Home | History | Annotate | Download | only in sctp
      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 /*
     23  * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
     24  * Use is subject to license terms.
     25  */
     26 
     27 #include <sys/types.h>
     28 #include <sys/systm.h>
     29 #include <sys/stream.h>
     30 #include <sys/cmn_err.h>
     31 #include <sys/kmem.h>
     32 #define	_SUN_TPI_VERSION 2
     33 #include <sys/tihdr.h>
     34 #include <sys/stropts.h>
     35 #include <sys/socket.h>
     36 #include <sys/random.h>
     37 #include <sys/policy.h>
     38 #include <sys/tsol/tndb.h>
     39 #include <sys/tsol/tnet.h>
     40 
     41 #include <netinet/in.h>
     42 #include <netinet/ip6.h>
     43 
     44 #include <inet/common.h>
     45 #include <inet/ip.h>
     46 #include <inet/ip6.h>
     47 #include <inet/ipclassifier.h>
     48 #include "sctp_impl.h"
     49 #include "sctp_asconf.h"
     50 #include "sctp_addr.h"
     51 
     52 /*
     53  * Returns 0 on success, EACCES on permission failure.
     54  */
     55 static int
     56 sctp_select_port(sctp_t *sctp, in_port_t *requested_port, int *user_specified)
     57 {
     58 	sctp_stack_t	*sctps = sctp->sctp_sctps;
     59 	conn_t		*connp = sctp->sctp_connp;
     60 
     61 	/*
     62 	 * Get a valid port (within the anonymous range and should not
     63 	 * be a privileged one) to use if the user has not given a port.
     64 	 * If multiple threads are here, they may all start with
     65 	 * with the same initial port. But, it should be fine as long as
     66 	 * sctp_bindi will ensure that no two threads will be assigned
     67 	 * the same port.
     68 	 */
     69 	if (*requested_port == 0) {
     70 		*requested_port = sctp_update_next_port(
     71 		    sctps->sctps_next_port_to_try,
     72 		    crgetzone(connp->conn_cred), sctps);
     73 		if (*requested_port == 0)
     74 			return (EACCES);
     75 		*user_specified = 0;
     76 	} else {
     77 		int i;
     78 		boolean_t priv = B_FALSE;
     79 
     80 		/*
     81 		 * If the requested_port is in the well-known privileged range,
     82 		 * verify that the stream was opened by a privileged user.
     83 		 * Note: No locks are held when inspecting sctp_g_*epriv_ports
     84 		 * but instead the code relies on:
     85 		 * - the fact that the address of the array and its size never
     86 		 *   changes
     87 		 * - the atomic assignment of the elements of the array
     88 		 */
     89 		if (*requested_port < sctps->sctps_smallest_nonpriv_port) {
     90 			priv = B_TRUE;
     91 		} else {
     92 			for (i = 0; i < sctps->sctps_g_num_epriv_ports; i++) {
     93 				if (*requested_port ==
     94 				    sctps->sctps_g_epriv_ports[i]) {
     95 					priv = B_TRUE;
     96 					break;
     97 				}
     98 			}
     99 		}
    100 		if (priv) {
    101 			/*
    102 			 * sctp_bind() should take a cred_t argument so that
    103 			 * we can use it here.
    104 			 */
    105 			if (secpolicy_net_privaddr(connp->conn_cred,
    106 			    *requested_port, IPPROTO_SCTP) != 0) {
    107 				dprint(1,
    108 				    ("sctp_bind(x): no prive for port %d",
    109 				    *requested_port));
    110 				return (EACCES);
    111 			}
    112 		}
    113 		*user_specified = 1;
    114 	}
    115 
    116 	return (0);
    117 }
    118 
    119 int
    120 sctp_listen(sctp_t *sctp)
    121 {
    122 	sctp_tf_t	*tf;
    123 	sctp_stack_t	*sctps = sctp->sctp_sctps;
    124 	conn_t		*connp = sctp->sctp_connp;
    125 
    126 	RUN_SCTP(sctp);
    127 	/*
    128 	 * TCP handles listen() increasing the backlog, need to check
    129 	 * if it should be handled here too
    130 	 */
    131 	if (sctp->sctp_state > SCTPS_BOUND ||
    132 	    (sctp->sctp_connp->conn_state_flags & CONN_CLOSING)) {
    133 		WAKE_SCTP(sctp);
    134 		return (EINVAL);
    135 	}
    136 
    137 	/* Do an anonymous bind for unbound socket doing listen(). */
    138 	if (sctp->sctp_nsaddrs == 0) {
    139 		struct sockaddr_storage ss;
    140 		int ret;
    141 
    142 		bzero(&ss, sizeof (ss));
    143 		ss.ss_family = connp->conn_family;
    144 
    145 		WAKE_SCTP(sctp);
    146 		if ((ret = sctp_bind(sctp, (struct sockaddr *)&ss,
    147 		    sizeof (ss))) != 0)
    148 			return (ret);
    149 		RUN_SCTP(sctp)
    150 	}
    151 
    152 	/* Cache things in the ixa without any refhold */
    153 	connp->conn_ixa->ixa_cred = connp->conn_cred;
    154 	connp->conn_ixa->ixa_cpid = connp->conn_cpid;
    155 	if (is_system_labeled())
    156 		connp->conn_ixa->ixa_tsl = crgetlabel(connp->conn_cred);
    157 
    158 	sctp->sctp_state = SCTPS_LISTEN;
    159 	(void) random_get_pseudo_bytes(sctp->sctp_secret, SCTP_SECRET_LEN);
    160 	sctp->sctp_last_secret_update = ddi_get_lbolt64();
    161 	bzero(sctp->sctp_old_secret, SCTP_SECRET_LEN);
    162 	tf = &sctps->sctps_listen_fanout[SCTP_LISTEN_HASH(
    163 	    ntohs(connp->conn_lport))];
    164 	sctp_listen_hash_insert(tf, sctp);
    165 	WAKE_SCTP(sctp);
    166 	return (0);
    167 }
    168 
    169 /*
    170  * Bind the sctp_t to a sockaddr, which includes an address and other
    171  * information, such as port or flowinfo.
    172  */
    173 int
    174 sctp_bind(sctp_t *sctp, struct sockaddr *sa, socklen_t len)
    175 {
    176 	int		user_specified;
    177 	boolean_t	bind_to_req_port_only;
    178 	in_port_t	requested_port;
    179 	in_port_t	allocated_port;
    180 	int		err = 0;
    181 	conn_t		*connp = sctp->sctp_connp;
    182 	uint_t		scope_id;
    183 	sin_t		*sin;
    184 	sin6_t		*sin6;
    185 
    186 	ASSERT(sctp != NULL);
    187 
    188 	RUN_SCTP(sctp);
    189 
    190 	if ((sctp->sctp_state >= SCTPS_BOUND) ||
    191 	    (sctp->sctp_connp->conn_state_flags & CONN_CLOSING) ||
    192 	    (sa == NULL || len == 0)) {
    193 		/*
    194 		 * Multiple binds not allowed for any SCTP socket
    195 		 * Also binding with null address is not supported.
    196 		 */
    197 		err = EINVAL;
    198 		goto done;
    199 	}
    200 
    201 	switch (sa->sa_family) {
    202 	case AF_INET:
    203 		sin = (sin_t *)sa;
    204 		if (len < sizeof (struct sockaddr_in) ||
    205 		    connp->conn_family == AF_INET6) {
    206 			err = EINVAL;
    207 			goto done;
    208 		}
    209 		requested_port = ntohs(sin->sin_port);
    210 		break;
    211 	case AF_INET6:
    212 		sin6 = (sin6_t *)sa;
    213 		if (len < sizeof (struct sockaddr_in6) ||
    214 		    connp->conn_family == AF_INET) {
    215 			err = EINVAL;
    216 			goto done;
    217 		}
    218 		requested_port = ntohs(sin6->sin6_port);
    219 		/* Set the flowinfo. */
    220 		connp->conn_flowinfo =
    221 		    sin6->sin6_flowinfo & ~IPV6_VERS_AND_FLOW_MASK;
    222 
    223 		scope_id = sin6->sin6_scope_id;
    224 		if (scope_id != 0 && IN6_IS_ADDR_LINKSCOPE(&sin6->sin6_addr)) {
    225 			connp->conn_ixa->ixa_flags |= IXAF_SCOPEID_SET;
    226 			connp->conn_ixa->ixa_scopeid = scope_id;
    227 			connp->conn_incoming_ifindex = scope_id;
    228 		} else {
    229 			connp->conn_ixa->ixa_flags &= ~IXAF_SCOPEID_SET;
    230 			connp->conn_incoming_ifindex = connp->conn_bound_if;
    231 		}
    232 		break;
    233 	default:
    234 		err = EAFNOSUPPORT;
    235 		goto done;
    236 	}
    237 	bind_to_req_port_only = requested_port == 0 ? B_FALSE : B_TRUE;
    238 
    239 	err = sctp_select_port(sctp, &requested_port, &user_specified);
    240 	if (err != 0)
    241 		goto done;
    242 
    243 	if ((err = sctp_bind_add(sctp, sa, 1, B_TRUE,
    244 	    user_specified == 1 ? htons(requested_port) : 0)) != 0) {
    245 		goto done;
    246 	}
    247 	err = sctp_bindi(sctp, requested_port, bind_to_req_port_only,
    248 	    user_specified, &allocated_port);
    249 	if (err != 0) {
    250 		sctp_free_saddrs(sctp);
    251 	} else {
    252 		ASSERT(sctp->sctp_state == SCTPS_BOUND);
    253 	}
    254 done:
    255 	WAKE_SCTP(sctp);
    256 	return (err);
    257 }
    258 
    259 /*
    260  * Perform bind/unbind operation of a list of addresses on a sctp_t
    261  */
    262 int
    263 sctp_bindx(sctp_t *sctp, const void *addrs, int addrcnt, int bindop)
    264 {
    265 	ASSERT(sctp != NULL);
    266 	ASSERT(addrs != NULL);
    267 	ASSERT(addrcnt > 0);
    268 
    269 	switch (bindop) {
    270 	case SCTP_BINDX_ADD_ADDR:
    271 		return (sctp_bind_add(sctp, addrs, addrcnt, B_FALSE,
    272 		    sctp->sctp_connp->conn_lport));
    273 	case SCTP_BINDX_REM_ADDR:
    274 		return (sctp_bind_del(sctp, addrs, addrcnt, B_FALSE));
    275 	default:
    276 		return (EINVAL);
    277 	}
    278 }
    279 
    280 /*
    281  * Add a list of addresses to a sctp_t.
    282  */
    283 int
    284 sctp_bind_add(sctp_t *sctp, const void *addrs, uint32_t addrcnt,
    285     boolean_t caller_hold_lock, in_port_t port)
    286 {
    287 	int		err = 0;
    288 	boolean_t	do_asconf = B_FALSE;
    289 	sctp_stack_t	*sctps = sctp->sctp_sctps;
    290 	conn_t		*connp = sctp->sctp_connp;
    291 
    292 	if (!caller_hold_lock)
    293 		RUN_SCTP(sctp);
    294 
    295 	if (sctp->sctp_state > SCTPS_ESTABLISHED ||
    296 	    (sctp->sctp_connp->conn_state_flags & CONN_CLOSING)) {
    297 		if (!caller_hold_lock)
    298 			WAKE_SCTP(sctp);
    299 		return (EINVAL);
    300 	}
    301 
    302 	if (sctp->sctp_state > SCTPS_LISTEN) {
    303 		/*
    304 		 * Let's do some checking here rather than undoing the
    305 		 * add later (for these reasons).
    306 		 */
    307 		if (!sctps->sctps_addip_enabled ||
    308 		    !sctp->sctp_understands_asconf ||
    309 		    !sctp->sctp_understands_addip) {
    310 			if (!caller_hold_lock)
    311 				WAKE_SCTP(sctp);
    312 			return (EINVAL);
    313 		}
    314 		do_asconf = B_TRUE;
    315 	}
    316 	/*
    317 	 * On a clustered node, for an inaddr_any bind, we will pass the list
    318 	 * of all the addresses in the global list, minus any address on the
    319 	 * loopback interface, and expect the clustering susbsystem to give us
    320 	 * the correct list for the 'port'. For explicit binds we give the
    321 	 * list of addresses  and the clustering module validates it for the
    322 	 * 'port'.
    323 	 *
    324 	 * On a non-clustered node, cl_sctp_check_addrs will be NULL and
    325 	 * we proceed as usual.
    326 	 */
    327 	if (cl_sctp_check_addrs != NULL) {
    328 		uchar_t		*addrlist = NULL;
    329 		size_t		size = 0;
    330 		int		unspec = 0;
    331 		boolean_t	do_listen;
    332 		uchar_t		*llist = NULL;
    333 		size_t		lsize = 0;
    334 
    335 		/*
    336 		 * If we are adding addresses after listening, but before
    337 		 * an association is established, we need to update the
    338 		 * clustering module with this info.
    339 		 */
    340 		do_listen = !do_asconf && sctp->sctp_state > SCTPS_BOUND &&
    341 		    cl_sctp_listen != NULL;
    342 
    343 		err = sctp_get_addrlist(sctp, addrs, &addrcnt, &addrlist,
    344 		    &unspec, &size);
    345 		if (err != 0) {
    346 			ASSERT(addrlist == NULL);
    347 			ASSERT(addrcnt == 0);
    348 			ASSERT(size == 0);
    349 			if (!caller_hold_lock)
    350 				WAKE_SCTP(sctp);
    351 			SCTP_KSTAT(sctps, sctp_cl_check_addrs);
    352 			return (err);
    353 		}
    354 		ASSERT(addrlist != NULL);
    355 		(*cl_sctp_check_addrs)(connp->conn_family, port, &addrlist,
    356 		    size, &addrcnt, unspec == 1);
    357 		if (addrcnt == 0) {
    358 			/* We free the list */
    359 			kmem_free(addrlist, size);
    360 			if (!caller_hold_lock)
    361 				WAKE_SCTP(sctp);
    362 			return (EINVAL);
    363 		}
    364 		if (do_listen) {
    365 			lsize = sizeof (in6_addr_t) * addrcnt;
    366 			llist = kmem_alloc(lsize, KM_SLEEP);
    367 		}
    368 		err = sctp_valid_addr_list(sctp, addrlist, addrcnt, llist,
    369 		    lsize);
    370 		if (err == 0 && do_listen) {
    371 			(*cl_sctp_listen)(connp->conn_family, llist,
    372 			    addrcnt, connp->conn_lport);
    373 			/* list will be freed by the clustering module */
    374 		} else if (err != 0 && llist != NULL) {
    375 			kmem_free(llist, lsize);
    376 		}
    377 		/* free the list we allocated */
    378 		kmem_free(addrlist, size);
    379 	} else {
    380 		err = sctp_valid_addr_list(sctp, addrs, addrcnt, NULL, 0);
    381 	}
    382 	if (err != 0) {
    383 		if (!caller_hold_lock)
    384 			WAKE_SCTP(sctp);
    385 		return (err);
    386 	}
    387 	/* Need to send  ASCONF messages */
    388 	if (do_asconf) {
    389 		err = sctp_add_ip(sctp, addrs, addrcnt);
    390 		if (err != 0) {
    391 			sctp_del_saddr_list(sctp, addrs, addrcnt, B_FALSE);
    392 			if (!caller_hold_lock)
    393 				WAKE_SCTP(sctp);
    394 			return (err);
    395 		}
    396 	}
    397 	if (!caller_hold_lock)
    398 		WAKE_SCTP(sctp);
    399 	return (0);
    400 }
    401 
    402 /*
    403  * Remove one or more addresses bound to the sctp_t.
    404  */
    405 int
    406 sctp_bind_del(sctp_t *sctp, const void *addrs, uint32_t addrcnt,
    407     boolean_t caller_hold_lock)
    408 {
    409 	int		error = 0;
    410 	boolean_t	do_asconf = B_FALSE;
    411 	uchar_t		*ulist = NULL;
    412 	size_t		usize = 0;
    413 	sctp_stack_t	*sctps = sctp->sctp_sctps;
    414 	conn_t		*connp = sctp->sctp_connp;
    415 
    416 	if (!caller_hold_lock)
    417 		RUN_SCTP(sctp);
    418 
    419 	if (sctp->sctp_state > SCTPS_ESTABLISHED ||
    420 	    (sctp->sctp_connp->conn_state_flags & CONN_CLOSING)) {
    421 		if (!caller_hold_lock)
    422 			WAKE_SCTP(sctp);
    423 		return (EINVAL);
    424 	}
    425 	/*
    426 	 * Fail the remove if we are beyond listen, but can't send this
    427 	 * to the peer.
    428 	 */
    429 	if (sctp->sctp_state > SCTPS_LISTEN) {
    430 		if (!sctps->sctps_addip_enabled ||
    431 		    !sctp->sctp_understands_asconf ||
    432 		    !sctp->sctp_understands_addip) {
    433 			if (!caller_hold_lock)
    434 				WAKE_SCTP(sctp);
    435 			return (EINVAL);
    436 		}
    437 		do_asconf = B_TRUE;
    438 	}
    439 
    440 	/* Can't delete the last address nor all of the addresses */
    441 	if (sctp->sctp_nsaddrs == 1 || addrcnt >= sctp->sctp_nsaddrs) {
    442 		if (!caller_hold_lock)
    443 			WAKE_SCTP(sctp);
    444 		return (EINVAL);
    445 	}
    446 
    447 	if (cl_sctp_unlisten != NULL && !do_asconf &&
    448 	    sctp->sctp_state > SCTPS_BOUND) {
    449 		usize = sizeof (in6_addr_t) * addrcnt;
    450 		ulist = kmem_alloc(usize, KM_SLEEP);
    451 	}
    452 
    453 	error = sctp_del_ip(sctp, addrs, addrcnt, ulist, usize);
    454 	if (error != 0) {
    455 		if (ulist != NULL)
    456 			kmem_free(ulist, usize);
    457 		if (!caller_hold_lock)
    458 			WAKE_SCTP(sctp);
    459 		return (error);
    460 	}
    461 	/* ulist will be non-NULL only if cl_sctp_unlisten is non-NULL */
    462 	if (ulist != NULL) {
    463 		ASSERT(cl_sctp_unlisten != NULL);
    464 		(*cl_sctp_unlisten)(connp->conn_family, ulist, addrcnt,
    465 		    connp->conn_lport);
    466 		/* ulist will be freed by the clustering module */
    467 	}
    468 	if (!caller_hold_lock)
    469 		WAKE_SCTP(sctp);
    470 	return (error);
    471 }
    472 
    473 /*
    474  * Returns 0 for success, errno value otherwise.
    475  *
    476  * If the "bind_to_req_port_only" parameter is set and the requested port
    477  * number is available, then set allocated_port to it.  If not available,
    478  * return an error.
    479  *
    480  * If the "bind_to_req_port_only" parameter is not set and the requested port
    481  * number is available, then set allocated_port to it.  If not available,
    482  * find the first anonymous port we can and set allocated_port to that.  If no
    483  * anonymous ports are available, return an error.
    484  *
    485  * In either case, when succeeding, update the sctp_t to record the port number
    486  * and insert it in the bind hash table.
    487  */
    488 int
    489 sctp_bindi(sctp_t *sctp, in_port_t port, boolean_t bind_to_req_port_only,
    490     int user_specified, in_port_t *allocated_port)
    491 {
    492 	/* number of times we have run around the loop */
    493 	int count = 0;
    494 	/* maximum number of times to run around the loop */
    495 	int loopmax;
    496 	sctp_stack_t	*sctps = sctp->sctp_sctps;
    497 	conn_t		*connp = sctp->sctp_connp;
    498 	zone_t *zone = crgetzone(connp->conn_cred);
    499 	zoneid_t zoneid = connp->conn_zoneid;
    500 
    501 	/*
    502 	 * Lookup for free addresses is done in a loop and "loopmax"
    503 	 * influences how long we spin in the loop
    504 	 */
    505 	if (bind_to_req_port_only) {
    506 		/*
    507 		 * If the requested port is busy, don't bother to look
    508 		 * for a new one. Setting loop maximum count to 1 has
    509 		 * that effect.
    510 		 */
    511 		loopmax = 1;
    512 	} else {
    513 		/*
    514 		 * If the requested port is busy, look for a free one
    515 		 * in the anonymous port range.
    516 		 * Set loopmax appropriately so that one does not look
    517 		 * forever in the case all of the anonymous ports are in use.
    518 		 */
    519 		loopmax = (sctps->sctps_largest_anon_port -
    520 		    sctps->sctps_smallest_anon_port + 1);
    521 	}
    522 	do {
    523 		uint16_t	lport;
    524 		sctp_tf_t	*tbf;
    525 		sctp_t		*lsctp;
    526 		int		addrcmp;
    527 
    528 		lport = htons(port);
    529 
    530 		/*
    531 		 * Ensure that the sctp_t is not currently in the bind hash.
    532 		 * Hold the lock on the hash bucket to ensure that
    533 		 * the duplicate check plus the insertion is an atomic
    534 		 * operation.
    535 		 *
    536 		 * This function does an inline lookup on the bind hash list
    537 		 * Make sure that we access only members of sctp_t
    538 		 * and that we don't look at sctp_sctp, since we are not
    539 		 * doing a SCTPB_REFHOLD. For more details please see the notes
    540 		 * in sctp_compress()
    541 		 */
    542 		sctp_bind_hash_remove(sctp);
    543 		tbf = &sctps->sctps_bind_fanout[SCTP_BIND_HASH(port)];
    544 		mutex_enter(&tbf->tf_lock);
    545 		for (lsctp = tbf->tf_sctp; lsctp != NULL;
    546 		    lsctp = lsctp->sctp_bind_hash) {
    547 			conn_t *lconnp = lsctp->sctp_connp;
    548 
    549 			if (lport != lconnp->conn_lport ||
    550 			    lsctp->sctp_state < SCTPS_BOUND)
    551 				continue;
    552 
    553 			/*
    554 			 * On a labeled system, we must treat bindings to ports
    555 			 * on shared IP addresses by sockets with MAC exemption
    556 			 * privilege as being in all zones, as there's
    557 			 * otherwise no way to identify the right receiver.
    558 			 */
    559 			if (lconnp->conn_zoneid != zoneid &&
    560 			    lconnp->conn_mac_mode == CONN_MAC_DEFAULT &&
    561 			    connp->conn_mac_mode == CONN_MAC_DEFAULT)
    562 				continue;
    563 
    564 			addrcmp = sctp_compare_saddrs(sctp, lsctp);
    565 			if (addrcmp != SCTP_ADDR_DISJOINT) {
    566 				if (!connp->conn_reuseaddr) {
    567 					/* in use */
    568 					break;
    569 				} else if (lsctp->sctp_state == SCTPS_BOUND ||
    570 				    lsctp->sctp_state == SCTPS_LISTEN) {
    571 					/*
    572 					 * socket option SO_REUSEADDR is set
    573 					 * on the binding sctp_t.
    574 					 *
    575 					 * We have found a match of IP source
    576 					 * address and source port, which is
    577 					 * refused regardless of the
    578 					 * SO_REUSEADDR setting, so we break.
    579 					 */
    580 					break;
    581 				}
    582 			}
    583 		}
    584 		if (lsctp != NULL) {
    585 			/* The port number is busy */
    586 			mutex_exit(&tbf->tf_lock);
    587 		} else {
    588 			if (is_system_labeled()) {
    589 				mlp_type_t addrtype, mlptype;
    590 				uint_t ipversion;
    591 
    592 				/*
    593 				 * On a labeled system we must check the type
    594 				 * of the binding requested by the user (either
    595 				 * MLP or SLP on shared and private addresses),
    596 				 * and that the user's requested binding
    597 				 * is permitted.
    598 				 */
    599 				if (connp->conn_family == AF_INET)
    600 					ipversion = IPV4_VERSION;
    601 				else
    602 					ipversion = IPV6_VERSION;
    603 
    604 				addrtype = tsol_mlp_addr_type(
    605 				    connp->conn_allzones ? ALL_ZONES :
    606 				    zone->zone_id,
    607 				    ipversion,
    608 				    connp->conn_family == AF_INET ?
    609 				    (void *)&sctp->sctp_ipha->ipha_src :
    610 				    (void *)&sctp->sctp_ip6h->ip6_src,
    611 				    sctps->sctps_netstack->netstack_ip);
    612 
    613 				/*
    614 				 * tsol_mlp_addr_type returns the possibilities
    615 				 * for the selected address.  Since all local
    616 				 * addresses are either private or shared, the
    617 				 * return value mlptSingle means "local address
    618 				 * not valid (interface not present)."
    619 				 */
    620 				if (addrtype == mlptSingle) {
    621 					mutex_exit(&tbf->tf_lock);
    622 					return (EADDRNOTAVAIL);
    623 				}
    624 				mlptype = tsol_mlp_port_type(zone, IPPROTO_SCTP,
    625 				    port, addrtype);
    626 				if (mlptype != mlptSingle) {
    627 					if (secpolicy_net_bindmlp(connp->
    628 					    conn_cred) != 0) {
    629 						mutex_exit(&tbf->tf_lock);
    630 						return (EACCES);
    631 					}
    632 					/*
    633 					 * If we're binding a shared MLP, then
    634 					 * make sure that this zone is the one
    635 					 * that owns that MLP.  Shared MLPs can
    636 					 * be owned by at most one zone.
    637 					 *
    638 					 * No need to handle exclusive-stack
    639 					 * zones since ALL_ZONES only applies
    640 					 * to the shared stack.
    641 					 */
    642 
    643 					if (mlptype == mlptShared &&
    644 					    addrtype == mlptShared &&
    645 					    connp->conn_zoneid !=
    646 					    tsol_mlp_findzone(IPPROTO_SCTP,
    647 					    lport)) {
    648 						mutex_exit(&tbf->tf_lock);
    649 						return (EACCES);
    650 					}
    651 					connp->conn_mlp_type = mlptype;
    652 				}
    653 			}
    654 			/*
    655 			 * This port is ours. Insert in fanout and mark as
    656 			 * bound to prevent others from getting the port
    657 			 * number.
    658 			 */
    659 			sctp->sctp_state = SCTPS_BOUND;
    660 			connp->conn_lport = lport;
    661 
    662 			ASSERT(&sctps->sctps_bind_fanout[
    663 			    SCTP_BIND_HASH(port)] == tbf);
    664 			sctp_bind_hash_insert(tbf, sctp, 1);
    665 
    666 			mutex_exit(&tbf->tf_lock);
    667 
    668 			/*
    669 			 * We don't want sctp_next_port_to_try to "inherit"
    670 			 * a port number supplied by the user in a bind.
    671 			 *
    672 			 * This is the only place where sctp_next_port_to_try
    673 			 * is updated. After the update, it may or may not
    674 			 * be in the valid range.
    675 			 */
    676 			if (user_specified == 0)
    677 				sctps->sctps_next_port_to_try = port + 1;
    678 
    679 			*allocated_port = port;
    680 
    681 			return (0);
    682 		}
    683 
    684 		if ((count == 0) && (user_specified)) {
    685 			/*
    686 			 * We may have to return an anonymous port. So
    687 			 * get one to start with.
    688 			 */
    689 			port = sctp_update_next_port(
    690 			    sctps->sctps_next_port_to_try,
    691 			    zone, sctps);
    692 			user_specified = 0;
    693 		} else {
    694 			port = sctp_update_next_port(port + 1, zone, sctps);
    695 		}
    696 		if (port == 0)
    697 			break;
    698 
    699 		/*
    700 		 * Don't let this loop run forever in the case where
    701 		 * all of the anonymous ports are in use.
    702 		 */
    703 	} while (++count < loopmax);
    704 
    705 	return (bind_to_req_port_only ? EADDRINUSE : EADDRNOTAVAIL);
    706 }
    707 
    708 /*
    709  * Don't let port fall into the privileged range.
    710  * Since the extra privileged ports can be arbitrary we also
    711  * ensure that we exclude those from consideration.
    712  * sctp_g_epriv_ports is not sorted thus we loop over it until
    713  * there are no changes.
    714  *
    715  * Note: No locks are held when inspecting sctp_g_*epriv_ports
    716  * but instead the code relies on:
    717  * - the fact that the address of the array and its size never changes
    718  * - the atomic assignment of the elements of the array
    719  */
    720 in_port_t
    721 sctp_update_next_port(in_port_t port, zone_t *zone, sctp_stack_t *sctps)
    722 {
    723 	int i;
    724 	boolean_t restart = B_FALSE;
    725 
    726 retry:
    727 	if (port < sctps->sctps_smallest_anon_port)
    728 		port = sctps->sctps_smallest_anon_port;
    729 
    730 	if (port > sctps->sctps_largest_anon_port) {
    731 		if (restart)
    732 			return (0);
    733 		restart = B_TRUE;
    734 		port = sctps->sctps_smallest_anon_port;
    735 	}
    736 
    737 	if (port < sctps->sctps_smallest_nonpriv_port)
    738 		port = sctps->sctps_smallest_nonpriv_port;
    739 
    740 	for (i = 0; i < sctps->sctps_g_num_epriv_ports; i++) {
    741 		if (port == sctps->sctps_g_epriv_ports[i]) {
    742 			port++;
    743 			/*
    744 			 * Make sure whether the port is in the
    745 			 * valid range.
    746 			 *
    747 			 * XXX Note that if sctp_g_epriv_ports contains
    748 			 * all the anonymous ports this will be an
    749 			 * infinite loop.
    750 			 */
    751 			goto retry;
    752 		}
    753 	}
    754 
    755 	if (is_system_labeled() &&
    756 	    (i = tsol_next_port(zone, port, IPPROTO_SCTP, B_TRUE)) != 0) {
    757 		port = i;
    758 		goto retry;
    759 	}
    760 
    761 	return (port);
    762 }
    763