sctp_pcb.c revision 208883
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 208883 2010-06-07 04:01:38Z 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:%u", sctp_ifnp->ifn_index);
521#endif
522			return (NULL);
523		}
524		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
525		sctp_ifnp->ifn_index = ifn_index;
526		sctp_ifnp->ifn_p = ifn;
527		sctp_ifnp->ifn_type = ifn_type;
528		sctp_ifnp->refcount = 0;
529		sctp_ifnp->vrf = vrf;
530		atomic_add_int(&vrf->refcount, 1);
531		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
532		if (if_name != NULL) {
533			memcpy(sctp_ifnp->ifn_name, if_name, SCTP_IFNAMSIZ);
534		} else {
535			memcpy(sctp_ifnp->ifn_name, "unknown", min(7, SCTP_IFNAMSIZ));
536		}
537		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
538		LIST_INIT(&sctp_ifnp->ifalist);
539		SCTP_IPI_ADDR_WLOCK();
540		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
541		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
542		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
543		new_ifn_af = 1;
544	}
545	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
546	if (sctp_ifap) {
547		/* Hmm, it already exists? */
548		if ((sctp_ifap->ifn_p) &&
549		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
550			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
551			    sctp_ifap->ifn_p->ifn_name, ifn_index,
552			    sctp_ifap);
553			if (new_ifn_af) {
554				/* Remove the created one that we don't want */
555				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
556			}
557			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
558				/* easy to solve, just switch back to active */
559				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
560				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
561				sctp_ifap->ifn_p = sctp_ifnp;
562				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
563			}
564	exit_stage_left:
565			SCTP_IPI_ADDR_WUNLOCK();
566			return (sctp_ifap);
567		} else {
568			if (sctp_ifap->ifn_p) {
569				/*
570				 * The last IFN gets the address, removee
571				 * the old one
572				 */
573				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
574				    sctp_ifap, sctp_ifap->ifn_p->ifn_name,
575				    sctp_ifap->ifn_p->ifn_index, if_name,
576				    ifn_index);
577				/* remove the address from the old ifn */
578				sctp_remove_ifa_from_ifn(sctp_ifap);
579				/* move the address over to the new ifn */
580				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
581				goto exit_stage_left;
582			} else {
583				/* repair ifnp which was NULL ? */
584				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
585				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
586				    sctp_ifnp, sctp_ifap);
587				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
588			}
589			goto exit_stage_left;
590		}
591	}
592	SCTP_IPI_ADDR_WUNLOCK();
593	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
594	if (sctp_ifap == NULL) {
595#ifdef INVARIANTS
596		panic("No memory for IFA");
597#endif
598		return (NULL);
599	}
600	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
601	sctp_ifap->ifn_p = sctp_ifnp;
602	atomic_add_int(&sctp_ifnp->refcount, 1);
603	sctp_ifap->vrf_id = vrf_id;
604	sctp_ifap->ifa = ifa;
605	memcpy(&sctp_ifap->address, addr, addr->sa_len);
606	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
607	sctp_ifap->flags = ifa_flags;
608	/* Set scope */
609	switch (sctp_ifap->address.sa.sa_family) {
610	case AF_INET:
611		{
612			struct sockaddr_in *sin;
613
614			sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
615			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
616			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
617				sctp_ifap->src_is_loop = 1;
618			}
619			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
620				sctp_ifap->src_is_priv = 1;
621			}
622			sctp_ifnp->num_v4++;
623			if (new_ifn_af)
624				new_ifn_af = AF_INET;
625			break;
626		}
627#ifdef INET6
628	case AF_INET6:
629		{
630			/* ok to use deprecated addresses? */
631			struct sockaddr_in6 *sin6;
632
633			sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
634			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
635			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
636				sctp_ifap->src_is_loop = 1;
637			}
638			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
639				sctp_ifap->src_is_priv = 1;
640			}
641			sctp_ifnp->num_v6++;
642			if (new_ifn_af)
643				new_ifn_af = AF_INET6;
644			break;
645		}
646#endif
647	default:
648		new_ifn_af = 0;
649		break;
650	}
651	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
652
653	if ((sctp_ifap->src_is_priv == 0) &&
654	    (sctp_ifap->src_is_loop == 0)) {
655		sctp_ifap->src_is_glob = 1;
656	}
657	SCTP_IPI_ADDR_WLOCK();
658	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
659	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
660	sctp_ifap->refcount = 1;
661	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
662	sctp_ifnp->ifa_count++;
663	vrf->total_ifa_count++;
664	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
665	if (new_ifn_af) {
666		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
667		sctp_ifnp->registered_af = new_ifn_af;
668	}
669	SCTP_IPI_ADDR_WUNLOCK();
670	if (dynamic_add) {
671		/*
672		 * Bump up the refcount so that when the timer completes it
673		 * will drop back down.
674		 */
675		struct sctp_laddr *wi;
676
677		atomic_add_int(&sctp_ifap->refcount, 1);
678		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
679		if (wi == NULL) {
680			/*
681			 * Gak, what can we do? We have lost an address
682			 * change can you say HOSED?
683			 */
684			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
685			/* Opps, must decrement the count */
686			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
687			    if_name);
688			return (NULL);
689		}
690		SCTP_INCR_LADDR_COUNT();
691		bzero(wi, sizeof(*wi));
692		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
693		wi->ifa = sctp_ifap;
694		wi->action = SCTP_ADD_IP_ADDRESS;
695
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
1013/*
1014 * rules for use
1015 *
1016 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1017 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1018 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1019 * NULL.
1020 */
1021
1022struct sctp_tcb *
1023sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1024    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1025{
1026	struct sctpasochead *head;
1027	struct sctp_inpcb *inp;
1028	struct sctp_tcb *stcb = NULL;
1029	struct sctp_nets *net;
1030	uint16_t rport;
1031
1032	inp = *inp_p;
1033	if (remote->sa_family == AF_INET) {
1034		rport = (((struct sockaddr_in *)remote)->sin_port);
1035	} else if (remote->sa_family == AF_INET6) {
1036		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1037	} else {
1038		return (NULL);
1039	}
1040	if (locked_tcb) {
1041		/*
1042		 * UN-lock so we can do proper locking here this occurs when
1043		 * called from load_addresses_from_init.
1044		 */
1045		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1046		SCTP_TCB_UNLOCK(locked_tcb);
1047	}
1048	SCTP_INP_INFO_RLOCK();
1049	if (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) {
1050		/*-
1051		 * Now either this guy is our listener or it's the
1052		 * connector. If it is the one that issued the connect, then
1053		 * it's only chance is to be the first TCB in the list. If
1054		 * it is the acceptor, then do the special_lookup to hash
1055		 * and find the real inp.
1056		 */
1057		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1058			/* to is peer addr, from is my addr */
1059			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1060			    netp, inp->def_vrf_id);
1061			if ((stcb != NULL) && (locked_tcb == NULL)) {
1062				/* we have a locked tcb, lower refcount */
1063				SCTP_INP_DECR_REF(inp);
1064			}
1065			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1066				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1067				SCTP_TCB_LOCK(locked_tcb);
1068				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1069				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1070			}
1071			SCTP_INP_INFO_RUNLOCK();
1072			return (stcb);
1073		} else {
1074			SCTP_INP_WLOCK(inp);
1075			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1076				goto null_return;
1077			}
1078			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1079			if (stcb == NULL) {
1080				goto null_return;
1081			}
1082			SCTP_TCB_LOCK(stcb);
1083
1084			if (stcb->rport != rport) {
1085				/* remote port does not match. */
1086				SCTP_TCB_UNLOCK(stcb);
1087				goto null_return;
1088			}
1089			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1090				SCTP_TCB_UNLOCK(stcb);
1091				goto null_return;
1092			}
1093			/* now look at the list of remote addresses */
1094			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1095#ifdef INVARIANTS
1096				if (net == (TAILQ_NEXT(net, sctp_next))) {
1097					panic("Corrupt net list");
1098				}
1099#endif
1100				if (net->ro._l_addr.sa.sa_family !=
1101				    remote->sa_family) {
1102					/* not the same family */
1103					continue;
1104				}
1105				switch (remote->sa_family) {
1106				case AF_INET:
1107					{
1108						struct sockaddr_in *sin,
1109						           *rsin;
1110
1111						sin = (struct sockaddr_in *)
1112						    &net->ro._l_addr;
1113						rsin = (struct sockaddr_in *)remote;
1114						if (sin->sin_addr.s_addr ==
1115						    rsin->sin_addr.s_addr) {
1116							/* found it */
1117							if (netp != NULL) {
1118								*netp = net;
1119							}
1120							if (locked_tcb == NULL) {
1121								SCTP_INP_DECR_REF(inp);
1122							} else if (locked_tcb != stcb) {
1123								SCTP_TCB_LOCK(locked_tcb);
1124							}
1125							if (locked_tcb) {
1126								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1127							}
1128							SCTP_INP_WUNLOCK(inp);
1129							SCTP_INP_INFO_RUNLOCK();
1130							return (stcb);
1131						}
1132						break;
1133					}
1134#ifdef INET6
1135				case AF_INET6:
1136					{
1137						struct sockaddr_in6 *sin6,
1138						            *rsin6;
1139
1140						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1141						rsin6 = (struct sockaddr_in6 *)remote;
1142						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1143						    rsin6)) {
1144							/* found it */
1145							if (netp != NULL) {
1146								*netp = net;
1147							}
1148							if (locked_tcb == NULL) {
1149								SCTP_INP_DECR_REF(inp);
1150							} else if (locked_tcb != stcb) {
1151								SCTP_TCB_LOCK(locked_tcb);
1152							}
1153							if (locked_tcb) {
1154								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1155							}
1156							SCTP_INP_WUNLOCK(inp);
1157							SCTP_INP_INFO_RUNLOCK();
1158							return (stcb);
1159						}
1160						break;
1161					}
1162#endif
1163				default:
1164					/* TSNH */
1165					break;
1166				}
1167			}
1168			SCTP_TCB_UNLOCK(stcb);
1169		}
1170	} else {
1171		SCTP_INP_WLOCK(inp);
1172		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1173			goto null_return;
1174		}
1175		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1176		    inp->sctp_hashmark)];
1177		if (head == NULL) {
1178			goto null_return;
1179		}
1180		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1181			if (stcb->rport != rport) {
1182				/* remote port does not match */
1183				continue;
1184			}
1185			SCTP_TCB_LOCK(stcb);
1186			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1187				SCTP_TCB_UNLOCK(stcb);
1188				continue;
1189			}
1190			/* now look at the list of remote addresses */
1191			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1192#ifdef INVARIANTS
1193				if (net == (TAILQ_NEXT(net, sctp_next))) {
1194					panic("Corrupt net list");
1195				}
1196#endif
1197				if (net->ro._l_addr.sa.sa_family !=
1198				    remote->sa_family) {
1199					/* not the same family */
1200					continue;
1201				}
1202				switch (remote->sa_family) {
1203				case AF_INET:
1204					{
1205						struct sockaddr_in *sin,
1206						           *rsin;
1207
1208						sin = (struct sockaddr_in *)
1209						    &net->ro._l_addr;
1210						rsin = (struct sockaddr_in *)remote;
1211						if (sin->sin_addr.s_addr ==
1212						    rsin->sin_addr.s_addr) {
1213							/* found it */
1214							if (netp != NULL) {
1215								*netp = net;
1216							}
1217							if (locked_tcb == NULL) {
1218								SCTP_INP_DECR_REF(inp);
1219							} else if (locked_tcb != stcb) {
1220								SCTP_TCB_LOCK(locked_tcb);
1221							}
1222							if (locked_tcb) {
1223								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1224							}
1225							SCTP_INP_WUNLOCK(inp);
1226							SCTP_INP_INFO_RUNLOCK();
1227							return (stcb);
1228						}
1229						break;
1230					}
1231#ifdef INET6
1232				case AF_INET6:
1233					{
1234						struct sockaddr_in6 *sin6,
1235						            *rsin6;
1236
1237						sin6 = (struct sockaddr_in6 *)
1238						    &net->ro._l_addr;
1239						rsin6 = (struct sockaddr_in6 *)remote;
1240						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1241						    rsin6)) {
1242							/* found it */
1243							if (netp != NULL) {
1244								*netp = net;
1245							}
1246							if (locked_tcb == NULL) {
1247								SCTP_INP_DECR_REF(inp);
1248							} else if (locked_tcb != stcb) {
1249								SCTP_TCB_LOCK(locked_tcb);
1250							}
1251							if (locked_tcb) {
1252								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1253							}
1254							SCTP_INP_WUNLOCK(inp);
1255							SCTP_INP_INFO_RUNLOCK();
1256							return (stcb);
1257						}
1258						break;
1259					}
1260#endif
1261				default:
1262					/* TSNH */
1263					break;
1264				}
1265			}
1266			SCTP_TCB_UNLOCK(stcb);
1267		}
1268	}
1269null_return:
1270	/* clean up for returning null */
1271	if (locked_tcb) {
1272		SCTP_TCB_LOCK(locked_tcb);
1273		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1274	}
1275	SCTP_INP_WUNLOCK(inp);
1276	SCTP_INP_INFO_RUNLOCK();
1277	/* not found */
1278	return (NULL);
1279}
1280
1281/*
1282 * Find an association for a specific endpoint using the association id given
1283 * out in the COMM_UP notification
1284 */
1285
1286struct sctp_tcb *
1287sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1288{
1289	/*
1290	 * Use my the assoc_id to find a endpoint
1291	 */
1292	struct sctpasochead *head;
1293	struct sctp_tcb *stcb;
1294	uint32_t id;
1295
1296	if (inp == NULL) {
1297		SCTP_PRINTF("TSNH ep_associd\n");
1298		return (NULL);
1299	}
1300	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1301		SCTP_PRINTF("TSNH ep_associd0\n");
1302		return (NULL);
1303	}
1304	id = (uint32_t) asoc_id;
1305	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1306	if (head == NULL) {
1307		/* invalid id TSNH */
1308		SCTP_PRINTF("TSNH ep_associd1\n");
1309		return (NULL);
1310	}
1311	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1312		if (stcb->asoc.assoc_id == id) {
1313			if (inp != stcb->sctp_ep) {
1314				/*
1315				 * some other guy has the same id active (id
1316				 * collision ??).
1317				 */
1318				SCTP_PRINTF("TSNH ep_associd2\n");
1319				continue;
1320			}
1321			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1322				continue;
1323			}
1324			if (want_lock) {
1325				SCTP_TCB_LOCK(stcb);
1326			}
1327			return (stcb);
1328		}
1329	}
1330	return (NULL);
1331}
1332
1333
1334struct sctp_tcb *
1335sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1336{
1337	struct sctp_tcb *stcb;
1338
1339	SCTP_INP_RLOCK(inp);
1340	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1341	SCTP_INP_RUNLOCK(inp);
1342	return (stcb);
1343}
1344
1345
1346static struct sctp_inpcb *
1347sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1348    uint16_t lport, uint32_t vrf_id)
1349{
1350	struct sctp_inpcb *inp;
1351	struct sockaddr_in *sin;
1352
1353#ifdef INET6
1354	struct sockaddr_in6 *sin6;
1355
1356#endif
1357	struct sctp_laddr *laddr;
1358
1359#ifdef INET6
1360	struct sockaddr_in6 *intf_addr6;
1361
1362#endif
1363
1364	int fnd;
1365
1366	/*
1367	 * Endpoing probe expects that the INP_INFO is locked.
1368	 */
1369	sin = NULL;
1370#ifdef INET6
1371	sin6 = NULL;
1372#endif
1373	switch (nam->sa_family) {
1374	case AF_INET:
1375		sin = (struct sockaddr_in *)nam;
1376		break;
1377#ifdef INET6
1378	case AF_INET6:
1379		sin6 = (struct sockaddr_in6 *)nam;
1380		break;
1381#endif
1382	default:
1383		/* unsupported family */
1384		return (NULL);
1385	}
1386
1387	if (head == NULL)
1388		return (NULL);
1389
1390	LIST_FOREACH(inp, head, sctp_hash) {
1391		SCTP_INP_RLOCK(inp);
1392		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1393			SCTP_INP_RUNLOCK(inp);
1394			continue;
1395		}
1396		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1397		    (inp->sctp_lport == lport)) {
1398			/* got it */
1399			if ((nam->sa_family == AF_INET) &&
1400			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1401			    SCTP_IPV6_V6ONLY(inp)) {
1402				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1403				SCTP_INP_RUNLOCK(inp);
1404				continue;
1405			}
1406			/* A V6 address and the endpoint is NOT bound V6 */
1407			if (nam->sa_family == AF_INET6 &&
1408			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1409				SCTP_INP_RUNLOCK(inp);
1410				continue;
1411			}
1412			/* does a VRF id match? */
1413			fnd = 0;
1414			if (inp->def_vrf_id == vrf_id)
1415				fnd = 1;
1416
1417			SCTP_INP_RUNLOCK(inp);
1418			if (!fnd)
1419				continue;
1420			return (inp);
1421		}
1422		SCTP_INP_RUNLOCK(inp);
1423	}
1424	if ((nam->sa_family == AF_INET) &&
1425	    (sin->sin_addr.s_addr == INADDR_ANY)) {
1426		/* Can't hunt for one that has no address specified */
1427		return (NULL);
1428	}
1429#ifdef INET6
1430	if ((nam->sa_family == AF_INET6) &&
1431	    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
1432		/* Can't hunt for one that has no address specified */
1433		return (NULL);
1434	}
1435#endif
1436	/*
1437	 * ok, not bound to all so see if we can find a EP bound to this
1438	 * address.
1439	 */
1440	LIST_FOREACH(inp, head, sctp_hash) {
1441		SCTP_INP_RLOCK(inp);
1442		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1443			SCTP_INP_RUNLOCK(inp);
1444			continue;
1445		}
1446		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1447			SCTP_INP_RUNLOCK(inp);
1448			continue;
1449		}
1450		/*
1451		 * Ok this could be a likely candidate, look at all of its
1452		 * addresses
1453		 */
1454		if (inp->sctp_lport != lport) {
1455			SCTP_INP_RUNLOCK(inp);
1456			continue;
1457		}
1458		/* does a VRF id match? */
1459		fnd = 0;
1460		if (inp->def_vrf_id == vrf_id)
1461			fnd = 1;
1462
1463		if (!fnd) {
1464			SCTP_INP_RUNLOCK(inp);
1465			continue;
1466		}
1467		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1468			if (laddr->ifa == NULL) {
1469				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1470				    __FUNCTION__);
1471				continue;
1472			}
1473			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1474			    laddr->ifa);
1475			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1476				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1477				continue;
1478			}
1479			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1480				/* possible, see if it matches */
1481				struct sockaddr_in *intf_addr;
1482
1483				intf_addr = &laddr->ifa->address.sin;
1484				switch (nam->sa_family) {
1485				case AF_INET:
1486					if (sin->sin_addr.s_addr ==
1487					    intf_addr->sin_addr.s_addr) {
1488						SCTP_INP_RUNLOCK(inp);
1489						return (inp);
1490					}
1491					break;
1492#ifdef INET6
1493				case AF_INET6:
1494					intf_addr6 = &laddr->ifa->address.sin6;
1495					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1496					    intf_addr6)) {
1497						SCTP_INP_RUNLOCK(inp);
1498						return (inp);
1499					}
1500					break;
1501#endif
1502				}
1503			}
1504		}
1505		SCTP_INP_RUNLOCK(inp);
1506	}
1507	return (NULL);
1508}
1509
1510
1511static struct sctp_inpcb *
1512sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1513{
1514	struct sctppcbhead *head;
1515	struct sctp_inpcb *t_inp;
1516	int fnd;
1517
1518	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1519	    SCTP_BASE_INFO(hashmark))];
1520	LIST_FOREACH(t_inp, head, sctp_hash) {
1521		if (t_inp->sctp_lport != lport) {
1522			continue;
1523		}
1524		/* is it in the VRF in question */
1525		fnd = 0;
1526		if (t_inp->def_vrf_id == vrf_id)
1527			fnd = 1;
1528		if (!fnd)
1529			continue;
1530
1531		/* This one is in use. */
1532		/* check the v6/v4 binding issue */
1533		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1534		    SCTP_IPV6_V6ONLY(t_inp)) {
1535			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1536				/* collision in V6 space */
1537				return (t_inp);
1538			} else {
1539				/* inp is BOUND_V4 no conflict */
1540				continue;
1541			}
1542		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1543			/* t_inp is bound v4 and v6, conflict always */
1544			return (t_inp);
1545		} else {
1546			/* t_inp is bound only V4 */
1547			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1548			    SCTP_IPV6_V6ONLY(inp)) {
1549				/* no conflict */
1550				continue;
1551			}
1552			/* else fall through to conflict */
1553		}
1554		return (t_inp);
1555	}
1556	return (NULL);
1557}
1558
1559
1560int
1561sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1562{
1563	/* For 1-2-1 with port reuse */
1564	struct sctppcbhead *head;
1565	struct sctp_inpcb *tinp;
1566
1567	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1568		/* only works with port reuse on */
1569		return (-1);
1570	}
1571	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1572		return (0);
1573	}
1574	SCTP_INP_RUNLOCK(inp);
1575	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1576	    SCTP_BASE_INFO(hashmark))];
1577	/* Kick out all non-listeners to the TCP hash */
1578	LIST_FOREACH(tinp, head, sctp_hash) {
1579		if (tinp->sctp_lport != inp->sctp_lport) {
1580			continue;
1581		}
1582		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1583			continue;
1584		}
1585		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1586			continue;
1587		}
1588		if (tinp->sctp_socket->so_qlimit) {
1589			continue;
1590		}
1591		SCTP_INP_WLOCK(tinp);
1592		LIST_REMOVE(tinp, sctp_hash);
1593		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1594		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1595		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1596		SCTP_INP_WUNLOCK(tinp);
1597	}
1598	SCTP_INP_WLOCK(inp);
1599	/* Pull from where he was */
1600	LIST_REMOVE(inp, sctp_hash);
1601	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1602	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1603	LIST_INSERT_HEAD(head, inp, sctp_hash);
1604	SCTP_INP_WUNLOCK(inp);
1605	SCTP_INP_RLOCK(inp);
1606	return (0);
1607}
1608
1609
1610struct sctp_inpcb *
1611sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1612    uint32_t vrf_id)
1613{
1614	/*
1615	 * First we check the hash table to see if someone has this port
1616	 * bound with just the port.
1617	 */
1618	struct sctp_inpcb *inp;
1619	struct sctppcbhead *head;
1620	struct sockaddr_in *sin;
1621	struct sockaddr_in6 *sin6;
1622	int lport;
1623
1624	if (nam->sa_family == AF_INET) {
1625		sin = (struct sockaddr_in *)nam;
1626		lport = ((struct sockaddr_in *)nam)->sin_port;
1627	} else if (nam->sa_family == AF_INET6) {
1628		sin6 = (struct sockaddr_in6 *)nam;
1629		lport = ((struct sockaddr_in6 *)nam)->sin6_port;
1630	} else {
1631		/* unsupported family */
1632		return (NULL);
1633	}
1634	/*
1635	 * I could cheat here and just cast to one of the types but we will
1636	 * do it right. It also provides the check against an Unsupported
1637	 * type too.
1638	 */
1639	/* Find the head of the ALLADDR chain */
1640	if (have_lock == 0) {
1641		SCTP_INP_INFO_RLOCK();
1642	}
1643	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1644	    SCTP_BASE_INFO(hashmark))];
1645	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1646
1647	/*
1648	 * If the TCP model exists it could be that the main listening
1649	 * endpoint is gone but there exists a connected socket for this guy
1650	 * yet. If so we can return the first one that we find. This may NOT
1651	 * be the correct one so the caller should be wary on the return
1652	 * INP. Currently the onlyc caller that sets this flag is in bindx
1653	 * where we are verifying that a user CAN bind the address. He
1654	 * either has bound it already, or someone else has, or its open to
1655	 * bind, so this is good enough.
1656	 */
1657	if (inp == NULL && find_tcp_pool) {
1658		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
1659		inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1660	}
1661	if (inp) {
1662		SCTP_INP_INCR_REF(inp);
1663	}
1664	if (have_lock == 0) {
1665		SCTP_INP_INFO_RUNLOCK();
1666	}
1667	return (inp);
1668}
1669
1670/*
1671 * Find an association for an endpoint with the pointer to whom you want to
1672 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1673 * need to change the *to to some other struct like a mbuf...
1674 */
1675struct sctp_tcb *
1676sctp_findassociation_addr_sa(struct sockaddr *to, struct sockaddr *from,
1677    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1678    uint32_t vrf_id)
1679{
1680	struct sctp_inpcb *inp = NULL;
1681	struct sctp_tcb *retval;
1682
1683	SCTP_INP_INFO_RLOCK();
1684	if (find_tcp_pool) {
1685		if (inp_p != NULL) {
1686			retval = sctp_tcb_special_locate(inp_p, from, to, netp,
1687			    vrf_id);
1688		} else {
1689			retval = sctp_tcb_special_locate(&inp, from, to, netp,
1690			    vrf_id);
1691		}
1692		if (retval != NULL) {
1693			SCTP_INP_INFO_RUNLOCK();
1694			return (retval);
1695		}
1696	}
1697	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1698	if (inp_p != NULL) {
1699		*inp_p = inp;
1700	}
1701	SCTP_INP_INFO_RUNLOCK();
1702
1703	if (inp == NULL) {
1704		return (NULL);
1705	}
1706	/*
1707	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1708	 * we now place the source address or from in the to of the find
1709	 * endpoint call. Since in reality this chain is used from the
1710	 * inbound packet side.
1711	 */
1712	if (inp_p != NULL) {
1713		retval = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1714		    NULL);
1715	} else {
1716		retval = sctp_findassociation_ep_addr(&inp, from, netp, to,
1717		    NULL);
1718	}
1719	return retval;
1720}
1721
1722
1723/*
1724 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1725 * find all addresses that the sender has specified in any address list. Each
1726 * address will be used to lookup the TCB and see if one exits.
1727 */
1728static struct sctp_tcb *
1729sctp_findassociation_special_addr(struct mbuf *m, int iphlen, int offset,
1730    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1731    struct sockaddr *dest)
1732{
1733	struct sockaddr_in sin4;
1734	struct sockaddr_in6 sin6;
1735	struct sctp_paramhdr *phdr, parm_buf;
1736	struct sctp_tcb *retval;
1737	uint32_t ptype, plen;
1738
1739	memset(&sin4, 0, sizeof(sin4));
1740	memset(&sin6, 0, sizeof(sin6));
1741	sin4.sin_len = sizeof(sin4);
1742	sin4.sin_family = AF_INET;
1743	sin4.sin_port = sh->src_port;
1744	sin6.sin6_len = sizeof(sin6);
1745	sin6.sin6_family = AF_INET6;
1746	sin6.sin6_port = sh->src_port;
1747
1748	retval = NULL;
1749	offset += sizeof(struct sctp_init_chunk);
1750
1751	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
1752	while (phdr != NULL) {
1753		/* now we must see if we want the parameter */
1754		ptype = ntohs(phdr->param_type);
1755		plen = ntohs(phdr->param_length);
1756		if (plen == 0) {
1757			break;
1758		}
1759		if (ptype == SCTP_IPV4_ADDRESS &&
1760		    plen == sizeof(struct sctp_ipv4addr_param)) {
1761			/* Get the rest of the address */
1762			struct sctp_ipv4addr_param ip4_parm, *p4;
1763
1764			phdr = sctp_get_next_param(m, offset,
1765			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
1766			if (phdr == NULL) {
1767				return (NULL);
1768			}
1769			p4 = (struct sctp_ipv4addr_param *)phdr;
1770			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
1771			/* look it up */
1772			retval = sctp_findassociation_ep_addr(inp_p,
1773			    (struct sockaddr *)&sin4, netp, dest, NULL);
1774			if (retval != NULL) {
1775				return (retval);
1776			}
1777		} else if (ptype == SCTP_IPV6_ADDRESS &&
1778		    plen == sizeof(struct sctp_ipv6addr_param)) {
1779			/* Get the rest of the address */
1780			struct sctp_ipv6addr_param ip6_parm, *p6;
1781
1782			phdr = sctp_get_next_param(m, offset,
1783			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
1784			if (phdr == NULL) {
1785				return (NULL);
1786			}
1787			p6 = (struct sctp_ipv6addr_param *)phdr;
1788			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
1789			/* look it up */
1790			retval = sctp_findassociation_ep_addr(inp_p,
1791			    (struct sockaddr *)&sin6, netp, dest, NULL);
1792			if (retval != NULL) {
1793				return (retval);
1794			}
1795		}
1796		offset += SCTP_SIZE32(plen);
1797		phdr = sctp_get_next_param(m, offset, &parm_buf,
1798		    sizeof(parm_buf));
1799	}
1800	return (NULL);
1801}
1802
1803static int
1804sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
1805{
1806	struct sctp_nets *net;
1807
1808	/*
1809	 * Simple question, the ports match, does the tcb own the to
1810	 * address?
1811	 */
1812	if ((stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1813		/* of course */
1814		return (1);
1815	}
1816	/* have to look at all bound addresses */
1817	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1818		if (net->ro._l_addr.sa.sa_family != to->sa_family) {
1819			/* not the same family, can't be a match */
1820			continue;
1821		}
1822		switch (to->sa_family) {
1823		case AF_INET:
1824			{
1825				struct sockaddr_in *sin, *rsin;
1826
1827				sin = (struct sockaddr_in *)&net->ro._l_addr;
1828				rsin = (struct sockaddr_in *)to;
1829				if (sin->sin_addr.s_addr ==
1830				    rsin->sin_addr.s_addr) {
1831					/* found it */
1832					return (1);
1833				}
1834				break;
1835			}
1836#ifdef INET6
1837		case AF_INET6:
1838			{
1839				struct sockaddr_in6 *sin6, *rsin6;
1840
1841				sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1842				rsin6 = (struct sockaddr_in6 *)to;
1843				if (SCTP6_ARE_ADDR_EQUAL(sin6,
1844				    rsin6)) {
1845					/* Update the endpoint pointer */
1846					return (1);
1847				}
1848				break;
1849			}
1850#endif
1851		default:
1852			/* TSNH */
1853			break;
1854		}
1855	}
1856	/* Nope, do not have the address ;-( */
1857	return (0);
1858}
1859
1860static struct sctp_tcb *
1861sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
1862    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
1863    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
1864{
1865	/*
1866	 * Use my vtag to hash. If we find it we then verify the source addr
1867	 * is in the assoc. If all goes well we save a bit on rec of a
1868	 * packet.
1869	 */
1870	struct sctpasochead *head;
1871	struct sctp_nets *net;
1872	struct sctp_tcb *stcb;
1873
1874	*netp = NULL;
1875	*inp_p = NULL;
1876	SCTP_INP_INFO_RLOCK();
1877	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
1878	    SCTP_BASE_INFO(hashasocmark))];
1879	if (head == NULL) {
1880		/* invalid vtag */
1881		SCTP_INP_INFO_RUNLOCK();
1882		return (NULL);
1883	}
1884	LIST_FOREACH(stcb, head, sctp_asocs) {
1885		SCTP_INP_RLOCK(stcb->sctp_ep);
1886		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1887			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1888			continue;
1889		}
1890		SCTP_TCB_LOCK(stcb);
1891		SCTP_INP_RUNLOCK(stcb->sctp_ep);
1892		if (stcb->asoc.my_vtag == vtag) {
1893			/* candidate */
1894			if (stcb->rport != rport) {
1895				SCTP_TCB_UNLOCK(stcb);
1896				continue;
1897			}
1898			if (stcb->sctp_ep->sctp_lport != lport) {
1899				SCTP_TCB_UNLOCK(stcb);
1900				continue;
1901			}
1902			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1903				SCTP_TCB_UNLOCK(stcb);
1904				continue;
1905			}
1906			/* RRS:Need toaddr check here */
1907			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
1908				/* Endpoint does not own this address */
1909				SCTP_TCB_UNLOCK(stcb);
1910				continue;
1911			}
1912			if (remote_tag) {
1913				/*
1914				 * If we have both vtags thats all we match
1915				 * on
1916				 */
1917				if (stcb->asoc.peer_vtag == remote_tag) {
1918					/*
1919					 * If both tags match we consider it
1920					 * conclusive and check NO
1921					 * source/destination addresses
1922					 */
1923					goto conclusive;
1924				}
1925			}
1926			if (skip_src_check) {
1927		conclusive:
1928				if (from) {
1929					net = sctp_findnet(stcb, from);
1930				} else {
1931					*netp = NULL;	/* unknown */
1932				}
1933				if (inp_p)
1934					*inp_p = stcb->sctp_ep;
1935				SCTP_INP_INFO_RUNLOCK();
1936				return (stcb);
1937			}
1938			net = sctp_findnet(stcb, from);
1939			if (net) {
1940				/* yep its him. */
1941				*netp = net;
1942				SCTP_STAT_INCR(sctps_vtagexpress);
1943				*inp_p = stcb->sctp_ep;
1944				SCTP_INP_INFO_RUNLOCK();
1945				return (stcb);
1946			} else {
1947				/*
1948				 * not him, this should only happen in rare
1949				 * cases so I peg it.
1950				 */
1951				SCTP_STAT_INCR(sctps_vtagbogus);
1952			}
1953		}
1954		SCTP_TCB_UNLOCK(stcb);
1955	}
1956	SCTP_INP_INFO_RUNLOCK();
1957	return (NULL);
1958}
1959
1960/*
1961 * Find an association with the pointer to the inbound IP packet. This can be
1962 * a IPv4 or IPv6 packet.
1963 */
1964struct sctp_tcb *
1965sctp_findassociation_addr(struct mbuf *m, int iphlen, int offset,
1966    struct sctphdr *sh, struct sctp_chunkhdr *ch,
1967    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
1968{
1969	int find_tcp_pool;
1970	struct ip *iph;
1971	struct sctp_tcb *retval;
1972	struct sockaddr_storage to_store, from_store;
1973	struct sockaddr *to = (struct sockaddr *)&to_store;
1974	struct sockaddr *from = (struct sockaddr *)&from_store;
1975	struct sctp_inpcb *inp;
1976
1977	iph = mtod(m, struct ip *);
1978	switch (iph->ip_v) {
1979	case IPVERSION:
1980		{
1981			/* its IPv4 */
1982			struct sockaddr_in *from4;
1983
1984			from4 = (struct sockaddr_in *)&from_store;
1985			bzero(from4, sizeof(*from4));
1986			from4->sin_family = AF_INET;
1987			from4->sin_len = sizeof(struct sockaddr_in);
1988			from4->sin_addr.s_addr = iph->ip_src.s_addr;
1989			from4->sin_port = sh->src_port;
1990			break;
1991		}
1992#ifdef INET6
1993	case IPV6_VERSION >> 4:
1994		{
1995			/* its IPv6 */
1996			struct ip6_hdr *ip6;
1997			struct sockaddr_in6 *from6;
1998
1999			ip6 = mtod(m, struct ip6_hdr *);
2000			from6 = (struct sockaddr_in6 *)&from_store;
2001			bzero(from6, sizeof(*from6));
2002			from6->sin6_family = AF_INET6;
2003			from6->sin6_len = sizeof(struct sockaddr_in6);
2004			from6->sin6_addr = ip6->ip6_src;
2005			from6->sin6_port = sh->src_port;
2006			/* Get the scopes in properly to the sin6 addr's */
2007			/* we probably don't need these operations */
2008			(void)sa6_recoverscope(from6);
2009			sa6_embedscope(from6, MODULE_GLOBAL(ip6_use_defzone));
2010			break;
2011		}
2012#endif
2013	default:
2014		/* Currently not supported. */
2015		return (NULL);
2016	}
2017
2018
2019	switch (iph->ip_v) {
2020	case IPVERSION:
2021		{
2022			/* its IPv4 */
2023			struct sockaddr_in *to4;
2024
2025			to4 = (struct sockaddr_in *)&to_store;
2026			bzero(to4, sizeof(*to4));
2027			to4->sin_family = AF_INET;
2028			to4->sin_len = sizeof(struct sockaddr_in);
2029			to4->sin_addr.s_addr = iph->ip_dst.s_addr;
2030			to4->sin_port = sh->dest_port;
2031			break;
2032		}
2033#ifdef INET6
2034	case IPV6_VERSION >> 4:
2035		{
2036			/* its IPv6 */
2037			struct ip6_hdr *ip6;
2038			struct sockaddr_in6 *to6;
2039
2040			ip6 = mtod(m, struct ip6_hdr *);
2041			to6 = (struct sockaddr_in6 *)&to_store;
2042			bzero(to6, sizeof(*to6));
2043			to6->sin6_family = AF_INET6;
2044			to6->sin6_len = sizeof(struct sockaddr_in6);
2045			to6->sin6_addr = ip6->ip6_dst;
2046			to6->sin6_port = sh->dest_port;
2047			/* Get the scopes in properly to the sin6 addr's */
2048			/* we probably don't need these operations */
2049			(void)sa6_recoverscope(to6);
2050			sa6_embedscope(to6, MODULE_GLOBAL(ip6_use_defzone));
2051			break;
2052		}
2053#endif
2054	default:
2055		/* TSNH */
2056		break;
2057	}
2058	if (sh->v_tag) {
2059		/* we only go down this path if vtag is non-zero */
2060		retval = sctp_findassoc_by_vtag(from, to, ntohl(sh->v_tag),
2061		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2062		if (retval) {
2063			return (retval);
2064		}
2065	}
2066	find_tcp_pool = 0;
2067	if ((ch->chunk_type != SCTP_INITIATION) &&
2068	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2069	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2070	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2071		/* Other chunk types go to the tcp pool. */
2072		find_tcp_pool = 1;
2073	}
2074	if (inp_p) {
2075		retval = sctp_findassociation_addr_sa(to, from, inp_p, netp,
2076		    find_tcp_pool, vrf_id);
2077		inp = *inp_p;
2078	} else {
2079		retval = sctp_findassociation_addr_sa(to, from, &inp, netp,
2080		    find_tcp_pool, vrf_id);
2081	}
2082	SCTPDBG(SCTP_DEBUG_PCB1, "retval:%p inp:%p\n", retval, inp);
2083	if (retval == NULL && inp) {
2084		/* Found a EP but not this address */
2085		if ((ch->chunk_type == SCTP_INITIATION) ||
2086		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2087			/*-
2088			 * special hook, we do NOT return linp or an
2089			 * association that is linked to an existing
2090			 * association that is under the TCP pool (i.e. no
2091			 * listener exists). The endpoint finding routine
2092			 * will always find a listner before examining the
2093			 * TCP pool.
2094			 */
2095			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2096				if (inp_p) {
2097					*inp_p = NULL;
2098				}
2099				return (NULL);
2100			}
2101			retval = sctp_findassociation_special_addr(m, iphlen,
2102			    offset, sh, &inp, netp, to);
2103			if (inp_p != NULL) {
2104				*inp_p = inp;
2105			}
2106		}
2107	}
2108	SCTPDBG(SCTP_DEBUG_PCB1, "retval is %p\n", retval);
2109	return (retval);
2110}
2111
2112/*
2113 * lookup an association by an ASCONF lookup address.
2114 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2115 */
2116struct sctp_tcb *
2117sctp_findassociation_ep_asconf(struct mbuf *m, int iphlen, int offset,
2118    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2119{
2120	struct sctp_tcb *stcb;
2121	struct sockaddr_in *sin;
2122
2123#ifdef INET6
2124	struct sockaddr_in6 *sin6;
2125
2126#endif
2127	struct sockaddr_storage local_store, remote_store;
2128	struct sockaddr *to;
2129	struct ip *iph;
2130
2131#ifdef INET6
2132	struct ip6_hdr *ip6;
2133
2134#endif
2135	struct sctp_paramhdr parm_buf, *phdr;
2136	int ptype;
2137	int zero_address = 0;
2138
2139
2140	memset(&local_store, 0, sizeof(local_store));
2141	memset(&remote_store, 0, sizeof(remote_store));
2142	to = (struct sockaddr *)&local_store;
2143	/* First get the destination address setup too. */
2144	iph = mtod(m, struct ip *);
2145	switch (iph->ip_v) {
2146	case IPVERSION:
2147		/* its IPv4 */
2148		sin = (struct sockaddr_in *)&local_store;
2149		sin->sin_family = AF_INET;
2150		sin->sin_len = sizeof(*sin);
2151		sin->sin_port = sh->dest_port;
2152		sin->sin_addr.s_addr = iph->ip_dst.s_addr;
2153		break;
2154#ifdef INET6
2155	case IPV6_VERSION >> 4:
2156		/* its IPv6 */
2157		ip6 = mtod(m, struct ip6_hdr *);
2158		sin6 = (struct sockaddr_in6 *)&local_store;
2159		sin6->sin6_family = AF_INET6;
2160		sin6->sin6_len = sizeof(*sin6);
2161		sin6->sin6_port = sh->dest_port;
2162		sin6->sin6_addr = ip6->ip6_dst;
2163		break;
2164#endif
2165	default:
2166		return NULL;
2167	}
2168
2169	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2170	    &parm_buf, sizeof(struct sctp_paramhdr));
2171	if (phdr == NULL) {
2172		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2173		    __FUNCTION__);
2174		return NULL;
2175	}
2176	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2177	/* get the correlation address */
2178	switch (ptype) {
2179#ifdef INET6
2180	case SCTP_IPV6_ADDRESS:
2181		{
2182			/* ipv6 address param */
2183			struct sctp_ipv6addr_param *p6, p6_buf;
2184
2185			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2186				return NULL;
2187			}
2188			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2189			    offset + sizeof(struct sctp_asconf_chunk),
2190			    &p6_buf.ph, sizeof(*p6));
2191			if (p6 == NULL) {
2192				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2193				    __FUNCTION__);
2194				return (NULL);
2195			}
2196			sin6 = (struct sockaddr_in6 *)&remote_store;
2197			sin6->sin6_family = AF_INET6;
2198			sin6->sin6_len = sizeof(*sin6);
2199			sin6->sin6_port = sh->src_port;
2200			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2201			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2202				zero_address = 1;
2203			break;
2204		}
2205#endif
2206	case SCTP_IPV4_ADDRESS:
2207		{
2208			/* ipv4 address param */
2209			struct sctp_ipv4addr_param *p4, p4_buf;
2210
2211			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2212				return NULL;
2213			}
2214			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2215			    offset + sizeof(struct sctp_asconf_chunk),
2216			    &p4_buf.ph, sizeof(*p4));
2217			if (p4 == NULL) {
2218				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2219				    __FUNCTION__);
2220				return (NULL);
2221			}
2222			sin = (struct sockaddr_in *)&remote_store;
2223			sin->sin_family = AF_INET;
2224			sin->sin_len = sizeof(*sin);
2225			sin->sin_port = sh->src_port;
2226			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2227			if (sin->sin_addr.s_addr == INADDR_ANY)
2228				zero_address = 1;
2229			break;
2230		}
2231	default:
2232		/* invalid address param type */
2233		return NULL;
2234	}
2235
2236	if (zero_address) {
2237		stcb = sctp_findassoc_by_vtag(NULL, to, ntohl(sh->v_tag), inp_p,
2238		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2239		/*
2240		 * printf("findassociation_ep_asconf: zero lookup address
2241		 * finds stcb 0x%x\n", (uint32_t)stcb);
2242		 */
2243	} else {
2244		stcb = sctp_findassociation_ep_addr(inp_p,
2245		    (struct sockaddr *)&remote_store, netp,
2246		    to, NULL);
2247	}
2248	return (stcb);
2249}
2250
2251
2252/*
2253 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2254 * addresses. This way if we don't get a bind we by default pick a ephemeral
2255 * port with all addresses bound.
2256 */
2257int
2258sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2259{
2260	/*
2261	 * we get called when a new endpoint starts up. We need to allocate
2262	 * the sctp_inpcb structure from the zone and init it. Mark it as
2263	 * unbound and find a port that we can use as an ephemeral with
2264	 * INADDR_ANY. If the user binds later no problem we can then add in
2265	 * the specific addresses. And setup the default parameters for the
2266	 * EP.
2267	 */
2268	int i, error;
2269	struct sctp_inpcb *inp;
2270	struct sctp_pcb *m;
2271	struct timeval time;
2272	sctp_sharedkey_t *null_key;
2273
2274	error = 0;
2275
2276	SCTP_INP_INFO_WLOCK();
2277	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2278	if (inp == NULL) {
2279		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2280		SCTP_INP_INFO_WUNLOCK();
2281		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2282		return (ENOBUFS);
2283	}
2284	/* zap it */
2285	bzero(inp, sizeof(*inp));
2286
2287	/* bump generations */
2288	/* setup socket pointers */
2289	inp->sctp_socket = so;
2290	inp->ip_inp.inp.inp_socket = so;
2291	inp->sctp_associd_counter = 1;
2292	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2293	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2294	/* init the small hash table we use to track asocid <-> tcb */
2295	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2296	if (inp->sctp_asocidhash == NULL) {
2297		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2298		SCTP_INP_INFO_WUNLOCK();
2299		return (ENOBUFS);
2300	}
2301#ifdef IPSEC
2302	{
2303		struct inpcbpolicy *pcb_sp = NULL;
2304
2305		error = ipsec_init_policy(so, &pcb_sp);
2306		/* Arrange to share the policy */
2307		inp->ip_inp.inp.inp_sp = pcb_sp;
2308		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2309	}
2310	if (error != 0) {
2311		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2312		SCTP_INP_INFO_WUNLOCK();
2313		return error;
2314	}
2315#endif				/* IPSEC */
2316	SCTP_INCR_EP_COUNT();
2317	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2318	SCTP_INP_INFO_WUNLOCK();
2319
2320	so->so_pcb = (caddr_t)inp;
2321
2322	if ((SCTP_SO_TYPE(so) == SOCK_DGRAM) ||
2323	    (SCTP_SO_TYPE(so) == SOCK_SEQPACKET)) {
2324		/* UDP style socket */
2325		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2326		    SCTP_PCB_FLAGS_UNBOUND);
2327		/* Be sure it is NON-BLOCKING IO for UDP */
2328		/* SCTP_SET_SO_NBIO(so); */
2329	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2330		/* TCP style socket */
2331		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2332		    SCTP_PCB_FLAGS_UNBOUND);
2333		/* Be sure we have blocking IO by default */
2334		SCTP_CLEAR_SO_NBIO(so);
2335	} else {
2336		/*
2337		 * unsupported socket type (RAW, etc)- in case we missed it
2338		 * in protosw
2339		 */
2340		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2341		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2342		return (EOPNOTSUPP);
2343	}
2344	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2345		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2346		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2347	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2348		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2349		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2350	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2351		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2352		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2353	}
2354	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2355	    &inp->sctp_hashmark);
2356	if (inp->sctp_tcbhash == NULL) {
2357		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2358		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2359		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2360		return (ENOBUFS);
2361	}
2362	inp->def_vrf_id = vrf_id;
2363
2364	SCTP_INP_INFO_WLOCK();
2365	SCTP_INP_LOCK_INIT(inp);
2366	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2367	SCTP_INP_READ_INIT(inp);
2368	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2369	/* lock the new ep */
2370	SCTP_INP_WLOCK(inp);
2371
2372	/* add it to the info area */
2373	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2374	SCTP_INP_INFO_WUNLOCK();
2375
2376	TAILQ_INIT(&inp->read_queue);
2377	LIST_INIT(&inp->sctp_addr_list);
2378
2379	LIST_INIT(&inp->sctp_asoc_list);
2380
2381#ifdef SCTP_TRACK_FREED_ASOCS
2382	/* TEMP CODE */
2383	LIST_INIT(&inp->sctp_asoc_free_list);
2384#endif
2385	/* Init the timer structure for signature change */
2386	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2387	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2388
2389	/* now init the actual endpoint default data */
2390	m = &inp->sctp_ep;
2391
2392	/* setup the base timeout information */
2393	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2394	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2395	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2396	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2397	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2398	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2399	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2400	/* all max/min max are in ms */
2401	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2402	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2403	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2404	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2405	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2406
2407	m->max_open_streams_intome = MAX_SCTP_STREAMS;
2408
2409	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2410	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2411	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2412	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2413	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2414	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2415	if ((SCTP_BASE_SYSCTL(sctp_default_cc_module) >= SCTP_CC_RFC2581) &&
2416	    (SCTP_BASE_SYSCTL(sctp_default_cc_module) <= SCTP_CC_HTCP)) {
2417		m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2418	} else {
2419		/* sysctl done with invalid value, set to 2581 */
2420		m->sctp_default_cc_module = SCTP_CC_RFC2581;
2421	}
2422	/* number of streams to pre-open on a association */
2423	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2424
2425	/* Add adaptation cookie */
2426	m->adaptation_layer_indicator = 0x504C5253;
2427
2428	/* seed random number generator */
2429	m->random_counter = 1;
2430	m->store_at = SCTP_SIGNATURE_SIZE;
2431	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2432	sctp_fill_random_store(m);
2433
2434	/* Minimum cookie size */
2435	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2436	    sizeof(struct sctp_state_cookie);
2437	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2438
2439	/* Setup the initial secret */
2440	(void)SCTP_GETTIME_TIMEVAL(&time);
2441	m->time_of_secret_change = time.tv_sec;
2442
2443	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2444		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2445	}
2446	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2447
2448	/* How long is a cookie good for ? */
2449	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2450	/*
2451	 * Initialize authentication parameters
2452	 */
2453	m->local_hmacs = sctp_default_supported_hmaclist();
2454	m->local_auth_chunks = sctp_alloc_chunklist();
2455	sctp_auth_set_default_chunks(m->local_auth_chunks);
2456	LIST_INIT(&m->shared_keys);
2457	/* add default NULL key as key id 0 */
2458	null_key = sctp_alloc_sharedkey();
2459	sctp_insert_sharedkey(&m->shared_keys, null_key);
2460	SCTP_INP_WUNLOCK(inp);
2461#ifdef SCTP_LOG_CLOSING
2462	sctp_log_closing(inp, NULL, 12);
2463#endif
2464	return (error);
2465}
2466
2467
2468void
2469sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2470    struct sctp_tcb *stcb)
2471{
2472	struct sctp_nets *net;
2473	uint16_t lport, rport;
2474	struct sctppcbhead *head;
2475	struct sctp_laddr *laddr, *oladdr;
2476
2477	atomic_add_int(&stcb->asoc.refcnt, 1);
2478	SCTP_TCB_UNLOCK(stcb);
2479	SCTP_INP_INFO_WLOCK();
2480	SCTP_INP_WLOCK(old_inp);
2481	SCTP_INP_WLOCK(new_inp);
2482	SCTP_TCB_LOCK(stcb);
2483	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2484
2485	new_inp->sctp_ep.time_of_secret_change =
2486	    old_inp->sctp_ep.time_of_secret_change;
2487	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2488	    sizeof(old_inp->sctp_ep.secret_key));
2489	new_inp->sctp_ep.current_secret_number =
2490	    old_inp->sctp_ep.current_secret_number;
2491	new_inp->sctp_ep.last_secret_number =
2492	    old_inp->sctp_ep.last_secret_number;
2493	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2494
2495	/* make it so new data pours into the new socket */
2496	stcb->sctp_socket = new_inp->sctp_socket;
2497	stcb->sctp_ep = new_inp;
2498
2499	/* Copy the port across */
2500	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2501	rport = stcb->rport;
2502	/* Pull the tcb from the old association */
2503	LIST_REMOVE(stcb, sctp_tcbhash);
2504	LIST_REMOVE(stcb, sctp_tcblist);
2505	if (stcb->asoc.in_asocid_hash) {
2506		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2507	}
2508	/* Now insert the new_inp into the TCP connected hash */
2509	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport),
2510	    SCTP_BASE_INFO(hashtcpmark))];
2511
2512	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2513	/* Its safe to access */
2514	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2515
2516	/* Now move the tcb into the endpoint list */
2517	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2518	/*
2519	 * Question, do we even need to worry about the ep-hash since we
2520	 * only have one connection? Probably not :> so lets get rid of it
2521	 * and not suck up any kernel memory in that.
2522	 */
2523	if (stcb->asoc.in_asocid_hash) {
2524		struct sctpasochead *lhd;
2525
2526		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2527		    new_inp->hashasocidmark)];
2528		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2529	}
2530	/* Ok. Let's restart timer. */
2531	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2532		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2533		    stcb, net);
2534	}
2535
2536	SCTP_INP_INFO_WUNLOCK();
2537	if (new_inp->sctp_tcbhash != NULL) {
2538		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2539		new_inp->sctp_tcbhash = NULL;
2540	}
2541	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2542		/* Subset bound, so copy in the laddr list from the old_inp */
2543		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2544			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2545			if (laddr == NULL) {
2546				/*
2547				 * Gak, what can we do? This assoc is really
2548				 * HOSED. We probably should send an abort
2549				 * here.
2550				 */
2551				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2552				continue;
2553			}
2554			SCTP_INCR_LADDR_COUNT();
2555			bzero(laddr, sizeof(*laddr));
2556			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2557			laddr->ifa = oladdr->ifa;
2558			atomic_add_int(&laddr->ifa->refcount, 1);
2559			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2560			    sctp_nxt_addr);
2561			new_inp->laddr_count++;
2562		}
2563	}
2564	/*
2565	 * Now any running timers need to be adjusted since we really don't
2566	 * care if they are running or not just blast in the new_inp into
2567	 * all of them.
2568	 */
2569
2570	stcb->asoc.hb_timer.ep = (void *)new_inp;
2571	stcb->asoc.dack_timer.ep = (void *)new_inp;
2572	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2573	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2574	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2575	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2576	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2577	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2578	/* now what about the nets? */
2579	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2580		net->pmtu_timer.ep = (void *)new_inp;
2581		net->rxt_timer.ep = (void *)new_inp;
2582		net->fr_timer.ep = (void *)new_inp;
2583	}
2584	SCTP_INP_WUNLOCK(new_inp);
2585	SCTP_INP_WUNLOCK(old_inp);
2586}
2587
2588
2589
2590
2591/* sctp_ifap is used to bypass normal local address validation checks */
2592int
2593sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2594    struct sctp_ifa *sctp_ifap, struct thread *p)
2595{
2596	/* bind a ep to a socket address */
2597	struct sctppcbhead *head;
2598	struct sctp_inpcb *inp, *inp_tmp;
2599	struct inpcb *ip_inp;
2600	int port_reuse_active = 0;
2601	int bindall;
2602	uint16_t lport;
2603	int error;
2604	uint32_t vrf_id;
2605
2606	lport = 0;
2607	error = 0;
2608	bindall = 1;
2609	inp = (struct sctp_inpcb *)so->so_pcb;
2610	ip_inp = (struct inpcb *)so->so_pcb;
2611#ifdef SCTP_DEBUG
2612	if (addr) {
2613		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port:%d\n",
2614		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2615		SCTPDBG(SCTP_DEBUG_PCB1, "Addr :");
2616		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2617	}
2618#endif
2619	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2620		/* already did a bind, subsequent binds NOT allowed ! */
2621		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2622		return (EINVAL);
2623	}
2624#ifdef INVARIANTS
2625	if (p == NULL)
2626		panic("null proc/thread");
2627#endif
2628	if (addr != NULL) {
2629		switch (addr->sa_family) {
2630		case AF_INET:
2631			{
2632				struct sockaddr_in *sin;
2633
2634				/* IPV6_V6ONLY socket? */
2635				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2636					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2637					return (EINVAL);
2638				}
2639				if (addr->sa_len != sizeof(*sin)) {
2640					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2641					return (EINVAL);
2642				}
2643				sin = (struct sockaddr_in *)addr;
2644				lport = sin->sin_port;
2645				/*
2646				 * For LOOPBACK the prison_local_ip4() call
2647				 * will transmute the ip address to the
2648				 * proper value.
2649				 */
2650				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2651					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2652					return (error);
2653				}
2654				if (sin->sin_addr.s_addr != INADDR_ANY) {
2655					bindall = 0;
2656				}
2657				break;
2658			}
2659#ifdef INET6
2660		case AF_INET6:
2661			{
2662				/*
2663				 * Only for pure IPv6 Address. (No IPv4
2664				 * Mapped!)
2665				 */
2666				struct sockaddr_in6 *sin6;
2667
2668				sin6 = (struct sockaddr_in6 *)addr;
2669
2670				if (addr->sa_len != sizeof(*sin6)) {
2671					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2672					return (EINVAL);
2673				}
2674				lport = sin6->sin6_port;
2675
2676				/*
2677				 * For LOOPBACK the prison_local_ip6() call
2678				 * will transmute the ipv6 address to the
2679				 * proper value.
2680				 */
2681				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2682				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2683					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2684					return (error);
2685				}
2686				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2687					bindall = 0;
2688					/* KAME hack: embed scopeid */
2689					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2690						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2691						return (EINVAL);
2692					}
2693				}
2694				/* this must be cleared for ifa_ifwithaddr() */
2695				sin6->sin6_scope_id = 0;
2696				break;
2697			}
2698#endif
2699		default:
2700			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2701			return (EAFNOSUPPORT);
2702		}
2703	}
2704	SCTP_INP_INFO_WLOCK();
2705	SCTP_INP_WLOCK(inp);
2706	/* Setup a vrf_id to be the default for the non-bind-all case. */
2707	vrf_id = inp->def_vrf_id;
2708
2709	/* increase our count due to the unlock we do */
2710	SCTP_INP_INCR_REF(inp);
2711	if (lport) {
2712		/*
2713		 * Did the caller specify a port? if so we must see if a ep
2714		 * already has this one bound.
2715		 */
2716		/* got to be root to get at low ports */
2717		if (ntohs(lport) < IPPORT_RESERVED) {
2718			if (p && (error =
2719			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2720			    )) {
2721				SCTP_INP_DECR_REF(inp);
2722				SCTP_INP_WUNLOCK(inp);
2723				SCTP_INP_INFO_WUNLOCK();
2724				return (error);
2725			}
2726		}
2727		if (p == NULL) {
2728			SCTP_INP_DECR_REF(inp);
2729			SCTP_INP_WUNLOCK(inp);
2730			SCTP_INP_INFO_WUNLOCK();
2731			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2732			return (error);
2733		}
2734		SCTP_INP_WUNLOCK(inp);
2735		if (bindall) {
2736			vrf_id = inp->def_vrf_id;
2737			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2738			if (inp_tmp != NULL) {
2739				/*
2740				 * lock guy returned and lower count note
2741				 * that we are not bound so inp_tmp should
2742				 * NEVER be inp. And it is this inp
2743				 * (inp_tmp) that gets the reference bump,
2744				 * so we must lower it.
2745				 */
2746				SCTP_INP_DECR_REF(inp_tmp);
2747				/* unlock info */
2748				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2749				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2750					/*
2751					 * Ok, must be one-2-one and
2752					 * allowing port re-use
2753					 */
2754					port_reuse_active = 1;
2755					goto continue_anyway;
2756				}
2757				SCTP_INP_DECR_REF(inp);
2758				SCTP_INP_INFO_WUNLOCK();
2759				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2760				return (EADDRINUSE);
2761			}
2762		} else {
2763			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2764			if (inp_tmp != NULL) {
2765				/*
2766				 * lock guy returned and lower count note
2767				 * that we are not bound so inp_tmp should
2768				 * NEVER be inp. And it is this inp
2769				 * (inp_tmp) that gets the reference bump,
2770				 * so we must lower it.
2771				 */
2772				SCTP_INP_DECR_REF(inp_tmp);
2773				/* unlock info */
2774				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2775				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2776					/*
2777					 * Ok, must be one-2-one and
2778					 * allowing port re-use
2779					 */
2780					port_reuse_active = 1;
2781					goto continue_anyway;
2782				}
2783				SCTP_INP_DECR_REF(inp);
2784				SCTP_INP_INFO_WUNLOCK();
2785				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2786				return (EADDRINUSE);
2787			}
2788		}
2789continue_anyway:
2790		SCTP_INP_WLOCK(inp);
2791		if (bindall) {
2792			/* verify that no lport is not used by a singleton */
2793			if ((port_reuse_active == 0) &&
2794			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))
2795			    ) {
2796				/* Sorry someone already has this one bound */
2797				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2798				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2799					port_reuse_active = 1;
2800				} else {
2801					SCTP_INP_DECR_REF(inp);
2802					SCTP_INP_WUNLOCK(inp);
2803					SCTP_INP_INFO_WUNLOCK();
2804					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2805					return (EADDRINUSE);
2806				}
2807			}
2808		}
2809	} else {
2810		uint16_t first, last, candidate;
2811		uint16_t count;
2812		int done;
2813
2814		if (ip_inp->inp_flags & INP_HIGHPORT) {
2815			first = MODULE_GLOBAL(ipport_hifirstauto);
2816			last = MODULE_GLOBAL(ipport_hilastauto);
2817		} else if (ip_inp->inp_flags & INP_LOWPORT) {
2818			if (p && (error =
2819			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2820			    )) {
2821				SCTP_INP_DECR_REF(inp);
2822				SCTP_INP_WUNLOCK(inp);
2823				SCTP_INP_INFO_WUNLOCK();
2824				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2825				return (error);
2826			}
2827			first = MODULE_GLOBAL(ipport_lowfirstauto);
2828			last = MODULE_GLOBAL(ipport_lowlastauto);
2829		} else {
2830			first = MODULE_GLOBAL(ipport_firstauto);
2831			last = MODULE_GLOBAL(ipport_lastauto);
2832		}
2833		if (first > last) {
2834			uint16_t temp;
2835
2836			temp = first;
2837			first = last;
2838			last = temp;
2839		}
2840		count = last - first + 1;	/* number of candidates */
2841		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
2842
2843		done = 0;
2844		while (!done) {
2845			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
2846				done = 1;
2847			}
2848			if (!done) {
2849				if (--count == 0) {
2850					SCTP_INP_DECR_REF(inp);
2851					SCTP_INP_WUNLOCK(inp);
2852					SCTP_INP_INFO_WUNLOCK();
2853					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2854					return (EADDRINUSE);
2855				}
2856				if (candidate == last)
2857					candidate = first;
2858				else
2859					candidate = candidate + 1;
2860			}
2861		}
2862		lport = htons(candidate);
2863	}
2864	SCTP_INP_DECR_REF(inp);
2865	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2866	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2867		/*
2868		 * this really should not happen. The guy did a non-blocking
2869		 * bind and then did a close at the same time.
2870		 */
2871		SCTP_INP_WUNLOCK(inp);
2872		SCTP_INP_INFO_WUNLOCK();
2873		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2874		return (EINVAL);
2875	}
2876	/* ok we look clear to give out this port, so lets setup the binding */
2877	if (bindall) {
2878		/* binding to all addresses, so just set in the proper flags */
2879		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
2880		/* set the automatic addr changes from kernel flag */
2881		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
2882			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2883			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2884		} else {
2885			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2886			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2887		}
2888		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
2889			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
2890		} else {
2891			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
2892		}
2893		/*
2894		 * set the automatic mobility_base from kernel flag (by
2895		 * micchie)
2896		 */
2897		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
2898			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
2899			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2900		} else {
2901			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
2902			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2903		}
2904		/*
2905		 * set the automatic mobility_fasthandoff from kernel flag
2906		 * (by micchie)
2907		 */
2908		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
2909			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
2910			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2911		} else {
2912			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
2913			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
2914		}
2915	} else {
2916		/*
2917		 * bind specific, make sure flags is off and add a new
2918		 * address structure to the sctp_addr_list inside the ep
2919		 * structure.
2920		 *
2921		 * We will need to allocate one and insert it at the head. The
2922		 * socketopt call can just insert new addresses in there as
2923		 * well. It will also have to do the embed scope kame hack
2924		 * too (before adding).
2925		 */
2926		struct sctp_ifa *ifa;
2927		struct sockaddr_storage store_sa;
2928
2929		memset(&store_sa, 0, sizeof(store_sa));
2930		if (addr->sa_family == AF_INET) {
2931			struct sockaddr_in *sin;
2932
2933			sin = (struct sockaddr_in *)&store_sa;
2934			memcpy(sin, addr, sizeof(struct sockaddr_in));
2935			sin->sin_port = 0;
2936		} else if (addr->sa_family == AF_INET6) {
2937			struct sockaddr_in6 *sin6;
2938
2939			sin6 = (struct sockaddr_in6 *)&store_sa;
2940			memcpy(sin6, addr, sizeof(struct sockaddr_in6));
2941			sin6->sin6_port = 0;
2942		}
2943		/*
2944		 * first find the interface with the bound address need to
2945		 * zero out the port to find the address! yuck! can't do
2946		 * this earlier since need port for sctp_pcb_findep()
2947		 */
2948		if (sctp_ifap != NULL)
2949			ifa = sctp_ifap;
2950		else {
2951			/*
2952			 * Note for BSD we hit here always other O/S's will
2953			 * pass things in via the sctp_ifap argument
2954			 * (Panda).
2955			 */
2956			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
2957			    vrf_id, SCTP_ADDR_NOT_LOCKED);
2958		}
2959		if (ifa == NULL) {
2960			/* Can't find an interface with that address */
2961			SCTP_INP_WUNLOCK(inp);
2962			SCTP_INP_INFO_WUNLOCK();
2963			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
2964			return (EADDRNOTAVAIL);
2965		}
2966		if (addr->sa_family == AF_INET6) {
2967			/* GAK, more FIXME IFA lock? */
2968			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
2969				/* Can't bind a non-existent addr. */
2970				SCTP_INP_WUNLOCK(inp);
2971				SCTP_INP_INFO_WUNLOCK();
2972				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2973				return (EINVAL);
2974			}
2975		}
2976		/* we're not bound all */
2977		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
2978		/* allow bindx() to send ASCONF's for binding changes */
2979		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2980		/* clear automatic addr changes from kernel flag */
2981		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2982
2983		/* add this address to the endpoint list */
2984		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
2985		if (error != 0) {
2986			SCTP_INP_WUNLOCK(inp);
2987			SCTP_INP_INFO_WUNLOCK();
2988			return (error);
2989		}
2990		inp->laddr_count++;
2991	}
2992	/* find the bucket */
2993	if (port_reuse_active) {
2994		/* Put it into tcp 1-2-1 hash */
2995		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport),
2996		    SCTP_BASE_INFO(hashtcpmark))];
2997		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
2998	} else {
2999		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
3000		    SCTP_BASE_INFO(hashmark))];
3001	}
3002	/* put it in the bucket */
3003	LIST_INSERT_HEAD(head, inp, sctp_hash);
3004	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3005	    head, ntohs(lport), port_reuse_active);
3006	/* set in the port */
3007	inp->sctp_lport = lport;
3008
3009	/* turn off just the unbound flag */
3010	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3011	SCTP_INP_WUNLOCK(inp);
3012	SCTP_INP_INFO_WUNLOCK();
3013	return (0);
3014}
3015
3016
3017static void
3018sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3019{
3020	struct sctp_iterator *it, *nit;
3021
3022	/*
3023	 * We enter with the only the ITERATOR_LOCK in place and a write
3024	 * lock on the inp_info stuff.
3025	 */
3026	it = sctp_it_ctl.cur_it;
3027	if (it && (it->vn != curvnet)) {
3028		/* Its not looking at our VNET */
3029		return;
3030	}
3031	if (it && (it->inp == inp)) {
3032		/*
3033		 * This is tricky and we hold the iterator lock, but when it
3034		 * returns and gets the lock (when we release it) the
3035		 * iterator will try to operate on inp. We need to stop that
3036		 * from happening. But of course the iterator has a
3037		 * reference on the stcb and inp. We can mark it and it will
3038		 * stop.
3039		 *
3040		 * If its a single iterator situation, we set the end iterator
3041		 * flag. Otherwise we set the iterator to go to the next
3042		 * inp.
3043		 *
3044		 */
3045		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3046			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3047		} else {
3048			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3049		}
3050	}
3051	/*
3052	 * Now go through and remove any single reference to our inp that
3053	 * may be still pending on the list
3054	 */
3055	SCTP_IPI_ITERATOR_WQ_LOCK();
3056	it = TAILQ_FIRST(&sctp_it_ctl.iteratorhead);
3057	while (it) {
3058		nit = TAILQ_NEXT(it, sctp_nxt_itr);
3059		if (it->vn != curvnet) {
3060			it = nit;
3061			continue;
3062		}
3063		if (it->inp == inp) {
3064			/* This one points to me is it inp specific? */
3065			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3066				/* Remove and free this one */
3067				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3068				    it, sctp_nxt_itr);
3069				if (it->function_atend != NULL) {
3070					(*it->function_atend) (it->pointer, it->val);
3071				}
3072				SCTP_FREE(it, SCTP_M_ITER);
3073			} else {
3074				it->inp = LIST_NEXT(it->inp, sctp_list);
3075			}
3076		}
3077		it = nit;
3078	}
3079	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3080}
3081
3082/* release sctp_inpcb unbind the port */
3083void
3084sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3085{
3086	/*
3087	 * Here we free a endpoint. We must find it (if it is in the Hash
3088	 * table) and remove it from there. Then we must also find it in the
3089	 * overall list and remove it from there. After all removals are
3090	 * complete then any timer has to be stopped. Then start the actual
3091	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3092	 * all associations. d) finally the ep itself.
3093	 */
3094	struct sctp_pcb *m;
3095	struct sctp_tcb *asoc, *nasoc;
3096	struct sctp_laddr *laddr, *nladdr;
3097	struct inpcb *ip_pcb;
3098	struct socket *so;
3099	int being_refed = 0;
3100	struct sctp_queued_to_read *sq;
3101
3102
3103	int cnt;
3104	sctp_sharedkey_t *shared_key;
3105
3106
3107#ifdef SCTP_LOG_CLOSING
3108	sctp_log_closing(inp, NULL, 0);
3109#endif
3110	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3111		/*
3112		 * Once we are in we can remove the flag from = 1 is only
3113		 * passed from the actual closing routines that are called
3114		 * via the sockets layer.
3115		 */
3116		SCTP_ITERATOR_LOCK();
3117		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3118		/* socket is gone, so no more wakeups allowed */
3119		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3120		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3121		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3122
3123		/* mark any iterators on the list or being processed */
3124		sctp_iterator_inp_being_freed(inp);
3125		SCTP_ITERATOR_UNLOCK();
3126	}
3127	so = inp->sctp_socket;
3128	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3129		/* been here before.. eeks.. get out of here */
3130		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3131#ifdef SCTP_LOG_CLOSING
3132		sctp_log_closing(inp, NULL, 1);
3133#endif
3134		return;
3135	}
3136	SCTP_ASOC_CREATE_LOCK(inp);
3137	SCTP_INP_INFO_WLOCK();
3138
3139	SCTP_INP_WLOCK(inp);
3140	/* First time through we have the socket lock, after that no more. */
3141	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3142	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3143
3144	if (inp->control) {
3145		sctp_m_freem(inp->control);
3146		inp->control = NULL;
3147	}
3148	if (inp->pkt) {
3149		sctp_m_freem(inp->pkt);
3150		inp->pkt = NULL;
3151	}
3152	m = &inp->sctp_ep;
3153	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3154					 * here but I will be nice :> (i.e.
3155					 * ip_pcb = ep;) */
3156	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3157		int cnt_in_sd;
3158
3159		cnt_in_sd = 0;
3160		for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
3161		    asoc = nasoc) {
3162			SCTP_TCB_LOCK(asoc);
3163			nasoc = LIST_NEXT(asoc, sctp_tcblist);
3164			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3165				/* Skip guys being freed */
3166				cnt_in_sd++;
3167				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3168					/*
3169					 * Special case - we did not start a
3170					 * kill timer on the asoc due to it
3171					 * was not closed. So go ahead and
3172					 * start it now.
3173					 */
3174					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3175					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3176				}
3177				SCTP_TCB_UNLOCK(asoc);
3178				continue;
3179			}
3180			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3181			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3182			    (asoc->asoc.total_output_queue_size == 0)) {
3183				/*
3184				 * If we have data in queue, we don't want
3185				 * to just free since the app may have done,
3186				 * send()/close or connect/send/close. And
3187				 * it wants the data to get across first.
3188				 */
3189				/* Just abandon things in the front states */
3190				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3191				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3192					cnt_in_sd++;
3193				}
3194				continue;
3195			}
3196			/* Disconnect the socket please */
3197			asoc->sctp_socket = NULL;
3198			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3199			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3200			    (asoc->asoc.control_pdapi) ||
3201			    (asoc->asoc.size_on_all_streams > 0) ||
3202			    (so && (so->so_rcv.sb_cc > 0))
3203			    ) {
3204				/* Left with Data unread */
3205				struct mbuf *op_err;
3206
3207				op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3208				    0, M_DONTWAIT, 1, MT_DATA);
3209				if (op_err) {
3210					/* Fill in the user initiated abort */
3211					struct sctp_paramhdr *ph;
3212					uint32_t *ippp;
3213
3214					SCTP_BUF_LEN(op_err) =
3215					    sizeof(struct sctp_paramhdr) + sizeof(uint32_t);
3216					ph = mtod(op_err,
3217					    struct sctp_paramhdr *);
3218					ph->param_type = htons(
3219					    SCTP_CAUSE_USER_INITIATED_ABT);
3220					ph->param_length = htons(SCTP_BUF_LEN(op_err));
3221					ippp = (uint32_t *) (ph + 1);
3222					*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3);
3223				}
3224				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3225#if defined(SCTP_PANIC_ON_ABORT)
3226				panic("inpcb_free does an abort");
3227#endif
3228				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3229				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3230				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3231				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3232					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3233				}
3234				if (sctp_free_assoc(inp, asoc,
3235				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3236					cnt_in_sd++;
3237				}
3238				continue;
3239			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3240				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3241				    (asoc->asoc.stream_queue_cnt == 0)
3242			    ) {
3243				if (asoc->asoc.locked_on_sending) {
3244					goto abort_anyway;
3245				}
3246				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3247				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3248					/*
3249					 * there is nothing queued to send,
3250					 * so I send shutdown
3251					 */
3252					sctp_send_shutdown(asoc, asoc->asoc.primary_destination);
3253					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3254					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3255						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3256					}
3257					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3258					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3259					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3260					    asoc->asoc.primary_destination);
3261					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3262					    asoc->asoc.primary_destination);
3263					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3264				}
3265			} else {
3266				/* mark into shutdown pending */
3267				struct sctp_stream_queue_pending *sp;
3268
3269				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3270				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3271				    asoc->asoc.primary_destination);
3272				if (asoc->asoc.locked_on_sending) {
3273					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3274					    sctp_streamhead);
3275					if (sp == NULL) {
3276						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3277						    asoc->asoc.locked_on_sending,
3278						    asoc->asoc.locked_on_sending->stream_no);
3279					} else {
3280						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3281							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3282					}
3283				}
3284				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3285				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3286				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3287					struct mbuf *op_err;
3288
3289			abort_anyway:
3290					op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3291					    0, M_DONTWAIT, 1, MT_DATA);
3292					if (op_err) {
3293						/*
3294						 * Fill in the user
3295						 * initiated abort
3296						 */
3297						struct sctp_paramhdr *ph;
3298						uint32_t *ippp;
3299
3300						SCTP_BUF_LEN(op_err) =
3301						    (sizeof(struct sctp_paramhdr) +
3302						    sizeof(uint32_t));
3303						ph = mtod(op_err,
3304						    struct sctp_paramhdr *);
3305						ph->param_type = htons(
3306						    SCTP_CAUSE_USER_INITIATED_ABT);
3307						ph->param_length = htons(SCTP_BUF_LEN(op_err));
3308						ippp = (uint32_t *) (ph + 1);
3309						*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5);
3310					}
3311					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3312#if defined(SCTP_PANIC_ON_ABORT)
3313					panic("inpcb_free does an abort");
3314#endif
3315
3316					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3317					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3318					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3319					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3320						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3321					}
3322					if (sctp_free_assoc(inp, asoc,
3323					    SCTP_PCBFREE_NOFORCE,
3324					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3325						cnt_in_sd++;
3326					}
3327					continue;
3328				} else {
3329					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3330				}
3331			}
3332			cnt_in_sd++;
3333			SCTP_TCB_UNLOCK(asoc);
3334		}
3335		/* now is there some left in our SHUTDOWN state? */
3336		if (cnt_in_sd) {
3337			SCTP_INP_WUNLOCK(inp);
3338			SCTP_ASOC_CREATE_UNLOCK(inp);
3339			SCTP_INP_INFO_WUNLOCK();
3340#ifdef SCTP_LOG_CLOSING
3341			sctp_log_closing(inp, NULL, 2);
3342#endif
3343			inp->sctp_socket = NULL;
3344			return;
3345		}
3346	}
3347	inp->sctp_socket = NULL;
3348	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3349	    SCTP_PCB_FLAGS_UNBOUND) {
3350		/*
3351		 * ok, this guy has been bound. It's port is somewhere in
3352		 * the SCTP_BASE_INFO(hash table). Remove it!
3353		 */
3354		LIST_REMOVE(inp, sctp_hash);
3355		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3356	}
3357	/*
3358	 * If there is a timer running to kill us, forget it, since it may
3359	 * have a contest on the INP lock.. which would cause us to die ...
3360	 */
3361	cnt = 0;
3362	for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
3363	    asoc = nasoc) {
3364		nasoc = LIST_NEXT(asoc, sctp_tcblist);
3365		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3366			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3367				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3368				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3369			}
3370			cnt++;
3371			continue;
3372		}
3373		/* Free associations that are NOT killing us */
3374		SCTP_TCB_LOCK(asoc);
3375		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3376		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3377			struct mbuf *op_err;
3378			uint32_t *ippp;
3379
3380			op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
3381			    0, M_DONTWAIT, 1, MT_DATA);
3382			if (op_err) {
3383				/* Fill in the user initiated abort */
3384				struct sctp_paramhdr *ph;
3385
3386				SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) +
3387				    sizeof(uint32_t));
3388				ph = mtod(op_err, struct sctp_paramhdr *);
3389				ph->param_type = htons(
3390				    SCTP_CAUSE_USER_INITIATED_ABT);
3391				ph->param_length = htons(SCTP_BUF_LEN(op_err));
3392				ippp = (uint32_t *) (ph + 1);
3393				*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7);
3394
3395			}
3396			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3397#if defined(SCTP_PANIC_ON_ABORT)
3398			panic("inpcb_free does an abort");
3399#endif
3400			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3401			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3402		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3403			cnt++;
3404			SCTP_TCB_UNLOCK(asoc);
3405			continue;
3406		}
3407		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3408		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3409			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3410		}
3411		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3412			cnt++;
3413		}
3414	}
3415	if (cnt) {
3416		/* Ok we have someone out there that will kill us */
3417		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3418		SCTP_INP_WUNLOCK(inp);
3419		SCTP_ASOC_CREATE_UNLOCK(inp);
3420		SCTP_INP_INFO_WUNLOCK();
3421#ifdef SCTP_LOG_CLOSING
3422		sctp_log_closing(inp, NULL, 3);
3423#endif
3424		return;
3425	}
3426	if (SCTP_INP_LOCK_CONTENDED(inp))
3427		being_refed++;
3428	if (SCTP_INP_READ_CONTENDED(inp))
3429		being_refed++;
3430	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3431		being_refed++;
3432
3433	if ((inp->refcount) ||
3434	    (being_refed) ||
3435	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3436		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3437		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3438		SCTP_INP_WUNLOCK(inp);
3439		SCTP_ASOC_CREATE_UNLOCK(inp);
3440		SCTP_INP_INFO_WUNLOCK();
3441#ifdef SCTP_LOG_CLOSING
3442		sctp_log_closing(inp, NULL, 4);
3443#endif
3444		return;
3445	}
3446	(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3447	inp->sctp_ep.signature_change.type = 0;
3448	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3449
3450#ifdef SCTP_LOG_CLOSING
3451	sctp_log_closing(inp, NULL, 5);
3452#endif
3453
3454	(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3455	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NONE;
3456	/* Clear the read queue */
3457	if ((inp->sctp_asocidhash) != NULL) {
3458		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3459		inp->sctp_asocidhash = NULL;
3460	}
3461	/* sa_ignore FREED_MEMORY */
3462	while ((sq = TAILQ_FIRST(&inp->read_queue)) != NULL) {
3463		/* Its only abandoned if it had data left */
3464		if (sq->length)
3465			SCTP_STAT_INCR(sctps_left_abandon);
3466
3467		TAILQ_REMOVE(&inp->read_queue, sq, next);
3468		sctp_free_remote_addr(sq->whoFrom);
3469		if (so)
3470			so->so_rcv.sb_cc -= sq->length;
3471		if (sq->data) {
3472			sctp_m_freem(sq->data);
3473			sq->data = NULL;
3474		}
3475		/*
3476		 * no need to free the net count, since at this point all
3477		 * assoc's are gone.
3478		 */
3479		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3480		SCTP_DECR_READQ_COUNT();
3481	}
3482	/* Now the sctp_pcb things */
3483	/*
3484	 * free each asoc if it is not already closed/free. we can't use the
3485	 * macro here since le_next will get freed as part of the
3486	 * sctp_free_assoc() call.
3487	 */
3488	cnt = 0;
3489	if (so) {
3490#ifdef IPSEC
3491		ipsec_delete_pcbpolicy(ip_pcb);
3492#endif				/* IPSEC */
3493
3494		/* Unlocks not needed since the socket is gone now */
3495	}
3496	if (ip_pcb->inp_options) {
3497		(void)sctp_m_free(ip_pcb->inp_options);
3498		ip_pcb->inp_options = 0;
3499	}
3500	if (ip_pcb->inp_moptions) {
3501		inp_freemoptions(ip_pcb->inp_moptions);
3502		ip_pcb->inp_moptions = 0;
3503	}
3504#ifdef INET6
3505	if (ip_pcb->inp_vflag & INP_IPV6) {
3506		struct in6pcb *in6p;
3507
3508		in6p = (struct in6pcb *)inp;
3509		ip6_freepcbopts(in6p->in6p_outputopts);
3510	}
3511#endif				/* INET6 */
3512	ip_pcb->inp_vflag = 0;
3513	/* free up authentication fields */
3514	if (inp->sctp_ep.local_auth_chunks != NULL)
3515		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3516	if (inp->sctp_ep.local_hmacs != NULL)
3517		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3518
3519	shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
3520	while (shared_key) {
3521		LIST_REMOVE(shared_key, next);
3522		sctp_free_sharedkey(shared_key);
3523		/* sa_ignore FREED_MEMORY */
3524		shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
3525	}
3526
3527	LIST_REMOVE(inp, sctp_list);
3528
3529	/*
3530	 * if we have an address list the following will free the list of
3531	 * ifaddr's that are set into this ep. Again macro limitations here,
3532	 * since the LIST_FOREACH could be a bad idea.
3533	 */
3534	for ((laddr = LIST_FIRST(&inp->sctp_addr_list)); laddr != NULL;
3535	    laddr = nladdr) {
3536		nladdr = LIST_NEXT(laddr, sctp_nxt_addr);
3537		sctp_remove_laddr(laddr);
3538	}
3539
3540#ifdef SCTP_TRACK_FREED_ASOCS
3541	/* TEMP CODE */
3542	for ((asoc = LIST_FIRST(&inp->sctp_asoc_free_list)); asoc != NULL;
3543	    asoc = nasoc) {
3544		nasoc = LIST_NEXT(asoc, sctp_tcblist);
3545		LIST_REMOVE(asoc, sctp_tcblist);
3546		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3547		SCTP_DECR_ASOC_COUNT();
3548	}
3549	/* *** END TEMP CODE *** */
3550#endif
3551	/* Now lets see about freeing the EP hash table. */
3552	if (inp->sctp_tcbhash != NULL) {
3553		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3554		inp->sctp_tcbhash = NULL;
3555	}
3556	/* Now we must put the ep memory back into the zone pool */
3557	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3558	SCTP_INP_LOCK_DESTROY(inp);
3559	SCTP_INP_READ_DESTROY(inp);
3560	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3561	SCTP_INP_INFO_WUNLOCK();
3562	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3563	SCTP_DECR_EP_COUNT();
3564}
3565
3566
3567struct sctp_nets *
3568sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3569{
3570	struct sctp_nets *net;
3571
3572	/* locate the address */
3573	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3574		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3575			return (net);
3576	}
3577	return (NULL);
3578}
3579
3580
3581int
3582sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3583{
3584	struct sctp_ifa *sctp_ifa;
3585
3586	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3587	if (sctp_ifa) {
3588		return (1);
3589	} else {
3590		return (0);
3591	}
3592}
3593
3594/*
3595 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3596 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3597 * stats of stuff.
3598 */
3599int
3600sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3601    int set_scope, int from)
3602{
3603	/*
3604	 * The following is redundant to the same lines in the
3605	 * sctp_aloc_assoc() but is needed since other's call the add
3606	 * address function
3607	 */
3608	struct sctp_nets *net, *netfirst;
3609	int addr_inscope;
3610
3611	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3612	    from);
3613	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3614
3615	netfirst = sctp_findnet(stcb, newaddr);
3616	if (netfirst) {
3617		/*
3618		 * Lie and return ok, we don't want to make the association
3619		 * go away for this behavior. It will happen in the TCP
3620		 * model in a connected socket. It does not reach the hash
3621		 * table until after the association is built so it can't be
3622		 * found. Mark as reachable, since the initial creation will
3623		 * have been cleared and the NOT_IN_ASSOC flag will have
3624		 * been added... and we don't want to end up removing it
3625		 * back out.
3626		 */
3627		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3628			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3629			    SCTP_ADDR_UNCONFIRMED);
3630		} else {
3631			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3632		}
3633
3634		return (0);
3635	}
3636	addr_inscope = 1;
3637	if (newaddr->sa_family == AF_INET) {
3638		struct sockaddr_in *sin;
3639
3640		sin = (struct sockaddr_in *)newaddr;
3641		if (sin->sin_addr.s_addr == 0) {
3642			/* Invalid address */
3643			return (-1);
3644		}
3645		/* zero out the bzero area */
3646		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3647
3648		/* assure len is set */
3649		sin->sin_len = sizeof(struct sockaddr_in);
3650		if (set_scope) {
3651#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3652			stcb->ipv4_local_scope = 1;
3653#else
3654			if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3655				stcb->asoc.ipv4_local_scope = 1;
3656			}
3657#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3658		} else {
3659			/* Validate the address is in scope */
3660			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3661			    (stcb->asoc.ipv4_local_scope == 0)) {
3662				addr_inscope = 0;
3663			}
3664		}
3665#ifdef INET6
3666	} else if (newaddr->sa_family == AF_INET6) {
3667		struct sockaddr_in6 *sin6;
3668
3669		sin6 = (struct sockaddr_in6 *)newaddr;
3670		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3671			/* Invalid address */
3672			return (-1);
3673		}
3674		/* assure len is set */
3675		sin6->sin6_len = sizeof(struct sockaddr_in6);
3676		if (set_scope) {
3677			if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3678				stcb->asoc.loopback_scope = 1;
3679				stcb->asoc.local_scope = 0;
3680				stcb->asoc.ipv4_local_scope = 1;
3681				stcb->asoc.site_scope = 1;
3682			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3683				/*
3684				 * If the new destination is a LINK_LOCAL we
3685				 * must have common site scope. Don't set
3686				 * the local scope since we may not share
3687				 * all links, only loopback can do this.
3688				 * Links on the local network would also be
3689				 * on our private network for v4 too.
3690				 */
3691				stcb->asoc.ipv4_local_scope = 1;
3692				stcb->asoc.site_scope = 1;
3693			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3694				/*
3695				 * If the new destination is SITE_LOCAL then
3696				 * we must have site scope in common.
3697				 */
3698				stcb->asoc.site_scope = 1;
3699			}
3700		} else {
3701			/* Validate the address is in scope */
3702			if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3703			    (stcb->asoc.loopback_scope == 0)) {
3704				addr_inscope = 0;
3705			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3706			    (stcb->asoc.local_scope == 0)) {
3707				addr_inscope = 0;
3708			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3709			    (stcb->asoc.site_scope == 0)) {
3710				addr_inscope = 0;
3711			}
3712		}
3713#endif
3714	} else {
3715		/* not supported family type */
3716		return (-1);
3717	}
3718	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3719	if (net == NULL) {
3720		return (-1);
3721	}
3722	SCTP_INCR_RADDR_COUNT();
3723	bzero(net, sizeof(*net));
3724	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3725	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3726	if (newaddr->sa_family == AF_INET) {
3727		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3728	} else if (newaddr->sa_family == AF_INET6) {
3729		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3730	}
3731	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3732	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3733		stcb->asoc.loopback_scope = 1;
3734		stcb->asoc.ipv4_local_scope = 1;
3735		stcb->asoc.local_scope = 0;
3736		stcb->asoc.site_scope = 1;
3737		addr_inscope = 1;
3738	}
3739	net->failure_threshold = stcb->asoc.def_net_failure;
3740	if (addr_inscope == 0) {
3741		net->dest_state = (SCTP_ADDR_REACHABLE |
3742		    SCTP_ADDR_OUT_OF_SCOPE);
3743	} else {
3744		if (from == SCTP_ADDR_IS_CONFIRMED)
3745			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3746			net->dest_state = SCTP_ADDR_REACHABLE;
3747		else
3748			net->dest_state = SCTP_ADDR_REACHABLE |
3749			    SCTP_ADDR_UNCONFIRMED;
3750	}
3751	/*
3752	 * We set this to 0, the timer code knows that this means its an
3753	 * initial value
3754	 */
3755	net->RTO = 0;
3756	net->RTO_measured = 0;
3757	stcb->asoc.numnets++;
3758	*(&net->ref_count) = 1;
3759	net->tos_flowlabel = 0;
3760	if (SCTP_BASE_SYSCTL(sctp_udp_tunneling_for_client_enable)) {
3761		net->port = htons(SCTP_BASE_SYSCTL(sctp_udp_tunneling_port));
3762	} else {
3763		net->port = 0;
3764	}
3765#ifdef INET
3766	if (newaddr->sa_family == AF_INET)
3767		net->tos_flowlabel = stcb->asoc.default_tos;
3768#endif
3769#ifdef INET6
3770	if (newaddr->sa_family == AF_INET6)
3771		net->tos_flowlabel = stcb->asoc.default_flowlabel;
3772#endif
3773	/* Init the timer structure */
3774	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3775	SCTP_OS_TIMER_INIT(&net->fr_timer.timer);
3776	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3777
3778	/* Now generate a route for this guy */
3779#ifdef INET6
3780	/* KAME hack: embed scopeid */
3781	if (newaddr->sa_family == AF_INET6) {
3782		struct sockaddr_in6 *sin6;
3783
3784		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3785		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3786		sin6->sin6_scope_id = 0;
3787	}
3788#endif
3789	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3790
3791	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3792		/* Get source address */
3793		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3794		    stcb,
3795		    (sctp_route_t *) & net->ro,
3796		    net,
3797		    0,
3798		    stcb->asoc.vrf_id);
3799		/* Now get the interface MTU */
3800		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3801			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3802		} else {
3803			net->mtu = 0;
3804		}
3805#ifdef SCTP_PRINT_FOR_B_AND_M
3806		SCTP_PRINTF("We have found an interface mtu of %d\n", net->mtu);
3807#endif
3808		if (net->mtu == 0) {
3809			/* Huh ?? */
3810			net->mtu = SCTP_DEFAULT_MTU;
3811		} else {
3812			uint32_t rmtu;
3813
3814			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3815#ifdef SCTP_PRINT_FOR_B_AND_M
3816			SCTP_PRINTF("The route mtu is %d\n", rmtu);
3817#endif
3818			if (rmtu == 0) {
3819				/*
3820				 * Start things off to match mtu of
3821				 * interface please.
3822				 */
3823				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3824				    net->ro.ro_rt, net->mtu);
3825			} else {
3826				/*
3827				 * we take the route mtu over the interface,
3828				 * since the route may be leading out the
3829				 * loopback, or a different interface.
3830				 */
3831				net->mtu = rmtu;
3832			}
3833		}
3834		if (from == SCTP_ALLOC_ASOC) {
3835#ifdef SCTP_PRINT_FOR_B_AND_M
3836			SCTP_PRINTF("New assoc sets mtu to :%d\n", net->mtu);
3837#endif
3838			stcb->asoc.smallest_mtu = net->mtu;
3839		}
3840	} else {
3841		net->mtu = stcb->asoc.smallest_mtu;
3842	}
3843#ifdef INET6
3844	if (newaddr->sa_family == AF_INET6) {
3845		struct sockaddr_in6 *sin6;
3846
3847		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3848		(void)sa6_recoverscope(sin6);
3849	}
3850#endif
3851	if (net->port) {
3852		net->mtu -= sizeof(struct udphdr);
3853	}
3854	if (stcb->asoc.smallest_mtu > net->mtu) {
3855#ifdef SCTP_PRINT_FOR_B_AND_M
3856		SCTP_PRINTF("new address mtu:%d smaller than smallest:%d\n",
3857		    net->mtu, stcb->asoc.smallest_mtu);
3858#endif
3859		stcb->asoc.smallest_mtu = net->mtu;
3860	}
3861	/* JRS - Use the congestion control given in the CC module */
3862	stcb->asoc.cc_functions.sctp_set_initial_cc_param(stcb, net);
3863
3864	/*
3865	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3866	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
3867	 */
3868	net->find_pseudo_cumack = 1;
3869	net->find_rtx_pseudo_cumack = 1;
3870	net->src_addr_selected = 0;
3871	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
3872	if (net->ro.ro_rt == NULL) {
3873		/* Since we have no route put it at the back */
3874		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3875	} else if (netfirst == NULL) {
3876		/* We are the first one in the pool. */
3877		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3878	} else if (netfirst->ro.ro_rt == NULL) {
3879		/*
3880		 * First one has NO route. Place this one ahead of the first
3881		 * one.
3882		 */
3883		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3884	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
3885		/*
3886		 * This one has a different interface than the one at the
3887		 * top of the list. Place it ahead.
3888		 */
3889		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3890	} else {
3891		/*
3892		 * Ok we have the same interface as the first one. Move
3893		 * forward until we find either a) one with a NULL route...
3894		 * insert ahead of that b) one with a different ifp.. insert
3895		 * after that. c) end of the list.. insert at the tail.
3896		 */
3897		struct sctp_nets *netlook;
3898
3899		do {
3900			netlook = TAILQ_NEXT(netfirst, sctp_next);
3901			if (netlook == NULL) {
3902				/* End of the list */
3903				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3904				break;
3905			} else if (netlook->ro.ro_rt == NULL) {
3906				/* next one has NO route */
3907				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
3908				break;
3909			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
3910				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
3911				    net, sctp_next);
3912				break;
3913			}
3914			/* Shift forward */
3915			netfirst = netlook;
3916		} while (netlook != NULL);
3917	}
3918
3919	/* got to have a primary set */
3920	if (stcb->asoc.primary_destination == 0) {
3921		stcb->asoc.primary_destination = net;
3922	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
3923		    (net->ro.ro_rt) &&
3924	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
3925		/* No route to current primary adopt new primary */
3926		stcb->asoc.primary_destination = net;
3927	}
3928	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, stcb->sctp_ep, stcb,
3929	    net);
3930	/* Validate primary is first */
3931	net = TAILQ_FIRST(&stcb->asoc.nets);
3932	if ((net != stcb->asoc.primary_destination) &&
3933	    (stcb->asoc.primary_destination)) {
3934		/*
3935		 * first one on the list is NOT the primary sctp_cmpaddr()
3936		 * is much more efficent if the primary is the first on the
3937		 * list, make it so.
3938		 */
3939		TAILQ_REMOVE(&stcb->asoc.nets,
3940		    stcb->asoc.primary_destination, sctp_next);
3941		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
3942		    stcb->asoc.primary_destination, sctp_next);
3943	}
3944	return (0);
3945}
3946
3947
3948static uint32_t
3949sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
3950{
3951	uint32_t id;
3952	struct sctpasochead *head;
3953	struct sctp_tcb *lstcb;
3954
3955	SCTP_INP_WLOCK(inp);
3956try_again:
3957	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3958		/* TSNH */
3959		SCTP_INP_WUNLOCK(inp);
3960		return (0);
3961	}
3962	/*
3963	 * We don't allow assoc id to be 0, this is needed otherwise if the
3964	 * id were to wrap we would have issues with some socket options.
3965	 */
3966	if (inp->sctp_associd_counter == 0) {
3967		inp->sctp_associd_counter++;
3968	}
3969	id = inp->sctp_associd_counter;
3970	inp->sctp_associd_counter++;
3971	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
3972	if (lstcb) {
3973		goto try_again;
3974	}
3975	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
3976	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
3977	stcb->asoc.in_asocid_hash = 1;
3978	SCTP_INP_WUNLOCK(inp);
3979	return id;
3980}
3981
3982/*
3983 * allocate an association and add it to the endpoint. The caller must be
3984 * careful to add all additional addresses once they are know right away or
3985 * else the assoc will be may experience a blackout scenario.
3986 */
3987struct sctp_tcb *
3988sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
3989    int *error, uint32_t override_tag, uint32_t vrf_id,
3990    struct thread *p
3991)
3992{
3993	/* note the p argument is only valid in unbound sockets */
3994
3995	struct sctp_tcb *stcb;
3996	struct sctp_association *asoc;
3997	struct sctpasochead *head;
3998	uint16_t rport;
3999	int err;
4000
4001	/*
4002	 * Assumption made here: Caller has done a
4003	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4004	 * address does not exist already.
4005	 */
4006	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4007		/* Hit max assoc, sorry no more */
4008		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4009		*error = ENOBUFS;
4010		return (NULL);
4011	}
4012	if (firstaddr == NULL) {
4013		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4014		*error = EINVAL;
4015		return (NULL);
4016	}
4017	SCTP_INP_RLOCK(inp);
4018	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4019	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4020	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4021		/*
4022		 * If its in the TCP pool, its NOT allowed to create an
4023		 * association. The parent listener needs to call
4024		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4025		 * off, or connected one does this.. its an error.
4026		 */
4027		SCTP_INP_RUNLOCK(inp);
4028		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4029		*error = EINVAL;
4030		return (NULL);
4031	}
4032	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4033#ifdef SCTP_DEBUG
4034	if (firstaddr) {
4035		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4036		SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4037		    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4038	} else {
4039		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4040	}
4041#endif				/* SCTP_DEBUG */
4042	if (firstaddr->sa_family == AF_INET) {
4043		struct sockaddr_in *sin;
4044
4045		sin = (struct sockaddr_in *)firstaddr;
4046		if ((sin->sin_port == 0) || (sin->sin_addr.s_addr == 0)) {
4047			/* Invalid address */
4048			SCTP_INP_RUNLOCK(inp);
4049			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4050			*error = EINVAL;
4051			return (NULL);
4052		}
4053		rport = sin->sin_port;
4054	} else if (firstaddr->sa_family == AF_INET6) {
4055		struct sockaddr_in6 *sin6;
4056
4057		sin6 = (struct sockaddr_in6 *)firstaddr;
4058		if ((sin6->sin6_port == 0) ||
4059		    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
4060			/* Invalid address */
4061			SCTP_INP_RUNLOCK(inp);
4062			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4063			*error = EINVAL;
4064			return (NULL);
4065		}
4066		rport = sin6->sin6_port;
4067	} else {
4068		/* not supported family type */
4069		SCTP_INP_RUNLOCK(inp);
4070		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4071		*error = EINVAL;
4072		return (NULL);
4073	}
4074	SCTP_INP_RUNLOCK(inp);
4075	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4076		/*
4077		 * If you have not performed a bind, then we need to do the
4078		 * ephemerial bind for you.
4079		 */
4080		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4081		    (struct sockaddr *)NULL,
4082		    (struct sctp_ifa *)NULL,
4083		    p
4084		    ))) {
4085			/* bind error, probably perm */
4086			*error = err;
4087			return (NULL);
4088		}
4089	}
4090	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4091	if (stcb == NULL) {
4092		/* out of memory? */
4093		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4094		*error = ENOMEM;
4095		return (NULL);
4096	}
4097	SCTP_INCR_ASOC_COUNT();
4098
4099	bzero(stcb, sizeof(*stcb));
4100	asoc = &stcb->asoc;
4101
4102	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4103	SCTP_TCB_LOCK_INIT(stcb);
4104	SCTP_TCB_SEND_LOCK_INIT(stcb);
4105	stcb->rport = rport;
4106	/* setup back pointer's */
4107	stcb->sctp_ep = inp;
4108	stcb->sctp_socket = inp->sctp_socket;
4109	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4110		/* failed */
4111		SCTP_TCB_LOCK_DESTROY(stcb);
4112		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4113		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4114		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4115		SCTP_DECR_ASOC_COUNT();
4116		*error = err;
4117		return (NULL);
4118	}
4119	/* and the port */
4120	SCTP_INP_INFO_WLOCK();
4121	SCTP_INP_WLOCK(inp);
4122	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4123		/* inpcb freed while alloc going on */
4124		SCTP_TCB_LOCK_DESTROY(stcb);
4125		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4126		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4127		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4128		SCTP_INP_WUNLOCK(inp);
4129		SCTP_INP_INFO_WUNLOCK();
4130		SCTP_DECR_ASOC_COUNT();
4131		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4132		*error = EINVAL;
4133		return (NULL);
4134	}
4135	SCTP_TCB_LOCK(stcb);
4136
4137	/* now that my_vtag is set, add it to the hash */
4138	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4139	/* put it in the bucket in the vtag hash of assoc's for the system */
4140	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4141	SCTP_INP_INFO_WUNLOCK();
4142
4143	if ((err = sctp_add_remote_addr(stcb, firstaddr, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4144		/* failure.. memory error? */
4145		if (asoc->strmout) {
4146			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4147			asoc->strmout = NULL;
4148		}
4149		if (asoc->mapping_array) {
4150			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4151			asoc->mapping_array = NULL;
4152		}
4153		if (asoc->nr_mapping_array) {
4154			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4155			asoc->nr_mapping_array = NULL;
4156		}
4157		SCTP_DECR_ASOC_COUNT();
4158		SCTP_TCB_LOCK_DESTROY(stcb);
4159		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4160		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4161		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4162		SCTP_INP_WUNLOCK(inp);
4163		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4164		*error = ENOBUFS;
4165		return (NULL);
4166	}
4167	/* Init all the timers */
4168	SCTP_OS_TIMER_INIT(&asoc->hb_timer.timer);
4169	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4170	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4171	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4172	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4173	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4174	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4175	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4176
4177	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4178	/* now file the port under the hash as well */
4179	if (inp->sctp_tcbhash != NULL) {
4180		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4181		    inp->sctp_hashmark)];
4182		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4183	}
4184	SCTP_INP_WUNLOCK(inp);
4185	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", stcb);
4186	return (stcb);
4187}
4188
4189
4190void
4191sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4192{
4193	struct sctp_association *asoc;
4194
4195	asoc = &stcb->asoc;
4196	asoc->numnets--;
4197	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4198	if (net == asoc->primary_destination) {
4199		/* Reset primary */
4200		struct sctp_nets *lnet;
4201
4202		lnet = TAILQ_FIRST(&asoc->nets);
4203		/*
4204		 * Mobility adaptation Ideally, if deleted destination is
4205		 * the primary, it becomes a fast retransmission trigger by
4206		 * the subsequent SET PRIMARY. (by micchie)
4207		 */
4208		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4209		    SCTP_MOBILITY_BASE) ||
4210		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4211		    SCTP_MOBILITY_FASTHANDOFF)) {
4212			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4213			if (asoc->deleted_primary != NULL) {
4214				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4215				goto out;
4216			}
4217			asoc->deleted_primary = net;
4218			atomic_add_int(&net->ref_count, 1);
4219			memset(&net->lastsa, 0, sizeof(net->lastsa));
4220			memset(&net->lastsv, 0, sizeof(net->lastsv));
4221			sctp_mobility_feature_on(stcb->sctp_ep,
4222			    SCTP_MOBILITY_PRIM_DELETED);
4223			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4224			    stcb->sctp_ep, stcb, NULL);
4225		}
4226out:
4227		/* Try to find a confirmed primary */
4228		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4229	}
4230	if (net == asoc->last_data_chunk_from) {
4231		/* Reset primary */
4232		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4233	}
4234	if (net == asoc->last_control_chunk_from) {
4235		/* Clear net */
4236		asoc->last_control_chunk_from = NULL;
4237	}
4238	sctp_free_remote_addr(net);
4239}
4240
4241/*
4242 * remove a remote endpoint address from an association, it will fail if the
4243 * address does not exist.
4244 */
4245int
4246sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4247{
4248	/*
4249	 * Here we need to remove a remote address. This is quite simple, we
4250	 * first find it in the list of address for the association
4251	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4252	 * on that item. Note we do not allow it to be removed if there are
4253	 * no other addresses.
4254	 */
4255	struct sctp_association *asoc;
4256	struct sctp_nets *net, *net_tmp;
4257
4258	asoc = &stcb->asoc;
4259
4260	/* locate the address */
4261	for (net = TAILQ_FIRST(&asoc->nets); net != NULL; net = net_tmp) {
4262		net_tmp = TAILQ_NEXT(net, sctp_next);
4263		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4264			continue;
4265		}
4266		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4267		    remaddr)) {
4268			/* we found the guy */
4269			if (asoc->numnets < 2) {
4270				/* Must have at LEAST two remote addresses */
4271				return (-1);
4272			} else {
4273				sctp_remove_net(stcb, net);
4274				return (0);
4275			}
4276		}
4277	}
4278	/* not found. */
4279	return (-2);
4280}
4281
4282void
4283sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4284{
4285	struct sctpvtaghead *chain;
4286	struct sctp_tagblock *twait_block;
4287	int found = 0;
4288	int i;
4289
4290	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4291	if (!LIST_EMPTY(chain)) {
4292		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4293			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4294				if ((twait_block->vtag_block[i].v_tag == tag) &&
4295				    (twait_block->vtag_block[i].lport == lport) &&
4296				    (twait_block->vtag_block[i].rport == rport)) {
4297					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4298					twait_block->vtag_block[i].v_tag = 0;
4299					twait_block->vtag_block[i].lport = 0;
4300					twait_block->vtag_block[i].rport = 0;
4301					found = 1;
4302					break;
4303				}
4304			}
4305			if (found)
4306				break;
4307		}
4308	}
4309}
4310
4311int
4312sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4313{
4314	struct sctpvtaghead *chain;
4315	struct sctp_tagblock *twait_block;
4316	int found = 0;
4317	int i;
4318
4319	SCTP_INP_INFO_WLOCK();
4320	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4321	if (!LIST_EMPTY(chain)) {
4322		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4323			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4324				if ((twait_block->vtag_block[i].v_tag == tag) &&
4325				    (twait_block->vtag_block[i].lport == lport) &&
4326				    (twait_block->vtag_block[i].rport == rport)) {
4327					found = 1;
4328					break;
4329				}
4330			}
4331			if (found)
4332				break;
4333		}
4334	}
4335	SCTP_INP_INFO_WUNLOCK();
4336	return (found);
4337}
4338
4339
4340void
4341sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4342{
4343	struct sctpvtaghead *chain;
4344	struct sctp_tagblock *twait_block;
4345	struct timeval now;
4346	int set, i;
4347
4348	if (time == 0) {
4349		/* Its disabled */
4350		return;
4351	}
4352	(void)SCTP_GETTIME_TIMEVAL(&now);
4353	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4354	set = 0;
4355	if (!LIST_EMPTY(chain)) {
4356		/* Block(s) present, lets find space, and expire on the fly */
4357		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4358			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4359				if ((twait_block->vtag_block[i].v_tag == 0) &&
4360				    !set) {
4361					twait_block->vtag_block[i].tv_sec_at_expire =
4362					    now.tv_sec + time;
4363					twait_block->vtag_block[i].v_tag = tag;
4364					twait_block->vtag_block[i].lport = lport;
4365					twait_block->vtag_block[i].rport = rport;
4366					set = 1;
4367				} else if ((twait_block->vtag_block[i].v_tag) &&
4368				    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4369					/* Audit expires this guy */
4370					twait_block->vtag_block[i].tv_sec_at_expire = 0;
4371					twait_block->vtag_block[i].v_tag = 0;
4372					twait_block->vtag_block[i].lport = 0;
4373					twait_block->vtag_block[i].rport = 0;
4374					if (set == 0) {
4375						/* Reuse it for my new tag */
4376						twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4377						twait_block->vtag_block[i].v_tag = tag;
4378						twait_block->vtag_block[i].lport = lport;
4379						twait_block->vtag_block[i].rport = rport;
4380						set = 1;
4381					}
4382				}
4383			}
4384			if (set) {
4385				/*
4386				 * We only do up to the block where we can
4387				 * place our tag for audits
4388				 */
4389				break;
4390			}
4391		}
4392	}
4393	/* Need to add a new block to chain */
4394	if (!set) {
4395		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4396		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4397		if (twait_block == NULL) {
4398#ifdef INVARIANTS
4399			panic("Can not alloc tagblock");
4400#endif
4401			return;
4402		}
4403		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4404		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4405		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4406		twait_block->vtag_block[0].v_tag = tag;
4407		twait_block->vtag_block[0].lport = lport;
4408		twait_block->vtag_block[0].rport = rport;
4409	}
4410}
4411
4412
4413static void
4414sctp_iterator_asoc_being_freed(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4415{
4416	struct sctp_iterator *it;
4417
4418	/*
4419	 * Unlock the tcb lock we do this so we avoid a dead lock scenario
4420	 * where the iterator is waiting on the TCB lock and the TCB lock is
4421	 * waiting on the iterator lock.
4422	 */
4423	it = stcb->asoc.stcb_starting_point_for_iterator;
4424	if (it == NULL) {
4425		return;
4426	}
4427	if (it->inp != stcb->sctp_ep) {
4428		/* hmm, focused on the wrong one? */
4429		return;
4430	}
4431	if (it->stcb != stcb) {
4432		return;
4433	}
4434	it->stcb = LIST_NEXT(stcb, sctp_tcblist);
4435	if (it->stcb == NULL) {
4436		/* done with all asoc's in this assoc */
4437		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
4438			it->inp = NULL;
4439		} else {
4440			it->inp = LIST_NEXT(inp, sctp_list);
4441		}
4442	}
4443}
4444
4445
4446/*-
4447 * Free the association after un-hashing the remote port. This
4448 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4449 * expect that the input to this function IS a locked TCB.
4450 * It will return 0, if it did NOT destroy the association (instead
4451 * it unlocks it. It will return NON-zero if it either destroyed the
4452 * association OR the association is already destroyed.
4453 */
4454int
4455sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4456{
4457	int i;
4458	struct sctp_association *asoc;
4459	struct sctp_nets *net, *prev;
4460	struct sctp_laddr *laddr;
4461	struct sctp_tmit_chunk *chk;
4462	struct sctp_asconf_addr *aparam;
4463	struct sctp_asconf_ack *aack;
4464	struct sctp_stream_reset_list *liste;
4465	struct sctp_queued_to_read *sq;
4466	struct sctp_stream_queue_pending *sp;
4467	sctp_sharedkey_t *shared_key;
4468	struct socket *so;
4469	int ccnt = 0;
4470	int cnt = 0;
4471
4472	/* first, lets purge the entry from the hash table. */
4473
4474#ifdef SCTP_LOG_CLOSING
4475	sctp_log_closing(inp, stcb, 6);
4476#endif
4477	if (stcb->asoc.state == 0) {
4478#ifdef SCTP_LOG_CLOSING
4479		sctp_log_closing(inp, NULL, 7);
4480#endif
4481		/* there is no asoc, really TSNH :-0 */
4482		return (1);
4483	}
4484	/* TEMP CODE */
4485	if (stcb->freed_from_where == 0) {
4486		/* Only record the first place free happened from */
4487		stcb->freed_from_where = from_location;
4488	}
4489	/* TEMP CODE */
4490
4491	asoc = &stcb->asoc;
4492	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4493	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4494		/* nothing around */
4495		so = NULL;
4496	else
4497		so = inp->sctp_socket;
4498
4499	/*
4500	 * We used timer based freeing if a reader or writer is in the way.
4501	 * So we first check if we are actually being called from a timer,
4502	 * if so we abort early if a reader or writer is still in the way.
4503	 */
4504	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4505	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4506		/*
4507		 * is it the timer driving us? if so are the reader/writers
4508		 * gone?
4509		 */
4510		if (stcb->asoc.refcnt) {
4511			/* nope, reader or writer in the way */
4512			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4513			/* no asoc destroyed */
4514			SCTP_TCB_UNLOCK(stcb);
4515#ifdef SCTP_LOG_CLOSING
4516			sctp_log_closing(inp, stcb, 8);
4517#endif
4518			return (0);
4519		}
4520	}
4521	/* now clean up any other timers */
4522	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
4523	asoc->hb_timer.self = NULL;
4524	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4525	asoc->dack_timer.self = NULL;
4526	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4527	/*-
4528	 * For stream reset we don't blast this unless
4529	 * it is a str-reset timer, it might be the
4530	 * free-asoc timer which we DON'T want to
4531	 * disturb.
4532	 */
4533	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4534		asoc->strreset_timer.self = NULL;
4535	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4536	asoc->asconf_timer.self = NULL;
4537	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4538	asoc->autoclose_timer.self = NULL;
4539	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4540	asoc->shut_guard_timer.self = NULL;
4541	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4542	asoc->delayed_event_timer.self = NULL;
4543	/* Mobility adaptation */
4544	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4545	asoc->delete_prim_timer.self = NULL;
4546	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4547		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
4548		net->fr_timer.self = NULL;
4549		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4550		net->rxt_timer.self = NULL;
4551		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4552		net->pmtu_timer.self = NULL;
4553	}
4554	/* Now the read queue needs to be cleaned up (only once) */
4555	cnt = 0;
4556	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4557		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4558		SCTP_INP_READ_LOCK(inp);
4559		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4560			if (sq->stcb == stcb) {
4561				sq->do_not_ref_stcb = 1;
4562				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4563				/*
4564				 * If there is no end, there never will be
4565				 * now.
4566				 */
4567				if (sq->end_added == 0) {
4568					/* Held for PD-API clear that. */
4569					sq->pdapi_aborted = 1;
4570					sq->held_length = 0;
4571					if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4572						/*
4573						 * Need to add a PD-API
4574						 * aborted indication.
4575						 * Setting the control_pdapi
4576						 * assures that it will be
4577						 * added right after this
4578						 * msg.
4579						 */
4580						uint32_t strseq;
4581
4582						stcb->asoc.control_pdapi = sq;
4583						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4584						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4585						    stcb,
4586						    SCTP_PARTIAL_DELIVERY_ABORTED,
4587						    (void *)&strseq,
4588						    SCTP_SO_LOCKED);
4589						stcb->asoc.control_pdapi = NULL;
4590					}
4591				}
4592				/* Add an end to wake them */
4593				sq->end_added = 1;
4594				cnt++;
4595			}
4596		}
4597		SCTP_INP_READ_UNLOCK(inp);
4598		if (stcb->block_entry) {
4599			cnt++;
4600			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4601			stcb->block_entry->error = ECONNRESET;
4602			stcb->block_entry = NULL;
4603		}
4604	}
4605	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4606		/*
4607		 * Someone holds a reference OR the socket is unaccepted
4608		 * yet.
4609		 */
4610		if ((stcb->asoc.refcnt) ||
4611		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4612		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4613			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4614			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4615		}
4616		SCTP_TCB_UNLOCK(stcb);
4617		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4618		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4619			/* nothing around */
4620			so = NULL;
4621		if (so) {
4622			/* Wake any reader/writers */
4623			sctp_sorwakeup(inp, so);
4624			sctp_sowwakeup(inp, so);
4625		}
4626#ifdef SCTP_LOG_CLOSING
4627		sctp_log_closing(inp, stcb, 9);
4628#endif
4629		/* no asoc destroyed */
4630		return (0);
4631	}
4632#ifdef SCTP_LOG_CLOSING
4633	sctp_log_closing(inp, stcb, 10);
4634#endif
4635	/*
4636	 * When I reach here, no others want to kill the assoc yet.. and I
4637	 * own the lock. Now its possible an abort comes in when I do the
4638	 * lock exchange below to grab all the locks to do the final take
4639	 * out. to prevent this we increment the count, which will start a
4640	 * timer and blow out above thus assuring us that we hold exclusive
4641	 * killing of the asoc. Note that after getting back the TCB lock we
4642	 * will go ahead and increment the counter back up and stop any
4643	 * timer a passing stranger may have started :-S
4644	 */
4645	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4646		atomic_add_int(&stcb->asoc.refcnt, 1);
4647
4648		SCTP_TCB_UNLOCK(stcb);
4649
4650		SCTP_ITERATOR_LOCK();
4651		SCTP_INP_INFO_WLOCK();
4652		SCTP_INP_WLOCK(inp);
4653		SCTP_TCB_LOCK(stcb);
4654	}
4655	/* Double check the GONE flag */
4656	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4657	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4658		/* nothing around */
4659		so = NULL;
4660
4661	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4662	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4663		/*
4664		 * For TCP type we need special handling when we are
4665		 * connected. We also include the peel'ed off ones to.
4666		 */
4667		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4668			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4669			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4670			if (so) {
4671				SOCK_LOCK(so);
4672				if (so->so_rcv.sb_cc == 0) {
4673					so->so_state &= ~(SS_ISCONNECTING |
4674					    SS_ISDISCONNECTING |
4675					    SS_ISCONFIRMING |
4676					    SS_ISCONNECTED);
4677				}
4678				socantrcvmore_locked(so);
4679				SOCK_UNLOCK(so);
4680				sctp_sowwakeup(inp, so);
4681				sctp_sorwakeup(inp, so);
4682				SCTP_SOWAKEUP(so);
4683			}
4684		}
4685	}
4686	/*
4687	 * Make it invalid too, that way if its about to run it will abort
4688	 * and return.
4689	 */
4690	sctp_iterator_asoc_being_freed(inp, stcb);
4691	/* re-increment the lock */
4692	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4693		atomic_add_int(&stcb->asoc.refcnt, -1);
4694	}
4695	asoc->state = 0;
4696	if (inp->sctp_tcbhash) {
4697		LIST_REMOVE(stcb, sctp_tcbhash);
4698	}
4699	if (stcb->asoc.in_asocid_hash) {
4700		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4701	}
4702	/* Now lets remove it from the list of ALL associations in the EP */
4703	LIST_REMOVE(stcb, sctp_tcblist);
4704	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4705		SCTP_INP_INCR_REF(inp);
4706		SCTP_INP_WUNLOCK(inp);
4707		SCTP_ITERATOR_UNLOCK();
4708	}
4709	/* pull from vtag hash */
4710	LIST_REMOVE(stcb, sctp_asocs);
4711	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4712	    inp->sctp_lport, stcb->rport);
4713
4714	/*
4715	 * Now restop the timers to be sure this is paranoia at is finest!
4716	 */
4717	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4718	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
4719	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4720	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4721	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4722	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4723	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4724	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4725	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4726		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
4727		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4728		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4729	}
4730
4731	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4732	prev = NULL;
4733	/*
4734	 * The chunk lists and such SHOULD be empty but we check them just
4735	 * in case.
4736	 */
4737	/* anything on the wheel needs to be removed */
4738	for (i = 0; i < asoc->streamoutcnt; i++) {
4739		struct sctp_stream_out *outs;
4740
4741		outs = &asoc->strmout[i];
4742		/* now clean up any chunks here */
4743		sp = TAILQ_FIRST(&outs->outqueue);
4744		while (sp) {
4745			TAILQ_REMOVE(&outs->outqueue, sp, next);
4746			if (sp->data) {
4747				sctp_m_freem(sp->data);
4748				sp->data = NULL;
4749				sp->tail_mbuf = NULL;
4750			}
4751			sctp_free_remote_addr(sp->net);
4752			sctp_free_spbufspace(stcb, asoc, sp);
4753			if (sp->holds_key_ref)
4754				sctp_auth_key_release(stcb, sp->auth_keyid);
4755			/* Free the zone stuff  */
4756			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_strmoq), sp);
4757			SCTP_DECR_STRMOQ_COUNT();
4758			/* sa_ignore FREED_MEMORY */
4759			sp = TAILQ_FIRST(&outs->outqueue);
4760		}
4761	}
4762
4763	/* sa_ignore FREED_MEMORY */
4764	while ((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) {
4765		TAILQ_REMOVE(&asoc->resetHead, liste, next_resp);
4766		SCTP_FREE(liste, SCTP_M_STRESET);
4767	}
4768
4769	sq = TAILQ_FIRST(&asoc->pending_reply_queue);
4770	while (sq) {
4771		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4772		if (sq->data) {
4773			sctp_m_freem(sq->data);
4774			sq->data = NULL;
4775		}
4776		sctp_free_remote_addr(sq->whoFrom);
4777		sq->whoFrom = NULL;
4778		sq->stcb = NULL;
4779		/* Free the ctl entry */
4780		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4781		SCTP_DECR_READQ_COUNT();
4782		/* sa_ignore FREED_MEMORY */
4783		sq = TAILQ_FIRST(&asoc->pending_reply_queue);
4784	}
4785
4786	chk = TAILQ_FIRST(&asoc->free_chunks);
4787	while (chk) {
4788		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4789		if (chk->data) {
4790			sctp_m_freem(chk->data);
4791			chk->data = NULL;
4792		}
4793		if (chk->holds_key_ref)
4794			sctp_auth_key_release(stcb, chk->auth_keyid);
4795		ccnt++;
4796		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4797		SCTP_DECR_CHK_COUNT();
4798		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4799		asoc->free_chunk_cnt--;
4800		/* sa_ignore FREED_MEMORY */
4801		chk = TAILQ_FIRST(&asoc->free_chunks);
4802	}
4803	/* pending send queue SHOULD be empty */
4804	if (!TAILQ_EMPTY(&asoc->send_queue)) {
4805		chk = TAILQ_FIRST(&asoc->send_queue);
4806		while (chk) {
4807			TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4808			if (chk->data) {
4809				sctp_m_freem(chk->data);
4810				chk->data = NULL;
4811			}
4812			if (chk->holds_key_ref)
4813				sctp_auth_key_release(stcb, chk->auth_keyid);
4814			ccnt++;
4815			sctp_free_remote_addr(chk->whoTo);
4816			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4817			SCTP_DECR_CHK_COUNT();
4818			/* sa_ignore FREED_MEMORY */
4819			chk = TAILQ_FIRST(&asoc->send_queue);
4820		}
4821	}
4822/*
4823  if (ccnt) {
4824  printf("Freed %d from send_queue\n", ccnt);
4825  ccnt = 0;
4826  }
4827*/
4828	/* sent queue SHOULD be empty */
4829	if (!TAILQ_EMPTY(&asoc->sent_queue)) {
4830		chk = TAILQ_FIRST(&asoc->sent_queue);
4831		while (chk) {
4832			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4833			if (chk->data) {
4834				sctp_m_freem(chk->data);
4835				chk->data = NULL;
4836			}
4837			if (chk->holds_key_ref)
4838				sctp_auth_key_release(stcb, chk->auth_keyid);
4839			ccnt++;
4840			sctp_free_remote_addr(chk->whoTo);
4841			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4842			SCTP_DECR_CHK_COUNT();
4843			/* sa_ignore FREED_MEMORY */
4844			chk = TAILQ_FIRST(&asoc->sent_queue);
4845		}
4846	}
4847/*
4848  if (ccnt) {
4849  printf("Freed %d from sent_queue\n", ccnt);
4850  ccnt = 0;
4851  }
4852*/
4853	/* control queue MAY not be empty */
4854	if (!TAILQ_EMPTY(&asoc->control_send_queue)) {
4855		chk = TAILQ_FIRST(&asoc->control_send_queue);
4856		while (chk) {
4857			TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
4858			if (chk->data) {
4859				sctp_m_freem(chk->data);
4860				chk->data = NULL;
4861			}
4862			if (chk->holds_key_ref)
4863				sctp_auth_key_release(stcb, chk->auth_keyid);
4864			ccnt++;
4865			sctp_free_remote_addr(chk->whoTo);
4866			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4867			SCTP_DECR_CHK_COUNT();
4868			/* sa_ignore FREED_MEMORY */
4869			chk = TAILQ_FIRST(&asoc->control_send_queue);
4870		}
4871	}
4872/*
4873  if (ccnt) {
4874  printf("Freed %d from ctrl_queue\n", ccnt);
4875  ccnt = 0;
4876  }
4877*/
4878
4879	/* ASCONF queue MAY not be empty */
4880	if (!TAILQ_EMPTY(&asoc->asconf_send_queue)) {
4881		chk = TAILQ_FIRST(&asoc->asconf_send_queue);
4882		while (chk) {
4883			TAILQ_REMOVE(&asoc->asconf_send_queue, 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_free_remote_addr(chk->whoTo);
4892			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4893			SCTP_DECR_CHK_COUNT();
4894			/* sa_ignore FREED_MEMORY */
4895			chk = TAILQ_FIRST(&asoc->asconf_send_queue);
4896		}
4897	}
4898/*
4899  if (ccnt) {
4900  printf("Freed %d from asconf_queue\n", ccnt);
4901  ccnt = 0;
4902  }
4903*/
4904	if (!TAILQ_EMPTY(&asoc->reasmqueue)) {
4905		chk = TAILQ_FIRST(&asoc->reasmqueue);
4906		while (chk) {
4907			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
4908			if (chk->data) {
4909				sctp_m_freem(chk->data);
4910				chk->data = NULL;
4911			}
4912			if (chk->holds_key_ref)
4913				sctp_auth_key_release(stcb, chk->auth_keyid);
4914			sctp_free_remote_addr(chk->whoTo);
4915			ccnt++;
4916			SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4917			SCTP_DECR_CHK_COUNT();
4918			/* sa_ignore FREED_MEMORY */
4919			chk = TAILQ_FIRST(&asoc->reasmqueue);
4920		}
4921	}
4922/*
4923  if (ccnt) {
4924  printf("Freed %d from reasm_queue\n", ccnt);
4925  ccnt = 0;
4926  }
4927*/
4928	if (asoc->mapping_array) {
4929		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4930		asoc->mapping_array = NULL;
4931	}
4932	if (asoc->nr_mapping_array) {
4933		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4934		asoc->nr_mapping_array = NULL;
4935	}
4936	/* the stream outs */
4937	if (asoc->strmout) {
4938		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4939		asoc->strmout = NULL;
4940	}
4941	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
4942	if (asoc->strmin) {
4943		struct sctp_queued_to_read *ctl;
4944
4945		for (i = 0; i < asoc->streamincnt; i++) {
4946			if (!TAILQ_EMPTY(&asoc->strmin[i].inqueue)) {
4947				/* We have somethings on the streamin queue */
4948				ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
4949				while (ctl) {
4950					TAILQ_REMOVE(&asoc->strmin[i].inqueue,
4951					    ctl, next);
4952					sctp_free_remote_addr(ctl->whoFrom);
4953					if (ctl->data) {
4954						sctp_m_freem(ctl->data);
4955						ctl->data = NULL;
4956					}
4957					/*
4958					 * We don't free the address here
4959					 * since all the net's were freed
4960					 * above.
4961					 */
4962					SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
4963					SCTP_DECR_READQ_COUNT();
4964					ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
4965				}
4966			}
4967		}
4968		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
4969		asoc->strmin = NULL;
4970	}
4971	asoc->streamincnt = 0;
4972	while (!TAILQ_EMPTY(&asoc->nets)) {
4973		/* sa_ignore FREED_MEMORY */
4974		net = TAILQ_FIRST(&asoc->nets);
4975		/* pull from list */
4976		if ((SCTP_BASE_INFO(ipi_count_raddr) == 0) || (prev == net)) {
4977#ifdef INVARIANTS
4978			panic("no net's left alloc'ed, or list points to itself");
4979#endif
4980			break;
4981		}
4982		prev = net;
4983		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4984		sctp_free_remote_addr(net);
4985	}
4986
4987	while (!LIST_EMPTY(&asoc->sctp_restricted_addrs)) {
4988		/* sa_ignore FREED_MEMORY */
4989		laddr = LIST_FIRST(&asoc->sctp_restricted_addrs);
4990		sctp_remove_laddr(laddr);
4991	}
4992
4993	/* pending asconf (address) parameters */
4994	while (!TAILQ_EMPTY(&asoc->asconf_queue)) {
4995		/* sa_ignore FREED_MEMORY */
4996		aparam = TAILQ_FIRST(&asoc->asconf_queue);
4997		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
4998		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
4999	}
5000	while (!TAILQ_EMPTY(&asoc->asconf_ack_sent)) {
5001		/* sa_ignore FREED_MEMORY */
5002		aack = TAILQ_FIRST(&asoc->asconf_ack_sent);
5003		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5004		if (aack->data != NULL) {
5005			sctp_m_freem(aack->data);
5006		}
5007		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5008	}
5009	/* clean up auth stuff */
5010	if (asoc->local_hmacs)
5011		sctp_free_hmaclist(asoc->local_hmacs);
5012	if (asoc->peer_hmacs)
5013		sctp_free_hmaclist(asoc->peer_hmacs);
5014
5015	if (asoc->local_auth_chunks)
5016		sctp_free_chunklist(asoc->local_auth_chunks);
5017	if (asoc->peer_auth_chunks)
5018		sctp_free_chunklist(asoc->peer_auth_chunks);
5019
5020	sctp_free_authinfo(&asoc->authinfo);
5021
5022	shared_key = LIST_FIRST(&asoc->shared_keys);
5023	while (shared_key) {
5024		LIST_REMOVE(shared_key, next);
5025		sctp_free_sharedkey(shared_key);
5026		/* sa_ignore FREED_MEMORY */
5027		shared_key = LIST_FIRST(&asoc->shared_keys);
5028	}
5029
5030	/* Insert new items here :> */
5031
5032	/* Get rid of LOCK */
5033	SCTP_TCB_LOCK_DESTROY(stcb);
5034	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5035	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5036		SCTP_INP_INFO_WUNLOCK();
5037		SCTP_INP_RLOCK(inp);
5038	}
5039#ifdef SCTP_TRACK_FREED_ASOCS
5040	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5041		/* now clean up the tasoc itself */
5042		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5043		SCTP_DECR_ASOC_COUNT();
5044	} else {
5045		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5046	}
5047#else
5048	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5049	SCTP_DECR_ASOC_COUNT();
5050#endif
5051	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5052		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5053			/*
5054			 * If its NOT the inp_free calling us AND sctp_close
5055			 * as been called, we call back...
5056			 */
5057			SCTP_INP_RUNLOCK(inp);
5058			/*
5059			 * This will start the kill timer (if we are the
5060			 * lastone) since we hold an increment yet. But this
5061			 * is the only safe way to do this since otherwise
5062			 * if the socket closes at the same time we are here
5063			 * we might collide in the cleanup.
5064			 */
5065			sctp_inpcb_free(inp,
5066			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5067			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5068			SCTP_INP_DECR_REF(inp);
5069			goto out_of;
5070		} else {
5071			/* The socket is still open. */
5072			SCTP_INP_DECR_REF(inp);
5073		}
5074	}
5075	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5076		SCTP_INP_RUNLOCK(inp);
5077	}
5078out_of:
5079	/* destroyed the asoc */
5080#ifdef SCTP_LOG_CLOSING
5081	sctp_log_closing(inp, NULL, 11);
5082#endif
5083	return (1);
5084}
5085
5086
5087
5088/*
5089 * determine if a destination is "reachable" based upon the addresses bound
5090 * to the current endpoint (e.g. only v4 or v6 currently bound)
5091 */
5092/*
5093 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5094 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5095 * types bound as its endpoint
5096 */
5097int
5098sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5099{
5100	struct sctp_inpcb *inp;
5101	int answer;
5102
5103	/*
5104	 * No locks here, the TCB, in all cases is already locked and an
5105	 * assoc is up. There is either a INP lock by the caller applied (in
5106	 * asconf case when deleting an address) or NOT in the HB case,
5107	 * however if HB then the INP increment is up and the INP will not
5108	 * be removed (on top of the fact that we have a TCB lock). So we
5109	 * only want to read the sctp_flags, which is either bound-all or
5110	 * not.. no protection needed since once an assoc is up you can't be
5111	 * changing your binding.
5112	 */
5113	inp = stcb->sctp_ep;
5114	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5115		/* if bound all, destination is not restricted */
5116		/*
5117		 * RRS: Question during lock work: Is this correct? If you
5118		 * are bound-all you still might need to obey the V4--V6
5119		 * flags??? IMO this bound-all stuff needs to be removed!
5120		 */
5121		return (1);
5122	}
5123	/* NOTE: all "scope" checks are done when local addresses are added */
5124	if (destaddr->sa_family == AF_INET6) {
5125		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5126	} else if (destaddr->sa_family == AF_INET) {
5127		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5128	} else {
5129		/* invalid family, so it's unreachable */
5130		answer = 0;
5131	}
5132	return (answer);
5133}
5134
5135/*
5136 * update the inp_vflags on an endpoint
5137 */
5138static void
5139sctp_update_ep_vflag(struct sctp_inpcb *inp)
5140{
5141	struct sctp_laddr *laddr;
5142
5143	/* first clear the flag */
5144	inp->ip_inp.inp.inp_vflag = 0;
5145	/* set the flag based on addresses on the ep list */
5146	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5147		if (laddr->ifa == NULL) {
5148			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5149			    __FUNCTION__);
5150			continue;
5151		}
5152		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5153			continue;
5154		}
5155		if (laddr->ifa->address.sa.sa_family == AF_INET6) {
5156			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5157		} else if (laddr->ifa->address.sa.sa_family == AF_INET) {
5158			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5159		}
5160	}
5161}
5162
5163/*
5164 * Add the address to the endpoint local address list There is nothing to be
5165 * done if we are bound to all addresses
5166 */
5167void
5168sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5169{
5170	struct sctp_laddr *laddr;
5171	int fnd, error = 0;
5172
5173	fnd = 0;
5174
5175	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5176		/* You are already bound to all. You have it already */
5177		return;
5178	}
5179	if (ifa->address.sa.sa_family == AF_INET6) {
5180		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5181			/* Can't bind a non-useable addr. */
5182			return;
5183		}
5184	}
5185	/* first, is it already present? */
5186	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5187		if (laddr->ifa == ifa) {
5188			fnd = 1;
5189			break;
5190		}
5191	}
5192
5193	if (fnd == 0) {
5194		/* Not in the ep list */
5195		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5196		if (error != 0)
5197			return;
5198		inp->laddr_count++;
5199		/* update inp_vflag flags */
5200		if (ifa->address.sa.sa_family == AF_INET6) {
5201			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5202		} else if (ifa->address.sa.sa_family == AF_INET) {
5203			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5204		}
5205	}
5206	return;
5207}
5208
5209
5210/*
5211 * select a new (hopefully reachable) destination net (should only be used
5212 * when we deleted an ep addr that is the only usable source address to reach
5213 * the destination net)
5214 */
5215static void
5216sctp_select_primary_destination(struct sctp_tcb *stcb)
5217{
5218	struct sctp_nets *net;
5219
5220	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5221		/* for now, we'll just pick the first reachable one we find */
5222		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5223			continue;
5224		if (sctp_destination_is_reachable(stcb,
5225		    (struct sockaddr *)&net->ro._l_addr)) {
5226			/* found a reachable destination */
5227			stcb->asoc.primary_destination = net;
5228		}
5229	}
5230	/* I can't there from here! ...we're gonna die shortly... */
5231}
5232
5233
5234/*
5235 * Delete the address from the endpoint local address list There is nothing
5236 * to be done if we are bound to all addresses
5237 */
5238void
5239sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5240{
5241	struct sctp_laddr *laddr;
5242	int fnd;
5243
5244	fnd = 0;
5245	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5246		/* You are already bound to all. You have it already */
5247		return;
5248	}
5249	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5250		if (laddr->ifa == ifa) {
5251			fnd = 1;
5252			break;
5253		}
5254	}
5255	if (fnd && (inp->laddr_count < 2)) {
5256		/* can't delete unless there are at LEAST 2 addresses */
5257		return;
5258	}
5259	if (fnd) {
5260		/*
5261		 * clean up any use of this address go through our
5262		 * associations and clear any last_used_address that match
5263		 * this one for each assoc, see if a new primary_destination
5264		 * is needed
5265		 */
5266		struct sctp_tcb *stcb;
5267
5268		/* clean up "next_addr_touse" */
5269		if (inp->next_addr_touse == laddr)
5270			/* delete this address */
5271			inp->next_addr_touse = NULL;
5272
5273		/* clean up "last_used_address" */
5274		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5275			struct sctp_nets *net;
5276
5277			SCTP_TCB_LOCK(stcb);
5278			if (stcb->asoc.last_used_address == laddr)
5279				/* delete this address */
5280				stcb->asoc.last_used_address = NULL;
5281			/*
5282			 * Now spin through all the nets and purge any ref
5283			 * to laddr
5284			 */
5285			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5286				if (net->ro._s_addr &&
5287				    (net->ro._s_addr->ifa == laddr->ifa)) {
5288					/* Yep, purge src address selected */
5289					sctp_rtentry_t *rt;
5290
5291					/* delete this address if cached */
5292					rt = net->ro.ro_rt;
5293					if (rt != NULL) {
5294						RTFREE(rt);
5295						net->ro.ro_rt = NULL;
5296					}
5297					sctp_free_ifa(net->ro._s_addr);
5298					net->ro._s_addr = NULL;
5299					net->src_addr_selected = 0;
5300				}
5301			}
5302			SCTP_TCB_UNLOCK(stcb);
5303		}		/* for each tcb */
5304		/* remove it from the ep list */
5305		sctp_remove_laddr(laddr);
5306		inp->laddr_count--;
5307		/* update inp_vflag flags */
5308		sctp_update_ep_vflag(inp);
5309	}
5310	return;
5311}
5312
5313/*
5314 * Add the address to the TCB local address restricted list.
5315 * This is a "pending" address list (eg. addresses waiting for an
5316 * ASCONF-ACK response) and cannot be used as a valid source address.
5317 */
5318void
5319sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5320{
5321	struct sctp_inpcb *inp;
5322	struct sctp_laddr *laddr;
5323	struct sctpladdr *list;
5324
5325	/*
5326	 * Assumes TCB is locked.. and possibly the INP. May need to
5327	 * confirm/fix that if we need it and is not the case.
5328	 */
5329	list = &stcb->asoc.sctp_restricted_addrs;
5330
5331	inp = stcb->sctp_ep;
5332	if (ifa->address.sa.sa_family == AF_INET6) {
5333		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5334			/* Can't bind a non-existent addr. */
5335			return;
5336		}
5337	}
5338	/* does the address already exist? */
5339	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5340		if (laddr->ifa == ifa) {
5341			return;
5342		}
5343	}
5344
5345	/* add to the list */
5346	(void)sctp_insert_laddr(list, ifa, 0);
5347	return;
5348}
5349
5350/*
5351 * insert an laddr entry with the given ifa for the desired list
5352 */
5353int
5354sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5355{
5356	struct sctp_laddr *laddr;
5357
5358	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5359	if (laddr == NULL) {
5360		/* out of memory? */
5361		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5362		return (EINVAL);
5363	}
5364	SCTP_INCR_LADDR_COUNT();
5365	bzero(laddr, sizeof(*laddr));
5366	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5367	laddr->ifa = ifa;
5368	laddr->action = act;
5369	atomic_add_int(&ifa->refcount, 1);
5370	/* insert it */
5371	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5372
5373	return (0);
5374}
5375
5376/*
5377 * Remove an laddr entry from the local address list (on an assoc)
5378 */
5379void
5380sctp_remove_laddr(struct sctp_laddr *laddr)
5381{
5382
5383	/* remove from the list */
5384	LIST_REMOVE(laddr, sctp_nxt_addr);
5385	sctp_free_ifa(laddr->ifa);
5386	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5387	SCTP_DECR_LADDR_COUNT();
5388}
5389
5390/*
5391 * Remove a local address from the TCB local address restricted list
5392 */
5393void
5394sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5395{
5396	struct sctp_inpcb *inp;
5397	struct sctp_laddr *laddr;
5398
5399	/*
5400	 * This is called by asconf work. It is assumed that a) The TCB is
5401	 * locked and b) The INP is locked. This is true in as much as I can
5402	 * trace through the entry asconf code where I did these locks.
5403	 * Again, the ASCONF code is a bit different in that it does lock
5404	 * the INP during its work often times. This must be since we don't
5405	 * want other proc's looking up things while what they are looking
5406	 * up is changing :-D
5407	 */
5408
5409	inp = stcb->sctp_ep;
5410	/* if subset bound and don't allow ASCONF's, can't delete last */
5411	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5412	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5413		if (stcb->sctp_ep->laddr_count < 2) {
5414			/* can't delete last address */
5415			return;
5416		}
5417	}
5418	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5419		/* remove the address if it exists */
5420		if (laddr->ifa == NULL)
5421			continue;
5422		if (laddr->ifa == ifa) {
5423			sctp_remove_laddr(laddr);
5424			return;
5425		}
5426	}
5427
5428	/* address not found! */
5429	return;
5430}
5431
5432/*
5433 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5434 */
5435/* sysctl */
5436static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5437static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5438
5439void
5440sctp_pcb_init()
5441{
5442	/*
5443	 * SCTP initialization for the PCB structures should be called by
5444	 * the sctp_init() funciton.
5445	 */
5446	int i;
5447	struct timeval tv;
5448
5449	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5450		/* error I was called twice */
5451		return;
5452	}
5453	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5454
5455#if defined(SCTP_LOCAL_TRACE_BUF)
5456	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5457#endif
5458	(void)SCTP_GETTIME_TIMEVAL(&tv);
5459#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5460	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5461	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5462#else
5463	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5464	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5465#endif
5466	/* init the empty list of (All) Endpoints */
5467	LIST_INIT(&SCTP_BASE_INFO(listhead));
5468
5469
5470	/* init the hash table of endpoints */
5471	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5472	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5473	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5474	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5475	    &SCTP_BASE_INFO(hashasocmark));
5476	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5477	    &SCTP_BASE_INFO(hashmark));
5478	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5479	    &SCTP_BASE_INFO(hashtcpmark));
5480	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5481
5482
5483	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5484	    &SCTP_BASE_INFO(hashvrfmark));
5485
5486	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5487	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5488	/* init the zones */
5489	/*
5490	 * FIX ME: Should check for NULL returns, but if it does fail we are
5491	 * doomed to panic anyways... add later maybe.
5492	 */
5493	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5494	    sizeof(struct sctp_inpcb), maxsockets);
5495
5496	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5497	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5498
5499	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5500	    sizeof(struct sctp_laddr),
5501	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5502
5503	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5504	    sizeof(struct sctp_nets),
5505	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5506
5507	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5508	    sizeof(struct sctp_tmit_chunk),
5509	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5510
5511	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5512	    sizeof(struct sctp_queued_to_read),
5513	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5514
5515	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5516	    sizeof(struct sctp_stream_queue_pending),
5517	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5518
5519	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5520	    sizeof(struct sctp_asconf),
5521	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5522
5523	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5524	    sizeof(struct sctp_asconf_ack),
5525	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5526
5527
5528	/* Master Lock INIT for info structure */
5529	SCTP_INP_INFO_LOCK_INIT();
5530	SCTP_STATLOG_INIT_LOCK();
5531
5532	SCTP_IPI_COUNT_INIT();
5533	SCTP_IPI_ADDR_INIT();
5534#ifdef SCTP_PACKET_LOGGING
5535	SCTP_IP_PKTLOG_INIT();
5536#endif
5537	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5538
5539	SCTP_WQ_ADDR_INIT();
5540	/* not sure if we need all the counts */
5541	SCTP_BASE_INFO(ipi_count_ep) = 0;
5542	/* assoc/tcb zone info */
5543	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5544	/* local addrlist zone info */
5545	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5546	/* remote addrlist zone info */
5547	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5548	/* chunk info */
5549	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5550
5551	/* socket queue zone info */
5552	SCTP_BASE_INFO(ipi_count_readq) = 0;
5553
5554	/* stream out queue cont */
5555	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5556
5557	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5558	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5559
5560	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5561
5562	/* Init the TIMEWAIT list */
5563	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5564		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5565	}
5566
5567	sctp_startup_iterator();
5568
5569	/*
5570	 * INIT the default VRF which for BSD is the only one, other O/S's
5571	 * may have more. But initially they must start with one and then
5572	 * add the VRF's as addresses are added.
5573	 */
5574	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5575
5576}
5577
5578/*
5579 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5580 */
5581void
5582sctp_pcb_finish(void)
5583{
5584	struct sctp_vrflist *vrf_bucket;
5585	struct sctp_vrf *vrf;
5586	struct sctp_ifn *ifn;
5587	struct sctp_ifa *ifa;
5588	struct sctpvtaghead *chain;
5589	struct sctp_tagblock *twait_block, *prev_twait_block;
5590	struct sctp_laddr *wi;
5591	int i;
5592
5593	/*
5594	 * Free BSD the it thread never exits but we do clean up. The only
5595	 * way freebsd reaches here if we have VRF's but we still add the
5596	 * ifdef to make it compile on old versions.
5597	 */
5598	{
5599		struct sctp_iterator *it, *nit;
5600
5601		SCTP_IPI_ITERATOR_WQ_LOCK();
5602		it = TAILQ_FIRST(&sctp_it_ctl.iteratorhead);
5603		while (it) {
5604			nit = TAILQ_NEXT(it, sctp_nxt_itr);
5605			if (it->vn != curvnet) {
5606				it = nit;
5607				continue;
5608			}
5609			TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
5610			    it, sctp_nxt_itr);
5611			if (it->function_atend != NULL) {
5612				(*it->function_atend) (it->pointer, it->val);
5613			}
5614			SCTP_FREE(it, SCTP_M_ITER);
5615			it = nit;
5616		}
5617		SCTP_IPI_ITERATOR_WQ_UNLOCK();
5618		SCTP_ITERATOR_LOCK();
5619		if ((sctp_it_ctl.cur_it) &&
5620		    (sctp_it_ctl.cur_it->vn == curvnet)) {
5621			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5622		}
5623		SCTP_ITERATOR_UNLOCK();
5624	}
5625
5626	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5627	SCTP_WQ_ADDR_LOCK();
5628	while ((wi = LIST_FIRST(&SCTP_BASE_INFO(addr_wq))) != NULL) {
5629		LIST_REMOVE(wi, sctp_nxt_addr);
5630		SCTP_DECR_LADDR_COUNT();
5631		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5632	}
5633	SCTP_WQ_ADDR_UNLOCK();
5634
5635	/*
5636	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5637	 * destroyed first).
5638	 */
5639	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5640	while ((vrf = LIST_FIRST(vrf_bucket)) != NULL) {
5641		while ((ifn = LIST_FIRST(&vrf->ifnlist)) != NULL) {
5642			while ((ifa = LIST_FIRST(&ifn->ifalist)) != NULL) {
5643				/* free the ifa */
5644				LIST_REMOVE(ifa, next_bucket);
5645				LIST_REMOVE(ifa, next_ifa);
5646				SCTP_FREE(ifa, SCTP_M_IFA);
5647			}
5648			/* free the ifn */
5649			LIST_REMOVE(ifn, next_bucket);
5650			LIST_REMOVE(ifn, next_ifn);
5651			SCTP_FREE(ifn, SCTP_M_IFN);
5652		}
5653		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5654		/* free the vrf */
5655		LIST_REMOVE(vrf, next_vrf);
5656		SCTP_FREE(vrf, SCTP_M_VRF);
5657	}
5658	/* free the vrf hashes */
5659	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5660	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5661
5662	/*
5663	 * free the TIMEWAIT list elements malloc'd in the function
5664	 * sctp_add_vtag_to_timewait()...
5665	 */
5666	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5667		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5668		if (!LIST_EMPTY(chain)) {
5669			prev_twait_block = NULL;
5670			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5671				if (prev_twait_block) {
5672					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5673				}
5674				prev_twait_block = twait_block;
5675			}
5676			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5677		}
5678	}
5679
5680	/* free the locks and mutexes */
5681#ifdef SCTP_PACKET_LOGGING
5682	SCTP_IP_PKTLOG_DESTROY();
5683#endif
5684	SCTP_IPI_ADDR_DESTROY();
5685	SCTP_STATLOG_DESTROY();
5686	SCTP_INP_INFO_LOCK_DESTROY();
5687
5688	SCTP_WQ_ADDR_DESTROY();
5689
5690	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
5691	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
5692	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
5693	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
5694	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
5695	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
5696	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
5697	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
5698	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
5699	/* Get rid of other stuff to */
5700	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
5701		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
5702	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
5703		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
5704	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
5705		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
5706
5707}
5708
5709
5710int
5711sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
5712    int iphlen, int offset, int limit, struct sctphdr *sh,
5713    struct sockaddr *altsa)
5714{
5715	/*
5716	 * grub through the INIT pulling addresses and loading them to the
5717	 * nets structure in the asoc. The from address in the mbuf should
5718	 * also be loaded (if it is not already). This routine can be called
5719	 * with either INIT or INIT-ACK's as long as the m points to the IP
5720	 * packet and the offset points to the beginning of the parameters.
5721	 */
5722	struct sctp_inpcb *inp, *l_inp;
5723	struct sctp_nets *net, *net_tmp;
5724	struct ip *iph;
5725	struct sctp_paramhdr *phdr, parm_buf;
5726	struct sctp_tcb *stcb_tmp;
5727	uint16_t ptype, plen;
5728	struct sockaddr *sa;
5729	struct sockaddr_storage dest_store;
5730	struct sockaddr *local_sa = (struct sockaddr *)&dest_store;
5731	struct sockaddr_in sin;
5732	struct sockaddr_in6 sin6;
5733	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
5734	struct sctp_auth_random *p_random = NULL;
5735	uint16_t random_len = 0;
5736	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
5737	struct sctp_auth_hmac_algo *hmacs = NULL;
5738	uint16_t hmacs_len = 0;
5739	uint8_t saw_asconf = 0;
5740	uint8_t saw_asconf_ack = 0;
5741	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
5742	struct sctp_auth_chunk_list *chunks = NULL;
5743	uint16_t num_chunks = 0;
5744	sctp_key_t *new_key;
5745	uint32_t keylen;
5746	int got_random = 0, got_hmacs = 0, got_chklist = 0;
5747
5748	/* First get the destination address setup too. */
5749	memset(&sin, 0, sizeof(sin));
5750	memset(&sin6, 0, sizeof(sin6));
5751
5752	sin.sin_family = AF_INET;
5753	sin.sin_len = sizeof(sin);
5754	sin.sin_port = stcb->rport;
5755
5756	sin6.sin6_family = AF_INET6;
5757	sin6.sin6_len = sizeof(struct sockaddr_in6);
5758	sin6.sin6_port = stcb->rport;
5759	if (altsa == NULL) {
5760		iph = mtod(m, struct ip *);
5761		switch (iph->ip_v) {
5762		case IPVERSION:
5763			{
5764				/* its IPv4 */
5765				struct sockaddr_in *sin_2;
5766
5767				sin_2 = (struct sockaddr_in *)(local_sa);
5768				memset(sin_2, 0, sizeof(sin));
5769				sin_2->sin_family = AF_INET;
5770				sin_2->sin_len = sizeof(sin);
5771				sin_2->sin_port = sh->dest_port;
5772				sin_2->sin_addr.s_addr = iph->ip_dst.s_addr;
5773				sin.sin_addr = iph->ip_src;
5774				sa = (struct sockaddr *)&sin;
5775				break;
5776			}
5777#ifdef INET6
5778		case IPV6_VERSION >> 4:
5779			{
5780				/* its IPv6 */
5781				struct ip6_hdr *ip6;
5782				struct sockaddr_in6 *sin6_2;
5783
5784				ip6 = mtod(m, struct ip6_hdr *);
5785				sin6_2 = (struct sockaddr_in6 *)(local_sa);
5786				memset(sin6_2, 0, sizeof(sin6));
5787				sin6_2->sin6_family = AF_INET6;
5788				sin6_2->sin6_len = sizeof(struct sockaddr_in6);
5789				sin6_2->sin6_port = sh->dest_port;
5790				sin6.sin6_addr = ip6->ip6_src;
5791				sa = (struct sockaddr *)&sin6;
5792				break;
5793			}
5794#endif
5795		default:
5796			sa = NULL;
5797			break;
5798		}
5799	} else {
5800		/*
5801		 * For cookies we use the src address NOT from the packet
5802		 * but from the original INIT
5803		 */
5804		sa = altsa;
5805	}
5806	/* Turn off ECN until we get through all params */
5807	stcb->asoc.ecn_allowed = 0;
5808	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5809		/* mark all addresses that we have currently on the list */
5810		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
5811	}
5812	/* does the source address already exist? if so skip it */
5813	l_inp = inp = stcb->sctp_ep;
5814
5815	atomic_add_int(&stcb->asoc.refcnt, 1);
5816	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, local_sa, stcb);
5817	atomic_add_int(&stcb->asoc.refcnt, -1);
5818
5819	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
5820		/* we must add the source address */
5821		/* no scope set here since we have a tcb already. */
5822		if ((sa->sa_family == AF_INET) &&
5823		    (stcb->asoc.ipv4_addr_legal)) {
5824			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
5825				return (-1);
5826			}
5827		} else if ((sa->sa_family == AF_INET6) &&
5828		    (stcb->asoc.ipv6_addr_legal)) {
5829			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
5830				return (-2);
5831			}
5832		}
5833	} else {
5834		if (net_tmp != NULL && stcb_tmp == stcb) {
5835			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
5836		} else if (stcb_tmp != stcb) {
5837			/* It belongs to another association? */
5838			if (stcb_tmp)
5839				SCTP_TCB_UNLOCK(stcb_tmp);
5840			return (-3);
5841		}
5842	}
5843	if (stcb->asoc.state == 0) {
5844		/* the assoc was freed? */
5845		return (-4);
5846	}
5847	/*
5848	 * peer must explicitly turn this on. This may have been initialized
5849	 * to be "on" in order to allow local addr changes while INIT's are
5850	 * in flight.
5851	 */
5852	stcb->asoc.peer_supports_asconf = 0;
5853	/* now we must go through each of the params. */
5854	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
5855	while (phdr) {
5856		ptype = ntohs(phdr->param_type);
5857		plen = ntohs(phdr->param_length);
5858		/*
5859		 * printf("ptype => %0x, plen => %d\n", (uint32_t)ptype,
5860		 * (int)plen);
5861		 */
5862		if (offset + plen > limit) {
5863			break;
5864		}
5865		if (plen == 0) {
5866			break;
5867		}
5868		if (ptype == SCTP_IPV4_ADDRESS) {
5869			if (stcb->asoc.ipv4_addr_legal) {
5870				struct sctp_ipv4addr_param *p4, p4_buf;
5871
5872				/* ok get the v4 address and check/add */
5873				phdr = sctp_get_next_param(m, offset,
5874				    (struct sctp_paramhdr *)&p4_buf,
5875				    sizeof(p4_buf));
5876				if (plen != sizeof(struct sctp_ipv4addr_param) ||
5877				    phdr == NULL) {
5878					return (-5);
5879				}
5880				p4 = (struct sctp_ipv4addr_param *)phdr;
5881				sin.sin_addr.s_addr = p4->addr;
5882				if (IN_MULTICAST(sin.sin_addr.s_addr)) {
5883					/* Skip multi-cast addresses */
5884					goto next_param;
5885				}
5886				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
5887				    (sin.sin_addr.s_addr == INADDR_ANY)) {
5888					goto next_param;
5889				}
5890				sa = (struct sockaddr *)&sin;
5891				inp = stcb->sctp_ep;
5892				atomic_add_int(&stcb->asoc.refcnt, 1);
5893				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
5894				    local_sa, stcb);
5895				atomic_add_int(&stcb->asoc.refcnt, -1);
5896
5897				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
5898				    inp == NULL) {
5899					/* we must add the source address */
5900					/*
5901					 * no scope set since we have a tcb
5902					 * already
5903					 */
5904
5905					/*
5906					 * we must validate the state again
5907					 * here
5908					 */
5909			add_it_now:
5910					if (stcb->asoc.state == 0) {
5911						/* the assoc was freed? */
5912						return (-7);
5913					}
5914					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
5915						return (-8);
5916					}
5917				} else if (stcb_tmp == stcb) {
5918					if (stcb->asoc.state == 0) {
5919						/* the assoc was freed? */
5920						return (-10);
5921					}
5922					if (net != NULL) {
5923						/* clear flag */
5924						net->dest_state &=
5925						    ~SCTP_ADDR_NOT_IN_ASSOC;
5926					}
5927				} else {
5928					/*
5929					 * strange, address is in another
5930					 * assoc? straighten out locks.
5931					 */
5932					if (stcb_tmp) {
5933						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
5934							/*
5935							 * in setup state we
5936							 * abort this guy
5937							 */
5938							sctp_abort_an_association(stcb_tmp->sctp_ep,
5939							    stcb_tmp, 1, NULL, 0);
5940							goto add_it_now;
5941						}
5942						SCTP_TCB_UNLOCK(stcb_tmp);
5943					}
5944					if (stcb->asoc.state == 0) {
5945						/* the assoc was freed? */
5946						return (-12);
5947					}
5948					return (-13);
5949				}
5950			}
5951		} else if (ptype == SCTP_IPV6_ADDRESS) {
5952			if (stcb->asoc.ipv6_addr_legal) {
5953				/* ok get the v6 address and check/add */
5954				struct sctp_ipv6addr_param *p6, p6_buf;
5955
5956				phdr = sctp_get_next_param(m, offset,
5957				    (struct sctp_paramhdr *)&p6_buf,
5958				    sizeof(p6_buf));
5959				if (plen != sizeof(struct sctp_ipv6addr_param) ||
5960				    phdr == NULL) {
5961					return (-14);
5962				}
5963				p6 = (struct sctp_ipv6addr_param *)phdr;
5964				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
5965				    sizeof(p6->addr));
5966				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
5967					/* Skip multi-cast addresses */
5968					goto next_param;
5969				}
5970				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
5971					/*
5972					 * Link local make no sense without
5973					 * scope
5974					 */
5975					goto next_param;
5976				}
5977				sa = (struct sockaddr *)&sin6;
5978				inp = stcb->sctp_ep;
5979				atomic_add_int(&stcb->asoc.refcnt, 1);
5980				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
5981				    local_sa, stcb);
5982				atomic_add_int(&stcb->asoc.refcnt, -1);
5983				if (stcb_tmp == NULL &&
5984				    (inp == stcb->sctp_ep || inp == NULL)) {
5985					/*
5986					 * we must validate the state again
5987					 * here
5988					 */
5989			add_it_now6:
5990					if (stcb->asoc.state == 0) {
5991						/* the assoc was freed? */
5992						return (-16);
5993					}
5994					/*
5995					 * we must add the address, no scope
5996					 * set
5997					 */
5998					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
5999						return (-17);
6000					}
6001				} else if (stcb_tmp == stcb) {
6002					/*
6003					 * we must validate the state again
6004					 * here
6005					 */
6006					if (stcb->asoc.state == 0) {
6007						/* the assoc was freed? */
6008						return (-19);
6009					}
6010					if (net != NULL) {
6011						/* clear flag */
6012						net->dest_state &=
6013						    ~SCTP_ADDR_NOT_IN_ASSOC;
6014					}
6015				} else {
6016					/*
6017					 * strange, address is in another
6018					 * assoc? straighten out locks.
6019					 */
6020					if (stcb_tmp)
6021						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6022							/*
6023							 * in setup state we
6024							 * abort this guy
6025							 */
6026							sctp_abort_an_association(stcb_tmp->sctp_ep,
6027							    stcb_tmp, 1, NULL, 0);
6028							goto add_it_now6;
6029						}
6030					SCTP_TCB_UNLOCK(stcb_tmp);
6031
6032					if (stcb->asoc.state == 0) {
6033						/* the assoc was freed? */
6034						return (-21);
6035					}
6036					return (-22);
6037				}
6038			}
6039		} else if (ptype == SCTP_ECN_CAPABLE) {
6040			stcb->asoc.ecn_allowed = 1;
6041		} else if (ptype == SCTP_ULP_ADAPTATION) {
6042			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6043				struct sctp_adaptation_layer_indication ai,
6044				                                *aip;
6045
6046				phdr = sctp_get_next_param(m, offset,
6047				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6048				aip = (struct sctp_adaptation_layer_indication *)phdr;
6049				if (aip) {
6050					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6051					stcb->asoc.adaptation_needed = 1;
6052				}
6053			}
6054		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6055			struct sctp_asconf_addr_param lstore, *fee;
6056			struct sctp_asconf_addrv4_param *fii;
6057			int lptype;
6058			struct sockaddr *lsa = NULL;
6059
6060			stcb->asoc.peer_supports_asconf = 1;
6061			if (plen > sizeof(lstore)) {
6062				return (-23);
6063			}
6064			phdr = sctp_get_next_param(m, offset,
6065			    (struct sctp_paramhdr *)&lstore,
6066			    min(plen, sizeof(lstore)));
6067			if (phdr == NULL) {
6068				return (-24);
6069			}
6070			fee = (struct sctp_asconf_addr_param *)phdr;
6071			lptype = ntohs(fee->addrp.ph.param_type);
6072			if (lptype == SCTP_IPV4_ADDRESS) {
6073				if (plen !=
6074				    sizeof(struct sctp_asconf_addrv4_param)) {
6075					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6076					    (int)sizeof(struct sctp_asconf_addrv4_param),
6077					    plen);
6078				} else {
6079					fii = (struct sctp_asconf_addrv4_param *)fee;
6080					sin.sin_addr.s_addr = fii->addrp.addr;
6081					lsa = (struct sockaddr *)&sin;
6082				}
6083			} else if (lptype == SCTP_IPV6_ADDRESS) {
6084				if (plen !=
6085				    sizeof(struct sctp_asconf_addr_param)) {
6086					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6087					    (int)sizeof(struct sctp_asconf_addr_param),
6088					    plen);
6089				} else {
6090					memcpy(sin6.sin6_addr.s6_addr,
6091					    fee->addrp.addr,
6092					    sizeof(fee->addrp.addr));
6093					lsa = (struct sockaddr *)&sin6;
6094				}
6095			}
6096			if (lsa) {
6097				(void)sctp_set_primary_addr(stcb, sa, NULL);
6098			}
6099		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6100			stcb->asoc.peer_supports_nat = 1;
6101		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6102			/* Peer supports pr-sctp */
6103			stcb->asoc.peer_supports_prsctp = 1;
6104		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6105			/* A supported extension chunk */
6106			struct sctp_supported_chunk_types_param *pr_supported;
6107			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6108			int num_ent, i;
6109
6110			phdr = sctp_get_next_param(m, offset,
6111			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6112			if (phdr == NULL) {
6113				return (-25);
6114			}
6115			stcb->asoc.peer_supports_asconf = 0;
6116			stcb->asoc.peer_supports_prsctp = 0;
6117			stcb->asoc.peer_supports_pktdrop = 0;
6118			stcb->asoc.peer_supports_strreset = 0;
6119			stcb->asoc.peer_supports_nr_sack = 0;
6120			stcb->asoc.peer_supports_auth = 0;
6121			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6122			num_ent = plen - sizeof(struct sctp_paramhdr);
6123			for (i = 0; i < num_ent; i++) {
6124				switch (pr_supported->chunk_types[i]) {
6125				case SCTP_ASCONF:
6126				case SCTP_ASCONF_ACK:
6127					stcb->asoc.peer_supports_asconf = 1;
6128					break;
6129				case SCTP_FORWARD_CUM_TSN:
6130					stcb->asoc.peer_supports_prsctp = 1;
6131					break;
6132				case SCTP_PACKET_DROPPED:
6133					stcb->asoc.peer_supports_pktdrop = 1;
6134					break;
6135				case SCTP_NR_SELECTIVE_ACK:
6136					if (SCTP_BASE_SYSCTL(sctp_nr_sack_on_off))
6137						stcb->asoc.peer_supports_nr_sack = 1;
6138					else
6139						stcb->asoc.peer_supports_nr_sack = 0;
6140					break;
6141				case SCTP_STREAM_RESET:
6142					stcb->asoc.peer_supports_strreset = 1;
6143					break;
6144				case SCTP_AUTHENTICATION:
6145					stcb->asoc.peer_supports_auth = 1;
6146					break;
6147				default:
6148					/* one I have not learned yet */
6149					break;
6150
6151				}
6152			}
6153		} else if (ptype == SCTP_ECN_NONCE_SUPPORTED) {
6154			/* Peer supports ECN-nonce */
6155			stcb->asoc.peer_supports_ecn_nonce = 1;
6156			stcb->asoc.ecn_nonce_allowed = 1;
6157		} else if (ptype == SCTP_RANDOM) {
6158			if (plen > sizeof(random_store))
6159				break;
6160			if (got_random) {
6161				/* already processed a RANDOM */
6162				goto next_param;
6163			}
6164			phdr = sctp_get_next_param(m, offset,
6165			    (struct sctp_paramhdr *)random_store,
6166			    min(sizeof(random_store), plen));
6167			if (phdr == NULL)
6168				return (-26);
6169			p_random = (struct sctp_auth_random *)phdr;
6170			random_len = plen - sizeof(*p_random);
6171			/* enforce the random length */
6172			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6173				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6174				return (-27);
6175			}
6176			got_random = 1;
6177		} else if (ptype == SCTP_HMAC_LIST) {
6178			int num_hmacs;
6179			int i;
6180
6181			if (plen > sizeof(hmacs_store))
6182				break;
6183			if (got_hmacs) {
6184				/* already processed a HMAC list */
6185				goto next_param;
6186			}
6187			phdr = sctp_get_next_param(m, offset,
6188			    (struct sctp_paramhdr *)hmacs_store,
6189			    min(plen, sizeof(hmacs_store)));
6190			if (phdr == NULL)
6191				return (-28);
6192			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6193			hmacs_len = plen - sizeof(*hmacs);
6194			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6195			/* validate the hmac list */
6196			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6197				return (-29);
6198			}
6199			if (stcb->asoc.peer_hmacs != NULL)
6200				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6201			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6202			if (stcb->asoc.peer_hmacs != NULL) {
6203				for (i = 0; i < num_hmacs; i++) {
6204					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6205					    ntohs(hmacs->hmac_ids[i]));
6206				}
6207			}
6208			got_hmacs = 1;
6209		} else if (ptype == SCTP_CHUNK_LIST) {
6210			int i;
6211
6212			if (plen > sizeof(chunks_store))
6213				break;
6214			if (got_chklist) {
6215				/* already processed a Chunks list */
6216				goto next_param;
6217			}
6218			phdr = sctp_get_next_param(m, offset,
6219			    (struct sctp_paramhdr *)chunks_store,
6220			    min(plen, sizeof(chunks_store)));
6221			if (phdr == NULL)
6222				return (-30);
6223			chunks = (struct sctp_auth_chunk_list *)phdr;
6224			num_chunks = plen - sizeof(*chunks);
6225			if (stcb->asoc.peer_auth_chunks != NULL)
6226				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6227			else
6228				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6229			for (i = 0; i < num_chunks; i++) {
6230				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6231				    stcb->asoc.peer_auth_chunks);
6232				/* record asconf/asconf-ack if listed */
6233				if (chunks->chunk_types[i] == SCTP_ASCONF)
6234					saw_asconf = 1;
6235				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6236					saw_asconf_ack = 1;
6237
6238			}
6239			got_chklist = 1;
6240		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6241			    (ptype == SCTP_STATE_COOKIE) ||
6242			    (ptype == SCTP_UNRECOG_PARAM) ||
6243			    (ptype == SCTP_COOKIE_PRESERVE) ||
6244			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6245			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6246			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6247			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6248		    (ptype == SCTP_SUCCESS_REPORT)) {
6249			 /* don't care */ ;
6250		} else {
6251			if ((ptype & 0x8000) == 0x0000) {
6252				/*
6253				 * must stop processing the rest of the
6254				 * param's. Any report bits were handled
6255				 * with the call to
6256				 * sctp_arethere_unrecognized_parameters()
6257				 * when the INIT or INIT-ACK was first seen.
6258				 */
6259				break;
6260			}
6261		}
6262
6263next_param:
6264		offset += SCTP_SIZE32(plen);
6265		if (offset >= limit) {
6266			break;
6267		}
6268		phdr = sctp_get_next_param(m, offset, &parm_buf,
6269		    sizeof(parm_buf));
6270	}
6271	/* Now check to see if we need to purge any addresses */
6272	for (net = TAILQ_FIRST(&stcb->asoc.nets); net != NULL; net = net_tmp) {
6273		net_tmp = TAILQ_NEXT(net, sctp_next);
6274		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6275		    SCTP_ADDR_NOT_IN_ASSOC) {
6276			/* This address has been removed from the asoc */
6277			/* remove and free it */
6278			stcb->asoc.numnets--;
6279			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6280			sctp_free_remote_addr(net);
6281			if (net == stcb->asoc.primary_destination) {
6282				stcb->asoc.primary_destination = NULL;
6283				sctp_select_primary_destination(stcb);
6284			}
6285		}
6286	}
6287	/* validate authentication required parameters */
6288	if (got_random && got_hmacs) {
6289		stcb->asoc.peer_supports_auth = 1;
6290	} else {
6291		stcb->asoc.peer_supports_auth = 0;
6292	}
6293	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6294		/* peer does not support auth but sent a chunks list? */
6295		return (-31);
6296	}
6297	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6298	    !stcb->asoc.peer_supports_auth) {
6299		/* peer supports asconf but not auth? */
6300		return (-32);
6301	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6302	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6303		return (-33);
6304	}
6305	/* concatenate the full random key */
6306	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6307	if (chunks != NULL) {
6308		keylen += sizeof(*chunks) + num_chunks;
6309	}
6310	new_key = sctp_alloc_key(keylen);
6311	if (new_key != NULL) {
6312		/* copy in the RANDOM */
6313		if (p_random != NULL) {
6314			keylen = sizeof(*p_random) + random_len;
6315			bcopy(p_random, new_key->key, keylen);
6316		}
6317		/* append in the AUTH chunks */
6318		if (chunks != NULL) {
6319			bcopy(chunks, new_key->key + keylen,
6320			    sizeof(*chunks) + num_chunks);
6321			keylen += sizeof(*chunks) + num_chunks;
6322		}
6323		/* append in the HMACs */
6324		if (hmacs != NULL) {
6325			bcopy(hmacs, new_key->key + keylen,
6326			    sizeof(*hmacs) + hmacs_len);
6327		}
6328	} else {
6329		/* failed to get memory for the key */
6330		return (-34);
6331	}
6332	if (stcb->asoc.authinfo.peer_random != NULL)
6333		sctp_free_key(stcb->asoc.authinfo.peer_random);
6334	stcb->asoc.authinfo.peer_random = new_key;
6335	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6336	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6337
6338	return (0);
6339}
6340
6341int
6342sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6343    struct sctp_nets *net)
6344{
6345	/* make sure the requested primary address exists in the assoc */
6346	if (net == NULL && sa)
6347		net = sctp_findnet(stcb, sa);
6348
6349	if (net == NULL) {
6350		/* didn't find the requested primary address! */
6351		return (-1);
6352	} else {
6353		/* set the primary address */
6354		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6355			/* Must be confirmed, so queue to set */
6356			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6357			return (0);
6358		}
6359		stcb->asoc.primary_destination = net;
6360		net->dest_state &= ~SCTP_ADDR_WAS_PRIMARY;
6361		net = TAILQ_FIRST(&stcb->asoc.nets);
6362		if (net != stcb->asoc.primary_destination) {
6363			/*
6364			 * first one on the list is NOT the primary
6365			 * sctp_cmpaddr() is much more efficent if the
6366			 * primary is the first on the list, make it so.
6367			 */
6368			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6369			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6370		}
6371		return (0);
6372	}
6373}
6374
6375int
6376sctp_is_vtag_good(struct sctp_inpcb *inp, uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now, int save_in_twait)
6377{
6378	/*
6379	 * This function serves two purposes. It will see if a TAG can be
6380	 * re-used and return 1 for yes it is ok and 0 for don't use that
6381	 * tag. A secondary function it will do is purge out old tags that
6382	 * can be removed.
6383	 */
6384	struct sctpvtaghead *chain;
6385	struct sctp_tagblock *twait_block;
6386	struct sctpasochead *head;
6387	struct sctp_tcb *stcb;
6388	int i;
6389
6390	SCTP_INP_INFO_RLOCK();
6391	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6392	    SCTP_BASE_INFO(hashasocmark))];
6393	if (head == NULL) {
6394		/* invalid vtag */
6395		goto skip_vtag_check;
6396	}
6397	LIST_FOREACH(stcb, head, sctp_asocs) {
6398		/*
6399		 * We choose not to lock anything here. TCB's can't be
6400		 * removed since we have the read lock, so they can't be
6401		 * freed on us, same thing for the INP. I may be wrong with
6402		 * this assumption, but we will go with it for now :-)
6403		 */
6404		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6405			continue;
6406		}
6407		if (stcb->asoc.my_vtag == tag) {
6408			/* candidate */
6409			if (stcb->rport != rport) {
6410				continue;
6411			}
6412			if (stcb->sctp_ep->sctp_lport != lport) {
6413				continue;
6414			}
6415			/* Its a used tag set */
6416			SCTP_INP_INFO_WUNLOCK();
6417			return (0);
6418		}
6419	}
6420skip_vtag_check:
6421
6422	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6423	/* Now what about timed wait ? */
6424	if (!LIST_EMPTY(chain)) {
6425		/*
6426		 * Block(s) are present, lets see if we have this tag in the
6427		 * list
6428		 */
6429		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6430			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6431				if (twait_block->vtag_block[i].v_tag == 0) {
6432					/* not used */
6433					continue;
6434				} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6435				    now->tv_sec) {
6436					/* Audit expires this guy */
6437					twait_block->vtag_block[i].tv_sec_at_expire = 0;
6438					twait_block->vtag_block[i].v_tag = 0;
6439					twait_block->vtag_block[i].lport = 0;
6440					twait_block->vtag_block[i].rport = 0;
6441				} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6442					    (twait_block->vtag_block[i].lport == lport) &&
6443				    (twait_block->vtag_block[i].rport == rport)) {
6444					/* Bad tag, sorry :< */
6445					SCTP_INP_INFO_WUNLOCK();
6446					return (0);
6447				}
6448			}
6449		}
6450	}
6451	SCTP_INP_INFO_RUNLOCK();
6452	return (1);
6453}
6454
6455
6456static sctp_assoc_t reneged_asoc_ids[256];
6457static uint8_t reneged_at = 0;
6458
6459
6460static void
6461sctp_drain_mbufs(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
6462{
6463	/*
6464	 * We must hunt this association for MBUF's past the cumack (i.e.
6465	 * out of order data that we can renege on).
6466	 */
6467	struct sctp_association *asoc;
6468	struct sctp_tmit_chunk *chk, *nchk;
6469	uint32_t cumulative_tsn_p1;
6470	struct sctp_queued_to_read *ctl, *nctl;
6471	int cnt, strmat;
6472	uint32_t gap, i;
6473	int fnd = 0;
6474
6475	/* We look for anything larger than the cum-ack + 1 */
6476
6477	asoc = &stcb->asoc;
6478	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6479		/* none we can reneg on. */
6480		return;
6481	}
6482	SCTP_STAT_INCR(sctps_protocol_drains_done);
6483	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6484	cnt = 0;
6485	/* First look in the re-assembly queue */
6486	chk = TAILQ_FIRST(&asoc->reasmqueue);
6487	while (chk) {
6488		/* Get the next one */
6489		nchk = TAILQ_NEXT(chk, sctp_next);
6490		if (compare_with_wrap(chk->rec.data.TSN_seq,
6491		    cumulative_tsn_p1, MAX_TSN)) {
6492			/* Yep it is above cum-ack */
6493			cnt++;
6494			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6495			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6496			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6497			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6498			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6499			if (chk->data) {
6500				sctp_m_freem(chk->data);
6501				chk->data = NULL;
6502			}
6503			sctp_free_a_chunk(stcb, chk);
6504		}
6505		chk = nchk;
6506	}
6507	/* Ok that was fun, now we will drain all the inbound streams? */
6508	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6509		ctl = TAILQ_FIRST(&asoc->strmin[strmat].inqueue);
6510		while (ctl) {
6511			nctl = TAILQ_NEXT(ctl, next);
6512			if (compare_with_wrap(ctl->sinfo_tsn,
6513			    cumulative_tsn_p1, MAX_TSN)) {
6514				/* Yep it is above cum-ack */
6515				cnt++;
6516				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6517				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6518				sctp_ucount_decr(asoc->cnt_on_all_streams);
6519				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6520				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6521				if (ctl->data) {
6522					sctp_m_freem(ctl->data);
6523					ctl->data = NULL;
6524				}
6525				sctp_free_remote_addr(ctl->whoFrom);
6526				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6527				SCTP_DECR_READQ_COUNT();
6528			}
6529			ctl = nctl;
6530		}
6531	}
6532	if (cnt) {
6533		/* We must back down to see what the new highest is */
6534		for (i = asoc->highest_tsn_inside_map;
6535		    (compare_with_wrap(i, asoc->mapping_array_base_tsn, MAX_TSN) || (i == asoc->mapping_array_base_tsn));
6536		    i--) {
6537			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6538			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6539				asoc->highest_tsn_inside_map = i;
6540				fnd = 1;
6541				break;
6542			}
6543		}
6544		if (!fnd) {
6545			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6546		}
6547		/*
6548		 * Question, should we go through the delivery queue? The
6549		 * only reason things are on here is the app not reading OR
6550		 * a p-d-api up. An attacker COULD send enough in to
6551		 * initiate the PD-API and then send a bunch of stuff to
6552		 * other streams... these would wind up on the delivery
6553		 * queue.. and then we would not get to them. But in order
6554		 * to do this I then have to back-track and un-deliver
6555		 * sequence numbers in streams.. el-yucko. I think for now
6556		 * we will NOT look at the delivery queue and leave it to be
6557		 * something to consider later. An alternative would be to
6558		 * abort the P-D-API with a notification and then deliver
6559		 * the data.... Or another method might be to keep track of
6560		 * how many times the situation occurs and if we see a
6561		 * possible attack underway just abort the association.
6562		 */
6563#ifdef SCTP_DEBUG
6564		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6565#endif
6566		/*
6567		 * Now do we need to find a new
6568		 * asoc->highest_tsn_inside_map?
6569		 */
6570		asoc->last_revoke_count = cnt;
6571		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6572		/* sa_ignore NO_NULL_CHK */
6573		sctp_send_sack(stcb);
6574		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6575		reneged_asoc_ids[reneged_at] = sctp_get_associd(stcb);
6576		reneged_at++;
6577	}
6578	/*
6579	 * Another issue, in un-setting the TSN's in the mapping array we
6580	 * DID NOT adjust the higest_tsn marker.  This will cause one of two
6581	 * things to occur. It may cause us to do extra work in checking for
6582	 * our mapping array movement. More importantly it may cause us to
6583	 * SACK every datagram. This may not be a bad thing though since we
6584	 * will recover once we get our cum-ack above and all this stuff we
6585	 * dumped recovered.
6586	 */
6587}
6588
6589void
6590sctp_drain()
6591{
6592	/*
6593	 * We must walk the PCB lists for ALL associations here. The system
6594	 * is LOW on MBUF's and needs help. This is where reneging will
6595	 * occur. We really hope this does NOT happen!
6596	 */
6597	VNET_ITERATOR_DECL(vnet_iter);
6598	VNET_LIST_RLOCK_NOSLEEP();
6599	VNET_FOREACH(vnet_iter) {
6600		CURVNET_SET(vnet_iter);
6601		struct sctp_inpcb *inp;
6602		struct sctp_tcb *stcb;
6603
6604		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6605		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6606#ifdef VIMAGE
6607			continue;
6608#else
6609			return;
6610#endif
6611		}
6612		SCTP_INP_INFO_RLOCK();
6613		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6614			/* For each endpoint */
6615			SCTP_INP_RLOCK(inp);
6616			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6617				/* For each association */
6618				SCTP_TCB_LOCK(stcb);
6619				sctp_drain_mbufs(inp, stcb);
6620				SCTP_TCB_UNLOCK(stcb);
6621			}
6622			SCTP_INP_RUNLOCK(inp);
6623		}
6624		SCTP_INP_INFO_RUNLOCK();
6625		CURVNET_RESTORE();
6626	}
6627	VNET_LIST_RUNLOCK_NOSLEEP();
6628}
6629
6630/*
6631 * start a new iterator
6632 * iterates through all endpoints and associations based on the pcb_state
6633 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6634 * assocs and "ef" (optional) is executed when the iterator completes.
6635 * "inpf" (optional) is executed for each new endpoint as it is being
6636 * iterated through. inpe (optional) is called when the inp completes
6637 * its way through all the stcbs.
6638 */
6639int
6640sctp_initiate_iterator(inp_func inpf,
6641    asoc_func af,
6642    inp_func inpe,
6643    uint32_t pcb_state,
6644    uint32_t pcb_features,
6645    uint32_t asoc_state,
6646    void *argp,
6647    uint32_t argi,
6648    end_func ef,
6649    struct sctp_inpcb *s_inp,
6650    uint8_t chunk_output_off)
6651{
6652	struct sctp_iterator *it = NULL;
6653
6654	if (af == NULL) {
6655		return (-1);
6656	}
6657	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6658	    SCTP_M_ITER);
6659	if (it == NULL) {
6660		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6661		return (ENOMEM);
6662	}
6663	memset(it, 0, sizeof(*it));
6664	it->function_assoc = af;
6665	it->function_inp = inpf;
6666	if (inpf)
6667		it->done_current_ep = 0;
6668	else
6669		it->done_current_ep = 1;
6670	it->function_atend = ef;
6671	it->pointer = argp;
6672	it->val = argi;
6673	it->pcb_flags = pcb_state;
6674	it->pcb_features = pcb_features;
6675	it->asoc_state = asoc_state;
6676	it->function_inp_end = inpe;
6677	it->no_chunk_output = chunk_output_off;
6678	it->vn = curvnet;
6679	if (s_inp) {
6680		it->inp = s_inp;
6681		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
6682	} else {
6683		SCTP_INP_INFO_RLOCK();
6684		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
6685
6686		SCTP_INP_INFO_RUNLOCK();
6687		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
6688
6689	}
6690	SCTP_IPI_ITERATOR_WQ_LOCK();
6691	if (it->inp) {
6692		SCTP_INP_INCR_REF(it->inp);
6693	}
6694	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
6695	if (sctp_it_ctl.iterator_running == 0) {
6696		sctp_wakeup_iterator();
6697	}
6698	SCTP_IPI_ITERATOR_WQ_UNLOCK();
6699	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
6700	return (0);
6701}
6702