xref: /titanic_41/usr/src/uts/common/inet/ip/spdsock.c (revision d61c759d7e296bb673b1d010970c16872c9ef561)
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 2010 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 #include <sys/dls.h>
46 
47 #include <sys/socket.h>
48 #include <netinet/in.h>
49 #include <net/pfkeyv2.h>
50 #include <net/pfpolicy.h>
51 
52 #include <inet/common.h>
53 #include <netinet/ip6.h>
54 #include <inet/ip.h>
55 #include <inet/ip6.h>
56 #include <inet/mi.h>
57 #include <inet/proto_set.h>
58 #include <inet/nd.h>
59 #include <inet/ip_if.h>
60 #include <inet/optcom.h>
61 #include <inet/ipsec_impl.h>
62 #include <inet/spdsock.h>
63 #include <inet/sadb.h>
64 #include <inet/iptun.h>
65 #include <inet/iptun/iptun_impl.h>
66 
67 #include <sys/isa_defs.h>
68 
69 #include <c2/audit.h>
70 
71 /*
72  * This is a transport provider for the PF_POLICY IPsec policy
73  * management socket, which provides a management interface into the
74  * SPD, allowing policy rules to be added, deleted, and queried.
75  *
76  * This effectively replaces the old private SIOC*IPSECONFIG ioctls
77  * with an extensible interface which will hopefully be public some
78  * day.
79  *
80  * See <net/pfpolicy.h> for more details on the protocol.
81  *
82  * We link against drv/ip and call directly into it to manipulate the
83  * SPD; see ipsec_impl.h for the policy data structures and spd.c for
84  * the code which maintains them.
85  *
86  * The MT model of this is QPAIR with the addition of some explicit
87  * locking to protect system-wide policy data structures.
88  */
89 
90 static vmem_t *spdsock_vmem;		/* for minor numbers. */
91 
92 #define	ALIGNED64(x) IS_P2ALIGNED((x), sizeof (uint64_t))
93 
94 /* Default structure copied into T_INFO_ACK messages (from rts.c...) */
95 static struct T_info_ack spdsock_g_t_info_ack = {
96 	T_INFO_ACK,
97 	T_INFINITE,	/* TSDU_size. Maximum size messages. */
98 	T_INVALID,	/* ETSDU_size. No expedited data. */
99 	T_INVALID,	/* CDATA_size. No connect data. */
100 	T_INVALID,	/* DDATA_size. No disconnect data. */
101 	0,		/* ADDR_size. */
102 	0,		/* OPT_size. No user-settable options */
103 	64 * 1024,	/* TIDU_size. spdsock allows maximum size messages. */
104 	T_COTS,		/* SERV_type. spdsock supports connection oriented. */
105 	TS_UNBND,	/* CURRENT_state. This is set from spdsock_state. */
106 	(XPG4_1)	/* Provider flags */
107 };
108 
109 /* Named Dispatch Parameter Management Structure */
110 typedef struct spdsockparam_s {
111 	uint_t	spdsock_param_min;
112 	uint_t	spdsock_param_max;
113 	uint_t	spdsock_param_value;
114 	char *spdsock_param_name;
115 } spdsockparam_t;
116 
117 /*
118  * Table of NDD variables supported by spdsock. These are loaded into
119  * spdsock_g_nd in spdsock_init_nd.
120  * All of these are alterable, within the min/max values given, at run time.
121  */
122 static	spdsockparam_t	lcl_param_arr[] = {
123 	/* min	max	value	name */
124 	{ 4096, 65536,	8192,	"spdsock_xmit_hiwat"},
125 	{ 0,	65536,	1024,	"spdsock_xmit_lowat"},
126 	{ 4096, 65536,	8192,	"spdsock_recv_hiwat"},
127 	{ 65536, 1024*1024*1024, 256*1024,	"spdsock_max_buf"},
128 	{ 0,	3,	0,	"spdsock_debug"},
129 };
130 #define	spds_xmit_hiwat	spds_params[0].spdsock_param_value
131 #define	spds_xmit_lowat	spds_params[1].spdsock_param_value
132 #define	spds_recv_hiwat	spds_params[2].spdsock_param_value
133 #define	spds_max_buf	spds_params[3].spdsock_param_value
134 #define	spds_debug		spds_params[4].spdsock_param_value
135 
136 #define	ss0dbg(a)	printf a
137 /* NOTE:  != 0 instead of > 0 so lint doesn't complain. */
138 #define	ss1dbg(spds, a)	if (spds->spds_debug != 0) printf a
139 #define	ss2dbg(spds, a)	if (spds->spds_debug > 1) printf a
140 #define	ss3dbg(spds, a)	if (spds->spds_debug > 2) printf a
141 
142 #define	RESET_SPDSOCK_DUMP_POLHEAD(ss, iph) { \
143 	ASSERT(RW_READ_HELD(&(iph)->iph_lock)); \
144 	(ss)->spdsock_dump_head = (iph); \
145 	(ss)->spdsock_dump_gen = (iph)->iph_gen; \
146 	(ss)->spdsock_dump_cur_type = 0; \
147 	(ss)->spdsock_dump_cur_af = IPSEC_AF_V4; \
148 	(ss)->spdsock_dump_cur_rule = NULL; \
149 	(ss)->spdsock_dump_count = 0; \
150 	(ss)->spdsock_dump_cur_chain = 0; \
151 }
152 
153 static int spdsock_close(queue_t *);
154 static int spdsock_open(queue_t *, dev_t *, int, int, cred_t *);
155 static void spdsock_wput(queue_t *, mblk_t *);
156 static void spdsock_wsrv(queue_t *);
157 static void spdsock_rsrv(queue_t *);
158 static void *spdsock_stack_init(netstackid_t stackid, netstack_t *ns);
159 static void spdsock_stack_fini(netstackid_t stackid, void *arg);
160 static void spdsock_loadcheck(void *);
161 static void spdsock_merge_algs(spd_stack_t *);
162 static void spdsock_flush_one(ipsec_policy_head_t *, netstack_t *);
163 static mblk_t *spdsock_dump_next_record(spdsock_t *);
164 static void update_iptun_policy(ipsec_tun_pol_t *);
165 
166 static struct module_info info = {
167 	5138, "spdsock", 1, INFPSZ, 512, 128
168 };
169 
170 static struct qinit rinit = {
171 	NULL, (pfi_t)spdsock_rsrv, spdsock_open, spdsock_close,
172 	NULL, &info
173 };
174 
175 static struct qinit winit = {
176 	(pfi_t)spdsock_wput, (pfi_t)spdsock_wsrv, NULL, NULL, NULL, &info
177 };
178 
179 struct streamtab spdsockinfo = {
180 	&rinit, &winit
181 };
182 
183 /* mapping from alg type to protocol number, as per RFC 2407 */
184 static const uint_t algproto[] = {
185 	PROTO_IPSEC_AH,
186 	PROTO_IPSEC_ESP,
187 };
188 
189 #define	NALGPROTOS	(sizeof (algproto) / sizeof (algproto[0]))
190 
191 /* mapping from kernel exec mode to spdsock exec mode */
192 static const uint_t execmodes[] = {
193 	SPD_ALG_EXEC_MODE_SYNC,
194 	SPD_ALG_EXEC_MODE_ASYNC
195 };
196 
197 #define	NEXECMODES	(sizeof (execmodes) / sizeof (execmodes[0]))
198 
199 #define	ALL_ACTIVE_POLHEADS ((ipsec_policy_head_t *)-1)
200 #define	ALL_INACTIVE_POLHEADS ((ipsec_policy_head_t *)-2)
201 
202 #define	ITP_NAME(itp) (itp != NULL ? itp->itp_name : NULL)
203 
204 /* ARGSUSED */
205 static int
206 spdsock_param_get(q, mp, cp, cr)
207 	queue_t	*q;
208 	mblk_t	*mp;
209 	caddr_t	cp;
210 	cred_t *cr;
211 {
212 	spdsockparam_t	*spdsockpa = (spdsockparam_t *)cp;
213 	uint_t value;
214 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
215 	spd_stack_t	*spds = ss->spdsock_spds;
216 
217 	mutex_enter(&spds->spds_param_lock);
218 	value = spdsockpa->spdsock_param_value;
219 	mutex_exit(&spds->spds_param_lock);
220 
221 	(void) mi_mpprintf(mp, "%u", value);
222 	return (0);
223 }
224 
225 /* This routine sets an NDD variable in a spdsockparam_t structure. */
226 /* ARGSUSED */
227 static int
228 spdsock_param_set(q, mp, value, cp, cr)
229 	queue_t	*q;
230 	mblk_t	*mp;
231 	char *value;
232 	caddr_t	cp;
233 	cred_t *cr;
234 {
235 	ulong_t	new_value;
236 	spdsockparam_t	*spdsockpa = (spdsockparam_t *)cp;
237 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
238 	spd_stack_t	*spds = ss->spdsock_spds;
239 
240 	/* Convert the value from a string into a long integer. */
241 	if (ddi_strtoul(value, NULL, 10, &new_value) != 0)
242 		return (EINVAL);
243 
244 	mutex_enter(&spds->spds_param_lock);
245 	/*
246 	 * Fail the request if the new value does not lie within the
247 	 * required bounds.
248 	 */
249 	if (new_value < spdsockpa->spdsock_param_min ||
250 	    new_value > spdsockpa->spdsock_param_max) {
251 		mutex_exit(&spds->spds_param_lock);
252 		return (EINVAL);
253 	}
254 
255 	/* Set the new value */
256 	spdsockpa->spdsock_param_value = new_value;
257 	mutex_exit(&spds->spds_param_lock);
258 
259 	return (0);
260 }
261 
262 /*
263  * Initialize at module load time
264  */
265 boolean_t
266 spdsock_ddi_init(void)
267 {
268 	spdsock_max_optsize = optcom_max_optsize(
269 	    spdsock_opt_obj.odb_opt_des_arr, spdsock_opt_obj.odb_opt_arr_cnt);
270 
271 	spdsock_vmem = vmem_create("spdsock", (void *)1, MAXMIN, 1,
272 	    NULL, NULL, NULL, 1, VM_SLEEP | VMC_IDENTIFIER);
273 
274 	/*
275 	 * We want to be informed each time a stack is created or
276 	 * destroyed in the kernel, so we can maintain the
277 	 * set of spd_stack_t's.
278 	 */
279 	netstack_register(NS_SPDSOCK, spdsock_stack_init, NULL,
280 	    spdsock_stack_fini);
281 
282 	return (B_TRUE);
283 }
284 
285 /*
286  * Walk through the param array specified registering each element with the
287  * named dispatch handler.
288  */
289 static boolean_t
290 spdsock_param_register(IDP *ndp, spdsockparam_t *ssp, int cnt)
291 {
292 	for (; cnt-- > 0; ssp++) {
293 		if (ssp->spdsock_param_name != NULL &&
294 		    ssp->spdsock_param_name[0]) {
295 			if (!nd_load(ndp,
296 			    ssp->spdsock_param_name,
297 			    spdsock_param_get, spdsock_param_set,
298 			    (caddr_t)ssp)) {
299 				nd_free(ndp);
300 				return (B_FALSE);
301 			}
302 		}
303 	}
304 	return (B_TRUE);
305 }
306 
307 /*
308  * Initialize for each stack instance
309  */
310 /* ARGSUSED */
311 static void *
312 spdsock_stack_init(netstackid_t stackid, netstack_t *ns)
313 {
314 	spd_stack_t	*spds;
315 	spdsockparam_t	*ssp;
316 
317 	spds = (spd_stack_t *)kmem_zalloc(sizeof (*spds), KM_SLEEP);
318 	spds->spds_netstack = ns;
319 
320 	ASSERT(spds->spds_g_nd == NULL);
321 
322 	ssp = (spdsockparam_t *)kmem_alloc(sizeof (lcl_param_arr), KM_SLEEP);
323 	spds->spds_params = ssp;
324 	bcopy(lcl_param_arr, ssp, sizeof (lcl_param_arr));
325 
326 	(void) spdsock_param_register(&spds->spds_g_nd, ssp,
327 	    A_CNT(lcl_param_arr));
328 
329 	mutex_init(&spds->spds_param_lock, NULL, MUTEX_DEFAULT, NULL);
330 	mutex_init(&spds->spds_alg_lock, NULL, MUTEX_DEFAULT, NULL);
331 
332 	return (spds);
333 }
334 
335 void
336 spdsock_ddi_destroy(void)
337 {
338 	vmem_destroy(spdsock_vmem);
339 
340 	netstack_unregister(NS_SPDSOCK);
341 }
342 
343 /* ARGSUSED */
344 static void
345 spdsock_stack_fini(netstackid_t stackid, void *arg)
346 {
347 	spd_stack_t *spds = (spd_stack_t *)arg;
348 
349 	freemsg(spds->spds_mp_algs);
350 	mutex_destroy(&spds->spds_param_lock);
351 	mutex_destroy(&spds->spds_alg_lock);
352 	nd_free(&spds->spds_g_nd);
353 	kmem_free(spds->spds_params, sizeof (lcl_param_arr));
354 	spds->spds_params = NULL;
355 
356 	kmem_free(spds, sizeof (*spds));
357 }
358 
359 /*
360  * NOTE: large quantities of this should be shared with keysock.
361  * Would be nice to combine some of this into a common module, but
362  * not possible given time pressures.
363  */
364 
365 /*
366  * High-level reality checking of extensions.
367  */
368 /* ARGSUSED */ /* XXX */
369 static boolean_t
370 ext_check(spd_ext_t *ext)
371 {
372 	spd_if_t *tunname = (spd_if_t *)ext;
373 	int i;
374 	char *idstr;
375 
376 	if (ext->spd_ext_type == SPD_EXT_TUN_NAME) {
377 		/* (NOTE:  Modified from SADB_EXT_IDENTITY..) */
378 
379 		/*
380 		 * Make sure the strings in these identities are
381 		 * null-terminated.  Let's "proactively" null-terminate the
382 		 * string at the last byte if it's not terminated sooner.
383 		 */
384 		i = SPD_64TO8(tunname->spd_if_len) - sizeof (spd_if_t);
385 		idstr = (char *)(tunname + 1);
386 		while (*idstr != '\0' && i > 0) {
387 			i--;
388 			idstr++;
389 		}
390 		if (i == 0) {
391 			/*
392 			 * I.e., if the bozo user didn't NULL-terminate the
393 			 * string...
394 			 */
395 			idstr--;
396 			*idstr = '\0';
397 		}
398 	}
399 	return (B_TRUE);	/* For now... */
400 }
401 
402 
403 
404 /* Return values for spdsock_get_ext(). */
405 #define	KGE_OK	0
406 #define	KGE_DUP	1
407 #define	KGE_UNK	2
408 #define	KGE_LEN	3
409 #define	KGE_CHK	4
410 
411 /*
412  * Parse basic extension headers and return in the passed-in pointer vector.
413  * Return values include:
414  *
415  *	KGE_OK	Everything's nice and parsed out.
416  *		If there are no extensions, place NULL in extv[0].
417  *	KGE_DUP	There is a duplicate extension.
418  *		First instance in appropriate bin.  First duplicate in
419  *		extv[0].
420  *	KGE_UNK	Unknown extension type encountered.  extv[0] contains
421  *		unknown header.
422  *	KGE_LEN	Extension length error.
423  *	KGE_CHK	High-level reality check failed on specific extension.
424  *
425  * My apologies for some of the pointer arithmetic in here.  I'm thinking
426  * like an assembly programmer, yet trying to make the compiler happy.
427  */
428 static int
429 spdsock_get_ext(spd_ext_t *extv[], spd_msg_t *basehdr, uint_t msgsize)
430 {
431 	bzero(extv, sizeof (spd_ext_t *) * (SPD_EXT_MAX + 1));
432 
433 	/* Use extv[0] as the "current working pointer". */
434 
435 	extv[0] = (spd_ext_t *)(basehdr + 1);
436 
437 	while (extv[0] < (spd_ext_t *)(((uint8_t *)basehdr) + msgsize)) {
438 		/* Check for unknown headers. */
439 		if (extv[0]->spd_ext_type == 0 ||
440 		    extv[0]->spd_ext_type > SPD_EXT_MAX)
441 			return (KGE_UNK);
442 
443 		/*
444 		 * Check length.  Use uint64_t because extlen is in units
445 		 * of 64-bit words.  If length goes beyond the msgsize,
446 		 * return an error.  (Zero length also qualifies here.)
447 		 */
448 		if (extv[0]->spd_ext_len == 0 ||
449 		    (void *)((uint64_t *)extv[0] + extv[0]->spd_ext_len) >
450 		    (void *)((uint8_t *)basehdr + msgsize))
451 			return (KGE_LEN);
452 
453 		/* Check for redundant headers. */
454 		if (extv[extv[0]->spd_ext_type] != NULL)
455 			return (KGE_DUP);
456 
457 		/*
458 		 * Reality check the extension if possible at the spdsock
459 		 * level.
460 		 */
461 		if (!ext_check(extv[0]))
462 			return (KGE_CHK);
463 
464 		/* If I make it here, assign the appropriate bin. */
465 		extv[extv[0]->spd_ext_type] = extv[0];
466 
467 		/* Advance pointer (See above for uint64_t ptr reasoning.) */
468 		extv[0] = (spd_ext_t *)
469 		    ((uint64_t *)extv[0] + extv[0]->spd_ext_len);
470 	}
471 
472 	/* Everything's cool. */
473 
474 	/*
475 	 * If extv[0] == NULL, then there are no extension headers in this
476 	 * message.  Ensure that this is the case.
477 	 */
478 	if (extv[0] == (spd_ext_t *)(basehdr + 1))
479 		extv[0] = NULL;
480 
481 	return (KGE_OK);
482 }
483 
484 static const int bad_ext_diag[] = {
485 	SPD_DIAGNOSTIC_MALFORMED_LCLPORT,
486 	SPD_DIAGNOSTIC_MALFORMED_REMPORT,
487 	SPD_DIAGNOSTIC_MALFORMED_PROTO,
488 	SPD_DIAGNOSTIC_MALFORMED_LCLADDR,
489 	SPD_DIAGNOSTIC_MALFORMED_REMADDR,
490 	SPD_DIAGNOSTIC_MALFORMED_ACTION,
491 	SPD_DIAGNOSTIC_MALFORMED_RULE,
492 	SPD_DIAGNOSTIC_MALFORMED_RULESET,
493 	SPD_DIAGNOSTIC_MALFORMED_ICMP_TYPECODE
494 };
495 
496 static const int dup_ext_diag[] = {
497 	SPD_DIAGNOSTIC_DUPLICATE_LCLPORT,
498 	SPD_DIAGNOSTIC_DUPLICATE_REMPORT,
499 	SPD_DIAGNOSTIC_DUPLICATE_PROTO,
500 	SPD_DIAGNOSTIC_DUPLICATE_LCLADDR,
501 	SPD_DIAGNOSTIC_DUPLICATE_REMADDR,
502 	SPD_DIAGNOSTIC_DUPLICATE_ACTION,
503 	SPD_DIAGNOSTIC_DUPLICATE_RULE,
504 	SPD_DIAGNOSTIC_DUPLICATE_RULESET,
505 	SPD_DIAGNOSTIC_DUPLICATE_ICMP_TYPECODE
506 };
507 
508 /*
509  * Transmit a PF_POLICY error message to the instance either pointed to
510  * by ks, the instance with serial number serial, or more, depending.
511  *
512  * The faulty message (or a reasonable facsimile thereof) is in mp.
513  * This function will free mp or recycle it for delivery, thereby causing
514  * the stream head to free it.
515  */
516 static void
517 spdsock_error(queue_t *q, mblk_t *mp, int error, int diagnostic)
518 {
519 	spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
520 
521 	ASSERT(mp->b_datap->db_type == M_DATA);
522 
523 	if (spmsg->spd_msg_type < SPD_MIN ||
524 	    spmsg->spd_msg_type > SPD_MAX)
525 		spmsg->spd_msg_type = SPD_RESERVED;
526 
527 	/*
528 	 * Strip out extension headers.
529 	 */
530 	ASSERT(mp->b_rptr + sizeof (*spmsg) <= mp->b_datap->db_lim);
531 	mp->b_wptr = mp->b_rptr + sizeof (*spmsg);
532 	spmsg->spd_msg_len = SPD_8TO64(sizeof (spd_msg_t));
533 	spmsg->spd_msg_errno = (uint8_t)error;
534 	spmsg->spd_msg_diagnostic = (uint16_t)diagnostic;
535 
536 	qreply(q, mp);
537 }
538 
539 static void
540 spdsock_diag(queue_t *q, mblk_t *mp, int diagnostic)
541 {
542 	spdsock_error(q, mp, EINVAL, diagnostic);
543 }
544 
545 static void
546 spd_echo(queue_t *q, mblk_t *mp)
547 {
548 	qreply(q, mp);
549 }
550 
551 /*
552  * Do NOT consume a reference to itp.
553  */
554 /*ARGSUSED*/
555 static void
556 spdsock_flush_node(ipsec_tun_pol_t *itp, void *cookie, netstack_t *ns)
557 {
558 	boolean_t active = (boolean_t)cookie;
559 	ipsec_policy_head_t *iph;
560 
561 	iph = active ? itp->itp_policy : itp->itp_inactive;
562 	IPPH_REFHOLD(iph);
563 	mutex_enter(&itp->itp_lock);
564 	spdsock_flush_one(iph, ns);  /* Releases iph refhold. */
565 	if (active)
566 		itp->itp_flags &= ~ITPF_PFLAGS;
567 	else
568 		itp->itp_flags &= ~ITPF_IFLAGS;
569 	mutex_exit(&itp->itp_lock);
570 	/* SPD_FLUSH is worth a tunnel MTU check. */
571 	update_iptun_policy(itp);
572 }
573 
574 /*
575  * Clear out one polhead.
576  */
577 static void
578 spdsock_flush_one(ipsec_policy_head_t *iph, netstack_t *ns)
579 {
580 	rw_enter(&iph->iph_lock, RW_WRITER);
581 	ipsec_polhead_flush(iph, ns);
582 	rw_exit(&iph->iph_lock);
583 	IPPH_REFRELE(iph, ns);
584 }
585 
586 static void
587 spdsock_flush(queue_t *q, ipsec_policy_head_t *iph, ipsec_tun_pol_t *itp,
588     mblk_t *mp)
589 {
590 	boolean_t active;
591 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
592 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
593 
594 	if (iph != ALL_ACTIVE_POLHEADS && iph != ALL_INACTIVE_POLHEADS) {
595 		spdsock_flush_one(iph, ns);
596 		if (audit_active) {
597 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
598 			cred_t *cr;
599 			pid_t cpid;
600 
601 			cr = msg_getcred(mp, &cpid);
602 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
603 			audit_pf_policy(SPD_FLUSH, cr, ns,
604 			    ITP_NAME(itp), active, 0, cpid);
605 		}
606 	} else {
607 		active = (iph == ALL_ACTIVE_POLHEADS);
608 
609 		/* First flush the global policy. */
610 		spdsock_flush_one(active ? ipsec_system_policy(ns) :
611 		    ipsec_inactive_policy(ns), ns);
612 		if (audit_active) {
613 			cred_t *cr;
614 			pid_t cpid;
615 
616 			cr = msg_getcred(mp, &cpid);
617 			audit_pf_policy(SPD_FLUSH, cr, ns, NULL,
618 			    active, 0, cpid);
619 		}
620 		/* Then flush every tunnel's appropriate one. */
621 		itp_walk(spdsock_flush_node, (void *)active, ns);
622 		if (audit_active) {
623 			cred_t *cr;
624 			pid_t cpid;
625 
626 			cr = msg_getcred(mp, &cpid);
627 			audit_pf_policy(SPD_FLUSH, cr, ns,
628 			    "all tunnels", active, 0, cpid);
629 		}
630 	}
631 
632 	spd_echo(q, mp);
633 }
634 
635 static boolean_t
636 spdsock_ext_to_sel(spd_ext_t **extv, ipsec_selkey_t *sel, int *diag)
637 {
638 	bzero(sel, sizeof (*sel));
639 
640 	if (extv[SPD_EXT_PROTO] != NULL) {
641 		struct spd_proto *pr =
642 		    (struct spd_proto *)extv[SPD_EXT_PROTO];
643 		sel->ipsl_proto = pr->spd_proto_number;
644 		sel->ipsl_valid |= IPSL_PROTOCOL;
645 	}
646 	if (extv[SPD_EXT_LCLPORT] != NULL) {
647 		struct spd_portrange *pr =
648 		    (struct spd_portrange *)extv[SPD_EXT_LCLPORT];
649 		sel->ipsl_lport = pr->spd_ports_minport;
650 		sel->ipsl_valid |= IPSL_LOCAL_PORT;
651 	}
652 	if (extv[SPD_EXT_REMPORT] != NULL) {
653 		struct spd_portrange *pr =
654 		    (struct spd_portrange *)extv[SPD_EXT_REMPORT];
655 		sel->ipsl_rport = pr->spd_ports_minport;
656 		sel->ipsl_valid |= IPSL_REMOTE_PORT;
657 	}
658 
659 	if (extv[SPD_EXT_ICMP_TYPECODE] != NULL) {
660 		struct spd_typecode *tc=
661 		    (struct spd_typecode *)extv[SPD_EXT_ICMP_TYPECODE];
662 
663 		sel->ipsl_valid |= IPSL_ICMP_TYPE;
664 		sel->ipsl_icmp_type = tc->spd_typecode_type;
665 		if (tc->spd_typecode_type_end < tc->spd_typecode_type)
666 			sel->ipsl_icmp_type_end = tc->spd_typecode_type;
667 		else
668 			sel->ipsl_icmp_type_end = tc->spd_typecode_type_end;
669 
670 		if (tc->spd_typecode_code != 255) {
671 			sel->ipsl_valid |= IPSL_ICMP_CODE;
672 			sel->ipsl_icmp_code = tc->spd_typecode_code;
673 			if (tc->spd_typecode_code_end < tc->spd_typecode_code)
674 				sel->ipsl_icmp_code_end = tc->spd_typecode_code;
675 			else
676 				sel->ipsl_icmp_code_end =
677 				    tc->spd_typecode_code_end;
678 		}
679 	}
680 #define	ADDR2SEL(sel, extv, field, pfield, extn, bit)			      \
681 	if ((extv)[(extn)] != NULL) {					      \
682 		uint_t addrlen;						      \
683 		struct spd_address *ap = 				      \
684 			(struct spd_address *)((extv)[(extn)]); 	      \
685 		addrlen = (ap->spd_address_af == AF_INET6) ? 		      \
686 			IPV6_ADDR_LEN : IP_ADDR_LEN;			      \
687 		if (SPD_64TO8(ap->spd_address_len) < 			      \
688 			(addrlen + sizeof (*ap))) {			      \
689 			*diag = SPD_DIAGNOSTIC_BAD_ADDR_LEN;		      \
690 			return (B_FALSE);				      \
691 		}							      \
692 		bcopy((ap+1), &((sel)->field), addrlen);		      \
693 		(sel)->pfield = ap->spd_address_prefixlen;		      \
694 		(sel)->ipsl_valid |= (bit);				      \
695 		(sel)->ipsl_valid |= (ap->spd_address_af == AF_INET6) ?	      \
696 			IPSL_IPV6 : IPSL_IPV4;				      \
697 	}
698 
699 	ADDR2SEL(sel, extv, ipsl_local, ipsl_local_pfxlen,
700 	    SPD_EXT_LCLADDR, IPSL_LOCAL_ADDR);
701 	ADDR2SEL(sel, extv, ipsl_remote, ipsl_remote_pfxlen,
702 	    SPD_EXT_REMADDR, IPSL_REMOTE_ADDR);
703 
704 	if ((sel->ipsl_valid & (IPSL_IPV6|IPSL_IPV4)) ==
705 	    (IPSL_IPV6|IPSL_IPV4)) {
706 		*diag = SPD_DIAGNOSTIC_MIXED_AF;
707 		return (B_FALSE);
708 	}
709 
710 #undef ADDR2SEL
711 
712 	return (B_TRUE);
713 }
714 
715 static boolean_t
716 spd_convert_type(uint32_t type, ipsec_act_t *act)
717 {
718 	switch (type) {
719 	case SPD_ACTTYPE_DROP:
720 		act->ipa_type = IPSEC_ACT_DISCARD;
721 		return (B_TRUE);
722 
723 	case SPD_ACTTYPE_PASS:
724 		act->ipa_type = IPSEC_ACT_CLEAR;
725 		return (B_TRUE);
726 
727 	case SPD_ACTTYPE_IPSEC:
728 		act->ipa_type = IPSEC_ACT_APPLY;
729 		return (B_TRUE);
730 	}
731 	return (B_FALSE);
732 }
733 
734 static boolean_t
735 spd_convert_flags(uint32_t flags, ipsec_act_t *act)
736 {
737 	/*
738 	 * Note use of !! for boolean canonicalization.
739 	 */
740 	act->ipa_apply.ipp_use_ah = !!(flags & SPD_APPLY_AH);
741 	act->ipa_apply.ipp_use_esp = !!(flags & SPD_APPLY_ESP);
742 	act->ipa_apply.ipp_use_espa = !!(flags & SPD_APPLY_ESPA);
743 	act->ipa_apply.ipp_use_se = !!(flags & SPD_APPLY_SE);
744 	act->ipa_apply.ipp_use_unique = !!(flags & SPD_APPLY_UNIQUE);
745 	return (B_TRUE);
746 }
747 
748 static void
749 spdsock_reset_act(ipsec_act_t *act)
750 {
751 	bzero(act, sizeof (*act));
752 	act->ipa_apply.ipp_espe_maxbits = IPSEC_MAX_KEYBITS;
753 	act->ipa_apply.ipp_espa_maxbits = IPSEC_MAX_KEYBITS;
754 	act->ipa_apply.ipp_ah_maxbits = IPSEC_MAX_KEYBITS;
755 }
756 
757 /*
758  * Sanity check action against reality, and shrink-wrap key sizes..
759  */
760 static boolean_t
761 spdsock_check_action(ipsec_act_t *act, boolean_t tunnel_polhead, int *diag,
762     spd_stack_t *spds)
763 {
764 	if (tunnel_polhead && act->ipa_apply.ipp_use_unique) {
765 		*diag = SPD_DIAGNOSTIC_ADD_INCON_FLAGS;
766 		return (B_FALSE);
767 	}
768 	if ((act->ipa_type != IPSEC_ACT_APPLY) &&
769 	    (act->ipa_apply.ipp_use_ah ||
770 	    act->ipa_apply.ipp_use_esp ||
771 	    act->ipa_apply.ipp_use_espa ||
772 	    act->ipa_apply.ipp_use_se ||
773 	    act->ipa_apply.ipp_use_unique)) {
774 		*diag = SPD_DIAGNOSTIC_ADD_INCON_FLAGS;
775 		return (B_FALSE);
776 	}
777 	if ((act->ipa_type == IPSEC_ACT_APPLY) &&
778 	    !act->ipa_apply.ipp_use_ah &&
779 	    !act->ipa_apply.ipp_use_esp) {
780 		*diag = SPD_DIAGNOSTIC_ADD_INCON_FLAGS;
781 		return (B_FALSE);
782 	}
783 	return (ipsec_check_action(act, diag, spds->spds_netstack));
784 }
785 
786 /*
787  * We may be short a few error checks here..
788  */
789 static boolean_t
790 spdsock_ext_to_actvec(spd_ext_t **extv, ipsec_act_t **actpp, uint_t *nactp,
791     int *diag, spd_stack_t *spds)
792 {
793 	struct spd_ext_actions *sactp =
794 	    (struct spd_ext_actions *)extv[SPD_EXT_ACTION];
795 	ipsec_act_t act, *actp, *endactp;
796 	struct spd_attribute *attrp, *endattrp;
797 	uint64_t *endp;
798 	int nact;
799 	boolean_t tunnel_polhead;
800 
801 	tunnel_polhead = (extv[SPD_EXT_TUN_NAME] != NULL &&
802 	    (((struct spd_rule *)extv[SPD_EXT_RULE])->spd_rule_flags &
803 	    SPD_RULE_FLAG_TUNNEL));
804 
805 	*actpp = NULL;
806 	*nactp = 0;
807 
808 	if (sactp == NULL) {
809 		*diag = SPD_DIAGNOSTIC_NO_ACTION_EXT;
810 		return (B_FALSE);
811 	}
812 
813 	/*
814 	 * Parse the "action" extension and convert into an action chain.
815 	 */
816 
817 	nact = sactp->spd_actions_count;
818 
819 	endp = (uint64_t *)sactp;
820 	endp += sactp->spd_actions_len;
821 	endattrp = (struct spd_attribute *)endp;
822 
823 	actp = kmem_alloc(sizeof (*actp) * nact, KM_NOSLEEP);
824 	if (actp == NULL) {
825 		*diag = SPD_DIAGNOSTIC_ADD_NO_MEM;
826 		return (B_FALSE);
827 	}
828 	*actpp = actp;
829 	*nactp = nact;
830 	endactp = actp + nact;
831 
832 	spdsock_reset_act(&act);
833 	attrp = (struct spd_attribute *)(&sactp[1]);
834 
835 	for (; attrp < endattrp; attrp++) {
836 		switch (attrp->spd_attr_tag) {
837 		case SPD_ATTR_NOP:
838 			break;
839 
840 		case SPD_ATTR_EMPTY:
841 			spdsock_reset_act(&act);
842 			break;
843 
844 		case SPD_ATTR_END:
845 			attrp = endattrp;
846 			/* FALLTHRU */
847 		case SPD_ATTR_NEXT:
848 			if (actp >= endactp) {
849 				*diag = SPD_DIAGNOSTIC_ADD_WRONG_ACT_COUNT;
850 				goto fail;
851 			}
852 			if (!spdsock_check_action(&act, tunnel_polhead,
853 			    diag, spds))
854 				goto fail;
855 			*actp++ = act;
856 			spdsock_reset_act(&act);
857 			break;
858 
859 		case SPD_ATTR_TYPE:
860 			if (!spd_convert_type(attrp->spd_attr_value, &act)) {
861 				*diag = SPD_DIAGNOSTIC_ADD_BAD_TYPE;
862 				goto fail;
863 			}
864 			break;
865 
866 		case SPD_ATTR_FLAGS:
867 			if (!tunnel_polhead && extv[SPD_EXT_TUN_NAME] != NULL) {
868 				/*
869 				 * Set "sa unique" for transport-mode
870 				 * tunnels whether we want to or not.
871 				 */
872 				attrp->spd_attr_value |= SPD_APPLY_UNIQUE;
873 			}
874 			if (!spd_convert_flags(attrp->spd_attr_value, &act)) {
875 				*diag = SPD_DIAGNOSTIC_ADD_BAD_FLAGS;
876 				goto fail;
877 			}
878 			break;
879 
880 		case SPD_ATTR_AH_AUTH:
881 			if (attrp->spd_attr_value == 0) {
882 				*diag = SPD_DIAGNOSTIC_UNSUPP_AH_ALG;
883 				goto fail;
884 			}
885 			act.ipa_apply.ipp_auth_alg = attrp->spd_attr_value;
886 			break;
887 
888 		case SPD_ATTR_ESP_ENCR:
889 			if (attrp->spd_attr_value == 0) {
890 				*diag = SPD_DIAGNOSTIC_UNSUPP_ESP_ENCR_ALG;
891 				goto fail;
892 			}
893 			act.ipa_apply.ipp_encr_alg = attrp->spd_attr_value;
894 			break;
895 
896 		case SPD_ATTR_ESP_AUTH:
897 			if (attrp->spd_attr_value == 0) {
898 				*diag = SPD_DIAGNOSTIC_UNSUPP_ESP_AUTH_ALG;
899 				goto fail;
900 			}
901 			act.ipa_apply.ipp_esp_auth_alg = attrp->spd_attr_value;
902 			break;
903 
904 		case SPD_ATTR_ENCR_MINBITS:
905 			act.ipa_apply.ipp_espe_minbits = attrp->spd_attr_value;
906 			break;
907 
908 		case SPD_ATTR_ENCR_MAXBITS:
909 			act.ipa_apply.ipp_espe_maxbits = attrp->spd_attr_value;
910 			break;
911 
912 		case SPD_ATTR_AH_MINBITS:
913 			act.ipa_apply.ipp_ah_minbits = attrp->spd_attr_value;
914 			break;
915 
916 		case SPD_ATTR_AH_MAXBITS:
917 			act.ipa_apply.ipp_ah_maxbits = attrp->spd_attr_value;
918 			break;
919 
920 		case SPD_ATTR_ESPA_MINBITS:
921 			act.ipa_apply.ipp_espa_minbits = attrp->spd_attr_value;
922 			break;
923 
924 		case SPD_ATTR_ESPA_MAXBITS:
925 			act.ipa_apply.ipp_espa_maxbits = attrp->spd_attr_value;
926 			break;
927 
928 		case SPD_ATTR_LIFE_SOFT_TIME:
929 		case SPD_ATTR_LIFE_HARD_TIME:
930 		case SPD_ATTR_LIFE_SOFT_BYTES:
931 		case SPD_ATTR_LIFE_HARD_BYTES:
932 			break;
933 
934 		case SPD_ATTR_KM_PROTO:
935 			act.ipa_apply.ipp_km_proto = attrp->spd_attr_value;
936 			break;
937 
938 		case SPD_ATTR_KM_COOKIE:
939 			act.ipa_apply.ipp_km_cookie = attrp->spd_attr_value;
940 			break;
941 
942 		case SPD_ATTR_REPLAY_DEPTH:
943 			act.ipa_apply.ipp_replay_depth = attrp->spd_attr_value;
944 			break;
945 		}
946 	}
947 	if (actp != endactp) {
948 		*diag = SPD_DIAGNOSTIC_ADD_WRONG_ACT_COUNT;
949 		goto fail;
950 	}
951 
952 	return (B_TRUE);
953 fail:
954 	ipsec_actvec_free(*actpp, nact);
955 	*actpp = NULL;
956 	return (B_FALSE);
957 }
958 
959 typedef struct
960 {
961 	ipsec_policy_t *pol;
962 	int dir;
963 } tmprule_t;
964 
965 static int
966 mkrule(ipsec_policy_head_t *iph, struct spd_rule *rule,
967     ipsec_selkey_t *sel, ipsec_act_t *actp, int nact, uint_t dir, uint_t af,
968     tmprule_t **rp, uint64_t *index, spd_stack_t *spds)
969 {
970 	ipsec_policy_t *pol;
971 
972 	sel->ipsl_valid &= ~(IPSL_IPV6|IPSL_IPV4);
973 	sel->ipsl_valid |= af;
974 
975 	pol = ipsec_policy_create(sel, actp, nact, rule->spd_rule_priority,
976 	    index, spds->spds_netstack);
977 	if (pol == NULL)
978 		return (ENOMEM);
979 
980 	(*rp)->pol = pol;
981 	(*rp)->dir = dir;
982 	(*rp)++;
983 
984 	if (!ipsec_check_policy(iph, pol, dir))
985 		return (EEXIST);
986 
987 	rule->spd_rule_index = pol->ipsp_index;
988 	return (0);
989 }
990 
991 static int
992 mkrulepair(ipsec_policy_head_t *iph, struct spd_rule *rule,
993     ipsec_selkey_t *sel, ipsec_act_t *actp, int nact, uint_t dir, uint_t afs,
994     tmprule_t **rp, uint64_t *index, spd_stack_t *spds)
995 {
996 	int error;
997 
998 	if (afs & IPSL_IPV4) {
999 		error = mkrule(iph, rule, sel, actp, nact, dir, IPSL_IPV4, rp,
1000 		    index, spds);
1001 		if (error != 0)
1002 			return (error);
1003 	}
1004 	if (afs & IPSL_IPV6) {
1005 		error = mkrule(iph, rule, sel, actp, nact, dir, IPSL_IPV6, rp,
1006 		    index, spds);
1007 		if (error != 0)
1008 			return (error);
1009 	}
1010 	return (0);
1011 }
1012 
1013 
1014 static void
1015 spdsock_addrule(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp,
1016     spd_ext_t **extv, ipsec_tun_pol_t *itp)
1017 {
1018 	ipsec_selkey_t sel;
1019 	ipsec_act_t *actp;
1020 	uint_t nact;
1021 	int diag = 0, error, afs;
1022 	struct spd_rule *rule = (struct spd_rule *)extv[SPD_EXT_RULE];
1023 	tmprule_t rules[4], *rulep = &rules[0];
1024 	boolean_t tunnel_mode, empty_itp, active;
1025 	uint64_t *index = (itp == NULL) ? NULL : &itp->itp_next_policy_index;
1026 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
1027 	spd_stack_t	*spds = ss->spdsock_spds;
1028 
1029 	if (rule == NULL) {
1030 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_NO_RULE_EXT);
1031 		if (audit_active) {
1032 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1033 			cred_t *cr;
1034 			pid_t cpid;
1035 
1036 			cr = msg_getcred(mp, &cpid);
1037 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1038 			audit_pf_policy(SPD_ADDRULE, cr,
1039 			    spds->spds_netstack, ITP_NAME(itp), active,
1040 			    SPD_DIAGNOSTIC_NO_RULE_EXT, cpid);
1041 		}
1042 		return;
1043 	}
1044 
1045 	tunnel_mode = (rule->spd_rule_flags & SPD_RULE_FLAG_TUNNEL);
1046 
1047 	if (itp != NULL) {
1048 		mutex_enter(&itp->itp_lock);
1049 		ASSERT(itp->itp_policy == iph || itp->itp_inactive == iph);
1050 		active = (itp->itp_policy == iph);
1051 		if (ITP_P_ISACTIVE(itp, iph)) {
1052 			/* Check for mix-and-match of tunnel/transport. */
1053 			if ((tunnel_mode && !ITP_P_ISTUNNEL(itp, iph)) ||
1054 			    (!tunnel_mode && ITP_P_ISTUNNEL(itp, iph))) {
1055 				mutex_exit(&itp->itp_lock);
1056 				spdsock_error(q, mp, EBUSY, 0);
1057 				return;
1058 			}
1059 			empty_itp = B_FALSE;
1060 		} else {
1061 			empty_itp = B_TRUE;
1062 			itp->itp_flags = active ? ITPF_P_ACTIVE : ITPF_I_ACTIVE;
1063 			if (tunnel_mode)
1064 				itp->itp_flags |= active ? ITPF_P_TUNNEL :
1065 				    ITPF_I_TUNNEL;
1066 		}
1067 	} else {
1068 		empty_itp = B_FALSE;
1069 	}
1070 
1071 	if (rule->spd_rule_index != 0) {
1072 		diag = SPD_DIAGNOSTIC_INVALID_RULE_INDEX;
1073 		error = EINVAL;
1074 		goto fail2;
1075 	}
1076 
1077 	if (!spdsock_ext_to_sel(extv, &sel, &diag)) {
1078 		error = EINVAL;
1079 		goto fail2;
1080 	}
1081 
1082 	if (itp != NULL) {
1083 		if (tunnel_mode) {
1084 			if (sel.ipsl_valid &
1085 			    (IPSL_REMOTE_PORT | IPSL_LOCAL_PORT)) {
1086 				itp->itp_flags |= active ?
1087 				    ITPF_P_PER_PORT_SECURITY :
1088 				    ITPF_I_PER_PORT_SECURITY;
1089 			}
1090 		} else {
1091 			/*
1092 			 * For now, we don't allow transport-mode on a tunnel
1093 			 * with ANY specific selectors.  Bail if we have such
1094 			 * a request.
1095 			 */
1096 			if (sel.ipsl_valid & IPSL_WILDCARD) {
1097 				diag = SPD_DIAGNOSTIC_NO_TUNNEL_SELECTORS;
1098 				error = EINVAL;
1099 				goto fail2;
1100 			}
1101 		}
1102 	}
1103 
1104 	if (!spdsock_ext_to_actvec(extv, &actp, &nact, &diag, spds)) {
1105 		error = EINVAL;
1106 		goto fail2;
1107 	}
1108 	/*
1109 	 * If no addresses were specified, add both.
1110 	 */
1111 	afs = sel.ipsl_valid & (IPSL_IPV6|IPSL_IPV4);
1112 	if (afs == 0)
1113 		afs = (IPSL_IPV6|IPSL_IPV4);
1114 
1115 	rw_enter(&iph->iph_lock, RW_WRITER);
1116 
1117 	if (rule->spd_rule_flags & SPD_RULE_FLAG_OUTBOUND) {
1118 		error = mkrulepair(iph, rule, &sel, actp, nact,
1119 		    IPSEC_TYPE_OUTBOUND, afs, &rulep, index, spds);
1120 		if (error != 0)
1121 			goto fail;
1122 	}
1123 
1124 	if (rule->spd_rule_flags & SPD_RULE_FLAG_INBOUND) {
1125 		error = mkrulepair(iph, rule, &sel, actp, nact,
1126 		    IPSEC_TYPE_INBOUND, afs, &rulep, index, spds);
1127 		if (error != 0)
1128 			goto fail;
1129 	}
1130 
1131 	while ((--rulep) >= &rules[0]) {
1132 		ipsec_enter_policy(iph, rulep->pol, rulep->dir,
1133 		    spds->spds_netstack);
1134 	}
1135 	rw_exit(&iph->iph_lock);
1136 	if (itp != NULL)
1137 		mutex_exit(&itp->itp_lock);
1138 
1139 	ipsec_actvec_free(actp, nact);
1140 	spd_echo(q, mp);
1141 	if (audit_active) {
1142 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1143 		cred_t *cr;
1144 		pid_t cpid;
1145 
1146 		cr = msg_getcred(mp, &cpid);
1147 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1148 		audit_pf_policy(SPD_ADDRULE, cr, spds->spds_netstack,
1149 		    ITP_NAME(itp), active, 0, cpid);
1150 	}
1151 	return;
1152 
1153 fail:
1154 	rw_exit(&iph->iph_lock);
1155 	while ((--rulep) >= &rules[0])
1156 		IPPOL_REFRELE(rulep->pol);
1157 	ipsec_actvec_free(actp, nact);
1158 fail2:
1159 	if (itp != NULL) {
1160 		if (empty_itp)
1161 			itp->itp_flags = 0;
1162 		mutex_exit(&itp->itp_lock);
1163 	}
1164 	spdsock_error(q, mp, error, diag);
1165 	if (audit_active) {
1166 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1167 		cred_t *cr;
1168 		pid_t cpid;
1169 
1170 		cr = msg_getcred(mp, &cpid);
1171 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1172 		audit_pf_policy(SPD_ADDRULE, cr, spds->spds_netstack,
1173 		    ITP_NAME(itp), active, error, cpid);
1174 	}
1175 }
1176 
1177 void
1178 spdsock_deleterule(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp,
1179     spd_ext_t **extv, ipsec_tun_pol_t *itp)
1180 {
1181 	ipsec_selkey_t sel;
1182 	struct spd_rule *rule = (struct spd_rule *)extv[SPD_EXT_RULE];
1183 	int err, diag = 0;
1184 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
1185 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
1186 
1187 	if (rule == NULL) {
1188 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_NO_RULE_EXT);
1189 		if (audit_active) {
1190 			boolean_t active;
1191 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1192 			cred_t *cr;
1193 			pid_t cpid;
1194 
1195 			cr = msg_getcred(mp, &cpid);
1196 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1197 			audit_pf_policy(SPD_DELETERULE, cr, ns,
1198 			    ITP_NAME(itp), active, SPD_DIAGNOSTIC_NO_RULE_EXT,
1199 			    cpid);
1200 		}
1201 		return;
1202 	}
1203 
1204 	/*
1205 	 * Must enter itp_lock first to avoid deadlock.  See tun.c's
1206 	 * set_sec_simple() for the other case of itp_lock and iph_lock.
1207 	 */
1208 	if (itp != NULL)
1209 		mutex_enter(&itp->itp_lock);
1210 
1211 	if (rule->spd_rule_index != 0) {
1212 		if (ipsec_policy_delete_index(iph, rule->spd_rule_index, ns) !=
1213 		    0) {
1214 			err = ESRCH;
1215 			goto fail;
1216 		}
1217 	} else {
1218 		if (!spdsock_ext_to_sel(extv, &sel, &diag)) {
1219 			err = EINVAL;	/* diag already set... */
1220 			goto fail;
1221 		}
1222 
1223 		if ((rule->spd_rule_flags & SPD_RULE_FLAG_INBOUND) &&
1224 		    !ipsec_policy_delete(iph, &sel, IPSEC_TYPE_INBOUND, ns)) {
1225 			err = ESRCH;
1226 			goto fail;
1227 		}
1228 
1229 		if ((rule->spd_rule_flags & SPD_RULE_FLAG_OUTBOUND) &&
1230 		    !ipsec_policy_delete(iph, &sel, IPSEC_TYPE_OUTBOUND, ns)) {
1231 			err = ESRCH;
1232 			goto fail;
1233 		}
1234 	}
1235 
1236 	if (itp != NULL) {
1237 		ASSERT(iph == itp->itp_policy || iph == itp->itp_inactive);
1238 		rw_enter(&iph->iph_lock, RW_READER);
1239 		if (avl_numnodes(&iph->iph_rulebyid) == 0) {
1240 			if (iph == itp->itp_policy)
1241 				itp->itp_flags &= ~ITPF_PFLAGS;
1242 			else
1243 				itp->itp_flags &= ~ITPF_IFLAGS;
1244 		}
1245 		/* Can exit locks in any order. */
1246 		rw_exit(&iph->iph_lock);
1247 		mutex_exit(&itp->itp_lock);
1248 	}
1249 	spd_echo(q, mp);
1250 	if (audit_active) {
1251 		boolean_t active;
1252 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1253 		cred_t *cr;
1254 		pid_t cpid;
1255 
1256 		cr = msg_getcred(mp, &cpid);
1257 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1258 		audit_pf_policy(SPD_DELETERULE, cr, ns, ITP_NAME(itp),
1259 		    active, 0, cpid);
1260 	}
1261 	return;
1262 fail:
1263 	if (itp != NULL)
1264 		mutex_exit(&itp->itp_lock);
1265 	spdsock_error(q, mp, err, diag);
1266 	if (audit_active) {
1267 		boolean_t active;
1268 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1269 		cred_t *cr;
1270 		pid_t cpid;
1271 
1272 		cr = msg_getcred(mp, &cpid);
1273 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1274 		audit_pf_policy(SPD_DELETERULE, cr, ns, ITP_NAME(itp),
1275 		    active, err, cpid);
1276 	}
1277 }
1278 
1279 /* Do NOT consume a reference to itp. */
1280 /* ARGSUSED */
1281 static void
1282 spdsock_flip_node(ipsec_tun_pol_t *itp, void *ignoreme, netstack_t *ns)
1283 {
1284 	mutex_enter(&itp->itp_lock);
1285 	ITPF_SWAP(itp->itp_flags);
1286 	ipsec_swap_policy(itp->itp_policy, itp->itp_inactive, ns);
1287 	mutex_exit(&itp->itp_lock);
1288 	/* SPD_FLIP is worth a tunnel MTU check. */
1289 	update_iptun_policy(itp);
1290 }
1291 
1292 void
1293 spdsock_flip(queue_t *q, mblk_t *mp, spd_if_t *tunname)
1294 {
1295 	char *tname;
1296 	ipsec_tun_pol_t *itp;
1297 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
1298 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
1299 
1300 	if (tunname != NULL) {
1301 		tname = (char *)tunname->spd_if_name;
1302 		if (*tname == '\0') {
1303 			/* can't fail */
1304 			ipsec_swap_global_policy(ns);
1305 			if (audit_active) {
1306 				boolean_t active;
1307 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1308 				cred_t *cr;
1309 				pid_t cpid;
1310 
1311 				cr = msg_getcred(mp, &cpid);
1312 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1313 				audit_pf_policy(SPD_FLIP, cr, ns,
1314 				    NULL, active, 0, cpid);
1315 			}
1316 			itp_walk(spdsock_flip_node, NULL, ns);
1317 			if (audit_active) {
1318 				boolean_t active;
1319 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1320 				cred_t *cr;
1321 				pid_t cpid;
1322 
1323 				cr = msg_getcred(mp, &cpid);
1324 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1325 				audit_pf_policy(SPD_FLIP, cr, ns,
1326 				    "all tunnels", active, 0, cpid);
1327 			}
1328 		} else {
1329 			itp = get_tunnel_policy(tname, ns);
1330 			if (itp == NULL) {
1331 				/* Better idea for "tunnel not found"? */
1332 				spdsock_error(q, mp, ESRCH, 0);
1333 				if (audit_active) {
1334 					boolean_t active;
1335 					spd_msg_t *spmsg =
1336 					    (spd_msg_t *)mp->b_rptr;
1337 					cred_t *cr;
1338 					pid_t cpid;
1339 
1340 					cr = msg_getcred(mp, &cpid);
1341 					active = (spmsg->spd_msg_spdid ==
1342 					    SPD_ACTIVE);
1343 					audit_pf_policy(SPD_FLIP, cr, ns,
1344 					    ITP_NAME(itp), active,
1345 					    ESRCH, cpid);
1346 				}
1347 				return;
1348 			}
1349 			spdsock_flip_node(itp, NULL, ns);
1350 			if (audit_active) {
1351 				boolean_t active;
1352 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1353 				cred_t *cr;
1354 				pid_t cpid;
1355 
1356 				cr = msg_getcred(mp, &cpid);
1357 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1358 				audit_pf_policy(SPD_FLIP, cr, ns,
1359 				    ITP_NAME(itp), active, 0, cpid);
1360 			}
1361 			ITP_REFRELE(itp, ns);
1362 		}
1363 	} else {
1364 		ipsec_swap_global_policy(ns);	/* can't fail */
1365 		if (audit_active) {
1366 			boolean_t active;
1367 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
1368 			cred_t *cr;
1369 			pid_t cpid;
1370 
1371 			cr = msg_getcred(mp, &cpid);
1372 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
1373 			audit_pf_policy(SPD_FLIP, cr,
1374 			    ns, NULL, active, 0, cpid);
1375 		}
1376 	}
1377 	spd_echo(q, mp);
1378 }
1379 
1380 /*
1381  * Unimplemented feature
1382  */
1383 /* ARGSUSED */
1384 static void
1385 spdsock_lookup(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp,
1386     spd_ext_t **extv, ipsec_tun_pol_t *itp)
1387 {
1388 	spdsock_error(q, mp, EINVAL, 0);
1389 }
1390 
1391 
1392 static mblk_t *
1393 spdsock_dump_ruleset(mblk_t *req, ipsec_policy_head_t *iph,
1394     uint32_t count, uint16_t error)
1395 {
1396 	size_t len = sizeof (spd_ruleset_ext_t) + sizeof (spd_msg_t);
1397 	spd_msg_t *msg;
1398 	spd_ruleset_ext_t *ruleset;
1399 	mblk_t *m = allocb(len, BPRI_HI);
1400 
1401 	ASSERT(RW_READ_HELD(&iph->iph_lock));
1402 
1403 	if (m == NULL) {
1404 		return (NULL);
1405 	}
1406 	msg = (spd_msg_t *)m->b_rptr;
1407 	ruleset = (spd_ruleset_ext_t *)(&msg[1]);
1408 
1409 	m->b_wptr = (uint8_t *)&ruleset[1];
1410 
1411 	*msg = *(spd_msg_t *)(req->b_rptr);
1412 	msg->spd_msg_len = SPD_8TO64(len);
1413 	msg->spd_msg_errno = error;
1414 
1415 	ruleset->spd_ruleset_len = SPD_8TO64(sizeof (*ruleset));
1416 	ruleset->spd_ruleset_type = SPD_EXT_RULESET;
1417 	ruleset->spd_ruleset_count = count;
1418 	ruleset->spd_ruleset_version = iph->iph_gen;
1419 	return (m);
1420 }
1421 
1422 static mblk_t *
1423 spdsock_dump_finish(spdsock_t *ss, int error)
1424 {
1425 	mblk_t *m;
1426 	ipsec_policy_head_t *iph = ss->spdsock_dump_head;
1427 	mblk_t *req = ss->spdsock_dump_req;
1428 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
1429 
1430 	rw_enter(&iph->iph_lock, RW_READER);
1431 	m = spdsock_dump_ruleset(req, iph, ss->spdsock_dump_count, error);
1432 	rw_exit(&iph->iph_lock);
1433 	IPPH_REFRELE(iph, ns);
1434 	if (ss->spdsock_itp != NULL) {
1435 		ITP_REFRELE(ss->spdsock_itp, ns);
1436 		ss->spdsock_itp = NULL;
1437 	}
1438 	ss->spdsock_dump_req = NULL;
1439 	freemsg(req);
1440 
1441 	return (m);
1442 }
1443 
1444 /*
1445  * Rule encoding functions.
1446  * We do a two-pass encode.
1447  * If base != NULL, fill in encoded rule part starting at base+offset.
1448  * Always return "offset" plus length of to-be-encoded data.
1449  */
1450 static uint_t
1451 spdsock_encode_typecode(uint8_t *base, uint_t offset, uint8_t type,
1452     uint8_t type_end, uint8_t code, uint8_t code_end)
1453 {
1454 	struct spd_typecode *tcp;
1455 
1456 	ASSERT(ALIGNED64(offset));
1457 
1458 	if (base != NULL) {
1459 		tcp = (struct spd_typecode *)(base + offset);
1460 		tcp->spd_typecode_len = SPD_8TO64(sizeof (*tcp));
1461 		tcp->spd_typecode_exttype = SPD_EXT_ICMP_TYPECODE;
1462 		tcp->spd_typecode_code = code;
1463 		tcp->spd_typecode_type = type;
1464 		tcp->spd_typecode_type_end = type_end;
1465 		tcp->spd_typecode_code_end = code_end;
1466 	}
1467 	offset += sizeof (*tcp);
1468 
1469 	ASSERT(ALIGNED64(offset));
1470 
1471 	return (offset);
1472 }
1473 
1474 static uint_t
1475 spdsock_encode_proto(uint8_t *base, uint_t offset, uint8_t proto)
1476 {
1477 	struct spd_proto *spp;
1478 
1479 	ASSERT(ALIGNED64(offset));
1480 
1481 	if (base != NULL) {
1482 		spp = (struct spd_proto *)(base + offset);
1483 		spp->spd_proto_len = SPD_8TO64(sizeof (*spp));
1484 		spp->spd_proto_exttype = SPD_EXT_PROTO;
1485 		spp->spd_proto_number = proto;
1486 		spp->spd_proto_reserved1 = 0;
1487 		spp->spd_proto_reserved2 = 0;
1488 	}
1489 	offset += sizeof (*spp);
1490 
1491 	ASSERT(ALIGNED64(offset));
1492 
1493 	return (offset);
1494 }
1495 
1496 static uint_t
1497 spdsock_encode_port(uint8_t *base, uint_t offset, uint16_t ext, uint16_t port)
1498 {
1499 	struct spd_portrange *spp;
1500 
1501 	ASSERT(ALIGNED64(offset));
1502 
1503 	if (base != NULL) {
1504 		spp = (struct spd_portrange *)(base + offset);
1505 		spp->spd_ports_len = SPD_8TO64(sizeof (*spp));
1506 		spp->spd_ports_exttype = ext;
1507 		spp->spd_ports_minport = port;
1508 		spp->spd_ports_maxport = port;
1509 	}
1510 	offset += sizeof (*spp);
1511 
1512 	ASSERT(ALIGNED64(offset));
1513 
1514 	return (offset);
1515 }
1516 
1517 static uint_t
1518 spdsock_encode_addr(uint8_t *base, uint_t offset, uint16_t ext,
1519     const ipsec_selkey_t *sel, const ipsec_addr_t *addr, uint_t pfxlen)
1520 {
1521 	struct spd_address *sae;
1522 	ipsec_addr_t *spdaddr;
1523 	uint_t start = offset;
1524 	uint_t addrlen;
1525 	uint_t af;
1526 
1527 	if (sel->ipsl_valid & IPSL_IPV4) {
1528 		af = AF_INET;
1529 		addrlen = IP_ADDR_LEN;
1530 	} else {
1531 		af = AF_INET6;
1532 		addrlen = IPV6_ADDR_LEN;
1533 	}
1534 
1535 	ASSERT(ALIGNED64(offset));
1536 
1537 	if (base != NULL) {
1538 		sae = (struct spd_address *)(base + offset);
1539 		sae->spd_address_exttype = ext;
1540 		sae->spd_address_af = af;
1541 		sae->spd_address_prefixlen = pfxlen;
1542 		sae->spd_address_reserved2 = 0;
1543 
1544 		spdaddr = (ipsec_addr_t *)(&sae[1]);
1545 		bcopy(addr, spdaddr, addrlen);
1546 	}
1547 	offset += sizeof (*sae);
1548 	addrlen = roundup(addrlen, sizeof (uint64_t));
1549 	offset += addrlen;
1550 
1551 	ASSERT(ALIGNED64(offset));
1552 
1553 	if (base != NULL)
1554 		sae->spd_address_len = SPD_8TO64(offset - start);
1555 	return (offset);
1556 }
1557 
1558 static uint_t
1559 spdsock_encode_sel(uint8_t *base, uint_t offset, const ipsec_sel_t *sel)
1560 {
1561 	const ipsec_selkey_t *selkey = &sel->ipsl_key;
1562 
1563 	if (selkey->ipsl_valid & IPSL_PROTOCOL)
1564 		offset = spdsock_encode_proto(base, offset, selkey->ipsl_proto);
1565 	if (selkey->ipsl_valid & IPSL_LOCAL_PORT)
1566 		offset = spdsock_encode_port(base, offset, SPD_EXT_LCLPORT,
1567 		    selkey->ipsl_lport);
1568 	if (selkey->ipsl_valid & IPSL_REMOTE_PORT)
1569 		offset = spdsock_encode_port(base, offset, SPD_EXT_REMPORT,
1570 		    selkey->ipsl_rport);
1571 	if (selkey->ipsl_valid & IPSL_REMOTE_ADDR)
1572 		offset = spdsock_encode_addr(base, offset, SPD_EXT_REMADDR,
1573 		    selkey, &selkey->ipsl_remote, selkey->ipsl_remote_pfxlen);
1574 	if (selkey->ipsl_valid & IPSL_LOCAL_ADDR)
1575 		offset = spdsock_encode_addr(base, offset, SPD_EXT_LCLADDR,
1576 		    selkey, &selkey->ipsl_local, selkey->ipsl_local_pfxlen);
1577 	if (selkey->ipsl_valid & IPSL_ICMP_TYPE) {
1578 		offset = spdsock_encode_typecode(base, offset,
1579 		    selkey->ipsl_icmp_type, selkey->ipsl_icmp_type_end,
1580 		    (selkey->ipsl_valid & IPSL_ICMP_CODE) ?
1581 		    selkey->ipsl_icmp_code : 255,
1582 		    (selkey->ipsl_valid & IPSL_ICMP_CODE) ?
1583 		    selkey->ipsl_icmp_code_end : 255);
1584 	}
1585 	return (offset);
1586 }
1587 
1588 static uint_t
1589 spdsock_encode_actattr(uint8_t *base, uint_t offset, uint32_t tag,
1590     uint32_t value)
1591 {
1592 	struct spd_attribute *attr;
1593 
1594 	ASSERT(ALIGNED64(offset));
1595 
1596 	if (base != NULL) {
1597 		attr = (struct spd_attribute *)(base + offset);
1598 		attr->spd_attr_tag = tag;
1599 		attr->spd_attr_value = value;
1600 	}
1601 	offset += sizeof (struct spd_attribute);
1602 
1603 	ASSERT(ALIGNED64(offset));
1604 
1605 	return (offset);
1606 }
1607 
1608 
1609 #define	EMIT(t, v) offset = spdsock_encode_actattr(base, offset, (t), (v))
1610 
1611 static uint_t
1612 spdsock_encode_action(uint8_t *base, uint_t offset, const ipsec_action_t *ap)
1613 {
1614 	const struct ipsec_act *act = &(ap->ipa_act);
1615 	uint_t flags;
1616 
1617 	EMIT(SPD_ATTR_EMPTY, 0);
1618 	switch (act->ipa_type) {
1619 	case IPSEC_ACT_DISCARD:
1620 	case IPSEC_ACT_REJECT:
1621 		EMIT(SPD_ATTR_TYPE, SPD_ACTTYPE_DROP);
1622 		break;
1623 	case IPSEC_ACT_BYPASS:
1624 	case IPSEC_ACT_CLEAR:
1625 		EMIT(SPD_ATTR_TYPE, SPD_ACTTYPE_PASS);
1626 		break;
1627 
1628 	case IPSEC_ACT_APPLY:
1629 		EMIT(SPD_ATTR_TYPE, SPD_ACTTYPE_IPSEC);
1630 		flags = 0;
1631 		if (act->ipa_apply.ipp_use_ah)
1632 			flags |= SPD_APPLY_AH;
1633 		if (act->ipa_apply.ipp_use_esp)
1634 			flags |= SPD_APPLY_ESP;
1635 		if (act->ipa_apply.ipp_use_espa)
1636 			flags |= SPD_APPLY_ESPA;
1637 		if (act->ipa_apply.ipp_use_se)
1638 			flags |= SPD_APPLY_SE;
1639 		if (act->ipa_apply.ipp_use_unique)
1640 			flags |= SPD_APPLY_UNIQUE;
1641 		EMIT(SPD_ATTR_FLAGS, flags);
1642 		if (flags & SPD_APPLY_AH) {
1643 			EMIT(SPD_ATTR_AH_AUTH, act->ipa_apply.ipp_auth_alg);
1644 			EMIT(SPD_ATTR_AH_MINBITS,
1645 			    act->ipa_apply.ipp_ah_minbits);
1646 			EMIT(SPD_ATTR_AH_MAXBITS,
1647 			    act->ipa_apply.ipp_ah_maxbits);
1648 		}
1649 		if (flags & SPD_APPLY_ESP) {
1650 			EMIT(SPD_ATTR_ESP_ENCR, act->ipa_apply.ipp_encr_alg);
1651 			EMIT(SPD_ATTR_ENCR_MINBITS,
1652 			    act->ipa_apply.ipp_espe_minbits);
1653 			EMIT(SPD_ATTR_ENCR_MAXBITS,
1654 			    act->ipa_apply.ipp_espe_maxbits);
1655 			if (flags & SPD_APPLY_ESPA) {
1656 				EMIT(SPD_ATTR_ESP_AUTH,
1657 				    act->ipa_apply.ipp_esp_auth_alg);
1658 				EMIT(SPD_ATTR_ESPA_MINBITS,
1659 				    act->ipa_apply.ipp_espa_minbits);
1660 				EMIT(SPD_ATTR_ESPA_MAXBITS,
1661 				    act->ipa_apply.ipp_espa_maxbits);
1662 			}
1663 		}
1664 		if (act->ipa_apply.ipp_km_proto != 0)
1665 			EMIT(SPD_ATTR_KM_PROTO, act->ipa_apply.ipp_km_proto);
1666 		if (act->ipa_apply.ipp_km_cookie != 0)
1667 			EMIT(SPD_ATTR_KM_PROTO, act->ipa_apply.ipp_km_cookie);
1668 		if (act->ipa_apply.ipp_replay_depth != 0)
1669 			EMIT(SPD_ATTR_REPLAY_DEPTH,
1670 			    act->ipa_apply.ipp_replay_depth);
1671 		/* Add more here */
1672 		break;
1673 	}
1674 
1675 	return (offset);
1676 }
1677 
1678 static uint_t
1679 spdsock_encode_action_list(uint8_t *base, uint_t offset,
1680     const ipsec_action_t *ap)
1681 {
1682 	struct spd_ext_actions *act;
1683 	uint_t nact = 0;
1684 	uint_t start = offset;
1685 
1686 	ASSERT(ALIGNED64(offset));
1687 
1688 	if (base != NULL) {
1689 		act = (struct spd_ext_actions *)(base + offset);
1690 		act->spd_actions_len = 0;
1691 		act->spd_actions_exttype = SPD_EXT_ACTION;
1692 		act->spd_actions_count = 0;
1693 		act->spd_actions_reserved = 0;
1694 	}
1695 
1696 	offset += sizeof (*act);
1697 
1698 	ASSERT(ALIGNED64(offset));
1699 
1700 	while (ap != NULL) {
1701 		offset = spdsock_encode_action(base, offset, ap);
1702 		ap = ap->ipa_next;
1703 		nact++;
1704 		if (ap != NULL) {
1705 			EMIT(SPD_ATTR_NEXT, 0);
1706 		}
1707 	}
1708 	EMIT(SPD_ATTR_END, 0);
1709 
1710 	ASSERT(ALIGNED64(offset));
1711 
1712 	if (base != NULL) {
1713 		act->spd_actions_count = nact;
1714 		act->spd_actions_len = SPD_8TO64(offset - start);
1715 	}
1716 
1717 	return (offset);
1718 }
1719 
1720 #undef EMIT
1721 
1722 /* ARGSUSED */
1723 static uint_t
1724 spdsock_rule_flags(uint_t dir, uint_t af)
1725 {
1726 	uint_t flags = 0;
1727 
1728 	if (dir == IPSEC_TYPE_INBOUND)
1729 		flags |= SPD_RULE_FLAG_INBOUND;
1730 	if (dir == IPSEC_TYPE_OUTBOUND)
1731 		flags |= SPD_RULE_FLAG_OUTBOUND;
1732 
1733 	return (flags);
1734 }
1735 
1736 
1737 static uint_t
1738 spdsock_encode_rule_head(uint8_t *base, uint_t offset, spd_msg_t *req,
1739     const ipsec_policy_t *rule, uint_t dir, uint_t af, char *name,
1740     boolean_t tunnel)
1741 {
1742 	struct spd_msg *spmsg;
1743 	struct spd_rule *spr;
1744 	spd_if_t *sid;
1745 
1746 	uint_t start = offset;
1747 
1748 	ASSERT(ALIGNED64(offset));
1749 
1750 	if (base != NULL) {
1751 		spmsg = (struct spd_msg *)(base + offset);
1752 		bzero(spmsg, sizeof (*spmsg));
1753 		spmsg->spd_msg_version = PF_POLICY_V1;
1754 		spmsg->spd_msg_type = SPD_DUMP;
1755 		spmsg->spd_msg_seq = req->spd_msg_seq;
1756 		spmsg->spd_msg_pid = req->spd_msg_pid;
1757 	}
1758 	offset += sizeof (struct spd_msg);
1759 
1760 	ASSERT(ALIGNED64(offset));
1761 
1762 	if (base != NULL) {
1763 		spr = (struct spd_rule *)(base + offset);
1764 		spr->spd_rule_type = SPD_EXT_RULE;
1765 		spr->spd_rule_priority = rule->ipsp_prio;
1766 		spr->spd_rule_flags = spdsock_rule_flags(dir, af);
1767 		if (tunnel)
1768 			spr->spd_rule_flags |= SPD_RULE_FLAG_TUNNEL;
1769 		spr->spd_rule_unused = 0;
1770 		spr->spd_rule_len = SPD_8TO64(sizeof (*spr));
1771 		spr->spd_rule_index = rule->ipsp_index;
1772 	}
1773 	offset += sizeof (struct spd_rule);
1774 
1775 	/*
1776 	 * If we have an interface name (i.e. if this policy head came from
1777 	 * a tunnel), add the SPD_EXT_TUN_NAME extension.
1778 	 */
1779 	if (name != NULL) {
1780 
1781 		ASSERT(ALIGNED64(offset));
1782 
1783 		if (base != NULL) {
1784 			sid = (spd_if_t *)(base + offset);
1785 			sid->spd_if_exttype = SPD_EXT_TUN_NAME;
1786 			sid->spd_if_len = SPD_8TO64(sizeof (spd_if_t) +
1787 			    roundup((strlen(name) - 4), 8));
1788 			(void) strlcpy((char *)sid->spd_if_name, name,
1789 			    LIFNAMSIZ);
1790 		}
1791 
1792 		offset += sizeof (spd_if_t) + roundup((strlen(name) - 4), 8);
1793 	}
1794 
1795 	offset = spdsock_encode_sel(base, offset, rule->ipsp_sel);
1796 	offset = spdsock_encode_action_list(base, offset, rule->ipsp_act);
1797 
1798 	ASSERT(ALIGNED64(offset));
1799 
1800 	if (base != NULL) {
1801 		spmsg->spd_msg_len = SPD_8TO64(offset - start);
1802 	}
1803 	return (offset);
1804 }
1805 
1806 /* ARGSUSED */
1807 static mblk_t *
1808 spdsock_encode_rule(mblk_t *req, const ipsec_policy_t *rule,
1809     uint_t dir, uint_t af, char *name, boolean_t tunnel)
1810 {
1811 	mblk_t *m;
1812 	uint_t len;
1813 	spd_msg_t *mreq = (spd_msg_t *)req->b_rptr;
1814 
1815 	/*
1816 	 * Figure out how much space we'll need.
1817 	 */
1818 	len = spdsock_encode_rule_head(NULL, 0, mreq, rule, dir, af, name,
1819 	    tunnel);
1820 
1821 	/*
1822 	 * Allocate mblk.
1823 	 */
1824 	m = allocb(len, BPRI_HI);
1825 	if (m == NULL)
1826 		return (NULL);
1827 
1828 	/*
1829 	 * Fill it in..
1830 	 */
1831 	m->b_wptr = m->b_rptr + len;
1832 	bzero(m->b_rptr, len);
1833 	(void) spdsock_encode_rule_head(m->b_rptr, 0, mreq, rule, dir, af,
1834 	    name, tunnel);
1835 	return (m);
1836 }
1837 
1838 static ipsec_policy_t *
1839 spdsock_dump_next_in_chain(spdsock_t *ss, ipsec_policy_head_t *iph,
1840     ipsec_policy_t *cur)
1841 {
1842 	ASSERT(RW_READ_HELD(&iph->iph_lock));
1843 
1844 	ss->spdsock_dump_count++;
1845 	ss->spdsock_dump_cur_rule = cur->ipsp_hash.hash_next;
1846 	return (cur);
1847 }
1848 
1849 static ipsec_policy_t *
1850 spdsock_dump_next_rule(spdsock_t *ss, ipsec_policy_head_t *iph)
1851 {
1852 	ipsec_policy_t *cur;
1853 	ipsec_policy_root_t *ipr;
1854 	int chain, nchains, type, af;
1855 
1856 	ASSERT(RW_READ_HELD(&iph->iph_lock));
1857 
1858 	cur = ss->spdsock_dump_cur_rule;
1859 
1860 	if (cur != NULL)
1861 		return (spdsock_dump_next_in_chain(ss, iph, cur));
1862 
1863 	type = ss->spdsock_dump_cur_type;
1864 
1865 next:
1866 	chain = ss->spdsock_dump_cur_chain;
1867 	ipr = &iph->iph_root[type];
1868 	nchains = ipr->ipr_nchains;
1869 
1870 	while (chain < nchains) {
1871 		cur = ipr->ipr_hash[chain].hash_head;
1872 		chain++;
1873 		if (cur != NULL) {
1874 			ss->spdsock_dump_cur_chain = chain;
1875 			return (spdsock_dump_next_in_chain(ss, iph, cur));
1876 		}
1877 	}
1878 	ss->spdsock_dump_cur_chain = nchains;
1879 
1880 	af = ss->spdsock_dump_cur_af;
1881 	while (af < IPSEC_NAF) {
1882 		cur = ipr->ipr_nonhash[af];
1883 		af++;
1884 		if (cur != NULL) {
1885 			ss->spdsock_dump_cur_af = af;
1886 			return (spdsock_dump_next_in_chain(ss, iph, cur));
1887 		}
1888 	}
1889 
1890 	type++;
1891 	if (type >= IPSEC_NTYPES)
1892 		return (NULL);
1893 
1894 	ss->spdsock_dump_cur_chain = 0;
1895 	ss->spdsock_dump_cur_type = type;
1896 	ss->spdsock_dump_cur_af = IPSEC_AF_V4;
1897 	goto next;
1898 
1899 }
1900 
1901 /*
1902  * If we're done with one policy head, but have more to go, we iterate through
1903  * another IPsec tunnel policy head (itp).  Return NULL if it is an error
1904  * worthy of returning EAGAIN via PF_POLICY.
1905  */
1906 static ipsec_tun_pol_t *
1907 spdsock_dump_iterate_next_tunnel(spdsock_t *ss, ipsec_stack_t *ipss)
1908 {
1909 	ipsec_tun_pol_t *itp;
1910 
1911 	ASSERT(RW_READ_HELD(&ipss->ipsec_tunnel_policy_lock));
1912 	if (ipss->ipsec_tunnel_policy_gen > ss->spdsock_dump_tun_gen) {
1913 		/* Oops, state of the tunnel polheads changed. */
1914 		itp = NULL;
1915 	} else if (ss->spdsock_itp == NULL) {
1916 		/* Just finished global, find first node. */
1917 		itp = avl_first(&ipss->ipsec_tunnel_policies);
1918 	} else {
1919 		/* We just finished current polhead, find the next one. */
1920 		itp = AVL_NEXT(&ipss->ipsec_tunnel_policies, ss->spdsock_itp);
1921 	}
1922 	if (itp != NULL) {
1923 		ITP_REFHOLD(itp);
1924 	}
1925 	if (ss->spdsock_itp != NULL) {
1926 		ITP_REFRELE(ss->spdsock_itp, ipss->ipsec_netstack);
1927 	}
1928 	ss->spdsock_itp = itp;
1929 	return (itp);
1930 }
1931 
1932 static mblk_t *
1933 spdsock_dump_next_record(spdsock_t *ss)
1934 {
1935 	ipsec_policy_head_t *iph;
1936 	ipsec_policy_t *rule;
1937 	mblk_t *m;
1938 	ipsec_tun_pol_t *itp;
1939 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
1940 	ipsec_stack_t *ipss = ns->netstack_ipsec;
1941 
1942 	iph = ss->spdsock_dump_head;
1943 
1944 	ASSERT(iph != NULL);
1945 
1946 	rw_enter(&iph->iph_lock, RW_READER);
1947 
1948 	if (iph->iph_gen != ss->spdsock_dump_gen) {
1949 		rw_exit(&iph->iph_lock);
1950 		return (spdsock_dump_finish(ss, EAGAIN));
1951 	}
1952 
1953 	while ((rule = spdsock_dump_next_rule(ss, iph)) == NULL) {
1954 		rw_exit(&iph->iph_lock);
1955 		if (--(ss->spdsock_dump_remaining_polheads) == 0)
1956 			return (spdsock_dump_finish(ss, 0));
1957 
1958 
1959 		/*
1960 		 * If we reach here, we have more policy heads (tunnel
1961 		 * entries) to dump.  Let's reset to a new policy head
1962 		 * and get some more rules.
1963 		 *
1964 		 * An empty policy head will have spdsock_dump_next_rule()
1965 		 * return NULL, and we loop (while dropping the number of
1966 		 * remaining polheads).  If we loop to 0, we finish.  We
1967 		 * keep looping until we hit 0 or until we have a rule to
1968 		 * encode.
1969 		 *
1970 		 * NOTE:  No need for ITP_REF*() macros here as we're only
1971 		 * going after and refholding the policy head itself.
1972 		 */
1973 		rw_enter(&ipss->ipsec_tunnel_policy_lock, RW_READER);
1974 		itp = spdsock_dump_iterate_next_tunnel(ss, ipss);
1975 		if (itp == NULL) {
1976 			rw_exit(&ipss->ipsec_tunnel_policy_lock);
1977 			return (spdsock_dump_finish(ss, EAGAIN));
1978 		}
1979 
1980 		/* Reset other spdsock_dump thingies. */
1981 		IPPH_REFRELE(ss->spdsock_dump_head, ns);
1982 		if (ss->spdsock_dump_active) {
1983 			ss->spdsock_dump_tunnel =
1984 			    itp->itp_flags & ITPF_P_TUNNEL;
1985 			iph = itp->itp_policy;
1986 		} else {
1987 			ss->spdsock_dump_tunnel =
1988 			    itp->itp_flags & ITPF_I_TUNNEL;
1989 			iph = itp->itp_inactive;
1990 		}
1991 		IPPH_REFHOLD(iph);
1992 		rw_exit(&ipss->ipsec_tunnel_policy_lock);
1993 
1994 		rw_enter(&iph->iph_lock, RW_READER);
1995 		RESET_SPDSOCK_DUMP_POLHEAD(ss, iph);
1996 	}
1997 
1998 	m = spdsock_encode_rule(ss->spdsock_dump_req, rule,
1999 	    ss->spdsock_dump_cur_type, ss->spdsock_dump_cur_af,
2000 	    (ss->spdsock_itp == NULL) ? NULL : ss->spdsock_itp->itp_name,
2001 	    ss->spdsock_dump_tunnel);
2002 	rw_exit(&iph->iph_lock);
2003 
2004 	if (m == NULL)
2005 		return (spdsock_dump_finish(ss, ENOMEM));
2006 	return (m);
2007 }
2008 
2009 /*
2010  * Dump records until we run into flow-control back-pressure.
2011  */
2012 static void
2013 spdsock_dump_some(queue_t *q, spdsock_t *ss)
2014 {
2015 	mblk_t *m, *dataind;
2016 
2017 	while ((ss->spdsock_dump_req != NULL) && canputnext(q)) {
2018 		m = spdsock_dump_next_record(ss);
2019 		if (m == NULL)
2020 			return;
2021 		dataind = allocb(sizeof (struct T_data_req), BPRI_HI);
2022 		if (dataind == NULL) {
2023 			freemsg(m);
2024 			return;
2025 		}
2026 		dataind->b_cont = m;
2027 		dataind->b_wptr += sizeof (struct T_data_req);
2028 		((struct T_data_ind *)dataind->b_rptr)->PRIM_type = T_DATA_IND;
2029 		((struct T_data_ind *)dataind->b_rptr)->MORE_flag = 0;
2030 		dataind->b_datap->db_type = M_PROTO;
2031 		putnext(q, dataind);
2032 	}
2033 }
2034 
2035 /*
2036  * Start dumping.
2037  * Format a start-of-dump record, and set up the stream and kick the rsrv
2038  * procedure to continue the job..
2039  */
2040 /* ARGSUSED */
2041 static void
2042 spdsock_dump(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp)
2043 {
2044 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2045 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
2046 	ipsec_stack_t *ipss = ns->netstack_ipsec;
2047 	mblk_t *mr;
2048 
2049 	/* spdsock_open() already set spdsock_itp to NULL. */
2050 	if (iph == ALL_ACTIVE_POLHEADS || iph == ALL_INACTIVE_POLHEADS) {
2051 		rw_enter(&ipss->ipsec_tunnel_policy_lock, RW_READER);
2052 		ss->spdsock_dump_remaining_polheads = 1 +
2053 		    avl_numnodes(&ipss->ipsec_tunnel_policies);
2054 		ss->spdsock_dump_tun_gen = ipss->ipsec_tunnel_policy_gen;
2055 		rw_exit(&ipss->ipsec_tunnel_policy_lock);
2056 		if (iph == ALL_ACTIVE_POLHEADS) {
2057 			iph = ipsec_system_policy(ns);
2058 			ss->spdsock_dump_active = B_TRUE;
2059 		} else {
2060 			iph = ipsec_inactive_policy(ns);
2061 			ss->spdsock_dump_active = B_FALSE;
2062 		}
2063 		ASSERT(ss->spdsock_itp == NULL);
2064 	} else {
2065 		ss->spdsock_dump_remaining_polheads = 1;
2066 	}
2067 
2068 	rw_enter(&iph->iph_lock, RW_READER);
2069 
2070 	mr = spdsock_dump_ruleset(mp, iph, 0, 0);
2071 
2072 	if (!mr) {
2073 		rw_exit(&iph->iph_lock);
2074 		spdsock_error(q, mp, ENOMEM, 0);
2075 		return;
2076 	}
2077 
2078 	ss->spdsock_dump_req = mp;
2079 	RESET_SPDSOCK_DUMP_POLHEAD(ss, iph);
2080 
2081 	rw_exit(&iph->iph_lock);
2082 
2083 	qreply(q, mr);
2084 	qenable(OTHERQ(q));
2085 }
2086 
2087 /* Do NOT consume a reference to ITP. */
2088 void
2089 spdsock_clone_node(ipsec_tun_pol_t *itp, void *ep, netstack_t *ns)
2090 {
2091 	int *errptr = (int *)ep;
2092 
2093 	if (*errptr != 0)
2094 		return;	/* We've failed already for some reason. */
2095 	mutex_enter(&itp->itp_lock);
2096 	ITPF_CLONE(itp->itp_flags);
2097 	*errptr = ipsec_copy_polhead(itp->itp_policy, itp->itp_inactive, ns);
2098 	mutex_exit(&itp->itp_lock);
2099 }
2100 
2101 void
2102 spdsock_clone(queue_t *q, mblk_t *mp, spd_if_t *tunname)
2103 {
2104 	int error;
2105 	char *tname;
2106 	ipsec_tun_pol_t *itp;
2107 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2108 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
2109 
2110 	if (tunname != NULL) {
2111 		tname = (char *)tunname->spd_if_name;
2112 		if (*tname == '\0') {
2113 			error = ipsec_clone_system_policy(ns);
2114 			if (audit_active) {
2115 				boolean_t active;
2116 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
2117 				cred_t *cr;
2118 				pid_t cpid;
2119 
2120 				cr = msg_getcred(mp, &cpid);
2121 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
2122 				audit_pf_policy(SPD_CLONE, cr, ns,
2123 				    NULL, active, error, cpid);
2124 			}
2125 			if (error == 0) {
2126 				itp_walk(spdsock_clone_node, &error, ns);
2127 				if (audit_active) {
2128 					boolean_t active;
2129 					spd_msg_t *spmsg =
2130 					    (spd_msg_t *)mp->b_rptr;
2131 					cred_t *cr;
2132 					pid_t cpid;
2133 
2134 					cr = msg_getcred(mp, &cpid);
2135 					active = (spmsg->spd_msg_spdid ==
2136 					    SPD_ACTIVE);
2137 					audit_pf_policy(SPD_CLONE, cr,
2138 					    ns, "all tunnels", active, 0,
2139 					    cpid);
2140 				}
2141 			}
2142 		} else {
2143 			itp = get_tunnel_policy(tname, ns);
2144 			if (itp == NULL) {
2145 				spdsock_error(q, mp, ENOENT, 0);
2146 				if (audit_active) {
2147 					boolean_t active;
2148 					spd_msg_t *spmsg =
2149 					    (spd_msg_t *)mp->b_rptr;
2150 					cred_t *cr;
2151 					pid_t cpid;
2152 
2153 					cr = msg_getcred(mp, &cpid);
2154 					active = (spmsg->spd_msg_spdid ==
2155 					    SPD_ACTIVE);
2156 					audit_pf_policy(SPD_CLONE, cr,
2157 					    ns, NULL, active, ENOENT, cpid);
2158 				}
2159 				return;
2160 			}
2161 			spdsock_clone_node(itp, &error, NULL);
2162 			if (audit_active) {
2163 				boolean_t active;
2164 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
2165 				cred_t *cr;
2166 				pid_t cpid;
2167 
2168 				cr = msg_getcred(mp, &cpid);
2169 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
2170 				audit_pf_policy(SPD_CLONE, cr, ns,
2171 				    ITP_NAME(itp), active, error, cpid);
2172 			}
2173 			ITP_REFRELE(itp, ns);
2174 		}
2175 	} else {
2176 		error = ipsec_clone_system_policy(ns);
2177 		if (audit_active) {
2178 			boolean_t active;
2179 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
2180 			cred_t *cr;
2181 			pid_t cpid;
2182 
2183 			cr = msg_getcred(mp, &cpid);
2184 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
2185 			audit_pf_policy(SPD_CLONE, cr, ns, NULL,
2186 			    active, error, cpid);
2187 		}
2188 	}
2189 
2190 	if (error != 0)
2191 		spdsock_error(q, mp, error, 0);
2192 	else
2193 		spd_echo(q, mp);
2194 }
2195 
2196 /*
2197  * Process a SPD_ALGLIST request. The caller expects separate alg entries
2198  * for AH authentication, ESP authentication, and ESP encryption.
2199  * The same distinction is then used when setting the min and max key
2200  * sizes when defining policies.
2201  */
2202 
2203 #define	SPDSOCK_AH_AUTH		0
2204 #define	SPDSOCK_ESP_AUTH	1
2205 #define	SPDSOCK_ESP_ENCR	2
2206 #define	SPDSOCK_NTYPES		3
2207 
2208 static const uint_t algattr[SPDSOCK_NTYPES] = {
2209 	SPD_ATTR_AH_AUTH,
2210 	SPD_ATTR_ESP_AUTH,
2211 	SPD_ATTR_ESP_ENCR
2212 };
2213 static const uint_t minbitsattr[SPDSOCK_NTYPES] = {
2214 	SPD_ATTR_AH_MINBITS,
2215 	SPD_ATTR_ESPA_MINBITS,
2216 	SPD_ATTR_ENCR_MINBITS
2217 };
2218 static const uint_t maxbitsattr[SPDSOCK_NTYPES] = {
2219 	SPD_ATTR_AH_MAXBITS,
2220 	SPD_ATTR_ESPA_MAXBITS,
2221 	SPD_ATTR_ENCR_MAXBITS
2222 };
2223 static const uint_t defbitsattr[SPDSOCK_NTYPES] = {
2224 	SPD_ATTR_AH_DEFBITS,
2225 	SPD_ATTR_ESPA_DEFBITS,
2226 	SPD_ATTR_ENCR_DEFBITS
2227 };
2228 static const uint_t incrbitsattr[SPDSOCK_NTYPES] = {
2229 	SPD_ATTR_AH_INCRBITS,
2230 	SPD_ATTR_ESPA_INCRBITS,
2231 	SPD_ATTR_ENCR_INCRBITS
2232 };
2233 
2234 #define	ATTRPERALG	6	/* fixed attributes per algs */
2235 
2236 void
2237 spdsock_alglist(queue_t *q, mblk_t *mp)
2238 {
2239 	uint_t algtype;
2240 	uint_t algidx;
2241 	uint_t algcount;
2242 	uint_t size;
2243 	mblk_t *m;
2244 	uint8_t *cur;
2245 	spd_msg_t *msg;
2246 	struct spd_ext_actions *act;
2247 	struct spd_attribute *attr;
2248 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2249 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
2250 
2251 	mutex_enter(&ipss->ipsec_alg_lock);
2252 	/*
2253 	 * The SPD client expects to receive separate entries for
2254 	 * AH authentication and ESP authentication supported algorithms.
2255 	 *
2256 	 * Don't return the "any" algorithms, if defined, as no
2257 	 * kernel policies can be set for these algorithms.
2258 	 */
2259 	algcount = 2 * ipss->ipsec_nalgs[IPSEC_ALG_AUTH] +
2260 	    ipss->ipsec_nalgs[IPSEC_ALG_ENCR];
2261 
2262 	if (ipss->ipsec_alglists[IPSEC_ALG_AUTH][SADB_AALG_NONE] != NULL)
2263 		algcount--;
2264 	if (ipss->ipsec_alglists[IPSEC_ALG_ENCR][SADB_EALG_NONE] != NULL)
2265 		algcount--;
2266 
2267 	/*
2268 	 * For each algorithm, we encode:
2269 	 * ALG / MINBITS / MAXBITS / DEFBITS / INCRBITS / {END, NEXT}
2270 	 */
2271 
2272 	size = sizeof (spd_msg_t) + sizeof (struct spd_ext_actions) +
2273 	    ATTRPERALG * sizeof (struct spd_attribute) * algcount;
2274 
2275 	ASSERT(ALIGNED64(size));
2276 
2277 	m = allocb(size, BPRI_HI);
2278 	if (m == NULL) {
2279 		mutex_exit(&ipss->ipsec_alg_lock);
2280 		spdsock_error(q, mp, ENOMEM, 0);
2281 		return;
2282 	}
2283 
2284 	m->b_wptr = m->b_rptr + size;
2285 	cur = m->b_rptr;
2286 
2287 	msg = (spd_msg_t *)cur;
2288 	bcopy(mp->b_rptr, cur, sizeof (*msg));
2289 
2290 	msg->spd_msg_len = SPD_8TO64(size);
2291 	msg->spd_msg_errno = 0;
2292 	msg->spd_msg_diagnostic = 0;
2293 
2294 	cur += sizeof (*msg);
2295 
2296 	act = (struct spd_ext_actions *)cur;
2297 	cur += sizeof (*act);
2298 
2299 	act->spd_actions_len = SPD_8TO64(size - sizeof (spd_msg_t));
2300 	act->spd_actions_exttype = SPD_EXT_ACTION;
2301 	act->spd_actions_count = algcount;
2302 	act->spd_actions_reserved = 0;
2303 
2304 	attr = (struct spd_attribute *)cur;
2305 
2306 #define	EMIT(tag, value) {					\
2307 		attr->spd_attr_tag = (tag); 			\
2308 		attr->spd_attr_value = (value); 		\
2309 		attr++;			  			\
2310 	}
2311 
2312 	/*
2313 	 * If you change the number of EMIT's here, change
2314 	 * ATTRPERALG above to match
2315 	 */
2316 #define	EMITALGATTRS(_type) {					\
2317 		EMIT(algattr[_type], algid); 		/* 1 */	\
2318 		EMIT(minbitsattr[_type], minbits);	/* 2 */	\
2319 		EMIT(maxbitsattr[_type], maxbits);	/* 3 */	\
2320 		EMIT(defbitsattr[_type], defbits);	/* 4 */	\
2321 		EMIT(incrbitsattr[_type], incr);	/* 5 */	\
2322 		EMIT(SPD_ATTR_NEXT, 0);			/* 6 */	\
2323 	}
2324 
2325 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
2326 		for (algidx = 0; algidx < ipss->ipsec_nalgs[algtype];
2327 		    algidx++) {
2328 			int algid = ipss->ipsec_sortlist[algtype][algidx];
2329 			ipsec_alginfo_t *alg =
2330 			    ipss->ipsec_alglists[algtype][algid];
2331 			uint_t minbits = alg->alg_minbits;
2332 			uint_t maxbits = alg->alg_maxbits;
2333 			uint_t defbits = alg->alg_default_bits;
2334 			uint_t incr = alg->alg_increment;
2335 
2336 			if (algtype == IPSEC_ALG_AUTH) {
2337 				if (algid == SADB_AALG_NONE)
2338 					continue;
2339 				EMITALGATTRS(SPDSOCK_AH_AUTH);
2340 				EMITALGATTRS(SPDSOCK_ESP_AUTH);
2341 			} else {
2342 				if (algid == SADB_EALG_NONE)
2343 					continue;
2344 				ASSERT(algtype == IPSEC_ALG_ENCR);
2345 				EMITALGATTRS(SPDSOCK_ESP_ENCR);
2346 			}
2347 		}
2348 	}
2349 
2350 	mutex_exit(&ipss->ipsec_alg_lock);
2351 
2352 #undef EMITALGATTRS
2353 #undef EMIT
2354 #undef ATTRPERALG
2355 
2356 	attr--;
2357 	attr->spd_attr_tag = SPD_ATTR_END;
2358 
2359 	freemsg(mp);
2360 	qreply(q, m);
2361 }
2362 
2363 /*
2364  * Process a SPD_DUMPALGS request.
2365  */
2366 
2367 #define	ATTRPERALG	9	/* fixed attributes per algs */
2368 
2369 void
2370 spdsock_dumpalgs(queue_t *q, mblk_t *mp)
2371 {
2372 	uint_t algtype;
2373 	uint_t algidx;
2374 	uint_t size;
2375 	mblk_t *m;
2376 	uint8_t *cur;
2377 	spd_msg_t *msg;
2378 	struct spd_ext_actions *act;
2379 	struct spd_attribute *attr;
2380 	ipsec_alginfo_t *alg;
2381 	uint_t algid;
2382 	uint_t i;
2383 	uint_t alg_size;
2384 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2385 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
2386 
2387 	mutex_enter(&ipss->ipsec_alg_lock);
2388 
2389 	/*
2390 	 * For each algorithm, we encode:
2391 	 * ALG / MINBITS / MAXBITS / DEFBITS / INCRBITS / {END, NEXT}
2392 	 *
2393 	 * ALG_ID / ALG_PROTO / ALG_INCRBITS / ALG_NKEYSIZES / ALG_KEYSIZE*
2394 	 * ALG_NBLOCKSIZES / ALG_BLOCKSIZE* / ALG_NPARAMS / ALG_PARAMS* /
2395 	 * ALG_MECHNAME / ALG_FLAGS / {END, NEXT}
2396 	 */
2397 
2398 	/*
2399 	 * Compute the size of the SPD message.
2400 	 */
2401 	size = sizeof (spd_msg_t) + sizeof (struct spd_ext_actions);
2402 
2403 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
2404 		for (algidx = 0; algidx < ipss->ipsec_nalgs[algtype];
2405 		    algidx++) {
2406 			algid = ipss->ipsec_sortlist[algtype][algidx];
2407 			alg = ipss->ipsec_alglists[algtype][algid];
2408 			alg_size = sizeof (struct spd_attribute) *
2409 			    (ATTRPERALG + alg->alg_nkey_sizes +
2410 			    alg->alg_nblock_sizes + alg->alg_nparams) +
2411 			    CRYPTO_MAX_MECH_NAME;
2412 			size += alg_size;
2413 		}
2414 	}
2415 
2416 	ASSERT(ALIGNED64(size));
2417 
2418 	m = allocb(size, BPRI_HI);
2419 	if (m == NULL) {
2420 		mutex_exit(&ipss->ipsec_alg_lock);
2421 		spdsock_error(q, mp, ENOMEM, 0);
2422 		return;
2423 	}
2424 
2425 	m->b_wptr = m->b_rptr + size;
2426 	cur = m->b_rptr;
2427 
2428 	msg = (spd_msg_t *)cur;
2429 	bcopy(mp->b_rptr, cur, sizeof (*msg));
2430 
2431 	msg->spd_msg_len = SPD_8TO64(size);
2432 	msg->spd_msg_errno = 0;
2433 	msg->spd_msg_type = SPD_ALGLIST;
2434 
2435 	msg->spd_msg_diagnostic = 0;
2436 
2437 	cur += sizeof (*msg);
2438 
2439 	act = (struct spd_ext_actions *)cur;
2440 	cur += sizeof (*act);
2441 
2442 	act->spd_actions_len = SPD_8TO64(size - sizeof (spd_msg_t));
2443 	act->spd_actions_exttype = SPD_EXT_ACTION;
2444 	act->spd_actions_count = ipss->ipsec_nalgs[IPSEC_ALG_AUTH] +
2445 	    ipss->ipsec_nalgs[IPSEC_ALG_ENCR];
2446 	act->spd_actions_reserved = 0;
2447 
2448 	/*
2449 	 * If there aren't any algorithms registered, return an empty message.
2450 	 * spdsock_get_ext() knows how to deal with this.
2451 	 */
2452 	if (act->spd_actions_count == 0) {
2453 		act->spd_actions_len = 0;
2454 		mutex_exit(&ipss->ipsec_alg_lock);
2455 		goto error;
2456 	}
2457 
2458 	attr = (struct spd_attribute *)cur;
2459 
2460 #define	EMIT(tag, value) {					\
2461 		attr->spd_attr_tag = (tag); 			\
2462 		attr->spd_attr_value = (value); 		\
2463 		attr++;			  			\
2464 	}
2465 
2466 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
2467 		for (algidx = 0; algidx < ipss->ipsec_nalgs[algtype];
2468 		    algidx++) {
2469 
2470 			algid = ipss->ipsec_sortlist[algtype][algidx];
2471 			alg = ipss->ipsec_alglists[algtype][algid];
2472 
2473 			/*
2474 			 * If you change the number of EMIT's here, change
2475 			 * ATTRPERALG above to match
2476 			 */
2477 			EMIT(SPD_ATTR_ALG_ID, algid);
2478 			EMIT(SPD_ATTR_ALG_PROTO, algproto[algtype]);
2479 			EMIT(SPD_ATTR_ALG_INCRBITS, alg->alg_increment);
2480 			EMIT(SPD_ATTR_ALG_NKEYSIZES, alg->alg_nkey_sizes);
2481 			for (i = 0; i < alg->alg_nkey_sizes; i++)
2482 				EMIT(SPD_ATTR_ALG_KEYSIZE,
2483 				    alg->alg_key_sizes[i]);
2484 
2485 			EMIT(SPD_ATTR_ALG_NBLOCKSIZES, alg->alg_nblock_sizes);
2486 			for (i = 0; i < alg->alg_nblock_sizes; i++)
2487 				EMIT(SPD_ATTR_ALG_BLOCKSIZE,
2488 				    alg->alg_block_sizes[i]);
2489 
2490 			EMIT(SPD_ATTR_ALG_NPARAMS, alg->alg_nparams);
2491 			for (i = 0; i < alg->alg_nparams; i++)
2492 				EMIT(SPD_ATTR_ALG_PARAMS,
2493 				    alg->alg_params[i]);
2494 
2495 			EMIT(SPD_ATTR_ALG_FLAGS, alg->alg_flags);
2496 
2497 			EMIT(SPD_ATTR_ALG_MECHNAME, CRYPTO_MAX_MECH_NAME);
2498 			bcopy(alg->alg_mech_name, attr, CRYPTO_MAX_MECH_NAME);
2499 			attr = (struct spd_attribute *)((char *)attr +
2500 			    CRYPTO_MAX_MECH_NAME);
2501 
2502 			EMIT(SPD_ATTR_NEXT, 0);
2503 		}
2504 	}
2505 
2506 	mutex_exit(&ipss->ipsec_alg_lock);
2507 
2508 #undef EMITALGATTRS
2509 #undef EMIT
2510 #undef ATTRPERALG
2511 
2512 	attr--;
2513 	attr->spd_attr_tag = SPD_ATTR_END;
2514 
2515 error:
2516 	freemsg(mp);
2517 	qreply(q, m);
2518 }
2519 
2520 /*
2521  * Do the actual work of processing an SPD_UPDATEALGS request. Can
2522  * be invoked either once IPsec is loaded on a cached request, or
2523  * when a request is received while IPsec is loaded.
2524  */
2525 static int
2526 spdsock_do_updatealg(spd_ext_t *extv[], spd_stack_t *spds)
2527 {
2528 	struct spd_ext_actions *actp;
2529 	struct spd_attribute *attr, *endattr;
2530 	uint64_t *start, *end;
2531 	ipsec_alginfo_t *alg = NULL;
2532 	ipsec_algtype_t alg_type = 0;
2533 	boolean_t skip_alg = B_TRUE, doing_proto = B_FALSE;
2534 	uint_t i, cur_key, cur_block, algid;
2535 	int diag = -1;
2536 
2537 	ASSERT(MUTEX_HELD(&spds->spds_alg_lock));
2538 
2539 	/* parse the message, building the list of algorithms */
2540 
2541 	actp = (struct spd_ext_actions *)extv[SPD_EXT_ACTION];
2542 	if (actp == NULL)
2543 		return (SPD_DIAGNOSTIC_NO_ACTION_EXT);
2544 
2545 	start = (uint64_t *)actp;
2546 	end = (start + actp->spd_actions_len);
2547 	endattr = (struct spd_attribute *)end;
2548 	attr = (struct spd_attribute *)&actp[1];
2549 
2550 	bzero(spds->spds_algs, IPSEC_NALGTYPES * IPSEC_MAX_ALGS *
2551 	    sizeof (ipsec_alginfo_t *));
2552 
2553 	alg = kmem_zalloc(sizeof (*alg), KM_SLEEP);
2554 
2555 #define	ALG_KEY_SIZES(a)   (((a)->alg_nkey_sizes + 1) * sizeof (uint16_t))
2556 #define	ALG_BLOCK_SIZES(a) (((a)->alg_nblock_sizes + 1) * sizeof (uint16_t))
2557 #define	ALG_PARAM_SIZES(a) (((a)->alg_nparams + 1) * sizeof (uint16_t))
2558 
2559 	while (attr < endattr) {
2560 		switch (attr->spd_attr_tag) {
2561 		case SPD_ATTR_NOP:
2562 		case SPD_ATTR_EMPTY:
2563 			break;
2564 		case SPD_ATTR_END:
2565 			attr = endattr;
2566 			/* FALLTHRU */
2567 		case SPD_ATTR_NEXT:
2568 			if (doing_proto) {
2569 				doing_proto = B_FALSE;
2570 				break;
2571 			}
2572 			if (skip_alg) {
2573 				ipsec_alg_free(alg);
2574 			} else {
2575 				ipsec_alg_free(
2576 				    spds->spds_algs[alg_type][alg->alg_id]);
2577 				spds->spds_algs[alg_type][alg->alg_id] =
2578 				    alg;
2579 			}
2580 			alg = kmem_zalloc(sizeof (*alg), KM_SLEEP);
2581 			break;
2582 
2583 		case SPD_ATTR_ALG_ID:
2584 			if (attr->spd_attr_value >= IPSEC_MAX_ALGS) {
2585 				ss1dbg(spds, ("spdsock_do_updatealg: "
2586 				    "invalid alg id %d\n",
2587 				    attr->spd_attr_value));
2588 				diag = SPD_DIAGNOSTIC_ALG_ID_RANGE;
2589 				goto bail;
2590 			}
2591 			alg->alg_id = attr->spd_attr_value;
2592 			break;
2593 
2594 		case SPD_ATTR_ALG_PROTO:
2595 			/* find the alg type */
2596 			for (i = 0; i < NALGPROTOS; i++)
2597 				if (algproto[i] == attr->spd_attr_value)
2598 					break;
2599 			skip_alg = (i == NALGPROTOS);
2600 			if (!skip_alg)
2601 				alg_type = i;
2602 			break;
2603 
2604 		case SPD_ATTR_ALG_INCRBITS:
2605 			alg->alg_increment = attr->spd_attr_value;
2606 			break;
2607 
2608 		case SPD_ATTR_ALG_NKEYSIZES:
2609 			if (alg->alg_key_sizes != NULL) {
2610 				kmem_free(alg->alg_key_sizes,
2611 				    ALG_KEY_SIZES(alg));
2612 			}
2613 			alg->alg_nkey_sizes = attr->spd_attr_value;
2614 			/*
2615 			 * Allocate room for the trailing zero key size
2616 			 * value as well.
2617 			 */
2618 			alg->alg_key_sizes = kmem_zalloc(ALG_KEY_SIZES(alg),
2619 			    KM_SLEEP);
2620 			cur_key = 0;
2621 			break;
2622 
2623 		case SPD_ATTR_ALG_KEYSIZE:
2624 			if (alg->alg_key_sizes == NULL ||
2625 			    cur_key >= alg->alg_nkey_sizes) {
2626 				ss1dbg(spds, ("spdsock_do_updatealg: "
2627 				    "too many key sizes\n"));
2628 				diag = SPD_DIAGNOSTIC_ALG_NUM_KEY_SIZES;
2629 				goto bail;
2630 			}
2631 			alg->alg_key_sizes[cur_key++] = attr->spd_attr_value;
2632 			break;
2633 
2634 		case SPD_ATTR_ALG_FLAGS:
2635 			/*
2636 			 * Flags (bit mask). The alg_flags element of
2637 			 * ipsecalg_flags_t is only 8 bits wide. The
2638 			 * user can set the VALID bit, but we will ignore it
2639 			 * and make the decision is the algorithm is valid.
2640 			 */
2641 			alg->alg_flags |= (uint8_t)attr->spd_attr_value;
2642 			break;
2643 
2644 		case SPD_ATTR_ALG_NBLOCKSIZES:
2645 			if (alg->alg_block_sizes != NULL) {
2646 				kmem_free(alg->alg_block_sizes,
2647 				    ALG_BLOCK_SIZES(alg));
2648 			}
2649 			alg->alg_nblock_sizes = attr->spd_attr_value;
2650 			/*
2651 			 * Allocate room for the trailing zero block size
2652 			 * value as well.
2653 			 */
2654 			alg->alg_block_sizes = kmem_zalloc(ALG_BLOCK_SIZES(alg),
2655 			    KM_SLEEP);
2656 			cur_block = 0;
2657 			break;
2658 
2659 		case SPD_ATTR_ALG_BLOCKSIZE:
2660 			if (alg->alg_block_sizes == NULL ||
2661 			    cur_block >= alg->alg_nblock_sizes) {
2662 				ss1dbg(spds, ("spdsock_do_updatealg: "
2663 				    "too many block sizes\n"));
2664 				diag = SPD_DIAGNOSTIC_ALG_NUM_BLOCK_SIZES;
2665 				goto bail;
2666 			}
2667 			alg->alg_block_sizes[cur_block++] =
2668 			    attr->spd_attr_value;
2669 			break;
2670 
2671 		case SPD_ATTR_ALG_NPARAMS:
2672 			if (alg->alg_params != NULL) {
2673 				kmem_free(alg->alg_params,
2674 				    ALG_PARAM_SIZES(alg));
2675 			}
2676 			alg->alg_nparams = attr->spd_attr_value;
2677 			/*
2678 			 * Allocate room for the trailing zero block size
2679 			 * value as well.
2680 			 */
2681 			alg->alg_params = kmem_zalloc(ALG_PARAM_SIZES(alg),
2682 			    KM_SLEEP);
2683 			cur_block = 0;
2684 			break;
2685 
2686 		case SPD_ATTR_ALG_PARAMS:
2687 			if (alg->alg_params == NULL ||
2688 			    cur_block >= alg->alg_nparams) {
2689 				ss1dbg(spds, ("spdsock_do_updatealg: "
2690 				    "too many params\n"));
2691 				diag = SPD_DIAGNOSTIC_ALG_NUM_BLOCK_SIZES;
2692 				goto bail;
2693 			}
2694 			/*
2695 			 * Array contains: iv_len, icv_len, salt_len
2696 			 * Any additional parameters are currently ignored.
2697 			 */
2698 			alg->alg_params[cur_block++] =
2699 			    attr->spd_attr_value;
2700 			break;
2701 
2702 		case SPD_ATTR_ALG_MECHNAME: {
2703 			char *mech_name;
2704 
2705 			if (attr->spd_attr_value > CRYPTO_MAX_MECH_NAME) {
2706 				ss1dbg(spds, ("spdsock_do_updatealg: "
2707 				    "mech name too long\n"));
2708 				diag = SPD_DIAGNOSTIC_ALG_MECH_NAME_LEN;
2709 				goto bail;
2710 			}
2711 			mech_name = (char *)(attr + 1);
2712 			bcopy(mech_name, alg->alg_mech_name,
2713 			    attr->spd_attr_value);
2714 			alg->alg_mech_name[CRYPTO_MAX_MECH_NAME-1] = '\0';
2715 			attr = (struct spd_attribute *)((char *)attr +
2716 			    attr->spd_attr_value);
2717 			break;
2718 		}
2719 
2720 		case SPD_ATTR_PROTO_ID:
2721 			doing_proto = B_TRUE;
2722 			for (i = 0; i < NALGPROTOS; i++) {
2723 				if (algproto[i] == attr->spd_attr_value) {
2724 					alg_type = i;
2725 					break;
2726 				}
2727 			}
2728 			break;
2729 
2730 		case SPD_ATTR_PROTO_EXEC_MODE:
2731 			if (!doing_proto)
2732 				break;
2733 			for (i = 0; i < NEXECMODES; i++) {
2734 				if (execmodes[i] == attr->spd_attr_value) {
2735 					spds->spds_algs_exec_mode[alg_type] = i;
2736 					break;
2737 				}
2738 			}
2739 			break;
2740 		}
2741 		attr++;
2742 	}
2743 
2744 #undef	ALG_KEY_SIZES
2745 #undef	ALG_BLOCK_SIZES
2746 #undef	ALG_PARAM_SIZES
2747 
2748 	/* update the algorithm tables */
2749 	spdsock_merge_algs(spds);
2750 bail:
2751 	/* cleanup */
2752 	ipsec_alg_free(alg);
2753 	for (alg_type = 0; alg_type < IPSEC_NALGTYPES; alg_type++)
2754 		for (algid = 0; algid < IPSEC_MAX_ALGS; algid++)
2755 		if (spds->spds_algs[alg_type][algid] != NULL)
2756 			ipsec_alg_free(spds->spds_algs[alg_type][algid]);
2757 	return (diag);
2758 }
2759 
2760 /*
2761  * Process an SPD_UPDATEALGS request. If IPsec is not loaded, queue
2762  * the request until IPsec loads. If IPsec is loaded, act on it
2763  * immediately.
2764  */
2765 
2766 static void
2767 spdsock_updatealg(queue_t *q, mblk_t *mp, spd_ext_t *extv[])
2768 {
2769 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2770 	spd_stack_t	*spds = ss->spdsock_spds;
2771 	ipsec_stack_t	*ipss = spds->spds_netstack->netstack_ipsec;
2772 
2773 	if (!ipsec_loaded(ipss)) {
2774 		/*
2775 		 * IPsec is not loaded, save request and return nicely,
2776 		 * the message will be processed once IPsec loads.
2777 		 */
2778 		mblk_t *new_mp;
2779 
2780 		/* last update message wins */
2781 		if ((new_mp = copymsg(mp)) == NULL) {
2782 			spdsock_error(q, mp, ENOMEM, 0);
2783 			return;
2784 		}
2785 		mutex_enter(&spds->spds_alg_lock);
2786 		bcopy(extv, spds->spds_extv_algs,
2787 		    sizeof (spd_ext_t *) * (SPD_EXT_MAX + 1));
2788 		if (spds->spds_mp_algs != NULL)
2789 			freemsg(spds->spds_mp_algs);
2790 		spds->spds_mp_algs = mp;
2791 		spds->spds_algs_pending = B_TRUE;
2792 		mutex_exit(&spds->spds_alg_lock);
2793 		if (audit_active) {
2794 			cred_t *cr;
2795 			pid_t cpid;
2796 
2797 			cr = msg_getcred(mp, &cpid);
2798 			audit_pf_policy(SPD_UPDATEALGS, cr,
2799 			    spds->spds_netstack, NULL, B_TRUE, EAGAIN,
2800 			    cpid);
2801 		}
2802 		spd_echo(q, new_mp);
2803 	} else {
2804 		/*
2805 		 * IPsec is loaded, act on the message immediately.
2806 		 */
2807 		int diag;
2808 
2809 		mutex_enter(&spds->spds_alg_lock);
2810 		diag = spdsock_do_updatealg(extv, spds);
2811 		if (diag == -1) {
2812 			/* Keep the lock held while we walk the SA tables. */
2813 			sadb_alg_update(IPSEC_ALG_ALL, 0, 0,
2814 			    spds->spds_netstack);
2815 			mutex_exit(&spds->spds_alg_lock);
2816 			spd_echo(q, mp);
2817 			if (audit_active) {
2818 				cred_t *cr;
2819 				pid_t cpid;
2820 
2821 				cr = msg_getcred(mp, &cpid);
2822 				audit_pf_policy(SPD_UPDATEALGS, cr,
2823 				    spds->spds_netstack, NULL, B_TRUE, 0,
2824 				    cpid);
2825 			}
2826 		} else {
2827 			mutex_exit(&spds->spds_alg_lock);
2828 			spdsock_diag(q, mp, diag);
2829 			if (audit_active) {
2830 				cred_t *cr;
2831 				pid_t cpid;
2832 
2833 				cr = msg_getcred(mp, &cpid);
2834 				audit_pf_policy(SPD_UPDATEALGS, cr,
2835 				    spds->spds_netstack, NULL, B_TRUE, diag,
2836 				    cpid);
2837 			}
2838 		}
2839 	}
2840 }
2841 
2842 /*
2843  * Find a tunnel instance (using the name to link ID mapping), and
2844  * update it after an IPsec change.  We need to do this always in case
2845  * we add policy AFTER plumbing a tunnel.  We also need to do this
2846  * because, as a side-effect, the tunnel's MTU is updated to reflect
2847  * any IPsec overhead in the itp's policy.
2848  */
2849 static void
2850 update_iptun_policy(ipsec_tun_pol_t *itp)
2851 {
2852 	datalink_id_t linkid;
2853 
2854 	if (dls_mgmt_get_linkid(itp->itp_name, &linkid) == 0)
2855 		iptun_set_policy(linkid, itp);
2856 }
2857 
2858 /*
2859  * Sort through the mess of polhead options to retrieve an appropriate one.
2860  * Returns NULL if we send an spdsock error.  Returns a valid pointer if we
2861  * found a valid polhead.  Returns ALL_ACTIVE_POLHEADS (aka. -1) or
2862  * ALL_INACTIVE_POLHEADS (aka. -2) if the operation calls for the operation to
2863  * act on ALL policy heads.
2864  */
2865 static ipsec_policy_head_t *
2866 get_appropriate_polhead(queue_t *q, mblk_t *mp, spd_if_t *tunname, int spdid,
2867     int msgtype, ipsec_tun_pol_t **itpp)
2868 {
2869 	ipsec_tun_pol_t *itp;
2870 	ipsec_policy_head_t *iph;
2871 	int errno;
2872 	char *tname;
2873 	boolean_t active;
2874 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2875 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
2876 	uint64_t gen;	/* Placeholder */
2877 
2878 	active = (spdid == SPD_ACTIVE);
2879 	*itpp = NULL;
2880 	if (!active && spdid != SPD_STANDBY) {
2881 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_BAD_SPDID);
2882 		return (NULL);
2883 	}
2884 
2885 	if (tunname != NULL) {
2886 		/* Acting on a tunnel's SPD. */
2887 		tname = (char *)tunname->spd_if_name;
2888 		if (*tname == '\0') {
2889 			/* Handle all-polhead cases here. */
2890 			if (msgtype != SPD_FLUSH && msgtype != SPD_DUMP) {
2891 				spdsock_diag(q, mp,
2892 				    SPD_DIAGNOSTIC_NOT_GLOBAL_OP);
2893 				return (NULL);
2894 			}
2895 			return (active ? ALL_ACTIVE_POLHEADS :
2896 			    ALL_INACTIVE_POLHEADS);
2897 		}
2898 
2899 		itp = get_tunnel_policy(tname, ns);
2900 		if (itp == NULL) {
2901 			if (msgtype != SPD_ADDRULE) {
2902 				/* "Tunnel not found" */
2903 				spdsock_error(q, mp, ENOENT, 0);
2904 				return (NULL);
2905 			}
2906 
2907 			errno = 0;
2908 			itp = create_tunnel_policy(tname, &errno, &gen, ns);
2909 			if (itp == NULL) {
2910 				/*
2911 				 * Something very bad happened, most likely
2912 				 * ENOMEM.  Return an indicator.
2913 				 */
2914 				spdsock_error(q, mp, errno, 0);
2915 				return (NULL);
2916 			}
2917 		}
2918 
2919 		/* Match up the itp to an iptun instance. */
2920 		update_iptun_policy(itp);
2921 
2922 		*itpp = itp;
2923 		/* For spdsock dump state, set the polhead's name. */
2924 		if (msgtype == SPD_DUMP) {
2925 			ITP_REFHOLD(itp);
2926 			ss->spdsock_itp = itp;
2927 			ss->spdsock_dump_tunnel = itp->itp_flags &
2928 			    (active ? ITPF_P_TUNNEL : ITPF_I_TUNNEL);
2929 		}
2930 	} else {
2931 		itp = NULL;
2932 		/* For spdsock dump state, indicate it's global policy. */
2933 		if (msgtype == SPD_DUMP)
2934 			ss->spdsock_itp = NULL;
2935 	}
2936 
2937 	if (active)
2938 		iph = (itp == NULL) ? ipsec_system_policy(ns) : itp->itp_policy;
2939 	else
2940 		iph = (itp == NULL) ? ipsec_inactive_policy(ns) :
2941 		    itp->itp_inactive;
2942 
2943 	ASSERT(iph != NULL);
2944 	if (itp != NULL) {
2945 		IPPH_REFHOLD(iph);
2946 	}
2947 
2948 	return (iph);
2949 }
2950 
2951 static void
2952 spdsock_parse(queue_t *q, mblk_t *mp)
2953 {
2954 	spd_msg_t *spmsg;
2955 	spd_ext_t *extv[SPD_EXT_MAX + 1];
2956 	uint_t msgsize;
2957 	ipsec_policy_head_t *iph;
2958 	ipsec_tun_pol_t *itp;
2959 	spd_if_t *tunname;
2960 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
2961 	spd_stack_t *spds = ss->spdsock_spds;
2962 	netstack_t *ns = spds->spds_netstack;
2963 	ipsec_stack_t *ipss = ns->netstack_ipsec;
2964 
2965 	/* Make sure nothing's below me. */
2966 	ASSERT(WR(q)->q_next == NULL);
2967 
2968 	spmsg = (spd_msg_t *)mp->b_rptr;
2969 
2970 	msgsize = SPD_64TO8(spmsg->spd_msg_len);
2971 
2972 	if (msgdsize(mp) != msgsize) {
2973 		/*
2974 		 * Message len incorrect w.r.t. actual size.  Send an error
2975 		 * (EMSGSIZE).	It may be necessary to massage things a
2976 		 * bit.	 For example, if the spd_msg_type is hosed,
2977 		 * I need to set it to SPD_RESERVED to get delivery to
2978 		 * do the right thing.	Then again, maybe just letting
2979 		 * the error delivery do the right thing.
2980 		 */
2981 		ss2dbg(spds,
2982 		    ("mblk (%lu) and base (%d) message sizes don't jibe.\n",
2983 		    msgdsize(mp), msgsize));
2984 		spdsock_error(q, mp, EMSGSIZE, SPD_DIAGNOSTIC_NONE);
2985 		return;
2986 	}
2987 
2988 	if (msgsize > (uint_t)(mp->b_wptr - mp->b_rptr)) {
2989 		/* Get all message into one mblk. */
2990 		if (pullupmsg(mp, -1) == 0) {
2991 			/*
2992 			 * Something screwy happened.
2993 			 */
2994 			ss3dbg(spds, ("spdsock_parse: pullupmsg() failed.\n"));
2995 			return;
2996 		} else {
2997 			spmsg = (spd_msg_t *)mp->b_rptr;
2998 		}
2999 	}
3000 
3001 	switch (spdsock_get_ext(extv, spmsg, msgsize)) {
3002 	case KGE_DUP:
3003 		/* Handle duplicate extension. */
3004 		ss1dbg(spds, ("Got duplicate extension of type %d.\n",
3005 		    extv[0]->spd_ext_type));
3006 		spdsock_diag(q, mp, dup_ext_diag[extv[0]->spd_ext_type]);
3007 		return;
3008 	case KGE_UNK:
3009 		/* Handle unknown extension. */
3010 		ss1dbg(spds, ("Got unknown extension of type %d.\n",
3011 		    extv[0]->spd_ext_type));
3012 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_UNKNOWN_EXT);
3013 		return;
3014 	case KGE_LEN:
3015 		/* Length error. */
3016 		ss1dbg(spds, ("Length %d on extension type %d overrun or 0.\n",
3017 		    extv[0]->spd_ext_len, extv[0]->spd_ext_type));
3018 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_BAD_EXTLEN);
3019 		return;
3020 	case KGE_CHK:
3021 		/* Reality check failed. */
3022 		ss1dbg(spds, ("Reality check failed on extension type %d.\n",
3023 		    extv[0]->spd_ext_type));
3024 		spdsock_diag(q, mp, bad_ext_diag[extv[0]->spd_ext_type]);
3025 		return;
3026 	default:
3027 		/* Default case is no errors. */
3028 		break;
3029 	}
3030 
3031 	/*
3032 	 * Special-case SPD_UPDATEALGS so as not to load IPsec.
3033 	 */
3034 	if (!ipsec_loaded(ipss) && spmsg->spd_msg_type != SPD_UPDATEALGS) {
3035 		spdsock_t *ss = (spdsock_t *)q->q_ptr;
3036 
3037 		ASSERT(ss != NULL);
3038 		ipsec_loader_loadnow(ipss);
3039 		ss->spdsock_timeout_arg = mp;
3040 		ss->spdsock_timeout = qtimeout(q, spdsock_loadcheck,
3041 		    q, LOADCHECK_INTERVAL);
3042 		return;
3043 	}
3044 
3045 	/* First check for messages that need no polheads at all. */
3046 	switch (spmsg->spd_msg_type) {
3047 	case SPD_UPDATEALGS:
3048 		spdsock_updatealg(q, mp, extv);
3049 		return;
3050 	case SPD_ALGLIST:
3051 		spdsock_alglist(q, mp);
3052 		return;
3053 	case SPD_DUMPALGS:
3054 		spdsock_dumpalgs(q, mp);
3055 		return;
3056 	}
3057 
3058 	/*
3059 	 * Then check for ones that need both primary/secondary polheads,
3060 	 * finding the appropriate tunnel policy if need be.
3061 	 */
3062 	tunname = (spd_if_t *)extv[SPD_EXT_TUN_NAME];
3063 	switch (spmsg->spd_msg_type) {
3064 	case SPD_FLIP:
3065 		spdsock_flip(q, mp, tunname);
3066 		return;
3067 	case SPD_CLONE:
3068 		spdsock_clone(q, mp, tunname);
3069 		return;
3070 	}
3071 
3072 	/*
3073 	 * Finally, find ones that operate on exactly one polhead, or
3074 	 * "all polheads" of a given type (active/inactive).
3075 	 */
3076 	iph = get_appropriate_polhead(q, mp, tunname, spmsg->spd_msg_spdid,
3077 	    spmsg->spd_msg_type, &itp);
3078 	if (iph == NULL)
3079 		return;
3080 
3081 	/* All-polheads-ready operations. */
3082 	switch (spmsg->spd_msg_type) {
3083 	case SPD_FLUSH:
3084 		if (itp != NULL) {
3085 			mutex_enter(&itp->itp_lock);
3086 			if (spmsg->spd_msg_spdid == SPD_ACTIVE)
3087 				itp->itp_flags &= ~ITPF_PFLAGS;
3088 			else
3089 				itp->itp_flags &= ~ITPF_IFLAGS;
3090 			mutex_exit(&itp->itp_lock);
3091 		}
3092 
3093 		spdsock_flush(q, iph, itp, mp);
3094 
3095 		if (itp != NULL) {
3096 			/* SPD_FLUSH is worth a tunnel MTU check. */
3097 			update_iptun_policy(itp);
3098 			ITP_REFRELE(itp, ns);
3099 		}
3100 		return;
3101 	case SPD_DUMP:
3102 		if (itp != NULL)
3103 			ITP_REFRELE(itp, ns);
3104 		spdsock_dump(q, iph, mp);
3105 		return;
3106 	}
3107 
3108 	if (iph == ALL_ACTIVE_POLHEADS || iph == ALL_INACTIVE_POLHEADS) {
3109 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_NOT_GLOBAL_OP);
3110 		return;
3111 	}
3112 
3113 	/* Single-polhead-only operations. */
3114 	switch (spmsg->spd_msg_type) {
3115 	case SPD_ADDRULE:
3116 		spdsock_addrule(q, iph, mp, extv, itp);
3117 		break;
3118 	case SPD_DELETERULE:
3119 		spdsock_deleterule(q, iph, mp, extv, itp);
3120 		break;
3121 	case SPD_LOOKUP:
3122 		spdsock_lookup(q, iph, mp, extv, itp);
3123 		break;
3124 	default:
3125 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_BAD_MSG_TYPE);
3126 		break;
3127 	}
3128 
3129 	IPPH_REFRELE(iph, ns);
3130 	if (itp != NULL) {
3131 		/* SPD_{ADD,DELETE}RULE are worth a tunnel MTU check. */
3132 		if (spmsg->spd_msg_type == SPD_ADDRULE ||
3133 		    spmsg->spd_msg_type == SPD_DELETERULE)
3134 			update_iptun_policy(itp);
3135 		ITP_REFRELE(itp, ns);
3136 	}
3137 }
3138 
3139 /*
3140  * If an algorithm mapping was received before IPsec was loaded, process it.
3141  * Called from the IPsec loader.
3142  */
3143 void
3144 spdsock_update_pending_algs(netstack_t *ns)
3145 {
3146 	spd_stack_t *spds = ns->netstack_spdsock;
3147 
3148 	mutex_enter(&spds->spds_alg_lock);
3149 	if (spds->spds_algs_pending) {
3150 		(void) spdsock_do_updatealg(spds->spds_extv_algs, spds);
3151 		spds->spds_algs_pending = B_FALSE;
3152 	}
3153 	mutex_exit(&spds->spds_alg_lock);
3154 }
3155 
3156 static void
3157 spdsock_loadcheck(void *arg)
3158 {
3159 	queue_t *q = (queue_t *)arg;
3160 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3161 	mblk_t *mp;
3162 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
3163 
3164 	ASSERT(ss != NULL);
3165 
3166 	ss->spdsock_timeout = 0;
3167 	mp = ss->spdsock_timeout_arg;
3168 	ASSERT(mp != NULL);
3169 	ss->spdsock_timeout_arg = NULL;
3170 	if (ipsec_failed(ipss))
3171 		spdsock_error(q, mp, EPROTONOSUPPORT, 0);
3172 	else
3173 		spdsock_parse(q, mp);
3174 }
3175 
3176 /*
3177  * Copy relevant state bits.
3178  */
3179 static void
3180 spdsock_copy_info(struct T_info_ack *tap, spdsock_t *ss)
3181 {
3182 	*tap = spdsock_g_t_info_ack;
3183 	tap->CURRENT_state = ss->spdsock_state;
3184 	tap->OPT_size = spdsock_max_optsize;
3185 }
3186 
3187 /*
3188  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
3189  * spdsock_wput.  Much of the T_CAPABILITY_ACK information is copied from
3190  * spdsock_g_t_info_ack.  The current state of the stream is copied from
3191  * spdsock_state.
3192  */
3193 static void
3194 spdsock_capability_req(queue_t *q, mblk_t *mp)
3195 {
3196 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3197 	t_uscalar_t cap_bits1;
3198 	struct T_capability_ack	*tcap;
3199 
3200 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
3201 
3202 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
3203 	    mp->b_datap->db_type, T_CAPABILITY_ACK);
3204 	if (mp == NULL)
3205 		return;
3206 
3207 	tcap = (struct T_capability_ack *)mp->b_rptr;
3208 	tcap->CAP_bits1 = 0;
3209 
3210 	if (cap_bits1 & TC1_INFO) {
3211 		spdsock_copy_info(&tcap->INFO_ack, ss);
3212 		tcap->CAP_bits1 |= TC1_INFO;
3213 	}
3214 
3215 	qreply(q, mp);
3216 }
3217 
3218 /*
3219  * This routine responds to T_INFO_REQ messages. It is called by
3220  * spdsock_wput_other.
3221  * Most of the T_INFO_ACK information is copied from spdsock_g_t_info_ack.
3222  * The current state of the stream is copied from spdsock_state.
3223  */
3224 static void
3225 spdsock_info_req(q, mp)
3226 	queue_t	*q;
3227 	mblk_t	*mp;
3228 {
3229 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
3230 	    T_INFO_ACK);
3231 	if (mp == NULL)
3232 		return;
3233 	spdsock_copy_info((struct T_info_ack *)mp->b_rptr,
3234 	    (spdsock_t *)q->q_ptr);
3235 	qreply(q, mp);
3236 }
3237 
3238 /*
3239  * spdsock_err_ack. This routine creates a
3240  * T_ERROR_ACK message and passes it
3241  * upstream.
3242  */
3243 static void
3244 spdsock_err_ack(q, mp, t_error, sys_error)
3245 	queue_t	*q;
3246 	mblk_t	*mp;
3247 	int	t_error;
3248 	int	sys_error;
3249 {
3250 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
3251 		qreply(q, mp);
3252 }
3253 
3254 /*
3255  * This routine retrieves the current status of socket options.
3256  * It returns the size of the option retrieved.
3257  */
3258 /* ARGSUSED */
3259 int
3260 spdsock_opt_get(queue_t *q, int level, int name, uchar_t *ptr)
3261 {
3262 	int *i1 = (int *)ptr;
3263 
3264 	switch (level) {
3265 	case SOL_SOCKET:
3266 		switch (name) {
3267 		case SO_TYPE:
3268 			*i1 = SOCK_RAW;
3269 			break;
3270 		/*
3271 		 * The following two items can be manipulated,
3272 		 * but changing them should do nothing.
3273 		 */
3274 		case SO_SNDBUF:
3275 			*i1 = (int)q->q_hiwat;
3276 			break;
3277 		case SO_RCVBUF:
3278 			*i1 = (int)(RD(q)->q_hiwat);
3279 			break;
3280 		}
3281 		break;
3282 	default:
3283 		return (0);
3284 	}
3285 	return (sizeof (int));
3286 }
3287 
3288 /*
3289  * This routine sets socket options.
3290  */
3291 /* ARGSUSED */
3292 int
3293 spdsock_opt_set(queue_t *q, uint_t mgmt_flags, int level, int name,
3294     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
3295     void *thisdg_attrs, cred_t *cr)
3296 {
3297 	int *i1 = (int *)invalp;
3298 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3299 	spd_stack_t	*spds = ss->spdsock_spds;
3300 
3301 	switch (level) {
3302 	case SOL_SOCKET:
3303 		switch (name) {
3304 		case SO_SNDBUF:
3305 			if (*i1 > spds->spds_max_buf)
3306 				return (ENOBUFS);
3307 			q->q_hiwat = *i1;
3308 			break;
3309 		case SO_RCVBUF:
3310 			if (*i1 > spds->spds_max_buf)
3311 				return (ENOBUFS);
3312 			RD(q)->q_hiwat = *i1;
3313 			(void) proto_set_rx_hiwat(RD(q), NULL, *i1);
3314 			break;
3315 		}
3316 		break;
3317 	}
3318 	return (0);
3319 }
3320 
3321 
3322 /*
3323  * Handle STREAMS messages.
3324  */
3325 static void
3326 spdsock_wput_other(queue_t *q, mblk_t *mp)
3327 {
3328 	struct iocblk *iocp;
3329 	int error;
3330 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3331 	spd_stack_t	*spds = ss->spdsock_spds;
3332 	cred_t		*cr;
3333 
3334 	switch (mp->b_datap->db_type) {
3335 	case M_PROTO:
3336 	case M_PCPROTO:
3337 		if ((mp->b_wptr - mp->b_rptr) < sizeof (long)) {
3338 			ss3dbg(spds, (
3339 			    "spdsock_wput_other: Not big enough M_PROTO\n"));
3340 			freemsg(mp);
3341 			return;
3342 		}
3343 		switch (((union T_primitives *)mp->b_rptr)->type) {
3344 		case T_CAPABILITY_REQ:
3345 			spdsock_capability_req(q, mp);
3346 			break;
3347 		case T_INFO_REQ:
3348 			spdsock_info_req(q, mp);
3349 			break;
3350 		case T_SVR4_OPTMGMT_REQ:
3351 		case T_OPTMGMT_REQ:
3352 			/*
3353 			 * All Solaris components should pass a db_credp
3354 			 * for this TPI message, hence we ASSERT.
3355 			 * But in case there is some other M_PROTO that looks
3356 			 * like a TPI message sent by some other kernel
3357 			 * component, we check and return an error.
3358 			 */
3359 			cr = msg_getcred(mp, NULL);
3360 			ASSERT(cr != NULL);
3361 			if (cr == NULL) {
3362 				spdsock_err_ack(q, mp, TSYSERR, EINVAL);
3363 				return;
3364 			}
3365 			if (((union T_primitives *)mp->b_rptr)->type ==
3366 			    T_SVR4_OPTMGMT_REQ) {
3367 				svr4_optcom_req(q, mp, cr, &spdsock_opt_obj);
3368 			} else {
3369 				tpi_optcom_req(q, mp, cr, &spdsock_opt_obj);
3370 			}
3371 			break;
3372 		case T_DATA_REQ:
3373 		case T_EXDATA_REQ:
3374 		case T_ORDREL_REQ:
3375 			/* Illegal for spdsock. */
3376 			freemsg(mp);
3377 			(void) putnextctl1(RD(q), M_ERROR, EPROTO);
3378 			break;
3379 		default:
3380 			/* Not supported by spdsock. */
3381 			spdsock_err_ack(q, mp, TNOTSUPPORT, 0);
3382 			break;
3383 		}
3384 		return;
3385 	case M_IOCTL:
3386 		iocp = (struct iocblk *)mp->b_rptr;
3387 		error = EINVAL;
3388 
3389 		switch (iocp->ioc_cmd) {
3390 		case ND_SET:
3391 		case ND_GET:
3392 			if (nd_getset(q, spds->spds_g_nd, mp)) {
3393 				qreply(q, mp);
3394 				return;
3395 			} else
3396 				error = ENOENT;
3397 			/* FALLTHRU */
3398 		default:
3399 			miocnak(q, mp, 0, error);
3400 			return;
3401 		}
3402 	case M_FLUSH:
3403 		if (*mp->b_rptr & FLUSHW) {
3404 			flushq(q, FLUSHALL);
3405 			*mp->b_rptr &= ~FLUSHW;
3406 		}
3407 		if (*mp->b_rptr & FLUSHR) {
3408 			qreply(q, mp);
3409 			return;
3410 		}
3411 		/* Else FALLTHRU */
3412 	}
3413 
3414 	/* If fell through, just black-hole the message. */
3415 	freemsg(mp);
3416 }
3417 
3418 static void
3419 spdsock_wput(queue_t *q, mblk_t *mp)
3420 {
3421 	uint8_t *rptr = mp->b_rptr;
3422 	mblk_t *mp1;
3423 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
3424 	spd_stack_t	*spds = ss->spdsock_spds;
3425 
3426 	/*
3427 	 * If we're dumping, defer processing other messages until the
3428 	 * dump completes.
3429 	 */
3430 	if (ss->spdsock_dump_req != NULL) {
3431 		if (!putq(q, mp))
3432 			freemsg(mp);
3433 		return;
3434 	}
3435 
3436 	switch (mp->b_datap->db_type) {
3437 	case M_DATA:
3438 		/*
3439 		 * Silently discard.
3440 		 */
3441 		ss2dbg(spds, ("raw M_DATA in spdsock.\n"));
3442 		freemsg(mp);
3443 		return;
3444 	case M_PROTO:
3445 	case M_PCPROTO:
3446 		if ((mp->b_wptr - rptr) >= sizeof (struct T_data_req)) {
3447 			if (((union T_primitives *)rptr)->type == T_DATA_REQ) {
3448 				if ((mp1 = mp->b_cont) == NULL) {
3449 					/* No data after T_DATA_REQ. */
3450 					ss2dbg(spds,
3451 					    ("No data after DATA_REQ.\n"));
3452 					freemsg(mp);
3453 					return;
3454 				}
3455 				freeb(mp);
3456 				mp = mp1;
3457 				ss2dbg(spds, ("T_DATA_REQ\n"));
3458 				break;	/* Out of switch. */
3459 			}
3460 		}
3461 		/* FALLTHRU */
3462 	default:
3463 		ss3dbg(spds, ("In default wput case (%d %d).\n",
3464 		    mp->b_datap->db_type, ((union T_primitives *)rptr)->type));
3465 		spdsock_wput_other(q, mp);
3466 		return;
3467 	}
3468 
3469 	/* I now have a PF_POLICY message in an M_DATA block. */
3470 	spdsock_parse(q, mp);
3471 }
3472 
3473 /*
3474  * Device open procedure, called when new queue pair created.
3475  * We are passed the read-side queue.
3476  */
3477 /* ARGSUSED */
3478 static int
3479 spdsock_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
3480 {
3481 	spdsock_t *ss;
3482 	queue_t *oq = OTHERQ(q);
3483 	minor_t ssminor;
3484 	netstack_t *ns;
3485 	spd_stack_t *spds;
3486 
3487 	if (secpolicy_ip_config(credp, B_FALSE) != 0)
3488 		return (EPERM);
3489 
3490 	if (q->q_ptr != NULL)
3491 		return (0);  /* Re-open of an already open instance. */
3492 
3493 	if (sflag & MODOPEN)
3494 		return (EINVAL);
3495 
3496 	ns = netstack_find_by_cred(credp);
3497 	ASSERT(ns != NULL);
3498 	spds = ns->netstack_spdsock;
3499 	ASSERT(spds != NULL);
3500 
3501 	ss2dbg(spds, ("Made it into PF_POLICY socket open.\n"));
3502 
3503 	ssminor = (minor_t)(uintptr_t)vmem_alloc(spdsock_vmem, 1, VM_NOSLEEP);
3504 	if (ssminor == 0) {
3505 		netstack_rele(spds->spds_netstack);
3506 		return (ENOMEM);
3507 	}
3508 	ss = kmem_zalloc(sizeof (spdsock_t), KM_NOSLEEP);
3509 	if (ss == NULL) {
3510 		vmem_free(spdsock_vmem, (void *)(uintptr_t)ssminor, 1);
3511 		netstack_rele(spds->spds_netstack);
3512 		return (ENOMEM);
3513 	}
3514 
3515 	ss->spdsock_minor = ssminor;
3516 	ss->spdsock_state = TS_UNBND;
3517 	ss->spdsock_dump_req = NULL;
3518 
3519 	ss->spdsock_spds = spds;
3520 
3521 	q->q_ptr = ss;
3522 	oq->q_ptr = ss;
3523 
3524 	q->q_hiwat = spds->spds_recv_hiwat;
3525 
3526 	oq->q_hiwat = spds->spds_xmit_hiwat;
3527 	oq->q_lowat = spds->spds_xmit_lowat;
3528 
3529 	qprocson(q);
3530 	(void) proto_set_rx_hiwat(q, NULL, spds->spds_recv_hiwat);
3531 
3532 	*devp = makedevice(getmajor(*devp), ss->spdsock_minor);
3533 	return (0);
3534 }
3535 
3536 /*
3537  * Read-side service procedure, invoked when we get back-enabled
3538  * when buffer space becomes available.
3539  *
3540  * Dump another chunk if we were dumping before; when we finish, kick
3541  * the write-side queue in case it's waiting for read queue space.
3542  */
3543 void
3544 spdsock_rsrv(queue_t *q)
3545 {
3546 	spdsock_t *ss = q->q_ptr;
3547 
3548 	if (ss->spdsock_dump_req != NULL)
3549 		spdsock_dump_some(q, ss);
3550 
3551 	if (ss->spdsock_dump_req == NULL)
3552 		qenable(OTHERQ(q));
3553 }
3554 
3555 /*
3556  * Write-side service procedure, invoked when we defer processing
3557  * if another message is received while a dump is in progress.
3558  */
3559 void
3560 spdsock_wsrv(queue_t *q)
3561 {
3562 	spdsock_t *ss = q->q_ptr;
3563 	mblk_t *mp;
3564 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
3565 
3566 	if (ss->spdsock_dump_req != NULL) {
3567 		qenable(OTHERQ(q));
3568 		return;
3569 	}
3570 
3571 	while ((mp = getq(q)) != NULL) {
3572 		if (ipsec_loaded(ipss)) {
3573 			spdsock_wput(q, mp);
3574 			if (ss->spdsock_dump_req != NULL)
3575 				return;
3576 		} else if (!ipsec_failed(ipss)) {
3577 			(void) putq(q, mp);
3578 		} else {
3579 			spdsock_error(q, mp, EPFNOSUPPORT, 0);
3580 		}
3581 	}
3582 }
3583 
3584 static int
3585 spdsock_close(queue_t *q)
3586 {
3587 	spdsock_t *ss = q->q_ptr;
3588 	spd_stack_t	*spds = ss->spdsock_spds;
3589 
3590 	qprocsoff(q);
3591 
3592 	/* Safe assumption. */
3593 	ASSERT(ss != NULL);
3594 
3595 	if (ss->spdsock_timeout != 0)
3596 		(void) quntimeout(q, ss->spdsock_timeout);
3597 
3598 	ss3dbg(spds, ("Driver close, PF_POLICY socket is going away.\n"));
3599 
3600 	vmem_free(spdsock_vmem, (void *)(uintptr_t)ss->spdsock_minor, 1);
3601 	netstack_rele(ss->spdsock_spds->spds_netstack);
3602 
3603 	kmem_free(ss, sizeof (spdsock_t));
3604 	return (0);
3605 }
3606 
3607 /*
3608  * Merge the IPsec algorithms tables with the received algorithm information.
3609  */
3610 void
3611 spdsock_merge_algs(spd_stack_t *spds)
3612 {
3613 	ipsec_alginfo_t *alg, *oalg;
3614 	ipsec_algtype_t algtype;
3615 	uint_t algidx, algid, nalgs;
3616 	crypto_mech_name_t *mechs;
3617 	uint_t mech_count, mech_idx;
3618 	netstack_t	*ns = spds->spds_netstack;
3619 	ipsec_stack_t	*ipss = ns->netstack_ipsec;
3620 
3621 	ASSERT(MUTEX_HELD(&spds->spds_alg_lock));
3622 
3623 	/*
3624 	 * Get the list of supported mechanisms from the crypto framework.
3625 	 * If a mechanism is supported by KCF, resolve its mechanism
3626 	 * id and mark it as being valid. This operation must be done
3627 	 * without holding alg_lock, since it can cause a provider
3628 	 * module to be loaded and the provider notification callback to
3629 	 * be invoked.
3630 	 */
3631 	mechs = crypto_get_mech_list(&mech_count, KM_SLEEP);
3632 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
3633 		for (algid = 0; algid < IPSEC_MAX_ALGS; algid++) {
3634 			int algflags = 0;
3635 			crypto_mech_type_t mt = CRYPTO_MECHANISM_INVALID;
3636 
3637 			alg = spds->spds_algs[algtype][algid];
3638 			if (alg == NULL)
3639 				continue;
3640 
3641 			/*
3642 			 * The NULL encryption algorithm is a special
3643 			 * case because there are no mechanisms, yet
3644 			 * the algorithm is still valid.
3645 			 */
3646 			if (alg->alg_id == SADB_EALG_NULL) {
3647 				alg->alg_mech_type = CRYPTO_MECHANISM_INVALID;
3648 				alg->alg_flags |= ALG_FLAG_VALID;
3649 				continue;
3650 			}
3651 
3652 			for (mech_idx = 0; mech_idx < mech_count; mech_idx++) {
3653 				if (strncmp(alg->alg_mech_name, mechs[mech_idx],
3654 				    CRYPTO_MAX_MECH_NAME) == 0) {
3655 					mt = crypto_mech2id(alg->alg_mech_name);
3656 					ASSERT(mt != CRYPTO_MECHANISM_INVALID);
3657 					algflags = ALG_FLAG_VALID;
3658 					break;
3659 				}
3660 			}
3661 			alg->alg_mech_type = mt;
3662 			alg->alg_flags |= algflags;
3663 		}
3664 	}
3665 
3666 	mutex_enter(&ipss->ipsec_alg_lock);
3667 
3668 	/*
3669 	 * For each algorithm currently defined, check if it is
3670 	 * present in the new tables created from the SPD_UPDATEALGS
3671 	 * message received from user-space.
3672 	 * Delete the algorithm entries that are currently defined
3673 	 * but not part of the new tables.
3674 	 */
3675 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
3676 		nalgs = ipss->ipsec_nalgs[algtype];
3677 		for (algidx = 0; algidx < nalgs; algidx++) {
3678 			algid = ipss->ipsec_sortlist[algtype][algidx];
3679 			if (spds->spds_algs[algtype][algid] == NULL)
3680 				ipsec_alg_unreg(algtype, algid, ns);
3681 		}
3682 	}
3683 
3684 	/*
3685 	 * For each algorithm we just received, check if it is
3686 	 * present in the currently defined tables. If it is, swap
3687 	 * the entry with the one we just allocated.
3688 	 * If the new algorithm is not in the current tables,
3689 	 * add it.
3690 	 */
3691 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
3692 		for (algid = 0; algid < IPSEC_MAX_ALGS; algid++) {
3693 			alg = spds->spds_algs[algtype][algid];
3694 			if (alg == NULL)
3695 				continue;
3696 
3697 			if ((oalg = ipss->ipsec_alglists[algtype][algid]) ==
3698 			    NULL) {
3699 				/*
3700 				 * New algorithm, add it to the algorithm
3701 				 * table.
3702 				 */
3703 				ipsec_alg_reg(algtype, alg, ns);
3704 			} else {
3705 				/*
3706 				 * Algorithm is already in the table. Swap
3707 				 * the existing entry with the new one.
3708 				 */
3709 				ipsec_alg_fix_min_max(alg, algtype, ns);
3710 				ipss->ipsec_alglists[algtype][algid] = alg;
3711 				ipsec_alg_free(oalg);
3712 			}
3713 			spds->spds_algs[algtype][algid] = NULL;
3714 		}
3715 	}
3716 
3717 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
3718 		ipss->ipsec_algs_exec_mode[algtype] =
3719 		    spds->spds_algs_exec_mode[algtype];
3720 	}
3721 
3722 	mutex_exit(&ipss->ipsec_alg_lock);
3723 
3724 	crypto_free_mech_list(mechs, mech_count);
3725 
3726 	ipsecah_algs_changed(ns);
3727 	ipsecesp_algs_changed(ns);
3728 }
3729