xref: /titanic_44/usr/src/uts/common/inet/sctp/sctp_addr.c (revision 07d06da50d310a325b457d6330165aebab1e0064)
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_ire.h>
45 #include <inet/ip_if.h>
46 #include <inet/ipclassifier.h>
47 #include <inet/sctp_ip.h>
48 #include "sctp_impl.h"
49 #include "sctp_addr.h"
50 
51 static void		sctp_ipif_inactive(sctp_ipif_t *);
52 static sctp_ipif_t	*sctp_lookup_ipif_addr(in6_addr_t *, boolean_t,
53 			    zoneid_t, boolean_t, uint_t, uint_t, boolean_t,
54 			    sctp_stack_t *);
55 static int		sctp_get_all_ipifs(sctp_t *, int);
56 static int		sctp_ipif_hash_insert(sctp_t *, sctp_ipif_t *, int,
57 			    boolean_t, boolean_t);
58 static void		sctp_ipif_hash_remove(sctp_t *, sctp_ipif_t *);
59 static void		sctp_fix_saddr(sctp_t *, in6_addr_t *);
60 static int		sctp_compare_ipif_list(sctp_ipif_hash_t *,
61 			    sctp_ipif_hash_t *);
62 static int		sctp_copy_ipifs(sctp_ipif_hash_t *, sctp_t *, int);
63 
64 #define	SCTP_ADDR4_HASH(addr)	\
65 	(((addr) ^ ((addr) >> 8) ^ ((addr) >> 16) ^ ((addr) >> 24)) &	\
66 	(SCTP_IPIF_HASH - 1))
67 
68 #define	SCTP_ADDR6_HASH(addr)	\
69 	(((addr).s6_addr32[3] ^						\
70 	(((addr).s6_addr32[3] ^ (addr).s6_addr32[2]) >> 12)) &		\
71 	(SCTP_IPIF_HASH - 1))
72 
73 #define	SCTP_IPIF_ADDR_HASH(addr, isv6)					\
74 	((isv6) ? SCTP_ADDR6_HASH((addr)) : 				\
75 	SCTP_ADDR4_HASH((addr)._S6_un._S6_u32[3]))
76 
77 #define	SCTP_IPIF_USABLE(sctp_ipif_state)	\
78 	((sctp_ipif_state) == SCTP_IPIFS_UP ||	\
79 	(sctp_ipif_state) ==  SCTP_IPIFS_DOWN)
80 
81 #define	SCTP_IPIF_DISCARD(sctp_ipif_flags)	\
82 	((sctp_ipif_flags) & (IPIF_PRIVATE | IPIF_DEPRECATED))
83 
84 #define	SCTP_IS_IPIF_LOOPBACK(ipif)		\
85 	((ipif)->sctp_ipif_ill->sctp_ill_flags & PHYI_LOOPBACK)
86 
87 #define	SCTP_IS_IPIF_LINKLOCAL(ipif)		\
88 	((ipif)->sctp_ipif_isv6 && 		\
89 	IN6_IS_ADDR_LINKLOCAL(&(ipif)->sctp_ipif_saddr))
90 
91 #define	SCTP_UNSUPP_AF(ipif, supp_af)	\
92 	((!(ipif)->sctp_ipif_isv6 && !((supp_af) & PARM_SUPP_V4)) ||	\
93 	((ipif)->sctp_ipif_isv6 && !((supp_af) & PARM_SUPP_V6)))
94 
95 #define	SCTP_IPIF_ZONE_MATCH(sctp, ipif) 				\
96 	IPCL_ZONE_MATCH((sctp)->sctp_connp, (ipif)->sctp_ipif_zoneid)
97 
98 #define	SCTP_ILL_HASH_FN(index)		((index) % SCTP_ILL_HASH)
99 #define	SCTP_ILL_TO_PHYINDEX(ill)	((ill)->ill_phyint->phyint_ifindex)
100 
101 /*
102  * SCTP Interface list manipulation functions, locking used.
103  */
104 
105 /*
106  * Delete an SCTP IPIF from the list if the refcount goes to 0 and it is
107  * marked as condemned. Also, check if the ILL needs to go away.
108  */
109 static void
110 sctp_ipif_inactive(sctp_ipif_t *sctp_ipif)
111 {
112 	sctp_ill_t	*sctp_ill;
113 	uint_t		hindex;
114 	uint_t		ill_index;
115 	sctp_stack_t	*sctps = sctp_ipif->sctp_ipif_ill->
116 	    sctp_ill_netstack->netstack_sctp;
117 
118 	rw_enter(&sctps->sctps_g_ills_lock, RW_READER);
119 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_WRITER);
120 
121 	hindex = SCTP_IPIF_ADDR_HASH(sctp_ipif->sctp_ipif_saddr,
122 	    sctp_ipif->sctp_ipif_isv6);
123 
124 	sctp_ill = sctp_ipif->sctp_ipif_ill;
125 	ASSERT(sctp_ill != NULL);
126 	ill_index = SCTP_ILL_HASH_FN(sctp_ill->sctp_ill_index);
127 	if (sctp_ipif->sctp_ipif_state != SCTP_IPIFS_CONDEMNED ||
128 	    sctp_ipif->sctp_ipif_refcnt != 0) {
129 		rw_exit(&sctps->sctps_g_ipifs_lock);
130 		rw_exit(&sctps->sctps_g_ills_lock);
131 		return;
132 	}
133 	list_remove(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list,
134 	    sctp_ipif);
135 	sctps->sctps_g_ipifs[hindex].ipif_count--;
136 	sctps->sctps_g_ipifs_count--;
137 	rw_destroy(&sctp_ipif->sctp_ipif_lock);
138 	kmem_free(sctp_ipif, sizeof (sctp_ipif_t));
139 
140 	(void) atomic_add_32_nv(&sctp_ill->sctp_ill_ipifcnt, -1);
141 	if (rw_tryupgrade(&sctps->sctps_g_ills_lock) != 0) {
142 		rw_downgrade(&sctps->sctps_g_ipifs_lock);
143 		if (sctp_ill->sctp_ill_ipifcnt == 0 &&
144 		    sctp_ill->sctp_ill_state == SCTP_ILLS_CONDEMNED) {
145 			list_remove(&sctps->sctps_g_ills[ill_index].
146 			    sctp_ill_list, (void *)sctp_ill);
147 			sctps->sctps_g_ills[ill_index].ill_count--;
148 			sctps->sctps_ills_count--;
149 			kmem_free(sctp_ill->sctp_ill_name,
150 			    sctp_ill->sctp_ill_name_length);
151 			kmem_free(sctp_ill, sizeof (sctp_ill_t));
152 		}
153 	}
154 	rw_exit(&sctps->sctps_g_ipifs_lock);
155 	rw_exit(&sctps->sctps_g_ills_lock);
156 }
157 
158 /*
159  * Lookup an SCTP IPIF given an IP address. Increments sctp_ipif refcnt.
160  * We are either looking for a IPIF with the given address before
161  * inserting it into the global list or looking for an IPIF for an
162  * address given an SCTP. In the former case we always check the zoneid,
163  * but for the latter case, check_zid could be B_FALSE if the connp
164  * for the sctp has conn_all_zones set. When looking for an address we
165  * give preference to one that is up, so even though we may find one that
166  * is not up we keep looking if there is one up, we hold the down addr
167  * in backup_ipif in case we don't find one that is up - i.e. we return
168  * the backup_ipif in that case. Note that if we are looking for. If we
169  * are specifically looking for an up address, then usable will be set
170  * to true.
171  */
172 static sctp_ipif_t *
173 sctp_lookup_ipif_addr(in6_addr_t *addr, boolean_t refhold, zoneid_t zoneid,
174     boolean_t check_zid, uint_t ifindex, uint_t seqid, boolean_t usable,
175     sctp_stack_t *sctps)
176 {
177 	int		j;
178 	sctp_ipif_t	*sctp_ipif;
179 	sctp_ipif_t	*backup_ipif = NULL;
180 	int		hindex;
181 
182 	hindex = SCTP_IPIF_ADDR_HASH(*addr, !IN6_IS_ADDR_V4MAPPED(addr));
183 
184 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_READER);
185 	if (sctps->sctps_g_ipifs[hindex].ipif_count == 0) {
186 		rw_exit(&sctps->sctps_g_ipifs_lock);
187 		return (NULL);
188 	}
189 	sctp_ipif = list_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list);
190 	for (j = 0; j < sctps->sctps_g_ipifs[hindex].ipif_count; j++) {
191 		rw_enter(&sctp_ipif->sctp_ipif_lock, RW_READER);
192 		if ((!check_zid ||
193 		    (sctp_ipif->sctp_ipif_zoneid == ALL_ZONES ||
194 		    zoneid == sctp_ipif->sctp_ipif_zoneid)) &&
195 		    (ifindex == 0 || ifindex ==
196 		    sctp_ipif->sctp_ipif_ill->sctp_ill_index) &&
197 		    ((seqid != 0 && seqid == sctp_ipif->sctp_ipif_id) ||
198 		    (IN6_ARE_ADDR_EQUAL(&sctp_ipif->sctp_ipif_saddr,
199 		    addr)))) {
200 			if (!usable || sctp_ipif->sctp_ipif_state ==
201 			    SCTP_IPIFS_UP) {
202 				rw_exit(&sctp_ipif->sctp_ipif_lock);
203 				if (refhold)
204 					SCTP_IPIF_REFHOLD(sctp_ipif);
205 				rw_exit(&sctps->sctps_g_ipifs_lock);
206 				return (sctp_ipif);
207 			} else if (sctp_ipif->sctp_ipif_state ==
208 			    SCTP_IPIFS_DOWN && backup_ipif == NULL) {
209 				backup_ipif = sctp_ipif;
210 			}
211 		}
212 		rw_exit(&sctp_ipif->sctp_ipif_lock);
213 		sctp_ipif = list_next(
214 		    &sctps->sctps_g_ipifs[hindex].sctp_ipif_list, sctp_ipif);
215 	}
216 	if (backup_ipif != NULL) {
217 		if (refhold)
218 			SCTP_IPIF_REFHOLD(backup_ipif);
219 		rw_exit(&sctps->sctps_g_ipifs_lock);
220 		return (backup_ipif);
221 	}
222 	rw_exit(&sctps->sctps_g_ipifs_lock);
223 	return (NULL);
224 }
225 
226 /*
227  * Populate the list with all the SCTP ipifs for a given ipversion.
228  * Increments sctp_ipif refcnt.
229  * Called with no locks held.
230  */
231 static int
232 sctp_get_all_ipifs(sctp_t *sctp, int sleep)
233 {
234 	sctp_ipif_t		*sctp_ipif;
235 	int			i;
236 	int			j;
237 	int			error = 0;
238 	sctp_stack_t		*sctps = sctp->sctp_sctps;
239 	boolean_t		isv6;
240 	conn_t			*connp = sctp->sctp_connp;
241 
242 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_READER);
243 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
244 		if (sctps->sctps_g_ipifs[i].ipif_count == 0)
245 			continue;
246 		sctp_ipif = list_head(&sctps->sctps_g_ipifs[i].sctp_ipif_list);
247 		for (j = 0; j < sctps->sctps_g_ipifs[i].ipif_count; j++) {
248 			rw_enter(&sctp_ipif->sctp_ipif_lock, RW_READER);
249 			isv6 = sctp_ipif->sctp_ipif_isv6;
250 			if (SCTP_IPIF_DISCARD(sctp_ipif->sctp_ipif_flags) ||
251 			    !SCTP_IPIF_USABLE(sctp_ipif->sctp_ipif_state) ||
252 			    !SCTP_IPIF_ZONE_MATCH(sctp, sctp_ipif) ||
253 			    SCTP_IS_ADDR_UNSPEC(!isv6,
254 			    sctp_ipif->sctp_ipif_saddr) ||
255 			    (connp->conn_family == AF_INET && isv6) ||
256 			    (connp->conn_ipv6_v6only && !isv6)) {
257 				rw_exit(&sctp_ipif->sctp_ipif_lock);
258 				sctp_ipif = list_next(
259 				    &sctps->sctps_g_ipifs[i].sctp_ipif_list,
260 				    sctp_ipif);
261 				continue;
262 			}
263 			rw_exit(&sctp_ipif->sctp_ipif_lock);
264 			SCTP_IPIF_REFHOLD(sctp_ipif);
265 			error = sctp_ipif_hash_insert(sctp, sctp_ipif, sleep,
266 			    B_FALSE, B_FALSE);
267 			if (error != 0 && error != EALREADY)
268 				goto free_stuff;
269 			sctp_ipif = list_next(
270 			    &sctps->sctps_g_ipifs[i].sctp_ipif_list,
271 			    sctp_ipif);
272 		}
273 	}
274 	rw_exit(&sctps->sctps_g_ipifs_lock);
275 	return (0);
276 free_stuff:
277 	rw_exit(&sctps->sctps_g_ipifs_lock);
278 	sctp_free_saddrs(sctp);
279 	return (ENOMEM);
280 }
281 
282 /*
283  * Given a list of address, fills in the list of SCTP ipifs if all the addresses
284  * are present in the SCTP interface list, return number of addresses filled
285  * or error. If the caller wants the list of addresses, it sends a pre-allocated
286  * buffer - list. Currently, this list is only used on a clustered node when
287  * the SCTP is in the listen state (from sctp_bind_add()). When called on a
288  * clustered node, the input is always a list of addresses (even if the
289  * original bind() was to INADDR_ANY).
290  * Called with no locks held.
291  */
292 int
293 sctp_valid_addr_list(sctp_t *sctp, const void *addrs, uint32_t addrcnt,
294     uchar_t *list, size_t lsize)
295 {
296 	struct sockaddr_in	*sin4;
297 	struct sockaddr_in6	*sin6;
298 	struct in_addr		*addr4;
299 	in6_addr_t		addr;
300 	int			cnt;
301 	int			err = 0;
302 	int			saddr_cnt = 0;
303 	sctp_ipif_t		*ipif;
304 	boolean_t		bind_to_all = B_FALSE;
305 	boolean_t		check_addrs = B_FALSE;
306 	boolean_t		check_lport = B_FALSE;
307 	uchar_t			*p = list;
308 	conn_t			*connp = sctp->sctp_connp;
309 
310 	/*
311 	 * Need to check for port and address depending on the state.
312 	 * After a socket is bound, we need to make sure that subsequent
313 	 * bindx() has correct port.  After an association is established,
314 	 * we need to check for changing the bound address to invalid
315 	 * addresses.
316 	 */
317 	if (sctp->sctp_state >= SCTPS_BOUND) {
318 		check_lport = B_TRUE;
319 		if (sctp->sctp_state > SCTPS_LISTEN)
320 			check_addrs = B_TRUE;
321 	}
322 
323 	if (sctp->sctp_conn_tfp != NULL)
324 		mutex_enter(&sctp->sctp_conn_tfp->tf_lock);
325 	if (sctp->sctp_listen_tfp != NULL)
326 		mutex_enter(&sctp->sctp_listen_tfp->tf_lock);
327 	for (cnt = 0; cnt < addrcnt; cnt++) {
328 		boolean_t	lookup_saddr = B_TRUE;
329 		uint_t		ifindex = 0;
330 
331 		switch (connp->conn_family) {
332 		case AF_INET:
333 			sin4 = (struct sockaddr_in *)addrs + cnt;
334 			if (sin4->sin_family != AF_INET || (check_lport &&
335 			    sin4->sin_port != connp->conn_lport)) {
336 				err = EINVAL;
337 				goto free_ret;
338 			}
339 			addr4 = &sin4->sin_addr;
340 			if (check_addrs &&
341 			    (addr4->s_addr == INADDR_ANY ||
342 			    addr4->s_addr == INADDR_BROADCAST ||
343 			    CLASSD(addr4->s_addr))) {
344 				err = EINVAL;
345 				goto free_ret;
346 			}
347 			IN6_INADDR_TO_V4MAPPED(addr4, &addr);
348 			if (!check_addrs && addr4->s_addr == INADDR_ANY) {
349 				lookup_saddr = B_FALSE;
350 				bind_to_all = B_TRUE;
351 			}
352 
353 			break;
354 		case AF_INET6:
355 			sin6 = (struct sockaddr_in6 *)addrs + cnt;
356 			if (sin6->sin6_family != AF_INET6 || (check_lport &&
357 			    sin6->sin6_port != connp->conn_lport)) {
358 				err = EINVAL;
359 				goto free_ret;
360 			}
361 			addr = sin6->sin6_addr;
362 			/* Contains the interface index */
363 			ifindex = sin6->sin6_scope_id;
364 			if (connp->conn_ipv6_v6only &&
365 			    IN6_IS_ADDR_V4MAPPED(&addr)) {
366 				err = EAFNOSUPPORT;
367 				goto free_ret;
368 			}
369 			if (check_addrs &&
370 			    (IN6_IS_ADDR_LINKLOCAL(&addr) ||
371 			    IN6_IS_ADDR_MULTICAST(&addr) ||
372 			    IN6_IS_ADDR_UNSPECIFIED(&addr))) {
373 				err = EINVAL;
374 				goto free_ret;
375 			}
376 			if (!check_addrs && IN6_IS_ADDR_UNSPECIFIED(&addr)) {
377 				lookup_saddr = B_FALSE;
378 				bind_to_all = B_TRUE;
379 			}
380 
381 			break;
382 		default:
383 			err = EAFNOSUPPORT;
384 			goto free_ret;
385 		}
386 		if (lookup_saddr) {
387 			ipif = sctp_lookup_ipif_addr(&addr, B_TRUE,
388 			    IPCL_ZONEID(connp), !connp->conn_allzones,
389 			    ifindex, 0, B_TRUE, sctp->sctp_sctps);
390 			if (ipif == NULL) {
391 				/* Address not in the list */
392 				err = EINVAL;
393 				goto free_ret;
394 			} else if (check_addrs && SCTP_IS_IPIF_LOOPBACK(ipif) &&
395 			    cl_sctp_check_addrs == NULL) {
396 				SCTP_IPIF_REFRELE(ipif);
397 				err = EINVAL;
398 				goto free_ret;
399 			}
400 		}
401 		if (!bind_to_all) {
402 			/*
403 			 * If an address is added after association setup,
404 			 * we need to wait for the peer to send us an ASCONF
405 			 * ACK before we can start using it.
406 			 * saddr_ipif_dontsrc will be reset (to 0) when we
407 			 * get the ASCONF ACK for this address.
408 			 */
409 			err = sctp_ipif_hash_insert(sctp, ipif, KM_SLEEP,
410 			    check_addrs ? B_TRUE : B_FALSE, B_FALSE);
411 			if (err != 0) {
412 				SCTP_IPIF_REFRELE(ipif);
413 				if (check_addrs && err == EALREADY)
414 					err = EADDRINUSE;
415 				goto free_ret;
416 			}
417 			saddr_cnt++;
418 			if (lsize >= sizeof (addr)) {
419 				bcopy(&addr, p, sizeof (addr));
420 				p += sizeof (addr);
421 				lsize -= sizeof (addr);
422 			}
423 		}
424 	}
425 	if (bind_to_all) {
426 		/*
427 		 * Free whatever we might have added before encountering
428 		 * inaddr_any.
429 		 */
430 		if (sctp->sctp_nsaddrs > 0) {
431 			sctp_free_saddrs(sctp);
432 			ASSERT(sctp->sctp_nsaddrs == 0);
433 		}
434 		err = sctp_get_all_ipifs(sctp, KM_SLEEP);
435 		if (err != 0)
436 			return (err);
437 		sctp->sctp_bound_to_all = 1;
438 	}
439 	if (sctp->sctp_listen_tfp != NULL)
440 		mutex_exit(&sctp->sctp_listen_tfp->tf_lock);
441 	if (sctp->sctp_conn_tfp != NULL)
442 		mutex_exit(&sctp->sctp_conn_tfp->tf_lock);
443 	return (0);
444 free_ret:
445 	if (saddr_cnt != 0)
446 		sctp_del_saddr_list(sctp, addrs, saddr_cnt, B_TRUE);
447 	if (sctp->sctp_listen_tfp != NULL)
448 		mutex_exit(&sctp->sctp_listen_tfp->tf_lock);
449 	if (sctp->sctp_conn_tfp != NULL)
450 		mutex_exit(&sctp->sctp_conn_tfp->tf_lock);
451 	return (err);
452 }
453 
454 static int
455 sctp_ipif_hash_insert(sctp_t *sctp, sctp_ipif_t *ipif, int sleep,
456     boolean_t dontsrc, boolean_t allow_dup)
457 {
458 	int			cnt;
459 	sctp_saddr_ipif_t	*ipif_obj;
460 	int			hindex;
461 
462 	hindex = SCTP_IPIF_ADDR_HASH(ipif->sctp_ipif_saddr,
463 	    ipif->sctp_ipif_isv6);
464 	ipif_obj = list_head(&sctp->sctp_saddrs[hindex].sctp_ipif_list);
465 	for (cnt = 0; cnt < sctp->sctp_saddrs[hindex].ipif_count; cnt++) {
466 		if (IN6_ARE_ADDR_EQUAL(&ipif_obj->saddr_ipifp->sctp_ipif_saddr,
467 		    &ipif->sctp_ipif_saddr)) {
468 			if (ipif->sctp_ipif_id !=
469 			    ipif_obj->saddr_ipifp->sctp_ipif_id &&
470 			    ipif_obj->saddr_ipifp->sctp_ipif_state ==
471 			    SCTP_IPIFS_DOWN && ipif->sctp_ipif_state ==
472 			    SCTP_IPIFS_UP) {
473 				SCTP_IPIF_REFRELE(ipif_obj->saddr_ipifp);
474 				ipif_obj->saddr_ipifp = ipif;
475 				ipif_obj->saddr_ipif_dontsrc = dontsrc ? 1 : 0;
476 				return (0);
477 			} else if (!allow_dup || ipif->sctp_ipif_id ==
478 			    ipif_obj->saddr_ipifp->sctp_ipif_id) {
479 				return (EALREADY);
480 			}
481 		}
482 		ipif_obj = list_next(&sctp->sctp_saddrs[hindex].sctp_ipif_list,
483 		    ipif_obj);
484 	}
485 	ipif_obj = kmem_zalloc(sizeof (sctp_saddr_ipif_t), sleep);
486 	if (ipif_obj == NULL) {
487 		/* Need to do something */
488 		return (ENOMEM);
489 	}
490 	ipif_obj->saddr_ipifp = ipif;
491 	ipif_obj->saddr_ipif_dontsrc = dontsrc ? 1 : 0;
492 	list_insert_tail(&sctp->sctp_saddrs[hindex].sctp_ipif_list, ipif_obj);
493 	sctp->sctp_saddrs[hindex].ipif_count++;
494 	sctp->sctp_nsaddrs++;
495 	return (0);
496 }
497 
498 /*
499  * Given a source address, walk through the peer address list to see
500  * if the source address is being used.  If it is, reset that.
501  * A cleared saddr will then make sctp_make_mp lookup the destination again
502  * and as part of that look for a new source.
503  */
504 static void
505 sctp_fix_saddr(sctp_t *sctp, in6_addr_t *saddr)
506 {
507 	sctp_faddr_t	*fp;
508 
509 	for (fp = sctp->sctp_faddrs; fp != NULL; fp = fp->next) {
510 		if (!IN6_ARE_ADDR_EQUAL(&fp->saddr, saddr))
511 			continue;
512 		V6_SET_ZERO(fp->saddr);
513 	}
514 }
515 
516 static void
517 sctp_ipif_hash_remove(sctp_t *sctp, sctp_ipif_t *ipif)
518 {
519 	int			cnt;
520 	sctp_saddr_ipif_t	*ipif_obj;
521 	int			hindex;
522 
523 	hindex = SCTP_IPIF_ADDR_HASH(ipif->sctp_ipif_saddr,
524 	    ipif->sctp_ipif_isv6);
525 	ipif_obj = list_head(&sctp->sctp_saddrs[hindex].sctp_ipif_list);
526 	for (cnt = 0; cnt < sctp->sctp_saddrs[hindex].ipif_count; cnt++) {
527 		if (IN6_ARE_ADDR_EQUAL(&ipif_obj->saddr_ipifp->sctp_ipif_saddr,
528 		    &ipif->sctp_ipif_saddr)) {
529 			list_remove(&sctp->sctp_saddrs[hindex].sctp_ipif_list,
530 			    ipif_obj);
531 			sctp->sctp_saddrs[hindex].ipif_count--;
532 			sctp->sctp_nsaddrs--;
533 			sctp_fix_saddr(sctp, &ipif->sctp_ipif_saddr);
534 			SCTP_IPIF_REFRELE(ipif_obj->saddr_ipifp);
535 			kmem_free(ipif_obj, sizeof (sctp_saddr_ipif_t));
536 			break;
537 		}
538 		ipif_obj = list_next(&sctp->sctp_saddrs[hindex].sctp_ipif_list,
539 		    ipif_obj);
540 	}
541 }
542 
543 static int
544 sctp_compare_ipif_list(sctp_ipif_hash_t *list1, sctp_ipif_hash_t *list2)
545 {
546 	int			i;
547 	int			j;
548 	sctp_saddr_ipif_t	*obj1;
549 	sctp_saddr_ipif_t	*obj2;
550 	int			overlap = 0;
551 
552 	obj1 = list_head(&list1->sctp_ipif_list);
553 	for (i = 0; i < list1->ipif_count; i++) {
554 		obj2 = list_head(&list2->sctp_ipif_list);
555 		for (j = 0; j < list2->ipif_count; j++) {
556 			if (IN6_ARE_ADDR_EQUAL(
557 			    &obj1->saddr_ipifp->sctp_ipif_saddr,
558 			    &obj2->saddr_ipifp->sctp_ipif_saddr)) {
559 				overlap++;
560 				break;
561 			}
562 			obj2 = list_next(&list2->sctp_ipif_list,
563 			    obj2);
564 		}
565 		obj1 = list_next(&list1->sctp_ipif_list, obj1);
566 	}
567 	return (overlap);
568 }
569 
570 int
571 sctp_compare_saddrs(sctp_t *sctp1, sctp_t *sctp2)
572 {
573 	int		i;
574 	int		overlap = 0;
575 
576 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
577 		overlap += sctp_compare_ipif_list(&sctp1->sctp_saddrs[i],
578 		    &sctp2->sctp_saddrs[i]);
579 	}
580 
581 	if (sctp1->sctp_nsaddrs == sctp2->sctp_nsaddrs &&
582 	    overlap == sctp1->sctp_nsaddrs) {
583 		return (SCTP_ADDR_EQUAL);
584 	}
585 
586 	if (overlap == sctp1->sctp_nsaddrs)
587 		return (SCTP_ADDR_SUBSET);
588 
589 	if (overlap > 0)
590 		return (SCTP_ADDR_OVERLAP);
591 
592 	return (SCTP_ADDR_DISJOINT);
593 }
594 
595 static int
596 sctp_copy_ipifs(sctp_ipif_hash_t *list1, sctp_t *sctp2, int sleep)
597 {
598 	int			i;
599 	sctp_saddr_ipif_t	*obj;
600 	int			error = 0;
601 
602 	obj = list_head(&list1->sctp_ipif_list);
603 	for (i = 0; i < list1->ipif_count; i++) {
604 		SCTP_IPIF_REFHOLD(obj->saddr_ipifp);
605 		error = sctp_ipif_hash_insert(sctp2, obj->saddr_ipifp, sleep,
606 		    B_FALSE, B_FALSE);
607 		ASSERT(error != EALREADY);
608 		if (error != 0)
609 			return (error);
610 		obj = list_next(&list1->sctp_ipif_list, obj);
611 	}
612 	return (error);
613 }
614 
615 int
616 sctp_dup_saddrs(sctp_t *sctp1, sctp_t *sctp2, int sleep)
617 {
618 	int	error = 0;
619 	int	i;
620 
621 	if (sctp1 == NULL || sctp1->sctp_bound_to_all == 1)
622 		return (sctp_get_all_ipifs(sctp2, sleep));
623 
624 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
625 		if (sctp1->sctp_saddrs[i].ipif_count == 0)
626 			continue;
627 		error = sctp_copy_ipifs(&sctp1->sctp_saddrs[i], sctp2, sleep);
628 		if (error != 0) {
629 			sctp_free_saddrs(sctp2);
630 			return (error);
631 		}
632 	}
633 	return (0);
634 }
635 
636 void
637 sctp_free_saddrs(sctp_t *sctp)
638 {
639 	int			i;
640 	int			l;
641 	sctp_saddr_ipif_t	*obj;
642 
643 	if (sctp->sctp_nsaddrs == 0)
644 		return;
645 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
646 		if (sctp->sctp_saddrs[i].ipif_count == 0)
647 			continue;
648 		obj = list_tail(&sctp->sctp_saddrs[i].sctp_ipif_list);
649 		for (l = 0; l < sctp->sctp_saddrs[i].ipif_count; l++) {
650 			list_remove(&sctp->sctp_saddrs[i].sctp_ipif_list, obj);
651 			SCTP_IPIF_REFRELE(obj->saddr_ipifp);
652 			sctp->sctp_nsaddrs--;
653 			kmem_free(obj, sizeof (sctp_saddr_ipif_t));
654 			obj = list_tail(&sctp->sctp_saddrs[i].sctp_ipif_list);
655 		}
656 		sctp->sctp_saddrs[i].ipif_count = 0;
657 	}
658 	if (sctp->sctp_bound_to_all == 1)
659 		sctp->sctp_bound_to_all = 0;
660 	ASSERT(sctp->sctp_nsaddrs == 0);
661 }
662 
663 /*
664  * Add/Delete the given ILL from the SCTP ILL list. Called with no locks
665  * held.
666  */
667 void
668 sctp_update_ill(ill_t *ill, int op)
669 {
670 	int		i;
671 	sctp_ill_t	*sctp_ill = NULL;
672 	uint_t		index;
673 	netstack_t	*ns = ill->ill_ipst->ips_netstack;
674 	sctp_stack_t	*sctps = ns->netstack_sctp;
675 
676 	rw_enter(&sctps->sctps_g_ills_lock, RW_WRITER);
677 
678 	index = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(ill));
679 	sctp_ill = list_head(&sctps->sctps_g_ills[index].sctp_ill_list);
680 	for (i = 0; i < sctps->sctps_g_ills[index].ill_count; i++) {
681 		if ((sctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(ill)) &&
682 		    (sctp_ill->sctp_ill_isv6 == ill->ill_isv6)) {
683 			break;
684 		}
685 		sctp_ill = list_next(&sctps->sctps_g_ills[index].sctp_ill_list,
686 		    sctp_ill);
687 	}
688 
689 	switch (op) {
690 	case SCTP_ILL_INSERT:
691 		if (sctp_ill != NULL) {
692 			/* Unmark it if it is condemned */
693 			if (sctp_ill->sctp_ill_state == SCTP_ILLS_CONDEMNED)
694 				sctp_ill->sctp_ill_state = 0;
695 			rw_exit(&sctps->sctps_g_ills_lock);
696 			return;
697 		}
698 		sctp_ill = kmem_zalloc(sizeof (sctp_ill_t), KM_NOSLEEP);
699 		/* Need to re-try? */
700 		if (sctp_ill == NULL) {
701 			cmn_err(CE_WARN, "sctp_update_ill: error adding "
702 			    "ILL %p to SCTP's ILL list", (void *)ill);
703 			rw_exit(&sctps->sctps_g_ills_lock);
704 			return;
705 		}
706 		sctp_ill->sctp_ill_name = kmem_zalloc(ill->ill_name_length,
707 		    KM_NOSLEEP);
708 		if (sctp_ill->sctp_ill_name == NULL) {
709 			cmn_err(CE_WARN, "sctp_update_ill: error adding "
710 			    "ILL %p to SCTP's ILL list", (void *)ill);
711 			kmem_free(sctp_ill, sizeof (sctp_ill_t));
712 			rw_exit(&sctps->sctps_g_ills_lock);
713 			return;
714 		}
715 		bcopy(ill->ill_name, sctp_ill->sctp_ill_name,
716 		    ill->ill_name_length);
717 		sctp_ill->sctp_ill_name_length = ill->ill_name_length;
718 		sctp_ill->sctp_ill_index = SCTP_ILL_TO_PHYINDEX(ill);
719 		sctp_ill->sctp_ill_flags = ill->ill_phyint->phyint_flags;
720 		sctp_ill->sctp_ill_netstack = ns;	/* No netstack_hold */
721 		sctp_ill->sctp_ill_isv6 = ill->ill_isv6;
722 		list_insert_tail(&sctps->sctps_g_ills[index].sctp_ill_list,
723 		    (void *)sctp_ill);
724 		sctps->sctps_g_ills[index].ill_count++;
725 		sctps->sctps_ills_count++;
726 
727 		break;
728 
729 	case SCTP_ILL_REMOVE:
730 
731 		if (sctp_ill == NULL) {
732 			rw_exit(&sctps->sctps_g_ills_lock);
733 			return;
734 		}
735 		if (sctp_ill->sctp_ill_ipifcnt == 0) {
736 			list_remove(&sctps->sctps_g_ills[index].sctp_ill_list,
737 			    (void *)sctp_ill);
738 			sctps->sctps_g_ills[index].ill_count--;
739 			sctps->sctps_ills_count--;
740 			kmem_free(sctp_ill->sctp_ill_name,
741 			    ill->ill_name_length);
742 			kmem_free(sctp_ill, sizeof (sctp_ill_t));
743 		} else {
744 			sctp_ill->sctp_ill_state = SCTP_ILLS_CONDEMNED;
745 		}
746 
747 		break;
748 	}
749 	rw_exit(&sctps->sctps_g_ills_lock);
750 }
751 
752 /*
753  * The ILL's index is being changed, just remove it from the old list,
754  * change the SCTP ILL's index and re-insert using the new index.
755  */
756 void
757 sctp_ill_reindex(ill_t *ill, uint_t orig_ill_index)
758 {
759 	sctp_ill_t	*sctp_ill = NULL;
760 	sctp_ill_t	*nxt_sill;
761 	uint_t		indx;
762 	uint_t		nindx;
763 	boolean_t	once = B_FALSE;
764 	netstack_t	*ns = ill->ill_ipst->ips_netstack;
765 	sctp_stack_t	*sctps = ns->netstack_sctp;
766 
767 	rw_enter(&sctps->sctps_g_ills_lock, RW_WRITER);
768 
769 	indx = SCTP_ILL_HASH_FN(orig_ill_index);
770 	nindx = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(ill));
771 	sctp_ill = list_head(&sctps->sctps_g_ills[indx].sctp_ill_list);
772 	while (sctp_ill != NULL) {
773 		nxt_sill = list_next(&sctps->sctps_g_ills[indx].sctp_ill_list,
774 		    sctp_ill);
775 		if (sctp_ill->sctp_ill_index == orig_ill_index) {
776 			sctp_ill->sctp_ill_index = SCTP_ILL_TO_PHYINDEX(ill);
777 			/*
778 			 * if the new index hashes to the same value, all's
779 			 * done.
780 			 */
781 			if (nindx != indx) {
782 				list_remove(
783 				    &sctps->sctps_g_ills[indx].sctp_ill_list,
784 				    (void *)sctp_ill);
785 				sctps->sctps_g_ills[indx].ill_count--;
786 				list_insert_tail(
787 				    &sctps->sctps_g_ills[nindx].sctp_ill_list,
788 				    (void *)sctp_ill);
789 				sctps->sctps_g_ills[nindx].ill_count++;
790 			}
791 			if (once)
792 				break;
793 			/* We might have one for v4 and for v6 */
794 			once = B_TRUE;
795 		}
796 		sctp_ill = nxt_sill;
797 	}
798 	rw_exit(&sctps->sctps_g_ills_lock);
799 }
800 
801 /* move ipif from f_ill to t_ill */
802 void
803 sctp_move_ipif(ipif_t *ipif, ill_t *f_ill, ill_t *t_ill)
804 {
805 	sctp_ill_t	*fsctp_ill = NULL;
806 	sctp_ill_t	*tsctp_ill = NULL;
807 	sctp_ipif_t	*sctp_ipif;
808 	uint_t		hindex;
809 	int		i;
810 	netstack_t	*ns = ipif->ipif_ill->ill_ipst->ips_netstack;
811 	sctp_stack_t	*sctps = ns->netstack_sctp;
812 
813 	rw_enter(&sctps->sctps_g_ills_lock, RW_READER);
814 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_READER);
815 
816 	hindex = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(f_ill));
817 	fsctp_ill = list_head(&sctps->sctps_g_ills[hindex].sctp_ill_list);
818 	for (i = 0; i < sctps->sctps_g_ills[hindex].ill_count; i++) {
819 		if (fsctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(f_ill) &&
820 		    fsctp_ill->sctp_ill_isv6 == f_ill->ill_isv6) {
821 			break;
822 		}
823 		fsctp_ill = list_next(
824 		    &sctps->sctps_g_ills[hindex].sctp_ill_list, fsctp_ill);
825 	}
826 
827 	hindex = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(t_ill));
828 	tsctp_ill = list_head(&sctps->sctps_g_ills[hindex].sctp_ill_list);
829 	for (i = 0; i < sctps->sctps_g_ills[hindex].ill_count; i++) {
830 		if (tsctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(t_ill) &&
831 		    tsctp_ill->sctp_ill_isv6 == t_ill->ill_isv6) {
832 			break;
833 		}
834 		tsctp_ill = list_next(
835 		    &sctps->sctps_g_ills[hindex].sctp_ill_list, tsctp_ill);
836 	}
837 
838 	hindex = SCTP_IPIF_ADDR_HASH(ipif->ipif_v6lcl_addr,
839 	    ipif->ipif_ill->ill_isv6);
840 	sctp_ipif = list_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list);
841 	for (i = 0; i < sctps->sctps_g_ipifs[hindex].ipif_count; i++) {
842 		if (sctp_ipif->sctp_ipif_id == ipif->ipif_seqid)
843 			break;
844 		sctp_ipif = list_next(
845 		    &sctps->sctps_g_ipifs[hindex].sctp_ipif_list, sctp_ipif);
846 	}
847 	/* Should be an ASSERT? */
848 	if (fsctp_ill == NULL || tsctp_ill == NULL || sctp_ipif == NULL) {
849 		ip1dbg(("sctp_move_ipif: error moving ipif %p from %p to %p\n",
850 		    (void *)ipif, (void *)f_ill, (void *)t_ill));
851 		rw_exit(&sctps->sctps_g_ipifs_lock);
852 		rw_exit(&sctps->sctps_g_ills_lock);
853 		return;
854 	}
855 	rw_enter(&sctp_ipif->sctp_ipif_lock, RW_WRITER);
856 	ASSERT(sctp_ipif->sctp_ipif_ill == fsctp_ill);
857 	sctp_ipif->sctp_ipif_ill = tsctp_ill;
858 	rw_exit(&sctp_ipif->sctp_ipif_lock);
859 	(void) atomic_add_32_nv(&fsctp_ill->sctp_ill_ipifcnt, -1);
860 	atomic_add_32(&tsctp_ill->sctp_ill_ipifcnt, 1);
861 	rw_exit(&sctps->sctps_g_ipifs_lock);
862 	rw_exit(&sctps->sctps_g_ills_lock);
863 }
864 
865 /*
866  * Walk the list of SCTPs and find each that has oipif in it's saddr list, and
867  * if so replace it with nipif.
868  */
869 void
870 sctp_update_saddrs(sctp_ipif_t *oipif, sctp_ipif_t *nipif, int idx,
871     sctp_stack_t *sctps)
872 {
873 	sctp_t			*sctp;
874 	sctp_t			*sctp_prev = NULL;
875 	sctp_saddr_ipif_t	*sobj;
876 	int			count;
877 
878 	mutex_enter(&sctps->sctps_g_lock);
879 	sctp = list_head(&sctps->sctps_g_list);
880 	while (sctp != NULL && oipif->sctp_ipif_refcnt > 0) {
881 		mutex_enter(&sctp->sctp_reflock);
882 		if (sctp->sctp_condemned ||
883 		    sctp->sctp_saddrs[idx].ipif_count <= 0) {
884 			mutex_exit(&sctp->sctp_reflock);
885 			sctp = list_next(&sctps->sctps_g_list, sctp);
886 			continue;
887 		}
888 		sctp->sctp_refcnt++;
889 		mutex_exit(&sctp->sctp_reflock);
890 		mutex_exit(&sctps->sctps_g_lock);
891 		if (sctp_prev != NULL)
892 			SCTP_REFRELE(sctp_prev);
893 
894 		RUN_SCTP(sctp);
895 		sobj = list_head(&sctp->sctp_saddrs[idx].sctp_ipif_list);
896 		for (count = 0; count <
897 		    sctp->sctp_saddrs[idx].ipif_count; count++) {
898 			if (sobj->saddr_ipifp == oipif) {
899 				SCTP_IPIF_REFHOLD(nipif);
900 				sobj->saddr_ipifp = nipif;
901 				ASSERT(oipif->sctp_ipif_refcnt > 0);
902 				/* We have the writer lock */
903 				oipif->sctp_ipif_refcnt--;
904 				/*
905 				 * Can't have more than one referring
906 				 * to the same sctp_ipif.
907 				 */
908 				break;
909 			}
910 			sobj = list_next(&sctp->sctp_saddrs[idx].sctp_ipif_list,
911 			    sobj);
912 		}
913 		WAKE_SCTP(sctp);
914 		sctp_prev = sctp;
915 		mutex_enter(&sctps->sctps_g_lock);
916 		sctp = list_next(&sctps->sctps_g_list, sctp);
917 	}
918 	mutex_exit(&sctps->sctps_g_lock);
919 	if (sctp_prev != NULL)
920 		SCTP_REFRELE(sctp_prev);
921 }
922 
923 /*
924  * Given an ipif, walk the hash list in the global ipif table and for
925  * any other SCTP ipif with the same address and non-zero reference, walk
926  * the SCTP list and update the saddr list, if required, to point to the
927  * new SCTP ipif. If it is a loopback interface, then there could be
928  * multiple interfaces with 127.0.0.1 if there are zones configured, so
929  * check the zoneid in addition to the address.
930  */
931 void
932 sctp_chk_and_updt_saddr(int hindex, sctp_ipif_t *ipif, sctp_stack_t *sctps)
933 {
934 	int		cnt;
935 	sctp_ipif_t	*sipif;
936 
937 	ASSERT(sctps->sctps_g_ipifs[hindex].ipif_count > 0);
938 	ASSERT(ipif->sctp_ipif_state == SCTP_IPIFS_UP);
939 
940 	sipif = list_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list);
941 	for (cnt = 0; cnt < sctps->sctps_g_ipifs[hindex].ipif_count; cnt++) {
942 		rw_enter(&sipif->sctp_ipif_lock, RW_WRITER);
943 		if (sipif->sctp_ipif_id != ipif->sctp_ipif_id &&
944 		    IN6_ARE_ADDR_EQUAL(&sipif->sctp_ipif_saddr,
945 		    &ipif->sctp_ipif_saddr) && sipif->sctp_ipif_refcnt > 0 &&
946 		    (!SCTP_IS_IPIF_LOOPBACK(ipif) || ipif->sctp_ipif_zoneid ==
947 		    sipif->sctp_ipif_zoneid)) {
948 			/*
949 			 * There can only be one address up at any time
950 			 * and we are here because ipif has been brought
951 			 * up.
952 			 */
953 			ASSERT(sipif->sctp_ipif_state != SCTP_IPIFS_UP);
954 			/*
955 			 * Someone has a reference to this we need to update to
956 			 * point to the new sipif.
957 			 */
958 			sctp_update_saddrs(sipif, ipif, hindex, sctps);
959 		}
960 		rw_exit(&sipif->sctp_ipif_lock);
961 		sipif = list_next(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list,
962 		    sipif);
963 	}
964 }
965 
966 /*
967  * Insert a new SCTP ipif using 'ipif'. v6addr is the address that existed
968  * prior to the current address in 'ipif'. Only when an existing address
969  * is changed on an IPIF, will v6addr be specified. If the IPIF already
970  * exists in the global SCTP ipif table, then we either removed it, if
971  * it doesn't have any existing reference, or mark it condemned otherwise.
972  * If an address is being brought up (IPIF_UP), then we need to scan
973  * the SCTP list to check if there is any SCTP that points to the *same*
974  * address on a different SCTP ipif and update in that case.
975  */
976 void
977 sctp_update_ipif_addr(ipif_t *ipif, in6_addr_t v6addr)
978 {
979 	ill_t		*ill = ipif->ipif_ill;
980 	int		i;
981 	sctp_ill_t	*sctp_ill;
982 	sctp_ill_t	*osctp_ill;
983 	sctp_ipif_t	*sctp_ipif = NULL;
984 	sctp_ipif_t	*osctp_ipif = NULL;
985 	uint_t		ill_index;
986 	int		hindex;
987 	sctp_stack_t	*sctps;
988 
989 	sctps = ipif->ipif_ill->ill_ipst->ips_netstack->netstack_sctp;
990 
991 	/* Index for new address */
992 	hindex = SCTP_IPIF_ADDR_HASH(ipif->ipif_v6lcl_addr, ill->ill_isv6);
993 
994 	/*
995 	 * The address on this IPIF is changing, we need to look for
996 	 * this old address and mark it condemned, before creating
997 	 * one for the new address.
998 	 */
999 	osctp_ipif = sctp_lookup_ipif_addr(&v6addr, B_FALSE,
1000 	    ipif->ipif_zoneid, B_TRUE, SCTP_ILL_TO_PHYINDEX(ill),
1001 	    ipif->ipif_seqid, B_FALSE, sctps);
1002 
1003 	rw_enter(&sctps->sctps_g_ills_lock, RW_READER);
1004 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_WRITER);
1005 
1006 	ill_index = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(ill));
1007 	sctp_ill = list_head(&sctps->sctps_g_ills[ill_index].sctp_ill_list);
1008 	for (i = 0; i < sctps->sctps_g_ills[ill_index].ill_count; i++) {
1009 		if (sctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(ill) &&
1010 		    sctp_ill->sctp_ill_isv6 == ill->ill_isv6) {
1011 			break;
1012 		}
1013 		sctp_ill = list_next(
1014 		    &sctps->sctps_g_ills[ill_index].sctp_ill_list, sctp_ill);
1015 	}
1016 
1017 	if (sctp_ill == NULL) {
1018 		ip1dbg(("sctp_update_ipif_addr: ill not found ..\n"));
1019 		rw_exit(&sctps->sctps_g_ipifs_lock);
1020 		rw_exit(&sctps->sctps_g_ills_lock);
1021 		return;
1022 	}
1023 
1024 	if (osctp_ipif != NULL) {
1025 
1026 		/* The address is the same? */
1027 		if (IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6lcl_addr, &v6addr)) {
1028 			boolean_t	chk_n_updt = B_FALSE;
1029 
1030 			rw_downgrade(&sctps->sctps_g_ipifs_lock);
1031 			rw_enter(&osctp_ipif->sctp_ipif_lock, RW_WRITER);
1032 			if (ipif->ipif_flags & IPIF_UP &&
1033 			    osctp_ipif->sctp_ipif_state != SCTP_IPIFS_UP) {
1034 				osctp_ipif->sctp_ipif_state = SCTP_IPIFS_UP;
1035 				chk_n_updt = B_TRUE;
1036 			} else {
1037 				osctp_ipif->sctp_ipif_state = SCTP_IPIFS_DOWN;
1038 			}
1039 			osctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1040 			rw_exit(&osctp_ipif->sctp_ipif_lock);
1041 			if (chk_n_updt) {
1042 				sctp_chk_and_updt_saddr(hindex, osctp_ipif,
1043 				    sctps);
1044 			}
1045 			rw_exit(&sctps->sctps_g_ipifs_lock);
1046 			rw_exit(&sctps->sctps_g_ills_lock);
1047 			return;
1048 		}
1049 		/*
1050 		 * We are effectively removing this address from the ILL.
1051 		 */
1052 		if (osctp_ipif->sctp_ipif_refcnt != 0) {
1053 			osctp_ipif->sctp_ipif_state = SCTP_IPIFS_CONDEMNED;
1054 		} else {
1055 			list_t		*ipif_list;
1056 			int		ohindex;
1057 
1058 			osctp_ill = osctp_ipif->sctp_ipif_ill;
1059 			/* hash index for the old one */
1060 			ohindex = SCTP_IPIF_ADDR_HASH(
1061 			    osctp_ipif->sctp_ipif_saddr,
1062 			    osctp_ipif->sctp_ipif_isv6);
1063 
1064 			ipif_list =
1065 			    &sctps->sctps_g_ipifs[ohindex].sctp_ipif_list;
1066 
1067 			list_remove(ipif_list, (void *)osctp_ipif);
1068 			sctps->sctps_g_ipifs[ohindex].ipif_count--;
1069 			sctps->sctps_g_ipifs_count--;
1070 			rw_destroy(&osctp_ipif->sctp_ipif_lock);
1071 			kmem_free(osctp_ipif, sizeof (sctp_ipif_t));
1072 			(void) atomic_add_32_nv(&osctp_ill->sctp_ill_ipifcnt,
1073 			    -1);
1074 		}
1075 	}
1076 
1077 	sctp_ipif = kmem_zalloc(sizeof (sctp_ipif_t), KM_NOSLEEP);
1078 	/* Try again? */
1079 	if (sctp_ipif == NULL) {
1080 		cmn_err(CE_WARN, "sctp_update_ipif_addr: error adding "
1081 		    "IPIF %p to SCTP's IPIF list", (void *)ipif);
1082 		rw_exit(&sctps->sctps_g_ipifs_lock);
1083 		rw_exit(&sctps->sctps_g_ills_lock);
1084 		return;
1085 	}
1086 	sctps->sctps_g_ipifs_count++;
1087 	rw_init(&sctp_ipif->sctp_ipif_lock, NULL, RW_DEFAULT, NULL);
1088 	sctp_ipif->sctp_ipif_saddr = ipif->ipif_v6lcl_addr;
1089 	sctp_ipif->sctp_ipif_ill = sctp_ill;
1090 	sctp_ipif->sctp_ipif_isv6 = ill->ill_isv6;
1091 	sctp_ipif->sctp_ipif_zoneid = ipif->ipif_zoneid;
1092 	sctp_ipif->sctp_ipif_id = ipif->ipif_seqid;
1093 	if (ipif->ipif_flags & IPIF_UP)
1094 		sctp_ipif->sctp_ipif_state = SCTP_IPIFS_UP;
1095 	else
1096 		sctp_ipif->sctp_ipif_state = SCTP_IPIFS_DOWN;
1097 	sctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1098 	/*
1099 	 * We add it to the head so that it is quicker to find good/recent
1100 	 * additions.
1101 	 */
1102 	list_insert_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list,
1103 	    (void *)sctp_ipif);
1104 	sctps->sctps_g_ipifs[hindex].ipif_count++;
1105 	atomic_add_32(&sctp_ill->sctp_ill_ipifcnt, 1);
1106 	if (sctp_ipif->sctp_ipif_state == SCTP_IPIFS_UP)
1107 		sctp_chk_and_updt_saddr(hindex, sctp_ipif, sctps);
1108 	rw_exit(&sctps->sctps_g_ipifs_lock);
1109 	rw_exit(&sctps->sctps_g_ills_lock);
1110 }
1111 
1112 /* Insert, Remove,  Mark up or Mark down the ipif */
1113 void
1114 sctp_update_ipif(ipif_t *ipif, int op)
1115 {
1116 	ill_t		*ill = ipif->ipif_ill;
1117 	int		i;
1118 	sctp_ill_t	*sctp_ill;
1119 	sctp_ipif_t	*sctp_ipif;
1120 	uint_t		ill_index;
1121 	uint_t		hindex;
1122 	netstack_t	*ns = ipif->ipif_ill->ill_ipst->ips_netstack;
1123 	sctp_stack_t	*sctps = ns->netstack_sctp;
1124 
1125 	ip2dbg(("sctp_update_ipif: %s %d\n", ill->ill_name, ipif->ipif_seqid));
1126 
1127 	rw_enter(&sctps->sctps_g_ills_lock, RW_READER);
1128 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_WRITER);
1129 
1130 	ill_index = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(ill));
1131 	sctp_ill = list_head(&sctps->sctps_g_ills[ill_index].sctp_ill_list);
1132 	for (i = 0; i < sctps->sctps_g_ills[ill_index].ill_count; i++) {
1133 		if (sctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(ill) &&
1134 		    sctp_ill->sctp_ill_isv6 == ill->ill_isv6) {
1135 			break;
1136 		}
1137 		sctp_ill = list_next(
1138 		    &sctps->sctps_g_ills[ill_index].sctp_ill_list, sctp_ill);
1139 	}
1140 	if (sctp_ill == NULL) {
1141 		rw_exit(&sctps->sctps_g_ipifs_lock);
1142 		rw_exit(&sctps->sctps_g_ills_lock);
1143 		return;
1144 	}
1145 
1146 	hindex = SCTP_IPIF_ADDR_HASH(ipif->ipif_v6lcl_addr,
1147 	    ipif->ipif_ill->ill_isv6);
1148 	sctp_ipif = list_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list);
1149 	for (i = 0; i < sctps->sctps_g_ipifs[hindex].ipif_count; i++) {
1150 		if (sctp_ipif->sctp_ipif_id == ipif->ipif_seqid) {
1151 			ASSERT(IN6_ARE_ADDR_EQUAL(&sctp_ipif->sctp_ipif_saddr,
1152 			    &ipif->ipif_v6lcl_addr));
1153 			break;
1154 		}
1155 		sctp_ipif = list_next(
1156 		    &sctps->sctps_g_ipifs[hindex].sctp_ipif_list,
1157 		    sctp_ipif);
1158 	}
1159 	if (sctp_ipif == NULL) {
1160 		ip1dbg(("sctp_update_ipif: null sctp_ipif for %d\n", op));
1161 		rw_exit(&sctps->sctps_g_ipifs_lock);
1162 		rw_exit(&sctps->sctps_g_ills_lock);
1163 		return;
1164 	}
1165 	ASSERT(sctp_ill == sctp_ipif->sctp_ipif_ill);
1166 	switch (op) {
1167 	case SCTP_IPIF_REMOVE:
1168 	{
1169 		list_t		*ipif_list;
1170 		list_t		*ill_list;
1171 
1172 		ill_list = &sctps->sctps_g_ills[ill_index].sctp_ill_list;
1173 		ipif_list = &sctps->sctps_g_ipifs[hindex].sctp_ipif_list;
1174 		if (sctp_ipif->sctp_ipif_refcnt != 0) {
1175 			sctp_ipif->sctp_ipif_state = SCTP_IPIFS_CONDEMNED;
1176 			rw_exit(&sctps->sctps_g_ipifs_lock);
1177 			rw_exit(&sctps->sctps_g_ills_lock);
1178 			return;
1179 		}
1180 		list_remove(ipif_list, (void *)sctp_ipif);
1181 		sctps->sctps_g_ipifs[hindex].ipif_count--;
1182 		sctps->sctps_g_ipifs_count--;
1183 		rw_destroy(&sctp_ipif->sctp_ipif_lock);
1184 		kmem_free(sctp_ipif, sizeof (sctp_ipif_t));
1185 		(void) atomic_add_32_nv(&sctp_ill->sctp_ill_ipifcnt, -1);
1186 		if (rw_tryupgrade(&sctps->sctps_g_ills_lock) != 0) {
1187 			rw_downgrade(&sctps->sctps_g_ipifs_lock);
1188 			if (sctp_ill->sctp_ill_ipifcnt == 0 &&
1189 			    sctp_ill->sctp_ill_state == SCTP_ILLS_CONDEMNED) {
1190 				list_remove(ill_list, (void *)sctp_ill);
1191 				sctps->sctps_ills_count--;
1192 				sctps->sctps_g_ills[ill_index].ill_count--;
1193 				kmem_free(sctp_ill->sctp_ill_name,
1194 				    sctp_ill->sctp_ill_name_length);
1195 				kmem_free(sctp_ill, sizeof (sctp_ill_t));
1196 			}
1197 		}
1198 		break;
1199 	}
1200 
1201 	case SCTP_IPIF_UP:
1202 
1203 		rw_downgrade(&sctps->sctps_g_ipifs_lock);
1204 		rw_enter(&sctp_ipif->sctp_ipif_lock, RW_WRITER);
1205 		sctp_ipif->sctp_ipif_state = SCTP_IPIFS_UP;
1206 		sctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1207 		rw_exit(&sctp_ipif->sctp_ipif_lock);
1208 		sctp_chk_and_updt_saddr(hindex, sctp_ipif,
1209 		    ipif->ipif_ill->ill_ipst->ips_netstack->netstack_sctp);
1210 
1211 		break;
1212 
1213 	case SCTP_IPIF_UPDATE:
1214 
1215 		rw_downgrade(&sctps->sctps_g_ipifs_lock);
1216 		rw_enter(&sctp_ipif->sctp_ipif_lock, RW_WRITER);
1217 		sctp_ipif->sctp_ipif_zoneid = ipif->ipif_zoneid;
1218 		sctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1219 		rw_exit(&sctp_ipif->sctp_ipif_lock);
1220 
1221 		break;
1222 
1223 	case SCTP_IPIF_DOWN:
1224 
1225 		rw_downgrade(&sctps->sctps_g_ipifs_lock);
1226 		rw_enter(&sctp_ipif->sctp_ipif_lock, RW_WRITER);
1227 		sctp_ipif->sctp_ipif_state = SCTP_IPIFS_DOWN;
1228 		sctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1229 		rw_exit(&sctp_ipif->sctp_ipif_lock);
1230 
1231 		break;
1232 	}
1233 	rw_exit(&sctps->sctps_g_ipifs_lock);
1234 	rw_exit(&sctps->sctps_g_ills_lock);
1235 }
1236 
1237 /*
1238  * SCTP source address list manipulaton, locking not used (except for
1239  * sctp locking by the caller.
1240  */
1241 
1242 /* Remove a specific saddr from the list */
1243 void
1244 sctp_del_saddr(sctp_t *sctp, sctp_saddr_ipif_t *sp)
1245 {
1246 	if (sctp->sctp_conn_tfp != NULL)
1247 		mutex_enter(&sctp->sctp_conn_tfp->tf_lock);
1248 
1249 	if (sctp->sctp_listen_tfp != NULL)
1250 		mutex_enter(&sctp->sctp_listen_tfp->tf_lock);
1251 
1252 	sctp_ipif_hash_remove(sctp, sp->saddr_ipifp);
1253 
1254 	if (sctp->sctp_bound_to_all == 1)
1255 		sctp->sctp_bound_to_all = 0;
1256 
1257 	if (sctp->sctp_conn_tfp != NULL)
1258 		mutex_exit(&sctp->sctp_conn_tfp->tf_lock);
1259 
1260 	if (sctp->sctp_listen_tfp != NULL)
1261 		mutex_exit(&sctp->sctp_listen_tfp->tf_lock);
1262 }
1263 
1264 /*
1265  * Delete source address from the existing list. No error checking done here
1266  * Called with no locks held.
1267  */
1268 void
1269 sctp_del_saddr_list(sctp_t *sctp, const void *addrs, int addcnt,
1270     boolean_t fanout_locked)
1271 {
1272 	struct sockaddr_in	*sin4;
1273 	struct sockaddr_in6	*sin6;
1274 	int			cnt;
1275 	in6_addr_t		addr;
1276 	sctp_ipif_t		*sctp_ipif;
1277 	int			ifindex = 0;
1278 	conn_t			*connp = sctp->sctp_connp;
1279 
1280 	ASSERT(sctp->sctp_nsaddrs >= addcnt);
1281 
1282 	if (!fanout_locked) {
1283 		if (sctp->sctp_conn_tfp != NULL)
1284 			mutex_enter(&sctp->sctp_conn_tfp->tf_lock);
1285 		if (sctp->sctp_listen_tfp != NULL)
1286 			mutex_enter(&sctp->sctp_listen_tfp->tf_lock);
1287 	}
1288 
1289 	for (cnt = 0; cnt < addcnt; cnt++) {
1290 		switch (connp->conn_family) {
1291 		case AF_INET:
1292 			sin4 = (struct sockaddr_in *)addrs + cnt;
1293 			IN6_INADDR_TO_V4MAPPED(&sin4->sin_addr, &addr);
1294 			break;
1295 
1296 		case AF_INET6:
1297 			sin6 = (struct sockaddr_in6 *)addrs + cnt;
1298 			addr = sin6->sin6_addr;
1299 			ifindex = sin6->sin6_scope_id;
1300 			break;
1301 		}
1302 		sctp_ipif = sctp_lookup_ipif_addr(&addr, B_FALSE,
1303 		    IPCL_ZONEID(connp), !connp->conn_allzones,
1304 		    ifindex, 0, B_TRUE, sctp->sctp_sctps);
1305 		ASSERT(sctp_ipif != NULL);
1306 		sctp_ipif_hash_remove(sctp, sctp_ipif);
1307 	}
1308 	if (sctp->sctp_bound_to_all == 1)
1309 		sctp->sctp_bound_to_all = 0;
1310 
1311 	if (!fanout_locked) {
1312 		if (sctp->sctp_conn_tfp != NULL)
1313 			mutex_exit(&sctp->sctp_conn_tfp->tf_lock);
1314 		if (sctp->sctp_listen_tfp != NULL)
1315 			mutex_exit(&sctp->sctp_listen_tfp->tf_lock);
1316 	}
1317 }
1318 
1319 /*
1320  * Given an address get the corresponding entry from the list
1321  * Called with no locks held.
1322  */
1323 sctp_saddr_ipif_t *
1324 sctp_saddr_lookup(sctp_t *sctp, in6_addr_t *addr, uint_t ifindex)
1325 {
1326 	int			cnt;
1327 	sctp_saddr_ipif_t	*ipif_obj;
1328 	int			hindex;
1329 	sctp_ipif_t		*sctp_ipif;
1330 
1331 	hindex = SCTP_IPIF_ADDR_HASH(*addr, !IN6_IS_ADDR_V4MAPPED(addr));
1332 	if (sctp->sctp_saddrs[hindex].ipif_count == 0)
1333 		return (NULL);
1334 
1335 	ipif_obj = list_head(&sctp->sctp_saddrs[hindex].sctp_ipif_list);
1336 	for (cnt = 0; cnt < sctp->sctp_saddrs[hindex].ipif_count; cnt++) {
1337 		sctp_ipif = ipif_obj->saddr_ipifp;
1338 		/*
1339 		 * Zone check shouldn't be needed.
1340 		 */
1341 		if (IN6_ARE_ADDR_EQUAL(addr, &sctp_ipif->sctp_ipif_saddr) &&
1342 		    (ifindex == 0 ||
1343 		    ifindex == sctp_ipif->sctp_ipif_ill->sctp_ill_index) &&
1344 		    SCTP_IPIF_USABLE(sctp_ipif->sctp_ipif_state)) {
1345 			return (ipif_obj);
1346 		}
1347 		ipif_obj = list_next(&sctp->sctp_saddrs[hindex].sctp_ipif_list,
1348 		    ipif_obj);
1349 	}
1350 	return (NULL);
1351 }
1352 
1353 /* Given an address, add it to the source address list */
1354 int
1355 sctp_saddr_add_addr(sctp_t *sctp, in6_addr_t *addr, uint_t ifindex)
1356 {
1357 	sctp_ipif_t		*sctp_ipif;
1358 	conn_t			*connp = sctp->sctp_connp;
1359 
1360 	sctp_ipif = sctp_lookup_ipif_addr(addr, B_TRUE, IPCL_ZONEID(connp),
1361 	    !connp->conn_allzones, ifindex, 0, B_TRUE, sctp->sctp_sctps);
1362 	if (sctp_ipif == NULL)
1363 		return (EINVAL);
1364 
1365 	if (sctp_ipif_hash_insert(sctp, sctp_ipif, KM_NOSLEEP, B_FALSE,
1366 	    B_FALSE) != 0) {
1367 		SCTP_IPIF_REFRELE(sctp_ipif);
1368 		return (EINVAL);
1369 	}
1370 	return (0);
1371 }
1372 
1373 /*
1374  * Remove or mark as dontsrc addresses that are currently not part of the
1375  * association. One would delete addresses when processing an INIT and
1376  * mark as dontsrc when processing an INIT-ACK.
1377  */
1378 void
1379 sctp_check_saddr(sctp_t *sctp, int supp_af, boolean_t delete,
1380     in6_addr_t *no_del_addr)
1381 {
1382 	int			i;
1383 	int			l;
1384 	sctp_saddr_ipif_t	*obj;
1385 	int			scanned = 0;
1386 	int			naddr;
1387 	int			nsaddr;
1388 	conn_t			*connp = sctp->sctp_connp;
1389 
1390 	ASSERT(!sctp->sctp_loopback && !sctp->sctp_linklocal && supp_af != 0);
1391 
1392 	/*
1393 	 * Irregardless of the supported address in the INIT, v4
1394 	 * must be supported.
1395 	 */
1396 	if (connp->conn_family == AF_INET)
1397 		supp_af = PARM_SUPP_V4;
1398 
1399 	nsaddr = sctp->sctp_nsaddrs;
1400 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
1401 		if (sctp->sctp_saddrs[i].ipif_count == 0)
1402 			continue;
1403 		obj = list_head(&sctp->sctp_saddrs[i].sctp_ipif_list);
1404 		naddr = sctp->sctp_saddrs[i].ipif_count;
1405 		for (l = 0; l < naddr; l++) {
1406 			sctp_ipif_t	*ipif;
1407 
1408 			ipif = obj->saddr_ipifp;
1409 			scanned++;
1410 
1411 			if (IN6_ARE_ADDR_EQUAL(&ipif->sctp_ipif_saddr,
1412 			    no_del_addr)) {
1413 				goto next_obj;
1414 			}
1415 
1416 			/*
1417 			 * Delete/mark dontsrc loopback/linklocal addresses and
1418 			 * unsupported address.
1419 			 * On a clustered node, we trust the clustering module
1420 			 * to do the right thing w.r.t loopback addresses, so
1421 			 * we ignore loopback addresses in this check.
1422 			 */
1423 			if ((SCTP_IS_IPIF_LOOPBACK(ipif) &&
1424 			    cl_sctp_check_addrs == NULL) ||
1425 			    SCTP_IS_IPIF_LINKLOCAL(ipif) ||
1426 			    SCTP_UNSUPP_AF(ipif, supp_af)) {
1427 				if (!delete) {
1428 					obj->saddr_ipif_unconfirmed = 1;
1429 					goto next_obj;
1430 				}
1431 				if (sctp->sctp_bound_to_all == 1)
1432 					sctp->sctp_bound_to_all = 0;
1433 				if (scanned < nsaddr) {
1434 					obj = list_next(&sctp->sctp_saddrs[i].
1435 					    sctp_ipif_list, obj);
1436 					sctp_ipif_hash_remove(sctp, ipif);
1437 					continue;
1438 				}
1439 				sctp_ipif_hash_remove(sctp, ipif);
1440 			}
1441 	next_obj:
1442 			if (scanned >= nsaddr)
1443 				return;
1444 			obj = list_next(&sctp->sctp_saddrs[i].sctp_ipif_list,
1445 			    obj);
1446 		}
1447 	}
1448 }
1449 
1450 
1451 /* Get the first valid address from the list. Called with no locks held */
1452 in6_addr_t
1453 sctp_get_valid_addr(sctp_t *sctp, boolean_t isv6, boolean_t *addr_set)
1454 {
1455 	int			i;
1456 	int			l;
1457 	sctp_saddr_ipif_t	*obj;
1458 	int			scanned = 0;
1459 	in6_addr_t		addr;
1460 
1461 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
1462 		if (sctp->sctp_saddrs[i].ipif_count == 0)
1463 			continue;
1464 		obj = list_head(&sctp->sctp_saddrs[i].sctp_ipif_list);
1465 		for (l = 0; l < sctp->sctp_saddrs[i].ipif_count; l++) {
1466 			sctp_ipif_t	*ipif;
1467 
1468 			ipif = obj->saddr_ipifp;
1469 			if (!SCTP_DONT_SRC(obj) &&
1470 			    ipif->sctp_ipif_isv6 == isv6 &&
1471 			    ipif->sctp_ipif_state == SCTP_IPIFS_UP) {
1472 				*addr_set = B_TRUE;
1473 				return (ipif->sctp_ipif_saddr);
1474 			}
1475 			scanned++;
1476 			if (scanned >= sctp->sctp_nsaddrs)
1477 				goto got_none;
1478 			obj = list_next(&sctp->sctp_saddrs[i].sctp_ipif_list,
1479 			    obj);
1480 		}
1481 	}
1482 got_none:
1483 	/* Need to double check this */
1484 	if (isv6 == B_TRUE)
1485 		addr =  ipv6_all_zeros;
1486 	else
1487 		IN6_IPADDR_TO_V4MAPPED(0, &addr);
1488 	*addr_set = B_FALSE;
1489 	return (addr);
1490 }
1491 
1492 /*
1493  * Return the list of local addresses of an association.  The parameter
1494  * myaddrs is supposed to be either (struct sockaddr_in *) or (struct
1495  * sockaddr_in6 *) depending on the address family.
1496  */
1497 int
1498 sctp_getmyaddrs(void *conn, void *myaddrs, int *addrcnt)
1499 {
1500 	int			i;
1501 	int			l;
1502 	sctp_saddr_ipif_t	*obj;
1503 	sctp_t			*sctp = (sctp_t *)conn;
1504 	conn_t			*connp = sctp->sctp_connp;
1505 	int			family = connp->conn_family;
1506 	int			max = *addrcnt;
1507 	size_t			added = 0;
1508 	struct sockaddr_in6	*sin6;
1509 	struct sockaddr_in	*sin4;
1510 	int			scanned = 0;
1511 	boolean_t		skip_lback = B_FALSE;
1512 	ip_xmit_attr_t		*ixa = connp->conn_ixa;
1513 
1514 	if (sctp->sctp_nsaddrs == 0)
1515 		return (EINVAL);
1516 
1517 	/*
1518 	 * Skip loopback addresses for non-loopback assoc., ignore
1519 	 * this on a clustered node.
1520 	 */
1521 	if (sctp->sctp_state >= SCTPS_ESTABLISHED && !sctp->sctp_loopback &&
1522 	    (cl_sctp_check_addrs == NULL)) {
1523 		skip_lback = B_TRUE;
1524 	}
1525 
1526 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
1527 		if (sctp->sctp_saddrs[i].ipif_count == 0)
1528 			continue;
1529 		obj = list_head(&sctp->sctp_saddrs[i].sctp_ipif_list);
1530 		for (l = 0; l < sctp->sctp_saddrs[i].ipif_count; l++) {
1531 			sctp_ipif_t	*ipif = obj->saddr_ipifp;
1532 			in6_addr_t	addr = ipif->sctp_ipif_saddr;
1533 
1534 			scanned++;
1535 			if ((ipif->sctp_ipif_state == SCTP_IPIFS_CONDEMNED) ||
1536 			    SCTP_DONT_SRC(obj) ||
1537 			    (SCTP_IS_IPIF_LOOPBACK(ipif) && skip_lback)) {
1538 				if (scanned >= sctp->sctp_nsaddrs)
1539 					goto done;
1540 				obj = list_next(&sctp->sctp_saddrs[i].
1541 				    sctp_ipif_list, obj);
1542 				continue;
1543 			}
1544 			switch (family) {
1545 			case AF_INET:
1546 				sin4 = (struct sockaddr_in *)myaddrs + added;
1547 				sin4->sin_family = AF_INET;
1548 				sin4->sin_port = connp->conn_lport;
1549 				IN6_V4MAPPED_TO_INADDR(&addr, &sin4->sin_addr);
1550 				break;
1551 
1552 			case AF_INET6:
1553 				sin6 = (struct sockaddr_in6 *)myaddrs + added;
1554 				sin6->sin6_family = AF_INET6;
1555 				sin6->sin6_port = connp->conn_lport;
1556 				sin6->sin6_addr = addr;
1557 				/*
1558 				 * Note that flowinfo is only returned for
1559 				 * getpeername just like for TCP and UDP.
1560 				 */
1561 				sin6->sin6_flowinfo = 0;
1562 
1563 				if (IN6_IS_ADDR_LINKSCOPE(&sin6->sin6_addr) &&
1564 				    (ixa->ixa_flags & IXAF_SCOPEID_SET))
1565 					sin6->sin6_scope_id = ixa->ixa_scopeid;
1566 				else
1567 					sin6->sin6_scope_id = 0;
1568 				sin6->__sin6_src_id = 0;
1569 				break;
1570 			}
1571 			added++;
1572 			if (added >= max || scanned >= sctp->sctp_nsaddrs)
1573 				goto done;
1574 			obj = list_next(&sctp->sctp_saddrs[i].sctp_ipif_list,
1575 			    obj);
1576 		}
1577 	}
1578 done:
1579 	*addrcnt = added;
1580 	return (0);
1581 }
1582 
1583 /*
1584  * Given the supported address family, walk through the source address list
1585  * and return the total length of the available addresses. If 'p' is not
1586  * null, construct the parameter list for the addresses in 'p'.
1587  * 'modify' will only be set when we want the source address list to
1588  * be modified. The source address list will be modified only when
1589  * generating an INIT chunk. For generating an INIT-ACK 'modify' will
1590  * be false since the 'sctp' will be that of the listener.
1591  */
1592 size_t
1593 sctp_saddr_info(sctp_t *sctp, int supp_af, uchar_t *p, boolean_t modify)
1594 {
1595 	int			i;
1596 	int			l;
1597 	sctp_saddr_ipif_t	*obj;
1598 	size_t			paramlen = 0;
1599 	sctp_parm_hdr_t		*hdr;
1600 	int			scanned = 0;
1601 	int			naddr;
1602 	int			nsaddr;
1603 	boolean_t		del_ll = B_FALSE;
1604 	boolean_t		del_lb = B_FALSE;
1605 
1606 
1607 	/*
1608 	 * On a clustered node don't bother changing anything
1609 	 * on the loopback interface.
1610 	 */
1611 	if (modify && !sctp->sctp_loopback && (cl_sctp_check_addrs == NULL))
1612 		del_lb = B_TRUE;
1613 
1614 	if (modify && !sctp->sctp_linklocal)
1615 		del_ll = B_TRUE;
1616 
1617 	nsaddr = sctp->sctp_nsaddrs;
1618 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
1619 		if (sctp->sctp_saddrs[i].ipif_count == 0)
1620 			continue;
1621 		obj = list_head(&sctp->sctp_saddrs[i].sctp_ipif_list);
1622 		naddr = sctp->sctp_saddrs[i].ipif_count;
1623 		for (l = 0; l < naddr; l++) {
1624 			in6_addr_t	addr;
1625 			sctp_ipif_t	*ipif;
1626 			boolean_t	ipif_lb;
1627 			boolean_t	ipif_ll;
1628 			boolean_t	unsupp_af;
1629 
1630 			ipif = obj->saddr_ipifp;
1631 			scanned++;
1632 
1633 			ipif_lb = SCTP_IS_IPIF_LOOPBACK(ipif);
1634 			ipif_ll = SCTP_IS_IPIF_LINKLOCAL(ipif);
1635 			unsupp_af = SCTP_UNSUPP_AF(ipif, supp_af);
1636 			/*
1637 			 * We need to either delete or skip loopback/linklocal
1638 			 * or unsupported addresses, if required.
1639 			 */
1640 			if ((ipif_ll && del_ll) || (ipif_lb && del_lb) ||
1641 			    (unsupp_af && modify)) {
1642 				if (sctp->sctp_bound_to_all == 1)
1643 					sctp->sctp_bound_to_all = 0;
1644 				if (scanned < nsaddr) {
1645 					obj = list_next(&sctp->sctp_saddrs[i].
1646 					    sctp_ipif_list, obj);
1647 					sctp_ipif_hash_remove(sctp, ipif);
1648 					continue;
1649 				}
1650 				sctp_ipif_hash_remove(sctp, ipif);
1651 				goto next_addr;
1652 			} else if (ipif_ll || unsupp_af ||
1653 			    (ipif_lb && (cl_sctp_check_addrs == NULL))) {
1654 				goto next_addr;
1655 			}
1656 
1657 			if (!SCTP_IPIF_USABLE(ipif->sctp_ipif_state))
1658 				goto next_addr;
1659 			if (p != NULL)
1660 				hdr = (sctp_parm_hdr_t *)(p + paramlen);
1661 			addr = ipif->sctp_ipif_saddr;
1662 			if (!ipif->sctp_ipif_isv6) {
1663 				struct in_addr	*v4;
1664 
1665 				if (p != NULL) {
1666 					hdr->sph_type = htons(PARM_ADDR4);
1667 					hdr->sph_len = htons(PARM_ADDR4_LEN);
1668 					v4 = (struct in_addr *)(hdr + 1);
1669 					IN6_V4MAPPED_TO_INADDR(&addr, v4);
1670 				}
1671 				paramlen += PARM_ADDR4_LEN;
1672 			} else {
1673 				if (p != NULL) {
1674 					hdr->sph_type = htons(PARM_ADDR6);
1675 					hdr->sph_len = htons(PARM_ADDR6_LEN);
1676 					bcopy(&addr, hdr + 1, sizeof (addr));
1677 				}
1678 				paramlen += PARM_ADDR6_LEN;
1679 			}
1680 next_addr:
1681 			if (scanned >= nsaddr)
1682 				return (paramlen);
1683 			obj = list_next(&sctp->sctp_saddrs[i].sctp_ipif_list,
1684 			    obj);
1685 		}
1686 	}
1687 	return (paramlen);
1688 }
1689 
1690 /*
1691  * This is used on a clustered node to obtain a list of addresses, the list
1692  * consists of sockaddr_in structs for v4 and sockaddr_in6 for v6. The list
1693  * is then passed onto the clustering module which sends back the correct
1694  * list based on the port info. Regardless of the input, i.e INADDR_ANY
1695  * or specific address(es), we create the list since it could be modified by
1696  * the clustering module. When given a list of addresses, we simply
1697  * create the list of sockaddr_in or sockaddr_in6 structs using those
1698  * addresses. If there is an INADDR_ANY in the input list, or if the
1699  * input is INADDR_ANY, we create a list of sockaddr_in or sockaddr_in6
1700  * structs consisting all the addresses in the global interface list
1701  * except those that are hosted on the loopback interface. We create
1702  * a list of sockaddr_in[6] structs just so that it can be directly input
1703  * to sctp_valid_addr_list() once the clustering module has processed it.
1704  */
1705 int
1706 sctp_get_addrlist(sctp_t *sctp, const void *addrs, uint32_t *addrcnt,
1707     uchar_t **addrlist, int *uspec, size_t *size)
1708 {
1709 	int			cnt;
1710 	int			icnt;
1711 	sctp_ipif_t		*sctp_ipif;
1712 	struct sockaddr_in	*s4;
1713 	struct sockaddr_in6	*s6;
1714 	uchar_t			*p;
1715 	int			err = 0;
1716 	sctp_stack_t		*sctps = sctp->sctp_sctps;
1717 	conn_t			*connp = sctp->sctp_connp;
1718 
1719 	*addrlist = NULL;
1720 	*size = 0;
1721 
1722 	/*
1723 	 * Create a list of sockaddr_in[6] structs using the input list.
1724 	 */
1725 	if (connp->conn_family == AF_INET) {
1726 		*size = sizeof (struct sockaddr_in) * *addrcnt;
1727 		*addrlist = kmem_zalloc(*size,  KM_SLEEP);
1728 		p = *addrlist;
1729 		for (cnt = 0; cnt < *addrcnt; cnt++) {
1730 			s4 = (struct sockaddr_in *)addrs + cnt;
1731 			/*
1732 			 * We need to create a list of all the available
1733 			 * addresses if there is an INADDR_ANY. However,
1734 			 * if we are beyond LISTEN, then this is invalid
1735 			 * (see sctp_valid_addr_list(). So, we just fail
1736 			 * it here rather than wait till it fails in
1737 			 * sctp_valid_addr_list().
1738 			 */
1739 			if (s4->sin_addr.s_addr == INADDR_ANY) {
1740 				kmem_free(*addrlist, *size);
1741 				*addrlist = NULL;
1742 				*size = 0;
1743 				if (sctp->sctp_state > SCTPS_LISTEN) {
1744 					*addrcnt = 0;
1745 					return (EINVAL);
1746 				}
1747 				if (uspec != NULL)
1748 					*uspec = 1;
1749 				goto get_all_addrs;
1750 			} else {
1751 				bcopy(s4, p, sizeof (*s4));
1752 				p += sizeof (*s4);
1753 			}
1754 		}
1755 	} else {
1756 		*size = sizeof (struct sockaddr_in6) * *addrcnt;
1757 		*addrlist = kmem_zalloc(*size, KM_SLEEP);
1758 		p = *addrlist;
1759 		for (cnt = 0; cnt < *addrcnt; cnt++) {
1760 			s6 = (struct sockaddr_in6 *)addrs + cnt;
1761 			/*
1762 			 * Comments for INADDR_ANY, above, apply here too.
1763 			 */
1764 			if (IN6_IS_ADDR_UNSPECIFIED(&s6->sin6_addr)) {
1765 				kmem_free(*addrlist, *size);
1766 				*size = 0;
1767 				*addrlist = NULL;
1768 				if (sctp->sctp_state > SCTPS_LISTEN) {
1769 					*addrcnt = 0;
1770 					return (EINVAL);
1771 				}
1772 				if (uspec != NULL)
1773 					*uspec = 1;
1774 				goto get_all_addrs;
1775 			} else {
1776 				bcopy(addrs, p, sizeof (*s6));
1777 				p += sizeof (*s6);
1778 			}
1779 		}
1780 	}
1781 	return (err);
1782 get_all_addrs:
1783 
1784 	/*
1785 	 * Allocate max possible size. We allocate the max. size here because
1786 	 * the clustering module could end up adding addresses to the list.
1787 	 * We allocate upfront so that the clustering module need to bother
1788 	 * re-sizing the list.
1789 	 */
1790 	if (connp->conn_family == AF_INET) {
1791 		*size = sizeof (struct sockaddr_in) *
1792 		    sctps->sctps_g_ipifs_count;
1793 	} else {
1794 		*size = sizeof (struct sockaddr_in6) *
1795 		    sctps->sctps_g_ipifs_count;
1796 	}
1797 	*addrlist = kmem_zalloc(*size, KM_SLEEP);
1798 	*addrcnt = 0;
1799 	p = *addrlist;
1800 	rw_enter(&sctps->sctps_g_ipifs_lock, RW_READER);
1801 
1802 	/*
1803 	 * Walk through the global interface list and add all addresses,
1804 	 * except those that are hosted on loopback interfaces.
1805 	 */
1806 	for (cnt = 0; cnt <  SCTP_IPIF_HASH; cnt++) {
1807 		if (sctps->sctps_g_ipifs[cnt].ipif_count == 0)
1808 			continue;
1809 		sctp_ipif = list_head(
1810 		    &sctps->sctps_g_ipifs[cnt].sctp_ipif_list);
1811 		for (icnt = 0;
1812 		    icnt < sctps->sctps_g_ipifs[cnt].ipif_count;
1813 		    icnt++) {
1814 			in6_addr_t	addr;
1815 
1816 			rw_enter(&sctp_ipif->sctp_ipif_lock, RW_READER);
1817 			addr = sctp_ipif->sctp_ipif_saddr;
1818 			if (SCTP_IPIF_DISCARD(sctp_ipif->sctp_ipif_flags) ||
1819 			    !SCTP_IPIF_USABLE(sctp_ipif->sctp_ipif_state) ||
1820 			    SCTP_IS_IPIF_LOOPBACK(sctp_ipif) ||
1821 			    SCTP_IS_IPIF_LINKLOCAL(sctp_ipif) ||
1822 			    !SCTP_IPIF_ZONE_MATCH(sctp, sctp_ipif) ||
1823 			    (connp->conn_family == AF_INET &&
1824 			    sctp_ipif->sctp_ipif_isv6) ||
1825 			    (sctp->sctp_connp->conn_ipv6_v6only &&
1826 			    !sctp_ipif->sctp_ipif_isv6)) {
1827 				rw_exit(&sctp_ipif->sctp_ipif_lock);
1828 				sctp_ipif = list_next(
1829 				    &sctps->sctps_g_ipifs[cnt].sctp_ipif_list,
1830 				    sctp_ipif);
1831 				continue;
1832 			}
1833 			rw_exit(&sctp_ipif->sctp_ipif_lock);
1834 			if (connp->conn_family == AF_INET) {
1835 				s4 = (struct sockaddr_in *)p;
1836 				IN6_V4MAPPED_TO_INADDR(&addr, &s4->sin_addr);
1837 				s4->sin_family = AF_INET;
1838 				p += sizeof (*s4);
1839 			} else {
1840 				s6 = (struct sockaddr_in6 *)p;
1841 				s6->sin6_addr = addr;
1842 				s6->sin6_family = AF_INET6;
1843 				s6->sin6_scope_id =
1844 				    sctp_ipif->sctp_ipif_ill->sctp_ill_index;
1845 				p += sizeof (*s6);
1846 			}
1847 			(*addrcnt)++;
1848 			sctp_ipif = list_next(
1849 			    &sctps->sctps_g_ipifs[cnt].sctp_ipif_list,
1850 			    sctp_ipif);
1851 		}
1852 	}
1853 	rw_exit(&sctps->sctps_g_ipifs_lock);
1854 	return (err);
1855 }
1856 
1857 /*
1858  * Get a list of addresses from the source address list. The  caller is
1859  * responsible for allocating sufficient buffer for this.
1860  */
1861 void
1862 sctp_get_saddr_list(sctp_t *sctp, uchar_t *p, size_t psize)
1863 {
1864 	int			cnt;
1865 	int			icnt;
1866 	sctp_saddr_ipif_t	*obj;
1867 	int			naddr;
1868 	int			scanned = 0;
1869 
1870 	for (cnt = 0; cnt < SCTP_IPIF_HASH; cnt++) {
1871 		if (sctp->sctp_saddrs[cnt].ipif_count == 0)
1872 			continue;
1873 		obj = list_head(&sctp->sctp_saddrs[cnt].sctp_ipif_list);
1874 		naddr = sctp->sctp_saddrs[cnt].ipif_count;
1875 		for (icnt = 0; icnt < naddr; icnt++) {
1876 			sctp_ipif_t	*ipif;
1877 
1878 			if (psize < sizeof (ipif->sctp_ipif_saddr))
1879 				return;
1880 
1881 			scanned++;
1882 			ipif = obj->saddr_ipifp;
1883 			bcopy(&ipif->sctp_ipif_saddr, p,
1884 			    sizeof (ipif->sctp_ipif_saddr));
1885 			p += sizeof (ipif->sctp_ipif_saddr);
1886 			psize -= sizeof (ipif->sctp_ipif_saddr);
1887 			if (scanned >= sctp->sctp_nsaddrs)
1888 				return;
1889 			obj = list_next(
1890 			    &sctp->sctp_saddrs[icnt].sctp_ipif_list,
1891 			    obj);
1892 		}
1893 	}
1894 }
1895 
1896 /*
1897  * Get a list of addresses from the remote address list. The  caller is
1898  * responsible for allocating sufficient buffer for this.
1899  */
1900 void
1901 sctp_get_faddr_list(sctp_t *sctp, uchar_t *p, size_t psize)
1902 {
1903 	sctp_faddr_t	*fp;
1904 
1905 	for (fp = sctp->sctp_faddrs; fp != NULL; fp = fp->next) {
1906 		if (psize < sizeof (fp->faddr))
1907 			return;
1908 		bcopy(&fp->faddr, p, sizeof (fp->faddr));
1909 		p += sizeof (fp->faddr);
1910 		psize -= sizeof (fp->faddr);
1911 	}
1912 }
1913 
1914 static void
1915 sctp_free_ills(sctp_stack_t *sctps)
1916 {
1917 	int			i;
1918 	int			l;
1919 	sctp_ill_t	*sctp_ill;
1920 
1921 	if (sctps->sctps_ills_count == 0)
1922 		return;
1923 
1924 	for (i = 0; i < SCTP_ILL_HASH; i++) {
1925 		sctp_ill = list_tail(&sctps->sctps_g_ills[i].sctp_ill_list);
1926 		for (l = 0; l < sctps->sctps_g_ills[i].ill_count; l++) {
1927 			ASSERT(sctp_ill->sctp_ill_ipifcnt == 0);
1928 			list_remove(&sctps->sctps_g_ills[i].sctp_ill_list,
1929 			    sctp_ill);
1930 			sctps->sctps_ills_count--;
1931 			kmem_free(sctp_ill->sctp_ill_name,
1932 			    sctp_ill->sctp_ill_name_length);
1933 			kmem_free(sctp_ill, sizeof (sctp_ill_t));
1934 			sctp_ill =
1935 			    list_tail(&sctps->sctps_g_ills[i].sctp_ill_list);
1936 		}
1937 		sctps->sctps_g_ills[i].ill_count = 0;
1938 	}
1939 	ASSERT(sctps->sctps_ills_count == 0);
1940 }
1941 
1942 static void
1943 sctp_free_ipifs(sctp_stack_t *sctps)
1944 {
1945 	int			i;
1946 	int			l;
1947 	sctp_ipif_t	*sctp_ipif;
1948 	sctp_ill_t	*sctp_ill;
1949 
1950 	if (sctps->sctps_g_ipifs_count == 0)
1951 		return;
1952 
1953 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
1954 		sctp_ipif = list_tail(&sctps->sctps_g_ipifs[i].sctp_ipif_list);
1955 		for (l = 0; l < sctps->sctps_g_ipifs[i].ipif_count; l++) {
1956 			sctp_ill = sctp_ipif->sctp_ipif_ill;
1957 
1958 			list_remove(&sctps->sctps_g_ipifs[i].sctp_ipif_list,
1959 			    sctp_ipif);
1960 			sctps->sctps_g_ipifs_count--;
1961 			(void) atomic_add_32_nv(&sctp_ill->sctp_ill_ipifcnt,
1962 			    -1);
1963 			kmem_free(sctp_ipif, sizeof (sctp_ipif_t));
1964 			sctp_ipif =
1965 			    list_tail(&sctps->sctps_g_ipifs[i].sctp_ipif_list);
1966 		}
1967 		sctps->sctps_g_ipifs[i].ipif_count = 0;
1968 	}
1969 	ASSERT(sctps->sctps_g_ipifs_count == 0);
1970 }
1971 
1972 
1973 /* Initialize the SCTP ILL list and lock */
1974 void
1975 sctp_saddr_init(sctp_stack_t *sctps)
1976 {
1977 	int	i;
1978 
1979 	sctps->sctps_g_ills = kmem_zalloc(sizeof (sctp_ill_hash_t) *
1980 	    SCTP_ILL_HASH, KM_SLEEP);
1981 	sctps->sctps_g_ipifs = kmem_zalloc(sizeof (sctp_ipif_hash_t) *
1982 	    SCTP_IPIF_HASH, KM_SLEEP);
1983 
1984 	rw_init(&sctps->sctps_g_ills_lock, NULL, RW_DEFAULT, NULL);
1985 	rw_init(&sctps->sctps_g_ipifs_lock, NULL, RW_DEFAULT, NULL);
1986 
1987 	for (i = 0; i < SCTP_ILL_HASH; i++) {
1988 		sctps->sctps_g_ills[i].ill_count = 0;
1989 		list_create(&sctps->sctps_g_ills[i].sctp_ill_list,
1990 		    sizeof (sctp_ill_t),
1991 		    offsetof(sctp_ill_t, sctp_ills));
1992 	}
1993 	for (i = 0; i < SCTP_IPIF_HASH; i++) {
1994 		sctps->sctps_g_ipifs[i].ipif_count = 0;
1995 		list_create(&sctps->sctps_g_ipifs[i].sctp_ipif_list,
1996 		    sizeof (sctp_ipif_t), offsetof(sctp_ipif_t, sctp_ipifs));
1997 	}
1998 }
1999 
2000 void
2001 sctp_saddr_fini(sctp_stack_t *sctps)
2002 {
2003 	int	i;
2004 
2005 	sctp_free_ipifs(sctps);
2006 	sctp_free_ills(sctps);
2007 
2008 	for (i = 0; i < SCTP_ILL_HASH; i++)
2009 		list_destroy(&sctps->sctps_g_ills[i].sctp_ill_list);
2010 	for (i = 0; i < SCTP_IPIF_HASH; i++)
2011 		list_destroy(&sctps->sctps_g_ipifs[i].sctp_ipif_list);
2012 
2013 	ASSERT(sctps->sctps_ills_count == 0 && sctps->sctps_g_ipifs_count == 0);
2014 	kmem_free(sctps->sctps_g_ills, sizeof (sctp_ill_hash_t) *
2015 	    SCTP_ILL_HASH);
2016 	sctps->sctps_g_ills = NULL;
2017 	kmem_free(sctps->sctps_g_ipifs, sizeof (sctp_ipif_hash_t) *
2018 	    SCTP_IPIF_HASH);
2019 	sctps->sctps_g_ipifs = NULL;
2020 	rw_destroy(&sctps->sctps_g_ills_lock);
2021 	rw_destroy(&sctps->sctps_g_ipifs_lock);
2022 }
2023