sctp_pcb.c revision 199437
1/*-
2 * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved.
3 *
4 * Redistribution and use in source and binary forms, with or without
5 * modification, are permitted provided that the following conditions are met:
6 *
7 * a) Redistributions of source code must retain the above copyright notice,
8 *   this list of conditions and the following disclaimer.
9 *
10 * b) Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in
12 *   the documentation and/or other materials provided with the distribution.
13 *
14 * c) Neither the name of Cisco Systems, Inc. nor the names of its
15 *    contributors may be used to endorse or promote products derived
16 *    from this software without specific prior written permission.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
20 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
22 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
28 * THE POSSIBILITY OF SUCH DAMAGE.
29 */
30
31/* $KAME: sctp_pcb.c,v 1.38 2005/03/06 16:04:18 itojun Exp $	 */
32
33#include <sys/cdefs.h>
34__FBSDID("$FreeBSD: head/sys/netinet/sctp_pcb.c 199437 2009-11-17 20:56:14Z tuexen $");
35
36#include <netinet/sctp_os.h>
37#include <sys/proc.h>
38#include <netinet/sctp_var.h>
39#include <netinet/sctp_sysctl.h>
40#include <netinet/sctp_pcb.h>
41#include <netinet/sctputil.h>
42#include <netinet/sctp.h>
43#include <netinet/sctp_header.h>
44#include <netinet/sctp_asconf.h>
45#include <netinet/sctp_output.h>
46#include <netinet/sctp_timer.h>
47#include <netinet/sctp_bsd_addr.h>
48#include <netinet/udp.h>
49
50
51VNET_DEFINE(struct sctp_base_info, system_base_info);
52
53/* FIX: we don't handle multiple link local scopes */
54/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
55#ifdef INET6
56int
57SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
58{
59	struct sockaddr_in6 tmp_a, tmp_b;
60
61	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
62	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
63		return 0;
64	}
65	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
66	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
67		return 0;
68	}
69	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
70}
71
72#endif
73
74void
75sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
76{
77	/*
78	 * We really don't need to lock this, but I will just because it
79	 * does not hurt.
80	 */
81	SCTP_INP_INFO_RLOCK();
82	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
83	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
84	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
85	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
86	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
87	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
88	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
89	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
90
91	SCTP_INP_INFO_RUNLOCK();
92}
93
94/*
95 * Addresses are added to VRF's (Virtual Router's). For BSD we
96 * have only the default VRF 0. We maintain a hash list of
97 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
98 * these has a list of addresses. When we add a new address
99 * to a VRF we lookup the ifn/ifn_index, if the ifn does
100 * not exist we create it and add it to the list of IFN's
101 * within the VRF. Once we have the sctp_ifn, we add the
102 * address to the list. So we look something like:
103 *
104 * hash-vrf-table
105 *   vrf-> ifn-> ifn -> ifn
106 *   vrf    |
107 *    ...   +--ifa-> ifa -> ifa
108 *   vrf
109 *
110 * We keep these seperate lists since the SCTP subsystem will
111 * point to these from its source address selection nets structure.
112 * When an address is deleted it does not happen right away on
113 * the SCTP side, it gets scheduled. What we do when a
114 * delete happens is immediately remove the address from
115 * the master list and decrement the refcount. As our
116 * addip iterator works through and frees the src address
117 * selection pointing to the sctp_ifa, eventually the refcount
118 * will reach 0 and we will delete it. Note that it is assumed
119 * that any locking on system level ifn/ifa is done at the
120 * caller of these functions and these routines will only
121 * lock the SCTP structures as they add or delete things.
122 *
123 * Other notes on VRF concepts.
124 *  - An endpoint can be in multiple VRF's
125 *  - An association lives within a VRF and only one VRF.
126 *  - Any incoming packet we can deduce the VRF for by
127 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
128 *  - Any downward send call or connect call must supply the
129 *    VRF via ancillary data or via some sort of set default
130 *    VRF socket option call (again for BSD no brainer since
131 *    the VRF is always 0).
132 *  - An endpoint may add multiple VRF's to it.
133 *  - Listening sockets can accept associations in any
134 *    of the VRF's they are in but the assoc will end up
135 *    in only one VRF (gotten from the packet or connect/send).
136 *
137 */
138
139struct sctp_vrf *
140sctp_allocate_vrf(int vrf_id)
141{
142	struct sctp_vrf *vrf = NULL;
143	struct sctp_vrflist *bucket;
144
145	/* First allocate the VRF structure */
146	vrf = sctp_find_vrf(vrf_id);
147	if (vrf) {
148		/* Already allocated */
149		return (vrf);
150	}
151	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
152	    SCTP_M_VRF);
153	if (vrf == NULL) {
154		/* No memory */
155#ifdef INVARIANTS
156		panic("No memory for VRF:%d", vrf_id);
157#endif
158		return (NULL);
159	}
160	/* setup the VRF */
161	memset(vrf, 0, sizeof(struct sctp_vrf));
162	vrf->vrf_id = vrf_id;
163	LIST_INIT(&vrf->ifnlist);
164	vrf->total_ifa_count = 0;
165	vrf->refcount = 0;
166	/* now also setup table ids */
167	SCTP_INIT_VRF_TABLEID(vrf);
168	/* Init the HASH of addresses */
169	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
170	    &vrf->vrf_addr_hashmark);
171	if (vrf->vrf_addr_hash == NULL) {
172		/* No memory */
173#ifdef INVARIANTS
174		panic("No memory for VRF:%d", vrf_id);
175#endif
176		SCTP_FREE(vrf, SCTP_M_VRF);
177		return (NULL);
178	}
179	/* Add it to the hash table */
180	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
181	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
182	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
183	return (vrf);
184}
185
186
187struct sctp_ifn *
188sctp_find_ifn(void *ifn, uint32_t ifn_index)
189{
190	struct sctp_ifn *sctp_ifnp;
191	struct sctp_ifnlist *hash_ifn_head;
192
193	/*
194	 * We assume the lock is held for the addresses if thats wrong
195	 * problems could occur :-)
196	 */
197	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
198	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
199		if (sctp_ifnp->ifn_index == ifn_index) {
200			return (sctp_ifnp);
201		}
202		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
203			return (sctp_ifnp);
204		}
205	}
206	return (NULL);
207}
208
209
210
211struct sctp_vrf *
212sctp_find_vrf(uint32_t vrf_id)
213{
214	struct sctp_vrflist *bucket;
215	struct sctp_vrf *liste;
216
217	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
218	LIST_FOREACH(liste, bucket, next_vrf) {
219		if (vrf_id == liste->vrf_id) {
220			return (liste);
221		}
222	}
223	return (NULL);
224}
225
226void
227sctp_free_vrf(struct sctp_vrf *vrf)
228{
229	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
230		if (vrf->vrf_addr_hash) {
231			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
232			vrf->vrf_addr_hash = NULL;
233		}
234		/* We zero'd the count */
235		LIST_REMOVE(vrf, next_vrf);
236		SCTP_FREE(vrf, SCTP_M_VRF);
237		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
238	}
239}
240
241void
242sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
243{
244	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
245		/* We zero'd the count */
246		if (sctp_ifnp->vrf) {
247			sctp_free_vrf(sctp_ifnp->vrf);
248		}
249		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
250		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
251	}
252}
253
254void
255sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
256{
257	struct sctp_ifn *sctp_ifnp;
258
259	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
260	if (sctp_ifnp != NULL) {
261		sctp_ifnp->ifn_mtu = mtu;
262	}
263}
264
265
266void
267sctp_free_ifa(struct sctp_ifa *sctp_ifap)
268{
269	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
270		/* We zero'd the count */
271		if (sctp_ifap->ifn_p) {
272			sctp_free_ifn(sctp_ifap->ifn_p);
273		}
274		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
275		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
276	}
277}
278
279static void
280sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
281{
282	struct sctp_ifn *found;
283
284	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
285	if (found == NULL) {
286		/* Not in the list.. sorry */
287		return;
288	}
289	if (hold_addr_lock == 0)
290		SCTP_IPI_ADDR_WLOCK();
291	LIST_REMOVE(sctp_ifnp, next_bucket);
292	LIST_REMOVE(sctp_ifnp, next_ifn);
293	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
294	    sctp_ifnp->registered_af);
295	if (hold_addr_lock == 0)
296		SCTP_IPI_ADDR_WUNLOCK();
297	/* Take away the reference, and possibly free it */
298	sctp_free_ifn(sctp_ifnp);
299}
300
301void
302sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
303    const char *if_name, uint32_t ifn_index)
304{
305	struct sctp_vrf *vrf;
306	struct sctp_ifa *sctp_ifap = NULL;
307
308	SCTP_IPI_ADDR_RLOCK();
309	vrf = sctp_find_vrf(vrf_id);
310	if (vrf == NULL) {
311		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
312		goto out;
313
314	}
315	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
316	if (sctp_ifap == NULL) {
317		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
318		goto out;
319	}
320	if (sctp_ifap->ifn_p == NULL) {
321		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
322		goto out;
323	}
324	if (if_name) {
325		int len1, len2;
326
327		len1 = strlen(if_name);
328		len2 = strlen(sctp_ifap->ifn_p->ifn_name);
329		if (len1 != len2) {
330			SCTPDBG(SCTP_DEBUG_PCB4, "IFN of ifa names different lenght %d vs %d - ignored\n",
331			    len1, len2);
332			goto out;
333		}
334		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) != 0) {
335			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
336			    sctp_ifap->ifn_p->ifn_name,
337			    if_name);
338			goto out;
339		}
340	} else {
341		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
342			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
343			    sctp_ifap->ifn_p->ifn_index, ifn_index);
344			goto out;
345		}
346	}
347
348	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
349	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
350out:
351	SCTP_IPI_ADDR_RUNLOCK();
352}
353
354void
355sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
356    const char *if_name, uint32_t ifn_index)
357{
358	struct sctp_vrf *vrf;
359	struct sctp_ifa *sctp_ifap = NULL;
360
361	SCTP_IPI_ADDR_RLOCK();
362	vrf = sctp_find_vrf(vrf_id);
363	if (vrf == NULL) {
364		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
365		goto out;
366
367	}
368	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
369	if (sctp_ifap == NULL) {
370		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
371		goto out;
372	}
373	if (sctp_ifap->ifn_p == NULL) {
374		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
375		goto out;
376	}
377	if (if_name) {
378		int len1, len2;
379
380		len1 = strlen(if_name);
381		len2 = strlen(sctp_ifap->ifn_p->ifn_name);
382		if (len1 != len2) {
383			SCTPDBG(SCTP_DEBUG_PCB4, "IFN of ifa names different lenght %d vs %d - ignored\n",
384			    len1, len2);
385			goto out;
386		}
387		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) != 0) {
388			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
389			    sctp_ifap->ifn_p->ifn_name,
390			    if_name);
391			goto out;
392		}
393	} else {
394		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
395			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
396			    sctp_ifap->ifn_p->ifn_index, ifn_index);
397			goto out;
398		}
399	}
400
401	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
402	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
403out:
404	SCTP_IPI_ADDR_RUNLOCK();
405}
406
407/*-
408 * Add an ifa to an ifn.
409 * Register the interface as necessary.
410 * NOTE: ADDR write lock MUST be held.
411 */
412static void
413sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
414{
415	int ifa_af;
416
417	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
418	sctp_ifap->ifn_p = sctp_ifnp;
419	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
420	/* update address counts */
421	sctp_ifnp->ifa_count++;
422	ifa_af = sctp_ifap->address.sa.sa_family;
423	if (ifa_af == AF_INET)
424		sctp_ifnp->num_v4++;
425	else
426		sctp_ifnp->num_v6++;
427	if (sctp_ifnp->ifa_count == 1) {
428		/* register the new interface */
429		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
430		sctp_ifnp->registered_af = ifa_af;
431	}
432}
433
434/*-
435 * Remove an ifa from its ifn.
436 * If no more addresses exist, remove the ifn too. Otherwise, re-register
437 * the interface based on the remaining address families left.
438 * NOTE: ADDR write lock MUST be held.
439 */
440static void
441sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
442{
443	uint32_t ifn_index;
444
445	LIST_REMOVE(sctp_ifap, next_ifa);
446	if (sctp_ifap->ifn_p) {
447		/* update address counts */
448		sctp_ifap->ifn_p->ifa_count--;
449		if (sctp_ifap->address.sa.sa_family == AF_INET6)
450			sctp_ifap->ifn_p->num_v6--;
451		else if (sctp_ifap->address.sa.sa_family == AF_INET)
452			sctp_ifap->ifn_p->num_v4--;
453
454		ifn_index = sctp_ifap->ifn_p->ifn_index;
455		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
456			/* remove the ifn, possibly freeing it */
457			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
458		} else {
459			/* re-register address family type, if needed */
460			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
461			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
462				SCTP_DEREGISTER_INTERFACE(ifn_index, AF_INET6);
463				SCTP_REGISTER_INTERFACE(ifn_index, AF_INET);
464				sctp_ifap->ifn_p->registered_af = AF_INET;
465			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
466			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
467				SCTP_DEREGISTER_INTERFACE(ifn_index, AF_INET);
468				SCTP_REGISTER_INTERFACE(ifn_index, AF_INET6);
469				sctp_ifap->ifn_p->registered_af = AF_INET6;
470			}
471			/* free the ifn refcount */
472			sctp_free_ifn(sctp_ifap->ifn_p);
473		}
474		sctp_ifap->ifn_p = NULL;
475	}
476}
477
478struct sctp_ifa *
479sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
480    uint32_t ifn_type, const char *if_name, void *ifa,
481    struct sockaddr *addr, uint32_t ifa_flags,
482    int dynamic_add)
483{
484	struct sctp_vrf *vrf;
485	struct sctp_ifn *sctp_ifnp = NULL;
486	struct sctp_ifa *sctp_ifap = NULL;
487	struct sctp_ifalist *hash_addr_head;
488	struct sctp_ifnlist *hash_ifn_head;
489	uint32_t hash_of_addr;
490	int new_ifn_af = 0;
491
492#ifdef SCTP_DEBUG
493	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
494	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
495#endif
496	SCTP_IPI_ADDR_WLOCK();
497	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
498	if (sctp_ifnp) {
499		vrf = sctp_ifnp->vrf;
500	} else {
501		vrf = sctp_find_vrf(vrf_id);
502		if (vrf == NULL) {
503			vrf = sctp_allocate_vrf(vrf_id);
504			if (vrf == NULL) {
505				SCTP_IPI_ADDR_WUNLOCK();
506				return (NULL);
507			}
508		}
509	}
510	if (sctp_ifnp == NULL) {
511		/*
512		 * build one and add it, can't hold lock until after malloc
513		 * done though.
514		 */
515		SCTP_IPI_ADDR_WUNLOCK();
516		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
517		    sizeof(struct sctp_ifn), SCTP_M_IFN);
518		if (sctp_ifnp == NULL) {
519#ifdef INVARIANTS
520			panic("No memory for IFN:%u", sctp_ifnp->ifn_index);
521#endif
522			return (NULL);
523		}
524		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
525		sctp_ifnp->ifn_index = ifn_index;
526		sctp_ifnp->ifn_p = ifn;
527		sctp_ifnp->ifn_type = ifn_type;
528		sctp_ifnp->refcount = 0;
529		sctp_ifnp->vrf = vrf;
530		atomic_add_int(&vrf->refcount, 1);
531		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
532		if (if_name != NULL) {
533			memcpy(sctp_ifnp->ifn_name, if_name, SCTP_IFNAMSIZ);
534		} else {
535			memcpy(sctp_ifnp->ifn_name, "unknown", min(7, SCTP_IFNAMSIZ));
536		}
537		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
538		LIST_INIT(&sctp_ifnp->ifalist);
539		SCTP_IPI_ADDR_WLOCK();
540		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
541		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
542		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
543		new_ifn_af = 1;
544	}
545	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
546	if (sctp_ifap) {
547		/* Hmm, it already exists? */
548		if ((sctp_ifap->ifn_p) &&
549		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
550			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
551			    sctp_ifap->ifn_p->ifn_name, ifn_index,
552			    sctp_ifap);
553			if (new_ifn_af) {
554				/* Remove the created one that we don't want */
555				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
556			}
557			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
558				/* easy to solve, just switch back to active */
559				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
560				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
561				sctp_ifap->ifn_p = sctp_ifnp;
562				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
563			}
564	exit_stage_left:
565			SCTP_IPI_ADDR_WUNLOCK();
566			return (sctp_ifap);
567		} else {
568			if (sctp_ifap->ifn_p) {
569				/*
570				 * The last IFN gets the address, removee
571				 * the old one
572				 */
573				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
574				    sctp_ifap, sctp_ifap->ifn_p->ifn_name,
575				    sctp_ifap->ifn_p->ifn_index, if_name,
576				    ifn_index);
577				/* remove the address from the old ifn */
578				sctp_remove_ifa_from_ifn(sctp_ifap);
579				/* move the address over to the new ifn */
580				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
581				goto exit_stage_left;
582			} else {
583				/* repair ifnp which was NULL ? */
584				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
585				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
586				    sctp_ifnp, sctp_ifap);
587				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
588			}
589			goto exit_stage_left;
590		}
591	}
592	SCTP_IPI_ADDR_WUNLOCK();
593	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
594	if (sctp_ifap == NULL) {
595#ifdef INVARIANTS
596		panic("No memory for IFA");
597#endif
598		return (NULL);
599	}
600	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
601	sctp_ifap->ifn_p = sctp_ifnp;
602	atomic_add_int(&sctp_ifnp->refcount, 1);
603	sctp_ifap->vrf_id = vrf_id;
604	sctp_ifap->ifa = ifa;
605	memcpy(&sctp_ifap->address, addr, addr->sa_len);
606	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
607	sctp_ifap->flags = ifa_flags;
608	/* Set scope */
609	switch (sctp_ifap->address.sa.sa_family) {
610	case AF_INET:
611		{
612			struct sockaddr_in *sin;
613
614			sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
615			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
616			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
617				sctp_ifap->src_is_loop = 1;
618			}
619			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
620				sctp_ifap->src_is_priv = 1;
621			}
622			sctp_ifnp->num_v4++;
623			if (new_ifn_af)
624				new_ifn_af = AF_INET;
625			break;
626		}
627#ifdef INET6
628	case AF_INET6:
629		{
630			/* ok to use deprecated addresses? */
631			struct sockaddr_in6 *sin6;
632
633			sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
634			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
635			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
636				sctp_ifap->src_is_loop = 1;
637			}
638			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
639				sctp_ifap->src_is_priv = 1;
640			}
641			sctp_ifnp->num_v6++;
642			if (new_ifn_af)
643				new_ifn_af = AF_INET6;
644			break;
645		}
646#endif
647	default:
648		new_ifn_af = 0;
649		break;
650	}
651	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
652
653	if ((sctp_ifap->src_is_priv == 0) &&
654	    (sctp_ifap->src_is_loop == 0)) {
655		sctp_ifap->src_is_glob = 1;
656	}
657	SCTP_IPI_ADDR_WLOCK();
658	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
659	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
660	sctp_ifap->refcount = 1;
661	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
662	sctp_ifnp->ifa_count++;
663	vrf->total_ifa_count++;
664	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
665	if (new_ifn_af) {
666		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
667		sctp_ifnp->registered_af = new_ifn_af;
668	}
669	SCTP_IPI_ADDR_WUNLOCK();
670	if (dynamic_add) {
671		/*
672		 * Bump up the refcount so that when the timer completes it
673		 * will drop back down.
674		 */
675		struct sctp_laddr *wi;
676
677		atomic_add_int(&sctp_ifap->refcount, 1);
678		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
679		if (wi == NULL) {
680			/*
681			 * Gak, what can we do? We have lost an address
682			 * change can you say HOSED?
683			 */
684			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
685			/* Opps, must decrement the count */
686			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
687			    if_name);
688			return (NULL);
689		}
690		SCTP_INCR_LADDR_COUNT();
691		bzero(wi, sizeof(*wi));
692		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
693		wi->ifa = sctp_ifap;
694		wi->action = SCTP_ADD_IP_ADDRESS;
695		SCTP_IPI_ITERATOR_WQ_LOCK();
696		/*
697		 * Should this really be a tailq? As it is we will process
698		 * the newest first :-0
699		 */
700		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
701		SCTP_IPI_ITERATOR_WQ_UNLOCK();
702		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
703		    (struct sctp_inpcb *)NULL,
704		    (struct sctp_tcb *)NULL,
705		    (struct sctp_nets *)NULL);
706	} else {
707		/* it's ready for use */
708		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
709	}
710	return (sctp_ifap);
711}
712
713void
714sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
715    uint32_t ifn_index, const char *if_name)
716{
717	struct sctp_vrf *vrf;
718	struct sctp_ifa *sctp_ifap = NULL;
719
720	SCTP_IPI_ADDR_WLOCK();
721	vrf = sctp_find_vrf(vrf_id);
722	if (vrf == NULL) {
723		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
724		goto out_now;
725	}
726#ifdef SCTP_DEBUG
727	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
728	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
729#endif
730	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
731	if (sctp_ifap) {
732		/* Validate the delete */
733		if (sctp_ifap->ifn_p) {
734			int valid = 0;
735
736			/*-
737			 * The name has priority over the ifn_index
738			 * if its given. We do this especially for
739			 * panda who might recycle indexes fast.
740			 */
741			if (if_name) {
742				int len1, len2;
743
744				len1 = min(SCTP_IFNAMSIZ, strlen(if_name));
745				len2 = min(SCTP_IFNAMSIZ, strlen(sctp_ifap->ifn_p->ifn_name));
746				if (len1 && len2 && (len1 == len2)) {
747					/* we can compare them */
748					if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) == 0) {
749						/*
750						 * They match its a correct
751						 * delete
752						 */
753						valid = 1;
754					}
755				}
756			}
757			if (!valid) {
758				/* last ditch check ifn_index */
759				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
760					valid = 1;
761				}
762			}
763			if (!valid) {
764				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
765				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
766				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
767				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
768				SCTP_IPI_ADDR_WUNLOCK();
769				return;
770			}
771		}
772		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", sctp_ifap);
773		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
774		sctp_ifap->localifa_flags |= SCTP_BEING_DELETED;
775		vrf->total_ifa_count--;
776		LIST_REMOVE(sctp_ifap, next_bucket);
777		sctp_remove_ifa_from_ifn(sctp_ifap);
778	}
779#ifdef SCTP_DEBUG
780	else {
781		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
782		    ifn_index);
783		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
784	}
785#endif
786
787out_now:
788	SCTP_IPI_ADDR_WUNLOCK();
789	if (sctp_ifap) {
790		struct sctp_laddr *wi;
791
792		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
793		if (wi == NULL) {
794			/*
795			 * Gak, what can we do? We have lost an address
796			 * change can you say HOSED?
797			 */
798			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
799
800			/* Oops, must decrement the count */
801			sctp_free_ifa(sctp_ifap);
802			return;
803		}
804		SCTP_INCR_LADDR_COUNT();
805		bzero(wi, sizeof(*wi));
806		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
807		wi->ifa = sctp_ifap;
808		wi->action = SCTP_DEL_IP_ADDRESS;
809		SCTP_IPI_ITERATOR_WQ_LOCK();
810		/*
811		 * Should this really be a tailq? As it is we will process
812		 * the newest first :-0
813		 */
814		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
815		SCTP_IPI_ITERATOR_WQ_UNLOCK();
816
817		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
818		    (struct sctp_inpcb *)NULL,
819		    (struct sctp_tcb *)NULL,
820		    (struct sctp_nets *)NULL);
821	}
822	return;
823}
824
825
826static struct sctp_tcb *
827sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
828    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
829{
830	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
831	/*
832	 * If we support the TCP model, then we must now dig through to see
833	 * if we can find our endpoint in the list of tcp ep's.
834	 */
835	uint16_t lport, rport;
836	struct sctppcbhead *ephead;
837	struct sctp_inpcb *inp;
838	struct sctp_laddr *laddr;
839	struct sctp_tcb *stcb;
840	struct sctp_nets *net;
841
842	if ((to == NULL) || (from == NULL)) {
843		return (NULL);
844	}
845	if (to->sa_family == AF_INET && from->sa_family == AF_INET) {
846		lport = ((struct sockaddr_in *)to)->sin_port;
847		rport = ((struct sockaddr_in *)from)->sin_port;
848	} else if (to->sa_family == AF_INET6 && from->sa_family == AF_INET6) {
849		lport = ((struct sockaddr_in6 *)to)->sin6_port;
850		rport = ((struct sockaddr_in6 *)from)->sin6_port;
851	} else {
852		return NULL;
853	}
854	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport), SCTP_BASE_INFO(hashtcpmark))];
855	/*
856	 * Ok now for each of the guys in this bucket we must look and see:
857	 * - Does the remote port match. - Does there single association's
858	 * addresses match this address (to). If so we update p_ep to point
859	 * to this ep and return the tcb from it.
860	 */
861	LIST_FOREACH(inp, ephead, sctp_hash) {
862		SCTP_INP_RLOCK(inp);
863		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
864			SCTP_INP_RUNLOCK(inp);
865			continue;
866		}
867		if (lport != inp->sctp_lport) {
868			SCTP_INP_RUNLOCK(inp);
869			continue;
870		}
871		if (inp->def_vrf_id != vrf_id) {
872			SCTP_INP_RUNLOCK(inp);
873			continue;
874		}
875		/* check to see if the ep has one of the addresses */
876		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
877			/* We are NOT bound all, so look further */
878			int match = 0;
879
880			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
881
882				if (laddr->ifa == NULL) {
883					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
884					continue;
885				}
886				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
887					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
888					continue;
889				}
890				if (laddr->ifa->address.sa.sa_family ==
891				    to->sa_family) {
892					/* see if it matches */
893					struct sockaddr_in *intf_addr, *sin;
894
895					intf_addr = &laddr->ifa->address.sin;
896					sin = (struct sockaddr_in *)to;
897					if (from->sa_family == AF_INET) {
898						if (sin->sin_addr.s_addr ==
899						    intf_addr->sin_addr.s_addr) {
900							match = 1;
901							break;
902						}
903					}
904#ifdef INET6
905					if (from->sa_family == AF_INET6) {
906						struct sockaddr_in6 *intf_addr6;
907						struct sockaddr_in6 *sin6;
908
909						sin6 = (struct sockaddr_in6 *)
910						    to;
911						intf_addr6 = &laddr->ifa->address.sin6;
912
913						if (SCTP6_ARE_ADDR_EQUAL(sin6,
914						    intf_addr6)) {
915							match = 1;
916							break;
917						}
918					}
919#endif
920				}
921			}
922			if (match == 0) {
923				/* This endpoint does not have this address */
924				SCTP_INP_RUNLOCK(inp);
925				continue;
926			}
927		}
928		/*
929		 * Ok if we hit here the ep has the address, does it hold
930		 * the tcb?
931		 */
932
933		stcb = LIST_FIRST(&inp->sctp_asoc_list);
934		if (stcb == NULL) {
935			SCTP_INP_RUNLOCK(inp);
936			continue;
937		}
938		SCTP_TCB_LOCK(stcb);
939		if (stcb->rport != rport) {
940			/* remote port does not match. */
941			SCTP_TCB_UNLOCK(stcb);
942			SCTP_INP_RUNLOCK(inp);
943			continue;
944		}
945		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
946			SCTP_TCB_UNLOCK(stcb);
947			SCTP_INP_RUNLOCK(inp);
948			continue;
949		}
950		/* Does this TCB have a matching address? */
951		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
952
953			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
954				/* not the same family, can't be a match */
955				continue;
956			}
957			switch (from->sa_family) {
958			case AF_INET:
959				{
960					struct sockaddr_in *sin, *rsin;
961
962					sin = (struct sockaddr_in *)&net->ro._l_addr;
963					rsin = (struct sockaddr_in *)from;
964					if (sin->sin_addr.s_addr ==
965					    rsin->sin_addr.s_addr) {
966						/* found it */
967						if (netp != NULL) {
968							*netp = net;
969						}
970						/*
971						 * Update the endpoint
972						 * pointer
973						 */
974						*inp_p = inp;
975						SCTP_INP_RUNLOCK(inp);
976						return (stcb);
977					}
978					break;
979				}
980#ifdef INET6
981			case AF_INET6:
982				{
983					struct sockaddr_in6 *sin6, *rsin6;
984
985					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
986					rsin6 = (struct sockaddr_in6 *)from;
987					if (SCTP6_ARE_ADDR_EQUAL(sin6,
988					    rsin6)) {
989						/* found it */
990						if (netp != NULL) {
991							*netp = net;
992						}
993						/*
994						 * Update the endpoint
995						 * pointer
996						 */
997						*inp_p = inp;
998						SCTP_INP_RUNLOCK(inp);
999						return (stcb);
1000					}
1001					break;
1002				}
1003#endif
1004			default:
1005				/* TSNH */
1006				break;
1007			}
1008		}
1009		SCTP_TCB_UNLOCK(stcb);
1010		SCTP_INP_RUNLOCK(inp);
1011	}
1012	return (NULL);
1013}
1014
1015/*
1016 * rules for use
1017 *
1018 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1019 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1020 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1021 * NULL.
1022 */
1023
1024struct sctp_tcb *
1025sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1026    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1027{
1028	struct sctpasochead *head;
1029	struct sctp_inpcb *inp;
1030	struct sctp_tcb *stcb = NULL;
1031	struct sctp_nets *net;
1032	uint16_t rport;
1033
1034	inp = *inp_p;
1035	if (remote->sa_family == AF_INET) {
1036		rport = (((struct sockaddr_in *)remote)->sin_port);
1037	} else if (remote->sa_family == AF_INET6) {
1038		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1039	} else {
1040		return (NULL);
1041	}
1042	if (locked_tcb) {
1043		/*
1044		 * UN-lock so we can do proper locking here this occurs when
1045		 * called from load_addresses_from_init.
1046		 */
1047		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1048		SCTP_TCB_UNLOCK(locked_tcb);
1049	}
1050	SCTP_INP_INFO_RLOCK();
1051	if (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) {
1052		/*-
1053		 * Now either this guy is our listener or it's the
1054		 * connector. If it is the one that issued the connect, then
1055		 * it's only chance is to be the first TCB in the list. If
1056		 * it is the acceptor, then do the special_lookup to hash
1057		 * and find the real inp.
1058		 */
1059		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1060			/* to is peer addr, from is my addr */
1061			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1062			    netp, inp->def_vrf_id);
1063			if ((stcb != NULL) && (locked_tcb == NULL)) {
1064				/* we have a locked tcb, lower refcount */
1065				SCTP_INP_DECR_REF(inp);
1066			}
1067			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1068				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1069				SCTP_TCB_LOCK(locked_tcb);
1070				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1071				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1072			}
1073			SCTP_INP_INFO_RUNLOCK();
1074			return (stcb);
1075		} else {
1076			SCTP_INP_WLOCK(inp);
1077			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1078				goto null_return;
1079			}
1080			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1081			if (stcb == NULL) {
1082				goto null_return;
1083			}
1084			SCTP_TCB_LOCK(stcb);
1085
1086			if (stcb->rport != rport) {
1087				/* remote port does not match. */
1088				SCTP_TCB_UNLOCK(stcb);
1089				goto null_return;
1090			}
1091			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1092				SCTP_TCB_UNLOCK(stcb);
1093				goto null_return;
1094			}
1095			/* now look at the list of remote addresses */
1096			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1097#ifdef INVARIANTS
1098				if (net == (TAILQ_NEXT(net, sctp_next))) {
1099					panic("Corrupt net list");
1100				}
1101#endif
1102				if (net->ro._l_addr.sa.sa_family !=
1103				    remote->sa_family) {
1104					/* not the same family */
1105					continue;
1106				}
1107				switch (remote->sa_family) {
1108				case AF_INET:
1109					{
1110						struct sockaddr_in *sin,
1111						           *rsin;
1112
1113						sin = (struct sockaddr_in *)
1114						    &net->ro._l_addr;
1115						rsin = (struct sockaddr_in *)remote;
1116						if (sin->sin_addr.s_addr ==
1117						    rsin->sin_addr.s_addr) {
1118							/* found it */
1119							if (netp != NULL) {
1120								*netp = net;
1121							}
1122							if (locked_tcb == NULL) {
1123								SCTP_INP_DECR_REF(inp);
1124							} else if (locked_tcb != stcb) {
1125								SCTP_TCB_LOCK(locked_tcb);
1126							}
1127							if (locked_tcb) {
1128								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1129							}
1130							SCTP_INP_WUNLOCK(inp);
1131							SCTP_INP_INFO_RUNLOCK();
1132							return (stcb);
1133						}
1134						break;
1135					}
1136#ifdef INET6
1137				case AF_INET6:
1138					{
1139						struct sockaddr_in6 *sin6,
1140						            *rsin6;
1141
1142						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1143						rsin6 = (struct sockaddr_in6 *)remote;
1144						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1145						    rsin6)) {
1146							/* found it */
1147							if (netp != NULL) {
1148								*netp = net;
1149							}
1150							if (locked_tcb == NULL) {
1151								SCTP_INP_DECR_REF(inp);
1152							} else if (locked_tcb != stcb) {
1153								SCTP_TCB_LOCK(locked_tcb);
1154							}
1155							if (locked_tcb) {
1156								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1157							}
1158							SCTP_INP_WUNLOCK(inp);
1159							SCTP_INP_INFO_RUNLOCK();
1160							return (stcb);
1161						}
1162						break;
1163					}
1164#endif
1165				default:
1166					/* TSNH */
1167					break;
1168				}
1169			}
1170			SCTP_TCB_UNLOCK(stcb);
1171		}
1172	} else {
1173		SCTP_INP_WLOCK(inp);
1174		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1175			goto null_return;
1176		}
1177		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1178		    inp->sctp_hashmark)];
1179		if (head == NULL) {
1180			goto null_return;
1181		}
1182		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1183			if (stcb->rport != rport) {
1184				/* remote port does not match */
1185				continue;
1186			}
1187			SCTP_TCB_LOCK(stcb);
1188			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1189				SCTP_TCB_UNLOCK(stcb);
1190				continue;
1191			}
1192			/* now look at the list of remote addresses */
1193			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1194#ifdef INVARIANTS
1195				if (net == (TAILQ_NEXT(net, sctp_next))) {
1196					panic("Corrupt net list");
1197				}
1198#endif
1199				if (net->ro._l_addr.sa.sa_family !=
1200				    remote->sa_family) {
1201					/* not the same family */
1202					continue;
1203				}
1204				switch (remote->sa_family) {
1205				case AF_INET:
1206					{
1207						struct sockaddr_in *sin,
1208						           *rsin;
1209
1210						sin = (struct sockaddr_in *)
1211						    &net->ro._l_addr;
1212						rsin = (struct sockaddr_in *)remote;
1213						if (sin->sin_addr.s_addr ==
1214						    rsin->sin_addr.s_addr) {
1215							/* found it */
1216							if (netp != NULL) {
1217								*netp = net;
1218							}
1219							if (locked_tcb == NULL) {
1220								SCTP_INP_DECR_REF(inp);
1221							} else if (locked_tcb != stcb) {
1222								SCTP_TCB_LOCK(locked_tcb);
1223							}
1224							if (locked_tcb) {
1225								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1226							}
1227							SCTP_INP_WUNLOCK(inp);
1228							SCTP_INP_INFO_RUNLOCK();
1229							return (stcb);
1230						}
1231						break;
1232					}
1233#ifdef INET6
1234				case AF_INET6:
1235					{
1236						struct sockaddr_in6 *sin6,
1237						            *rsin6;
1238
1239						sin6 = (struct sockaddr_in6 *)
1240						    &net->ro._l_addr;
1241						rsin6 = (struct sockaddr_in6 *)remote;
1242						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1243						    rsin6)) {
1244							/* found it */
1245							if (netp != NULL) {
1246								*netp = net;
1247							}
1248							if (locked_tcb == NULL) {
1249								SCTP_INP_DECR_REF(inp);
1250							} else if (locked_tcb != stcb) {
1251								SCTP_TCB_LOCK(locked_tcb);
1252							}
1253							if (locked_tcb) {
1254								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1255							}
1256							SCTP_INP_WUNLOCK(inp);
1257							SCTP_INP_INFO_RUNLOCK();
1258							return (stcb);
1259						}
1260						break;
1261					}
1262#endif
1263				default:
1264					/* TSNH */
1265					break;
1266				}
1267			}
1268			SCTP_TCB_UNLOCK(stcb);
1269		}
1270	}
1271null_return:
1272	/* clean up for returning null */
1273	if (locked_tcb) {
1274		SCTP_TCB_LOCK(locked_tcb);
1275		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1276	}
1277	SCTP_INP_WUNLOCK(inp);
1278	SCTP_INP_INFO_RUNLOCK();
1279	/* not found */
1280	return (NULL);
1281}
1282
1283/*
1284 * Find an association for a specific endpoint using the association id given
1285 * out in the COMM_UP notification
1286 */
1287
1288struct sctp_tcb *
1289sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1290{
1291	/*
1292	 * Use my the assoc_id to find a endpoint
1293	 */
1294	struct sctpasochead *head;
1295	struct sctp_tcb *stcb;
1296	uint32_t id;
1297
1298	if (inp == NULL) {
1299		SCTP_PRINTF("TSNH ep_associd\n");
1300		return (NULL);
1301	}
1302	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1303		SCTP_PRINTF("TSNH ep_associd0\n");
1304		return (NULL);
1305	}
1306	id = (uint32_t) asoc_id;
1307	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1308	if (head == NULL) {
1309		/* invalid id TSNH */
1310		SCTP_PRINTF("TSNH ep_associd1\n");
1311		return (NULL);
1312	}
1313	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1314		if (stcb->asoc.assoc_id == id) {
1315			if (inp != stcb->sctp_ep) {
1316				/*
1317				 * some other guy has the same id active (id
1318				 * collision ??).
1319				 */
1320				SCTP_PRINTF("TSNH ep_associd2\n");
1321				continue;
1322			}
1323			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1324				continue;
1325			}
1326			if (want_lock) {
1327				SCTP_TCB_LOCK(stcb);
1328			}
1329			return (stcb);
1330		}
1331	}
1332	return (NULL);
1333}
1334
1335
1336struct sctp_tcb *
1337sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1338{
1339	struct sctp_tcb *stcb;
1340
1341	SCTP_INP_RLOCK(inp);
1342	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1343	SCTP_INP_RUNLOCK(inp);
1344	return (stcb);
1345}
1346
1347
1348static struct sctp_inpcb *
1349sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1350    uint16_t lport, uint32_t vrf_id)
1351{
1352	struct sctp_inpcb *inp;
1353	struct sockaddr_in *sin;
1354
1355#ifdef INET6
1356	struct sockaddr_in6 *sin6;
1357
1358#endif
1359	struct sctp_laddr *laddr;
1360
1361#ifdef INET6
1362	struct sockaddr_in6 *intf_addr6;
1363
1364#endif
1365
1366	int fnd;
1367
1368	/*
1369	 * Endpoing probe expects that the INP_INFO is locked.
1370	 */
1371	sin = NULL;
1372#ifdef INET6
1373	sin6 = NULL;
1374#endif
1375	switch (nam->sa_family) {
1376	case AF_INET:
1377		sin = (struct sockaddr_in *)nam;
1378		break;
1379#ifdef INET6
1380	case AF_INET6:
1381		sin6 = (struct sockaddr_in6 *)nam;
1382		break;
1383#endif
1384	default:
1385		/* unsupported family */
1386		return (NULL);
1387	}
1388
1389	if (head == NULL)
1390		return (NULL);
1391
1392	LIST_FOREACH(inp, head, sctp_hash) {
1393		SCTP_INP_RLOCK(inp);
1394		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1395			SCTP_INP_RUNLOCK(inp);
1396			continue;
1397		}
1398		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1399		    (inp->sctp_lport == lport)) {
1400			/* got it */
1401			if ((nam->sa_family == AF_INET) &&
1402			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1403			    SCTP_IPV6_V6ONLY(inp)) {
1404				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1405				SCTP_INP_RUNLOCK(inp);
1406				continue;
1407			}
1408			/* A V6 address and the endpoint is NOT bound V6 */
1409			if (nam->sa_family == AF_INET6 &&
1410			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1411				SCTP_INP_RUNLOCK(inp);
1412				continue;
1413			}
1414			/* does a VRF id match? */
1415			fnd = 0;
1416			if (inp->def_vrf_id == vrf_id)
1417				fnd = 1;
1418
1419			SCTP_INP_RUNLOCK(inp);
1420			if (!fnd)
1421				continue;
1422			return (inp);
1423		}
1424		SCTP_INP_RUNLOCK(inp);
1425	}
1426	if ((nam->sa_family == AF_INET) &&
1427	    (sin->sin_addr.s_addr == INADDR_ANY)) {
1428		/* Can't hunt for one that has no address specified */
1429		return (NULL);
1430	}
1431#ifdef INET6
1432	if ((nam->sa_family == AF_INET6) &&
1433	    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
1434		/* Can't hunt for one that has no address specified */
1435		return (NULL);
1436	}
1437#endif
1438	/*
1439	 * ok, not bound to all so see if we can find a EP bound to this
1440	 * address.
1441	 */
1442	LIST_FOREACH(inp, head, sctp_hash) {
1443		SCTP_INP_RLOCK(inp);
1444		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1445			SCTP_INP_RUNLOCK(inp);
1446			continue;
1447		}
1448		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1449			SCTP_INP_RUNLOCK(inp);
1450			continue;
1451		}
1452		/*
1453		 * Ok this could be a likely candidate, look at all of its
1454		 * addresses
1455		 */
1456		if (inp->sctp_lport != lport) {
1457			SCTP_INP_RUNLOCK(inp);
1458			continue;
1459		}
1460		/* does a VRF id match? */
1461		fnd = 0;
1462		if (inp->def_vrf_id == vrf_id)
1463			fnd = 1;
1464
1465		if (!fnd) {
1466			SCTP_INP_RUNLOCK(inp);
1467			continue;
1468		}
1469		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1470			if (laddr->ifa == NULL) {
1471				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1472				    __FUNCTION__);
1473				continue;
1474			}
1475			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1476			    laddr->ifa);
1477			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1478				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1479				continue;
1480			}
1481			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1482				/* possible, see if it matches */
1483				struct sockaddr_in *intf_addr;
1484
1485				intf_addr = &laddr->ifa->address.sin;
1486				switch (nam->sa_family) {
1487				case AF_INET:
1488					if (sin->sin_addr.s_addr ==
1489					    intf_addr->sin_addr.s_addr) {
1490						SCTP_INP_RUNLOCK(inp);
1491						return (inp);
1492					}
1493					break;
1494#ifdef INET6
1495				case AF_INET6:
1496					intf_addr6 = &laddr->ifa->address.sin6;
1497					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1498					    intf_addr6)) {
1499						SCTP_INP_RUNLOCK(inp);
1500						return (inp);
1501					}
1502					break;
1503#endif
1504				}
1505			}
1506		}
1507		SCTP_INP_RUNLOCK(inp);
1508	}
1509	return (NULL);
1510}
1511
1512
1513static struct sctp_inpcb *
1514sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1515{
1516	struct sctppcbhead *head;
1517	struct sctp_inpcb *t_inp;
1518	int fnd;
1519
1520	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1521	    SCTP_BASE_INFO(hashmark))];
1522	LIST_FOREACH(t_inp, head, sctp_hash) {
1523		if (t_inp->sctp_lport != lport) {
1524			continue;
1525		}
1526		/* is it in the VRF in question */
1527		fnd = 0;
1528		if (t_inp->def_vrf_id == vrf_id)
1529			fnd = 1;
1530		if (!fnd)
1531			continue;
1532
1533		/* This one is in use. */
1534		/* check the v6/v4 binding issue */
1535		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1536		    SCTP_IPV6_V6ONLY(t_inp)) {
1537			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1538				/* collision in V6 space */
1539				return (t_inp);
1540			} else {
1541				/* inp is BOUND_V4 no conflict */
1542				continue;
1543			}
1544		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1545			/* t_inp is bound v4 and v6, conflict always */
1546			return (t_inp);
1547		} else {
1548			/* t_inp is bound only V4 */
1549			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1550			    SCTP_IPV6_V6ONLY(inp)) {
1551				/* no conflict */
1552				continue;
1553			}
1554			/* else fall through to conflict */
1555		}
1556		return (t_inp);
1557	}
1558	return (NULL);
1559}
1560
1561
1562int
1563sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1564{
1565	/* For 1-2-1 with port reuse */
1566	struct sctppcbhead *head;
1567	struct sctp_inpcb *tinp;
1568
1569	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1570		/* only works with port reuse on */
1571		return (-1);
1572	}
1573	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1574		return (0);
1575	}
1576	SCTP_INP_RUNLOCK(inp);
1577	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1578	    SCTP_BASE_INFO(hashmark))];
1579	/* Kick out all non-listeners to the TCP hash */
1580	LIST_FOREACH(tinp, head, sctp_hash) {
1581		if (tinp->sctp_lport != inp->sctp_lport) {
1582			continue;
1583		}
1584		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1585			continue;
1586		}
1587		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1588			continue;
1589		}
1590		if (tinp->sctp_socket->so_qlimit) {
1591			continue;
1592		}
1593		SCTP_INP_WLOCK(tinp);
1594		LIST_REMOVE(tinp, sctp_hash);
1595		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1596		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1597		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1598		SCTP_INP_WUNLOCK(tinp);
1599	}
1600	SCTP_INP_WLOCK(inp);
1601	/* Pull from where he was */
1602	LIST_REMOVE(inp, sctp_hash);
1603	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1604	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1605	LIST_INSERT_HEAD(head, inp, sctp_hash);
1606	SCTP_INP_WUNLOCK(inp);
1607	SCTP_INP_RLOCK(inp);
1608	return (0);
1609}
1610
1611
1612struct sctp_inpcb *
1613sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1614    uint32_t vrf_id)
1615{
1616	/*
1617	 * First we check the hash table to see if someone has this port
1618	 * bound with just the port.
1619	 */
1620	struct sctp_inpcb *inp;
1621	struct sctppcbhead *head;
1622	struct sockaddr_in *sin;
1623	struct sockaddr_in6 *sin6;
1624	int lport;
1625
1626	if (nam->sa_family == AF_INET) {
1627		sin = (struct sockaddr_in *)nam;
1628		lport = ((struct sockaddr_in *)nam)->sin_port;
1629	} else if (nam->sa_family == AF_INET6) {
1630		sin6 = (struct sockaddr_in6 *)nam;
1631		lport = ((struct sockaddr_in6 *)nam)->sin6_port;
1632	} else {
1633		/* unsupported family */
1634		return (NULL);
1635	}
1636	/*
1637	 * I could cheat here and just cast to one of the types but we will
1638	 * do it right. It also provides the check against an Unsupported
1639	 * type too.
1640	 */
1641	/* Find the head of the ALLADDR chain */
1642	if (have_lock == 0) {
1643		SCTP_INP_INFO_RLOCK();
1644	}
1645	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1646	    SCTP_BASE_INFO(hashmark))];
1647	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1648
1649	/*
1650	 * If the TCP model exists it could be that the main listening
1651	 * endpoint is gone but there exists a connected socket for this guy
1652	 * yet. If so we can return the first one that we find. This may NOT
1653	 * be the correct one so the caller should be wary on the return
1654	 * INP. Currently the onlyc caller that sets this flag is in bindx
1655	 * where we are verifying that a user CAN bind the address. He
1656	 * either has bound it already, or someone else has, or its open to
1657	 * bind, so this is good enough.
1658	 */
1659	if (inp == NULL && find_tcp_pool) {
1660		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
1661		inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1662	}
1663	if (inp) {
1664		SCTP_INP_INCR_REF(inp);
1665	}
1666	if (have_lock == 0) {
1667		SCTP_INP_INFO_RUNLOCK();
1668	}
1669	return (inp);
1670}
1671
1672/*
1673 * Find an association for an endpoint with the pointer to whom you want to
1674 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1675 * need to change the *to to some other struct like a mbuf...
1676 */
1677struct sctp_tcb *
1678sctp_findassociation_addr_sa(struct sockaddr *to, struct sockaddr *from,
1679    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1680    uint32_t vrf_id)
1681{
1682	struct sctp_inpcb *inp = NULL;
1683	struct sctp_tcb *retval;
1684
1685	SCTP_INP_INFO_RLOCK();
1686	if (find_tcp_pool) {
1687		if (inp_p != NULL) {
1688			retval = sctp_tcb_special_locate(inp_p, from, to, netp,
1689			    vrf_id);
1690		} else {
1691			retval = sctp_tcb_special_locate(&inp, from, to, netp,
1692			    vrf_id);
1693		}
1694		if (retval != NULL) {
1695			SCTP_INP_INFO_RUNLOCK();
1696			return (retval);
1697		}
1698	}
1699	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1700	if (inp_p != NULL) {
1701		*inp_p = inp;
1702	}
1703	SCTP_INP_INFO_RUNLOCK();
1704
1705	if (inp == NULL) {
1706		return (NULL);
1707	}
1708	/*
1709	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1710	 * we now place the source address or from in the to of the find
1711	 * endpoint call. Since in reality this chain is used from the
1712	 * inbound packet side.
1713	 */
1714	if (inp_p != NULL) {
1715		retval = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1716		    NULL);
1717	} else {
1718		retval = sctp_findassociation_ep_addr(&inp, from, netp, to,
1719		    NULL);
1720	}
1721	return retval;
1722}
1723
1724
1725/*
1726 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1727 * find all addresses that the sender has specified in any address list. Each
1728 * address will be used to lookup the TCB and see if one exits.
1729 */
1730static struct sctp_tcb *
1731sctp_findassociation_special_addr(struct mbuf *m, int iphlen, int offset,
1732    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1733    struct sockaddr *dest)
1734{
1735	struct sockaddr_in sin4;
1736	struct sockaddr_in6 sin6;
1737	struct sctp_paramhdr *phdr, parm_buf;
1738	struct sctp_tcb *retval;
1739	uint32_t ptype, plen;
1740
1741	memset(&sin4, 0, sizeof(sin4));
1742	memset(&sin6, 0, sizeof(sin6));
1743	sin4.sin_len = sizeof(sin4);
1744	sin4.sin_family = AF_INET;
1745	sin4.sin_port = sh->src_port;
1746	sin6.sin6_len = sizeof(sin6);
1747	sin6.sin6_family = AF_INET6;
1748	sin6.sin6_port = sh->src_port;
1749
1750	retval = NULL;
1751	offset += sizeof(struct sctp_init_chunk);
1752
1753	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
1754	while (phdr != NULL) {
1755		/* now we must see if we want the parameter */
1756		ptype = ntohs(phdr->param_type);
1757		plen = ntohs(phdr->param_length);
1758		if (plen == 0) {
1759			break;
1760		}
1761		if (ptype == SCTP_IPV4_ADDRESS &&
1762		    plen == sizeof(struct sctp_ipv4addr_param)) {
1763			/* Get the rest of the address */
1764			struct sctp_ipv4addr_param ip4_parm, *p4;
1765
1766			phdr = sctp_get_next_param(m, offset,
1767			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
1768			if (phdr == NULL) {
1769				return (NULL);
1770			}
1771			p4 = (struct sctp_ipv4addr_param *)phdr;
1772			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
1773			/* look it up */
1774			retval = sctp_findassociation_ep_addr(inp_p,
1775			    (struct sockaddr *)&sin4, netp, dest, NULL);
1776			if (retval != NULL) {
1777				return (retval);
1778			}
1779		} else if (ptype == SCTP_IPV6_ADDRESS &&
1780		    plen == sizeof(struct sctp_ipv6addr_param)) {
1781			/* Get the rest of the address */
1782			struct sctp_ipv6addr_param ip6_parm, *p6;
1783
1784			phdr = sctp_get_next_param(m, offset,
1785			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
1786			if (phdr == NULL) {
1787				return (NULL);
1788			}
1789			p6 = (struct sctp_ipv6addr_param *)phdr;
1790			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
1791			/* look it up */
1792			retval = sctp_findassociation_ep_addr(inp_p,
1793			    (struct sockaddr *)&sin6, netp, dest, NULL);
1794			if (retval != NULL) {
1795				return (retval);
1796			}
1797		}
1798		offset += SCTP_SIZE32(plen);
1799		phdr = sctp_get_next_param(m, offset, &parm_buf,
1800		    sizeof(parm_buf));
1801	}
1802	return (NULL);
1803}
1804
1805static int
1806sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
1807{
1808	struct sctp_nets *net;
1809
1810	/*
1811	 * Simple question, the ports match, does the tcb own the to
1812	 * address?
1813	 */
1814	if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1815		/* of course */
1816		return (1);
1817	}
1818	/* have to look at all bound addresses */
1819	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1820		if (net->ro._l_addr.sa.sa_family != to->sa_family) {
1821			/* not the same family, can't be a match */
1822			continue;
1823		}
1824		switch (to->sa_family) {
1825		case AF_INET:
1826			{
1827				struct sockaddr_in *sin, *rsin;
1828
1829				sin = (struct sockaddr_in *)&net->ro._l_addr;
1830				rsin = (struct sockaddr_in *)to;
1831				if (sin->sin_addr.s_addr ==
1832				    rsin->sin_addr.s_addr) {
1833					/* found it */
1834					return (1);
1835				}
1836				break;
1837			}
1838#ifdef INET6
1839		case AF_INET6:
1840			{
1841				struct sockaddr_in6 *sin6, *rsin6;
1842
1843				sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1844				rsin6 = (struct sockaddr_in6 *)to;
1845				if (SCTP6_ARE_ADDR_EQUAL(sin6,
1846				    rsin6)) {
1847					/* Update the endpoint pointer */
1848					return (1);
1849				}
1850				break;
1851			}
1852#endif
1853		default:
1854			/* TSNH */
1855			break;
1856		}
1857	}
1858	/* Nope, do not have the address ;-( */
1859	return (0);
1860}
1861
1862static struct sctp_tcb *
1863sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
1864    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
1865    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
1866{
1867	/*
1868	 * Use my vtag to hash. If we find it we then verify the source addr
1869	 * is in the assoc. If all goes well we save a bit on rec of a
1870	 * packet.
1871	 */
1872	struct sctpasochead *head;
1873	struct sctp_nets *net;
1874	struct sctp_tcb *stcb;
1875
1876	*netp = NULL;
1877	*inp_p = NULL;
1878	SCTP_INP_INFO_RLOCK();
1879	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
1880	    SCTP_BASE_INFO(hashasocmark))];
1881	if (head == NULL) {
1882		/* invalid vtag */
1883		SCTP_INP_INFO_RUNLOCK();
1884		return (NULL);
1885	}
1886	LIST_FOREACH(stcb, head, sctp_asocs) {
1887		SCTP_INP_RLOCK(stcb->sctp_ep);
1888		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1889			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1890			continue;
1891		}
1892		SCTP_TCB_LOCK(stcb);
1893		SCTP_INP_RUNLOCK(stcb->sctp_ep);
1894		if (stcb->asoc.my_vtag == vtag) {
1895			/* candidate */
1896			if (stcb->rport != rport) {
1897				SCTP_TCB_UNLOCK(stcb);
1898				continue;
1899			}
1900			if (stcb->sctp_ep->sctp_lport != lport) {
1901				SCTP_TCB_UNLOCK(stcb);
1902				continue;
1903			}
1904			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1905				SCTP_TCB_UNLOCK(stcb);
1906				continue;
1907			}
1908			/* RRS:Need toaddr check here */
1909			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
1910				/* Endpoint does not own this address */
1911				SCTP_TCB_UNLOCK(stcb);
1912				continue;
1913			}
1914			if (remote_tag) {
1915				/*
1916				 * If we have both vtags thats all we match
1917				 * on
1918				 */
1919				if (stcb->asoc.peer_vtag == remote_tag) {
1920					/*
1921					 * If both tags match we consider it
1922					 * conclusive and check NO
1923					 * source/destination addresses
1924					 */
1925					goto conclusive;
1926				}
1927			}
1928			if (skip_src_check) {
1929		conclusive:
1930				if (from) {
1931					net = sctp_findnet(stcb, from);
1932				} else {
1933					*netp = NULL;	/* unknown */
1934				}
1935				if (inp_p)
1936					*inp_p = stcb->sctp_ep;
1937				SCTP_INP_INFO_RUNLOCK();
1938				return (stcb);
1939			}
1940			net = sctp_findnet(stcb, from);
1941			if (net) {
1942				/* yep its him. */
1943				*netp = net;
1944				SCTP_STAT_INCR(sctps_vtagexpress);
1945				*inp_p = stcb->sctp_ep;
1946				SCTP_INP_INFO_RUNLOCK();
1947				return (stcb);
1948			} else {
1949				/*
1950				 * not him, this should only happen in rare
1951				 * cases so I peg it.
1952				 */
1953				SCTP_STAT_INCR(sctps_vtagbogus);
1954			}
1955		}
1956		SCTP_TCB_UNLOCK(stcb);
1957	}
1958	SCTP_INP_INFO_RUNLOCK();
1959	return (NULL);
1960}
1961
1962/*
1963 * Find an association with the pointer to the inbound IP packet. This can be
1964 * a IPv4 or IPv6 packet.
1965 */
1966struct sctp_tcb *
1967sctp_findassociation_addr(struct mbuf *m, int iphlen, int offset,
1968    struct sctphdr *sh, struct sctp_chunkhdr *ch,
1969    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
1970{
1971	int find_tcp_pool;
1972	struct ip *iph;
1973	struct sctp_tcb *retval;
1974	struct sockaddr_storage to_store, from_store;
1975	struct sockaddr *to = (struct sockaddr *)&to_store;
1976	struct sockaddr *from = (struct sockaddr *)&from_store;
1977	struct sctp_inpcb *inp;
1978
1979	iph = mtod(m, struct ip *);
1980	switch (iph->ip_v) {
1981	case IPVERSION:
1982		{
1983			/* its IPv4 */
1984			struct sockaddr_in *from4;
1985
1986			from4 = (struct sockaddr_in *)&from_store;
1987			bzero(from4, sizeof(*from4));
1988			from4->sin_family = AF_INET;
1989			from4->sin_len = sizeof(struct sockaddr_in);
1990			from4->sin_addr.s_addr = iph->ip_src.s_addr;
1991			from4->sin_port = sh->src_port;
1992			break;
1993		}
1994#ifdef INET6
1995	case IPV6_VERSION >> 4:
1996		{
1997			/* its IPv6 */
1998			struct ip6_hdr *ip6;
1999			struct sockaddr_in6 *from6;
2000
2001			ip6 = mtod(m, struct ip6_hdr *);
2002			from6 = (struct sockaddr_in6 *)&from_store;
2003			bzero(from6, sizeof(*from6));
2004			from6->sin6_family = AF_INET6;
2005			from6->sin6_len = sizeof(struct sockaddr_in6);
2006			from6->sin6_addr = ip6->ip6_src;
2007			from6->sin6_port = sh->src_port;
2008			/* Get the scopes in properly to the sin6 addr's */
2009			/* we probably don't need these operations */
2010			(void)sa6_recoverscope(from6);
2011			sa6_embedscope(from6, MODULE_GLOBAL(ip6_use_defzone));
2012			break;
2013		}
2014#endif
2015	default:
2016		/* Currently not supported. */
2017		return (NULL);
2018	}
2019
2020
2021	switch (iph->ip_v) {
2022	case IPVERSION:
2023		{
2024			/* its IPv4 */
2025			struct sockaddr_in *to4;
2026
2027			to4 = (struct sockaddr_in *)&to_store;
2028			bzero(to4, sizeof(*to4));
2029			to4->sin_family = AF_INET;
2030			to4->sin_len = sizeof(struct sockaddr_in);
2031			to4->sin_addr.s_addr = iph->ip_dst.s_addr;
2032			to4->sin_port = sh->dest_port;
2033			break;
2034		}
2035#ifdef INET6
2036	case IPV6_VERSION >> 4:
2037		{
2038			/* its IPv6 */
2039			struct ip6_hdr *ip6;
2040			struct sockaddr_in6 *to6;
2041
2042			ip6 = mtod(m, struct ip6_hdr *);
2043			to6 = (struct sockaddr_in6 *)&to_store;
2044			bzero(to6, sizeof(*to6));
2045			to6->sin6_family = AF_INET6;
2046			to6->sin6_len = sizeof(struct sockaddr_in6);
2047			to6->sin6_addr = ip6->ip6_dst;
2048			to6->sin6_port = sh->dest_port;
2049			/* Get the scopes in properly to the sin6 addr's */
2050			/* we probably don't need these operations */
2051			(void)sa6_recoverscope(to6);
2052			sa6_embedscope(to6, MODULE_GLOBAL(ip6_use_defzone));
2053			break;
2054		}
2055#endif
2056	default:
2057		/* TSNH */
2058		break;
2059	}
2060	if (sh->v_tag) {
2061		/* we only go down this path if vtag is non-zero */
2062		retval = sctp_findassoc_by_vtag(from, to, ntohl(sh->v_tag),
2063		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2064		if (retval) {
2065			return (retval);
2066		}
2067	}
2068	find_tcp_pool = 0;
2069	if ((ch->chunk_type != SCTP_INITIATION) &&
2070	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2071	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2072	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2073		/* Other chunk types go to the tcp pool. */
2074		find_tcp_pool = 1;
2075	}
2076	if (inp_p) {
2077		retval = sctp_findassociation_addr_sa(to, from, inp_p, netp,
2078		    find_tcp_pool, vrf_id);
2079		inp = *inp_p;
2080	} else {
2081		retval = sctp_findassociation_addr_sa(to, from, &inp, netp,
2082		    find_tcp_pool, vrf_id);
2083	}
2084	SCTPDBG(SCTP_DEBUG_PCB1, "retval:%p inp:%p\n", retval, inp);
2085	if (retval == NULL && inp) {
2086		/* Found a EP but not this address */
2087		if ((ch->chunk_type == SCTP_INITIATION) ||
2088		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2089			/*-
2090			 * special hook, we do NOT return linp or an
2091			 * association that is linked to an existing
2092			 * association that is under the TCP pool (i.e. no
2093			 * listener exists). The endpoint finding routine
2094			 * will always find a listner before examining the
2095			 * TCP pool.
2096			 */
2097			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2098				if (inp_p) {
2099					*inp_p = NULL;
2100				}
2101				return (NULL);
2102			}
2103			retval = sctp_findassociation_special_addr(m, iphlen,
2104			    offset, sh, &inp, netp, to);
2105			if (inp_p != NULL) {
2106				*inp_p = inp;
2107			}
2108		}
2109	}
2110	SCTPDBG(SCTP_DEBUG_PCB1, "retval is %p\n", retval);
2111	return (retval);
2112}
2113
2114/*
2115 * lookup an association by an ASCONF lookup address.
2116 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2117 */
2118struct sctp_tcb *
2119sctp_findassociation_ep_asconf(struct mbuf *m, int iphlen, int offset,
2120    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2121{
2122	struct sctp_tcb *stcb;
2123	struct sockaddr_in *sin;
2124
2125#ifdef INET6
2126	struct sockaddr_in6 *sin6;
2127
2128#endif
2129	struct sockaddr_storage local_store, remote_store;
2130	struct sockaddr *to;
2131	struct ip *iph;
2132
2133#ifdef INET6
2134	struct ip6_hdr *ip6;
2135
2136#endif
2137	struct sctp_paramhdr parm_buf, *phdr;
2138	int ptype;
2139	int zero_address = 0;
2140
2141
2142	memset(&local_store, 0, sizeof(local_store));
2143	memset(&remote_store, 0, sizeof(remote_store));
2144	to = (struct sockaddr *)&local_store;
2145	/* First get the destination address setup too. */
2146	iph = mtod(m, struct ip *);
2147	switch (iph->ip_v) {
2148	case IPVERSION:
2149		/* its IPv4 */
2150		sin = (struct sockaddr_in *)&local_store;
2151		sin->sin_family = AF_INET;
2152		sin->sin_len = sizeof(*sin);
2153		sin->sin_port = sh->dest_port;
2154		sin->sin_addr.s_addr = iph->ip_dst.s_addr;
2155		break;
2156#ifdef INET6
2157	case IPV6_VERSION >> 4:
2158		/* its IPv6 */
2159		ip6 = mtod(m, struct ip6_hdr *);
2160		sin6 = (struct sockaddr_in6 *)&local_store;
2161		sin6->sin6_family = AF_INET6;
2162		sin6->sin6_len = sizeof(*sin6);
2163		sin6->sin6_port = sh->dest_port;
2164		sin6->sin6_addr = ip6->ip6_dst;
2165		break;
2166#endif
2167	default:
2168		return NULL;
2169	}
2170
2171	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2172	    &parm_buf, sizeof(struct sctp_paramhdr));
2173	if (phdr == NULL) {
2174		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2175		    __FUNCTION__);
2176		return NULL;
2177	}
2178	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2179	/* get the correlation address */
2180	switch (ptype) {
2181#ifdef INET6
2182	case SCTP_IPV6_ADDRESS:
2183		{
2184			/* ipv6 address param */
2185			struct sctp_ipv6addr_param *p6, p6_buf;
2186
2187			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2188				return NULL;
2189			}
2190			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2191			    offset + sizeof(struct sctp_asconf_chunk),
2192			    &p6_buf.ph, sizeof(*p6));
2193			if (p6 == NULL) {
2194				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2195				    __FUNCTION__);
2196				return (NULL);
2197			}
2198			sin6 = (struct sockaddr_in6 *)&remote_store;
2199			sin6->sin6_family = AF_INET6;
2200			sin6->sin6_len = sizeof(*sin6);
2201			sin6->sin6_port = sh->src_port;
2202			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2203			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2204				zero_address = 1;
2205			break;
2206		}
2207#endif
2208	case SCTP_IPV4_ADDRESS:
2209		{
2210			/* ipv4 address param */
2211			struct sctp_ipv4addr_param *p4, p4_buf;
2212
2213			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2214				return NULL;
2215			}
2216			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2217			    offset + sizeof(struct sctp_asconf_chunk),
2218			    &p4_buf.ph, sizeof(*p4));
2219			if (p4 == NULL) {
2220				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2221				    __FUNCTION__);
2222				return (NULL);
2223			}
2224			sin = (struct sockaddr_in *)&remote_store;
2225			sin->sin_family = AF_INET;
2226			sin->sin_len = sizeof(*sin);
2227			sin->sin_port = sh->src_port;
2228			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2229			if (sin->sin_addr.s_addr == INADDR_ANY)
2230				zero_address = 1;
2231			break;
2232		}
2233	default:
2234		/* invalid address param type */
2235		return NULL;
2236	}
2237
2238	if (zero_address) {
2239		stcb = sctp_findassoc_by_vtag(NULL, to, ntohl(sh->v_tag), inp_p,
2240		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2241		/*
2242		 * printf("findassociation_ep_asconf: zero lookup address
2243		 * finds stcb 0x%x\n", (uint32_t)stcb);
2244		 */
2245	} else {
2246		stcb = sctp_findassociation_ep_addr(inp_p,
2247		    (struct sockaddr *)&remote_store, netp,
2248		    to, NULL);
2249	}
2250	return (stcb);
2251}
2252
2253
2254/*
2255 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2256 * addresses. This way if we don't get a bind we by default pick a ephemeral
2257 * port with all addresses bound.
2258 */
2259int
2260sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2261{
2262	/*
2263	 * we get called when a new endpoint starts up. We need to allocate
2264	 * the sctp_inpcb structure from the zone and init it. Mark it as
2265	 * unbound and find a port that we can use as an ephemeral with
2266	 * INADDR_ANY. If the user binds later no problem we can then add in
2267	 * the specific addresses. And setup the default parameters for the
2268	 * EP.
2269	 */
2270	int i, error;
2271	struct sctp_inpcb *inp;
2272	struct sctp_pcb *m;
2273	struct timeval time;
2274	sctp_sharedkey_t *null_key;
2275
2276	error = 0;
2277
2278	SCTP_INP_INFO_WLOCK();
2279	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2280	if (inp == NULL) {
2281		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2282		SCTP_INP_INFO_WUNLOCK();
2283		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2284		return (ENOBUFS);
2285	}
2286	/* zap it */
2287	bzero(inp, sizeof(*inp));
2288
2289	/* bump generations */
2290	/* setup socket pointers */
2291	inp->sctp_socket = so;
2292	inp->ip_inp.inp.inp_socket = so;
2293	inp->sctp_associd_counter = 1;
2294	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2295	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2296	/* init the small hash table we use to track asocid <-> tcb */
2297	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2298	if (inp->sctp_asocidhash == NULL) {
2299		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2300		SCTP_INP_INFO_WUNLOCK();
2301		return error;
2302	}
2303#ifdef IPSEC
2304	{
2305		struct inpcbpolicy *pcb_sp = NULL;
2306
2307		error = ipsec_init_policy(so, &pcb_sp);
2308		/* Arrange to share the policy */
2309		inp->ip_inp.inp.inp_sp = pcb_sp;
2310		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2311	}
2312	if (error != 0) {
2313		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2314		SCTP_INP_INFO_WUNLOCK();
2315		return error;
2316	}
2317#endif				/* IPSEC */
2318	SCTP_INCR_EP_COUNT();
2319	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2320	SCTP_INP_INFO_WUNLOCK();
2321
2322	so->so_pcb = (caddr_t)inp;
2323
2324	if ((SCTP_SO_TYPE(so) == SOCK_DGRAM) ||
2325	    (SCTP_SO_TYPE(so) == SOCK_SEQPACKET)) {
2326		/* UDP style socket */
2327		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2328		    SCTP_PCB_FLAGS_UNBOUND);
2329		/* Be sure it is NON-BLOCKING IO for UDP */
2330		/* SCTP_SET_SO_NBIO(so); */
2331	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2332		/* TCP style socket */
2333		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2334		    SCTP_PCB_FLAGS_UNBOUND);
2335		/* Be sure we have blocking IO by default */
2336		SCTP_CLEAR_SO_NBIO(so);
2337	} else {
2338		/*
2339		 * unsupported socket type (RAW, etc)- in case we missed it
2340		 * in protosw
2341		 */
2342		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2343		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2344		return (EOPNOTSUPP);
2345	}
2346	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2347		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2348		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2349	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2350		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2351		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2352	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2353		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2354		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2355	}
2356	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2357	    &inp->sctp_hashmark);
2358	if (inp->sctp_tcbhash == NULL) {
2359		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2360		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2361		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2362		return (ENOBUFS);
2363	}
2364	inp->def_vrf_id = vrf_id;
2365
2366	SCTP_INP_INFO_WLOCK();
2367	SCTP_INP_LOCK_INIT(inp);
2368	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2369	SCTP_INP_READ_INIT(inp);
2370	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2371	/* lock the new ep */
2372	SCTP_INP_WLOCK(inp);
2373
2374	/* add it to the info area */
2375	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2376	SCTP_INP_INFO_WUNLOCK();
2377
2378	TAILQ_INIT(&inp->read_queue);
2379	LIST_INIT(&inp->sctp_addr_list);
2380
2381	LIST_INIT(&inp->sctp_asoc_list);
2382
2383#ifdef SCTP_TRACK_FREED_ASOCS
2384	/* TEMP CODE */
2385	LIST_INIT(&inp->sctp_asoc_free_list);
2386#endif
2387	/* Init the timer structure for signature change */
2388	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2389	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2390
2391	/* now init the actual endpoint default data */
2392	m = &inp->sctp_ep;
2393
2394	/* setup the base timeout information */
2395	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2396	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2397	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2398	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2399	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2400	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2401	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2402	/* all max/min max are in ms */
2403	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2404	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2405	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2406	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2407	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2408
2409	m->max_open_streams_intome = MAX_SCTP_STREAMS;
2410
2411	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2412	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2413	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2414	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2415	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2416	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2417	if ((SCTP_BASE_SYSCTL(sctp_default_cc_module) >= SCTP_CC_RFC2581) &&
2418	    (SCTP_BASE_SYSCTL(sctp_default_cc_module) <= SCTP_CC_HTCP)) {
2419		m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2420	} else {
2421		/* sysctl done with invalid value, set to 2581 */
2422		m->sctp_default_cc_module = SCTP_CC_RFC2581;
2423	}
2424	/* number of streams to pre-open on a association */
2425	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2426
2427	/* Add adaptation cookie */
2428	m->adaptation_layer_indicator = 0x504C5253;
2429
2430	/* seed random number generator */
2431	m->random_counter = 1;
2432	m->store_at = SCTP_SIGNATURE_SIZE;
2433	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2434	sctp_fill_random_store(m);
2435
2436	/* Minimum cookie size */
2437	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2438	    sizeof(struct sctp_state_cookie);
2439	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2440
2441	/* Setup the initial secret */
2442	(void)SCTP_GETTIME_TIMEVAL(&time);
2443	m->time_of_secret_change = time.tv_sec;
2444
2445	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2446		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2447	}
2448	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2449
2450	/* How long is a cookie good for ? */
2451	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2452	/*
2453	 * Initialize authentication parameters
2454	 */
2455	m->local_hmacs = sctp_default_supported_hmaclist();
2456	m->local_auth_chunks = sctp_alloc_chunklist();
2457	sctp_auth_set_default_chunks(m->local_auth_chunks);
2458	LIST_INIT(&m->shared_keys);
2459	/* add default NULL key as key id 0 */
2460	null_key = sctp_alloc_sharedkey();
2461	sctp_insert_sharedkey(&m->shared_keys, null_key);
2462	SCTP_INP_WUNLOCK(inp);
2463#ifdef SCTP_LOG_CLOSING
2464	sctp_log_closing(inp, NULL, 12);
2465#endif
2466	return (error);
2467}
2468
2469
2470void
2471sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2472    struct sctp_tcb *stcb)
2473{
2474	struct sctp_nets *net;
2475	uint16_t lport, rport;
2476	struct sctppcbhead *head;
2477	struct sctp_laddr *laddr, *oladdr;
2478
2479	atomic_add_int(&stcb->asoc.refcnt, 1);
2480	SCTP_TCB_UNLOCK(stcb);
2481	SCTP_INP_INFO_WLOCK();
2482	SCTP_INP_WLOCK(old_inp);
2483	SCTP_INP_WLOCK(new_inp);
2484	SCTP_TCB_LOCK(stcb);
2485	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2486
2487	new_inp->sctp_ep.time_of_secret_change =
2488	    old_inp->sctp_ep.time_of_secret_change;
2489	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2490	    sizeof(old_inp->sctp_ep.secret_key));
2491	new_inp->sctp_ep.current_secret_number =
2492	    old_inp->sctp_ep.current_secret_number;
2493	new_inp->sctp_ep.last_secret_number =
2494	    old_inp->sctp_ep.last_secret_number;
2495	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2496
2497	/* make it so new data pours into the new socket */
2498	stcb->sctp_socket = new_inp->sctp_socket;
2499	stcb->sctp_ep = new_inp;
2500
2501	/* Copy the port across */
2502	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2503	rport = stcb->rport;
2504	/* Pull the tcb from the old association */
2505	LIST_REMOVE(stcb, sctp_tcbhash);
2506	LIST_REMOVE(stcb, sctp_tcblist);
2507	if (stcb->asoc.in_asocid_hash) {
2508		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2509	}
2510	/* Now insert the new_inp into the TCP connected hash */
2511	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport),
2512	    SCTP_BASE_INFO(hashtcpmark))];
2513
2514	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2515	/* Its safe to access */
2516	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2517
2518	/* Now move the tcb into the endpoint list */
2519	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2520	/*
2521	 * Question, do we even need to worry about the ep-hash since we
2522	 * only have one connection? Probably not :> so lets get rid of it
2523	 * and not suck up any kernel memory in that.
2524	 */
2525	if (stcb->asoc.in_asocid_hash) {
2526		struct sctpasochead *lhd;
2527
2528		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2529		    new_inp->hashasocidmark)];
2530		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2531	}
2532	/* Ok. Let's restart timer. */
2533	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2534		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2535		    stcb, net);
2536	}
2537
2538	SCTP_INP_INFO_WUNLOCK();
2539	if (new_inp->sctp_tcbhash != NULL) {
2540		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2541		new_inp->sctp_tcbhash = NULL;
2542	}
2543	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2544		/* Subset bound, so copy in the laddr list from the old_inp */
2545		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2546			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2547			if (laddr == NULL) {
2548				/*
2549				 * Gak, what can we do? This assoc is really
2550				 * HOSED. We probably should send an abort
2551				 * here.
2552				 */
2553				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2554				continue;
2555			}
2556			SCTP_INCR_LADDR_COUNT();
2557			bzero(laddr, sizeof(*laddr));
2558			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2559			laddr->ifa = oladdr->ifa;
2560			atomic_add_int(&laddr->ifa->refcount, 1);
2561			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2562			    sctp_nxt_addr);
2563			new_inp->laddr_count++;
2564		}
2565	}
2566	/*
2567	 * Now any running timers need to be adjusted since we really don't
2568	 * care if they are running or not just blast in the new_inp into
2569	 * all of them.
2570	 */
2571
2572	stcb->asoc.hb_timer.ep = (void *)new_inp;
2573	stcb->asoc.dack_timer.ep = (void *)new_inp;
2574	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2575	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2576	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2577	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2578	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2579	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2580	/* now what about the nets? */
2581	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2582		net->pmtu_timer.ep = (void *)new_inp;
2583		net->rxt_timer.ep = (void *)new_inp;
2584		net->fr_timer.ep = (void *)new_inp;
2585	}
2586	SCTP_INP_WUNLOCK(new_inp);
2587	SCTP_INP_WUNLOCK(old_inp);
2588}
2589
2590
2591
2592
2593/* sctp_ifap is used to bypass normal local address validation checks */
2594int
2595sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2596    struct sctp_ifa *sctp_ifap, struct thread *p)
2597{
2598	/* bind a ep to a socket address */
2599	struct sctppcbhead *head;
2600	struct sctp_inpcb *inp, *inp_tmp;
2601	struct inpcb *ip_inp;
2602	int port_reuse_active = 0;
2603	int bindall;
2604	uint16_t lport;
2605	int error;
2606	uint32_t vrf_id;
2607
2608	lport = 0;
2609	error = 0;
2610	bindall = 1;
2611	inp = (struct sctp_inpcb *)so->so_pcb;
2612	ip_inp = (struct inpcb *)so->so_pcb;
2613#ifdef SCTP_DEBUG
2614	if (addr) {
2615		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port:%d\n",
2616		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2617		SCTPDBG(SCTP_DEBUG_PCB1, "Addr :");
2618		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2619	}
2620#endif
2621	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2622		/* already did a bind, subsequent binds NOT allowed ! */
2623		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2624		return (EINVAL);
2625	}
2626#ifdef INVARIANTS
2627	if (p == NULL)
2628		panic("null proc/thread");
2629#endif
2630	if (addr != NULL) {
2631		switch (addr->sa_family) {
2632		case AF_INET:
2633			{
2634				struct sockaddr_in *sin;
2635
2636				/* IPV6_V6ONLY socket? */
2637				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2638					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2639					return (EINVAL);
2640				}
2641				if (addr->sa_len != sizeof(*sin)) {
2642					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2643					return (EINVAL);
2644				}
2645				sin = (struct sockaddr_in *)addr;
2646				lport = sin->sin_port;
2647				/*
2648				 * For LOOPBACK the prison_local_ip4() call
2649				 * will transmute the ip address to the
2650				 * proper value.
2651				 */
2652				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2653					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2654					return (error);
2655				}
2656				if (sin->sin_addr.s_addr != INADDR_ANY) {
2657					bindall = 0;
2658				}
2659				break;
2660			}
2661#ifdef INET6
2662		case AF_INET6:
2663			{
2664				/*
2665				 * Only for pure IPv6 Address. (No IPv4
2666				 * Mapped!)
2667				 */
2668				struct sockaddr_in6 *sin6;
2669
2670				sin6 = (struct sockaddr_in6 *)addr;
2671
2672				if (addr->sa_len != sizeof(*sin6)) {
2673					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2674					return (EINVAL);
2675				}
2676				lport = sin6->sin6_port;
2677
2678				/*
2679				 * For LOOPBACK the prison_local_ip6() call
2680				 * will transmute the ipv6 address to the
2681				 * proper value.
2682				 */
2683				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2684				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2685					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2686					return (error);
2687				}
2688				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2689					bindall = 0;
2690					/* KAME hack: embed scopeid */
2691					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2692						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2693						return (EINVAL);
2694					}
2695				}
2696				/* this must be cleared for ifa_ifwithaddr() */
2697				sin6->sin6_scope_id = 0;
2698				break;
2699			}
2700#endif
2701		default:
2702			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2703			return (EAFNOSUPPORT);
2704		}
2705	}
2706	SCTP_INP_INFO_WLOCK();
2707	SCTP_INP_WLOCK(inp);
2708	/* Setup a vrf_id to be the default for the non-bind-all case. */
2709	vrf_id = inp->def_vrf_id;
2710
2711	/* increase our count due to the unlock we do */
2712	SCTP_INP_INCR_REF(inp);
2713	if (lport) {
2714		/*
2715		 * Did the caller specify a port? if so we must see if a ep
2716		 * already has this one bound.
2717		 */
2718		/* got to be root to get at low ports */
2719		if (ntohs(lport) < IPPORT_RESERVED) {
2720			if (p && (error =
2721			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2722			    )) {
2723				SCTP_INP_DECR_REF(inp);
2724				SCTP_INP_WUNLOCK(inp);
2725				SCTP_INP_INFO_WUNLOCK();
2726				return (error);
2727			}
2728		}
2729		if (p == NULL) {
2730			SCTP_INP_DECR_REF(inp);
2731			SCTP_INP_WUNLOCK(inp);
2732			SCTP_INP_INFO_WUNLOCK();
2733			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2734			return (error);
2735		}
2736		SCTP_INP_WUNLOCK(inp);
2737		if (bindall) {
2738			vrf_id = inp->def_vrf_id;
2739			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2740			if (inp_tmp != NULL) {
2741				/*
2742				 * lock guy returned and lower count note
2743				 * that we are not bound so inp_tmp should
2744				 * NEVER be inp. And it is this inp
2745				 * (inp_tmp) that gets the reference bump,
2746				 * so we must lower it.
2747				 */
2748				SCTP_INP_DECR_REF(inp_tmp);
2749				/* unlock info */
2750				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2751				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2752					/*
2753					 * Ok, must be one-2-one and
2754					 * allowing port re-use
2755					 */
2756					port_reuse_active = 1;
2757					goto continue_anyway;
2758				}
2759				SCTP_INP_DECR_REF(inp);
2760				SCTP_INP_INFO_WUNLOCK();
2761				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2762				return (EADDRINUSE);
2763			}
2764		} else {
2765			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2766			if (inp_tmp != NULL) {
2767				/*
2768				 * lock guy returned and lower count note
2769				 * that we are not bound so inp_tmp should
2770				 * NEVER be inp. And it is this inp
2771				 * (inp_tmp) that gets the reference bump,
2772				 * so we must lower it.
2773				 */
2774				SCTP_INP_DECR_REF(inp_tmp);
2775				/* unlock info */
2776				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2777				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2778					/*
2779					 * Ok, must be one-2-one and
2780					 * allowing port re-use
2781					 */
2782					port_reuse_active = 1;
2783					goto continue_anyway;
2784				}
2785				SCTP_INP_DECR_REF(inp);
2786				SCTP_INP_INFO_WUNLOCK();
2787				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2788				return (EADDRINUSE);
2789			}
2790		}
2791continue_anyway:
2792		SCTP_INP_WLOCK(inp);
2793		if (bindall) {
2794			/* verify that no lport is not used by a singleton */
2795			if ((port_reuse_active == 0) &&
2796			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))
2797			    ) {
2798				/* Sorry someone already has this one bound */
2799				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2800				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2801					port_reuse_active = 1;
2802				} else {
2803					SCTP_INP_DECR_REF(inp);
2804					SCTP_INP_WUNLOCK(inp);
2805					SCTP_INP_INFO_WUNLOCK();
2806					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2807					return (EADDRINUSE);
2808				}
2809			}
2810		}
2811	} else {
2812		uint16_t first, last, candidate;
2813		uint16_t count;
2814		int done;
2815
2816		if (ip_inp->inp_flags & INP_HIGHPORT) {
2817			first = MODULE_GLOBAL(ipport_hifirstauto);
2818			last = MODULE_GLOBAL(ipport_hilastauto);
2819		} else if (ip_inp->inp_flags & INP_LOWPORT) {
2820			if (p && (error =
2821			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2822			    )) {
2823				SCTP_INP_DECR_REF(inp);
2824				SCTP_INP_WUNLOCK(inp);
2825				SCTP_INP_INFO_WUNLOCK();
2826				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2827				return (error);
2828			}
2829			first = MODULE_GLOBAL(ipport_lowfirstauto);
2830			last = MODULE_GLOBAL(ipport_lowlastauto);
2831		} else {
2832			first = MODULE_GLOBAL(ipport_firstauto);
2833			last = MODULE_GLOBAL(ipport_lastauto);
2834		}
2835		if (first > last) {
2836			uint16_t temp;
2837
2838			temp = first;
2839			first = last;
2840			last = temp;
2841		}
2842		count = last - first + 1;	/* number of candidates */
2843		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
2844
2845		done = 0;
2846		while (!done) {
2847			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
2848				done = 1;
2849			}
2850			if (!done) {
2851				if (--count == 0) {
2852					SCTP_INP_DECR_REF(inp);
2853					SCTP_INP_WUNLOCK(inp);
2854					SCTP_INP_INFO_WUNLOCK();
2855					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2856					return (EADDRINUSE);
2857				}
2858				if (candidate == last)
2859					candidate = first;
2860				else
2861					candidate = candidate + 1;
2862			}
2863		}
2864		lport = htons(candidate);
2865	}
2866	SCTP_INP_DECR_REF(inp);
2867	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2868	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2869		/*
2870		 * this really should not happen. The guy did a non-blocking
2871		 * bind and then did a close at the same time.
2872		 */
2873		SCTP_INP_WUNLOCK(inp);
2874		SCTP_INP_INFO_WUNLOCK();
2875		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2876		return (EINVAL);
2877	}
2878	/* ok we look clear to give out this port, so lets setup the binding */
2879	if (bindall) {
2880		/* binding to all addresses, so just set in the proper flags */
2881		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
2882		/* set the automatic addr changes from kernel flag */
2883		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
2884			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2885			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2886		} else {
2887			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2888			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2889		}
2890		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
2891			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
2892		} else {
2893			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
2894		}
2895		/*
2896		 * set the automatic mobility_base from kernel flag (by
2897		 * micchie)
2898		 */
2899		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
2900			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
2901			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2902		} else {
2903			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
2904			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2905		}
2906		/*
2907		 * set the automatic mobility_fasthandoff from kernel flag
2908		 * (by micchie)
2909		 */
2910		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
2911			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
2912			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2913		} else {
2914			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
2915			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2916		}
2917	} else {
2918		/*
2919		 * bind specific, make sure flags is off and add a new
2920		 * address structure to the sctp_addr_list inside the ep
2921		 * structure.
2922		 *
2923		 * We will need to allocate one and insert it at the head. The
2924		 * socketopt call can just insert new addresses in there as
2925		 * well. It will also have to do the embed scope kame hack
2926		 * too (before adding).
2927		 */
2928		struct sctp_ifa *ifa;
2929		struct sockaddr_storage store_sa;
2930
2931		memset(&store_sa, 0, sizeof(store_sa));
2932		if (addr->sa_family == AF_INET) {
2933			struct sockaddr_in *sin;
2934
2935			sin = (struct sockaddr_in *)&store_sa;
2936			memcpy(sin, addr, sizeof(struct sockaddr_in));
2937			sin->sin_port = 0;
2938		} else if (addr->sa_family == AF_INET6) {
2939			struct sockaddr_in6 *sin6;
2940
2941			sin6 = (struct sockaddr_in6 *)&store_sa;
2942			memcpy(sin6, addr, sizeof(struct sockaddr_in6));
2943			sin6->sin6_port = 0;
2944		}
2945		/*
2946		 * first find the interface with the bound address need to
2947		 * zero out the port to find the address! yuck! can't do
2948		 * this earlier since need port for sctp_pcb_findep()
2949		 */
2950		if (sctp_ifap != NULL)
2951			ifa = sctp_ifap;
2952		else {
2953			/*
2954			 * Note for BSD we hit here always other O/S's will
2955			 * pass things in via the sctp_ifap argument
2956			 * (Panda).
2957			 */
2958			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
2959			    vrf_id, SCTP_ADDR_NOT_LOCKED);
2960		}
2961		if (ifa == NULL) {
2962			/* Can't find an interface with that address */
2963			SCTP_INP_WUNLOCK(inp);
2964			SCTP_INP_INFO_WUNLOCK();
2965			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
2966			return (EADDRNOTAVAIL);
2967		}
2968		if (addr->sa_family == AF_INET6) {
2969			/* GAK, more FIXME IFA lock? */
2970			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
2971				/* Can't bind a non-existent addr. */
2972				SCTP_INP_WUNLOCK(inp);
2973				SCTP_INP_INFO_WUNLOCK();
2974				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2975				return (EINVAL);
2976			}
2977		}
2978		/* we're not bound all */
2979		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
2980		/* allow bindx() to send ASCONF's for binding changes */
2981		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2982		/* clear automatic addr changes from kernel flag */
2983		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2984
2985		/* add this address to the endpoint list */
2986		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
2987		if (error != 0) {
2988			SCTP_INP_WUNLOCK(inp);
2989			SCTP_INP_INFO_WUNLOCK();
2990			return (error);
2991		}
2992		inp->laddr_count++;
2993	}
2994	/* find the bucket */
2995	if (port_reuse_active) {
2996		/* Put it into tcp 1-2-1 hash */
2997		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport),
2998		    SCTP_BASE_INFO(hashtcpmark))];
2999		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3000	} else {
3001		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
3002		    SCTP_BASE_INFO(hashmark))];
3003	}
3004	/* put it in the bucket */
3005	LIST_INSERT_HEAD(head, inp, sctp_hash);
3006	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3007	    head, ntohs(lport), port_reuse_active);
3008	/* set in the port */
3009	inp->sctp_lport = lport;
3010
3011	/* turn off just the unbound flag */
3012	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3013	SCTP_INP_WUNLOCK(inp);
3014	SCTP_INP_INFO_WUNLOCK();
3015	return (0);
3016}
3017
3018
3019static void
3020sctp_iterator_inp_being_freed(struct sctp_inpcb *inp, struct sctp_inpcb *inp_next)
3021{
3022	struct sctp_iterator *it;
3023
3024	/*
3025	 * We enter with the only the ITERATOR_LOCK in place and a write
3026	 * lock on the inp_info stuff.
3027	 */
3028
3029	/*
3030	 * Go through all iterators, we must do this since it is possible
3031	 * that some iterator does NOT have the lock, but is waiting for it.
3032	 * And the one that had the lock has either moved in the last
3033	 * iteration or we just cleared it above. We need to find all of
3034	 * those guys. The list of iterators should never be very big
3035	 * though.
3036	 */
3037	TAILQ_FOREACH(it, &SCTP_BASE_INFO(iteratorhead), sctp_nxt_itr) {
3038		if (it == inp->inp_starting_point_for_iterator)
3039			/* skip this guy, he's special */
3040			continue;
3041		if (it->inp == inp) {
3042			/*
3043			 * This is tricky and we DON'T lock the iterator.
3044			 * Reason is he's running but waiting for me since
3045			 * inp->inp_starting_point_for_iterator has the lock
3046			 * on me (the guy above we skipped). This tells us
3047			 * its is not running but waiting for
3048			 * inp->inp_starting_point_for_iterator to be
3049			 * released by the guy that does have our INP in a
3050			 * lock.
3051			 */
3052			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3053				it->inp = NULL;
3054				it->stcb = NULL;
3055			} else {
3056				/* set him up to do the next guy not me */
3057				it->inp = inp_next;
3058				it->stcb = NULL;
3059			}
3060		}
3061	}
3062	it = inp->inp_starting_point_for_iterator;
3063	if (it) {
3064		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3065			it->inp = NULL;
3066		} else {
3067			it->inp = inp_next;
3068		}
3069		it->stcb = NULL;
3070	}
3071}
3072
3073/* release sctp_inpcb unbind the port */
3074void
3075sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3076{
3077	/*
3078	 * Here we free a endpoint. We must find it (if it is in the Hash
3079	 * table) and remove it from there. Then we must also find it in the
3080	 * overall list and remove it from there. After all removals are
3081	 * complete then any timer has to be stopped. Then start the actual
3082	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3083	 * all associations. d) finally the ep itself.
3084	 */
3085	struct sctp_pcb *m;
3086	struct sctp_inpcb *inp_save;
3087	struct sctp_tcb *asoc, *nasoc;
3088	struct sctp_laddr *laddr, *nladdr;
3089	struct inpcb *ip_pcb;
3090	struct socket *so;
3091
3092	struct sctp_queued_to_read *sq;
3093
3094
3095	int cnt;
3096	sctp_sharedkey_t *shared_key;
3097
3098
3099#ifdef SCTP_LOG_CLOSING
3100	sctp_log_closing(inp, NULL, 0);
3101#endif
3102	SCTP_ITERATOR_LOCK();
3103	so = inp->sctp_socket;
3104	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3105		/* been here before.. eeks.. get out of here */
3106		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3107		SCTP_ITERATOR_UNLOCK();
3108#ifdef SCTP_LOG_CLOSING
3109		sctp_log_closing(inp, NULL, 1);
3110#endif
3111		return;
3112	}
3113	SCTP_ASOC_CREATE_LOCK(inp);
3114	SCTP_INP_INFO_WLOCK();
3115
3116	SCTP_INP_WLOCK(inp);
3117	/* First time through we have the socket lock, after that no more. */
3118	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3119		/*
3120		 * Once we are in we can remove the flag from = 1 is only
3121		 * passed from the actual closing routines that are called
3122		 * via the sockets layer.
3123		 */
3124		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3125		/* socket is gone, so no more wakeups allowed */
3126		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3127		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3128		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3129	}
3130	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3131	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3132
3133	if (inp->control) {
3134		sctp_m_freem(inp->control);
3135		inp->control = NULL;
3136	}
3137	if (inp->pkt) {
3138		sctp_m_freem(inp->pkt);
3139		inp->pkt = NULL;
3140	}
3141	m = &inp->sctp_ep;
3142	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3143					 * here but I will be nice :> (i.e.
3144					 * ip_pcb = ep;) */
3145	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3146		int cnt_in_sd;
3147
3148		cnt_in_sd = 0;
3149		for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
3150		    asoc = nasoc) {
3151			SCTP_TCB_LOCK(asoc);
3152			nasoc = LIST_NEXT(asoc, sctp_tcblist);
3153			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3154				/* Skip guys being freed */
3155				/* asoc->sctp_socket = NULL; FIXME MT */
3156				cnt_in_sd++;
3157				SCTP_TCB_UNLOCK(asoc);
3158				continue;
3159			}
3160			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3161			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3162			    (asoc->asoc.total_output_queue_size == 0)) {
3163				/*
3164				 * If we have data in queue, we don't want
3165				 * to just free since the app may have done,
3166				 * send()/close or connect/send/close. And
3167				 * it wants the data to get across first.
3168				 */
3169				/* Just abandon things in the front states */
3170				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3171				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3172					cnt_in_sd++;
3173				}
3174				continue;
3175			}
3176			/* Disconnect the socket please */
3177			asoc->sctp_socket = NULL;
3178			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3179			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3180			    (asoc->asoc.control_pdapi) ||
3181			    (asoc->asoc.size_on_all_streams > 0) ||
3182			    (so && (so->so_rcv.sb_cc > 0))
3183			    ) {
3184				/* Left with Data unread */
3185				struct mbuf *op_err;
3186
3187				op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3188				    0, M_DONTWAIT, 1, MT_DATA);
3189				if (op_err) {
3190					/* Fill in the user initiated abort */
3191					struct sctp_paramhdr *ph;
3192					uint32_t *ippp;
3193
3194					SCTP_BUF_LEN(op_err) =
3195					    sizeof(struct sctp_paramhdr) + sizeof(uint32_t);
3196					ph = mtod(op_err,
3197					    struct sctp_paramhdr *);
3198					ph->param_type = htons(
3199					    SCTP_CAUSE_USER_INITIATED_ABT);
3200					ph->param_length = htons(SCTP_BUF_LEN(op_err));
3201					ippp = (uint32_t *) (ph + 1);
3202					*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3);
3203				}
3204				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3205#if defined(SCTP_PANIC_ON_ABORT)
3206				panic("inpcb_free does an abort");
3207#endif
3208				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3209				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3210				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3211				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3212					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3213				}
3214				if (sctp_free_assoc(inp, asoc,
3215				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3216					cnt_in_sd++;
3217				}
3218				continue;
3219			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3220				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3221				    (asoc->asoc.stream_queue_cnt == 0)
3222			    ) {
3223				if (asoc->asoc.locked_on_sending) {
3224					goto abort_anyway;
3225				}
3226				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3227				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3228					/*
3229					 * there is nothing queued to send,
3230					 * so I send shutdown
3231					 */
3232					sctp_send_shutdown(asoc, asoc->asoc.primary_destination);
3233					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3234					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3235						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3236					}
3237					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3238					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3239					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3240					    asoc->asoc.primary_destination);
3241					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3242					    asoc->asoc.primary_destination);
3243					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3244				}
3245			} else {
3246				/* mark into shutdown pending */
3247				struct sctp_stream_queue_pending *sp;
3248
3249				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3250				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3251				    asoc->asoc.primary_destination);
3252				if (asoc->asoc.locked_on_sending) {
3253					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3254					    sctp_streamhead);
3255					if (sp == NULL) {
3256						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3257						    asoc->asoc.locked_on_sending,
3258						    asoc->asoc.locked_on_sending->stream_no);
3259					} else {
3260						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3261							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3262					}
3263				}
3264				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3265				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3266				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3267					struct mbuf *op_err;
3268
3269			abort_anyway:
3270					op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3271					    0, M_DONTWAIT, 1, MT_DATA);
3272					if (op_err) {
3273						/*
3274						 * Fill in the user
3275						 * initiated abort
3276						 */
3277						struct sctp_paramhdr *ph;
3278						uint32_t *ippp;
3279
3280						SCTP_BUF_LEN(op_err) =
3281						    (sizeof(struct sctp_paramhdr) +
3282						    sizeof(uint32_t));
3283						ph = mtod(op_err,
3284						    struct sctp_paramhdr *);
3285						ph->param_type = htons(
3286						    SCTP_CAUSE_USER_INITIATED_ABT);
3287						ph->param_length = htons(SCTP_BUF_LEN(op_err));
3288						ippp = (uint32_t *) (ph + 1);
3289						*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5);
3290					}
3291					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3292#if defined(SCTP_PANIC_ON_ABORT)
3293					panic("inpcb_free does an abort");
3294#endif
3295
3296					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3297					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3298					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3299					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3300						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3301					}
3302					if (sctp_free_assoc(inp, asoc,
3303					    SCTP_PCBFREE_NOFORCE,
3304					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3305						cnt_in_sd++;
3306					}
3307					continue;
3308				} else {
3309					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3310				}
3311			}
3312			cnt_in_sd++;
3313			SCTP_TCB_UNLOCK(asoc);
3314		}
3315		/* now is there some left in our SHUTDOWN state? */
3316		if (cnt_in_sd) {
3317			SCTP_INP_WUNLOCK(inp);
3318			SCTP_ASOC_CREATE_UNLOCK(inp);
3319			SCTP_INP_INFO_WUNLOCK();
3320			SCTP_ITERATOR_UNLOCK();
3321#ifdef SCTP_LOG_CLOSING
3322			sctp_log_closing(inp, NULL, 2);
3323#endif
3324			inp->sctp_socket = NULL;
3325			return;
3326		}
3327	}
3328	inp->sctp_socket = NULL;
3329	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3330	    SCTP_PCB_FLAGS_UNBOUND) {
3331		/*
3332		 * ok, this guy has been bound. It's port is somewhere in
3333		 * the SCTP_BASE_INFO(hash table). Remove it!
3334		 */
3335		LIST_REMOVE(inp, sctp_hash);
3336		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3337	}
3338	/*
3339	 * If there is a timer running to kill us, forget it, since it may
3340	 * have a contest on the INP lock.. which would cause us to die ...
3341	 */
3342	cnt = 0;
3343	for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
3344	    asoc = nasoc) {
3345		nasoc = LIST_NEXT(asoc, sctp_tcblist);
3346		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3347			cnt++;
3348			continue;
3349		}
3350		/* Free associations that are NOT killing us */
3351		SCTP_TCB_LOCK(asoc);
3352		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3353		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3354			struct mbuf *op_err;
3355			uint32_t *ippp;
3356
3357			op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3358			    0, M_DONTWAIT, 1, MT_DATA);
3359			if (op_err) {
3360				/* Fill in the user initiated abort */
3361				struct sctp_paramhdr *ph;
3362
3363				SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) +
3364				    sizeof(uint32_t));
3365				ph = mtod(op_err, struct sctp_paramhdr *);
3366				ph->param_type = htons(
3367				    SCTP_CAUSE_USER_INITIATED_ABT);
3368				ph->param_length = htons(SCTP_BUF_LEN(op_err));
3369				ippp = (uint32_t *) (ph + 1);
3370				*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7);
3371
3372			}
3373			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3374#if defined(SCTP_PANIC_ON_ABORT)
3375			panic("inpcb_free does an abort");
3376#endif
3377			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3378			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3379		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3380			cnt++;
3381			SCTP_TCB_UNLOCK(asoc);
3382			continue;
3383		}
3384		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3385		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3386			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3387		}
3388		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3389			cnt++;
3390		}
3391	}
3392	if (cnt) {
3393		/* Ok we have someone out there that will kill us */
3394		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3395		SCTP_INP_WUNLOCK(inp);
3396		SCTP_ASOC_CREATE_UNLOCK(inp);
3397		SCTP_INP_INFO_WUNLOCK();
3398		SCTP_ITERATOR_UNLOCK();
3399#ifdef SCTP_LOG_CLOSING
3400		sctp_log_closing(inp, NULL, 3);
3401#endif
3402		return;
3403	}
3404	if ((inp->refcount) || (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3405		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3406		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3407		SCTP_INP_WUNLOCK(inp);
3408		SCTP_ASOC_CREATE_UNLOCK(inp);
3409		SCTP_INP_INFO_WUNLOCK();
3410		SCTP_ITERATOR_UNLOCK();
3411#ifdef SCTP_LOG_CLOSING
3412		sctp_log_closing(inp, NULL, 4);
3413#endif
3414		return;
3415	}
3416	(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3417	inp->sctp_ep.signature_change.type = 0;
3418	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3419
3420#ifdef SCTP_LOG_CLOSING
3421	sctp_log_closing(inp, NULL, 5);
3422#endif
3423
3424	(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3425	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NONE;
3426	/* Clear the read queue */
3427	if ((inp->sctp_asocidhash) != NULL) {
3428		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3429		inp->sctp_asocidhash = NULL;
3430	}
3431	/* sa_ignore FREED_MEMORY */
3432	while ((sq = TAILQ_FIRST(&inp->read_queue)) != NULL) {
3433		/* Its only abandoned if it had data left */
3434		if (sq->length)
3435			SCTP_STAT_INCR(sctps_left_abandon);
3436
3437		TAILQ_REMOVE(&inp->read_queue, sq, next);
3438		sctp_free_remote_addr(sq->whoFrom);
3439		if (so)
3440			so->so_rcv.sb_cc -= sq->length;
3441		if (sq->data) {
3442			sctp_m_freem(sq->data);
3443			sq->data = NULL;
3444		}
3445		/*
3446		 * no need to free the net count, since at this point all
3447		 * assoc's are gone.
3448		 */
3449		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3450		SCTP_DECR_READQ_COUNT();
3451	}
3452	/* Now the sctp_pcb things */
3453	/*
3454	 * free each asoc if it is not already closed/free. we can't use the
3455	 * macro here since le_next will get freed as part of the
3456	 * sctp_free_assoc() call.
3457	 */
3458	cnt = 0;
3459	if (so) {
3460#ifdef IPSEC
3461		ipsec_delete_pcbpolicy(ip_pcb);
3462#endif				/* IPSEC */
3463
3464		/* Unlocks not needed since the socket is gone now */
3465	}
3466	if (ip_pcb->inp_options) {
3467		(void)sctp_m_free(ip_pcb->inp_options);
3468		ip_pcb->inp_options = 0;
3469	}
3470	if (ip_pcb->inp_moptions) {
3471		inp_freemoptions(ip_pcb->inp_moptions);
3472		ip_pcb->inp_moptions = 0;
3473	}
3474#ifdef INET6
3475	if (ip_pcb->inp_vflag & INP_IPV6) {
3476		struct in6pcb *in6p;
3477
3478		in6p = (struct in6pcb *)inp;
3479		ip6_freepcbopts(in6p->in6p_outputopts);
3480	}
3481#endif				/* INET6 */
3482	ip_pcb->inp_vflag = 0;
3483	/* free up authentication fields */
3484	if (inp->sctp_ep.local_auth_chunks != NULL)
3485		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3486	if (inp->sctp_ep.local_hmacs != NULL)
3487		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3488
3489	shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
3490	while (shared_key) {
3491		LIST_REMOVE(shared_key, next);
3492		sctp_free_sharedkey(shared_key);
3493		/* sa_ignore FREED_MEMORY */
3494		shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
3495	}
3496
3497	inp_save = LIST_NEXT(inp, sctp_list);
3498	LIST_REMOVE(inp, sctp_list);
3499
3500	/* fix any iterators only after out of the list */
3501	sctp_iterator_inp_being_freed(inp, inp_save);
3502	/*
3503	 * if we have an address list the following will free the list of
3504	 * ifaddr's that are set into this ep. Again macro limitations here,
3505	 * since the LIST_FOREACH could be a bad idea.
3506	 */
3507	for ((laddr = LIST_FIRST(&inp->sctp_addr_list)); laddr != NULL;
3508	    laddr = nladdr) {
3509		nladdr = LIST_NEXT(laddr, sctp_nxt_addr);
3510		sctp_remove_laddr(laddr);
3511	}
3512
3513#ifdef SCTP_TRACK_FREED_ASOCS
3514	/* TEMP CODE */
3515	for ((asoc = LIST_FIRST(&inp->sctp_asoc_free_list)); asoc != NULL;
3516	    asoc = nasoc) {
3517		nasoc = LIST_NEXT(asoc, sctp_tcblist);
3518		LIST_REMOVE(asoc, sctp_tcblist);
3519		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3520		SCTP_DECR_ASOC_COUNT();
3521	}
3522	/* *** END TEMP CODE *** */
3523#endif
3524	/* Now lets see about freeing the EP hash table. */
3525	if (inp->sctp_tcbhash != NULL) {
3526		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3527		inp->sctp_tcbhash = NULL;
3528	}
3529	/* Now we must put the ep memory back into the zone pool */
3530	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3531	SCTP_INP_LOCK_DESTROY(inp);
3532	SCTP_INP_READ_DESTROY(inp);
3533	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3534	SCTP_INP_INFO_WUNLOCK();
3535	SCTP_ITERATOR_UNLOCK();
3536	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3537	SCTP_DECR_EP_COUNT();
3538}
3539
3540
3541struct sctp_nets *
3542sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3543{
3544	struct sctp_nets *net;
3545
3546	/* locate the address */
3547	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3548		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3549			return (net);
3550	}
3551	return (NULL);
3552}
3553
3554
3555int
3556sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3557{
3558	struct sctp_ifa *sctp_ifa;
3559
3560	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3561	if (sctp_ifa) {
3562		return (1);
3563	} else {
3564		return (0);
3565	}
3566}
3567
3568/*
3569 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3570 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3571 * stats of stuff.
3572 */
3573int
3574sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3575    int set_scope, int from)
3576{
3577	/*
3578	 * The following is redundant to the same lines in the
3579	 * sctp_aloc_assoc() but is needed since other's call the add
3580	 * address function
3581	 */
3582	struct sctp_nets *net, *netfirst;
3583	int addr_inscope;
3584
3585	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3586	    from);
3587	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3588
3589	netfirst = sctp_findnet(stcb, newaddr);
3590	if (netfirst) {
3591		/*
3592		 * Lie and return ok, we don't want to make the association
3593		 * go away for this behavior. It will happen in the TCP
3594		 * model in a connected socket. It does not reach the hash
3595		 * table until after the association is built so it can't be
3596		 * found. Mark as reachable, since the initial creation will
3597		 * have been cleared and the NOT_IN_ASSOC flag will have
3598		 * been added... and we don't want to end up removing it
3599		 * back out.
3600		 */
3601		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3602			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3603			    SCTP_ADDR_UNCONFIRMED);
3604		} else {
3605			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3606		}
3607
3608		return (0);
3609	}
3610	addr_inscope = 1;
3611	if (newaddr->sa_family == AF_INET) {
3612		struct sockaddr_in *sin;
3613
3614		sin = (struct sockaddr_in *)newaddr;
3615		if (sin->sin_addr.s_addr == 0) {
3616			/* Invalid address */
3617			return (-1);
3618		}
3619		/* zero out the bzero area */
3620		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3621
3622		/* assure len is set */
3623		sin->sin_len = sizeof(struct sockaddr_in);
3624		if (set_scope) {
3625#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3626			stcb->ipv4_local_scope = 1;
3627#else
3628			if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3629				stcb->asoc.ipv4_local_scope = 1;
3630			}
3631#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3632		} else {
3633			/* Validate the address is in scope */
3634			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3635			    (stcb->asoc.ipv4_local_scope == 0)) {
3636				addr_inscope = 0;
3637			}
3638		}
3639#ifdef INET6
3640	} else if (newaddr->sa_family == AF_INET6) {
3641		struct sockaddr_in6 *sin6;
3642
3643		sin6 = (struct sockaddr_in6 *)newaddr;
3644		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3645			/* Invalid address */
3646			return (-1);
3647		}
3648		/* assure len is set */
3649		sin6->sin6_len = sizeof(struct sockaddr_in6);
3650		if (set_scope) {
3651			if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3652				stcb->asoc.loopback_scope = 1;
3653				stcb->asoc.local_scope = 0;
3654				stcb->asoc.ipv4_local_scope = 1;
3655				stcb->asoc.site_scope = 1;
3656			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3657				/*
3658				 * If the new destination is a LINK_LOCAL we
3659				 * must have common site scope. Don't set
3660				 * the local scope since we may not share
3661				 * all links, only loopback can do this.
3662				 * Links on the local network would also be
3663				 * on our private network for v4 too.
3664				 */
3665				stcb->asoc.ipv4_local_scope = 1;
3666				stcb->asoc.site_scope = 1;
3667			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3668				/*
3669				 * If the new destination is SITE_LOCAL then
3670				 * we must have site scope in common.
3671				 */
3672				stcb->asoc.site_scope = 1;
3673			}
3674		} else {
3675			/* Validate the address is in scope */
3676			if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3677			    (stcb->asoc.loopback_scope == 0)) {
3678				addr_inscope = 0;
3679			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3680			    (stcb->asoc.local_scope == 0)) {
3681				addr_inscope = 0;
3682			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3683			    (stcb->asoc.site_scope == 0)) {
3684				addr_inscope = 0;
3685			}
3686		}
3687#endif
3688	} else {
3689		/* not supported family type */
3690		return (-1);
3691	}
3692	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3693	if (net == NULL) {
3694		return (-1);
3695	}
3696	SCTP_INCR_RADDR_COUNT();
3697	bzero(net, sizeof(*net));
3698	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3699	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3700	if (newaddr->sa_family == AF_INET) {
3701		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3702	} else if (newaddr->sa_family == AF_INET6) {
3703		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3704	}
3705	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3706	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3707		stcb->asoc.loopback_scope = 1;
3708		stcb->asoc.ipv4_local_scope = 1;
3709		stcb->asoc.local_scope = 0;
3710		stcb->asoc.site_scope = 1;
3711		addr_inscope = 1;
3712	}
3713	net->failure_threshold = stcb->asoc.def_net_failure;
3714	if (addr_inscope == 0) {
3715		net->dest_state = (SCTP_ADDR_REACHABLE |
3716		    SCTP_ADDR_OUT_OF_SCOPE);
3717	} else {
3718		if (from == SCTP_ADDR_IS_CONFIRMED)
3719			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3720			net->dest_state = SCTP_ADDR_REACHABLE;
3721		else
3722			net->dest_state = SCTP_ADDR_REACHABLE |
3723			    SCTP_ADDR_UNCONFIRMED;
3724	}
3725	/*
3726	 * We set this to 0, the timer code knows that this means its an
3727	 * initial value
3728	 */
3729	net->RTO = 0;
3730	net->RTO_measured = 0;
3731	stcb->asoc.numnets++;
3732	*(&net->ref_count) = 1;
3733	net->tos_flowlabel = 0;
3734	if (SCTP_BASE_SYSCTL(sctp_udp_tunneling_for_client_enable)) {
3735		net->port = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port));
3736	} else {
3737		net->port = 0;
3738	}
3739#ifdef INET
3740	if (newaddr->sa_family == AF_INET)
3741		net->tos_flowlabel = stcb->asoc.default_tos;
3742#endif
3743#ifdef INET6
3744	if (newaddr->sa_family == AF_INET6)
3745		net->tos_flowlabel = stcb->asoc.default_flowlabel;
3746#endif
3747	/* Init the timer structure */
3748	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3749	SCTP_OS_TIMER_INIT(&net->fr_timer.timer);
3750	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3751
3752	/* Now generate a route for this guy */
3753#ifdef INET6
3754	/* KAME hack: embed scopeid */
3755	if (newaddr->sa_family == AF_INET6) {
3756		struct sockaddr_in6 *sin6;
3757
3758		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3759		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3760		sin6->sin6_scope_id = 0;
3761	}
3762#endif
3763	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3764
3765	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3766		/* Get source address */
3767		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3768		    stcb,
3769		    (sctp_route_t *) & net->ro,
3770		    net,
3771		    0,
3772		    stcb->asoc.vrf_id);
3773		/* Now get the interface MTU */
3774		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3775			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3776		} else {
3777			net->mtu = 0;
3778		}
3779#ifdef SCTP_PRINT_FOR_B_AND_M
3780		SCTP_PRINTF("We have found an interface mtu of %d\n", net->mtu);
3781#endif
3782		if (net->mtu == 0) {
3783			/* Huh ?? */
3784			net->mtu = SCTP_DEFAULT_MTU;
3785		} else {
3786			uint32_t rmtu;
3787
3788			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3789#ifdef SCTP_PRINT_FOR_B_AND_M
3790			SCTP_PRINTF("The route mtu is %d\n", rmtu);
3791#endif
3792			if (rmtu == 0) {
3793				/*
3794				 * Start things off to match mtu of
3795				 * interface please.
3796				 */
3797				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3798				    net->ro.ro_rt, net->mtu);
3799			} else {
3800				/*
3801				 * we take the route mtu over the interface,
3802				 * since the route may be leading out the
3803				 * loopback, or a different interface.
3804				 */
3805				net->mtu = rmtu;
3806			}
3807		}
3808		if (from == SCTP_ALLOC_ASOC) {
3809#ifdef SCTP_PRINT_FOR_B_AND_M
3810			SCTP_PRINTF("New assoc sets mtu to :%d\n", net->mtu);
3811#endif
3812			stcb->asoc.smallest_mtu = net->mtu;
3813		}
3814	} else {
3815		net->mtu = stcb->asoc.smallest_mtu;
3816	}
3817#ifdef INET6
3818	if (newaddr->sa_family == AF_INET6) {
3819		struct sockaddr_in6 *sin6;
3820
3821		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3822		(void)sa6_recoverscope(sin6);
3823	}
3824#endif
3825	if (net->port) {
3826		net->mtu -= sizeof(struct udphdr);
3827	}
3828	if (stcb->asoc.smallest_mtu > net->mtu) {
3829#ifdef SCTP_PRINT_FOR_B_AND_M
3830		SCTP_PRINTF("new address mtu:%d smaller than smallest:%d\n",
3831		    net->mtu, stcb->asoc.smallest_mtu);
3832#endif
3833		stcb->asoc.smallest_mtu = net->mtu;
3834	}
3835	/* JRS - Use the congestion control given in the CC module */
3836	stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net);
3837
3838	/*
3839	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3840	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
3841	 */
3842	net->find_pseudo_cumack = 1;
3843	net->find_rtx_pseudo_cumack = 1;
3844	net->src_addr_selected = 0;
3845	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
3846	if (net->ro.ro_rt == NULL) {
3847		/* Since we have no route put it at the back */
3848		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3849	} else if (netfirst == NULL) {
3850		/* We are the first one in the pool. */
3851		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3852	} else if (netfirst->ro.ro_rt == NULL) {
3853		/*
3854		 * First one has NO route. Place this one ahead of the first
3855		 * one.
3856		 */
3857		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3858	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
3859		/*
3860		 * This one has a different interface than the one at the
3861		 * top of the list. Place it ahead.
3862		 */
3863		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3864	} else {
3865		/*
3866		 * Ok we have the same interface as the first one. Move
3867		 * forward until we find either a) one with a NULL route...
3868		 * insert ahead of that b) one with a different ifp.. insert
3869		 * after that. c) end of the list.. insert at the tail.
3870		 */
3871		struct sctp_nets *netlook;
3872
3873		do {
3874			netlook = TAILQ_NEXT(netfirst, sctp_next);
3875			if (netlook == NULL) {
3876				/* End of the list */
3877				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3878				break;
3879			} else if (netlook->ro.ro_rt == NULL) {
3880				/* next one has NO route */
3881				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
3882				break;
3883			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
3884				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
3885				    net, sctp_next);
3886				break;
3887			}
3888			/* Shift forward */
3889			netfirst = netlook;
3890		} while (netlook != NULL);
3891	}
3892
3893	/* got to have a primary set */
3894	if (stcb->asoc.primary_destination == 0) {
3895		stcb->asoc.primary_destination = net;
3896	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
3897		    (net->ro.ro_rt) &&
3898	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
3899		/* No route to current primary adopt new primary */
3900		stcb->asoc.primary_destination = net;
3901	}
3902	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, stcb->sctp_ep, stcb,
3903	    net);
3904	/* Validate primary is first */
3905	net = TAILQ_FIRST(&stcb->asoc.nets);
3906	if ((net != stcb->asoc.primary_destination) &&
3907	    (stcb->asoc.primary_destination)) {
3908		/*
3909		 * first one on the list is NOT the primary sctp_cmpaddr()
3910		 * is much more efficent if the primary is the first on the
3911		 * list, make it so.
3912		 */
3913		TAILQ_REMOVE(&stcb->asoc.nets,
3914		    stcb->asoc.primary_destination, sctp_next);
3915		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
3916		    stcb->asoc.primary_destination, sctp_next);
3917	}
3918	return (0);
3919}
3920
3921
3922static uint32_t
3923sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
3924{
3925	uint32_t id;
3926	struct sctpasochead *head;
3927	struct sctp_tcb *lstcb;
3928
3929	SCTP_INP_WLOCK(inp);
3930try_again:
3931	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3932		/* TSNH */
3933		SCTP_INP_WUNLOCK(inp);
3934		return (0);
3935	}
3936	/*
3937	 * We don't allow assoc id to be 0, this is needed otherwise if the
3938	 * id were to wrap we would have issues with some socket options.
3939	 */
3940	if (inp->sctp_associd_counter == 0) {
3941		inp->sctp_associd_counter++;
3942	}
3943	id = inp->sctp_associd_counter;
3944	inp->sctp_associd_counter++;
3945	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
3946	if (lstcb) {
3947		goto try_again;
3948	}
3949	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
3950	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
3951	stcb->asoc.in_asocid_hash = 1;
3952	SCTP_INP_WUNLOCK(inp);
3953	return id;
3954}
3955
3956/*
3957 * allocate an association and add it to the endpoint. The caller must be
3958 * careful to add all additional addresses once they are know right away or
3959 * else the assoc will be may experience a blackout scenario.
3960 */
3961struct sctp_tcb *
3962sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
3963    int for_a_init, int *error, uint32_t override_tag, uint32_t vrf_id,
3964    struct thread *p
3965)
3966{
3967	/* note the p argument is only valid in unbound sockets */
3968
3969	struct sctp_tcb *stcb;
3970	struct sctp_association *asoc;
3971	struct sctpasochead *head;
3972	uint16_t rport;
3973	int err;
3974
3975	/*
3976	 * Assumption made here: Caller has done a
3977	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
3978	 * address does not exist already.
3979	 */
3980	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
3981		/* Hit max assoc, sorry no more */
3982		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
3983		*error = ENOBUFS;
3984		return (NULL);
3985	}
3986	if (firstaddr == NULL) {
3987		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3988		*error = EINVAL;
3989		return (NULL);
3990	}
3991	SCTP_INP_RLOCK(inp);
3992	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
3993	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
3994	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
3995		/*
3996		 * If its in the TCP pool, its NOT allowed to create an
3997		 * association. The parent listener needs to call
3998		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
3999		 * off, or connected one does this.. its an error.
4000		 */
4001		SCTP_INP_RUNLOCK(inp);
4002		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4003		*error = EINVAL;
4004		return (NULL);
4005	}
4006	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4007#ifdef SCTP_DEBUG
4008	if (firstaddr) {
4009		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4010		SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4011		    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4012	} else {
4013		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4014	}
4015#endif				/* SCTP_DEBUG */
4016	if (firstaddr->sa_family == AF_INET) {
4017		struct sockaddr_in *sin;
4018
4019		sin = (struct sockaddr_in *)firstaddr;
4020		if ((sin->sin_port == 0) || (sin->sin_addr.s_addr == 0)) {
4021			/* Invalid address */
4022			SCTP_INP_RUNLOCK(inp);
4023			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4024			*error = EINVAL;
4025			return (NULL);
4026		}
4027		rport = sin->sin_port;
4028	} else if (firstaddr->sa_family == AF_INET6) {
4029		struct sockaddr_in6 *sin6;
4030
4031		sin6 = (struct sockaddr_in6 *)firstaddr;
4032		if ((sin6->sin6_port == 0) ||
4033		    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
4034			/* Invalid address */
4035			SCTP_INP_RUNLOCK(inp);
4036			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4037			*error = EINVAL;
4038			return (NULL);
4039		}
4040		rport = sin6->sin6_port;
4041	} else {
4042		/* not supported family type */
4043		SCTP_INP_RUNLOCK(inp);
4044		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4045		*error = EINVAL;
4046		return (NULL);
4047	}
4048	SCTP_INP_RUNLOCK(inp);
4049	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4050		/*
4051		 * If you have not performed a bind, then we need to do the
4052		 * ephemerial bind for you.
4053		 */
4054		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4055		    (struct sockaddr *)NULL,
4056		    (struct sctp_ifa *)NULL,
4057		    p
4058		    ))) {
4059			/* bind error, probably perm */
4060			*error = err;
4061			return (NULL);
4062		}
4063	}
4064	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4065	if (stcb == NULL) {
4066		/* out of memory? */
4067		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4068		*error = ENOMEM;
4069		return (NULL);
4070	}
4071	SCTP_INCR_ASOC_COUNT();
4072
4073	bzero(stcb, sizeof(*stcb));
4074	asoc = &stcb->asoc;
4075
4076	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4077	SCTP_TCB_LOCK_INIT(stcb);
4078	SCTP_TCB_SEND_LOCK_INIT(stcb);
4079	stcb->rport = rport;
4080	/* setup back pointer's */
4081	stcb->sctp_ep = inp;
4082	stcb->sctp_socket = inp->sctp_socket;
4083	if ((err = sctp_init_asoc(inp, stcb, for_a_init, override_tag, vrf_id))) {
4084		/* failed */
4085		SCTP_TCB_LOCK_DESTROY(stcb);
4086		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4087		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4088		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4089		SCTP_DECR_ASOC_COUNT();
4090		*error = err;
4091		return (NULL);
4092	}
4093	/* and the port */
4094	SCTP_INP_INFO_WLOCK();
4095	SCTP_INP_WLOCK(inp);
4096	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4097		/* inpcb freed while alloc going on */
4098		SCTP_TCB_LOCK_DESTROY(stcb);
4099		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4100		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4101		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4102		SCTP_INP_WUNLOCK(inp);
4103		SCTP_INP_INFO_WUNLOCK();
4104		SCTP_DECR_ASOC_COUNT();
4105		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4106		*error = EINVAL;
4107		return (NULL);
4108	}
4109	SCTP_TCB_LOCK(stcb);
4110
4111	/* now that my_vtag is set, add it to the hash */
4112	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4113	/* put it in the bucket in the vtag hash of assoc's for the system */
4114	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4115	SCTP_INP_INFO_WUNLOCK();
4116
4117	if ((err = sctp_add_remote_addr(stcb, firstaddr, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4118		/* failure.. memory error? */
4119		if (asoc->strmout) {
4120			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4121			asoc->strmout = NULL;
4122		}
4123		if (asoc->mapping_array) {
4124			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4125			asoc->mapping_array = NULL;
4126		}
4127		if (asoc->nr_mapping_array) {
4128			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4129			asoc->nr_mapping_array = NULL;
4130		}
4131		SCTP_DECR_ASOC_COUNT();
4132		SCTP_TCB_LOCK_DESTROY(stcb);
4133		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4134		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4135		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4136		SCTP_INP_WUNLOCK(inp);
4137		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4138		*error = ENOBUFS;
4139		return (NULL);
4140	}
4141	/* Init all the timers */
4142	SCTP_OS_TIMER_INIT(&asoc->hb_timer.timer);
4143	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4144	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4145	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4146	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4147	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4148	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4149	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4150
4151	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4152	/* now file the port under the hash as well */
4153	if (inp->sctp_tcbhash != NULL) {
4154		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4155		    inp->sctp_hashmark)];
4156		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4157	}
4158	SCTP_INP_WUNLOCK(inp);
4159	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", stcb);
4160	return (stcb);
4161}
4162
4163
4164void
4165sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4166{
4167	struct sctp_association *asoc;
4168
4169	asoc = &stcb->asoc;
4170	asoc->numnets--;
4171	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4172	if (net == asoc->primary_destination) {
4173		/* Reset primary */
4174		struct sctp_nets *lnet;
4175
4176		lnet = TAILQ_FIRST(&asoc->nets);
4177		/*
4178		 * Mobility adaptation Ideally, if deleted destination is
4179		 * the primary, it becomes a fast retransmission trigger by
4180		 * the subsequent SET PRIMARY. (by micchie)
4181		 */
4182		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4183		    SCTP_MOBILITY_BASE) ||
4184		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4185		    SCTP_MOBILITY_FASTHANDOFF)) {
4186			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4187			if (asoc->deleted_primary != NULL) {
4188				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4189				goto out;
4190			}
4191			asoc->deleted_primary = net;
4192			atomic_add_int(&net->ref_count, 1);
4193			memset(&net->lastsa, 0, sizeof(net->lastsa));
4194			memset(&net->lastsv, 0, sizeof(net->lastsv));
4195			sctp_mobility_feature_on(stcb->sctp_ep,
4196			    SCTP_MOBILITY_PRIM_DELETED);
4197			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4198			    stcb->sctp_ep, stcb, NULL);
4199		}
4200out:
4201		/* Try to find a confirmed primary */
4202		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4203	}
4204	if (net == asoc->last_data_chunk_from) {
4205		/* Reset primary */
4206		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4207	}
4208	if (net == asoc->last_control_chunk_from) {
4209		/* Clear net */
4210		asoc->last_control_chunk_from = NULL;
4211	}
4212	sctp_free_remote_addr(net);
4213}
4214
4215/*
4216 * remove a remote endpoint address from an association, it will fail if the
4217 * address does not exist.
4218 */
4219int
4220sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4221{
4222	/*
4223	 * Here we need to remove a remote address. This is quite simple, we
4224	 * first find it in the list of address for the association
4225	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4226	 * on that item. Note we do not allow it to be removed if there are
4227	 * no other addresses.
4228	 */
4229	struct sctp_association *asoc;
4230	struct sctp_nets *net, *net_tmp;
4231
4232	asoc = &stcb->asoc;
4233
4234	/* locate the address */
4235	for (net = TAILQ_FIRST(&asoc->nets); net != NULL; net = net_tmp) {
4236		net_tmp = TAILQ_NEXT(net, sctp_next);
4237		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4238			continue;
4239		}
4240		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4241		    remaddr)) {
4242			/* we found the guy */
4243			if (asoc->numnets < 2) {
4244				/* Must have at LEAST two remote addresses */
4245				return (-1);
4246			} else {
4247				sctp_remove_net(stcb, net);
4248				return (0);
4249			}
4250		}
4251	}
4252	/* not found. */
4253	return (-2);
4254}
4255
4256void
4257sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4258{
4259	struct sctpvtaghead *chain;
4260	struct sctp_tagblock *twait_block;
4261	int found = 0;
4262	int i;
4263
4264	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4265	if (!LIST_EMPTY(chain)) {
4266		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4267			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4268				if ((twait_block->vtag_block[i].v_tag == tag) &&
4269				    (twait_block->vtag_block[i].lport == lport) &&
4270				    (twait_block->vtag_block[i].rport == rport)) {
4271					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4272					twait_block->vtag_block[i].v_tag = 0;
4273					twait_block->vtag_block[i].lport = 0;
4274					twait_block->vtag_block[i].rport = 0;
4275					found = 1;
4276					break;
4277				}
4278			}
4279			if (found)
4280				break;
4281		}
4282	}
4283}
4284
4285int
4286sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4287{
4288	struct sctpvtaghead *chain;
4289	struct sctp_tagblock *twait_block;
4290	int found = 0;
4291	int i;
4292
4293	SCTP_INP_INFO_WLOCK();
4294	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4295	if (!LIST_EMPTY(chain)) {
4296		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4297			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4298				if ((twait_block->vtag_block[i].v_tag == tag) &&
4299				    (twait_block->vtag_block[i].lport == lport) &&
4300				    (twait_block->vtag_block[i].rport == rport)) {
4301					found = 1;
4302					break;
4303				}
4304			}
4305			if (found)
4306				break;
4307		}
4308	}
4309	SCTP_INP_INFO_WUNLOCK();
4310	return (found);
4311}
4312
4313
4314void
4315sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4316{
4317	struct sctpvtaghead *chain;
4318	struct sctp_tagblock *twait_block;
4319	struct timeval now;
4320	int set, i;
4321
4322	if (time == 0) {
4323		/* Its disabled */
4324		return;
4325	}
4326	(void)SCTP_GETTIME_TIMEVAL(&now);
4327	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4328	set = 0;
4329	if (!LIST_EMPTY(chain)) {
4330		/* Block(s) present, lets find space, and expire on the fly */
4331		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4332			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4333				if ((twait_block->vtag_block[i].v_tag == 0) &&
4334				    !set) {
4335					twait_block->vtag_block[i].tv_sec_at_expire =
4336					    now.tv_sec + time;
4337					twait_block->vtag_block[i].v_tag = tag;
4338					twait_block->vtag_block[i].lport = lport;
4339					twait_block->vtag_block[i].rport = rport;
4340					set = 1;
4341				} else if ((twait_block->vtag_block[i].v_tag) &&
4342				    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4343					/* Audit expires this guy */
4344					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4345					twait_block->vtag_block[i].v_tag = 0;
4346					twait_block->vtag_block[i].lport = 0;
4347					twait_block->vtag_block[i].rport = 0;
4348					if (set == 0) {
4349						/* Reuse it for my new tag */
4350						twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4351						twait_block->vtag_block[i].v_tag = tag;
4352						twait_block->vtag_block[i].lport = lport;
4353						twait_block->vtag_block[i].rport = rport;
4354						set = 1;
4355					}
4356				}
4357			}
4358			if (set) {
4359				/*
4360				 * We only do up to the block where we can
4361				 * place our tag for audits
4362				 */
4363				break;
4364			}
4365		}
4366	}
4367	/* Need to add a new block to chain */
4368	if (!set) {
4369		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4370		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4371		if (twait_block == NULL) {
4372#ifdef INVARIANTS
4373			panic("Can not alloc tagblock");
4374#endif
4375			return;
4376		}
4377		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4378		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4379		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4380		twait_block->vtag_block[0].v_tag = tag;
4381		twait_block->vtag_block[0].lport = lport;
4382		twait_block->vtag_block[0].rport = rport;
4383	}
4384}
4385
4386
4387static void
4388sctp_iterator_asoc_being_freed(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4389{
4390	struct sctp_iterator *it;
4391
4392	/*
4393	 * Unlock the tcb lock we do this so we avoid a dead lock scenario
4394	 * where the iterator is waiting on the TCB lock and the TCB lock is
4395	 * waiting on the iterator lock.
4396	 */
4397	it = stcb->asoc.stcb_starting_point_for_iterator;
4398	if (it == NULL) {
4399		return;
4400	}
4401	if (it->inp != stcb->sctp_ep) {
4402		/* hmm, focused on the wrong one? */
4403		return;
4404	}
4405	if (it->stcb != stcb) {
4406		return;
4407	}
4408	it->stcb = LIST_NEXT(stcb, sctp_tcblist);
4409	if (it->stcb == NULL) {
4410		/* done with all asoc's in this assoc */
4411		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
4412			it->inp = NULL;
4413		} else {
4414			it->inp = LIST_NEXT(inp, sctp_list);
4415		}
4416	}
4417}
4418
4419
4420/*-
4421 * Free the association after un-hashing the remote port. This
4422 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4423 * expect that the input to this function IS a locked TCB.
4424 * It will return 0, if it did NOT destroy the association (instead
4425 * it unlocks it. It will return NON-zero if it either destroyed the
4426 * association OR the association is already destroyed.
4427 */
4428int
4429sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4430{
4431	int i;
4432	struct sctp_association *asoc;
4433	struct sctp_nets *net, *prev;
4434	struct sctp_laddr *laddr;
4435	struct sctp_tmit_chunk *chk;
4436	struct sctp_asconf_addr *aparam;
4437	struct sctp_asconf_ack *aack;
4438	struct sctp_stream_reset_list *liste;
4439	struct sctp_queued_to_read *sq;
4440	struct sctp_stream_queue_pending *sp;
4441	sctp_sharedkey_t *shared_key;
4442	struct socket *so;
4443	int ccnt = 0;
4444	int cnt = 0;
4445
4446	/* first, lets purge the entry from the hash table. */
4447
4448#ifdef SCTP_LOG_CLOSING
4449	sctp_log_closing(inp, stcb, 6);
4450#endif
4451	if (stcb->asoc.state == 0) {
4452#ifdef SCTP_LOG_CLOSING
4453		sctp_log_closing(inp, NULL, 7);
4454#endif
4455		/* there is no asoc, really TSNH :-0 */
4456		return (1);
4457	}
4458	/* TEMP CODE */
4459	if (stcb->freed_from_where == 0) {
4460		/* Only record the first place free happened from */
4461		stcb->freed_from_where = from_location;
4462	}
4463	/* TEMP CODE */
4464
4465	asoc = &stcb->asoc;
4466	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4467	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4468		/* nothing around */
4469		so = NULL;
4470	else
4471		so = inp->sctp_socket;
4472
4473	/*
4474	 * We used timer based freeing if a reader or writer is in the way.
4475	 * So we first check if we are actually being called from a timer,
4476	 * if so we abort early if a reader or writer is still in the way.
4477	 */
4478	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4479	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4480		/*
4481		 * is it the timer driving us? if so are the reader/writers
4482		 * gone?
4483		 */
4484		if (stcb->asoc.refcnt) {
4485			/* nope, reader or writer in the way */
4486			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4487			/* no asoc destroyed */
4488			SCTP_TCB_UNLOCK(stcb);
4489#ifdef SCTP_LOG_CLOSING
4490			sctp_log_closing(inp, stcb, 8);
4491#endif
4492			return (0);
4493		}
4494	}
4495	/* now clean up any other timers */
4496	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
4497	asoc->hb_timer.self = NULL;
4498	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4499	asoc->dack_timer.self = NULL;
4500	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4501	/*-
4502	 * For stream reset we don't blast this unless
4503	 * it is a str-reset timer, it might be the
4504	 * free-asoc timer which we DON'T want to
4505	 * disturb.
4506	 */
4507	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4508		asoc->strreset_timer.self = NULL;
4509	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4510	asoc->asconf_timer.self = NULL;
4511	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4512	asoc->autoclose_timer.self = NULL;
4513	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4514	asoc->shut_guard_timer.self = NULL;
4515	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4516	asoc->delayed_event_timer.self = NULL;
4517	/* Mobility adaptation */
4518	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4519	asoc->delete_prim_timer.self = NULL;
4520	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4521		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
4522		net->fr_timer.self = NULL;
4523		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4524		net->rxt_timer.self = NULL;
4525		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4526		net->pmtu_timer.self = NULL;
4527	}
4528	/* Now the read queue needs to be cleaned up (only once) */
4529	cnt = 0;
4530	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4531		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4532		SCTP_INP_READ_LOCK(inp);
4533		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4534			if (sq->stcb == stcb) {
4535				sq->do_not_ref_stcb = 1;
4536				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4537				/*
4538				 * If there is no end, there never will be
4539				 * now.
4540				 */
4541				if (sq->end_added == 0) {
4542					/* Held for PD-API clear that. */
4543					sq->pdapi_aborted = 1;
4544					sq->held_length = 0;
4545					if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4546						/*
4547						 * Need to add a PD-API
4548						 * aborted indication.
4549						 * Setting the control_pdapi
4550						 * assures that it will be
4551						 * added right after this
4552						 * msg.
4553						 */
4554						uint32_t strseq;
4555
4556						stcb->asoc.control_pdapi = sq;
4557						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4558						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4559						    stcb,
4560						    SCTP_PARTIAL_DELIVERY_ABORTED,
4561						    (void *)&strseq,
4562						    SCTP_SO_LOCKED);
4563						stcb->asoc.control_pdapi = NULL;
4564					}
4565				}
4566				/* Add an end to wake them */
4567				sq->end_added = 1;
4568				cnt++;
4569			}
4570		}
4571		SCTP_INP_READ_UNLOCK(inp);
4572		if (stcb->block_entry) {
4573			cnt++;
4574			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4575			stcb->block_entry->error = ECONNRESET;
4576			stcb->block_entry = NULL;
4577		}
4578	}
4579	if (stcb->asoc.refcnt) {
4580		/*
4581		 * reader or writer in the way, we have hopefully given him
4582		 * something to chew on above.
4583		 */
4584		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4585		SCTP_TCB_UNLOCK(stcb);
4586		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4587		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4588			/* nothing around */
4589			so = NULL;
4590		if (so) {
4591			/* Wake any reader/writers */
4592			sctp_sorwakeup(inp, so);
4593			sctp_sowwakeup(inp, so);
4594		}
4595#ifdef SCTP_LOG_CLOSING
4596		sctp_log_closing(inp, stcb, 9);
4597#endif
4598		/* no asoc destroyed */
4599		return (0);
4600	}
4601#ifdef SCTP_LOG_CLOSING
4602	sctp_log_closing(inp, stcb, 10);
4603#endif
4604	/*
4605	 * When I reach here, no others want to kill the assoc yet.. and I
4606	 * own the lock. Now its possible an abort comes in when I do the
4607	 * lock exchange below to grab all the locks to do the final take
4608	 * out. to prevent this we increment the count, which will start a
4609	 * timer and blow out above thus assuring us that we hold exclusive
4610	 * killing of the asoc. Note that after getting back the TCB lock we
4611	 * will go ahead and increment the counter back up and stop any
4612	 * timer a passing stranger may have started :-S
4613	 */
4614	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4615		atomic_add_int(&stcb->asoc.refcnt, 1);
4616
4617		SCTP_TCB_UNLOCK(stcb);
4618
4619		SCTP_ITERATOR_LOCK();
4620		SCTP_INP_INFO_WLOCK();
4621		SCTP_INP_WLOCK(inp);
4622		SCTP_TCB_LOCK(stcb);
4623	}
4624	/* Double check the GONE flag */
4625	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4626	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4627		/* nothing around */
4628		so = NULL;
4629
4630	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4631	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4632		/*
4633		 * For TCP type we need special handling when we are
4634		 * connected. We also include the peel'ed off ones to.
4635		 */
4636		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4637			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4638			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4639			if (so) {
4640				SOCK_LOCK(so);
4641				if (so->so_rcv.sb_cc == 0) {
4642					so->so_state &= ~(SS_ISCONNECTING |
4643					    SS_ISDISCONNECTING |
4644					    SS_ISCONFIRMING |
4645					    SS_ISCONNECTED);
4646				}
4647				SOCK_UNLOCK(so);
4648				socantrcvmore(so);
4649				sctp_sowwakeup(inp, so);
4650				sctp_sorwakeup(inp, so);
4651				SCTP_SOWAKEUP(so);
4652			}
4653		}
4654	}
4655	/*
4656	 * Make it invalid too, that way if its about to run it will abort
4657	 * and return.
4658	 */
4659	sctp_iterator_asoc_being_freed(inp, stcb);
4660	/* re-increment the lock */
4661	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4662		atomic_add_int(&stcb->asoc.refcnt, -1);
4663	}
4664	asoc->state = 0;
4665	if (inp->sctp_tcbhash) {
4666		LIST_REMOVE(stcb, sctp_tcbhash);
4667	}
4668	if (stcb->asoc.in_asocid_hash) {
4669		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4670	}
4671	/* Now lets remove it from the list of ALL associations in the EP */
4672	LIST_REMOVE(stcb, sctp_tcblist);
4673	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4674		SCTP_INP_INCR_REF(inp);
4675		SCTP_INP_WUNLOCK(inp);
4676		SCTP_ITERATOR_UNLOCK();
4677	}
4678	/* pull from vtag hash */
4679	LIST_REMOVE(stcb, sctp_asocs);
4680	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4681	    inp->sctp_lport, stcb->rport);
4682
4683	/*
4684	 * Now restop the timers to be sure - this is paranoia at is finest!
4685	 */
4686	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4687	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
4688	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4689	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4690	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4691	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4692	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4693	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4694	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4695		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
4696		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4697		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4698	}
4699
4700	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4701	prev = NULL;
4702	/*
4703	 * The chunk lists and such SHOULD be empty but we check them just
4704	 * in case.
4705	 */
4706	/* anything on the wheel needs to be removed */
4707	for (i = 0; i < asoc->streamoutcnt; i++) {
4708		struct sctp_stream_out *outs;
4709
4710		outs = &asoc->strmout[i];
4711		/* now clean up any chunks here */
4712		sp = TAILQ_FIRST(&outs->outqueue);
4713		while (sp) {
4714			TAILQ_REMOVE(&outs->outqueue, sp, next);
4715			if (sp->data) {
4716				sctp_m_freem(sp->data);
4717				sp->data = NULL;
4718				sp->tail_mbuf = NULL;
4719			}
4720			sctp_free_remote_addr(sp->net);
4721			sctp_free_spbufspace(stcb, asoc, sp);
4722			if (sp->holds_key_ref)
4723				sctp_auth_key_release(stcb, sp->auth_keyid);
4724			/* Free the zone stuff  */
4725			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_strmoq), sp);
4726			SCTP_DECR_STRMOQ_COUNT();
4727			/* sa_ignore FREED_MEMORY */
4728			sp = TAILQ_FIRST(&outs->outqueue);
4729		}
4730	}
4731
4732	/* sa_ignore FREED_MEMORY */
4733	while ((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) {
4734		TAILQ_REMOVE(&asoc->resetHead, liste, next_resp);
4735		SCTP_FREE(liste, SCTP_M_STRESET);
4736	}
4737
4738	sq = TAILQ_FIRST(&asoc->pending_reply_queue);
4739	while (sq) {
4740		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4741		if (sq->data) {
4742			sctp_m_freem(sq->data);
4743			sq->data = NULL;
4744		}
4745		sctp_free_remote_addr(sq->whoFrom);
4746		sq->whoFrom = NULL;
4747		sq->stcb = NULL;
4748		/* Free the ctl entry */
4749		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4750		SCTP_DECR_READQ_COUNT();
4751		/* sa_ignore FREED_MEMORY */
4752		sq = TAILQ_FIRST(&asoc->pending_reply_queue);
4753	}
4754
4755	chk = TAILQ_FIRST(&asoc->free_chunks);
4756	while (chk) {
4757		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4758		if (chk->data) {
4759			sctp_m_freem(chk->data);
4760			chk->data = NULL;
4761		}
4762		if (chk->holds_key_ref)
4763			sctp_auth_key_release(stcb, chk->auth_keyid);
4764		ccnt++;
4765		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4766		SCTP_DECR_CHK_COUNT();
4767		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4768		asoc->free_chunk_cnt--;
4769		/* sa_ignore FREED_MEMORY */
4770		chk = TAILQ_FIRST(&asoc->free_chunks);
4771	}
4772	/* pending send queue SHOULD be empty */
4773	if (!TAILQ_EMPTY(&asoc->send_queue)) {
4774		chk = TAILQ_FIRST(&asoc->send_queue);
4775		while (chk) {
4776			TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4777			if (chk->data) {
4778				sctp_m_freem(chk->data);
4779				chk->data = NULL;
4780			}
4781			if (chk->holds_key_ref)
4782				sctp_auth_key_release(stcb, chk->auth_keyid);
4783			ccnt++;
4784			sctp_free_remote_addr(chk->whoTo);
4785			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4786			SCTP_DECR_CHK_COUNT();
4787			/* sa_ignore FREED_MEMORY */
4788			chk = TAILQ_FIRST(&asoc->send_queue);
4789		}
4790	}
4791/*
4792  if (ccnt) {
4793  printf("Freed %d from send_queue\n", ccnt);
4794  ccnt = 0;
4795  }
4796*/
4797	/* sent queue SHOULD be empty */
4798	if (!TAILQ_EMPTY(&asoc->sent_queue)) {
4799		chk = TAILQ_FIRST(&asoc->sent_queue);
4800		while (chk) {
4801			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4802			if (chk->data) {
4803				sctp_m_freem(chk->data);
4804				chk->data = NULL;
4805			}
4806			if (chk->holds_key_ref)
4807				sctp_auth_key_release(stcb, chk->auth_keyid);
4808			ccnt++;
4809			sctp_free_remote_addr(chk->whoTo);
4810			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4811			SCTP_DECR_CHK_COUNT();
4812			/* sa_ignore FREED_MEMORY */
4813			chk = TAILQ_FIRST(&asoc->sent_queue);
4814		}
4815	}
4816/*
4817  if (ccnt) {
4818  printf("Freed %d from sent_queue\n", ccnt);
4819  ccnt = 0;
4820  }
4821*/
4822	/* control queue MAY not be empty */
4823	if (!TAILQ_EMPTY(&asoc->control_send_queue)) {
4824		chk = TAILQ_FIRST(&asoc->control_send_queue);
4825		while (chk) {
4826			TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
4827			if (chk->data) {
4828				sctp_m_freem(chk->data);
4829				chk->data = NULL;
4830			}
4831			if (chk->holds_key_ref)
4832				sctp_auth_key_release(stcb, chk->auth_keyid);
4833			ccnt++;
4834			sctp_free_remote_addr(chk->whoTo);
4835			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4836			SCTP_DECR_CHK_COUNT();
4837			/* sa_ignore FREED_MEMORY */
4838			chk = TAILQ_FIRST(&asoc->control_send_queue);
4839		}
4840	}
4841/*
4842  if (ccnt) {
4843  printf("Freed %d from ctrl_queue\n", ccnt);
4844  ccnt = 0;
4845  }
4846*/
4847
4848	/* ASCONF queue MAY not be empty */
4849	if (!TAILQ_EMPTY(&asoc->asconf_send_queue)) {
4850		chk = TAILQ_FIRST(&asoc->asconf_send_queue);
4851		while (chk) {
4852			TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
4853			if (chk->data) {
4854				sctp_m_freem(chk->data);
4855				chk->data = NULL;
4856			}
4857			if (chk->holds_key_ref)
4858				sctp_auth_key_release(stcb, chk->auth_keyid);
4859			ccnt++;
4860			sctp_free_remote_addr(chk->whoTo);
4861			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4862			SCTP_DECR_CHK_COUNT();
4863			/* sa_ignore FREED_MEMORY */
4864			chk = TAILQ_FIRST(&asoc->asconf_send_queue);
4865		}
4866	}
4867/*
4868  if (ccnt) {
4869  printf("Freed %d from asconf_queue\n", ccnt);
4870  ccnt = 0;
4871  }
4872*/
4873	if (!TAILQ_EMPTY(&asoc->reasmqueue)) {
4874		chk = TAILQ_FIRST(&asoc->reasmqueue);
4875		while (chk) {
4876			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
4877			if (chk->data) {
4878				sctp_m_freem(chk->data);
4879				chk->data = NULL;
4880			}
4881			if (chk->holds_key_ref)
4882				sctp_auth_key_release(stcb, chk->auth_keyid);
4883			sctp_free_remote_addr(chk->whoTo);
4884			ccnt++;
4885			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4886			SCTP_DECR_CHK_COUNT();
4887			/* sa_ignore FREED_MEMORY */
4888			chk = TAILQ_FIRST(&asoc->reasmqueue);
4889		}
4890	}
4891/*
4892  if (ccnt) {
4893  printf("Freed %d from reasm_queue\n", ccnt);
4894  ccnt = 0;
4895  }
4896*/
4897	if (asoc->mapping_array) {
4898		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4899		asoc->mapping_array = NULL;
4900	}
4901	if (asoc->nr_mapping_array) {
4902		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4903		asoc->nr_mapping_array = NULL;
4904	}
4905	/* the stream outs */
4906	if (asoc->strmout) {
4907		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4908		asoc->strmout = NULL;
4909	}
4910	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
4911	if (asoc->strmin) {
4912		struct sctp_queued_to_read *ctl;
4913
4914		for (i = 0; i < asoc->streamincnt; i++) {
4915			if (!TAILQ_EMPTY(&asoc->strmin[i].inqueue)) {
4916				/* We have somethings on the streamin queue */
4917				ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
4918				while (ctl) {
4919					TAILQ_REMOVE(&asoc->strmin[i].inqueue,
4920					    ctl, next);
4921					sctp_free_remote_addr(ctl->whoFrom);
4922					if (ctl->data) {
4923						sctp_m_freem(ctl->data);
4924						ctl->data = NULL;
4925					}
4926					/*
4927					 * We don't free the address here
4928					 * since all the net's were freed
4929					 * above.
4930					 */
4931					SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
4932					SCTP_DECR_READQ_COUNT();
4933					ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
4934				}
4935			}
4936		}
4937		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
4938		asoc->strmin = NULL;
4939	}
4940	asoc->streamincnt = 0;
4941	while (!TAILQ_EMPTY(&asoc->nets)) {
4942		/* sa_ignore FREED_MEMORY */
4943		net = TAILQ_FIRST(&asoc->nets);
4944		/* pull from list */
4945		if ((SCTP_BASE_INFO(ipi_count_raddr) == 0) || (prev == net)) {
4946#ifdef INVARIANTS
4947			panic("no net's left alloc'ed, or list points to itself");
4948#endif
4949			break;
4950		}
4951		prev = net;
4952		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4953		sctp_free_remote_addr(net);
4954	}
4955
4956	while (!LIST_EMPTY(&asoc->sctp_restricted_addrs)) {
4957		/* sa_ignore FREED_MEMORY */
4958		laddr = LIST_FIRST(&asoc->sctp_restricted_addrs);
4959		sctp_remove_laddr(laddr);
4960	}
4961
4962	/* pending asconf (address) parameters */
4963	while (!TAILQ_EMPTY(&asoc->asconf_queue)) {
4964		/* sa_ignore FREED_MEMORY */
4965		aparam = TAILQ_FIRST(&asoc->asconf_queue);
4966		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
4967		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
4968	}
4969	while (!TAILQ_EMPTY(&asoc->asconf_ack_sent)) {
4970		/* sa_ignore FREED_MEMORY */
4971		aack = TAILQ_FIRST(&asoc->asconf_ack_sent);
4972		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
4973		if (aack->data != NULL) {
4974			sctp_m_freem(aack->data);
4975		}
4976		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
4977	}
4978	/* clean up auth stuff */
4979	if (asoc->local_hmacs)
4980		sctp_free_hmaclist(asoc->local_hmacs);
4981	if (asoc->peer_hmacs)
4982		sctp_free_hmaclist(asoc->peer_hmacs);
4983
4984	if (asoc->local_auth_chunks)
4985		sctp_free_chunklist(asoc->local_auth_chunks);
4986	if (asoc->peer_auth_chunks)
4987		sctp_free_chunklist(asoc->peer_auth_chunks);
4988
4989	sctp_free_authinfo(&asoc->authinfo);
4990
4991	shared_key = LIST_FIRST(&asoc->shared_keys);
4992	while (shared_key) {
4993		LIST_REMOVE(shared_key, next);
4994		sctp_free_sharedkey(shared_key);
4995		/* sa_ignore FREED_MEMORY */
4996		shared_key = LIST_FIRST(&asoc->shared_keys);
4997	}
4998
4999	/* Insert new items here :> */
5000
5001	/* Get rid of LOCK */
5002	SCTP_TCB_LOCK_DESTROY(stcb);
5003	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5004	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5005		SCTP_INP_INFO_WUNLOCK();
5006		SCTP_INP_RLOCK(inp);
5007	}
5008#ifdef SCTP_TRACK_FREED_ASOCS
5009	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5010		/* now clean up the tasoc itself */
5011		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5012		SCTP_DECR_ASOC_COUNT();
5013	} else {
5014		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5015	}
5016#else
5017	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5018	SCTP_DECR_ASOC_COUNT();
5019#endif
5020	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5021		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5022			/*
5023			 * If its NOT the inp_free calling us AND sctp_close
5024			 * as been called, we call back...
5025			 */
5026			SCTP_INP_RUNLOCK(inp);
5027			/*
5028			 * This will start the kill timer (if we are the
5029			 * lastone) since we hold an increment yet. But this
5030			 * is the only safe way to do this since otherwise
5031			 * if the socket closes at the same time we are here
5032			 * we might collide in the cleanup.
5033			 */
5034			sctp_inpcb_free(inp,
5035			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5036			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5037			SCTP_INP_DECR_REF(inp);
5038			goto out_of;
5039		} else {
5040			/* The socket is still open. */
5041			SCTP_INP_DECR_REF(inp);
5042		}
5043	}
5044	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5045		SCTP_INP_RUNLOCK(inp);
5046	}
5047out_of:
5048	/* destroyed the asoc */
5049#ifdef SCTP_LOG_CLOSING
5050	sctp_log_closing(inp, NULL, 11);
5051#endif
5052	return (1);
5053}
5054
5055
5056
5057/*
5058 * determine if a destination is "reachable" based upon the addresses bound
5059 * to the current endpoint (e.g. only v4 or v6 currently bound)
5060 */
5061/*
5062 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5063 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5064 * types bound as its endpoint
5065 */
5066int
5067sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5068{
5069	struct sctp_inpcb *inp;
5070	int answer;
5071
5072	/*
5073	 * No locks here, the TCB, in all cases is already locked and an
5074	 * assoc is up. There is either a INP lock by the caller applied (in
5075	 * asconf case when deleting an address) or NOT in the HB case,
5076	 * however if HB then the INP increment is up and the INP will not
5077	 * be removed (on top of the fact that we have a TCB lock). So we
5078	 * only want to read the sctp_flags, which is either bound-all or
5079	 * not.. no protection needed since once an assoc is up you can't be
5080	 * changing your binding.
5081	 */
5082	inp = stcb->sctp_ep;
5083	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5084		/* if bound all, destination is not restricted */
5085		/*
5086		 * RRS: Question during lock work: Is this correct? If you
5087		 * are bound-all you still might need to obey the V4--V6
5088		 * flags??? IMO this bound-all stuff needs to be removed!
5089		 */
5090		return (1);
5091	}
5092	/* NOTE: all "scope" checks are done when local addresses are added */
5093	if (destaddr->sa_family == AF_INET6) {
5094		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5095	} else if (destaddr->sa_family == AF_INET) {
5096		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5097	} else {
5098		/* invalid family, so it's unreachable */
5099		answer = 0;
5100	}
5101	return (answer);
5102}
5103
5104/*
5105 * update the inp_vflags on an endpoint
5106 */
5107static void
5108sctp_update_ep_vflag(struct sctp_inpcb *inp)
5109{
5110	struct sctp_laddr *laddr;
5111
5112	/* first clear the flag */
5113	inp->ip_inp.inp.inp_vflag = 0;
5114	/* set the flag based on addresses on the ep list */
5115	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5116		if (laddr->ifa == NULL) {
5117			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5118			    __FUNCTION__);
5119			continue;
5120		}
5121		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5122			continue;
5123		}
5124		if (laddr->ifa->address.sa.sa_family == AF_INET6) {
5125			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5126		} else if (laddr->ifa->address.sa.sa_family == AF_INET) {
5127			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5128		}
5129	}
5130}
5131
5132/*
5133 * Add the address to the endpoint local address list There is nothing to be
5134 * done if we are bound to all addresses
5135 */
5136void
5137sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5138{
5139	struct sctp_laddr *laddr;
5140	int fnd, error = 0;
5141
5142	fnd = 0;
5143
5144	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5145		/* You are already bound to all. You have it already */
5146		return;
5147	}
5148	if (ifa->address.sa.sa_family == AF_INET6) {
5149		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5150			/* Can't bind a non-useable addr. */
5151			return;
5152		}
5153	}
5154	/* first, is it already present? */
5155	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5156		if (laddr->ifa == ifa) {
5157			fnd = 1;
5158			break;
5159		}
5160	}
5161
5162	if (fnd == 0) {
5163		/* Not in the ep list */
5164		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5165		if (error != 0)
5166			return;
5167		inp->laddr_count++;
5168		/* update inp_vflag flags */
5169		if (ifa->address.sa.sa_family == AF_INET6) {
5170			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5171		} else if (ifa->address.sa.sa_family == AF_INET) {
5172			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5173		}
5174	}
5175	return;
5176}
5177
5178
5179/*
5180 * select a new (hopefully reachable) destination net (should only be used
5181 * when we deleted an ep addr that is the only usable source address to reach
5182 * the destination net)
5183 */
5184static void
5185sctp_select_primary_destination(struct sctp_tcb *stcb)
5186{
5187	struct sctp_nets *net;
5188
5189	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5190		/* for now, we'll just pick the first reachable one we find */
5191		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5192			continue;
5193		if (sctp_destination_is_reachable(stcb,
5194		    (struct sockaddr *)&net->ro._l_addr)) {
5195			/* found a reachable destination */
5196			stcb->asoc.primary_destination = net;
5197		}
5198	}
5199	/* I can't there from here! ...we're gonna die shortly... */
5200}
5201
5202
5203/*
5204 * Delete the address from the endpoint local address list There is nothing
5205 * to be done if we are bound to all addresses
5206 */
5207void
5208sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5209{
5210	struct sctp_laddr *laddr;
5211	int fnd;
5212
5213	fnd = 0;
5214	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5215		/* You are already bound to all. You have it already */
5216		return;
5217	}
5218	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5219		if (laddr->ifa == ifa) {
5220			fnd = 1;
5221			break;
5222		}
5223	}
5224	if (fnd && (inp->laddr_count < 2)) {
5225		/* can't delete unless there are at LEAST 2 addresses */
5226		return;
5227	}
5228	if (fnd) {
5229		/*
5230		 * clean up any use of this address go through our
5231		 * associations and clear any last_used_address that match
5232		 * this one for each assoc, see if a new primary_destination
5233		 * is needed
5234		 */
5235		struct sctp_tcb *stcb;
5236
5237		/* clean up "next_addr_touse" */
5238		if (inp->next_addr_touse == laddr)
5239			/* delete this address */
5240			inp->next_addr_touse = NULL;
5241
5242		/* clean up "last_used_address" */
5243		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5244			struct sctp_nets *net;
5245
5246			SCTP_TCB_LOCK(stcb);
5247			if (stcb->asoc.last_used_address == laddr)
5248				/* delete this address */
5249				stcb->asoc.last_used_address = NULL;
5250			/*
5251			 * Now spin through all the nets and purge any ref
5252			 * to laddr
5253			 */
5254			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5255				if (net->ro._s_addr &&
5256				    (net->ro._s_addr->ifa == laddr->ifa)) {
5257					/* Yep, purge src address selected */
5258					sctp_rtentry_t *rt;
5259
5260					/* delete this address if cached */
5261					rt = net->ro.ro_rt;
5262					if (rt != NULL) {
5263						RTFREE(rt);
5264						net->ro.ro_rt = NULL;
5265					}
5266					sctp_free_ifa(net->ro._s_addr);
5267					net->ro._s_addr = NULL;
5268					net->src_addr_selected = 0;
5269				}
5270			}
5271			SCTP_TCB_UNLOCK(stcb);
5272		}		/* for each tcb */
5273		/* remove it from the ep list */
5274		sctp_remove_laddr(laddr);
5275		inp->laddr_count--;
5276		/* update inp_vflag flags */
5277		sctp_update_ep_vflag(inp);
5278	}
5279	return;
5280}
5281
5282/*
5283 * Add the address to the TCB local address restricted list.
5284 * This is a "pending" address list (eg. addresses waiting for an
5285 * ASCONF-ACK response) and cannot be used as a valid source address.
5286 */
5287void
5288sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5289{
5290	struct sctp_inpcb *inp;
5291	struct sctp_laddr *laddr;
5292	struct sctpladdr *list;
5293
5294	/*
5295	 * Assumes TCB is locked.. and possibly the INP. May need to
5296	 * confirm/fix that if we need it and is not the case.
5297	 */
5298	list = &stcb->asoc.sctp_restricted_addrs;
5299
5300	inp = stcb->sctp_ep;
5301	if (ifa->address.sa.sa_family == AF_INET6) {
5302		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5303			/* Can't bind a non-existent addr. */
5304			return;
5305		}
5306	}
5307	/* does the address already exist? */
5308	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5309		if (laddr->ifa == ifa) {
5310			return;
5311		}
5312	}
5313
5314	/* add to the list */
5315	(void)sctp_insert_laddr(list, ifa, 0);
5316	return;
5317}
5318
5319/*
5320 * insert an laddr entry with the given ifa for the desired list
5321 */
5322int
5323sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5324{
5325	struct sctp_laddr *laddr;
5326
5327	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5328	if (laddr == NULL) {
5329		/* out of memory? */
5330		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5331		return (EINVAL);
5332	}
5333	SCTP_INCR_LADDR_COUNT();
5334	bzero(laddr, sizeof(*laddr));
5335	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5336	laddr->ifa = ifa;
5337	laddr->action = act;
5338	atomic_add_int(&ifa->refcount, 1);
5339	/* insert it */
5340	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5341
5342	return (0);
5343}
5344
5345/*
5346 * Remove an laddr entry from the local address list (on an assoc)
5347 */
5348void
5349sctp_remove_laddr(struct sctp_laddr *laddr)
5350{
5351
5352	/* remove from the list */
5353	LIST_REMOVE(laddr, sctp_nxt_addr);
5354	sctp_free_ifa(laddr->ifa);
5355	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5356	SCTP_DECR_LADDR_COUNT();
5357}
5358
5359/*
5360 * Remove a local address from the TCB local address restricted list
5361 */
5362void
5363sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5364{
5365	struct sctp_inpcb *inp;
5366	struct sctp_laddr *laddr;
5367
5368	/*
5369	 * This is called by asconf work. It is assumed that a) The TCB is
5370	 * locked and b) The INP is locked. This is true in as much as I can
5371	 * trace through the entry asconf code where I did these locks.
5372	 * Again, the ASCONF code is a bit different in that it does lock
5373	 * the INP during its work often times. This must be since we don't
5374	 * want other proc's looking up things while what they are looking
5375	 * up is changing :-D
5376	 */
5377
5378	inp = stcb->sctp_ep;
5379	/* if subset bound and don't allow ASCONF's, can't delete last */
5380	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5381	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5382		if (stcb->sctp_ep->laddr_count < 2) {
5383			/* can't delete last address */
5384			return;
5385		}
5386	}
5387	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5388		/* remove the address if it exists */
5389		if (laddr->ifa == NULL)
5390			continue;
5391		if (laddr->ifa == ifa) {
5392			sctp_remove_laddr(laddr);
5393			return;
5394		}
5395	}
5396
5397	/* address not found! */
5398	return;
5399}
5400
5401/*
5402 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5403 */
5404/* sysctl */
5405static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5406static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5407
5408void
5409sctp_pcb_init()
5410{
5411	/*
5412	 * SCTP initialization for the PCB structures should be called by
5413	 * the sctp_init() funciton.
5414	 */
5415	int i;
5416	struct timeval tv;
5417
5418	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5419		/* error I was called twice */
5420		return;
5421	}
5422	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5423
5424#if defined(SCTP_LOCAL_TRACE_BUF)
5425	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5426#endif
5427	(void)SCTP_GETTIME_TIMEVAL(&tv);
5428	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5429	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5430	/* init the empty list of (All) Endpoints */
5431	LIST_INIT(&SCTP_BASE_INFO(listhead));
5432
5433	/* init the iterator head */
5434	TAILQ_INIT(&SCTP_BASE_INFO(iteratorhead));
5435
5436	/* init the hash table of endpoints */
5437	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5438	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5439	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5440	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5441	    &SCTP_BASE_INFO(hashasocmark));
5442	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5443	    &SCTP_BASE_INFO(hashmark));
5444	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5445	    &SCTP_BASE_INFO(hashtcpmark));
5446	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5447
5448
5449	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5450	    &SCTP_BASE_INFO(hashvrfmark));
5451
5452	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5453	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5454	/* init the zones */
5455	/*
5456	 * FIX ME: Should check for NULL returns, but if it does fail we are
5457	 * doomed to panic anyways... add later maybe.
5458	 */
5459	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5460	    sizeof(struct sctp_inpcb), maxsockets);
5461
5462	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5463	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5464
5465	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5466	    sizeof(struct sctp_laddr),
5467	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5468
5469	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5470	    sizeof(struct sctp_nets),
5471	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5472
5473	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5474	    sizeof(struct sctp_tmit_chunk),
5475	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5476
5477	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5478	    sizeof(struct sctp_queued_to_read),
5479	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5480
5481	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5482	    sizeof(struct sctp_stream_queue_pending),
5483	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5484
5485	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5486	    sizeof(struct sctp_asconf),
5487	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5488
5489	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5490	    sizeof(struct sctp_asconf_ack),
5491	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5492
5493
5494	/* Master Lock INIT for info structure */
5495	SCTP_INP_INFO_LOCK_INIT();
5496	SCTP_STATLOG_INIT_LOCK();
5497	SCTP_ITERATOR_LOCK_INIT();
5498
5499	SCTP_IPI_COUNT_INIT();
5500	SCTP_IPI_ADDR_INIT();
5501	SCTP_IPI_ITERATOR_WQ_INIT();
5502#ifdef SCTP_PACKET_LOGGING
5503	SCTP_IP_PKTLOG_INIT();
5504#endif
5505	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5506
5507	/* not sure if we need all the counts */
5508	SCTP_BASE_INFO(ipi_count_ep) = 0;
5509	/* assoc/tcb zone info */
5510	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5511	/* local addrlist zone info */
5512	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5513	/* remote addrlist zone info */
5514	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5515	/* chunk info */
5516	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5517
5518	/* socket queue zone info */
5519	SCTP_BASE_INFO(ipi_count_readq) = 0;
5520
5521	/* stream out queue cont */
5522	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5523
5524	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5525	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5526
5527	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5528
5529	/* Init the TIMEWAIT list */
5530	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5531		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait[i]));
5532	}
5533
5534#if defined(SCTP_USE_THREAD_BASED_ITERATOR)
5535	SCTP_BASE_INFO(iterator_running) = 0;
5536	SCTP_BASE_INFO(threads_must_exit) = 0;
5537	sctp_startup_iterator();
5538#endif
5539
5540	/*
5541	 * INIT the default VRF which for BSD is the only one, other O/S's
5542	 * may have more. But initially they must start with one and then
5543	 * add the VRF's as addresses are added.
5544	 */
5545	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5546
5547}
5548
5549/*
5550 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5551 */
5552void
5553sctp_pcb_finish(void)
5554{
5555	struct sctp_vrflist *vrf_bucket;
5556	struct sctp_vrf *vrf;
5557	struct sctp_ifn *ifn;
5558	struct sctp_ifa *ifa;
5559	struct sctpvtaghead *chain;
5560	struct sctp_tagblock *twait_block, *prev_twait_block;
5561	struct sctp_laddr *wi;
5562	struct sctp_iterator *it;
5563	int i;
5564
5565#if defined(SCTP_USE_THREAD_BASED_ITERATOR)
5566	SCTP_BASE_INFO(threads_must_exit) = 1;
5567	/* Wake the thread up so it will exit now */
5568	sctp_wakeup_iterator();
5569
5570#endif
5571	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5572	SCTP_IPI_ITERATOR_WQ_LOCK();
5573	while ((wi = LIST_FIRST(&SCTP_BASE_INFO(addr_wq))) != NULL) {
5574		LIST_REMOVE(wi, sctp_nxt_addr);
5575		SCTP_DECR_LADDR_COUNT();
5576		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5577	}
5578	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5579	while ((it = TAILQ_FIRST(&SCTP_BASE_INFO(iteratorhead))) != NULL) {
5580		if (it->function_atend != NULL) {
5581			(*it->function_atend) (it->pointer, it->val);
5582		}
5583		TAILQ_REMOVE(&SCTP_BASE_INFO(iteratorhead), it, sctp_nxt_itr);
5584		SCTP_FREE(it, SCTP_M_ITER);
5585	}
5586
5587	/*
5588	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5589	 * destroyed first).
5590	 */
5591	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5592	while ((vrf = LIST_FIRST(vrf_bucket)) != NULL) {
5593		while ((ifn = LIST_FIRST(&vrf->ifnlist)) != NULL) {
5594			while ((ifa = LIST_FIRST(&ifn->ifalist)) != NULL) {
5595				/* free the ifa */
5596				LIST_REMOVE(ifa, next_bucket);
5597				LIST_REMOVE(ifa, next_ifa);
5598				SCTP_FREE(ifa, SCTP_M_IFA);
5599			}
5600			/* free the ifn */
5601			LIST_REMOVE(ifn, next_bucket);
5602			LIST_REMOVE(ifn, next_ifn);
5603			SCTP_FREE(ifn, SCTP_M_IFN);
5604		}
5605		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5606		/* free the vrf */
5607		LIST_REMOVE(vrf, next_vrf);
5608		SCTP_FREE(vrf, SCTP_M_VRF);
5609	}
5610	/* free the vrf hashes */
5611	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5612	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5613
5614	/*
5615	 * free the TIMEWAIT list elements malloc'd in the function
5616	 * sctp_add_vtag_to_timewait()...
5617	 */
5618	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5619		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5620		if (!LIST_EMPTY(chain)) {
5621			prev_twait_block = NULL;
5622			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5623				if (prev_twait_block) {
5624					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5625				}
5626				prev_twait_block = twait_block;
5627			}
5628			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5629		}
5630	}
5631
5632	/* free the locks and mutexes */
5633#ifdef SCTP_PACKET_LOGGING
5634	SCTP_IP_PKTLOG_DESTROY();
5635#endif
5636	SCTP_IPI_ADDR_DESTROY();
5637	SCTP_ITERATOR_LOCK_DESTROY();
5638	SCTP_STATLOG_DESTROY();
5639	SCTP_INP_INFO_LOCK_DESTROY();
5640
5641	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
5642	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
5643	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
5644	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
5645	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
5646	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
5647	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
5648	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
5649	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
5650	/* Get rid of other stuff to */
5651	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
5652		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
5653	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
5654		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
5655	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
5656		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
5657
5658}
5659
5660
5661int
5662sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
5663    int iphlen, int offset, int limit, struct sctphdr *sh,
5664    struct sockaddr *altsa)
5665{
5666	/*
5667	 * grub through the INIT pulling addresses and loading them to the
5668	 * nets structure in the asoc. The from address in the mbuf should
5669	 * also be loaded (if it is not already). This routine can be called
5670	 * with either INIT or INIT-ACK's as long as the m points to the IP
5671	 * packet and the offset points to the beginning of the parameters.
5672	 */
5673	struct sctp_inpcb *inp, *l_inp;
5674	struct sctp_nets *net, *net_tmp;
5675	struct ip *iph;
5676	struct sctp_paramhdr *phdr, parm_buf;
5677	struct sctp_tcb *stcb_tmp;
5678	uint16_t ptype, plen;
5679	struct sockaddr *sa;
5680	struct sockaddr_storage dest_store;
5681	struct sockaddr *local_sa = (struct sockaddr *)&dest_store;
5682	struct sockaddr_in sin;
5683	struct sockaddr_in6 sin6;
5684	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
5685	struct sctp_auth_random *p_random = NULL;
5686	uint16_t random_len = 0;
5687	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
5688	struct sctp_auth_hmac_algo *hmacs = NULL;
5689	uint16_t hmacs_len = 0;
5690	uint8_t saw_asconf = 0;
5691	uint8_t saw_asconf_ack = 0;
5692	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
5693	struct sctp_auth_chunk_list *chunks = NULL;
5694	uint16_t num_chunks = 0;
5695	sctp_key_t *new_key;
5696	uint32_t keylen;
5697	int got_random = 0, got_hmacs = 0, got_chklist = 0;
5698
5699	/* First get the destination address setup too. */
5700	memset(&sin, 0, sizeof(sin));
5701	memset(&sin6, 0, sizeof(sin6));
5702
5703	sin.sin_family = AF_INET;
5704	sin.sin_len = sizeof(sin);
5705	sin.sin_port = stcb->rport;
5706
5707	sin6.sin6_family = AF_INET6;
5708	sin6.sin6_len = sizeof(struct sockaddr_in6);
5709	sin6.sin6_port = stcb->rport;
5710	if (altsa == NULL) {
5711		iph = mtod(m, struct ip *);
5712		switch (iph->ip_v) {
5713		case IPVERSION:
5714			{
5715				/* its IPv4 */
5716				struct sockaddr_in *sin_2;
5717
5718				sin_2 = (struct sockaddr_in *)(local_sa);
5719				memset(sin_2, 0, sizeof(sin));
5720				sin_2->sin_family = AF_INET;
5721				sin_2->sin_len = sizeof(sin);
5722				sin_2->sin_port = sh->dest_port;
5723				sin_2->sin_addr.s_addr = iph->ip_dst.s_addr;
5724				sin.sin_addr = iph->ip_src;
5725				sa = (struct sockaddr *)&sin;
5726				break;
5727			}
5728#ifdef INET6
5729		case IPV6_VERSION >> 4:
5730			{
5731				/* its IPv6 */
5732				struct ip6_hdr *ip6;
5733				struct sockaddr_in6 *sin6_2;
5734
5735				ip6 = mtod(m, struct ip6_hdr *);
5736				sin6_2 = (struct sockaddr_in6 *)(local_sa);
5737				memset(sin6_2, 0, sizeof(sin6));
5738				sin6_2->sin6_family = AF_INET6;
5739				sin6_2->sin6_len = sizeof(struct sockaddr_in6);
5740				sin6_2->sin6_port = sh->dest_port;
5741				sin6.sin6_addr = ip6->ip6_src;
5742				sa = (struct sockaddr *)&sin6;
5743				break;
5744			}
5745#endif
5746		default:
5747			sa = NULL;
5748			break;
5749		}
5750	} else {
5751		/*
5752		 * For cookies we use the src address NOT from the packet
5753		 * but from the original INIT
5754		 */
5755		sa = altsa;
5756	}
5757	/* Turn off ECN until we get through all params */
5758	stcb->asoc.ecn_allowed = 0;
5759	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5760		/* mark all addresses that we have currently on the list */
5761		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
5762	}
5763	/* does the source address already exist? if so skip it */
5764	l_inp = inp = stcb->sctp_ep;
5765
5766	atomic_add_int(&stcb->asoc.refcnt, 1);
5767	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, local_sa, stcb);
5768	atomic_add_int(&stcb->asoc.refcnt, -1);
5769
5770	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
5771		/* we must add the source address */
5772		/* no scope set here since we have a tcb already. */
5773		if ((sa->sa_family == AF_INET) &&
5774		    (stcb->asoc.ipv4_addr_legal)) {
5775			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
5776				return (-1);
5777			}
5778		} else if ((sa->sa_family == AF_INET6) &&
5779		    (stcb->asoc.ipv6_addr_legal)) {
5780			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
5781				return (-2);
5782			}
5783		}
5784	} else {
5785		if (net_tmp != NULL && stcb_tmp == stcb) {
5786			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
5787		} else if (stcb_tmp != stcb) {
5788			/* It belongs to another association? */
5789			if (stcb_tmp)
5790				SCTP_TCB_UNLOCK(stcb_tmp);
5791			return (-3);
5792		}
5793	}
5794	if (stcb->asoc.state == 0) {
5795		/* the assoc was freed? */
5796		return (-4);
5797	}
5798	/*
5799	 * peer must explicitly turn this on. This may have been initialized
5800	 * to be "on" in order to allow local addr changes while INIT's are
5801	 * in flight.
5802	 */
5803	stcb->asoc.peer_supports_asconf = 0;
5804	/* now we must go through each of the params. */
5805	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
5806	while (phdr) {
5807		ptype = ntohs(phdr->param_type);
5808		plen = ntohs(phdr->param_length);
5809		/*
5810		 * printf("ptype => %0x, plen => %d\n", (uint32_t)ptype,
5811		 * (int)plen);
5812		 */
5813		if (offset + plen > limit) {
5814			break;
5815		}
5816		if (plen == 0) {
5817			break;
5818		}
5819		if (ptype == SCTP_IPV4_ADDRESS) {
5820			if (stcb->asoc.ipv4_addr_legal) {
5821				struct sctp_ipv4addr_param *p4, p4_buf;
5822
5823				/* ok get the v4 address and check/add */
5824				phdr = sctp_get_next_param(m, offset,
5825				    (struct sctp_paramhdr *)&p4_buf,
5826				    sizeof(p4_buf));
5827				if (plen != sizeof(struct sctp_ipv4addr_param) ||
5828				    phdr == NULL) {
5829					return (-5);
5830				}
5831				p4 = (struct sctp_ipv4addr_param *)phdr;
5832				sin.sin_addr.s_addr = p4->addr;
5833				if (IN_MULTICAST(sin.sin_addr.s_addr)) {
5834					/* Skip multi-cast addresses */
5835					goto next_param;
5836				}
5837				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
5838				    (sin.sin_addr.s_addr == INADDR_ANY)) {
5839					goto next_param;
5840				}
5841				sa = (struct sockaddr *)&sin;
5842				inp = stcb->sctp_ep;
5843				atomic_add_int(&stcb->asoc.refcnt, 1);
5844				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
5845				    local_sa, stcb);
5846				atomic_add_int(&stcb->asoc.refcnt, -1);
5847
5848				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
5849				    inp == NULL) {
5850					/* we must add the source address */
5851					/*
5852					 * no scope set since we have a tcb
5853					 * already
5854					 */
5855
5856					/*
5857					 * we must validate the state again
5858					 * here
5859					 */
5860			add_it_now:
5861					if (stcb->asoc.state == 0) {
5862						/* the assoc was freed? */
5863						return (-7);
5864					}
5865					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
5866						return (-8);
5867					}
5868				} else if (stcb_tmp == stcb) {
5869					if (stcb->asoc.state == 0) {
5870						/* the assoc was freed? */
5871						return (-10);
5872					}
5873					if (net != NULL) {
5874						/* clear flag */
5875						net->dest_state &=
5876						    ~SCTP_ADDR_NOT_IN_ASSOC;
5877					}
5878				} else {
5879					/*
5880					 * strange, address is in another
5881					 * assoc? straighten out locks.
5882					 */
5883					if (stcb_tmp) {
5884						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
5885							/*
5886							 * in setup state we
5887							 * abort this guy
5888							 */
5889							sctp_abort_an_association(stcb_tmp->sctp_ep,
5890							    stcb_tmp, 1, NULL, 0);
5891							goto add_it_now;
5892						}
5893						SCTP_TCB_UNLOCK(stcb_tmp);
5894					}
5895					if (stcb->asoc.state == 0) {
5896						/* the assoc was freed? */
5897						return (-12);
5898					}
5899					return (-13);
5900				}
5901			}
5902		} else if (ptype == SCTP_IPV6_ADDRESS) {
5903			if (stcb->asoc.ipv6_addr_legal) {
5904				/* ok get the v6 address and check/add */
5905				struct sctp_ipv6addr_param *p6, p6_buf;
5906
5907				phdr = sctp_get_next_param(m, offset,
5908				    (struct sctp_paramhdr *)&p6_buf,
5909				    sizeof(p6_buf));
5910				if (plen != sizeof(struct sctp_ipv6addr_param) ||
5911				    phdr == NULL) {
5912					return (-14);
5913				}
5914				p6 = (struct sctp_ipv6addr_param *)phdr;
5915				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
5916				    sizeof(p6->addr));
5917				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
5918					/* Skip multi-cast addresses */
5919					goto next_param;
5920				}
5921				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
5922					/*
5923					 * Link local make no sense without
5924					 * scope
5925					 */
5926					goto next_param;
5927				}
5928				sa = (struct sockaddr *)&sin6;
5929				inp = stcb->sctp_ep;
5930				atomic_add_int(&stcb->asoc.refcnt, 1);
5931				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
5932				    local_sa, stcb);
5933				atomic_add_int(&stcb->asoc.refcnt, -1);
5934				if (stcb_tmp == NULL &&
5935				    (inp == stcb->sctp_ep || inp == NULL)) {
5936					/*
5937					 * we must validate the state again
5938					 * here
5939					 */
5940			add_it_now6:
5941					if (stcb->asoc.state == 0) {
5942						/* the assoc was freed? */
5943						return (-16);
5944					}
5945					/*
5946					 * we must add the address, no scope
5947					 * set
5948					 */
5949					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
5950						return (-17);
5951					}
5952				} else if (stcb_tmp == stcb) {
5953					/*
5954					 * we must validate the state again
5955					 * here
5956					 */
5957					if (stcb->asoc.state == 0) {
5958						/* the assoc was freed? */
5959						return (-19);
5960					}
5961					if (net != NULL) {
5962						/* clear flag */
5963						net->dest_state &=
5964						    ~SCTP_ADDR_NOT_IN_ASSOC;
5965					}
5966				} else {
5967					/*
5968					 * strange, address is in another
5969					 * assoc? straighten out locks.
5970					 */
5971					if (stcb_tmp)
5972						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
5973							/*
5974							 * in setup state we
5975							 * abort this guy
5976							 */
5977							sctp_abort_an_association(stcb_tmp->sctp_ep,
5978							    stcb_tmp, 1, NULL, 0);
5979							goto add_it_now6;
5980						}
5981					SCTP_TCB_UNLOCK(stcb_tmp);
5982
5983					if (stcb->asoc.state == 0) {
5984						/* the assoc was freed? */
5985						return (-21);
5986					}
5987					return (-22);
5988				}
5989			}
5990		} else if (ptype == SCTP_ECN_CAPABLE) {
5991			stcb->asoc.ecn_allowed = 1;
5992		} else if (ptype == SCTP_ULP_ADAPTATION) {
5993			if (stcb->asoc.state != SCTP_STATE_OPEN) {
5994				struct sctp_adaptation_layer_indication ai,
5995				                                *aip;
5996
5997				phdr = sctp_get_next_param(m, offset,
5998				    (struct sctp_paramhdr *)&ai, sizeof(ai));
5999				aip = (struct sctp_adaptation_layer_indication *)phdr;
6000				if (aip) {
6001					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6002					stcb->asoc.adaptation_needed = 1;
6003				}
6004			}
6005		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6006			struct sctp_asconf_addr_param lstore, *fee;
6007			struct sctp_asconf_addrv4_param *fii;
6008			int lptype;
6009			struct sockaddr *lsa = NULL;
6010
6011			stcb->asoc.peer_supports_asconf = 1;
6012			if (plen > sizeof(lstore)) {
6013				return (-23);
6014			}
6015			phdr = sctp_get_next_param(m, offset,
6016			    (struct sctp_paramhdr *)&lstore,
6017			    min(plen, sizeof(lstore)));
6018			if (phdr == NULL) {
6019				return (-24);
6020			}
6021			fee = (struct sctp_asconf_addr_param *)phdr;
6022			lptype = ntohs(fee->addrp.ph.param_type);
6023			if (lptype == SCTP_IPV4_ADDRESS) {
6024				if (plen !=
6025				    sizeof(struct sctp_asconf_addrv4_param)) {
6026					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6027					    (int)sizeof(struct sctp_asconf_addrv4_param),
6028					    plen);
6029				} else {
6030					fii = (struct sctp_asconf_addrv4_param *)fee;
6031					sin.sin_addr.s_addr = fii->addrp.addr;
6032					lsa = (struct sockaddr *)&sin;
6033				}
6034			} else if (lptype == SCTP_IPV6_ADDRESS) {
6035				if (plen !=
6036				    sizeof(struct sctp_asconf_addr_param)) {
6037					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6038					    (int)sizeof(struct sctp_asconf_addr_param),
6039					    plen);
6040				} else {
6041					memcpy(sin6.sin6_addr.s6_addr,
6042					    fee->addrp.addr,
6043					    sizeof(fee->addrp.addr));
6044					lsa = (struct sockaddr *)&sin6;
6045				}
6046			}
6047			if (lsa) {
6048				(void)sctp_set_primary_addr(stcb, sa, NULL);
6049			}
6050		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6051			stcb->asoc.peer_supports_nat = 1;
6052		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6053			/* Peer supports pr-sctp */
6054			stcb->asoc.peer_supports_prsctp = 1;
6055		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6056			/* A supported extension chunk */
6057			struct sctp_supported_chunk_types_param *pr_supported;
6058			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6059			int num_ent, i;
6060
6061			phdr = sctp_get_next_param(m, offset,
6062			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6063			if (phdr == NULL) {
6064				return (-25);
6065			}
6066			stcb->asoc.peer_supports_asconf = 0;
6067			stcb->asoc.peer_supports_prsctp = 0;
6068			stcb->asoc.peer_supports_pktdrop = 0;
6069			stcb->asoc.peer_supports_strreset = 0;
6070			stcb->asoc.peer_supports_nr_sack = 0;
6071			stcb->asoc.peer_supports_auth = 0;
6072			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6073			num_ent = plen - sizeof(struct sctp_paramhdr);
6074			for (i = 0; i < num_ent; i++) {
6075				switch (pr_supported->chunk_types[i]) {
6076				case SCTP_ASCONF:
6077				case SCTP_ASCONF_ACK:
6078					stcb->asoc.peer_supports_asconf = 1;
6079					break;
6080				case SCTP_FORWARD_CUM_TSN:
6081					stcb->asoc.peer_supports_prsctp = 1;
6082					break;
6083				case SCTP_PACKET_DROPPED:
6084					stcb->asoc.peer_supports_pktdrop = 1;
6085					break;
6086				case SCTP_NR_SELECTIVE_ACK:
6087					if (SCTP_BASE_SYSCTL(sctp_nr_sack_on_off))
6088						stcb->asoc.peer_supports_nr_sack = 1;
6089					else
6090						stcb->asoc.peer_supports_nr_sack = 0;
6091					break;
6092				case SCTP_STREAM_RESET:
6093					stcb->asoc.peer_supports_strreset = 1;
6094					break;
6095				case SCTP_AUTHENTICATION:
6096					stcb->asoc.peer_supports_auth = 1;
6097					break;
6098				default:
6099					/* one I have not learned yet */
6100					break;
6101
6102				}
6103			}
6104		} else if (ptype == SCTP_ECN_NONCE_SUPPORTED) {
6105			/* Peer supports ECN-nonce */
6106			stcb->asoc.peer_supports_ecn_nonce = 1;
6107			stcb->asoc.ecn_nonce_allowed = 1;
6108		} else if (ptype == SCTP_RANDOM) {
6109			if (plen > sizeof(random_store))
6110				break;
6111			if (got_random) {
6112				/* already processed a RANDOM */
6113				goto next_param;
6114			}
6115			phdr = sctp_get_next_param(m, offset,
6116			    (struct sctp_paramhdr *)random_store,
6117			    min(sizeof(random_store), plen));
6118			if (phdr == NULL)
6119				return (-26);
6120			p_random = (struct sctp_auth_random *)phdr;
6121			random_len = plen - sizeof(*p_random);
6122			/* enforce the random length */
6123			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6124				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6125				return (-27);
6126			}
6127			got_random = 1;
6128		} else if (ptype == SCTP_HMAC_LIST) {
6129			int num_hmacs;
6130			int i;
6131
6132			if (plen > sizeof(hmacs_store))
6133				break;
6134			if (got_hmacs) {
6135				/* already processed a HMAC list */
6136				goto next_param;
6137			}
6138			phdr = sctp_get_next_param(m, offset,
6139			    (struct sctp_paramhdr *)hmacs_store,
6140			    min(plen, sizeof(hmacs_store)));
6141			if (phdr == NULL)
6142				return (-28);
6143			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6144			hmacs_len = plen - sizeof(*hmacs);
6145			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6146			/* validate the hmac list */
6147			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6148				return (-29);
6149			}
6150			if (stcb->asoc.peer_hmacs != NULL)
6151				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6152			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6153			if (stcb->asoc.peer_hmacs != NULL) {
6154				for (i = 0; i < num_hmacs; i++) {
6155					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6156					    ntohs(hmacs->hmac_ids[i]));
6157				}
6158			}
6159			got_hmacs = 1;
6160		} else if (ptype == SCTP_CHUNK_LIST) {
6161			int i;
6162
6163			if (plen > sizeof(chunks_store))
6164				break;
6165			if (got_chklist) {
6166				/* already processed a Chunks list */
6167				goto next_param;
6168			}
6169			phdr = sctp_get_next_param(m, offset,
6170			    (struct sctp_paramhdr *)chunks_store,
6171			    min(plen, sizeof(chunks_store)));
6172			if (phdr == NULL)
6173				return (-30);
6174			chunks = (struct sctp_auth_chunk_list *)phdr;
6175			num_chunks = plen - sizeof(*chunks);
6176			if (stcb->asoc.peer_auth_chunks != NULL)
6177				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6178			else
6179				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6180			for (i = 0; i < num_chunks; i++) {
6181				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6182				    stcb->asoc.peer_auth_chunks);
6183				/* record asconf/asconf-ack if listed */
6184				if (chunks->chunk_types[i] == SCTP_ASCONF)
6185					saw_asconf = 1;
6186				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6187					saw_asconf_ack = 1;
6188
6189			}
6190			got_chklist = 1;
6191		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6192			    (ptype == SCTP_STATE_COOKIE) ||
6193			    (ptype == SCTP_UNRECOG_PARAM) ||
6194			    (ptype == SCTP_COOKIE_PRESERVE) ||
6195			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6196			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6197			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6198			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6199		    (ptype == SCTP_SUCCESS_REPORT)) {
6200			 /* don't care */ ;
6201		} else {
6202			if ((ptype & 0x8000) == 0x0000) {
6203				/*
6204				 * must stop processing the rest of the
6205				 * param's. Any report bits were handled
6206				 * with the call to
6207				 * sctp_arethere_unrecognized_parameters()
6208				 * when the INIT or INIT-ACK was first seen.
6209				 */
6210				break;
6211			}
6212		}
6213
6214next_param:
6215		offset += SCTP_SIZE32(plen);
6216		if (offset >= limit) {
6217			break;
6218		}
6219		phdr = sctp_get_next_param(m, offset, &parm_buf,
6220		    sizeof(parm_buf));
6221	}
6222	/* Now check to see if we need to purge any addresses */
6223	for (net = TAILQ_FIRST(&stcb->asoc.nets); net != NULL; net = net_tmp) {
6224		net_tmp = TAILQ_NEXT(net, sctp_next);
6225		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6226		    SCTP_ADDR_NOT_IN_ASSOC) {
6227			/* This address has been removed from the asoc */
6228			/* remove and free it */
6229			stcb->asoc.numnets--;
6230			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6231			sctp_free_remote_addr(net);
6232			if (net == stcb->asoc.primary_destination) {
6233				stcb->asoc.primary_destination = NULL;
6234				sctp_select_primary_destination(stcb);
6235			}
6236		}
6237	}
6238	/* validate authentication required parameters */
6239	if (got_random && got_hmacs) {
6240		stcb->asoc.peer_supports_auth = 1;
6241	} else {
6242		stcb->asoc.peer_supports_auth = 0;
6243	}
6244	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6245		/* peer does not support auth but sent a chunks list? */
6246		return (-31);
6247	}
6248	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6249	    !stcb->asoc.peer_supports_auth) {
6250		/* peer supports asconf but not auth? */
6251		return (-32);
6252	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6253	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6254		return (-33);
6255	}
6256	/* concatenate the full random key */
6257#ifdef SCTP_AUTH_DRAFT_04
6258	keylen = random_len;
6259	new_key = sctp_alloc_key(keylen);
6260	if (new_key != NULL) {
6261		/* copy in the RANDOM */
6262		if (p_random != NULL)
6263			bcopy(p_random->random_data, new_key->key, random_len);
6264	}
6265#else
6266	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6267	if (chunks != NULL) {
6268		keylen += sizeof(*chunks) + num_chunks;
6269	}
6270	new_key = sctp_alloc_key(keylen);
6271	if (new_key != NULL) {
6272		/* copy in the RANDOM */
6273		if (p_random != NULL) {
6274			keylen = sizeof(*p_random) + random_len;
6275			bcopy(p_random, new_key->key, keylen);
6276		}
6277		/* append in the AUTH chunks */
6278		if (chunks != NULL) {
6279			bcopy(chunks, new_key->key + keylen,
6280			    sizeof(*chunks) + num_chunks);
6281			keylen += sizeof(*chunks) + num_chunks;
6282		}
6283		/* append in the HMACs */
6284		if (hmacs != NULL) {
6285			bcopy(hmacs, new_key->key + keylen,
6286			    sizeof(*hmacs) + hmacs_len);
6287		}
6288	}
6289#endif
6290	else {
6291		/* failed to get memory for the key */
6292		return (-34);
6293	}
6294	if (stcb->asoc.authinfo.peer_random != NULL)
6295		sctp_free_key(stcb->asoc.authinfo.peer_random);
6296	stcb->asoc.authinfo.peer_random = new_key;
6297#ifdef SCTP_AUTH_DRAFT_04
6298	/* don't include the chunks and hmacs for draft -04 */
6299	stcb->asoc.authinfo.peer_random->keylen = random_len;
6300#endif
6301	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6302	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6303
6304	return (0);
6305}
6306
6307int
6308sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6309    struct sctp_nets *net)
6310{
6311	/* make sure the requested primary address exists in the assoc */
6312	if (net == NULL && sa)
6313		net = sctp_findnet(stcb, sa);
6314
6315	if (net == NULL) {
6316		/* didn't find the requested primary address! */
6317		return (-1);
6318	} else {
6319		/* set the primary address */
6320		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6321			/* Must be confirmed, so queue to set */
6322			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6323			return (0);
6324		}
6325		stcb->asoc.primary_destination = net;
6326		net->dest_state &= ~SCTP_ADDR_WAS_PRIMARY;
6327		net = TAILQ_FIRST(&stcb->asoc.nets);
6328		if (net != stcb->asoc.primary_destination) {
6329			/*
6330			 * first one on the list is NOT the primary
6331			 * sctp_cmpaddr() is much more efficent if the
6332			 * primary is the first on the list, make it so.
6333			 */
6334			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6335			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6336		}
6337		return (0);
6338	}
6339}
6340
6341int
6342sctp_is_vtag_good(struct sctp_inpcb *inp, uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now, int save_in_twait)
6343{
6344	/*
6345	 * This function serves two purposes. It will see if a TAG can be
6346	 * re-used and return 1 for yes it is ok and 0 for don't use that
6347	 * tag. A secondary function it will do is purge out old tags that
6348	 * can be removed.
6349	 */
6350	struct sctpvtaghead *chain;
6351	struct sctp_tagblock *twait_block;
6352	struct sctpasochead *head;
6353	struct sctp_tcb *stcb;
6354	int i;
6355
6356	SCTP_INP_INFO_RLOCK();
6357	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6358	    SCTP_BASE_INFO(hashasocmark))];
6359	if (head == NULL) {
6360		/* invalid vtag */
6361		goto skip_vtag_check;
6362	}
6363	LIST_FOREACH(stcb, head, sctp_asocs) {
6364		/*
6365		 * We choose not to lock anything here. TCB's can't be
6366		 * removed since we have the read lock, so they can't be
6367		 * freed on us, same thing for the INP. I may be wrong with
6368		 * this assumption, but we will go with it for now :-)
6369		 */
6370		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6371			continue;
6372		}
6373		if (stcb->asoc.my_vtag == tag) {
6374			/* candidate */
6375			if (stcb->rport != rport) {
6376				continue;
6377			}
6378			if (stcb->sctp_ep->sctp_lport != lport) {
6379				continue;
6380			}
6381			/* Its a used tag set */
6382			SCTP_INP_INFO_WUNLOCK();
6383			return (0);
6384		}
6385	}
6386skip_vtag_check:
6387
6388	chain = &SCTP_BASE_INFO(vtag_timewait[(tag % SCTP_STACK_VTAG_HASH_SIZE))];
6389	/* Now what about timed wait ? */
6390	if (!LIST_EMPTY(chain)) {
6391		/*
6392		 * Block(s) are present, lets see if we have this tag in the
6393		 * list
6394		 */
6395		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6396			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6397				if (twait_block->vtag_block[i].v_tag == 0) {
6398					/* not used */
6399					continue;
6400				} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6401				    now->tv_sec) {
6402					/* Audit expires this guy */
6403					twait_block->vtag_block[i].tv_sec_at_expire = 0;
6404					twait_block->vtag_block[i].v_tag = 0;
6405					twait_block->vtag_block[i].lport = 0;
6406					twait_block->vtag_block[i].rport = 0;
6407				} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6408					    (twait_block->vtag_block[i].lport == lport) &&
6409				    (twait_block->vtag_block[i].rport == rport)) {
6410					/* Bad tag, sorry :< */
6411					SCTP_INP_INFO_WUNLOCK();
6412					return (0);
6413				}
6414			}
6415		}
6416	}
6417	SCTP_INP_INFO_RUNLOCK();
6418	return (1);
6419}
6420
6421
6422static sctp_assoc_t reneged_asoc_ids[256];
6423static uint8_t reneged_at = 0;
6424
6425
6426static void
6427sctp_drain_mbufs(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
6428{
6429	/*
6430	 * We must hunt this association for MBUF's past the cumack (i.e.
6431	 * out of order data that we can renege on).
6432	 */
6433	struct sctp_association *asoc;
6434	struct sctp_tmit_chunk *chk, *nchk;
6435	uint32_t cumulative_tsn_p1, tsn;
6436	struct sctp_queued_to_read *ctl, *nctl;
6437	int cnt, strmat, gap;
6438
6439	/* We look for anything larger than the cum-ack + 1 */
6440
6441	asoc = &stcb->asoc;
6442	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6443		/* none we can reneg on. */
6444		return;
6445	}
6446	SCTP_STAT_INCR(sctps_protocol_drains_done);
6447	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6448	cnt = 0;
6449	/* First look in the re-assembly queue */
6450	chk = TAILQ_FIRST(&asoc->reasmqueue);
6451	while (chk) {
6452		/* Get the next one */
6453		nchk = TAILQ_NEXT(chk, sctp_next);
6454		if (compare_with_wrap(chk->rec.data.TSN_seq,
6455		    cumulative_tsn_p1, MAX_TSN)) {
6456			/* Yep it is above cum-ack */
6457			cnt++;
6458			tsn = chk->rec.data.TSN_seq;
6459			if (tsn >= asoc->mapping_array_base_tsn) {
6460				gap = tsn - asoc->mapping_array_base_tsn;
6461			} else {
6462				gap = (MAX_TSN - asoc->mapping_array_base_tsn) +
6463				    tsn + 1;
6464			}
6465			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6466			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6467			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6468			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6469			if (chk->data) {
6470				sctp_m_freem(chk->data);
6471				chk->data = NULL;
6472			}
6473			sctp_free_a_chunk(stcb, chk);
6474		}
6475		chk = nchk;
6476	}
6477	/* Ok that was fun, now we will drain all the inbound streams? */
6478	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6479		ctl = TAILQ_FIRST(&asoc->strmin[strmat].inqueue);
6480		while (ctl) {
6481			nctl = TAILQ_NEXT(ctl, next);
6482			if (compare_with_wrap(ctl->sinfo_tsn,
6483			    cumulative_tsn_p1, MAX_TSN)) {
6484				/* Yep it is above cum-ack */
6485				cnt++;
6486				tsn = ctl->sinfo_tsn;
6487				if (tsn >= asoc->mapping_array_base_tsn) {
6488					gap = tsn -
6489					    asoc->mapping_array_base_tsn;
6490				} else {
6491					gap = (MAX_TSN -
6492					    asoc->mapping_array_base_tsn) +
6493					    tsn + 1;
6494				}
6495				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6496				sctp_ucount_decr(asoc->cnt_on_all_streams);
6497
6498				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array,
6499				    gap);
6500				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue,
6501				    ctl, next);
6502				if (ctl->data) {
6503					sctp_m_freem(ctl->data);
6504					ctl->data = NULL;
6505				}
6506				sctp_free_remote_addr(ctl->whoFrom);
6507				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6508				SCTP_DECR_READQ_COUNT();
6509			}
6510			ctl = nctl;
6511		}
6512	}
6513	/*
6514	 * Question, should we go through the delivery queue? The only
6515	 * reason things are on here is the app not reading OR a p-d-api up.
6516	 * An attacker COULD send enough in to initiate the PD-API and then
6517	 * send a bunch of stuff to other streams... these would wind up on
6518	 * the delivery queue.. and then we would not get to them. But in
6519	 * order to do this I then have to back-track and un-deliver
6520	 * sequence numbers in streams.. el-yucko. I think for now we will
6521	 * NOT look at the delivery queue and leave it to be something to
6522	 * consider later. An alternative would be to abort the P-D-API with
6523	 * a notification and then deliver the data.... Or another method
6524	 * might be to keep track of how many times the situation occurs and
6525	 * if we see a possible attack underway just abort the association.
6526	 */
6527#ifdef SCTP_DEBUG
6528	if (cnt) {
6529		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6530	}
6531#endif
6532	if (cnt) {
6533		/*
6534		 * Now do we need to find a new
6535		 * asoc->highest_tsn_inside_map?
6536		 */
6537		if (asoc->highest_tsn_inside_map >= asoc->mapping_array_base_tsn) {
6538			gap = asoc->highest_tsn_inside_map - asoc->mapping_array_base_tsn;
6539		} else {
6540			gap = (MAX_TSN - asoc->mapping_array_base_tsn) +
6541			    asoc->highest_tsn_inside_map + 1;
6542		}
6543		if (gap >= (asoc->mapping_array_size << 3)) {
6544			/*
6545			 * Something bad happened or cum-ack and high were
6546			 * behind the base, but if so earlier checks should
6547			 * have found NO data... wierd... we will start at
6548			 * end of mapping array.
6549			 */
6550			SCTP_PRINTF("Gap was larger than array?? %d set to max:%d maparraymax:%x\n",
6551			    (int)gap,
6552			    (int)(asoc->mapping_array_size << 3),
6553			    (int)asoc->highest_tsn_inside_map);
6554			gap = asoc->mapping_array_size << 3;
6555		}
6556		while (gap > 0) {
6557			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6558				/* found the new highest */
6559				asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn + gap;
6560				if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) {
6561					sctp_log_map(0, 8, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT);
6562				}
6563				break;
6564			}
6565			gap--;
6566		}
6567		if (gap == 0) {
6568			/* Nothing left in map */
6569			memset(asoc->mapping_array, 0, asoc->mapping_array_size);
6570			asoc->mapping_array_base_tsn = asoc->cumulative_tsn + 1;
6571			asoc->highest_tsn_inside_map = asoc->cumulative_tsn;
6572			if (SCTP_BASE_SYSCTL(sctp_logging_level) & SCTP_MAP_LOGGING_ENABLE) {
6573				sctp_log_map(0, 9, asoc->highest_tsn_inside_map, SCTP_MAP_SLIDE_RESULT);
6574			}
6575		}
6576		asoc->last_revoke_count = cnt;
6577		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6578		/* sa_ignore NO_NULL_CHK */
6579		sctp_send_sack(stcb);
6580		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6581		reneged_asoc_ids[reneged_at] = sctp_get_associd(stcb);
6582		reneged_at++;
6583	}
6584	/*
6585	 * Another issue, in un-setting the TSN's in the mapping array we
6586	 * DID NOT adjust the higest_tsn marker.  This will cause one of two
6587	 * things to occur. It may cause us to do extra work in checking for
6588	 * our mapping array movement. More importantly it may cause us to
6589	 * SACK every datagram. This may not be a bad thing though since we
6590	 * will recover once we get our cum-ack above and all this stuff we
6591	 * dumped recovered.
6592	 */
6593}
6594
6595void
6596sctp_drain()
6597{
6598	/*
6599	 * We must walk the PCB lists for ALL associations here. The system
6600	 * is LOW on MBUF's and needs help. This is where reneging will
6601	 * occur. We really hope this does NOT happen!
6602	 */
6603	VNET_ITERATOR_DECL(vnet_iter);
6604	VNET_LIST_RLOCK_NOSLEEP();
6605	VNET_FOREACH(vnet_iter) {
6606		CURVNET_SET(vnet_iter);
6607		struct sctp_inpcb *inp;
6608		struct sctp_tcb *stcb;
6609
6610		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6611		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6612#ifdef VIMAGE
6613			continue;
6614#else
6615			return;
6616#endif
6617		}
6618		SCTP_INP_INFO_RLOCK();
6619		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6620			/* For each endpoint */
6621			SCTP_INP_RLOCK(inp);
6622			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6623				/* For each association */
6624				SCTP_TCB_LOCK(stcb);
6625				sctp_drain_mbufs(inp, stcb);
6626				SCTP_TCB_UNLOCK(stcb);
6627			}
6628			SCTP_INP_RUNLOCK(inp);
6629		}
6630		SCTP_INP_INFO_RUNLOCK();
6631		CURVNET_RESTORE();
6632	}
6633	VNET_LIST_RUNLOCK_NOSLEEP();
6634}
6635
6636/*
6637 * start a new iterator
6638 * iterates through all endpoints and associations based on the pcb_state
6639 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6640 * assocs and "ef" (optional) is executed when the iterator completes.
6641 * "inpf" (optional) is executed for each new endpoint as it is being
6642 * iterated through. inpe (optional) is called when the inp completes
6643 * its way through all the stcbs.
6644 */
6645int
6646sctp_initiate_iterator(inp_func inpf,
6647    asoc_func af,
6648    inp_func inpe,
6649    uint32_t pcb_state,
6650    uint32_t pcb_features,
6651    uint32_t asoc_state,
6652    void *argp,
6653    uint32_t argi,
6654    end_func ef,
6655    struct sctp_inpcb *s_inp,
6656    uint8_t chunk_output_off)
6657{
6658	struct sctp_iterator *it = NULL;
6659
6660	if (af == NULL) {
6661		return (-1);
6662	}
6663	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6664	    SCTP_M_ITER);
6665	if (it == NULL) {
6666		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6667		return (ENOMEM);
6668	}
6669	memset(it, 0, sizeof(*it));
6670	it->function_assoc = af;
6671	it->function_inp = inpf;
6672	if (inpf)
6673		it->done_current_ep = 0;
6674	else
6675		it->done_current_ep = 1;
6676	it->function_atend = ef;
6677	it->pointer = argp;
6678	it->val = argi;
6679	it->pcb_flags = pcb_state;
6680	it->pcb_features = pcb_features;
6681	it->asoc_state = asoc_state;
6682	it->function_inp_end = inpe;
6683	it->no_chunk_output = chunk_output_off;
6684	if (s_inp) {
6685		it->inp = s_inp;
6686		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
6687	} else {
6688		SCTP_INP_INFO_RLOCK();
6689		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
6690
6691		SCTP_INP_INFO_RUNLOCK();
6692		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
6693
6694	}
6695	SCTP_IPI_ITERATOR_WQ_LOCK();
6696	if (it->inp) {
6697		SCTP_INP_INCR_REF(it->inp);
6698	}
6699	TAILQ_INSERT_TAIL(&SCTP_BASE_INFO(iteratorhead), it, sctp_nxt_itr);
6700#if defined(SCTP_USE_THREAD_BASED_ITERATOR)
6701	if (SCTP_BASE_INFO(iterator_running) == 0) {
6702		sctp_wakeup_iterator();
6703	}
6704	SCTP_IPI_ITERATOR_WQ_UNLOCK();
6705#else
6706	if (it->inp)
6707		SCTP_INP_DECR_REF(it->inp);
6708	SCTP_IPI_ITERATOR_WQ_UNLOCK();
6709	/* Init the timer */
6710	SCTP_OS_TIMER_INIT(&it->tmr.timer);
6711	/* add to the list of all iterators */
6712	sctp_timer_start(SCTP_TIMER_TYPE_ITERATOR, (struct sctp_inpcb *)it,
6713	    NULL, NULL);
6714#endif
6715	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
6716	return (0);
6717}
6718