xref: /titanic_51/usr/src/uts/common/inet/ip/spdsock.c (revision 7ff836697c120cb94bd30d5c2204eb9b74718e4c)
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 2009 Sun Microsystems, Inc.  All rights reserved.
23  * Use is subject to license terms.
24  */
25 
26 #include <sys/param.h>
27 #include <sys/types.h>
28 #include <sys/stream.h>
29 #include <sys/strsubr.h>
30 #include <sys/strsun.h>
31 #include <sys/stropts.h>
32 #include <sys/zone.h>
33 #include <sys/vnode.h>
34 #include <sys/sysmacros.h>
35 #define	_SUN_TPI_VERSION 2
36 #include <sys/tihdr.h>
37 #include <sys/ddi.h>
38 #include <sys/sunddi.h>
39 #include <sys/mkdev.h>
40 #include <sys/debug.h>
41 #include <sys/kmem.h>
42 #include <sys/cmn_err.h>
43 #include <sys/suntpi.h>
44 #include <sys/policy.h>
45 
46 #include <sys/socket.h>
47 #include <netinet/in.h>
48 #include <net/pfkeyv2.h>
49 #include <net/pfpolicy.h>
50 
51 #include <inet/common.h>
52 #include <netinet/ip6.h>
53 #include <inet/ip.h>
54 #include <inet/ip6.h>
55 #include <inet/mi.h>
56 #include <inet/proto_set.h>
57 #include <inet/nd.h>
58 #include <inet/ip_if.h>
59 #include <inet/tun.h>
60 #include <inet/optcom.h>
61 #include <inet/ipsec_info.h>
62 #include <inet/ipsec_impl.h>
63 #include <inet/spdsock.h>
64 #include <inet/sadb.h>
65 
66 #include <sys/isa_defs.h>
67 
68 #include <c2/audit.h>
69 
70 /*
71  * This is a transport provider for the PF_POLICY IPsec policy
72  * management socket, which provides a management interface into the
73  * SPD, allowing policy rules to be added, deleted, and queried.
74  *
75  * This effectively replaces the old private SIOC*IPSECONFIG ioctls
76  * with an extensible interface which will hopefully be public some
77  * day.
78  *
79  * See <net/pfpolicy.h> for more details on the protocol.
80  *
81  * We link against drv/ip and call directly into it to manipulate the
82  * SPD; see ipsec_impl.h for the policy data structures and spd.c for
83  * the code which maintains them.
84  *
85  * The MT model of this is QPAIR with the addition of some explicit
86  * locking to protect system-wide policy data structures.
87  */
88 
89 static vmem_t *spdsock_vmem;		/* for minor numbers. */
90 
91 #define	ALIGNED64(x) IS_P2ALIGNED((x), sizeof (uint64_t))
92 
93 /* Default structure copied into T_INFO_ACK messages (from rts.c...) */
94 static struct T_info_ack spdsock_g_t_info_ack = {
95 	T_INFO_ACK,
96 	T_INFINITE,	/* TSDU_size. Maximum size messages. */
97 	T_INVALID,	/* ETSDU_size. No expedited data. */
98 	T_INVALID,	/* CDATA_size. No connect data. */
99 	T_INVALID,	/* DDATA_size. No disconnect data. */
100 	0,		/* ADDR_size. */
101 	0,		/* OPT_size. No user-settable options */
102 	64 * 1024,	/* TIDU_size. spdsock allows maximum size messages. */
103 	T_COTS,		/* SERV_type. spdsock supports connection oriented. */
104 	TS_UNBND,	/* CURRENT_state. This is set from spdsock_state. */
105 	(XPG4_1)	/* Provider flags */
106 };
107 
108 /* Named Dispatch Parameter Management Structure */
109 typedef struct spdsockparam_s {
110 	uint_t	spdsock_param_min;
111 	uint_t	spdsock_param_max;
112 	uint_t	spdsock_param_value;
113 	char *spdsock_param_name;
114 } spdsockparam_t;
115 
116 /*
117  * Table of NDD variables supported by spdsock. These are loaded into
118  * spdsock_g_nd in spdsock_init_nd.
119  * All of these are alterable, within the min/max values given, at run time.
120  */
121 static	spdsockparam_t	lcl_param_arr[] = {
122 	/* min	max	value	name */
123 	{ 4096, 65536,	8192,	"spdsock_xmit_hiwat"},
124 	{ 0,	65536,	1024,	"spdsock_xmit_lowat"},
125 	{ 4096, 65536,	8192,	"spdsock_recv_hiwat"},
126 	{ 65536, 1024*1024*1024, 256*1024,	"spdsock_max_buf"},
127 	{ 0,	3,	0,	"spdsock_debug"},
128 };
129 #define	spds_xmit_hiwat	spds_params[0].spdsock_param_value
130 #define	spds_xmit_lowat	spds_params[1].spdsock_param_value
131 #define	spds_recv_hiwat	spds_params[2].spdsock_param_value
132 #define	spds_max_buf	spds_params[3].spdsock_param_value
133 #define	spds_debug		spds_params[4].spdsock_param_value
134 
135 #define	ss0dbg(a)	printf a
136 /* NOTE:  != 0 instead of > 0 so lint doesn't complain. */
137 #define	ss1dbg(spds, a)	if (spds->spds_debug != 0) printf a
138 #define	ss2dbg(spds, a)	if (spds->spds_debug > 1) printf a
139 #define	ss3dbg(spds, a)	if (spds->spds_debug > 2) printf a
140 
141 #define	RESET_SPDSOCK_DUMP_POLHEAD(ss, iph) { \
142 	ASSERT(RW_READ_HELD(&(iph)->iph_lock)); \
143 	(ss)->spdsock_dump_head = (iph); \
144 	(ss)->spdsock_dump_gen = (iph)->iph_gen; \
145 	(ss)->spdsock_dump_cur_type = 0; \
146 	(ss)->spdsock_dump_cur_af = IPSEC_AF_V4; \
147 	(ss)->spdsock_dump_cur_rule = NULL; \
148 	(ss)->spdsock_dump_count = 0; \
149 	(ss)->spdsock_dump_cur_chain = 0; \
150 }
151 
152 static int spdsock_close(queue_t *);
153 static int spdsock_open(queue_t *, dev_t *, int, int, cred_t *);
154 static void spdsock_wput(queue_t *, mblk_t *);
155 static void spdsock_wsrv(queue_t *);
156 static void spdsock_rsrv(queue_t *);
157 static void *spdsock_stack_init(netstackid_t stackid, netstack_t *ns);
158 static void spdsock_stack_fini(netstackid_t stackid, void *arg);
159 static void spdsock_loadcheck(void *);
160 static void spdsock_merge_algs(spd_stack_t *);
161 static void spdsock_flush_one(ipsec_policy_head_t *, netstack_t *);
162 static mblk_t *spdsock_dump_next_record(spdsock_t *);
163 
164 static struct module_info info = {
165 	5138, "spdsock", 1, INFPSZ, 512, 128
166 };
167 
168 static struct qinit rinit = {
169 	NULL, (pfi_t)spdsock_rsrv, spdsock_open, spdsock_close,
170 	NULL, &info
171 };
172 
173 static struct qinit winit = {
174 	(pfi_t)spdsock_wput, (pfi_t)spdsock_wsrv, NULL, NULL, NULL, &info
175 };
176 
177 struct streamtab spdsockinfo = {
178 	&rinit, &winit
179 };
180 
181 /* mapping from alg type to protocol number, as per RFC 2407 */
182 static const uint_t algproto[] = {
183 	PROTO_IPSEC_AH,
184 	PROTO_IPSEC_ESP,
185 };
186 
187 #define	NALGPROTOS	(sizeof (algproto) / sizeof (algproto[0]))
188 
189 /* mapping from kernel exec mode to spdsock exec mode */
190 static const uint_t execmodes[] = {
191 	SPD_ALG_EXEC_MODE_SYNC,
192 	SPD_ALG_EXEC_MODE_ASYNC
193 };
194 
195 #define	NEXECMODES	(sizeof (execmodes) / sizeof (execmodes[0]))
196 
197 #define	ALL_ACTIVE_POLHEADS ((ipsec_policy_head_t *)-1)
198 #define	ALL_INACTIVE_POLHEADS ((ipsec_policy_head_t *)-2)
199 
200 #define	ITP_NAME(itp) (itp != NULL ? itp->itp_name : NULL)
201 
202 /* ARGSUSED */
203 static int
204 spdsock_param_get(q, mp, cp, cr)
205 	queue_t	*q;
206 	mblk_t	*mp;
207 	caddr_t	cp;
208 	cred_t *cr;
209 {
210 	spdsockparam_t	*spdsockpa = (spdsockparam_t *)cp;
211 	uint_t value;
212 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
213 	spd_stack_t	*spds = ss->spdsock_spds;
214 
215 	mutex_enter(&spds->spds_param_lock);
216 	value = spdsockpa->spdsock_param_value;
217 	mutex_exit(&spds->spds_param_lock);
218 
219 	(void) mi_mpprintf(mp, "%u", value);
220 	return (0);
221 }
222 
223 /* This routine sets an NDD variable in a spdsockparam_t structure. */
224 /* ARGSUSED */
225 static int
226 spdsock_param_set(q, mp, value, cp, cr)
227 	queue_t	*q;
228 	mblk_t	*mp;
229 	char *value;
230 	caddr_t	cp;
231 	cred_t *cr;
232 {
233 	ulong_t	new_value;
234 	spdsockparam_t	*spdsockpa = (spdsockparam_t *)cp;
235 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
236 	spd_stack_t	*spds = ss->spdsock_spds;
237 
238 	/* Convert the value from a string into a long integer. */
239 	if (ddi_strtoul(value, NULL, 10, &new_value) != 0)
240 		return (EINVAL);
241 
242 	mutex_enter(&spds->spds_param_lock);
243 	/*
244 	 * Fail the request if the new value does not lie within the
245 	 * required bounds.
246 	 */
247 	if (new_value < spdsockpa->spdsock_param_min ||
248 	    new_value > spdsockpa->spdsock_param_max) {
249 		mutex_exit(&spds->spds_param_lock);
250 		return (EINVAL);
251 	}
252 
253 	/* Set the new value */
254 	spdsockpa->spdsock_param_value = new_value;
255 	mutex_exit(&spds->spds_param_lock);
256 
257 	return (0);
258 }
259 
260 /*
261  * Initialize at module load time
262  */
263 boolean_t
264 spdsock_ddi_init(void)
265 {
266 	spdsock_max_optsize = optcom_max_optsize(
267 	    spdsock_opt_obj.odb_opt_des_arr, spdsock_opt_obj.odb_opt_arr_cnt);
268 
269 	spdsock_vmem = vmem_create("spdsock", (void *)1, MAXMIN, 1,
270 	    NULL, NULL, NULL, 1, VM_SLEEP | VMC_IDENTIFIER);
271 
272 	/*
273 	 * We want to be informed each time a stack is created or
274 	 * destroyed in the kernel, so we can maintain the
275 	 * set of spd_stack_t's.
276 	 */
277 	netstack_register(NS_SPDSOCK, spdsock_stack_init, NULL,
278 	    spdsock_stack_fini);
279 
280 	return (B_TRUE);
281 }
282 
283 /*
284  * Walk through the param array specified registering each element with the
285  * named dispatch handler.
286  */
287 static boolean_t
288 spdsock_param_register(IDP *ndp, spdsockparam_t *ssp, int cnt)
289 {
290 	for (; cnt-- > 0; ssp++) {
291 		if (ssp->spdsock_param_name != NULL &&
292 		    ssp->spdsock_param_name[0]) {
293 			if (!nd_load(ndp,
294 			    ssp->spdsock_param_name,
295 			    spdsock_param_get, spdsock_param_set,
296 			    (caddr_t)ssp)) {
297 				nd_free(ndp);
298 				return (B_FALSE);
299 			}
300 		}
301 	}
302 	return (B_TRUE);
303 }
304 
305 /*
306  * Initialize for each stack instance
307  */
308 /* ARGSUSED */
309 static void *
310 spdsock_stack_init(netstackid_t stackid, netstack_t *ns)
311 {
312 	spd_stack_t	*spds;
313 	spdsockparam_t	*ssp;
314 
315 	spds = (spd_stack_t *)kmem_zalloc(sizeof (*spds), KM_SLEEP);
316 	spds->spds_netstack = ns;
317 
318 	ASSERT(spds->spds_g_nd == NULL);
319 
320 	ssp = (spdsockparam_t *)kmem_alloc(sizeof (lcl_param_arr), KM_SLEEP);
321 	spds->spds_params = ssp;
322 	bcopy(lcl_param_arr, ssp, sizeof (lcl_param_arr));
323 
324 	(void) spdsock_param_register(&spds->spds_g_nd, ssp,
325 	    A_CNT(lcl_param_arr));
326 
327 	mutex_init(&spds->spds_param_lock, NULL, MUTEX_DEFAULT, NULL);
328 	mutex_init(&spds->spds_alg_lock, NULL, MUTEX_DEFAULT, NULL);
329 
330 	return (spds);
331 }
332 
333 void
334 spdsock_ddi_destroy(void)
335 {
336 	vmem_destroy(spdsock_vmem);
337 
338 	netstack_unregister(NS_SPDSOCK);
339 }
340 
341 /* ARGSUSED */
342 static void
343 spdsock_stack_fini(netstackid_t stackid, void *arg)
344 {
345 	spd_stack_t *spds = (spd_stack_t *)arg;
346 
347 	freemsg(spds->spds_mp_algs);
348 	mutex_destroy(&spds->spds_param_lock);
349 	mutex_destroy(&spds->spds_alg_lock);
350 	nd_free(&spds->spds_g_nd);
351 	kmem_free(spds->spds_params, sizeof (lcl_param_arr));
352 	spds->spds_params = NULL;
353 
354 	kmem_free(spds, sizeof (*spds));
355 }
356 
357 /*
358  * NOTE: large quantities of this should be shared with keysock.
359  * Would be nice to combine some of this into a common module, but
360  * not possible given time pressures.
361  */
362 
363 /*
364  * High-level reality checking of extensions.
365  */
366 /* ARGSUSED */ /* XXX */
367 static boolean_t
368 ext_check(spd_ext_t *ext)
369 {
370 	spd_if_t *tunname = (spd_if_t *)ext;
371 	int i;
372 	char *idstr;
373 
374 	if (ext->spd_ext_type == SPD_EXT_TUN_NAME) {
375 		/* (NOTE:  Modified from SADB_EXT_IDENTITY..) */
376 
377 		/*
378 		 * Make sure the strings in these identities are
379 		 * null-terminated.  Let's "proactively" null-terminate the
380 		 * string at the last byte if it's not terminated sooner.
381 		 */
382 		i = SPD_64TO8(tunname->spd_if_len) - sizeof (spd_if_t);
383 		idstr = (char *)(tunname + 1);
384 		while (*idstr != '\0' && i > 0) {
385 			i--;
386 			idstr++;
387 		}
388 		if (i == 0) {
389 			/*
390 			 * I.e., if the bozo user didn't NULL-terminate the
391 			 * string...
392 			 */
393 			idstr--;
394 			*idstr = '\0';
395 		}
396 	}
397 	return (B_TRUE);	/* For now... */
398 }
399 
400 
401 
402 /* Return values for spdsock_get_ext(). */
403 #define	KGE_OK	0
404 #define	KGE_DUP	1
405 #define	KGE_UNK	2
406 #define	KGE_LEN	3
407 #define	KGE_CHK	4
408 
409 /*
410  * Parse basic extension headers and return in the passed-in pointer vector.
411  * Return values include:
412  *
413  *	KGE_OK	Everything's nice and parsed out.
414  *		If there are no extensions, place NULL in extv[0].
415  *	KGE_DUP	There is a duplicate extension.
416  *		First instance in appropriate bin.  First duplicate in
417  *		extv[0].
418  *	KGE_UNK	Unknown extension type encountered.  extv[0] contains
419  *		unknown header.
420  *	KGE_LEN	Extension length error.
421  *	KGE_CHK	High-level reality check failed on specific extension.
422  *
423  * My apologies for some of the pointer arithmetic in here.  I'm thinking
424  * like an assembly programmer, yet trying to make the compiler happy.
425  */
426 static int
427 spdsock_get_ext(spd_ext_t *extv[], spd_msg_t *basehdr, uint_t msgsize)
428 {
429 	bzero(extv, sizeof (spd_ext_t *) * (SPD_EXT_MAX + 1));
430 
431 	/* Use extv[0] as the "current working pointer". */
432 
433 	extv[0] = (spd_ext_t *)(basehdr + 1);
434 
435 	while (extv[0] < (spd_ext_t *)(((uint8_t *)basehdr) + msgsize)) {
436 		/* Check for unknown headers. */
437 		if (extv[0]->spd_ext_type == 0 ||
438 		    extv[0]->spd_ext_type > SPD_EXT_MAX)
439 			return (KGE_UNK);
440 
441 		/*
442 		 * Check length.  Use uint64_t because extlen is in units
443 		 * of 64-bit words.  If length goes beyond the msgsize,
444 		 * return an error.  (Zero length also qualifies here.)
445 		 */
446 		if (extv[0]->spd_ext_len == 0 ||
447 		    (void *)((uint64_t *)extv[0] + extv[0]->spd_ext_len) >
448 		    (void *)((uint8_t *)basehdr + msgsize))
449 			return (KGE_LEN);
450 
451 		/* Check for redundant headers. */
452 		if (extv[extv[0]->spd_ext_type] != NULL)
453 			return (KGE_DUP);
454 
455 		/*
456 		 * Reality check the extension if possible at the spdsock
457 		 * level.
458 		 */
459 		if (!ext_check(extv[0]))
460 			return (KGE_CHK);
461 
462 		/* If I make it here, assign the appropriate bin. */
463 		extv[extv[0]->spd_ext_type] = extv[0];
464 
465 		/* Advance pointer (See above for uint64_t ptr reasoning.) */
466 		extv[0] = (spd_ext_t *)
467 		    ((uint64_t *)extv[0] + extv[0]->spd_ext_len);
468 	}
469 
470 	/* Everything's cool. */
471 
472 	/*
473 	 * If extv[0] == NULL, then there are no extension headers in this
474 	 * message.  Ensure that this is the case.
475 	 */
476 	if (extv[0] == (spd_ext_t *)(basehdr + 1))
477 		extv[0] = NULL;
478 
479 	return (KGE_OK);
480 }
481 
482 static const int bad_ext_diag[] = {
483 	SPD_DIAGNOSTIC_MALFORMED_LCLPORT,
484 	SPD_DIAGNOSTIC_MALFORMED_REMPORT,
485 	SPD_DIAGNOSTIC_MALFORMED_PROTO,
486 	SPD_DIAGNOSTIC_MALFORMED_LCLADDR,
487 	SPD_DIAGNOSTIC_MALFORMED_REMADDR,
488 	SPD_DIAGNOSTIC_MALFORMED_ACTION,
489 	SPD_DIAGNOSTIC_MALFORMED_RULE,
490 	SPD_DIAGNOSTIC_MALFORMED_RULESET,
491 	SPD_DIAGNOSTIC_MALFORMED_ICMP_TYPECODE
492 };
493 
494 static const int dup_ext_diag[] = {
495 	SPD_DIAGNOSTIC_DUPLICATE_LCLPORT,
496 	SPD_DIAGNOSTIC_DUPLICATE_REMPORT,
497 	SPD_DIAGNOSTIC_DUPLICATE_PROTO,
498 	SPD_DIAGNOSTIC_DUPLICATE_LCLADDR,
499 	SPD_DIAGNOSTIC_DUPLICATE_REMADDR,
500 	SPD_DIAGNOSTIC_DUPLICATE_ACTION,
501 	SPD_DIAGNOSTIC_DUPLICATE_RULE,
502 	SPD_DIAGNOSTIC_DUPLICATE_RULESET,
503 	SPD_DIAGNOSTIC_DUPLICATE_ICMP_TYPECODE
504 };
505 
506 /*
507  * Transmit a PF_POLICY error message to the instance either pointed to
508  * by ks, the instance with serial number serial, or more, depending.
509  *
510  * The faulty message (or a reasonable facsimile thereof) is in mp.
511  * This function will free mp or recycle it for delivery, thereby causing
512  * the stream head to free it.
513  */
514 static void
515 spdsock_error(queue_t *q, mblk_t *mp, int error, int diagnostic)
516 {
517 	spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
518 
519 	ASSERT(mp->b_datap->db_type == M_DATA);
520 
521 	if (spmsg->spd_msg_type < SPD_MIN ||
522 	    spmsg->spd_msg_type > SPD_MAX)
523 		spmsg->spd_msg_type = SPD_RESERVED;
524 
525 	/*
526 	 * Strip out extension headers.
527 	 */
528 	ASSERT(mp->b_rptr + sizeof (*spmsg) <= mp->b_datap->db_lim);
529 	mp->b_wptr = mp->b_rptr + sizeof (*spmsg);
530 	spmsg->spd_msg_len = SPD_8TO64(sizeof (spd_msg_t));
531 	spmsg->spd_msg_errno = (uint8_t)error;
532 	spmsg->spd_msg_diagnostic = (uint16_t)diagnostic;
533 
534 	qreply(q, mp);
535 }
536 
537 static void
538 spdsock_diag(queue_t *q, mblk_t *mp, int diagnostic)
539 {
540 	spdsock_error(q, mp, EINVAL, diagnostic);
541 }
542 
543 static void
544 spd_echo(queue_t *q, mblk_t *mp)
545 {
546 	qreply(q, mp);
547 }
548 
549 /*
550  * Do NOT consume a reference to itp.
551  */
552 /*ARGSUSED*/
553 static void
554 spdsock_flush_node(ipsec_tun_pol_t *itp, void *cookie, netstack_t *ns)
555 {
556 	boolean_t active = (boolean_t)cookie;
557 	ipsec_policy_head_t *iph;
558 
559 	iph = active ? itp->itp_policy : itp->itp_inactive;
560 	IPPH_REFHOLD(iph);
561 	mutex_enter(&itp->itp_lock);
562 	spdsock_flush_one(iph, ns);
563 	if (active)
564 		itp->itp_flags &= ~ITPF_PFLAGS;
565 	else
566 		itp->itp_flags &= ~ITPF_IFLAGS;
567 	mutex_exit(&itp->itp_lock);
568 }
569 
570 /*
571  * Clear out one polhead.
572  */
573 static void
574 spdsock_flush_one(ipsec_policy_head_t *iph, netstack_t *ns)
575 {
576 	rw_enter(&iph->iph_lock, RW_WRITER);
577 	ipsec_polhead_flush(iph, ns);
578 	rw_exit(&iph->iph_lock);
579 	IPPH_REFRELE(iph, ns);
580 }
581 
582 static void
583 spdsock_flush(queue_t *q, ipsec_policy_head_t *iph, ipsec_tun_pol_t *itp,
584     mblk_t *mp)
585 {
586 	boolean_t active;
587 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
588 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
589 
590 	if (iph != ALL_ACTIVE_POLHEADS && iph != ALL_INACTIVE_POLHEADS) {
591 		spdsock_flush_one(iph, ns);
592 		if (audit_active) {
593 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
594 			cred_t *cr;
595 			pid_t cpid;
596 
597 			cr = msg_getcred(mp, &cpid);
598 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
599 			audit_pf_policy(SPD_FLUSH, cr, ns,
600 			    ITP_NAME(itp), active, 0, cpid);
601 		}
602 	} else {
603 		active = (iph == ALL_ACTIVE_POLHEADS);
604 
605 		/* First flush the global policy. */
606 		spdsock_flush_one(active ? ipsec_system_policy(ns) :
607 		    ipsec_inactive_policy(ns), ns);
608 		if (audit_active) {
609 			cred_t *cr;
610 			pid_t cpid;
611 
612 			cr = msg_getcred(mp, &cpid);
613 			audit_pf_policy(SPD_FLUSH, cr, ns, NULL,
614 			    active, 0, cpid);
615 		}
616 		/* Then flush every tunnel's appropriate one. */
617 		itp_walk(spdsock_flush_node, (void *)active, ns);
618 		if (audit_active) {
619 			cred_t *cr;
620 			pid_t cpid;
621 
622 			cr = msg_getcred(mp, &cpid);
623 			audit_pf_policy(SPD_FLUSH, cr, ns,
624 			    "all tunnels", active, 0, cpid);
625 		}
626 	}
627 
628 	spd_echo(q, mp);
629 }
630 
631 static boolean_t
632 spdsock_ext_to_sel(spd_ext_t **extv, ipsec_selkey_t *sel, int *diag)
633 {
634 	bzero(sel, sizeof (*sel));
635 
636 	if (extv[SPD_EXT_PROTO] != NULL) {
637 		struct spd_proto *pr =
638 		    (struct spd_proto *)extv[SPD_EXT_PROTO];
639 		sel->ipsl_proto = pr->spd_proto_number;
640 		sel->ipsl_valid |= IPSL_PROTOCOL;
641 	}
642 	if (extv[SPD_EXT_LCLPORT] != NULL) {
643 		struct spd_portrange *pr =
644 		    (struct spd_portrange *)extv[SPD_EXT_LCLPORT];
645 		sel->ipsl_lport = pr->spd_ports_minport;
646 		sel->ipsl_valid |= IPSL_LOCAL_PORT;
647 	}
648 	if (extv[SPD_EXT_REMPORT] != NULL) {
649 		struct spd_portrange *pr =
650 		    (struct spd_portrange *)extv[SPD_EXT_REMPORT];
651 		sel->ipsl_rport = pr->spd_ports_minport;
652 		sel->ipsl_valid |= IPSL_REMOTE_PORT;
653 	}
654 
655 	if (extv[SPD_EXT_ICMP_TYPECODE] != NULL) {
656 		struct spd_typecode *tc=
657 		    (struct spd_typecode *)extv[SPD_EXT_ICMP_TYPECODE];
658 
659 		sel->ipsl_valid |= IPSL_ICMP_TYPE;
660 		sel->ipsl_icmp_type = tc->spd_typecode_type;
661 		if (tc->spd_typecode_type_end < tc->spd_typecode_type)
662 			sel->ipsl_icmp_type_end = tc->spd_typecode_type;
663 		else
664 			sel->ipsl_icmp_type_end = tc->spd_typecode_type_end;
665 
666 		if (tc->spd_typecode_code != 255) {
667 			sel->ipsl_valid |= IPSL_ICMP_CODE;
668 			sel->ipsl_icmp_code = tc->spd_typecode_code;
669 			if (tc->spd_typecode_code_end < tc->spd_typecode_code)
670 				sel->ipsl_icmp_code_end = tc->spd_typecode_code;
671 			else
672 				sel->ipsl_icmp_code_end =
673 				    tc->spd_typecode_code_end;
674 		}
675 	}
676 #define	ADDR2SEL(sel, extv, field, pfield, extn, bit)			      \
677 	if ((extv)[(extn)] != NULL) {					      \
678 		uint_t addrlen;						      \
679 		struct spd_address *ap = 				      \
680 			(struct spd_address *)((extv)[(extn)]); 	      \
681 		addrlen = (ap->spd_address_af == AF_INET6) ? 		      \
682 			IPV6_ADDR_LEN : IP_ADDR_LEN;			      \
683 		if (SPD_64TO8(ap->spd_address_len) < 			      \
684 			(addrlen + sizeof (*ap))) {			      \
685 			*diag = SPD_DIAGNOSTIC_BAD_ADDR_LEN;		      \
686 			return (B_FALSE);				      \
687 		}							      \
688 		bcopy((ap+1), &((sel)->field), addrlen);		      \
689 		(sel)->pfield = ap->spd_address_prefixlen;		      \
690 		(sel)->ipsl_valid |= (bit);				      \
691 		(sel)->ipsl_valid |= (ap->spd_address_af == AF_INET6) ?	      \
692 			IPSL_IPV6 : IPSL_IPV4;				      \
693 	}
694 
695 	ADDR2SEL(sel, extv, ipsl_local, ipsl_local_pfxlen,
696 	    SPD_EXT_LCLADDR, IPSL_LOCAL_ADDR);
697 	ADDR2SEL(sel, extv, ipsl_remote, ipsl_remote_pfxlen,
698 	    SPD_EXT_REMADDR, IPSL_REMOTE_ADDR);
699 
700 	if ((sel->ipsl_valid & (IPSL_IPV6|IPSL_IPV4)) ==
701 	    (IPSL_IPV6|IPSL_IPV4)) {
702 		*diag = SPD_DIAGNOSTIC_MIXED_AF;
703 		return (B_FALSE);
704 	}
705 
706 #undef ADDR2SEL
707 
708 	return (B_TRUE);
709 }
710 
711 static boolean_t
712 spd_convert_type(uint32_t type, ipsec_act_t *act)
713 {
714 	switch (type) {
715 	case SPD_ACTTYPE_DROP:
716 		act->ipa_type = IPSEC_ACT_DISCARD;
717 		return (B_TRUE);
718 
719 	case SPD_ACTTYPE_PASS:
720 		act->ipa_type = IPSEC_ACT_CLEAR;
721 		return (B_TRUE);
722 
723 	case SPD_ACTTYPE_IPSEC:
724 		act->ipa_type = IPSEC_ACT_APPLY;
725 		return (B_TRUE);
726 	}
727 	return (B_FALSE);
728 }
729 
730 static boolean_t
731 spd_convert_flags(uint32_t flags, ipsec_act_t *act)
732 {
733 	/*
734 	 * Note use of !! for boolean canonicalization.
735 	 */
736 	act->ipa_apply.ipp_use_ah = !!(flags & SPD_APPLY_AH);
737 	act->ipa_apply.ipp_use_esp = !!(flags & SPD_APPLY_ESP);
738 	act->ipa_apply.ipp_use_espa = !!(flags & SPD_APPLY_ESPA);
739 	act->ipa_apply.ipp_use_se = !!(flags & SPD_APPLY_SE);
740 	act->ipa_apply.ipp_use_unique = !!(flags & SPD_APPLY_UNIQUE);
741 	return (B_TRUE);
742 }
743 
744 static void
745 spdsock_reset_act(ipsec_act_t *act)
746 {
747 	bzero(act, sizeof (*act));
748 	act->ipa_apply.ipp_espe_maxbits = IPSEC_MAX_KEYBITS;
749 	act->ipa_apply.ipp_espa_maxbits = IPSEC_MAX_KEYBITS;
750 	act->ipa_apply.ipp_ah_maxbits = IPSEC_MAX_KEYBITS;
751 }
752 
753 /*
754  * Sanity check action against reality, and shrink-wrap key sizes..
755  */
756 static boolean_t
757 spdsock_check_action(ipsec_act_t *act, boolean_t tunnel_polhead, int *diag,
758     spd_stack_t *spds)
759 {
760 	if (tunnel_polhead && act->ipa_apply.ipp_use_unique) {
761 		*diag = SPD_DIAGNOSTIC_ADD_INCON_FLAGS;
762 		return (B_FALSE);
763 	}
764 	if ((act->ipa_type != IPSEC_ACT_APPLY) &&
765 	    (act->ipa_apply.ipp_use_ah ||
766 	    act->ipa_apply.ipp_use_esp ||
767 	    act->ipa_apply.ipp_use_espa ||
768 	    act->ipa_apply.ipp_use_se ||
769 	    act->ipa_apply.ipp_use_unique)) {
770 		*diag = SPD_DIAGNOSTIC_ADD_INCON_FLAGS;
771 		return (B_FALSE);
772 	}
773 	if ((act->ipa_type == IPSEC_ACT_APPLY) &&
774 	    !act->ipa_apply.ipp_use_ah &&
775 	    !act->ipa_apply.ipp_use_esp) {
776 		*diag = SPD_DIAGNOSTIC_ADD_INCON_FLAGS;
777 		return (B_FALSE);
778 	}
779 	return (ipsec_check_action(act, diag, spds->spds_netstack));
780 }
781 
782 /*
783  * We may be short a few error checks here..
784  */
785 static boolean_t
786 spdsock_ext_to_actvec(spd_ext_t **extv, ipsec_act_t **actpp, uint_t *nactp,
787     int *diag, spd_stack_t *spds)
788 {
789 	struct spd_ext_actions *sactp =
790 	    (struct spd_ext_actions *)extv[SPD_EXT_ACTION];
791 	ipsec_act_t act, *actp, *endactp;
792 	struct spd_attribute *attrp, *endattrp;
793 	uint64_t *endp;
794 	int nact;
795 	boolean_t tunnel_polhead;
796 
797 	tunnel_polhead = (extv[SPD_EXT_TUN_NAME] != NULL &&
798 	    (((struct spd_rule *)extv[SPD_EXT_RULE])->spd_rule_flags &
799 	    SPD_RULE_FLAG_TUNNEL));
800 
801 	*actpp = NULL;
802 	*nactp = 0;
803 
804 	if (sactp == NULL) {
805 		*diag = SPD_DIAGNOSTIC_NO_ACTION_EXT;
806 		return (B_FALSE);
807 	}
808 
809 	/*
810 	 * Parse the "action" extension and convert into an action chain.
811 	 */
812 
813 	nact = sactp->spd_actions_count;
814 
815 	endp = (uint64_t *)sactp;
816 	endp += sactp->spd_actions_len;
817 	endattrp = (struct spd_attribute *)endp;
818 
819 	actp = kmem_alloc(sizeof (*actp) * nact, KM_NOSLEEP);
820 	if (actp == NULL) {
821 		*diag = SPD_DIAGNOSTIC_ADD_NO_MEM;
822 		return (B_FALSE);
823 	}
824 	*actpp = actp;
825 	*nactp = nact;
826 	endactp = actp + nact;
827 
828 	spdsock_reset_act(&act);
829 	attrp = (struct spd_attribute *)(&sactp[1]);
830 
831 	for (; attrp < endattrp; attrp++) {
832 		switch (attrp->spd_attr_tag) {
833 		case SPD_ATTR_NOP:
834 			break;
835 
836 		case SPD_ATTR_EMPTY:
837 			spdsock_reset_act(&act);
838 			break;
839 
840 		case SPD_ATTR_END:
841 			attrp = endattrp;
842 			/* FALLTHRU */
843 		case SPD_ATTR_NEXT:
844 			if (actp >= endactp) {
845 				*diag = SPD_DIAGNOSTIC_ADD_WRONG_ACT_COUNT;
846 				goto fail;
847 			}
848 			if (!spdsock_check_action(&act, tunnel_polhead,
849 			    diag, spds))
850 				goto fail;
851 			*actp++ = act;
852 			spdsock_reset_act(&act);
853 			break;
854 
855 		case SPD_ATTR_TYPE:
856 			if (!spd_convert_type(attrp->spd_attr_value, &act)) {
857 				*diag = SPD_DIAGNOSTIC_ADD_BAD_TYPE;
858 				goto fail;
859 			}
860 			break;
861 
862 		case SPD_ATTR_FLAGS:
863 			if (!tunnel_polhead && extv[SPD_EXT_TUN_NAME] != NULL) {
864 				/*
865 				 * Set "sa unique" for transport-mode
866 				 * tunnels whether we want to or not.
867 				 */
868 				attrp->spd_attr_value |= SPD_APPLY_UNIQUE;
869 			}
870 			if (!spd_convert_flags(attrp->spd_attr_value, &act)) {
871 				*diag = SPD_DIAGNOSTIC_ADD_BAD_FLAGS;
872 				goto fail;
873 			}
874 			break;
875 
876 		case SPD_ATTR_AH_AUTH:
877 			if (attrp->spd_attr_value == 0) {
878 				*diag = SPD_DIAGNOSTIC_UNSUPP_AH_ALG;
879 				goto fail;
880 			}
881 			act.ipa_apply.ipp_auth_alg = attrp->spd_attr_value;
882 			break;
883 
884 		case SPD_ATTR_ESP_ENCR:
885 			if (attrp->spd_attr_value == 0) {
886 				*diag = SPD_DIAGNOSTIC_UNSUPP_ESP_ENCR_ALG;
887 				goto fail;
888 			}
889 			act.ipa_apply.ipp_encr_alg = attrp->spd_attr_value;
890 			break;
891 
892 		case SPD_ATTR_ESP_AUTH:
893 			if (attrp->spd_attr_value == 0) {
894 				*diag = SPD_DIAGNOSTIC_UNSUPP_ESP_AUTH_ALG;
895 				goto fail;
896 			}
897 			act.ipa_apply.ipp_esp_auth_alg = attrp->spd_attr_value;
898 			break;
899 
900 		case SPD_ATTR_ENCR_MINBITS:
901 			act.ipa_apply.ipp_espe_minbits = attrp->spd_attr_value;
902 			break;
903 
904 		case SPD_ATTR_ENCR_MAXBITS:
905 			act.ipa_apply.ipp_espe_maxbits = attrp->spd_attr_value;
906 			break;
907 
908 		case SPD_ATTR_AH_MINBITS:
909 			act.ipa_apply.ipp_ah_minbits = attrp->spd_attr_value;
910 			break;
911 
912 		case SPD_ATTR_AH_MAXBITS:
913 			act.ipa_apply.ipp_ah_maxbits = attrp->spd_attr_value;
914 			break;
915 
916 		case SPD_ATTR_ESPA_MINBITS:
917 			act.ipa_apply.ipp_espa_minbits = attrp->spd_attr_value;
918 			break;
919 
920 		case SPD_ATTR_ESPA_MAXBITS:
921 			act.ipa_apply.ipp_espa_maxbits = attrp->spd_attr_value;
922 			break;
923 
924 		case SPD_ATTR_LIFE_SOFT_TIME:
925 		case SPD_ATTR_LIFE_HARD_TIME:
926 		case SPD_ATTR_LIFE_SOFT_BYTES:
927 		case SPD_ATTR_LIFE_HARD_BYTES:
928 			break;
929 
930 		case SPD_ATTR_KM_PROTO:
931 			act.ipa_apply.ipp_km_proto = attrp->spd_attr_value;
932 			break;
933 
934 		case SPD_ATTR_KM_COOKIE:
935 			act.ipa_apply.ipp_km_cookie = attrp->spd_attr_value;
936 			break;
937 
938 		case SPD_ATTR_REPLAY_DEPTH:
939 			act.ipa_apply.ipp_replay_depth = attrp->spd_attr_value;
940 			break;
941 		}
942 	}
943 	if (actp != endactp) {
944 		*diag = SPD_DIAGNOSTIC_ADD_WRONG_ACT_COUNT;
945 		goto fail;
946 	}
947 
948 	return (B_TRUE);
949 fail:
950 	ipsec_actvec_free(*actpp, nact);
951 	*actpp = NULL;
952 	return (B_FALSE);
953 }
954 
955 typedef struct
956 {
957 	ipsec_policy_t *pol;
958 	int dir;
959 } tmprule_t;
960 
961 static int
962 mkrule(ipsec_policy_head_t *iph, struct spd_rule *rule,
963     ipsec_selkey_t *sel, ipsec_act_t *actp, int nact, uint_t dir, uint_t af,
964     tmprule_t **rp, uint64_t *index, spd_stack_t *spds)
965 {
966 	ipsec_policy_t *pol;
967 
968 	sel->ipsl_valid &= ~(IPSL_IPV6|IPSL_IPV4);
969 	sel->ipsl_valid |= af;
970 
971 	pol = ipsec_policy_create(sel, actp, nact, rule->spd_rule_priority,
972 	    index, spds->spds_netstack);
973 	if (pol == NULL)
974 		return (ENOMEM);
975 
976 	(*rp)->pol = pol;
977 	(*rp)->dir = dir;
978 	(*rp)++;
979 
980 	if (!ipsec_check_policy(iph, pol, dir))
981 		return (EEXIST);
982 
983 	rule->spd_rule_index = pol->ipsp_index;
984 	return (0);
985 }
986 
987 static int
988 mkrulepair(ipsec_policy_head_t *iph, struct spd_rule *rule,
989     ipsec_selkey_t *sel, ipsec_act_t *actp, int nact, uint_t dir, uint_t afs,
990     tmprule_t **rp, uint64_t *index, spd_stack_t *spds)
991 {
992 	int error;
993 
994 	if (afs & IPSL_IPV4) {
995 		error = mkrule(iph, rule, sel, actp, nact, dir, IPSL_IPV4, rp,
996 		    index, spds);
997 		if (error != 0)
998 			return (error);
999 	}
1000 	if (afs & IPSL_IPV6) {
1001 		error = mkrule(iph, rule, sel, actp, nact, dir, IPSL_IPV6, rp,
1002 		    index, spds);
1003 		if (error != 0)
1004 			return (error);
1005 	}
1006 	return (0);
1007 }
1008 
1009 
1010 static void
1011 spdsock_addrule(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp,
1012     spd_ext_t **extv, ipsec_tun_pol_t *itp)
1013 {
1014 	ipsec_selkey_t sel;
1015 	ipsec_act_t *actp;
1016 	uint_t nact;
1017 	int diag = 0, error, afs;
1018 	struct spd_rule *rule = (struct spd_rule *)extv[SPD_EXT_RULE];
1019 	tmprule_t rules[4], *rulep = &rules[0];
1020 	boolean_t tunnel_mode, empty_itp, active;
1021 	uint64_t *index = (itp == NULL) ? NULL : &itp->itp_next_policy_index;
1022 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
1023 	spd_stack_t	*spds = ss->spdsock_spds;
1024 
1025 	if (rule == NULL) {
1026 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_NO_RULE_EXT);
1027 		if (audit_active) {
1028 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1029 			cred_t *cr;
1030 			pid_t cpid;
1031 
1032 			cr = msg_getcred(mp, &cpid);
1033 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1034 			audit_pf_policy(SPD_ADDRULE, cr,
1035 			    spds->spds_netstack, ITP_NAME(itp), active,
1036 			    SPD_DIAGNOSTIC_NO_RULE_EXT, cpid);
1037 		}
1038 		return;
1039 	}
1040 
1041 	tunnel_mode = (rule->spd_rule_flags & SPD_RULE_FLAG_TUNNEL);
1042 
1043 	if (itp != NULL) {
1044 		mutex_enter(&itp->itp_lock);
1045 		ASSERT(itp->itp_policy == iph || itp->itp_inactive == iph);
1046 		active = (itp->itp_policy == iph);
1047 		if (ITP_P_ISACTIVE(itp, iph)) {
1048 			/* Check for mix-and-match of tunnel/transport. */
1049 			if ((tunnel_mode && !ITP_P_ISTUNNEL(itp, iph)) ||
1050 			    (!tunnel_mode && ITP_P_ISTUNNEL(itp, iph))) {
1051 				mutex_exit(&itp->itp_lock);
1052 				spdsock_error(q, mp, EBUSY, 0);
1053 				return;
1054 			}
1055 			empty_itp = B_FALSE;
1056 		} else {
1057 			empty_itp = B_TRUE;
1058 			itp->itp_flags = active ? ITPF_P_ACTIVE : ITPF_I_ACTIVE;
1059 			if (tunnel_mode)
1060 				itp->itp_flags |= active ? ITPF_P_TUNNEL :
1061 				    ITPF_I_TUNNEL;
1062 		}
1063 	} else {
1064 		empty_itp = B_FALSE;
1065 	}
1066 
1067 	if (rule->spd_rule_index != 0) {
1068 		diag = SPD_DIAGNOSTIC_INVALID_RULE_INDEX;
1069 		error = EINVAL;
1070 		goto fail2;
1071 	}
1072 
1073 	if (!spdsock_ext_to_sel(extv, &sel, &diag)) {
1074 		error = EINVAL;
1075 		goto fail2;
1076 	}
1077 
1078 	if (itp != NULL) {
1079 		if (tunnel_mode) {
1080 			if (sel.ipsl_valid &
1081 			    (IPSL_REMOTE_PORT | IPSL_LOCAL_PORT)) {
1082 				itp->itp_flags |= active ?
1083 				    ITPF_P_PER_PORT_SECURITY :
1084 				    ITPF_I_PER_PORT_SECURITY;
1085 			}
1086 		} else {
1087 			/*
1088 			 * For now, we don't allow transport-mode on a tunnel
1089 			 * with ANY specific selectors.  Bail if we have such
1090 			 * a request.
1091 			 */
1092 			if (sel.ipsl_valid & IPSL_WILDCARD) {
1093 				diag = SPD_DIAGNOSTIC_NO_TUNNEL_SELECTORS;
1094 				error = EINVAL;
1095 				goto fail2;
1096 			}
1097 		}
1098 	}
1099 
1100 	if (!spdsock_ext_to_actvec(extv, &actp, &nact, &diag, spds)) {
1101 		error = EINVAL;
1102 		goto fail2;
1103 	}
1104 	/*
1105 	 * If no addresses were specified, add both.
1106 	 */
1107 	afs = sel.ipsl_valid & (IPSL_IPV6|IPSL_IPV4);
1108 	if (afs == 0)
1109 		afs = (IPSL_IPV6|IPSL_IPV4);
1110 
1111 	rw_enter(&iph->iph_lock, RW_WRITER);
1112 
1113 	if (rule->spd_rule_flags & SPD_RULE_FLAG_OUTBOUND) {
1114 		error = mkrulepair(iph, rule, &sel, actp, nact,
1115 		    IPSEC_TYPE_OUTBOUND, afs, &rulep, index, spds);
1116 		if (error != 0)
1117 			goto fail;
1118 	}
1119 
1120 	if (rule->spd_rule_flags & SPD_RULE_FLAG_INBOUND) {
1121 		error = mkrulepair(iph, rule, &sel, actp, nact,
1122 		    IPSEC_TYPE_INBOUND, afs, &rulep, index, spds);
1123 		if (error != 0)
1124 			goto fail;
1125 	}
1126 
1127 	while ((--rulep) >= &rules[0]) {
1128 		ipsec_enter_policy(iph, rulep->pol, rulep->dir,
1129 		    spds->spds_netstack);
1130 	}
1131 	rw_exit(&iph->iph_lock);
1132 	if (itp != NULL)
1133 		mutex_exit(&itp->itp_lock);
1134 
1135 	ipsec_actvec_free(actp, nact);
1136 	spd_echo(q, mp);
1137 	if (audit_active) {
1138 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1139 		cred_t *cr;
1140 		pid_t cpid;
1141 
1142 		cr = msg_getcred(mp, &cpid);
1143 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1144 		audit_pf_policy(SPD_ADDRULE, cr, spds->spds_netstack,
1145 		    ITP_NAME(itp), active, 0, cpid);
1146 	}
1147 	return;
1148 
1149 fail:
1150 	rw_exit(&iph->iph_lock);
1151 	while ((--rulep) >= &rules[0]) {
1152 		IPPOL_REFRELE(rulep->pol, spds->spds_netstack);
1153 	}
1154 	ipsec_actvec_free(actp, nact);
1155 fail2:
1156 	if (itp != NULL) {
1157 		if (empty_itp)
1158 			itp->itp_flags = 0;
1159 		mutex_exit(&itp->itp_lock);
1160 	}
1161 	spdsock_error(q, mp, error, diag);
1162 	if (audit_active) {
1163 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1164 		cred_t *cr;
1165 		pid_t cpid;
1166 
1167 		cr = msg_getcred(mp, &cpid);
1168 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1169 		audit_pf_policy(SPD_ADDRULE, cr, spds->spds_netstack,
1170 		    ITP_NAME(itp), active, error, cpid);
1171 	}
1172 }
1173 
1174 void
1175 spdsock_deleterule(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp,
1176     spd_ext_t **extv, ipsec_tun_pol_t *itp)
1177 {
1178 	ipsec_selkey_t sel;
1179 	struct spd_rule *rule = (struct spd_rule *)extv[SPD_EXT_RULE];
1180 	int err, diag = 0;
1181 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
1182 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
1183 
1184 	if (rule == NULL) {
1185 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_NO_RULE_EXT);
1186 		if (audit_active) {
1187 			boolean_t active;
1188 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1189 			cred_t *cr;
1190 			pid_t cpid;
1191 
1192 			cr = msg_getcred(mp, &cpid);
1193 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1194 			audit_pf_policy(SPD_DELETERULE, cr, ns,
1195 			    ITP_NAME(itp), active, SPD_DIAGNOSTIC_NO_RULE_EXT,
1196 			    cpid);
1197 		}
1198 		return;
1199 	}
1200 
1201 	/*
1202 	 * Must enter itp_lock first to avoid deadlock.  See tun.c's
1203 	 * set_sec_simple() for the other case of itp_lock and iph_lock.
1204 	 */
1205 	if (itp != NULL)
1206 		mutex_enter(&itp->itp_lock);
1207 
1208 	if (rule->spd_rule_index != 0) {
1209 		if (ipsec_policy_delete_index(iph, rule->spd_rule_index, ns) !=
1210 		    0) {
1211 			err = ESRCH;
1212 			goto fail;
1213 		}
1214 	} else {
1215 		if (!spdsock_ext_to_sel(extv, &sel, &diag)) {
1216 			err = EINVAL;	/* diag already set... */
1217 			goto fail;
1218 		}
1219 
1220 		if ((rule->spd_rule_flags & SPD_RULE_FLAG_INBOUND) &&
1221 		    !ipsec_policy_delete(iph, &sel, IPSEC_TYPE_INBOUND, ns)) {
1222 			err = ESRCH;
1223 			goto fail;
1224 		}
1225 
1226 		if ((rule->spd_rule_flags & SPD_RULE_FLAG_OUTBOUND) &&
1227 		    !ipsec_policy_delete(iph, &sel, IPSEC_TYPE_OUTBOUND, ns)) {
1228 			err = ESRCH;
1229 			goto fail;
1230 		}
1231 	}
1232 
1233 	if (itp != NULL) {
1234 		ASSERT(iph == itp->itp_policy || iph == itp->itp_inactive);
1235 		rw_enter(&iph->iph_lock, RW_READER);
1236 		if (avl_numnodes(&iph->iph_rulebyid) == 0) {
1237 			if (iph == itp->itp_policy)
1238 				itp->itp_flags &= ~ITPF_PFLAGS;
1239 			else
1240 				itp->itp_flags &= ~ITPF_IFLAGS;
1241 		}
1242 		/* Can exit locks in any order. */
1243 		rw_exit(&iph->iph_lock);
1244 		mutex_exit(&itp->itp_lock);
1245 	}
1246 	spd_echo(q, mp);
1247 	if (audit_active) {
1248 		boolean_t active;
1249 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1250 		cred_t *cr;
1251 		pid_t cpid;
1252 
1253 		cr = msg_getcred(mp, &cpid);
1254 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1255 		audit_pf_policy(SPD_DELETERULE, cr, ns, ITP_NAME(itp),
1256 		    active, 0, cpid);
1257 	}
1258 	return;
1259 fail:
1260 	if (itp != NULL)
1261 		mutex_exit(&itp->itp_lock);
1262 	spdsock_error(q, mp, err, diag);
1263 	if (audit_active) {
1264 		boolean_t active;
1265 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1266 		cred_t *cr;
1267 		pid_t cpid;
1268 
1269 		cr = msg_getcred(mp, &cpid);
1270 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1271 		audit_pf_policy(SPD_DELETERULE, cr, ns, ITP_NAME(itp),
1272 		    active, err, cpid);
1273 	}
1274 }
1275 
1276 /* Do NOT consume a reference to itp. */
1277 /* ARGSUSED */
1278 static void
1279 spdsock_flip_node(ipsec_tun_pol_t *itp, void *ignoreme, netstack_t *ns)
1280 {
1281 	mutex_enter(&itp->itp_lock);
1282 	ITPF_SWAP(itp->itp_flags);
1283 	ipsec_swap_policy(itp->itp_policy, itp->itp_inactive, ns);
1284 	mutex_exit(&itp->itp_lock);
1285 }
1286 
1287 void
1288 spdsock_flip(queue_t *q, mblk_t *mp, spd_if_t *tunname)
1289 {
1290 	char *tname;
1291 	ipsec_tun_pol_t *itp;
1292 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
1293 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
1294 
1295 	if (tunname != NULL) {
1296 		tname = (char *)tunname->spd_if_name;
1297 		if (*tname == '\0') {
1298 			/* can't fail */
1299 			ipsec_swap_global_policy(ns);
1300 			if (audit_active) {
1301 				boolean_t active;
1302 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1303 				cred_t *cr;
1304 				pid_t cpid;
1305 
1306 				cr = msg_getcred(mp, &cpid);
1307 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1308 				audit_pf_policy(SPD_FLIP, cr, ns,
1309 				    NULL, active, 0, cpid);
1310 			}
1311 			itp_walk(spdsock_flip_node, NULL, ns);
1312 			if (audit_active) {
1313 				boolean_t active;
1314 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1315 				cred_t *cr;
1316 				pid_t cpid;
1317 
1318 				cr = msg_getcred(mp, &cpid);
1319 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1320 				audit_pf_policy(SPD_FLIP, cr, ns,
1321 				    "all tunnels", active, 0, cpid);
1322 			}
1323 		} else {
1324 			itp = get_tunnel_policy(tname, ns);
1325 			if (itp == NULL) {
1326 				/* Better idea for "tunnel not found"? */
1327 				spdsock_error(q, mp, ESRCH, 0);
1328 				if (audit_active) {
1329 					boolean_t active;
1330 					spd_msg_t *spmsg =
1331 					    (spd_msg_t *)mp->b_rptr;
1332 					cred_t *cr;
1333 					pid_t cpid;
1334 
1335 					cr = msg_getcred(mp, &cpid);
1336 					active = (spmsg->spd_msg_spdid ==
1337 					    SPD_ACTIVE);
1338 					audit_pf_policy(SPD_FLIP, cr, ns,
1339 					    ITP_NAME(itp), active,
1340 					    ESRCH, cpid);
1341 				}
1342 				return;
1343 			}
1344 			spdsock_flip_node(itp, NULL, NULL);
1345 			if (audit_active) {
1346 				boolean_t active;
1347 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1348 				cred_t *cr;
1349 				pid_t cpid;
1350 
1351 				cr = msg_getcred(mp, &cpid);
1352 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1353 				audit_pf_policy(SPD_FLIP, cr, ns,
1354 				    ITP_NAME(itp), active, 0, cpid);
1355 			}
1356 			ITP_REFRELE(itp, ns);
1357 		}
1358 	} else {
1359 		ipsec_swap_global_policy(ns);	/* can't fail */
1360 		if (audit_active) {
1361 			boolean_t active;
1362 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1363 			cred_t *cr;
1364 			pid_t cpid;
1365 
1366 			cr = msg_getcred(mp, &cpid);
1367 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1368 			audit_pf_policy(SPD_FLIP, cr,
1369 			    ns, NULL, active, 0, cpid);
1370 		}
1371 	}
1372 	spd_echo(q, mp);
1373 }
1374 
1375 /*
1376  * Unimplemented feature
1377  */
1378 /* ARGSUSED */
1379 static void
1380 spdsock_lookup(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp,
1381     spd_ext_t **extv, ipsec_tun_pol_t *itp)
1382 {
1383 	spdsock_error(q, mp, EINVAL, 0);
1384 }
1385 
1386 
1387 static mblk_t *
1388 spdsock_dump_ruleset(mblk_t *req, ipsec_policy_head_t *iph,
1389     uint32_t count, uint16_t error)
1390 {
1391 	size_t len = sizeof (spd_ruleset_ext_t) + sizeof (spd_msg_t);
1392 	spd_msg_t *msg;
1393 	spd_ruleset_ext_t *ruleset;
1394 	mblk_t *m = allocb(len, BPRI_HI);
1395 
1396 	ASSERT(RW_READ_HELD(&iph->iph_lock));
1397 
1398 	if (m == NULL) {
1399 		return (NULL);
1400 	}
1401 	msg = (spd_msg_t *)m->b_rptr;
1402 	ruleset = (spd_ruleset_ext_t *)(&msg[1]);
1403 
1404 	m->b_wptr = (uint8_t *)&ruleset[1];
1405 
1406 	*msg = *(spd_msg_t *)(req->b_rptr);
1407 	msg->spd_msg_len = SPD_8TO64(len);
1408 	msg->spd_msg_errno = error;
1409 
1410 	ruleset->spd_ruleset_len = SPD_8TO64(sizeof (*ruleset));
1411 	ruleset->spd_ruleset_type = SPD_EXT_RULESET;
1412 	ruleset->spd_ruleset_count = count;
1413 	ruleset->spd_ruleset_version = iph->iph_gen;
1414 	return (m);
1415 }
1416 
1417 static mblk_t *
1418 spdsock_dump_finish(spdsock_t *ss, int error)
1419 {
1420 	mblk_t *m;
1421 	ipsec_policy_head_t *iph = ss->spdsock_dump_head;
1422 	mblk_t *req = ss->spdsock_dump_req;
1423 
1424 	rw_enter(&iph->iph_lock, RW_READER);
1425 	m = spdsock_dump_ruleset(req, iph, ss->spdsock_dump_count, error);
1426 	rw_exit(&iph->iph_lock);
1427 	IPPH_REFRELE(iph, ss->spdsock_spds->spds_netstack);
1428 	ss->spdsock_dump_req = NULL;
1429 	freemsg(req);
1430 
1431 	return (m);
1432 }
1433 
1434 /*
1435  * Rule encoding functions.
1436  * We do a two-pass encode.
1437  * If base != NULL, fill in encoded rule part starting at base+offset.
1438  * Always return "offset" plus length of to-be-encoded data.
1439  */
1440 static uint_t
1441 spdsock_encode_typecode(uint8_t *base, uint_t offset, uint8_t type,
1442     uint8_t type_end, uint8_t code, uint8_t code_end)
1443 {
1444 	struct spd_typecode *tcp;
1445 
1446 	ASSERT(ALIGNED64(offset));
1447 
1448 	if (base != NULL) {
1449 		tcp = (struct spd_typecode *)(base + offset);
1450 		tcp->spd_typecode_len = SPD_8TO64(sizeof (*tcp));
1451 		tcp->spd_typecode_exttype = SPD_EXT_ICMP_TYPECODE;
1452 		tcp->spd_typecode_code = code;
1453 		tcp->spd_typecode_type = type;
1454 		tcp->spd_typecode_type_end = type_end;
1455 		tcp->spd_typecode_code_end = code_end;
1456 	}
1457 	offset += sizeof (*tcp);
1458 
1459 	ASSERT(ALIGNED64(offset));
1460 
1461 	return (offset);
1462 }
1463 
1464 static uint_t
1465 spdsock_encode_proto(uint8_t *base, uint_t offset, uint8_t proto)
1466 {
1467 	struct spd_proto *spp;
1468 
1469 	ASSERT(ALIGNED64(offset));
1470 
1471 	if (base != NULL) {
1472 		spp = (struct spd_proto *)(base + offset);
1473 		spp->spd_proto_len = SPD_8TO64(sizeof (*spp));
1474 		spp->spd_proto_exttype = SPD_EXT_PROTO;
1475 		spp->spd_proto_number = proto;
1476 		spp->spd_proto_reserved1 = 0;
1477 		spp->spd_proto_reserved2 = 0;
1478 	}
1479 	offset += sizeof (*spp);
1480 
1481 	ASSERT(ALIGNED64(offset));
1482 
1483 	return (offset);
1484 }
1485 
1486 static uint_t
1487 spdsock_encode_port(uint8_t *base, uint_t offset, uint16_t ext, uint16_t port)
1488 {
1489 	struct spd_portrange *spp;
1490 
1491 	ASSERT(ALIGNED64(offset));
1492 
1493 	if (base != NULL) {
1494 		spp = (struct spd_portrange *)(base + offset);
1495 		spp->spd_ports_len = SPD_8TO64(sizeof (*spp));
1496 		spp->spd_ports_exttype = ext;
1497 		spp->spd_ports_minport = port;
1498 		spp->spd_ports_maxport = port;
1499 	}
1500 	offset += sizeof (*spp);
1501 
1502 	ASSERT(ALIGNED64(offset));
1503 
1504 	return (offset);
1505 }
1506 
1507 static uint_t
1508 spdsock_encode_addr(uint8_t *base, uint_t offset, uint16_t ext,
1509     const ipsec_selkey_t *sel, const ipsec_addr_t *addr, uint_t pfxlen)
1510 {
1511 	struct spd_address *sae;
1512 	ipsec_addr_t *spdaddr;
1513 	uint_t start = offset;
1514 	uint_t addrlen;
1515 	uint_t af;
1516 
1517 	if (sel->ipsl_valid & IPSL_IPV4) {
1518 		af = AF_INET;
1519 		addrlen = IP_ADDR_LEN;
1520 	} else {
1521 		af = AF_INET6;
1522 		addrlen = IPV6_ADDR_LEN;
1523 	}
1524 
1525 	ASSERT(ALIGNED64(offset));
1526 
1527 	if (base != NULL) {
1528 		sae = (struct spd_address *)(base + offset);
1529 		sae->spd_address_exttype = ext;
1530 		sae->spd_address_af = af;
1531 		sae->spd_address_prefixlen = pfxlen;
1532 		sae->spd_address_reserved2 = 0;
1533 
1534 		spdaddr = (ipsec_addr_t *)(&sae[1]);
1535 		bcopy(addr, spdaddr, addrlen);
1536 	}
1537 	offset += sizeof (*sae);
1538 	addrlen = roundup(addrlen, sizeof (uint64_t));
1539 	offset += addrlen;
1540 
1541 	ASSERT(ALIGNED64(offset));
1542 
1543 	if (base != NULL)
1544 		sae->spd_address_len = SPD_8TO64(offset - start);
1545 	return (offset);
1546 }
1547 
1548 static uint_t
1549 spdsock_encode_sel(uint8_t *base, uint_t offset, const ipsec_sel_t *sel)
1550 {
1551 	const ipsec_selkey_t *selkey = &sel->ipsl_key;
1552 
1553 	if (selkey->ipsl_valid & IPSL_PROTOCOL)
1554 		offset = spdsock_encode_proto(base, offset, selkey->ipsl_proto);
1555 	if (selkey->ipsl_valid & IPSL_LOCAL_PORT)
1556 		offset = spdsock_encode_port(base, offset, SPD_EXT_LCLPORT,
1557 		    selkey->ipsl_lport);
1558 	if (selkey->ipsl_valid & IPSL_REMOTE_PORT)
1559 		offset = spdsock_encode_port(base, offset, SPD_EXT_REMPORT,
1560 		    selkey->ipsl_rport);
1561 	if (selkey->ipsl_valid & IPSL_REMOTE_ADDR)
1562 		offset = spdsock_encode_addr(base, offset, SPD_EXT_REMADDR,
1563 		    selkey, &selkey->ipsl_remote, selkey->ipsl_remote_pfxlen);
1564 	if (selkey->ipsl_valid & IPSL_LOCAL_ADDR)
1565 		offset = spdsock_encode_addr(base, offset, SPD_EXT_LCLADDR,
1566 		    selkey, &selkey->ipsl_local, selkey->ipsl_local_pfxlen);
1567 	if (selkey->ipsl_valid & IPSL_ICMP_TYPE) {
1568 		offset = spdsock_encode_typecode(base, offset,
1569 		    selkey->ipsl_icmp_type, selkey->ipsl_icmp_type_end,
1570 		    (selkey->ipsl_valid & IPSL_ICMP_CODE) ?
1571 		    selkey->ipsl_icmp_code : 255,
1572 		    (selkey->ipsl_valid & IPSL_ICMP_CODE) ?
1573 		    selkey->ipsl_icmp_code_end : 255);
1574 	}
1575 	return (offset);
1576 }
1577 
1578 static uint_t
1579 spdsock_encode_actattr(uint8_t *base, uint_t offset, uint32_t tag,
1580     uint32_t value)
1581 {
1582 	struct spd_attribute *attr;
1583 
1584 	ASSERT(ALIGNED64(offset));
1585 
1586 	if (base != NULL) {
1587 		attr = (struct spd_attribute *)(base + offset);
1588 		attr->spd_attr_tag = tag;
1589 		attr->spd_attr_value = value;
1590 	}
1591 	offset += sizeof (struct spd_attribute);
1592 
1593 	ASSERT(ALIGNED64(offset));
1594 
1595 	return (offset);
1596 }
1597 
1598 
1599 #define	EMIT(t, v) offset = spdsock_encode_actattr(base, offset, (t), (v))
1600 
1601 static uint_t
1602 spdsock_encode_action(uint8_t *base, uint_t offset, const ipsec_action_t *ap)
1603 {
1604 	const struct ipsec_act *act = &(ap->ipa_act);
1605 	uint_t flags;
1606 
1607 	EMIT(SPD_ATTR_EMPTY, 0);
1608 	switch (act->ipa_type) {
1609 	case IPSEC_ACT_DISCARD:
1610 	case IPSEC_ACT_REJECT:
1611 		EMIT(SPD_ATTR_TYPE, SPD_ACTTYPE_DROP);
1612 		break;
1613 	case IPSEC_ACT_BYPASS:
1614 	case IPSEC_ACT_CLEAR:
1615 		EMIT(SPD_ATTR_TYPE, SPD_ACTTYPE_PASS);
1616 		break;
1617 
1618 	case IPSEC_ACT_APPLY:
1619 		EMIT(SPD_ATTR_TYPE, SPD_ACTTYPE_IPSEC);
1620 		flags = 0;
1621 		if (act->ipa_apply.ipp_use_ah)
1622 			flags |= SPD_APPLY_AH;
1623 		if (act->ipa_apply.ipp_use_esp)
1624 			flags |= SPD_APPLY_ESP;
1625 		if (act->ipa_apply.ipp_use_espa)
1626 			flags |= SPD_APPLY_ESPA;
1627 		if (act->ipa_apply.ipp_use_se)
1628 			flags |= SPD_APPLY_SE;
1629 		if (act->ipa_apply.ipp_use_unique)
1630 			flags |= SPD_APPLY_UNIQUE;
1631 		EMIT(SPD_ATTR_FLAGS, flags);
1632 		if (flags & SPD_APPLY_AH) {
1633 			EMIT(SPD_ATTR_AH_AUTH, act->ipa_apply.ipp_auth_alg);
1634 			EMIT(SPD_ATTR_AH_MINBITS,
1635 			    act->ipa_apply.ipp_ah_minbits);
1636 			EMIT(SPD_ATTR_AH_MAXBITS,
1637 			    act->ipa_apply.ipp_ah_maxbits);
1638 		}
1639 		if (flags & SPD_APPLY_ESP) {
1640 			EMIT(SPD_ATTR_ESP_ENCR, act->ipa_apply.ipp_encr_alg);
1641 			EMIT(SPD_ATTR_ENCR_MINBITS,
1642 			    act->ipa_apply.ipp_espe_minbits);
1643 			EMIT(SPD_ATTR_ENCR_MAXBITS,
1644 			    act->ipa_apply.ipp_espe_maxbits);
1645 			if (flags & SPD_APPLY_ESPA) {
1646 				EMIT(SPD_ATTR_ESP_AUTH,
1647 				    act->ipa_apply.ipp_esp_auth_alg);
1648 				EMIT(SPD_ATTR_ESPA_MINBITS,
1649 				    act->ipa_apply.ipp_espa_minbits);
1650 				EMIT(SPD_ATTR_ESPA_MAXBITS,
1651 				    act->ipa_apply.ipp_espa_maxbits);
1652 			}
1653 		}
1654 		if (act->ipa_apply.ipp_km_proto != 0)
1655 			EMIT(SPD_ATTR_KM_PROTO, act->ipa_apply.ipp_km_proto);
1656 		if (act->ipa_apply.ipp_km_cookie != 0)
1657 			EMIT(SPD_ATTR_KM_PROTO, act->ipa_apply.ipp_km_cookie);
1658 		if (act->ipa_apply.ipp_replay_depth != 0)
1659 			EMIT(SPD_ATTR_REPLAY_DEPTH,
1660 			    act->ipa_apply.ipp_replay_depth);
1661 		/* Add more here */
1662 		break;
1663 	}
1664 
1665 	return (offset);
1666 }
1667 
1668 static uint_t
1669 spdsock_encode_action_list(uint8_t *base, uint_t offset,
1670     const ipsec_action_t *ap)
1671 {
1672 	struct spd_ext_actions *act;
1673 	uint_t nact = 0;
1674 	uint_t start = offset;
1675 
1676 	ASSERT(ALIGNED64(offset));
1677 
1678 	if (base != NULL) {
1679 		act = (struct spd_ext_actions *)(base + offset);
1680 		act->spd_actions_len = 0;
1681 		act->spd_actions_exttype = SPD_EXT_ACTION;
1682 		act->spd_actions_count = 0;
1683 		act->spd_actions_reserved = 0;
1684 	}
1685 
1686 	offset += sizeof (*act);
1687 
1688 	ASSERT(ALIGNED64(offset));
1689 
1690 	while (ap != NULL) {
1691 		offset = spdsock_encode_action(base, offset, ap);
1692 		ap = ap->ipa_next;
1693 		nact++;
1694 		if (ap != NULL) {
1695 			EMIT(SPD_ATTR_NEXT, 0);
1696 		}
1697 	}
1698 	EMIT(SPD_ATTR_END, 0);
1699 
1700 	ASSERT(ALIGNED64(offset));
1701 
1702 	if (base != NULL) {
1703 		act->spd_actions_count = nact;
1704 		act->spd_actions_len = SPD_8TO64(offset - start);
1705 	}
1706 
1707 	return (offset);
1708 }
1709 
1710 #undef EMIT
1711 
1712 /* ARGSUSED */
1713 static uint_t
1714 spdsock_rule_flags(uint_t dir, uint_t af)
1715 {
1716 	uint_t flags = 0;
1717 
1718 	if (dir == IPSEC_TYPE_INBOUND)
1719 		flags |= SPD_RULE_FLAG_INBOUND;
1720 	if (dir == IPSEC_TYPE_OUTBOUND)
1721 		flags |= SPD_RULE_FLAG_OUTBOUND;
1722 
1723 	return (flags);
1724 }
1725 
1726 
1727 static uint_t
1728 spdsock_encode_rule_head(uint8_t *base, uint_t offset, spd_msg_t *req,
1729     const ipsec_policy_t *rule, uint_t dir, uint_t af, char *name,
1730     boolean_t tunnel)
1731 {
1732 	struct spd_msg *spmsg;
1733 	struct spd_rule *spr;
1734 	spd_if_t *sid;
1735 
1736 	uint_t start = offset;
1737 
1738 	ASSERT(ALIGNED64(offset));
1739 
1740 	if (base != NULL) {
1741 		spmsg = (struct spd_msg *)(base + offset);
1742 		bzero(spmsg, sizeof (*spmsg));
1743 		spmsg->spd_msg_version = PF_POLICY_V1;
1744 		spmsg->spd_msg_type = SPD_DUMP;
1745 		spmsg->spd_msg_seq = req->spd_msg_seq;
1746 		spmsg->spd_msg_pid = req->spd_msg_pid;
1747 	}
1748 	offset += sizeof (struct spd_msg);
1749 
1750 	ASSERT(ALIGNED64(offset));
1751 
1752 	if (base != NULL) {
1753 		spr = (struct spd_rule *)(base + offset);
1754 		spr->spd_rule_type = SPD_EXT_RULE;
1755 		spr->spd_rule_priority = rule->ipsp_prio;
1756 		spr->spd_rule_flags = spdsock_rule_flags(dir, af);
1757 		if (tunnel)
1758 			spr->spd_rule_flags |= SPD_RULE_FLAG_TUNNEL;
1759 		spr->spd_rule_unused = 0;
1760 		spr->spd_rule_len = SPD_8TO64(sizeof (*spr));
1761 		spr->spd_rule_index = rule->ipsp_index;
1762 	}
1763 	offset += sizeof (struct spd_rule);
1764 
1765 	/*
1766 	 * If we have an interface name (i.e. if this policy head came from
1767 	 * a tunnel), add the SPD_EXT_TUN_NAME extension.
1768 	 */
1769 	if (name != NULL) {
1770 
1771 		ASSERT(ALIGNED64(offset));
1772 
1773 		if (base != NULL) {
1774 			sid = (spd_if_t *)(base + offset);
1775 			sid->spd_if_exttype = SPD_EXT_TUN_NAME;
1776 			sid->spd_if_len = SPD_8TO64(sizeof (spd_if_t) +
1777 			    roundup((strlen(name) - 4), 8));
1778 			(void) strlcpy((char *)sid->spd_if_name, name,
1779 			    LIFNAMSIZ);
1780 		}
1781 
1782 		offset += sizeof (spd_if_t) + roundup((strlen(name) - 4), 8);
1783 	}
1784 
1785 	offset = spdsock_encode_sel(base, offset, rule->ipsp_sel);
1786 	offset = spdsock_encode_action_list(base, offset, rule->ipsp_act);
1787 
1788 	ASSERT(ALIGNED64(offset));
1789 
1790 	if (base != NULL) {
1791 		spmsg->spd_msg_len = SPD_8TO64(offset - start);
1792 	}
1793 	return (offset);
1794 }
1795 
1796 /* ARGSUSED */
1797 static mblk_t *
1798 spdsock_encode_rule(mblk_t *req, const ipsec_policy_t *rule,
1799     uint_t dir, uint_t af, char *name, boolean_t tunnel)
1800 {
1801 	mblk_t *m;
1802 	uint_t len;
1803 	spd_msg_t *mreq = (spd_msg_t *)req->b_rptr;
1804 
1805 	/*
1806 	 * Figure out how much space we'll need.
1807 	 */
1808 	len = spdsock_encode_rule_head(NULL, 0, mreq, rule, dir, af, name,
1809 	    tunnel);
1810 
1811 	/*
1812 	 * Allocate mblk.
1813 	 */
1814 	m = allocb(len, BPRI_HI);
1815 	if (m == NULL)
1816 		return (NULL);
1817 
1818 	/*
1819 	 * Fill it in..
1820 	 */
1821 	m->b_wptr = m->b_rptr + len;
1822 	bzero(m->b_rptr, len);
1823 	(void) spdsock_encode_rule_head(m->b_rptr, 0, mreq, rule, dir, af,
1824 	    name, tunnel);
1825 	return (m);
1826 }
1827 
1828 static ipsec_policy_t *
1829 spdsock_dump_next_in_chain(spdsock_t *ss, ipsec_policy_head_t *iph,
1830     ipsec_policy_t *cur)
1831 {
1832 	ASSERT(RW_READ_HELD(&iph->iph_lock));
1833 
1834 	ss->spdsock_dump_count++;
1835 	ss->spdsock_dump_cur_rule = cur->ipsp_hash.hash_next;
1836 	return (cur);
1837 }
1838 
1839 static ipsec_policy_t *
1840 spdsock_dump_next_rule(spdsock_t *ss, ipsec_policy_head_t *iph)
1841 {
1842 	ipsec_policy_t *cur;
1843 	ipsec_policy_root_t *ipr;
1844 	int chain, nchains, type, af;
1845 
1846 	ASSERT(RW_READ_HELD(&iph->iph_lock));
1847 
1848 	cur = ss->spdsock_dump_cur_rule;
1849 
1850 	if (cur != NULL)
1851 		return (spdsock_dump_next_in_chain(ss, iph, cur));
1852 
1853 	type = ss->spdsock_dump_cur_type;
1854 
1855 next:
1856 	chain = ss->spdsock_dump_cur_chain;
1857 	ipr = &iph->iph_root[type];
1858 	nchains = ipr->ipr_nchains;
1859 
1860 	while (chain < nchains) {
1861 		cur = ipr->ipr_hash[chain].hash_head;
1862 		chain++;
1863 		if (cur != NULL) {
1864 			ss->spdsock_dump_cur_chain = chain;
1865 			return (spdsock_dump_next_in_chain(ss, iph, cur));
1866 		}
1867 	}
1868 	ss->spdsock_dump_cur_chain = nchains;
1869 
1870 	af = ss->spdsock_dump_cur_af;
1871 	while (af < IPSEC_NAF) {
1872 		cur = ipr->ipr_nonhash[af];
1873 		af++;
1874 		if (cur != NULL) {
1875 			ss->spdsock_dump_cur_af = af;
1876 			return (spdsock_dump_next_in_chain(ss, iph, cur));
1877 		}
1878 	}
1879 
1880 	type++;
1881 	if (type >= IPSEC_NTYPES)
1882 		return (NULL);
1883 
1884 	ss->spdsock_dump_cur_chain = 0;
1885 	ss->spdsock_dump_cur_type = type;
1886 	ss->spdsock_dump_cur_af = IPSEC_AF_V4;
1887 	goto next;
1888 
1889 }
1890 
1891 /*
1892  * If we're done with one policy head, but have more to go, we iterate through
1893  * another IPsec tunnel policy head (itp).  Return NULL if it is an error
1894  * worthy of returning EAGAIN via PF_POLICY.
1895  */
1896 static ipsec_tun_pol_t *
1897 spdsock_dump_iterate_next_tunnel(spdsock_t *ss, ipsec_stack_t *ipss)
1898 {
1899 	ipsec_tun_pol_t *itp;
1900 
1901 	ASSERT(RW_READ_HELD(&ipss->ipsec_tunnel_policy_lock));
1902 	if (ipss->ipsec_tunnel_policy_gen > ss->spdsock_dump_tun_gen) {
1903 		/* Oops, state of the tunnel polheads changed. */
1904 		itp = NULL;
1905 	} else if (ss->spdsock_itp == NULL) {
1906 		/* Just finished global, find first node. */
1907 		itp = avl_first(&ipss->ipsec_tunnel_policies);
1908 	} else {
1909 		/* We just finished current polhead, find the next one. */
1910 		itp = AVL_NEXT(&ipss->ipsec_tunnel_policies, ss->spdsock_itp);
1911 	}
1912 	if (itp != NULL) {
1913 		ITP_REFHOLD(itp);
1914 	}
1915 	if (ss->spdsock_itp != NULL) {
1916 		ITP_REFRELE(ss->spdsock_itp, ipss->ipsec_netstack);
1917 	}
1918 	ss->spdsock_itp = itp;
1919 	return (itp);
1920 }
1921 
1922 static mblk_t *
1923 spdsock_dump_next_record(spdsock_t *ss)
1924 {
1925 	ipsec_policy_head_t *iph;
1926 	ipsec_policy_t *rule;
1927 	mblk_t *m;
1928 	ipsec_tun_pol_t *itp;
1929 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
1930 	ipsec_stack_t *ipss = ns->netstack_ipsec;
1931 
1932 	iph = ss->spdsock_dump_head;
1933 
1934 	ASSERT(iph != NULL);
1935 
1936 	rw_enter(&iph->iph_lock, RW_READER);
1937 
1938 	if (iph->iph_gen != ss->spdsock_dump_gen) {
1939 		rw_exit(&iph->iph_lock);
1940 		return (spdsock_dump_finish(ss, EAGAIN));
1941 	}
1942 
1943 	while ((rule = spdsock_dump_next_rule(ss, iph)) == NULL) {
1944 		rw_exit(&iph->iph_lock);
1945 		if (--(ss->spdsock_dump_remaining_polheads) == 0)
1946 			return (spdsock_dump_finish(ss, 0));
1947 
1948 
1949 		/*
1950 		 * If we reach here, we have more policy heads (tunnel
1951 		 * entries) to dump.  Let's reset to a new policy head
1952 		 * and get some more rules.
1953 		 *
1954 		 * An empty policy head will have spdsock_dump_next_rule()
1955 		 * return NULL, and we loop (while dropping the number of
1956 		 * remaining polheads).  If we loop to 0, we finish.  We
1957 		 * keep looping until we hit 0 or until we have a rule to
1958 		 * encode.
1959 		 *
1960 		 * NOTE:  No need for ITP_REF*() macros here as we're only
1961 		 * going after and refholding the policy head itself.
1962 		 */
1963 		rw_enter(&ipss->ipsec_tunnel_policy_lock, RW_READER);
1964 		itp = spdsock_dump_iterate_next_tunnel(ss, ipss);
1965 		if (itp == NULL) {
1966 			rw_exit(&ipss->ipsec_tunnel_policy_lock);
1967 			return (spdsock_dump_finish(ss, EAGAIN));
1968 		}
1969 
1970 		/* Reset other spdsock_dump thingies. */
1971 		IPPH_REFRELE(ss->spdsock_dump_head, ns);
1972 		if (ss->spdsock_dump_active) {
1973 			ss->spdsock_dump_tunnel =
1974 			    itp->itp_flags & ITPF_P_TUNNEL;
1975 			iph = itp->itp_policy;
1976 		} else {
1977 			ss->spdsock_dump_tunnel =
1978 			    itp->itp_flags & ITPF_I_TUNNEL;
1979 			iph = itp->itp_inactive;
1980 		}
1981 		IPPH_REFHOLD(iph);
1982 		rw_exit(&ipss->ipsec_tunnel_policy_lock);
1983 
1984 		rw_enter(&iph->iph_lock, RW_READER);
1985 		RESET_SPDSOCK_DUMP_POLHEAD(ss, iph);
1986 	}
1987 
1988 	m = spdsock_encode_rule(ss->spdsock_dump_req, rule,
1989 	    ss->spdsock_dump_cur_type, ss->spdsock_dump_cur_af,
1990 	    (ss->spdsock_itp == NULL) ? NULL : ss->spdsock_itp->itp_name,
1991 	    ss->spdsock_dump_tunnel);
1992 	rw_exit(&iph->iph_lock);
1993 
1994 	if (m == NULL)
1995 		return (spdsock_dump_finish(ss, ENOMEM));
1996 	return (m);
1997 }
1998 
1999 /*
2000  * Dump records until we run into flow-control back-pressure.
2001  */
2002 static void
2003 spdsock_dump_some(queue_t *q, spdsock_t *ss)
2004 {
2005 	mblk_t *m, *dataind;
2006 
2007 	while ((ss->spdsock_dump_req != NULL) && canputnext(q)) {
2008 		m = spdsock_dump_next_record(ss);
2009 		if (m == NULL)
2010 			return;
2011 		dataind = allocb(sizeof (struct T_data_req), BPRI_HI);
2012 		if (dataind == NULL) {
2013 			freemsg(m);
2014 			return;
2015 		}
2016 		dataind->b_cont = m;
2017 		dataind->b_wptr += sizeof (struct T_data_req);
2018 		((struct T_data_ind *)dataind->b_rptr)->PRIM_type = T_DATA_IND;
2019 		((struct T_data_ind *)dataind->b_rptr)->MORE_flag = 0;
2020 		dataind->b_datap->db_type = M_PROTO;
2021 		putnext(q, dataind);
2022 	}
2023 }
2024 
2025 /*
2026  * Start dumping.
2027  * Format a start-of-dump record, and set up the stream and kick the rsrv
2028  * procedure to continue the job..
2029  */
2030 /* ARGSUSED */
2031 static void
2032 spdsock_dump(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp)
2033 {
2034 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2035 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
2036 	ipsec_stack_t *ipss = ns->netstack_ipsec;
2037 	mblk_t *mr;
2038 
2039 	/* spdsock_open() already set spdsock_itp to NULL. */
2040 	if (iph == ALL_ACTIVE_POLHEADS || iph == ALL_INACTIVE_POLHEADS) {
2041 		rw_enter(&ipss->ipsec_tunnel_policy_lock, RW_READER);
2042 		ss->spdsock_dump_remaining_polheads = 1 +
2043 		    avl_numnodes(&ipss->ipsec_tunnel_policies);
2044 		ss->spdsock_dump_tun_gen = ipss->ipsec_tunnel_policy_gen;
2045 		rw_exit(&ipss->ipsec_tunnel_policy_lock);
2046 		if (iph == ALL_ACTIVE_POLHEADS) {
2047 			iph = ipsec_system_policy(ns);
2048 			ss->spdsock_dump_active = B_TRUE;
2049 		} else {
2050 			iph = ipsec_inactive_policy(ns);
2051 			ss->spdsock_dump_active = B_FALSE;
2052 		}
2053 		ASSERT(ss->spdsock_itp == NULL);
2054 	} else {
2055 		ss->spdsock_dump_remaining_polheads = 1;
2056 	}
2057 
2058 	rw_enter(&iph->iph_lock, RW_READER);
2059 
2060 	mr = spdsock_dump_ruleset(mp, iph, 0, 0);
2061 
2062 	if (!mr) {
2063 		rw_exit(&iph->iph_lock);
2064 		spdsock_error(q, mp, ENOMEM, 0);
2065 		return;
2066 	}
2067 
2068 	ss->spdsock_dump_req = mp;
2069 	RESET_SPDSOCK_DUMP_POLHEAD(ss, iph);
2070 
2071 	rw_exit(&iph->iph_lock);
2072 
2073 	qreply(q, mr);
2074 	qenable(OTHERQ(q));
2075 }
2076 
2077 /* Do NOT consume a reference to ITP. */
2078 void
2079 spdsock_clone_node(ipsec_tun_pol_t *itp, void *ep, netstack_t *ns)
2080 {
2081 	int *errptr = (int *)ep;
2082 
2083 	if (*errptr != 0)
2084 		return;	/* We've failed already for some reason. */
2085 	mutex_enter(&itp->itp_lock);
2086 	ITPF_CLONE(itp->itp_flags);
2087 	*errptr = ipsec_copy_polhead(itp->itp_policy, itp->itp_inactive, ns);
2088 	mutex_exit(&itp->itp_lock);
2089 }
2090 
2091 void
2092 spdsock_clone(queue_t *q, mblk_t *mp, spd_if_t *tunname)
2093 {
2094 	int error;
2095 	char *tname;
2096 	ipsec_tun_pol_t *itp;
2097 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2098 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
2099 
2100 	if (tunname != NULL) {
2101 		tname = (char *)tunname->spd_if_name;
2102 		if (*tname == '\0') {
2103 			error = ipsec_clone_system_policy(ns);
2104 			if (audit_active) {
2105 				boolean_t active;
2106 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
2107 				cred_t *cr;
2108 				pid_t cpid;
2109 
2110 				cr = msg_getcred(mp, &cpid);
2111 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
2112 				audit_pf_policy(SPD_CLONE, cr, ns,
2113 				    NULL, active, error, cpid);
2114 			}
2115 			if (error == 0) {
2116 				itp_walk(spdsock_clone_node, &error, ns);
2117 				if (audit_active) {
2118 					boolean_t active;
2119 					spd_msg_t *spmsg =
2120 					    (spd_msg_t *)mp->b_rptr;
2121 					cred_t *cr;
2122 					pid_t cpid;
2123 
2124 					cr = msg_getcred(mp, &cpid);
2125 					active = (spmsg->spd_msg_spdid ==
2126 					    SPD_ACTIVE);
2127 					audit_pf_policy(SPD_CLONE, cr,
2128 					    ns, "all tunnels", active, 0,
2129 					    cpid);
2130 				}
2131 			}
2132 		} else {
2133 			itp = get_tunnel_policy(tname, ns);
2134 			if (itp == NULL) {
2135 				spdsock_error(q, mp, ENOENT, 0);
2136 				if (audit_active) {
2137 					boolean_t active;
2138 					spd_msg_t *spmsg =
2139 					    (spd_msg_t *)mp->b_rptr;
2140 					cred_t *cr;
2141 					pid_t cpid;
2142 
2143 					cr = msg_getcred(mp, &cpid);
2144 					active = (spmsg->spd_msg_spdid ==
2145 					    SPD_ACTIVE);
2146 					audit_pf_policy(SPD_CLONE, cr,
2147 					    ns, ITP_NAME(itp), active, ENOENT,
2148 					    cpid);
2149 				}
2150 				return;
2151 			}
2152 			spdsock_clone_node(itp, &error, NULL);
2153 			ITP_REFRELE(itp, ns);
2154 			if (audit_active) {
2155 				boolean_t active;
2156 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
2157 				cred_t *cr;
2158 				pid_t cpid;
2159 
2160 				cr = msg_getcred(mp, &cpid);
2161 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
2162 				audit_pf_policy(SPD_CLONE, cr, ns,
2163 				    ITP_NAME(itp), active, error, cpid);
2164 			}
2165 		}
2166 	} else {
2167 		error = ipsec_clone_system_policy(ns);
2168 		if (audit_active) {
2169 			boolean_t active;
2170 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
2171 			cred_t *cr;
2172 			pid_t cpid;
2173 
2174 			cr = msg_getcred(mp, &cpid);
2175 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
2176 			audit_pf_policy(SPD_CLONE, cr, ns, NULL,
2177 			    active, error, cpid);
2178 		}
2179 	}
2180 
2181 	if (error != 0)
2182 		spdsock_error(q, mp, error, 0);
2183 	else
2184 		spd_echo(q, mp);
2185 }
2186 
2187 /*
2188  * Process a SPD_ALGLIST request. The caller expects separate alg entries
2189  * for AH authentication, ESP authentication, and ESP encryption.
2190  * The same distinction is then used when setting the min and max key
2191  * sizes when defining policies.
2192  */
2193 
2194 #define	SPDSOCK_AH_AUTH		0
2195 #define	SPDSOCK_ESP_AUTH	1
2196 #define	SPDSOCK_ESP_ENCR	2
2197 #define	SPDSOCK_NTYPES		3
2198 
2199 static const uint_t algattr[SPDSOCK_NTYPES] = {
2200 	SPD_ATTR_AH_AUTH,
2201 	SPD_ATTR_ESP_AUTH,
2202 	SPD_ATTR_ESP_ENCR
2203 };
2204 static const uint_t minbitsattr[SPDSOCK_NTYPES] = {
2205 	SPD_ATTR_AH_MINBITS,
2206 	SPD_ATTR_ESPA_MINBITS,
2207 	SPD_ATTR_ENCR_MINBITS
2208 };
2209 static const uint_t maxbitsattr[SPDSOCK_NTYPES] = {
2210 	SPD_ATTR_AH_MAXBITS,
2211 	SPD_ATTR_ESPA_MAXBITS,
2212 	SPD_ATTR_ENCR_MAXBITS
2213 };
2214 static const uint_t defbitsattr[SPDSOCK_NTYPES] = {
2215 	SPD_ATTR_AH_DEFBITS,
2216 	SPD_ATTR_ESPA_DEFBITS,
2217 	SPD_ATTR_ENCR_DEFBITS
2218 };
2219 static const uint_t incrbitsattr[SPDSOCK_NTYPES] = {
2220 	SPD_ATTR_AH_INCRBITS,
2221 	SPD_ATTR_ESPA_INCRBITS,
2222 	SPD_ATTR_ENCR_INCRBITS
2223 };
2224 
2225 #define	ATTRPERALG	6	/* fixed attributes per algs */
2226 
2227 void
2228 spdsock_alglist(queue_t *q, mblk_t *mp)
2229 {
2230 	uint_t algtype;
2231 	uint_t algidx;
2232 	uint_t algcount;
2233 	uint_t size;
2234 	mblk_t *m;
2235 	uint8_t *cur;
2236 	spd_msg_t *msg;
2237 	struct spd_ext_actions *act;
2238 	struct spd_attribute *attr;
2239 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2240 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
2241 
2242 	mutex_enter(&ipss->ipsec_alg_lock);
2243 	/*
2244 	 * The SPD client expects to receive separate entries for
2245 	 * AH authentication and ESP authentication supported algorithms.
2246 	 *
2247 	 * Don't return the "any" algorithms, if defined, as no
2248 	 * kernel policies can be set for these algorithms.
2249 	 */
2250 	algcount = 2 * ipss->ipsec_nalgs[IPSEC_ALG_AUTH] +
2251 	    ipss->ipsec_nalgs[IPSEC_ALG_ENCR];
2252 
2253 	if (ipss->ipsec_alglists[IPSEC_ALG_AUTH][SADB_AALG_NONE] != NULL)
2254 		algcount--;
2255 	if (ipss->ipsec_alglists[IPSEC_ALG_ENCR][SADB_EALG_NONE] != NULL)
2256 		algcount--;
2257 
2258 	/*
2259 	 * For each algorithm, we encode:
2260 	 * ALG / MINBITS / MAXBITS / DEFBITS / INCRBITS / {END, NEXT}
2261 	 */
2262 
2263 	size = sizeof (spd_msg_t) + sizeof (struct spd_ext_actions) +
2264 	    ATTRPERALG * sizeof (struct spd_attribute) * algcount;
2265 
2266 	ASSERT(ALIGNED64(size));
2267 
2268 	m = allocb(size, BPRI_HI);
2269 	if (m == NULL) {
2270 		mutex_exit(&ipss->ipsec_alg_lock);
2271 		spdsock_error(q, mp, ENOMEM, 0);
2272 		return;
2273 	}
2274 
2275 	m->b_wptr = m->b_rptr + size;
2276 	cur = m->b_rptr;
2277 
2278 	msg = (spd_msg_t *)cur;
2279 	bcopy(mp->b_rptr, cur, sizeof (*msg));
2280 
2281 	msg->spd_msg_len = SPD_8TO64(size);
2282 	msg->spd_msg_errno = 0;
2283 	msg->spd_msg_diagnostic = 0;
2284 
2285 	cur += sizeof (*msg);
2286 
2287 	act = (struct spd_ext_actions *)cur;
2288 	cur += sizeof (*act);
2289 
2290 	act->spd_actions_len = SPD_8TO64(size - sizeof (spd_msg_t));
2291 	act->spd_actions_exttype = SPD_EXT_ACTION;
2292 	act->spd_actions_count = algcount;
2293 	act->spd_actions_reserved = 0;
2294 
2295 	attr = (struct spd_attribute *)cur;
2296 
2297 #define	EMIT(tag, value) {					\
2298 		attr->spd_attr_tag = (tag); 			\
2299 		attr->spd_attr_value = (value); 		\
2300 		attr++;			  			\
2301 	}
2302 
2303 	/*
2304 	 * If you change the number of EMIT's here, change
2305 	 * ATTRPERALG above to match
2306 	 */
2307 #define	EMITALGATTRS(_type) {					\
2308 		EMIT(algattr[_type], algid); 		/* 1 */	\
2309 		EMIT(minbitsattr[_type], minbits);	/* 2 */	\
2310 		EMIT(maxbitsattr[_type], maxbits);	/* 3 */	\
2311 		EMIT(defbitsattr[_type], defbits);	/* 4 */	\
2312 		EMIT(incrbitsattr[_type], incr);	/* 5 */	\
2313 		EMIT(SPD_ATTR_NEXT, 0);			/* 6 */	\
2314 	}
2315 
2316 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
2317 		for (algidx = 0; algidx < ipss->ipsec_nalgs[algtype];
2318 		    algidx++) {
2319 			int algid = ipss->ipsec_sortlist[algtype][algidx];
2320 			ipsec_alginfo_t *alg =
2321 			    ipss->ipsec_alglists[algtype][algid];
2322 			uint_t minbits = alg->alg_minbits;
2323 			uint_t maxbits = alg->alg_maxbits;
2324 			uint_t defbits = alg->alg_default_bits;
2325 			uint_t incr = alg->alg_increment;
2326 
2327 			if (algtype == IPSEC_ALG_AUTH) {
2328 				if (algid == SADB_AALG_NONE)
2329 					continue;
2330 				EMITALGATTRS(SPDSOCK_AH_AUTH);
2331 				EMITALGATTRS(SPDSOCK_ESP_AUTH);
2332 			} else {
2333 				if (algid == SADB_EALG_NONE)
2334 					continue;
2335 				ASSERT(algtype == IPSEC_ALG_ENCR);
2336 				EMITALGATTRS(SPDSOCK_ESP_ENCR);
2337 			}
2338 		}
2339 	}
2340 
2341 	mutex_exit(&ipss->ipsec_alg_lock);
2342 
2343 #undef EMITALGATTRS
2344 #undef EMIT
2345 #undef ATTRPERALG
2346 
2347 	attr--;
2348 	attr->spd_attr_tag = SPD_ATTR_END;
2349 
2350 	freemsg(mp);
2351 	qreply(q, m);
2352 }
2353 
2354 /*
2355  * Process a SPD_DUMPALGS request.
2356  */
2357 
2358 #define	ATTRPERALG	7	/* fixed attributes per algs */
2359 
2360 void
2361 spdsock_dumpalgs(queue_t *q, mblk_t *mp)
2362 {
2363 	uint_t algtype;
2364 	uint_t algidx;
2365 	uint_t size;
2366 	mblk_t *m;
2367 	uint8_t *cur;
2368 	spd_msg_t *msg;
2369 	struct spd_ext_actions *act;
2370 	struct spd_attribute *attr;
2371 	ipsec_alginfo_t *alg;
2372 	uint_t algid;
2373 	uint_t i;
2374 	uint_t alg_size;
2375 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2376 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
2377 
2378 	mutex_enter(&ipss->ipsec_alg_lock);
2379 
2380 	/*
2381 	 * For each algorithm, we encode:
2382 	 * ALG / MINBITS / MAXBITS / DEFBITS / INCRBITS / {END, NEXT}
2383 	 *
2384 	 * ALG_ID / ALG_PROTO / ALG_INCRBITS / ALG_NKEYSIZES / ALG_KEYSIZE*
2385 	 * ALG_NBLOCKSIZES / ALG_BLOCKSIZE* / ALG_MECHNAME / {END, NEXT}
2386 	 */
2387 
2388 	/*
2389 	 * Compute the size of the SPD message.
2390 	 */
2391 	size = sizeof (spd_msg_t) + sizeof (struct spd_ext_actions);
2392 
2393 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
2394 		for (algidx = 0; algidx < ipss->ipsec_nalgs[algtype];
2395 		    algidx++) {
2396 			algid = ipss->ipsec_sortlist[algtype][algidx];
2397 			alg = ipss->ipsec_alglists[algtype][algid];
2398 			alg_size = sizeof (struct spd_attribute) *
2399 			    (ATTRPERALG + alg->alg_nkey_sizes +
2400 			    alg->alg_nblock_sizes) + CRYPTO_MAX_MECH_NAME;
2401 			size += alg_size;
2402 		}
2403 	}
2404 
2405 	ASSERT(ALIGNED64(size));
2406 
2407 	m = allocb(size, BPRI_HI);
2408 	if (m == NULL) {
2409 		mutex_exit(&ipss->ipsec_alg_lock);
2410 		spdsock_error(q, mp, ENOMEM, 0);
2411 		return;
2412 	}
2413 
2414 	m->b_wptr = m->b_rptr + size;
2415 	cur = m->b_rptr;
2416 
2417 	msg = (spd_msg_t *)cur;
2418 	bcopy(mp->b_rptr, cur, sizeof (*msg));
2419 
2420 	msg->spd_msg_len = SPD_8TO64(size);
2421 	msg->spd_msg_errno = 0;
2422 	msg->spd_msg_type = SPD_ALGLIST;
2423 
2424 	msg->spd_msg_diagnostic = 0;
2425 
2426 	cur += sizeof (*msg);
2427 
2428 	act = (struct spd_ext_actions *)cur;
2429 	cur += sizeof (*act);
2430 
2431 	act->spd_actions_len = SPD_8TO64(size - sizeof (spd_msg_t));
2432 	act->spd_actions_exttype = SPD_EXT_ACTION;
2433 	act->spd_actions_count = ipss->ipsec_nalgs[IPSEC_ALG_AUTH] +
2434 	    ipss->ipsec_nalgs[IPSEC_ALG_ENCR];
2435 	act->spd_actions_reserved = 0;
2436 
2437 	/*
2438 	 * If there aren't any algorithms registered, return an empty message.
2439 	 * spdsock_get_ext() knows how to deal with this.
2440 	 */
2441 	if (act->spd_actions_count == 0) {
2442 		act->spd_actions_len = 0;
2443 		mutex_exit(&ipss->ipsec_alg_lock);
2444 		goto error;
2445 	}
2446 
2447 	attr = (struct spd_attribute *)cur;
2448 
2449 #define	EMIT(tag, value) {					\
2450 		attr->spd_attr_tag = (tag); 			\
2451 		attr->spd_attr_value = (value); 		\
2452 		attr++;			  			\
2453 	}
2454 
2455 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
2456 		for (algidx = 0; algidx < ipss->ipsec_nalgs[algtype];
2457 		    algidx++) {
2458 
2459 			algid = ipss->ipsec_sortlist[algtype][algidx];
2460 			alg = ipss->ipsec_alglists[algtype][algid];
2461 
2462 			/*
2463 			 * If you change the number of EMIT's here, change
2464 			 * ATTRPERALG above to match
2465 			 */
2466 			EMIT(SPD_ATTR_ALG_ID, algid);
2467 			EMIT(SPD_ATTR_ALG_PROTO, algproto[algtype]);
2468 			EMIT(SPD_ATTR_ALG_INCRBITS, alg->alg_increment);
2469 
2470 			EMIT(SPD_ATTR_ALG_NKEYSIZES, alg->alg_nkey_sizes);
2471 			for (i = 0; i < alg->alg_nkey_sizes; i++)
2472 				EMIT(SPD_ATTR_ALG_KEYSIZE,
2473 				    alg->alg_key_sizes[i]);
2474 
2475 			EMIT(SPD_ATTR_ALG_NBLOCKSIZES, alg->alg_nblock_sizes);
2476 			for (i = 0; i < alg->alg_nblock_sizes; i++)
2477 				EMIT(SPD_ATTR_ALG_BLOCKSIZE,
2478 				    alg->alg_block_sizes[i]);
2479 
2480 			EMIT(SPD_ATTR_ALG_MECHNAME, CRYPTO_MAX_MECH_NAME);
2481 			bcopy(alg->alg_mech_name, attr, CRYPTO_MAX_MECH_NAME);
2482 			attr = (struct spd_attribute *)((char *)attr +
2483 			    CRYPTO_MAX_MECH_NAME);
2484 
2485 			EMIT(SPD_ATTR_NEXT, 0);
2486 		}
2487 	}
2488 
2489 	mutex_exit(&ipss->ipsec_alg_lock);
2490 
2491 #undef EMITALGATTRS
2492 #undef EMIT
2493 #undef ATTRPERALG
2494 
2495 	attr--;
2496 	attr->spd_attr_tag = SPD_ATTR_END;
2497 
2498 error:
2499 	freemsg(mp);
2500 	qreply(q, m);
2501 }
2502 
2503 /*
2504  * Do the actual work of processing an SPD_UPDATEALGS request. Can
2505  * be invoked either once IPsec is loaded on a cached request, or
2506  * when a request is received while IPsec is loaded.
2507  */
2508 static void
2509 spdsock_do_updatealg(spd_ext_t *extv[], int *diag, spd_stack_t *spds)
2510 {
2511 	struct spd_ext_actions *actp;
2512 	struct spd_attribute *attr, *endattr;
2513 	uint64_t *start, *end;
2514 	ipsec_alginfo_t *alg = NULL;
2515 	ipsec_algtype_t alg_type = 0;
2516 	boolean_t skip_alg = B_TRUE, doing_proto = B_FALSE;
2517 	uint_t i, cur_key, cur_block, algid;
2518 
2519 	*diag = -1;
2520 	ASSERT(MUTEX_HELD(&spds->spds_alg_lock));
2521 
2522 	/* parse the message, building the list of algorithms */
2523 
2524 	actp = (struct spd_ext_actions *)extv[SPD_EXT_ACTION];
2525 	if (actp == NULL) {
2526 		*diag = SPD_DIAGNOSTIC_NO_ACTION_EXT;
2527 		return;
2528 	}
2529 
2530 	start = (uint64_t *)actp;
2531 	end = (start + actp->spd_actions_len);
2532 	endattr = (struct spd_attribute *)end;
2533 	attr = (struct spd_attribute *)&actp[1];
2534 
2535 	bzero(spds->spds_algs, IPSEC_NALGTYPES * IPSEC_MAX_ALGS *
2536 	    sizeof (ipsec_alginfo_t *));
2537 
2538 	alg = kmem_zalloc(sizeof (*alg), KM_SLEEP);
2539 
2540 #define	ALG_KEY_SIZES(a)   (((a)->alg_nkey_sizes + 1) * sizeof (uint16_t))
2541 #define	ALG_BLOCK_SIZES(a) (((a)->alg_nblock_sizes + 1) * sizeof (uint16_t))
2542 
2543 	while (attr < endattr) {
2544 		switch (attr->spd_attr_tag) {
2545 		case SPD_ATTR_NOP:
2546 		case SPD_ATTR_EMPTY:
2547 			break;
2548 		case SPD_ATTR_END:
2549 			attr = endattr;
2550 			/* FALLTHRU */
2551 		case SPD_ATTR_NEXT:
2552 			if (doing_proto) {
2553 				doing_proto = B_FALSE;
2554 				break;
2555 			}
2556 			if (skip_alg) {
2557 				ipsec_alg_free(alg);
2558 			} else {
2559 				ipsec_alg_free(
2560 				    spds->spds_algs[alg_type][alg->alg_id]);
2561 				spds->spds_algs[alg_type][alg->alg_id] =
2562 				    alg;
2563 			}
2564 			alg = kmem_zalloc(sizeof (*alg), KM_SLEEP);
2565 			break;
2566 
2567 		case SPD_ATTR_ALG_ID:
2568 			if (attr->spd_attr_value >= IPSEC_MAX_ALGS) {
2569 				ss1dbg(spds, ("spdsock_do_updatealg: "
2570 				    "invalid alg id %d\n",
2571 				    attr->spd_attr_value));
2572 				*diag = SPD_DIAGNOSTIC_ALG_ID_RANGE;
2573 				goto bail;
2574 			}
2575 			alg->alg_id = attr->spd_attr_value;
2576 			break;
2577 
2578 		case SPD_ATTR_ALG_PROTO:
2579 			/* find the alg type */
2580 			for (i = 0; i < NALGPROTOS; i++)
2581 				if (algproto[i] == attr->spd_attr_value)
2582 					break;
2583 			skip_alg = (i == NALGPROTOS);
2584 			if (!skip_alg)
2585 				alg_type = i;
2586 			break;
2587 
2588 		case SPD_ATTR_ALG_INCRBITS:
2589 			alg->alg_increment = attr->spd_attr_value;
2590 			break;
2591 
2592 		case SPD_ATTR_ALG_NKEYSIZES:
2593 			if (alg->alg_key_sizes != NULL) {
2594 				kmem_free(alg->alg_key_sizes,
2595 				    ALG_KEY_SIZES(alg));
2596 			}
2597 			alg->alg_nkey_sizes = attr->spd_attr_value;
2598 			/*
2599 			 * Allocate room for the trailing zero key size
2600 			 * value as well.
2601 			 */
2602 			alg->alg_key_sizes = kmem_zalloc(ALG_KEY_SIZES(alg),
2603 			    KM_SLEEP);
2604 			cur_key = 0;
2605 			break;
2606 
2607 		case SPD_ATTR_ALG_KEYSIZE:
2608 			if (alg->alg_key_sizes == NULL ||
2609 			    cur_key >= alg->alg_nkey_sizes) {
2610 				ss1dbg(spds, ("spdsock_do_updatealg: "
2611 				    "too many key sizes\n"));
2612 				*diag = SPD_DIAGNOSTIC_ALG_NUM_KEY_SIZES;
2613 				goto bail;
2614 			}
2615 			alg->alg_key_sizes[cur_key++] = attr->spd_attr_value;
2616 			break;
2617 
2618 		case SPD_ATTR_ALG_NBLOCKSIZES:
2619 			if (alg->alg_block_sizes != NULL) {
2620 				kmem_free(alg->alg_block_sizes,
2621 				    ALG_BLOCK_SIZES(alg));
2622 			}
2623 			alg->alg_nblock_sizes = attr->spd_attr_value;
2624 			/*
2625 			 * Allocate room for the trailing zero block size
2626 			 * value as well.
2627 			 */
2628 			alg->alg_block_sizes = kmem_zalloc(ALG_BLOCK_SIZES(alg),
2629 			    KM_SLEEP);
2630 			cur_block = 0;
2631 			break;
2632 
2633 		case SPD_ATTR_ALG_BLOCKSIZE:
2634 			if (alg->alg_block_sizes == NULL ||
2635 			    cur_block >= alg->alg_nblock_sizes) {
2636 				ss1dbg(spds, ("spdsock_do_updatealg: "
2637 				    "too many block sizes\n"));
2638 				*diag = SPD_DIAGNOSTIC_ALG_NUM_BLOCK_SIZES;
2639 				goto bail;
2640 			}
2641 			alg->alg_block_sizes[cur_block++] =
2642 			    attr->spd_attr_value;
2643 			break;
2644 
2645 		case SPD_ATTR_ALG_MECHNAME: {
2646 			char *mech_name;
2647 
2648 			if (attr->spd_attr_value > CRYPTO_MAX_MECH_NAME) {
2649 				ss1dbg(spds, ("spdsock_do_updatealg: "
2650 				    "mech name too long\n"));
2651 				*diag = SPD_DIAGNOSTIC_ALG_MECH_NAME_LEN;
2652 				goto bail;
2653 			}
2654 			mech_name = (char *)(attr + 1);
2655 			bcopy(mech_name, alg->alg_mech_name,
2656 			    attr->spd_attr_value);
2657 			alg->alg_mech_name[CRYPTO_MAX_MECH_NAME-1] = '\0';
2658 			attr = (struct spd_attribute *)((char *)attr +
2659 			    attr->spd_attr_value);
2660 			break;
2661 		}
2662 
2663 		case SPD_ATTR_PROTO_ID:
2664 			doing_proto = B_TRUE;
2665 			for (i = 0; i < NALGPROTOS; i++) {
2666 				if (algproto[i] == attr->spd_attr_value) {
2667 					alg_type = i;
2668 					break;
2669 				}
2670 			}
2671 			break;
2672 
2673 		case SPD_ATTR_PROTO_EXEC_MODE:
2674 			if (!doing_proto)
2675 				break;
2676 			for (i = 0; i < NEXECMODES; i++) {
2677 				if (execmodes[i] == attr->spd_attr_value) {
2678 					spds->spds_algs_exec_mode[alg_type] = i;
2679 					break;
2680 				}
2681 			}
2682 			break;
2683 		}
2684 		attr++;
2685 	}
2686 
2687 #undef	ALG_KEY_SIZES
2688 #undef	ALG_BLOCK_SIZES
2689 
2690 	/* update the algorithm tables */
2691 	spdsock_merge_algs(spds);
2692 bail:
2693 	/* cleanup */
2694 	ipsec_alg_free(alg);
2695 	for (alg_type = 0; alg_type < IPSEC_NALGTYPES; alg_type++)
2696 		for (algid = 0; algid < IPSEC_MAX_ALGS; algid++)
2697 		if (spds->spds_algs[alg_type][algid] != NULL)
2698 			ipsec_alg_free(spds->spds_algs[alg_type][algid]);
2699 }
2700 
2701 /*
2702  * Process an SPD_UPDATEALGS request. If IPsec is not loaded, queue
2703  * the request until IPsec loads. If IPsec is loaded, act on it
2704  * immediately.
2705  */
2706 
2707 static void
2708 spdsock_updatealg(queue_t *q, mblk_t *mp, spd_ext_t *extv[])
2709 {
2710 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2711 	spd_stack_t	*spds = ss->spdsock_spds;
2712 	ipsec_stack_t	*ipss = spds->spds_netstack->netstack_ipsec;
2713 
2714 	if (!ipsec_loaded(ipss)) {
2715 		/*
2716 		 * IPsec is not loaded, save request and return nicely,
2717 		 * the message will be processed once IPsec loads.
2718 		 */
2719 		mblk_t *new_mp;
2720 
2721 		/* last update message wins */
2722 		if ((new_mp = copymsg(mp)) == NULL) {
2723 			spdsock_error(q, mp, ENOMEM, 0);
2724 			return;
2725 		}
2726 		mutex_enter(&spds->spds_alg_lock);
2727 		bcopy(extv, spds->spds_extv_algs,
2728 		    sizeof (spd_ext_t *) * (SPD_EXT_MAX + 1));
2729 		if (spds->spds_mp_algs != NULL)
2730 			freemsg(spds->spds_mp_algs);
2731 		spds->spds_mp_algs = mp;
2732 		spds->spds_algs_pending = B_TRUE;
2733 		mutex_exit(&spds->spds_alg_lock);
2734 		if (audit_active) {
2735 			cred_t *cr;
2736 			pid_t cpid;
2737 
2738 			cr = msg_getcred(mp, &cpid);
2739 			audit_pf_policy(SPD_UPDATEALGS, cr,
2740 			    spds->spds_netstack, NULL, B_TRUE, EAGAIN,
2741 			    cpid);
2742 		}
2743 		spd_echo(q, new_mp);
2744 	} else {
2745 		/*
2746 		 * IPsec is loaded, act on the message immediately.
2747 		 */
2748 		int diag;
2749 
2750 		mutex_enter(&spds->spds_alg_lock);
2751 		spdsock_do_updatealg(extv, &diag, spds);
2752 		mutex_exit(&spds->spds_alg_lock);
2753 		if (diag == -1) {
2754 			spd_echo(q, mp);
2755 			if (audit_active) {
2756 				cred_t *cr;
2757 				pid_t cpid;
2758 
2759 				cr = msg_getcred(mp, &cpid);
2760 				audit_pf_policy(SPD_UPDATEALGS, cr,
2761 				    spds->spds_netstack, NULL, B_TRUE, 0,
2762 				    cpid);
2763 			}
2764 		} else {
2765 			spdsock_diag(q, mp, diag);
2766 			if (audit_active) {
2767 				cred_t *cr;
2768 				pid_t cpid;
2769 
2770 				cr = msg_getcred(mp, &cpid);
2771 				audit_pf_policy(SPD_UPDATEALGS, cr,
2772 				    spds->spds_netstack, NULL, B_TRUE, diag,
2773 				    cpid);
2774 			}
2775 		}
2776 	}
2777 }
2778 
2779 /*
2780  * With a reference-held ill, dig down and find an instance of "tun", and
2781  * assign its tunnel policy pointer, while reference-holding it.  Also,
2782  * release ill's refrence when finished.
2783  *
2784  * We'll be messing with q_next, so be VERY careful.
2785  */
2786 static void
2787 find_tun_and_set_itp(ill_t *ill, ipsec_tun_pol_t *itp)
2788 {
2789 	queue_t *q;
2790 	tun_t *tun;
2791 
2792 	/* Don't bother if this ill is going away. */
2793 	if (ill->ill_flags & ILL_CONDEMNED) {
2794 		ill_refrele(ill);
2795 		return;
2796 	}
2797 
2798 
2799 	q = ill->ill_wq;
2800 	claimstr(q);	/* Lighter-weight than freezestr(). */
2801 
2802 	do {
2803 		/* Use strcmp() because "tun" is bounded. */
2804 		if (strcmp(q->q_qinfo->qi_minfo->mi_idname, "tun") == 0) {
2805 			/* Aha!  Got it. */
2806 			tun = (tun_t *)q->q_ptr;
2807 			if (tun != NULL) {
2808 				mutex_enter(&tun->tun_lock);
2809 				if (tun->tun_itp != itp) {
2810 					ASSERT(tun->tun_itp == NULL);
2811 					ITP_REFHOLD(itp);
2812 					tun->tun_itp = itp;
2813 				}
2814 				mutex_exit(&tun->tun_lock);
2815 				goto release_and_return;
2816 			}
2817 			/*
2818 			 * Else assume this is some other module named "tun"
2819 			 * and move on, hoping we find one that actually has
2820 			 * something in q_ptr.
2821 			 */
2822 		}
2823 		q = q->q_next;
2824 	} while (q != NULL);
2825 
2826 release_and_return:
2827 	releasestr(ill->ill_wq);
2828 	ill_refrele(ill);
2829 }
2830 
2831 /*
2832  * Sort through the mess of polhead options to retrieve an appropriate one.
2833  * Returns NULL if we send an spdsock error.  Returns a valid pointer if we
2834  * found a valid polhead.  Returns ALL_ACTIVE_POLHEADS (aka. -1) or
2835  * ALL_INACTIVE_POLHEADS (aka. -2) if the operation calls for the operation to
2836  * act on ALL policy heads.
2837  */
2838 static ipsec_policy_head_t *
2839 get_appropriate_polhead(queue_t *q, mblk_t *mp, spd_if_t *tunname, int spdid,
2840     int msgtype, ipsec_tun_pol_t **itpp)
2841 {
2842 	ipsec_tun_pol_t *itp;
2843 	ipsec_policy_head_t *iph;
2844 	int errno;
2845 	char *tname;
2846 	boolean_t active;
2847 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2848 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
2849 	uint64_t gen;	/* Placeholder */
2850 	ill_t *v4, *v6;
2851 
2852 	active = (spdid == SPD_ACTIVE);
2853 	*itpp = NULL;
2854 	if (!active && spdid != SPD_STANDBY) {
2855 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_BAD_SPDID);
2856 		return (NULL);
2857 	}
2858 
2859 	if (tunname != NULL) {
2860 		/* Acting on a tunnel's SPD. */
2861 		tname = (char *)tunname->spd_if_name;
2862 		if (*tname == '\0') {
2863 			/* Handle all-polhead cases here. */
2864 			if (msgtype != SPD_FLUSH && msgtype != SPD_DUMP) {
2865 				spdsock_diag(q, mp,
2866 				    SPD_DIAGNOSTIC_NOT_GLOBAL_OP);
2867 				return (NULL);
2868 			}
2869 			return (active ? ALL_ACTIVE_POLHEADS :
2870 			    ALL_INACTIVE_POLHEADS);
2871 		}
2872 
2873 		itp = get_tunnel_policy(tname, ns);
2874 		if (itp == NULL) {
2875 			if (msgtype != SPD_ADDRULE) {
2876 				/* "Tunnel not found" */
2877 				spdsock_error(q, mp, ENOENT, 0);
2878 				return (NULL);
2879 			}
2880 
2881 			errno = 0;
2882 			itp = create_tunnel_policy(tname, &errno, &gen, ns);
2883 			if (itp == NULL) {
2884 				/*
2885 				 * Something very bad happened, most likely
2886 				 * ENOMEM.  Return an indicator.
2887 				 */
2888 				spdsock_error(q, mp, errno, 0);
2889 				return (NULL);
2890 			}
2891 		}
2892 		/*
2893 		 * Troll the plumbed tunnels and see if we have a
2894 		 * match.  We need to do this always in case we add
2895 		 * policy AFTER plumbing a tunnel.
2896 		 */
2897 		v4 = ill_lookup_on_name(tname, B_FALSE, B_FALSE, NULL,
2898 		    NULL, NULL, &errno, NULL, ns->netstack_ip);
2899 		if (v4 != NULL)
2900 			find_tun_and_set_itp(v4, itp);
2901 		v6 = ill_lookup_on_name(tname, B_FALSE, B_TRUE, NULL,
2902 		    NULL, NULL, &errno, NULL, ns->netstack_ip);
2903 		if (v6 != NULL)
2904 			find_tun_and_set_itp(v6, itp);
2905 		ASSERT(itp != NULL);
2906 		*itpp = itp;
2907 		/* For spdsock dump state, set the polhead's name. */
2908 		if (msgtype == SPD_DUMP) {
2909 			ITP_REFHOLD(itp);
2910 			ss->spdsock_itp = itp;
2911 			ss->spdsock_dump_tunnel = itp->itp_flags &
2912 			    (active ? ITPF_P_TUNNEL : ITPF_I_TUNNEL);
2913 		}
2914 	} else {
2915 		itp = NULL;
2916 		/* For spdsock dump state, indicate it's global policy. */
2917 		if (msgtype == SPD_DUMP)
2918 			ss->spdsock_itp = NULL;
2919 	}
2920 
2921 	if (active)
2922 		iph = (itp == NULL) ? ipsec_system_policy(ns) : itp->itp_policy;
2923 	else
2924 		iph = (itp == NULL) ? ipsec_inactive_policy(ns) :
2925 		    itp->itp_inactive;
2926 
2927 	ASSERT(iph != NULL);
2928 	if (itp != NULL) {
2929 		IPPH_REFHOLD(iph);
2930 	}
2931 
2932 	return (iph);
2933 }
2934 
2935 static void
2936 spdsock_parse(queue_t *q, mblk_t *mp)
2937 {
2938 	spd_msg_t *spmsg;
2939 	spd_ext_t *extv[SPD_EXT_MAX + 1];
2940 	uint_t msgsize;
2941 	ipsec_policy_head_t *iph;
2942 	ipsec_tun_pol_t *itp;
2943 	spd_if_t *tunname;
2944 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2945 	spd_stack_t *spds = ss->spdsock_spds;
2946 	netstack_t *ns = spds->spds_netstack;
2947 	ipsec_stack_t *ipss = ns->netstack_ipsec;
2948 
2949 	/* Make sure nothing's below me. */
2950 	ASSERT(WR(q)->q_next == NULL);
2951 
2952 	spmsg = (spd_msg_t *)mp->b_rptr;
2953 
2954 	msgsize = SPD_64TO8(spmsg->spd_msg_len);
2955 
2956 	if (msgdsize(mp) != msgsize) {
2957 		/*
2958 		 * Message len incorrect w.r.t. actual size.  Send an error
2959 		 * (EMSGSIZE).	It may be necessary to massage things a
2960 		 * bit.	 For example, if the spd_msg_type is hosed,
2961 		 * I need to set it to SPD_RESERVED to get delivery to
2962 		 * do the right thing.	Then again, maybe just letting
2963 		 * the error delivery do the right thing.
2964 		 */
2965 		ss2dbg(spds,
2966 		    ("mblk (%lu) and base (%d) message sizes don't jibe.\n",
2967 		    msgdsize(mp), msgsize));
2968 		spdsock_error(q, mp, EMSGSIZE, SPD_DIAGNOSTIC_NONE);
2969 		return;
2970 	}
2971 
2972 	if (msgsize > (uint_t)(mp->b_wptr - mp->b_rptr)) {
2973 		/* Get all message into one mblk. */
2974 		if (pullupmsg(mp, -1) == 0) {
2975 			/*
2976 			 * Something screwy happened.
2977 			 */
2978 			ss3dbg(spds, ("spdsock_parse: pullupmsg() failed.\n"));
2979 			return;
2980 		} else {
2981 			spmsg = (spd_msg_t *)mp->b_rptr;
2982 		}
2983 	}
2984 
2985 	switch (spdsock_get_ext(extv, spmsg, msgsize)) {
2986 	case KGE_DUP:
2987 		/* Handle duplicate extension. */
2988 		ss1dbg(spds, ("Got duplicate extension of type %d.\n",
2989 		    extv[0]->spd_ext_type));
2990 		spdsock_diag(q, mp, dup_ext_diag[extv[0]->spd_ext_type]);
2991 		return;
2992 	case KGE_UNK:
2993 		/* Handle unknown extension. */
2994 		ss1dbg(spds, ("Got unknown extension of type %d.\n",
2995 		    extv[0]->spd_ext_type));
2996 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_UNKNOWN_EXT);
2997 		return;
2998 	case KGE_LEN:
2999 		/* Length error. */
3000 		ss1dbg(spds, ("Length %d on extension type %d overrun or 0.\n",
3001 		    extv[0]->spd_ext_len, extv[0]->spd_ext_type));
3002 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_BAD_EXTLEN);
3003 		return;
3004 	case KGE_CHK:
3005 		/* Reality check failed. */
3006 		ss1dbg(spds, ("Reality check failed on extension type %d.\n",
3007 		    extv[0]->spd_ext_type));
3008 		spdsock_diag(q, mp, bad_ext_diag[extv[0]->spd_ext_type]);
3009 		return;
3010 	default:
3011 		/* Default case is no errors. */
3012 		break;
3013 	}
3014 
3015 	/*
3016 	 * Special-case SPD_UPDATEALGS so as not to load IPsec.
3017 	 */
3018 	if (!ipsec_loaded(ipss) && spmsg->spd_msg_type != SPD_UPDATEALGS) {
3019 		spdsock_t *ss = (spdsock_t *)q->q_ptr;
3020 
3021 		ASSERT(ss != NULL);
3022 		ipsec_loader_loadnow(ipss);
3023 		ss->spdsock_timeout_arg = mp;
3024 		ss->spdsock_timeout = qtimeout(q, spdsock_loadcheck,
3025 		    q, LOADCHECK_INTERVAL);
3026 		return;
3027 	}
3028 
3029 	/* First check for messages that need no polheads at all. */
3030 	switch (spmsg->spd_msg_type) {
3031 	case SPD_UPDATEALGS:
3032 		spdsock_updatealg(q, mp, extv);
3033 		return;
3034 	case SPD_ALGLIST:
3035 		spdsock_alglist(q, mp);
3036 		return;
3037 	case SPD_DUMPALGS:
3038 		spdsock_dumpalgs(q, mp);
3039 		return;
3040 	}
3041 
3042 	/*
3043 	 * Then check for ones that need both primary/secondary polheads,
3044 	 * finding the appropriate tunnel policy if need be.
3045 	 */
3046 	tunname = (spd_if_t *)extv[SPD_EXT_TUN_NAME];
3047 	switch (spmsg->spd_msg_type) {
3048 	case SPD_FLIP:
3049 		spdsock_flip(q, mp, tunname);
3050 		return;
3051 	case SPD_CLONE:
3052 		spdsock_clone(q, mp, tunname);
3053 		return;
3054 	}
3055 
3056 	/*
3057 	 * Finally, find ones that operate on exactly one polhead, or
3058 	 * "all polheads" of a given type (active/inactive).
3059 	 */
3060 	iph = get_appropriate_polhead(q, mp, tunname, spmsg->spd_msg_spdid,
3061 	    spmsg->spd_msg_type, &itp);
3062 	if (iph == NULL)
3063 		return;
3064 
3065 	/* All-polheads-ready operations. */
3066 	switch (spmsg->spd_msg_type) {
3067 	case SPD_FLUSH:
3068 		if (itp != NULL) {
3069 			mutex_enter(&itp->itp_lock);
3070 			if (spmsg->spd_msg_spdid == SPD_ACTIVE)
3071 				itp->itp_flags &= ~ITPF_PFLAGS;
3072 			else
3073 				itp->itp_flags &= ~ITPF_IFLAGS;
3074 			mutex_exit(&itp->itp_lock);
3075 			ITP_REFRELE(itp, ns);
3076 		}
3077 		spdsock_flush(q, iph, itp, mp);
3078 		return;
3079 	case SPD_DUMP:
3080 		if (itp != NULL)
3081 			ITP_REFRELE(itp, ns);
3082 		spdsock_dump(q, iph, mp);
3083 		return;
3084 	}
3085 
3086 	if (iph == ALL_ACTIVE_POLHEADS || iph == ALL_INACTIVE_POLHEADS) {
3087 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_NOT_GLOBAL_OP);
3088 		return;
3089 	}
3090 
3091 	/* Single-polhead-only operations. */
3092 	switch (spmsg->spd_msg_type) {
3093 	case SPD_ADDRULE:
3094 		spdsock_addrule(q, iph, mp, extv, itp);
3095 		break;
3096 	case SPD_DELETERULE:
3097 		spdsock_deleterule(q, iph, mp, extv, itp);
3098 		break;
3099 	case SPD_LOOKUP:
3100 		spdsock_lookup(q, iph, mp, extv, itp);
3101 		break;
3102 	default:
3103 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_BAD_MSG_TYPE);
3104 		break;
3105 	}
3106 
3107 	IPPH_REFRELE(iph, ns);
3108 	if (itp != NULL)
3109 		ITP_REFRELE(itp, ns);
3110 }
3111 
3112 /*
3113  * If an algorithm mapping was received before IPsec was loaded, process it.
3114  * Called from the IPsec loader.
3115  */
3116 void
3117 spdsock_update_pending_algs(netstack_t *ns)
3118 {
3119 	spd_stack_t *spds = ns->netstack_spdsock;
3120 
3121 	mutex_enter(&spds->spds_alg_lock);
3122 	if (spds->spds_algs_pending) {
3123 		int diag;
3124 
3125 		spdsock_do_updatealg(spds->spds_extv_algs, &diag,
3126 		    spds);
3127 		spds->spds_algs_pending = B_FALSE;
3128 	}
3129 	mutex_exit(&spds->spds_alg_lock);
3130 }
3131 
3132 static void
3133 spdsock_loadcheck(void *arg)
3134 {
3135 	queue_t *q = (queue_t *)arg;
3136 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3137 	mblk_t *mp;
3138 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
3139 
3140 	ASSERT(ss != NULL);
3141 
3142 	ss->spdsock_timeout = 0;
3143 	mp = ss->spdsock_timeout_arg;
3144 	ASSERT(mp != NULL);
3145 	ss->spdsock_timeout_arg = NULL;
3146 	if (ipsec_failed(ipss))
3147 		spdsock_error(q, mp, EPROTONOSUPPORT, 0);
3148 	else
3149 		spdsock_parse(q, mp);
3150 }
3151 
3152 /*
3153  * Copy relevant state bits.
3154  */
3155 static void
3156 spdsock_copy_info(struct T_info_ack *tap, spdsock_t *ss)
3157 {
3158 	*tap = spdsock_g_t_info_ack;
3159 	tap->CURRENT_state = ss->spdsock_state;
3160 	tap->OPT_size = spdsock_max_optsize;
3161 }
3162 
3163 /*
3164  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
3165  * spdsock_wput.  Much of the T_CAPABILITY_ACK information is copied from
3166  * spdsock_g_t_info_ack.  The current state of the stream is copied from
3167  * spdsock_state.
3168  */
3169 static void
3170 spdsock_capability_req(queue_t *q, mblk_t *mp)
3171 {
3172 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3173 	t_uscalar_t cap_bits1;
3174 	struct T_capability_ack	*tcap;
3175 
3176 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
3177 
3178 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
3179 	    mp->b_datap->db_type, T_CAPABILITY_ACK);
3180 	if (mp == NULL)
3181 		return;
3182 
3183 	tcap = (struct T_capability_ack *)mp->b_rptr;
3184 	tcap->CAP_bits1 = 0;
3185 
3186 	if (cap_bits1 & TC1_INFO) {
3187 		spdsock_copy_info(&tcap->INFO_ack, ss);
3188 		tcap->CAP_bits1 |= TC1_INFO;
3189 	}
3190 
3191 	qreply(q, mp);
3192 }
3193 
3194 /*
3195  * This routine responds to T_INFO_REQ messages. It is called by
3196  * spdsock_wput_other.
3197  * Most of the T_INFO_ACK information is copied from spdsock_g_t_info_ack.
3198  * The current state of the stream is copied from spdsock_state.
3199  */
3200 static void
3201 spdsock_info_req(q, mp)
3202 	queue_t	*q;
3203 	mblk_t	*mp;
3204 {
3205 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
3206 	    T_INFO_ACK);
3207 	if (mp == NULL)
3208 		return;
3209 	spdsock_copy_info((struct T_info_ack *)mp->b_rptr,
3210 	    (spdsock_t *)q->q_ptr);
3211 	qreply(q, mp);
3212 }
3213 
3214 /*
3215  * spdsock_err_ack. This routine creates a
3216  * T_ERROR_ACK message and passes it
3217  * upstream.
3218  */
3219 static void
3220 spdsock_err_ack(q, mp, t_error, sys_error)
3221 	queue_t	*q;
3222 	mblk_t	*mp;
3223 	int	t_error;
3224 	int	sys_error;
3225 {
3226 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
3227 		qreply(q, mp);
3228 }
3229 
3230 /*
3231  * This routine retrieves the current status of socket options.
3232  * It returns the size of the option retrieved.
3233  */
3234 /* ARGSUSED */
3235 int
3236 spdsock_opt_get(queue_t *q, int level, int name, uchar_t *ptr)
3237 {
3238 	int *i1 = (int *)ptr;
3239 
3240 	switch (level) {
3241 	case SOL_SOCKET:
3242 		switch (name) {
3243 		case SO_TYPE:
3244 			*i1 = SOCK_RAW;
3245 			break;
3246 		/*
3247 		 * The following two items can be manipulated,
3248 		 * but changing them should do nothing.
3249 		 */
3250 		case SO_SNDBUF:
3251 			*i1 = (int)q->q_hiwat;
3252 			break;
3253 		case SO_RCVBUF:
3254 			*i1 = (int)(RD(q)->q_hiwat);
3255 			break;
3256 		}
3257 		break;
3258 	default:
3259 		return (0);
3260 	}
3261 	return (sizeof (int));
3262 }
3263 
3264 /*
3265  * This routine sets socket options.
3266  */
3267 /* ARGSUSED */
3268 int
3269 spdsock_opt_set(queue_t *q, uint_t mgmt_flags, int level, int name,
3270     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3271     void *thisdg_attrs, cred_t *cr, mblk_t *mblk)
3272 {
3273 	int *i1 = (int *)invalp;
3274 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3275 	spd_stack_t	*spds = ss->spdsock_spds;
3276 
3277 	switch (level) {
3278 	case SOL_SOCKET:
3279 		switch (name) {
3280 		case SO_SNDBUF:
3281 			if (*i1 > spds->spds_max_buf)
3282 				return (ENOBUFS);
3283 			q->q_hiwat = *i1;
3284 			break;
3285 		case SO_RCVBUF:
3286 			if (*i1 > spds->spds_max_buf)
3287 				return (ENOBUFS);
3288 			RD(q)->q_hiwat = *i1;
3289 			(void) proto_set_rx_hiwat(RD(q), NULL, *i1);
3290 			break;
3291 		}
3292 		break;
3293 	}
3294 	return (0);
3295 }
3296 
3297 
3298 /*
3299  * Handle STREAMS messages.
3300  */
3301 static void
3302 spdsock_wput_other(queue_t *q, mblk_t *mp)
3303 {
3304 	struct iocblk *iocp;
3305 	int error;
3306 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3307 	spd_stack_t	*spds = ss->spdsock_spds;
3308 	cred_t		*cr;
3309 
3310 	switch (mp->b_datap->db_type) {
3311 	case M_PROTO:
3312 	case M_PCPROTO:
3313 		if ((mp->b_wptr - mp->b_rptr) < sizeof (long)) {
3314 			ss3dbg(spds, (
3315 			    "spdsock_wput_other: Not big enough M_PROTO\n"));
3316 			freemsg(mp);
3317 			return;
3318 		}
3319 		switch (((union T_primitives *)mp->b_rptr)->type) {
3320 		case T_CAPABILITY_REQ:
3321 			spdsock_capability_req(q, mp);
3322 			break;
3323 		case T_INFO_REQ:
3324 			spdsock_info_req(q, mp);
3325 			break;
3326 		case T_SVR4_OPTMGMT_REQ:
3327 		case T_OPTMGMT_REQ:
3328 			/*
3329 			 * All Solaris components should pass a db_credp
3330 			 * for this TPI message, hence we ASSERT.
3331 			 * But in case there is some other M_PROTO that looks
3332 			 * like a TPI message sent by some other kernel
3333 			 * component, we check and return an error.
3334 			 */
3335 			cr = msg_getcred(mp, NULL);
3336 			ASSERT(cr != NULL);
3337 			if (cr == NULL) {
3338 				spdsock_err_ack(q, mp, TSYSERR, EINVAL);
3339 				return;
3340 			}
3341 			if (((union T_primitives *)mp->b_rptr)->type ==
3342 			    T_SVR4_OPTMGMT_REQ) {
3343 				(void) svr4_optcom_req(q, mp, cr,
3344 				    &spdsock_opt_obj, B_FALSE);
3345 			} else {
3346 				(void) tpi_optcom_req(q, mp, cr,
3347 				    &spdsock_opt_obj, B_FALSE);
3348 			}
3349 			break;
3350 		case T_DATA_REQ:
3351 		case T_EXDATA_REQ:
3352 		case T_ORDREL_REQ:
3353 			/* Illegal for spdsock. */
3354 			freemsg(mp);
3355 			(void) putnextctl1(RD(q), M_ERROR, EPROTO);
3356 			break;
3357 		default:
3358 			/* Not supported by spdsock. */
3359 			spdsock_err_ack(q, mp, TNOTSUPPORT, 0);
3360 			break;
3361 		}
3362 		return;
3363 	case M_IOCTL:
3364 		iocp = (struct iocblk *)mp->b_rptr;
3365 		error = EINVAL;
3366 
3367 		switch (iocp->ioc_cmd) {
3368 		case ND_SET:
3369 		case ND_GET:
3370 			if (nd_getset(q, spds->spds_g_nd, mp)) {
3371 				qreply(q, mp);
3372 				return;
3373 			} else
3374 				error = ENOENT;
3375 			/* FALLTHRU */
3376 		default:
3377 			miocnak(q, mp, 0, error);
3378 			return;
3379 		}
3380 	case M_FLUSH:
3381 		if (*mp->b_rptr & FLUSHW) {
3382 			flushq(q, FLUSHALL);
3383 			*mp->b_rptr &= ~FLUSHW;
3384 		}
3385 		if (*mp->b_rptr & FLUSHR) {
3386 			qreply(q, mp);
3387 			return;
3388 		}
3389 		/* Else FALLTHRU */
3390 	}
3391 
3392 	/* If fell through, just black-hole the message. */
3393 	freemsg(mp);
3394 }
3395 
3396 static void
3397 spdsock_wput(queue_t *q, mblk_t *mp)
3398 {
3399 	uint8_t *rptr = mp->b_rptr;
3400 	mblk_t *mp1;
3401 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3402 	spd_stack_t	*spds = ss->spdsock_spds;
3403 
3404 	/*
3405 	 * If we're dumping, defer processing other messages until the
3406 	 * dump completes.
3407 	 */
3408 	if (ss->spdsock_dump_req != NULL) {
3409 		if (!putq(q, mp))
3410 			freemsg(mp);
3411 		return;
3412 	}
3413 
3414 	switch (mp->b_datap->db_type) {
3415 	case M_DATA:
3416 		/*
3417 		 * Silently discard.
3418 		 */
3419 		ss2dbg(spds, ("raw M_DATA in spdsock.\n"));
3420 		freemsg(mp);
3421 		return;
3422 	case M_PROTO:
3423 	case M_PCPROTO:
3424 		if ((mp->b_wptr - rptr) >= sizeof (struct T_data_req)) {
3425 			if (((union T_primitives *)rptr)->type == T_DATA_REQ) {
3426 				if ((mp1 = mp->b_cont) == NULL) {
3427 					/* No data after T_DATA_REQ. */
3428 					ss2dbg(spds,
3429 					    ("No data after DATA_REQ.\n"));
3430 					freemsg(mp);
3431 					return;
3432 				}
3433 				freeb(mp);
3434 				mp = mp1;
3435 				ss2dbg(spds, ("T_DATA_REQ\n"));
3436 				break;	/* Out of switch. */
3437 			}
3438 		}
3439 		/* FALLTHRU */
3440 	default:
3441 		ss3dbg(spds, ("In default wput case (%d %d).\n",
3442 		    mp->b_datap->db_type, ((union T_primitives *)rptr)->type));
3443 		spdsock_wput_other(q, mp);
3444 		return;
3445 	}
3446 
3447 	/* I now have a PF_POLICY message in an M_DATA block. */
3448 	spdsock_parse(q, mp);
3449 }
3450 
3451 /*
3452  * Device open procedure, called when new queue pair created.
3453  * We are passed the read-side queue.
3454  */
3455 /* ARGSUSED */
3456 static int
3457 spdsock_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
3458 {
3459 	spdsock_t *ss;
3460 	queue_t *oq = OTHERQ(q);
3461 	minor_t ssminor;
3462 	netstack_t *ns;
3463 	spd_stack_t *spds;
3464 
3465 	if (secpolicy_ip_config(credp, B_FALSE) != 0)
3466 		return (EPERM);
3467 
3468 	if (q->q_ptr != NULL)
3469 		return (0);  /* Re-open of an already open instance. */
3470 
3471 	if (sflag & MODOPEN)
3472 		return (EINVAL);
3473 
3474 	ns = netstack_find_by_cred(credp);
3475 	ASSERT(ns != NULL);
3476 	spds = ns->netstack_spdsock;
3477 	ASSERT(spds != NULL);
3478 
3479 	ss2dbg(spds, ("Made it into PF_POLICY socket open.\n"));
3480 
3481 	ssminor = (minor_t)(uintptr_t)vmem_alloc(spdsock_vmem, 1, VM_NOSLEEP);
3482 	if (ssminor == 0) {
3483 		netstack_rele(spds->spds_netstack);
3484 		return (ENOMEM);
3485 	}
3486 	ss = kmem_zalloc(sizeof (spdsock_t), KM_NOSLEEP);
3487 	if (ss == NULL) {
3488 		vmem_free(spdsock_vmem, (void *)(uintptr_t)ssminor, 1);
3489 		netstack_rele(spds->spds_netstack);
3490 		return (ENOMEM);
3491 	}
3492 
3493 	ss->spdsock_minor = ssminor;
3494 	ss->spdsock_state = TS_UNBND;
3495 	ss->spdsock_dump_req = NULL;
3496 
3497 	ss->spdsock_spds = spds;
3498 
3499 	q->q_ptr = ss;
3500 	oq->q_ptr = ss;
3501 
3502 	q->q_hiwat = spds->spds_recv_hiwat;
3503 
3504 	oq->q_hiwat = spds->spds_xmit_hiwat;
3505 	oq->q_lowat = spds->spds_xmit_lowat;
3506 
3507 	qprocson(q);
3508 	(void) proto_set_rx_hiwat(q, NULL, spds->spds_recv_hiwat);
3509 
3510 	*devp = makedevice(getmajor(*devp), ss->spdsock_minor);
3511 	return (0);
3512 }
3513 
3514 /*
3515  * Read-side service procedure, invoked when we get back-enabled
3516  * when buffer space becomes available.
3517  *
3518  * Dump another chunk if we were dumping before; when we finish, kick
3519  * the write-side queue in case it's waiting for read queue space.
3520  */
3521 void
3522 spdsock_rsrv(queue_t *q)
3523 {
3524 	spdsock_t *ss = q->q_ptr;
3525 
3526 	if (ss->spdsock_dump_req != NULL)
3527 		spdsock_dump_some(q, ss);
3528 
3529 	if (ss->spdsock_dump_req == NULL)
3530 		qenable(OTHERQ(q));
3531 }
3532 
3533 /*
3534  * Write-side service procedure, invoked when we defer processing
3535  * if another message is received while a dump is in progress.
3536  */
3537 void
3538 spdsock_wsrv(queue_t *q)
3539 {
3540 	spdsock_t *ss = q->q_ptr;
3541 	mblk_t *mp;
3542 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
3543 
3544 	if (ss->spdsock_dump_req != NULL) {
3545 		qenable(OTHERQ(q));
3546 		return;
3547 	}
3548 
3549 	while ((mp = getq(q)) != NULL) {
3550 		if (ipsec_loaded(ipss)) {
3551 			spdsock_wput(q, mp);
3552 			if (ss->spdsock_dump_req != NULL)
3553 				return;
3554 		} else if (!ipsec_failed(ipss)) {
3555 			(void) putq(q, mp);
3556 		} else {
3557 			spdsock_error(q, mp, EPFNOSUPPORT, 0);
3558 		}
3559 	}
3560 }
3561 
3562 static int
3563 spdsock_close(queue_t *q)
3564 {
3565 	spdsock_t *ss = q->q_ptr;
3566 	spd_stack_t	*spds = ss->spdsock_spds;
3567 
3568 	qprocsoff(q);
3569 
3570 	/* Safe assumption. */
3571 	ASSERT(ss != NULL);
3572 
3573 	if (ss->spdsock_timeout != 0)
3574 		(void) quntimeout(q, ss->spdsock_timeout);
3575 
3576 	ss3dbg(spds, ("Driver close, PF_POLICY socket is going away.\n"));
3577 
3578 	vmem_free(spdsock_vmem, (void *)(uintptr_t)ss->spdsock_minor, 1);
3579 	netstack_rele(ss->spdsock_spds->spds_netstack);
3580 
3581 	kmem_free(ss, sizeof (spdsock_t));
3582 	return (0);
3583 }
3584 
3585 /*
3586  * Merge the IPsec algorithms tables with the received algorithm information.
3587  */
3588 void
3589 spdsock_merge_algs(spd_stack_t *spds)
3590 {
3591 	ipsec_alginfo_t *alg, *oalg;
3592 	ipsec_algtype_t algtype;
3593 	uint_t algidx, algid, nalgs;
3594 	crypto_mech_name_t *mechs;
3595 	uint_t mech_count, mech_idx;
3596 	netstack_t	*ns = spds->spds_netstack;
3597 	ipsec_stack_t	*ipss = ns->netstack_ipsec;
3598 
3599 	ASSERT(MUTEX_HELD(&spds->spds_alg_lock));
3600 
3601 	/*
3602 	 * Get the list of supported mechanisms from the crypto framework.
3603 	 * If a mechanism is supported by KCF, resolve its mechanism
3604 	 * id and mark it as being valid. This operation must be done
3605 	 * without holding alg_lock, since it can cause a provider
3606 	 * module to be loaded and the provider notification callback to
3607 	 * be invoked.
3608 	 */
3609 	mechs = crypto_get_mech_list(&mech_count, KM_SLEEP);
3610 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
3611 		for (algid = 0; algid < IPSEC_MAX_ALGS; algid++) {
3612 			int algflags = 0;
3613 			crypto_mech_type_t mt = CRYPTO_MECHANISM_INVALID;
3614 
3615 			alg = spds->spds_algs[algtype][algid];
3616 			if (alg == NULL)
3617 				continue;
3618 
3619 			/*
3620 			 * The NULL encryption algorithm is a special
3621 			 * case because there are no mechanisms, yet
3622 			 * the algorithm is still valid.
3623 			 */
3624 			if (alg->alg_id == SADB_EALG_NULL) {
3625 				alg->alg_mech_type = CRYPTO_MECHANISM_INVALID;
3626 				alg->alg_flags = ALG_FLAG_VALID;
3627 				continue;
3628 			}
3629 
3630 			for (mech_idx = 0; mech_idx < mech_count; mech_idx++) {
3631 				if (strncmp(alg->alg_mech_name, mechs[mech_idx],
3632 				    CRYPTO_MAX_MECH_NAME) == 0) {
3633 					mt = crypto_mech2id(alg->alg_mech_name);
3634 					ASSERT(mt != CRYPTO_MECHANISM_INVALID);
3635 					algflags = ALG_FLAG_VALID;
3636 					break;
3637 				}
3638 			}
3639 			alg->alg_mech_type = mt;
3640 			alg->alg_flags = algflags;
3641 		}
3642 	}
3643 
3644 	mutex_enter(&ipss->ipsec_alg_lock);
3645 
3646 	/*
3647 	 * For each algorithm currently defined, check if it is
3648 	 * present in the new tables created from the SPD_UPDATEALGS
3649 	 * message received from user-space.
3650 	 * Delete the algorithm entries that are currently defined
3651 	 * but not part of the new tables.
3652 	 */
3653 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
3654 		nalgs = ipss->ipsec_nalgs[algtype];
3655 		for (algidx = 0; algidx < nalgs; algidx++) {
3656 			algid = ipss->ipsec_sortlist[algtype][algidx];
3657 			if (spds->spds_algs[algtype][algid] == NULL)
3658 				ipsec_alg_unreg(algtype, algid, ns);
3659 		}
3660 	}
3661 
3662 	/*
3663 	 * For each algorithm we just received, check if it is
3664 	 * present in the currently defined tables. If it is, swap
3665 	 * the entry with the one we just allocated.
3666 	 * If the new algorithm is not in the current tables,
3667 	 * add it.
3668 	 */
3669 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
3670 		for (algid = 0; algid < IPSEC_MAX_ALGS; algid++) {
3671 			alg = spds->spds_algs[algtype][algid];
3672 			if (alg == NULL)
3673 				continue;
3674 
3675 			if ((oalg = ipss->ipsec_alglists[algtype][algid]) ==
3676 			    NULL) {
3677 				/*
3678 				 * New algorithm, add it to the algorithm
3679 				 * table.
3680 				 */
3681 				ipsec_alg_reg(algtype, alg, ns);
3682 			} else {
3683 				/*
3684 				 * Algorithm is already in the table. Swap
3685 				 * the existing entry with the new one.
3686 				 */
3687 				ipsec_alg_fix_min_max(alg, algtype, ns);
3688 				ipss->ipsec_alglists[algtype][algid] = alg;
3689 				ipsec_alg_free(oalg);
3690 			}
3691 			spds->spds_algs[algtype][algid] = NULL;
3692 		}
3693 	}
3694 
3695 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
3696 		ipss->ipsec_algs_exec_mode[algtype] =
3697 		    spds->spds_algs_exec_mode[algtype];
3698 	}
3699 
3700 	mutex_exit(&ipss->ipsec_alg_lock);
3701 
3702 	crypto_free_mech_list(mechs, mech_count);
3703 
3704 	ipsecah_algs_changed(ns);
3705 	ipsecesp_algs_changed(ns);
3706 }
3707