Home | History | Annotate | Download | only in ip
      1 /*
      2  * CDDL HEADER START
      3  *
      4  * The contents of this file are subject to the terms of the
      5  * Common Development and Distribution License (the "License").
      6  * You may not use this file except in compliance with the License.
      7  *
      8  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
      9  * or http://www.opensolaris.org/os/licensing.
     10  * See the License for the specific language governing permissions
     11  * and limitations under the License.
     12  *
     13  * When distributing Covered Code, include this CDDL HEADER in each
     14  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
     15  * If applicable, add the following below this CDDL HEADER, with the
     16  * fields enclosed by brackets "[]" replaced with your own identifying
     17  * information: Portions Copyright [yyyy] [name of copyright owner]
     18  *
     19  * CDDL HEADER END
     20  */
     21 /*
     22  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
     23  * Use is subject to license terms.
     24  */
     25 
     26 #include <sys/param.h>
     27 #include <sys/types.h>
     28 #include <sys/stream.h>
     29 #include <sys/strsubr.h>
     30 #include <sys/strsun.h>
     31 #include <sys/stropts.h>
     32 #include <sys/zone.h>
     33 #include <sys/vnode.h>
     34 #include <sys/sysmacros.h>
     35 #define	_SUN_TPI_VERSION 2
     36 #include <sys/tihdr.h>
     37 #include <sys/ddi.h>
     38 #include <sys/sunddi.h>
     39 #include <sys/mkdev.h>
     40 #include <sys/debug.h>
     41 #include <sys/kmem.h>
     42 #include <sys/cmn_err.h>
     43 #include <sys/suntpi.h>
     44 #include <sys/policy.h>
     45 #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 
    165 static struct module_info info = {
    166 	5138, "spdsock", 1, INFPSZ, 512, 128
    167 };
    168 
    169 static struct qinit rinit = {
    170 	NULL, (pfi_t)spdsock_rsrv, spdsock_open, spdsock_close,
    171 	NULL, &info
    172 };
    173 
    174 static struct qinit winit = {
    175 	(pfi_t)spdsock_wput, (pfi_t)spdsock_wsrv, NULL, NULL, NULL, &info
    176 };
    177 
    178 struct streamtab spdsockinfo = {
    179 	&rinit, &winit
    180 };
    181 
    182 /* mapping from alg type to protocol number, as per RFC 2407 */
    183 static const uint_t algproto[] = {
    184 	PROTO_IPSEC_AH,
    185 	PROTO_IPSEC_ESP,
    186 };
    187 
    188 #define	NALGPROTOS	(sizeof (algproto) / sizeof (algproto[0]))
    189 
    190 /* mapping from kernel exec mode to spdsock exec mode */
    191 static const uint_t execmodes[] = {
    192 	SPD_ALG_EXEC_MODE_SYNC,
    193 	SPD_ALG_EXEC_MODE_ASYNC
    194 };
    195 
    196 #define	NEXECMODES	(sizeof (execmodes) / sizeof (execmodes[0]))
    197 
    198 #define	ALL_ACTIVE_POLHEADS ((ipsec_policy_head_t *)-1)
    199 #define	ALL_INACTIVE_POLHEADS ((ipsec_policy_head_t *)-2)
    200 
    201 #define	ITP_NAME(itp) (itp != NULL ? itp->itp_name : NULL)
    202 
    203 /* ARGSUSED */
    204 static int
    205 spdsock_param_get(q, mp, cp, cr)
    206 	queue_t	*q;
    207 	mblk_t	*mp;
    208 	caddr_t	cp;
    209 	cred_t *cr;
    210 {
    211 	spdsockparam_t	*spdsockpa = (spdsockparam_t *)cp;
    212 	uint_t value;
    213 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
    214 	spd_stack_t	*spds = ss->spdsock_spds;
    215 
    216 	mutex_enter(&spds->spds_param_lock);
    217 	value = spdsockpa->spdsock_param_value;
    218 	mutex_exit(&spds->spds_param_lock);
    219 
    220 	(void) mi_mpprintf(mp, "%u", value);
    221 	return (0);
    222 }
    223 
    224 /* This routine sets an NDD variable in a spdsockparam_t structure. */
    225 /* ARGSUSED */
    226 static int
    227 spdsock_param_set(q, mp, value, cp, cr)
    228 	queue_t	*q;
    229 	mblk_t	*mp;
    230 	char *value;
    231 	caddr_t	cp;
    232 	cred_t *cr;
    233 {
    234 	ulong_t	new_value;
    235 	spdsockparam_t	*spdsockpa = (spdsockparam_t *)cp;
    236 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
    237 	spd_stack_t	*spds = ss->spdsock_spds;
    238 
    239 	/* Convert the value from a string into a long integer. */
    240 	if (ddi_strtoul(value, NULL, 10, &new_value) != 0)
    241 		return (EINVAL);
    242 
    243 	mutex_enter(&spds->spds_param_lock);
    244 	/*
    245 	 * Fail the request if the new value does not lie within the
    246 	 * required bounds.
    247 	 */
    248 	if (new_value < spdsockpa->spdsock_param_min ||
    249 	    new_value > spdsockpa->spdsock_param_max) {
    250 		mutex_exit(&spds->spds_param_lock);
    251 		return (EINVAL);
    252 	}
    253 
    254 	/* Set the new value */
    255 	spdsockpa->spdsock_param_value = new_value;
    256 	mutex_exit(&spds->spds_param_lock);
    257 
    258 	return (0);
    259 }
    260 
    261 /*
    262  * Initialize at module load time
    263  */
    264 boolean_t
    265 spdsock_ddi_init(void)
    266 {
    267 	spdsock_max_optsize = optcom_max_optsize(
    268 	    spdsock_opt_obj.odb_opt_des_arr, spdsock_opt_obj.odb_opt_arr_cnt);
    269 
    270 	spdsock_vmem = vmem_create("spdsock", (void *)1, MAXMIN, 1,
    271 	    NULL, NULL, NULL, 1, VM_SLEEP | VMC_IDENTIFIER);
    272 
    273 	/*
    274 	 * We want to be informed each time a stack is created or
    275 	 * destroyed in the kernel, so we can maintain the
    276 	 * set of spd_stack_t's.
    277 	 */
    278 	netstack_register(NS_SPDSOCK, spdsock_stack_init, NULL,
    279 	    spdsock_stack_fini);
    280 
    281 	return (B_TRUE);
    282 }
    283 
    284 /*
    285  * Walk through the param array specified registering each element with the
    286  * named dispatch handler.
    287  */
    288 static boolean_t
    289 spdsock_param_register(IDP *ndp, spdsockparam_t *ssp, int cnt)
    290 {
    291 	for (; cnt-- > 0; ssp++) {
    292 		if (ssp->spdsock_param_name != NULL &&
    293 		    ssp->spdsock_param_name[0]) {
    294 			if (!nd_load(ndp,
    295 			    ssp->spdsock_param_name,
    296 			    spdsock_param_get, spdsock_param_set,
    297 			    (caddr_t)ssp)) {
    298 				nd_free(ndp);
    299 				return (B_FALSE);
    300 			}
    301 		}
    302 	}
    303 	return (B_TRUE);
    304 }
    305 
    306 /*
    307  * Initialize for each stack instance
    308  */
    309 /* ARGSUSED */
    310 static void *
    311 spdsock_stack_init(netstackid_t stackid, netstack_t *ns)
    312 {
    313 	spd_stack_t	*spds;
    314 	spdsockparam_t	*ssp;
    315 
    316 	spds = (spd_stack_t *)kmem_zalloc(sizeof (*spds), KM_SLEEP);
    317 	spds->spds_netstack = ns;
    318 
    319 	ASSERT(spds->spds_g_nd == NULL);
    320 
    321 	ssp = (spdsockparam_t *)kmem_alloc(sizeof (lcl_param_arr), KM_SLEEP);
    322 	spds->spds_params = ssp;
    323 	bcopy(lcl_param_arr, ssp, sizeof (lcl_param_arr));
    324 
    325 	(void) spdsock_param_register(&spds->spds_g_nd, ssp,
    326 	    A_CNT(lcl_param_arr));
    327 
    328 	mutex_init(&spds->spds_param_lock, NULL, MUTEX_DEFAULT, NULL);
    329 	mutex_init(&spds->spds_alg_lock, NULL, MUTEX_DEFAULT, NULL);
    330 
    331 	return (spds);
    332 }
    333 
    334 void
    335 spdsock_ddi_destroy(void)
    336 {
    337 	vmem_destroy(spdsock_vmem);
    338 
    339 	netstack_unregister(NS_SPDSOCK);
    340 }
    341 
    342 /* ARGSUSED */
    343 static void
    344 spdsock_stack_fini(netstackid_t stackid, void *arg)
    345 {
    346 	spd_stack_t *spds = (spd_stack_t *)arg;
    347 
    348 	freemsg(spds->spds_mp_algs);
    349 	mutex_destroy(&spds->spds_param_lock);
    350 	mutex_destroy(&spds->spds_alg_lock);
    351 	nd_free(&spds->spds_g_nd);
    352 	kmem_free(spds->spds_params, sizeof (lcl_param_arr));
    353 	spds->spds_params = NULL;
    354 
    355 	kmem_free(spds, sizeof (*spds));
    356 }
    357 
    358 /*
    359  * NOTE: large quantities of this should be shared with keysock.
    360  * Would be nice to combine some of this into a common module, but
    361  * not possible given time pressures.
    362  */
    363 
    364 /*
    365  * High-level reality checking of extensions.
    366  */
    367 /* ARGSUSED */ /* XXX */
    368 static boolean_t
    369 ext_check(spd_ext_t *ext)
    370 {
    371 	spd_if_t *tunname = (spd_if_t *)ext;
    372 	int i;
    373 	char *idstr;
    374 
    375 	if (ext->spd_ext_type == SPD_EXT_TUN_NAME) {
    376 		/* (NOTE:  Modified from SADB_EXT_IDENTITY..) */
    377 
    378 		/*
    379 		 * Make sure the strings in these identities are
    380 		 * null-terminated.  Let's "proactively" null-terminate the
    381 		 * string at the last byte if it's not terminated sooner.
    382 		 */
    383 		i = SPD_64TO8(tunname->spd_if_len) - sizeof (spd_if_t);
    384 		idstr = (char *)(tunname + 1);
    385 		while (*idstr != '\0' && i > 0) {
    386 			i--;
    387 			idstr++;
    388 		}
    389 		if (i == 0) {
    390 			/*
    391 			 * I.e., if the bozo user didn't NULL-terminate the
    392 			 * string...
    393 			 */
    394 			idstr--;
    395 			*idstr = '\0';
    396 		}
    397 	}
    398 	return (B_TRUE);	/* For now... */
    399 }
    400 
    401 
    402 
    403 /* Return values for spdsock_get_ext(). */
    404 #define	KGE_OK	0
    405 #define	KGE_DUP	1
    406 #define	KGE_UNK	2
    407 #define	KGE_LEN	3
    408 #define	KGE_CHK	4
    409 
    410 /*
    411  * Parse basic extension headers and return in the passed-in pointer vector.
    412  * Return values include:
    413  *
    414  *	KGE_OK	Everything's nice and parsed out.
    415  *		If there are no extensions, place NULL in extv[0].
    416  *	KGE_DUP	There is a duplicate extension.
    417  *		First instance in appropriate bin.  First duplicate in
    418  *		extv[0].
    419  *	KGE_UNK	Unknown extension type encountered.  extv[0] contains
    420  *		unknown header.
    421  *	KGE_LEN	Extension length error.
    422  *	KGE_CHK	High-level reality check failed on specific extension.
    423  *
    424  * My apologies for some of the pointer arithmetic in here.  I'm thinking
    425  * like an assembly programmer, yet trying to make the compiler happy.
    426  */
    427 static int
    428 spdsock_get_ext(spd_ext_t *extv[], spd_msg_t *basehdr, uint_t msgsize)
    429 {
    430 	bzero(extv, sizeof (spd_ext_t *) * (SPD_EXT_MAX + 1));
    431 
    432 	/* Use extv[0] as the "current working pointer". */
    433 
    434 	extv[0] = (spd_ext_t *)(basehdr + 1);
    435 
    436 	while (extv[0] < (spd_ext_t *)(((uint8_t *)basehdr) + msgsize)) {
    437 		/* Check for unknown headers. */
    438 		if (extv[0]->spd_ext_type == 0 ||
    439 		    extv[0]->spd_ext_type > SPD_EXT_MAX)
    440 			return (KGE_UNK);
    441 
    442 		/*
    443 		 * Check length.  Use uint64_t because extlen is in units
    444 		 * of 64-bit words.  If length goes beyond the msgsize,
    445 		 * return an error.  (Zero length also qualifies here.)
    446 		 */
    447 		if (extv[0]->spd_ext_len == 0 ||
    448 		    (void *)((uint64_t *)extv[0] + extv[0]->spd_ext_len) >
    449 		    (void *)((uint8_t *)basehdr + msgsize))
    450 			return (KGE_LEN);
    451 
    452 		/* Check for redundant headers. */
    453 		if (extv[extv[0]->spd_ext_type] != NULL)
    454 			return (KGE_DUP);
    455 
    456 		/*
    457 		 * Reality check the extension if possible at the spdsock
    458 		 * level.
    459 		 */
    460 		if (!ext_check(extv[0]))
    461 			return (KGE_CHK);
    462 
    463 		/* If I make it here, assign the appropriate bin. */
    464 		extv[extv[0]->spd_ext_type] = extv[0];
    465 
    466 		/* Advance pointer (See above for uint64_t ptr reasoning.) */
    467 		extv[0] = (spd_ext_t *)
    468 		    ((uint64_t *)extv[0] + extv[0]->spd_ext_len);
    469 	}
    470 
    471 	/* Everything's cool. */
    472 
    473 	/*
    474 	 * If extv[0] == NULL, then there are no extension headers in this
    475 	 * message.  Ensure that this is the case.
    476 	 */
    477 	if (extv[0] == (spd_ext_t *)(basehdr + 1))
    478 		extv[0] = NULL;
    479 
    480 	return (KGE_OK);
    481 }
    482 
    483 static const int bad_ext_diag[] = {
    484 	SPD_DIAGNOSTIC_MALFORMED_LCLPORT,
    485 	SPD_DIAGNOSTIC_MALFORMED_REMPORT,
    486 	SPD_DIAGNOSTIC_MALFORMED_PROTO,
    487 	SPD_DIAGNOSTIC_MALFORMED_LCLADDR,
    488 	SPD_DIAGNOSTIC_MALFORMED_REMADDR,
    489 	SPD_DIAGNOSTIC_MALFORMED_ACTION,
    490 	SPD_DIAGNOSTIC_MALFORMED_RULE,
    491 	SPD_DIAGNOSTIC_MALFORMED_RULESET,
    492 	SPD_DIAGNOSTIC_MALFORMED_ICMP_TYPECODE
    493 };
    494 
    495 static const int dup_ext_diag[] = {
    496 	SPD_DIAGNOSTIC_DUPLICATE_LCLPORT,
    497 	SPD_DIAGNOSTIC_DUPLICATE_REMPORT,
    498 	SPD_DIAGNOSTIC_DUPLICATE_PROTO,
    499 	SPD_DIAGNOSTIC_DUPLICATE_LCLADDR,
    500 	SPD_DIAGNOSTIC_DUPLICATE_REMADDR,
    501 	SPD_DIAGNOSTIC_DUPLICATE_ACTION,
    502 	SPD_DIAGNOSTIC_DUPLICATE_RULE,
    503 	SPD_DIAGNOSTIC_DUPLICATE_RULESET,
    504 	SPD_DIAGNOSTIC_DUPLICATE_ICMP_TYPECODE
    505 };
    506 
    507 /*
    508  * Transmit a PF_POLICY error message to the instance either pointed to
    509  * by ks, the instance with serial number serial, or more, depending.
    510  *
    511  * The faulty message (or a reasonable facsimile thereof) is in mp.
    512  * This function will free mp or recycle it for delivery, thereby causing
    513  * the stream head to free it.
    514  */
    515 static void
    516 spdsock_error(queue_t *q, mblk_t *mp, int error, int diagnostic)
    517 {
    518 	spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
    519 
    520 	ASSERT(mp->b_datap->db_type == M_DATA);
    521 
    522 	if (spmsg->spd_msg_type < SPD_MIN ||
    523 	    spmsg->spd_msg_type > SPD_MAX)
    524 		spmsg->spd_msg_type = SPD_RESERVED;
    525 
    526 	/*
    527 	 * Strip out extension headers.
    528 	 */
    529 	ASSERT(mp->b_rptr + sizeof (*spmsg) <= mp->b_datap->db_lim);
    530 	mp->b_wptr = mp->b_rptr + sizeof (*spmsg);
    531 	spmsg->spd_msg_len = SPD_8TO64(sizeof (spd_msg_t));
    532 	spmsg->spd_msg_errno = (uint8_t)error;
    533 	spmsg->spd_msg_diagnostic = (uint16_t)diagnostic;
    534 
    535 	qreply(q, mp);
    536 }
    537 
    538 static void
    539 spdsock_diag(queue_t *q, mblk_t *mp, int diagnostic)
    540 {
    541 	spdsock_error(q, mp, EINVAL, diagnostic);
    542 }
    543 
    544 static void
    545 spd_echo(queue_t *q, mblk_t *mp)
    546 {
    547 	qreply(q, mp);
    548 }
    549 
    550 /*
    551  * Do NOT consume a reference to itp.
    552  */
    553 /*ARGSUSED*/
    554 static void
    555 spdsock_flush_node(ipsec_tun_pol_t *itp, void *cookie, netstack_t *ns)
    556 {
    557 	boolean_t active = (boolean_t)cookie;
    558 	ipsec_policy_head_t *iph;
    559 
    560 	iph = active ? itp->itp_policy : itp->itp_inactive;
    561 	IPPH_REFHOLD(iph);
    562 	mutex_enter(&itp->itp_lock);
    563 	spdsock_flush_one(iph, ns);
    564 	if (active)
    565 		itp->itp_flags &= ~ITPF_PFLAGS;
    566 	else
    567 		itp->itp_flags &= ~ITPF_IFLAGS;
    568 	mutex_exit(&itp->itp_lock);
    569 }
    570 
    571 /*
    572  * Clear out one polhead.
    573  */
    574 static void
    575 spdsock_flush_one(ipsec_policy_head_t *iph, netstack_t *ns)
    576 {
    577 	rw_enter(&iph->iph_lock, RW_WRITER);
    578 	ipsec_polhead_flush(iph, ns);
    579 	rw_exit(&iph->iph_lock);
    580 	IPPH_REFRELE(iph, ns);
    581 }
    582 
    583 static void
    584 spdsock_flush(queue_t *q, ipsec_policy_head_t *iph, ipsec_tun_pol_t *itp,
    585     mblk_t *mp)
    586 {
    587 	boolean_t active;
    588 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
    589 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
    590 
    591 	if (iph != ALL_ACTIVE_POLHEADS && iph != ALL_INACTIVE_POLHEADS) {
    592 		spdsock_flush_one(iph, ns);
    593 		if (audit_active) {
    594 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
    595 			cred_t *cr;
    596 			pid_t cpid;
    597 
    598 			cr = msg_getcred(mp, &cpid);
    599 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
    600 			audit_pf_policy(SPD_FLUSH, cr, ns,
    601 			    ITP_NAME(itp), active, 0, cpid);
    602 		}
    603 	} else {
    604 		active = (iph == ALL_ACTIVE_POLHEADS);
    605 
    606 		/* First flush the global policy. */
    607 		spdsock_flush_one(active ? ipsec_system_policy(ns) :
    608 		    ipsec_inactive_policy(ns), ns);
    609 		if (audit_active) {
    610 			cred_t *cr;
    611 			pid_t cpid;
    612 
    613 			cr = msg_getcred(mp, &cpid);
    614 			audit_pf_policy(SPD_FLUSH, cr, ns, NULL,
    615 			    active, 0, cpid);
    616 		}
    617 		/* Then flush every tunnel's appropriate one. */
    618 		itp_walk(spdsock_flush_node, (void *)active, ns);
    619 		if (audit_active) {
    620 			cred_t *cr;
    621 			pid_t cpid;
    622 
    623 			cr = msg_getcred(mp, &cpid);
    624 			audit_pf_policy(SPD_FLUSH, cr, ns,
    625 			    "all tunnels", active, 0, cpid);
    626 		}
    627 	}
    628 
    629 	spd_echo(q, mp);
    630 }
    631 
    632 static boolean_t
    633 spdsock_ext_to_sel(spd_ext_t **extv, ipsec_selkey_t *sel, int *diag)
    634 {
    635 	bzero(sel, sizeof (*sel));
    636 
    637 	if (extv[SPD_EXT_PROTO] != NULL) {
    638 		struct spd_proto *pr =
    639 		    (struct spd_proto *)extv[SPD_EXT_PROTO];
    640 		sel->ipsl_proto = pr->spd_proto_number;
    641 		sel->ipsl_valid |= IPSL_PROTOCOL;
    642 	}
    643 	if (extv[SPD_EXT_LCLPORT] != NULL) {
    644 		struct spd_portrange *pr =
    645 		    (struct spd_portrange *)extv[SPD_EXT_LCLPORT];
    646 		sel->ipsl_lport = pr->spd_ports_minport;
    647 		sel->ipsl_valid |= IPSL_LOCAL_PORT;
    648 	}
    649 	if (extv[SPD_EXT_REMPORT] != NULL) {
    650 		struct spd_portrange *pr =
    651 		    (struct spd_portrange *)extv[SPD_EXT_REMPORT];
    652 		sel->ipsl_rport = pr->spd_ports_minport;
    653 		sel->ipsl_valid |= IPSL_REMOTE_PORT;
    654 	}
    655 
    656 	if (extv[SPD_EXT_ICMP_TYPECODE] != NULL) {
    657 		struct spd_typecode *tc=
    658 		    (struct spd_typecode *)extv[SPD_EXT_ICMP_TYPECODE];
    659 
    660 		sel->ipsl_valid |= IPSL_ICMP_TYPE;
    661 		sel->ipsl_icmp_type = tc->spd_typecode_type;
    662 		if (tc->spd_typecode_type_end < tc->spd_typecode_type)
    663 			sel->ipsl_icmp_type_end = tc->spd_typecode_type;
    664 		else
    665 			sel->ipsl_icmp_type_end = tc->spd_typecode_type_end;
    666 
    667 		if (tc->spd_typecode_code != 255) {
    668 			sel->ipsl_valid |= IPSL_ICMP_CODE;
    669 			sel->ipsl_icmp_code = tc->spd_typecode_code;
    670 			if (tc->spd_typecode_code_end < tc->spd_typecode_code)
    671 				sel->ipsl_icmp_code_end = tc->spd_typecode_code;
    672 			else
    673 				sel->ipsl_icmp_code_end =
    674 				    tc->spd_typecode_code_end;
    675 		}
    676 	}
    677 #define	ADDR2SEL(sel, extv, field, pfield, extn, bit)			      \
    678 	if ((extv)[(extn)] != NULL) {					      \
    679 		uint_t addrlen;						      \
    680 		struct spd_address *ap = 				      \
    681 			(struct spd_address *)((extv)[(extn)]); 	      \
    682 		addrlen = (ap->spd_address_af == AF_INET6) ? 		      \
    683 			IPV6_ADDR_LEN : IP_ADDR_LEN;			      \
    684 		if (SPD_64TO8(ap->spd_address_len) < 			      \
    685 			(addrlen + sizeof (*ap))) {			      \
    686 			*diag = SPD_DIAGNOSTIC_BAD_ADDR_LEN;		      \
    687 			return (B_FALSE);				      \
    688 		}							      \
    689 		bcopy((ap+1), &((sel)->field), addrlen);		      \
    690 		(sel)->pfield = ap->spd_address_prefixlen;		      \
    691 		(sel)->ipsl_valid |= (bit);				      \
    692 		(sel)->ipsl_valid |= (ap->spd_address_af == AF_INET6) ?	      \
    693 			IPSL_IPV6 : IPSL_IPV4;				      \
    694 	}
    695 
    696 	ADDR2SEL(sel, extv, ipsl_local, ipsl_local_pfxlen,
    697 	    SPD_EXT_LCLADDR, IPSL_LOCAL_ADDR);
    698 	ADDR2SEL(sel, extv, ipsl_remote, ipsl_remote_pfxlen,
    699 	    SPD_EXT_REMADDR, IPSL_REMOTE_ADDR);
    700 
    701 	if ((sel->ipsl_valid & (IPSL_IPV6|IPSL_IPV4)) ==
    702 	    (IPSL_IPV6|IPSL_IPV4)) {
    703 		*diag = SPD_DIAGNOSTIC_MIXED_AF;
    704 		return (B_FALSE);
    705 	}
    706 
    707 #undef ADDR2SEL
    708 
    709 	return (B_TRUE);
    710 }
    711 
    712 static boolean_t
    713 spd_convert_type(uint32_t type, ipsec_act_t *act)
    714 {
    715 	switch (type) {
    716 	case SPD_ACTTYPE_DROP:
    717 		act->ipa_type = IPSEC_ACT_DISCARD;
    718 		return (B_TRUE);
    719 
    720 	case SPD_ACTTYPE_PASS:
    721 		act->ipa_type = IPSEC_ACT_CLEAR;
    722 		return (B_TRUE);
    723 
    724 	case SPD_ACTTYPE_IPSEC:
    725 		act->ipa_type = IPSEC_ACT_APPLY;
    726 		return (B_TRUE);
    727 	}
    728 	return (B_FALSE);
    729 }
    730 
    731 static boolean_t
    732 spd_convert_flags(uint32_t flags, ipsec_act_t *act)
    733 {
    734 	/*
    735 	 * Note use of !! for boolean canonicalization.
    736 	 */
    737 	act->ipa_apply.ipp_use_ah = !!(flags & SPD_APPLY_AH);
    738 	act->ipa_apply.ipp_use_esp = !!(flags & SPD_APPLY_ESP);
    739 	act->ipa_apply.ipp_use_espa = !!(flags & SPD_APPLY_ESPA);
    740 	act->ipa_apply.ipp_use_se = !!(flags & SPD_APPLY_SE);
    741 	act->ipa_apply.ipp_use_unique = !!(flags & SPD_APPLY_UNIQUE);
    742 	return (B_TRUE);
    743 }
    744 
    745 static void
    746 spdsock_reset_act(ipsec_act_t *act)
    747 {
    748 	bzero(act, sizeof (*act));
    749 	act->ipa_apply.ipp_espe_maxbits = IPSEC_MAX_KEYBITS;
    750 	act->ipa_apply.ipp_espa_maxbits = IPSEC_MAX_KEYBITS;
    751 	act->ipa_apply.ipp_ah_maxbits = IPSEC_MAX_KEYBITS;
    752 }
    753 
    754 /*
    755  * Sanity check action against reality, and shrink-wrap key sizes..
    756  */
    757 static boolean_t
    758 spdsock_check_action(ipsec_act_t *act, boolean_t tunnel_polhead, int *diag,
    759     spd_stack_t *spds)
    760 {
    761 	if (tunnel_polhead && act->ipa_apply.ipp_use_unique) {
    762 		*diag = SPD_DIAGNOSTIC_ADD_INCON_FLAGS;
    763 		return (B_FALSE);
    764 	}
    765 	if ((act->ipa_type != IPSEC_ACT_APPLY) &&
    766 	    (act->ipa_apply.ipp_use_ah ||
    767 	    act->ipa_apply.ipp_use_esp ||
    768 	    act->ipa_apply.ipp_use_espa ||
    769 	    act->ipa_apply.ipp_use_se ||
    770 	    act->ipa_apply.ipp_use_unique)) {
    771 		*diag = SPD_DIAGNOSTIC_ADD_INCON_FLAGS;
    772 		return (B_FALSE);
    773 	}
    774 	if ((act->ipa_type == IPSEC_ACT_APPLY) &&
    775 	    !act->ipa_apply.ipp_use_ah &&
    776 	    !act->ipa_apply.ipp_use_esp) {
    777 		*diag = SPD_DIAGNOSTIC_ADD_INCON_FLAGS;
    778 		return (B_FALSE);
    779 	}
    780 	return (ipsec_check_action(act, diag, spds->spds_netstack));
    781 }
    782 
    783 /*
    784  * We may be short a few error checks here..
    785  */
    786 static boolean_t
    787 spdsock_ext_to_actvec(spd_ext_t **extv, ipsec_act_t **actpp, uint_t *nactp,
    788     int *diag, spd_stack_t *spds)
    789 {
    790 	struct spd_ext_actions *sactp =
    791 	    (struct spd_ext_actions *)extv[SPD_EXT_ACTION];
    792 	ipsec_act_t act, *actp, *endactp;
    793 	struct spd_attribute *attrp, *endattrp;
    794 	uint64_t *endp;
    795 	int nact;
    796 	boolean_t tunnel_polhead;
    797 
    798 	tunnel_polhead = (extv[SPD_EXT_TUN_NAME] != NULL &&
    799 	    (((struct spd_rule *)extv[SPD_EXT_RULE])->spd_rule_flags &
    800 	    SPD_RULE_FLAG_TUNNEL));
    801 
    802 	*actpp = NULL;
    803 	*nactp = 0;
    804 
    805 	if (sactp == NULL) {
    806 		*diag = SPD_DIAGNOSTIC_NO_ACTION_EXT;
    807 		return (B_FALSE);
    808 	}
    809 
    810 	/*
    811 	 * Parse the "action" extension and convert into an action chain.
    812 	 */
    813 
    814 	nact = sactp->spd_actions_count;
    815 
    816 	endp = (uint64_t *)sactp;
    817 	endp += sactp->spd_actions_len;
    818 	endattrp = (struct spd_attribute *)endp;
    819 
    820 	actp = kmem_alloc(sizeof (*actp) * nact, KM_NOSLEEP);
    821 	if (actp == NULL) {
    822 		*diag = SPD_DIAGNOSTIC_ADD_NO_MEM;
    823 		return (B_FALSE);
    824 	}
    825 	*actpp = actp;
    826 	*nactp = nact;
    827 	endactp = actp + nact;
    828 
    829 	spdsock_reset_act(&act);
    830 	attrp = (struct spd_attribute *)(&sactp[1]);
    831 
    832 	for (; attrp < endattrp; attrp++) {
    833 		switch (attrp->spd_attr_tag) {
    834 		case SPD_ATTR_NOP:
    835 			break;
    836 
    837 		case SPD_ATTR_EMPTY:
    838 			spdsock_reset_act(&act);
    839 			break;
    840 
    841 		case SPD_ATTR_END:
    842 			attrp = endattrp;
    843 			/* FALLTHRU */
    844 		case SPD_ATTR_NEXT:
    845 			if (actp >= endactp) {
    846 				*diag = SPD_DIAGNOSTIC_ADD_WRONG_ACT_COUNT;
    847 				goto fail;
    848 			}
    849 			if (!spdsock_check_action(&act, tunnel_polhead,
    850 			    diag, spds))
    851 				goto fail;
    852 			*actp++ = act;
    853 			spdsock_reset_act(&act);
    854 			break;
    855 
    856 		case SPD_ATTR_TYPE:
    857 			if (!spd_convert_type(attrp->spd_attr_value, &act)) {
    858 				*diag = SPD_DIAGNOSTIC_ADD_BAD_TYPE;
    859 				goto fail;
    860 			}
    861 			break;
    862 
    863 		case SPD_ATTR_FLAGS:
    864 			if (!tunnel_polhead && extv[SPD_EXT_TUN_NAME] != NULL) {
    865 				/*
    866 				 * Set "sa unique" for transport-mode
    867 				 * tunnels whether we want to or not.
    868 				 */
    869 				attrp->spd_attr_value |= SPD_APPLY_UNIQUE;
    870 			}
    871 			if (!spd_convert_flags(attrp->spd_attr_value, &act)) {
    872 				*diag = SPD_DIAGNOSTIC_ADD_BAD_FLAGS;
    873 				goto fail;
    874 			}
    875 			break;
    876 
    877 		case SPD_ATTR_AH_AUTH:
    878 			if (attrp->spd_attr_value == 0) {
    879 				*diag = SPD_DIAGNOSTIC_UNSUPP_AH_ALG;
    880 				goto fail;
    881 			}
    882 			act.ipa_apply.ipp_auth_alg = attrp->spd_attr_value;
    883 			break;
    884 
    885 		case SPD_ATTR_ESP_ENCR:
    886 			if (attrp->spd_attr_value == 0) {
    887 				*diag = SPD_DIAGNOSTIC_UNSUPP_ESP_ENCR_ALG;
    888 				goto fail;
    889 			}
    890 			act.ipa_apply.ipp_encr_alg = attrp->spd_attr_value;
    891 			break;
    892 
    893 		case SPD_ATTR_ESP_AUTH:
    894 			if (attrp->spd_attr_value == 0) {
    895 				*diag = SPD_DIAGNOSTIC_UNSUPP_ESP_AUTH_ALG;
    896 				goto fail;
    897 			}
    898 			act.ipa_apply.ipp_esp_auth_alg = attrp->spd_attr_value;
    899 			break;
    900 
    901 		case SPD_ATTR_ENCR_MINBITS:
    902 			act.ipa_apply.ipp_espe_minbits = attrp->spd_attr_value;
    903 			break;
    904 
    905 		case SPD_ATTR_ENCR_MAXBITS:
    906 			act.ipa_apply.ipp_espe_maxbits = attrp->spd_attr_value;
    907 			break;
    908 
    909 		case SPD_ATTR_AH_MINBITS:
    910 			act.ipa_apply.ipp_ah_minbits = attrp->spd_attr_value;
    911 			break;
    912 
    913 		case SPD_ATTR_AH_MAXBITS:
    914 			act.ipa_apply.ipp_ah_maxbits = attrp->spd_attr_value;
    915 			break;
    916 
    917 		case SPD_ATTR_ESPA_MINBITS:
    918 			act.ipa_apply.ipp_espa_minbits = attrp->spd_attr_value;
    919 			break;
    920 
    921 		case SPD_ATTR_ESPA_MAXBITS:
    922 			act.ipa_apply.ipp_espa_maxbits = attrp->spd_attr_value;
    923 			break;
    924 
    925 		case SPD_ATTR_LIFE_SOFT_TIME:
    926 		case SPD_ATTR_LIFE_HARD_TIME:
    927 		case SPD_ATTR_LIFE_SOFT_BYTES:
    928 		case SPD_ATTR_LIFE_HARD_BYTES:
    929 			break;
    930 
    931 		case SPD_ATTR_KM_PROTO:
    932 			act.ipa_apply.ipp_km_proto = attrp->spd_attr_value;
    933 			break;
    934 
    935 		case SPD_ATTR_KM_COOKIE:
    936 			act.ipa_apply.ipp_km_cookie = attrp->spd_attr_value;
    937 			break;
    938 
    939 		case SPD_ATTR_REPLAY_DEPTH:
    940 			act.ipa_apply.ipp_replay_depth = attrp->spd_attr_value;
    941 			break;
    942 		}
    943 	}
    944 	if (actp != endactp) {
    945 		*diag = SPD_DIAGNOSTIC_ADD_WRONG_ACT_COUNT;
    946 		goto fail;
    947 	}
    948 
    949 	return (B_TRUE);
    950 fail:
    951 	ipsec_actvec_free(*actpp, nact);
    952 	*actpp = NULL;
    953 	return (B_FALSE);
    954 }
    955 
    956 typedef struct
    957 {
    958 	ipsec_policy_t *pol;
    959 	int dir;
    960 } tmprule_t;
    961 
    962 static int
    963 mkrule(ipsec_policy_head_t *iph, struct spd_rule *rule,
    964     ipsec_selkey_t *sel, ipsec_act_t *actp, int nact, uint_t dir, uint_t af,
    965     tmprule_t **rp, uint64_t *index, spd_stack_t *spds)
    966 {
    967 	ipsec_policy_t *pol;
    968 
    969 	sel->ipsl_valid &= ~(IPSL_IPV6|IPSL_IPV4);
    970 	sel->ipsl_valid |= af;
    971 
    972 	pol = ipsec_policy_create(sel, actp, nact, rule->spd_rule_priority,
    973 	    index, spds->spds_netstack);
    974 	if (pol == NULL)
    975 		return (ENOMEM);
    976 
    977 	(*rp)->pol = pol;
    978 	(*rp)->dir = dir;
    979 	(*rp)++;
    980 
    981 	if (!ipsec_check_policy(iph, pol, dir))
    982 		return (EEXIST);
    983 
    984 	rule->spd_rule_index = pol->ipsp_index;
    985 	return (0);
    986 }
    987 
    988 static int
    989 mkrulepair(ipsec_policy_head_t *iph, struct spd_rule *rule,
    990     ipsec_selkey_t *sel, ipsec_act_t *actp, int nact, uint_t dir, uint_t afs,
    991     tmprule_t **rp, uint64_t *index, spd_stack_t *spds)
    992 {
    993 	int error;
    994 
    995 	if (afs & IPSL_IPV4) {
    996 		error = mkrule(iph, rule, sel, actp, nact, dir, IPSL_IPV4, rp,
    997 		    index, spds);
    998 		if (error != 0)
    999 			return (error);
   1000 	}
   1001 	if (afs & IPSL_IPV6) {
   1002 		error = mkrule(iph, rule, sel, actp, nact, dir, IPSL_IPV6, rp,
   1003 		    index, spds);
   1004 		if (error != 0)
   1005 			return (error);
   1006 	}
   1007 	return (0);
   1008 }
   1009 
   1010 
   1011 static void
   1012 spdsock_addrule(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp,
   1013     spd_ext_t **extv, ipsec_tun_pol_t *itp)
   1014 {
   1015 	ipsec_selkey_t sel;
   1016 	ipsec_act_t *actp;
   1017 	uint_t nact;
   1018 	int diag = 0, error, afs;
   1019 	struct spd_rule *rule = (struct spd_rule *)extv[SPD_EXT_RULE];
   1020 	tmprule_t rules[4], *rulep = &rules[0];
   1021 	boolean_t tunnel_mode, empty_itp, active;
   1022 	uint64_t *index = (itp == NULL) ? NULL : &itp->itp_next_policy_index;
   1023 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   1024 	spd_stack_t	*spds = ss->spdsock_spds;
   1025 
   1026 	if (rule == NULL) {
   1027 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_NO_RULE_EXT);
   1028 		if (audit_active) {
   1029 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   1030 			cred_t *cr;
   1031 			pid_t cpid;
   1032 
   1033 			cr = msg_getcred(mp, &cpid);
   1034 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   1035 			audit_pf_policy(SPD_ADDRULE, cr,
   1036 			    spds->spds_netstack, ITP_NAME(itp), active,
   1037 			    SPD_DIAGNOSTIC_NO_RULE_EXT, cpid);
   1038 		}
   1039 		return;
   1040 	}
   1041 
   1042 	tunnel_mode = (rule->spd_rule_flags & SPD_RULE_FLAG_TUNNEL);
   1043 
   1044 	if (itp != NULL) {
   1045 		mutex_enter(&itp->itp_lock);
   1046 		ASSERT(itp->itp_policy == iph || itp->itp_inactive == iph);
   1047 		active = (itp->itp_policy == iph);
   1048 		if (ITP_P_ISACTIVE(itp, iph)) {
   1049 			/* Check for mix-and-match of tunnel/transport. */
   1050 			if ((tunnel_mode && !ITP_P_ISTUNNEL(itp, iph)) ||
   1051 			    (!tunnel_mode && ITP_P_ISTUNNEL(itp, iph))) {
   1052 				mutex_exit(&itp->itp_lock);
   1053 				spdsock_error(q, mp, EBUSY, 0);
   1054 				return;
   1055 			}
   1056 			empty_itp = B_FALSE;
   1057 		} else {
   1058 			empty_itp = B_TRUE;
   1059 			itp->itp_flags = active ? ITPF_P_ACTIVE : ITPF_I_ACTIVE;
   1060 			if (tunnel_mode)
   1061 				itp->itp_flags |= active ? ITPF_P_TUNNEL :
   1062 				    ITPF_I_TUNNEL;
   1063 		}
   1064 	} else {
   1065 		empty_itp = B_FALSE;
   1066 	}
   1067 
   1068 	if (rule->spd_rule_index != 0) {
   1069 		diag = SPD_DIAGNOSTIC_INVALID_RULE_INDEX;
   1070 		error = EINVAL;
   1071 		goto fail2;
   1072 	}
   1073 
   1074 	if (!spdsock_ext_to_sel(extv, &sel, &diag)) {
   1075 		error = EINVAL;
   1076 		goto fail2;
   1077 	}
   1078 
   1079 	if (itp != NULL) {
   1080 		if (tunnel_mode) {
   1081 			if (sel.ipsl_valid &
   1082 			    (IPSL_REMOTE_PORT | IPSL_LOCAL_PORT)) {
   1083 				itp->itp_flags |= active ?
   1084 				    ITPF_P_PER_PORT_SECURITY :
   1085 				    ITPF_I_PER_PORT_SECURITY;
   1086 			}
   1087 		} else {
   1088 			/*
   1089 			 * For now, we don't allow transport-mode on a tunnel
   1090 			 * with ANY specific selectors.  Bail if we have such
   1091 			 * a request.
   1092 			 */
   1093 			if (sel.ipsl_valid & IPSL_WILDCARD) {
   1094 				diag = SPD_DIAGNOSTIC_NO_TUNNEL_SELECTORS;
   1095 				error = EINVAL;
   1096 				goto fail2;
   1097 			}
   1098 		}
   1099 	}
   1100 
   1101 	if (!spdsock_ext_to_actvec(extv, &actp, &nact, &diag, spds)) {
   1102 		error = EINVAL;
   1103 		goto fail2;
   1104 	}
   1105 	/*
   1106 	 * If no addresses were specified, add both.
   1107 	 */
   1108 	afs = sel.ipsl_valid & (IPSL_IPV6|IPSL_IPV4);
   1109 	if (afs == 0)
   1110 		afs = (IPSL_IPV6|IPSL_IPV4);
   1111 
   1112 	rw_enter(&iph->iph_lock, RW_WRITER);
   1113 
   1114 	if (rule->spd_rule_flags & SPD_RULE_FLAG_OUTBOUND) {
   1115 		error = mkrulepair(iph, rule, &sel, actp, nact,
   1116 		    IPSEC_TYPE_OUTBOUND, afs, &rulep, index, spds);
   1117 		if (error != 0)
   1118 			goto fail;
   1119 	}
   1120 
   1121 	if (rule->spd_rule_flags & SPD_RULE_FLAG_INBOUND) {
   1122 		error = mkrulepair(iph, rule, &sel, actp, nact,
   1123 		    IPSEC_TYPE_INBOUND, afs, &rulep, index, spds);
   1124 		if (error != 0)
   1125 			goto fail;
   1126 	}
   1127 
   1128 	while ((--rulep) >= &rules[0]) {
   1129 		ipsec_enter_policy(iph, rulep->pol, rulep->dir,
   1130 		    spds->spds_netstack);
   1131 	}
   1132 	rw_exit(&iph->iph_lock);
   1133 	if (itp != NULL)
   1134 		mutex_exit(&itp->itp_lock);
   1135 
   1136 	ipsec_actvec_free(actp, nact);
   1137 	spd_echo(q, mp);
   1138 	if (audit_active) {
   1139 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   1140 		cred_t *cr;
   1141 		pid_t cpid;
   1142 
   1143 		cr = msg_getcred(mp, &cpid);
   1144 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   1145 		audit_pf_policy(SPD_ADDRULE, cr, spds->spds_netstack,
   1146 		    ITP_NAME(itp), active, 0, cpid);
   1147 	}
   1148 	return;
   1149 
   1150 fail:
   1151 	rw_exit(&iph->iph_lock);
   1152 	while ((--rulep) >= &rules[0])
   1153 		IPPOL_REFRELE(rulep->pol);
   1154 	ipsec_actvec_free(actp, nact);
   1155 fail2:
   1156 	if (itp != NULL) {
   1157 		if (empty_itp)
   1158 			itp->itp_flags = 0;
   1159 		mutex_exit(&itp->itp_lock);
   1160 	}
   1161 	spdsock_error(q, mp, error, diag);
   1162 	if (audit_active) {
   1163 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   1164 		cred_t *cr;
   1165 		pid_t cpid;
   1166 
   1167 		cr = msg_getcred(mp, &cpid);
   1168 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   1169 		audit_pf_policy(SPD_ADDRULE, cr, spds->spds_netstack,
   1170 		    ITP_NAME(itp), active, error, cpid);
   1171 	}
   1172 }
   1173 
   1174 void
   1175 spdsock_deleterule(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp,
   1176     spd_ext_t **extv, ipsec_tun_pol_t *itp)
   1177 {
   1178 	ipsec_selkey_t sel;
   1179 	struct spd_rule *rule = (struct spd_rule *)extv[SPD_EXT_RULE];
   1180 	int err, diag = 0;
   1181 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   1182 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
   1183 
   1184 	if (rule == NULL) {
   1185 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_NO_RULE_EXT);
   1186 		if (audit_active) {
   1187 			boolean_t active;
   1188 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   1189 			cred_t *cr;
   1190 			pid_t cpid;
   1191 
   1192 			cr = msg_getcred(mp, &cpid);
   1193 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   1194 			audit_pf_policy(SPD_DELETERULE, cr, ns,
   1195 			    ITP_NAME(itp), active, SPD_DIAGNOSTIC_NO_RULE_EXT,
   1196 			    cpid);
   1197 		}
   1198 		return;
   1199 	}
   1200 
   1201 	/*
   1202 	 * Must enter itp_lock first to avoid deadlock.  See tun.c's
   1203 	 * set_sec_simple() for the other case of itp_lock and iph_lock.
   1204 	 */
   1205 	if (itp != NULL)
   1206 		mutex_enter(&itp->itp_lock);
   1207 
   1208 	if (rule->spd_rule_index != 0) {
   1209 		if (ipsec_policy_delete_index(iph, rule->spd_rule_index, ns) !=
   1210 		    0) {
   1211 			err = ESRCH;
   1212 			goto fail;
   1213 		}
   1214 	} else {
   1215 		if (!spdsock_ext_to_sel(extv, &sel, &diag)) {
   1216 			err = EINVAL;	/* diag already set... */
   1217 			goto fail;
   1218 		}
   1219 
   1220 		if ((rule->spd_rule_flags & SPD_RULE_FLAG_INBOUND) &&
   1221 		    !ipsec_policy_delete(iph, &sel, IPSEC_TYPE_INBOUND, ns)) {
   1222 			err = ESRCH;
   1223 			goto fail;
   1224 		}
   1225 
   1226 		if ((rule->spd_rule_flags & SPD_RULE_FLAG_OUTBOUND) &&
   1227 		    !ipsec_policy_delete(iph, &sel, IPSEC_TYPE_OUTBOUND, ns)) {
   1228 			err = ESRCH;
   1229 			goto fail;
   1230 		}
   1231 	}
   1232 
   1233 	if (itp != NULL) {
   1234 		ASSERT(iph == itp->itp_policy || iph == itp->itp_inactive);
   1235 		rw_enter(&iph->iph_lock, RW_READER);
   1236 		if (avl_numnodes(&iph->iph_rulebyid) == 0) {
   1237 			if (iph == itp->itp_policy)
   1238 				itp->itp_flags &= ~ITPF_PFLAGS;
   1239 			else
   1240 				itp->itp_flags &= ~ITPF_IFLAGS;
   1241 		}
   1242 		/* Can exit locks in any order. */
   1243 		rw_exit(&iph->iph_lock);
   1244 		mutex_exit(&itp->itp_lock);
   1245 	}
   1246 	spd_echo(q, mp);
   1247 	if (audit_active) {
   1248 		boolean_t active;
   1249 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   1250 		cred_t *cr;
   1251 		pid_t cpid;
   1252 
   1253 		cr = msg_getcred(mp, &cpid);
   1254 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   1255 		audit_pf_policy(SPD_DELETERULE, cr, ns, ITP_NAME(itp),
   1256 		    active, 0, cpid);
   1257 	}
   1258 	return;
   1259 fail:
   1260 	if (itp != NULL)
   1261 		mutex_exit(&itp->itp_lock);
   1262 	spdsock_error(q, mp, err, diag);
   1263 	if (audit_active) {
   1264 		boolean_t active;
   1265 		spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   1266 		cred_t *cr;
   1267 		pid_t cpid;
   1268 
   1269 		cr = msg_getcred(mp, &cpid);
   1270 		active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   1271 		audit_pf_policy(SPD_DELETERULE, cr, ns, ITP_NAME(itp),
   1272 		    active, err, cpid);
   1273 	}
   1274 }
   1275 
   1276 /* Do NOT consume a reference to itp. */
   1277 /* ARGSUSED */
   1278 static void
   1279 spdsock_flip_node(ipsec_tun_pol_t *itp, void *ignoreme, netstack_t *ns)
   1280 {
   1281 	mutex_enter(&itp->itp_lock);
   1282 	ITPF_SWAP(itp->itp_flags);
   1283 	ipsec_swap_policy(itp->itp_policy, itp->itp_inactive, ns);
   1284 	mutex_exit(&itp->itp_lock);
   1285 }
   1286 
   1287 void
   1288 spdsock_flip(queue_t *q, mblk_t *mp, spd_if_t *tunname)
   1289 {
   1290 	char *tname;
   1291 	ipsec_tun_pol_t *itp;
   1292 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   1293 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
   1294 
   1295 	if (tunname != NULL) {
   1296 		tname = (char *)tunname->spd_if_name;
   1297 		if (*tname == '\0') {
   1298 			/* can't fail */
   1299 			ipsec_swap_global_policy(ns);
   1300 			if (audit_active) {
   1301 				boolean_t active;
   1302 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   1303 				cred_t *cr;
   1304 				pid_t cpid;
   1305 
   1306 				cr = msg_getcred(mp, &cpid);
   1307 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   1308 				audit_pf_policy(SPD_FLIP, cr, ns,
   1309 				    NULL, active, 0, cpid);
   1310 			}
   1311 			itp_walk(spdsock_flip_node, NULL, ns);
   1312 			if (audit_active) {
   1313 				boolean_t active;
   1314 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   1315 				cred_t *cr;
   1316 				pid_t cpid;
   1317 
   1318 				cr = msg_getcred(mp, &cpid);
   1319 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   1320 				audit_pf_policy(SPD_FLIP, cr, ns,
   1321 				    "all tunnels", active, 0, cpid);
   1322 			}
   1323 		} else {
   1324 			itp = get_tunnel_policy(tname, ns);
   1325 			if (itp == NULL) {
   1326 				/* Better idea for "tunnel not found"? */
   1327 				spdsock_error(q, mp, ESRCH, 0);
   1328 				if (audit_active) {
   1329 					boolean_t active;
   1330 					spd_msg_t *spmsg =
   1331 					    (spd_msg_t *)mp->b_rptr;
   1332 					cred_t *cr;
   1333 					pid_t cpid;
   1334 
   1335 					cr = msg_getcred(mp, &cpid);
   1336 					active = (spmsg->spd_msg_spdid ==
   1337 					    SPD_ACTIVE);
   1338 					audit_pf_policy(SPD_FLIP, cr, ns,
   1339 					    ITP_NAME(itp), active,
   1340 					    ESRCH, cpid);
   1341 				}
   1342 				return;
   1343 			}
   1344 			spdsock_flip_node(itp, NULL, NULL);
   1345 			if (audit_active) {
   1346 				boolean_t active;
   1347 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   1348 				cred_t *cr;
   1349 				pid_t cpid;
   1350 
   1351 				cr = msg_getcred(mp, &cpid);
   1352 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   1353 				audit_pf_policy(SPD_FLIP, cr, ns,
   1354 				    ITP_NAME(itp), active, 0, cpid);
   1355 			}
   1356 			ITP_REFRELE(itp, ns);
   1357 		}
   1358 	} else {
   1359 		ipsec_swap_global_policy(ns);	/* can't fail */
   1360 		if (audit_active) {
   1361 			boolean_t active;
   1362 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   1363 			cred_t *cr;
   1364 			pid_t cpid;
   1365 
   1366 			cr = msg_getcred(mp, &cpid);
   1367 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   1368 			audit_pf_policy(SPD_FLIP, cr,
   1369 			    ns, NULL, active, 0, cpid);
   1370 		}
   1371 	}
   1372 	spd_echo(q, mp);
   1373 }
   1374 
   1375 /*
   1376  * Unimplemented feature
   1377  */
   1378 /* ARGSUSED */
   1379 static void
   1380 spdsock_lookup(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp,
   1381     spd_ext_t **extv, ipsec_tun_pol_t *itp)
   1382 {
   1383 	spdsock_error(q, mp, EINVAL, 0);
   1384 }
   1385 
   1386 
   1387 static mblk_t *
   1388 spdsock_dump_ruleset(mblk_t *req, ipsec_policy_head_t *iph,
   1389     uint32_t count, uint16_t error)
   1390 {
   1391 	size_t len = sizeof (spd_ruleset_ext_t) + sizeof (spd_msg_t);
   1392 	spd_msg_t *msg;
   1393 	spd_ruleset_ext_t *ruleset;
   1394 	mblk_t *m = allocb(len, BPRI_HI);
   1395 
   1396 	ASSERT(RW_READ_HELD(&iph->iph_lock));
   1397 
   1398 	if (m == NULL) {
   1399 		return (NULL);
   1400 	}
   1401 	msg = (spd_msg_t *)m->b_rptr;
   1402 	ruleset = (spd_ruleset_ext_t *)(&msg[1]);
   1403 
   1404 	m->b_wptr = (uint8_t *)&ruleset[1];
   1405 
   1406 	*msg = *(spd_msg_t *)(req->b_rptr);
   1407 	msg->spd_msg_len = SPD_8TO64(len);
   1408 	msg->spd_msg_errno = error;
   1409 
   1410 	ruleset->spd_ruleset_len = SPD_8TO64(sizeof (*ruleset));
   1411 	ruleset->spd_ruleset_type = SPD_EXT_RULESET;
   1412 	ruleset->spd_ruleset_count = count;
   1413 	ruleset->spd_ruleset_version = iph->iph_gen;
   1414 	return (m);
   1415 }
   1416 
   1417 static mblk_t *
   1418 spdsock_dump_finish(spdsock_t *ss, int error)
   1419 {
   1420 	mblk_t *m;
   1421 	ipsec_policy_head_t *iph = ss->spdsock_dump_head;
   1422 	mblk_t *req = ss->spdsock_dump_req;
   1423 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
   1424 
   1425 	rw_enter(&iph->iph_lock, RW_READER);
   1426 	m = spdsock_dump_ruleset(req, iph, ss->spdsock_dump_count, error);
   1427 	rw_exit(&iph->iph_lock);
   1428 	IPPH_REFRELE(iph, ns);
   1429 	if (ss->spdsock_itp != NULL) {
   1430 		ITP_REFRELE(ss->spdsock_itp, ns);
   1431 		ss->spdsock_itp = NULL;
   1432 	}
   1433 	ss->spdsock_dump_req = NULL;
   1434 	freemsg(req);
   1435 
   1436 	return (m);
   1437 }
   1438 
   1439 /*
   1440  * Rule encoding functions.
   1441  * We do a two-pass encode.
   1442  * If base != NULL, fill in encoded rule part starting at base+offset.
   1443  * Always return "offset" plus length of to-be-encoded data.
   1444  */
   1445 static uint_t
   1446 spdsock_encode_typecode(uint8_t *base, uint_t offset, uint8_t type,
   1447     uint8_t type_end, uint8_t code, uint8_t code_end)
   1448 {
   1449 	struct spd_typecode *tcp;
   1450 
   1451 	ASSERT(ALIGNED64(offset));
   1452 
   1453 	if (base != NULL) {
   1454 		tcp = (struct spd_typecode *)(base + offset);
   1455 		tcp->spd_typecode_len = SPD_8TO64(sizeof (*tcp));
   1456 		tcp->spd_typecode_exttype = SPD_EXT_ICMP_TYPECODE;
   1457 		tcp->spd_typecode_code = code;
   1458 		tcp->spd_typecode_type = type;
   1459 		tcp->spd_typecode_type_end = type_end;
   1460 		tcp->spd_typecode_code_end = code_end;
   1461 	}
   1462 	offset += sizeof (*tcp);
   1463 
   1464 	ASSERT(ALIGNED64(offset));
   1465 
   1466 	return (offset);
   1467 }
   1468 
   1469 static uint_t
   1470 spdsock_encode_proto(uint8_t *base, uint_t offset, uint8_t proto)
   1471 {
   1472 	struct spd_proto *spp;
   1473 
   1474 	ASSERT(ALIGNED64(offset));
   1475 
   1476 	if (base != NULL) {
   1477 		spp = (struct spd_proto *)(base + offset);
   1478 		spp->spd_proto_len = SPD_8TO64(sizeof (*spp));
   1479 		spp->spd_proto_exttype = SPD_EXT_PROTO;
   1480 		spp->spd_proto_number = proto;
   1481 		spp->spd_proto_reserved1 = 0;
   1482 		spp->spd_proto_reserved2 = 0;
   1483 	}
   1484 	offset += sizeof (*spp);
   1485 
   1486 	ASSERT(ALIGNED64(offset));
   1487 
   1488 	return (offset);
   1489 }
   1490 
   1491 static uint_t
   1492 spdsock_encode_port(uint8_t *base, uint_t offset, uint16_t ext, uint16_t port)
   1493 {
   1494 	struct spd_portrange *spp;
   1495 
   1496 	ASSERT(ALIGNED64(offset));
   1497 
   1498 	if (base != NULL) {
   1499 		spp = (struct spd_portrange *)(base + offset);
   1500 		spp->spd_ports_len = SPD_8TO64(sizeof (*spp));
   1501 		spp->spd_ports_exttype = ext;
   1502 		spp->spd_ports_minport = port;
   1503 		spp->spd_ports_maxport = port;
   1504 	}
   1505 	offset += sizeof (*spp);
   1506 
   1507 	ASSERT(ALIGNED64(offset));
   1508 
   1509 	return (offset);
   1510 }
   1511 
   1512 static uint_t
   1513 spdsock_encode_addr(uint8_t *base, uint_t offset, uint16_t ext,
   1514     const ipsec_selkey_t *sel, const ipsec_addr_t *addr, uint_t pfxlen)
   1515 {
   1516 	struct spd_address *sae;
   1517 	ipsec_addr_t *spdaddr;
   1518 	uint_t start = offset;
   1519 	uint_t addrlen;
   1520 	uint_t af;
   1521 
   1522 	if (sel->ipsl_valid & IPSL_IPV4) {
   1523 		af = AF_INET;
   1524 		addrlen = IP_ADDR_LEN;
   1525 	} else {
   1526 		af = AF_INET6;
   1527 		addrlen = IPV6_ADDR_LEN;
   1528 	}
   1529 
   1530 	ASSERT(ALIGNED64(offset));
   1531 
   1532 	if (base != NULL) {
   1533 		sae = (struct spd_address *)(base + offset);
   1534 		sae->spd_address_exttype = ext;
   1535 		sae->spd_address_af = af;
   1536 		sae->spd_address_prefixlen = pfxlen;
   1537 		sae->spd_address_reserved2 = 0;
   1538 
   1539 		spdaddr = (ipsec_addr_t *)(&sae[1]);
   1540 		bcopy(addr, spdaddr, addrlen);
   1541 	}
   1542 	offset += sizeof (*sae);
   1543 	addrlen = roundup(addrlen, sizeof (uint64_t));
   1544 	offset += addrlen;
   1545 
   1546 	ASSERT(ALIGNED64(offset));
   1547 
   1548 	if (base != NULL)
   1549 		sae->spd_address_len = SPD_8TO64(offset - start);
   1550 	return (offset);
   1551 }
   1552 
   1553 static uint_t
   1554 spdsock_encode_sel(uint8_t *base, uint_t offset, const ipsec_sel_t *sel)
   1555 {
   1556 	const ipsec_selkey_t *selkey = &sel->ipsl_key;
   1557 
   1558 	if (selkey->ipsl_valid & IPSL_PROTOCOL)
   1559 		offset = spdsock_encode_proto(base, offset, selkey->ipsl_proto);
   1560 	if (selkey->ipsl_valid & IPSL_LOCAL_PORT)
   1561 		offset = spdsock_encode_port(base, offset, SPD_EXT_LCLPORT,
   1562 		    selkey->ipsl_lport);
   1563 	if (selkey->ipsl_valid & IPSL_REMOTE_PORT)
   1564 		offset = spdsock_encode_port(base, offset, SPD_EXT_REMPORT,
   1565 		    selkey->ipsl_rport);
   1566 	if (selkey->ipsl_valid & IPSL_REMOTE_ADDR)
   1567 		offset = spdsock_encode_addr(base, offset, SPD_EXT_REMADDR,
   1568 		    selkey, &selkey->ipsl_remote, selkey->ipsl_remote_pfxlen);
   1569 	if (selkey->ipsl_valid & IPSL_LOCAL_ADDR)
   1570 		offset = spdsock_encode_addr(base, offset, SPD_EXT_LCLADDR,
   1571 		    selkey, &selkey->ipsl_local, selkey->ipsl_local_pfxlen);
   1572 	if (selkey->ipsl_valid & IPSL_ICMP_TYPE) {
   1573 		offset = spdsock_encode_typecode(base, offset,
   1574 		    selkey->ipsl_icmp_type, selkey->ipsl_icmp_type_end,
   1575 		    (selkey->ipsl_valid & IPSL_ICMP_CODE) ?
   1576 		    selkey->ipsl_icmp_code : 255,
   1577 		    (selkey->ipsl_valid & IPSL_ICMP_CODE) ?
   1578 		    selkey->ipsl_icmp_code_end : 255);
   1579 	}
   1580 	return (offset);
   1581 }
   1582 
   1583 static uint_t
   1584 spdsock_encode_actattr(uint8_t *base, uint_t offset, uint32_t tag,
   1585     uint32_t value)
   1586 {
   1587 	struct spd_attribute *attr;
   1588 
   1589 	ASSERT(ALIGNED64(offset));
   1590 
   1591 	if (base != NULL) {
   1592 		attr = (struct spd_attribute *)(base + offset);
   1593 		attr->spd_attr_tag = tag;
   1594 		attr->spd_attr_value = value;
   1595 	}
   1596 	offset += sizeof (struct spd_attribute);
   1597 
   1598 	ASSERT(ALIGNED64(offset));
   1599 
   1600 	return (offset);
   1601 }
   1602 
   1603 
   1604 #define	EMIT(t, v) offset = spdsock_encode_actattr(base, offset, (t), (v))
   1605 
   1606 static uint_t
   1607 spdsock_encode_action(uint8_t *base, uint_t offset, const ipsec_action_t *ap)
   1608 {
   1609 	const struct ipsec_act *act = &(ap->ipa_act);
   1610 	uint_t flags;
   1611 
   1612 	EMIT(SPD_ATTR_EMPTY, 0);
   1613 	switch (act->ipa_type) {
   1614 	case IPSEC_ACT_DISCARD:
   1615 	case IPSEC_ACT_REJECT:
   1616 		EMIT(SPD_ATTR_TYPE, SPD_ACTTYPE_DROP);
   1617 		break;
   1618 	case IPSEC_ACT_BYPASS:
   1619 	case IPSEC_ACT_CLEAR:
   1620 		EMIT(SPD_ATTR_TYPE, SPD_ACTTYPE_PASS);
   1621 		break;
   1622 
   1623 	case IPSEC_ACT_APPLY:
   1624 		EMIT(SPD_ATTR_TYPE, SPD_ACTTYPE_IPSEC);
   1625 		flags = 0;
   1626 		if (act->ipa_apply.ipp_use_ah)
   1627 			flags |= SPD_APPLY_AH;
   1628 		if (act->ipa_apply.ipp_use_esp)
   1629 			flags |= SPD_APPLY_ESP;
   1630 		if (act->ipa_apply.ipp_use_espa)
   1631 			flags |= SPD_APPLY_ESPA;
   1632 		if (act->ipa_apply.ipp_use_se)
   1633 			flags |= SPD_APPLY_SE;
   1634 		if (act->ipa_apply.ipp_use_unique)
   1635 			flags |= SPD_APPLY_UNIQUE;
   1636 		EMIT(SPD_ATTR_FLAGS, flags);
   1637 		if (flags & SPD_APPLY_AH) {
   1638 			EMIT(SPD_ATTR_AH_AUTH, act->ipa_apply.ipp_auth_alg);
   1639 			EMIT(SPD_ATTR_AH_MINBITS,
   1640 			    act->ipa_apply.ipp_ah_minbits);
   1641 			EMIT(SPD_ATTR_AH_MAXBITS,
   1642 			    act->ipa_apply.ipp_ah_maxbits);
   1643 		}
   1644 		if (flags & SPD_APPLY_ESP) {
   1645 			EMIT(SPD_ATTR_ESP_ENCR, act->ipa_apply.ipp_encr_alg);
   1646 			EMIT(SPD_ATTR_ENCR_MINBITS,
   1647 			    act->ipa_apply.ipp_espe_minbits);
   1648 			EMIT(SPD_ATTR_ENCR_MAXBITS,
   1649 			    act->ipa_apply.ipp_espe_maxbits);
   1650 			if (flags & SPD_APPLY_ESPA) {
   1651 				EMIT(SPD_ATTR_ESP_AUTH,
   1652 				    act->ipa_apply.ipp_esp_auth_alg);
   1653 				EMIT(SPD_ATTR_ESPA_MINBITS,
   1654 				    act->ipa_apply.ipp_espa_minbits);
   1655 				EMIT(SPD_ATTR_ESPA_MAXBITS,
   1656 				    act->ipa_apply.ipp_espa_maxbits);
   1657 			}
   1658 		}
   1659 		if (act->ipa_apply.ipp_km_proto != 0)
   1660 			EMIT(SPD_ATTR_KM_PROTO, act->ipa_apply.ipp_km_proto);
   1661 		if (act->ipa_apply.ipp_km_cookie != 0)
   1662 			EMIT(SPD_ATTR_KM_PROTO, act->ipa_apply.ipp_km_cookie);
   1663 		if (act->ipa_apply.ipp_replay_depth != 0)
   1664 			EMIT(SPD_ATTR_REPLAY_DEPTH,
   1665 			    act->ipa_apply.ipp_replay_depth);
   1666 		/* Add more here */
   1667 		break;
   1668 	}
   1669 
   1670 	return (offset);
   1671 }
   1672 
   1673 static uint_t
   1674 spdsock_encode_action_list(uint8_t *base, uint_t offset,
   1675     const ipsec_action_t *ap)
   1676 {
   1677 	struct spd_ext_actions *act;
   1678 	uint_t nact = 0;
   1679 	uint_t start = offset;
   1680 
   1681 	ASSERT(ALIGNED64(offset));
   1682 
   1683 	if (base != NULL) {
   1684 		act = (struct spd_ext_actions *)(base + offset);
   1685 		act->spd_actions_len = 0;
   1686 		act->spd_actions_exttype = SPD_EXT_ACTION;
   1687 		act->spd_actions_count = 0;
   1688 		act->spd_actions_reserved = 0;
   1689 	}
   1690 
   1691 	offset += sizeof (*act);
   1692 
   1693 	ASSERT(ALIGNED64(offset));
   1694 
   1695 	while (ap != NULL) {
   1696 		offset = spdsock_encode_action(base, offset, ap);
   1697 		ap = ap->ipa_next;
   1698 		nact++;
   1699 		if (ap != NULL) {
   1700 			EMIT(SPD_ATTR_NEXT, 0);
   1701 		}
   1702 	}
   1703 	EMIT(SPD_ATTR_END, 0);
   1704 
   1705 	ASSERT(ALIGNED64(offset));
   1706 
   1707 	if (base != NULL) {
   1708 		act->spd_actions_count = nact;
   1709 		act->spd_actions_len = SPD_8TO64(offset - start);
   1710 	}
   1711 
   1712 	return (offset);
   1713 }
   1714 
   1715 #undef EMIT
   1716 
   1717 /* ARGSUSED */
   1718 static uint_t
   1719 spdsock_rule_flags(uint_t dir, uint_t af)
   1720 {
   1721 	uint_t flags = 0;
   1722 
   1723 	if (dir == IPSEC_TYPE_INBOUND)
   1724 		flags |= SPD_RULE_FLAG_INBOUND;
   1725 	if (dir == IPSEC_TYPE_OUTBOUND)
   1726 		flags |= SPD_RULE_FLAG_OUTBOUND;
   1727 
   1728 	return (flags);
   1729 }
   1730 
   1731 
   1732 static uint_t
   1733 spdsock_encode_rule_head(uint8_t *base, uint_t offset, spd_msg_t *req,
   1734     const ipsec_policy_t *rule, uint_t dir, uint_t af, char *name,
   1735     boolean_t tunnel)
   1736 {
   1737 	struct spd_msg *spmsg;
   1738 	struct spd_rule *spr;
   1739 	spd_if_t *sid;
   1740 
   1741 	uint_t start = offset;
   1742 
   1743 	ASSERT(ALIGNED64(offset));
   1744 
   1745 	if (base != NULL) {
   1746 		spmsg = (struct spd_msg *)(base + offset);
   1747 		bzero(spmsg, sizeof (*spmsg));
   1748 		spmsg->spd_msg_version = PF_POLICY_V1;
   1749 		spmsg->spd_msg_type = SPD_DUMP;
   1750 		spmsg->spd_msg_seq = req->spd_msg_seq;
   1751 		spmsg->spd_msg_pid = req->spd_msg_pid;
   1752 	}
   1753 	offset += sizeof (struct spd_msg);
   1754 
   1755 	ASSERT(ALIGNED64(offset));
   1756 
   1757 	if (base != NULL) {
   1758 		spr = (struct spd_rule *)(base + offset);
   1759 		spr->spd_rule_type = SPD_EXT_RULE;
   1760 		spr->spd_rule_priority = rule->ipsp_prio;
   1761 		spr->spd_rule_flags = spdsock_rule_flags(dir, af);
   1762 		if (tunnel)
   1763 			spr->spd_rule_flags |= SPD_RULE_FLAG_TUNNEL;
   1764 		spr->spd_rule_unused = 0;
   1765 		spr->spd_rule_len = SPD_8TO64(sizeof (*spr));
   1766 		spr->spd_rule_index = rule->ipsp_index;
   1767 	}
   1768 	offset += sizeof (struct spd_rule);
   1769 
   1770 	/*
   1771 	 * If we have an interface name (i.e. if this policy head came from
   1772 	 * a tunnel), add the SPD_EXT_TUN_NAME extension.
   1773 	 */
   1774 	if (name != NULL) {
   1775 
   1776 		ASSERT(ALIGNED64(offset));
   1777 
   1778 		if (base != NULL) {
   1779 			sid = (spd_if_t *)(base + offset);
   1780 			sid->spd_if_exttype = SPD_EXT_TUN_NAME;
   1781 			sid->spd_if_len = SPD_8TO64(sizeof (spd_if_t) +
   1782 			    roundup((strlen(name) - 4), 8));
   1783 			(void) strlcpy((char *)sid->spd_if_name, name,
   1784 			    LIFNAMSIZ);
   1785 		}
   1786 
   1787 		offset += sizeof (spd_if_t) + roundup((strlen(name) - 4), 8);
   1788 	}
   1789 
   1790 	offset = spdsock_encode_sel(base, offset, rule->ipsp_sel);
   1791 	offset = spdsock_encode_action_list(base, offset, rule->ipsp_act);
   1792 
   1793 	ASSERT(ALIGNED64(offset));
   1794 
   1795 	if (base != NULL) {
   1796 		spmsg->spd_msg_len = SPD_8TO64(offset - start);
   1797 	}
   1798 	return (offset);
   1799 }
   1800 
   1801 /* ARGSUSED */
   1802 static mblk_t *
   1803 spdsock_encode_rule(mblk_t *req, const ipsec_policy_t *rule,
   1804     uint_t dir, uint_t af, char *name, boolean_t tunnel)
   1805 {
   1806 	mblk_t *m;
   1807 	uint_t len;
   1808 	spd_msg_t *mreq = (spd_msg_t *)req->b_rptr;
   1809 
   1810 	/*
   1811 	 * Figure out how much space we'll need.
   1812 	 */
   1813 	len = spdsock_encode_rule_head(NULL, 0, mreq, rule, dir, af, name,
   1814 	    tunnel);
   1815 
   1816 	/*
   1817 	 * Allocate mblk.
   1818 	 */
   1819 	m = allocb(len, BPRI_HI);
   1820 	if (m == NULL)
   1821 		return (NULL);
   1822 
   1823 	/*
   1824 	 * Fill it in..
   1825 	 */
   1826 	m->b_wptr = m->b_rptr + len;
   1827 	bzero(m->b_rptr, len);
   1828 	(void) spdsock_encode_rule_head(m->b_rptr, 0, mreq, rule, dir, af,
   1829 	    name, tunnel);
   1830 	return (m);
   1831 }
   1832 
   1833 static ipsec_policy_t *
   1834 spdsock_dump_next_in_chain(spdsock_t *ss, ipsec_policy_head_t *iph,
   1835     ipsec_policy_t *cur)
   1836 {
   1837 	ASSERT(RW_READ_HELD(&iph->iph_lock));
   1838 
   1839 	ss->spdsock_dump_count++;
   1840 	ss->spdsock_dump_cur_rule = cur->ipsp_hash.hash_next;
   1841 	return (cur);
   1842 }
   1843 
   1844 static ipsec_policy_t *
   1845 spdsock_dump_next_rule(spdsock_t *ss, ipsec_policy_head_t *iph)
   1846 {
   1847 	ipsec_policy_t *cur;
   1848 	ipsec_policy_root_t *ipr;
   1849 	int chain, nchains, type, af;
   1850 
   1851 	ASSERT(RW_READ_HELD(&iph->iph_lock));
   1852 
   1853 	cur = ss->spdsock_dump_cur_rule;
   1854 
   1855 	if (cur != NULL)
   1856 		return (spdsock_dump_next_in_chain(ss, iph, cur));
   1857 
   1858 	type = ss->spdsock_dump_cur_type;
   1859 
   1860 next:
   1861 	chain = ss->spdsock_dump_cur_chain;
   1862 	ipr = &iph->iph_root[type];
   1863 	nchains = ipr->ipr_nchains;
   1864 
   1865 	while (chain < nchains) {
   1866 		cur = ipr->ipr_hash[chain].hash_head;
   1867 		chain++;
   1868 		if (cur != NULL) {
   1869 			ss->spdsock_dump_cur_chain = chain;
   1870 			return (spdsock_dump_next_in_chain(ss, iph, cur));
   1871 		}
   1872 	}
   1873 	ss->spdsock_dump_cur_chain = nchains;
   1874 
   1875 	af = ss->spdsock_dump_cur_af;
   1876 	while (af < IPSEC_NAF) {
   1877 		cur = ipr->ipr_nonhash[af];
   1878 		af++;
   1879 		if (cur != NULL) {
   1880 			ss->spdsock_dump_cur_af = af;
   1881 			return (spdsock_dump_next_in_chain(ss, iph, cur));
   1882 		}
   1883 	}
   1884 
   1885 	type++;
   1886 	if (type >= IPSEC_NTYPES)
   1887 		return (NULL);
   1888 
   1889 	ss->spdsock_dump_cur_chain = 0;
   1890 	ss->spdsock_dump_cur_type = type;
   1891 	ss->spdsock_dump_cur_af = IPSEC_AF_V4;
   1892 	goto next;
   1893 
   1894 }
   1895 
   1896 /*
   1897  * If we're done with one policy head, but have more to go, we iterate through
   1898  * another IPsec tunnel policy head (itp).  Return NULL if it is an error
   1899  * worthy of returning EAGAIN via PF_POLICY.
   1900  */
   1901 static ipsec_tun_pol_t *
   1902 spdsock_dump_iterate_next_tunnel(spdsock_t *ss, ipsec_stack_t *ipss)
   1903 {
   1904 	ipsec_tun_pol_t *itp;
   1905 
   1906 	ASSERT(RW_READ_HELD(&ipss->ipsec_tunnel_policy_lock));
   1907 	if (ipss->ipsec_tunnel_policy_gen > ss->spdsock_dump_tun_gen) {
   1908 		/* Oops, state of the tunnel polheads changed. */
   1909 		itp = NULL;
   1910 	} else if (ss->spdsock_itp == NULL) {
   1911 		/* Just finished global, find first node. */
   1912 		itp = avl_first(&ipss->ipsec_tunnel_policies);
   1913 	} else {
   1914 		/* We just finished current polhead, find the next one. */
   1915 		itp = AVL_NEXT(&ipss->ipsec_tunnel_policies, ss->spdsock_itp);
   1916 	}
   1917 	if (itp != NULL) {
   1918 		ITP_REFHOLD(itp);
   1919 	}
   1920 	if (ss->spdsock_itp != NULL) {
   1921 		ITP_REFRELE(ss->spdsock_itp, ipss->ipsec_netstack);
   1922 	}
   1923 	ss->spdsock_itp = itp;
   1924 	return (itp);
   1925 }
   1926 
   1927 static mblk_t *
   1928 spdsock_dump_next_record(spdsock_t *ss)
   1929 {
   1930 	ipsec_policy_head_t *iph;
   1931 	ipsec_policy_t *rule;
   1932 	mblk_t *m;
   1933 	ipsec_tun_pol_t *itp;
   1934 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
   1935 	ipsec_stack_t *ipss = ns->netstack_ipsec;
   1936 
   1937 	iph = ss->spdsock_dump_head;
   1938 
   1939 	ASSERT(iph != NULL);
   1940 
   1941 	rw_enter(&iph->iph_lock, RW_READER);
   1942 
   1943 	if (iph->iph_gen != ss->spdsock_dump_gen) {
   1944 		rw_exit(&iph->iph_lock);
   1945 		return (spdsock_dump_finish(ss, EAGAIN));
   1946 	}
   1947 
   1948 	while ((rule = spdsock_dump_next_rule(ss, iph)) == NULL) {
   1949 		rw_exit(&iph->iph_lock);
   1950 		if (--(ss->spdsock_dump_remaining_polheads) == 0)
   1951 			return (spdsock_dump_finish(ss, 0));
   1952 
   1953 
   1954 		/*
   1955 		 * If we reach here, we have more policy heads (tunnel
   1956 		 * entries) to dump.  Let's reset to a new policy head
   1957 		 * and get some more rules.
   1958 		 *
   1959 		 * An empty policy head will have spdsock_dump_next_rule()
   1960 		 * return NULL, and we loop (while dropping the number of
   1961 		 * remaining polheads).  If we loop to 0, we finish.  We
   1962 		 * keep looping until we hit 0 or until we have a rule to
   1963 		 * encode.
   1964 		 *
   1965 		 * NOTE:  No need for ITP_REF*() macros here as we're only
   1966 		 * going after and refholding the policy head itself.
   1967 		 */
   1968 		rw_enter(&ipss->ipsec_tunnel_policy_lock, RW_READER);
   1969 		itp = spdsock_dump_iterate_next_tunnel(ss, ipss);
   1970 		if (itp == NULL) {
   1971 			rw_exit(&ipss->ipsec_tunnel_policy_lock);
   1972 			return (spdsock_dump_finish(ss, EAGAIN));
   1973 		}
   1974 
   1975 		/* Reset other spdsock_dump thingies. */
   1976 		IPPH_REFRELE(ss->spdsock_dump_head, ns);
   1977 		if (ss->spdsock_dump_active) {
   1978 			ss->spdsock_dump_tunnel =
   1979 			    itp->itp_flags & ITPF_P_TUNNEL;
   1980 			iph = itp->itp_policy;
   1981 		} else {
   1982 			ss->spdsock_dump_tunnel =
   1983 			    itp->itp_flags & ITPF_I_TUNNEL;
   1984 			iph = itp->itp_inactive;
   1985 		}
   1986 		IPPH_REFHOLD(iph);
   1987 		rw_exit(&ipss->ipsec_tunnel_policy_lock);
   1988 
   1989 		rw_enter(&iph->iph_lock, RW_READER);
   1990 		RESET_SPDSOCK_DUMP_POLHEAD(ss, iph);
   1991 	}
   1992 
   1993 	m = spdsock_encode_rule(ss->spdsock_dump_req, rule,
   1994 	    ss->spdsock_dump_cur_type, ss->spdsock_dump_cur_af,
   1995 	    (ss->spdsock_itp == NULL) ? NULL : ss->spdsock_itp->itp_name,
   1996 	    ss->spdsock_dump_tunnel);
   1997 	rw_exit(&iph->iph_lock);
   1998 
   1999 	if (m == NULL)
   2000 		return (spdsock_dump_finish(ss, ENOMEM));
   2001 	return (m);
   2002 }
   2003 
   2004 /*
   2005  * Dump records until we run into flow-control back-pressure.
   2006  */
   2007 static void
   2008 spdsock_dump_some(queue_t *q, spdsock_t *ss)
   2009 {
   2010 	mblk_t *m, *dataind;
   2011 
   2012 	while ((ss->spdsock_dump_req != NULL) && canputnext(q)) {
   2013 		m = spdsock_dump_next_record(ss);
   2014 		if (m == NULL)
   2015 			return;
   2016 		dataind = allocb(sizeof (struct T_data_req), BPRI_HI);
   2017 		if (dataind == NULL) {
   2018 			freemsg(m);
   2019 			return;
   2020 		}
   2021 		dataind->b_cont = m;
   2022 		dataind->b_wptr += sizeof (struct T_data_req);
   2023 		((struct T_data_ind *)dataind->b_rptr)->PRIM_type = T_DATA_IND;
   2024 		((struct T_data_ind *)dataind->b_rptr)->MORE_flag = 0;
   2025 		dataind->b_datap->db_type = M_PROTO;
   2026 		putnext(q, dataind);
   2027 	}
   2028 }
   2029 
   2030 /*
   2031  * Start dumping.
   2032  * Format a start-of-dump record, and set up the stream and kick the rsrv
   2033  * procedure to continue the job..
   2034  */
   2035 /* ARGSUSED */
   2036 static void
   2037 spdsock_dump(queue_t *q, ipsec_policy_head_t *iph, mblk_t *mp)
   2038 {
   2039 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   2040 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
   2041 	ipsec_stack_t *ipss = ns->netstack_ipsec;
   2042 	mblk_t *mr;
   2043 
   2044 	/* spdsock_open() already set spdsock_itp to NULL. */
   2045 	if (iph == ALL_ACTIVE_POLHEADS || iph == ALL_INACTIVE_POLHEADS) {
   2046 		rw_enter(&ipss->ipsec_tunnel_policy_lock, RW_READER);
   2047 		ss->spdsock_dump_remaining_polheads = 1 +
   2048 		    avl_numnodes(&ipss->ipsec_tunnel_policies);
   2049 		ss->spdsock_dump_tun_gen = ipss->ipsec_tunnel_policy_gen;
   2050 		rw_exit(&ipss->ipsec_tunnel_policy_lock);
   2051 		if (iph == ALL_ACTIVE_POLHEADS) {
   2052 			iph = ipsec_system_policy(ns);
   2053 			ss->spdsock_dump_active = B_TRUE;
   2054 		} else {
   2055 			iph = ipsec_inactive_policy(ns);
   2056 			ss->spdsock_dump_active = B_FALSE;
   2057 		}
   2058 		ASSERT(ss->spdsock_itp == NULL);
   2059 	} else {
   2060 		ss->spdsock_dump_remaining_polheads = 1;
   2061 	}
   2062 
   2063 	rw_enter(&iph->iph_lock, RW_READER);
   2064 
   2065 	mr = spdsock_dump_ruleset(mp, iph, 0, 0);
   2066 
   2067 	if (!mr) {
   2068 		rw_exit(&iph->iph_lock);
   2069 		spdsock_error(q, mp, ENOMEM, 0);
   2070 		return;
   2071 	}
   2072 
   2073 	ss->spdsock_dump_req = mp;
   2074 	RESET_SPDSOCK_DUMP_POLHEAD(ss, iph);
   2075 
   2076 	rw_exit(&iph->iph_lock);
   2077 
   2078 	qreply(q, mr);
   2079 	qenable(OTHERQ(q));
   2080 }
   2081 
   2082 /* Do NOT consume a reference to ITP. */
   2083 void
   2084 spdsock_clone_node(ipsec_tun_pol_t *itp, void *ep, netstack_t *ns)
   2085 {
   2086 	int *errptr = (int *)ep;
   2087 
   2088 	if (*errptr != 0)
   2089 		return;	/* We've failed already for some reason. */
   2090 	mutex_enter(&itp->itp_lock);
   2091 	ITPF_CLONE(itp->itp_flags);
   2092 	*errptr = ipsec_copy_polhead(itp->itp_policy, itp->itp_inactive, ns);
   2093 	mutex_exit(&itp->itp_lock);
   2094 }
   2095 
   2096 void
   2097 spdsock_clone(queue_t *q, mblk_t *mp, spd_if_t *tunname)
   2098 {
   2099 	int error;
   2100 	char *tname;
   2101 	ipsec_tun_pol_t *itp;
   2102 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   2103 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
   2104 
   2105 	if (tunname != NULL) {
   2106 		tname = (char *)tunname->spd_if_name;
   2107 		if (*tname == '\0') {
   2108 			error = ipsec_clone_system_policy(ns);
   2109 			if (audit_active) {
   2110 				boolean_t active;
   2111 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   2112 				cred_t *cr;
   2113 				pid_t cpid;
   2114 
   2115 				cr = msg_getcred(mp, &cpid);
   2116 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   2117 				audit_pf_policy(SPD_CLONE, cr, ns,
   2118 				    NULL, active, error, cpid);
   2119 			}
   2120 			if (error == 0) {
   2121 				itp_walk(spdsock_clone_node, &error, ns);
   2122 				if (audit_active) {
   2123 					boolean_t active;
   2124 					spd_msg_t *spmsg =
   2125 					    (spd_msg_t *)mp->b_rptr;
   2126 					cred_t *cr;
   2127 					pid_t cpid;
   2128 
   2129 					cr = msg_getcred(mp, &cpid);
   2130 					active = (spmsg->spd_msg_spdid ==
   2131 					    SPD_ACTIVE);
   2132 					audit_pf_policy(SPD_CLONE, cr,
   2133 					    ns, "all tunnels", active, 0,
   2134 					    cpid);
   2135 				}
   2136 			}
   2137 		} else {
   2138 			itp = get_tunnel_policy(tname, ns);
   2139 			if (itp == NULL) {
   2140 				spdsock_error(q, mp, ENOENT, 0);
   2141 				if (audit_active) {
   2142 					boolean_t active;
   2143 					spd_msg_t *spmsg =
   2144 					    (spd_msg_t *)mp->b_rptr;
   2145 					cred_t *cr;
   2146 					pid_t cpid;
   2147 
   2148 					cr = msg_getcred(mp, &cpid);
   2149 					active = (spmsg->spd_msg_spdid ==
   2150 					    SPD_ACTIVE);
   2151 					audit_pf_policy(SPD_CLONE, cr,
   2152 					    ns, NULL, active, ENOENT, cpid);
   2153 				}
   2154 				return;
   2155 			}
   2156 			spdsock_clone_node(itp, &error, NULL);
   2157 			if (audit_active) {
   2158 				boolean_t active;
   2159 				spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   2160 				cred_t *cr;
   2161 				pid_t cpid;
   2162 
   2163 				cr = msg_getcred(mp, &cpid);
   2164 				active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   2165 				audit_pf_policy(SPD_CLONE, cr, ns,
   2166 				    ITP_NAME(itp), active, error, cpid);
   2167 			}
   2168 			ITP_REFRELE(itp, ns);
   2169 		}
   2170 	} else {
   2171 		error = ipsec_clone_system_policy(ns);
   2172 		if (audit_active) {
   2173 			boolean_t active;
   2174 			spd_msg_t *spmsg = (spd_msg_t *)mp->b_rptr;
   2175 			cred_t *cr;
   2176 			pid_t cpid;
   2177 
   2178 			cr = msg_getcred(mp, &cpid);
   2179 			active = (spmsg->spd_msg_spdid == SPD_ACTIVE);
   2180 			audit_pf_policy(SPD_CLONE, cr, ns, NULL,
   2181 			    active, error, cpid);
   2182 		}
   2183 	}
   2184 
   2185 	if (error != 0)
   2186 		spdsock_error(q, mp, error, 0);
   2187 	else
   2188 		spd_echo(q, mp);
   2189 }
   2190 
   2191 /*
   2192  * Process a SPD_ALGLIST request. The caller expects separate alg entries
   2193  * for AH authentication, ESP authentication, and ESP encryption.
   2194  * The same distinction is then used when setting the min and max key
   2195  * sizes when defining policies.
   2196  */
   2197 
   2198 #define	SPDSOCK_AH_AUTH		0
   2199 #define	SPDSOCK_ESP_AUTH	1
   2200 #define	SPDSOCK_ESP_ENCR	2
   2201 #define	SPDSOCK_NTYPES		3
   2202 
   2203 static const uint_t algattr[SPDSOCK_NTYPES] = {
   2204 	SPD_ATTR_AH_AUTH,
   2205 	SPD_ATTR_ESP_AUTH,
   2206 	SPD_ATTR_ESP_ENCR
   2207 };
   2208 static const uint_t minbitsattr[SPDSOCK_NTYPES] = {
   2209 	SPD_ATTR_AH_MINBITS,
   2210 	SPD_ATTR_ESPA_MINBITS,
   2211 	SPD_ATTR_ENCR_MINBITS
   2212 };
   2213 static const uint_t maxbitsattr[SPDSOCK_NTYPES] = {
   2214 	SPD_ATTR_AH_MAXBITS,
   2215 	SPD_ATTR_ESPA_MAXBITS,
   2216 	SPD_ATTR_ENCR_MAXBITS
   2217 };
   2218 static const uint_t defbitsattr[SPDSOCK_NTYPES] = {
   2219 	SPD_ATTR_AH_DEFBITS,
   2220 	SPD_ATTR_ESPA_DEFBITS,
   2221 	SPD_ATTR_ENCR_DEFBITS
   2222 };
   2223 static const uint_t incrbitsattr[SPDSOCK_NTYPES] = {
   2224 	SPD_ATTR_AH_INCRBITS,
   2225 	SPD_ATTR_ESPA_INCRBITS,
   2226 	SPD_ATTR_ENCR_INCRBITS
   2227 };
   2228 
   2229 #define	ATTRPERALG	6	/* fixed attributes per algs */
   2230 
   2231 void
   2232 spdsock_alglist(queue_t *q, mblk_t *mp)
   2233 {
   2234 	uint_t algtype;
   2235 	uint_t algidx;
   2236 	uint_t algcount;
   2237 	uint_t size;
   2238 	mblk_t *m;
   2239 	uint8_t *cur;
   2240 	spd_msg_t *msg;
   2241 	struct spd_ext_actions *act;
   2242 	struct spd_attribute *attr;
   2243 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   2244 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
   2245 
   2246 	mutex_enter(&ipss->ipsec_alg_lock);
   2247 	/*
   2248 	 * The SPD client expects to receive separate entries for
   2249 	 * AH authentication and ESP authentication supported algorithms.
   2250 	 *
   2251 	 * Don't return the "any" algorithms, if defined, as no
   2252 	 * kernel policies can be set for these algorithms.
   2253 	 */
   2254 	algcount = 2 * ipss->ipsec_nalgs[IPSEC_ALG_AUTH] +
   2255 	    ipss->ipsec_nalgs[IPSEC_ALG_ENCR];
   2256 
   2257 	if (ipss->ipsec_alglists[IPSEC_ALG_AUTH][SADB_AALG_NONE] != NULL)
   2258 		algcount--;
   2259 	if (ipss->ipsec_alglists[IPSEC_ALG_ENCR][SADB_EALG_NONE] != NULL)
   2260 		algcount--;
   2261 
   2262 	/*
   2263 	 * For each algorithm, we encode:
   2264 	 * ALG / MINBITS / MAXBITS / DEFBITS / INCRBITS / {END, NEXT}
   2265 	 */
   2266 
   2267 	size = sizeof (spd_msg_t) + sizeof (struct spd_ext_actions) +
   2268 	    ATTRPERALG * sizeof (struct spd_attribute) * algcount;
   2269 
   2270 	ASSERT(ALIGNED64(size));
   2271 
   2272 	m = allocb(size, BPRI_HI);
   2273 	if (m == NULL) {
   2274 		mutex_exit(&ipss->ipsec_alg_lock);
   2275 		spdsock_error(q, mp, ENOMEM, 0);
   2276 		return;
   2277 	}
   2278 
   2279 	m->b_wptr = m->b_rptr + size;
   2280 	cur = m->b_rptr;
   2281 
   2282 	msg = (spd_msg_t *)cur;
   2283 	bcopy(mp->b_rptr, cur, sizeof (*msg));
   2284 
   2285 	msg->spd_msg_len = SPD_8TO64(size);
   2286 	msg->spd_msg_errno = 0;
   2287 	msg->spd_msg_diagnostic = 0;
   2288 
   2289 	cur += sizeof (*msg);
   2290 
   2291 	act = (struct spd_ext_actions *)cur;
   2292 	cur += sizeof (*act);
   2293 
   2294 	act->spd_actions_len = SPD_8TO64(size - sizeof (spd_msg_t));
   2295 	act->spd_actions_exttype = SPD_EXT_ACTION;
   2296 	act->spd_actions_count = algcount;
   2297 	act->spd_actions_reserved = 0;
   2298 
   2299 	attr = (struct spd_attribute *)cur;
   2300 
   2301 #define	EMIT(tag, value) {					\
   2302 		attr->spd_attr_tag = (tag); 			\
   2303 		attr->spd_attr_value = (value); 		\
   2304 		attr++;			  			\
   2305 	}
   2306 
   2307 	/*
   2308 	 * If you change the number of EMIT's here, change
   2309 	 * ATTRPERALG above to match
   2310 	 */
   2311 #define	EMITALGATTRS(_type) {					\
   2312 		EMIT(algattr[_type], algid); 		/* 1 */	\
   2313 		EMIT(minbitsattr[_type], minbits);	/* 2 */	\
   2314 		EMIT(maxbitsattr[_type], maxbits);	/* 3 */	\
   2315 		EMIT(defbitsattr[_type], defbits);	/* 4 */	\
   2316 		EMIT(incrbitsattr[_type], incr);	/* 5 */	\
   2317 		EMIT(SPD_ATTR_NEXT, 0);			/* 6 */	\
   2318 	}
   2319 
   2320 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
   2321 		for (algidx = 0; algidx < ipss->ipsec_nalgs[algtype];
   2322 		    algidx++) {
   2323 			int algid = ipss->ipsec_sortlist[algtype][algidx];
   2324 			ipsec_alginfo_t *alg =
   2325 			    ipss->ipsec_alglists[algtype][algid];
   2326 			uint_t minbits = alg->alg_minbits;
   2327 			uint_t maxbits = alg->alg_maxbits;
   2328 			uint_t defbits = alg->alg_default_bits;
   2329 			uint_t incr = alg->alg_increment;
   2330 
   2331 			if (algtype == IPSEC_ALG_AUTH) {
   2332 				if (algid == SADB_AALG_NONE)
   2333 					continue;
   2334 				EMITALGATTRS(SPDSOCK_AH_AUTH);
   2335 				EMITALGATTRS(SPDSOCK_ESP_AUTH);
   2336 			} else {
   2337 				if (algid == SADB_EALG_NONE)
   2338 					continue;
   2339 				ASSERT(algtype == IPSEC_ALG_ENCR);
   2340 				EMITALGATTRS(SPDSOCK_ESP_ENCR);
   2341 			}
   2342 		}
   2343 	}
   2344 
   2345 	mutex_exit(&ipss->ipsec_alg_lock);
   2346 
   2347 #undef EMITALGATTRS
   2348 #undef EMIT
   2349 #undef ATTRPERALG
   2350 
   2351 	attr--;
   2352 	attr->spd_attr_tag = SPD_ATTR_END;
   2353 
   2354 	freemsg(mp);
   2355 	qreply(q, m);
   2356 }
   2357 
   2358 /*
   2359  * Process a SPD_DUMPALGS request.
   2360  */
   2361 
   2362 #define	ATTRPERALG	9	/* fixed attributes per algs */
   2363 
   2364 void
   2365 spdsock_dumpalgs(queue_t *q, mblk_t *mp)
   2366 {
   2367 	uint_t algtype;
   2368 	uint_t algidx;
   2369 	uint_t size;
   2370 	mblk_t *m;
   2371 	uint8_t *cur;
   2372 	spd_msg_t *msg;
   2373 	struct spd_ext_actions *act;
   2374 	struct spd_attribute *attr;
   2375 	ipsec_alginfo_t *alg;
   2376 	uint_t algid;
   2377 	uint_t i;
   2378 	uint_t alg_size;
   2379 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   2380 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
   2381 
   2382 	mutex_enter(&ipss->ipsec_alg_lock);
   2383 
   2384 	/*
   2385 	 * For each algorithm, we encode:
   2386 	 * ALG / MINBITS / MAXBITS / DEFBITS / INCRBITS / {END, NEXT}
   2387 	 *
   2388 	 * ALG_ID / ALG_PROTO / ALG_INCRBITS / ALG_NKEYSIZES / ALG_KEYSIZE*
   2389 	 * ALG_NBLOCKSIZES / ALG_BLOCKSIZE* / ALG_NPARAMS / ALG_PARAMS* /
   2390 	 * ALG_MECHNAME / ALG_FLAGS / {END, NEXT}
   2391 	 */
   2392 
   2393 	/*
   2394 	 * Compute the size of the SPD message.
   2395 	 */
   2396 	size = sizeof (spd_msg_t) + sizeof (struct spd_ext_actions);
   2397 
   2398 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
   2399 		for (algidx = 0; algidx < ipss->ipsec_nalgs[algtype];
   2400 		    algidx++) {
   2401 			algid = ipss->ipsec_sortlist[algtype][algidx];
   2402 			alg = ipss->ipsec_alglists[algtype][algid];
   2403 			alg_size = sizeof (struct spd_attribute) *
   2404 			    (ATTRPERALG + alg->alg_nkey_sizes +
   2405 			    alg->alg_nblock_sizes + alg->alg_nparams) +
   2406 			    CRYPTO_MAX_MECH_NAME;
   2407 			size += alg_size;
   2408 		}
   2409 	}
   2410 
   2411 	ASSERT(ALIGNED64(size));
   2412 
   2413 	m = allocb(size, BPRI_HI);
   2414 	if (m == NULL) {
   2415 		mutex_exit(&ipss->ipsec_alg_lock);
   2416 		spdsock_error(q, mp, ENOMEM, 0);
   2417 		return;
   2418 	}
   2419 
   2420 	m->b_wptr = m->b_rptr + size;
   2421 	cur = m->b_rptr;
   2422 
   2423 	msg = (spd_msg_t *)cur;
   2424 	bcopy(mp->b_rptr, cur, sizeof (*msg));
   2425 
   2426 	msg->spd_msg_len = SPD_8TO64(size);
   2427 	msg->spd_msg_errno = 0;
   2428 	msg->spd_msg_type = SPD_ALGLIST;
   2429 
   2430 	msg->spd_msg_diagnostic = 0;
   2431 
   2432 	cur += sizeof (*msg);
   2433 
   2434 	act = (struct spd_ext_actions *)cur;
   2435 	cur += sizeof (*act);
   2436 
   2437 	act->spd_actions_len = SPD_8TO64(size - sizeof (spd_msg_t));
   2438 	act->spd_actions_exttype = SPD_EXT_ACTION;
   2439 	act->spd_actions_count = ipss->ipsec_nalgs[IPSEC_ALG_AUTH] +
   2440 	    ipss->ipsec_nalgs[IPSEC_ALG_ENCR];
   2441 	act->spd_actions_reserved = 0;
   2442 
   2443 	/*
   2444 	 * If there aren't any algorithms registered, return an empty message.
   2445 	 * spdsock_get_ext() knows how to deal with this.
   2446 	 */
   2447 	if (act->spd_actions_count == 0) {
   2448 		act->spd_actions_len = 0;
   2449 		mutex_exit(&ipss->ipsec_alg_lock);
   2450 		goto error;
   2451 	}
   2452 
   2453 	attr = (struct spd_attribute *)cur;
   2454 
   2455 #define	EMIT(tag, value) {					\
   2456 		attr->spd_attr_tag = (tag); 			\
   2457 		attr->spd_attr_value = (value); 		\
   2458 		attr++;			  			\
   2459 	}
   2460 
   2461 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
   2462 		for (algidx = 0; algidx < ipss->ipsec_nalgs[algtype];
   2463 		    algidx++) {
   2464 
   2465 			algid = ipss->ipsec_sortlist[algtype][algidx];
   2466 			alg = ipss->ipsec_alglists[algtype][algid];
   2467 
   2468 			/*
   2469 			 * If you change the number of EMIT's here, change
   2470 			 * ATTRPERALG above to match
   2471 			 */
   2472 			EMIT(SPD_ATTR_ALG_ID, algid);
   2473 			EMIT(SPD_ATTR_ALG_PROTO, algproto[algtype]);
   2474 			EMIT(SPD_ATTR_ALG_INCRBITS, alg->alg_increment);
   2475 			EMIT(SPD_ATTR_ALG_NKEYSIZES, alg->alg_nkey_sizes);
   2476 			for (i = 0; i < alg->alg_nkey_sizes; i++)
   2477 				EMIT(SPD_ATTR_ALG_KEYSIZE,
   2478 				    alg->alg_key_sizes[i]);
   2479 
   2480 			EMIT(SPD_ATTR_ALG_NBLOCKSIZES, alg->alg_nblock_sizes);
   2481 			for (i = 0; i < alg->alg_nblock_sizes; i++)
   2482 				EMIT(SPD_ATTR_ALG_BLOCKSIZE,
   2483 				    alg->alg_block_sizes[i]);
   2484 
   2485 			EMIT(SPD_ATTR_ALG_NPARAMS, alg->alg_nparams);
   2486 			for (i = 0; i < alg->alg_nparams; i++)
   2487 				EMIT(SPD_ATTR_ALG_PARAMS,
   2488 				    alg->alg_params[i]);
   2489 
   2490 			EMIT(SPD_ATTR_ALG_FLAGS, alg->alg_flags);
   2491 
   2492 			EMIT(SPD_ATTR_ALG_MECHNAME, CRYPTO_MAX_MECH_NAME);
   2493 			bcopy(alg->alg_mech_name, attr, CRYPTO_MAX_MECH_NAME);
   2494 			attr = (struct spd_attribute *)((char *)attr +
   2495 			    CRYPTO_MAX_MECH_NAME);
   2496 
   2497 			EMIT(SPD_ATTR_NEXT, 0);
   2498 		}
   2499 	}
   2500 
   2501 	mutex_exit(&ipss->ipsec_alg_lock);
   2502 
   2503 #undef EMITALGATTRS
   2504 #undef EMIT
   2505 #undef ATTRPERALG
   2506 
   2507 	attr--;
   2508 	attr->spd_attr_tag = SPD_ATTR_END;
   2509 
   2510 error:
   2511 	freemsg(mp);
   2512 	qreply(q, m);
   2513 }
   2514 
   2515 /*
   2516  * Do the actual work of processing an SPD_UPDATEALGS request. Can
   2517  * be invoked either once IPsec is loaded on a cached request, or
   2518  * when a request is received while IPsec is loaded.
   2519  */
   2520 static int
   2521 spdsock_do_updatealg(spd_ext_t *extv[], spd_stack_t *spds)
   2522 {
   2523 	struct spd_ext_actions *actp;
   2524 	struct spd_attribute *attr, *endattr;
   2525 	uint64_t *start, *end;
   2526 	ipsec_alginfo_t *alg = NULL;
   2527 	ipsec_algtype_t alg_type = 0;
   2528 	boolean_t skip_alg = B_TRUE, doing_proto = B_FALSE;
   2529 	uint_t i, cur_key, cur_block, algid;
   2530 	int diag = -1;
   2531 
   2532 	ASSERT(MUTEX_HELD(&spds->spds_alg_lock));
   2533 
   2534 	/* parse the message, building the list of algorithms */
   2535 
   2536 	actp = (struct spd_ext_actions *)extv[SPD_EXT_ACTION];
   2537 	if (actp == NULL)
   2538 		return (SPD_DIAGNOSTIC_NO_ACTION_EXT);
   2539 
   2540 	start = (uint64_t *)actp;
   2541 	end = (start + actp->spd_actions_len);
   2542 	endattr = (struct spd_attribute *)end;
   2543 	attr = (struct spd_attribute *)&actp[1];
   2544 
   2545 	bzero(spds->spds_algs, IPSEC_NALGTYPES * IPSEC_MAX_ALGS *
   2546 	    sizeof (ipsec_alginfo_t *));
   2547 
   2548 	alg = kmem_zalloc(sizeof (*alg), KM_SLEEP);
   2549 
   2550 #define	ALG_KEY_SIZES(a)   (((a)->alg_nkey_sizes + 1) * sizeof (uint16_t))
   2551 #define	ALG_BLOCK_SIZES(a) (((a)->alg_nblock_sizes + 1) * sizeof (uint16_t))
   2552 #define	ALG_PARAM_SIZES(a) (((a)->alg_nparams + 1) * sizeof (uint16_t))
   2553 
   2554 	while (attr < endattr) {
   2555 		switch (attr->spd_attr_tag) {
   2556 		case SPD_ATTR_NOP:
   2557 		case SPD_ATTR_EMPTY:
   2558 			break;
   2559 		case SPD_ATTR_END:
   2560 			attr = endattr;
   2561 			/* FALLTHRU */
   2562 		case SPD_ATTR_NEXT:
   2563 			if (doing_proto) {
   2564 				doing_proto = B_FALSE;
   2565 				break;
   2566 			}
   2567 			if (skip_alg) {
   2568 				ipsec_alg_free(alg);
   2569 			} else {
   2570 				ipsec_alg_free(
   2571 				    spds->spds_algs[alg_type][alg->alg_id]);
   2572 				spds->spds_algs[alg_type][alg->alg_id] =
   2573 				    alg;
   2574 			}
   2575 			alg = kmem_zalloc(sizeof (*alg), KM_SLEEP);
   2576 			break;
   2577 
   2578 		case SPD_ATTR_ALG_ID:
   2579 			if (attr->spd_attr_value >= IPSEC_MAX_ALGS) {
   2580 				ss1dbg(spds, ("spdsock_do_updatealg: "
   2581 				    "invalid alg id %d\n",
   2582 				    attr->spd_attr_value));
   2583 				diag = SPD_DIAGNOSTIC_ALG_ID_RANGE;
   2584 				goto bail;
   2585 			}
   2586 			alg->alg_id = attr->spd_attr_value;
   2587 			break;
   2588 
   2589 		case SPD_ATTR_ALG_PROTO:
   2590 			/* find the alg type */
   2591 			for (i = 0; i < NALGPROTOS; i++)
   2592 				if (algproto[i] == attr->spd_attr_value)
   2593 					break;
   2594 			skip_alg = (i == NALGPROTOS);
   2595 			if (!skip_alg)
   2596 				alg_type = i;
   2597 			break;
   2598 
   2599 		case SPD_ATTR_ALG_INCRBITS:
   2600 			alg->alg_increment = attr->spd_attr_value;
   2601 			break;
   2602 
   2603 		case SPD_ATTR_ALG_NKEYSIZES:
   2604 			if (alg->alg_key_sizes != NULL) {
   2605 				kmem_free(alg->alg_key_sizes,
   2606 				    ALG_KEY_SIZES(alg));
   2607 			}
   2608 			alg->alg_nkey_sizes = attr->spd_attr_value;
   2609 			/*
   2610 			 * Allocate room for the trailing zero key size
   2611 			 * value as well.
   2612 			 */
   2613 			alg->alg_key_sizes = kmem_zalloc(ALG_KEY_SIZES(alg),
   2614 			    KM_SLEEP);
   2615 			cur_key = 0;
   2616 			break;
   2617 
   2618 		case SPD_ATTR_ALG_KEYSIZE:
   2619 			if (alg->alg_key_sizes == NULL ||
   2620 			    cur_key >= alg->alg_nkey_sizes) {
   2621 				ss1dbg(spds, ("spdsock_do_updatealg: "
   2622 				    "too many key sizes\n"));
   2623 				diag = SPD_DIAGNOSTIC_ALG_NUM_KEY_SIZES;
   2624 				goto bail;
   2625 			}
   2626 			alg->alg_key_sizes[cur_key++] = attr->spd_attr_value;
   2627 			break;
   2628 
   2629 		case SPD_ATTR_ALG_FLAGS:
   2630 			/*
   2631 			 * Flags (bit mask). The alg_flags element of
   2632 			 * ipsecalg_flags_t is only 8 bits wide. The
   2633 			 * user can set the VALID bit, but we will ignore it
   2634 			 * and make the decision is the algorithm is valid.
   2635 			 */
   2636 			alg->alg_flags |= (uint8_t)attr->spd_attr_value;
   2637 			break;
   2638 
   2639 		case SPD_ATTR_ALG_NBLOCKSIZES:
   2640 			if (alg->alg_block_sizes != NULL) {
   2641 				kmem_free(alg->alg_block_sizes,
   2642 				    ALG_BLOCK_SIZES(alg));
   2643 			}
   2644 			alg->alg_nblock_sizes = attr->spd_attr_value;
   2645 			/*
   2646 			 * Allocate room for the trailing zero block size
   2647 			 * value as well.
   2648 			 */
   2649 			alg->alg_block_sizes = kmem_zalloc(ALG_BLOCK_SIZES(alg),
   2650 			    KM_SLEEP);
   2651 			cur_block = 0;
   2652 			break;
   2653 
   2654 		case SPD_ATTR_ALG_BLOCKSIZE:
   2655 			if (alg->alg_block_sizes == NULL ||
   2656 			    cur_block >= alg->alg_nblock_sizes) {
   2657 				ss1dbg(spds, ("spdsock_do_updatealg: "
   2658 				    "too many block sizes\n"));
   2659 				diag = SPD_DIAGNOSTIC_ALG_NUM_BLOCK_SIZES;
   2660 				goto bail;
   2661 			}
   2662 			alg->alg_block_sizes[cur_block++] =
   2663 			    attr->spd_attr_value;
   2664 			break;
   2665 
   2666 		case SPD_ATTR_ALG_NPARAMS:
   2667 			if (alg->alg_params != NULL) {
   2668 				kmem_free(alg->alg_params,
   2669 				    ALG_PARAM_SIZES(alg));
   2670 			}
   2671 			alg->alg_nparams = attr->spd_attr_value;
   2672 			/*
   2673 			 * Allocate room for the trailing zero block size
   2674 			 * value as well.
   2675 			 */
   2676 			alg->alg_params = kmem_zalloc(ALG_PARAM_SIZES(alg),
   2677 			    KM_SLEEP);
   2678 			cur_block = 0;
   2679 			break;
   2680 
   2681 		case SPD_ATTR_ALG_PARAMS:
   2682 			if (alg->alg_params == NULL ||
   2683 			    cur_block >= alg->alg_nparams) {
   2684 				ss1dbg(spds, ("spdsock_do_updatealg: "
   2685 				    "too many params\n"));
   2686 				diag = SPD_DIAGNOSTIC_ALG_NUM_BLOCK_SIZES;
   2687 				goto bail;
   2688 			}
   2689 			/*
   2690 			 * Array contains: iv_len, icv_len, salt_len
   2691 			 * Any additional parameters are currently ignored.
   2692 			 */
   2693 			alg->alg_params[cur_block++] =
   2694 			    attr->spd_attr_value;
   2695 			break;
   2696 
   2697 		case SPD_ATTR_ALG_MECHNAME: {
   2698 			char *mech_name;
   2699 
   2700 			if (attr->spd_attr_value > CRYPTO_MAX_MECH_NAME) {
   2701 				ss1dbg(spds, ("spdsock_do_updatealg: "
   2702 				    "mech name too long\n"));
   2703 				diag = SPD_DIAGNOSTIC_ALG_MECH_NAME_LEN;
   2704 				goto bail;
   2705 			}
   2706 			mech_name = (char *)(attr + 1);
   2707 			bcopy(mech_name, alg->alg_mech_name,
   2708 			    attr->spd_attr_value);
   2709 			alg->alg_mech_name[CRYPTO_MAX_MECH_NAME-1] = '\0';
   2710 			attr = (struct spd_attribute *)((char *)attr +
   2711 			    attr->spd_attr_value);
   2712 			break;
   2713 		}
   2714 
   2715 		case SPD_ATTR_PROTO_ID:
   2716 			doing_proto = B_TRUE;
   2717 			for (i = 0; i < NALGPROTOS; i++) {
   2718 				if (algproto[i] == attr->spd_attr_value) {
   2719 					alg_type = i;
   2720 					break;
   2721 				}
   2722 			}
   2723 			break;
   2724 
   2725 		case SPD_ATTR_PROTO_EXEC_MODE:
   2726 			if (!doing_proto)
   2727 				break;
   2728 			for (i = 0; i < NEXECMODES; i++) {
   2729 				if (execmodes[i] == attr->spd_attr_value) {
   2730 					spds->spds_algs_exec_mode[alg_type] = i;
   2731 					break;
   2732 				}
   2733 			}
   2734 			break;
   2735 		}
   2736 		attr++;
   2737 	}
   2738 
   2739 #undef	ALG_KEY_SIZES
   2740 #undef	ALG_BLOCK_SIZES
   2741 #undef	ALG_PARAM_SIZES
   2742 
   2743 	/* update the algorithm tables */
   2744 	spdsock_merge_algs(spds);
   2745 bail:
   2746 	/* cleanup */
   2747 	ipsec_alg_free(alg);
   2748 	for (alg_type = 0; alg_type < IPSEC_NALGTYPES; alg_type++)
   2749 		for (algid = 0; algid < IPSEC_MAX_ALGS; algid++)
   2750 		if (spds->spds_algs[alg_type][algid] != NULL)
   2751 			ipsec_alg_free(spds->spds_algs[alg_type][algid]);
   2752 	return (diag);
   2753 }
   2754 
   2755 /*
   2756  * Process an SPD_UPDATEALGS request. If IPsec is not loaded, queue
   2757  * the request until IPsec loads. If IPsec is loaded, act on it
   2758  * immediately.
   2759  */
   2760 
   2761 static void
   2762 spdsock_updatealg(queue_t *q, mblk_t *mp, spd_ext_t *extv[])
   2763 {
   2764 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   2765 	spd_stack_t	*spds = ss->spdsock_spds;
   2766 	ipsec_stack_t	*ipss = spds->spds_netstack->netstack_ipsec;
   2767 
   2768 	if (!ipsec_loaded(ipss)) {
   2769 		/*
   2770 		 * IPsec is not loaded, save request and return nicely,
   2771 		 * the message will be processed once IPsec loads.
   2772 		 */
   2773 		mblk_t *new_mp;
   2774 
   2775 		/* last update message wins */
   2776 		if ((new_mp = copymsg(mp)) == NULL) {
   2777 			spdsock_error(q, mp, ENOMEM, 0);
   2778 			return;
   2779 		}
   2780 		mutex_enter(&spds->spds_alg_lock);
   2781 		bcopy(extv, spds->spds_extv_algs,
   2782 		    sizeof (spd_ext_t *) * (SPD_EXT_MAX + 1));
   2783 		if (spds->spds_mp_algs != NULL)
   2784 			freemsg(spds->spds_mp_algs);
   2785 		spds->spds_mp_algs = mp;
   2786 		spds->spds_algs_pending = B_TRUE;
   2787 		mutex_exit(&spds->spds_alg_lock);
   2788 		if (audit_active) {
   2789 			cred_t *cr;
   2790 			pid_t cpid;
   2791 
   2792 			cr = msg_getcred(mp, &cpid);
   2793 			audit_pf_policy(SPD_UPDATEALGS, cr,
   2794 			    spds->spds_netstack, NULL, B_TRUE, EAGAIN,
   2795 			    cpid);
   2796 		}
   2797 		spd_echo(q, new_mp);
   2798 	} else {
   2799 		/*
   2800 		 * IPsec is loaded, act on the message immediately.
   2801 		 */
   2802 		int diag;
   2803 
   2804 		mutex_enter(&spds->spds_alg_lock);
   2805 		diag = spdsock_do_updatealg(extv, spds);
   2806 		if (diag == -1) {
   2807 			/* Keep the lock held while we walk the SA tables. */
   2808 			sadb_alg_update(IPSEC_ALG_ALL, 0, 0,
   2809 			    spds->spds_netstack);
   2810 			mutex_exit(&spds->spds_alg_lock);
   2811 			spd_echo(q, mp);
   2812 			if (audit_active) {
   2813 				cred_t *cr;
   2814 				pid_t cpid;
   2815 
   2816 				cr = msg_getcred(mp, &cpid);
   2817 				audit_pf_policy(SPD_UPDATEALGS, cr,
   2818 				    spds->spds_netstack, NULL, B_TRUE, 0,
   2819 				    cpid);
   2820 			}
   2821 		} else {
   2822 			mutex_exit(&spds->spds_alg_lock);
   2823 			spdsock_diag(q, mp, diag);
   2824 			if (audit_active) {
   2825 				cred_t *cr;
   2826 				pid_t cpid;
   2827 
   2828 				cr = msg_getcred(mp, &cpid);
   2829 				audit_pf_policy(SPD_UPDATEALGS, cr,
   2830 				    spds->spds_netstack, NULL, B_TRUE, diag,
   2831 				    cpid);
   2832 			}
   2833 		}
   2834 	}
   2835 }
   2836 
   2837 /*
   2838  * Sort through the mess of polhead options to retrieve an appropriate one.
   2839  * Returns NULL if we send an spdsock error.  Returns a valid pointer if we
   2840  * found a valid polhead.  Returns ALL_ACTIVE_POLHEADS (aka. -1) or
   2841  * ALL_INACTIVE_POLHEADS (aka. -2) if the operation calls for the operation to
   2842  * act on ALL policy heads.
   2843  */
   2844 static ipsec_policy_head_t *
   2845 get_appropriate_polhead(queue_t *q, mblk_t *mp, spd_if_t *tunname, int spdid,
   2846     int msgtype, ipsec_tun_pol_t **itpp)
   2847 {
   2848 	ipsec_tun_pol_t *itp;
   2849 	ipsec_policy_head_t *iph;
   2850 	int errno;
   2851 	char *tname;
   2852 	boolean_t active;
   2853 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   2854 	netstack_t *ns = ss->spdsock_spds->spds_netstack;
   2855 	uint64_t gen;	/* Placeholder */
   2856 	datalink_id_t linkid;
   2857 
   2858 	active = (spdid == SPD_ACTIVE);
   2859 	*itpp = NULL;
   2860 	if (!active && spdid != SPD_STANDBY) {
   2861 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_BAD_SPDID);
   2862 		return (NULL);
   2863 	}
   2864 
   2865 	if (tunname != NULL) {
   2866 		/* Acting on a tunnel's SPD. */
   2867 		tname = (char *)tunname->spd_if_name;
   2868 		if (*tname == '\0') {
   2869 			/* Handle all-polhead cases here. */
   2870 			if (msgtype != SPD_FLUSH && msgtype != SPD_DUMP) {
   2871 				spdsock_diag(q, mp,
   2872 				    SPD_DIAGNOSTIC_NOT_GLOBAL_OP);
   2873 				return (NULL);
   2874 			}
   2875 			return (active ? ALL_ACTIVE_POLHEADS :
   2876 			    ALL_INACTIVE_POLHEADS);
   2877 		}
   2878 
   2879 		itp = get_tunnel_policy(tname, ns);
   2880 		if (itp == NULL) {
   2881 			if (msgtype != SPD_ADDRULE) {
   2882 				/* "Tunnel not found" */
   2883 				spdsock_error(q, mp, ENOENT, 0);
   2884 				return (NULL);
   2885 			}
   2886 
   2887 			errno = 0;
   2888 			itp = create_tunnel_policy(tname, &errno, &gen, ns);
   2889 			if (itp == NULL) {
   2890 				/*
   2891 				 * Something very bad happened, most likely
   2892 				 * ENOMEM.  Return an indicator.
   2893 				 */
   2894 				spdsock_error(q, mp, errno, 0);
   2895 				return (NULL);
   2896 			}
   2897 		}
   2898 		/*
   2899 		 * Troll the plumbed tunnels and see if we have a match.  We
   2900 		 * need to do this always in case we add policy AFTER plumbing
   2901 		 * a tunnel.
   2902 		 */
   2903 		if (dls_mgmt_get_linkid(tname, &linkid) == 0)
   2904 			iptun_set_policy(linkid, itp);
   2905 
   2906 		*itpp = itp;
   2907 		/* For spdsock dump state, set the polhead's name. */
   2908 		if (msgtype == SPD_DUMP) {
   2909 			ITP_REFHOLD(itp);
   2910 			ss->spdsock_itp = itp;
   2911 			ss->spdsock_dump_tunnel = itp->itp_flags &
   2912 			    (active ? ITPF_P_TUNNEL : ITPF_I_TUNNEL);
   2913 		}
   2914 	} else {
   2915 		itp = NULL;
   2916 		/* For spdsock dump state, indicate it's global policy. */
   2917 		if (msgtype == SPD_DUMP)
   2918 			ss->spdsock_itp = NULL;
   2919 	}
   2920 
   2921 	if (active)
   2922 		iph = (itp == NULL) ? ipsec_system_policy(ns) : itp->itp_policy;
   2923 	else
   2924 		iph = (itp == NULL) ? ipsec_inactive_policy(ns) :
   2925 		    itp->itp_inactive;
   2926 
   2927 	ASSERT(iph != NULL);
   2928 	if (itp != NULL) {
   2929 		IPPH_REFHOLD(iph);
   2930 	}
   2931 
   2932 	return (iph);
   2933 }
   2934 
   2935 static void
   2936 spdsock_parse(queue_t *q, mblk_t *mp)
   2937 {
   2938 	spd_msg_t *spmsg;
   2939 	spd_ext_t *extv[SPD_EXT_MAX + 1];
   2940 	uint_t msgsize;
   2941 	ipsec_policy_head_t *iph;
   2942 	ipsec_tun_pol_t *itp;
   2943 	spd_if_t *tunname;
   2944 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   2945 	spd_stack_t *spds = ss->spdsock_spds;
   2946 	netstack_t *ns = spds->spds_netstack;
   2947 	ipsec_stack_t *ipss = ns->netstack_ipsec;
   2948 
   2949 	/* Make sure nothing's below me. */
   2950 	ASSERT(WR(q)->q_next == NULL);
   2951 
   2952 	spmsg = (spd_msg_t *)mp->b_rptr;
   2953 
   2954 	msgsize = SPD_64TO8(spmsg->spd_msg_len);
   2955 
   2956 	if (msgdsize(mp) != msgsize) {
   2957 		/*
   2958 		 * Message len incorrect w.r.t. actual size.  Send an error
   2959 		 * (EMSGSIZE).	It may be necessary to massage things a
   2960 		 * bit.	 For example, if the spd_msg_type is hosed,
   2961 		 * I need to set it to SPD_RESERVED to get delivery to
   2962 		 * do the right thing.	Then again, maybe just letting
   2963 		 * the error delivery do the right thing.
   2964 		 */
   2965 		ss2dbg(spds,
   2966 		    ("mblk (%lu) and base (%d) message sizes don't jibe.\n",
   2967 		    msgdsize(mp), msgsize));
   2968 		spdsock_error(q, mp, EMSGSIZE, SPD_DIAGNOSTIC_NONE);
   2969 		return;
   2970 	}
   2971 
   2972 	if (msgsize > (uint_t)(mp->b_wptr - mp->b_rptr)) {
   2973 		/* Get all message into one mblk. */
   2974 		if (pullupmsg(mp, -1) == 0) {
   2975 			/*
   2976 			 * Something screwy happened.
   2977 			 */
   2978 			ss3dbg(spds, ("spdsock_parse: pullupmsg() failed.\n"));
   2979 			return;
   2980 		} else {
   2981 			spmsg = (spd_msg_t *)mp->b_rptr;
   2982 		}
   2983 	}
   2984 
   2985 	switch (spdsock_get_ext(extv, spmsg, msgsize)) {
   2986 	case KGE_DUP:
   2987 		/* Handle duplicate extension. */
   2988 		ss1dbg(spds, ("Got duplicate extension of type %d.\n",
   2989 		    extv[0]->spd_ext_type));
   2990 		spdsock_diag(q, mp, dup_ext_diag[extv[0]->spd_ext_type]);
   2991 		return;
   2992 	case KGE_UNK:
   2993 		/* Handle unknown extension. */
   2994 		ss1dbg(spds, ("Got unknown extension of type %d.\n",
   2995 		    extv[0]->spd_ext_type));
   2996 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_UNKNOWN_EXT);
   2997 		return;
   2998 	case KGE_LEN:
   2999 		/* Length error. */
   3000 		ss1dbg(spds, ("Length %d on extension type %d overrun or 0.\n",
   3001 		    extv[0]->spd_ext_len, extv[0]->spd_ext_type));
   3002 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_BAD_EXTLEN);
   3003 		return;
   3004 	case KGE_CHK:
   3005 		/* Reality check failed. */
   3006 		ss1dbg(spds, ("Reality check failed on extension type %d.\n",
   3007 		    extv[0]->spd_ext_type));
   3008 		spdsock_diag(q, mp, bad_ext_diag[extv[0]->spd_ext_type]);
   3009 		return;
   3010 	default:
   3011 		/* Default case is no errors. */
   3012 		break;
   3013 	}
   3014 
   3015 	/*
   3016 	 * Special-case SPD_UPDATEALGS so as not to load IPsec.
   3017 	 */
   3018 	if (!ipsec_loaded(ipss) && spmsg->spd_msg_type != SPD_UPDATEALGS) {
   3019 		spdsock_t *ss = (spdsock_t *)q->q_ptr;
   3020 
   3021 		ASSERT(ss != NULL);
   3022 		ipsec_loader_loadnow(ipss);
   3023 		ss->spdsock_timeout_arg = mp;
   3024 		ss->spdsock_timeout = qtimeout(q, spdsock_loadcheck,
   3025 		    q, LOADCHECK_INTERVAL);
   3026 		return;
   3027 	}
   3028 
   3029 	/* First check for messages that need no polheads at all. */
   3030 	switch (spmsg->spd_msg_type) {
   3031 	case SPD_UPDATEALGS:
   3032 		spdsock_updatealg(q, mp, extv);
   3033 		return;
   3034 	case SPD_ALGLIST:
   3035 		spdsock_alglist(q, mp);
   3036 		return;
   3037 	case SPD_DUMPALGS:
   3038 		spdsock_dumpalgs(q, mp);
   3039 		return;
   3040 	}
   3041 
   3042 	/*
   3043 	 * Then check for ones that need both primary/secondary polheads,
   3044 	 * finding the appropriate tunnel policy if need be.
   3045 	 */
   3046 	tunname = (spd_if_t *)extv[SPD_EXT_TUN_NAME];
   3047 	switch (spmsg->spd_msg_type) {
   3048 	case SPD_FLIP:
   3049 		spdsock_flip(q, mp, tunname);
   3050 		return;
   3051 	case SPD_CLONE:
   3052 		spdsock_clone(q, mp, tunname);
   3053 		return;
   3054 	}
   3055 
   3056 	/*
   3057 	 * Finally, find ones that operate on exactly one polhead, or
   3058 	 * "all polheads" of a given type (active/inactive).
   3059 	 */
   3060 	iph = get_appropriate_polhead(q, mp, tunname, spmsg->spd_msg_spdid,
   3061 	    spmsg->spd_msg_type, &itp);
   3062 	if (iph == NULL)
   3063 		return;
   3064 
   3065 	/* All-polheads-ready operations. */
   3066 	switch (spmsg->spd_msg_type) {
   3067 	case SPD_FLUSH:
   3068 		if (itp != NULL) {
   3069 			mutex_enter(&itp->itp_lock);
   3070 			if (spmsg->spd_msg_spdid == SPD_ACTIVE)
   3071 				itp->itp_flags &= ~ITPF_PFLAGS;
   3072 			else
   3073 				itp->itp_flags &= ~ITPF_IFLAGS;
   3074 			mutex_exit(&itp->itp_lock);
   3075 			ITP_REFRELE(itp, ns);
   3076 		}
   3077 		spdsock_flush(q, iph, itp, mp);
   3078 		return;
   3079 	case SPD_DUMP:
   3080 		if (itp != NULL)
   3081 			ITP_REFRELE(itp, ns);
   3082 		spdsock_dump(q, iph, mp);
   3083 		return;
   3084 	}
   3085 
   3086 	if (iph == ALL_ACTIVE_POLHEADS || iph == ALL_INACTIVE_POLHEADS) {
   3087 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_NOT_GLOBAL_OP);
   3088 		return;
   3089 	}
   3090 
   3091 	/* Single-polhead-only operations. */
   3092 	switch (spmsg->spd_msg_type) {
   3093 	case SPD_ADDRULE:
   3094 		spdsock_addrule(q, iph, mp, extv, itp);
   3095 		break;
   3096 	case SPD_DELETERULE:
   3097 		spdsock_deleterule(q, iph, mp, extv, itp);
   3098 		break;
   3099 	case SPD_LOOKUP:
   3100 		spdsock_lookup(q, iph, mp, extv, itp);
   3101 		break;
   3102 	default:
   3103 		spdsock_diag(q, mp, SPD_DIAGNOSTIC_BAD_MSG_TYPE);
   3104 		break;
   3105 	}
   3106 
   3107 	IPPH_REFRELE(iph, ns);
   3108 	if (itp != NULL)
   3109 		ITP_REFRELE(itp, ns);
   3110 }
   3111 
   3112 /*
   3113  * If an algorithm mapping was received before IPsec was loaded, process it.
   3114  * Called from the IPsec loader.
   3115  */
   3116 void
   3117 spdsock_update_pending_algs(netstack_t *ns)
   3118 {
   3119 	spd_stack_t *spds = ns->netstack_spdsock;
   3120 
   3121 	mutex_enter(&spds->spds_alg_lock);
   3122 	if (spds->spds_algs_pending) {
   3123 		(void) spdsock_do_updatealg(spds->spds_extv_algs, spds);
   3124 		spds->spds_algs_pending = B_FALSE;
   3125 	}
   3126 	mutex_exit(&spds->spds_alg_lock);
   3127 }
   3128 
   3129 static void
   3130 spdsock_loadcheck(void *arg)
   3131 {
   3132 	queue_t *q = (queue_t *)arg;
   3133 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   3134 	mblk_t *mp;
   3135 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
   3136 
   3137 	ASSERT(ss != NULL);
   3138 
   3139 	ss->spdsock_timeout = 0;
   3140 	mp = ss->spdsock_timeout_arg;
   3141 	ASSERT(mp != NULL);
   3142 	ss->spdsock_timeout_arg = NULL;
   3143 	if (ipsec_failed(ipss))
   3144 		spdsock_error(q, mp, EPROTONOSUPPORT, 0);
   3145 	else
   3146 		spdsock_parse(q, mp);
   3147 }
   3148 
   3149 /*
   3150  * Copy relevant state bits.
   3151  */
   3152 static void
   3153 spdsock_copy_info(struct T_info_ack *tap, spdsock_t *ss)
   3154 {
   3155 	*tap = spdsock_g_t_info_ack;
   3156 	tap->CURRENT_state = ss->spdsock_state;
   3157 	tap->OPT_size = spdsock_max_optsize;
   3158 }
   3159 
   3160 /*
   3161  * This routine responds to T_CAPABILITY_REQ messages.  It is called by
   3162  * spdsock_wput.  Much of the T_CAPABILITY_ACK information is copied from
   3163  * spdsock_g_t_info_ack.  The current state of the stream is copied from
   3164  * spdsock_state.
   3165  */
   3166 static void
   3167 spdsock_capability_req(queue_t *q, mblk_t *mp)
   3168 {
   3169 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   3170 	t_uscalar_t cap_bits1;
   3171 	struct T_capability_ack	*tcap;
   3172 
   3173 	cap_bits1 = ((struct T_capability_req *)mp->b_rptr)->CAP_bits1;
   3174 
   3175 	mp = tpi_ack_alloc(mp, sizeof (struct T_capability_ack),
   3176 	    mp->b_datap->db_type, T_CAPABILITY_ACK);
   3177 	if (mp == NULL)
   3178 		return;
   3179 
   3180 	tcap = (struct T_capability_ack *)mp->b_rptr;
   3181 	tcap->CAP_bits1 = 0;
   3182 
   3183 	if (cap_bits1 & TC1_INFO) {
   3184 		spdsock_copy_info(&tcap->INFO_ack, ss);
   3185 		tcap->CAP_bits1 |= TC1_INFO;
   3186 	}
   3187 
   3188 	qreply(q, mp);
   3189 }
   3190 
   3191 /*
   3192  * This routine responds to T_INFO_REQ messages. It is called by
   3193  * spdsock_wput_other.
   3194  * Most of the T_INFO_ACK information is copied from spdsock_g_t_info_ack.
   3195  * The current state of the stream is copied from spdsock_state.
   3196  */
   3197 static void
   3198 spdsock_info_req(q, mp)
   3199 	queue_t	*q;
   3200 	mblk_t	*mp;
   3201 {
   3202 	mp = tpi_ack_alloc(mp, sizeof (struct T_info_ack), M_PCPROTO,
   3203 	    T_INFO_ACK);
   3204 	if (mp == NULL)
   3205 		return;
   3206 	spdsock_copy_info((struct T_info_ack *)mp->b_rptr,
   3207 	    (spdsock_t *)q->q_ptr);
   3208 	qreply(q, mp);
   3209 }
   3210 
   3211 /*
   3212  * spdsock_err_ack. This routine creates a
   3213  * T_ERROR_ACK message and passes it
   3214  * upstream.
   3215  */
   3216 static void
   3217 spdsock_err_ack(q, mp, t_error, sys_error)
   3218 	queue_t	*q;
   3219 	mblk_t	*mp;
   3220 	int	t_error;
   3221 	int	sys_error;
   3222 {
   3223 	if ((mp = mi_tpi_err_ack_alloc(mp, t_error, sys_error)) != NULL)
   3224 		qreply(q, mp);
   3225 }
   3226 
   3227 /*
   3228  * This routine retrieves the current status of socket options.
   3229  * It returns the size of the option retrieved.
   3230  */
   3231 /* ARGSUSED */
   3232 int
   3233 spdsock_opt_get(queue_t *q, int level, int name, uchar_t *ptr)
   3234 {
   3235 	int *i1 = (int *)ptr;
   3236 
   3237 	switch (level) {
   3238 	case SOL_SOCKET:
   3239 		switch (name) {
   3240 		case SO_TYPE:
   3241 			*i1 = SOCK_RAW;
   3242 			break;
   3243 		/*
   3244 		 * The following two items can be manipulated,
   3245 		 * but changing them should do nothing.
   3246 		 */
   3247 		case SO_SNDBUF:
   3248 			*i1 = (int)q->q_hiwat;
   3249 			break;
   3250 		case SO_RCVBUF:
   3251 			*i1 = (int)(RD(q)->q_hiwat);
   3252 			break;
   3253 		}
   3254 		break;
   3255 	default:
   3256 		return (0);
   3257 	}
   3258 	return (sizeof (int));
   3259 }
   3260 
   3261 /*
   3262  * This routine sets socket options.
   3263  */
   3264 /* ARGSUSED */
   3265 int
   3266 spdsock_opt_set(queue_t *q, uint_t mgmt_flags, int level, int name,
   3267     uint_t inlen, uchar_t *invalp, uint_t *outlenp, uchar_t *outvalp,
   3268     void *thisdg_attrs, cred_t *cr)
   3269 {
   3270 	int *i1 = (int *)invalp;
   3271 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   3272 	spd_stack_t	*spds = ss->spdsock_spds;
   3273 
   3274 	switch (level) {
   3275 	case SOL_SOCKET:
   3276 		switch (name) {
   3277 		case SO_SNDBUF:
   3278 			if (*i1 > spds->spds_max_buf)
   3279 				return (ENOBUFS);
   3280 			q->q_hiwat = *i1;
   3281 			break;
   3282 		case SO_RCVBUF:
   3283 			if (*i1 > spds->spds_max_buf)
   3284 				return (ENOBUFS);
   3285 			RD(q)->q_hiwat = *i1;
   3286 			(void) proto_set_rx_hiwat(RD(q), NULL, *i1);
   3287 			break;
   3288 		}
   3289 		break;
   3290 	}
   3291 	return (0);
   3292 }
   3293 
   3294 
   3295 /*
   3296  * Handle STREAMS messages.
   3297  */
   3298 static void
   3299 spdsock_wput_other(queue_t *q, mblk_t *mp)
   3300 {
   3301 	struct iocblk *iocp;
   3302 	int error;
   3303 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   3304 	spd_stack_t	*spds = ss->spdsock_spds;
   3305 	cred_t		*cr;
   3306 
   3307 	switch (mp->b_datap->db_type) {
   3308 	case M_PROTO:
   3309 	case M_PCPROTO:
   3310 		if ((mp->b_wptr - mp->b_rptr) < sizeof (long)) {
   3311 			ss3dbg(spds, (
   3312 			    "spdsock_wput_other: Not big enough M_PROTO\n"));
   3313 			freemsg(mp);
   3314 			return;
   3315 		}
   3316 		switch (((union T_primitives *)mp->b_rptr)->type) {
   3317 		case T_CAPABILITY_REQ:
   3318 			spdsock_capability_req(q, mp);
   3319 			break;
   3320 		case T_INFO_REQ:
   3321 			spdsock_info_req(q, mp);
   3322 			break;
   3323 		case T_SVR4_OPTMGMT_REQ:
   3324 		case T_OPTMGMT_REQ:
   3325 			/*
   3326 			 * All Solaris components should pass a db_credp
   3327 			 * for this TPI message, hence we ASSERT.
   3328 			 * But in case there is some other M_PROTO that looks
   3329 			 * like a TPI message sent by some other kernel
   3330 			 * component, we check and return an error.
   3331 			 */
   3332 			cr = msg_getcred(mp, NULL);
   3333 			ASSERT(cr != NULL);
   3334 			if (cr == NULL) {
   3335 				spdsock_err_ack(q, mp, TSYSERR, EINVAL);
   3336 				return;
   3337 			}
   3338 			if (((union T_primitives *)mp->b_rptr)->type ==
   3339 			    T_SVR4_OPTMGMT_REQ) {
   3340 				svr4_optcom_req(q, mp, cr, &spdsock_opt_obj);
   3341 			} else {
   3342 				tpi_optcom_req(q, mp, cr, &spdsock_opt_obj);
   3343 			}
   3344 			break;
   3345 		case T_DATA_REQ:
   3346 		case T_EXDATA_REQ:
   3347 		case T_ORDREL_REQ:
   3348 			/* Illegal for spdsock. */
   3349 			freemsg(mp);
   3350 			(void) putnextctl1(RD(q), M_ERROR, EPROTO);
   3351 			break;
   3352 		default:
   3353 			/* Not supported by spdsock. */
   3354 			spdsock_err_ack(q, mp, TNOTSUPPORT, 0);
   3355 			break;
   3356 		}
   3357 		return;
   3358 	case M_IOCTL:
   3359 		iocp = (struct iocblk *)mp->b_rptr;
   3360 		error = EINVAL;
   3361 
   3362 		switch (iocp->ioc_cmd) {
   3363 		case ND_SET:
   3364 		case ND_GET:
   3365 			if (nd_getset(q, spds->spds_g_nd, mp)) {
   3366 				qreply(q, mp);
   3367 				return;
   3368 			} else
   3369 				error = ENOENT;
   3370 			/* FALLTHRU */
   3371 		default:
   3372 			miocnak(q, mp, 0, error);
   3373 			return;
   3374 		}
   3375 	case M_FLUSH:
   3376 		if (*mp->b_rptr & FLUSHW) {
   3377 			flushq(q, FLUSHALL);
   3378 			*mp->b_rptr &= ~FLUSHW;
   3379 		}
   3380 		if (*mp->b_rptr & FLUSHR) {
   3381 			qreply(q, mp);
   3382 			return;
   3383 		}
   3384 		/* Else FALLTHRU */
   3385 	}
   3386 
   3387 	/* If fell through, just black-hole the message. */
   3388 	freemsg(mp);
   3389 }
   3390 
   3391 static void
   3392 spdsock_wput(queue_t *q, mblk_t *mp)
   3393 {
   3394 	uint8_t *rptr = mp->b_rptr;
   3395 	mblk_t *mp1;
   3396 	spdsock_t *ss = (spdsock_t *)q->q_ptr;
   3397 	spd_stack_t	*spds = ss->spdsock_spds;
   3398 
   3399 	/*
   3400 	 * If we're dumping, defer processing other messages until the
   3401 	 * dump completes.
   3402 	 */
   3403 	if (ss->spdsock_dump_req != NULL) {
   3404 		if (!putq(q, mp))
   3405 			freemsg(mp);
   3406 		return;
   3407 	}
   3408 
   3409 	switch (mp->b_datap->db_type) {
   3410 	case M_DATA:
   3411 		/*
   3412 		 * Silently discard.
   3413 		 */
   3414 		ss2dbg(spds, ("raw M_DATA in spdsock.\n"));
   3415 		freemsg(mp);
   3416 		return;
   3417 	case M_PROTO:
   3418 	case M_PCPROTO:
   3419 		if ((mp->b_wptr - rptr) >= sizeof (struct T_data_req)) {
   3420 			if (((union T_primitives *)rptr)->type == T_DATA_REQ) {
   3421 				if ((mp1 = mp->b_cont) == NULL) {
   3422 					/* No data after T_DATA_REQ. */
   3423 					ss2dbg(spds,
   3424 					    ("No data after DATA_REQ.\n"));
   3425 					freemsg(mp);
   3426 					return;
   3427 				}
   3428 				freeb(mp);
   3429 				mp = mp1;
   3430 				ss2dbg(spds, ("T_DATA_REQ\n"));
   3431 				break;	/* Out of switch. */
   3432 			}
   3433 		}
   3434 		/* FALLTHRU */
   3435 	default:
   3436 		ss3dbg(spds, ("In default wput case (%d %d).\n",
   3437 		    mp->b_datap->db_type, ((union T_primitives *)rptr)->type));
   3438 		spdsock_wput_other(q, mp);
   3439 		return;
   3440 	}
   3441 
   3442 	/* I now have a PF_POLICY message in an M_DATA block. */
   3443 	spdsock_parse(q, mp);
   3444 }
   3445 
   3446 /*
   3447  * Device open procedure, called when new queue pair created.
   3448  * We are passed the read-side queue.
   3449  */
   3450 /* ARGSUSED */
   3451 static int
   3452 spdsock_open(queue_t *q, dev_t *devp, int flag, int sflag, cred_t *credp)
   3453 {
   3454 	spdsock_t *ss;
   3455 	queue_t *oq = OTHERQ(q);
   3456 	minor_t ssminor;
   3457 	netstack_t *ns;
   3458 	spd_stack_t *spds;
   3459 
   3460 	if (secpolicy_ip_config(credp, B_FALSE) != 0)
   3461 		return (EPERM);
   3462 
   3463 	if (q->q_ptr != NULL)
   3464 		return (0);  /* Re-open of an already open instance. */
   3465 
   3466 	if (sflag & MODOPEN)
   3467 		return (EINVAL);
   3468 
   3469 	ns = netstack_find_by_cred(credp);
   3470 	ASSERT(ns != NULL);
   3471 	spds = ns->netstack_spdsock;
   3472 	ASSERT(spds != NULL);
   3473 
   3474 	ss2dbg(spds, ("Made it into PF_POLICY socket open.\n"));
   3475 
   3476 	ssminor = (minor_t)(uintptr_t)vmem_alloc(spdsock_vmem, 1, VM_NOSLEEP);
   3477 	if (ssminor == 0) {
   3478 		netstack_rele(spds->spds_netstack);
   3479 		return (ENOMEM);
   3480 	}
   3481 	ss = kmem_zalloc(sizeof (spdsock_t), KM_NOSLEEP);
   3482 	if (ss == NULL) {
   3483 		vmem_free(spdsock_vmem, (void *)(uintptr_t)ssminor, 1);
   3484 		netstack_rele(spds->spds_netstack);
   3485 		return (ENOMEM);
   3486 	}
   3487 
   3488 	ss->spdsock_minor = ssminor;
   3489 	ss->spdsock_state = TS_UNBND;
   3490 	ss->spdsock_dump_req = NULL;
   3491 
   3492 	ss->spdsock_spds = spds;
   3493 
   3494 	q->q_ptr = ss;
   3495 	oq->q_ptr = ss;
   3496 
   3497 	q->q_hiwat = spds->spds_recv_hiwat;
   3498 
   3499 	oq->q_hiwat = spds->spds_xmit_hiwat;
   3500 	oq->q_lowat = spds->spds_xmit_lowat;
   3501 
   3502 	qprocson(q);
   3503 	(void) proto_set_rx_hiwat(q, NULL, spds->spds_recv_hiwat);
   3504 
   3505 	*devp = makedevice(getmajor(*devp), ss->spdsock_minor);
   3506 	return (0);
   3507 }
   3508 
   3509 /*
   3510  * Read-side service procedure, invoked when we get back-enabled
   3511  * when buffer space becomes available.
   3512  *
   3513  * Dump another chunk if we were dumping before; when we finish, kick
   3514  * the write-side queue in case it's waiting for read queue space.
   3515  */
   3516 void
   3517 spdsock_rsrv(queue_t *q)
   3518 {
   3519 	spdsock_t *ss = q->q_ptr;
   3520 
   3521 	if (ss->spdsock_dump_req != NULL)
   3522 		spdsock_dump_some(q, ss);
   3523 
   3524 	if (ss->spdsock_dump_req == NULL)
   3525 		qenable(OTHERQ(q));
   3526 }
   3527 
   3528 /*
   3529  * Write-side service procedure, invoked when we defer processing
   3530  * if another message is received while a dump is in progress.
   3531  */
   3532 void
   3533 spdsock_wsrv(queue_t *q)
   3534 {
   3535 	spdsock_t *ss = q->q_ptr;
   3536 	mblk_t *mp;
   3537 	ipsec_stack_t *ipss = ss->spdsock_spds->spds_netstack->netstack_ipsec;
   3538 
   3539 	if (ss->spdsock_dump_req != NULL) {
   3540 		qenable(OTHERQ(q));
   3541 		return;
   3542 	}
   3543 
   3544 	while ((mp = getq(q)) != NULL) {
   3545 		if (ipsec_loaded(ipss)) {
   3546 			spdsock_wput(q, mp);
   3547 			if (ss->spdsock_dump_req != NULL)
   3548 				return;
   3549 		} else if (!ipsec_failed(ipss)) {
   3550 			(void) putq(q, mp);
   3551 		} else {
   3552 			spdsock_error(q, mp, EPFNOSUPPORT, 0);
   3553 		}
   3554 	}
   3555 }
   3556 
   3557 static int
   3558 spdsock_close(queue_t *q)
   3559 {
   3560 	spdsock_t *ss = q->q_ptr;
   3561 	spd_stack_t	*spds = ss->spdsock_spds;
   3562 
   3563 	qprocsoff(q);
   3564 
   3565 	/* Safe assumption. */
   3566 	ASSERT(ss != NULL);
   3567 
   3568 	if (ss->spdsock_timeout != 0)
   3569 		(void) quntimeout(q, ss->spdsock_timeout);
   3570 
   3571 	ss3dbg(spds, ("Driver close, PF_POLICY socket is going away.\n"));
   3572 
   3573 	vmem_free(spdsock_vmem, (void *)(uintptr_t)ss->spdsock_minor, 1);
   3574 	netstack_rele(ss->spdsock_spds->spds_netstack);
   3575 
   3576 	kmem_free(ss, sizeof (spdsock_t));
   3577 	return (0);
   3578 }
   3579 
   3580 /*
   3581  * Merge the IPsec algorithms tables with the received algorithm information.
   3582  */
   3583 void
   3584 spdsock_merge_algs(spd_stack_t *spds)
   3585 {
   3586 	ipsec_alginfo_t *alg, *oalg;
   3587 	ipsec_algtype_t algtype;
   3588 	uint_t algidx, algid, nalgs;
   3589 	crypto_mech_name_t *mechs;
   3590 	uint_t mech_count, mech_idx;
   3591 	netstack_t	*ns = spds->spds_netstack;
   3592 	ipsec_stack_t	*ipss = ns->netstack_ipsec;
   3593 
   3594 	ASSERT(MUTEX_HELD(&spds->spds_alg_lock));
   3595 
   3596 	/*
   3597 	 * Get the list of supported mechanisms from the crypto framework.
   3598 	 * If a mechanism is supported by KCF, resolve its mechanism
   3599 	 * id and mark it as being valid. This operation must be done
   3600 	 * without holding alg_lock, since it can cause a provider
   3601 	 * module to be loaded and the provider notification callback to
   3602 	 * be invoked.
   3603 	 */
   3604 	mechs = crypto_get_mech_list(&mech_count, KM_SLEEP);
   3605 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
   3606 		for (algid = 0; algid < IPSEC_MAX_ALGS; algid++) {
   3607 			int algflags = 0;
   3608 			crypto_mech_type_t mt = CRYPTO_MECHANISM_INVALID;
   3609 
   3610 			alg = spds->spds_algs[algtype][algid];
   3611 			if (alg == NULL)
   3612 				continue;
   3613 
   3614 			/*
   3615 			 * The NULL encryption algorithm is a special
   3616 			 * case because there are no mechanisms, yet
   3617 			 * the algorithm is still valid.
   3618 			 */
   3619 			if (alg->alg_id == SADB_EALG_NULL) {
   3620 				alg->alg_mech_type = CRYPTO_MECHANISM_INVALID;
   3621 				alg->alg_flags |= ALG_FLAG_VALID;
   3622 				continue;
   3623 			}
   3624 
   3625 			for (mech_idx = 0; mech_idx < mech_count; mech_idx++) {
   3626 				if (strncmp(alg->alg_mech_name, mechs[mech_idx],
   3627 				    CRYPTO_MAX_MECH_NAME) == 0) {
   3628 					mt = crypto_mech2id(alg->alg_mech_name);
   3629 					ASSERT(mt != CRYPTO_MECHANISM_INVALID);
   3630 					algflags = ALG_FLAG_VALID;
   3631 					break;
   3632 				}
   3633 			}
   3634 			alg->alg_mech_type = mt;
   3635 			alg->alg_flags |= algflags;
   3636 		}
   3637 	}
   3638 
   3639 	mutex_enter(&ipss->ipsec_alg_lock);
   3640 
   3641 	/*
   3642 	 * For each algorithm currently defined, check if it is
   3643 	 * present in the new tables created from the SPD_UPDATEALGS
   3644 	 * message received from user-space.
   3645 	 * Delete the algorithm entries that are currently defined
   3646 	 * but not part of the new tables.
   3647 	 */
   3648 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
   3649 		nalgs = ipss->ipsec_nalgs[algtype];
   3650 		for (algidx = 0; algidx < nalgs; algidx++) {
   3651 			algid = ipss->ipsec_sortlist[algtype][algidx];
   3652 			if (spds->spds_algs[algtype][algid] == NULL)
   3653 				ipsec_alg_unreg(algtype, algid, ns);
   3654 		}
   3655 	}
   3656 
   3657 	/*
   3658 	 * For each algorithm we just received, check if it is
   3659 	 * present in the currently defined tables. If it is, swap
   3660 	 * the entry with the one we just allocated.
   3661 	 * If the new algorithm is not in the current tables,
   3662 	 * add it.
   3663 	 */
   3664 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
   3665 		for (algid = 0; algid < IPSEC_MAX_ALGS; algid++) {
   3666 			alg = spds->spds_algs[algtype][algid];
   3667 			if (alg == NULL)
   3668 				continue;
   3669 
   3670 			if ((oalg = ipss->ipsec_alglists[algtype][algid]) ==
   3671 			    NULL) {
   3672 				/*
   3673 				 * New algorithm, add it to the algorithm
   3674 				 * table.
   3675 				 */
   3676 				ipsec_alg_reg(algtype, alg, ns);
   3677 			} else {
   3678 				/*
   3679 				 * Algorithm is already in the table. Swap
   3680 				 * the existing entry with the new one.
   3681 				 */
   3682 				ipsec_alg_fix_min_max(alg, algtype, ns);
   3683 				ipss->ipsec_alglists[algtype][algid] = alg;
   3684 				ipsec_alg_free(oalg);
   3685 			}
   3686 			spds->spds_algs[algtype][algid] = NULL;
   3687 		}
   3688 	}
   3689 
   3690 	for (algtype = 0; algtype < IPSEC_NALGTYPES; algtype++) {
   3691 		ipss->ipsec_algs_exec_mode[algtype] =
   3692 		    spds->spds_algs_exec_mode[algtype];
   3693 	}
   3694 
   3695 	mutex_exit(&ipss->ipsec_alg_lock);
   3696 
   3697 	crypto_free_mech_list(mechs, mech_count);
   3698 
   3699 	ipsecah_algs_changed(ns);
   3700 	ipsecesp_algs_changed(ns);
   3701 }
   3702