sctp_pcb.c revision 169382
1178786Skmacy/*-
2178786Skmacy * Copyright (c) 2001-2007, by Cisco Systems, Inc. All rights reserved.
3178786Skmacy *
4178786Skmacy * Redistribution and use in source and binary forms, with or without
5178786Skmacy * modification, are permitted provided that the following conditions are met:
6178786Skmacy *
7178786Skmacy * a) Redistributions of source code must retain the above copyright notice,
8178786Skmacy *   this list of conditions and the following disclaimer.
9178786Skmacy *
10178786Skmacy * b) Redistributions in binary form must reproduce the above copyright
11178786Skmacy *    notice, this list of conditions and the following disclaimer in
12178786Skmacy *   the documentation and/or other materials provided with the distribution.
13178786Skmacy *
14178786Skmacy * c) Neither the name of Cisco Systems, Inc. nor the names of its
15178786Skmacy *    contributors may be used to endorse or promote products derived
16178786Skmacy *    from this software without specific prior written permission.
17178786Skmacy *
18178786Skmacy * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
19178786Skmacy * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
20178786Skmacy * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21178786Skmacy * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
22178786Skmacy * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
23178786Skmacy * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
24178786Skmacy * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
25178786Skmacy * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
26178786Skmacy * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
27178786Skmacy * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
28178786Skmacy * THE POSSIBILITY OF SUCH DAMAGE.
29178786Skmacy */
30178786Skmacy
31178786Skmacy/* $KAME: sctp_pcb.c,v 1.38 2005/03/06 16:04:18 itojun Exp $	 */
32178786Skmacy
33178786Skmacy#include <sys/cdefs.h>
34178786Skmacy__FBSDID("$FreeBSD: head/sys/netinet/sctp_pcb.c 169382 2007-05-08 17:01:12Z rrs $");
35178786Skmacy
36178786Skmacy#include <netinet/sctp_os.h>
37178786Skmacy#include <sys/proc.h>
38178786Skmacy#include <netinet/sctp_var.h>
39178786Skmacy#include <netinet/sctp_sysctl.h>
40178786Skmacy#include <netinet/sctp_pcb.h>
41178786Skmacy#include <netinet/sctputil.h>
42178786Skmacy#include <netinet/sctp.h>
43178786Skmacy#include <netinet/sctp_header.h>
44178786Skmacy#include <netinet/sctp_asconf.h>
45178786Skmacy#include <netinet/sctp_output.h>
46178786Skmacy#include <netinet/sctp_timer.h>
47178786Skmacy#include <netinet/sctp_bsd_addr.h>
48237263Snp
49237263Snp
50178786Skmacystruct sctp_epinfo sctppcbinfo;
51237263Snp
52237263Snp/* FIX: we don't handle multiple link local scopes */
53178786Skmacy/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
54178786Skmacyint
55178786SkmacySCTP6_ARE_ADDR_EQUAL(struct in6_addr *a, struct in6_addr *b)
56178786Skmacy{
57178786Skmacy	struct in6_addr tmp_a, tmp_b;
58178786Skmacy
59178786Skmacy	/* use a copy of a and b */
60178786Skmacy	tmp_a = *a;
61178786Skmacy	tmp_b = *b;
62178786Skmacy	in6_clearscope(&tmp_a);
63178786Skmacy	in6_clearscope(&tmp_b);
64237263Snp	return (IN6_ARE_ADDR_EQUAL(&tmp_a, &tmp_b));
65178786Skmacy}
66178786Skmacy
67178786Skmacyvoid
68178786Skmacysctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
69178786Skmacy{
70178786Skmacy	/*
71178786Skmacy	 * We really don't need to lock this, but I will just because it
72178786Skmacy	 * does not hurt.
73178786Skmacy	 */
74178786Skmacy	SCTP_INP_INFO_RLOCK();
75178786Skmacy	spcb->ep_count = sctppcbinfo.ipi_count_ep;
76178786Skmacy	spcb->asoc_count = sctppcbinfo.ipi_count_asoc;
77178786Skmacy	spcb->laddr_count = sctppcbinfo.ipi_count_laddr;
78178786Skmacy	spcb->raddr_count = sctppcbinfo.ipi_count_raddr;
79178786Skmacy	spcb->chk_count = sctppcbinfo.ipi_count_chunk;
80178786Skmacy	spcb->readq_count = sctppcbinfo.ipi_count_readq;
81178786Skmacy	spcb->stream_oque = sctppcbinfo.ipi_count_strmoq;
82178786Skmacy	spcb->free_chunks = sctppcbinfo.ipi_free_chunks;
83178786Skmacy
84178786Skmacy	SCTP_INP_INFO_RUNLOCK();
85178786Skmacy}
86178786Skmacy
87178786Skmacy/*
88178786Skmacy * Addresses are added to VRF's (Virtual Router's). For BSD we
89237263Snp * have only the default VRF 0. We maintain a hash list of
90178786Skmacy * VRF's. Each VRF has its own list of sctp_ifn's. Each of
91178786Skmacy * these has a list of addresses. When we add a new address
92178786Skmacy * to a VRF we lookup the ifn/ifn_index, if the ifn does
93178786Skmacy * not exist we create it and add it to the list of IFN's
94178786Skmacy * within the VRF. Once we have the sctp_ifn, we add the
95178786Skmacy * address to the list. So we look something like:
96178786Skmacy *
97178786Skmacy * hash-vrf-table
98178786Skmacy *   vrf-> ifn-> ifn -> ifn
99178786Skmacy *   vrf    |
100178786Skmacy *    ...   +--ifa-> ifa -> ifa
101178786Skmacy *   vrf
102178786Skmacy *
103178786Skmacy * We keep these seperate lists since the SCTP subsystem will
104178786Skmacy * point to these from its source address selection nets structure.
105178786Skmacy * When an address is deleted it does not happen right away on
106178786Skmacy * the SCTP side, it gets scheduled. What we do when a
107178786Skmacy * delete happens is immediately remove the address from
108178786Skmacy * the master list and decrement the refcount. As our
109178786Skmacy * addip iterator works through and frees the src address
110178786Skmacy * selection pointing to the sctp_ifa, eventually the refcount
111178786Skmacy * will reach 0 and we will delete it. Note that it is assumed
112178786Skmacy * that any locking on system level ifn/ifa is done at the
113178786Skmacy * caller of these functions and these routines will only
114178786Skmacy * lock the SCTP structures as they add or delete things.
115178786Skmacy *
116178786Skmacy * Other notes on VRF concepts.
117178786Skmacy *  - An endpoint can be in multiple VRF's
118178786Skmacy *  - An association lives within a VRF and only one VRF.
119178786Skmacy *  - Any incoming packet we can deduce the VRF for by
120178786Skmacy *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
121178786Skmacy *  - Any downward send call or connect call must supply the
122178786Skmacy *    VRF via ancillary data or via some sort of set default
123178786Skmacy *    VRF socket option call (again for BSD no brainer since
124178786Skmacy *    the VRF is always 0).
125178786Skmacy *  - An endpoint may add multiple VRF's to it.
126178786Skmacy *  - Listening sockets can accept associations in any
127178786Skmacy *    of the VRF's they are in but the assoc will end up
128178786Skmacy *    in only one VRF (gotten from the packet or connect/send).
129178786Skmacy *
130178786Skmacy */
131178786Skmacy
132178786Skmacystruct sctp_vrf *
133178786Skmacysctp_allocate_vrf(int vrf_id)
134178786Skmacy{
135178786Skmacy	struct sctp_vrf *vrf = NULL;
136178786Skmacy	struct sctp_vrflist *bucket;
137178786Skmacy
138178786Skmacy	/* First allocate the VRF structure */
139178786Skmacy	vrf = sctp_find_vrf(vrf_id);
140237263Snp	if (vrf) {
141178786Skmacy		/* Already allocated */
142178786Skmacy		return (vrf);
143178786Skmacy	}
144178786Skmacy	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
145178786Skmacy	    "SCTP_VRF");
146178786Skmacy	if (vrf == NULL) {
147178786Skmacy		/* No memory */
148178786Skmacy#ifdef INVARIANTS
149237263Snp		panic("No memory for VRF:%d", vrf_id);
150237263Snp#endif
151178786Skmacy		return (NULL);
152178786Skmacy	}
153237263Snp	/* setup the VRF */
154178786Skmacy	memset(vrf, 0, sizeof(struct sctp_vrf));
155178786Skmacy	vrf->vrf_id = vrf_id;
156237263Snp	LIST_INIT(&vrf->ifnlist);
157178786Skmacy	vrf->total_ifa_count = 0;
158178786Skmacy	/* Init the HASH of addresses */
159178786Skmacy	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
160178786Skmacy	    &vrf->vrf_addr_hashmark);
161237263Snp	if (vrf->vrf_addr_hash == NULL) {
162237263Snp		/* No memory */
163178786Skmacy#ifdef INVARIANTS
164178786Skmacy		panic("No memory for VRF:%d", vrf_id);
165237263Snp#endif
166237263Snp		SCTP_FREE(vrf);
167178786Skmacy		return (NULL);
168237263Snp	}
169237263Snp	vrf->vrf_ifn_hash = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
170178786Skmacy	    &vrf->vrf_ifn_hashmark);
171178786Skmacy	if (vrf->vrf_ifn_hash == NULL) {
172178786Skmacy		/* No memory */
173178786Skmacy#ifdef INVARIANTS
174178786Skmacy		panic("No memory for VRF:%d", vrf_id);
175178786Skmacy#endif
176178786Skmacy		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
177178786Skmacy		SCTP_FREE(vrf);
178237263Snp		return (NULL);
179178786Skmacy	}
180178786Skmacy	/* Add it to the hash table */
181178786Skmacy	bucket = &sctppcbinfo.sctp_vrfhash[(vrf_id & sctppcbinfo.hashvrfmark)];
182178786Skmacy	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
183178786Skmacy	atomic_add_int(&sctppcbinfo.ipi_count_vrfs, 1);
184178786Skmacy	return (vrf);
185178786Skmacy}
186178786Skmacy
187178786Skmacy
188178786Skmacystruct sctp_ifn *
189178786Skmacysctp_find_ifn(struct sctp_vrf *vrf, void *ifn, uint32_t ifn_index)
190178786Skmacy{
191178786Skmacy	struct sctp_ifn *sctp_ifnp;
192178786Skmacy	struct sctp_ifnlist *hash_ifn_head;
193178786Skmacy
194178786Skmacy	/*
195178786Skmacy	 * We assume the lock is held for the addresses if thats wrong
196178786Skmacy	 * problems could occur :-)
197178786Skmacy	 */
198178786Skmacy	hash_ifn_head = &vrf->vrf_ifn_hash[(ifn_index & vrf->vrf_ifn_hashmark)];
199178786Skmacy	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
200178786Skmacy		if (sctp_ifnp->ifn_index == ifn_index) {
201178786Skmacy			return (sctp_ifnp);
202178786Skmacy		}
203178786Skmacy		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
204178786Skmacy			return (sctp_ifnp);
205178786Skmacy		}
206237263Snp	}
207178786Skmacy	return (NULL);
208178786Skmacy}
209178786Skmacy
210178786Skmacy
211178786Skmacy
212178786Skmacystruct sctp_vrf *
213178786Skmacysctp_find_vrf(uint32_t vrf_id)
214178786Skmacy{
215178786Skmacy	struct sctp_vrflist *bucket;
216178786Skmacy	struct sctp_vrf *liste;
217178786Skmacy
218178786Skmacy	bucket = &sctppcbinfo.sctp_vrfhash[(vrf_id & sctppcbinfo.hashvrfmark)];
219178786Skmacy	LIST_FOREACH(liste, bucket, next_vrf) {
220178786Skmacy		if (vrf_id == liste->vrf_id) {
221178786Skmacy			return (liste);
222178786Skmacy		}
223178786Skmacy	}
224178786Skmacy	return (NULL);
225178786Skmacy}
226178786Skmacy
227178786Skmacy
228178786Skmacyvoid
229178786Skmacysctp_free_ifn(struct sctp_ifn *sctp_ifnp)
230178786Skmacy{
231178786Skmacy	int ret;
232178786Skmacy
233178786Skmacy	ret = atomic_fetchadd_int(&sctp_ifnp->refcount, -1);
234178786Skmacy	if (ret == 1) {
235178786Skmacy		/* We zero'd the count */
236178786Skmacy		SCTP_FREE(sctp_ifnp);
237178786Skmacy		atomic_subtract_int(&sctppcbinfo.ipi_count_ifns, 1);
238178786Skmacy	}
239178786Skmacy}
240178786Skmacy
241178786Skmacyvoid
242178786Skmacysctp_update_ifn_mtu(uint32_t vrf_id, uint32_t ifn_index, uint32_t mtu)
243178786Skmacy{
244178786Skmacy	struct sctp_ifn *sctp_ifnp;
245178786Skmacy	struct sctp_vrf *vrf;
246178786Skmacy
247178786Skmacy	vrf = sctp_find_vrf(vrf_id);
248178786Skmacy	if (vrf == NULL)
249178786Skmacy		return;
250178786Skmacy	sctp_ifnp = sctp_find_ifn(vrf, (void *)NULL, ifn_index);
251178786Skmacy	if (sctp_ifnp != NULL) {
252178786Skmacy		sctp_ifnp->ifn_mtu = mtu;
253178786Skmacy	}
254178786Skmacy}
255178786Skmacy
256178786Skmacy
257178786Skmacyvoid
258178786Skmacysctp_free_ifa(struct sctp_ifa *sctp_ifap)
259178786Skmacy{
260178786Skmacy	int ret;
261178786Skmacy
262178786Skmacy	ret = atomic_fetchadd_int(&sctp_ifap->refcount, -1);
263178786Skmacy	if (ret == 1) {
264178786Skmacy		/* We zero'd the count */
265178786Skmacy		SCTP_FREE(sctp_ifap);
266178786Skmacy		atomic_subtract_int(&sctppcbinfo.ipi_count_ifas, 1);
267178786Skmacy	}
268178786Skmacy}
269178786Skmacy
270178786Skmacystatic void
271178786Skmacysctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
272237263Snp{
273178786Skmacy	struct sctp_ifn *found;
274178786Skmacy
275	found = sctp_find_ifn(sctp_ifnp->vrf, sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
276	if (found == NULL) {
277		/* Not in the list.. sorry */
278		return;
279	}
280	if (hold_addr_lock == 0)
281		SCTP_IPI_ADDR_LOCK();
282	LIST_REMOVE(sctp_ifnp, next_bucket);
283	LIST_REMOVE(sctp_ifnp, next_ifn);
284	if (hold_addr_lock == 0)
285		SCTP_IPI_ADDR_UNLOCK();
286	/* Take away the reference, and possibly free it */
287	sctp_free_ifn(sctp_ifnp);
288}
289
290
291struct sctp_ifa *
292sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
293    uint32_t ifn_type, const char *if_name,
294    void *ifa, struct sockaddr *addr, uint32_t ifa_flags,
295    int dynamic_add)
296{
297	struct sctp_vrf *vrf;
298	struct sctp_ifn *sctp_ifnp = NULL;
299	struct sctp_ifa *sctp_ifap = NULL;
300	struct sctp_ifalist *hash_addr_head;
301	struct sctp_ifnlist *hash_ifn_head;
302	uint32_t hash_of_addr;
303
304	/* How granular do we need the locks to be here? */
305	SCTP_IPI_ADDR_LOCK();
306	vrf = sctp_find_vrf(vrf_id);
307	if (vrf == NULL) {
308		vrf = sctp_allocate_vrf(vrf_id);
309		if (vrf == NULL) {
310			SCTP_IPI_ADDR_UNLOCK();
311			return (NULL);
312		}
313	}
314	sctp_ifnp = sctp_find_ifn(vrf, ifn, ifn_index);
315	if (sctp_ifnp == NULL) {
316		/*
317		 * build one and add it, can't hold lock until after malloc
318		 * done though.
319		 */
320		SCTP_IPI_ADDR_UNLOCK();
321		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *, sizeof(struct sctp_ifn), "SCTP_IFN");
322		if (sctp_ifnp == NULL) {
323#ifdef INVARIANTS
324			panic("No memory for IFN:%u", sctp_ifnp->ifn_index);
325#endif
326			return (NULL);
327		}
328		sctp_ifnp->ifn_index = ifn_index;
329		sctp_ifnp->ifn_p = ifn;
330		sctp_ifnp->ifn_type = ifn_type;
331		sctp_ifnp->ifa_count = 0;
332		sctp_ifnp->refcount = 1;
333		sctp_ifnp->vrf = vrf;
334		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index);
335		if (if_name != NULL) {
336			memcpy(sctp_ifnp->ifn_name, if_name, SCTP_IFNAMSIZ);
337		} else {
338			memcpy(sctp_ifnp->ifn_name, "unknown", min(7, SCTP_IFNAMSIZ));
339		}
340		hash_ifn_head = &vrf->vrf_ifn_hash[(ifn_index & vrf->vrf_ifn_hashmark)];
341		LIST_INIT(&sctp_ifnp->ifalist);
342		SCTP_IPI_ADDR_LOCK();
343		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
344		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
345		atomic_add_int(&sctppcbinfo.ipi_count_ifns, 1);
346	}
347	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, 1);
348	if (sctp_ifap) {
349		/* Hmm, it already exists? */
350		if ((sctp_ifap->ifn_p) &&
351		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
352			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
353				/* easy to solve, just switch back to active */
354				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
355				sctp_ifap->ifn_p = sctp_ifnp;
356		exit_stage_left:
357				SCTP_IPI_ADDR_UNLOCK();
358				return (sctp_ifap);
359			} else {
360				goto exit_stage_left;
361			}
362		} else {
363			if (sctp_ifap->ifn_p) {
364				/*
365				 * The first IFN gets the address,
366				 * duplicates are ignored.
367				 */
368				goto exit_stage_left;
369			} else {
370				/* repair ifnp which was NULL ? */
371				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
372				sctp_ifap->ifn_p = sctp_ifnp;
373				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
374			}
375			goto exit_stage_left;
376		}
377	}
378	SCTP_IPI_ADDR_UNLOCK();
379	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), "SCTP_IFA");
380	if (sctp_ifap == NULL) {
381#ifdef INVARIANTS
382		panic("No memory for IFA");
383#endif
384		return (NULL);
385	}
386	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
387	sctp_ifap->ifn_p = sctp_ifnp;
388	atomic_add_int(&sctp_ifnp->refcount, 1);
389
390	sctp_ifap->ifa = ifa;
391	memcpy(&sctp_ifap->address, addr, addr->sa_len);
392	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
393	sctp_ifap->flags = ifa_flags;
394	/* Set scope */
395	if (sctp_ifap->address.sa.sa_family == AF_INET) {
396		struct sockaddr_in *sin;
397
398		sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
399		if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
400		    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
401			sctp_ifap->src_is_loop = 1;
402		}
403		if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
404			sctp_ifap->src_is_priv = 1;
405		}
406	} else if (sctp_ifap->address.sa.sa_family == AF_INET6) {
407		/* ok to use deprecated addresses? */
408		struct sockaddr_in6 *sin6;
409
410		sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
411		if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
412		    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
413			sctp_ifap->src_is_loop = 1;
414		}
415		if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
416			sctp_ifap->src_is_priv = 1;
417		}
418	}
419	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
420
421	if ((sctp_ifap->src_is_priv == 0) &&
422	    (sctp_ifap->src_is_loop == 0)) {
423		sctp_ifap->src_is_glob = 1;
424	}
425	SCTP_IPI_ADDR_LOCK();
426	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
427	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
428	sctp_ifap->refcount = 1;
429	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
430	sctp_ifnp->ifa_count++;
431	vrf->total_ifa_count++;
432	atomic_add_int(&sctppcbinfo.ipi_count_ifas, 1);
433	SCTP_IPI_ADDR_UNLOCK();
434	if (dynamic_add) {
435		/*
436		 * Bump up the refcount so that when the timer completes it
437		 * will drop back down.
438		 */
439		struct sctp_laddr *wi;
440
441		atomic_add_int(&sctp_ifap->refcount, 1);
442		wi = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
443		if (wi == NULL) {
444			/*
445			 * Gak, what can we do? We have lost an address
446			 * change can you say HOSED?
447			 */
448#ifdef SCTP_DEBUG
449			if (sctp_debug_on & SCTP_DEBUG_PCB1) {
450				printf("Lost and address change ???\n");
451			}
452#endif				/* SCTP_DEBUG */
453			/* Opps, must decrement the count */
454			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index);
455			return (NULL);
456		}
457		SCTP_INCR_LADDR_COUNT();
458		bzero(wi, sizeof(*wi));
459		wi->ifa = sctp_ifap;
460		wi->action = SCTP_ADD_IP_ADDRESS;
461		SCTP_IPI_ITERATOR_WQ_LOCK();
462		/*
463		 * Should this really be a tailq? As it is we will process
464		 * the newest first :-0
465		 */
466		LIST_INSERT_HEAD(&sctppcbinfo.addr_wq, wi, sctp_nxt_addr);
467		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
468		    (struct sctp_inpcb *)NULL,
469		    (struct sctp_tcb *)NULL,
470		    (struct sctp_nets *)NULL);
471		SCTP_IPI_ITERATOR_WQ_UNLOCK();
472	} else {
473		/* it's ready for use */
474		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
475	}
476	return (sctp_ifap);
477}
478
479void
480sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
481    uint32_t ifn_index)
482{
483	struct sctp_vrf *vrf;
484	struct sctp_ifa *sctp_ifap = NULL;
485
486	SCTP_IPI_ADDR_LOCK();
487
488	vrf = sctp_find_vrf(vrf_id);
489	if (vrf == NULL) {
490#ifdef SCTP_DEBUG
491		printf("Can't find vrf_id:%d\n", vrf_id);
492#endif
493		goto out_now;
494	}
495	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, 1);
496	if (sctp_ifap) {
497		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
498		sctp_ifap->localifa_flags |= SCTP_BEING_DELETED;
499		vrf->total_ifa_count--;
500		LIST_REMOVE(sctp_ifap, next_bucket);
501		LIST_REMOVE(sctp_ifap, next_ifa);
502		if (sctp_ifap->ifn_p) {
503			sctp_ifap->ifn_p->ifa_count--;
504			if (SCTP_LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
505				sctp_delete_ifn(sctp_ifap->ifn_p, 1);
506			}
507			sctp_free_ifn(sctp_ifap->ifn_p);
508			sctp_ifap->ifn_p = NULL;
509		}
510	}
511#ifdef SCTP_DEBUG
512	else {
513		printf("Del Addr-ifn:%d Could not find address:",
514		    ifn_index);
515		sctp_print_address(addr);
516	}
517#endif
518out_now:
519	SCTP_IPI_ADDR_UNLOCK();
520	if (sctp_ifap) {
521		struct sctp_laddr *wi;
522
523		wi = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
524		if (wi == NULL) {
525			/*
526			 * Gak, what can we do? We have lost an address
527			 * change can you say HOSED?
528			 */
529#ifdef SCTP_DEBUG
530			if (sctp_debug_on & SCTP_DEBUG_PCB1) {
531				printf("Lost and address change ???\n");
532			}
533#endif				/* SCTP_DEBUG */
534
535			/* Opps, must decrement the count */
536			sctp_free_ifa(sctp_ifap);
537			return;
538		}
539		SCTP_INCR_LADDR_COUNT();
540		bzero(wi, sizeof(*wi));
541		wi->ifa = sctp_ifap;
542		wi->action = SCTP_DEL_IP_ADDRESS;
543		SCTP_IPI_ITERATOR_WQ_LOCK();
544		/*
545		 * Should this really be a tailq? As it is we will process
546		 * the newest first :-0
547		 */
548		LIST_INSERT_HEAD(&sctppcbinfo.addr_wq, wi, sctp_nxt_addr);
549		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
550		    (struct sctp_inpcb *)NULL,
551		    (struct sctp_tcb *)NULL,
552		    (struct sctp_nets *)NULL);
553		SCTP_IPI_ITERATOR_WQ_UNLOCK();
554	}
555	return;
556}
557
558
559
560static struct sctp_tcb *
561sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
562    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
563{
564	/**** ASSUMSES THE CALLER holds the INP_INFO_RLOCK */
565	/*
566	 * If we support the TCP model, then we must now dig through to see
567	 * if we can find our endpoint in the list of tcp ep's.
568	 */
569	uint16_t lport, rport;
570	struct sctppcbhead *ephead;
571	struct sctp_inpcb *inp;
572	struct sctp_laddr *laddr;
573	struct sctp_tcb *stcb;
574	struct sctp_nets *net;
575
576	if ((to == NULL) || (from == NULL)) {
577		return (NULL);
578	}
579	if (to->sa_family == AF_INET && from->sa_family == AF_INET) {
580		lport = ((struct sockaddr_in *)to)->sin_port;
581		rport = ((struct sockaddr_in *)from)->sin_port;
582	} else if (to->sa_family == AF_INET6 && from->sa_family == AF_INET6) {
583		lport = ((struct sockaddr_in6 *)to)->sin6_port;
584		rport = ((struct sockaddr_in6 *)from)->sin6_port;
585	} else {
586		return NULL;
587	}
588	ephead = &sctppcbinfo.sctp_tcpephash[SCTP_PCBHASH_ALLADDR(
589	    (lport + rport), sctppcbinfo.hashtcpmark)];
590	/*
591	 * Ok now for each of the guys in this bucket we must look and see:
592	 * - Does the remote port match. - Does there single association's
593	 * addresses match this address (to). If so we update p_ep to point
594	 * to this ep and return the tcb from it.
595	 */
596	LIST_FOREACH(inp, ephead, sctp_hash) {
597		SCTP_INP_RLOCK(inp);
598		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
599			SCTP_INP_RUNLOCK(inp);
600			continue;
601		}
602		if (lport != inp->sctp_lport) {
603			SCTP_INP_RUNLOCK(inp);
604			continue;
605		}
606		if (inp->def_vrf_id != vrf_id) {
607			SCTP_INP_RUNLOCK(inp);
608			continue;
609		}
610		/* check to see if the ep has one of the addresses */
611		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
612			/* We are NOT bound all, so look further */
613			int match = 0;
614
615			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
616
617				if (laddr->ifa == NULL) {
618#ifdef SCTP_DEBUG
619					if (sctp_debug_on & SCTP_DEBUG_PCB1) {
620						printf("An ounce of prevention is worth a pound of cure\n");
621					}
622#endif
623					continue;
624				}
625				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
626#ifdef SCTP_DEBUG
627					if (sctp_debug_on & SCTP_DEBUG_PCB1) {
628						printf("ifa being deleted\n");
629					}
630#endif
631					continue;
632				}
633				if (laddr->ifa->address.sa.sa_family ==
634				    to->sa_family) {
635					/* see if it matches */
636					struct sockaddr_in *intf_addr, *sin;
637
638					intf_addr = &laddr->ifa->address.sin;
639					sin = (struct sockaddr_in *)to;
640					if (from->sa_family == AF_INET) {
641						if (sin->sin_addr.s_addr ==
642						    intf_addr->sin_addr.s_addr) {
643							match = 1;
644							break;
645						}
646					} else {
647						struct sockaddr_in6 *intf_addr6;
648						struct sockaddr_in6 *sin6;
649
650						sin6 = (struct sockaddr_in6 *)
651						    to;
652						intf_addr6 = &laddr->ifa->address.sin6;
653
654						if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
655						    &intf_addr6->sin6_addr)) {
656							match = 1;
657							break;
658						}
659					}
660				}
661			}
662			if (match == 0) {
663				/* This endpoint does not have this address */
664				SCTP_INP_RUNLOCK(inp);
665				continue;
666			}
667		}
668		/*
669		 * Ok if we hit here the ep has the address, does it hold
670		 * the tcb?
671		 */
672
673		stcb = LIST_FIRST(&inp->sctp_asoc_list);
674		if (stcb == NULL) {
675			SCTP_INP_RUNLOCK(inp);
676			continue;
677		}
678		SCTP_TCB_LOCK(stcb);
679		if (stcb->rport != rport) {
680			/* remote port does not match. */
681			SCTP_TCB_UNLOCK(stcb);
682			SCTP_INP_RUNLOCK(inp);
683			continue;
684		}
685		/* Does this TCB have a matching address? */
686		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
687
688			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
689				/* not the same family, can't be a match */
690				continue;
691			}
692			if (from->sa_family == AF_INET) {
693				struct sockaddr_in *sin, *rsin;
694
695				sin = (struct sockaddr_in *)&net->ro._l_addr;
696				rsin = (struct sockaddr_in *)from;
697				if (sin->sin_addr.s_addr ==
698				    rsin->sin_addr.s_addr) {
699					/* found it */
700					if (netp != NULL) {
701						*netp = net;
702					}
703					/* Update the endpoint pointer */
704					*inp_p = inp;
705					SCTP_INP_RUNLOCK(inp);
706					return (stcb);
707				}
708			} else {
709				struct sockaddr_in6 *sin6, *rsin6;
710
711				sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
712				rsin6 = (struct sockaddr_in6 *)from;
713				if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
714				    &rsin6->sin6_addr)) {
715					/* found it */
716					if (netp != NULL) {
717						*netp = net;
718					}
719					/* Update the endpoint pointer */
720					*inp_p = inp;
721					SCTP_INP_RUNLOCK(inp);
722					return (stcb);
723				}
724			}
725		}
726		SCTP_TCB_UNLOCK(stcb);
727		SCTP_INP_RUNLOCK(inp);
728	}
729	return (NULL);
730}
731
732/*
733 * rules for use
734 *
735 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
736 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
737 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
738 * NULL.
739 */
740
741struct sctp_tcb *
742sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
743    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
744{
745	struct sctpasochead *head;
746	struct sctp_inpcb *inp;
747	struct sctp_tcb *stcb = NULL;
748	struct sctp_nets *net;
749	uint16_t rport;
750
751	inp = *inp_p;
752	if (remote->sa_family == AF_INET) {
753		rport = (((struct sockaddr_in *)remote)->sin_port);
754	} else if (remote->sa_family == AF_INET6) {
755		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
756	} else {
757		return (NULL);
758	}
759	if (locked_tcb) {
760		/*
761		 * UN-lock so we can do proper locking here this occurs when
762		 * called from load_addresses_from_init.
763		 */
764		SCTP_TCB_UNLOCK(locked_tcb);
765	}
766	SCTP_INP_INFO_RLOCK();
767	if (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) {
768		/*-
769		 * Now either this guy is our listener or it's the
770		 * connector. If it is the one that issued the connect, then
771		 * it's only chance is to be the first TCB in the list. If
772		 * it is the acceptor, then do the special_lookup to hash
773		 * and find the real inp.
774		 */
775		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
776			/* to is peer addr, from is my addr */
777			stcb = sctp_tcb_special_locate(inp_p, remote, local,
778			    netp, inp->def_vrf_id);
779			if ((stcb != NULL) && (locked_tcb == NULL)) {
780				/* we have a locked tcb, lower refcount */
781				SCTP_INP_WLOCK(inp);
782				SCTP_INP_DECR_REF(inp);
783				SCTP_INP_WUNLOCK(inp);
784			}
785			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
786				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
787				SCTP_TCB_LOCK(locked_tcb);
788				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
789			}
790			SCTP_INP_INFO_RUNLOCK();
791			return (stcb);
792		} else {
793			SCTP_INP_WLOCK(inp);
794			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
795				goto null_return;
796			}
797			stcb = LIST_FIRST(&inp->sctp_asoc_list);
798			if (stcb == NULL) {
799				goto null_return;
800			}
801			SCTP_TCB_LOCK(stcb);
802			if (stcb->rport != rport) {
803				/* remote port does not match. */
804				SCTP_TCB_UNLOCK(stcb);
805				goto null_return;
806			}
807			/* now look at the list of remote addresses */
808			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
809#ifdef INVARIANTS
810				if (net == (TAILQ_NEXT(net, sctp_next))) {
811					panic("Corrupt net list");
812				}
813#endif
814				if (net->ro._l_addr.sa.sa_family !=
815				    remote->sa_family) {
816					/* not the same family */
817					continue;
818				}
819				if (remote->sa_family == AF_INET) {
820					struct sockaddr_in *sin, *rsin;
821
822					sin = (struct sockaddr_in *)
823					    &net->ro._l_addr;
824					rsin = (struct sockaddr_in *)remote;
825					if (sin->sin_addr.s_addr ==
826					    rsin->sin_addr.s_addr) {
827						/* found it */
828						if (netp != NULL) {
829							*netp = net;
830						}
831						if (locked_tcb == NULL) {
832							SCTP_INP_DECR_REF(inp);
833						} else if (locked_tcb != stcb) {
834							SCTP_TCB_LOCK(locked_tcb);
835						}
836						SCTP_INP_WUNLOCK(inp);
837						SCTP_INP_INFO_RUNLOCK();
838						return (stcb);
839					}
840				} else if (remote->sa_family == AF_INET6) {
841					struct sockaddr_in6 *sin6, *rsin6;
842
843					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
844					rsin6 = (struct sockaddr_in6 *)remote;
845					if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
846					    &rsin6->sin6_addr)) {
847						/* found it */
848						if (netp != NULL) {
849							*netp = net;
850						}
851						if (locked_tcb == NULL) {
852							SCTP_INP_DECR_REF(inp);
853						} else if (locked_tcb != stcb) {
854							SCTP_TCB_LOCK(locked_tcb);
855						}
856						SCTP_INP_WUNLOCK(inp);
857						SCTP_INP_INFO_RUNLOCK();
858						return (stcb);
859					}
860				}
861			}
862			SCTP_TCB_UNLOCK(stcb);
863		}
864	} else {
865		SCTP_INP_WLOCK(inp);
866		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
867			goto null_return;
868		}
869		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
870		    inp->sctp_hashmark)];
871		if (head == NULL) {
872			goto null_return;
873		}
874		LIST_FOREACH(stcb, head, sctp_tcbhash) {
875			if (stcb->rport != rport) {
876				/* remote port does not match */
877				continue;
878			}
879			/* now look at the list of remote addresses */
880			SCTP_TCB_LOCK(stcb);
881			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
882#ifdef INVARIANTS
883				if (net == (TAILQ_NEXT(net, sctp_next))) {
884					panic("Corrupt net list");
885				}
886#endif
887				if (net->ro._l_addr.sa.sa_family !=
888				    remote->sa_family) {
889					/* not the same family */
890					continue;
891				}
892				if (remote->sa_family == AF_INET) {
893					struct sockaddr_in *sin, *rsin;
894
895					sin = (struct sockaddr_in *)
896					    &net->ro._l_addr;
897					rsin = (struct sockaddr_in *)remote;
898					if (sin->sin_addr.s_addr ==
899					    rsin->sin_addr.s_addr) {
900						/* found it */
901						if (netp != NULL) {
902							*netp = net;
903						}
904						if (locked_tcb == NULL) {
905							SCTP_INP_DECR_REF(inp);
906						} else if (locked_tcb != stcb) {
907							SCTP_TCB_LOCK(locked_tcb);
908						}
909						SCTP_INP_WUNLOCK(inp);
910						SCTP_INP_INFO_RUNLOCK();
911						return (stcb);
912					}
913				} else if (remote->sa_family == AF_INET6) {
914					struct sockaddr_in6 *sin6, *rsin6;
915
916					sin6 = (struct sockaddr_in6 *)
917					    &net->ro._l_addr;
918					rsin6 = (struct sockaddr_in6 *)remote;
919					if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
920					    &rsin6->sin6_addr)) {
921						/* found it */
922						if (netp != NULL) {
923							*netp = net;
924						}
925						if (locked_tcb == NULL) {
926							SCTP_INP_DECR_REF(inp);
927						} else if (locked_tcb != stcb) {
928							SCTP_TCB_LOCK(locked_tcb);
929						}
930						SCTP_INP_WUNLOCK(inp);
931						SCTP_INP_INFO_RUNLOCK();
932						return (stcb);
933					}
934				}
935			}
936			SCTP_TCB_UNLOCK(stcb);
937		}
938	}
939null_return:
940	/* clean up for returning null */
941	if (locked_tcb) {
942		SCTP_TCB_LOCK(locked_tcb);
943	}
944	SCTP_INP_WUNLOCK(inp);
945	SCTP_INP_INFO_RUNLOCK();
946	/* not found */
947	return (NULL);
948}
949
950/*
951 * Find an association for a specific endpoint using the association id given
952 * out in the COMM_UP notification
953 */
954
955struct sctp_tcb *
956sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
957{
958	/*
959	 * Use my the assoc_id to find a endpoint
960	 */
961	struct sctpasochead *head;
962	struct sctp_tcb *stcb;
963	uint32_t id;
964
965	if (asoc_id == 0 || inp == NULL) {
966		return (NULL);
967	}
968	SCTP_INP_INFO_RLOCK();
969	id = (uint32_t) asoc_id;
970	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(id,
971	    sctppcbinfo.hashasocmark)];
972	if (head == NULL) {
973		/* invalid id TSNH */
974		SCTP_INP_INFO_RUNLOCK();
975		return (NULL);
976	}
977	LIST_FOREACH(stcb, head, sctp_asocs) {
978		SCTP_INP_RLOCK(stcb->sctp_ep);
979		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
980			SCTP_INP_RUNLOCK(stcb->sctp_ep);
981			SCTP_INP_INFO_RUNLOCK();
982			return (NULL);
983		}
984		if (stcb->asoc.assoc_id == id) {
985			/* candidate */
986			if (inp != stcb->sctp_ep) {
987				/*
988				 * some other guy has the same id active (id
989				 * collision ??).
990				 */
991				SCTP_INP_RUNLOCK(stcb->sctp_ep);
992				continue;
993			}
994			if (want_lock) {
995				SCTP_TCB_LOCK(stcb);
996			}
997			SCTP_INP_RUNLOCK(stcb->sctp_ep);
998			SCTP_INP_INFO_RUNLOCK();
999			return (stcb);
1000		}
1001		SCTP_INP_RUNLOCK(stcb->sctp_ep);
1002	}
1003	/* Ok if we missed here, lets try the restart hash */
1004	head = &sctppcbinfo.sctp_restarthash[SCTP_PCBHASH_ASOC(id, sctppcbinfo.hashrestartmark)];
1005	if (head == NULL) {
1006		/* invalid id TSNH */
1007		SCTP_INP_INFO_RUNLOCK();
1008		return (NULL);
1009	}
1010	LIST_FOREACH(stcb, head, sctp_tcbrestarhash) {
1011		SCTP_INP_RLOCK(stcb->sctp_ep);
1012		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1013			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1014			SCTP_INP_INFO_RUNLOCK();
1015			return (NULL);
1016		}
1017		SCTP_TCB_LOCK(stcb);
1018		SCTP_INP_RUNLOCK(stcb->sctp_ep);
1019		if (stcb->asoc.assoc_id == id) {
1020			/* candidate */
1021			if (inp != stcb->sctp_ep) {
1022				/*
1023				 * some other guy has the same id active (id
1024				 * collision ??).
1025				 */
1026				SCTP_TCB_UNLOCK(stcb);
1027				continue;
1028			}
1029			SCTP_INP_INFO_RUNLOCK();
1030			return (stcb);
1031		}
1032		SCTP_TCB_UNLOCK(stcb);
1033	}
1034	SCTP_INP_INFO_RUNLOCK();
1035	return (NULL);
1036}
1037
1038
1039static struct sctp_inpcb *
1040sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1041    uint16_t lport, uint32_t vrf_id)
1042{
1043	struct sctp_inpcb *inp;
1044	struct sockaddr_in *sin;
1045	struct sockaddr_in6 *sin6;
1046	struct sctp_laddr *laddr;
1047	int fnd;
1048
1049	/*
1050	 * Endpoing probe expects that the INP_INFO is locked.
1051	 */
1052	if (nam->sa_family == AF_INET) {
1053		sin = (struct sockaddr_in *)nam;
1054		sin6 = NULL;
1055	} else if (nam->sa_family == AF_INET6) {
1056		sin6 = (struct sockaddr_in6 *)nam;
1057		sin = NULL;
1058	} else {
1059		/* unsupported family */
1060		return (NULL);
1061	}
1062	if (head == NULL)
1063		return (NULL);
1064	LIST_FOREACH(inp, head, sctp_hash) {
1065		SCTP_INP_RLOCK(inp);
1066		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1067			SCTP_INP_RUNLOCK(inp);
1068			continue;
1069		}
1070		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1071		    (inp->sctp_lport == lport)) {
1072			/* got it */
1073			if ((nam->sa_family == AF_INET) &&
1074			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1075			    SCTP_IPV6_V6ONLY(inp)) {
1076				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1077				SCTP_INP_RUNLOCK(inp);
1078				continue;
1079			}
1080			/* A V6 address and the endpoint is NOT bound V6 */
1081			if (nam->sa_family == AF_INET6 &&
1082			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1083				SCTP_INP_RUNLOCK(inp);
1084				continue;
1085			}
1086			/* does a VRF id match? */
1087			fnd = 0;
1088			if (inp->def_vrf_id == vrf_id)
1089				fnd = 1;
1090
1091			SCTP_INP_RUNLOCK(inp);
1092			if (!fnd)
1093				continue;
1094			return (inp);
1095		}
1096		SCTP_INP_RUNLOCK(inp);
1097	}
1098
1099	if ((nam->sa_family == AF_INET) &&
1100	    (sin->sin_addr.s_addr == INADDR_ANY)) {
1101		/* Can't hunt for one that has no address specified */
1102		return (NULL);
1103	} else if ((nam->sa_family == AF_INET6) &&
1104	    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
1105		/* Can't hunt for one that has no address specified */
1106		return (NULL);
1107	}
1108	/*
1109	 * ok, not bound to all so see if we can find a EP bound to this
1110	 * address.
1111	 */
1112	LIST_FOREACH(inp, head, sctp_hash) {
1113		SCTP_INP_RLOCK(inp);
1114		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1115			SCTP_INP_RUNLOCK(inp);
1116			continue;
1117		}
1118		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1119			SCTP_INP_RUNLOCK(inp);
1120			continue;
1121		}
1122		/*
1123		 * Ok this could be a likely candidate, look at all of its
1124		 * addresses
1125		 */
1126		if (inp->sctp_lport != lport) {
1127			SCTP_INP_RUNLOCK(inp);
1128			continue;
1129		}
1130		/* does a VRF id match? */
1131		fnd = 0;
1132		if (inp->def_vrf_id == vrf_id)
1133			fnd = 1;
1134
1135		if (!fnd) {
1136			SCTP_INP_RUNLOCK(inp);
1137			continue;
1138		}
1139		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1140			if (laddr->ifa == NULL) {
1141#ifdef SCTP_DEBUG
1142				if (sctp_debug_on & SCTP_DEBUG_PCB1) {
1143					printf("An ounce of prevention is worth a pound of cure\n");
1144				}
1145#endif
1146				continue;
1147			}
1148#ifdef SCTP_DEBUG
1149			if (sctp_debug_on & SCTP_DEBUG_PCB1) {
1150				printf("Ok laddr->ifa:%p is possible, ",
1151				    laddr->ifa);
1152			}
1153#endif
1154			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1155#ifdef SCTP_DEBUG
1156				if (sctp_debug_on & SCTP_DEBUG_PCB1) {
1157					printf("Huh IFA being deleted\n");
1158				}
1159#endif
1160				continue;
1161			}
1162			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1163				/* possible, see if it matches */
1164				struct sockaddr_in *intf_addr;
1165
1166				intf_addr = &laddr->ifa->address.sin;
1167				if (nam->sa_family == AF_INET) {
1168					if (sin->sin_addr.s_addr ==
1169					    intf_addr->sin_addr.s_addr) {
1170						SCTP_INP_RUNLOCK(inp);
1171						return (inp);
1172					}
1173				} else if (nam->sa_family == AF_INET6) {
1174					struct sockaddr_in6 *intf_addr6;
1175
1176					intf_addr6 = &laddr->ifa->address.sin6;
1177					if (SCTP6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
1178					    &intf_addr6->sin6_addr)) {
1179						SCTP_INP_RUNLOCK(inp);
1180						return (inp);
1181					}
1182				}
1183			}
1184		}
1185		SCTP_INP_RUNLOCK(inp);
1186	}
1187	return (NULL);
1188}
1189
1190
1191struct sctp_inpcb *
1192sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock, uint32_t vrf_id)
1193{
1194	/*
1195	 * First we check the hash table to see if someone has this port
1196	 * bound with just the port.
1197	 */
1198	struct sctp_inpcb *inp;
1199	struct sctppcbhead *head;
1200	struct sockaddr_in *sin;
1201	struct sockaddr_in6 *sin6;
1202	int lport;
1203
1204	if (nam->sa_family == AF_INET) {
1205		sin = (struct sockaddr_in *)nam;
1206		lport = ((struct sockaddr_in *)nam)->sin_port;
1207	} else if (nam->sa_family == AF_INET6) {
1208		sin6 = (struct sockaddr_in6 *)nam;
1209		lport = ((struct sockaddr_in6 *)nam)->sin6_port;
1210	} else {
1211		/* unsupported family */
1212		return (NULL);
1213	}
1214	/*
1215	 * I could cheat here and just cast to one of the types but we will
1216	 * do it right. It also provides the check against an Unsupported
1217	 * type too.
1218	 */
1219	/* Find the head of the ALLADDR chain */
1220	if (have_lock == 0) {
1221		SCTP_INP_INFO_RLOCK();
1222
1223	}
1224	head = &sctppcbinfo.sctp_ephash[SCTP_PCBHASH_ALLADDR(lport,
1225	    sctppcbinfo.hashmark)];
1226	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1227
1228	/*
1229	 * If the TCP model exists it could be that the main listening
1230	 * endpoint is gone but there exists a connected socket for this guy
1231	 * yet. If so we can return the first one that we find. This may NOT
1232	 * be the correct one but the sctp_findassociation_ep_addr has
1233	 * further code to look at all TCP models.
1234	 */
1235	if (inp == NULL && find_tcp_pool) {
1236		unsigned int i;
1237
1238		for (i = 0; i < sctppcbinfo.hashtblsize; i++) {
1239			/*
1240			 * This is real gross, but we do NOT have a remote
1241			 * port at this point depending on who is calling.
1242			 * We must therefore look for ANY one that matches
1243			 * our local port :/
1244			 */
1245			head = &sctppcbinfo.sctp_tcpephash[i];
1246			if (LIST_FIRST(head)) {
1247				inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1248				if (inp) {
1249					/* Found one */
1250					break;
1251				}
1252			}
1253		}
1254	}
1255	if (inp) {
1256		SCTP_INP_INCR_REF(inp);
1257	}
1258	if (have_lock == 0) {
1259		SCTP_INP_INFO_RUNLOCK();
1260	}
1261	return (inp);
1262}
1263
1264/*
1265 * Find an association for an endpoint with the pointer to whom you want to
1266 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1267 * need to change the *to to some other struct like a mbuf...
1268 */
1269struct sctp_tcb *
1270sctp_findassociation_addr_sa(struct sockaddr *to, struct sockaddr *from,
1271    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool, uint32_t vrf_id)
1272{
1273	struct sctp_inpcb *inp = NULL;
1274	struct sctp_tcb *retval;
1275
1276	SCTP_INP_INFO_RLOCK();
1277	if (find_tcp_pool) {
1278		if (inp_p != NULL) {
1279			retval = sctp_tcb_special_locate(inp_p, from, to, netp, vrf_id);
1280		} else {
1281			retval = sctp_tcb_special_locate(&inp, from, to, netp, vrf_id);
1282		}
1283		if (retval != NULL) {
1284			SCTP_INP_INFO_RUNLOCK();
1285			return (retval);
1286		}
1287	}
1288	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1289	if (inp_p != NULL) {
1290		*inp_p = inp;
1291	}
1292	SCTP_INP_INFO_RUNLOCK();
1293
1294	if (inp == NULL) {
1295		return (NULL);
1296	}
1297	/*
1298	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1299	 * we now place the source address or from in the to of the find
1300	 * endpoint call. Since in reality this chain is used from the
1301	 * inbound packet side.
1302	 */
1303	if (inp_p != NULL) {
1304		retval = sctp_findassociation_ep_addr(inp_p, from, netp, to, NULL);
1305	} else {
1306		retval = sctp_findassociation_ep_addr(&inp, from, netp, to, NULL);
1307	}
1308	return retval;
1309}
1310
1311
1312/*
1313 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1314 * find all addresses that the sender has specified in any address list. Each
1315 * address will be used to lookup the TCB and see if one exits.
1316 */
1317static struct sctp_tcb *
1318sctp_findassociation_special_addr(struct mbuf *m, int iphlen, int offset,
1319    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1320    struct sockaddr *dest)
1321{
1322	struct sockaddr_in sin4;
1323	struct sockaddr_in6 sin6;
1324	struct sctp_paramhdr *phdr, parm_buf;
1325	struct sctp_tcb *retval;
1326	uint32_t ptype, plen;
1327
1328	memset(&sin4, 0, sizeof(sin4));
1329	memset(&sin6, 0, sizeof(sin6));
1330	sin4.sin_len = sizeof(sin4);
1331	sin4.sin_family = AF_INET;
1332	sin4.sin_port = sh->src_port;
1333	sin6.sin6_len = sizeof(sin6);
1334	sin6.sin6_family = AF_INET6;
1335	sin6.sin6_port = sh->src_port;
1336
1337	retval = NULL;
1338	offset += sizeof(struct sctp_init_chunk);
1339
1340	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
1341	while (phdr != NULL) {
1342		/* now we must see if we want the parameter */
1343		ptype = ntohs(phdr->param_type);
1344		plen = ntohs(phdr->param_length);
1345		if (plen == 0) {
1346			break;
1347		}
1348		if (ptype == SCTP_IPV4_ADDRESS &&
1349		    plen == sizeof(struct sctp_ipv4addr_param)) {
1350			/* Get the rest of the address */
1351			struct sctp_ipv4addr_param ip4_parm, *p4;
1352
1353			phdr = sctp_get_next_param(m, offset,
1354			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
1355			if (phdr == NULL) {
1356				return (NULL);
1357			}
1358			p4 = (struct sctp_ipv4addr_param *)phdr;
1359			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
1360			/* look it up */
1361			retval = sctp_findassociation_ep_addr(inp_p,
1362			    (struct sockaddr *)&sin4, netp, dest, NULL);
1363			if (retval != NULL) {
1364				return (retval);
1365			}
1366		} else if (ptype == SCTP_IPV6_ADDRESS &&
1367		    plen == sizeof(struct sctp_ipv6addr_param)) {
1368			/* Get the rest of the address */
1369			struct sctp_ipv6addr_param ip6_parm, *p6;
1370
1371			phdr = sctp_get_next_param(m, offset,
1372			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
1373			if (phdr == NULL) {
1374				return (NULL);
1375			}
1376			p6 = (struct sctp_ipv6addr_param *)phdr;
1377			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
1378			/* look it up */
1379			retval = sctp_findassociation_ep_addr(inp_p,
1380			    (struct sockaddr *)&sin6, netp, dest, NULL);
1381			if (retval != NULL) {
1382				return (retval);
1383			}
1384		}
1385		offset += SCTP_SIZE32(plen);
1386		phdr = sctp_get_next_param(m, offset, &parm_buf,
1387		    sizeof(parm_buf));
1388	}
1389	return (NULL);
1390}
1391
1392
1393static struct sctp_tcb *
1394sctp_findassoc_by_vtag(struct sockaddr *from, uint32_t vtag,
1395    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
1396    uint16_t lport, int skip_src_check)
1397{
1398	/*
1399	 * Use my vtag to hash. If we find it we then verify the source addr
1400	 * is in the assoc. If all goes well we save a bit on rec of a
1401	 * packet.
1402	 */
1403	struct sctpasochead *head;
1404	struct sctp_nets *net;
1405	struct sctp_tcb *stcb;
1406
1407	*netp = NULL;
1408	*inp_p = NULL;
1409	SCTP_INP_INFO_RLOCK();
1410	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(vtag,
1411	    sctppcbinfo.hashasocmark)];
1412	if (head == NULL) {
1413		/* invalid vtag */
1414		SCTP_INP_INFO_RUNLOCK();
1415		return (NULL);
1416	}
1417	LIST_FOREACH(stcb, head, sctp_asocs) {
1418		SCTP_INP_RLOCK(stcb->sctp_ep);
1419		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1420			SCTP_INP_RUNLOCK(stcb->sctp_ep);
1421			continue;
1422		}
1423		SCTP_TCB_LOCK(stcb);
1424		SCTP_INP_RUNLOCK(stcb->sctp_ep);
1425		if (stcb->asoc.my_vtag == vtag) {
1426			/* candidate */
1427			if (stcb->rport != rport) {
1428				/*
1429				 * we could remove this if vtags are unique
1430				 * across the system.
1431				 */
1432				SCTP_TCB_UNLOCK(stcb);
1433				continue;
1434			}
1435			if (stcb->sctp_ep->sctp_lport != lport) {
1436				/*
1437				 * we could remove this if vtags are unique
1438				 * across the system.
1439				 */
1440				SCTP_TCB_UNLOCK(stcb);
1441				continue;
1442			}
1443			if (skip_src_check) {
1444				*netp = NULL;	/* unknown */
1445				*inp_p = stcb->sctp_ep;
1446				SCTP_INP_INFO_RUNLOCK();
1447				return (stcb);
1448			}
1449			net = sctp_findnet(stcb, from);
1450			if (net) {
1451				/* yep its him. */
1452				*netp = net;
1453				SCTP_STAT_INCR(sctps_vtagexpress);
1454				*inp_p = stcb->sctp_ep;
1455				SCTP_INP_INFO_RUNLOCK();
1456				return (stcb);
1457			} else {
1458				/*
1459				 * not him, this should only happen in rare
1460				 * cases so I peg it.
1461				 */
1462				SCTP_STAT_INCR(sctps_vtagbogus);
1463			}
1464		}
1465		SCTP_TCB_UNLOCK(stcb);
1466	}
1467	SCTP_INP_INFO_RUNLOCK();
1468	return (NULL);
1469}
1470
1471/*
1472 * Find an association with the pointer to the inbound IP packet. This can be
1473 * a IPv4 or IPv6 packet.
1474 */
1475struct sctp_tcb *
1476sctp_findassociation_addr(struct mbuf *m, int iphlen, int offset,
1477    struct sctphdr *sh, struct sctp_chunkhdr *ch,
1478    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
1479{
1480	int find_tcp_pool;
1481	struct ip *iph;
1482	struct sctp_tcb *retval;
1483	struct sockaddr_storage to_store, from_store;
1484	struct sockaddr *to = (struct sockaddr *)&to_store;
1485	struct sockaddr *from = (struct sockaddr *)&from_store;
1486	struct sctp_inpcb *inp;
1487
1488	iph = mtod(m, struct ip *);
1489	if (iph->ip_v == IPVERSION) {
1490		/* its IPv4 */
1491		struct sockaddr_in *from4;
1492
1493		from4 = (struct sockaddr_in *)&from_store;
1494		bzero(from4, sizeof(*from4));
1495		from4->sin_family = AF_INET;
1496		from4->sin_len = sizeof(struct sockaddr_in);
1497		from4->sin_addr.s_addr = iph->ip_src.s_addr;
1498		from4->sin_port = sh->src_port;
1499	} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
1500		/* its IPv6 */
1501		struct ip6_hdr *ip6;
1502		struct sockaddr_in6 *from6;
1503
1504		ip6 = mtod(m, struct ip6_hdr *);
1505		from6 = (struct sockaddr_in6 *)&from_store;
1506		bzero(from6, sizeof(*from6));
1507		from6->sin6_family = AF_INET6;
1508		from6->sin6_len = sizeof(struct sockaddr_in6);
1509		from6->sin6_addr = ip6->ip6_src;
1510		from6->sin6_port = sh->src_port;
1511		/* Get the scopes in properly to the sin6 addr's */
1512		/* we probably don't need these operations */
1513		(void)sa6_recoverscope(from6);
1514		sa6_embedscope(from6, ip6_use_defzone);
1515	} else {
1516		/* Currently not supported. */
1517		return (NULL);
1518	}
1519	if (sh->v_tag) {
1520		/* we only go down this path if vtag is non-zero */
1521		retval = sctp_findassoc_by_vtag(from, ntohl(sh->v_tag),
1522		    inp_p, netp, sh->src_port, sh->dest_port, 0);
1523		if (retval) {
1524			return (retval);
1525		}
1526	}
1527	if (iph->ip_v == IPVERSION) {
1528		/* its IPv4 */
1529		struct sockaddr_in *to4;
1530
1531		to4 = (struct sockaddr_in *)&to_store;
1532		bzero(to4, sizeof(*to4));
1533		to4->sin_family = AF_INET;
1534		to4->sin_len = sizeof(struct sockaddr_in);
1535		to4->sin_addr.s_addr = iph->ip_dst.s_addr;
1536		to4->sin_port = sh->dest_port;
1537	} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
1538		/* its IPv6 */
1539		struct ip6_hdr *ip6;
1540		struct sockaddr_in6 *to6;
1541
1542		ip6 = mtod(m, struct ip6_hdr *);
1543		to6 = (struct sockaddr_in6 *)&to_store;
1544		bzero(to6, sizeof(*to6));
1545		to6->sin6_family = AF_INET6;
1546		to6->sin6_len = sizeof(struct sockaddr_in6);
1547		to6->sin6_addr = ip6->ip6_dst;
1548		to6->sin6_port = sh->dest_port;
1549		/* Get the scopes in properly to the sin6 addr's */
1550		/* we probably don't need these operations */
1551		(void)sa6_recoverscope(to6);
1552		sa6_embedscope(to6, ip6_use_defzone);
1553	}
1554	find_tcp_pool = 0;
1555	if ((ch->chunk_type != SCTP_INITIATION) &&
1556	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
1557	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
1558	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
1559		/* Other chunk types go to the tcp pool. */
1560		find_tcp_pool = 1;
1561	}
1562	if (inp_p) {
1563		retval = sctp_findassociation_addr_sa(to, from, inp_p, netp,
1564		    find_tcp_pool, vrf_id);
1565		inp = *inp_p;
1566	} else {
1567		retval = sctp_findassociation_addr_sa(to, from, &inp, netp,
1568		    find_tcp_pool, vrf_id);
1569	}
1570#ifdef SCTP_DEBUG
1571	if (sctp_debug_on & SCTP_DEBUG_PCB1) {
1572		printf("retval:%p inp:%p\n", retval, inp);
1573	}
1574#endif
1575	if (retval == NULL && inp) {
1576		/* Found a EP but not this address */
1577		if ((ch->chunk_type == SCTP_INITIATION) ||
1578		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
1579			/*-
1580			 * special hook, we do NOT return linp or an
1581			 * association that is linked to an existing
1582			 * association that is under the TCP pool (i.e. no
1583			 * listener exists). The endpoint finding routine
1584			 * will always find a listner before examining the
1585			 * TCP pool.
1586			 */
1587			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
1588				if (inp_p) {
1589					*inp_p = NULL;
1590				}
1591				return (NULL);
1592			}
1593			retval = sctp_findassociation_special_addr(m, iphlen,
1594			    offset, sh, &inp, netp, to);
1595			if (inp_p != NULL) {
1596				*inp_p = inp;
1597			}
1598		}
1599	}
1600#ifdef SCTP_DEBUG
1601	if (sctp_debug_on & SCTP_DEBUG_PCB1) {
1602		printf("retval is %p\n", retval);
1603	}
1604#endif
1605	return (retval);
1606}
1607
1608/*
1609 * lookup an association by an ASCONF lookup address.
1610 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
1611 */
1612struct sctp_tcb *
1613sctp_findassociation_ep_asconf(struct mbuf *m, int iphlen, int offset,
1614    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp)
1615{
1616	struct sctp_tcb *stcb;
1617	struct sockaddr_in *sin;
1618	struct sockaddr_in6 *sin6;
1619	struct sockaddr_storage local_store, remote_store;
1620	struct ip *iph;
1621	struct sctp_paramhdr parm_buf, *phdr;
1622	int ptype;
1623	int zero_address = 0;
1624
1625
1626	memset(&local_store, 0, sizeof(local_store));
1627	memset(&remote_store, 0, sizeof(remote_store));
1628
1629	/* First get the destination address setup too. */
1630	iph = mtod(m, struct ip *);
1631	if (iph->ip_v == IPVERSION) {
1632		/* its IPv4 */
1633		sin = (struct sockaddr_in *)&local_store;
1634		sin->sin_family = AF_INET;
1635		sin->sin_len = sizeof(*sin);
1636		sin->sin_port = sh->dest_port;
1637		sin->sin_addr.s_addr = iph->ip_dst.s_addr;
1638	} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
1639		/* its IPv6 */
1640		struct ip6_hdr *ip6;
1641
1642		ip6 = mtod(m, struct ip6_hdr *);
1643		sin6 = (struct sockaddr_in6 *)&local_store;
1644		sin6->sin6_family = AF_INET6;
1645		sin6->sin6_len = sizeof(*sin6);
1646		sin6->sin6_port = sh->dest_port;
1647		sin6->sin6_addr = ip6->ip6_dst;
1648	} else {
1649		return NULL;
1650	}
1651
1652	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
1653	    &parm_buf, sizeof(struct sctp_paramhdr));
1654	if (phdr == NULL) {
1655#ifdef SCTP_DEBUG
1656		if (sctp_debug_on & SCTP_DEBUG_INPUT3) {
1657			printf("findassociation_ep_asconf: failed to get asconf lookup addr\n");
1658		}
1659#endif				/* SCTP_DEBUG */
1660		return NULL;
1661	}
1662	ptype = (int)((uint32_t) ntohs(phdr->param_type));
1663	/* get the correlation address */
1664	if (ptype == SCTP_IPV6_ADDRESS) {
1665		/* ipv6 address param */
1666		struct sctp_ipv6addr_param *p6, p6_buf;
1667
1668		if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
1669			return NULL;
1670		}
1671		p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
1672		    offset + sizeof(struct sctp_asconf_chunk),
1673		    &p6_buf.ph, sizeof(*p6));
1674		if (p6 == NULL) {
1675#ifdef SCTP_DEBUG
1676			if (sctp_debug_on & SCTP_DEBUG_INPUT3) {
1677				printf("findassociation_ep_asconf: failed to get asconf v6 lookup addr\n");
1678			}
1679#endif				/* SCTP_DEBUG */
1680			return (NULL);
1681		}
1682		sin6 = (struct sockaddr_in6 *)&remote_store;
1683		sin6->sin6_family = AF_INET6;
1684		sin6->sin6_len = sizeof(*sin6);
1685		sin6->sin6_port = sh->src_port;
1686		memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
1687		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
1688			zero_address = 1;
1689	} else if (ptype == SCTP_IPV4_ADDRESS) {
1690		/* ipv4 address param */
1691		struct sctp_ipv4addr_param *p4, p4_buf;
1692
1693		if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
1694			return NULL;
1695		}
1696		p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
1697		    offset + sizeof(struct sctp_asconf_chunk),
1698		    &p4_buf.ph, sizeof(*p4));
1699		if (p4 == NULL) {
1700#ifdef SCTP_DEBUG
1701			if (sctp_debug_on & SCTP_DEBUG_INPUT3) {
1702				printf("findassociation_ep_asconf: failed to get asconf v4 lookup addr\n");
1703			}
1704#endif				/* SCTP_DEBUG */
1705			return (NULL);
1706		}
1707		sin = (struct sockaddr_in *)&remote_store;
1708		sin->sin_family = AF_INET;
1709		sin->sin_len = sizeof(*sin);
1710		sin->sin_port = sh->src_port;
1711		memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
1712		if (sin->sin_addr.s_addr == INADDR_ANY)
1713			zero_address = 1;
1714	} else {
1715		/* invalid address param type */
1716		return NULL;
1717	}
1718
1719	if (zero_address) {
1720		stcb = sctp_findassoc_by_vtag(NULL, ntohl(sh->v_tag), inp_p,
1721		    netp, sh->src_port, sh->dest_port, 1);
1722		/*
1723		 * printf("findassociation_ep_asconf: zero lookup address
1724		 * finds stcb 0x%x\n", (uint32_t)stcb);
1725		 */
1726	} else {
1727		stcb = sctp_findassociation_ep_addr(inp_p,
1728		    (struct sockaddr *)&remote_store, netp,
1729		    (struct sockaddr *)&local_store, NULL);
1730	}
1731	return (stcb);
1732}
1733
1734
1735/*
1736 * allocate a sctp_inpcb and setup a temporary binding to a port/all
1737 * addresses. This way if we don't get a bind we by default pick a ephemeral
1738 * port with all addresses bound.
1739 */
1740int
1741sctp_inpcb_alloc(struct socket *so)
1742{
1743	/*
1744	 * we get called when a new endpoint starts up. We need to allocate
1745	 * the sctp_inpcb structure from the zone and init it. Mark it as
1746	 * unbound and find a port that we can use as an ephemeral with
1747	 * INADDR_ANY. If the user binds later no problem we can then add in
1748	 * the specific addresses. And setup the default parameters for the
1749	 * EP.
1750	 */
1751	int i, error;
1752	struct sctp_inpcb *inp;
1753	struct sctp_pcb *m;
1754	struct timeval time;
1755	sctp_sharedkey_t *null_key;
1756
1757	error = 0;
1758
1759	SCTP_INP_INFO_WLOCK();
1760	inp = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_ep, struct sctp_inpcb);
1761	if (inp == NULL) {
1762		printf("Out of SCTP-INPCB structures - no resources\n");
1763		SCTP_INP_INFO_WUNLOCK();
1764		return (ENOBUFS);
1765	}
1766	/* zap it */
1767	bzero(inp, sizeof(*inp));
1768
1769	/* bump generations */
1770	/* setup socket pointers */
1771	inp->sctp_socket = so;
1772	inp->ip_inp.inp.inp_socket = so;
1773
1774	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
1775	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
1776
1777#ifdef IPSEC
1778	{
1779		struct inpcbpolicy *pcb_sp = NULL;
1780
1781		error = ipsec_init_pcbpolicy(so, &pcb_sp);
1782		/* Arrange to share the policy */
1783		inp->ip_inp.inp.inp_sp = pcb_sp;
1784		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
1785	}
1786	if (error != 0) {
1787		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
1788		SCTP_INP_INFO_WUNLOCK();
1789		return error;
1790	}
1791#endif				/* IPSEC */
1792	SCTP_INCR_EP_COUNT();
1793	inp->ip_inp.inp.inp_ip_ttl = ip_defttl;
1794	SCTP_INP_INFO_WUNLOCK();
1795
1796	so->so_pcb = (caddr_t)inp;
1797
1798	if ((SCTP_SO_TYPE(so) == SOCK_DGRAM) ||
1799	    (SCTP_SO_TYPE(so) == SOCK_SEQPACKET)) {
1800		/* UDP style socket */
1801		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
1802		    SCTP_PCB_FLAGS_UNBOUND);
1803		/* Be sure it is NON-BLOCKING IO for UDP */
1804		/* SCTP_SET_SO_NBIO(so); */
1805	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
1806		/* TCP style socket */
1807		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
1808		    SCTP_PCB_FLAGS_UNBOUND);
1809		/* Be sure we have blocking IO by default */
1810		SCTP_CLEAR_SO_NBIO(so);
1811	} else {
1812		/*
1813		 * unsupported socket type (RAW, etc)- in case we missed it
1814		 * in protosw
1815		 */
1816		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
1817		return (EOPNOTSUPP);
1818	}
1819	inp->sctp_tcbhash = SCTP_HASH_INIT(sctp_pcbtblsize,
1820	    &inp->sctp_hashmark);
1821	if (inp->sctp_tcbhash == NULL) {
1822		printf("Out of SCTP-INPCB->hashinit - no resources\n");
1823		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
1824		return (ENOBUFS);
1825	}
1826	inp->def_vrf_id = SCTP_DEFAULT_VRFID;
1827	inp->def_table_id = SCTP_DEFAULT_TABLEID;
1828
1829	SCTP_INP_INFO_WLOCK();
1830	SCTP_INP_LOCK_INIT(inp);
1831	SCTP_INP_READ_INIT(inp);
1832	SCTP_ASOC_CREATE_LOCK_INIT(inp);
1833	/* lock the new ep */
1834	SCTP_INP_WLOCK(inp);
1835
1836	/* add it to the info area */
1837	LIST_INSERT_HEAD(&sctppcbinfo.listhead, inp, sctp_list);
1838	SCTP_INP_INFO_WUNLOCK();
1839
1840	TAILQ_INIT(&inp->read_queue);
1841	LIST_INIT(&inp->sctp_addr_list);
1842
1843	LIST_INIT(&inp->sctp_asoc_list);
1844
1845#ifdef SCTP_TRACK_FREED_ASOCS
1846	/* TEMP CODE */
1847	LIST_INIT(&inp->sctp_asoc_free_list);
1848#endif
1849	/* Init the timer structure for signature change */
1850	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
1851	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
1852
1853	/* now init the actual endpoint default data */
1854	m = &inp->sctp_ep;
1855
1856	/* setup the base timeout information */
1857	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
1858	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
1859	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(sctp_delayed_sack_time_default);
1860	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(sctp_heartbeat_interval_default);
1861	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(sctp_pmtu_raise_time_default);
1862	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(sctp_shutdown_guard_time_default);
1863	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(sctp_secret_lifetime_default);
1864	/* all max/min max are in ms */
1865	m->sctp_maxrto = sctp_rto_max_default;
1866	m->sctp_minrto = sctp_rto_min_default;
1867	m->initial_rto = sctp_rto_initial_default;
1868	m->initial_init_rto_max = sctp_init_rto_max_default;
1869	m->sctp_sack_freq = sctp_sack_freq_default;
1870
1871	m->max_open_streams_intome = MAX_SCTP_STREAMS;
1872
1873	m->max_init_times = sctp_init_rtx_max_default;
1874	m->max_send_times = sctp_assoc_rtx_max_default;
1875	m->def_net_failure = sctp_path_rtx_max_default;
1876	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
1877	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
1878	m->max_burst = sctp_max_burst_default;
1879	/* number of streams to pre-open on a association */
1880	m->pre_open_stream_count = sctp_nr_outgoing_streams_default;
1881
1882	/* Add adaptation cookie */
1883	m->adaptation_layer_indicator = 0x504C5253;
1884
1885	/* seed random number generator */
1886	m->random_counter = 1;
1887	m->store_at = SCTP_SIGNATURE_SIZE;
1888	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
1889	sctp_fill_random_store(m);
1890
1891	/* Minimum cookie size */
1892	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
1893	    sizeof(struct sctp_state_cookie);
1894	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
1895
1896	/* Setup the initial secret */
1897	(void)SCTP_GETTIME_TIMEVAL(&time);
1898	m->time_of_secret_change = time.tv_sec;
1899
1900	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
1901		m->secret_key[0][i] = sctp_select_initial_TSN(m);
1902	}
1903	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
1904
1905	/* How long is a cookie good for ? */
1906	m->def_cookie_life = sctp_valid_cookie_life_default;
1907	/*
1908	 * Initialize authentication parameters
1909	 */
1910	m->local_hmacs = sctp_default_supported_hmaclist();
1911	m->local_auth_chunks = sctp_alloc_chunklist();
1912	sctp_auth_set_default_chunks(m->local_auth_chunks);
1913	LIST_INIT(&m->shared_keys);
1914	/* add default NULL key as key id 0 */
1915	null_key = sctp_alloc_sharedkey();
1916	sctp_insert_sharedkey(&m->shared_keys, null_key);
1917	SCTP_INP_WUNLOCK(inp);
1918#ifdef SCTP_LOG_CLOSING
1919	sctp_log_closing(inp, NULL, 12);
1920#endif
1921	return (error);
1922}
1923
1924
1925void
1926sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
1927    struct sctp_tcb *stcb)
1928{
1929	struct sctp_nets *net;
1930	uint16_t lport, rport;
1931	struct sctppcbhead *head;
1932	struct sctp_laddr *laddr, *oladdr;
1933
1934	SCTP_TCB_UNLOCK(stcb);
1935	SCTP_INP_INFO_WLOCK();
1936	SCTP_INP_WLOCK(old_inp);
1937	SCTP_INP_WLOCK(new_inp);
1938	SCTP_TCB_LOCK(stcb);
1939
1940	new_inp->sctp_ep.time_of_secret_change =
1941	    old_inp->sctp_ep.time_of_secret_change;
1942	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
1943	    sizeof(old_inp->sctp_ep.secret_key));
1944	new_inp->sctp_ep.current_secret_number =
1945	    old_inp->sctp_ep.current_secret_number;
1946	new_inp->sctp_ep.last_secret_number =
1947	    old_inp->sctp_ep.last_secret_number;
1948	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
1949
1950	/* make it so new data pours into the new socket */
1951	stcb->sctp_socket = new_inp->sctp_socket;
1952	stcb->sctp_ep = new_inp;
1953
1954	/* Copy the port across */
1955	lport = new_inp->sctp_lport = old_inp->sctp_lport;
1956	rport = stcb->rport;
1957	/* Pull the tcb from the old association */
1958	LIST_REMOVE(stcb, sctp_tcbhash);
1959	LIST_REMOVE(stcb, sctp_tcblist);
1960
1961	/* Now insert the new_inp into the TCP connected hash */
1962	head = &sctppcbinfo.sctp_tcpephash[SCTP_PCBHASH_ALLADDR((lport + rport),
1963	    sctppcbinfo.hashtcpmark)];
1964
1965	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
1966	/* Its safe to access */
1967	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
1968
1969	/* Now move the tcb into the endpoint list */
1970	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
1971	/*
1972	 * Question, do we even need to worry about the ep-hash since we
1973	 * only have one connection? Probably not :> so lets get rid of it
1974	 * and not suck up any kernel memory in that.
1975	 */
1976
1977	/* Ok. Let's restart timer. */
1978	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1979		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
1980		    stcb, net);
1981	}
1982
1983	SCTP_INP_INFO_WUNLOCK();
1984	if (new_inp->sctp_tcbhash != NULL) {
1985		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
1986		new_inp->sctp_tcbhash = NULL;
1987	}
1988	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1989		/* Subset bound, so copy in the laddr list from the old_inp */
1990		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
1991			laddr = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
1992			if (laddr == NULL) {
1993				/*
1994				 * Gak, what can we do? This assoc is really
1995				 * HOSED. We probably should send an abort
1996				 * here.
1997				 */
1998#ifdef SCTP_DEBUG
1999				if (sctp_debug_on & SCTP_DEBUG_PCB1) {
2000					printf("Association hosed in TCP model, out of laddr memory\n");
2001				}
2002#endif				/* SCTP_DEBUG */
2003				continue;
2004			}
2005			SCTP_INCR_LADDR_COUNT();
2006			bzero(laddr, sizeof(*laddr));
2007			laddr->ifa = oladdr->ifa;
2008			atomic_add_int(&laddr->ifa->refcount, 1);
2009			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2010			    sctp_nxt_addr);
2011			new_inp->laddr_count++;
2012		}
2013	}
2014	/*
2015	 * Now any running timers need to be adjusted since we really don't
2016	 * care if they are running or not just blast in the new_inp into
2017	 * all of them.
2018	 */
2019
2020	stcb->asoc.hb_timer.ep = (void *)new_inp;
2021	stcb->asoc.dack_timer.ep = (void *)new_inp;
2022	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2023	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2024	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2025	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2026	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2027	/* now what about the nets? */
2028	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2029		net->pmtu_timer.ep = (void *)new_inp;
2030		net->rxt_timer.ep = (void *)new_inp;
2031		net->fr_timer.ep = (void *)new_inp;
2032	}
2033	SCTP_INP_WUNLOCK(new_inp);
2034	SCTP_INP_WUNLOCK(old_inp);
2035}
2036
2037static int
2038sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
2039{
2040	struct sctppcbhead *head;
2041	struct sctp_inpcb *t_inp;
2042	int fnd;
2043
2044	head = &sctppcbinfo.sctp_ephash[SCTP_PCBHASH_ALLADDR(lport,
2045	    sctppcbinfo.hashmark)];
2046
2047	LIST_FOREACH(t_inp, head, sctp_hash) {
2048		if (t_inp->sctp_lport != lport) {
2049			continue;
2050		}
2051		/* is it in the VRF in question */
2052		fnd = 0;
2053		if (t_inp->def_vrf_id == vrf_id)
2054			fnd = 1;
2055		if (!fnd)
2056			continue;
2057
2058		/* This one is in use. */
2059		/* check the v6/v4 binding issue */
2060		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
2061		    SCTP_IPV6_V6ONLY(t_inp)) {
2062			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
2063				/* collision in V6 space */
2064				return (1);
2065			} else {
2066				/* inp is BOUND_V4 no conflict */
2067				continue;
2068			}
2069		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
2070			/* t_inp is bound v4 and v6, conflict always */
2071			return (1);
2072		} else {
2073			/* t_inp is bound only V4 */
2074			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
2075			    SCTP_IPV6_V6ONLY(inp)) {
2076				/* no conflict */
2077				continue;
2078			}
2079			/* else fall through to conflict */
2080		}
2081		return (1);
2082	}
2083	return (0);
2084}
2085
2086
2087
2088int
2089sctp_inpcb_bind(struct socket *so, struct sockaddr *addr, struct thread *p)
2090{
2091	/* bind a ep to a socket address */
2092	struct sctppcbhead *head;
2093	struct sctp_inpcb *inp, *inp_tmp;
2094	struct inpcb *ip_inp;
2095	int bindall;
2096	uint16_t lport;
2097	int error;
2098	uint32_t vrf_id;
2099
2100	lport = 0;
2101	error = 0;
2102	bindall = 1;
2103	inp = (struct sctp_inpcb *)so->so_pcb;
2104	ip_inp = (struct inpcb *)so->so_pcb;
2105#ifdef SCTP_DEBUG
2106	if (sctp_debug_on & SCTP_DEBUG_PCB1) {
2107		if (addr) {
2108			printf("Bind called port:%d\n",
2109			    ntohs(((struct sockaddr_in *)addr)->sin_port));
2110			printf("Addr :");
2111			sctp_print_address(addr);
2112		}
2113	}
2114#endif				/* SCTP_DEBUG */
2115	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2116		/* already did a bind, subsequent binds NOT allowed ! */
2117		return (EINVAL);
2118	}
2119	if (addr != NULL) {
2120		if (addr->sa_family == AF_INET) {
2121			struct sockaddr_in *sin;
2122
2123			/* IPV6_V6ONLY socket? */
2124			if (SCTP_IPV6_V6ONLY(ip_inp)) {
2125				return (EINVAL);
2126			}
2127			if (addr->sa_len != sizeof(*sin))
2128				return (EINVAL);
2129
2130			sin = (struct sockaddr_in *)addr;
2131			lport = sin->sin_port;
2132
2133			if (sin->sin_addr.s_addr != INADDR_ANY) {
2134				bindall = 0;
2135			}
2136		} else if (addr->sa_family == AF_INET6) {
2137			/* Only for pure IPv6 Address. (No IPv4 Mapped!) */
2138			struct sockaddr_in6 *sin6;
2139
2140			sin6 = (struct sockaddr_in6 *)addr;
2141
2142			if (addr->sa_len != sizeof(*sin6))
2143				return (EINVAL);
2144
2145			lport = sin6->sin6_port;
2146			if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2147				bindall = 0;
2148				/* KAME hack: embed scopeid */
2149				if (sa6_embedscope(sin6, ip6_use_defzone) != 0)
2150					return (EINVAL);
2151			}
2152			/* this must be cleared for ifa_ifwithaddr() */
2153			sin6->sin6_scope_id = 0;
2154		} else {
2155			return (EAFNOSUPPORT);
2156		}
2157	}
2158	/*
2159	 * Setup a vrf_id to be the default for the non-bind-all case.
2160	 */
2161	vrf_id = inp->def_vrf_id;
2162
2163	SCTP_INP_INFO_WLOCK();
2164	SCTP_INP_WLOCK(inp);
2165	/* increase our count due to the unlock we do */
2166	SCTP_INP_INCR_REF(inp);
2167	if (lport) {
2168		/*
2169		 * Did the caller specify a port? if so we must see if a ep
2170		 * already has this one bound.
2171		 */
2172		/* got to be root to get at low ports */
2173		if (ntohs(lport) < IPPORT_RESERVED) {
2174			if (p && (error =
2175			    priv_check_cred(p->td_ucred,
2176			    PRIV_NETINET_RESERVEDPORT,
2177			    SUSER_ALLOWJAIL
2178			    )
2179			    )) {
2180				SCTP_INP_DECR_REF(inp);
2181				SCTP_INP_WUNLOCK(inp);
2182				SCTP_INP_INFO_WUNLOCK();
2183				return (error);
2184			}
2185		}
2186		if (p == NULL) {
2187			SCTP_INP_DECR_REF(inp);
2188			SCTP_INP_WUNLOCK(inp);
2189			SCTP_INP_INFO_WUNLOCK();
2190			return (error);
2191		}
2192		SCTP_INP_WUNLOCK(inp);
2193		if (bindall) {
2194			vrf_id = inp->def_vrf_id;
2195			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2196			if (inp_tmp != NULL) {
2197				/*
2198				 * lock guy returned and lower count note
2199				 * that we are not bound so inp_tmp should
2200				 * NEVER be inp. And it is this inp
2201				 * (inp_tmp) that gets the reference bump,
2202				 * so we must lower it.
2203				 */
2204				SCTP_INP_DECR_REF(inp_tmp);
2205				SCTP_INP_DECR_REF(inp);
2206				/* unlock info */
2207				SCTP_INP_INFO_WUNLOCK();
2208				return (EADDRNOTAVAIL);
2209			}
2210		} else {
2211			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2212			if (inp_tmp != NULL) {
2213				/*
2214				 * lock guy returned and lower count note
2215				 * that we are not bound so inp_tmp should
2216				 * NEVER be inp. And it is this inp
2217				 * (inp_tmp) that gets the reference bump,
2218				 * so we must lower it.
2219				 */
2220				SCTP_INP_DECR_REF(inp_tmp);
2221				SCTP_INP_DECR_REF(inp);
2222				/* unlock info */
2223				SCTP_INP_INFO_WUNLOCK();
2224				return (EADDRNOTAVAIL);
2225			}
2226		}
2227		SCTP_INP_WLOCK(inp);
2228		if (bindall) {
2229			/* verify that no lport is not used by a singleton */
2230			if (sctp_isport_inuse(inp, lport, vrf_id)) {
2231				/* Sorry someone already has this one bound */
2232				SCTP_INP_DECR_REF(inp);
2233				SCTP_INP_WUNLOCK(inp);
2234				SCTP_INP_INFO_WUNLOCK();
2235				return (EADDRNOTAVAIL);
2236			}
2237		}
2238	} else {
2239		/*
2240		 * get any port but lets make sure no one has any address
2241		 * with this port bound
2242		 */
2243
2244		/*
2245		 * setup the inp to the top (I could use the union but this
2246		 * is just as easy
2247		 */
2248		uint32_t port_guess;
2249		uint16_t port_attempt;
2250		int not_done = 1;
2251		int not_found = 1;
2252
2253		while (not_done) {
2254			port_guess = sctp_select_initial_TSN(&inp->sctp_ep);
2255			port_attempt = (port_guess & 0x0000ffff);
2256			if (port_attempt == 0) {
2257				goto next_half;
2258			}
2259			if (port_attempt < IPPORT_RESERVED) {
2260				port_attempt += IPPORT_RESERVED;
2261			}
2262			vrf_id = inp->def_vrf_id;
2263			if (sctp_isport_inuse(inp, htons(port_attempt), vrf_id) == 1) {
2264				/* got a port we can use */
2265				not_found = 0;
2266				break;
2267			}
2268			if (not_found == 1) {
2269				/* We can use this port */
2270				not_done = 0;
2271				continue;
2272			}
2273			/* try upper half */
2274	next_half:
2275			port_attempt = ((port_guess >> 16) & 0x0000ffff);
2276			if (port_attempt == 0) {
2277				goto last_try;
2278			}
2279			if (port_attempt < IPPORT_RESERVED) {
2280				port_attempt += IPPORT_RESERVED;
2281			}
2282			vrf_id = inp->def_vrf_id;
2283			if (sctp_isport_inuse(inp, htons(port_attempt), vrf_id) == 1) {
2284				/* got a port we can use */
2285				not_found = 0;
2286				break;
2287			}
2288			if (not_found == 1) {
2289				/* We can use this port */
2290				not_done = 0;
2291				continue;
2292			}
2293			/* try two half's added together */
2294	last_try:
2295			port_attempt = (((port_guess >> 16) & 0x0000ffff) +
2296			    (port_guess & 0x0000ffff));
2297			if (port_attempt == 0) {
2298				/* get a new random number */
2299				continue;
2300			}
2301			if (port_attempt < IPPORT_RESERVED) {
2302				port_attempt += IPPORT_RESERVED;
2303			}
2304			vrf_id = inp->def_vrf_id;
2305			if (sctp_isport_inuse(inp, htons(port_attempt), vrf_id) == 1) {
2306				/* got a port we can use */
2307				not_found = 0;
2308				break;
2309			}
2310			if (not_found == 1) {
2311				/* We can use this port */
2312				not_done = 0;
2313				continue;
2314			}
2315		}
2316		/* we don't get out of the loop until we have a port */
2317		lport = htons(port_attempt);
2318	}
2319	SCTP_INP_DECR_REF(inp);
2320	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2321	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2322		/*
2323		 * this really should not happen. The guy did a non-blocking
2324		 * bind and then did a close at the same time.
2325		 */
2326		SCTP_INP_WUNLOCK(inp);
2327		SCTP_INP_INFO_WUNLOCK();
2328		return (EINVAL);
2329	}
2330	/* ok we look clear to give out this port, so lets setup the binding */
2331	if (bindall) {
2332		/* binding to all addresses, so just set in the proper flags */
2333		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
2334		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2335		/* set the automatic addr changes from kernel flag */
2336		if (sctp_auto_asconf == 0) {
2337			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2338		} else {
2339			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2340		}
2341	} else {
2342		/*
2343		 * bind specific, make sure flags is off and add a new
2344		 * address structure to the sctp_addr_list inside the ep
2345		 * structure.
2346		 *
2347		 * We will need to allocate one and insert it at the head. The
2348		 * socketopt call can just insert new addresses in there as
2349		 * well. It will also have to do the embed scope kame hack
2350		 * too (before adding).
2351		 */
2352		struct sctp_ifa *ifa;
2353		struct sockaddr_storage store_sa;
2354
2355		memset(&store_sa, 0, sizeof(store_sa));
2356		if (addr->sa_family == AF_INET) {
2357			struct sockaddr_in *sin;
2358
2359			sin = (struct sockaddr_in *)&store_sa;
2360			memcpy(sin, addr, sizeof(struct sockaddr_in));
2361			sin->sin_port = 0;
2362		} else if (addr->sa_family == AF_INET6) {
2363			struct sockaddr_in6 *sin6;
2364
2365			sin6 = (struct sockaddr_in6 *)&store_sa;
2366			memcpy(sin6, addr, sizeof(struct sockaddr_in6));
2367			sin6->sin6_port = 0;
2368		}
2369		/*
2370		 * first find the interface with the bound address need to
2371		 * zero out the port to find the address! yuck! can't do
2372		 * this earlier since need port for sctp_pcb_findep()
2373		 */
2374		ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa, vrf_id, 0);
2375		if (ifa == NULL) {
2376			/* Can't find an interface with that address */
2377			SCTP_INP_WUNLOCK(inp);
2378			SCTP_INP_INFO_WUNLOCK();
2379			return (EADDRNOTAVAIL);
2380		}
2381		if (addr->sa_family == AF_INET6) {
2382			/* GAK, more FIXME IFA lock? */
2383			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
2384				/* Can't bind a non-existent addr. */
2385				SCTP_INP_WUNLOCK(inp);
2386				SCTP_INP_INFO_WUNLOCK();
2387				return (EINVAL);
2388			}
2389		}
2390		/* we're not bound all */
2391		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
2392		/* set the automatic addr changes from kernel flag */
2393		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
2394		if (sctp_auto_asconf == 0) {
2395			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2396		} else {
2397			/*
2398			 * allow bindx() to send ASCONF's for binding
2399			 * changes
2400			 */
2401			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
2402		}
2403
2404		/* add this address to the endpoint list */
2405		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
2406		if (error != 0) {
2407			SCTP_INP_WUNLOCK(inp);
2408			SCTP_INP_INFO_WUNLOCK();
2409			return (error);
2410		}
2411		inp->laddr_count++;
2412	}
2413	/* find the bucket */
2414	head = &sctppcbinfo.sctp_ephash[SCTP_PCBHASH_ALLADDR(lport,
2415	    sctppcbinfo.hashmark)];
2416	/* put it in the bucket */
2417	LIST_INSERT_HEAD(head, inp, sctp_hash);
2418#ifdef SCTP_DEBUG
2419	if (sctp_debug_on & SCTP_DEBUG_PCB1) {
2420		printf("Main hash to bind at head:%p, bound port:%d\n", head, ntohs(lport));
2421	}
2422#endif
2423	/* set in the port */
2424	inp->sctp_lport = lport;
2425
2426	/* turn off just the unbound flag */
2427	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2428	SCTP_INP_WUNLOCK(inp);
2429	SCTP_INP_INFO_WUNLOCK();
2430	return (0);
2431}
2432
2433
2434static void
2435sctp_iterator_inp_being_freed(struct sctp_inpcb *inp, struct sctp_inpcb *inp_next)
2436{
2437	struct sctp_iterator *it;
2438
2439	/*
2440	 * We enter with the only the ITERATOR_LOCK in place and a write
2441	 * lock on the inp_info stuff.
2442	 */
2443
2444	/*
2445	 * Go through all iterators, we must do this since it is possible
2446	 * that some iterator does NOT have the lock, but is waiting for it.
2447	 * And the one that had the lock has either moved in the last
2448	 * iteration or we just cleared it above. We need to find all of
2449	 * those guys. The list of iterators should never be very big
2450	 * though.
2451	 */
2452	TAILQ_FOREACH(it, &sctppcbinfo.iteratorhead, sctp_nxt_itr) {
2453		if (it == inp->inp_starting_point_for_iterator)
2454			/* skip this guy, he's special */
2455			continue;
2456		if (it->inp == inp) {
2457			/*
2458			 * This is tricky and we DON'T lock the iterator.
2459			 * Reason is he's running but waiting for me since
2460			 * inp->inp_starting_point_for_iterator has the lock
2461			 * on me (the guy above we skipped). This tells us
2462			 * its is not running but waiting for
2463			 * inp->inp_starting_point_for_iterator to be
2464			 * released by the guy that does have our INP in a
2465			 * lock.
2466			 */
2467			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
2468				it->inp = NULL;
2469				it->stcb = NULL;
2470			} else {
2471				/* set him up to do the next guy not me */
2472				it->inp = inp_next;
2473				it->stcb = NULL;
2474			}
2475		}
2476	}
2477	it = inp->inp_starting_point_for_iterator;
2478	if (it) {
2479		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
2480			it->inp = NULL;
2481		} else {
2482			it->inp = inp_next;
2483		}
2484		it->stcb = NULL;
2485	}
2486}
2487
2488/* release sctp_inpcb unbind the port */
2489void
2490sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
2491{
2492	/*
2493	 * Here we free a endpoint. We must find it (if it is in the Hash
2494	 * table) and remove it from there. Then we must also find it in the
2495	 * overall list and remove it from there. After all removals are
2496	 * complete then any timer has to be stopped. Then start the actual
2497	 * freeing. a) Any local lists. b) Any associations. c) The hash of
2498	 * all associations. d) finally the ep itself.
2499	 */
2500	struct sctp_pcb *m;
2501	struct sctp_inpcb *inp_save;
2502	struct sctp_tcb *asoc, *nasoc;
2503	struct sctp_laddr *laddr, *nladdr;
2504	struct inpcb *ip_pcb;
2505	struct socket *so;
2506
2507	struct sctp_queued_to_read *sq;
2508
2509
2510	int cnt;
2511	sctp_sharedkey_t *shared_key;
2512
2513
2514#ifdef SCTP_LOG_CLOSING
2515	sctp_log_closing(inp, NULL, 0);
2516#endif
2517
2518	SCTP_ITERATOR_LOCK();
2519	so = inp->sctp_socket;
2520	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2521		/* been here before.. eeks.. get out of here */
2522		printf("This conflict in free SHOULD not be happening!\n");
2523		SCTP_ITERATOR_UNLOCK();
2524#ifdef SCTP_LOG_CLOSING
2525		sctp_log_closing(inp, NULL, 1);
2526#endif
2527		return;
2528	}
2529	SCTP_ASOC_CREATE_LOCK(inp);
2530	SCTP_INP_INFO_WLOCK();
2531
2532	SCTP_INP_WLOCK(inp);
2533	/*
2534	 * First time through we have the socket lock, after that no more.
2535	 */
2536	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
2537		/*
2538		 * Once we are in we can remove the flag from = 1 is only
2539		 * passed from the actual closing routines that are called
2540		 * via the sockets layer.
2541		 */
2542		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
2543	}
2544	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
2545	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
2546
2547	if (inp->control) {
2548		sctp_m_freem(inp->control);
2549		inp->control = NULL;
2550	}
2551	if (inp->pkt) {
2552		sctp_m_freem(inp->pkt);
2553		inp->pkt = NULL;
2554	}
2555	m = &inp->sctp_ep;
2556	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
2557					 * here but I will be nice :> (i.e.
2558					 * ip_pcb = ep;) */
2559	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
2560		int cnt_in_sd;
2561
2562		cnt_in_sd = 0;
2563		for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
2564		    asoc = nasoc) {
2565			nasoc = LIST_NEXT(asoc, sctp_tcblist);
2566			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2567				/* Skip guys being freed */
2568				asoc->sctp_socket = NULL;
2569				cnt_in_sd++;
2570				continue;
2571			}
2572			if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
2573			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) {
2574				/* Just abandon things in the front states */
2575				if (asoc->asoc.total_output_queue_size == 0) {
2576					sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_2);
2577					continue;
2578				}
2579			}
2580			SCTP_TCB_LOCK(asoc);
2581			/* Disconnect the socket please */
2582			asoc->sctp_socket = NULL;
2583			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
2584			if ((asoc->asoc.size_on_reasm_queue > 0) ||
2585			    (asoc->asoc.control_pdapi) ||
2586			    (asoc->asoc.size_on_all_streams > 0) ||
2587			    (so && (so->so_rcv.sb_cc > 0))
2588			    ) {
2589				/* Left with Data unread */
2590				struct mbuf *op_err;
2591
2592				op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
2593				    0, M_DONTWAIT, 1, MT_DATA);
2594				if (op_err) {
2595					/* Fill in the user initiated abort */
2596					struct sctp_paramhdr *ph;
2597					uint32_t *ippp;
2598
2599					SCTP_BUF_LEN(op_err) =
2600					    sizeof(struct sctp_paramhdr) + sizeof(uint32_t);
2601					ph = mtod(op_err,
2602					    struct sctp_paramhdr *);
2603					ph->param_type = htons(
2604					    SCTP_CAUSE_USER_INITIATED_ABT);
2605					ph->param_length = htons(SCTP_BUF_LEN(op_err));
2606					ippp = (uint32_t *) (ph + 1);
2607					*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_3);
2608				}
2609				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
2610				sctp_send_abort_tcb(asoc, op_err);
2611				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
2612				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2613				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2614					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2615				}
2616				sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4);
2617				continue;
2618			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
2619				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
2620				    (asoc->asoc.stream_queue_cnt == 0)
2621			    ) {
2622				if (asoc->asoc.locked_on_sending) {
2623					goto abort_anyway;
2624				}
2625				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
2626				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
2627					/*
2628					 * there is nothing queued to send,
2629					 * so I send shutdown
2630					 */
2631					sctp_send_shutdown(asoc, asoc->asoc.primary_destination);
2632					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2633					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2634						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2635					}
2636					asoc->asoc.state = SCTP_STATE_SHUTDOWN_SENT;
2637					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
2638					    asoc->asoc.primary_destination);
2639					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
2640					    asoc->asoc.primary_destination);
2641					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR);
2642				}
2643			} else {
2644				/* mark into shutdown pending */
2645				struct sctp_stream_queue_pending *sp;
2646
2647				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
2648				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
2649				    asoc->asoc.primary_destination);
2650				if (asoc->asoc.locked_on_sending) {
2651					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
2652					    sctp_streamhead);
2653					if (sp == NULL) {
2654						printf("Error, sp is NULL, locked on sending is %p strm:%d\n",
2655						    asoc->asoc.locked_on_sending,
2656						    asoc->asoc.locked_on_sending->stream_no);
2657					} else {
2658						if ((sp->length == 0) && (sp->msg_is_complete == 0))
2659							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
2660					}
2661				}
2662				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
2663				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
2664				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
2665					struct mbuf *op_err;
2666
2667			abort_anyway:
2668					op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
2669					    0, M_DONTWAIT, 1, MT_DATA);
2670					if (op_err) {
2671						/*
2672						 * Fill in the user
2673						 * initiated abort
2674						 */
2675						struct sctp_paramhdr *ph;
2676						uint32_t *ippp;
2677
2678						SCTP_BUF_LEN(op_err) =
2679						    (sizeof(struct sctp_paramhdr) +
2680						    sizeof(uint32_t));
2681						ph = mtod(op_err,
2682						    struct sctp_paramhdr *);
2683						ph->param_type = htons(
2684						    SCTP_CAUSE_USER_INITIATED_ABT);
2685						ph->param_length = htons(SCTP_BUF_LEN(op_err));
2686						ippp = (uint32_t *) (ph + 1);
2687						*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_5);
2688					}
2689					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
2690					sctp_send_abort_tcb(asoc, op_err);
2691					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
2692					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2693					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2694						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2695					}
2696					sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_6);
2697					continue;
2698				}
2699			}
2700			cnt_in_sd++;
2701			SCTP_TCB_UNLOCK(asoc);
2702		}
2703		/* now is there some left in our SHUTDOWN state? */
2704		if (cnt_in_sd) {
2705			SCTP_INP_WUNLOCK(inp);
2706			SCTP_ASOC_CREATE_UNLOCK(inp);
2707			SCTP_INP_INFO_WUNLOCK();
2708			SCTP_ITERATOR_UNLOCK();
2709#ifdef SCTP_LOG_CLOSING
2710			sctp_log_closing(inp, NULL, 2);
2711#endif
2712			return;
2713		}
2714	}
2715	inp->sctp_socket = NULL;
2716	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
2717	    SCTP_PCB_FLAGS_UNBOUND) {
2718		/*
2719		 * ok, this guy has been bound. It's port is somewhere in
2720		 * the sctppcbinfo hash table. Remove it!
2721		 */
2722		LIST_REMOVE(inp, sctp_hash);
2723		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
2724	}
2725	/*
2726	 * If there is a timer running to kill us, forget it, since it may
2727	 * have a contest on the INP lock.. which would cause us to die ...
2728	 */
2729	cnt = 0;
2730	for ((asoc = LIST_FIRST(&inp->sctp_asoc_list)); asoc != NULL;
2731	    asoc = nasoc) {
2732		nasoc = LIST_NEXT(asoc, sctp_tcblist);
2733		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2734			cnt++;
2735			continue;
2736		}
2737		/* Free associations that are NOT killing us */
2738		SCTP_TCB_LOCK(asoc);
2739		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
2740		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
2741			struct mbuf *op_err;
2742			uint32_t *ippp;
2743
2744			op_err = sctp_get_mbuf_for_msg((sizeof(struct sctp_paramhdr) + sizeof(uint32_t)),
2745			    0, M_DONTWAIT, 1, MT_DATA);
2746			if (op_err) {
2747				/* Fill in the user initiated abort */
2748				struct sctp_paramhdr *ph;
2749
2750				SCTP_BUF_LEN(op_err) = (sizeof(struct sctp_paramhdr) +
2751				    sizeof(uint32_t));
2752				ph = mtod(op_err, struct sctp_paramhdr *);
2753				ph->param_type = htons(
2754				    SCTP_CAUSE_USER_INITIATED_ABT);
2755				ph->param_length = htons(SCTP_BUF_LEN(op_err));
2756				ippp = (uint32_t *) (ph + 1);
2757				*ippp = htonl(SCTP_FROM_SCTP_PCB + SCTP_LOC_7);
2758
2759			}
2760			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
2761			sctp_send_abort_tcb(asoc, op_err);
2762			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
2763		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2764			cnt++;
2765			SCTP_TCB_UNLOCK(asoc);
2766			continue;
2767		}
2768		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
2769		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
2770			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
2771		}
2772		sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8);
2773	}
2774	if (cnt) {
2775		/* Ok we have someone out there that will kill us */
2776		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2777		SCTP_INP_WUNLOCK(inp);
2778		SCTP_ASOC_CREATE_UNLOCK(inp);
2779		SCTP_INP_INFO_WUNLOCK();
2780		SCTP_ITERATOR_UNLOCK();
2781#ifdef SCTP_LOG_CLOSING
2782		sctp_log_closing(inp, NULL, 3);
2783#endif
2784		return;
2785	}
2786	if ((inp->refcount) || (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
2787		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2788		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
2789		SCTP_INP_WUNLOCK(inp);
2790		SCTP_ASOC_CREATE_UNLOCK(inp);
2791		SCTP_INP_INFO_WUNLOCK();
2792		SCTP_ITERATOR_UNLOCK();
2793#ifdef SCTP_LOG_CLOSING
2794		sctp_log_closing(inp, NULL, 4);
2795#endif
2796		return;
2797	}
2798	(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2799	inp->sctp_ep.signature_change.type = 0;
2800	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
2801
2802#ifdef SCTP_LOG_CLOSING
2803	sctp_log_closing(inp, NULL, 5);
2804#endif
2805
2806	(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
2807	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NONE;
2808	/* Clear the read queue */
2809	while ((sq = TAILQ_FIRST(&inp->read_queue)) != NULL) {
2810		/* Its only abandoned if it had data left */
2811		if (sq->length)
2812			SCTP_STAT_INCR(sctps_left_abandon);
2813
2814		TAILQ_REMOVE(&inp->read_queue, sq, next);
2815		sctp_free_remote_addr(sq->whoFrom);
2816		if (so)
2817			so->so_rcv.sb_cc -= sq->length;
2818		if (sq->data) {
2819			sctp_m_freem(sq->data);
2820			sq->data = NULL;
2821		}
2822		/*
2823		 * no need to free the net count, since at this point all
2824		 * assoc's are gone.
2825		 */
2826		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, sq);
2827		SCTP_DECR_READQ_COUNT();
2828	}
2829	/* Now the sctp_pcb things */
2830	/*
2831	 * free each asoc if it is not already closed/free. we can't use the
2832	 * macro here since le_next will get freed as part of the
2833	 * sctp_free_assoc() call.
2834	 */
2835	cnt = 0;
2836	if (so) {
2837#ifdef IPSEC
2838		ipsec4_delete_pcbpolicy(ip_pcb);
2839#endif				/* IPSEC */
2840
2841		/* Unlocks not needed since the socket is gone now */
2842	}
2843	if (ip_pcb->inp_options) {
2844		(void)sctp_m_free(ip_pcb->inp_options);
2845		ip_pcb->inp_options = 0;
2846	}
2847	if (ip_pcb->inp_moptions) {
2848		ip_freemoptions(ip_pcb->inp_moptions);
2849		ip_pcb->inp_moptions = 0;
2850	}
2851#ifdef INET6
2852	if (ip_pcb->inp_vflag & INP_IPV6) {
2853		struct in6pcb *in6p;
2854
2855		in6p = (struct in6pcb *)inp;
2856		ip6_freepcbopts(in6p->in6p_outputopts);
2857	}
2858#endif				/* INET6 */
2859	ip_pcb->inp_vflag = 0;
2860	/* free up authentication fields */
2861	if (inp->sctp_ep.local_auth_chunks != NULL)
2862		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
2863	if (inp->sctp_ep.local_hmacs != NULL)
2864		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
2865
2866	shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
2867	while (shared_key) {
2868		LIST_REMOVE(shared_key, next);
2869		sctp_free_sharedkey(shared_key);
2870		shared_key = LIST_FIRST(&inp->sctp_ep.shared_keys);
2871	}
2872
2873	inp_save = LIST_NEXT(inp, sctp_list);
2874	LIST_REMOVE(inp, sctp_list);
2875
2876	/* fix any iterators only after out of the list */
2877	sctp_iterator_inp_being_freed(inp, inp_save);
2878	/*
2879	 * if we have an address list the following will free the list of
2880	 * ifaddr's that are set into this ep. Again macro limitations here,
2881	 * since the LIST_FOREACH could be a bad idea.
2882	 */
2883	for ((laddr = LIST_FIRST(&inp->sctp_addr_list)); laddr != NULL;
2884	    laddr = nladdr) {
2885		nladdr = LIST_NEXT(laddr, sctp_nxt_addr);
2886		sctp_remove_laddr(laddr);
2887	}
2888
2889#ifdef SCTP_TRACK_FREED_ASOCS
2890	/* TEMP CODE */
2891	for ((asoc = LIST_FIRST(&inp->sctp_asoc_free_list)); asoc != NULL;
2892	    asoc = nasoc) {
2893		nasoc = LIST_NEXT(asoc, sctp_tcblist);
2894		LIST_REMOVE(asoc, sctp_tcblist);
2895		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, asoc);
2896		SCTP_DECR_ASOC_COUNT();
2897	}
2898	/* *** END TEMP CODE *** */
2899#endif
2900	/* Now lets see about freeing the EP hash table. */
2901	if (inp->sctp_tcbhash != NULL) {
2902		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
2903		inp->sctp_tcbhash = NULL;
2904	}
2905	/* Now we must put the ep memory back into the zone pool */
2906	SCTP_INP_LOCK_DESTROY(inp);
2907	SCTP_INP_READ_DESTROY(inp);
2908	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
2909	SCTP_INP_INFO_WUNLOCK();
2910
2911	SCTP_ITERATOR_UNLOCK();
2912
2913	SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_ep, inp);
2914	SCTP_DECR_EP_COUNT();
2915
2916}
2917
2918
2919struct sctp_nets *
2920sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
2921{
2922	struct sctp_nets *net;
2923
2924	/* locate the address */
2925	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2926		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
2927			return (net);
2928	}
2929	return (NULL);
2930}
2931
2932
2933/*
2934 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
2935 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
2936 * stats of stuff.
2937 */
2938int
2939sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
2940{
2941	struct sctp_ifa *sctp_ifa;
2942
2943	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, 0);
2944	if (sctp_ifa) {
2945		return (1);
2946	} else {
2947		return (0);
2948	}
2949}
2950
2951void
2952sctp_set_initial_cc_param(struct sctp_tcb *stcb, struct sctp_nets *net)
2953{
2954	net->cwnd = min((net->mtu * 4), max((2 * net->mtu), SCTP_INITIAL_CWND));
2955	/* we always get at LEAST 2 MTU's */
2956	if (net->cwnd < (2 * net->mtu)) {
2957		net->cwnd = 2 * net->mtu;
2958	}
2959	net->ssthresh = stcb->asoc.peers_rwnd;
2960}
2961
2962int
2963sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
2964    int set_scope, int from)
2965{
2966	/*
2967	 * The following is redundant to the same lines in the
2968	 * sctp_aloc_assoc() but is needed since other's call the add
2969	 * address function
2970	 */
2971	struct sctp_nets *net, *netfirst;
2972	int addr_inscope;
2973
2974#ifdef SCTP_DEBUG
2975	if (sctp_debug_on & SCTP_DEBUG_PCB1) {
2976		printf("Adding an address (from:%d) to the peer: ", from);
2977		sctp_print_address(newaddr);
2978	}
2979#endif
2980
2981	netfirst = sctp_findnet(stcb, newaddr);
2982	if (netfirst) {
2983		/*
2984		 * Lie and return ok, we don't want to make the association
2985		 * go away for this behavior. It will happen in the TCP
2986		 * model in a connected socket. It does not reach the hash
2987		 * table until after the association is built so it can't be
2988		 * found. Mark as reachable, since the initial creation will
2989		 * have been cleared and the NOT_IN_ASSOC flag will have
2990		 * been added... and we don't want to end up removing it
2991		 * back out.
2992		 */
2993		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
2994			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
2995			    SCTP_ADDR_UNCONFIRMED);
2996		} else {
2997			netfirst->dest_state = SCTP_ADDR_REACHABLE;
2998		}
2999
3000		return (0);
3001	}
3002	addr_inscope = 1;
3003	if (newaddr->sa_family == AF_INET) {
3004		struct sockaddr_in *sin;
3005
3006		sin = (struct sockaddr_in *)newaddr;
3007		if (sin->sin_addr.s_addr == 0) {
3008			/* Invalid address */
3009			return (-1);
3010		}
3011		/* zero out the bzero area */
3012		memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3013
3014		/* assure len is set */
3015		sin->sin_len = sizeof(struct sockaddr_in);
3016		if (set_scope) {
3017#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3018			stcb->ipv4_local_scope = 1;
3019#else
3020			if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3021				stcb->asoc.ipv4_local_scope = 1;
3022			}
3023#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3024		} else {
3025			/* Validate the address is in scope */
3026			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3027			    (stcb->asoc.ipv4_local_scope == 0)) {
3028				addr_inscope = 0;
3029			}
3030		}
3031	} else if (newaddr->sa_family == AF_INET6) {
3032		struct sockaddr_in6 *sin6;
3033
3034		sin6 = (struct sockaddr_in6 *)newaddr;
3035		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3036			/* Invalid address */
3037			return (-1);
3038		}
3039		/* assure len is set */
3040		sin6->sin6_len = sizeof(struct sockaddr_in6);
3041		if (set_scope) {
3042			if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3043				stcb->asoc.loopback_scope = 1;
3044				stcb->asoc.local_scope = 0;
3045				stcb->asoc.ipv4_local_scope = 1;
3046				stcb->asoc.site_scope = 1;
3047			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3048				/*
3049				 * If the new destination is a LINK_LOCAL we
3050				 * must have common site scope. Don't set
3051				 * the local scope since we may not share
3052				 * all links, only loopback can do this.
3053				 * Links on the local network would also be
3054				 * on our private network for v4 too.
3055				 */
3056				stcb->asoc.ipv4_local_scope = 1;
3057				stcb->asoc.site_scope = 1;
3058			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3059				/*
3060				 * If the new destination is SITE_LOCAL then
3061				 * we must have site scope in common.
3062				 */
3063				stcb->asoc.site_scope = 1;
3064			}
3065		} else {
3066			/* Validate the address is in scope */
3067			if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3068			    (stcb->asoc.loopback_scope == 0)) {
3069				addr_inscope = 0;
3070			} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3071			    (stcb->asoc.local_scope == 0)) {
3072				addr_inscope = 0;
3073			} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3074			    (stcb->asoc.site_scope == 0)) {
3075				addr_inscope = 0;
3076			}
3077		}
3078	} else {
3079		/* not supported family type */
3080		return (-1);
3081	}
3082	net = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_net, struct sctp_nets);
3083	if (net == NULL) {
3084		return (-1);
3085	}
3086	SCTP_INCR_RADDR_COUNT();
3087	bzero(net, sizeof(*net));
3088	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3089	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3090	if (newaddr->sa_family == AF_INET) {
3091		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3092	} else if (newaddr->sa_family == AF_INET6) {
3093		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3094	}
3095	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3096	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3097		stcb->asoc.loopback_scope = 1;
3098		stcb->asoc.ipv4_local_scope = 1;
3099		stcb->asoc.local_scope = 0;
3100		stcb->asoc.site_scope = 1;
3101		addr_inscope = 1;
3102	}
3103	net->failure_threshold = stcb->asoc.def_net_failure;
3104	if (addr_inscope == 0) {
3105		net->dest_state = (SCTP_ADDR_REACHABLE |
3106		    SCTP_ADDR_OUT_OF_SCOPE);
3107	} else {
3108		if (from == SCTP_ADDR_IS_CONFIRMED)
3109			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3110			net->dest_state = SCTP_ADDR_REACHABLE;
3111		else
3112			net->dest_state = SCTP_ADDR_REACHABLE |
3113			    SCTP_ADDR_UNCONFIRMED;
3114	}
3115	net->RTO = stcb->asoc.initial_rto;
3116	stcb->asoc.numnets++;
3117	*(&net->ref_count) = 1;
3118	net->tos_flowlabel = 0;
3119#ifdef INET
3120	if (newaddr->sa_family == AF_INET)
3121		net->tos_flowlabel = stcb->asoc.default_tos;
3122#endif
3123#ifdef INET6
3124	if (newaddr->sa_family == AF_INET6)
3125		net->tos_flowlabel = stcb->asoc.default_flowlabel;
3126#endif
3127	/* Init the timer structure */
3128	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3129	SCTP_OS_TIMER_INIT(&net->fr_timer.timer);
3130	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3131
3132	/* Now generate a route for this guy */
3133	/* KAME hack: embed scopeid */
3134	if (newaddr->sa_family == AF_INET6) {
3135		struct sockaddr_in6 *sin6;
3136
3137		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3138		(void)sa6_embedscope(sin6, ip6_use_defzone);
3139		sin6->sin6_scope_id = 0;
3140	}
3141	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id,
3142	    stcb->asoc.table_id);
3143
3144	if (newaddr->sa_family == AF_INET6) {
3145		struct sockaddr_in6 *sin6;
3146
3147		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3148		(void)sa6_recoverscope(sin6);
3149	}
3150	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3151		/* Get source address */
3152		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3153		    stcb,
3154		    (sctp_route_t *) & net->ro,
3155		    net,
3156		    0,
3157		    stcb->asoc.vrf_id);
3158		/* Now get the interface MTU */
3159		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3160			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3161		} else {
3162			net->mtu = 0;
3163		}
3164#ifdef SCTP_PRINT_FOR_B_AND_M
3165		printf("We have found an interface mtu of %d\n", net->mtu);
3166#endif
3167		if (net->mtu == 0) {
3168			/* Huh ?? */
3169			net->mtu = SCTP_DEFAULT_MTU;
3170		} else {
3171			uint32_t rmtu;
3172
3173			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3174#ifdef SCTP_PRINT_FOR_B_AND_M
3175			printf("The route mtu is %d\n", rmtu);
3176#endif
3177			if (rmtu == 0) {
3178				/*
3179				 * Start things off to match mtu of
3180				 * interface please.
3181				 */
3182				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3183				    net->ro.ro_rt, net->mtu);
3184			} else {
3185				/*
3186				 * we take the route mtu over the interface,
3187				 * since the route may be leading out the
3188				 * loopback, or a different interface.
3189				 */
3190				net->mtu = rmtu;
3191			}
3192		}
3193		if (from == SCTP_ALLOC_ASOC) {
3194#ifdef SCTP_PRINT_FOR_B_AND_M
3195			printf("New assoc sets mtu to :%d\n",
3196			    net->mtu);
3197#endif
3198			stcb->asoc.smallest_mtu = net->mtu;
3199		}
3200	} else {
3201		net->mtu = stcb->asoc.smallest_mtu;
3202	}
3203	if (stcb->asoc.smallest_mtu > net->mtu) {
3204#ifdef SCTP_PRINT_FOR_B_AND_M
3205		printf("new address mtu:%d smaller than smallest:%d\n",
3206		    net->mtu, stcb->asoc.smallest_mtu);
3207#endif
3208		stcb->asoc.smallest_mtu = net->mtu;
3209	}
3210	/*
3211	 * We take the max of the burst limit times a MTU or the
3212	 * INITIAL_CWND. We then limit this to 4 MTU's of sending.
3213	 */
3214	sctp_set_initial_cc_param(stcb, net);
3215
3216
3217#if defined(SCTP_CWND_MONITOR) || defined(SCTP_CWND_LOGGING)
3218	sctp_log_cwnd(stcb, net, 0, SCTP_CWND_INITIALIZATION);
3219#endif
3220
3221	/*
3222	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3223	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
3224	 */
3225	net->find_pseudo_cumack = 1;
3226	net->find_rtx_pseudo_cumack = 1;
3227	net->src_addr_selected = 0;
3228	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
3229	if (net->ro.ro_rt == NULL) {
3230		/* Since we have no route put it at the back */
3231		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3232	} else if (netfirst == NULL) {
3233		/* We are the first one in the pool. */
3234		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3235	} else if (netfirst->ro.ro_rt == NULL) {
3236		/*
3237		 * First one has NO route. Place this one ahead of the first
3238		 * one.
3239		 */
3240		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3241	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
3242		/*
3243		 * This one has a different interface than the one at the
3244		 * top of the list. Place it ahead.
3245		 */
3246		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3247	} else {
3248		/*
3249		 * Ok we have the same interface as the first one. Move
3250		 * forward until we find either a) one with a NULL route...
3251		 * insert ahead of that b) one with a different ifp.. insert
3252		 * after that. c) end of the list.. insert at the tail.
3253		 */
3254		struct sctp_nets *netlook;
3255
3256		do {
3257			netlook = TAILQ_NEXT(netfirst, sctp_next);
3258			if (netlook == NULL) {
3259				/* End of the list */
3260				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3261				break;
3262			} else if (netlook->ro.ro_rt == NULL) {
3263				/* next one has NO route */
3264				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
3265				break;
3266			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
3267				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
3268				    net, sctp_next);
3269				break;
3270			}
3271			/* Shift forward */
3272			netfirst = netlook;
3273		} while (netlook != NULL);
3274	}
3275
3276	/* got to have a primary set */
3277	if (stcb->asoc.primary_destination == 0) {
3278		stcb->asoc.primary_destination = net;
3279	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
3280		    (net->ro.ro_rt) &&
3281	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
3282		/* No route to current primary adopt new primary */
3283		stcb->asoc.primary_destination = net;
3284	}
3285	sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, stcb->sctp_ep, stcb,
3286	    net);
3287	/* Validate primary is first */
3288	net = TAILQ_FIRST(&stcb->asoc.nets);
3289	if ((net != stcb->asoc.primary_destination) &&
3290	    (stcb->asoc.primary_destination)) {
3291		/*
3292		 * first one on the list is NOT the primary sctp_cmpaddr()
3293		 * is much more efficent if the primary is the first on the
3294		 * list, make it so.
3295		 */
3296		TAILQ_REMOVE(&stcb->asoc.nets,
3297		    stcb->asoc.primary_destination, sctp_next);
3298		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
3299		    stcb->asoc.primary_destination, sctp_next);
3300	}
3301	return (0);
3302}
3303
3304
3305/*
3306 * allocate an association and add it to the endpoint. The caller must be
3307 * careful to add all additional addresses once they are know right away or
3308 * else the assoc will be may experience a blackout scenario.
3309 */
3310struct sctp_tcb *
3311sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
3312    int for_a_init, int *error, uint32_t override_tag, uint32_t vrf_id)
3313{
3314	struct sctp_tcb *stcb;
3315	struct sctp_association *asoc;
3316	struct sctpasochead *head;
3317	uint16_t rport;
3318	int err;
3319
3320	/*
3321	 * Assumption made here: Caller has done a
3322	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
3323	 * address does not exist already.
3324	 */
3325	if (sctppcbinfo.ipi_count_asoc >= SCTP_MAX_NUM_OF_ASOC) {
3326		/* Hit max assoc, sorry no more */
3327		*error = ENOBUFS;
3328		return (NULL);
3329	}
3330	SCTP_INP_RLOCK(inp);
3331	if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
3332		/*
3333		 * If its in the TCP pool, its NOT allowed to create an
3334		 * association. The parent listener needs to call
3335		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
3336		 * off, or connected one does this.. its an error.
3337		 */
3338		SCTP_INP_RUNLOCK(inp);
3339		*error = EINVAL;
3340		return (NULL);
3341	}
3342#ifdef SCTP_DEBUG
3343	if (sctp_debug_on & SCTP_DEBUG_PCB3) {
3344		printf("Allocate an association for peer:");
3345		if (firstaddr) {
3346			sctp_print_address(firstaddr);
3347			printf("Port:%d\n",
3348			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
3349		} else
3350			printf("None\n");
3351	}
3352#endif				/* SCTP_DEBUG */
3353	if (firstaddr->sa_family == AF_INET) {
3354		struct sockaddr_in *sin;
3355
3356		sin = (struct sockaddr_in *)firstaddr;
3357		if ((sin->sin_port == 0) || (sin->sin_addr.s_addr == 0)) {
3358			/* Invalid address */
3359			SCTP_INP_RUNLOCK(inp);
3360			*error = EINVAL;
3361			return (NULL);
3362		}
3363		rport = sin->sin_port;
3364	} else if (firstaddr->sa_family == AF_INET6) {
3365		struct sockaddr_in6 *sin6;
3366
3367		sin6 = (struct sockaddr_in6 *)firstaddr;
3368		if ((sin6->sin6_port == 0) ||
3369		    (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))) {
3370			/* Invalid address */
3371			SCTP_INP_RUNLOCK(inp);
3372			*error = EINVAL;
3373			return (NULL);
3374		}
3375		rport = sin6->sin6_port;
3376	} else {
3377		/* not supported family type */
3378		SCTP_INP_RUNLOCK(inp);
3379		*error = EINVAL;
3380		return (NULL);
3381	}
3382	SCTP_INP_RUNLOCK(inp);
3383	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
3384		/*
3385		 * If you have not performed a bind, then we need to do the
3386		 * ephemerial bind for you.
3387		 */
3388		if ((err = sctp_inpcb_bind(inp->sctp_socket,
3389		    (struct sockaddr *)NULL,
3390		    (struct thread *)NULL
3391		    ))) {
3392			/* bind error, probably perm */
3393			*error = err;
3394			return (NULL);
3395		}
3396	}
3397	stcb = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_asoc, struct sctp_tcb);
3398	if (stcb == NULL) {
3399		/* out of memory? */
3400		*error = ENOMEM;
3401		return (NULL);
3402	}
3403	SCTP_INCR_ASOC_COUNT();
3404
3405	bzero(stcb, sizeof(*stcb));
3406	asoc = &stcb->asoc;
3407	SCTP_TCB_LOCK_INIT(stcb);
3408	SCTP_TCB_SEND_LOCK_INIT(stcb);
3409	/* setup back pointer's */
3410	stcb->sctp_ep = inp;
3411	stcb->sctp_socket = inp->sctp_socket;
3412	if ((err = sctp_init_asoc(inp, asoc, for_a_init, override_tag, vrf_id))) {
3413		/* failed */
3414		SCTP_TCB_LOCK_DESTROY(stcb);
3415		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
3416		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
3417		SCTP_DECR_ASOC_COUNT();
3418		*error = err;
3419		return (NULL);
3420	}
3421	/* and the port */
3422	stcb->rport = rport;
3423	SCTP_INP_INFO_WLOCK();
3424	SCTP_INP_WLOCK(inp);
3425	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
3426		/* inpcb freed while alloc going on */
3427		SCTP_TCB_LOCK_DESTROY(stcb);
3428		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
3429		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
3430		SCTP_INP_WUNLOCK(inp);
3431		SCTP_INP_INFO_WUNLOCK();
3432		SCTP_DECR_ASOC_COUNT();
3433		*error = EINVAL;
3434		return (NULL);
3435	}
3436	SCTP_TCB_LOCK(stcb);
3437
3438	/* now that my_vtag is set, add it to the hash */
3439	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag,
3440	    sctppcbinfo.hashasocmark)];
3441	/* put it in the bucket in the vtag hash of assoc's for the system */
3442	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
3443	SCTP_INP_INFO_WUNLOCK();
3444
3445	if ((err = sctp_add_remote_addr(stcb, firstaddr, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
3446		/* failure.. memory error? */
3447		if (asoc->strmout)
3448			SCTP_FREE(asoc->strmout);
3449		if (asoc->mapping_array)
3450			SCTP_FREE(asoc->mapping_array);
3451
3452		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
3453		SCTP_DECR_ASOC_COUNT();
3454		SCTP_TCB_LOCK_DESTROY(stcb);
3455		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
3456		SCTP_INP_WUNLOCK(inp);
3457		*error = ENOBUFS;
3458		return (NULL);
3459	}
3460	/* Init all the timers */
3461	SCTP_OS_TIMER_INIT(&asoc->hb_timer.timer);
3462	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
3463	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
3464	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
3465	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
3466	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
3467	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
3468
3469	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
3470	/* now file the port under the hash as well */
3471	if (inp->sctp_tcbhash != NULL) {
3472		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
3473		    inp->sctp_hashmark)];
3474		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
3475	}
3476	SCTP_INP_WUNLOCK(inp);
3477#ifdef SCTP_DEBUG
3478	if (sctp_debug_on & SCTP_DEBUG_PCB1) {
3479		printf("Association %p now allocated\n", stcb);
3480	}
3481#endif
3482	return (stcb);
3483}
3484
3485
3486void
3487sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
3488{
3489	struct sctp_association *asoc;
3490
3491	asoc = &stcb->asoc;
3492	asoc->numnets--;
3493	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
3494	if (net == asoc->primary_destination) {
3495		/* Reset primary */
3496		struct sctp_nets *lnet;
3497
3498		lnet = TAILQ_FIRST(&asoc->nets);
3499		/* Try to find a confirmed primary */
3500		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
3501	}
3502	if (net == asoc->last_data_chunk_from) {
3503		/* Reset primary */
3504		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
3505	}
3506	if (net == asoc->last_control_chunk_from) {
3507		/* Clear net */
3508		asoc->last_control_chunk_from = NULL;
3509	}
3510	sctp_free_remote_addr(net);
3511}
3512
3513/*
3514 * remove a remote endpoint address from an association, it will fail if the
3515 * address does not exist.
3516 */
3517int
3518sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
3519{
3520	/*
3521	 * Here we need to remove a remote address. This is quite simple, we
3522	 * first find it in the list of address for the association
3523	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
3524	 * on that item. Note we do not allow it to be removed if there are
3525	 * no other addresses.
3526	 */
3527	struct sctp_association *asoc;
3528	struct sctp_nets *net, *net_tmp;
3529
3530	asoc = &stcb->asoc;
3531
3532	/* locate the address */
3533	for (net = TAILQ_FIRST(&asoc->nets); net != NULL; net = net_tmp) {
3534		net_tmp = TAILQ_NEXT(net, sctp_next);
3535		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
3536			continue;
3537		}
3538		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
3539		    remaddr)) {
3540			/* we found the guy */
3541			if (asoc->numnets < 2) {
3542				/* Must have at LEAST two remote addresses */
3543				return (-1);
3544			} else {
3545				sctp_remove_net(stcb, net);
3546				return (0);
3547			}
3548		}
3549	}
3550	/* not found. */
3551	return (-2);
3552}
3553
3554
3555void
3556sctp_add_vtag_to_timewait(struct sctp_inpcb *inp, uint32_t tag, uint32_t time)
3557{
3558	struct sctpvtaghead *chain;
3559	struct sctp_tagblock *twait_block;
3560	struct timeval now;
3561	int set, i;
3562
3563	(void)SCTP_GETTIME_TIMEVAL(&now);
3564	chain = &sctppcbinfo.vtag_timewait[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
3565	set = 0;
3566	if (!SCTP_LIST_EMPTY(chain)) {
3567		/* Block(s) present, lets find space, and expire on the fly */
3568		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
3569			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
3570				if ((twait_block->vtag_block[i].v_tag == 0) &&
3571				    !set) {
3572					twait_block->vtag_block[i].tv_sec_at_expire =
3573					    now.tv_sec + time;
3574					twait_block->vtag_block[i].v_tag = tag;
3575					set = 1;
3576				} else if ((twait_block->vtag_block[i].v_tag) &&
3577					    ((long)twait_block->vtag_block[i].tv_sec_at_expire >
3578				    now.tv_sec)) {
3579					/* Audit expires this guy */
3580					twait_block->vtag_block[i].tv_sec_at_expire = 0;
3581					twait_block->vtag_block[i].v_tag = 0;
3582					if (set == 0) {
3583						/* Reuse it for my new tag */
3584						twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + SCTP_TIME_WAIT;
3585						twait_block->vtag_block[0].v_tag = tag;
3586						set = 1;
3587					}
3588				}
3589			}
3590			if (set) {
3591				/*
3592				 * We only do up to the block where we can
3593				 * place our tag for audits
3594				 */
3595				break;
3596			}
3597		}
3598	}
3599	/* Need to add a new block to chain */
3600	if (!set) {
3601		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
3602		    sizeof(struct sctp_tagblock), "TimeWait");
3603		if (twait_block == NULL) {
3604			return;
3605		}
3606		memset(twait_block, 0, sizeof(struct sctp_tagblock));
3607		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
3608		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec +
3609		    SCTP_TIME_WAIT;
3610		twait_block->vtag_block[0].v_tag = tag;
3611	}
3612}
3613
3614
3615static void
3616sctp_iterator_asoc_being_freed(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
3617{
3618	struct sctp_iterator *it;
3619
3620	/*
3621	 * Unlock the tcb lock we do this so we avoid a dead lock scenario
3622	 * where the iterator is waiting on the TCB lock and the TCB lock is
3623	 * waiting on the iterator lock.
3624	 */
3625	it = stcb->asoc.stcb_starting_point_for_iterator;
3626	if (it == NULL) {
3627		return;
3628	}
3629	if (it->inp != stcb->sctp_ep) {
3630		/* hmm, focused on the wrong one? */
3631		return;
3632	}
3633	if (it->stcb != stcb) {
3634		return;
3635	}
3636	it->stcb = LIST_NEXT(stcb, sctp_tcblist);
3637	if (it->stcb == NULL) {
3638		/* done with all asoc's in this assoc */
3639		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3640			it->inp = NULL;
3641		} else {
3642			it->inp = LIST_NEXT(inp, sctp_list);
3643		}
3644	}
3645}
3646
3647
3648/*
3649 * Free the association after un-hashing the remote port.
3650 */
3651int
3652sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
3653{
3654	int i;
3655	struct sctp_association *asoc;
3656	struct sctp_nets *net, *prev;
3657	struct sctp_laddr *laddr;
3658	struct sctp_tmit_chunk *chk;
3659	struct sctp_asconf_addr *aparam;
3660	struct sctp_stream_reset_list *liste;
3661	struct sctp_queued_to_read *sq;
3662	struct sctp_stream_queue_pending *sp;
3663	sctp_sharedkey_t *shared_key;
3664	struct socket *so;
3665	int ccnt = 0;
3666	int cnt = 0;
3667
3668	/* first, lets purge the entry from the hash table. */
3669
3670#ifdef SCTP_LOG_CLOSING
3671	sctp_log_closing(inp, stcb, 6);
3672#endif
3673	if (stcb->asoc.state == 0) {
3674#ifdef SCTP_LOG_CLOSING
3675		sctp_log_closing(inp, NULL, 7);
3676#endif
3677		/* there is no asoc, really TSNH :-0 */
3678		return (1);
3679	}
3680	/* TEMP CODE */
3681	if (stcb->freed_from_where == 0) {
3682		/* Only record the first place free happened from */
3683		stcb->freed_from_where = from_location;
3684	}
3685	/* TEMP CODE */
3686
3687	asoc = &stcb->asoc;
3688	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
3689	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
3690		/* nothing around */
3691		so = NULL;
3692	else
3693		so = inp->sctp_socket;
3694
3695	/*
3696	 * We used timer based freeing if a reader or writer is in the way.
3697	 * So we first check if we are actually being called from a timer,
3698	 * if so we abort early if a reader or writer is still in the way.
3699	 */
3700	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
3701	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
3702		/*
3703		 * is it the timer driving us? if so are the reader/writers
3704		 * gone?
3705		 */
3706		if (stcb->asoc.refcnt) {
3707			/* nope, reader or writer in the way */
3708			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
3709			/* no asoc destroyed */
3710			SCTP_TCB_UNLOCK(stcb);
3711#ifdef SCTP_LOG_CLOSING
3712			sctp_log_closing(inp, stcb, 8);
3713#endif
3714			return (0);
3715		}
3716	}
3717	/* now clean up any other timers */
3718	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
3719	asoc->hb_timer.self = NULL;
3720	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
3721	asoc->dack_timer.self = NULL;
3722	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
3723	/*-
3724	 * For stream reset we don't blast this unless
3725	 * it is a str-reset timer, it might be the
3726	 * free-asoc timer which we DON'T want to
3727	 * disturb.
3728	 */
3729	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
3730		asoc->strreset_timer.self = NULL;
3731	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
3732	asoc->asconf_timer.self = NULL;
3733	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
3734	asoc->autoclose_timer.self = NULL;
3735	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
3736	asoc->shut_guard_timer.self = NULL;
3737	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
3738	asoc->delayed_event_timer.self = NULL;
3739	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
3740		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
3741		net->fr_timer.self = NULL;
3742		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
3743		net->rxt_timer.self = NULL;
3744		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
3745		net->pmtu_timer.self = NULL;
3746	}
3747	/* Now the read queue needs to be cleaned up (only once) */
3748	cnt = 0;
3749	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
3750		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
3751		SCTP_INP_READ_LOCK(inp);
3752		TAILQ_FOREACH(sq, &inp->read_queue, next) {
3753			if (sq->stcb == stcb) {
3754				sq->do_not_ref_stcb = 1;
3755				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
3756				/*
3757				 * If there is no end, there never will be
3758				 * now.
3759				 */
3760				if (sq->end_added == 0) {
3761					/* Held for PD-API clear that. */
3762					sq->pdapi_aborted = 1;
3763					sq->held_length = 0;
3764					if (sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PDAPIEVNT)) {
3765						/*
3766						 * Need to add a PD-API
3767						 * aborted indication.
3768						 * Setting the control_pdapi
3769						 * assures that it will be
3770						 * added right after this
3771						 * msg.
3772						 */
3773						uint32_t strseq;
3774
3775						stcb->asoc.control_pdapi = sq;
3776						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
3777						sctp_notify_partial_delivery_indication(stcb,
3778						    SCTP_PARTIAL_DELIVERY_ABORTED, 1, strseq);
3779						stcb->asoc.control_pdapi = NULL;
3780					}
3781				}
3782				/* Add an end to wake them */
3783				sq->end_added = 1;
3784				cnt++;
3785			}
3786		}
3787		SCTP_INP_READ_UNLOCK(inp);
3788		if (stcb->block_entry) {
3789			cnt++;
3790			stcb->block_entry->error = ECONNRESET;
3791			stcb->block_entry = NULL;
3792		}
3793	}
3794	if ((from_inpcbfree != SCTP_PCBFREE_FORCE) && (stcb->asoc.refcnt)) {
3795		/*
3796		 * reader or writer in the way, we have hopefully given him
3797		 * something to chew on above.
3798		 */
3799		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
3800		SCTP_TCB_UNLOCK(stcb);
3801		if (so) {
3802			SCTP_INP_RLOCK(inp);
3803			if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
3804			    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
3805				/* nothing around */
3806				so = NULL;
3807			if (so) {
3808				/* Wake any reader/writers */
3809				sctp_sorwakeup(inp, so);
3810				sctp_sowwakeup(inp, so);
3811			}
3812			SCTP_INP_RUNLOCK(inp);
3813
3814		}
3815#ifdef SCTP_LOG_CLOSING
3816		sctp_log_closing(inp, stcb, 9);
3817#endif
3818		/* no asoc destroyed */
3819		return (0);
3820	}
3821#ifdef SCTP_LOG_CLOSING
3822	sctp_log_closing(inp, stcb, 10);
3823#endif
3824	/*
3825	 * When I reach here, no others want to kill the assoc yet.. and I
3826	 * own the lock. Now its possible an abort comes in when I do the
3827	 * lock exchange below to grab all the locks to do the final take
3828	 * out. to prevent this we increment the count, which will start a
3829	 * timer and blow out above thus assuring us that we hold exclusive
3830	 * killing of the asoc. Note that after getting back the TCB lock we
3831	 * will go ahead and increment the counter back up and stop any
3832	 * timer a passing stranger may have started :-S
3833	 */
3834	if (from_inpcbfree == SCTP_NORMAL_PROC) {
3835		atomic_add_int(&stcb->asoc.refcnt, 1);
3836
3837		SCTP_TCB_UNLOCK(stcb);
3838
3839		SCTP_ITERATOR_LOCK();
3840		SCTP_INP_INFO_WLOCK();
3841		SCTP_INP_WLOCK(inp);
3842		SCTP_TCB_LOCK(stcb);
3843	}
3844	/* Double check the GONE flag */
3845	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
3846	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
3847		/* nothing around */
3848		so = NULL;
3849
3850	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
3851	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
3852		/*
3853		 * For TCP type we need special handling when we are
3854		 * connected. We also include the peel'ed off ones to.
3855		 */
3856		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
3857			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
3858			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
3859			if (so) {
3860				SOCK_LOCK(so);
3861				if (so->so_rcv.sb_cc == 0) {
3862					so->so_state &= ~(SS_ISCONNECTING |
3863					    SS_ISDISCONNECTING |
3864					    SS_ISCONFIRMING |
3865					    SS_ISCONNECTED);
3866				}
3867				SOCK_UNLOCK(so);
3868				sctp_sowwakeup(inp, so);
3869				sctp_sorwakeup(inp, so);
3870				SCTP_SOWAKEUP(so);
3871			}
3872		}
3873	}
3874	/*
3875	 * Make it invalid too, that way if its about to run it will abort
3876	 * and return.
3877	 */
3878	sctp_iterator_asoc_being_freed(inp, stcb);
3879	/* re-increment the lock */
3880	if (from_inpcbfree == SCTP_NORMAL_PROC) {
3881		atomic_add_int(&stcb->asoc.refcnt, -1);
3882	}
3883	asoc->state = 0;
3884	if (inp->sctp_tcbhash) {
3885		LIST_REMOVE(stcb, sctp_tcbhash);
3886	}
3887	if (stcb->asoc.in_restart_hash) {
3888		LIST_REMOVE(stcb, sctp_tcbrestarhash);
3889	}
3890	/* Now lets remove it from the list of ALL associations in the EP */
3891	LIST_REMOVE(stcb, sctp_tcblist);
3892	if (from_inpcbfree == SCTP_NORMAL_PROC) {
3893		SCTP_INP_INCR_REF(inp);
3894		SCTP_INP_WUNLOCK(inp);
3895		SCTP_ITERATOR_UNLOCK();
3896	}
3897	/* pull from vtag hash */
3898	LIST_REMOVE(stcb, sctp_asocs);
3899	sctp_add_vtag_to_timewait(inp, asoc->my_vtag, SCTP_TIME_WAIT);
3900
3901
3902	/*
3903	 * Now restop the timers to be sure - this is paranoia at is finest!
3904	 */
3905	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
3906	(void)SCTP_OS_TIMER_STOP(&asoc->hb_timer.timer);
3907	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
3908	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
3909	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
3910	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
3911	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
3912	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
3913
3914	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
3915		(void)SCTP_OS_TIMER_STOP(&net->fr_timer.timer);
3916		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
3917		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
3918	}
3919
3920	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
3921	prev = NULL;
3922	/*
3923	 * The chunk lists and such SHOULD be empty but we check them just
3924	 * in case.
3925	 */
3926	/* anything on the wheel needs to be removed */
3927	for (i = 0; i < asoc->streamoutcnt; i++) {
3928		struct sctp_stream_out *outs;
3929
3930		outs = &asoc->strmout[i];
3931		/* now clean up any chunks here */
3932		sp = TAILQ_FIRST(&outs->outqueue);
3933		while (sp) {
3934			TAILQ_REMOVE(&outs->outqueue, sp, next);
3935			if (sp->data) {
3936				sctp_m_freem(sp->data);
3937				sp->data = NULL;
3938				sp->tail_mbuf = NULL;
3939			}
3940			sctp_free_remote_addr(sp->net);
3941			sctp_free_spbufspace(stcb, asoc, sp);
3942			/* Free the zone stuff  */
3943			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_strmoq, sp);
3944			SCTP_DECR_STRMOQ_COUNT();
3945			sp = TAILQ_FIRST(&outs->outqueue);
3946		}
3947	}
3948
3949	while ((liste = TAILQ_FIRST(&asoc->resetHead)) != NULL) {
3950		TAILQ_REMOVE(&asoc->resetHead, liste, next_resp);
3951		SCTP_FREE(liste);
3952	}
3953
3954	sq = TAILQ_FIRST(&asoc->pending_reply_queue);
3955	while (sq) {
3956		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
3957		if (sq->data) {
3958			sctp_m_freem(sq->data);
3959			sq->data = NULL;
3960		}
3961		sctp_free_remote_addr(sq->whoFrom);
3962		sq->whoFrom = NULL;
3963		sq->stcb = NULL;
3964		/* Free the ctl entry */
3965		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, sq);
3966		SCTP_DECR_READQ_COUNT();
3967		sq = TAILQ_FIRST(&asoc->pending_reply_queue);
3968	}
3969
3970	chk = TAILQ_FIRST(&asoc->free_chunks);
3971	while (chk) {
3972		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
3973		if (chk->data) {
3974			sctp_m_freem(chk->data);
3975			chk->data = NULL;
3976		}
3977		ccnt++;
3978		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
3979		SCTP_DECR_CHK_COUNT();
3980		atomic_subtract_int(&sctppcbinfo.ipi_free_chunks, 1);
3981		asoc->free_chunk_cnt--;
3982		chk = TAILQ_FIRST(&asoc->free_chunks);
3983	}
3984	/* pending send queue SHOULD be empty */
3985	if (!TAILQ_EMPTY(&asoc->send_queue)) {
3986		chk = TAILQ_FIRST(&asoc->send_queue);
3987		while (chk) {
3988			TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
3989			if (chk->data) {
3990				sctp_m_freem(chk->data);
3991				chk->data = NULL;
3992			}
3993			ccnt++;
3994			sctp_free_remote_addr(chk->whoTo);
3995			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
3996			SCTP_DECR_CHK_COUNT();
3997			chk = TAILQ_FIRST(&asoc->send_queue);
3998		}
3999	}
4000/*
4001  if(ccnt) {
4002  printf("Freed %d from send_queue\n", ccnt);
4003  ccnt = 0;
4004  }
4005*/
4006	/* sent queue SHOULD be empty */
4007	if (!TAILQ_EMPTY(&asoc->sent_queue)) {
4008		chk = TAILQ_FIRST(&asoc->sent_queue);
4009		while (chk) {
4010			TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4011			if (chk->data) {
4012				sctp_m_freem(chk->data);
4013				chk->data = NULL;
4014			}
4015			ccnt++;
4016			sctp_free_remote_addr(chk->whoTo);
4017			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4018			SCTP_DECR_CHK_COUNT();
4019			chk = TAILQ_FIRST(&asoc->sent_queue);
4020		}
4021	}
4022/*
4023  if(ccnt) {
4024  printf("Freed %d from sent_queue\n", ccnt);
4025  ccnt = 0;
4026  }
4027*/
4028	/* control queue MAY not be empty */
4029	if (!TAILQ_EMPTY(&asoc->control_send_queue)) {
4030		chk = TAILQ_FIRST(&asoc->control_send_queue);
4031		while (chk) {
4032			TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
4033			if (chk->data) {
4034				sctp_m_freem(chk->data);
4035				chk->data = NULL;
4036			}
4037			ccnt++;
4038			sctp_free_remote_addr(chk->whoTo);
4039			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4040			SCTP_DECR_CHK_COUNT();
4041			chk = TAILQ_FIRST(&asoc->control_send_queue);
4042		}
4043	}
4044/*
4045  if(ccnt) {
4046  printf("Freed %d from ctrl_queue\n", ccnt);
4047  ccnt = 0;
4048  }
4049*/
4050	if (!TAILQ_EMPTY(&asoc->reasmqueue)) {
4051		chk = TAILQ_FIRST(&asoc->reasmqueue);
4052		while (chk) {
4053			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
4054			if (chk->data) {
4055				sctp_m_freem(chk->data);
4056				chk->data = NULL;
4057			}
4058			sctp_free_remote_addr(chk->whoTo);
4059			ccnt++;
4060			SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_chunk, chk);
4061			SCTP_DECR_CHK_COUNT();
4062			chk = TAILQ_FIRST(&asoc->reasmqueue);
4063		}
4064	}
4065/*
4066  if(ccnt) {
4067  printf("Freed %d from reasm_queue\n", ccnt);
4068  ccnt = 0;
4069  }
4070*/
4071	if (asoc->mapping_array) {
4072		SCTP_FREE(asoc->mapping_array);
4073		asoc->mapping_array = NULL;
4074	}
4075	/* the stream outs */
4076	if (asoc->strmout) {
4077		SCTP_FREE(asoc->strmout);
4078		asoc->strmout = NULL;
4079	}
4080	asoc->streamoutcnt = 0;
4081	if (asoc->strmin) {
4082		struct sctp_queued_to_read *ctl;
4083
4084		for (i = 0; i < asoc->streamincnt; i++) {
4085			if (!TAILQ_EMPTY(&asoc->strmin[i].inqueue)) {
4086				/* We have somethings on the streamin queue */
4087				ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
4088				while (ctl) {
4089					TAILQ_REMOVE(&asoc->strmin[i].inqueue,
4090					    ctl, next);
4091					sctp_free_remote_addr(ctl->whoFrom);
4092					if (ctl->data) {
4093						sctp_m_freem(ctl->data);
4094						ctl->data = NULL;
4095					}
4096					/*
4097					 * We don't free the address here
4098					 * since all the net's were freed
4099					 * above.
4100					 */
4101					SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, ctl);
4102					SCTP_DECR_READQ_COUNT();
4103					ctl = TAILQ_FIRST(&asoc->strmin[i].inqueue);
4104				}
4105			}
4106		}
4107		SCTP_FREE(asoc->strmin);
4108		asoc->strmin = NULL;
4109	}
4110	asoc->streamincnt = 0;
4111	while (!TAILQ_EMPTY(&asoc->nets)) {
4112		net = TAILQ_FIRST(&asoc->nets);
4113		/* pull from list */
4114		if ((sctppcbinfo.ipi_count_raddr == 0) || (prev == net)) {
4115#ifdef INVARIANTS
4116			panic("no net's left alloc'ed, or list points to itself");
4117#endif
4118			break;
4119		}
4120		prev = net;
4121		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4122		sctp_free_remote_addr(net);
4123	}
4124
4125	while (!SCTP_LIST_EMPTY(&asoc->sctp_restricted_addrs)) {
4126		laddr = LIST_FIRST(&asoc->sctp_restricted_addrs);
4127		sctp_remove_laddr(laddr);
4128	}
4129
4130	/* pending asconf (address) parameters */
4131	while (!TAILQ_EMPTY(&asoc->asconf_queue)) {
4132		aparam = TAILQ_FIRST(&asoc->asconf_queue);
4133		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
4134		SCTP_FREE(aparam);
4135	}
4136	if (asoc->last_asconf_ack_sent != NULL) {
4137		sctp_m_freem(asoc->last_asconf_ack_sent);
4138		asoc->last_asconf_ack_sent = NULL;
4139	}
4140	/* clean up auth stuff */
4141	if (asoc->local_hmacs)
4142		sctp_free_hmaclist(asoc->local_hmacs);
4143	if (asoc->peer_hmacs)
4144		sctp_free_hmaclist(asoc->peer_hmacs);
4145
4146	if (asoc->local_auth_chunks)
4147		sctp_free_chunklist(asoc->local_auth_chunks);
4148	if (asoc->peer_auth_chunks)
4149		sctp_free_chunklist(asoc->peer_auth_chunks);
4150
4151	sctp_free_authinfo(&asoc->authinfo);
4152
4153	shared_key = LIST_FIRST(&asoc->shared_keys);
4154	while (shared_key) {
4155		LIST_REMOVE(shared_key, next);
4156		sctp_free_sharedkey(shared_key);
4157		shared_key = LIST_FIRST(&asoc->shared_keys);
4158	}
4159
4160	/* Insert new items here :> */
4161
4162	/* Get rid of LOCK */
4163	SCTP_TCB_LOCK_DESTROY(stcb);
4164	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4165	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4166		SCTP_INP_INFO_WUNLOCK();
4167		SCTP_INP_RLOCK(inp);
4168	}
4169#ifdef SCTP_TRACK_FREED_ASOCS
4170	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
4171		/* now clean up the tasoc itself */
4172		SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
4173		SCTP_DECR_ASOC_COUNT();
4174	} else {
4175		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
4176	}
4177#else
4178	SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_asoc, stcb);
4179	SCTP_DECR_ASOC_COUNT();
4180#endif
4181	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4182		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
4183			/*
4184			 * If its NOT the inp_free calling us AND sctp_close
4185			 * as been called, we call back...
4186			 */
4187			SCTP_INP_RUNLOCK(inp);
4188			/*
4189			 * This will start the kill timer (if we are the
4190			 * lastone) since we hold an increment yet. But this
4191			 * is the only safe way to do this since otherwise
4192			 * if the socket closes at the same time we are here
4193			 * we might collide in the cleanup.
4194			 */
4195			sctp_inpcb_free(inp,
4196			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
4197			    SCTP_CALLED_DIRECTLY_NOCMPSET);
4198			SCTP_INP_DECR_REF(inp);
4199			goto out_of;
4200		} else {
4201			/* The socket is still open. */
4202			SCTP_INP_DECR_REF(inp);
4203		}
4204	}
4205	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4206		SCTP_INP_RUNLOCK(inp);
4207	}
4208out_of:
4209	/* destroyed the asoc */
4210#ifdef SCTP_LOG_CLOSING
4211	sctp_log_closing(inp, NULL, 11);
4212#endif
4213	return (1);
4214}
4215
4216
4217
4218/*
4219 * determine if a destination is "reachable" based upon the addresses bound
4220 * to the current endpoint (e.g. only v4 or v6 currently bound)
4221 */
4222/*
4223 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
4224 * assoc level v4/v6 flags, as the assoc *may* not have the same address
4225 * types bound as its endpoint
4226 */
4227int
4228sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
4229{
4230	struct sctp_inpcb *inp;
4231	int answer;
4232
4233	/*
4234	 * No locks here, the TCB, in all cases is already locked and an
4235	 * assoc is up. There is either a INP lock by the caller applied (in
4236	 * asconf case when deleting an address) or NOT in the HB case,
4237	 * however if HB then the INP increment is up and the INP will not
4238	 * be removed (on top of the fact that we have a TCB lock). So we
4239	 * only want to read the sctp_flags, which is either bound-all or
4240	 * not.. no protection needed since once an assoc is up you can't be
4241	 * changing your binding.
4242	 */
4243	inp = stcb->sctp_ep;
4244	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
4245		/* if bound all, destination is not restricted */
4246		/*
4247		 * RRS: Question during lock work: Is this correct? If you
4248		 * are bound-all you still might need to obey the V4--V6
4249		 * flags??? IMO this bound-all stuff needs to be removed!
4250		 */
4251		return (1);
4252	}
4253	/* NOTE: all "scope" checks are done when local addresses are added */
4254	if (destaddr->sa_family == AF_INET6) {
4255		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
4256	} else if (destaddr->sa_family == AF_INET) {
4257		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
4258	} else {
4259		/* invalid family, so it's unreachable */
4260		answer = 0;
4261	}
4262	return (answer);
4263}
4264
4265/*
4266 * update the inp_vflags on an endpoint
4267 */
4268static void
4269sctp_update_ep_vflag(struct sctp_inpcb *inp)
4270{
4271	struct sctp_laddr *laddr;
4272
4273	/* first clear the flag */
4274	inp->ip_inp.inp.inp_vflag = 0;
4275	/* set the flag based on addresses on the ep list */
4276	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
4277		if (laddr->ifa == NULL) {
4278#ifdef SCTP_DEBUG
4279			if (sctp_debug_on & SCTP_DEBUG_PCB1) {
4280				printf("An ounce of prevention is worth a pound of cure\n");
4281			}
4282#endif				/* SCTP_DEBUG */
4283			continue;
4284		}
4285		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
4286			continue;
4287		}
4288		if (laddr->ifa->address.sa.sa_family == AF_INET6) {
4289			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
4290		} else if (laddr->ifa->address.sa.sa_family == AF_INET) {
4291			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
4292		}
4293	}
4294}
4295
4296/*
4297 * Add the address to the endpoint local address list There is nothing to be
4298 * done if we are bound to all addresses
4299 */
4300int
4301sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
4302{
4303	struct sctp_laddr *laddr;
4304	int fnd, error;
4305
4306	fnd = 0;
4307
4308	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
4309		/* You are already bound to all. You have it already */
4310		return (0);
4311	}
4312	if (ifa->address.sa.sa_family == AF_INET6) {
4313		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
4314			/* Can't bind a non-useable addr. */
4315			return (-1);
4316		}
4317	}
4318	/* first, is it already present? */
4319	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
4320		if (laddr->ifa == ifa) {
4321			fnd = 1;
4322			break;
4323		}
4324	}
4325
4326	if (fnd == 0) {
4327		/* Not in the ep list */
4328		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
4329		if (error != 0)
4330			return (error);
4331		inp->laddr_count++;
4332		/* update inp_vflag flags */
4333		if (ifa->address.sa.sa_family == AF_INET6) {
4334			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
4335		} else if (ifa->address.sa.sa_family == AF_INET) {
4336			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
4337		}
4338	}
4339	return (0);
4340}
4341
4342
4343/*
4344 * select a new (hopefully reachable) destination net (should only be used
4345 * when we deleted an ep addr that is the only usable source address to reach
4346 * the destination net)
4347 */
4348static void
4349sctp_select_primary_destination(struct sctp_tcb *stcb)
4350{
4351	struct sctp_nets *net;
4352
4353	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
4354		/* for now, we'll just pick the first reachable one we find */
4355		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
4356			continue;
4357		if (sctp_destination_is_reachable(stcb,
4358		    (struct sockaddr *)&net->ro._l_addr)) {
4359			/* found a reachable destination */
4360			stcb->asoc.primary_destination = net;
4361		}
4362	}
4363	/* I can't there from here! ...we're gonna die shortly... */
4364}
4365
4366
4367/*
4368 * Delete the address from the endpoint local address list There is nothing
4369 * to be done if we are bound to all addresses
4370 */
4371int
4372sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
4373{
4374	struct sctp_laddr *laddr;
4375	int fnd;
4376
4377	fnd = 0;
4378	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
4379		/* You are already bound to all. You have it already */
4380		return (EINVAL);
4381	}
4382	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
4383		if (laddr->ifa == ifa) {
4384			fnd = 1;
4385			break;
4386		}
4387	}
4388	if (fnd && (inp->laddr_count < 2)) {
4389		/* can't delete unless there are at LEAST 2 addresses */
4390		return (-1);
4391	}
4392	if (fnd) {
4393		/*
4394		 * clean up any use of this address go through our
4395		 * associations and clear any last_used_address that match
4396		 * this one for each assoc, see if a new primary_destination
4397		 * is needed
4398		 */
4399		struct sctp_tcb *stcb;
4400
4401		/* clean up "next_addr_touse" */
4402		if (inp->next_addr_touse == laddr)
4403			/* delete this address */
4404			inp->next_addr_touse = NULL;
4405
4406		/* clean up "last_used_address" */
4407		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
4408			struct sctp_nets *net;
4409
4410			SCTP_TCB_LOCK(stcb);
4411			if (stcb->asoc.last_used_address == laddr)
4412				/* delete this address */
4413				stcb->asoc.last_used_address = NULL;
4414			/*
4415			 * Now spin through all the nets and purge any ref
4416			 * to laddr
4417			 */
4418			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
4419				if (net->ro._s_addr &&
4420				    (net->ro._s_addr->ifa == laddr->ifa)) {
4421					/* Yep, purge src address selected */
4422					sctp_rtentry_t *rt;
4423
4424					/* delete this address if cached */
4425					rt = net->ro.ro_rt;
4426					if (rt != NULL) {
4427						RTFREE(rt);
4428						net->ro.ro_rt = NULL;
4429					}
4430					sctp_free_ifa(net->ro._s_addr);
4431					net->ro._s_addr = NULL;
4432					net->src_addr_selected = 0;
4433				}
4434			}
4435			SCTP_TCB_UNLOCK(stcb);
4436		}		/* for each tcb */
4437		/* remove it from the ep list */
4438		sctp_remove_laddr(laddr);
4439		inp->laddr_count--;
4440		/* update inp_vflag flags */
4441		sctp_update_ep_vflag(inp);
4442	}
4443	return (0);
4444}
4445
4446/*
4447 * Add the addr to the TCB local address list For the BOUNDALL or dynamic
4448 * case, this is a "pending" address list (eg. addresses waiting for an
4449 * ASCONF-ACK response) For the subset binding, static case, this is a
4450 * "valid" address list
4451 */
4452int
4453sctp_add_local_addr_assoc(struct sctp_tcb *stcb, struct sctp_ifa *ifa, int restricted_list)
4454{
4455	struct sctp_inpcb *inp;
4456	struct sctp_laddr *laddr;
4457	struct sctpladdr *list;
4458	int error;
4459
4460	/*
4461	 * Assumes TCB is locked.. and possibly the INP. May need to
4462	 * confirm/fix that if we need it and is not the case.
4463	 */
4464	list = &stcb->asoc.sctp_restricted_addrs;
4465
4466	inp = stcb->sctp_ep;
4467	if (ifa->address.sa.sa_family == AF_INET6) {
4468		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
4469			/* Can't bind a non-existent addr. */
4470			return (-1);
4471		}
4472	}
4473	/* does the address already exist? */
4474	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
4475		if (laddr->ifa == ifa) {
4476			return (-1);
4477		}
4478	}
4479
4480	/* add to the list */
4481	error = sctp_insert_laddr(list, ifa, 0);
4482	if (error != 0)
4483		return (error);
4484	return (0);
4485}
4486
4487/*
4488 * insert an laddr entry with the given ifa for the desired list
4489 */
4490int
4491sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
4492{
4493	struct sctp_laddr *laddr;
4494
4495	laddr = SCTP_ZONE_GET(sctppcbinfo.ipi_zone_laddr, struct sctp_laddr);
4496	if (laddr == NULL) {
4497		/* out of memory? */
4498		return (EINVAL);
4499	}
4500	SCTP_INCR_LADDR_COUNT();
4501	bzero(laddr, sizeof(*laddr));
4502	laddr->ifa = ifa;
4503	laddr->action = act;
4504	atomic_add_int(&ifa->refcount, 1);
4505	/* insert it */
4506	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
4507
4508	return (0);
4509}
4510
4511/*
4512 * Remove an laddr entry from the local address list (on an assoc)
4513 */
4514void
4515sctp_remove_laddr(struct sctp_laddr *laddr)
4516{
4517
4518	/* remove from the list */
4519	LIST_REMOVE(laddr, sctp_nxt_addr);
4520	sctp_free_ifa(laddr->ifa);
4521	SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_laddr, laddr);
4522	SCTP_DECR_LADDR_COUNT();
4523}
4524
4525/*
4526 * Remove an address from the TCB local address list
4527 */
4528int
4529sctp_del_local_addr_assoc(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
4530{
4531	struct sctp_inpcb *inp;
4532	struct sctp_laddr *laddr;
4533
4534	/*
4535	 * This is called by asconf work. It is assumed that a) The TCB is
4536	 * locked and b) The INP is locked. This is true in as much as I can
4537	 * trace through the entry asconf code where I did these locks.
4538	 * Again, the ASCONF code is a bit different in that it does lock
4539	 * the INP during its work often times. This must be since we don't
4540	 * want other proc's looking up things while what they are looking
4541	 * up is changing :-D
4542	 */
4543
4544	inp = stcb->sctp_ep;
4545	/* if subset bound and don't allow ASCONF's, can't delete last */
4546	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
4547	    (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF) == 0)) {
4548		if (stcb->asoc.numnets < 2) {
4549			/* can't delete last address */
4550			return (-1);
4551		}
4552	}
4553	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
4554		/* remove the address if it exists */
4555		if (laddr->ifa == NULL)
4556			continue;
4557		if (laddr->ifa == ifa) {
4558			sctp_remove_laddr(laddr);
4559			return (0);
4560		}
4561	}
4562
4563	/* address not found! */
4564	return (-1);
4565}
4566
4567static char sctp_pcb_initialized = 0;
4568
4569/*
4570 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
4571 */
4572/* sysctl */
4573static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
4574static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
4575
4576void
4577sctp_pcb_init()
4578{
4579	/*
4580	 * SCTP initialization for the PCB structures should be called by
4581	 * the sctp_init() funciton.
4582	 */
4583	int i;
4584
4585	if (sctp_pcb_initialized != 0) {
4586		/* error I was called twice */
4587		return;
4588	}
4589	sctp_pcb_initialized = 1;
4590
4591	bzero(&sctpstat, sizeof(struct sctpstat));
4592	(void)SCTP_GETTIME_TIMEVAL(&sctpstat.sctps_discontinuitytime);
4593	/* init the empty list of (All) Endpoints */
4594	LIST_INIT(&sctppcbinfo.listhead);
4595
4596	/* init the iterator head */
4597	TAILQ_INIT(&sctppcbinfo.iteratorhead);
4598
4599	/* init the hash table of endpoints */
4600	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &sctp_hashtblsize);
4601	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &sctp_pcbtblsize);
4602	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &sctp_chunkscale);
4603	sctppcbinfo.sctp_asochash = SCTP_HASH_INIT((sctp_hashtblsize * 31),
4604	    &sctppcbinfo.hashasocmark);
4605	sctppcbinfo.sctp_ephash = SCTP_HASH_INIT(sctp_hashtblsize,
4606	    &sctppcbinfo.hashmark);
4607	sctppcbinfo.sctp_tcpephash = SCTP_HASH_INIT(sctp_hashtblsize,
4608	    &sctppcbinfo.hashtcpmark);
4609	sctppcbinfo.hashtblsize = sctp_hashtblsize;
4610
4611	/* init the small hash table we use to track restarted asoc's */
4612	sctppcbinfo.sctp_restarthash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE,
4613	    &sctppcbinfo.hashrestartmark);
4614
4615
4616	sctppcbinfo.sctp_vrfhash = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
4617	    &sctppcbinfo.hashvrfmark);
4618
4619	/* init the zones */
4620	/*
4621	 * FIX ME: Should check for NULL returns, but if it does fail we are
4622	 * doomed to panic anyways... add later maybe.
4623	 */
4624	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_ep, "sctp_ep",
4625	    sizeof(struct sctp_inpcb), maxsockets);
4626
4627	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_asoc, "sctp_asoc",
4628	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
4629
4630	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_laddr, "sctp_laddr",
4631	    sizeof(struct sctp_laddr),
4632	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
4633
4634	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_net, "sctp_raddr",
4635	    sizeof(struct sctp_nets),
4636	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
4637
4638	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_chunk, "sctp_chunk",
4639	    sizeof(struct sctp_tmit_chunk),
4640	    (sctp_max_number_of_assoc * sctp_chunkscale));
4641
4642	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_readq, "sctp_readq",
4643	    sizeof(struct sctp_queued_to_read),
4644	    (sctp_max_number_of_assoc * sctp_chunkscale));
4645
4646	SCTP_ZONE_INIT(sctppcbinfo.ipi_zone_strmoq, "sctp_stream_msg_out",
4647	    sizeof(struct sctp_stream_queue_pending),
4648	    (sctp_max_number_of_assoc * sctp_chunkscale));
4649
4650	/* Master Lock INIT for info structure */
4651	SCTP_INP_INFO_LOCK_INIT();
4652	SCTP_STATLOG_INIT_LOCK();
4653	SCTP_ITERATOR_LOCK_INIT();
4654
4655	SCTP_IPI_COUNT_INIT();
4656	SCTP_IPI_ADDR_INIT();
4657	SCTP_IPI_ITERATOR_WQ_INIT();
4658
4659	LIST_INIT(&sctppcbinfo.addr_wq);
4660
4661	/* not sure if we need all the counts */
4662	sctppcbinfo.ipi_count_ep = 0;
4663	/* assoc/tcb zone info */
4664	sctppcbinfo.ipi_count_asoc = 0;
4665	/* local addrlist zone info */
4666	sctppcbinfo.ipi_count_laddr = 0;
4667	/* remote addrlist zone info */
4668	sctppcbinfo.ipi_count_raddr = 0;
4669	/* chunk info */
4670	sctppcbinfo.ipi_count_chunk = 0;
4671
4672	/* socket queue zone info */
4673	sctppcbinfo.ipi_count_readq = 0;
4674
4675	/* stream out queue cont */
4676	sctppcbinfo.ipi_count_strmoq = 0;
4677
4678	sctppcbinfo.ipi_free_strmoq = 0;
4679	sctppcbinfo.ipi_free_chunks = 0;
4680
4681	SCTP_OS_TIMER_INIT(&sctppcbinfo.addr_wq_timer.timer);
4682
4683	/* Init the TIMEWAIT list */
4684	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
4685		LIST_INIT(&sctppcbinfo.vtag_timewait[i]);
4686	}
4687
4688#if defined(SCTP_USE_THREAD_BASED_ITERATOR)
4689	sctppcbinfo.iterator_running = 0;
4690	sctp_startup_iterator();
4691#endif
4692
4693	/*
4694	 * INIT the default VRF which for BSD is the only one, other O/S's
4695	 * may have more. But initially they must start with one and then
4696	 * add the VRF's as addresses are added.
4697	 */
4698	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
4699
4700}
4701
4702
4703int
4704sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
4705    int iphlen, int offset, int limit, struct sctphdr *sh,
4706    struct sockaddr *altsa)
4707{
4708	/*
4709	 * grub through the INIT pulling addresses and loading them to the
4710	 * nets structure in the asoc. The from address in the mbuf should
4711	 * also be loaded (if it is not already). This routine can be called
4712	 * with either INIT or INIT-ACK's as long as the m points to the IP
4713	 * packet and the offset points to the beginning of the parameters.
4714	 */
4715	struct sctp_inpcb *inp, *l_inp;
4716	struct sctp_nets *net, *net_tmp;
4717	struct ip *iph;
4718	struct sctp_paramhdr *phdr, parm_buf;
4719	struct sctp_tcb *stcb_tmp;
4720	uint16_t ptype, plen;
4721	struct sockaddr *sa;
4722	struct sockaddr_storage dest_store;
4723	struct sockaddr *local_sa = (struct sockaddr *)&dest_store;
4724	struct sockaddr_in sin;
4725	struct sockaddr_in6 sin6;
4726	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
4727	struct sctp_auth_random *p_random = NULL;
4728	uint16_t random_len = 0;
4729	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
4730	struct sctp_auth_hmac_algo *hmacs = NULL;
4731	uint16_t hmacs_len = 0;
4732	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
4733	struct sctp_auth_chunk_list *chunks = NULL;
4734	uint16_t num_chunks = 0;
4735	sctp_key_t *new_key;
4736	uint32_t keylen;
4737	int got_random = 0, got_hmacs = 0, got_chklist = 0;
4738
4739	/* First get the destination address setup too. */
4740	memset(&sin, 0, sizeof(sin));
4741	memset(&sin6, 0, sizeof(sin6));
4742
4743	sin.sin_family = AF_INET;
4744	sin.sin_len = sizeof(sin);
4745	sin.sin_port = stcb->rport;
4746
4747	sin6.sin6_family = AF_INET6;
4748	sin6.sin6_len = sizeof(struct sockaddr_in6);
4749	sin6.sin6_port = stcb->rport;
4750	if (altsa == NULL) {
4751		iph = mtod(m, struct ip *);
4752		if (iph->ip_v == IPVERSION) {
4753			/* its IPv4 */
4754			struct sockaddr_in *sin_2;
4755
4756			sin_2 = (struct sockaddr_in *)(local_sa);
4757			memset(sin_2, 0, sizeof(sin));
4758			sin_2->sin_family = AF_INET;
4759			sin_2->sin_len = sizeof(sin);
4760			sin_2->sin_port = sh->dest_port;
4761			sin_2->sin_addr.s_addr = iph->ip_dst.s_addr;
4762			sin.sin_addr = iph->ip_src;
4763			sa = (struct sockaddr *)&sin;
4764		} else if (iph->ip_v == (IPV6_VERSION >> 4)) {
4765			/* its IPv6 */
4766			struct ip6_hdr *ip6;
4767			struct sockaddr_in6 *sin6_2;
4768
4769			ip6 = mtod(m, struct ip6_hdr *);
4770			sin6_2 = (struct sockaddr_in6 *)(local_sa);
4771			memset(sin6_2, 0, sizeof(sin6));
4772			sin6_2->sin6_family = AF_INET6;
4773			sin6_2->sin6_len = sizeof(struct sockaddr_in6);
4774			sin6_2->sin6_port = sh->dest_port;
4775			sin6.sin6_addr = ip6->ip6_src;
4776			sa = (struct sockaddr *)&sin6;
4777		} else {
4778			sa = NULL;
4779		}
4780	} else {
4781		/*
4782		 * For cookies we use the src address NOT from the packet
4783		 * but from the original INIT
4784		 */
4785		sa = altsa;
4786	}
4787	/* Turn off ECN until we get through all params */
4788	stcb->asoc.ecn_allowed = 0;
4789	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
4790		/* mark all addresses that we have currently on the list */
4791		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
4792	}
4793	/* does the source address already exist? if so skip it */
4794	l_inp = inp = stcb->sctp_ep;
4795
4796	atomic_add_int(&stcb->asoc.refcnt, 1);
4797	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, local_sa, stcb);
4798	atomic_add_int(&stcb->asoc.refcnt, -1);
4799
4800	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
4801		/* we must add the source address */
4802		/* no scope set here since we have a tcb already. */
4803		if ((sa->sa_family == AF_INET) &&
4804		    (stcb->asoc.ipv4_addr_legal)) {
4805			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
4806				return (-1);
4807			}
4808		} else if ((sa->sa_family == AF_INET6) &&
4809		    (stcb->asoc.ipv6_addr_legal)) {
4810			if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
4811				return (-2);
4812			}
4813		}
4814	} else {
4815		if (net_tmp != NULL && stcb_tmp == stcb) {
4816			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
4817		} else if (stcb_tmp != stcb) {
4818			/* It belongs to another association? */
4819			SCTP_TCB_UNLOCK(stcb_tmp);
4820			return (-3);
4821		}
4822	}
4823	if (stcb->asoc.state == 0) {
4824		/* the assoc was freed? */
4825		return (-4);
4826	}
4827	/* now we must go through each of the params. */
4828	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
4829	while (phdr) {
4830		ptype = ntohs(phdr->param_type);
4831		plen = ntohs(phdr->param_length);
4832		/*
4833		 * printf("ptype => %0x, plen => %d\n", (uint32_t)ptype,
4834		 * (int)plen);
4835		 */
4836		if (offset + plen > limit) {
4837			break;
4838		}
4839		if (plen == 0) {
4840			break;
4841		}
4842		if (ptype == SCTP_IPV4_ADDRESS) {
4843			if (stcb->asoc.ipv4_addr_legal) {
4844				struct sctp_ipv4addr_param *p4, p4_buf;
4845
4846				/* ok get the v4 address and check/add */
4847				phdr = sctp_get_next_param(m, offset,
4848				    (struct sctp_paramhdr *)&p4_buf, sizeof(p4_buf));
4849				if (plen != sizeof(struct sctp_ipv4addr_param) ||
4850				    phdr == NULL) {
4851					return (-5);
4852				}
4853				p4 = (struct sctp_ipv4addr_param *)phdr;
4854				sin.sin_addr.s_addr = p4->addr;
4855				if (IN_MULTICAST(sin.sin_addr.s_addr)) {
4856					/* Skip multi-cast addresses */
4857					goto next_param;
4858				}
4859				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
4860				    (sin.sin_addr.s_addr == INADDR_ANY)) {
4861					goto next_param;
4862				}
4863				sa = (struct sockaddr *)&sin;
4864				inp = stcb->sctp_ep;
4865				atomic_add_int(&stcb->asoc.refcnt, 1);
4866				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
4867				    local_sa, stcb);
4868				atomic_add_int(&stcb->asoc.refcnt, -1);
4869
4870				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
4871				    inp == NULL) {
4872					/* we must add the source address */
4873					/*
4874					 * no scope set since we have a tcb
4875					 * already
4876					 */
4877
4878					/*
4879					 * we must validate the state again
4880					 * here
4881					 */
4882					if (stcb->asoc.state == 0) {
4883						/* the assoc was freed? */
4884						return (-7);
4885					}
4886					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
4887						return (-8);
4888					}
4889				} else if (stcb_tmp == stcb) {
4890					if (stcb->asoc.state == 0) {
4891						/* the assoc was freed? */
4892						return (-10);
4893					}
4894					if (net != NULL) {
4895						/* clear flag */
4896						net->dest_state &=
4897						    ~SCTP_ADDR_NOT_IN_ASSOC;
4898					}
4899				} else {
4900					/*
4901					 * strange, address is in another
4902					 * assoc? straighten out locks.
4903					 */
4904					if (stcb->asoc.state == 0) {
4905						/* the assoc was freed? */
4906						return (-12);
4907					}
4908					return (-13);
4909				}
4910			}
4911		} else if (ptype == SCTP_IPV6_ADDRESS) {
4912			if (stcb->asoc.ipv6_addr_legal) {
4913				/* ok get the v6 address and check/add */
4914				struct sctp_ipv6addr_param *p6, p6_buf;
4915
4916				phdr = sctp_get_next_param(m, offset,
4917				    (struct sctp_paramhdr *)&p6_buf, sizeof(p6_buf));
4918				if (plen != sizeof(struct sctp_ipv6addr_param) ||
4919				    phdr == NULL) {
4920					return (-14);
4921				}
4922				p6 = (struct sctp_ipv6addr_param *)phdr;
4923				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
4924				    sizeof(p6->addr));
4925				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
4926					/* Skip multi-cast addresses */
4927					goto next_param;
4928				}
4929				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
4930					/*
4931					 * Link local make no sense without
4932					 * scope
4933					 */
4934					goto next_param;
4935				}
4936				sa = (struct sockaddr *)&sin6;
4937				inp = stcb->sctp_ep;
4938				atomic_add_int(&stcb->asoc.refcnt, 1);
4939				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
4940				    local_sa, stcb);
4941				atomic_add_int(&stcb->asoc.refcnt, -1);
4942				if (stcb_tmp == NULL && (inp == stcb->sctp_ep ||
4943				    inp == NULL)) {
4944					/*
4945					 * we must validate the state again
4946					 * here
4947					 */
4948					if (stcb->asoc.state == 0) {
4949						/* the assoc was freed? */
4950						return (-16);
4951					}
4952					/*
4953					 * we must add the address, no scope
4954					 * set
4955					 */
4956					if (sctp_add_remote_addr(stcb, sa, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
4957						return (-17);
4958					}
4959				} else if (stcb_tmp == stcb) {
4960					/*
4961					 * we must validate the state again
4962					 * here
4963					 */
4964					if (stcb->asoc.state == 0) {
4965						/* the assoc was freed? */
4966						return (-19);
4967					}
4968					if (net != NULL) {
4969						/* clear flag */
4970						net->dest_state &=
4971						    ~SCTP_ADDR_NOT_IN_ASSOC;
4972					}
4973				} else {
4974					/*
4975					 * strange, address is in another
4976					 * assoc? straighten out locks.
4977					 */
4978					if (stcb->asoc.state == 0) {
4979						/* the assoc was freed? */
4980						return (-21);
4981					}
4982					return (-22);
4983				}
4984			}
4985		} else if (ptype == SCTP_ECN_CAPABLE) {
4986			stcb->asoc.ecn_allowed = 1;
4987		} else if (ptype == SCTP_ULP_ADAPTATION) {
4988			if (stcb->asoc.state != SCTP_STATE_OPEN) {
4989				struct sctp_adaptation_layer_indication ai,
4990				                                *aip;
4991
4992				phdr = sctp_get_next_param(m, offset,
4993				    (struct sctp_paramhdr *)&ai, sizeof(ai));
4994				aip = (struct sctp_adaptation_layer_indication *)phdr;
4995				if (aip) {
4996					sctp_ulp_notify(SCTP_NOTIFY_ADAPTATION_INDICATION,
4997					    stcb, ntohl(aip->indication), NULL);
4998				}
4999			}
5000		} else if (ptype == SCTP_SET_PRIM_ADDR) {
5001			struct sctp_asconf_addr_param lstore, *fee;
5002			struct sctp_asconf_addrv4_param *fii;
5003			int lptype;
5004			struct sockaddr *lsa = NULL;
5005
5006			stcb->asoc.peer_supports_asconf = 1;
5007			if (plen > sizeof(lstore)) {
5008				return (-23);
5009			}
5010			phdr = sctp_get_next_param(m, offset,
5011			    (struct sctp_paramhdr *)&lstore, min(plen, sizeof(lstore)));
5012			if (phdr == NULL) {
5013				return (-24);
5014			}
5015			fee = (struct sctp_asconf_addr_param *)phdr;
5016			lptype = ntohs(fee->addrp.ph.param_type);
5017			if (lptype == SCTP_IPV4_ADDRESS) {
5018				if (plen !=
5019				    sizeof(struct sctp_asconf_addrv4_param)) {
5020					printf("Sizeof setprim in init/init ack not %d but %d - ignored\n",
5021					    (int)sizeof(struct sctp_asconf_addrv4_param),
5022					    plen);
5023				} else {
5024					fii = (struct sctp_asconf_addrv4_param *)fee;
5025					sin.sin_addr.s_addr = fii->addrp.addr;
5026					lsa = (struct sockaddr *)&sin;
5027				}
5028			} else if (lptype == SCTP_IPV6_ADDRESS) {
5029				if (plen !=
5030				    sizeof(struct sctp_asconf_addr_param)) {
5031					printf("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
5032					    (int)sizeof(struct sctp_asconf_addr_param),
5033					    plen);
5034				} else {
5035					memcpy(sin6.sin6_addr.s6_addr,
5036					    fee->addrp.addr,
5037					    sizeof(fee->addrp.addr));
5038					lsa = (struct sockaddr *)&sin6;
5039				}
5040			}
5041			if (lsa) {
5042				sctp_set_primary_addr(stcb, sa, NULL);
5043			}
5044		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
5045			/* Peer supports pr-sctp */
5046			stcb->asoc.peer_supports_prsctp = 1;
5047		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
5048			/* A supported extension chunk */
5049			struct sctp_supported_chunk_types_param *pr_supported;
5050			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
5051			int num_ent, i;
5052
5053			phdr = sctp_get_next_param(m, offset,
5054			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
5055			if (phdr == NULL) {
5056				return (-25);
5057			}
5058			stcb->asoc.peer_supports_asconf = 0;
5059			stcb->asoc.peer_supports_prsctp = 0;
5060			stcb->asoc.peer_supports_pktdrop = 0;
5061			stcb->asoc.peer_supports_strreset = 0;
5062			stcb->asoc.peer_supports_auth = 0;
5063			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
5064			num_ent = plen - sizeof(struct sctp_paramhdr);
5065			for (i = 0; i < num_ent; i++) {
5066				switch (pr_supported->chunk_types[i]) {
5067				case SCTP_ASCONF:
5068				case SCTP_ASCONF_ACK:
5069					stcb->asoc.peer_supports_asconf = 1;
5070					break;
5071				case SCTP_FORWARD_CUM_TSN:
5072					stcb->asoc.peer_supports_prsctp = 1;
5073					break;
5074				case SCTP_PACKET_DROPPED:
5075					stcb->asoc.peer_supports_pktdrop = 1;
5076					break;
5077				case SCTP_STREAM_RESET:
5078					stcb->asoc.peer_supports_strreset = 1;
5079					break;
5080				case SCTP_AUTHENTICATION:
5081					stcb->asoc.peer_supports_auth = 1;
5082					break;
5083				default:
5084					/* one I have not learned yet */
5085					break;
5086
5087				}
5088			}
5089		} else if (ptype == SCTP_ECN_NONCE_SUPPORTED) {
5090			/* Peer supports ECN-nonce */
5091			stcb->asoc.peer_supports_ecn_nonce = 1;
5092			stcb->asoc.ecn_nonce_allowed = 1;
5093		} else if (ptype == SCTP_RANDOM) {
5094			if (plen > sizeof(random_store))
5095				break;
5096			if (got_random) {
5097				/* already processed a RANDOM */
5098				goto next_param;
5099			}
5100			phdr = sctp_get_next_param(m, offset,
5101			    (struct sctp_paramhdr *)random_store,
5102			    min(sizeof(random_store), plen));
5103			if (phdr == NULL)
5104				return (-26);
5105			p_random = (struct sctp_auth_random *)phdr;
5106			random_len = plen - sizeof(*p_random);
5107			/* enforce the random length */
5108			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
5109#ifdef SCTP_DEBUG
5110				if (sctp_debug_on & SCTP_DEBUG_AUTH1)
5111					printf("SCTP: invalid RANDOM len\n");
5112#endif
5113				return (-27);
5114			}
5115			got_random = 1;
5116		} else if (ptype == SCTP_HMAC_LIST) {
5117			int num_hmacs;
5118			int i;
5119
5120			if (plen > sizeof(hmacs_store))
5121				break;
5122			if (got_hmacs) {
5123				/* already processed a HMAC list */
5124				goto next_param;
5125			}
5126			phdr = sctp_get_next_param(m, offset,
5127			    (struct sctp_paramhdr *)hmacs_store,
5128			    min(plen, sizeof(hmacs_store)));
5129			if (phdr == NULL)
5130				return (-28);
5131			hmacs = (struct sctp_auth_hmac_algo *)phdr;
5132			hmacs_len = plen - sizeof(*hmacs);
5133			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
5134			/* validate the hmac list */
5135			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
5136				return (-29);
5137			}
5138			if (stcb->asoc.peer_hmacs != NULL)
5139				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
5140			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
5141			if (stcb->asoc.peer_hmacs != NULL) {
5142				for (i = 0; i < num_hmacs; i++) {
5143					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
5144					    ntohs(hmacs->hmac_ids[i]));
5145				}
5146			}
5147			got_hmacs = 1;
5148		} else if (ptype == SCTP_CHUNK_LIST) {
5149			int i;
5150
5151			if (plen > sizeof(chunks_store))
5152				break;
5153			if (got_chklist) {
5154				/* already processed a Chunks list */
5155				goto next_param;
5156			}
5157			phdr = sctp_get_next_param(m, offset,
5158			    (struct sctp_paramhdr *)chunks_store,
5159			    min(plen, sizeof(chunks_store)));
5160			if (phdr == NULL)
5161				return (-30);
5162			chunks = (struct sctp_auth_chunk_list *)phdr;
5163			num_chunks = plen - sizeof(*chunks);
5164			if (stcb->asoc.peer_auth_chunks != NULL)
5165				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
5166			else
5167				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
5168			for (i = 0; i < num_chunks; i++) {
5169				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
5170				    stcb->asoc.peer_auth_chunks);
5171			}
5172			got_chklist = 1;
5173		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
5174			    (ptype == SCTP_STATE_COOKIE) ||
5175			    (ptype == SCTP_UNRECOG_PARAM) ||
5176			    (ptype == SCTP_COOKIE_PRESERVE) ||
5177			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
5178			    (ptype == SCTP_ADD_IP_ADDRESS) ||
5179			    (ptype == SCTP_DEL_IP_ADDRESS) ||
5180			    (ptype == SCTP_ERROR_CAUSE_IND) ||
5181		    (ptype == SCTP_SUCCESS_REPORT)) {
5182			 /* don't care */ ;
5183		} else {
5184			if ((ptype & 0x8000) == 0x0000) {
5185				/*
5186				 * must stop processing the rest of the
5187				 * param's. Any report bits were handled
5188				 * with the call to
5189				 * sctp_arethere_unrecognized_parameters()
5190				 * when the INIT or INIT-ACK was first seen.
5191				 */
5192				break;
5193			}
5194		}
5195next_param:
5196		offset += SCTP_SIZE32(plen);
5197		if (offset >= limit) {
5198			break;
5199		}
5200		phdr = sctp_get_next_param(m, offset, &parm_buf,
5201		    sizeof(parm_buf));
5202	}
5203	/* Now check to see if we need to purge any addresses */
5204	for (net = TAILQ_FIRST(&stcb->asoc.nets); net != NULL; net = net_tmp) {
5205		net_tmp = TAILQ_NEXT(net, sctp_next);
5206		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
5207		    SCTP_ADDR_NOT_IN_ASSOC) {
5208			/* This address has been removed from the asoc */
5209			/* remove and free it */
5210			stcb->asoc.numnets--;
5211			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
5212			sctp_free_remote_addr(net);
5213			if (net == stcb->asoc.primary_destination) {
5214				stcb->asoc.primary_destination = NULL;
5215				sctp_select_primary_destination(stcb);
5216			}
5217		}
5218	}
5219	/* validate authentication required parameters */
5220	if (got_random && got_hmacs) {
5221		stcb->asoc.peer_supports_auth = 1;
5222	} else {
5223		stcb->asoc.peer_supports_auth = 0;
5224	}
5225	if (!stcb->asoc.peer_supports_auth && got_chklist) {
5226		/* peer does not support auth but sent a chunks list? */
5227		return (-31);
5228	}
5229	if (!sctp_asconf_auth_nochk && stcb->asoc.peer_supports_asconf &&
5230	    !stcb->asoc.peer_supports_auth) {
5231		/* peer supports asconf but not auth? */
5232		return (-32);
5233	}
5234	/* concatenate the full random key */
5235#ifdef SCTP_AUTH_DRAFT_04
5236	keylen = random_len;
5237	new_key = sctp_alloc_key(keylen);
5238	if (new_key != NULL) {
5239		/* copy in the RANDOM */
5240		if (p_random != NULL)
5241			bcopy(p_random->random_data, new_key->key, random_len);
5242	}
5243#else
5244	keylen = sizeof(*p_random) + random_len + sizeof(*chunks) + num_chunks +
5245	    sizeof(*hmacs) + hmacs_len;
5246	new_key = sctp_alloc_key(keylen);
5247	if (new_key != NULL) {
5248		/* copy in the RANDOM */
5249		if (p_random != NULL) {
5250			keylen = sizeof(*p_random) + random_len;
5251			bcopy(p_random, new_key->key, keylen);
5252		}
5253		/* append in the AUTH chunks */
5254		if (chunks != NULL) {
5255			bcopy(chunks, new_key->key + keylen,
5256			    sizeof(*chunks) + num_chunks);
5257			keylen += sizeof(*chunks) + num_chunks;
5258		}
5259		/* append in the HMACs */
5260		if (hmacs != NULL) {
5261			bcopy(hmacs, new_key->key + keylen,
5262			    sizeof(*hmacs) + hmacs_len);
5263		}
5264	}
5265#endif
5266	else {
5267		/* failed to get memory for the key */
5268		return (-33);
5269	}
5270	if (stcb->asoc.authinfo.peer_random != NULL)
5271		sctp_free_key(stcb->asoc.authinfo.peer_random);
5272	stcb->asoc.authinfo.peer_random = new_key;
5273#ifdef SCTP_AUTH_DRAFT_04
5274	/* don't include the chunks and hmacs for draft -04 */
5275	stcb->asoc.authinfo.peer_random->keylen = random_len;
5276#endif
5277	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
5278	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
5279
5280	return (0);
5281}
5282
5283int
5284sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
5285    struct sctp_nets *net)
5286{
5287	/* make sure the requested primary address exists in the assoc */
5288	if (net == NULL && sa)
5289		net = sctp_findnet(stcb, sa);
5290
5291	if (net == NULL) {
5292		/* didn't find the requested primary address! */
5293		return (-1);
5294	} else {
5295		/* set the primary address */
5296		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
5297			/* Must be confirmed, so queue to set */
5298			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
5299			return (0);
5300		}
5301		stcb->asoc.primary_destination = net;
5302		net->dest_state &= ~SCTP_ADDR_WAS_PRIMARY;
5303		net = TAILQ_FIRST(&stcb->asoc.nets);
5304		if (net != stcb->asoc.primary_destination) {
5305			/*
5306			 * first one on the list is NOT the primary
5307			 * sctp_cmpaddr() is much more efficent if the
5308			 * primary is the first on the list, make it so.
5309			 */
5310			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
5311			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
5312		}
5313		return (0);
5314	}
5315}
5316
5317
5318int
5319sctp_is_vtag_good(struct sctp_inpcb *inp, uint32_t tag, struct timeval *now)
5320{
5321	/*
5322	 * This function serves two purposes. It will see if a TAG can be
5323	 * re-used and return 1 for yes it is ok and 0 for don't use that
5324	 * tag. A secondary function it will do is purge out old tags that
5325	 * can be removed.
5326	 */
5327	struct sctpasochead *head;
5328	struct sctpvtaghead *chain;
5329	struct sctp_tagblock *twait_block;
5330	struct sctp_tcb *stcb;
5331	int i;
5332
5333	SCTP_INP_INFO_WLOCK();
5334	chain = &sctppcbinfo.vtag_timewait[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
5335	/* First is the vtag in use ? */
5336
5337	head = &sctppcbinfo.sctp_asochash[SCTP_PCBHASH_ASOC(tag,
5338	    sctppcbinfo.hashasocmark)];
5339	if (head == NULL) {
5340		goto check_restart;
5341	}
5342	LIST_FOREACH(stcb, head, sctp_asocs) {
5343
5344		if (stcb->asoc.my_vtag == tag) {
5345			/*
5346			 * We should remove this if and return 0 always if
5347			 * we want vtags unique across all endpoints. For
5348			 * now within a endpoint is ok.
5349			 */
5350			if (inp == stcb->sctp_ep) {
5351				/* bad tag, in use */
5352				SCTP_INP_INFO_WUNLOCK();
5353				return (0);
5354			}
5355		}
5356	}
5357check_restart:
5358	/* Now lets check the restart hash */
5359	head = &sctppcbinfo.sctp_restarthash[SCTP_PCBHASH_ASOC(tag,
5360	    sctppcbinfo.hashrestartmark)];
5361	if (head == NULL) {
5362		goto check_time_wait;
5363	}
5364	LIST_FOREACH(stcb, head, sctp_tcbrestarhash) {
5365		if (stcb->asoc.assoc_id == tag) {
5366			/* candidate */
5367			if (inp == stcb->sctp_ep) {
5368				/* bad tag, in use */
5369				SCTP_INP_INFO_WUNLOCK();
5370				return (0);
5371			}
5372		}
5373	}
5374check_time_wait:
5375	/* Now what about timed wait ? */
5376	if (!SCTP_LIST_EMPTY(chain)) {
5377		/*
5378		 * Block(s) are present, lets see if we have this tag in the
5379		 * list
5380		 */
5381		LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5382			for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
5383				if (twait_block->vtag_block[i].v_tag == 0) {
5384					/* not used */
5385					continue;
5386				} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire >
5387				    now->tv_sec) {
5388					/* Audit expires this guy */
5389					twait_block->vtag_block[i].tv_sec_at_expire = 0;
5390					twait_block->vtag_block[i].v_tag = 0;
5391				} else if (twait_block->vtag_block[i].v_tag ==
5392				    tag) {
5393					/* Bad tag, sorry :< */
5394					SCTP_INP_INFO_WUNLOCK();
5395					return (0);
5396				}
5397			}
5398		}
5399	}
5400	/* Not found, ok to use the tag */
5401	SCTP_INP_INFO_WUNLOCK();
5402	return (1);
5403}
5404
5405
5406static sctp_assoc_t reneged_asoc_ids[256];
5407static uint8_t reneged_at = 0;
5408
5409
5410static void
5411sctp_drain_mbufs(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
5412{
5413	/*
5414	 * We must hunt this association for MBUF's past the cumack (i.e.
5415	 * out of order data that we can renege on).
5416	 */
5417	struct sctp_association *asoc;
5418	struct sctp_tmit_chunk *chk, *nchk;
5419	uint32_t cumulative_tsn_p1, tsn;
5420	struct sctp_queued_to_read *ctl, *nctl;
5421	int cnt, strmat, gap;
5422
5423	/* We look for anything larger than the cum-ack + 1 */
5424
5425	SCTP_STAT_INCR(sctps_protocol_drain_calls);
5426	if (sctp_do_drain == 0) {
5427		return;
5428	}
5429	asoc = &stcb->asoc;
5430	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
5431		/* none we can reneg on. */
5432		return;
5433	}
5434	SCTP_STAT_INCR(sctps_protocol_drains_done);
5435	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
5436	cnt = 0;
5437	/* First look in the re-assembly queue */
5438	chk = TAILQ_FIRST(&asoc->reasmqueue);
5439	while (chk) {
5440		/* Get the next one */
5441		nchk = TAILQ_NEXT(chk, sctp_next);
5442		if (compare_with_wrap(chk->rec.data.TSN_seq,
5443		    cumulative_tsn_p1, MAX_TSN)) {
5444			/* Yep it is above cum-ack */
5445			cnt++;
5446			tsn = chk->rec.data.TSN_seq;
5447			if (tsn >= asoc->mapping_array_base_tsn) {
5448				gap = tsn - asoc->mapping_array_base_tsn;
5449			} else {
5450				gap = (MAX_TSN - asoc->mapping_array_base_tsn) +
5451				    tsn + 1;
5452			}
5453			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
5454			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
5455			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
5456			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5457			if (chk->data) {
5458				sctp_m_freem(chk->data);
5459				chk->data = NULL;
5460			}
5461			sctp_free_remote_addr(chk->whoTo);
5462			sctp_free_a_chunk(stcb, chk);
5463		}
5464		chk = nchk;
5465	}
5466	/* Ok that was fun, now we will drain all the inbound streams? */
5467	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
5468		ctl = TAILQ_FIRST(&asoc->strmin[strmat].inqueue);
5469		while (ctl) {
5470			nctl = TAILQ_NEXT(ctl, next);
5471			if (compare_with_wrap(ctl->sinfo_tsn,
5472			    cumulative_tsn_p1, MAX_TSN)) {
5473				/* Yep it is above cum-ack */
5474				cnt++;
5475				tsn = ctl->sinfo_tsn;
5476				if (tsn >= asoc->mapping_array_base_tsn) {
5477					gap = tsn -
5478					    asoc->mapping_array_base_tsn;
5479				} else {
5480					gap = (MAX_TSN -
5481					    asoc->mapping_array_base_tsn) +
5482					    tsn + 1;
5483				}
5484				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
5485				sctp_ucount_decr(asoc->cnt_on_all_streams);
5486
5487				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array,
5488				    gap);
5489				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue,
5490				    ctl, next);
5491				if (ctl->data) {
5492					sctp_m_freem(ctl->data);
5493					ctl->data = NULL;
5494				}
5495				sctp_free_remote_addr(ctl->whoFrom);
5496				SCTP_ZONE_FREE(sctppcbinfo.ipi_zone_readq, ctl);
5497				SCTP_DECR_READQ_COUNT();
5498			}
5499			ctl = nctl;
5500		}
5501	}
5502	/*
5503	 * Question, should we go through the delivery queue? The only
5504	 * reason things are on here is the app not reading OR a p-d-api up.
5505	 * An attacker COULD send enough in to initiate the PD-API and then
5506	 * send a bunch of stuff to other streams... these would wind up on
5507	 * the delivery queue.. and then we would not get to them. But in
5508	 * order to do this I then have to back-track and un-deliver
5509	 * sequence numbers in streams.. el-yucko. I think for now we will
5510	 * NOT look at the delivery queue and leave it to be something to
5511	 * consider later. An alternative would be to abort the P-D-API with
5512	 * a notification and then deliver the data.... Or another method
5513	 * might be to keep track of how many times the situation occurs and
5514	 * if we see a possible attack underway just abort the association.
5515	 */
5516#ifdef SCTP_DEBUG
5517	if (sctp_debug_on & SCTP_DEBUG_PCB1) {
5518		if (cnt) {
5519			printf("Freed %d chunks from reneg harvest\n", cnt);
5520		}
5521	}
5522#endif				/* SCTP_DEBUG */
5523	if (cnt) {
5524		/*
5525		 * Now do we need to find a new
5526		 * asoc->highest_tsn_inside_map?
5527		 */
5528		if (asoc->highest_tsn_inside_map >= asoc->mapping_array_base_tsn) {
5529			gap = asoc->highest_tsn_inside_map - asoc->mapping_array_base_tsn;
5530		} else {
5531			gap = (MAX_TSN - asoc->mapping_array_base_tsn) +
5532			    asoc->highest_tsn_inside_map + 1;
5533		}
5534		if (gap >= (asoc->mapping_array_size << 3)) {
5535			/*
5536			 * Something bad happened or cum-ack and high were
5537			 * behind the base, but if so earlier checks should
5538			 * have found NO data... wierd... we will start at
5539			 * end of mapping array.
5540			 */
5541			printf("Gap was larger than array?? %d set to max:%d maparraymax:%x\n",
5542			    (int)gap,
5543			    (int)(asoc->mapping_array_size << 3),
5544			    (int)asoc->highest_tsn_inside_map);
5545			gap = asoc->mapping_array_size << 3;
5546		}
5547		while (gap > 0) {
5548			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
5549				/* found the new highest */
5550				asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn + gap;
5551				break;
5552			}
5553			gap--;
5554		}
5555		if (gap == 0) {
5556			/* Nothing left in map */
5557			memset(asoc->mapping_array, 0, asoc->mapping_array_size);
5558			asoc->mapping_array_base_tsn = asoc->cumulative_tsn + 1;
5559			asoc->highest_tsn_inside_map = asoc->cumulative_tsn;
5560		}
5561		asoc->last_revoke_count = cnt;
5562		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
5563		sctp_send_sack(stcb);
5564		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN);
5565		reneged_asoc_ids[reneged_at] = sctp_get_associd(stcb);
5566		reneged_at++;
5567	}
5568	/*
5569	 * Another issue, in un-setting the TSN's in the mapping array we
5570	 * DID NOT adjust the higest_tsn marker.  This will cause one of two
5571	 * things to occur. It may cause us to do extra work in checking for
5572	 * our mapping array movement. More importantly it may cause us to
5573	 * SACK every datagram. This may not be a bad thing though since we
5574	 * will recover once we get our cum-ack above and all this stuff we
5575	 * dumped recovered.
5576	 */
5577}
5578
5579void
5580sctp_drain()
5581{
5582	/*
5583	 * We must walk the PCB lists for ALL associations here. The system
5584	 * is LOW on MBUF's and needs help. This is where reneging will
5585	 * occur. We really hope this does NOT happen!
5586	 */
5587	struct sctp_inpcb *inp;
5588	struct sctp_tcb *stcb;
5589
5590	SCTP_INP_INFO_RLOCK();
5591	LIST_FOREACH(inp, &sctppcbinfo.listhead, sctp_list) {
5592		/* For each endpoint */
5593		SCTP_INP_RLOCK(inp);
5594		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5595			/* For each association */
5596			SCTP_TCB_LOCK(stcb);
5597			sctp_drain_mbufs(inp, stcb);
5598			SCTP_TCB_UNLOCK(stcb);
5599		}
5600		SCTP_INP_RUNLOCK(inp);
5601	}
5602	SCTP_INP_INFO_RUNLOCK();
5603}
5604
5605/*
5606 * start a new iterator
5607 * iterates through all endpoints and associations based on the pcb_state
5608 * flags and asoc_state.  "af" (mandatory) is executed for all matching
5609 * assocs and "ef" (optional) is executed when the iterator completes.
5610 * "inpf" (optional) is executed for each new endpoint as it is being
5611 * iterated through. inpe (optional) is called when the inp completes
5612 * its way through all the stcbs.
5613 */
5614int
5615sctp_initiate_iterator(inp_func inpf,
5616    asoc_func af,
5617    inp_func inpe,
5618    uint32_t pcb_state,
5619    uint32_t pcb_features,
5620    uint32_t asoc_state,
5621    void *argp,
5622    uint32_t argi,
5623    end_func ef,
5624    struct sctp_inpcb *s_inp,
5625    uint8_t chunk_output_off)
5626{
5627	struct sctp_iterator *it = NULL;
5628
5629	if (af == NULL) {
5630		return (-1);
5631	}
5632	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
5633	    "Iterator");
5634	if (it == NULL) {
5635		return (ENOMEM);
5636	}
5637	memset(it, 0, sizeof(*it));
5638	it->function_assoc = af;
5639	it->function_inp = inpf;
5640	if (inpf)
5641		it->done_current_ep = 0;
5642	else
5643		it->done_current_ep = 1;
5644	it->function_atend = ef;
5645	it->pointer = argp;
5646	it->val = argi;
5647	it->pcb_flags = pcb_state;
5648	it->pcb_features = pcb_features;
5649	it->asoc_state = asoc_state;
5650	it->function_inp_end = inpe;
5651	it->no_chunk_output = chunk_output_off;
5652	if (s_inp) {
5653		it->inp = s_inp;
5654		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
5655	} else {
5656		SCTP_INP_INFO_RLOCK();
5657		it->inp = LIST_FIRST(&sctppcbinfo.listhead);
5658
5659		SCTP_INP_INFO_RUNLOCK();
5660		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
5661
5662	}
5663	SCTP_IPI_ITERATOR_WQ_LOCK();
5664	if (it->inp)
5665		SCTP_INP_INCR_REF(it->inp);
5666	TAILQ_INSERT_TAIL(&sctppcbinfo.iteratorhead, it, sctp_nxt_itr);
5667#if defined(SCTP_USE_THREAD_BASED_ITERATOR)
5668	if (sctppcbinfo.iterator_running == 0) {
5669		sctp_wakeup_iterator();
5670	}
5671	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5672#else
5673	if (it->inp)
5674		SCTP_INP_DECR_REF(it->inp);
5675	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5676	/* Init the timer */
5677	SCTP_OS_TIMER_INIT(&it->tmr.timer);
5678	/* add to the list of all iterators */
5679	sctp_timer_start(SCTP_TIMER_TYPE_ITERATOR, (struct sctp_inpcb *)it,
5680	    NULL, NULL);
5681#endif
5682	return (0);
5683}
5684