xref: /titanic_50/usr/src/uts/common/inet/ipsec_impl.h (revision 672986541be54a7a471bb088e60780c37e371d7e)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License (the "License").
6  * You may not use this file except in compliance with the License.
7  *
8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9  * or http://www.opensolaris.org/os/licensing.
10  * See the License for the specific language governing permissions
11  * and limitations under the License.
12  *
13  * When distributing Covered Code, include this CDDL HEADER in each
14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15  * If applicable, add the following below this CDDL HEADER, with the
16  * fields enclosed by brackets "[]" replaced with your own identifying
17  * information: Portions Copyright [yyyy] [name of copyright owner]
18  *
19  * CDDL HEADER END
20  */
21 /*
22  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #ifndef _INET_IPSEC_IMPL_H
27 #define	_INET_IPSEC_IMPL_H
28 
29 #pragma ident	"%Z%%M%	%I%	%E% SMI"
30 
31 #include <inet/ip.h>
32 #include <inet/ipdrop.h>
33 
34 #ifdef	__cplusplus
35 extern "C" {
36 #endif
37 
38 #define	IPSEC_CONF_SRC_ADDRESS	0	/* Source Address */
39 #define	IPSEC_CONF_SRC_PORT		1	/* Source Port */
40 #define	IPSEC_CONF_DST_ADDRESS	2	/* Dest Address */
41 #define	IPSEC_CONF_DST_PORT		3	/* Dest Port */
42 #define	IPSEC_CONF_SRC_MASK		4	/* Source Address Mask */
43 #define	IPSEC_CONF_DST_MASK		5	/* Destination Address Mask */
44 #define	IPSEC_CONF_ULP			6	/* Upper layer Port */
45 #define	IPSEC_CONF_IPSEC_PROT	7	/* AH or ESP or AH_ESP */
46 #define	IPSEC_CONF_IPSEC_AALGS	8	/* Auth Algorithms - MD5 etc. */
47 #define	IPSEC_CONF_IPSEC_EALGS	9	/* Encr Algorithms - DES etc. */
48 #define	IPSEC_CONF_IPSEC_EAALGS	10	/* Encr Algorithms - MD5 etc. */
49 #define	IPSEC_CONF_IPSEC_SA		11	/* Shared or unique SA */
50 #define	IPSEC_CONF_IPSEC_DIR 		12	/* Direction of traffic */
51 #define	IPSEC_CONF_ICMP_TYPE 		13	/* ICMP type */
52 #define	IPSEC_CONF_ICMP_CODE 		14	/* ICMP code */
53 #define	IPSEC_CONF_NEGOTIATE		15	/* Negotiation */
54 #define	IPSEC_CONF_TUNNEL		16	/* Tunnel */
55 
56 /* Type of an entry */
57 
58 #define	IPSEC_NTYPES			0x02
59 #define	IPSEC_TYPE_OUTBOUND		0x00
60 #define	IPSEC_TYPE_INBOUND		0x01
61 
62 /* Policy */
63 #define	IPSEC_POLICY_APPLY	0x01
64 #define	IPSEC_POLICY_DISCARD	0x02
65 #define	IPSEC_POLICY_BYPASS	0x03
66 
67 /* Shared or unique SA */
68 #define	IPSEC_SHARED_SA		0x01
69 #define	IPSEC_UNIQUE_SA		0x02
70 
71 /* IPSEC protocols and combinations */
72 #define	IPSEC_AH_ONLY		0x01
73 #define	IPSEC_ESP_ONLY		0x02
74 #define	IPSEC_AH_ESP		0x03
75 
76 #ifdef _KERNEL
77 
78 #include <inet/common.h>
79 #include <netinet/ip6.h>
80 #include <netinet/icmp6.h>
81 #include <net/pfkeyv2.h>
82 #include <inet/ip.h>
83 #include <inet/sadb.h>
84 #include <inet/ipsecah.h>
85 #include <inet/ipsecesp.h>
86 #include <sys/crypto/common.h>
87 #include <sys/crypto/api.h>
88 #include <sys/avl.h>
89 
90 /*
91  * Maximum number of authentication algorithms (can be indexed by one byte
92  * per PF_KEY and the IKE IPsec DOI.
93  */
94 #define	MAX_AALGS 256
95 
96 /*
97  * IPsec task queue constants.
98  */
99 #define	IPSEC_TASKQ_MIN 10
100 #define	IPSEC_TASKQ_MAX 20
101 
102 /*
103  * So we can access IPsec global variables that live in keysock.c.
104  */
105 extern boolean_t keysock_extended_reg(netstack_t *);
106 extern uint32_t keysock_next_seq(netstack_t *);
107 
108 /*
109  * Locking for ipsec policy rules:
110  *
111  * policy heads: system policy is static; per-conn polheads are dynamic,
112  * and refcounted (and inherited); use atomic refcounts and "don't let
113  * go with both hands".
114  *
115  * policy: refcounted; references from polhead, ipsec_out
116  *
117  * actions: refcounted; referenced from: action hash table, policy, ipsec_out
118  * selectors: refcounted; referenced from: selector hash table, policy.
119  */
120 
121 /*
122  * the following are inspired by, but not directly based on,
123  * some of the sys/queue.h type-safe pseudo-polymorphic macros
124  * found in BSD.
125  *
126  * XXX If we use these more generally, we'll have to make the names
127  * less generic (HASH_* will probably clobber other namespaces).
128  */
129 
130 #define	HASH_LOCK(table, hash) \
131 	mutex_enter(&(table)[hash].hash_lock)
132 #define	HASH_UNLOCK(table, hash) \
133 	mutex_exit(&(table)[hash].hash_lock)
134 
135 #define	HASH_LOCKED(table, hash) \
136 	MUTEX_HELD(&(table)[hash].hash_lock)
137 
138 #define	HASH_ITERATE(var, field, table, hash) 		\
139 	var = table[hash].hash_head; var != NULL; var = var->field.hash_next
140 
141 #define	HASH_NEXT(var, field) 		\
142 	(var)->field.hash_next
143 
144 #define	HASH_INSERT(var, field, table, hash)			\
145 {								\
146 	ASSERT(HASH_LOCKED(table, hash));			\
147 	(var)->field.hash_next = (table)[hash].hash_head;	\
148 	(var)->field.hash_pp = &(table)[hash].hash_head;	\
149 	(table)[hash].hash_head = var;				\
150 	if ((var)->field.hash_next != NULL)			\
151 		(var)->field.hash_next->field.hash_pp = 	\
152 			&((var)->field.hash_next); 		\
153 }
154 
155 
156 #define	HASH_UNCHAIN(var, field, table, hash)			\
157 {								\
158 	ASSERT(MUTEX_HELD(&(table)[hash].hash_lock));		\
159 	HASHLIST_UNCHAIN(var, field);				\
160 }
161 
162 #define	HASHLIST_INSERT(var, field, head)			\
163 {								\
164 	(var)->field.hash_next = head;				\
165 	(var)->field.hash_pp = &(head);				\
166 	head = var;						\
167 	if ((var)->field.hash_next != NULL)			\
168 		(var)->field.hash_next->field.hash_pp = 	\
169 			&((var)->field.hash_next); 		\
170 }
171 
172 #define	HASHLIST_UNCHAIN(var, field) 				\
173 {								\
174 	*var->field.hash_pp = var->field.hash_next;		\
175 	if (var->field.hash_next)				\
176 		var->field.hash_next->field.hash_pp = 		\
177 			var->field.hash_pp;			\
178 	HASH_NULL(var, field);					\
179 }
180 
181 
182 #define	HASH_NULL(var, field) 					\
183 {								\
184 	var->field.hash_next = NULL;				\
185 	var->field.hash_pp = NULL;				\
186 }
187 
188 #define	HASH_LINK(fieldname, type)				\
189 	struct {						\
190 		type *hash_next;				\
191 		type **hash_pp;					\
192 	} fieldname
193 
194 
195 #define	HASH_HEAD(tag)						\
196 	struct {						\
197 		struct tag *hash_head;				\
198 		kmutex_t hash_lock;				\
199 	}
200 
201 
202 typedef struct ipsec_policy_s ipsec_policy_t;
203 
204 typedef HASH_HEAD(ipsec_policy_s) ipsec_policy_hash_t;
205 
206 /*
207  * When adding new fields to ipsec_prot_t, make sure to update
208  * ipsec_in_to_out_action() as well as other code in spd.c
209  */
210 
211 typedef struct ipsec_prot
212 {
213 	unsigned int
214 		ipp_use_ah : 1,
215 		ipp_use_esp : 1,
216 		ipp_use_se : 1,
217 		ipp_use_unique : 1,
218 		ipp_use_espa : 1,
219 		ipp_pad : 27;
220 	uint8_t		ipp_auth_alg;		 /* DOI number */
221 	uint8_t		ipp_encr_alg;		 /* DOI number */
222 	uint8_t		ipp_esp_auth_alg;	 /* DOI number */
223 	uint16_t 	ipp_ah_minbits;		 /* AH: min keylen */
224 	uint16_t 	ipp_ah_maxbits;		 /* AH: max keylen */
225 	uint16_t	ipp_espe_minbits;	 /* ESP encr: min keylen */
226 	uint16_t	ipp_espe_maxbits;	 /* ESP encr: max keylen */
227 	uint16_t	ipp_espa_minbits;	 /* ESP auth: min keylen */
228 	uint16_t	ipp_espa_maxbits;	 /* ESP auth: max keylen */
229 	uint32_t	ipp_km_proto;		 /* key mgmt protocol */
230 	uint32_t	ipp_km_cookie;		 /* key mgmt cookie */
231 	uint32_t	ipp_replay_depth;	 /* replay window */
232 	/* XXX add lifetimes */
233 } ipsec_prot_t;
234 
235 #define	IPSEC_MAX_KEYBITS (0xffff)
236 
237 /*
238  * An individual policy action, possibly a member of a chain.
239  *
240  * Action chains may be shared between multiple policy rules.
241  *
242  * With one exception (IPSEC_POLICY_LOG), a chain consists of an
243  * ordered list of alternative ways to handle a packet.
244  *
245  * All actions are also "interned" into a hash table (to allow
246  * multiple rules with the same action chain to share one copy in
247  * memory).
248  */
249 
250 typedef struct ipsec_act
251 {
252 	uint8_t		ipa_type;
253 	uint8_t		ipa_log;
254 	union
255 	{
256 		ipsec_prot_t	ipau_apply;
257 		uint8_t		ipau_reject_type;
258 		uint32_t	ipau_resolve_id; /* magic cookie */
259 		uint8_t		ipau_log_type;
260 	} ipa_u;
261 #define	ipa_apply ipa_u.ipau_apply
262 #define	ipa_reject_type ipa_u.ipau_reject_type
263 #define	ipa_log_type ipa_u.ipau_log_type
264 #define	ipa_resolve_type ipa_u.ipau_resolve_type
265 } ipsec_act_t;
266 
267 #define	IPSEC_ACT_APPLY		0x01 /* match IPSEC_POLICY_APPLY */
268 #define	IPSEC_ACT_DISCARD	0x02 /* match IPSEC_POLICY_DISCARD */
269 #define	IPSEC_ACT_BYPASS	0x03 /* match IPSEC_POLICY_BYPASS */
270 #define	IPSEC_ACT_REJECT	0x04
271 #define	IPSEC_ACT_CLEAR		0x05
272 
273 typedef struct ipsec_action_s
274 {
275 	HASH_LINK(ipa_hash, struct ipsec_action_s);
276 	struct ipsec_action_s	*ipa_next;	/* next alternative */
277 	uint32_t		ipa_refs;		/* refcount */
278 	ipsec_act_t		ipa_act;
279 	/*
280 	 * The following bits are equivalent to an OR of bits included in the
281 	 * ipau_apply fields of this and subsequent actions in an
282 	 * action chain; this is an optimization for the sake of
283 	 * ipsec_out_process() in ip.c and a few other places.
284 	 */
285 	unsigned int
286 		ipa_hval: 8,
287 		ipa_allow_clear:1,		/* rule allows cleartext? */
288 		ipa_want_ah:1,			/* an action wants ah */
289 		ipa_want_esp:1,			/* an action wants esp */
290 		ipa_want_se:1,			/* an action wants se */
291 		ipa_want_unique:1,		/* want unique sa's */
292 		ipa_pad:19;
293 	uint32_t		ipa_ovhd;	/* per-packet encap ovhd */
294 } ipsec_action_t;
295 
296 #define	IPACT_REFHOLD(ipa) {			\
297 	atomic_add_32(&(ipa)->ipa_refs, 1);	\
298 	ASSERT((ipa)->ipa_refs != 0);	\
299 }
300 #define	IPACT_REFRELE(ipa) {					\
301 	ASSERT((ipa)->ipa_refs != 0);				\
302 	membar_exit();						\
303 	if (atomic_add_32_nv(&(ipa)->ipa_refs, -1) == 0)	\
304 		ipsec_action_free(ipa);				\
305 	(ipa) = 0;						\
306 }
307 
308 /*
309  * For now, use a trivially sized hash table for actions.
310  * In the future we can add the structure canonicalization necessary
311  * to get the hash function to behave correctly..
312  */
313 #define	IPSEC_ACTION_HASH_SIZE 1
314 
315 /*
316  * Merged address structure, for cheezy address-family independent
317  * matches in policy code.
318  */
319 
320 typedef union ipsec_addr
321 {
322 	in6_addr_t	ipsad_v6;
323 	in_addr_t	ipsad_v4;
324 } ipsec_addr_t;
325 
326 /*
327  * ipsec selector set, as used by the kernel policy structures.
328  * Note that that we specify "local" and "remote"
329  * rather than "source" and "destination", which allows the selectors
330  * for symmetric policy rules to be shared between inbound and
331  * outbound rules.
332  *
333  * "local" means "destination" on inbound, and "source" on outbound.
334  * "remote" means "source" on inbound, and "destination" on outbound.
335  * XXX if we add a fifth policy enforcement point for forwarded packets,
336  * what do we do?
337  *
338  * The ipsl_valid mask is not done as a bitfield; this is so we
339  * can use "ffs()" to find the "most interesting" valid tag.
340  *
341  * XXX should we have multiple types for space-conservation reasons?
342  * (v4 vs v6?  prefix vs. range)?
343  */
344 
345 typedef struct ipsec_selkey
346 {
347 	uint32_t	ipsl_valid;		/* bitmask of valid entries */
348 #define	IPSL_REMOTE_ADDR		0x00000001
349 #define	IPSL_LOCAL_ADDR			0x00000002
350 #define	IPSL_REMOTE_PORT		0x00000004
351 #define	IPSL_LOCAL_PORT			0x00000008
352 #define	IPSL_PROTOCOL			0x00000010
353 #define	IPSL_ICMP_TYPE			0x00000020
354 #define	IPSL_ICMP_CODE			0x00000040
355 #define	IPSL_IPV6			0x00000080
356 #define	IPSL_IPV4			0x00000100
357 
358 #define	IPSL_WILDCARD			0x0000007f
359 
360 	ipsec_addr_t	ipsl_local;
361 	ipsec_addr_t	ipsl_remote;
362 	uint16_t	ipsl_lport;
363 	uint16_t	ipsl_rport;
364 	/*
365 	 * ICMP type and code selectors. Both have an end value to
366 	 * specify ranges, or * and *_end are equal for a single
367 	 * value
368 	 */
369 	uint8_t		ipsl_icmp_type;
370 	uint8_t		ipsl_icmp_type_end;
371 	uint8_t		ipsl_icmp_code;
372 	uint8_t		ipsl_icmp_code_end;
373 
374 	uint8_t		ipsl_proto;		/* ip payload type */
375 	uint8_t		ipsl_local_pfxlen;	/* #bits of prefix */
376 	uint8_t		ipsl_remote_pfxlen;	/* #bits of prefix */
377 	uint8_t		ipsl_mbz;
378 
379 	/* Insert new elements above this line */
380 	uint32_t	ipsl_pol_hval;
381 	uint32_t	ipsl_sel_hval;
382 } ipsec_selkey_t;
383 
384 typedef struct ipsec_sel
385 {
386 	HASH_LINK(ipsl_hash, struct ipsec_sel);
387 	uint32_t	ipsl_refs;		/* # refs to this sel */
388 	ipsec_selkey_t	ipsl_key;		/* actual selector guts */
389 } ipsec_sel_t;
390 
391 /*
392  * One policy rule.  This will be linked into a single hash chain bucket in
393  * the parent rule structure.  If the selector is simple enough to
394  * allow hashing, it gets filed under ipsec_policy_root_t->ipr_hash.
395  * Otherwise it goes onto a linked list in ipsec_policy_root_t->ipr_nonhash[af]
396  *
397  * In addition, we file the rule into an avl tree keyed by the rule index.
398  * (Duplicate rules are permitted; the comparison function breaks ties).
399  */
400 struct ipsec_policy_s
401 {
402 	HASH_LINK(ipsp_hash, struct ipsec_policy_s);
403 	avl_node_t		ipsp_byid;
404 	uint64_t		ipsp_index;	/* unique id */
405 	uint32_t		ipsp_prio; 	/* rule priority */
406 	uint32_t		ipsp_refs;
407 	ipsec_sel_t		*ipsp_sel;	/* selector set (shared) */
408 	ipsec_action_t		*ipsp_act; 	/* action (may be shared) */
409 };
410 
411 #define	IPPOL_REFHOLD(ipp) {			\
412 	atomic_add_32(&(ipp)->ipsp_refs, 1);	\
413 	ASSERT((ipp)->ipsp_refs != 0);		\
414 }
415 #define	IPPOL_REFRELE(ipp, ns) {				\
416 	ASSERT((ipp)->ipsp_refs != 0);				\
417 	membar_exit();						\
418 	if (atomic_add_32_nv(&(ipp)->ipsp_refs, -1) == 0)	\
419 		ipsec_policy_free(ipp, ns);			\
420 	(ipp) = 0;						\
421 }
422 
423 #define	IPPOL_UNCHAIN(php, ip, ns)					\
424 	HASHLIST_UNCHAIN((ip), ipsp_hash);				\
425 	avl_remove(&(php)->iph_rulebyid, (ip));				\
426 	IPPOL_REFRELE(ip, ns);
427 
428 /*
429  * Policy ruleset.  One per (protocol * direction) for system policy.
430  */
431 
432 #define	IPSEC_AF_V4	0
433 #define	IPSEC_AF_V6	1
434 #define	IPSEC_NAF	2
435 
436 typedef struct ipsec_policy_root_s
437 {
438 	ipsec_policy_t		*ipr_nonhash[IPSEC_NAF];
439 	int			ipr_nchains;
440 	ipsec_policy_hash_t 	*ipr_hash;
441 } ipsec_policy_root_t;
442 
443 /*
444  * Policy head.  One for system policy; there may also be one present
445  * on ill_t's with interface-specific policy, as well as one present
446  * for sockets with per-socket policy allocated.
447  */
448 
449 typedef struct ipsec_policy_head_s
450 {
451 	uint32_t	iph_refs;
452 	krwlock_t	iph_lock;
453 	uint64_t	iph_gen; /* generation number */
454 	ipsec_policy_root_t iph_root[IPSEC_NTYPES];
455 	avl_tree_t	iph_rulebyid;
456 } ipsec_policy_head_t;
457 
458 #define	IPPH_REFHOLD(iph) {			\
459 	atomic_add_32(&(iph)->iph_refs, 1);	\
460 	ASSERT((iph)->iph_refs != 0);		\
461 }
462 #define	IPPH_REFRELE(iph, ns) {					\
463 	ASSERT((iph)->iph_refs != 0);				\
464 	membar_exit();						\
465 	if (atomic_add_32_nv(&(iph)->iph_refs, -1) == 0)	\
466 		ipsec_polhead_free(iph, ns);			\
467 	(iph) = 0;						\
468 }
469 
470 /*
471  * IPsec fragment related structures
472  */
473 
474 typedef struct ipsec_fragcache_entry {
475 	struct ipsec_fragcache_entry *itpfe_next;	/* hash list chain */
476 	mblk_t *itpfe_fraglist;			/* list of fragments */
477 	time_t itpfe_exp;			/* time when entry is stale */
478 	int itpfe_depth;			/* # of fragments in list */
479 	ipsec_addr_t itpfe_frag_src;
480 	ipsec_addr_t itpfe_frag_dst;
481 #define	itpfe_src itpfe_frag_src.ipsad_v4
482 #define	itpfe_src6 itpfe_frag_src.ipsad_v6
483 #define	itpfe_dst itpfe_frag_dst.ipsad_v4
484 #define	itpfe_dst6 itpfe_frag_dst.ipsad_v6
485 	uint32_t itpfe_id;			/* IP datagram ID */
486 	uint8_t itpfe_proto;			/* IP Protocol */
487 	uint8_t itpfe_last;			/* Last packet */
488 } ipsec_fragcache_entry_t;
489 
490 typedef struct ipsec_fragcache {
491 	kmutex_t itpf_lock;
492 	struct ipsec_fragcache_entry **itpf_ptr;
493 	struct ipsec_fragcache_entry *itpf_freelist;
494 	time_t itpf_expire_hint;	/* time when oldest entry is stale */
495 } ipsec_fragcache_t;
496 
497 /*
498  * Tunnel policies.  We keep a minature of the transport-mode/global policy
499  * per each tunnel instance.
500  *
501  * People who need both an itp held down AND one of its polheads need to
502  * first lock the itp, THEN the polhead, otherwise deadlock WILL occur.
503  */
504 typedef struct ipsec_tun_pol_s {
505 	avl_node_t itp_node;
506 	kmutex_t itp_lock;
507 	uint64_t itp_next_policy_index;
508 	ipsec_policy_head_t *itp_policy;
509 	ipsec_policy_head_t *itp_inactive;
510 	uint32_t itp_flags;
511 	uint32_t itp_refcnt;
512 	char itp_name[LIFNAMSIZ];
513 	ipsec_fragcache_t itp_fragcache;
514 } ipsec_tun_pol_t;
515 /* NOTE - Callers (tun code) synchronize their own instances for these flags. */
516 #define	ITPF_P_ACTIVE 0x1	/* Are we using IPsec right now? */
517 #define	ITPF_P_TUNNEL 0x2	/* Negotiate tunnel-mode */
518 /* Optimization -> Do we have per-port security entries in this polhead? */
519 #define	ITPF_P_PER_PORT_SECURITY 0x4
520 #define	ITPF_PFLAGS 0x7
521 #define	ITPF_SHIFT 3
522 
523 #define	ITPF_I_ACTIVE 0x8	/* Is the inactive using IPsec right now? */
524 #define	ITPF_I_TUNNEL 0x10	/* Negotiate tunnel-mode (on inactive) */
525 /* Optimization -> Do we have per-port security entries in this polhead? */
526 #define	ITPF_I_PER_PORT_SECURITY 0x20
527 #define	ITPF_IFLAGS 0x38
528 
529 /* NOTE:  f cannot be an expression. */
530 #define	ITPF_CLONE(f) (f) = (((f) & ITPF_PFLAGS) | \
531 	    (((f) & ITPF_PFLAGS) << ITPF_SHIFT));
532 #define	ITPF_SWAP(f) (f) = ((((f) & ITPF_PFLAGS) << ITPF_SHIFT) | \
533 	    (((f) & ITPF_IFLAGS) >> ITPF_SHIFT))
534 
535 #define	ITP_P_ISACTIVE(itp, iph) ((itp)->itp_flags & \
536 	(((itp)->itp_policy == (iph)) ? ITPF_P_ACTIVE : ITPF_I_ACTIVE))
537 
538 #define	ITP_P_ISTUNNEL(itp, iph) ((itp)->itp_flags & \
539 	(((itp)->itp_policy == (iph)) ? ITPF_P_TUNNEL : ITPF_I_TUNNEL))
540 
541 #define	ITP_P_ISPERPORT(itp, iph) ((itp)->itp_flags & \
542 	(((itp)->itp_policy == (iph)) ? ITPF_P_PER_PORT_SECURITY : \
543 	ITPF_I_PER_PORT_SECURITY))
544 
545 #define	ITP_REFHOLD(itp) { \
546 	atomic_add_32(&((itp)->itp_refcnt), 1);	\
547 	ASSERT((itp)->itp_refcnt != 0); \
548 }
549 
550 #define	ITP_REFRELE(itp, ns) { \
551 	ASSERT((itp)->itp_refcnt != 0); \
552 	membar_exit(); \
553 	if (atomic_add_32_nv(&((itp)->itp_refcnt), -1) == 0) \
554 		itp_free(itp, ns); \
555 }
556 
557 /*
558  * Certificate identity.
559  */
560 
561 typedef struct ipsid_s
562 {
563 	struct ipsid_s *ipsid_next;
564 	struct ipsid_s **ipsid_ptpn;
565 	uint32_t	ipsid_refcnt;
566 	int		ipsid_type;	/* id type */
567 	char 		*ipsid_cid;	/* certificate id string */
568 } ipsid_t;
569 
570 /*
571  * ipsid_t reference hold/release macros, just like ipsa versions.
572  */
573 
574 #define	IPSID_REFHOLD(ipsid) {			\
575 	atomic_add_32(&(ipsid)->ipsid_refcnt, 1);	\
576 	ASSERT((ipsid)->ipsid_refcnt != 0);	\
577 }
578 
579 /*
580  * Decrement the reference count on the ID.  Someone else will clean up
581  * after us later.
582  */
583 
584 #define	IPSID_REFRELE(ipsid) {					\
585 	membar_exit();						\
586 	atomic_add_32(&(ipsid)->ipsid_refcnt, -1);		\
587 }
588 
589 struct ipsec_out_s;
590 
591 /*
592  * Following are the estimates of what the maximum AH and ESP header size
593  * would be. This is used to tell the upper layer the right value of MSS
594  * it should use without consulting AH/ESP. If the size is something
595  * different from this, ULP will learn the right one through
596  * ICMP_FRAGMENTATION_NEEDED messages generated locally.
597  *
598  * AH : 12 bytes of constant header + 12 bytes of ICV checksum (MD5/SHA1).
599  *
600  * ESP : 8 bytes of constant header + 16 bytes of IV + 12 bytes ICV +
601  * 2 bytes of trailer + 15 bytes pad = 53
602  *
603  * Note that for ESP, this estimate is overly pessimistic; however, a
604  * more accurate estimate needs to know the exact amount of space
605  * which will be available to ESP so it can just leave 2 bytes free in
606  * the last cipherblock for the ESP inner trailer, and that
607  * information is not available at the right moment in the current
608  * stack.
609  */
610 #define	IPSEC_MAX_AH_HDR_SIZE   (24)
611 #define	IPSEC_MAX_ESP_HDR_SIZE  (53)
612 
613 /* Alternate, when we know the crypto block size */
614 #define	IPSEC_BASE_ESP_HDR_SIZE(sa) (4 + 4 + 12 + 1 + 2 * (sa)->ipsa_iv_len)
615 #define	IPSEC_DEF_BLOCKSIZE	(8) /* safe default */
616 
617 /*
618  * Identity hash table.
619  *
620  * Identities are refcounted and "interned" into the hash table.
621  * Only references coming from other objects (SA's, latching state)
622  * are counted in ipsid_refcnt.
623  *
624  * Locking: IPSID_REFHOLD is safe only when (a) the object's hash bucket
625  * is locked, (b) we know that the refcount must be > 0.
626  *
627  * The ipsid_next and ipsid_ptpn fields are only to be referenced or
628  * modified when the bucket lock is held; in particular, we only
629  * delete objects while holding the bucket lock, and we only increase
630  * the refcount from 0 to 1 while the bucket lock is held.
631  */
632 
633 #define	IPSID_HASHSIZE 64
634 
635 typedef struct ipsif_s
636 {
637 	ipsid_t *ipsif_head;
638 	kmutex_t ipsif_lock;
639 } ipsif_t;
640 
641 
642 /*
643  * IPSEC stack instances
644  */
645 struct ipsec_stack {
646 	netstack_t		*ipsec_netstack;	/* Common netstack */
647 
648 	/* Packet dropper for IP IPsec processing failures */
649 	ipdropper_t		ipsec_dropper;
650 
651 /* From spd.c */
652 	/*
653 	 * Policy rule index generator.  We assume this won't wrap in the
654 	 * lifetime of a system.  If we make 2^20 policy changes per second,
655 	 * this will last 2^44 seconds, or roughly 500,000 years, so we don't
656 	 * have to worry about reusing policy index values.
657 	 */
658 	uint64_t		ipsec_next_policy_index;
659 
660 	HASH_HEAD(ipsec_action_s) ipsec_action_hash[IPSEC_ACTION_HASH_SIZE];
661 	HASH_HEAD(ipsec_sel)	  *ipsec_sel_hash;
662 	uint32_t		ipsec_spd_hashsize;
663 
664 	ipsif_t			ipsec_ipsid_buckets[IPSID_HASHSIZE];
665 
666 	/*
667 	 * Active & Inactive system policy roots
668 	 */
669 	ipsec_policy_head_t	ipsec_system_policy;
670 	ipsec_policy_head_t	ipsec_inactive_policy;
671 
672 	/* Packet dropper for generic SPD drops. */
673 	ipdropper_t		ipsec_spd_dropper;
674 	krwlock_t		ipsec_itp_get_byaddr_rw_lock;
675 	ipsec_tun_pol_t		*(*ipsec_itp_get_byaddr)
676 	    (uint32_t *, uint32_t *, int, netstack_t *);
677 
678 /* ipdrop.c */
679 	kstat_t			*ipsec_ip_drop_kstat;
680 	struct ip_dropstats	*ipsec_ip_drop_types;
681 
682 /* spd.c */
683 	/*
684 	 * Have a counter for every possible policy message in
685 	 * ipsec_policy_failure_msgs
686 	 */
687 	uint32_t		ipsec_policy_failure_count[IPSEC_POLICY_MAX];
688 	/* Time since last ipsec policy failure that printed a message. */
689 	hrtime_t		ipsec_policy_failure_last;
690 
691 /* ip_spd.c */
692 	/* stats */
693 	kstat_t			*ipsec_ksp;
694 	struct ipsec_kstats_s	*ipsec_kstats;
695 
696 /* sadb.c */
697 	/* Packet dropper for generic SADB drops. */
698 	ipdropper_t		ipsec_sadb_dropper;
699 
700 /* spd.c */
701 	boolean_t		ipsec_inbound_v4_policy_present;
702 	boolean_t		ipsec_outbound_v4_policy_present;
703 	boolean_t		ipsec_inbound_v6_policy_present;
704 	boolean_t		ipsec_outbound_v6_policy_present;
705 
706 /* spd.c */
707 	/*
708 	 * Because policy needs to know what algorithms are supported, keep the
709 	 * lists of algorithms here.
710 	 */
711 	kmutex_t 		ipsec_alg_lock;
712 
713 	uint8_t			ipsec_nalgs[IPSEC_NALGTYPES];
714 	ipsec_alginfo_t	*ipsec_alglists[IPSEC_NALGTYPES][IPSEC_MAX_ALGS];
715 
716 	uint8_t		ipsec_sortlist[IPSEC_NALGTYPES][IPSEC_MAX_ALGS];
717 
718 	int		ipsec_algs_exec_mode[IPSEC_NALGTYPES];
719 
720 	uint32_t 	ipsec_tun_spd_hashsize;
721 	/*
722 	 * Tunnel policies - AVL tree indexed by tunnel name.
723 	 */
724 	krwlock_t 	ipsec_tunnel_policy_lock;
725 	uint64_t	ipsec_tunnel_policy_gen;
726 	avl_tree_t	ipsec_tunnel_policies;
727 
728 /* ipsec_loader.c */
729 	kmutex_t	ipsec_loader_lock;
730 	int		ipsec_loader_state;
731 	int		ipsec_loader_sig;
732 	kt_did_t	ipsec_loader_tid;
733 	kcondvar_t	ipsec_loader_sig_cv;	/* For loader_sig conditions. */
734 
735 };
736 typedef struct ipsec_stack ipsec_stack_t;
737 
738 /* Handle the kstat_create in ip_drop_init() failing */
739 #define	DROPPER(_ipss, _dropper) \
740 	(((_ipss)->ipsec_ip_drop_types == NULL) ? NULL : \
741 	&((_ipss)->ipsec_ip_drop_types->_dropper))
742 
743 /*
744  * Loader states..
745  */
746 #define	IPSEC_LOADER_WAIT	0
747 #define	IPSEC_LOADER_FAILED	-1
748 #define	IPSEC_LOADER_SUCCEEDED	1
749 
750 /*
751  * ipsec_loader entrypoints.
752  */
753 extern void ipsec_loader_init(ipsec_stack_t *);
754 extern void ipsec_loader_start(ipsec_stack_t *);
755 extern void ipsec_loader_destroy(ipsec_stack_t *);
756 extern void ipsec_loader_loadnow(ipsec_stack_t *);
757 extern boolean_t ipsec_loader_wait(queue_t *q, ipsec_stack_t *);
758 extern boolean_t ipsec_loaded(ipsec_stack_t *);
759 extern boolean_t ipsec_failed(ipsec_stack_t *);
760 
761 /*
762  * callback from ipsec_loader to ip
763  */
764 extern void ip_ipsec_load_complete(ipsec_stack_t *);
765 
766 /*
767  * ipsec policy entrypoints (spd.c)
768  */
769 
770 extern void ipsec_policy_g_destroy(void);
771 extern void ipsec_policy_g_init(void);
772 
773 extern int ipsec_alloc_table(ipsec_policy_head_t *, int, int, boolean_t,
774     netstack_t *);
775 extern void ipsec_polhead_init(ipsec_policy_head_t *, int);
776 extern void ipsec_polhead_destroy(ipsec_policy_head_t *);
777 extern void ipsec_polhead_free_table(ipsec_policy_head_t *);
778 extern mblk_t *ipsec_check_global_policy(mblk_t *, conn_t *, ipha_t *,
779 		    ip6_t *, boolean_t, netstack_t *);
780 extern mblk_t *ipsec_check_inbound_policy(mblk_t *, conn_t *, ipha_t *, ip6_t *,
781     boolean_t);
782 
783 extern boolean_t ipsec_in_to_out(mblk_t *, ipha_t *, ip6_t *);
784 extern void ipsec_log_policy_failure(int, char *, ipha_t *, ip6_t *, boolean_t,
785 		    netstack_t *);
786 extern boolean_t ipsec_inbound_accept_clear(mblk_t *, ipha_t *, ip6_t *);
787 extern int ipsec_conn_cache_policy(conn_t *, boolean_t);
788 extern mblk_t *ipsec_alloc_ipsec_out(netstack_t *);
789 extern mblk_t	*ipsec_attach_ipsec_out(mblk_t **, conn_t *, ipsec_policy_t *,
790     uint8_t, netstack_t *);
791 extern mblk_t	*ipsec_init_ipsec_out(mblk_t *, mblk_t **, conn_t *,
792     ipsec_policy_t *, uint8_t, netstack_t *);
793 struct ipsec_in_s;
794 extern ipsec_action_t *ipsec_in_to_out_action(struct ipsec_in_s *);
795 extern boolean_t ipsec_check_ipsecin_latch(struct ipsec_in_s *, mblk_t *,
796     struct ipsec_latch_s *, ipha_t *, ip6_t *, const char **, kstat_named_t **,
797     conn_t *);
798 extern void ipsec_latch_inbound(ipsec_latch_t *ipl, struct ipsec_in_s *ii);
799 
800 extern void ipsec_policy_free(ipsec_policy_t *, netstack_t *);
801 extern void ipsec_action_free(ipsec_action_t *);
802 extern void ipsec_polhead_free(ipsec_policy_head_t *, netstack_t *);
803 extern ipsec_policy_head_t *ipsec_polhead_split(ipsec_policy_head_t *,
804     netstack_t *);
805 extern ipsec_policy_head_t *ipsec_polhead_create(void);
806 extern ipsec_policy_head_t *ipsec_system_policy(netstack_t *);
807 extern ipsec_policy_head_t *ipsec_inactive_policy(netstack_t *);
808 extern void ipsec_swap_policy(ipsec_policy_head_t *, ipsec_policy_head_t *,
809     netstack_t *);
810 extern void ipsec_swap_global_policy(netstack_t *);
811 
812 extern int ipsec_clone_system_policy(netstack_t *);
813 extern ipsec_policy_t *ipsec_policy_create(ipsec_selkey_t *,
814     const ipsec_act_t *, int, int, uint64_t *, netstack_t *);
815 extern boolean_t ipsec_policy_delete(ipsec_policy_head_t *,
816     ipsec_selkey_t *, int, netstack_t *);
817 extern int ipsec_policy_delete_index(ipsec_policy_head_t *, uint64_t,
818     netstack_t *);
819 extern void ipsec_polhead_flush(ipsec_policy_head_t *, netstack_t *);
820 extern int ipsec_copy_polhead(ipsec_policy_head_t *, ipsec_policy_head_t *,
821     netstack_t *);
822 extern void ipsec_actvec_from_req(ipsec_req_t *, ipsec_act_t **, uint_t *,
823     netstack_t *);
824 extern void ipsec_actvec_free(ipsec_act_t *, uint_t);
825 extern int ipsec_req_from_head(ipsec_policy_head_t *, ipsec_req_t *, int);
826 extern mblk_t *ipsec_construct_inverse_acquire(sadb_msg_t *, sadb_ext_t **,
827     netstack_t *);
828 extern mblk_t *ip_wput_attach_policy(mblk_t *, ipha_t *, ip6_t *, ire_t *,
829     conn_t *, boolean_t, zoneid_t);
830 extern mblk_t	*ip_wput_ire_parse_ipsec_out(mblk_t *, ipha_t *, ip6_t *,
831     ire_t *, conn_t *, boolean_t, zoneid_t);
832 extern ipsec_policy_t *ipsec_find_policy(int, conn_t *,
833     struct ipsec_out_s *, ipsec_selector_t *, netstack_t *);
834 extern ipsid_t *ipsid_lookup(int, char *, netstack_t *);
835 extern boolean_t ipsid_equal(ipsid_t *, ipsid_t *);
836 extern void ipsid_gc(netstack_t *);
837 extern void ipsec_latch_ids(ipsec_latch_t *, ipsid_t *, ipsid_t *);
838 
839 extern void ipsec_config_flush(netstack_t *);
840 extern boolean_t ipsec_check_policy(ipsec_policy_head_t *, ipsec_policy_t *,
841     int);
842 extern void ipsec_enter_policy(ipsec_policy_head_t *, ipsec_policy_t *, int,
843     netstack_t *);
844 extern boolean_t ipsec_check_action(ipsec_act_t *, int *, netstack_t *);
845 
846 extern mblk_t *ipsec_out_tag(mblk_t *, mblk_t *, netstack_t *);
847 extern mblk_t *ipsec_in_tag(mblk_t *, mblk_t *, netstack_t *);
848 extern mblk_t *ip_copymsg(mblk_t *mp);
849 
850 extern void iplatch_free(ipsec_latch_t *, netstack_t *);
851 extern ipsec_latch_t *iplatch_create(void);
852 extern int ipsec_set_req(cred_t *, conn_t *, ipsec_req_t *);
853 
854 extern void ipsec_insert_always(avl_tree_t *tree, void *new_node);
855 
856 extern int32_t ipsec_act_ovhd(const ipsec_act_t *act);
857 
858 
859 extern boolean_t iph_ipvN(ipsec_policy_head_t *, boolean_t);
860 
861 /*
862  * Tunnel-support SPD functions and variables.
863  */
864 struct tun_s;	/* Defined in inet/tun.h. */
865 extern boolean_t ipsec_tun_inbound(mblk_t *, mblk_t **,  ipsec_tun_pol_t *,
866     ipha_t *, ip6_t *, ipha_t *, ip6_t *, int, netstack_t *);
867 extern mblk_t *ipsec_tun_outbound(mblk_t *, struct tun_s *, ipha_t *,
868     ip6_t *, ipha_t *, ip6_t *, int, netstack_t *);
869 extern void itp_free(ipsec_tun_pol_t *, netstack_t *);
870 extern ipsec_tun_pol_t *create_tunnel_policy(char *, int *, uint64_t *,
871     netstack_t *);
872 extern ipsec_tun_pol_t *get_tunnel_policy(char *, netstack_t *);
873 extern void itp_unlink(ipsec_tun_pol_t *, netstack_t *);
874 extern void itp_walk(void (*)(ipsec_tun_pol_t *, void *, netstack_t *),
875     void *, netstack_t *);
876 
877 extern ipsec_tun_pol_t *itp_get_byaddr_dummy(uint32_t *, uint32_t *,
878     int, netstack_t *);
879 
880 /*
881  * IPsec AH/ESP functions called from IP.
882  */
883 
884 extern void ipsecah_in_assocfailure(mblk_t *, char, ushort_t, char *,
885     uint32_t, void *, int, ipsecah_stack_t *);
886 extern void ipsecesp_in_assocfailure(mblk_t *, char, ushort_t, char *,
887     uint32_t, void *, int, ipsecesp_stack_t *);
888 
889 /*
890  * Algorithm management helper functions.
891  */
892 extern boolean_t ipsec_valid_key_size(uint16_t, ipsec_alginfo_t *);
893 
894 /*
895  * Per-socket policy, for now, takes precedence... this priority value
896  * insures it.
897  */
898 #define	IPSEC_PRIO_SOCKET		0x1000000
899 
900 /* DDI initialization functions. */
901 extern	boolean_t    ipsecesp_ddi_init(void);
902 extern	boolean_t    ipsecah_ddi_init(void);
903 extern	boolean_t    keysock_ddi_init(void);
904 extern	boolean_t    spdsock_ddi_init(void);
905 
906 extern	void    ipsecesp_ddi_destroy(void);
907 extern	void    ipsecah_ddi_destroy(void);
908 extern	void	keysock_ddi_destroy(void);
909 extern	void    spdsock_ddi_destroy(void);
910 
911 /*
912  * AH- and ESP-specific functions that are called directly by other modules.
913  */
914 extern void ipsecah_fill_defs(struct sadb_x_ecomb *, netstack_t *);
915 extern void ipsecesp_fill_defs(struct sadb_x_ecomb *, netstack_t *);
916 extern void ipsecah_algs_changed(netstack_t *);
917 extern void ipsecesp_algs_changed(netstack_t *);
918 extern void ipsecesp_init_funcs(ipsa_t *);
919 extern void ipsecah_init_funcs(ipsa_t *);
920 extern ipsec_status_t ipsecah_icmp_error(mblk_t *);
921 extern ipsec_status_t ipsecesp_icmp_error(mblk_t *);
922 
923 /*
924  * Wrapper for putnext() to ipsec accelerated interface.
925  */
926 extern void ipsec_hw_putnext(queue_t *, mblk_t *);
927 
928 /*
929  * spdsock functions that are called directly by IP.
930  */
931 extern void spdsock_update_pending_algs(netstack_t *);
932 
933 /*
934  * IP functions that are called from AH and ESP.
935  */
936 extern boolean_t ipsec_outbound_sa(mblk_t *, uint_t);
937 extern esph_t *ipsec_inbound_esp_sa(mblk_t *, netstack_t *);
938 extern ah_t *ipsec_inbound_ah_sa(mblk_t *, netstack_t *);
939 extern ipsec_policy_t *ipsec_find_policy_head(ipsec_policy_t *,
940     ipsec_policy_head_t *, int, ipsec_selector_t *, netstack_t *);
941 
942 /*
943  * IP dropper init/destroy.
944  */
945 void ip_drop_init(ipsec_stack_t *);
946 void ip_drop_destroy(ipsec_stack_t *);
947 
948 /*
949  * NAT-Traversal cleanup
950  */
951 extern void nattymod_clean_ipif(ipif_t *);
952 
953 /*
954  * Common functions
955  */
956 extern boolean_t ip_addr_match(uint8_t *, int, in6_addr_t *);
957 
958 /*
959  * AH and ESP counters types.
960  */
961 typedef uint32_t ah_counter;
962 typedef uint32_t esp_counter;
963 
964 #endif /* _KERNEL */
965 
966 #ifdef	__cplusplus
967 }
968 #endif
969 
970 #endif	/* _INET_IPSEC_IMPL_H */
971