sctp_pcb.c revision 212225
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 212225 2010-09-05 13:41:45Z rrs $");
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");
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
696		SCTP_WQ_ADDR_LOCK();
697		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
698		SCTP_WQ_ADDR_UNLOCK();
699
700		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
701		    (struct sctp_inpcb *)NULL,
702		    (struct sctp_tcb *)NULL,
703		    (struct sctp_nets *)NULL);
704	} else {
705		/* it's ready for use */
706		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
707	}
708	return (sctp_ifap);
709}
710
711void
712sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
713    uint32_t ifn_index, const char *if_name)
714{
715	struct sctp_vrf *vrf;
716	struct sctp_ifa *sctp_ifap = NULL;
717
718	SCTP_IPI_ADDR_WLOCK();
719	vrf = sctp_find_vrf(vrf_id);
720	if (vrf == NULL) {
721		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
722		goto out_now;
723	}
724#ifdef SCTP_DEBUG
725	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
726	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
727#endif
728	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
729	if (sctp_ifap) {
730		/* Validate the delete */
731		if (sctp_ifap->ifn_p) {
732			int valid = 0;
733
734			/*-
735			 * The name has priority over the ifn_index
736			 * if its given. We do this especially for
737			 * panda who might recycle indexes fast.
738			 */
739			if (if_name) {
740				int len1, len2;
741
742				len1 = min(SCTP_IFNAMSIZ, strlen(if_name));
743				len2 = min(SCTP_IFNAMSIZ, strlen(sctp_ifap->ifn_p->ifn_name));
744				if (len1 && len2 && (len1 == len2)) {
745					/* we can compare them */
746					if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, len1) == 0) {
747						/*
748						 * They match its a correct
749						 * delete
750						 */
751						valid = 1;
752					}
753				}
754			}
755			if (!valid) {
756				/* last ditch check ifn_index */
757				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
758					valid = 1;
759				}
760			}
761			if (!valid) {
762				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
763				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
764				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
765				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
766				SCTP_IPI_ADDR_WUNLOCK();
767				return;
768			}
769		}
770		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", sctp_ifap);
771		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
772		sctp_ifap->localifa_flags |= SCTP_BEING_DELETED;
773		vrf->total_ifa_count--;
774		LIST_REMOVE(sctp_ifap, next_bucket);
775		sctp_remove_ifa_from_ifn(sctp_ifap);
776	}
777#ifdef SCTP_DEBUG
778	else {
779		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
780		    ifn_index);
781		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
782	}
783#endif
784
785out_now:
786	SCTP_IPI_ADDR_WUNLOCK();
787	if (sctp_ifap) {
788		struct sctp_laddr *wi;
789
790		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
791		if (wi == NULL) {
792			/*
793			 * Gak, what can we do? We have lost an address
794			 * change can you say HOSED?
795			 */
796			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
797
798			/* Oops, must decrement the count */
799			sctp_free_ifa(sctp_ifap);
800			return;
801		}
802		SCTP_INCR_LADDR_COUNT();
803		bzero(wi, sizeof(*wi));
804		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
805		wi->ifa = sctp_ifap;
806		wi->action = SCTP_DEL_IP_ADDRESS;
807		SCTP_WQ_ADDR_LOCK();
808		/*
809		 * Should this really be a tailq? As it is we will process
810		 * the newest first :-0
811		 */
812		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
813		SCTP_WQ_ADDR_UNLOCK();
814
815		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
816		    (struct sctp_inpcb *)NULL,
817		    (struct sctp_tcb *)NULL,
818		    (struct sctp_nets *)NULL);
819	}
820	return;
821}
822
823
824static struct sctp_tcb *
825sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
826    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
827{
828	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
829	/*
830	 * If we support the TCP model, then we must now dig through to see
831	 * if we can find our endpoint in the list of tcp ep's.
832	 */
833	uint16_t lport, rport;
834	struct sctppcbhead *ephead;
835	struct sctp_inpcb *inp;
836	struct sctp_laddr *laddr;
837	struct sctp_tcb *stcb;
838	struct sctp_nets *net;
839
840	if ((to == NULL) || (from == NULL)) {
841		return (NULL);
842	}
843	if (to->sa_family == AF_INET && from->sa_family == AF_INET) {
844		lport = ((struct sockaddr_in *)to)->sin_port;
845		rport = ((struct sockaddr_in *)from)->sin_port;
846	} else if (to->sa_family == AF_INET6 && from->sa_family == AF_INET6) {
847		lport = ((struct sockaddr_in6 *)to)->sin6_port;
848		rport = ((struct sockaddr_in6 *)from)->sin6_port;
849	} else {
850		return NULL;
851	}
852	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport), SCTP_BASE_INFO(hashtcpmark))];
853	/*
854	 * Ok now for each of the guys in this bucket we must look and see:
855	 * - Does the remote port match. - Does there single association's
856	 * addresses match this address (to). If so we update p_ep to point
857	 * to this ep and return the tcb from it.
858	 */
859	LIST_FOREACH(inp, ephead, sctp_hash) {
860		SCTP_INP_RLOCK(inp);
861		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
862			SCTP_INP_RUNLOCK(inp);
863			continue;
864		}
865		if (lport != inp->sctp_lport) {
866			SCTP_INP_RUNLOCK(inp);
867			continue;
868		}
869		if (inp->def_vrf_id != vrf_id) {
870			SCTP_INP_RUNLOCK(inp);
871			continue;
872		}
873		/* check to see if the ep has one of the addresses */
874		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
875			/* We are NOT bound all, so look further */
876			int match = 0;
877
878			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
879
880				if (laddr->ifa == NULL) {
881					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
882					continue;
883				}
884				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
885					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
886					continue;
887				}
888				if (laddr->ifa->address.sa.sa_family ==
889				    to->sa_family) {
890					/* see if it matches */
891					struct sockaddr_in *intf_addr, *sin;
892
893					intf_addr = &laddr->ifa->address.sin;
894					sin = (struct sockaddr_in *)to;
895					if (from->sa_family == AF_INET) {
896						if (sin->sin_addr.s_addr ==
897						    intf_addr->sin_addr.s_addr) {
898							match = 1;
899							break;
900						}
901					}
902#ifdef INET6
903					if (from->sa_family == AF_INET6) {
904						struct sockaddr_in6 *intf_addr6;
905						struct sockaddr_in6 *sin6;
906
907						sin6 = (struct sockaddr_in6 *)
908						    to;
909						intf_addr6 = &laddr->ifa->address.sin6;
910
911						if (SCTP6_ARE_ADDR_EQUAL(sin6,
912						    intf_addr6)) {
913							match = 1;
914							break;
915						}
916					}
917#endif
918				}
919			}
920			if (match == 0) {
921				/* This endpoint does not have this address */
922				SCTP_INP_RUNLOCK(inp);
923				continue;
924			}
925		}
926		/*
927		 * Ok if we hit here the ep has the address, does it hold
928		 * the tcb?
929		 */
930
931		stcb = LIST_FIRST(&inp->sctp_asoc_list);
932		if (stcb == NULL) {
933			SCTP_INP_RUNLOCK(inp);
934			continue;
935		}
936		SCTP_TCB_LOCK(stcb);
937		if (stcb->rport != rport) {
938			/* remote port does not match. */
939			SCTP_TCB_UNLOCK(stcb);
940			SCTP_INP_RUNLOCK(inp);
941			continue;
942		}
943		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
944			SCTP_TCB_UNLOCK(stcb);
945			SCTP_INP_RUNLOCK(inp);
946			continue;
947		}
948		/* Does this TCB have a matching address? */
949		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
950
951			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
952				/* not the same family, can't be a match */
953				continue;
954			}
955			switch (from->sa_family) {
956			case AF_INET:
957				{
958					struct sockaddr_in *sin, *rsin;
959
960					sin = (struct sockaddr_in *)&net->ro._l_addr;
961					rsin = (struct sockaddr_in *)from;
962					if (sin->sin_addr.s_addr ==
963					    rsin->sin_addr.s_addr) {
964						/* found it */
965						if (netp != NULL) {
966							*netp = net;
967						}
968						/*
969						 * Update the endpoint
970						 * pointer
971						 */
972						*inp_p = inp;
973						SCTP_INP_RUNLOCK(inp);
974						return (stcb);
975					}
976					break;
977				}
978#ifdef INET6
979			case AF_INET6:
980				{
981					struct sockaddr_in6 *sin6, *rsin6;
982
983					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
984					rsin6 = (struct sockaddr_in6 *)from;
985					if (SCTP6_ARE_ADDR_EQUAL(sin6,
986					    rsin6)) {
987						/* found it */
988						if (netp != NULL) {
989							*netp = net;
990						}
991						/*
992						 * Update the endpoint
993						 * pointer
994						 */
995						*inp_p = inp;
996						SCTP_INP_RUNLOCK(inp);
997						return (stcb);
998					}
999					break;
1000				}
1001#endif
1002			default:
1003				/* TSNH */
1004				break;
1005			}
1006		}
1007		SCTP_TCB_UNLOCK(stcb);
1008		SCTP_INP_RUNLOCK(inp);
1009	}
1010	return (NULL);
1011}
1012
1013static int
1014sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
1015{
1016	int loopback_scope, ipv4_local_scope, local_scope, site_scope;
1017	int ipv4_addr_legal, ipv6_addr_legal;
1018	struct sctp_vrf *vrf;
1019	struct sctp_ifn *sctp_ifn;
1020	struct sctp_ifa *sctp_ifa;
1021
1022	loopback_scope = stcb->asoc.loopback_scope;
1023	ipv4_local_scope = stcb->asoc.ipv4_local_scope;
1024	local_scope = stcb->asoc.local_scope;
1025	site_scope = stcb->asoc.site_scope;
1026	ipv4_addr_legal = ipv6_addr_legal = 0;
1027	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1028		ipv6_addr_legal = 1;
1029		if (SCTP_IPV6_V6ONLY(stcb->sctp_ep) == 0) {
1030			ipv4_addr_legal = 1;
1031		}
1032	} else {
1033		ipv4_addr_legal = 1;
1034	}
1035
1036	SCTP_IPI_ADDR_RLOCK();
1037	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
1038	if (vrf == NULL) {
1039		/* no vrf, no addresses */
1040		SCTP_IPI_ADDR_RUNLOCK();
1041		return (0);
1042	}
1043	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
1044		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
1045			if ((loopback_scope == 0) &&
1046			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
1047				continue;
1048			}
1049			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
1050				if (sctp_is_addr_restricted(stcb, sctp_ifa))
1051					continue;
1052				switch (sctp_ifa->address.sa.sa_family) {
1053#ifdef INET
1054				case AF_INET:
1055					if (ipv4_addr_legal) {
1056						struct sockaddr_in *sin,
1057						           *rsin;
1058
1059						sin = &sctp_ifa->address.sin;
1060						rsin = (struct sockaddr_in *)to;
1061						if ((ipv4_local_scope == 0) &&
1062						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
1063							continue;
1064						}
1065						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
1066							SCTP_IPI_ADDR_RUNLOCK();
1067							return (1);
1068						}
1069					}
1070					break;
1071#endif
1072#ifdef INET6
1073				case AF_INET6:
1074					if (ipv6_addr_legal) {
1075						struct sockaddr_in6 *sin6,
1076						            *rsin6;
1077
1078						sin6 = &sctp_ifa->address.sin6;
1079						rsin6 = (struct sockaddr_in6 *)to;
1080						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
1081							if (local_scope == 0)
1082								continue;
1083							if (sin6->sin6_scope_id == 0) {
1084								if (sa6_recoverscope(sin6) != 0)
1085									continue;
1086							}
1087						}
1088						if ((site_scope == 0) &&
1089						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
1090							continue;
1091						}
1092						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
1093							SCTP_IPI_ADDR_RUNLOCK();
1094							return (1);
1095						}
1096					}
1097					break;
1098#endif
1099				default:
1100					/* TSNH */
1101					break;
1102				}
1103			}
1104		}
1105	} else {
1106		struct sctp_laddr *laddr;
1107
1108		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
1109			if (sctp_is_addr_restricted(stcb, laddr->ifa)) {
1110				continue;
1111			}
1112			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
1113				continue;
1114			}
1115			switch (to->sa_family) {
1116#ifdef INET
1117			case AF_INET:
1118				{
1119					struct sockaddr_in *sin, *rsin;
1120
1121					sin = (struct sockaddr_in *)&laddr->ifa->address.sin;
1122					rsin = (struct sockaddr_in *)to;
1123					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
1124						SCTP_IPI_ADDR_RUNLOCK();
1125						return (1);
1126					}
1127					break;
1128				}
1129#endif
1130#ifdef INET6
1131			case AF_INET6:
1132				{
1133					struct sockaddr_in6 *sin6, *rsin6;
1134
1135					sin6 = (struct sockaddr_in6 *)&laddr->ifa->address.sin6;
1136					rsin6 = (struct sockaddr_in6 *)to;
1137					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
1138						SCTP_IPI_ADDR_RUNLOCK();
1139						return (1);
1140					}
1141					break;
1142				}
1143
1144#endif
1145			default:
1146				/* TSNH */
1147				break;
1148			}
1149
1150		}
1151	}
1152	SCTP_IPI_ADDR_RUNLOCK();
1153	return (0);
1154}
1155
1156/*
1157 * rules for use
1158 *
1159 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1160 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1161 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1162 * NULL.
1163 */
1164
1165struct sctp_tcb *
1166sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1167    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1168{
1169	struct sctpasochead *head;
1170	struct sctp_inpcb *inp;
1171	struct sctp_tcb *stcb = NULL;
1172	struct sctp_nets *net;
1173	uint16_t rport;
1174
1175	inp = *inp_p;
1176	if (remote->sa_family == AF_INET) {
1177		rport = (((struct sockaddr_in *)remote)->sin_port);
1178	} else if (remote->sa_family == AF_INET6) {
1179		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1180	} else {
1181		return (NULL);
1182	}
1183	if (locked_tcb) {
1184		/*
1185		 * UN-lock so we can do proper locking here this occurs when
1186		 * called from load_addresses_from_init.
1187		 */
1188		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1189		SCTP_TCB_UNLOCK(locked_tcb);
1190	}
1191	SCTP_INP_INFO_RLOCK();
1192	if (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) {
1193		/*-
1194		 * Now either this guy is our listener or it's the
1195		 * connector. If it is the one that issued the connect, then
1196		 * it's only chance is to be the first TCB in the list. If
1197		 * it is the acceptor, then do the special_lookup to hash
1198		 * and find the real inp.
1199		 */
1200		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1201			/* to is peer addr, from is my addr */
1202			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1203			    netp, inp->def_vrf_id);
1204			if ((stcb != NULL) && (locked_tcb == NULL)) {
1205				/* we have a locked tcb, lower refcount */
1206				SCTP_INP_DECR_REF(inp);
1207			}
1208			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1209				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1210				SCTP_TCB_LOCK(locked_tcb);
1211				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1212				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1213			}
1214			SCTP_INP_INFO_RUNLOCK();
1215			return (stcb);
1216		} else {
1217			SCTP_INP_WLOCK(inp);
1218			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1219				goto null_return;
1220			}
1221			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1222			if (stcb == NULL) {
1223				goto null_return;
1224			}
1225			SCTP_TCB_LOCK(stcb);
1226
1227			if (stcb->rport != rport) {
1228				/* remote port does not match. */
1229				SCTP_TCB_UNLOCK(stcb);
1230				goto null_return;
1231			}
1232			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1233				SCTP_TCB_UNLOCK(stcb);
1234				goto null_return;
1235			}
1236			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1237				SCTP_TCB_UNLOCK(stcb);
1238				goto null_return;
1239			}
1240			/* now look at the list of remote addresses */
1241			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1242#ifdef INVARIANTS
1243				if (net == (TAILQ_NEXT(net, sctp_next))) {
1244					panic("Corrupt net list");
1245				}
1246#endif
1247				if (net->ro._l_addr.sa.sa_family !=
1248				    remote->sa_family) {
1249					/* not the same family */
1250					continue;
1251				}
1252				switch (remote->sa_family) {
1253				case AF_INET:
1254					{
1255						struct sockaddr_in *sin,
1256						           *rsin;
1257
1258						sin = (struct sockaddr_in *)
1259						    &net->ro._l_addr;
1260						rsin = (struct sockaddr_in *)remote;
1261						if (sin->sin_addr.s_addr ==
1262						    rsin->sin_addr.s_addr) {
1263							/* found it */
1264							if (netp != NULL) {
1265								*netp = net;
1266							}
1267							if (locked_tcb == NULL) {
1268								SCTP_INP_DECR_REF(inp);
1269							} else if (locked_tcb != stcb) {
1270								SCTP_TCB_LOCK(locked_tcb);
1271							}
1272							if (locked_tcb) {
1273								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1274							}
1275							SCTP_INP_WUNLOCK(inp);
1276							SCTP_INP_INFO_RUNLOCK();
1277							return (stcb);
1278						}
1279						break;
1280					}
1281#ifdef INET6
1282				case AF_INET6:
1283					{
1284						struct sockaddr_in6 *sin6,
1285						            *rsin6;
1286
1287						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1288						rsin6 = (struct sockaddr_in6 *)remote;
1289						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1290						    rsin6)) {
1291							/* found it */
1292							if (netp != NULL) {
1293								*netp = net;
1294							}
1295							if (locked_tcb == NULL) {
1296								SCTP_INP_DECR_REF(inp);
1297							} else if (locked_tcb != stcb) {
1298								SCTP_TCB_LOCK(locked_tcb);
1299							}
1300							if (locked_tcb) {
1301								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1302							}
1303							SCTP_INP_WUNLOCK(inp);
1304							SCTP_INP_INFO_RUNLOCK();
1305							return (stcb);
1306						}
1307						break;
1308					}
1309#endif
1310				default:
1311					/* TSNH */
1312					break;
1313				}
1314			}
1315			SCTP_TCB_UNLOCK(stcb);
1316		}
1317	} else {
1318		SCTP_INP_WLOCK(inp);
1319		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1320			goto null_return;
1321		}
1322		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1323		    inp->sctp_hashmark)];
1324		if (head == NULL) {
1325			goto null_return;
1326		}
1327		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1328			if (stcb->rport != rport) {
1329				/* remote port does not match */
1330				continue;
1331			}
1332			SCTP_TCB_LOCK(stcb);
1333			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1334				SCTP_TCB_UNLOCK(stcb);
1335				continue;
1336			}
1337			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1338				SCTP_TCB_UNLOCK(stcb);
1339				continue;
1340			}
1341			/* now look at the list of remote addresses */
1342			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1343#ifdef INVARIANTS
1344				if (net == (TAILQ_NEXT(net, sctp_next))) {
1345					panic("Corrupt net list");
1346				}
1347#endif
1348				if (net->ro._l_addr.sa.sa_family !=
1349				    remote->sa_family) {
1350					/* not the same family */
1351					continue;
1352				}
1353				switch (remote->sa_family) {
1354				case AF_INET:
1355					{
1356						struct sockaddr_in *sin,
1357						           *rsin;
1358
1359						sin = (struct sockaddr_in *)
1360						    &net->ro._l_addr;
1361						rsin = (struct sockaddr_in *)remote;
1362						if (sin->sin_addr.s_addr ==
1363						    rsin->sin_addr.s_addr) {
1364							/* found it */
1365							if (netp != NULL) {
1366								*netp = net;
1367							}
1368							if (locked_tcb == NULL) {
1369								SCTP_INP_DECR_REF(inp);
1370							} else if (locked_tcb != stcb) {
1371								SCTP_TCB_LOCK(locked_tcb);
1372							}
1373							if (locked_tcb) {
1374								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1375							}
1376							SCTP_INP_WUNLOCK(inp);
1377							SCTP_INP_INFO_RUNLOCK();
1378							return (stcb);
1379						}
1380						break;
1381					}
1382#ifdef INET6
1383				case AF_INET6:
1384					{
1385						struct sockaddr_in6 *sin6,
1386						            *rsin6;
1387
1388						sin6 = (struct sockaddr_in6 *)
1389						    &net->ro._l_addr;
1390						rsin6 = (struct sockaddr_in6 *)remote;
1391						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1392						    rsin6)) {
1393							/* found it */
1394							if (netp != NULL) {
1395								*netp = net;
1396							}
1397							if (locked_tcb == NULL) {
1398								SCTP_INP_DECR_REF(inp);
1399							} else if (locked_tcb != stcb) {
1400								SCTP_TCB_LOCK(locked_tcb);
1401							}
1402							if (locked_tcb) {
1403								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1404							}
1405							SCTP_INP_WUNLOCK(inp);
1406							SCTP_INP_INFO_RUNLOCK();
1407							return (stcb);
1408						}
1409						break;
1410					}
1411#endif
1412				default:
1413					/* TSNH */
1414					break;
1415				}
1416			}
1417			SCTP_TCB_UNLOCK(stcb);
1418		}
1419	}
1420null_return:
1421	/* clean up for returning null */
1422	if (locked_tcb) {
1423		SCTP_TCB_LOCK(locked_tcb);
1424		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1425	}
1426	SCTP_INP_WUNLOCK(inp);
1427	SCTP_INP_INFO_RUNLOCK();
1428	/* not found */
1429	return (NULL);
1430}
1431
1432/*
1433 * Find an association for a specific endpoint using the association id given
1434 * out in the COMM_UP notification
1435 */
1436
1437struct sctp_tcb *
1438sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1439{
1440	/*
1441	 * Use my the assoc_id to find a endpoint
1442	 */
1443	struct sctpasochead *head;
1444	struct sctp_tcb *stcb;
1445	uint32_t id;
1446
1447	if (inp == NULL) {
1448		SCTP_PRINTF("TSNH ep_associd\n");
1449		return (NULL);
1450	}
1451	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1452		SCTP_PRINTF("TSNH ep_associd0\n");
1453		return (NULL);
1454	}
1455	id = (uint32_t) asoc_id;
1456	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1457	if (head == NULL) {
1458		/* invalid id TSNH */
1459		SCTP_PRINTF("TSNH ep_associd1\n");
1460		return (NULL);
1461	}
1462	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1463		if (stcb->asoc.assoc_id == id) {
1464			if (inp != stcb->sctp_ep) {
1465				/*
1466				 * some other guy has the same id active (id
1467				 * collision ??).
1468				 */
1469				SCTP_PRINTF("TSNH ep_associd2\n");
1470				continue;
1471			}
1472			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1473				continue;
1474			}
1475			if (want_lock) {
1476				SCTP_TCB_LOCK(stcb);
1477			}
1478			return (stcb);
1479		}
1480	}
1481	return (NULL);
1482}
1483
1484
1485struct sctp_tcb *
1486sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1487{
1488	struct sctp_tcb *stcb;
1489
1490	SCTP_INP_RLOCK(inp);
1491	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1492	SCTP_INP_RUNLOCK(inp);
1493	return (stcb);
1494}
1495
1496
1497static struct sctp_inpcb *
1498sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1499    uint16_t lport, uint32_t vrf_id)
1500{
1501	struct sctp_inpcb *inp;
1502	struct sockaddr_in *sin;
1503
1504#ifdef INET6
1505	struct sockaddr_in6 *sin6;
1506
1507#endif
1508	struct sctp_laddr *laddr;
1509
1510#ifdef INET6
1511	struct sockaddr_in6 *intf_addr6;
1512
1513#endif
1514
1515	int fnd;
1516
1517	/*
1518	 * Endpoing probe expects that the INP_INFO is locked.
1519	 */
1520	sin = NULL;
1521#ifdef INET6
1522	sin6 = NULL;
1523#endif
1524	switch (nam->sa_family) {
1525	case AF_INET:
1526		sin = (struct sockaddr_in *)nam;
1527		break;
1528#ifdef INET6
1529	case AF_INET6:
1530		sin6 = (struct sockaddr_in6 *)nam;
1531		break;
1532#endif
1533	default:
1534		/* unsupported family */
1535		return (NULL);
1536	}
1537
1538	if (head == NULL)
1539		return (NULL);
1540
1541	LIST_FOREACH(inp, head, sctp_hash) {
1542		SCTP_INP_RLOCK(inp);
1543		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1544			SCTP_INP_RUNLOCK(inp);
1545			continue;
1546		}
1547		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1548		    (inp->sctp_lport == lport)) {
1549			/* got it */
1550			if ((nam->sa_family == AF_INET) &&
1551			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1552			    SCTP_IPV6_V6ONLY(inp)) {
1553				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1554				SCTP_INP_RUNLOCK(inp);
1555				continue;
1556			}
1557			/* A V6 address and the endpoint is NOT bound V6 */
1558			if (nam->sa_family == AF_INET6 &&
1559			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1560				SCTP_INP_RUNLOCK(inp);
1561				continue;
1562			}
1563			/* does a VRF id match? */
1564			fnd = 0;
1565			if (inp->def_vrf_id == vrf_id)
1566				fnd = 1;
1567
1568			SCTP_INP_RUNLOCK(inp);
1569			if (!fnd)
1570				continue;
1571			return (inp);
1572		}
1573		SCTP_INP_RUNLOCK(inp);
1574	}
1575	if ((nam->sa_family == AF_INET) &&
1576	    (sin->sin_addr.s_addr == INADDR_ANY)) {
1577		/* Can't hunt for one that has no address specified */
1578		return (NULL);
1579	}
1580#ifdef INET6
1581	if ((nam->sa_family == AF_INET6) &&
1582	    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
1583		/* Can't hunt for one that has no address specified */
1584		return (NULL);
1585	}
1586#endif
1587	/*
1588	 * ok, not bound to all so see if we can find a EP bound to this
1589	 * address.
1590	 */
1591	LIST_FOREACH(inp, head, sctp_hash) {
1592		SCTP_INP_RLOCK(inp);
1593		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1594			SCTP_INP_RUNLOCK(inp);
1595			continue;
1596		}
1597		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1598			SCTP_INP_RUNLOCK(inp);
1599			continue;
1600		}
1601		/*
1602		 * Ok this could be a likely candidate, look at all of its
1603		 * addresses
1604		 */
1605		if (inp->sctp_lport != lport) {
1606			SCTP_INP_RUNLOCK(inp);
1607			continue;
1608		}
1609		/* does a VRF id match? */
1610		fnd = 0;
1611		if (inp->def_vrf_id == vrf_id)
1612			fnd = 1;
1613
1614		if (!fnd) {
1615			SCTP_INP_RUNLOCK(inp);
1616			continue;
1617		}
1618		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1619			if (laddr->ifa == NULL) {
1620				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1621				    __FUNCTION__);
1622				continue;
1623			}
1624			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1625			    laddr->ifa);
1626			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1627				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1628				continue;
1629			}
1630			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1631				/* possible, see if it matches */
1632				struct sockaddr_in *intf_addr;
1633
1634				intf_addr = &laddr->ifa->address.sin;
1635				switch (nam->sa_family) {
1636				case AF_INET:
1637					if (sin->sin_addr.s_addr ==
1638					    intf_addr->sin_addr.s_addr) {
1639						SCTP_INP_RUNLOCK(inp);
1640						return (inp);
1641					}
1642					break;
1643#ifdef INET6
1644				case AF_INET6:
1645					intf_addr6 = &laddr->ifa->address.sin6;
1646					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1647					    intf_addr6)) {
1648						SCTP_INP_RUNLOCK(inp);
1649						return (inp);
1650					}
1651					break;
1652#endif
1653				}
1654			}
1655		}
1656		SCTP_INP_RUNLOCK(inp);
1657	}
1658	return (NULL);
1659}
1660
1661
1662static struct sctp_inpcb *
1663sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1664{
1665	struct sctppcbhead *head;
1666	struct sctp_inpcb *t_inp;
1667	int fnd;
1668
1669	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1670	    SCTP_BASE_INFO(hashmark))];
1671	LIST_FOREACH(t_inp, head, sctp_hash) {
1672		if (t_inp->sctp_lport != lport) {
1673			continue;
1674		}
1675		/* is it in the VRF in question */
1676		fnd = 0;
1677		if (t_inp->def_vrf_id == vrf_id)
1678			fnd = 1;
1679		if (!fnd)
1680			continue;
1681
1682		/* This one is in use. */
1683		/* check the v6/v4 binding issue */
1684		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1685		    SCTP_IPV6_V6ONLY(t_inp)) {
1686			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1687				/* collision in V6 space */
1688				return (t_inp);
1689			} else {
1690				/* inp is BOUND_V4 no conflict */
1691				continue;
1692			}
1693		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1694			/* t_inp is bound v4 and v6, conflict always */
1695			return (t_inp);
1696		} else {
1697			/* t_inp is bound only V4 */
1698			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1699			    SCTP_IPV6_V6ONLY(inp)) {
1700				/* no conflict */
1701				continue;
1702			}
1703			/* else fall through to conflict */
1704		}
1705		return (t_inp);
1706	}
1707	return (NULL);
1708}
1709
1710
1711int
1712sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1713{
1714	/* For 1-2-1 with port reuse */
1715	struct sctppcbhead *head;
1716	struct sctp_inpcb *tinp;
1717
1718	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1719		/* only works with port reuse on */
1720		return (-1);
1721	}
1722	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1723		return (0);
1724	}
1725	SCTP_INP_RUNLOCK(inp);
1726	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1727	    SCTP_BASE_INFO(hashmark))];
1728	/* Kick out all non-listeners to the TCP hash */
1729	LIST_FOREACH(tinp, head, sctp_hash) {
1730		if (tinp->sctp_lport != inp->sctp_lport) {
1731			continue;
1732		}
1733		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1734			continue;
1735		}
1736		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1737			continue;
1738		}
1739		if (tinp->sctp_socket->so_qlimit) {
1740			continue;
1741		}
1742		SCTP_INP_WLOCK(tinp);
1743		LIST_REMOVE(tinp, sctp_hash);
1744		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1745		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1746		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1747		SCTP_INP_WUNLOCK(tinp);
1748	}
1749	SCTP_INP_WLOCK(inp);
1750	/* Pull from where he was */
1751	LIST_REMOVE(inp, sctp_hash);
1752	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1753	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1754	LIST_INSERT_HEAD(head, inp, sctp_hash);
1755	SCTP_INP_WUNLOCK(inp);
1756	SCTP_INP_RLOCK(inp);
1757	return (0);
1758}
1759
1760
1761struct sctp_inpcb *
1762sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1763    uint32_t vrf_id)
1764{
1765	/*
1766	 * First we check the hash table to see if someone has this port
1767	 * bound with just the port.
1768	 */
1769	struct sctp_inpcb *inp;
1770	struct sctppcbhead *head;
1771	struct sockaddr_in *sin;
1772	struct sockaddr_in6 *sin6;
1773	int lport;
1774
1775	if (nam->sa_family == AF_INET) {
1776		sin = (struct sockaddr_in *)nam;
1777		lport = ((struct sockaddr_in *)nam)->sin_port;
1778	} else if (nam->sa_family == AF_INET6) {
1779		sin6 = (struct sockaddr_in6 *)nam;
1780		lport = ((struct sockaddr_in6 *)nam)->sin6_port;
1781	} else {
1782		/* unsupported family */
1783		return (NULL);
1784	}
1785	/*
1786	 * I could cheat here and just cast to one of the types but we will
1787	 * do it right. It also provides the check against an Unsupported
1788	 * type too.
1789	 */
1790	/* Find the head of the ALLADDR chain */
1791	if (have_lock == 0) {
1792		SCTP_INP_INFO_RLOCK();
1793	}
1794	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1795	    SCTP_BASE_INFO(hashmark))];
1796	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1797
1798	/*
1799	 * If the TCP model exists it could be that the main listening
1800	 * endpoint is gone but there exists a connected socket for this guy
1801	 * yet. If so we can return the first one that we find. This may NOT
1802	 * be the correct one so the caller should be wary on the return
1803	 * INP. Currently the onlyc caller that sets this flag is in bindx
1804	 * where we are verifying that a user CAN bind the address. He
1805	 * either has bound it already, or someone else has, or its open to
1806	 * bind, so this is good enough.
1807	 */
1808	if (inp == NULL && find_tcp_pool) {
1809		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
1810		inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1811	}
1812	if (inp) {
1813		SCTP_INP_INCR_REF(inp);
1814	}
1815	if (have_lock == 0) {
1816		SCTP_INP_INFO_RUNLOCK();
1817	}
1818	return (inp);
1819}
1820
1821/*
1822 * Find an association for an endpoint with the pointer to whom you want to
1823 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1824 * need to change the *to to some other struct like a mbuf...
1825 */
1826struct sctp_tcb *
1827sctp_findassociation_addr_sa(struct sockaddr *to, struct sockaddr *from,
1828    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1829    uint32_t vrf_id)
1830{
1831	struct sctp_inpcb *inp = NULL;
1832	struct sctp_tcb *retval;
1833
1834	SCTP_INP_INFO_RLOCK();
1835	if (find_tcp_pool) {
1836		if (inp_p != NULL) {
1837			retval = sctp_tcb_special_locate(inp_p, from, to, netp,
1838			    vrf_id);
1839		} else {
1840			retval = sctp_tcb_special_locate(&inp, from, to, netp,
1841			    vrf_id);
1842		}
1843		if (retval != NULL) {
1844			SCTP_INP_INFO_RUNLOCK();
1845			return (retval);
1846		}
1847	}
1848	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1849	if (inp_p != NULL) {
1850		*inp_p = inp;
1851	}
1852	SCTP_INP_INFO_RUNLOCK();
1853
1854	if (inp == NULL) {
1855		return (NULL);
1856	}
1857	/*
1858	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1859	 * we now place the source address or from in the to of the find
1860	 * endpoint call. Since in reality this chain is used from the
1861	 * inbound packet side.
1862	 */
1863	if (inp_p != NULL) {
1864		retval = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1865		    NULL);
1866	} else {
1867		retval = sctp_findassociation_ep_addr(&inp, from, netp, to,
1868		    NULL);
1869	}
1870	return retval;
1871}
1872
1873
1874/*
1875 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1876 * find all addresses that the sender has specified in any address list. Each
1877 * address will be used to lookup the TCB and see if one exits.
1878 */
1879static struct sctp_tcb *
1880sctp_findassociation_special_addr(struct mbuf *m, int iphlen, int offset,
1881    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1882    struct sockaddr *dest)
1883{
1884	struct sockaddr_in sin4;
1885	struct sockaddr_in6 sin6;
1886	struct sctp_paramhdr *phdr, parm_buf;
1887	struct sctp_tcb *retval;
1888	uint32_t ptype, plen;
1889
1890	memset(&sin4, 0, sizeof(sin4));
1891	memset(&sin6, 0, sizeof(sin6));
1892	sin4.sin_len = sizeof(sin4);
1893	sin4.sin_family = AF_INET;
1894	sin4.sin_port = sh->src_port;
1895	sin6.sin6_len = sizeof(sin6);
1896	sin6.sin6_family = AF_INET6;
1897	sin6.sin6_port = sh->src_port;
1898
1899	retval = NULL;
1900	offset += sizeof(struct sctp_init_chunk);
1901
1902	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
1903	while (phdr != NULL) {
1904		/* now we must see if we want the parameter */
1905		ptype = ntohs(phdr->param_type);
1906		plen = ntohs(phdr->param_length);
1907		if (plen == 0) {
1908			break;
1909		}
1910		if (ptype == SCTP_IPV4_ADDRESS &&
1911		    plen == sizeof(struct sctp_ipv4addr_param)) {
1912			/* Get the rest of the address */
1913			struct sctp_ipv4addr_param ip4_parm, *p4;
1914
1915			phdr = sctp_get_next_param(m, offset,
1916			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
1917			if (phdr == NULL) {
1918				return (NULL);
1919			}
1920			p4 = (struct sctp_ipv4addr_param *)phdr;
1921			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
1922			/* look it up */
1923			retval = sctp_findassociation_ep_addr(inp_p,
1924			    (struct sockaddr *)&sin4, netp, dest, NULL);
1925			if (retval != NULL) {
1926				return (retval);
1927			}
1928		} else if (ptype == SCTP_IPV6_ADDRESS &&
1929		    plen == sizeof(struct sctp_ipv6addr_param)) {
1930			/* Get the rest of the address */
1931			struct sctp_ipv6addr_param ip6_parm, *p6;
1932
1933			phdr = sctp_get_next_param(m, offset,
1934			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
1935			if (phdr == NULL) {
1936				return (NULL);
1937			}
1938			p6 = (struct sctp_ipv6addr_param *)phdr;
1939			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
1940			/* look it up */
1941			retval = sctp_findassociation_ep_addr(inp_p,
1942			    (struct sockaddr *)&sin6, netp, dest, NULL);
1943			if (retval != NULL) {
1944				return (retval);
1945			}
1946		}
1947		offset += SCTP_SIZE32(plen);
1948		phdr = sctp_get_next_param(m, offset, &parm_buf,
1949		    sizeof(parm_buf));
1950	}
1951	return (NULL);
1952}
1953
1954static struct sctp_tcb *
1955sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
1956    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
1957    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
1958{
1959	/*
1960	 * Use my vtag to hash. If we find it we then verify the source addr
1961	 * is in the assoc. If all goes well we save a bit on rec of a
1962	 * packet.
1963	 */
1964	struct sctpasochead *head;
1965	struct sctp_nets *net;
1966	struct sctp_tcb *stcb;
1967
1968	*netp = NULL;
1969	*inp_p = NULL;
1970	SCTP_INP_INFO_RLOCK();
1971	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
1972	    SCTP_BASE_INFO(hashasocmark))];
1973	if (head == NULL) {
1974		/* invalid vtag */
1975		SCTP_INP_INFO_RUNLOCK();
1976		return (NULL);
1977	}
1978	LIST_FOREACH(stcb, head, sctp_asocs) {
1979		SCTP_INP_RLOCK(stcb->sctp_ep);
1980		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1981			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1982			continue;
1983		}
1984		SCTP_TCB_LOCK(stcb);
1985		SCTP_INP_RUNLOCK(stcb->sctp_ep);
1986		if (stcb->asoc.my_vtag == vtag) {
1987			/* candidate */
1988			if (stcb->rport != rport) {
1989				SCTP_TCB_UNLOCK(stcb);
1990				continue;
1991			}
1992			if (stcb->sctp_ep->sctp_lport != lport) {
1993				SCTP_TCB_UNLOCK(stcb);
1994				continue;
1995			}
1996			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1997				SCTP_TCB_UNLOCK(stcb);
1998				continue;
1999			}
2000			/* RRS:Need toaddr check here */
2001			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2002				/* Endpoint does not own this address */
2003				SCTP_TCB_UNLOCK(stcb);
2004				continue;
2005			}
2006			if (remote_tag) {
2007				/*
2008				 * If we have both vtags thats all we match
2009				 * on
2010				 */
2011				if (stcb->asoc.peer_vtag == remote_tag) {
2012					/*
2013					 * If both tags match we consider it
2014					 * conclusive and check NO
2015					 * source/destination addresses
2016					 */
2017					goto conclusive;
2018				}
2019			}
2020			if (skip_src_check) {
2021		conclusive:
2022				if (from) {
2023					net = sctp_findnet(stcb, from);
2024				} else {
2025					*netp = NULL;	/* unknown */
2026				}
2027				if (inp_p)
2028					*inp_p = stcb->sctp_ep;
2029				SCTP_INP_INFO_RUNLOCK();
2030				return (stcb);
2031			}
2032			net = sctp_findnet(stcb, from);
2033			if (net) {
2034				/* yep its him. */
2035				*netp = net;
2036				SCTP_STAT_INCR(sctps_vtagexpress);
2037				*inp_p = stcb->sctp_ep;
2038				SCTP_INP_INFO_RUNLOCK();
2039				return (stcb);
2040			} else {
2041				/*
2042				 * not him, this should only happen in rare
2043				 * cases so I peg it.
2044				 */
2045				SCTP_STAT_INCR(sctps_vtagbogus);
2046			}
2047		}
2048		SCTP_TCB_UNLOCK(stcb);
2049	}
2050	SCTP_INP_INFO_RUNLOCK();
2051	return (NULL);
2052}
2053
2054/*
2055 * Find an association with the pointer to the inbound IP packet. This can be
2056 * a IPv4 or IPv6 packet.
2057 */
2058struct sctp_tcb *
2059sctp_findassociation_addr(struct mbuf *m, int iphlen, int offset,
2060    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2061    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2062{
2063	int find_tcp_pool;
2064	struct ip *iph;
2065	struct sctp_tcb *retval;
2066	struct sockaddr_storage to_store, from_store;
2067	struct sockaddr *to = (struct sockaddr *)&to_store;
2068	struct sockaddr *from = (struct sockaddr *)&from_store;
2069	struct sctp_inpcb *inp;
2070
2071	iph = mtod(m, struct ip *);
2072	switch (iph->ip_v) {
2073	case IPVERSION:
2074		{
2075			/* its IPv4 */
2076			struct sockaddr_in *from4;
2077
2078			from4 = (struct sockaddr_in *)&from_store;
2079			bzero(from4, sizeof(*from4));
2080			from4->sin_family = AF_INET;
2081			from4->sin_len = sizeof(struct sockaddr_in);
2082			from4->sin_addr.s_addr = iph->ip_src.s_addr;
2083			from4->sin_port = sh->src_port;
2084			break;
2085		}
2086#ifdef INET6
2087	case IPV6_VERSION >> 4:
2088		{
2089			/* its IPv6 */
2090			struct ip6_hdr *ip6;
2091			struct sockaddr_in6 *from6;
2092
2093			ip6 = mtod(m, struct ip6_hdr *);
2094			from6 = (struct sockaddr_in6 *)&from_store;
2095			bzero(from6, sizeof(*from6));
2096			from6->sin6_family = AF_INET6;
2097			from6->sin6_len = sizeof(struct sockaddr_in6);
2098			from6->sin6_addr = ip6->ip6_src;
2099			from6->sin6_port = sh->src_port;
2100			/* Get the scopes in properly to the sin6 addr's */
2101			/* we probably don't need these operations */
2102			(void)sa6_recoverscope(from6);
2103			sa6_embedscope(from6, MODULE_GLOBAL(ip6_use_defzone));
2104			break;
2105		}
2106#endif
2107	default:
2108		/* Currently not supported. */
2109		return (NULL);
2110	}
2111
2112
2113	switch (iph->ip_v) {
2114	case IPVERSION:
2115		{
2116			/* its IPv4 */
2117			struct sockaddr_in *to4;
2118
2119			to4 = (struct sockaddr_in *)&to_store;
2120			bzero(to4, sizeof(*to4));
2121			to4->sin_family = AF_INET;
2122			to4->sin_len = sizeof(struct sockaddr_in);
2123			to4->sin_addr.s_addr = iph->ip_dst.s_addr;
2124			to4->sin_port = sh->dest_port;
2125			break;
2126		}
2127#ifdef INET6
2128	case IPV6_VERSION >> 4:
2129		{
2130			/* its IPv6 */
2131			struct ip6_hdr *ip6;
2132			struct sockaddr_in6 *to6;
2133
2134			ip6 = mtod(m, struct ip6_hdr *);
2135			to6 = (struct sockaddr_in6 *)&to_store;
2136			bzero(to6, sizeof(*to6));
2137			to6->sin6_family = AF_INET6;
2138			to6->sin6_len = sizeof(struct sockaddr_in6);
2139			to6->sin6_addr = ip6->ip6_dst;
2140			to6->sin6_port = sh->dest_port;
2141			/* Get the scopes in properly to the sin6 addr's */
2142			/* we probably don't need these operations */
2143			(void)sa6_recoverscope(to6);
2144			sa6_embedscope(to6, MODULE_GLOBAL(ip6_use_defzone));
2145			break;
2146		}
2147#endif
2148	default:
2149		/* TSNH */
2150		break;
2151	}
2152	if (sh->v_tag) {
2153		/* we only go down this path if vtag is non-zero */
2154		retval = sctp_findassoc_by_vtag(from, to, ntohl(sh->v_tag),
2155		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2156		if (retval) {
2157			return (retval);
2158		}
2159	}
2160	find_tcp_pool = 0;
2161	if ((ch->chunk_type != SCTP_INITIATION) &&
2162	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2163	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2164	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2165		/* Other chunk types go to the tcp pool. */
2166		find_tcp_pool = 1;
2167	}
2168	if (inp_p) {
2169		retval = sctp_findassociation_addr_sa(to, from, inp_p, netp,
2170		    find_tcp_pool, vrf_id);
2171		inp = *inp_p;
2172	} else {
2173		retval = sctp_findassociation_addr_sa(to, from, &inp, netp,
2174		    find_tcp_pool, vrf_id);
2175	}
2176	SCTPDBG(SCTP_DEBUG_PCB1, "retval:%p inp:%p\n", retval, inp);
2177	if (retval == NULL && inp) {
2178		/* Found a EP but not this address */
2179		if ((ch->chunk_type == SCTP_INITIATION) ||
2180		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2181			/*-
2182			 * special hook, we do NOT return linp or an
2183			 * association that is linked to an existing
2184			 * association that is under the TCP pool (i.e. no
2185			 * listener exists). The endpoint finding routine
2186			 * will always find a listner before examining the
2187			 * TCP pool.
2188			 */
2189			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2190				if (inp_p) {
2191					*inp_p = NULL;
2192				}
2193				return (NULL);
2194			}
2195			retval = sctp_findassociation_special_addr(m, iphlen,
2196			    offset, sh, &inp, netp, to);
2197			if (inp_p != NULL) {
2198				*inp_p = inp;
2199			}
2200		}
2201	}
2202	SCTPDBG(SCTP_DEBUG_PCB1, "retval is %p\n", retval);
2203	return (retval);
2204}
2205
2206/*
2207 * lookup an association by an ASCONF lookup address.
2208 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2209 */
2210struct sctp_tcb *
2211sctp_findassociation_ep_asconf(struct mbuf *m, int iphlen, int offset,
2212    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2213{
2214	struct sctp_tcb *stcb;
2215	struct sockaddr_in *sin;
2216
2217#ifdef INET6
2218	struct sockaddr_in6 *sin6;
2219
2220#endif
2221	struct sockaddr_storage local_store, remote_store;
2222	struct sockaddr *to;
2223	struct ip *iph;
2224
2225#ifdef INET6
2226	struct ip6_hdr *ip6;
2227
2228#endif
2229	struct sctp_paramhdr parm_buf, *phdr;
2230	int ptype;
2231	int zero_address = 0;
2232
2233
2234	memset(&local_store, 0, sizeof(local_store));
2235	memset(&remote_store, 0, sizeof(remote_store));
2236	to = (struct sockaddr *)&local_store;
2237	/* First get the destination address setup too. */
2238	iph = mtod(m, struct ip *);
2239	switch (iph->ip_v) {
2240	case IPVERSION:
2241		/* its IPv4 */
2242		sin = (struct sockaddr_in *)&local_store;
2243		sin->sin_family = AF_INET;
2244		sin->sin_len = sizeof(*sin);
2245		sin->sin_port = sh->dest_port;
2246		sin->sin_addr.s_addr = iph->ip_dst.s_addr;
2247		break;
2248#ifdef INET6
2249	case IPV6_VERSION >> 4:
2250		/* its IPv6 */
2251		ip6 = mtod(m, struct ip6_hdr *);
2252		sin6 = (struct sockaddr_in6 *)&local_store;
2253		sin6->sin6_family = AF_INET6;
2254		sin6->sin6_len = sizeof(*sin6);
2255		sin6->sin6_port = sh->dest_port;
2256		sin6->sin6_addr = ip6->ip6_dst;
2257		break;
2258#endif
2259	default:
2260		return NULL;
2261	}
2262
2263	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2264	    &parm_buf, sizeof(struct sctp_paramhdr));
2265	if (phdr == NULL) {
2266		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2267		    __FUNCTION__);
2268		return NULL;
2269	}
2270	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2271	/* get the correlation address */
2272	switch (ptype) {
2273#ifdef INET6
2274	case SCTP_IPV6_ADDRESS:
2275		{
2276			/* ipv6 address param */
2277			struct sctp_ipv6addr_param *p6, p6_buf;
2278
2279			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2280				return NULL;
2281			}
2282			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2283			    offset + sizeof(struct sctp_asconf_chunk),
2284			    &p6_buf.ph, sizeof(*p6));
2285			if (p6 == NULL) {
2286				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2287				    __FUNCTION__);
2288				return (NULL);
2289			}
2290			sin6 = (struct sockaddr_in6 *)&remote_store;
2291			sin6->sin6_family = AF_INET6;
2292			sin6->sin6_len = sizeof(*sin6);
2293			sin6->sin6_port = sh->src_port;
2294			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2295			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2296				zero_address = 1;
2297			break;
2298		}
2299#endif
2300	case SCTP_IPV4_ADDRESS:
2301		{
2302			/* ipv4 address param */
2303			struct sctp_ipv4addr_param *p4, p4_buf;
2304
2305			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2306				return NULL;
2307			}
2308			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2309			    offset + sizeof(struct sctp_asconf_chunk),
2310			    &p4_buf.ph, sizeof(*p4));
2311			if (p4 == NULL) {
2312				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2313				    __FUNCTION__);
2314				return (NULL);
2315			}
2316			sin = (struct sockaddr_in *)&remote_store;
2317			sin->sin_family = AF_INET;
2318			sin->sin_len = sizeof(*sin);
2319			sin->sin_port = sh->src_port;
2320			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2321			if (sin->sin_addr.s_addr == INADDR_ANY)
2322				zero_address = 1;
2323			break;
2324		}
2325	default:
2326		/* invalid address param type */
2327		return NULL;
2328	}
2329
2330	if (zero_address) {
2331		stcb = sctp_findassoc_by_vtag(NULL, to, ntohl(sh->v_tag), inp_p,
2332		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2333		/*
2334		 * printf("findassociation_ep_asconf: zero lookup address
2335		 * finds stcb 0x%x\n", (uint32_t)stcb);
2336		 */
2337	} else {
2338		stcb = sctp_findassociation_ep_addr(inp_p,
2339		    (struct sockaddr *)&remote_store, netp,
2340		    to, NULL);
2341	}
2342	return (stcb);
2343}
2344
2345
2346/*
2347 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2348 * addresses. This way if we don't get a bind we by default pick a ephemeral
2349 * port with all addresses bound.
2350 */
2351int
2352sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2353{
2354	/*
2355	 * we get called when a new endpoint starts up. We need to allocate
2356	 * the sctp_inpcb structure from the zone and init it. Mark it as
2357	 * unbound and find a port that we can use as an ephemeral with
2358	 * INADDR_ANY. If the user binds later no problem we can then add in
2359	 * the specific addresses. And setup the default parameters for the
2360	 * EP.
2361	 */
2362	int i, error;
2363	struct sctp_inpcb *inp;
2364	struct sctp_pcb *m;
2365	struct timeval time;
2366	sctp_sharedkey_t *null_key;
2367
2368	error = 0;
2369
2370	SCTP_INP_INFO_WLOCK();
2371	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2372	if (inp == NULL) {
2373		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2374		SCTP_INP_INFO_WUNLOCK();
2375		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2376		return (ENOBUFS);
2377	}
2378	/* zap it */
2379	bzero(inp, sizeof(*inp));
2380
2381	/* bump generations */
2382	/* setup socket pointers */
2383	inp->sctp_socket = so;
2384	inp->ip_inp.inp.inp_socket = so;
2385	inp->sctp_associd_counter = 1;
2386	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2387	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2388	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2389	/* init the small hash table we use to track asocid <-> tcb */
2390	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2391	if (inp->sctp_asocidhash == NULL) {
2392		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2393		SCTP_INP_INFO_WUNLOCK();
2394		return (ENOBUFS);
2395	}
2396#ifdef IPSEC
2397	{
2398		struct inpcbpolicy *pcb_sp = NULL;
2399
2400		error = ipsec_init_policy(so, &pcb_sp);
2401		/* Arrange to share the policy */
2402		inp->ip_inp.inp.inp_sp = pcb_sp;
2403		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2404	}
2405	if (error != 0) {
2406		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2407		SCTP_INP_INFO_WUNLOCK();
2408		return error;
2409	}
2410#endif				/* IPSEC */
2411	SCTP_INCR_EP_COUNT();
2412	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2413	SCTP_INP_INFO_WUNLOCK();
2414
2415	so->so_pcb = (caddr_t)inp;
2416
2417	if ((SCTP_SO_TYPE(so) == SOCK_DGRAM) ||
2418	    (SCTP_SO_TYPE(so) == SOCK_SEQPACKET)) {
2419		/* UDP style socket */
2420		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2421		    SCTP_PCB_FLAGS_UNBOUND);
2422		/* Be sure it is NON-BLOCKING IO for UDP */
2423		/* SCTP_SET_SO_NBIO(so); */
2424	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2425		/* TCP style socket */
2426		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2427		    SCTP_PCB_FLAGS_UNBOUND);
2428		/* Be sure we have blocking IO by default */
2429		SCTP_CLEAR_SO_NBIO(so);
2430	} else {
2431		/*
2432		 * unsupported socket type (RAW, etc)- in case we missed it
2433		 * in protosw
2434		 */
2435		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2436		so->so_pcb = NULL;
2437		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2438		return (EOPNOTSUPP);
2439	}
2440	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2441		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2442		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2443	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2444		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2445		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2446	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2447		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2448		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2449	}
2450	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2451	    &inp->sctp_hashmark);
2452	if (inp->sctp_tcbhash == NULL) {
2453		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2454		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2455		so->so_pcb = NULL;
2456		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2457		return (ENOBUFS);
2458	}
2459	inp->def_vrf_id = vrf_id;
2460
2461	SCTP_INP_INFO_WLOCK();
2462	SCTP_INP_LOCK_INIT(inp);
2463	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2464	SCTP_INP_READ_INIT(inp);
2465	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2466	/* lock the new ep */
2467	SCTP_INP_WLOCK(inp);
2468
2469	/* add it to the info area */
2470	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2471	SCTP_INP_INFO_WUNLOCK();
2472
2473	TAILQ_INIT(&inp->read_queue);
2474	LIST_INIT(&inp->sctp_addr_list);
2475
2476	LIST_INIT(&inp->sctp_asoc_list);
2477
2478#ifdef SCTP_TRACK_FREED_ASOCS
2479	/* TEMP CODE */
2480	LIST_INIT(&inp->sctp_asoc_free_list);
2481#endif
2482	/* Init the timer structure for signature change */
2483	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2484	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2485
2486	/* now init the actual endpoint default data */
2487	m = &inp->sctp_ep;
2488
2489	/* setup the base timeout information */
2490	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2491	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2492	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2493	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2494	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2495	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2496	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2497	/* all max/min max are in ms */
2498	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2499	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2500	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2501	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2502	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2503
2504	m->max_open_streams_intome = MAX_SCTP_STREAMS;
2505
2506	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2507	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2508	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2509	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2510	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2511	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2512	if ((SCTP_BASE_SYSCTL(sctp_default_cc_module) >= SCTP_CC_RFC2581) &&
2513	    (SCTP_BASE_SYSCTL(sctp_default_cc_module) <= SCTP_CC_HTCP)) {
2514		m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2515	} else {
2516		/* sysctl done with invalid value, set to 2581 */
2517		m->sctp_default_cc_module = SCTP_CC_RFC2581;
2518	}
2519	/* number of streams to pre-open on a association */
2520	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2521
2522	/* Add adaptation cookie */
2523	m->adaptation_layer_indicator = 0x504C5253;
2524
2525	/* seed random number generator */
2526	m->random_counter = 1;
2527	m->store_at = SCTP_SIGNATURE_SIZE;
2528	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2529	sctp_fill_random_store(m);
2530
2531	/* Minimum cookie size */
2532	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2533	    sizeof(struct sctp_state_cookie);
2534	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2535
2536	/* Setup the initial secret */
2537	(void)SCTP_GETTIME_TIMEVAL(&time);
2538	m->time_of_secret_change = time.tv_sec;
2539
2540	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2541		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2542	}
2543	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2544
2545	/* How long is a cookie good for ? */
2546	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2547	/*
2548	 * Initialize authentication parameters
2549	 */
2550	m->local_hmacs = sctp_default_supported_hmaclist();
2551	m->local_auth_chunks = sctp_alloc_chunklist();
2552	sctp_auth_set_default_chunks(m->local_auth_chunks);
2553	LIST_INIT(&m->shared_keys);
2554	/* add default NULL key as key id 0 */
2555	null_key = sctp_alloc_sharedkey();
2556	sctp_insert_sharedkey(&m->shared_keys, null_key);
2557	SCTP_INP_WUNLOCK(inp);
2558#ifdef SCTP_LOG_CLOSING
2559	sctp_log_closing(inp, NULL, 12);
2560#endif
2561	return (error);
2562}
2563
2564
2565void
2566sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2567    struct sctp_tcb *stcb)
2568{
2569	struct sctp_nets *net;
2570	uint16_t lport, rport;
2571	struct sctppcbhead *head;
2572	struct sctp_laddr *laddr, *oladdr;
2573
2574	atomic_add_int(&stcb->asoc.refcnt, 1);
2575	SCTP_TCB_UNLOCK(stcb);
2576	SCTP_INP_INFO_WLOCK();
2577	SCTP_INP_WLOCK(old_inp);
2578	SCTP_INP_WLOCK(new_inp);
2579	SCTP_TCB_LOCK(stcb);
2580	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2581
2582	new_inp->sctp_ep.time_of_secret_change =
2583	    old_inp->sctp_ep.time_of_secret_change;
2584	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2585	    sizeof(old_inp->sctp_ep.secret_key));
2586	new_inp->sctp_ep.current_secret_number =
2587	    old_inp->sctp_ep.current_secret_number;
2588	new_inp->sctp_ep.last_secret_number =
2589	    old_inp->sctp_ep.last_secret_number;
2590	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2591
2592	/* make it so new data pours into the new socket */
2593	stcb->sctp_socket = new_inp->sctp_socket;
2594	stcb->sctp_ep = new_inp;
2595
2596	/* Copy the port across */
2597	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2598	rport = stcb->rport;
2599	/* Pull the tcb from the old association */
2600	LIST_REMOVE(stcb, sctp_tcbhash);
2601	LIST_REMOVE(stcb, sctp_tcblist);
2602	if (stcb->asoc.in_asocid_hash) {
2603		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2604	}
2605	/* Now insert the new_inp into the TCP connected hash */
2606	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport),
2607	    SCTP_BASE_INFO(hashtcpmark))];
2608
2609	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2610	/* Its safe to access */
2611	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2612
2613	/* Now move the tcb into the endpoint list */
2614	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2615	/*
2616	 * Question, do we even need to worry about the ep-hash since we
2617	 * only have one connection? Probably not :> so lets get rid of it
2618	 * and not suck up any kernel memory in that.
2619	 */
2620	if (stcb->asoc.in_asocid_hash) {
2621		struct sctpasochead *lhd;
2622
2623		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2624		    new_inp->hashasocidmark)];
2625		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2626	}
2627	/* Ok. Let's restart timer. */
2628	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2629		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2630		    stcb, net);
2631	}
2632
2633	SCTP_INP_INFO_WUNLOCK();
2634	if (new_inp->sctp_tcbhash != NULL) {
2635		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2636		new_inp->sctp_tcbhash = NULL;
2637	}
2638	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2639		/* Subset bound, so copy in the laddr list from the old_inp */
2640		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2641			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2642			if (laddr == NULL) {
2643				/*
2644				 * Gak, what can we do? This assoc is really
2645				 * HOSED. We probably should send an abort
2646				 * here.
2647				 */
2648				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2649				continue;
2650			}
2651			SCTP_INCR_LADDR_COUNT();
2652			bzero(laddr, sizeof(*laddr));
2653			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2654			laddr->ifa = oladdr->ifa;
2655			atomic_add_int(&laddr->ifa->refcount, 1);
2656			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2657			    sctp_nxt_addr);
2658			new_inp->laddr_count++;
2659		}
2660	}
2661	/*
2662	 * Now any running timers need to be adjusted since we really don't
2663	 * care if they are running or not just blast in the new_inp into
2664	 * all of them.
2665	 */
2666
2667	stcb->asoc.hb_timer.ep = (void *)new_inp;
2668	stcb->asoc.dack_timer.ep = (void *)new_inp;
2669	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2670	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2671	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2672	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2673	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2674	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2675	/* now what about the nets? */
2676	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2677		net->pmtu_timer.ep = (void *)new_inp;
2678		net->rxt_timer.ep = (void *)new_inp;
2679		net->fr_timer.ep = (void *)new_inp;
2680	}
2681	SCTP_INP_WUNLOCK(new_inp);
2682	SCTP_INP_WUNLOCK(old_inp);
2683}
2684
2685
2686
2687
2688/* sctp_ifap is used to bypass normal local address validation checks */
2689int
2690sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2691    struct sctp_ifa *sctp_ifap, struct thread *p)
2692{
2693	/* bind a ep to a socket address */
2694	struct sctppcbhead *head;
2695	struct sctp_inpcb *inp, *inp_tmp;
2696	struct inpcb *ip_inp;
2697	int port_reuse_active = 0;
2698	int bindall;
2699	uint16_t lport;
2700	int error;
2701	uint32_t vrf_id;
2702
2703	lport = 0;
2704	error = 0;
2705	bindall = 1;
2706	inp = (struct sctp_inpcb *)so->so_pcb;
2707	ip_inp = (struct inpcb *)so->so_pcb;
2708#ifdef SCTP_DEBUG
2709	if (addr) {
2710		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port:%d\n",
2711		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2712		SCTPDBG(SCTP_DEBUG_PCB1, "Addr :");
2713		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2714	}
2715#endif
2716	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2717		/* already did a bind, subsequent binds NOT allowed ! */
2718		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2719		return (EINVAL);
2720	}
2721#ifdef INVARIANTS
2722	if (p == NULL)
2723		panic("null proc/thread");
2724#endif
2725	if (addr != NULL) {
2726		switch (addr->sa_family) {
2727		case AF_INET:
2728			{
2729				struct sockaddr_in *sin;
2730
2731				/* IPV6_V6ONLY socket? */
2732				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2733					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2734					return (EINVAL);
2735				}
2736				if (addr->sa_len != sizeof(*sin)) {
2737					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2738					return (EINVAL);
2739				}
2740				sin = (struct sockaddr_in *)addr;
2741				lport = sin->sin_port;
2742				/*
2743				 * For LOOPBACK the prison_local_ip4() call
2744				 * will transmute the ip address to the
2745				 * proper value.
2746				 */
2747				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2748					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2749					return (error);
2750				}
2751				if (sin->sin_addr.s_addr != INADDR_ANY) {
2752					bindall = 0;
2753				}
2754				break;
2755			}
2756#ifdef INET6
2757		case AF_INET6:
2758			{
2759				/*
2760				 * Only for pure IPv6 Address. (No IPv4
2761				 * Mapped!)
2762				 */
2763				struct sockaddr_in6 *sin6;
2764
2765				sin6 = (struct sockaddr_in6 *)addr;
2766
2767				if (addr->sa_len != sizeof(*sin6)) {
2768					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2769					return (EINVAL);
2770				}
2771				lport = sin6->sin6_port;
2772
2773				/*
2774				 * For LOOPBACK the prison_local_ip6() call
2775				 * will transmute the ipv6 address to the
2776				 * proper value.
2777				 */
2778				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2779				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2780					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2781					return (error);
2782				}
2783				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2784					bindall = 0;
2785					/* KAME hack: embed scopeid */
2786					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2787						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2788						return (EINVAL);
2789					}
2790				}
2791				/* this must be cleared for ifa_ifwithaddr() */
2792				sin6->sin6_scope_id = 0;
2793				break;
2794			}
2795#endif
2796		default:
2797			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2798			return (EAFNOSUPPORT);
2799		}
2800	}
2801	SCTP_INP_INFO_WLOCK();
2802	SCTP_INP_WLOCK(inp);
2803	/* Setup a vrf_id to be the default for the non-bind-all case. */
2804	vrf_id = inp->def_vrf_id;
2805
2806	/* increase our count due to the unlock we do */
2807	SCTP_INP_INCR_REF(inp);
2808	if (lport) {
2809		/*
2810		 * Did the caller specify a port? if so we must see if a ep
2811		 * already has this one bound.
2812		 */
2813		/* got to be root to get at low ports */
2814		if (ntohs(lport) < IPPORT_RESERVED) {
2815			if (p && (error =
2816			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2817			    )) {
2818				SCTP_INP_DECR_REF(inp);
2819				SCTP_INP_WUNLOCK(inp);
2820				SCTP_INP_INFO_WUNLOCK();
2821				return (error);
2822			}
2823		}
2824		if (p == NULL) {
2825			SCTP_INP_DECR_REF(inp);
2826			SCTP_INP_WUNLOCK(inp);
2827			SCTP_INP_INFO_WUNLOCK();
2828			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2829			return (error);
2830		}
2831		SCTP_INP_WUNLOCK(inp);
2832		if (bindall) {
2833			vrf_id = inp->def_vrf_id;
2834			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2835			if (inp_tmp != NULL) {
2836				/*
2837				 * lock guy returned and lower count note
2838				 * that we are not bound so inp_tmp should
2839				 * NEVER be inp. And it is this inp
2840				 * (inp_tmp) that gets the reference bump,
2841				 * so we must lower it.
2842				 */
2843				SCTP_INP_DECR_REF(inp_tmp);
2844				/* unlock info */
2845				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2846				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2847					/*
2848					 * Ok, must be one-2-one and
2849					 * allowing port re-use
2850					 */
2851					port_reuse_active = 1;
2852					goto continue_anyway;
2853				}
2854				SCTP_INP_DECR_REF(inp);
2855				SCTP_INP_INFO_WUNLOCK();
2856				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2857				return (EADDRINUSE);
2858			}
2859		} else {
2860			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2861			if (inp_tmp != NULL) {
2862				/*
2863				 * lock guy returned and lower count note
2864				 * that we are not bound so inp_tmp should
2865				 * NEVER be inp. And it is this inp
2866				 * (inp_tmp) that gets the reference bump,
2867				 * so we must lower it.
2868				 */
2869				SCTP_INP_DECR_REF(inp_tmp);
2870				/* unlock info */
2871				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2872				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2873					/*
2874					 * Ok, must be one-2-one and
2875					 * allowing port re-use
2876					 */
2877					port_reuse_active = 1;
2878					goto continue_anyway;
2879				}
2880				SCTP_INP_DECR_REF(inp);
2881				SCTP_INP_INFO_WUNLOCK();
2882				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2883				return (EADDRINUSE);
2884			}
2885		}
2886continue_anyway:
2887		SCTP_INP_WLOCK(inp);
2888		if (bindall) {
2889			/* verify that no lport is not used by a singleton */
2890			if ((port_reuse_active == 0) &&
2891			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))
2892			    ) {
2893				/* Sorry someone already has this one bound */
2894				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2895				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2896					port_reuse_active = 1;
2897				} else {
2898					SCTP_INP_DECR_REF(inp);
2899					SCTP_INP_WUNLOCK(inp);
2900					SCTP_INP_INFO_WUNLOCK();
2901					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2902					return (EADDRINUSE);
2903				}
2904			}
2905		}
2906	} else {
2907		uint16_t first, last, candidate;
2908		uint16_t count;
2909		int done;
2910
2911		if (ip_inp->inp_flags & INP_HIGHPORT) {
2912			first = MODULE_GLOBAL(ipport_hifirstauto);
2913			last = MODULE_GLOBAL(ipport_hilastauto);
2914		} else if (ip_inp->inp_flags & INP_LOWPORT) {
2915			if (p && (error =
2916			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2917			    )) {
2918				SCTP_INP_DECR_REF(inp);
2919				SCTP_INP_WUNLOCK(inp);
2920				SCTP_INP_INFO_WUNLOCK();
2921				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2922				return (error);
2923			}
2924			first = MODULE_GLOBAL(ipport_lowfirstauto);
2925			last = MODULE_GLOBAL(ipport_lowlastauto);
2926		} else {
2927			first = MODULE_GLOBAL(ipport_firstauto);
2928			last = MODULE_GLOBAL(ipport_lastauto);
2929		}
2930		if (first > last) {
2931			uint16_t temp;
2932
2933			temp = first;
2934			first = last;
2935			last = temp;
2936		}
2937		count = last - first + 1;	/* number of candidates */
2938		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
2939
2940		done = 0;
2941		while (!done) {
2942			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
2943				done = 1;
2944			}
2945			if (!done) {
2946				if (--count == 0) {
2947					SCTP_INP_DECR_REF(inp);
2948					SCTP_INP_WUNLOCK(inp);
2949					SCTP_INP_INFO_WUNLOCK();
2950					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2951					return (EADDRINUSE);
2952				}
2953				if (candidate == last)
2954					candidate = first;
2955				else
2956					candidate = candidate + 1;
2957			}
2958		}
2959		lport = htons(candidate);
2960	}
2961	SCTP_INP_DECR_REF(inp);
2962	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2963	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2964		/*
2965		 * this really should not happen. The guy did a non-blocking
2966		 * bind and then did a close at the same time.
2967		 */
2968		SCTP_INP_WUNLOCK(inp);
2969		SCTP_INP_INFO_WUNLOCK();
2970		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2971		return (EINVAL);
2972	}
2973	/* ok we look clear to give out this port, so lets setup the binding */
2974	if (bindall) {
2975		/* binding to all addresses, so just set in the proper flags */
2976		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
2977		/* set the automatic addr changes from kernel flag */
2978		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
2979			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2980			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2981		} else {
2982			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2983			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2984		}
2985		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
2986			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
2987		} else {
2988			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
2989		}
2990		/*
2991		 * set the automatic mobility_base from kernel flag (by
2992		 * micchie)
2993		 */
2994		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
2995			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
2996			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2997		} else {
2998			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
2999			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3000		}
3001		/*
3002		 * set the automatic mobility_fasthandoff from kernel flag
3003		 * (by micchie)
3004		 */
3005		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3006			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3007			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3008		} else {
3009			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3010			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3011		}
3012	} else {
3013		/*
3014		 * bind specific, make sure flags is off and add a new
3015		 * address structure to the sctp_addr_list inside the ep
3016		 * structure.
3017		 *
3018		 * We will need to allocate one and insert it at the head. The
3019		 * socketopt call can just insert new addresses in there as
3020		 * well. It will also have to do the embed scope kame hack
3021		 * too (before adding).
3022		 */
3023		struct sctp_ifa *ifa;
3024		struct sockaddr_storage store_sa;
3025
3026		memset(&store_sa, 0, sizeof(store_sa));
3027		if (addr->sa_family == AF_INET) {
3028			struct sockaddr_in *sin;
3029
3030			sin = (struct sockaddr_in *)&store_sa;
3031			memcpy(sin, addr, sizeof(struct sockaddr_in));
3032			sin->sin_port = 0;
3033		} else if (addr->sa_family == AF_INET6) {
3034			struct sockaddr_in6 *sin6;
3035
3036			sin6 = (struct sockaddr_in6 *)&store_sa;
3037			memcpy(sin6, addr, sizeof(struct sockaddr_in6));
3038			sin6->sin6_port = 0;
3039		}
3040		/*
3041		 * first find the interface with the bound address need to
3042		 * zero out the port to find the address! yuck! can't do
3043		 * this earlier since need port for sctp_pcb_findep()
3044		 */
3045		if (sctp_ifap != NULL)
3046			ifa = sctp_ifap;
3047		else {
3048			/*
3049			 * Note for BSD we hit here always other O/S's will
3050			 * pass things in via the sctp_ifap argument
3051			 * (Panda).
3052			 */
3053			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
3054			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3055		}
3056		if (ifa == NULL) {
3057			/* Can't find an interface with that address */
3058			SCTP_INP_WUNLOCK(inp);
3059			SCTP_INP_INFO_WUNLOCK();
3060			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3061			return (EADDRNOTAVAIL);
3062		}
3063		if (addr->sa_family == AF_INET6) {
3064			/* GAK, more FIXME IFA lock? */
3065			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3066				/* Can't bind a non-existent addr. */
3067				SCTP_INP_WUNLOCK(inp);
3068				SCTP_INP_INFO_WUNLOCK();
3069				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3070				return (EINVAL);
3071			}
3072		}
3073		/* we're not bound all */
3074		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3075		/* allow bindx() to send ASCONF's for binding changes */
3076		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3077		/* clear automatic addr changes from kernel flag */
3078		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3079
3080		/* add this address to the endpoint list */
3081		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3082		if (error != 0) {
3083			SCTP_INP_WUNLOCK(inp);
3084			SCTP_INP_INFO_WUNLOCK();
3085			return (error);
3086		}
3087		inp->laddr_count++;
3088	}
3089	/* find the bucket */
3090	if (port_reuse_active) {
3091		/* Put it into tcp 1-2-1 hash */
3092		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport),
3093		    SCTP_BASE_INFO(hashtcpmark))];
3094		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3095	} else {
3096		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
3097		    SCTP_BASE_INFO(hashmark))];
3098	}
3099	/* put it in the bucket */
3100	LIST_INSERT_HEAD(head, inp, sctp_hash);
3101	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3102	    head, ntohs(lport), port_reuse_active);
3103	/* set in the port */
3104	inp->sctp_lport = lport;
3105
3106	/* turn off just the unbound flag */
3107	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3108	SCTP_INP_WUNLOCK(inp);
3109	SCTP_INP_INFO_WUNLOCK();
3110	return (0);
3111}
3112
3113
3114static void
3115sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3116{
3117	struct sctp_iterator *it, *nit;
3118
3119	/*
3120	 * We enter with the only the ITERATOR_LOCK in place and a write
3121	 * lock on the inp_info stuff.
3122	 */
3123	it = sctp_it_ctl.cur_it;
3124	if (it && (it->vn != curvnet)) {
3125		/* Its not looking at our VNET */
3126		return;
3127	}
3128	if (it && (it->inp == inp)) {
3129		/*
3130		 * This is tricky and we hold the iterator lock, but when it
3131		 * returns and gets the lock (when we release it) the
3132		 * iterator will try to operate on inp. We need to stop that
3133		 * from happening. But of course the iterator has a
3134		 * reference on the stcb and inp. We can mark it and it will
3135		 * stop.
3136		 *
3137		 * If its a single iterator situation, we set the end iterator
3138		 * flag. Otherwise we set the iterator to go to the next
3139		 * inp.
3140		 *
3141		 */
3142		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3143			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3144		} else {
3145			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3146		}
3147	}
3148	/*
3149	 * Now go through and remove any single reference to our inp that
3150	 * may be still pending on the list
3151	 */
3152	SCTP_IPI_ITERATOR_WQ_LOCK();
3153	it = TAILQ_FIRST(&sctp_it_ctl.iteratorhead);
3154	while (it) {
3155		nit = TAILQ_NEXT(it, sctp_nxt_itr);
3156		if (it->vn != curvnet) {
3157			it = nit;
3158			continue;
3159		}
3160		if (it->inp == inp) {
3161			/* This one points to me is it inp specific? */
3162			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3163				/* Remove and free this one */
3164				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3165				    it, sctp_nxt_itr);
3166				if (it->function_atend != NULL) {
3167					(*it->function_atend) (it->pointer, it->val);
3168				}
3169				SCTP_FREE(it, SCTP_M_ITER);
3170			} else {
3171				it->inp = LIST_NEXT(it->inp, sctp_list);
3172				if (it->inp) {
3173					SCTP_INP_INCR_REF(it->inp);
3174				}
3175			}
3176			/*
3177			 * When its put in the refcnt is incremented so decr
3178			 * it
3179			 */
3180			SCTP_INP_DECR_REF(inp);
3181		}
3182		it = nit;
3183	}
3184	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3185}
3186
3187/* release sctp_inpcb unbind the port */
3188void
3189sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3190{
3191	/*
3192	 * Here we free a endpoint. We must find it (if it is in the Hash
3193	 * table) and remove it from there. Then we must also find it in the
3194	 * overall list and remove it from there. After all removals are
3195	 * complete then any timer has to be stopped. Then start the actual
3196	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3197	 * all associations. d) finally the ep itself.
3198	 */
3199	struct sctp_pcb *m;
3200	struct sctp_tcb *asoc, *nasoc;
3201	struct sctp_laddr *laddr, *nladdr;
3202	struct inpcb *ip_pcb;
3203	struct socket *so;
3204	int being_refed = 0;
3205	struct sctp_queued_to_read *sq;
3206
3207
3208	int cnt;
3209	sctp_sharedkey_t *shared_key;
3210
3211
3212#ifdef SCTP_LOG_CLOSING
3213	sctp_log_closing(inp, NULL, 0);
3214#endif
3215	SCTP_ITERATOR_LOCK();
3216	/* mark any iterators on the list or being processed */
3217	sctp_iterator_inp_being_freed(inp);
3218	SCTP_ITERATOR_UNLOCK();
3219	so = inp->sctp_socket;
3220	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3221		/* been here before.. eeks.. get out of here */
3222		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3223#ifdef SCTP_LOG_CLOSING
3224		sctp_log_closing(inp, NULL, 1);
3225#endif
3226		return;
3227	}
3228	SCTP_ASOC_CREATE_LOCK(inp);
3229	SCTP_INP_INFO_WLOCK();
3230
3231	SCTP_INP_WLOCK(inp);
3232	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3233		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3234		/* socket is gone, so no more wakeups allowed */
3235		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3236		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3237		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3238
3239	}
3240	/* First time through we have the socket lock, after that no more. */
3241	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3242	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3243
3244	if (inp->control) {
3245		sctp_m_freem(inp->control);
3246		inp->control = NULL;
3247	}
3248	if (inp->pkt) {
3249		sctp_m_freem(inp->pkt);
3250		inp->pkt = NULL;
3251	}
3252	m = &inp->sctp_ep;
3253	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3254					 * here but I will be nice :> (i.e.
3255					 * ip_pcb = ep;) */
3256	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3257		int cnt_in_sd;
3258
3259		cnt_in_sd = 0;
3260		for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
3261		    asoc = nasoc) {
3262			SCTP_TCB_LOCK(asoc);
3263			nasoc = LIST_NEXT(asoc, sctp_tcblist);
3264			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3265				/* Skip guys being freed */
3266				cnt_in_sd++;
3267				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3268					/*
3269					 * Special case - we did not start a
3270					 * kill timer on the asoc due to it
3271					 * was not closed. So go ahead and
3272					 * start it now.
3273					 */
3274					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3275					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3276				}
3277				SCTP_TCB_UNLOCK(asoc);
3278				continue;
3279			}
3280			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3281			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3282			    (asoc->asoc.total_output_queue_size == 0)) {
3283				/*
3284				 * If we have data in queue, we don't want
3285				 * to just free since the app may have done,
3286				 * send()/close or connect/send/close. And
3287				 * it wants the data to get across first.
3288				 */
3289				/* Just abandon things in the front states */
3290				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3291				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3292					cnt_in_sd++;
3293				}
3294				continue;
3295			}
3296			/* Disconnect the socket please */
3297			asoc->sctp_socket = NULL;
3298			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3299			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3300			    (asoc->asoc.control_pdapi) ||
3301			    (asoc->asoc.size_on_all_streams > 0) ||
3302			    (so && (so->so_rcv.sb_cc > 0))
3303			    ) {
3304				/* Left with Data unread */
3305				struct mbuf *op_err;
3306
3307				op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3308				    0, M_DONTWAIT, 1, MT_DATA);
3309				if (op_err) {
3310					/* Fill in the user initiated abort */
3311					struct sctp_paramhdr *ph;
3312					uint32_t *ippp;
3313
3314					SCTP_BUF_LEN(op_err) =
3315					    sizeof(struct sctp_paramhdr) + sizeof(uint32_t);
3316					ph = mtod(op_err,
3317					    struct sctp_paramhdr *);
3318					ph->param_type = htons(
3319					    SCTP_CAUSE_USER_INITIATED_ABT);
3320					ph->param_length = htons(SCTP_BUF_LEN(op_err));
3321					ippp = (uint32_t *) (ph + 1);
3322					*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3);
3323				}
3324				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3325#if defined(SCTP_PANIC_ON_ABORT)
3326				panic("inpcb_free does an abort");
3327#endif
3328				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3329				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3330				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3331				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3332					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3333				}
3334				if (sctp_free_assoc(inp, asoc,
3335				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3336					cnt_in_sd++;
3337				}
3338				continue;
3339			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3340				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3341				    (asoc->asoc.stream_queue_cnt == 0)
3342			    ) {
3343				if (asoc->asoc.locked_on_sending) {
3344					goto abort_anyway;
3345				}
3346				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3347				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3348					/*
3349					 * there is nothing queued to send,
3350					 * so I send shutdown
3351					 */
3352					sctp_send_shutdown(asoc, asoc->asoc.primary_destination);
3353					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3354					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3355						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3356					}
3357					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3358					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3359					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3360					    asoc->asoc.primary_destination);
3361					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3362					    asoc->asoc.primary_destination);
3363					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3364				}
3365			} else {
3366				/* mark into shutdown pending */
3367				struct sctp_stream_queue_pending *sp;
3368
3369				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3370				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3371				    asoc->asoc.primary_destination);
3372				if (asoc->asoc.locked_on_sending) {
3373					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3374					    sctp_streamhead);
3375					if (sp == NULL) {
3376						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3377						    asoc->asoc.locked_on_sending,
3378						    asoc->asoc.locked_on_sending->stream_no);
3379					} else {
3380						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3381							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3382					}
3383				}
3384				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3385				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3386				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3387					struct mbuf *op_err;
3388
3389			abort_anyway:
3390					op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3391					    0, M_DONTWAIT, 1, MT_DATA);
3392					if (op_err) {
3393						/*
3394						 * Fill in the user
3395						 * initiated abort
3396						 */
3397						struct sctp_paramhdr *ph;
3398						uint32_t *ippp;
3399
3400						SCTP_BUF_LEN(op_err) =
3401						    (sizeof(struct sctp_paramhdr) +
3402						    sizeof(uint32_t));
3403						ph = mtod(op_err,
3404						    struct sctp_paramhdr *);
3405						ph->param_type = htons(
3406						    SCTP_CAUSE_USER_INITIATED_ABT);
3407						ph->param_length = htons(SCTP_BUF_LEN(op_err));
3408						ippp = (uint32_t *) (ph + 1);
3409						*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5);
3410					}
3411					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3412#if defined(SCTP_PANIC_ON_ABORT)
3413					panic("inpcb_free does an abort");
3414#endif
3415
3416					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3417					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3418					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3419					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3420						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3421					}
3422					if (sctp_free_assoc(inp, asoc,
3423					    SCTP_PCBFREE_NOFORCE,
3424					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3425						cnt_in_sd++;
3426					}
3427					continue;
3428				} else {
3429					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3430				}
3431			}
3432			cnt_in_sd++;
3433			SCTP_TCB_UNLOCK(asoc);
3434		}
3435		/* now is there some left in our SHUTDOWN state? */
3436		if (cnt_in_sd) {
3437#ifdef SCTP_LOG_CLOSING
3438			sctp_log_closing(inp, NULL, 2);
3439#endif
3440			inp->sctp_socket = NULL;
3441			SCTP_INP_WUNLOCK(inp);
3442			SCTP_ASOC_CREATE_UNLOCK(inp);
3443			SCTP_INP_INFO_WUNLOCK();
3444			return;
3445		}
3446	}
3447	inp->sctp_socket = NULL;
3448	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3449	    SCTP_PCB_FLAGS_UNBOUND) {
3450		/*
3451		 * ok, this guy has been bound. It's port is somewhere in
3452		 * the SCTP_BASE_INFO(hash table). Remove it!
3453		 */
3454		LIST_REMOVE(inp, sctp_hash);
3455		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3456	}
3457	/*
3458	 * If there is a timer running to kill us, forget it, since it may
3459	 * have a contest on the INP lock.. which would cause us to die ...
3460	 */
3461	cnt = 0;
3462	for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
3463	    asoc = nasoc) {
3464		nasoc = LIST_NEXT(asoc, sctp_tcblist);
3465		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3466			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3467				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3468				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3469			}
3470			cnt++;
3471			continue;
3472		}
3473		/* Free associations that are NOT killing us */
3474		SCTP_TCB_LOCK(asoc);
3475		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3476		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3477			struct mbuf *op_err;
3478			uint32_t *ippp;
3479
3480			op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3481			    0, M_DONTWAIT, 1, MT_DATA);
3482			if (op_err) {
3483				/* Fill in the user initiated abort */
3484				struct sctp_paramhdr *ph;
3485
3486				SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) +
3487				    sizeof(uint32_t));
3488				ph = mtod(op_err, struct sctp_paramhdr *);
3489				ph->param_type = htons(
3490				    SCTP_CAUSE_USER_INITIATED_ABT);
3491				ph->param_length = htons(SCTP_BUF_LEN(op_err));
3492				ippp = (uint32_t *) (ph + 1);
3493				*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7);
3494
3495			}
3496			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3497#if defined(SCTP_PANIC_ON_ABORT)
3498			panic("inpcb_free does an abort");
3499#endif
3500			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3501			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3502		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3503			cnt++;
3504			SCTP_TCB_UNLOCK(asoc);
3505			continue;
3506		}
3507		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3508		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3509			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3510		}
3511		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3512			cnt++;
3513		}
3514	}
3515	if (cnt) {
3516		/* Ok we have someone out there that will kill us */
3517		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3518#ifdef SCTP_LOG_CLOSING
3519		sctp_log_closing(inp, NULL, 3);
3520#endif
3521		SCTP_INP_WUNLOCK(inp);
3522		SCTP_ASOC_CREATE_UNLOCK(inp);
3523		SCTP_INP_INFO_WUNLOCK();
3524		return;
3525	}
3526	if (SCTP_INP_LOCK_CONTENDED(inp))
3527		being_refed++;
3528	if (SCTP_INP_READ_CONTENDED(inp))
3529		being_refed++;
3530	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3531		being_refed++;
3532
3533	if ((inp->refcount) ||
3534	    (being_refed) ||
3535	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3536		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3537#ifdef SCTP_LOG_CLOSING
3538		sctp_log_closing(inp, NULL, 4);
3539#endif
3540		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3541		SCTP_INP_WUNLOCK(inp);
3542		SCTP_ASOC_CREATE_UNLOCK(inp);
3543		SCTP_INP_INFO_WUNLOCK();
3544		return;
3545	}
3546	inp->sctp_ep.signature_change.type = 0;
3547	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3548	/*
3549	 * Remove it from the list .. last thing we need a lock for.
3550	 */
3551	LIST_REMOVE(inp, sctp_list);
3552	SCTP_INP_WUNLOCK(inp);
3553	SCTP_ASOC_CREATE_UNLOCK(inp);
3554	SCTP_INP_INFO_WUNLOCK();
3555	/*
3556	 * Now we release all locks. Since this INP cannot be found anymore
3557	 * except possbily by the kill timer that might be running. We call
3558	 * the drain function here. It should hit the case were it sees the
3559	 * ACTIVE flag cleared and exit out freeing us to proceed and
3560	 * destroy everything.
3561	 */
3562	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3563		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3564	} else {
3565		/* Probably un-needed */
3566		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3567	}
3568
3569#ifdef SCTP_LOG_CLOSING
3570	sctp_log_closing(inp, NULL, 5);
3571#endif
3572
3573
3574	if ((inp->sctp_asocidhash) != NULL) {
3575		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3576		inp->sctp_asocidhash = NULL;
3577	}
3578	/* sa_ignore FREED_MEMORY */
3579	while ((sq = TAILQ_FIRST(&inp->read_queue)) != NULL) {
3580		/* Its only abandoned if it had data left */
3581		if (sq->length)
3582			SCTP_STAT_INCR(sctps_left_abandon);
3583
3584		TAILQ_REMOVE(&inp->read_queue, sq, next);
3585		sctp_free_remote_addr(sq->whoFrom);
3586		if (so)
3587			so->so_rcv.sb_cc -= sq->length;
3588		if (sq->data) {
3589			sctp_m_freem(sq->data);
3590			sq->data = NULL;
3591		}
3592		/*
3593		 * no need to free the net count, since at this point all
3594		 * assoc's are gone.
3595		 */
3596		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3597		SCTP_DECR_READQ_COUNT();
3598	}
3599	/* Now the sctp_pcb things */
3600	/*
3601	 * free each asoc if it is not already closed/free. we can't use the
3602	 * macro here since le_next will get freed as part of the
3603	 * sctp_free_assoc() call.
3604	 */
3605	cnt = 0;
3606	if (so) {
3607#ifdef IPSEC
3608		ipsec_delete_pcbpolicy(ip_pcb);
3609#endif				/* IPSEC */
3610
3611		/* Unlocks not needed since the socket is gone now */
3612	}
3613	if (ip_pcb->inp_options) {
3614		(void)sctp_m_free(ip_pcb->inp_options);
3615		ip_pcb->inp_options = 0;
3616	}
3617	if (ip_pcb->inp_moptions) {
3618		inp_freemoptions(ip_pcb->inp_moptions);
3619		ip_pcb->inp_moptions = 0;
3620	}
3621#ifdef INET6
3622	if (ip_pcb->inp_vflag & INP_IPV6) {
3623		struct in6pcb *in6p;
3624
3625		in6p = (struct in6pcb *)inp;
3626		ip6_freepcbopts(in6p->in6p_outputopts);
3627	}
3628#endif				/* INET6 */
3629	ip_pcb->inp_vflag = 0;
3630	/* free up authentication fields */
3631	if (inp->sctp_ep.local_auth_chunks != NULL)
3632		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3633	if (inp->sctp_ep.local_hmacs != NULL)
3634		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3635
3636	shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
3637	while (shared_key) {
3638		LIST_REMOVE(shared_key, next);
3639		sctp_free_sharedkey(shared_key);
3640		/* sa_ignore FREED_MEMORY */
3641		shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
3642	}
3643
3644	/*
3645	 * if we have an address list the following will free the list of
3646	 * ifaddr's that are set into this ep. Again macro limitations here,
3647	 * since the LIST_FOREACH could be a bad idea.
3648	 */
3649	for ((laddr = LIST_FIRST(&inp->sctp_addr_list)); laddr != NULL;
3650	    laddr = nladdr) {
3651		nladdr = LIST_NEXT(laddr, sctp_nxt_addr);
3652		sctp_remove_laddr(laddr);
3653	}
3654
3655#ifdef SCTP_TRACK_FREED_ASOCS
3656	/* TEMP CODE */
3657	for ((asoc = LIST_FIRST(&inp->sctp_asoc_free_list)); asoc != NULL;
3658	    asoc = nasoc) {
3659		nasoc = LIST_NEXT(asoc, sctp_tcblist);
3660		LIST_REMOVE(asoc, sctp_tcblist);
3661		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3662		SCTP_DECR_ASOC_COUNT();
3663	}
3664	/* *** END TEMP CODE *** */
3665#endif
3666	/* Now lets see about freeing the EP hash table. */
3667	if (inp->sctp_tcbhash != NULL) {
3668		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3669		inp->sctp_tcbhash = NULL;
3670	}
3671	/* Now we must put the ep memory back into the zone pool */
3672	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3673	SCTP_INP_LOCK_DESTROY(inp);
3674	SCTP_INP_READ_DESTROY(inp);
3675	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3676	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3677	SCTP_DECR_EP_COUNT();
3678}
3679
3680
3681struct sctp_nets *
3682sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3683{
3684	struct sctp_nets *net;
3685
3686	/* locate the address */
3687	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3688		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3689			return (net);
3690	}
3691	return (NULL);
3692}
3693
3694
3695int
3696sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3697{
3698	struct sctp_ifa *sctp_ifa;
3699
3700	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3701	if (sctp_ifa) {
3702		return (1);
3703	} else {
3704		return (0);
3705	}
3706}
3707
3708/*
3709 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3710 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3711 * stats of stuff.
3712 */
3713int
3714sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3715    int set_scope, int from)
3716{
3717	/*
3718	 * The following is redundant to the same lines in the
3719	 * sctp_aloc_assoc() but is needed since other's call the add
3720	 * address function
3721	 */
3722	struct sctp_nets *net, *netfirst;
3723	int addr_inscope;
3724
3725	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3726	    from);
3727	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3728
3729	netfirst = sctp_findnet(stcb, newaddr);
3730	if (netfirst) {
3731		/*
3732		 * Lie and return ok, we don't want to make the association
3733		 * go away for this behavior. It will happen in the TCP
3734		 * model in a connected socket. It does not reach the hash
3735		 * table until after the association is built so it can't be
3736		 * found. Mark as reachable, since the initial creation will
3737		 * have been cleared and the NOT_IN_ASSOC flag will have
3738		 * been added... and we don't want to end up removing it
3739		 * back out.
3740		 */
3741		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3742			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3743			    SCTP_ADDR_UNCONFIRMED);
3744		} else {
3745			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3746		}
3747
3748		return (0);
3749	}
3750	addr_inscope = 1;
3751	if (newaddr->sa_family == AF_INET) {
3752		struct sockaddr_in *sin;
3753
3754		sin = (struct sockaddr_in *)newaddr;
3755		if (sin->sin_addr.s_addr == 0) {
3756			/* Invalid address */
3757			return (-1);
3758		}
3759		/* zero out the bzero area */
3760		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3761
3762		/* assure len is set */
3763		sin->sin_len = sizeof(struct sockaddr_in);
3764		if (set_scope) {
3765#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3766			stcb->ipv4_local_scope = 1;
3767#else
3768			if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3769				stcb->asoc.ipv4_local_scope = 1;
3770			}
3771#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3772		} else {
3773			/* Validate the address is in scope */
3774			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3775			    (stcb->asoc.ipv4_local_scope == 0)) {
3776				addr_inscope = 0;
3777			}
3778		}
3779#ifdef INET6
3780	} else if (newaddr->sa_family == AF_INET6) {
3781		struct sockaddr_in6 *sin6;
3782
3783		sin6 = (struct sockaddr_in6 *)newaddr;
3784		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3785			/* Invalid address */
3786			return (-1);
3787		}
3788		/* assure len is set */
3789		sin6->sin6_len = sizeof(struct sockaddr_in6);
3790		if (set_scope) {
3791			if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3792				stcb->asoc.loopback_scope = 1;
3793				stcb->asoc.local_scope = 0;
3794				stcb->asoc.ipv4_local_scope = 1;
3795				stcb->asoc.site_scope = 1;
3796			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3797				/*
3798				 * If the new destination is a LINK_LOCAL we
3799				 * must have common site scope. Don't set
3800				 * the local scope since we may not share
3801				 * all links, only loopback can do this.
3802				 * Links on the local network would also be
3803				 * on our private network for v4 too.
3804				 */
3805				stcb->asoc.ipv4_local_scope = 1;
3806				stcb->asoc.site_scope = 1;
3807			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3808				/*
3809				 * If the new destination is SITE_LOCAL then
3810				 * we must have site scope in common.
3811				 */
3812				stcb->asoc.site_scope = 1;
3813			}
3814		} else {
3815			/* Validate the address is in scope */
3816			if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3817			    (stcb->asoc.loopback_scope == 0)) {
3818				addr_inscope = 0;
3819			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3820			    (stcb->asoc.local_scope == 0)) {
3821				addr_inscope = 0;
3822			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3823			    (stcb->asoc.site_scope == 0)) {
3824				addr_inscope = 0;
3825			}
3826		}
3827#endif
3828	} else {
3829		/* not supported family type */
3830		return (-1);
3831	}
3832	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3833	if (net == NULL) {
3834		return (-1);
3835	}
3836	SCTP_INCR_RADDR_COUNT();
3837	bzero(net, sizeof(*net));
3838	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3839	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3840	if (newaddr->sa_family == AF_INET) {
3841		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3842	} else if (newaddr->sa_family == AF_INET6) {
3843		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3844	}
3845	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3846	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3847		stcb->asoc.loopback_scope = 1;
3848		stcb->asoc.ipv4_local_scope = 1;
3849		stcb->asoc.local_scope = 0;
3850		stcb->asoc.site_scope = 1;
3851		addr_inscope = 1;
3852	}
3853	net->failure_threshold = stcb->asoc.def_net_failure;
3854	if (addr_inscope == 0) {
3855		net->dest_state = (SCTP_ADDR_REACHABLE |
3856		    SCTP_ADDR_OUT_OF_SCOPE);
3857	} else {
3858		if (from == SCTP_ADDR_IS_CONFIRMED)
3859			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3860			net->dest_state = SCTP_ADDR_REACHABLE;
3861		else
3862			net->dest_state = SCTP_ADDR_REACHABLE |
3863			    SCTP_ADDR_UNCONFIRMED;
3864	}
3865	/*
3866	 * We set this to 0, the timer code knows that this means its an
3867	 * initial value
3868	 */
3869	net->RTO = 0;
3870	net->RTO_measured = 0;
3871	stcb->asoc.numnets++;
3872	*(&net->ref_count) = 1;
3873	net->tos_flowlabel = 0;
3874	if (SCTP_BASE_SYSCTL(sctp_udp_tunneling_for_client_enable)) {
3875		net->port = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port));
3876	} else {
3877		net->port = 0;
3878	}
3879#ifdef INET
3880	if (newaddr->sa_family == AF_INET)
3881		net->tos_flowlabel = stcb->asoc.default_tos;
3882#endif
3883#ifdef INET6
3884	if (newaddr->sa_family == AF_INET6)
3885		net->tos_flowlabel = stcb->asoc.default_flowlabel;
3886#endif
3887	/* Init the timer structure */
3888	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3889	SCTP_OS_TIMER_INIT(&net->fr_timer.timer);
3890	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3891
3892	/* Now generate a route for this guy */
3893#ifdef INET6
3894	/* KAME hack: embed scopeid */
3895	if (newaddr->sa_family == AF_INET6) {
3896		struct sockaddr_in6 *sin6;
3897
3898		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3899		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3900		sin6->sin6_scope_id = 0;
3901	}
3902#endif
3903	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3904
3905	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3906		/* Get source address */
3907		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3908		    stcb,
3909		    (sctp_route_t *) & net->ro,
3910		    net,
3911		    0,
3912		    stcb->asoc.vrf_id);
3913		/* Now get the interface MTU */
3914		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3915			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3916		} else {
3917			net->mtu = 0;
3918		}
3919#ifdef SCTP_PRINT_FOR_B_AND_M
3920		SCTP_PRINTF("We have found an interface mtu of %d\n", net->mtu);
3921#endif
3922		if (net->mtu == 0) {
3923			/* Huh ?? */
3924			net->mtu = SCTP_DEFAULT_MTU;
3925		} else {
3926			uint32_t rmtu;
3927
3928			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3929#ifdef SCTP_PRINT_FOR_B_AND_M
3930			SCTP_PRINTF("The route mtu is %d\n", rmtu);
3931#endif
3932			if (rmtu == 0) {
3933				/*
3934				 * Start things off to match mtu of
3935				 * interface please.
3936				 */
3937				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3938				    net->ro.ro_rt, net->mtu);
3939			} else {
3940				/*
3941				 * we take the route mtu over the interface,
3942				 * since the route may be leading out the
3943				 * loopback, or a different interface.
3944				 */
3945				net->mtu = rmtu;
3946			}
3947		}
3948		if (from == SCTP_ALLOC_ASOC) {
3949#ifdef SCTP_PRINT_FOR_B_AND_M
3950			SCTP_PRINTF("New assoc sets mtu to :%d\n", net->mtu);
3951#endif
3952			stcb->asoc.smallest_mtu = net->mtu;
3953		}
3954	} else {
3955		net->mtu = stcb->asoc.smallest_mtu;
3956	}
3957#ifdef INET6
3958	if (newaddr->sa_family == AF_INET6) {
3959		struct sockaddr_in6 *sin6;
3960
3961		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3962		(void)sa6_recoverscope(sin6);
3963	}
3964#endif
3965	if (net->port) {
3966		net->mtu -= sizeof(struct udphdr);
3967	}
3968	if (stcb->asoc.smallest_mtu > net->mtu) {
3969#ifdef SCTP_PRINT_FOR_B_AND_M
3970		SCTP_PRINTF("new address mtu:%d smaller than smallest:%d\n",
3971		    net->mtu, stcb->asoc.smallest_mtu);
3972#endif
3973		stcb->asoc.smallest_mtu = net->mtu;
3974	}
3975	/* JRS - Use the congestion control given in the CC module */
3976	stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net);
3977
3978	/*
3979	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3980	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
3981	 */
3982	net->find_pseudo_cumack = 1;
3983	net->find_rtx_pseudo_cumack = 1;
3984	net->src_addr_selected = 0;
3985	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
3986	if (net->ro.ro_rt == NULL) {
3987		/* Since we have no route put it at the back */
3988		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3989	} else if (netfirst == NULL) {
3990		/* We are the first one in the pool. */
3991		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3992	} else if (netfirst->ro.ro_rt == NULL) {
3993		/*
3994		 * First one has NO route. Place this one ahead of the first
3995		 * one.
3996		 */
3997		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3998	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
3999		/*
4000		 * This one has a different interface than the one at the
4001		 * top of the list. Place it ahead.
4002		 */
4003		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
4004	} else {
4005		/*
4006		 * Ok we have the same interface as the first one. Move
4007		 * forward until we find either a) one with a NULL route...
4008		 * insert ahead of that b) one with a different ifp.. insert
4009		 * after that. c) end of the list.. insert at the tail.
4010		 */
4011		struct sctp_nets *netlook;
4012
4013		do {
4014			netlook = TAILQ_NEXT(netfirst, sctp_next);
4015			if (netlook == NULL) {
4016				/* End of the list */
4017				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4018				break;
4019			} else if (netlook->ro.ro_rt == NULL) {
4020				/* next one has NO route */
4021				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4022				break;
4023			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4024				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4025				    net, sctp_next);
4026				break;
4027			}
4028			/* Shift forward */
4029			netfirst = netlook;
4030		} while (netlook != NULL);
4031	}
4032
4033	/* got to have a primary set */
4034	if (stcb->asoc.primary_destination == 0) {
4035		stcb->asoc.primary_destination = net;
4036	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4037		    (net->ro.ro_rt) &&
4038	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4039		/* No route to current primary adopt new primary */
4040		stcb->asoc.primary_destination = net;
4041	}
4042	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, stcb->sctp_ep, stcb,
4043	    net);
4044	/* Validate primary is first */
4045	net = TAILQ_FIRST(&stcb->asoc.nets);
4046	if ((net != stcb->asoc.primary_destination) &&
4047	    (stcb->asoc.primary_destination)) {
4048		/*
4049		 * first one on the list is NOT the primary sctp_cmpaddr()
4050		 * is much more efficent if the primary is the first on the
4051		 * list, make it so.
4052		 */
4053		TAILQ_REMOVE(&stcb->asoc.nets,
4054		    stcb->asoc.primary_destination, sctp_next);
4055		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4056		    stcb->asoc.primary_destination, sctp_next);
4057	}
4058	return (0);
4059}
4060
4061
4062static uint32_t
4063sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4064{
4065	uint32_t id;
4066	struct sctpasochead *head;
4067	struct sctp_tcb *lstcb;
4068
4069	SCTP_INP_WLOCK(inp);
4070try_again:
4071	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4072		/* TSNH */
4073		SCTP_INP_WUNLOCK(inp);
4074		return (0);
4075	}
4076	/*
4077	 * We don't allow assoc id to be 0, this is needed otherwise if the
4078	 * id were to wrap we would have issues with some socket options.
4079	 */
4080	if (inp->sctp_associd_counter == 0) {
4081		inp->sctp_associd_counter++;
4082	}
4083	id = inp->sctp_associd_counter;
4084	inp->sctp_associd_counter++;
4085	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4086	if (lstcb) {
4087		goto try_again;
4088	}
4089	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4090	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4091	stcb->asoc.in_asocid_hash = 1;
4092	SCTP_INP_WUNLOCK(inp);
4093	return id;
4094}
4095
4096/*
4097 * allocate an association and add it to the endpoint. The caller must be
4098 * careful to add all additional addresses once they are know right away or
4099 * else the assoc will be may experience a blackout scenario.
4100 */
4101struct sctp_tcb *
4102sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4103    int *error, uint32_t override_tag, uint32_t vrf_id,
4104    struct thread *p
4105)
4106{
4107	/* note the p argument is only valid in unbound sockets */
4108
4109	struct sctp_tcb *stcb;
4110	struct sctp_association *asoc;
4111	struct sctpasochead *head;
4112	uint16_t rport;
4113	int err;
4114
4115	/*
4116	 * Assumption made here: Caller has done a
4117	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4118	 * address does not exist already.
4119	 */
4120	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4121		/* Hit max assoc, sorry no more */
4122		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4123		*error = ENOBUFS;
4124		return (NULL);
4125	}
4126	if (firstaddr == NULL) {
4127		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4128		*error = EINVAL;
4129		return (NULL);
4130	}
4131	SCTP_INP_RLOCK(inp);
4132	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4133	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4134	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4135		/*
4136		 * If its in the TCP pool, its NOT allowed to create an
4137		 * association. The parent listener needs to call
4138		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4139		 * off, or connected one does this.. its an error.
4140		 */
4141		SCTP_INP_RUNLOCK(inp);
4142		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4143		*error = EINVAL;
4144		return (NULL);
4145	}
4146	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4147#ifdef SCTP_DEBUG
4148	if (firstaddr) {
4149		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4150		SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4151		    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4152	} else {
4153		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4154	}
4155#endif				/* SCTP_DEBUG */
4156	if (firstaddr->sa_family == AF_INET) {
4157		struct sockaddr_in *sin;
4158
4159		sin = (struct sockaddr_in *)firstaddr;
4160		if ((sin->sin_port == 0) || (sin->sin_addr.s_addr == 0)) {
4161			/* Invalid address */
4162			SCTP_INP_RUNLOCK(inp);
4163			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4164			*error = EINVAL;
4165			return (NULL);
4166		}
4167		rport = sin->sin_port;
4168	} else if (firstaddr->sa_family == AF_INET6) {
4169		struct sockaddr_in6 *sin6;
4170
4171		sin6 = (struct sockaddr_in6 *)firstaddr;
4172		if ((sin6->sin6_port == 0) ||
4173		    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
4174			/* Invalid address */
4175			SCTP_INP_RUNLOCK(inp);
4176			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4177			*error = EINVAL;
4178			return (NULL);
4179		}
4180		rport = sin6->sin6_port;
4181	} else {
4182		/* not supported family type */
4183		SCTP_INP_RUNLOCK(inp);
4184		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4185		*error = EINVAL;
4186		return (NULL);
4187	}
4188	SCTP_INP_RUNLOCK(inp);
4189	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4190		/*
4191		 * If you have not performed a bind, then we need to do the
4192		 * ephemerial bind for you.
4193		 */
4194		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4195		    (struct sockaddr *)NULL,
4196		    (struct sctp_ifa *)NULL,
4197		    p
4198		    ))) {
4199			/* bind error, probably perm */
4200			*error = err;
4201			return (NULL);
4202		}
4203	}
4204	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4205	if (stcb == NULL) {
4206		/* out of memory? */
4207		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4208		*error = ENOMEM;
4209		return (NULL);
4210	}
4211	SCTP_INCR_ASOC_COUNT();
4212
4213	bzero(stcb, sizeof(*stcb));
4214	asoc = &stcb->asoc;
4215
4216	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4217	SCTP_TCB_LOCK_INIT(stcb);
4218	SCTP_TCB_SEND_LOCK_INIT(stcb);
4219	stcb->rport = rport;
4220	/* setup back pointer's */
4221	stcb->sctp_ep = inp;
4222	stcb->sctp_socket = inp->sctp_socket;
4223	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4224		/* failed */
4225		SCTP_TCB_LOCK_DESTROY(stcb);
4226		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4227		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4228		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4229		SCTP_DECR_ASOC_COUNT();
4230		*error = err;
4231		return (NULL);
4232	}
4233	/* and the port */
4234	SCTP_INP_INFO_WLOCK();
4235	SCTP_INP_WLOCK(inp);
4236	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4237		/* inpcb freed while alloc going on */
4238		SCTP_TCB_LOCK_DESTROY(stcb);
4239		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4240		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4241		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4242		SCTP_INP_WUNLOCK(inp);
4243		SCTP_INP_INFO_WUNLOCK();
4244		SCTP_DECR_ASOC_COUNT();
4245		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4246		*error = EINVAL;
4247		return (NULL);
4248	}
4249	SCTP_TCB_LOCK(stcb);
4250
4251	/* now that my_vtag is set, add it to the hash */
4252	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4253	/* put it in the bucket in the vtag hash of assoc's for the system */
4254	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4255	SCTP_INP_INFO_WUNLOCK();
4256
4257	if ((err = sctp_add_remote_addr(stcb, firstaddr, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4258		/* failure.. memory error? */
4259		if (asoc->strmout) {
4260			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4261			asoc->strmout = NULL;
4262		}
4263		if (asoc->mapping_array) {
4264			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4265			asoc->mapping_array = NULL;
4266		}
4267		if (asoc->nr_mapping_array) {
4268			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4269			asoc->nr_mapping_array = NULL;
4270		}
4271		SCTP_DECR_ASOC_COUNT();
4272		SCTP_TCB_LOCK_DESTROY(stcb);
4273		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4274		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4275		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4276		SCTP_INP_WUNLOCK(inp);
4277		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4278		*error = ENOBUFS;
4279		return (NULL);
4280	}
4281	/* Init all the timers */
4282	SCTP_OS_TIMER_INIT(&asoc->hb_timer.timer);
4283	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4284	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4285	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4286	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4287	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4288	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4289	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4290
4291	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4292	/* now file the port under the hash as well */
4293	if (inp->sctp_tcbhash != NULL) {
4294		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4295		    inp->sctp_hashmark)];
4296		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4297	}
4298	SCTP_INP_WUNLOCK(inp);
4299	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", stcb);
4300	return (stcb);
4301}
4302
4303
4304void
4305sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4306{
4307	struct sctp_association *asoc;
4308
4309	asoc = &stcb->asoc;
4310	asoc->numnets--;
4311	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4312	if (net == asoc->primary_destination) {
4313		/* Reset primary */
4314		struct sctp_nets *lnet;
4315
4316		lnet = TAILQ_FIRST(&asoc->nets);
4317		/*
4318		 * Mobility adaptation Ideally, if deleted destination is
4319		 * the primary, it becomes a fast retransmission trigger by
4320		 * the subsequent SET PRIMARY. (by micchie)
4321		 */
4322		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4323		    SCTP_MOBILITY_BASE) ||
4324		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4325		    SCTP_MOBILITY_FASTHANDOFF)) {
4326			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4327			if (asoc->deleted_primary != NULL) {
4328				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4329				goto out;
4330			}
4331			asoc->deleted_primary = net;
4332			atomic_add_int(&net->ref_count, 1);
4333			memset(&net->lastsa, 0, sizeof(net->lastsa));
4334			memset(&net->lastsv, 0, sizeof(net->lastsv));
4335			sctp_mobility_feature_on(stcb->sctp_ep,
4336			    SCTP_MOBILITY_PRIM_DELETED);
4337			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4338			    stcb->sctp_ep, stcb, NULL);
4339		}
4340out:
4341		/* Try to find a confirmed primary */
4342		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4343	}
4344	if (net == asoc->last_data_chunk_from) {
4345		/* Reset primary */
4346		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4347	}
4348	if (net == asoc->last_control_chunk_from) {
4349		/* Clear net */
4350		asoc->last_control_chunk_from = NULL;
4351	}
4352	sctp_free_remote_addr(net);
4353}
4354
4355/*
4356 * remove a remote endpoint address from an association, it will fail if the
4357 * address does not exist.
4358 */
4359int
4360sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4361{
4362	/*
4363	 * Here we need to remove a remote address. This is quite simple, we
4364	 * first find it in the list of address for the association
4365	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4366	 * on that item. Note we do not allow it to be removed if there are
4367	 * no other addresses.
4368	 */
4369	struct sctp_association *asoc;
4370	struct sctp_nets *net, *net_tmp;
4371
4372	asoc = &stcb->asoc;
4373
4374	/* locate the address */
4375	for (net = TAILQ_FIRST(&asoc->nets); net != NULL; net = net_tmp) {
4376		net_tmp = TAILQ_NEXT(net, sctp_next);
4377		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4378			continue;
4379		}
4380		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4381		    remaddr)) {
4382			/* we found the guy */
4383			if (asoc->numnets < 2) {
4384				/* Must have at LEAST two remote addresses */
4385				return (-1);
4386			} else {
4387				sctp_remove_net(stcb, net);
4388				return (0);
4389			}
4390		}
4391	}
4392	/* not found. */
4393	return (-2);
4394}
4395
4396void
4397sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4398{
4399	struct sctpvtaghead *chain;
4400	struct sctp_tagblock *twait_block;
4401	int found = 0;
4402	int i;
4403
4404	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4405	if (!LIST_EMPTY(chain)) {
4406		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4407			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4408				if ((twait_block->vtag_block[i].v_tag == tag) &&
4409				    (twait_block->vtag_block[i].lport == lport) &&
4410				    (twait_block->vtag_block[i].rport == rport)) {
4411					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4412					twait_block->vtag_block[i].v_tag = 0;
4413					twait_block->vtag_block[i].lport = 0;
4414					twait_block->vtag_block[i].rport = 0;
4415					found = 1;
4416					break;
4417				}
4418			}
4419			if (found)
4420				break;
4421		}
4422	}
4423}
4424
4425int
4426sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4427{
4428	struct sctpvtaghead *chain;
4429	struct sctp_tagblock *twait_block;
4430	int found = 0;
4431	int i;
4432
4433	SCTP_INP_INFO_WLOCK();
4434	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4435	if (!LIST_EMPTY(chain)) {
4436		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4437			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4438				if ((twait_block->vtag_block[i].v_tag == tag) &&
4439				    (twait_block->vtag_block[i].lport == lport) &&
4440				    (twait_block->vtag_block[i].rport == rport)) {
4441					found = 1;
4442					break;
4443				}
4444			}
4445			if (found)
4446				break;
4447		}
4448	}
4449	SCTP_INP_INFO_WUNLOCK();
4450	return (found);
4451}
4452
4453
4454void
4455sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4456{
4457	struct sctpvtaghead *chain;
4458	struct sctp_tagblock *twait_block;
4459	struct timeval now;
4460	int set, i;
4461
4462	if (time == 0) {
4463		/* Its disabled */
4464		return;
4465	}
4466	(void)SCTP_GETTIME_TIMEVAL(&now);
4467	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4468	set = 0;
4469	if (!LIST_EMPTY(chain)) {
4470		/* Block(s) present, lets find space, and expire on the fly */
4471		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4472			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4473				if ((twait_block->vtag_block[i].v_tag == 0) &&
4474				    !set) {
4475					twait_block->vtag_block[i].tv_sec_at_expire =
4476					    now.tv_sec + time;
4477					twait_block->vtag_block[i].v_tag = tag;
4478					twait_block->vtag_block[i].lport = lport;
4479					twait_block->vtag_block[i].rport = rport;
4480					set = 1;
4481				} else if ((twait_block->vtag_block[i].v_tag) &&
4482				    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4483					/* Audit expires this guy */
4484					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4485					twait_block->vtag_block[i].v_tag = 0;
4486					twait_block->vtag_block[i].lport = 0;
4487					twait_block->vtag_block[i].rport = 0;
4488					if (set == 0) {
4489						/* Reuse it for my new tag */
4490						twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4491						twait_block->vtag_block[i].v_tag = tag;
4492						twait_block->vtag_block[i].lport = lport;
4493						twait_block->vtag_block[i].rport = rport;
4494						set = 1;
4495					}
4496				}
4497			}
4498			if (set) {
4499				/*
4500				 * We only do up to the block where we can
4501				 * place our tag for audits
4502				 */
4503				break;
4504			}
4505		}
4506	}
4507	/* Need to add a new block to chain */
4508	if (!set) {
4509		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4510		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4511		if (twait_block == NULL) {
4512#ifdef INVARIANTS
4513			panic("Can not alloc tagblock");
4514#endif
4515			return;
4516		}
4517		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4518		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4519		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4520		twait_block->vtag_block[0].v_tag = tag;
4521		twait_block->vtag_block[0].lport = lport;
4522		twait_block->vtag_block[0].rport = rport;
4523	}
4524}
4525
4526
4527
4528/*-
4529 * Free the association after un-hashing the remote port. This
4530 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4531 * expect that the input to this function IS a locked TCB.
4532 * It will return 0, if it did NOT destroy the association (instead
4533 * it unlocks it. It will return NON-zero if it either destroyed the
4534 * association OR the association is already destroyed.
4535 */
4536int
4537sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4538{
4539	int i;
4540	struct sctp_association *asoc;
4541	struct sctp_nets *net, *prev;
4542	struct sctp_laddr *laddr;
4543	struct sctp_tmit_chunk *chk;
4544	struct sctp_asconf_addr *aparam;
4545	struct sctp_asconf_ack *aack;
4546	struct sctp_stream_reset_list *liste;
4547	struct sctp_queued_to_read *sq;
4548	struct sctp_stream_queue_pending *sp;
4549	sctp_sharedkey_t *shared_key;
4550	struct socket *so;
4551	int ccnt = 0;
4552	int cnt = 0;
4553
4554	/* first, lets purge the entry from the hash table. */
4555
4556#ifdef SCTP_LOG_CLOSING
4557	sctp_log_closing(inp, stcb, 6);
4558#endif
4559	if (stcb->asoc.state == 0) {
4560#ifdef SCTP_LOG_CLOSING
4561		sctp_log_closing(inp, NULL, 7);
4562#endif
4563		/* there is no asoc, really TSNH :-0 */
4564		return (1);
4565	}
4566	/* TEMP CODE */
4567	if (stcb->freed_from_where == 0) {
4568		/* Only record the first place free happened from */
4569		stcb->freed_from_where = from_location;
4570	}
4571	/* TEMP CODE */
4572
4573	asoc = &stcb->asoc;
4574	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4575	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4576		/* nothing around */
4577		so = NULL;
4578	else
4579		so = inp->sctp_socket;
4580
4581	/*
4582	 * We used timer based freeing if a reader or writer is in the way.
4583	 * So we first check if we are actually being called from a timer,
4584	 * if so we abort early if a reader or writer is still in the way.
4585	 */
4586	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4587	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4588		/*
4589		 * is it the timer driving us? if so are the reader/writers
4590		 * gone?
4591		 */
4592		if (stcb->asoc.refcnt) {
4593			/* nope, reader or writer in the way */
4594			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4595			/* no asoc destroyed */
4596			SCTP_TCB_UNLOCK(stcb);
4597#ifdef SCTP_LOG_CLOSING
4598			sctp_log_closing(inp, stcb, 8);
4599#endif
4600			return (0);
4601		}
4602	}
4603	/* now clean up any other timers */
4604	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
4605	asoc->hb_timer.self = NULL;
4606	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4607	asoc->dack_timer.self = NULL;
4608	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4609	/*-
4610	 * For stream reset we don't blast this unless
4611	 * it is a str-reset timer, it might be the
4612	 * free-asoc timer which we DON'T want to
4613	 * disturb.
4614	 */
4615	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4616		asoc->strreset_timer.self = NULL;
4617	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4618	asoc->asconf_timer.self = NULL;
4619	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4620	asoc->autoclose_timer.self = NULL;
4621	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4622	asoc->shut_guard_timer.self = NULL;
4623	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4624	asoc->delayed_event_timer.self = NULL;
4625	/* Mobility adaptation */
4626	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4627	asoc->delete_prim_timer.self = NULL;
4628	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4629		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
4630		net->fr_timer.self = NULL;
4631		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4632		net->rxt_timer.self = NULL;
4633		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4634		net->pmtu_timer.self = NULL;
4635	}
4636	/* Now the read queue needs to be cleaned up (only once) */
4637	cnt = 0;
4638	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4639		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4640		SCTP_INP_READ_LOCK(inp);
4641		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4642			if (sq->stcb == stcb) {
4643				sq->do_not_ref_stcb = 1;
4644				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4645				/*
4646				 * If there is no end, there never will be
4647				 * now.
4648				 */
4649				if (sq->end_added == 0) {
4650					/* Held for PD-API clear that. */
4651					sq->pdapi_aborted = 1;
4652					sq->held_length = 0;
4653					if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4654						/*
4655						 * Need to add a PD-API
4656						 * aborted indication.
4657						 * Setting the control_pdapi
4658						 * assures that it will be
4659						 * added right after this
4660						 * msg.
4661						 */
4662						uint32_t strseq;
4663
4664						stcb->asoc.control_pdapi = sq;
4665						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4666						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4667						    stcb,
4668						    SCTP_PARTIAL_DELIVERY_ABORTED,
4669						    (void *)&strseq,
4670						    SCTP_SO_LOCKED);
4671						stcb->asoc.control_pdapi = NULL;
4672					}
4673				}
4674				/* Add an end to wake them */
4675				sq->end_added = 1;
4676				cnt++;
4677			}
4678		}
4679		SCTP_INP_READ_UNLOCK(inp);
4680		if (stcb->block_entry) {
4681			cnt++;
4682			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4683			stcb->block_entry->error = ECONNRESET;
4684			stcb->block_entry = NULL;
4685		}
4686	}
4687	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4688		/*
4689		 * Someone holds a reference OR the socket is unaccepted
4690		 * yet.
4691		 */
4692		if ((stcb->asoc.refcnt) ||
4693		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4694		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4695			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4696			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4697		}
4698		SCTP_TCB_UNLOCK(stcb);
4699		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4700		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4701			/* nothing around */
4702			so = NULL;
4703		if (so) {
4704			/* Wake any reader/writers */
4705			sctp_sorwakeup(inp, so);
4706			sctp_sowwakeup(inp, so);
4707		}
4708#ifdef SCTP_LOG_CLOSING
4709		sctp_log_closing(inp, stcb, 9);
4710#endif
4711		/* no asoc destroyed */
4712		return (0);
4713	}
4714#ifdef SCTP_LOG_CLOSING
4715	sctp_log_closing(inp, stcb, 10);
4716#endif
4717	/*
4718	 * When I reach here, no others want to kill the assoc yet.. and I
4719	 * own the lock. Now its possible an abort comes in when I do the
4720	 * lock exchange below to grab all the locks to do the final take
4721	 * out. to prevent this we increment the count, which will start a
4722	 * timer and blow out above thus assuring us that we hold exclusive
4723	 * killing of the asoc. Note that after getting back the TCB lock we
4724	 * will go ahead and increment the counter back up and stop any
4725	 * timer a passing stranger may have started :-S
4726	 */
4727	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4728		atomic_add_int(&stcb->asoc.refcnt, 1);
4729
4730		SCTP_TCB_UNLOCK(stcb);
4731		SCTP_INP_INFO_WLOCK();
4732		SCTP_INP_WLOCK(inp);
4733		SCTP_TCB_LOCK(stcb);
4734	}
4735	/* Double check the GONE flag */
4736	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4737	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4738		/* nothing around */
4739		so = NULL;
4740
4741	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4742	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4743		/*
4744		 * For TCP type we need special handling when we are
4745		 * connected. We also include the peel'ed off ones to.
4746		 */
4747		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4748			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4749			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4750			if (so) {
4751				SOCK_LOCK(so);
4752				if (so->so_rcv.sb_cc == 0) {
4753					so->so_state &= ~(SS_ISCONNECTING |
4754					    SS_ISDISCONNECTING |
4755					    SS_ISCONFIRMING |
4756					    SS_ISCONNECTED);
4757				}
4758				socantrcvmore_locked(so);
4759				sctp_sowwakeup(inp, so);
4760				sctp_sorwakeup(inp, so);
4761				SCTP_SOWAKEUP(so);
4762			}
4763		}
4764	}
4765	/*
4766	 * Make it invalid too, that way if its about to run it will abort
4767	 * and return.
4768	 */
4769	/* re-increment the lock */
4770	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4771		atomic_add_int(&stcb->asoc.refcnt, -1);
4772	}
4773	if (stcb->asoc.refcnt) {
4774		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4775		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4776		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4777			SCTP_INP_INFO_WUNLOCK();
4778			SCTP_INP_WUNLOCK(inp);
4779		}
4780		SCTP_TCB_UNLOCK(stcb);
4781		return (0);
4782	}
4783	asoc->state = 0;
4784	if (inp->sctp_tcbhash) {
4785		LIST_REMOVE(stcb, sctp_tcbhash);
4786	}
4787	if (stcb->asoc.in_asocid_hash) {
4788		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4789	}
4790	/* Now lets remove it from the list of ALL associations in the EP */
4791	LIST_REMOVE(stcb, sctp_tcblist);
4792	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4793		SCTP_INP_INCR_REF(inp);
4794		SCTP_INP_WUNLOCK(inp);
4795	}
4796	/* pull from vtag hash */
4797	LIST_REMOVE(stcb, sctp_asocs);
4798	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4799	    inp->sctp_lport, stcb->rport);
4800
4801	/*
4802	 * Now restop the timers to be sure this is paranoia at is finest!
4803	 */
4804	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4805	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
4806	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4807	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4808	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4809	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4810	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4811	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4812	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4813		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
4814		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4815		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4816	}
4817
4818	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4819	prev = NULL;
4820	/*
4821	 * The chunk lists and such SHOULD be empty but we check them just
4822	 * in case.
4823	 */
4824	/* anything on the wheel needs to be removed */
4825	for (i = 0; i < asoc->streamoutcnt; i++) {
4826		struct sctp_stream_out *outs;
4827
4828		outs = &asoc->strmout[i];
4829		/* now clean up any chunks here */
4830		sp = TAILQ_FIRST(&outs->outqueue);
4831		while (sp) {
4832			TAILQ_REMOVE(&outs->outqueue, sp, next);
4833			if (sp->data) {
4834				if (so) {
4835					/* Still an open socket - report */
4836					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4837					    SCTP_NOTIFY_DATAGRAM_UNSENT,
4838					    (void *)sp, 0);
4839				}
4840				if (sp->data) {
4841					sctp_m_freem(sp->data);
4842					sp->data = NULL;
4843					sp->tail_mbuf = NULL;
4844				}
4845			}
4846			sctp_free_remote_addr(sp->net);
4847			sctp_free_spbufspace(stcb, asoc, sp);
4848			if (sp->holds_key_ref)
4849				sctp_auth_key_release(stcb, sp->auth_keyid);
4850			/* Free the zone stuff  */
4851			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_strmoq), sp);
4852			SCTP_DECR_STRMOQ_COUNT();
4853			/* sa_ignore FREED_MEMORY */
4854			sp = TAILQ_FIRST(&outs->outqueue);
4855		}
4856	}
4857
4858	/* sa_ignore FREED_MEMORY */
4859	while ((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) {
4860		TAILQ_REMOVE(&asoc->resetHead, liste, next_resp);
4861		SCTP_FREE(liste, SCTP_M_STRESET);
4862	}
4863
4864	sq = TAILQ_FIRST(&asoc->pending_reply_queue);
4865	while (sq) {
4866		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4867		if (sq->data) {
4868			sctp_m_freem(sq->data);
4869			sq->data = NULL;
4870		}
4871		sctp_free_remote_addr(sq->whoFrom);
4872		sq->whoFrom = NULL;
4873		sq->stcb = NULL;
4874		/* Free the ctl entry */
4875		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4876		SCTP_DECR_READQ_COUNT();
4877		/* sa_ignore FREED_MEMORY */
4878		sq = TAILQ_FIRST(&asoc->pending_reply_queue);
4879	}
4880
4881	chk = TAILQ_FIRST(&asoc->free_chunks);
4882	while (chk) {
4883		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4884		if (chk->data) {
4885			sctp_m_freem(chk->data);
4886			chk->data = NULL;
4887		}
4888		if (chk->holds_key_ref)
4889			sctp_auth_key_release(stcb, chk->auth_keyid);
4890		ccnt++;
4891		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4892		SCTP_DECR_CHK_COUNT();
4893		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4894		asoc->free_chunk_cnt--;
4895		/* sa_ignore FREED_MEMORY */
4896		chk = TAILQ_FIRST(&asoc->free_chunks);
4897	}
4898	/* pending send queue SHOULD be empty */
4899	if (!TAILQ_EMPTY(&asoc->send_queue)) {
4900		chk = TAILQ_FIRST(&asoc->send_queue);
4901		while (chk) {
4902			TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4903			if (chk->data) {
4904				if (so) {
4905					/* Still a socket? */
4906					sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb,
4907					    SCTP_NOTIFY_DATAGRAM_UNSENT, chk, 0);
4908				}
4909				if (chk->data) {
4910					sctp_m_freem(chk->data);
4911					chk->data = NULL;
4912				}
4913			}
4914			if (chk->holds_key_ref)
4915				sctp_auth_key_release(stcb, chk->auth_keyid);
4916			ccnt++;
4917			sctp_free_remote_addr(chk->whoTo);
4918			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4919			SCTP_DECR_CHK_COUNT();
4920			/* sa_ignore FREED_MEMORY */
4921			chk = TAILQ_FIRST(&asoc->send_queue);
4922		}
4923	}
4924/*
4925  if (ccnt) {
4926  printf("Freed %d from send_queue\n", ccnt);
4927  ccnt = 0;
4928  }
4929*/
4930	/* sent queue SHOULD be empty */
4931	if (!TAILQ_EMPTY(&asoc->sent_queue)) {
4932		chk = TAILQ_FIRST(&asoc->sent_queue);
4933		while (chk) {
4934			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4935			if (chk->data) {
4936				if (so) {
4937					/* Still a socket? */
4938					sctp_ulp_notify(SCTP_NOTIFY_DG_FAIL, stcb,
4939					    SCTP_NOTIFY_DATAGRAM_SENT, chk, 0);
4940				}
4941				if (chk->data) {
4942					sctp_m_freem(chk->data);
4943					chk->data = NULL;
4944				}
4945			}
4946			if (chk->holds_key_ref)
4947				sctp_auth_key_release(stcb, chk->auth_keyid);
4948			ccnt++;
4949			sctp_free_remote_addr(chk->whoTo);
4950			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4951			SCTP_DECR_CHK_COUNT();
4952			/* sa_ignore FREED_MEMORY */
4953			chk = TAILQ_FIRST(&asoc->sent_queue);
4954		}
4955	}
4956/*
4957  if (ccnt) {
4958  printf("Freed %d from sent_queue\n", ccnt);
4959  ccnt = 0;
4960  }
4961*/
4962	/* control queue MAY not be empty */
4963	if (!TAILQ_EMPTY(&asoc->control_send_queue)) {
4964		chk = TAILQ_FIRST(&asoc->control_send_queue);
4965		while (chk) {
4966			TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
4967			if (chk->data) {
4968				sctp_m_freem(chk->data);
4969				chk->data = NULL;
4970			}
4971			if (chk->holds_key_ref)
4972				sctp_auth_key_release(stcb, chk->auth_keyid);
4973			ccnt++;
4974			sctp_free_remote_addr(chk->whoTo);
4975			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4976			SCTP_DECR_CHK_COUNT();
4977			/* sa_ignore FREED_MEMORY */
4978			chk = TAILQ_FIRST(&asoc->control_send_queue);
4979		}
4980	}
4981/*
4982  if (ccnt) {
4983  printf("Freed %d from ctrl_queue\n", ccnt);
4984  ccnt = 0;
4985  }
4986*/
4987
4988	/* ASCONF queue MAY not be empty */
4989	if (!TAILQ_EMPTY(&asoc->asconf_send_queue)) {
4990		chk = TAILQ_FIRST(&asoc->asconf_send_queue);
4991		while (chk) {
4992			TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
4993			if (chk->data) {
4994				sctp_m_freem(chk->data);
4995				chk->data = NULL;
4996			}
4997			if (chk->holds_key_ref)
4998				sctp_auth_key_release(stcb, chk->auth_keyid);
4999			ccnt++;
5000			sctp_free_remote_addr(chk->whoTo);
5001			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5002			SCTP_DECR_CHK_COUNT();
5003			/* sa_ignore FREED_MEMORY */
5004			chk = TAILQ_FIRST(&asoc->asconf_send_queue);
5005		}
5006	}
5007/*
5008  if (ccnt) {
5009  printf("Freed %d from asconf_queue\n", ccnt);
5010  ccnt = 0;
5011  }
5012*/
5013	if (!TAILQ_EMPTY(&asoc->reasmqueue)) {
5014		chk = TAILQ_FIRST(&asoc->reasmqueue);
5015		while (chk) {
5016			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5017			if (chk->data) {
5018				sctp_m_freem(chk->data);
5019				chk->data = NULL;
5020			}
5021			if (chk->holds_key_ref)
5022				sctp_auth_key_release(stcb, chk->auth_keyid);
5023			sctp_free_remote_addr(chk->whoTo);
5024			ccnt++;
5025			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5026			SCTP_DECR_CHK_COUNT();
5027			/* sa_ignore FREED_MEMORY */
5028			chk = TAILQ_FIRST(&asoc->reasmqueue);
5029		}
5030	}
5031/*
5032  if (ccnt) {
5033  printf("Freed %d from reasm_queue\n", ccnt);
5034  ccnt = 0;
5035  }
5036*/
5037	if (asoc->mapping_array) {
5038		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5039		asoc->mapping_array = NULL;
5040	}
5041	if (asoc->nr_mapping_array) {
5042		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5043		asoc->nr_mapping_array = NULL;
5044	}
5045	/* the stream outs */
5046	if (asoc->strmout) {
5047		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5048		asoc->strmout = NULL;
5049	}
5050	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5051	if (asoc->strmin) {
5052		struct sctp_queued_to_read *ctl;
5053
5054		for (i = 0; i < asoc->streamincnt; i++) {
5055			if (!TAILQ_EMPTY(&asoc->strmin[i].inqueue)) {
5056				/* We have somethings on the streamin queue */
5057				ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
5058				while (ctl) {
5059					TAILQ_REMOVE(&asoc->strmin[i].inqueue,
5060					    ctl, next);
5061					sctp_free_remote_addr(ctl->whoFrom);
5062					if (ctl->data) {
5063						sctp_m_freem(ctl->data);
5064						ctl->data = NULL;
5065					}
5066					/*
5067					 * We don't free the address here
5068					 * since all the net's were freed
5069					 * above.
5070					 */
5071					SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5072					SCTP_DECR_READQ_COUNT();
5073					ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
5074				}
5075			}
5076		}
5077		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5078		asoc->strmin = NULL;
5079	}
5080	asoc->streamincnt = 0;
5081	while (!TAILQ_EMPTY(&asoc->nets)) {
5082		/* sa_ignore FREED_MEMORY */
5083		net = TAILQ_FIRST(&asoc->nets);
5084		/* pull from list */
5085		if ((SCTP_BASE_INFO(ipi_count_raddr) == 0) || (prev == net)) {
5086#ifdef INVARIANTS
5087			panic("no net's left alloc'ed, or list points to itself");
5088#endif
5089			break;
5090		}
5091		prev = net;
5092		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5093		sctp_free_remote_addr(net);
5094	}
5095
5096	while (!LIST_EMPTY(&asoc->sctp_restricted_addrs)) {
5097		/* sa_ignore FREED_MEMORY */
5098		laddr = LIST_FIRST(&asoc->sctp_restricted_addrs);
5099		sctp_remove_laddr(laddr);
5100	}
5101
5102	/* pending asconf (address) parameters */
5103	while (!TAILQ_EMPTY(&asoc->asconf_queue)) {
5104		/* sa_ignore FREED_MEMORY */
5105		aparam = TAILQ_FIRST(&asoc->asconf_queue);
5106		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5107		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5108	}
5109	while (!TAILQ_EMPTY(&asoc->asconf_ack_sent)) {
5110		/* sa_ignore FREED_MEMORY */
5111		aack = TAILQ_FIRST(&asoc->asconf_ack_sent);
5112		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5113		if (aack->data != NULL) {
5114			sctp_m_freem(aack->data);
5115		}
5116		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5117	}
5118	/* clean up auth stuff */
5119	if (asoc->local_hmacs)
5120		sctp_free_hmaclist(asoc->local_hmacs);
5121	if (asoc->peer_hmacs)
5122		sctp_free_hmaclist(asoc->peer_hmacs);
5123
5124	if (asoc->local_auth_chunks)
5125		sctp_free_chunklist(asoc->local_auth_chunks);
5126	if (asoc->peer_auth_chunks)
5127		sctp_free_chunklist(asoc->peer_auth_chunks);
5128
5129	sctp_free_authinfo(&asoc->authinfo);
5130
5131	shared_key = LIST_FIRST(&asoc->shared_keys);
5132	while (shared_key) {
5133		LIST_REMOVE(shared_key, next);
5134		sctp_free_sharedkey(shared_key);
5135		/* sa_ignore FREED_MEMORY */
5136		shared_key = LIST_FIRST(&asoc->shared_keys);
5137	}
5138
5139	/* Insert new items here :> */
5140
5141	/* Get rid of LOCK */
5142	SCTP_TCB_LOCK_DESTROY(stcb);
5143	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5144	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5145		SCTP_INP_INFO_WUNLOCK();
5146		SCTP_INP_RLOCK(inp);
5147	}
5148#ifdef SCTP_TRACK_FREED_ASOCS
5149	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5150		/* now clean up the tasoc itself */
5151		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5152		SCTP_DECR_ASOC_COUNT();
5153	} else {
5154		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5155	}
5156#else
5157	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5158	SCTP_DECR_ASOC_COUNT();
5159#endif
5160	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5161		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5162			/*
5163			 * If its NOT the inp_free calling us AND sctp_close
5164			 * as been called, we call back...
5165			 */
5166			SCTP_INP_RUNLOCK(inp);
5167			/*
5168			 * This will start the kill timer (if we are the
5169			 * lastone) since we hold an increment yet. But this
5170			 * is the only safe way to do this since otherwise
5171			 * if the socket closes at the same time we are here
5172			 * we might collide in the cleanup.
5173			 */
5174			sctp_inpcb_free(inp,
5175			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5176			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5177			SCTP_INP_DECR_REF(inp);
5178			goto out_of;
5179		} else {
5180			/* The socket is still open. */
5181			SCTP_INP_DECR_REF(inp);
5182		}
5183	}
5184	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5185		SCTP_INP_RUNLOCK(inp);
5186	}
5187out_of:
5188	/* destroyed the asoc */
5189#ifdef SCTP_LOG_CLOSING
5190	sctp_log_closing(inp, NULL, 11);
5191#endif
5192	return (1);
5193}
5194
5195
5196
5197/*
5198 * determine if a destination is "reachable" based upon the addresses bound
5199 * to the current endpoint (e.g. only v4 or v6 currently bound)
5200 */
5201/*
5202 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5203 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5204 * types bound as its endpoint
5205 */
5206int
5207sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5208{
5209	struct sctp_inpcb *inp;
5210	int answer;
5211
5212	/*
5213	 * No locks here, the TCB, in all cases is already locked and an
5214	 * assoc is up. There is either a INP lock by the caller applied (in
5215	 * asconf case when deleting an address) or NOT in the HB case,
5216	 * however if HB then the INP increment is up and the INP will not
5217	 * be removed (on top of the fact that we have a TCB lock). So we
5218	 * only want to read the sctp_flags, which is either bound-all or
5219	 * not.. no protection needed since once an assoc is up you can't be
5220	 * changing your binding.
5221	 */
5222	inp = stcb->sctp_ep;
5223	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5224		/* if bound all, destination is not restricted */
5225		/*
5226		 * RRS: Question during lock work: Is this correct? If you
5227		 * are bound-all you still might need to obey the V4--V6
5228		 * flags??? IMO this bound-all stuff needs to be removed!
5229		 */
5230		return (1);
5231	}
5232	/* NOTE: all "scope" checks are done when local addresses are added */
5233	if (destaddr->sa_family == AF_INET6) {
5234		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5235	} else if (destaddr->sa_family == AF_INET) {
5236		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5237	} else {
5238		/* invalid family, so it's unreachable */
5239		answer = 0;
5240	}
5241	return (answer);
5242}
5243
5244/*
5245 * update the inp_vflags on an endpoint
5246 */
5247static void
5248sctp_update_ep_vflag(struct sctp_inpcb *inp)
5249{
5250	struct sctp_laddr *laddr;
5251
5252	/* first clear the flag */
5253	inp->ip_inp.inp.inp_vflag = 0;
5254	/* set the flag based on addresses on the ep list */
5255	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5256		if (laddr->ifa == NULL) {
5257			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5258			    __FUNCTION__);
5259			continue;
5260		}
5261		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5262			continue;
5263		}
5264		if (laddr->ifa->address.sa.sa_family == AF_INET6) {
5265			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5266		} else if (laddr->ifa->address.sa.sa_family == AF_INET) {
5267			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5268		}
5269	}
5270}
5271
5272/*
5273 * Add the address to the endpoint local address list There is nothing to be
5274 * done if we are bound to all addresses
5275 */
5276void
5277sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5278{
5279	struct sctp_laddr *laddr;
5280	int fnd, error = 0;
5281
5282	fnd = 0;
5283
5284	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5285		/* You are already bound to all. You have it already */
5286		return;
5287	}
5288	if (ifa->address.sa.sa_family == AF_INET6) {
5289		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5290			/* Can't bind a non-useable addr. */
5291			return;
5292		}
5293	}
5294	/* first, is it already present? */
5295	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5296		if (laddr->ifa == ifa) {
5297			fnd = 1;
5298			break;
5299		}
5300	}
5301
5302	if (fnd == 0) {
5303		/* Not in the ep list */
5304		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5305		if (error != 0)
5306			return;
5307		inp->laddr_count++;
5308		/* update inp_vflag flags */
5309		if (ifa->address.sa.sa_family == AF_INET6) {
5310			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5311		} else if (ifa->address.sa.sa_family == AF_INET) {
5312			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5313		}
5314	}
5315	return;
5316}
5317
5318
5319/*
5320 * select a new (hopefully reachable) destination net (should only be used
5321 * when we deleted an ep addr that is the only usable source address to reach
5322 * the destination net)
5323 */
5324static void
5325sctp_select_primary_destination(struct sctp_tcb *stcb)
5326{
5327	struct sctp_nets *net;
5328
5329	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5330		/* for now, we'll just pick the first reachable one we find */
5331		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5332			continue;
5333		if (sctp_destination_is_reachable(stcb,
5334		    (struct sockaddr *)&net->ro._l_addr)) {
5335			/* found a reachable destination */
5336			stcb->asoc.primary_destination = net;
5337		}
5338	}
5339	/* I can't there from here! ...we're gonna die shortly... */
5340}
5341
5342
5343/*
5344 * Delete the address from the endpoint local address list There is nothing
5345 * to be done if we are bound to all addresses
5346 */
5347void
5348sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5349{
5350	struct sctp_laddr *laddr;
5351	int fnd;
5352
5353	fnd = 0;
5354	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5355		/* You are already bound to all. You have it already */
5356		return;
5357	}
5358	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5359		if (laddr->ifa == ifa) {
5360			fnd = 1;
5361			break;
5362		}
5363	}
5364	if (fnd && (inp->laddr_count < 2)) {
5365		/* can't delete unless there are at LEAST 2 addresses */
5366		return;
5367	}
5368	if (fnd) {
5369		/*
5370		 * clean up any use of this address go through our
5371		 * associations and clear any last_used_address that match
5372		 * this one for each assoc, see if a new primary_destination
5373		 * is needed
5374		 */
5375		struct sctp_tcb *stcb;
5376
5377		/* clean up "next_addr_touse" */
5378		if (inp->next_addr_touse == laddr)
5379			/* delete this address */
5380			inp->next_addr_touse = NULL;
5381
5382		/* clean up "last_used_address" */
5383		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5384			struct sctp_nets *net;
5385
5386			SCTP_TCB_LOCK(stcb);
5387			if (stcb->asoc.last_used_address == laddr)
5388				/* delete this address */
5389				stcb->asoc.last_used_address = NULL;
5390			/*
5391			 * Now spin through all the nets and purge any ref
5392			 * to laddr
5393			 */
5394			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5395				if (net->ro._s_addr &&
5396				    (net->ro._s_addr->ifa == laddr->ifa)) {
5397					/* Yep, purge src address selected */
5398					sctp_rtentry_t *rt;
5399
5400					/* delete this address if cached */
5401					rt = net->ro.ro_rt;
5402					if (rt != NULL) {
5403						RTFREE(rt);
5404						net->ro.ro_rt = NULL;
5405					}
5406					sctp_free_ifa(net->ro._s_addr);
5407					net->ro._s_addr = NULL;
5408					net->src_addr_selected = 0;
5409				}
5410			}
5411			SCTP_TCB_UNLOCK(stcb);
5412		}		/* for each tcb */
5413		/* remove it from the ep list */
5414		sctp_remove_laddr(laddr);
5415		inp->laddr_count--;
5416		/* update inp_vflag flags */
5417		sctp_update_ep_vflag(inp);
5418	}
5419	return;
5420}
5421
5422/*
5423 * Add the address to the TCB local address restricted list.
5424 * This is a "pending" address list (eg. addresses waiting for an
5425 * ASCONF-ACK response) and cannot be used as a valid source address.
5426 */
5427void
5428sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5429{
5430	struct sctp_inpcb *inp;
5431	struct sctp_laddr *laddr;
5432	struct sctpladdr *list;
5433
5434	/*
5435	 * Assumes TCB is locked.. and possibly the INP. May need to
5436	 * confirm/fix that if we need it and is not the case.
5437	 */
5438	list = &stcb->asoc.sctp_restricted_addrs;
5439
5440	inp = stcb->sctp_ep;
5441	if (ifa->address.sa.sa_family == AF_INET6) {
5442		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5443			/* Can't bind a non-existent addr. */
5444			return;
5445		}
5446	}
5447	/* does the address already exist? */
5448	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5449		if (laddr->ifa == ifa) {
5450			return;
5451		}
5452	}
5453
5454	/* add to the list */
5455	(void)sctp_insert_laddr(list, ifa, 0);
5456	return;
5457}
5458
5459/*
5460 * insert an laddr entry with the given ifa for the desired list
5461 */
5462int
5463sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5464{
5465	struct sctp_laddr *laddr;
5466
5467	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5468	if (laddr == NULL) {
5469		/* out of memory? */
5470		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5471		return (EINVAL);
5472	}
5473	SCTP_INCR_LADDR_COUNT();
5474	bzero(laddr, sizeof(*laddr));
5475	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5476	laddr->ifa = ifa;
5477	laddr->action = act;
5478	atomic_add_int(&ifa->refcount, 1);
5479	/* insert it */
5480	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5481
5482	return (0);
5483}
5484
5485/*
5486 * Remove an laddr entry from the local address list (on an assoc)
5487 */
5488void
5489sctp_remove_laddr(struct sctp_laddr *laddr)
5490{
5491
5492	/* remove from the list */
5493	LIST_REMOVE(laddr, sctp_nxt_addr);
5494	sctp_free_ifa(laddr->ifa);
5495	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5496	SCTP_DECR_LADDR_COUNT();
5497}
5498
5499/*
5500 * Remove a local address from the TCB local address restricted list
5501 */
5502void
5503sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5504{
5505	struct sctp_inpcb *inp;
5506	struct sctp_laddr *laddr;
5507
5508	/*
5509	 * This is called by asconf work. It is assumed that a) The TCB is
5510	 * locked and b) The INP is locked. This is true in as much as I can
5511	 * trace through the entry asconf code where I did these locks.
5512	 * Again, the ASCONF code is a bit different in that it does lock
5513	 * the INP during its work often times. This must be since we don't
5514	 * want other proc's looking up things while what they are looking
5515	 * up is changing :-D
5516	 */
5517
5518	inp = stcb->sctp_ep;
5519	/* if subset bound and don't allow ASCONF's, can't delete last */
5520	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5521	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5522		if (stcb->sctp_ep->laddr_count < 2) {
5523			/* can't delete last address */
5524			return;
5525		}
5526	}
5527	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5528		/* remove the address if it exists */
5529		if (laddr->ifa == NULL)
5530			continue;
5531		if (laddr->ifa == ifa) {
5532			sctp_remove_laddr(laddr);
5533			return;
5534		}
5535	}
5536
5537	/* address not found! */
5538	return;
5539}
5540
5541/*
5542 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5543 */
5544/* sysctl */
5545static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5546static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5547
5548void
5549sctp_pcb_init()
5550{
5551	/*
5552	 * SCTP initialization for the PCB structures should be called by
5553	 * the sctp_init() funciton.
5554	 */
5555	int i;
5556	struct timeval tv;
5557
5558	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5559		/* error I was called twice */
5560		return;
5561	}
5562	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5563
5564#if defined(SCTP_LOCAL_TRACE_BUF)
5565	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5566#endif
5567	(void)SCTP_GETTIME_TIMEVAL(&tv);
5568#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5569	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5570	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5571#else
5572	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5573	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5574#endif
5575	/* init the empty list of (All) Endpoints */
5576	LIST_INIT(&SCTP_BASE_INFO(listhead));
5577
5578
5579	/* init the hash table of endpoints */
5580	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5581	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5582	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5583	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5584	    &SCTP_BASE_INFO(hashasocmark));
5585	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5586	    &SCTP_BASE_INFO(hashmark));
5587	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5588	    &SCTP_BASE_INFO(hashtcpmark));
5589	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5590
5591
5592	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5593	    &SCTP_BASE_INFO(hashvrfmark));
5594
5595	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5596	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5597	/* init the zones */
5598	/*
5599	 * FIX ME: Should check for NULL returns, but if it does fail we are
5600	 * doomed to panic anyways... add later maybe.
5601	 */
5602	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5603	    sizeof(struct sctp_inpcb), maxsockets);
5604
5605	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5606	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5607
5608	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5609	    sizeof(struct sctp_laddr),
5610	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5611
5612	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5613	    sizeof(struct sctp_nets),
5614	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5615
5616	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5617	    sizeof(struct sctp_tmit_chunk),
5618	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5619
5620	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5621	    sizeof(struct sctp_queued_to_read),
5622	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5623
5624	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5625	    sizeof(struct sctp_stream_queue_pending),
5626	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5627
5628	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5629	    sizeof(struct sctp_asconf),
5630	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5631
5632	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5633	    sizeof(struct sctp_asconf_ack),
5634	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5635
5636
5637	/* Master Lock INIT for info structure */
5638	SCTP_INP_INFO_LOCK_INIT();
5639	SCTP_STATLOG_INIT_LOCK();
5640
5641	SCTP_IPI_COUNT_INIT();
5642	SCTP_IPI_ADDR_INIT();
5643#ifdef SCTP_PACKET_LOGGING
5644	SCTP_IP_PKTLOG_INIT();
5645#endif
5646	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5647
5648	SCTP_WQ_ADDR_INIT();
5649	/* not sure if we need all the counts */
5650	SCTP_BASE_INFO(ipi_count_ep) = 0;
5651	/* assoc/tcb zone info */
5652	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5653	/* local addrlist zone info */
5654	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5655	/* remote addrlist zone info */
5656	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5657	/* chunk info */
5658	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5659
5660	/* socket queue zone info */
5661	SCTP_BASE_INFO(ipi_count_readq) = 0;
5662
5663	/* stream out queue cont */
5664	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5665
5666	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5667	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5668
5669	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5670
5671	/* Init the TIMEWAIT list */
5672	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5673		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5674	}
5675
5676	sctp_startup_iterator();
5677
5678	/*
5679	 * INIT the default VRF which for BSD is the only one, other O/S's
5680	 * may have more. But initially they must start with one and then
5681	 * add the VRF's as addresses are added.
5682	 */
5683	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5684
5685}
5686
5687/*
5688 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5689 */
5690void
5691sctp_pcb_finish(void)
5692{
5693	struct sctp_vrflist *vrf_bucket;
5694	struct sctp_vrf *vrf;
5695	struct sctp_ifn *ifn;
5696	struct sctp_ifa *ifa;
5697	struct sctpvtaghead *chain;
5698	struct sctp_tagblock *twait_block, *prev_twait_block;
5699	struct sctp_laddr *wi;
5700	int i;
5701
5702	/*
5703	 * Free BSD the it thread never exits but we do clean up. The only
5704	 * way freebsd reaches here if we have VRF's but we still add the
5705	 * ifdef to make it compile on old versions.
5706	 */
5707	{
5708		struct sctp_iterator *it, *nit;
5709
5710		SCTP_IPI_ITERATOR_WQ_LOCK();
5711		it = TAILQ_FIRST(&sctp_it_ctl.iteratorhead);
5712		while (it) {
5713			nit = TAILQ_NEXT(it, sctp_nxt_itr);
5714			if (it->vn != curvnet) {
5715				it = nit;
5716				continue;
5717			}
5718			TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
5719			    it, sctp_nxt_itr);
5720			if (it->function_atend != NULL) {
5721				(*it->function_atend) (it->pointer, it->val);
5722			}
5723			SCTP_FREE(it, SCTP_M_ITER);
5724			it = nit;
5725		}
5726		SCTP_IPI_ITERATOR_WQ_UNLOCK();
5727		SCTP_ITERATOR_LOCK();
5728		if ((sctp_it_ctl.cur_it) &&
5729		    (sctp_it_ctl.cur_it->vn == curvnet)) {
5730			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5731		}
5732		SCTP_ITERATOR_UNLOCK();
5733	}
5734
5735	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5736	SCTP_WQ_ADDR_LOCK();
5737	while ((wi = LIST_FIRST(&SCTP_BASE_INFO(addr_wq))) != NULL) {
5738		LIST_REMOVE(wi, sctp_nxt_addr);
5739		SCTP_DECR_LADDR_COUNT();
5740		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5741	}
5742	SCTP_WQ_ADDR_UNLOCK();
5743
5744	/*
5745	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5746	 * destroyed first).
5747	 */
5748	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5749	while ((vrf = LIST_FIRST(vrf_bucket)) != NULL) {
5750		while ((ifn = LIST_FIRST(&vrf->ifnlist)) != NULL) {
5751			while ((ifa = LIST_FIRST(&ifn->ifalist)) != NULL) {
5752				/* free the ifa */
5753				LIST_REMOVE(ifa, next_bucket);
5754				LIST_REMOVE(ifa, next_ifa);
5755				SCTP_FREE(ifa, SCTP_M_IFA);
5756			}
5757			/* free the ifn */
5758			LIST_REMOVE(ifn, next_bucket);
5759			LIST_REMOVE(ifn, next_ifn);
5760			SCTP_FREE(ifn, SCTP_M_IFN);
5761		}
5762		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5763		/* free the vrf */
5764		LIST_REMOVE(vrf, next_vrf);
5765		SCTP_FREE(vrf, SCTP_M_VRF);
5766	}
5767	/* free the vrf hashes */
5768	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5769	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5770
5771	/*
5772	 * free the TIMEWAIT list elements malloc'd in the function
5773	 * sctp_add_vtag_to_timewait()...
5774	 */
5775	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5776		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5777		if (!LIST_EMPTY(chain)) {
5778			prev_twait_block = NULL;
5779			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5780				if (prev_twait_block) {
5781					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5782				}
5783				prev_twait_block = twait_block;
5784			}
5785			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5786		}
5787	}
5788
5789	/* free the locks and mutexes */
5790#ifdef SCTP_PACKET_LOGGING
5791	SCTP_IP_PKTLOG_DESTROY();
5792#endif
5793	SCTP_IPI_ADDR_DESTROY();
5794	SCTP_STATLOG_DESTROY();
5795	SCTP_INP_INFO_LOCK_DESTROY();
5796
5797	SCTP_WQ_ADDR_DESTROY();
5798
5799	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
5800	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
5801	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
5802	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
5803	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
5804	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
5805	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
5806	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
5807	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
5808	/* Get rid of other stuff to */
5809	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
5810		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
5811	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
5812		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
5813	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
5814		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
5815
5816}
5817
5818
5819int
5820sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
5821    int iphlen, int offset, int limit, struct sctphdr *sh,
5822    struct sockaddr *altsa)
5823{
5824	/*
5825	 * grub through the INIT pulling addresses and loading them to the
5826	 * nets structure in the asoc. The from address in the mbuf should
5827	 * also be loaded (if it is not already). This routine can be called
5828	 * with either INIT or INIT-ACK's as long as the m points to the IP
5829	 * packet and the offset points to the beginning of the parameters.
5830	 */
5831	struct sctp_inpcb *inp, *l_inp;
5832	struct sctp_nets *net, *net_tmp;
5833	struct ip *iph;
5834	struct sctp_paramhdr *phdr, parm_buf;
5835	struct sctp_tcb *stcb_tmp;
5836	uint16_t ptype, plen;
5837	struct sockaddr *sa;
5838	struct sockaddr_storage dest_store;
5839	struct sockaddr *local_sa = (struct sockaddr *)&dest_store;
5840	struct sockaddr_in sin;
5841	struct sockaddr_in6 sin6;
5842	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
5843	struct sctp_auth_random *p_random = NULL;
5844	uint16_t random_len = 0;
5845	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
5846	struct sctp_auth_hmac_algo *hmacs = NULL;
5847	uint16_t hmacs_len = 0;
5848	uint8_t saw_asconf = 0;
5849	uint8_t saw_asconf_ack = 0;
5850	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
5851	struct sctp_auth_chunk_list *chunks = NULL;
5852	uint16_t num_chunks = 0;
5853	sctp_key_t *new_key;
5854	uint32_t keylen;
5855	int got_random = 0, got_hmacs = 0, got_chklist = 0;
5856
5857	/* First get the destination address setup too. */
5858	memset(&sin, 0, sizeof(sin));
5859	memset(&sin6, 0, sizeof(sin6));
5860
5861	sin.sin_family = AF_INET;
5862	sin.sin_len = sizeof(sin);
5863	sin.sin_port = stcb->rport;
5864
5865	sin6.sin6_family = AF_INET6;
5866	sin6.sin6_len = sizeof(struct sockaddr_in6);
5867	sin6.sin6_port = stcb->rport;
5868	if (altsa == NULL) {
5869		iph = mtod(m, struct ip *);
5870		switch (iph->ip_v) {
5871		case IPVERSION:
5872			{
5873				/* its IPv4 */
5874				struct sockaddr_in *sin_2;
5875
5876				sin_2 = (struct sockaddr_in *)(local_sa);
5877				memset(sin_2, 0, sizeof(sin));
5878				sin_2->sin_family = AF_INET;
5879				sin_2->sin_len = sizeof(sin);
5880				sin_2->sin_port = sh->dest_port;
5881				sin_2->sin_addr.s_addr = iph->ip_dst.s_addr;
5882				sin.sin_addr = iph->ip_src;
5883				sa = (struct sockaddr *)&sin;
5884				break;
5885			}
5886#ifdef INET6
5887		case IPV6_VERSION >> 4:
5888			{
5889				/* its IPv6 */
5890				struct ip6_hdr *ip6;
5891				struct sockaddr_in6 *sin6_2;
5892
5893				ip6 = mtod(m, struct ip6_hdr *);
5894				sin6_2 = (struct sockaddr_in6 *)(local_sa);
5895				memset(sin6_2, 0, sizeof(sin6));
5896				sin6_2->sin6_family = AF_INET6;
5897				sin6_2->sin6_len = sizeof(struct sockaddr_in6);
5898				sin6_2->sin6_port = sh->dest_port;
5899				sin6.sin6_addr = ip6->ip6_src;
5900				sa = (struct sockaddr *)&sin6;
5901				break;
5902			}
5903#endif
5904		default:
5905			return (-1);
5906			break;
5907		}
5908	} else {
5909		/*
5910		 * For cookies we use the src address NOT from the packet
5911		 * but from the original INIT
5912		 */
5913		sa = altsa;
5914	}
5915	/* Turn off ECN until we get through all params */
5916	stcb->asoc.ecn_allowed = 0;
5917	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5918		/* mark all addresses that we have currently on the list */
5919		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
5920	}
5921	/* does the source address already exist? if so skip it */
5922	l_inp = inp = stcb->sctp_ep;
5923
5924	atomic_add_int(&stcb->asoc.refcnt, 1);
5925	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, local_sa, stcb);
5926	atomic_add_int(&stcb->asoc.refcnt, -1);
5927
5928	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
5929		/* we must add the source address */
5930		/* no scope set here since we have a tcb already. */
5931		if ((sa->sa_family == AF_INET) &&
5932		    (stcb->asoc.ipv4_addr_legal)) {
5933			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
5934				return (-1);
5935			}
5936		} else if ((sa->sa_family == AF_INET6) &&
5937		    (stcb->asoc.ipv6_addr_legal)) {
5938			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
5939				return (-2);
5940			}
5941		}
5942	} else {
5943		if (net_tmp != NULL && stcb_tmp == stcb) {
5944			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
5945		} else if (stcb_tmp != stcb) {
5946			/* It belongs to another association? */
5947			if (stcb_tmp)
5948				SCTP_TCB_UNLOCK(stcb_tmp);
5949			return (-3);
5950		}
5951	}
5952	if (stcb->asoc.state == 0) {
5953		/* the assoc was freed? */
5954		return (-4);
5955	}
5956	/*
5957	 * peer must explicitly turn this on. This may have been initialized
5958	 * to be "on" in order to allow local addr changes while INIT's are
5959	 * in flight.
5960	 */
5961	stcb->asoc.peer_supports_asconf = 0;
5962	/* now we must go through each of the params. */
5963	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
5964	while (phdr) {
5965		ptype = ntohs(phdr->param_type);
5966		plen = ntohs(phdr->param_length);
5967		/*
5968		 * printf("ptype => %0x, plen => %d\n", (uint32_t)ptype,
5969		 * (int)plen);
5970		 */
5971		if (offset + plen > limit) {
5972			break;
5973		}
5974		if (plen == 0) {
5975			break;
5976		}
5977		if (ptype == SCTP_IPV4_ADDRESS) {
5978			if (stcb->asoc.ipv4_addr_legal) {
5979				struct sctp_ipv4addr_param *p4, p4_buf;
5980
5981				/* ok get the v4 address and check/add */
5982				phdr = sctp_get_next_param(m, offset,
5983				    (struct sctp_paramhdr *)&p4_buf,
5984				    sizeof(p4_buf));
5985				if (plen != sizeof(struct sctp_ipv4addr_param) ||
5986				    phdr == NULL) {
5987					return (-5);
5988				}
5989				p4 = (struct sctp_ipv4addr_param *)phdr;
5990				sin.sin_addr.s_addr = p4->addr;
5991				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
5992					/* Skip multi-cast addresses */
5993					goto next_param;
5994				}
5995				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
5996				    (sin.sin_addr.s_addr == INADDR_ANY)) {
5997					goto next_param;
5998				}
5999				sa = (struct sockaddr *)&sin;
6000				inp = stcb->sctp_ep;
6001				atomic_add_int(&stcb->asoc.refcnt, 1);
6002				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6003				    local_sa, stcb);
6004				atomic_add_int(&stcb->asoc.refcnt, -1);
6005
6006				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6007				    inp == NULL) {
6008					/* we must add the source address */
6009					/*
6010					 * no scope set since we have a tcb
6011					 * already
6012					 */
6013
6014					/*
6015					 * we must validate the state again
6016					 * here
6017					 */
6018			add_it_now:
6019					if (stcb->asoc.state == 0) {
6020						/* the assoc was freed? */
6021						return (-7);
6022					}
6023					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6024						return (-8);
6025					}
6026				} else if (stcb_tmp == stcb) {
6027					if (stcb->asoc.state == 0) {
6028						/* the assoc was freed? */
6029						return (-10);
6030					}
6031					if (net != NULL) {
6032						/* clear flag */
6033						net->dest_state &=
6034						    ~SCTP_ADDR_NOT_IN_ASSOC;
6035					}
6036				} else {
6037					/*
6038					 * strange, address is in another
6039					 * assoc? straighten out locks.
6040					 */
6041					if (stcb_tmp) {
6042						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6043							/*
6044							 * in setup state we
6045							 * abort this guy
6046							 */
6047							sctp_abort_an_association(stcb_tmp->sctp_ep,
6048							    stcb_tmp, 1, NULL, 0);
6049							goto add_it_now;
6050						}
6051						SCTP_TCB_UNLOCK(stcb_tmp);
6052					}
6053					if (stcb->asoc.state == 0) {
6054						/* the assoc was freed? */
6055						return (-12);
6056					}
6057					return (-13);
6058				}
6059			}
6060		} else if (ptype == SCTP_IPV6_ADDRESS) {
6061			if (stcb->asoc.ipv6_addr_legal) {
6062				/* ok get the v6 address and check/add */
6063				struct sctp_ipv6addr_param *p6, p6_buf;
6064
6065				phdr = sctp_get_next_param(m, offset,
6066				    (struct sctp_paramhdr *)&p6_buf,
6067				    sizeof(p6_buf));
6068				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6069				    phdr == NULL) {
6070					return (-14);
6071				}
6072				p6 = (struct sctp_ipv6addr_param *)phdr;
6073				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6074				    sizeof(p6->addr));
6075				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6076					/* Skip multi-cast addresses */
6077					goto next_param;
6078				}
6079				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6080					/*
6081					 * Link local make no sense without
6082					 * scope
6083					 */
6084					goto next_param;
6085				}
6086				sa = (struct sockaddr *)&sin6;
6087				inp = stcb->sctp_ep;
6088				atomic_add_int(&stcb->asoc.refcnt, 1);
6089				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6090				    local_sa, stcb);
6091				atomic_add_int(&stcb->asoc.refcnt, -1);
6092				if (stcb_tmp == NULL &&
6093				    (inp == stcb->sctp_ep || inp == NULL)) {
6094					/*
6095					 * we must validate the state again
6096					 * here
6097					 */
6098			add_it_now6:
6099					if (stcb->asoc.state == 0) {
6100						/* the assoc was freed? */
6101						return (-16);
6102					}
6103					/*
6104					 * we must add the address, no scope
6105					 * set
6106					 */
6107					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6108						return (-17);
6109					}
6110				} else if (stcb_tmp == stcb) {
6111					/*
6112					 * we must validate the state again
6113					 * here
6114					 */
6115					if (stcb->asoc.state == 0) {
6116						/* the assoc was freed? */
6117						return (-19);
6118					}
6119					if (net != NULL) {
6120						/* clear flag */
6121						net->dest_state &=
6122						    ~SCTP_ADDR_NOT_IN_ASSOC;
6123					}
6124				} else {
6125					/*
6126					 * strange, address is in another
6127					 * assoc? straighten out locks.
6128					 */
6129					if (stcb_tmp)
6130						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6131							/*
6132							 * in setup state we
6133							 * abort this guy
6134							 */
6135							sctp_abort_an_association(stcb_tmp->sctp_ep,
6136							    stcb_tmp, 1, NULL, 0);
6137							goto add_it_now6;
6138						}
6139					SCTP_TCB_UNLOCK(stcb_tmp);
6140
6141					if (stcb->asoc.state == 0) {
6142						/* the assoc was freed? */
6143						return (-21);
6144					}
6145					return (-22);
6146				}
6147			}
6148		} else if (ptype == SCTP_ECN_CAPABLE) {
6149			stcb->asoc.ecn_allowed = 1;
6150		} else if (ptype == SCTP_ULP_ADAPTATION) {
6151			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6152				struct sctp_adaptation_layer_indication ai,
6153				                                *aip;
6154
6155				phdr = sctp_get_next_param(m, offset,
6156				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6157				aip = (struct sctp_adaptation_layer_indication *)phdr;
6158				if (aip) {
6159					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6160					stcb->asoc.adaptation_needed = 1;
6161				}
6162			}
6163		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6164			struct sctp_asconf_addr_param lstore, *fee;
6165			struct sctp_asconf_addrv4_param *fii;
6166			int lptype;
6167			struct sockaddr *lsa = NULL;
6168
6169			stcb->asoc.peer_supports_asconf = 1;
6170			if (plen > sizeof(lstore)) {
6171				return (-23);
6172			}
6173			phdr = sctp_get_next_param(m, offset,
6174			    (struct sctp_paramhdr *)&lstore,
6175			    min(plen, sizeof(lstore)));
6176			if (phdr == NULL) {
6177				return (-24);
6178			}
6179			fee = (struct sctp_asconf_addr_param *)phdr;
6180			lptype = ntohs(fee->addrp.ph.param_type);
6181			if (lptype == SCTP_IPV4_ADDRESS) {
6182				if (plen !=
6183				    sizeof(struct sctp_asconf_addrv4_param)) {
6184					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6185					    (int)sizeof(struct sctp_asconf_addrv4_param),
6186					    plen);
6187				} else {
6188					fii = (struct sctp_asconf_addrv4_param *)fee;
6189					sin.sin_addr.s_addr = fii->addrp.addr;
6190					lsa = (struct sockaddr *)&sin;
6191				}
6192			} else if (lptype == SCTP_IPV6_ADDRESS) {
6193				if (plen !=
6194				    sizeof(struct sctp_asconf_addr_param)) {
6195					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6196					    (int)sizeof(struct sctp_asconf_addr_param),
6197					    plen);
6198				} else {
6199					memcpy(sin6.sin6_addr.s6_addr,
6200					    fee->addrp.addr,
6201					    sizeof(fee->addrp.addr));
6202					lsa = (struct sockaddr *)&sin6;
6203				}
6204			}
6205			if (lsa) {
6206				(void)sctp_set_primary_addr(stcb, sa, NULL);
6207			}
6208		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6209			stcb->asoc.peer_supports_nat = 1;
6210		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6211			/* Peer supports pr-sctp */
6212			stcb->asoc.peer_supports_prsctp = 1;
6213		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6214			/* A supported extension chunk */
6215			struct sctp_supported_chunk_types_param *pr_supported;
6216			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6217			int num_ent, i;
6218
6219			phdr = sctp_get_next_param(m, offset,
6220			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6221			if (phdr == NULL) {
6222				return (-25);
6223			}
6224			stcb->asoc.peer_supports_asconf = 0;
6225			stcb->asoc.peer_supports_prsctp = 0;
6226			stcb->asoc.peer_supports_pktdrop = 0;
6227			stcb->asoc.peer_supports_strreset = 0;
6228			stcb->asoc.peer_supports_nr_sack = 0;
6229			stcb->asoc.peer_supports_auth = 0;
6230			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6231			num_ent = plen - sizeof(struct sctp_paramhdr);
6232			for (i = 0; i < num_ent; i++) {
6233				switch (pr_supported->chunk_types[i]) {
6234				case SCTP_ASCONF:
6235				case SCTP_ASCONF_ACK:
6236					stcb->asoc.peer_supports_asconf = 1;
6237					break;
6238				case SCTP_FORWARD_CUM_TSN:
6239					stcb->asoc.peer_supports_prsctp = 1;
6240					break;
6241				case SCTP_PACKET_DROPPED:
6242					stcb->asoc.peer_supports_pktdrop = 1;
6243					break;
6244				case SCTP_NR_SELECTIVE_ACK:
6245					stcb->asoc.peer_supports_nr_sack = 1;
6246					break;
6247				case SCTP_STREAM_RESET:
6248					stcb->asoc.peer_supports_strreset = 1;
6249					break;
6250				case SCTP_AUTHENTICATION:
6251					stcb->asoc.peer_supports_auth = 1;
6252					break;
6253				default:
6254					/* one I have not learned yet */
6255					break;
6256
6257				}
6258			}
6259		} else if (ptype == SCTP_ECN_NONCE_SUPPORTED) {
6260			/* Peer supports ECN-nonce */
6261			stcb->asoc.peer_supports_ecn_nonce = 1;
6262			stcb->asoc.ecn_nonce_allowed = 1;
6263		} else if (ptype == SCTP_RANDOM) {
6264			if (plen > sizeof(random_store))
6265				break;
6266			if (got_random) {
6267				/* already processed a RANDOM */
6268				goto next_param;
6269			}
6270			phdr = sctp_get_next_param(m, offset,
6271			    (struct sctp_paramhdr *)random_store,
6272			    min(sizeof(random_store), plen));
6273			if (phdr == NULL)
6274				return (-26);
6275			p_random = (struct sctp_auth_random *)phdr;
6276			random_len = plen - sizeof(*p_random);
6277			/* enforce the random length */
6278			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6279				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6280				return (-27);
6281			}
6282			got_random = 1;
6283		} else if (ptype == SCTP_HMAC_LIST) {
6284			int num_hmacs;
6285			int i;
6286
6287			if (plen > sizeof(hmacs_store))
6288				break;
6289			if (got_hmacs) {
6290				/* already processed a HMAC list */
6291				goto next_param;
6292			}
6293			phdr = sctp_get_next_param(m, offset,
6294			    (struct sctp_paramhdr *)hmacs_store,
6295			    min(plen, sizeof(hmacs_store)));
6296			if (phdr == NULL)
6297				return (-28);
6298			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6299			hmacs_len = plen - sizeof(*hmacs);
6300			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6301			/* validate the hmac list */
6302			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6303				return (-29);
6304			}
6305			if (stcb->asoc.peer_hmacs != NULL)
6306				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6307			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6308			if (stcb->asoc.peer_hmacs != NULL) {
6309				for (i = 0; i < num_hmacs; i++) {
6310					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6311					    ntohs(hmacs->hmac_ids[i]));
6312				}
6313			}
6314			got_hmacs = 1;
6315		} else if (ptype == SCTP_CHUNK_LIST) {
6316			int i;
6317
6318			if (plen > sizeof(chunks_store))
6319				break;
6320			if (got_chklist) {
6321				/* already processed a Chunks list */
6322				goto next_param;
6323			}
6324			phdr = sctp_get_next_param(m, offset,
6325			    (struct sctp_paramhdr *)chunks_store,
6326			    min(plen, sizeof(chunks_store)));
6327			if (phdr == NULL)
6328				return (-30);
6329			chunks = (struct sctp_auth_chunk_list *)phdr;
6330			num_chunks = plen - sizeof(*chunks);
6331			if (stcb->asoc.peer_auth_chunks != NULL)
6332				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6333			else
6334				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6335			for (i = 0; i < num_chunks; i++) {
6336				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6337				    stcb->asoc.peer_auth_chunks);
6338				/* record asconf/asconf-ack if listed */
6339				if (chunks->chunk_types[i] == SCTP_ASCONF)
6340					saw_asconf = 1;
6341				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6342					saw_asconf_ack = 1;
6343
6344			}
6345			got_chklist = 1;
6346		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6347			    (ptype == SCTP_STATE_COOKIE) ||
6348			    (ptype == SCTP_UNRECOG_PARAM) ||
6349			    (ptype == SCTP_COOKIE_PRESERVE) ||
6350			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6351			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6352			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6353			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6354		    (ptype == SCTP_SUCCESS_REPORT)) {
6355			 /* don't care */ ;
6356		} else {
6357			if ((ptype & 0x8000) == 0x0000) {
6358				/*
6359				 * must stop processing the rest of the
6360				 * param's. Any report bits were handled
6361				 * with the call to
6362				 * sctp_arethere_unrecognized_parameters()
6363				 * when the INIT or INIT-ACK was first seen.
6364				 */
6365				break;
6366			}
6367		}
6368
6369next_param:
6370		offset += SCTP_SIZE32(plen);
6371		if (offset >= limit) {
6372			break;
6373		}
6374		phdr = sctp_get_next_param(m, offset, &parm_buf,
6375		    sizeof(parm_buf));
6376	}
6377	/* Now check to see if we need to purge any addresses */
6378	for (net = TAILQ_FIRST(&stcb->asoc.nets); net != NULL; net = net_tmp) {
6379		net_tmp = TAILQ_NEXT(net, sctp_next);
6380		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6381		    SCTP_ADDR_NOT_IN_ASSOC) {
6382			/* This address has been removed from the asoc */
6383			/* remove and free it */
6384			stcb->asoc.numnets--;
6385			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6386			sctp_free_remote_addr(net);
6387			if (net == stcb->asoc.primary_destination) {
6388				stcb->asoc.primary_destination = NULL;
6389				sctp_select_primary_destination(stcb);
6390			}
6391		}
6392	}
6393	/* validate authentication required parameters */
6394	if (got_random && got_hmacs) {
6395		stcb->asoc.peer_supports_auth = 1;
6396	} else {
6397		stcb->asoc.peer_supports_auth = 0;
6398	}
6399	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6400		/* peer does not support auth but sent a chunks list? */
6401		return (-31);
6402	}
6403	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6404	    !stcb->asoc.peer_supports_auth) {
6405		/* peer supports asconf but not auth? */
6406		return (-32);
6407	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6408	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6409		return (-33);
6410	}
6411	/* concatenate the full random key */
6412	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6413	if (chunks != NULL) {
6414		keylen += sizeof(*chunks) + num_chunks;
6415	}
6416	new_key = sctp_alloc_key(keylen);
6417	if (new_key != NULL) {
6418		/* copy in the RANDOM */
6419		if (p_random != NULL) {
6420			keylen = sizeof(*p_random) + random_len;
6421			bcopy(p_random, new_key->key, keylen);
6422		}
6423		/* append in the AUTH chunks */
6424		if (chunks != NULL) {
6425			bcopy(chunks, new_key->key + keylen,
6426			    sizeof(*chunks) + num_chunks);
6427			keylen += sizeof(*chunks) + num_chunks;
6428		}
6429		/* append in the HMACs */
6430		if (hmacs != NULL) {
6431			bcopy(hmacs, new_key->key + keylen,
6432			    sizeof(*hmacs) + hmacs_len);
6433		}
6434	} else {
6435		/* failed to get memory for the key */
6436		return (-34);
6437	}
6438	if (stcb->asoc.authinfo.peer_random != NULL)
6439		sctp_free_key(stcb->asoc.authinfo.peer_random);
6440	stcb->asoc.authinfo.peer_random = new_key;
6441	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6442	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6443
6444	return (0);
6445}
6446
6447int
6448sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6449    struct sctp_nets *net)
6450{
6451	/* make sure the requested primary address exists in the assoc */
6452	if (net == NULL && sa)
6453		net = sctp_findnet(stcb, sa);
6454
6455	if (net == NULL) {
6456		/* didn't find the requested primary address! */
6457		return (-1);
6458	} else {
6459		/* set the primary address */
6460		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6461			/* Must be confirmed, so queue to set */
6462			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6463			return (0);
6464		}
6465		stcb->asoc.primary_destination = net;
6466		net->dest_state &= ~SCTP_ADDR_WAS_PRIMARY;
6467		net = TAILQ_FIRST(&stcb->asoc.nets);
6468		if (net != stcb->asoc.primary_destination) {
6469			/*
6470			 * first one on the list is NOT the primary
6471			 * sctp_cmpaddr() is much more efficent if the
6472			 * primary is the first on the list, make it so.
6473			 */
6474			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6475			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6476		}
6477		return (0);
6478	}
6479}
6480
6481int
6482sctp_is_vtag_good(struct sctp_inpcb *inp, uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now, int save_in_twait)
6483{
6484	/*
6485	 * This function serves two purposes. It will see if a TAG can be
6486	 * re-used and return 1 for yes it is ok and 0 for don't use that
6487	 * tag. A secondary function it will do is purge out old tags that
6488	 * can be removed.
6489	 */
6490	struct sctpvtaghead *chain;
6491	struct sctp_tagblock *twait_block;
6492	struct sctpasochead *head;
6493	struct sctp_tcb *stcb;
6494	int i;
6495
6496	SCTP_INP_INFO_RLOCK();
6497	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6498	    SCTP_BASE_INFO(hashasocmark))];
6499	if (head == NULL) {
6500		/* invalid vtag */
6501		goto skip_vtag_check;
6502	}
6503	LIST_FOREACH(stcb, head, sctp_asocs) {
6504		/*
6505		 * We choose not to lock anything here. TCB's can't be
6506		 * removed since we have the read lock, so they can't be
6507		 * freed on us, same thing for the INP. I may be wrong with
6508		 * this assumption, but we will go with it for now :-)
6509		 */
6510		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6511			continue;
6512		}
6513		if (stcb->asoc.my_vtag == tag) {
6514			/* candidate */
6515			if (stcb->rport != rport) {
6516				continue;
6517			}
6518			if (stcb->sctp_ep->sctp_lport != lport) {
6519				continue;
6520			}
6521			/* Its a used tag set */
6522			SCTP_INP_INFO_RUNLOCK();
6523			return (0);
6524		}
6525	}
6526skip_vtag_check:
6527
6528	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6529	/* Now what about timed wait ? */
6530	if (!LIST_EMPTY(chain)) {
6531		/*
6532		 * Block(s) are present, lets see if we have this tag in the
6533		 * list
6534		 */
6535		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6536			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6537				if (twait_block->vtag_block[i].v_tag == 0) {
6538					/* not used */
6539					continue;
6540				} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6541				    now->tv_sec) {
6542					/* Audit expires this guy */
6543					twait_block->vtag_block[i].tv_sec_at_expire = 0;
6544					twait_block->vtag_block[i].v_tag = 0;
6545					twait_block->vtag_block[i].lport = 0;
6546					twait_block->vtag_block[i].rport = 0;
6547				} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6548					    (twait_block->vtag_block[i].lport == lport) &&
6549				    (twait_block->vtag_block[i].rport == rport)) {
6550					/* Bad tag, sorry :< */
6551					SCTP_INP_INFO_RUNLOCK();
6552					return (0);
6553				}
6554			}
6555		}
6556	}
6557	SCTP_INP_INFO_RUNLOCK();
6558	return (1);
6559}
6560
6561
6562static sctp_assoc_t reneged_asoc_ids[256];
6563static uint8_t reneged_at = 0;
6564
6565
6566static void
6567sctp_drain_mbufs(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
6568{
6569	/*
6570	 * We must hunt this association for MBUF's past the cumack (i.e.
6571	 * out of order data that we can renege on).
6572	 */
6573	struct sctp_association *asoc;
6574	struct sctp_tmit_chunk *chk, *nchk;
6575	uint32_t cumulative_tsn_p1;
6576	struct sctp_queued_to_read *ctl, *nctl;
6577	int cnt, strmat;
6578	uint32_t gap, i;
6579	int fnd = 0;
6580
6581	/* We look for anything larger than the cum-ack + 1 */
6582
6583	asoc = &stcb->asoc;
6584	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6585		/* none we can reneg on. */
6586		return;
6587	}
6588	SCTP_STAT_INCR(sctps_protocol_drains_done);
6589	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6590	cnt = 0;
6591	/* First look in the re-assembly queue */
6592	chk = TAILQ_FIRST(&asoc->reasmqueue);
6593	while (chk) {
6594		/* Get the next one */
6595		nchk = TAILQ_NEXT(chk, sctp_next);
6596		if (compare_with_wrap(chk->rec.data.TSN_seq,
6597		    cumulative_tsn_p1, MAX_TSN)) {
6598			/* Yep it is above cum-ack */
6599			cnt++;
6600			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6601			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6602			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6603			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6604			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6605			if (chk->data) {
6606				sctp_m_freem(chk->data);
6607				chk->data = NULL;
6608			}
6609			sctp_free_a_chunk(stcb, chk);
6610		}
6611		chk = nchk;
6612	}
6613	/* Ok that was fun, now we will drain all the inbound streams? */
6614	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6615		ctl = TAILQ_FIRST(&asoc->strmin[strmat].inqueue);
6616		while (ctl) {
6617			nctl = TAILQ_NEXT(ctl, next);
6618			if (compare_with_wrap(ctl->sinfo_tsn,
6619			    cumulative_tsn_p1, MAX_TSN)) {
6620				/* Yep it is above cum-ack */
6621				cnt++;
6622				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6623				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6624				sctp_ucount_decr(asoc->cnt_on_all_streams);
6625				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6626				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6627				if (ctl->data) {
6628					sctp_m_freem(ctl->data);
6629					ctl->data = NULL;
6630				}
6631				sctp_free_remote_addr(ctl->whoFrom);
6632				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6633				SCTP_DECR_READQ_COUNT();
6634			}
6635			ctl = nctl;
6636		}
6637	}
6638	if (cnt) {
6639		/* We must back down to see what the new highest is */
6640		for (i = asoc->highest_tsn_inside_map;
6641		    (compare_with_wrap(i, asoc->mapping_array_base_tsn, MAX_TSN) || (i == asoc->mapping_array_base_tsn));
6642		    i--) {
6643			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6644			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6645				asoc->highest_tsn_inside_map = i;
6646				fnd = 1;
6647				break;
6648			}
6649		}
6650		if (!fnd) {
6651			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6652		}
6653		/*
6654		 * Question, should we go through the delivery queue? The
6655		 * only reason things are on here is the app not reading OR
6656		 * a p-d-api up. An attacker COULD send enough in to
6657		 * initiate the PD-API and then send a bunch of stuff to
6658		 * other streams... these would wind up on the delivery
6659		 * queue.. and then we would not get to them. But in order
6660		 * to do this I then have to back-track and un-deliver
6661		 * sequence numbers in streams.. el-yucko. I think for now
6662		 * we will NOT look at the delivery queue and leave it to be
6663		 * something to consider later. An alternative would be to
6664		 * abort the P-D-API with a notification and then deliver
6665		 * the data.... Or another method might be to keep track of
6666		 * how many times the situation occurs and if we see a
6667		 * possible attack underway just abort the association.
6668		 */
6669#ifdef SCTP_DEBUG
6670		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6671#endif
6672		/*
6673		 * Now do we need to find a new
6674		 * asoc->highest_tsn_inside_map?
6675		 */
6676		asoc->last_revoke_count = cnt;
6677		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6678		/* sa_ignore NO_NULL_CHK */
6679		sctp_send_sack(stcb);
6680		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6681		reneged_asoc_ids[reneged_at] = sctp_get_associd(stcb);
6682		reneged_at++;
6683	}
6684	/*
6685	 * Another issue, in un-setting the TSN's in the mapping array we
6686	 * DID NOT adjust the higest_tsn marker.  This will cause one of two
6687	 * things to occur. It may cause us to do extra work in checking for
6688	 * our mapping array movement. More importantly it may cause us to
6689	 * SACK every datagram. This may not be a bad thing though since we
6690	 * will recover once we get our cum-ack above and all this stuff we
6691	 * dumped recovered.
6692	 */
6693}
6694
6695void
6696sctp_drain()
6697{
6698	/*
6699	 * We must walk the PCB lists for ALL associations here. The system
6700	 * is LOW on MBUF's and needs help. This is where reneging will
6701	 * occur. We really hope this does NOT happen!
6702	 */
6703	VNET_ITERATOR_DECL(vnet_iter);
6704	VNET_LIST_RLOCK_NOSLEEP();
6705	VNET_FOREACH(vnet_iter) {
6706		CURVNET_SET(vnet_iter);
6707		struct sctp_inpcb *inp;
6708		struct sctp_tcb *stcb;
6709
6710		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6711		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6712#ifdef VIMAGE
6713			continue;
6714#else
6715			return;
6716#endif
6717		}
6718		SCTP_INP_INFO_RLOCK();
6719		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6720			/* For each endpoint */
6721			SCTP_INP_RLOCK(inp);
6722			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6723				/* For each association */
6724				SCTP_TCB_LOCK(stcb);
6725				sctp_drain_mbufs(inp, stcb);
6726				SCTP_TCB_UNLOCK(stcb);
6727			}
6728			SCTP_INP_RUNLOCK(inp);
6729		}
6730		SCTP_INP_INFO_RUNLOCK();
6731		CURVNET_RESTORE();
6732	}
6733	VNET_LIST_RUNLOCK_NOSLEEP();
6734}
6735
6736/*
6737 * start a new iterator
6738 * iterates through all endpoints and associations based on the pcb_state
6739 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6740 * assocs and "ef" (optional) is executed when the iterator completes.
6741 * "inpf" (optional) is executed for each new endpoint as it is being
6742 * iterated through. inpe (optional) is called when the inp completes
6743 * its way through all the stcbs.
6744 */
6745int
6746sctp_initiate_iterator(inp_func inpf,
6747    asoc_func af,
6748    inp_func inpe,
6749    uint32_t pcb_state,
6750    uint32_t pcb_features,
6751    uint32_t asoc_state,
6752    void *argp,
6753    uint32_t argi,
6754    end_func ef,
6755    struct sctp_inpcb *s_inp,
6756    uint8_t chunk_output_off)
6757{
6758	struct sctp_iterator *it = NULL;
6759
6760	if (af == NULL) {
6761		return (-1);
6762	}
6763	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6764	    SCTP_M_ITER);
6765	if (it == NULL) {
6766		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6767		return (ENOMEM);
6768	}
6769	memset(it, 0, sizeof(*it));
6770	it->function_assoc = af;
6771	it->function_inp = inpf;
6772	if (inpf)
6773		it->done_current_ep = 0;
6774	else
6775		it->done_current_ep = 1;
6776	it->function_atend = ef;
6777	it->pointer = argp;
6778	it->val = argi;
6779	it->pcb_flags = pcb_state;
6780	it->pcb_features = pcb_features;
6781	it->asoc_state = asoc_state;
6782	it->function_inp_end = inpe;
6783	it->no_chunk_output = chunk_output_off;
6784	it->vn = curvnet;
6785	if (s_inp) {
6786		/* Assume lock is held here */
6787		it->inp = s_inp;
6788		SCTP_INP_INCR_REF(it->inp);
6789		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
6790	} else {
6791		SCTP_INP_INFO_RLOCK();
6792		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
6793		if (it->inp) {
6794			SCTP_INP_INCR_REF(it->inp);
6795		}
6796		SCTP_INP_INFO_RUNLOCK();
6797		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
6798
6799	}
6800	SCTP_IPI_ITERATOR_WQ_LOCK();
6801
6802	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
6803	if (sctp_it_ctl.iterator_running == 0) {
6804		sctp_wakeup_iterator();
6805	}
6806	SCTP_IPI_ITERATOR_WQ_UNLOCK();
6807	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
6808	return (0);
6809}
6810