xref: /illumos-gate/usr/src/uts/common/inet/sctp/sctp_addr.c (revision 3afe87ebb25691cb6d158edaa34a6fb9b703a691)
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/types.h>
27 #include <sys/systm.h>
28 #include <sys/stream.h>
29 #include <sys/cmn_err.h>
30 #include <sys/ddi.h>
31 #include <sys/sunddi.h>
32 #include <sys/kmem.h>
33 #include <sys/socket.h>
34 #include <sys/sysmacros.h>
35 #include <sys/list.h>
36 
37 #include <netinet/in.h>
38 #include <netinet/ip6.h>
39 #include <netinet/sctp.h>
40 
41 #include <inet/common.h>
42 #include <inet/ip.h>
43 #include <inet/ip6.h>
44 #include <inet/ip_if.h>
45 #include <inet/ipclassifier.h>
46 #include <inet/sctp_ip.h>
47 #include "sctp_impl.h"
48 #include "sctp_addr.h"
49 
50 static void		sctp_ipif_inactive(sctp_ipif_t *);
51 static sctp_ipif_t	*sctp_lookup_ipif_addr(in6_addr_t *, boolean_t,
52 			    zoneid_t, boolean_t, uint_t, uint_t, boolean_t,
53 			    sctp_stack_t *);
54 static int		sctp_get_all_ipifs(sctp_t *, int);
55 static int		sctp_ipif_hash_insert(sctp_t *, sctp_ipif_t *, int,
56 			    boolean_t, boolean_t);
57 static void		sctp_ipif_hash_remove(sctp_t *, sctp_ipif_t *);
58 static void		sctp_fix_saddr(sctp_t *, in6_addr_t *);
59 static int		sctp_compare_ipif_list(sctp_ipif_hash_t *,
60 			    sctp_ipif_hash_t *);
61 static int		sctp_copy_ipifs(sctp_ipif_hash_t *, sctp_t *, int);
62 
63 #define	SCTP_ADDR4_HASH(addr)	\
64 	(((addr) ^ ((addr) >> 8) ^ ((addr) >> 16) ^ ((addr) >> 24)) &	\
65 	(SCTP_IPIF_HASH - 1))
66 
67 #define	SCTP_ADDR6_HASH(addr)	\
68 	(((addr).s6_addr32[3] ^						\
69 	(((addr).s6_addr32[3] ^ (addr).s6_addr32[2]) >> 12)) &		\
70 	(SCTP_IPIF_HASH - 1))
71 
72 #define	SCTP_IPIF_ADDR_HASH(addr, isv6)					\
73 	((isv6) ? SCTP_ADDR6_HASH((addr)) : 				\
74 	SCTP_ADDR4_HASH((addr)._S6_un._S6_u32[3]))
75 
76 #define	SCTP_IPIF_USABLE(sctp_ipif_state)	\
77 	((sctp_ipif_state) == SCTP_IPIFS_UP ||	\
78 	(sctp_ipif_state) ==  SCTP_IPIFS_DOWN)
79 
80 #define	SCTP_IPIF_DISCARD(sctp_ipif_flags)	\
81 	((sctp_ipif_flags) & (IPIF_PRIVATE | IPIF_DEPRECATED))
82 
83 #define	SCTP_IS_IPIF_LOOPBACK(ipif)		\
84 	((ipif)->sctp_ipif_ill->sctp_ill_flags & PHYI_LOOPBACK)
85 
86 #define	SCTP_IS_IPIF_LINKLOCAL(ipif)		\
87 	((ipif)->sctp_ipif_isv6 && 		\
88 	IN6_IS_ADDR_LINKLOCAL(&(ipif)->sctp_ipif_saddr))
89 
90 #define	SCTP_UNSUPP_AF(ipif, supp_af)	\
91 	((!(ipif)->sctp_ipif_isv6 && !((supp_af) & PARM_SUPP_V4)) ||	\
92 	((ipif)->sctp_ipif_isv6 && !((supp_af) & PARM_SUPP_V6)))
93 
94 #define	SCTP_IPIF_ZONE_MATCH(sctp, ipif) 				\
95 	IPCL_ZONE_MATCH((sctp)->sctp_connp, (ipif)->sctp_ipif_zoneid)
96 
97 #define	SCTP_ILL_HASH_FN(index)		((index) % SCTP_ILL_HASH)
98 #define	SCTP_ILL_TO_PHYINDEX(ill)	((ill)->ill_phyint->phyint_ifindex)
99 
100 /*
101  * SCTP Interface list manipulation functions, locking used.
102  */
103 
104 /*
105  * Delete an SCTP IPIF from the list if the refcount goes to 0 and it is
106  * marked as condemned. Also, check if the ILL needs to go away.
107  */
108 static void
109 sctp_ipif_inactive(sctp_ipif_t *sctp_ipif)
110 {
111 	sctp_ill_t	*sctp_ill;
112 	uint_t		hindex;
113 	uint_t		ill_index;
114 	sctp_stack_t	*sctps = sctp_ipif->sctp_ipif_ill->
115 	    sctp_ill_netstack->netstack_sctp;
116 
117 	rw_enter(&sctps->sctps_g_ills_lock, RW_READER);
118 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_WRITER);
119 
120 	hindex = SCTP_IPIF_ADDR_HASH(sctp_ipif->sctp_ipif_saddr,
121 	    sctp_ipif->sctp_ipif_isv6);
122 
123 	sctp_ill = sctp_ipif->sctp_ipif_ill;
124 	ASSERT(sctp_ill != NULL);
125 	ill_index = SCTP_ILL_HASH_FN(sctp_ill->sctp_ill_index);
126 	if (sctp_ipif->sctp_ipif_state != SCTP_IPIFS_CONDEMNED ||
127 	    sctp_ipif->sctp_ipif_refcnt != 0) {
128 		rw_exit(&sctps->sctps_g_ipifs_lock);
129 		rw_exit(&sctps->sctps_g_ills_lock);
130 		return;
131 	}
132 	list_remove(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list,
133 	    sctp_ipif);
134 	sctps->sctps_g_ipifs[hindex].ipif_count--;
135 	sctps->sctps_g_ipifs_count--;
136 	rw_destroy(&sctp_ipif->sctp_ipif_lock);
137 	kmem_free(sctp_ipif, sizeof (sctp_ipif_t));
138 
139 	(void) atomic_add_32_nv(&sctp_ill->sctp_ill_ipifcnt, -1);
140 	if (rw_tryupgrade(&sctps->sctps_g_ills_lock) != 0) {
141 		rw_downgrade(&sctps->sctps_g_ipifs_lock);
142 		if (sctp_ill->sctp_ill_ipifcnt == 0 &&
143 		    sctp_ill->sctp_ill_state == SCTP_ILLS_CONDEMNED) {
144 			list_remove(&sctps->sctps_g_ills[ill_index].
145 			    sctp_ill_list, (void *)sctp_ill);
146 			sctps->sctps_g_ills[ill_index].ill_count--;
147 			sctps->sctps_ills_count--;
148 			kmem_free(sctp_ill->sctp_ill_name,
149 			    sctp_ill->sctp_ill_name_length);
150 			kmem_free(sctp_ill, sizeof (sctp_ill_t));
151 		}
152 	}
153 	rw_exit(&sctps->sctps_g_ipifs_lock);
154 	rw_exit(&sctps->sctps_g_ills_lock);
155 }
156 
157 /*
158  * Lookup an SCTP IPIF given an IP address. Increments sctp_ipif refcnt.
159  * We are either looking for a IPIF with the given address before
160  * inserting it into the global list or looking for an IPIF for an
161  * address given an SCTP. In the former case we always check the zoneid,
162  * but for the latter case, check_zid could be B_FALSE if the connp
163  * for the sctp has conn_all_zones set. When looking for an address we
164  * give preference to one that is up, so even though we may find one that
165  * is not up we keep looking if there is one up, we hold the down addr
166  * in backup_ipif in case we don't find one that is up - i.e. we return
167  * the backup_ipif in that case. Note that if we are looking for. If we
168  * are specifically looking for an up address, then usable will be set
169  * to true.
170  */
171 static sctp_ipif_t *
172 sctp_lookup_ipif_addr(in6_addr_t *addr, boolean_t refhold, zoneid_t zoneid,
173     boolean_t check_zid, uint_t ifindex, uint_t seqid, boolean_t usable,
174     sctp_stack_t *sctps)
175 {
176 	int		j;
177 	sctp_ipif_t	*sctp_ipif;
178 	sctp_ipif_t	*backup_ipif = NULL;
179 	int		hindex;
180 
181 	hindex = SCTP_IPIF_ADDR_HASH(*addr, !IN6_IS_ADDR_V4MAPPED(addr));
182 
183 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_READER);
184 	if (sctps->sctps_g_ipifs[hindex].ipif_count == 0) {
185 		rw_exit(&sctps->sctps_g_ipifs_lock);
186 		return (NULL);
187 	}
188 	sctp_ipif = list_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list);
189 	for (j = 0; j < sctps->sctps_g_ipifs[hindex].ipif_count; j++) {
190 		rw_enter(&sctp_ipif->sctp_ipif_lock, RW_READER);
191 		if ((!check_zid ||
192 		    (sctp_ipif->sctp_ipif_zoneid == ALL_ZONES ||
193 		    zoneid == sctp_ipif->sctp_ipif_zoneid)) &&
194 		    (ifindex == 0 || ifindex ==
195 		    sctp_ipif->sctp_ipif_ill->sctp_ill_index) &&
196 		    ((seqid != 0 && seqid == sctp_ipif->sctp_ipif_id) ||
197 		    (IN6_ARE_ADDR_EQUAL(&sctp_ipif->sctp_ipif_saddr,
198 		    addr)))) {
199 			if (!usable || sctp_ipif->sctp_ipif_state ==
200 			    SCTP_IPIFS_UP) {
201 				rw_exit(&sctp_ipif->sctp_ipif_lock);
202 				if (refhold)
203 					SCTP_IPIF_REFHOLD(sctp_ipif);
204 				rw_exit(&sctps->sctps_g_ipifs_lock);
205 				return (sctp_ipif);
206 			} else if (sctp_ipif->sctp_ipif_state ==
207 			    SCTP_IPIFS_DOWN && backup_ipif == NULL) {
208 				backup_ipif = sctp_ipif;
209 			}
210 		}
211 		rw_exit(&sctp_ipif->sctp_ipif_lock);
212 		sctp_ipif = list_next(
213 		    &sctps->sctps_g_ipifs[hindex].sctp_ipif_list, sctp_ipif);
214 	}
215 	if (backup_ipif != NULL) {
216 		if (refhold)
217 			SCTP_IPIF_REFHOLD(backup_ipif);
218 		rw_exit(&sctps->sctps_g_ipifs_lock);
219 		return (backup_ipif);
220 	}
221 	rw_exit(&sctps->sctps_g_ipifs_lock);
222 	return (NULL);
223 }
224 
225 /*
226  * Populate the list with all the SCTP ipifs for a given ipversion.
227  * Increments sctp_ipif refcnt.
228  * Called with no locks held.
229  */
230 static int
231 sctp_get_all_ipifs(sctp_t *sctp, int sleep)
232 {
233 	sctp_ipif_t		*sctp_ipif;
234 	int			i;
235 	int			j;
236 	int			error = 0;
237 	sctp_stack_t	*sctps = sctp->sctp_sctps;
238 
239 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_READER);
240 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
241 		if (sctps->sctps_g_ipifs[i].ipif_count == 0)
242 			continue;
243 		sctp_ipif = list_head(&sctps->sctps_g_ipifs[i].sctp_ipif_list);
244 		for (j = 0; j < sctps->sctps_g_ipifs[i].ipif_count; j++) {
245 			rw_enter(&sctp_ipif->sctp_ipif_lock, RW_READER);
246 			if (SCTP_IPIF_DISCARD(sctp_ipif->sctp_ipif_flags) ||
247 			    !SCTP_IPIF_USABLE(sctp_ipif->sctp_ipif_state) ||
248 			    !SCTP_IPIF_ZONE_MATCH(sctp, sctp_ipif) ||
249 			    (sctp->sctp_ipversion == IPV4_VERSION &&
250 			    sctp_ipif->sctp_ipif_isv6) ||
251 			    (sctp->sctp_connp->conn_ipv6_v6only &&
252 			    !sctp_ipif->sctp_ipif_isv6)) {
253 				rw_exit(&sctp_ipif->sctp_ipif_lock);
254 				sctp_ipif = list_next(
255 				    &sctps->sctps_g_ipifs[i].sctp_ipif_list,
256 				    sctp_ipif);
257 				continue;
258 			}
259 			rw_exit(&sctp_ipif->sctp_ipif_lock);
260 			SCTP_IPIF_REFHOLD(sctp_ipif);
261 			error = sctp_ipif_hash_insert(sctp, sctp_ipif, sleep,
262 			    B_FALSE, B_FALSE);
263 			if (error != 0 && error != EALREADY)
264 				goto free_stuff;
265 			sctp_ipif = list_next(
266 			    &sctps->sctps_g_ipifs[i].sctp_ipif_list,
267 			    sctp_ipif);
268 		}
269 	}
270 	rw_exit(&sctps->sctps_g_ipifs_lock);
271 	return (0);
272 free_stuff:
273 	rw_exit(&sctps->sctps_g_ipifs_lock);
274 	sctp_free_saddrs(sctp);
275 	return (ENOMEM);
276 }
277 
278 /*
279  * Given a list of address, fills in the list of SCTP ipifs if all the addresses
280  * are present in the SCTP interface list, return number of addresses filled
281  * or error. If the caller wants the list of addresses, it sends a pre-allocated
282  * buffer - list. Currently, this list is only used on a clustered node when
283  * the SCTP is in the listen state (from sctp_bind_add()). When called on a
284  * clustered node, the input is always a list of addresses (even if the
285  * original bind() was to INADDR_ANY).
286  * Called with no locks held.
287  */
288 int
289 sctp_valid_addr_list(sctp_t *sctp, const void *addrs, uint32_t addrcnt,
290     uchar_t *list, size_t lsize)
291 {
292 	struct sockaddr_in	*sin4;
293 	struct sockaddr_in6	*sin6;
294 	struct in_addr		*addr4;
295 	in6_addr_t		addr;
296 	int			cnt;
297 	int			err = 0;
298 	int			saddr_cnt = 0;
299 	sctp_ipif_t		*ipif;
300 	boolean_t		bind_to_all = B_FALSE;
301 	boolean_t		check_addrs = B_FALSE;
302 	boolean_t		check_lport = B_FALSE;
303 	uchar_t			*p = list;
304 
305 	/*
306 	 * Need to check for port and address depending on the state.
307 	 * After a socket is bound, we need to make sure that subsequent
308 	 * bindx() has correct port.  After an association is established,
309 	 * we need to check for changing the bound address to invalid
310 	 * addresses.
311 	 */
312 	if (sctp->sctp_state >= SCTPS_BOUND) {
313 		check_lport = B_TRUE;
314 		if (sctp->sctp_state > SCTPS_LISTEN)
315 			check_addrs = B_TRUE;
316 	}
317 
318 	if (sctp->sctp_conn_tfp != NULL)
319 		mutex_enter(&sctp->sctp_conn_tfp->tf_lock);
320 	if (sctp->sctp_listen_tfp != NULL)
321 		mutex_enter(&sctp->sctp_listen_tfp->tf_lock);
322 	for (cnt = 0; cnt < addrcnt; cnt++) {
323 		boolean_t	lookup_saddr = B_TRUE;
324 		uint_t		ifindex = 0;
325 
326 		switch (sctp->sctp_family) {
327 		case AF_INET:
328 			sin4 = (struct sockaddr_in *)addrs + cnt;
329 			if (sin4->sin_family != AF_INET || (check_lport &&
330 			    sin4->sin_port != sctp->sctp_lport)) {
331 				err = EINVAL;
332 				goto free_ret;
333 			}
334 			addr4 = &sin4->sin_addr;
335 			if (check_addrs &&
336 			    (addr4->s_addr == INADDR_ANY ||
337 			    addr4->s_addr == INADDR_BROADCAST ||
338 			    CLASSD(addr4->s_addr))) {
339 				err = EINVAL;
340 				goto free_ret;
341 			}
342 			IN6_INADDR_TO_V4MAPPED(addr4, &addr);
343 			if (!check_addrs && addr4->s_addr == INADDR_ANY) {
344 				lookup_saddr = B_FALSE;
345 				bind_to_all = B_TRUE;
346 			}
347 
348 			break;
349 		case AF_INET6:
350 			sin6 = (struct sockaddr_in6 *)addrs + cnt;
351 			if (sin6->sin6_family != AF_INET6 || (check_lport &&
352 			    sin6->sin6_port != sctp->sctp_lport)) {
353 				err = EINVAL;
354 				goto free_ret;
355 			}
356 			addr = sin6->sin6_addr;
357 			/* Contains the interface index */
358 			ifindex = sin6->sin6_scope_id;
359 			if (sctp->sctp_connp->conn_ipv6_v6only &&
360 			    IN6_IS_ADDR_V4MAPPED(&addr)) {
361 				err = EAFNOSUPPORT;
362 				goto free_ret;
363 			}
364 			if (check_addrs &&
365 			    (IN6_IS_ADDR_LINKLOCAL(&addr) ||
366 			    IN6_IS_ADDR_MULTICAST(&addr) ||
367 			    IN6_IS_ADDR_UNSPECIFIED(&addr))) {
368 				err = EINVAL;
369 				goto free_ret;
370 			}
371 			if (!check_addrs && IN6_IS_ADDR_UNSPECIFIED(&addr)) {
372 				lookup_saddr = B_FALSE;
373 				bind_to_all = B_TRUE;
374 			}
375 
376 			break;
377 		default:
378 			err = EAFNOSUPPORT;
379 			goto free_ret;
380 		}
381 		if (lookup_saddr) {
382 			ipif = sctp_lookup_ipif_addr(&addr, B_TRUE,
383 			    sctp->sctp_zoneid, !sctp->sctp_connp->conn_allzones,
384 			    ifindex, 0, B_TRUE, sctp->sctp_sctps);
385 			if (ipif == NULL) {
386 				/* Address not in the list */
387 				err = EINVAL;
388 				goto free_ret;
389 			} else if (check_addrs && SCTP_IS_IPIF_LOOPBACK(ipif) &&
390 			    cl_sctp_check_addrs == NULL) {
391 				SCTP_IPIF_REFRELE(ipif);
392 				err = EINVAL;
393 				goto free_ret;
394 			}
395 		}
396 		if (!bind_to_all) {
397 			/*
398 			 * If an address is added after association setup,
399 			 * we need to wait for the peer to send us an ASCONF
400 			 * ACK before we can start using it.
401 			 * saddr_ipif_dontsrc will be reset (to 0) when we
402 			 * get the ASCONF ACK for this address.
403 			 */
404 			err = sctp_ipif_hash_insert(sctp, ipif, KM_SLEEP,
405 			    check_addrs ? B_TRUE : B_FALSE, B_FALSE);
406 			if (err != 0) {
407 				SCTP_IPIF_REFRELE(ipif);
408 				if (check_addrs && err == EALREADY)
409 					err = EADDRINUSE;
410 				goto free_ret;
411 			}
412 			saddr_cnt++;
413 			if (lsize >= sizeof (addr)) {
414 				bcopy(&addr, p, sizeof (addr));
415 				p += sizeof (addr);
416 				lsize -= sizeof (addr);
417 			}
418 		}
419 	}
420 	if (bind_to_all) {
421 		/*
422 		 * Free whatever we might have added before encountering
423 		 * inaddr_any.
424 		 */
425 		if (sctp->sctp_nsaddrs > 0) {
426 			sctp_free_saddrs(sctp);
427 			ASSERT(sctp->sctp_nsaddrs == 0);
428 		}
429 		err = sctp_get_all_ipifs(sctp, KM_SLEEP);
430 		if (err != 0)
431 			return (err);
432 		sctp->sctp_bound_to_all = 1;
433 	}
434 	if (sctp->sctp_listen_tfp != NULL)
435 		mutex_exit(&sctp->sctp_listen_tfp->tf_lock);
436 	if (sctp->sctp_conn_tfp != NULL)
437 		mutex_exit(&sctp->sctp_conn_tfp->tf_lock);
438 	return (0);
439 free_ret:
440 	if (saddr_cnt != 0)
441 		sctp_del_saddr_list(sctp, addrs, saddr_cnt, B_TRUE);
442 	if (sctp->sctp_listen_tfp != NULL)
443 		mutex_exit(&sctp->sctp_listen_tfp->tf_lock);
444 	if (sctp->sctp_conn_tfp != NULL)
445 		mutex_exit(&sctp->sctp_conn_tfp->tf_lock);
446 	return (err);
447 }
448 
449 static int
450 sctp_ipif_hash_insert(sctp_t *sctp, sctp_ipif_t *ipif, int sleep,
451     boolean_t dontsrc, boolean_t allow_dup)
452 {
453 	int			cnt;
454 	sctp_saddr_ipif_t	*ipif_obj;
455 	int			hindex;
456 
457 	hindex = SCTP_IPIF_ADDR_HASH(ipif->sctp_ipif_saddr,
458 	    ipif->sctp_ipif_isv6);
459 	ipif_obj = list_head(&sctp->sctp_saddrs[hindex].sctp_ipif_list);
460 	for (cnt = 0; cnt < sctp->sctp_saddrs[hindex].ipif_count; cnt++) {
461 		if (IN6_ARE_ADDR_EQUAL(&ipif_obj->saddr_ipifp->sctp_ipif_saddr,
462 		    &ipif->sctp_ipif_saddr)) {
463 			if (ipif->sctp_ipif_id !=
464 			    ipif_obj->saddr_ipifp->sctp_ipif_id &&
465 			    ipif_obj->saddr_ipifp->sctp_ipif_state ==
466 			    SCTP_IPIFS_DOWN && ipif->sctp_ipif_state ==
467 			    SCTP_IPIFS_UP) {
468 				SCTP_IPIF_REFRELE(ipif_obj->saddr_ipifp);
469 				ipif_obj->saddr_ipifp = ipif;
470 				ipif_obj->saddr_ipif_dontsrc = dontsrc ? 1 : 0;
471 				return (0);
472 			} else if (!allow_dup || ipif->sctp_ipif_id ==
473 			    ipif_obj->saddr_ipifp->sctp_ipif_id) {
474 				return (EALREADY);
475 			}
476 		}
477 		ipif_obj = list_next(&sctp->sctp_saddrs[hindex].sctp_ipif_list,
478 		    ipif_obj);
479 	}
480 	ipif_obj = kmem_zalloc(sizeof (sctp_saddr_ipif_t), sleep);
481 	if (ipif_obj == NULL) {
482 		/* Need to do something */
483 		return (ENOMEM);
484 	}
485 	ipif_obj->saddr_ipifp = ipif;
486 	ipif_obj->saddr_ipif_dontsrc = dontsrc ? 1 : 0;
487 	list_insert_tail(&sctp->sctp_saddrs[hindex].sctp_ipif_list, ipif_obj);
488 	sctp->sctp_saddrs[hindex].ipif_count++;
489 	sctp->sctp_nsaddrs++;
490 	return (0);
491 }
492 
493 /*
494  * Given a source address, walk through the peer address list to see
495  * if the source address is being used.  If it is, reset that.
496  */
497 static void
498 sctp_fix_saddr(sctp_t *sctp, in6_addr_t *saddr)
499 {
500 	sctp_faddr_t	*fp;
501 
502 	for (fp = sctp->sctp_faddrs; fp != NULL; fp = fp->next) {
503 		if (!IN6_ARE_ADDR_EQUAL(&fp->saddr, saddr))
504 			continue;
505 		if (fp->ire != NULL) {
506 			IRE_REFRELE_NOTR(fp->ire);
507 			fp->ire = NULL;
508 		}
509 		V6_SET_ZERO(fp->saddr);
510 	}
511 }
512 
513 static void
514 sctp_ipif_hash_remove(sctp_t *sctp, sctp_ipif_t *ipif)
515 {
516 	int			cnt;
517 	sctp_saddr_ipif_t	*ipif_obj;
518 	int			hindex;
519 
520 	hindex = SCTP_IPIF_ADDR_HASH(ipif->sctp_ipif_saddr,
521 	    ipif->sctp_ipif_isv6);
522 	ipif_obj = list_head(&sctp->sctp_saddrs[hindex].sctp_ipif_list);
523 	for (cnt = 0; cnt < sctp->sctp_saddrs[hindex].ipif_count; cnt++) {
524 		if (IN6_ARE_ADDR_EQUAL(&ipif_obj->saddr_ipifp->sctp_ipif_saddr,
525 		    &ipif->sctp_ipif_saddr)) {
526 			list_remove(&sctp->sctp_saddrs[hindex].sctp_ipif_list,
527 			    ipif_obj);
528 			sctp->sctp_saddrs[hindex].ipif_count--;
529 			sctp->sctp_nsaddrs--;
530 			sctp_fix_saddr(sctp, &ipif->sctp_ipif_saddr);
531 			SCTP_IPIF_REFRELE(ipif_obj->saddr_ipifp);
532 			kmem_free(ipif_obj, sizeof (sctp_saddr_ipif_t));
533 			break;
534 		}
535 		ipif_obj = list_next(&sctp->sctp_saddrs[hindex].sctp_ipif_list,
536 		    ipif_obj);
537 	}
538 }
539 
540 static int
541 sctp_compare_ipif_list(sctp_ipif_hash_t *list1, sctp_ipif_hash_t *list2)
542 {
543 	int			i;
544 	int			j;
545 	sctp_saddr_ipif_t	*obj1;
546 	sctp_saddr_ipif_t	*obj2;
547 	int			overlap = 0;
548 
549 	obj1 = list_head(&list1->sctp_ipif_list);
550 	for (i = 0; i < list1->ipif_count; i++) {
551 		obj2 = list_head(&list2->sctp_ipif_list);
552 		for (j = 0; j < list2->ipif_count; j++) {
553 			if (IN6_ARE_ADDR_EQUAL(
554 			    &obj1->saddr_ipifp->sctp_ipif_saddr,
555 			    &obj2->saddr_ipifp->sctp_ipif_saddr)) {
556 				overlap++;
557 				break;
558 			}
559 			obj2 = list_next(&list2->sctp_ipif_list,
560 			    obj2);
561 		}
562 		obj1 = list_next(&list1->sctp_ipif_list, obj1);
563 	}
564 	return (overlap);
565 }
566 
567 int
568 sctp_compare_saddrs(sctp_t *sctp1, sctp_t *sctp2)
569 {
570 	int		i;
571 	int		overlap = 0;
572 
573 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
574 		overlap += sctp_compare_ipif_list(&sctp1->sctp_saddrs[i],
575 		    &sctp2->sctp_saddrs[i]);
576 	}
577 
578 	if (sctp1->sctp_nsaddrs == sctp2->sctp_nsaddrs &&
579 	    overlap == sctp1->sctp_nsaddrs) {
580 		return (SCTP_ADDR_EQUAL);
581 	}
582 
583 	if (overlap == sctp1->sctp_nsaddrs)
584 		return (SCTP_ADDR_SUBSET);
585 
586 	if (overlap > 0)
587 		return (SCTP_ADDR_OVERLAP);
588 
589 	return (SCTP_ADDR_DISJOINT);
590 }
591 
592 static int
593 sctp_copy_ipifs(sctp_ipif_hash_t *list1, sctp_t *sctp2, int sleep)
594 {
595 	int			i;
596 	sctp_saddr_ipif_t	*obj;
597 	int			error = 0;
598 
599 	obj = list_head(&list1->sctp_ipif_list);
600 	for (i = 0; i < list1->ipif_count; i++) {
601 		SCTP_IPIF_REFHOLD(obj->saddr_ipifp);
602 		error = sctp_ipif_hash_insert(sctp2, obj->saddr_ipifp, sleep,
603 		    B_FALSE, B_FALSE);
604 		ASSERT(error != EALREADY);
605 		if (error != 0)
606 			return (error);
607 		obj = list_next(&list1->sctp_ipif_list, obj);
608 	}
609 	return (error);
610 }
611 
612 int
613 sctp_dup_saddrs(sctp_t *sctp1, sctp_t *sctp2, int sleep)
614 {
615 	int	error = 0;
616 	int	i;
617 
618 	if (sctp1 == NULL || sctp1->sctp_bound_to_all == 1)
619 		return (sctp_get_all_ipifs(sctp2, sleep));
620 
621 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
622 		if (sctp1->sctp_saddrs[i].ipif_count == 0)
623 			continue;
624 		error = sctp_copy_ipifs(&sctp1->sctp_saddrs[i], sctp2, sleep);
625 		if (error != 0) {
626 			sctp_free_saddrs(sctp2);
627 			return (error);
628 		}
629 	}
630 	return (0);
631 }
632 
633 void
634 sctp_free_saddrs(sctp_t *sctp)
635 {
636 	int			i;
637 	int			l;
638 	sctp_saddr_ipif_t	*obj;
639 
640 	if (sctp->sctp_nsaddrs == 0)
641 		return;
642 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
643 		if (sctp->sctp_saddrs[i].ipif_count == 0)
644 			continue;
645 		obj = list_tail(&sctp->sctp_saddrs[i].sctp_ipif_list);
646 		for (l = 0; l < sctp->sctp_saddrs[i].ipif_count; l++) {
647 			list_remove(&sctp->sctp_saddrs[i].sctp_ipif_list, obj);
648 			SCTP_IPIF_REFRELE(obj->saddr_ipifp);
649 			sctp->sctp_nsaddrs--;
650 			kmem_free(obj, sizeof (sctp_saddr_ipif_t));
651 			obj = list_tail(&sctp->sctp_saddrs[i].sctp_ipif_list);
652 		}
653 		sctp->sctp_saddrs[i].ipif_count = 0;
654 	}
655 	if (sctp->sctp_bound_to_all == 1)
656 		sctp->sctp_bound_to_all = 0;
657 	ASSERT(sctp->sctp_nsaddrs == 0);
658 }
659 
660 /*
661  * Add/Delete the given ILL from the SCTP ILL list. Called with no locks
662  * held.
663  */
664 void
665 sctp_update_ill(ill_t *ill, int op)
666 {
667 	int		i;
668 	sctp_ill_t	*sctp_ill = NULL;
669 	uint_t		index;
670 	netstack_t	*ns = ill->ill_ipst->ips_netstack;
671 	sctp_stack_t	*sctps = ns->netstack_sctp;
672 
673 	rw_enter(&sctps->sctps_g_ills_lock, RW_WRITER);
674 
675 	index = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(ill));
676 	sctp_ill = list_head(&sctps->sctps_g_ills[index].sctp_ill_list);
677 	for (i = 0; i < sctps->sctps_g_ills[index].ill_count; i++) {
678 		if ((sctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(ill)) &&
679 		    (sctp_ill->sctp_ill_isv6 == ill->ill_isv6)) {
680 			break;
681 		}
682 		sctp_ill = list_next(&sctps->sctps_g_ills[index].sctp_ill_list,
683 		    sctp_ill);
684 	}
685 
686 	switch (op) {
687 	case SCTP_ILL_INSERT:
688 		if (sctp_ill != NULL) {
689 			/* Unmark it if it is condemned */
690 			if (sctp_ill->sctp_ill_state == SCTP_ILLS_CONDEMNED)
691 				sctp_ill->sctp_ill_state = 0;
692 			rw_exit(&sctps->sctps_g_ills_lock);
693 			return;
694 		}
695 		sctp_ill = kmem_zalloc(sizeof (sctp_ill_t), KM_NOSLEEP);
696 		/* Need to re-try? */
697 		if (sctp_ill == NULL) {
698 			cmn_err(CE_WARN, "sctp_update_ill: error adding "
699 			    "ILL %p to SCTP's ILL list", (void *)ill);
700 			rw_exit(&sctps->sctps_g_ills_lock);
701 			return;
702 		}
703 		sctp_ill->sctp_ill_name = kmem_zalloc(ill->ill_name_length,
704 		    KM_NOSLEEP);
705 		if (sctp_ill->sctp_ill_name == NULL) {
706 			cmn_err(CE_WARN, "sctp_update_ill: error adding "
707 			    "ILL %p to SCTP's ILL list", (void *)ill);
708 			kmem_free(sctp_ill, sizeof (sctp_ill_t));
709 			rw_exit(&sctps->sctps_g_ills_lock);
710 			return;
711 		}
712 		bcopy(ill->ill_name, sctp_ill->sctp_ill_name,
713 		    ill->ill_name_length);
714 		sctp_ill->sctp_ill_name_length = ill->ill_name_length;
715 		sctp_ill->sctp_ill_index = SCTP_ILL_TO_PHYINDEX(ill);
716 		sctp_ill->sctp_ill_flags = ill->ill_phyint->phyint_flags;
717 		sctp_ill->sctp_ill_netstack = ns;	/* No netstack_hold */
718 		sctp_ill->sctp_ill_isv6 = ill->ill_isv6;
719 		list_insert_tail(&sctps->sctps_g_ills[index].sctp_ill_list,
720 		    (void *)sctp_ill);
721 		sctps->sctps_g_ills[index].ill_count++;
722 		sctps->sctps_ills_count++;
723 
724 		break;
725 
726 	case SCTP_ILL_REMOVE:
727 
728 		if (sctp_ill == NULL) {
729 			rw_exit(&sctps->sctps_g_ills_lock);
730 			return;
731 		}
732 		if (sctp_ill->sctp_ill_ipifcnt == 0) {
733 			list_remove(&sctps->sctps_g_ills[index].sctp_ill_list,
734 			    (void *)sctp_ill);
735 			sctps->sctps_g_ills[index].ill_count--;
736 			sctps->sctps_ills_count--;
737 			kmem_free(sctp_ill->sctp_ill_name,
738 			    ill->ill_name_length);
739 			kmem_free(sctp_ill, sizeof (sctp_ill_t));
740 		} else {
741 			sctp_ill->sctp_ill_state = SCTP_ILLS_CONDEMNED;
742 		}
743 
744 		break;
745 	}
746 	rw_exit(&sctps->sctps_g_ills_lock);
747 }
748 
749 /*
750  * The ILL's index is being changed, just remove it from the old list,
751  * change the SCTP ILL's index and re-insert using the new index.
752  */
753 void
754 sctp_ill_reindex(ill_t *ill, uint_t orig_ill_index)
755 {
756 	sctp_ill_t	*sctp_ill = NULL;
757 	sctp_ill_t	*nxt_sill;
758 	uint_t		indx;
759 	uint_t		nindx;
760 	boolean_t	once = B_FALSE;
761 	netstack_t	*ns = ill->ill_ipst->ips_netstack;
762 	sctp_stack_t	*sctps = ns->netstack_sctp;
763 
764 	rw_enter(&sctps->sctps_g_ills_lock, RW_WRITER);
765 
766 	indx = SCTP_ILL_HASH_FN(orig_ill_index);
767 	nindx = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(ill));
768 	sctp_ill = list_head(&sctps->sctps_g_ills[indx].sctp_ill_list);
769 	while (sctp_ill != NULL) {
770 		nxt_sill = list_next(&sctps->sctps_g_ills[indx].sctp_ill_list,
771 		    sctp_ill);
772 		if (sctp_ill->sctp_ill_index == orig_ill_index) {
773 			sctp_ill->sctp_ill_index = SCTP_ILL_TO_PHYINDEX(ill);
774 			/*
775 			 * if the new index hashes to the same value, all's
776 			 * done.
777 			 */
778 			if (nindx != indx) {
779 				list_remove(
780 				    &sctps->sctps_g_ills[indx].sctp_ill_list,
781 				    (void *)sctp_ill);
782 				sctps->sctps_g_ills[indx].ill_count--;
783 				list_insert_tail(
784 				    &sctps->sctps_g_ills[nindx].sctp_ill_list,
785 				    (void *)sctp_ill);
786 				sctps->sctps_g_ills[nindx].ill_count++;
787 			}
788 			if (once)
789 				break;
790 			/* We might have one for v4 and for v6 */
791 			once = B_TRUE;
792 		}
793 		sctp_ill = nxt_sill;
794 	}
795 	rw_exit(&sctps->sctps_g_ills_lock);
796 }
797 
798 /* move ipif from f_ill to t_ill */
799 void
800 sctp_move_ipif(ipif_t *ipif, ill_t *f_ill, ill_t *t_ill)
801 {
802 	sctp_ill_t	*fsctp_ill = NULL;
803 	sctp_ill_t	*tsctp_ill = NULL;
804 	sctp_ipif_t	*sctp_ipif;
805 	uint_t		hindex;
806 	int		i;
807 	netstack_t	*ns = ipif->ipif_ill->ill_ipst->ips_netstack;
808 	sctp_stack_t	*sctps = ns->netstack_sctp;
809 
810 	rw_enter(&sctps->sctps_g_ills_lock, RW_READER);
811 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_READER);
812 
813 	hindex = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(f_ill));
814 	fsctp_ill = list_head(&sctps->sctps_g_ills[hindex].sctp_ill_list);
815 	for (i = 0; i < sctps->sctps_g_ills[hindex].ill_count; i++) {
816 		if (fsctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(f_ill) &&
817 		    fsctp_ill->sctp_ill_isv6 == f_ill->ill_isv6) {
818 			break;
819 		}
820 		fsctp_ill = list_next(
821 		    &sctps->sctps_g_ills[hindex].sctp_ill_list, fsctp_ill);
822 	}
823 
824 	hindex = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(t_ill));
825 	tsctp_ill = list_head(&sctps->sctps_g_ills[hindex].sctp_ill_list);
826 	for (i = 0; i < sctps->sctps_g_ills[hindex].ill_count; i++) {
827 		if (tsctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(t_ill) &&
828 		    tsctp_ill->sctp_ill_isv6 == t_ill->ill_isv6) {
829 			break;
830 		}
831 		tsctp_ill = list_next(
832 		    &sctps->sctps_g_ills[hindex].sctp_ill_list, tsctp_ill);
833 	}
834 
835 	hindex = SCTP_IPIF_ADDR_HASH(ipif->ipif_v6lcl_addr,
836 	    ipif->ipif_ill->ill_isv6);
837 	sctp_ipif = list_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list);
838 	for (i = 0; i < sctps->sctps_g_ipifs[hindex].ipif_count; i++) {
839 		if (sctp_ipif->sctp_ipif_id == ipif->ipif_seqid)
840 			break;
841 		sctp_ipif = list_next(
842 		    &sctps->sctps_g_ipifs[hindex].sctp_ipif_list, sctp_ipif);
843 	}
844 	/* Should be an ASSERT? */
845 	if (fsctp_ill == NULL || tsctp_ill == NULL || sctp_ipif == NULL) {
846 		ip1dbg(("sctp_move_ipif: error moving ipif %p from %p to %p\n",
847 		    (void *)ipif, (void *)f_ill, (void *)t_ill));
848 		rw_exit(&sctps->sctps_g_ipifs_lock);
849 		rw_exit(&sctps->sctps_g_ills_lock);
850 		return;
851 	}
852 	rw_enter(&sctp_ipif->sctp_ipif_lock, RW_WRITER);
853 	ASSERT(sctp_ipif->sctp_ipif_ill == fsctp_ill);
854 	sctp_ipif->sctp_ipif_ill = tsctp_ill;
855 	rw_exit(&sctp_ipif->sctp_ipif_lock);
856 	(void) atomic_add_32_nv(&fsctp_ill->sctp_ill_ipifcnt, -1);
857 	atomic_add_32(&tsctp_ill->sctp_ill_ipifcnt, 1);
858 	rw_exit(&sctps->sctps_g_ipifs_lock);
859 	rw_exit(&sctps->sctps_g_ills_lock);
860 }
861 
862 /*
863  * Walk the list of SCTPs and find each that has oipif in it's saddr list, and
864  * if so replace it with nipif.
865  */
866 void
867 sctp_update_saddrs(sctp_ipif_t *oipif, sctp_ipif_t *nipif, int idx,
868     sctp_stack_t *sctps)
869 {
870 	sctp_t			*sctp;
871 	sctp_t			*sctp_prev = NULL;
872 	sctp_saddr_ipif_t	*sobj;
873 	int			count;
874 
875 	sctp = sctps->sctps_gsctp;
876 	mutex_enter(&sctps->sctps_g_lock);
877 	while (sctp != NULL && oipif->sctp_ipif_refcnt > 0) {
878 		mutex_enter(&sctp->sctp_reflock);
879 		if (sctp->sctp_condemned ||
880 		    sctp->sctp_saddrs[idx].ipif_count <= 0) {
881 			mutex_exit(&sctp->sctp_reflock);
882 			sctp = list_next(&sctps->sctps_g_list, sctp);
883 			continue;
884 		}
885 		sctp->sctp_refcnt++;
886 		mutex_exit(&sctp->sctp_reflock);
887 		mutex_exit(&sctps->sctps_g_lock);
888 		if (sctp_prev != NULL)
889 			SCTP_REFRELE(sctp_prev);
890 
891 		RUN_SCTP(sctp);
892 		sobj = list_head(&sctp->sctp_saddrs[idx].sctp_ipif_list);
893 		for (count = 0; count <
894 		    sctp->sctp_saddrs[idx].ipif_count; count++) {
895 			if (sobj->saddr_ipifp == oipif) {
896 				SCTP_IPIF_REFHOLD(nipif);
897 				sobj->saddr_ipifp = nipif;
898 				ASSERT(oipif->sctp_ipif_refcnt > 0);
899 				/* We have the writer lock */
900 				oipif->sctp_ipif_refcnt--;
901 				/*
902 				 * Can't have more than one referring
903 				 * to the same sctp_ipif.
904 				 */
905 				break;
906 			}
907 			sobj = list_next(&sctp->sctp_saddrs[idx].sctp_ipif_list,
908 			    sobj);
909 		}
910 		WAKE_SCTP(sctp);
911 		sctp_prev = sctp;
912 		mutex_enter(&sctps->sctps_g_lock);
913 		sctp = list_next(&sctps->sctps_g_list, sctp);
914 	}
915 	mutex_exit(&sctps->sctps_g_lock);
916 	if (sctp_prev != NULL)
917 		SCTP_REFRELE(sctp_prev);
918 }
919 
920 /*
921  * Given an ipif, walk the hash list in the global ipif table and for
922  * any other SCTP ipif with the same address and non-zero reference, walk
923  * the SCTP list and update the saddr list, if required, to point to the
924  * new SCTP ipif.
925  */
926 void
927 sctp_chk_and_updt_saddr(int hindex, sctp_ipif_t *ipif, sctp_stack_t *sctps)
928 {
929 	int		cnt;
930 	sctp_ipif_t	*sipif;
931 
932 	ASSERT(sctps->sctps_g_ipifs[hindex].ipif_count > 0);
933 	ASSERT(ipif->sctp_ipif_state == SCTP_IPIFS_UP);
934 
935 	sipif = list_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list);
936 	for (cnt = 0; cnt < sctps->sctps_g_ipifs[hindex].ipif_count; cnt++) {
937 		rw_enter(&sipif->sctp_ipif_lock, RW_WRITER);
938 		if (sipif->sctp_ipif_id != ipif->sctp_ipif_id &&
939 		    IN6_ARE_ADDR_EQUAL(&sipif->sctp_ipif_saddr,
940 		    &ipif->sctp_ipif_saddr) && sipif->sctp_ipif_refcnt > 0) {
941 			/*
942 			 * There can only be one address up at any time
943 			 * and we are here because ipif has been brought
944 			 * up.
945 			 */
946 			ASSERT(sipif->sctp_ipif_state != SCTP_IPIFS_UP);
947 			/*
948 			 * Someone has a reference to this we need to update to
949 			 * point to the new sipif.
950 			 */
951 			sctp_update_saddrs(sipif, ipif, hindex, sctps);
952 		}
953 		rw_exit(&sipif->sctp_ipif_lock);
954 		sipif = list_next(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list,
955 		    sipif);
956 	}
957 }
958 
959 /*
960  * Insert a new SCTP ipif using 'ipif'. v6addr is the address that existed
961  * prior to the current address in 'ipif'. Only when an existing address
962  * is changed on an IPIF, will v6addr be specified. If the IPIF already
963  * exists in the global SCTP ipif table, then we either removed it, if
964  * it doesn't have any existing reference, or mark it condemned otherwise.
965  * If an address is being brought up (IPIF_UP), then we need to scan
966  * the SCTP list to check if there is any SCTP that points to the *same*
967  * address on a different SCTP ipif and update in that case.
968  */
969 void
970 sctp_update_ipif_addr(ipif_t *ipif, in6_addr_t v6addr)
971 {
972 	ill_t		*ill = ipif->ipif_ill;
973 	int		i;
974 	sctp_ill_t	*sctp_ill;
975 	sctp_ill_t	*osctp_ill;
976 	sctp_ipif_t	*sctp_ipif = NULL;
977 	sctp_ipif_t	*osctp_ipif = NULL;
978 	uint_t		ill_index;
979 	int		hindex;
980 	sctp_stack_t	*sctps;
981 
982 
983 	sctps = ipif->ipif_ill->ill_ipst->ips_netstack->netstack_sctp;
984 
985 	/* Index for new address */
986 	hindex = SCTP_IPIF_ADDR_HASH(ipif->ipif_v6lcl_addr, ill->ill_isv6);
987 
988 	/*
989 	 * The address on this IPIF is changing, we need to look for
990 	 * this old address and mark it condemned, before creating
991 	 * one for the new address.
992 	 */
993 	osctp_ipif = sctp_lookup_ipif_addr(&v6addr, B_FALSE,
994 	    ipif->ipif_zoneid, B_TRUE, SCTP_ILL_TO_PHYINDEX(ill),
995 	    ipif->ipif_seqid, B_FALSE, sctps);
996 
997 	rw_enter(&sctps->sctps_g_ills_lock, RW_READER);
998 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_WRITER);
999 
1000 	ill_index = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(ill));
1001 	sctp_ill = list_head(&sctps->sctps_g_ills[ill_index].sctp_ill_list);
1002 	for (i = 0; i < sctps->sctps_g_ills[ill_index].ill_count; i++) {
1003 		if (sctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(ill) &&
1004 		    sctp_ill->sctp_ill_isv6 == ill->ill_isv6) {
1005 			break;
1006 		}
1007 		sctp_ill = list_next(
1008 		    &sctps->sctps_g_ills[ill_index].sctp_ill_list, sctp_ill);
1009 	}
1010 
1011 	if (sctp_ill == NULL) {
1012 		ip1dbg(("sctp_update_ipif_addr: ill not found ..\n"));
1013 		rw_exit(&sctps->sctps_g_ipifs_lock);
1014 		rw_exit(&sctps->sctps_g_ills_lock);
1015 		return;
1016 	}
1017 
1018 	if (osctp_ipif != NULL) {
1019 
1020 		/* The address is the same? */
1021 		if (IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6lcl_addr, &v6addr)) {
1022 			boolean_t	chk_n_updt = B_FALSE;
1023 
1024 			rw_downgrade(&sctps->sctps_g_ipifs_lock);
1025 			rw_enter(&osctp_ipif->sctp_ipif_lock, RW_WRITER);
1026 			if (ipif->ipif_flags & IPIF_UP &&
1027 			    osctp_ipif->sctp_ipif_state != SCTP_IPIFS_UP) {
1028 				osctp_ipif->sctp_ipif_state = SCTP_IPIFS_UP;
1029 				chk_n_updt = B_TRUE;
1030 			} else {
1031 				osctp_ipif->sctp_ipif_state = SCTP_IPIFS_DOWN;
1032 			}
1033 			osctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1034 			rw_exit(&osctp_ipif->sctp_ipif_lock);
1035 			if (chk_n_updt) {
1036 				sctp_chk_and_updt_saddr(hindex, osctp_ipif,
1037 				    sctps);
1038 			}
1039 			rw_exit(&sctps->sctps_g_ipifs_lock);
1040 			rw_exit(&sctps->sctps_g_ills_lock);
1041 			return;
1042 		}
1043 		/*
1044 		 * We are effectively removing this address from the ILL.
1045 		 */
1046 		if (osctp_ipif->sctp_ipif_refcnt != 0) {
1047 			osctp_ipif->sctp_ipif_state = SCTP_IPIFS_CONDEMNED;
1048 		} else {
1049 			list_t		*ipif_list;
1050 			int		ohindex;
1051 
1052 			osctp_ill = osctp_ipif->sctp_ipif_ill;
1053 			/* hash index for the old one */
1054 			ohindex = SCTP_IPIF_ADDR_HASH(
1055 			    osctp_ipif->sctp_ipif_saddr,
1056 			    osctp_ipif->sctp_ipif_isv6);
1057 
1058 			ipif_list =
1059 			    &sctps->sctps_g_ipifs[ohindex].sctp_ipif_list;
1060 
1061 			list_remove(ipif_list, (void *)osctp_ipif);
1062 			sctps->sctps_g_ipifs[ohindex].ipif_count--;
1063 			sctps->sctps_g_ipifs_count--;
1064 			rw_destroy(&osctp_ipif->sctp_ipif_lock);
1065 			kmem_free(osctp_ipif, sizeof (sctp_ipif_t));
1066 			(void) atomic_add_32_nv(&osctp_ill->sctp_ill_ipifcnt,
1067 			    -1);
1068 		}
1069 	}
1070 
1071 	sctp_ipif = kmem_zalloc(sizeof (sctp_ipif_t), KM_NOSLEEP);
1072 	/* Try again? */
1073 	if (sctp_ipif == NULL) {
1074 		cmn_err(CE_WARN, "sctp_update_ipif_addr: error adding "
1075 		    "IPIF %p to SCTP's IPIF list", (void *)ipif);
1076 		rw_exit(&sctps->sctps_g_ipifs_lock);
1077 		rw_exit(&sctps->sctps_g_ills_lock);
1078 		return;
1079 	}
1080 	sctps->sctps_g_ipifs_count++;
1081 	rw_init(&sctp_ipif->sctp_ipif_lock, NULL, RW_DEFAULT, NULL);
1082 	sctp_ipif->sctp_ipif_saddr = ipif->ipif_v6lcl_addr;
1083 	sctp_ipif->sctp_ipif_ill = sctp_ill;
1084 	sctp_ipif->sctp_ipif_isv6 = ill->ill_isv6;
1085 	sctp_ipif->sctp_ipif_zoneid = ipif->ipif_zoneid;
1086 	sctp_ipif->sctp_ipif_id = ipif->ipif_seqid;
1087 	if (ipif->ipif_flags & IPIF_UP)
1088 		sctp_ipif->sctp_ipif_state = SCTP_IPIFS_UP;
1089 	else
1090 		sctp_ipif->sctp_ipif_state = SCTP_IPIFS_DOWN;
1091 	sctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1092 	/*
1093 	 * We add it to the head so that it is quicker to find good/recent
1094 	 * additions.
1095 	 */
1096 	list_insert_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list,
1097 	    (void *)sctp_ipif);
1098 	sctps->sctps_g_ipifs[hindex].ipif_count++;
1099 	atomic_add_32(&sctp_ill->sctp_ill_ipifcnt, 1);
1100 	if (sctp_ipif->sctp_ipif_state == SCTP_IPIFS_UP)
1101 		sctp_chk_and_updt_saddr(hindex, sctp_ipif, sctps);
1102 	rw_exit(&sctps->sctps_g_ipifs_lock);
1103 	rw_exit(&sctps->sctps_g_ills_lock);
1104 }
1105 
1106 /* Insert, Remove,  Mark up or Mark down the ipif */
1107 void
1108 sctp_update_ipif(ipif_t *ipif, int op)
1109 {
1110 	ill_t		*ill = ipif->ipif_ill;
1111 	int		i;
1112 	sctp_ill_t	*sctp_ill;
1113 	sctp_ipif_t	*sctp_ipif;
1114 	uint_t		ill_index;
1115 	uint_t		hindex;
1116 	netstack_t	*ns = ipif->ipif_ill->ill_ipst->ips_netstack;
1117 	sctp_stack_t	*sctps = ns->netstack_sctp;
1118 
1119 	ip2dbg(("sctp_update_ipif: %s %d\n", ill->ill_name, ipif->ipif_seqid));
1120 
1121 	rw_enter(&sctps->sctps_g_ills_lock, RW_READER);
1122 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_WRITER);
1123 
1124 	ill_index = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(ill));
1125 	sctp_ill = list_head(&sctps->sctps_g_ills[ill_index].sctp_ill_list);
1126 	for (i = 0; i < sctps->sctps_g_ills[ill_index].ill_count; i++) {
1127 		if (sctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(ill) &&
1128 		    sctp_ill->sctp_ill_isv6 == ill->ill_isv6) {
1129 			break;
1130 		}
1131 		sctp_ill = list_next(
1132 		    &sctps->sctps_g_ills[ill_index].sctp_ill_list, sctp_ill);
1133 	}
1134 	if (sctp_ill == NULL) {
1135 		rw_exit(&sctps->sctps_g_ipifs_lock);
1136 		rw_exit(&sctps->sctps_g_ills_lock);
1137 		return;
1138 	}
1139 
1140 	hindex = SCTP_IPIF_ADDR_HASH(ipif->ipif_v6lcl_addr,
1141 	    ipif->ipif_ill->ill_isv6);
1142 	sctp_ipif = list_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list);
1143 	for (i = 0; i < sctps->sctps_g_ipifs[hindex].ipif_count; i++) {
1144 		if (sctp_ipif->sctp_ipif_id == ipif->ipif_seqid) {
1145 			ASSERT(IN6_ARE_ADDR_EQUAL(&sctp_ipif->sctp_ipif_saddr,
1146 			    &ipif->ipif_v6lcl_addr));
1147 			break;
1148 		}
1149 		sctp_ipif = list_next(
1150 		    &sctps->sctps_g_ipifs[hindex].sctp_ipif_list,
1151 		    sctp_ipif);
1152 	}
1153 	if (sctp_ipif == NULL) {
1154 		ip1dbg(("sctp_update_ipif: null sctp_ipif for %d\n", op));
1155 		rw_exit(&sctps->sctps_g_ipifs_lock);
1156 		rw_exit(&sctps->sctps_g_ills_lock);
1157 		return;
1158 	}
1159 	ASSERT(sctp_ill == sctp_ipif->sctp_ipif_ill);
1160 	switch (op) {
1161 	case SCTP_IPIF_REMOVE:
1162 	{
1163 		list_t		*ipif_list;
1164 		list_t		*ill_list;
1165 
1166 		ill_list = &sctps->sctps_g_ills[ill_index].sctp_ill_list;
1167 		ipif_list = &sctps->sctps_g_ipifs[hindex].sctp_ipif_list;
1168 		if (sctp_ipif->sctp_ipif_refcnt != 0) {
1169 			sctp_ipif->sctp_ipif_state = SCTP_IPIFS_CONDEMNED;
1170 			rw_exit(&sctps->sctps_g_ipifs_lock);
1171 			rw_exit(&sctps->sctps_g_ills_lock);
1172 			return;
1173 		}
1174 		list_remove(ipif_list, (void *)sctp_ipif);
1175 		sctps->sctps_g_ipifs[hindex].ipif_count--;
1176 		sctps->sctps_g_ipifs_count--;
1177 		rw_destroy(&sctp_ipif->sctp_ipif_lock);
1178 		kmem_free(sctp_ipif, sizeof (sctp_ipif_t));
1179 		(void) atomic_add_32_nv(&sctp_ill->sctp_ill_ipifcnt, -1);
1180 		if (rw_tryupgrade(&sctps->sctps_g_ills_lock) != 0) {
1181 			rw_downgrade(&sctps->sctps_g_ipifs_lock);
1182 			if (sctp_ill->sctp_ill_ipifcnt == 0 &&
1183 			    sctp_ill->sctp_ill_state == SCTP_ILLS_CONDEMNED) {
1184 				list_remove(ill_list, (void *)sctp_ill);
1185 				sctps->sctps_ills_count--;
1186 				sctps->sctps_g_ills[ill_index].ill_count--;
1187 				kmem_free(sctp_ill->sctp_ill_name,
1188 				    sctp_ill->sctp_ill_name_length);
1189 				kmem_free(sctp_ill, sizeof (sctp_ill_t));
1190 			}
1191 		}
1192 		break;
1193 	}
1194 
1195 	case SCTP_IPIF_UP:
1196 
1197 		rw_downgrade(&sctps->sctps_g_ipifs_lock);
1198 		rw_enter(&sctp_ipif->sctp_ipif_lock, RW_WRITER);
1199 		sctp_ipif->sctp_ipif_state = SCTP_IPIFS_UP;
1200 		sctp_ipif->sctp_ipif_mtu = ipif->ipif_mtu;
1201 		sctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1202 		rw_exit(&sctp_ipif->sctp_ipif_lock);
1203 		sctp_chk_and_updt_saddr(hindex, sctp_ipif,
1204 		    ipif->ipif_ill->ill_ipst->ips_netstack->netstack_sctp);
1205 
1206 		break;
1207 
1208 	case SCTP_IPIF_UPDATE:
1209 
1210 		rw_downgrade(&sctps->sctps_g_ipifs_lock);
1211 		rw_enter(&sctp_ipif->sctp_ipif_lock, RW_WRITER);
1212 		sctp_ipif->sctp_ipif_mtu = ipif->ipif_mtu;
1213 		sctp_ipif->sctp_ipif_zoneid = ipif->ipif_zoneid;
1214 		sctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1215 		rw_exit(&sctp_ipif->sctp_ipif_lock);
1216 
1217 		break;
1218 
1219 	case SCTP_IPIF_DOWN:
1220 
1221 		rw_downgrade(&sctps->sctps_g_ipifs_lock);
1222 		rw_enter(&sctp_ipif->sctp_ipif_lock, RW_WRITER);
1223 		sctp_ipif->sctp_ipif_state = SCTP_IPIFS_DOWN;
1224 		sctp_ipif->sctp_ipif_mtu = ipif->ipif_mtu;
1225 		sctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1226 		rw_exit(&sctp_ipif->sctp_ipif_lock);
1227 
1228 		break;
1229 	}
1230 	rw_exit(&sctps->sctps_g_ipifs_lock);
1231 	rw_exit(&sctps->sctps_g_ills_lock);
1232 }
1233 
1234 /*
1235  * SCTP source address list manipulaton, locking not used (except for
1236  * sctp locking by the caller.
1237  */
1238 
1239 /* Remove a specific saddr from the list */
1240 void
1241 sctp_del_saddr(sctp_t *sctp, sctp_saddr_ipif_t *sp)
1242 {
1243 	if (sctp->sctp_conn_tfp != NULL)
1244 		mutex_enter(&sctp->sctp_conn_tfp->tf_lock);
1245 
1246 	if (sctp->sctp_listen_tfp != NULL)
1247 		mutex_enter(&sctp->sctp_listen_tfp->tf_lock);
1248 
1249 	sctp_ipif_hash_remove(sctp, sp->saddr_ipifp);
1250 
1251 	if (sctp->sctp_bound_to_all == 1)
1252 		sctp->sctp_bound_to_all = 0;
1253 
1254 	if (sctp->sctp_conn_tfp != NULL)
1255 		mutex_exit(&sctp->sctp_conn_tfp->tf_lock);
1256 
1257 	if (sctp->sctp_listen_tfp != NULL)
1258 		mutex_exit(&sctp->sctp_listen_tfp->tf_lock);
1259 }
1260 
1261 /*
1262  * Delete source address from the existing list. No error checking done here
1263  * Called with no locks held.
1264  */
1265 void
1266 sctp_del_saddr_list(sctp_t *sctp, const void *addrs, int addcnt,
1267     boolean_t fanout_locked)
1268 {
1269 	struct sockaddr_in	*sin4;
1270 	struct sockaddr_in6	*sin6;
1271 	int			cnt;
1272 	in6_addr_t		addr;
1273 	sctp_ipif_t		*sctp_ipif;
1274 	int			ifindex = 0;
1275 
1276 	ASSERT(sctp->sctp_nsaddrs >= addcnt);
1277 
1278 	if (!fanout_locked) {
1279 		if (sctp->sctp_conn_tfp != NULL)
1280 			mutex_enter(&sctp->sctp_conn_tfp->tf_lock);
1281 		if (sctp->sctp_listen_tfp != NULL)
1282 			mutex_enter(&sctp->sctp_listen_tfp->tf_lock);
1283 	}
1284 
1285 	for (cnt = 0; cnt < addcnt; cnt++) {
1286 		switch (sctp->sctp_family) {
1287 		case AF_INET:
1288 			sin4 = (struct sockaddr_in *)addrs + cnt;
1289 			IN6_INADDR_TO_V4MAPPED(&sin4->sin_addr, &addr);
1290 			break;
1291 
1292 		case AF_INET6:
1293 			sin6 = (struct sockaddr_in6 *)addrs + cnt;
1294 			addr = sin6->sin6_addr;
1295 			ifindex = sin6->sin6_scope_id;
1296 			break;
1297 		}
1298 		sctp_ipif = sctp_lookup_ipif_addr(&addr, B_FALSE,
1299 		    sctp->sctp_zoneid, !sctp->sctp_connp->conn_allzones,
1300 		    ifindex, 0, B_TRUE, sctp->sctp_sctps);
1301 		ASSERT(sctp_ipif != NULL);
1302 		sctp_ipif_hash_remove(sctp, sctp_ipif);
1303 	}
1304 	if (sctp->sctp_bound_to_all == 1)
1305 		sctp->sctp_bound_to_all = 0;
1306 
1307 	if (!fanout_locked) {
1308 		if (sctp->sctp_conn_tfp != NULL)
1309 			mutex_exit(&sctp->sctp_conn_tfp->tf_lock);
1310 		if (sctp->sctp_listen_tfp != NULL)
1311 			mutex_exit(&sctp->sctp_listen_tfp->tf_lock);
1312 	}
1313 }
1314 
1315 /*
1316  * Given an address get the corresponding entry from the list
1317  * Called with no locks held.
1318  */
1319 sctp_saddr_ipif_t *
1320 sctp_saddr_lookup(sctp_t *sctp, in6_addr_t *addr, uint_t ifindex)
1321 {
1322 	int			cnt;
1323 	sctp_saddr_ipif_t	*ipif_obj;
1324 	int			hindex;
1325 	sctp_ipif_t		*sctp_ipif;
1326 
1327 	hindex = SCTP_IPIF_ADDR_HASH(*addr, !IN6_IS_ADDR_V4MAPPED(addr));
1328 	if (sctp->sctp_saddrs[hindex].ipif_count == 0)
1329 		return (NULL);
1330 
1331 	ipif_obj = list_head(&sctp->sctp_saddrs[hindex].sctp_ipif_list);
1332 	for (cnt = 0; cnt < sctp->sctp_saddrs[hindex].ipif_count; cnt++) {
1333 		sctp_ipif = ipif_obj->saddr_ipifp;
1334 		/*
1335 		 * Zone check shouldn't be needed.
1336 		 */
1337 		if (IN6_ARE_ADDR_EQUAL(addr, &sctp_ipif->sctp_ipif_saddr) &&
1338 		    (ifindex == 0 ||
1339 		    ifindex == sctp_ipif->sctp_ipif_ill->sctp_ill_index) &&
1340 		    SCTP_IPIF_USABLE(sctp_ipif->sctp_ipif_state)) {
1341 			return (ipif_obj);
1342 		}
1343 		ipif_obj = list_next(&sctp->sctp_saddrs[hindex].sctp_ipif_list,
1344 		    ipif_obj);
1345 	}
1346 	return (NULL);
1347 }
1348 
1349 /* Given an address, add it to the source address list */
1350 int
1351 sctp_saddr_add_addr(sctp_t *sctp, in6_addr_t *addr, uint_t ifindex)
1352 {
1353 	sctp_ipif_t		*sctp_ipif;
1354 
1355 	sctp_ipif = sctp_lookup_ipif_addr(addr, B_TRUE, sctp->sctp_zoneid,
1356 	    !sctp->sctp_connp->conn_allzones, ifindex, 0, B_TRUE,
1357 	    sctp->sctp_sctps);
1358 	if (sctp_ipif == NULL)
1359 		return (EINVAL);
1360 
1361 	if (sctp_ipif_hash_insert(sctp, sctp_ipif, KM_NOSLEEP, B_FALSE,
1362 	    B_FALSE) != 0) {
1363 		SCTP_IPIF_REFRELE(sctp_ipif);
1364 		return (EINVAL);
1365 	}
1366 	return (0);
1367 }
1368 
1369 /*
1370  * Remove or mark as dontsrc addresses that are currently not part of the
1371  * association. One would delete addresses when processing an INIT and
1372  * mark as dontsrc when processing an INIT-ACK.
1373  */
1374 void
1375 sctp_check_saddr(sctp_t *sctp, int supp_af, boolean_t delete,
1376     in6_addr_t *no_del_addr)
1377 {
1378 	int			i;
1379 	int			l;
1380 	sctp_saddr_ipif_t	*obj;
1381 	int			scanned = 0;
1382 	int			naddr;
1383 	int			nsaddr;
1384 
1385 	ASSERT(!sctp->sctp_loopback && !sctp->sctp_linklocal && supp_af != 0);
1386 
1387 	/*
1388 	 * Irregardless of the supported address in the INIT, v4
1389 	 * must be supported.
1390 	 */
1391 	if (sctp->sctp_family == AF_INET)
1392 		supp_af = PARM_SUPP_V4;
1393 
1394 	nsaddr = sctp->sctp_nsaddrs;
1395 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
1396 		if (sctp->sctp_saddrs[i].ipif_count == 0)
1397 			continue;
1398 		obj = list_head(&sctp->sctp_saddrs[i].sctp_ipif_list);
1399 		naddr = sctp->sctp_saddrs[i].ipif_count;
1400 		for (l = 0; l < naddr; l++) {
1401 			sctp_ipif_t	*ipif;
1402 
1403 			ipif = obj->saddr_ipifp;
1404 			scanned++;
1405 
1406 			if (IN6_ARE_ADDR_EQUAL(&ipif->sctp_ipif_saddr,
1407 			    no_del_addr)) {
1408 				goto next_obj;
1409 			}
1410 
1411 			/*
1412 			 * Delete/mark dontsrc loopback/linklocal addresses and
1413 			 * unsupported address.
1414 			 * On a clustered node, we trust the clustering module
1415 			 * to do the right thing w.r.t loopback addresses, so
1416 			 * we ignore loopback addresses in this check.
1417 			 */
1418 			if ((SCTP_IS_IPIF_LOOPBACK(ipif) &&
1419 			    cl_sctp_check_addrs == NULL) ||
1420 			    SCTP_IS_IPIF_LINKLOCAL(ipif) ||
1421 			    SCTP_UNSUPP_AF(ipif, supp_af)) {
1422 				if (!delete) {
1423 					obj->saddr_ipif_unconfirmed = 1;
1424 					goto next_obj;
1425 				}
1426 				if (sctp->sctp_bound_to_all == 1)
1427 					sctp->sctp_bound_to_all = 0;
1428 				if (scanned < nsaddr) {
1429 					obj = list_next(&sctp->sctp_saddrs[i].
1430 					    sctp_ipif_list, obj);
1431 					sctp_ipif_hash_remove(sctp, ipif);
1432 					continue;
1433 				}
1434 				sctp_ipif_hash_remove(sctp, ipif);
1435 			}
1436 	next_obj:
1437 			if (scanned >= nsaddr)
1438 				return;
1439 			obj = list_next(&sctp->sctp_saddrs[i].sctp_ipif_list,
1440 			    obj);
1441 		}
1442 	}
1443 }
1444 
1445 
1446 /* Get the first valid address from the list. Called with no locks held */
1447 in6_addr_t
1448 sctp_get_valid_addr(sctp_t *sctp, boolean_t isv6, boolean_t *addr_set)
1449 {
1450 	int			i;
1451 	int			l;
1452 	sctp_saddr_ipif_t	*obj;
1453 	int			scanned = 0;
1454 	in6_addr_t		addr;
1455 
1456 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
1457 		if (sctp->sctp_saddrs[i].ipif_count == 0)
1458 			continue;
1459 		obj = list_head(&sctp->sctp_saddrs[i].sctp_ipif_list);
1460 		for (l = 0; l < sctp->sctp_saddrs[i].ipif_count; l++) {
1461 			sctp_ipif_t	*ipif;
1462 
1463 			ipif = obj->saddr_ipifp;
1464 			if (!SCTP_DONT_SRC(obj) &&
1465 			    ipif->sctp_ipif_isv6 == isv6 &&
1466 			    ipif->sctp_ipif_state == SCTP_IPIFS_UP) {
1467 				*addr_set = B_TRUE;
1468 				return (ipif->sctp_ipif_saddr);
1469 			}
1470 			scanned++;
1471 			if (scanned >= sctp->sctp_nsaddrs)
1472 				goto got_none;
1473 			obj = list_next(&sctp->sctp_saddrs[i].sctp_ipif_list,
1474 			    obj);
1475 		}
1476 	}
1477 got_none:
1478 	/* Need to double check this */
1479 	if (isv6 == B_TRUE)
1480 		addr =  ipv6_all_zeros;
1481 	else
1482 		IN6_IPADDR_TO_V4MAPPED(0, &addr);
1483 	*addr_set = B_FALSE;
1484 	return (addr);
1485 }
1486 
1487 /*
1488  * Return the list of local addresses of an association.  The parameter
1489  * myaddrs is supposed to be either (struct sockaddr_in *) or (struct
1490  * sockaddr_in6 *) depending on the address family.
1491  */
1492 int
1493 sctp_getmyaddrs(void *conn, void *myaddrs, int *addrcnt)
1494 {
1495 	int			i;
1496 	int			l;
1497 	sctp_saddr_ipif_t	*obj;
1498 	sctp_t			*sctp = (sctp_t *)conn;
1499 	int			family = sctp->sctp_family;
1500 	int			max = *addrcnt;
1501 	size_t			added = 0;
1502 	struct sockaddr_in6	*sin6;
1503 	struct sockaddr_in	*sin4;
1504 	int			scanned = 0;
1505 	boolean_t		skip_lback = B_FALSE;
1506 
1507 	if (sctp->sctp_nsaddrs == 0)
1508 		return (EINVAL);
1509 
1510 	/*
1511 	 * Skip loopback addresses for non-loopback assoc., ignore
1512 	 * this on a clustered node.
1513 	 */
1514 	if (sctp->sctp_state >= SCTPS_ESTABLISHED && !sctp->sctp_loopback &&
1515 	    (cl_sctp_check_addrs == NULL)) {
1516 		skip_lback = B_TRUE;
1517 	}
1518 
1519 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
1520 		if (sctp->sctp_saddrs[i].ipif_count == 0)
1521 			continue;
1522 		obj = list_head(&sctp->sctp_saddrs[i].sctp_ipif_list);
1523 		for (l = 0; l < sctp->sctp_saddrs[i].ipif_count; l++) {
1524 			sctp_ipif_t	*ipif = obj->saddr_ipifp;
1525 			in6_addr_t	addr = ipif->sctp_ipif_saddr;
1526 
1527 			scanned++;
1528 			if ((ipif->sctp_ipif_state == SCTP_IPIFS_CONDEMNED) ||
1529 			    SCTP_DONT_SRC(obj) ||
1530 			    (SCTP_IS_IPIF_LOOPBACK(ipif) && skip_lback)) {
1531 				if (scanned >= sctp->sctp_nsaddrs)
1532 					goto done;
1533 				obj = list_next(&sctp->sctp_saddrs[i].
1534 				    sctp_ipif_list, obj);
1535 				continue;
1536 			}
1537 			switch (family) {
1538 			case AF_INET:
1539 				sin4 = (struct sockaddr_in *)myaddrs + added;
1540 				sin4->sin_family = AF_INET;
1541 				sin4->sin_port = sctp->sctp_lport;
1542 				IN6_V4MAPPED_TO_INADDR(&addr, &sin4->sin_addr);
1543 				break;
1544 
1545 			case AF_INET6:
1546 				sin6 = (struct sockaddr_in6 *)myaddrs + added;
1547 				sin6->sin6_family = AF_INET6;
1548 				sin6->sin6_port = sctp->sctp_lport;
1549 				sin6->sin6_addr = addr;
1550 				break;
1551 			}
1552 			added++;
1553 			if (added >= max || scanned >= sctp->sctp_nsaddrs)
1554 				goto done;
1555 			obj = list_next(&sctp->sctp_saddrs[i].sctp_ipif_list,
1556 			    obj);
1557 		}
1558 	}
1559 done:
1560 	*addrcnt = added;
1561 	return (0);
1562 }
1563 
1564 /*
1565  * Given the supported address family, walk through the source address list
1566  * and return the total length of the available addresses. If 'p' is not
1567  * null, construct the parameter list for the addresses in 'p'.
1568  * 'modify' will only be set when we want the source address list to
1569  * be modified. The source address list will be modified only when
1570  * generating an INIT chunk. For generating an INIT-ACK 'modify' will
1571  * be false since the 'sctp' will be that of the listener.
1572  */
1573 size_t
1574 sctp_saddr_info(sctp_t *sctp, int supp_af, uchar_t *p, boolean_t modify)
1575 {
1576 	int			i;
1577 	int			l;
1578 	sctp_saddr_ipif_t	*obj;
1579 	size_t			paramlen = 0;
1580 	sctp_parm_hdr_t		*hdr;
1581 	int			scanned = 0;
1582 	int			naddr;
1583 	int			nsaddr;
1584 	boolean_t		del_ll = B_FALSE;
1585 	boolean_t		del_lb = B_FALSE;
1586 
1587 
1588 	/*
1589 	 * On a clustered node don't bother changing anything
1590 	 * on the loopback interface.
1591 	 */
1592 	if (modify && !sctp->sctp_loopback && (cl_sctp_check_addrs == NULL))
1593 		del_lb = B_TRUE;
1594 
1595 	if (modify && !sctp->sctp_linklocal)
1596 		del_ll = B_TRUE;
1597 
1598 	nsaddr = sctp->sctp_nsaddrs;
1599 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
1600 		if (sctp->sctp_saddrs[i].ipif_count == 0)
1601 			continue;
1602 		obj = list_head(&sctp->sctp_saddrs[i].sctp_ipif_list);
1603 		naddr = sctp->sctp_saddrs[i].ipif_count;
1604 		for (l = 0; l < naddr; l++) {
1605 			in6_addr_t	addr;
1606 			sctp_ipif_t	*ipif;
1607 			boolean_t	ipif_lb;
1608 			boolean_t	ipif_ll;
1609 			boolean_t	unsupp_af;
1610 
1611 			ipif = obj->saddr_ipifp;
1612 			scanned++;
1613 
1614 			ipif_lb = SCTP_IS_IPIF_LOOPBACK(ipif);
1615 			ipif_ll = SCTP_IS_IPIF_LINKLOCAL(ipif);
1616 			unsupp_af = SCTP_UNSUPP_AF(ipif, supp_af);
1617 			/*
1618 			 * We need to either delete or skip loopback/linklocal
1619 			 * or unsupported addresses, if required.
1620 			 */
1621 			if ((ipif_ll && del_ll) || (ipif_lb && del_lb) ||
1622 			    (unsupp_af && modify)) {
1623 				if (sctp->sctp_bound_to_all == 1)
1624 					sctp->sctp_bound_to_all = 0;
1625 				if (scanned < nsaddr) {
1626 					obj = list_next(&sctp->sctp_saddrs[i].
1627 					    sctp_ipif_list, obj);
1628 					sctp_ipif_hash_remove(sctp, ipif);
1629 					continue;
1630 				}
1631 				sctp_ipif_hash_remove(sctp, ipif);
1632 				goto next_addr;
1633 			} else if (ipif_ll || unsupp_af ||
1634 			    (ipif_lb && (cl_sctp_check_addrs == NULL))) {
1635 				goto next_addr;
1636 			}
1637 
1638 			if (!SCTP_IPIF_USABLE(ipif->sctp_ipif_state))
1639 				goto next_addr;
1640 			if (p != NULL)
1641 				hdr = (sctp_parm_hdr_t *)(p + paramlen);
1642 			addr = ipif->sctp_ipif_saddr;
1643 			if (!ipif->sctp_ipif_isv6) {
1644 				struct in_addr	*v4;
1645 
1646 				if (p != NULL) {
1647 					hdr->sph_type = htons(PARM_ADDR4);
1648 					hdr->sph_len = htons(PARM_ADDR4_LEN);
1649 					v4 = (struct in_addr *)(hdr + 1);
1650 					IN6_V4MAPPED_TO_INADDR(&addr, v4);
1651 				}
1652 				paramlen += PARM_ADDR4_LEN;
1653 			} else {
1654 				if (p != NULL) {
1655 					hdr->sph_type = htons(PARM_ADDR6);
1656 					hdr->sph_len = htons(PARM_ADDR6_LEN);
1657 					bcopy(&addr, hdr + 1, sizeof (addr));
1658 				}
1659 				paramlen += PARM_ADDR6_LEN;
1660 			}
1661 next_addr:
1662 			if (scanned >= nsaddr)
1663 				return (paramlen);
1664 			obj = list_next(&sctp->sctp_saddrs[i].sctp_ipif_list,
1665 			    obj);
1666 		}
1667 	}
1668 	return (paramlen);
1669 }
1670 
1671 /*
1672  * This is used on a clustered node to obtain a list of addresses, the list
1673  * consists of sockaddr_in structs for v4 and sockaddr_in6 for v6. The list
1674  * is then passed onto the clustering module which sends back the correct
1675  * list based on the port info. Regardless of the input, i.e INADDR_ANY
1676  * or specific address(es), we create the list since it could be modified by
1677  * the clustering module. When given a list of addresses, we simply
1678  * create the list of sockaddr_in or sockaddr_in6 structs using those
1679  * addresses. If there is an INADDR_ANY in the input list, or if the
1680  * input is INADDR_ANY, we create a list of sockaddr_in or sockaddr_in6
1681  * structs consisting all the addresses in the global interface list
1682  * except those that are hosted on the loopback interface. We create
1683  * a list of sockaddr_in[6] structs just so that it can be directly input
1684  * to sctp_valid_addr_list() once the clustering module has processed it.
1685  */
1686 int
1687 sctp_get_addrlist(sctp_t *sctp, const void *addrs, uint32_t *addrcnt,
1688     uchar_t **addrlist, int *uspec, size_t *size)
1689 {
1690 	int			cnt;
1691 	int			icnt;
1692 	sctp_ipif_t		*sctp_ipif;
1693 	struct sockaddr_in	*s4;
1694 	struct sockaddr_in6	*s6;
1695 	uchar_t			*p;
1696 	int			err = 0;
1697 	sctp_stack_t		*sctps = sctp->sctp_sctps;
1698 
1699 	*addrlist = NULL;
1700 	*size = 0;
1701 
1702 	/*
1703 	 * Create a list of sockaddr_in[6] structs using the input list.
1704 	 */
1705 	if (sctp->sctp_family == AF_INET) {
1706 		*size = sizeof (struct sockaddr_in) * *addrcnt;
1707 		*addrlist = kmem_zalloc(*size,  KM_SLEEP);
1708 		p = *addrlist;
1709 		for (cnt = 0; cnt < *addrcnt; cnt++) {
1710 			s4 = (struct sockaddr_in *)addrs + cnt;
1711 			/*
1712 			 * We need to create a list of all the available
1713 			 * addresses if there is an INADDR_ANY. However,
1714 			 * if we are beyond LISTEN, then this is invalid
1715 			 * (see sctp_valid_addr_list(). So, we just fail
1716 			 * it here rather than wait till it fails in
1717 			 * sctp_valid_addr_list().
1718 			 */
1719 			if (s4->sin_addr.s_addr == INADDR_ANY) {
1720 				kmem_free(*addrlist, *size);
1721 				*addrlist = NULL;
1722 				*size = 0;
1723 				if (sctp->sctp_state > SCTPS_LISTEN) {
1724 					*addrcnt = 0;
1725 					return (EINVAL);
1726 				}
1727 				if (uspec != NULL)
1728 					*uspec = 1;
1729 				goto get_all_addrs;
1730 			} else {
1731 				bcopy(s4, p, sizeof (*s4));
1732 				p += sizeof (*s4);
1733 			}
1734 		}
1735 	} else {
1736 		*size = sizeof (struct sockaddr_in6) * *addrcnt;
1737 		*addrlist = kmem_zalloc(*size, KM_SLEEP);
1738 		p = *addrlist;
1739 		for (cnt = 0; cnt < *addrcnt; cnt++) {
1740 			s6 = (struct sockaddr_in6 *)addrs + cnt;
1741 			/*
1742 			 * Comments for INADDR_ANY, above, apply here too.
1743 			 */
1744 			if (IN6_IS_ADDR_UNSPECIFIED(&s6->sin6_addr)) {
1745 				kmem_free(*addrlist, *size);
1746 				*size = 0;
1747 				*addrlist = NULL;
1748 				if (sctp->sctp_state > SCTPS_LISTEN) {
1749 					*addrcnt = 0;
1750 					return (EINVAL);
1751 				}
1752 				if (uspec != NULL)
1753 					*uspec = 1;
1754 				goto get_all_addrs;
1755 			} else {
1756 				bcopy(addrs, p, sizeof (*s6));
1757 				p += sizeof (*s6);
1758 			}
1759 		}
1760 	}
1761 	return (err);
1762 get_all_addrs:
1763 
1764 	/*
1765 	 * Allocate max possible size. We allocate the max. size here because
1766 	 * the clustering module could end up adding addresses to the list.
1767 	 * We allocate upfront so that the clustering module need to bother
1768 	 * re-sizing the list.
1769 	 */
1770 	if (sctp->sctp_family == AF_INET) {
1771 		*size = sizeof (struct sockaddr_in) *
1772 		    sctps->sctps_g_ipifs_count;
1773 	} else {
1774 		*size = sizeof (struct sockaddr_in6) *
1775 		    sctps->sctps_g_ipifs_count;
1776 	}
1777 	*addrlist = kmem_zalloc(*size, KM_SLEEP);
1778 	*addrcnt = 0;
1779 	p = *addrlist;
1780 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_READER);
1781 
1782 	/*
1783 	 * Walk through the global interface list and add all addresses,
1784 	 * except those that are hosted on loopback interfaces.
1785 	 */
1786 	for (cnt = 0; cnt <  SCTP_IPIF_HASH; cnt++) {
1787 		if (sctps->sctps_g_ipifs[cnt].ipif_count == 0)
1788 			continue;
1789 		sctp_ipif = list_head(
1790 		    &sctps->sctps_g_ipifs[cnt].sctp_ipif_list);
1791 		for (icnt = 0;
1792 		    icnt < sctps->sctps_g_ipifs[cnt].ipif_count;
1793 		    icnt++) {
1794 			in6_addr_t	addr;
1795 
1796 			rw_enter(&sctp_ipif->sctp_ipif_lock, RW_READER);
1797 			addr = sctp_ipif->sctp_ipif_saddr;
1798 			if (SCTP_IPIF_DISCARD(sctp_ipif->sctp_ipif_flags) ||
1799 			    !SCTP_IPIF_USABLE(sctp_ipif->sctp_ipif_state) ||
1800 			    SCTP_IS_IPIF_LOOPBACK(sctp_ipif) ||
1801 			    SCTP_IS_IPIF_LINKLOCAL(sctp_ipif) ||
1802 			    !SCTP_IPIF_ZONE_MATCH(sctp, sctp_ipif) ||
1803 			    (sctp->sctp_ipversion == IPV4_VERSION &&
1804 			    sctp_ipif->sctp_ipif_isv6) ||
1805 			    (sctp->sctp_connp->conn_ipv6_v6only &&
1806 			    !sctp_ipif->sctp_ipif_isv6)) {
1807 				rw_exit(&sctp_ipif->sctp_ipif_lock);
1808 				sctp_ipif = list_next(
1809 				    &sctps->sctps_g_ipifs[cnt].sctp_ipif_list,
1810 				    sctp_ipif);
1811 				continue;
1812 			}
1813 			rw_exit(&sctp_ipif->sctp_ipif_lock);
1814 			if (sctp->sctp_family == AF_INET) {
1815 				s4 = (struct sockaddr_in *)p;
1816 				IN6_V4MAPPED_TO_INADDR(&addr, &s4->sin_addr);
1817 				s4->sin_family = AF_INET;
1818 				p += sizeof (*s4);
1819 			} else {
1820 				s6 = (struct sockaddr_in6 *)p;
1821 				s6->sin6_addr = addr;
1822 				s6->sin6_family = AF_INET6;
1823 				s6->sin6_scope_id =
1824 				    sctp_ipif->sctp_ipif_ill->sctp_ill_index;
1825 				p += sizeof (*s6);
1826 			}
1827 			(*addrcnt)++;
1828 			sctp_ipif = list_next(
1829 			    &sctps->sctps_g_ipifs[cnt].sctp_ipif_list,
1830 			    sctp_ipif);
1831 		}
1832 	}
1833 	rw_exit(&sctps->sctps_g_ipifs_lock);
1834 	return (err);
1835 }
1836 
1837 /*
1838  * Get a list of addresses from the source address list. The  caller is
1839  * responsible for allocating sufficient buffer for this.
1840  */
1841 void
1842 sctp_get_saddr_list(sctp_t *sctp, uchar_t *p, size_t psize)
1843 {
1844 	int			cnt;
1845 	int			icnt;
1846 	sctp_saddr_ipif_t	*obj;
1847 	int			naddr;
1848 	int			scanned = 0;
1849 
1850 	for (cnt = 0; cnt < SCTP_IPIF_HASH; cnt++) {
1851 		if (sctp->sctp_saddrs[cnt].ipif_count == 0)
1852 			continue;
1853 		obj = list_head(&sctp->sctp_saddrs[cnt].sctp_ipif_list);
1854 		naddr = sctp->sctp_saddrs[cnt].ipif_count;
1855 		for (icnt = 0; icnt < naddr; icnt++) {
1856 			sctp_ipif_t	*ipif;
1857 
1858 			if (psize < sizeof (ipif->sctp_ipif_saddr))
1859 				return;
1860 
1861 			scanned++;
1862 			ipif = obj->saddr_ipifp;
1863 			bcopy(&ipif->sctp_ipif_saddr, p,
1864 			    sizeof (ipif->sctp_ipif_saddr));
1865 			p += sizeof (ipif->sctp_ipif_saddr);
1866 			psize -= sizeof (ipif->sctp_ipif_saddr);
1867 			if (scanned >= sctp->sctp_nsaddrs)
1868 				return;
1869 			obj = list_next(
1870 			    &sctp->sctp_saddrs[icnt].sctp_ipif_list,
1871 			    obj);
1872 		}
1873 	}
1874 }
1875 
1876 /*
1877  * Get a list of addresses from the remote address list. The  caller is
1878  * responsible for allocating sufficient buffer for this.
1879  */
1880 void
1881 sctp_get_faddr_list(sctp_t *sctp, uchar_t *p, size_t psize)
1882 {
1883 	sctp_faddr_t	*fp;
1884 
1885 	for (fp = sctp->sctp_faddrs; fp != NULL; fp = fp->next) {
1886 		if (psize < sizeof (fp->faddr))
1887 			return;
1888 		bcopy(&fp->faddr, p, sizeof (fp->faddr));
1889 		p += sizeof (fp->faddr);
1890 		psize -= sizeof (fp->faddr);
1891 	}
1892 }
1893 
1894 static void
1895 sctp_free_ills(sctp_stack_t *sctps)
1896 {
1897 	int			i;
1898 	int			l;
1899 	sctp_ill_t	*sctp_ill;
1900 
1901 	if (sctps->sctps_ills_count == 0)
1902 		return;
1903 
1904 	for (i = 0; i < SCTP_ILL_HASH; i++) {
1905 		sctp_ill = list_tail(&sctps->sctps_g_ills[i].sctp_ill_list);
1906 		for (l = 0; l < sctps->sctps_g_ills[i].ill_count; l++) {
1907 			ASSERT(sctp_ill->sctp_ill_ipifcnt == 0);
1908 			list_remove(&sctps->sctps_g_ills[i].sctp_ill_list,
1909 			    sctp_ill);
1910 			sctps->sctps_ills_count--;
1911 			kmem_free(sctp_ill->sctp_ill_name,
1912 			    sctp_ill->sctp_ill_name_length);
1913 			kmem_free(sctp_ill, sizeof (sctp_ill_t));
1914 			sctp_ill =
1915 			    list_tail(&sctps->sctps_g_ills[i].sctp_ill_list);
1916 		}
1917 		sctps->sctps_g_ills[i].ill_count = 0;
1918 	}
1919 	ASSERT(sctps->sctps_ills_count == 0);
1920 }
1921 
1922 static void
1923 sctp_free_ipifs(sctp_stack_t *sctps)
1924 {
1925 	int			i;
1926 	int			l;
1927 	sctp_ipif_t	*sctp_ipif;
1928 	sctp_ill_t	*sctp_ill;
1929 
1930 	if (sctps->sctps_g_ipifs_count == 0)
1931 		return;
1932 
1933 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
1934 		sctp_ipif = list_tail(&sctps->sctps_g_ipifs[i].sctp_ipif_list);
1935 		for (l = 0; l < sctps->sctps_g_ipifs[i].ipif_count; l++) {
1936 			sctp_ill = sctp_ipif->sctp_ipif_ill;
1937 
1938 			list_remove(&sctps->sctps_g_ipifs[i].sctp_ipif_list,
1939 			    sctp_ipif);
1940 			sctps->sctps_g_ipifs_count--;
1941 			(void) atomic_add_32_nv(&sctp_ill->sctp_ill_ipifcnt,
1942 			    -1);
1943 			kmem_free(sctp_ipif, sizeof (sctp_ipif_t));
1944 			sctp_ipif =
1945 			    list_tail(&sctps->sctps_g_ipifs[i].sctp_ipif_list);
1946 		}
1947 		sctps->sctps_g_ipifs[i].ipif_count = 0;
1948 	}
1949 	ASSERT(sctps->sctps_g_ipifs_count == 0);
1950 }
1951 
1952 
1953 /* Initialize the SCTP ILL list and lock */
1954 void
1955 sctp_saddr_init(sctp_stack_t *sctps)
1956 {
1957 	int	i;
1958 
1959 	sctps->sctps_g_ills = kmem_zalloc(sizeof (sctp_ill_hash_t) *
1960 	    SCTP_ILL_HASH, KM_SLEEP);
1961 	sctps->sctps_g_ipifs = kmem_zalloc(sizeof (sctp_ipif_hash_t) *
1962 	    SCTP_IPIF_HASH, KM_SLEEP);
1963 
1964 	rw_init(&sctps->sctps_g_ills_lock, NULL, RW_DEFAULT, NULL);
1965 	rw_init(&sctps->sctps_g_ipifs_lock, NULL, RW_DEFAULT, NULL);
1966 
1967 	for (i = 0; i < SCTP_ILL_HASH; i++) {
1968 		sctps->sctps_g_ills[i].ill_count = 0;
1969 		list_create(&sctps->sctps_g_ills[i].sctp_ill_list,
1970 		    sizeof (sctp_ill_t),
1971 		    offsetof(sctp_ill_t, sctp_ills));
1972 	}
1973 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
1974 		sctps->sctps_g_ipifs[i].ipif_count = 0;
1975 		list_create(&sctps->sctps_g_ipifs[i].sctp_ipif_list,
1976 		    sizeof (sctp_ipif_t), offsetof(sctp_ipif_t, sctp_ipifs));
1977 	}
1978 }
1979 
1980 void
1981 sctp_saddr_fini(sctp_stack_t *sctps)
1982 {
1983 	int	i;
1984 
1985 	sctp_free_ipifs(sctps);
1986 	sctp_free_ills(sctps);
1987 
1988 	for (i = 0; i < SCTP_ILL_HASH; i++)
1989 		list_destroy(&sctps->sctps_g_ills[i].sctp_ill_list);
1990 	for (i = 0; i < SCTP_IPIF_HASH; i++)
1991 		list_destroy(&sctps->sctps_g_ipifs[i].sctp_ipif_list);
1992 
1993 	ASSERT(sctps->sctps_ills_count == 0 && sctps->sctps_g_ipifs_count == 0);
1994 	kmem_free(sctps->sctps_g_ills, sizeof (sctp_ill_hash_t) *
1995 	    SCTP_ILL_HASH);
1996 	sctps->sctps_g_ills = NULL;
1997 	kmem_free(sctps->sctps_g_ipifs, sizeof (sctp_ipif_hash_t) *
1998 	    SCTP_IPIF_HASH);
1999 	sctps->sctps_g_ipifs = NULL;
2000 	rw_destroy(&sctps->sctps_g_ills_lock);
2001 	rw_destroy(&sctps->sctps_g_ipifs_lock);
2002 }
2003