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 (c) 2004, 2010, Oracle and/or its affiliates. All rights reserved.
23 */
24
25 #include <sys/types.h>
26 #include <sys/systm.h>
27 #include <sys/stream.h>
28 #include <sys/cmn_err.h>
29 #include <sys/ddi.h>
30 #include <sys/sunddi.h>
31 #include <sys/kmem.h>
32 #include <sys/socket.h>
33 #include <sys/sysmacros.h>
34 #include <sys/list.h>
35
36 #include <netinet/in.h>
37 #include <netinet/ip6.h>
38 #include <netinet/sctp.h>
39
40 #include <inet/common.h>
41 #include <inet/ip.h>
42 #include <inet/ip6.h>
43 #include <inet/ip_ire.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 boolean_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_dec_32_nv(&sctp_ill->sctp_ill_ipifcnt);
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 rw_enter(&sctp->sctp_saddrs[hindex].ipif_hash_lock, RW_WRITER);
465 ipif_obj = list_head(&sctp->sctp_saddrs[hindex].sctp_ipif_list);
466 for (cnt = 0; cnt < sctp->sctp_saddrs[hindex].ipif_count; cnt++) {
467 if (IN6_ARE_ADDR_EQUAL(&ipif_obj->saddr_ipifp->sctp_ipif_saddr,
468 &ipif->sctp_ipif_saddr)) {
469 if (ipif->sctp_ipif_id !=
470 ipif_obj->saddr_ipifp->sctp_ipif_id &&
471 ipif_obj->saddr_ipifp->sctp_ipif_state ==
472 SCTP_IPIFS_DOWN && ipif->sctp_ipif_state ==
473 SCTP_IPIFS_UP) {
474 SCTP_IPIF_REFRELE(ipif_obj->saddr_ipifp);
475 ipif_obj->saddr_ipifp = ipif;
476 ipif_obj->saddr_ipif_dontsrc = dontsrc ? 1 : 0;
477 rw_exit(
478 &sctp->sctp_saddrs[hindex].ipif_hash_lock);
479 return (0);
480 } else if (!allow_dup || ipif->sctp_ipif_id ==
481 ipif_obj->saddr_ipifp->sctp_ipif_id) {
482 rw_exit(
483 &sctp->sctp_saddrs[hindex].ipif_hash_lock);
484 return (EALREADY);
485 }
486 }
487 ipif_obj = list_next(&sctp->sctp_saddrs[hindex].sctp_ipif_list,
488 ipif_obj);
489 }
490 ipif_obj = kmem_zalloc(sizeof (sctp_saddr_ipif_t), sleep);
491 if (ipif_obj == NULL) {
492 rw_exit(&sctp->sctp_saddrs[hindex].ipif_hash_lock);
493 /* Need to do something */
494 return (ENOMEM);
495 }
496 ipif_obj->saddr_ipifp = ipif;
497 ipif_obj->saddr_ipif_dontsrc = dontsrc ? 1 : 0;
498 list_insert_tail(&sctp->sctp_saddrs[hindex].sctp_ipif_list, ipif_obj);
499 sctp->sctp_saddrs[hindex].ipif_count++;
500 sctp->sctp_nsaddrs++;
501 rw_exit(&sctp->sctp_saddrs[hindex].ipif_hash_lock);
502 return (0);
503 }
504
505 /*
506 * Given a source address, walk through the peer address list to see
507 * if the source address is being used. If it is, reset that.
508 * A cleared saddr will then make sctp_make_mp lookup the destination again
509 * and as part of that look for a new source.
510 */
511 static void
512 sctp_fix_saddr(sctp_t *sctp, in6_addr_t *saddr)
513 {
514 sctp_faddr_t *fp;
515
516 for (fp = sctp->sctp_faddrs; fp != NULL; fp = fp->sf_next) {
517 if (!IN6_ARE_ADDR_EQUAL(&fp->sf_saddr, saddr))
518 continue;
519 V6_SET_ZERO(fp->sf_saddr);
520 }
521 }
522
523 static void
524 sctp_ipif_hash_remove(sctp_t *sctp, sctp_ipif_t *ipif, boolean_t locked)
525 {
526 int cnt;
527 sctp_saddr_ipif_t *ipif_obj;
528 int hindex;
529
530 hindex = SCTP_IPIF_ADDR_HASH(ipif->sctp_ipif_saddr,
531 ipif->sctp_ipif_isv6);
532 if (!locked)
533 rw_enter(&sctp->sctp_saddrs[hindex].ipif_hash_lock, RW_WRITER);
534 ipif_obj = list_head(&sctp->sctp_saddrs[hindex].sctp_ipif_list);
535 for (cnt = 0; cnt < sctp->sctp_saddrs[hindex].ipif_count; cnt++) {
536 if (IN6_ARE_ADDR_EQUAL(&ipif_obj->saddr_ipifp->sctp_ipif_saddr,
537 &ipif->sctp_ipif_saddr)) {
538 list_remove(&sctp->sctp_saddrs[hindex].sctp_ipif_list,
539 ipif_obj);
540 sctp->sctp_saddrs[hindex].ipif_count--;
541 sctp->sctp_nsaddrs--;
542 sctp_fix_saddr(sctp, &ipif->sctp_ipif_saddr);
543 SCTP_IPIF_REFRELE(ipif_obj->saddr_ipifp);
544 kmem_free(ipif_obj, sizeof (sctp_saddr_ipif_t));
545 break;
546 }
547 ipif_obj = list_next(&sctp->sctp_saddrs[hindex].sctp_ipif_list,
548 ipif_obj);
549 }
550 if (!locked)
551 rw_exit(&sctp->sctp_saddrs[hindex].ipif_hash_lock);
552 }
553
554 static int
555 sctp_compare_ipif_list(sctp_ipif_hash_t *list1, sctp_ipif_hash_t *list2)
556 {
557 int i;
558 int j;
559 sctp_saddr_ipif_t *obj1;
560 sctp_saddr_ipif_t *obj2;
561 int overlap = 0;
562
563 rw_enter(&list1->ipif_hash_lock, RW_READER);
564 rw_enter(&list2->ipif_hash_lock, RW_READER);
565 obj1 = list_head(&list1->sctp_ipif_list);
566 for (i = 0; i < list1->ipif_count; i++) {
567 obj2 = list_head(&list2->sctp_ipif_list);
568 for (j = 0; j < list2->ipif_count; j++) {
569 if (IN6_ARE_ADDR_EQUAL(
570 &obj1->saddr_ipifp->sctp_ipif_saddr,
571 &obj2->saddr_ipifp->sctp_ipif_saddr)) {
572 overlap++;
573 break;
574 }
575 obj2 = list_next(&list2->sctp_ipif_list,
576 obj2);
577 }
578 obj1 = list_next(&list1->sctp_ipif_list, obj1);
579 }
580 rw_exit(&list1->ipif_hash_lock);
581 rw_exit(&list2->ipif_hash_lock);
582 return (overlap);
583 }
584
585 int
586 sctp_compare_saddrs(sctp_t *sctp1, sctp_t *sctp2)
587 {
588 int i;
589 int overlap = 0;
590
591 for (i = 0; i < SCTP_IPIF_HASH; i++) {
592 overlap += sctp_compare_ipif_list(&sctp1->sctp_saddrs[i],
593 &sctp2->sctp_saddrs[i]);
594 }
595
596 if (sctp1->sctp_nsaddrs == sctp2->sctp_nsaddrs &&
597 overlap == sctp1->sctp_nsaddrs) {
598 return (SCTP_ADDR_EQUAL);
599 }
600
601 if (overlap == sctp1->sctp_nsaddrs)
602 return (SCTP_ADDR_SUBSET);
603
604 if (overlap > 0)
605 return (SCTP_ADDR_OVERLAP);
606
607 return (SCTP_ADDR_DISJOINT);
608 }
609
610 static int
611 sctp_copy_ipifs(sctp_ipif_hash_t *list1, sctp_t *sctp2, int sleep)
612 {
613 int i;
614 sctp_saddr_ipif_t *obj;
615 int error = 0;
616
617 rw_enter(&list1->ipif_hash_lock, RW_READER);
618 obj = list_head(&list1->sctp_ipif_list);
619 for (i = 0; i < list1->ipif_count; i++) {
620 SCTP_IPIF_REFHOLD(obj->saddr_ipifp);
621 error = sctp_ipif_hash_insert(sctp2, obj->saddr_ipifp, sleep,
622 B_FALSE, B_FALSE);
623 ASSERT(error != EALREADY);
624 if (error != 0) {
625 rw_exit(&list1->ipif_hash_lock);
626 return (error);
627 }
628 obj = list_next(&list1->sctp_ipif_list, obj);
629 }
630 rw_exit(&list1->ipif_hash_lock);
631 return (error);
632 }
633
634 int
635 sctp_dup_saddrs(sctp_t *sctp1, sctp_t *sctp2, int sleep)
636 {
637 int error = 0;
638 int i;
639
640 if (sctp1 == NULL || sctp1->sctp_bound_to_all == 1)
641 return (sctp_get_all_ipifs(sctp2, sleep));
642
643 for (i = 0; i < SCTP_IPIF_HASH; i++) {
644 rw_enter(&sctp1->sctp_saddrs[i].ipif_hash_lock, RW_READER);
645 if (sctp1->sctp_saddrs[i].ipif_count == 0) {
646 rw_exit(&sctp1->sctp_saddrs[i].ipif_hash_lock);
647 continue;
648 }
649 error = sctp_copy_ipifs(&sctp1->sctp_saddrs[i], sctp2, sleep);
650 if (error != 0) {
651 rw_exit(&sctp1->sctp_saddrs[i].ipif_hash_lock);
652 sctp_free_saddrs(sctp2);
653 return (error);
654 }
655 rw_exit(&sctp1->sctp_saddrs[i].ipif_hash_lock);
656 }
657 return (0);
658 }
659
660 void
661 sctp_free_saddrs(sctp_t *sctp)
662 {
663 int i;
664 int l;
665 sctp_saddr_ipif_t *obj;
666
667 if (sctp->sctp_nsaddrs == 0)
668 return;
669 for (i = 0; i < SCTP_IPIF_HASH; i++) {
670 rw_enter(&sctp->sctp_saddrs[i].ipif_hash_lock, RW_WRITER);
671 if (sctp->sctp_saddrs[i].ipif_count == 0) {
672 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
673 continue;
674 }
675 obj = list_tail(&sctp->sctp_saddrs[i].sctp_ipif_list);
676 for (l = 0; l < sctp->sctp_saddrs[i].ipif_count; l++) {
677 list_remove(&sctp->sctp_saddrs[i].sctp_ipif_list, obj);
678 SCTP_IPIF_REFRELE(obj->saddr_ipifp);
679 sctp->sctp_nsaddrs--;
680 kmem_free(obj, sizeof (sctp_saddr_ipif_t));
681 obj = list_tail(&sctp->sctp_saddrs[i].sctp_ipif_list);
682 }
683 sctp->sctp_saddrs[i].ipif_count = 0;
684 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
685 }
686 if (sctp->sctp_bound_to_all == 1)
687 sctp->sctp_bound_to_all = 0;
688 ASSERT(sctp->sctp_nsaddrs == 0);
689 }
690
691 /*
692 * Add/Delete the given ILL from the SCTP ILL list. Called with no locks
693 * held.
694 */
695 void
696 sctp_update_ill(ill_t *ill, int op)
697 {
698 int i;
699 sctp_ill_t *sctp_ill = NULL;
700 uint_t index;
701 netstack_t *ns = ill->ill_ipst->ips_netstack;
702 sctp_stack_t *sctps = ns->netstack_sctp;
703
704 rw_enter(&sctps->sctps_g_ills_lock, RW_WRITER);
705
706 index = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(ill));
707 sctp_ill = list_head(&sctps->sctps_g_ills[index].sctp_ill_list);
708 for (i = 0; i < sctps->sctps_g_ills[index].ill_count; i++) {
709 if ((sctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(ill)) &&
710 (sctp_ill->sctp_ill_isv6 == ill->ill_isv6)) {
711 break;
712 }
713 sctp_ill = list_next(&sctps->sctps_g_ills[index].sctp_ill_list,
714 sctp_ill);
715 }
716
717 switch (op) {
718 case SCTP_ILL_INSERT:
719 if (sctp_ill != NULL) {
720 /* Unmark it if it is condemned */
721 if (sctp_ill->sctp_ill_state == SCTP_ILLS_CONDEMNED)
722 sctp_ill->sctp_ill_state = 0;
723 rw_exit(&sctps->sctps_g_ills_lock);
724 return;
725 }
726 sctp_ill = kmem_zalloc(sizeof (sctp_ill_t), KM_NOSLEEP);
727 /* Need to re-try? */
728 if (sctp_ill == NULL) {
729 cmn_err(CE_WARN, "sctp_update_ill: error adding "
730 "ILL %p to SCTP's ILL list", (void *)ill);
731 rw_exit(&sctps->sctps_g_ills_lock);
732 return;
733 }
734 sctp_ill->sctp_ill_name = kmem_zalloc(ill->ill_name_length,
735 KM_NOSLEEP);
736 if (sctp_ill->sctp_ill_name == NULL) {
737 cmn_err(CE_WARN, "sctp_update_ill: error adding "
738 "ILL %p to SCTP's ILL list", (void *)ill);
739 kmem_free(sctp_ill, sizeof (sctp_ill_t));
740 rw_exit(&sctps->sctps_g_ills_lock);
741 return;
742 }
743 bcopy(ill->ill_name, sctp_ill->sctp_ill_name,
744 ill->ill_name_length);
745 sctp_ill->sctp_ill_name_length = ill->ill_name_length;
746 sctp_ill->sctp_ill_index = SCTP_ILL_TO_PHYINDEX(ill);
747 sctp_ill->sctp_ill_flags = ill->ill_phyint->phyint_flags;
748 sctp_ill->sctp_ill_netstack = ns; /* No netstack_hold */
749 sctp_ill->sctp_ill_isv6 = ill->ill_isv6;
750 list_insert_tail(&sctps->sctps_g_ills[index].sctp_ill_list,
751 (void *)sctp_ill);
752 sctps->sctps_g_ills[index].ill_count++;
753 sctps->sctps_ills_count++;
754
755 break;
756
757 case SCTP_ILL_REMOVE:
758
759 if (sctp_ill == NULL) {
760 rw_exit(&sctps->sctps_g_ills_lock);
761 return;
762 }
763 if (sctp_ill->sctp_ill_ipifcnt == 0) {
764 list_remove(&sctps->sctps_g_ills[index].sctp_ill_list,
765 (void *)sctp_ill);
766 sctps->sctps_g_ills[index].ill_count--;
767 sctps->sctps_ills_count--;
768 kmem_free(sctp_ill->sctp_ill_name,
769 ill->ill_name_length);
770 kmem_free(sctp_ill, sizeof (sctp_ill_t));
771 } else {
772 sctp_ill->sctp_ill_state = SCTP_ILLS_CONDEMNED;
773 }
774
775 break;
776 }
777 rw_exit(&sctps->sctps_g_ills_lock);
778 }
779
780 /*
781 * The ILL's index is being changed, just remove it from the old list,
782 * change the SCTP ILL's index and re-insert using the new index.
783 */
784 void
785 sctp_ill_reindex(ill_t *ill, uint_t orig_ill_index)
786 {
787 sctp_ill_t *sctp_ill = NULL;
788 sctp_ill_t *nxt_sill;
789 uint_t indx;
790 uint_t nindx;
791 boolean_t once = B_FALSE;
792 netstack_t *ns = ill->ill_ipst->ips_netstack;
793 sctp_stack_t *sctps = ns->netstack_sctp;
794
795 rw_enter(&sctps->sctps_g_ills_lock, RW_WRITER);
796
797 indx = SCTP_ILL_HASH_FN(orig_ill_index);
798 nindx = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(ill));
799 sctp_ill = list_head(&sctps->sctps_g_ills[indx].sctp_ill_list);
800 while (sctp_ill != NULL) {
801 nxt_sill = list_next(&sctps->sctps_g_ills[indx].sctp_ill_list,
802 sctp_ill);
803 if (sctp_ill->sctp_ill_index == orig_ill_index) {
804 sctp_ill->sctp_ill_index = SCTP_ILL_TO_PHYINDEX(ill);
805 /*
806 * if the new index hashes to the same value, all's
807 * done.
808 */
809 if (nindx != indx) {
810 list_remove(
811 &sctps->sctps_g_ills[indx].sctp_ill_list,
812 (void *)sctp_ill);
813 sctps->sctps_g_ills[indx].ill_count--;
814 list_insert_tail(
815 &sctps->sctps_g_ills[nindx].sctp_ill_list,
816 (void *)sctp_ill);
817 sctps->sctps_g_ills[nindx].ill_count++;
818 }
819 if (once)
820 break;
821 /* We might have one for v4 and for v6 */
822 once = B_TRUE;
823 }
824 sctp_ill = nxt_sill;
825 }
826 rw_exit(&sctps->sctps_g_ills_lock);
827 }
828
829 /* move ipif from f_ill to t_ill */
830 void
831 sctp_move_ipif(ipif_t *ipif, ill_t *f_ill, ill_t *t_ill)
832 {
833 sctp_ill_t *fsctp_ill = NULL;
834 sctp_ill_t *tsctp_ill = NULL;
835 sctp_ipif_t *sctp_ipif;
836 uint_t hindex;
837 int i;
838 netstack_t *ns = ipif->ipif_ill->ill_ipst->ips_netstack;
839 sctp_stack_t *sctps = ns->netstack_sctp;
840
841 rw_enter(&sctps->sctps_g_ills_lock, RW_READER);
842 rw_enter(&sctps->sctps_g_ipifs_lock, RW_READER);
843
844 hindex = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(f_ill));
845 fsctp_ill = list_head(&sctps->sctps_g_ills[hindex].sctp_ill_list);
846 for (i = 0; i < sctps->sctps_g_ills[hindex].ill_count; i++) {
847 if (fsctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(f_ill) &&
848 fsctp_ill->sctp_ill_isv6 == f_ill->ill_isv6) {
849 break;
850 }
851 fsctp_ill = list_next(
852 &sctps->sctps_g_ills[hindex].sctp_ill_list, fsctp_ill);
853 }
854
855 hindex = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(t_ill));
856 tsctp_ill = list_head(&sctps->sctps_g_ills[hindex].sctp_ill_list);
857 for (i = 0; i < sctps->sctps_g_ills[hindex].ill_count; i++) {
858 if (tsctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(t_ill) &&
859 tsctp_ill->sctp_ill_isv6 == t_ill->ill_isv6) {
860 break;
861 }
862 tsctp_ill = list_next(
863 &sctps->sctps_g_ills[hindex].sctp_ill_list, tsctp_ill);
864 }
865
866 hindex = SCTP_IPIF_ADDR_HASH(ipif->ipif_v6lcl_addr,
867 ipif->ipif_ill->ill_isv6);
868 sctp_ipif = list_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list);
869 for (i = 0; i < sctps->sctps_g_ipifs[hindex].ipif_count; i++) {
870 if (sctp_ipif->sctp_ipif_id == ipif->ipif_seqid)
871 break;
872 sctp_ipif = list_next(
873 &sctps->sctps_g_ipifs[hindex].sctp_ipif_list, sctp_ipif);
874 }
875 /* Should be an ASSERT? */
876 if (fsctp_ill == NULL || tsctp_ill == NULL || sctp_ipif == NULL) {
877 ip1dbg(("sctp_move_ipif: error moving ipif %p from %p to %p\n",
878 (void *)ipif, (void *)f_ill, (void *)t_ill));
879 rw_exit(&sctps->sctps_g_ipifs_lock);
880 rw_exit(&sctps->sctps_g_ills_lock);
881 return;
882 }
883 rw_enter(&sctp_ipif->sctp_ipif_lock, RW_WRITER);
884 ASSERT(sctp_ipif->sctp_ipif_ill == fsctp_ill);
885 sctp_ipif->sctp_ipif_ill = tsctp_ill;
886 rw_exit(&sctp_ipif->sctp_ipif_lock);
887 (void) atomic_dec_32_nv(&fsctp_ill->sctp_ill_ipifcnt);
888 atomic_inc_32(&tsctp_ill->sctp_ill_ipifcnt);
889 rw_exit(&sctps->sctps_g_ipifs_lock);
890 rw_exit(&sctps->sctps_g_ills_lock);
891 }
892
893 /*
894 * Walk the list of SCTPs and find each that has oipif in it's saddr list, and
895 * if so replace it with nipif.
896 */
897 void
898 sctp_update_saddrs(sctp_ipif_t *oipif, sctp_ipif_t *nipif, int idx,
899 sctp_stack_t *sctps)
900 {
901 sctp_t *sctp;
902 sctp_t *sctp_prev = NULL;
903 sctp_saddr_ipif_t *sobj;
904 int count;
905
906 mutex_enter(&sctps->sctps_g_lock);
907 sctp = list_head(&sctps->sctps_g_list);
908 while (sctp != NULL && oipif->sctp_ipif_refcnt > 0) {
909 mutex_enter(&sctp->sctp_reflock);
910 if (sctp->sctp_condemned ||
911 sctp->sctp_saddrs[idx].ipif_count <= 0) {
912 mutex_exit(&sctp->sctp_reflock);
913 sctp = list_next(&sctps->sctps_g_list, sctp);
914 continue;
915 }
916 sctp->sctp_refcnt++;
917 mutex_exit(&sctp->sctp_reflock);
918 mutex_exit(&sctps->sctps_g_lock);
919 if (sctp_prev != NULL)
920 SCTP_REFRELE(sctp_prev);
921
922 RUN_SCTP(sctp);
923 sobj = list_head(&sctp->sctp_saddrs[idx].sctp_ipif_list);
924 for (count = 0; count <
925 sctp->sctp_saddrs[idx].ipif_count; count++) {
926 if (sobj->saddr_ipifp == oipif) {
927 SCTP_IPIF_REFHOLD(nipif);
928 sobj->saddr_ipifp = nipif;
929 ASSERT(oipif->sctp_ipif_refcnt > 0);
930 /* We have the writer lock */
931 oipif->sctp_ipif_refcnt--;
932 /*
933 * Can't have more than one referring
934 * to the same sctp_ipif.
935 */
936 break;
937 }
938 sobj = list_next(&sctp->sctp_saddrs[idx].sctp_ipif_list,
939 sobj);
940 }
941 WAKE_SCTP(sctp);
942 sctp_prev = sctp;
943 mutex_enter(&sctps->sctps_g_lock);
944 sctp = list_next(&sctps->sctps_g_list, sctp);
945 }
946 mutex_exit(&sctps->sctps_g_lock);
947 if (sctp_prev != NULL)
948 SCTP_REFRELE(sctp_prev);
949 }
950
951 /*
952 * Given an ipif, walk the hash list in the global ipif table and for
953 * any other SCTP ipif with the same address and non-zero reference, walk
954 * the SCTP list and update the saddr list, if required, to point to the
955 * new SCTP ipif. If it is a loopback interface, then there could be
956 * multiple interfaces with 127.0.0.1 if there are zones configured, so
957 * check the zoneid in addition to the address.
958 */
959 void
960 sctp_chk_and_updt_saddr(int hindex, sctp_ipif_t *ipif, sctp_stack_t *sctps)
961 {
962 int cnt;
963 sctp_ipif_t *sipif;
964
965 ASSERT(sctps->sctps_g_ipifs[hindex].ipif_count > 0);
966 ASSERT(ipif->sctp_ipif_state == SCTP_IPIFS_UP);
967
968 sipif = list_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list);
969 for (cnt = 0; cnt < sctps->sctps_g_ipifs[hindex].ipif_count; cnt++) {
970 rw_enter(&sipif->sctp_ipif_lock, RW_WRITER);
971 if (sipif->sctp_ipif_id != ipif->sctp_ipif_id &&
972 IN6_ARE_ADDR_EQUAL(&sipif->sctp_ipif_saddr,
973 &ipif->sctp_ipif_saddr) && sipif->sctp_ipif_refcnt > 0 &&
974 (!SCTP_IS_IPIF_LOOPBACK(ipif) || ipif->sctp_ipif_zoneid ==
975 sipif->sctp_ipif_zoneid)) {
976 /*
977 * There can only be one address up at any time
978 * and we are here because ipif has been brought
979 * up.
980 */
981 ASSERT(sipif->sctp_ipif_state != SCTP_IPIFS_UP);
982 /*
983 * Someone has a reference to this we need to update to
984 * point to the new sipif.
985 */
986 sctp_update_saddrs(sipif, ipif, hindex, sctps);
987 }
988 rw_exit(&sipif->sctp_ipif_lock);
989 sipif = list_next(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list,
990 sipif);
991 }
992 }
993
994 /*
995 * Insert a new SCTP ipif using 'ipif'. v6addr is the address that existed
996 * prior to the current address in 'ipif'. Only when an existing address
997 * is changed on an IPIF, will v6addr be specified. If the IPIF already
998 * exists in the global SCTP ipif table, then we either removed it, if
999 * it doesn't have any existing reference, or mark it condemned otherwise.
1000 * If an address is being brought up (IPIF_UP), then we need to scan
1001 * the SCTP list to check if there is any SCTP that points to the *same*
1002 * address on a different SCTP ipif and update in that case.
1003 */
1004 void
1005 sctp_update_ipif_addr(ipif_t *ipif, in6_addr_t v6addr)
1006 {
1007 ill_t *ill = ipif->ipif_ill;
1008 int i;
1009 sctp_ill_t *sctp_ill;
1010 sctp_ill_t *osctp_ill;
1011 sctp_ipif_t *sctp_ipif = NULL;
1012 sctp_ipif_t *osctp_ipif = NULL;
1013 uint_t ill_index;
1014 int hindex;
1015 sctp_stack_t *sctps;
1016
1017 sctps = ipif->ipif_ill->ill_ipst->ips_netstack->netstack_sctp;
1018
1019 /* Index for new address */
1020 hindex = SCTP_IPIF_ADDR_HASH(ipif->ipif_v6lcl_addr, ill->ill_isv6);
1021
1022 /*
1023 * The address on this IPIF is changing, we need to look for
1024 * this old address and mark it condemned, before creating
1025 * one for the new address.
1026 */
1027 osctp_ipif = sctp_lookup_ipif_addr(&v6addr, B_FALSE,
1028 ipif->ipif_zoneid, B_TRUE, SCTP_ILL_TO_PHYINDEX(ill),
1029 ipif->ipif_seqid, B_FALSE, sctps);
1030
1031 rw_enter(&sctps->sctps_g_ills_lock, RW_READER);
1032 rw_enter(&sctps->sctps_g_ipifs_lock, RW_WRITER);
1033
1034 ill_index = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(ill));
1035 sctp_ill = list_head(&sctps->sctps_g_ills[ill_index].sctp_ill_list);
1036 for (i = 0; i < sctps->sctps_g_ills[ill_index].ill_count; i++) {
1037 if (sctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(ill) &&
1038 sctp_ill->sctp_ill_isv6 == ill->ill_isv6) {
1039 break;
1040 }
1041 sctp_ill = list_next(
1042 &sctps->sctps_g_ills[ill_index].sctp_ill_list, sctp_ill);
1043 }
1044
1045 if (sctp_ill == NULL) {
1046 ip1dbg(("sctp_update_ipif_addr: ill not found ..\n"));
1047 rw_exit(&sctps->sctps_g_ipifs_lock);
1048 rw_exit(&sctps->sctps_g_ills_lock);
1049 return;
1050 }
1051
1052 if (osctp_ipif != NULL) {
1053
1054 /* The address is the same? */
1055 if (IN6_ARE_ADDR_EQUAL(&ipif->ipif_v6lcl_addr, &v6addr)) {
1056 boolean_t chk_n_updt = B_FALSE;
1057
1058 rw_downgrade(&sctps->sctps_g_ipifs_lock);
1059 rw_enter(&osctp_ipif->sctp_ipif_lock, RW_WRITER);
1060 if (ipif->ipif_flags & IPIF_UP &&
1061 osctp_ipif->sctp_ipif_state != SCTP_IPIFS_UP) {
1062 osctp_ipif->sctp_ipif_state = SCTP_IPIFS_UP;
1063 chk_n_updt = B_TRUE;
1064 } else {
1065 osctp_ipif->sctp_ipif_state = SCTP_IPIFS_DOWN;
1066 }
1067 osctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1068 rw_exit(&osctp_ipif->sctp_ipif_lock);
1069 if (chk_n_updt) {
1070 sctp_chk_and_updt_saddr(hindex, osctp_ipif,
1071 sctps);
1072 }
1073 rw_exit(&sctps->sctps_g_ipifs_lock);
1074 rw_exit(&sctps->sctps_g_ills_lock);
1075 return;
1076 }
1077 /*
1078 * We are effectively removing this address from the ILL.
1079 */
1080 if (osctp_ipif->sctp_ipif_refcnt != 0) {
1081 osctp_ipif->sctp_ipif_state = SCTP_IPIFS_CONDEMNED;
1082 } else {
1083 list_t *ipif_list;
1084 int ohindex;
1085
1086 osctp_ill = osctp_ipif->sctp_ipif_ill;
1087 /* hash index for the old one */
1088 ohindex = SCTP_IPIF_ADDR_HASH(
1089 osctp_ipif->sctp_ipif_saddr,
1090 osctp_ipif->sctp_ipif_isv6);
1091
1092 ipif_list =
1093 &sctps->sctps_g_ipifs[ohindex].sctp_ipif_list;
1094
1095 list_remove(ipif_list, (void *)osctp_ipif);
1096 sctps->sctps_g_ipifs[ohindex].ipif_count--;
1097 sctps->sctps_g_ipifs_count--;
1098 rw_destroy(&osctp_ipif->sctp_ipif_lock);
1099 kmem_free(osctp_ipif, sizeof (sctp_ipif_t));
1100 (void) atomic_dec_32_nv(&osctp_ill->sctp_ill_ipifcnt);
1101 }
1102 }
1103
1104 sctp_ipif = kmem_zalloc(sizeof (sctp_ipif_t), KM_NOSLEEP);
1105 /* Try again? */
1106 if (sctp_ipif == NULL) {
1107 cmn_err(CE_WARN, "sctp_update_ipif_addr: error adding "
1108 "IPIF %p to SCTP's IPIF list", (void *)ipif);
1109 rw_exit(&sctps->sctps_g_ipifs_lock);
1110 rw_exit(&sctps->sctps_g_ills_lock);
1111 return;
1112 }
1113 sctps->sctps_g_ipifs_count++;
1114 rw_init(&sctp_ipif->sctp_ipif_lock, NULL, RW_DEFAULT, NULL);
1115 sctp_ipif->sctp_ipif_saddr = ipif->ipif_v6lcl_addr;
1116 sctp_ipif->sctp_ipif_ill = sctp_ill;
1117 sctp_ipif->sctp_ipif_isv6 = ill->ill_isv6;
1118 sctp_ipif->sctp_ipif_zoneid = ipif->ipif_zoneid;
1119 sctp_ipif->sctp_ipif_id = ipif->ipif_seqid;
1120 if (ipif->ipif_flags & IPIF_UP)
1121 sctp_ipif->sctp_ipif_state = SCTP_IPIFS_UP;
1122 else
1123 sctp_ipif->sctp_ipif_state = SCTP_IPIFS_DOWN;
1124 sctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1125 /*
1126 * We add it to the head so that it is quicker to find good/recent
1127 * additions.
1128 */
1129 list_insert_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list,
1130 (void *)sctp_ipif);
1131 sctps->sctps_g_ipifs[hindex].ipif_count++;
1132 atomic_inc_32(&sctp_ill->sctp_ill_ipifcnt);
1133 if (sctp_ipif->sctp_ipif_state == SCTP_IPIFS_UP)
1134 sctp_chk_and_updt_saddr(hindex, sctp_ipif, sctps);
1135 rw_exit(&sctps->sctps_g_ipifs_lock);
1136 rw_exit(&sctps->sctps_g_ills_lock);
1137 }
1138
1139 /* Insert, Remove, Mark up or Mark down the ipif */
1140 void
1141 sctp_update_ipif(ipif_t *ipif, int op)
1142 {
1143 ill_t *ill = ipif->ipif_ill;
1144 int i;
1145 sctp_ill_t *sctp_ill;
1146 sctp_ipif_t *sctp_ipif;
1147 uint_t ill_index;
1148 uint_t hindex;
1149 netstack_t *ns = ipif->ipif_ill->ill_ipst->ips_netstack;
1150 sctp_stack_t *sctps = ns->netstack_sctp;
1151
1152 ip2dbg(("sctp_update_ipif: %s %d\n", ill->ill_name, ipif->ipif_seqid));
1153
1154 rw_enter(&sctps->sctps_g_ills_lock, RW_READER);
1155 rw_enter(&sctps->sctps_g_ipifs_lock, RW_WRITER);
1156
1157 ill_index = SCTP_ILL_HASH_FN(SCTP_ILL_TO_PHYINDEX(ill));
1158 sctp_ill = list_head(&sctps->sctps_g_ills[ill_index].sctp_ill_list);
1159 for (i = 0; i < sctps->sctps_g_ills[ill_index].ill_count; i++) {
1160 if (sctp_ill->sctp_ill_index == SCTP_ILL_TO_PHYINDEX(ill) &&
1161 sctp_ill->sctp_ill_isv6 == ill->ill_isv6) {
1162 break;
1163 }
1164 sctp_ill = list_next(
1165 &sctps->sctps_g_ills[ill_index].sctp_ill_list, sctp_ill);
1166 }
1167 if (sctp_ill == NULL) {
1168 rw_exit(&sctps->sctps_g_ipifs_lock);
1169 rw_exit(&sctps->sctps_g_ills_lock);
1170 return;
1171 }
1172
1173 hindex = SCTP_IPIF_ADDR_HASH(ipif->ipif_v6lcl_addr,
1174 ipif->ipif_ill->ill_isv6);
1175 sctp_ipif = list_head(&sctps->sctps_g_ipifs[hindex].sctp_ipif_list);
1176 for (i = 0; i < sctps->sctps_g_ipifs[hindex].ipif_count; i++) {
1177 if (sctp_ipif->sctp_ipif_id == ipif->ipif_seqid) {
1178 ASSERT(IN6_ARE_ADDR_EQUAL(&sctp_ipif->sctp_ipif_saddr,
1179 &ipif->ipif_v6lcl_addr));
1180 break;
1181 }
1182 sctp_ipif = list_next(
1183 &sctps->sctps_g_ipifs[hindex].sctp_ipif_list,
1184 sctp_ipif);
1185 }
1186 if (sctp_ipif == NULL) {
1187 ip1dbg(("sctp_update_ipif: null sctp_ipif for %d\n", op));
1188 rw_exit(&sctps->sctps_g_ipifs_lock);
1189 rw_exit(&sctps->sctps_g_ills_lock);
1190 return;
1191 }
1192 ASSERT(sctp_ill == sctp_ipif->sctp_ipif_ill);
1193 switch (op) {
1194 case SCTP_IPIF_REMOVE:
1195 {
1196 list_t *ipif_list;
1197 list_t *ill_list;
1198
1199 ill_list = &sctps->sctps_g_ills[ill_index].sctp_ill_list;
1200 ipif_list = &sctps->sctps_g_ipifs[hindex].sctp_ipif_list;
1201 if (sctp_ipif->sctp_ipif_refcnt != 0) {
1202 sctp_ipif->sctp_ipif_state = SCTP_IPIFS_CONDEMNED;
1203 rw_exit(&sctps->sctps_g_ipifs_lock);
1204 rw_exit(&sctps->sctps_g_ills_lock);
1205 return;
1206 }
1207 list_remove(ipif_list, (void *)sctp_ipif);
1208 sctps->sctps_g_ipifs[hindex].ipif_count--;
1209 sctps->sctps_g_ipifs_count--;
1210 rw_destroy(&sctp_ipif->sctp_ipif_lock);
1211 kmem_free(sctp_ipif, sizeof (sctp_ipif_t));
1212 (void) atomic_dec_32_nv(&sctp_ill->sctp_ill_ipifcnt);
1213 if (rw_tryupgrade(&sctps->sctps_g_ills_lock) != 0) {
1214 rw_downgrade(&sctps->sctps_g_ipifs_lock);
1215 if (sctp_ill->sctp_ill_ipifcnt == 0 &&
1216 sctp_ill->sctp_ill_state == SCTP_ILLS_CONDEMNED) {
1217 list_remove(ill_list, (void *)sctp_ill);
1218 sctps->sctps_ills_count--;
1219 sctps->sctps_g_ills[ill_index].ill_count--;
1220 kmem_free(sctp_ill->sctp_ill_name,
1221 sctp_ill->sctp_ill_name_length);
1222 kmem_free(sctp_ill, sizeof (sctp_ill_t));
1223 }
1224 }
1225 break;
1226 }
1227
1228 case SCTP_IPIF_UP:
1229
1230 rw_downgrade(&sctps->sctps_g_ipifs_lock);
1231 rw_enter(&sctp_ipif->sctp_ipif_lock, RW_WRITER);
1232 sctp_ipif->sctp_ipif_state = SCTP_IPIFS_UP;
1233 sctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1234 rw_exit(&sctp_ipif->sctp_ipif_lock);
1235 sctp_chk_and_updt_saddr(hindex, sctp_ipif,
1236 ipif->ipif_ill->ill_ipst->ips_netstack->netstack_sctp);
1237
1238 break;
1239
1240 case SCTP_IPIF_UPDATE:
1241
1242 rw_downgrade(&sctps->sctps_g_ipifs_lock);
1243 rw_enter(&sctp_ipif->sctp_ipif_lock, RW_WRITER);
1244 sctp_ipif->sctp_ipif_zoneid = ipif->ipif_zoneid;
1245 sctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1246 rw_exit(&sctp_ipif->sctp_ipif_lock);
1247
1248 break;
1249
1250 case SCTP_IPIF_DOWN:
1251
1252 rw_downgrade(&sctps->sctps_g_ipifs_lock);
1253 rw_enter(&sctp_ipif->sctp_ipif_lock, RW_WRITER);
1254 sctp_ipif->sctp_ipif_state = SCTP_IPIFS_DOWN;
1255 sctp_ipif->sctp_ipif_flags = ipif->ipif_flags;
1256 rw_exit(&sctp_ipif->sctp_ipif_lock);
1257
1258 break;
1259 }
1260 rw_exit(&sctps->sctps_g_ipifs_lock);
1261 rw_exit(&sctps->sctps_g_ills_lock);
1262 }
1263
1264 /*
1265 * SCTP source address list manipulaton, locking not used (except for
1266 * sctp locking by the caller.
1267 */
1268
1269 /* Remove a specific saddr from the list */
1270 void
1271 sctp_del_saddr(sctp_t *sctp, sctp_saddr_ipif_t *sp)
1272 {
1273 if (sctp->sctp_conn_tfp != NULL)
1274 mutex_enter(&sctp->sctp_conn_tfp->tf_lock);
1275
1276 if (sctp->sctp_listen_tfp != NULL)
1277 mutex_enter(&sctp->sctp_listen_tfp->tf_lock);
1278
1279 sctp_ipif_hash_remove(sctp, sp->saddr_ipifp, B_FALSE);
1280
1281 if (sctp->sctp_bound_to_all == 1)
1282 sctp->sctp_bound_to_all = 0;
1283
1284 if (sctp->sctp_conn_tfp != NULL)
1285 mutex_exit(&sctp->sctp_conn_tfp->tf_lock);
1286
1287 if (sctp->sctp_listen_tfp != NULL)
1288 mutex_exit(&sctp->sctp_listen_tfp->tf_lock);
1289 }
1290
1291 /*
1292 * Delete source address from the existing list. No error checking done here
1293 * Called with no locks held.
1294 */
1295 void
1296 sctp_del_saddr_list(sctp_t *sctp, const void *addrs, int addcnt,
1297 boolean_t fanout_locked)
1298 {
1299 struct sockaddr_in *sin4;
1300 struct sockaddr_in6 *sin6;
1301 int cnt;
1302 in6_addr_t addr;
1303 sctp_ipif_t *sctp_ipif;
1304 int ifindex = 0;
1305 conn_t *connp = sctp->sctp_connp;
1306
1307 ASSERT(sctp->sctp_nsaddrs >= addcnt);
1308
1309 if (!fanout_locked) {
1310 if (sctp->sctp_conn_tfp != NULL)
1311 mutex_enter(&sctp->sctp_conn_tfp->tf_lock);
1312 if (sctp->sctp_listen_tfp != NULL)
1313 mutex_enter(&sctp->sctp_listen_tfp->tf_lock);
1314 }
1315
1316 for (cnt = 0; cnt < addcnt; cnt++) {
1317 switch (connp->conn_family) {
1318 case AF_INET:
1319 sin4 = (struct sockaddr_in *)addrs + cnt;
1320 IN6_INADDR_TO_V4MAPPED(&sin4->sin_addr, &addr);
1321 break;
1322
1323 case AF_INET6:
1324 sin6 = (struct sockaddr_in6 *)addrs + cnt;
1325 addr = sin6->sin6_addr;
1326 ifindex = sin6->sin6_scope_id;
1327 break;
1328 }
1329 sctp_ipif = sctp_lookup_ipif_addr(&addr, B_FALSE,
1330 IPCL_ZONEID(connp), !connp->conn_allzones,
1331 ifindex, 0, B_TRUE, sctp->sctp_sctps);
1332 ASSERT(sctp_ipif != NULL);
1333 sctp_ipif_hash_remove(sctp, sctp_ipif, B_FALSE);
1334 }
1335 if (sctp->sctp_bound_to_all == 1)
1336 sctp->sctp_bound_to_all = 0;
1337
1338 if (!fanout_locked) {
1339 if (sctp->sctp_conn_tfp != NULL)
1340 mutex_exit(&sctp->sctp_conn_tfp->tf_lock);
1341 if (sctp->sctp_listen_tfp != NULL)
1342 mutex_exit(&sctp->sctp_listen_tfp->tf_lock);
1343 }
1344 }
1345
1346 /*
1347 * Given an address get the corresponding entry from the list
1348 * Called with no locks held.
1349 */
1350 sctp_saddr_ipif_t *
1351 sctp_saddr_lookup(sctp_t *sctp, in6_addr_t *addr, uint_t ifindex)
1352 {
1353 int cnt;
1354 sctp_saddr_ipif_t *ipif_obj;
1355 int hindex;
1356 sctp_ipif_t *sctp_ipif;
1357
1358 hindex = SCTP_IPIF_ADDR_HASH(*addr, !IN6_IS_ADDR_V4MAPPED(addr));
1359 rw_enter(&sctp->sctp_saddrs[hindex].ipif_hash_lock, RW_READER);
1360 if (sctp->sctp_saddrs[hindex].ipif_count == 0) {
1361 rw_exit(&sctp->sctp_saddrs[hindex].ipif_hash_lock);
1362 return (NULL);
1363 }
1364
1365 ipif_obj = list_head(&sctp->sctp_saddrs[hindex].sctp_ipif_list);
1366 for (cnt = 0; cnt < sctp->sctp_saddrs[hindex].ipif_count; cnt++) {
1367 sctp_ipif = ipif_obj->saddr_ipifp;
1368 /*
1369 * Zone check shouldn't be needed.
1370 */
1371 if (IN6_ARE_ADDR_EQUAL(addr, &sctp_ipif->sctp_ipif_saddr) &&
1372 (ifindex == 0 ||
1373 ifindex == sctp_ipif->sctp_ipif_ill->sctp_ill_index) &&
1374 SCTP_IPIF_USABLE(sctp_ipif->sctp_ipif_state)) {
1375 rw_exit(&sctp->sctp_saddrs[hindex].ipif_hash_lock);
1376 return (ipif_obj);
1377 }
1378 ipif_obj = list_next(&sctp->sctp_saddrs[hindex].sctp_ipif_list,
1379 ipif_obj);
1380 }
1381 rw_exit(&sctp->sctp_saddrs[hindex].ipif_hash_lock);
1382 return (NULL);
1383 }
1384
1385 /* Given an address, add it to the source address list */
1386 int
1387 sctp_saddr_add_addr(sctp_t *sctp, in6_addr_t *addr, uint_t ifindex)
1388 {
1389 sctp_ipif_t *sctp_ipif;
1390 conn_t *connp = sctp->sctp_connp;
1391
1392 sctp_ipif = sctp_lookup_ipif_addr(addr, B_TRUE, IPCL_ZONEID(connp),
1393 !connp->conn_allzones, ifindex, 0, B_TRUE, sctp->sctp_sctps);
1394 if (sctp_ipif == NULL)
1395 return (EINVAL);
1396
1397 if (sctp_ipif_hash_insert(sctp, sctp_ipif, KM_NOSLEEP, B_FALSE,
1398 B_FALSE) != 0) {
1399 SCTP_IPIF_REFRELE(sctp_ipif);
1400 return (EINVAL);
1401 }
1402 return (0);
1403 }
1404
1405 /*
1406 * Remove or mark as dontsrc addresses that are currently not part of the
1407 * association. One would delete addresses when processing an INIT and
1408 * mark as dontsrc when processing an INIT-ACK.
1409 */
1410 void
1411 sctp_check_saddr(sctp_t *sctp, int supp_af, boolean_t delete,
1412 in6_addr_t *no_del_addr)
1413 {
1414 int i;
1415 int l;
1416 sctp_saddr_ipif_t *obj;
1417 int scanned = 0;
1418 int naddr;
1419 int nsaddr;
1420 conn_t *connp = sctp->sctp_connp;
1421
1422 ASSERT(!sctp->sctp_loopback && !sctp->sctp_linklocal && supp_af != 0);
1423
1424 /*
1425 * Irregardless of the supported address in the INIT, v4
1426 * must be supported.
1427 */
1428 if (connp->conn_family == AF_INET)
1429 supp_af = PARM_SUPP_V4;
1430
1431 nsaddr = sctp->sctp_nsaddrs;
1432 for (i = 0; i < SCTP_IPIF_HASH; i++) {
1433 rw_enter(&sctp->sctp_saddrs[i].ipif_hash_lock, RW_WRITER);
1434 if (sctp->sctp_saddrs[i].ipif_count == 0) {
1435 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1436 continue;
1437 }
1438 obj = list_head(&sctp->sctp_saddrs[i].sctp_ipif_list);
1439 naddr = sctp->sctp_saddrs[i].ipif_count;
1440 for (l = 0; l < naddr; l++) {
1441 sctp_ipif_t *ipif;
1442
1443 ipif = obj->saddr_ipifp;
1444 scanned++;
1445
1446 if (IN6_ARE_ADDR_EQUAL(&ipif->sctp_ipif_saddr,
1447 no_del_addr)) {
1448 goto next_obj;
1449 }
1450
1451 /*
1452 * Delete/mark dontsrc loopback/linklocal addresses and
1453 * unsupported address.
1454 * On a clustered node, we trust the clustering module
1455 * to do the right thing w.r.t loopback addresses, so
1456 * we ignore loopback addresses in this check.
1457 */
1458 if ((SCTP_IS_IPIF_LOOPBACK(ipif) &&
1459 cl_sctp_check_addrs == NULL) ||
1460 SCTP_IS_IPIF_LINKLOCAL(ipif) ||
1461 SCTP_UNSUPP_AF(ipif, supp_af)) {
1462 if (!delete) {
1463 obj->saddr_ipif_unconfirmed = 1;
1464 goto next_obj;
1465 }
1466 if (sctp->sctp_bound_to_all == 1)
1467 sctp->sctp_bound_to_all = 0;
1468 if (scanned < nsaddr) {
1469 obj = list_next(&sctp->sctp_saddrs[i].
1470 sctp_ipif_list, obj);
1471 sctp_ipif_hash_remove(sctp, ipif,
1472 B_TRUE);
1473 continue;
1474 }
1475 sctp_ipif_hash_remove(sctp, ipif, B_TRUE);
1476 }
1477 next_obj:
1478 if (scanned >= nsaddr) {
1479 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1480 return;
1481 }
1482 obj = list_next(&sctp->sctp_saddrs[i].sctp_ipif_list,
1483 obj);
1484 }
1485 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1486 }
1487 }
1488
1489
1490 /* Get the first valid address from the list. Called with no locks held */
1491 in6_addr_t
1492 sctp_get_valid_addr(sctp_t *sctp, boolean_t isv6, boolean_t *addr_set)
1493 {
1494 int i;
1495 int l;
1496 sctp_saddr_ipif_t *obj;
1497 int scanned = 0;
1498 in6_addr_t addr;
1499
1500 for (i = 0; i < SCTP_IPIF_HASH; i++) {
1501 rw_enter(&sctp->sctp_saddrs[i].ipif_hash_lock, RW_READER);
1502 if (sctp->sctp_saddrs[i].ipif_count == 0) {
1503 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1504 continue;
1505 }
1506 obj = list_head(&sctp->sctp_saddrs[i].sctp_ipif_list);
1507 for (l = 0; l < sctp->sctp_saddrs[i].ipif_count; l++) {
1508 sctp_ipif_t *ipif;
1509
1510 ipif = obj->saddr_ipifp;
1511 if (!SCTP_DONT_SRC(obj) &&
1512 ipif->sctp_ipif_isv6 == isv6 &&
1513 ipif->sctp_ipif_state == SCTP_IPIFS_UP) {
1514 *addr_set = B_TRUE;
1515 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1516 return (ipif->sctp_ipif_saddr);
1517 }
1518 scanned++;
1519 if (scanned >= sctp->sctp_nsaddrs) {
1520 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1521 goto got_none;
1522 }
1523 obj = list_next(&sctp->sctp_saddrs[i].sctp_ipif_list,
1524 obj);
1525 }
1526 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1527 }
1528 got_none:
1529 /* Need to double check this */
1530 if (isv6 == B_TRUE)
1531 addr = ipv6_all_zeros;
1532 else
1533 IN6_IPADDR_TO_V4MAPPED(0, &addr);
1534 *addr_set = B_FALSE;
1535 return (addr);
1536 }
1537
1538 /*
1539 * Return the list of local addresses of an association. The parameter
1540 * myaddrs is supposed to be either (struct sockaddr_in *) or (struct
1541 * sockaddr_in6 *) depending on the address family.
1542 */
1543 int
1544 sctp_getmyaddrs(void *conn, void *myaddrs, int *addrcnt)
1545 {
1546 int i;
1547 int l;
1548 sctp_saddr_ipif_t *obj;
1549 sctp_t *sctp = (sctp_t *)conn;
1550 conn_t *connp = sctp->sctp_connp;
1551 int family = connp->conn_family;
1552 int max = *addrcnt;
1553 size_t added = 0;
1554 struct sockaddr_in6 *sin6;
1555 struct sockaddr_in *sin4;
1556 int scanned = 0;
1557 boolean_t skip_lback = B_FALSE;
1558 ip_xmit_attr_t *ixa = connp->conn_ixa;
1559
1560 if (sctp->sctp_nsaddrs == 0)
1561 return (EINVAL);
1562
1563 /*
1564 * Skip loopback addresses for non-loopback assoc., ignore
1565 * this on a clustered node.
1566 */
1567 if (sctp->sctp_state >= SCTPS_ESTABLISHED && !sctp->sctp_loopback &&
1568 (cl_sctp_check_addrs == NULL)) {
1569 skip_lback = B_TRUE;
1570 }
1571
1572 for (i = 0; i < SCTP_IPIF_HASH; i++) {
1573 rw_enter(&sctp->sctp_saddrs[i].ipif_hash_lock, RW_READER);
1574 if (sctp->sctp_saddrs[i].ipif_count == 0) {
1575 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1576 continue;
1577 }
1578 obj = list_head(&sctp->sctp_saddrs[i].sctp_ipif_list);
1579 for (l = 0; l < sctp->sctp_saddrs[i].ipif_count; l++) {
1580 sctp_ipif_t *ipif = obj->saddr_ipifp;
1581 in6_addr_t addr = ipif->sctp_ipif_saddr;
1582
1583 scanned++;
1584 if ((ipif->sctp_ipif_state == SCTP_IPIFS_CONDEMNED) ||
1585 SCTP_DONT_SRC(obj) ||
1586 (SCTP_IS_IPIF_LOOPBACK(ipif) && skip_lback)) {
1587 if (scanned >= sctp->sctp_nsaddrs) {
1588 rw_exit(&sctp->
1589 sctp_saddrs[i].ipif_hash_lock);
1590 goto done;
1591 }
1592 obj = list_next(&sctp->sctp_saddrs[i].
1593 sctp_ipif_list, obj);
1594 continue;
1595 }
1596 switch (family) {
1597 case AF_INET:
1598 sin4 = (struct sockaddr_in *)myaddrs + added;
1599 sin4->sin_family = AF_INET;
1600 sin4->sin_port = connp->conn_lport;
1601 IN6_V4MAPPED_TO_INADDR(&addr, &sin4->sin_addr);
1602 break;
1603
1604 case AF_INET6:
1605 sin6 = (struct sockaddr_in6 *)myaddrs + added;
1606 sin6->sin6_family = AF_INET6;
1607 sin6->sin6_port = connp->conn_lport;
1608 sin6->sin6_addr = addr;
1609 /*
1610 * Note that flowinfo is only returned for
1611 * getpeername just like for TCP and UDP.
1612 */
1613 sin6->sin6_flowinfo = 0;
1614
1615 if (IN6_IS_ADDR_LINKSCOPE(&sin6->sin6_addr) &&
1616 (ixa->ixa_flags & IXAF_SCOPEID_SET))
1617 sin6->sin6_scope_id = ixa->ixa_scopeid;
1618 else
1619 sin6->sin6_scope_id = 0;
1620 sin6->__sin6_src_id = 0;
1621 break;
1622 }
1623 added++;
1624 if (added >= max || scanned >= sctp->sctp_nsaddrs) {
1625 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1626 goto done;
1627 }
1628 obj = list_next(&sctp->sctp_saddrs[i].sctp_ipif_list,
1629 obj);
1630 }
1631 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1632 }
1633 done:
1634 *addrcnt = added;
1635 return (0);
1636 }
1637
1638 /*
1639 * Given the supported address family, walk through the source address list
1640 * and return the total length of the available addresses. If 'p' is not
1641 * null, construct the parameter list for the addresses in 'p'.
1642 * 'modify' will only be set when we want the source address list to
1643 * be modified. The source address list will be modified only when
1644 * generating an INIT chunk. For generating an INIT-ACK 'modify' will
1645 * be false since the 'sctp' will be that of the listener.
1646 */
1647 size_t
1648 sctp_saddr_info(sctp_t *sctp, int supp_af, uchar_t *p, boolean_t modify)
1649 {
1650 int i;
1651 int l;
1652 sctp_saddr_ipif_t *obj;
1653 size_t paramlen = 0;
1654 sctp_parm_hdr_t *hdr;
1655 int scanned = 0;
1656 int naddr;
1657 int nsaddr;
1658 boolean_t del_ll = B_FALSE;
1659 boolean_t del_lb = B_FALSE;
1660
1661
1662 /*
1663 * On a clustered node don't bother changing anything
1664 * on the loopback interface.
1665 */
1666 if (modify && !sctp->sctp_loopback && (cl_sctp_check_addrs == NULL))
1667 del_lb = B_TRUE;
1668
1669 if (modify && !sctp->sctp_linklocal)
1670 del_ll = B_TRUE;
1671
1672 nsaddr = sctp->sctp_nsaddrs;
1673 for (i = 0; i < SCTP_IPIF_HASH; i++) {
1674 rw_enter(&sctp->sctp_saddrs[i].ipif_hash_lock, RW_WRITER);
1675 if (sctp->sctp_saddrs[i].ipif_count == 0) {
1676 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1677 continue;
1678 }
1679 obj = list_head(&sctp->sctp_saddrs[i].sctp_ipif_list);
1680 naddr = sctp->sctp_saddrs[i].ipif_count;
1681 for (l = 0; l < naddr; l++) {
1682 in6_addr_t addr;
1683 sctp_ipif_t *ipif;
1684 boolean_t ipif_lb;
1685 boolean_t ipif_ll;
1686 boolean_t unsupp_af;
1687
1688 ipif = obj->saddr_ipifp;
1689 scanned++;
1690
1691 ipif_lb = SCTP_IS_IPIF_LOOPBACK(ipif);
1692 ipif_ll = SCTP_IS_IPIF_LINKLOCAL(ipif);
1693 unsupp_af = SCTP_UNSUPP_AF(ipif, supp_af);
1694 /*
1695 * We need to either delete or skip loopback/linklocal
1696 * or unsupported addresses, if required.
1697 */
1698 if ((ipif_ll && del_ll) || (ipif_lb && del_lb) ||
1699 (unsupp_af && modify)) {
1700 if (sctp->sctp_bound_to_all == 1)
1701 sctp->sctp_bound_to_all = 0;
1702 if (scanned < nsaddr) {
1703 obj = list_next(&sctp->sctp_saddrs[i].
1704 sctp_ipif_list, obj);
1705 sctp_ipif_hash_remove(sctp, ipif,
1706 B_TRUE);
1707 continue;
1708 }
1709 sctp_ipif_hash_remove(sctp, ipif, B_TRUE);
1710
1711 goto next_addr;
1712 } else if (ipif_ll || unsupp_af ||
1713 (ipif_lb && (cl_sctp_check_addrs == NULL))) {
1714 goto next_addr;
1715 }
1716
1717 if (!SCTP_IPIF_USABLE(ipif->sctp_ipif_state))
1718 goto next_addr;
1719 if (p != NULL)
1720 hdr = (sctp_parm_hdr_t *)(p + paramlen);
1721 addr = ipif->sctp_ipif_saddr;
1722 if (!ipif->sctp_ipif_isv6) {
1723 struct in_addr *v4;
1724
1725 if (p != NULL) {
1726 hdr->sph_type = htons(PARM_ADDR4);
1727 hdr->sph_len = htons(PARM_ADDR4_LEN);
1728 v4 = (struct in_addr *)(hdr + 1);
1729 IN6_V4MAPPED_TO_INADDR(&addr, v4);
1730 }
1731 paramlen += PARM_ADDR4_LEN;
1732 } else {
1733 if (p != NULL) {
1734 hdr->sph_type = htons(PARM_ADDR6);
1735 hdr->sph_len = htons(PARM_ADDR6_LEN);
1736 bcopy(&addr, hdr + 1, sizeof (addr));
1737 }
1738 paramlen += PARM_ADDR6_LEN;
1739 }
1740 next_addr:
1741 if (scanned >= nsaddr) {
1742 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1743 return (paramlen);
1744 }
1745 obj = list_next(&sctp->sctp_saddrs[i].sctp_ipif_list,
1746 obj);
1747 }
1748 rw_exit(&sctp->sctp_saddrs[i].ipif_hash_lock);
1749 }
1750 return (paramlen);
1751 }
1752
1753 /*
1754 * This is used on a clustered node to obtain a list of addresses, the list
1755 * consists of sockaddr_in structs for v4 and sockaddr_in6 for v6. The list
1756 * is then passed onto the clustering module which sends back the correct
1757 * list based on the port info. Regardless of the input, i.e INADDR_ANY
1758 * or specific address(es), we create the list since it could be modified by
1759 * the clustering module. When given a list of addresses, we simply
1760 * create the list of sockaddr_in or sockaddr_in6 structs using those
1761 * addresses. If there is an INADDR_ANY in the input list, or if the
1762 * input is INADDR_ANY, we create a list of sockaddr_in or sockaddr_in6
1763 * structs consisting all the addresses in the global interface list
1764 * except those that are hosted on the loopback interface. We create
1765 * a list of sockaddr_in[6] structs just so that it can be directly input
1766 * to sctp_valid_addr_list() once the clustering module has processed it.
1767 */
1768 int
1769 sctp_get_addrlist(sctp_t *sctp, const void *addrs, uint32_t *addrcnt,
1770 uchar_t **addrlist, int *uspec, size_t *size)
1771 {
1772 int cnt;
1773 int icnt;
1774 sctp_ipif_t *sctp_ipif;
1775 struct sockaddr_in *s4;
1776 struct sockaddr_in6 *s6;
1777 uchar_t *p;
1778 int err = 0;
1779 sctp_stack_t *sctps = sctp->sctp_sctps;
1780 conn_t *connp = sctp->sctp_connp;
1781
1782 *addrlist = NULL;
1783 *size = 0;
1784
1785 /*
1786 * Create a list of sockaddr_in[6] structs using the input list.
1787 */
1788 if (connp->conn_family == AF_INET) {
1789 *size = sizeof (struct sockaddr_in) * *addrcnt;
1790 *addrlist = kmem_zalloc(*size, KM_SLEEP);
1791 p = *addrlist;
1792 for (cnt = 0; cnt < *addrcnt; cnt++) {
1793 s4 = (struct sockaddr_in *)addrs + cnt;
1794 /*
1795 * We need to create a list of all the available
1796 * addresses if there is an INADDR_ANY. However,
1797 * if we are beyond LISTEN, then this is invalid
1798 * (see sctp_valid_addr_list(). So, we just fail
1799 * it here rather than wait till it fails in
1800 * sctp_valid_addr_list().
1801 */
1802 if (s4->sin_addr.s_addr == INADDR_ANY) {
1803 kmem_free(*addrlist, *size);
1804 *addrlist = NULL;
1805 *size = 0;
1806 if (sctp->sctp_state > SCTPS_LISTEN) {
1807 *addrcnt = 0;
1808 return (EINVAL);
1809 }
1810 if (uspec != NULL)
1811 *uspec = 1;
1812 goto get_all_addrs;
1813 } else {
1814 bcopy(s4, p, sizeof (*s4));
1815 p += sizeof (*s4);
1816 }
1817 }
1818 } else {
1819 *size = sizeof (struct sockaddr_in6) * *addrcnt;
1820 *addrlist = kmem_zalloc(*size, KM_SLEEP);
1821 p = *addrlist;
1822 for (cnt = 0; cnt < *addrcnt; cnt++) {
1823 s6 = (struct sockaddr_in6 *)addrs + cnt;
1824 /*
1825 * Comments for INADDR_ANY, above, apply here too.
1826 */
1827 if (IN6_IS_ADDR_UNSPECIFIED(&s6->sin6_addr)) {
1828 kmem_free(*addrlist, *size);
1829 *size = 0;
1830 *addrlist = NULL;
1831 if (sctp->sctp_state > SCTPS_LISTEN) {
1832 *addrcnt = 0;
1833 return (EINVAL);
1834 }
1835 if (uspec != NULL)
1836 *uspec = 1;
1837 goto get_all_addrs;
1838 } else {
1839 bcopy(addrs, p, sizeof (*s6));
1840 p += sizeof (*s6);
1841 }
1842 }
1843 }
1844 return (err);
1845 get_all_addrs:
1846
1847 /*
1848 * Allocate max possible size. We allocate the max. size here because
1849 * the clustering module could end up adding addresses to the list.
1850 * We allocate upfront so that the clustering module need to bother
1851 * re-sizing the list.
1852 */
1853 if (connp->conn_family == AF_INET) {
1854 *size = sizeof (struct sockaddr_in) *
1855 sctps->sctps_g_ipifs_count;
1856 } else {
1857 *size = sizeof (struct sockaddr_in6) *
1858 sctps->sctps_g_ipifs_count;
1859 }
1860 *addrlist = kmem_zalloc(*size, KM_SLEEP);
1861 *addrcnt = 0;
1862 p = *addrlist;
1863 rw_enter(&sctps->sctps_g_ipifs_lock, RW_READER);
1864
1865 /*
1866 * Walk through the global interface list and add all addresses,
1867 * except those that are hosted on loopback interfaces.
1868 */
1869 for (cnt = 0; cnt < SCTP_IPIF_HASH; cnt++) {
1870 if (sctps->sctps_g_ipifs[cnt].ipif_count == 0)
1871 continue;
1872 sctp_ipif = list_head(
1873 &sctps->sctps_g_ipifs[cnt].sctp_ipif_list);
1874 for (icnt = 0;
1875 icnt < sctps->sctps_g_ipifs[cnt].ipif_count;
1876 icnt++) {
1877 in6_addr_t addr;
1878
1879 rw_enter(&sctp_ipif->sctp_ipif_lock, RW_READER);
1880 addr = sctp_ipif->sctp_ipif_saddr;
1881 if (SCTP_IPIF_DISCARD(sctp_ipif->sctp_ipif_flags) ||
1882 !SCTP_IPIF_USABLE(sctp_ipif->sctp_ipif_state) ||
1883 SCTP_IS_IPIF_LOOPBACK(sctp_ipif) ||
1884 SCTP_IS_IPIF_LINKLOCAL(sctp_ipif) ||
1885 !SCTP_IPIF_ZONE_MATCH(sctp, sctp_ipif) ||
1886 (connp->conn_family == AF_INET &&
1887 sctp_ipif->sctp_ipif_isv6) ||
1888 (sctp->sctp_connp->conn_ipv6_v6only &&
1889 !sctp_ipif->sctp_ipif_isv6)) {
1890 rw_exit(&sctp_ipif->sctp_ipif_lock);
1891 sctp_ipif = list_next(
1892 &sctps->sctps_g_ipifs[cnt].sctp_ipif_list,
1893 sctp_ipif);
1894 continue;
1895 }
1896 rw_exit(&sctp_ipif->sctp_ipif_lock);
1897 if (connp->conn_family == AF_INET) {
1898 s4 = (struct sockaddr_in *)p;
1899 IN6_V4MAPPED_TO_INADDR(&addr, &s4->sin_addr);
1900 s4->sin_family = AF_INET;
1901 p += sizeof (*s4);
1902 } else {
1903 s6 = (struct sockaddr_in6 *)p;
1904 s6->sin6_addr = addr;
1905 s6->sin6_family = AF_INET6;
1906 s6->sin6_scope_id =
1907 sctp_ipif->sctp_ipif_ill->sctp_ill_index;
1908 p += sizeof (*s6);
1909 }
1910 (*addrcnt)++;
1911 sctp_ipif = list_next(
1912 &sctps->sctps_g_ipifs[cnt].sctp_ipif_list,
1913 sctp_ipif);
1914 }
1915 }
1916 rw_exit(&sctps->sctps_g_ipifs_lock);
1917 return (err);
1918 }
1919
1920 /*
1921 * Get a list of addresses from the source address list. The caller is
1922 * responsible for allocating sufficient buffer for this.
1923 */
1924 void
1925 sctp_get_saddr_list(sctp_t *sctp, uchar_t *p, size_t psize)
1926 {
1927 int cnt;
1928 int icnt;
1929 sctp_saddr_ipif_t *obj;
1930 int naddr;
1931 int scanned = 0;
1932
1933 for (cnt = 0; cnt < SCTP_IPIF_HASH; cnt++) {
1934 rw_enter(&sctp->sctp_saddrs[cnt].ipif_hash_lock, RW_READER);
1935 if (sctp->sctp_saddrs[cnt].ipif_count == 0) {
1936 rw_exit(&sctp->sctp_saddrs[cnt].ipif_hash_lock);
1937 continue;
1938 }
1939 obj = list_head(&sctp->sctp_saddrs[cnt].sctp_ipif_list);
1940 naddr = sctp->sctp_saddrs[cnt].ipif_count;
1941 for (icnt = 0; icnt < naddr; icnt++) {
1942 sctp_ipif_t *ipif;
1943
1944 if (psize < sizeof (ipif->sctp_ipif_saddr)) {
1945 rw_exit(&sctp->sctp_saddrs[cnt].ipif_hash_lock);
1946 return;
1947 }
1948
1949 scanned++;
1950 ipif = obj->saddr_ipifp;
1951 bcopy(&ipif->sctp_ipif_saddr, p,
1952 sizeof (ipif->sctp_ipif_saddr));
1953 p += sizeof (ipif->sctp_ipif_saddr);
1954 psize -= sizeof (ipif->sctp_ipif_saddr);
1955 if (scanned >= sctp->sctp_nsaddrs) {
1956 rw_exit(&sctp->sctp_saddrs[cnt].ipif_hash_lock);
1957 return;
1958 }
1959 obj = list_next(
1960 &sctp->sctp_saddrs[icnt].sctp_ipif_list,
1961 obj);
1962 }
1963 rw_exit(&sctp->sctp_saddrs[cnt].ipif_hash_lock);
1964 }
1965 }
1966
1967 /*
1968 * Get a list of addresses from the remote address list. The caller is
1969 * responsible for allocating sufficient buffer for this.
1970 */
1971 void
1972 sctp_get_faddr_list(sctp_t *sctp, uchar_t *p, size_t psize)
1973 {
1974 sctp_faddr_t *fp;
1975
1976 for (fp = sctp->sctp_faddrs; fp != NULL; fp = fp->sf_next) {
1977 if (psize < sizeof (fp->sf_faddr))
1978 return;
1979 bcopy(&fp->sf_faddr, p, sizeof (fp->sf_faddr));
1980 p += sizeof (fp->sf_faddr);
1981 psize -= sizeof (fp->sf_faddr);
1982 }
1983 }
1984
1985 static void
1986 sctp_free_ills(sctp_stack_t *sctps)
1987 {
1988 int i;
1989 int l;
1990 sctp_ill_t *sctp_ill;
1991
1992 if (sctps->sctps_ills_count == 0)
1993 return;
1994
1995 for (i = 0; i < SCTP_ILL_HASH; i++) {
1996 sctp_ill = list_tail(&sctps->sctps_g_ills[i].sctp_ill_list);
1997 for (l = 0; l < sctps->sctps_g_ills[i].ill_count; l++) {
1998 ASSERT(sctp_ill->sctp_ill_ipifcnt == 0);
1999 list_remove(&sctps->sctps_g_ills[i].sctp_ill_list,
2000 sctp_ill);
2001 sctps->sctps_ills_count--;
2002 kmem_free(sctp_ill->sctp_ill_name,
2003 sctp_ill->sctp_ill_name_length);
2004 kmem_free(sctp_ill, sizeof (sctp_ill_t));
2005 sctp_ill =
2006 list_tail(&sctps->sctps_g_ills[i].sctp_ill_list);
2007 }
2008 sctps->sctps_g_ills[i].ill_count = 0;
2009 }
2010 ASSERT(sctps->sctps_ills_count == 0);
2011 }
2012
2013 static void
2014 sctp_free_ipifs(sctp_stack_t *sctps)
2015 {
2016 int i;
2017 int l;
2018 sctp_ipif_t *sctp_ipif;
2019 sctp_ill_t *sctp_ill;
2020
2021 if (sctps->sctps_g_ipifs_count == 0)
2022 return;
2023
2024 for (i = 0; i < SCTP_IPIF_HASH; i++) {
2025 sctp_ipif = list_tail(&sctps->sctps_g_ipifs[i].sctp_ipif_list);
2026 for (l = 0; l < sctps->sctps_g_ipifs[i].ipif_count; l++) {
2027 sctp_ill = sctp_ipif->sctp_ipif_ill;
2028
2029 list_remove(&sctps->sctps_g_ipifs[i].sctp_ipif_list,
2030 sctp_ipif);
2031 sctps->sctps_g_ipifs_count--;
2032 (void) atomic_dec_32_nv(&sctp_ill->sctp_ill_ipifcnt);
2033 kmem_free(sctp_ipif, sizeof (sctp_ipif_t));
2034 sctp_ipif =
2035 list_tail(&sctps->sctps_g_ipifs[i].sctp_ipif_list);
2036 }
2037 sctps->sctps_g_ipifs[i].ipif_count = 0;
2038 }
2039 ASSERT(sctps->sctps_g_ipifs_count == 0);
2040 }
2041
2042
2043 /* Initialize the SCTP ILL list and lock */
2044 void
2045 sctp_saddr_init(sctp_stack_t *sctps)
2046 {
2047 int i;
2048
2049 sctps->sctps_g_ills = kmem_zalloc(sizeof (sctp_ill_hash_t) *
2050 SCTP_ILL_HASH, KM_SLEEP);
2051 sctps->sctps_g_ipifs = kmem_zalloc(sizeof (sctp_ipif_hash_t) *
2052 SCTP_IPIF_HASH, KM_SLEEP);
2053
2054 rw_init(&sctps->sctps_g_ills_lock, NULL, RW_DEFAULT, NULL);
2055 rw_init(&sctps->sctps_g_ipifs_lock, NULL, RW_DEFAULT, NULL);
2056
2057 for (i = 0; i < SCTP_ILL_HASH; i++) {
2058 sctps->sctps_g_ills[i].ill_count = 0;
2059 list_create(&sctps->sctps_g_ills[i].sctp_ill_list,
2060 sizeof (sctp_ill_t),
2061 offsetof(sctp_ill_t, sctp_ills));
2062 }
2063 for (i = 0; i < SCTP_IPIF_HASH; i++) {
2064 sctps->sctps_g_ipifs[i].ipif_count = 0;
2065 list_create(&sctps->sctps_g_ipifs[i].sctp_ipif_list,
2066 sizeof (sctp_ipif_t), offsetof(sctp_ipif_t, sctp_ipifs));
2067 }
2068 }
2069
2070 void
2071 sctp_saddr_fini(sctp_stack_t *sctps)
2072 {
2073 int i;
2074
2075 sctp_free_ipifs(sctps);
2076 sctp_free_ills(sctps);
2077
2078 for (i = 0; i < SCTP_ILL_HASH; i++)
2079 list_destroy(&sctps->sctps_g_ills[i].sctp_ill_list);
2080 for (i = 0; i < SCTP_IPIF_HASH; i++)
2081 list_destroy(&sctps->sctps_g_ipifs[i].sctp_ipif_list);
2082
2083 ASSERT(sctps->sctps_ills_count == 0 && sctps->sctps_g_ipifs_count == 0);
2084 kmem_free(sctps->sctps_g_ills, sizeof (sctp_ill_hash_t) *
2085 SCTP_ILL_HASH);
2086 sctps->sctps_g_ills = NULL;
2087 kmem_free(sctps->sctps_g_ipifs, sizeof (sctp_ipif_hash_t) *
2088 SCTP_IPIF_HASH);
2089 sctps->sctps_g_ipifs = NULL;
2090 rw_destroy(&sctps->sctps_g_ills_lock);
2091 rw_destroy(&sctps->sctps_g_ipifs_lock);
2092 }