_uptime;
4948		SPLIST_WRITER_FOREACH(sp, dir) {
4949			KASSERTMSG(sp->state != IPSEC_SPSTATE_DEAD,
4950			    "sp->state=%u", sp->state);
4951
4952			if (sp->lifetime == 0 && sp->validtime == 0)
4953				continue;
4954
4955			if ((sp->lifetime && now - sp->created > sp->lifetime) ||
4956			    (sp->validtime && now - sp->lastused > sp->validtime)) {
4957				key_unlink_sp(sp);
4958				mutex_exit(&key_spd.lock);
4959				key_spdexpire(sp);
4960				key_destroy_sp(sp);
4961				goto retry;
4962			}
4963		}
4964		mutex_exit(&key_spd.lock);
4965	}
4966
4967    retry_socksplist:
4968	mutex_enter(&key_spd.lock);
4969	SOCKSPLIST_WRITER_FOREACH(sp) {
4970		if (sp->state != IPSEC_SPSTATE_DEAD)
4971			continue;
4972
4973		key_unlink_sp(sp);
4974		mutex_exit(&key_spd.lock);
4975		key_destroy_sp(sp);
4976		goto retry_socksplist;
4977	}
4978	mutex_exit(&key_spd.lock);
4979}
4980
4981static void
4982key_timehandler_sad(void)
4983{
4984	struct secashead *sah;
4985	int s;
4986	volatile time_t now;
4987
4988restart:
4989	mutex_enter(&key_sad.lock);
4990	SAHLIST_WRITER_FOREACH(sah) {
4991		/* If sah has been dead and has no sav, then delete it */
4992		if (sah->state == SADB_SASTATE_DEAD &&
4993		    !key_sah_has_sav(sah)) {
4994			key_unlink_sah(sah);
4995			mutex_exit(&key_sad.lock);
4996			key_destroy_sah(sah);
4997			goto restart;
4998		}
4999	}
5000	mutex_exit(&key_sad.lock);
5001
5002	s = pserialize_read_enter();
5003	SAHLIST_READER_FOREACH(sah) {
5004		struct secasvar *sav;
5005
5006		key_sah_ref(sah);
5007		pserialize_read_exit(s);
5008
5009		/* if LARVAL entry doesn't become MATURE, delete it. */
5010		mutex_enter(&key_sad.lock);
5011	restart_sav_LARVAL:
5012		/*
5013		 * Same as key_timehandler_spd(), set time_uptime here.
5014		 */
5015		now = time_uptime;
5016		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_LARVAL) {
5017			if (now - sav->created > key_larval_lifetime) {
5018				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
5019				goto restart_sav_LARVAL;
5020			}
5021		}
5022		mutex_exit(&key_sad.lock);
5023
5024		/*
5025		 * check MATURE entry to start to send expire message
5026		 * whether or not.
5027		 */
5028	restart_sav_MATURE:
5029		mutex_enter(&key_sad.lock);
5030		/*
5031		 * ditto
5032		 */
5033		now = time_uptime;
5034		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_MATURE) {
5035			/* we don't need to check. */
5036			if (sav->lft_s == NULL)
5037				continue;
5038
5039			/* sanity check */
5040			KASSERT(sav->lft_c != NULL);
5041
5042			/* check SOFT lifetime */
5043			if (sav->lft_s->sadb_lifetime_addtime != 0 &&
5044			    now - sav->created > sav->lft_s->sadb_lifetime_addtime) {
5045				/*
5046				 * check SA to be used whether or not.
5047				 * when SA hasn't been used, delete it.
5048				 */
5049				if (sav->lft_c->sadb_lifetime_usetime == 0) {
5050					key_sa_chgstate(sav, SADB_SASTATE_DEAD);
5051					mutex_exit(&key_sad.lock);
5052				} else {
5053					key_sa_chgstate(sav, SADB_SASTATE_DYING);
5054					mutex_exit(&key_sad.lock);
5055					/*
5056					 * XXX If we keep to send expire
5057					 * message in the status of
5058					 * DYING. Do remove below code.
5059					 */
5060					key_expire(sav);
5061				}
5062				goto restart_sav_MATURE;
5063			}
5064			/* check SOFT lifetime by bytes */
5065			/*
5066			 * XXX I don't know the way to delete this SA
5067			 * when new SA is installed.  Caution when it's
5068			 * installed too big lifetime by time.
5069			 */
5070			else {
5071				uint64_t lft_c_bytes = 0;
5072				lifetime_counters_t sum = {0};
5073
5074				percpu_foreach_xcall(sav->lft_c_counters_percpu,
5075				    XC_HIGHPRI_IPL(IPL_SOFTNET),
5076				    key_sum_lifetime_counters, sum);
5077				lft_c_bytes = sum[LIFETIME_COUNTER_BYTES];
5078
5079				if (sav->lft_s->sadb_lifetime_bytes == 0 ||
5080				    sav->lft_s->sadb_lifetime_bytes >= lft_c_bytes)
5081					continue;
5082
5083				key_sa_chgstate(sav, SADB_SASTATE_DYING);
5084				mutex_exit(&key_sad.lock);
5085				/*
5086				 * XXX If we keep to send expire
5087				 * message in the status of
5088				 * DYING. Do remove below code.
5089				 */
5090				key_expire(sav);
5091				goto restart_sav_MATURE;
5092			}
5093		}
5094		mutex_exit(&key_sad.lock);
5095
5096		/* check DYING entry to change status to DEAD. */
5097		mutex_enter(&key_sad.lock);
5098	restart_sav_DYING:
5099		/*
5100		 * ditto
5101		 */
5102		now = time_uptime;
5103		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_DYING) {
5104			/* we don't need to check. */
5105			if (sav->lft_h == NULL)
5106				continue;
5107
5108			/* sanity check */
5109			KASSERT(sav->lft_c != NULL);
5110
5111			if (sav->lft_h->sadb_lifetime_addtime != 0 &&
5112			    now - sav->created > sav->lft_h->sadb_lifetime_addtime) {
5113				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
5114				goto restart_sav_DYING;
5115			}
5116#if 0	/* XXX Should we keep to send expire message until HARD lifetime ? */
5117			else if (sav->lft_s != NULL
5118			      && sav->lft_s->sadb_lifetime_addtime != 0
5119			      && now - sav->created > sav->lft_s->sadb_lifetime_addtime) {
5120				/*
5121				 * XXX: should be checked to be
5122				 * installed the valid SA.
5123				 */
5124
5125				/*
5126				 * If there is no SA then sending
5127				 * expire message.
5128				 */
5129				key_expire(sav);
5130			}
5131#endif
5132			/* check HARD lifetime by bytes */
5133			else {
5134				uint64_t lft_c_bytes = 0;
5135				lifetime_counters_t sum = {0};
5136
5137				percpu_foreach_xcall(sav->lft_c_counters_percpu,
5138				    XC_HIGHPRI_IPL(IPL_SOFTNET),
5139				    key_sum_lifetime_counters, sum);
5140				lft_c_bytes = sum[LIFETIME_COUNTER_BYTES];
5141
5142				if (sav->lft_h->sadb_lifetime_bytes == 0 ||
5143				    sav->lft_h->sadb_lifetime_bytes >= lft_c_bytes)
5144					continue;
5145
5146				key_sa_chgstate(sav, SADB_SASTATE_DEAD);
5147				goto restart_sav_DYING;
5148			}
5149		}
5150		mutex_exit(&key_sad.lock);
5151
5152		/* delete entry in DEAD */
5153	restart_sav_DEAD:
5154		mutex_enter(&key_sad.lock);
5155		SAVLIST_WRITER_FOREACH(sav, sah, SADB_SASTATE_DEAD) {
5156			key_unlink_sav(sav);
5157			mutex_exit(&key_sad.lock);
5158			key_destroy_sav(sav);
5159			goto restart_sav_DEAD;
5160		}
5161		mutex_exit(&key_sad.lock);
5162
5163		s = pserialize_read_enter();
5164		key_sah_unref(sah);
5165	}
5166	pserialize_read_exit(s);
5167}
5168
5169static void
5170key_timehandler_acq(void)
5171{
5172#ifndef IPSEC_NONBLOCK_ACQUIRE
5173	struct secacq *acq, *nextacq;
5174	volatile time_t now;
5175
5176    restart:
5177	mutex_enter(&key_misc.lock);
5178	/*
5179	 * Same as key_timehandler_spd(), set time_uptime here.
5180	 */
5181	now = time_uptime;
5182	LIST_FOREACH_SAFE(acq, &key_misc.acqlist, chain, nextacq) {
5183		if (now - acq->created > key_blockacq_lifetime) {
5184			LIST_REMOVE(acq, chain);
5185			mutex_exit(&key_misc.lock);
5186			kmem_free(acq, sizeof(*acq));
5187			goto restart;
5188		}
5189	}
5190	mutex_exit(&key_misc.lock);
5191#endif
5192}
5193
5194static void
5195key_timehandler_spacq(void)
5196{
5197#ifdef notyet
5198	struct secspacq *acq, *nextacq;
5199	time_t now = time_uptime;
5200
5201	LIST_FOREACH_SAFE(acq, &key_misc.spacqlist, chain, nextacq) {
5202		if (now - acq->created > key_blockacq_lifetime) {
5203			KASSERT(__LIST_CHAINED(acq));
5204			LIST_REMOVE(acq, chain);
5205			kmem_free(acq, sizeof(*acq));
5206		}
5207	}
5208#endif
5209}
5210
5211static unsigned int key_timehandler_work_enqueued = 0;
5212
5213/*
5214 * time handler.
5215 * scanning SPD and SAD to check status for each entries,
5216 * and do to remove or to expire.
5217 */
5218static void
5219key_timehandler_work(struct work *wk, void *arg)
5220{
5221
5222	/* We can allow enqueuing another work at this point */
5223	atomic_swap_uint(&key_timehandler_work_enqueued, 0);
5224
5225	key_timehandler_spd();
5226	key_timehandler_sad();
5227	key_timehandler_acq();
5228	key_timehandler_spacq();
5229
5230	key_acquire_sendup_pending_mbuf();
5231
5232	/* do exchange to tick time !! */
5233	callout_reset(&key_timehandler_ch, hz, key_timehandler, NULL);
5234
5235	return;
5236}
5237
5238static void
5239key_timehandler(void *arg)
5240{
5241
5242	/* Avoid enqueuing another work when one is already enqueued */
5243	if (atomic_swap_uint(&key_timehandler_work_enqueued, 1) == 1)
5244		return;
5245
5246	workqueue_enqueue(key_timehandler_wq, &key_timehandler_wk, NULL);
5247}
5248
5249u_long
5250key_random(void)
5251{
5252	u_long value;
5253
5254	key_randomfill(&value, sizeof(value));
5255	return value;
5256}
5257
5258void
5259key_randomfill(void *p, size_t l)
5260{
5261
5262	cprng_fast(p, l);
5263}
5264
5265/*
5266 * map SADB_SATYPE_* to IPPROTO_*.
5267 * if satype == SADB_SATYPE then satype is mapped to ~0.
5268 * OUT:
5269 *	0: invalid satype.
5270 */
5271static u_int16_t
5272key_satype2proto(u_int8_t satype)
5273{
5274	switch (satype) {
5275	case SADB_SATYPE_UNSPEC:
5276		return IPSEC_PROTO_ANY;
5277	case SADB_SATYPE_AH:
5278		return IPPROTO_AH;
5279	case SADB_SATYPE_ESP:
5280		return IPPROTO_ESP;
5281	case SADB_X_SATYPE_IPCOMP:
5282		return IPPROTO_IPCOMP;
5283	case SADB_X_SATYPE_TCPSIGNATURE:
5284		return IPPROTO_TCP;
5285	default:
5286		return 0;
5287	}
5288	/* NOTREACHED */
5289}
5290
5291/*
5292 * map IPPROTO_* to SADB_SATYPE_*
5293 * OUT:
5294 *	0: invalid protocol type.
5295 */
5296static u_int8_t
5297key_proto2satype(u_int16_t proto)
5298{
5299	switch (proto) {
5300	case IPPROTO_AH:
5301		return SADB_SATYPE_AH;
5302	case IPPROTO_ESP:
5303		return SADB_SATYPE_ESP;
5304	case IPPROTO_IPCOMP:
5305		return SADB_X_SATYPE_IPCOMP;
5306	case IPPROTO_TCP:
5307		return SADB_X_SATYPE_TCPSIGNATURE;
5308	default:
5309		return 0;
5310	}
5311	/* NOTREACHED */
5312}
5313
5314static int
5315key_setsecasidx(int proto, int mode, int reqid,
5316    const struct sockaddr *src, const struct sockaddr *dst,
5317    struct secasindex * saidx)
5318{
5319	const union sockaddr_union *src_u = (const union sockaddr_union *)src;
5320	const union sockaddr_union *dst_u = (const union sockaddr_union *)dst;
5321
5322	/* sa len safety check */
5323	if (key_checksalen(src_u) != 0)
5324		return -1;
5325	if (key_checksalen(dst_u) != 0)
5326		return -1;
5327
5328	memset(saidx, 0, sizeof(*saidx));
5329	saidx->proto = proto;
5330	saidx->mode = mode;
5331	saidx->reqid = reqid;
5332	memcpy(&saidx->src, src_u, src_u->sa.sa_len);
5333	memcpy(&saidx->dst, dst_u, dst_u->sa.sa_len);
5334
5335	key_porttosaddr(&((saidx)->src), 0);
5336	key_porttosaddr(&((saidx)->dst), 0);
5337	return 0;
5338}
5339
5340static void
5341key_init_spidx_bymsghdr(struct secpolicyindex *spidx,
5342    const struct sadb_msghdr *mhp)
5343{
5344	const struct sadb_address *src0, *dst0;
5345	const struct sockaddr *src, *dst;
5346	const struct sadb_x_policy *xpl0;
5347
5348	src0 = mhp->ext[SADB_EXT_ADDRESS_SRC];
5349	dst0 = mhp->ext[SADB_EXT_ADDRESS_DST];
5350	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5351	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5352	xpl0 = mhp->ext[SADB_X_EXT_POLICY];
5353
5354	memset(spidx, 0, sizeof(*spidx));
5355	spidx->dir = xpl0->sadb_x_policy_dir;
5356	spidx->prefs = src0->sadb_address_prefixlen;
5357	spidx->prefd = dst0->sadb_address_prefixlen;
5358	spidx->ul_proto = src0->sadb_address_proto;
5359	/* XXX boundary check against sa_len */
5360	memcpy(&spidx->src, src, src->sa_len);
5361	memcpy(&spidx->dst, dst, dst->sa_len);
5362}
5363
5364/* %%% PF_KEY */
5365/*
5366 * SADB_GETSPI processing is to receive
5367 *	<base, (SA2), src address, dst address, (SPI range)>
5368 * from the IKMPd, to assign a unique spi value, to hang on the INBOUND
5369 * tree with the status of LARVAL, and send
5370 *	<base, SA(*), address(SD)>
5371 * to the IKMPd.
5372 *
5373 * IN:	mhp: pointer to the pointer to each header.
5374 * OUT:	NULL if fail.
5375 *	other if success, return pointer to the message to send.
5376 */
5377static int
5378key_api_getspi(struct socket *so, struct mbuf *m,
5379	   const struct sadb_msghdr *mhp)
5380{
5381	const struct sockaddr *src, *dst;
5382	struct secasindex saidx;
5383	struct secashead *sah;
5384	struct secasvar *newsav;
5385	u_int8_t proto;
5386	u_int32_t spi;
5387	u_int8_t mode;
5388	u_int16_t reqid;
5389	int error;
5390
5391	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
5392	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
5393		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5394		return key_senderror(so, m, EINVAL);
5395	}
5396	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
5397	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
5398		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5399		return key_senderror(so, m, EINVAL);
5400	}
5401	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
5402		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
5403		mode = sa2->sadb_x_sa2_mode;
5404		reqid = sa2->sadb_x_sa2_reqid;
5405	} else {
5406		mode = IPSEC_MODE_ANY;
5407		reqid = 0;
5408	}
5409
5410	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5411	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5412
5413	/* map satype to proto */
5414	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
5415	if (proto == 0) {
5416		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
5417		return key_senderror(so, m, EINVAL);
5418	}
5419
5420
5421	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
5422	if (error != 0)
5423		return key_senderror(so, m, EINVAL);
5424
5425	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
5426	if (error != 0)
5427		return key_senderror(so, m, EINVAL);
5428
5429	/* SPI allocation */
5430	spi = key_do_getnewspi(mhp->ext[SADB_EXT_SPIRANGE], &saidx);
5431	if (spi == 0)
5432		return key_senderror(so, m, EINVAL);
5433
5434	/* get a SA index */
5435	sah = key_getsah_ref(&saidx, CMP_REQID);
5436	if (sah == NULL) {
5437		/* create a new SA index */
5438		sah = key_newsah(&saidx);
5439		if (sah == NULL) {
5440			IPSECLOG(LOG_DEBUG, "No more memory.\n");
5441			return key_senderror(so, m, ENOBUFS);
5442		}
5443	}
5444
5445	/* get a new SA */
5446	/* XXX rewrite */
5447	newsav = KEY_NEWSAV(m, mhp, &error, proto);
5448	if (newsav == NULL) {
5449		key_sah_unref(sah);
5450		/* XXX don't free new SA index allocated in above. */
5451		return key_senderror(so, m, error);
5452	}
5453
5454	/* set spi */
5455	newsav->spi = htonl(spi);
5456
5457	/* Add to sah#savlist */
5458	key_init_sav(newsav);
5459	newsav->sah = sah;
5460	newsav->state = SADB_SASTATE_LARVAL;
5461	mutex_enter(&key_sad.lock);
5462	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_LARVAL, newsav);
5463	mutex_exit(&key_sad.lock);
5464	key_validate_savlist(sah, SADB_SASTATE_LARVAL);
5465
5466	key_sah_unref(sah);
5467
5468#ifndef IPSEC_NONBLOCK_ACQUIRE
5469	/* delete the entry in key_misc.acqlist */
5470	if (mhp->msg->sadb_msg_seq != 0) {
5471		struct secacq *acq;
5472		mutex_enter(&key_misc.lock);
5473		acq = key_getacqbyseq(mhp->msg->sadb_msg_seq);
5474		if (acq != NULL) {
5475			/* reset counter in order to deletion by timehandler. */
5476			acq->created = time_uptime;
5477			acq->count = 0;
5478		}
5479		mutex_exit(&key_misc.lock);
5480	}
5481#endif
5482
5483    {
5484	struct mbuf *n, *nn;
5485	struct sadb_sa *m_sa;
5486	int off, len;
5487
5488	CTASSERT(PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
5489	    PFKEY_ALIGN8(sizeof(struct sadb_sa)) <= MCLBYTES);
5490
5491	/* create new sadb_msg to reply. */
5492	len = PFKEY_ALIGN8(sizeof(struct sadb_msg)) +
5493	    PFKEY_ALIGN8(sizeof(struct sadb_sa));
5494
5495	n = key_alloc_mbuf_simple(len, M_WAITOK);
5496	n->m_len = len;
5497	n->m_next = NULL;
5498	off = 0;
5499
5500	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
5501	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
5502
5503	m_sa = (struct sadb_sa *)(mtod(n, char *) + off);
5504	m_sa->sadb_sa_len = PFKEY_UNIT64(sizeof(struct sadb_sa));
5505	m_sa->sadb_sa_exttype = SADB_EXT_SA;
5506	m_sa->sadb_sa_spi = htonl(spi);
5507	off += PFKEY_ALIGN8(sizeof(struct sadb_sa));
5508
5509	KASSERTMSG(off == len, "length inconsistency");
5510
5511	n->m_next = key_gather_mbuf(m, mhp, 0, 2, SADB_EXT_ADDRESS_SRC,
5512	    SADB_EXT_ADDRESS_DST);
5513
5514	KASSERT(n->m_len >= sizeof(struct sadb_msg));
5515
5516	n->m_pkthdr.len = 0;
5517	for (nn = n; nn; nn = nn->m_next)
5518		n->m_pkthdr.len += nn->m_len;
5519
5520	key_fill_replymsg(n, newsav->seq);
5521	m_freem(m);
5522	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
5523    }
5524}
5525
5526/*
5527 * allocating new SPI
5528 * called by key_api_getspi().
5529 * OUT:
5530 *	0:	failure.
5531 *	others: success.
5532 */
5533static u_int32_t
5534key_do_getnewspi(const struct sadb_spirange *spirange,
5535		 const struct secasindex *saidx)
5536{
5537	u_int32_t newspi;
5538	u_int32_t spmin, spmax;
5539	int count = key_spi_trycnt;
5540
5541	/* set spi range to allocate */
5542	if (spirange != NULL) {
5543		spmin = spirange->sadb_spirange_min;
5544		spmax = spirange->sadb_spirange_max;
5545	} else {
5546		spmin = key_spi_minval;
5547		spmax = key_spi_maxval;
5548	}
5549	/* IPCOMP needs 2-byte SPI */
5550	if (saidx->proto == IPPROTO_IPCOMP) {
5551		u_int32_t t;
5552		if (spmin >= 0x10000)
5553			spmin = 0xffff;
5554		if (spmax >= 0x10000)
5555			spmax = 0xffff;
5556		if (spmin > spmax) {
5557			t = spmin; spmin = spmax; spmax = t;
5558		}
5559	}
5560
5561	if (spmin == spmax) {
5562		if (key_checkspidup(saidx, htonl(spmin))) {
5563			IPSECLOG(LOG_DEBUG, "SPI %u exists already.\n", spmin);
5564			return 0;
5565		}
5566
5567		count--; /* taking one cost. */
5568		newspi = spmin;
5569
5570	} else {
5571
5572		/* init SPI */
5573		newspi = 0;
5574
5575		/* when requesting to allocate spi ranged */
5576		while (count--) {
5577			/* generate pseudo-random SPI value ranged. */
5578			newspi = spmin + (key_random() % (spmax - spmin + 1));
5579
5580			if (!key_checkspidup(saidx, htonl(newspi)))
5581				break;
5582		}
5583
5584		if (count == 0 || newspi == 0) {
5585			IPSECLOG(LOG_DEBUG, "to allocate spi is failed.\n");
5586			return 0;
5587		}
5588	}
5589
5590	/* statistics */
5591	keystat.getspi_count =
5592	    (keystat.getspi_count + key_spi_trycnt - count) / 2;
5593
5594	return newspi;
5595}
5596
5597static int
5598key_handle_natt_info(struct secasvar *sav,
5599      		     const struct sadb_msghdr *mhp)
5600{
5601	const char *msg = "?" ;
5602	struct sadb_x_nat_t_type *type;
5603	struct sadb_x_nat_t_port *sport, *dport;
5604	struct sadb_address *iaddr, *raddr;
5605	struct sadb_x_nat_t_frag *frag;
5606
5607	if (mhp->ext[SADB_X_EXT_NAT_T_TYPE] == NULL ||
5608	    mhp->ext[SADB_X_EXT_NAT_T_SPORT] == NULL ||
5609	    mhp->ext[SADB_X_EXT_NAT_T_DPORT] == NULL)
5610		return 0;
5611
5612	if (mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) {
5613		msg = "TYPE";
5614		goto bad;
5615	}
5616
5617	if (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) {
5618		msg = "SPORT";
5619		goto bad;
5620	}
5621
5622	if (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport)) {
5623		msg = "DPORT";
5624		goto bad;
5625	}
5626
5627	if (mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL) {
5628		IPSECLOG(LOG_DEBUG, "NAT-T OAi present\n");
5629		if (mhp->extlen[SADB_X_EXT_NAT_T_OAI] < sizeof(*iaddr)) {
5630			msg = "OAI";
5631			goto bad;
5632		}
5633	}
5634
5635	if (mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL) {
5636		IPSECLOG(LOG_DEBUG, "NAT-T OAr present\n");
5637		if (mhp->extlen[SADB_X_EXT_NAT_T_OAR] < sizeof(*raddr)) {
5638			msg = "OAR";
5639			goto bad;
5640		}
5641	}
5642
5643	if (mhp->ext[SADB_X_EXT_NAT_T_FRAG] != NULL) {
5644	    if (mhp->extlen[SADB_X_EXT_NAT_T_FRAG] < sizeof(*frag)) {
5645		    msg = "FRAG";
5646		    goto bad;
5647	    }
5648	}
5649
5650	type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5651	sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5652	dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5653	iaddr = mhp->ext[SADB_X_EXT_NAT_T_OAI];
5654	raddr = mhp->ext[SADB_X_EXT_NAT_T_OAR];
5655	frag = mhp->ext[SADB_X_EXT_NAT_T_FRAG];
5656
5657	IPSECLOG(LOG_DEBUG, "type %d, sport = %d, dport = %d\n",
5658	    type->sadb_x_nat_t_type_type,
5659	    ntohs(sport->sadb_x_nat_t_port_port),
5660	    ntohs(dport->sadb_x_nat_t_port_port));
5661
5662	sav->natt_type = type->sadb_x_nat_t_type_type;
5663	key_porttosaddr(&sav->sah->saidx.src, sport->sadb_x_nat_t_port_port);
5664	key_porttosaddr(&sav->sah->saidx.dst, dport->sadb_x_nat_t_port_port);
5665	if (frag)
5666		sav->esp_frag = frag->sadb_x_nat_t_frag_fraglen;
5667	else
5668		sav->esp_frag = IP_MAXPACKET;
5669
5670	return 0;
5671bad:
5672	IPSECLOG(LOG_DEBUG, "invalid message %s\n", msg);
5673	__USE(msg);
5674	return -1;
5675}
5676
5677/* Just update the IPSEC_NAT_T ports if present */
5678static int
5679key_set_natt_ports(union sockaddr_union *src, union sockaddr_union *dst,
5680      		     const struct sadb_msghdr *mhp)
5681{
5682	if (mhp->ext[SADB_X_EXT_NAT_T_OAI] != NULL)
5683		IPSECLOG(LOG_DEBUG, "NAT-T OAi present\n");
5684	if (mhp->ext[SADB_X_EXT_NAT_T_OAR] != NULL)
5685		IPSECLOG(LOG_DEBUG, "NAT-T OAr present\n");
5686
5687	if ((mhp->ext[SADB_X_EXT_NAT_T_TYPE] != NULL) &&
5688	    (mhp->ext[SADB_X_EXT_NAT_T_SPORT] != NULL) &&
5689	    (mhp->ext[SADB_X_EXT_NAT_T_DPORT] != NULL)) {
5690		struct sadb_x_nat_t_type *type;
5691		struct sadb_x_nat_t_port *sport;
5692		struct sadb_x_nat_t_port *dport;
5693
5694		if ((mhp->extlen[SADB_X_EXT_NAT_T_TYPE] < sizeof(*type)) ||
5695		    (mhp->extlen[SADB_X_EXT_NAT_T_SPORT] < sizeof(*sport)) ||
5696		    (mhp->extlen[SADB_X_EXT_NAT_T_DPORT] < sizeof(*dport))) {
5697			IPSECLOG(LOG_DEBUG, "invalid message\n");
5698			return -1;
5699		}
5700
5701		type = mhp->ext[SADB_X_EXT_NAT_T_TYPE];
5702		sport = mhp->ext[SADB_X_EXT_NAT_T_SPORT];
5703		dport = mhp->ext[SADB_X_EXT_NAT_T_DPORT];
5704
5705		key_porttosaddr(src, sport->sadb_x_nat_t_port_port);
5706		key_porttosaddr(dst, dport->sadb_x_nat_t_port_port);
5707
5708		IPSECLOG(LOG_DEBUG, "type %d, sport = %d, dport = %d\n",
5709		    type->sadb_x_nat_t_type_type,
5710		    ntohs(sport->sadb_x_nat_t_port_port),
5711		    ntohs(dport->sadb_x_nat_t_port_port));
5712	}
5713
5714	return 0;
5715}
5716
5717
5718/*
5719 * SADB_UPDATE processing
5720 * receive
5721 *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5722 *       key(AE), (identity(SD),) (sensitivity)>
5723 * from the ikmpd, and update a secasvar entry whose status is SADB_SASTATE_LARVAL.
5724 * and send
5725 *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5726 *       (identity(SD),) (sensitivity)>
5727 * to the ikmpd.
5728 *
5729 * m will always be freed.
5730 */
5731static int
5732key_api_update(struct socket *so, struct mbuf *m, const struct sadb_msghdr *mhp)
5733{
5734	struct sadb_sa *sa0;
5735	const struct sockaddr *src, *dst;
5736	struct secasindex saidx;
5737	struct secashead *sah;
5738	struct secasvar *sav, *newsav, *oldsav;
5739	u_int16_t proto;
5740	u_int8_t mode;
5741	u_int16_t reqid;
5742	int error;
5743
5744	/* map satype to proto */
5745	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
5746	if (proto == 0) {
5747		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
5748		return key_senderror(so, m, EINVAL);
5749	}
5750
5751	if (mhp->ext[SADB_EXT_SA] == NULL ||
5752	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
5753	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
5754	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
5755	     mhp->ext[SADB_EXT_KEY_ENCRYPT] == NULL) ||
5756	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
5757	     mhp->ext[SADB_EXT_KEY_AUTH] == NULL) ||
5758	    (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL &&
5759	     mhp->ext[SADB_EXT_LIFETIME_SOFT] == NULL) ||
5760	    (mhp->ext[SADB_EXT_LIFETIME_HARD] == NULL &&
5761	     mhp->ext[SADB_EXT_LIFETIME_SOFT] != NULL)) {
5762		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5763		return key_senderror(so, m, EINVAL);
5764	}
5765	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
5766	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
5767	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
5768		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
5769		return key_senderror(so, m, EINVAL);
5770	}
5771	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
5772		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
5773		mode = sa2->sadb_x_sa2_mode;
5774		reqid = sa2->sadb_x_sa2_reqid;
5775	} else {
5776		mode = IPSEC_MODE_ANY;
5777		reqid = 0;
5778	}
5779	/* XXX boundary checking for other extensions */
5780
5781	sa0 = mhp->ext[SADB_EXT_SA];
5782	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
5783	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
5784
5785	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
5786	if (error != 0)
5787		return key_senderror(so, m, EINVAL);
5788
5789	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
5790	if (error != 0)
5791		return key_senderror(so, m, EINVAL);
5792
5793	/* get a SA header */
5794	sah = key_getsah_ref(&saidx, CMP_REQID);
5795	if (sah == NULL) {
5796		IPSECLOG(LOG_DEBUG, "no SA index found.\n");
5797		return key_senderror(so, m, ENOENT);
5798	}
5799
5800	/* set spidx if there */
5801	/* XXX rewrite */
5802	error = key_setident(sah, m, mhp);
5803	if (error)
5804		goto error_sah;
5805
5806	/* find a SA with sequence number. */
5807#ifdef IPSEC_DOSEQCHECK
5808	if (mhp->msg->sadb_msg_seq != 0) {
5809		sav = key_getsavbyseq(sah, mhp->msg->sadb_msg_seq);
5810		if (sav == NULL) {
5811			IPSECLOG(LOG_DEBUG,
5812			    "no larval SA with sequence %u exists.\n",
5813			    mhp->msg->sadb_msg_seq);
5814			error = ENOENT;
5815			goto error_sah;
5816		}
5817	}
5818#else
5819	sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
5820	if (sav == NULL) {
5821		IPSECLOG(LOG_DEBUG, "no such a SA found (spi:%u)\n",
5822		    (u_int32_t)ntohl(sa0->sadb_sa_spi));
5823		error = EINVAL;
5824		goto error_sah;
5825	}
5826#endif
5827
5828	/* validity check */
5829	if (sav->sah->saidx.proto != proto) {
5830		IPSECLOG(LOG_DEBUG, "protocol mismatched (DB=%u param=%u)\n",
5831		    sav->sah->saidx.proto, proto);
5832		error = EINVAL;
5833		goto error;
5834	}
5835#ifdef IPSEC_DOSEQCHECK
5836	if (sav->spi != sa0->sadb_sa_spi) {
5837		IPSECLOG(LOG_DEBUG, "SPI mismatched (DB:%u param:%u)\n",
5838		    (u_int32_t)ntohl(sav->spi),
5839		    (u_int32_t)ntohl(sa0->sadb_sa_spi));
5840		error = EINVAL;
5841		goto error;
5842	}
5843#endif
5844	if (sav->pid != mhp->msg->sadb_msg_pid) {
5845		IPSECLOG(LOG_DEBUG, "pid mismatched (DB:%u param:%u)\n",
5846		    sav->pid, mhp->msg->sadb_msg_pid);
5847		error = EINVAL;
5848		goto error;
5849	}
5850
5851	/*
5852	 * Allocate a new SA instead of modifying the existing SA directly
5853	 * to avoid race conditions.
5854	 */
5855	newsav = kmem_zalloc(sizeof(struct secasvar), KM_SLEEP);
5856
5857	/* copy sav values */
5858	newsav->spi = sav->spi;
5859	newsav->seq = sav->seq;
5860	newsav->created = sav->created;
5861	newsav->pid = sav->pid;
5862	newsav->sah = sav->sah;
5863 	KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
5864	    "DP from %s:%u update SA:%p to SA:%p spi=%#x proto=%d\n",
5865	    __func__, __LINE__, sav, newsav,
5866	    ntohl(newsav->spi), proto);
5867
5868	error = key_setsaval(newsav, m, mhp);
5869	if (error) {
5870		kmem_free(newsav, sizeof(*newsav));
5871		goto error;
5872	}
5873
5874	error = key_handle_natt_info(newsav, mhp);
5875	if (error != 0) {
5876		key_delsav(newsav);
5877		goto error;
5878	}
5879
5880	error = key_init_xform(newsav);
5881	if (error != 0) {
5882		key_delsav(newsav);
5883		goto error;
5884	}
5885
5886	/* Add to sah#savlist */
5887	key_init_sav(newsav);
5888	newsav->state = SADB_SASTATE_MATURE;
5889	mutex_enter(&key_sad.lock);
5890	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_MATURE, newsav);
5891	SAVLUT_WRITER_INSERT_HEAD(newsav);
5892	mutex_exit(&key_sad.lock);
5893	key_validate_savlist(sah, SADB_SASTATE_MATURE);
5894
5895	/*
5896	 * We need to lookup and remove the sav atomically, so get it again
5897	 * here by a special API while we have a reference to it.
5898	 */
5899	oldsav = key_lookup_and_remove_sav(sah, sa0->sadb_sa_spi, sav);
5900	KASSERT(oldsav == NULL || oldsav == sav);
5901	/* We can release the reference because of oldsav */
5902	KEY_SA_UNREF(&sav);
5903	if (oldsav == NULL) {
5904		/* Someone has already removed the sav.  Nothing to do. */
5905	} else {
5906		key_wait_sav(oldsav);
5907		key_destroy_sav(oldsav);
5908		oldsav = NULL;
5909	}
5910	sav = NULL;
5911
5912	key_sah_unref(sah);
5913	sah = NULL;
5914
5915    {
5916	struct mbuf *n;
5917
5918	/* set msg buf from mhp */
5919	n = key_getmsgbuf_x1(m, mhp);
5920	if (n == NULL) {
5921		IPSECLOG(LOG_DEBUG, "No more memory.\n");
5922		return key_senderror(so, m, ENOBUFS);
5923	}
5924
5925	m_freem(m);
5926	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
5927    }
5928error:
5929	KEY_SA_UNREF(&sav);
5930error_sah:
5931	key_sah_unref(sah);
5932	return key_senderror(so, m, error);
5933}
5934
5935/*
5936 * search SAD with sequence for a SA which state is SADB_SASTATE_LARVAL.
5937 * only called by key_api_update().
5938 * OUT:
5939 *	NULL	: not found
5940 *	others	: found, pointer to a SA.
5941 */
5942#ifdef IPSEC_DOSEQCHECK
5943static struct secasvar *
5944key_getsavbyseq(struct secashead *sah, u_int32_t seq)
5945{
5946	struct secasvar *sav;
5947	u_int state;
5948	int s;
5949
5950	state = SADB_SASTATE_LARVAL;
5951
5952	/* search SAD with sequence number ? */
5953	s = pserialize_read_enter();
5954	SAVLIST_READER_FOREACH(sav, sah, state) {
5955		KEY_CHKSASTATE(state, sav->state);
5956
5957		if (sav->seq == seq) {
5958			SA_ADDREF(sav);
5959			KEYDEBUG_PRINTF(KEYDEBUG_IPSEC_STAMP,
5960			    "DP cause refcnt++:%d SA:%p\n",
5961			    key_sa_refcnt(sav), sav);
5962			break;
5963		}
5964	}
5965	pserialize_read_exit(s);
5966
5967	return sav;
5968}
5969#endif
5970
5971/*
5972 * SADB_ADD processing
5973 * add an entry to SA database, when received
5974 *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5975 *       key(AE), (identity(SD),) (sensitivity)>
5976 * from the ikmpd,
5977 * and send
5978 *   <base, SA, (SA2), (lifetime(HSC),) address(SD), (address(P),)
5979 *       (identity(SD),) (sensitivity)>
5980 * to the ikmpd.
5981 *
5982 * IGNORE identity and sensitivity messages.
5983 *
5984 * m will always be freed.
5985 */
5986static int
5987key_api_add(struct socket *so, struct mbuf *m,
5988	const struct sadb_msghdr *mhp)
5989{
5990	struct sadb_sa *sa0;
5991	const struct sockaddr *src, *dst;
5992	struct secasindex saidx;
5993	struct secashead *sah;
5994	struct secasvar *newsav;
5995	u_int16_t proto;
5996	u_int8_t mode;
5997	u_int16_t reqid;
5998	int error;
5999
6000	/* map satype to proto */
6001	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
6002	if (proto == 0) {
6003		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
6004		return key_senderror(so, m, EINVAL);
6005	}
6006
6007	if (mhp->ext[SADB_EXT_SA] == NULL ||
6008	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
6009	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
6010	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_ESP &&
6011	     mhp->ext[SADB_EXT_KEY_ENCRYPT] == NULL) ||
6012	    (mhp->msg->sadb_msg_satype == SADB_SATYPE_AH &&
6013	     mhp->ext[SADB_EXT_KEY_AUTH] == NULL) ||
6014	    (mhp->ext[SADB_EXT_LIFETIME_HARD] != NULL &&
6015	     mhp->ext[SADB_EXT_LIFETIME_SOFT] == NULL) ||
6016	    (mhp->ext[SADB_EXT_LIFETIME_HARD] == NULL &&
6017	     mhp->ext[SADB_EXT_LIFETIME_SOFT] != NULL)) {
6018		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6019		return key_senderror(so, m, EINVAL);
6020	}
6021	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
6022	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
6023	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
6024		/* XXX need more */
6025		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6026		return key_senderror(so, m, EINVAL);
6027	}
6028	if (mhp->ext[SADB_X_EXT_SA2] != NULL) {
6029		const struct sadb_x_sa2 *sa2 = mhp->ext[SADB_X_EXT_SA2];
6030		mode = sa2->sadb_x_sa2_mode;
6031		reqid = sa2->sadb_x_sa2_reqid;
6032	} else {
6033		mode = IPSEC_MODE_ANY;
6034		reqid = 0;
6035	}
6036
6037	sa0 = mhp->ext[SADB_EXT_SA];
6038	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6039	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6040
6041	error = key_setsecasidx(proto, mode, reqid, src, dst, &saidx);
6042	if (error != 0)
6043		return key_senderror(so, m, EINVAL);
6044
6045	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6046	if (error != 0)
6047		return key_senderror(so, m, EINVAL);
6048
6049	/* get a SA header */
6050	sah = key_getsah_ref(&saidx, CMP_REQID);
6051	if (sah == NULL) {
6052		/* create a new SA header */
6053		sah = key_newsah(&saidx);
6054		if (sah == NULL) {
6055			IPSECLOG(LOG_DEBUG, "No more memory.\n");
6056			return key_senderror(so, m, ENOBUFS);
6057		}
6058	}
6059
6060	/* set spidx if there */
6061	/* XXX rewrite */
6062	error = key_setident(sah, m, mhp);
6063	if (error)
6064		goto error;
6065
6066    {
6067	struct secasvar *sav;
6068
6069	/* We can create new SA only if SPI is differenct. */
6070	sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
6071	if (sav != NULL) {
6072		KEY_SA_UNREF(&sav);
6073		IPSECLOG(LOG_DEBUG, "SA already exists.\n");
6074		error = EEXIST;
6075		goto error;
6076	}
6077    }
6078
6079	/* create new SA entry. */
6080	newsav = KEY_NEWSAV(m, mhp, &error, proto);
6081	if (newsav == NULL)
6082		goto error;
6083	newsav->sah = sah;
6084
6085	error = key_handle_natt_info(newsav, mhp);
6086	if (error != 0) {
6087		key_delsav(newsav);
6088		error = EINVAL;
6089		goto error;
6090	}
6091
6092	error = key_init_xform(newsav);
6093	if (error != 0) {
6094		key_delsav(newsav);
6095		goto error;
6096	}
6097
6098	/* Add to sah#savlist */
6099	key_init_sav(newsav);
6100	newsav->state = SADB_SASTATE_MATURE;
6101	mutex_enter(&key_sad.lock);
6102	SAVLIST_WRITER_INSERT_TAIL(sah, SADB_SASTATE_MATURE, newsav);
6103	SAVLUT_WRITER_INSERT_HEAD(newsav);
6104	mutex_exit(&key_sad.lock);
6105	key_validate_savlist(sah, SADB_SASTATE_MATURE);
6106
6107	key_sah_unref(sah);
6108	sah = NULL;
6109
6110	/*
6111	 * don't call key_freesav() here, as we would like to keep the SA
6112	 * in the database on success.
6113	 */
6114
6115    {
6116	struct mbuf *n;
6117
6118	/* set msg buf from mhp */
6119	n = key_getmsgbuf_x1(m, mhp);
6120	if (n == NULL) {
6121		IPSECLOG(LOG_DEBUG, "No more memory.\n");
6122		return key_senderror(so, m, ENOBUFS);
6123	}
6124
6125	m_freem(m);
6126	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6127    }
6128error:
6129	key_sah_unref(sah);
6130	return key_senderror(so, m, error);
6131}
6132
6133/* m is retained */
6134static int
6135key_setident(struct secashead *sah, struct mbuf *m,
6136	     const struct sadb_msghdr *mhp)
6137{
6138	const struct sadb_ident *idsrc, *iddst;
6139	int idsrclen, iddstlen;
6140
6141	KASSERT(!cpu_softintr_p());
6142	KASSERT(sah != NULL);
6143	KASSERT(m != NULL);
6144	KASSERT(mhp != NULL);
6145	KASSERT(mhp->msg != NULL);
6146
6147	/*
6148	 * Can be called with an existing sah from key_api_update().
6149	 */
6150	if (sah->idents != NULL) {
6151		kmem_free(sah->idents, sah->idents_len);
6152		sah->idents = NULL;
6153		sah->idents_len = 0;
6154	}
6155	if (sah->identd != NULL) {
6156		kmem_free(sah->identd, sah->identd_len);
6157		sah->identd = NULL;
6158		sah->identd_len = 0;
6159	}
6160
6161	/* don't make buffer if not there */
6162	if (mhp->ext[SADB_EXT_IDENTITY_SRC] == NULL &&
6163	    mhp->ext[SADB_EXT_IDENTITY_DST] == NULL) {
6164		sah->idents = NULL;
6165		sah->identd = NULL;
6166		return 0;
6167	}
6168
6169	if (mhp->ext[SADB_EXT_IDENTITY_SRC] == NULL ||
6170	    mhp->ext[SADB_EXT_IDENTITY_DST] == NULL) {
6171		IPSECLOG(LOG_DEBUG, "invalid identity.\n");
6172		return EINVAL;
6173	}
6174
6175	idsrc = mhp->ext[SADB_EXT_IDENTITY_SRC];
6176	iddst = mhp->ext[SADB_EXT_IDENTITY_DST];
6177	idsrclen = mhp->extlen[SADB_EXT_IDENTITY_SRC];
6178	iddstlen = mhp->extlen[SADB_EXT_IDENTITY_DST];
6179
6180	/* validity check */
6181	if (idsrc->sadb_ident_type != iddst->sadb_ident_type) {
6182		IPSECLOG(LOG_DEBUG, "ident type mismatched src %u, dst %u.\n",
6183		    idsrc->sadb_ident_type, iddst->sadb_ident_type);
6184		/*
6185		 * Some VPN appliances(e.g. NetScreen) can send different
6186		 * identifier types on IDii and IDir, so be able to allow
6187		 * such message.
6188		 */
6189		if (!ipsec_allow_different_idtype) {
6190			return EINVAL;
6191		}
6192	}
6193
6194	switch (idsrc->sadb_ident_type) {
6195	case SADB_IDENTTYPE_PREFIX:
6196	case SADB_IDENTTYPE_FQDN:
6197	case SADB_IDENTTYPE_USERFQDN:
6198	default:
6199		/* XXX do nothing */
6200		sah->idents = NULL;
6201		sah->identd = NULL;
6202	 	return 0;
6203	}
6204
6205	/* make structure */
6206	sah->idents = kmem_alloc(idsrclen, KM_SLEEP);
6207	sah->idents_len = idsrclen;
6208	sah->identd = kmem_alloc(iddstlen, KM_SLEEP);
6209	sah->identd_len = iddstlen;
6210	memcpy(sah->idents, idsrc, idsrclen);
6211	memcpy(sah->identd, iddst, iddstlen);
6212
6213	return 0;
6214}
6215
6216/*
6217 * m will not be freed on return. It never return NULL.
6218 * it is caller's responsibility to free the result.
6219 */
6220static struct mbuf *
6221key_getmsgbuf_x1(struct mbuf *m, const struct sadb_msghdr *mhp)
6222{
6223	struct mbuf *n;
6224
6225	KASSERT(m != NULL);
6226	KASSERT(mhp != NULL);
6227	KASSERT(mhp->msg != NULL);
6228
6229	/* create new sadb_msg to reply. */
6230	n = key_gather_mbuf(m, mhp, 1, 15, SADB_EXT_RESERVED,
6231	    SADB_EXT_SA, SADB_X_EXT_SA2,
6232	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST,
6233	    SADB_EXT_LIFETIME_HARD, SADB_EXT_LIFETIME_SOFT,
6234	    SADB_EXT_IDENTITY_SRC, SADB_EXT_IDENTITY_DST,
6235	    SADB_X_EXT_NAT_T_TYPE, SADB_X_EXT_NAT_T_SPORT,
6236	    SADB_X_EXT_NAT_T_DPORT, SADB_X_EXT_NAT_T_OAI,
6237	    SADB_X_EXT_NAT_T_OAR, SADB_X_EXT_NAT_T_FRAG);
6238
6239	KASSERT(n->m_len >= sizeof(struct sadb_msg));
6240
6241	mtod(n, struct sadb_msg *)->sadb_msg_errno = 0;
6242	mtod(n, struct sadb_msg *)->sadb_msg_len =
6243	    PFKEY_UNIT64(n->m_pkthdr.len);
6244
6245	return n;
6246}
6247
6248static int key_delete_all (struct socket *, struct mbuf *,
6249			   const struct sadb_msghdr *, u_int16_t);
6250
6251/*
6252 * SADB_DELETE processing
6253 * receive
6254 *   <base, SA(*), address(SD)>
6255 * from the ikmpd, and set SADB_SASTATE_DEAD,
6256 * and send,
6257 *   <base, SA(*), address(SD)>
6258 * to the ikmpd.
6259 *
6260 * m will always be freed.
6261 */
6262static int
6263key_api_delete(struct socket *so, struct mbuf *m,
6264	   const struct sadb_msghdr *mhp)
6265{
6266	struct sadb_sa *sa0;
6267	const struct sockaddr *src, *dst;
6268	struct secasindex saidx;
6269	struct secashead *sah;
6270	struct secasvar *sav = NULL;
6271	u_int16_t proto;
6272	int error;
6273
6274	/* map satype to proto */
6275	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
6276	if (proto == 0) {
6277		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
6278		return key_senderror(so, m, EINVAL);
6279	}
6280
6281	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
6282	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
6283		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6284		return key_senderror(so, m, EINVAL);
6285	}
6286
6287	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
6288	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
6289		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6290		return key_senderror(so, m, EINVAL);
6291	}
6292
6293	if (mhp->ext[SADB_EXT_SA] == NULL) {
6294		/*
6295		 * Caller wants us to delete all non-LARVAL SAs
6296		 * that match the src/dst.  This is used during
6297		 * IKE INITIAL-CONTACT.
6298		 */
6299		IPSECLOG(LOG_DEBUG, "doing delete all.\n");
6300		return key_delete_all(so, m, mhp, proto);
6301	} else if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa)) {
6302		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6303		return key_senderror(so, m, EINVAL);
6304	}
6305
6306	sa0 = mhp->ext[SADB_EXT_SA];
6307	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6308	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6309
6310	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
6311	if (error != 0)
6312		return key_senderror(so, m, EINVAL);
6313
6314	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6315	if (error != 0)
6316		return key_senderror(so, m, EINVAL);
6317
6318	/* get a SA header */
6319	sah = key_getsah_ref(&saidx, CMP_HEAD);
6320	if (sah != NULL) {
6321		/* get a SA with SPI. */
6322		sav = key_lookup_and_remove_sav(sah, sa0->sadb_sa_spi, NULL);
6323		key_sah_unref(sah);
6324	}
6325
6326	if (sav == NULL) {
6327		IPSECLOG(LOG_DEBUG, "no SA found.\n");
6328		return key_senderror(so, m, ENOENT);
6329	}
6330
6331	key_wait_sav(sav);
6332	key_destroy_sav(sav);
6333	sav = NULL;
6334
6335    {
6336	struct mbuf *n;
6337
6338	/* create new sadb_msg to reply. */
6339	n = key_gather_mbuf(m, mhp, 1, 4, SADB_EXT_RESERVED,
6340	    SADB_EXT_SA, SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6341
6342	key_fill_replymsg(n, 0);
6343	m_freem(m);
6344	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6345    }
6346}
6347
6348/*
6349 * delete all SAs for src/dst.  Called from key_api_delete().
6350 */
6351static int
6352key_delete_all(struct socket *so, struct mbuf *m,
6353	       const struct sadb_msghdr *mhp, u_int16_t proto)
6354{
6355	const struct sockaddr *src, *dst;
6356	struct secasindex saidx;
6357	struct secashead *sah;
6358	struct secasvar *sav;
6359	u_int state;
6360	int error;
6361
6362	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6363	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6364
6365	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
6366	if (error != 0)
6367		return key_senderror(so, m, EINVAL);
6368
6369	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6370	if (error != 0)
6371		return key_senderror(so, m, EINVAL);
6372
6373	sah = key_getsah_ref(&saidx, CMP_HEAD);
6374	if (sah != NULL) {
6375		/* Delete all non-LARVAL SAs. */
6376		SASTATE_ALIVE_FOREACH(state) {
6377			if (state == SADB_SASTATE_LARVAL)
6378				continue;
6379		restart:
6380			mutex_enter(&key_sad.lock);
6381			SAVLIST_WRITER_FOREACH(sav, sah, state) {
6382				sav->state = SADB_SASTATE_DEAD;
6383				key_unlink_sav(sav);
6384				mutex_exit(&key_sad.lock);
6385				key_destroy_sav(sav);
6386				goto restart;
6387			}
6388			mutex_exit(&key_sad.lock);
6389		}
6390		key_sah_unref(sah);
6391	}
6392    {
6393	struct mbuf *n;
6394
6395	/* create new sadb_msg to reply. */
6396	n = key_gather_mbuf(m, mhp, 1, 3, SADB_EXT_RESERVED,
6397	    SADB_EXT_ADDRESS_SRC, SADB_EXT_ADDRESS_DST);
6398
6399	key_fill_replymsg(n, 0);
6400	m_freem(m);
6401	return key_sendup_mbuf(so, n, KEY_SENDUP_ALL);
6402    }
6403}
6404
6405/*
6406 * SADB_GET processing
6407 * receive
6408 *   <base, SA(*), address(SD)>
6409 * from the ikmpd, and get a SP and a SA to respond,
6410 * and send,
6411 *   <base, SA, (lifetime(HSC),) address(SD), (address(P),) key(AE),
6412 *       (identity(SD),) (sensitivity)>
6413 * to the ikmpd.
6414 *
6415 * m will always be freed.
6416 */
6417static int
6418key_api_get(struct socket *so, struct mbuf *m,
6419	const struct sadb_msghdr *mhp)
6420{
6421	struct sadb_sa *sa0;
6422	const struct sockaddr *src, *dst;
6423	struct secasindex saidx;
6424	struct secasvar *sav = NULL;
6425	u_int16_t proto;
6426	int error;
6427
6428	/* map satype to proto */
6429	if ((proto = key_satype2proto(mhp->msg->sadb_msg_satype)) == 0) {
6430		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
6431		return key_senderror(so, m, EINVAL);
6432	}
6433
6434	if (mhp->ext[SADB_EXT_SA] == NULL ||
6435	    mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
6436	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL) {
6437		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6438		return key_senderror(so, m, EINVAL);
6439	}
6440	if (mhp->extlen[SADB_EXT_SA] < sizeof(struct sadb_sa) ||
6441	    mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
6442	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address)) {
6443		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
6444		return key_senderror(so, m, EINVAL);
6445	}
6446
6447	sa0 = mhp->ext[SADB_EXT_SA];
6448	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
6449	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
6450
6451	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
6452	if (error != 0)
6453		return key_senderror(so, m, EINVAL);
6454
6455	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
6456	if (error != 0)
6457		return key_senderror(so, m, EINVAL);
6458
6459	/* get a SA header */
6460    {
6461	struct secashead *sah;
6462	int s = pserialize_read_enter();
6463
6464	sah = key_getsah(&saidx, CMP_HEAD);
6465	if (sah != NULL) {
6466		/* get a SA with SPI. */
6467		sav = key_getsavbyspi(sah, sa0->sadb_sa_spi);
6468	}
6469	pserialize_read_exit(s);
6470    }
6471	if (sav == NULL) {
6472		IPSECLOG(LOG_DEBUG, "no SA found.\n");
6473		return key_senderror(so, m, ENOENT);
6474	}
6475
6476    {
6477	struct mbuf *n;
6478	u_int8_t satype;
6479
6480	/* map proto to satype */
6481	satype = key_proto2satype(sav->sah->saidx.proto);
6482	if (satype == 0) {
6483		KEY_SA_UNREF(&sav);
6484		IPSECLOG(LOG_DEBUG, "there was invalid proto in SAD.\n");
6485		return key_senderror(so, m, EINVAL);
6486	}
6487
6488	/* create new sadb_msg to reply. */
6489	n = key_setdumpsa(sav, SADB_GET, satype, mhp->msg->sadb_msg_seq,
6490	    mhp->msg->sadb_msg_pid);
6491	KEY_SA_UNREF(&sav);
6492	m_freem(m);
6493	return key_sendup_mbuf(so, n, KEY_SENDUP_ONE);
6494    }
6495}
6496
6497/* XXX make it sysctl-configurable? */
6498static void
6499key_getcomb_setlifetime(struct sadb_comb *comb)
6500{
6501
6502	comb->sadb_comb_soft_allocations = 1;
6503	comb->sadb_comb_hard_allocations = 1;
6504	comb->sadb_comb_soft_bytes = 0;
6505	comb->sadb_comb_hard_bytes = 0;
6506	comb->sadb_comb_hard_addtime = 86400;	/* 1 day */
6507	comb->sadb_comb_soft_addtime = comb->sadb_comb_hard_addtime * 80 / 100;
6508	comb->sadb_comb_hard_usetime = 28800;	/* 8 hours */
6509	comb->sadb_comb_soft_usetime = comb->sadb_comb_hard_usetime * 80 / 100;
6510}
6511
6512/*
6513 * XXX reorder combinations by preference
6514 * XXX no idea if the user wants ESP authentication or not
6515 */
6516static struct mbuf *
6517key_getcomb_esp(int mflag)
6518{
6519	struct sadb_comb *comb;
6520	const struct enc_xform *algo;
6521	struct mbuf *result = NULL, *m, *n;
6522	int encmin;
6523	int i, off, o;
6524	int totlen;
6525	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6526
6527	m = NULL;
6528	for (i = 1; i <= SADB_EALG_MAX; i++) {
6529		algo = esp_algorithm_lookup(i);
6530		if (algo == NULL)
6531			continue;
6532
6533		/* discard algorithms with key size smaller than system min */
6534		if (_BITS(algo->maxkey) < ipsec_esp_keymin)
6535			continue;
6536		if (_BITS(algo->minkey) < ipsec_esp_keymin)
6537			encmin = ipsec_esp_keymin;
6538		else
6539			encmin = _BITS(algo->minkey);
6540
6541		if (ipsec_esp_auth)
6542			m = key_getcomb_ah(mflag);
6543		else {
6544			KASSERTMSG(l <= MLEN,
6545			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
6546			MGET(m, mflag, MT_DATA);
6547			if (m) {
6548				m_align(m, l);
6549				m->m_len = l;
6550				m->m_next = NULL;
6551				memset(mtod(m, void *), 0, m->m_len);
6552			}
6553		}
6554		if (!m)
6555			goto fail;
6556
6557		totlen = 0;
6558		for (n = m; n; n = n->m_next)
6559			totlen += n->m_len;
6560		KASSERTMSG((totlen % l) == 0, "totlen=%u, l=%u", totlen, l);
6561
6562		for (off = 0; off < totlen; off += l) {
6563			n = m_pulldown(m, off, l, &o);
6564			if (!n) {
6565				/* m is already freed */
6566				goto fail;
6567			}
6568			comb = (struct sadb_comb *)(mtod(n, char *) + o);
6569			memset(comb, 0, sizeof(*comb));
6570			key_getcomb_setlifetime(comb);
6571			comb->sadb_comb_encrypt = i;
6572			comb->sadb_comb_encrypt_minbits = encmin;
6573			comb->sadb_comb_encrypt_maxbits = _BITS(algo->maxkey);
6574		}
6575
6576		if (!result)
6577			result = m;
6578		else
6579			m_cat(result, m);
6580	}
6581
6582	return result;
6583
6584 fail:
6585	if (result)
6586		m_freem(result);
6587	return NULL;
6588}
6589
6590static void
6591key_getsizes_ah(const struct auth_hash *ah, int alg,
6592	        u_int16_t* ksmin, u_int16_t* ksmax)
6593{
6594	*ksmin = *ksmax = ah->keysize;
6595	if (ah->keysize == 0) {
6596		/*
6597		 * Transform takes arbitrary key size but algorithm
6598		 * key size is restricted.  Enforce this here.
6599		 */
6600		switch (alg) {
6601		case SADB_X_AALG_MD5:	*ksmin = *ksmax = 16; break;
6602		case SADB_X_AALG_SHA:	*ksmin = *ksmax = 20; break;
6603		case SADB_X_AALG_NULL:	*ksmin = 0; *ksmax = 256; break;
6604		default:
6605			IPSECLOG(LOG_DEBUG, "unknown AH algorithm %u\n", alg);
6606			break;
6607		}
6608	}
6609}
6610
6611/*
6612 * XXX reorder combinations by preference
6613 */
6614static struct mbuf *
6615key_getcomb_ah(int mflag)
6616{
6617	struct sadb_comb *comb;
6618	const struct auth_hash *algo;
6619	struct mbuf *m;
6620	u_int16_t minkeysize, maxkeysize;
6621	int i;
6622	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6623
6624	m = NULL;
6625	for (i = 1; i <= SADB_AALG_MAX; i++) {
6626#if 1
6627		/* we prefer HMAC algorithms, not old algorithms */
6628		if (i != SADB_AALG_SHA1HMAC &&
6629		    i != SADB_AALG_MD5HMAC &&
6630		    i != SADB_X_AALG_SHA2_256 &&
6631		    i != SADB_X_AALG_SHA2_384 &&
6632		    i != SADB_X_AALG_SHA2_512)
6633			continue;
6634#endif
6635		algo = ah_algorithm_lookup(i);
6636		if (!algo)
6637			continue;
6638		key_getsizes_ah(algo, i, &minkeysize, &maxkeysize);
6639		/* discard algorithms with key size smaller than system min */
6640		if (_BITS(minkeysize) < ipsec_ah_keymin)
6641			continue;
6642
6643		if (!m) {
6644			KASSERTMSG(l <= MLEN,
6645			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
6646			MGET(m, mflag, MT_DATA);
6647			if (m) {
6648				m_align(m, l);
6649				m->m_len = l;
6650				m->m_next = NULL;
6651			}
6652		} else
6653			M_PREPEND(m, l, mflag);
6654		if (!m)
6655			return NULL;
6656
6657		if (m->m_len < sizeof(struct sadb_comb)) {
6658			m = m_pullup(m, sizeof(struct sadb_comb));
6659			if (m == NULL)
6660				return NULL;
6661		}
6662
6663		comb = mtod(m, struct sadb_comb *);
6664		memset(comb, 0, sizeof(*comb));
6665		key_getcomb_setlifetime(comb);
6666		comb->sadb_comb_auth = i;
6667		comb->sadb_comb_auth_minbits = _BITS(minkeysize);
6668		comb->sadb_comb_auth_maxbits = _BITS(maxkeysize);
6669	}
6670
6671	return m;
6672}
6673
6674/*
6675 * not really an official behavior.  discussed in pf_key@inner.net in Sep2000.
6676 * XXX reorder combinations by preference
6677 */
6678static struct mbuf *
6679key_getcomb_ipcomp(int mflag)
6680{
6681	struct sadb_comb *comb;
6682	const struct comp_algo *algo;
6683	struct mbuf *m;
6684	int i;
6685	const int l = PFKEY_ALIGN8(sizeof(struct sadb_comb));
6686
6687	m = NULL;
6688	for (i = 1; i <= SADB_X_CALG_MAX; i++) {
6689		algo = ipcomp_algorithm_lookup(i);
6690		if (!algo)
6691			continue;
6692
6693		if (!m) {
6694			KASSERTMSG(l <= MLEN,
6695			    "l=%u > MLEN=%lu", l, (u_long) MLEN);
6696			MGET(m, mflag, MT_DATA);
6697			if (m) {
6698				m_align(m, l);
6699				m->m_len = l;
6700				m->m_next = NULL;
6701			}
6702		} else
6703			M_PREPEND(m, l, mflag);
6704		if (!m)
6705			return NULL;
6706
6707		if (m->m_len < sizeof(struct sadb_comb)) {
6708			m = m_pullup(m, sizeof(struct sadb_comb));
6709			if (m == NULL)
6710				return NULL;
6711		}
6712
6713		comb = mtod(m, struct sadb_comb *);
6714		memset(comb, 0, sizeof(*comb));
6715		key_getcomb_setlifetime(comb);
6716		comb->sadb_comb_encrypt = i;
6717		/* what should we set into sadb_comb_*_{min,max}bits? */
6718	}
6719
6720	return m;
6721}
6722
6723/*
6724 * XXX no way to pass mode (transport/tunnel) to userland
6725 * XXX replay checking?
6726 * XXX sysctl interface to ipsec_{ah,esp}_keymin
6727 */
6728static struct mbuf *
6729key_getprop(const struct secasindex *saidx, int mflag)
6730{
6731	struct sadb_prop *prop;
6732	struct mbuf *m, *n;
6733	const int l = PFKEY_ALIGN8(sizeof(struct sadb_prop));
6734	int totlen;
6735
6736	switch (saidx->proto)  {
6737	case IPPROTO_ESP:
6738		m = key_getcomb_esp(mflag);
6739		break;
6740	case IPPROTO_AH:
6741		m = key_getcomb_ah(mflag);
6742		break;
6743	case IPPROTO_IPCOMP:
6744		m = key_getcomb_ipcomp(mflag);
6745		break;
6746	default:
6747		return NULL;
6748	}
6749
6750	if (!m)
6751		return NULL;
6752	M_PREPEND(m, l, mflag);
6753	if (!m)
6754		return NULL;
6755
6756	totlen = 0;
6757	for (n = m; n; n = n->m_next)
6758		totlen += n->m_len;
6759
6760	prop = mtod(m, struct sadb_prop *);
6761	memset(prop, 0, sizeof(*prop));
6762	prop->sadb_prop_len = PFKEY_UNIT64(totlen);
6763	prop->sadb_prop_exttype = SADB_EXT_PROPOSAL;
6764	prop->sadb_prop_replay = 32;	/* XXX */
6765
6766	return m;
6767}
6768
6769/*
6770 * SADB_ACQUIRE processing called by key_checkrequest() and key_api_acquire().
6771 * send
6772 *   <base, SA, address(SD), (address(P)), x_policy,
6773 *       (identity(SD),) (sensitivity,) proposal>
6774 * to KMD, and expect to receive
6775 *   <base> with SADB_ACQUIRE if error occurred,
6776 * or
6777 *   <base, src address, dst address, (SPI range)> with SADB_GETSPI
6778 * from KMD by PF_KEY.
6779 *
6780 * XXX x_policy is outside of RFC2367 (KAME extension).
6781 * XXX sensitivity is not supported.
6782 * XXX for ipcomp, RFC2367 does not define how to fill in proposal.
6783 * see comment for key_getcomb_ipcomp().
6784 *
6785 * OUT:
6786 *    0     : succeed
6787 *    others: error number
6788 */
6789static int
6790key_acquire(const struct secasindex *saidx, const struct secpolicy *sp, int mflag)
6791{
6792	struct mbuf *result = NULL, *m;
6793#ifndef IPSEC_NONBLOCK_ACQUIRE
6794	struct secacq *newacq;
6795#endif
6796	u_int8_t satype;
6797	int error = -1;
6798	u_int32_t seq;
6799
6800	/* sanity check */
6801	KASSERT(saidx != NULL);
6802	satype = key_proto2satype(saidx->proto);
6803	KASSERTMSG(satype != 0, "null satype, protocol %u", saidx->proto);
6804
6805#ifndef IPSEC_NONBLOCK_ACQUIRE
6806	/*
6807	 * We never do anything about acquiring SA.  There is another
6808	 * solution that kernel blocks to send SADB_ACQUIRE message until
6809	 * getting something message from IKEd.  In later case, to be
6810	 * managed with ACQUIRING list.
6811	 */
6812	/* Get an entry to check whether sending message or not. */
6813	mutex_enter(&key_misc.lock);
6814	newacq = key_getacq(saidx);
6815	if (newacq != NULL) {
6816		if (key_blockacq_count < newacq->count) {
6817			/* reset counter and do send message. */
6818			newacq->count = 0;
6819		} else {
6820			/* increment counter and do nothing. */
6821			newacq->count++;
6822			mutex_exit(&key_misc.lock);
6823			return 0;
6824		}
6825	} else {
6826		/* make new entry for blocking to send SADB_ACQUIRE. */
6827		newacq = key_newacq(saidx);
6828		if (newacq == NULL) {
6829			mutex_exit(&key_misc.lock);
6830			return ENOBUFS;
6831		}
6832
6833		/* add to key_misc.acqlist */
6834		LIST_INSERT_HEAD(&key_misc.acqlist, newacq, chain);
6835	}
6836
6837	seq = newacq->seq;
6838	mutex_exit(&key_misc.lock);
6839#else
6840	seq = (acq_seq = (acq_seq == ~0 ? 1 : ++acq_seq));
6841#endif
6842	m = key_setsadbmsg(SADB_ACQUIRE, 0, satype, seq, 0, 0, mflag);
6843	if (!m) {
6844		error = ENOBUFS;
6845		goto fail;
6846	}
6847	result = m;
6848
6849	/* set sadb_address for saidx's. */
6850	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &saidx->src.sa, FULLMASK,
6851	    IPSEC_ULPROTO_ANY, mflag);
6852	if (!m) {
6853		error = ENOBUFS;
6854		goto fail;
6855	}
6856	m_cat(result, m);
6857
6858	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &saidx->dst.sa, FULLMASK,
6859	    IPSEC_ULPROTO_ANY, mflag);
6860	if (!m) {
6861		error = ENOBUFS;
6862		goto fail;
6863	}
6864	m_cat(result, m);
6865
6866	/* XXX proxy address (optional) */
6867
6868	/* set sadb_x_policy */
6869	if (sp) {
6870		m = key_setsadbxpolicy(sp->policy, sp->spidx.dir, sp->id,
6871		    mflag);
6872		if (!m) {
6873			error = ENOBUFS;
6874			goto fail;
6875		}
6876		m_cat(result, m);
6877	}
6878
6879	/* XXX identity (optional) */
6880#if 0
6881	if (idexttype && fqdn) {
6882		/* create identity extension (FQDN) */
6883		struct sadb_ident *id;
6884		int fqdnlen;
6885
6886		fqdnlen = strlen(fqdn) + 1;	/* +1 for terminating-NUL */
6887		id = (struct sadb_ident *)p;
6888		memset(id, 0, sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6889		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(fqdnlen));
6890		id->sadb_ident_exttype = idexttype;
6891		id->sadb_ident_type = SADB_IDENTTYPE_FQDN;
6892		memcpy(id + 1, fqdn, fqdnlen);
6893		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(fqdnlen);
6894	}
6895
6896	if (idexttype) {
6897		/* create identity extension (USERFQDN) */
6898		struct sadb_ident *id;
6899		int userfqdnlen;
6900
6901		if (userfqdn) {
6902			/* +1 for terminating-NUL */
6903			userfqdnlen = strlen(userfqdn) + 1;
6904		} else
6905			userfqdnlen = 0;
6906		id = (struct sadb_ident *)p;
6907		memset(id, 0, sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6908		id->sadb_ident_len = PFKEY_UNIT64(sizeof(*id) + PFKEY_ALIGN8(userfqdnlen));
6909		id->sadb_ident_exttype = idexttype;
6910		id->sadb_ident_type = SADB_IDENTTYPE_USERFQDN;
6911		/* XXX is it correct? */
6912		if (curlwp)
6913			id->sadb_ident_id = kauth_cred_getuid(curlwp->l_cred);
6914		if (userfqdn && userfqdnlen)
6915			memcpy(id + 1, userfqdn, userfqdnlen);
6916		p += sizeof(struct sadb_ident) + PFKEY_ALIGN8(userfqdnlen);
6917	}
6918#endif
6919
6920	/* XXX sensitivity (optional) */
6921
6922	/* create proposal/combination extension */
6923	m = key_getprop(saidx, mflag);
6924#if 0
6925	/*
6926	 * spec conformant: always attach proposal/combination extension,
6927	 * the problem is that we have no way to attach it for ipcomp,
6928	 * due to the way sadb_comb is declared in RFC2367.
6929	 */
6930	if (!m) {
6931		error = ENOBUFS;
6932		goto fail;
6933	}
6934	m_cat(result, m);
6935#else
6936	/*
6937	 * outside of spec; make proposal/combination extension optional.
6938	 */
6939	if (m)
6940		m_cat(result, m);
6941#endif
6942
6943	KASSERT(result->m_flags & M_PKTHDR);
6944	KASSERT(result->m_len >= sizeof(struct sadb_msg));
6945
6946	result->m_pkthdr.len = 0;
6947	for (m = result; m; m = m->m_next)
6948		result->m_pkthdr.len += m->m_len;
6949
6950	mtod(result, struct sadb_msg *)->sadb_msg_len =
6951	    PFKEY_UNIT64(result->m_pkthdr.len);
6952
6953	/*
6954	 * Called from key_api_acquire that must come from userland, so
6955	 * we can call key_sendup_mbuf immediately.
6956	 */
6957	if (mflag == M_WAITOK)
6958		return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
6959	/*
6960	 * XXX we cannot call key_sendup_mbuf directly here because
6961	 * it can cause a deadlock:
6962	 * - We have a reference to an SP (and an SA) here
6963	 * - key_sendup_mbuf will try to take key_so_mtx
6964	 * - Some other thread may try to localcount_drain to the SP with
6965	 *   holding key_so_mtx in say key_api_spdflush
6966	 * - In this case localcount_drain never return because key_sendup_mbuf
6967	 *   that has stuck on key_so_mtx never release a reference to the SP
6968	 *
6969	 * So defer key_sendup_mbuf to the timer.
6970	 */
6971	return key_acquire_sendup_mbuf_later(result);
6972
6973 fail:
6974	if (result)
6975		m_freem(result);
6976	return error;
6977}
6978
6979static struct mbuf *key_acquire_mbuf_head = NULL;
6980static unsigned key_acquire_mbuf_count = 0;
6981#define KEY_ACQUIRE_MBUF_MAX	10
6982
6983static void
6984key_acquire_sendup_pending_mbuf(void)
6985{
6986	struct mbuf *m, *prev;
6987	int error;
6988
6989again:
6990	prev = NULL;
6991	mutex_enter(&key_misc.lock);
6992	m = key_acquire_mbuf_head;
6993	/* Get an earliest mbuf (one at the tail of the list) */
6994	while (m != NULL) {
6995		if (m->m_nextpkt == NULL) {
6996			if (prev != NULL)
6997				prev->m_nextpkt = NULL;
6998			if (m == key_acquire_mbuf_head)
6999				key_acquire_mbuf_head = NULL;
7000			key_acquire_mbuf_count--;
7001			break;
7002		}
7003		prev = m;
7004		m = m->m_nextpkt;
7005	}
7006	mutex_exit(&key_misc.lock);
7007
7008	if (m == NULL)
7009		return;
7010
7011	m->m_nextpkt = NULL;
7012	error = key_sendup_mbuf(NULL, m, KEY_SENDUP_REGISTERED);
7013	if (error != 0)
7014		IPSECLOG(LOG_WARNING, "key_sendup_mbuf failed (error=%d)\n",
7015		    error);
7016
7017	if (prev != NULL)
7018		goto again;
7019}
7020
7021static int
7022key_acquire_sendup_mbuf_later(struct mbuf *m)
7023{
7024
7025	mutex_enter(&key_misc.lock);
7026	/* Avoid queuing too much mbufs */
7027	if (key_acquire_mbuf_count >= KEY_ACQUIRE_MBUF_MAX) {
7028		mutex_exit(&key_misc.lock);
7029		m_freem(m);
7030		return ENOBUFS; /* XXX */
7031	}
7032	/* Enqueue mbuf at the head of the list */
7033	m->m_nextpkt = key_acquire_mbuf_head;
7034	key_acquire_mbuf_head = m;
7035	key_acquire_mbuf_count++;
7036	mutex_exit(&key_misc.lock);
7037
7038	/* Kick the timer */
7039	key_timehandler(NULL);
7040
7041	return 0;
7042}
7043
7044#ifndef IPSEC_NONBLOCK_ACQUIRE
7045static struct secacq *
7046key_newacq(const struct secasindex *saidx)
7047{
7048	struct secacq *newacq;
7049
7050	/* get new entry */
7051	newacq = kmem_intr_zalloc(sizeof(struct secacq), KM_NOSLEEP);
7052	if (newacq == NULL) {
7053		IPSECLOG(LOG_DEBUG, "No more memory.\n");
7054		return NULL;
7055	}
7056
7057	/* copy secindex */
7058	memcpy(&newacq->saidx, saidx, sizeof(newacq->saidx));
7059	newacq->seq = (acq_seq == ~0 ? 1 : ++acq_seq);
7060	newacq->created = time_uptime;
7061	newacq->count = 0;
7062
7063	return newacq;
7064}
7065
7066static struct secacq *
7067key_getacq(const struct secasindex *saidx)
7068{
7069	struct secacq *acq;
7070
7071	KASSERT(mutex_owned(&key_misc.lock));
7072
7073	LIST_FOREACH(acq, &key_misc.acqlist, chain) {
7074		if (key_saidx_match(saidx, &acq->saidx, CMP_EXACTLY))
7075			return acq;
7076	}
7077
7078	return NULL;
7079}
7080
7081static struct secacq *
7082key_getacqbyseq(u_int32_t seq)
7083{
7084	struct secacq *acq;
7085
7086	KASSERT(mutex_owned(&key_misc.lock));
7087
7088	LIST_FOREACH(acq, &key_misc.acqlist, chain) {
7089		if (acq->seq == seq)
7090			return acq;
7091	}
7092
7093	return NULL;
7094}
7095#endif
7096
7097#ifdef notyet
7098static struct secspacq *
7099key_newspacq(const struct secpolicyindex *spidx)
7100{
7101	struct secspacq *acq;
7102
7103	/* get new entry */
7104	acq = kmem_intr_zalloc(sizeof(struct secspacq), KM_NOSLEEP);
7105	if (acq == NULL) {
7106		IPSECLOG(LOG_DEBUG, "No more memory.\n");
7107		return NULL;
7108	}
7109
7110	/* copy secindex */
7111	memcpy(&acq->spidx, spidx, sizeof(acq->spidx));
7112	acq->created = time_uptime;
7113	acq->count = 0;
7114
7115	return acq;
7116}
7117
7118static struct secspacq *
7119key_getspacq(const struct secpolicyindex *spidx)
7120{
7121	struct secspacq *acq;
7122
7123	LIST_FOREACH(acq, &key_misc.spacqlist, chain) {
7124		if (key_spidx_match_exactly(spidx, &acq->spidx))
7125			return acq;
7126	}
7127
7128	return NULL;
7129}
7130#endif /* notyet */
7131
7132/*
7133 * SADB_ACQUIRE processing,
7134 * in first situation, is receiving
7135 *   <base>
7136 * from the ikmpd, and clear sequence of its secasvar entry.
7137 *
7138 * In second situation, is receiving
7139 *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
7140 * from a user land process, and return
7141 *   <base, address(SD), (address(P),) (identity(SD),) (sensitivity,) proposal>
7142 * to the socket.
7143 *
7144 * m will always be freed.
7145 */
7146static int
7147key_api_acquire(struct socket *so, struct mbuf *m,
7148      	     const struct sadb_msghdr *mhp)
7149{
7150	const struct sockaddr *src, *dst;
7151	struct secasindex saidx;
7152	u_int16_t proto;
7153	int error;
7154
7155	/*
7156	 * Error message from KMd.
7157	 * We assume that if error was occurred in IKEd, the length of PFKEY
7158	 * message is equal to the size of sadb_msg structure.
7159	 * We do not raise error even if error occurred in this function.
7160	 */
7161	if (mhp->msg->sadb_msg_len == PFKEY_UNIT64(sizeof(struct sadb_msg))) {
7162#ifndef IPSEC_NONBLOCK_ACQUIRE
7163		struct secacq *acq;
7164
7165		/* check sequence number */
7166		if (mhp->msg->sadb_msg_seq == 0) {
7167			IPSECLOG(LOG_DEBUG, "must specify sequence number.\n");
7168			m_freem(m);
7169			return 0;
7170		}
7171
7172		mutex_enter(&key_misc.lock);
7173		acq = key_getacqbyseq(mhp->msg->sadb_msg_seq);
7174		if (acq == NULL) {
7175			mutex_exit(&key_misc.lock);
7176			/*
7177			 * the specified larval SA is already gone, or we got
7178			 * a bogus sequence number.  we can silently ignore it.
7179			 */
7180			m_freem(m);
7181			return 0;
7182		}
7183
7184		/* reset acq counter in order to deletion by timehander. */
7185		acq->created = time_uptime;
7186		acq->count = 0;
7187		mutex_exit(&key_misc.lock);
7188#endif
7189		m_freem(m);
7190		return 0;
7191	}
7192
7193	/*
7194	 * This message is from user land.
7195	 */
7196
7197	/* map satype to proto */
7198	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
7199	if (proto == 0) {
7200		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
7201		return key_senderror(so, m, EINVAL);
7202	}
7203
7204	if (mhp->ext[SADB_EXT_ADDRESS_SRC] == NULL ||
7205	    mhp->ext[SADB_EXT_ADDRESS_DST] == NULL ||
7206	    mhp->ext[SADB_EXT_PROPOSAL] == NULL) {
7207		/* error */
7208		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
7209		return key_senderror(so, m, EINVAL);
7210	}
7211	if (mhp->extlen[SADB_EXT_ADDRESS_SRC] < sizeof(struct sadb_address) ||
7212	    mhp->extlen[SADB_EXT_ADDRESS_DST] < sizeof(struct sadb_address) ||
7213	    mhp->extlen[SADB_EXT_PROPOSAL] < sizeof(struct sadb_prop)) {
7214		/* error */
7215		IPSECLOG(LOG_DEBUG, "invalid message is passed.\n");
7216		return key_senderror(so, m, EINVAL);
7217	}
7218
7219	src = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_SRC);
7220	dst = key_msghdr_get_sockaddr(mhp, SADB_EXT_ADDRESS_DST);
7221
7222	error = key_setsecasidx(proto, IPSEC_MODE_ANY, 0, src, dst, &saidx);
7223	if (error != 0)
7224		return key_senderror(so, m, EINVAL);
7225
7226	error = key_set_natt_ports(&saidx.src, &saidx.dst, mhp);
7227	if (error != 0)
7228		return key_senderror(so, m, EINVAL);
7229
7230	/* get a SA index */
7231    {
7232	struct secashead *sah;
7233	int s = pserialize_read_enter();
7234
7235	sah = key_getsah(&saidx, CMP_MODE_REQID);
7236	if (sah != NULL) {
7237		pserialize_read_exit(s);
7238		IPSECLOG(LOG_DEBUG, "a SA exists already.\n");
7239		return key_senderror(so, m, EEXIST);
7240	}
7241	pserialize_read_exit(s);
7242    }
7243
7244	error = key_acquire(&saidx, NULL, M_WAITOK);
7245	if (error != 0) {
7246		IPSECLOG(LOG_DEBUG, "error %d returned from key_acquire.\n",
7247		    error);
7248		return key_senderror(so, m, error);
7249	}
7250
7251	return key_sendup_mbuf(so, m, KEY_SENDUP_REGISTERED);
7252}
7253
7254/*
7255 * SADB_REGISTER processing.
7256 * If SATYPE_UNSPEC has been passed as satype, only return sabd_supported.
7257 * receive
7258 *   <base>
7259 * from the ikmpd, and register a socket to send PF_KEY messages,
7260 * and send
7261 *   <base, supported>
7262 * to KMD by PF_KEY.
7263 * If socket is detached, must free from regnode.
7264 *
7265 * m will always be freed.
7266 */
7267static int
7268key_api_register(struct socket *so, struct mbuf *m,
7269	     const struct sadb_msghdr *mhp)
7270{
7271	struct secreg *reg, *newreg = 0;
7272
7273	/* check for invalid register message */
7274	if (mhp->msg->sadb_msg_satype >= __arraycount(key_misc.reglist))
7275		return key_senderror(so, m, EINVAL);
7276
7277	/* When SATYPE_UNSPEC is specified, only return sabd_supported. */
7278	if (mhp->msg->sadb_msg_satype == SADB_SATYPE_UNSPEC)
7279		goto setmsg;
7280
7281	/* Allocate regnode in advance, out of mutex */
7282	newreg = kmem_zalloc(sizeof(*newreg), KM_SLEEP);
7283
7284	/* check whether existing or not */
7285	mutex_enter(&key_misc.lock);
7286	LIST_FOREACH(reg, &key_misc.reglist[mhp->msg->sadb_msg_satype], chain) {
7287		if (reg->so == so) {
7288			IPSECLOG(LOG_DEBUG, "socket exists already.\n");
7289			mutex_exit(&key_misc.lock);
7290			kmem_free(newreg, sizeof(*newreg));
7291			return key_senderror(so, m, EEXIST);
7292		}
7293	}
7294
7295	newreg->so = so;
7296	((struct keycb *)sotorawcb(so))->kp_registered++;
7297
7298	/* add regnode to key_misc.reglist. */
7299	LIST_INSERT_HEAD(&key_misc.reglist[mhp->msg->sadb_msg_satype], newreg, chain);
7300	mutex_exit(&key_misc.lock);
7301
7302  setmsg:
7303    {
7304	struct mbuf *n;
7305	struct sadb_supported *sup;
7306	u_int len, alen, elen;
7307	int off;
7308	int i;
7309	struct sadb_alg *alg;
7310
7311	/* create new sadb_msg to reply. */
7312	alen = 0;
7313	for (i = 1; i <= SADB_AALG_MAX; i++) {
7314		if (ah_algorithm_lookup(i))
7315			alen += sizeof(struct sadb_alg);
7316	}
7317	if (alen)
7318		alen += sizeof(struct sadb_supported);
7319	elen = 0;
7320	for (i = 1; i <= SADB_EALG_MAX; i++) {
7321		if (esp_algorithm_lookup(i))
7322			elen += sizeof(struct sadb_alg);
7323	}
7324	if (elen)
7325		elen += sizeof(struct sadb_supported);
7326
7327	len = sizeof(struct sadb_msg) + alen + elen;
7328
7329	if (len > MCLBYTES)
7330		return key_senderror(so, m, ENOBUFS);
7331
7332	n = key_alloc_mbuf_simple(len, M_WAITOK);
7333	n->m_pkthdr.len = n->m_len = len;
7334	n->m_next = NULL;
7335	off = 0;
7336
7337	m_copydata(m, 0, sizeof(struct sadb_msg), mtod(n, char *) + off);
7338	key_fill_replymsg(n, 0);
7339
7340	off += PFKEY_ALIGN8(sizeof(struct sadb_msg));
7341
7342	/* for authentication algorithm */
7343	if (alen) {
7344		sup = (struct sadb_supported *)(mtod(n, char *) + off);
7345		sup->sadb_supported_len = PFKEY_UNIT64(alen);
7346		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_AUTH;
7347		sup->sadb_supported_reserved = 0;
7348		off += PFKEY_ALIGN8(sizeof(*sup));
7349
7350		for (i = 1; i <= SADB_AALG_MAX; i++) {
7351			const struct auth_hash *aalgo;
7352			u_int16_t minkeysize, maxkeysize;
7353
7354			aalgo = ah_algorithm_lookup(i);
7355			if (!aalgo)
7356				continue;
7357			alg = (struct sadb_alg *)(mtod(n, char *) + off);
7358			alg->sadb_alg_id = i;
7359			alg->sadb_alg_ivlen = 0;
7360			key_getsizes_ah(aalgo, i, &minkeysize, &maxkeysize);
7361			alg->sadb_alg_minbits = _BITS(minkeysize);
7362			alg->sadb_alg_maxbits = _BITS(maxkeysize);
7363			alg->sadb_alg_reserved = 0;
7364			off += PFKEY_ALIGN8(sizeof(*alg));
7365		}
7366	}
7367
7368	/* for encryption algorithm */
7369	if (elen) {
7370		sup = (struct sadb_supported *)(mtod(n, char *) + off);
7371		sup->sadb_supported_len = PFKEY_UNIT64(elen);
7372		sup->sadb_supported_exttype = SADB_EXT_SUPPORTED_ENCRYPT;
7373		sup->sadb_supported_reserved = 0;
7374		off += PFKEY_ALIGN8(sizeof(*sup));
7375
7376		for (i = 1; i <= SADB_EALG_MAX; i++) {
7377			const struct enc_xform *ealgo;
7378
7379			ealgo = esp_algorithm_lookup(i);
7380			if (!ealgo)
7381				continue;
7382			alg = (struct sadb_alg *)(mtod(n, char *) + off);
7383			alg->sadb_alg_id = i;
7384			alg->sadb_alg_ivlen = ealgo->blocksize;
7385			alg->sadb_alg_minbits = _BITS(ealgo->minkey);
7386			alg->sadb_alg_maxbits = _BITS(ealgo->maxkey);
7387			alg->sadb_alg_reserved = 0;
7388			off += PFKEY_ALIGN8(sizeof(struct sadb_alg));
7389		}
7390	}
7391
7392	KASSERTMSG(off == len, "length inconsistency");
7393
7394	m_freem(m);
7395	return key_sendup_mbuf(so, n, KEY_SENDUP_REGISTERED);
7396    }
7397}
7398
7399/*
7400 * free secreg entry registered.
7401 * XXX: I want to do free a socket marked done SADB_RESIGER to socket.
7402 */
7403void
7404key_freereg(struct socket *so)
7405{
7406	struct secreg *reg;
7407	int i;
7408
7409	KASSERT(!cpu_softintr_p());
7410	KASSERT(so != NULL);
7411
7412	/*
7413	 * check whether existing or not.
7414	 * check all type of SA, because there is a potential that
7415	 * one socket is registered to multiple type of SA.
7416	 */
7417	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
7418		mutex_enter(&key_misc.lock);
7419		LIST_FOREACH(reg, &key_misc.reglist[i], chain) {
7420			if (reg->so == so) {
7421				LIST_REMOVE(reg, chain);
7422				break;
7423			}
7424		}
7425		mutex_exit(&key_misc.lock);
7426		if (reg != NULL)
7427			kmem_free(reg, sizeof(*reg));
7428	}
7429
7430	return;
7431}
7432
7433/*
7434 * SADB_EXPIRE processing
7435 * send
7436 *   <base, SA, SA2, lifetime(C and one of HS), address(SD)>
7437 * to KMD by PF_KEY.
7438 * NOTE: We send only soft lifetime extension.
7439 *
7440 * OUT:	0	: succeed
7441 *	others	: error number
7442 */
7443static int
7444key_expire(struct secasvar *sav)
7445{
7446	int s;
7447	int satype;
7448	struct mbuf *result = NULL, *m;
7449	int len;
7450	int error = -1;
7451	struct sadb_lifetime *lt;
7452	lifetime_counters_t sum = {0};
7453
7454	/* XXX: Why do we lock ? */
7455	s = splsoftnet();	/*called from softclock()*/
7456
7457	KASSERT(sav != NULL);
7458
7459	satype = key_proto2satype(sav->sah->saidx.proto);
7460	KASSERTMSG(satype != 0, "invalid proto is passed");
7461
7462	/* set msg header */
7463	m = key_setsadbmsg(SADB_EXPIRE, 0, satype, sav->seq, 0, key_sa_refcnt(sav),
7464	    M_WAITOK);
7465	result = m;
7466
7467	/* create SA extension */
7468	m = key_setsadbsa(sav);
7469	m_cat(result, m);
7470
7471	/* create SA extension */
7472	m = key_setsadbxsa2(sav->sah->saidx.mode,
7473	    sav->replay ? sav->replay->count : 0, sav->sah->saidx.reqid);
7474	m_cat(result, m);
7475
7476	/* create lifetime extension (current and soft) */
7477	len = PFKEY_ALIGN8(sizeof(*lt)) * 2;
7478	m = key_alloc_mbuf(len, M_WAITOK);
7479	KASSERT(m->m_next == NULL);
7480
7481	memset(mtod(m, void *), 0, len);
7482	lt = mtod(m, struct sadb_lifetime *);
7483	lt->sadb_lifetime_len = PFKEY_UNIT64(sizeof(struct sadb_lifetime));
7484	lt->sadb_lifetime_exttype = SADB_EXT_LIFETIME_CURRENT;
7485	percpu_foreach_xcall(sav->lft_c_counters_percpu,
7486	    XC_HIGHPRI_IPL(IPL_SOFTNET), key_sum_lifetime_counters, sum);
7487	lt->sadb_lifetime_allocations = sum[LIFETIME_COUNTER_ALLOCATIONS];
7488	lt->sadb_lifetime_bytes = sum[LIFETIME_COUNTER_BYTES];
7489	lt->sadb_lifetime_addtime =
7490	    time_mono_to_wall(sav->lft_c->sadb_lifetime_addtime);
7491	lt->sadb_lifetime_usetime =
7492	    time_mono_to_wall(sav->lft_c->sadb_lifetime_usetime);
7493	lt = (struct sadb_lifetime *)(mtod(m, char *) + len / 2);
7494	memcpy(lt, sav->lft_s, sizeof(*lt));
7495	m_cat(result, m);
7496
7497	/* set sadb_address for source */
7498	m = key_setsadbaddr(SADB_EXT_ADDRESS_SRC, &sav->sah->saidx.src.sa,
7499	    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
7500	m_cat(result, m);
7501
7502	/* set sadb_address for destination */
7503	m = key_setsadbaddr(SADB_EXT_ADDRESS_DST, &sav->sah->saidx.dst.sa,
7504	    FULLMASK, IPSEC_ULPROTO_ANY, M_WAITOK);
7505	m_cat(result, m);
7506
7507	if ((result->m_flags & M_PKTHDR) == 0) {
7508		error = EINVAL;
7509		goto fail;
7510	}
7511
7512	if (result->m_len < sizeof(struct sadb_msg)) {
7513		result = m_pullup(result, sizeof(struct sadb_msg));
7514		if (result == NULL) {
7515			error = ENOBUFS;
7516			goto fail;
7517		}
7518	}
7519
7520	result->m_pkthdr.len = 0;
7521	for (m = result; m; m = m->m_next)
7522		result->m_pkthdr.len += m->m_len;
7523
7524	mtod(result, struct sadb_msg *)->sadb_msg_len =
7525	    PFKEY_UNIT64(result->m_pkthdr.len);
7526
7527	splx(s);
7528	return key_sendup_mbuf(NULL, result, KEY_SENDUP_REGISTERED);
7529
7530 fail:
7531	if (result)
7532		m_freem(result);
7533	splx(s);
7534	return error;
7535}
7536
7537/*
7538 * SADB_FLUSH processing
7539 * receive
7540 *   <base>
7541 * from the ikmpd, and free all entries in secastree.
7542 * and send,
7543 *   <base>
7544 * to the ikmpd.
7545 * NOTE: to do is only marking SADB_SASTATE_DEAD.
7546 *
7547 * m will always be freed.
7548 */
7549static int
7550key_api_flush(struct socket *so, struct mbuf *m,
7551          const struct sadb_msghdr *mhp)
7552{
7553	struct sadb_msg *newmsg;
7554	struct secashead *sah;
7555	struct secasvar *sav;
7556	u_int16_t proto;
7557	u_int8_t state;
7558	int s;
7559
7560	/* map satype to proto */
7561	proto = key_satype2proto(mhp->msg->sadb_msg_satype);
7562	if (proto == 0) {
7563		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
7564		return key_senderror(so, m, EINVAL);
7565	}
7566
7567	/* no SATYPE specified, i.e. flushing all SA. */
7568	s = pserialize_read_enter();
7569	SAHLIST_READER_FOREACH(sah) {
7570		if (mhp->msg->sadb_msg_satype != SADB_SATYPE_UNSPEC &&
7571		    proto != sah->saidx.proto)
7572			continue;
7573
7574		key_sah_ref(sah);
7575		pserialize_read_exit(s);
7576
7577		SASTATE_ALIVE_FOREACH(state) {
7578		restart:
7579			mutex_enter(&key_sad.lock);
7580			SAVLIST_WRITER_FOREACH(sav, sah, state) {
7581				sav->state = SADB_SASTATE_DEAD;
7582				key_unlink_sav(sav);
7583				mutex_exit(&key_sad.lock);
7584				key_destroy_sav(sav);
7585				goto restart;
7586			}
7587			mutex_exit(&key_sad.lock);
7588		}
7589
7590		s = pserialize_read_enter();
7591		sah->state = SADB_SASTATE_DEAD;
7592		key_sah_unref(sah);
7593	}
7594	pserialize_read_exit(s);
7595
7596	if (m->m_len < sizeof(struct sadb_msg) ||
7597	    sizeof(struct sadb_msg) > m->m_len + M_TRAILINGSPACE(m)) {
7598		IPSECLOG(LOG_DEBUG, "No more memory.\n");
7599		return key_senderror(so, m, ENOBUFS);
7600	}
7601
7602	if (m->m_next)
7603		m_freem(m->m_next);
7604	m->m_next = NULL;
7605	m->m_pkthdr.len = m->m_len = sizeof(struct sadb_msg);
7606	newmsg = mtod(m, struct sadb_msg *);
7607	newmsg->sadb_msg_errno = 0;
7608	newmsg->sadb_msg_len = PFKEY_UNIT64(m->m_pkthdr.len);
7609
7610	return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7611}
7612
7613
7614static struct mbuf *
7615key_setdump_chain(u_int8_t req_satype, int *errorp, int *lenp, pid_t pid)
7616{
7617	struct secashead *sah;
7618	struct secasvar *sav;
7619	u_int16_t proto;
7620	u_int8_t satype;
7621	u_int8_t state;
7622	int cnt;
7623	struct mbuf *m, *n, *prev;
7624
7625	KASSERT(mutex_owned(&key_sad.lock));
7626
7627	*lenp = 0;
7628
7629	/* map satype to proto */
7630	proto = key_satype2proto(req_satype);
7631	if (proto == 0) {
7632		*errorp = EINVAL;
7633		return (NULL);
7634	}
7635
7636	/* count sav entries to be sent to userland. */
7637	cnt = 0;
7638	SAHLIST_WRITER_FOREACH(sah) {
7639		if (req_satype != SADB_SATYPE_UNSPEC &&
7640		    proto != sah->saidx.proto)
7641			continue;
7642
7643		SASTATE_ANY_FOREACH(state) {
7644			SAVLIST_WRITER_FOREACH(sav, sah, state) {
7645				cnt++;
7646			}
7647		}
7648	}
7649
7650	if (cnt == 0) {
7651		*errorp = ENOENT;
7652		return (NULL);
7653	}
7654
7655	/* send this to the userland, one at a time. */
7656	m = NULL;
7657	prev = m;
7658	SAHLIST_WRITER_FOREACH(sah) {
7659		if (req_satype != SADB_SATYPE_UNSPEC &&
7660		    proto != sah->saidx.proto)
7661			continue;
7662
7663		/* map proto to satype */
7664		satype = key_proto2satype(sah->saidx.proto);
7665		if (satype == 0) {
7666			m_freem(m);
7667			*errorp = EINVAL;
7668			return (NULL);
7669		}
7670
7671		SASTATE_ANY_FOREACH(state) {
7672			SAVLIST_WRITER_FOREACH(sav, sah, state) {
7673				n = key_setdumpsa(sav, SADB_DUMP, satype,
7674				    --cnt, pid);
7675				if (!m)
7676					m = n;
7677				else
7678					prev->m_nextpkt = n;
7679				prev = n;
7680			}
7681		}
7682	}
7683
7684	if (!m) {
7685		*errorp = EINVAL;
7686		return (NULL);
7687	}
7688
7689	if ((m->m_flags & M_PKTHDR) != 0) {
7690		m->m_pkthdr.len = 0;
7691		for (n = m; n; n = n->m_next)
7692			m->m_pkthdr.len += n->m_len;
7693	}
7694
7695	*errorp = 0;
7696	return (m);
7697}
7698
7699/*
7700 * SADB_DUMP processing
7701 * dump all entries including status of DEAD in SAD.
7702 * receive
7703 *   <base>
7704 * from the ikmpd, and dump all secasvar leaves
7705 * and send,
7706 *   <base> .....
7707 * to the ikmpd.
7708 *
7709 * m will always be freed.
7710 */
7711static int
7712key_api_dump(struct socket *so, struct mbuf *m0,
7713	 const struct sadb_msghdr *mhp)
7714{
7715	u_int16_t proto;
7716	u_int8_t satype;
7717	struct mbuf *n;
7718	int error, len, ok;
7719
7720	/* map satype to proto */
7721	satype = mhp->msg->sadb_msg_satype;
7722	proto = key_satype2proto(satype);
7723	if (proto == 0) {
7724		IPSECLOG(LOG_DEBUG, "invalid satype is passed.\n");
7725		return key_senderror(so, m0, EINVAL);
7726	}
7727
7728	/*
7729	 * If the requestor has insufficient socket-buffer space
7730	 * for the entire chain, nobody gets any response to the DUMP.
7731	 * XXX For now, only the requestor ever gets anything.
7732	 * Moreover, if the requestor has any space at all, they receive
7733	 * the entire chain, otherwise the request is refused with ENOBUFS.
7734	 */
7735	if (sbspace(&so->so_rcv) <= 0) {
7736		return key_senderror(so, m0, ENOBUFS);
7737	}
7738
7739	mutex_enter(&key_sad.lock);
7740	n = key_setdump_chain(satype, &error, &len, mhp->msg->sadb_msg_pid);
7741	mutex_exit(&key_sad.lock);
7742
7743	if (n == NULL) {
7744		return key_senderror(so, m0, ENOENT);
7745	}
7746	{
7747		uint64_t *ps = PFKEY_STAT_GETREF();
7748		ps[PFKEY_STAT_IN_TOTAL]++;
7749		ps[PFKEY_STAT_IN_BYTES] += len;
7750		PFKEY_STAT_PUTREF();
7751	}
7752
7753	/*
7754	 * PF_KEY DUMP responses are no longer broadcast to all PF_KEY sockets.
7755	 * The requestor receives either the entire chain, or an
7756	 * error message with ENOBUFS.
7757	 *
7758	 * sbappendaddrchain() takes the chain of entries, one
7759	 * packet-record per SPD entry, prepends the key_src sockaddr
7760	 * to each packet-record, links the sockaddr mbufs into a new
7761	 * list of records, then   appends the entire resulting
7762	 * list to the requesting socket.
7763	 */
7764	ok = sbappendaddrchain(&so->so_rcv, (struct sockaddr *)&key_src, n,
7765	    SB_PRIO_ONESHOT_OVERFLOW);
7766
7767	if (!ok) {
7768		PFKEY_STATINC(PFKEY_STAT_IN_NOMEM);
7769		m_freem(n);
7770		return key_senderror(so, m0, ENOBUFS);
7771	}
7772
7773	m_freem(m0);
7774	return 0;
7775}
7776
7777/*
7778 * SADB_X_PROMISC processing
7779 *
7780 * m will always be freed.
7781 */
7782static int
7783key_api_promisc(struct socket *so, struct mbuf *m,
7784	    const struct sadb_msghdr *mhp)
7785{
7786	int olen;
7787
7788	olen = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
7789
7790	if (olen < sizeof(struct sadb_msg)) {
7791#if 1
7792		return key_senderror(so, m, EINVAL);
7793#else
7794		m_freem(m);
7795		return 0;
7796#endif
7797	} else if (olen == sizeof(struct sadb_msg)) {
7798		/* enable/disable promisc mode */
7799		struct keycb *kp = (struct keycb *)sotorawcb(so);
7800		if (kp == NULL)
7801			return key_senderror(so, m, EINVAL);
7802		mhp->msg->sadb_msg_errno = 0;
7803		switch (mhp->msg->sadb_msg_satype) {
7804		case 0:
7805		case 1:
7806			kp->kp_promisc = mhp->msg->sadb_msg_satype;
7807			break;
7808		default:
7809			return key_senderror(so, m, EINVAL);
7810		}
7811
7812		/* send the original message back to everyone */
7813		mhp->msg->sadb_msg_errno = 0;
7814		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7815	} else {
7816		/* send packet as is */
7817
7818		m_adj(m, PFKEY_ALIGN8(sizeof(struct sadb_msg)));
7819
7820		/* TODO: if sadb_msg_seq is specified, send to specific pid */
7821		return key_sendup_mbuf(so, m, KEY_SENDUP_ALL);
7822	}
7823}
7824
7825static int (*key_api_typesw[]) (struct socket *, struct mbuf *,
7826		const struct sadb_msghdr *) = {
7827	NULL,			/* SADB_RESERVED */
7828	key_api_getspi,		/* SADB_GETSPI */
7829	key_api_update,		/* SADB_UPDATE */
7830	key_api_add,		/* SADB_ADD */
7831	key_api_delete,		/* SADB_DELETE */
7832	key_api_get,		/* SADB_GET */
7833	key_api_acquire,	/* SADB_ACQUIRE */
7834	key_api_register,	/* SADB_REGISTER */
7835	NULL,			/* SADB_EXPIRE */
7836	key_api_flush,		/* SADB_FLUSH */
7837	key_api_dump,		/* SADB_DUMP */
7838	key_api_promisc,	/* SADB_X_PROMISC */
7839	NULL,			/* SADB_X_PCHANGE */
7840	key_api_spdadd,		/* SADB_X_SPDUPDATE */
7841	key_api_spdadd,		/* SADB_X_SPDADD */
7842	key_api_spddelete,	/* SADB_X_SPDDELETE */
7843	key_api_spdget,		/* SADB_X_SPDGET */
7844	NULL,			/* SADB_X_SPDACQUIRE */
7845	key_api_spddump,	/* SADB_X_SPDDUMP */
7846	key_api_spdflush,	/* SADB_X_SPDFLUSH */
7847	key_api_spdadd,		/* SADB_X_SPDSETIDX */
7848	NULL,			/* SADB_X_SPDEXPIRE */
7849	key_api_spddelete2,	/* SADB_X_SPDDELETE2 */
7850	key_api_nat_map,	/* SADB_X_NAT_T_NEW_MAPPING */
7851};
7852
7853/*
7854 * parse sadb_msg buffer to process PFKEYv2,
7855 * and create a data to response if needed.
7856 * I think to be dealed with mbuf directly.
7857 * IN:
7858 *     msgp  : pointer to pointer to a received buffer pulluped.
7859 *             This is rewrited to response.
7860 *     so    : pointer to socket.
7861 * OUT:
7862 *    length for buffer to send to user process.
7863 */
7864int
7865key_parse(struct mbuf *m, struct socket *so)
7866{
7867	struct sadb_msg *msg;
7868	struct sadb_msghdr mh;
7869	u_int orglen;
7870	int error;
7871
7872	KASSERT(m != NULL);
7873	KASSERT(so != NULL);
7874
7875#if 0	/*kdebug_sadb assumes msg in linear buffer*/
7876	if (KEYDEBUG_ON(KEYDEBUG_KEY_DUMP)) {
7877		kdebug_sadb("passed sadb_msg", msg);
7878	}
7879#endif
7880
7881	if (m->m_len < sizeof(struct sadb_msg)) {
7882		m = m_pullup(m, sizeof(struct sadb_msg));
7883		if (!m)
7884			return ENOBUFS;
7885	}
7886	msg = mtod(m, struct sadb_msg *);
7887	orglen = PFKEY_UNUNIT64(msg->sadb_msg_len);
7888
7889	if ((m->m_flags & M_PKTHDR) == 0 ||
7890	    m->m_pkthdr.len != orglen) {
7891		IPSECLOG(LOG_DEBUG, "invalid message length.\n");
7892		PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
7893		error = EINVAL;
7894		goto senderror;
7895	}
7896
7897	if (msg->sadb_msg_version != PF_KEY_V2) {
7898		IPSECLOG(LOG_DEBUG, "PF_KEY version %u is mismatched.\n",
7899		    msg->sadb_msg_version);
7900		PFKEY_STATINC(PFKEY_STAT_OUT_INVVER);
7901		error = EINVAL;
7902		goto senderror;
7903	}
7904
7905	if (msg->sadb_msg_type > SADB_MAX) {
7906		IPSECLOG(LOG_DEBUG, "invalid type %u is passed.\n",
7907		    msg->sadb_msg_type);
7908		PFKEY_STATINC(PFKEY_STAT_OUT_INVMSGTYPE);
7909		error = EINVAL;
7910		goto senderror;
7911	}
7912
7913	/* for old-fashioned code - should be nuked */
7914	if (m->m_pkthdr.len > MCLBYTES) {
7915		m_freem(m);
7916		return ENOBUFS;
7917	}
7918	if (m->m_next) {
7919		struct mbuf *n;
7920
7921		n = key_alloc_mbuf_simple(m->m_pkthdr.len, M_WAITOK);
7922
7923		m_copydata(m, 0, m->m_pkthdr.len, mtod(n, void *));
7924		n->m_pkthdr.len = n->m_len = m->m_pkthdr.len;
7925		n->m_next = NULL;
7926		m_freem(m);
7927		m = n;
7928	}
7929
7930	/* align the mbuf chain so that extensions are in contiguous region. */
7931	error = key_align(m, &mh);
7932	if (error)
7933		return error;
7934
7935	if (m->m_next) {	/*XXX*/
7936		m_freem(m);
7937		return ENOBUFS;
7938	}
7939
7940	msg = mh.msg;
7941
7942	/* check SA type */
7943	switch (msg->sadb_msg_satype) {
7944	case SADB_SATYPE_UNSPEC:
7945		switch (msg->sadb_msg_type) {
7946		case SADB_GETSPI:
7947		case SADB_UPDATE:
7948		case SADB_ADD:
7949		case SADB_DELETE:
7950		case SADB_GET:
7951		case SADB_ACQUIRE:
7952		case SADB_EXPIRE:
7953			IPSECLOG(LOG_DEBUG,
7954			    "must specify satype when msg type=%u.\n",
7955			    msg->sadb_msg_type);
7956			PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7957			error = EINVAL;
7958			goto senderror;
7959		}
7960		break;
7961	case SADB_SATYPE_AH:
7962	case SADB_SATYPE_ESP:
7963	case SADB_X_SATYPE_IPCOMP:
7964	case SADB_X_SATYPE_TCPSIGNATURE:
7965		switch (msg->sadb_msg_type) {
7966		case SADB_X_SPDADD:
7967		case SADB_X_SPDDELETE:
7968		case SADB_X_SPDGET:
7969		case SADB_X_SPDDUMP:
7970		case SADB_X_SPDFLUSH:
7971		case SADB_X_SPDSETIDX:
7972		case SADB_X_SPDUPDATE:
7973		case SADB_X_SPDDELETE2:
7974			IPSECLOG(LOG_DEBUG, "illegal satype=%u\n",
7975			    msg->sadb_msg_type);
7976			PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7977			error = EINVAL;
7978			goto senderror;
7979		}
7980		break;
7981	case SADB_SATYPE_RSVP:
7982	case SADB_SATYPE_OSPFV2:
7983	case SADB_SATYPE_RIPV2:
7984	case SADB_SATYPE_MIP:
7985		IPSECLOG(LOG_DEBUG, "type %u isn't supported.\n",
7986		    msg->sadb_msg_satype);
7987		PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7988		error = EOPNOTSUPP;
7989		goto senderror;
7990	case 1:	/* XXX: What does it do? */
7991		if (msg->sadb_msg_type == SADB_X_PROMISC)
7992			break;
7993		/*FALLTHROUGH*/
7994	default:
7995		IPSECLOG(LOG_DEBUG, "invalid type %u is passed.\n",
7996		    msg->sadb_msg_satype);
7997		PFKEY_STATINC(PFKEY_STAT_OUT_INVSATYPE);
7998		error = EINVAL;
7999		goto senderror;
8000	}
8001
8002	/* check field of upper layer protocol and address family */
8003	if (mh.ext[SADB_EXT_ADDRESS_SRC] != NULL &&
8004	    mh.ext[SADB_EXT_ADDRESS_DST] != NULL) {
8005		const struct sadb_address *src0, *dst0;
8006		const struct sockaddr *sa0, *da0;
8007		u_int plen;
8008
8009		src0 = mh.ext[SADB_EXT_ADDRESS_SRC];
8010		dst0 = mh.ext[SADB_EXT_ADDRESS_DST];
8011		sa0 = key_msghdr_get_sockaddr(&mh, SADB_EXT_ADDRESS_SRC);
8012		da0 = key_msghdr_get_sockaddr(&mh, SADB_EXT_ADDRESS_DST);
8013
8014		/* check upper layer protocol */
8015		if (src0->sadb_address_proto != dst0->sadb_address_proto) {
8016			IPSECLOG(LOG_DEBUG,
8017			    "upper layer protocol mismatched src %u, dst %u.\n",
8018			    src0->sadb_address_proto, dst0->sadb_address_proto);
8019
8020			goto invaddr;
8021		}
8022
8023		/* check family */
8024		if (sa0->sa_family != da0->sa_family) {
8025			IPSECLOG(LOG_DEBUG,
8026			    "address family mismatched src %u, dst %u.\n",
8027			    sa0->sa_family, da0->sa_family);
8028			goto invaddr;
8029		}
8030		if (sa0->sa_len != da0->sa_len) {
8031			IPSECLOG(LOG_DEBUG,
8032			    "address size mismatched src %u, dst %u.\n",
8033			    sa0->sa_len, da0->sa_len);
8034			goto invaddr;
8035		}
8036
8037		switch (sa0->sa_family) {
8038		case AF_INET:
8039			if (sa0->sa_len != sizeof(struct sockaddr_in)) {
8040				IPSECLOG(LOG_DEBUG,
8041				    "address size mismatched %u != %zu.\n",
8042				    sa0->sa_len, sizeof(struct sockaddr_in));
8043				goto invaddr;
8044			}
8045			break;
8046		case AF_INET6:
8047			if (sa0->sa_len != sizeof(struct sockaddr_in6)) {
8048				IPSECLOG(LOG_DEBUG,
8049				    "address size mismatched %u != %zu.\n",
8050				    sa0->sa_len, sizeof(struct sockaddr_in6));
8051				goto invaddr;
8052			}
8053			break;
8054		default:
8055			IPSECLOG(LOG_DEBUG, "unsupported address family %u.\n",
8056			    sa0->sa_family);
8057			error = EAFNOSUPPORT;
8058			goto senderror;
8059		}
8060		plen = key_sabits(sa0);
8061
8062		/* check max prefix length */
8063		if (src0->sadb_address_prefixlen > plen ||
8064		    dst0->sadb_address_prefixlen > plen) {
8065			IPSECLOG(LOG_DEBUG, "illegal prefixlen.\n");
8066			goto invaddr;
8067		}
8068
8069		/*
8070		 * prefixlen == 0 is valid because there can be a case when
8071		 * all addresses are matched.
8072		 */
8073	}
8074
8075	if (msg->sadb_msg_type >= __arraycount(key_api_typesw) ||
8076	    key_api_typesw[msg->sadb_msg_type] == NULL) {
8077		PFKEY_STATINC(PFKEY_STAT_OUT_INVMSGTYPE);
8078		error = EINVAL;
8079		goto senderror;
8080	}
8081
8082	return (*key_api_typesw[msg->sadb_msg_type])(so, m, &mh);
8083
8084invaddr:
8085	error = EINVAL;
8086senderror:
8087	PFKEY_STATINC(PFKEY_STAT_OUT_INVADDR);
8088	return key_senderror(so, m, error);
8089}
8090
8091static int
8092key_senderror(struct socket *so, struct mbuf *m, int code)
8093{
8094	struct sadb_msg *msg;
8095
8096	KASSERT(m->m_len >= sizeof(struct sadb_msg));
8097
8098	if (so == NULL) {
8099		/*
8100		 * This means the request comes from kernel.
8101		 * As the request comes from kernel, it is unnecessary to
8102		 * send message to userland. Just return errcode directly.
8103		 */
8104		m_freem(m);
8105		return code;
8106	}
8107
8108	msg = mtod(m, struct sadb_msg *);
8109	msg->sadb_msg_errno = code;
8110	return key_sendup_mbuf(so, m, KEY_SENDUP_ONE);
8111}
8112
8113/*
8114 * set the pointer to each header into message buffer.
8115 * m will be freed on error.
8116 * XXX larger-than-MCLBYTES extension?
8117 */
8118static int
8119key_align(struct mbuf *m, struct sadb_msghdr *mhp)
8120{
8121	struct mbuf *n;
8122	struct sadb_ext *ext;
8123	size_t off, end;
8124	int extlen;
8125	int toff;
8126
8127	KASSERT(m != NULL);
8128	KASSERT(mhp != NULL);
8129	KASSERT(m->m_len >= sizeof(struct sadb_msg));
8130
8131	/* initialize */
8132	memset(mhp, 0, sizeof(*mhp));
8133
8134	mhp->msg = mtod(m, struct sadb_msg *);
8135	mhp->ext[0] = mhp->msg;	/*XXX backward compat */
8136
8137	end = PFKEY_UNUNIT64(mhp->msg->sadb_msg_len);
8138	extlen = end;	/*just in case extlen is not updated*/
8139	for (off = sizeof(struct sadb_msg); off < end; off += extlen) {
8140		n = m_pulldown(m, off, sizeof(struct sadb_ext), &toff);
8141		if (!n) {
8142			/* m is already freed */
8143			return ENOBUFS;
8144		}
8145		ext = (struct sadb_ext *)(mtod(n, char *) + toff);
8146
8147		/* set pointer */
8148		switch (ext->sadb_ext_type) {
8149		case SADB_EXT_SA:
8150		case SADB_EXT_ADDRESS_SRC:
8151		case SADB_EXT_ADDRESS_DST:
8152		case SADB_EXT_ADDRESS_PROXY:
8153		case SADB_EXT_LIFETIME_CURRENT:
8154		case SADB_EXT_LIFETIME_HARD:
8155		case SADB_EXT_LIFETIME_SOFT:
8156		case SADB_EXT_KEY_AUTH:
8157		case SADB_EXT_KEY_ENCRYPT:
8158		case SADB_EXT_IDENTITY_SRC:
8159		case SADB_EXT_IDENTITY_DST:
8160		case SADB_EXT_SENSITIVITY:
8161		case SADB_EXT_PROPOSAL:
8162		case SADB_EXT_SUPPORTED_AUTH:
8163		case SADB_EXT_SUPPORTED_ENCRYPT:
8164		case SADB_EXT_SPIRANGE:
8165		case SADB_X_EXT_POLICY:
8166		case SADB_X_EXT_SA2:
8167		case SADB_X_EXT_NAT_T_TYPE:
8168		case SADB_X_EXT_NAT_T_SPORT:
8169		case SADB_X_EXT_NAT_T_DPORT:
8170		case SADB_X_EXT_NAT_T_OAI:
8171		case SADB_X_EXT_NAT_T_OAR:
8172		case SADB_X_EXT_NAT_T_FRAG:
8173			/* duplicate check */
8174			/*
8175			 * XXX Are there duplication payloads of either
8176			 * KEY_AUTH or KEY_ENCRYPT ?
8177			 */
8178			if (mhp->ext[ext->sadb_ext_type] != NULL) {
8179				IPSECLOG(LOG_DEBUG,
8180				    "duplicate ext_type %u is passed.\n",
8181				    ext->sadb_ext_type);
8182				m_freem(m);
8183				PFKEY_STATINC(PFKEY_STAT_OUT_DUPEXT);
8184				return EINVAL;
8185			}
8186			break;
8187		default:
8188			IPSECLOG(LOG_DEBUG, "invalid ext_type %u is passed.\n",
8189			    ext->sadb_ext_type);
8190			m_freem(m);
8191			PFKEY_STATINC(PFKEY_STAT_OUT_INVEXTTYPE);
8192			return EINVAL;
8193		}
8194
8195		extlen = PFKEY_UNUNIT64(ext->sadb_ext_len);
8196
8197		if (key_validate_ext(ext, extlen)) {
8198			m_freem(m);
8199			PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
8200			return EINVAL;
8201		}
8202
8203		n = m_pulldown(m, off, extlen, &toff);
8204		if (!n) {
8205			/* m is already freed */
8206			return ENOBUFS;
8207		}
8208		ext = (struct sadb_ext *)(mtod(n, char *) + toff);
8209
8210		mhp->ext[ext->sadb_ext_type] = ext;
8211		mhp->extoff[ext->sadb_ext_type] = off;
8212		mhp->extlen[ext->sadb_ext_type] = extlen;
8213	}
8214
8215	if (off != end) {
8216		m_freem(m);
8217		PFKEY_STATINC(PFKEY_STAT_OUT_INVLEN);
8218		return EINVAL;
8219	}
8220
8221	return 0;
8222}
8223
8224static int
8225key_validate_ext(const struct sadb_ext *ext, int len)
8226{
8227	const struct sockaddr *sa;
8228	enum { NONE, ADDR } checktype = NONE;
8229	int baselen = 0;
8230	const int sal = offsetof(struct sockaddr, sa_len) + sizeof(sa->sa_len);
8231
8232	if (len != PFKEY_UNUNIT64(ext->sadb_ext_len))
8233		return EINVAL;
8234
8235	/* if it does not match minimum/maximum length, bail */
8236	if (ext->sadb_ext_type >= __arraycount(minsize) ||
8237	    ext->sadb_ext_type >= __arraycount(maxsize))
8238		return EINVAL;
8239	if (!minsize[ext->sadb_ext_type] || len < minsize[ext->sadb_ext_type])
8240		return EINVAL;
8241	if (maxsize[ext->sadb_ext_type] && len > maxsize[ext->sadb_ext_type])
8242		return EINVAL;
8243
8244	/* more checks based on sadb_ext_type XXX need more */
8245	switch (ext->sadb_ext_type) {
8246	case SADB_EXT_ADDRESS_SRC:
8247	case SADB_EXT_ADDRESS_DST:
8248	case SADB_EXT_ADDRESS_PROXY:
8249		baselen = PFKEY_ALIGN8(sizeof(struct sadb_address));
8250		checktype = ADDR;
8251		break;
8252	case SADB_EXT_IDENTITY_SRC:
8253	case SADB_EXT_IDENTITY_DST:
8254		if (((const struct sadb_ident *)ext)->sadb_ident_type ==
8255		    SADB_X_IDENTTYPE_ADDR) {
8256			baselen = PFKEY_ALIGN8(sizeof(struct sadb_ident));
8257			checktype = ADDR;
8258		} else
8259			checktype = NONE;
8260		break;
8261	default:
8262		checktype = NONE;
8263		break;
8264	}
8265
8266	switch (checktype) {
8267	case NONE:
8268		break;
8269	case ADDR:
8270		sa = (const struct sockaddr *)(((const u_int8_t*)ext)+baselen);
8271		if (len < baselen + sal)
8272			return EINVAL;
8273		if (baselen + PFKEY_ALIGN8(sa->sa_len) != len)
8274			return EINVAL;
8275		break;
8276	}
8277
8278	return 0;
8279}
8280
8281static int
8282key_do_init(void)
8283{
8284	int i, error;
8285
8286	mutex_init(&key_misc.lock, MUTEX_DEFAULT, IPL_NONE);
8287
8288	mutex_init(&key_spd.lock, MUTEX_DEFAULT, IPL_NONE);
8289	cv_init(&key_spd.cv_lc, "key_sp_lc");
8290	key_spd.psz = pserialize_create();
8291	cv_init(&key_spd.cv_psz, "key_sp_psz");
8292	key_spd.psz_performing = false;
8293
8294	mutex_init(&key_sad.lock, MUTEX_DEFAULT, IPL_NONE);
8295	cv_init(&key_sad.cv_lc, "key_sa_lc");
8296	key_sad.psz = pserialize_create();
8297	cv_init(&key_sad.cv_psz, "key_sa_psz");
8298	key_sad.psz_performing = false;
8299
8300	pfkeystat_percpu = percpu_alloc(sizeof(uint64_t) * PFKEY_NSTATS);
8301
8302	callout_init(&key_timehandler_ch, CALLOUT_MPSAFE);
8303	error = workqueue_create(&key_timehandler_wq, "key_timehandler",
8304	    key_timehandler_work, NULL, PRI_SOFTNET, IPL_SOFTNET, WQ_MPSAFE);
8305	if (error != 0)
8306		panic("%s: workqueue_create failed (%d)\n", __func__, error);
8307
8308	for (i = 0; i < IPSEC_DIR_MAX; i++) {
8309		PSLIST_INIT(&key_spd.splist[i]);
8310	}
8311
8312	PSLIST_INIT(&key_spd.socksplist);
8313
8314	key_sad.sahlists = hashinit(SAHHASH_NHASH, HASH_PSLIST, true,
8315	    &key_sad.sahlistmask);
8316	key_sad.savlut = hashinit(SAVLUT_NHASH, HASH_PSLIST, true,
8317	    &key_sad.savlutmask);
8318
8319	for (i = 0; i <= SADB_SATYPE_MAX; i++) {
8320		LIST_INIT(&key_misc.reglist[i]);
8321	}
8322
8323#ifndef IPSEC_NONBLOCK_ACQUIRE
8324	LIST_INIT(&key_misc.acqlist);
8325#endif
8326#ifdef notyet
8327	LIST_INIT(&key_misc.spacqlist);
8328#endif
8329
8330	/* system default */
8331	ip4_def_policy.policy = IPSEC_POLICY_NONE;
8332	ip4_def_policy.state = IPSEC_SPSTATE_ALIVE;
8333	localcount_init(&ip4_def_policy.localcount);
8334
8335#ifdef INET6
8336	ip6_def_policy.policy = IPSEC_POLICY_NONE;
8337	ip6_def_policy.state = IPSEC_SPSTATE_ALIVE;
8338	localcount_init(&ip6_def_policy.localcount);
8339#endif
8340
8341	callout_reset(&key_timehandler_ch, hz, key_timehandler, NULL);
8342
8343	/* initialize key statistics */
8344	keystat.getspi_count = 1;
8345
8346	aprint_verbose("IPsec: Initialized Security Association Processing.\n");
8347
8348	return (0);
8349}
8350
8351void
8352key_init(void)
8353{
8354	static ONCE_DECL(key_init_once);
8355
8356	sysctl_net_keyv2_setup(NULL);
8357	sysctl_net_key_compat_setup(NULL);
8358
8359	RUN_ONCE(&key_init_once, key_do_init);
8360
8361	key_init_so();
8362}
8363
8364/*
8365 * XXX: maybe This function is called after INBOUND IPsec processing.
8366 *
8367 * Special check for tunnel-mode packets.
8368 * We must make some checks for consistency between inner and outer IP header.
8369 *
8370 * xxx more checks to be provided
8371 */
8372int
8373key_checktunnelsanity(
8374    struct secasvar *sav,
8375    u_int family,
8376    void *src,
8377    void *dst
8378)
8379{
8380
8381	/* XXX: check inner IP header */
8382
8383	return 1;
8384}
8385
8386#if 0
8387#define hostnamelen	strlen(hostname)
8388
8389/*
8390 * Get FQDN for the host.
8391 * If the administrator configured hostname (by hostname(1)) without
8392 * domain name, returns nothing.
8393 */
8394static const char *
8395key_getfqdn(void)
8396{
8397	int i;
8398	int hasdot;
8399	static char fqdn[MAXHOSTNAMELEN + 1];
8400
8401	if (!hostnamelen)
8402		return NULL;
8403
8404	/* check if it comes with domain name. */
8405	hasdot = 0;
8406	for (i = 0; i < hostnamelen; i++) {
8407		if (hostname[i] == '.')
8408			hasdot++;
8409	}
8410	if (!hasdot)
8411		return NULL;
8412
8413	/* NOTE: hostname may not be NUL-terminated. */
8414	memset(fqdn, 0, sizeof(fqdn));
8415	memcpy(fqdn, hostname, hostnamelen);
8416	fqdn[hostnamelen] = '\0';
8417	return fqdn;
8418}
8419
8420/*
8421 * get username@FQDN for the host/user.
8422 */
8423static const char *
8424key_getuserfqdn(void)
8425{
8426	const char *host;
8427	static char userfqdn[MAXHOSTNAMELEN + MAXLOGNAME + 2];
8428	struct proc *p = curproc;
8429	char *q;
8430
8431	if (!p || !p->p_pgrp || !p->p_pgrp->pg_session)
8432		return NULL;
8433	if (!(host = key_getfqdn()))
8434		return NULL;
8435
8436	/* NOTE: s_login may not be-NUL terminated. */
8437	memset(userfqdn, 0, sizeof(userfqdn));
8438	memcpy(userfqdn, Mp->p_pgrp->pg_session->s_login, AXLOGNAME);
8439	userfqdn[MAXLOGNAME] = '\0';	/* safeguard */
8440	q = userfqdn + strlen(userfqdn);
8441	*q++ = '@';
8442	memcpy(q, host, strlen(host));
8443	q += strlen(host);
8444	*q++ = '\0';
8445
8446	return userfqdn;
8447}
8448#endif
8449
8450/* record data transfer on SA, and update timestamps */
8451void
8452key_sa_recordxfer(struct secasvar *sav, struct mbuf *m)
8453{
8454	lifetime_counters_t *counters;
8455
8456	KASSERT(sav != NULL);
8457	KASSERT(sav->lft_c != NULL);
8458	KASSERT(m != NULL);
8459
8460	counters = percpu_getref(sav->lft_c_counters_percpu);
8461
8462	/*
8463	 * XXX Currently, there is a difference of bytes size
8464	 * between inbound and outbound processing.
8465	 */
8466	(*counters)[LIFETIME_COUNTER_BYTES] += m->m_pkthdr.len;
8467	/* to check bytes lifetime is done in key_timehandler(). */
8468
8469	/*
8470	 * We use the number of packets as the unit of
8471	 * sadb_lifetime_allocations.  We increment the variable
8472	 * whenever {esp,ah}_{in,out}put is called.
8473	 */
8474	(*counters)[LIFETIME_COUNTER_ALLOCATIONS]++;
8475	/* XXX check for expires? */
8476
8477	percpu_putref(sav->lft_c_counters_percpu);
8478
8479	/*
8480	 * NOTE: We record CURRENT sadb_lifetime_usetime by using wall clock,
8481	 * in seconds.  HARD and SOFT lifetime are measured by the time
8482	 * difference (again in seconds) from sadb_lifetime_usetime.
8483	 *
8484	 *	usetime
8485	 *	v     expire   expire
8486	 * -----+-----+--------+---> t
8487	 *	<--------------> HARD
8488	 *	<-----> SOFT
8489	 */
8490	sav->lft_c->sadb_lifetime_usetime = time_uptime;
8491	/* XXX check for expires? */
8492
8493	return;
8494}
8495
8496/* dumb version */
8497void
8498key_sa_routechange(struct sockaddr *dst)
8499{
8500	struct secashead *sah;
8501	int s;
8502
8503	s = pserialize_read_enter();
8504	SAHLIST_READER_FOREACH(sah) {
8505		struct route *ro;
8506		const struct sockaddr *sa;
8507
8508		key_sah_ref(sah);
8509		pserialize_read_exit(s);
8510
8511		ro = &sah->sa_route;
8512		sa = rtcache_getdst(ro);
8513		if (sa != NULL && dst->sa_len == sa->sa_len &&
8514		    memcmp(dst, sa, dst->sa_len) == 0)
8515			rtcache_free(ro);
8516
8517		s = pserialize_read_enter();
8518		key_sah_unref(sah);
8519	}
8520	pserialize_read_exit(s);
8521
8522	return;
8523}
8524
8525static void
8526key_sa_chgstate(struct secasvar *sav, u_int8_t state)
8527{
8528	struct secasvar *_sav;
8529
8530	ASSERT_SLEEPABLE();
8531	KASSERT(mutex_owned(&key_sad.lock));
8532
8533	if (sav->state == state)
8534		return;
8535
8536	key_unlink_sav(sav);
8537	localcount_fini(&sav->localcount);
8538	SAVLIST_ENTRY_DESTROY(sav);
8539	key_init_sav(sav);
8540
8541	sav->state = state;
8542	if (!SADB_SASTATE_USABLE_P(sav)) {
8543		/* We don't need to care about the order */
8544		SAVLIST_WRITER_INSERT_HEAD(sav->sah, state, sav);
8545		return;
8546	}
8547	/*
8548	 * Sort the list by lft_c->sadb_lifetime_addtime
8549	 * in ascending order.
8550	 */
8551	SAVLIST_WRITER_FOREACH(_sav, sav->sah, state) {
8552		if (_sav->lft_c->sadb_lifetime_addtime >
8553		    sav->lft_c->sadb_lifetime_addtime) {
8554			SAVLIST_WRITER_INSERT_BEFORE(_sav, sav);
8555			break;
8556		}
8557	}
8558	if (_sav == NULL) {
8559		SAVLIST_WRITER_INSERT_TAIL(sav->sah, state, sav);
8560	}
8561
8562	SAVLUT_WRITER_INSERT_HEAD(sav);
8563
8564	key_validate_savlist(sav->sah, state);
8565}
8566
8567/* XXX too much? */
8568static struct mbuf *
8569key_alloc_mbuf(int l, int mflag)
8570{
8571	struct mbuf *m = NULL, *n;
8572	int len, t;
8573
8574	KASSERT(mflag == M_NOWAIT || (mflag == M_WAITOK && !cpu_softintr_p()));
8575
8576	len = l;
8577	while (len > 0) {
8578		MGET(n, mflag, MT_DATA);
8579		if (n && len > MLEN) {
8580			MCLGET(n, mflag);
8581			if ((n->m_flags & M_EXT) == 0) {
8582				m_freem(n);
8583				n = NULL;
8584			}
8585		}
8586		if (!n) {
8587			m_freem(m);
8588			return NULL;
8589		}
8590
8591		n->m_next = NULL;
8592		n->m_len = 0;
8593		n->m_len = M_TRAILINGSPACE(n);
8594		/* use the bottom of mbuf, hoping we can prepend afterwards */
8595		if (n->m_len > len) {
8596			t = (n->m_len - len) & ~(sizeof(long) - 1);
8597			n->m_data += t;
8598			n->m_len = len;
8599		}
8600
8601		len -= n->m_len;
8602
8603		if (m)
8604			m_cat(m, n);
8605		else
8606			m = n;
8607	}
8608
8609	return m;
8610}
8611
8612static struct mbuf *
8613key_setdump(u_int8_t req_satype, int *errorp, uint32_t pid)
8614{
8615	struct secashead *sah;
8616	struct secasvar *sav;
8617	u_int16_t proto;
8618	u_int8_t satype;
8619	u_int8_t state;
8620	int cnt;
8621	struct mbuf *m, *n;
8622
8623	KASSERT(mutex_owned(&key_sad.lock));
8624
8625	/* map satype to proto */
8626	proto = key_satype2proto(req_satype);
8627	if (proto == 0) {
8628		*errorp = EINVAL;
8629		return (NULL);
8630	}
8631
8632	/* count sav entries to be sent to the userland. */
8633	cnt = 0;
8634	SAHLIST_WRITER_FOREACH(sah) {
8635		if (req_satype != SADB_SATYPE_UNSPEC &&
8636		    proto != sah->saidx.proto)
8637			continue;
8638
8639		SASTATE_ANY_FOREACH(state) {
8640			SAVLIST_WRITER_FOREACH(sav, sah, state) {
8641				cnt++;
8642			}
8643		}
8644	}
8645
8646	if (cnt == 0) {
8647		*errorp = ENOENT;
8648		return (NULL);
8649	}
8650
8651	/* send this to the userland, one at a time. */
8652	m = NULL;
8653	SAHLIST_WRITER_FOREACH(sah) {
8654		if (req_satype != SADB_SATYPE_UNSPEC &&
8655		    proto != sah->saidx.proto)
8656			continue;
8657
8658		/* map proto to satype */
8659		satype = key_proto2satype(sah->saidx.proto);
8660		if (satype == 0) {
8661			m_freem(m);
8662			*errorp = EINVAL;
8663			return (NULL);
8664		}
8665
8666		SASTATE_ANY_FOREACH(state) {
8667			SAVLIST_WRITER_FOREACH(sav, sah, state) {
8668				n = key_setdumpsa(sav, SADB_DUMP, satype,
8669				    --cnt, pid);
8670				if (!m)
8671					m = n;
8672				else
8673					m_cat(m, n);
8674			}
8675		}
8676	}
8677
8678	if (!m) {
8679		*errorp = EINVAL;
8680		return (NULL);
8681	}
8682
8683	if ((m->m_flags & M_PKTHDR) != 0) {
8684		m->m_pkthdr.len = 0;
8685		for (n = m; n; n = n->m_next)
8686			m->m_pkthdr.len += n->m_len;
8687	}
8688
8689	*errorp = 0;
8690	return (m);
8691}
8692
8693static struct mbuf *
8694key_setspddump(int *errorp, pid_t pid)
8695{
8696	struct secpolicy *sp;
8697	int cnt;
8698	u_int dir;
8699	struct mbuf *m, *n;
8700
8701	KASSERT(mutex_owned(&key_spd.lock));
8702
8703	/* search SPD entry and get buffer size. */
8704	cnt = 0;
8705	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
8706		SPLIST_WRITER_FOREACH(sp, dir) {
8707			cnt++;
8708		}
8709	}
8710
8711	if (cnt == 0) {
8712		*errorp = ENOENT;
8713		return (NULL);
8714	}
8715
8716	m = NULL;
8717	for (dir = 0; dir < IPSEC_DIR_MAX; dir++) {
8718		SPLIST_WRITER_FOREACH(sp, dir) {
8719			--cnt;
8720			n = key_setdumpsp(sp, SADB_X_SPDDUMP, cnt, pid);
8721
8722			if (!m)
8723				m = n;
8724			else {
8725				m->m_pkthdr.len += n->m_pkthdr.len;
8726				m_cat(m, n);
8727			}
8728		}
8729	}
8730
8731	*errorp = 0;
8732	return (m);
8733}
8734
8735int
8736key_get_used(void) {
8737	return !SPLIST_READER_EMPTY(IPSEC_DIR_INBOUND) ||
8738	    !SPLIST_READER_EMPTY(IPSEC_DIR_OUTBOUND) ||
8739	    !SOCKSPLIST_READER_EMPTY();
8740}
8741
8742void
8743key_update_used(void)
8744{
8745	switch (ipsec_enabled) {
8746	default:
8747	case 0:
8748#ifdef notyet
8749		/* XXX: racy */
8750		ipsec_used = 0;
8751#endif
8752		break;
8753	case 1:
8754#ifndef notyet
8755		/* XXX: racy */
8756		if (!ipsec_used)
8757#endif
8758		ipsec_used = key_get_used();
8759		break;
8760	case 2:
8761		ipsec_used = 1;
8762		break;
8763	}
8764}
8765
8766static inline void
8767key_savlut_writer_insert_head(struct secasvar *sav)
8768{
8769	uint32_t hash_key;
8770	uint32_t hash;
8771
8772	KASSERT(mutex_owned(&key_sad.lock));
8773	KASSERT(!sav->savlut_added);
8774
8775	hash_key = sav->spi;
8776
8777	hash = key_savluthash(&sav->sah->saidx.dst.sa,
8778	    sav->sah->saidx.proto, hash_key, key_sad.savlutmask);
8779
8780	PSLIST_WRITER_INSERT_HEAD(&key_sad.savlut[hash], sav,
8781	    pslist_entry_savlut);
8782	sav->savlut_added = true;
8783}
8784
8785/*
8786 * Calculate hash using protocol, source address,
8787 * and destination address included in saidx.
8788 */
8789static inline uint32_t
8790key_saidxhash(const struct secasindex *saidx, u_long mask)
8791{
8792	uint32_t hash32;
8793	const struct sockaddr_in *sin;
8794	const struct sockaddr_in6 *sin6;
8795
8796	hash32 = saidx->proto;
8797
8798	switch (saidx->src.sa.sa_family) {
8799	case AF_INET:
8800		sin = &saidx->src.sin;
8801		hash32 = hash32_buf(&sin->sin_addr,
8802		    sizeof(sin->sin_addr), hash32);
8803		sin = &saidx->dst.sin;
8804		hash32 = hash32_buf(&sin->sin_addr,
8805		    sizeof(sin->sin_addr), hash32 << 1);
8806		break;
8807	case AF_INET6:
8808		sin6 = &saidx->src.sin6;
8809		hash32 = hash32_buf(&sin6->sin6_addr,
8810		    sizeof(sin6->sin6_addr), hash32);
8811		sin6 = &saidx->dst.sin6;
8812		hash32 = hash32_buf(&sin6->sin6_addr,
8813		    sizeof(sin6->sin6_addr), hash32 << 1);
8814		break;
8815	default:
8816		hash32 = 0;
8817		break;
8818	}
8819
8820	return hash32 & mask;
8821}
8822
8823/*
8824 * Calculate hash using destination address, protocol,
8825 * and spi. Those parameter depend on the search of
8826 * key_lookup_sa().
8827 */
8828static uint32_t
8829key_savluthash(const struct sockaddr *dst, uint32_t proto,
8830    uint32_t spi, u_long mask)
8831{
8832	uint32_t hash32;
8833	const struct sockaddr_in *sin;
8834	const struct sockaddr_in6 *sin6;
8835
8836	hash32 = hash32_buf(&proto, sizeof(proto), spi);
8837
8838	switch(dst->sa_family) {
8839	case AF_INET:
8840		sin = satocsin(dst);
8841		hash32 = hash32_buf(&sin->sin_addr,
8842		    sizeof(sin->sin_addr), hash32);
8843		break;
8844	case AF_INET6:
8845		sin6 = satocsin6(dst);
8846		hash32 = hash32_buf(&sin6->sin6_addr,
8847		    sizeof(sin6->sin6_addr), hash32);
8848		break;
8849	default:
8850		hash32 = 0;
8851	}
8852
8853	return hash32 & mask;
8854}
8855
8856static int
8857sysctl_net_key_dumpsa(SYSCTLFN_ARGS)
8858{
8859	struct mbuf *m, *n;
8860	int err2 = 0;
8861	char *p, *ep;
8862	size_t len;
8863	int error;
8864
8865	if (newp)
8866		return (EPERM);
8867	if (namelen != 1)
8868		return (EINVAL);
8869
8870	mutex_enter(&key_sad.lock);
8871	m = key_setdump(name[0], &error, l->l_proc->p_pid);
8872	mutex_exit(&key_sad.lock);
8873	if (!m)
8874		return (error);
8875	if (!oldp)
8876		*oldlenp = m->m_pkthdr.len;
8877	else {
8878		p = oldp;
8879		if (*oldlenp < m->m_pkthdr.len) {
8880			err2 = ENOMEM;
8881			ep = p + *oldlenp;
8882		} else {
8883			*oldlenp = m->m_pkthdr.len;
8884			ep = p + m->m_pkthdr.len;
8885		}
8886		for (n = m; n; n = n->m_next) {
8887			len =  (ep - p < n->m_len) ?
8888				ep - p : n->m_len;
8889			error = copyout(mtod(n, const void *), p, len);
8890			p += len;
8891			if (error)
8892				break;
8893		}
8894		if (error == 0)
8895			error = err2;
8896	}
8897	m_freem(m);
8898
8899	return (error);
8900}
8901
8902static int
8903sysctl_net_key_dumpsp(SYSCTLFN_ARGS)
8904{
8905	struct mbuf *m, *n;
8906	int err2 = 0;
8907	char *p, *ep;
8908	size_t len;
8909	int error;
8910
8911	if (newp)
8912		return (EPERM);
8913	if (namelen != 0)
8914		return (EINVAL);
8915
8916	mutex_enter(&key_spd.lock);
8917	m = key_setspddump(&error, l->l_proc->p_pid);
8918	mutex_exit(&key_spd.lock);
8919	if (!m)
8920		return (error);
8921	if (!oldp)
8922		*oldlenp = m->m_pkthdr.len;
8923	else {
8924		p = oldp;
8925		if (*oldlenp < m->m_pkthdr.len) {
8926			err2 = ENOMEM;
8927			ep = p + *oldlenp;
8928		} else {
8929			*oldlenp = m->m_pkthdr.len;
8930			ep = p + m->m_pkthdr.len;
8931		}
8932		for (n = m; n; n = n->m_next) {
8933			len = (ep - p < n->m_len) ? ep - p : n->m_len;
8934			error = copyout(mtod(n, const void *), p, len);
8935			p += len;
8936			if (error)
8937				break;
8938		}
8939		if (error == 0)
8940			error = err2;
8941	}
8942	m_freem(m);
8943
8944	return (error);
8945}
8946
8947/*
8948 * Create sysctl tree for native IPSEC key knobs, originally
8949 * under name "net.keyv2"  * with MIB number { CTL_NET, PF_KEY_V2. }.
8950 * However, sysctl(8) never checked for nodes under { CTL_NET, PF_KEY_V2 };
8951 * and in any case the part of our sysctl namespace used for dumping the
8952 * SPD and SA database  *HAS* to be compatible with the KAME sysctl
8953 * namespace, for API reasons.
8954 *
8955 * Pending a consensus on the right way  to fix this, add a level of
8956 * indirection in how we number the `native' IPSEC key nodes;
8957 * and (as requested by Andrew Brown)  move registration of the
8958 * KAME-compatible names  to a separate function.
8959 */
8960#if 0
8961#  define IPSEC_PFKEY PF_KEY_V2
8962# define IPSEC_PFKEY_NAME "keyv2"
8963#else
8964#  define IPSEC_PFKEY PF_KEY
8965# define IPSEC_PFKEY_NAME "key"
8966#endif
8967
8968static int
8969sysctl_net_key_stats(SYSCTLFN_ARGS)
8970{
8971
8972	return (NETSTAT_SYSCTL(pfkeystat_percpu, PFKEY_NSTATS));
8973}
8974
8975static void
8976sysctl_net_keyv2_setup(struct sysctllog **clog)
8977{
8978
8979	sysctl_createv(clog, 0, NULL, NULL,
8980		       CTLFLAG_PERMANENT,
8981		       CTLTYPE_NODE, IPSEC_PFKEY_NAME, NULL,
8982		       NULL, 0, NULL, 0,
8983		       CTL_NET, IPSEC_PFKEY, CTL_EOL);
8984
8985	sysctl_createv(clog, 0, NULL, NULL,
8986		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8987		       CTLTYPE_INT, "debug", NULL,
8988		       NULL, 0, &key_debug_level, 0,
8989		       CTL_NET, IPSEC_PFKEY, KEYCTL_DEBUG_LEVEL, CTL_EOL);
8990	sysctl_createv(clog, 0, NULL, NULL,
8991		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8992		       CTLTYPE_INT, "spi_try", NULL,
8993		       NULL, 0, &key_spi_trycnt, 0,
8994		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_TRY, CTL_EOL);
8995	sysctl_createv(clog, 0, NULL, NULL,
8996		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
8997		       CTLTYPE_INT, "spi_min_value", NULL,
8998		       NULL, 0, &key_spi_minval, 0,
8999		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_MIN_VALUE, CTL_EOL);
9000	sysctl_createv(clog, 0, NULL, NULL,
9001		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9002		       CTLTYPE_INT, "spi_max_value", NULL,
9003		       NULL, 0, &key_spi_maxval, 0,
9004		       CTL_NET, IPSEC_PFKEY, KEYCTL_SPI_MAX_VALUE, CTL_EOL);
9005	sysctl_createv(clog, 0, NULL, NULL,
9006		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9007		       CTLTYPE_INT, "random_int", NULL,
9008		       NULL, 0, &key_int_random, 0,
9009		       CTL_NET, IPSEC_PFKEY, KEYCTL_RANDOM_INT, CTL_EOL);
9010	sysctl_createv(clog, 0, NULL, NULL,
9011		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9012		       CTLTYPE_INT, "larval_lifetime", NULL,
9013		       NULL, 0, &key_larval_lifetime, 0,
9014		       CTL_NET, IPSEC_PFKEY, KEYCTL_LARVAL_LIFETIME, CTL_EOL);
9015	sysctl_createv(clog, 0, NULL, NULL,
9016		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9017		       CTLTYPE_INT, "blockacq_count", NULL,
9018		       NULL, 0, &key_blockacq_count, 0,
9019		       CTL_NET, IPSEC_PFKEY, KEYCTL_BLOCKACQ_COUNT, CTL_EOL);
9020	sysctl_createv(clog, 0, NULL, NULL,
9021		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9022		       CTLTYPE_INT, "blockacq_lifetime", NULL,
9023		       NULL, 0, &key_blockacq_lifetime, 0,
9024		       CTL_NET, IPSEC_PFKEY, KEYCTL_BLOCKACQ_LIFETIME, CTL_EOL);
9025	sysctl_createv(clog, 0, NULL, NULL,
9026		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9027		       CTLTYPE_INT, "esp_keymin", NULL,
9028		       NULL, 0, &ipsec_esp_keymin, 0,
9029		       CTL_NET, IPSEC_PFKEY, KEYCTL_ESP_KEYMIN, CTL_EOL);
9030	sysctl_createv(clog, 0, NULL, NULL,
9031		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9032		       CTLTYPE_INT, "prefered_oldsa", NULL,
9033		       NULL, 0, &key_prefered_oldsa, 0,
9034		       CTL_NET, PF_KEY, KEYCTL_PREFERED_OLDSA, CTL_EOL);
9035	sysctl_createv(clog, 0, NULL, NULL,
9036		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9037		       CTLTYPE_INT, "esp_auth", NULL,
9038		       NULL, 0, &ipsec_esp_auth, 0,
9039		       CTL_NET, IPSEC_PFKEY, KEYCTL_ESP_AUTH, CTL_EOL);
9040	sysctl_createv(clog, 0, NULL, NULL,
9041		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9042		       CTLTYPE_INT, "ah_keymin", NULL,
9043		       NULL, 0, &ipsec_ah_keymin, 0,
9044		       CTL_NET, IPSEC_PFKEY, KEYCTL_AH_KEYMIN, CTL_EOL);
9045	sysctl_createv(clog, 0, NULL, NULL,
9046		       CTLFLAG_PERMANENT,
9047		       CTLTYPE_STRUCT, "stats",
9048		       SYSCTL_DESCR("PF_KEY statistics"),
9049		       sysctl_net_key_stats, 0, NULL, 0,
9050		       CTL_NET, IPSEC_PFKEY, CTL_CREATE, CTL_EOL);
9051	sysctl_createv(clog, 0, NULL, NULL,
9052		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
9053		       CTLTYPE_BOOL, "allow_different_idtype", NULL,
9054		       NULL, 0, &ipsec_allow_different_idtype, 0,
9055		       CTL_NET, IPSEC_PFKEY, KEYCTL_ALLOW_DIFFERENT_IDTYPE, CTL_EOL);
9056}
9057
9058/*
9059 * Register sysctl names used by setkey(8). For historical reasons,
9060 * and to share a single API, these names appear under { CTL_NET, PF_KEY }
9061 * for both IPSEC and KAME IPSEC.
9062 */
9063static void
9064sysctl_net_key_compat_setup(struct sysctllog **clog)
9065{
9066
9067	sysctl_createv(clog, 0, NULL, NULL,
9068		       CTLFLAG_PERMANENT,
9069		       CTLTYPE_NODE, "key", NULL,
9070		       NULL, 0, NULL, 0,
9071		       CTL_NET, PF_KEY, CTL_EOL);
9072
9073	/* Register the net.key.dump{sa,sp} nodes used by setkey(8). */
9074	sysctl_createv(clog, 0, NULL, NULL,
9075		       CTLFLAG_PERMANENT,
9076		       CTLTYPE_STRUCT, "dumpsa", NULL,
9077		       sysctl_net_key_dumpsa, 0, NULL, 0,
9078		       CTL_NET, PF_KEY, KEYCTL_DUMPSA, CTL_EOL);
9079	sysctl_createv(clog, 0, NULL, NULL,
9080		       CTLFLAG_PERMANENT,
9081		       CTLTYPE_STRUCT, "dumpsp", NULL,
9082		       sysctl_net_key_dumpsp, 0, NULL, 0,
9083		       CTL_NET, PF_KEY, KEYCTL_DUMPSP, CTL_EOL);
9084}
9085