xref: /titanic_41/usr/src/uts/common/inet/ip/spdsock.c (revision 24a1f0af9f770e0e795ef1fa1c6dece8dd8dc959)
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 2008 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #pragma ident	"%Z%%M%	%I%	%E% SMI"
27 
28 #include <sys/param.h>
29 #include <sys/types.h>
30 #include <sys/stream.h>
31 #include <sys/strsubr.h>
32 #include <sys/strsun.h>
33 #include <sys/stropts.h>
34 #include <sys/zone.h>
35 #include <sys/vnode.h>
36 #include <sys/sysmacros.h>
37 #define	_SUN_TPI_VERSION 2
38 #include <sys/tihdr.h>
39 #include <sys/ddi.h>
40 #include <sys/sunddi.h>
41 #include <sys/mkdev.h>
42 #include <sys/debug.h>
43 #include <sys/kmem.h>
44 #include <sys/cmn_err.h>
45 #include <sys/suntpi.h>
46 #include <sys/policy.h>
47 
48 #include <sys/socket.h>
49 #include <netinet/in.h>
50 #include <net/pfkeyv2.h>
51 #include <net/pfpolicy.h>
52 
53 #include <inet/common.h>
54 #include <netinet/ip6.h>
55 #include <inet/ip.h>
56 #include <inet/ip6.h>
57 #include <inet/mi.h>
58 #include <inet/nd.h>
59 #include <inet/ip_if.h>
60 #include <inet/tun.h>
61 #include <inet/optcom.h>
62 #include <inet/ipsec_info.h>
63 #include <inet/ipsec_impl.h>
64 #include <inet/spdsock.h>
65 #include <inet/sadb.h>
66 
67 #include <sys/isa_defs.h>
68 
69 #include <c2/audit.h>
70 
71 /*
72  * This is a transport provider for the PF_POLICY IPsec policy
73  * management socket, which provides a management interface into the
74  * SPD, allowing policy rules to be added, deleted, and queried.
75  *
76  * This effectively replaces the old private SIOC*IPSECONFIG ioctls
77  * with an extensible interface which will hopefully be public some
78  * day.
79  *
80  * See <net/pfpolicy.h> for more details on the protocol.
81  *
82  * We link against drv/ip and call directly into it to manipulate the
83  * SPD; see ipsec_impl.h for the policy data structures and spd.c for
84  * the code which maintains them.
85  *
86  * The MT model of this is QPAIR with the addition of some explicit
87  * locking to protect system-wide policy data structures.
88  */
89 
90 static vmem_t *spdsock_vmem;		/* for minor numbers. */
91 
92 #define	ALIGNED64(x) IS_P2ALIGNED((x), sizeof (uint64_t))
93 
94 /* Default structure copied into T_INFO_ACK messages (from rts.c...) */
95 static struct T_info_ack spdsock_g_t_info_ack = {
96 	T_INFO_ACK,
97 	T_INFINITE,	/* TSDU_size. Maximum size messages. */
98 	T_INVALID,	/* ETSDU_size. No expedited data. */
99 	T_INVALID,	/* CDATA_size. No connect data. */
100 	T_INVALID,	/* DDATA_size. No disconnect data. */
101 	0,		/* ADDR_size. */
102 	0,		/* OPT_size. No user-settable options */
103 	64 * 1024,	/* TIDU_size. spdsock allows maximum size messages. */
104 	T_COTS,		/* SERV_type. spdsock supports connection oriented. */
105 	TS_UNBND,	/* CURRENT_state. This is set from spdsock_state. */
106 	(XPG4_1)	/* Provider flags */
107 };
108 
109 /* Named Dispatch Parameter Management Structure */
110 typedef struct spdsockparam_s {
111 	uint_t	spdsock_param_min;
112 	uint_t	spdsock_param_max;
113 	uint_t	spdsock_param_value;
114 	char *spdsock_param_name;
115 } spdsockparam_t;
116 
117 /*
118  * Table of NDD variables supported by spdsock. These are loaded into
119  * spdsock_g_nd in spdsock_init_nd.
120  * All of these are alterable, within the min/max values given, at run time.
121  */
122 static	spdsockparam_t	lcl_param_arr[] = {
123 	/* min	max	value	name */
124 	{ 4096, 65536,	8192,	"spdsock_xmit_hiwat"},
125 	{ 0,	65536,	1024,	"spdsock_xmit_lowat"},
126 	{ 4096, 65536,	8192,	"spdsock_recv_hiwat"},
127 	{ 65536, 1024*1024*1024, 256*1024,	"spdsock_max_buf"},
128 	{ 0,	3,	0,	"spdsock_debug"},
129 };
130 #define	spds_xmit_hiwat	spds_params[0].spdsock_param_value
131 #define	spds_xmit_lowat	spds_params[1].spdsock_param_value
132 #define	spds_recv_hiwat	spds_params[2].spdsock_param_value
133 #define	spds_max_buf	spds_params[3].spdsock_param_value
134 #define	spds_debug		spds_params[4].spdsock_param_value
135 
136 #define	ss0dbg(a)	printf a
137 /* NOTE:  != 0 instead of > 0 so lint doesn't complain. */
138 #define	ss1dbg(spds, a)	if (spds->spds_debug != 0) printf a
139 #define	ss2dbg(spds, a)	if (spds->spds_debug > 1) printf a
140 #define	ss3dbg(spds, a)	if (spds->spds_debug > 2) printf a
141 
142 #define	RESET_SPDSOCK_DUMP_POLHEAD(ss, iph) { \
143 	ASSERT(RW_READ_HELD(&(iph)->iph_lock)); \
144 	(ss)->spdsock_dump_head = (iph); \
145 	(ss)->spdsock_dump_gen = (iph)->iph_gen; \
146 	(ss)->spdsock_dump_cur_type = 0; \
147 	(ss)->spdsock_dump_cur_af = IPSEC_AF_V4; \
148 	(ss)->spdsock_dump_cur_rule = NULL; \
149 	(ss)->spdsock_dump_count = 0; \
150 	(ss)->spdsock_dump_cur_chain = 0; \
151 }
152 
153 static int spdsock_close(queue_t *);
154 static int spdsock_open(queue_t *, dev_t *, int, int, cred_t *);
155 static void spdsock_wput(queue_t *, mblk_t *);
156 static void spdsock_wsrv(queue_t *);
157 static void spdsock_rsrv(queue_t *);
158 static void *spdsock_stack_init(netstackid_t stackid, netstack_t *ns);
159 static void spdsock_stack_fini(netstackid_t stackid, void *arg);
160 static void spdsock_loadcheck(void *);
161 static void spdsock_merge_algs(spd_stack_t *);
162 static void spdsock_flush_one(ipsec_policy_head_t *, netstack_t *);
163 static mblk_t *spdsock_dump_next_record(spdsock_t *);
164 
165 static struct module_info info = {
166 	5138, "spdsock", 1, INFPSZ, 512, 128
167 };
168 
169 static struct qinit rinit = {
170 	NULL, (pfi_t)spdsock_rsrv, spdsock_open, spdsock_close,
171 	NULL, &info
172 };
173 
174 static struct qinit winit = {
175 	(pfi_t)spdsock_wput, (pfi_t)spdsock_wsrv, NULL, NULL, NULL, &info
176 };
177 
178 struct streamtab spdsockinfo = {
179 	&rinit, &winit
180 };
181 
182 /* mapping from alg type to protocol number, as per RFC 2407 */
183 static const uint_t algproto[] = {
184 	PROTO_IPSEC_AH,
185 	PROTO_IPSEC_ESP,
186 };
187 
188 #define	NALGPROTOS	(sizeof (algproto) / sizeof (algproto[0]))
189 
190 /* mapping from kernel exec mode to spdsock exec mode */
191 static const uint_t execmodes[] = {
192 	SPD_ALG_EXEC_MODE_SYNC,
193 	SPD_ALG_EXEC_MODE_ASYNC
194 };
195 
196 #define	NEXECMODES	(sizeof (execmodes) / sizeof (execmodes[0]))
197 
198 #define	ALL_ACTIVE_POLHEADS ((ipsec_policy_head_t *)-1)
199 #define	ALL_INACTIVE_POLHEADS ((ipsec_policy_head_t *)-2)
200 
201 #define	ITP_NAME(itp) (itp != NULL ? itp->itp_name : NULL)
202 
203 /* ARGSUSED */
204 static int
205 spdsock_param_get(q, mp, cp, cr)
206 	queue_t	*q;
207 	mblk_t	*mp;
208 	caddr_t	cp;
209 	cred_t *cr;
210 {
211 	spdsockparam_t	*spdsockpa = (spdsockparam_t *)cp;
212 	uint_t value;
213 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
214 	spd_stack_t	*spds = ss->spdsock_spds;
215 
216 	mutex_enter(&spds->spds_param_lock);
217 	value = spdsockpa->spdsock_param_value;
218 	mutex_exit(&spds->spds_param_lock);
219 
220 	(void) mi_mpprintf(mp, "%u", value);
221 	return (0);
222 }
223 
224 /* This routine sets an NDD variable in a spdsockparam_t structure. */
225 /* ARGSUSED */
226 static int
227 spdsock_param_set(q, mp, value, cp, cr)
228 	queue_t	*q;
229 	mblk_t	*mp;
230 	char *value;
231 	caddr_t	cp;
232 	cred_t *cr;
233 {
234 	ulong_t	new_value;
235 	spdsockparam_t	*spdsockpa = (spdsockparam_t *)cp;
236 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
237 	spd_stack_t	*spds = ss->spdsock_spds;
238 
239 	/* Convert the value from a string into a long integer. */
240 	if (ddi_strtoul(value, NULL, 10, &new_value) != 0)
241 		return (EINVAL);
242 
243 	mutex_enter(&spds->spds_param_lock);
244 	/*
245 	 * Fail the request if the new value does not lie within the
246 	 * required bounds.
247 	 */
248 	if (new_value < spdsockpa->spdsock_param_min ||
249 	    new_value > spdsockpa->spdsock_param_max) {
250 		mutex_exit(&spds->spds_param_lock);
251 		return (EINVAL);
252 	}
253 
254 	/* Set the new value */
255 	spdsockpa->spdsock_param_value = new_value;
256 	mutex_exit(&spds->spds_param_lock);
257 
258 	return (0);
259 }
260 
261 /*
262  * Initialize at module load time
263  */
264 boolean_t
265 spdsock_ddi_init(void)
266 {
267 	spdsock_max_optsize = optcom_max_optsize(
268 	    spdsock_opt_obj.odb_opt_des_arr, spdsock_opt_obj.odb_opt_arr_cnt);
269 
270 	spdsock_vmem = vmem_create("spdsock", (void *)1, MAXMIN, 1,
271 	    NULL, NULL, NULL, 1, VM_SLEEP | VMC_IDENTIFIER);
272 
273 	/*
274 	 * We want to be informed each time a stack is created or
275 	 * destroyed in the kernel, so we can maintain the
276 	 * set of spd_stack_t's.
277 	 */
278 	netstack_register(NS_SPDSOCK, spdsock_stack_init, NULL,
279 	    spdsock_stack_fini);
280 
281 	return (B_TRUE);
282 }
283 
284 /*
285  * Walk through the param array specified registering each element with the
286  * named dispatch handler.
287  */
288 static boolean_t
289 spdsock_param_register(IDP *ndp, spdsockparam_t *ssp, int cnt)
290 {
291 	for (; cnt-- > 0; ssp++) {
292 		if (ssp->spdsock_param_name != NULL &&
293 		    ssp->spdsock_param_name[0]) {
294 			if (!nd_load(ndp,
295 			    ssp->spdsock_param_name,
296 			    spdsock_param_get, spdsock_param_set,
297 			    (caddr_t)ssp)) {
298 				nd_free(ndp);
299 				return (B_FALSE);
300 			}
301 		}
302 	}
303 	return (B_TRUE);
304 }
305 
306 /*
307  * Initialize for each stack instance
308  */
309 /* ARGSUSED */
310 static void *
311 spdsock_stack_init(netstackid_t stackid, netstack_t *ns)
312 {
313 	spd_stack_t	*spds;
314 	spdsockparam_t	*ssp;
315 
316 	spds = (spd_stack_t *)kmem_zalloc(sizeof (*spds), KM_SLEEP);
317 	spds->spds_netstack = ns;
318 
319 	ASSERT(spds->spds_g_nd == NULL);
320 
321 	ssp = (spdsockparam_t *)kmem_alloc(sizeof (lcl_param_arr), KM_SLEEP);
322 	spds->spds_params = ssp;
323 	bcopy(lcl_param_arr, ssp, sizeof (lcl_param_arr));
324 
325 	(void) spdsock_param_register(&spds->spds_g_nd, ssp,
326 	    A_CNT(lcl_param_arr));
327 
328 	mutex_init(&spds->spds_param_lock, NULL, MUTEX_DEFAULT, NULL);
329 	mutex_init(&spds->spds_alg_lock, NULL, MUTEX_DEFAULT, NULL);
330 
331 	return (spds);
332 }
333 
334 void
335 spdsock_ddi_destroy(void)
336 {
337 	vmem_destroy(spdsock_vmem);
338 
339 	netstack_unregister(NS_SPDSOCK);
340 }
341 
342 /* ARGSUSED */
343 static void
344 spdsock_stack_fini(netstackid_t stackid, void *arg)
345 {
346 	spd_stack_t *spds = (spd_stack_t *)arg;
347 
348 	freemsg(spds->spds_mp_algs);
349 	mutex_destroy(&spds->spds_param_lock);
350 	mutex_destroy(&spds->spds_alg_lock);
351 	nd_free(&spds->spds_g_nd);
352 	kmem_free(spds->spds_params, sizeof (lcl_param_arr));
353 	spds->spds_params = NULL;
354 
355 	kmem_free(spds, sizeof (*spds));
356 }
357 
358 /*
359  * NOTE: large quantities of this should be shared with keysock.
360  * Would be nice to combine some of this into a common module, but
361  * not possible given time pressures.
362  */
363 
364 /*
365  * High-level reality checking of extensions.
366  */
367 /* ARGSUSED */ /* XXX */
368 static boolean_t
369 ext_check(spd_ext_t *ext)
370 {
371 	spd_if_t *tunname = (spd_if_t *)ext;
372 	int i;
373 	char *idstr;
374 
375 	if (ext->spd_ext_type == SPD_EXT_TUN_NAME) {
376 		/* (NOTE:  Modified from SADB_EXT_IDENTITY..) */
377 
378 		/*
379 		 * Make sure the strings in these identities are
380 		 * null-terminated.  Let's "proactively" null-terminate the
381 		 * string at the last byte if it's not terminated sooner.
382 		 */
383 		i = SPD_64TO8(tunname->spd_if_len) - sizeof (spd_if_t);
384 		idstr = (char *)(tunname + 1);
385 		while (*idstr != '\0' && i > 0) {
386 			i--;
387 			idstr++;
388 		}
389 		if (i == 0) {
390 			/*
391 			 * I.e., if the bozo user didn't NULL-terminate the
392 			 * string...
393 			 */
394 			idstr--;
395 			*idstr = '\0';
396 		}
397 	}
398 	return (B_TRUE);	/* For now... */
399 }
400 
401 
402 
403 /* Return values for spdsock_get_ext(). */
404 #define	KGE_OK	0
405 #define	KGE_DUP	1
406 #define	KGE_UNK	2
407 #define	KGE_LEN	3
408 #define	KGE_CHK	4
409 
410 /*
411  * Parse basic extension headers and return in the passed-in pointer vector.
412  * Return values include:
413  *
414  *	KGE_OK	Everything's nice and parsed out.
415  *		If there are no extensions, place NULL in extv[0].
416  *	KGE_DUP	There is a duplicate extension.
417  *		First instance in appropriate bin.  First duplicate in
418  *		extv[0].
419  *	KGE_UNK	Unknown extension type encountered.  extv[0] contains
420  *		unknown header.
421  *	KGE_LEN	Extension length error.
422  *	KGE_CHK	High-level reality check failed on specific extension.
423  *
424  * My apologies for some of the pointer arithmetic in here.  I'm thinking
425  * like an assembly programmer, yet trying to make the compiler happy.
426  */
427 static int
428 spdsock_get_ext(spd_ext_t *extv[], spd_msg_t *basehdr, uint_t msgsize)
429 {
430 	bzero(extv, sizeof (spd_ext_t *) * (SPD_EXT_MAX + 1));
431 
432 	/* Use extv[0] as the "current working pointer". */
433 
434 	extv[0] = (spd_ext_t *)(basehdr + 1);
435 
436 	while (extv[0] < (spd_ext_t *)(((uint8_t *)basehdr) + msgsize)) {
437 		/* Check for unknown headers. */
438 		if (extv[0]->spd_ext_type == 0 ||
439 		    extv[0]->spd_ext_type > SPD_EXT_MAX)
440 			return (KGE_UNK);
441 
442 		/*
443 		 * Check length.  Use uint64_t because extlen is in units
444 		 * of 64-bit words.  If length goes beyond the msgsize,
445 		 * return an error.  (Zero length also qualifies here.)
446 		 */
447 		if (extv[0]->spd_ext_len == 0 ||
448 		    (void *)((uint64_t *)extv[0] + extv[0]->spd_ext_len) >
449 		    (void *)((uint8_t *)basehdr + msgsize))
450 			return (KGE_LEN);
451 
452 		/* Check for redundant headers. */
453 		if (extv[extv[0]->spd_ext_type] != NULL)
454 			return (KGE_DUP);
455 
456 		/*
457 		 * Reality check the extension if possible at the spdsock
458 		 * level.
459 		 */
460 		if (!ext_check(extv[0]))
461 			return (KGE_CHK);
462 
463 		/* If I make it here, assign the appropriate bin. */
464 		extv[extv[0]->spd_ext_type] = extv[0];
465 
466 		/* Advance pointer (See above for uint64_t ptr reasoning.) */
467 		extv[0] = (spd_ext_t *)
468 		    ((uint64_t *)extv[0] + extv[0]->spd_ext_len);
469 	}
470 
471 	/* Everything's cool. */
472 
473 	/*
474 	 * If extv[0] == NULL, then there are no extension headers in this
475 	 * message.  Ensure that this is the case.
476 	 */
477 	if (extv[0] == (spd_ext_t *)(basehdr + 1))
478 		extv[0] = NULL;
479 
480 	return (KGE_OK);
481 }
482 
483 static const int bad_ext_diag[] = {
484 	SPD_DIAGNOSTIC_MALFORMED_LCLPORT,
485 	SPD_DIAGNOSTIC_MALFORMED_REMPORT,
486 	SPD_DIAGNOSTIC_MALFORMED_PROTO,
487 	SPD_DIAGNOSTIC_MALFORMED_LCLADDR,
488 	SPD_DIAGNOSTIC_MALFORMED_REMADDR,
489 	SPD_DIAGNOSTIC_MALFORMED_ACTION,
490 	SPD_DIAGNOSTIC_MALFORMED_RULE,
491 	SPD_DIAGNOSTIC_MALFORMED_RULESET,
492 	SPD_DIAGNOSTIC_MALFORMED_ICMP_TYPECODE
493 };
494 
495 static const int dup_ext_diag[] = {
496 	SPD_DIAGNOSTIC_DUPLICATE_LCLPORT,
497 	SPD_DIAGNOSTIC_DUPLICATE_REMPORT,
498 	SPD_DIAGNOSTIC_DUPLICATE_PROTO,
499 	SPD_DIAGNOSTIC_DUPLICATE_LCLADDR,
500 	SPD_DIAGNOSTIC_DUPLICATE_REMADDR,
501 	SPD_DIAGNOSTIC_DUPLICATE_ACTION,
502 	SPD_DIAGNOSTIC_DUPLICATE_RULE,
503 	SPD_DIAGNOSTIC_DUPLICATE_RULESET,
504 	SPD_DIAGNOSTIC_DUPLICATE_ICMP_TYPECODE
505 };
506 
507 /*
508  * Transmit a PF_POLICY error message to the instance either pointed to
509  * by ks, the instance with serial number serial, or more, depending.
510  *
511  * The faulty message (or a reasonable facsimile thereof) is in mp.
512  * This function will free mp or recycle it for delivery, thereby causing
513  * the stream head to free it.
514  */
515 static void
516 spdsock_error(queue_t *q, mblk_t *mp, int error, int diagnostic)
517 {
518 	spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
519 
520 	ASSERT(mp->b_datap->db_type == M_DATA);
521 
522 	if (spmsg->spd_msg_type < SPD_MIN ||
523 	    spmsg->spd_msg_type > SPD_MAX)
524 		spmsg->spd_msg_type = SPD_RESERVED;
525 
526 	/*
527 	 * Strip out extension headers.
528 	 */
529 	ASSERT(mp->b_rptr + sizeof (*spmsg) <= mp->b_datap->db_lim);
530 	mp->b_wptr = mp->b_rptr + sizeof (*spmsg);
531 	spmsg->spd_msg_len = SPD_8TO64(sizeof (spd_msg_t));
532 	spmsg->spd_msg_errno = (uint8_t)error;
533 	spmsg->spd_msg_diagnostic = (uint16_t)diagnostic;
534 
535 	qreply(q, mp);
536 }
537 
538 static void
539 spdsock_diag(queue_t *q, mblk_t *mp, int diagnostic)
540 {
541 	spdsock_error(q, mp, EINVAL, diagnostic);
542 }
543 
544 static void
545 spd_echo(queue_t *q, mblk_t *mp)
546 {
547 	qreply(q, mp);
548 }
549 
550 /*
551  * Do NOT consume a reference to itp.
552  */
553 /*ARGSUSED*/
554 static void
555 spdsock_flush_node(ipsec_tun_pol_t *itp, void *cookie, netstack_t *ns)
556 {
557 	boolean_t active = (boolean_t)cookie;
558 	ipsec_policy_head_t *iph;
559 
560 	iph = active ? itp->itp_policy : itp->itp_inactive;
561 	IPPH_REFHOLD(iph);
562 	mutex_enter(&itp->itp_lock);
563 	spdsock_flush_one(iph, ns);
564 	if (active)
565 		itp->itp_flags &= ~ITPF_PFLAGS;
566 	else
567 		itp->itp_flags &= ~ITPF_IFLAGS;
568 	mutex_exit(&itp->itp_lock);
569 }
570 
571 /*
572  * Clear out one polhead.
573  */
574 static void
575 spdsock_flush_one(ipsec_policy_head_t *iph, netstack_t *ns)
576 {
577 	rw_enter(&iph->iph_lock, RW_WRITER);
578 	ipsec_polhead_flush(iph, ns);
579 	rw_exit(&iph->iph_lock);
580 	IPPH_REFRELE(iph, ns);
581 }
582 
583 static void
584 spdsock_flush(queue_t *q, ipsec_policy_head_t *iph, ipsec_tun_pol_t *itp,
585     mblk_t *mp)
586 {
587 	boolean_t active;
588 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
589 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
590 
591 	if (iph != ALL_ACTIVE_POLHEADS && iph != ALL_INACTIVE_POLHEADS) {
592 		spdsock_flush_one(iph, ns);
593 		if (audit_active) {
594 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
595 
596 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
597 			audit_pf_policy(SPD_FLUSH, DB_CRED(mp), ns,
598 			    ITP_NAME(itp), active, 0, DB_CPID(mp));
599 		}
600 	} else {
601 		active = (iph == ALL_ACTIVE_POLHEADS);
602 
603 		/* First flush the global policy. */
604 		spdsock_flush_one(active ? ipsec_system_policy(ns) :
605 		    ipsec_inactive_policy(ns), ns);
606 		if (audit_active) {
607 			audit_pf_policy(SPD_FLUSH, DB_CRED(mp), ns, NULL,
608 			    active, 0, DB_CPID(mp));
609 		}
610 		/* Then flush every tunnel's appropriate one. */
611 		itp_walk(spdsock_flush_node, (void *)active, ns);
612 		if (audit_active)
613 			audit_pf_policy(SPD_FLUSH, DB_CRED(mp), ns,
614 			    "all tunnels", active, 0, DB_CPID(mp));
615 	}
616 
617 	spd_echo(q, mp);
618 }
619 
620 static boolean_t
621 spdsock_ext_to_sel(spd_ext_t **extv, ipsec_selkey_t *sel, int *diag)
622 {
623 	bzero(sel, sizeof (*sel));
624 
625 	if (extv[SPD_EXT_PROTO] != NULL) {
626 		struct spd_proto *pr =
627 		    (struct spd_proto *)extv[SPD_EXT_PROTO];
628 		sel->ipsl_proto = pr->spd_proto_number;
629 		sel->ipsl_valid |= IPSL_PROTOCOL;
630 	}
631 	if (extv[SPD_EXT_LCLPORT] != NULL) {
632 		struct spd_portrange *pr =
633 		    (struct spd_portrange *)extv[SPD_EXT_LCLPORT];
634 		sel->ipsl_lport = pr->spd_ports_minport;
635 		sel->ipsl_valid |= IPSL_LOCAL_PORT;
636 	}
637 	if (extv[SPD_EXT_REMPORT] != NULL) {
638 		struct spd_portrange *pr =
639 		    (struct spd_portrange *)extv[SPD_EXT_REMPORT];
640 		sel->ipsl_rport = pr->spd_ports_minport;
641 		sel->ipsl_valid |= IPSL_REMOTE_PORT;
642 	}
643 
644 	if (extv[SPD_EXT_ICMP_TYPECODE] != NULL) {
645 		struct spd_typecode *tc=
646 		    (struct spd_typecode *)extv[SPD_EXT_ICMP_TYPECODE];
647 
648 		sel->ipsl_valid |= IPSL_ICMP_TYPE;
649 		sel->ipsl_icmp_type = tc->spd_typecode_type;
650 		if (tc->spd_typecode_type_end < tc->spd_typecode_type)
651 			sel->ipsl_icmp_type_end = tc->spd_typecode_type;
652 		else
653 			sel->ipsl_icmp_type_end = tc->spd_typecode_type_end;
654 
655 		if (tc->spd_typecode_code != 255) {
656 			sel->ipsl_valid |= IPSL_ICMP_CODE;
657 			sel->ipsl_icmp_code = tc->spd_typecode_code;
658 			if (tc->spd_typecode_code_end < tc->spd_typecode_code)
659 				sel->ipsl_icmp_code_end = tc->spd_typecode_code;
660 			else
661 				sel->ipsl_icmp_code_end =
662 				    tc->spd_typecode_code_end;
663 		}
664 	}
665 #define	ADDR2SEL(sel, extv, field, pfield, extn, bit)			      \
666 	if ((extv)[(extn)] != NULL) {					      \
667 		uint_t addrlen;						      \
668 		struct spd_address *ap = 				      \
669 			(struct spd_address *)((extv)[(extn)]); 	      \
670 		addrlen = (ap->spd_address_af == AF_INET6) ? 		      \
671 			IPV6_ADDR_LEN : IP_ADDR_LEN;			      \
672 		if (SPD_64TO8(ap->spd_address_len) < 			      \
673 			(addrlen + sizeof (*ap))) {			      \
674 			*diag = SPD_DIAGNOSTIC_BAD_ADDR_LEN;		      \
675 			return (B_FALSE);				      \
676 		}							      \
677 		bcopy((ap+1), &((sel)->field), addrlen);		      \
678 		(sel)->pfield = ap->spd_address_prefixlen;		      \
679 		(sel)->ipsl_valid |= (bit);				      \
680 		(sel)->ipsl_valid |= (ap->spd_address_af == AF_INET6) ?	      \
681 			IPSL_IPV6 : IPSL_IPV4;				      \
682 	}
683 
684 	ADDR2SEL(sel, extv, ipsl_local, ipsl_local_pfxlen,
685 	    SPD_EXT_LCLADDR, IPSL_LOCAL_ADDR);
686 	ADDR2SEL(sel, extv, ipsl_remote, ipsl_remote_pfxlen,
687 	    SPD_EXT_REMADDR, IPSL_REMOTE_ADDR);
688 
689 	if ((sel->ipsl_valid & (IPSL_IPV6|IPSL_IPV4)) ==
690 	    (IPSL_IPV6|IPSL_IPV4)) {
691 		*diag = SPD_DIAGNOSTIC_MIXED_AF;
692 		return (B_FALSE);
693 	}
694 
695 #undef ADDR2SEL
696 
697 	return (B_TRUE);
698 }
699 
700 static boolean_t
701 spd_convert_type(uint32_t type, ipsec_act_t *act)
702 {
703 	switch (type) {
704 	case SPD_ACTTYPE_DROP:
705 		act->ipa_type = IPSEC_ACT_DISCARD;
706 		return (B_TRUE);
707 
708 	case SPD_ACTTYPE_PASS:
709 		act->ipa_type = IPSEC_ACT_CLEAR;
710 		return (B_TRUE);
711 
712 	case SPD_ACTTYPE_IPSEC:
713 		act->ipa_type = IPSEC_ACT_APPLY;
714 		return (B_TRUE);
715 	}
716 	return (B_FALSE);
717 }
718 
719 static boolean_t
720 spd_convert_flags(uint32_t flags, ipsec_act_t *act)
721 {
722 	/*
723 	 * Note use of !! for boolean canonicalization.
724 	 */
725 	act->ipa_apply.ipp_use_ah = !!(flags & SPD_APPLY_AH);
726 	act->ipa_apply.ipp_use_esp = !!(flags & SPD_APPLY_ESP);
727 	act->ipa_apply.ipp_use_espa = !!(flags & SPD_APPLY_ESPA);
728 	act->ipa_apply.ipp_use_se = !!(flags & SPD_APPLY_SE);
729 	act->ipa_apply.ipp_use_unique = !!(flags & SPD_APPLY_UNIQUE);
730 	return (B_TRUE);
731 }
732 
733 static void
734 spdsock_reset_act(ipsec_act_t *act)
735 {
736 	bzero(act, sizeof (*act));
737 	act->ipa_apply.ipp_espe_maxbits = IPSEC_MAX_KEYBITS;
738 	act->ipa_apply.ipp_espa_maxbits = IPSEC_MAX_KEYBITS;
739 	act->ipa_apply.ipp_ah_maxbits = IPSEC_MAX_KEYBITS;
740 }
741 
742 /*
743  * Sanity check action against reality, and shrink-wrap key sizes..
744  */
745 static boolean_t
746 spdsock_check_action(ipsec_act_t *act, boolean_t tunnel_polhead, int *diag,
747     spd_stack_t *spds)
748 {
749 	if (tunnel_polhead && act->ipa_apply.ipp_use_unique) {
750 		*diag = SPD_DIAGNOSTIC_ADD_INCON_FLAGS;
751 		return (B_FALSE);
752 	}
753 	if ((act->ipa_type != IPSEC_ACT_APPLY) &&
754 	    (act->ipa_apply.ipp_use_ah ||
755 	    act->ipa_apply.ipp_use_esp ||
756 	    act->ipa_apply.ipp_use_espa ||
757 	    act->ipa_apply.ipp_use_se ||
758 	    act->ipa_apply.ipp_use_unique)) {
759 		*diag = SPD_DIAGNOSTIC_ADD_INCON_FLAGS;
760 		return (B_FALSE);
761 	}
762 	if ((act->ipa_type == IPSEC_ACT_APPLY) &&
763 	    !act->ipa_apply.ipp_use_ah &&
764 	    !act->ipa_apply.ipp_use_esp) {
765 		*diag = SPD_DIAGNOSTIC_ADD_INCON_FLAGS;
766 		return (B_FALSE);
767 	}
768 	return (ipsec_check_action(act, diag, spds->spds_netstack));
769 }
770 
771 /*
772  * We may be short a few error checks here..
773  */
774 static boolean_t
775 spdsock_ext_to_actvec(spd_ext_t **extv, ipsec_act_t **actpp, uint_t *nactp,
776     int *diag, spd_stack_t *spds)
777 {
778 	struct spd_ext_actions *sactp =
779 	    (struct spd_ext_actions *)extv[SPD_EXT_ACTION];
780 	ipsec_act_t act, *actp, *endactp;
781 	struct spd_attribute *attrp, *endattrp;
782 	uint64_t *endp;
783 	int nact;
784 	boolean_t tunnel_polhead;
785 
786 	tunnel_polhead = (extv[SPD_EXT_TUN_NAME] != NULL &&
787 	    (((struct spd_rule *)extv[SPD_EXT_RULE])->spd_rule_flags &
788 	    SPD_RULE_FLAG_TUNNEL));
789 
790 	*actpp = NULL;
791 	*nactp = 0;
792 
793 	if (sactp == NULL) {
794 		*diag = SPD_DIAGNOSTIC_NO_ACTION_EXT;
795 		return (B_FALSE);
796 	}
797 
798 	/*
799 	 * Parse the "action" extension and convert into an action chain.
800 	 */
801 
802 	nact = sactp->spd_actions_count;
803 
804 	endp = (uint64_t *)sactp;
805 	endp += sactp->spd_actions_len;
806 	endattrp = (struct spd_attribute *)endp;
807 
808 	actp = kmem_alloc(sizeof (*actp) * nact, KM_NOSLEEP);
809 	if (actp == NULL) {
810 		*diag = SPD_DIAGNOSTIC_ADD_NO_MEM;
811 		return (B_FALSE);
812 	}
813 	*actpp = actp;
814 	*nactp = nact;
815 	endactp = actp + nact;
816 
817 	spdsock_reset_act(&act);
818 	attrp = (struct spd_attribute *)(&sactp[1]);
819 
820 	for (; attrp < endattrp; attrp++) {
821 		switch (attrp->spd_attr_tag) {
822 		case SPD_ATTR_NOP:
823 			break;
824 
825 		case SPD_ATTR_EMPTY:
826 			spdsock_reset_act(&act);
827 			break;
828 
829 		case SPD_ATTR_END:
830 			attrp = endattrp;
831 			/* FALLTHRU */
832 		case SPD_ATTR_NEXT:
833 			if (actp >= endactp) {
834 				*diag = SPD_DIAGNOSTIC_ADD_WRONG_ACT_COUNT;
835 				goto fail;
836 			}
837 			if (!spdsock_check_action(&act, tunnel_polhead,
838 			    diag, spds))
839 				goto fail;
840 			*actp++ = act;
841 			spdsock_reset_act(&act);
842 			break;
843 
844 		case SPD_ATTR_TYPE:
845 			if (!spd_convert_type(attrp->spd_attr_value, &act)) {
846 				*diag = SPD_DIAGNOSTIC_ADD_BAD_TYPE;
847 				goto fail;
848 			}
849 			break;
850 
851 		case SPD_ATTR_FLAGS:
852 			if (!tunnel_polhead && extv[SPD_EXT_TUN_NAME] != NULL) {
853 				/*
854 				 * Set "sa unique" for transport-mode
855 				 * tunnels whether we want to or not.
856 				 */
857 				attrp->spd_attr_value |= SPD_APPLY_UNIQUE;
858 			}
859 			if (!spd_convert_flags(attrp->spd_attr_value, &act)) {
860 				*diag = SPD_DIAGNOSTIC_ADD_BAD_FLAGS;
861 				goto fail;
862 			}
863 			break;
864 
865 		case SPD_ATTR_AH_AUTH:
866 			if (attrp->spd_attr_value == 0) {
867 				*diag = SPD_DIAGNOSTIC_UNSUPP_AH_ALG;
868 				goto fail;
869 			}
870 			act.ipa_apply.ipp_auth_alg = attrp->spd_attr_value;
871 			break;
872 
873 		case SPD_ATTR_ESP_ENCR:
874 			if (attrp->spd_attr_value == 0) {
875 				*diag = SPD_DIAGNOSTIC_UNSUPP_ESP_ENCR_ALG;
876 				goto fail;
877 			}
878 			act.ipa_apply.ipp_encr_alg = attrp->spd_attr_value;
879 			break;
880 
881 		case SPD_ATTR_ESP_AUTH:
882 			if (attrp->spd_attr_value == 0) {
883 				*diag = SPD_DIAGNOSTIC_UNSUPP_ESP_AUTH_ALG;
884 				goto fail;
885 			}
886 			act.ipa_apply.ipp_esp_auth_alg = attrp->spd_attr_value;
887 			break;
888 
889 		case SPD_ATTR_ENCR_MINBITS:
890 			act.ipa_apply.ipp_espe_minbits = attrp->spd_attr_value;
891 			break;
892 
893 		case SPD_ATTR_ENCR_MAXBITS:
894 			act.ipa_apply.ipp_espe_maxbits = attrp->spd_attr_value;
895 			break;
896 
897 		case SPD_ATTR_AH_MINBITS:
898 			act.ipa_apply.ipp_ah_minbits = attrp->spd_attr_value;
899 			break;
900 
901 		case SPD_ATTR_AH_MAXBITS:
902 			act.ipa_apply.ipp_ah_maxbits = attrp->spd_attr_value;
903 			break;
904 
905 		case SPD_ATTR_ESPA_MINBITS:
906 			act.ipa_apply.ipp_espa_minbits = attrp->spd_attr_value;
907 			break;
908 
909 		case SPD_ATTR_ESPA_MAXBITS:
910 			act.ipa_apply.ipp_espa_maxbits = attrp->spd_attr_value;
911 			break;
912 
913 		case SPD_ATTR_LIFE_SOFT_TIME:
914 		case SPD_ATTR_LIFE_HARD_TIME:
915 		case SPD_ATTR_LIFE_SOFT_BYTES:
916 		case SPD_ATTR_LIFE_HARD_BYTES:
917 			break;
918 
919 		case SPD_ATTR_KM_PROTO:
920 			act.ipa_apply.ipp_km_proto = attrp->spd_attr_value;
921 			break;
922 
923 		case SPD_ATTR_KM_COOKIE:
924 			act.ipa_apply.ipp_km_cookie = attrp->spd_attr_value;
925 			break;
926 
927 		case SPD_ATTR_REPLAY_DEPTH:
928 			act.ipa_apply.ipp_replay_depth = attrp->spd_attr_value;
929 			break;
930 		}
931 	}
932 	if (actp != endactp) {
933 		*diag = SPD_DIAGNOSTIC_ADD_WRONG_ACT_COUNT;
934 		goto fail;
935 	}
936 
937 	return (B_TRUE);
938 fail:
939 	ipsec_actvec_free(*actpp, nact);
940 	*actpp = NULL;
941 	return (B_FALSE);
942 }
943 
944 typedef struct
945 {
946 	ipsec_policy_t *pol;
947 	int dir;
948 } tmprule_t;
949 
950 static int
951 mkrule(ipsec_policy_head_t *iph, struct spd_rule *rule,
952     ipsec_selkey_t *sel, ipsec_act_t *actp, int nact, uint_t dir, uint_t af,
953     tmprule_t **rp, uint64_t *index, spd_stack_t *spds)
954 {
955 	ipsec_policy_t *pol;
956 
957 	sel->ipsl_valid &= ~(IPSL_IPV6|IPSL_IPV4);
958 	sel->ipsl_valid |= af;
959 
960 	pol = ipsec_policy_create(sel, actp, nact, rule->spd_rule_priority,
961 	    index, spds->spds_netstack);
962 	if (pol == NULL)
963 		return (ENOMEM);
964 
965 	(*rp)->pol = pol;
966 	(*rp)->dir = dir;
967 	(*rp)++;
968 
969 	if (!ipsec_check_policy(iph, pol, dir))
970 		return (EEXIST);
971 
972 	rule->spd_rule_index = pol->ipsp_index;
973 	return (0);
974 }
975 
976 static int
977 mkrulepair(ipsec_policy_head_t *iph, struct spd_rule *rule,
978     ipsec_selkey_t *sel, ipsec_act_t *actp, int nact, uint_t dir, uint_t afs,
979     tmprule_t **rp, uint64_t *index, spd_stack_t *spds)
980 {
981 	int error;
982 
983 	if (afs & IPSL_IPV4) {
984 		error = mkrule(iph, rule, sel, actp, nact, dir, IPSL_IPV4, rp,
985 		    index, spds);
986 		if (error != 0)
987 			return (error);
988 	}
989 	if (afs & IPSL_IPV6) {
990 		error = mkrule(iph, rule, sel, actp, nact, dir, IPSL_IPV6, rp,
991 		    index, spds);
992 		if (error != 0)
993 			return (error);
994 	}
995 	return (0);
996 }
997 
998 
999 static void
1000 spdsock_addrule(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp,
1001     spd_ext_t **extv, ipsec_tun_pol_t *itp)
1002 {
1003 	ipsec_selkey_t sel;
1004 	ipsec_act_t *actp;
1005 	uint_t nact;
1006 	int diag = 0, error, afs;
1007 	struct spd_rule *rule = (struct spd_rule *)extv[SPD_EXT_RULE];
1008 	tmprule_t rules[4], *rulep = &rules[0];
1009 	boolean_t tunnel_mode, empty_itp, active;
1010 	uint64_t *index = (itp == NULL) ? NULL : &itp->itp_next_policy_index;
1011 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
1012 	spd_stack_t	*spds = ss->spdsock_spds;
1013 
1014 	if (rule == NULL) {
1015 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_NO_RULE_EXT);
1016 		if (audit_active) {
1017 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1018 
1019 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1020 			audit_pf_policy(SPD_ADDRULE, DB_CRED(mp),
1021 			    spds->spds_netstack, ITP_NAME(itp), active,
1022 			    SPD_DIAGNOSTIC_NO_RULE_EXT, DB_CPID(mp));
1023 		}
1024 		return;
1025 	}
1026 
1027 	tunnel_mode = (rule->spd_rule_flags & SPD_RULE_FLAG_TUNNEL);
1028 
1029 	if (itp != NULL) {
1030 		mutex_enter(&itp->itp_lock);
1031 		ASSERT(itp->itp_policy == iph || itp->itp_inactive == iph);
1032 		active = (itp->itp_policy == iph);
1033 		if (ITP_P_ISACTIVE(itp, iph)) {
1034 			/* Check for mix-and-match of tunnel/transport. */
1035 			if ((tunnel_mode && !ITP_P_ISTUNNEL(itp, iph)) ||
1036 			    (!tunnel_mode && ITP_P_ISTUNNEL(itp, iph))) {
1037 				mutex_exit(&itp->itp_lock);
1038 				spdsock_error(q, mp, EBUSY, 0);
1039 				return;
1040 			}
1041 			empty_itp = B_FALSE;
1042 		} else {
1043 			empty_itp = B_TRUE;
1044 			itp->itp_flags = active ? ITPF_P_ACTIVE : ITPF_I_ACTIVE;
1045 			if (tunnel_mode)
1046 				itp->itp_flags |= active ? ITPF_P_TUNNEL :
1047 				    ITPF_I_TUNNEL;
1048 		}
1049 	} else {
1050 		empty_itp = B_FALSE;
1051 	}
1052 
1053 	if (rule->spd_rule_index != 0) {
1054 		diag = SPD_DIAGNOSTIC_INVALID_RULE_INDEX;
1055 		error = EINVAL;
1056 		goto fail2;
1057 	}
1058 
1059 	if (!spdsock_ext_to_sel(extv, &sel, &diag)) {
1060 		error = EINVAL;
1061 		goto fail2;
1062 	}
1063 
1064 	if (itp != NULL) {
1065 		if (tunnel_mode) {
1066 			if (sel.ipsl_valid &
1067 			    (IPSL_REMOTE_PORT | IPSL_LOCAL_PORT)) {
1068 				itp->itp_flags |= active ?
1069 				    ITPF_P_PER_PORT_SECURITY :
1070 				    ITPF_I_PER_PORT_SECURITY;
1071 			}
1072 		} else {
1073 			/*
1074 			 * For now, we don't allow transport-mode on a tunnel
1075 			 * with ANY specific selectors.  Bail if we have such
1076 			 * a request.
1077 			 */
1078 			if (sel.ipsl_valid & IPSL_WILDCARD) {
1079 				diag = SPD_DIAGNOSTIC_NO_TUNNEL_SELECTORS;
1080 				error = EINVAL;
1081 				goto fail2;
1082 			}
1083 		}
1084 	}
1085 
1086 	if (!spdsock_ext_to_actvec(extv, &actp, &nact, &diag, spds)) {
1087 		error = EINVAL;
1088 		goto fail2;
1089 	}
1090 	/*
1091 	 * If no addresses were specified, add both.
1092 	 */
1093 	afs = sel.ipsl_valid & (IPSL_IPV6|IPSL_IPV4);
1094 	if (afs == 0)
1095 		afs = (IPSL_IPV6|IPSL_IPV4);
1096 
1097 	rw_enter(&iph->iph_lock, RW_WRITER);
1098 
1099 	if (rule->spd_rule_flags & SPD_RULE_FLAG_OUTBOUND) {
1100 		error = mkrulepair(iph, rule, &sel, actp, nact,
1101 		    IPSEC_TYPE_OUTBOUND, afs, &rulep, index, spds);
1102 		if (error != 0)
1103 			goto fail;
1104 	}
1105 
1106 	if (rule->spd_rule_flags & SPD_RULE_FLAG_INBOUND) {
1107 		error = mkrulepair(iph, rule, &sel, actp, nact,
1108 		    IPSEC_TYPE_INBOUND, afs, &rulep, index, spds);
1109 		if (error != 0)
1110 			goto fail;
1111 	}
1112 
1113 	while ((--rulep) >= &rules[0]) {
1114 		ipsec_enter_policy(iph, rulep->pol, rulep->dir,
1115 		    spds->spds_netstack);
1116 	}
1117 	rw_exit(&iph->iph_lock);
1118 	if (itp != NULL)
1119 		mutex_exit(&itp->itp_lock);
1120 
1121 	ipsec_actvec_free(actp, nact);
1122 	spd_echo(q, mp);
1123 	if (audit_active) {
1124 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1125 
1126 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1127 		audit_pf_policy(SPD_ADDRULE, DB_CRED(mp), spds->spds_netstack,
1128 		    ITP_NAME(itp), active, 0, DB_CPID(mp));
1129 	}
1130 	return;
1131 
1132 fail:
1133 	rw_exit(&iph->iph_lock);
1134 	while ((--rulep) >= &rules[0]) {
1135 		IPPOL_REFRELE(rulep->pol, spds->spds_netstack);
1136 	}
1137 	ipsec_actvec_free(actp, nact);
1138 fail2:
1139 	if (itp != NULL) {
1140 		if (empty_itp)
1141 			itp->itp_flags = 0;
1142 		mutex_exit(&itp->itp_lock);
1143 	}
1144 	spdsock_error(q, mp, error, diag);
1145 	if (audit_active) {
1146 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1147 
1148 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1149 		audit_pf_policy(SPD_ADDRULE, DB_CRED(mp), spds->spds_netstack,
1150 		    ITP_NAME(itp), active, error, DB_CPID(mp));
1151 	}
1152 }
1153 
1154 void
1155 spdsock_deleterule(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp,
1156     spd_ext_t **extv, ipsec_tun_pol_t *itp)
1157 {
1158 	ipsec_selkey_t sel;
1159 	struct spd_rule *rule = (struct spd_rule *)extv[SPD_EXT_RULE];
1160 	int err, diag = 0;
1161 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
1162 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
1163 
1164 	if (rule == NULL) {
1165 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_NO_RULE_EXT);
1166 		if (audit_active) {
1167 			boolean_t active;
1168 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1169 
1170 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1171 			audit_pf_policy(SPD_DELETERULE, DB_CRED(mp), ns,
1172 			    ITP_NAME(itp), active, SPD_DIAGNOSTIC_NO_RULE_EXT,
1173 			    DB_CPID(mp));
1174 		}
1175 		return;
1176 	}
1177 
1178 	/*
1179 	 * Must enter itp_lock first to avoid deadlock.  See tun.c's
1180 	 * set_sec_simple() for the other case of itp_lock and iph_lock.
1181 	 */
1182 	if (itp != NULL)
1183 		mutex_enter(&itp->itp_lock);
1184 
1185 	if (rule->spd_rule_index != 0) {
1186 		if (ipsec_policy_delete_index(iph, rule->spd_rule_index, ns) !=
1187 		    0) {
1188 			err = ESRCH;
1189 			goto fail;
1190 		}
1191 	} else {
1192 		if (!spdsock_ext_to_sel(extv, &sel, &diag)) {
1193 			err = EINVAL;	/* diag already set... */
1194 			goto fail;
1195 		}
1196 
1197 		if ((rule->spd_rule_flags & SPD_RULE_FLAG_INBOUND) &&
1198 		    !ipsec_policy_delete(iph, &sel, IPSEC_TYPE_INBOUND, ns)) {
1199 			err = ESRCH;
1200 			goto fail;
1201 		}
1202 
1203 		if ((rule->spd_rule_flags & SPD_RULE_FLAG_OUTBOUND) &&
1204 		    !ipsec_policy_delete(iph, &sel, IPSEC_TYPE_OUTBOUND, ns)) {
1205 			err = ESRCH;
1206 			goto fail;
1207 		}
1208 	}
1209 
1210 	if (itp != NULL) {
1211 		ASSERT(iph == itp->itp_policy || iph == itp->itp_inactive);
1212 		rw_enter(&iph->iph_lock, RW_READER);
1213 		if (avl_numnodes(&iph->iph_rulebyid) == 0) {
1214 			if (iph == itp->itp_policy)
1215 				itp->itp_flags &= ~ITPF_PFLAGS;
1216 			else
1217 				itp->itp_flags &= ~ITPF_IFLAGS;
1218 		}
1219 		/* Can exit locks in any order. */
1220 		rw_exit(&iph->iph_lock);
1221 		mutex_exit(&itp->itp_lock);
1222 	}
1223 	spd_echo(q, mp);
1224 	if (audit_active) {
1225 		boolean_t active;
1226 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1227 
1228 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1229 		audit_pf_policy(SPD_DELETERULE, DB_CRED(mp), ns, ITP_NAME(itp),
1230 		    active, 0, DB_CPID(mp));
1231 	}
1232 	return;
1233 fail:
1234 	if (itp != NULL)
1235 		mutex_exit(&itp->itp_lock);
1236 	spdsock_error(q, mp, err, diag);
1237 	if (audit_active) {
1238 		boolean_t active;
1239 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1240 
1241 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1242 		audit_pf_policy(SPD_DELETERULE, DB_CRED(mp), ns, ITP_NAME(itp),
1243 		    active, err, DB_CPID(mp));
1244 	}
1245 }
1246 
1247 /* Do NOT consume a reference to itp. */
1248 /* ARGSUSED */
1249 static void
1250 spdsock_flip_node(ipsec_tun_pol_t *itp, void *ignoreme, netstack_t *ns)
1251 {
1252 	mutex_enter(&itp->itp_lock);
1253 	ITPF_SWAP(itp->itp_flags);
1254 	ipsec_swap_policy(itp->itp_policy, itp->itp_inactive, ns);
1255 	mutex_exit(&itp->itp_lock);
1256 }
1257 
1258 void
1259 spdsock_flip(queue_t *q, mblk_t *mp, spd_if_t *tunname)
1260 {
1261 	char *tname;
1262 	ipsec_tun_pol_t *itp;
1263 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
1264 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
1265 
1266 	if (tunname != NULL) {
1267 		tname = (char *)tunname->spd_if_name;
1268 		if (*tname == '\0') {
1269 			/* can't fail */
1270 			ipsec_swap_global_policy(ns);
1271 			if (audit_active) {
1272 				boolean_t active;
1273 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1274 
1275 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1276 				audit_pf_policy(SPD_FLIP, DB_CRED(mp),
1277 				    ns, NULL, active, 0, DB_CPID(mp));
1278 			}
1279 			itp_walk(spdsock_flip_node, NULL, ns);
1280 			if (audit_active) {
1281 				boolean_t active;
1282 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1283 
1284 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1285 				audit_pf_policy(SPD_FLIP, DB_CRED(mp), ns,
1286 				    "all tunnels", active, 0, DB_CPID(mp));
1287 			}
1288 		} else {
1289 			itp = get_tunnel_policy(tname, ns);
1290 			if (itp == NULL) {
1291 				/* Better idea for "tunnel not found"? */
1292 				spdsock_error(q, mp, ESRCH, 0);
1293 				if (audit_active) {
1294 					boolean_t active;
1295 					spd_msg_t *spmsg =
1296 					    (spd_msg_t *)mp->b_rptr;
1297 
1298 					active = (spmsg->spd_msg_spdid ==
1299 					    SPD_ACTIVE);
1300 					audit_pf_policy(SPD_FLIP, DB_CRED(mp),
1301 					    ns, ITP_NAME(itp), active,
1302 					    ESRCH, DB_CPID(mp));
1303 				}
1304 				return;
1305 			}
1306 			spdsock_flip_node(itp, NULL, NULL);
1307 			if (audit_active) {
1308 				boolean_t active;
1309 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1310 
1311 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1312 				audit_pf_policy(SPD_FLIP, DB_CRED(mp), ns,
1313 				    ITP_NAME(itp), active, 0, DB_CPID(mp));
1314 			}
1315 			ITP_REFRELE(itp, ns);
1316 		}
1317 	} else {
1318 		ipsec_swap_global_policy(ns);	/* can't fail */
1319 		if (audit_active) {
1320 			boolean_t active;
1321 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1322 
1323 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1324 			audit_pf_policy(SPD_FLIP, DB_CRED(mp),
1325 			    ns, NULL, active, 0, DB_CPID(mp));
1326 		}
1327 	}
1328 	spd_echo(q, mp);
1329 }
1330 
1331 /*
1332  * Unimplemented feature
1333  */
1334 /* ARGSUSED */
1335 static void
1336 spdsock_lookup(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp,
1337     spd_ext_t **extv, ipsec_tun_pol_t *itp)
1338 {
1339 	spdsock_error(q, mp, EINVAL, 0);
1340 }
1341 
1342 
1343 static mblk_t *
1344 spdsock_dump_ruleset(mblk_t *req, ipsec_policy_head_t *iph,
1345     uint32_t count, uint16_t error)
1346 {
1347 	size_t len = sizeof (spd_ruleset_ext_t) + sizeof (spd_msg_t);
1348 	spd_msg_t *msg;
1349 	spd_ruleset_ext_t *ruleset;
1350 	mblk_t *m = allocb(len, BPRI_HI);
1351 
1352 	ASSERT(RW_READ_HELD(&iph->iph_lock));
1353 
1354 	if (m == NULL) {
1355 		return (NULL);
1356 	}
1357 	msg = (spd_msg_t *)m->b_rptr;
1358 	ruleset = (spd_ruleset_ext_t *)(&msg[1]);
1359 
1360 	m->b_wptr = (uint8_t *)&ruleset[1];
1361 
1362 	*msg = *(spd_msg_t *)(req->b_rptr);
1363 	msg->spd_msg_len = SPD_8TO64(len);
1364 	msg->spd_msg_errno = error;
1365 
1366 	ruleset->spd_ruleset_len = SPD_8TO64(sizeof (*ruleset));
1367 	ruleset->spd_ruleset_type = SPD_EXT_RULESET;
1368 	ruleset->spd_ruleset_count = count;
1369 	ruleset->spd_ruleset_version = iph->iph_gen;
1370 	return (m);
1371 }
1372 
1373 static mblk_t *
1374 spdsock_dump_finish(spdsock_t *ss, int error)
1375 {
1376 	mblk_t *m;
1377 	ipsec_policy_head_t *iph = ss->spdsock_dump_head;
1378 	mblk_t *req = ss->spdsock_dump_req;
1379 
1380 	rw_enter(&iph->iph_lock, RW_READER);
1381 	m = spdsock_dump_ruleset(req, iph, ss->spdsock_dump_count, error);
1382 	rw_exit(&iph->iph_lock);
1383 	IPPH_REFRELE(iph, ss->spdsock_spds->spds_netstack);
1384 	ss->spdsock_dump_req = NULL;
1385 	freemsg(req);
1386 
1387 	return (m);
1388 }
1389 
1390 /*
1391  * Rule encoding functions.
1392  * We do a two-pass encode.
1393  * If base != NULL, fill in encoded rule part starting at base+offset.
1394  * Always return "offset" plus length of to-be-encoded data.
1395  */
1396 static uint_t
1397 spdsock_encode_typecode(uint8_t *base, uint_t offset, uint8_t type,
1398     uint8_t type_end, uint8_t code, uint8_t code_end)
1399 {
1400 	struct spd_typecode *tcp;
1401 
1402 	ASSERT(ALIGNED64(offset));
1403 
1404 	if (base != NULL) {
1405 		tcp = (struct spd_typecode *)(base + offset);
1406 		tcp->spd_typecode_len = SPD_8TO64(sizeof (*tcp));
1407 		tcp->spd_typecode_exttype = SPD_EXT_ICMP_TYPECODE;
1408 		tcp->spd_typecode_code = code;
1409 		tcp->spd_typecode_type = type;
1410 		tcp->spd_typecode_type_end = type_end;
1411 		tcp->spd_typecode_code_end = code_end;
1412 	}
1413 	offset += sizeof (*tcp);
1414 
1415 	ASSERT(ALIGNED64(offset));
1416 
1417 	return (offset);
1418 }
1419 
1420 static uint_t
1421 spdsock_encode_proto(uint8_t *base, uint_t offset, uint8_t proto)
1422 {
1423 	struct spd_proto *spp;
1424 
1425 	ASSERT(ALIGNED64(offset));
1426 
1427 	if (base != NULL) {
1428 		spp = (struct spd_proto *)(base + offset);
1429 		spp->spd_proto_len = SPD_8TO64(sizeof (*spp));
1430 		spp->spd_proto_exttype = SPD_EXT_PROTO;
1431 		spp->spd_proto_number = proto;
1432 		spp->spd_proto_reserved1 = 0;
1433 		spp->spd_proto_reserved2 = 0;
1434 	}
1435 	offset += sizeof (*spp);
1436 
1437 	ASSERT(ALIGNED64(offset));
1438 
1439 	return (offset);
1440 }
1441 
1442 static uint_t
1443 spdsock_encode_port(uint8_t *base, uint_t offset, uint16_t ext, uint16_t port)
1444 {
1445 	struct spd_portrange *spp;
1446 
1447 	ASSERT(ALIGNED64(offset));
1448 
1449 	if (base != NULL) {
1450 		spp = (struct spd_portrange *)(base + offset);
1451 		spp->spd_ports_len = SPD_8TO64(sizeof (*spp));
1452 		spp->spd_ports_exttype = ext;
1453 		spp->spd_ports_minport = port;
1454 		spp->spd_ports_maxport = port;
1455 	}
1456 	offset += sizeof (*spp);
1457 
1458 	ASSERT(ALIGNED64(offset));
1459 
1460 	return (offset);
1461 }
1462 
1463 static uint_t
1464 spdsock_encode_addr(uint8_t *base, uint_t offset, uint16_t ext,
1465     const ipsec_selkey_t *sel, const ipsec_addr_t *addr, uint_t pfxlen)
1466 {
1467 	struct spd_address *sae;
1468 	ipsec_addr_t *spdaddr;
1469 	uint_t start = offset;
1470 	uint_t addrlen;
1471 	uint_t af;
1472 
1473 	if (sel->ipsl_valid & IPSL_IPV4) {
1474 		af = AF_INET;
1475 		addrlen = IP_ADDR_LEN;
1476 	} else {
1477 		af = AF_INET6;
1478 		addrlen = IPV6_ADDR_LEN;
1479 	}
1480 
1481 	ASSERT(ALIGNED64(offset));
1482 
1483 	if (base != NULL) {
1484 		sae = (struct spd_address *)(base + offset);
1485 		sae->spd_address_exttype = ext;
1486 		sae->spd_address_af = af;
1487 		sae->spd_address_prefixlen = pfxlen;
1488 		sae->spd_address_reserved2 = 0;
1489 
1490 		spdaddr = (ipsec_addr_t *)(&sae[1]);
1491 		bcopy(addr, spdaddr, addrlen);
1492 	}
1493 	offset += sizeof (*sae);
1494 	addrlen = roundup(addrlen, sizeof (uint64_t));
1495 	offset += addrlen;
1496 
1497 	ASSERT(ALIGNED64(offset));
1498 
1499 	if (base != NULL)
1500 		sae->spd_address_len = SPD_8TO64(offset - start);
1501 	return (offset);
1502 }
1503 
1504 static uint_t
1505 spdsock_encode_sel(uint8_t *base, uint_t offset, const ipsec_sel_t *sel)
1506 {
1507 	const ipsec_selkey_t *selkey = &sel->ipsl_key;
1508 
1509 	if (selkey->ipsl_valid & IPSL_PROTOCOL)
1510 		offset = spdsock_encode_proto(base, offset, selkey->ipsl_proto);
1511 	if (selkey->ipsl_valid & IPSL_LOCAL_PORT)
1512 		offset = spdsock_encode_port(base, offset, SPD_EXT_LCLPORT,
1513 		    selkey->ipsl_lport);
1514 	if (selkey->ipsl_valid & IPSL_REMOTE_PORT)
1515 		offset = spdsock_encode_port(base, offset, SPD_EXT_REMPORT,
1516 		    selkey->ipsl_rport);
1517 	if (selkey->ipsl_valid & IPSL_REMOTE_ADDR)
1518 		offset = spdsock_encode_addr(base, offset, SPD_EXT_REMADDR,
1519 		    selkey, &selkey->ipsl_remote, selkey->ipsl_remote_pfxlen);
1520 	if (selkey->ipsl_valid & IPSL_LOCAL_ADDR)
1521 		offset = spdsock_encode_addr(base, offset, SPD_EXT_LCLADDR,
1522 		    selkey, &selkey->ipsl_local, selkey->ipsl_local_pfxlen);
1523 	if (selkey->ipsl_valid & IPSL_ICMP_TYPE) {
1524 		offset = spdsock_encode_typecode(base, offset,
1525 		    selkey->ipsl_icmp_type, selkey->ipsl_icmp_type_end,
1526 		    (selkey->ipsl_valid & IPSL_ICMP_CODE) ?
1527 		    selkey->ipsl_icmp_code : 255,
1528 		    (selkey->ipsl_valid & IPSL_ICMP_CODE) ?
1529 		    selkey->ipsl_icmp_code_end : 255);
1530 	}
1531 	return (offset);
1532 }
1533 
1534 static uint_t
1535 spdsock_encode_actattr(uint8_t *base, uint_t offset, uint32_t tag,
1536     uint32_t value)
1537 {
1538 	struct spd_attribute *attr;
1539 
1540 	ASSERT(ALIGNED64(offset));
1541 
1542 	if (base != NULL) {
1543 		attr = (struct spd_attribute *)(base + offset);
1544 		attr->spd_attr_tag = tag;
1545 		attr->spd_attr_value = value;
1546 	}
1547 	offset += sizeof (struct spd_attribute);
1548 
1549 	ASSERT(ALIGNED64(offset));
1550 
1551 	return (offset);
1552 }
1553 
1554 
1555 #define	EMIT(t, v) offset = spdsock_encode_actattr(base, offset, (t), (v))
1556 
1557 static uint_t
1558 spdsock_encode_action(uint8_t *base, uint_t offset, const ipsec_action_t *ap)
1559 {
1560 	const struct ipsec_act *act = &(ap->ipa_act);
1561 	uint_t flags;
1562 
1563 	EMIT(SPD_ATTR_EMPTY, 0);
1564 	switch (act->ipa_type) {
1565 	case IPSEC_ACT_DISCARD:
1566 	case IPSEC_ACT_REJECT:
1567 		EMIT(SPD_ATTR_TYPE, SPD_ACTTYPE_DROP);
1568 		break;
1569 	case IPSEC_ACT_BYPASS:
1570 	case IPSEC_ACT_CLEAR:
1571 		EMIT(SPD_ATTR_TYPE, SPD_ACTTYPE_PASS);
1572 		break;
1573 
1574 	case IPSEC_ACT_APPLY:
1575 		EMIT(SPD_ATTR_TYPE, SPD_ACTTYPE_IPSEC);
1576 		flags = 0;
1577 		if (act->ipa_apply.ipp_use_ah)
1578 			flags |= SPD_APPLY_AH;
1579 		if (act->ipa_apply.ipp_use_esp)
1580 			flags |= SPD_APPLY_ESP;
1581 		if (act->ipa_apply.ipp_use_espa)
1582 			flags |= SPD_APPLY_ESPA;
1583 		if (act->ipa_apply.ipp_use_se)
1584 			flags |= SPD_APPLY_SE;
1585 		if (act->ipa_apply.ipp_use_unique)
1586 			flags |= SPD_APPLY_UNIQUE;
1587 		EMIT(SPD_ATTR_FLAGS, flags);
1588 		if (flags & SPD_APPLY_AH) {
1589 			EMIT(SPD_ATTR_AH_AUTH, act->ipa_apply.ipp_auth_alg);
1590 			EMIT(SPD_ATTR_AH_MINBITS,
1591 			    act->ipa_apply.ipp_ah_minbits);
1592 			EMIT(SPD_ATTR_AH_MAXBITS,
1593 			    act->ipa_apply.ipp_ah_maxbits);
1594 		}
1595 		if (flags & SPD_APPLY_ESP) {
1596 			EMIT(SPD_ATTR_ESP_ENCR, act->ipa_apply.ipp_encr_alg);
1597 			EMIT(SPD_ATTR_ENCR_MINBITS,
1598 			    act->ipa_apply.ipp_espe_minbits);
1599 			EMIT(SPD_ATTR_ENCR_MAXBITS,
1600 			    act->ipa_apply.ipp_espe_maxbits);
1601 			if (flags & SPD_APPLY_ESPA) {
1602 				EMIT(SPD_ATTR_ESP_AUTH,
1603 				    act->ipa_apply.ipp_esp_auth_alg);
1604 				EMIT(SPD_ATTR_ESPA_MINBITS,
1605 				    act->ipa_apply.ipp_espa_minbits);
1606 				EMIT(SPD_ATTR_ESPA_MAXBITS,
1607 				    act->ipa_apply.ipp_espa_maxbits);
1608 			}
1609 		}
1610 		if (act->ipa_apply.ipp_km_proto != 0)
1611 			EMIT(SPD_ATTR_KM_PROTO, act->ipa_apply.ipp_km_proto);
1612 		if (act->ipa_apply.ipp_km_cookie != 0)
1613 			EMIT(SPD_ATTR_KM_PROTO, act->ipa_apply.ipp_km_cookie);
1614 		if (act->ipa_apply.ipp_replay_depth != 0)
1615 			EMIT(SPD_ATTR_REPLAY_DEPTH,
1616 			    act->ipa_apply.ipp_replay_depth);
1617 		/* Add more here */
1618 		break;
1619 	}
1620 
1621 	return (offset);
1622 }
1623 
1624 static uint_t
1625 spdsock_encode_action_list(uint8_t *base, uint_t offset,
1626     const ipsec_action_t *ap)
1627 {
1628 	struct spd_ext_actions *act;
1629 	uint_t nact = 0;
1630 	uint_t start = offset;
1631 
1632 	ASSERT(ALIGNED64(offset));
1633 
1634 	if (base != NULL) {
1635 		act = (struct spd_ext_actions *)(base + offset);
1636 		act->spd_actions_len = 0;
1637 		act->spd_actions_exttype = SPD_EXT_ACTION;
1638 		act->spd_actions_count = 0;
1639 		act->spd_actions_reserved = 0;
1640 	}
1641 
1642 	offset += sizeof (*act);
1643 
1644 	ASSERT(ALIGNED64(offset));
1645 
1646 	while (ap != NULL) {
1647 		offset = spdsock_encode_action(base, offset, ap);
1648 		ap = ap->ipa_next;
1649 		nact++;
1650 		if (ap != NULL) {
1651 			EMIT(SPD_ATTR_NEXT, 0);
1652 		}
1653 	}
1654 	EMIT(SPD_ATTR_END, 0);
1655 
1656 	ASSERT(ALIGNED64(offset));
1657 
1658 	if (base != NULL) {
1659 		act->spd_actions_count = nact;
1660 		act->spd_actions_len = SPD_8TO64(offset - start);
1661 	}
1662 
1663 	return (offset);
1664 }
1665 
1666 #undef EMIT
1667 
1668 /* ARGSUSED */
1669 static uint_t
1670 spdsock_rule_flags(uint_t dir, uint_t af)
1671 {
1672 	uint_t flags = 0;
1673 
1674 	if (dir == IPSEC_TYPE_INBOUND)
1675 		flags |= SPD_RULE_FLAG_INBOUND;
1676 	if (dir == IPSEC_TYPE_OUTBOUND)
1677 		flags |= SPD_RULE_FLAG_OUTBOUND;
1678 
1679 	return (flags);
1680 }
1681 
1682 
1683 static uint_t
1684 spdsock_encode_rule_head(uint8_t *base, uint_t offset, spd_msg_t *req,
1685     const ipsec_policy_t *rule, uint_t dir, uint_t af, char *name,
1686     boolean_t tunnel)
1687 {
1688 	struct spd_msg *spmsg;
1689 	struct spd_rule *spr;
1690 	spd_if_t *sid;
1691 
1692 	uint_t start = offset;
1693 
1694 	ASSERT(ALIGNED64(offset));
1695 
1696 	if (base != NULL) {
1697 		spmsg = (struct spd_msg *)(base + offset);
1698 		bzero(spmsg, sizeof (*spmsg));
1699 		spmsg->spd_msg_version = PF_POLICY_V1;
1700 		spmsg->spd_msg_type = SPD_DUMP;
1701 		spmsg->spd_msg_seq = req->spd_msg_seq;
1702 		spmsg->spd_msg_pid = req->spd_msg_pid;
1703 	}
1704 	offset += sizeof (struct spd_msg);
1705 
1706 	ASSERT(ALIGNED64(offset));
1707 
1708 	if (base != NULL) {
1709 		spr = (struct spd_rule *)(base + offset);
1710 		spr->spd_rule_type = SPD_EXT_RULE;
1711 		spr->spd_rule_priority = rule->ipsp_prio;
1712 		spr->spd_rule_flags = spdsock_rule_flags(dir, af);
1713 		if (tunnel)
1714 			spr->spd_rule_flags |= SPD_RULE_FLAG_TUNNEL;
1715 		spr->spd_rule_unused = 0;
1716 		spr->spd_rule_len = SPD_8TO64(sizeof (*spr));
1717 		spr->spd_rule_index = rule->ipsp_index;
1718 	}
1719 	offset += sizeof (struct spd_rule);
1720 
1721 	/*
1722 	 * If we have an interface name (i.e. if this policy head came from
1723 	 * a tunnel), add the SPD_EXT_TUN_NAME extension.
1724 	 */
1725 	if (name != NULL) {
1726 
1727 		ASSERT(ALIGNED64(offset));
1728 
1729 		if (base != NULL) {
1730 			sid = (spd_if_t *)(base + offset);
1731 			sid->spd_if_exttype = SPD_EXT_TUN_NAME;
1732 			sid->spd_if_len = SPD_8TO64(sizeof (spd_if_t) +
1733 			    roundup((strlen(name) - 4), 8));
1734 			(void) strlcpy((char *)sid->spd_if_name, name,
1735 			    LIFNAMSIZ);
1736 		}
1737 
1738 		offset += sizeof (spd_if_t) + roundup((strlen(name) - 4), 8);
1739 	}
1740 
1741 	offset = spdsock_encode_sel(base, offset, rule->ipsp_sel);
1742 	offset = spdsock_encode_action_list(base, offset, rule->ipsp_act);
1743 
1744 	ASSERT(ALIGNED64(offset));
1745 
1746 	if (base != NULL) {
1747 		spmsg->spd_msg_len = SPD_8TO64(offset - start);
1748 	}
1749 	return (offset);
1750 }
1751 
1752 /* ARGSUSED */
1753 static mblk_t *
1754 spdsock_encode_rule(mblk_t *req, const ipsec_policy_t *rule,
1755     uint_t dir, uint_t af, char *name, boolean_t tunnel)
1756 {
1757 	mblk_t *m;
1758 	uint_t len;
1759 	spd_msg_t *mreq = (spd_msg_t *)req->b_rptr;
1760 
1761 	/*
1762 	 * Figure out how much space we'll need.
1763 	 */
1764 	len = spdsock_encode_rule_head(NULL, 0, mreq, rule, dir, af, name,
1765 	    tunnel);
1766 
1767 	/*
1768 	 * Allocate mblk.
1769 	 */
1770 	m = allocb(len, BPRI_HI);
1771 	if (m == NULL)
1772 		return (NULL);
1773 
1774 	/*
1775 	 * Fill it in..
1776 	 */
1777 	m->b_wptr = m->b_rptr + len;
1778 	bzero(m->b_rptr, len);
1779 	(void) spdsock_encode_rule_head(m->b_rptr, 0, mreq, rule, dir, af,
1780 	    name, tunnel);
1781 	return (m);
1782 }
1783 
1784 static ipsec_policy_t *
1785 spdsock_dump_next_in_chain(spdsock_t *ss, ipsec_policy_head_t *iph,
1786     ipsec_policy_t *cur)
1787 {
1788 	ASSERT(RW_READ_HELD(&iph->iph_lock));
1789 
1790 	ss->spdsock_dump_count++;
1791 	ss->spdsock_dump_cur_rule = cur->ipsp_hash.hash_next;
1792 	return (cur);
1793 }
1794 
1795 static ipsec_policy_t *
1796 spdsock_dump_next_rule(spdsock_t *ss, ipsec_policy_head_t *iph)
1797 {
1798 	ipsec_policy_t *cur;
1799 	ipsec_policy_root_t *ipr;
1800 	int chain, nchains, type, af;
1801 
1802 	ASSERT(RW_READ_HELD(&iph->iph_lock));
1803 
1804 	cur = ss->spdsock_dump_cur_rule;
1805 
1806 	if (cur != NULL)
1807 		return (spdsock_dump_next_in_chain(ss, iph, cur));
1808 
1809 	type = ss->spdsock_dump_cur_type;
1810 
1811 next:
1812 	chain = ss->spdsock_dump_cur_chain;
1813 	ipr = &iph->iph_root[type];
1814 	nchains = ipr->ipr_nchains;
1815 
1816 	while (chain < nchains) {
1817 		cur = ipr->ipr_hash[chain].hash_head;
1818 		chain++;
1819 		if (cur != NULL) {
1820 			ss->spdsock_dump_cur_chain = chain;
1821 			return (spdsock_dump_next_in_chain(ss, iph, cur));
1822 		}
1823 	}
1824 	ss->spdsock_dump_cur_chain = nchains;
1825 
1826 	af = ss->spdsock_dump_cur_af;
1827 	while (af < IPSEC_NAF) {
1828 		cur = ipr->ipr_nonhash[af];
1829 		af++;
1830 		if (cur != NULL) {
1831 			ss->spdsock_dump_cur_af = af;
1832 			return (spdsock_dump_next_in_chain(ss, iph, cur));
1833 		}
1834 	}
1835 
1836 	type++;
1837 	if (type >= IPSEC_NTYPES)
1838 		return (NULL);
1839 
1840 	ss->spdsock_dump_cur_chain = 0;
1841 	ss->spdsock_dump_cur_type = type;
1842 	ss->spdsock_dump_cur_af = IPSEC_AF_V4;
1843 	goto next;
1844 
1845 }
1846 
1847 /*
1848  * If we're done with one policy head, but have more to go, we iterate through
1849  * another IPsec tunnel policy head (itp).  Return NULL if it is an error
1850  * worthy of returning EAGAIN via PF_POLICY.
1851  */
1852 static ipsec_tun_pol_t *
1853 spdsock_dump_iterate_next_tunnel(spdsock_t *ss, ipsec_stack_t *ipss)
1854 {
1855 	ipsec_tun_pol_t *itp;
1856 
1857 	ASSERT(RW_READ_HELD(&ipss->ipsec_tunnel_policy_lock));
1858 	if (ipss->ipsec_tunnel_policy_gen > ss->spdsock_dump_tun_gen) {
1859 		/* Oops, state of the tunnel polheads changed. */
1860 		itp = NULL;
1861 	} else if (ss->spdsock_itp == NULL) {
1862 		/* Just finished global, find first node. */
1863 		itp = avl_first(&ipss->ipsec_tunnel_policies);
1864 	} else {
1865 		/* We just finished current polhead, find the next one. */
1866 		itp = AVL_NEXT(&ipss->ipsec_tunnel_policies, ss->spdsock_itp);
1867 	}
1868 	if (itp != NULL) {
1869 		ITP_REFHOLD(itp);
1870 	}
1871 	if (ss->spdsock_itp != NULL) {
1872 		ITP_REFRELE(ss->spdsock_itp, ipss->ipsec_netstack);
1873 	}
1874 	ss->spdsock_itp = itp;
1875 	return (itp);
1876 }
1877 
1878 static mblk_t *
1879 spdsock_dump_next_record(spdsock_t *ss)
1880 {
1881 	ipsec_policy_head_t *iph;
1882 	ipsec_policy_t *rule;
1883 	mblk_t *m;
1884 	ipsec_tun_pol_t *itp;
1885 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
1886 	ipsec_stack_t *ipss = ns->netstack_ipsec;
1887 
1888 	iph = ss->spdsock_dump_head;
1889 
1890 	ASSERT(iph != NULL);
1891 
1892 	rw_enter(&iph->iph_lock, RW_READER);
1893 
1894 	if (iph->iph_gen != ss->spdsock_dump_gen) {
1895 		rw_exit(&iph->iph_lock);
1896 		return (spdsock_dump_finish(ss, EAGAIN));
1897 	}
1898 
1899 	while ((rule = spdsock_dump_next_rule(ss, iph)) == NULL) {
1900 		rw_exit(&iph->iph_lock);
1901 		if (--(ss->spdsock_dump_remaining_polheads) == 0)
1902 			return (spdsock_dump_finish(ss, 0));
1903 
1904 
1905 		/*
1906 		 * If we reach here, we have more policy heads (tunnel
1907 		 * entries) to dump.  Let's reset to a new policy head
1908 		 * and get some more rules.
1909 		 *
1910 		 * An empty policy head will have spdsock_dump_next_rule()
1911 		 * return NULL, and we loop (while dropping the number of
1912 		 * remaining polheads).  If we loop to 0, we finish.  We
1913 		 * keep looping until we hit 0 or until we have a rule to
1914 		 * encode.
1915 		 *
1916 		 * NOTE:  No need for ITP_REF*() macros here as we're only
1917 		 * going after and refholding the policy head itself.
1918 		 */
1919 		rw_enter(&ipss->ipsec_tunnel_policy_lock, RW_READER);
1920 		itp = spdsock_dump_iterate_next_tunnel(ss, ipss);
1921 		if (itp == NULL) {
1922 			rw_exit(&ipss->ipsec_tunnel_policy_lock);
1923 			return (spdsock_dump_finish(ss, EAGAIN));
1924 		}
1925 
1926 		/* Reset other spdsock_dump thingies. */
1927 		IPPH_REFRELE(ss->spdsock_dump_head, ns);
1928 		if (ss->spdsock_dump_active) {
1929 			ss->spdsock_dump_tunnel =
1930 			    itp->itp_flags & ITPF_P_TUNNEL;
1931 			iph = itp->itp_policy;
1932 		} else {
1933 			ss->spdsock_dump_tunnel =
1934 			    itp->itp_flags & ITPF_I_TUNNEL;
1935 			iph = itp->itp_inactive;
1936 		}
1937 		IPPH_REFHOLD(iph);
1938 		rw_exit(&ipss->ipsec_tunnel_policy_lock);
1939 
1940 		rw_enter(&iph->iph_lock, RW_READER);
1941 		RESET_SPDSOCK_DUMP_POLHEAD(ss, iph);
1942 	}
1943 
1944 	m = spdsock_encode_rule(ss->spdsock_dump_req, rule,
1945 	    ss->spdsock_dump_cur_type, ss->spdsock_dump_cur_af,
1946 	    (ss->spdsock_itp == NULL) ? NULL : ss->spdsock_itp->itp_name,
1947 	    ss->spdsock_dump_tunnel);
1948 	rw_exit(&iph->iph_lock);
1949 
1950 	if (m == NULL)
1951 		return (spdsock_dump_finish(ss, ENOMEM));
1952 	return (m);
1953 }
1954 
1955 /*
1956  * Dump records until we run into flow-control back-pressure.
1957  */
1958 static void
1959 spdsock_dump_some(queue_t *q, spdsock_t *ss)
1960 {
1961 	mblk_t *m, *dataind;
1962 
1963 	while ((ss->spdsock_dump_req != NULL) && canputnext(q)) {
1964 		m = spdsock_dump_next_record(ss);
1965 		if (m == NULL)
1966 			return;
1967 		dataind = allocb(sizeof (struct T_data_req), BPRI_HI);
1968 		if (dataind == NULL) {
1969 			freemsg(m);
1970 			return;
1971 		}
1972 		dataind->b_cont = m;
1973 		dataind->b_wptr += sizeof (struct T_data_req);
1974 		((struct T_data_ind *)dataind->b_rptr)->PRIM_type = T_DATA_IND;
1975 		((struct T_data_ind *)dataind->b_rptr)->MORE_flag = 0;
1976 		dataind->b_datap->db_type = M_PROTO;
1977 		putnext(q, dataind);
1978 	}
1979 }
1980 
1981 /*
1982  * Start dumping.
1983  * Format a start-of-dump record, and set up the stream and kick the rsrv
1984  * procedure to continue the job..
1985  */
1986 /* ARGSUSED */
1987 static void
1988 spdsock_dump(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp)
1989 {
1990 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
1991 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
1992 	ipsec_stack_t *ipss = ns->netstack_ipsec;
1993 	mblk_t *mr;
1994 
1995 	/* spdsock_open() already set spdsock_itp to NULL. */
1996 	if (iph == ALL_ACTIVE_POLHEADS || iph == ALL_INACTIVE_POLHEADS) {
1997 		rw_enter(&ipss->ipsec_tunnel_policy_lock, RW_READER);
1998 		ss->spdsock_dump_remaining_polheads = 1 +
1999 		    avl_numnodes(&ipss->ipsec_tunnel_policies);
2000 		ss->spdsock_dump_tun_gen = ipss->ipsec_tunnel_policy_gen;
2001 		rw_exit(&ipss->ipsec_tunnel_policy_lock);
2002 		if (iph == ALL_ACTIVE_POLHEADS) {
2003 			iph = ipsec_system_policy(ns);
2004 			ss->spdsock_dump_active = B_TRUE;
2005 		} else {
2006 			iph = ipsec_inactive_policy(ns);
2007 			ss->spdsock_dump_active = B_FALSE;
2008 		}
2009 		ASSERT(ss->spdsock_itp == NULL);
2010 	} else {
2011 		ss->spdsock_dump_remaining_polheads = 1;
2012 	}
2013 
2014 	rw_enter(&iph->iph_lock, RW_READER);
2015 
2016 	mr = spdsock_dump_ruleset(mp, iph, 0, 0);
2017 
2018 	if (!mr) {
2019 		rw_exit(&iph->iph_lock);
2020 		spdsock_error(q, mp, ENOMEM, 0);
2021 		return;
2022 	}
2023 
2024 	ss->spdsock_dump_req = mp;
2025 	RESET_SPDSOCK_DUMP_POLHEAD(ss, iph);
2026 
2027 	rw_exit(&iph->iph_lock);
2028 
2029 	qreply(q, mr);
2030 	qenable(OTHERQ(q));
2031 }
2032 
2033 /* Do NOT consume a reference to ITP. */
2034 void
2035 spdsock_clone_node(ipsec_tun_pol_t *itp, void *ep, netstack_t *ns)
2036 {
2037 	int *errptr = (int *)ep;
2038 
2039 	if (*errptr != 0)
2040 		return;	/* We've failed already for some reason. */
2041 	mutex_enter(&itp->itp_lock);
2042 	ITPF_CLONE(itp->itp_flags);
2043 	*errptr = ipsec_copy_polhead(itp->itp_policy, itp->itp_inactive, ns);
2044 	mutex_exit(&itp->itp_lock);
2045 }
2046 
2047 void
2048 spdsock_clone(queue_t *q, mblk_t *mp, spd_if_t *tunname)
2049 {
2050 	int error;
2051 	char *tname;
2052 	ipsec_tun_pol_t *itp;
2053 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2054 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
2055 
2056 	if (tunname != NULL) {
2057 		tname = (char *)tunname->spd_if_name;
2058 		if (*tname == '\0') {
2059 			error = ipsec_clone_system_policy(ns);
2060 			if (audit_active) {
2061 				boolean_t active;
2062 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
2063 
2064 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
2065 				audit_pf_policy(SPD_CLONE, DB_CRED(mp), ns,
2066 				    NULL, active, error, DB_CPID(mp));
2067 			}
2068 			if (error == 0) {
2069 				itp_walk(spdsock_clone_node, &error, ns);
2070 				if (audit_active) {
2071 					boolean_t active;
2072 					spd_msg_t *spmsg =
2073 					    (spd_msg_t *)mp->b_rptr;
2074 
2075 					active = (spmsg->spd_msg_spdid ==
2076 					    SPD_ACTIVE);
2077 					audit_pf_policy(SPD_CLONE, DB_CRED(mp),
2078 					    ns, "all tunnels", active, 0,
2079 					    DB_CPID(mp));
2080 				}
2081 			}
2082 		} else {
2083 			itp = get_tunnel_policy(tname, ns);
2084 			if (itp == NULL) {
2085 				spdsock_error(q, mp, ENOENT, 0);
2086 				if (audit_active) {
2087 					boolean_t active;
2088 					spd_msg_t *spmsg =
2089 					    (spd_msg_t *)mp->b_rptr;
2090 
2091 					active = (spmsg->spd_msg_spdid ==
2092 					    SPD_ACTIVE);
2093 					audit_pf_policy(SPD_CLONE, DB_CRED(mp),
2094 					    ns, ITP_NAME(itp), active, ENOENT,
2095 					    DB_CPID(mp));
2096 				}
2097 				return;
2098 			}
2099 			spdsock_clone_node(itp, &error, NULL);
2100 			ITP_REFRELE(itp, ns);
2101 			if (audit_active) {
2102 				boolean_t active;
2103 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
2104 
2105 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
2106 				audit_pf_policy(SPD_CLONE, DB_CRED(mp), ns,
2107 				    ITP_NAME(itp), active, error, DB_CPID(mp));
2108 			}
2109 		}
2110 	} else {
2111 		error = ipsec_clone_system_policy(ns);
2112 		if (audit_active) {
2113 			boolean_t active;
2114 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
2115 
2116 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
2117 			audit_pf_policy(SPD_CLONE, DB_CRED(mp), ns, NULL,
2118 			    active, error, DB_CPID(mp));
2119 		}
2120 	}
2121 
2122 	if (error != 0)
2123 		spdsock_error(q, mp, error, 0);
2124 	else
2125 		spd_echo(q, mp);
2126 }
2127 
2128 /*
2129  * Process a SPD_ALGLIST request. The caller expects separate alg entries
2130  * for AH authentication, ESP authentication, and ESP encryption.
2131  * The same distinction is then used when setting the min and max key
2132  * sizes when defining policies.
2133  */
2134 
2135 #define	SPDSOCK_AH_AUTH		0
2136 #define	SPDSOCK_ESP_AUTH	1
2137 #define	SPDSOCK_ESP_ENCR	2
2138 #define	SPDSOCK_NTYPES		3
2139 
2140 static const uint_t algattr[SPDSOCK_NTYPES] = {
2141 	SPD_ATTR_AH_AUTH,
2142 	SPD_ATTR_ESP_AUTH,
2143 	SPD_ATTR_ESP_ENCR
2144 };
2145 static const uint_t minbitsattr[SPDSOCK_NTYPES] = {
2146 	SPD_ATTR_AH_MINBITS,
2147 	SPD_ATTR_ESPA_MINBITS,
2148 	SPD_ATTR_ENCR_MINBITS
2149 };
2150 static const uint_t maxbitsattr[SPDSOCK_NTYPES] = {
2151 	SPD_ATTR_AH_MAXBITS,
2152 	SPD_ATTR_ESPA_MAXBITS,
2153 	SPD_ATTR_ENCR_MAXBITS
2154 };
2155 static const uint_t defbitsattr[SPDSOCK_NTYPES] = {
2156 	SPD_ATTR_AH_DEFBITS,
2157 	SPD_ATTR_ESPA_DEFBITS,
2158 	SPD_ATTR_ENCR_DEFBITS
2159 };
2160 static const uint_t incrbitsattr[SPDSOCK_NTYPES] = {
2161 	SPD_ATTR_AH_INCRBITS,
2162 	SPD_ATTR_ESPA_INCRBITS,
2163 	SPD_ATTR_ENCR_INCRBITS
2164 };
2165 
2166 #define	ATTRPERALG	6	/* fixed attributes per algs */
2167 
2168 void
2169 spdsock_alglist(queue_t *q, mblk_t *mp)
2170 {
2171 	uint_t algtype;
2172 	uint_t algidx;
2173 	uint_t algcount;
2174 	uint_t size;
2175 	mblk_t *m;
2176 	uint8_t *cur;
2177 	spd_msg_t *msg;
2178 	struct spd_ext_actions *act;
2179 	struct spd_attribute *attr;
2180 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2181 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
2182 
2183 	mutex_enter(&ipss->ipsec_alg_lock);
2184 	/*
2185 	 * The SPD client expects to receive separate entries for
2186 	 * AH authentication and ESP authentication supported algorithms.
2187 	 *
2188 	 * Don't return the "any" algorithms, if defined, as no
2189 	 * kernel policies can be set for these algorithms.
2190 	 */
2191 	algcount = 2 * ipss->ipsec_nalgs[IPSEC_ALG_AUTH] +
2192 	    ipss->ipsec_nalgs[IPSEC_ALG_ENCR];
2193 
2194 	if (ipss->ipsec_alglists[IPSEC_ALG_AUTH][SADB_AALG_NONE] != NULL)
2195 		algcount--;
2196 	if (ipss->ipsec_alglists[IPSEC_ALG_ENCR][SADB_EALG_NONE] != NULL)
2197 		algcount--;
2198 
2199 	/*
2200 	 * For each algorithm, we encode:
2201 	 * ALG / MINBITS / MAXBITS / DEFBITS / INCRBITS / {END, NEXT}
2202 	 */
2203 
2204 	size = sizeof (spd_msg_t) + sizeof (struct spd_ext_actions) +
2205 	    ATTRPERALG * sizeof (struct spd_attribute) * algcount;
2206 
2207 	ASSERT(ALIGNED64(size));
2208 
2209 	m = allocb(size, BPRI_HI);
2210 	if (m == NULL) {
2211 		mutex_exit(&ipss->ipsec_alg_lock);
2212 		spdsock_error(q, mp, ENOMEM, 0);
2213 		return;
2214 	}
2215 
2216 	m->b_wptr = m->b_rptr + size;
2217 	cur = m->b_rptr;
2218 
2219 	msg = (spd_msg_t *)cur;
2220 	bcopy(mp->b_rptr, cur, sizeof (*msg));
2221 
2222 	msg->spd_msg_len = SPD_8TO64(size);
2223 	msg->spd_msg_errno = 0;
2224 	msg->spd_msg_diagnostic = 0;
2225 
2226 	cur += sizeof (*msg);
2227 
2228 	act = (struct spd_ext_actions *)cur;
2229 	cur += sizeof (*act);
2230 
2231 	act->spd_actions_len = SPD_8TO64(size - sizeof (spd_msg_t));
2232 	act->spd_actions_exttype = SPD_EXT_ACTION;
2233 	act->spd_actions_count = algcount;
2234 	act->spd_actions_reserved = 0;
2235 
2236 	attr = (struct spd_attribute *)cur;
2237 
2238 #define	EMIT(tag, value) {					\
2239 		attr->spd_attr_tag = (tag); 			\
2240 		attr->spd_attr_value = (value); 		\
2241 		attr++;			  			\
2242 	}
2243 
2244 	/*
2245 	 * If you change the number of EMIT's here, change
2246 	 * ATTRPERALG above to match
2247 	 */
2248 #define	EMITALGATTRS(_type) {					\
2249 		EMIT(algattr[_type], algid); 		/* 1 */	\
2250 		EMIT(minbitsattr[_type], minbits);	/* 2 */	\
2251 		EMIT(maxbitsattr[_type], maxbits);	/* 3 */	\
2252 		EMIT(defbitsattr[_type], defbits);	/* 4 */	\
2253 		EMIT(incrbitsattr[_type], incr);	/* 5 */	\
2254 		EMIT(SPD_ATTR_NEXT, 0);			/* 6 */	\
2255 	}
2256 
2257 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
2258 		for (algidx = 0; algidx < ipss->ipsec_nalgs[algtype];
2259 		    algidx++) {
2260 			int algid = ipss->ipsec_sortlist[algtype][algidx];
2261 			ipsec_alginfo_t *alg =
2262 			    ipss->ipsec_alglists[algtype][algid];
2263 			uint_t minbits = alg->alg_minbits;
2264 			uint_t maxbits = alg->alg_maxbits;
2265 			uint_t defbits = alg->alg_default_bits;
2266 			uint_t incr = alg->alg_increment;
2267 
2268 			if (algtype == IPSEC_ALG_AUTH) {
2269 				if (algid == SADB_AALG_NONE)
2270 					continue;
2271 				EMITALGATTRS(SPDSOCK_AH_AUTH);
2272 				EMITALGATTRS(SPDSOCK_ESP_AUTH);
2273 			} else {
2274 				if (algid == SADB_EALG_NONE)
2275 					continue;
2276 				ASSERT(algtype == IPSEC_ALG_ENCR);
2277 				EMITALGATTRS(SPDSOCK_ESP_ENCR);
2278 			}
2279 		}
2280 	}
2281 
2282 	mutex_exit(&ipss->ipsec_alg_lock);
2283 
2284 #undef EMITALGATTRS
2285 #undef EMIT
2286 #undef ATTRPERALG
2287 
2288 	attr--;
2289 	attr->spd_attr_tag = SPD_ATTR_END;
2290 
2291 	freemsg(mp);
2292 	qreply(q, m);
2293 }
2294 
2295 /*
2296  * Process a SPD_DUMPALGS request.
2297  */
2298 
2299 #define	ATTRPERALG	7	/* fixed attributes per algs */
2300 
2301 void
2302 spdsock_dumpalgs(queue_t *q, mblk_t *mp)
2303 {
2304 	uint_t algtype;
2305 	uint_t algidx;
2306 	uint_t size;
2307 	mblk_t *m;
2308 	uint8_t *cur;
2309 	spd_msg_t *msg;
2310 	struct spd_ext_actions *act;
2311 	struct spd_attribute *attr;
2312 	ipsec_alginfo_t *alg;
2313 	uint_t algid;
2314 	uint_t i;
2315 	uint_t alg_size;
2316 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2317 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
2318 
2319 	mutex_enter(&ipss->ipsec_alg_lock);
2320 
2321 	/*
2322 	 * For each algorithm, we encode:
2323 	 * ALG / MINBITS / MAXBITS / DEFBITS / INCRBITS / {END, NEXT}
2324 	 *
2325 	 * ALG_ID / ALG_PROTO / ALG_INCRBITS / ALG_NKEYSIZES / ALG_KEYSIZE*
2326 	 * ALG_NBLOCKSIZES / ALG_BLOCKSIZE* / ALG_MECHNAME / {END, NEXT}
2327 	 */
2328 
2329 	/*
2330 	 * Compute the size of the SPD message.
2331 	 */
2332 	size = sizeof (spd_msg_t) + sizeof (struct spd_ext_actions);
2333 
2334 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
2335 		for (algidx = 0; algidx < ipss->ipsec_nalgs[algtype];
2336 		    algidx++) {
2337 			algid = ipss->ipsec_sortlist[algtype][algidx];
2338 			alg = ipss->ipsec_alglists[algtype][algid];
2339 			alg_size = sizeof (struct spd_attribute) *
2340 			    (ATTRPERALG + alg->alg_nkey_sizes +
2341 			    alg->alg_nblock_sizes) + CRYPTO_MAX_MECH_NAME;
2342 			size += alg_size;
2343 		}
2344 	}
2345 
2346 	ASSERT(ALIGNED64(size));
2347 
2348 	m = allocb(size, BPRI_HI);
2349 	if (m == NULL) {
2350 		mutex_exit(&ipss->ipsec_alg_lock);
2351 		spdsock_error(q, mp, ENOMEM, 0);
2352 		return;
2353 	}
2354 
2355 	m->b_wptr = m->b_rptr + size;
2356 	cur = m->b_rptr;
2357 
2358 	msg = (spd_msg_t *)cur;
2359 	bcopy(mp->b_rptr, cur, sizeof (*msg));
2360 
2361 	msg->spd_msg_len = SPD_8TO64(size);
2362 	msg->spd_msg_errno = 0;
2363 	msg->spd_msg_diagnostic = 0;
2364 
2365 	cur += sizeof (*msg);
2366 
2367 	act = (struct spd_ext_actions *)cur;
2368 	cur += sizeof (*act);
2369 
2370 	act->spd_actions_len = SPD_8TO64(size - sizeof (spd_msg_t));
2371 	act->spd_actions_exttype = SPD_EXT_ACTION;
2372 	act->spd_actions_count = ipss->ipsec_nalgs[IPSEC_ALG_AUTH] +
2373 	    ipss->ipsec_nalgs[IPSEC_ALG_ENCR];
2374 	act->spd_actions_reserved = 0;
2375 
2376 	attr = (struct spd_attribute *)cur;
2377 
2378 #define	EMIT(tag, value) {					\
2379 		attr->spd_attr_tag = (tag); 			\
2380 		attr->spd_attr_value = (value); 		\
2381 		attr++;			  			\
2382 	}
2383 
2384 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
2385 		for (algidx = 0; algidx < ipss->ipsec_nalgs[algtype];
2386 		    algidx++) {
2387 
2388 			algid = ipss->ipsec_sortlist[algtype][algidx];
2389 			alg = ipss->ipsec_alglists[algtype][algid];
2390 
2391 			/*
2392 			 * If you change the number of EMIT's here, change
2393 			 * ATTRPERALG above to match
2394 			 */
2395 			EMIT(SPD_ATTR_ALG_ID, algid);
2396 			EMIT(SPD_ATTR_ALG_PROTO, algproto[algtype]);
2397 			EMIT(SPD_ATTR_ALG_INCRBITS, alg->alg_increment);
2398 
2399 			EMIT(SPD_ATTR_ALG_NKEYSIZES, alg->alg_nkey_sizes);
2400 			for (i = 0; i < alg->alg_nkey_sizes; i++)
2401 				EMIT(SPD_ATTR_ALG_KEYSIZE,
2402 				    alg->alg_key_sizes[i]);
2403 
2404 			EMIT(SPD_ATTR_ALG_NBLOCKSIZES, alg->alg_nblock_sizes);
2405 			for (i = 0; i < alg->alg_nblock_sizes; i++)
2406 				EMIT(SPD_ATTR_ALG_BLOCKSIZE,
2407 				    alg->alg_block_sizes[i]);
2408 
2409 			EMIT(SPD_ATTR_ALG_MECHNAME, CRYPTO_MAX_MECH_NAME);
2410 			bcopy(alg->alg_mech_name, attr, CRYPTO_MAX_MECH_NAME);
2411 			attr = (struct spd_attribute *)((char *)attr +
2412 			    CRYPTO_MAX_MECH_NAME);
2413 
2414 			EMIT(SPD_ATTR_NEXT, 0);
2415 		}
2416 	}
2417 
2418 	mutex_exit(&ipss->ipsec_alg_lock);
2419 
2420 #undef EMITALGATTRS
2421 #undef EMIT
2422 #undef ATTRPERALG
2423 
2424 	attr--;
2425 	attr->spd_attr_tag = SPD_ATTR_END;
2426 
2427 	freemsg(mp);
2428 	qreply(q, m);
2429 }
2430 
2431 /*
2432  * Do the actual work of processing an SPD_UPDATEALGS request. Can
2433  * be invoked either once IPsec is loaded on a cached request, or
2434  * when a request is received while IPsec is loaded.
2435  */
2436 static void
2437 spdsock_do_updatealg(spd_ext_t *extv[], int *diag, spd_stack_t *spds)
2438 {
2439 	struct spd_ext_actions *actp;
2440 	struct spd_attribute *attr, *endattr;
2441 	uint64_t *start, *end;
2442 	ipsec_alginfo_t *alg = NULL;
2443 	ipsec_algtype_t alg_type = 0;
2444 	boolean_t skip_alg = B_TRUE, doing_proto = B_FALSE;
2445 	uint_t i, cur_key, cur_block, algid;
2446 
2447 	*diag = -1;
2448 	ASSERT(MUTEX_HELD(&spds->spds_alg_lock));
2449 
2450 	/* parse the message, building the list of algorithms */
2451 
2452 	actp = (struct spd_ext_actions *)extv[SPD_EXT_ACTION];
2453 	if (actp == NULL) {
2454 		*diag = SPD_DIAGNOSTIC_NO_ACTION_EXT;
2455 		return;
2456 	}
2457 
2458 	start = (uint64_t *)actp;
2459 	end = (start + actp->spd_actions_len);
2460 	endattr = (struct spd_attribute *)end;
2461 	attr = (struct spd_attribute *)&actp[1];
2462 
2463 	bzero(spds->spds_algs, IPSEC_NALGTYPES * IPSEC_MAX_ALGS *
2464 	    sizeof (ipsec_alginfo_t *));
2465 
2466 	alg = kmem_zalloc(sizeof (*alg), KM_SLEEP);
2467 
2468 #define	ALG_KEY_SIZES(a)   (((a)->alg_nkey_sizes + 1) * sizeof (uint16_t))
2469 #define	ALG_BLOCK_SIZES(a) (((a)->alg_nblock_sizes + 1) * sizeof (uint16_t))
2470 
2471 	while (attr < endattr) {
2472 		switch (attr->spd_attr_tag) {
2473 		case SPD_ATTR_NOP:
2474 		case SPD_ATTR_EMPTY:
2475 			break;
2476 		case SPD_ATTR_END:
2477 			attr = endattr;
2478 			/* FALLTHRU */
2479 		case SPD_ATTR_NEXT:
2480 			if (doing_proto) {
2481 				doing_proto = B_FALSE;
2482 				break;
2483 			}
2484 			if (skip_alg) {
2485 				ipsec_alg_free(alg);
2486 			} else {
2487 				ipsec_alg_free(
2488 				    spds->spds_algs[alg_type][alg->alg_id]);
2489 				spds->spds_algs[alg_type][alg->alg_id] =
2490 				    alg;
2491 			}
2492 			alg = kmem_zalloc(sizeof (*alg), KM_SLEEP);
2493 			break;
2494 
2495 		case SPD_ATTR_ALG_ID:
2496 			if (attr->spd_attr_value >= IPSEC_MAX_ALGS) {
2497 				ss1dbg(spds, ("spdsock_do_updatealg: "
2498 				    "invalid alg id %d\n",
2499 				    attr->spd_attr_value));
2500 				*diag = SPD_DIAGNOSTIC_ALG_ID_RANGE;
2501 				goto bail;
2502 			}
2503 			alg->alg_id = attr->spd_attr_value;
2504 			break;
2505 
2506 		case SPD_ATTR_ALG_PROTO:
2507 			/* find the alg type */
2508 			for (i = 0; i < NALGPROTOS; i++)
2509 				if (algproto[i] == attr->spd_attr_value)
2510 					break;
2511 			skip_alg = (i == NALGPROTOS);
2512 			if (!skip_alg)
2513 				alg_type = i;
2514 			break;
2515 
2516 		case SPD_ATTR_ALG_INCRBITS:
2517 			alg->alg_increment = attr->spd_attr_value;
2518 			break;
2519 
2520 		case SPD_ATTR_ALG_NKEYSIZES:
2521 			if (alg->alg_key_sizes != NULL) {
2522 				kmem_free(alg->alg_key_sizes,
2523 				    ALG_KEY_SIZES(alg));
2524 			}
2525 			alg->alg_nkey_sizes = attr->spd_attr_value;
2526 			/*
2527 			 * Allocate room for the trailing zero key size
2528 			 * value as well.
2529 			 */
2530 			alg->alg_key_sizes = kmem_zalloc(ALG_KEY_SIZES(alg),
2531 			    KM_SLEEP);
2532 			cur_key = 0;
2533 			break;
2534 
2535 		case SPD_ATTR_ALG_KEYSIZE:
2536 			if (alg->alg_key_sizes == NULL ||
2537 			    cur_key >= alg->alg_nkey_sizes) {
2538 				ss1dbg(spds, ("spdsock_do_updatealg: "
2539 				    "too many key sizes\n"));
2540 				*diag = SPD_DIAGNOSTIC_ALG_NUM_KEY_SIZES;
2541 				goto bail;
2542 			}
2543 			alg->alg_key_sizes[cur_key++] = attr->spd_attr_value;
2544 			break;
2545 
2546 		case SPD_ATTR_ALG_NBLOCKSIZES:
2547 			if (alg->alg_block_sizes != NULL) {
2548 				kmem_free(alg->alg_block_sizes,
2549 				    ALG_BLOCK_SIZES(alg));
2550 			}
2551 			alg->alg_nblock_sizes = attr->spd_attr_value;
2552 			/*
2553 			 * Allocate room for the trailing zero block size
2554 			 * value as well.
2555 			 */
2556 			alg->alg_block_sizes = kmem_zalloc(ALG_BLOCK_SIZES(alg),
2557 			    KM_SLEEP);
2558 			cur_block = 0;
2559 			break;
2560 
2561 		case SPD_ATTR_ALG_BLOCKSIZE:
2562 			if (alg->alg_block_sizes == NULL ||
2563 			    cur_block >= alg->alg_nblock_sizes) {
2564 				ss1dbg(spds, ("spdsock_do_updatealg: "
2565 				    "too many block sizes\n"));
2566 				*diag = SPD_DIAGNOSTIC_ALG_NUM_BLOCK_SIZES;
2567 				goto bail;
2568 			}
2569 			alg->alg_block_sizes[cur_block++] =
2570 			    attr->spd_attr_value;
2571 			break;
2572 
2573 		case SPD_ATTR_ALG_MECHNAME: {
2574 			char *mech_name;
2575 
2576 			if (attr->spd_attr_value > CRYPTO_MAX_MECH_NAME) {
2577 				ss1dbg(spds, ("spdsock_do_updatealg: "
2578 				    "mech name too long\n"));
2579 				*diag = SPD_DIAGNOSTIC_ALG_MECH_NAME_LEN;
2580 				goto bail;
2581 			}
2582 			mech_name = (char *)(attr + 1);
2583 			bcopy(mech_name, alg->alg_mech_name,
2584 			    attr->spd_attr_value);
2585 			alg->alg_mech_name[CRYPTO_MAX_MECH_NAME-1] = '\0';
2586 			attr = (struct spd_attribute *)((char *)attr +
2587 			    attr->spd_attr_value);
2588 			break;
2589 		}
2590 
2591 		case SPD_ATTR_PROTO_ID:
2592 			doing_proto = B_TRUE;
2593 			for (i = 0; i < NALGPROTOS; i++) {
2594 				if (algproto[i] == attr->spd_attr_value) {
2595 					alg_type = i;
2596 					break;
2597 				}
2598 			}
2599 			break;
2600 
2601 		case SPD_ATTR_PROTO_EXEC_MODE:
2602 			if (!doing_proto)
2603 				break;
2604 			for (i = 0; i < NEXECMODES; i++) {
2605 				if (execmodes[i] == attr->spd_attr_value) {
2606 					spds->spds_algs_exec_mode[alg_type] = i;
2607 					break;
2608 				}
2609 			}
2610 			break;
2611 		}
2612 		attr++;
2613 	}
2614 
2615 #undef	ALG_KEY_SIZES
2616 #undef	ALG_BLOCK_SIZES
2617 
2618 	/* update the algorithm tables */
2619 	spdsock_merge_algs(spds);
2620 bail:
2621 	/* cleanup */
2622 	ipsec_alg_free(alg);
2623 	for (alg_type = 0; alg_type < IPSEC_NALGTYPES; alg_type++)
2624 		for (algid = 0; algid < IPSEC_MAX_ALGS; algid++)
2625 		if (spds->spds_algs[alg_type][algid] != NULL)
2626 			ipsec_alg_free(spds->spds_algs[alg_type][algid]);
2627 }
2628 
2629 /*
2630  * Process an SPD_UPDATEALGS request. If IPsec is not loaded, queue
2631  * the request until IPsec loads. If IPsec is loaded, act on it
2632  * immediately.
2633  */
2634 
2635 static void
2636 spdsock_updatealg(queue_t *q, mblk_t *mp, spd_ext_t *extv[])
2637 {
2638 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2639 	spd_stack_t	*spds = ss->spdsock_spds;
2640 	ipsec_stack_t	*ipss = spds->spds_netstack->netstack_ipsec;
2641 
2642 	if (!ipsec_loaded(ipss)) {
2643 		/*
2644 		 * IPsec is not loaded, save request and return nicely,
2645 		 * the message will be processed once IPsec loads.
2646 		 */
2647 		mblk_t *new_mp;
2648 
2649 		/* last update message wins */
2650 		if ((new_mp = copymsg(mp)) == NULL) {
2651 			spdsock_error(q, mp, ENOMEM, 0);
2652 			return;
2653 		}
2654 		mutex_enter(&spds->spds_alg_lock);
2655 		bcopy(extv, spds->spds_extv_algs,
2656 		    sizeof (spd_ext_t *) * (SPD_EXT_MAX + 1));
2657 		if (spds->spds_mp_algs != NULL)
2658 			freemsg(spds->spds_mp_algs);
2659 		spds->spds_mp_algs = mp;
2660 		spds->spds_algs_pending = B_TRUE;
2661 		mutex_exit(&spds->spds_alg_lock);
2662 		if (audit_active)
2663 			audit_pf_policy(SPD_UPDATEALGS, DB_CRED(mp),
2664 			    spds->spds_netstack, NULL, B_TRUE, EAGAIN,
2665 			    DB_CPID(mp));
2666 		spd_echo(q, new_mp);
2667 	} else {
2668 		/*
2669 		 * IPsec is loaded, act on the message immediately.
2670 		 */
2671 		int diag;
2672 
2673 		mutex_enter(&spds->spds_alg_lock);
2674 		spdsock_do_updatealg(extv, &diag, spds);
2675 		mutex_exit(&spds->spds_alg_lock);
2676 		if (diag == -1) {
2677 			spd_echo(q, mp);
2678 		if (audit_active)
2679 			audit_pf_policy(SPD_UPDATEALGS, DB_CRED(mp),
2680 			    spds->spds_netstack, NULL, B_TRUE, 0,
2681 			    DB_CPID(mp));
2682 		} else {
2683 			spdsock_diag(q, mp, diag);
2684 		if (audit_active)
2685 			audit_pf_policy(SPD_UPDATEALGS, DB_CRED(mp),
2686 			    spds->spds_netstack, NULL, B_TRUE, diag,
2687 			    DB_CPID(mp));
2688 		}
2689 	}
2690 }
2691 
2692 /*
2693  * With a reference-held ill, dig down and find an instance of "tun", and
2694  * assign its tunnel policy pointer, while reference-holding it.  Also,
2695  * release ill's refrence when finished.
2696  *
2697  * We'll be messing with q_next, so be VERY careful.
2698  */
2699 static void
2700 find_tun_and_set_itp(ill_t *ill, ipsec_tun_pol_t *itp)
2701 {
2702 	queue_t *q;
2703 	tun_t *tun;
2704 
2705 	/* Don't bother if this ill is going away. */
2706 	if (ill->ill_flags & ILL_CONDEMNED) {
2707 		ill_refrele(ill);
2708 		return;
2709 	}
2710 
2711 
2712 	q = ill->ill_wq;
2713 	claimstr(q);	/* Lighter-weight than freezestr(). */
2714 
2715 	do {
2716 		/* Use strcmp() because "tun" is bounded. */
2717 		if (strcmp(q->q_qinfo->qi_minfo->mi_idname, "tun") == 0) {
2718 			/* Aha!  Got it. */
2719 			tun = (tun_t *)q->q_ptr;
2720 			if (tun != NULL) {
2721 				mutex_enter(&tun->tun_lock);
2722 				if (tun->tun_itp != itp) {
2723 					ASSERT(tun->tun_itp == NULL);
2724 					ITP_REFHOLD(itp);
2725 					tun->tun_itp = itp;
2726 				}
2727 				mutex_exit(&tun->tun_lock);
2728 				goto release_and_return;
2729 			}
2730 			/*
2731 			 * Else assume this is some other module named "tun"
2732 			 * and move on, hoping we find one that actually has
2733 			 * something in q_ptr.
2734 			 */
2735 		}
2736 		q = q->q_next;
2737 	} while (q != NULL);
2738 
2739 release_and_return:
2740 	releasestr(ill->ill_wq);
2741 	ill_refrele(ill);
2742 }
2743 
2744 /*
2745  * Sort through the mess of polhead options to retrieve an appropriate one.
2746  * Returns NULL if we send an spdsock error.  Returns a valid pointer if we
2747  * found a valid polhead.  Returns ALL_ACTIVE_POLHEADS (aka. -1) or
2748  * ALL_INACTIVE_POLHEADS (aka. -2) if the operation calls for the operation to
2749  * act on ALL policy heads.
2750  */
2751 static ipsec_policy_head_t *
2752 get_appropriate_polhead(queue_t *q, mblk_t *mp, spd_if_t *tunname, int spdid,
2753     int msgtype, ipsec_tun_pol_t **itpp)
2754 {
2755 	ipsec_tun_pol_t *itp;
2756 	ipsec_policy_head_t *iph;
2757 	int errno;
2758 	char *tname;
2759 	boolean_t active;
2760 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2761 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
2762 	uint64_t gen;	/* Placeholder */
2763 	ill_t *v4, *v6;
2764 
2765 	active = (spdid == SPD_ACTIVE);
2766 	*itpp = NULL;
2767 	if (!active && spdid != SPD_STANDBY) {
2768 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_BAD_SPDID);
2769 		return (NULL);
2770 	}
2771 
2772 	if (tunname != NULL) {
2773 		/* Acting on a tunnel's SPD. */
2774 		tname = (char *)tunname->spd_if_name;
2775 		if (*tname == '\0') {
2776 			/* Handle all-polhead cases here. */
2777 			if (msgtype != SPD_FLUSH && msgtype != SPD_DUMP) {
2778 				spdsock_diag(q, mp,
2779 				    SPD_DIAGNOSTIC_NOT_GLOBAL_OP);
2780 				return (NULL);
2781 			}
2782 			return (active ? ALL_ACTIVE_POLHEADS :
2783 			    ALL_INACTIVE_POLHEADS);
2784 		}
2785 
2786 		itp = get_tunnel_policy(tname, ns);
2787 		if (itp == NULL) {
2788 			if (msgtype != SPD_ADDRULE) {
2789 				/* "Tunnel not found" */
2790 				spdsock_error(q, mp, ENOENT, 0);
2791 				return (NULL);
2792 			}
2793 
2794 			errno = 0;
2795 			itp = create_tunnel_policy(tname, &errno, &gen, ns);
2796 			if (itp == NULL) {
2797 				/*
2798 				 * Something very bad happened, most likely
2799 				 * ENOMEM.  Return an indicator.
2800 				 */
2801 				spdsock_error(q, mp, errno, 0);
2802 				return (NULL);
2803 			}
2804 		}
2805 		/*
2806 		 * Troll the plumbed tunnels and see if we have a
2807 		 * match.  We need to do this always in case we add
2808 		 * policy AFTER plumbing a tunnel.
2809 		 */
2810 		v4 = ill_lookup_on_name(tname, B_FALSE, B_FALSE, NULL,
2811 		    NULL, NULL, &errno, NULL, ns->netstack_ip);
2812 		if (v4 != NULL)
2813 			find_tun_and_set_itp(v4, itp);
2814 		v6 = ill_lookup_on_name(tname, B_FALSE, B_TRUE, NULL,
2815 		    NULL, NULL, &errno, NULL, ns->netstack_ip);
2816 		if (v6 != NULL)
2817 			find_tun_and_set_itp(v6, itp);
2818 		ASSERT(itp != NULL);
2819 		*itpp = itp;
2820 		/* For spdsock dump state, set the polhead's name. */
2821 		if (msgtype == SPD_DUMP) {
2822 			ITP_REFHOLD(itp);
2823 			ss->spdsock_itp = itp;
2824 			ss->spdsock_dump_tunnel = itp->itp_flags &
2825 			    (active ? ITPF_P_TUNNEL : ITPF_I_TUNNEL);
2826 		}
2827 	} else {
2828 		itp = NULL;
2829 		/* For spdsock dump state, indicate it's global policy. */
2830 		if (msgtype == SPD_DUMP)
2831 			ss->spdsock_itp = NULL;
2832 	}
2833 
2834 	if (active)
2835 		iph = (itp == NULL) ? ipsec_system_policy(ns) : itp->itp_policy;
2836 	else
2837 		iph = (itp == NULL) ? ipsec_inactive_policy(ns) :
2838 		    itp->itp_inactive;
2839 
2840 	ASSERT(iph != NULL);
2841 	if (itp != NULL) {
2842 		IPPH_REFHOLD(iph);
2843 	}
2844 
2845 	return (iph);
2846 }
2847 
2848 static void
2849 spdsock_parse(queue_t *q, mblk_t *mp)
2850 {
2851 	spd_msg_t *spmsg;
2852 	spd_ext_t *extv[SPD_EXT_MAX + 1];
2853 	uint_t msgsize;
2854 	ipsec_policy_head_t *iph;
2855 	ipsec_tun_pol_t *itp;
2856 	spd_if_t *tunname;
2857 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2858 	spd_stack_t *spds = ss->spdsock_spds;
2859 	netstack_t *ns = spds->spds_netstack;
2860 	ipsec_stack_t *ipss = ns->netstack_ipsec;
2861 
2862 	/* Make sure nothing's below me. */
2863 	ASSERT(WR(q)->q_next == NULL);
2864 
2865 	spmsg = (spd_msg_t *)mp->b_rptr;
2866 
2867 	msgsize = SPD_64TO8(spmsg->spd_msg_len);
2868 
2869 	if (msgdsize(mp) != msgsize) {
2870 		/*
2871 		 * Message len incorrect w.r.t. actual size.  Send an error
2872 		 * (EMSGSIZE).	It may be necessary to massage things a
2873 		 * bit.	 For example, if the spd_msg_type is hosed,
2874 		 * I need to set it to SPD_RESERVED to get delivery to
2875 		 * do the right thing.	Then again, maybe just letting
2876 		 * the error delivery do the right thing.
2877 		 */
2878 		ss2dbg(spds,
2879 		    ("mblk (%lu) and base (%d) message sizes don't jibe.\n",
2880 		    msgdsize(mp), msgsize));
2881 		spdsock_error(q, mp, EMSGSIZE, SPD_DIAGNOSTIC_NONE);
2882 		return;
2883 	}
2884 
2885 	if (msgsize > (uint_t)(mp->b_wptr - mp->b_rptr)) {
2886 		/* Get all message into one mblk. */
2887 		if (pullupmsg(mp, -1) == 0) {
2888 			/*
2889 			 * Something screwy happened.
2890 			 */
2891 			ss3dbg(spds, ("spdsock_parse: pullupmsg() failed.\n"));
2892 			return;
2893 		} else {
2894 			spmsg = (spd_msg_t *)mp->b_rptr;
2895 		}
2896 	}
2897 
2898 	switch (spdsock_get_ext(extv, spmsg, msgsize)) {
2899 	case KGE_DUP:
2900 		/* Handle duplicate extension. */
2901 		ss1dbg(spds, ("Got duplicate extension of type %d.\n",
2902 		    extv[0]->spd_ext_type));
2903 		spdsock_diag(q, mp, dup_ext_diag[extv[0]->spd_ext_type]);
2904 		return;
2905 	case KGE_UNK:
2906 		/* Handle unknown extension. */
2907 		ss1dbg(spds, ("Got unknown extension of type %d.\n",
2908 		    extv[0]->spd_ext_type));
2909 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_UNKNOWN_EXT);
2910 		return;
2911 	case KGE_LEN:
2912 		/* Length error. */
2913 		ss1dbg(spds, ("Length %d on extension type %d overrun or 0.\n",
2914 		    extv[0]->spd_ext_len, extv[0]->spd_ext_type));
2915 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_BAD_EXTLEN);
2916 		return;
2917 	case KGE_CHK:
2918 		/* Reality check failed. */
2919 		ss1dbg(spds, ("Reality check failed on extension type %d.\n",
2920 		    extv[0]->spd_ext_type));
2921 		spdsock_diag(q, mp, bad_ext_diag[extv[0]->spd_ext_type]);
2922 		return;
2923 	default:
2924 		/* Default case is no errors. */
2925 		break;
2926 	}
2927 
2928 	/*
2929 	 * Special-case SPD_UPDATEALGS so as not to load IPsec.
2930 	 */
2931 	if (!ipsec_loaded(ipss) && spmsg->spd_msg_type != SPD_UPDATEALGS) {
2932 		spdsock_t *ss = (spdsock_t *)q->q_ptr;
2933 
2934 		ASSERT(ss != NULL);
2935 		ipsec_loader_loadnow(ipss);
2936 		ss->spdsock_timeout_arg = mp;
2937 		ss->spdsock_timeout = qtimeout(q, spdsock_loadcheck,
2938 		    q, LOADCHECK_INTERVAL);
2939 		return;
2940 	}
2941 
2942 	/* First check for messages that need no polheads at all. */
2943 	switch (spmsg->spd_msg_type) {
2944 	case SPD_UPDATEALGS:
2945 		spdsock_updatealg(q, mp, extv);
2946 		return;
2947 	case SPD_ALGLIST:
2948 		spdsock_alglist(q, mp);
2949 		return;
2950 	case SPD_DUMPALGS:
2951 		spdsock_dumpalgs(q, mp);
2952 		return;
2953 	}
2954 
2955 	/*
2956 	 * Then check for ones that need both primary/secondary polheads,
2957 	 * finding the appropriate tunnel policy if need be.
2958 	 */
2959 	tunname = (spd_if_t *)extv[SPD_EXT_TUN_NAME];
2960 	switch (spmsg->spd_msg_type) {
2961 	case SPD_FLIP:
2962 		spdsock_flip(q, mp, tunname);
2963 		return;
2964 	case SPD_CLONE:
2965 		spdsock_clone(q, mp, tunname);
2966 		return;
2967 	}
2968 
2969 	/*
2970 	 * Finally, find ones that operate on exactly one polhead, or
2971 	 * "all polheads" of a given type (active/inactive).
2972 	 */
2973 	iph = get_appropriate_polhead(q, mp, tunname, spmsg->spd_msg_spdid,
2974 	    spmsg->spd_msg_type, &itp);
2975 	if (iph == NULL)
2976 		return;
2977 
2978 	/* All-polheads-ready operations. */
2979 	switch (spmsg->spd_msg_type) {
2980 	case SPD_FLUSH:
2981 		if (itp != NULL) {
2982 			mutex_enter(&itp->itp_lock);
2983 			if (spmsg->spd_msg_spdid == SPD_ACTIVE)
2984 				itp->itp_flags &= ~ITPF_PFLAGS;
2985 			else
2986 				itp->itp_flags &= ~ITPF_IFLAGS;
2987 			mutex_exit(&itp->itp_lock);
2988 			ITP_REFRELE(itp, ns);
2989 		}
2990 		spdsock_flush(q, iph, itp, mp);
2991 		return;
2992 	case SPD_DUMP:
2993 		if (itp != NULL)
2994 			ITP_REFRELE(itp, ns);
2995 		spdsock_dump(q, iph, mp);
2996 		return;
2997 	}
2998 
2999 	if (iph == ALL_ACTIVE_POLHEADS || iph == ALL_INACTIVE_POLHEADS) {
3000 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_NOT_GLOBAL_OP);
3001 		return;
3002 	}
3003 
3004 	/* Single-polhead-only operations. */
3005 	switch (spmsg->spd_msg_type) {
3006 	case SPD_ADDRULE:
3007 		spdsock_addrule(q, iph, mp, extv, itp);
3008 		break;
3009 	case SPD_DELETERULE:
3010 		spdsock_deleterule(q, iph, mp, extv, itp);
3011 		break;
3012 	case SPD_LOOKUP:
3013 		spdsock_lookup(q, iph, mp, extv, itp);
3014 		break;
3015 	default:
3016 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_BAD_MSG_TYPE);
3017 		break;
3018 	}
3019 
3020 	IPPH_REFRELE(iph, ns);
3021 	if (itp != NULL)
3022 		ITP_REFRELE(itp, ns);
3023 }
3024 
3025 /*
3026  * If an algorithm mapping was received before IPsec was loaded, process it.
3027  * Called from the IPsec loader.
3028  */
3029 void
3030 spdsock_update_pending_algs(netstack_t *ns)
3031 {
3032 	spd_stack_t *spds = ns->netstack_spdsock;
3033 
3034 	mutex_enter(&spds->spds_alg_lock);
3035 	if (spds->spds_algs_pending) {
3036 		int diag;
3037 
3038 		spdsock_do_updatealg(spds->spds_extv_algs, &diag,
3039 		    spds);
3040 		spds->spds_algs_pending = B_FALSE;
3041 	}
3042 	mutex_exit(&spds->spds_alg_lock);
3043 }
3044 
3045 static void
3046 spdsock_loadcheck(void *arg)
3047 {
3048 	queue_t *q = (queue_t *)arg;
3049 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3050 	mblk_t *mp;
3051 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
3052 
3053 	ASSERT(ss != NULL);
3054 
3055 	ss->spdsock_timeout = 0;
3056 	mp = ss->spdsock_timeout_arg;
3057 	ASSERT(mp != NULL);
3058 	ss->spdsock_timeout_arg = NULL;
3059 	if (ipsec_failed(ipss))
3060 		spdsock_error(q, mp, EPROTONOSUPPORT, 0);
3061 	else
3062 		spdsock_parse(q, mp);
3063 }
3064 
3065 /*
3066  * Copy relevant state bits.
3067  */
3068 static void
3069 spdsock_copy_info(struct T_info_ack *tap, spdsock_t *ss)
3070 {
3071 	*tap = spdsock_g_t_info_ack;
3072 	tap->CURRENT_state = ss->spdsock_state;
3073 	tap->OPT_size = spdsock_max_optsize;
3074 }
3075 
3076 /*
3077  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
3078  * spdsock_wput.  Much of the T_CAPABILITY_ACK information is copied from
3079  * spdsock_g_t_info_ack.  The current state of the stream is copied from
3080  * spdsock_state.
3081  */
3082 static void
3083 spdsock_capability_req(queue_t *q, mblk_t *mp)
3084 {
3085 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3086 	t_uscalar_t cap_bits1;
3087 	struct T_capability_ack	*tcap;
3088 
3089 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
3090 
3091 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
3092 	    mp->b_datap->db_type, T_CAPABILITY_ACK);
3093 	if (mp == NULL)
3094 		return;
3095 
3096 	tcap = (struct T_capability_ack *)mp->b_rptr;
3097 	tcap->CAP_bits1 = 0;
3098 
3099 	if (cap_bits1 & TC1_INFO) {
3100 		spdsock_copy_info(&tcap->INFO_ack, ss);
3101 		tcap->CAP_bits1 |= TC1_INFO;
3102 	}
3103 
3104 	qreply(q, mp);
3105 }
3106 
3107 /*
3108  * This routine responds to T_INFO_REQ messages. It is called by
3109  * spdsock_wput_other.
3110  * Most of the T_INFO_ACK information is copied from spdsock_g_t_info_ack.
3111  * The current state of the stream is copied from spdsock_state.
3112  */
3113 static void
3114 spdsock_info_req(q, mp)
3115 	queue_t	*q;
3116 	mblk_t	*mp;
3117 {
3118 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
3119 	    T_INFO_ACK);
3120 	if (mp == NULL)
3121 		return;
3122 	spdsock_copy_info((struct T_info_ack *)mp->b_rptr,
3123 	    (spdsock_t *)q->q_ptr);
3124 	qreply(q, mp);
3125 }
3126 
3127 /*
3128  * spdsock_err_ack. This routine creates a
3129  * T_ERROR_ACK message and passes it
3130  * upstream.
3131  */
3132 static void
3133 spdsock_err_ack(q, mp, t_error, sys_error)
3134 	queue_t	*q;
3135 	mblk_t	*mp;
3136 	int	t_error;
3137 	int	sys_error;
3138 {
3139 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
3140 		qreply(q, mp);
3141 }
3142 
3143 /*
3144  * This routine retrieves the current status of socket options.
3145  * It returns the size of the option retrieved.
3146  */
3147 /* ARGSUSED */
3148 int
3149 spdsock_opt_get(queue_t *q, int level, int name, uchar_t *ptr)
3150 {
3151 	int *i1 = (int *)ptr;
3152 
3153 	switch (level) {
3154 	case SOL_SOCKET:
3155 		switch (name) {
3156 		case SO_TYPE:
3157 			*i1 = SOCK_RAW;
3158 			break;
3159 		/*
3160 		 * The following two items can be manipulated,
3161 		 * but changing them should do nothing.
3162 		 */
3163 		case SO_SNDBUF:
3164 			*i1 = (int)q->q_hiwat;
3165 			break;
3166 		case SO_RCVBUF:
3167 			*i1 = (int)(RD(q)->q_hiwat);
3168 			break;
3169 		}
3170 		break;
3171 	default:
3172 		return (0);
3173 	}
3174 	return (sizeof (int));
3175 }
3176 
3177 /*
3178  * This routine sets socket options.
3179  */
3180 /* ARGSUSED */
3181 int
3182 spdsock_opt_set(queue_t *q, uint_t mgmt_flags, int level, int name,
3183     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3184     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
3185 {
3186 	int *i1 = (int *)invalp;
3187 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3188 	spd_stack_t	*spds = ss->spdsock_spds;
3189 
3190 	switch (level) {
3191 	case SOL_SOCKET:
3192 		switch (name) {
3193 		case SO_SNDBUF:
3194 			if (*i1 > spds->spds_max_buf)
3195 				return (ENOBUFS);
3196 			q->q_hiwat = *i1;
3197 			break;
3198 		case SO_RCVBUF:
3199 			if (*i1 > spds->spds_max_buf)
3200 				return (ENOBUFS);
3201 			RD(q)->q_hiwat = *i1;
3202 			(void) mi_set_sth_hiwat(RD(q), *i1);
3203 			break;
3204 		}
3205 		break;
3206 	}
3207 	return (0);
3208 }
3209 
3210 
3211 /*
3212  * Handle STREAMS messages.
3213  */
3214 static void
3215 spdsock_wput_other(queue_t *q, mblk_t *mp)
3216 {
3217 	struct iocblk *iocp;
3218 	int error;
3219 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3220 	spd_stack_t	*spds = ss->spdsock_spds;
3221 	cred_t		*cr;
3222 
3223 	switch (mp->b_datap->db_type) {
3224 	case M_PROTO:
3225 	case M_PCPROTO:
3226 		if ((mp->b_wptr - mp->b_rptr) < sizeof (long)) {
3227 			ss3dbg(spds, (
3228 			    "spdsock_wput_other: Not big enough M_PROTO\n"));
3229 			freemsg(mp);
3230 			return;
3231 		}
3232 		cr = zone_get_kcred(netstackid_to_zoneid(
3233 		    spds->spds_netstack->netstack_stackid));
3234 		ASSERT(cr != NULL);
3235 
3236 		switch (((union T_primitives *)mp->b_rptr)->type) {
3237 		case T_CAPABILITY_REQ:
3238 			spdsock_capability_req(q, mp);
3239 			break;
3240 		case T_INFO_REQ:
3241 			spdsock_info_req(q, mp);
3242 			break;
3243 		case T_SVR4_OPTMGMT_REQ:
3244 			(void) svr4_optcom_req(q, mp, DB_CREDDEF(mp, cr),
3245 			    &spdsock_opt_obj, B_FALSE);
3246 			break;
3247 		case T_OPTMGMT_REQ:
3248 			(void) tpi_optcom_req(q, mp, DB_CREDDEF(mp, cr),
3249 			    &spdsock_opt_obj, B_FALSE);
3250 			break;
3251 		case T_DATA_REQ:
3252 		case T_EXDATA_REQ:
3253 		case T_ORDREL_REQ:
3254 			/* Illegal for spdsock. */
3255 			freemsg(mp);
3256 			(void) putnextctl1(RD(q), M_ERROR, EPROTO);
3257 			break;
3258 		default:
3259 			/* Not supported by spdsock. */
3260 			spdsock_err_ack(q, mp, TNOTSUPPORT, 0);
3261 			break;
3262 		}
3263 		crfree(cr);
3264 		return;
3265 	case M_IOCTL:
3266 		iocp = (struct iocblk *)mp->b_rptr;
3267 		error = EINVAL;
3268 
3269 		switch (iocp->ioc_cmd) {
3270 		case ND_SET:
3271 		case ND_GET:
3272 			if (nd_getset(q, spds->spds_g_nd, mp)) {
3273 				qreply(q, mp);
3274 				return;
3275 			} else
3276 				error = ENOENT;
3277 			/* FALLTHRU */
3278 		default:
3279 			miocnak(q, mp, 0, error);
3280 			return;
3281 		}
3282 	case M_FLUSH:
3283 		if (*mp->b_rptr & FLUSHW) {
3284 			flushq(q, FLUSHALL);
3285 			*mp->b_rptr &= ~FLUSHW;
3286 		}
3287 		if (*mp->b_rptr & FLUSHR) {
3288 			qreply(q, mp);
3289 			return;
3290 		}
3291 		/* Else FALLTHRU */
3292 	}
3293 
3294 	/* If fell through, just black-hole the message. */
3295 	freemsg(mp);
3296 }
3297 
3298 static void
3299 spdsock_wput(queue_t *q, mblk_t *mp)
3300 {
3301 	uint8_t *rptr = mp->b_rptr;
3302 	mblk_t *mp1;
3303 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3304 	spd_stack_t	*spds = ss->spdsock_spds;
3305 
3306 	/*
3307 	 * If we're dumping, defer processing other messages until the
3308 	 * dump completes.
3309 	 */
3310 	if (ss->spdsock_dump_req != NULL) {
3311 		if (!putq(q, mp))
3312 			freemsg(mp);
3313 		return;
3314 	}
3315 
3316 	switch (mp->b_datap->db_type) {
3317 	case M_DATA:
3318 		/*
3319 		 * Silently discard.
3320 		 */
3321 		ss2dbg(spds, ("raw M_DATA in spdsock.\n"));
3322 		freemsg(mp);
3323 		return;
3324 	case M_PROTO:
3325 	case M_PCPROTO:
3326 		if ((mp->b_wptr - rptr) >= sizeof (struct T_data_req)) {
3327 			if (((union T_primitives *)rptr)->type == T_DATA_REQ) {
3328 				if ((mp1 = mp->b_cont) == NULL) {
3329 					/* No data after T_DATA_REQ. */
3330 					ss2dbg(spds,
3331 					    ("No data after DATA_REQ.\n"));
3332 					freemsg(mp);
3333 					return;
3334 				}
3335 				freeb(mp);
3336 				mp = mp1;
3337 				ss2dbg(spds, ("T_DATA_REQ\n"));
3338 				break;	/* Out of switch. */
3339 			}
3340 		}
3341 		/* FALLTHRU */
3342 	default:
3343 		ss3dbg(spds, ("In default wput case (%d %d).\n",
3344 		    mp->b_datap->db_type, ((union T_primitives *)rptr)->type));
3345 		spdsock_wput_other(q, mp);
3346 		return;
3347 	}
3348 
3349 	/* I now have a PF_POLICY message in an M_DATA block. */
3350 	spdsock_parse(q, mp);
3351 }
3352 
3353 /*
3354  * Device open procedure, called when new queue pair created.
3355  * We are passed the read-side queue.
3356  */
3357 /* ARGSUSED */
3358 static int
3359 spdsock_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
3360 {
3361 	spdsock_t *ss;
3362 	queue_t *oq = OTHERQ(q);
3363 	minor_t ssminor;
3364 	netstack_t *ns;
3365 	spd_stack_t *spds;
3366 
3367 	if (secpolicy_ip_config(credp, B_FALSE) != 0)
3368 		return (EPERM);
3369 
3370 	if (q->q_ptr != NULL)
3371 		return (0);  /* Re-open of an already open instance. */
3372 
3373 	if (sflag & MODOPEN)
3374 		return (EINVAL);
3375 
3376 	ns = netstack_find_by_cred(credp);
3377 	ASSERT(ns != NULL);
3378 	spds = ns->netstack_spdsock;
3379 	ASSERT(spds != NULL);
3380 
3381 	ss2dbg(spds, ("Made it into PF_POLICY socket open.\n"));
3382 
3383 	ssminor = (minor_t)(uintptr_t)vmem_alloc(spdsock_vmem, 1, VM_NOSLEEP);
3384 	if (ssminor == 0) {
3385 		netstack_rele(spds->spds_netstack);
3386 		return (ENOMEM);
3387 	}
3388 	ss = kmem_zalloc(sizeof (spdsock_t), KM_NOSLEEP);
3389 	if (ss == NULL) {
3390 		vmem_free(spdsock_vmem, (void *)(uintptr_t)ssminor, 1);
3391 		netstack_rele(spds->spds_netstack);
3392 		return (ENOMEM);
3393 	}
3394 
3395 	ss->spdsock_minor = ssminor;
3396 	ss->spdsock_state = TS_UNBND;
3397 	ss->spdsock_dump_req = NULL;
3398 
3399 	ss->spdsock_spds = spds;
3400 
3401 	q->q_ptr = ss;
3402 	oq->q_ptr = ss;
3403 
3404 	q->q_hiwat = spds->spds_recv_hiwat;
3405 
3406 	oq->q_hiwat = spds->spds_xmit_hiwat;
3407 	oq->q_lowat = spds->spds_xmit_lowat;
3408 
3409 	qprocson(q);
3410 	(void) mi_set_sth_hiwat(q, spds->spds_recv_hiwat);
3411 
3412 	*devp = makedevice(getmajor(*devp), ss->spdsock_minor);
3413 	return (0);
3414 }
3415 
3416 /*
3417  * Read-side service procedure, invoked when we get back-enabled
3418  * when buffer space becomes available.
3419  *
3420  * Dump another chunk if we were dumping before; when we finish, kick
3421  * the write-side queue in case it's waiting for read queue space.
3422  */
3423 void
3424 spdsock_rsrv(queue_t *q)
3425 {
3426 	spdsock_t *ss = q->q_ptr;
3427 
3428 	if (ss->spdsock_dump_req != NULL)
3429 		spdsock_dump_some(q, ss);
3430 
3431 	if (ss->spdsock_dump_req == NULL)
3432 		qenable(OTHERQ(q));
3433 }
3434 
3435 /*
3436  * Write-side service procedure, invoked when we defer processing
3437  * if another message is received while a dump is in progress.
3438  */
3439 void
3440 spdsock_wsrv(queue_t *q)
3441 {
3442 	spdsock_t *ss = q->q_ptr;
3443 	mblk_t *mp;
3444 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
3445 
3446 	if (ss->spdsock_dump_req != NULL) {
3447 		qenable(OTHERQ(q));
3448 		return;
3449 	}
3450 
3451 	while ((mp = getq(q)) != NULL) {
3452 		if (ipsec_loaded(ipss)) {
3453 			spdsock_wput(q, mp);
3454 			if (ss->spdsock_dump_req != NULL)
3455 				return;
3456 		} else if (!ipsec_failed(ipss)) {
3457 			(void) putq(q, mp);
3458 		} else {
3459 			spdsock_error(q, mp, EPFNOSUPPORT, 0);
3460 		}
3461 	}
3462 }
3463 
3464 static int
3465 spdsock_close(queue_t *q)
3466 {
3467 	spdsock_t *ss = q->q_ptr;
3468 	spd_stack_t	*spds = ss->spdsock_spds;
3469 
3470 	qprocsoff(q);
3471 
3472 	/* Safe assumption. */
3473 	ASSERT(ss != NULL);
3474 
3475 	if (ss->spdsock_timeout != 0)
3476 		(void) quntimeout(q, ss->spdsock_timeout);
3477 
3478 	ss3dbg(spds, ("Driver close, PF_POLICY socket is going away.\n"));
3479 
3480 	vmem_free(spdsock_vmem, (void *)(uintptr_t)ss->spdsock_minor, 1);
3481 	netstack_rele(ss->spdsock_spds->spds_netstack);
3482 
3483 	kmem_free(ss, sizeof (spdsock_t));
3484 	return (0);
3485 }
3486 
3487 /*
3488  * Merge the IPsec algorithms tables with the received algorithm information.
3489  */
3490 void
3491 spdsock_merge_algs(spd_stack_t *spds)
3492 {
3493 	ipsec_alginfo_t *alg, *oalg;
3494 	ipsec_algtype_t algtype;
3495 	uint_t algidx, algid, nalgs;
3496 	crypto_mech_name_t *mechs;
3497 	uint_t mech_count, mech_idx;
3498 	netstack_t	*ns = spds->spds_netstack;
3499 	ipsec_stack_t	*ipss = ns->netstack_ipsec;
3500 
3501 	ASSERT(MUTEX_HELD(&spds->spds_alg_lock));
3502 
3503 	/*
3504 	 * Get the list of supported mechanisms from the crypto framework.
3505 	 * If a mechanism is supported by KCF, resolve its mechanism
3506 	 * id and mark it as being valid. This operation must be done
3507 	 * without holding alg_lock, since it can cause a provider
3508 	 * module to be loaded and the provider notification callback to
3509 	 * be invoked.
3510 	 */
3511 	mechs = crypto_get_mech_list(&mech_count, KM_SLEEP);
3512 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
3513 		for (algid = 0; algid < IPSEC_MAX_ALGS; algid++) {
3514 			int algflags = 0;
3515 			crypto_mech_type_t mt = CRYPTO_MECHANISM_INVALID;
3516 
3517 			alg = spds->spds_algs[algtype][algid];
3518 			if (alg == NULL)
3519 				continue;
3520 
3521 			/*
3522 			 * The NULL encryption algorithm is a special
3523 			 * case because there are no mechanisms, yet
3524 			 * the algorithm is still valid.
3525 			 */
3526 			if (alg->alg_id == SADB_EALG_NULL) {
3527 				alg->alg_mech_type = CRYPTO_MECHANISM_INVALID;
3528 				alg->alg_flags = ALG_FLAG_VALID;
3529 				continue;
3530 			}
3531 
3532 			for (mech_idx = 0; mech_idx < mech_count; mech_idx++) {
3533 				if (strncmp(alg->alg_mech_name, mechs[mech_idx],
3534 				    CRYPTO_MAX_MECH_NAME) == 0) {
3535 					mt = crypto_mech2id(alg->alg_mech_name);
3536 					ASSERT(mt != CRYPTO_MECHANISM_INVALID);
3537 					algflags = ALG_FLAG_VALID;
3538 					break;
3539 				}
3540 			}
3541 			alg->alg_mech_type = mt;
3542 			alg->alg_flags = algflags;
3543 		}
3544 	}
3545 
3546 	mutex_enter(&ipss->ipsec_alg_lock);
3547 
3548 	/*
3549 	 * For each algorithm currently defined, check if it is
3550 	 * present in the new tables created from the SPD_UPDATEALGS
3551 	 * message received from user-space.
3552 	 * Delete the algorithm entries that are currently defined
3553 	 * but not part of the new tables.
3554 	 */
3555 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
3556 		nalgs = ipss->ipsec_nalgs[algtype];
3557 		for (algidx = 0; algidx < nalgs; algidx++) {
3558 			algid = ipss->ipsec_sortlist[algtype][algidx];
3559 			if (spds->spds_algs[algtype][algid] == NULL)
3560 				ipsec_alg_unreg(algtype, algid, ns);
3561 		}
3562 	}
3563 
3564 	/*
3565 	 * For each algorithm we just received, check if it is
3566 	 * present in the currently defined tables. If it is, swap
3567 	 * the entry with the one we just allocated.
3568 	 * If the new algorithm is not in the current tables,
3569 	 * add it.
3570 	 */
3571 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
3572 		for (algid = 0; algid < IPSEC_MAX_ALGS; algid++) {
3573 			alg = spds->spds_algs[algtype][algid];
3574 			if (alg == NULL)
3575 				continue;
3576 
3577 			if ((oalg = ipss->ipsec_alglists[algtype][algid]) ==
3578 			    NULL) {
3579 				/*
3580 				 * New algorithm, add it to the algorithm
3581 				 * table.
3582 				 */
3583 				ipsec_alg_reg(algtype, alg, ns);
3584 			} else {
3585 				/*
3586 				 * Algorithm is already in the table. Swap
3587 				 * the existing entry with the new one.
3588 				 */
3589 				ipsec_alg_fix_min_max(alg, algtype, ns);
3590 				ipss->ipsec_alglists[algtype][algid] = alg;
3591 				ipsec_alg_free(oalg);
3592 			}
3593 			spds->spds_algs[algtype][algid] = NULL;
3594 		}
3595 	}
3596 
3597 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
3598 		ipss->ipsec_algs_exec_mode[algtype] =
3599 		    spds->spds_algs_exec_mode[algtype];
3600 	}
3601 
3602 	mutex_exit(&ipss->ipsec_alg_lock);
3603 
3604 	crypto_free_mech_list(mechs, mech_count);
3605 
3606 	ipsecah_algs_changed(ns);
3607 	ipsecesp_algs_changed(ns);
3608 }
3609