sctp_pcb.c revision 267735
1/*-
2 * Copyright (c) 2001-2008, by Cisco Systems, Inc. All rights reserved.
3 * Copyright (c) 2008-2012, by Randall Stewart. All rights reserved.
4 * Copyright (c) 2008-2012, by Michael Tuexen. All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions are met:
8 *
9 * a) Redistributions of source code must retain the above copyright notice,
10 *    this list of conditions and the following disclaimer.
11 *
12 * b) Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in
14 *    the documentation and/or other materials provided with the distribution.
15 *
16 * c) Neither the name of Cisco Systems, Inc. nor the names of its
17 *    contributors may be used to endorse or promote products derived
18 *    from this software without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
22 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
24 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
30 * THE POSSIBILITY OF SUCH DAMAGE.
31 */
32
33#include <sys/cdefs.h>
34__FBSDID("$FreeBSD: stable/10/sys/netinet/sctp_pcb.c 267735 2014-06-22 16:50:46Z tuexen $");
35
36#include <netinet/sctp_os.h>
37#include <sys/proc.h>
38#include <netinet/sctp_var.h>
39#include <netinet/sctp_sysctl.h>
40#include <netinet/sctp_pcb.h>
41#include <netinet/sctputil.h>
42#include <netinet/sctp.h>
43#include <netinet/sctp_header.h>
44#include <netinet/sctp_asconf.h>
45#include <netinet/sctp_output.h>
46#include <netinet/sctp_timer.h>
47#include <netinet/sctp_bsd_addr.h>
48#include <netinet/sctp_dtrace_define.h>
49#include <netinet/udp.h>
50#ifdef INET6
51#include <netinet6/ip6_var.h>
52#endif
53#include <sys/sched.h>
54#include <sys/smp.h>
55#include <sys/unistd.h>
56
57
58VNET_DEFINE(struct sctp_base_info, system_base_info);
59
60/* FIX: we don't handle multiple link local scopes */
61/* "scopeless" replacement IN6_ARE_ADDR_EQUAL */
62#ifdef INET6
63int
64SCTP6_ARE_ADDR_EQUAL(struct sockaddr_in6 *a, struct sockaddr_in6 *b)
65{
66	struct sockaddr_in6 tmp_a, tmp_b;
67
68	memcpy(&tmp_a, a, sizeof(struct sockaddr_in6));
69	if (sa6_embedscope(&tmp_a, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
70		return (0);
71	}
72	memcpy(&tmp_b, b, sizeof(struct sockaddr_in6));
73	if (sa6_embedscope(&tmp_b, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
74		return (0);
75	}
76	return (IN6_ARE_ADDR_EQUAL(&tmp_a.sin6_addr, &tmp_b.sin6_addr));
77}
78
79#endif
80
81void
82sctp_fill_pcbinfo(struct sctp_pcbinfo *spcb)
83{
84	/*
85	 * We really don't need to lock this, but I will just because it
86	 * does not hurt.
87	 */
88	SCTP_INP_INFO_RLOCK();
89	spcb->ep_count = SCTP_BASE_INFO(ipi_count_ep);
90	spcb->asoc_count = SCTP_BASE_INFO(ipi_count_asoc);
91	spcb->laddr_count = SCTP_BASE_INFO(ipi_count_laddr);
92	spcb->raddr_count = SCTP_BASE_INFO(ipi_count_raddr);
93	spcb->chk_count = SCTP_BASE_INFO(ipi_count_chunk);
94	spcb->readq_count = SCTP_BASE_INFO(ipi_count_readq);
95	spcb->stream_oque = SCTP_BASE_INFO(ipi_count_strmoq);
96	spcb->free_chunks = SCTP_BASE_INFO(ipi_free_chunks);
97	SCTP_INP_INFO_RUNLOCK();
98}
99
100/*-
101 * Addresses are added to VRF's (Virtual Router's). For BSD we
102 * have only the default VRF 0. We maintain a hash list of
103 * VRF's. Each VRF has its own list of sctp_ifn's. Each of
104 * these has a list of addresses. When we add a new address
105 * to a VRF we lookup the ifn/ifn_index, if the ifn does
106 * not exist we create it and add it to the list of IFN's
107 * within the VRF. Once we have the sctp_ifn, we add the
108 * address to the list. So we look something like:
109 *
110 * hash-vrf-table
111 *   vrf-> ifn-> ifn -> ifn
112 *   vrf    |
113 *    ...   +--ifa-> ifa -> ifa
114 *   vrf
115 *
116 * We keep these separate lists since the SCTP subsystem will
117 * point to these from its source address selection nets structure.
118 * When an address is deleted it does not happen right away on
119 * the SCTP side, it gets scheduled. What we do when a
120 * delete happens is immediately remove the address from
121 * the master list and decrement the refcount. As our
122 * addip iterator works through and frees the src address
123 * selection pointing to the sctp_ifa, eventually the refcount
124 * will reach 0 and we will delete it. Note that it is assumed
125 * that any locking on system level ifn/ifa is done at the
126 * caller of these functions and these routines will only
127 * lock the SCTP structures as they add or delete things.
128 *
129 * Other notes on VRF concepts.
130 *  - An endpoint can be in multiple VRF's
131 *  - An association lives within a VRF and only one VRF.
132 *  - Any incoming packet we can deduce the VRF for by
133 *    looking at the mbuf/pak inbound (for BSD its VRF=0 :D)
134 *  - Any downward send call or connect call must supply the
135 *    VRF via ancillary data or via some sort of set default
136 *    VRF socket option call (again for BSD no brainer since
137 *    the VRF is always 0).
138 *  - An endpoint may add multiple VRF's to it.
139 *  - Listening sockets can accept associations in any
140 *    of the VRF's they are in but the assoc will end up
141 *    in only one VRF (gotten from the packet or connect/send).
142 *
143 */
144
145struct sctp_vrf *
146sctp_allocate_vrf(int vrf_id)
147{
148	struct sctp_vrf *vrf = NULL;
149	struct sctp_vrflist *bucket;
150
151	/* First allocate the VRF structure */
152	vrf = sctp_find_vrf(vrf_id);
153	if (vrf) {
154		/* Already allocated */
155		return (vrf);
156	}
157	SCTP_MALLOC(vrf, struct sctp_vrf *, sizeof(struct sctp_vrf),
158	    SCTP_M_VRF);
159	if (vrf == NULL) {
160		/* No memory */
161#ifdef INVARIANTS
162		panic("No memory for VRF:%d", vrf_id);
163#endif
164		return (NULL);
165	}
166	/* setup the VRF */
167	memset(vrf, 0, sizeof(struct sctp_vrf));
168	vrf->vrf_id = vrf_id;
169	LIST_INIT(&vrf->ifnlist);
170	vrf->total_ifa_count = 0;
171	vrf->refcount = 0;
172	/* now also setup table ids */
173	SCTP_INIT_VRF_TABLEID(vrf);
174	/* Init the HASH of addresses */
175	vrf->vrf_addr_hash = SCTP_HASH_INIT(SCTP_VRF_ADDR_HASH_SIZE,
176	    &vrf->vrf_addr_hashmark);
177	if (vrf->vrf_addr_hash == NULL) {
178		/* No memory */
179#ifdef INVARIANTS
180		panic("No memory for VRF:%d", vrf_id);
181#endif
182		SCTP_FREE(vrf, SCTP_M_VRF);
183		return (NULL);
184	}
185	/* Add it to the hash table */
186	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
187	LIST_INSERT_HEAD(bucket, vrf, next_vrf);
188	atomic_add_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
189	return (vrf);
190}
191
192
193struct sctp_ifn *
194sctp_find_ifn(void *ifn, uint32_t ifn_index)
195{
196	struct sctp_ifn *sctp_ifnp;
197	struct sctp_ifnlist *hash_ifn_head;
198
199	/*
200	 * We assume the lock is held for the addresses if that's wrong
201	 * problems could occur :-)
202	 */
203	hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
204	LIST_FOREACH(sctp_ifnp, hash_ifn_head, next_bucket) {
205		if (sctp_ifnp->ifn_index == ifn_index) {
206			return (sctp_ifnp);
207		}
208		if (sctp_ifnp->ifn_p && ifn && (sctp_ifnp->ifn_p == ifn)) {
209			return (sctp_ifnp);
210		}
211	}
212	return (NULL);
213}
214
215
216struct sctp_vrf *
217sctp_find_vrf(uint32_t vrf_id)
218{
219	struct sctp_vrflist *bucket;
220	struct sctp_vrf *liste;
221
222	bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(vrf_id & SCTP_BASE_INFO(hashvrfmark))];
223	LIST_FOREACH(liste, bucket, next_vrf) {
224		if (vrf_id == liste->vrf_id) {
225			return (liste);
226		}
227	}
228	return (NULL);
229}
230
231
232void
233sctp_free_vrf(struct sctp_vrf *vrf)
234{
235	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&vrf->refcount)) {
236		if (vrf->vrf_addr_hash) {
237			SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
238			vrf->vrf_addr_hash = NULL;
239		}
240		/* We zero'd the count */
241		LIST_REMOVE(vrf, next_vrf);
242		SCTP_FREE(vrf, SCTP_M_VRF);
243		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_vrfs), 1);
244	}
245}
246
247
248void
249sctp_free_ifn(struct sctp_ifn *sctp_ifnp)
250{
251	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifnp->refcount)) {
252		/* We zero'd the count */
253		if (sctp_ifnp->vrf) {
254			sctp_free_vrf(sctp_ifnp->vrf);
255		}
256		SCTP_FREE(sctp_ifnp, SCTP_M_IFN);
257		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
258	}
259}
260
261
262void
263sctp_update_ifn_mtu(uint32_t ifn_index, uint32_t mtu)
264{
265	struct sctp_ifn *sctp_ifnp;
266
267	sctp_ifnp = sctp_find_ifn((void *)NULL, ifn_index);
268	if (sctp_ifnp != NULL) {
269		sctp_ifnp->ifn_mtu = mtu;
270	}
271}
272
273
274void
275sctp_free_ifa(struct sctp_ifa *sctp_ifap)
276{
277	if (SCTP_DECREMENT_AND_CHECK_REFCOUNT(&sctp_ifap->refcount)) {
278		/* We zero'd the count */
279		if (sctp_ifap->ifn_p) {
280			sctp_free_ifn(sctp_ifap->ifn_p);
281		}
282		SCTP_FREE(sctp_ifap, SCTP_M_IFA);
283		atomic_subtract_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
284	}
285}
286
287
288static void
289sctp_delete_ifn(struct sctp_ifn *sctp_ifnp, int hold_addr_lock)
290{
291	struct sctp_ifn *found;
292
293	found = sctp_find_ifn(sctp_ifnp->ifn_p, sctp_ifnp->ifn_index);
294	if (found == NULL) {
295		/* Not in the list.. sorry */
296		return;
297	}
298	if (hold_addr_lock == 0)
299		SCTP_IPI_ADDR_WLOCK();
300	LIST_REMOVE(sctp_ifnp, next_bucket);
301	LIST_REMOVE(sctp_ifnp, next_ifn);
302	SCTP_DEREGISTER_INTERFACE(sctp_ifnp->ifn_index,
303	    sctp_ifnp->registered_af);
304	if (hold_addr_lock == 0)
305		SCTP_IPI_ADDR_WUNLOCK();
306	/* Take away the reference, and possibly free it */
307	sctp_free_ifn(sctp_ifnp);
308}
309
310
311void
312sctp_mark_ifa_addr_down(uint32_t vrf_id, struct sockaddr *addr,
313    const char *if_name, uint32_t ifn_index)
314{
315	struct sctp_vrf *vrf;
316	struct sctp_ifa *sctp_ifap;
317
318	SCTP_IPI_ADDR_RLOCK();
319	vrf = sctp_find_vrf(vrf_id);
320	if (vrf == NULL) {
321		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
322		goto out;
323
324	}
325	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
326	if (sctp_ifap == NULL) {
327		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
328		goto out;
329	}
330	if (sctp_ifap->ifn_p == NULL) {
331		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
332		goto out;
333	}
334	if (if_name) {
335		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
336			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
337			    sctp_ifap->ifn_p->ifn_name, if_name);
338			goto out;
339		}
340	} else {
341		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
342			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
343			    sctp_ifap->ifn_p->ifn_index, ifn_index);
344			goto out;
345		}
346	}
347
348	sctp_ifap->localifa_flags &= (~SCTP_ADDR_VALID);
349	sctp_ifap->localifa_flags |= SCTP_ADDR_IFA_UNUSEABLE;
350out:
351	SCTP_IPI_ADDR_RUNLOCK();
352}
353
354
355void
356sctp_mark_ifa_addr_up(uint32_t vrf_id, struct sockaddr *addr,
357    const char *if_name, uint32_t ifn_index)
358{
359	struct sctp_vrf *vrf;
360	struct sctp_ifa *sctp_ifap;
361
362	SCTP_IPI_ADDR_RLOCK();
363	vrf = sctp_find_vrf(vrf_id);
364	if (vrf == NULL) {
365		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
366		goto out;
367
368	}
369	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
370	if (sctp_ifap == NULL) {
371		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find sctp_ifap for address\n");
372		goto out;
373	}
374	if (sctp_ifap->ifn_p == NULL) {
375		SCTPDBG(SCTP_DEBUG_PCB4, "IFA has no IFN - can't mark unuseable\n");
376		goto out;
377	}
378	if (if_name) {
379		if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) != 0) {
380			SCTPDBG(SCTP_DEBUG_PCB4, "IFN %s of IFA not the same as %s\n",
381			    sctp_ifap->ifn_p->ifn_name, if_name);
382			goto out;
383		}
384	} else {
385		if (sctp_ifap->ifn_p->ifn_index != ifn_index) {
386			SCTPDBG(SCTP_DEBUG_PCB4, "IFA owned by ifn_index:%d down command for ifn_index:%d - ignored\n",
387			    sctp_ifap->ifn_p->ifn_index, ifn_index);
388			goto out;
389		}
390	}
391
392	sctp_ifap->localifa_flags &= (~SCTP_ADDR_IFA_UNUSEABLE);
393	sctp_ifap->localifa_flags |= SCTP_ADDR_VALID;
394out:
395	SCTP_IPI_ADDR_RUNLOCK();
396}
397
398
399/*-
400 * Add an ifa to an ifn.
401 * Register the interface as necessary.
402 * NOTE: ADDR write lock MUST be held.
403 */
404static void
405sctp_add_ifa_to_ifn(struct sctp_ifn *sctp_ifnp, struct sctp_ifa *sctp_ifap)
406{
407	int ifa_af;
408
409	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
410	sctp_ifap->ifn_p = sctp_ifnp;
411	atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
412	/* update address counts */
413	sctp_ifnp->ifa_count++;
414	ifa_af = sctp_ifap->address.sa.sa_family;
415	switch (ifa_af) {
416#ifdef INET
417	case AF_INET:
418		sctp_ifnp->num_v4++;
419		break;
420#endif
421#ifdef INET6
422	case AF_INET6:
423		sctp_ifnp->num_v6++;
424		break;
425#endif
426	default:
427		break;
428	}
429	if (sctp_ifnp->ifa_count == 1) {
430		/* register the new interface */
431		SCTP_REGISTER_INTERFACE(sctp_ifnp->ifn_index, ifa_af);
432		sctp_ifnp->registered_af = ifa_af;
433	}
434}
435
436
437/*-
438 * Remove an ifa from its ifn.
439 * If no more addresses exist, remove the ifn too. Otherwise, re-register
440 * the interface based on the remaining address families left.
441 * NOTE: ADDR write lock MUST be held.
442 */
443static void
444sctp_remove_ifa_from_ifn(struct sctp_ifa *sctp_ifap)
445{
446	LIST_REMOVE(sctp_ifap, next_ifa);
447	if (sctp_ifap->ifn_p) {
448		/* update address counts */
449		sctp_ifap->ifn_p->ifa_count--;
450		switch (sctp_ifap->address.sa.sa_family) {
451#ifdef INET
452		case AF_INET:
453			sctp_ifap->ifn_p->num_v4--;
454			break;
455#endif
456#ifdef INET6
457		case AF_INET6:
458			sctp_ifap->ifn_p->num_v6--;
459			break;
460#endif
461		default:
462			break;
463		}
464
465		if (LIST_EMPTY(&sctp_ifap->ifn_p->ifalist)) {
466			/* remove the ifn, possibly freeing it */
467			sctp_delete_ifn(sctp_ifap->ifn_p, SCTP_ADDR_LOCKED);
468		} else {
469			/* re-register address family type, if needed */
470			if ((sctp_ifap->ifn_p->num_v6 == 0) &&
471			    (sctp_ifap->ifn_p->registered_af == AF_INET6)) {
472				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
473				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
474				sctp_ifap->ifn_p->registered_af = AF_INET;
475			} else if ((sctp_ifap->ifn_p->num_v4 == 0) &&
476			    (sctp_ifap->ifn_p->registered_af == AF_INET)) {
477				SCTP_DEREGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET);
478				SCTP_REGISTER_INTERFACE(sctp_ifap->ifn_p->ifn_index, AF_INET6);
479				sctp_ifap->ifn_p->registered_af = AF_INET6;
480			}
481			/* free the ifn refcount */
482			sctp_free_ifn(sctp_ifap->ifn_p);
483		}
484		sctp_ifap->ifn_p = NULL;
485	}
486}
487
488
489struct sctp_ifa *
490sctp_add_addr_to_vrf(uint32_t vrf_id, void *ifn, uint32_t ifn_index,
491    uint32_t ifn_type, const char *if_name, void *ifa,
492    struct sockaddr *addr, uint32_t ifa_flags,
493    int dynamic_add)
494{
495	struct sctp_vrf *vrf;
496	struct sctp_ifn *sctp_ifnp = NULL;
497	struct sctp_ifa *sctp_ifap = NULL;
498	struct sctp_ifalist *hash_addr_head;
499	struct sctp_ifnlist *hash_ifn_head;
500	uint32_t hash_of_addr;
501	int new_ifn_af = 0;
502
503#ifdef SCTP_DEBUG
504	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: adding address: ", vrf_id);
505	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
506#endif
507	SCTP_IPI_ADDR_WLOCK();
508	sctp_ifnp = sctp_find_ifn(ifn, ifn_index);
509	if (sctp_ifnp) {
510		vrf = sctp_ifnp->vrf;
511	} else {
512		vrf = sctp_find_vrf(vrf_id);
513		if (vrf == NULL) {
514			vrf = sctp_allocate_vrf(vrf_id);
515			if (vrf == NULL) {
516				SCTP_IPI_ADDR_WUNLOCK();
517				return (NULL);
518			}
519		}
520	}
521	if (sctp_ifnp == NULL) {
522		/*
523		 * build one and add it, can't hold lock until after malloc
524		 * done though.
525		 */
526		SCTP_IPI_ADDR_WUNLOCK();
527		SCTP_MALLOC(sctp_ifnp, struct sctp_ifn *,
528		    sizeof(struct sctp_ifn), SCTP_M_IFN);
529		if (sctp_ifnp == NULL) {
530#ifdef INVARIANTS
531			panic("No memory for IFN");
532#endif
533			return (NULL);
534		}
535		memset(sctp_ifnp, 0, sizeof(struct sctp_ifn));
536		sctp_ifnp->ifn_index = ifn_index;
537		sctp_ifnp->ifn_p = ifn;
538		sctp_ifnp->ifn_type = ifn_type;
539		sctp_ifnp->refcount = 0;
540		sctp_ifnp->vrf = vrf;
541		atomic_add_int(&vrf->refcount, 1);
542		sctp_ifnp->ifn_mtu = SCTP_GATHER_MTU_FROM_IFN_INFO(ifn, ifn_index, addr->sa_family);
543		if (if_name != NULL) {
544			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", if_name);
545		} else {
546			snprintf(sctp_ifnp->ifn_name, SCTP_IFNAMSIZ, "%s", "unknown");
547		}
548		hash_ifn_head = &SCTP_BASE_INFO(vrf_ifn_hash)[(ifn_index & SCTP_BASE_INFO(vrf_ifn_hashmark))];
549		LIST_INIT(&sctp_ifnp->ifalist);
550		SCTP_IPI_ADDR_WLOCK();
551		LIST_INSERT_HEAD(hash_ifn_head, sctp_ifnp, next_bucket);
552		LIST_INSERT_HEAD(&vrf->ifnlist, sctp_ifnp, next_ifn);
553		atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifns), 1);
554		new_ifn_af = 1;
555	}
556	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
557	if (sctp_ifap) {
558		/* Hmm, it already exists? */
559		if ((sctp_ifap->ifn_p) &&
560		    (sctp_ifap->ifn_p->ifn_index == ifn_index)) {
561			SCTPDBG(SCTP_DEBUG_PCB4, "Using existing ifn %s (0x%x) for ifa %p\n",
562			    sctp_ifap->ifn_p->ifn_name, ifn_index,
563			    (void *)sctp_ifap);
564			if (new_ifn_af) {
565				/* Remove the created one that we don't want */
566				sctp_delete_ifn(sctp_ifnp, SCTP_ADDR_LOCKED);
567			}
568			if (sctp_ifap->localifa_flags & SCTP_BEING_DELETED) {
569				/* easy to solve, just switch back to active */
570				SCTPDBG(SCTP_DEBUG_PCB4, "Clearing deleted ifa flag\n");
571				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
572				sctp_ifap->ifn_p = sctp_ifnp;
573				atomic_add_int(&sctp_ifap->ifn_p->refcount, 1);
574			}
575	exit_stage_left:
576			SCTP_IPI_ADDR_WUNLOCK();
577			return (sctp_ifap);
578		} else {
579			if (sctp_ifap->ifn_p) {
580				/*
581				 * The last IFN gets the address, remove the
582				 * old one
583				 */
584				SCTPDBG(SCTP_DEBUG_PCB4, "Moving ifa %p from %s (0x%x) to %s (0x%x)\n",
585				    (void *)sctp_ifap, sctp_ifap->ifn_p->ifn_name,
586				    sctp_ifap->ifn_p->ifn_index, if_name,
587				    ifn_index);
588				/* remove the address from the old ifn */
589				sctp_remove_ifa_from_ifn(sctp_ifap);
590				/* move the address over to the new ifn */
591				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
592				goto exit_stage_left;
593			} else {
594				/* repair ifnp which was NULL ? */
595				sctp_ifap->localifa_flags = SCTP_ADDR_VALID;
596				SCTPDBG(SCTP_DEBUG_PCB4, "Repairing ifn %p for ifa %p\n",
597				    (void *)sctp_ifnp, (void *)sctp_ifap);
598				sctp_add_ifa_to_ifn(sctp_ifnp, sctp_ifap);
599			}
600			goto exit_stage_left;
601		}
602	}
603	SCTP_IPI_ADDR_WUNLOCK();
604	SCTP_MALLOC(sctp_ifap, struct sctp_ifa *, sizeof(struct sctp_ifa), SCTP_M_IFA);
605	if (sctp_ifap == NULL) {
606#ifdef INVARIANTS
607		panic("No memory for IFA");
608#endif
609		return (NULL);
610	}
611	memset(sctp_ifap, 0, sizeof(struct sctp_ifa));
612	sctp_ifap->ifn_p = sctp_ifnp;
613	atomic_add_int(&sctp_ifnp->refcount, 1);
614	sctp_ifap->vrf_id = vrf_id;
615	sctp_ifap->ifa = ifa;
616	memcpy(&sctp_ifap->address, addr, addr->sa_len);
617	sctp_ifap->localifa_flags = SCTP_ADDR_VALID | SCTP_ADDR_DEFER_USE;
618	sctp_ifap->flags = ifa_flags;
619	/* Set scope */
620	switch (sctp_ifap->address.sa.sa_family) {
621#ifdef INET
622	case AF_INET:
623		{
624			struct sockaddr_in *sin;
625
626			sin = (struct sockaddr_in *)&sctp_ifap->address.sin;
627			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
628			    (IN4_ISLOOPBACK_ADDRESS(&sin->sin_addr))) {
629				sctp_ifap->src_is_loop = 1;
630			}
631			if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr))) {
632				sctp_ifap->src_is_priv = 1;
633			}
634			sctp_ifnp->num_v4++;
635			if (new_ifn_af)
636				new_ifn_af = AF_INET;
637			break;
638		}
639#endif
640#ifdef INET6
641	case AF_INET6:
642		{
643			/* ok to use deprecated addresses? */
644			struct sockaddr_in6 *sin6;
645
646			sin6 = (struct sockaddr_in6 *)&sctp_ifap->address.sin6;
647			if (SCTP_IFN_IS_IFT_LOOP(sctp_ifap->ifn_p) ||
648			    (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr))) {
649				sctp_ifap->src_is_loop = 1;
650			}
651			if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
652				sctp_ifap->src_is_priv = 1;
653			}
654			sctp_ifnp->num_v6++;
655			if (new_ifn_af)
656				new_ifn_af = AF_INET6;
657			break;
658		}
659#endif
660	default:
661		new_ifn_af = 0;
662		break;
663	}
664	hash_of_addr = sctp_get_ifa_hash_val(&sctp_ifap->address.sa);
665
666	if ((sctp_ifap->src_is_priv == 0) &&
667	    (sctp_ifap->src_is_loop == 0)) {
668		sctp_ifap->src_is_glob = 1;
669	}
670	SCTP_IPI_ADDR_WLOCK();
671	hash_addr_head = &vrf->vrf_addr_hash[(hash_of_addr & vrf->vrf_addr_hashmark)];
672	LIST_INSERT_HEAD(hash_addr_head, sctp_ifap, next_bucket);
673	sctp_ifap->refcount = 1;
674	LIST_INSERT_HEAD(&sctp_ifnp->ifalist, sctp_ifap, next_ifa);
675	sctp_ifnp->ifa_count++;
676	vrf->total_ifa_count++;
677	atomic_add_int(&SCTP_BASE_INFO(ipi_count_ifas), 1);
678	if (new_ifn_af) {
679		SCTP_REGISTER_INTERFACE(ifn_index, new_ifn_af);
680		sctp_ifnp->registered_af = new_ifn_af;
681	}
682	SCTP_IPI_ADDR_WUNLOCK();
683	if (dynamic_add) {
684		/*
685		 * Bump up the refcount so that when the timer completes it
686		 * will drop back down.
687		 */
688		struct sctp_laddr *wi;
689
690		atomic_add_int(&sctp_ifap->refcount, 1);
691		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
692		if (wi == NULL) {
693			/*
694			 * Gak, what can we do? We have lost an address
695			 * change can you say HOSED?
696			 */
697			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
698			/* Opps, must decrement the count */
699			sctp_del_addr_from_vrf(vrf_id, addr, ifn_index,
700			    if_name);
701			return (NULL);
702		}
703		SCTP_INCR_LADDR_COUNT();
704		bzero(wi, sizeof(*wi));
705		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
706		wi->ifa = sctp_ifap;
707		wi->action = SCTP_ADD_IP_ADDRESS;
708
709		SCTP_WQ_ADDR_LOCK();
710		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
711		SCTP_WQ_ADDR_UNLOCK();
712
713		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
714		    (struct sctp_inpcb *)NULL,
715		    (struct sctp_tcb *)NULL,
716		    (struct sctp_nets *)NULL);
717	} else {
718		/* it's ready for use */
719		sctp_ifap->localifa_flags &= ~SCTP_ADDR_DEFER_USE;
720	}
721	return (sctp_ifap);
722}
723
724void
725sctp_del_addr_from_vrf(uint32_t vrf_id, struct sockaddr *addr,
726    uint32_t ifn_index, const char *if_name)
727{
728	struct sctp_vrf *vrf;
729	struct sctp_ifa *sctp_ifap = NULL;
730
731	SCTP_IPI_ADDR_WLOCK();
732	vrf = sctp_find_vrf(vrf_id);
733	if (vrf == NULL) {
734		SCTPDBG(SCTP_DEBUG_PCB4, "Can't find vrf_id 0x%x\n", vrf_id);
735		goto out_now;
736	}
737#ifdef SCTP_DEBUG
738	SCTPDBG(SCTP_DEBUG_PCB4, "vrf_id 0x%x: deleting address:", vrf_id);
739	SCTPDBG_ADDR(SCTP_DEBUG_PCB4, addr);
740#endif
741	sctp_ifap = sctp_find_ifa_by_addr(addr, vrf->vrf_id, SCTP_ADDR_LOCKED);
742	if (sctp_ifap) {
743		/* Validate the delete */
744		if (sctp_ifap->ifn_p) {
745			int valid = 0;
746
747			/*-
748			 * The name has priority over the ifn_index
749			 * if its given. We do this especially for
750			 * panda who might recycle indexes fast.
751			 */
752			if (if_name) {
753				if (strncmp(if_name, sctp_ifap->ifn_p->ifn_name, SCTP_IFNAMSIZ) == 0) {
754					/* They match its a correct delete */
755					valid = 1;
756				}
757			}
758			if (!valid) {
759				/* last ditch check ifn_index */
760				if (ifn_index == sctp_ifap->ifn_p->ifn_index) {
761					valid = 1;
762				}
763			}
764			if (!valid) {
765				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s does not match addresses\n",
766				    ifn_index, ((if_name == NULL) ? "NULL" : if_name));
767				SCTPDBG(SCTP_DEBUG_PCB4, "ifn:%d ifname:%s - ignoring delete\n",
768				    sctp_ifap->ifn_p->ifn_index, sctp_ifap->ifn_p->ifn_name);
769				SCTP_IPI_ADDR_WUNLOCK();
770				return;
771			}
772		}
773		SCTPDBG(SCTP_DEBUG_PCB4, "Deleting ifa %p\n", (void *)sctp_ifap);
774		sctp_ifap->localifa_flags &= SCTP_ADDR_VALID;
775		/*
776		 * We don't set the flag. This means that the structure will
777		 * hang around in EP's that have bound specific to it until
778		 * they close. This gives us TCP like behavior if someone
779		 * removes an address (or for that matter adds it right
780		 * back).
781		 */
782		/* sctp_ifap->localifa_flags |= SCTP_BEING_DELETED; */
783		vrf->total_ifa_count--;
784		LIST_REMOVE(sctp_ifap, next_bucket);
785		sctp_remove_ifa_from_ifn(sctp_ifap);
786	}
787#ifdef SCTP_DEBUG
788	else {
789		SCTPDBG(SCTP_DEBUG_PCB4, "Del Addr-ifn:%d Could not find address:",
790		    ifn_index);
791		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
792	}
793#endif
794
795out_now:
796	SCTP_IPI_ADDR_WUNLOCK();
797	if (sctp_ifap) {
798		struct sctp_laddr *wi;
799
800		wi = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
801		if (wi == NULL) {
802			/*
803			 * Gak, what can we do? We have lost an address
804			 * change can you say HOSED?
805			 */
806			SCTPDBG(SCTP_DEBUG_PCB4, "Lost an address change?\n");
807
808			/* Oops, must decrement the count */
809			sctp_free_ifa(sctp_ifap);
810			return;
811		}
812		SCTP_INCR_LADDR_COUNT();
813		bzero(wi, sizeof(*wi));
814		(void)SCTP_GETTIME_TIMEVAL(&wi->start_time);
815		wi->ifa = sctp_ifap;
816		wi->action = SCTP_DEL_IP_ADDRESS;
817		SCTP_WQ_ADDR_LOCK();
818		/*
819		 * Should this really be a tailq? As it is we will process
820		 * the newest first :-0
821		 */
822		LIST_INSERT_HEAD(&SCTP_BASE_INFO(addr_wq), wi, sctp_nxt_addr);
823		SCTP_WQ_ADDR_UNLOCK();
824
825		sctp_timer_start(SCTP_TIMER_TYPE_ADDR_WQ,
826		    (struct sctp_inpcb *)NULL,
827		    (struct sctp_tcb *)NULL,
828		    (struct sctp_nets *)NULL);
829	}
830	return;
831}
832
833
834static int
835sctp_does_stcb_own_this_addr(struct sctp_tcb *stcb, struct sockaddr *to)
836{
837	int loopback_scope;
838
839#if defined(INET)
840	int ipv4_local_scope, ipv4_addr_legal;
841
842#endif
843#if defined(INET6)
844	int local_scope, site_scope, ipv6_addr_legal;
845
846#endif
847	struct sctp_vrf *vrf;
848	struct sctp_ifn *sctp_ifn;
849	struct sctp_ifa *sctp_ifa;
850
851	loopback_scope = stcb->asoc.scope.loopback_scope;
852#if defined(INET)
853	ipv4_local_scope = stcb->asoc.scope.ipv4_local_scope;
854	ipv4_addr_legal = stcb->asoc.scope.ipv4_addr_legal;
855#endif
856#if defined(INET6)
857	local_scope = stcb->asoc.scope.local_scope;
858	site_scope = stcb->asoc.scope.site_scope;
859	ipv6_addr_legal = stcb->asoc.scope.ipv6_addr_legal;
860#endif
861
862	SCTP_IPI_ADDR_RLOCK();
863	vrf = sctp_find_vrf(stcb->asoc.vrf_id);
864	if (vrf == NULL) {
865		/* no vrf, no addresses */
866		SCTP_IPI_ADDR_RUNLOCK();
867		return (0);
868	}
869	if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
870		LIST_FOREACH(sctp_ifn, &vrf->ifnlist, next_ifn) {
871			if ((loopback_scope == 0) &&
872			    SCTP_IFN_IS_IFT_LOOP(sctp_ifn)) {
873				continue;
874			}
875			LIST_FOREACH(sctp_ifa, &sctp_ifn->ifalist, next_ifa) {
876				if (sctp_is_addr_restricted(stcb, sctp_ifa) &&
877				    (!sctp_is_addr_pending(stcb, sctp_ifa))) {
878					/*
879					 * We allow pending addresses, where
880					 * we have sent an asconf-add to be
881					 * considered valid.
882					 */
883					continue;
884				}
885				if (sctp_ifa->address.sa.sa_family != to->sa_family) {
886					continue;
887				}
888				switch (sctp_ifa->address.sa.sa_family) {
889#ifdef INET
890				case AF_INET:
891					if (ipv4_addr_legal) {
892						struct sockaddr_in *sin,
893						           *rsin;
894
895						sin = &sctp_ifa->address.sin;
896						rsin = (struct sockaddr_in *)to;
897						if ((ipv4_local_scope == 0) &&
898						    IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
899							continue;
900						}
901						if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
902							SCTP_IPI_ADDR_RUNLOCK();
903							return (1);
904						}
905					}
906					break;
907#endif
908#ifdef INET6
909				case AF_INET6:
910					if (ipv6_addr_legal) {
911						struct sockaddr_in6 *sin6,
912						            *rsin6;
913
914						sin6 = &sctp_ifa->address.sin6;
915						rsin6 = (struct sockaddr_in6 *)to;
916						if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
917							if (local_scope == 0)
918								continue;
919							if (sin6->sin6_scope_id == 0) {
920								if (sa6_recoverscope(sin6) != 0)
921									continue;
922							}
923						}
924						if ((site_scope == 0) &&
925						    (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))) {
926							continue;
927						}
928						if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
929							SCTP_IPI_ADDR_RUNLOCK();
930							return (1);
931						}
932					}
933					break;
934#endif
935				default:
936					/* TSNH */
937					break;
938				}
939			}
940		}
941	} else {
942		struct sctp_laddr *laddr;
943
944		LIST_FOREACH(laddr, &stcb->sctp_ep->sctp_addr_list, sctp_nxt_addr) {
945			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
946				SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
947				continue;
948			}
949			if (sctp_is_addr_restricted(stcb, laddr->ifa) &&
950			    (!sctp_is_addr_pending(stcb, laddr->ifa))) {
951				/*
952				 * We allow pending addresses, where we have
953				 * sent an asconf-add to be considered
954				 * valid.
955				 */
956				continue;
957			}
958			if (laddr->ifa->address.sa.sa_family != to->sa_family) {
959				continue;
960			}
961			switch (to->sa_family) {
962#ifdef INET
963			case AF_INET:
964				{
965					struct sockaddr_in *sin, *rsin;
966
967					sin = (struct sockaddr_in *)&laddr->ifa->address.sin;
968					rsin = (struct sockaddr_in *)to;
969					if (sin->sin_addr.s_addr == rsin->sin_addr.s_addr) {
970						SCTP_IPI_ADDR_RUNLOCK();
971						return (1);
972					}
973					break;
974				}
975#endif
976#ifdef INET6
977			case AF_INET6:
978				{
979					struct sockaddr_in6 *sin6, *rsin6;
980
981					sin6 = (struct sockaddr_in6 *)&laddr->ifa->address.sin6;
982					rsin6 = (struct sockaddr_in6 *)to;
983					if (SCTP6_ARE_ADDR_EQUAL(sin6, rsin6)) {
984						SCTP_IPI_ADDR_RUNLOCK();
985						return (1);
986					}
987					break;
988				}
989
990#endif
991			default:
992				/* TSNH */
993				break;
994			}
995
996		}
997	}
998	SCTP_IPI_ADDR_RUNLOCK();
999	return (0);
1000}
1001
1002
1003static struct sctp_tcb *
1004sctp_tcb_special_locate(struct sctp_inpcb **inp_p, struct sockaddr *from,
1005    struct sockaddr *to, struct sctp_nets **netp, uint32_t vrf_id)
1006{
1007	/**** ASSUMES THE CALLER holds the INP_INFO_RLOCK */
1008	/*
1009	 * If we support the TCP model, then we must now dig through to see
1010	 * if we can find our endpoint in the list of tcp ep's.
1011	 */
1012	uint16_t lport, rport;
1013	struct sctppcbhead *ephead;
1014	struct sctp_inpcb *inp;
1015	struct sctp_laddr *laddr;
1016	struct sctp_tcb *stcb;
1017	struct sctp_nets *net;
1018
1019	if ((to == NULL) || (from == NULL)) {
1020		return (NULL);
1021	}
1022	switch (to->sa_family) {
1023#ifdef INET
1024	case AF_INET:
1025		if (from->sa_family == AF_INET) {
1026			lport = ((struct sockaddr_in *)to)->sin_port;
1027			rport = ((struct sockaddr_in *)from)->sin_port;
1028		} else {
1029			return (NULL);
1030		}
1031		break;
1032#endif
1033#ifdef INET6
1034	case AF_INET6:
1035		if (from->sa_family == AF_INET6) {
1036			lport = ((struct sockaddr_in6 *)to)->sin6_port;
1037			rport = ((struct sockaddr_in6 *)from)->sin6_port;
1038		} else {
1039			return (NULL);
1040		}
1041		break;
1042#endif
1043	default:
1044		return (NULL);
1045	}
1046	ephead = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
1047	/*
1048	 * Ok now for each of the guys in this bucket we must look and see:
1049	 * - Does the remote port match. - Does there single association's
1050	 * addresses match this address (to). If so we update p_ep to point
1051	 * to this ep and return the tcb from it.
1052	 */
1053	LIST_FOREACH(inp, ephead, sctp_hash) {
1054		SCTP_INP_RLOCK(inp);
1055		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1056			SCTP_INP_RUNLOCK(inp);
1057			continue;
1058		}
1059		if (lport != inp->sctp_lport) {
1060			SCTP_INP_RUNLOCK(inp);
1061			continue;
1062		}
1063		if (inp->def_vrf_id != vrf_id) {
1064			SCTP_INP_RUNLOCK(inp);
1065			continue;
1066		}
1067		/* check to see if the ep has one of the addresses */
1068		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
1069			/* We are NOT bound all, so look further */
1070			int match = 0;
1071
1072			LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1073
1074				if (laddr->ifa == NULL) {
1075					SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n", __FUNCTION__);
1076					continue;
1077				}
1078				if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1079					SCTPDBG(SCTP_DEBUG_PCB1, "ifa being deleted\n");
1080					continue;
1081				}
1082				if (laddr->ifa->address.sa.sa_family ==
1083				    to->sa_family) {
1084					/* see if it matches */
1085#ifdef INET
1086					if (from->sa_family == AF_INET) {
1087						struct sockaddr_in *intf_addr,
1088						           *sin;
1089
1090						intf_addr = &laddr->ifa->address.sin;
1091						sin = (struct sockaddr_in *)to;
1092						if (sin->sin_addr.s_addr ==
1093						    intf_addr->sin_addr.s_addr) {
1094							match = 1;
1095							break;
1096						}
1097					}
1098#endif
1099#ifdef INET6
1100					if (from->sa_family == AF_INET6) {
1101						struct sockaddr_in6 *intf_addr6;
1102						struct sockaddr_in6 *sin6;
1103
1104						sin6 = (struct sockaddr_in6 *)
1105						    to;
1106						intf_addr6 = &laddr->ifa->address.sin6;
1107
1108						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1109						    intf_addr6)) {
1110							match = 1;
1111							break;
1112						}
1113					}
1114#endif
1115				}
1116			}
1117			if (match == 0) {
1118				/* This endpoint does not have this address */
1119				SCTP_INP_RUNLOCK(inp);
1120				continue;
1121			}
1122		}
1123		/*
1124		 * Ok if we hit here the ep has the address, does it hold
1125		 * the tcb?
1126		 */
1127		/* XXX: Why don't we TAILQ_FOREACH through sctp_asoc_list? */
1128		stcb = LIST_FIRST(&inp->sctp_asoc_list);
1129		if (stcb == NULL) {
1130			SCTP_INP_RUNLOCK(inp);
1131			continue;
1132		}
1133		SCTP_TCB_LOCK(stcb);
1134		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1135			SCTP_TCB_UNLOCK(stcb);
1136			SCTP_INP_RUNLOCK(inp);
1137			continue;
1138		}
1139		if (stcb->rport != rport) {
1140			/* remote port does not match. */
1141			SCTP_TCB_UNLOCK(stcb);
1142			SCTP_INP_RUNLOCK(inp);
1143			continue;
1144		}
1145		if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1146			SCTP_TCB_UNLOCK(stcb);
1147			SCTP_INP_RUNLOCK(inp);
1148			continue;
1149		}
1150		if (!sctp_does_stcb_own_this_addr(stcb, to)) {
1151			SCTP_TCB_UNLOCK(stcb);
1152			SCTP_INP_RUNLOCK(inp);
1153			continue;
1154		}
1155		/* Does this TCB have a matching address? */
1156		TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1157
1158			if (net->ro._l_addr.sa.sa_family != from->sa_family) {
1159				/* not the same family, can't be a match */
1160				continue;
1161			}
1162			switch (from->sa_family) {
1163#ifdef INET
1164			case AF_INET:
1165				{
1166					struct sockaddr_in *sin, *rsin;
1167
1168					sin = (struct sockaddr_in *)&net->ro._l_addr;
1169					rsin = (struct sockaddr_in *)from;
1170					if (sin->sin_addr.s_addr ==
1171					    rsin->sin_addr.s_addr) {
1172						/* found it */
1173						if (netp != NULL) {
1174							*netp = net;
1175						}
1176						/*
1177						 * Update the endpoint
1178						 * pointer
1179						 */
1180						*inp_p = inp;
1181						SCTP_INP_RUNLOCK(inp);
1182						return (stcb);
1183					}
1184					break;
1185				}
1186#endif
1187#ifdef INET6
1188			case AF_INET6:
1189				{
1190					struct sockaddr_in6 *sin6, *rsin6;
1191
1192					sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1193					rsin6 = (struct sockaddr_in6 *)from;
1194					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1195					    rsin6)) {
1196						/* found it */
1197						if (netp != NULL) {
1198							*netp = net;
1199						}
1200						/*
1201						 * Update the endpoint
1202						 * pointer
1203						 */
1204						*inp_p = inp;
1205						SCTP_INP_RUNLOCK(inp);
1206						return (stcb);
1207					}
1208					break;
1209				}
1210#endif
1211			default:
1212				/* TSNH */
1213				break;
1214			}
1215		}
1216		SCTP_TCB_UNLOCK(stcb);
1217		SCTP_INP_RUNLOCK(inp);
1218	}
1219	return (NULL);
1220}
1221
1222
1223/*
1224 * rules for use
1225 *
1226 * 1) If I return a NULL you must decrement any INP ref cnt. 2) If I find an
1227 * stcb, both will be locked (locked_tcb and stcb) but decrement will be done
1228 * (if locked == NULL). 3) Decrement happens on return ONLY if locked ==
1229 * NULL.
1230 */
1231
1232struct sctp_tcb *
1233sctp_findassociation_ep_addr(struct sctp_inpcb **inp_p, struct sockaddr *remote,
1234    struct sctp_nets **netp, struct sockaddr *local, struct sctp_tcb *locked_tcb)
1235{
1236	struct sctpasochead *head;
1237	struct sctp_inpcb *inp;
1238	struct sctp_tcb *stcb = NULL;
1239	struct sctp_nets *net;
1240	uint16_t rport;
1241
1242	inp = *inp_p;
1243	switch (remote->sa_family) {
1244#ifdef INET
1245	case AF_INET:
1246		rport = (((struct sockaddr_in *)remote)->sin_port);
1247		break;
1248#endif
1249#ifdef INET6
1250	case AF_INET6:
1251		rport = (((struct sockaddr_in6 *)remote)->sin6_port);
1252		break;
1253#endif
1254	default:
1255		return (NULL);
1256	}
1257	if (locked_tcb) {
1258		/*
1259		 * UN-lock so we can do proper locking here this occurs when
1260		 * called from load_addresses_from_init.
1261		 */
1262		atomic_add_int(&locked_tcb->asoc.refcnt, 1);
1263		SCTP_TCB_UNLOCK(locked_tcb);
1264	}
1265	SCTP_INP_INFO_RLOCK();
1266	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
1267	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
1268		/*-
1269		 * Now either this guy is our listener or it's the
1270		 * connector. If it is the one that issued the connect, then
1271		 * it's only chance is to be the first TCB in the list. If
1272		 * it is the acceptor, then do the special_lookup to hash
1273		 * and find the real inp.
1274		 */
1275		if ((inp->sctp_socket) && (inp->sctp_socket->so_qlimit)) {
1276			/* to is peer addr, from is my addr */
1277			stcb = sctp_tcb_special_locate(inp_p, remote, local,
1278			    netp, inp->def_vrf_id);
1279			if ((stcb != NULL) && (locked_tcb == NULL)) {
1280				/* we have a locked tcb, lower refcount */
1281				SCTP_INP_DECR_REF(inp);
1282			}
1283			if ((locked_tcb != NULL) && (locked_tcb != stcb)) {
1284				SCTP_INP_RLOCK(locked_tcb->sctp_ep);
1285				SCTP_TCB_LOCK(locked_tcb);
1286				atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1287				SCTP_INP_RUNLOCK(locked_tcb->sctp_ep);
1288			}
1289			SCTP_INP_INFO_RUNLOCK();
1290			return (stcb);
1291		} else {
1292			SCTP_INP_WLOCK(inp);
1293			if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1294				goto null_return;
1295			}
1296			stcb = LIST_FIRST(&inp->sctp_asoc_list);
1297			if (stcb == NULL) {
1298				goto null_return;
1299			}
1300			SCTP_TCB_LOCK(stcb);
1301
1302			if (stcb->rport != rport) {
1303				/* remote port does not match. */
1304				SCTP_TCB_UNLOCK(stcb);
1305				goto null_return;
1306			}
1307			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1308				SCTP_TCB_UNLOCK(stcb);
1309				goto null_return;
1310			}
1311			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1312				SCTP_TCB_UNLOCK(stcb);
1313				goto null_return;
1314			}
1315			/* now look at the list of remote addresses */
1316			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1317#ifdef INVARIANTS
1318				if (net == (TAILQ_NEXT(net, sctp_next))) {
1319					panic("Corrupt net list");
1320				}
1321#endif
1322				if (net->ro._l_addr.sa.sa_family !=
1323				    remote->sa_family) {
1324					/* not the same family */
1325					continue;
1326				}
1327				switch (remote->sa_family) {
1328#ifdef INET
1329				case AF_INET:
1330					{
1331						struct sockaddr_in *sin,
1332						           *rsin;
1333
1334						sin = (struct sockaddr_in *)
1335						    &net->ro._l_addr;
1336						rsin = (struct sockaddr_in *)remote;
1337						if (sin->sin_addr.s_addr ==
1338						    rsin->sin_addr.s_addr) {
1339							/* found it */
1340							if (netp != NULL) {
1341								*netp = net;
1342							}
1343							if (locked_tcb == NULL) {
1344								SCTP_INP_DECR_REF(inp);
1345							} else if (locked_tcb != stcb) {
1346								SCTP_TCB_LOCK(locked_tcb);
1347							}
1348							if (locked_tcb) {
1349								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1350							}
1351							SCTP_INP_WUNLOCK(inp);
1352							SCTP_INP_INFO_RUNLOCK();
1353							return (stcb);
1354						}
1355						break;
1356					}
1357#endif
1358#ifdef INET6
1359				case AF_INET6:
1360					{
1361						struct sockaddr_in6 *sin6,
1362						            *rsin6;
1363
1364						sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
1365						rsin6 = (struct sockaddr_in6 *)remote;
1366						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1367						    rsin6)) {
1368							/* found it */
1369							if (netp != NULL) {
1370								*netp = net;
1371							}
1372							if (locked_tcb == NULL) {
1373								SCTP_INP_DECR_REF(inp);
1374							} else if (locked_tcb != stcb) {
1375								SCTP_TCB_LOCK(locked_tcb);
1376							}
1377							if (locked_tcb) {
1378								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1379							}
1380							SCTP_INP_WUNLOCK(inp);
1381							SCTP_INP_INFO_RUNLOCK();
1382							return (stcb);
1383						}
1384						break;
1385					}
1386#endif
1387				default:
1388					/* TSNH */
1389					break;
1390				}
1391			}
1392			SCTP_TCB_UNLOCK(stcb);
1393		}
1394	} else {
1395		SCTP_INP_WLOCK(inp);
1396		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1397			goto null_return;
1398		}
1399		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(rport,
1400		    inp->sctp_hashmark)];
1401		if (head == NULL) {
1402			goto null_return;
1403		}
1404		LIST_FOREACH(stcb, head, sctp_tcbhash) {
1405			if (stcb->rport != rport) {
1406				/* remote port does not match */
1407				continue;
1408			}
1409			SCTP_TCB_LOCK(stcb);
1410			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1411				SCTP_TCB_UNLOCK(stcb);
1412				continue;
1413			}
1414			if (local && !sctp_does_stcb_own_this_addr(stcb, local)) {
1415				SCTP_TCB_UNLOCK(stcb);
1416				continue;
1417			}
1418			/* now look at the list of remote addresses */
1419			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
1420#ifdef INVARIANTS
1421				if (net == (TAILQ_NEXT(net, sctp_next))) {
1422					panic("Corrupt net list");
1423				}
1424#endif
1425				if (net->ro._l_addr.sa.sa_family !=
1426				    remote->sa_family) {
1427					/* not the same family */
1428					continue;
1429				}
1430				switch (remote->sa_family) {
1431#ifdef INET
1432				case AF_INET:
1433					{
1434						struct sockaddr_in *sin,
1435						           *rsin;
1436
1437						sin = (struct sockaddr_in *)
1438						    &net->ro._l_addr;
1439						rsin = (struct sockaddr_in *)remote;
1440						if (sin->sin_addr.s_addr ==
1441						    rsin->sin_addr.s_addr) {
1442							/* found it */
1443							if (netp != NULL) {
1444								*netp = net;
1445							}
1446							if (locked_tcb == NULL) {
1447								SCTP_INP_DECR_REF(inp);
1448							} else if (locked_tcb != stcb) {
1449								SCTP_TCB_LOCK(locked_tcb);
1450							}
1451							if (locked_tcb) {
1452								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1453							}
1454							SCTP_INP_WUNLOCK(inp);
1455							SCTP_INP_INFO_RUNLOCK();
1456							return (stcb);
1457						}
1458						break;
1459					}
1460#endif
1461#ifdef INET6
1462				case AF_INET6:
1463					{
1464						struct sockaddr_in6 *sin6,
1465						            *rsin6;
1466
1467						sin6 = (struct sockaddr_in6 *)
1468						    &net->ro._l_addr;
1469						rsin6 = (struct sockaddr_in6 *)remote;
1470						if (SCTP6_ARE_ADDR_EQUAL(sin6,
1471						    rsin6)) {
1472							/* found it */
1473							if (netp != NULL) {
1474								*netp = net;
1475							}
1476							if (locked_tcb == NULL) {
1477								SCTP_INP_DECR_REF(inp);
1478							} else if (locked_tcb != stcb) {
1479								SCTP_TCB_LOCK(locked_tcb);
1480							}
1481							if (locked_tcb) {
1482								atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1483							}
1484							SCTP_INP_WUNLOCK(inp);
1485							SCTP_INP_INFO_RUNLOCK();
1486							return (stcb);
1487						}
1488						break;
1489					}
1490#endif
1491				default:
1492					/* TSNH */
1493					break;
1494				}
1495			}
1496			SCTP_TCB_UNLOCK(stcb);
1497		}
1498	}
1499null_return:
1500	/* clean up for returning null */
1501	if (locked_tcb) {
1502		SCTP_TCB_LOCK(locked_tcb);
1503		atomic_subtract_int(&locked_tcb->asoc.refcnt, 1);
1504	}
1505	SCTP_INP_WUNLOCK(inp);
1506	SCTP_INP_INFO_RUNLOCK();
1507	/* not found */
1508	return (NULL);
1509}
1510
1511
1512/*
1513 * Find an association for a specific endpoint using the association id given
1514 * out in the COMM_UP notification
1515 */
1516struct sctp_tcb *
1517sctp_findasoc_ep_asocid_locked(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1518{
1519	/*
1520	 * Use my the assoc_id to find a endpoint
1521	 */
1522	struct sctpasochead *head;
1523	struct sctp_tcb *stcb;
1524	uint32_t id;
1525
1526	if (inp == NULL) {
1527		SCTP_PRINTF("TSNH ep_associd\n");
1528		return (NULL);
1529	}
1530	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1531		SCTP_PRINTF("TSNH ep_associd0\n");
1532		return (NULL);
1533	}
1534	id = (uint32_t) asoc_id;
1535	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
1536	if (head == NULL) {
1537		/* invalid id TSNH */
1538		SCTP_PRINTF("TSNH ep_associd1\n");
1539		return (NULL);
1540	}
1541	LIST_FOREACH(stcb, head, sctp_tcbasocidhash) {
1542		if (stcb->asoc.assoc_id == id) {
1543			if (inp != stcb->sctp_ep) {
1544				/*
1545				 * some other guy has the same id active (id
1546				 * collision ??).
1547				 */
1548				SCTP_PRINTF("TSNH ep_associd2\n");
1549				continue;
1550			}
1551			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
1552				continue;
1553			}
1554			if (want_lock) {
1555				SCTP_TCB_LOCK(stcb);
1556			}
1557			return (stcb);
1558		}
1559	}
1560	return (NULL);
1561}
1562
1563
1564struct sctp_tcb *
1565sctp_findassociation_ep_asocid(struct sctp_inpcb *inp, sctp_assoc_t asoc_id, int want_lock)
1566{
1567	struct sctp_tcb *stcb;
1568
1569	SCTP_INP_RLOCK(inp);
1570	stcb = sctp_findasoc_ep_asocid_locked(inp, asoc_id, want_lock);
1571	SCTP_INP_RUNLOCK(inp);
1572	return (stcb);
1573}
1574
1575
1576/*
1577 * Endpoint probe expects that the INP_INFO is locked.
1578 */
1579static struct sctp_inpcb *
1580sctp_endpoint_probe(struct sockaddr *nam, struct sctppcbhead *head,
1581    uint16_t lport, uint32_t vrf_id)
1582{
1583	struct sctp_inpcb *inp;
1584	struct sctp_laddr *laddr;
1585
1586#ifdef INET
1587	struct sockaddr_in *sin;
1588
1589#endif
1590#ifdef INET6
1591	struct sockaddr_in6 *sin6;
1592	struct sockaddr_in6 *intf_addr6;
1593
1594#endif
1595	int fnd;
1596
1597#ifdef INET
1598	sin = NULL;
1599#endif
1600#ifdef INET6
1601	sin6 = NULL;
1602#endif
1603	switch (nam->sa_family) {
1604#ifdef INET
1605	case AF_INET:
1606		sin = (struct sockaddr_in *)nam;
1607		break;
1608#endif
1609#ifdef INET6
1610	case AF_INET6:
1611		sin6 = (struct sockaddr_in6 *)nam;
1612		break;
1613#endif
1614	default:
1615		/* unsupported family */
1616		return (NULL);
1617	}
1618
1619	if (head == NULL)
1620		return (NULL);
1621
1622	LIST_FOREACH(inp, head, sctp_hash) {
1623		SCTP_INP_RLOCK(inp);
1624		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1625			SCTP_INP_RUNLOCK(inp);
1626			continue;
1627		}
1628		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) &&
1629		    (inp->sctp_lport == lport)) {
1630			/* got it */
1631#ifdef INET
1632			if ((nam->sa_family == AF_INET) &&
1633			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1634			    SCTP_IPV6_V6ONLY(inp)) {
1635				/* IPv4 on a IPv6 socket with ONLY IPv6 set */
1636				SCTP_INP_RUNLOCK(inp);
1637				continue;
1638			}
1639#endif
1640#ifdef INET6
1641			/* A V6 address and the endpoint is NOT bound V6 */
1642			if (nam->sa_family == AF_INET6 &&
1643			    (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) == 0) {
1644				SCTP_INP_RUNLOCK(inp);
1645				continue;
1646			}
1647#endif
1648			/* does a VRF id match? */
1649			fnd = 0;
1650			if (inp->def_vrf_id == vrf_id)
1651				fnd = 1;
1652
1653			SCTP_INP_RUNLOCK(inp);
1654			if (!fnd)
1655				continue;
1656			return (inp);
1657		}
1658		SCTP_INP_RUNLOCK(inp);
1659	}
1660	switch (nam->sa_family) {
1661#ifdef INET
1662	case AF_INET:
1663		if (sin->sin_addr.s_addr == INADDR_ANY) {
1664			/* Can't hunt for one that has no address specified */
1665			return (NULL);
1666		}
1667		break;
1668#endif
1669#ifdef INET6
1670	case AF_INET6:
1671		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1672			/* Can't hunt for one that has no address specified */
1673			return (NULL);
1674		}
1675		break;
1676#endif
1677	default:
1678		break;
1679	}
1680	/*
1681	 * ok, not bound to all so see if we can find a EP bound to this
1682	 * address.
1683	 */
1684	LIST_FOREACH(inp, head, sctp_hash) {
1685		SCTP_INP_RLOCK(inp);
1686		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1687			SCTP_INP_RUNLOCK(inp);
1688			continue;
1689		}
1690		if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL)) {
1691			SCTP_INP_RUNLOCK(inp);
1692			continue;
1693		}
1694		/*
1695		 * Ok this could be a likely candidate, look at all of its
1696		 * addresses
1697		 */
1698		if (inp->sctp_lport != lport) {
1699			SCTP_INP_RUNLOCK(inp);
1700			continue;
1701		}
1702		/* does a VRF id match? */
1703		fnd = 0;
1704		if (inp->def_vrf_id == vrf_id)
1705			fnd = 1;
1706
1707		if (!fnd) {
1708			SCTP_INP_RUNLOCK(inp);
1709			continue;
1710		}
1711		LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
1712			if (laddr->ifa == NULL) {
1713				SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
1714				    __FUNCTION__);
1715				continue;
1716			}
1717			SCTPDBG(SCTP_DEBUG_PCB1, "Ok laddr->ifa:%p is possible, ",
1718			    (void *)laddr->ifa);
1719			if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
1720				SCTPDBG(SCTP_DEBUG_PCB1, "Huh IFA being deleted\n");
1721				continue;
1722			}
1723			if (laddr->ifa->address.sa.sa_family == nam->sa_family) {
1724				/* possible, see if it matches */
1725				switch (nam->sa_family) {
1726#ifdef INET
1727				case AF_INET:
1728					if (sin->sin_addr.s_addr ==
1729					    laddr->ifa->address.sin.sin_addr.s_addr) {
1730						SCTP_INP_RUNLOCK(inp);
1731						return (inp);
1732					}
1733					break;
1734#endif
1735#ifdef INET6
1736				case AF_INET6:
1737					intf_addr6 = &laddr->ifa->address.sin6;
1738					if (SCTP6_ARE_ADDR_EQUAL(sin6,
1739					    intf_addr6)) {
1740						SCTP_INP_RUNLOCK(inp);
1741						return (inp);
1742					}
1743					break;
1744#endif
1745				}
1746			}
1747		}
1748		SCTP_INP_RUNLOCK(inp);
1749	}
1750	return (NULL);
1751}
1752
1753
1754static struct sctp_inpcb *
1755sctp_isport_inuse(struct sctp_inpcb *inp, uint16_t lport, uint32_t vrf_id)
1756{
1757	struct sctppcbhead *head;
1758	struct sctp_inpcb *t_inp;
1759	int fnd;
1760
1761	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1762	    SCTP_BASE_INFO(hashmark))];
1763	LIST_FOREACH(t_inp, head, sctp_hash) {
1764		if (t_inp->sctp_lport != lport) {
1765			continue;
1766		}
1767		/* is it in the VRF in question */
1768		fnd = 0;
1769		if (t_inp->def_vrf_id == vrf_id)
1770			fnd = 1;
1771		if (!fnd)
1772			continue;
1773
1774		/* This one is in use. */
1775		/* check the v6/v4 binding issue */
1776		if ((t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1777		    SCTP_IPV6_V6ONLY(t_inp)) {
1778			if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1779				/* collision in V6 space */
1780				return (t_inp);
1781			} else {
1782				/* inp is BOUND_V4 no conflict */
1783				continue;
1784			}
1785		} else if (t_inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) {
1786			/* t_inp is bound v4 and v6, conflict always */
1787			return (t_inp);
1788		} else {
1789			/* t_inp is bound only V4 */
1790			if ((inp->sctp_flags & SCTP_PCB_FLAGS_BOUND_V6) &&
1791			    SCTP_IPV6_V6ONLY(inp)) {
1792				/* no conflict */
1793				continue;
1794			}
1795			/* else fall through to conflict */
1796		}
1797		return (t_inp);
1798	}
1799	return (NULL);
1800}
1801
1802
1803int
1804sctp_swap_inpcb_for_listen(struct sctp_inpcb *inp)
1805{
1806	/* For 1-2-1 with port reuse */
1807	struct sctppcbhead *head;
1808	struct sctp_inpcb *tinp;
1809
1810	if (sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) {
1811		/* only works with port reuse on */
1812		return (-1);
1813	}
1814	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) == 0) {
1815		return (0);
1816	}
1817	SCTP_INP_RUNLOCK(inp);
1818	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport,
1819	    SCTP_BASE_INFO(hashmark))];
1820	/* Kick out all non-listeners to the TCP hash */
1821	LIST_FOREACH(tinp, head, sctp_hash) {
1822		if (tinp->sctp_lport != inp->sctp_lport) {
1823			continue;
1824		}
1825		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
1826			continue;
1827		}
1828		if (tinp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
1829			continue;
1830		}
1831		if (tinp->sctp_socket->so_qlimit) {
1832			continue;
1833		}
1834		SCTP_INP_WLOCK(tinp);
1835		LIST_REMOVE(tinp, sctp_hash);
1836		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(tinp->sctp_lport, SCTP_BASE_INFO(hashtcpmark))];
1837		tinp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
1838		LIST_INSERT_HEAD(head, tinp, sctp_hash);
1839		SCTP_INP_WUNLOCK(tinp);
1840	}
1841	SCTP_INP_WLOCK(inp);
1842	/* Pull from where he was */
1843	LIST_REMOVE(inp, sctp_hash);
1844	inp->sctp_flags &= ~SCTP_PCB_FLAGS_IN_TCPPOOL;
1845	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(inp->sctp_lport, SCTP_BASE_INFO(hashmark))];
1846	LIST_INSERT_HEAD(head, inp, sctp_hash);
1847	SCTP_INP_WUNLOCK(inp);
1848	SCTP_INP_RLOCK(inp);
1849	return (0);
1850}
1851
1852
1853struct sctp_inpcb *
1854sctp_pcb_findep(struct sockaddr *nam, int find_tcp_pool, int have_lock,
1855    uint32_t vrf_id)
1856{
1857	/*
1858	 * First we check the hash table to see if someone has this port
1859	 * bound with just the port.
1860	 */
1861	struct sctp_inpcb *inp;
1862	struct sctppcbhead *head;
1863	int lport;
1864	unsigned int i;
1865
1866#ifdef INET
1867	struct sockaddr_in *sin;
1868
1869#endif
1870#ifdef INET6
1871	struct sockaddr_in6 *sin6;
1872
1873#endif
1874
1875	switch (nam->sa_family) {
1876#ifdef INET
1877	case AF_INET:
1878		sin = (struct sockaddr_in *)nam;
1879		lport = sin->sin_port;
1880		break;
1881#endif
1882#ifdef INET6
1883	case AF_INET6:
1884		sin6 = (struct sockaddr_in6 *)nam;
1885		lport = sin6->sin6_port;
1886		break;
1887#endif
1888	default:
1889		return (NULL);
1890	}
1891	/*
1892	 * I could cheat here and just cast to one of the types but we will
1893	 * do it right. It also provides the check against an Unsupported
1894	 * type too.
1895	 */
1896	/* Find the head of the ALLADDR chain */
1897	if (have_lock == 0) {
1898		SCTP_INP_INFO_RLOCK();
1899	}
1900	head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport,
1901	    SCTP_BASE_INFO(hashmark))];
1902	inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1903
1904	/*
1905	 * If the TCP model exists it could be that the main listening
1906	 * endpoint is gone but there still exists a connected socket for
1907	 * this guy. If so we can return the first one that we find. This
1908	 * may NOT be the correct one so the caller should be wary on the
1909	 * returned INP. Currently the only caller that sets find_tcp_pool
1910	 * is in bindx where we are verifying that a user CAN bind the
1911	 * address. He either has bound it already, or someone else has, or
1912	 * its open to bind, so this is good enough.
1913	 */
1914	if (inp == NULL && find_tcp_pool) {
1915		for (i = 0; i < SCTP_BASE_INFO(hashtcpmark) + 1; i++) {
1916			head = &SCTP_BASE_INFO(sctp_tcpephash)[i];
1917			inp = sctp_endpoint_probe(nam, head, lport, vrf_id);
1918			if (inp) {
1919				break;
1920			}
1921		}
1922	}
1923	if (inp) {
1924		SCTP_INP_INCR_REF(inp);
1925	}
1926	if (have_lock == 0) {
1927		SCTP_INP_INFO_RUNLOCK();
1928	}
1929	return (inp);
1930}
1931
1932
1933/*
1934 * Find an association for an endpoint with the pointer to whom you want to
1935 * send to and the endpoint pointer. The address can be IPv4 or IPv6. We may
1936 * need to change the *to to some other struct like a mbuf...
1937 */
1938struct sctp_tcb *
1939sctp_findassociation_addr_sa(struct sockaddr *from, struct sockaddr *to,
1940    struct sctp_inpcb **inp_p, struct sctp_nets **netp, int find_tcp_pool,
1941    uint32_t vrf_id)
1942{
1943	struct sctp_inpcb *inp = NULL;
1944	struct sctp_tcb *stcb;
1945
1946	SCTP_INP_INFO_RLOCK();
1947	if (find_tcp_pool) {
1948		if (inp_p != NULL) {
1949			stcb = sctp_tcb_special_locate(inp_p, from, to, netp,
1950			    vrf_id);
1951		} else {
1952			stcb = sctp_tcb_special_locate(&inp, from, to, netp,
1953			    vrf_id);
1954		}
1955		if (stcb != NULL) {
1956			SCTP_INP_INFO_RUNLOCK();
1957			return (stcb);
1958		}
1959	}
1960	inp = sctp_pcb_findep(to, 0, 1, vrf_id);
1961	if (inp_p != NULL) {
1962		*inp_p = inp;
1963	}
1964	SCTP_INP_INFO_RUNLOCK();
1965	if (inp == NULL) {
1966		return (NULL);
1967	}
1968	/*
1969	 * ok, we have an endpoint, now lets find the assoc for it (if any)
1970	 * we now place the source address or from in the to of the find
1971	 * endpoint call. Since in reality this chain is used from the
1972	 * inbound packet side.
1973	 */
1974	if (inp_p != NULL) {
1975		stcb = sctp_findassociation_ep_addr(inp_p, from, netp, to,
1976		    NULL);
1977	} else {
1978		stcb = sctp_findassociation_ep_addr(&inp, from, netp, to,
1979		    NULL);
1980	}
1981	return (stcb);
1982}
1983
1984
1985/*
1986 * This routine will grub through the mbuf that is a INIT or INIT-ACK and
1987 * find all addresses that the sender has specified in any address list. Each
1988 * address will be used to lookup the TCB and see if one exits.
1989 */
1990static struct sctp_tcb *
1991sctp_findassociation_special_addr(struct mbuf *m, int offset,
1992    struct sctphdr *sh, struct sctp_inpcb **inp_p, struct sctp_nets **netp,
1993    struct sockaddr *dst)
1994{
1995	struct sctp_paramhdr *phdr, parm_buf;
1996
1997#if defined(INET) || defined(INET6)
1998	struct sctp_tcb *stcb;
1999	uint16_t ptype;
2000
2001#endif
2002	uint16_t plen;
2003
2004#ifdef INET
2005	struct sockaddr_in sin4;
2006
2007#endif
2008#ifdef INET6
2009	struct sockaddr_in6 sin6;
2010
2011#endif
2012
2013#ifdef INET
2014	memset(&sin4, 0, sizeof(sin4));
2015	sin4.sin_len = sizeof(sin4);
2016	sin4.sin_family = AF_INET;
2017	sin4.sin_port = sh->src_port;
2018#endif
2019#ifdef INET6
2020	memset(&sin6, 0, sizeof(sin6));
2021	sin6.sin6_len = sizeof(sin6);
2022	sin6.sin6_family = AF_INET6;
2023	sin6.sin6_port = sh->src_port;
2024#endif
2025
2026	offset += sizeof(struct sctp_init_chunk);
2027
2028	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
2029	while (phdr != NULL) {
2030		/* now we must see if we want the parameter */
2031#if defined(INET) || defined(INET6)
2032		ptype = ntohs(phdr->param_type);
2033#endif
2034		plen = ntohs(phdr->param_length);
2035		if (plen == 0) {
2036			break;
2037		}
2038#ifdef INET
2039		if (ptype == SCTP_IPV4_ADDRESS &&
2040		    plen == sizeof(struct sctp_ipv4addr_param)) {
2041			/* Get the rest of the address */
2042			struct sctp_ipv4addr_param ip4_parm, *p4;
2043
2044			phdr = sctp_get_next_param(m, offset,
2045			    (struct sctp_paramhdr *)&ip4_parm, min(plen, sizeof(ip4_parm)));
2046			if (phdr == NULL) {
2047				return (NULL);
2048			}
2049			p4 = (struct sctp_ipv4addr_param *)phdr;
2050			memcpy(&sin4.sin_addr, &p4->addr, sizeof(p4->addr));
2051			/* look it up */
2052			stcb = sctp_findassociation_ep_addr(inp_p,
2053			    (struct sockaddr *)&sin4, netp, dst, NULL);
2054			if (stcb != NULL) {
2055				return (stcb);
2056			}
2057		}
2058#endif
2059#ifdef INET6
2060		if (ptype == SCTP_IPV6_ADDRESS &&
2061		    plen == sizeof(struct sctp_ipv6addr_param)) {
2062			/* Get the rest of the address */
2063			struct sctp_ipv6addr_param ip6_parm, *p6;
2064
2065			phdr = sctp_get_next_param(m, offset,
2066			    (struct sctp_paramhdr *)&ip6_parm, min(plen, sizeof(ip6_parm)));
2067			if (phdr == NULL) {
2068				return (NULL);
2069			}
2070			p6 = (struct sctp_ipv6addr_param *)phdr;
2071			memcpy(&sin6.sin6_addr, &p6->addr, sizeof(p6->addr));
2072			/* look it up */
2073			stcb = sctp_findassociation_ep_addr(inp_p,
2074			    (struct sockaddr *)&sin6, netp, dst, NULL);
2075			if (stcb != NULL) {
2076				return (stcb);
2077			}
2078		}
2079#endif
2080		offset += SCTP_SIZE32(plen);
2081		phdr = sctp_get_next_param(m, offset, &parm_buf,
2082		    sizeof(parm_buf));
2083	}
2084	return (NULL);
2085}
2086
2087static struct sctp_tcb *
2088sctp_findassoc_by_vtag(struct sockaddr *from, struct sockaddr *to, uint32_t vtag,
2089    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint16_t rport,
2090    uint16_t lport, int skip_src_check, uint32_t vrf_id, uint32_t remote_tag)
2091{
2092	/*
2093	 * Use my vtag to hash. If we find it we then verify the source addr
2094	 * is in the assoc. If all goes well we save a bit on rec of a
2095	 * packet.
2096	 */
2097	struct sctpasochead *head;
2098	struct sctp_nets *net;
2099	struct sctp_tcb *stcb;
2100
2101	SCTP_INP_INFO_RLOCK();
2102	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(vtag,
2103	    SCTP_BASE_INFO(hashasocmark))];
2104	if (head == NULL) {
2105		/* invalid vtag */
2106		SCTP_INP_INFO_RUNLOCK();
2107		return (NULL);
2108	}
2109	LIST_FOREACH(stcb, head, sctp_asocs) {
2110		SCTP_INP_RLOCK(stcb->sctp_ep);
2111		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
2112			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2113			continue;
2114		}
2115		if (stcb->sctp_ep->def_vrf_id != vrf_id) {
2116			SCTP_INP_RUNLOCK(stcb->sctp_ep);
2117			continue;
2118		}
2119		SCTP_TCB_LOCK(stcb);
2120		SCTP_INP_RUNLOCK(stcb->sctp_ep);
2121		if (stcb->asoc.my_vtag == vtag) {
2122			/* candidate */
2123			if (stcb->rport != rport) {
2124				SCTP_TCB_UNLOCK(stcb);
2125				continue;
2126			}
2127			if (stcb->sctp_ep->sctp_lport != lport) {
2128				SCTP_TCB_UNLOCK(stcb);
2129				continue;
2130			}
2131			if (stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
2132				SCTP_TCB_UNLOCK(stcb);
2133				continue;
2134			}
2135			/* RRS:Need toaddr check here */
2136			if (sctp_does_stcb_own_this_addr(stcb, to) == 0) {
2137				/* Endpoint does not own this address */
2138				SCTP_TCB_UNLOCK(stcb);
2139				continue;
2140			}
2141			if (remote_tag) {
2142				/*
2143				 * If we have both vtags that's all we match
2144				 * on
2145				 */
2146				if (stcb->asoc.peer_vtag == remote_tag) {
2147					/*
2148					 * If both tags match we consider it
2149					 * conclusive and check NO
2150					 * source/destination addresses
2151					 */
2152					goto conclusive;
2153				}
2154			}
2155			if (skip_src_check) {
2156		conclusive:
2157				if (from) {
2158					*netp = sctp_findnet(stcb, from);
2159				} else {
2160					*netp = NULL;	/* unknown */
2161				}
2162				if (inp_p)
2163					*inp_p = stcb->sctp_ep;
2164				SCTP_INP_INFO_RUNLOCK();
2165				return (stcb);
2166			}
2167			net = sctp_findnet(stcb, from);
2168			if (net) {
2169				/* yep its him. */
2170				*netp = net;
2171				SCTP_STAT_INCR(sctps_vtagexpress);
2172				*inp_p = stcb->sctp_ep;
2173				SCTP_INP_INFO_RUNLOCK();
2174				return (stcb);
2175			} else {
2176				/*
2177				 * not him, this should only happen in rare
2178				 * cases so I peg it.
2179				 */
2180				SCTP_STAT_INCR(sctps_vtagbogus);
2181			}
2182		}
2183		SCTP_TCB_UNLOCK(stcb);
2184	}
2185	SCTP_INP_INFO_RUNLOCK();
2186	return (NULL);
2187}
2188
2189
2190/*
2191 * Find an association with the pointer to the inbound IP packet. This can be
2192 * a IPv4 or IPv6 packet.
2193 */
2194struct sctp_tcb *
2195sctp_findassociation_addr(struct mbuf *m, int offset,
2196    struct sockaddr *src, struct sockaddr *dst,
2197    struct sctphdr *sh, struct sctp_chunkhdr *ch,
2198    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2199{
2200	int find_tcp_pool;
2201	struct sctp_tcb *stcb;
2202	struct sctp_inpcb *inp;
2203
2204	if (sh->v_tag) {
2205		/* we only go down this path if vtag is non-zero */
2206		stcb = sctp_findassoc_by_vtag(src, dst, ntohl(sh->v_tag),
2207		    inp_p, netp, sh->src_port, sh->dest_port, 0, vrf_id, 0);
2208		if (stcb) {
2209			return (stcb);
2210		}
2211	}
2212	find_tcp_pool = 0;
2213	if ((ch->chunk_type != SCTP_INITIATION) &&
2214	    (ch->chunk_type != SCTP_INITIATION_ACK) &&
2215	    (ch->chunk_type != SCTP_COOKIE_ACK) &&
2216	    (ch->chunk_type != SCTP_COOKIE_ECHO)) {
2217		/* Other chunk types go to the tcp pool. */
2218		find_tcp_pool = 1;
2219	}
2220	if (inp_p) {
2221		stcb = sctp_findassociation_addr_sa(src, dst, inp_p, netp,
2222		    find_tcp_pool, vrf_id);
2223		inp = *inp_p;
2224	} else {
2225		stcb = sctp_findassociation_addr_sa(src, dst, &inp, netp,
2226		    find_tcp_pool, vrf_id);
2227	}
2228	SCTPDBG(SCTP_DEBUG_PCB1, "stcb:%p inp:%p\n", (void *)stcb, (void *)inp);
2229	if (stcb == NULL && inp) {
2230		/* Found a EP but not this address */
2231		if ((ch->chunk_type == SCTP_INITIATION) ||
2232		    (ch->chunk_type == SCTP_INITIATION_ACK)) {
2233			/*-
2234			 * special hook, we do NOT return linp or an
2235			 * association that is linked to an existing
2236			 * association that is under the TCP pool (i.e. no
2237			 * listener exists). The endpoint finding routine
2238			 * will always find a listener before examining the
2239			 * TCP pool.
2240			 */
2241			if (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) {
2242				if (inp_p) {
2243					*inp_p = NULL;
2244				}
2245				return (NULL);
2246			}
2247			stcb = sctp_findassociation_special_addr(m,
2248			    offset, sh, &inp, netp, dst);
2249			if (inp_p != NULL) {
2250				*inp_p = inp;
2251			}
2252		}
2253	}
2254	SCTPDBG(SCTP_DEBUG_PCB1, "stcb is %p\n", (void *)stcb);
2255	return (stcb);
2256}
2257
2258/*
2259 * lookup an association by an ASCONF lookup address.
2260 * if the lookup address is 0.0.0.0 or ::0, use the vtag to do the lookup
2261 */
2262struct sctp_tcb *
2263sctp_findassociation_ep_asconf(struct mbuf *m, int offset,
2264    struct sockaddr *dst, struct sctphdr *sh,
2265    struct sctp_inpcb **inp_p, struct sctp_nets **netp, uint32_t vrf_id)
2266{
2267	struct sctp_tcb *stcb;
2268	struct sockaddr_storage remote_store;
2269	struct sctp_paramhdr parm_buf, *phdr;
2270	int ptype;
2271	int zero_address = 0;
2272
2273#ifdef INET
2274	struct sockaddr_in *sin;
2275
2276#endif
2277#ifdef INET6
2278	struct sockaddr_in6 *sin6;
2279
2280#endif
2281
2282	memset(&remote_store, 0, sizeof(remote_store));
2283	phdr = sctp_get_next_param(m, offset + sizeof(struct sctp_asconf_chunk),
2284	    &parm_buf, sizeof(struct sctp_paramhdr));
2285	if (phdr == NULL) {
2286		SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf lookup addr\n",
2287		    __FUNCTION__);
2288		return NULL;
2289	}
2290	ptype = (int)((uint32_t) ntohs(phdr->param_type));
2291	/* get the correlation address */
2292	switch (ptype) {
2293#ifdef INET6
2294	case SCTP_IPV6_ADDRESS:
2295		{
2296			/* ipv6 address param */
2297			struct sctp_ipv6addr_param *p6, p6_buf;
2298
2299			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv6addr_param)) {
2300				return NULL;
2301			}
2302			p6 = (struct sctp_ipv6addr_param *)sctp_get_next_param(m,
2303			    offset + sizeof(struct sctp_asconf_chunk),
2304			    &p6_buf.ph, sizeof(*p6));
2305			if (p6 == NULL) {
2306				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v6 lookup addr\n",
2307				    __FUNCTION__);
2308				return (NULL);
2309			}
2310			sin6 = (struct sockaddr_in6 *)&remote_store;
2311			sin6->sin6_family = AF_INET6;
2312			sin6->sin6_len = sizeof(*sin6);
2313			sin6->sin6_port = sh->src_port;
2314			memcpy(&sin6->sin6_addr, &p6->addr, sizeof(struct in6_addr));
2315			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
2316				zero_address = 1;
2317			break;
2318		}
2319#endif
2320#ifdef INET
2321	case SCTP_IPV4_ADDRESS:
2322		{
2323			/* ipv4 address param */
2324			struct sctp_ipv4addr_param *p4, p4_buf;
2325
2326			if (ntohs(phdr->param_length) != sizeof(struct sctp_ipv4addr_param)) {
2327				return NULL;
2328			}
2329			p4 = (struct sctp_ipv4addr_param *)sctp_get_next_param(m,
2330			    offset + sizeof(struct sctp_asconf_chunk),
2331			    &p4_buf.ph, sizeof(*p4));
2332			if (p4 == NULL) {
2333				SCTPDBG(SCTP_DEBUG_INPUT3, "%s: failed to get asconf v4 lookup addr\n",
2334				    __FUNCTION__);
2335				return (NULL);
2336			}
2337			sin = (struct sockaddr_in *)&remote_store;
2338			sin->sin_family = AF_INET;
2339			sin->sin_len = sizeof(*sin);
2340			sin->sin_port = sh->src_port;
2341			memcpy(&sin->sin_addr, &p4->addr, sizeof(struct in_addr));
2342			if (sin->sin_addr.s_addr == INADDR_ANY)
2343				zero_address = 1;
2344			break;
2345		}
2346#endif
2347	default:
2348		/* invalid address param type */
2349		return NULL;
2350	}
2351
2352	if (zero_address) {
2353		stcb = sctp_findassoc_by_vtag(NULL, dst, ntohl(sh->v_tag), inp_p,
2354		    netp, sh->src_port, sh->dest_port, 1, vrf_id, 0);
2355		if (stcb != NULL) {
2356			SCTP_INP_DECR_REF(*inp_p);
2357		}
2358	} else {
2359		stcb = sctp_findassociation_ep_addr(inp_p,
2360		    (struct sockaddr *)&remote_store, netp,
2361		    dst, NULL);
2362	}
2363	return (stcb);
2364}
2365
2366
2367/*
2368 * allocate a sctp_inpcb and setup a temporary binding to a port/all
2369 * addresses. This way if we don't get a bind we by default pick a ephemeral
2370 * port with all addresses bound.
2371 */
2372int
2373sctp_inpcb_alloc(struct socket *so, uint32_t vrf_id)
2374{
2375	/*
2376	 * we get called when a new endpoint starts up. We need to allocate
2377	 * the sctp_inpcb structure from the zone and init it. Mark it as
2378	 * unbound and find a port that we can use as an ephemeral with
2379	 * INADDR_ANY. If the user binds later no problem we can then add in
2380	 * the specific addresses. And setup the default parameters for the
2381	 * EP.
2382	 */
2383	int i, error;
2384	struct sctp_inpcb *inp;
2385	struct sctp_pcb *m;
2386	struct timeval time;
2387	sctp_sharedkey_t *null_key;
2388
2389	error = 0;
2390
2391	SCTP_INP_INFO_WLOCK();
2392	inp = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_ep), struct sctp_inpcb);
2393	if (inp == NULL) {
2394		SCTP_PRINTF("Out of SCTP-INPCB structures - no resources\n");
2395		SCTP_INP_INFO_WUNLOCK();
2396		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2397		return (ENOBUFS);
2398	}
2399	/* zap it */
2400	bzero(inp, sizeof(*inp));
2401
2402	/* bump generations */
2403	/* setup socket pointers */
2404	inp->sctp_socket = so;
2405	inp->ip_inp.inp.inp_socket = so;
2406#ifdef INET6
2407	if (INP_SOCKAF(so) == AF_INET6) {
2408		if (MODULE_GLOBAL(ip6_auto_flowlabel)) {
2409			inp->ip_inp.inp.inp_flags |= IN6P_AUTOFLOWLABEL;
2410		}
2411		if (MODULE_GLOBAL(ip6_v6only)) {
2412			inp->ip_inp.inp.inp_flags |= IN6P_IPV6_V6ONLY;
2413		}
2414	}
2415#endif
2416	inp->sctp_associd_counter = 1;
2417	inp->partial_delivery_point = SCTP_SB_LIMIT_RCV(so) >> SCTP_PARTIAL_DELIVERY_SHIFT;
2418	inp->sctp_frag_point = SCTP_DEFAULT_MAXSEGMENT;
2419	inp->sctp_cmt_on_off = SCTP_BASE_SYSCTL(sctp_cmt_on_off);
2420	inp->sctp_ecn_enable = SCTP_BASE_SYSCTL(sctp_ecn_enable);
2421	/* init the small hash table we use to track asocid <-> tcb */
2422	inp->sctp_asocidhash = SCTP_HASH_INIT(SCTP_STACK_VTAG_HASH_SIZE, &inp->hashasocidmark);
2423	if (inp->sctp_asocidhash == NULL) {
2424		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2425		SCTP_INP_INFO_WUNLOCK();
2426		return (ENOBUFS);
2427	}
2428#ifdef IPSEC
2429	{
2430		struct inpcbpolicy *pcb_sp = NULL;
2431
2432		error = ipsec_init_policy(so, &pcb_sp);
2433		/* Arrange to share the policy */
2434		inp->ip_inp.inp.inp_sp = pcb_sp;
2435		((struct in6pcb *)(&inp->ip_inp.inp))->in6p_sp = pcb_sp;
2436	}
2437	if (error != 0) {
2438		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2439		SCTP_INP_INFO_WUNLOCK();
2440		return error;
2441	}
2442#endif				/* IPSEC */
2443	SCTP_INCR_EP_COUNT();
2444	inp->ip_inp.inp.inp_ip_ttl = MODULE_GLOBAL(ip_defttl);
2445	SCTP_INP_INFO_WUNLOCK();
2446
2447	so->so_pcb = (caddr_t)inp;
2448
2449	if (SCTP_SO_TYPE(so) == SOCK_SEQPACKET) {
2450		/* UDP style socket */
2451		inp->sctp_flags = (SCTP_PCB_FLAGS_UDPTYPE |
2452		    SCTP_PCB_FLAGS_UNBOUND);
2453		/* Be sure it is NON-BLOCKING IO for UDP */
2454		/* SCTP_SET_SO_NBIO(so); */
2455	} else if (SCTP_SO_TYPE(so) == SOCK_STREAM) {
2456		/* TCP style socket */
2457		inp->sctp_flags = (SCTP_PCB_FLAGS_TCPTYPE |
2458		    SCTP_PCB_FLAGS_UNBOUND);
2459		/* Be sure we have blocking IO by default */
2460		SCTP_CLEAR_SO_NBIO(so);
2461	} else {
2462		/*
2463		 * unsupported socket type (RAW, etc)- in case we missed it
2464		 * in protosw
2465		 */
2466		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EOPNOTSUPP);
2467		so->so_pcb = NULL;
2468		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2469		return (EOPNOTSUPP);
2470	}
2471	if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_1) {
2472		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2473		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2474	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_2) {
2475		sctp_feature_on(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2476		sctp_feature_on(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2477	} else if (SCTP_BASE_SYSCTL(sctp_default_frag_interleave) == SCTP_FRAG_LEVEL_0) {
2478		sctp_feature_off(inp, SCTP_PCB_FLAGS_FRAG_INTERLEAVE);
2479		sctp_feature_off(inp, SCTP_PCB_FLAGS_INTERLEAVE_STRMS);
2480	}
2481	inp->sctp_tcbhash = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_pcbtblsize),
2482	    &inp->sctp_hashmark);
2483	if (inp->sctp_tcbhash == NULL) {
2484		SCTP_PRINTF("Out of SCTP-INPCB->hashinit - no resources\n");
2485		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
2486		so->so_pcb = NULL;
2487		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
2488		return (ENOBUFS);
2489	}
2490	inp->def_vrf_id = vrf_id;
2491
2492	SCTP_INP_INFO_WLOCK();
2493	SCTP_INP_LOCK_INIT(inp);
2494	INP_LOCK_INIT(&inp->ip_inp.inp, "inp", "sctpinp");
2495	SCTP_INP_READ_INIT(inp);
2496	SCTP_ASOC_CREATE_LOCK_INIT(inp);
2497	/* lock the new ep */
2498	SCTP_INP_WLOCK(inp);
2499
2500	/* add it to the info area */
2501	LIST_INSERT_HEAD(&SCTP_BASE_INFO(listhead), inp, sctp_list);
2502	SCTP_INP_INFO_WUNLOCK();
2503
2504	TAILQ_INIT(&inp->read_queue);
2505	LIST_INIT(&inp->sctp_addr_list);
2506
2507	LIST_INIT(&inp->sctp_asoc_list);
2508
2509#ifdef SCTP_TRACK_FREED_ASOCS
2510	/* TEMP CODE */
2511	LIST_INIT(&inp->sctp_asoc_free_list);
2512#endif
2513	/* Init the timer structure for signature change */
2514	SCTP_OS_TIMER_INIT(&inp->sctp_ep.signature_change.timer);
2515	inp->sctp_ep.signature_change.type = SCTP_TIMER_TYPE_NEWCOOKIE;
2516
2517	/* now init the actual endpoint default data */
2518	m = &inp->sctp_ep;
2519
2520	/* setup the base timeout information */
2521	m->sctp_timeoutticks[SCTP_TIMER_SEND] = SEC_TO_TICKS(SCTP_SEND_SEC);	/* needed ? */
2522	m->sctp_timeoutticks[SCTP_TIMER_INIT] = SEC_TO_TICKS(SCTP_INIT_SEC);	/* needed ? */
2523	m->sctp_timeoutticks[SCTP_TIMER_RECV] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_delayed_sack_time_default));
2524	m->sctp_timeoutticks[SCTP_TIMER_HEARTBEAT] = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_heartbeat_interval_default));
2525	m->sctp_timeoutticks[SCTP_TIMER_PMTU] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_pmtu_raise_time_default));
2526	m->sctp_timeoutticks[SCTP_TIMER_MAXSHUTDOWN] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_shutdown_guard_time_default));
2527	m->sctp_timeoutticks[SCTP_TIMER_SIGNATURE] = SEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_secret_lifetime_default));
2528	/* all max/min max are in ms */
2529	m->sctp_maxrto = SCTP_BASE_SYSCTL(sctp_rto_max_default);
2530	m->sctp_minrto = SCTP_BASE_SYSCTL(sctp_rto_min_default);
2531	m->initial_rto = SCTP_BASE_SYSCTL(sctp_rto_initial_default);
2532	m->initial_init_rto_max = SCTP_BASE_SYSCTL(sctp_init_rto_max_default);
2533	m->sctp_sack_freq = SCTP_BASE_SYSCTL(sctp_sack_freq_default);
2534	m->max_init_times = SCTP_BASE_SYSCTL(sctp_init_rtx_max_default);
2535	m->max_send_times = SCTP_BASE_SYSCTL(sctp_assoc_rtx_max_default);
2536	m->def_net_failure = SCTP_BASE_SYSCTL(sctp_path_rtx_max_default);
2537	m->def_net_pf_threshold = SCTP_BASE_SYSCTL(sctp_path_pf_threshold);
2538	m->sctp_sws_sender = SCTP_SWS_SENDER_DEF;
2539	m->sctp_sws_receiver = SCTP_SWS_RECEIVER_DEF;
2540	m->max_burst = SCTP_BASE_SYSCTL(sctp_max_burst_default);
2541	m->fr_max_burst = SCTP_BASE_SYSCTL(sctp_fr_max_burst_default);
2542
2543	m->sctp_default_cc_module = SCTP_BASE_SYSCTL(sctp_default_cc_module);
2544	m->sctp_default_ss_module = SCTP_BASE_SYSCTL(sctp_default_ss_module);
2545	m->max_open_streams_intome = SCTP_BASE_SYSCTL(sctp_nr_incoming_streams_default);
2546	/* number of streams to pre-open on a association */
2547	m->pre_open_stream_count = SCTP_BASE_SYSCTL(sctp_nr_outgoing_streams_default);
2548
2549	/* Add adaptation cookie */
2550	m->adaptation_layer_indicator = 0;
2551	m->adaptation_layer_indicator_provided = 0;
2552
2553	/* seed random number generator */
2554	m->random_counter = 1;
2555	m->store_at = SCTP_SIGNATURE_SIZE;
2556	SCTP_READ_RANDOM(m->random_numbers, sizeof(m->random_numbers));
2557	sctp_fill_random_store(m);
2558
2559	/* Minimum cookie size */
2560	m->size_of_a_cookie = (sizeof(struct sctp_init_msg) * 2) +
2561	    sizeof(struct sctp_state_cookie);
2562	m->size_of_a_cookie += SCTP_SIGNATURE_SIZE;
2563
2564	/* Setup the initial secret */
2565	(void)SCTP_GETTIME_TIMEVAL(&time);
2566	m->time_of_secret_change = time.tv_sec;
2567
2568	for (i = 0; i < SCTP_NUMBER_OF_SECRETS; i++) {
2569		m->secret_key[0][i] = sctp_select_initial_TSN(m);
2570	}
2571	sctp_timer_start(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL);
2572
2573	/* How long is a cookie good for ? */
2574	m->def_cookie_life = MSEC_TO_TICKS(SCTP_BASE_SYSCTL(sctp_valid_cookie_life_default));
2575	/*
2576	 * Initialize authentication parameters
2577	 */
2578	m->local_hmacs = sctp_default_supported_hmaclist();
2579	m->local_auth_chunks = sctp_alloc_chunklist();
2580	m->default_dscp = 0;
2581#ifdef INET6
2582	m->default_flowlabel = 0;
2583#endif
2584	m->port = 0;		/* encapsulation disabled by default */
2585	sctp_auth_set_default_chunks(m->local_auth_chunks);
2586	LIST_INIT(&m->shared_keys);
2587	/* add default NULL key as key id 0 */
2588	null_key = sctp_alloc_sharedkey();
2589	sctp_insert_sharedkey(&m->shared_keys, null_key);
2590	SCTP_INP_WUNLOCK(inp);
2591#ifdef SCTP_LOG_CLOSING
2592	sctp_log_closing(inp, NULL, 12);
2593#endif
2594	return (error);
2595}
2596
2597
2598void
2599sctp_move_pcb_and_assoc(struct sctp_inpcb *old_inp, struct sctp_inpcb *new_inp,
2600    struct sctp_tcb *stcb)
2601{
2602	struct sctp_nets *net;
2603	uint16_t lport, rport;
2604	struct sctppcbhead *head;
2605	struct sctp_laddr *laddr, *oladdr;
2606
2607	atomic_add_int(&stcb->asoc.refcnt, 1);
2608	SCTP_TCB_UNLOCK(stcb);
2609	SCTP_INP_INFO_WLOCK();
2610	SCTP_INP_WLOCK(old_inp);
2611	SCTP_INP_WLOCK(new_inp);
2612	SCTP_TCB_LOCK(stcb);
2613	atomic_subtract_int(&stcb->asoc.refcnt, 1);
2614
2615	new_inp->sctp_ep.time_of_secret_change =
2616	    old_inp->sctp_ep.time_of_secret_change;
2617	memcpy(new_inp->sctp_ep.secret_key, old_inp->sctp_ep.secret_key,
2618	    sizeof(old_inp->sctp_ep.secret_key));
2619	new_inp->sctp_ep.current_secret_number =
2620	    old_inp->sctp_ep.current_secret_number;
2621	new_inp->sctp_ep.last_secret_number =
2622	    old_inp->sctp_ep.last_secret_number;
2623	new_inp->sctp_ep.size_of_a_cookie = old_inp->sctp_ep.size_of_a_cookie;
2624
2625	/* make it so new data pours into the new socket */
2626	stcb->sctp_socket = new_inp->sctp_socket;
2627	stcb->sctp_ep = new_inp;
2628
2629	/* Copy the port across */
2630	lport = new_inp->sctp_lport = old_inp->sctp_lport;
2631	rport = stcb->rport;
2632	/* Pull the tcb from the old association */
2633	LIST_REMOVE(stcb, sctp_tcbhash);
2634	LIST_REMOVE(stcb, sctp_tcblist);
2635	if (stcb->asoc.in_asocid_hash) {
2636		LIST_REMOVE(stcb, sctp_tcbasocidhash);
2637	}
2638	/* Now insert the new_inp into the TCP connected hash */
2639	head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR((lport | rport), SCTP_BASE_INFO(hashtcpmark))];
2640
2641	LIST_INSERT_HEAD(head, new_inp, sctp_hash);
2642	/* Its safe to access */
2643	new_inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
2644
2645	/* Now move the tcb into the endpoint list */
2646	LIST_INSERT_HEAD(&new_inp->sctp_asoc_list, stcb, sctp_tcblist);
2647	/*
2648	 * Question, do we even need to worry about the ep-hash since we
2649	 * only have one connection? Probably not :> so lets get rid of it
2650	 * and not suck up any kernel memory in that.
2651	 */
2652	if (stcb->asoc.in_asocid_hash) {
2653		struct sctpasochead *lhd;
2654
2655		lhd = &new_inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(stcb->asoc.assoc_id,
2656		    new_inp->hashasocidmark)];
2657		LIST_INSERT_HEAD(lhd, stcb, sctp_tcbasocidhash);
2658	}
2659	/* Ok. Let's restart timer. */
2660	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2661		sctp_timer_start(SCTP_TIMER_TYPE_PATHMTURAISE, new_inp,
2662		    stcb, net);
2663	}
2664
2665	SCTP_INP_INFO_WUNLOCK();
2666	if (new_inp->sctp_tcbhash != NULL) {
2667		SCTP_HASH_FREE(new_inp->sctp_tcbhash, new_inp->sctp_hashmark);
2668		new_inp->sctp_tcbhash = NULL;
2669	}
2670	if ((new_inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) {
2671		/* Subset bound, so copy in the laddr list from the old_inp */
2672		LIST_FOREACH(oladdr, &old_inp->sctp_addr_list, sctp_nxt_addr) {
2673			laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
2674			if (laddr == NULL) {
2675				/*
2676				 * Gak, what can we do? This assoc is really
2677				 * HOSED. We probably should send an abort
2678				 * here.
2679				 */
2680				SCTPDBG(SCTP_DEBUG_PCB1, "Association hosed in TCP model, out of laddr memory\n");
2681				continue;
2682			}
2683			SCTP_INCR_LADDR_COUNT();
2684			bzero(laddr, sizeof(*laddr));
2685			(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
2686			laddr->ifa = oladdr->ifa;
2687			atomic_add_int(&laddr->ifa->refcount, 1);
2688			LIST_INSERT_HEAD(&new_inp->sctp_addr_list, laddr,
2689			    sctp_nxt_addr);
2690			new_inp->laddr_count++;
2691			if (oladdr == stcb->asoc.last_used_address) {
2692				stcb->asoc.last_used_address = laddr;
2693			}
2694		}
2695	}
2696	/*
2697	 * Now any running timers need to be adjusted since we really don't
2698	 * care if they are running or not just blast in the new_inp into
2699	 * all of them.
2700	 */
2701
2702	stcb->asoc.dack_timer.ep = (void *)new_inp;
2703	stcb->asoc.asconf_timer.ep = (void *)new_inp;
2704	stcb->asoc.strreset_timer.ep = (void *)new_inp;
2705	stcb->asoc.shut_guard_timer.ep = (void *)new_inp;
2706	stcb->asoc.autoclose_timer.ep = (void *)new_inp;
2707	stcb->asoc.delayed_event_timer.ep = (void *)new_inp;
2708	stcb->asoc.delete_prim_timer.ep = (void *)new_inp;
2709	/* now what about the nets? */
2710	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
2711		net->pmtu_timer.ep = (void *)new_inp;
2712		net->hb_timer.ep = (void *)new_inp;
2713		net->rxt_timer.ep = (void *)new_inp;
2714	}
2715	SCTP_INP_WUNLOCK(new_inp);
2716	SCTP_INP_WUNLOCK(old_inp);
2717}
2718
2719
2720
2721
2722/* sctp_ifap is used to bypass normal local address validation checks */
2723int
2724sctp_inpcb_bind(struct socket *so, struct sockaddr *addr,
2725    struct sctp_ifa *sctp_ifap, struct thread *p)
2726{
2727	/* bind a ep to a socket address */
2728	struct sctppcbhead *head;
2729	struct sctp_inpcb *inp, *inp_tmp;
2730	struct inpcb *ip_inp;
2731	int port_reuse_active = 0;
2732	int bindall;
2733	uint16_t lport;
2734	int error;
2735	uint32_t vrf_id;
2736
2737	lport = 0;
2738	bindall = 1;
2739	inp = (struct sctp_inpcb *)so->so_pcb;
2740	ip_inp = (struct inpcb *)so->so_pcb;
2741#ifdef SCTP_DEBUG
2742	if (addr) {
2743		SCTPDBG(SCTP_DEBUG_PCB1, "Bind called port: %d\n",
2744		    ntohs(((struct sockaddr_in *)addr)->sin_port));
2745		SCTPDBG(SCTP_DEBUG_PCB1, "Addr: ");
2746		SCTPDBG_ADDR(SCTP_DEBUG_PCB1, addr);
2747	}
2748#endif
2749	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) == 0) {
2750		/* already did a bind, subsequent binds NOT allowed ! */
2751		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2752		return (EINVAL);
2753	}
2754#ifdef INVARIANTS
2755	if (p == NULL)
2756		panic("null proc/thread");
2757#endif
2758	if (addr != NULL) {
2759		switch (addr->sa_family) {
2760#ifdef INET
2761		case AF_INET:
2762			{
2763				struct sockaddr_in *sin;
2764
2765				/* IPV6_V6ONLY socket? */
2766				if (SCTP_IPV6_V6ONLY(ip_inp)) {
2767					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2768					return (EINVAL);
2769				}
2770				if (addr->sa_len != sizeof(*sin)) {
2771					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2772					return (EINVAL);
2773				}
2774				sin = (struct sockaddr_in *)addr;
2775				lport = sin->sin_port;
2776				/*
2777				 * For LOOPBACK the prison_local_ip4() call
2778				 * will transmute the ip address to the
2779				 * proper value.
2780				 */
2781				if (p && (error = prison_local_ip4(p->td_ucred, &sin->sin_addr)) != 0) {
2782					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2783					return (error);
2784				}
2785				if (sin->sin_addr.s_addr != INADDR_ANY) {
2786					bindall = 0;
2787				}
2788				break;
2789			}
2790#endif
2791#ifdef INET6
2792		case AF_INET6:
2793			{
2794				/*
2795				 * Only for pure IPv6 Address. (No IPv4
2796				 * Mapped!)
2797				 */
2798				struct sockaddr_in6 *sin6;
2799
2800				sin6 = (struct sockaddr_in6 *)addr;
2801
2802				if (addr->sa_len != sizeof(*sin6)) {
2803					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2804					return (EINVAL);
2805				}
2806				lport = sin6->sin6_port;
2807				/*
2808				 * For LOOPBACK the prison_local_ip6() call
2809				 * will transmute the ipv6 address to the
2810				 * proper value.
2811				 */
2812				if (p && (error = prison_local_ip6(p->td_ucred, &sin6->sin6_addr,
2813				    (SCTP_IPV6_V6ONLY(inp) != 0))) != 0) {
2814					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2815					return (error);
2816				}
2817				if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
2818					bindall = 0;
2819					/* KAME hack: embed scopeid */
2820					if (sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone)) != 0) {
2821						SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2822						return (EINVAL);
2823					}
2824				}
2825				/* this must be cleared for ifa_ifwithaddr() */
2826				sin6->sin6_scope_id = 0;
2827				break;
2828			}
2829#endif
2830		default:
2831			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EAFNOSUPPORT);
2832			return (EAFNOSUPPORT);
2833		}
2834	}
2835	SCTP_INP_INFO_WLOCK();
2836	SCTP_INP_WLOCK(inp);
2837	/* Setup a vrf_id to be the default for the non-bind-all case. */
2838	vrf_id = inp->def_vrf_id;
2839
2840	/* increase our count due to the unlock we do */
2841	SCTP_INP_INCR_REF(inp);
2842	if (lport) {
2843		/*
2844		 * Did the caller specify a port? if so we must see if an ep
2845		 * already has this one bound.
2846		 */
2847		/* got to be root to get at low ports */
2848		if (ntohs(lport) < IPPORT_RESERVED) {
2849			if (p && (error =
2850			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2851			    )) {
2852				SCTP_INP_DECR_REF(inp);
2853				SCTP_INP_WUNLOCK(inp);
2854				SCTP_INP_INFO_WUNLOCK();
2855				return (error);
2856			}
2857		}
2858		SCTP_INP_WUNLOCK(inp);
2859		if (bindall) {
2860			vrf_id = inp->def_vrf_id;
2861			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2862			if (inp_tmp != NULL) {
2863				/*
2864				 * lock guy returned and lower count note
2865				 * that we are not bound so inp_tmp should
2866				 * NEVER be inp. And it is this inp
2867				 * (inp_tmp) that gets the reference bump,
2868				 * so we must lower it.
2869				 */
2870				SCTP_INP_DECR_REF(inp_tmp);
2871				/* unlock info */
2872				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2873				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2874					/*
2875					 * Ok, must be one-2-one and
2876					 * allowing port re-use
2877					 */
2878					port_reuse_active = 1;
2879					goto continue_anyway;
2880				}
2881				SCTP_INP_DECR_REF(inp);
2882				SCTP_INP_INFO_WUNLOCK();
2883				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2884				return (EADDRINUSE);
2885			}
2886		} else {
2887			inp_tmp = sctp_pcb_findep(addr, 0, 1, vrf_id);
2888			if (inp_tmp != NULL) {
2889				/*
2890				 * lock guy returned and lower count note
2891				 * that we are not bound so inp_tmp should
2892				 * NEVER be inp. And it is this inp
2893				 * (inp_tmp) that gets the reference bump,
2894				 * so we must lower it.
2895				 */
2896				SCTP_INP_DECR_REF(inp_tmp);
2897				/* unlock info */
2898				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2899				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2900					/*
2901					 * Ok, must be one-2-one and
2902					 * allowing port re-use
2903					 */
2904					port_reuse_active = 1;
2905					goto continue_anyway;
2906				}
2907				SCTP_INP_DECR_REF(inp);
2908				SCTP_INP_INFO_WUNLOCK();
2909				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2910				return (EADDRINUSE);
2911			}
2912		}
2913continue_anyway:
2914		SCTP_INP_WLOCK(inp);
2915		if (bindall) {
2916			/* verify that no lport is not used by a singleton */
2917			if ((port_reuse_active == 0) &&
2918			    (inp_tmp = sctp_isport_inuse(inp, lport, vrf_id))) {
2919				/* Sorry someone already has this one bound */
2920				if ((sctp_is_feature_on(inp, SCTP_PCB_FLAGS_PORTREUSE)) &&
2921				    (sctp_is_feature_on(inp_tmp, SCTP_PCB_FLAGS_PORTREUSE))) {
2922					port_reuse_active = 1;
2923				} else {
2924					SCTP_INP_DECR_REF(inp);
2925					SCTP_INP_WUNLOCK(inp);
2926					SCTP_INP_INFO_WUNLOCK();
2927					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2928					return (EADDRINUSE);
2929				}
2930			}
2931		}
2932	} else {
2933		uint16_t first, last, candidate;
2934		uint16_t count;
2935		int done;
2936
2937		if (ip_inp->inp_flags & INP_HIGHPORT) {
2938			first = MODULE_GLOBAL(ipport_hifirstauto);
2939			last = MODULE_GLOBAL(ipport_hilastauto);
2940		} else if (ip_inp->inp_flags & INP_LOWPORT) {
2941			if (p && (error =
2942			    priv_check(p, PRIV_NETINET_RESERVEDPORT)
2943			    )) {
2944				SCTP_INP_DECR_REF(inp);
2945				SCTP_INP_WUNLOCK(inp);
2946				SCTP_INP_INFO_WUNLOCK();
2947				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, error);
2948				return (error);
2949			}
2950			first = MODULE_GLOBAL(ipport_lowfirstauto);
2951			last = MODULE_GLOBAL(ipport_lowlastauto);
2952		} else {
2953			first = MODULE_GLOBAL(ipport_firstauto);
2954			last = MODULE_GLOBAL(ipport_lastauto);
2955		}
2956		if (first > last) {
2957			uint16_t temp;
2958
2959			temp = first;
2960			first = last;
2961			last = temp;
2962		}
2963		count = last - first + 1;	/* number of candidates */
2964		candidate = first + sctp_select_initial_TSN(&inp->sctp_ep) % (count);
2965
2966		done = 0;
2967		while (!done) {
2968			if (sctp_isport_inuse(inp, htons(candidate), inp->def_vrf_id) == NULL) {
2969				done = 1;
2970			}
2971			if (!done) {
2972				if (--count == 0) {
2973					SCTP_INP_DECR_REF(inp);
2974					SCTP_INP_WUNLOCK(inp);
2975					SCTP_INP_INFO_WUNLOCK();
2976					SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRINUSE);
2977					return (EADDRINUSE);
2978				}
2979				if (candidate == last)
2980					candidate = first;
2981				else
2982					candidate = candidate + 1;
2983			}
2984		}
2985		lport = htons(candidate);
2986	}
2987	SCTP_INP_DECR_REF(inp);
2988	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE |
2989	    SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
2990		/*
2991		 * this really should not happen. The guy did a non-blocking
2992		 * bind and then did a close at the same time.
2993		 */
2994		SCTP_INP_WUNLOCK(inp);
2995		SCTP_INP_INFO_WUNLOCK();
2996		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
2997		return (EINVAL);
2998	}
2999	/* ok we look clear to give out this port, so lets setup the binding */
3000	if (bindall) {
3001		/* binding to all addresses, so just set in the proper flags */
3002		inp->sctp_flags |= SCTP_PCB_FLAGS_BOUNDALL;
3003		/* set the automatic addr changes from kernel flag */
3004		if (SCTP_BASE_SYSCTL(sctp_auto_asconf) == 0) {
3005			sctp_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3006			sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3007		} else {
3008			sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3009			sctp_feature_on(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3010		}
3011		if (SCTP_BASE_SYSCTL(sctp_multiple_asconfs) == 0) {
3012			sctp_feature_off(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3013		} else {
3014			sctp_feature_on(inp, SCTP_PCB_FLAGS_MULTIPLE_ASCONFS);
3015		}
3016		/*
3017		 * set the automatic mobility_base from kernel flag (by
3018		 * micchie)
3019		 */
3020		if (SCTP_BASE_SYSCTL(sctp_mobility_base) == 0) {
3021			sctp_mobility_feature_off(inp, SCTP_MOBILITY_BASE);
3022			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3023		} else {
3024			sctp_mobility_feature_on(inp, SCTP_MOBILITY_BASE);
3025			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3026		}
3027		/*
3028		 * set the automatic mobility_fasthandoff from kernel flag
3029		 * (by micchie)
3030		 */
3031		if (SCTP_BASE_SYSCTL(sctp_mobility_fasthandoff) == 0) {
3032			sctp_mobility_feature_off(inp, SCTP_MOBILITY_FASTHANDOFF);
3033			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3034		} else {
3035			sctp_mobility_feature_on(inp, SCTP_MOBILITY_FASTHANDOFF);
3036			sctp_mobility_feature_off(inp, SCTP_MOBILITY_PRIM_DELETED);
3037		}
3038	} else {
3039		/*
3040		 * bind specific, make sure flags is off and add a new
3041		 * address structure to the sctp_addr_list inside the ep
3042		 * structure.
3043		 *
3044		 * We will need to allocate one and insert it at the head. The
3045		 * socketopt call can just insert new addresses in there as
3046		 * well. It will also have to do the embed scope kame hack
3047		 * too (before adding).
3048		 */
3049		struct sctp_ifa *ifa;
3050		struct sockaddr_storage store_sa;
3051
3052		memset(&store_sa, 0, sizeof(store_sa));
3053		switch (addr->sa_family) {
3054#ifdef INET
3055		case AF_INET:
3056			{
3057				struct sockaddr_in *sin;
3058
3059				sin = (struct sockaddr_in *)&store_sa;
3060				memcpy(sin, addr, sizeof(struct sockaddr_in));
3061				sin->sin_port = 0;
3062				break;
3063			}
3064#endif
3065#ifdef INET6
3066		case AF_INET6:
3067			{
3068				struct sockaddr_in6 *sin6;
3069
3070				sin6 = (struct sockaddr_in6 *)&store_sa;
3071				memcpy(sin6, addr, sizeof(struct sockaddr_in6));
3072				sin6->sin6_port = 0;
3073				break;
3074			}
3075#endif
3076		default:
3077			break;
3078		}
3079		/*
3080		 * first find the interface with the bound address need to
3081		 * zero out the port to find the address! yuck! can't do
3082		 * this earlier since need port for sctp_pcb_findep()
3083		 */
3084		if (sctp_ifap != NULL) {
3085			ifa = sctp_ifap;
3086		} else {
3087			/*
3088			 * Note for BSD we hit here always other O/S's will
3089			 * pass things in via the sctp_ifap argument
3090			 * (Panda).
3091			 */
3092			ifa = sctp_find_ifa_by_addr((struct sockaddr *)&store_sa,
3093			    vrf_id, SCTP_ADDR_NOT_LOCKED);
3094		}
3095		if (ifa == NULL) {
3096			/* Can't find an interface with that address */
3097			SCTP_INP_WUNLOCK(inp);
3098			SCTP_INP_INFO_WUNLOCK();
3099			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EADDRNOTAVAIL);
3100			return (EADDRNOTAVAIL);
3101		}
3102#ifdef INET6
3103		if (addr->sa_family == AF_INET6) {
3104			/* GAK, more FIXME IFA lock? */
3105			if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
3106				/* Can't bind a non-existent addr. */
3107				SCTP_INP_WUNLOCK(inp);
3108				SCTP_INP_INFO_WUNLOCK();
3109				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
3110				return (EINVAL);
3111			}
3112		}
3113#endif
3114		/* we're not bound all */
3115		inp->sctp_flags &= ~SCTP_PCB_FLAGS_BOUNDALL;
3116		/* allow bindx() to send ASCONF's for binding changes */
3117		sctp_feature_on(inp, SCTP_PCB_FLAGS_DO_ASCONF);
3118		/* clear automatic addr changes from kernel flag */
3119		sctp_feature_off(inp, SCTP_PCB_FLAGS_AUTO_ASCONF);
3120
3121		/* add this address to the endpoint list */
3122		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, 0);
3123		if (error != 0) {
3124			SCTP_INP_WUNLOCK(inp);
3125			SCTP_INP_INFO_WUNLOCK();
3126			return (error);
3127		}
3128		inp->laddr_count++;
3129	}
3130	/* find the bucket */
3131	if (port_reuse_active) {
3132		/* Put it into tcp 1-2-1 hash */
3133		head = &SCTP_BASE_INFO(sctp_tcpephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashtcpmark))];
3134		inp->sctp_flags |= SCTP_PCB_FLAGS_IN_TCPPOOL;
3135	} else {
3136		head = &SCTP_BASE_INFO(sctp_ephash)[SCTP_PCBHASH_ALLADDR(lport, SCTP_BASE_INFO(hashmark))];
3137	}
3138	/* put it in the bucket */
3139	LIST_INSERT_HEAD(head, inp, sctp_hash);
3140	SCTPDBG(SCTP_DEBUG_PCB1, "Main hash to bind at head:%p, bound port:%d - in tcp_pool=%d\n",
3141	    (void *)head, ntohs(lport), port_reuse_active);
3142	/* set in the port */
3143	inp->sctp_lport = lport;
3144
3145	/* turn off just the unbound flag */
3146	inp->sctp_flags &= ~SCTP_PCB_FLAGS_UNBOUND;
3147	SCTP_INP_WUNLOCK(inp);
3148	SCTP_INP_INFO_WUNLOCK();
3149	return (0);
3150}
3151
3152
3153static void
3154sctp_iterator_inp_being_freed(struct sctp_inpcb *inp)
3155{
3156	struct sctp_iterator *it, *nit;
3157
3158	/*
3159	 * We enter with the only the ITERATOR_LOCK in place and a write
3160	 * lock on the inp_info stuff.
3161	 */
3162	it = sctp_it_ctl.cur_it;
3163	if (it && (it->vn != curvnet)) {
3164		/* Its not looking at our VNET */
3165		return;
3166	}
3167	if (it && (it->inp == inp)) {
3168		/*
3169		 * This is tricky and we hold the iterator lock, but when it
3170		 * returns and gets the lock (when we release it) the
3171		 * iterator will try to operate on inp. We need to stop that
3172		 * from happening. But of course the iterator has a
3173		 * reference on the stcb and inp. We can mark it and it will
3174		 * stop.
3175		 *
3176		 * If its a single iterator situation, we set the end iterator
3177		 * flag. Otherwise we set the iterator to go to the next
3178		 * inp.
3179		 *
3180		 */
3181		if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3182			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
3183		} else {
3184			sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_INP;
3185		}
3186	}
3187	/*
3188	 * Now go through and remove any single reference to our inp that
3189	 * may be still pending on the list
3190	 */
3191	SCTP_IPI_ITERATOR_WQ_LOCK();
3192	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
3193		if (it->vn != curvnet) {
3194			continue;
3195		}
3196		if (it->inp == inp) {
3197			/* This one points to me is it inp specific? */
3198			if (it->iterator_flags & SCTP_ITERATOR_DO_SINGLE_INP) {
3199				/* Remove and free this one */
3200				TAILQ_REMOVE(&sctp_it_ctl.iteratorhead,
3201				    it, sctp_nxt_itr);
3202				if (it->function_atend != NULL) {
3203					(*it->function_atend) (it->pointer, it->val);
3204				}
3205				SCTP_FREE(it, SCTP_M_ITER);
3206			} else {
3207				it->inp = LIST_NEXT(it->inp, sctp_list);
3208				if (it->inp) {
3209					SCTP_INP_INCR_REF(it->inp);
3210				}
3211			}
3212			/*
3213			 * When its put in the refcnt is incremented so decr
3214			 * it
3215			 */
3216			SCTP_INP_DECR_REF(inp);
3217		}
3218	}
3219	SCTP_IPI_ITERATOR_WQ_UNLOCK();
3220}
3221
3222/* release sctp_inpcb unbind the port */
3223void
3224sctp_inpcb_free(struct sctp_inpcb *inp, int immediate, int from)
3225{
3226	/*
3227	 * Here we free a endpoint. We must find it (if it is in the Hash
3228	 * table) and remove it from there. Then we must also find it in the
3229	 * overall list and remove it from there. After all removals are
3230	 * complete then any timer has to be stopped. Then start the actual
3231	 * freeing. a) Any local lists. b) Any associations. c) The hash of
3232	 * all associations. d) finally the ep itself.
3233	 */
3234	struct sctp_tcb *asoc, *nasoc;
3235	struct sctp_laddr *laddr, *nladdr;
3236	struct inpcb *ip_pcb;
3237	struct socket *so;
3238	int being_refed = 0;
3239	struct sctp_queued_to_read *sq, *nsq;
3240	int cnt;
3241	sctp_sharedkey_t *shared_key, *nshared_key;
3242
3243
3244#ifdef SCTP_LOG_CLOSING
3245	sctp_log_closing(inp, NULL, 0);
3246#endif
3247	SCTP_ITERATOR_LOCK();
3248	/* mark any iterators on the list or being processed */
3249	sctp_iterator_inp_being_freed(inp);
3250	SCTP_ITERATOR_UNLOCK();
3251	so = inp->sctp_socket;
3252	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
3253		/* been here before.. eeks.. get out of here */
3254		SCTP_PRINTF("This conflict in free SHOULD not be happening! from %d, imm %d\n", from, immediate);
3255#ifdef SCTP_LOG_CLOSING
3256		sctp_log_closing(inp, NULL, 1);
3257#endif
3258		return;
3259	}
3260	SCTP_ASOC_CREATE_LOCK(inp);
3261	SCTP_INP_INFO_WLOCK();
3262
3263	SCTP_INP_WLOCK(inp);
3264	if (from == SCTP_CALLED_AFTER_CMPSET_OFCLOSE) {
3265		inp->sctp_flags &= ~SCTP_PCB_FLAGS_CLOSE_IP;
3266		/* socket is gone, so no more wakeups allowed */
3267		inp->sctp_flags |= SCTP_PCB_FLAGS_DONT_WAKE;
3268		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEINPUT;
3269		inp->sctp_flags &= ~SCTP_PCB_FLAGS_WAKEOUTPUT;
3270
3271	}
3272	/* First time through we have the socket lock, after that no more. */
3273	sctp_timer_stop(SCTP_TIMER_TYPE_NEWCOOKIE, inp, NULL, NULL,
3274	    SCTP_FROM_SCTP_PCB + SCTP_LOC_1);
3275
3276	if (inp->control) {
3277		sctp_m_freem(inp->control);
3278		inp->control = NULL;
3279	}
3280	if (inp->pkt) {
3281		sctp_m_freem(inp->pkt);
3282		inp->pkt = NULL;
3283	}
3284	ip_pcb = &inp->ip_inp.inp;	/* we could just cast the main pointer
3285					 * here but I will be nice :> (i.e.
3286					 * ip_pcb = ep;) */
3287	if (immediate == SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE) {
3288		int cnt_in_sd;
3289
3290		cnt_in_sd = 0;
3291		LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3292			SCTP_TCB_LOCK(asoc);
3293			if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3294				/* Skip guys being freed */
3295				cnt_in_sd++;
3296				if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3297					/*
3298					 * Special case - we did not start a
3299					 * kill timer on the asoc due to it
3300					 * was not closed. So go ahead and
3301					 * start it now.
3302					 */
3303					asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3304					sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3305				}
3306				SCTP_TCB_UNLOCK(asoc);
3307				continue;
3308			}
3309			if (((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_WAIT) ||
3310			    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_COOKIE_ECHOED)) &&
3311			    (asoc->asoc.total_output_queue_size == 0)) {
3312				/*
3313				 * If we have data in queue, we don't want
3314				 * to just free since the app may have done,
3315				 * send()/close or connect/send/close. And
3316				 * it wants the data to get across first.
3317				 */
3318				/* Just abandon things in the front states */
3319				if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_NOFORCE,
3320				    SCTP_FROM_SCTP_PCB + SCTP_LOC_2) == 0) {
3321					cnt_in_sd++;
3322				}
3323				continue;
3324			}
3325			/* Disconnect the socket please */
3326			asoc->sctp_socket = NULL;
3327			asoc->asoc.state |= SCTP_STATE_CLOSED_SOCKET;
3328			if ((asoc->asoc.size_on_reasm_queue > 0) ||
3329			    (asoc->asoc.control_pdapi) ||
3330			    (asoc->asoc.size_on_all_streams > 0) ||
3331			    (so && (so->so_rcv.sb_cc > 0))) {
3332				/* Left with Data unread */
3333				struct mbuf *op_err;
3334
3335				op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3336				asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_3;
3337				sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3338				SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3339				if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3340				    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3341					SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3342				}
3343				if (sctp_free_assoc(inp, asoc,
3344				    SCTP_PCBFREE_NOFORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_4) == 0) {
3345					cnt_in_sd++;
3346				}
3347				continue;
3348			} else if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3349				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3350			    (asoc->asoc.stream_queue_cnt == 0)) {
3351				if (asoc->asoc.locked_on_sending) {
3352					goto abort_anyway;
3353				}
3354				if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_SENT) &&
3355				    (SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_SHUTDOWN_ACK_SENT)) {
3356					struct sctp_nets *netp;
3357
3358					/*
3359					 * there is nothing queued to send,
3360					 * so I send shutdown
3361					 */
3362					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3363					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3364						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3365					}
3366					SCTP_SET_STATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_SENT);
3367					SCTP_CLEAR_SUBSTATE(&asoc->asoc, SCTP_STATE_SHUTDOWN_PENDING);
3368					sctp_stop_timers_for_shutdown(asoc);
3369					if (asoc->asoc.alternate) {
3370						netp = asoc->asoc.alternate;
3371					} else {
3372						netp = asoc->asoc.primary_destination;
3373					}
3374					sctp_send_shutdown(asoc, netp);
3375					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWN, asoc->sctp_ep, asoc,
3376					    netp);
3377					sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3378					    asoc->asoc.primary_destination);
3379					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_SHUT_TMR, SCTP_SO_LOCKED);
3380				}
3381			} else {
3382				/* mark into shutdown pending */
3383				struct sctp_stream_queue_pending *sp;
3384
3385				asoc->asoc.state |= SCTP_STATE_SHUTDOWN_PENDING;
3386				sctp_timer_start(SCTP_TIMER_TYPE_SHUTDOWNGUARD, asoc->sctp_ep, asoc,
3387				    asoc->asoc.primary_destination);
3388				if (asoc->asoc.locked_on_sending) {
3389					sp = TAILQ_LAST(&((asoc->asoc.locked_on_sending)->outqueue),
3390					    sctp_streamhead);
3391					if (sp == NULL) {
3392						SCTP_PRINTF("Error, sp is NULL, locked on sending is %p strm:%d\n",
3393						    (void *)asoc->asoc.locked_on_sending,
3394						    asoc->asoc.locked_on_sending->stream_no);
3395					} else {
3396						if ((sp->length == 0) && (sp->msg_is_complete == 0))
3397							asoc->asoc.state |= SCTP_STATE_PARTIAL_MSG_LEFT;
3398					}
3399				}
3400				if (TAILQ_EMPTY(&asoc->asoc.send_queue) &&
3401				    TAILQ_EMPTY(&asoc->asoc.sent_queue) &&
3402				    (asoc->asoc.state & SCTP_STATE_PARTIAL_MSG_LEFT)) {
3403					struct mbuf *op_err;
3404
3405			abort_anyway:
3406					op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3407					asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_5;
3408					sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3409					SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3410					if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3411					    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3412						SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3413					}
3414					if (sctp_free_assoc(inp, asoc,
3415					    SCTP_PCBFREE_NOFORCE,
3416					    SCTP_FROM_SCTP_PCB + SCTP_LOC_6) == 0) {
3417						cnt_in_sd++;
3418					}
3419					continue;
3420				} else {
3421					sctp_chunk_output(inp, asoc, SCTP_OUTPUT_FROM_CLOSING, SCTP_SO_LOCKED);
3422				}
3423			}
3424			cnt_in_sd++;
3425			SCTP_TCB_UNLOCK(asoc);
3426		}
3427		/* now is there some left in our SHUTDOWN state? */
3428		if (cnt_in_sd) {
3429#ifdef SCTP_LOG_CLOSING
3430			sctp_log_closing(inp, NULL, 2);
3431#endif
3432			inp->sctp_socket = NULL;
3433			SCTP_INP_WUNLOCK(inp);
3434			SCTP_ASOC_CREATE_UNLOCK(inp);
3435			SCTP_INP_INFO_WUNLOCK();
3436			return;
3437		}
3438	}
3439	inp->sctp_socket = NULL;
3440	if ((inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) !=
3441	    SCTP_PCB_FLAGS_UNBOUND) {
3442		/*
3443		 * ok, this guy has been bound. It's port is somewhere in
3444		 * the SCTP_BASE_INFO(hash table). Remove it!
3445		 */
3446		LIST_REMOVE(inp, sctp_hash);
3447		inp->sctp_flags |= SCTP_PCB_FLAGS_UNBOUND;
3448	}
3449	/*
3450	 * If there is a timer running to kill us, forget it, since it may
3451	 * have a contest on the INP lock.. which would cause us to die ...
3452	 */
3453	cnt = 0;
3454	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_list, sctp_tcblist, nasoc) {
3455		SCTP_TCB_LOCK(asoc);
3456		if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3457			if (asoc->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE) {
3458				asoc->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
3459				sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, asoc, NULL);
3460			}
3461			cnt++;
3462			SCTP_TCB_UNLOCK(asoc);
3463			continue;
3464		}
3465		/* Free associations that are NOT killing us */
3466		if ((SCTP_GET_STATE(&asoc->asoc) != SCTP_STATE_COOKIE_WAIT) &&
3467		    ((asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0)) {
3468			struct mbuf *op_err;
3469
3470			op_err = sctp_generate_cause(SCTP_CAUSE_USER_INITIATED_ABT, "");
3471			asoc->sctp_ep->last_abort_code = SCTP_FROM_SCTP_PCB + SCTP_LOC_7;
3472			sctp_send_abort_tcb(asoc, op_err, SCTP_SO_LOCKED);
3473			SCTP_STAT_INCR_COUNTER32(sctps_aborted);
3474		} else if (asoc->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) {
3475			cnt++;
3476			SCTP_TCB_UNLOCK(asoc);
3477			continue;
3478		}
3479		if ((SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_OPEN) ||
3480		    (SCTP_GET_STATE(&asoc->asoc) == SCTP_STATE_SHUTDOWN_RECEIVED)) {
3481			SCTP_STAT_DECR_GAUGE32(sctps_currestab);
3482		}
3483		if (sctp_free_assoc(inp, asoc, SCTP_PCBFREE_FORCE, SCTP_FROM_SCTP_PCB + SCTP_LOC_8) == 0) {
3484			cnt++;
3485		}
3486	}
3487	if (cnt) {
3488		/* Ok we have someone out there that will kill us */
3489		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3490#ifdef SCTP_LOG_CLOSING
3491		sctp_log_closing(inp, NULL, 3);
3492#endif
3493		SCTP_INP_WUNLOCK(inp);
3494		SCTP_ASOC_CREATE_UNLOCK(inp);
3495		SCTP_INP_INFO_WUNLOCK();
3496		return;
3497	}
3498	if (SCTP_INP_LOCK_CONTENDED(inp))
3499		being_refed++;
3500	if (SCTP_INP_READ_CONTENDED(inp))
3501		being_refed++;
3502	if (SCTP_ASOC_CREATE_LOCK_CONTENDED(inp))
3503		being_refed++;
3504
3505	if ((inp->refcount) ||
3506	    (being_refed) ||
3507	    (inp->sctp_flags & SCTP_PCB_FLAGS_CLOSE_IP)) {
3508		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3509#ifdef SCTP_LOG_CLOSING
3510		sctp_log_closing(inp, NULL, 4);
3511#endif
3512		sctp_timer_start(SCTP_TIMER_TYPE_INPKILL, inp, NULL, NULL);
3513		SCTP_INP_WUNLOCK(inp);
3514		SCTP_ASOC_CREATE_UNLOCK(inp);
3515		SCTP_INP_INFO_WUNLOCK();
3516		return;
3517	}
3518	inp->sctp_ep.signature_change.type = 0;
3519	inp->sctp_flags |= SCTP_PCB_FLAGS_SOCKET_ALLGONE;
3520	/*
3521	 * Remove it from the list .. last thing we need a lock for.
3522	 */
3523	LIST_REMOVE(inp, sctp_list);
3524	SCTP_INP_WUNLOCK(inp);
3525	SCTP_ASOC_CREATE_UNLOCK(inp);
3526	SCTP_INP_INFO_WUNLOCK();
3527	/*
3528	 * Now we release all locks. Since this INP cannot be found anymore
3529	 * except possibly by the kill timer that might be running. We call
3530	 * the drain function here. It should hit the case were it sees the
3531	 * ACTIVE flag cleared and exit out freeing us to proceed and
3532	 * destroy everything.
3533	 */
3534	if (from != SCTP_CALLED_FROM_INPKILL_TIMER) {
3535		(void)SCTP_OS_TIMER_STOP_DRAIN(&inp->sctp_ep.signature_change.timer);
3536	} else {
3537		/* Probably un-needed */
3538		(void)SCTP_OS_TIMER_STOP(&inp->sctp_ep.signature_change.timer);
3539	}
3540
3541#ifdef SCTP_LOG_CLOSING
3542	sctp_log_closing(inp, NULL, 5);
3543#endif
3544
3545
3546	if ((inp->sctp_asocidhash) != NULL) {
3547		SCTP_HASH_FREE(inp->sctp_asocidhash, inp->hashasocidmark);
3548		inp->sctp_asocidhash = NULL;
3549	}
3550	/* sa_ignore FREED_MEMORY */
3551	TAILQ_FOREACH_SAFE(sq, &inp->read_queue, next, nsq) {
3552		/* Its only abandoned if it had data left */
3553		if (sq->length)
3554			SCTP_STAT_INCR(sctps_left_abandon);
3555
3556		TAILQ_REMOVE(&inp->read_queue, sq, next);
3557		sctp_free_remote_addr(sq->whoFrom);
3558		if (so)
3559			so->so_rcv.sb_cc -= sq->length;
3560		if (sq->data) {
3561			sctp_m_freem(sq->data);
3562			sq->data = NULL;
3563		}
3564		/*
3565		 * no need to free the net count, since at this point all
3566		 * assoc's are gone.
3567		 */
3568		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
3569		SCTP_DECR_READQ_COUNT();
3570	}
3571	/* Now the sctp_pcb things */
3572	/*
3573	 * free each asoc if it is not already closed/free. we can't use the
3574	 * macro here since le_next will get freed as part of the
3575	 * sctp_free_assoc() call.
3576	 */
3577	if (so) {
3578#ifdef IPSEC
3579		ipsec_delete_pcbpolicy(ip_pcb);
3580#endif				/* IPSEC */
3581
3582		/* Unlocks not needed since the socket is gone now */
3583	}
3584	if (ip_pcb->inp_options) {
3585		(void)sctp_m_free(ip_pcb->inp_options);
3586		ip_pcb->inp_options = 0;
3587	}
3588#ifdef INET6
3589	if (ip_pcb->inp_vflag & INP_IPV6) {
3590		struct in6pcb *in6p;
3591
3592		in6p = (struct in6pcb *)inp;
3593		ip6_freepcbopts(in6p->in6p_outputopts);
3594	}
3595#endif				/* INET6 */
3596	ip_pcb->inp_vflag = 0;
3597	/* free up authentication fields */
3598	if (inp->sctp_ep.local_auth_chunks != NULL)
3599		sctp_free_chunklist(inp->sctp_ep.local_auth_chunks);
3600	if (inp->sctp_ep.local_hmacs != NULL)
3601		sctp_free_hmaclist(inp->sctp_ep.local_hmacs);
3602
3603	LIST_FOREACH_SAFE(shared_key, &inp->sctp_ep.shared_keys, next, nshared_key) {
3604		LIST_REMOVE(shared_key, next);
3605		sctp_free_sharedkey(shared_key);
3606		/* sa_ignore FREED_MEMORY */
3607	}
3608
3609	/*
3610	 * if we have an address list the following will free the list of
3611	 * ifaddr's that are set into this ep. Again macro limitations here,
3612	 * since the LIST_FOREACH could be a bad idea.
3613	 */
3614	LIST_FOREACH_SAFE(laddr, &inp->sctp_addr_list, sctp_nxt_addr, nladdr) {
3615		sctp_remove_laddr(laddr);
3616	}
3617
3618#ifdef SCTP_TRACK_FREED_ASOCS
3619	/* TEMP CODE */
3620	LIST_FOREACH_SAFE(asoc, &inp->sctp_asoc_free_list, sctp_tcblist, nasoc) {
3621		LIST_REMOVE(asoc, sctp_tcblist);
3622		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), asoc);
3623		SCTP_DECR_ASOC_COUNT();
3624	}
3625	/* *** END TEMP CODE *** */
3626#endif
3627	/* Now lets see about freeing the EP hash table. */
3628	if (inp->sctp_tcbhash != NULL) {
3629		SCTP_HASH_FREE(inp->sctp_tcbhash, inp->sctp_hashmark);
3630		inp->sctp_tcbhash = NULL;
3631	}
3632	/* Now we must put the ep memory back into the zone pool */
3633	INP_LOCK_DESTROY(&inp->ip_inp.inp);
3634	SCTP_INP_LOCK_DESTROY(inp);
3635	SCTP_INP_READ_DESTROY(inp);
3636	SCTP_ASOC_CREATE_LOCK_DESTROY(inp);
3637	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_ep), inp);
3638	SCTP_DECR_EP_COUNT();
3639}
3640
3641
3642struct sctp_nets *
3643sctp_findnet(struct sctp_tcb *stcb, struct sockaddr *addr)
3644{
3645	struct sctp_nets *net;
3646
3647	/* locate the address */
3648	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
3649		if (sctp_cmpaddr(addr, (struct sockaddr *)&net->ro._l_addr))
3650			return (net);
3651	}
3652	return (NULL);
3653}
3654
3655
3656int
3657sctp_is_address_on_local_host(struct sockaddr *addr, uint32_t vrf_id)
3658{
3659	struct sctp_ifa *sctp_ifa;
3660
3661	sctp_ifa = sctp_find_ifa_by_addr(addr, vrf_id, SCTP_ADDR_NOT_LOCKED);
3662	if (sctp_ifa) {
3663		return (1);
3664	} else {
3665		return (0);
3666	}
3667}
3668
3669/*
3670 * add's a remote endpoint address, done with the INIT/INIT-ACK as well as
3671 * when a ASCONF arrives that adds it. It will also initialize all the cwnd
3672 * stats of stuff.
3673 */
3674int
3675sctp_add_remote_addr(struct sctp_tcb *stcb, struct sockaddr *newaddr,
3676    struct sctp_nets **netp, int set_scope, int from)
3677{
3678	/*
3679	 * The following is redundant to the same lines in the
3680	 * sctp_aloc_assoc() but is needed since others call the add address
3681	 * function
3682	 */
3683	struct sctp_nets *net, *netfirst;
3684	int addr_inscope;
3685
3686	SCTPDBG(SCTP_DEBUG_PCB1, "Adding an address (from:%d) to the peer: ",
3687	    from);
3688	SCTPDBG_ADDR(SCTP_DEBUG_PCB1, newaddr);
3689
3690	netfirst = sctp_findnet(stcb, newaddr);
3691	if (netfirst) {
3692		/*
3693		 * Lie and return ok, we don't want to make the association
3694		 * go away for this behavior. It will happen in the TCP
3695		 * model in a connected socket. It does not reach the hash
3696		 * table until after the association is built so it can't be
3697		 * found. Mark as reachable, since the initial creation will
3698		 * have been cleared and the NOT_IN_ASSOC flag will have
3699		 * been added... and we don't want to end up removing it
3700		 * back out.
3701		 */
3702		if (netfirst->dest_state & SCTP_ADDR_UNCONFIRMED) {
3703			netfirst->dest_state = (SCTP_ADDR_REACHABLE |
3704			    SCTP_ADDR_UNCONFIRMED);
3705		} else {
3706			netfirst->dest_state = SCTP_ADDR_REACHABLE;
3707		}
3708
3709		return (0);
3710	}
3711	addr_inscope = 1;
3712	switch (newaddr->sa_family) {
3713#ifdef INET
3714	case AF_INET:
3715		{
3716			struct sockaddr_in *sin;
3717
3718			sin = (struct sockaddr_in *)newaddr;
3719			if (sin->sin_addr.s_addr == 0) {
3720				/* Invalid address */
3721				return (-1);
3722			}
3723			/* zero out the bzero area */
3724			memset(&sin->sin_zero, 0, sizeof(sin->sin_zero));
3725
3726			/* assure len is set */
3727			sin->sin_len = sizeof(struct sockaddr_in);
3728			if (set_scope) {
3729#ifdef SCTP_DONT_DO_PRIVADDR_SCOPE
3730				stcb->asoc.scope.ipv4_local_scope = 1;
3731#else
3732				if (IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) {
3733					stcb->asoc.scope.ipv4_local_scope = 1;
3734				}
3735#endif				/* SCTP_DONT_DO_PRIVADDR_SCOPE */
3736			} else {
3737				/* Validate the address is in scope */
3738				if ((IN4_ISPRIVATE_ADDRESS(&sin->sin_addr)) &&
3739				    (stcb->asoc.scope.ipv4_local_scope == 0)) {
3740					addr_inscope = 0;
3741				}
3742			}
3743			break;
3744		}
3745#endif
3746#ifdef INET6
3747	case AF_INET6:
3748		{
3749			struct sockaddr_in6 *sin6;
3750
3751			sin6 = (struct sockaddr_in6 *)newaddr;
3752			if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
3753				/* Invalid address */
3754				return (-1);
3755			}
3756			/* assure len is set */
3757			sin6->sin6_len = sizeof(struct sockaddr_in6);
3758			if (set_scope) {
3759				if (sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id)) {
3760					stcb->asoc.scope.loopback_scope = 1;
3761					stcb->asoc.scope.local_scope = 0;
3762					stcb->asoc.scope.ipv4_local_scope = 1;
3763					stcb->asoc.scope.site_scope = 1;
3764				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr)) {
3765					/*
3766					 * If the new destination is a
3767					 * LINK_LOCAL we must have common
3768					 * site scope. Don't set the local
3769					 * scope since we may not share all
3770					 * links, only loopback can do this.
3771					 * Links on the local network would
3772					 * also be on our private network
3773					 * for v4 too.
3774					 */
3775					stcb->asoc.scope.ipv4_local_scope = 1;
3776					stcb->asoc.scope.site_scope = 1;
3777				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr)) {
3778					/*
3779					 * If the new destination is
3780					 * SITE_LOCAL then we must have site
3781					 * scope in common.
3782					 */
3783					stcb->asoc.scope.site_scope = 1;
3784				}
3785			} else {
3786				/* Validate the address is in scope */
3787				if (IN6_IS_ADDR_LOOPBACK(&sin6->sin6_addr) &&
3788				    (stcb->asoc.scope.loopback_scope == 0)) {
3789					addr_inscope = 0;
3790				} else if (IN6_IS_ADDR_LINKLOCAL(&sin6->sin6_addr) &&
3791				    (stcb->asoc.scope.local_scope == 0)) {
3792					addr_inscope = 0;
3793				} else if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr) &&
3794				    (stcb->asoc.scope.site_scope == 0)) {
3795					addr_inscope = 0;
3796				}
3797			}
3798			break;
3799		}
3800#endif
3801	default:
3802		/* not supported family type */
3803		return (-1);
3804	}
3805	net = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_net), struct sctp_nets);
3806	if (net == NULL) {
3807		return (-1);
3808	}
3809	SCTP_INCR_RADDR_COUNT();
3810	bzero(net, sizeof(struct sctp_nets));
3811	(void)SCTP_GETTIME_TIMEVAL(&net->start_time);
3812	memcpy(&net->ro._l_addr, newaddr, newaddr->sa_len);
3813	switch (newaddr->sa_family) {
3814#ifdef INET
3815	case AF_INET:
3816		((struct sockaddr_in *)&net->ro._l_addr)->sin_port = stcb->rport;
3817		break;
3818#endif
3819#ifdef INET6
3820	case AF_INET6:
3821		((struct sockaddr_in6 *)&net->ro._l_addr)->sin6_port = stcb->rport;
3822		break;
3823#endif
3824	default:
3825		break;
3826	}
3827	net->addr_is_local = sctp_is_address_on_local_host(newaddr, stcb->asoc.vrf_id);
3828	if (net->addr_is_local && ((set_scope || (from == SCTP_ADDR_IS_CONFIRMED)))) {
3829		stcb->asoc.scope.loopback_scope = 1;
3830		stcb->asoc.scope.ipv4_local_scope = 1;
3831		stcb->asoc.scope.local_scope = 0;
3832		stcb->asoc.scope.site_scope = 1;
3833		addr_inscope = 1;
3834	}
3835	net->failure_threshold = stcb->asoc.def_net_failure;
3836	net->pf_threshold = stcb->asoc.def_net_pf_threshold;
3837	if (addr_inscope == 0) {
3838		net->dest_state = (SCTP_ADDR_REACHABLE |
3839		    SCTP_ADDR_OUT_OF_SCOPE);
3840	} else {
3841		if (from == SCTP_ADDR_IS_CONFIRMED)
3842			/* SCTP_ADDR_IS_CONFIRMED is passed by connect_x */
3843			net->dest_state = SCTP_ADDR_REACHABLE;
3844		else
3845			net->dest_state = SCTP_ADDR_REACHABLE |
3846			    SCTP_ADDR_UNCONFIRMED;
3847	}
3848	/*
3849	 * We set this to 0, the timer code knows that this means its an
3850	 * initial value
3851	 */
3852	net->rto_needed = 1;
3853	net->RTO = 0;
3854	net->RTO_measured = 0;
3855	stcb->asoc.numnets++;
3856	net->ref_count = 1;
3857	net->cwr_window_tsn = net->last_cwr_tsn = stcb->asoc.sending_seq - 1;
3858	net->port = stcb->asoc.port;
3859	net->dscp = stcb->asoc.default_dscp;
3860#ifdef INET6
3861	net->flowlabel = stcb->asoc.default_flowlabel;
3862#endif
3863	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DONOT_HEARTBEAT)) {
3864		net->dest_state |= SCTP_ADDR_NOHB;
3865	} else {
3866		net->dest_state &= ~SCTP_ADDR_NOHB;
3867	}
3868	if (sctp_stcb_is_feature_on(stcb->sctp_ep, stcb, SCTP_PCB_FLAGS_DO_NOT_PMTUD)) {
3869		net->dest_state |= SCTP_ADDR_NO_PMTUD;
3870	} else {
3871		net->dest_state &= ~SCTP_ADDR_NO_PMTUD;
3872	}
3873	net->heart_beat_delay = stcb->asoc.heart_beat_delay;
3874	/* Init the timer structure */
3875	SCTP_OS_TIMER_INIT(&net->rxt_timer.timer);
3876	SCTP_OS_TIMER_INIT(&net->pmtu_timer.timer);
3877	SCTP_OS_TIMER_INIT(&net->hb_timer.timer);
3878
3879	/* Now generate a route for this guy */
3880#ifdef INET6
3881	/* KAME hack: embed scopeid */
3882	if (newaddr->sa_family == AF_INET6) {
3883		struct sockaddr_in6 *sin6;
3884
3885		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3886		(void)sa6_embedscope(sin6, MODULE_GLOBAL(ip6_use_defzone));
3887		sin6->sin6_scope_id = 0;
3888	}
3889#endif
3890	SCTP_RTALLOC((sctp_route_t *) & net->ro, stcb->asoc.vrf_id);
3891
3892	if (SCTP_ROUTE_HAS_VALID_IFN(&net->ro)) {
3893		/* Get source address */
3894		net->ro._s_addr = sctp_source_address_selection(stcb->sctp_ep,
3895		    stcb,
3896		    (sctp_route_t *) & net->ro,
3897		    net,
3898		    0,
3899		    stcb->asoc.vrf_id);
3900		/* Now get the interface MTU */
3901		if (net->ro._s_addr && net->ro._s_addr->ifn_p) {
3902			net->mtu = SCTP_GATHER_MTU_FROM_INTFC(net->ro._s_addr->ifn_p);
3903		}
3904		if (net->mtu > 0) {
3905			uint32_t rmtu;
3906
3907			rmtu = SCTP_GATHER_MTU_FROM_ROUTE(net->ro._s_addr, &net->ro._l_addr.sa, net->ro.ro_rt);
3908			if (rmtu == 0) {
3909				/*
3910				 * Start things off to match mtu of
3911				 * interface please.
3912				 */
3913				SCTP_SET_MTU_OF_ROUTE(&net->ro._l_addr.sa,
3914				    net->ro.ro_rt, net->mtu);
3915			} else {
3916				/*
3917				 * we take the route mtu over the interface,
3918				 * since the route may be leading out the
3919				 * loopback, or a different interface.
3920				 */
3921				net->mtu = rmtu;
3922			}
3923		}
3924	}
3925	if (net->mtu == 0) {
3926		switch (newaddr->sa_family) {
3927#ifdef INET
3928		case AF_INET:
3929			net->mtu = SCTP_DEFAULT_MTU;
3930			break;
3931#endif
3932#ifdef INET6
3933		case AF_INET6:
3934			net->mtu = 1280;
3935			break;
3936#endif
3937		default:
3938			break;
3939		}
3940	}
3941	if (net->port) {
3942		net->mtu -= (uint32_t) sizeof(struct udphdr);
3943	}
3944	if (from == SCTP_ALLOC_ASOC) {
3945		stcb->asoc.smallest_mtu = net->mtu;
3946	}
3947	if (stcb->asoc.smallest_mtu > net->mtu) {
3948		stcb->asoc.smallest_mtu = net->mtu;
3949	}
3950#ifdef INET6
3951	if (newaddr->sa_family == AF_INET6) {
3952		struct sockaddr_in6 *sin6;
3953
3954		sin6 = (struct sockaddr_in6 *)&net->ro._l_addr;
3955		(void)sa6_recoverscope(sin6);
3956	}
3957#endif
3958
3959	/* JRS - Use the congestion control given in the CC module */
3960	if (stcb->asoc.cc_functions.sctp_set_initial_cc_param != NULL)
3961		(*stcb->asoc.cc_functions.sctp_set_initial_cc_param) (stcb, net);
3962
3963	/*
3964	 * CMT: CUC algo - set find_pseudo_cumack to TRUE (1) at beginning
3965	 * of assoc (2005/06/27, iyengar@cis.udel.edu)
3966	 */
3967	net->find_pseudo_cumack = 1;
3968	net->find_rtx_pseudo_cumack = 1;
3969	net->src_addr_selected = 0;
3970	/* Choose an initial flowid. */
3971	net->flowid = stcb->asoc.my_vtag ^
3972	    ntohs(stcb->rport) ^
3973	    ntohs(stcb->sctp_ep->sctp_lport);
3974#ifdef INVARIANTS
3975	net->flowidset = 1;
3976#endif
3977	if (netp) {
3978		*netp = net;
3979	}
3980	netfirst = TAILQ_FIRST(&stcb->asoc.nets);
3981	if (net->ro.ro_rt == NULL) {
3982		/* Since we have no route put it at the back */
3983		TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
3984	} else if (netfirst == NULL) {
3985		/* We are the first one in the pool. */
3986		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3987	} else if (netfirst->ro.ro_rt == NULL) {
3988		/*
3989		 * First one has NO route. Place this one ahead of the first
3990		 * one.
3991		 */
3992		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3993	} else if (net->ro.ro_rt->rt_ifp != netfirst->ro.ro_rt->rt_ifp) {
3994		/*
3995		 * This one has a different interface than the one at the
3996		 * top of the list. Place it ahead.
3997		 */
3998		TAILQ_INSERT_HEAD(&stcb->asoc.nets, net, sctp_next);
3999	} else {
4000		/*
4001		 * Ok we have the same interface as the first one. Move
4002		 * forward until we find either a) one with a NULL route...
4003		 * insert ahead of that b) one with a different ifp.. insert
4004		 * after that. c) end of the list.. insert at the tail.
4005		 */
4006		struct sctp_nets *netlook;
4007
4008		do {
4009			netlook = TAILQ_NEXT(netfirst, sctp_next);
4010			if (netlook == NULL) {
4011				/* End of the list */
4012				TAILQ_INSERT_TAIL(&stcb->asoc.nets, net, sctp_next);
4013				break;
4014			} else if (netlook->ro.ro_rt == NULL) {
4015				/* next one has NO route */
4016				TAILQ_INSERT_BEFORE(netfirst, net, sctp_next);
4017				break;
4018			} else if (netlook->ro.ro_rt->rt_ifp != net->ro.ro_rt->rt_ifp) {
4019				TAILQ_INSERT_AFTER(&stcb->asoc.nets, netlook,
4020				    net, sctp_next);
4021				break;
4022			}
4023			/* Shift forward */
4024			netfirst = netlook;
4025		} while (netlook != NULL);
4026	}
4027
4028	/* got to have a primary set */
4029	if (stcb->asoc.primary_destination == 0) {
4030		stcb->asoc.primary_destination = net;
4031	} else if ((stcb->asoc.primary_destination->ro.ro_rt == NULL) &&
4032		    (net->ro.ro_rt) &&
4033	    ((net->dest_state & SCTP_ADDR_UNCONFIRMED) == 0)) {
4034		/* No route to current primary adopt new primary */
4035		stcb->asoc.primary_destination = net;
4036	}
4037	/* Validate primary is first */
4038	net = TAILQ_FIRST(&stcb->asoc.nets);
4039	if ((net != stcb->asoc.primary_destination) &&
4040	    (stcb->asoc.primary_destination)) {
4041		/*
4042		 * first one on the list is NOT the primary sctp_cmpaddr()
4043		 * is much more efficient if the primary is the first on the
4044		 * list, make it so.
4045		 */
4046		TAILQ_REMOVE(&stcb->asoc.nets,
4047		    stcb->asoc.primary_destination, sctp_next);
4048		TAILQ_INSERT_HEAD(&stcb->asoc.nets,
4049		    stcb->asoc.primary_destination, sctp_next);
4050	}
4051	return (0);
4052}
4053
4054
4055static uint32_t
4056sctp_aloc_a_assoc_id(struct sctp_inpcb *inp, struct sctp_tcb *stcb)
4057{
4058	uint32_t id;
4059	struct sctpasochead *head;
4060	struct sctp_tcb *lstcb;
4061
4062	SCTP_INP_WLOCK(inp);
4063try_again:
4064	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
4065		/* TSNH */
4066		SCTP_INP_WUNLOCK(inp);
4067		return (0);
4068	}
4069	/*
4070	 * We don't allow assoc id to be one of SCTP_FUTURE_ASSOC,
4071	 * SCTP_CURRENT_ASSOC and SCTP_ALL_ASSOC.
4072	 */
4073	if (inp->sctp_associd_counter <= SCTP_ALL_ASSOC) {
4074		inp->sctp_associd_counter = SCTP_ALL_ASSOC + 1;
4075	}
4076	id = inp->sctp_associd_counter;
4077	inp->sctp_associd_counter++;
4078	lstcb = sctp_findasoc_ep_asocid_locked(inp, (sctp_assoc_t) id, 0);
4079	if (lstcb) {
4080		goto try_again;
4081	}
4082	head = &inp->sctp_asocidhash[SCTP_PCBHASH_ASOC(id, inp->hashasocidmark)];
4083	LIST_INSERT_HEAD(head, stcb, sctp_tcbasocidhash);
4084	stcb->asoc.in_asocid_hash = 1;
4085	SCTP_INP_WUNLOCK(inp);
4086	return id;
4087}
4088
4089/*
4090 * allocate an association and add it to the endpoint. The caller must be
4091 * careful to add all additional addresses once they are know right away or
4092 * else the assoc will be may experience a blackout scenario.
4093 */
4094struct sctp_tcb *
4095sctp_aloc_assoc(struct sctp_inpcb *inp, struct sockaddr *firstaddr,
4096    int *error, uint32_t override_tag, uint32_t vrf_id,
4097    struct thread *p
4098)
4099{
4100	/* note the p argument is only valid in unbound sockets */
4101
4102	struct sctp_tcb *stcb;
4103	struct sctp_association *asoc;
4104	struct sctpasochead *head;
4105	uint16_t rport;
4106	int err;
4107
4108	/*
4109	 * Assumption made here: Caller has done a
4110	 * sctp_findassociation_ep_addr(ep, addr's); to make sure the
4111	 * address does not exist already.
4112	 */
4113	if (SCTP_BASE_INFO(ipi_count_asoc) >= SCTP_MAX_NUM_OF_ASOC) {
4114		/* Hit max assoc, sorry no more */
4115		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4116		*error = ENOBUFS;
4117		return (NULL);
4118	}
4119	if (firstaddr == NULL) {
4120		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4121		*error = EINVAL;
4122		return (NULL);
4123	}
4124	SCTP_INP_RLOCK(inp);
4125	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) &&
4126	    ((sctp_is_feature_off(inp, SCTP_PCB_FLAGS_PORTREUSE)) ||
4127	    (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED))) {
4128		/*
4129		 * If its in the TCP pool, its NOT allowed to create an
4130		 * association. The parent listener needs to call
4131		 * sctp_aloc_assoc.. or the one-2-many socket. If a peeled
4132		 * off, or connected one does this.. its an error.
4133		 */
4134		SCTP_INP_RUNLOCK(inp);
4135		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4136		*error = EINVAL;
4137		return (NULL);
4138	}
4139	if ((inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL) ||
4140	    (inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE)) {
4141		if ((inp->sctp_flags & SCTP_PCB_FLAGS_WAS_CONNECTED) ||
4142		    (inp->sctp_flags & SCTP_PCB_FLAGS_WAS_ABORTED)) {
4143			SCTP_INP_RUNLOCK(inp);
4144			SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4145			*error = EINVAL;
4146			return (NULL);
4147		}
4148	}
4149	SCTPDBG(SCTP_DEBUG_PCB3, "Allocate an association for peer:");
4150#ifdef SCTP_DEBUG
4151	if (firstaddr) {
4152		SCTPDBG_ADDR(SCTP_DEBUG_PCB3, firstaddr);
4153		switch (firstaddr->sa_family) {
4154#ifdef INET
4155		case AF_INET:
4156			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4157			    ntohs(((struct sockaddr_in *)firstaddr)->sin_port));
4158			break;
4159#endif
4160#ifdef INET6
4161		case AF_INET6:
4162			SCTPDBG(SCTP_DEBUG_PCB3, "Port:%d\n",
4163			    ntohs(((struct sockaddr_in6 *)firstaddr)->sin6_port));
4164			break;
4165#endif
4166		default:
4167			break;
4168		}
4169	} else {
4170		SCTPDBG(SCTP_DEBUG_PCB3, "None\n");
4171	}
4172#endif				/* SCTP_DEBUG */
4173	switch (firstaddr->sa_family) {
4174#ifdef INET
4175	case AF_INET:
4176		{
4177			struct sockaddr_in *sin;
4178
4179			sin = (struct sockaddr_in *)firstaddr;
4180			if ((ntohs(sin->sin_port) == 0) ||
4181			    (sin->sin_addr.s_addr == INADDR_ANY) ||
4182			    (sin->sin_addr.s_addr == INADDR_BROADCAST) ||
4183			    IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
4184				/* Invalid address */
4185				SCTP_INP_RUNLOCK(inp);
4186				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4187				*error = EINVAL;
4188				return (NULL);
4189			}
4190			rport = sin->sin_port;
4191			break;
4192		}
4193#endif
4194#ifdef INET6
4195	case AF_INET6:
4196		{
4197			struct sockaddr_in6 *sin6;
4198
4199			sin6 = (struct sockaddr_in6 *)firstaddr;
4200			if ((ntohs(sin6->sin6_port) == 0) ||
4201			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
4202			    IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
4203				/* Invalid address */
4204				SCTP_INP_RUNLOCK(inp);
4205				SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4206				*error = EINVAL;
4207				return (NULL);
4208			}
4209			rport = sin6->sin6_port;
4210			break;
4211		}
4212#endif
4213	default:
4214		/* not supported family type */
4215		SCTP_INP_RUNLOCK(inp);
4216		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4217		*error = EINVAL;
4218		return (NULL);
4219	}
4220	SCTP_INP_RUNLOCK(inp);
4221	if (inp->sctp_flags & SCTP_PCB_FLAGS_UNBOUND) {
4222		/*
4223		 * If you have not performed a bind, then we need to do the
4224		 * ephemeral bind for you.
4225		 */
4226		if ((err = sctp_inpcb_bind(inp->sctp_socket,
4227		    (struct sockaddr *)NULL,
4228		    (struct sctp_ifa *)NULL,
4229		    p
4230		    ))) {
4231			/* bind error, probably perm */
4232			*error = err;
4233			return (NULL);
4234		}
4235	}
4236	stcb = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_asoc), struct sctp_tcb);
4237	if (stcb == NULL) {
4238		/* out of memory? */
4239		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
4240		*error = ENOMEM;
4241		return (NULL);
4242	}
4243	SCTP_INCR_ASOC_COUNT();
4244
4245	bzero(stcb, sizeof(*stcb));
4246	asoc = &stcb->asoc;
4247
4248	asoc->assoc_id = sctp_aloc_a_assoc_id(inp, stcb);
4249	SCTP_TCB_LOCK_INIT(stcb);
4250	SCTP_TCB_SEND_LOCK_INIT(stcb);
4251	stcb->rport = rport;
4252	/* setup back pointer's */
4253	stcb->sctp_ep = inp;
4254	stcb->sctp_socket = inp->sctp_socket;
4255	if ((err = sctp_init_asoc(inp, stcb, override_tag, vrf_id))) {
4256		/* failed */
4257		SCTP_TCB_LOCK_DESTROY(stcb);
4258		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4259		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4260		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4261		SCTP_DECR_ASOC_COUNT();
4262		*error = err;
4263		return (NULL);
4264	}
4265	/* and the port */
4266	SCTP_INP_INFO_WLOCK();
4267	SCTP_INP_WLOCK(inp);
4268	if (inp->sctp_flags & (SCTP_PCB_FLAGS_SOCKET_GONE | SCTP_PCB_FLAGS_SOCKET_ALLGONE)) {
4269		/* inpcb freed while alloc going on */
4270		SCTP_TCB_LOCK_DESTROY(stcb);
4271		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4272		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4273		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4274		SCTP_INP_WUNLOCK(inp);
4275		SCTP_INP_INFO_WUNLOCK();
4276		SCTP_DECR_ASOC_COUNT();
4277		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
4278		*error = EINVAL;
4279		return (NULL);
4280	}
4281	SCTP_TCB_LOCK(stcb);
4282
4283	/* now that my_vtag is set, add it to the hash */
4284	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(stcb->asoc.my_vtag, SCTP_BASE_INFO(hashasocmark))];
4285	/* put it in the bucket in the vtag hash of assoc's for the system */
4286	LIST_INSERT_HEAD(head, stcb, sctp_asocs);
4287	SCTP_INP_INFO_WUNLOCK();
4288
4289	if ((err = sctp_add_remote_addr(stcb, firstaddr, NULL, SCTP_DO_SETSCOPE, SCTP_ALLOC_ASOC))) {
4290		/* failure.. memory error? */
4291		if (asoc->strmout) {
4292			SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
4293			asoc->strmout = NULL;
4294		}
4295		if (asoc->mapping_array) {
4296			SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
4297			asoc->mapping_array = NULL;
4298		}
4299		if (asoc->nr_mapping_array) {
4300			SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
4301			asoc->nr_mapping_array = NULL;
4302		}
4303		SCTP_DECR_ASOC_COUNT();
4304		SCTP_TCB_UNLOCK(stcb);
4305		SCTP_TCB_LOCK_DESTROY(stcb);
4306		SCTP_TCB_SEND_LOCK_DESTROY(stcb);
4307		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4308		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
4309		SCTP_INP_WUNLOCK(inp);
4310		SCTP_LTRACE_ERR_RET(inp, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOBUFS);
4311		*error = ENOBUFS;
4312		return (NULL);
4313	}
4314	/* Init all the timers */
4315	SCTP_OS_TIMER_INIT(&asoc->dack_timer.timer);
4316	SCTP_OS_TIMER_INIT(&asoc->strreset_timer.timer);
4317	SCTP_OS_TIMER_INIT(&asoc->asconf_timer.timer);
4318	SCTP_OS_TIMER_INIT(&asoc->shut_guard_timer.timer);
4319	SCTP_OS_TIMER_INIT(&asoc->autoclose_timer.timer);
4320	SCTP_OS_TIMER_INIT(&asoc->delayed_event_timer.timer);
4321	SCTP_OS_TIMER_INIT(&asoc->delete_prim_timer.timer);
4322
4323	LIST_INSERT_HEAD(&inp->sctp_asoc_list, stcb, sctp_tcblist);
4324	/* now file the port under the hash as well */
4325	if (inp->sctp_tcbhash != NULL) {
4326		head = &inp->sctp_tcbhash[SCTP_PCBHASH_ALLADDR(stcb->rport,
4327		    inp->sctp_hashmark)];
4328		LIST_INSERT_HEAD(head, stcb, sctp_tcbhash);
4329	}
4330	SCTP_INP_WUNLOCK(inp);
4331	SCTPDBG(SCTP_DEBUG_PCB1, "Association %p now allocated\n", (void *)stcb);
4332	return (stcb);
4333}
4334
4335
4336void
4337sctp_remove_net(struct sctp_tcb *stcb, struct sctp_nets *net)
4338{
4339	struct sctp_association *asoc;
4340
4341	asoc = &stcb->asoc;
4342	asoc->numnets--;
4343	TAILQ_REMOVE(&asoc->nets, net, sctp_next);
4344	if (net == asoc->primary_destination) {
4345		/* Reset primary */
4346		struct sctp_nets *lnet;
4347
4348		lnet = TAILQ_FIRST(&asoc->nets);
4349		/*
4350		 * Mobility adaptation Ideally, if deleted destination is
4351		 * the primary, it becomes a fast retransmission trigger by
4352		 * the subsequent SET PRIMARY. (by micchie)
4353		 */
4354		if (sctp_is_mobility_feature_on(stcb->sctp_ep,
4355		    SCTP_MOBILITY_BASE) ||
4356		    sctp_is_mobility_feature_on(stcb->sctp_ep,
4357		    SCTP_MOBILITY_FASTHANDOFF)) {
4358			SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: primary dst is deleting\n");
4359			if (asoc->deleted_primary != NULL) {
4360				SCTPDBG(SCTP_DEBUG_ASCONF1, "remove_net: deleted primary may be already stored\n");
4361				goto out;
4362			}
4363			asoc->deleted_primary = net;
4364			atomic_add_int(&net->ref_count, 1);
4365			memset(&net->lastsa, 0, sizeof(net->lastsa));
4366			memset(&net->lastsv, 0, sizeof(net->lastsv));
4367			sctp_mobility_feature_on(stcb->sctp_ep,
4368			    SCTP_MOBILITY_PRIM_DELETED);
4369			sctp_timer_start(SCTP_TIMER_TYPE_PRIM_DELETED,
4370			    stcb->sctp_ep, stcb, NULL);
4371		}
4372out:
4373		/* Try to find a confirmed primary */
4374		asoc->primary_destination = sctp_find_alternate_net(stcb, lnet, 0);
4375	}
4376	if (net == asoc->last_data_chunk_from) {
4377		/* Reset primary */
4378		asoc->last_data_chunk_from = TAILQ_FIRST(&asoc->nets);
4379	}
4380	if (net == asoc->last_control_chunk_from) {
4381		/* Clear net */
4382		asoc->last_control_chunk_from = NULL;
4383	}
4384	if (net == stcb->asoc.alternate) {
4385		sctp_free_remote_addr(stcb->asoc.alternate);
4386		stcb->asoc.alternate = NULL;
4387	}
4388	sctp_free_remote_addr(net);
4389}
4390
4391/*
4392 * remove a remote endpoint address from an association, it will fail if the
4393 * address does not exist.
4394 */
4395int
4396sctp_del_remote_addr(struct sctp_tcb *stcb, struct sockaddr *remaddr)
4397{
4398	/*
4399	 * Here we need to remove a remote address. This is quite simple, we
4400	 * first find it in the list of address for the association
4401	 * (tasoc->asoc.nets) and then if it is there, we do a LIST_REMOVE
4402	 * on that item. Note we do not allow it to be removed if there are
4403	 * no other addresses.
4404	 */
4405	struct sctp_association *asoc;
4406	struct sctp_nets *net, *nnet;
4407
4408	asoc = &stcb->asoc;
4409
4410	/* locate the address */
4411	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
4412		if (net->ro._l_addr.sa.sa_family != remaddr->sa_family) {
4413			continue;
4414		}
4415		if (sctp_cmpaddr((struct sockaddr *)&net->ro._l_addr,
4416		    remaddr)) {
4417			/* we found the guy */
4418			if (asoc->numnets < 2) {
4419				/* Must have at LEAST two remote addresses */
4420				return (-1);
4421			} else {
4422				sctp_remove_net(stcb, net);
4423				return (0);
4424			}
4425		}
4426	}
4427	/* not found. */
4428	return (-2);
4429}
4430
4431void
4432sctp_delete_from_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4433{
4434	struct sctpvtaghead *chain;
4435	struct sctp_tagblock *twait_block;
4436	int found = 0;
4437	int i;
4438
4439	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4440	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4441		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4442			if ((twait_block->vtag_block[i].v_tag == tag) &&
4443			    (twait_block->vtag_block[i].lport == lport) &&
4444			    (twait_block->vtag_block[i].rport == rport)) {
4445				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4446				twait_block->vtag_block[i].v_tag = 0;
4447				twait_block->vtag_block[i].lport = 0;
4448				twait_block->vtag_block[i].rport = 0;
4449				found = 1;
4450				break;
4451			}
4452		}
4453		if (found)
4454			break;
4455	}
4456}
4457
4458int
4459sctp_is_in_timewait(uint32_t tag, uint16_t lport, uint16_t rport)
4460{
4461	struct sctpvtaghead *chain;
4462	struct sctp_tagblock *twait_block;
4463	int found = 0;
4464	int i;
4465
4466	SCTP_INP_INFO_WLOCK();
4467	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4468	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4469		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4470			if ((twait_block->vtag_block[i].v_tag == tag) &&
4471			    (twait_block->vtag_block[i].lport == lport) &&
4472			    (twait_block->vtag_block[i].rport == rport)) {
4473				found = 1;
4474				break;
4475			}
4476		}
4477		if (found)
4478			break;
4479	}
4480	SCTP_INP_INFO_WUNLOCK();
4481	return (found);
4482}
4483
4484
4485void
4486sctp_add_vtag_to_timewait(uint32_t tag, uint32_t time, uint16_t lport, uint16_t rport)
4487{
4488	struct sctpvtaghead *chain;
4489	struct sctp_tagblock *twait_block;
4490	struct timeval now;
4491	int set, i;
4492
4493	if (time == 0) {
4494		/* Its disabled */
4495		return;
4496	}
4497	(void)SCTP_GETTIME_TIMEVAL(&now);
4498	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
4499	set = 0;
4500	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
4501		/* Block(s) present, lets find space, and expire on the fly */
4502		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
4503			if ((twait_block->vtag_block[i].v_tag == 0) &&
4504			    !set) {
4505				twait_block->vtag_block[i].tv_sec_at_expire =
4506				    now.tv_sec + time;
4507				twait_block->vtag_block[i].v_tag = tag;
4508				twait_block->vtag_block[i].lport = lport;
4509				twait_block->vtag_block[i].rport = rport;
4510				set = 1;
4511			} else if ((twait_block->vtag_block[i].v_tag) &&
4512			    ((long)twait_block->vtag_block[i].tv_sec_at_expire < now.tv_sec)) {
4513				/* Audit expires this guy */
4514				twait_block->vtag_block[i].tv_sec_at_expire = 0;
4515				twait_block->vtag_block[i].v_tag = 0;
4516				twait_block->vtag_block[i].lport = 0;
4517				twait_block->vtag_block[i].rport = 0;
4518				if (set == 0) {
4519					/* Reuse it for my new tag */
4520					twait_block->vtag_block[i].tv_sec_at_expire = now.tv_sec + time;
4521					twait_block->vtag_block[i].v_tag = tag;
4522					twait_block->vtag_block[i].lport = lport;
4523					twait_block->vtag_block[i].rport = rport;
4524					set = 1;
4525				}
4526			}
4527		}
4528		if (set) {
4529			/*
4530			 * We only do up to the block where we can place our
4531			 * tag for audits
4532			 */
4533			break;
4534		}
4535	}
4536	/* Need to add a new block to chain */
4537	if (!set) {
4538		SCTP_MALLOC(twait_block, struct sctp_tagblock *,
4539		    sizeof(struct sctp_tagblock), SCTP_M_TIMW);
4540		if (twait_block == NULL) {
4541#ifdef INVARIANTS
4542			panic("Can not alloc tagblock");
4543#endif
4544			return;
4545		}
4546		memset(twait_block, 0, sizeof(struct sctp_tagblock));
4547		LIST_INSERT_HEAD(chain, twait_block, sctp_nxt_tagblock);
4548		twait_block->vtag_block[0].tv_sec_at_expire = now.tv_sec + time;
4549		twait_block->vtag_block[0].v_tag = tag;
4550		twait_block->vtag_block[0].lport = lport;
4551		twait_block->vtag_block[0].rport = rport;
4552	}
4553}
4554
4555
4556
4557/*-
4558 * Free the association after un-hashing the remote port. This
4559 * function ALWAYS returns holding NO LOCK on the stcb. It DOES
4560 * expect that the input to this function IS a locked TCB.
4561 * It will return 0, if it did NOT destroy the association (instead
4562 * it unlocks it. It will return NON-zero if it either destroyed the
4563 * association OR the association is already destroyed.
4564 */
4565int
4566sctp_free_assoc(struct sctp_inpcb *inp, struct sctp_tcb *stcb, int from_inpcbfree, int from_location)
4567{
4568	int i;
4569	struct sctp_association *asoc;
4570	struct sctp_nets *net, *nnet;
4571	struct sctp_laddr *laddr, *naddr;
4572	struct sctp_tmit_chunk *chk, *nchk;
4573	struct sctp_asconf_addr *aparam, *naparam;
4574	struct sctp_asconf_ack *aack, *naack;
4575	struct sctp_stream_reset_list *strrst, *nstrrst;
4576	struct sctp_queued_to_read *sq, *nsq;
4577	struct sctp_stream_queue_pending *sp, *nsp;
4578	sctp_sharedkey_t *shared_key, *nshared_key;
4579	struct socket *so;
4580
4581	/* first, lets purge the entry from the hash table. */
4582
4583#ifdef SCTP_LOG_CLOSING
4584	sctp_log_closing(inp, stcb, 6);
4585#endif
4586	if (stcb->asoc.state == 0) {
4587#ifdef SCTP_LOG_CLOSING
4588		sctp_log_closing(inp, NULL, 7);
4589#endif
4590		/* there is no asoc, really TSNH :-0 */
4591		return (1);
4592	}
4593	if (stcb->asoc.alternate) {
4594		sctp_free_remote_addr(stcb->asoc.alternate);
4595		stcb->asoc.alternate = NULL;
4596	}
4597	/* TEMP CODE */
4598	if (stcb->freed_from_where == 0) {
4599		/* Only record the first place free happened from */
4600		stcb->freed_from_where = from_location;
4601	}
4602	/* TEMP CODE */
4603
4604	asoc = &stcb->asoc;
4605	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4606	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4607		/* nothing around */
4608		so = NULL;
4609	else
4610		so = inp->sctp_socket;
4611
4612	/*
4613	 * We used timer based freeing if a reader or writer is in the way.
4614	 * So we first check if we are actually being called from a timer,
4615	 * if so we abort early if a reader or writer is still in the way.
4616	 */
4617	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) &&
4618	    (from_inpcbfree == SCTP_NORMAL_PROC)) {
4619		/*
4620		 * is it the timer driving us? if so are the reader/writers
4621		 * gone?
4622		 */
4623		if (stcb->asoc.refcnt) {
4624			/* nope, reader or writer in the way */
4625			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4626			/* no asoc destroyed */
4627			SCTP_TCB_UNLOCK(stcb);
4628#ifdef SCTP_LOG_CLOSING
4629			sctp_log_closing(inp, stcb, 8);
4630#endif
4631			return (0);
4632		}
4633	}
4634	/* now clean up any other timers */
4635	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4636	asoc->dack_timer.self = NULL;
4637	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4638	/*-
4639	 * For stream reset we don't blast this unless
4640	 * it is a str-reset timer, it might be the
4641	 * free-asoc timer which we DON'T want to
4642	 * disturb.
4643	 */
4644	if (asoc->strreset_timer.type == SCTP_TIMER_TYPE_STRRESET)
4645		asoc->strreset_timer.self = NULL;
4646	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4647	asoc->asconf_timer.self = NULL;
4648	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4649	asoc->autoclose_timer.self = NULL;
4650	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4651	asoc->shut_guard_timer.self = NULL;
4652	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4653	asoc->delayed_event_timer.self = NULL;
4654	/* Mobility adaptation */
4655	(void)SCTP_OS_TIMER_STOP(&asoc->delete_prim_timer.timer);
4656	asoc->delete_prim_timer.self = NULL;
4657	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4658		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4659		net->rxt_timer.self = NULL;
4660		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4661		net->pmtu_timer.self = NULL;
4662		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4663		net->hb_timer.self = NULL;
4664	}
4665	/* Now the read queue needs to be cleaned up (only once) */
4666	if ((stcb->asoc.state & SCTP_STATE_ABOUT_TO_BE_FREED) == 0) {
4667		stcb->asoc.state |= SCTP_STATE_ABOUT_TO_BE_FREED;
4668		SCTP_INP_READ_LOCK(inp);
4669		TAILQ_FOREACH(sq, &inp->read_queue, next) {
4670			if (sq->stcb == stcb) {
4671				sq->do_not_ref_stcb = 1;
4672				sq->sinfo_cumtsn = stcb->asoc.cumulative_tsn;
4673				/*
4674				 * If there is no end, there never will be
4675				 * now.
4676				 */
4677				if (sq->end_added == 0) {
4678					/* Held for PD-API clear that. */
4679					sq->pdapi_aborted = 1;
4680					sq->held_length = 0;
4681					if (sctp_stcb_is_feature_on(inp, stcb, SCTP_PCB_FLAGS_PDAPIEVNT) && (so != NULL)) {
4682						/*
4683						 * Need to add a PD-API
4684						 * aborted indication.
4685						 * Setting the control_pdapi
4686						 * assures that it will be
4687						 * added right after this
4688						 * msg.
4689						 */
4690						uint32_t strseq;
4691
4692						stcb->asoc.control_pdapi = sq;
4693						strseq = (sq->sinfo_stream << 16) | sq->sinfo_ssn;
4694						sctp_ulp_notify(SCTP_NOTIFY_PARTIAL_DELVIERY_INDICATION,
4695						    stcb,
4696						    SCTP_PARTIAL_DELIVERY_ABORTED,
4697						    (void *)&strseq,
4698						    SCTP_SO_LOCKED);
4699						stcb->asoc.control_pdapi = NULL;
4700					}
4701				}
4702				/* Add an end to wake them */
4703				sq->end_added = 1;
4704			}
4705		}
4706		SCTP_INP_READ_UNLOCK(inp);
4707		if (stcb->block_entry) {
4708			SCTP_LTRACE_ERR_RET(inp, stcb, NULL, SCTP_FROM_SCTP_PCB, ECONNRESET);
4709			stcb->block_entry->error = ECONNRESET;
4710			stcb->block_entry = NULL;
4711		}
4712	}
4713	if ((stcb->asoc.refcnt) || (stcb->asoc.state & SCTP_STATE_IN_ACCEPT_QUEUE)) {
4714		/*
4715		 * Someone holds a reference OR the socket is unaccepted
4716		 * yet.
4717		 */
4718		if ((stcb->asoc.refcnt) ||
4719		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4720		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE)) {
4721			stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4722			sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4723		}
4724		SCTP_TCB_UNLOCK(stcb);
4725		if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4726		    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4727			/* nothing around */
4728			so = NULL;
4729		if (so) {
4730			/* Wake any reader/writers */
4731			sctp_sorwakeup(inp, so);
4732			sctp_sowwakeup(inp, so);
4733		}
4734#ifdef SCTP_LOG_CLOSING
4735		sctp_log_closing(inp, stcb, 9);
4736#endif
4737		/* no asoc destroyed */
4738		return (0);
4739	}
4740#ifdef SCTP_LOG_CLOSING
4741	sctp_log_closing(inp, stcb, 10);
4742#endif
4743	/*
4744	 * When I reach here, no others want to kill the assoc yet.. and I
4745	 * own the lock. Now its possible an abort comes in when I do the
4746	 * lock exchange below to grab all the locks to do the final take
4747	 * out. to prevent this we increment the count, which will start a
4748	 * timer and blow out above thus assuring us that we hold exclusive
4749	 * killing of the asoc. Note that after getting back the TCB lock we
4750	 * will go ahead and increment the counter back up and stop any
4751	 * timer a passing stranger may have started :-S
4752	 */
4753	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4754		atomic_add_int(&stcb->asoc.refcnt, 1);
4755
4756		SCTP_TCB_UNLOCK(stcb);
4757		SCTP_INP_INFO_WLOCK();
4758		SCTP_INP_WLOCK(inp);
4759		SCTP_TCB_LOCK(stcb);
4760	}
4761	/* Double check the GONE flag */
4762	if ((inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) ||
4763	    (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE))
4764		/* nothing around */
4765		so = NULL;
4766
4767	if ((inp->sctp_flags & SCTP_PCB_FLAGS_TCPTYPE) ||
4768	    (inp->sctp_flags & SCTP_PCB_FLAGS_IN_TCPPOOL)) {
4769		/*
4770		 * For TCP type we need special handling when we are
4771		 * connected. We also include the peel'ed off ones to.
4772		 */
4773		if (inp->sctp_flags & SCTP_PCB_FLAGS_CONNECTED) {
4774			inp->sctp_flags &= ~SCTP_PCB_FLAGS_CONNECTED;
4775			inp->sctp_flags |= SCTP_PCB_FLAGS_WAS_CONNECTED;
4776			if (so) {
4777				SOCK_LOCK(so);
4778				if (so->so_rcv.sb_cc == 0) {
4779					so->so_state &= ~(SS_ISCONNECTING |
4780					    SS_ISDISCONNECTING |
4781					    SS_ISCONFIRMING |
4782					    SS_ISCONNECTED);
4783				}
4784				socantrcvmore_locked(so);
4785				sctp_sowwakeup(inp, so);
4786				sctp_sorwakeup(inp, so);
4787				SCTP_SOWAKEUP(so);
4788			}
4789		}
4790	}
4791	/*
4792	 * Make it invalid too, that way if its about to run it will abort
4793	 * and return.
4794	 */
4795	/* re-increment the lock */
4796	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4797		atomic_add_int(&stcb->asoc.refcnt, -1);
4798	}
4799	if (stcb->asoc.refcnt) {
4800		stcb->asoc.state &= ~SCTP_STATE_IN_ACCEPT_QUEUE;
4801		sctp_timer_start(SCTP_TIMER_TYPE_ASOCKILL, inp, stcb, NULL);
4802		if (from_inpcbfree == SCTP_NORMAL_PROC) {
4803			SCTP_INP_INFO_WUNLOCK();
4804			SCTP_INP_WUNLOCK(inp);
4805		}
4806		SCTP_TCB_UNLOCK(stcb);
4807		return (0);
4808	}
4809	asoc->state = 0;
4810	if (inp->sctp_tcbhash) {
4811		LIST_REMOVE(stcb, sctp_tcbhash);
4812	}
4813	if (stcb->asoc.in_asocid_hash) {
4814		LIST_REMOVE(stcb, sctp_tcbasocidhash);
4815	}
4816	/* Now lets remove it from the list of ALL associations in the EP */
4817	LIST_REMOVE(stcb, sctp_tcblist);
4818	if (from_inpcbfree == SCTP_NORMAL_PROC) {
4819		SCTP_INP_INCR_REF(inp);
4820		SCTP_INP_WUNLOCK(inp);
4821	}
4822	/* pull from vtag hash */
4823	LIST_REMOVE(stcb, sctp_asocs);
4824	sctp_add_vtag_to_timewait(asoc->my_vtag, SCTP_BASE_SYSCTL(sctp_vtag_time_wait),
4825	    inp->sctp_lport, stcb->rport);
4826
4827	/*
4828	 * Now restop the timers to be sure this is paranoia at is finest!
4829	 */
4830	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4831	(void)SCTP_OS_TIMER_STOP(&asoc->dack_timer.timer);
4832	(void)SCTP_OS_TIMER_STOP(&asoc->strreset_timer.timer);
4833	(void)SCTP_OS_TIMER_STOP(&asoc->asconf_timer.timer);
4834	(void)SCTP_OS_TIMER_STOP(&asoc->shut_guard_timer.timer);
4835	(void)SCTP_OS_TIMER_STOP(&asoc->autoclose_timer.timer);
4836	(void)SCTP_OS_TIMER_STOP(&asoc->delayed_event_timer.timer);
4837	TAILQ_FOREACH(net, &asoc->nets, sctp_next) {
4838		(void)SCTP_OS_TIMER_STOP(&net->rxt_timer.timer);
4839		(void)SCTP_OS_TIMER_STOP(&net->pmtu_timer.timer);
4840		(void)SCTP_OS_TIMER_STOP(&net->hb_timer.timer);
4841	}
4842
4843	asoc->strreset_timer.type = SCTP_TIMER_TYPE_NONE;
4844	/*
4845	 * The chunk lists and such SHOULD be empty but we check them just
4846	 * in case.
4847	 */
4848	/* anything on the wheel needs to be removed */
4849	for (i = 0; i < asoc->streamoutcnt; i++) {
4850		struct sctp_stream_out *outs;
4851
4852		outs = &asoc->strmout[i];
4853		/* now clean up any chunks here */
4854		TAILQ_FOREACH_SAFE(sp, &outs->outqueue, next, nsp) {
4855			TAILQ_REMOVE(&outs->outqueue, sp, next);
4856			sctp_free_spbufspace(stcb, asoc, sp);
4857			if (sp->data) {
4858				if (so) {
4859					/* Still an open socket - report */
4860					sctp_ulp_notify(SCTP_NOTIFY_SPECIAL_SP_FAIL, stcb,
4861					    0, (void *)sp, SCTP_SO_LOCKED);
4862				}
4863				if (sp->data) {
4864					sctp_m_freem(sp->data);
4865					sp->data = NULL;
4866					sp->tail_mbuf = NULL;
4867					sp->length = 0;
4868				}
4869			}
4870			if (sp->net) {
4871				sctp_free_remote_addr(sp->net);
4872				sp->net = NULL;
4873			}
4874			sctp_free_a_strmoq(stcb, sp, SCTP_SO_LOCKED);
4875		}
4876	}
4877	/* sa_ignore FREED_MEMORY */
4878	TAILQ_FOREACH_SAFE(strrst, &asoc->resetHead, next_resp, nstrrst) {
4879		TAILQ_REMOVE(&asoc->resetHead, strrst, next_resp);
4880		SCTP_FREE(strrst, SCTP_M_STRESET);
4881	}
4882	TAILQ_FOREACH_SAFE(sq, &asoc->pending_reply_queue, next, nsq) {
4883		TAILQ_REMOVE(&asoc->pending_reply_queue, sq, next);
4884		if (sq->data) {
4885			sctp_m_freem(sq->data);
4886			sq->data = NULL;
4887		}
4888		sctp_free_remote_addr(sq->whoFrom);
4889		sq->whoFrom = NULL;
4890		sq->stcb = NULL;
4891		/* Free the ctl entry */
4892		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), sq);
4893		SCTP_DECR_READQ_COUNT();
4894		/* sa_ignore FREED_MEMORY */
4895	}
4896	TAILQ_FOREACH_SAFE(chk, &asoc->free_chunks, sctp_next, nchk) {
4897		TAILQ_REMOVE(&asoc->free_chunks, chk, sctp_next);
4898		if (chk->data) {
4899			sctp_m_freem(chk->data);
4900			chk->data = NULL;
4901		}
4902		if (chk->holds_key_ref)
4903			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4904		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4905		SCTP_DECR_CHK_COUNT();
4906		atomic_subtract_int(&SCTP_BASE_INFO(ipi_free_chunks), 1);
4907		asoc->free_chunk_cnt--;
4908		/* sa_ignore FREED_MEMORY */
4909	}
4910	/* pending send queue SHOULD be empty */
4911	TAILQ_FOREACH_SAFE(chk, &asoc->send_queue, sctp_next, nchk) {
4912		if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4913			asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4914#ifdef INVARIANTS
4915		} else {
4916			panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4917#endif
4918		}
4919		TAILQ_REMOVE(&asoc->send_queue, chk, sctp_next);
4920		if (chk->data) {
4921			if (so) {
4922				/* Still a socket? */
4923				sctp_ulp_notify(SCTP_NOTIFY_UNSENT_DG_FAIL, stcb,
4924				    0, chk, SCTP_SO_LOCKED);
4925			}
4926			if (chk->data) {
4927				sctp_m_freem(chk->data);
4928				chk->data = NULL;
4929			}
4930		}
4931		if (chk->holds_key_ref)
4932			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4933		if (chk->whoTo) {
4934			sctp_free_remote_addr(chk->whoTo);
4935			chk->whoTo = NULL;
4936		}
4937		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4938		SCTP_DECR_CHK_COUNT();
4939		/* sa_ignore FREED_MEMORY */
4940	}
4941	/* sent queue SHOULD be empty */
4942	TAILQ_FOREACH_SAFE(chk, &asoc->sent_queue, sctp_next, nchk) {
4943		if (chk->sent != SCTP_DATAGRAM_NR_ACKED) {
4944			if (asoc->strmout[chk->rec.data.stream_number].chunks_on_queues > 0) {
4945				asoc->strmout[chk->rec.data.stream_number].chunks_on_queues--;
4946#ifdef INVARIANTS
4947			} else {
4948				panic("No chunks on the queues for sid %u.", chk->rec.data.stream_number);
4949#endif
4950			}
4951		}
4952		TAILQ_REMOVE(&asoc->sent_queue, chk, sctp_next);
4953		if (chk->data) {
4954			if (so) {
4955				/* Still a socket? */
4956				sctp_ulp_notify(SCTP_NOTIFY_SENT_DG_FAIL, stcb,
4957				    0, chk, SCTP_SO_LOCKED);
4958			}
4959			if (chk->data) {
4960				sctp_m_freem(chk->data);
4961				chk->data = NULL;
4962			}
4963		}
4964		if (chk->holds_key_ref)
4965			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4966		sctp_free_remote_addr(chk->whoTo);
4967		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4968		SCTP_DECR_CHK_COUNT();
4969		/* sa_ignore FREED_MEMORY */
4970	}
4971#ifdef INVARIANTS
4972	for (i = 0; i < stcb->asoc.streamoutcnt; i++) {
4973		if (stcb->asoc.strmout[i].chunks_on_queues > 0) {
4974			panic("%u chunks left for stream %u.", stcb->asoc.strmout[i].chunks_on_queues, i);
4975		}
4976	}
4977#endif
4978	/* control queue MAY not be empty */
4979	TAILQ_FOREACH_SAFE(chk, &asoc->control_send_queue, sctp_next, nchk) {
4980		TAILQ_REMOVE(&asoc->control_send_queue, chk, sctp_next);
4981		if (chk->data) {
4982			sctp_m_freem(chk->data);
4983			chk->data = NULL;
4984		}
4985		if (chk->holds_key_ref)
4986			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
4987		sctp_free_remote_addr(chk->whoTo);
4988		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
4989		SCTP_DECR_CHK_COUNT();
4990		/* sa_ignore FREED_MEMORY */
4991	}
4992	/* ASCONF queue MAY not be empty */
4993	TAILQ_FOREACH_SAFE(chk, &asoc->asconf_send_queue, sctp_next, nchk) {
4994		TAILQ_REMOVE(&asoc->asconf_send_queue, chk, sctp_next);
4995		if (chk->data) {
4996			sctp_m_freem(chk->data);
4997			chk->data = NULL;
4998		}
4999		if (chk->holds_key_ref)
5000			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5001		sctp_free_remote_addr(chk->whoTo);
5002		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5003		SCTP_DECR_CHK_COUNT();
5004		/* sa_ignore FREED_MEMORY */
5005	}
5006	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
5007		TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
5008		if (chk->data) {
5009			sctp_m_freem(chk->data);
5010			chk->data = NULL;
5011		}
5012		if (chk->holds_key_ref)
5013			sctp_auth_key_release(stcb, chk->auth_keyid, SCTP_SO_LOCKED);
5014		sctp_free_remote_addr(chk->whoTo);
5015		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_chunk), chk);
5016		SCTP_DECR_CHK_COUNT();
5017		/* sa_ignore FREED_MEMORY */
5018	}
5019
5020	if (asoc->mapping_array) {
5021		SCTP_FREE(asoc->mapping_array, SCTP_M_MAP);
5022		asoc->mapping_array = NULL;
5023	}
5024	if (asoc->nr_mapping_array) {
5025		SCTP_FREE(asoc->nr_mapping_array, SCTP_M_MAP);
5026		asoc->nr_mapping_array = NULL;
5027	}
5028	/* the stream outs */
5029	if (asoc->strmout) {
5030		SCTP_FREE(asoc->strmout, SCTP_M_STRMO);
5031		asoc->strmout = NULL;
5032	}
5033	asoc->strm_realoutsize = asoc->streamoutcnt = 0;
5034	if (asoc->strmin) {
5035		struct sctp_queued_to_read *ctl, *nctl;
5036
5037		for (i = 0; i < asoc->streamincnt; i++) {
5038			TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[i].inqueue, next, nctl) {
5039				TAILQ_REMOVE(&asoc->strmin[i].inqueue, ctl, next);
5040				sctp_free_remote_addr(ctl->whoFrom);
5041				if (ctl->data) {
5042					sctp_m_freem(ctl->data);
5043					ctl->data = NULL;
5044				}
5045				/*
5046				 * We don't free the address here since all
5047				 * the net's were freed above.
5048				 */
5049				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
5050				SCTP_DECR_READQ_COUNT();
5051			}
5052		}
5053		SCTP_FREE(asoc->strmin, SCTP_M_STRMI);
5054		asoc->strmin = NULL;
5055	}
5056	asoc->streamincnt = 0;
5057	TAILQ_FOREACH_SAFE(net, &asoc->nets, sctp_next, nnet) {
5058#ifdef INVARIANTS
5059		if (SCTP_BASE_INFO(ipi_count_raddr) == 0) {
5060			panic("no net's left alloc'ed, or list points to itself");
5061		}
5062#endif
5063		TAILQ_REMOVE(&asoc->nets, net, sctp_next);
5064		sctp_free_remote_addr(net);
5065	}
5066	LIST_FOREACH_SAFE(laddr, &asoc->sctp_restricted_addrs, sctp_nxt_addr, naddr) {
5067		/* sa_ignore FREED_MEMORY */
5068		sctp_remove_laddr(laddr);
5069	}
5070
5071	/* pending asconf (address) parameters */
5072	TAILQ_FOREACH_SAFE(aparam, &asoc->asconf_queue, next, naparam) {
5073		/* sa_ignore FREED_MEMORY */
5074		TAILQ_REMOVE(&asoc->asconf_queue, aparam, next);
5075		SCTP_FREE(aparam, SCTP_M_ASC_ADDR);
5076	}
5077	TAILQ_FOREACH_SAFE(aack, &asoc->asconf_ack_sent, next, naack) {
5078		/* sa_ignore FREED_MEMORY */
5079		TAILQ_REMOVE(&asoc->asconf_ack_sent, aack, next);
5080		if (aack->data != NULL) {
5081			sctp_m_freem(aack->data);
5082		}
5083		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asconf_ack), aack);
5084	}
5085	/* clean up auth stuff */
5086	if (asoc->local_hmacs)
5087		sctp_free_hmaclist(asoc->local_hmacs);
5088	if (asoc->peer_hmacs)
5089		sctp_free_hmaclist(asoc->peer_hmacs);
5090
5091	if (asoc->local_auth_chunks)
5092		sctp_free_chunklist(asoc->local_auth_chunks);
5093	if (asoc->peer_auth_chunks)
5094		sctp_free_chunklist(asoc->peer_auth_chunks);
5095
5096	sctp_free_authinfo(&asoc->authinfo);
5097
5098	LIST_FOREACH_SAFE(shared_key, &asoc->shared_keys, next, nshared_key) {
5099		LIST_REMOVE(shared_key, next);
5100		sctp_free_sharedkey(shared_key);
5101		/* sa_ignore FREED_MEMORY */
5102	}
5103
5104	/* Insert new items here :> */
5105
5106	/* Get rid of LOCK */
5107	SCTP_TCB_UNLOCK(stcb);
5108	SCTP_TCB_LOCK_DESTROY(stcb);
5109	SCTP_TCB_SEND_LOCK_DESTROY(stcb);
5110	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5111		SCTP_INP_INFO_WUNLOCK();
5112		SCTP_INP_RLOCK(inp);
5113	}
5114#ifdef SCTP_TRACK_FREED_ASOCS
5115	if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5116		/* now clean up the tasoc itself */
5117		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5118		SCTP_DECR_ASOC_COUNT();
5119	} else {
5120		LIST_INSERT_HEAD(&inp->sctp_asoc_free_list, stcb, sctp_tcblist);
5121	}
5122#else
5123	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_asoc), stcb);
5124	SCTP_DECR_ASOC_COUNT();
5125#endif
5126	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5127		if (inp->sctp_flags & SCTP_PCB_FLAGS_SOCKET_GONE) {
5128			/*
5129			 * If its NOT the inp_free calling us AND sctp_close
5130			 * as been called, we call back...
5131			 */
5132			SCTP_INP_RUNLOCK(inp);
5133			/*
5134			 * This will start the kill timer (if we are the
5135			 * last one) since we hold an increment yet. But
5136			 * this is the only safe way to do this since
5137			 * otherwise if the socket closes at the same time
5138			 * we are here we might collide in the cleanup.
5139			 */
5140			sctp_inpcb_free(inp,
5141			    SCTP_FREE_SHOULD_USE_GRACEFUL_CLOSE,
5142			    SCTP_CALLED_DIRECTLY_NOCMPSET);
5143			SCTP_INP_DECR_REF(inp);
5144			goto out_of;
5145		} else {
5146			/* The socket is still open. */
5147			SCTP_INP_DECR_REF(inp);
5148		}
5149	}
5150	if (from_inpcbfree == SCTP_NORMAL_PROC) {
5151		SCTP_INP_RUNLOCK(inp);
5152	}
5153out_of:
5154	/* destroyed the asoc */
5155#ifdef SCTP_LOG_CLOSING
5156	sctp_log_closing(inp, NULL, 11);
5157#endif
5158	return (1);
5159}
5160
5161
5162
5163/*
5164 * determine if a destination is "reachable" based upon the addresses bound
5165 * to the current endpoint (e.g. only v4 or v6 currently bound)
5166 */
5167/*
5168 * FIX: if we allow assoc-level bindx(), then this needs to be fixed to use
5169 * assoc level v4/v6 flags, as the assoc *may* not have the same address
5170 * types bound as its endpoint
5171 */
5172int
5173sctp_destination_is_reachable(struct sctp_tcb *stcb, struct sockaddr *destaddr)
5174{
5175	struct sctp_inpcb *inp;
5176	int answer;
5177
5178	/*
5179	 * No locks here, the TCB, in all cases is already locked and an
5180	 * assoc is up. There is either a INP lock by the caller applied (in
5181	 * asconf case when deleting an address) or NOT in the HB case,
5182	 * however if HB then the INP increment is up and the INP will not
5183	 * be removed (on top of the fact that we have a TCB lock). So we
5184	 * only want to read the sctp_flags, which is either bound-all or
5185	 * not.. no protection needed since once an assoc is up you can't be
5186	 * changing your binding.
5187	 */
5188	inp = stcb->sctp_ep;
5189	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5190		/* if bound all, destination is not restricted */
5191		/*
5192		 * RRS: Question during lock work: Is this correct? If you
5193		 * are bound-all you still might need to obey the V4--V6
5194		 * flags??? IMO this bound-all stuff needs to be removed!
5195		 */
5196		return (1);
5197	}
5198	/* NOTE: all "scope" checks are done when local addresses are added */
5199	switch (destaddr->sa_family) {
5200#ifdef INET6
5201	case AF_INET6:
5202		answer = inp->ip_inp.inp.inp_vflag & INP_IPV6;
5203		break;
5204#endif
5205#ifdef INET
5206	case AF_INET:
5207		answer = inp->ip_inp.inp.inp_vflag & INP_IPV4;
5208		break;
5209#endif
5210	default:
5211		/* invalid family, so it's unreachable */
5212		answer = 0;
5213		break;
5214	}
5215	return (answer);
5216}
5217
5218/*
5219 * update the inp_vflags on an endpoint
5220 */
5221static void
5222sctp_update_ep_vflag(struct sctp_inpcb *inp)
5223{
5224	struct sctp_laddr *laddr;
5225
5226	/* first clear the flag */
5227	inp->ip_inp.inp.inp_vflag = 0;
5228	/* set the flag based on addresses on the ep list */
5229	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5230		if (laddr->ifa == NULL) {
5231			SCTPDBG(SCTP_DEBUG_PCB1, "%s: NULL ifa\n",
5232			    __FUNCTION__);
5233			continue;
5234		}
5235		if (laddr->ifa->localifa_flags & SCTP_BEING_DELETED) {
5236			continue;
5237		}
5238		switch (laddr->ifa->address.sa.sa_family) {
5239#ifdef INET6
5240		case AF_INET6:
5241			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5242			break;
5243#endif
5244#ifdef INET
5245		case AF_INET:
5246			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5247			break;
5248#endif
5249		default:
5250			break;
5251		}
5252	}
5253}
5254
5255/*
5256 * Add the address to the endpoint local address list There is nothing to be
5257 * done if we are bound to all addresses
5258 */
5259void
5260sctp_add_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa, uint32_t action)
5261{
5262	struct sctp_laddr *laddr;
5263	int fnd, error = 0;
5264
5265	fnd = 0;
5266
5267	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5268		/* You are already bound to all. You have it already */
5269		return;
5270	}
5271#ifdef INET6
5272	if (ifa->address.sa.sa_family == AF_INET6) {
5273		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5274			/* Can't bind a non-useable addr. */
5275			return;
5276		}
5277	}
5278#endif
5279	/* first, is it already present? */
5280	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5281		if (laddr->ifa == ifa) {
5282			fnd = 1;
5283			break;
5284		}
5285	}
5286
5287	if (fnd == 0) {
5288		/* Not in the ep list */
5289		error = sctp_insert_laddr(&inp->sctp_addr_list, ifa, action);
5290		if (error != 0)
5291			return;
5292		inp->laddr_count++;
5293		/* update inp_vflag flags */
5294		switch (ifa->address.sa.sa_family) {
5295#ifdef INET6
5296		case AF_INET6:
5297			inp->ip_inp.inp.inp_vflag |= INP_IPV6;
5298			break;
5299#endif
5300#ifdef INET
5301		case AF_INET:
5302			inp->ip_inp.inp.inp_vflag |= INP_IPV4;
5303			break;
5304#endif
5305		default:
5306			break;
5307		}
5308	}
5309	return;
5310}
5311
5312
5313/*
5314 * select a new (hopefully reachable) destination net (should only be used
5315 * when we deleted an ep addr that is the only usable source address to reach
5316 * the destination net)
5317 */
5318static void
5319sctp_select_primary_destination(struct sctp_tcb *stcb)
5320{
5321	struct sctp_nets *net;
5322
5323	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5324		/* for now, we'll just pick the first reachable one we find */
5325		if (net->dest_state & SCTP_ADDR_UNCONFIRMED)
5326			continue;
5327		if (sctp_destination_is_reachable(stcb,
5328		    (struct sockaddr *)&net->ro._l_addr)) {
5329			/* found a reachable destination */
5330			stcb->asoc.primary_destination = net;
5331		}
5332	}
5333	/* I can't there from here! ...we're gonna die shortly... */
5334}
5335
5336
5337/*
5338 * Delete the address from the endpoint local address list There is nothing
5339 * to be done if we are bound to all addresses
5340 */
5341void
5342sctp_del_local_addr_ep(struct sctp_inpcb *inp, struct sctp_ifa *ifa)
5343{
5344	struct sctp_laddr *laddr;
5345	int fnd;
5346
5347	fnd = 0;
5348	if (inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) {
5349		/* You are already bound to all. You have it already */
5350		return;
5351	}
5352	LIST_FOREACH(laddr, &inp->sctp_addr_list, sctp_nxt_addr) {
5353		if (laddr->ifa == ifa) {
5354			fnd = 1;
5355			break;
5356		}
5357	}
5358	if (fnd && (inp->laddr_count < 2)) {
5359		/* can't delete unless there are at LEAST 2 addresses */
5360		return;
5361	}
5362	if (fnd) {
5363		/*
5364		 * clean up any use of this address go through our
5365		 * associations and clear any last_used_address that match
5366		 * this one for each assoc, see if a new primary_destination
5367		 * is needed
5368		 */
5369		struct sctp_tcb *stcb;
5370
5371		/* clean up "next_addr_touse" */
5372		if (inp->next_addr_touse == laddr)
5373			/* delete this address */
5374			inp->next_addr_touse = NULL;
5375
5376		/* clean up "last_used_address" */
5377		LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
5378			struct sctp_nets *net;
5379
5380			SCTP_TCB_LOCK(stcb);
5381			if (stcb->asoc.last_used_address == laddr)
5382				/* delete this address */
5383				stcb->asoc.last_used_address = NULL;
5384			/*
5385			 * Now spin through all the nets and purge any ref
5386			 * to laddr
5387			 */
5388			TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
5389				if (net->ro._s_addr &&
5390				    (net->ro._s_addr->ifa == laddr->ifa)) {
5391					/* Yep, purge src address selected */
5392					sctp_rtentry_t *rt;
5393
5394					/* delete this address if cached */
5395					rt = net->ro.ro_rt;
5396					if (rt != NULL) {
5397						RTFREE(rt);
5398						net->ro.ro_rt = NULL;
5399					}
5400					sctp_free_ifa(net->ro._s_addr);
5401					net->ro._s_addr = NULL;
5402					net->src_addr_selected = 0;
5403				}
5404			}
5405			SCTP_TCB_UNLOCK(stcb);
5406		}		/* for each tcb */
5407		/* remove it from the ep list */
5408		sctp_remove_laddr(laddr);
5409		inp->laddr_count--;
5410		/* update inp_vflag flags */
5411		sctp_update_ep_vflag(inp);
5412	}
5413	return;
5414}
5415
5416/*
5417 * Add the address to the TCB local address restricted list.
5418 * This is a "pending" address list (eg. addresses waiting for an
5419 * ASCONF-ACK response) and cannot be used as a valid source address.
5420 */
5421void
5422sctp_add_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5423{
5424	struct sctp_laddr *laddr;
5425	struct sctpladdr *list;
5426
5427	/*
5428	 * Assumes TCB is locked.. and possibly the INP. May need to
5429	 * confirm/fix that if we need it and is not the case.
5430	 */
5431	list = &stcb->asoc.sctp_restricted_addrs;
5432
5433#ifdef INET6
5434	if (ifa->address.sa.sa_family == AF_INET6) {
5435		if (ifa->localifa_flags & SCTP_ADDR_IFA_UNUSEABLE) {
5436			/* Can't bind a non-existent addr. */
5437			return;
5438		}
5439	}
5440#endif
5441	/* does the address already exist? */
5442	LIST_FOREACH(laddr, list, sctp_nxt_addr) {
5443		if (laddr->ifa == ifa) {
5444			return;
5445		}
5446	}
5447
5448	/* add to the list */
5449	(void)sctp_insert_laddr(list, ifa, 0);
5450	return;
5451}
5452
5453/*
5454 * insert an laddr entry with the given ifa for the desired list
5455 */
5456int
5457sctp_insert_laddr(struct sctpladdr *list, struct sctp_ifa *ifa, uint32_t act)
5458{
5459	struct sctp_laddr *laddr;
5460
5461	laddr = SCTP_ZONE_GET(SCTP_BASE_INFO(ipi_zone_laddr), struct sctp_laddr);
5462	if (laddr == NULL) {
5463		/* out of memory? */
5464		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, EINVAL);
5465		return (EINVAL);
5466	}
5467	SCTP_INCR_LADDR_COUNT();
5468	bzero(laddr, sizeof(*laddr));
5469	(void)SCTP_GETTIME_TIMEVAL(&laddr->start_time);
5470	laddr->ifa = ifa;
5471	laddr->action = act;
5472	atomic_add_int(&ifa->refcount, 1);
5473	/* insert it */
5474	LIST_INSERT_HEAD(list, laddr, sctp_nxt_addr);
5475
5476	return (0);
5477}
5478
5479/*
5480 * Remove an laddr entry from the local address list (on an assoc)
5481 */
5482void
5483sctp_remove_laddr(struct sctp_laddr *laddr)
5484{
5485
5486	/* remove from the list */
5487	LIST_REMOVE(laddr, sctp_nxt_addr);
5488	sctp_free_ifa(laddr->ifa);
5489	SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), laddr);
5490	SCTP_DECR_LADDR_COUNT();
5491}
5492
5493/*
5494 * Remove a local address from the TCB local address restricted list
5495 */
5496void
5497sctp_del_local_addr_restricted(struct sctp_tcb *stcb, struct sctp_ifa *ifa)
5498{
5499	struct sctp_inpcb *inp;
5500	struct sctp_laddr *laddr;
5501
5502	/*
5503	 * This is called by asconf work. It is assumed that a) The TCB is
5504	 * locked and b) The INP is locked. This is true in as much as I can
5505	 * trace through the entry asconf code where I did these locks.
5506	 * Again, the ASCONF code is a bit different in that it does lock
5507	 * the INP during its work often times. This must be since we don't
5508	 * want other proc's looking up things while what they are looking
5509	 * up is changing :-D
5510	 */
5511
5512	inp = stcb->sctp_ep;
5513	/* if subset bound and don't allow ASCONF's, can't delete last */
5514	if (((inp->sctp_flags & SCTP_PCB_FLAGS_BOUNDALL) == 0) &&
5515	    sctp_is_feature_off(inp, SCTP_PCB_FLAGS_DO_ASCONF)) {
5516		if (stcb->sctp_ep->laddr_count < 2) {
5517			/* can't delete last address */
5518			return;
5519		}
5520	}
5521	LIST_FOREACH(laddr, &stcb->asoc.sctp_restricted_addrs, sctp_nxt_addr) {
5522		/* remove the address if it exists */
5523		if (laddr->ifa == NULL)
5524			continue;
5525		if (laddr->ifa == ifa) {
5526			sctp_remove_laddr(laddr);
5527			return;
5528		}
5529	}
5530
5531	/* address not found! */
5532	return;
5533}
5534
5535/*
5536 * Temporarily remove for __APPLE__ until we use the Tiger equivalents
5537 */
5538/* sysctl */
5539static int sctp_max_number_of_assoc = SCTP_MAX_NUM_OF_ASOC;
5540static int sctp_scale_up_for_address = SCTP_SCALE_FOR_ADDR;
5541
5542
5543
5544#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5545struct sctp_mcore_ctrl *sctp_mcore_workers = NULL;
5546int *sctp_cpuarry = NULL;
5547void
5548sctp_queue_to_mcore(struct mbuf *m, int off, int cpu_to_use)
5549{
5550	/* Queue a packet to a processor for the specified core */
5551	struct sctp_mcore_queue *qent;
5552	struct sctp_mcore_ctrl *wkq;
5553	int need_wake = 0;
5554
5555	if (sctp_mcore_workers == NULL) {
5556		/* Something went way bad during setup */
5557		sctp_input_with_port(m, off, 0);
5558		return;
5559	}
5560	SCTP_MALLOC(qent, struct sctp_mcore_queue *,
5561	    (sizeof(struct sctp_mcore_queue)),
5562	    SCTP_M_MCORE);
5563	if (qent == NULL) {
5564		/* This is trouble  */
5565		sctp_input_with_port(m, off, 0);
5566		return;
5567	}
5568	qent->vn = curvnet;
5569	qent->m = m;
5570	qent->off = off;
5571	qent->v6 = 0;
5572	wkq = &sctp_mcore_workers[cpu_to_use];
5573	SCTP_MCORE_QLOCK(wkq);
5574
5575	TAILQ_INSERT_TAIL(&wkq->que, qent, next);
5576	if (wkq->running == 0) {
5577		need_wake = 1;
5578	}
5579	SCTP_MCORE_QUNLOCK(wkq);
5580	if (need_wake) {
5581		wakeup(&wkq->running);
5582	}
5583}
5584
5585static void
5586sctp_mcore_thread(void *arg)
5587{
5588
5589	struct sctp_mcore_ctrl *wkq;
5590	struct sctp_mcore_queue *qent;
5591
5592	wkq = (struct sctp_mcore_ctrl *)arg;
5593	struct mbuf *m;
5594	int off, v6;
5595
5596	/* Wait for first tickle */
5597	SCTP_MCORE_LOCK(wkq);
5598	wkq->running = 0;
5599	msleep(&wkq->running,
5600	    &wkq->core_mtx,
5601	    0, "wait for pkt", 0);
5602	SCTP_MCORE_UNLOCK(wkq);
5603
5604	/* Bind to our cpu */
5605	thread_lock(curthread);
5606	sched_bind(curthread, wkq->cpuid);
5607	thread_unlock(curthread);
5608
5609	/* Now lets start working */
5610	SCTP_MCORE_LOCK(wkq);
5611	/* Now grab lock and go */
5612	for (;;) {
5613		SCTP_MCORE_QLOCK(wkq);
5614skip_sleep:
5615		wkq->running = 1;
5616		qent = TAILQ_FIRST(&wkq->que);
5617		if (qent) {
5618			TAILQ_REMOVE(&wkq->que, qent, next);
5619			SCTP_MCORE_QUNLOCK(wkq);
5620			CURVNET_SET(qent->vn);
5621			m = qent->m;
5622			off = qent->off;
5623			v6 = qent->v6;
5624			SCTP_FREE(qent, SCTP_M_MCORE);
5625			if (v6 == 0) {
5626				sctp_input_with_port(m, off, 0);
5627			} else {
5628				SCTP_PRINTF("V6 not yet supported\n");
5629				sctp_m_freem(m);
5630			}
5631			CURVNET_RESTORE();
5632			SCTP_MCORE_QLOCK(wkq);
5633		}
5634		wkq->running = 0;
5635		if (!TAILQ_EMPTY(&wkq->que)) {
5636			goto skip_sleep;
5637		}
5638		SCTP_MCORE_QUNLOCK(wkq);
5639		msleep(&wkq->running,
5640		    &wkq->core_mtx,
5641		    0, "wait for pkt", 0);
5642	}
5643}
5644
5645static void
5646sctp_startup_mcore_threads(void)
5647{
5648	int i, cpu;
5649
5650	if (mp_ncpus == 1)
5651		return;
5652
5653	if (sctp_mcore_workers != NULL) {
5654		/*
5655		 * Already been here in some previous vnet?
5656		 */
5657		return;
5658	}
5659	SCTP_MALLOC(sctp_mcore_workers, struct sctp_mcore_ctrl *,
5660	    ((mp_maxid + 1) * sizeof(struct sctp_mcore_ctrl)),
5661	    SCTP_M_MCORE);
5662	if (sctp_mcore_workers == NULL) {
5663		/* TSNH I hope */
5664		return;
5665	}
5666	memset(sctp_mcore_workers, 0, ((mp_maxid + 1) *
5667	    sizeof(struct sctp_mcore_ctrl)));
5668	/* Init the structures */
5669	for (i = 0; i <= mp_maxid; i++) {
5670		TAILQ_INIT(&sctp_mcore_workers[i].que);
5671		SCTP_MCORE_LOCK_INIT(&sctp_mcore_workers[i]);
5672		SCTP_MCORE_QLOCK_INIT(&sctp_mcore_workers[i]);
5673		sctp_mcore_workers[i].cpuid = i;
5674	}
5675	if (sctp_cpuarry == NULL) {
5676		SCTP_MALLOC(sctp_cpuarry, int *,
5677		    (mp_ncpus * sizeof(int)),
5678		    SCTP_M_MCORE);
5679		i = 0;
5680		CPU_FOREACH(cpu) {
5681			sctp_cpuarry[i] = cpu;
5682			i++;
5683		}
5684	}
5685	/* Now start them all */
5686	CPU_FOREACH(cpu) {
5687		(void)kproc_create(sctp_mcore_thread,
5688		    (void *)&sctp_mcore_workers[cpu],
5689		    &sctp_mcore_workers[cpu].thread_proc,
5690		    RFPROC,
5691		    SCTP_KTHREAD_PAGES,
5692		    SCTP_MCORE_NAME);
5693
5694	}
5695}
5696
5697#endif
5698
5699void
5700sctp_pcb_init()
5701{
5702	/*
5703	 * SCTP initialization for the PCB structures should be called by
5704	 * the sctp_init() funciton.
5705	 */
5706	int i;
5707	struct timeval tv;
5708
5709	if (SCTP_BASE_VAR(sctp_pcb_initialized) != 0) {
5710		/* error I was called twice */
5711		return;
5712	}
5713	SCTP_BASE_VAR(sctp_pcb_initialized) = 1;
5714
5715#if defined(SCTP_LOCAL_TRACE_BUF)
5716	bzero(&SCTP_BASE_SYSCTL(sctp_log), sizeof(struct sctp_log));
5717#endif
5718#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5719	SCTP_MALLOC(SCTP_BASE_STATS, struct sctpstat *,
5720	    ((mp_maxid + 1) * sizeof(struct sctpstat)),
5721	    SCTP_M_MCORE);
5722#endif
5723	(void)SCTP_GETTIME_TIMEVAL(&tv);
5724#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5725	bzero(SCTP_BASE_STATS, (sizeof(struct sctpstat) * (mp_maxid + 1)));
5726	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_sec = (uint32_t) tv.tv_sec;
5727	SCTP_BASE_STATS[PCPU_GET(cpuid)].sctps_discontinuitytime.tv_usec = (uint32_t) tv.tv_usec;
5728#else
5729	bzero(&SCTP_BASE_STATS, sizeof(struct sctpstat));
5730	SCTP_BASE_STAT(sctps_discontinuitytime).tv_sec = (uint32_t) tv.tv_sec;
5731	SCTP_BASE_STAT(sctps_discontinuitytime).tv_usec = (uint32_t) tv.tv_usec;
5732#endif
5733	/* init the empty list of (All) Endpoints */
5734	LIST_INIT(&SCTP_BASE_INFO(listhead));
5735
5736
5737	/* init the hash table of endpoints */
5738	TUNABLE_INT_FETCH("net.inet.sctp.tcbhashsize", &SCTP_BASE_SYSCTL(sctp_hashtblsize));
5739	TUNABLE_INT_FETCH("net.inet.sctp.pcbhashsize", &SCTP_BASE_SYSCTL(sctp_pcbtblsize));
5740	TUNABLE_INT_FETCH("net.inet.sctp.chunkscale", &SCTP_BASE_SYSCTL(sctp_chunkscale));
5741	SCTP_BASE_INFO(sctp_asochash) = SCTP_HASH_INIT((SCTP_BASE_SYSCTL(sctp_hashtblsize) * 31),
5742	    &SCTP_BASE_INFO(hashasocmark));
5743	SCTP_BASE_INFO(sctp_ephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5744	    &SCTP_BASE_INFO(hashmark));
5745	SCTP_BASE_INFO(sctp_tcpephash) = SCTP_HASH_INIT(SCTP_BASE_SYSCTL(sctp_hashtblsize),
5746	    &SCTP_BASE_INFO(hashtcpmark));
5747	SCTP_BASE_INFO(hashtblsize) = SCTP_BASE_SYSCTL(sctp_hashtblsize);
5748
5749
5750	SCTP_BASE_INFO(sctp_vrfhash) = SCTP_HASH_INIT(SCTP_SIZE_OF_VRF_HASH,
5751	    &SCTP_BASE_INFO(hashvrfmark));
5752
5753	SCTP_BASE_INFO(vrf_ifn_hash) = SCTP_HASH_INIT(SCTP_VRF_IFN_HASH_SIZE,
5754	    &SCTP_BASE_INFO(vrf_ifn_hashmark));
5755	/* init the zones */
5756	/*
5757	 * FIX ME: Should check for NULL returns, but if it does fail we are
5758	 * doomed to panic anyways... add later maybe.
5759	 */
5760	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_ep), "sctp_ep",
5761	    sizeof(struct sctp_inpcb), maxsockets);
5762
5763	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asoc), "sctp_asoc",
5764	    sizeof(struct sctp_tcb), sctp_max_number_of_assoc);
5765
5766	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_laddr), "sctp_laddr",
5767	    sizeof(struct sctp_laddr),
5768	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5769
5770	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_net), "sctp_raddr",
5771	    sizeof(struct sctp_nets),
5772	    (sctp_max_number_of_assoc * sctp_scale_up_for_address));
5773
5774	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_chunk), "sctp_chunk",
5775	    sizeof(struct sctp_tmit_chunk),
5776	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5777
5778	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_readq), "sctp_readq",
5779	    sizeof(struct sctp_queued_to_read),
5780	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5781
5782	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_strmoq), "sctp_stream_msg_out",
5783	    sizeof(struct sctp_stream_queue_pending),
5784	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5785
5786	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf), "sctp_asconf",
5787	    sizeof(struct sctp_asconf),
5788	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5789
5790	SCTP_ZONE_INIT(SCTP_BASE_INFO(ipi_zone_asconf_ack), "sctp_asconf_ack",
5791	    sizeof(struct sctp_asconf_ack),
5792	    (sctp_max_number_of_assoc * SCTP_BASE_SYSCTL(sctp_chunkscale)));
5793
5794
5795	/* Master Lock INIT for info structure */
5796	SCTP_INP_INFO_LOCK_INIT();
5797	SCTP_STATLOG_INIT_LOCK();
5798
5799	SCTP_IPI_COUNT_INIT();
5800	SCTP_IPI_ADDR_INIT();
5801#ifdef SCTP_PACKET_LOGGING
5802	SCTP_IP_PKTLOG_INIT();
5803#endif
5804	LIST_INIT(&SCTP_BASE_INFO(addr_wq));
5805
5806	SCTP_WQ_ADDR_INIT();
5807	/* not sure if we need all the counts */
5808	SCTP_BASE_INFO(ipi_count_ep) = 0;
5809	/* assoc/tcb zone info */
5810	SCTP_BASE_INFO(ipi_count_asoc) = 0;
5811	/* local addrlist zone info */
5812	SCTP_BASE_INFO(ipi_count_laddr) = 0;
5813	/* remote addrlist zone info */
5814	SCTP_BASE_INFO(ipi_count_raddr) = 0;
5815	/* chunk info */
5816	SCTP_BASE_INFO(ipi_count_chunk) = 0;
5817
5818	/* socket queue zone info */
5819	SCTP_BASE_INFO(ipi_count_readq) = 0;
5820
5821	/* stream out queue cont */
5822	SCTP_BASE_INFO(ipi_count_strmoq) = 0;
5823
5824	SCTP_BASE_INFO(ipi_free_strmoq) = 0;
5825	SCTP_BASE_INFO(ipi_free_chunks) = 0;
5826
5827	SCTP_OS_TIMER_INIT(&SCTP_BASE_INFO(addr_wq_timer.timer));
5828
5829	/* Init the TIMEWAIT list */
5830	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5831		LIST_INIT(&SCTP_BASE_INFO(vtag_timewait)[i]);
5832	}
5833	sctp_startup_iterator();
5834
5835#if defined(__FreeBSD__) && defined(SCTP_MCORE_INPUT) && defined(SMP)
5836	sctp_startup_mcore_threads();
5837#endif
5838
5839	/*
5840	 * INIT the default VRF which for BSD is the only one, other O/S's
5841	 * may have more. But initially they must start with one and then
5842	 * add the VRF's as addresses are added.
5843	 */
5844	sctp_init_vrf_list(SCTP_DEFAULT_VRF);
5845}
5846
5847/*
5848 * Assumes that the SCTP_BASE_INFO() lock is NOT held.
5849 */
5850void
5851sctp_pcb_finish(void)
5852{
5853	struct sctp_vrflist *vrf_bucket;
5854	struct sctp_vrf *vrf, *nvrf;
5855	struct sctp_ifn *ifn, *nifn;
5856	struct sctp_ifa *ifa, *nifa;
5857	struct sctpvtaghead *chain;
5858	struct sctp_tagblock *twait_block, *prev_twait_block;
5859	struct sctp_laddr *wi, *nwi;
5860	int i;
5861	struct sctp_iterator *it, *nit;
5862
5863	/*
5864	 * In FreeBSD the iterator thread never exits but we do clean up.
5865	 * The only way FreeBSD reaches here is if we have VRF's but we
5866	 * still add the ifdef to make it compile on old versions.
5867	 */
5868	SCTP_IPI_ITERATOR_WQ_LOCK();
5869	TAILQ_FOREACH_SAFE(it, &sctp_it_ctl.iteratorhead, sctp_nxt_itr, nit) {
5870		if (it->vn != curvnet) {
5871			continue;
5872		}
5873		TAILQ_REMOVE(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
5874		if (it->function_atend != NULL) {
5875			(*it->function_atend) (it->pointer, it->val);
5876		}
5877		SCTP_FREE(it, SCTP_M_ITER);
5878	}
5879	SCTP_IPI_ITERATOR_WQ_UNLOCK();
5880	SCTP_ITERATOR_LOCK();
5881	if ((sctp_it_ctl.cur_it) &&
5882	    (sctp_it_ctl.cur_it->vn == curvnet)) {
5883		sctp_it_ctl.iterator_flags |= SCTP_ITERATOR_STOP_CUR_IT;
5884	}
5885	SCTP_ITERATOR_UNLOCK();
5886	SCTP_OS_TIMER_STOP(&SCTP_BASE_INFO(addr_wq_timer.timer));
5887	SCTP_WQ_ADDR_LOCK();
5888	LIST_FOREACH_SAFE(wi, &SCTP_BASE_INFO(addr_wq), sctp_nxt_addr, nwi) {
5889		LIST_REMOVE(wi, sctp_nxt_addr);
5890		SCTP_DECR_LADDR_COUNT();
5891		SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_laddr), wi);
5892	}
5893	SCTP_WQ_ADDR_UNLOCK();
5894
5895	/*
5896	 * free the vrf/ifn/ifa lists and hashes (be sure address monitor is
5897	 * destroyed first).
5898	 */
5899	vrf_bucket = &SCTP_BASE_INFO(sctp_vrfhash)[(SCTP_DEFAULT_VRFID & SCTP_BASE_INFO(hashvrfmark))];
5900	LIST_FOREACH_SAFE(vrf, vrf_bucket, next_vrf, nvrf) {
5901		LIST_FOREACH_SAFE(ifn, &vrf->ifnlist, next_ifn, nifn) {
5902			LIST_FOREACH_SAFE(ifa, &ifn->ifalist, next_ifa, nifa) {
5903				/* free the ifa */
5904				LIST_REMOVE(ifa, next_bucket);
5905				LIST_REMOVE(ifa, next_ifa);
5906				SCTP_FREE(ifa, SCTP_M_IFA);
5907			}
5908			/* free the ifn */
5909			LIST_REMOVE(ifn, next_bucket);
5910			LIST_REMOVE(ifn, next_ifn);
5911			SCTP_FREE(ifn, SCTP_M_IFN);
5912		}
5913		SCTP_HASH_FREE(vrf->vrf_addr_hash, vrf->vrf_addr_hashmark);
5914		/* free the vrf */
5915		LIST_REMOVE(vrf, next_vrf);
5916		SCTP_FREE(vrf, SCTP_M_VRF);
5917	}
5918	/* free the vrf hashes */
5919	SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_vrfhash), SCTP_BASE_INFO(hashvrfmark));
5920	SCTP_HASH_FREE(SCTP_BASE_INFO(vrf_ifn_hash), SCTP_BASE_INFO(vrf_ifn_hashmark));
5921
5922	/*
5923	 * free the TIMEWAIT list elements malloc'd in the function
5924	 * sctp_add_vtag_to_timewait()...
5925	 */
5926	for (i = 0; i < SCTP_STACK_VTAG_HASH_SIZE; i++) {
5927		chain = &SCTP_BASE_INFO(vtag_timewait)[i];
5928		if (!LIST_EMPTY(chain)) {
5929			prev_twait_block = NULL;
5930			LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
5931				if (prev_twait_block) {
5932					SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5933				}
5934				prev_twait_block = twait_block;
5935			}
5936			SCTP_FREE(prev_twait_block, SCTP_M_TIMW);
5937		}
5938	}
5939
5940	/* free the locks and mutexes */
5941#ifdef SCTP_PACKET_LOGGING
5942	SCTP_IP_PKTLOG_DESTROY();
5943#endif
5944	SCTP_IPI_ADDR_DESTROY();
5945	SCTP_STATLOG_DESTROY();
5946	SCTP_INP_INFO_LOCK_DESTROY();
5947
5948	SCTP_WQ_ADDR_DESTROY();
5949
5950	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_ep));
5951	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asoc));
5952	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_laddr));
5953	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_net));
5954	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_chunk));
5955	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_readq));
5956	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_strmoq));
5957	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf));
5958	SCTP_ZONE_DESTROY(SCTP_BASE_INFO(ipi_zone_asconf_ack));
5959	/* Get rid of other stuff to */
5960	if (SCTP_BASE_INFO(sctp_asochash) != NULL)
5961		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_asochash), SCTP_BASE_INFO(hashasocmark));
5962	if (SCTP_BASE_INFO(sctp_ephash) != NULL)
5963		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_ephash), SCTP_BASE_INFO(hashmark));
5964	if (SCTP_BASE_INFO(sctp_tcpephash) != NULL)
5965		SCTP_HASH_FREE(SCTP_BASE_INFO(sctp_tcpephash), SCTP_BASE_INFO(hashtcpmark));
5966#if defined(__FreeBSD__) && defined(SMP) && defined(SCTP_USE_PERCPU_STAT)
5967	SCTP_FREE(SCTP_BASE_STATS, SCTP_M_MCORE);
5968#endif
5969}
5970
5971
5972int
5973sctp_load_addresses_from_init(struct sctp_tcb *stcb, struct mbuf *m,
5974    int offset, int limit,
5975    struct sockaddr *src, struct sockaddr *dst,
5976    struct sockaddr *altsa)
5977{
5978	/*
5979	 * grub through the INIT pulling addresses and loading them to the
5980	 * nets structure in the asoc. The from address in the mbuf should
5981	 * also be loaded (if it is not already). This routine can be called
5982	 * with either INIT or INIT-ACK's as long as the m points to the IP
5983	 * packet and the offset points to the beginning of the parameters.
5984	 */
5985	struct sctp_inpcb *inp;
5986	struct sctp_nets *net, *nnet, *net_tmp;
5987	struct sctp_paramhdr *phdr, parm_buf;
5988	struct sctp_tcb *stcb_tmp;
5989	uint16_t ptype, plen;
5990	struct sockaddr *sa;
5991	uint8_t random_store[SCTP_PARAM_BUFFER_SIZE];
5992	struct sctp_auth_random *p_random = NULL;
5993	uint16_t random_len = 0;
5994	uint8_t hmacs_store[SCTP_PARAM_BUFFER_SIZE];
5995	struct sctp_auth_hmac_algo *hmacs = NULL;
5996	uint16_t hmacs_len = 0;
5997	uint8_t saw_asconf = 0;
5998	uint8_t saw_asconf_ack = 0;
5999	uint8_t chunks_store[SCTP_PARAM_BUFFER_SIZE];
6000	struct sctp_auth_chunk_list *chunks = NULL;
6001	uint16_t num_chunks = 0;
6002	sctp_key_t *new_key;
6003	uint32_t keylen;
6004	int got_random = 0, got_hmacs = 0, got_chklist = 0;
6005	uint8_t ecn_allowed;
6006
6007#ifdef INET
6008	struct sockaddr_in sin;
6009
6010#endif
6011#ifdef INET6
6012	struct sockaddr_in6 sin6;
6013
6014#endif
6015
6016	/* First get the destination address setup too. */
6017#ifdef INET
6018	memset(&sin, 0, sizeof(sin));
6019	sin.sin_family = AF_INET;
6020	sin.sin_len = sizeof(sin);
6021	sin.sin_port = stcb->rport;
6022#endif
6023#ifdef INET6
6024	memset(&sin6, 0, sizeof(sin6));
6025	sin6.sin6_family = AF_INET6;
6026	sin6.sin6_len = sizeof(struct sockaddr_in6);
6027	sin6.sin6_port = stcb->rport;
6028#endif
6029	if (altsa) {
6030		sa = altsa;
6031	} else {
6032		sa = src;
6033	}
6034	/* Turn off ECN until we get through all params */
6035	ecn_allowed = 0;
6036	TAILQ_FOREACH(net, &stcb->asoc.nets, sctp_next) {
6037		/* mark all addresses that we have currently on the list */
6038		net->dest_state |= SCTP_ADDR_NOT_IN_ASSOC;
6039	}
6040	/* does the source address already exist? if so skip it */
6041	inp = stcb->sctp_ep;
6042	atomic_add_int(&stcb->asoc.refcnt, 1);
6043	stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net_tmp, dst, stcb);
6044	atomic_add_int(&stcb->asoc.refcnt, -1);
6045
6046	if ((stcb_tmp == NULL && inp == stcb->sctp_ep) || inp == NULL) {
6047		/* we must add the source address */
6048		/* no scope set here since we have a tcb already. */
6049		switch (sa->sa_family) {
6050#ifdef INET
6051		case AF_INET:
6052			if (stcb->asoc.scope.ipv4_addr_legal) {
6053				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_2)) {
6054					return (-1);
6055				}
6056			}
6057			break;
6058#endif
6059#ifdef INET6
6060		case AF_INET6:
6061			if (stcb->asoc.scope.ipv6_addr_legal) {
6062				if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_3)) {
6063					return (-2);
6064				}
6065			}
6066			break;
6067#endif
6068		default:
6069			break;
6070		}
6071	} else {
6072		if (net_tmp != NULL && stcb_tmp == stcb) {
6073			net_tmp->dest_state &= ~SCTP_ADDR_NOT_IN_ASSOC;
6074		} else if (stcb_tmp != stcb) {
6075			/* It belongs to another association? */
6076			if (stcb_tmp)
6077				SCTP_TCB_UNLOCK(stcb_tmp);
6078			return (-3);
6079		}
6080	}
6081	if (stcb->asoc.state == 0) {
6082		/* the assoc was freed? */
6083		return (-4);
6084	}
6085	/*
6086	 * peer must explicitly turn this on. This may have been initialized
6087	 * to be "on" in order to allow local addr changes while INIT's are
6088	 * in flight.
6089	 */
6090	stcb->asoc.peer_supports_asconf = 0;
6091	/* now we must go through each of the params. */
6092	phdr = sctp_get_next_param(m, offset, &parm_buf, sizeof(parm_buf));
6093	while (phdr) {
6094		ptype = ntohs(phdr->param_type);
6095		plen = ntohs(phdr->param_length);
6096		/*
6097		 * SCTP_PRINTF("ptype => %0x, plen => %d\n",
6098		 * (uint32_t)ptype, (int)plen);
6099		 */
6100		if (offset + plen > limit) {
6101			break;
6102		}
6103		if (plen == 0) {
6104			break;
6105		}
6106#ifdef INET
6107		if (ptype == SCTP_IPV4_ADDRESS) {
6108			if (stcb->asoc.scope.ipv4_addr_legal) {
6109				struct sctp_ipv4addr_param *p4, p4_buf;
6110
6111				/* ok get the v4 address and check/add */
6112				phdr = sctp_get_next_param(m, offset,
6113				    (struct sctp_paramhdr *)&p4_buf,
6114				    sizeof(p4_buf));
6115				if (plen != sizeof(struct sctp_ipv4addr_param) ||
6116				    phdr == NULL) {
6117					return (-5);
6118				}
6119				p4 = (struct sctp_ipv4addr_param *)phdr;
6120				sin.sin_addr.s_addr = p4->addr;
6121				if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) {
6122					/* Skip multi-cast addresses */
6123					goto next_param;
6124				}
6125				if ((sin.sin_addr.s_addr == INADDR_BROADCAST) ||
6126				    (sin.sin_addr.s_addr == INADDR_ANY)) {
6127					goto next_param;
6128				}
6129				sa = (struct sockaddr *)&sin;
6130				inp = stcb->sctp_ep;
6131				atomic_add_int(&stcb->asoc.refcnt, 1);
6132				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6133				    dst, stcb);
6134				atomic_add_int(&stcb->asoc.refcnt, -1);
6135
6136				if ((stcb_tmp == NULL && inp == stcb->sctp_ep) ||
6137				    inp == NULL) {
6138					/* we must add the source address */
6139					/*
6140					 * no scope set since we have a tcb
6141					 * already
6142					 */
6143
6144					/*
6145					 * we must validate the state again
6146					 * here
6147					 */
6148			add_it_now:
6149					if (stcb->asoc.state == 0) {
6150						/* the assoc was freed? */
6151						return (-7);
6152					}
6153					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_4)) {
6154						return (-8);
6155					}
6156				} else if (stcb_tmp == stcb) {
6157					if (stcb->asoc.state == 0) {
6158						/* the assoc was freed? */
6159						return (-10);
6160					}
6161					if (net != NULL) {
6162						/* clear flag */
6163						net->dest_state &=
6164						    ~SCTP_ADDR_NOT_IN_ASSOC;
6165					}
6166				} else {
6167					/*
6168					 * strange, address is in another
6169					 * assoc? straighten out locks.
6170					 */
6171					if (stcb_tmp) {
6172						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6173							/*
6174							 * in setup state we
6175							 * abort this guy
6176							 */
6177							sctp_abort_an_association(stcb_tmp->sctp_ep,
6178							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6179							goto add_it_now;
6180						}
6181						SCTP_TCB_UNLOCK(stcb_tmp);
6182					}
6183					if (stcb->asoc.state == 0) {
6184						/* the assoc was freed? */
6185						return (-12);
6186					}
6187					return (-13);
6188				}
6189			}
6190		} else
6191#endif
6192#ifdef INET6
6193		if (ptype == SCTP_IPV6_ADDRESS) {
6194			if (stcb->asoc.scope.ipv6_addr_legal) {
6195				/* ok get the v6 address and check/add */
6196				struct sctp_ipv6addr_param *p6, p6_buf;
6197
6198				phdr = sctp_get_next_param(m, offset,
6199				    (struct sctp_paramhdr *)&p6_buf,
6200				    sizeof(p6_buf));
6201				if (plen != sizeof(struct sctp_ipv6addr_param) ||
6202				    phdr == NULL) {
6203					return (-14);
6204				}
6205				p6 = (struct sctp_ipv6addr_param *)phdr;
6206				memcpy((caddr_t)&sin6.sin6_addr, p6->addr,
6207				    sizeof(p6->addr));
6208				if (IN6_IS_ADDR_MULTICAST(&sin6.sin6_addr)) {
6209					/* Skip multi-cast addresses */
6210					goto next_param;
6211				}
6212				if (IN6_IS_ADDR_LINKLOCAL(&sin6.sin6_addr)) {
6213					/*
6214					 * Link local make no sense without
6215					 * scope
6216					 */
6217					goto next_param;
6218				}
6219				sa = (struct sockaddr *)&sin6;
6220				inp = stcb->sctp_ep;
6221				atomic_add_int(&stcb->asoc.refcnt, 1);
6222				stcb_tmp = sctp_findassociation_ep_addr(&inp, sa, &net,
6223				    dst, stcb);
6224				atomic_add_int(&stcb->asoc.refcnt, -1);
6225				if (stcb_tmp == NULL &&
6226				    (inp == stcb->sctp_ep || inp == NULL)) {
6227					/*
6228					 * we must validate the state again
6229					 * here
6230					 */
6231			add_it_now6:
6232					if (stcb->asoc.state == 0) {
6233						/* the assoc was freed? */
6234						return (-16);
6235					}
6236					/*
6237					 * we must add the address, no scope
6238					 * set
6239					 */
6240					if (sctp_add_remote_addr(stcb, sa, NULL, SCTP_DONOT_SETSCOPE, SCTP_LOAD_ADDR_5)) {
6241						return (-17);
6242					}
6243				} else if (stcb_tmp == stcb) {
6244					/*
6245					 * we must validate the state again
6246					 * here
6247					 */
6248					if (stcb->asoc.state == 0) {
6249						/* the assoc was freed? */
6250						return (-19);
6251					}
6252					if (net != NULL) {
6253						/* clear flag */
6254						net->dest_state &=
6255						    ~SCTP_ADDR_NOT_IN_ASSOC;
6256					}
6257				} else {
6258					/*
6259					 * strange, address is in another
6260					 * assoc? straighten out locks.
6261					 */
6262					if (stcb_tmp)
6263						if (SCTP_GET_STATE(&stcb_tmp->asoc) & SCTP_STATE_COOKIE_WAIT) {
6264							/*
6265							 * in setup state we
6266							 * abort this guy
6267							 */
6268							sctp_abort_an_association(stcb_tmp->sctp_ep,
6269							    stcb_tmp, NULL, SCTP_SO_NOT_LOCKED);
6270							goto add_it_now6;
6271						}
6272					SCTP_TCB_UNLOCK(stcb_tmp);
6273
6274					if (stcb->asoc.state == 0) {
6275						/* the assoc was freed? */
6276						return (-21);
6277					}
6278					return (-22);
6279				}
6280			}
6281		} else
6282#endif
6283		if (ptype == SCTP_ECN_CAPABLE) {
6284			ecn_allowed = 1;
6285		} else if (ptype == SCTP_ULP_ADAPTATION) {
6286			if (stcb->asoc.state != SCTP_STATE_OPEN) {
6287				struct sctp_adaptation_layer_indication ai,
6288				                                *aip;
6289
6290				phdr = sctp_get_next_param(m, offset,
6291				    (struct sctp_paramhdr *)&ai, sizeof(ai));
6292				aip = (struct sctp_adaptation_layer_indication *)phdr;
6293				if (aip) {
6294					stcb->asoc.peers_adaptation = ntohl(aip->indication);
6295					stcb->asoc.adaptation_needed = 1;
6296				}
6297			}
6298		} else if (ptype == SCTP_SET_PRIM_ADDR) {
6299			struct sctp_asconf_addr_param lstore, *fee;
6300			int lptype;
6301			struct sockaddr *lsa = NULL;
6302
6303#ifdef INET
6304			struct sctp_asconf_addrv4_param *fii;
6305
6306#endif
6307
6308			stcb->asoc.peer_supports_asconf = 1;
6309			if (plen > sizeof(lstore)) {
6310				return (-23);
6311			}
6312			phdr = sctp_get_next_param(m, offset,
6313			    (struct sctp_paramhdr *)&lstore,
6314			    min(plen, sizeof(lstore)));
6315			if (phdr == NULL) {
6316				return (-24);
6317			}
6318			fee = (struct sctp_asconf_addr_param *)phdr;
6319			lptype = ntohs(fee->addrp.ph.param_type);
6320			switch (lptype) {
6321#ifdef INET
6322			case SCTP_IPV4_ADDRESS:
6323				if (plen !=
6324				    sizeof(struct sctp_asconf_addrv4_param)) {
6325					SCTP_PRINTF("Sizeof setprim in init/init ack not %d but %d - ignored\n",
6326					    (int)sizeof(struct sctp_asconf_addrv4_param),
6327					    plen);
6328				} else {
6329					fii = (struct sctp_asconf_addrv4_param *)fee;
6330					sin.sin_addr.s_addr = fii->addrp.addr;
6331					lsa = (struct sockaddr *)&sin;
6332				}
6333				break;
6334#endif
6335#ifdef INET6
6336			case SCTP_IPV6_ADDRESS:
6337				if (plen !=
6338				    sizeof(struct sctp_asconf_addr_param)) {
6339					SCTP_PRINTF("Sizeof setprim (v6) in init/init ack not %d but %d - ignored\n",
6340					    (int)sizeof(struct sctp_asconf_addr_param),
6341					    plen);
6342				} else {
6343					memcpy(sin6.sin6_addr.s6_addr,
6344					    fee->addrp.addr,
6345					    sizeof(fee->addrp.addr));
6346					lsa = (struct sockaddr *)&sin6;
6347				}
6348				break;
6349#endif
6350			default:
6351				break;
6352			}
6353			if (lsa) {
6354				(void)sctp_set_primary_addr(stcb, sa, NULL);
6355			}
6356		} else if (ptype == SCTP_HAS_NAT_SUPPORT) {
6357			stcb->asoc.peer_supports_nat = 1;
6358		} else if (ptype == SCTP_PRSCTP_SUPPORTED) {
6359			/* Peer supports pr-sctp */
6360			stcb->asoc.peer_supports_prsctp = 1;
6361		} else if (ptype == SCTP_SUPPORTED_CHUNK_EXT) {
6362			/* A supported extension chunk */
6363			struct sctp_supported_chunk_types_param *pr_supported;
6364			uint8_t local_store[SCTP_PARAM_BUFFER_SIZE];
6365			int num_ent, i;
6366
6367			phdr = sctp_get_next_param(m, offset,
6368			    (struct sctp_paramhdr *)&local_store, min(sizeof(local_store), plen));
6369			if (phdr == NULL) {
6370				return (-25);
6371			}
6372			stcb->asoc.peer_supports_asconf = 0;
6373			stcb->asoc.peer_supports_prsctp = 0;
6374			stcb->asoc.peer_supports_pktdrop = 0;
6375			stcb->asoc.peer_supports_strreset = 0;
6376			stcb->asoc.peer_supports_nr_sack = 0;
6377			stcb->asoc.peer_supports_auth = 0;
6378			pr_supported = (struct sctp_supported_chunk_types_param *)phdr;
6379			num_ent = plen - sizeof(struct sctp_paramhdr);
6380			for (i = 0; i < num_ent; i++) {
6381				switch (pr_supported->chunk_types[i]) {
6382				case SCTP_ASCONF:
6383				case SCTP_ASCONF_ACK:
6384					stcb->asoc.peer_supports_asconf = 1;
6385					break;
6386				case SCTP_FORWARD_CUM_TSN:
6387					stcb->asoc.peer_supports_prsctp = 1;
6388					break;
6389				case SCTP_PACKET_DROPPED:
6390					stcb->asoc.peer_supports_pktdrop = 1;
6391					break;
6392				case SCTP_NR_SELECTIVE_ACK:
6393					stcb->asoc.peer_supports_nr_sack = 1;
6394					break;
6395				case SCTP_STREAM_RESET:
6396					stcb->asoc.peer_supports_strreset = 1;
6397					break;
6398				case SCTP_AUTHENTICATION:
6399					stcb->asoc.peer_supports_auth = 1;
6400					break;
6401				default:
6402					/* one I have not learned yet */
6403					break;
6404
6405				}
6406			}
6407		} else if (ptype == SCTP_RANDOM) {
6408			if (plen > sizeof(random_store))
6409				break;
6410			if (got_random) {
6411				/* already processed a RANDOM */
6412				goto next_param;
6413			}
6414			phdr = sctp_get_next_param(m, offset,
6415			    (struct sctp_paramhdr *)random_store,
6416			    min(sizeof(random_store), plen));
6417			if (phdr == NULL)
6418				return (-26);
6419			p_random = (struct sctp_auth_random *)phdr;
6420			random_len = plen - sizeof(*p_random);
6421			/* enforce the random length */
6422			if (random_len != SCTP_AUTH_RANDOM_SIZE_REQUIRED) {
6423				SCTPDBG(SCTP_DEBUG_AUTH1, "SCTP: invalid RANDOM len\n");
6424				return (-27);
6425			}
6426			got_random = 1;
6427		} else if (ptype == SCTP_HMAC_LIST) {
6428			int num_hmacs;
6429			int i;
6430
6431			if (plen > sizeof(hmacs_store))
6432				break;
6433			if (got_hmacs) {
6434				/* already processed a HMAC list */
6435				goto next_param;
6436			}
6437			phdr = sctp_get_next_param(m, offset,
6438			    (struct sctp_paramhdr *)hmacs_store,
6439			    min(plen, sizeof(hmacs_store)));
6440			if (phdr == NULL)
6441				return (-28);
6442			hmacs = (struct sctp_auth_hmac_algo *)phdr;
6443			hmacs_len = plen - sizeof(*hmacs);
6444			num_hmacs = hmacs_len / sizeof(hmacs->hmac_ids[0]);
6445			/* validate the hmac list */
6446			if (sctp_verify_hmac_param(hmacs, num_hmacs)) {
6447				return (-29);
6448			}
6449			if (stcb->asoc.peer_hmacs != NULL)
6450				sctp_free_hmaclist(stcb->asoc.peer_hmacs);
6451			stcb->asoc.peer_hmacs = sctp_alloc_hmaclist(num_hmacs);
6452			if (stcb->asoc.peer_hmacs != NULL) {
6453				for (i = 0; i < num_hmacs; i++) {
6454					(void)sctp_auth_add_hmacid(stcb->asoc.peer_hmacs,
6455					    ntohs(hmacs->hmac_ids[i]));
6456				}
6457			}
6458			got_hmacs = 1;
6459		} else if (ptype == SCTP_CHUNK_LIST) {
6460			int i;
6461
6462			if (plen > sizeof(chunks_store))
6463				break;
6464			if (got_chklist) {
6465				/* already processed a Chunks list */
6466				goto next_param;
6467			}
6468			phdr = sctp_get_next_param(m, offset,
6469			    (struct sctp_paramhdr *)chunks_store,
6470			    min(plen, sizeof(chunks_store)));
6471			if (phdr == NULL)
6472				return (-30);
6473			chunks = (struct sctp_auth_chunk_list *)phdr;
6474			num_chunks = plen - sizeof(*chunks);
6475			if (stcb->asoc.peer_auth_chunks != NULL)
6476				sctp_clear_chunklist(stcb->asoc.peer_auth_chunks);
6477			else
6478				stcb->asoc.peer_auth_chunks = sctp_alloc_chunklist();
6479			for (i = 0; i < num_chunks; i++) {
6480				(void)sctp_auth_add_chunk(chunks->chunk_types[i],
6481				    stcb->asoc.peer_auth_chunks);
6482				/* record asconf/asconf-ack if listed */
6483				if (chunks->chunk_types[i] == SCTP_ASCONF)
6484					saw_asconf = 1;
6485				if (chunks->chunk_types[i] == SCTP_ASCONF_ACK)
6486					saw_asconf_ack = 1;
6487
6488			}
6489			got_chklist = 1;
6490		} else if ((ptype == SCTP_HEARTBEAT_INFO) ||
6491			    (ptype == SCTP_STATE_COOKIE) ||
6492			    (ptype == SCTP_UNRECOG_PARAM) ||
6493			    (ptype == SCTP_COOKIE_PRESERVE) ||
6494			    (ptype == SCTP_SUPPORTED_ADDRTYPE) ||
6495			    (ptype == SCTP_ADD_IP_ADDRESS) ||
6496			    (ptype == SCTP_DEL_IP_ADDRESS) ||
6497			    (ptype == SCTP_ERROR_CAUSE_IND) ||
6498		    (ptype == SCTP_SUCCESS_REPORT)) {
6499			 /* don't care */ ;
6500		} else {
6501			if ((ptype & 0x8000) == 0x0000) {
6502				/*
6503				 * must stop processing the rest of the
6504				 * param's. Any report bits were handled
6505				 * with the call to
6506				 * sctp_arethere_unrecognized_parameters()
6507				 * when the INIT or INIT-ACK was first seen.
6508				 */
6509				break;
6510			}
6511		}
6512
6513next_param:
6514		offset += SCTP_SIZE32(plen);
6515		if (offset >= limit) {
6516			break;
6517		}
6518		phdr = sctp_get_next_param(m, offset, &parm_buf,
6519		    sizeof(parm_buf));
6520	}
6521	/* Now check to see if we need to purge any addresses */
6522	TAILQ_FOREACH_SAFE(net, &stcb->asoc.nets, sctp_next, nnet) {
6523		if ((net->dest_state & SCTP_ADDR_NOT_IN_ASSOC) ==
6524		    SCTP_ADDR_NOT_IN_ASSOC) {
6525			/* This address has been removed from the asoc */
6526			/* remove and free it */
6527			stcb->asoc.numnets--;
6528			TAILQ_REMOVE(&stcb->asoc.nets, net, sctp_next);
6529			sctp_free_remote_addr(net);
6530			if (net == stcb->asoc.primary_destination) {
6531				stcb->asoc.primary_destination = NULL;
6532				sctp_select_primary_destination(stcb);
6533			}
6534		}
6535	}
6536	if (ecn_allowed == 0) {
6537		stcb->asoc.ecn_allowed = 0;
6538	}
6539	/* validate authentication required parameters */
6540	if (got_random && got_hmacs) {
6541		stcb->asoc.peer_supports_auth = 1;
6542	} else {
6543		stcb->asoc.peer_supports_auth = 0;
6544	}
6545	if (!stcb->asoc.peer_supports_auth && got_chklist) {
6546		/* peer does not support auth but sent a chunks list? */
6547		return (-31);
6548	}
6549	if (!SCTP_BASE_SYSCTL(sctp_asconf_auth_nochk) && stcb->asoc.peer_supports_asconf &&
6550	    !stcb->asoc.peer_supports_auth) {
6551		/* peer supports asconf but not auth? */
6552		return (-32);
6553	} else if ((stcb->asoc.peer_supports_asconf) && (stcb->asoc.peer_supports_auth) &&
6554	    ((saw_asconf == 0) || (saw_asconf_ack == 0))) {
6555		return (-33);
6556	}
6557	/* concatenate the full random key */
6558	keylen = sizeof(*p_random) + random_len + sizeof(*hmacs) + hmacs_len;
6559	if (chunks != NULL) {
6560		keylen += sizeof(*chunks) + num_chunks;
6561	}
6562	new_key = sctp_alloc_key(keylen);
6563	if (new_key != NULL) {
6564		/* copy in the RANDOM */
6565		if (p_random != NULL) {
6566			keylen = sizeof(*p_random) + random_len;
6567			bcopy(p_random, new_key->key, keylen);
6568		}
6569		/* append in the AUTH chunks */
6570		if (chunks != NULL) {
6571			bcopy(chunks, new_key->key + keylen,
6572			    sizeof(*chunks) + num_chunks);
6573			keylen += sizeof(*chunks) + num_chunks;
6574		}
6575		/* append in the HMACs */
6576		if (hmacs != NULL) {
6577			bcopy(hmacs, new_key->key + keylen,
6578			    sizeof(*hmacs) + hmacs_len);
6579		}
6580	} else {
6581		/* failed to get memory for the key */
6582		return (-34);
6583	}
6584	if (stcb->asoc.authinfo.peer_random != NULL)
6585		sctp_free_key(stcb->asoc.authinfo.peer_random);
6586	stcb->asoc.authinfo.peer_random = new_key;
6587	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.assoc_keyid);
6588	sctp_clear_cachedkeys(stcb, stcb->asoc.authinfo.recv_keyid);
6589
6590	return (0);
6591}
6592
6593int
6594sctp_set_primary_addr(struct sctp_tcb *stcb, struct sockaddr *sa,
6595    struct sctp_nets *net)
6596{
6597	/* make sure the requested primary address exists in the assoc */
6598	if (net == NULL && sa)
6599		net = sctp_findnet(stcb, sa);
6600
6601	if (net == NULL) {
6602		/* didn't find the requested primary address! */
6603		return (-1);
6604	} else {
6605		/* set the primary address */
6606		if (net->dest_state & SCTP_ADDR_UNCONFIRMED) {
6607			/* Must be confirmed, so queue to set */
6608			net->dest_state |= SCTP_ADDR_REQ_PRIMARY;
6609			return (0);
6610		}
6611		stcb->asoc.primary_destination = net;
6612		if (!(net->dest_state & SCTP_ADDR_PF) && (stcb->asoc.alternate)) {
6613			sctp_free_remote_addr(stcb->asoc.alternate);
6614			stcb->asoc.alternate = NULL;
6615		}
6616		net = TAILQ_FIRST(&stcb->asoc.nets);
6617		if (net != stcb->asoc.primary_destination) {
6618			/*
6619			 * first one on the list is NOT the primary
6620			 * sctp_cmpaddr() is much more efficient if the
6621			 * primary is the first on the list, make it so.
6622			 */
6623			TAILQ_REMOVE(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6624			TAILQ_INSERT_HEAD(&stcb->asoc.nets, stcb->asoc.primary_destination, sctp_next);
6625		}
6626		return (0);
6627	}
6628}
6629
6630int
6631sctp_is_vtag_good(uint32_t tag, uint16_t lport, uint16_t rport, struct timeval *now)
6632{
6633	/*
6634	 * This function serves two purposes. It will see if a TAG can be
6635	 * re-used and return 1 for yes it is ok and 0 for don't use that
6636	 * tag. A secondary function it will do is purge out old tags that
6637	 * can be removed.
6638	 */
6639	struct sctpvtaghead *chain;
6640	struct sctp_tagblock *twait_block;
6641	struct sctpasochead *head;
6642	struct sctp_tcb *stcb;
6643	int i;
6644
6645	SCTP_INP_INFO_RLOCK();
6646	head = &SCTP_BASE_INFO(sctp_asochash)[SCTP_PCBHASH_ASOC(tag,
6647	    SCTP_BASE_INFO(hashasocmark))];
6648	if (head == NULL) {
6649		/* invalid vtag */
6650		goto skip_vtag_check;
6651	}
6652	LIST_FOREACH(stcb, head, sctp_asocs) {
6653		/*
6654		 * We choose not to lock anything here. TCB's can't be
6655		 * removed since we have the read lock, so they can't be
6656		 * freed on us, same thing for the INP. I may be wrong with
6657		 * this assumption, but we will go with it for now :-)
6658		 */
6659		if (stcb->sctp_ep->sctp_flags & SCTP_PCB_FLAGS_SOCKET_ALLGONE) {
6660			continue;
6661		}
6662		if (stcb->asoc.my_vtag == tag) {
6663			/* candidate */
6664			if (stcb->rport != rport) {
6665				continue;
6666			}
6667			if (stcb->sctp_ep->sctp_lport != lport) {
6668				continue;
6669			}
6670			/* Its a used tag set */
6671			SCTP_INP_INFO_RUNLOCK();
6672			return (0);
6673		}
6674	}
6675skip_vtag_check:
6676
6677	chain = &SCTP_BASE_INFO(vtag_timewait)[(tag % SCTP_STACK_VTAG_HASH_SIZE)];
6678	/* Now what about timed wait ? */
6679	LIST_FOREACH(twait_block, chain, sctp_nxt_tagblock) {
6680		/*
6681		 * Block(s) are present, lets see if we have this tag in the
6682		 * list
6683		 */
6684		for (i = 0; i < SCTP_NUMBER_IN_VTAG_BLOCK; i++) {
6685			if (twait_block->vtag_block[i].v_tag == 0) {
6686				/* not used */
6687				continue;
6688			} else if ((long)twait_block->vtag_block[i].tv_sec_at_expire <
6689			    now->tv_sec) {
6690				/* Audit expires this guy */
6691				twait_block->vtag_block[i].tv_sec_at_expire = 0;
6692				twait_block->vtag_block[i].v_tag = 0;
6693				twait_block->vtag_block[i].lport = 0;
6694				twait_block->vtag_block[i].rport = 0;
6695			} else if ((twait_block->vtag_block[i].v_tag == tag) &&
6696				    (twait_block->vtag_block[i].lport == lport) &&
6697			    (twait_block->vtag_block[i].rport == rport)) {
6698				/* Bad tag, sorry :< */
6699				SCTP_INP_INFO_RUNLOCK();
6700				return (0);
6701			}
6702		}
6703	}
6704	SCTP_INP_INFO_RUNLOCK();
6705	return (1);
6706}
6707
6708static void
6709sctp_drain_mbufs(struct sctp_tcb *stcb)
6710{
6711	/*
6712	 * We must hunt this association for MBUF's past the cumack (i.e.
6713	 * out of order data that we can renege on).
6714	 */
6715	struct sctp_association *asoc;
6716	struct sctp_tmit_chunk *chk, *nchk;
6717	uint32_t cumulative_tsn_p1;
6718	struct sctp_queued_to_read *ctl, *nctl;
6719	int cnt, strmat;
6720	uint32_t gap, i;
6721	int fnd = 0;
6722
6723	/* We look for anything larger than the cum-ack + 1 */
6724
6725	asoc = &stcb->asoc;
6726	if (asoc->cumulative_tsn == asoc->highest_tsn_inside_map) {
6727		/* none we can reneg on. */
6728		return;
6729	}
6730	SCTP_STAT_INCR(sctps_protocol_drains_done);
6731	cumulative_tsn_p1 = asoc->cumulative_tsn + 1;
6732	cnt = 0;
6733	/* First look in the re-assembly queue */
6734	TAILQ_FOREACH_SAFE(chk, &asoc->reasmqueue, sctp_next, nchk) {
6735		if (SCTP_TSN_GT(chk->rec.data.TSN_seq, cumulative_tsn_p1)) {
6736			/* Yep it is above cum-ack */
6737			cnt++;
6738			SCTP_CALC_TSN_TO_GAP(gap, chk->rec.data.TSN_seq, asoc->mapping_array_base_tsn);
6739			asoc->size_on_reasm_queue = sctp_sbspace_sub(asoc->size_on_reasm_queue, chk->send_size);
6740			sctp_ucount_decr(asoc->cnt_on_reasm_queue);
6741			SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6742			TAILQ_REMOVE(&asoc->reasmqueue, chk, sctp_next);
6743			if (chk->data) {
6744				sctp_m_freem(chk->data);
6745				chk->data = NULL;
6746			}
6747			sctp_free_a_chunk(stcb, chk, SCTP_SO_NOT_LOCKED);
6748		}
6749	}
6750	/* Ok that was fun, now we will drain all the inbound streams? */
6751	for (strmat = 0; strmat < asoc->streamincnt; strmat++) {
6752		TAILQ_FOREACH_SAFE(ctl, &asoc->strmin[strmat].inqueue, next, nctl) {
6753			if (SCTP_TSN_GT(ctl->sinfo_tsn, cumulative_tsn_p1)) {
6754				/* Yep it is above cum-ack */
6755				cnt++;
6756				SCTP_CALC_TSN_TO_GAP(gap, ctl->sinfo_tsn, asoc->mapping_array_base_tsn);
6757				asoc->size_on_all_streams = sctp_sbspace_sub(asoc->size_on_all_streams, ctl->length);
6758				sctp_ucount_decr(asoc->cnt_on_all_streams);
6759				SCTP_UNSET_TSN_PRESENT(asoc->mapping_array, gap);
6760				TAILQ_REMOVE(&asoc->strmin[strmat].inqueue, ctl, next);
6761				if (ctl->data) {
6762					sctp_m_freem(ctl->data);
6763					ctl->data = NULL;
6764				}
6765				sctp_free_remote_addr(ctl->whoFrom);
6766				SCTP_ZONE_FREE(SCTP_BASE_INFO(ipi_zone_readq), ctl);
6767				SCTP_DECR_READQ_COUNT();
6768			}
6769		}
6770	}
6771	if (cnt) {
6772		/* We must back down to see what the new highest is */
6773		for (i = asoc->highest_tsn_inside_map; SCTP_TSN_GE(i, asoc->mapping_array_base_tsn); i--) {
6774			SCTP_CALC_TSN_TO_GAP(gap, i, asoc->mapping_array_base_tsn);
6775			if (SCTP_IS_TSN_PRESENT(asoc->mapping_array, gap)) {
6776				asoc->highest_tsn_inside_map = i;
6777				fnd = 1;
6778				break;
6779			}
6780		}
6781		if (!fnd) {
6782			asoc->highest_tsn_inside_map = asoc->mapping_array_base_tsn - 1;
6783		}
6784		/*
6785		 * Question, should we go through the delivery queue? The
6786		 * only reason things are on here is the app not reading OR
6787		 * a p-d-api up. An attacker COULD send enough in to
6788		 * initiate the PD-API and then send a bunch of stuff to
6789		 * other streams... these would wind up on the delivery
6790		 * queue.. and then we would not get to them. But in order
6791		 * to do this I then have to back-track and un-deliver
6792		 * sequence numbers in streams.. el-yucko. I think for now
6793		 * we will NOT look at the delivery queue and leave it to be
6794		 * something to consider later. An alternative would be to
6795		 * abort the P-D-API with a notification and then deliver
6796		 * the data.... Or another method might be to keep track of
6797		 * how many times the situation occurs and if we see a
6798		 * possible attack underway just abort the association.
6799		 */
6800#ifdef SCTP_DEBUG
6801		SCTPDBG(SCTP_DEBUG_PCB1, "Freed %d chunks from reneg harvest\n", cnt);
6802#endif
6803		/*
6804		 * Now do we need to find a new
6805		 * asoc->highest_tsn_inside_map?
6806		 */
6807		asoc->last_revoke_count = cnt;
6808		(void)SCTP_OS_TIMER_STOP(&stcb->asoc.dack_timer.timer);
6809		/* sa_ignore NO_NULL_CHK */
6810		sctp_send_sack(stcb, SCTP_SO_NOT_LOCKED);
6811		sctp_chunk_output(stcb->sctp_ep, stcb, SCTP_OUTPUT_FROM_DRAIN, SCTP_SO_NOT_LOCKED);
6812	}
6813	/*
6814	 * Another issue, in un-setting the TSN's in the mapping array we
6815	 * DID NOT adjust the highest_tsn marker.  This will cause one of
6816	 * two things to occur. It may cause us to do extra work in checking
6817	 * for our mapping array movement. More importantly it may cause us
6818	 * to SACK every datagram. This may not be a bad thing though since
6819	 * we will recover once we get our cum-ack above and all this stuff
6820	 * we dumped recovered.
6821	 */
6822}
6823
6824void
6825sctp_drain()
6826{
6827	/*
6828	 * We must walk the PCB lists for ALL associations here. The system
6829	 * is LOW on MBUF's and needs help. This is where reneging will
6830	 * occur. We really hope this does NOT happen!
6831	 */
6832	VNET_ITERATOR_DECL(vnet_iter);
6833	VNET_LIST_RLOCK_NOSLEEP();
6834	VNET_FOREACH(vnet_iter) {
6835		CURVNET_SET(vnet_iter);
6836		struct sctp_inpcb *inp;
6837		struct sctp_tcb *stcb;
6838
6839		SCTP_STAT_INCR(sctps_protocol_drain_calls);
6840		if (SCTP_BASE_SYSCTL(sctp_do_drain) == 0) {
6841#ifdef VIMAGE
6842			continue;
6843#else
6844			return;
6845#endif
6846		}
6847		SCTP_INP_INFO_RLOCK();
6848		LIST_FOREACH(inp, &SCTP_BASE_INFO(listhead), sctp_list) {
6849			/* For each endpoint */
6850			SCTP_INP_RLOCK(inp);
6851			LIST_FOREACH(stcb, &inp->sctp_asoc_list, sctp_tcblist) {
6852				/* For each association */
6853				SCTP_TCB_LOCK(stcb);
6854				sctp_drain_mbufs(stcb);
6855				SCTP_TCB_UNLOCK(stcb);
6856			}
6857			SCTP_INP_RUNLOCK(inp);
6858		}
6859		SCTP_INP_INFO_RUNLOCK();
6860		CURVNET_RESTORE();
6861	}
6862	VNET_LIST_RUNLOCK_NOSLEEP();
6863}
6864
6865/*
6866 * start a new iterator
6867 * iterates through all endpoints and associations based on the pcb_state
6868 * flags and asoc_state.  "af" (mandatory) is executed for all matching
6869 * assocs and "ef" (optional) is executed when the iterator completes.
6870 * "inpf" (optional) is executed for each new endpoint as it is being
6871 * iterated through. inpe (optional) is called when the inp completes
6872 * its way through all the stcbs.
6873 */
6874int
6875sctp_initiate_iterator(inp_func inpf,
6876    asoc_func af,
6877    inp_func inpe,
6878    uint32_t pcb_state,
6879    uint32_t pcb_features,
6880    uint32_t asoc_state,
6881    void *argp,
6882    uint32_t argi,
6883    end_func ef,
6884    struct sctp_inpcb *s_inp,
6885    uint8_t chunk_output_off)
6886{
6887	struct sctp_iterator *it = NULL;
6888
6889	if (af == NULL) {
6890		return (-1);
6891	}
6892	SCTP_MALLOC(it, struct sctp_iterator *, sizeof(struct sctp_iterator),
6893	    SCTP_M_ITER);
6894	if (it == NULL) {
6895		SCTP_LTRACE_ERR_RET(NULL, NULL, NULL, SCTP_FROM_SCTP_PCB, ENOMEM);
6896		return (ENOMEM);
6897	}
6898	memset(it, 0, sizeof(*it));
6899	it->function_assoc = af;
6900	it->function_inp = inpf;
6901	if (inpf)
6902		it->done_current_ep = 0;
6903	else
6904		it->done_current_ep = 1;
6905	it->function_atend = ef;
6906	it->pointer = argp;
6907	it->val = argi;
6908	it->pcb_flags = pcb_state;
6909	it->pcb_features = pcb_features;
6910	it->asoc_state = asoc_state;
6911	it->function_inp_end = inpe;
6912	it->no_chunk_output = chunk_output_off;
6913	it->vn = curvnet;
6914	if (s_inp) {
6915		/* Assume lock is held here */
6916		it->inp = s_inp;
6917		SCTP_INP_INCR_REF(it->inp);
6918		it->iterator_flags = SCTP_ITERATOR_DO_SINGLE_INP;
6919	} else {
6920		SCTP_INP_INFO_RLOCK();
6921		it->inp = LIST_FIRST(&SCTP_BASE_INFO(listhead));
6922		if (it->inp) {
6923			SCTP_INP_INCR_REF(it->inp);
6924		}
6925		SCTP_INP_INFO_RUNLOCK();
6926		it->iterator_flags = SCTP_ITERATOR_DO_ALL_INP;
6927
6928	}
6929	SCTP_IPI_ITERATOR_WQ_LOCK();
6930
6931	TAILQ_INSERT_TAIL(&sctp_it_ctl.iteratorhead, it, sctp_nxt_itr);
6932	if (sctp_it_ctl.iterator_running == 0) {
6933		sctp_wakeup_iterator();
6934	}
6935	SCTP_IPI_ITERATOR_WQ_UNLOCK();
6936	/* sa_ignore MEMLEAK {memory is put on the tailq for the iterator} */
6937	return (0);
6938}
6939